1 2021-10-20 Alex Coplan <alex.coplan@arm.com>
3 * calls.c (initialize_argument_information): Remove some dead
4 code, remove handling for function_arg returning const_int.
5 * doc/tm.texi: Delete documentation for unused target hooks.
6 * doc/tm.texi.in: Likewise.
7 * target.def (load_bounds_for_arg): Delete.
8 (store_bounds_for_arg): Delete.
9 (load_returned_bounds): Delete.
10 (store_returned_bounds): Delete.
11 * targhooks.c (default_load_bounds_for_arg): Delete.
12 (default_store_bounds_for_arg): Delete.
13 (default_load_returned_bounds): Delete.
14 (default_store_returned_bounds): Delete.
15 * targhooks.h (default_load_bounds_for_arg): Delete.
16 (default_store_bounds_for_arg): Delete.
17 (default_load_returned_bounds): Delete.
18 (default_store_returned_bounds): Delete.
20 2021-10-20 Jonathan Wakely <jwakely@redhat.com>
22 * doc/extend.texi (Basic Asm): Clarify that asm is not an
24 * doc/invoke.texi (-fno-asm): Fix description for C++.
26 2021-10-20 Jonathan Wakely <jwakely@redhat.com>
28 * doc/install.texi: Remove link to old.html
30 2021-10-20 Tamar Christina <tamar.christina@arm.com>
32 * config/aarch64/aarch64-simd.md (*aarch64_cmtst_same_<mode>): New.
34 2021-10-20 Tamar Christina <tamar.christina@arm.com>
36 * config/aarch64/aarch64-simd.md (*aarch64_narrow_trunc<mode>): New.
38 2021-10-20 Tamar Christina <tamar.christina@arm.com>
40 * config/aarch64/aarch64-simd.md (aarch64_simd_ashr<mode>): Add case cmp
42 * config/aarch64/constraints.md (D1): New.
44 2021-10-20 Tamar Christina <tamar.christina@arm.com>
46 * config/aarch64/aarch64-simd.md
47 (*aarch64_<srn_op>topbits_shuffle<mode>_le): New.
48 (*aarch64_topbits_shuffle<mode>_le): New.
49 (*aarch64_<srn_op>topbits_shuffle<mode>_be): New.
50 (*aarch64_topbits_shuffle<mode>_be): New.
51 * config/aarch64/predicates.md
52 (aarch64_simd_shift_imm_vec_exact_top): New.
54 2021-10-20 Tamar Christina <tamar.christina@arm.com>
56 * config/aarch64/aarch64-simd.md (*aarch64_<srn_op>shrn<mode>_vect,
57 *aarch64_<srn_op>shrn<mode>2_vect_le,
58 *aarch64_<srn_op>shrn<mode>2_vect_be): New.
59 * config/aarch64/iterators.md (srn_op): New.
61 2021-10-20 Chung-Lin Tang <cltang@codesourcery.com>
63 * omp-low.c (omp_copy_decl_2): For !ctx, use record_vars to add new copy
65 (scan_sharing_clauses): Place copy of OMP_CLAUSE_IN_REDUCTION decl in
66 ctx->outer instead of ctx.
68 2021-10-20 Martin Liska <mliska@suse.cz>
71 2021-10-19 Martin Liska <mliska@suse.cz>
74 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
75 * system.h (strip_whilespaces): New function.
77 2021-10-20 Martin Liska <mliska@suse.cz>
80 2021-10-19 Martin Liska <mliska@suse.cz>
83 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
86 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
88 * config/aarch64/aarch64.c (aarch64_case_values_threshold):
89 Change to 8 with -Os, 11 otherwise.
91 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
93 * config/aarch64/aarch64.c (neoversev1_tunings):
94 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
95 (neoversen2_tunings): Likewise.
97 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
100 * config/aarch64/aarch64.md (frint_pattern): Update comment.
101 * config/aarch64/aarch64-simd-builtins.def: Change frintn to roundeven.
102 * config/aarch64/arm_fp16.h: Change frintn to roundeven.
103 * config/aarch64/arm_neon.h: Likewise.
104 * config/aarch64/iterators.md (frint_pattern): Use roundeven for FRINTN.
106 2021-10-20 Martin Liska <mliska@suse.cz>
108 * config/arm/arm.c (arm_unwind_emit_sequence): Do not declare
109 already declared global variable.
110 (arm_unwind_emit_set): Use out_file as function argument.
111 (arm_unwind_emit): Likewise.
112 * config/darwin.c (machopic_output_data_section_indirection): Likewise.
113 (machopic_output_stub_indirection): Likewise.
114 (machopic_output_indirection): Likewise.
115 (machopic_finish): Likewise.
116 * config/i386/i386.c (ix86_asm_output_function_label): Likewise.
117 * config/i386/winnt.c (i386_pe_seh_unwind_emit): Likewise.
118 * config/ia64/ia64.c (process_epilogue): Likewise.
119 (process_cfa_adjust_cfa): Likewise.
120 (process_cfa_register): Likewise.
121 (process_cfa_offset): Likewise.
122 (ia64_asm_unwind_emit): Likewise.
123 * config/s390/s390.c (s390_asm_output_function_label): Likewise.
125 2021-10-20 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
127 * config/aarch64/aarch64-builtins.c
128 (aarch64_general_gimple_fold_builtin):
129 lower vld1 and vst1 variants of the neon builtins
130 * config/aarch64/aarch64-protos.h:
131 (aarch64_general_gimple_fold_builtin): Add gsi parameter.
132 * config/aarch64/aarch64.c (aarch64_general_gimple_fold_builtin):
135 2021-10-20 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
137 * match.pd: Generate IFN_TRUNC.
139 2021-10-20 Richard Biener <rguenther@suse.de>
141 PR tree-optimization/102853
142 * tree-data-ref.c (split_constant_offset_1): Bail out
143 immediately if the expression traps on overflow.
145 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
147 * tree-ssa-threadbackward.c (back_threader::~back_threader): Remove.
149 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
151 * tree-ssa-threadupdate.c (back_jt_path_registry::adjust_paths_after_duplication):
152 Remove superflous debugging message.
153 (back_jt_path_registry::duplicate_thread_path): Same.
155 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
157 * tree-ssa-threadbackward.c (back_threader_registry::back_threader_registry):
159 (back_threader_registry::register_path): Remove m_threaded_paths.
161 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
163 PR tree-optimization/102814
164 * doc/invoke.texi: Document --param=max-fsm-thread-length.
165 * params.opt: Add --param=max-fsm-thread-length.
166 * tree-ssa-threadbackward.c
167 (back_threader_profitability::profitable_path_p): Fail on paths
168 longer than max-fsm-thread-length.
170 2021-10-20 Eric Botcazou <ebotcazou@adacore.com>
173 * cfgexpand.c (expand_gimple_basic_block): Disregard a final debug
174 statement to reset the current location for the outgoing edges.
176 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
177 Richard Biener <rguenther@suse.de>
179 * tree-ssa-threadupdate.c (cancel_thread): Dump threading reason
180 on the same line as the threading cancellation.
181 (jt_path_registry::cancel_invalid_paths): Avoid rotating loops.
182 Avoid threading through loop headers where the path remains in the
185 2021-10-20 Siddhesh Poyarekar <siddhesh@gotplt.org>
187 * tree-object-size.c (unknown): Make into a function. Adjust
189 (unknown_object_size): Simplify implementation.
191 2021-10-20 Hongtao Liu <hongtao.liu@intel.com>
192 Kewen Lin <linkw@linux.ibm.com>
194 * doc/sourcebuild.texi (Effective-Target Keywords): Document
195 vect_slp_v2qi_store, vect_slp_v4qi_store, vect_slp_v8qi_store,
196 vect_slp_v16qi_store, vect_slp_v2hi_store,
197 vect_slp_v4hi_store, vect_slp_v2si_store, vect_slp_v4si_store.
199 2021-10-19 Jonathan Wakely <jwakely@redhat.com>
201 * doc/extend.texi (Basic PowerPC Built-in Functions): Fix typo.
203 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
207 * config/rs6000/emmintrin.h: Guard POWER8 intrinsics.
208 * config/rs6000/pmmintrin.h: Same.
209 * config/rs6000/smmintrin.h: Same.
210 * config/rs6000/tmmintrin.h: Same.
212 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
214 * config.gcc (extra_headers): Add nmmintrin.h.
216 2021-10-19 Richard Biener <rguenther@suse.de>
218 * tree-vectorizer.h (vect_supportable_dr_alignment): Add
219 misalignment parameter.
220 * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
221 Do not change DR_MISALIGNMENT in place, instead pass the
222 adjusted misalignment to vect_supportable_dr_alignment.
223 (vect_peeling_supportable): Likewise.
224 (vect_peeling_hash_get_lowest_cost): Adjust.
225 (vect_enhance_data_refs_alignment): Likewise.
226 (vect_vfa_access_size): Likewise.
227 (vect_supportable_dr_alignment): Add misalignment
228 parameter and simplify.
229 * tree-vect-stmts.c (get_negative_load_store_type): Adjust.
230 (get_group_load_store_type): Likewise.
231 (get_load_store_type): Likewise.
233 2021-10-19 Clément Chigot <clement.chigot@atos.net>
235 * config/rs6000/rs6000.c (rs6000_xcoff_file_end): Move
236 __tls_get_addr reference to .text csect.
238 2021-10-19 Martin Liska <mliska@suse.cz>
241 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
244 2021-10-19 Richard Biener <rguenther@suse.de>
246 * tree-vectorizer.h (vect_get_store_cost): Adjust signature.
247 (vect_get_load_cost): Likewise.
248 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
249 alignment support scheme and misalignment as arguments
251 (vect_get_peeling_costs_all_drs): Compute that info here
252 and note that we shouldn't need to.
253 * tree-vect-stmts.c (vect_model_store_cost): Get
254 alignment support scheme and misalignment as arguments.
255 (vect_get_store_cost): Likewise.
256 (vect_model_load_cost): Likewise.
257 (vect_get_load_cost): Likewise.
258 (vectorizable_store): Pass down alignment support scheme
259 and misalignment to costing.
260 (vectorizable_load): Likewise.
262 2021-10-19 Richard Biener <rguenther@suse.de>
264 * tree-vect-stmts.c (get_negative_load_store_type): Add
265 offset output parameter and initialize it.
266 (get_group_load_store_type): Likewise.
267 (get_load_store_type): Likewise.
268 (vectorizable_store): Use offset as computed by
270 (vectorizable_load): Likewise.
272 2021-10-19 Richard Biener <rguenther@suse.de>
274 PR tree-optimization/102827
275 * tree-if-conv.c (predicate_statements): Add pe parameter
276 and use that edge to insert invariant stmts on.
277 (combine_blocks): Pass through pe.
278 (tree_if_conversion): Compute the edge to insert invariant
279 stmts on and pass it along.
281 2021-10-19 Roger Sayle <roger@nextmovesoftware.com>
284 * config/bfin/bfin.md (addsubv2hi3, subaddv2hi3, ssaddsubv2hi3,
285 sssubaddv2hi3): Swap the order of operators in vec_concat.
287 2021-10-19 Xionghu Luo <luoxhu@linux.ibm.com>
289 * config/rs6000/altivec.md (*altivec_vmrghb_internal): Delete.
290 (altivec_vmrghb_direct): New.
291 (*altivec_vmrghh_internal): Delete.
292 (altivec_vmrghh_direct): New.
293 (*altivec_vmrghw_internal): Delete.
294 (altivec_vmrghw_direct_<mode>): New.
295 (altivec_vmrghw_direct): Delete.
296 (*altivec_vmrglb_internal): Delete.
297 (altivec_vmrglb_direct): New.
298 (*altivec_vmrglh_internal): Delete.
299 (altivec_vmrglh_direct): New.
300 (*altivec_vmrglw_internal): Delete.
301 (altivec_vmrglw_direct_<mode>): New.
302 (altivec_vmrglw_direct): Delete.
303 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
304 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):
306 * config/rs6000/vsx.md (vsx_xxmrghw_<mode>): Adjust.
307 (vsx_xxmrglw_<mode>): Adjust.
309 2021-10-19 Aldy Hernandez <aldyh@redhat.com>
311 * passes.def: Change threading comment before pass_ccp pass.
313 2021-10-19 Haochen Gui <guihaoc@gcc.gnu.org>
315 * config/rs6000/rs6000-call.c (altivec_expand_lxvr_builtin):
316 Modify the expansion for sign extension. All extensions are done
317 within VSX registers.
319 2021-10-19 Richard Biener <rguenther@suse.de>
321 * tree-vect-stmts.c (get_group_load_store_type): Add
322 misalignment output parameter and initialize it.
323 (get_group_load_store_type): Likewise.
324 (vectorizable_store): Remove now redundant queries.
325 (vectorizable_load): Likewise.
327 2021-10-19 Richard Biener <rguenther@suse.de>
329 * tree-vectorizer.h (vect_supportable_dr_alignment): Remove
330 check_aligned argument.
331 * tree-vect-data-refs.c (vect_supportable_dr_alignment):
333 (vect_peeling_hash_insert): Add supportable_if_not_aligned
334 argument and do not call vect_supportable_dr_alignment here.
335 (vect_peeling_supportable): Adjust.
336 (vect_enhance_data_refs_alignment): Compute whether the
337 access is supported with different alignment here and
338 pass that down to vect_peeling_hash_insert.
339 (vect_vfa_access_size): Adjust.
340 * tree-vect-stmts.c (vect_get_store_cost): Likewise.
341 (vect_get_load_cost): Likewise.
342 (get_negative_load_store_type): Likewise.
343 (get_group_load_store_type): Likewise.
344 (get_load_store_type): Likewise.
346 2021-10-19 Martin Liska <mliska@suse.cz>
349 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
350 * system.h (strip_whilespaces): New function.
352 2021-10-19 dianhong xu <dianhong.xu@intel.com>
354 * config/i386/avx512fp16intrin.h:
355 (_mm512_set1_pch): New intrinsic.
356 * config/i386/avx512fp16vlintrin.h:
357 (_mm256_set1_pch): New intrinsic.
358 (_mm_set1_pch): Ditto.
360 2021-10-18 Andrew MacLeod <amacleod@redhat.com>
362 PR tree-optimization/102796
363 * gimple-range.cc (gimple_ranger::range_on_edge): Process EH edges
364 normally. Return get_tree_range for non gimple_range_ssa_p names.
365 (gimple_ranger::range_of_stmt): Use get_tree_range for non
366 gimple_range_ssa_p names.
368 2021-10-18 Uroš Bizjak <ubizjak@gmail.com>
371 * config/i386/i386.c (ix86_print_operand_address):
372 Error out for non-address_operand asm operands.
374 2021-10-18 Richard Biener <rguenther@suse.de>
376 * tree-vect-data-refs.c (vect_peeling_hash_insert): Do
377 not auto-convert dr_alignment_support to bool.
378 (vect_peeling_supportable): Likewise.
379 (vect_enhance_data_refs_alignment): Likewise.
380 (vect_supportable_dr_alignment): Commonize read/write case.
381 * tree-vect-stmts.c (vect_get_store_cost): Use
382 dr_alignment_support, not int, for the vect_supportable_dr_alignment
384 (vect_get_load_cost): Likewise.
386 2021-10-18 Siddhesh Poyarekar <siddhesh@gotplt.org>
388 * tree-object-size.c (object_sizes_execute): Consolidate LHS
389 null check and do it early.
391 2021-10-18 Richard Biener <rguenther@suse.de>
393 * tree-vect-stmts.c (vectorizable_store): Use the
394 computed alignment scheme instead of querying
397 2021-10-18 Richard Biener <rguenther@suse.de>
399 * tree-vect-stmts.c (vectorizable_store): Do not recompute
400 alignment scheme already determined by get_load_store_type.
402 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
404 * tree-ssa-threadbackward.c (class pass_thread_jumps_full):
405 Clone corresponding pass.
407 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
409 * combine.c (recog_for_combine): For an unrecognized move/set of
410 a constant, try force_const_mem to place it in the constant pool.
411 * cse.c (constant_pool_entries_cost, constant_pool_entries_regcost):
412 Delete global variables (that are no longer assigned a cost value).
413 (cse_insn): Simplify logic for deciding whether to place a folded
414 constant in the constant pool using force_const_mem.
415 (cse_main): Remove zero initialization of constant_pool_entries_cost
416 and constant_pool_entries_regcost.
417 * config/i386/i386.c (ix86_rtx_costs): Make memory accesses
418 fractionally more expensive, when optimizing for speed.
420 2021-10-18 Martin Liska <mliska@suse.cz>
422 PR gcov-profile/102746
423 PR gcov-profile/102747
424 * gcov.c (main): Return return_code.
425 (output_gcov_file): Mark return_code when error happens.
426 (generate_results): Likewise.
427 (read_graph_file): Likewise.
428 (read_count_file): Likewise.
430 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
432 * config/bfin/bfin.md (define_constants): Remove UNSPEC_ONES.
433 (define_insn "ones"): Replace UNSPEC_ONES with a truncate of
434 a popcount, allowing compile-time evaluation/simplification.
435 (popcountsi2, popcounthi2): New expanders using a "ones" insn.
437 2021-10-18 Richard Biener <rguenther@suse.de>
439 PR tree-optimization/102788
440 * tree-vect-patterns.c (vect_init_pattern_stmt): Allow
442 (vect_pattern_recog_1): Likewise.
443 (vect_recog_bool_pattern): Continue matching the pattern
444 even if we do not have a vector type for a conversion
447 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
449 * simplify-rtx.c (simplify_const_unary_operation) [SS_NEG, SS_ABS]:
450 Evalute SS_NEG and SS_ABS of a constant argument.
452 2021-10-18 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
455 * gimple-match-head.c (try_conditional_simplification): Add case for single operand.
456 * internal-fn.def: Add entry for COND_NEG internal function.
457 * internal-fn.c (FOR_EACH_CODE_MAPPING): Add entry for
458 NEGATE_EXPR, COND_NEG mapping.
459 * optabs.def: Add entry for cond_neg_optab.
460 * match.pd (UNCOND_UNARY, COND_UNARY): New operator lists.
461 (vec_cond COND (foo A) B) -> (IFN_COND_FOO COND A B): New pattern.
462 (vec_cond COND B (foo A)) -> (IFN_COND_FOO ~COND A B): Likewise.
464 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
466 * tree-ssa-strlen.c (class strlen_pass): Rename from
468 (handle_builtin_stxncpy_strncat): Move to strlen_pass.
469 (handle_assign): Same.
470 (adjust_last_stmt): Same.
471 (maybe_warn_overflow): Same.
472 (handle_builtin_strlen): Same.
473 (handle_builtin_strchr): Same.
474 (handle_builtin_strcpy): Same.
475 (handle_builtin_strncat): Same.
476 (handle_builtin_stxncpy_strncat): Same.
477 (handle_builtin_memcpy): Same.
478 (handle_builtin_strcat): Same.
479 (handle_alloc_call): Same.
480 (handle_builtin_memset): Same.
481 (handle_builtin_memcmp): Same.
482 (get_len_or_size): Same.
483 (strxcmp_eqz_result): Same.
484 (handle_builtin_string_cmp): Same.
485 (handle_pointer_plus): Same.
486 (count_nonzero_bytes_addr): Same.
487 (count_nonzero_bytes): Same.
488 (handle_store): Same.
489 (strlen_check_and_optimize_call): Same.
490 (handle_integral_assign): Same.
491 (check_and_optimize_stmt): Same.
492 (printf_strlen_execute): Rename strlen_dom_walker to strlen_pass.
494 2021-10-18 Richard Biener <rguenther@suse.de>
496 PR tree-optimization/102798
497 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
498 Only copy points-to info to newly generated SSA names.
500 2021-10-18 Martin Liska <mliska@suse.cz>
502 * dbgcnt.c (dbg_cnt_process_opt): Remove unused but set variable.
503 * gcov.c (get_cycles_count): Likewise.
504 * lto-compress.c (lto_compression_zlib): Likewise.
505 (lto_uncompression_zlib): Likewise.
506 * targhooks.c (default_pch_valid_p): Likewise.
508 2021-10-17 Aldy Hernandez <aldyh@redhat.com>
510 * tree-pass.h (make_pass_thread_jumps_full): New.
511 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Inline.
512 (try_thread_blocks): Add resolve and speed arguments.
513 (pass_thread_jumps::execute): Inline.
514 (do_early_thread_jumps): New.
515 (do_thread_jumps): New.
516 (make_pass_thread_jumps): Move.
517 (pass_early_thread_jumps::gate): Inline.
518 (pass_early_thread_jumps::execute): Inline.
519 (class pass_thread_jumps_full): New.
521 2021-10-16 Piotr Kubaj <pkubaj@FreeBSD.org>
523 * configure.ac: Treat powerpc64*-*-freebsd* the same as
524 powerpc64-*-freebsd*.
525 * configure: Regenerate.
527 2021-10-16 H.J. Lu <hjl.tools@gmail.com>
529 * value-query.cc (get_ssa_name_ptr_info_nonnull): Change
530 set_ptr_nonull to set_ptr_nonnull in comments.
532 2021-10-16 Jan Hubicka <hubicka@ucw.cz>
534 PR tree-optimization/102720
535 * tree-ssa-structalias.c (compute_points_to_sets): Fix producing
536 of call used and clobbered sets.
538 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
540 * config/darwin.h (LINK_COMMAND_SPEC_A): Update 'r' handling to
541 skip gomp and itm when r or nodefaultlibs is given.
542 (DSYMUTIL_SPEC): Do not call dsymutil for '-r' link lines.
543 Update ordering of exclusions, remove duplicate 'v' addition
544 (collect2 will add this from the main command line).
546 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
548 * config/darwin-driver.c (darwin_driver_init): Revise comments, handle
549 filelist and framework options in specs instead of code.
550 * config/darwin.h (SUBTARGET_DRIVER_SELF_SPECS): Update to handle link
551 specs that are really driver ones.
552 (DARWIN_CC1_SPEC): Likewise.
553 (CPP_SPEC): Likewise.
554 (SYSROOT_SPEC): Append space.
555 (LINK_SYSROOT_SPEC): Remove most driver link specs.
556 (STANDARD_STARTFILE_PREFIX_2): Update link-related specs.
557 (STARTFILE_SPEC): Likewise.
558 (ASM_MMACOSX_VERSION_MIN_SPEC): Fix line wrap.
559 (ASM_SPEC): Update driver-related specs.
560 (ASM_FINAL_SPEC): Likewise.
561 * config/darwin.opt: Remove now unused option aliases.
562 * config/i386/darwin.h (EXTRA_ASM_OPTS): Ensure space after opt.
563 (ASM_SPEC): Update driver-related specs.
565 2021-10-15 Roger Sayle <roger@nextmovesoftware.com>
567 * config/i386/i386.c (ix86_hardreg_mov_ok): For vector modes,
568 allow standard_sse_constant_p immediate constants.
570 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
572 * config.gcc: Add tm-dwarf2.h to tm_d-file.
574 2021-10-15 Andrew MacLeod <amacleod@redhat.com>
576 * gimple-range-fold.h (gimple_range_ssa_p): Don't process names
577 that occur in abnormal phis.
578 * gimple-range.cc (gimple_ranger::range_on_edge): Return false for
579 abnormal and EH edges.
580 * gimple-ssa-evrp.c (rvrp_folder::value_of_expr): Ditto.
581 (rvrp_folder::value_on_edge): Ditto.
582 (rvrp_folder::value_of_stmt): Ditto.
583 (hybrid_folder::value_of_expr): Ditto for ranger queries.
584 (hybrid_folder::value_on_edge): Ditto.
585 (hybrid_folder::value_of_stmt): Ditto.
586 * value-query.cc (gimple_range_global): Always return a range if
587 the type is supported.
589 2021-10-15 John David Anglin <danglin@gcc.gnu.org>
591 * config/pa/pa.md: Consistently use "rG" constraint for copy
592 instruction in move patterns.
594 2021-10-15 Andrew Stubbs <ams@codesourcery.com>
596 * config/gcn/gcn-hsa.h (S_FIJI): Set unconditionally.
599 * config/gcn/gcn.c: Hard code SRAM ECC settings for old architectures.
600 * config/gcn/mkoffload.c (ELFABIVERSION_AMDGPU_HSA): Rename to ...
601 (ELFABIVERSION_AMDGPU_HSA_V3): ... this.
602 (ELFABIVERSION_AMDGPU_HSA_V4): New.
603 (SET_SRAM_ECC_UNSUPPORTED): New.
604 (copy_early_debug_info): Create elf flags to match the other objects.
605 (main): Just let the attribute flags pass through.
607 2021-10-15 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
609 * tree-loop-distribution.c (reduction_var_overflows_first):
610 Pass the type of reduction_var as first argument as it is also
611 done for the load type.
612 (loop_distribution::transform_reduction_loop): Add missing
613 TREE_TYPE while determining precission of reduction_var.
615 2021-10-15 Richard Biener <rguenther@suse.de>
617 * defaults.h (PREFERRED_DEBUGGING_TYPE): Choose DWARF2_DEBUG
619 * toplev.c (process_options): Warn when STABS debugging is
620 enabled but not the preferred format.
621 * config/pa/som.h (PREFERRED_DEBUGGING_TYPE): Define to
623 * config/pdp11/pdp11.h (PREFERRED_DEBUGGING_TYPE): Likewise.
625 2021-10-15 Richard Biener <rguenther@suse.de>
628 * tree-inline.c (copy_bb): Avoid underflowing nargs.
630 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
632 * config/i386/i386-expand.c (ix86_expand_vec_perm): Convert
633 HFmode input operand to HImode.
634 (ix86_vectorize_vec_perm_const): Likewise.
635 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1_hf):
637 (*avx512f_permvar_truncv8siv8hi_1_hf):
640 2021-10-15 Richard Biener <rguenther@suse.de>
643 * expmed.c (store_bit_field_1): Ensure a LHS subreg would
644 not create a paradoxical subreg.
646 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
648 * config/i386/i386-expand.c (ix86_expand_vector_init):
649 For half_vector concat for HFmode, handle them like HImode.
651 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
653 * tree-ssa-threadbackward.c (class back_threader): Add m_resolve.
654 (back_threader::back_threader): Same.
655 (back_threader::resolve_phi): Try to solve without looking back if
657 (back_threader::find_paths_to_names): Same.
658 (try_thread_blocks): Pass resolve argument to back threader.
659 (pass_early_thread_jumps::execute): Same.
661 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
663 * doc/invoke.texi: Remove max-fsm-thread-length,
664 max-fsm-thread-paths, and fsm-maximum-phi-arguments.
666 * tree-ssa-threadbackward.c (back_threader::back_threader): Remove
668 (back_threader_registry::back_threader_registry): Same.
669 (back_threader_profitability::profitable_path_p): Remove
670 param_max_fsm_thread-length.
671 (back_threader_registry::register_path): Remove
672 m_max_allowable_paths.
674 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
676 * tree-ssa-threadbackward.c (class back_threader): Make m_imports
678 (back_threader::~back_threader): Do not release m_path.
680 2021-10-14 Andrew MacLeod <amacleod@redhat.com>
682 PR tree-optimization/102738
683 * vr-values.c (simplify_using_ranges::simplify): Handle RSHIFT_EXPR.
685 2021-10-14 Kwok Cheung Yeung <kcy@codesourcery.com>
687 * omp-general.c (omp_check_context_selector): Move from c-omp.c.
688 (omp_mark_declare_variant): Move from c-omp.c.
689 (omp_context_name_list_prop): Update for Fortran strings.
690 * omp-general.h (omp_check_context_selector): New prototype.
691 (omp_mark_declare_variant): New prototype.
693 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
695 * config/rs6000/rs6000.c (rs6000_density_test): Move early
696 exit test further up the function.
698 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
700 * config/arm/arm.c (arm_add_stmt_cost): Delete.
701 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
703 2021-10-14 Martin Jambor <mjambor@suse.cz>
705 * doc/invoke.texi (Optimize Options): Add entry for
706 ipa-cp-recursive-freq-factor.
708 2021-10-14 Tamar Christina <tamar.christina@arm.com>
710 * match.pd: New rule.
712 2021-10-14 Jan Hubicka <hubicka@ucw.cz>
715 * ipa-modref-tree.h (modref_access_node::update2):
716 Also check that parm_offset is unchanged.
717 (modref_ref_node::insert_access): Fix updating of
720 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
722 * tree-ssa-threadbackward.c (back_threader::resolve_phi): Add
725 2021-10-14 Richard Biener <rguenther@suse.de>
727 PR tree-optimization/102659
728 * tree-if-conv.c (if_convertible_gimple_assign_stmt_p): Also
729 rewrite pointer typed undefined overflow operations.
730 (predicate_statements): Likewise. Make sure to emit invariant
731 conversions in the preheader.
732 * tree-vectorizer.c (vect_loop_vectorized_call): Look through
733 non-empty preheaders.
734 * tree-data-ref.c (dr_analyze_indices): Strip useless
735 conversions to the MEM_REF base type.
737 2021-10-14 Martin Liska <mliska@suse.cz>
739 * common.opt: Stop using AUTODETECT_VALUE
740 and use EnabledBy where possible.
741 * opts.c: Enable OPT_fvar_tracking with optimize >= 1.
742 * toplev.c (AUTODETECT_VALUE): Remove macro.
743 (process_options): Simplify by using EnabledBy and
744 OPT_fvar_tracking. Use OPTION_SET_P macro instead of
747 2021-10-14 Jonathan Wright <jonathan.wright@arm.com>
749 * config/aarch64/arm_neon.h (vld1_s8_x3): Use signed type for
751 (vld1_s32_x3): Likewise.
753 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
755 PR tree-optimization/102736
756 PR tree-optimization/102736
757 * gimple-range-path.cc (path_range_query::range_on_path_entry):
758 Assert that the requested range is defined outside the path.
759 (path_range_query::ssa_range_in_phi): Do not call
760 range_on_path_entry for SSA names that are defined within the
763 2021-10-14 Iain Sandoe <iain@sandoe.co.uk>
765 * config/darwin-driver.c (darwin_find_version_from_kernel):
766 Quote internal identifiers and avoid contractions in
768 (darwin_default_min_version): Likewise.
769 (darwin_driver_init): Likewise.
771 2021-10-14 Martin Jambor <mjambor@suse.cz>
773 * params.opt (ipa-cp-recursive-freq-factor): New.
774 * ipa-cp.c (ipcp_value): Switch to inline initialization. New members
775 scc_no, self_recursion_generated_level, same_scc and
776 self_recursion_generated_p.
777 (ipcp_lattice::add_value): Replaced parameter unlimited with
778 same_lat_gen_level, usit it determine limit of values and store it to
780 (ipcp_lattice<valtype>::print): Dump the new fileds.
781 (allocate_and_init_ipcp_value): Take same_lat_gen_level as a new
782 parameter and store it to the new value.
783 (self_recursively_generated_p): Removed.
784 (propagate_vals_across_arith_jfunc): Use self_recursion_generated_p
785 instead of self_recursively_generated_p, store self generation level
787 (value_topo_info<valtype>::add_val): Set scc_no.
788 (value_topo_info<valtype>::propagate_effects): Multiply frequencies of
789 recursively feeding values and self generated values by appropriate
792 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
794 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Remove
795 redundant test for flag_vect_cost_model.
797 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
799 * bitmap.c (debug): New overloaded function for auto_bitmaps.
800 * bitmap.h (debug): Same.
802 2021-10-14 Hongyu Wang <hongyu.wang@intel.com>
804 * config/i386/avx512fp16intrin.h (_mm512_mask_fcmadd_pch):
806 (_mm512_mask3_fcmadd_pch): Likewise.
807 (_mm512_mask_fmadd_pch): Likewise
808 (_mm512_mask3_fmadd_pch): Likewise
809 (_mm512_mask_fcmadd_round_pch): Likewise
810 (_mm512_mask3_fcmadd_round_pch): Likewise
811 (_mm512_mask_fmadd_round_pch): Likewise
812 (_mm512_mask3_fmadd_round_pch): Likewise
813 (_mm_mask_fcmadd_sch): Likewise
814 (_mm_mask3_fcmadd_sch): Likewise
815 (_mm_mask_fmadd_sch): Likewise
816 (_mm_mask3_fmadd_sch): Likewise
817 (_mm_mask_fcmadd_round_sch): Likewise
818 (_mm_mask3_fcmadd_round_sch): Likewise
819 (_mm_mask_fmadd_round_sch): Likewise
820 (_mm_mask3_fmadd_round_sch): Likewise
821 (_mm_fcmadd_round_sch): Likewise
822 * config/i386/avx512fp16vlintrin.h (_mm_mask_fmadd_pch):
824 (_mm_mask3_fmadd_pch): Likewise
825 (_mm256_mask_fmadd_pch): Likewise
826 (_mm256_mask3_fmadd_pch): Likewise
827 (_mm_mask_fcmadd_pch): Likewise
828 (_mm_mask3_fcmadd_pch): Likewise
829 (_mm256_mask_fcmadd_pch): Likewise
830 (_mm256_mask3_fcmadd_pch): Likewise
831 * config/i386/i386-builtin.def: Add mask3 builtin for complex
832 fma, and adjust mask_builtin to corresponding expander.
833 * config/i386/i386-expand.c (ix86_expand_round_builtin):
834 Skip eraseing embedded rounding for expanders that emits
836 * config/i386/sse.md (complexmove): New mode_attr.
837 (<avx512>_fmaddc_<mode>_mask1<round_expand_name>): New expander.
838 (<avx512>_fcmaddc_<mode>_mask1<round_expand_name>): Likewise.
839 (avx512fp16_fmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
840 (avx512fp16_fcmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
841 (avx512fp16_fcmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
842 (avx512fp16_fmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
843 * config/i386/subst.md (round_embedded_complex): New subst.
845 2021-10-14 John David Anglin <danglin@gcc.gnu.org>
847 * config/pa/pa.md (cbranchsf4): Disable if TARGET_SOFT_FLOAT.
848 (cbranchdf4): Likewise.
849 Add missing move patterns for TARGET_SOFT_FLOAT.
851 2021-10-13 Roger Sayle <roger@nextmovesoftware.com>
853 * config/i386/i386-expand.c (ix86_expand_vector_move): Use a
854 pseudo intermediate when moving a SUBREG into a hard register,
855 by checking ix86_hardreg_mov_ok.
856 (ix86_expand_vector_extract): Store zero-extended SImode
857 intermediate in a pseudo, then set target using a SUBREG_PROMOTED
859 * config/i386/sse.md (mov<VMOVE>_internal): Prevent CSE creating
860 complex (SUBREG) sets of (vector) hard registers before reload, by
861 checking ix86_hardreg_mov_ok.
863 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
865 * ctfc.h (enum ctf_dtu_d_union_enum): Remove redundant comma.
867 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
869 * dwarf2ctf.c (gen_ctf_array_type): Fix typo in comment.
871 2021-10-13 Martin Sebor <msebor@redhat.com>
874 * pointer-query.cc (compute_objsize_r): Handle named address spaces.
876 2021-10-13 Iain Sandoe <iain@sandoe.co.uk>
878 * collect2.c (is_lto_object_file): Release simple-object
879 resources, close files.
881 2021-10-13 Andre Vieira <andre.simoesdiasvieira@arm.com>
883 * config/arm/arm.c (thumb2_legitimate_address_p): Use VALID_MVE_MODE
884 when checking mve addressing modes.
885 (mve_vector_mem_operand): Fix the way we handle pre, post and offset
887 (arm_print_operand): Fix printing of POST_ and PRE_MODIFY.
888 * config/arm/mve.md: Use mve_memory_operand predicate everywhere where
889 there is a single Ux constraint.
891 2021-10-13 John David Anglin <danglin@gcc.gnu.org>
893 * config/pa/pa.md (muldi3): Add support for inlining 64-bit
894 multiplication on 32-bit PA 1.1 and 2.0 targets.
896 2021-10-13 Martin Liska <mliska@suse.cz>
898 PR gcov-profile/90364
899 * coverage.c (build_info): Emit checksum to the global variable.
900 (build_info_type): Add new field for checksum.
901 (coverage_obj_finish): Pass object_checksum.
902 (coverage_init): Use 0 as checksum for .gcno files.
903 * gcov-dump.c (dump_gcov_file): Dump also new checksum field.
904 * gcov.c (read_graph_file): Read also checksum.
905 * doc/invoke.texi: Document the behaviour change.
907 2021-10-13 Richard Biener <rguenther@suse.de>
909 * gimple-iterator.h (gsi_iterator_update): Add GSI_LAST_NEW_STMT,
910 start at integer value 2.
911 * gimple-iterator.c (gsi_insert_seq_nodes_before): Update
912 the iterator for GSI_LAST_NEW_STMT.
913 (gsi_insert_seq_nodes_after): Likewise.
914 * tree-if-conv.c (predicate_statements): Use GSI_LAST_NEW_STMT.
915 * tree-ssa.c (execute_update_addresses_taken): Correct bogus
916 arguments to gsi_replace.
918 2021-10-13 Martin Liska <mliska@suse.cz>
921 * common.opt: Use EnabledBy instead of detection in
922 finish_options and process_options.
923 * opts.c (finish_options): Remove handling of
924 x_flag_unroll_all_loops.
925 * toplev.c (process_options): Likewise for flag_web and
926 flag_rename_registers.
928 2021-10-13 Richard Biener <rguenther@suse.de>
930 PR tree-optimization/102659
931 * tree-if-conv.c (need_to_rewrite_undefined): New flag.
932 (if_convertible_gimple_assign_stmt_p): Mark the loop for
933 rewrite when stmts with undefined behavior on integer
935 (combine_blocks): Predicate also when we need to rewrite stmts.
936 (predicate_statements): Rewrite affected stmts to something
937 with well-defined behavior on overflow.
938 (tree_if_conversion): Initialize need_to_rewrite_undefined.
940 2021-10-13 Richard Biener <rguenther@suse.de>
943 * ipa-sra.c (ptr_parm_has_nonarg_uses): Fix volatileness
946 2021-10-13 Jose E. Marchesi <jose.marchesi@oracle.com>
948 * dwarf2ctf.c: Fix typo in comment.
950 2021-10-12 Bill Schmidt <wschmidt@linux.ibm.com>
953 * config/rs6000/altivec.h (vec_cpsgn): Swap operand order.
954 * config/rs6000/rs6000-overload.def (VEC_COPYSIGN): Use SKIP to
955 avoid generating an automatic #define of vec_cpsgn. Use the
956 correct built-in for V4SFmode that doesn't depend on VSX.
958 2021-10-12 Uroš Bizjak <ubizjak@gmail.com>
962 * config/i386/i386.md (*add<mode>_1_slp): Rewrite as
963 define_insn_and_split pattern. Add alternative 1 and split it
964 post reload to insert operand 1 into the low part of operand 0.
965 (*sub<mode>_1_slp): Ditto.
966 (*and<mode>_1_slp): Ditto.
967 (*<any_or:code><mode>_1_slp): Ditto.
968 (*ashl<mode>3_1_slp): Ditto.
969 (*<any_shiftrt:insn><mode>3_1_slp): Ditto.
970 (*<any_rotate:insn><mode>3_1_slp): Ditto.
971 (*neg<mode>_1_slp): New insn_and_split pattern.
972 (*one_cmpl<mode>_1_slp): Ditto.
974 2021-10-12 David Edelsohn <dje.gcc@gmail.com>
976 * doc/install.texi: Update MinGW and mingw-64 Binaries
979 2021-10-12 Daniel Le Duc Khoi Nguyen <greenrecyclebin@gmail.com>
981 * doc/extend.texi (Common Variable Attributes): Fix typos in
982 alloc_size documentation.
984 2021-10-12 Richard Biener <rguenther@suse.de>
986 PR tree-optimization/102696
987 * tree-vect-slp.c (vect_build_slp_tree_2): Properly mark
988 the tree fatally failed when we reject a BIT_FIELD_REF.
990 2021-10-12 Richard Biener <rguenther@suse.de>
992 PR tree-optimization/102572
993 * tree-vect-stmts.c (vect_build_gather_load_calls): When
994 gathering the vectorized defs for the mask pass in the
995 desired mask vector type so invariants will be handled
998 2021-10-12 Tamar Christina <tamar.christina@arm.com>
1000 * config/aarch64/aarch64-sve.md (*fcm<cmp_op><mode>_bic_combine,
1001 *fcm<cmp_op><mode>_nor_combine, *fcmuo<mode>_bic_combine,
1002 *fcmuo<mode>_nor_combine): New.
1004 2021-10-12 Eric Botcazou <ebotcazou@adacore.com>
1007 * config/sparc/sparc-modes.def (OI): New integer mode.
1009 2021-10-12 Jakub Jelinek <jakub@redhat.com>
1011 * gimple-fold.h (clear_padding_type_may_have_padding_p): Declare.
1012 * gimple-fold.c (clear_padding_type_may_have_padding_p): No longer
1015 2021-10-12 Jakub Jelinek <jakub@redhat.com>
1017 * tree-vectorizer.h (loop_cost_model): New function.
1018 (unlimited_cost_model): Use it.
1019 * tree-vect-loop.c (vect_analyze_loop_costing): Use loop_cost_model
1020 call instead of flag_vect_cost_model.
1021 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
1022 (vect_prune_runtime_alias_test_list): Likewise. Also use it instead
1023 of flag_simd_cost_model.
1025 2021-10-12 liuhongt <hongtao.liu@intel.com>
1028 * config/i386/i386-expand.c (emit_reduc_half): Handle
1030 * config/i386/mmx.md (reduc_<code>_scal_v4qi): New expander.
1031 (reduc_plus_scal_v4qi): Ditto.
1033 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
1035 * config/rs6000/smmintrin.h (_mm_cmpeq_epi64, _mm_cmpgt_epi64,
1036 _mm_mullo_epi32, _mm_mul_epi32, _mm_packus_epi32): New.
1037 * config/rs6000/nmmintrin.h: Copy from i386, tweak to suit.
1039 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
1041 * config/rs6000/smmintrin.h (_mm_cvtepi8_epi16, _mm_cvtepi8_epi32,
1042 _mm_cvtepi8_epi64, _mm_cvtepi16_epi32, _mm_cvtepi16_epi64,
1043 _mm_cvtepi32_epi64, _mm_cvtepu8_epi16, _mm_cvtepu8_epi32,
1044 _mm_cvtepu8_epi64, _mm_cvtepu16_epi32, _mm_cvtepu16_epi64,
1045 _mm_cvtepu32_epi64): New.
1047 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
1049 * config/rs6000/smmintrin.h (_mm_test_all_zeros,
1050 _mm_test_all_ones, _mm_test_mix_ones_zeros): Rewrite as macro.
1052 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
1054 * config/rs6000/smmintrin.h (_mm_min_epi8, _mm_min_epu16,
1055 _mm_min_epi32, _mm_min_epu32, _mm_max_epi8, _mm_max_epu16,
1056 _mm_max_epi32, _mm_max_epu32): New.
1058 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
1060 * ipa-modref-tree.h (struct modref_access_node): Revert
1062 (struct modref_ref_node): Likewise.
1064 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
1066 * ipa-modref-tree.h (modref_tree::global_access_p): New member
1069 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
1070 ignore_stores_eaf_flags): Move to ipa-modref.h
1071 (remove_useless_eaf_flags): Remove early exit on NOCLOBBER.
1072 (modref_summary::global_memory_read_p): New member function.
1073 (modref_summary::global_memory_written_p): New member function.
1074 * ipa-modref.h (modref_summary::global_memory_read_p,
1075 modref_summary::global_memory_written_p): Declare.
1076 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
1077 ignore_stores_eaf_flags): move here.
1078 * tree-ssa-structalias.c: Include ipa-modref-tree.h, ipa-modref.h
1080 (handle_rhs_call): Rewrite.
1081 (handle_call_arg): New function.
1082 (determine_global_memory_access): New function.
1083 (handle_const_call): Remove
1084 (handle_pure_call): Remove
1085 (find_func_aliases_for_call): Update use of handle_rhs_call.
1086 (compute_points_to_sets): Handle global memory acccesses
1089 2021-10-11 Diane Meirowitz <diane.meirowitz@oracle.com>
1091 * doc/invoke.texi: Add link to UndefinedBehaviorSanitizer
1092 documentation, mention UBSAN_OPTIONS, similar to what is done
1093 for AddressSanitizer.
1095 2021-10-11 Richard Biener <rguenther@suse.de>
1097 PR middle-end/102683
1098 * internal-fn.c (expand_DEFERRED_INIT): Check for mode
1099 availability before building an integer type for storage
1102 2021-10-11 Richard Biener <rguenther@suse.de>
1104 PR middle-end/101480
1105 * gimple.c (gimple_call_fnspec): Do not mark operator new/delete
1108 2021-10-11 Martin Liska <mliska@suse.cz>
1110 * common.opt: Remove Init(2) for some options.
1111 * toplev.c (process_options): Do not use AUTODETECT_VALUE, but
1112 use rather OPTION_SET_P.
1114 2021-10-11 Martin Liska <mliska@suse.cz>
1116 * common.opt: Remove usage of IRA_REGION_AUTODETECT.
1117 * flag-types.h (enum ira_region): Likewise.
1118 * toplev.c (process_options): Use OPTION_SET_P instead of
1119 IRA_REGION_AUTODETECT.
1121 2021-10-11 Jakub Jelinek <jakub@redhat.com>
1123 * omp-low.c (omp_runtime_api_call): Handle omp_get_max_teams,
1124 omp_[sg]et_teams_thread_limit and omp_set_num_teams.
1126 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
1128 * config/s390/s390-protos.h (s390_rawmemchr): Add prototype.
1129 * config/s390/s390.c (s390_rawmemchr): New function.
1130 * config/s390/s390.md (rawmemchr<SINT:mode>): New expander.
1131 * config/s390/vector.md (@vec_vfees<mode>): Basically a copy of
1132 the pattern vfees<mode> from vx-builtins.md.
1133 * config/s390/vx-builtins.md (*vfees<mode>): Remove.
1135 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
1137 * builtins.c (get_memory_rtx): Change to external linkage.
1138 * builtins.h (get_memory_rtx): Add function prototype.
1139 * doc/md.texi (rawmemchr<mode>): Document.
1140 * internal-fn.c (expand_RAWMEMCHR): Define.
1141 * internal-fn.def (RAWMEMCHR): Add.
1142 * optabs.def (rawmemchr_optab): Add.
1143 * tree-loop-distribution.c (find_single_drs): Change return code
1144 behaviour by also returning true if no single store was found
1146 (loop_distribution::classify_partition): Respect the new return
1147 code behaviour of function find_single_drs.
1148 (loop_distribution::execute): Call new function
1149 transform_reduction_loop in order to replace rawmemchr or strlen
1150 like loops by calls into builtins.
1151 (generate_reduction_builtin_1): New function.
1152 (generate_rawmemchr_builtin): New function.
1153 (generate_strlen_builtin_1): New function.
1154 (generate_strlen_builtin): New function.
1155 (generate_strlen_builtin_using_rawmemchr): New function.
1156 (reduction_var_overflows_first): New function.
1157 (determine_reduction_stmt_1): New function.
1158 (determine_reduction_stmt): New function.
1159 (loop_distribution::transform_reduction_loop): New function.
1161 2021-10-11 Martin Liska <mliska@suse.cz>
1163 * tree.c (cl_option_hasher::hash): Use cl_optimization_hash
1164 and remove legacy hashing code.
1166 2021-10-11 Kito Cheng <kito.cheng@sifive.com>
1169 * builtins.c (maybe_emit_call_builtin___clear_cache): Allow
1170 CONST_INT for BEGIN and END, and use gcc_assert rather than
1173 2021-10-10 Jakub Jelinek <jakub@redhat.com>
1176 * var-tracking.c (add_stores): For cselib_sp_derived_value_p values
1177 use MO_VAL_SET if loc is not sp.
1179 2021-10-10 Andrew Pinski <apinski@marvell.com>
1181 PR tree-optimization/102622
1182 * match.pd: Swap the order of a?pow2cst:0 and a?-1:0 transformations.
1183 Swap the order of a?0:pow2cst and a?0:-1 transformations.
1185 2021-10-09 liuhongt <hongtao.liu@intel.com>
1188 * config/i386/i386-expand.c (ix86_valid_mask_cmp_mode): Handle
1190 (ix86_use_mask_cmp_p): Ditto.
1191 (ix86_expand_sse_movcc): Ditto.
1192 * config/i386/i386.md (setcc_hf_mask): New define_insn.
1193 (movhf_mask): Ditto.
1194 (UNSPEC_MOVCC_MASK): New unspec.
1195 * config/i386/sse.md (UNSPEC_PCMP): Move to i386.md.
1197 2021-10-08 Vladimir N. Makarov <vmakarov@redhat.com>
1199 PR rtl-optimization/102627
1200 * lra-constraints.c (split_reg): Use at least natural mode of hard reg.
1202 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
1204 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Grow
1207 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
1209 * value-range.cc (irange::debug): New.
1210 * value-range.h (irange::debug): New.
1212 2021-10-08 Richard Sandiford <richard.sandiford@arm.com>
1214 PR tree-optimization/102385
1215 * predict.h (change_edge_frequency): Declare.
1216 * predict.c (change_edge_frequency): New function.
1217 * tree-ssa-loop-manip.h (tree_transform_and_unroll_loop): Remove
1219 (tree_unroll_loop): Likewise.
1220 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Update accordingly.
1221 * tree-predcom.c (pcom_worker::tree_predictive_commoning_loop):
1223 * tree-ssa-loop-prefetch.c (loop_prefetch_arrays): Likewise.
1224 * tree-ssa-loop-manip.c (tree_unroll_loop): Likewise.
1225 (tree_transform_and_unroll_loop): Likewise. Use single_dom_exit
1226 to retrieve the exit edges. Make all the old profile update code
1227 conditional on !single_loop_p -- the case it was written for --
1228 and use a different approach for the single-loop case.
1230 2021-10-08 Martin Liska <mliska@suse.cz>
1232 * config/alpha/alpha.c (alpha_option_override): Use new macro
1234 * config/arc/arc.c (arc_override_options): Likewise.
1235 * config/arm/arm.c (arm_option_override): Likewise.
1236 * config/bfin/bfin.c (bfin_load_pic_reg): Likewise.
1237 * config/c6x/c6x.c (c6x_option_override): Likewise.
1238 * config/csky/csky.c: Likewise.
1239 * config/darwin.c (darwin_override_options): Likewise.
1240 * config/frv/frv.c (frv_option_override): Likewise.
1241 * config/i386/djgpp.h: Likewise.
1242 * config/i386/i386.c (ix86_stack_protect_guard): Likewise.
1243 (ix86_max_noce_ifcvt_seq_cost): Likewise.
1244 * config/ia64/ia64.c (ia64_option_override): Likewise.
1245 (ia64_override_options_after_change): Likewise.
1246 * config/m32c/m32c.c (m32c_option_override): Likewise.
1247 * config/m32r/m32r.c (m32r_init): Likewise.
1248 * config/m68k/m68k.c (m68k_option_override): Likewise.
1249 * config/microblaze/microblaze.c (microblaze_option_override): Likewise.
1250 * config/mips/mips.c (mips_option_override): Likewise.
1251 * config/nios2/nios2.c (nios2_option_override): Likewise.
1252 * config/nvptx/nvptx.c (nvptx_option_override): Likewise.
1253 * config/pa/pa.c (pa_option_override): Likewise.
1254 * config/riscv/riscv.c (riscv_option_override): Likewise.
1255 * config/rs6000/aix71.h: Likewise.
1256 * config/rs6000/aix72.h: Likewise.
1257 * config/rs6000/aix73.h: Likewise.
1258 * config/rs6000/rs6000.c (darwin_rs6000_override_options): Likewise.
1259 (rs6000_override_options_after_change): Likewise.
1260 (rs6000_linux64_override_options): Likewise.
1261 (glibc_supports_ieee_128bit): Likewise.
1262 (rs6000_option_override_internal): Likewise.
1263 (rs6000_file_start): Likewise.
1264 (rs6000_darwin_file_start): Likewise.
1265 * config/rs6000/rtems.h: Likewise.
1266 * config/rs6000/sysv4.h: Likewise.
1267 * config/rs6000/vxworks.h (SUB3TARGET_OVERRIDE_OPTIONS): Likewise.
1268 * config/s390/s390.c (s390_option_override): Likewise.
1269 * config/sh/linux.h: Likewise.
1270 * config/sh/netbsd-elf.h (while): Likewise.
1271 * config/sh/sh.c (sh_option_override): Likewise.
1272 * config/sol2.c (solaris_override_options): Likewise.
1273 * config/sparc/sparc.c (sparc_option_override): Likewise.
1274 * config/tilegx/tilegx.c (tilegx_option_override): Likewise.
1275 * config/visium/visium.c (visium_option_override): Likewise.
1276 * config/vxworks.c (vxworks_override_options): Likewise.
1277 * lto-opts.c (lto_write_options): Likewise.
1278 * omp-expand.c (expand_omp_simd): Likewise.
1279 * omp-general.c (omp_max_vf): Likewise.
1280 * omp-offload.c (oacc_xform_loop): Likewise.
1281 * opts.h (OPTION_SET_P): Likewise.
1282 * targhooks.c (default_max_noce_ifcvt_seq_cost): Likewise.
1283 * toplev.c (process_options): Likewise.
1284 * tree-predcom.c: Likewise.
1285 * tree-sra.c (analyze_all_variable_accesses): Likewise.
1287 2021-10-08 liuhongt <hongtao.liu@intel.com>
1290 * config/i386/i386.c (ix86_optab_supported_p):
1291 Return true for HFmode.
1292 * match.pd: Simplify (_Float16) ceil ((double) x) to
1293 __builtin_ceilf16 (a) when a is _Float16 type and
1294 direct_internal_fn_supported_p.
1296 2021-10-08 liuhongt <hongtao.liu@intel.com>
1299 * config/i386/i386-expand.c (emit_reduc_half): Hanlde V4HImode.
1300 * config/i386/mmx.md (reduc_plus_scal_v4hi): New.
1301 (reduc_<code>_scal_v4hi): New.
1303 2021-10-08 liuhongt <hongtao.liu@intel.com>
1305 * common.opt (ftree-vectorize): Add Var(flag_tree_vectorize).
1306 * doc/invoke.texi (Options That Control Optimization): Update
1308 * opts.c (default_options_table): Enable auto-vectorization at
1309 O2 with very-cheap cost model.
1310 (finish_options): Use cheap cost model for
1311 explicit -ftree{,-loop}-vectorize.
1313 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
1315 * ctfc.c (ctfc_delete_container): Free hash table contents.
1317 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
1319 * toplev.c (process_options): Do not warn for GNU GIMPLE.
1321 2021-10-07 Siddhesh Poyarekar <siddhesh@gotplt.org>
1323 * tree-object-size.c (addr_object_size,
1324 compute_builtin_object_size): Drop PDECL and POFF arguments.
1325 (addr_object_size): Adjust calls.
1326 * tree-object-size.h (compute_builtin_object_size): Drop PDECL
1329 2021-10-07 Roger Sayle <roger@nextmovesoftware.com>
1331 * rtl.def (SMUL_HIGHPART, UMUL_HIGHPART): New RTX codes for
1332 representing signed and unsigned high-part multiplication resp.
1333 * simplify-rtx.c (simplify_binary_operation_1) [SMUL_HIGHPART,
1334 UMUL_HIGHPART]: Simplify high-part multiplications by zero.
1335 [SS_PLUS, US_PLUS, SS_MINUS, US_MINUS, SS_MULT, US_MULT,
1336 SS_DIV, US_DIV]: Similar simplifications for saturating
1338 (simplify_const_binary_operation) [SS_PLUS, US_PLUS, SS_MINUS,
1339 US_MINUS, SS_MULT, US_MULT, SMUL_HIGHPART, UMUL_HIGHPART]:
1340 Implement compile-time evaluation for constant operands.
1341 * dwarf2out.c (mem_loc_descriptor): Skip SMUL_HIGHPART and
1343 * doc/rtl.texi (smul_highpart, umul_highpart): Document RTX codes.
1344 * doc/md.texi (smul@var{m}3_highpart, umul@var{m3}_highpart):
1345 Mention the new smul_highpart and umul_highpart RTX codes.
1346 * doc/invoke.texi: Silence @xref "compilation" warnings.
1348 2021-10-07 Martin Jambor <mjambor@suse.cz>
1351 * ipa-prop.c (ipa_edge_args_sum_t::duplicate): Also handle the
1352 case when the source reference description corresponds to a
1353 referance taken in a function src->caller is inlined to.
1355 2021-10-07 Jan Hubicka <hubicka@ucw.cz>
1358 * ipa-modref-tree.h (modref_access_node::contains_p): Handle offsets
1360 (modref_access_node::try_merge_with): Add sanity check that there
1361 are no redundant entries in the list.
1363 2021-10-07 Richard Biener <rguenther@suse.de>
1365 PR tree-optimization/102608
1366 * tree-ssa-sccvn.c (visit_stmt): Drop .DEFERRED_INIT to
1369 2021-10-07 Martin Liska <mliska@suse.cz>
1371 * toplev.c (toplev::main): Make
1372 save_opt_decoded_options a pointer type
1373 * toplev.h: Likewise.
1375 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
1377 * config/gcn/gcn-valu.md (gather<mode>_insn_2offsets<exec>): Apply
1378 HAVE_GCN_ASM_GLOBAL_LOAD_FIXED.
1379 (scatter<mode>_insn_2offsets<exec_scatter>): Likewise.
1381 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
1383 * config/gcn/gcn-hsa.h (SRAMOPT): Include the whole option string.
1384 Adjust for new -msram-ecc=any behaviour.
1385 (ASM_SPEC): Adjust -mxnack and -msram-ecc usage.
1386 * config/gcn/gcn.c (output_file_start): Implement -msram-ecc=any.
1387 * config/gcn/mkoffload.c (EF_AMDGPU_XNACK): Rename to ...
1388 (EF_AMDGPU_XNACK_V3): ... this.
1389 (EF_AMDGPU_SRAM_ECC): Rename to ...
1390 (EF_AMDGPU_SRAM_ECC_V3): ... this.
1391 (EF_AMDGPU_FEATURE_XNACK_V4): New.
1392 (EF_AMDGPU_FEATURE_XNACK_UNSUPPORTED_V4): New.
1393 (EF_AMDGPU_FEATURE_XNACK_ANY_V4): New.
1394 (EF_AMDGPU_FEATURE_XNACK_OFF_V4): New.
1395 (EF_AMDGPU_FEATURE_XNACK_ON_V4): New.
1396 (EF_AMDGPU_FEATURE_SRAMECC_V4): New.
1397 (EF_AMDGPU_FEATURE_SRAMECC_UNSUPPORTED_V4): New.
1398 (EF_AMDGPU_FEATURE_SRAMECC_ANY_V4): New.
1399 (EF_AMDGPU_FEATURE_SRAMECC_OFF_V4): New.
1400 (EF_AMDGPU_FEATURE_SRAMECC_ON_V4): New.
1401 (SET_XNACK_ON): New.
1402 (SET_XNACK_OFF): New.
1404 (SET_SRAM_ECC_ON): New.
1405 (SET_SRAM_ECC_ANY): New.
1406 (SET_SRAM_ECC_OFF): New.
1407 (TEST_SRAM_ECC_ANY): New.
1408 (TEST_SRAM_ECC_ON): New.
1409 (main): Implement HSACOv4 and -msram-ecc=any.
1411 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
1413 * config.in: Regenerate.
1414 * config/gcn/gcn-hsa.h (X_FIJI): New macro.
1418 (A_FIJI): Rename to ...
1420 (A_900): Rename to ...
1422 (A_906): Rename to ...
1424 (A_908): Rename to ...
1426 (SRAMOPT): New macro.
1427 (ASM_SPEC): Adjust xnack option usage.
1428 * config/gcn/gcn.c (output_file_start): Adjust amdgcn_target usage.
1429 * configure: Regenerate.
1430 * configure.ac: Detect LLVM assembler dialect.
1432 2021-10-07 Richard Biener <rguenther@suse.de>
1434 * tree-pretty-print.c (dump_generic_node): Do not elide
1435 printing '&' when dumping with -gimple.
1437 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
1439 * gimple-range-cache.cc (non_null_ref::adjust_range): Call new
1441 * gimple-range-fold.cc (adjust_pointer_diff_expr): Ditto.
1442 (adjust_imagpart_expr): Ditto.
1443 * value-range.cc (irange::irange_intersect): Call new routine if
1444 RHS is a single pair.
1445 (irange::intersect): New wide_int version.
1446 * value-range.h (class irange): New prototype.
1448 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
1450 * gimple-range-edge.cc (gimple_outgoing_range::gimple_outgoing_range):
1451 Add parameter to limit size when recognizing switches.
1452 (gimple_outgoing_range::edge_range_p): Check size limit.
1453 * gimple-range-edge.h (gimple_outgoing_range): Add size field.
1454 * gimple-range-gori.cc (gori_map::calculate_gori): Ignore switches
1455 that exceed the size limit.
1456 (gori_compute::gori_compute): Add initializer.
1457 * params.opt (evrp-switch-limit): New.
1458 * doc/invoke.texi: Update docs.
1460 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
1462 * value-range.h (irange::set_varying): Use TYPE_MIN_VALUE and
1463 TYPE_MAX_VALUE instead of creating new trees when possible.
1465 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
1467 * gimple-range-cache.cc (non_null_ref::adjust_range): Check for
1468 zero and non-zero more efficently.
1470 2021-10-06 Richard Biener <rguenther@suse.de>
1473 * dumpfile.h (TDF_GIMPLE_VAL): New.
1474 (dump_flag): Re-order and adjust TDF_* flags. Make
1475 the enum uint32_t. Use std::underlying_type in the
1477 (optgroup_flag): Likewise for the operator overloads.
1478 * tree-pretty-print.c (dump_generic_node): Wrap ADDR_EXPR
1479 in _Literal if TDF_GIMPLE_VAL.
1480 * gimple-pretty-print.c (dump_gimple_assign): Add
1481 TDF_GIMPLE_VAL to flags when dumping operands where only
1482 is_gimple_val are allowed.
1483 (dump_gimple_cond): Likewise.
1485 2021-10-06 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
1487 * gimple-isel.cc (gimple_expand_vec_cond_expr): Remove redundant if
1490 2021-10-05 qing zhao <qing.zhao@oracle.com>
1492 PR middle-end/102359
1493 * gimplify.c (gimplify_decl_expr): Not add initialization for an
1494 auto variable when it has been initialized by frontend.
1496 2021-10-05 Aldy Hernandez <aldyh@redhat.com>
1498 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
1501 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
1503 * common/config/avr/avr-common.c (avr_handle_option): Mark
1504 argument as ATTRIBUTE_UNUSED.
1506 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
1508 * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC):
1509 Undefine before redefinition.
1511 2021-10-05 Richard Biener <rguenther@suse.de>
1513 * toplev.c (no_backend): Remove global var.
1514 (process_options): Pass in no_backend, move post_options
1515 langhook call to toplev::main.
1516 (do_compile): Pass in no_backend, move process_options call
1518 (toplev::run_self_tests): Check no_backend at the caller.
1519 (toplev::main): Call post_options and process_options
1520 split out from do_compile, do self-tests only if
1521 no_backend is initialized.
1523 2021-10-05 Richard Biener <rguenther@suse.de>
1525 * tree-cfg.c (dump_function_to_file): Dump the UID of the
1526 function as part of the name when requested.
1527 * tree-pretty-print.c (dump_function_name): Dump the UID when
1528 requested and the langhook produced the actual name.
1530 2021-10-05 Richard Biener <rguenther@suse.de>
1532 PR middle-end/102587
1533 PR middle-end/102285
1534 * internal-fn.c (expand_DEFERRED_INIT): Fall back to
1535 zero-initialization as last resort, use the constant
1536 size as given by the DEFERRED_INIT argument to build
1539 2021-10-04 Marek Polacek <polacek@redhat.com>
1542 * doc/invoke.texi: Document -Warray-compare.
1544 2021-10-04 Richard Biener <rguenther@suse.de>
1546 * gimplify.c (is_var_need_auto_init): DECL_HARD_REGISTER
1547 variables are not to be initialized.
1549 2021-10-04 Richard Biener <rguenther@suse.de>
1551 * expr.h (non_mem_decl_p): Declare.
1552 (mem_ref_refers_to_non_mem_p): Likewise.
1553 * expr.c (non_mem_decl_p): Export.
1554 (mem_ref_refers_to_non_mem_p): Likewise.
1555 * internal-fn.c (expand_DEFERRED_INIT): Do not expand the LHS
1556 but check the base with mem_ref_refers_to_non_mem_p
1559 2021-10-04 Richard Biener <rguenther@suse.de>
1561 PR tree-optimization/102570
1562 * tree-ssa-sccvn.h (vn_reference_op_struct): Document
1563 we are using clique for the internal function code.
1564 * tree-ssa-sccvn.c (vn_reference_op_eq): Compare the
1565 internal function code.
1566 (print_vn_reference_ops): Print the internal function code.
1567 (vn_reference_op_compute_hash): Hash it.
1568 (copy_reference_ops_from_call): Record it.
1569 (visit_stmt): Remove the restriction around internal function
1571 (fully_constant_vn_reference_p): Use fold_const_call and handle
1573 (vn_reference_eq): Compare call return types.
1574 * tree-ssa-pre.c (create_expression_by_pieces): Handle
1575 generating calls to internal functions.
1576 (compute_avail): Remove the restriction around internal function
1579 2021-10-04 Aldy Hernandez <aldyh@redhat.com>
1581 PR tree-optimization/102560
1582 * gimple-ssa-warn-alloca.c (alloca_call_type): Remove static
1583 marker for invalid_range.
1585 2021-10-04 Richard Biener <rguenther@suse.de>
1587 PR middle-end/102587
1588 * internal-fn.c (expand_DEFERRED_INIT): Guard register
1589 initialization path an avoid initializing VLA registers
1592 2021-10-04 Eric Botcazou <ebotcazou@adacore.com>
1594 * config/rs6000/vxworks.h (TARGET_INIT_LIBFUNCS): Delete.
1596 2021-10-03 Martin Liska <mliska@suse.cz>
1598 * toplev.c (toplev::main): Check opt_index if it is a part
1601 2021-10-02 Aldy Hernandez <aldyh@redhat.com>
1603 PR tree-optimization/102563
1604 * range-op.cc (operator_lshift::op1_range): Do not clobber
1607 2021-10-02 Martin Liska <mliska@suse.cz>
1609 * toplev.c (toplev::main): save_decoded_options[0] is program
1610 name and so it should be skipped.
1612 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
1614 PR tree-optimization/102546
1615 * range-op.cc (operator_lshift::op1_range): Teach range-ops that
1616 X << Y is non-zero implies X is also non-zero.
1618 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1620 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
1622 * config/aarch64/aarch64-tune.md: Regenerate.
1623 * doc/invoke.texi: Update docs.
1625 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1627 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
1629 * config/aarch64/aarch64-tune.md: Regenerate.
1630 * doc/invoke.texi: Update docs.
1632 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1634 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
1636 * config/aarch64/aarch64-tune.md: Regenerate.
1637 * doc/invoke.texi: Update docs.
1639 2021-10-01 Martin Sebor <msebor@redhat.com>
1642 * doc/invoke.texi (-Waddress): Update.
1643 * gengtype.c (write_types): Avoid -Waddress.
1644 * poly-int.h (POLY_SET_COEFF): Avoid using null.
1646 2021-10-01 John David Anglin <danglin@gcc.gnu.org>
1649 * config/pa/pa.c (pa_option_override): Default to dwarf version 4
1652 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1654 * config/aarch64/aarch64.h (AARCH64_FL_V9): Update value.
1656 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
1658 * gimple-range-path.cc (path_range_query::compute_ranges): Use
1660 * gimple-range-path.h (class path_range_query): Remove shadowed
1662 (path_range_query::get_path_oracle): New.
1664 2021-10-01 Jakub Jelinek <jakub@redhat.com>
1665 Richard Biener <rguenther@suse.de>
1668 * doc/invoke.texi (-fsanitize=integer-divide-by-zero): Remove
1669 INT_MIN / -1 division detection from here ...
1670 (-fsanitize=signed-integer-overflow): ... and add it here.
1672 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1674 * config/aarch64/aarch64-arches.def (AARCH64_ARCH): Added
1676 * config/aarch64/aarch64.h (AARCH64_FL_V9): New.
1677 (AARCH64_FL_FOR_ARCH9): New flags for Armv9-A.
1678 (AARCH64_ISA_V9): New ISA flag.
1679 * doc/invoke.texi: Update docs.
1681 2021-10-01 Martin Liska <mliska@suse.cz>
1683 * toplev.c (toplev::main): Save decoded optimization options.
1684 * toplev.h (save_opt_decoded_options): New.
1685 * doc/extend.texi: Be more clear about optimize and target
1688 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
1690 * explow.c: Include langhooks.h.
1691 (set_stack_check_libfunc): Build a proper function type.
1693 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
1696 * config/i386/i386.c (legitimate_pic_address_disp_p): For PE-COFF do
1697 not return true for external weak function symbols in medium model.
1699 2021-10-01 Jakub Jelinek <jakub@redhat.com>
1701 * tree.h (OMP_CLAUSE_ORDER_REPRODUCIBLE): Define.
1702 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_ORDER>: Print
1703 reproducible: for OMP_CLAUSE_ORDER_REPRODUCIBLE.
1704 * omp-general.c (omp_extract_for_data): If OMP_CLAUSE_ORDER is seen
1705 without OMP_CLAUSE_ORDER_UNCONSTRAINED, overwrite sched_kind to
1706 OMP_CLAUSE_SCHEDULE_STATIC.
1708 2021-10-01 Richard Biener <rguenther@suse.de>
1710 PR middle-end/102518
1711 * tree-inline.c (setup_one_parameter): Avoid substituting
1712 an invariant into contexts where a GIMPLE register is not valid.
1714 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1716 * config/arm/arm-cpus.in: Add Cortex-R52+ CPU.
1717 * config/arm/arm-tables.opt: Regenerate.
1718 * config/arm/arm-tune.md: Regenerate.
1719 * doc/invoke.texi: Update docs.
1721 2021-09-30 Uroš Bizjak <ubizjak@gmail.com>
1724 * config/i386/i386.md
1725 (sign_extend:WIDE (any_logic:NARROW (memory, immediate)) splitters):
1728 2021-09-30 Tobias Burnus <tobias@codesourcery.com>
1730 * omp-low.c (omp_runtime_api_call): Add omp_aligned_{,c}alloc and
1731 omp_{c,re}alloc, fix omp_alloc/omp_free.
1733 2021-09-30 Martin Liska <mliska@suse.cz>
1735 * defaults.h (ASM_OUTPUT_ASCII): Do not hide global variable
1736 asm_out_file and stream directly to MYFILE.
1738 2021-09-30 Richard Biener <rguenther@suse.de>
1740 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
1741 Restore and fix condition under which we apply npeel to
1742 the DRs misalignment value.
1744 2021-09-30 Richard Biener <rguenther@suse.de>
1746 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
1747 Fix npeel check for variable amount of peeling.
1749 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
1751 * lto-wrapper.c (run_gcc): Plug snprintf overflow.
1753 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
1755 * gimple-range.cc (gimple_ranger::debug): New.
1756 * gimple-range.h (class gimple_ranger): Add debug.
1758 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
1760 PR middle-end/102519
1761 * tree-vrp.c (hybrid_threader::~hybrid_threader): Free m_query.
1763 2021-09-29 Indu Bhagat <indu.bhagat@oracle.com>
1766 * btfout.c (GTY): Add GTY (()) albeit for cosmetic only purpose.
1767 (btf_finalize): Empty the hash_map btf_var_ids.
1769 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
1771 * tree-vrp.c (thread_through_all_blocks): Return bool.
1772 (execute_vrp_threader): Return TODO_* flags.
1773 (pass_data_vrp_threader): Set todo_flags_finish to 0.
1775 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
1777 * timevar.def (TV_TREE_VRP_THREADER): New.
1778 * tree-vrp.c: Use TV_TREE_VRP_THREADER for VRP threader pass.
1780 2021-09-29 David Faust <david.faust@oracle.com>
1782 * config.gcc (bpf-*-*): Do not overwrite extra_headers.
1784 2021-09-29 Jonathan Wright <jonathan.wright@arm.com>
1786 * config/aarch64/aarch64-builtins.c (TYPES_BINOP_PPU): Define
1787 new type qualifier enum.
1788 (TYPES_TERNOP_SSSU): Likewise.
1789 (TYPES_TERNOP_PPPU): Likewise.
1790 * config/aarch64/aarch64-simd-builtins.def: Define PPU, SSU,
1791 PPPU and SSSU builtin generator macros for qtbl1 and qtbx1
1793 * config/aarch64/arm_neon.h (vqtbl1_p8): Use type-qualified
1794 builtin and remove casts.
1795 (vqtbl1_s8): Likewise.
1796 (vqtbl1q_p8): Likewise.
1797 (vqtbl1q_s8): Likewise.
1798 (vqtbx1_s8): Likewise.
1799 (vqtbx1_p8): Likewise.
1800 (vqtbx1q_s8): Likewise.
1801 (vqtbx1q_p8): Likewise.
1802 (vtbl1_p8): Likewise.
1803 (vtbl2_p8): Likewise.
1804 (vtbx2_p8): Likewise.
1806 2021-09-29 Richard Biener <rguenther@suse.de>
1808 * tree-vect-data-refs.c (vect_dr_misalign_for_aligned_access):
1810 (vect_update_misalignment_for_peel): Use it to update
1811 misaligned to the value necessary for an aligned access.
1812 (vect_get_peeling_costs_all_drs): Likewise.
1813 (vect_enhance_data_refs_alignment): Likewise.
1815 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1817 * config/aarch64/aarch64.c (aarch64_expand_cpymem): Count number of
1818 emitted operations and adjust heuristic for code size.
1820 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1822 * config/aarch64/aarch64.c (aarch64_expand_setmem): Count number of
1823 emitted operations and adjust heuristic for code size.
1825 2021-09-29 Jakub Jelinek <jakub@redhat.com>
1827 PR middle-end/102504
1828 * gimplify.c (gimplify_scan_omp_clauses): Use omp_check_private even
1829 in OMP_SCOPE clauses, not just on worksharing construct clauses.
1831 2021-09-28 Geng Qi <gengqi@linux.alibaba.com>
1833 * config/riscv/riscv.md (mulv<mode>4): Call gen_smul<mode>3_highpart.
1834 (<u>mulditi3): Call <su>muldi3_highpart.
1835 (<u>muldi3_highpart): Rename to <su>muldi3_highpart.
1836 (<u>mulsidi3): Call <su>mulsi3_highpart.
1837 (<u>mulsi3_highpart): Rename to <su>mulsi3_highpart.
1839 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
1841 * config/darwin.h (DSYMUTIL_SPEC): Recognize D sources.
1843 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
1845 * config/rs6000/darwin.h (FIXED_R13): Add for PPC64.
1846 (FIRST_SAVED_GP_REGNO): Save from R13 even when it is one
1849 2021-09-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1851 * config/aarch64/aarch64.h (AARCH64_FL_LS64): Define
1852 (AARCH64_FL_V8_7): Likewise.
1853 (AARCH64_FL_FOR_ARCH8_7): Likewise.
1854 * config/aarch64/aarch64-arches.def (armv8.7-a): Define.
1855 * config/aarch64/aarch64-option-extensions.def (ls64): Define.
1856 * doc/invoke.texi: Document the above.
1858 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
1860 * dbgcnt.c (dbg_cnt_counter): New.
1861 * dbgcnt.h (dbg_cnt_counter): New.
1862 * dumpfile.c (dump_options): Add entry for TDF_THREADING.
1863 * dumpfile.h (enum dump_flag): Add TDF_THREADING.
1864 * gimple-range-path.cc (DEBUG_SOLVER): Use TDF_THREADING.
1865 * tree-ssa-threadupdate.c (dump_jump_thread_path): Dump out
1868 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
1870 * cfgcleanup.c (pass_jump::execute): Check
1871 flag_expensive_optimizations.
1872 (pass_jump_after_combine::gate): Same.
1873 * doc/invoke.texi (-fthread-jumps): Enable for -O1.
1874 * opts.c (default_options_table): Enable -fthread-jumps at -O1.
1875 * tree-ssa-threadupdate.c
1876 (fwd_jt_path_registry::remove_jump_threads_including): Bail unless
1879 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
1881 * tree-ssa-reassoc.c (biased_names): New global.
1882 (propagate_bias_p): New function.
1883 (loop_carried_phi): Remove.
1884 (propagate_rank): Propagate bias along single uses.
1885 (get_rank): Update biased_names when needed.
1887 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
1889 * passes.def (pass_reassoc): Rename parameter to early_p.
1890 * tree-ssa-reassoc.c (reassoc_bias_loop_carried_phi_ranks_p):
1892 (phi_rank): Don't bias loop-carried phi ranks
1893 before vectorization pass.
1894 (execute_reassoc): Add bias_loop_carried_phi_ranks_p parameter.
1895 (pass_reassoc::pass_reassoc): Add bias_loop_carried_phi_ranks_p
1897 (pass_reassoc::set_param): Set bias_loop_carried_phi_ranks_p
1899 (pass_reassoc::execute): Pass bias_loop_carried_phi_ranks_p to
1901 (pass_reassoc::bias_loop_carried_phi_ranks_p): New member.
1903 2021-09-28 Jakub Jelinek <jakub@redhat.com>
1906 * config/i386/i386.c (standard_80387_constant_p): Don't recognize
1907 special 80387 instruction XFmode constants if flag_rounding_math.
1909 2021-09-28 Richard Biener <rguenther@suse.de>
1911 PR tree-optimization/100112
1912 * tree-ssa-sccvn.c (visit_reference_op_load): Record the
1913 referece into the hashtable twice in case last_vuse is
1914 different from the original vuse on the stmt.
1916 2021-09-28 Jakub Jelinek <jakub@redhat.com>
1918 PR middle-end/102492
1919 * gimplify.c (gimplify_adjust_omp_clauses_1): Don't call the
1920 omp_finish_clause langhook on implicitly added OMP_CLAUSE_PRIVATE
1921 clauses on SIMD constructs.
1923 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
1925 PR tree-optimization/102511
1926 * gimple-range-path.cc (path_range_query::range_on_path_entry):
1927 Return VARYING when nothing found.
1929 2021-09-28 Hongyu Wang <hongyu.wang@intel.com>
1932 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): Add
1934 (VALID_SSE2_REG_VHF_MODE): Add V4HFmode and V2HFmode.
1935 (VALID_MMX_REG_MODE): Add V4HFmode.
1936 (SSE_REG_MODE_P): Replace VALID_AVX512FP16_REG_MODE with
1937 vector mode condition.
1938 * config/i386/i386.c (classify_argument): Parse V4HF/V2HF
1940 (function_arg_32): Add V4HFmode.
1941 (function_arg_advance_32): Likewise.
1942 * config/i386/i386.md (mode): Add V4HF/V2HF.
1943 (MODE_SIZE): Likewise.
1944 * config/i386/mmx.md (MMXMODE): Add V4HF mode.
1945 (V_32): Add V2HF mode.
1946 (VHF_32_64): New mode iterator.
1947 (*mov<mode>_internal): Adjust sse alternatives to support
1949 (*mov<mode>_internal): Adjust sse alternatives to support
1951 (<insn><mode>3): New define_insn for add/sub/mul/div.
1953 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
1955 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Check
1957 (pass_early_thread_jumps::gate): Same.
1958 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
1959 Return if !flag_thread_jumps.
1960 * tree-ssa-threadupdate.c
1961 (jt_path_registry::register_jump_thread): Assert that
1962 flag_thread_jumps is true.
1964 2021-09-28 liuhongt <hongtao.liu@intel.com>
1967 (simplify_context::simplify_binary_operation_1): Relax
1968 condition of simplifying (vec_concat:M (vec_select op0
1969 index0)(vec_select op1 index1)) to allow different modes
1970 between op0 and M, but have same inner mode.
1972 2021-09-28 liuhongt <hongtao.liu@intel.com>
1974 * config/i386/i386-expand.c (emit_reduc_half): Handle
1975 V8HF/V16HF/V32HFmode.
1976 * config/i386/sse.md (REDUC_SSE_PLUS_MODE): Add V8HF.
1977 (REDUC_SSE_SMINMAX_MODE): Ditto.
1978 (REDUC_PLUS_MODE): Add V16HF and V32HF.
1979 (REDUC_SMINMAX_MODE): Ditto.
1981 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
1983 * gimple-range-path.cc
1984 (path_range_query::precompute_ranges_in_block): Rename to...
1985 (path_range_query::compute_ranges_in_block): ...this.
1986 (path_range_query::precompute_ranges): Rename to...
1987 (path_range_query::compute_ranges): ...this.
1988 (path_range_query::precompute_relations): Rename to...
1989 (path_range_query::compute_relations): ...this.
1990 (path_range_query::precompute_phi_relations): Rename to...
1991 (path_range_query::compute_phi_relations): ...this.
1992 * gimple-range-path.h: Rename precompute* to compute*.
1993 * tree-ssa-threadbackward.c
1994 (back_threader::find_taken_edge_switch): Same.
1995 (back_threader::find_taken_edge_cond): Same.
1996 * tree-ssa-threadedge.c
1997 (hybrid_jt_simplifier::compute_ranges_from_state): Same.
1998 (hybrid_jt_state::register_equivs_stmt): Inline...
1999 * tree-ssa-threadedge.h: ...here.
2001 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
2003 * tree-vrp.c (lhs_of_dominating_assert): Remove.
2004 (class vrp_jt_state): Remove.
2005 (class vrp_jt_simplifier): Remove.
2006 (vrp_jt_simplifier::simplify): Remove.
2007 (class vrp_jump_threader): Remove.
2008 (vrp_jump_threader::vrp_jump_threader): Remove.
2009 (vrp_jump_threader::~vrp_jump_threader): Remove.
2010 (vrp_jump_threader::before_dom_children): Remove.
2011 (vrp_jump_threader::after_dom_children): Remove.
2013 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
2015 * passes.def (pass_vrp_threader): New.
2016 * tree-pass.h (make_pass_vrp_threader): Add make_pass_vrp_threader.
2017 * tree-ssa-threadedge.c (hybrid_jt_state::register_equivs_stmt): New.
2018 (hybrid_jt_simplifier::hybrid_jt_simplifier): New.
2019 (hybrid_jt_simplifier::simplify): New.
2020 (hybrid_jt_simplifier::compute_ranges_from_state): New.
2021 * tree-ssa-threadedge.h (class hybrid_jt_state): New.
2022 (class hybrid_jt_simplifier): New.
2023 * tree-vrp.c (execute_vrp): Remove ASSERT_EXPR based jump
2025 (class hybrid_threader): New.
2026 (hybrid_threader::hybrid_threader): New.
2027 (hybrid_threader::~hybrid_threader): New.
2028 (hybrid_threader::before_dom_children): New.
2029 (hybrid_threader::after_dom_children): New.
2030 (execute_vrp_threader): New.
2031 (class pass_vrp_threader): New.
2032 (make_pass_vrp_threader): New.
2034 2021-09-27 Martin Liska <mliska@suse.cz>
2036 * output.h (enum section_flag): New.
2037 (SECTION_FORGET): Remove.
2038 (SECTION_ENTSIZE): Make it (1UL << 8) - 1.
2039 (SECTION_STYLE_MASK): Define it based on other enum
2041 * varasm.c (switch_to_section): Remove unused handling of
2044 2021-09-27 Martin Liska <mliska@suse.cz>
2046 * common.opt: Add new variable flag_default_complex_method.
2047 * opts.c (finish_options): Handle flags related to
2048 x_flag_complex_method.
2049 * toplev.c (process_options): Remove option handling related
2050 to flag_complex_method.
2052 2021-09-27 Richard Biener <rguenther@suse.de>
2054 PR middle-end/102450
2055 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid using
2056 type_for_size, instead use int_mode_for_size.
2058 2021-09-27 Andrew Pinski <apinski@marvell.com>
2061 * gimplify.c (gimplify_save_expr): Return early
2062 if the type of val is error_mark_node.
2064 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
2066 * tree-ssanames.c (ssa_name_has_boolean_range): Use
2069 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
2071 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
2072 vrp_visit_cond_stmt.
2073 * tree-ssa-dom.c (cprop_operand): Convert to range_query API.
2074 (cprop_into_stmt): Same.
2075 (dom_opt_dom_walker::optimize_stmt): Same.
2077 2021-09-27 Richard Biener <rguenther@suse.de>
2079 PR tree-optimization/97351
2080 PR tree-optimization/97352
2081 PR tree-optimization/82426
2082 * tree-vectorizer.h (dr_misalignment): Add vector type
2084 (aligned_access_p): Likewise.
2085 (known_alignment_for_access_p): Likewise.
2086 (vect_supportable_dr_alignment): Likewise.
2087 (vect_known_alignment_in_bytes): Likewise. Refactor.
2088 (DR_MISALIGNMENT): Remove.
2089 (vect_update_shared_vectype): Likewise.
2090 * tree-vect-data-refs.c (dr_misalignment): Refactor, handle
2091 a vector type with larger alignment requirement and apply
2092 the negative step adjustment here.
2093 (vect_calculate_target_alignment): Remove.
2094 (vect_compute_data_ref_alignment): Get explicit vector type
2095 argument, do not apply a negative step alignment adjustment
2097 (vect_slp_analyze_node_alignment): Re-analyze alignment
2098 when we re-visit the DR with a bigger desired alignment but
2099 keep more precise results from smaller alignments.
2100 * tree-vect-slp.c (vect_update_shared_vectype): Remove.
2101 (vect_slp_analyze_node_operations_1): Do not update the
2102 shared vector type on stmts.
2103 * tree-vect-stmts.c (vect_analyze_stmt): Push/pop the
2104 vector type of an SLP node to the representative stmt-info.
2105 (vect_transform_stmt): Likewise.
2107 2021-09-27 liuhongt <hongtao.liu@intel.com>
2110 2021-09-09 liuhongt <hongtao.liu@intel.com>
2113 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
2114 (reduc_plus_scal_v4sf): .. this, New define_expand.
2115 (reduc_plus_scal_v2df): .. and this, New define_expand.
2117 2021-09-26 liuhongt <hongtao.liu@intel.com>
2119 * doc/extend.texi (Half-Precision): Remove storage only
2120 description for _Float16 w/o avx512fp16.
2122 2021-09-25 Dimitar Dimitrov <dimitar@dinux.eu>
2124 * config/pru/constraints.md (Rrio): New constraint.
2125 * config/pru/predicates.md (regio_operand): New predicate.
2126 * config/pru/pru-pragma.c (pru_register_pragmas): Register
2127 the __regio_symbol address space.
2128 * config/pru/pru-protos.h (pru_symref2ioregno): Declaration.
2129 * config/pru/pru.c (pru_symref2ioregno): New helper function.
2130 (pru_legitimate_address_p): Remove.
2131 (pru_addr_space_legitimate_address_p): Use the address space
2133 (pru_nongeneric_pointer_addrspace): New helper function.
2134 (pru_insert_attributes): New function to validate __regio_symbol
2136 (TARGET_INSERT_ATTRIBUTES): New macro.
2137 (TARGET_LEGITIMATE_ADDRESS_P): Remove.
2138 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): New macro.
2139 * config/pru/pru.h (enum reg_class): Add REGIO_REGS class.
2140 * config/pru/pru.md (*regio_readsi): New pattern to read I/O
2142 (*regio_nozext_writesi): New pattern to write to I/O registers.
2143 (*regio_zext_write_r30<EQS0:mode>): Ditto.
2144 * doc/extend.texi: Document the new PRU Named Address Space.
2146 2021-09-24 Patrick Palka <ppalka@redhat.com>
2150 * real.c (encode_ieee_double): Avoid unwanted sign extension.
2151 (encode_ieee_quad): Likewise.
2153 2021-09-24 Vladimir Makarov <vmakarov@redhat.com>
2155 PR rtl-optimization/102147
2156 * ira-build.c (ira_conflict_vector_profitable_p): Make
2157 profitability calculation independent of host compiler pointer and
2160 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
2162 * gimple-range-path.cc (path_range_query::path_range_query):
2163 Move debugging header...
2164 (path_range_query::precompute_ranges): ...here.
2165 (path_range_query::internal_range_of_expr): Do not call
2166 range_on_path_entry if NAME is defined in the current block.
2168 2021-09-24 Richard Biener <rguenther@suse.de>
2170 * cfghooks.c (verify_flow_info): Verify unallocated BB and
2171 edge flags are not set.
2173 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
2175 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
2177 (jt_path_registry::register_jump_thread): Call
2178 cancel_invalid_paths.
2179 * tree-ssa-threadupdate.h (class jt_path_registry): Add
2180 cancel_invalid_paths.
2182 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
2184 PR tree-optimization/102400
2185 * tree-ssa-sccvn.c (vn_reference_insert_pieces): Initialize
2186 result_vdef to zero value.
2188 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
2190 PR tree-optimization/102451
2191 * tree-ssa-dse.c (delete_dead_or_redundant_call): Record bb of stmt
2194 2021-09-24 Hongyu Wang <hongyu.wang@intel.com>
2196 * config/i386/sse.md (cond_<insn><mode>): Extend to support
2198 (cond_mul<mode>): Likewise.
2199 (cond_div<mode>): Likewise.
2200 (cond_<code><mode>): Likewise.
2201 (cond_fma<mode>): Likewise.
2202 (cond_fms<mode>): Likewise.
2203 (cond_fnma<mode>): Likewise.
2204 (cond_fnms<mode>): Likewise.
2206 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
2208 PR tree-optimization/102463
2209 * gimple-range-fold.cc (fold_using_range::relation_fold_and_or): If
2210 there is no range-ops handler, don't look for a relation.
2212 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
2214 * gimple-range-cache.cc (ranger_cache::ranger_cache): Take
2215 non-executable_edge flag as parameter.
2216 * gimple-range-cache.h (ranger_cache): Adjust prototype.
2217 * gimple-range-gori.cc (gori_compute::gori_compute): Take
2218 non-executable_edge flag as parameter.
2219 (gori_compute::outgoing_edge_range_p): Check new flag.
2220 * gimple-range-gori.h (gori_compute): Adjust prototype.
2221 * gimple-range.cc (gimple_ranger::gimple_ranger): Create new flag.
2222 (gimple_ranger::range_on_edge): Check new flag.
2223 * gimple-range.h (gimple_ranger::non_executable_edge_flag): New.
2224 * gimple-ssa-evrp.c (rvrp_folder): Pass ranger flag to simplifer.
2225 (hybrid_folder::hybrid_folder): Set ranger non-executable flag value.
2226 (hybrid_folder::fold_stmt): Set flag value in the simplifer.
2227 * vr-values.c (simplify_using_ranges::set_and_propagate_unexecutable):
2228 Use not_executable flag if provided inmstead of EDGE_EXECUTABLE.
2229 (simplify_using_ranges::simplify_switch_using_ranges): Clear
2230 EDGE_EXECUTABLE like it originally did.
2231 (simplify_using_ranges::cleanup_edges_and_switches): Clear any
2232 NON_EXECUTABLE flags.
2233 (simplify_using_ranges::simplify_using_ranges): Adjust.
2234 * vr-values.h (class simplify_using_ranges): Adjust.
2235 (simplify_using_ranges::set_range_query): Add non-executable flag param.
2237 2021-09-23 Bill Schmidt <wschmidt@linux.ibm.com>
2240 * config/rs6000/rs6000-call.c (rs6000_aggregate_candidate): Detect
2241 zero-width bit fields and return indicator.
2242 (rs6000_discover_homogeneous_aggregate): Diagnose when the
2243 presence of a zero-width bit field changes parameter passing in
2246 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
2248 * gimple-range-fold.cc (fold_using_range::range_of_phi):
2249 Remove dominator check.
2251 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
2253 * gimple-range-path.cc (path_range_query::precompute_relations):
2254 Hoist edge calculations before using EDGE_SUCC.
2256 2021-09-23 Jonathan Wakely <jwakely@redhat.com>
2258 * configure.ac: Fix --with-multilib-list description.
2259 * configure: Regenerate.
2261 2021-09-23 Richard Biener <rguenther@suse.de>
2263 PR tree-optimization/102448
2264 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
2265 Clear alignment info copied from DR_PTR_INFO.
2267 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
2269 * config/i386/i386-expand.c (ix86_use_mask_cmp_p): Enable
2271 * config/i386/sse.md (sseintvecmodelower): Add HF vector modes.
2272 (<avx512>_store<mode>_mask): Extend to support HF vector modes.
2273 (vec_cmp<mode><avx512fmaskmodelower>): Likewise.
2274 (vcond_mask_<mode><avx512fmaskmodelower>): Likewise.
2275 (vcond<mode><mode>): New expander.
2276 (vcond<mode><sseintvecmodelower>): Likewise.
2277 (vcond<sseintvecmodelower><mode>): Likewise.
2278 (vcondu<mode><sseintvecmodelower>): Likewise.
2280 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
2282 * config/i386/sse.md (extend<ssePHmodelower><mode>2):
2284 (extendv4hf<mode>2): Likewise.
2285 (extendv2hfv2df2): Likewise.
2286 (trunc<mode><ssePHmodelower>2): Likewise.
2287 (avx512fp16_vcvt<castmode>2ph_<mode>): Rename to ...
2288 (trunc<mode>v4hf2): ... this, and drop constraints.
2289 (avx512fp16_vcvtpd2ph_v2df): Rename to ...
2290 (truncv2dfv2hf2): ... this, and likewise.
2292 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
2294 * config/i386/sse.md (float<floatunssuffix><mode><ssePHmodelower>2):
2296 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>):
2298 (float<floatunssuffix><mode>v4hf2): ... this, and drop constraints.
2299 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Rename to ...
2300 (float<floatunssuffix>v2div2hf2): ... this, and likewise.
2302 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
2304 * config/i386/i386.md (fix<fixunssuffix>_trunchf<mode>2): New expander.
2305 (fixuns_trunchfhi2): Likewise.
2306 (*fixuns_trunchfsi2zext): New define_insn.
2307 * config/i386/sse.md (ssePHmodelower): New mode_attr.
2308 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
2309 New expander for same element vector fix_truncate.
2310 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
2311 Likewise for V4HF to V4SI/V4DI fix_truncate.
2312 (fix<fixunssuffix>_truncv2hfv2di2):
2313 Likeise for V2HF to V2DI fix_truncate.
2315 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
2317 * config/i386/i386.md (<code>hf3): New expander.
2319 2021-09-23 liuhongt <hongtao.liu@intel.com>
2321 * config/i386/sse.md (FMAMODEM): extend to handle FP16.
2322 (VFH_SF_AVX512VL): Extend to handle HFmode.
2323 (VF_SF_AVX512VL): Deleted.
2325 2021-09-23 liuhongt <hongtao.liu@intel.com>
2327 * config/i386/i386.md (rinthf2): New expander.
2328 (nearbyinthf2): New expander.
2330 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
2332 * tree-ssa-dom.c (class dom_jump_threader_simplifier): Rename...
2333 (class dom_jt_state): ...this and provide virtual overrides.
2334 (dom_jt_state::register_equiv): New.
2335 (class dom_jt_simplifier): Rename from
2336 dom_jump_threader_simplifier.
2337 (dom_jump_threader_simplifier::simplify): Rename...
2338 (dom_jt_simplifier::simplify): ...to this.
2339 (pass_dominator::execute): Use dom_jt_simplifier and
2341 * tree-ssa-threadedge.c (jump_threader::jump_threader):
2343 (jt_state::register_equivs_stmt): Abstract out...
2344 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
2346 (jump_threader::thread_around_empty_blocks): Update state.
2347 (jump_threader::thread_through_normal_block): Same.
2348 (jt_state::jt_state): Remove.
2349 (jt_state::push): Remove pass specific bits. Keep block vector
2351 (jt_state::append_path): New.
2352 (jt_state::pop): Remove pass specific bits.
2353 (jt_state::register_equiv): Same.
2354 (jt_state::record_ranges_from_stmt): Same.
2355 (jt_state::register_equivs_on_edge): Same. Rename...
2356 (jt_state::register_equivs_edge): ...to this.
2357 (jt_state::dump): New.
2358 (jt_state::debug): New.
2359 (jump_threader_simplifier::simplify): Remove.
2360 (jt_state::get_path): New.
2361 * tree-ssa-threadedge.h (class jt_simplifier): Make into a base
2362 class. Expose common functionality as virtual methods.
2363 (class jump_threader_simplifier): Same. Rename...
2364 (class jt_simplifier): ...to this.
2365 * tree-vrp.c (class vrp_jump_threader_simplifier): Rename...
2366 (class vrp_jt_simplifier): ...to this. Provide pass specific
2368 (class vrp_jt_state): New.
2369 (vrp_jump_threader_simplifier::simplify): Rename...
2370 (vrp_jt_simplifier::simplify): ...to this. Inline code from
2371 what used to be the base class.
2372 (vrp_jump_threader::vrp_jump_threader): Use vrp_jt_state and
2375 2021-09-22 Tobias Burnus <tobias@codesourcery.com>
2378 * doc/invoke.texi (-Wno-missing-include-dirs.): Document Fortran
2381 2021-09-22 Roger Sayle <roger@nextmovesoftware.com>
2382 Richard Biener <rguenther@suse.de>
2384 * match.pd (negation simplifications): Implement some negation
2385 folding transformations from fold-const.c's fold_negate_expr.
2386 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Add a SIMPLIFY
2387 argument, to control whether the op should be simplified prior
2388 to looking up/assigning a value number.
2389 (vn_nary_build_or_lookup): Update call to vn_nary_build_or_lookup_1.
2390 (vn_nary_simplify): Likewise.
2391 (visit_nary_op): Likewise, but when constructing a NEGATE_EXPR
2392 now call vn_nary_build_or_lookup_1 disabling simplification.
2394 2021-09-22 Jiufu Guo <guojiufu@linux.ibm.com>
2396 PR tree-optimization/102087
2397 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
2398 Update bound/cmp/control for niter.
2400 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
2402 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
2403 Move check for non-empty BB here.
2404 (fur_source::register_outgoing_edges): ...from here.
2406 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
2408 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
2409 Remove call to improve_range_with_equivs.
2410 (path_range_query::improve_range_with_equivs): Remove
2411 * gimple-range-path.h: Remove improve_range_with_equivs.
2413 2021-09-22 dianhong xu <dianhong.xu@intel.com>
2415 * config/i386/avx512fp16intrin.h:
2416 (_mm512_mask_blend_ph): New intrinsic.
2417 (_mm512_permutex2var_ph): Ditto.
2418 (_mm512_permutexvar_ph): Ditto.
2419 * config/i386/avx512fp16vlintrin.h:
2420 (_mm256_mask_blend_ph): New intrinsic.
2421 (_mm256_permutex2var_ph): Ditto.
2422 (_mm256_permutexvar_ph): Ditto.
2423 (_mm_mask_blend_ph): Ditto.
2424 (_mm_permutex2var_ph): Ditto.
2425 (_mm_permutexvar_ph): Ditto.
2427 2021-09-22 dianhong xu <dianhong.xu@intel.com>
2429 * config/i386/avx512fp16intrin.h: Add new intrinsics.
2430 (_mm512_conj_pch): New intrinsic.
2431 (_mm512_mask_conj_pch): Ditto.
2432 (_mm512_maskz_conj_pch): Ditto.
2433 * config/i386/avx512fp16vlintrin.h: Add new intrinsics.
2434 (_mm256_conj_pch): New intrinsic.
2435 (_mm256_mask_conj_pch): Ditto.
2436 (_mm256_maskz_conj_pch): Ditto.
2437 (_mm_conj_pch): Ditto.
2438 (_mm_mask_conj_pch): Ditto.
2439 (_mm_maskz_conj_pch): Ditto.
2441 2021-09-22 dianhong xu <dianhong.xu@intel.com>
2443 * config/i386/avx512fp16intrin.h (_MM512_REDUCE_OP): New macro
2444 (_mm512_reduce_add_ph): New intrinsic.
2445 (_mm512_reduce_mul_ph): Ditto.
2446 (_mm512_reduce_min_ph): Ditto.
2447 (_mm512_reduce_max_ph): Ditto.
2448 * config/i386/avx512fp16vlintrin.h
2449 (_MM256_REDUCE_OP/_MM_REDUCE_OP): New macro.
2450 (_mm256_reduce_add_ph): New intrinsic.
2451 (_mm256_reduce_mul_ph): Ditto.
2452 (_mm256_reduce_min_ph): Ditto.
2453 (_mm256_reduce_max_ph): Ditto.
2454 (_mm_reduce_add_ph): Ditto.
2455 (_mm_reduce_mul_ph): Ditto.
2456 (_mm_reduce_min_ph): Ditto.
2457 (_mm_reduce_max_ph): Ditto.
2459 2021-09-22 dianhong xu <dianhong.xu@intel.com>
2461 * config/i386/avx512fp16intrin.h (__m512h_u, __m256h_u,
2462 __m128h_u): New typedef.
2463 (_mm512_load_ph): New intrinsic.
2464 (_mm256_load_ph): Ditto.
2465 (_mm_load_ph): Ditto.
2466 (_mm512_loadu_ph): Ditto.
2467 (_mm256_loadu_ph): Ditto.
2468 (_mm_loadu_ph): Ditto.
2469 (_mm512_store_ph): Ditto.
2470 (_mm256_store_ph): Ditto.
2471 (_mm_store_ph): Ditto.
2472 (_mm512_storeu_ph): Ditto.
2473 (_mm256_storeu_ph): Ditto.
2474 (_mm_storeu_ph): Ditto.
2475 (_mm512_abs_ph): Ditto.
2476 * config/i386/avx512fp16vlintrin.h
2477 (_mm_abs_ph): Ditto.
2478 (_mm256_abs_ph): Ditto.
2480 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
2482 * config/s390/tpf.md (prologue_tpf, epilogue_tpf): Add cc clobber.
2484 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
2487 * config/s390/s390.c (s390_expand_insv): Emit a normal move if it
2488 is actually a full copy of the source operand into the target.
2489 Don't emit a strict low part move if source and target mode match.
2491 2021-09-22 Jakub Jelinek <jakub@redhat.com>
2493 PR middle-end/102415
2494 * omp-expand.c (expand_omp_single): If region->exit is NULL,
2495 assert region->entry is GIMPLE_OMP_SCOPE region and return.
2497 2021-09-22 Jakub Jelinek <jakub@redhat.com>
2499 * tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
2500 * tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
2501 arguments from 2 to 3.
2502 * tree-pretty-print.c (dump_omp_clause): Print allocator() around
2503 allocate clause allocator and print align if present.
2504 * omp-low.c (scan_sharing_clauses): Force allocate_map entry even
2505 for omp_default_mem_alloc if align modifier is present. If align
2506 modifier is present, use TREE_LIST to encode both allocator and
2508 (lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
2509 Handle align modifier on allocator clause if present.
2511 2021-09-22 liuhongt <hongtao.liu@intel.com>
2513 * config/i386/i386.md (define_attr "isa"): Add
2515 (define_attr "enabled"): Correspond fma_or_avx512vl to
2516 TARGET_FMA || TARGET_AVX512VL.
2517 * config/i386/mmx.md (fmav2sf4): Extend to AVX512 fma.
2522 2021-09-22 liuhongt <hongtao.liu@intel.com>
2524 * config/i386/i386.md (cstorehf3): New define_expand.
2526 2021-09-22 liuhongt <hongtao.liu@intel.com>
2528 * config/i386/i386.md (<rounding_insn>hf2): New expander.
2529 (sse4_1_round<mode>2): Extend from MODEF to MODEFH.
2530 * config/i386/sse.md (*sse4_1_round<ssescalarmodesuffix>):
2531 Extend from VF_128 to VFH_128.
2533 2021-09-22 liuhongt <hongtao.liu@intel.com>
2535 * config/i386/i386-features.c (i386-features.c): Handle
2537 * config/i386/i386.md (sqrthf2): New expander.
2538 (*sqrthf2): New define_insn.
2539 * config/i386/sse.md
2540 (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
2543 2021-09-22 liuhongt <hongtao.liu@intel.com>
2545 * config/i386/avx512fp16intrin.h (_mm_mask_fcmadd_sch):
2547 (_mm_mask3_fcmadd_sch): Likewise.
2548 (_mm_maskz_fcmadd_sch): Likewise.
2549 (_mm_fcmadd_sch): Likewise.
2550 (_mm_mask_fmadd_sch): Likewise.
2551 (_mm_mask3_fmadd_sch): Likewise.
2552 (_mm_maskz_fmadd_sch): Likewise.
2553 (_mm_fmadd_sch): Likewise.
2554 (_mm_mask_fcmadd_round_sch): Likewise.
2555 (_mm_mask3_fcmadd_round_sch): Likewise.
2556 (_mm_maskz_fcmadd_round_sch): Likewise.
2557 (_mm_fcmadd_round_sch): Likewise.
2558 (_mm_mask_fmadd_round_sch): Likewise.
2559 (_mm_mask3_fmadd_round_sch): Likewise.
2560 (_mm_maskz_fmadd_round_sch): Likewise.
2561 (_mm_fmadd_round_sch): Likewise.
2562 (_mm_fcmul_sch): Likewise.
2563 (_mm_mask_fcmul_sch): Likewise.
2564 (_mm_maskz_fcmul_sch): Likewise.
2565 (_mm_fmul_sch): Likewise.
2566 (_mm_mask_fmul_sch): Likewise.
2567 (_mm_maskz_fmul_sch): Likewise.
2568 (_mm_fcmul_round_sch): Likewise.
2569 (_mm_mask_fcmul_round_sch): Likewise.
2570 (_mm_maskz_fcmul_round_sch): Likewise.
2571 (_mm_fmul_round_sch): Likewise.
2572 (_mm_mask_fmul_round_sch): Likewise.
2573 (_mm_maskz_fmul_round_sch): Likewise.
2574 * config/i386/i386-builtin.def: Add corresponding new builtins.
2575 * config/i386/sse.md
2576 (avx512fp16_fmaddcsh_v8hf_maskz<round_expand_name>): New expander.
2577 (avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>): Ditto.
2578 (avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>):
2580 (avx512fp16_<complexopname>sh_v8hf_mask<round_name>): Ditto.
2581 (avx512fp16_<complexopname>sh_v8hf<mask_scalarc_name><round_scalarcz_name>):
2583 * config/i386/subst.md (mask_scalarcz_name): New.
2584 (mask_scalarc_name): Ditto.
2585 (mask_scalarc_operand3): Ditto.
2586 (mask_scalarcz_operand4): Ditto.
2587 (round_scalarcz_name): Ditto.
2588 (round_scalarc_mask_operand3): Ditto.
2589 (round_scalarcz_mask_operand4): Ditto.
2590 (round_scalarc_mask_op3): Ditto.
2591 (round_scalarcz_mask_op4): Ditto.
2592 (round_scalarcz_constraint): Ditto.
2593 (round_scalarcz_nimm_predicate): Ditto.
2594 (mask_scalarcz): Ditto.
2595 (mask_scalarc): Ditto.
2596 (round_scalarcz): Ditto.
2598 2021-09-22 liuhongt <hongtao.liu@intel.com>
2600 * config/i386/avx512fp16intrin.h (_mm512_fcmadd_pch):
2602 (_mm512_mask_fcmadd_pch): Likewise.
2603 (_mm512_mask3_fcmadd_pch): Likewise.
2604 (_mm512_maskz_fcmadd_pch): Likewise.
2605 (_mm512_fmadd_pch): Likewise.
2606 (_mm512_mask_fmadd_pch): Likewise.
2607 (_mm512_mask3_fmadd_pch): Likewise.
2608 (_mm512_maskz_fmadd_pch): Likewise.
2609 (_mm512_fcmadd_round_pch): Likewise.
2610 (_mm512_mask_fcmadd_round_pch): Likewise.
2611 (_mm512_mask3_fcmadd_round_pch): Likewise.
2612 (_mm512_maskz_fcmadd_round_pch): Likewise.
2613 (_mm512_fmadd_round_pch): Likewise.
2614 (_mm512_mask_fmadd_round_pch): Likewise.
2615 (_mm512_mask3_fmadd_round_pch): Likewise.
2616 (_mm512_maskz_fmadd_round_pch): Likewise.
2617 (_mm512_fcmul_pch): Likewise.
2618 (_mm512_mask_fcmul_pch): Likewise.
2619 (_mm512_maskz_fcmul_pch): Likewise.
2620 (_mm512_fmul_pch): Likewise.
2621 (_mm512_mask_fmul_pch): Likewise.
2622 (_mm512_maskz_fmul_pch): Likewise.
2623 (_mm512_fcmul_round_pch): Likewise.
2624 (_mm512_mask_fcmul_round_pch): Likewise.
2625 (_mm512_maskz_fcmul_round_pch): Likewise.
2626 (_mm512_fmul_round_pch): Likewise.
2627 (_mm512_mask_fmul_round_pch): Likewise.
2628 (_mm512_maskz_fmul_round_pch): Likewise.
2629 * config/i386/avx512fp16vlintrin.h (_mm_fmadd_pch):
2631 (_mm_mask_fmadd_pch): Likewise.
2632 (_mm_mask3_fmadd_pch): Likewise.
2633 (_mm_maskz_fmadd_pch): Likewise.
2634 (_mm256_fmadd_pch): Likewise.
2635 (_mm256_mask_fmadd_pch): Likewise.
2636 (_mm256_mask3_fmadd_pch): Likewise.
2637 (_mm256_maskz_fmadd_pch): Likewise.
2638 (_mm_fcmadd_pch): Likewise.
2639 (_mm_mask_fcmadd_pch): Likewise.
2640 (_mm_mask3_fcmadd_pch): Likewise.
2641 (_mm_maskz_fcmadd_pch): Likewise.
2642 (_mm256_fcmadd_pch): Likewise.
2643 (_mm256_mask_fcmadd_pch): Likewise.
2644 (_mm256_mask3_fcmadd_pch): Likewise.
2645 (_mm256_maskz_fcmadd_pch): Likewise.
2646 (_mm_fmul_pch): Likewise.
2647 (_mm_mask_fmul_pch): Likewise.
2648 (_mm_maskz_fmul_pch): Likewise.
2649 (_mm256_fmul_pch): Likewise.
2650 (_mm256_mask_fmul_pch): Likewise.
2651 (_mm256_maskz_fmul_pch): Likewise.
2652 (_mm_fcmul_pch): Likewise.
2653 (_mm_mask_fcmul_pch): Likewise.
2654 (_mm_maskz_fcmul_pch): Likewise.
2655 (_mm256_fcmul_pch): Likewise.
2656 (_mm256_mask_fcmul_pch): Likewise.
2657 (_mm256_maskz_fcmul_pch): Likewise.
2658 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF_V8HF_V8HF,
2659 V8HF_FTYPE_V16HF_V16HF_V16HF, V16HF_FTYPE_V16HF_V16HF_V16HF_UQI,
2660 V32HF_FTYPE_V32HF_V32HF_V32HF_INT,
2661 V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT): Add new builtin types.
2662 * config/i386/i386-builtin.def: Add new builtins.
2663 * config/i386/i386-expand.c: Handle new builtin types.
2664 * config/i386/subst.md (SUBST_CV): New.
2665 (maskc_name): Ditto.
2666 (maskc_operand3): Ditto.
2668 (sdc_maskz_name): Ditto.
2669 (sdc_mask_op4): Ditto.
2670 (sdc_mask_op5): Ditto.
2671 (sdc_mask_mode512bit_condition): Ditto.
2673 (round_maskc_operand3): Ditto.
2674 (round_sdc_mask_operand4): Ditto.
2675 (round_maskc_op3): Ditto.
2676 (round_sdc_mask_op4): Ditto.
2677 (round_saeonly_sdc_mask_operand5): Ditto.
2678 * config/i386/sse.md (unspec): Add complex fma unspecs.
2679 (avx512fmaskcmode): New.
2680 (UNSPEC_COMPLEX_F_C_MA): Ditto.
2681 (UNSPEC_COMPLEX_F_C_MUL): Ditto.
2682 (complexopname): Ditto.
2683 (<avx512>_fmaddc_<mode>_maskz<round_expand_name>): New expander.
2684 (<avx512>_fcmaddc_<mode>_maskz<round_expand_name>): Ditto.
2685 (fma_<complexopname>_<mode><sdc_maskz_name><round_name>): New
2687 (<avx512>_<complexopname>_<mode>_mask<round_name>): Ditto.
2688 (<avx512>_<complexopname>_<mode><maskc_name><round_name>): Ditto.
2690 2021-09-22 Kewen Lin <linkw@linux.ibm.com>
2692 * config/rs6000/rs6000.opt (rs6000-density-pct-threshold,
2693 rs6000-density-size-threshold, rs6000-density-penalty,
2694 rs6000-density-load-pct-threshold,
2695 rs6000-density-load-num-threshold): New parameter.
2696 * config/rs6000/rs6000.c (rs6000_density_test): Adjust with
2697 corresponding parameters.
2699 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2701 * gimple-range-path.cc (path_range_query::defined_outside_path):
2703 (path_range_query::range_on_path_entry): New.
2704 (path_range_query::internal_range_of_expr): Resolve unknowns
2706 (path_range_query::improve_range_with_equivs): New.
2707 (path_range_query::ssa_range_in_phi): Resolve unknowns with
2709 * gimple-range-path.h (class path_range_query): Add
2710 defined_outside_path, range_on_path_entry, and
2711 improve_range_with_equivs.
2713 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2715 * gimple-range-path.cc (path_range_query::add_to_imports): New.
2716 (path_range_query::add_copies_to_imports): New.
2717 (path_range_query::precompute_ranges): Call
2718 add_copies_to_imports.
2719 * gimple-range-path.h (class path_range_query): Add prototypes
2720 for add_copies_to_imports and add_to_imports.
2722 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2724 * gimple-range-path.cc (path_range_query::range_defined_in_block):
2725 Remove useless code.
2727 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2729 * gimple-range-fold.h (class fur_source): Make oracle protected.
2730 * gimple-range-path.cc (path_range_query::path_range_query): Add
2731 resolve argument. Initialize oracle.
2732 (path_range_query::~path_range_query): Delete oracle.
2733 (path_range_query::range_of_stmt): Adapt to use relations.
2734 (path_range_query::precompute_ranges): Pre-compute relations.
2735 (class jt_fur_source): New
2736 (jt_fur_source::jt_fur_source): New.
2737 (jt_fur_source::register_relation): New.
2738 (jt_fur_source::query_relation): New.
2739 (path_range_query::precompute_relations): New.
2740 (path_range_query::precompute_phi_relations): New.
2741 * gimple-range-path.h (path_range_query): Add resolve argument.
2742 Add oracle, precompute_relations, precompute_phi_relations.
2743 * tree-ssa-threadbackward.c (back_threader::back_threader): Pass
2744 resolve argument to solver.
2746 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2748 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
2749 Rename postfold_gcond_edges to register_outgoing_edges and
2751 (fold_using_range::postfold_gcond_edges): Rename...
2752 (fur_source::register_outgoing_edges): ...to this.
2753 * gimple-range-fold.h (postfold_gcond_edges): Rename to
2754 register_outgoing_edges and move to fur_source.
2756 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2758 * gimple-range-fold.cc (fold_using_range::range_of_phi): Check
2759 dom_info_available_p.
2761 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2763 * gimple-range-cache.cc (non_null_ref::non_null_ref): Use create
2764 and quick_grow_cleared instead of safe_grow_cleared.
2766 2021-09-21 Thomas Schwinge <thomas@codesourcery.com>
2769 * omp-oacc-neuter-broadcast.cc (oacc_do_neutering): Evaluate
2772 2021-09-21 Richard Earnshaw <rearnsha@arm.com>
2774 * configure.ac: Detect when the assembler supports new-style
2775 architecture extensions.
2776 * common/config/arm/arm-common.c (arm_rewrite_mcpu): Return
2777 the full CPU string if the assembler can grok it.
2778 (arm_rewrite_march): Likewise but for the architecture.
2779 * config.in: Regenerate.
2780 * configure: Regenerate.
2782 2021-09-21 Richard Biener <rguenther@suse.de>
2784 PR tree-optimization/102421
2785 * tree-vect-loop.c (vect_dissolve_slp_only_groups): Copy and
2786 adjust alignment info.
2788 2021-09-21 Kewen Lin <linkw@linux.ibm.com>
2790 * ipa-fnsummary.c (ipa_fn_summary_write): Remove inconsistent
2791 bitfield stream out.
2793 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
2795 * gimple-range-fold.cc (fold_using_range::range_of_phi): Ignore
2796 undefined edges, apply an equivalence if appropriate.
2797 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Return
2798 UNDEFINED if EDGE_EXECUTABLE is not set.
2799 * gimple-range.cc (gimple_ranger::gimple_ranger): Set all edges
2800 as EXECUTABLE upon startup.
2801 (gimple_ranger::range_on_edge): Return UNDEFINED for edges without
2802 EDGE_EXECUTABLE set.
2803 * vr-values.c (set_and_propagate_unexecutable): New.
2804 (simplify_using_ranges::fold_cond): Call set_and_propagate.
2805 (simplify_using_ranges::simplify_switch_using_ranges): Ditto.
2806 * vr-values.h: Add prototype.
2808 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
2810 * value-relation.cc (equiv_oracle::register_initial_def): New.
2811 (equiv_oracle::register_relation): Call register_initial_def.
2812 (equiv_oracle::add_equiv_to_block): New. Split register_relation.
2813 (relation_oracle::register_stmt): Check def block of PHI arguments.
2814 * value-relation.h (equiv_oracle): Add new prototypes.
2816 2021-09-20 Matthias Kretz <m.kretz@gsi.de>
2818 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
2819 Define __RECIPROCAL_MATH__, __NO_SIGNED_ZEROS__,
2820 __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__, and
2821 __ROUNDING_MATH__ according to their corresponding flags.
2822 * doc/cpp.texi: Document __RECIPROCAL_MATH__,
2823 __NO_SIGNED_ZEROS__, __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__,
2824 and __ROUNDING_MATH__.
2826 2021-09-20 Richard Biener <rguenther@suse.de>
2828 * tree-vect-stmts.c (vectorizable_load): Use the vectype
2831 2021-09-20 Richard Biener <rguenther@suse.de>
2833 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
2834 Do not compute alignment of the vectorized access here.
2836 2021-09-20 Richard Biener <rguenther@suse.de>
2838 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
2839 Store -1 for runtime alias peeling iterations.
2841 2021-09-20 Richard Biener <rguenther@suse.de>
2843 * config.gcc: Obsolete hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11*.
2845 2021-09-20 Thomas Schwinge <thomas@codesourcery.com>
2847 * input.c (string_concat_db::record_string_concatenation)
2848 (string_concat_db::get_string_concatenation): Skip for
2849 'RESERVED_LOCATION_P'.
2851 2021-09-20 Richard Biener <rguenther@suse.de>
2853 PR tree-optimization/65206
2854 * tree-data-ref.h (struct data_reference): Add alt_indices,
2856 * tree-data-ref.c (free_data_ref): Release alt_indices.
2857 (dr_analyze_indices): Work on struct indices and get DR_REF as tree.
2858 (create_data_ref): Adjust.
2859 (initialize_data_dependence_relation): Split into head
2860 and tail. When the base objects fail to match up try
2861 again with pointer-based analysis of indices.
2862 * tree-vectorizer.c (vec_info_shared::check_datarefs): Do
2863 not compare the lazily computed alternate set of indices.
2865 2021-09-20 Iain Sandoe <iain@sandoe.co.uk>
2867 * gcc.c: Test for execute OK when we find the
2868 programs for assembler linker and dsymutil and those
2869 were specified at configure-time.
2871 2021-09-19 Martin Sebor <msebor@redhat.com>
2873 PR middle-end/102403
2874 * gimple-predicate-analysis.cc (predicate::init_from_control_deps):
2875 Correct a function pre/postcondition.
2877 2021-09-19 Martin Sebor <msebor@redhat.com>
2879 PR middle-end/102243
2880 * tree-ssa-strlen.c (get_range): Handle null cfun.
2882 2021-09-19 Iain Sandoe <iain@sandoe.co.uk>
2884 * config/darwin.h (LINK_COMMAND_SPEC_A): Use Darwin10
2885 unwinder shim as a convenience library.
2887 2021-09-19 Andrew Pinski <apinski@marvell.com>
2889 * doc/install.texi: Add note about
2890 binutils 2.35 is required for LTO usage.
2892 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
2894 * tree-ssa-threadbackward.c
2895 (back_threader_registry::register_path): Use push_edge.
2896 * tree-ssa-threadedge.c
2897 (jump_threader::thread_around_empty_blocks): Same.
2898 (jump_threader::thread_through_normal_block): Same.
2899 (jump_threader::thread_across_edge): Same. Also, use auto_bitmap.
2901 * tree-ssa-threadupdate.c
2902 (jt_path_registry::allocate_thread_edge): Remove.
2903 (jt_path_registry::push_edge): New.
2904 (dump_jump_thread_path): Make static.
2905 * tree-ssa-threadupdate.h (allocate_thread_edge): Remove.
2908 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
2910 * gimple-range-path.cc (path_range_query::path_range_query): Add
2912 (path_range_query::dump): Remove extern declaration of dump_ranger.
2913 * gimple-range-trace.cc (dump_ranger): Add DEBUG_FUNCTION marker.
2914 * gimple-range-trace.h (dump_ranger): Add prototype.
2916 2021-09-19 John Ericson <git@JohnEricson.me>
2918 * gcc.c (find_a_program): New function, factored out of...
2919 (find_a_file): Here.
2920 (execute): Use find_a_program when looking for programs rather
2923 2021-09-19 Matwey V. Kornilov <matwey.kornilov@gmail.com>
2925 * config/avr/avr-mcus.def: Add atmega324pb.
2926 * doc/avr-mmcu.texi: Corresponding changes.
2928 2021-09-19 Roger Sayle <roger@nextmovesoftware.com>
2931 * match.pd (cmp @0 REAL_CST@1): When @0 is also REAL_CST, apply
2932 the same transformations as to @1. For comparisons against NaN,
2933 don't check HONOR_SNANS but confirm that neither operand is a
2936 2021-09-19 Benjamin Peterson <benjamin@locrian.net>
2938 * attribs.c (make_unique_name): Delete.
2939 * attribs.h (make_unique_name): Delete.
2941 2021-09-19 Andrew Pinski <apinski@marvell.com>
2943 * lra-constraints.c (check_and_process_move): Assert
2944 that dclass and sclass are greater than or equal to NO_REGS.
2946 2021-09-18 Jakub Jelinek <jakub@redhat.com>
2948 * tree.h (OMP_CLAUSE_ORDER_UNCONSTRAINED): Define.
2949 * tree-pretty-print.c (dump_omp_clause): Print unconstrained:
2950 for OMP_CLAUSE_ORDER_UNCONSTRAINED.
2952 2021-09-18 liuhongt <hongtao.liu@intel.com>
2954 * config/i386/i386-features.c (remove_partial_avx_dependency):
2955 Restrict TARGET_USE_VECTOR_FP_CONVERTS and
2956 TARGET_USE_VECTOR_CONVERTS to conversion instructions only.
2958 2021-09-18 Jakub Jelinek <jakub@redhat.com>
2960 * gimplify.c (omp_default_clause): For C/C++ default({,first}private),
2961 if file/namespace scope variable doesn't have predetermined sharing,
2962 treat it as if there was default(none).
2964 2021-09-18 liuhongt <hongtao.liu@intel.com>
2966 * config/i386/avx512fp16intrin.h (_mm_fmadd_sh):
2968 (_mm_mask_fmadd_sh): Likewise.
2969 (_mm_mask3_fmadd_sh): Likewise.
2970 (_mm_maskz_fmadd_sh): Likewise.
2971 (_mm_fmadd_round_sh): Likewise.
2972 (_mm_mask_fmadd_round_sh): Likewise.
2973 (_mm_mask3_fmadd_round_sh): Likewise.
2974 (_mm_maskz_fmadd_round_sh): Likewise.
2975 (_mm_fnmadd_sh): Likewise.
2976 (_mm_mask_fnmadd_sh): Likewise.
2977 (_mm_mask3_fnmadd_sh): Likewise.
2978 (_mm_maskz_fnmadd_sh): Likewise.
2979 (_mm_fnmadd_round_sh): Likewise.
2980 (_mm_mask_fnmadd_round_sh): Likewise.
2981 (_mm_mask3_fnmadd_round_sh): Likewise.
2982 (_mm_maskz_fnmadd_round_sh): Likewise.
2983 (_mm_fmsub_sh): Likewise.
2984 (_mm_mask_fmsub_sh): Likewise.
2985 (_mm_mask3_fmsub_sh): Likewise.
2986 (_mm_maskz_fmsub_sh): Likewise.
2987 (_mm_fmsub_round_sh): Likewise.
2988 (_mm_mask_fmsub_round_sh): Likewise.
2989 (_mm_mask3_fmsub_round_sh): Likewise.
2990 (_mm_maskz_fmsub_round_sh): Likewise.
2991 (_mm_fnmsub_sh): Likewise.
2992 (_mm_mask_fnmsub_sh): Likewise.
2993 (_mm_mask3_fnmsub_sh): Likewise.
2994 (_mm_maskz_fnmsub_sh): Likewise.
2995 (_mm_fnmsub_round_sh): Likewise.
2996 (_mm_mask_fnmsub_round_sh): Likewise.
2997 (_mm_mask3_fnmsub_round_sh): Likewise.
2998 (_mm_maskz_fnmsub_round_sh): Likewise.
2999 * config/i386/i386-builtin-types.def
3000 (V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT): New builtin type.
3001 * config/i386/i386-builtin.def: Add new builtins.
3002 * config/i386/i386-expand.c: Handle new builtin type.
3003 * config/i386/sse.md (fmai_vmfmadd_<mode><round_name>):
3004 Ajdust to support FP16.
3005 (fmai_vmfmsub_<mode><round_name>): Ditto.
3006 (fmai_vmfnmadd_<mode><round_name>): Ditto.
3007 (fmai_vmfnmsub_<mode><round_name>): Ditto.
3008 (*fmai_fmadd_<mode>): Ditto.
3009 (*fmai_fmsub_<mode>): Ditto.
3010 (*fmai_fnmadd_<mode><round_name>): Ditto.
3011 (*fmai_fnmsub_<mode><round_name>): Ditto.
3012 (avx512f_vmfmadd_<mode>_mask<round_name>): Ditto.
3013 (avx512f_vmfmadd_<mode>_mask3<round_name>): Ditto.
3014 (avx512f_vmfmadd_<mode>_maskz<round_expand_name>): Ditto.
3015 (avx512f_vmfmadd_<mode>_maskz_1<round_name>): Ditto.
3016 (*avx512f_vmfmsub_<mode>_mask<round_name>): Ditto.
3017 (avx512f_vmfmsub_<mode>_mask3<round_name>): Ditto.
3018 (*avx512f_vmfmsub_<mode>_maskz_1<round_name>): Ditto.
3019 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
3020 (*avx512f_vmfnmsub_<mode>_mask3<round_name>): Ditto.
3021 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
3022 (*avx512f_vmfnmadd_<mode>_mask<round_name>): Renamed to ...
3023 (avx512f_vmfnmadd_<mode>_mask<round_name>) ... this, and
3024 adjust to support FP16.
3025 (avx512f_vmfnmadd_<mode>_mask3<round_name>): Ditto.
3026 (avx512f_vmfnmadd_<mode>_maskz_1<round_name>): Ditto.
3027 (avx512f_vmfnmadd_<mode>_maskz<round_expand_name>): New
3030 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
3032 * config/i386/sse.md (avx512fmaskmodelower): Extend to support
3034 (maskload<mode><avx512fmaskmodelower>): Ditto.
3035 (maskstore<mode><avx512fmaskmodelower>): Ditto.
3037 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
3039 * config/i386/i386-expand.c (ix86_expand_fp_absneg_operator):
3041 (ix86_expand_copysign): Ditto.
3042 (ix86_expand_xorsign): Ditto.
3043 * config/i386/i386.c (ix86_build_const_vector): Handle HF vector
3045 (ix86_build_signbit_mask): Ditto.
3046 (ix86_can_change_mode_class): Ditto.
3047 * config/i386/i386.md
3048 (SSEMODEF): Add HFmode.
3049 (ssevecmodef): Ditto.
3050 (<code>hf2): New define_expand.
3051 (*<code>hf2_1): New define_insn_and_split.
3052 (copysign<mode>): Extend to support HFmode under AVX512FP16.
3053 (xorsign<mode>): Ditto.
3054 * config/i386/sse.md (VFB): New mode iterator.
3055 (VFB_128_256): Ditto.
3057 (sseintvecmode2): Support HF vector mode.
3058 (<code><mode>2): Use new mode iterator.
3059 (*<code><mode>2): Ditto.
3060 (copysign<mode>3): Ditto.
3061 (xorsign<mode>3): Ditto.
3062 (<code><mode>3<mask_name>): Ditto.
3063 (<code><mode>3<mask_name>): Ditto.
3064 (<sse>_andnot<mode>3<mask_name>): Adjust for HF vector mode.
3065 (<sse>_andnot<mode>3<mask_name>): Ditto.
3066 (*<code><mode>3<mask_name>): Ditto.
3067 (*<code><mode>3<mask_name>): Ditto.
3069 2021-09-18 liuhongt <hongtao.liu@intel.com>
3071 * config/i386/avx512fp16intrin.h (_mm512_mask_fmadd_ph):
3073 (_mm512_mask3_fmadd_ph): Likewise.
3074 (_mm512_maskz_fmadd_ph): Likewise.
3075 (_mm512_fmadd_round_ph): Likewise.
3076 (_mm512_mask_fmadd_round_ph): Likewise.
3077 (_mm512_mask3_fmadd_round_ph): Likewise.
3078 (_mm512_maskz_fmadd_round_ph): Likewise.
3079 (_mm512_fnmadd_ph): Likewise.
3080 (_mm512_mask_fnmadd_ph): Likewise.
3081 (_mm512_mask3_fnmadd_ph): Likewise.
3082 (_mm512_maskz_fnmadd_ph): Likewise.
3083 (_mm512_fnmadd_round_ph): Likewise.
3084 (_mm512_mask_fnmadd_round_ph): Likewise.
3085 (_mm512_mask3_fnmadd_round_ph): Likewise.
3086 (_mm512_maskz_fnmadd_round_ph): Likewise.
3087 (_mm512_fmsub_ph): Likewise.
3088 (_mm512_mask_fmsub_ph): Likewise.
3089 (_mm512_mask3_fmsub_ph): Likewise.
3090 (_mm512_maskz_fmsub_ph): Likewise.
3091 (_mm512_fmsub_round_ph): Likewise.
3092 (_mm512_mask_fmsub_round_ph): Likewise.
3093 (_mm512_mask3_fmsub_round_ph): Likewise.
3094 (_mm512_maskz_fmsub_round_ph): Likewise.
3095 (_mm512_fnmsub_ph): Likewise.
3096 (_mm512_mask_fnmsub_ph): Likewise.
3097 (_mm512_mask3_fnmsub_ph): Likewise.
3098 (_mm512_maskz_fnmsub_ph): Likewise.
3099 (_mm512_fnmsub_round_ph): Likewise.
3100 (_mm512_mask_fnmsub_round_ph): Likewise.
3101 (_mm512_mask3_fnmsub_round_ph): Likewise.
3102 (_mm512_maskz_fnmsub_round_ph): Likewise.
3103 * config/i386/avx512fp16vlintrin.h (_mm256_fmadd_ph):
3105 (_mm256_mask_fmadd_ph): Likewise.
3106 (_mm256_mask3_fmadd_ph): Likewise.
3107 (_mm256_maskz_fmadd_ph): Likewise.
3108 (_mm_fmadd_ph): Likewise.
3109 (_mm_mask_fmadd_ph): Likewise.
3110 (_mm_mask3_fmadd_ph): Likewise.
3111 (_mm_maskz_fmadd_ph): Likewise.
3112 (_mm256_fnmadd_ph): Likewise.
3113 (_mm256_mask_fnmadd_ph): Likewise.
3114 (_mm256_mask3_fnmadd_ph): Likewise.
3115 (_mm256_maskz_fnmadd_ph): Likewise.
3116 (_mm_fnmadd_ph): Likewise.
3117 (_mm_mask_fnmadd_ph): Likewise.
3118 (_mm_mask3_fnmadd_ph): Likewise.
3119 (_mm_maskz_fnmadd_ph): Likewise.
3120 (_mm256_fmsub_ph): Likewise.
3121 (_mm256_mask_fmsub_ph): Likewise.
3122 (_mm256_mask3_fmsub_ph): Likewise.
3123 (_mm256_maskz_fmsub_ph): Likewise.
3124 (_mm_fmsub_ph): Likewise.
3125 (_mm_mask_fmsub_ph): Likewise.
3126 (_mm_mask3_fmsub_ph): Likewise.
3127 (_mm_maskz_fmsub_ph): Likewise.
3128 (_mm256_fnmsub_ph): Likewise.
3129 (_mm256_mask_fnmsub_ph): Likewise.
3130 (_mm256_mask3_fnmsub_ph): Likewise.
3131 (_mm256_maskz_fnmsub_ph): Likewise.
3132 (_mm_fnmsub_ph): Likewise.
3133 (_mm_mask_fnmsub_ph): Likewise.
3134 (_mm_mask3_fnmsub_ph): Likewise.
3135 (_mm_maskz_fnmsub_ph): Likewise.
3136 * config/i386/i386-builtin.def: Add corresponding new builtins.
3137 * config/i386/sse.md
3138 (<avx512>_fmadd_<mode>_maskz<round_expand_name>): Adjust to
3139 support HF vector modes.
3140 (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
3142 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
3143 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
3144 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
3145 (<avx512>_fmadd_<mode>_mask<round_name>): Ditto.
3146 (<avx512>_fmadd_<mode>_mask3<round_name>): Ditto.
3147 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Ditto.
3148 (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>):
3150 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
3151 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
3152 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
3153 (<avx512>_fmsub_<mode>_mask<round_name>): Ditto.
3154 (<avx512>_fmsub_<mode>_mask3<round_name>): Ditto.
3155 (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>):
3157 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
3158 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
3159 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
3160 (<avx512>_fnmadd_<mode>_mask<round_name>): Ditto.
3161 (<avx512>_fnmadd_<mode>_mask3<round_name>): Ditto.
3162 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Ditto.
3163 (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>):
3165 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
3166 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
3167 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
3168 (<avx512>_fnmsub_<mode>_mask<round_name>): Ditto.
3169 (<avx512>_fnmsub_<mode>_mask3<round_name>): Ditto.
3171 2021-09-18 liuhongt <hongtao.liu@intel.com>
3173 * config/i386/avx512fp16intrin.h (_mm512_fmaddsub_ph):
3175 (_mm512_mask_fmaddsub_ph): Likewise.
3176 (_mm512_mask3_fmaddsub_ph): Likewise.
3177 (_mm512_maskz_fmaddsub_ph): Likewise.
3178 (_mm512_fmaddsub_round_ph): Likewise.
3179 (_mm512_mask_fmaddsub_round_ph): Likewise.
3180 (_mm512_mask3_fmaddsub_round_ph): Likewise.
3181 (_mm512_maskz_fmaddsub_round_ph): Likewise.
3182 (_mm512_mask_fmsubadd_ph): Likewise.
3183 (_mm512_mask3_fmsubadd_ph): Likewise.
3184 (_mm512_maskz_fmsubadd_ph): Likewise.
3185 (_mm512_fmsubadd_round_ph): Likewise.
3186 (_mm512_mask_fmsubadd_round_ph): Likewise.
3187 (_mm512_mask3_fmsubadd_round_ph): Likewise.
3188 (_mm512_maskz_fmsubadd_round_ph): Likewise.
3189 * config/i386/avx512fp16vlintrin.h (_mm256_fmaddsub_ph):
3191 (_mm256_mask_fmaddsub_ph): Likewise.
3192 (_mm256_mask3_fmaddsub_ph): Likewise.
3193 (_mm256_maskz_fmaddsub_ph): Likewise.
3194 (_mm_fmaddsub_ph): Likewise.
3195 (_mm_mask_fmaddsub_ph): Likewise.
3196 (_mm_mask3_fmaddsub_ph): Likewise.
3197 (_mm_maskz_fmaddsub_ph): Likewise.
3198 (_mm256_fmsubadd_ph): Likewise.
3199 (_mm256_mask_fmsubadd_ph): Likewise.
3200 (_mm256_mask3_fmsubadd_ph): Likewise.
3201 (_mm256_maskz_fmsubadd_ph): Likewise.
3202 (_mm_fmsubadd_ph): Likewise.
3203 (_mm_mask_fmsubadd_ph): Likewise.
3204 (_mm_mask3_fmsubadd_ph): Likewise.
3205 (_mm_maskz_fmsubadd_ph): Likewise.
3206 * config/i386/i386-builtin.def: Add corresponding new builtins.
3207 * config/i386/sse.md (VFH_SF_AVX512VL): New mode iterator.
3208 * (<avx512>_fmsubadd_<mode>_maskz<round_expand_name>): New expander.
3209 * (<avx512>_fmaddsub_<mode>_maskz<round_expand_name>): Use
3211 * (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
3213 * (<avx512>_fmaddsub_<mode>_mask<round_name>): Ditto.
3214 * (<avx512>_fmaddsub_<mode>_mask3<round_name>): Ditto.
3215 * (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
3217 * (<avx512>_fmsubadd_<mode>_mask<round_name>): Ditto.
3218 * (<avx512>_fmsubadd_<mode>_mask3<round_name>): Ditto.
3220 2021-09-18 liuhongt <hongtao.liu@intel.com>
3223 * config/i386/i386.c (ix86_print_operand): Handle
3224 V8HF/V16HF/V32HFmode.
3225 * config/i386/i386.h (VALID_BCST_MODE_P): Add HFmode.
3226 * config/i386/sse.md (avx512bcst): Remove.
3228 2021-09-17 Martin Sebor <msebor@redhat.com>
3230 * Makefile.in (OBJS): Add gimple-predicate-analysis.o.
3231 * tree-ssa-uninit.c (max_phi_args): Move to gimple-predicate-analysis.
3232 (MASK_SET_BIT, MASK_TEST_BIT, MASK_EMPTY): Same.
3233 (check_defs): Add comment.
3234 (can_skip_redundant_opnd): Update comment.
3235 (compute_uninit_opnds_pos): Adjust to namespace change.
3236 (find_pdom): Move to gimple-predicate-analysis.cc.
3238 (struct uninit_undef_val_t): New.
3239 (is_non_loop_exit_postdominating): Move to gimple-predicate-analysis.cc.
3240 (find_control_equiv_block): Same.
3241 (MAX_NUM_CHAINS, MAX_CHAIN_LEN, MAX_POSTDOM_CHECK): Same.
3242 (MAX_SWITCH_CASES): Same.
3243 (compute_control_dep_chain): Same.
3244 (find_uninit_use): Use predicate analyzer.
3245 (struct pred_info): Move to gimple-predicate-analysis.
3246 (convert_control_dep_chain_into_preds): Same.
3247 (find_predicates): Same.
3248 (collect_phi_def_edges): Same.
3249 (warn_uninitialized_phi): Use predicate analyzer.
3250 (find_def_preds): Move to gimple-predicate-analysis.
3251 (dump_pred_info): Same.
3252 (dump_pred_chain): Same.
3253 (dump_predicates): Same.
3254 (destroy_predicate_vecs): Remove.
3255 (execute_late_warn_uninitialized): New.
3256 (get_cmp_code): Move to gimple-predicate-analysis.
3257 (is_value_included_in): Same.
3258 (value_sat_pred_p): Same.
3259 (find_matching_predicate_in_rest_chains): Same.
3260 (is_use_properly_guarded): Same.
3261 (prune_uninit_phi_opnds): Same.
3262 (find_var_cmp_const): Same.
3263 (use_pred_not_overlap_with_undef_path_pred): Same.
3264 (pred_equal_p): Same.
3265 (is_neq_relop_p): Same.
3266 (is_neq_zero_form_p): Same.
3267 (pred_expr_equal_p): Same.
3268 (is_pred_expr_subset_of): Same.
3269 (is_pred_chain_subset_of): Same.
3270 (is_included_in): Same.
3271 (is_superset_of): Same.
3273 (simplify_pred): Same.
3274 (simplify_preds_2): Same.
3275 (simplify_preds_3): Same.
3276 (simplify_preds_4): Same.
3277 (simplify_preds): Same.
3279 (push_to_worklist): Same.
3280 (get_pred_info_from_cmp): Same.
3281 (is_degenerated_phi): Same.
3282 (normalize_one_pred_1): Same.
3283 (normalize_one_pred): Same.
3284 (normalize_one_pred_chain): Same.
3285 (normalize_preds): Same.
3286 (can_one_predicate_be_invalidated_p): Same.
3287 (can_chain_union_be_invalidated_p): Same.
3288 (uninit_uses_cannot_happen): Same.
3289 (pass_late_warn_uninitialized::execute): Define.
3290 * gimple-predicate-analysis.cc: New file.
3291 * gimple-predicate-analysis.h: New file.
3293 2021-09-17 Julian Brown <julian@codesourcery.com>
3295 * config/gcn/gcn.c (gimple.h): Include.
3296 (gcn_fork_join): Emit barrier for worker-level joins.
3297 * omp-oacc-neuter-broadcast.cc (find_local_vars_to_propagate): Add
3298 writes_gang_private bitmap parameter. Set bit for blocks
3299 containing gang-private variable writes.
3300 (worker_single_simple): Don't emit barrier after predicated block.
3301 (worker_single_copy): Don't emit barrier if we're not broadcasting
3302 anything and the block contains no gang-private writes.
3303 (neuter_worker_single): Don't predicate blocks that only contain
3304 NOPs or internal marker functions. Pass has_gang_private_write
3305 argument to worker_single_copy.
3306 (oacc_do_neutering): Add writes_gang_private bitmap handling.
3308 2021-09-17 Julian Brown <julian@codesourcery.com>
3310 * config/gcn/gcn-protos.h
3311 (gcn_goacc_create_worker_broadcast_record): Update prototype.
3312 * config/gcn/gcn-tree.c (gcn_goacc_get_worker_red_decl): Use
3313 preallocated block of LDS memory. Do not cache/share decls for
3314 reduction temporaries between invocations.
3315 (gcn_goacc_reduction_teardown): Unshare VAR on second use.
3316 (gcn_goacc_create_worker_broadcast_record): Add OFFSET parameter
3317 and return temporary LDS space at that offset. Return pointer in
3319 * config/gcn/gcn.c (acc_lds_size, gang_private_hwm, lds_allocs):
3321 (ACC_LDS_SIZE): Define as acc_lds_size.
3322 (gcn_init_machine_status): Don't initialise lds_allocated,
3323 lds_allocs, reduc_decls fields of machine function struct.
3324 (gcn_option_override): Handle default size for gang-private
3325 variables and -mgang-private-size option.
3326 (gcn_expand_prologue): Use LDS_SIZE instead of LDS_SIZE-1 when
3327 initialising M0_REG.
3328 (gcn_shared_mem_layout): New function.
3329 (gcn_print_lds_decl): Update comment. Use global lds_allocs map and
3330 gang_private_hwm variable.
3331 (TARGET_GOACC_SHARED_MEM_LAYOUT): Define target hook.
3332 * config/gcn/gcn.h (machine_function): Remove lds_allocated,
3333 lds_allocs, reduc_decls. Add reduction_base, reduction_limit.
3334 * config/gcn/gcn.opt (gang_private_size_opt): New global.
3335 (mgang-private-size=): New option.
3336 * doc/tm.texi.in (TARGET_GOACC_SHARED_MEM_LAYOUT): Place
3338 * doc/tm.texi: Regenerate.
3339 * omp-oacc-neuter-broadcast.cc (targhooks.h, diagnostic-core.h):
3341 (build_sender_ref): Handle sender_decl being pointer.
3342 (worker_single_copy): Add PLACEMENT and ISOLATE_BROADCASTS
3343 parameters. Pass placement argument to
3344 create_worker_broadcast_record hook invocations. Handle
3345 sender_decl being pointer and isolate_broadcasts inserting extra
3347 (blk_offset_map_t): Add typedef.
3348 (neuter_worker_single): Add BLK_OFFSET_MAP parameter. Pass
3349 preallocated range to worker_single_copy call.
3350 (dfs_broadcast_reachable_1): New function.
3351 (idx_decl_pair_t, used_range_vec_t): New typedefs.
3352 (sort_size_descending): New function.
3353 (addr_range): New class.
3354 (splay_tree_compare_addr_range, splay_tree_free_key)
3355 (first_fit_range, merge_ranges_1, merge_ranges): New functions.
3356 (execute_omp_oacc_neuter_broadcast): Rename to...
3357 (oacc_do_neutering): ... this. Add BOUNDS_LO, BOUNDS_HI
3358 parameters. Arrange layout of shared memory for broadcast
3360 (execute_omp_oacc_neuter_broadcast): New function.
3361 (pass_omp_oacc_neuter_broadcast::gate): Remove num_workers==1
3362 handling from here. Enable pass for all OpenACC routines in order
3363 to call shared memory-layout hook.
3364 * target.def (create_worker_broadcast_record): Add OFFSET
3366 (shared_mem_layout): New hook.
3368 2021-09-17 Julian Brown <julian@codesourcery.com>
3369 Thomas Schwinge <thomas@codesourcery.com>
3371 * omp-oacc-neuter-broadcast.cc
3372 (pass_omp_oacc_neuter_broadcast::gate): Disable if num_workers is
3374 (execute_omp_oacc_neuter_broadcast): Adjust.
3376 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
3378 * value-relation.cc (class equiv_chain): Move to header file.
3379 (path_oracle::path_oracle): New.
3380 (path_oracle::~path_oracle): New.
3381 (path_oracle::register_relation): New.
3382 (path_oracle::query_relation): New.
3383 (path_oracle::reset_path): New.
3384 (path_oracle::dump): New.
3385 * value-relation.h (class equiv_chain): Move to here.
3386 (class path_oracle): New.
3388 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
3390 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a DOM
3392 * gimple-range-fold.cc (fur_depend::register_relation): Use
3393 register_stmt/edge routines.
3394 * value-relation.cc (equiv_chain::find): Relocate from equiv_oracle.
3395 (equiv_oracle::equiv_oracle): Create self equivalence cache.
3396 (equiv_oracle::~equiv_oracle): Release same.
3397 (equiv_oracle::equiv_set): Return entry from self equiv cache if there
3398 are no equivalences.
3399 (equiv_oracle::find_equiv_block): Move list find to equiv_chain.
3400 (equiv_oracle::register_relation): Rename from register_equiv.
3401 (relation_chain_head::find_relation): Relocate from dom_oracle.
3402 (relation_oracle::register_stmt): New.
3403 (relation_oracle::register_edge): New.
3404 (dom_oracle::*): Rename from relation_oracle.
3405 (dom_oracle::register_relation): Adjust to call equiv_oracle.
3406 (dom_oracle::set_one_relation): Split from register_relation.
3407 (dom_oracle::register_transitives): Consolidate 2 methods.
3408 (dom_oracle::find_relation_block): Move core to relation_chain.
3409 (dom_oracle::query_relation): Rename from find_relation_dom and adjust.
3410 * value-relation.h (class relation_oracle): New pure virtual base.
3411 (class equiv_oracle): Inherit from relation_oracle and adjust.
3412 (class dom_oracle): Rename from old relation_oracle and adjust.
3414 2021-09-17 Martin Sebor <msebor@redhat.com>
3416 PR middle-end/102200
3417 * pointer-query.cc (access_ref::inform_access): Handle MIN/MAX_EXPR.
3418 (handle_min_max_size): Change argument. Store original SSA_NAME for
3419 operands to potentially distinct (sub)objects.
3420 (compute_objsize_r): Adjust call to the above.
3422 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
3424 * config/rs6000/rs6000.c (rs6000-builtins.h): New include.
3425 (rs6000_new_builtin_vectorized_function): New function.
3426 (rs6000_new_builtin_md_vectorized_function): Likewise.
3427 (rs6000_builtin_vectorized_function): Call
3428 rs6000_new_builtin_vectorized_function.
3429 (rs6000_builtin_md_vectorized_function): Call
3430 rs6000_new_builtin_md_vectorized_function.
3432 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
3434 * config/rs6000/rs6000-builtin-new.def (ASSEMBLE_ACC): Add mmaint flag.
3435 (ASSEMBLE_PAIR): Likewise.
3436 (BUILD_ACC): Likewise.
3437 (DISASSEMBLE_ACC): Likewise.
3438 (DISASSEMBLE_PAIR): Likewise.
3439 (PMXVBF16GER2): Likewise.
3440 (PMXVBF16GER2NN): Likewise.
3441 (PMXVBF16GER2NP): Likewise.
3442 (PMXVBF16GER2PN): Likewise.
3443 (PMXVBF16GER2PP): Likewise.
3444 (PMXVF16GER2): Likewise.
3445 (PMXVF16GER2NN): Likewise.
3446 (PMXVF16GER2NP): Likewise.
3447 (PMXVF16GER2PN): Likewise.
3448 (PMXVF16GER2PP): Likewise.
3449 (PMXVF32GER): Likewise.
3450 (PMXVF32GERNN): Likewise.
3451 (PMXVF32GERNP): Likewise.
3452 (PMXVF32GERPN): Likewise.
3453 (PMXVF32GERPP): Likewise.
3454 (PMXVF64GER): Likewise.
3455 (PMXVF64GERNN): Likewise.
3456 (PMXVF64GERNP): Likewise.
3457 (PMXVF64GERPN): Likewise.
3458 (PMXVF64GERPP): Likewise.
3459 (PMXVI16GER2): Likewise.
3460 (PMXVI16GER2PP): Likewise.
3461 (PMXVI16GER2S): Likewise.
3462 (PMXVI16GER2SPP): Likewise.
3463 (PMXVI4GER8): Likewise.
3464 (PMXVI4GER8PP): Likewise.
3465 (PMXVI8GER4): Likewise.
3466 (PMXVI8GER4PP): Likewise.
3467 (PMXVI8GER4SPP): Likewise.
3468 (XVBF16GER2): Likewise.
3469 (XVBF16GER2NN): Likewise.
3470 (XVBF16GER2NP): Likewise.
3471 (XVBF16GER2PN): Likewise.
3472 (XVBF16GER2PP): Likewise.
3473 (XVF16GER2): Likewise.
3474 (XVF16GER2NN): Likewise.
3475 (XVF16GER2NP): Likewise.
3476 (XVF16GER2PN): Likewise.
3477 (XVF16GER2PP): Likewise.
3478 (XVF32GER): Likewise.
3479 (XVF32GERNN): Likewise.
3480 (XVF32GERNP): Likewise.
3481 (XVF32GERPN): Likewise.
3482 (XVF32GERPP): Likewise.
3483 (XVF64GER): Likewise.
3484 (XVF64GERNN): Likewise.
3485 (XVF64GERNP): Likewise.
3486 (XVF64GERPN): Likewise.
3487 (XVF64GERPP): Likewise.
3488 (XVI16GER2): Likewise.
3489 (XVI16GER2PP): Likewise.
3490 (XVI16GER2S): Likewise.
3491 (XVI16GER2SPP): Likewise.
3492 (XVI4GER8): Likewise.
3493 (XVI4GER8PP): Likewise.
3494 (XVI8GER4): Likewise.
3495 (XVI8GER4PP): Likewise.
3496 (XVI8GER4SPP): Likewise.
3497 (XXMFACC): Likewise.
3498 (XXMTACC): Likewise.
3499 (XXSETACCZ): Likewise.
3500 (ASSEMBLE_PAIR_V): Likewise.
3501 (BUILD_PAIR): Likewise.
3502 (DISASSEMBLE_PAIR_V): Likewise.
3505 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_mma_builtin):
3506 Handle RS6000_BIF_LXVP and RS6000_BIF_STXVP.
3507 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add ismmaint.
3508 (parse_bif_attrs): Handle ismmaint.
3509 (write_decls): Add bif_mmaint_bit and bif_is_mmaint.
3510 (write_bif_static_init): Handle ismmaint.
3512 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
3514 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin): New
3516 (rs6000_gimple_fold_builtin): Call rs6000_gimple_fold_new_builtin.
3517 (rs6000_new_builtin_valid_without_lhs): New function.
3518 (rs6000_gimple_fold_new_mma_builtin): Likewise.
3519 (rs6000_gimple_fold_new_builtin): Likewise.
3521 2021-09-17 Thomas Schwinge <thomas@codesourcery.com>
3523 * hash-table.h (hash_table<Descriptor, Lazy, Allocator>::expand):
3524 Destruct stale Value objects.
3525 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
3528 2021-09-17 Roger Sayle <roger@nextmovesoftware.com>
3531 * match.pd (shift optimizations): Disable recent sign-changing
3532 optimization for shifts by zero, these will be folded later.
3534 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
3536 * config/rs6000/rs6000-builtin-new.def (__builtin_mffsl): Move from
3537 [power9] to [always].
3539 2021-09-17 Richard Biener <rguenther@suse.de>
3541 * tree-vect-stmts.c (vectorizable_load): Do not frob
3544 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
3546 * config/i386/i386-features.c (remove_partial_avx_dependency):
3547 Also check TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY and
3548 and TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY before generating
3550 * config/i386/i386.h (TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY):
3552 (TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
3553 * config/i386/i386.md (SSE FP to FP splitters): Replace
3554 TARGET_SSE_PARTIAL_REG_DEPENDENCY with
3555 TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY.
3556 (SSE INT to FP splitter): Replace TARGET_SSE_PARTIAL_REG_DEPENDENCY
3557 with TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY.
3558 * config/i386/x86-tune.def
3559 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): New.
3560 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
3562 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
3565 * config/i386/i386-features.c (remove_partial_avx_dependency):
3566 Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
3567 before generating vxorps.
3569 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
3571 * config/i386/i386-options.c (processor_cost_table): Use
3572 tremont_cost for Tremont.
3573 * config/i386/x86-tune-costs.h (tremont_memcpy): New.
3574 (tremont_memset): Likewise.
3575 (tremont_cost): Likewise.
3576 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
3579 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
3581 * common/config/i386/i386-common.c: Use Haswell scheduling model
3583 * config/i386/i386.c (ix86_sched_init_global): Prepare for Tremont
3585 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Tremont
3587 (ix86_adjust_cost): Handle Tremont.
3588 * config/i386/x86-tune.def (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY):
3590 (X86_TUNE_USE_LEAVE): Likewise.
3591 (X86_TUNE_PUSH_MEMORY): Likewise.
3592 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
3593 (X86_TUNE_USE_CLTD): Likewise.
3594 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
3595 (X86_TUNE_AVOID_MFENCE): Likewise.
3596 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
3597 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
3598 (X86_TUNE_ACCUMULATE_OUTGOING_ARGS): Disable for Tremont.
3599 (X86_TUNE_FOUR_JUMP_LIMIT): Likewise.
3600 (X86_TUNE_OPT_AGU): Likewise.
3601 (X86_TUNE_AVOID_LEA_FOR_ADDR): Likewise.
3602 (X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE): Likewise.
3603 (X86_TUNE_EXPAND_ABS): Likewise.
3604 (X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS): Likewise.
3605 (X86_TUNE_SLOW_PSHUFB): Likewise.
3607 2021-09-17 Eric Botcazou <ebotcazou@adacore.com>
3609 PR rtl-optimization/102306
3610 * combine.c (try_combine): Abort the combination if we are about to
3611 duplicate volatile references.
3613 2021-09-17 liuhongt <hongtao.liu@intel.com>
3615 * config/i386/avx512fp16intrin.h (_mm_undefined_ph):
3617 (_mm256_undefined_ph): Likewise.
3618 (_mm512_undefined_ph): Likewise.
3619 (_mm_cvtsh_h): Likewise.
3620 (_mm256_cvtsh_h): Likewise.
3621 (_mm512_cvtsh_h): Likewise.
3622 (_mm512_castph_ps): Likewise.
3623 (_mm512_castph_pd): Likewise.
3624 (_mm512_castph_si512): Likewise.
3625 (_mm512_castph512_ph128): Likewise.
3626 (_mm512_castph512_ph256): Likewise.
3627 (_mm512_castph128_ph512): Likewise.
3628 (_mm512_castph256_ph512): Likewise.
3629 (_mm512_zextph128_ph512): Likewise.
3630 (_mm512_zextph256_ph512): Likewise.
3631 (_mm512_castps_ph): Likewise.
3632 (_mm512_castpd_ph): Likewise.
3633 (_mm512_castsi512_ph): Likewise.
3634 * config/i386/avx512fp16vlintrin.h (_mm_castph_ps):
3636 (_mm256_castph_ps): Likewise.
3637 (_mm_castph_pd): Likewise.
3638 (_mm256_castph_pd): Likewise.
3639 (_mm_castph_si128): Likewise.
3640 (_mm256_castph_si256): Likewise.
3641 (_mm_castps_ph): Likewise.
3642 (_mm256_castps_ph): Likewise.
3643 (_mm_castpd_ph): Likewise.
3644 (_mm256_castpd_ph): Likewise.
3645 (_mm_castsi128_ph): Likewise.
3646 (_mm256_castsi256_ph): Likewise.
3647 (_mm256_castph256_ph128): Likewise.
3648 (_mm256_castph128_ph256): Likewise.
3649 (_mm256_zextph128_ph256): Likewise.
3651 2021-09-17 liuhongt <hongtao.liu@intel.com>
3653 * config/i386/avx512fp16intrin.h (_mm_cvtsh_ss):
3655 (_mm_mask_cvtsh_ss): Likewise.
3656 (_mm_maskz_cvtsh_ss): Likewise.
3657 (_mm_cvtsh_sd): Likewise.
3658 (_mm_mask_cvtsh_sd): Likewise.
3659 (_mm_maskz_cvtsh_sd): Likewise.
3660 (_mm_cvt_roundsh_ss): Likewise.
3661 (_mm_mask_cvt_roundsh_ss): Likewise.
3662 (_mm_maskz_cvt_roundsh_ss): Likewise.
3663 (_mm_cvt_roundsh_sd): Likewise.
3664 (_mm_mask_cvt_roundsh_sd): Likewise.
3665 (_mm_maskz_cvt_roundsh_sd): Likewise.
3666 (_mm_cvtss_sh): Likewise.
3667 (_mm_mask_cvtss_sh): Likewise.
3668 (_mm_maskz_cvtss_sh): Likewise.
3669 (_mm_cvtsd_sh): Likewise.
3670 (_mm_mask_cvtsd_sh): Likewise.
3671 (_mm_maskz_cvtsd_sh): Likewise.
3672 (_mm_cvt_roundss_sh): Likewise.
3673 (_mm_mask_cvt_roundss_sh): Likewise.
3674 (_mm_maskz_cvt_roundss_sh): Likewise.
3675 (_mm_cvt_roundsd_sh): Likewise.
3676 (_mm_mask_cvt_roundsd_sh): Likewise.
3677 (_mm_maskz_cvt_roundsd_sh): Likewise.
3678 * config/i386/i386-builtin-types.def
3679 (V8HF_FTYPE_V2DF_V8HF_V8HF_UQI_INT,
3680 V8HF_FTYPE_V4SF_V8HF_V8HF_UQI_INT,
3681 V2DF_FTYPE_V8HF_V2DF_V2DF_UQI_INT,
3682 V4SF_FTYPE_V8HF_V4SF_V4SF_UQI_INT): Add new builtin types.
3683 * config/i386/i386-builtin.def: Add corrresponding new builtins.
3684 * config/i386/i386-expand.c: Handle new builtin types.
3685 * config/i386/sse.md (VF48_128): New mode iterator.
3686 (avx512fp16_vcvtsh2<ssescalarmodesuffix><mask_scalar_name><round_saeonly_scalar_name>):
3688 (avx512fp16_vcvt<ssescalarmodesuffix>2sh<mask_scalar_name><round_scalar_name>):
3691 2021-09-17 liuhongt <hongtao.liu@intel.com>
3693 * config/i386/avx512fp16intrin.h (_mm512_cvtph_pd):
3695 (_mm512_mask_cvtph_pd): Likewise.
3696 (_mm512_maskz_cvtph_pd): Likewise.
3697 (_mm512_cvt_roundph_pd): Likewise.
3698 (_mm512_mask_cvt_roundph_pd): Likewise.
3699 (_mm512_maskz_cvt_roundph_pd): Likewise.
3700 (_mm512_cvtxph_ps): Likewise.
3701 (_mm512_mask_cvtxph_ps): Likewise.
3702 (_mm512_maskz_cvtxph_ps): Likewise.
3703 (_mm512_cvtx_roundph_ps): Likewise.
3704 (_mm512_mask_cvtx_roundph_ps): Likewise.
3705 (_mm512_maskz_cvtx_roundph_ps): Likewise.
3706 (_mm512_cvtxps_ph): Likewise.
3707 (_mm512_mask_cvtxps_ph): Likewise.
3708 (_mm512_maskz_cvtxps_ph): Likewise.
3709 (_mm512_cvtx_roundps_ph): Likewise.
3710 (_mm512_mask_cvtx_roundps_ph): Likewise.
3711 (_mm512_maskz_cvtx_roundps_ph): Likewise.
3712 (_mm512_cvtpd_ph): Likewise.
3713 (_mm512_mask_cvtpd_ph): Likewise.
3714 (_mm512_maskz_cvtpd_ph): Likewise.
3715 (_mm512_cvt_roundpd_ph): Likewise.
3716 (_mm512_mask_cvt_roundpd_ph): Likewise.
3717 (_mm512_maskz_cvt_roundpd_ph): Likewise.
3718 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_pd):
3720 (_mm_mask_cvtph_pd): Likewise.
3721 (_mm_maskz_cvtph_pd): Likewise.
3722 (_mm256_cvtph_pd): Likewise.
3723 (_mm256_mask_cvtph_pd): Likewise.
3724 (_mm256_maskz_cvtph_pd): Likewise.
3725 (_mm_cvtxph_ps): Likewise.
3726 (_mm_mask_cvtxph_ps): Likewise.
3727 (_mm_maskz_cvtxph_ps): Likewise.
3728 (_mm256_cvtxph_ps): Likewise.
3729 (_mm256_mask_cvtxph_ps): Likewise.
3730 (_mm256_maskz_cvtxph_ps): Likewise.
3731 (_mm_cvtxps_ph): Likewise.
3732 (_mm_mask_cvtxps_ph): Likewise.
3733 (_mm_maskz_cvtxps_ph): Likewise.
3734 (_mm256_cvtxps_ph): Likewise.
3735 (_mm256_mask_cvtxps_ph): Likewise.
3736 (_mm256_maskz_cvtxps_ph): Likewise.
3737 (_mm_cvtpd_ph): Likewise.
3738 (_mm_mask_cvtpd_ph): Likewise.
3739 (_mm_maskz_cvtpd_ph): Likewise.
3740 (_mm256_cvtpd_ph): Likewise.
3741 (_mm256_mask_cvtpd_ph): Likewise.
3742 (_mm256_maskz_cvtpd_ph): Likewise.
3743 * config/i386/i386-builtin.def: Add corresponding new builtins.
3744 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3745 * config/i386/i386-expand.c: Handle new builtin types.
3746 * config/i386/sse.md
3747 (VF4_128_8_256): New.
3748 (VF48H_AVX512VL): Ditto.
3749 (ssePHmode): Add HF vector modes.
3750 (castmode): Add new convertable modes.
3753 (avx512fp16_vcvt<castmode>2ph_<mode><mask_name><round_name>): Ditto.
3754 (avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
3755 (*avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
3756 (avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
3757 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
3758 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask_1): Ditto.
3759 (avx512fp16_float_extend_ph<mode>2<mask_name><round_saeonly_name>):
3761 (avx512fp16_float_extend_ph<mode>2<mask_name>): Ditto.
3762 (*avx512fp16_float_extend_ph<mode>2_load<mask_name>): Ditto.
3763 (avx512fp16_float_extend_phv2df2<mask_name>): Ditto.
3764 (*avx512fp16_float_extend_phv2df2_load<mask_name>): Ditto.
3766 2021-09-17 liuhongt <hongtao.liu@intel.com>
3768 * config/i386/avx512fp16intrin.h (_mm_cvttsh_i32):
3770 (_mm_cvttsh_u32): Likewise.
3771 (_mm_cvtt_roundsh_i32): Likewise.
3772 (_mm_cvtt_roundsh_u32): Likewise.
3773 (_mm_cvttsh_i64): Likewise.
3774 (_mm_cvttsh_u64): Likewise.
3775 (_mm_cvtt_roundsh_i64): Likewise.
3776 (_mm_cvtt_roundsh_u64): Likewise.
3777 * config/i386/i386-builtin.def: Add corresponding new builtins.
3778 * config/i386/sse.md
3779 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<round_saeonly_name>):
3782 2021-09-17 liuhongt <hongtao.liu@intel.com>
3784 * config/i386/avx512fp16intrin.h (_mm512_cvttph_epi32):
3786 (_mm512_mask_cvttph_epi32): Likewise.
3787 (_mm512_maskz_cvttph_epi32): Likewise.
3788 (_mm512_cvtt_roundph_epi32): Likewise.
3789 (_mm512_mask_cvtt_roundph_epi32): Likewise.
3790 (_mm512_maskz_cvtt_roundph_epi32): Likewise.
3791 (_mm512_cvttph_epu32): Likewise.
3792 (_mm512_mask_cvttph_epu32): Likewise.
3793 (_mm512_maskz_cvttph_epu32): Likewise.
3794 (_mm512_cvtt_roundph_epu32): Likewise.
3795 (_mm512_mask_cvtt_roundph_epu32): Likewise.
3796 (_mm512_maskz_cvtt_roundph_epu32): Likewise.
3797 (_mm512_cvttph_epi64): Likewise.
3798 (_mm512_mask_cvttph_epi64): Likewise.
3799 (_mm512_maskz_cvttph_epi64): Likewise.
3800 (_mm512_cvtt_roundph_epi64): Likewise.
3801 (_mm512_mask_cvtt_roundph_epi64): Likewise.
3802 (_mm512_maskz_cvtt_roundph_epi64): Likewise.
3803 (_mm512_cvttph_epu64): Likewise.
3804 (_mm512_mask_cvttph_epu64): Likewise.
3805 (_mm512_maskz_cvttph_epu64): Likewise.
3806 (_mm512_cvtt_roundph_epu64): Likewise.
3807 (_mm512_mask_cvtt_roundph_epu64): Likewise.
3808 (_mm512_maskz_cvtt_roundph_epu64): Likewise.
3809 (_mm512_cvttph_epi16): Likewise.
3810 (_mm512_mask_cvttph_epi16): Likewise.
3811 (_mm512_maskz_cvttph_epi16): Likewise.
3812 (_mm512_cvtt_roundph_epi16): Likewise.
3813 (_mm512_mask_cvtt_roundph_epi16): Likewise.
3814 (_mm512_maskz_cvtt_roundph_epi16): Likewise.
3815 (_mm512_cvttph_epu16): Likewise.
3816 (_mm512_mask_cvttph_epu16): Likewise.
3817 (_mm512_maskz_cvttph_epu16): Likewise.
3818 (_mm512_cvtt_roundph_epu16): Likewise.
3819 (_mm512_mask_cvtt_roundph_epu16): Likewise.
3820 (_mm512_maskz_cvtt_roundph_epu16): Likewise.
3821 * config/i386/avx512fp16vlintrin.h (_mm_cvttph_epi32):
3823 (_mm_mask_cvttph_epi32): Likewise.
3824 (_mm_maskz_cvttph_epi32): Likewise.
3825 (_mm256_cvttph_epi32): Likewise.
3826 (_mm256_mask_cvttph_epi32): Likewise.
3827 (_mm256_maskz_cvttph_epi32): Likewise.
3828 (_mm_cvttph_epu32): Likewise.
3829 (_mm_mask_cvttph_epu32): Likewise.
3830 (_mm_maskz_cvttph_epu32): Likewise.
3831 (_mm256_cvttph_epu32): Likewise.
3832 (_mm256_mask_cvttph_epu32): Likewise.
3833 (_mm256_maskz_cvttph_epu32): Likewise.
3834 (_mm_cvttph_epi64): Likewise.
3835 (_mm_mask_cvttph_epi64): Likewise.
3836 (_mm_maskz_cvttph_epi64): Likewise.
3837 (_mm256_cvttph_epi64): Likewise.
3838 (_mm256_mask_cvttph_epi64): Likewise.
3839 (_mm256_maskz_cvttph_epi64): Likewise.
3840 (_mm_cvttph_epu64): Likewise.
3841 (_mm_mask_cvttph_epu64): Likewise.
3842 (_mm_maskz_cvttph_epu64): Likewise.
3843 (_mm256_cvttph_epu64): Likewise.
3844 (_mm256_mask_cvttph_epu64): Likewise.
3845 (_mm256_maskz_cvttph_epu64): Likewise.
3846 (_mm_cvttph_epi16): Likewise.
3847 (_mm_mask_cvttph_epi16): Likewise.
3848 (_mm_maskz_cvttph_epi16): Likewise.
3849 (_mm256_cvttph_epi16): Likewise.
3850 (_mm256_mask_cvttph_epi16): Likewise.
3851 (_mm256_maskz_cvttph_epi16): Likewise.
3852 (_mm_cvttph_epu16): Likewise.
3853 (_mm_mask_cvttph_epu16): Likewise.
3854 (_mm_maskz_cvttph_epu16): Likewise.
3855 (_mm256_cvttph_epu16): Likewise.
3856 (_mm256_mask_cvttph_epu16): Likewise.
3857 (_mm256_maskz_cvttph_epu16): Likewise.
3858 * config/i386/i386-builtin.def: Add new builtins.
3859 * config/i386/sse.md
3860 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name><round_saeonly_name>):
3862 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name>): Ditto.
3863 (*avx512fp16_fix<fixunssuffix>_trunc<mode>2_load<mask_name>): Ditto.
3864 (avx512fp16_fix<fixunssuffix>_truncv2di2<mask_name>): Ditto.
3865 (avx512fp16_fix<fixunssuffix>_truncv2di2_load<mask_name>): Ditto.
3867 2021-09-17 liuhongt <hongtao.liu@intel.com>
3869 * config/i386/avx512fp16intrin.h (_mm_cvtsh_i32): New intrinsic.
3870 (_mm_cvtsh_u32): Likewise.
3871 (_mm_cvt_roundsh_i32): Likewise.
3872 (_mm_cvt_roundsh_u32): Likewise.
3873 (_mm_cvtsh_i64): Likewise.
3874 (_mm_cvtsh_u64): Likewise.
3875 (_mm_cvt_roundsh_i64): Likewise.
3876 (_mm_cvt_roundsh_u64): Likewise.
3877 (_mm_cvti32_sh): Likewise.
3878 (_mm_cvtu32_sh): Likewise.
3879 (_mm_cvt_roundi32_sh): Likewise.
3880 (_mm_cvt_roundu32_sh): Likewise.
3881 (_mm_cvti64_sh): Likewise.
3882 (_mm_cvtu64_sh): Likewise.
3883 (_mm_cvt_roundi64_sh): Likewise.
3884 (_mm_cvt_roundu64_sh): Likewise.
3885 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3886 * config/i386/i386-builtin.def: Add corresponding new builtins.
3887 * config/i386/i386-expand.c (ix86_expand_round_builtin):
3888 Handle new builtin types.
3889 * config/i386/sse.md
3890 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix><round_name>):
3892 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix>_2): Likewise.
3893 (avx512fp16_vcvt<floatsuffix>si2sh<rex64namesuffix><round_name>): Likewise.
3895 2021-09-16 Bill Schmidt <wschmidt@linux.ibm.com>
3897 * config/rs6000/rs6000-c.c (rs6000-builtins.h): New include.
3898 (altivec_resolve_new_overloaded_builtin): New forward decl.
3899 (rs6000_new_builtin_type_compatible): New function.
3900 (altivec_resolve_overloaded_builtin): Call
3901 altivec_resolve_new_overloaded_builtin.
3902 (altivec_build_new_resolved_builtin): New function.
3903 (altivec_resolve_new_overloaded_builtin): Likewise.
3904 * config/rs6000/rs6000-call.c (rs6000_new_builtin_is_supported):
3906 * config/rs6000/rs6000-gen-builtins.c (write_decls): Remove _p from
3907 name of rs6000_new_builtin_is_supported.
3909 2021-09-16 Uroš Bizjak <ubizjak@gmail.com>
3911 * config/i386/i386-protos.h (ix86_decompose_address):
3912 Change return type to bool.
3913 * config/i386/i386.c (ix86_decompose_address): Ditto.
3915 2021-09-16 Tobias Burnus <tobias@codesourcery.com>
3918 * config/rs6000/t-rs6000 (build/rs6000-gen-builtins.o, build/rbtree.o):
3919 Added 'build/' to target, use build/%.o rule.
3920 (build/rs6000-gen-builtins$(build_exeext)): Add 'build/' and
3921 '$(build_exeext)' to target and 'build/' for the *.o files.
3922 (rs6000-builtins.c): Update for those changes; run rs6000-gen-builtins
3925 2021-09-16 Martin Jambor <mjambor@suse.cz>
3927 * cgraph.c (cgraph_node::dump): Do not check caller count sums if
3928 the body has been removed. Remove trailing whitespace.
3930 2021-09-16 Richard Biener <rguenther@suse.de>
3932 PR middle-end/102360
3933 * internal-fn.c (expand_DEFERRED_INIT): Make pattern-init
3934 of non-memory more robust.
3936 2021-09-16 Daniel Cederman <cederman@gaisler.com>
3938 * config/sparc/sparc-opts.h (enum sparc_processor_type): Add LEON5
3939 * config/sparc/sparc.c (struct processor_costs): Add LEON5 costs
3940 (leon5_adjust_cost): Increase cost of store with data dependency
3941 on ALU instruction and FPU anti-dependencies.
3942 (sparc_option_override): Add LEON5 costs
3943 (sparc_adjust_cost): Add LEON5 cost adjustments
3944 * config/sparc/sparc.h: Add LEON5
3945 * config/sparc/sparc.md: Include LEON5 scheduling information
3946 * config/sparc/sparc.opt: Add LEON5
3947 * doc/invoke.texi: Add LEON5
3948 * config/sparc/leon5.md: New file.
3950 2021-09-16 Daniel Cederman <cederman@gaisler.com>
3952 * config/sparc/sparc.md (stack_protect_set32): Add NOP to prevent
3953 sensitive sequence for B2BST errata workaround.
3955 2021-09-16 Daniel Cederman <cederman@gaisler.com>
3957 * config/sparc/sparc.c (sparc_do_work_around_errata): Do not begin
3958 functions with atomic instruction in the UT700 errata workaround.
3960 2021-09-16 Daniel Cederman <cederman@gaisler.com>
3962 * config/sparc/sparc.c (next_active_non_empty_insn): New function
3963 that returns next active non empty assembly instruction.
3964 (sparc_do_work_around_errata): Use new function.
3966 2021-09-16 Daniel Cederman <cederman@gaisler.com>
3968 * config/sparc/sparc.c (store_insn_p): Add predicate for store
3970 (load_insn_p): Add predicate for load attributes.
3971 (sparc_do_work_around_errata): Use new predicates.
3973 2021-09-16 Andreas Larsson <andreas@gaisler.com>
3975 * config/sparc/sparc.c (dump_target_flag_bits): Print bit names for
3978 2021-09-16 Martin Liska <mliska@suse.cz>
3980 * config/mips/netbsd.h: Fix typo in name of a macro.
3982 2021-09-16 liuhongt <hongtao.liu@intel.com>
3984 PR middle-end/102080
3985 * match.pd: Check mask type when doing cond_op related gimple
3987 * tree.c (is_truth_type_for): New function.
3988 * tree.h (is_truth_type_for): New declaration.
3990 2021-09-16 liuhongt <hongtao.liu@intel.com>
3992 * config/i386/avx512fp16intrin.h (_mm512_cvtepi32_ph): New
3994 (_mm512_mask_cvtepi32_ph): Likewise.
3995 (_mm512_maskz_cvtepi32_ph): Likewise.
3996 (_mm512_cvt_roundepi32_ph): Likewise.
3997 (_mm512_mask_cvt_roundepi32_ph): Likewise.
3998 (_mm512_maskz_cvt_roundepi32_ph): Likewise.
3999 (_mm512_cvtepu32_ph): Likewise.
4000 (_mm512_mask_cvtepu32_ph): Likewise.
4001 (_mm512_maskz_cvtepu32_ph): Likewise.
4002 (_mm512_cvt_roundepu32_ph): Likewise.
4003 (_mm512_mask_cvt_roundepu32_ph): Likewise.
4004 (_mm512_maskz_cvt_roundepu32_ph): Likewise.
4005 (_mm512_cvtepi64_ph): Likewise.
4006 (_mm512_mask_cvtepi64_ph): Likewise.
4007 (_mm512_maskz_cvtepi64_ph): Likewise.
4008 (_mm512_cvt_roundepi64_ph): Likewise.
4009 (_mm512_mask_cvt_roundepi64_ph): Likewise.
4010 (_mm512_maskz_cvt_roundepi64_ph): Likewise.
4011 (_mm512_cvtepu64_ph): Likewise.
4012 (_mm512_mask_cvtepu64_ph): Likewise.
4013 (_mm512_maskz_cvtepu64_ph): Likewise.
4014 (_mm512_cvt_roundepu64_ph): Likewise.
4015 (_mm512_mask_cvt_roundepu64_ph): Likewise.
4016 (_mm512_maskz_cvt_roundepu64_ph): Likewise.
4017 (_mm512_cvtepi16_ph): Likewise.
4018 (_mm512_mask_cvtepi16_ph): Likewise.
4019 (_mm512_maskz_cvtepi16_ph): Likewise.
4020 (_mm512_cvt_roundepi16_ph): Likewise.
4021 (_mm512_mask_cvt_roundepi16_ph): Likewise.
4022 (_mm512_maskz_cvt_roundepi16_ph): Likewise.
4023 (_mm512_cvtepu16_ph): Likewise.
4024 (_mm512_mask_cvtepu16_ph): Likewise.
4025 (_mm512_maskz_cvtepu16_ph): Likewise.
4026 (_mm512_cvt_roundepu16_ph): Likewise.
4027 (_mm512_mask_cvt_roundepu16_ph): Likewise.
4028 (_mm512_maskz_cvt_roundepu16_ph): Likewise.
4029 * config/i386/avx512fp16vlintrin.h (_mm_cvtepi32_ph): New
4031 (_mm_mask_cvtepi32_ph): Likewise.
4032 (_mm_maskz_cvtepi32_ph): Likewise.
4033 (_mm256_cvtepi32_ph): Likewise.
4034 (_mm256_mask_cvtepi32_ph): Likewise.
4035 (_mm256_maskz_cvtepi32_ph): Likewise.
4036 (_mm_cvtepu32_ph): Likewise.
4037 (_mm_mask_cvtepu32_ph): Likewise.
4038 (_mm_maskz_cvtepu32_ph): Likewise.
4039 (_mm256_cvtepu32_ph): Likewise.
4040 (_mm256_mask_cvtepu32_ph): Likewise.
4041 (_mm256_maskz_cvtepu32_ph): Likewise.
4042 (_mm_cvtepi64_ph): Likewise.
4043 (_mm_mask_cvtepi64_ph): Likewise.
4044 (_mm_maskz_cvtepi64_ph): Likewise.
4045 (_mm256_cvtepi64_ph): Likewise.
4046 (_mm256_mask_cvtepi64_ph): Likewise.
4047 (_mm256_maskz_cvtepi64_ph): Likewise.
4048 (_mm_cvtepu64_ph): Likewise.
4049 (_mm_mask_cvtepu64_ph): Likewise.
4050 (_mm_maskz_cvtepu64_ph): Likewise.
4051 (_mm256_cvtepu64_ph): Likewise.
4052 (_mm256_mask_cvtepu64_ph): Likewise.
4053 (_mm256_maskz_cvtepu64_ph): Likewise.
4054 (_mm_cvtepi16_ph): Likewise.
4055 (_mm_mask_cvtepi16_ph): Likewise.
4056 (_mm_maskz_cvtepi16_ph): Likewise.
4057 (_mm256_cvtepi16_ph): Likewise.
4058 (_mm256_mask_cvtepi16_ph): Likewise.
4059 (_mm256_maskz_cvtepi16_ph): Likewise.
4060 (_mm_cvtepu16_ph): Likewise.
4061 (_mm_mask_cvtepu16_ph): Likewise.
4062 (_mm_maskz_cvtepu16_ph): Likewise.
4063 (_mm256_cvtepu16_ph): Likewise.
4064 (_mm256_mask_cvtepu16_ph): Likewise.
4065 (_mm256_maskz_cvtepu16_ph): Likewise.
4066 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
4067 * config/i386/i386-builtin.def: Add corresponding new builtins.
4068 * config/i386/i386-expand.c
4069 (ix86_expand_args_builtin): Handle new builtin types.
4070 (ix86_expand_round_builtin): Ditto.
4071 * config/i386/i386-modes.def: Declare V2HF and V6HF.
4072 * config/i386/sse.md (VI2H_AVX512VL): New.
4074 (sseintvecmode): Add HF vector modes.
4075 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode><mask_name><round_name>):
4077 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
4078 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
4079 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
4080 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
4081 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask_1): Ditto.
4082 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
4083 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
4084 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
4085 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
4086 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask_1): Ditto.
4087 * config/i386/subst.md (round_qq2phsuff): New subst_attr.
4089 2021-09-16 liuhongt <hongtao.liu@intel.com>
4091 * config/i386/avx512fp16intrin.h (_mm512_cvtph_epi32):
4093 (_mm512_mask_cvtph_epi32): Likewise.
4094 (_mm512_maskz_cvtph_epi32): Likewise.
4095 (_mm512_cvt_roundph_epi32): Likewise.
4096 (_mm512_mask_cvt_roundph_epi32): Likewise.
4097 (_mm512_maskz_cvt_roundph_epi32): Likewise.
4098 (_mm512_cvtph_epu32): Likewise.
4099 (_mm512_mask_cvtph_epu32): Likewise.
4100 (_mm512_maskz_cvtph_epu32): Likewise.
4101 (_mm512_cvt_roundph_epu32): Likewise.
4102 (_mm512_mask_cvt_roundph_epu32): Likewise.
4103 (_mm512_maskz_cvt_roundph_epu32): Likewise.
4104 (_mm512_cvtph_epi64): Likewise.
4105 (_mm512_mask_cvtph_epi64): Likewise.
4106 (_mm512_maskz_cvtph_epi64): Likewise.
4107 (_mm512_cvt_roundph_epi64): Likewise.
4108 (_mm512_mask_cvt_roundph_epi64): Likewise.
4109 (_mm512_maskz_cvt_roundph_epi64): Likewise.
4110 (_mm512_cvtph_epu64): Likewise.
4111 (_mm512_mask_cvtph_epu64): Likewise.
4112 (_mm512_maskz_cvtph_epu64): Likewise.
4113 (_mm512_cvt_roundph_epu64): Likewise.
4114 (_mm512_mask_cvt_roundph_epu64): Likewise.
4115 (_mm512_maskz_cvt_roundph_epu64): Likewise.
4116 (_mm512_cvtph_epi16): Likewise.
4117 (_mm512_mask_cvtph_epi16): Likewise.
4118 (_mm512_maskz_cvtph_epi16): Likewise.
4119 (_mm512_cvt_roundph_epi16): Likewise.
4120 (_mm512_mask_cvt_roundph_epi16): Likewise.
4121 (_mm512_maskz_cvt_roundph_epi16): Likewise.
4122 (_mm512_cvtph_epu16): Likewise.
4123 (_mm512_mask_cvtph_epu16): Likewise.
4124 (_mm512_maskz_cvtph_epu16): Likewise.
4125 (_mm512_cvt_roundph_epu16): Likewise.
4126 (_mm512_mask_cvt_roundph_epu16): Likewise.
4127 (_mm512_maskz_cvt_roundph_epu16): Likewise.
4128 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_epi32):
4130 (_mm_mask_cvtph_epi32): Likewise.
4131 (_mm_maskz_cvtph_epi32): Likewise.
4132 (_mm256_cvtph_epi32): Likewise.
4133 (_mm256_mask_cvtph_epi32): Likewise.
4134 (_mm256_maskz_cvtph_epi32): Likewise.
4135 (_mm_cvtph_epu32): Likewise.
4136 (_mm_mask_cvtph_epu32): Likewise.
4137 (_mm_maskz_cvtph_epu32): Likewise.
4138 (_mm256_cvtph_epu32): Likewise.
4139 (_mm256_mask_cvtph_epu32): Likewise.
4140 (_mm256_maskz_cvtph_epu32): Likewise.
4141 (_mm_cvtph_epi64): Likewise.
4142 (_mm_mask_cvtph_epi64): Likewise.
4143 (_mm_maskz_cvtph_epi64): Likewise.
4144 (_mm256_cvtph_epi64): Likewise.
4145 (_mm256_mask_cvtph_epi64): Likewise.
4146 (_mm256_maskz_cvtph_epi64): Likewise.
4147 (_mm_cvtph_epu64): Likewise.
4148 (_mm_mask_cvtph_epu64): Likewise.
4149 (_mm_maskz_cvtph_epu64): Likewise.
4150 (_mm256_cvtph_epu64): Likewise.
4151 (_mm256_mask_cvtph_epu64): Likewise.
4152 (_mm256_maskz_cvtph_epu64): Likewise.
4153 (_mm_cvtph_epi16): Likewise.
4154 (_mm_mask_cvtph_epi16): Likewise.
4155 (_mm_maskz_cvtph_epi16): Likewise.
4156 (_mm256_cvtph_epi16): Likewise.
4157 (_mm256_mask_cvtph_epi16): Likewise.
4158 (_mm256_maskz_cvtph_epi16): Likewise.
4159 (_mm_cvtph_epu16): Likewise.
4160 (_mm_mask_cvtph_epu16): Likewise.
4161 (_mm_maskz_cvtph_epu16): Likewise.
4162 (_mm256_cvtph_epu16): Likewise.
4163 (_mm256_mask_cvtph_epu16): Likewise.
4164 (_mm256_maskz_cvtph_epu16): Likewise.
4165 * config/i386/i386-builtin-types.def: Add new builtin types.
4166 * config/i386/i386-builtin.def: Add new builtins.
4167 * config/i386/i386-expand.c
4168 (ix86_expand_args_builtin): Handle new builtin types.
4169 (ix86_expand_round_builtin): Ditto.
4170 * config/i386/sse.md (sseintconvert): New.
4172 (UNSPEC_US_FIX_NOTRUNC): Ditto.
4173 (sseintconvertsignprefix): Ditto.
4174 (avx512fp16_vcvtph2<sseintconvertsignprefix><sseintconvert>_<mode><mask_name><round_name>):
4177 2021-09-16 liuhongt <hongtao.liu@intel.com>
4179 * config/i386/avx512fp16intrin.h: (_mm_cvtsi16_si128):
4181 (_mm_cvtsi128_si16): Likewise.
4182 (_mm_mask_load_sh): Likewise.
4183 (_mm_maskz_load_sh): Likewise.
4184 (_mm_mask_store_sh): Likewise.
4185 (_mm_move_sh): Likewise.
4186 (_mm_mask_move_sh): Likewise.
4187 (_mm_maskz_move_sh): Likewise.
4188 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
4189 * config/i386/i386-builtin.def: Add corresponding new builtins.
4190 * config/i386/i386-expand.c
4191 (ix86_expand_special_args_builtin): Handle new builtin types.
4192 (ix86_expand_vector_init_one_nonzero): Adjust for FP16 target.
4193 * config/i386/sse.md (VI2F): New mode iterator.
4194 (vec_set<mode>_0): Use new mode iterator.
4195 (avx512f_mov<ssescalarmodelower>_mask): Adjust for HF vector mode.
4196 (avx512f_store<mode>_mask): Ditto.
4198 2021-09-16 Kewen Lin <linkw@linux.ibm.com>
4200 * config/rs6000/rs6000.opt (-mtoc-fusion): Remove.
4202 2021-09-15 David Edelsohn <dje.gcc@gmail.com>
4204 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
4205 Proceed if no symbol summary or the symbol alias flag is false.
4207 2021-09-15 Jakub Jelinek <jakub@redhat.com>
4211 * varasm.c (output_constructor_regular_field): Instead of assertion
4212 that array_size_for_constructor result is equal to size of
4213 TREE_TYPE (local->val) in bytes, assert that the type size is greater
4214 or equal to array_size_for_constructor result and use type size as
4217 2021-09-15 Martin Liska <mliska@suse.cz>
4220 * config/i386/vxworks.h: Use new macro TARGET_CPU_P.
4222 2021-09-15 Martin Liska <mliska@suse.cz>
4225 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
4226 Check that we have a symbol summary for a symbol.
4228 2021-09-15 Richard Biener <rguenther@suse.de>
4231 * config/rs6000/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
4232 to inherit from elfos.h
4234 2021-09-15 liuhongt <hongtao.liu@intel.com>
4237 * config/i386/i386-expand.c
4238 (ix86_expand_vector_init_interleave): Use puncklwd to pack 2
4240 (ix86_expand_vector_set): Use blendw instead of pinsrw.
4241 * config/i386/i386.c (ix86_can_change_mode_class): Adjust for
4242 AVX512FP16 which supports 16bit vector load.
4243 * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>):
4245 (avx512bw_interleave_high<mode><mask_name>): .. this, and
4246 extend to V32HFmode.
4247 (avx2_interleave_highv16hi<mask_name>): Rename to ..
4248 (avx2_interleave_high<mode><mask_name>): .. this, and extend
4250 (vec_interleave_highv8hi<mask_name>): Rename to ..
4251 (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode.
4252 (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>):
4254 (<mask_codefor>avx512bw_interleave_low<mode><mask_name>):
4255 this, and extend to V32HFmode.
4256 (avx2_interleave_lowv16hi<mask_name>): Rename to ..
4257 (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode.
4258 (vec_interleave_lowv8hi<mask_name>): Rename to ..
4259 (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode.
4260 (sse4_1_pblendw): Rename to ..
4261 (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode.
4262 (avx2_pblendph): New define_expand.
4263 (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use
4264 sseintmodesuffix instead of ssemodesuffix.
4265 (blendsuf): New mode attr.
4267 2021-09-15 Richard Biener <rguenther@suse.de>
4269 * tree-vectorizer.h (dr_misalignment): Move out of line.
4270 (dr_target_alignment): New.
4271 (DR_TARGET_ALIGNMENT): Wrap dr_target_alignment.
4272 (set_dr_target_alignment): New.
4273 (SET_DR_TARGET_ALIGNMENT): Wrap set_dr_target_alignment.
4274 * tree-vect-data-refs.c (dr_misalignment): Compute and
4275 return the group members misalignment.
4276 (vect_compute_data_ref_alignment): Use SET_DR_TARGET_ALIGNMENT.
4277 (vect_analyze_data_refs_alignment): Compute alignment only
4278 for the first element of a DR group.
4279 (vect_slp_analyze_node_alignment): Likewise.
4281 2021-09-15 Hongyu Wang <hongyu.wang@intel.com>
4283 * config/i386/avx512fp16intrin.h: Adjust all builtin calls.
4284 * config/i386/avx512fp16vlintrin.h: Likewise.
4285 * config/i386/i386-builtin.def: Adjust builtin name and
4286 enumeration to match AVX512F style.
4288 2021-09-15 Richard Biener <rguenther@suse.de>
4290 PR tree-optimization/102318
4291 * tree-vect-loop.c (vect_transform_cycle_phi): Revert
4292 previous change and do the mode conversion separately from
4293 the sign conversion.
4295 2021-09-15 Hongtao Liu <hongtao.liu@intel.com>
4296 Peter Cordes <peter@cordes.ca>
4299 * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
4300 (*vec_extract<mode><ssescalarmodelower>_valign): Output
4301 vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
4304 2021-09-15 Richard Biener <rguenther@suse.de>
4306 * config.gcc: Remove vax-*-openbsd* configuration.
4308 2021-09-15 Richard Biener <rguenther@suse.de>
4310 * config.gcc: Remove m68k-openbsd.
4312 2021-09-15 Max Filippov <jcmvbkbc@gmail.com>
4315 * config/xtensa/t-xtensa (TM_H): Add include/xtensa-config.h.
4317 2021-09-14 Peter Bergner <bergner@linux.ibm.com>
4319 * config/rs6000/mma.md (unspec): Delete UNSPEC_MMA_XXSETACCZ.
4320 (unspecv): Add UNSPECV_MMA_XXSETACCZ.
4321 (*mma_xxsetaccz): Delete.
4322 (mma_xxsetaccz): Change to define_insn. Remove operand 1.
4323 Use UNSPECV_MMA_XXSETACCZ. Update comment.
4324 * config/rs6000/rs6000.c (rs6000_rtx_costs): Use UNSPECV_MMA_XXSETACCZ.
4326 2021-09-14 Iain Sandoe <iain@sandoe.co.uk>
4328 * Makefile.in: Remove variables related to applying no-PIE
4329 to the exes on $build.
4330 * configure: Regenerate.
4331 * configure.ac: Remove configuration related to applying
4332 no-PIE to the exes on $build.
4334 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
4336 * config/arc/arc.md (doloop_end): Add missing mode.
4337 (loop_end): Likewise.
4339 2021-09-14 Jakub Jelinek <jakub@redhat.com>
4341 * gimplify.c (goa_stabilize_expr): Add depth argument, propagate
4342 it to recursive calls, for depth above 7 just gimplify or return.
4343 Perform a test even for MODIFY_EXPR, ADDR_EXPR, COMPOUND_EXPR with
4344 __builtin_clear_padding and TARGET_EXPR.
4345 (gimplify_omp_atomic): Adjust goa_stabilize_expr callers.
4347 2021-09-14 liuhongt <hongtao.liu@intel.com>
4349 * config/i386/avx512fp16intrin.h (_mm_fpclass_sh_mask):
4351 (_mm_mask_fpclass_sh_mask): Likewise.
4352 (_mm512_mask_fpclass_ph_mask): Likewise.
4353 (_mm512_fpclass_ph_mask): Likewise.
4354 (_mm_getexp_sh): Likewise.
4355 (_mm_mask_getexp_sh): Likewise.
4356 (_mm_maskz_getexp_sh): Likewise.
4357 (_mm512_getexp_ph): Likewise.
4358 (_mm512_mask_getexp_ph): Likewise.
4359 (_mm512_maskz_getexp_ph): Likewise.
4360 (_mm_getexp_round_sh): Likewise.
4361 (_mm_mask_getexp_round_sh): Likewise.
4362 (_mm_maskz_getexp_round_sh): Likewise.
4363 (_mm512_getexp_round_ph): Likewise.
4364 (_mm512_mask_getexp_round_ph): Likewise.
4365 (_mm512_maskz_getexp_round_ph): Likewise.
4366 (_mm_getmant_sh): Likewise.
4367 (_mm_mask_getmant_sh): Likewise.
4368 (_mm_maskz_getmant_sh): Likewise.
4369 (_mm512_getmant_ph): Likewise.
4370 (_mm512_mask_getmant_ph): Likewise.
4371 (_mm512_maskz_getmant_ph): Likewise.
4372 (_mm_getmant_round_sh): Likewise.
4373 (_mm_mask_getmant_round_sh): Likewise.
4374 (_mm_maskz_getmant_round_sh): Likewise.
4375 (_mm512_getmant_round_ph): Likewise.
4376 (_mm512_mask_getmant_round_ph): Likewise.
4377 (_mm512_maskz_getmant_round_ph): Likewise.
4378 * config/i386/avx512fp16vlintrin.h (_mm_mask_fpclass_ph_mask):
4380 (_mm_fpclass_ph_mask): Likewise.
4381 (_mm256_mask_fpclass_ph_mask): Likewise.
4382 (_mm256_fpclass_ph_mask): Likewise.
4383 (_mm256_getexp_ph): Likewise.
4384 (_mm256_mask_getexp_ph): Likewise.
4385 (_mm256_maskz_getexp_ph): Likewise.
4386 (_mm_getexp_ph): Likewise.
4387 (_mm_mask_getexp_ph): Likewise.
4388 (_mm_maskz_getexp_ph): Likewise.
4389 (_mm256_getmant_ph): Likewise.
4390 (_mm256_mask_getmant_ph): Likewise.
4391 (_mm256_maskz_getmant_ph): Likewise.
4392 (_mm_getmant_ph): Likewise.
4393 (_mm_mask_getmant_ph): Likewise.
4394 (_mm_maskz_getmant_ph): Likewise.
4395 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
4396 * config/i386/i386-builtin.def: Add corresponding new builtins.
4397 * config/i386/i386-expand.c
4398 (ix86_expand_args_builtin): Handle new builtin types.
4399 (ix86_expand_round_builtin): Ditto.
4400 * config/i386/sse.md (vecmemsuffix): Add HF vector modes.
4401 (<avx512>_getexp<mode><mask_name><round_saeonly_name>): Adjust
4402 to support HF vector modes.
4403 (avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name):
4405 (avx512dq_fpclass<mode><mask_scalar_merge_name>): Ditto.
4406 (avx512dq_vmfpclass<mode><mask_scalar_merge_name>): Ditto.
4407 (<avx512>_getmant<mode><mask_name><round_saeonly_name>): Ditto.
4408 (avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4411 2021-09-14 liuhongt <hongtao.liu@intel.com>
4413 * config/i386/avx512fp16intrin.h (_mm512_reduce_ph):
4415 (_mm512_mask_reduce_ph): Likewise.
4416 (_mm512_maskz_reduce_ph): Likewise.
4417 (_mm512_reduce_round_ph): Likewise.
4418 (_mm512_mask_reduce_round_ph): Likewise.
4419 (_mm512_maskz_reduce_round_ph): Likewise.
4420 (_mm_reduce_sh): Likewise.
4421 (_mm_mask_reduce_sh): Likewise.
4422 (_mm_maskz_reduce_sh): Likewise.
4423 (_mm_reduce_round_sh): Likewise.
4424 (_mm_mask_reduce_round_sh): Likewise.
4425 (_mm_maskz_reduce_round_sh): Likewise.
4426 (_mm512_roundscale_ph): Likewise.
4427 (_mm512_mask_roundscale_ph): Likewise.
4428 (_mm512_maskz_roundscale_ph): Likewise.
4429 (_mm512_roundscale_round_ph): Likewise.
4430 (_mm512_mask_roundscale_round_ph): Likewise.
4431 (_mm512_maskz_roundscale_round_ph): Likewise.
4432 (_mm_roundscale_sh): Likewise.
4433 (_mm_mask_roundscale_sh): Likewise.
4434 (_mm_maskz_roundscale_sh): Likewise.
4435 (_mm_roundscale_round_sh): Likewise.
4436 (_mm_mask_roundscale_round_sh): Likewise.
4437 (_mm_maskz_roundscale_round_sh): Likewise.
4438 * config/i386/avx512fp16vlintrin.h: (_mm_reduce_ph):
4440 (_mm_mask_reduce_ph): Likewise.
4441 (_mm_maskz_reduce_ph): Likewise.
4442 (_mm256_reduce_ph): Likewise.
4443 (_mm256_mask_reduce_ph): Likewise.
4444 (_mm256_maskz_reduce_ph): Likewise.
4445 (_mm_roundscale_ph): Likewise.
4446 (_mm_mask_roundscale_ph): Likewise.
4447 (_mm_maskz_roundscale_ph): Likewise.
4448 (_mm256_roundscale_ph): Likewise.
4449 (_mm256_mask_roundscale_ph): Likewise.
4450 (_mm256_maskz_roundscale_ph): Likewise.
4451 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
4452 * config/i386/i386-builtin.def: Add corresponding new builtins.
4453 * config/i386/i386-expand.c
4454 (ix86_expand_args_builtin): Handle new builtin types.
4455 (ix86_expand_round_builtin): Ditto.
4456 * config/i386/sse.md (<mask_codefor>reducep<mode><mask_name>):
4458 (<mask_codefor>reducep<mode><mask_name><round_saeonly_name>):
4459 ... this, and adjust for round operands.
4460 (reduces<mode><mask_scalar_name>): Likewise, with ...
4461 (reduces<mode><mask_scalar_name><round_saeonly_scalar_name):
4463 (<avx512>_rndscale<mode><mask_name><round_saeonly_name>):
4464 Adjust for HF vector modes.
4465 (avx512f_rndscale<mode><mask_scalar_name><round_saeonly_scalar_name>):
4467 (*avx512f_rndscale<mode><round_saeonly_name>): Ditto.
4469 2021-09-14 liuhongt <hongtao.liu@intel.com>
4471 * config/i386/avx512fp16intrin.h: (_mm512_rcp_ph):
4473 (_mm512_mask_rcp_ph): Likewise.
4474 (_mm512_maskz_rcp_ph): Likewise.
4475 (_mm_rcp_sh): Likewise.
4476 (_mm_mask_rcp_sh): Likewise.
4477 (_mm_maskz_rcp_sh): Likewise.
4478 (_mm512_scalef_ph): Likewise.
4479 (_mm512_mask_scalef_ph): Likewise.
4480 (_mm512_maskz_scalef_ph): Likewise.
4481 (_mm512_scalef_round_ph): Likewise.
4482 (_mm512_mask_scalef_round_ph): Likewise.
4483 (_mm512_maskz_scalef_round_ph): Likewise.
4484 (_mm_scalef_sh): Likewise.
4485 (_mm_mask_scalef_sh): Likewise.
4486 (_mm_maskz_scalef_sh): Likewise.
4487 (_mm_scalef_round_sh): Likewise.
4488 (_mm_mask_scalef_round_sh): Likewise.
4489 (_mm_maskz_scalef_round_sh): Likewise.
4490 * config/i386/avx512fp16vlintrin.h (_mm_rcp_ph):
4492 (_mm256_rcp_ph): Likewise.
4493 (_mm_mask_rcp_ph): Likewise.
4494 (_mm256_mask_rcp_ph): Likewise.
4495 (_mm_maskz_rcp_ph): Likewise.
4496 (_mm256_maskz_rcp_ph): Likewise.
4497 (_mm_scalef_ph): Likewise.
4498 (_mm256_scalef_ph): Likewise.
4499 (_mm_mask_scalef_ph): Likewise.
4500 (_mm256_mask_scalef_ph): Likewise.
4501 (_mm_maskz_scalef_ph): Likewise.
4502 (_mm256_maskz_scalef_ph): Likewise.
4503 * config/i386/i386-builtin.def: Add new builtins.
4504 * config/i386/sse.md (VFH_AVX512VL): New.
4505 (avx512fp16_rcp<mode>2<mask_name>): Ditto.
4506 (avx512fp16_vmrcpv8hf2<mask_scalar_name>): Ditto.
4507 (avx512f_vmscalef<mode><mask_scalar_name><round_scalar_name>):
4508 Adjust to support HF vector modes.
4509 (<avx512>_scalef<mode><mask_name><round_name>): Ditto.
4511 2021-09-14 liuhongt <hongtao.liu@intel.com>
4513 * config/i386/avx512fp16intrin.h: (_mm512_sqrt_ph):
4515 (_mm512_mask_sqrt_ph): Likewise.
4516 (_mm512_maskz_sqrt_ph): Likewise.
4517 (_mm512_sqrt_round_ph): Likewise.
4518 (_mm512_mask_sqrt_round_ph): Likewise.
4519 (_mm512_maskz_sqrt_round_ph): Likewise.
4520 (_mm512_rsqrt_ph): Likewise.
4521 (_mm512_mask_rsqrt_ph): Likewise.
4522 (_mm512_maskz_rsqrt_ph): Likewise.
4523 (_mm_rsqrt_sh): Likewise.
4524 (_mm_mask_rsqrt_sh): Likewise.
4525 (_mm_maskz_rsqrt_sh): Likewise.
4526 (_mm_sqrt_sh): Likewise.
4527 (_mm_mask_sqrt_sh): Likewise.
4528 (_mm_maskz_sqrt_sh): Likewise.
4529 (_mm_sqrt_round_sh): Likewise.
4530 (_mm_mask_sqrt_round_sh): Likewise.
4531 (_mm_maskz_sqrt_round_sh): Likewise.
4532 * config/i386/avx512fp16vlintrin.h (_mm_sqrt_ph): New intrinsic.
4533 (_mm256_sqrt_ph): Likewise.
4534 (_mm_mask_sqrt_ph): Likewise.
4535 (_mm256_mask_sqrt_ph): Likewise.
4536 (_mm_maskz_sqrt_ph): Likewise.
4537 (_mm256_maskz_sqrt_ph): Likewise.
4538 (_mm_rsqrt_ph): Likewise.
4539 (_mm256_rsqrt_ph): Likewise.
4540 (_mm_mask_rsqrt_ph): Likewise.
4541 (_mm256_mask_rsqrt_ph): Likewise.
4542 (_mm_maskz_rsqrt_ph): Likewise.
4543 (_mm256_maskz_rsqrt_ph): Likewise.
4544 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
4545 * config/i386/i386-builtin.def: Add corresponding new builtins.
4546 * config/i386/i386-expand.c
4547 (ix86_expand_args_builtin): Handle new builtins.
4548 (ix86_expand_round_builtin): Ditto.
4549 * config/i386/sse.md (VF_AVX512FP16VL): New.
4550 (sqrt<mode>2): Adjust for HF vector modes.
4551 (<sse>_sqrt<mode>2<mask_name><round_name>): Likewise.
4552 (<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
4554 (<sse>_rsqrt<mode>2<mask_name>): New.
4555 (avx512fp16_vmrsqrtv8hf2<mask_scalar_name>): Likewise.
4557 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
4560 * diagnostic-spec.c (warning_suppressed_at, copy_warning): Handle
4561 'RESERVED_LOCATION_P' locations.
4562 * warning-control.cc (get_nowarn_spec, suppress_warning)
4563 (copy_warning): Likewise.
4565 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
4567 * diagnostic-spec.h (typedef xint_hash_t): Use 'location_t' instead of...
4568 (typedef key_type_t): ... this. Remove.
4569 (nowarn_map): Document.
4570 * diagnostic-spec.c (nowarn_map): Likewise.
4571 * warning-control.cc (convert_to_key): Evolve functions into...
4572 (get_location): ... these. Adjust all users.
4574 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
4576 * warning-control.cc (copy_warning): Remove 'nowarn_map' setup.
4578 2021-09-13 Jason Merrill <jason@redhat.com>
4580 * params.opt: Add destructive-interference-size and
4581 constructive-interference-size.
4582 * doc/invoke.texi: Document them.
4583 * config/aarch64/aarch64.c (aarch64_override_options_internal):
4585 * config/arm/arm.c (arm_option_override): Set them.
4586 * config/i386/i386-options.c (ix86_option_override_internal):
4589 2021-09-13 Martin Liska <mliska@suse.cz>
4590 H.J. Lu <hjl.tools@gmail.com>
4593 * common/config/i386/cpuinfo.h (cpu_indicator_init): Add support
4594 for x86-64 micro levels for __builtin_cpu_supports.
4595 * common/config/i386/i386-cpuinfo.h (enum feature_priority):
4596 Add priorities for the micro-arch levels.
4597 (enum processor_features): Add new features.
4598 * common/config/i386/i386-isas.h: Add micro-arch features.
4599 * config/i386/i386-builtins.c (get_builtin_code_for_version):
4600 Support the micro-arch levels by callsing
4601 __builtin_cpu_supports.
4602 * doc/extend.texi: Document that the levels are support by
4603 __builtin_cpu_supports.
4605 2021-09-13 Andrew Pinski <apinski@marvell.com>
4608 * config/aarch64/aarch64-builtins.c (aarch64_fold_builtin_lane_check):
4610 (aarch64_general_fold_builtin): Handle AARCH64_SIMD_BUILTIN_LANE_CHECK.
4611 (aarch64_general_gimple_fold_builtin): Likewise.
4613 2021-09-13 Andrew Pinski <apinski@marvell.com>
4615 * config.gcc: Add m32r-*-linux* and m32rle-*-linux*
4616 to the Unsupported targets list.
4617 Remove support for m32r-*-linux* and m32rle-*-linux*.
4618 * config/m32r/linux.h: Removed.
4619 * config/m32r/t-linux: Removed.
4621 2021-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4624 * config/aarch64/aarch64.c (aarch64_classify_address): Don't allow
4625 register index for SVE predicate modes.
4627 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
4629 * tree-ssa-threadbackward.c
4630 (back_threader_profitability::profitable_path_p): Remove FSM
4632 (back_threader_registry::register_path): Same.
4633 * tree-ssa-threadedge.c
4634 (jump_threader::simplify_control_stmt_condition): Same.
4635 * tree-ssa-threadupdate.c (jt_path_registry::jt_path_registry):
4636 Add backedge_threads argument.
4637 (fwd_jt_path_registry::fwd_jt_path_registry): Pass
4638 backedge_threads argument.
4639 (back_jt_path_registry::back_jt_path_registry): Same.
4640 (dump_jump_thread_path): Adjust for FSM removal.
4641 (back_jt_path_registry::rewire_first_differing_edge): Same.
4642 (back_jt_path_registry::adjust_paths_after_duplication): Same.
4643 (back_jt_path_registry::update_cfg): Same.
4644 (jt_path_registry::register_jump_thread): Same.
4645 * tree-ssa-threadupdate.h (enum jump_thread_edge_type): Remove
4647 (class back_jt_path_registry): Add backedge_threads to
4650 2021-09-13 Martin Liska <mliska@suse.cz>
4653 * asan.h (sanitize_coverage_p): Handle when fn == NULL.
4655 2021-09-13 H.J. Lu <hjl.tools@gmail.com>
4658 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): New.
4659 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
4660 (MOVE_MAX): Check TARGET_AVX256_MOVE_BY_PIECES and
4661 TARGET_AVX256_STORE_BY_PIECES instead of
4662 TARGET_AVX256_SPLIT_UNALIGNED_LOAD and
4663 TARGET_AVX256_SPLIT_UNALIGNED_STORE.
4664 (STORE_MAX_PIECES): Check TARGET_AVX256_STORE_BY_PIECES instead
4665 of TARGET_AVX256_SPLIT_UNALIGNED_STORE.
4666 * config/i386/x86-tune.def (X86_TUNE_AVX256_MOVE_BY_PIECES): New.
4667 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
4669 2021-09-13 liuhongt <hongtao.liu@intel.com>
4672 * expmed.c (extract_bit_field_using_extv): Use
4673 gen_lowpart_if_possible instead of gen_lowpart to avoid ICE.
4675 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
4677 * Makefile.in (OBJS): Add value-pointer-equiv.o.
4678 * gimple-ssa-evrp.c (class ssa_equiv_stack): Move to
4679 value-pointer-equiv.*.
4680 (ssa_equiv_stack::ssa_equiv_stack): Same.
4681 (ssa_equiv_stack::enter): Same.
4682 (ssa_equiv_stack::leave): Same.
4683 (ssa_equiv_stack::push_replacement): Same.
4684 (ssa_equiv_stack::get_replacement): Same.
4685 (is_pointer_ssa): Same.
4686 (class pointer_equiv_analyzer): Same.
4687 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
4688 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Same.
4689 (pointer_equiv_analyzer::set_global_equiv): Same.
4690 (pointer_equiv_analyzer::set_cond_equiv): Same.
4691 (pointer_equiv_analyzer::get_equiv): Same.
4692 (pointer_equiv_analyzer::enter): Same.
4693 (pointer_equiv_analyzer::leave): Same.
4694 (pointer_equiv_analyzer::get_equiv_expr): Same.
4695 (pta_valueize): Same.
4696 (pointer_equiv_analyzer::visit_stmt): Same.
4697 (pointer_equiv_analyzer::visit_edge): Same.
4698 (hybrid_folder::value_of_expr): Same.
4699 (hybrid_folder::value_on_edge): Same.
4700 * value-pointer-equiv.cc: New file.
4701 * value-pointer-equiv.h: New file.
4703 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
4706 * gimple-fold.c (gimple_fold_builtin_memory_op): Allow folding
4707 memcpy if the size is not more than MOVE_MAX * MOVE_RATIO.
4709 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
4712 * config/arm/arm.md (movmisaligndi): New define_expand.
4713 * config/arm/vec-common.md (movmisalign<mode>): Iterate over VDQ mode.
4715 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
4718 * emit-rtl.c (gen_highpart): Use adjust_address to handle
4719 MEM rather than calling simplify_gen_subreg.
4721 2021-09-13 Jan-Benedict Glaw <jbglaw@ług-owl.de>
4723 * config/alpha/vms.h (INIT_CUMULATIVE_ARGS): Wrap multi-statment
4724 define into a block.
4726 2021-09-13 Richard Biener <rguenther@suse.de>
4728 * config/darwin.h (DARWIN_PREFER_DWARF): Do not define.
4729 * config/i386/darwin.h (PREFERRED_DEBUGGING_TYPE): Do not
4730 change based on DARWIN_PREFER_DWARF not being defined.
4732 2021-09-13 Richard Biener <rguenther@suse.de>
4734 * config/i386/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
4735 to inherit from elfos.h
4737 2021-09-13 Richard Biener <rguenther@suse.de>
4739 * config.gcc: Add cr16-*-* to the list of obsoleted targets.
4741 2021-09-13 Richard Biener <rguenther@suse.de>
4743 * config/avr/elf.h (PREFERRED_DEBUGGING_TYPE): Remove
4744 override, pick up DWARF2_DEBUG define from elfos.h
4746 2021-09-13 Richard Biener <rguenther@suse.de>
4748 * config/rx/rx.h (PREFERRED_DEBUGGING_TYPE): Always define to
4751 2021-09-13 Richard Biener <rguenther@suse.de>
4753 * config/alpha/vms.h (PREFERRED_DEBUGGING_TYPE): Define to
4756 2021-09-13 Richard Biener <rguenther@suse.de>
4758 * config/i386/cygming.h: Always default to DWARF2 debugging.
4759 Do not define DBX_DEBUGGING_INFO, that's done via dbxcoff.h
4761 * doc/install.texi: Document binutils 2.16 as minimum
4762 requirement for mingw.
4764 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
4766 * config/rs6000/rs6000.c (struct rs6000_cost_data): New members
4767 nstmts, nloads and extra_ctor_cost.
4768 (rs6000_density_test): Add load density related heuristics. Do
4769 extra costing on vector construction statements if need.
4770 (rs6000_init_cost): Init new members.
4771 (rs6000_update_target_cost_per_stmt): New function.
4772 (rs6000_add_stmt_cost): Factor vect_nonmem hunk out to function
4773 rs6000_update_target_cost_per_stmt and call it.
4775 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
4777 * config/rs6000/rs6000.c (struct rs6000_cost_data): Remove typedef.
4778 (rs6000_init_cost): Adjust.
4780 2021-09-13 liuhongt <hongtao.liu@intel.com>
4782 * config/i386/i386.md: (UNSPEC_COPYSIGN): Remove.
4783 (UNSPEC_XORSIGN): Ditto.
4785 2021-09-12 Roger Sayle <roger@nextmovesoftware.com>
4787 * expr.c (convert_move): Preserve SUBREG_PROMOTED_VAR_P when
4788 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
4791 2021-09-11 Aldy Hernandez <aldyh@redhat.com>
4793 * tree-ssa-threadbackward.c (class back_threader_registry): Use
4794 back_jt_path_registry.
4795 * tree-ssa-threadedge.c (jump_threader::jump_threader): Use
4796 fwd_jt_path_registry.
4797 * tree-ssa-threadedge.h (class jump_threader): Same..
4798 * tree-ssa-threadupdate.c
4799 (jump_thread_path_registry::jump_thread_path_registry): Rename...
4800 (jt_path_registry::jt_path_registry): ...to this.
4801 (jump_thread_path_registry::~jump_thread_path_registry): Rename...
4802 (jt_path_registry::~jt_path_registry): ...this.
4803 (fwd_jt_path_registry::fwd_jt_path_registry): New.
4804 (fwd_jt_path_registry::~fwd_jt_path_registry): New.
4805 (jump_thread_path_registry::allocate_thread_edge): Rename...
4806 (jt_path_registry::allocate_thread_edge): ...to this.
4807 (jump_thread_path_registry::allocate_thread_path): Rename...
4808 (jt_path_registry::allocate_thread_path): ...to this.
4809 (jump_thread_path_registry::lookup_redirection_data): Rename...
4810 (fwd_jt_path_registry::lookup_redirection_data): ...to this.
4811 (jump_thread_path_registry::thread_block_1): Rename...
4812 (fwd_jt_path_registry::thread_block_1): ...to this.
4813 (jump_thread_path_registry::thread_block): Rename...
4814 (fwd_jt_path_registry::thread_block): ...to this.
4815 (jt_path_registry::thread_through_loop_header): Rename...
4816 (fwd_jt_path_registry::thread_through_loop_header): ...to this.
4817 (jump_thread_path_registry::mark_threaded_blocks): Rename...
4818 (fwd_jt_path_registry::mark_threaded_blocks): ...to this.
4819 (jump_thread_path_registry::debug_path): Rename...
4820 (jt_path_registry::debug_path): ...to this.
4821 (jump_thread_path_registry::dump): Rename...
4822 (jt_path_registry::debug): ...to this.
4823 (jump_thread_path_registry::rewire_first_differing_edge): Rename...
4824 (back_jt_path_registry::rewire_first_differing_edge): ...to this.
4825 (jump_thread_path_registry::adjust_paths_after_duplication): Rename...
4826 (back_jt_path_registry::adjust_paths_after_duplication): ...to this.
4827 (jump_thread_path_registry::duplicate_thread_path): Rename...
4828 (back_jt_path_registry::duplicate_thread_path): ...to this. Also,
4829 drop ill-formed candidates.
4830 (jump_thread_path_registry::remove_jump_threads_including): Rename...
4831 (fwd_jt_path_registry::remove_jump_threads_including): ...to this.
4832 (jt_path_registry::thread_through_all_blocks): New.
4833 (back_jt_path_registry::update_cfg): New.
4834 (fwd_jt_path_registry::update_cfg): New.
4835 (jump_thread_path_registry::register_jump_thread): Rename...
4836 (jt_path_registry::register_jump_thread): ...to this.
4837 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
4839 (class jt_path_registry): ...here.
4840 (class fwd_jt_path_registry): New.
4841 (class back_jt_path_registry): New.
4843 2021-09-10 liuhongt <hongtao.liu@intel.com>
4846 2021-09-01 liuhongt <hongtao.liu@intel.com>
4848 * emit-rtl.c (validate_subreg): Get rid of all float-int
4851 2021-09-10 Jakub Jelinek <jakub@redhat.com>
4853 * tree-core.h (enum omp_memory_order): Add OMP_MEMORY_ORDER_MASK,
4854 OMP_FAIL_MEMORY_ORDER_UNSPECIFIED, OMP_FAIL_MEMORY_ORDER_RELAXED,
4855 OMP_FAIL_MEMORY_ORDER_ACQUIRE, OMP_FAIL_MEMORY_ORDER_RELEASE,
4856 OMP_FAIL_MEMORY_ORDER_ACQ_REL, OMP_FAIL_MEMORY_ORDER_SEQ_CST and
4857 OMP_FAIL_MEMORY_ORDER_MASK enumerators.
4858 (OMP_FAIL_MEMORY_ORDER_SHIFT): Define.
4859 * gimple-pretty-print.c (dump_gimple_omp_atomic_load,
4860 dump_gimple_omp_atomic_store): Print [weak] for weak atomic
4862 * gimple.h (enum gf_mask): Change GF_OMP_ATOMIC_MEMORY_ORDER
4863 to 6-bit mask, adjust GF_OMP_ATOMIC_NEED_VALUE value and add
4865 (gimple_omp_atomic_weak_p, gimple_omp_atomic_set_weak): New inline
4867 * tree.h (OMP_ATOMIC_WEAK): Define.
4868 * tree-pretty-print.c (dump_omp_atomic_memory_order): Adjust for
4869 fail memory order being encoded in the same enum and also print
4870 fail clause if present.
4871 (dump_generic_node): Print weak clause if OMP_ATOMIC_WEAK.
4872 * gimplify.c (goa_stabilize_expr): Add target_expr and rhs arguments,
4873 handle pre_p == NULL case as a test mode that only returns value
4874 but doesn't change gimplify nor change anything otherwise, adjust
4875 recursive calls, add MODIFY_EXPR, ADDR_EXPR, COND_EXPR, TARGET_EXPR
4876 and CALL_EXPR handling, adjust COMPOUND_EXPR handling for
4877 __builtin_clear_padding calls, for !rhs gimplify as lvalue rather
4879 (gimplify_omp_atomic): Adjust goa_stabilize_expr caller. Handle
4880 COND_EXPR rhs. Set weak flag on gimple load/store for
4882 * omp-expand.c (omp_memory_order_to_fail_memmodel): New function.
4883 (omp_memory_order_to_memmodel): Adjust for fail clause encoded
4885 (expand_omp_atomic_cas): New function.
4886 (expand_omp_atomic_pipeline): Use omp_memory_order_to_fail_memmodel
4888 (expand_omp_atomic): Attempt to optimize atomic compare and exchange
4889 using expand_omp_atomic_cas.
4891 2021-09-10 Aldy Hernandez <aldyh@redhat.com>
4892 Michael Matz <matz@suse.de>
4894 * tree-pass.h (PROP_loop_opts_done): New.
4895 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
4896 Intersect with global range.
4897 * tree-ssa-loop.c (tree_ssa_loop_done): Set PROP_loop_opts_done.
4898 * tree-ssa-threadbackward.c
4899 (back_threader_profitability::profitable_path_p): Disable
4900 threading through latches until after loop optimizations have run.
4902 2021-09-10 David Faust <david.faust@oracle.com>
4904 * doc/invoke.texi: Document BPF -mcpu, -mjmpext, -mjmp32 and -malu32
4907 2021-09-10 David Faust <david.faust@oracle.com>
4909 * config/bpf/bpf-opts.h (bpf_isa_version): New enum.
4910 * config/bpf/bpf-protos.h (bpf_expand_cbranch): New.
4911 * config/bpf/bpf.c (bpf_option_override): Handle -mcpu option.
4912 (bpf_expand_cbranch): New function.
4913 * config/bpf/bpf.md (AM mode iterator): Conditionalize support for SI
4915 (zero_extendsidi2): Only use mov32 instruction if it is available.
4916 (SIM mode iterator): Conditionalize support for SI mode.
4917 (JM mode iterator): New.
4918 (cbranchdi4): Update name, use new JM iterator. Use bpf_expand_cbranch.
4919 (*branch_on_di): Update name, use new JM iterator.
4920 * config/bpf/bpf.opt: (mjmpext): New option.
4924 (bpf_isa): New enum.
4926 2021-09-10 David Faust <david.faust@oracle.com>
4928 * config/bpf/bpf.md (zero_extendhidi2): Add new output template
4929 for register-to-register extensions.
4930 (zero_extendqidi2): Likewise.
4932 2021-09-10 Richard Biener <rguenther@suse.de>
4934 PR middle-end/102273
4935 * internal-fn.c (expand_DEFERRED_INIT): Always expand non-SSA vars.
4937 2021-09-10 Richard Biener <rguenther@suse.de>
4939 PR middle-end/102269
4940 * gimplify.c (is_var_need_auto_init): Empty types do not need
4943 2021-09-10 Richard Biener <rguenther@suse.de>
4945 * configure.ac (--with-stabs): Remove.
4946 * configure: Regenerate.
4947 * doc/install.texi: Remove --with-stabs documentation.
4949 2021-09-10 liuhongt <hongtao.liu@intel.com>
4951 * config/i386/avx512fp16intrin.h: (_mm512_cmp_ph_mask):
4953 (_mm512_mask_cmp_ph_mask): Likewise.
4954 (_mm512_cmp_round_ph_mask): Likewise.
4955 (_mm512_mask_cmp_round_ph_mask): Likewise.
4956 (_mm_cmp_sh_mask): Likewise.
4957 (_mm_mask_cmp_sh_mask): Likewise.
4958 (_mm_cmp_round_sh_mask): Likewise.
4959 (_mm_mask_cmp_round_sh_mask): Likewise.
4960 (_mm_comieq_sh): Likewise.
4961 (_mm_comilt_sh): Likewise.
4962 (_mm_comile_sh): Likewise.
4963 (_mm_comigt_sh): Likewise.
4964 (_mm_comige_sh): Likewise.
4965 (_mm_comineq_sh): Likewise.
4966 (_mm_ucomieq_sh): Likewise.
4967 (_mm_ucomilt_sh): Likewise.
4968 (_mm_ucomile_sh): Likewise.
4969 (_mm_ucomigt_sh): Likewise.
4970 (_mm_ucomige_sh): Likewise.
4971 (_mm_ucomineq_sh): Likewise.
4972 (_mm_comi_round_sh): Likewise.
4973 (_mm_comi_sh): Likewise.
4974 * config/i386/avx512fp16vlintrin.h (_mm_cmp_ph_mask): New intrinsic.
4975 (_mm_mask_cmp_ph_mask): Likewise.
4976 (_mm256_cmp_ph_mask): Likewise.
4977 (_mm256_mask_cmp_ph_mask): Likewise.
4978 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
4979 * config/i386/i386-builtin.def: Add corresponding new builtins.
4980 * config/i386/i386-expand.c
4981 (ix86_expand_args_builtin): Handle new builtin types.
4982 (ix86_expand_round_builtin): Ditto.
4983 * config/i386/i386.md (ssevecmode): Add HF mode.
4984 (MODEFH): New mode iterator.
4985 * config/i386/sse.md
4986 (V48H_AVX512VL): New mode iterator to support HF vector modes.
4987 Ajdust corresponding description.
4988 (ssecmpintprefix): New.
4989 (VI12_AVX512VL): Adjust to support HF vector modes.
4990 (cmp_imm_predicate): Likewise.
4991 (<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
4993 (avx512f_vmcmp<mode>3<round_saeonly_name>): Likewise.
4994 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Likewise.
4995 (<sse>_<unord>comi<round_saeonly_name>): Likewise.
4997 2021-09-10 liuhongt <hongtao.liu@intel.com>
4999 * config/i386/avx512fp16intrin.h: (_mm512_max_ph): New intrinsic.
5000 (_mm512_mask_max_ph): Likewise.
5001 (_mm512_maskz_max_ph): Likewise.
5002 (_mm512_min_ph): Likewise.
5003 (_mm512_mask_min_ph): Likewise.
5004 (_mm512_maskz_min_ph): Likewise.
5005 (_mm512_max_round_ph): Likewise.
5006 (_mm512_mask_max_round_ph): Likewise.
5007 (_mm512_maskz_max_round_ph): Likewise.
5008 (_mm512_min_round_ph): Likewise.
5009 (_mm512_mask_min_round_ph): Likewise.
5010 (_mm512_maskz_min_round_ph): Likewise.
5011 (_mm_max_sh): Likewise.
5012 (_mm_mask_max_sh): Likewise.
5013 (_mm_maskz_max_sh): Likewise.
5014 (_mm_min_sh): Likewise.
5015 (_mm_mask_min_sh): Likewise.
5016 (_mm_maskz_min_sh): Likewise.
5017 (_mm_max_round_sh): Likewise.
5018 (_mm_mask_max_round_sh): Likewise.
5019 (_mm_maskz_max_round_sh): Likewise.
5020 (_mm_min_round_sh): Likewise.
5021 (_mm_mask_min_round_sh): Likewise.
5022 (_mm_maskz_min_round_sh): Likewise.
5023 * config/i386/avx512fp16vlintrin.h (_mm_max_ph): New intrinsic.
5024 (_mm256_max_ph): Likewise.
5025 (_mm_mask_max_ph): Likewise.
5026 (_mm256_mask_max_ph): Likewise.
5027 (_mm_maskz_max_ph): Likewise.
5028 (_mm256_maskz_max_ph): Likewise.
5029 (_mm_min_ph): Likewise.
5030 (_mm256_min_ph): Likewise.
5031 (_mm_mask_min_ph): Likewise.
5032 (_mm256_mask_min_ph): Likewise.
5033 (_mm_maskz_min_ph): Likewise.
5034 (_mm256_maskz_min_ph): Likewise.
5035 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
5036 * config/i386/i386-builtin.def: Add corresponding new builtins.
5037 * config/i386/i386-expand.c
5038 (ix86_expand_args_builtin): Handle new builtin types.
5039 * config/i386/sse.md
5040 (<code><mode>3<mask_name><round_saeonly_name>): Adjust to
5041 support HF vector modes.
5042 (*<code><mode>3<mask_name><round_saeonly_name>): Likewise.
5043 (ieee_<ieee_maxmin><mode>3<mask_name><round_saeonly_name>):
5045 (<sse>_vm<code><mode>3<mask_scalar_name><round_saeonly_scalar_name>):
5047 * config/i386/subst.md (round_saeonly_mode512bit_condition):
5048 Adjust for HF vector modes.
5050 2021-09-10 Liu, Hongtao <hongtao.liu@intel.com>
5052 * config/i386/avx512fp16intrin.h (_mm_add_sh): New intrinsic.
5053 (_mm_mask_add_sh): Likewise.
5054 (_mm_maskz_add_sh): Likewise.
5055 (_mm_sub_sh): Likewise.
5056 (_mm_mask_sub_sh): Likewise.
5057 (_mm_maskz_sub_sh): Likewise.
5058 (_mm_mul_sh): Likewise.
5059 (_mm_mask_mul_sh): Likewise.
5060 (_mm_maskz_mul_sh): Likewise.
5061 (_mm_div_sh): Likewise.
5062 (_mm_mask_div_sh): Likewise.
5063 (_mm_maskz_div_sh): Likewise.
5064 (_mm_add_round_sh): Likewise.
5065 (_mm_mask_add_round_sh): Likewise.
5066 (_mm_maskz_add_round_sh): Likewise.
5067 (_mm_sub_round_sh): Likewise.
5068 (_mm_mask_sub_round_sh): Likewise.
5069 (_mm_maskz_sub_round_sh): Likewise.
5070 (_mm_mul_round_sh): Likewise.
5071 (_mm_mask_mul_round_sh): Likewise.
5072 (_mm_maskz_mul_round_sh): Likewise.
5073 (_mm_div_round_sh): Likewise.
5074 (_mm_mask_div_round_sh): Likewise.
5075 (_mm_maskz_div_round_sh): Likewise.
5076 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
5077 * config/i386/i386-builtin.def: Add corresponding new builtins.
5078 * config/i386/i386-expand.c
5079 (ix86_expand_round_builtin): Handle new builtins.
5080 * config/i386/sse.md (VF_128): Change description.
5081 (<sse>_vm<plusminus_insn><mode>3<mask_scalar_name><round_scalar_name>):
5082 Adjust to support HF vector modes.
5083 (<sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name><round_scalar_name>):
5086 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
5088 * config/i386/i386-expand.c
5089 (ix86_avx256_split_vector_move_misalign): Handle V16HF mode.
5090 * config/i386/i386.c
5091 (ix86_preferred_simd_mode): Handle HF mode.
5092 * config/i386/sse.md (V_256H): New mode iterator.
5093 (avx_vextractf128<mode>): Use it.
5094 (VEC_INIT_MODE): Align vector HFmode condition to vector
5095 HImodes since there're no real HF instruction used.
5096 (VEC_INIT_HALF_MODE): Ditto.
5098 (VIHF_AVX512BW): Ditto.
5099 (*vec_extracthf): Ditto.
5100 (VEC_EXTRACT_MODE): Ditto.
5102 2021-09-10 Richard Biener <rguenther@suse.de>
5105 * config/dbx.h: Remove.
5106 * config/dbxcoff.h: Do not define PREFERRED_DEBUGGING_TYPE.
5107 * config/lynx.h: Likewise.
5109 2021-09-10 liuhongt <hongtao.liu@intel.com>
5111 * config/i386/i386-expand.c (ix86_expand_copysign): Expand
5112 right into ANDNOT + AND + IOR, using paradoxical subregs.
5113 (ix86_split_copysign_const): Remove.
5114 (ix86_split_copysign_var): Ditto.
5115 * config/i386/i386-protos.h (ix86_split_copysign_const): Dotto.
5116 (ix86_split_copysign_var): Ditto.
5117 * config/i386/i386.md (@copysign<mode>3_const): Ditto.
5118 (@copysign<mode>3_var): Ditto.
5120 2021-09-09 qing zhao <qing.zhao@oracle.com>
5122 * builtins.c (expand_builtin_memset): Make external visible.
5123 * builtins.h (expand_builtin_memset): Declare extern.
5124 * common.opt (ftrivial-auto-var-init=): New option.
5125 * doc/extend.texi: Document the uninitialized attribute.
5126 * doc/invoke.texi: Document -ftrivial-auto-var-init.
5127 * flag-types.h (enum auto_init_type): New enumerated type
5129 * gimple-fold.c (clear_padding_type): Add one new parameter.
5130 (clear_padding_union): Likewise.
5131 (clear_padding_emit_loop): Likewise.
5132 (clear_type_padding_in_mask): Likewise.
5133 (gimple_fold_builtin_clear_padding): Handle this new parameter.
5134 * gimplify.c (gimple_add_init_for_auto_var): New function.
5135 (gimple_add_padding_init_for_auto_var): New function.
5136 (is_var_need_auto_init): New function.
5137 (gimplify_decl_expr): Add initialization to automatic variables per
5139 (gimplify_call_expr): Add one new parameter for call to
5140 __builtin_clear_padding.
5141 (gimplify_init_constructor): Add padding initialization in the end.
5142 * internal-fn.c (INIT_PATTERN_VALUE): New macro.
5143 (expand_DEFERRED_INIT): New function.
5144 * internal-fn.def (DEFERRED_INIT): New internal function.
5145 * tree-cfg.c (verify_gimple_call): Verify calls to .DEFERRED_INIT.
5146 * tree-sra.c (generate_subtree_deferred_init): New function.
5147 (scan_function): Avoid setting cannot_scalarize_away_bitmap for
5148 calls to .DEFERRED_INIT.
5149 (sra_modify_deferred_init): New function.
5150 (sra_modify_function_body): Handle calls to DEFERRED_INIT specially.
5151 * tree-ssa-structalias.c (find_func_aliases_for_call): Likewise.
5152 * tree-ssa-uninit.c (warn_uninit): Handle calls to DEFERRED_INIT
5154 (check_defs): Likewise.
5155 (warn_uninitialized_vars): Likewise.
5156 * tree-ssa.c (ssa_undefined_value_p): Likewise.
5157 * tree.c (build_common_builtin_nodes): Build tree node for
5158 BUILT_IN_CLEAR_PADDING when needed.
5160 2021-09-09 Richard Biener <rguenther@suse.de>
5162 * tree-ssa-loop-im.c (fill_always_executed_in_1): Walk
5165 2021-09-09 Richard Biener <rguenther@suse.de>
5167 * tree-ssa-loop-im.c (fill_always_executed_in_1): Integrate
5168 DOM walk from get_loop_body_in_dom_order using a worklist
5171 2021-09-09 liuhongt <hongtao.liu@intel.com>
5173 * config.gcc: Add avx512fp16vlintrin.h.
5174 * config/i386/avx512fp16intrin.h: (_mm512_add_ph): New intrinsic.
5175 (_mm512_mask_add_ph): Likewise.
5176 (_mm512_maskz_add_ph): Likewise.
5177 (_mm512_sub_ph): Likewise.
5178 (_mm512_mask_sub_ph): Likewise.
5179 (_mm512_maskz_sub_ph): Likewise.
5180 (_mm512_mul_ph): Likewise.
5181 (_mm512_mask_mul_ph): Likewise.
5182 (_mm512_maskz_mul_ph): Likewise.
5183 (_mm512_div_ph): Likewise.
5184 (_mm512_mask_div_ph): Likewise.
5185 (_mm512_maskz_div_ph): Likewise.
5186 (_mm512_add_round_ph): Likewise.
5187 (_mm512_mask_add_round_ph): Likewise.
5188 (_mm512_maskz_add_round_ph): Likewise.
5189 (_mm512_sub_round_ph): Likewise.
5190 (_mm512_mask_sub_round_ph): Likewise.
5191 (_mm512_maskz_sub_round_ph): Likewise.
5192 (_mm512_mul_round_ph): Likewise.
5193 (_mm512_mask_mul_round_ph): Likewise.
5194 (_mm512_maskz_mul_round_ph): Likewise.
5195 (_mm512_div_round_ph): Likewise.
5196 (_mm512_mask_div_round_ph): Likewise.
5197 (_mm512_maskz_div_round_ph): Likewise.
5198 * config/i386/avx512fp16vlintrin.h: New header.
5199 * config/i386/i386-builtin-types.def (V16HF, V8HF, V32HF):
5200 Add new builtin types.
5201 * config/i386/i386-builtin.def: Add corresponding builtins.
5202 * config/i386/i386-expand.c
5203 (ix86_expand_args_builtin): Handle new builtin types.
5204 (ix86_expand_round_builtin): Likewise.
5205 * config/i386/immintrin.h: Include avx512fp16vlintrin.h
5206 * config/i386/sse.md (VFH): New mode_iterator.
5208 (avx512fmaskmode): Add HF vector modes.
5209 (avx512fmaskhalfmode): Likewise.
5210 (<plusminus_insn><mode>3<mask_name><round_name>): Adjust to for
5212 (*<plusminus_insn><mode>3<mask_name><round_name>): Likewise.
5213 (mul<mode>3<mask_name><round_name>): Likewise.
5214 (*mul<mode>3<mask_name><round_name>): Likewise.
5215 (div<mode>3): Likewise.
5216 (<sse>_div<mode>3<mask_name><round_name>): Likewise.
5217 * config/i386/subst.md (SUBST_V): Add HF vector modes.
5218 (SUBST_A): Likewise.
5219 (round_mode512bit_condition): Adjust for V32HFmode.
5221 2021-09-09 liuhongt <hongtao.liu@intel.com>
5224 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
5225 (reduc_plus_scal_v4sf): .. this, New define_expand.
5226 (reduc_plus_scal_v2df): .. and this, New define_expand.
5228 2021-09-09 liuhongt <hongtao.liu@intel.com>
5231 * config/i386/sse.md (*vec_extract<mode><ssescalarmodelower>_valign):
5234 2021-09-08 Jonathan Wakely <jwakely@redhat.com>
5237 * doc/trouble.texi (Copy Assignment): Fix description of
5238 behaviour and fix code in example.
5240 2021-09-08 Segher Boessenkool <segher@kernel.crashing.org>
5243 * config/rs6000/rs6000-logue.c (rs6000_emit_epilogue): For ELFv2 use
5244 r11 instead of r12 for restoring CR.
5246 2021-09-08 Jakub Jelinek <jakub@redhat.com>
5247 liuhongt <hongtao.liu@intel.com>
5250 * config/i386/i386.md (@xorsign<mode>3_1): Remove.
5251 * config/i386/i386-expand.c (ix86_expand_xorsign): Expand right away
5252 into AND with mask and XOR, using paradoxical subregs.
5253 (ix86_split_xorsign): Remove.
5254 * config/i386/i386-protos.h (ix86_split_xorsign): Remove.
5256 2021-09-08 Di Zhao <dizhao@os.amperecomputing.com>
5258 * tree-ssa-sccvn.c (vn_nary_op_insert_into): fix result compare
5260 2021-09-08 Jakub Jelinek <jakub@redhat.com>
5263 * config/i386/i386.md (xorsign<mode>3): If operands[1] is equal to
5264 operands[2], emit abs<mode>2 instead.
5265 (@xorsign<mode>3_1): Add early-clobbers for output operand, enable
5266 first alternative even for avx, add another alternative with
5267 =&Yv <- 0, Yv, Yvm constraints.
5268 * config/i386/i386-expand.c (ix86_split_xorsign): If op0 is equal
5269 to op1, emit vpandn instead.
5271 2021-09-08 liuhongt <hongtao.liu@intel.com>
5273 * config/i386/avx512fp16intrin.h (_mm_set_ph): New intrinsic.
5274 (_mm256_set_ph): Likewise.
5275 (_mm512_set_ph): Likewise.
5276 (_mm_setr_ph): Likewise.
5277 (_mm256_setr_ph): Likewise.
5278 (_mm512_setr_ph): Likewise.
5279 (_mm_set1_ph): Likewise.
5280 (_mm256_set1_ph): Likewise.
5281 (_mm512_set1_ph): Likewise.
5282 (_mm_setzero_ph): Likewise.
5283 (_mm256_setzero_ph): Likewise.
5284 (_mm512_setzero_ph): Likewise.
5285 (_mm_set_sh): Likewise.
5286 (_mm_load_sh): Likewise.
5287 (_mm_store_sh): Likewise.
5288 * config/i386/i386-builtin-types.def (V8HF): New type.
5289 (DEF_FUNCTION_TYPE (V8HF, V8HI)): New builtin function type
5290 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
5291 Support vector HFmodes.
5292 (ix86_expand_vector_init_one_nonzero): Likewise.
5293 (ix86_expand_vector_init_one_var): Likewise.
5294 (ix86_expand_vector_init_interleave): Likewise.
5295 (ix86_expand_vector_init_general): Likewise.
5296 (ix86_expand_vector_set): Likewise.
5297 (ix86_expand_vector_extract): Likewise.
5298 (ix86_expand_vector_init_concat): Likewise.
5299 (ix86_expand_sse_movcc): Handle vector HFmodes.
5300 (ix86_expand_vector_set_var): Ditto.
5301 * config/i386/i386-modes.def: Add HF vector modes in comment.
5302 * config/i386/i386.c (classify_argument): Add HF vector modes.
5303 (ix86_hard_regno_mode_ok): Allow HF vector modes for AVX512FP16.
5304 (ix86_vector_mode_supported_p): Likewise.
5305 (ix86_set_reg_reg_cost): Handle vector HFmode.
5306 (ix86_get_ssemov): Handle vector HFmode.
5307 (function_arg_advance_64): Pass unamed V16HFmode and V32HFmode
5309 (function_arg_advance_32): Pass V8HF/V16HF/V32HF by sse reg for 32bit
5311 (function_arg_advance_32): Ditto.
5312 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): New.
5313 (VALID_AVX256_REG_OR_OI_MODE): Rename to ..
5314 (VALID_AVX256_REG_OR_OI_VHF_MODE): .. this, and add V16HF.
5315 (VALID_SSE2_REG_VHF_MODE): New.
5316 (VALID_AVX512VL_128_REG_MODE): Add V8HF and TImode.
5317 (SSE_REG_MODE_P): Add vector HFmode.
5318 * config/i386/i386.md (mode): Add HF vector modes.
5319 (MODE_SIZE): Likewise.
5320 (ssemodesuffix): Add ph suffix for HF vector modes.
5321 * config/i386/sse.md (VFH_128): New mode iterator.
5322 (VMOVE): Adjust for HF vector modes.
5324 (V_256_512): Likewise.
5326 (avx512fmaskmode): Likewise.
5327 (shuffletype): Likewise.
5328 (sseinsnmode): Likewise.
5329 (ssedoublevecmode): Likewise.
5330 (ssehalfvecmode): Likewise.
5331 (ssehalfvecmodelower): Likewise.
5332 (ssePScmode): Likewise.
5333 (ssescalarmode): Likewise.
5334 (ssescalarmodelower): Likewise.
5335 (sseintprefix): Likewise.
5337 (bcstscalarsuff): Likewise.
5338 (xtg_mode): Likewise.
5339 (VI12HF_AVX512VL): New mode_iterator.
5340 (VF_AVX512FP16): Likewise.
5342 (VIHF_256): Likewise.
5343 (VIHF_AVX512BW): Likewise.
5344 (V16_256): Likewise.
5345 (V32_512): Likewise.
5346 (sseintmodesuffix): New mode_attr.
5347 (sse): Add scalar and vector HFmodes.
5348 (ssescalarmode): Add vector HFmode mapping.
5349 (ssescalarmodesuffix): Add sh suffix for HFmode.
5350 (*<sse>_vm<insn><mode>3): Use VFH_128.
5351 (*<sse>_vm<multdiv_mnemonic><mode>3): Likewise.
5352 (*ieee_<ieee_maxmin><mode>3): Likewise.
5353 (<avx512>_blendm<mode>): New define_insn.
5354 (vec_setv8hf): New define_expand.
5355 (vec_set<mode>_0): New define_insn for HF vector set.
5356 (*avx512fp16_movsh): Likewise.
5357 (avx512fp16_movsh): Likewise.
5358 (vec_extract_lo_v32hi): Rename to ...
5359 (vec_extract_lo_<mode>): ... this, and adjust to allow HF
5361 (vec_extract_hi_v32hi): Likewise.
5362 (vec_extract_hi_<mode>): Likewise.
5363 (vec_extract_lo_v16hi): Likewise.
5364 (vec_extract_lo_<mode>): Likewise.
5365 (vec_extract_hi_v16hi): Likewise.
5366 (vec_extract_hi_<mode>): Likewise.
5367 (vec_set_hi_v16hi): Likewise.
5368 (vec_set_hi_<mode>): Likewise.
5369 (vec_set_lo_v16hi): Likewise.
5370 (vec_set_lo_<mode>): Likewise.
5371 (*vec_extract<mode>_0): New define_insn_and_split for HF
5373 (*vec_extracthf): New define_insn.
5374 (VEC_EXTRACT_MODE): Add HF vector modes.
5375 (PINSR_MODE): Add V8HF.
5376 (sse2p4_1): Likewise.
5377 (pinsr_evex_isa): Likewise.
5378 (<sse2p4_1>_pinsr<ssemodesuffix>): Adjust to support
5379 insert for V8HFmode.
5380 (pbroadcast_evex_isa): Add HF vector modes.
5381 (AVX2_VEC_DUP_MODE): Likewise.
5382 (VEC_INIT_MODE): Likewise.
5383 (VEC_INIT_HALF_MODE): Likewise.
5384 (avx2_pbroadcast<mode>): Adjust to support HF vector mode
5386 (avx2_pbroadcast<mode>_1): Likewise.
5387 (<avx512>_vec_dup<mode>_1): Likewise.
5388 (<avx512>_vec_dup<mode><mask_name>): Likewise.
5389 (<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>):
5392 2021-09-08 Guo, Xuepeng <xuepeng.guo@intel.com>
5393 H.J. Lu <hongjiu.lu@intel.com>
5394 Liu Hongtao <hongtao.liu@intel.com>
5395 Wang Hongyu <hongyu.wang@intel.com>
5396 Xu Dianhong <dianhong.xu@intel.com>
5398 * common/config/i386/cpuinfo.h (get_available_features):
5399 Detect FEATURE_AVX512FP16.
5400 * common/config/i386/i386-common.c
5401 (OPTION_MASK_ISA_AVX512FP16_SET,
5402 OPTION_MASK_ISA_AVX512FP16_UNSET,
5403 OPTION_MASK_ISA2_AVX512FP16_SET,
5404 OPTION_MASK_ISA2_AVX512FP16_UNSET): New.
5405 (OPTION_MASK_ISA2_AVX512BW_UNSET,
5406 OPTION_MASK_ISA2_AVX512BF16_UNSET): Add AVX512FP16.
5407 (ix86_handle_option): Handle -mavx512fp16.
5408 * common/config/i386/i386-cpuinfo.h (enum processor_features):
5409 Add FEATURE_AVX512FP16.
5410 * common/config/i386/i386-isas.h: Add entry for AVX512FP16.
5411 * config.gcc: Add avx512fp16intrin.h.
5412 * config/i386/avx512fp16intrin.h: New intrinsic header.
5413 * config/i386/cpuid.h: Add bit_AVX512FP16.
5414 * config/i386/i386-builtin-types.def: (FLOAT16): New primitive type.
5415 * config/i386/i386-builtins.c: Support _Float16 type for i386
5417 (ix86_register_float16_builtin_type): New function.
5418 (ix86_float16_type_node): New.
5419 * config/i386/i386-c.c (ix86_target_macros_internal): Define
5421 * config/i386/i386-expand.c (ix86_expand_branch): Support
5423 (ix86_prepare_fp_compare_args): Adjust TARGET_SSE_MATH &&
5424 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
5425 (ix86_expand_fp_movcc): Ditto.
5426 * config/i386/i386-isa.def: Add PTA define for AVX512FP16.
5427 * config/i386/i386-options.c (isa2_opts): Add -mavx512fp16.
5428 (ix86_valid_target_attribute_inner_p): Add avx512fp16 attribute.
5429 * config/i386/i386.c (ix86_get_ssemov): Use
5430 vmovdqu16/vmovw/vmovsh for HFmode/HImode scalar or vector.
5431 (ix86_get_excess_precision): Use
5432 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when TARGET_AVX512FP16
5434 (sse_store_index): Use SFmode cost for HFmode cost.
5435 (inline_memory_move_cost): Add HFmode, and perfer SSE cost over
5436 GPR cost for HFmode.
5437 (ix86_hard_regno_mode_ok): Allow HImode in sse register.
5438 (ix86_mangle_type): Add manlging for _Float16 type.
5439 (inline_secondary_memory_needed): No memory is needed for
5440 16bit movement between gpr and sse reg under
5442 (ix86_multiplication_cost): Adjust TARGET_SSE_MATH &&
5443 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
5444 (ix86_division_cost): Ditto.
5445 (ix86_rtx_costs): Ditto.
5446 (ix86_add_stmt_cost): Ditto.
5447 (ix86_optab_supported_p): Ditto.
5448 * config/i386/i386.h (VALID_AVX512F_SCALAR_MODE): Add HFmode.
5449 (SSE_FLOAT_MODE_SSEMATH_OR_HF_P): Add HFmode.
5450 (PTA_SAPPHIRERAPIDS): Add PTA_AVX512FP16.
5451 * config/i386/i386.md (mode): Add HFmode.
5452 (MODE_SIZE): Add HFmode.
5453 (isa): Add avx512fp16.
5454 (enabled): Handle avx512fp16.
5455 (ssemodesuffix): Add sh suffix for HFmode.
5456 (comm): Add mult, div.
5457 (plusminusmultdiv): New code iterator.
5458 (insn): Add mult, div.
5459 (*movhf_internal): Adjust for avx512fp16 instruction.
5460 (*movhi_internal): Ditto.
5461 (*cmpi<unord>hf): New define_insn for HFmode.
5462 (*ieee_s<ieee_maxmin>hf3): Likewise.
5463 (extendhf<mode>2): Likewise.
5464 (trunc<mode>hf2): Likewise.
5465 (float<floatunssuffix><mode>hf2): Likewise.
5466 (*<insn>hf): Likewise.
5467 (cbranchhf4): New expander.
5468 (movhfcc): Likewise.
5469 (<insn>hf3): Likewise.
5472 * config/i386/i386.opt: Add mavx512fp16.
5473 * config/i386/immintrin.h: Include avx512fp16intrin.h.
5474 * doc/invoke.texi: Add mavx512fp16.
5475 * doc/extend.texi: Add avx512fp16 Usage Notes.
5477 2021-09-08 liuhongt <hongtao.liu@intel.com>
5479 * common.opt: Support -fexcess-precision=16.
5480 * config/aarch64/aarch64.c (aarch64_excess_precision): Return
5481 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when
5482 EXCESS_PRECISION_TYPE_FLOAT16.
5483 * config/arm/arm.c (arm_excess_precision): Ditto.
5484 * config/i386/i386.c (ix86_get_excess_precision): Ditto.
5485 * config/m68k/m68k.c (m68k_excess_precision): Issue an error
5486 when EXCESS_PRECISION_TYPE_FLOAT16.
5487 * config/s390/s390.c (s390_excess_precision): Ditto.
5488 * coretypes.h (enum excess_precision_type): Add
5489 EXCESS_PRECISION_TYPE_FLOAT16.
5490 * doc/tm.texi (TARGET_C_EXCESS_PRECISION): Update documents.
5491 * doc/tm.texi.in (TARGET_C_EXCESS_PRECISION): Ditto.
5492 * doc/extend.texi (Half-Precision): Document
5493 -fexcess-precision=16.
5494 * flag-types.h (enum excess_precision): Add
5495 EXCESS_PRECISION_FLOAT16.
5496 * target.def (excess_precision): Update document.
5497 * tree.c (excess_precision_type): Set excess_precision_type to
5498 EXCESS_PRECISION_FLOAT16 when -fexcess-precision=16.
5500 2021-09-08 liuhongt <hongtao.liu@intel.com>
5502 * doc/extend.texi: (@node Floating Types): Adjust the wording.
5503 (@node Half-Precision): Ditto.
5505 2021-09-07 Takayuki 'January June' Suwa <jjsuwa_sys3175@yahoo.co.jp>
5508 * config/xtensa/xtensa.c (xtensa_emit_move_sequence): Add
5509 'CONST_INT_P (src)' to the condition of the block that tries to
5510 eliminate literal when loading integer contant.
5512 2021-09-07 David Faust <david.faust@oracle.com>
5514 * doc/extend.texi (BPF Type Attributes) New node.
5515 Document new preserve_access_index attribute.
5516 Document new preserve_access_index builtin.
5517 * doc/invoke.texi: Document -mco-re and -mno-co-re options.
5519 2021-09-07 David Faust <david.faust@oracle.com>
5521 * config/bpf/bpf.c: Adjust includes.
5522 (bpf_handle_preserve_access_index_attribute): New function.
5523 (bpf_attribute_table): Use it here.
5524 (bpf_builtins): Add BPF_BUILTIN_PRESERVE_ACCESS_INDEX.
5525 (bpf_option_override): Handle "-mco-re" option.
5526 (bpf_asm_init_sections): New.
5527 (TARGET_ASM_INIT_SECTIONS): Redefine.
5528 (bpf_file_end): New.
5529 (TARGET_ASM_FILE_END): Redefine.
5530 (bpf_init_builtins): Add "__builtin_preserve_access_index".
5531 (bpf_core_compute, bpf_core_get_index): New.
5532 (is_attr_preserve_access): New.
5533 (bpf_expand_builtin): Handle new builtins.
5534 (bpf_core_newdecl, bpf_core_is_maybe_aggregate_access): New.
5535 (bpf_core_walk): New.
5536 (bpf_resolve_overloaded_builtin): New.
5537 (TARGET_RESOLVE_OVERLOADED_BUILTIN): Redefine.
5539 (pass_bpf_core_attr): New RTL pass.
5540 * config/bpf/bpf-passes.def: New file.
5541 * config/bpf/bpf-protos.h (make_pass_bpf_core_attr): New.
5542 * config/bpf/coreout.c: New file.
5543 * config/bpf/coreout.h: Likewise.
5544 * config/bpf/t-bpf (TM_H): Add $(srcdir)/config/bpf/coreout.h.
5545 (coreout.o): New rule.
5546 (PASSES_EXTRA): Add $(srcdir)/config/bpf/bpf-passes.def.
5547 * config.gcc (bpf): Add coreout.h to extra_headers.
5548 Add coreout.o to extra_objs.
5549 Add $(srcdir)/config/bpf/coreout.c to target_gtfiles.
5551 2021-09-07 David Faust <david.faust@oracle.com>
5553 * btfout.c (get_btf_id): Function is no longer static.
5554 * ctfc.h: Expose it here.
5556 2021-09-07 David Faust <david.faust@oracle.com>
5558 * ctfc.c (ctf_lookup_tree_type): New function.
5561 2021-09-07 David Faust <david.faust@oracle.com>
5563 * ctfc.c (ctf_dtd_lookup): Function is no longer static.
5564 * ctfc.h: Analogous change.
5566 2021-09-07 David Faust <david.faust@oracle.com>
5568 * dwarf2out.c (lookup_type_die): Function is no longer static.
5569 * dwarf2out.h: Expose it here.
5571 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
5573 * dwarf2ctf.c (ctf_debug_finalize): Make it static.
5574 (ctf_debug_early_finish): New definition.
5575 (ctf_debug_finish): Likewise.
5576 * dwarf2ctf.h (ctf_debug_finalize): Remove declaration.
5577 (ctf_debug_early_finish): New declaration.
5578 (ctf_debug_finish): Likewise.
5579 * dwarf2out.c (dwarf2out_finish): Invoke ctf_debug_finish.
5580 (dwarf2out_early_finish): Invoke ctf_debug_early_finish.
5582 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
5584 * config/bpf/bpf.c (bpf_option_override): For BPF backend, disable LTO
5585 support when compiling for CO-RE.
5586 * config/bpf/bpf.opt: Add new command line option -mco-re.
5588 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
5590 * flag-types.h (enum debug_info_type): Add new enum
5591 DINFO_TYPE_BTF_WITH_CORE.
5592 (BTF_WITH_CORE_DEBUG): New bitmask.
5593 * flags.h (btf_with_core_debuginfo_p): New declaration.
5594 * opts.c (btf_with_core_debuginfo_p): New definition.
5596 2021-09-07 Jason Merrill <jason@redhat.com>
5598 * tree.h (error_operand_p): Change to inline function.
5600 2021-09-07 Aldy Hernandez <aldyh@redhat.com>
5602 * tree-ssa-threadedge.c (forwarder_block_p): Rename to...
5603 (empty_block_with_phis_p): ...this.
5604 (potentially_threadable_block): Same.
5605 (jump_threader::thread_through_normal_block): Same.
5607 2021-09-07 Eric Botcazou <ebotcazou@adacore.com>
5610 * dwarf2out.c (mark_base_types): New overloaded function.
5611 (dwarf2out_early_finish): Invoke it on the COMDAT type list as well
5612 as the compilation unit, and call move_marked_base_types afterward.
5614 2021-09-07 H.J. Lu <hjl.tools@gmail.com>
5617 * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
5619 (ix86_expand_vector_convert_uns_vsivsf): Likewise.
5621 2021-09-07 Richard Biener <rguenther@suse.de>
5623 PR tree-optimization/102226
5624 * tree-vect-loop.c (vect_transform_cycle_phi): Record
5625 the converted value for the epilogue PHI use.
5627 2021-09-07 Martin Liska <mliska@suse.cz>
5629 PR gcov-profile/80223
5630 * ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
5631 options, do not inline when no_profile_instrument_function
5632 attributes are different in early inliner. It's fine to inline
5633 it after PGO instrumentation.
5635 2021-09-07 Richard Biener <rguenther@suse.de>
5637 PR tree-optimization/101555
5638 * tree-ssa-pre.c (translate_vuse_through_block): Do not
5639 perform an alias walk to determine the validity of the
5640 mem at the start of the block which is already guaranteed
5641 by means of prune_clobbered_mems.
5642 (phi_translate_1): Pass edge to translate_vuse_through_block.
5644 2021-09-07 Xionghu Luo <luoxhu@linux.ibm.com>
5647 * config/rs6000/rs6000.md (fmod<mode>3): New define_expand.
5648 (remainder<mode>3): Likewise.
5650 2021-09-07 YunQiang Su <yunqiang.su@cipunited.com>
5652 * config/mips/mips.c (mips_file_start): add .module for
5655 2021-09-06 Roger Sayle <roger@nextmovesoftware.com>
5657 * wide-int.cc (wi::clz): Reorder tests to ensure the result
5658 is zero for all negative values.
5660 2021-09-06 Tobias Burnus <tobias@codesourcery.com>
5662 * doc/invoke.texi (-foffload-options): Fix @opindex.
5664 2021-09-06 H.J. Lu <hjl.tools@gmail.com>
5667 * config/i386/i386-expand.c (ix86_split_xorsign): Use operands[2].
5668 * config/i386/i386.md (@xorsign<mode>3_1): Add non-destructive
5669 source alternative for AVX.
5671 2021-09-06 liuhongt <hongtao.liu@intel.com>
5673 PR middle-end/102182
5674 * optabs.c (expand_fix): Add from1 to avoid from being
5677 2021-09-06 Eric Botcazou <ebotcazou@adacore.com>
5679 * dwarf2out.c (modified_type_die): Deal with all array types earlier
5680 and use local variable consistently throughout the function.
5682 2021-09-06 Jakub Jelinek <jakub@redhat.com>
5684 PR tree-optimization/102207
5685 * match.pd: Don't demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
5686 were promoted from signed to wider unsigned type.
5688 2021-09-06 Andrew Pinski <apinski@marvell.com>
5690 PR tree-optimization/63184
5691 * match.pd: Add simplification of pointer_diff of two pointer_plus
5692 with addr_expr in the first operand of each pointer_plus.
5693 Add simplificatoin of ne/eq of two pointer_plus with addr_expr
5694 in the first operand of each pointer_plus.
5696 2021-09-06 Richard Biener <rguenther@suse.de>
5698 PR tree-optimization/102176
5699 * tree-vect-slp.c (vect_slp_gather_vectorized_scalar_stmts):
5701 (vect_bb_slp_scalar_cost): Use the computed set of
5702 vectorized scalar stmts instead of relying on the out-of-date
5703 and not accurate PURE_SLP_STMT.
5704 (vect_bb_vectorization_profitable_p): Compute the set
5705 of vectorized scalar stmts.
5707 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
5709 * gimple-range-path.cc (path_range_query::range_of_stmt): Remove
5710 GIMPLE_COND special casing.
5711 (path_range_query::range_defined_in_block): Use range_of_stmt
5712 instead of calling fold_range directly.
5714 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
5716 * gimple-range-path.cc (path_range_query::range_of_expr): Set
5717 m_undefined_path when appropriate.
5718 (path_range_query::internal_range_of_expr): Copy from range_of_expr.
5719 (path_range_query::unreachable_path_p): New.
5720 (path_range_query::precompute_ranges): Set m_undefined_path.
5721 * gimple-range-path.h (path_range_query::unreachable_path_p): New.
5722 (path_range_query::internal_range_of_expr): New.
5723 * tree-ssa-threadbackward.c (back_threader::find_taken_edge_cond):
5724 Use unreachable_path_p.
5726 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
5728 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
5729 Remove argument and call find_taken_edge.
5730 (back_threader::resolve_phi): Do not calculate taken edge before
5731 calling maybe_register_path.
5732 (back_threader::find_paths_to_names): Same.
5734 2021-09-05 Jeff Law <jlaw@localhost.localdomain>
5736 * config/h8300/h8300.md (QHSI2 mode iterator): New mode iterator.
5737 * config/h8300/testcompare.md (store_c): Update name, use new
5739 (store_neg_c, store_shifted_c): New patterns.
5741 2021-09-03 Segher Boessenkool <segher@kernel.crashing.org>
5744 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): On ELFv2 use r11
5745 instead of r12 for CR save, in all cases.
5747 2021-09-03 Andrew Pinski <apinski@marvell.com>
5749 * config/aarch64/aarch64-sve-builtins.cc (register_vector_type):
5750 Handle error_mark_node as the type of the type_decl.
5752 2021-09-03 Andrew Pinski <apinski@marvell.com>
5754 * config/aarch64/aarch64-builtins.c (struct aarch64_simd_type_info):
5756 (aarch64_simd_types): Likewise.
5757 (aarch64_simd_intOI_type_node): Likewise.
5758 (aarch64_simd_intCI_type_node): Likewise.
5759 (aarch64_simd_intXI_type_node): Likewise.
5760 * config/aarch64/aarch64.h (aarch64_fp16_type_node): Likewise.
5761 (aarch64_fp16_ptr_type_node): Likewise.
5762 (aarch64_bf16_type_node): Likewise.
5763 (aarch64_bf16_ptr_type_node): Likewise.
5765 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5767 * range-op.cc (operator_minus::op1_op2_relation_effect): Abstract
5769 (minus_op1_op2_relation_effect): ...here.
5770 (class operator_pointer_diff): New.
5771 (operator_pointer_diff::op1_op2_relation_effect): Call
5772 minus_op1_op2_relation_effect.
5773 (integral_table::integral_table): Add entry for POINTER_DIFF_EXPR.
5775 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5777 * tree-ssa-threadbackward.c (back_threader::thread_through_all_blocks):
5778 Add may_peel_loop_headers.
5779 (back_threader_registry::thread_through_all_blocks): Same.
5780 (try_thread_blocks): Pass may_peel_loop_headers argument.
5781 (pass_early_thread_jumps::execute): Same.
5783 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5785 * tree-ssa-threadedge.c (has_phis_p): New.
5786 (forwarder_block_p): New.
5787 (potentially_threadable_block): Call forwarder_block_p.
5788 (jump_threader::thread_around_empty_blocks): Call has_phis_p.
5789 (jump_threader::thread_through_normal_block): Call
5792 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5794 * tree-ssa-threadbackward.c (back_threader::dump): New.
5795 (back_threader::debug): New.
5796 (back_threader_profitability::profitable_path_p): Dump blocks
5797 even if we are bailing early.
5799 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5801 * tree-ssa-threadupdate.c (cancel_thread): New.
5802 (jump_thread_path_registry::thread_block_1): Use cancel_thread.
5803 (jump_thread_path_registry::mark_threaded_blocks): Same.
5804 (jump_thread_path_registry::register_jump_thread): Same.
5806 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5808 * tree-ssa-threadedge.c (jt_state::push): Only call methods for
5809 which objects are available.
5810 (jt_state::pop): Same.
5811 (jt_state::register_equiv): Same.
5812 (jt_state::register_equivs_on_edge): Same.
5814 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5816 * tree-ssa-threadedge.c (jump_threader::thread_across_edge):
5817 Move pop until after a thread is registered.
5819 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5821 * tree-ssa-threadupdate.c (debug): New.
5823 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5825 * gimple-range-trace.cc (push_dump_file::push_dump_file): New.
5826 (push_dump_file::~push_dump_file): New.
5827 (dump_ranger): Change dump_file temporarily while dumping
5829 * gimple-range-trace.h (class push_dump_file): New.
5831 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5833 * gimple-range-trace.cc (debug_seed_ranger): Remove static.
5834 (dump_ranger): Dump function name.
5836 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5838 * gimple-range-path.cc (path_range_query::range_defined_in_block):
5839 Adjust for non-null.
5840 (path_range_query::adjust_for_non_null_uses): New.
5841 (path_range_query::precompute_ranges): Call
5842 adjust_for_non_null_uses.
5843 * gimple-range-path.h: Add m_non_null and
5844 adjust_for_non_null_uses.
5846 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5848 * gimple-range-path.cc (path_range_query::dump): Dump path
5850 (path_range_query::precompute_ranges): Dump entire path.
5852 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5854 * value-relation.cc (relation_oracle::debug): New.
5855 * value-relation.h (relation_oracle::debug): New.
5857 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5859 * tree-ssa-loop-ch.c: Remove unnecessary include file.
5861 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5863 * gimple-range-fold.cc (fold_using_range::postfold_gcond_edges):
5864 Skip statements with no defining BB.
5865 * gimple-range-path.cc (path_range_query::range_defined_in_block):
5866 Do not get confused by statements with no defining BB.
5868 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5870 * gimple-range-fold.cc (adjust_imagpart_expr): Move from
5871 gimple_range_adjustment. Add support for constants.
5872 (adjust_realpart_expr): New.
5873 (gimple_range_adjustment): Move IMAGPART_EXPR code to
5874 adjust_imagpart_expr.
5875 * range-op.cc (integral_table::integral_table): Add entry for
5878 2021-09-03 Jakub Jelinek <jakub@redhat.com>
5880 * omp-expand.c (expand_omp_atomic_pipeline): Use
5881 IFN_ATOMIC_COMPARE_EXCHANGE instead of
5882 BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_? so that memory order
5885 2021-09-03 Jakub Jelinek <jakub@redhat.com>
5888 * tree.h (DECL_FIELD_ABI_IGNORED): Changed into rvalue only macro
5889 that is false if DECL_BIT_FIELD.
5890 (SET_DECL_FIELD_ABI_IGNORED, DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD,
5891 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD): Define.
5892 * tree-streamer-out.c (pack_ts_decl_common_value_fields): For
5893 DECL_BIT_FIELD stream DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead
5894 of DECL_FIELD_ABI_IGNORED.
5895 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
5896 SET_DECL_FIELD_ABI_IGNORED instead of writing to
5897 DECL_FIELD_ABI_IGNORED and for DECL_BIT_FIELD use
5898 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead.
5899 * lto-streamer-out.c (hash_tree): For DECL_BIT_FIELD hash
5900 DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead of DECL_FIELD_ABI_IGNORED.
5902 2021-09-03 liuhongt <hongtao.liu@intel.com>
5905 * config/i386/amxbf16intrin.h : Remove macro check for __AMX_BF16__.
5906 * config/i386/amxint8intrin.h : Remove macro check for __AMX_INT8__.
5907 * config/i386/amxtileintrin.h : Remove macro check for __AMX_TILE__.
5909 2021-09-02 Martin Sebor <msebor@redhat.com>
5911 PR tree-optimization/17506
5913 * tree-ssa-uninit.c (warn_uninit): Remove conditional guarding note.
5915 2021-09-02 Richard Biener <rguenther@suse.de>
5917 * tree-ssa-loop-im.c (fill_always_executed_in_1): Refine
5918 fix for PR78185 and continue processing when leaving
5921 2021-09-02 Jakub Jelinek <jakub@redhat.com>
5923 PR tree-optimization/99591
5924 * match.pd: Demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
5927 2021-09-02 Richard Biener <rguenther@suse.de>
5930 2021-09-02 Richard Biener <rguenther@suse.de>
5932 PR tree-optimization/102155
5933 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
5934 over a part of the RPO array and do not recurse here.
5935 Dump blocks marked as always executed.
5936 (fill_always_executed_in): Walk over the RPO array and
5937 process loops whose header we run into.
5938 (loop_invariant_motion_in_fun): Compute the first RPO
5939 using rev_post_order_and_mark_dfs_back_seme in iteration
5940 order and pass that to fill_always_executed_in.
5942 2021-09-02 liuhongt <hongtao.liu@intel.com>
5944 * config/i386/i386-modes.def (FLOAT_MODE): Define ieee HFmode.
5945 * config/i386/i386.c (enum x86_64_reg_class): Add
5947 (merge_classes): Handle X86_64_SSEHF_CLASS.
5948 (examine_argument): Ditto.
5949 (construct_container): Ditto.
5950 (classify_argument): Ditto, and set HFmode/HCmode to
5952 (function_value_32): Return _FLoat16/Complex Float16 by
5954 (function_value_64): Return _Float16/Complex Float16 by SSE
5956 (ix86_print_operand): Handle CONST_DOUBLE HFmode.
5957 (ix86_secondary_reload): Require gpr as intermediate register
5958 to store _Float16 from sse register when sse4 is not
5960 (ix86_libgcc_floating_mode_supported_p): Enable _FLoat16 under
5962 (ix86_scalar_mode_supported_p): Ditto.
5963 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Defined.
5964 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add HFmode.
5965 (VALID_INT_MODE_P): Add HFmode and HCmode.
5966 * config/i386/i386.md (*pushhf_rex64): New define_insn.
5968 (*movhf_internal): Ditto.
5969 * doc/extend.texi (Half-Precision Floating Point): Documemt
5972 2021-09-02 Richard Biener <rguenther@suse.de>
5974 PR tree-optimization/102155
5975 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
5976 over a part of the RPO array and do not recurse here.
5977 Dump blocks marked as always executed.
5978 (fill_always_executed_in): Walk over the RPO array and
5979 process loops whose header we run into.
5980 (loop_invariant_motion_in_fun): Compute the first RPO
5981 using rev_post_order_and_mark_dfs_back_seme in iteration
5982 order and pass that to fill_always_executed_in.
5984 2021-09-02 YunQiang Su <syq@debian.org>
5987 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
5989 * config/mips/mips.c (mips_module_isa_name): New.
5990 mips_file_start: add .module mipsREV to all asm output
5992 2021-09-01 Jeff Law <jlaw@localhost.localdomain>
5994 PR tree-optimization/102152
5995 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Reduce a vector
5996 comparison to a scalar comparison before calling
5997 update_stmt_if_modified.
5999 2021-09-01 Andrew Pinski <apinski@marvell.com>
6002 * config/aarch64/aarch64.c (aarch64_expand_setmem):
6003 Check STRICT_ALIGNMENT before creating an overlapping
6006 2021-09-01 Martin Sebor <msebor@redhat.com>
6008 * gimple-ssa-warn-access.cc (get_size_range): Add argument.
6009 (check_access): Pass additional argument.
6010 (check_memop_access): Remove template and make a member function.
6011 (maybe_check_dealloc_call): Make a pass_waccess member function.
6012 (class pass_waccess): Add, rename, and remove members.
6013 (pass_waccess::pass_waccess): Adjust to name change.
6014 (pass_waccess::~pass_waccess): Same.
6015 (check_alloca): Make a member function.
6016 (check_alloc_size_call): Same.
6017 (check_strcat): Same.
6018 (check_strncat): Same.
6019 (check_stxcpy): Same.
6020 (check_stxncpy): Same.
6021 (check_strncmp): Same.
6022 (maybe_warn_rdwr_sizes): Rename...
6023 (pass_waccess::maybe_check_access_sizes): ...to this.
6024 (pass_waccess::check_call): Adjust to name changes.
6025 (pass_waccess::maybe_check_dealloc_call): Make a pass_waccess member
6027 (pass_waccess::execute): Adjust to name changes.
6028 * gimple-ssa-warn-access.h (check_memop_access): Remove.
6029 * pointer-query.cc (access_ref::phi): Handle null pointer.
6030 (access_ref::inform_access): Same.
6031 (pointer_query::put_ref): Modify a cached value, not a copy of it.
6032 (pointer_query::dump): New function.
6033 (compute_objsize_r): Avoid overwriting access_ref::bndrng. Cache
6035 * pointer-query.h (pointer_query::dump): Declare.
6036 * tree-ssa-strlen.c (get_range): Simplify. Use function query.
6037 (dump_strlen_info): Use function query.
6038 (printf_strlen_execute): Factor code out into pointer_query::put_ref.
6040 2021-09-01 Thomas Schwinge <thomas@codesourcery.com>
6042 * tree.c (walk_tree_1) <OMP_CLAUSE>: Simplify.
6044 2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
6046 * doc/extend.texi: Document unavailable attribute.
6047 * print-tree.c (print_node): Handle unavailable attribute.
6048 * tree-core.h (struct tree_base): Add a bit to carry unavailability.
6049 * tree.c (error_unavailable_use): New.
6050 * tree.h (TREE_UNAVAILABLE): New.
6051 (error_unavailable_use): New.
6053 2021-09-01 Jakub Jelinek <jakub@redhat.com>
6055 PR tree-optimization/102124
6056 * tree-vect-patterns.c (vect_recog_widen_op_pattern): For ORIG_CODE
6057 MINUS_EXPR, if itype is unsigned with smaller precision than type,
6058 add an extra cast to signed variant of itype to ensure sign-extension.
6060 2021-09-01 Martin Liska <mliska@suse.cz>
6062 * graph.c (draw_cfg_node_succ_edges): Do not color fallthru
6063 edges and rather use colors for TRUE and FALSE edges.
6065 2021-09-01 Richard Biener <rguenther@suse.de>
6067 PR tree-optimization/93491
6068 * tree-ssa-pre.c (compute_avail): Set BB_MAY_NOTRETURN
6069 after processing the stmt itself. Do not consider
6070 pure functions possibly not returning. Properly avoid
6071 adding possibly trapping calls to EXP_GEN when there's
6072 a preceeding possibly not returning call.
6073 * tree-ssa-sccvn.c (vn_reference_may_trap): Conservatively
6076 2021-09-01 Richard Biener <rguenther@suse.de>
6078 PR tree-optimization/102139
6079 * tree-vectorizer.h (vec_base_alignments): Adjust hash-map
6080 type to record a std::pair of the stmt-info and the innermost
6082 (dr_vec_info::group): New member.
6083 * tree-vect-data-refs.c (vect_record_base_alignment): Adjust.
6084 (vect_compute_data_ref_alignment): Verify the recorded
6085 base alignment can be used.
6086 (data_ref_pair): Remove.
6087 (dr_group_sort_cmp): Adjust.
6088 (vect_analyze_data_ref_accesses): Store the group-ID in the
6089 dr_vec_info and operate on a vector of dr_vec_infos.
6091 2021-09-01 YunQiang Su <yunqiang.su@cipunited.com>
6093 * read-md.c (md_reader::handle_enum): support value assignation.
6094 * doc/md.texi: record define_c_enum value assignation support.
6096 2021-09-01 Jakub Jelinek <jakub@redhat.com>
6098 PR tree-optimization/102141
6099 * gimple-ssa-store-merging.c (bswap_view_convert): Add BEFORE
6100 argument. If false, emit stmts after gsi instead of before, and
6102 (bswap_replace): Adjust callers. When converting output of bswap,
6103 emit VIEW_CONVERT prepratation stmts after a copy of gsi instead
6106 2021-09-01 liuhongt <hongtao.liu@intel.com>
6108 * emit-rtl.c (validate_subreg): Get rid of all float-int
6111 2021-09-01 liuhongt <hongtao.liu@intel.com>
6114 2021-08-30 liuhongt <hongtao.liu@intel.com>
6116 * expmed.c (extract_bit_field_1): Make sure we're playing with
6117 integral modes before call extract_integral_bit_field.
6118 (extract_integral_bit_field): Add a parameter of type
6119 scalar_int_mode which corresponds to of tmode.
6120 And call extract_and_convert_fixed_bit_field instead of
6121 extract_fixed_bit_field and convert_extracted_bit_field.
6122 (extract_and_convert_fixed_bit_field): New function, it's a
6123 combination of extract_fixed_bit_field and
6124 convert_extracted_bit_field.
6126 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
6128 * tree.c (walk_tree_1) <OMP_CLAUSE_TILE>: Handle three operands.
6130 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
6132 * omp-general.h (omp_is_reference): Rename to...
6133 (omp_privatize_by_reference): ... this. Adjust all users...
6134 * omp-general.c: ... here, ...
6135 * gimplify.c: ... here, ...
6136 * omp-expand.c: ... here, ...
6137 * omp-low.c: ... here.
6139 2021-08-31 Martin Sebor <msebor@redhat.com>
6141 * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Test
6142 pointer element for equality to zero, not that of the cotaining
6145 2021-08-31 Martin Sebor <msebor@redhat.com>
6147 * gcc-rich-location.h (gcc_rich_location): Make ctor explicit.
6149 2021-08-31 Martin Sebor <msebor@redhat.com>
6151 * function.h (function): Add comments.
6152 (get_range_query): Same. Add attribute returns nonnull.
6154 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
6156 * expr.c (convert_modes): Don't use subreg_promoted_mode on a
6157 SUBREG if it can't be guaranteed to a SUBREG_PROMOTED_VAR_P set.
6158 Instead use the standard (safer) is_a <scalar_int_mode> idiom.
6160 2021-08-31 Jeff Law <jlaw@localhost.localdomain>
6162 * config.gcc (cris-*-elf, cris-*-none): Remove dbxelf.h from
6164 (m32r-*-elf, m32rle-*-elf, m32r-*-linux): Likewise.
6165 (mn10300-*-*, am33_2.0-*-linux*): Likewise.
6166 (xtensa*-*-elf, xtensa*-*-linux, xtensa*-*-uclinux): Likewise.
6167 (m32c-*-elf*, m32c-*-rtems*): Likewise.
6168 * config/cris/cris.h (DBX_NO_XREFS): Remove.
6169 (DBX_CONTIN_LENGTH, DBX_CONTIN_CHAR): Likewise.
6170 * config/m32r/m32r.h (DBXOUT_SOURCE_LINE): Likewise.
6171 (DBX_DEBUGGING_INFO, DBX_CONTIN_LENGTH): Likewise.
6172 * config/mn10300/mn10300.h (DEFAULT_GDB_EXTENSIONS): Likewise.
6173 * config/mn10300/linux.h (DBX_REGISTER_NAMES): Likewise.
6175 2021-08-31 Marcel Vollweiler <marcel@codesourcery.com>
6177 * gimplify.c (gimplify_scan_omp_clauses): Error handling. 'ancestor' only
6178 allowed on target constructs and only with particular other clauses.
6179 * omp-expand.c (expand_omp_target): Output of 'sorry, not supported' if
6181 * omp-low.c (check_omp_nesting_restrictions): Error handling. No nested OpenMP
6182 structs when 'ancestor' is used.
6183 (scan_omp_1_stmt): No usage of OpenMP runtime routines in a target region when
6185 * tree-pretty-print.c (dump_omp_clause): Append 'ancestor'.
6186 * tree.h (OMP_CLAUSE_DEVICE_ANCESTOR): Define macro.
6188 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
6190 * expr.c (convert_modes): Preserve SUBREG_PROMOTED_VAR_P when
6191 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
6193 * simplify-rtx.c (simplify_unary_operation_1) [SIGN_EXTEND]:
6194 Likewise, preserve SUBREG_PROMOTED_VAR_P when creating a (wider)
6195 partial subreg from a SUBREG_PROMOTED_VAR_P subreg. Generate
6196 SIGN_EXTEND of the SUBREG_REG when a subreg would be paradoxical.
6197 [ZERO_EXTEND]: Likewise, preserve SUBREG_PROMOTED_VAR_P when
6198 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
6199 subreg. Generate ZERO_EXTEND of the SUBREG_REG when a subreg
6200 would be paradoxical.
6202 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
6204 * combine.c (combine_simplify_rtx): Avoid converting an explicit
6205 TRUNCATE into a lowpart SUBREG on !TRULY_NOOP_TRUNCATION targets.
6206 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
6208 2021-08-31 Richard Biener <rguenther@suse.de>
6210 PR tree-optimization/102142
6211 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Fix
6212 condition under which to unset the visited flag.
6214 2021-08-31 Richard Biener <rguenther@suse.de>
6216 PR middle-end/102129
6217 * tree-ssa-ter.c (find_replaceable_in_bb): Do not move
6218 possibly trapping expressions across calls.
6220 2021-08-31 Jakub Jelinek <jakub@redhat.com>
6222 PR tree-optimization/102134
6223 * tree-ssa-ccp.c (bit_value_binop) <case RSHIFT_EXPR>: If sgn is
6224 UNSIGNED and r1val | r1mask has MSB set, ensure lzcount doesn't
6227 2021-08-31 Andrew Pinski <apinski@marvell.com>
6230 * collect-utils.c (setup_signals): New declaration.
6231 * collect-utils.h (setup_signals): New function.
6232 * collect2.c (handler): Delete.
6233 (main): Instead of manually setting up the signals,
6234 just call setup_signals.
6235 * lto-wrapper.c (main): Likewise.
6237 2021-08-31 Andrew Pinski <apinski@marvell.com>
6240 * config/i386/i386-protos.h (x86_output_aligned_bss):
6241 Change align argument to unsigned type.
6242 (x86_elf_aligned_decl_common): Likewise.
6243 * config/i386/i386.c (x86_elf_aligned_decl_common): Likewise.
6244 (x86_output_aligned_bss): Likewise.
6246 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
6248 * config/mips/mips.c (mips_module_isa_name): New.
6249 mips_file_start: add .module mipsREV to all asm output
6251 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
6253 * config/mips/mips.h (struct mips_cpu_info): define enum mips_isa;
6254 use enum instead of int for 'isa' member.
6255 * config.gcc, config/mips/mips.c, config/mips/mips-cpus.def,
6256 config/mips/netbsd.h: replace hardcoded numbers with enum.
6258 2021-08-31 liuhongt <hongtao.liu@intel.com>
6260 * config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
6261 define_split to define_insn_and_split.
6262 (*avx2_eq<mode>3): Removed.
6263 (<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
6264 (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
6265 (*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
6267 (*avx2_gt<mode>3): Removed.
6268 (<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
6269 define_insn to define_expand, and adjust pattern.
6270 (UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.
6272 2021-08-30 David Malcolm <dmalcolm@redhat.com>
6275 * Makefile.in (ANALYZER_OBJS): Add analyzer/call-info.o.
6277 2021-08-30 Jason Merrill <jason@redhat.com>
6279 * doc/invoke.texi: Document -Wmissing-requires.
6281 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
6283 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Remove
6284 TARGET_EXTRA_BUILTINS guard.
6286 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
6288 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Change
6289 initialization of V2DI_type_node and unsigned_V2DI_type_node.
6291 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
6293 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Use the new
6294 decl when new_builtins_are_live.
6295 * config/rs6000/rs6000-builtin-new.def (__builtin_cfstring): New
6298 2021-08-30 Pat Haugen <pthaugen@linux.ibm.com>
6300 * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add
6301 OPTION_MASK_P10_FUSION_2STORE.
6302 (POWERPC_MASKS): Likewise.
6303 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
6304 store fusion for Power10.
6305 (is_fusable_store): New.
6306 (power10_sched_reorder): Likewise.
6307 (rs6000_sched_reorder): Do Power10 specific reordering.
6308 (rs6000_sched_reorder2): Likewise.
6309 * config/rs6000/rs6000.opt: Add new option.
6311 2021-08-30 Richard Biener <rguenther@suse.de>
6313 PR tree-optimization/102128
6314 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
6315 Move scanning for if-converted scalar code to the caller
6316 and instead delay clearing the visited flag for profitable
6318 (vect_slp_region): Cost all subgraphs before scheduling.
6319 For if-converted BB vectorization scan for scalar COND_EXPRs
6320 and do not vectorize if any found and the cost model is
6323 2021-08-30 Richard Biener <rguenther@suse.de>
6325 * common.opt (fexceptions): Mark
6326 EnabledBy(fnon-call-exceptions).
6327 * doc/invoke.texi (fnon-call-exceptions): Document this
6328 enables -fexceptions.
6330 2021-08-30 Sebastian Huber <sebastian.huber@embedded-brains.de>
6332 * tsystem.h (abort): Define abort() if inhibit_libc is defined and it
6333 is not already defined.
6335 2021-08-30 liuhongt <hongtao.liu@intel.com>
6337 * expmed.c (extract_bit_field_1): Make sure we're playing with
6338 integral modes before call extract_integral_bit_field.
6339 (extract_integral_bit_field): Add a parameter of type
6340 scalar_int_mode which corresponds to of tmode.
6341 And call extract_and_convert_fixed_bit_field instead of
6342 extract_fixed_bit_field and convert_extracted_bit_field.
6343 (extract_and_convert_fixed_bit_field): New function, it's a
6344 combination of extract_fixed_bit_field and
6345 convert_extracted_bit_field.
6347 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
6349 * config/darwin.c (darwin_libc_has_function): Do not run
6350 the checks for x86 or modern Darwin. Make sure that there
6351 is a value set for darwin_macosx_version_min before testing.
6353 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
6355 * config/i386/darwin.h (CLEAR_INSN_CACHE): New.
6357 2021-08-28 Jan Hubicka <hubicka@ucw.cz>
6359 * ipa-modref-tree.h (modref_access_node::merge): Break out
6360 logic combining offsets and logic merging ranges to ...
6361 (modref_access_node::combined_offsets): ... here
6362 (modref_access_node::update2): ... here
6363 (modref_access_node::closer_pair_p): New member function.
6364 (modref_access_node::forced_merge): New member function.
6365 (modre_ref_node::insert): Do merging when table is full.
6367 2021-08-28 YunQiang Su <yunqiang.su@cipunited.com>
6370 * config.gcc: MIPS: use N64 ABI by default if the triple end
6371 with -gnuabi64, which is used by Debian since 2013.
6373 2021-08-28 Alexandre Oliva <oliva@adacore.com>
6375 * ipa-modref.c (analyze_function): Skip debug stmts.
6376 * tree-inline.c (estimate_num_insn): Consider builtins even
6377 without a cgraph_node.
6379 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
6381 * config/h8300/bitfield.md (cstore<mode>4): Remove expander.
6382 * config/h8300/h8300.c (h8300_expand_branch): Remove function.
6383 * config/h8300/h8300-protos.h (h8300_expadn_branch): Remove prototype.
6384 * config/h8300/h8300.md (eqne): New code iterator.
6385 (geultu, geultu_to_c): Similarly.
6386 * config/h8300/testcompare.md (cstore<mode>4): Dummy expander.
6387 (store_c_<mode>, store_c_i_<mode>): New define_insn_and_splits
6388 (cmp<mode>_c): New pattern
6390 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
6392 * tree-ssa-dom.c (reduce_vector_comparison_to_scalar_comparison): New
6394 (dom_opt_dom_walker::optimize_stmt): Use it.
6396 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
6398 * config/darwin.c (finalize_ctors): Add a section-start linker-
6400 (finalize_dtors): Likewise.
6401 * config/darwin.h (MIN_LD64_INIT_TERM_START_LABELS): New.
6403 2021-08-27 Bill Schmidt <wschmidt@linux.ibm.com>
6405 * config/rs6000/rs6000-call.c (rs6000-builtins.h): New #include.
6406 (rs6000_init_builtins): Call rs6000_init_generated_builtins. Skip the
6407 old initialization logic when new builtins are enabled.
6408 * config/rs6000/rs6000-gen-builtins.c (write_decls): Rename
6409 rs6000_autoinit_builtins to rs6000_init_generated_builtins.
6410 (write_init_file): Likewise.
6412 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
6414 * configure.ac (darwin2[[0-9]]* | darwin19*): Alter use of
6415 gcc_GAS_CHECK_FEATURE to remove an extraneous parameter.
6416 (amdgcn-* | gcn-*) Likewise.
6418 2021-08-27 Anthony Sharp <anthonysharp15@gmail.com>
6420 * symbol-summary.h: Added missing template keyword.
6422 2021-08-27 Richard Biener <rguenther@suse.de>
6424 PR tree-optimization/45178
6425 * tree-ssa-dce.c (find_obviously_necessary_stmts): For
6426 infinite loops without exit do not mark control dependent
6427 edges of the latch necessary.
6429 2021-08-27 konglin1 <lingling.kong@intel.com>
6432 * config/i386/sse.md: (<avx512>scattersi<mode>): Add mask operand to
6434 (<avx512>scattersi<mode>): Likewise.
6435 (*avx512f_scattersi<VI48F:mode>): Merge mask operand to set_dest.
6436 (*avx512f_scatterdi<VI48F:mode>): Likewise
6438 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
6440 * config/rs6000/rs6000.c (rs6000_builtin_md_vectorized_function): Add
6441 support for built-in functions MISC_BUILTIN_DIVWE, MISC_BUILTIN_DIVWEU,
6442 MISC_BUILTIN_DIVDE, MISC_BUILTIN_DIVDEU, P10_BUILTIN_CFUGED,
6443 P10_BUILTIN_CNTLZDM, P10_BUILTIN_CNTTZDM, P10_BUILTIN_PDEPD and
6444 P10_BUILTIN_PEXTD on Power10.
6446 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
6448 * config/rs6000/rs6000-call.c (builtin_function_type): Add unsigned
6449 signedness for some Power10 bifs.
6451 2021-08-27 David Edelsohn <dje.gcc@gmail.com>
6454 * config/rs6000/rs6000.c (rs6000_adjust_field_align): Use
6455 computed alignment if the entire struct has attribute packed.
6457 2021-08-27 liuhongt <hongtao.liu@intel.com>
6461 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
6462 IX86_BUILTIN_SHUFPD512, IX86_BUILTIN_SHUFPS512,
6463 IX86_BUILTIN_SHUFPD256, IX86_BUILTIN_SHUFPS,
6464 IX86_BUILTIN_SHUFPS256.
6465 (ix86_masked_all_ones): New function.
6467 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
6469 * config/i386/i386.md (*btr<mode>_1): Call force_reg unconditionally.
6470 (conditional moves with memory inputs splitters): Ditto.
6471 * config/i386/sse.md (one_cmpl<mode>2): Simplify.
6473 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
6475 * ipa-modref-tree.h (modref_access_node::try_merge_with): Restart
6476 search after merging.
6478 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
6480 * config/rs6000/rs6000-overload.def: Add remaining overloads.
6482 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
6484 * config/rs6000/rs6000-builtin-new.def: Add cell stanza.
6486 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
6488 * config/rs6000/rs6000-builtin-new.def: Add ieee128-hw, dfp,
6489 crypto, and htm stanzas.
6491 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
6493 * config/rs6000/rs6000-builtin-new.def: Add mma stanza.
6495 2021-08-26 Martin Sebor <msebor@redhat.com>
6497 * tree-ssa-uninit.c (warn_uninit): Refactor and simplify.
6498 (warn_uninit_phi_uses): Remove argument from calls to warn_uninit.
6499 (warn_uninitialized_vars): Same. Reduce visibility of locals.
6500 (warn_uninitialized_phi): Same.
6502 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
6504 * tree-ssa-ccp.c (get_individual_bits): Helper function to
6505 extract the individual bits from a widest_int constant (mask).
6506 (gray_code_bit_flips): New read-only table for effiently
6507 enumerating permutations/combinations of bits.
6508 (bit_value_binop) [LROTATE_EXPR, RROTATE_EXPR]: Handle rotates
6509 by unknown counts that are guaranteed less than the target
6510 precision and four or fewer unknown bits by enumeration.
6511 [LSHIFT_EXPR, RSHIFT_EXPR]: Likewise, also handle shifts by
6512 enumeration under the same conditions. Handle remaining
6513 shifts as a mask based upon the minimum possible shift value.
6515 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
6516 Richard Biener <rguenther@suse.de>
6518 * match.pd (shift transformations): Remove a redundant
6519 !POINTER_TYPE_P check.
6521 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
6524 * config/i386/i386.md (cmove reg-reg move elimination peephole2s):
6525 Set all_regs to true in the call to replace_rtx.
6527 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
6529 * ipa-modref-tree.c (test_insert_search_collapse): Update test.
6530 * ipa-modref-tree.h (modref_base_node::insert): Be smarter when
6531 hiting --param modref-max-refs limit.
6532 (modref_tree:insert_base): Be smarter when hitting
6533 --param modref-max-bases limit. Add new parameter REF.
6534 (modref_tree:insert): Update.
6535 (modref_tree:merge): Update.
6536 * ipa-modref.c (read_modref_records): Update.
6538 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
6540 * params.opt: (modref-max-adjustments): Add full stop.
6542 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
6544 * ipa-modref-tree.h (modref_ref_node::verify): New member
6546 (modref_ref_node::insert): Use it.
6547 (modref_ref_node::try_mere_with): Fix off by one error.
6549 2021-08-26 Martin Liska <mliska@suse.cz>
6550 Stefan Kneifel <stefan.kneifel@bluewin.ch>
6552 * cgraph.h (create_version_clone_with_body): Add new parameter.
6553 * cgraphclones.c: Likewise.
6554 * multiple_target.c (create_dispatcher_calls): Do not use
6556 (create_target_clone): Likewise here.
6558 2021-08-26 Jonathan Yong <10walls@gmail.com>
6560 * doc/extend.texi: Add note about reserved priorities
6561 to the constructor attribute.
6563 2021-08-25 Martin Sebor <msebor@redhat.com>
6565 * gimple-range-cache.cc (ssa_global_cache::dump): Avoid printing
6566 range table header alone.
6567 * gimple-range.cc (gimple_ranger::export_global_ranges): Same.
6569 2021-08-25 Jan Hubicka <hubicka@ucw.cz>
6571 * doc/invoke.texi: Document --param modref-max-adjustments.
6572 * ipa-modref-tree.c (test_insert_search_collapse): Update.
6573 (test_merge): Update.
6574 * ipa-modref-tree.h (struct modref_access_node): Add adjustments;
6575 (modref_access_node::operator==): Fix handling of access ranges.
6576 (modref_access_node::contains): Constify parameter; handle also
6577 mismatched parm offsets.
6578 (modref_access_node::update): New function.
6579 (modref_access_node::merge): New function.
6580 (unspecified_modref_access_node): Update constructor.
6581 (modref_ref_node::insert_access): Add record_adjustments parameter;
6583 (modref_ref_node::try_merge_with): New private function.
6584 (modref_tree::insert): New record_adjustments parameter.
6585 (modref_tree::merge): New record_adjustments parameter.
6586 (modref_tree::copy_from): Update.
6587 * ipa-modref.c (dump_access): Dump adjustments field.
6588 (get_access): Update constructor.
6589 (record_access): Update call of insert.
6590 (record_access_lto): Update call of insert.
6591 (merge_call_side_effects): Add record_adjustments parameter.
6592 (get_access_for_fnspec): Update.
6593 (process_fnspec): Update.
6594 (analyze_call): Update.
6595 (analyze_function): Update.
6596 (read_modref_records): Update.
6597 (ipa_merge_modref_summary_after_inlining): Update.
6598 (propagate_unknown_call): Update.
6599 (modref_propagate_in_scc): Update.
6600 * params.opt (param-max-modref-adjustments=): New.
6602 2021-08-25 Michael Meissner <meissner@linux.ibm.com>
6604 * config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
6606 (xxspltiw_v4si): Use vecperm type attribute.
6607 (xxspltiw_v4si_inst): Use vecperm type attribute.
6608 (xxspltiw_v4sf_inst): Likewise.
6609 (xxspltidp_v2df): Use vecperm type attribute. Use
6610 UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
6611 (xxspltidp_v2df_inst): Likewise.
6612 (xxsplti32dx_v4si): Use vecperm type attribute.
6613 (xxsplti32dx_v4si_inst): Likewise.
6614 (xxsplti32dx_v4sf_inst): Likewise.
6615 (xxblend_<mode>): Likewise.
6616 (xxpermx): Likewise.
6617 (xxpermx_inst): Likewise.
6620 2021-08-25 Lewis Hyatt <lhyatt@gmail.com>
6623 * coretypes.h (typedef diagnostic_input_charset_callback): Declare.
6624 * diagnostic.c (diagnostic_initialize_input_context): New function.
6625 * diagnostic.h (diagnostic_initialize_input_context): Declare.
6626 * input.c (default_charset_callback): New function.
6627 (file_cache::initialize_input_context): New function.
6628 (file_cache_slot::create): Added ability to convert the input
6629 according to the input context.
6630 (file_cache::file_cache): Initialize the new input context.
6631 (class file_cache_slot): Added new m_alloc_offset member.
6632 (file_cache_slot::file_cache_slot): Initialize the new member.
6633 (file_cache_slot::~file_cache_slot): Handle potentially offset buffer.
6634 (file_cache_slot::maybe_grow): Likewise.
6635 (file_cache_slot::needs_read_p): Handle NULL fp, which is now possible.
6636 (file_cache_slot::get_next_line): Likewise.
6637 * input.h (class file_cache): Added input context member.
6639 2021-08-25 Richard Biener <rguenther@suse.de>
6641 PR tree-optimization/102046
6642 * tree-vect-slp.c (vect_build_slp_tree_2): Conservatively
6643 update ->any_pattern when swapping operands.
6645 2021-08-25 Hongyu Wang <hongyu.wang@intel.com>
6648 * config/i386/i386.c (ix86_live_on_entry): Adjust comment.
6649 (ix86_decompose_address): Remove retval check for ASHIFT,
6650 allow non-canonical zero extend if AND mask covers ASHIFT
6652 (ix86_legitimate_address_p): Adjust condition for decompose.
6653 (ix86_rtx_costs): Adjust cost for lea with non-canonical
6655 Co-Authored by: Uros Bizjak <ubizjak@gmail.com>
6657 2021-08-25 Jiufu Guo <guojiufu@linux.ibm.com>
6659 PR tree-optimization/101145
6660 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
6662 (number_of_iterations_lt): Invoke above function.
6663 (adjust_cond_for_loop_until_wrap):
6664 Merge to number_of_iterations_until_wrap.
6665 (number_of_iterations_cond): Update invokes for
6666 adjust_cond_for_loop_until_wrap and number_of_iterations_lt.
6668 2021-08-25 konglin1 <lingling.kong@intel.com>
6671 * config/i386/avx512dqintrin.h (_mm512_fpclass_ps_mask): Fix
6673 (_mm512_mask_fpclass_ps_mask): Ditto.
6675 2021-08-25 Kewen Lin <linkw@linux.ibm.com>
6677 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Remove.
6678 (vec_unpacku_hi_v8hi): Likewise.
6679 (vec_unpacku_lo_v16qi): Likewise.
6680 (vec_unpacku_lo_v8hi): Likewise.
6681 (vec_unpacku_hi_<VP_small_lc>): New define_expand.
6682 (vec_unpacku_lo_<VP_small_lc>): Likewise.
6684 2021-08-24 David Edelsohn <dje.gcc@gmail.com>
6686 * config/rs6000/aix.h (SYSTEM_IMPLICIT_EXTERN_C): Delete.
6687 * config/rs6000/aix71.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
6688 * config/rs6000/aix72.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
6689 * config/rs6000/aix73.h (TARGET_AIX_VERSION): Increase to 73.
6691 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
6693 PR middle-end/102031
6694 * simplify-rtx.c (simplify_truncation): When comparing precisions
6695 use "subreg_prec" variable, not "subreg_mode".
6697 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
6699 * config/rs6000/rs6000-builtin-new.def: Add power10 and power10-64
6702 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
6704 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Initialize
6705 various pointer type nodes.
6706 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add enum
6707 values for various pointer types.
6708 (ptr_V16QI_type_node): New macro.
6709 (ptr_V1TI_type_node): New macro.
6710 (ptr_V2DI_type_node): New macro.
6711 (ptr_V2DF_type_node): New macro.
6712 (ptr_V4SI_type_node): New macro.
6713 (ptr_V4SF_type_node): New macro.
6714 (ptr_V8HI_type_node): New macro.
6715 (ptr_unsigned_V16QI_type_node): New macro.
6716 (ptr_unsigned_V1TI_type_node): New macro.
6717 (ptr_unsigned_V8HI_type_node): New macro.
6718 (ptr_unsigned_V4SI_type_node): New macro.
6719 (ptr_unsigned_V2DI_type_node): New macro.
6720 (ptr_bool_V16QI_type_node): New macro.
6721 (ptr_bool_V8HI_type_node): New macro.
6722 (ptr_bool_V4SI_type_node): New macro.
6723 (ptr_bool_V2DI_type_node): New macro.
6724 (ptr_bool_V1TI_type_node): New macro.
6725 (ptr_pixel_type_node): New macro.
6726 (ptr_intQI_type_node): New macro.
6727 (ptr_uintQI_type_node): New macro.
6728 (ptr_intHI_type_node): New macro.
6729 (ptr_uintHI_type_node): New macro.
6730 (ptr_intSI_type_node): New macro.
6731 (ptr_uintSI_type_node): New macro.
6732 (ptr_intDI_type_node): New macro.
6733 (ptr_uintDI_type_node): New macro.
6734 (ptr_intTI_type_node): New macro.
6735 (ptr_uintTI_type_node): New macro.
6736 (ptr_long_integer_type_node): New macro.
6737 (ptr_long_unsigned_type_node): New macro.
6738 (ptr_float_type_node): New macro.
6739 (ptr_double_type_node): New macro.
6740 (ptr_long_double_type_node): New macro.
6741 (ptr_dfloat64_type_node): New macro.
6742 (ptr_dfloat128_type_node): New macro.
6743 (ptr_ieee128_type_node): New macro.
6744 (ptr_ibm128_type_node): New macro.
6745 (ptr_vector_pair_type_node): New macro.
6746 (ptr_vector_quad_type_node): New macro.
6747 (ptr_long_long_integer_type_node): New macro.
6748 (ptr_long_long_unsigned_type_node): New macro.
6750 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
6752 * config/rs6000/rs6000-builtin-new.def: Add power9-vector, power9,
6753 and power9-64 stanzas.
6755 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
6756 Tom de Vries <tdevries@suse.de>
6758 * config.gcc (nvptx-*-*): Define {c,c++}_target_objs.
6759 * config/nvptx/nvptx-protos.h (nvptx_cpu_cpp_builtins): Prototype.
6760 * config/nvptx/nvptx.h (TARGET_CPU_CPP_BUILTINS): Implement with
6761 a call to the new nvptx_cpu_cpp_builtins function in nvptx-c.c.
6762 * config/nvptx/t-nvptx (nvptx-c.o): New rule.
6763 * config/nvptx/nvptx-c.c: New source file.
6764 (nvptx_cpu_cpp_builtins): Move implementation here.
6766 2021-08-24 Martin Sebor <msebor@redhat.com>
6768 PR middle-end/101600
6769 PR middle-end/101977
6770 * gimple-ssa-warn-access.cc (maybe_warn_for_bound): Tighten up
6771 the phrasing of a warning.
6772 (check_access): Use the remaining size after subtracting any offset
6773 rather than the whole object size.
6774 * pointer-query.cc (access_ref::get_ref): Clear BASE0 flag if it's
6775 clear for any nonnull PHI argument.
6776 (compute_objsize): Clear argument.
6778 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
6780 * config/rs6000/rs6000-builtin-new.def: Add power8-vector stanza.
6782 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
6784 * config/rs6000/rs6000-builtin-new.def: Add power7 and power7-64
6787 2021-08-24 Andrew MacLeod <amacleod@redhat.com>
6789 * value-relation.cc (rr_transitive_table): New.
6790 (relation_transitive): New.
6791 (value_relation::swap): Remove.
6792 (value_relation::apply_transitive): New.
6793 (relation_oracle::relation_oracle): Allocate a new tmp bitmap.
6794 (relation_oracle::register_relation): Call register_transitives.
6795 (relation_oracle::register_transitives): New.
6796 * value-relation.h (relation_oracle): Add new temporary bitmap and
6799 2021-08-24 H.J. Lu <hjl.tools@gmail.com>
6802 * config/i386/i386-expand.c (ix86_expand_vector_move): Broadcast
6803 from integer to a pseudo vector register.
6805 2021-08-24 Richard Biener <rguenther@suse.de>
6807 PR tree-optimization/100089
6808 * tree-vectorizer.h (vect_slp_bb): Rename to ...
6809 (vect_slp_if_converted_bb): ... this and get the original
6810 loop as new argument.
6811 * tree-vectorizer.c (try_vectorize_loop_1): Revert previous fix,
6812 pass original loop to vect_slp_if_converted_bb.
6813 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
6814 If orig_loop was passed scan the not vectorized stmts
6815 for COND_EXPRs and force not profitable if found.
6816 (vect_slp_region): Pass down all SLP instances to costing
6817 if orig_loop was specified.
6818 (vect_slp_bbs): Pass through orig_loop.
6819 (vect_slp_bb): Rename to ...
6820 (vect_slp_if_converted_bb): ... this and get the original
6821 loop as new argument.
6822 (vect_slp_function): Adjust.
6824 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
6827 * config/arm/arm.md (attribute arch): Add fix_vlldm.
6828 (arch_enabled): Use it.
6829 * config/arm/vfp.md (lazy_store_multiple_insn): Add alternative to
6830 use when erratum mitigation is needed.
6832 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
6835 * config/arm/arm.opt (mfix-cmse-cve-2021-35465): New option.
6836 * doc/invoke.texi (Arm Options): Document it.
6837 * config/arm/arm-cpus.in (quirk_vlldm): New feature bit.
6838 (ALL_QUIRKS): Add quirk_vlldm.
6839 (cortex-m33): Add quirk_vlldm.
6840 (cortex-m35p, cortex-m55): Likewise.
6841 * config/arm/arm.c (arm_option_override): Enable fix_vlldm if
6842 targetting an affected CPU and not explicitly controlled on
6845 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
6847 * config/arm/vfp.md (lazy_store_multiple_insn): Rewrite as valid RTL.
6848 (lazy_load_multiple_insn): Likewise.
6850 2021-08-24 liuhongt <hongtao.liu@intel.com>
6853 * config/i386/sse.md (<avx512>_vternlog<mode><sd_maskz_name>):
6854 Enable avx512 embedded broadcast.
6855 (*<avx512>_vternlog<mode>_all): Ditto.
6856 (<avx512>_vternlog<mode>_mask): Ditto.
6858 2021-08-24 liuhongt <hongtao.liu@intel.com>
6861 * config/i386/i386.c (ix86_rtx_costs): Define cost for
6863 * config/i386/i386.h (STRIP_UNARY): New macro.
6864 * config/i386/predicates.md (reg_or_notreg_operand): New
6866 * config/i386/sse.md (*<avx512>_vternlog<mode>_all): New define_insn.
6867 (*<avx512>_vternlog<mode>_1): New pre_reload
6868 define_insn_and_split.
6869 (*<avx512>_vternlog<mode>_2): Ditto.
6870 (*<avx512>_vternlog<mode>_3): Ditto.
6871 (any_logic1,any_logic2): New code iterator.
6872 (logic_op): New code attribute.
6873 (ternlogsuffix): Extend to VNxDF and VNxSF.
6875 2021-08-24 Richard Biener <rguenther@suse.de>
6877 * doc/invoke.texi (vect-inner-loop-cost-factor): Adjust.
6878 * params.opt (--param vect-inner-loop-cost-factor): Adjust
6880 * tree-vect-loop.c (vect_analyze_loop_form): Initialize
6881 inner_loop_cost_factor to the minimum of the estimated number
6882 of iterations of the inner loop and vect-inner-loop-cost-factor.
6884 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
6885 Richard Biener <rguenther@suse.de>
6887 * config/i386/i386-features.c (compute_convert_gain): Provide
6888 more accurate values for CONST_INT, when optimizing for size.
6889 * config/i386/i386.c (COSTS_N_BYTES): Move definition from here...
6890 * config/i386/i386.h (COSTS_N_BYTES): to here.
6892 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
6893 Jakub Jelinek <jakub@redhat.com>
6895 PR middle-end/102029
6896 * match.pd (shift transformations): Add an additional check for
6897 !POINTER_TYPE_P in the recently added left shift transformation.
6899 2021-08-24 liuhongt <hongtao.liu@intel.com>
6901 PR tree-optimization/100089
6902 * tree-vectorizer.c (try_vectorize_loop_1): Disable slp in
6903 loop vectorizer when cost model is very-cheap.
6905 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
6907 * config/rs6000/rs6000-gen-builtins.c (parse_bif_entry): Don't call
6908 asprintf, which is not available on AIX.
6910 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
6912 * config.gcc (target_gtfiles): Add ./rs6000-builtins.h.
6913 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Set.
6915 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
6917 * config.gcc (powerpc*-*-*): Add rs6000-builtins.o to extra_objs.
6918 * config/rs6000/rs6000-gen-builtins.c (main): Close init_file
6920 * config/rs6000/t-rs6000 (rs6000-gen-builtins.o): New target.
6921 (rbtree.o): Likewise.
6922 (rs6000-gen-builtins): Likewise.
6923 (rs6000-builtins.c): Likewise.
6924 (rs6000-builtins.h): Likewise.
6925 (rs6000.o): Add dependency.
6926 (EXTRA_HEADERS): Add rs6000-vecdefines.h.
6927 (rs6000-vecdefines.h): New target.
6928 (rs6000-builtins.o): Likewise.
6929 (rs6000-call.o): Add rs6000-builtins.h as a dependency.
6930 (rs6000-c.o): Likewise.
6932 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
6935 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace):
6936 Diagnose buffer overrun.
6937 (safe_inc_pos): Fix overrun detection.
6938 (match_identifier): Diagnose buffer overrun.
6939 (match_integer): Likewise.
6940 (match_to_right_bracket): Likewise.
6942 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
6944 * ipa-modref-tree.h (modref_access_node::range_info_useful_p):
6945 Improve range compare.
6946 (modref_access_node::contains): New member function.
6947 (modref_access_node::search): Remove.
6948 (modref_access_node::insert): Be smarter about subaccesses.
6950 2021-08-23 Thomas Schwinge <thomas@codesourcery.com>
6952 * config/i386/i386-options.c (ix86_omp_device_kind_arch_isa)
6953 <omp_device_arch> [ACCEL_COMPILER]: Match "intel_mic".
6954 * config/i386/t-omp-device (omp-device-properties-i386) <arch>:
6957 2021-08-23 Jeff Law <jlaw@localhost.localdomain>
6959 * config/h8300/h8300-protos.h (h8300_expand_epilogue): Add new
6961 * config/h8300/jumpcall.md (call, call_value): Restrict to
6962 !SIBLING_CALL_P cases.
6963 (subcall, sibcall_value): New patterns & expanders.
6964 * config/h8300/proepi.md (epilogue): Pass new argument to
6965 h8300_expand_epilogue.
6966 (sibcall_epilogue): New expander.
6967 * config/h8300/h8300.c (h8300_expand_epilogue): Handle sibcall
6969 (h8300_ok_for_sibcall_p): New function.
6970 (TARGET_FUNCTION_OK_FOR_SIBCALL): define.
6972 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
6974 * simplify-rtx.c (simplify_unary_operation_1): [TRUNCATE]:
6975 Handle case where the operand is already the desired mode.
6977 2021-08-23 Richard Biener <rguenther@suse.de>
6980 * tree-ssa-structalias.c (ipa_pta_execute): Check in_other_partition
6981 in addition to has_gimple_body.
6983 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
6985 PR middle-end/101949
6986 * ipa-modref.c (analyze_ssa_name_flags): Fix merging of
6989 2021-08-23 Martin Liska <mliska@suse.cz>
6991 * doc/invoke.texi: Put the option out of -mxl-mode-app-model
6994 2021-08-23 Richard Biener <rguenther@suse.de>
6996 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6997 Properly scale the inner loop cost only once.
6999 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
7001 * tree-ssa-ccp.c (bit_value_binop) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]:
7002 Provide bounds for unsigned (and signed with non-negative operands)
7003 division and modulus.
7005 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
7007 * simplify-rtx.c (simplify_truncation): Generalize simplification
7008 of (truncate:A (subreg:B X)).
7009 (simplify_unary_operation_1) [FLOAT_TRUNCATE, FLOAT_EXTEND,
7010 SIGN_EXTEND, ZERO_EXTEND]: Handle cases where the operand
7011 already has the desired machine mode.
7012 (test_scalar_int_ops): Add tests that useless extensions and
7013 truncations are optimized away.
7014 (test_scalar_int_ext_ops): New self-test function to confirm
7015 that truncations of extensions are correctly simplified.
7016 (test_scalar_int_ext_ops2): New self-test function to check
7017 truncations of truncations, extensions of extensions, and
7018 truncations of extensions.
7019 (test_scalar_ops): Call the above two functions with a
7020 representative sampling of integer machine modes.
7022 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
7024 * match.pd (shift transformations): Change the sign of an
7025 LSHIFT_EXPR if it reduces the number of explicit conversions.
7027 2021-08-23 Jakub Jelinek <jakub@redhat.com>
7029 PR tree-optimization/86723
7030 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
7031 cast64_to_32 argument, set *cast64_to_32 to false, unless n is
7032 non-memory permutation of 64-bit src which only has bytes of
7033 0 or [5..8] and n->range is 4.
7034 (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
7035 find_bswap_or_nop_finalize caller, support bswap with some bytes
7036 zeroed, as long as at least two bytes are not zeroed.
7037 (bswap_replace): Add mask argument and handle masking of bswap
7039 (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
7040 caller, punt if cast64_to_32 or mask is not all ones.
7041 (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
7042 caller, for now punt if cast64_to_32.
7044 2021-08-23 Richard Biener <rguenther@suse.de>
7046 PR tree-optimization/79334
7047 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Record
7048 a type also for COMPONENT_REFs.
7049 (vn_reference_may_trap): Check ARRAY_REF with constant index
7050 against the array domain.
7052 2021-08-23 liuhongt <hongtao.liu@intel.com>
7055 * config/i386/sse.md (*avx512f_pshufb_truncv8hiv8qi_1): Add
7056 TARGET_AVX512BW to condition.
7058 2021-08-23 Jakub Jelinek <jakub@redhat.com>
7061 * dwarf2out.c (gen_variable_die): Add DW_AT_location for global
7062 register variables already during early_dwarf if possible.
7064 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
7066 * config/arm/arm_mve.h: Fix __arm_vctp16q return type.
7068 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
7071 * config/arm/arm.opt: Fix typo.
7072 * config/arm/t-rmprofile: Fix typo.
7074 2021-08-23 Jakub Jelinek <jakub@redhat.com>
7076 * tree.h (OMP_CLAUSE_GRAINSIZE_STRICT): Define.
7077 (OMP_CLAUSE_NUM_TASKS_STRICT): Define.
7078 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_GRAINSIZE,
7079 case OMP_CLAUSE_NUM_TASKS>: Print strict: modifier.
7080 * omp-expand.c (expand_task_call): Use GOMP_TASK_FLAG_STRICT in iflags
7081 if either grainsize or num_tasks clause has the strict modifier.
7083 2021-08-23 Martin Liska <mliska@suse.cz>
7085 * dbgcnt.def (DEBUG_COUNTER): New counter.
7086 * gimple.c (gimple_call_arg_flags): Use it in IPA PTA.
7088 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
7090 * ipa-modref.c (analyze_ssa_name_flags): Improve handling of return slot.
7092 2021-08-23 Xi Ruoyao <xry111@mengyan1223.wang>
7095 * config/mips/mips-protos.h (mips_msa_output_shift_immediate):
7097 * config/mips/mips.c (mips_msa_output_shift_immediate): New
7099 * config/mips/mips-msa.md (vashl<mode>3, vashr<mode>3,
7100 vlshr<mode>3): Call it.
7102 2021-08-22 Jan Hubicka <hubicka@ucw.cz>
7103 Martin Liska <mliska@suse.cz>
7105 PR middle-end/101949
7106 * ipa-modref.c (analyze_ssa_name_flags): Indirect call implies
7109 2021-08-21 Dragan Mladjenovic <OT_Dragan.Mladjenovic@mediatek.com>
7111 * config/mips/mips.c (mips_function_rodata_section,
7112 TARGET_ASM_FUNCTION_RODATA_SECTION): Removed.
7114 2021-08-21 John David Anglin <danglin@gcc.gnu.org>
7116 * config/pa/pa.c (pa_asm_output_aligned_common): Remove warning.
7118 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
7120 * configure.ac (thread-local storage support): Remove tls_first_major
7121 and tls_first_minor. Use "$conftest_s" to check support.
7122 * configure: Regenerate.
7124 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
7126 * configure.ac: Fixup formatting.
7128 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
7130 * acinclude.m4 (gcc_GAS_CHECK_FEATURE): Remove third argument and ...
7131 * configure.ac: ... update all callers.
7133 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
7136 * acinclude.m4 (_gcc_COMPUTE_GAS_VERSION, _gcc_GAS_VERSION_GTE_IFELSE)
7137 (gcc_GAS_VERSION_GTE_IFELSE): Remove.
7138 (gcc_GAS_CHECK_FEATURE): Do not handle in-tree case specially.
7139 * configure.ac: Remove gcc_cv_gas_major_version, gcc_cv_gas_minor_version.
7140 Remove remaining checks for in-tree assembler.
7141 * configure: Regenerate.
7143 2021-08-20 Jeff Law <jlaw@localhost.localdomain>
7145 * config/h8300/h8300.c (shift_alg_hi): Improve arithmetic shift right
7146 by 15 bits for H8/300H and H8/S. Improve logical shifts by 12
7148 (shift_alg_si): Improve arithmetic right shift by 28-30 bits for
7149 H8/300H. Improve arithmetic shift right by 15 bits for H8/S.
7150 Improve logical shifts by 27 bits for H8/S.
7151 (get_shift_alg): Corresponding changes.
7152 (h8300_option_override): Revert to loops for -Os when profitable.
7154 2021-08-20 Richard Biener <rguenther@suse.de>
7156 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
7158 (vect_analyze_data_ref_accesses): Likewise. Assign the BB
7159 index as group_id when dataref_groups were not computed.
7160 * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
7161 we advace to the next BB.
7163 2021-08-20 Jakub Jelinek <jakub@redhat.com>
7165 * omp-builtins.def (BUILT_IN_GOMP_WARNING, BUILT_IN_GOMP_ERROR): New
7168 2021-08-20 Martin Liska <mliska@suse.cz>
7170 PR gcov-profile/89961
7171 * gcov.c (make_gcov_file_name): Rewrite using std::string.
7172 (mangle_name): Simplify, do not used the second argument.
7173 (strip_extention): New function.
7174 (get_md5sum): Likewise.
7175 (get_gcov_intermediate_filename): Handle properly -p and -x
7177 (output_gcov_file): Use string type.
7178 (generate_results): Likewise.
7179 (md5sum_to_hex): Remove.
7181 2021-08-20 Michael Meissner <meissner@linux.ibm.com>
7183 * config/rs6000/altivec.md (UNSPEC_XXEVAL): Move to vsx.md.
7184 (UNSPEC_XXSPLTIW): Move to vsx.md.
7185 (UNSPEC_XXSPLTID): Move to vsx.md.
7186 (UNSPEC_XXSPLTI32DX): Move to vsx.md.
7187 (UNSPEC_XXBLEND): Move to vsx.md.
7188 (UNSPEC_XXPERMX): Move to vsx.md.
7189 (VM3): Move to vsx.md.
7190 (VM3_char): Move to vsx.md.
7191 (xxspltiw_v4si): Move to vsx.md.
7192 (xxspltiw_v4sf): Move to vsx.md.
7193 (xxspltiw_v4sf_inst): Move to vsx.md.
7194 (xxspltidp_v2df): Move to vsx.md.
7195 (xxspltidp_v2df_inst): Move to vsx.md.
7196 (xxsplti32dx_v4si_inst): Move to vsx.md.
7197 (xxsplti32dx_v4sf): Move to vsx.md.
7198 (xxsplti32dx_v4sf_inst): Move to vsx.md.
7199 (xxblend_<mode>): Move to vsx.md.
7200 (xxpermx): Move to vsx.md.
7201 (xxpermx_inst): Move to vsx.md.
7202 * config/rs6000/vsx.md (UNSPEC_XXEVAL): Move from altivec.md.
7203 (UNSPEC_XXSPLTIW): Move from altivec.md.
7204 (UNSPEC_XXSPLTID): Move from altivec.md.
7205 (UNSPEC_XXSPLTI32DX): Move from altivec.md.
7206 (UNSPEC_XXBLEND): Move from altivec.md.
7207 (UNSPEC_XXPERMX): Move from altivec.md.
7208 (VM3): Move from altivec.md.
7209 (VM3_char): Move from altivec.md.
7210 (xxspltiw_v4si): Move from altivec.md.
7211 (xxspltiw_v4sf): Move from altivec.md.
7212 (xxspltiw_v4sf_inst): Move from altivec.md.
7213 (xxspltidp_v2df): Move from altivec.md.
7214 (xxspltidp_v2df_inst): Move from altivec.md.
7215 (xxsplti32dx_v4si_inst): Move from altivec.md.
7216 (xxsplti32dx_v4sf): Move from altivec.md.
7217 (xxsplti32dx_v4sf_inst): Move from altivec.md.
7218 (xxblend_<mode>): Move from altivec.md.
7219 (xxpermx): Move from altivec.md.
7220 (xxpermx_inst): Move from altivec.md.
7222 2021-08-19 Roger Sayle <roger@nextmovesoftware.com>
7224 * tree-vect-generic.c (expand_vector_operations_1): Use either
7225 gimplify_build1 or gimplify_build2 instead of gimple_build_assign
7226 when constructing scalar splat expressions.
7228 2021-08-19 Peter Bergner <bergner@linux.ibm.com>
7231 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Cast
7232 pointer to __vector_pair *.
7234 2021-08-19 Martin Sebor <msebor@redhat.com>
7236 * gimple-range.cc: Add comments.
7237 * gimple-range.h: Same.
7239 2021-08-19 Martin Sebor <msebor@redhat.com>
7241 PR middle-end/101984
7242 * gimple-ssa-warn-access.cc (pass_waccess::execute): Also call
7245 2021-08-19 Jeff Law <jlaw@localhost.localdomain>
7247 * config.gcc (h8300-*-elf*): Do not include dbxelf.h.
7248 (h8300-*-linux*, v850-*-rtems*, v850*-elf*): Likewise.
7249 * config/v850/v850.h (DEFAULT_GDB_EXTENSIONS): Remove.
7251 2021-08-19 Jakub Jelinek <jakub@redhat.com>
7253 PR middle-end/101950
7254 * optabs.c (expand_clrsb_using_clz): New function.
7255 (expand_unop): Use it as another clrsb expansion fallback.
7257 2021-08-19 liuhongt <hongtao.liu@intel.com>
7260 2021-07-28 liuhongt <hongtao.liu@intel.com>
7263 * config/i386/i386.h (processor_costs): Add new member
7265 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
7266 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
7267 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
7268 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
7269 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
7270 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
7271 generic_cost, core_cost): Initialize integer_to_sse same value
7273 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
7274 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7275 Use integer_to_sse instead of sse_op to calculate the cost of
7278 2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
7280 * config.gcc: Include rpath.opt for Darwin.
7281 * config/darwin.h (DRIVER_SELF_SPECS): Handle -rpath.
7283 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
7286 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
7289 2021-08-18 Jonathan Wright <jonathan.wright@arm.com>
7291 * config/aarch64/arm_neon.h (vld3_lane_f64): Use float RTL
7292 pattern and type cast.
7293 (vld4_lane_f32): Use float RTL pattern.
7294 (vld4q_lane_f64): Use float type cast.
7296 2021-08-18 Jan Hubicka <hubicka@ucw.cz>
7298 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Check also
7301 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
7303 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor): Extend.
7304 (test_map_of_type_with_ctor_and_dtor_expand): Add function.
7305 (hash_map_tests_c_tests): Call it.
7307 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
7309 * ggc.h (enum ggc_collect): New.
7310 (ggc_collect): Use it.
7311 * ggc-page.c: Adjust.
7312 * ggc-common.c: Likewise.
7313 * ggc-tests.c: Likewise.
7314 * read-rtl-function.c: Likewise.
7315 * selftest-run-tests.c: Likewise.
7316 * doc/gty.texi (Invoking the garbage collector): Likewise.
7318 2021-08-18 liuhongt <hongtao.liu@intel.com>
7321 * config/i386/i386.h (TARGET_V2DF_REDUCTION_PREFER_HADDPD):
7323 * config/i386/sse.md (*sse3_haddv2df3_low): Add
7324 TARGET_V2DF_REDUCTION_PREFER_HADDPD.
7325 (*sse3_hsubv2df3_low): Ditto.
7326 * config/i386/x86-tune.def
7327 (X86_TUNE_V2DF_REDUCTION_PREFER_HADDPD): New tune.
7329 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
7331 * gimple-range-gori.cc (gori_compute::gori_compute): Enable tracing.
7332 (gori_compute::compute_operand_range): Add tracing.
7333 (gori_compute::logical_combine): Ditto.
7334 (gori_compute::compute_logical_operands): Ditto.
7335 (gori_compute::compute_operand1_range): Ditto.
7336 (gori_compute::compute_operand2_range): Ditto.
7337 (gori_compute::outgoing_edge_range_p): Ditto.
7338 * gimple-range-gori.h (class gori_compute): Add range_tracer.
7340 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
7342 * flag-types.h (enum evrp_mode): Adjust evrp-mode values.
7343 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Relocate from.
7344 * gimple-range-trace.h (DEBUG_RANGE_CACHE): Here.
7345 * params.opt (--param=evrp-mode): Adjust options.
7347 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
7349 * Makefile.in (OBJS): Add gimple-range-trace.o.
7350 * gimple-range-cache.h (enable_new_values): Remove unused prototype.
7351 * gimple-range-fold.cc: Adjust headers.
7352 * gimple-range-trace.cc: New.
7353 * gimple-range-trace.h: New.
7354 * gimple-range.cc (gimple_ranger::gimple_ranger): Enable tracer.
7355 (gimple_ranger::range_of_expr): Add tracing.
7356 (gimple_ranger::range_on_entry): Ditto.
7357 (gimple_ranger::range_on_exit): Ditto.
7358 (gimple_ranger::range_on_edge): Ditto.
7359 (gimple_ranger::fold_range_internal): Ditto.
7360 (gimple_ranger::dump_bb): Do not calculate edge range twice.
7361 (trace_ranger::*): Remove.
7362 (enable_ranger): Never create a trace_ranger.
7363 (debug_seed_ranger): Move to gimple-range-trace.cc.
7364 (dump_ranger): Ditto.
7365 (debug_ranger): Ditto.
7366 * gimple-range.h: Include gimple-range-trace.h.
7367 (range_on_entry, range_on_exit): No longer virtual.
7368 (class trace_ranger): Remove.
7369 (DEBUG_RANGE_CACHE): Move to gimple-range-trace.h.
7371 2021-08-17 Martin Sebor <msebor@redhat.com>
7373 PR middle-end/101854
7374 * builtins.c (expand_builtin_alloca): Move warning code to check_alloca
7375 in gimple-ssa-warn-access.cc.
7376 * calls.c (alloc_max_size): Move code to check_alloca.
7377 (get_size_range): Move to pointer-query.cc.
7378 (maybe_warn_alloc_args_overflow): Move to gimple-ssa-warn-access.cc.
7379 (get_attr_nonstring_decl): Move to tree.c.
7380 (fntype_argno_type): Move to gimple-ssa-warn-access.cc.
7381 (append_attrname): Same.
7382 (maybe_warn_rdwr_sizes): Same.
7383 (initialize_argument_information): Move code to
7384 gimple-ssa-warn-access.cc.
7385 * calls.h (maybe_warn_alloc_args_overflow): Move to
7386 gimple-ssa-warn-access.h.
7387 (get_attr_nonstring_decl): Move to tree.h.
7388 (maybe_warn_nonstring_arg): Move to gimple-ssa-warn-access.h.
7389 (enum size_range_flags): Move to pointer-query.h.
7390 (get_size_range): Same.
7391 * gimple-ssa-warn-access.cc (has_location): Remove unused overload
7392 to avoid Clang -Wunused-function.
7393 (get_size_range): Declare static.
7394 (maybe_emit_free_warning): Rename...
7395 (maybe_check_dealloc_call): ...to this for consistency.
7396 (class pass_waccess): Add members.
7397 (pass_waccess::~pass_waccess): Defined.
7398 (alloc_max_size): Move here from calls.c.
7399 (maybe_warn_alloc_args_overflow): Same.
7400 (check_alloca): New function.
7401 (check_alloc_size_call): New function.
7402 (check_strncat): Handle another warning flag.
7403 (pass_waccess::check_builtin): Handle alloca.
7404 (fntype_argno_type): Move here from calls.c.
7405 (append_attrname): Same.
7406 (maybe_warn_rdwr_sizes): Same.
7407 (pass_waccess::check_call): Define.
7408 (check_nonstring_args): New function.
7409 (pass_waccess::check): Call new member functions.
7410 (pass_waccess::execute): Enable ranger.
7411 * gimple-ssa-warn-access.h (get_size_range): Move here from calls.h.
7412 (maybe_warn_nonstring_arg): Same.
7413 * gimple-ssa-warn-restrict.c: Remove #include.
7414 * pointer-query.cc (get_size_range): Move here from calls.c.
7415 * pointer-query.h (enum size_range_flags): Same.
7416 (get_size_range): Same.
7417 * tree.c (get_attr_nonstring_decl): Move here from calls.c.
7418 * tree.h (get_attr_nonstring_decl): Move here from calls.h.
7420 2021-08-17 Thomas Schwinge <thomas@codesourcery.com>
7422 * ggc.h (ggc_collect): Add 'force_collect' parameter.
7423 * ggc-page.c (ggc_collect): Use that one instead of global
7424 'ggc_force_collect'. Adjust all users.
7425 * doc/gty.texi (Invoking the garbage collector): Update.
7426 * ggc-internal.h (ggc_force_collect): Remove.
7427 * ggc-common.c (ggc_force_collect): Likewise.
7428 * selftest.h (forcibly_ggc_collect): Remove.
7429 * ggc-tests.c (selftest::forcibly_ggc_collect): Likewise.
7430 * read-rtl-function.c (test_loading_labels): Adjust.
7431 * selftest-run-tests.c (run_tests): Likewise.
7433 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
7435 * config/darwin.c (darwin_file_end): Reset and reclaim the
7436 section names table at the end of compile.
7438 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
7441 * config.in: Regenerate.
7442 * config/i386/darwin.h (EXTRA_ASM_OPTS): New
7443 (ASM_SPEC): Pass options to disable branch shortening where
7445 * configure: Regenerate.
7446 * configure.ac: Detect versions of 'as' that support the
7447 optimisation which has the bug.
7449 2021-08-17 Richard Biener <rguenther@suse.de>
7451 * optabs-query.c (supports_vec_gather_load_p): Also check
7453 (supports_vec_scatter_store_p): Likewise.
7454 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Fall
7455 back to masked variants if non-masked are not supported.
7456 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
7457 When we need to use masked gather/scatter but do not have
7458 a mask set up a constant true one.
7459 * tree-vect-stmts.c (vect_check_scalar_mask): Also allow
7462 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
7464 * tree-ssa-ccp.c (bit_value_binop) [MINUS_EXPR]: Use same
7465 algorithm as PLUS_EXPR to improve subtraction bit bounds.
7466 [POINTER_DIFF_EXPR]: Treat as synonymous with MINUS_EXPR.
7468 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
7470 * tree-ssa-ccp.c (bit_value_mult_const): New helper function to
7471 calculate the mask-value pair result of a multiplication by an
7473 (bit_value_binop) [MULT_EXPR]: Call it from here for
7474 multiplications by (sparse) non-negative constants.
7476 2021-08-17 Christophe Lyon <christophe.lyon@foss.st.com>
7479 * config.gcc (gcc_cv_initfini_array): Leave undefined for
7480 uclinuxfdpiceabi targets.
7482 2021-08-17 Alexandre Oliva <oliva@adacore.com>
7484 * tree-inline.c (maybe_move_debug_stmts_to_successors): Don't
7485 reverse debug stmts.
7487 2021-08-17 Alexandre Oliva <oliva@adacore.com>
7489 * tree-cfg.c (dump_function_to_file): Use fun, not cfun.
7491 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
7493 * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete.
7494 (__LD4Q_LANE_FUNC): Likewise.
7495 (vld4_lane_u8): Define without macro.
7496 (vld4_lane_u16): Likewise.
7497 (vld4_lane_u32): Likewise.
7498 (vld4_lane_u64): Likewise.
7499 (vld4_lane_s8): Likewise.
7500 (vld4_lane_s16): Likewise.
7501 (vld4_lane_s32): Likewise.
7502 (vld4_lane_s64): Likewise.
7503 (vld4_lane_f16): Likewise.
7504 (vld4_lane_f32): Likewise.
7505 (vld4_lane_f64): Likewise.
7506 (vld4_lane_p8): Likewise.
7507 (vld4_lane_p16): Likewise.
7508 (vld4_lane_p64): Likewise.
7509 (vld4q_lane_u8): Likewise.
7510 (vld4q_lane_u16): Likewise.
7511 (vld4q_lane_u32): Likewise.
7512 (vld4q_lane_u64): Likewise.
7513 (vld4q_lane_s8): Likewise.
7514 (vld4q_lane_s16): Likewise.
7515 (vld4q_lane_s32): Likewise.
7516 (vld4q_lane_s64): Likewise.
7517 (vld4q_lane_f16): Likewise.
7518 (vld4q_lane_f32): Likewise.
7519 (vld4q_lane_f64): Likewise.
7520 (vld4q_lane_p8): Likewise.
7521 (vld4q_lane_p16): Likewise.
7522 (vld4q_lane_p64): Likewise.
7523 (vld4_lane_bf16): Likewise.
7524 (vld4q_lane_bf16): Likewise.
7526 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
7528 * config/aarch64/arm_neon.h (__LD3_LANE_FUNC): Delete.
7529 (__LD3Q_LANE_FUNC): Delete.
7530 (vld3_lane_u8): Define without macro.
7531 (vld3_lane_u16): Likewise.
7532 (vld3_lane_u32): Likewise.
7533 (vld3_lane_u64): Likewise.
7534 (vld3_lane_s8): Likewise.
7535 (vld3_lane_s16): Likewise.
7536 (vld3_lane_s32): Likewise.
7537 (vld3_lane_s64): Likewise.
7538 (vld3_lane_f16): Likewise.
7539 (vld3_lane_f32): Likewise.
7540 (vld3_lane_f64): Likewise.
7541 (vld3_lane_p8): Likewise.
7542 (vld3_lane_p16): Likewise.
7543 (vld3_lane_p64): Likewise.
7544 (vld3q_lane_u8): Likewise.
7545 (vld3q_lane_u16): Likewise.
7546 (vld3q_lane_u32): Likewise.
7547 (vld3q_lane_u64): Likewise.
7548 (vld3q_lane_s8): Likewise.
7549 (vld3q_lane_s16): Likewise.
7550 (vld3q_lane_s32): Likewise.
7551 (vld3q_lane_s64): Likewise.
7552 (vld3q_lane_f16): Likewise.
7553 (vld3q_lane_f32): Likewise.
7554 (vld3q_lane_f64): Likewise.
7555 (vld3q_lane_p8): Likewise.
7556 (vld3q_lane_p16): Likewise.
7557 (vld3q_lane_p64): Likewise.
7558 (vld3_lane_bf16): Likewise.
7559 (vld3q_lane_bf16): Likewise.
7561 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
7563 * config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Delete.
7564 (__LD2Q_LANE_FUNC): Likewise.
7565 (vld2_lane_u8): Define without macro.
7566 (vld2_lane_u16): Likewise.
7567 (vld2_lane_u32): Likewise.
7568 (vld2_lane_u64): Likewise.
7569 (vld2_lane_s8): Likewise.
7570 (vld2_lane_s16): Likewise.
7571 (vld2_lane_s32): Likewise.
7572 (vld2_lane_s64): Likewise.
7573 (vld2_lane_f16): Likewise.
7574 (vld2_lane_f32): Likewise.
7575 (vld2_lane_f64): Likewise.
7576 (vld2_lane_p8): Likewise.
7577 (vld2_lane_p16): Likewise.
7578 (vld2_lane_p64): Likewise.
7579 (vld2q_lane_u8): Likewise.
7580 (vld2q_lane_u16): Likewise.
7581 (vld2q_lane_u32): Likewise.
7582 (vld2q_lane_u64): Likewise.
7583 (vld2q_lane_s8): Likewise.
7584 (vld2q_lane_s16): Likewise.
7585 (vld2q_lane_s32): Likewise.
7586 (vld2q_lane_s64): Likewise.
7587 (vld2q_lane_f16): Likewise.
7588 (vld2q_lane_f32): Likewise.
7589 (vld2q_lane_f64): Likewise.
7590 (vld2q_lane_p8): Likewise.
7591 (vld2q_lane_p16): Likewise.
7592 (vld2q_lane_p64): Likewise.
7593 (vld2_lane_bf16): Likewise.
7594 (vld2q_lane_bf16): Likewise.
7596 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
7598 * haifa-sched.c (advance_one_cycle): Output more context-synchronization
7601 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
7603 * haifa-sched.c (enum rfs_decision, rfs_str): Add RFS_AUTOPREF.
7604 (rank_for_schedule): Use it.
7606 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
7608 PR rtl-optimization/91598
7609 * haifa-sched.c (autopref_rank_for_schedule): Prioritize "irrelevant"
7610 insns after memory reads and before memory writes.
7612 2021-08-17 Alistair_Lee <alistair.lee@arm.com>
7614 * rtl.h (CONST_VECTOR_P): New macro.
7615 * config/aarch64/aarch64.c (aarch64_get_sve_pred_bits): Use RTL
7616 code testing macros.
7617 (aarch64_ptrue_all_mode): Likewise.
7618 (aarch64_expand_mov_immediate): Likewise.
7619 (aarch64_const_vec_all_in_range_p): Likewise.
7620 (aarch64_rtx_costs): Likewise.
7621 (aarch64_legitimate_constant_p): Likewise.
7622 (aarch64_simd_valid_immediate): Likewise.
7623 (aarch64_simd_make_constant): Likewise.
7624 (aarch64_convert_mult_to_shift): Likewise.
7625 (aarch64_expand_sve_vec_perm): Likewise.
7626 (aarch64_vec_fpconst_pow_of_2): Likewise.
7628 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
7630 PR tree-optimization/101938
7631 * range-op.cc (operator_abs::op1_range): Special case
7632 -TYPE_MIN_VALUE for flag_wrapv.
7634 2021-08-17 Kewen Lin <linkw@linux.ibm.com>
7636 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Add the cost for
7639 2021-08-17 Jakub Jelinek <jakub@redhat.com>
7641 * tree.def (OMP_SCOPE): New tree code.
7642 * tree.h (OMP_SCOPE_BODY, OMP_SCOPE_CLAUSES): Define.
7643 * tree-nested.c (convert_nonlocal_reference_stmt,
7644 convert_local_reference_stmt, convert_gimple_call): Handle
7646 * tree-pretty-print.c (dump_generic_node): Handle OMP_SCOPE.
7647 * gimple.def (GIMPLE_OMP_SCOPE): New gimple code.
7648 * gimple.c (gimple_build_omp_scope): New function.
7649 (gimple_copy): Handle GIMPLE_OMP_SCOPE.
7650 * gimple.h (gimple_build_omp_scope): Declare.
7651 (gimple_has_substatements): Handle GIMPLE_OMP_SCOPE.
7652 (gimple_omp_scope_clauses, gimple_omp_scope_clauses_ptr,
7653 gimple_omp_scope_set_clauses): New inline functions.
7654 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_SCOPE.
7655 * gimple-pretty-print.c (dump_gimple_omp_scope): New function.
7656 (pp_gimple_stmt_1): Handle GIMPLE_OMP_SCOPE.
7657 * gimple-walk.c (walk_gimple_stmt): Likewise.
7658 * gimple-low.c (lower_stmt): Likewise.
7659 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
7660 (gimplify_scan_omp_clauses): For task reductions, handle OMP_SCOPE
7661 like ORT_WORKSHARE constructs. Adjust diagnostics for %<scope%>
7662 allowing task reductions. Reject inscan reductions on scope.
7663 (omp_find_stores_stmt): Handle GIMPLE_OMP_SCOPE.
7664 (gimplify_omp_workshare, gimplify_expr): Handle OMP_SCOPE.
7665 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_SCOPE.
7666 (estimate_num_insns): Likewise.
7667 * omp-low.c (build_outer_var_ref): Look through GIMPLE_OMP_SCOPE
7668 contexts if var isn't privatized there.
7669 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_SCOPE.
7670 (scan_omp_1_stmt): Likewise.
7671 (maybe_add_implicit_barrier_cancel): Look through outer
7673 (lower_omp_scope): New function.
7674 (lower_omp_task_reductions): Handle OMP_SCOPE.
7675 (lower_omp_1): Handle GIMPLE_OMP_SCOPE.
7676 (diagnose_sb_1, diagnose_sb_2): Likewise.
7677 * omp-expand.c (expand_omp_single): Support also GIMPLE_OMP_SCOPE.
7678 (expand_omp): Handle GIMPLE_OMP_SCOPE.
7679 (omp_make_gimple_edges): Likewise.
7680 * omp-builtins.def (BUILT_IN_GOMP_SCOPE_START): New built-in.
7682 2021-08-17 Richard Biener <rguenther@suse.de>
7684 PR tree-optimization/101925
7685 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Set
7686 reverse on COMPONENT_REF and ARRAY_REF according to
7687 what reverse_storage_order_for_component_p does.
7688 (vn_reference_eq): Compare reversed on reference ops.
7689 (reverse_storage_order_for_component_p): New overload.
7690 (vn_reference_lookup_3): Check reverse_storage_order_for_component_p
7691 on the reference looked up.
7693 2021-08-17 Jeff Law <jlaw@localhost.localdomain>
7695 * config/h8300/h8300.c (shift_alg_si): Avoid loops for most SImode
7697 (h8300_option_override): Use loops on H8/S more often when optimizing
7699 (get_shift_alg): Handle new "special" cases on H8/S. Simplify
7700 accordingly. Handle various arithmetic right shifts with special
7701 sequences that we couldn't handle before.
7703 2021-08-16 Jeff Law <jlaw@localhost.localdomain>
7705 * config.gcc (rl78-*-elf*): Do not include dbxelf.h.
7707 2021-08-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
7709 * config/sparc/rtemself.h (SPARC_GCOV_TYPE_SIZE): Define.
7710 * config/sparc/sparc.c (sparc_gcov_type_size): New.
7711 (TARGET_GCOV_TYPE_SIZE): Redefine if SPARC_GCOV_TYPE_SIZE is defined.
7712 * coverage.c (get_gcov_type): Use targetm.gcov_type_size().
7713 * doc/tm.texi (TARGET_GCOV_TYPE_SIZE): Add hook under "Misc".
7714 * doc/tm.texi.in: Regenerate.
7715 * target.def (gcov_type_size): New target hook.
7716 * targhooks.c (default_gcov_type_size): New.
7717 * targhooks.h (default_gcov_type_size): Declare.
7718 * tree-profile.c (gimple_gen_edge_profiler): Use precision of
7720 (gimple_gen_time_profiler): Likewise.
7722 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
7724 * dwarf2out.c (add_scalar_info): Deal with DW_AT_data_bit_offset.
7726 2021-08-16 Tobias Burnus <tobias@codesourcery.com>
7728 PR middle-end/101931
7729 * omp-low.c (omp_runtime_api_call): Update for routines
7730 added in the meanwhile.
7732 2021-08-16 Martin Liska <mliska@suse.cz>
7734 PR tree-optimization/100393
7735 * tree-switch-conversion.c (group_cluster::dump): Use
7736 get_comparison_count.
7737 (jump_table_cluster::find_jump_tables): Pre-compute number of
7738 comparisons and then decrement it. Cache also max_ratio.
7739 (jump_table_cluster::can_be_handled): Change signature.
7740 * tree-switch-conversion.h (get_comparison_count): New.
7742 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
7744 * dwarf2out.c (add_data_member_location_attribute): Use GNAT
7745 encodings only when -fgnat-encodings=all is specified.
7746 (add_bound_info): Likewise.
7747 (add_byte_size_attribute): Likewise.
7748 (gen_member_die): Likewise.
7750 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
7752 * omp-oacc-neuter-broadcast.cc
7753 (execute_omp_oacc_neuter_broadcast): Plug 'par' memory leak.
7755 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
7757 * omp-oacc-neuter-broadcast.cc
7758 (execute_omp_oacc_neuter_broadcast): Clarify memory management for
7761 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
7763 * omp-oacc-neuter-broadcast.cc (field_map): Move variable into...
7764 (execute_omp_oacc_neuter_broadcast): ... here.
7765 (install_var_field, build_receiver_ref, build_sender_ref): Take
7766 'field_map_t *' parameter. Adjust all users.
7767 (worker_single_copy, neuter_worker_single): Take a
7768 'record_field_map_t *' parameter. Adjust all users.
7770 2021-08-16 liuhongt <hongtao.liu@intel.com>
7773 * config/i386/i386.md (ldexp<mode>3): Force operands[1] to
7776 2021-08-16 Martin Liska <mliska@suse.cz>
7779 * multiple_target.c (create_dispatcher_calls): Make default
7780 function local only if it is a definition.
7782 2021-08-16 Martin Liska <mliska@suse.cz>
7785 * ipa-icf-gimple.c (func_checker::compare_ssa_name): Do not
7786 consider equal SSA_NAMEs when one is a param.
7788 2021-08-16 liuhongt <hongtao.liu@intel.com>
7791 * config/i386/i386-expand.c (ix86_expand_vec_perm_vpermt2):
7792 Support vpermi2b for V32QI/V16QImode.
7793 (ix86_extract_perm_from_pool_constant): New function.
7794 (ix86_expand_vec_one_operand_perm_avx512): Support
7795 vpermw/vpermb under TARGET_AVX512BW/TARGET_AVX512VBMI.
7796 (expand_vec_perm_1): Adjust comments for upper.
7797 * config/i386/i386-protos.h (ix86_extract_perm_from_pool_constant):
7799 * config/i386/predicates.md (permvar_truncate_operand): New predicate.
7800 (pshufb_truncv4siv4hi_operand): Ditto.
7801 (pshufb_truncv8hiv8qi_operand): Ditto.
7802 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1):
7803 New pre_reload define_insn_and_split.
7804 (*avx512f_permvar_truncv8siv8hi_1): Ditto.
7805 (*avx512f_vpermvar_truncv8div8si_1): Ditto.
7806 (*avx512f_permvar_truncv32hiv32qi_1): Ditto.
7807 (*avx512f_permvar_truncv16hiv16qi_1): Ditto.
7808 (*avx512f_permvar_truncv4div4si_1): Ditto.
7809 (*avx512f_pshufb_truncv8hiv8qi_1): Ditto.
7810 (*avx512f_pshufb_truncv4siv4hi_1): Ditto.
7811 (*avx512f_pshufd_truncv2div2si_1): Ditto.
7813 2021-08-16 Kito Cheng <kito.cheng@sifive.com>
7815 * config/riscv/multilib-generator: Support code model option for
7817 * doc/install.texi: Add document of new option for
7818 --with-multilib-generator.
7820 2021-08-15 Clément Chigot <clement.chigot@atos.net>
7822 * config/rs6000/rs6000.c (xcoff_tls_exec_model_detected): New.
7823 (rs6000_legitimize_tls_address_aix): Use it.
7824 (rs6000_xcoff_file_end): Add ".ref __tls_get_addr" when
7825 xcoff_tls_exec_model_detected is true.
7827 2021-08-15 Jeff Law <jlaw@localhost.localdomain>
7829 * config/h8300/h8300.c (shift_alg_si): Retune H8/300H shifts
7830 to allow a bit more code growth, saving many dozens of cycles.
7831 (h8300_option_override): Adjus shift_alg_si if optimizing for
7833 (get_shift_alg): Use special + inline shifts for residuals
7836 2021-08-14 Stafford Horne <shorne@gmail.com>
7839 * config/or1k/or1k-opts.h: New file.
7840 * config/or1k/or1k.c (or1k_legitimize_address_1, print_reloc):
7841 Support generating gotha relocations if -mcmodel=large is
7843 * config/or1k/or1k.h (TARGET_CMODEL_SMALL, TARGET_CMODEL_LARGE):
7845 * config/or1k/or1k.opt (mcmodel=): New option.
7846 * doc/invoke.texi (OpenRISC Options): Document mcmodel.
7848 2021-08-14 Martin Sebor <msebor@redhat.com>
7850 PR middle-end/101791
7851 * gimple-ssa-warn-access.cc (new_delete_mismatch_p): Use new argument
7852 to valid_new_delete_pair_p.
7853 * tree.c (valid_new_delete_pair_p): Add argument.
7854 * tree.h (valid_new_delete_pair_p): Same.
7856 2021-08-14 Jakub Jelinek <jakub@redhat.com>
7859 * config/i386/i386-expand.c (expand_vec_perm_broadcast_1)
7860 <case E_V64QImode>: For this mode assert
7861 !TARGET_AVX512BW || d->perm[0] rather than !TARGET_AVX2 || d->perm[0].
7863 2021-08-13 Michael Meissner <meissner@linux.ibm.com>
7866 * config/rs6000/altivec.md (xxeval): Use register_predicate
7867 instead of altivec_register_predicate.
7869 2021-08-13 Martin Sebor <msebor@redhat.com>
7871 PR middle-end/101734
7872 * tree-ssa-uninit.c (maybe_warn_read_write_only): New function.
7873 (maybe_warn_operand): Call it.
7875 2021-08-13 Martin Liska <mliska@suse.cz>
7878 * attribs.c (decl_attributes): Make naked functions "noipa"
7881 2021-08-13 Martin Liska <mliska@suse.cz>
7884 * symtab.c (symtab_node::noninterposable_alias): Do not create
7885 local aliases for target_clone functions as the clonning pass
7888 2021-08-13 Martin Liska <mliska@suse.cz>
7890 * opts.c (LIVE_PATCHING_OPTION): Define.
7891 (control_options_for_live_patching): Use it in error messages.
7893 2021-08-13 Jan Hubicka <hubicka@ucw.cz>
7895 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOREAD.
7896 (implicit_const_eaf_flags, implicit_pure_eaf_flags,
7897 ignore_stores_eaf_flags): New constants.
7898 (remove_useless_eaf_flags): New function.
7899 (eaf_flags_useful_p): Use it.
7900 (deref_flags): Add EAF_NOT_RETURNED if flag is unused;
7902 (modref_lattice::init): Add EAF_NOREAD.
7903 (modref_lattice::add_escape_point): Do not reacord escape point if
7905 (modref_lattice::merge): EAF_NOESCAPE implies EAF_NODIRECTESCAPE;
7906 use remove_useless_eaf_flags.
7907 (modref_lattice::merge_deref): Use ignore_stores_eaf_flags.
7908 (modref_lattice::merge_direct_load): Add EAF_NOREAD
7909 (analyze_ssa_name_flags): Fix handling EAF_NOT_RETURNED
7910 (analyze_parms): Use remove_useless_eaf_flags.
7911 (ipa_merge_modref_summary_after_inlining): Use ignore_stores_eaf_flags.
7912 (modref_merge_call_site_flags): Add caller and ecf_flags parameter;
7913 use remove_useless_eaf_flags.
7914 (modref_propagate_flags_in_scc): Update.
7915 * ipa-modref.h: Turn eaf_flags_t back to char.
7916 * tree-core.h (EAF_NOT_RETURNED): Fix.
7917 (EAF_NOREAD): New constant
7918 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1): Check for
7920 * tree-ssa-structalias.c (handle_rhs_call): Handle new flags.
7921 (handle_pure_call): Likewise.
7923 2021-08-12 Jakub Jelinek <jakub@redhat.com>
7925 * tree.def (OMP_MASKED): New tree code.
7926 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_FILTER.
7927 * tree.h (OMP_MASKED_BODY, OMP_MASKED_CLAUSES, OMP_MASKED_COMBINED,
7928 OMP_CLAUSE_FILTER_EXPR): Define.
7929 * tree.c (omp_clause_num_ops): Add OMP_CLAUSE_FILTER entry.
7930 (omp_clause_code_name): Likewise.
7931 (walk_tree_1): Handle OMP_CLAUSE_FILTER.
7932 * tree-nested.c (convert_nonlocal_omp_clauses,
7933 convert_local_omp_clauses): Handle OMP_CLAUSE_FILTER.
7934 (convert_nonlocal_reference_stmt, convert_local_reference_stmt,
7935 convert_gimple_call): Handle GIMPLE_OMP_MASTER.
7936 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_FILTER.
7937 (dump_generic_node): Handle OMP_MASTER.
7938 * gimple.def (GIMPLE_OMP_MASKED): New gimple code.
7939 * gimple.c (gimple_build_omp_masked): New function.
7940 (gimple_copy): Handle GIMPLE_OMP_MASKED.
7941 * gimple.h (gimple_build_omp_masked): Declare.
7942 (gimple_has_substatements): Handle GIMPLE_OMP_MASKED.
7943 (gimple_omp_masked_clauses, gimple_omp_masked_clauses_ptr,
7944 gimple_omp_masked_set_clauses): New inline functions.
7945 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_MASKED.
7946 * gimple-pretty-print.c (dump_gimple_omp_masked): New function.
7947 (pp_gimple_stmt_1): Handle GIMPLE_OMP_MASKED.
7948 * gimple-walk.c (walk_gimple_stmt): Likewise.
7949 * gimple-low.c (lower_stmt): Likewise.
7950 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
7951 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_FILTER. For clauses
7952 that take one expression rather than decl or constant, force
7953 gimplification of that into a SSA_NAME or temporary unless min
7955 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_FILTER.
7956 (gimplify_expr): Handle OMP_MASKED.
7957 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_MASKED.
7958 (estimate_num_insns): Likewise.
7959 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_FILTER.
7960 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_MASKED. Adjust
7961 diagnostics for existence of masked construct.
7962 (scan_omp_1_stmt, lower_omp_master, lower_omp_1, diagnose_sb_1,
7963 diagnose_sb_2): Handle GIMPLE_OMP_MASKED.
7964 * omp-expand.c (expand_omp_synch, expand_omp, omp_make_gimple_edges):
7967 2021-08-12 Uroš Bizjak <ubizjak@gmail.com>
7970 * config/i386/i386.md (avx512f_scalef<mode>2): New insn pattern.
7971 (ldexp<mode>3): Use avx512f_scalef<mode>2.
7972 (UNSPEC_SCALEF): Move from ...
7973 * config/i386/sse.md (UNSPEC_SCALEF): ... here.
7975 2021-08-12 Jan Hubicka <hubicka@ucw.cz>
7977 * ipa-split.c (consider_split): Fix condition testing void functions.
7979 2021-08-12 Aldy Hernandez <aldyh@redhat.com>
7981 * doc/invoke.texi: Remove docs for threader-mode param.
7982 * flag-types.h (enum threader_mode): Remove.
7983 * params.opt: Remove threader-mode param.
7984 * tree-ssa-threadbackward.c (class back_threader): Remove
7985 path_is_unreachable_p.
7986 Make find_paths private.
7987 Add maybe_thread and thread_through_all_blocks.
7988 Remove reference marker for m_registry.
7989 Remove reference marker for m_profit.
7990 (back_threader::back_threader): Adjust for registry and profit not
7992 (dump_path): Move down.
7994 (class thread_jumps): Remove.
7995 (class back_threader_registry): Remove m_all_paths.
7997 (thread_jumps::thread_through_all_blocks): Move to back_threader
7999 (fsm_find_thread_path): Remove
8000 (back_threader::maybe_thread): New.
8001 (back_threader::thread_through_all_blocks): Move from
8003 (back_threader_registry::back_threader_registry): Remove
8005 (back_threader_registry::~back_threader_registry): Remove.
8006 (thread_jumps::find_taken_edge): Remove.
8007 (thread_jumps::check_subpath_and_update_thread_path): Remove.
8008 (thread_jumps::maybe_register_path): Remove.
8009 (thread_jumps::handle_phi): Remove.
8010 (handle_assignment_p): Remove.
8011 (thread_jumps::handle_assignment): Remove.
8012 (thread_jumps::fsm_find_control_statement_thread_paths): Remove.
8013 (thread_jumps::find_jump_threads_backwards): Remove.
8014 (thread_jumps::find_jump_threads_backwards_with_ranger): Remove.
8015 (try_thread_blocks): Rename find_jump_threads_backwards to
8017 (pass_early_thread_jumps::execute): Same.
8019 2021-08-12 Tobias Burnus <tobias@codesourcery.com>
8021 * tree-core.h (omp_clause_proc_bind_kind): Add
8022 OMP_CLAUSE_PROC_BIND_PRIMARY.
8023 * tree-pretty-print.c (dump_omp_clause): Add TODO comment to
8024 change 'master' to 'primary' in proc_bind for OpenMP 5.1.
8026 2021-08-12 Claudiu Zissulescu <claziss@synopsys.com>
8028 * common/config/arc/arc-common.c (arc_option_init_struct): Remove
8029 fno-common reference.
8030 * config/arc/arc.c (arc_override_options): Remove overriding of
8033 2021-08-12 Jakub Jelinek <jakub@redhat.com>
8036 * config/i386/i386-expand.c (ix86_expand_vec_one_operand_perm_avx512):
8037 If d->testing_p, return true after performing checks instead of
8038 actually expanding the insn.
8039 (expand_vec_perm_broadcast_1): Handle V32HImode - assert
8040 !TARGET_AVX512BW and return false.
8042 2021-08-12 Eric Botcazou <ebotcazou@gcc.gnu.org>
8044 * configure.ac (PE linker --disable-dynamicbase support): New check.
8045 * configure: Regenerate.
8046 * config.in: Likewise.
8047 * config/i386/mingw32.h (LINK_SPEC_DISABLE_DYNAMICBASE): New define.
8048 (LINK_SPEC): Use it.
8049 * config/i386/mingw-w64.h (LINK_SPEC_DISABLE_DYNAMICBASE): Likewise.
8050 (LINK_SPEC): Likewise.
8052 2021-08-12 liuhongt <hongtao.liu@intel.com>
8055 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_2): New
8056 post_reload define_insn_and_split.
8057 (*avx512bw_zero_extendv32qiv32hi2_2): Ditto.
8058 (*sse4_1_zero_extendv8qiv8hi2_4): Ditto.
8059 (*avx512f_zero_extendv16hiv16si2_2): Ditto.
8060 (*avx2_zero_extendv8hiv8si2_2): Ditto.
8061 (*sse4_1_zero_extendv4hiv4si2_4): Ditto.
8062 (*avx512f_zero_extendv8siv8di2_2): Ditto.
8063 (*avx2_zero_extendv4siv4di2_2): Ditto.
8064 (*sse4_1_zero_extendv2siv2di2_4): Ditto.
8065 (VI248_256, VI248_512, VI148_512, VI148_256, VI148_128): New
8068 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
8070 * config/rs6000/rs6000-builtin-new.def: Add always, power5, and
8073 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
8075 * config/rs6000/rs6000-builtin-new.def: Add vsx stanza.
8077 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
8079 * config/rs6000/rs6000-builtin-new.def: Finish altivec stanza.
8080 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
8081 initialization of pcvoid_type_node here...
8082 (altivec_init_builtins): ...from here.
8083 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add
8084 RS6000_BTI_const_ptr_void.
8085 (pcvoid_type_node): New macro.
8087 2021-08-11 Richard Biener <rguenther@suse.de>
8090 * tree-ssa-forwprop.c (pass_forwprop::execute): Do not decompose
8091 hard-register accesses.
8093 2021-08-11 Richard Biener <rguenther@suse.de>
8095 * tree-ssa-operands.c (operands_scanner::get_expr_operands):
8096 Do not look at COMPONENT_REF FIELD_DECLs TREE_THIS_VOLATILE
8097 to determine has_volatile_ops.
8099 2021-08-11 Eric Botcazou <ebotcazou@gcc.gnu.org>
8101 * cfgexpand.c (expand_used_vars): Reuse attribs local variable.
8103 2021-08-11 Jan Hubicka <hubicka@ucw.cz>
8104 Alexandre Oliva <oliva@adacore.com>
8106 * ipa-modref.c (modref_lattice::dump): Fix escape_point's min_flags
8108 (modref_lattice::merge_deref): Fix handling of indirect scape points.
8109 (update_escape_summary_1): Likewise.
8110 (update_escape_summary): Likewise.
8111 (ipa_merge_modref_summary_after_inlining): Likewise.
8113 2021-08-11 Richard Biener <rguenther@suse.de>
8115 PR middle-end/101858
8116 * fold-const.c (fold_binary_loc): Guard simplification
8117 of X < (cast) (1 << Y) to integer types.
8119 2021-08-11 Richard Biener <rguenther@suse.de>
8121 PR tree-optimization/101861
8122 * tree-vect-stmts.c (vectorizable_load): Fix error in
8123 previous change with regard to gather vectorization.
8125 2021-08-11 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
8128 * config/arm/arm_neon.h (vdup_n_s8): Replace call to builtin
8130 (vdup_n_s16): Likewise.
8131 (vdup_n_s32): Likewise.
8132 (vdup_n_s64): Likewise.
8133 (vdup_n_u8): Likewise.
8134 (vdup_n_u16): Likewise.
8135 (vdup_n_u32): Likewise.
8136 (vdup_n_u64): Likewise.
8137 (vdup_n_p8): Likewise.
8138 (vdup_n_p16): Likewise.
8139 (vdup_n_p64): Likewise.
8140 (vdup_n_f16): Likewise.
8141 (vdup_n_f32): Likewise.
8142 (vdupq_n_s8): Likewise.
8143 (vdupq_n_s16): Likewise.
8144 (vdupq_n_s32): Likewise.
8145 (vdupq_n_s64): Likewise.
8146 (vdupq_n_u8): Likewise.
8147 (vdupq_n_u16): Likewise.
8148 (vdupq_n_u32): Likewise.
8149 (vdupq_n_u64): Likewise.
8150 (vdupq_n_p8): Likewise.
8151 (vdupq_n_p16): Likewise.
8152 (vdupq_n_p64): Likewise.
8153 (vdupq_n_f16): Likewise.
8154 (vdupq_n_f32): Likewise.
8155 (vmov_n_s8): Replace call to builtin with call to corresponding
8157 (vmov_n_s16): Likewise.
8158 (vmov_n_s32): Likewise.
8159 (vmov_n_s64): Likewise.
8160 (vmov_n_u8): Likewise.
8161 (vmov_n_u16): Likewise.
8162 (vmov_n_u32): Likewise.
8163 (vmov_n_u64): Likewise.
8164 (vmov_n_p8): Likewise.
8165 (vmov_n_p16): Likewise.
8166 (vmov_n_f16): Likewise.
8167 (vmov_n_f32): Likewise.
8168 (vmovq_n_s8): Likewise.
8169 (vmovq_n_s16): Likewise.
8170 (vmovq_n_s32): Likewise.
8171 (vmovq_n_s64): Likewise.
8172 (vmovq_n_u8): Likewise.
8173 (vmovq_n_u16): Likewise.
8174 (vmovq_n_u32): Likewise.
8175 (vmovq_n_u64): Likewise.
8176 (vmovq_n_p8): Likewise.
8177 (vmovq_n_p16): Likewise.
8178 (vmovq_n_f16): Likewise.
8179 (vmovq_n_f32): Likewise.
8180 * config/arm/arm_neon_builtins.def: Remove entries for vdup_n.
8182 2021-08-11 liuhongt <hongtao.liu@intel.com>
8185 * config/i386/i386.md (ldexp<mode>3): Extend to vscalefs[sd]
8186 when TARGET_AVX512F and TARGET_SSE_MATH.
8188 2021-08-10 Jakub Jelinek <jakub@redhat.com>
8191 * config/i386/i386-expand.c (expand_vec_perm_even_odd): Return false
8192 for V32HImode if !TARGET_AVX512BW.
8193 (ix86_vectorize_vec_perm_const) <case E_V32HImode, case E_V64QImode>:
8194 If !TARGET_AVX512BW and TARGET_AVX512F and d.testing_p, don't fail
8195 early, but actually check the permutation.
8197 2021-08-10 Richard Biener <rguenther@suse.de>
8199 PR tree-optimization/101809
8200 * tree-vect-stmts.c (get_load_store_type): Allow emulated
8201 gathers with offset vector nunits being a constant multiple
8202 of the data vector nunits.
8203 (vect_get_gather_scatter_ops): Use the appropriate nunits
8204 for the offset vector defs.
8205 (vectorizable_store): Adjust call to
8206 vect_get_gather_scatter_ops.
8207 (vectorizable_load): Likewise. Handle the case of less
8208 offset vectors than data vectors.
8210 2021-08-10 Jakub Jelinek <jakub@redhat.com>
8213 * config/i386/sse.md (*avx512f_shuf_<shuffletype>64x2_1<mask_name>_1,
8214 *avx512f_shuf_<shuffletype>32x4_1<mask_name>_1): New define_insn
8217 2021-08-10 Richard Biener <rguenther@suse.de>
8219 PR tree-optimization/101801
8220 PR tree-optimization/101819
8221 * tree-vectorizer.h (vect_emulated_vector_p): Declare.
8222 * tree-vect-loop.c (vect_emulated_vector_p): New function.
8223 (vectorizable_reduction): Re-instantiate a check for emulated
8225 * tree-vect-stmts.c (vectorizable_shift): Likewise.
8226 (vectorizable_operation): Likewise. Cost emulated vector
8227 operations according to the scalar sequence synthesized by
8230 2021-08-10 Richard Biener <rguenther@suse.de>
8232 PR middle-end/101824
8233 * tree-nested.c (get_frame_field): Mark the COMPONENT_REF as
8234 volatile in case the variable was.
8236 2021-08-10 H.J. Lu <hjl.tools@gmail.com>
8239 * config/i386/constraints.md (BC): Document for integer SSE
8240 constant all bits set operand.
8241 (BF): New constraint for const floating-point all bits set
8243 * config/i386/i386.c (standard_sse_constant_p): Likewise.
8244 (standard_sse_constant_opcode): Likewise.
8245 * config/i386/sse.md (sseconstm1): New mode attribute.
8246 (mov<mode>_internal): Replace BC with <sseconstm1>.
8248 2021-08-10 liuhongt <hongtao.liu@intel.com>
8250 * config/i386/sse.md (cond_<insn><mode>): New expander.
8251 (VI248_AVX512VLBW): New mode iterator.
8252 * config/i386/predicates.md
8253 (nonimmediate_or_const_vec_dup_operand): New predicate.
8255 2021-08-09 Andrew MacLeod <amacleod@redhat.com>
8257 PR tree-optimization/101741
8258 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Check
8259 type of parameter for toupper/tolower.
8261 2021-08-09 Martin Jambor <mjambor@suse.cz>
8264 * ipa-prop.c (propagate_controlled_uses): Removed a spurious space.
8266 2021-08-09 Pat Haugen <pthaugen@linux.ibm.com>
8268 * config/rs6000/rs6000.c (is_load_insn1): Verify destination is a
8270 (is_store_insn1): Verify source is a register.
8272 2021-08-09 Uroš Bizjak <ubizjak@gmail.com>
8275 * config/i386/mmx.md (<any_logic:code>v2sf3):
8276 Rename from *mmx_<any_logic:code>v2sf3
8278 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
8280 * config/nvptx/nvptx.c: Cross-reference parts adapted in
8281 'gcc/omp-oacc-neuter-broadcast.cc'.
8282 * omp-low.c: Likewise.
8283 * omp-oacc-neuter-broadcast.cc: Cross-reference parts adapted from
8286 2021-08-09 Julian Brown <julian@codesourcery.com>
8287 Kwok Cheung Yeung <kcy@codesourcery.com>
8288 Thomas Schwinge <thomas@codesourcery.com>
8290 * config/gcn/gcn.c (gcn_init_builtins): Override decls for
8291 BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START,
8292 BUILT_IN_GOACC_SINGLE_COPY_END and BUILT_IN_GOACC_BARRIER.
8293 (gcn_goacc_validate_dims): Turn on worker partitioning unconditionally.
8294 (gcn_fork_join): Update comment.
8295 * config/gcn/gcn.opt (flag_worker_partitioning): Remove.
8296 (macc_experimental_workers): Remove unused option.
8298 2021-08-09 Julian Brown <julian@codesourcery.com>
8299 Nathan Sidwell <nathan@codesourcery.com> (via 'gcc/config/nvptx/nvptx.c' master)
8300 Kwok Cheung Yeung <kcy@codesourcery.com>
8301 Thomas Schwinge <thomas@codesourcery.com>
8303 * Makefile.in (OBJS): Add omp-oacc-neuter-broadcast.o.
8304 * doc/tm.texi.in (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD):
8305 Add documentation hook.
8306 * doc/tm.texi: Regenerate.
8307 * omp-oacc-neuter-broadcast.cc: New file.
8308 * omp-builtins.def (BUILT_IN_GOACC_BARRIER)
8309 (BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START)
8310 (BUILT_IN_GOACC_SINGLE_COPY_END): New builtins.
8311 * passes.def (pass_omp_oacc_neuter_broadcast): Add pass.
8312 * target.def (goacc.create_worker_broadcast_record): Add target
8314 * tree-pass.h (make_pass_omp_oacc_neuter_broadcast): Add
8316 * config/gcn/gcn-protos.h (gcn_goacc_adjust_propagation_record):
8317 Rename prototype to...
8318 (gcn_goacc_create_worker_broadcast_record): ... this.
8319 * config/gcn/gcn-tree.c (gcn_goacc_adjust_propagation_record): Rename
8321 (gcn_goacc_create_worker_broadcast_record): ... this.
8322 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_PROPAGATION_RECORD):
8324 (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD): ... this.
8326 2021-08-09 Tejas Belagod <tejas.belagod@arm.com>
8329 * config/aarch64/aarch64-simd.md (vlshr<mode>3, vashr<mode>3): Use
8332 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
8334 * Makefile.in (GTFILES): Remove '$(srcdir)/omp-offload.c'.
8336 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
8338 * builtins.def (DEF_GOACC_BUILTIN, DEF_GOMP_BUILTIN): Don't
8339 consider '-foffload-abi'.
8340 * common.opt (-foffload-abi): Remove 'Var', 'Init'.
8341 * opts.c (common_handle_option) <-foffload-abi> [ACCEL_COMPILER]:
8344 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
8346 * optc-gen.awk: Sanity check that 'Init' doesn't appear without
8349 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
8351 * omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE): Remove.
8353 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
8355 * doc/gty.texi (Files): Update.
8357 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
8359 * doc/gty.texi (Files): Fix GTY header file example.
8361 2021-08-09 Roger Sayle <roger@nextmovesoftware.com>
8363 * tree-ssa-ccp.c (value_mask_to_min_max): Helper function to
8364 determine the upper and lower bounds from a mask-value pair.
8365 (bit_value_unop) [ABS_EXPR, ABSU_EXPR]: Add support for
8366 absolute value and unsigned absolute value expressions.
8367 (bit_value_binop): Initialize *VAL's precision.
8368 [LT_EXPR, LE_EXPR]: Use value_mask_to_min_max to determine
8369 upper and lower bounds of operands. Add LE_EXPR/GE_EXPR
8370 support when the operands are unknown but potentially equal.
8371 [MIN_EXPR, MAX_EXPR]: Support minimum/maximum expressions.
8373 2021-08-09 Bin Cheng <bin.cheng@linux.alibaba.com>
8375 * config/aarch64/aarch64.md
8376 (*extend<SHORT:mode><GPI:mode>2_aarch64): Use %<GPI:w>0.
8378 2021-08-08 Sergei Trofimovich <siarheit@google.com>
8380 * lra-constraints.c: Fix s/otput/output/ typo.
8382 2021-08-06 Martin Sebor <msebor@redhat.com>
8384 * builtins.c (expand_builtin_memchr): Move to gimple-ssa-warn-access.cc.
8385 (expand_builtin_strcat): Same.
8386 (expand_builtin_stpncpy): Same.
8387 (expand_builtin_strncat): Same.
8388 (check_read_access): Same.
8389 (check_memop_access): Same.
8390 (expand_builtin_strlen): Move checks to gimple-ssa-warn-access.cc.
8391 (expand_builtin_strnlen): Same.
8392 (expand_builtin_memcpy): Same.
8393 (expand_builtin_memmove): Same.
8394 (expand_builtin_mempcpy): Same.
8395 (expand_builtin_strcpy): Same.
8396 (expand_builtin_strcpy_args): Same.
8397 (expand_builtin_stpcpy_1): Same.
8398 (expand_builtin_strncpy): Same.
8399 (expand_builtin_memset): Same.
8400 (expand_builtin_bzero): Same.
8401 (expand_builtin_strcmp): Same.
8402 (expand_builtin_strncmp): Same.
8403 (expand_builtin): Remove handlers.
8404 (fold_builtin_strlen): Add a comment.
8405 * builtins.h (check_access): Move to gimple-ssa-warn-access.cc.
8406 * calls.c (maybe_warn_nonstring_arg): Same.
8407 * diagnostic-spec.c (nowarn_spec_t::nowarn_spec_t): Add warning option.
8408 * gimple-fold.c (gimple_fold_builtin_strcpy): Pass argument to callee.
8409 (gimple_fold_builtin_stpcpy): Same.
8410 * gimple-ssa-warn-access.cc (has_location): New function.
8411 (get_location): Same.
8412 (get_callee_fndecl): Same.
8415 (warn_string_no_nul): Define.
8416 (unterminated_array): Same.
8417 (check_nul_terminated_array): Same.
8418 (maybe_warn_nonstring_arg): Same.
8419 (maybe_warn_for_bound): Same.
8420 (warn_for_access): Same.
8421 (check_access): Same.
8422 (check_memop_access): Same.
8423 (check_read_access): Same.
8424 (warn_dealloc_offset): Use helper functions.
8425 (maybe_emit_free_warning): Same.
8426 (class pass_waccess): Add members.
8427 (check_strcat): New function.
8428 (check_strncat): New function.
8429 (check_stxcpy): New function.
8430 (check_stxncpy): New function.
8431 (check_strncmp): New function.
8432 (pass_waccess::check_builtin): New function.
8433 (pass_waccess::check): Call it.
8434 * gimple-ssa-warn-access.h (warn_string_no_nul): Move here from
8436 (maybe_warn_for_bound): Same.
8437 (check_access): Same.
8438 (check_memop_access): Same.
8439 (check_read_access): Same.
8440 * pointer-query.h (struct access_data): Define a ctor overload.
8442 2021-08-06 Richard Biener <rguenther@suse.de>
8444 PR tree-optimization/101801
8445 * tree-vectorizer.h (vect_worthwhile_without_simd_p): Rename...
8446 (vect_can_vectorize_without_simd_p): ... to this.
8447 * tree-vect-loop.c (vect_worthwhile_without_simd_p): Rename...
8448 (vect_can_vectorize_without_simd_p): ... to this and fold
8449 in vect_min_worthwhile_factor.
8450 (vect_min_worthwhile_factor): Remove.
8451 (vectorizable_reduction): Adjust and remove the cost part.
8452 * tree-vect-stmts.c (vectorizable_shift): Likewise.
8453 (vectorizable_operation): Likewise.
8455 2021-08-06 Uroš Bizjak <ubizjak@gmail.com>
8458 * config/i386/i386.md (cmove reg-to-reg move elimination peephole2s):
8459 Add general_gr_operand predicate to operand 3.
8461 2021-08-06 Roger Sayle <roger@nextmovesoftware.com>
8463 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern): Use
8464 CFN_BUILT_IN_CLRSB* instead of BUILT_IN_CLRSB* for consistency.
8466 2021-08-06 Tamar Christina <tamar.christina@arm.com>
8468 * config/aarch64/aarch64-sve-builtins.cc (register_svpattern,
8469 register_svprfop): Pass vec<> by pointer.
8470 * langhooks-def.h (lhd_simulate_enum_decl): Likewise.
8471 * langhooks.c (lhd_simulate_enum_decl): Likewise.
8472 * langhooks.h (struct lang_hooks_for_types): Likewise.
8474 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
8476 * config/aarch64/arm_neon.h (vst1_bf16_x2): Use
8477 __builtin_memcpy instead of constructing an additional
8478 __builtin_aarch64_simd_oi one vector at a time.
8479 (vst1q_bf16_x2): Likewise.
8480 (vst1_bf16_x3): Use __builtin_memcpy instead of constructing
8481 an additional __builtin_aarch64_simd_ci one vector at a time.
8482 (vst1q_bf16_x3): Likewise.
8483 (vst1_bf16_x4): Use __builtin_memcpy instead of a union.
8484 (vst1q_bf16_x4): Likewise.
8485 (vst2_bf16): Use __builtin_memcpy instead of constructing an
8486 additional __builtin_aarch64_simd_oi one vector at a time.
8487 (vst2q_bf16): Likewise.
8488 (vst3_bf16): Use __builtin_memcpy instead of constructing an
8489 additional __builtin_aarch64_simd_ci mode one vector at a
8491 (vst3q_bf16): Likewise.
8492 (vst4_bf16): Use __builtin_memcpy instead of constructing an
8493 additional __builtin_aarch64_simd_xi one vector at a time.
8494 (vst4q_bf16): Likewise.
8496 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
8498 * config/aarch64/arm_neon.h (__ST2_LANE_FUNC): Delete.
8499 (__ST2Q_LANE_FUNC): Delete.
8500 (vst2_lane_f16): Use __builtin_memcpy to copy vector
8501 structure instead of constructing __builtin_aarch64_simd_oi
8502 one vector at a time.
8503 (vst2_lane_f32): Likewise.
8504 (vst2_lane_f64): Likewise.
8505 (vst2_lane_p8): Likewise.
8506 (vst2_lane_p16): Likewise.
8507 (vst2_lane_p64): Likewise.
8508 (vst2_lane_s8): Likewise.
8509 (vst2_lane_s16): Likewise.
8510 (vst2_lane_s32): Likewise.
8511 (vst2_lane_s64): Likewise.
8512 (vst2_lane_u8): Likewise.
8513 (vst2_lane_u16): Likewise.
8514 (vst2_lane_u32): Likewise.
8515 (vst2_lane_u64): Likewise.
8516 (vst2_lane_bf16): Likewise.
8517 (vst2q_lane_f16): Use __builtin_memcpy to copy vector
8518 structure instead of using a union.
8519 (vst2q_lane_f32): Likewise.
8520 (vst2q_lane_f64): Likewise.
8521 (vst2q_lane_p8): Likewise.
8522 (vst2q_lane_p16): Likewise.
8523 (vst2q_lane_p64): Likewise.
8524 (vst2q_lane_s8): Likewise.
8525 (vst2q_lane_s16): Likewise.
8526 (vst2q_lane_s32): Likewise.
8527 (vst2q_lane_s64): Likewise.
8528 (vst2q_lane_u8): Likewise.
8529 (vst2q_lane_u16): Likewise.
8530 (vst2q_lane_u32): Likewise.
8531 (vst2q_lane_u64): Likewise.
8532 (vst2q_lane_bf16): Likewise.
8534 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
8536 * config/aarch64/arm_neon.h (__ST3_LANE_FUNC): Delete.
8537 (__ST3Q_LANE_FUNC): Delete.
8538 (vst3_lane_f16): Use __builtin_memcpy to copy vector
8539 structure instead of constructing __builtin_aarch64_simd_ci
8540 one vector at a time.
8541 (vst3_lane_f32): Likewise.
8542 (vst3_lane_f64): Likewise.
8543 (vst3_lane_p8): Likewise.
8544 (vst3_lane_p16): Likewise.
8545 (vst3_lane_p64): Likewise.
8546 (vst3_lane_s8): Likewise.
8547 (vst3_lane_s16): Likewise.
8548 (vst3_lane_s32): Likewise.
8549 (vst3_lane_s64): Likewise.
8550 (vst3_lane_u8): Likewise.
8551 (vst3_lane_u16): Likewise.
8552 (vst3_lane_u32): Likewise.
8553 (vst3_lane_u64): Likewise.
8554 (vst3_lane_bf16): Likewise.
8555 (vst3q_lane_f16): Use __builtin_memcpy to copy vector
8556 structure instead of using a union.
8557 (vst3q_lane_f32): Likewise.
8558 (vst3q_lane_f64): Likewise.
8559 (vst3q_lane_p8): Likewise.
8560 (vst3q_lane_p16): Likewise.
8561 (vst3q_lane_p64): Likewise.
8562 (vst3q_lane_s8): Likewise.
8563 (vst3q_lane_s16): Likewise.
8564 (vst3q_lane_s32): Likewise.
8565 (vst3q_lane_s64): Likewise.
8566 (vst3q_lane_u8): Likewise.
8567 (vst3q_lane_u16): Likewise.
8568 (vst3q_lane_u32): Likewise.
8569 (vst3q_lane_u64): Likewise.
8570 (vst3q_lane_bf16): Likewise.
8572 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
8574 * config/aarch64/arm_neon.h (__ST4_LANE_FUNC): Delete.
8575 (__ST4Q_LANE_FUNC): Delete.
8576 (vst4_lane_f16): Use __builtin_memcpy to copy vector
8577 structure instead of constructing __builtin_aarch64_simd_xi
8578 one vector at a time.
8579 (vst4_lane_f32): Likewise.
8580 (vst4_lane_f64): Likewise.
8581 (vst4_lane_p8): Likewise.
8582 (vst4_lane_p16): Likewise.
8583 (vst4_lane_p64): Likewise.
8584 (vst4_lane_s8): Likewise.
8585 (vst4_lane_s16): Likewise.
8586 (vst4_lane_s32): Likewise.
8587 (vst4_lane_s64): Likewise.
8588 (vst4_lane_u8): Likewise.
8589 (vst4_lane_u16): Likewise.
8590 (vst4_lane_u32): Likewise.
8591 (vst4_lane_u64): Likewise.
8592 (vst4_lane_bf16): Likewise.
8593 (vst4q_lane_f16): Use __builtin_memcpy to copy vector
8594 structure instead of using a union.
8595 (vst4q_lane_f32): Likewise.
8596 (vst4q_lane_f64): Likewise.
8597 (vst4q_lane_p8): Likewise.
8598 (vst4q_lane_p16): Likewise.
8599 (vst4q_lane_p64): Likewise.
8600 (vst4q_lane_s8): Likewise.
8601 (vst4q_lane_s16): Likewise.
8602 (vst4q_lane_s32): Likewise.
8603 (vst4q_lane_s64): Likewise.
8604 (vst4q_lane_u8): Likewise.
8605 (vst4q_lane_u16): Likewise.
8606 (vst4q_lane_u32): Likewise.
8607 (vst4q_lane_u64): Likewise.
8608 (vst4q_lane_bf16): Likewise.
8610 2021-08-06 Martin Liska <mliska@suse.cz>
8612 * config/rs6000/rs6000.c (rs6000_option_override_internal): When
8613 a target option is restored, it can have
8614 rs6000_long_double_type_size set to FLOAT_PRECISION_TFmode
8615 and error should not be emitted.
8617 2021-08-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
8619 * gcov-io.h (gcov_write): Declare.
8620 * gcov-io.c (gcov_write): New.
8621 (gcov_write_counter): Remove.
8622 (gcov_write_tag_length): Likewise.
8623 (gcov_write_summary): Replace gcov_write_tag_length() with calls to
8624 gcov_write_unsigned().
8625 * doc/invoke.texi (fprofile-info-section): Mention
8626 __gcov_info_to_gdca().
8628 2021-08-06 Martin Sebor <msebor@redhat.com>
8630 * dominance.c (prune_bbs_to_update_dominators): Adjust by-value vec
8631 arguments to by-reference.
8632 (iterate_fix_dominators): Same.
8633 * dominance.h (iterate_fix_dominators): Same.
8634 * ipa-prop.h: Call auto_vec::to_vec_legacy.
8635 * tree-data-ref.c (dump_data_dependence_relation): Adjust by-value vec
8636 arguments to by-reference.
8637 (debug_data_dependence_relation): Same.
8638 (dump_data_dependence_relations): Same.
8639 * tree-data-ref.h (debug_data_dependence_relation): Same.
8640 (dump_data_dependence_relations): Same.
8641 * tree-predcom.c (dump_chains): Same.
8642 (initialize_root_vars_lm): Same.
8643 (determine_unroll_factor): Same.
8644 (replace_phis_by_defined_names): Same.
8645 (insert_init_seqs): Same.
8646 (pcom_worker::tree_predictive_commoning_loop): Call
8647 auto_vec::to_vec_legacy.
8648 * tree-ssa-pre.c (insert_into_preds_of_block): Adjust by-value vec
8649 arguments to by-reference.
8650 * tree-ssa-threadbackward.c (populate_worklist): Same.
8651 (back_threader::resolve_def): Same.
8652 * tree-vect-data-refs.c (vect_check_nonzero_value): Same.
8653 (vect_enhance_data_refs_alignment): Same.
8654 (vect_check_lower_bound): Same.
8655 (vect_prune_runtime_alias_test_list): Same.
8656 (vect_permute_store_chain): Same.
8657 * tree-vect-slp-patterns.c (vect_normalize_conj_loc): Same.
8658 * tree-vect-stmts.c (vect_create_vectorized_demotion_stmts): Same.
8659 * tree-vectorizer.h (vect_permute_store_chain): Same.
8660 * vec.c (test_init): New function.
8661 (vec_c_tests): Call new function.
8662 * vec.h (vec): Declare ctors, dtor, and assignment.
8663 (auto_vec::vec_to_legacy): New function.
8664 (vec::copy): Adjust initialization.
8666 2021-08-05 H.J. Lu <hjl.tools@gmail.com>
8669 * config/i386/i386.c (ix86_can_inline_p): Ignore MASK_80387 if
8670 callee only uses GPRs.
8671 * config/i386/ia32intrin.h: Revert commit 5463cee2770.
8672 * config/i386/serializeintrin.h: Revert commit 71958f740f1.
8673 * config/i386/x86gprintrin.h: Add
8674 #pragma GCC target("general-regs-only") and #pragma GCC pop_options
8675 to disable non-GPR ISAs.
8677 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
8679 PR middle-end/101787
8680 * doc/md.texi (cond_ashl, cond_ashr, cond_lshr): Document.
8682 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
8684 * tree-vectorizer.h (vect_is_store_elt_extraction, vect_is_reduction)
8685 (vect_reduc_type, vect_embedded_comparison_type, vect_comparison_type)
8686 (vect_is_extending_load, vect_is_integer_truncation): New functions,
8687 moved from aarch64.c but given different names.
8688 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction)
8689 (aarch64_is_reduction, aarch64_reduc_type)
8690 (aarch64_embedded_comparison_type, aarch64_comparison_type)
8691 (aarch64_extending_load_p, aarch64_integer_truncation_p): Delete
8692 in favor of the above. Update callers accordingly.
8694 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
8697 * config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
8698 writing .cpu directive in asm output.
8699 * config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
8700 (arm_last_printed_arch_string): Delete.
8701 (arm_last-printed_fpu_string): Delete.
8702 (arm_configure_build_target): If use of floating-point/SIMD is
8703 disabled, remove all fp/simd related features from the target ISA.
8704 (last_arm_targ_options): New variable.
8705 (arm_print_asm_arch_directives): Add new parameters. Change order
8706 of emitted directives and handle all cases here.
8707 (arm_file_start): Always call arm_print_asm_arch_directives, move
8708 all generation of .arch/.arch_extension here.
8709 (arm_file_end): Call arm_print_asm_arch.
8710 (arm_declare_function_name): Call arm_print_asm_arch_directives
8711 instead of printing .arch/.fpu directives directly.
8713 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
8715 * config/arm/arm.c (arm_configure_build_target): Don't call
8716 arm_option_reconfigure_globals.
8717 (arm_option_restore): Call arm_option_reconfigure_globals after
8718 reconfiguring the target.
8719 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
8721 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
8723 * config/arm/arm.c (arm_configure_build_target): Ensure the target's
8724 arch_name is always set.
8726 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
8728 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
8729 of vec_select high-half from being added into Neon subtract
8732 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
8734 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
8735 of vec_select high-half from being added into Neon add cost.
8737 2021-08-05 Kewen Lin <linkw@linux.ibm.com>
8739 * cfgloop.h (loops_list::loops_list): Add one optional argument
8740 root and adjust accordingly, update loop tree walking and factor
8742 * cfgloop.c (loops_list::walk_loop_tree): ... this. New function.
8744 2021-08-05 Eric Botcazou <ebotcazou@gcc.gnu.org>
8746 PR tree-optimization/101626
8747 * tree-sra.c (propagate_subaccesses_from_rhs): Do not set the
8748 reverse scalar storage order on a pointer or vector component.
8750 2021-08-05 liuhongt <hongtao.liu@intel.com>
8752 * config/i386/sse.md (cond_<code><mode>): New expander.
8754 2021-08-05 liuhongt <hongtao.liu@intel.com>
8756 * config/i386/sse.md (cond_<code><mode>): New expander.
8758 2021-08-05 liuhongt <hongtao.liu@intel.com>
8760 * config/i386/sse.md (cond_<code><mode>): New expander.
8762 2021-08-04 David Malcolm <dmalcolm@redhat.com>
8765 * Makefile.in (ANALYZER_OBJS): Add analyzer/region-model-asm.o.
8767 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
8770 * config/i386/i386.h (STORE_MAX_PIECES): Allow 16/32/64 bytes
8771 only if TARGET_INTER_UNIT_MOVES_TO_VEC is true.
8773 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
8776 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
8777 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
8778 data with SSE register from one memory location to another.
8780 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
8782 * config/s390/s390.c (expand_perm_with_vpdi): New function.
8783 (vectorize_vec_perm_const_1): Call expand_perm_with_vpdi.
8784 * config/s390/vector.md (*vpdi1<mode>, @vpdi1<mode>): Enable a
8785 parameterized expander.
8786 (*vpdi4<mode>, @vpdi4<mode>): Likewise.
8788 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
8790 * config/s390/s390.c (MAX_VECT_LEN): Define macro.
8791 (struct expand_vec_perm_d): Define struct.
8792 (expand_perm_with_merge): New function.
8793 (vectorize_vec_perm_const_1): New function.
8794 (s390_vectorize_vec_perm_const): New function.
8795 (TARGET_VECTORIZE_VEC_PERM_CONST): Define target macro.
8797 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
8799 * config/s390/vector.md (V_HW_64): Remove mode iterator.
8800 (*vec_load_pair<mode>): Use V_HW_2 instead of V_HW_64.
8801 * config/s390/vx-builtins.md
8802 (vec_scatter_element<V_HW_2:mode>_SI): Use V_HW_2 instead of
8805 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
8807 * config/s390/s390.md (UNSPEC_VEC_PERMI): Remove constant
8809 * config/s390/vector.md (*vpdi1<mode>, *vpdi4<mode>): New pattern
8811 * config/s390/vx-builtins.md (*vec_permi<mode>): Emit generic rtx
8812 instead of an unspec.
8814 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
8816 * config/s390/s390-modes.def: Add more vector modes to support
8817 concatenation of two vectors.
8818 * config/s390/s390-protos.h (s390_expand_merge_perm_const): Add
8820 (s390_expand_merge): Likewise.
8821 * config/s390/s390.c (s390_expand_merge_perm_const): New function.
8822 (s390_expand_merge): New function.
8823 * config/s390/s390.md (UNSPEC_VEC_MERGEH, UNSPEC_VEC_MERGEL):
8824 Remove constant definitions.
8825 * config/s390/vector.md (V_HW_2): Add mode iterators.
8826 (VI_HW_4, V_HW_4): Rename VI_HW_4 to V_HW_4.
8827 (vec_2x_nelts, vec_2x_wide): New mode attributes.
8828 (*vmrhb, *vmrlb, *vmrhh, *vmrlh, *vmrhf, *vmrlf, *vmrhg, *vmrlg):
8829 New pattern definitions.
8830 (vec_widen_umult_lo_<mode>, vec_widen_umult_hi_<mode>)
8831 (vec_widen_smult_lo_<mode>, vec_widen_smult_hi_<mode>)
8832 (vec_unpacks_lo_v4sf, vec_unpacks_hi_v4sf, vec_unpacks_lo_v2df)
8833 (vec_unpacks_hi_v2df): Adjust expanders to emit non-unspec RTX for
8835 * config/s390/vx-builtins.md (V_HW_4): Remove mode iterator. Now
8837 (vec_mergeh<mode>, vec_mergel<mode>): Use s390_expand_merge to
8838 emit vec merge pattern.
8840 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
8842 * config/aarch64/aarch64.c (aarch64_strip_extend_vec_half):
8844 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent cost of
8845 vec_select high-half from being added into Neon multiply
8847 * rtlanal.c (vec_series_highpart_p): Define.
8848 * rtlanal.h (vec_series_highpart_p): Declare.
8850 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
8852 * config/aarch64/aarch64.c (aarch64_strip_duplicate_vec_elt):
8854 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent
8855 vec_select cost from being added into Neon multiply cost.
8857 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
8859 * tree-vect-loop.c (vect_better_loop_vinfo_p): Detect cases in
8860 which old_loop_vinfo is an epilogue loop that handles a constant
8861 number of iterations.
8863 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
8865 * tree-vect-loop.c (vect_analyze_loop): Print a dump message
8866 when a reanalyzed loop fails to be cheaper than the current
8869 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
8871 * config/aarch64/aarch64.c: Fix a typo.
8873 2021-08-04 Vincent Lefèvre <vincent-gcc@vinc17.net>
8875 PR gcov-profile/101773
8876 * gcov-io.c (gcov_close): Check return code of a fclose.
8878 2021-08-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
8881 * dwarf2out.c (dwarf2out_assembly_start): Emit a dummy
8882 .file statement when needed.
8884 2021-08-04 Richard Biener <rguenther@suse.de>
8886 * tree-vect-data-refs.c (vect_check_gather_scatter):
8887 Include widening conversions only when the result is
8888 still handed by native gather or the current offset
8889 size not already matches the data size.
8890 Also succeed analysis in case there's no native support,
8891 noted by a IFN_LAST ifn and a NULL decl.
8892 (vect_analyze_data_refs): Always consider gathers.
8893 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
8894 Test for no IFN gather rather than decl gather.
8895 * tree-vect-stmts.c (vect_model_load_cost): Pass in the
8896 gather-scatter info and cost emulated gathers accordingly.
8897 (vect_truncate_gather_scatter_offset): Properly test for
8899 (vect_use_strided_gather_scatters_p): Likewise.
8900 (get_load_store_type): Handle emulated gathers and its
8902 (vectorizable_load): Likewise. Emulate them by extracting
8903 scalar offsets, doing scalar loads and a vector construct.
8905 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
8908 * expr.c (op_by_pieces_d::op_by_pieces_d): Add a max_pieces
8909 argument to set m_max_size.
8910 (move_by_pieces_d): Pass MOVE_MAX_PIECES to op_by_pieces_d.
8911 (store_by_pieces_d): Pass STORE_MAX_PIECES to op_by_pieces_d.
8912 (compare_by_pieces_d): Pass COMPARE_MAX_PIECES to op_by_pieces_d.
8914 2021-08-04 Roger Sayle <roger@nextmovesoftware.com>
8915 Marc Glisse <marc.glisse@inria.fr>
8917 * match.pd (bit_ior, bit_xor): Canonicalize (X*C1)|(X*C2) and
8918 (X*C1)^(X*C2) as X*(C1+C2), and related variants, using
8919 tree_nonzero_bits to ensure that operands are bit-wise disjoint.
8921 2021-08-04 Richard Biener <rguenther@suse.de>
8923 * tree-ssa-forwprop.c (pass_forwprop::execute): Split
8924 out code to decompose vector loads ...
8925 (optimize_vector_load): ... here. Generalize it to
8926 handle intermediate widening and TARGET_MEM_REF loads
8927 and apply it to loads with a supported vector mode as well.
8929 2021-08-04 Richard Biener <rguenther@suse.de>
8931 PR tree-optimization/101756
8932 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Make sure
8933 the result of the reduction epilogue is compatible to the original
8936 2021-08-04 liuhongt <hongtao.liu@intel.com>
8939 * config/i386/i386.md (peephole2): Refine predicate from
8940 register_operand to general_reg_operand.
8942 2021-08-04 Aldy Hernandez <aldyh@redhat.com>
8944 * gimple-range-path.h (path_range_query::dump): Mark override.
8946 2021-08-04 Richard Biener <rguenther@suse.de>
8948 PR tree-optimization/101769
8949 * tree-tailcall.c (eliminate_tail_call): Add the created loop
8950 for the first recursion and return it via the new output parameter.
8951 (optimize_tail_call): Pass through new output param.
8952 (tree_optimize_tail_calls_1): After creating all latches,
8953 add the created loop to the loop tree. Do not mark loops for fixup.
8955 2021-08-04 Martin Liska <mliska@suse.cz>
8957 * doc/invoke.texi: Document threader-mode param.
8959 2021-08-04 liuhongt <hongtao.liu@intel.com>
8961 * config/i386/sse.md (cond_fma<mode>): New expander.
8962 (cond_fms<mode>): Ditto.
8963 (cond_fnma<mode>): Ditto.
8964 (cond_fnms<mode>): Ditto.
8966 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
8968 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode>): Use && instead of &.
8970 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
8972 * config/rs6000/constraints.md: Remove "e" from the list of available
8973 constraint characters.
8975 2021-08-03 Eugene Rozenfeld <erozen@microsoft.com>
8977 PR gcov-profile/71672
8978 * auto-profile.c (afdo_indirect_call): Fix setup of the historgram value for indirect calls.
8980 2021-08-03 Paul A. Clarke <pc@us.ibm.com>
8982 * config/rs6000/smmintrin.h (_mm_minpos_epu16): New.
8984 2021-08-03 H.J. Lu <hjl.tools@gmail.com>
8986 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): In 64-bit mode,
8987 try XMM31 to avoid vzeroupper.
8989 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8991 * doc/invoke.texi: Document -mtune=neoverse-512tvb and
8992 -mcpu=neoverse-512tvb.
8993 * config/aarch64/aarch64-cores.def (neoverse-512tvb): New entry.
8994 * config/aarch64/aarch64-tune.md: Regenerate.
8995 * config/aarch64/aarch64.c (neoverse512tvb_sve_vector_cost)
8996 (neoverse512tvb_sve_issue_info, neoverse512tvb_vec_issue_info)
8997 (neoverse512tvb_vector_cost, neoverse512tvb_tunings): New structures.
8998 (aarch64_adjust_body_cost_sve): Handle -mtune=neoverse-512tvb.
8999 (aarch64_adjust_body_cost): Likewise.
9001 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
9003 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Only
9004 record issue information for operations that occur in the
9007 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
9009 * config/aarch64/aarch64.c (aarch64_multiply_add_p): Add a vec_flags
9010 parameter. Detect cases in which an Advanced SIMD MLA would almost
9011 certainly require a MOV.
9012 (aarch64_count_ops): Update accordingly.
9014 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
9016 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction): New
9017 function, split out from...
9018 (aarch64_detect_vector_stmt_subtype): ...here.
9019 (aarch64_add_stmt_cost): Treat extracting element 0 as free.
9021 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
9023 * config/aarch64/aarch64-protos.h (sve_vec_cost):
9024 Add gather_load_x32_cost and gather_load_x64_cost.
9025 * config/aarch64/aarch64.c (generic_sve_vector_cost)
9026 (a64fx_sve_vector_cost, neoversev1_sve_vector_cost): Update
9027 accordingly, using the values given by the scalar_load * number
9028 of elements calculation that we used previously.
9029 (aarch64_detect_vector_stmt_subtype): Use the new fields.
9031 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
9033 * config/aarch64/aarch64.c (aarch64_adjust_body_cost_sve): New
9034 function, split out from...
9035 (aarch64_adjust_body_cost): ...here.
9037 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
9039 * config/aarch64/fractional-cost.h: New file.
9040 * config/aarch64/aarch64.c: Include <algorithm> (indirectly)
9041 and cost_fraction.h.
9042 (vec_cost_fraction): New typedef.
9043 (aarch64_detect_scalar_stmt_subtype): Use it for statement costs.
9044 (aarch64_detect_vector_stmt_subtype): Likewise.
9045 (aarch64_sve_adjust_stmt_cost, aarch64_adjust_stmt_cost): Likewise.
9046 (aarch64_estimate_min_cycles_per_iter): Use vec_cost_fraction
9048 (aarch64_adjust_body_cost): Likewise.
9049 (aarch64_test_cost_fraction): New function.
9050 (aarch64_run_selftests): Call it.
9052 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
9054 * config/aarch64/aarch64-protos.h (tune_params::sve_width): Turn
9056 * config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Update
9058 (aarch64_estimated_poly_value): Likewise. Use the least significant
9059 set bit for the minimum and likely values. Use the most significant
9060 set bit for the maximum value.
9062 2021-08-03 liuhongt <hongtao.liu@intel.com>
9064 * config/i386/sse.md (cond_<insn><mode>): New expander.
9065 (cond_mul<mode>): Ditto.
9067 2021-08-03 Kewen Lin <linkw@linux.ibm.com>
9069 * tree-cfg.c (move_sese_region_to_fn): Fix typos on dloop.
9071 2021-08-03 liuhongt <hongtao.liu@intel.com>
9073 * config/i386/sse.md (cond_<insn><mode>):New expander.
9074 (cond_mul<mode>): Ditto.
9075 (cond_div<mode>): Ditto.
9077 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
9079 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Also
9080 check stack_realign_needed for stack realignment.
9081 (ix86_legitimate_constant_p): Always allow CONST_WIDE_INT smaller
9082 than the largest integer supported by vector register.
9083 * config/i386/i386.h (MAX_MOVE_MAX): New. Set to 64.
9084 (MOVE_MAX): Set to bytes of the largest integer supported by
9086 (STORE_MAX_PIECES): New.
9088 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
9090 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
9091 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
9092 data from one memory location to another.
9094 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
9097 * config/i386/i386.c (TARGET_GEN_MEMSET_SCRATCH_RTX): New.
9099 2021-08-02 Aldy Hernandez <aldyh@redhat.com>
9101 PR tree-optimization/101724
9102 * params.opt: Remove --param=threader-iterative.
9103 * tree-ssa-threadbackward.c (pass_thread_jumps::execute): Remove
9106 2021-08-02 Tom de Vries <tdevries@suse.de>
9108 PR middle-end/101665
9109 * doc/extend.texi (nonnull attribute): Improve documentation.
9111 2021-08-02 Andrew Pinski <apinski@marvell.com>
9113 PR rtl-optimization/101683
9114 * rtlanal.c (may_trap_p_1): Handle UNSIGNED_FIX.
9116 2021-08-02 Roger Sayle <roger@nextmovesoftware.com>
9118 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern):
9119 Renamed from cond_removal_in_popcount_clz_ctz_pattern.
9120 Add support for BSWAP, FFS, PARITY and CLRSB builtins.
9121 (tree_ssa_phiop_worker): Update call to function above.
9123 2021-08-01 H.J. Lu <hjl.tools@gmail.com>
9126 * config/i386/i386.md (bsr_rex64_1_zext): New.
9127 (combine splitter for constant - clzll): Replace gen_bsr_rex64_1
9128 with gen_bsr_rex64_1_zext.
9130 2021-07-31 Jakub Jelinek <jakub@redhat.com>
9133 * config/i386/i386.md (bsr_rex64_1, bsr_1, bsr_zext_1): New
9134 define_insn patterns.
9135 (*bsr_rex64_2, *bsr_2): New define_insn_and_split patterns.
9136 Add combine splitters for constant - clz.
9137 (clz<mode>2): Use a temporary pseudo for bsr result.
9139 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
9141 * config/rs6000/smmintrin.h (_mm_floor_pd, _mm_floor_ps,
9142 _mm_floor_sd, _mm_floor_ss): New.
9144 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
9146 * config/rs6000/smmintrin.h (_mm_ceil_pd, _mm_ceil_ps,
9147 _mm_ceil_sd, _mm_ceil_ss): New.
9149 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
9151 * config/rs6000/smmintrin.h (_mm_blend_pd, _mm_blendv_pd,
9152 _mm_blend_ps, _mm_blendv_ps): New.
9154 2021-07-30 Roger Sayle <roger@nextmovesoftware.com>
9155 Uroš Bizjak <ubizjak@gmail.com>
9157 * config/i386/i386.md (*dec_cmov<mode>): New define_insn_and_split
9158 to generate a conditional move using the carry flag after sub $1.
9159 (peephole2): Eliminate a register-to-register move by inverting
9160 the condition of a conditional move.
9162 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
9164 * config/mmix/mmix.md ("call", "call_value", "*call_real")
9165 ("*call_value_real"): Don't generate rtx mentioning the generic
9166 operands 1 and 2 to "call", and similarly for "call_value".
9167 * config/mmix/mmix.c (mmix_print_operand_punct_valid_p)
9168 (mmix_print_operand): Use '!' instead of 'p'.
9170 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
9172 * doc/md.texi (call): Correct information about operand 2.
9173 * config/mmix/mmix.md ("call", "call_value"): Remove fixed FIXMEs.
9175 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
9177 * range-op.cc (operator_trunc_mod::wi_fold): Fold constants.
9179 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
9181 * range-op.cc (operator_div::wi_fold): Return UNDEFINED for [0, 0] divisor.
9183 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
9185 * gimple-range-cache.cc (*::set_bb_range): Change const basic_block to
9187 (*::get_bb_range): Ditto.
9188 (*::bb_range_p): Ditto.
9189 * gimple-range-cache.h: Change prototypes.
9191 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
9194 * builtins.c (builtin_memcpy_read_str): Change the mode argument
9195 from scalar_int_mode to fixed_size_mode.
9196 (builtin_strncpy_read_str): Likewise.
9197 (gen_memset_value_from_prev): New function.
9198 (builtin_memset_read_str): Change the mode argument from
9199 scalar_int_mode to fixed_size_mode. Use gen_memset_value_from_prev
9200 and support CONST_VECTOR.
9201 (builtin_memset_gen_str): Likewise.
9202 (try_store_by_multiple_pieces): Use by_pieces_constfn to declare
9204 * builtins.h (builtin_strncpy_read_str): Replace scalar_int_mode
9205 with fixed_size_mode.
9206 (builtin_memset_read_str): Likewise.
9207 * expr.c (widest_int_mode_for_size): Renamed to ...
9208 (widest_fixed_size_mode_for_size): Add a bool argument to
9209 indicate if QI vector mode can be used.
9210 (by_pieces_ninsns): Call widest_fixed_size_mode_for_size
9211 instead of widest_int_mode_for_size.
9212 (pieces_addr::adjust): Change the mode argument from
9213 scalar_int_mode to fixed_size_mode.
9214 (op_by_pieces_d): Make m_len read-only. Add a bool member,
9215 m_qi_vector_mode, to indicate that QI vector mode can be used.
9216 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
9217 initialize m_qi_vector_mode. Call widest_fixed_size_mode_for_size
9218 instead of widest_int_mode_for_size.
9219 (op_by_pieces_d::get_usable_mode): Change the mode argument from
9220 scalar_int_mode to fixed_size_mode. Call
9221 widest_fixed_size_mode_for_size instead of
9222 widest_int_mode_for_size.
9223 (op_by_pieces_d::smallest_fixed_size_mode_for_size): New member
9224 function to return the smallest integer or QI vector mode.
9225 (op_by_pieces_d::run): Call widest_fixed_size_mode_for_size
9226 instead of widest_int_mode_for_size. Call
9227 smallest_fixed_size_mode_for_size instead of
9228 smallest_int_mode_for_size.
9229 (store_by_pieces_d::store_by_pieces_d): Add a bool argument to
9230 indicate that QI vector mode can be used and pass it to
9231 op_by_pieces_d::op_by_pieces_d.
9232 (can_store_by_pieces): Call widest_fixed_size_mode_for_size
9233 instead of widest_int_mode_for_size. Pass memsetp to
9234 widest_fixed_size_mode_for_size to support QI vector mode.
9235 Allow all CONST_VECTORs for memset if vec_duplicate is supported.
9236 (store_by_pieces): Pass memsetp to
9237 store_by_pieces_d::store_by_pieces_d.
9238 (clear_by_pieces_1): Removed.
9239 (clear_by_pieces): Replace clear_by_pieces_1 with
9240 builtin_memset_read_str and pass true to store_by_pieces_d to
9241 support vector mode broadcast.
9242 (string_cst_read_str): Change the mode argument from
9243 scalar_int_mode to fixed_size_mode.
9244 * expr.h (by_pieces_constfn): Change scalar_int_mode to
9246 (by_pieces_prev): Likewise.
9247 * rtl.h (lowpart_subreg_regno): New.
9248 * rtlanal.c (lowpart_subreg_regno): New. A wrapper around
9249 simplify_subreg_regno.
9250 * target.def (gen_memset_scratch_rtx): New hook.
9251 * doc/tm.texi.in: Add TARGET_GEN_MEMSET_SCRATCH_RTX.
9252 * doc/tm.texi: Regenerated.
9254 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
9257 * config/mips/mips.c (mips_atomic_assign_expand_fenv): Use
9258 TARGET_EXPR instead of MODIFY_EXPR.
9260 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
9263 * config/mips/mips-protos.h (mips_expand_vec_cmp_expr): Declare.
9264 * config/mips/mips.c (mips_expand_vec_cmp_expr): New function.
9265 * config/mips/mips-msa.md (vec_cmp<MSA:mode><mode_i>): New
9267 (vec_cmpu<IMSA:mode><mode_i>): New expander.
9269 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
9272 * config/i386/i386-options.c (ix86_option_override_internal):
9273 Don't enable LZCNT/POPCNT if they have been disabled explicitly.
9275 2021-07-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
9278 * config/arm/arm_neon.h (vld1_p64): Replace call to builtin by
9279 explicitly dereferencing __a.
9280 (vld1_s64): Likewise.
9281 (vld1_u64): Likewise.
9282 * config/arm/arm_neon_builtins.def (vld1): Remove entry for di
9283 and change to VAR13.
9285 2021-07-30 Aldy Hernandez <aldyh@redhat.com>
9287 * gimple-loop-versioning.cc (lv_dom_walker::lv_dom_walker): Remove
9288 use of m_range_analyzer.
9289 (loop_versioning::lv_dom_walker::before_dom_children): Same.
9290 (loop_versioning::lv_dom_walker::after_dom_children): Remove.
9291 (loop_versioning::prune_loop_conditions): Replace vr_values use
9292 with range_query interface.
9293 (pass_loop_versioning::execute): Use ranger.
9295 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
9298 * ipa-devirt.c (ipa_odr_read_section): Compare the precision of
9299 enum values, and emit a warning if they mismatch.
9301 2021-07-30 Kewen Lin <linkw@linux.ibm.com>
9303 * cfgloop.h (as_const): New function.
9304 (class loop_iterator): Rename to ...
9305 (class loops_list): ... this.
9306 (loop_iterator::next): Rename to ...
9307 (loops_list::Iter::fill_curr_loop): ... this and adjust.
9308 (loop_iterator::loop_iterator): Rename to ...
9309 (loops_list::loops_list): ... this and adjust.
9310 (loops_list::Iter): New class.
9311 (loops_list::iterator): New type.
9312 (loops_list::const_iterator): New type.
9313 (loops_list::begin): New function.
9314 (loops_list::end): Likewise.
9315 (loops_list::begin const): Likewise.
9316 (loops_list::end const): Likewise.
9317 (FOR_EACH_LOOP): Remove.
9318 (FOR_EACH_LOOP_FN): Remove.
9319 * cfgloop.c (flow_loops_dump): Adjust FOR_EACH_LOOP* with range-based
9320 for loop with loops_list instance.
9321 (sort_sibling_loops): Likewise.
9322 (disambiguate_loops_with_multiple_latches): Likewise.
9323 (verify_loop_structure): Likewise.
9324 * cfgloopmanip.c (create_preheaders): Likewise.
9325 (force_single_succ_latches): Likewise.
9326 * config/aarch64/falkor-tag-collision-avoidance.c
9327 (execute_tag_collision_avoidance): Likewise.
9328 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Likewise.
9329 * config/s390/s390.c (s390_adjust_loops): Likewise.
9330 * doc/loop.texi: Likewise.
9331 * gimple-loop-interchange.cc (pass_linterchange::execute): Likewise.
9332 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
9333 * gimple-loop-versioning.cc (loop_versioning::analyze_blocks): Likewise.
9334 (loop_versioning::make_versioning_decisions): Likewise.
9335 * gimple-ssa-split-paths.c (split_paths): Likewise.
9336 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
9337 * graphite.c (canonicalize_loop_form): Likewise.
9338 (graphite_transform_loops): Likewise.
9339 * ipa-fnsummary.c (analyze_function_body): Likewise.
9340 * ipa-pure-const.c (analyze_function): Likewise.
9341 * loop-doloop.c (doloop_optimize_loops): Likewise.
9342 * loop-init.c (loop_optimizer_finalize): Likewise.
9343 (fix_loop_structure): Likewise.
9344 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
9345 (move_loop_invariants): Likewise.
9346 * loop-unroll.c (decide_unrolling): Likewise.
9347 (unroll_loops): Likewise.
9348 * modulo-sched.c (sms_schedule): Likewise.
9349 * predict.c (predict_loops): Likewise.
9350 (pass_profile::execute): Likewise.
9351 * profile.c (branch_prob): Likewise.
9352 * sel-sched-ir.c (sel_finish_pipelining): Likewise.
9353 (sel_find_rgns): Likewise.
9354 * tree-cfg.c (replace_loop_annotate): Likewise.
9355 (replace_uses_by): Likewise.
9356 (move_sese_region_to_fn): Likewise.
9357 * tree-if-conv.c (pass_if_conversion::execute): Likewise.
9358 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
9359 * tree-parloops.c (parallelize_loops): Likewise.
9360 * tree-predcom.c (tree_predictive_commoning): Likewise.
9361 * tree-scalar-evolution.c (scev_initialize): Likewise.
9362 (scev_reset): Likewise.
9363 * tree-ssa-dce.c (find_obviously_necessary_stmts): Likewise.
9364 * tree-ssa-live.c (remove_unused_locals): Likewise.
9365 * tree-ssa-loop-ch.c (ch_base::copy_headers): Likewise.
9366 * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
9367 (tree_ssa_lim_initialize): Likewise.
9368 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Likewise.
9369 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize): Likewise.
9370 * tree-ssa-loop-manip.c (get_loops_exits): Likewise.
9371 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations): Likewise.
9372 (free_numbers_of_iterations_estimates): Likewise.
9373 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Likewise.
9374 * tree-ssa-loop-split.c (tree_ssa_split_loops): Likewise.
9375 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Likewise.
9376 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
9377 (pass_scev_cprop::execute): Likewise.
9378 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
9379 * tree-ssa-sccvn.c (do_rpo_vn): Likewise.
9380 * tree-ssa-threadupdate.c
9381 (jump_thread_path_registry::thread_through_all_blocks): Likewise.
9382 * tree-vectorizer.c (vectorize_loops): Likewise.
9383 * tree-vrp.c (vrp_asserts::find_assert_locations): Likewise.
9385 2021-07-29 Hans-Peter Nilsson <hp@bitrange.com>
9387 * config/mmix/mmix.c (mmix_function_arg_1): Avoid
9388 generating a VOIDmode register for e.g the
9389 function_arg_info::end_marker.
9391 2021-07-29 Jeff Law <jeffreyalaw@gmail.com>
9393 * config/h8300/h8300-modes.def: Add CCZ, CCV and CCC, drop CCZNV.
9394 * config/h8300/h8300.md (H8cc mode iterator): Add CCZ.
9395 (cc mode_attr): Similarly.
9396 (ccz subst_attr): Similarly.
9397 * config/h8300/jumpcall.md: Add new patterns for branch-on-bit.
9398 * config/h8300/testcompare.md: Remove various cc0 based patterns
9399 that had been commented out. Add pattern to set CCZ from a bit
9402 2021-07-29 Thomas Schwinge <thomas@codesourcery.com>
9403 Julian Brown <julian@codesourcery.com>
9404 Kwok Cheung Yeung <kcy@codesourcery.com>
9406 * omp-offload.c (oacc_loop_xform_head_tail, oacc_loop_process):
9407 'update_stmt' after modification.
9408 (pass_oacc_loop_designation): New function, extracted out of...
9409 (pass_oacc_device_lower): ... this.
9410 (pass_data_oacc_loop_designation, pass_oacc_loop_designation)
9411 (make_pass_oacc_loop_designation): New
9412 * passes.def: Add it.
9413 * tree-parloops.c (create_parallel_loop): Adjust.
9414 * tree-pass.h (make_pass_oacc_loop_designation): New.
9416 2021-07-29 Aldy Hernandez <aldyh@redhat.com>
9418 * flag-types.h (enum threader_mode): New.
9419 * params.opt: Add entry for --param=threader-mode.
9420 * tree-ssa-threadbackward.c (THREADER_ITERATIVE_MODE): New.
9421 (class back_threader): New.
9422 (back_threader::back_threader): New.
9423 (back_threader::~back_threader): New.
9424 (back_threader::maybe_register_path): New.
9425 (back_threader::find_taken_edge): New.
9426 (back_threader::find_taken_edge_switch): New.
9427 (back_threader::find_taken_edge_cond): New.
9428 (back_threader::resolve_def): New.
9429 (back_threader::resolve_phi): New.
9430 (back_threader::find_paths_to_names): New.
9431 (back_threader::find_paths): New.
9434 (thread_jumps::find_jump_threads_backwards): Call ranger threader.
9435 (thread_jumps::find_jump_threads_backwards_with_ranger): New.
9436 (pass_thread_jumps::execute): Abstract out code...
9437 (try_thread_blocks): ...here.
9438 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
9439 Abstract out threading candidate code to...
9440 (single_succ_to_potentially_threadable_block): ...here.
9441 * tree-ssa-threadedge.h (single_succ_to_potentially_threadable_block):
9443 * tree-ssa-threadupdate.c (register_jump_thread): Return boolean.
9444 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
9445 Return bool from register_jump_thread.
9447 2021-07-29 Andreas Krebbel <krebbel@linux.ibm.com>
9449 * target.def: in0 and in1 do not need to be registers.
9450 * doc/tm.texi: Regenerate.
9452 2021-07-29 liuhongt <hongtao.liu@intel.com>
9455 * config/i386/i386.c (ix86_widen_mult_cost): New function.
9456 (ix86_add_stmt_cost): Use ix86_widen_mult_cost for
9459 2021-07-29 Jiufu Guo <guojiufu@linux.ibm.com>
9462 * config/rs6000/rs6000.c (TARGET_PREFERRED_DOLOOP_MODE): New hook.
9463 (rs6000_preferred_doloop_mode): New hook.
9464 * doc/tm.texi: Regenerate.
9465 * doc/tm.texi.in: Add hook preferred_doloop_mode.
9466 * target.def (preferred_doloop_mode): New hook.
9467 * targhooks.c (default_preferred_doloop_mode): New hook.
9468 * targhooks.h (default_preferred_doloop_mode): New hook.
9469 * tree-ssa-loop-ivopts.c (compute_doloop_base_on_mode): New function.
9470 (add_iv_candidate_for_doloop): Call targetm.preferred_doloop_mode
9471 and compute_doloop_base_on_mode.
9473 2021-07-28 Martin Sebor <msebor@redhat.com>
9475 PR middle-end/101494
9476 * tree-ssa-uninit.c (maybe_warn_operand): Correct object offset
9477 and size computation.
9479 2021-07-28 Martin Sebor <msebor@redhat.com>
9481 PR middle-end/101601
9482 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Remove
9484 Handle pointers to functions.
9486 2021-07-28 Martin Sebor <msebor@redhat.com>
9488 * Makefile.in (OBJS): Add gimple-ssa-warn-access.o and pointer-query.o.
9489 * attribs.h (fndecl_dealloc_argno): Move fndecl_dealloc_argno to tree.h.
9490 * builtins.c (compute_objsize_r): Move to pointer-query.cc.
9491 (access_ref::access_ref): Same.
9492 (access_ref::phi): Same.
9493 (access_ref::get_ref): Same.
9494 (access_ref::size_remaining): Same.
9495 (access_ref::offset_in_range): Same.
9496 (access_ref::add_offset): Same.
9497 (access_ref::inform_access): Same.
9498 (ssa_name_limit_t::visit_phi): Same.
9499 (ssa_name_limit_t::leave_phi): Same.
9500 (ssa_name_limit_t::next): Same.
9501 (ssa_name_limit_t::next_phi): Same.
9502 (ssa_name_limit_t::~ssa_name_limit_t): Same.
9503 (pointer_query::pointer_query): Same.
9504 (pointer_query::get_ref): Same.
9505 (pointer_query::put_ref): Same.
9506 (pointer_query::flush_cache): Same.
9507 (warn_string_no_nul): Move to gimple-ssa-warn-access.cc.
9508 (check_nul_terminated_array): Same.
9509 (unterminated_array): Same.
9510 (maybe_warn_for_bound): Same.
9511 (check_read_access): Same.
9512 (warn_for_access): Same.
9513 (get_size_range): Same.
9514 (check_access): Same.
9515 (gimple_call_alloc_size): Move to tree.c.
9516 (gimple_parm_array_size): Move to pointer-query.cc.
9517 (get_offset_range): Same.
9518 (gimple_call_return_array): Same.
9519 (handle_min_max_size): Same.
9520 (handle_array_ref): Same.
9521 (handle_mem_ref): Same.
9522 (compute_objsize): Same.
9523 (gimple_call_alloc_p): Move to gimple-ssa-warn-access.cc.
9524 (call_dealloc_argno): Same.
9525 (fndecl_dealloc_argno): Same.
9526 (new_delete_mismatch_p): Same.
9527 (matching_alloc_calls_p): Same.
9528 (warn_dealloc_offset): Same.
9529 (maybe_emit_free_warning): Same.
9530 * builtins.h (check_nul_terminated_array): Move to
9531 gimple-ssa-warn-access.h.
9532 (check_nul_terminated_array): Same.
9533 (warn_string_no_nul): Same.
9534 (unterminated_array): Same.
9535 (class ssa_name_limit_t): Same.
9536 (class pointer_query): Same.
9537 (struct access_ref): Same.
9538 (class range_query): Same.
9539 (struct access_data): Same.
9540 (gimple_call_alloc_size): Same.
9541 (gimple_parm_array_size): Same.
9542 (compute_objsize): Same.
9543 (class access_data): Same.
9544 (maybe_emit_free_warning): Same.
9545 * calls.c (initialize_argument_information): Remove call to
9546 maybe_emit_free_warning.
9547 * gimple-array-bounds.cc: Include new header..
9548 * gimple-fold.c: Same.
9549 * gimple-ssa-sprintf.c: Same.
9550 * gimple-ssa-warn-restrict.c: Same.
9551 * passes.def: Add pass_warn_access.
9552 * tree-pass.h (make_pass_warn_access): Declare.
9553 * tree-ssa-strlen.c: Include new headers.
9554 * tree.c (fndecl_dealloc_argno): Move here from builtins.c.
9555 * tree.h (fndecl_dealloc_argno): Move here from attribs.h.
9556 * gimple-ssa-warn-access.cc: New file.
9557 * gimple-ssa-warn-access.h: New file.
9558 * pointer-query.cc: New file.
9559 * pointer-query.h: New file.
9561 2021-07-28 Jakub Jelinek <jakub@redhat.com>
9563 PR middle-end/101624
9564 * ubsan.c (maybe_instrument_pointer_overflow,
9565 instrument_object_size): Only test DECL_REGISTER on VAR_DECLs,
9566 PARM_DECLs or RESULT_DECLs.
9567 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
9569 2021-07-28 Jakub Jelinek <jakub@redhat.com>
9571 PR middle-end/101642
9572 * match.pd (bswap16 (x) == bswap16 (y)): Cast both operands
9573 to type of bswap16 for comparison.
9574 (bswap16 (x) == cst): Cast bswap16 operand to type of cst.
9576 2021-07-28 Richard Biener <rguenther@suse.de>
9578 PR tree-optimization/101615
9579 * tree-vect-slp.c (vect_optimize_slp): Materialize permutes
9580 at CTOR SLP graph entries.
9582 2021-07-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9584 * config/aarch64/aarch64.md (*extend<SHORT:mode><GPI:mode>2_aarch64):
9585 Add "r,w" alternative.
9587 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
9590 * config/i386/i386.c (ix86_avx_u128_mode_needed): Don't set
9591 AVX_U128_DIRTY when all bits are zero.
9593 2021-07-28 Richard Biener <rguenther@suse.de>
9595 PR tree-optimization/101615
9596 * tree-vect-slp.c (vect_optimize_slp): Pre-existing vector
9597 external nodes cannot be permuted so make them perm_out 0.
9599 2021-07-28 Andrew Stubbs <ams@codesourcery.com>
9602 * config.in: Regenerate.
9603 * config/gcn/gcn-hsa.h (A_FIJI): New define.
9604 (A_900): New define.
9605 (A_906): New define.
9606 (A_908): New define.
9607 (ASM_SPEC): Use A_FIJI, A_900, A_906 and A_908.
9608 * config/gcn/gcn.c (output_file_start): Adjust attributes according
9609 to the assembler capabilities.
9610 * config/gcn/mkoffload.c (main): Likewise.
9611 * configure: Regenerate.
9612 * configure.ac: Add tests for LLVM assembler attribute features.
9614 2021-07-28 Andrew MacLeod <amacleod@redhat.com>
9616 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Check for
9617 cond_false and cond_true on branches.
9619 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
9621 * config/aarch64/aarch64.c (aarch64_gen_adjusted_ldpstp): use
9624 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
9626 * alias.c (init_alias_analysis): Don't skip prologue/epilogue.
9628 2021-07-28 Jakub Jelinek <jakub@redhat.com>
9631 * config/i386/sse.md (vashr<mode>3): Split into vashrv8di3 expander
9632 and vashrv4di3 expander, where the latter requires just TARGET_AVX2
9633 and has special !TARGET_AVX512VL expansion.
9634 (vashrv2di3<mask_name>): Rename to ...
9635 (vashrv2di3): ... this. Change condition to TARGET_XOP || TARGET_AVX2
9636 and add special !TARGET_XOP && !TARGET_AVX512VL expansion.
9638 2021-07-28 Martin Uecker <muecker@gwdg.de>
9640 * calls.c (maybe_warn_rdwr_sizes): Correct argument
9641 numbers in warning that were switched.
9643 2021-07-28 Kewen Lin <linkw@linux.ibm.com>
9645 PR tree-optimization/101596
9646 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Fix wrong check
9647 by using new_type's precision instead.
9649 2021-07-28 liuhongt <hongtao.liu@intel.com>
9652 * config/i386/i386.h (processor_costs): Add new member
9654 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
9655 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
9656 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
9657 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
9658 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
9659 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
9660 generic_cost, core_cost): Initialize integer_to_sse same value
9662 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
9663 * config/i386/i386.c (ix86_builtin_vectorization_cost):
9664 Use integer_to_sse instead of sse_op to calculate the cost of
9667 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
9669 * config/rs6000/rs6000-gen-builtins.c (write_ovld_static_init): New
9671 (write_init_file): Call write_ovld_static_init.
9673 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
9675 * config/rs6000/rs6000-gen-builtins.c (write_bif_static_init): New
9677 (write_init_file): Call write_bif_static_init.
9679 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
9681 * config/rs6000/rs6000-gen-builtins.c (typemap): New struct.
9682 (TYPE_MAP_SIZE): New macro.
9683 (type_map): New initialized variable.
9684 (typemap_cmp): New function.
9685 (write_type_node): Likewise.
9686 (write_fntype_init): Implement.
9688 2021-07-27 Martin Sebor <msebor@redhat.com>
9690 PR tree-optimization/101584
9691 * tree-ssa-uninit.c (builtin_call_nomodifying_p): New function.
9692 (check_defs): Call it.
9694 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
9696 * tree-ssa-dom.c (dom_jump_threader_simplifier):
9697 Put avail_exprs_stack in the class, instead of passing it to
9698 jump_threader_simplifier.
9699 (dom_jump_threader_simplifier::simplify): Add state argument.
9700 (dom_opt_dom_walker): Add state.
9701 (pass_dominator::execute): Pass state to threader.
9702 (dom_opt_dom_walker::before_dom_children): Use state.
9703 * tree-ssa-threadedge.c (jump_threader::jump_threader): Replace
9705 (jump_threader::record_temporary_equivalences_from_phis):
9706 Register equivalences through the state variable.
9707 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
9708 Record ranges in a statement through the state variable.
9709 (jump_threader::simplify_control_stmt_condition): Pass state to
9711 (jump_threader::simplify_control_stmt_condition_1): Same.
9712 (jump_threader::thread_around_empty_blocks): Remove obsolete
9714 (jump_threader::thread_through_normal_block): Record equivalences
9715 on edge through the state variable.
9716 (jump_threader::thread_across_edge): Abstract state pushing.
9717 (jt_state::jt_state): New.
9718 (jt_state::push): New.
9719 (jt_state::pop): New.
9720 (jt_state::register_equiv): New.
9721 (jt_state::record_ranges_from_stmt): New.
9722 (jt_state::register_equivs_on_edge): New.
9723 (jump_threader_simplifier::jump_threader_simplifier): Move from
9725 (jump_threader_simplifier::simplify): Add state argument.
9726 * tree-ssa-threadedge.h (class jt_state): New.
9727 (class jump_threader): Add state to constructor.
9728 (class jump_threader_simplifier): Add state to simplify. Remove
9729 avail_exprs_stack from class.
9730 * tree-vrp.c (vrp_jump_threader_simplifier::simplify): Add state
9732 (vrp_jump_threader::vrp_jump_threader): Add state.
9733 (vrp_jump_threader::~vrp_jump_threader): Cleanup state.
9735 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
9737 * Makefile.in (OBJS): Add gimple-range-path.o.
9738 * gimple-range-path.cc: New file.
9739 * gimple-range-path.h: New file.
9741 2021-07-27 Jonathan Wright <jonathan.wright@arm.com>
9743 * config/aarch64/aarch64-simd.md: Push sign/zero-extension
9744 inside vec_duplicate for all patterns.
9745 * simplify-rtx.c (simplify_context::simplify_unary_operation_1):
9746 Push sign/zero-extension inside vec_duplicate.
9748 2021-07-27 Richard Biener <rguenther@suse.de>
9750 PR tree-optimization/101573
9751 * tree-ssa-uninit.c (warn_uninit_phi_uses): New function
9752 looking at uninitialized PHI arg defs in some constrained cases.
9753 (warn_uninitialized_vars): Call it.
9754 (execute_early_warn_uninitialized): Calculate dominators.
9756 2021-07-27 Richard Biener <rguenther@suse.de>
9758 PR tree-optimization/39821
9759 * tree-vect-stmts.c (vect_model_promotion_demotion_cost): Use
9760 vector_stmt for widening arithmetic.
9761 (vectorizable_conversion): Adjust.
9763 2021-07-27 Martin Jambor <mjambor@suse.cz>
9765 * cgraph.h (ipa_replace_map): New field force_load_ref.
9766 * ipa-prop.h (ipa_param_descriptor): Reduce precision of move_cost,
9767 aded new flag load_dereferenced, adjusted comments.
9768 (ipa_get_param_dereferenced): New function.
9769 (ipa_set_param_dereferenced): Likewise.
9770 * cgraphclones.c (cgraph_node::create_virtual_clone): Follow it.
9771 * ipa-cp.c: Include gimple.h.
9772 (ipcp_discover_new_direct_edges): Take into account dereferenced flag.
9773 (get_replacement_map): New parameter force_load_ref, set the
9774 appropriate flag in ipa_replace_map if set.
9775 (struct symbol_and_index_together): New type.
9776 (adjust_refs_in_act_callers): New function.
9777 (adjust_references_in_caller): Likewise.
9778 (create_specialized_node): When appropriate, call
9779 adjust_references_in_caller and force only load references.
9780 * ipa-prop.c (load_from_dereferenced_name): New function.
9781 (ipa_analyze_controlled_uses): Also detect loads from a
9782 dereference, harden testing of call statements.
9783 (ipa_write_node_info): Stream the dereferenced flag.
9784 (ipa_read_node_info): Likewise.
9785 (ipa_set_jf_constant): Also create refdesc when jump function
9786 references a variable.
9787 (cgraph_node_for_jfunc): Rename to symtab_node_for_jfunc, work
9788 also on references of variables and return a symtab_node. Adjust
9790 (propagate_controlled_uses): Also remove references to VAR_DECLs.
9792 2021-07-27 Jakub Jelinek <jakub@redhat.com>
9794 PR middle-end/101586
9795 * gimple-fold.c (clear_padding_type): Ignore FIELD_DECLs with byte
9796 positions above or equal to sz except for diagnostics of flexible
9799 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
9801 PR tree-optimization/78888
9802 * gimple-range-fold.cc (get_letter_range): New.
9803 (fold_using_range::range_of_builtin_call): Call get_letter_range.
9805 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
9807 PR tree-optimization/78888
9808 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Add cases
9809 for CFN_BUILT_IN_TOUPPER and CFN_BUILT_IN_TOLOWER.
9811 2021-07-26 Roger Sayle <roger@nextmovesoftware.com>
9812 Marc Glisse <marc.glisse@inria.fr>
9814 * match.pd (rotate): Simplify equality/inequality of rotations.
9815 (bswap): Simplify equality/inequality tests of byte swapping.
9817 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
9819 * range-op.cc (operator_bitwise_xor::op1_op2_relation_effect):
9822 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
9824 * range-op.cc (operator_lshift::fold_range): Pass rel to
9825 base class fold_range.
9826 (operator_rshift::fold_range): Same.
9828 2021-07-26 Ashimida <ashimida@linux.alibaba.com>
9831 * toplev.h (min_align_loops_log): Remove declaration.
9832 (min_align_jumps_log, min_align_labels_log): Likewise.
9833 (min_align_functions_log): Likewise.
9835 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
9837 * tree-vrp.c (vrp_simplify_cond_using_ranges): Rename vr_values
9839 (execute_vrp): Abstract out simplification of conditionals...
9840 (simplify_casted_conds): ...here.
9842 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
9844 * gimple-array-bounds.cc (array_bounds_checker::get_value_range):
9845 Add gimple argument.
9846 (array_bounds_checker::check_array_ref): Same.
9847 (array_bounds_checker::check_addr_expr): Same.
9848 (array_bounds_checker::check_array_bounds): Pass statement to
9849 check_array_bounds and check_addr_expr.
9850 * gimple-array-bounds.h (check_array_bounds): Add gimple argument.
9851 (check_addr_expr): Same.
9852 (get_value_range): Same.
9854 2021-07-26 Tamar Christina <tamar.christina@arm.com>
9856 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): Rename to..
9857 (sdot_prod, udot_prod): ... This.
9858 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): Merged
9860 (<sur>dot_prod<vsi2qi>): ... this.
9861 (aarch64_<sur>dot_lane<vsi2qi>, aarch64_<sur>dot_laneq<vsi2qi>):
9862 Change operands order.
9863 (<sur>sadv16qi): Use new operands order.
9864 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32,
9865 vdotq_s32): Use new RTL ordering.
9867 2021-07-26 Tamar Christina <tamar.christina@arm.com>
9869 * config/aarch64/aarch64-builtins.c (TYPES_TERNOP_SUSS,
9870 aarch64_types_ternop_suss_qualifiers): New.
9871 * config/aarch64/aarch64-simd-builtins.def (usdot_prod): Use it.
9872 * config/aarch64/aarch64-simd.md (usdot_prod<vsi2qi>): Re-organize RTL.
9873 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Use it.
9875 2021-07-23 Jakub Jelinek <jakub@redhat.com>
9877 PR rtl-optimization/101562
9878 * expmed.c (store_integral_bit_field): Only use movstrict_optab
9879 if the operand isn't paradoxical.
9881 2021-07-23 Aldy Hernandez <aldyh@redhat.com>
9883 * gimple-array-bounds.h (class array_bounds_checker): Change
9884 ranges type to range_query.
9886 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9888 * config/aarch64/arm_neon.h (vst1_s64_x2): Use
9889 __builtin_memcpy instead of constructing
9890 __builtin_aarch64_simd_oi one vector at a time.
9891 (vst1_u64_x2): Likewise.
9892 (vst1_f64_x2): Likewise.
9893 (vst1_s8_x2): Likewise.
9894 (vst1_p8_x2): Likewise.
9895 (vst1_s16_x2): Likewise.
9896 (vst1_p16_x2): Likewise.
9897 (vst1_s32_x2): Likewise.
9898 (vst1_u8_x2): Likewise.
9899 (vst1_u16_x2): Likewise.
9900 (vst1_u32_x2): Likewise.
9901 (vst1_f16_x2): Likewise.
9902 (vst1_f32_x2): Likewise.
9903 (vst1_p64_x2): Likewise.
9904 (vst1q_s8_x2): Likewise.
9905 (vst1q_p8_x2): Likewise.
9906 (vst1q_s16_x2): Likewise.
9907 (vst1q_p16_x2): Likewise.
9908 (vst1q_s32_x2): Likewise.
9909 (vst1q_s64_x2): Likewise.
9910 (vst1q_u8_x2): Likewise.
9911 (vst1q_u16_x2): Likewise.
9912 (vst1q_u32_x2): Likewise.
9913 (vst1q_u64_x2): Likewise.
9914 (vst1q_f16_x2): Likewise.
9915 (vst1q_f32_x2): Likewise.
9916 (vst1q_f64_x2): Likewise.
9917 (vst1q_p64_x2): Likewise.
9919 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9921 * config/aarch64/arm_neon.h (vst1_s64_x3): Use
9922 __builtin_memcpy instead of constructing
9923 __builtin_aarch64_simd_ci one vector at a time.
9924 (vst1_u64_x3): Likewise.
9925 (vst1_f64_x3): Likewise.
9926 (vst1_s8_x3): Likewise.
9927 (vst1_p8_x3): Likewise.
9928 (vst1_s16_x3): Likewise.
9929 (vst1_p16_x3): Likewise.
9930 (vst1_s32_x3): Likewise.
9931 (vst1_u8_x3): Likewise.
9932 (vst1_u16_x3): Likewise.
9933 (vst1_u32_x3): Likewise.
9934 (vst1_f16_x3): Likewise.
9935 (vst1_f32_x3): Likewise.
9936 (vst1_p64_x3): Likewise.
9937 (vst1q_s8_x3): Likewise.
9938 (vst1q_p8_x3): Likewise.
9939 (vst1q_s16_x3): Likewise.
9940 (vst1q_p16_x3): Likewise.
9941 (vst1q_s32_x3): Likewise.
9942 (vst1q_s64_x3): Likewise.
9943 (vst1q_u8_x3): Likewise.
9944 (vst1q_u16_x3): Likewise.
9945 (vst1q_u32_x3): Likewise.
9946 (vst1q_u64_x3): Likewise.
9947 (vst1q_f16_x3): Likewise.
9948 (vst1q_f32_x3): Likewise.
9949 (vst1q_f64_x3): Likewise.
9950 (vst1q_p64_x3): Likewise.
9952 2021-07-23 H.J. Lu <hjl.tools@gmail.com>
9955 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): Don't return
9956 hard register when LRA is in progress.
9958 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9960 * config/aarch64/arm_neon.h (vst1_s8_x4): Use
9961 __builtin_memcpy instead of using a union.
9962 (vst1q_s8_x4): Likewise.
9963 (vst1_s16_x4): Likewise.
9964 (vst1q_s16_x4): Likewise.
9965 (vst1_s32_x4): Likewise.
9966 (vst1q_s32_x4): Likewise.
9967 (vst1_u8_x4): Likewise.
9968 (vst1q_u8_x4): Likewise.
9969 (vst1_u16_x4): Likewise.
9970 (vst1q_u16_x4): Likewise.
9971 (vst1_u32_x4): Likewise.
9972 (vst1q_u32_x4): Likewise.
9973 (vst1_f16_x4): Likewise.
9974 (vst1q_f16_x4): Likewise.
9975 (vst1_f32_x4): Likewise.
9976 (vst1q_f32_x4): Likewise.
9977 (vst1_p8_x4): Likewise.
9978 (vst1q_p8_x4): Likewise.
9979 (vst1_p16_x4): Likewise.
9980 (vst1q_p16_x4): Likewise.
9981 (vst1_s64_x4): Likewise.
9982 (vst1_u64_x4): Likewise.
9983 (vst1_p64_x4): Likewise.
9984 (vst1q_s64_x4): Likewise.
9985 (vst1q_u64_x4): Likewise.
9986 (vst1q_p64_x4): Likewise.
9987 (vst1_f64_x4): Likewise.
9988 (vst1q_f64_x4): Likewise.
9990 2021-07-23 Jonathan Wrightt <jonathan.wright@arm.com>
9992 * config/aarch64/arm_neon.h (vst2_s64): Use __builtin_memcpy
9993 instead of constructing __builtin_aarch64_simd_oi one vector
9995 (vst2_u64): Likewise.
9996 (vst2_f64): Likewise.
9997 (vst2_s8): Likewise.
9998 (vst2_p8): Likewise.
9999 (vst2_s16): Likewise.
10000 (vst2_p16): Likewise.
10001 (vst2_s32): Likewise.
10002 (vst2_u8): Likewise.
10003 (vst2_u16): Likewise.
10004 (vst2_u32): Likewise.
10005 (vst2_f16): Likewise.
10006 (vst2_f32): Likewise.
10007 (vst2_p64): Likewise.
10008 (vst2q_s8): Likewise.
10009 (vst2q_p8): Likewise.
10010 (vst2q_s16): Likewise.
10011 (vst2q_p16): Likewise.
10012 (vst2q_s32): Likewise.
10013 (vst2q_s64): Likewise.
10014 (vst2q_u8): Likewise.
10015 (vst2q_u16): Likewise.
10016 (vst2q_u32): Likewise.
10017 (vst2q_u64): Likewise.
10018 (vst2q_f16): Likewise.
10019 (vst2q_f32): Likewise.
10020 (vst2q_f64): Likewise.
10021 (vst2q_p64): Likewise.
10023 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
10025 * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy
10026 instead of constructing __builtin_aarch64_simd_ci one vector
10028 (vst3_u64): Likewise.
10029 (vst3_f64): Likewise.
10030 (vst3_s8): Likewise.
10031 (vst3_p8): Likewise.
10032 (vst3_s16): Likewise.
10033 (vst3_p16): Likewise.
10034 (vst3_s32): Likewise.
10035 (vst3_u8): Likewise.
10036 (vst3_u16): Likewise.
10037 (vst3_u32): Likewise.
10038 (vst3_f16): Likewise.
10039 (vst3_f32): Likewise.
10040 (vst3_p64): Likewise.
10041 (vst3q_s8): Likewise.
10042 (vst3q_p8): Likewise.
10043 (vst3q_s16): Likewise.
10044 (vst3q_p16): Likewise.
10045 (vst3q_s32): Likewise.
10046 (vst3q_s64): Likewise.
10047 (vst3q_u8): Likewise.
10048 (vst3q_u16): Likewise.
10049 (vst3q_u32): Likewise.
10050 (vst3q_u64): Likewise.
10051 (vst3q_f16): Likewise.
10052 (vst3q_f32): Likewise.
10053 (vst3q_f64): Likewise.
10054 (vst3q_p64): Likewise.
10056 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
10058 * config/aarch64/arm_neon.h (vst4_s64): Use __builtin_memcpy
10059 instead of constructing __builtin_aarch64_simd_xi one vector
10061 (vst4_u64): Likewise.
10062 (vst4_f64): Likewise.
10063 (vst4_s8): Likewise.
10064 (vst4_p8): Likewise.
10065 (vst4_s16): Likewise.
10066 (vst4_p16): Likewise.
10067 (vst4_s32): Likewise.
10068 (vst4_u8): Likewise.
10069 (vst4_u16): Likewise.
10070 (vst4_u32): Likewise.
10071 (vst4_f16): Likewise.
10072 (vst4_f32): Likewise.
10073 (vst4_p64): Likewise.
10074 (vst4q_s8): Likewise.
10075 (vst4q_p8): Likewise.
10076 (vst4q_s16): Likewise.
10077 (vst4q_p16): Likewise.
10078 (vst4q_s32): Likewise.
10079 (vst4q_s64): Likewise.
10080 (vst4q_u8): Likewise.
10081 (vst4q_u16): Likewise.
10082 (vst4q_u32): Likewise.
10083 (vst4q_u64): Likewise.
10084 (vst4q_f16): Likewise.
10085 (vst4q_f32): Likewise.
10086 (vst4q_f64): Likewise.
10087 (vst4q_p64): Likewise.
10089 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
10091 * config/aarch64/arm_neon.h (vtbx4_s8): Use __builtin_memcpy
10092 instead of constructing __builtin_aarch64_simd_oi one vector
10094 (vtbx4_u8): Likewise.
10095 (vtbx4_p8): Likewise.
10097 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
10099 * config/aarch64/arm_neon.h (vtbl3_s8): Use __builtin_memcpy
10100 instead of constructing __builtin_aarch64_simd_oi one vector
10102 (vtbl3_u8): Likewise.
10103 (vtbl3_p8): Likewise.
10104 (vtbl4_s8): Likewise.
10105 (vtbl4_u8): Likewise.
10106 (vtbl4_p8): Likewise.
10108 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
10110 * config/aarch64/arm_neon.h (vqtbx2_s8): Use __builtin_memcpy
10111 instead of constructing __builtin_aarch64_simd_oi one vector
10113 (vqtbx2_u8): Likewise.
10114 (vqtbx2_p8): Likewise.
10115 (vqtbx2q_s8): Likewise.
10116 (vqtbx2q_u8): Likewise.
10117 (vqtbx2q_p8): Likewise.
10118 (vqtbx3_s8): Use __builtin_memcpy instead of constructing
10119 __builtin_aarch64_simd_ci one vector at a time.
10120 (vqtbx3_u8): Likewise.
10121 (vqtbx3_p8): Likewise.
10122 (vqtbx3q_s8): Likewise.
10123 (vqtbx3q_u8): Likewise.
10124 (vqtbx3q_p8): Likewise.
10125 (vqtbx4_s8): Use __builtin_memcpy instead of constructing
10126 __builtin_aarch64_simd_xi one vector at a time.
10127 (vqtbx4_u8): Likewise.
10128 (vqtbx4_p8): Likewise.
10129 (vqtbx4q_s8): Likewise.
10130 (vqtbx4q_u8): Likewise.
10131 (vqtbx4q_p8): Likewise.
10133 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
10135 * config/aarch64/arm_neon.h (vqtbl2_s8): Use __builtin_memcpy
10136 instead of constructing __builtin_aarch64_simd_oi one vector
10138 (vqtbl2_u8): Likewise.
10139 (vqtbl2_p8): Likewise.
10140 (vqtbl2q_s8): Likewise.
10141 (vqtbl2q_u8): Likewise.
10142 (vqtbl2q_p8): Likewise.
10143 (vqtbl3_s8): Use __builtin_memcpy instead of constructing
10144 __builtin_aarch64_simd_ci one vector at a time.
10145 (vqtbl3_u8): Likewise.
10146 (vqtbl3_p8): Likewise.
10147 (vqtbl3q_s8): Likewise.
10148 (vqtbl3q_u8): Likewise.
10149 (vqtbl3q_p8): Likewise.
10150 (vqtbl4_s8): Use __builtin_memcpy instead of constructing
10151 __builtin_aarch64_simd_xi one vector at a time.
10152 (vqtbl4_u8): Likewise.
10153 (vqtbl4_p8): Likewise.
10154 (vqtbl4q_s8): Likewise.
10155 (vqtbl4q_u8): Likewise.
10156 (vqtbl4q_p8): Likewise.
10158 2021-07-23 Haochen Gui <guihaoc@gcc.gnu.org>
10161 * config/rs6000/rs6000.md (cstore<mode>4): Fix wrong fall through.
10163 2021-07-22 Andrew Pinski <apinski@marvell.com>
10165 PR tree-optimization/10153
10166 * tree-tailcall.c (create_tailcall_accumulator):
10167 Don't call fold_convert as the type should be correct already.
10168 (tree_optimize_tail_calls_1): Use build_{one,zero}_cst instead
10169 of integer_{one,zero}_node for the call of create_tailcall_accumulator.
10171 2021-07-22 Aldy Hernandez <aldyh@redhat.com>
10173 * gimple-range-cache.cc (non_null_ref::adjust_range): Replace
10174 varying_p check for null/non-null check.
10176 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
10178 PR tree-optimization/101511
10179 * value-relation.cc (relation_oracle::query_relation): Check if ssa1
10180 is in ssa2's equiv set, and don't trap if so.
10182 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
10184 PR tree-optimization/101497
10185 * gimple-range-fold.cc (fold_using_range::range_of_cond_expr): Check
10188 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
10190 PR tree-optimization/101496
10191 * vr-values.c (simplify_using_ranges::fold_cond): Call range_of_stmt
10192 first, then vrp_visit_cond_Stmt.
10194 2021-07-22 liuhongt <hongtao.liu@intel.com>
10196 * config/i386/i386-expand.c
10197 (ix86_broadcast_from_integer_constant): Rename to ..
10198 (ix86_broadcast_from_constant): .. this, and extend it to
10200 (ix86_expand_vector_move): Extend to float mode.
10201 * config/i386/i386-features.c
10202 (replace_constant_pool_with_broadcast): Remove.
10203 (remove_partial_avx_dependency_gate): Ditto.
10204 (constant_pool_broadcast): Ditto.
10205 (class pass_constant_pool_broadcast): Ditto.
10206 (make_pass_constant_pool_broadcast): Ditto.
10207 (remove_partial_avx_dependency): Adjust gate.
10208 * config/i386/i386-passes.def: Remove pass_constant_pool_broadcast.
10209 * config/i386/i386-protos.h
10210 (make_pass_constant_pool_broadcast): Remove.
10212 2021-07-22 liuhongt <hongtao.liu@intel.com>
10214 * config/i386/constraints.md (Wb): New constraint.
10216 * config/i386/i386.md (*ashlhi3_1): Extend to avx512 mask
10218 (*ashlqi3_1): Ditto.
10219 (*<insn><mode>3_1): Split to ..
10220 (*ashr<mode>3_1): this, ...
10221 (*lshr<mode>3_1): and this, also extend this pattern to avx512
10223 (*<insn><mode>3_1): Split to ..
10224 (*ashr<mode>3_1): this, ...
10225 (*lshrqi3_1): and this, also extend this pattern to avx512
10227 (*lshrhi3_1): And this, also extend this pattern to avx512
10229 * config/i386/sse.md (k<code><mode>): New define_split after
10230 it to convert generic shift pattern to mask shift ones.
10232 2021-07-21 Thomas Schwinge <thomas@codesourcery.com>
10233 Joseph Myers <joseph@codesourcery.com>
10234 Cesar Philippidis <cesar@codesourcery.com>
10236 * tree-core.h (omp_clause_code): Add 'OMP_CLAUSE_NOHOST'.
10237 * tree.c (omp_clause_num_ops, omp_clause_code_name, walk_tree_1):
10239 * tree-pretty-print.c (dump_omp_clause): Likewise.
10240 * omp-general.c (oacc_verify_routine_clauses): Likewise.
10241 * gimplify.c (gimplify_scan_omp_clauses)
10242 (gimplify_adjust_omp_clauses): Likewise.
10243 * tree-nested.c (convert_nonlocal_omp_clauses)
10244 (convert_local_omp_clauses): Likewise.
10245 * omp-low.c (scan_sharing_clauses): Likewise.
10246 * omp-offload.c (execute_oacc_device_lower): Update.
10248 2021-07-21 Martin Sebor <msebor@redhat.com>
10250 * tree-ssa-alias.c (walk_aliased_vdefs_1): Fix typos in a comment.
10252 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
10254 * config/rs6000/rs6000-gen-builtins.c (write_init_bif_table):
10257 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
10259 * config/rs6000/rs6000-gen-builtins.c (write_fntype): New
10261 (write_fntype_init): New stub function.
10262 (write_init_bif_table): Likewise.
10263 (write_init_ovld_table): New function.
10264 (write_init_file): Implement.
10266 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
10268 * config/rs6000/rs6000-gen-builtins.c
10269 (write_autogenerated_header): New function.
10270 (write_decls): Likewise.
10271 (write_extern_fntype): New callback function.
10272 (write_header_file): Implement.
10274 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
10276 * config/rs6000/rs6000-gen-builtins.c (write_defines_file):
10279 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
10281 * config/rs6000/rs6000-gen-builtins.c (complete_vector_type): New
10283 (complete_base_type): Likewise.
10284 (construct_fntype_id): Likewise.
10285 (parse_bif_entry): Call contruct_fntype_id.
10286 (parse_ovld_entry): Likewise.
10288 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
10290 * config/rs6000/rs6000-gen-builtins.c (ovld_stanza): New struct.
10291 (MAXOVLDSTANZAS): New macro.
10292 (ovld_stanzas): New variable.
10293 (curr_ovld_stanza): Likewise.
10294 (MAXOVLDS): New macro.
10295 (ovlddata): New struct.
10296 (ovlds): New variable.
10297 (curr_ovld): Likewise.
10298 (max_ovld_args): Likewise.
10299 (parse_ovld_entry): New function.
10300 (parse_ovld_stanza): Likewise.
10301 (parse_ovld): Implement.
10303 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
10305 * config/rs6000/rs6000-gen-builtins.c (parse_bif_attrs):
10308 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
10310 * config/rs6000/rs6000-gen-builtins.c (parse_args): New function.
10311 (parse_prototype): Implement.
10313 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
10315 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): New enum.
10316 (curr_bif_stanza): New variable.
10317 (stanza_entry): New struct.
10318 (stanza_map): New initialized variable.
10319 (enable_string): Likewise.
10320 (fnkinds): New enum.
10321 (typelist): New struct.
10322 (attrinfo): Likewise.
10323 (MAXRESTROPNDS): New macro.
10324 (prototype): New struct.
10325 (MAXBIFS): New macro.
10326 (bifdata): New struct.
10327 (bifs): New variable.
10328 (curr_bif): Likewise.
10329 (bif_order): Likewise.
10330 (bif_index): Likewise.
10331 (fatal): New function.
10332 (stanza_name_to_stanza): Likewise.
10333 (parse_bif_attrs): New stub function.
10334 (parse_prototype): Likewise.
10335 (parse_bif_entry): New function.
10336 (parse_bif_stanza): Likewise.
10337 (parse_bif): Implement.
10338 (set_bif_order): New function.
10339 (create_bif_order): Implement.
10341 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
10343 * config/rs6000/rs6000-gen-builtins.c (rbtree.h): New #include.
10344 (num_bifs): New variable.
10345 (num_ovld_stanzas): Likewise.
10346 (num_ovlds): Likewise.
10347 (parse_codes): New enum.
10348 (bif_rbt): New variable.
10349 (ovld_rbt): Likewise.
10350 (fntype_rbt): Likewise.
10351 (bifo_rbt): Likewise.
10352 (parse_bif): New stub function.
10353 (create_bif_order): Likewise.
10354 (parse_ovld): Likewise.
10355 (write_header_file): Likewise.
10356 (write_init_file): Likewise.
10357 (write_defines_file): Likewise.
10358 (delete_output_files): New function.
10361 2021-07-21 H.J. Lu <hjl.tools@gmail.com>
10364 * config/i386/i386-builtin.def: Remove OPTION_MASK_ISA_SSE4_2
10365 from CRC32 _builtin functions.
10367 2021-07-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
10369 * coverage.c (build_gcov_info_var_registration): Mark the object placed
10370 in the linker set as referenced so that it does not get optimized away.
10372 2021-07-21 Kito Cheng <kito.cheng@sifive.com>
10375 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
10377 * config.gcc (riscv*-*-*): Detect which python is available.
10379 2021-07-21 Jakub Jelinek <jakub@redhat.com>
10381 PR middle-end/101535
10382 * gimplify.c (omp_check_private): Properly skip ORT_TARGET_DATA
10383 contexts in which decl isn't privatized and for ORT_TARGET return
10384 false if decl is mapped.
10386 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
10388 * gimple-loop-jam.c: Include tree-ssa-sccvn.h.
10389 (tree_loop_unroll_and_jam): Run value-numbering on a loop that
10390 has been successfully unrolled.
10392 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
10394 * tree-ssa-loop-manip.c (determine_exit_conditions): Return a null
10395 exit condition if no tail loop is needed, and if the original exit
10396 condition should therefore be kept as-is.
10397 (tree_transform_and_unroll_loop): Handle that case here too.
10399 2021-07-21 Kewen Lin <linkw@linux.ibm.com>
10401 * tree-data-ref.c (free_dependence_relations): Adjust to pass vec
10403 (free_data_refs): Likewise.
10404 * tree-data-ref.h (free_dependence_relations): Likewise.
10405 (free_data_refs): Likewise.
10406 * tree-predcom.c (struct chain): Use auto_vec instead of vec for
10408 (struct component): Likewise.
10409 (pcom_worker::pcom_worker): Adjust for auto_vec and renaming changes.
10410 (pcom_worker::~pcom_worker): Likewise.
10411 (pcom_worker::release_chain): Adjust as auto_vec changes.
10412 (pcom_worker::loop): Rename to ...
10413 (pcom_worker::m_loop): ... this.
10414 (pcom_worker::datarefs): Rename to ...
10415 (pcom_worker::m_datarefs): ... this. Use auto_vec instead of vec.
10416 (pcom_worker::dependences): Rename to ...
10417 (pcom_worker::m_dependences): ... this. Use auto_vec instead of vec.
10418 (pcom_worker::chains): Rename to ...
10419 (pcom_worker::m_chains): ... this. Use auto_vec instead of vec.
10420 (pcom_worker::looparound_phis): Rename to ...
10421 (pcom_worker::m_looparound_phis): ... this. Use auto_vec instead of
10423 (pcom_worker::cache): Rename to ...
10424 (pcom_worker::m_cache): ... this. Use auto_vec instead of vec.
10425 (pcom_worker::release_chain): Adjust for auto_vec changes.
10426 (pcom_worker::release_chains): Adjust for auto_vec and renaming
10428 (release_component): Remove.
10429 (release_components): Adjust for release_component removal.
10430 (component_of): Adjust to use vec.
10431 (merge_comps): Likewise.
10432 (pcom_worker::aff_combination_dr_offset): Adjust for renaming changes.
10433 (pcom_worker::determine_offset): Likewise.
10434 (class comp_ptrs): Remove.
10435 (pcom_worker::split_data_refs_to_components): Adjust for renaming
10436 changes, for comp_ptrs removal with auto_vec.
10437 (pcom_worker::suitable_component_p): Adjust for renaming changes.
10438 (pcom_worker::filter_suitable_components): Adjust for release_component
10440 (pcom_worker::valid_initializer_p): Adjust for renaming changes.
10441 (pcom_worker::find_looparound_phi): Likewise.
10442 (pcom_worker::add_looparound_copies): Likewise.
10443 (pcom_worker::determine_roots_comp): Likewise.
10444 (pcom_worker::single_nonlooparound_use): Likewise.
10445 (pcom_worker::execute_pred_commoning_chain): Likewise.
10446 (pcom_worker::execute_pred_commoning): Likewise.
10447 (pcom_worker::try_combine_chains): Likewise.
10448 (pcom_worker::prepare_initializers_chain): Likewise.
10449 (pcom_worker::prepare_initializers): Likewise.
10450 (pcom_worker::prepare_finalizers_chain): Likewise.
10451 (pcom_worker::prepare_finalizers): Likewise.
10452 (pcom_worker::tree_predictive_commoning_loop): Likewise.
10454 2021-07-20 Martin Sebor <msebor@redhat.com>
10456 PR middle-end/101397
10457 * builtins.c (gimple_call_return_array): Add argument. Correct
10458 offsets for memchr, mempcpy, stpcpy, and stpncpy.
10459 (compute_objsize_r): Adjust offset computation for argument returning
10462 2021-07-20 Martin Sebor <msebor@redhat.com>
10464 PR middle-end/101300
10465 * tree-ssa-uninit.c (check_defs): Handle UBSAN built-ins.
10467 2021-07-20 Jeff Law <jlaw@localhost.localdomain>
10469 * function.c (assign_parm_setup_block): Use adjust_address instead
10470 of change_address to preserve MEM_EXPR and friends.
10472 2021-07-20 Martin Sebor <msebor@redhat.com>
10474 * cfgloop.h (single_likely_exit): Adjust by-value argument to
10475 by-const-reference.
10476 * cfgloopanal.c (single_likely_exit): Same.
10477 * cgraph.h (struct cgraph_node): Same.
10478 * cgraphclones.c (cgraph_node::create_virtual_clone): Same.
10479 * genautomata.c (merge_states): Same.
10480 * genextract.c (VEC_char_to_string): Same.
10481 * genmatch.c (dt_node::gen_kids_1): Same.
10482 (walk_captures): Adjust by-value argument to by-reference.
10483 * gimple-ssa-store-merging.c (check_no_overlap): Adjust by-value argument
10484 to by-const-reference.
10485 * gimple.c (gimple_build_call_vec): Same.
10486 (gimple_build_call_internal_vec): Same.
10487 (gimple_build_switch): Same.
10488 (sort_case_labels): Same.
10489 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
10491 * gimple.h (gimple_build_call_vec): Adjust by-value argument to
10492 by-const-reference.
10493 (gimple_build_call_internal_vec): Same.
10494 (gimple_build_switch): Same.
10495 (sort_case_labels): Same.
10496 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
10498 * haifa-sched.c (calc_priorities): Adjust by-value argument to
10499 by-const-reference.
10500 (sched_init_luids): Same.
10501 (haifa_init_h_i_d): Same.
10502 * ipa-cp.c (ipa_get_indirect_edge_target_1): Same.
10503 (adjust_callers_for_value_intersection): Adjust by-value argument to
10505 (find_more_scalar_values_for_callers_subset): Adjust by-value argument to
10506 by-const-reference.
10507 (find_more_contexts_for_caller_subset): Same.
10508 (find_aggregate_values_for_callers_subset): Same.
10509 (copy_useful_known_contexts): Same.
10510 * ipa-fnsummary.c (remap_edge_summaries): Same.
10511 (remap_freqcounting_predicate): Same.
10512 * ipa-inline.c (add_new_edges_to_heap): Adjust by-value argument to
10514 * ipa-predicate.c (predicate::remap_after_inlining): Adjust by-value argument
10515 to by-const-reference.
10516 * ipa-predicate.h (predicate::remap_after_inlining): Same.
10517 * ipa-prop.c (ipa_find_agg_cst_for_param): Same.
10518 * ipa-prop.h (ipa_find_agg_cst_for_param): Same.
10519 * ira-build.c (ira_loop_tree_body_rev_postorder): Same.
10520 * read-rtl.c (add_overload_instance): Same.
10521 * rtl.h (native_decode_rtx): Same.
10522 (native_decode_vector_rtx): Same.
10523 * sched-int.h (sched_init_luids): Same.
10524 (haifa_init_h_i_d): Same.
10525 * simplify-rtx.c (native_decode_vector_rtx): Same.
10526 (native_decode_rtx): Same.
10527 * tree-call-cdce.c (gen_shrink_wrap_conditions): Same.
10528 (shrink_wrap_one_built_in_call_with_conds): Same.
10529 (shrink_wrap_conditional_dead_built_in_calls): Same.
10530 * tree-data-ref.c (create_runtime_alias_checks): Same.
10531 (compute_all_dependences): Same.
10532 * tree-data-ref.h (compute_all_dependences): Same.
10533 (create_runtime_alias_checks): Same.
10534 (index_in_loop_nest): Same.
10535 * tree-if-conv.c (mask_exists): Same.
10536 * tree-loop-distribution.c (class loop_distribution): Same.
10537 (loop_distribution::create_rdg_vertices): Same.
10538 (dump_rdg_partitions): Same.
10539 (debug_rdg_partitions): Same.
10540 (partition_contains_all_rw): Same.
10541 (loop_distribution::distribute_loop): Same.
10542 * tree-parloops.c (oacc_entry_exit_ok_1): Same.
10543 (oacc_entry_exit_single_gang): Same.
10544 * tree-ssa-loop-im.c (hoist_memory_references): Same.
10545 (loop_suitable_for_sm): Same.
10546 * tree-ssa-loop-niter.c (bound_index): Same.
10547 * tree-ssa-reassoc.c (update_ops): Same.
10548 (swap_ops_for_binary_stmt): Same.
10549 (rewrite_expr_tree): Same.
10550 (rewrite_expr_tree_parallel): Same.
10551 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Same.
10552 * tree-ssa-sccvn.h (ao_ref_init_from_vn_reference): Same.
10553 * tree-ssa-structalias.c (process_all_all_constraints): Same.
10554 (make_constraints_to): Same.
10555 (handle_lhs_call): Same.
10556 (find_func_aliases_for_builtin_call): Same.
10557 (sort_fieldstack): Same.
10558 (check_for_overlaps): Same.
10559 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Same.
10560 (vect_create_cond_for_unequal_addrs): Same.
10561 (vect_create_cond_for_lower_bounds): Same.
10562 (vect_create_cond_for_alias_checks): Same.
10563 * tree-vect-slp-patterns.c (vect_validate_multiplication): Same.
10564 * tree-vect-slp.c (vect_analyze_slp_instance): Same.
10565 (vect_make_slp_decision): Same.
10566 (vect_slp_bbs): Same.
10567 (duplicate_and_interleave): Same.
10568 (vect_transform_slp_perm_load): Same.
10569 (vect_schedule_slp): Same.
10570 * tree-vectorizer.h (vect_transform_slp_perm_load): Same.
10571 (vect_schedule_slp): Same.
10572 (duplicate_and_interleave): Same.
10573 * tree.c (build_vector_from_ctor): Same.
10574 (build_vector): Same.
10575 (check_vector_cst): Same.
10576 (check_vector_cst_duplicate): Same.
10577 (check_vector_cst_fill): Same.
10578 (check_vector_cst_stepped): Same.
10579 * tree.h (build_vector_from_ctor): Same.
10581 2021-07-20 Jakub Jelinek <jakub@redhat.com>
10584 * config/rs6000/rs6000-protos.h (easy_altivec_constant): Change return
10585 type from bool to int.
10586 * config/rs6000/rs6000.c (vspltis_constant): Fix up handling the
10587 EASY_VECTOR_MSB case if either step or copies is not 1.
10588 (vspltis_shifted): Fix comment typo.
10589 (easy_altivec_constant): Change return type from bool to int, instead
10590 of returning true return byte size of the element mode that should be
10591 used to synthetize the constant.
10592 * config/rs6000/predicates.md (easy_vector_constant_msb): Require
10593 that vspltis_shifted is 0, handle the case where easy_altivec_constant
10594 assumes using different vector mode from CONST_VECTOR's mode.
10595 * config/rs6000/altivec.md (easy_vector_constant_msb splitter): Use
10596 easy_altivec_constant to determine mode in which -1 >> -1 should be
10597 performed, use rs6000_expand_vector_init instead of gen_vec_initv4sisi.
10599 2021-07-20 Richard Biener <rguenther@suse.de>
10602 * dwarf2out.h (dwarf_file_data): Add key member.
10603 * dwarf2out.c (dwarf_file_hasher::equal): Compare key.
10604 (dwarf_file_hasher::hash): Hash key.
10605 (lookup_filename): Remap the filename and store it in the
10606 filename member of dwarf_file_data when creating a new
10608 (file_name_acquire): Do not remap the filename again.
10609 (maybe_emit_file): Likewise.
10611 2021-07-20 Jonathan Wright <jonathan.wright@arm.com>
10613 * config/aarch64/aarch64-simd-builtins.def: Use two variant
10614 generators for all TBL/TBX intrinsics and rename to
10615 consistent forms: qtbl[1234] or qtbx[1234].
10616 * config/aarch64/aarch64-simd.md (aarch64_tbl1<mode>):
10618 (aarch64_qtbl1<mode>): This.
10619 (aarch64_tbx1<mode>): Rename to...
10620 (aarch64_qtbx1<mode>): This.
10621 (aarch64_tbl2v16qi): Delete.
10622 (aarch64_tbl3<mode>): Rename to...
10623 (aarch64_qtbl2<mode>): This.
10624 (aarch64_tbx4<mode>): Rename to...
10625 (aarch64_qtbx2<mode>): This.
10626 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_1): Use
10627 renamed qtbl1 and qtbl2 RTL patterns.
10628 * config/aarch64/arm_neon.h (vqtbl1_p8): Use renamed qtbl1
10630 (vqtbl1_s8): Likewise.
10631 (vqtbl1_u8): Likewise.
10632 (vqtbl1q_p8): Likewise.
10633 (vqtbl1q_s8): Likewise.
10634 (vqtbl1q_u8): Likewise.
10635 (vqtbx1_s8): Use renamed qtbx1 RTL pattern.
10636 (vqtbx1_u8): Likewise.
10637 (vqtbx1_p8): Likewise.
10638 (vqtbx1q_s8): Likewise.
10639 (vqtbx1q_u8): Likewise.
10640 (vqtbx1q_p8): Likewise.
10641 (vtbl1_s8): Use renamed qtbl1 RTL pattern.
10642 (vtbl1_u8): Likewise.
10643 (vtbl1_p8): Likewise.
10644 (vtbl2_s8): Likewise
10645 (vtbl2_u8): Likewise.
10646 (vtbl2_p8): Likewise.
10647 (vtbl3_s8): Use renamed qtbl2 RTL pattern.
10648 (vtbl3_u8): Likewise.
10649 (vtbl3_p8): Likewise.
10650 (vtbl4_s8): Likewise.
10651 (vtbl4_u8): Likewise.
10652 (vtbl4_p8): Likewise.
10653 (vtbx2_s8): Use renamed qtbx2 RTL pattern.
10654 (vtbx2_u8): Likewise.
10655 (vtbx2_p8): Likewise.
10656 (vqtbl2_s8): Use renamed qtbl2 RTL pattern.
10657 (vqtbl2_u8): Likewise.
10658 (vqtbl2_p8): Likewise.
10659 (vqtbl2q_s8): Likewise.
10660 (vqtbl2q_u8): Likewise.
10661 (vqtbl2q_p8): Likewise.
10662 (vqtbx2_s8): Use renamed qtbx2 RTL pattern.
10663 (vqtbx2_u8): Likewise.
10664 (vqtbx2_p8): Likewise.
10665 (vqtbx2q_s8): Likewise.
10666 (vqtbx2q_u8): Likewise.
10667 (vqtbx2q_p8): Likewise.
10668 (vtbx4_s8): Likewise.
10669 (vtbx4_u8): Likewise.
10670 (vtbx4_p8): Likewise.
10672 2021-07-20 Uroš Bizjak <ubizjak@gmail.com>
10675 * config/i386/sync.md (define_peephole2 atomic_storedi_fpu):
10677 (define_peephole2 atomic_loaddi_fpu): Ditto.
10679 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
10681 * config.gcc (riscv*-*-*): Detect which python is available.
10683 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
10685 * config/rs6000/vsx.md (mulhs_<mode>): Rename to...
10686 (smul<mode>3_highpart): ... this.
10687 (mulhu_<mode>): Rename to...
10688 (umul<mode>3_highpart): ... this.
10689 * config/rs6000/rs6000-builtin.def (MULHS_V2DI, MULHS_V4SI,
10690 MULHU_V2DI, MULHU_V4SI): Adjust.
10692 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
10694 PR tree-optimization/100696
10695 * internal-fn.c (first_commutative_argument): Add info for IFN_MULH.
10696 * internal-fn.def (IFN_MULH): New internal function.
10697 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Add support to
10698 recog normal multiply highpart as IFN_MULH.
10699 * config/i386/i386.c (ix86_add_stmt_cost): Adjust for combined
10702 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
10704 * config/elfos.h (CTF_DEBUGGING_INFO): New definition.
10705 (BTF_DEBUGGING_INFO): Likewise.
10706 * doc/tm.texi.in: Document the new macros.
10707 * doc/tm.texi: Regenerated.
10708 * toplev.c: Guard initialization of debug hooks.
10710 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
10712 * flags.h (ctf_debuginfo_p): New function declaration.
10713 * opts.c (ctf_debuginfo_p): New function definition.
10715 2021-07-19 Andrew Stubbs <ams@codesourcery.com>
10718 * config/gcn/gcn-hsa.h (DRIVER_SELF_SPECS): New.
10719 (ASM_SPEC): Set -mattr for xnack and sram-ecc.
10720 * config/gcn/gcn-opts.h (enum sram_ecc_type): New.
10721 * config/gcn/gcn-valu.md: Add a warning comment.
10722 * config/gcn/gcn.c (gcn_option_override): Add "sorry" for -mxnack.
10723 (output_file_start): Add xnack and sram-ecc state to ".amdgcn_target".
10724 * config/gcn/gcn.md: Add a warning comment.
10725 * config/gcn/gcn.opt: Add -mxnack and -msram-ecc.
10726 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): Remove
10728 (EF_AMDGPU_XNACK): New.
10729 (EF_AMDGPU_SRAM_ECC): New.
10731 (copy_early_debug_info): Use elf_flags.
10732 (main): Handle -mxnack and -msram-ecc options.
10733 * doc/invoke.texi: Document -mxnack and -msram-ecc.
10735 2021-07-19 Andrew Pinski <apinski@marvell.com>
10738 * config/aarch64/aarch64.md (csneg3_uxtw_insn): Rename to ...
10739 (*cs<neg_not_cs>3_uxtw_insn4): and extend to NEG_NOT.
10741 2021-07-19 Richard Biener <rguenther@suse.de>
10743 PR tree-optimization/101505
10744 * tree-vect-patterns.c (vect_determine_precisions): Walk
10745 PHIs also for loop vectorization.
10747 2021-07-19 Richard Biener <rguenther@suse.de>
10749 * gimple.h (gimple_expr_type): Remove.
10750 * doc/gimple.texi: Remove gimple_expr_type documentation.
10752 2021-07-19 Richard Biener <rguenther@suse.de>
10754 * tree-ssa-sccvn.c (vn_reference_eq): Handle NULL vr->type.
10755 (ao_ref_init_from_vn_reference): Likewise.
10756 (fully_constant_reference): Likewise.
10757 (vn_reference_lookup_call): Do not set vr->type to random
10759 * tree-ssa-pre.c (compute_avail): Do not try to PRE calls
10761 * tree-vect-generic.c (expand_vector_piecewise): Pass in
10762 whether we expanded parallel.
10763 (expand_vector_parallel): Adjust.
10764 (expand_vector_addition): Likewise.
10765 (expand_vector_comparison): Likewise.
10766 (expand_vector_operation): Likewise.
10767 (expand_vector_scalar_condition): Likewise.
10768 (expand_vector_conversion): Likewise.
10770 2021-07-19 Richard Biener <rguenther@suse.de>
10772 * tree-vrp.c (register_edge_assert_for_2): Use the
10774 (vrp_folder::fold_predicate_in): Likewise.
10775 * vr-values.c (gimple_assign_nonzero_p): Likewise.
10776 (vr_values::extract_range_from_comparison): Likewise.
10777 (vr_values::extract_range_from_ubsan_builtin): Use the
10778 type of the first operand.
10779 (vr_values::extract_range_basic): Push down type
10780 computation, use the appropriate LHS.
10781 (vr_values::extract_range_from_assignment): Use the
10784 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
10787 * common/config/i386/i386-common.c (ix86_handle_option): For
10788 -mgeneral-regs-only, enable the GPR only instructions which are
10789 enabled implicitly by SSE ISAs unless they have been disabled
10792 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
10795 * config/i386/i386.c (ix86_check_avx_upper_stores): Moved before
10796 ix86_avx_u128_mode_needed.
10797 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY if callee
10798 returns AVX register.
10800 2021-07-17 Jan Hubicka <hubicka@ucw.cz>
10802 * tree-ssa-structalias.c (handle_rhs_call): Support EAF_NOT_RETURNED.
10803 (handle_const_call): Liekise
10804 (handle_pure_call): Liekise
10806 2021-07-17 Andrew MacLeod <amacleod@redhat.com>
10808 PR tree-optimization/96542
10809 * range-op.cc (range_operator::wi_fold_in_parts): New.
10810 (range_operator::fold_range): Call wi_fold_in_parts.
10811 (operator_lshift::wi_fold): Fix broken lshift by [0,0].
10812 * range-op.h (wi_fold_in_parts): Add prototype.
10814 2021-07-16 David Malcolm <dmalcolm@redhat.com>
10816 * doc/analyzer.texi: Add __analyzer_dump_state.
10818 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10820 * config/rs6000/rbtree.c: New file.
10821 * config/rs6000/rbtree.h: New file.
10823 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10825 * config/rs6000/rs6000-gen-builtins.c (restriction): New enum.
10826 (typeinfo): Add restr field.
10827 (match_bracketed_pair): New function.
10828 (match_const_restriction): Implement.
10830 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10832 * config/rs6000/rs6000-gen-builtins.c (match_basetype): Implement.
10834 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10836 * config/rs6000/rs6000-gen-builtins.c (void_status): New enum.
10837 (basetype): Likewise.
10838 (typeinfo): Likewise.
10839 (handle_pointer): New function.
10840 (match_basetype): New stub function.
10841 (match_const_restriction): Likewise.
10842 (match_type): New function.
10844 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10846 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace): New
10848 (advance_line): Likewise.
10849 (safe_inc_pos): Likewise.
10850 (match_identifier): Likewise.
10851 (match_integer): Likewise.
10852 (match_to_right_bracket): Likewise.
10854 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10856 * config/rs6000/rs6000-gen-builtins.c (bif_file): New variable.
10857 (ovld_file): Likewise.
10858 (header_file): Likewise.
10859 (init_file): Likewise.
10860 (defines_file): Likewise.
10861 (pgm_path): Likewise.
10862 (bif_path): Likewise.
10863 (ovld_path): Likewise.
10864 (header_path): Likewise.
10865 (init_path): Likewise.
10866 (defines_path): Likewise.
10867 (LINELEN): New macro.
10868 (linebuf): New variable.
10872 (bif_diag): New function.
10873 (ovld_diag): Likewise.
10875 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10877 * config/rs6000/rs6000-builtin-new.def: New.
10878 * config/rs6000/rs6000-overload.def: New.
10880 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10882 * config/rs6000/rs6000-gen-builtins.c: New.
10884 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10886 * Makefile.in (EXTRA_GTYPE_DEPS): New variable.
10887 (s-gtype): Depend on EXTRA_GTYPE_DEPS.
10888 * gengtype-state.c (state_writer::write_state_file_list): Add a
10889 parameter to the fileslist expression for the number of build
10891 (read_state_files_list): Detect build headers and strip the
10892 initial "./" or ".\" from their names.
10893 * gengtype.c (build_headers): New global variable.
10894 (num_build_headers): Likewise.
10895 (open_base_files): Emit #include for each build header.
10896 (main): Detect and count build headers.
10897 * gengtype.h (build_headers): New extern variable.
10898 (num_build_headers): Likewise.
10900 2021-07-16 Richard Biener <rguenther@suse.de>
10902 * gimple-ssa-store-merging.c (verify_symbolic_number_p): Use
10903 the type of the LHS.
10904 (find_bswap_or_nop_1): Likewise.
10905 (find_bswap_or_nop): Likewise.
10906 * tree-vectorizer.h (vect_get_smallest_scalar_type): Adjust
10908 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
10909 Remove unused parameters, pass in the scalar type. Fix
10910 internal store function handling.
10911 * tree-vect-stmts.c (vect_analyze_stmt): Remove assert.
10912 (vect_get_vector_types_for_stmt): Move down check for
10913 existing vector stmt after we've determined a scalar type.
10914 Pass down the used scalar type to vect_get_smallest_scalar_type.
10915 * tree-vect-generic.c (expand_vector_condition): Use
10916 the type of the LHS.
10917 (expand_vector_scalar_condition): Likewise.
10918 (expand_vector_operations_1): Likewise.
10919 * tree-vect-patterns.c (vect_widened_op_tree): Likewise.
10920 (vect_recog_dot_prod_pattern): Likewise.
10921 (vect_recog_sad_pattern): Likewise.
10922 (vect_recog_widen_op_pattern): Likewise.
10923 (vect_recog_widen_sum_pattern): Likewise.
10924 (vect_recog_mixed_size_cond_pattern): Likewise.
10926 2021-07-16 Jan Hubicka <hubicka@ucw.cz>
10928 * ipa-modref.c (struct escape_entry): Use eaf_fleags_t.
10929 (dump_eaf_flags): Dump EAF_NOT_RETURNED
10930 (eaf_flags_useful_p): Use eaf_fleags_t; handle const functions
10931 and EAF_NOT_RETURNED.
10932 (modref_summary::useful_p): Likewise.
10933 (modref_summary_lto::useful_p): Likewise.
10934 (struct) modref_summary_lto: Use eaf_fleags_t.
10935 (deref_flags): Handle EAF_NOT_RETURNED.
10936 (struct escape_point): Use min_flags.
10937 (modref_lattice::init): Add EAF_NOT_RETURNED.
10938 (merge_call_lhs_flags): Ignore EAF_NOT_RETURNED functions
10939 (analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
10941 (analyze_parms): Also analyze const functions; update conition on
10943 (modref_write): Update streaming.
10944 (read_section): Update streaming.
10945 (remap_arg_flags): Use eaf_flags_t.
10946 (modref_merge_call_site_flags): Hanlde EAF_NOT_RETURNED.
10947 * ipa-modref.h: (eaf_flags_t): New typedef.
10948 (struct modref_summary): Use eaf_flags_t.
10949 * tree-core.h (EAF_NOT_RETURNED): New constant.
10951 2021-07-16 Richard Biener <rguenther@suse.de>
10953 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Use
10954 the type of the LHS.
10955 (gimple_assign_nonnegative_warnv_p): Likewise.
10956 (gimple_call_nonnegative_warnv_p): Likewise. Return false
10957 if the call has no LHS.
10958 * gimple.c (gimple_could_trap_p_1): Use the type of the LHS.
10959 * tree-eh.c (stmt_could_throw_1_p): Likewise.
10960 * tree-inline.c (insert_init_stmt): Likewise.
10961 * tree-ssa-loop-niter.c (get_val_for): Likewise.
10962 * tree-outof-ssa.c (ssa_is_replaceable_p): Use the type of
10964 * tree-ssa-sccvn.c (init_vn_nary_op_from_stmt): Take a
10965 gassign *. Use the type of the lhs.
10966 (vn_nary_op_lookup_stmt): Adjust.
10967 (vn_nary_op_insert_stmt): Likewise.
10969 2021-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
10971 * config/s390/predicates.md (bras_sym_operand): Accept all
10972 functions in 64-bit mode, use UNSPEC_PLT31.
10973 (larl_operand): Use UNSPEC_PLT31.
10974 * config/s390/s390.c (s390_loadrelative_operand_p): Likewise.
10975 (legitimize_pic_address): Likewise.
10976 (s390_emit_tls_call_insn): Mark __tls_get_offset as function,
10978 (s390_delegitimize_address): Use UNSPEC_PLT31.
10979 (s390_output_addr_const_extra): Likewise.
10980 (print_operand): Add @PLT to TLS calls, handle %K.
10981 (s390_function_profiler): Mark __fentry__/_mcount as function,
10982 use %K, use UNSPEC_PLT31.
10983 (s390_output_mi_thunk): Use only UNSPEC_GOT, use %K.
10984 (s390_emit_call): Use UNSPEC_PLT31.
10985 (s390_emit_tpf_eh_return): Mark __tpf_eh_return as function.
10986 * config/s390/s390.md (UNSPEC_PLT31): Rename from UNSPEC_PLT.
10987 (*movdi_64): Use %K.
10988 (reload_base_64): Likewise.
10989 (*sibcall_brc): Likewise.
10990 (*sibcall_brcl): Likewise.
10991 (*sibcall_value_brc): Likewise.
10992 (*sibcall_value_brcl): Likewise.
10994 (*brasl): Likewise.
10995 (*bras_r): Likewise.
10996 (*brasl_r): Likewise.
10997 (*bras_tls): Likewise.
10998 (*brasl_tls): Likewise.
10999 (main_base_64): Likewise.
11000 (reload_base_64): Likewise.
11001 (@split_stack_call<mode>): Likewise.
11003 2021-07-16 Richard Biener <rguenther@suse.de>
11005 PR tree-optimization/101467
11006 * tree-vect-stmts.c (vect_gen_while): Properly guard
11007 make_temp_ssa_name usage.
11009 2021-07-16 Cooper Qu <cooper.qu@linux.alibaba.com>
11011 * config.gcc: Don't use forked print-sysroot-suffix.sh and
11012 t-sysroot-suffix for C-SKY.
11013 * config/csky/print-sysroot-suffix.sh: Delete.
11014 * config/csky/t-csky-linux: Delete.
11015 * config/csky/t-sysroot-suffix: Define MULTILIB_DIRNAMES
11016 instead of CSKY_MULTILIB_DIRNAMES.
11018 2021-07-16 Richard Biener <rguenther@suse.de>
11020 * tree-vect-loop.c (vect_transform_cycle_phi): Correct sign
11021 conversion issues with the partial reduction of the reused
11022 vector accumulator.
11024 2021-07-16 Richard Biener <rguenther@suse.de>
11026 * config/i386/i386-options.c (ix86_option_override_internal): Set
11027 param_vect_partial_vector_usage to zero if not set.
11029 2021-07-15 Uroš Bizjak <ubizjak@gmail.com>
11032 * config/i386/i386.h (VALID_SSE_REG_MODE): Add TDmode.
11033 (VALID_INT_MODE_P): Add SDmode and DDmode.
11034 Add TDmode for TARGET_64BIT.
11035 (VALID_DFP_MODE_P): Remove.
11036 * config/i386/i386.c (ix86_hard_regno_mode_ok):
11037 Do not use VALID_DFP_MODE_P.
11039 2021-07-15 Andrew MacLeod <amacleod@redhat.com>
11041 * gimple-range-fold.cc (adjust_pointer_diff_expr): Use
11043 (fold_using_range::fold_stmt): Ditto.
11044 (fold_using_range::range_of_range_op): Ditto.
11045 (fold_using_range::range_of_phi): Ditto.
11046 (fold_using_range::range_of_call): Ditto.
11047 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
11048 (fold_using_range::range_of_builtin_call): Ditto.
11049 (fold_using_range::range_of_cond_expr): Ditto.
11050 * gimple-range-fold.h (gimple_range_type): New.
11052 2021-07-15 Martin Sebor <msebor@redhat.com>
11054 PR middle-end/97027
11055 * tree-ssa-strlen.c (handle_assign): New function.
11056 (maybe_warn_overflow): Add argument.
11057 (nonzero_bytes_for_type): New function.
11058 (count_nonzero_bytes): Handle more tree types. Call
11059 nonzero_bytes_for_tye.
11060 (count_nonzero_bytes): Handle types.
11061 (handle_store): Handle stores from function calls.
11062 (strlen_check_and_optimize_call): Move code to handle_assign. Call
11063 it for assignments from function calls.
11065 2021-07-15 David Malcolm <dmalcolm@redhat.com>
11070 * doc/invoke.texi: Add -Wanalyzer-use-of-uninitialized-value.
11072 2021-07-15 David Malcolm <dmalcolm@redhat.com>
11074 * doc/invoke.texi (-fdump-analyzer-exploded-paths): New.
11076 2021-07-15 Martin Sebor <msebor@redhat.com>
11080 * fold-const.c (operand_compare::operand_equal_p): Handle OEP_DECL_NAME.
11081 (operand_compare::verify_hash_value): Same.
11082 * tree-core.h (OEP_DECL_NAME): New.
11084 2021-07-15 Martin Jambor <mjambor@suse.cz>
11086 * profile-count.h (profile_count::value): Change the return type to
11088 * gimple-pretty-print.c (dump_gimple_bb_header): Adjust print
11090 * tree-cfg.c (dump_function_to_file): Likewise.
11092 2021-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
11095 * config/rs6000/rs6000-p8swap.c (has_part_mult): New.
11096 (rs6000_analyze_swaps): Insns containing a subreg of a mult are
11099 2021-07-15 Richard Biener <rguenther@suse.de>
11101 * tree-vectorizer.h (vect_gen_while): Match up with
11102 vect_gen_while_not.
11103 * tree-vect-stmts.c (vect_gen_while): Adjust API to that
11104 of vect_gen_while_not.
11105 (vect_gen_while_not): Adjust.
11106 * tree-vect-loop-manip.c (vect_set_loop_controls_directly): Likewise.
11108 2021-07-15 Aldy Hernandez <aldyh@redhat.com>
11110 * gimple-range-cache.cc (non_null_ref::adjust_range): New.
11111 (ranger_cache::range_of_def): Call adjust_range.
11112 (ranger_cache::entry_range): Same.
11113 * gimple-range-cache.h (non_null_ref::adjust_range): New.
11114 * gimple-range.cc (gimple_ranger::range_of_expr): Call
11116 (gimple_ranger::range_on_entry): Same.
11118 2021-07-15 Tamar Christina <tamar.christina@arm.com>
11121 2021-07-14 Tamar Christina <tamar.christina@arm.com>
11123 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
11125 2021-07-15 Tamar Christina <tamar.christina@arm.com>
11128 2021-07-14 Tamar Christina <tamar.christina@arm.com>
11130 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
11131 (sdot_prod, udot_prod): ...These.
11132 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
11133 (aarch64_<sur>dot<vsi2qi>): Rename to...
11134 (<sur>dot_prod<vsi2qi>): ...This.
11135 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
11138 2021-07-15 Jakub Jelinek <jakub@redhat.com>
11140 PR middle-end/101437
11141 * gimplify.c (gimplify_expr): Throw away volatile reads from empty
11142 types even if they have non-BLKmode TYPE_MODE.
11144 2021-07-15 Richard Biener <rguenther@suse.de>
11147 * gcc.c (process_command): Process -gtoggle like process_options
11148 would after parsing options.
11150 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
11152 * cfgexpand.c (expand_asm_loc): Adjust.
11153 (expand_asm_stmt): Likewise.
11154 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Likewise.
11155 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
11156 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
11157 * config/avr/avr.c (avr_md_asm_adjust): Likewise.
11158 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
11159 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
11160 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
11161 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
11162 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
11163 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
11164 * config/s390/s390.c (s390_md_asm_adjust): Likewise.
11165 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
11166 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
11167 * doc/tm.texi: Regenerate.
11168 * target.def: Add location argument to md_asm_adjust.
11170 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
11172 * tree-diagnostic.c (diagnostic_report_current_function): Use the
11173 diagnostic's location, not input_location.
11175 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
11177 * cfgexpand.c (tree_conflicts_with_clobbers_p): Pass location to
11179 (expand_asm_stmt): Likewise.
11181 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
11183 * config/rs6000/rs6000.c (adjacent_mem_locations): Return the lower
11184 addressed memory rtx, if any.
11185 (rs6000_split_multireg_move): Fix code formatting.
11186 Handle MMA build built-ins with operands in adjacent memory locations.
11188 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
11190 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Move to later
11193 2021-07-14 Jason Merrill <jason@redhat.com>
11195 * sel-sched-ir.h (get_all_loop_exits): Use auto_vec.
11197 2021-07-14 Jason Merrill <jason@redhat.com>
11199 * doc/invoke.texi: -fdelete-dead-exceptions is on by default for
11202 2021-07-14 Tamar Christina <tamar.christina@arm.com>
11204 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
11205 Remove erroneous line.
11207 2021-07-14 Andrew MacLeod <amacleod@redhat.com>
11209 * params.opt (param_evrp_mode): Change default.
11211 2021-07-14 Tamar Christina <tamar.christina@arm.com>
11213 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
11214 (sdot_prod, udot_prod): ...These.
11215 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
11216 (aarch64_<sur>dot<vsi2qi>): Rename to...
11217 (<sur>dot_prod<vsi2qi>): ...This.
11218 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
11221 2021-07-14 Tamar Christina <tamar.christina@arm.com>
11223 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
11225 2021-07-14 Tamar Christina <tamar.christina@arm.com>
11227 * doc/sourcebuild.texi (arm_v8_2a_i8mm_neon_hw): Document.
11229 2021-07-14 Tamar Christina <tamar.christina@arm.com>
11231 * config/arm/neon.md (usdot_prod<vsi2qi>): New.
11233 2021-07-14 Tamar Christina <tamar.christina@arm.com>
11235 * config/aarch64/aarch64-simd.md (aarch64_usdot<vsi2qi>): Rename to...
11236 (usdot_prod<vsi2qi>): ... This.
11237 * config/aarch64/aarch64-simd-builtins.def (usdot): Rename to...
11238 (usdot_prod): ...This.
11239 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Likewise.
11240 * config/aarch64/aarch64-sve.md (@aarch64_<sur>dot_prod<vsi2qi>):
11242 (@<sur>dot_prod<vsi2qi>): ...This.
11243 * config/aarch64/aarch64-sve-builtins-base.cc
11244 (svusdot_impl::expand): Use it.
11246 2021-07-14 Tamar Christina <tamar.christina@arm.com>
11248 * optabs.def (usdot_prod_optab): New.
11249 * doc/md.texi: Document it and clarify other dot prod optabs.
11250 * optabs-tree.h (enum optab_subtype): Add optab_vector_mixed_sign.
11251 * optabs-tree.c (optab_for_tree_code): Support usdot_prod_optab.
11252 * optabs.c (expand_widen_pattern_expr): Likewise.
11253 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
11254 * tree-vect-loop.c (vectorizable_reduction): Query dot-product kind.
11255 * tree-vect-patterns.c (vect_supportable_direct_optab_p): Take optional
11257 (vect_widened_op_tree): Optionally ignore
11259 (vect_recog_dot_prod_pattern): Support usdot_prod_optab.
11261 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
11264 * config/i386/driver-i386.c (host_detect_local_cpu): Check
11265 "arch [32|64]" and "tune [32|64]" for 32-bit and 64-bit codegen.
11266 Enable UINTR only for 64-bit codegen.
11267 * config/i386/i386-options.c
11268 (ix86_option_override_internal::DEF_PTA): Skip PTA_UINTR if not
11270 * config/i386/i386.h (ARCH_ARG): New.
11271 (CC1_CPU_SPEC): Pass "[arch|tune] 32" for 32-bit codegen and
11272 "[arch|tune] 64" for 64-bit codegen.
11274 2021-07-14 Richard Biener <rguenther@suse.de>
11276 PR tree-optimization/101445
11277 * tree-vect-stmts.c (vectorizable_load): Do the gap adjustment
11278 of the IV in the correct direction for negative stride
11281 2021-07-14 Jakub Jelinek <jakub@redhat.com>
11284 * godump.c (godump_str_hash): New type.
11285 (godump_container::pot_dummy_types): Use string_hash instead of
11286 ptr_hash in the hash_set.
11288 2021-07-14 Richard Biener <rguenther@suse.de>
11290 * tree-vect-loop.c (vect_find_reusable_accumulator): Handle
11291 vector types where the old vector type has a multiple of
11292 the new vector type elements.
11293 (vect_create_partial_epilog): New function, split out from...
11294 (vect_create_epilog_for_reduction): ... here.
11295 (vect_transform_cycle_phi): Reduce the re-used accumulator
11296 to the new vector type.
11298 2021-07-14 Alexandre Oliva <oliva@adacore.com>
11300 * tree-ssa-alias.c (attr_fnspec::verify): Fix index in
11301 non-'t'-sized arg check.
11303 2021-07-14 Alexandre Oliva <oliva@adacore.com>
11305 * tree-cfg.c (cleanup_dead_labels_eh): Update
11306 post_landing_pad label upon change of landing pad block's
11308 (cleanup_dead_labels): Check that a removed label is not that
11311 2021-07-13 Jonathan Wright <jonathan.wright@arm.com>
11313 * combine.c (combine_simplify_rtx): Add vec_select -> subreg
11315 * config/aarch64/aarch64.md (*zero_extend<SHORT:mode><GPI:mode>2_aarch64):
11316 Add Neon to general purpose register case for zero-extend
11318 * config/arm/vfp.md (*arm_movsi_vfp): Remove "*" from *t -> r
11319 case to prevent some cases opting to go through memory.
11320 * cse.c (fold_rtx): Add vec_select -> subreg simplification.
11321 * rtl.c (rtvec_series_p): Define predicate to determine
11322 whether a vector contains a linear series of integers.
11323 * rtl.h (rtvec_series_p): Define.
11324 * rtlanal.c (vec_series_lowpart_p): Define predicate to
11325 determine if a vector selection is equivalent to the low part
11327 * rtlanal.h (vec_series_lowpart_p): Define.
11328 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
11329 Add vec_select -> subreg simplification.
11331 2021-07-13 Paul A. Clarke <pc@us.ibm.com>
11333 * config/rs6000/smmintrin.h (_mm_testz_si128, _mm_testc_si128,
11334 _mm_testnzc_si128, _mm_test_all_ones, _mm_test_all_zeros,
11335 _mm_test_mix_ones_zeros): New.
11337 2021-07-13 Roger Sayle <roger@nextmovesoftware.com>
11338 Richard Biener <rguenther@suse.de>
11340 * gimple.c (gimple_could_trap_p_1): Make S argument a
11341 "const gimple*". Preserve constness in call to
11342 gimple_asm_volatile_p.
11343 (gimple_could_trap_p): Make S argument a "const gimple*".
11344 * gimple.h (gimple_could_trap_p_1, gimple_could_trap_p):
11345 Update function prototypes.
11347 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
11349 * tree-vectorizer.h (vect_reusable_accumulator): New structure.
11350 (_loop_vec_info::main_loop_edge): New field.
11351 (_loop_vec_info::skip_main_loop_edge): Likewise.
11352 (_loop_vec_info::skip_this_loop_edge): Likewise.
11353 (_loop_vec_info::reusable_accumulators): Likewise.
11354 (_stmt_vec_info::reduc_scalar_results): Likewise.
11355 (_stmt_vec_info::reused_accumulator): Likewise.
11356 (vect_get_main_loop_result): Declare.
11357 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize
11358 reduc_scalar_inputs.
11359 (vec_info::free_stmt_vec_info): Free reduc_scalar_inputs.
11360 * tree-vect-loop-manip.c (vect_get_main_loop_result): New function.
11361 (vect_do_peeling): Fill an epilogue loop's main_loop_edge,
11362 skip_main_loop_edge and skip_this_loop_edge fields.
11363 * tree-vect-loop.c (INCLUDE_ALGORITHM): Define.
11364 (vect_emit_reduction_init_stmts): New function.
11365 (get_initial_def_for_reduction): Use it.
11366 (get_initial_defs_for_reduction): Likewise. Change the vinfo
11367 parameter to a loop_vec_info.
11368 (vect_create_epilog_for_reduction): Store the scalar results
11369 in the reduc_info. If an epilogue loop is reusing an accumulator
11370 from the main loop, and if the epilogue loop can also be skipped,
11371 try to place the reduction code in the join block. Record
11372 accumulators that could potentially be reused by epilogue loops.
11373 (vect_transform_cycle_phi): When vectorizing epilogue loops,
11374 try to reuse accumulators from the main loop. Record the initial
11375 value in reduc_info for non-SLP reductions too.
11377 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
11379 * tree-vect-loop.c (get_initial_def_for_reduction): Remove
11380 adjustment handling. Take the neutral value as an argument,
11381 in place of the code argument.
11382 (vect_transform_cycle_phi): Update accordingly. Handle the
11383 initial values of cond reductions separately from code reductions.
11384 Choose the adjustment here rather than in
11385 get_initial_def_for_reduction. Sink the splat of vec_initial_def.
11387 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
11389 * tree-vect-loop.c (neutral_op_for_slp_reduction): Replace with...
11390 (neutral_op_for_reduction): ...this, providing a more general
11392 (vect_create_epilog_for_reduction): Update accordingly.
11393 (vectorizable_reduction): Likewise.
11394 (vect_transform_cycle_phi): Likewise.
11396 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
11398 * tree-vect-loop.c (get_initial_def_for_reduction): Take the
11399 reduc_info instead of the original stmt_vec_info.
11400 (vect_transform_cycle_phi): Update accordingly.
11402 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
11404 * tree-vect-loop.c (get_initial_defs_for_reduction): Take the
11405 reduc_info as an additional parameter.
11406 (vect_transform_cycle_phi): Update accordingly.
11408 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
11410 * tree-vectorizer.h: Include tree-ssa-operands.h.
11411 (vect_phi_initial_value): New function.
11412 * tree-vect-loop.c (neutral_op_for_slp_reduction): Use it.
11413 (get_initial_defs_for_reduction, info_for_reduction): Likewise.
11414 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
11415 (vect_transform_cycle_phi, vectorizable_induction): Likewise.
11417 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
11419 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
11420 the phi results to vectype after creating them. Remove later
11421 conversion code that thus becomes redundant.
11423 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
11425 * tree-vect-loop.c (vect_create_epilog_for_reduction): Replace
11426 the new_phis vector with a reduc_inputs vector. Combine handling
11427 of reduction chains and ncopies > 1.
11429 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
11431 * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
11432 scalar_results to group_size elements after reducing down from
11433 N*group_size elements. Construct an array_slice of the live-out
11434 stmts and assert that there is one stmt per scalar result.
11436 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
11438 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
11439 nested_in_vect_loop and use double_reduc everywhere. Remove dead
11440 assignment to "loop".
11442 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
11444 * internal-fn.c (vectorized_internal_fn_supported_p): Handle
11445 vector types first. For scalar types, consider both the preferred
11446 vector mode and the alternative vector modes.
11447 * optabs-query.c (can_vec_mask_load_store_p): Use the same
11448 structure as above, in particular using related_vector_mode
11449 for modes provided by autovectorize_vector_modes.
11451 2021-07-13 Jakub Jelinek <jakub@redhat.com>
11452 Richard Biener <rguenther@suse.de>
11454 PR tree-optimization/101419
11455 * tree-pass.h (PROP_objsz): Define.
11456 (make_pass_early_object_sizes): Declare.
11457 * passes.def (pass_all_early_optimizations): Rename pass_object_sizes
11458 there to pass_early_object_sizes, drop parameter.
11459 (pass_all_optimizations): Move pass_object_sizes right after pass_ccp,
11460 drop parameter, move pass_post_ipa_warn right after that.
11461 * tree-object-size.c (pass_object_sizes::execute): Rename to...
11462 (object_sizes_execute): ... this. Add insert_min_max_p argument.
11463 (pass_data_object_sizes): Move after object_sizes_execute.
11464 (pass_object_sizes): Likewise. In execute method call
11465 object_sizes_execute, drop set_pass_param method and insert_min_max_p
11466 non-static data member and its initializer in the ctor.
11467 (pass_data_early_object_sizes, pass_early_object_sizes,
11468 make_pass_early_object_sizes): New.
11469 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Use
11470 (cfun->curr_properties & PROP_objsz) instead of cfun->after_inlining.
11472 2021-07-13 Kito Cheng <kito.cheng@sifive.com>
11475 * config/riscv/constraints.md ("S"): Update description and remove
11477 * doc/md.texi (Machine Constraints): Document the 'S' constraints
11480 2021-07-13 Richard Biener <rguenther@suse.de>
11483 2021-07-12 Richard Biener <rguenther@suse.de>
11485 * tree-vect-slp.c (vect_slp_region): Show the number of
11486 SLP graph entries in the optimization message.
11488 2021-07-13 Michael Meissner <meissner@linux.ibm.com>
11490 * config/rs6000/altivec.md (xxspltiw_v4sf): Change local variable
11492 * config/rs6000/rs6000-protos.h (rs6000_const_f32_to_i32): Change
11493 return type to long.
11494 * config/rs6000/rs6000.c (rs6000_const_f32_to_i32): Change return
11497 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
11499 * gimple-range-fold.cc (fold_using_range::range_of_builtin_ubsan_call):
11500 Query relation between the 2 operands and use it.
11502 2021-07-12 Sergei Trofimovich <siarheit@google.com>
11504 * doc/cfg.texi: Fix s/ei_safe_safe/ei_safe_edge/ typo.
11506 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
11509 * config/i386/predicates.md (vec_setm_sse41_operand):
11510 Rename from vec_setm_operand.
11511 (vec_setm_avx2_operand): New predicate.
11512 * config/i386/sse.md (vec_set<V_128:mode>): Use V_128 mode iterator.
11513 Use vec_setm_sse41_operand as operand 2 predicate.
11514 (vec_set<V_256_512:mode): New expander.
11515 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_sse41_operand
11516 as operand 2 predicate.
11518 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
11520 PR tree-optimization/101335
11521 * range-op.cc (operator_cast::lhs_op1_relation): Delete.
11523 2021-07-12 Andrew Pinski <apinski@marvell.com>
11525 * tree-ssa-phiopt.c (match_simplify_replacement): Move
11526 insert of the sequence before the movement of the
11527 statement. Check if to see if the statement is used
11528 outside of the original phi to see if we should move it.
11530 2021-07-12 Richard Biener <rguenther@suse.de>
11532 * dump-context.h (debug_dump_context::debug_dump_context):
11533 Add FILE * parameter defaulted to stderr.
11534 * dumpfile.c (debug_dump_context::debug_dump_context): Adjust.
11535 * tree-vect-slp.c (dot_slp_tree): New functions.
11537 2021-07-12 Richard Biener <rguenther@suse.de>
11539 PR tree-optimization/101373
11540 * tree-ssa-pre.c (prune_clobbered_mems): Also prune trapping
11541 references when the BB may not return.
11542 (compute_avail): Pass in the function we're working on and
11543 replace cfun references with it. Externally throwing
11544 const calls also possibly terminate the function.
11545 (pass_pre::execute): Pass down the function we're working on.
11546 * gcse.c (compute_hash_table_work): Externally throwing
11547 const/pure calls also need record_last_mem_set_info.
11548 * postreload-gcse.c (record_opr_changes): Looping or externally
11549 throwing const/pure calls also need record_last_mem_set_info.
11551 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
11553 * recog.c (memory_address_addr_space_p): Change the type to bool.
11554 Return true/false instead of 1/0.
11555 (offsettable_memref_p): Ditto.
11556 (offsettable_nonstrict_memref_p): Ditto.
11557 (offsettable_address_addr_space_p): Ditto.
11558 Change the type of addressp indirect function to bool.
11559 * recog.h (memory_address_addr_space_p): Change the type to bool.
11560 (strict_memory_address_addr_space_p): Ditto.
11561 (offsettable_memref_p): Ditto.
11562 (offsettable_nonstrict_memref_p): Ditto.
11563 (offsettable_address_addr_space_p): Ditto.
11564 * reload.c (maybe_memory_address_addr_space_p): Ditto.
11565 (strict_memory_address_addr_space_p): Change the type to bool.
11566 Return true/false instead of 1/0.
11567 (maybe_memory_address_addr_space_p): Change the type to bool.
11569 2021-07-12 Richard Biener <rguenther@suse.de>
11571 * tree-vect-slp.c (vect_slp_region): Show the number of
11572 SLP graph entries in the optimization message.
11574 2021-07-12 Richard Biener <rguenther@suse.de>
11576 PR tree-optimization/101394
11577 * tree-ssa-pre.c (do_pre_regular_insertion): Avoid inserting
11578 copies from abnormals for a full redundancy.
11580 2021-07-12 Richard Biener <rguenther@suse.de>
11582 PR middle-end/101423
11583 * gimple.c (gimple_could_trap_p_1): Internal function calls
11585 * tree-eh.c (tree_could_trap_p): Likewise.
11587 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
11590 * config/arm/arm_neon.h (vmul_n_u32): Replace call to builtin with
11592 (vmulq_n_u32): Likewise.
11593 (vmul_n_f32): Gate __a * __b on __FAST_MATH__.
11594 (vmulq_n_f32): Likewise.
11595 (vmul_n_f16): Likewise.
11596 (vmulq_n_f16): Likewise.
11598 2021-07-12 Martin Liska <mliska@suse.cz>
11600 PR sanitizer/101425
11601 * gcc.c (check_offload_target_name): Call
11602 candidates_list_and_hint only if we have a candidate.
11604 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
11607 * config/arm/neon.md (vec_init): Move to ...
11608 * config/arm/vec-common.md (vec_init): ... here.
11609 Change the pattern's mode to VDQX and gate it on VALID_MVE_MODE.
11611 2021-07-12 Roger Sayle <roger@nextmovesoftware.com>
11613 PR tree-optimization/101403
11614 * match.pd ((T)bswap(X)>>C): Correctly handle cases where
11615 signedness of the shift is not the same as the signedness of
11616 the type extension.
11618 2021-07-09 Roger Sayle <roger@nextmovesoftware.com>
11619 Uroš Bizjak <ubizjak@gmail.com>
11621 * config/i386/i386.md (*divmodsi4_const): Optimize SImode
11622 divmod of a constant numerator with new define_insn_and_split.
11624 2021-07-09 Iain Sandoe <iain@sandoe.co.uk>
11627 * config/i386/i386-expand.c (ix86_expand_call): If a call is
11628 to a non-local-binding, or local but to a public symbol, then
11629 assume that it might be indirected via the lazy symbol binder.
11630 Mark R10 and R10 as clobbered in that case.
11632 2021-07-09 Eric Botcazou <ebotcazou@adacore.com>
11635 * gcc.c (ASM_DEBUG_DWARF_OPTION): Set again to --gdwarf2 in
11636 the case where HAVE_AS_WORKING_DWARF_N_FLAG is not defined
11637 and HAVE_LD_BROKEN_PE_DWARF5 is defined.
11639 2021-07-09 Uroš Bizjak <ubizjak@gmail.com>
11641 * config/i386/i386.md (*udivmodsi4_pow2_zext_1): Limit the
11642 log2 range of operands[3] to [1,31].
11643 (*udivmodsi4_pow2_zext_2): Ditto. Correct insn RTX pattern.
11645 2021-07-09 Sergei Trofimovich <siarheit@google.com>
11647 * doc/md.texi: Don't split @smallexample in multiple @groups.
11649 2021-07-09 Sergei Trofimovich <siarheit@google.com>
11651 * doc/md.texi: Add missing 'see' word.
11653 2021-07-09 Andrew Pinski <apinski@marvell.com>
11655 * tree-ssa-phiopt.c (phiopt_early_allow): Change arguments
11656 to take sequence and gimple_match_op. Accept the case where
11657 op is a SSA_NAME and one statement in the sequence.
11658 Also allow constants.
11659 (gimple_simplify_phiopt): Always pass a sequence to resimplify.
11660 Update call to phiopt_early_allow. Discard the sequence if not
11663 2021-07-09 Xi Ruoyao <xry111@mengyan1223.wang>
11668 * config/mips/mips.c (mips_const_insns): Use MSA_SUPPORTED_MODE_P
11669 instead of ISA_HAS_MSA.
11670 (mips_expand_vec_unpack): Likewise.
11671 (mips_expand_vector_init): Likewise.
11673 2021-07-09 Kewen Lin <linkw@linux.ibm.com>
11675 * config/rs6000/vsx.md (mods_<mode>): Rename to...
11676 (mod<mode>3): ... this.
11677 (modu_<mode>): Rename to...
11678 (umod<mode>3): ... this.
11679 * config/rs6000/rs6000-builtin.def (MODS_V2DI, MODS_V4SI, MODU_V2DI,
11680 MODU_V4SI): Adjust.
11682 2021-07-08 Jeff Law <jeffreyalaw@gmail.com>
11684 * config/h8300/shiftrotate.md (variable shifts): Expose condition
11685 code handling for the test before the loop.
11687 2021-07-08 Martin Jambor <mjambor@suse.cz>
11690 * ipa-sra.c (class isra_call_summary): New member
11691 m_before_any_store, initialize it in the constructor.
11692 (isra_call_summary::dump): Dump the new field.
11693 (ipa_sra_call_summaries::duplicate): Copy it.
11694 (process_scan_results): Set it.
11695 (isra_write_edge_summary): Stream it.
11696 (isra_read_edge_summary): Likewise.
11697 (param_splitting_across_edge): Only override
11698 safe_to_import_accesses if m_before_any_store is set.
11700 2021-07-08 Martin Sebor <msebor@redhat.com>
11702 PR bootstrap/101374
11703 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref):
11704 Use Object Size Type 0 instead of 1.
11706 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
11708 * tree-vect-loop.c (vectorizable_reduction): Remove always-true
11711 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
11713 * match.pd: Simplify an extend-operate-truncate sequence involving
11716 2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
11717 Richard Biener <rguenther@suse.de>
11719 PR tree-optimization/40210
11720 * match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
11721 (x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
11722 when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
11724 2021-07-08 Uroš Bizjak <ubizjak@gmail.com>
11727 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
11729 * config/i386/mmx.md (V_32): New mode iterator.
11730 (mov<V_32:mode>): Use V_32 mode iterator.
11731 (*mov<V_32:mode>_internal): Ditto.
11732 (*push<V_32:mode>2_rex64): Ditto.
11733 (*push<V_32:mode>2): Ditto.
11734 (movmisalign<V_32:mode>): Ditto.
11735 (mmx_<any_shiftrt:insn>v1si3): New insn pattern.
11736 (sse4_1_<any_extend:code>v2qiv2hi2): Ditto.
11737 (vec_unpacks_lo_v4qi): New expander.
11738 (vec_unpacks_hi_v4qi): Ditto.
11739 (vec_unpacku_lo_v4qi): Ditto.
11740 (vec_unpacku_hi_v4qi): Ditto.
11741 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add V1SImode.
11742 (VALID_INT_MODE_P): Ditto.
11744 2021-07-08 Michael Meissner <meissner@linux.ibm.com>
11747 * config/rs6000/rs6000.md (udivti3): New insn.
11748 (divti3): New insn.
11749 (umodti3): New insn.
11750 (modti3): New insn.
11752 2021-07-07 Martin Sebor <msebor@redhat.com>
11754 PR tree-optimization/100137
11755 PR tree-optimization/99121
11756 PR tree-optimization/97027
11757 * builtins.c (access_ref::access_ref): Also set offmax.
11758 (access_ref::offset_in_range): Define new function.
11759 (access_ref::add_offset): Set offmax.
11760 (access_ref::inform_access): Handle access_none.
11761 (handle_mem_ref): Clear ostype.
11762 (compute_objsize_r): Handle ASSERT_EXPR.
11763 * builtins.h (struct access_ref): Add offmax member.
11764 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Use
11765 compute_objsize() and simplify.
11767 2021-07-07 Peter Bergner <bergner@linux.ibm.com>
11769 * config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
11770 and VSX_BUILTIN_STXVP.
11772 2021-07-07 Martin Sebor <msebor@redhat.com>
11775 * config/aarch64/aarch64.c (aarch64_simd_lane_bounds): Remove
11776 a stray %K from error_at() missed in r12-2088.
11778 2021-07-07 Richard Biener <rguenther@suse.de>
11780 PR tree-optimization/99728
11781 * tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
11783 (mem_refs_may_alias_p): Add assert we handled aggregate
11785 (sm_seq_valid_bb): Give up when running into aggregate copies.
11786 (ref_indep_loop_p): Handle aggregate copies as never
11787 being invariant themselves but allow other refs to be
11788 disambiguated against them.
11789 (can_sm_ref_p): Do not try to apply store-motion to aggregate
11792 2021-07-06 Indu Bhagat <indu.bhagat@oracle.com>
11795 * dwarf2ctf.c (ctf_get_AT_data_member_location): Multiply by 8 to get
11798 2021-07-06 Martin Sebor <msebor@redhat.com>
11800 * gimple-pretty-print.c (percent_G_format): Remove.
11801 * tree-diagnostic.c (default_tree_printer): Remove calls.
11802 * tree-pretty-print.c (percent_K_format): Remove.
11803 * tree-pretty-print.h (percent_K_format): Remove.
11805 2021-07-06 Martin Sebor <msebor@redhat.com>
11807 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_builtin):
11808 Remove %K and use error_at.
11809 (aarch64_expand_fcmla_builtin): Same.
11810 (aarch64_expand_builtin_tme): Same.
11811 (aarch64_expand_builtin_memtag): Same.
11812 * config/arm/arm-builtins.c (arm_expand_acle_builtin): Same.
11813 (arm_expand_builtin): Same.
11814 * config/arm/arm.c (bounds_check): Same.
11816 2021-07-06 Martin Sebor <msebor@redhat.com>
11818 * builtins.c (warn_string_no_nul): Remove %G.
11819 (maybe_warn_for_bound): Same.
11820 (warn_for_access): Same.
11821 (check_access): Same.
11822 (check_strncat_sizes): Same.
11823 (expand_builtin_strncat): Same.
11824 (expand_builtin_strncmp): Same.
11825 (expand_builtin): Same.
11826 (expand_builtin_object_size): Same.
11827 (warn_dealloc_offset): Same.
11828 (maybe_emit_free_warning): Same.
11829 * calls.c (maybe_warn_alloc_args_overflow): Same.
11830 (maybe_warn_nonstring_arg): Same.
11831 (maybe_warn_rdwr_sizes): Same.
11832 * expr.c (expand_expr_real_1): Remove %K.
11833 * gimple-fold.c (gimple_fold_builtin_strncpy): Remove %G.
11834 (gimple_fold_builtin_strncat): Same.
11835 * gimple-ssa-sprintf.c (format_directive): Same.
11836 (handle_printf_call): Same.
11837 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Same.
11838 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
11839 (maybe_diag_access_bounds): Same. Call gimple_location.
11840 (check_bounds_or_overlap): Same.
11841 * trans-mem.c (ipa_tm_scan_irr_block): Remove %K. Simplify.
11842 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Remove %G.
11843 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
11844 (maybe_diag_stxncpy_trunc): Same.
11845 (handle_builtin_stxncpy_strncat): Same.
11846 (maybe_warn_pointless_strcmp): Same.
11847 * tree-ssa-uninit.c (maybe_warn_operand): Same.
11849 2021-07-06 Uroš Bizjak <ubizjak@gmail.com>
11852 * config/i386/predicates.md (vec_setm_operand): Enable
11853 register_operand for TARGET_SSE4_1.
11854 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_operand
11855 as operand 2 predicate. Call ix86_expand_vector_set_var
11856 for non-constant index operand.
11857 (vec_setv4qi): Use vec_setm_mmx_operand as operand 2 predicate.
11858 Call ix86_expand_vector_set_var for non-constant index operand.
11860 2021-07-06 Jeff Law <jeffreyalaw@gmail.com>
11862 * config/h8300/jumpcall.md (*branch): When possible, generate
11863 the comparison in CCZN mode.
11864 * config/h8300/predicates.md (simple_memory_operand): Reject all
11865 auto-increment addressing modes.
11867 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
11869 PR bootstrap/100246
11870 * config/i386/i386.h (struct stringop_algs): Define a CTOR for
11873 2021-07-06 Richard Biener <rguenther@suse.de>
11875 * doc/md.texi (vec_fmaddsub<mode>4): Document.
11876 (vec_fmsubadd<mode>4): Likewise.
11877 * optabs.def (vec_fmaddsub$a4): Add.
11878 (vec_fmsubadd$a4): Likewise.
11879 * internal-fn.def (IFN_VEC_FMADDSUB): Add.
11880 (IFN_VEC_FMSUBADD): Likewise.
11881 * tree-vect-slp-patterns.c (addsub_pattern::recognize):
11882 Refactor to handle IFN_VEC_FMADDSUB and IFN_VEC_FMSUBADD.
11883 (addsub_pattern::build): Likewise.
11884 * tree-vect-slp.c (vect_optimize_slp): CFN_VEC_FMADDSUB
11885 and CFN_VEC_FMSUBADD are not transparent for permutes.
11886 * config/i386/sse.md (vec_fmaddsub<mode>4): New expander.
11887 (vec_fmsubadd<mode>4): Likewise.
11889 2021-07-06 Richard Biener <rguenther@suse.de>
11891 * doc/invoke.texi (fmove-loop-stores): Document.
11892 * common.opt (fmove-loop-stores): New option.
11893 * opts.c (default_options_table): Enable -fmove-loop-stores
11894 at -O1 but not -Og.
11895 * tree-ssa-loop-im.c (pass_lim::execute): Pass
11896 flag_move_loop_stores instead of true to
11897 loop_invariant_motion_in_fun.
11899 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
11901 * doc/install.texi: Document --with-dsymutil.
11903 2021-07-06 Andrew Pinski <apinski@marvell.com>
11905 PR tree-optimization/101256
11906 * dbgcnt.def (phiopt_edge_range): New counter.
11907 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
11908 Check to make sure the new name is defined in the same
11909 bb as the conditional before duplicating range info.
11910 Also add debug counter.
11912 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
11914 PR rtl-optimization/100328
11915 * config/i386/i386-options.c (ix86_option_override_internal):
11916 Set param_ira_consider_dup_in_all_alts to 0.
11918 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
11920 PR rtl-optimization/100328
11921 * doc/invoke.texi (ira-consider-dup-in-all-alts): Document new
11923 * ira.c (ira_get_dup_out_num): Adjust as parameter
11924 param_ira_consider_dup_in_all_alts.
11925 * params.opt (ira-consider-dup-in-all-alts): New.
11926 * ira-conflicts.c (process_regs_for_copy): Add one parameter
11927 single_input_op_has_cstr_p.
11928 (get_freq_for_shuffle_copy): New function.
11929 (add_insn_allocno_copies): Adjust as single_input_op_has_cstr_p.
11930 * ira-int.h (ira_get_dup_out_num): Add one bool parameter.
11932 2021-07-05 Jeff Law <jeffreyalaw@gmail.com>
11934 * config/h8300/shiftrotate.md (shift-by-variable patterns): Update to
11935 generate condition code aware RTL directly.
11937 2021-07-05 Andrew Pinski <apinski@marvell.com>
11939 PR tree-optimization/101039
11940 * match.pd (A CMP 0 ? A : -A): New patterns.
11941 * tree-ssa-phiopt.c (abs_replacement): Delete function.
11942 (tree_ssa_phiopt_worker): Don't call abs_replacement.
11943 Update comment about abs_replacement.
11945 2021-07-05 Andrew Pinski <apinski@marvell.com>
11947 * tree-ssa-phiopt.c (gimple_simplify_phiopt):
11948 If "A ? B : C" fails to simplify, try "(!A) ? C : B".
11950 2021-07-05 Andrew Pinski <apinski@marvell.com>
11952 * tree-ssa-phiopt.c (match_simplify_replacement):
11953 Add early_p argument. Call gimple_simplify_phiopt
11954 instead of gimple_simplify.
11955 (tree_ssa_phiopt_worker): Update call to
11956 match_simplify_replacement and allow unconditionally.
11957 (phiopt_early_allow): New function.
11958 (gimple_simplify_phiopt): New function.
11960 2021-07-05 Andrew Pinski <apinski@marvell.com>
11962 PR middle-end/101237
11963 * fold-const.c (negate_expr_p): Remove call to element_mode
11964 and TREE_MODE/TREE_TYPE when calling HONOR_SIGNED_ZEROS,
11965 HONOR_SIGN_DEPENDENT_ROUNDING, and HONOR_SNANS.
11966 (fold_negate_expr_1): Likewise.
11967 (const_unop): Likewise.
11968 (fold_cond_expr_with_comparison): Likewise.
11969 (fold_binary_loc): Likewise.
11970 (fold_ternary_loc): Likewise.
11971 (tree_call_nonnegative_warnv_p): Likewise.
11972 * match.pd (-(A + B) -> (-B) - A): Likewise.
11974 2021-07-05 Iain Sandoe <iain@sandoe.co.uk>
11976 * configure.ac: Handle --with-dsymutil in the same way as we
11977 do for the assembler and linker. (DEFAULT_DSYMUTIL): New.
11978 Extract the type and version for the dsymutil configured or
11979 found by the default searches.
11980 * config.in: Regenerated.
11981 * configure: Regenerated.
11982 * collect2.c (do_dsymutil): Handle locating dsymutil in the
11983 same way as for the assembler and linker.
11984 * config/darwin.h (DSYMUTIL): Delete.
11985 * gcc.c: Report a configured dsymutil correctly.
11986 * exec-tool.in: Allow for dsymutil.
11988 2021-07-05 Uroš Bizjak <ubizjak@gmail.com>
11990 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
11991 Handle V4QI and V2HI modes.
11992 (expand_vec_perm_blend): Allow 4-byte vector modes with TARGET_SSE4_1.
11993 Handle V4QI mode. Emit mmx_pblendvb32 for 4-byte modes.
11994 (expand_vec_perm_pshufb): Rewrite to use switch statemets.
11995 Handle 4-byte dual operands with TARGET_XOP and single operands
11996 with TARGET_SSSE3. Emit mmx_ppermv32 for TARGET_XOP and
11997 mmx_pshufbv4qi3 for TARGET_SSSE3.
11998 (expand_vec_perm_pblendv): Allow 4-byte vector modes with TARGET_SSE4_1.
11999 (expand_vec_perm_interleave2): Allow 4-byte vector modes.
12000 (expand_vec_perm_pshufb2): Allow 4-byte vector modes with TARGET_SSSE3.
12001 (expand_vec_perm_even_odd_1): Handle V4QI mode.
12002 (expand_vec_perm_broadcast_1): Handle V4QI mode.
12003 (ix86_vectorize_vec_perm_const): Handle V4QI mode.
12004 * config/i386/mmx.md (mmx_ppermv32): New insn pattern.
12005 (mmx_pshufbv4qi3): Ditto.
12006 (*mmx_pblendw32): Ditto.
12007 (*mmx_pblendw64): Rename from *mmx_pblendw.
12008 (mmx_punpckhbw_low): New insn_and_split pattern.
12009 (mmx_punpcklbw_low): Ditto.
12011 2021-07-05 Richard Biener <rguenther@suse.de>
12013 * tree-vect-loop-manip.c (vect_loop_versioning): Do not
12014 set LOOP_C_INFINITE on the vectorized loop.
12016 2021-07-05 Richard Biener <rguenther@suse.de>
12018 PR middle-end/101291
12019 * cfgloopmanip.c (loop_version): Set the loop copy of the
12020 versioned loop to the new loop.
12022 2021-07-04 Iain Sandoe <iain@sandoe.co.uk>
12025 * config.gcc: Ensure that Darwin biarch definitions are
12026 added before i386.h.
12027 * config/i386/darwin.h (TARGET_64BIT): Remove.
12028 (PR80556_WORKAROUND): New.
12029 (REAL_LIBGCC_SPEC): Amend to use PR80556_WORKAROUND.
12030 (DARWIN_SUBARCH_SPEC): New.
12031 * config/i386/darwin32-biarch.h (TARGET_64BIT_DEFAULT,
12032 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
12033 (REAL_LIBGCC_SPEC): Remove.
12034 * config/i386/darwin64-biarch.h (TARGET_64BIT_DEFAULT,
12035 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
12036 (REAL_LIBGCC_SPEC): Remove.
12038 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
12040 PR middle-end/101294
12041 * expr.c (store_constructor): Don't use vec_duplicate on vector.
12043 2021-07-02 Martin Sebor <msebor@redhat.com>
12045 PR middle-end/98871
12046 PR middle-end/98512
12047 * diagnostic.c (get_any_inlining_info): New.
12048 (update_effective_level_from_pragmas): Handle inlining context.
12049 (diagnostic_enabled): Same.
12050 (diagnostic_report_diagnostic): Same.
12051 * diagnostic.h (struct diagnostic_info): Add ctor.
12052 (struct diagnostic_context): Add new member.
12053 * tree-diagnostic.c (set_inlining_locations): New.
12054 (tree_diagnostics_defaults): Set new callback pointer.
12056 2021-07-02 Peter Bergner <bergner@linux.ibm.com>
12058 * config/rs6000/rs6000-builtin.def (BU_MMA_PAIR_LD, BU_MMA_PAIR_ST):
12060 (__builtin_vsx_lxvp, __builtin_vsx_stxvp): New built-ins.
12061 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Expand
12062 lxvp and stxvp built-ins.
12063 (mma_init_builtins): Handle lxvp and stxvp built-ins.
12064 (builtin_function_type): Likewise.
12065 * doc/extend.texi (__builtin_vsx_lxvp, __builtin_mma_stxvp): Document.
12067 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
12069 * config/h8300/h8300-protos.h (compute_a_shift_cc): Accept
12070 additional argument for the code.
12071 * config/h8300/h8300.c (compute_a_shift_cc): Accept additional
12072 argument for the code. Just return if the ZN bits are useful or
12073 not rather than the old style CC_* enums.
12074 * config/h8300/shiftrotate.md (shiftqi_noscratch): Move before
12075 more generic shiftqi patterns.
12076 (shifthi_noscratch, shiftsi_noscratch): Similarly.
12077 (shiftqi_noscratch_set_flags): New pattern.
12078 (shifthi_noscratch_set_flags, shiftsi_noscratch_set_flags): Likewise.
12080 2021-07-02 Andrew MacLeod <amacleod@redhat.com>
12082 PR tree-optimization/101223
12083 * range-op.cc (build_lt): Add -1 for signed values.
12084 (built_gt): Subtract -1 for signed values.
12086 2021-07-02 David Faust <david.faust@oracle.com>
12088 * btfout.c (get_btf_kind): Support BTF_KIND_FLOAT.
12089 (btf_asm_type): Likewise.
12091 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
12093 * config/h8300/h8300-protos.h (output_a_shift): Make first argument
12094 an array of rtx rather than a pointer to rtx. Add code argument.
12095 (compute_a_shift_length): Similarly.
12096 * config/h8300/h8300.c (h8300_shift_costs): Adjust now that the
12097 shift itself isn't an operand. Create dummy operand[0] to carry
12098 a mode and pass a suitable rtx code to compute_a_shift_length.
12099 (get_shift_alg): Adjust operand number of clobber in output templates.
12100 (output_a_shift): Make first argument an array of rtx rather than
12101 a pointer to rtx. Add code argument for the type of shift.
12102 Adjust now that the shift itself is no longer an operand.
12103 (compute_a_shift_length): Similarly.
12104 * config/h8300/shiftrotate.md (shiftqi, shifthi, shiftsi): Use an
12105 iterator rather than nshift_operator.
12106 (shiftqi_noscratch, shifthi_noscratch, shiftsi_noscratch): Likewise.
12107 (shiftqi_clobber_flags): Adjust to API changes in output_a_shift
12108 and compute_a_shift_length.
12109 (shiftqi_noscratch_clobber_flags): Likewise.
12110 (shifthi_noscratch_clobber_flags): Likewise.
12111 (shiftsi_noscratch_clobber_flags): Likewise.
12113 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
12116 * config/darwin.h (DSYMUTIL_SPEC): Do not try to run
12117 dsymutil for BTF/CTF.
12119 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
12122 * config/darwin.h (CTF_INFO_SECTION_NAME): Update the
12123 segment to include BTF.
12124 (BTF_INFO_SECTION_NAME): New.
12126 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
12128 * config/m32r/m32r-protos.h (call_operand): Adjust return type.
12129 (small_data_operand, memreg_operand, small_insn_p): Likewise.
12130 * config/m32r/m32r.c (call_operand): Adjust return type.
12131 (small_data_operand, memreg_operand): Likewise.
12133 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
12135 * config/frv/frv-protos.h (integer_register_operand): Adjust return
12137 (frv_load_operand, gpr_or_fpr_operand, gpr_no_subreg_operand): Likewise.
12138 (fpr_or_int6_operand, gpr_or_int_operand); Likewise.
12139 (gpr_or_int12_operand, gpr_or_int10_operand); Likewise.
12140 (move_source_operand, move_destination_operand): Likewise.
12141 (condexec_source_operand, condexec_dest_operand): Likewise.
12142 (lr_operand, gpr_or_memory_operand, fpr_or_memory_operand): Likewise.
12143 (reg_or_0_operand, fcc_operand, icc_operand, cc_operand): Likewise.
12144 (fcr_operand, icr_operand, cr_operand, call_operand): Likewise.
12145 (fpr_operand, even_reg_operand, odd_reg_operand): Likewise.
12146 (even_gpr_operand, odd_gpr_operand, quad_fpr_operand): Likewise.
12147 (even_fpr_operand, odd_fpr_operand): Likewise.
12148 (dbl_memory_one_insn_operand, dbl_memory_two_insn_operand): Likewise.
12149 (int12_operand, int6_operand, int5_operand, uint5_operand): Likewise.
12150 (uint4_operand, uint1_operand, int_2word_operand): Likewise
12151 (upper_int16_operand, uint16_operand, symbolic_operand): Likewise.
12152 (relational_operator, float_relational_operator): Likewise.
12153 (ccr_eqne_operator, minmax_operator): Likewise.
12154 (condexec_si_binary_operator, condexec_si_media_operator): Likewise.
12155 (condexec_si_divide_operator, condexec_si_unary_operator): Likewise.
12156 (condexec_sf_conv_operator, condexec_sf_add_operator): Likewise.
12157 (intop_compare_operator, acc_operand, even_acc_operand): Likewise.
12158 (quad_acc_operand, accg_operand): Likewise.
12160 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
12162 * config/stormy16/stormy16-protos.h (xstormy16_below_100_symbol): Change
12163 return type to a bool.
12164 (nonimmediate_nonstack_operand): Likewise.
12165 (xstormy16_splittable_below100_operand): Likewise.
12166 * config/stormy16/stormy16.c (xstormy16_below_100_symbol): Fix
12168 (xstormy16_splittable_below100_operand): Likewise.
12170 2021-07-02 Richard Biener <rguenther@suse.de>
12172 PR tree-optimization/101293
12173 * tree-ssa-loop-im.c (mem_ref_hasher::equal): Compare MEM_REF bases
12174 with combined offsets.
12175 (gather_mem_refs_stmt): Hash MEM_REFs as if their offset were
12176 combined with the rest of the offset.
12178 2021-07-02 Eric Botcazou <ebotcazou@adacore.com>
12180 * config/i386/i386.c (asm_preferred_eh_data_format): Always use the
12181 PIC encodings for PE-COFF targets.
12183 2021-07-02 Jakub Jelinek <jakub@redhat.com>
12186 * config/i386/i386-expand.c (ix86_broadcast_from_integer_constant):
12187 Return nullptr for TImode inner mode.
12189 2021-07-02 Richard Biener <rguenther@suse.de>
12191 PR tree-optimization/101280
12192 PR tree-optimization/101173
12193 * gimple-loop-interchange.cc
12194 (tree_loop_interchange::valid_data_dependences): Properly
12195 guard all dependence checks with DDR_REVERSED_P or its
12198 2021-07-02 Hongyu Wang <hongyu.wang@intel.com>
12200 * config/i386/i386-expand.c (ix86_expand_builtin):
12201 Add branch to clear odata when ZF is set for asedecenc_expand
12202 and wideaesdecenc_expand.
12204 2021-07-02 Eugene Rozenfeld <erozen@microsoft.com>
12206 * config/i386/gcc-auto-profile: regenerate
12208 2021-07-02 liuhongt <hongtao.liu@intel.com>
12210 * config/i386/sse.md (trunc<mode><pmov_dst_4>2): Refined to ..
12211 (trunc<mode><pmov_dst_4_lower>2): this.
12213 2021-07-01 David Malcolm <dmalcolm@redhat.com>
12215 * diagnostic.h (diagnostic_context::m_file_cache): New field.
12216 * input.c (class fcache): Rename to...
12217 (class file_cache_slot): ...this, making most members private and
12218 prefixing fields with "m_".
12219 (file_cache_slot::get_file_path): New accessor.
12220 (file_cache_slot::get_use_count): New accessor.
12221 (file_cache_slot::missing_trailing_newline_p): New accessor.
12222 (file_cache_slot::inc_use_count): New.
12223 (fcache_buffer_size): Move to...
12224 (file_cache_slot::buffer_size): ...here.
12225 (fcache_line_record_size): Move to...
12226 (file_cache_slot::line_record_size): ...here.
12227 (fcache_tab): Delete, in favor of global_dc->m_file_cache.
12228 (fcache_tab_size): Move to file_cache::num_file_slots.
12229 (diagnostic_file_cache_init): Update for move of fcache_tab
12230 to global_dc->m_file_cache.
12231 (diagnostic_file_cache_fini): Likewise.
12232 (lookup_file_in_cache_tab): Convert to...
12233 (file_cache::lookup_file): ...this.
12234 (diagnostics_file_cache_forcibly_evict_file): Update for move of
12235 fcache_tab to global_dc->m_file_cache, moving most of
12236 implementation to...
12237 (file_cache::forcibly_evict_file): ...this new function and...
12238 (file_cache_slot::evict): ...this new function.
12239 (evicted_cache_tab_entry): Convert to...
12240 (file_cache::evicted_cache_tab_entry): ...this.
12241 (add_file_to_cache_tab): Convert to...
12242 (file_cache::add_file): ...this, moving bulk of implementation
12244 (file_cache_slot::create): ..this new function.
12245 (file_cache::file_cache): New.
12246 (file_cache::~file_cache): New.
12247 (lookup_or_add_file_to_cache_tab): Convert to...
12248 (file_cache::lookup_or_add_file): ..this new function.
12249 (fcache::fcache): Rename to...
12250 (file_cache_slot::file_cache_slot): ...this, adding "m_" prefixes
12252 (fcache::~fcache): Rename to...
12253 (file_cache_slot::~file_cache_slot): ...this, adding "m_" prefixes
12255 (needs_read): Convert to...
12256 (file_cache_slot::needs_read_p): ...this.
12257 (needs_grow): Convert to...
12258 (file_cache_slot::needs_grow_p): ...this.
12259 (maybe_grow): Convert to...
12260 (file_cache_slot::maybe_grow): ...this.
12261 (read_data): Convert to...
12262 (file_cache_slot::read_data): ...this.
12263 (maybe_read_data): Convert to...
12264 (file_cache_slot::maybe_read_data): ...this.
12265 (get_next_line): Convert to...
12266 (file_cache_slot::get_next_line): ...this.
12267 (goto_next_line): Convert to...
12268 (file_cache_slot::goto_next_line): ...this.
12269 (read_line_num): Convert to...
12270 (file_cache_slot::read_line_num): ...this.
12271 (location_get_source_line): Update for moving of globals to
12272 global_dc->m_file_cache.
12273 (location_missing_trailing_newline): Likewise.
12274 * input.h (class file_cache_slot): New forward decl.
12275 (class file_cache): New.
12277 2021-07-01 Michael Meissner <meissner@linux.ibm.com>
12279 * config/rs6000/rs6000.c (rs6000_maybe_emit_fp_cmove): Add IEEE
12280 128-bit floating point conditional move support.
12281 (have_compare_and_set_mask): Add IEEE 128-bit floating point
12283 * config/rs6000/rs6000.md (mov<mode>cc, IEEE128 iterator): New insn.
12284 (mov<mode>cc_p10, IEEE128 iterator): New insn.
12285 (mov<mode>cc_invert_p10, IEEE128 iterator): New insn.
12286 (fpmask<mode>, IEEE128 iterator): New insn.
12287 (xxsel<mode>, IEEE128 iterator): New insn.
12289 2021-07-01 Iain Sandoe <iain@sandoe.co.uk>
12292 * config/darwin.h (CTF_INFO_SECTION_NAME): New.
12294 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
12296 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
12298 * config/i386/i386-protos.h (ix86_expand_vector_init_duplicate):
12300 * config/i386/sse.md (INT_BROADCAST_MODE): New mode iterator.
12301 (vec_duplicate<mode>): New expander.
12303 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
12306 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
12308 (ix86_byte_broadcast): New function.
12309 (ix86_convert_const_wide_int_to_broadcast): Likewise.
12310 (ix86_expand_move): Convert CONST_WIDE_INT to broadcast if mode
12311 size is 16 bytes or bigger.
12312 (ix86_broadcast_from_integer_constant): New function.
12313 (ix86_expand_vector_move): Convert CONST_WIDE_INT and CONST_VECTOR
12314 to broadcast if mode size is 16 bytes or bigger.
12315 * config/i386/i386-protos.h (ix86_gen_scratch_sse_rtx): New
12317 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): New function.
12319 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
12321 * config/i386/predicates.md (ix86_endbr_immediate_operand):
12322 Return true/false instead of 1/0.
12323 (movq_parallel): Ditto.
12325 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
12327 * recog.c (general_operand): Return true/false instead of 1/0.
12328 (register_operand): Ditto.
12329 (immediate_operand): Ditto.
12330 (const_int_operand): Ditto.
12331 (const_scalar_int_operand): Ditto.
12332 (const_double_operand): Ditto.
12333 (push_operand): Ditto.
12334 (pop_operand): Ditto.
12335 (memory_operand): Ditto.
12336 (indirect_operand): Ditto.
12338 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
12340 * genpreds.c (write_predicate_subfunction):
12341 Change the type of written subfunction to bool.
12342 (write_one_predicate_function):
12343 Change the type of written function to bool.
12344 (write_tm_preds_h): Ditto.
12345 * recog.h (*insn_operand_predicate_fn): Change the type to bool.
12346 * recog.c (general_operand): Change the type to bool.
12347 (address_operand): Ditto.
12348 (register_operand): Ditto.
12349 (pmode_register_operand): Ditto.
12350 (scratch_operand): Ditto.
12351 (immediate_operand): Ditto.
12352 (const_int_operand): Ditto.
12353 (const_scalar_int_operand): Ditto.
12354 (const_double_operand): Ditto.
12355 (nonimmediate_operand): Ditto.
12356 (nonmemory_operand): Ditto.
12357 (push_operand): Ditto.
12358 (pop_operand): Ditto.
12359 (memory_operand): Ditto.
12360 (indirect_operand): Ditto.
12361 (ordered_comparison_operator): Ditto.
12362 (comparison_operator): Ditto.
12363 * config/i386/i386-expand.c (ix86_expand_sse_cmp):
12364 Change the type of indirect predicate function to bool.
12365 * config/rs6000/rs6000.c (easy_vector_constant):
12366 Change the type to bool.
12367 * config/mips/mips-protos.h (m16_based_address_p):
12368 Change the type of operand 3 to bool.
12370 2021-07-01 Richard Biener <rguenther@suse.de>
12372 PR tree-optimization/101280
12373 PR tree-optimization/101173
12374 * gimple-loop-interchange.cc
12375 (tree_loop_interchange::valid_data_dependences): Revert
12376 previous change and instead correctly handle DDR_REVERSED_P
12379 2021-07-01 Richard Biener <rguenther@suse.de>
12381 PR tree-optimization/101278
12382 * tree-ssa-dse.c (dse_classify_store): First check for
12383 uses, then ignore stmt for chaining purposes.
12385 2021-07-01 Richard Biener <rguenther@suse.de>
12387 PR tree-optimization/100778
12388 * tree-vect-slp.c (vect_schedule_slp_node): Do not place trapping
12389 vectorized ops ahead of their scalar BB.
12391 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
12394 * config/i386/i386.md (*nabs<dwi>2_doubleword):
12395 New insn_and_split pattern.
12396 (*nabs<dwi>2_1): Ditto.
12397 * config/i386/i386-features.c
12398 (general_scalar_chain::compute_convert_gain):
12399 Handle (NEG (ABS (...))) RTX. Rewrite src code
12400 scanner as switch statement.
12401 (general_scalar_chain::convert_insn):
12402 Handle (NEG (ABS (...))) RTX.
12403 (general_scalar_to_vector_candidate_p):
12404 Detect (NEG (ABS (...))) RTX. Reorder case statements
12405 for (AND (NOT (...) ...)) fallthrough.
12407 2021-07-01 Richard Biener <rguenther@suse.de>
12409 PR tree-optimization/101178
12410 * tree-vect-slp.c (slpg_vertex::materialize): Remove.
12411 (slpg::perm_in): Add.
12412 (slpg::get_perm_in): Remove.
12413 (slpg::get_perm_materialized): Add.
12414 (vect_optimize_slp): Handle VEC_PERM nodes more optimally
12415 during permute propagation and materialization.
12417 2021-07-01 Jakub Jelinek <jakub@redhat.com>
12420 * dwarf2out.c (loc_list_from_tree_1): Handle COMPOUND_LITERAL_EXPR.
12422 2021-07-01 Jakub Jelinek <jakub@redhat.com>
12424 PR middle-end/94366
12425 * omp-low.c (lower_rec_input_clauses): Rename is_fp_and_or to
12426 is_truth_op, set it for TRUTH_*IF_EXPR regardless of new_var's type,
12427 use boolean_type_node instead of integer_type_node as NE_EXPR type.
12428 (lower_reduction_clauses): Likewise.
12430 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
12432 * config/gcn/gcn.c: Include dwarf2.h.
12433 (gcn_addr_space_debug): New function.
12434 (TARGET_ADDR_SPACE_DEBUG): New hook.
12436 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
12438 * common/config/gcn/gcn-common.c
12439 (gcn_option_optimization_table): Change OPT_fomit_frame_pointer to -O3.
12440 * config/gcn/gcn.c (gcn_expand_prologue): Prefer the frame pointer
12442 (gcn_expand_prologue): Prefer the frame pointer when emitting CFI.
12443 (gcn_frame_pointer_rqd): New function.
12444 (TARGET_FRAME_POINTER_REQUIRED): New hook.
12446 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
12448 * config/gcn/gcn.c (move_callee_saved_registers): Emit CFI notes for
12449 prologue register saves.
12450 (gcn_debug_unwind_info): Use UI_DWARF2.
12451 (gcn_dwarf_register_number): Map DWARF_LINK_REGISTER to DWARF PC.
12452 (gcn_dwarf_register_span): DWARF_LINK_REGISTER doesn't span.
12453 * config/gcn/gcn.h: (DWARF_FRAME_RETURN_COLUMN): New define.
12454 (DWARF_LINK_REGISTER): New define.
12455 (FIRST_PSEUDO_REGISTER): Increment.
12456 (FIXED_REGISTERS): Add entry for DWARF_LINK_REGISTER.
12457 (CALL_USED_REGISTERS): Likewise.
12458 (REGISTER_NAMES): Likewise.
12460 2021-06-30 Richard Biener <rguenther@suse.de>
12462 PR tree-optimization/101267
12463 * tree-vect-stmts.c (vect_check_scalar_mask): Adjust
12464 API and use SLP compatible interface of vect_is_simple_use.
12465 Reject not vectorized SLP defs for callers that do not support
12467 (vect_check_store_rhs): Handle masked stores and pass down
12468 the appropriate operator index.
12469 (vectorizable_call): Adjust.
12470 (vectorizable_store): Likewise.
12471 (vectorizable_load): Likewise. Handle SLP pecularity of
12473 (vect_is_simple_use): Remove special-casing of masked stores.
12475 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
12477 * common.opt (foffload): Remove help as Driver only.
12478 * gcc.c (display_help): Add -foffload.
12480 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
12482 * gcc.c (close_at_file, execute): Replace alloca by XALLOCAVEC.
12483 (check_offload_target_name): Fix splitting OFFLOAD_TARGETS into
12484 a candidate list; better inform no offload target is configured
12485 and fix hint extraction when passed target is not '\0' at [len].
12486 * common.opt (foffload): Add tailing '.'.
12487 (foffload-options): Likewise; fix flag name in the help string.
12489 2021-06-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
12492 * config/arm/arm_neon.h: Move vabs intrinsics before vcage_f32.
12493 (vcage_f32): Gate comparison on __FAST_MATH__.
12494 (vcageq_f32): Likewise.
12495 (vcale_f32): Likewise.
12496 (vcaleq_f32): Likewise.
12497 (vcagt_f32): Likewise.
12498 (vcagtq_f32): Likewise.
12499 (vcalt_f32): Likewise.
12500 (vcaltq_f32): Likewise.
12501 (vcage_f16): Likewise.
12502 (vcageq_f16): Likewise.
12503 (vcale_f16): Likewise.
12504 (vcaleq_f16): Likewise.
12505 (vcagt_f16): Likewise.
12506 (vcagtq_f16): Likewise.
12507 (vcalt_f16): Likewise.
12508 (vcaltq_f16): Likewise.
12510 2021-06-30 Richard Biener <rguenther@suse.de>
12512 PR tree-optimization/101264
12513 * tree-vect-slp.c (vect_optimize_slp): Propagate the
12514 computed perm_in to all "any" permute successors
12515 we cannot de-duplicate immediately.
12517 2021-06-30 liuhongt <hongtao.liu@intel.com>
12520 * config/i386/sse.md
12521 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>):
12523 (avx512f_sfixupimm<mode><maskz_scalar_name><round_saeonly_name>):
12525 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>"): Refined.
12526 * config/i386/subst.md (maskz_scalar): New define_subst.
12527 (maskz_scalar_name): New subst_attr.
12528 (maskz_scalar_op5): Ditto.
12529 (round_saeonly_maskz_scalar_op5): Ditto.
12530 (round_saeonly_maskz_scalar_operand5): Ditto.
12532 2021-06-30 David Edelsohn <dje.gcc@gmail.com>
12534 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
12535 Increase code CSECT alignment to at least 32 bytes.
12536 * config/rs6000/xcoff.h (TEXT_SECTION_ASM_OP): Add 32 byte
12537 alignment designation.
12539 2021-06-29 Sergei Trofimovich <siarheit@google.com>
12541 * doc/generic.texi: Fix s/net yet/not yet/ typo.
12543 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
12545 PR tree-optimization/101254
12546 * range-op.cc (operator_minus::op1_op2_relation_effect): Check for
12547 wrapping/non-wrapping when setting the result range.
12549 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
12551 * value-query.cc (gimple_range_global): Allow phis.
12553 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
12555 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): Use stmt.
12556 (simplify_using_ranges::op_with_boolean_value_range_p): Add a
12557 statement for location context.
12558 (check_for_binary_op_overflow): Ditto.
12559 (simplify_using_ranges::get_vr_for_comparison): Ditto.
12560 (simplify_using_ranges::compare_name_with_value): Ditto.
12561 (simplify_using_ranges::compare_names): Ditto.
12562 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Ditto.
12563 (simplify_using_ranges::simplify_truth_ops_using_ranges): Ditto.
12564 (simplify_using_ranges::simplify_min_or_max_using_ranges): Ditto.
12565 (simplify_using_ranges::simplify_internal_call_using_ranges): Ditto.
12566 (simplify_using_ranges::two_valued_val_range_p): Ditto.
12567 (simplify_using_ranges::simplify): Ditto.
12568 * vr-values.h: Adjust prototypes.
12570 2021-06-29 Uroš Bizjak <ubizjak@gmail.com>
12573 * config/i386/mmx.md (vec_addsubv2sf3): New insn pattern.
12575 2021-06-29 Julian Brown <julian@codesourcery.com>
12577 * config/gcn/gcn.c (gcn_init_libfuncs): New function.
12578 (TARGET_INIT_LIBFUNCS): Define target hook using above function.
12579 * config/gcn/gcn.h (UNITS_PER_WORD): Define to 8 for IN_LIBGCC2, 4
12581 (LIBGCC2_UNITS_PER_WORD, BITS_PER_WORD): Remove definitions.
12582 (MAX_FIXED_MODE_SIZE): Change to 128.
12584 2021-06-29 Julian Brown <julian@codesourcery.com>
12586 * config/gcn/gcn.md (UNSPEC_FLBIT_INT): New unspec constant.
12587 (s_mnemonic): Add clrsb.
12588 (gcn_flbit<mode>_int): Add insn pattern for SImode/DImode.
12589 (clrsb<mode>2): Add expander for SImode/DImode.
12591 2021-06-29 Julian Brown <julian@codesourcery.com>
12593 * config/gcn/gcn.md (<su>mulsidi3, <su>mulsidi3_reg, <su>mulsidi3_imm,
12594 muldi3): Add patterns.
12596 2021-06-29 Julian Brown <julian@codesourcery.com>
12598 * config/gcn/gcn.md (<su>mulsi3_highpart): Change to expander.
12599 (<su>mulsi3_highpart_reg, <su>mulsi3_highpart_imm): New patterns.
12601 2021-06-29 Julian Brown <julian@codesourcery.com>
12603 * config/gcn/gcn.md (mulsi3): Make s_mulk_i32 variant clobber SCC.
12605 2021-06-29 Joseph Myers <joseph@codesourcery.com>
12607 * btfout.c, ctfout.c: Include "memmodel.h".
12609 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
12611 * gcc.c (check_offload_target_name): Cast len argument to
12612 %q.*s to 'int'; avoid -Wstringop-truncation warning.
12614 2021-06-29 Richard Biener <rguenther@suse.de>
12616 * tree-vect-slp.c (vect_optimize_slp): Forward propagate
12617 to "any" permute nodes and relax "any" permute proapgation
12618 during iterative backward propagation.
12620 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
12623 * common.opt (-foffload=): Update description.
12624 (-foffload-options=): New.
12625 * doc/invoke.texi (C Language Options): Document
12626 -foffload and -foffload-options.
12627 * gcc.c (check_offload_target_name): New, split off from
12628 handle_foffload_option.
12629 (check_foffload_target_names): New.
12630 (handle_foffload_option): Handle -foffload=default.
12631 (driver_handle_option): Update for -foffload-options.
12632 * lto-opts.c (lto_write_options): Use -foffload-options
12633 instead of -foffload.
12634 * lto-wrapper.c (merge_and_complain, append_offload_options):
12636 * opts.c (common_handle_option): Likewise.
12638 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
12640 * doc/invoke.texi (C Language Options): Sort options
12641 alphabetically in optlist and also the description itself.
12642 Remove leftover -fallow-single-precision from and add missing
12643 -fgnu-tm to the optlist.
12645 2021-06-29 Richard Biener <rguenther@suse.de>
12647 * tree-vect-slp.c (slpg_vertex::visited): Remove.
12648 (vect_slp_perms_eq): Handle -1 permutes.
12649 (vect_optimize_slp): Rewrite permute propagation.
12651 2021-06-29 Jakub Jelinek <jakub@redhat.com>
12654 * match.pd ((intptr_t)x eq/ne CST to x eq/ne (typeof x) CST): Don't
12655 perform the optimization in GENERIC when sanitizing and x has a
12658 2021-06-29 Richard Biener <rguenther@suse.de>
12660 PR tree-optimization/101242
12661 * tree-vect-slp.c (vect_slp_build_vertices): Force-add
12662 PHIs with not represented initial values as leafs.
12664 2021-06-29 Jan-Benedict Glaw <jbglaw@getslash.de>
12666 * config/pdp11/pdp11.h (ASM_OUTPUT_SKIP): Fix signedness warning.
12667 * config/pdp11/pdp11.c (pdp11_asm_print_operand_punct_valid_p): Remove
12668 "register" keyword.
12669 (pdp11_initial_elimination_offset) Remove unused variable.
12670 (pdp11_cmp_length) Ditto.
12671 (pdp11_insn_cost): Ditto, and fix signedness warning.
12673 2021-06-29 David Edelsohn <dje.gcc@gmail.com>
12675 * btfout.c: Include tm_p.h.
12678 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
12680 * config/bpf/bpf.c (bpf_expand_prologue): Do not mark insns as
12682 (bpf_expand_epilogue): Likewise.
12683 * config/bpf/bpf.h (DWARF2_FRAME_INFO): Define to 0.
12684 Do not define DBX_DEBUGGING_INFO.
12686 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
12688 * doc/invoke.texi: Document the CTF and BTF debug info options.
12690 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
12691 David Faust <david.faust@oracle.com>
12692 Jose E. Marchesi <jose.marchesi@oracle.com>
12693 Weimin Pan <weimin.pan@oracle.com>
12695 * Makefile.in: Add ctfc.*, ctfout.c and btfout.c files to
12696 GTFILES. Add new object files.
12697 * common.opt: Add CTF and BTF debug info options.
12698 * btfout.c: New file.
12699 * ctfc.c: Likewise.
12700 * ctfc.h: Likewise.
12701 * ctfout.c: Likewise.
12702 * dwarf2ctf.c: Likewise.
12703 * dwarf2ctf.h: Likewise.
12704 * dwarf2cfi.c (dwarf2out_do_frame): Acknowledge CTF_DEBUG and
12706 * dwarf2out.c (dwarf2out_source_line): Likewise.
12707 (dwarf2out_finish): Skip emitting DWARF if CTF or BTF are to
12709 (debug_format_do_cu): New function.
12710 (dwarf2out_early_finish): Traverse DIEs and emit CTF/BTF for
12712 Include dwarf2ctf.c.
12713 * final.c (dwarf2_debug_info_emitted_p): Acknowledge DWARF-based debug
12715 * flag-types.h (enum debug_info_type): Add CTF_DEBUG and BTF_DEBUG.
12716 (CTF_DEBUG): New bitmask.
12717 (BTF_DEBUG): Likewise.
12718 (enum ctf_debug_info_levels): New enum.
12719 * gengtype.c (open_base_files): Handle ctfc.h.
12720 (main): Handle uint32_t type.
12721 * flags.h (btf_debuginfo_p): New definition.
12722 (dwarf_based_debuginfo_p): Likewise.
12723 * opts.c (debug_type_names): Add entries for CTF and BTF.
12724 (btf_debuginfo_p): New function.
12725 (dwarf_based_debuginfo_p): Likewise.
12726 (common_handle_option): Handle -gctfN and -gbtf options.
12727 (set_debug_level): Set CTF_DEBUG, BTF_DEBUG whenever appropriate.
12728 * toplev.c (process_options): Inform the user and ignore -gctfLEVEL if
12731 2021-06-28 Jose E. Marchesi <jose.marchesi@oracle.com>
12733 * dwarf2out.c (AT_class): Function is no longer static.
12734 (AT_int): Likewise.
12735 (AT_unsigned): Likewise.
12736 (AT_loc): Likewise.
12737 (get_AT): Likewise.
12738 (get_AT_string): Likewise.
12739 (get_AT_flag): Likewise.
12740 (get_AT_unsigned): Likewise.
12741 (get_AT_ref): Likewise.
12742 (new_die_raw): Likewise.
12743 (lookup_decl_die): Likewise.
12744 (base_type_die): Likewise.
12745 (add_name_attribute): Likewise.
12746 (add_AT_int): Likewise.
12747 (add_AT_unsigned): Likewise.
12748 (add_AT_loc): Likewise.
12749 (dw_get_die_tag): New function.
12750 (dw_get_die_child): Likewise.
12751 (dw_get_die_sib): Likewise.
12752 (struct dwarf_file_data): Move from here to dwarf2out.h
12753 (struct dw_attr_struct): Likewise.
12754 * dwarf2out.h: Analogous changes.
12756 2021-06-28 Martin Jambor <mjambor@suse.cz>
12759 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
12760 members m_dead_stmts and m_dead_ssas.
12761 * ipa-param-manipulation.c
12762 (ipa_param_body_adjustments::mark_dead_statements): New function.
12763 (ipa_param_body_adjustments::common_initialization): Call it on
12764 all removed but not split parameters.
12765 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
12767 (ipa_param_body_adjustments::modify_call_stmt): Remove arguments that
12769 * tree-inline.c (remap_gimple_stmt): Do not copy dead statements, reset
12770 dead debug statements.
12771 (copy_phis_for_bb): Do not copy dead PHI nodes.
12773 2021-06-28 Martin Jambor <mjambor@suse.cz>
12776 * symtab-clones.h (clone_info): Removed member param_adjustments.
12777 * ipa-param-manipulation.h: Adjust initial comment to reflect how we
12778 deal with pass-through splits now.
12779 (ipa_param_performed_split): Removed.
12780 (ipa_param_adjustments::modify_call): Adjusted parameters.
12781 (class ipa_param_body_adjustments): Adjusted parameters of
12782 register_replacement, modify_gimple_stmt and modify_call_stmt.
12783 (ipa_verify_edge_has_no_modifications): Declare.
12784 (ipa_edge_modifications_finalize): Declare.
12785 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Remove
12786 performed_splits processing, pas only edge to padjs->modify_call,
12787 check that call arguments were not modified if they should not have
12789 * cgraphclones.c (cgraph_node::create_clone): Do not copy performed
12791 * ipa-param-manipulation.c (struct pass_through_split_map): New type.
12792 (ipa_edge_modification_info): Likewise.
12793 (ipa_edge_modification_sum): Likewise.
12794 (ipa_edge_modifications): New edge summary.
12795 (ipa_verify_edge_has_no_modifications): New function.
12796 (transitive_split_p): Removed.
12797 (transitive_split_map): Likewise.
12798 (init_transitive_splits): Likewise.
12799 (ipa_param_adjustments::modify_call): Adjusted to use the new edge
12800 summary instead of performed_splits.
12801 (ipa_param_body_adjustments::register_replacement): Drop dummy
12802 parameter, set base_index of the created ipa_param_body_replacement.
12803 (phi_arg_will_live_p): New function.
12804 (ipa_param_body_adjustments::common_initialization): Do not create
12805 IPA_SRA dummy decls.
12806 (simple_tree_swap_info): Removed.
12807 (remap_split_decl_to_dummy): Likewise.
12808 (record_argument_state_1): New function.
12809 (record_argument_state): Likewise.
12810 (ipa_param_body_adjustments::modify_call_stmt): New parameter
12811 orig_stmt. Do not work with dummy decls, save necessary info about
12812 changes to ipa_edge_modifications.
12813 (ipa_param_body_adjustments::modify_gimple_stmt): New parameter
12814 orig_stmt, pass it to modify_call_stmt.
12815 (ipa_param_body_adjustments::modify_cfun_body): Adjust call to
12816 modify_gimple_stmt.
12817 (ipa_edge_modifications_finalize): New function.
12818 * tree-inline.c (remap_gimple_stmt): Pass original statement to
12819 modify_gimple_stmt.
12820 (copy_phis_for_bb): Do not copy dead PHI nodes.
12821 (expand_call_inline): Do not remap performed_splits.
12822 (update_clone_info): Likewise.
12823 * toplev.c: Include ipa-param-manipulation.h.
12824 (toplev::finalize): Call ipa_edge_modifications_finalize.
12826 2021-06-28 Andrew Pinski <apinski@marvell.com>
12828 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Duplicate range
12829 info if we're the only things setting the target PHI.
12830 (value_replacement): Don't duplicate range here.
12831 (minmax_replacement): Likewise.
12833 2021-06-28 Richard Biener <rguenther@suse.de>
12835 PR tree-optimization/101229
12836 * gimple-walk.c (gimple_walk_op): Handle PHIs.
12838 2021-06-28 Martin Liska <mliska@suse.cz>
12840 * config/v850/v850.c (construct_dispose_instruction): Allocate
12842 (construct_prepare_instruction): Likewise.
12844 2021-06-28 Martin Liska <mliska@suse.cz>
12846 * config/v850/v850.c (v850_option_override): Build default
12848 (v850_can_inline_p): New. Allow MASK_PROLOG_FUNCTION to be
12849 ignored for inlining.
12850 (TARGET_CAN_INLINE_P): New.
12852 2021-06-28 Richard Biener <rguenther@suse.de>
12854 PR tree-optimization/101207
12855 * tree-vect-slp.c (vect_optimize_slp): Do BB reduction
12856 permute eliding for load permutations properly.
12858 2021-06-28 Richard Biener <rguenther@suse.de>
12860 PR tree-optimization/101173
12861 * gimple-loop-interchange.cc
12862 (tree_loop_interchange::valid_data_dependences): Disallow outer
12863 loop dependence distance of zero.
12865 2021-06-28 liuhongt <hongtao.liu@intel.com>
12868 * config/i386/sse.md (*avx_cmp<mode>3_lt): New
12869 define_insn_and_split.
12870 (*avx_cmp<mode>3_ltint): Ditto.
12871 (*avx2_pcmp<mode>3_3): Ditto.
12872 (*avx2_pcmp<mode>3_4): Ditto.
12873 (*avx2_pcmp<mode>3_5): Ditto.
12875 2021-06-28 liuhongt <hongtao.liu@intel.com>
12877 * config/i386/i386-builtin.def (IX86_BUILTIN_BLENDVPD256,
12878 IX86_BUILTIN_BLENDVPS256, IX86_BUILTIN_PBLENDVB256,
12879 IX86_BUILTIN_BLENDVPD, IX86_BUILTIN_BLENDVPS,
12880 IX86_BUILTIN_PBLENDVB128): Replace icode with
12882 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold blendv
12884 * config/i386/sse.md (*<sse4_1_avx2>_pblendvb_lt_subreg_not):
12885 New pre_reload splitter.
12887 2021-06-27 Andrew Pinski <apinski@marvell.com>
12889 PR middle-end/101230
12890 * fold-const.c (fold_ternary_loc): Check
12891 the return value of invert_tree_comparison.
12893 2021-06-27 David Edelsohn <dje.gcc@gmail.com>
12895 * config.gcc: Add SPDX License Identifier.
12896 (powerpc-ibm-aix789): Default to aix73.h.
12897 (powerpc-ibm-aix7.2.*.*): New stanza.
12898 * config/rs6000/aix72.h: Add SPDX License Identifier.
12899 * config/rs6000/aix73.h: New file.
12901 2021-06-26 Jason Merrill <jason@redhat.com>
12903 * except.c: #include "dwarf2.h" instead of "dwarf2out.h".
12905 2021-06-26 Andrew Pinski <apinski@marvell.com>
12907 * genmatch.c (lower_cond): Copy for_subst_vec
12908 for the simplify also.
12909 (lower): Swap the order for lower_for and lower_cond.
12911 2021-06-26 Andrew Pinski <apinski@marvell.com>
12913 * tree-ssa-phiopt.c (match_simplify_replacement): Reset
12914 flow senatitive info on the moved ssa set.
12916 2021-06-26 Andrew Pinski <apinski@marvell.com>
12918 * fold-const.c (fold_cond_expr_with_comparison):
12919 Exand arg0 into comp_code, arg00, and arg01.
12920 (fold_ternary_loc): Use invert_tree_comparison
12921 instead of fold_invert_truthvalue for the case
12922 where we have A CMP B ? C : A.
12924 2021-06-25 Martin Sebor <msebor@redhat.com>
12926 PR middle-end/101216
12927 * calls.c (maybe_warn_rdwr_sizes): Use the no_warning constant.
12929 2021-06-25 Jeff Law <jeffreyalaw@gmail.com>
12931 * config/h8300/h8300.c (select_cc_mode): Handle ASHIFTRT and LSHIFTRT.
12933 2021-06-25 Richard Biener <rguenther@suse.de>
12935 PR tree-optimization/101202
12936 * tree-vect-slp.c (vect_optimize_slp): Explicitely handle
12939 2021-06-25 Richard Biener <rguenther@suse.de>
12941 * tree-vect-slp-patterns.c (addsub_pattern::build): Copy
12942 STMT_VINFO_REDUC_DEF from the original representative.
12944 2021-06-25 Martin Sebor <msebor@redhat.com>
12946 * builtins.c (warn_string_no_nul): Replace uses of TREE_NO_WARNING,
12947 gimple_no_warning_p and gimple_set_no_warning with
12948 warning_suppressed_p, and suppress_warning.
12950 (maybe_warn_for_bound): Same.
12951 (warn_for_access): Same.
12952 (check_access): Same.
12953 (expand_builtin_strncmp): Same.
12954 (fold_builtin_varargs): Same.
12955 * calls.c (maybe_warn_nonstring_arg): Same.
12956 (maybe_warn_rdwr_sizes): Same.
12957 * cfgexpand.c (expand_call_stmt): Same.
12958 * cgraphunit.c (check_global_declaration): Same.
12959 * fold-const.c (fold_undefer_overflow_warnings): Same.
12960 (fold_truth_not_expr): Same.
12961 (fold_unary_loc): Same.
12962 (fold_checksum_tree): Same.
12963 * gimple-array-bounds.cc (array_bounds_checker::check_array_ref): Same.
12964 (array_bounds_checker::check_mem_ref): Same.
12965 (array_bounds_checker::check_addr_expr): Same.
12966 (array_bounds_checker::check_array_bounds): Same.
12967 * gimple-expr.c (copy_var_decl): Same.
12968 * gimple-fold.c (gimple_fold_builtin_strcpy): Same.
12969 (gimple_fold_builtin_strncat): Same.
12970 (gimple_fold_builtin_stxcpy_chk): Same.
12971 (gimple_fold_builtin_stpcpy): Same.
12972 (gimple_fold_builtin_sprintf): Same.
12973 (fold_stmt_1): Same.
12974 * gimple-ssa-isolate-paths.c (diag_returned_locals): Same.
12975 * gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Same.
12976 * gimple-ssa-sprintf.c (handle_printf_call): Same.
12977 * gimple-ssa-store-merging.c (imm_store_chain_info::output_merged_store): Same.
12978 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
12979 * gimple-ssa-warn-restrict.h: Adjust declarations.
12980 (maybe_diag_access_bounds): Replace uses of TREE_NO_WARNING,
12981 gimple_no_warning_p and gimple_set_no_warning with
12982 warning_suppressed_p, and suppress_warning.
12983 (check_call): Same.
12984 (check_bounds_or_overlap): Same.
12985 * gimple.c (gimple_build_call_from_tree): Same.
12986 * gimplify.c (gimplify_return_expr): Same.
12987 (gimplify_cond_expr): Same.
12988 (gimplify_modify_expr_complex_part): Same.
12989 (gimplify_modify_expr): Same.
12990 (gimple_push_cleanup): Same.
12991 (gimplify_expr): Same.
12992 * omp-expand.c (expand_omp_for_generic): Same.
12993 (expand_omp_taskloop_for_outer): Same.
12994 * omp-low.c (lower_rec_input_clauses): Same.
12995 (lower_lastprivate_clauses): Same.
12996 (lower_send_clauses): Same.
12997 (lower_omp_target): Same.
12998 * tree-cfg.c (pass_warn_function_return::execute): Same.
12999 * tree-complex.c (create_one_component_var): Same.
13000 * tree-inline.c (remap_gimple_op_r): Same.
13001 (copy_tree_body_r): Same.
13002 (declare_return_variable): Same.
13003 (expand_call_inline): Same.
13004 * tree-nested.c (lookup_field_for_decl): Same.
13005 * tree-sra.c (create_access_replacement): Same.
13006 (generate_subtree_copies): Same.
13007 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Same.
13008 * tree-ssa-forwprop.c (combine_cond_expr_cond): Same.
13009 * tree-ssa-loop-ch.c (ch_base::copy_headers): Same.
13010 * tree-ssa-loop-im.c (execute_sm): Same.
13011 * tree-ssa-phiopt.c (cond_store_replacement): Same.
13012 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
13013 (handle_builtin_strcpy): Same.
13014 (maybe_diag_stxncpy_trunc): Same.
13015 (handle_builtin_stxncpy_strncat): Same.
13016 (handle_builtin_strcat): Same.
13017 * tree-ssa-uninit.c (get_no_uninit_warning): Same.
13018 (set_no_uninit_warning): Same.
13019 (uninit_undefined_value_p): Same.
13020 (warn_uninit): Same.
13021 (maybe_warn_operand): Same.
13022 * tree-vrp.c (compare_values_warnv): Same.
13023 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr): Same.
13024 (test_for_singularity): Same.
13025 * gimple.h (warning_suppressed_p): New function.
13026 (suppress_warning): Same.
13027 (copy_no_warning): Same.
13028 (gimple_set_block): Call gimple_set_location.
13029 (gimple_set_location): Call copy_warning.
13031 2021-06-25 Martin Sebor <msebor@redhat.com>
13033 * tree.h (warning_suppressed_at, copy_warning,
13034 warning_suppressed_p, suppress_warning): New functions.
13036 2021-06-25 Martin Sebor <msebor@redhat.com>
13038 * Makefile.in (OBJS-libcommon): Add diagnostic-spec.o.
13039 * gengtype.c (open_base_files): Add diagnostic-spec.h.
13040 * diagnostic-spec.c: New file.
13041 * diagnostic-spec.h: New file.
13042 * tree.h (no_warning, all_warnings, suppress_warning_at): New
13044 * warning-control.cc: New file.
13046 2021-06-25 liuhongt <hongtao.liu@intel.com>
13049 * config/i386/i386.c (x86_order_regs_for_local_alloc):
13052 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
13054 PR tree-optimization/101189
13055 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Pass
13056 LHS range of condition to postfold routine.
13057 (fold_using_range::postfold_gcond_edges): Only process the TRUE or
13058 FALSE edge if the LHS range supports it being taken.
13059 * gimple-range-fold.h (postfold_gcond_edges): Add range parameter.
13061 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
13063 * value-relation.cc (equiv_oracle::dump): Do not dump NULL blocks.
13064 (relation_oracle::find_relation_block): Check correct bitmap.
13065 (relation_oracle::dump): Do not dump NULL blocks.
13067 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
13069 * gimple-range-cache.cc (ranger_cache::propagate_cache): Call
13070 range_on_edge instead of manually calculating.
13072 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
13074 * range-op.cc: Fix comment.
13076 2021-06-24 Uroš Bizjak <ubizjak@gmail.com>
13079 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
13080 Handle V8QI and V4HI modes.
13081 * config/i386/mmx.md (sse4_1_<any_extend:code>v4qiv4hi2):
13083 (sse4_1_<any_extend:code>v4qiv4hi2): Ditto.
13084 (mmxpackmode): New mode attribute.
13085 (vec_pack_trunc_<mmxpackmode:mode>): New expander.
13086 (mmxunpackmode): New mode attribute.
13087 (vec_unpacks_lo_<mmxunpackmode:mode>): New expander.
13088 (vec_unpacks_hi_<mmxunpackmode:mode>): Ditto.
13089 (vec_unpacku_lo_<mmxunpackmode:mode>): Ditto.
13090 (vec_unpacku_hi_<mmxunpackmode:mode>): Ditto.
13091 * config/i386/i386.md (extsuffix): Move from ...
13092 * config/i386/sse.md: ... here.
13094 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
13096 * dwarf2out.c (dwarf2out_assembly_start): Emit .file 0 marker here..
13097 (dwarf2out_finish): ...instead of here.
13099 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
13101 * configure.ac (--gdwarf-5 option): Use objdump instead of readelf.
13102 (working --gdwarf-4/--gdwarf-5 for all sources): Likewise.
13103 (--gdwarf-4 not refusing generated .debug_line): Adjust for Windows.
13104 * configure: Regenerate.
13106 2021-06-24 Richard Biener <rguenther@suse.de>
13108 * config/i386/sse.md (vec_addsubv4df3, vec_addsubv2df3,
13109 vec_addsubv8sf3, vec_addsubv4sf3): Merge into ...
13110 (vec_addsub<mode>3): ... using a new addsub_cst mode attribute.
13112 2021-06-24 Richard Biener <rguenther@suse.de>
13114 * config/i386/sse.md (avx_addsubv4df3): Rename to
13116 (avx_addsubv8sf3): Rename to vec_addsubv8sf3.
13117 (sse3_addsubv2df3): Rename to vec_addsubv2df3.
13118 (sse3_addsubv4sf3): Rename to vec_addsubv4sf3.
13119 * config/i386/i386-builtin.def: Adjust.
13120 * internal-fn.def (VEC_ADDSUB): New internal optab fn.
13121 * optabs.def (vec_addsub_optab): New optab.
13122 * tree-vect-slp-patterns.c (class addsub_pattern): New.
13123 (slp_patterns): Add addsub_pattern.
13124 * tree-vect-slp.c (vect_optimize_slp): Disable propagation
13125 across CFN_VEC_ADDSUB.
13126 * tree-vectorizer.h (vect_pattern::vect_pattern): Make
13128 * doc/md.texi (vec_addsub<mode>3): Document.
13130 2021-06-24 Jakub Jelinek <jakub@redhat.com>
13132 PR middle-end/101170
13133 * df-scan.c (df_ref_record): For paradoxical big-endian SUBREGs
13134 where regno + subreg_regno_offset wraps around use 0 as starting
13137 2021-06-24 Jakub Jelinek <jakub@redhat.com>
13139 PR middle-end/101172
13140 * stor-layout.c (finish_bitfield_representative): If nextf has
13141 error_mark_node type, set repr type to error_mark_node too.
13143 2021-06-24 Ilya Leoshkevich <iii@linux.ibm.com>
13145 * config/s390/s390.c (s390_function_profiler): Ignore labelno
13147 * config/s390/s390.h (NO_PROFILE_COUNTERS): Define.
13149 2021-06-24 Richard Biener <rguenther@suse.de>
13151 * tree-vect-slp.c (vect_optimize_slp): Do not propagate
13152 across operations that have different semantics on different
13155 2021-06-24 Jakub Jelinek <jakub@redhat.com>
13157 * tree.h (OMP_CLAUSE_MAP_IN_REDUCTION): Document meaning for OpenMP.
13158 * gimplify.c (gimplify_scan_omp_clauses): For OpenMP map clauses
13159 with OMP_CLAUSE_MAP_IN_REDUCTION flag partially defer gimplification
13160 of non-decl OMP_CLAUSE_DECL. For OMP_CLAUSE_IN_REDUCTION on
13161 OMP_TARGET user outer_ctx instead of ctx for placeholders and
13162 initializer/combiner gimplification.
13163 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_MAP_IN_REDUCTION
13164 on target constructs.
13165 (lower_rec_input_clauses): Likewise.
13166 (lower_omp_target): Likewise.
13167 * omp-expand.c (expand_omp_target): Temporarily ignore nowait clause
13168 on target if in_reduction is present.
13170 2021-06-24 Kewen Lin <linkw@linux.ibm.com>
13172 * tree-predcom.c (class pcom_worker): New class.
13173 (release_chain): Renamed to...
13174 (pcom_worker::release_chain): ...this.
13175 (release_chains): Renamed to...
13176 (pcom_worker::release_chains): ...this.
13177 (aff_combination_dr_offset): Renamed to...
13178 (pcom_worker::aff_combination_dr_offset): ...this.
13179 (determine_offset): Renamed to...
13180 (pcom_worker::determine_offset): ...this.
13181 (class comp_ptrs): New class.
13182 (split_data_refs_to_components): Renamed to...
13183 (pcom_worker::split_data_refs_to_components): ...this,
13184 and update with class comp_ptrs.
13185 (suitable_component_p): Renamed to...
13186 (pcom_worker::suitable_component_p): ...this.
13187 (filter_suitable_components): Renamed to...
13188 (pcom_worker::filter_suitable_components): ...this.
13189 (valid_initializer_p): Renamed to...
13190 (pcom_worker::valid_initializer_p): ...this.
13191 (find_looparound_phi): Renamed to...
13192 (pcom_worker::find_looparound_phi): ...this.
13193 (add_looparound_copies): Renamed to...
13194 (pcom_worker::add_looparound_copies): ...this.
13195 (determine_roots_comp): Renamed to...
13196 (pcom_worker::determine_roots_comp): ...this.
13197 (determine_roots): Renamed to...
13198 (pcom_worker::determine_roots): ...this.
13199 (single_nonlooparound_use): Renamed to...
13200 (pcom_worker::single_nonlooparound_use): ...this.
13201 (remove_stmt): Renamed to...
13202 (pcom_worker::remove_stmt): ...this.
13203 (execute_pred_commoning_chain): Renamed to...
13204 (pcom_worker::execute_pred_commoning_chain): ...this.
13205 (execute_pred_commoning): Renamed to...
13206 (pcom_worker::execute_pred_commoning): ...this.
13207 (struct epcc_data): New member worker.
13208 (execute_pred_commoning_cbck): Call execute_pred_commoning
13209 with pcom_worker pointer.
13210 (find_use_stmt): Renamed to...
13211 (pcom_worker::find_use_stmt): ...this.
13212 (find_associative_operation_root): Renamed to...
13213 (pcom_worker::find_associative_operation_root): ...this.
13214 (find_common_use_stmt): Renamed to...
13215 (pcom_worker::find_common_use_stmt): ...this.
13216 (combinable_refs_p): Renamed to...
13217 (pcom_worker::combinable_refs_p): ...this.
13218 (reassociate_to_the_same_stmt): Renamed to...
13219 (pcom_worker::reassociate_to_the_same_stmt): ...this.
13220 (stmt_combining_refs): Renamed to...
13221 (pcom_worker::stmt_combining_refs): ...this.
13222 (combine_chains): Renamed to...
13223 (pcom_worker::combine_chains): ...this.
13224 (try_combine_chains): Renamed to...
13225 (pcom_worker::try_combine_chains): ...this.
13226 (prepare_initializers_chain): Renamed to...
13227 (pcom_worker::prepare_initializers_chain): ...this.
13228 (prepare_initializers): Renamed to...
13229 (pcom_worker::prepare_initializers): ...this.
13230 (prepare_finalizers_chain): Renamed to...
13231 (pcom_worker::prepare_finalizers_chain): ...this.
13232 (prepare_finalizers): Renamed to...
13233 (pcom_worker::prepare_finalizers): ...this.
13234 (tree_predictive_commoning_loop): Renamed to...
13235 (pcom_worker::tree_predictive_commoning_loop): ...this, adjust
13236 some calls and remove some cleanup code.
13237 (tree_predictive_commoning): Adjusted to use pcom_worker instance.
13238 (static variable looparound_phis): Remove.
13239 (static variable name_expansions): Remove.
13241 2021-06-24 Richard Biener <rguenther@suse.de>
13243 * tree-vect-slp.c (slpg_vertex): New struct.
13244 (vect_slp_build_vertices): Adjust.
13245 (vect_optimize_slp): Likewise. Maintain an outgoing permute
13246 and a materialized one.
13248 2021-06-24 Richard Biener <rguenther@suse.de>
13250 PR tree-optimization/101105
13251 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
13252 Only ignore steps when they are equal or scalar order is preserved.
13254 2021-06-24 liuhongt <hongtao.liu@intel.com>
13257 * config/i386/i386-expand.c (ix86_expand_vec_interleave):
13258 Adjust comments for ix86_expand_vecop_qihi2.
13259 (ix86_expand_vecmul_qihi): Renamed to ..
13260 (ix86_expand_vecop_qihi2): Adjust function prototype to
13261 support shift operation, add static to definition.
13262 (ix86_expand_vec_shift_qihi_constant): Add static to definition.
13263 (ix86_expand_vecop_qihi): Call ix86_expand_vecop_qihi2 and
13264 ix86_expand_vec_shift_qihi_constant.
13265 * config/i386/i386-protos.h (ix86_expand_vecmul_qihi): Deleted.
13266 (ix86_expand_vec_shift_qihi_constant): Deleted.
13267 * config/i386/sse.md (VI12_256_512_AVX512VL): New mode
13269 (mulv8qi3): Call ix86_expand_vecop_qihi directly, add
13270 condition TARGET_64BIT.
13271 (mul<mode>3): Ditto.
13272 (<insn><mode>3): Ditto.
13273 (vlshr<mode>3): Extend to support avx512 vlshr.
13274 (v<insn><mode>3): New expander for
13276 (v<insn>v8qi3): Ditto.
13277 (vashrv8hi3<mask_name>): Renamed to ..
13278 (vashr<mode>3): And extend to support V16QImode for avx512.
13279 (vashrv16qi3): Deleted.
13280 (vashrv2di3<mask_name>): Extend expander to support avx512
13283 2021-06-23 Dimitar Dimitrov <dimitar@dinux.eu>
13285 * doc/lto.texi (Design Overview): Update that slim objects are
13288 2021-06-23 Aaron Sawdey <acsawdey@linux.ibm.com>
13290 * config/rs6000/rs6000-cpus.def: Take OPTION_MASK_PCREL_OPT out
13291 of OTHER_POWER10_MASKS so it will not be enabled by default.
13293 2021-06-23 Richard Biener <rguenther@suse.de>
13294 Martin Jambor <mjambor@suse.cz>
13296 * tree-inline.c (setup_one_parameter): Set TREE_READONLY of the
13297 param replacement unconditionally. Adjust comment.
13299 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
13301 * Makefile.in (OBJS): Add gimple-range-fold.o
13302 * gimple-range-fold.cc: New.
13303 * gimple-range-fold.h: New.
13304 * gimple-range-gori.cc (gimple_range_calc_op1): Move to here.
13305 (gimple_range_calc_op2): Ditto.
13306 * gimple-range-gori.h: Move prototypes to here.
13307 * gimple-range.cc: Adjust include files.
13308 (fur_source:fur_source): Relocate to gimple-range-fold.cc.
13309 (fur_source::get_operand): Ditto.
13310 (fur_source::get_phi_operand): Ditto.
13311 (fur_source::query_relation): Ditto.
13312 (fur_source::register_relation): Ditto.
13313 (class fur_edge): Ditto.
13314 (fur_edge::fur_edge): Ditto.
13315 (fur_edge::get_operand): Ditto.
13316 (fur_edge::get_phi_operand): Ditto.
13317 (fur_stmt::fur_stmt): Ditto.
13318 (fur_stmt::get_operand): Ditto.
13319 (fur_stmt::get_phi_operand): Ditto.
13320 (fur_stmt::query_relation): Ditto.
13321 (class fur_depend): Relocate to gimple-range-fold.h.
13322 (fur_depend::fur_depend): Relocate to gimple-range-fold.cc.
13323 (fur_depend::register_relation): Ditto.
13324 (fur_depend::register_relation): Ditto.
13325 (class fur_list): Ditto.
13326 (fur_list::fur_list): Ditto.
13327 (fur_list::get_operand): Ditto.
13328 (fur_list::get_phi_operand): Ditto.
13329 (fold_range): Ditto.
13330 (adjust_pointer_diff_expr): Ditto.
13331 (gimple_range_adjustment): Ditto.
13332 (gimple_range_base_of_assignment): Ditto.
13333 (gimple_range_operand1): Ditto.
13334 (gimple_range_operand2): Ditto.
13335 (gimple_range_calc_op1): Relocate to gimple-range-gori.cc.
13336 (gimple_range_calc_op2): Ditto.
13337 (fold_using_range::fold_stmt): Relocate to gimple-range-fold.cc.
13338 (fold_using_range::range_of_range_op): Ditto.
13339 (fold_using_range::range_of_address): Ditto.
13340 (fold_using_range::range_of_phi): Ditto.
13341 (fold_using_range::range_of_call): Ditto.
13342 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
13343 (fold_using_range::range_of_builtin_call): Ditto.
13344 (fold_using_range::range_of_cond_expr): Ditto.
13345 (fold_using_range::range_of_ssa_name_with_loop_info): Ditto.
13346 (fold_using_range::relation_fold_and_or): Ditto.
13347 (fold_using_range::postfold_gcond_edges): Ditto.
13348 * gimple-range.h: Add gimple-range-fold.h to include files. Change
13349 GIMPLE_RANGE_STMT_H to GIMPLE_RANGE_H.
13350 (gimple_range_handler): Relocate to gimple-range-fold.h.
13351 (gimple_range_ssa_p): Ditto.
13352 (range_compatible_p): Ditto.
13353 (class fur_source): Ditto.
13354 (class fur_stmt): Ditto.
13355 (class fold_using_range): Ditto.
13356 (gimple_range_calc_op1): Relocate to gimple-range-gori.h
13357 (gimple_range_calc_op2): Ditto.
13359 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
13361 PR tree-optimization/101148
13362 PR tree-optimization/101014
13363 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
13364 (ranger_cache::~ranger_cache): Adjust.
13365 (ranger_cache::block_range): Check if propagation disallowed.
13366 (ranger_cache::propagate_cache): Disallow propagation if new value
13367 can't be stored properly.
13368 * gimple-range-cache.h (ranger_cache::m_propfail): New member.
13370 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
13372 * gimple-range-cache.cc (class ssa_block_ranges): Adjust prototype.
13373 (sbr_vector::set_bb_range): Return true.
13374 (class sbr_sparse_bitmap): Adjust.
13375 (sbr_sparse_bitmap::set_bb_range): Return value.
13376 (block_range_cache::set_bb_range): Return value.
13377 (ranger_cache::propagate_cache): Use return value to print msg.
13378 * gimple-range-cache.h (class block_range_cache): Adjust.
13380 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
13382 * gimple-range.cc (dump_bb): Use range_on_edge from the cache.
13384 2021-06-23 Jeff Law <jeffreyalaw@gmail.com>
13386 * config/h8300/logical.md (<code><mode>3<ccnz>): Use <cczn>
13387 so this pattern can be used for test/compare removal. Pass
13388 current insn to compute_logical_op_length and output_logical_op.
13389 * config/h8300/h8300.c (compute_logical_op_cc): Remove.
13390 (h8300_and_costs): Add argument to compute_logical_op_length.
13391 (output_logical_op): Add new argument. Use it to determine if the
13392 condition codes are used and adjust the output accordingly.
13393 (compute_logical_op_length): Add new argument and update length
13394 computations when condition codes are used.
13395 * config/h8300/h8300-protos.h (compute_logical_op_length): Update
13397 (output_logical_op): Likewise.
13399 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
13402 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
13403 Handle 64bit modes for TARGET_XOP. Use indirect gen_* functions.
13404 * config/i386/mmx.md (mmx_ppermv64): New insn pattern.
13405 * config/i386/i386.md (unspec): Move UNSPEC_XOP_PERMUTE from ...
13406 * config/i386/sse.md (unspec): ... here.
13408 2021-06-23 Martin Liska <mliska@suse.cz>
13411 * optc-save-gen.awk: Put back arm_fp16_format to
13414 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
13417 * config/i386/i386.md (bsr_rex64): Add zero-flag setting RTX.
13420 (clz<mode>2): Update RTX pattern for additions.
13422 2021-06-23 Jakub Jelinek <jakub@redhat.com>
13424 PR middle-end/101167
13425 * omp-low.c (lower_omp_regimplify_p): Regimplify also PARM_DECLs
13426 and RESULT_DECLs that have DECL_HAS_VALUE_EXPR_P set.
13428 2021-06-22 Sergei Trofimovich <siarheit@google.com>
13430 * doc/rtl.texi: drop unbalanced parenthesis.
13432 2021-06-22 Richard Biener <rguenther@suse.de>
13434 PR middle-end/101156
13435 * gimplify.c (gimplify_expr): Remove premature incorrect
13438 2021-06-22 Jakub Jelinek <jakub@redhat.com>
13440 PR tree-optimization/101159
13441 * tree-vect-patterns.c (vect_recog_popcount_pattern): Fix some
13444 2021-06-22 Jakub Jelinek <jakub@redhat.com>
13446 PR middle-end/101160
13447 * function.c (assign_parms): For decl_result with TYPE_EMPTY_P type
13448 clear crtl->return_rtx instead of keeping it referencing a pseudo.
13450 2021-06-22 Jakub Jelinek <jakub@redhat.com>
13451 Andrew Pinski <apinski@marvell.com>
13453 PR tree-optimization/101162
13454 * fold-const.c (range_check_type): Handle OFFSET_TYPE like pointer
13457 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
13459 * range-op.cc (range_relational_tests): New.
13460 (range_op_tests): Call range_relational_tests.
13462 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
13464 * range-op.cc (operator_cast::lhs_op1_relation): New.
13465 (operator_identity::lhs_op1_relation): Mew.
13467 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
13469 * range-op.cc (operator_minus::op1_op2_relation_effect): New.
13471 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
13473 * range-op.cc (operator_plus::lhs_op1_relation): New.
13474 (operator_plus::lhs_op2_relation): New.
13476 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
13478 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a
13479 relation_oracle if dominators exist.
13480 (ranger_cache::~ranger_cache): Dispose of oracle.
13481 (ranger_cache::dump_bb): Dump oracle.
13482 * gimple-range.cc (fur_source::fur_source): New.
13483 (fur_source::get_operand): Use mmeber query.
13484 (fur_source::get_phi_operand): Use member_query.
13485 (fur_source::query_relation): New.
13486 (fur_source::register_dependency): Delete.
13487 (fur_source::register_relation): New.
13488 (fur_edge::fur_edge): Adjust.
13489 (fur_edge::get_phi_operand): Fix comment.
13490 (fur_edge::query): Delete.
13491 (fur_stmt::fur_stmt): Adjust.
13492 (fur_stmt::query): Delete.
13493 (fur_depend::fur_depend): Adjust.
13494 (fur_depend::register_relation): New.
13495 (fur_depend::register_relation): New.
13496 (fur_list::fur_list): Adjust.
13497 (fur_list::get_operand): Use member query.
13498 (fold_using_range::range_of_range_op): Process and query relations.
13499 (fold_using_range::range_of_address): Adjust dependency call.
13500 (fold_using_range::range_of_phi): Ditto.
13501 (gimple_ranger::gimple_ranger): New. Use ranger_ache oracle.
13502 (fold_using_range::relation_fold_and_or): New.
13503 (fold_using_range::postfold_gcond_edges): New.
13504 * gimple-range.h (class gimple_ranger): Adjust.
13505 (class fur_source): Adjust members.
13506 (class fur_stmt): Ditto.
13507 (class fold_using_range): Ditto.
13509 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
13511 * range-op.cc (range_operator::wi_fold): Apply relation effect.
13512 (range_operator::fold_range): Adjust and apply relation effect.
13513 (*::fold_range): Add relation parameters.
13514 (*::op1_range): Ditto.
13515 (*::op2_range): Ditto.
13516 (range_operator::lhs_op1_relation): New.
13517 (range_operator::lhs_op2_relation): New.
13518 (range_operator::op1_op2_relation): New.
13519 (range_operator::op1_op2_relation_effect): New.
13520 (relop_early_resolve): New.
13521 (operator_equal::op1_op2_relation): New.
13522 (operator_equal::fold_range): Call relop_early_resolve.
13523 (operator_not_equal::op1_op2_relation): New.
13524 (operator_not_equal::fold_range): Call relop_early_resolve.
13525 (operator_lt::op1_op2_relation): New.
13526 (operator_lt::fold_range): Call relop_early_resolve.
13527 (operator_le::op1_op2_relation): New.
13528 (operator_le::fold_range): Call relop_early_resolve.
13529 (operator_gt::op1_op2_relation): New.
13530 (operator_gt::fold_range): Call relop_early_resolve.
13531 (operator_ge::op1_op2_relation): New.
13532 (operator_ge::fold_range): Call relop_early_resolve.
13533 * range-op.h (class range_operator): Adjust parameters and methods.
13535 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
13537 * Makefile.in (OBJS): Add value-relation.o.
13538 * gimple-range.h: Adjust include files.
13539 * tree-data-ref.c: Adjust include file order.
13540 * value-query.cc (range_query::get_value_range): Default to no oracle.
13541 (range_query::query_relation): New.
13542 (range_query::query_relation): New.
13543 * value-query.h (class range_query): Adjust.
13544 * value-relation.cc: New.
13545 * value-relation.h: New.
13547 2021-06-22 Richard Biener <rguenther@suse.de>
13549 PR tree-optimization/101151
13550 * tree-ssa-sink.c (statement_sink_location): Expand irreducible
13553 2021-06-22 Jojo R <rjiejie@linux.alibaba.com>
13555 * config/riscv/riscv.c (thead_c906_tune_info): New.
13556 (riscv_tune_info_table): Use new tune.
13558 2021-06-22 Richard Biener <rguenther@suse.de>
13560 PR tree-optimization/101158
13561 * tree-vect-slp.c (vect_build_slp_tree_1): Move same operand
13562 checking after checking for matching operation.
13564 2021-06-22 Richard Biener <rguenther@suse.de>
13566 PR tree-optimization/101159
13567 * tree-vect-patterns.c (vect_recog_popcount_pattern): Add
13568 missing NULL vectype check.
13570 2021-06-22 Richard Biener <rguenther@suse.de>
13572 PR tree-optimization/101154
13573 * tree-vect-slp.c (vect_build_slp_tree_2): Fix out-of-bound access.
13575 2021-06-22 Jakub Jelinek <jakub@redhat.com>
13578 * config/i386/i386-protos.h (ix86_last_zero_store_uid): Declare.
13579 * config/i386/i386-expand.c (ix86_last_zero_store_uid): New variable.
13580 * config/i386/i386.c (ix86_expand_prologue): Clear it.
13581 * config/i386/i386.md (peephole2s for 1/2/4 stores of const0_rtx):
13582 Remove "" from match_operand. Emit new insns using emit_move_insn and
13583 set ix86_last_zero_store_uid to INSN_UID of the last store.
13584 Add peephole2s for 1/2/4 stores of const0_rtx following previous
13587 2021-06-22 Martin Liska <mliska@suse.cz>
13589 * auto-profile.c (AUTO_PROFILE_VERSION): Bump as string format
13592 2021-06-22 Martin Liska <mliska@suse.cz>
13594 * gcov-io.h: Remove padding entries.
13596 2021-06-22 liuhongt <hongtao.liu@intel.com>
13598 PR tree-optimization/97770
13599 * tree-vect-patterns.c (vect_recog_popcount_pattern):
13601 (vect_recog_func vect_vect_recog_func_ptrs): Add new pattern.
13603 2021-06-22 liuhongt <hongtao.liu@intel.com>
13606 * config/i386/i386-builtin.def (BDESC): Adjust builtin name.
13607 * config/i386/sse.md (<avx512>_expand<mode>_mask): Rename to ..
13608 (expand<mode>_mask): this ..
13609 (*expand<mode>_mask): New pre_reload splitter to transform
13610 v{,p}expand* to vmov* when mask is zero, all ones, or has all
13611 ones in it's lower part, otherwise still generate
13614 2021-06-22 liuhongt <hongtao.liu@intel.com>
13617 * config/i386/i386-expand.c
13618 (ix86_expand_special_args_builtin): Keep constm1_operand only
13619 if it satisfies insn's operand predicate.
13621 2021-06-21 Jason Merrill <jason@redhat.com>
13624 * df-scan.c (df_ref_record): Check that regno < endregno.
13625 * function.c (assign_parms, expand_function_end): Do nothing with a
13626 TYPE_EMPTY_P result.
13628 2021-06-21 Richard Biener <rguenther@suse.de>
13630 PR tree-optimization/101120
13631 * tree-vect-data-refs.c (bump_vector_ptr): Fold the
13633 * tree-vect-slp.c (vect_transform_slp_perm_load): Add
13634 DR chain DCE capability.
13635 * tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
13636 * tree-vect-stmts.c (vectorizable_load): Remove unused
13637 loads in the DR chain for SLP.
13639 2021-06-21 Jakub Jelinek <jakub@redhat.com>
13641 PR inline-asm/100785
13642 * gimplify.c (gimplify_asm_expr): Don't diagnose errors if
13643 output or input operands were already error_mark_node.
13644 * cfgexpand.c (expand_asm_stmt): If errors are emitted,
13645 remove all inputs, outputs and clobbers from the asm and
13646 set template to "".
13648 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
13650 * config/arm/arm_neon.h (vceq_s8): Replace builtin with __a == __b.
13651 (vceq_s16): Likewise.
13652 (vceq_s32): Likewise.
13653 (vceq_u8): Likewise.
13654 (vceq_u16): Likewise.
13655 (vceq_u32): Likewise.
13656 (vceq_p8): Likewise.
13657 (vceqq_s8): Likewise.
13658 (vceqq_s16): Likewise.
13659 (vceqq_s32): Likewise.
13660 (vceqq_u8): Likewise.
13661 (vceqq_u16): Likewise.
13662 (vceqq_u32): Likewise.
13663 (vceqq_p8): Likewise.
13664 (vceq_f32): Gate __a == __b on __FAST_MATH__.
13665 (vceqq_f32): Likewise.
13666 (vceq_f16): Likewise.
13667 (vceqq_f16): Likewise.
13669 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
13672 * config/arm/iterators.md (NEON_VACMP): Remove.
13673 * config/arm/neon.md (neon_vca<cmp_op><mode>): Use GLTE instead of GTGE
13675 (neon_vca<cmp_op><mode>_insn): Likewise.
13676 (neon_vca<cmp_op_unsp><mode>_insn_unspec): Use NEON_VAGLTE instead of
13679 2021-06-21 Richard Biener <rguenther@suse.de>
13681 PR tree-optimization/101121
13682 * tree-vect-slp.c (vect_build_slp_tree_2): To not fail fatally
13683 when we just lack a stmt with the desired op when doing permutation.
13684 (vect_build_slp_tree): When caching a failed SLP build attempt
13685 assert that at least one lane is marked as not matching.
13687 2021-06-21 liuhongt <hongtao.liu@intel.com>
13690 * config/i386/i386.md: (*anddi_1): Disparage slightly the mask
13691 register alternative.
13692 (*and<mode>_1): Ditto.
13694 (*andn<mode>_1): Ditto.
13695 (*<code><mode>_1): Ditto.
13696 (*<code>qi_1): Ditto.
13697 (*one_cmpl<mode>2_1): Ditto.
13698 (*one_cmplsi2_1_zext): Ditto.
13699 (*one_cmplqi2_1): Ditto.
13700 * config/i386/i386.c (x86_order_regs_for_local_alloc): Change
13701 the order of mask registers to be before general registers.
13703 2021-06-21 Roger Sayle <roger@nextmovesoftware.com>
13706 * config/i386/i386.md: New define_peephole2s to shrink writing
13707 1, 2 or 4 consecutive zeros to memory when optimizing for size.
13709 2021-06-18 Jeff Law <jeffreyalaw@gmail.com>
13711 * config/h8300/h8300.c (h8300_select_cc_mode): Handle SYMBOL_REF.
13712 * config/h8300/logical.md (<code><mode>3 logcial expander): Generate
13713 more efficient code when the source can be trivially simplified.
13715 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
13717 * gimple-range-cache.cc (ranger_cache::range_of_def): Calculate
13718 a range if global is not available.
13719 (ranger_cache::entry_range): Fallback to range_of_def.
13720 * gimple-range-cache.h (range_of_def): Adjust prototype.
13722 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
13724 PR tree-optimization/101014
13725 * gimple-range-cache.cc (ranger_cache::ranger_cache): Remove poor
13727 (ranger_cache::~ranger_cache): Ditto.
13728 (ranger_cache::enable_new_values): Delete.
13729 (ranger_cache::push_poor_value): Delete.
13730 (ranger_cache::range_of_def): Remove poor value processing.
13731 (ranger_cache::entry_range): Ditto.
13732 (ranger_cache::fill_block_cache): Ditto.
13733 * gimple-range-cache.h (class ranger_cache): Remove poor value members.
13734 * gimple-range.cc (gimple_ranger::range_of_expr): Remove call.
13735 * gimple-range.h (class gimple_ranger): Adjust.
13737 2021-06-18 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
13740 * common/config/arm/arm-common.c (arm_canon_arch_option_1): New function
13741 derived from arm_canon_arch.
13742 (arm_canon_arch_option): Call it.
13743 (arm_canon_arch_multilib_option): New function.
13744 * config/arm/arm-cpus.in (IGNORE_FOR_MULTILIB): New fgroup.
13745 * config/arm/arm.h (arm_canon_arch_multilib_option): New prototype.
13746 (CANON_ARCH_MULTILIB_SPEC_FUNCTION): New macro.
13747 (MULTILIB_ARCH_CANONICAL_SPECS): New macro.
13748 (DRIVER_SELF_SPECS): Add MULTILIB_ARCH_CANONICAL_SPECS.
13749 * config/arm/arm.opt (mlibarch): New option.
13750 * config/arm/t-rmprofile (MULTILIB_MATCHES): For armv8*-m, replace use
13751 of march on RHS with mlibarch.
13753 2021-06-18 Marcel Vollweiler <marcel@codesourcery.com>
13755 * config.in: Regenerate.
13756 * config/gcn/gcn.c (print_operand_address): Fix for global_load assembler
13758 * configure: Regenerate.
13759 * configure.ac: Fix for global_load assembler functions.
13761 2021-06-18 Richard Biener <rguenther@suse.de>
13763 PR tree-optimization/101112
13764 * tree-vect-slp.c (vect_slp_linearize_chain): Fix condition
13765 to lookup a pattern stmt def.
13767 2021-06-18 Jakub Jelinek <jakub@redhat.com>
13769 PR middle-end/101062
13770 * stor-layout.c (finish_bitfield_layout): Don't add bitfield
13771 representatives in QUAL_UNION_TYPE.
13773 2021-06-18 Andrew Pinski <apinski@marvell.com>
13775 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
13776 Add counting of how many times it is done.
13777 (factor_out_conditional_conversion): Likewise.
13778 (match_simplify_replacement): Likewise.
13779 (value_replacement): Likewise.
13780 (spaceship_replacement): Likewise.
13781 (cond_store_replacement): Likewise.
13782 (cond_if_else_store_replacement_1): Likewise.
13783 (hoist_adjacent_loads): Likewise.
13785 2021-06-18 Andrew Pinski <apinski@marvell.com>
13787 * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
13788 types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and CNONJ_EXPR.
13789 (verify_gimple_assign_binary): Reject point and offset types on
13790 MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
13791 FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
13792 FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.
13794 2021-06-18 Michael Meissner <meissner@linux.ibm.com>
13796 * config/rs6000/rs6000.c (rs6000_emit_minmax): Add support for ISA
13797 3.1 IEEE 128-bit floating point xsmaxcqp/xsmincqp instructions.
13798 * config/rs6000/rs6000.md (s<minmax><mode>3, IEEE128 iterator):
13801 2021-06-17 Aaron Sawdey <acsawdey@linux.ibm.com>
13803 * config/rs6000/genfusion.pl (gen_logical_addsubf): Add
13804 earlyclobber to alts 0/1.
13805 (gen_addadd): Add earlyclobber to alts 0/1.
13806 * config/rs6000/fusion.md: Regenerate file.
13808 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13810 * cfgloopanal.c (get_loop_hot_path): Make path an auto_vec.
13812 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
13814 * gimple-range-cache.cc: Comment cleanups.
13815 * gimple-range-gori.cc: Comment cleanups.
13816 * gimple-range.cc: Comment/spacing cleanups
13817 * value-range.h: Comment cleanups.
13819 2021-06-17 H.J. Lu <hjl.tools@gmail.com>
13822 * calls.c (expand_call): Replace PUSH_ARGS with
13823 targetm.calls.push_argument (0).
13824 (emit_library_call_value_1): Likewise.
13825 * defaults.h (PUSH_ARGS): Removed.
13826 (PUSH_ARGS_REVERSED): Replace PUSH_ARGS with
13827 targetm.calls.push_argument (0).
13828 * expr.c (block_move_libcall_safe_for_call_parm): Likewise.
13829 (emit_push_insn): Pass the number bytes to push to
13830 targetm.calls.push_argument and pass 0 if ARGS_ADDR is 0.
13831 * hooks.c (hook_bool_uint_true): New.
13832 * hooks.h (hook_bool_uint_true): Likewise.
13833 * rtlanal.c (nonzero_bits1): Replace PUSH_ARGS with
13834 targetm.calls.push_argument (0).
13835 * target.def (push_argument): Add a targetm.calls hook.
13836 * targhooks.c (default_push_argument): New.
13837 * targhooks.h (default_push_argument): Likewise.
13838 * config/bpf/bpf.h (PUSH_ARGS): Removed.
13839 * config/cr16/cr16.c (TARGET_PUSH_ARGUMENT): New.
13840 * config/cr16/cr16.h (PUSH_ARGS): Removed.
13841 * config/i386/i386.c (ix86_push_argument): New.
13842 (TARGET_PUSH_ARGUMENT): Likewise.
13843 * config/i386/i386.h (PUSH_ARGS): Removed.
13844 * config/m32c/m32c.c (TARGET_PUSH_ARGUMENT): New.
13845 * config/m32c/m32c.h (PUSH_ARGS): Removed.
13846 * config/nios2/nios2.h (PUSH_ARGS): Likewise.
13847 * config/pru/pru.h (PUSH_ARGS): Likewise.
13848 * doc/tm.texi.in: Remove PUSH_ARGS documentation. Add
13849 TARGET_PUSH_ARGUMENT hook.
13850 * doc/tm.texi: Regenerated.
13852 2021-06-17 Uroš Bizjak <ubizjak@gmail.com>
13855 * config/i386/i386-expand.c (expand_vector_set_var):
13856 Handle V2FS mode remapping. Pass TARGET_MMX_WITH_SSE to
13857 ix86_expand_vector_init_duplicate.
13858 (ix86_expand_vector_init_duplicate): Emit insv_1 for
13859 QImode for !TARGET_PARTIAL_REG_STALL.
13860 * config/i386/predicates.md (vec_setm_mmx_operand): New predicate.
13861 * config/i386/mmx.md (vec_setv2sf): Use vec_setm_mmx_operand
13862 as operand 2 predicate. Call ix86_expand_vector_set_var
13863 for non-constant index operand.
13864 (vec_setv2si): Ditto.
13865 (vec_setv4hi): Ditto.
13866 (vec_setv8qi): ditto.
13868 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
13870 PR tree-optimization/100790
13871 * gimple-range.cc (range_of_builtin_call): Cleanup clz and ctz
13874 2021-06-17 Martin Liska <mliska@suse.cz>
13876 * doc/invoke.texi: Use consistently -O1 instead of -O.
13878 2021-06-17 Martin Liska <mliska@suse.cz>
13880 * gcov-io.h: Update documentation entry about string format.
13882 2021-06-17 Marius Hillenbrand <mhillen@linux.ibm.com>
13885 * config/s390/vecintrin.h (vec_doublee): Fix to use
13886 __builtin_s390_vflls.
13887 (vec_floate): Fix to use __builtin_s390_vflrd.
13889 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13891 * dominance.c (get_dominated_to_depth): Return auto_vec<basic_block>.
13892 * dominance.h (get_dominated_to_depth): Likewise.
13893 (get_all_dominated_blocks): Likewise.
13894 * cfgcleanup.c (delete_unreachable_blocks): Adjust.
13895 * gcse.c (hoist_code): Likewise.
13896 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
13897 * tree-parloops.c (oacc_entry_exit_ok): Likewise.
13898 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Likewise.
13899 * tree-ssa-phiprop.c (pass_phiprop::execute): Likewise.
13901 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13903 * dominance.c (get_dominated_by_region): Return auto_vec<basic_block>.
13904 * dominance.h (get_dominated_by_region): Likewise.
13905 * tree-cfg.c (gimple_duplicate_sese_region): Adjust.
13906 (gimple_duplicate_sese_tail): Likewise.
13907 (move_sese_region_to_fn): Likewise.
13909 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13911 * dominance.c (get_dominated_by): Return auto_vec<basic_block>.
13912 * dominance.h (get_dominated_by): Likewise.
13913 * auto-profile.c (afdo_find_equiv_class): Adjust.
13914 * cfgloopmanip.c (duplicate_loop_to_header_edge): Likewise.
13915 * loop-unroll.c (unroll_loop_runtime_iterations): Likewise.
13916 * tree-cfg.c (test_linear_chain): Likewise.
13917 (test_diamond): Likewise.
13919 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13921 * cfgloop.h (get_loop_hot_path): Return auto_vec<basic_block>.
13922 * cfgloopanal.c (get_loop_hot_path): Likewise.
13923 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
13925 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13927 * cgraph.c (cgraph_node::collect_callers): Return
13928 auto_vec<cgraph_edge *>.
13929 * cgraph.h (cgraph_node::collect_callers): Likewise.
13930 * ipa-cp.c (create_specialized_node): Adjust.
13931 (decide_about_value): Likewise.
13932 (decide_whether_version_node): Likewise.
13933 * ipa-sra.c (process_isra_node_results): Likewise.
13935 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13937 * vec.h (vl_ptr>::using_auto_storage): Handle null m_vec.
13938 (auto_vec<T, 0>::auto_vec): Define move constructor, and delete copy
13940 (auto_vec<T, 0>::operator=): Define move assignment and delete copy
13943 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
13945 * gimple-range.cc (debug_seed_ranger): New.
13946 (dump_ranger): New.
13947 (debug_ranger): New.
13949 2021-06-17 Richard Biener <rguenther@suse.de>
13951 PR tree-optimization/54400
13952 * tree-vectorizer.h (enum slp_instance_kind): Add
13953 slp_inst_kind_bb_reduc.
13954 (reduction_fn_for_scalar_code): Declare.
13955 * tree-vect-data-refs.c (vect_slp_analyze_instance_dependence):
13956 Check SLP_INSTANCE_KIND instead of looking at the
13958 (vect_slp_analyze_instance_alignment): Likewise.
13959 * tree-vect-loop.c (reduction_fn_for_scalar_code): Export.
13960 * tree-vect-slp.c (vect_slp_linearize_chain): Split out
13961 chain linearization from vect_build_slp_tree_2 and generalize
13962 for the use of BB reduction vectorization.
13963 (vect_build_slp_tree_2): Adjust accordingly.
13964 (vect_optimize_slp): Elide permutes at the root of BB reduction
13966 (vectorizable_bb_reduc_epilogue): New function.
13967 (vect_slp_prune_covered_roots): Likewise.
13968 (vect_slp_analyze_operations): Use them.
13969 (vect_slp_check_for_constructors): Recognize associatable
13970 chains for BB reduction vectorization.
13971 (vectorize_slp_instance_root_stmt): Generate code for the
13972 BB reduction epilogue.
13974 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
13976 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Check with
13978 (gori_compute::may_recompute_p): New.
13979 (gori_compute::outgoing_edge_range_p): Perform recomputations.
13980 * gimple-range-gori.h (class gori_compute): Add prototype.
13982 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
13984 * gimple-range-cache.cc (ranger_cache::range_on_edge): Always return
13985 true when a range can be calculated.
13986 * gimple-range.cc (gimple_ranger::dump_bb): Check has_edge_range_p.
13988 2021-06-16 Martin Sebor <msebor@redhat.com>
13990 * doc/invoke.texi (-Wmismatched-dealloc, -Wmismatched-new-delete):
13991 Correct documented defaults.
13993 2021-06-16 Andrew MacLeod <amacleod@redhat.com>
13995 * gimple-range-cache.cc (ranger_cache::ranger_cache): Initialize
13996 m_new_value_p directly.
13998 2021-06-16 Uroš Bizjak <ubizjak@gmail.com>
14001 * config/i386/i386-expand.c (expand_vec_perm_2perm_pblendv):
14002 Handle 64bit modes for TARGET_SSE4_1.
14003 (expand_vec_perm_pshufb2): Handle 64bit modes for TARGET_SSSE3.
14004 (expand_vec_perm_even_odd_pack): Handle V4HI mode.
14005 (expand_vec_perm_even_odd_1) <case E_V4HImode>: Expand via
14006 expand_vec_perm_pshufb2 for TARGET_SSSE3 and via
14007 expand_vec_perm_even_odd_pack for TARGET_SSE4_1.
14008 * config/i386/mmx.md (mmx_packusdw): New insn pattern.
14010 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
14012 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn<mode>):
14013 Change to an expander that emits the correct instruction
14014 depending on endianness.
14015 (aarch64_<sur><addsub>hn<mode>_insn_le): Define.
14016 (aarch64_<sur><addsub>hn<mode>_insn_be): Define.
14018 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
14020 * config/aarch64/aarch64-simd-builtins.def: Split generator
14021 for aarch64_<su>qmovn builtins into scalar and vector
14023 * config/aarch64/aarch64-simd.md (aarch64_<su>qmovn<mode>_insn_le):
14025 (aarch64_<su>qmovn<mode>_insn_be): Define.
14026 (aarch64_<su>qmovn<mode>): Split into scalar and vector
14027 variants. Change vector variant to an expander that emits the
14028 correct instruction depending on endianness.
14030 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
14032 * config/aarch64/aarch64-simd-builtins.def: Split generator
14033 for aarch64_sqmovun builtins into scalar and vector variants.
14034 * config/aarch64/aarch64-simd.md (aarch64_sqmovun<mode>):
14035 Split into scalar and vector variants. Change vector variant
14036 to an expander that emits the correct instruction depending
14038 (aarch64_sqmovun<mode>_insn_le): Define.
14039 (aarch64_sqmovun<mode>_insn_be): Define.
14041 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
14043 * config/aarch64/aarch64-simd.md (aarch64_xtn<mode>_insn_le):
14044 Define - modeling zero-high-half semantics.
14045 (aarch64_xtn<mode>): Change to an expander that emits the
14046 appropriate instruction depending on endianness.
14047 (aarch64_xtn<mode>_insn_be): Define - modeling zero-high-half
14049 (aarch64_xtn2<mode>_le): Rename to...
14050 (aarch64_xtn2<mode>_insn_le): This.
14051 (aarch64_xtn2<mode>_be): Rename to...
14052 (aarch64_xtn2<mode>_insn_be): This.
14053 (vec_pack_trunc_<mode>): Emit truncation instruction instead
14055 * config/aarch64/iterators.md (Vnarrowd): Add Vnarrowd mode
14056 attribute iterator.
14058 2021-06-16 Martin Jambor <mjambor@suse.cz>
14060 PR tree-optimization/100453
14061 * tree-sra.c (create_access): Disqualify any const candidates
14062 which are written to.
14063 (sra_modify_expr): Do not store sub-replacements back to a const base.
14064 (handle_unscalarized_data_in_subtree): Likewise.
14065 (sra_modify_assign): Likewise. Earlier, use TREE_READONLy test
14066 instead of constant_decl_p.
14068 2021-06-16 Jakub Jelinek <jakub@redhat.com>
14070 PR middle-end/101062
14071 * stor-layout.c (finish_bitfield_representative): For fields in unions
14072 assume nextf is always NULL.
14073 (finish_bitfield_layout): Compute bit field representatives also in
14074 unions, but handle it as if each bitfield was the only field in the
14077 2021-06-16 Richard Biener <rguenther@suse.de>
14079 PR tree-optimization/101088
14080 * tree-ssa-loop-im.c (sm_seq_valid_bb): Only look for
14081 supported refs on edges. Do not assert same ref but
14082 different kind stores are unsuported but mark them so.
14083 (hoist_memory_references): Only look for supported refs
14086 2021-06-16 Roger Sayle <roger@nextmovesoftware.com>
14088 PR rtl-optimization/46235
14089 * config/i386/i386.md: New define_split for bt followed by cmov.
14090 (*bt<mode>_setcqi): New define_insn_and_split for bt followed by setc.
14091 (*bt<mode>_setncqi): New define_insn_and_split for bt then setnc.
14092 (*bt<mode>_setnc<mode>): New define_insn_and_split for bt followed
14093 by setnc with zero extension.
14095 2021-06-16 Richard Biener <rguenther@suse.de>
14097 PR tree-optimization/101083
14098 * tree-vect-slp.c (vect_slp_build_two_operator_nodes): Get
14099 vectype as argument.
14100 (vect_build_slp_tree_2): Adjust.
14102 2021-06-15 Martin Sebor <msebor@redhat.com>
14104 PR middle-end/100876
14105 * builtins.c: (gimple_call_return_array): Account for size_t
14106 mangling as either unsigned int or unsigned long
14108 2021-06-15 Jeff Law <jeffreyalaw@gmail.com>
14110 * compare-elim.c (try_eliminate_compare): Run DCE to clean things
14111 up before eliminating comparisons.
14113 2021-06-15 Aldy Hernandez <aldyh@redhat.com>
14115 * range-op.cc (operator_bitwise_or::wi_fold): Make sure
14116 nonzero|X is nonzero.
14117 (range_op_bitwise_and_tests): Add tests for above.
14119 2021-06-15 Carl Love <cel@us.ibm.com>
14122 * config/rs6000/rs6000-builtin.def (VCMPEQUT): Fix the ICODE for the
14124 (VRLQ, VSLQ, VSRQ, VSRAQ): Remove unused BU_P10_OVERLOAD_2
14127 2021-06-15 Tobias Burnus <tobias@codesourcery.com>
14130 * gimplify.c (enum gimplify_defaultmap_kind): Add GDMK_SCALAR_TARGET.
14131 (struct gimplify_omp_ctx): Extend defaultmap array by one.
14132 (new_omp_context): Init defaultmap[GDMK_SCALAR_TARGET].
14133 (omp_notice_variable): Update type classification for Fortran.
14134 (gimplify_scan_omp_clauses): Update calls for new argument; handle
14135 GDMK_SCALAR_TARGET; for Fortran, GDMK_POINTER avoid GOVD_MAP_0LEN_ARRAY.
14136 * langhooks-def.h (lhd_omp_scalar_p): Add 'ptr_ok' argument.
14137 * langhooks.c (lhd_omp_scalar_p): Likewise.
14138 (LANG_HOOKS_OMP_ALLOCATABLE_P, LANG_HOOKS_OMP_SCALAR_TARGET_P): New.
14139 (LANG_HOOKS_DECLS): Add them.
14140 * langhooks.h (struct lang_hooks_for_decls): Add new hooks, update
14141 omp_scalar_p pointer type to include the new bool argument.
14143 2021-06-15 David Malcolm <dmalcolm@redhat.com>
14145 * doc/analyzer.texi
14146 (Special Functions for Debugging the Analyzer): Add
14147 __analyzer_dump_capacity.
14149 2021-06-15 Jakub Jelinek <jakub@redhat.com>
14152 * expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
14153 case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.
14155 2021-06-15 Richard Biener <rguenther@suse.de>
14157 * cfgloopanal.c (mark_irreducible_loops): Use a dominance
14158 check to identify loop latches.
14159 * cfgloop.c (verify_loop_structure): Likewise.
14160 * loop-init.c (apply_loop_flags): Allow marked irreducible
14161 regions even with multiple latches.
14162 * predict.c (rebuild_frequencies): Simplify.
14164 2021-06-15 Richard Biener <rguenther@suse.de>
14166 * tree-ssa-threadupdate.c
14167 (jump_thread_path_registry::mark_threaded_blocks): Assert we
14168 have marked irreducible regions.
14170 2021-06-14 Martin Sebor <msebor@redhat.com>
14173 * builtins.c (gimple_call_return_array): Check for attribute fn spec.
14174 Handle calls to placement new.
14175 (ndecl_dealloc_argno): Avoid placement delete.
14177 2021-06-14 Peter Bergner <bergner@linux.ibm.com>
14180 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
14181 create_tmp_reg_or_ssa_name().
14183 2021-06-14 Andrew MacLeod <amacleod@redhat.com>
14185 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
14186 (ranger_cache::enable_new_values): Set to specified value and
14187 return the old value.
14188 (ranger_cache::disable_new_values): Delete.
14189 (ranger_cache::fill_block_cache): Disable non 1st order derived
14191 * gimple-range-cache.h (ranger_cache): Adjust prototypes.
14192 * gimple-range.cc (gimple_ranger::range_of_expr): Adjust.
14194 2021-06-14 Uroš Bizjak <ubizjak@gmail.com>
14197 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
14198 Return true early when testing with V2HImode.
14199 * config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.
14201 2021-06-14 Christophe Lyon <christophe.lyon@linaro.org>
14203 * config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
14204 (mve_vec_unpack<US>_hi_<mode>): New pattern.
14205 (@mve_vec_pack_trunc_lo_<mode>): New pattern.
14206 (mve_vmovntq_<supf><mode>): Prefix with '@'.
14207 * config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
14209 (vec_unpack<US>_lo_<mode>): Likewise.
14210 (vec_pack_trunc_<mode>): Rename to
14211 neon_quad_vec_pack_trunc_<mode>.
14212 * config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
14214 (vec_unpack<US>_lo_<mode>): New.
14215 (vec_pack_trunc_<mode>): New.
14217 2021-06-14 Richard Biener <rguenther@suse.de>
14219 PR tree-optimization/100934
14220 * tree-ssa-dom.c (pass_dominator::execute): Properly
14221 mark irreducible regions.
14223 2021-06-14 Martin Liska <mliska@suse.cz>
14225 * doc/invoke.texi: Put r{...} on the same line as @item.
14227 2021-06-14 Martin Liska <mliska@suse.cz>
14229 * doc/invoke.texi: Add missing newline.
14231 2021-06-14 Martin Liska <mliska@suse.cz>
14233 * doc/invoke.texi: Remove '+' charasters.
14235 2021-06-14 Claudiu Zissulescu <claziss@synopsys.com>
14237 * config.gcc (arc): Add support for with_cpu option.
14238 * config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.
14240 2021-06-14 Richard Biener <rguenther@suse.de>
14242 PR tree-optimization/101031
14243 * tree-ssa-strlen.c (maybe_invalidate): Increment max_size
14244 instead of size when accounting for a possibly string
14247 2021-06-14 Martin Liska <mliska@suse.cz>
14249 * gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].
14251 2021-06-14 Aldy Hernandez <aldyh@redhat.com>
14253 * value-query.cc (gimple_range_global): Call get_range_global
14254 if called after inlining.
14256 2021-06-13 Uroš Bizjak <ubizjak@gmail.com>
14259 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
14260 Emit constant permutation insn directly from here.
14262 2021-06-13 Trevor Saunders <tbsaunde@tbsaunde.org>
14264 * attribs.c (find_attribute_namespace): Iterate over vec<> with
14266 * auto-profile.c (afdo_find_equiv_class): Likewise.
14267 * gcc.c (do_specs_vec): Likewise.
14268 (do_spec_1): Likewise.
14269 (driver::set_up_specs): Likewise.
14270 * gimple-loop-jam.c (any_access_function_variant_p): Likewise.
14271 * gimple-ssa-store-merging.c (compatible_load_p): Likewise.
14272 (imm_store_chain_info::try_coalesce_bswap): Likewise.
14273 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
14274 (get_location_for_stmts): Likewise.
14275 * graphite-poly.c (print_iteration_domains): Likewise.
14276 (free_poly_bb): Likewise.
14277 (remove_gbbs_in_scop): Likewise.
14278 (free_scop): Likewise.
14279 (dump_gbb_cases): Likewise.
14280 (dump_gbb_conditions): Likewise.
14281 (print_pdrs): Likewise.
14282 (print_scop): Likewise.
14283 * ifcvt.c (cond_move_process_if_block): Likewise.
14284 * lower-subreg.c (decompose_multiword_subregs): Likewise.
14285 * regcprop.c (pass_cprop_hardreg::execute): Likewise.
14286 * sanopt.c (sanitize_rewrite_addressable_params): Likewise.
14287 * sel-sched-dump.c (dump_insn_vector): Likewise.
14288 * store-motion.c (store_ops_ok): Likewise.
14289 (store_killed_in_insn): Likewise.
14290 * timevar.c (timer::named_items::print): Likewise.
14291 * tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
14292 (cleanup_tree_cfg_noloop): Likewise.
14293 * tree-data-ref.c (dump_data_references): Likewise.
14294 (print_dir_vectors): Likewise.
14295 (print_dist_vectors): Likewise.
14296 (dump_data_dependence_relations): Likewise.
14297 (dump_dist_dir_vectors): Likewise.
14298 (dump_ddrs): Likewise.
14299 (create_runtime_alias_checks): Likewise.
14300 (free_subscripts): Likewise.
14301 (save_dist_v): Likewise.
14302 (save_dir_v): Likewise.
14303 (invariant_access_functions): Likewise.
14304 (same_access_functions): Likewise.
14305 (access_functions_are_affine_or_constant_p): Likewise.
14306 (find_data_references_in_stmt): Likewise.
14307 (graphite_find_data_references_in_stmt): Likewise.
14308 (free_dependence_relations): Likewise.
14309 (free_data_refs): Likewise.
14310 * tree-inline.c (copy_debug_stmts): Likewise.
14311 * tree-into-ssa.c (dump_currdefs): Likewise.
14312 (rewrite_update_phi_arguments): Likewise.
14313 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
14314 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
14316 (vect_slp_analyze_node_dependences): Likewise.
14317 (vect_slp_analyze_instance_dependence): Likewise.
14318 (vect_record_base_alignments): Likewise.
14319 (vect_get_peeling_costs_all_drs): Likewise.
14320 (vect_peeling_supportable): Likewise.
14321 * tree-vectorizer.c (vec_info::~vec_info): Likewise.
14322 (vec_info::free_stmt_vec_infos): Likewise.
14324 2021-06-13 Jeff Law <jeffreyalaw@gmail.com>
14326 * config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
14327 (andqi3_1<cczn>): Removed.
14328 (<ors>qi3_1): Do not split for IOR/XOR a single bit.
14329 (H8/SX bit logicals): Split out from other patterns.
14330 * config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
14331 mulqihi3_const_clobber_flags.
14332 (mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly
14334 2021-06-13 H.J. Lu <hjl.tools@gmail.com>
14337 * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
14338 to true if red zone is used.
14339 (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
14340 ix86_red_zone_used.
14341 * config/i386/i386.h (machine_function): Add red_zone_used.
14342 (ix86_red_zone_size): Removed.
14343 (ix86_red_zone_used): New.
14344 * config/i386/i386.md (peephole2 patterns): Replace
14345 ix86_red_zone_size with ix86_red_zone_used.
14347 2021-06-12 Jason Merrill <jason@redhat.com>
14349 * doc/extend.texi (unused variable attribute): Applies to
14350 structure fields as well.
14352 2021-06-12 Eugene Rozenfeld <erozen@microsoft.com>
14354 * auto-profile.c (read_profile): fix a typo in an error string
14356 2021-06-11 Thomas Schwinge <thomas@codesourcery.com>
14358 * tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
14360 * tree-pretty-print.c (dump_omp_clauses): Update.
14361 (dump_generic_node) <OMP_CLAUSE>: Use it.
14363 2021-06-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
14366 * config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
14367 int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
14368 the polymorphic variants matching code.
14369 (__arm_vld1q_z): Likewise.
14370 (__arm_vld2q): Likewise.
14371 (__arm_vld4q): Likewise.
14372 (__arm_vldrbq_gather_offset): Likewise.
14373 (__arm_vldrbq_gather_offset_z): Likewise.
14375 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
14377 PR tree-optimization/96392
14378 * fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.
14380 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
14382 PR tree-optimization/96392
14383 * fold-const.c (fold_real_zero_addition_p): Take both arguments
14384 of the addition or subtraction, not just the zero. Use this
14385 other argument in tests for signaling NaNs and signed zeros.
14386 (tree_expr_maybe_real_minus_zero_p): New predicate.
14387 * fold-const.h (fold_real_zero_addition_p): Update prototype.
14388 (tree_expr_maybe_real_minus_zero_p): New function prototype.
14389 * match.pd: Update calls to fold_real_zero_addition_p.
14390 Replace HONOR_NANS with tree_expr_maybe_nan_p.
14391 Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
14392 Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
14393 * tree-ssa-reassoc.c (eliminate_using_constants): Update
14394 call to fold_real_zero_addition_p.
14396 2021-06-11 Richard Biener <rguenther@suse.de>
14398 PR tree-optimization/101025
14399 * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
14400 all refs that require dependence checking.
14402 2021-06-11 Richard Biener <rguenther@suse.de>
14404 PR tree-optimization/101028
14405 * tree-vect-slp.c (vect_build_slp_tree_2): When SLP
14406 reassoc discovery fails fatally, mark appropriate lanes
14409 2021-06-11 Richard Biener <rguenther@suse.de>
14411 PR tree-optimization/101026
14412 * tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
14413 have a representative for the associated chain nodes.
14415 2021-06-11 Jakub Jelinek <jakub@redhat.com>
14417 PR rtl-optimization/101008
14418 * simplify-rtx.c (relational_result): New function.
14419 (simplify_logical_relational_operation,
14420 simplify_relational_operation): Use it.
14422 2021-06-11 Jakub Jelinek <jakub@redhat.com>
14425 * config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.
14427 2021-06-11 Uroš Bizjak <ubizjak@gmail.com>
14430 * config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
14431 false if the permutation can be implemented with constant
14432 permutation instruction in wider mode.
14433 (canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
14434 Handle V8QImode and V4HImode.
14436 2021-06-11 Martin Liska <mliska@suse.cz>
14438 PR gcov-profile/100788
14439 * common.opt: Add new option.
14440 * coverage.c (coverage_begin_function): Emit warning instead on
14441 the internal compiler error.
14442 * doc/invoke.texi: Document the option.
14443 * toplev.c (process_options): Enable it by default.
14445 2021-06-11 Richard Biener <rguenther@suse.de>
14447 PR middle-end/101009
14448 * tree-data-ref.c (build_classic_dist_vector_1): Make sure
14449 to set *init_b to true when we encounter a constant equal
14451 (compute_affine_dependence): Also dump the actual DR_REF.
14453 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
14455 PR tree-optimization/100984
14456 * gimple-ssa-evrp.c (ssa_equiv_stack): Use auto_vec for
14457 replacements table.
14458 (ssa_equiv_stack::~ssa_equiv_stack): Remove.
14460 2021-06-11 Kewen Lin <linkw@linux.ibm.com>
14462 * config/rs6000/rs6000.md
14463 (floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
14464 define_insn_and_split.
14466 2021-06-11 Richard Biener <rguenther@suse.de>
14468 * tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
14469 to sort operands of the associative chain.
14471 2021-06-11 Richard Biener <rguenther@suse.de>
14473 * system.h (gcc_stablesort_r): Declare.
14474 * sort.cc (gcc_sort_r): Support stable sort.
14475 (gcc_stablesort_r): Define.
14476 * vec.h (vec<>::stablesort): Add.
14478 2021-06-10 Uroš Bizjak <ubizjak@gmail.com>
14481 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
14482 Handle V2SF mode. Emit SHUFPS to fixup unpack-high for V2SF mode.
14483 (expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
14484 (expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
14485 (expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
14486 (expand_vec_perm_interleave2): Handle 64bit modes.
14487 (expand_vec_perm_even_odd_pack): Handle V8QI mode.
14488 (expand_vec_perm_even_odd_1): Ditto.
14489 (ix86_vectorize_vec_perm_const): Ditto.
14490 * config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
14491 * config/i386/sse.md: ... here.
14492 * config/i386/mmx.md (*vec_interleave_lowv2sf):
14493 New insn_and_split pattern.
14494 (*vec_interleave_highv2sf): Ditto.
14495 (mmx_pshufbv8qi3): New insn pattern.
14496 (*mmx_pblendw): Ditto.
14498 2021-06-10 Peter Bergner <bergner@linux.ibm.com>
14500 * config/rs6000/rs6000-builtin.def (build_pair): New built-in.
14501 (build_acc): Likewise.
14502 * config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
14503 source operands in little-endian mode.
14504 (rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
14505 (mma_init_builtins): Likewise.
14506 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
14507 ordering for the MMA assemble and build source operands.
14508 * doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
14510 (__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
14513 2021-06-10 Jeff Law <jeffreyalaw@gmail.com>
14515 * config/h8300/h8300.c (select_cc_mode): Handle MEM. Use
14517 * config/h8300/extensions.md: Replace _clobber_flags patterns
14520 2021-06-10 Robin Dapp <rdapp@linux.ibm.com>
14522 * config/s390/vector.md (vcond_mask_<mode><mode>): Change to
14523 (vcond_mask_<mode><tointvec>): this.
14525 2021-06-10 Andrew Stubbs <ams@codesourcery.com>
14526 Thomas Schwinge <thomas@codesourcery.com>
14528 * omp-builtins.def (BUILT_IN_GOACC_ENTER_EXIT_DATA): Split into...
14529 (BUILT_IN_GOACC_ENTER_DATA, BUILT_IN_GOACC_EXIT_DATA): ... these.
14530 * gimple.h (enum gf_mask): Split
14531 'GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA' into
14532 'GF_OMP_TARGET_KIND_OACC_ENTER_DATA' and
14533 'GF_OMP_TARGET_KIND_OACC_EXIT_DATA'.
14534 (is_gimple_omp_oacc): Update.
14535 * gimple-pretty-print.c (dump_gimple_omp_target): Likewise.
14536 * gimplify.c (gimplify_omp_target_update): Likewise.
14537 * omp-expand.c (expand_omp_target, build_omp_regions_1)
14538 (omp_make_gimple_edges): Likewise.
14539 * omp-low.c (check_omp_nesting_restrictions, lower_omp_target):
14542 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
14544 * value-query.cc (value_query::value_on_edge): Rename name to
14546 (range_query::range_on_edge): Same.
14547 (range_query::value_of_expr): Same.
14548 (range_query::value_on_edge): Same.
14549 * value-query.h (class value_query): Same.
14550 (class range_query): Same.
14552 2021-06-10 Richard Biener <rguenther@suse.de>
14554 PR tree-optimization/101003
14555 * tree-vect-slp.c (vect_build_slp_tree_2): Appropriately
14556 use the pattern stmt defs when linearizing a chain.
14558 2021-06-10 Jakub Jelinek <jakub@redhat.com>
14561 * ifcvt.c (noce_get_alt_condition, noce_try_abs): Use
14562 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
14564 2021-06-10 Clement Chigot <clement.chigot@atos.net>
14566 * config/rs6000/aix71.h (ASM_CPU_SPEC): Add Power10 directive.
14567 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
14569 2021-06-09 Andrew Pinski <apinski@marvell.com>
14571 PR tree-optimization/100925
14572 * match.pd (a ? CST1 : CST2): Limit transformations
14573 that would produce a negative to integeral types only.
14574 Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.
14576 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
14579 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
14581 * doc/tm.texi: Correctly update.
14583 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
14585 * doc/tm.texi: Correctly update.
14587 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
14590 * doc/tm.texi.in (Trampolines): Add a missing blank line.
14592 2021-06-09 Paul Eggert <eggert@cs.ucla.edu>
14595 * doc/invoke.texi (Code Gen Options); Document that -fno-trampolines
14596 and -ftrampolines work only with Ada.
14597 * doc/tm.texi.in (Trampolines): Likewise.
14598 * doc/tm.texi: Regenerated.
14600 2021-06-09 Carl Love <cel@us.ibm.com>
14602 * config/rs6000/altivec.h (vec_signextll, vec_signexti, vec_signextq):
14603 Add define for new builtins.
14604 * config/rs6000/altivec.md(altivec_vreveti2): Add define_expand.
14605 * config/rs6000/rs6000-builtin.def (VSIGNEXTI, VSIGNEXTLL): Add
14606 overloaded builtin definitions.
14607 (VSIGNEXTSB2W, VSIGNEXTSH2W, VSIGNEXTSB2D, VSIGNEXTSH2D,VSIGNEXTSW2D,
14608 VSIGNEXTSD2Q): Add builtin expansions.
14609 (SIGNEXT): Add P10 overload definition.
14610 * config/rs6000/rs6000-call.c (P9V_BUILTIN_VEC_VSIGNEXTI, P9V_BUILTIN_VEC_VSIGNEXTLL,
14611 P10_BUILTIN_VEC_SIGNEXT): Add overloaded argument definitions.
14612 * config/rs6000/vsx.md (vsx_sign_extend_v2di_v1ti): Add define_insn.
14613 (vsignextend_v2di_v1ti, vsignextend_qi_<mode>, vsignextend_hi_<mode>,
14614 vsignextend_si_v2di)[VIlong]: Add define_expand.
14615 Make define_insn vsx_sign_extend_si_v2di visible.
14616 * doc/extend.texi: Add documentation for the vec_signexti,
14617 vec_signextll builtins and vec_signextq.
14619 2021-06-09 Carl Love <cel@us.ibm.com>
14621 * config/rs6000/rs6000.c (__fixkfti, __fixunskfti, __floattikf,
14622 __floatuntikf): Names changed to __fixkfti_sw, __fixunskfti_sw,
14623 __floattikf_sw, __floatuntikf_sw respectively.
14624 * config/rs6000/rs6000.md (floatti<mode>2, floatunsti<mode>2,
14625 fix_trunc<mode>ti2, fixuns_trunc<mode>ti2): Add
14626 define_insn for mode IEEE 128.
14628 2021-06-09 Carl Love <cel@us.ibm.com>
14630 * config/rs6000/altivec.md (altivec_vslq, altivec_vsrq):
14631 Rename to altivec_vslq_<mode>, altivec_vsrq_<mode>, mode VEC_TI.
14632 * config/rs6000/vector.md (VEC_TI): Was named VSX_TI in vsx.md.
14633 (vashlv1ti3): Change to vashl<mode>3, mode VEC_TI.
14634 (vlshrv1ti3): Change to vlshr<mode>3, mode VEC_TI.
14635 * config/rs6000/vsx.md (VSX_TI): Remove define_mode_iterator. Update
14636 uses of VSX_TI to VEC_TI.
14638 2021-06-09 Carl Love <cel@us.ibm.com>
14640 * config/rs6000/dfp.md (floattitd2, fixtdti2): New define_insns.
14642 2021-06-09 Carl Love <cel@us.ibm.com>
14644 * config/rs6000/altivec.h (vec_dive, vec_mod): Add define for new
14646 * config/rs6000/altivec.md (UNSPEC_VMULEUD, UNSPEC_VMULESD,
14647 UNSPEC_VMULOUD, UNSPEC_VMULOSD): New unspecs.
14648 (altivec_eqv1ti, altivec_gtv1ti, altivec_gtuv1ti, altivec_vmuleud,
14649 altivec_vmuloud, altivec_vmulesd, altivec_vmulosd, altivec_vrlq,
14650 altivec_vrlqmi, altivec_vrlqmi_inst, altivec_vrlqnm,
14651 altivec_vrlqnm_inst, altivec_vslq, altivec_vsrq, altivec_vsraq,
14652 altivec_vcmpequt_p, altivec_vcmpgtst_p, altivec_vcmpgtut_p): New
14654 (vec_widen_umult_even_v2di, vec_widen_smult_even_v2di,
14655 vec_widen_umult_odd_v2di, vec_widen_smult_odd_v2di, altivec_vrlqmi,
14656 altivec_vrlqnm): New define_expands.
14657 * config/rs6000/rs6000-builtin.def (VCMPEQUT_P, VCMPGTST_P,
14658 VCMPGTUT_P): Add macro expansions.
14659 (BU_P10V_AV_P): Add builtin predicate definition.
14660 (VCMPGTUT, VCMPGTST, VCMPEQUT, CMPNET, CMPGE_1TI,
14661 CMPGE_U1TI, CMPLE_1TI, CMPLE_U1TI, VNOR_V1TI_UNS, VNOR_V1TI, VCMPNET_P,
14662 VCMPAET_P, VMULEUD, VMULESD, VMULOUD, VMULOSD, VRLQ,
14663 VSLQ, VSRQ, VSRAQ, VRLQNM, DIV_V1TI, UDIV_V1TI, DIVES_V1TI, DIVEU_V1TI,
14664 MODS_V1TI, MODU_V1TI, VRLQMI): New macro expansions.
14665 (VRLQ, VSLQ, VSRQ, VSRAQ, DIVE, MOD): New overload expansions.
14666 * config/rs6000/rs6000-call.c (P10_BUILTIN_VCMPEQUT,
14667 P10V_BUILTIN_CMPGE_1TI, P10V_BUILTIN_CMPGE_U1TI,
14668 P10V_BUILTIN_VCMPGTUT, P10V_BUILTIN_VCMPGTST,
14669 P10V_BUILTIN_CMPLE_1TI, P10V_BUILTIN_VCMPLE_U1TI,
14670 P10V_BUILTIN_DIV_V1TI, P10V_BUILTIN_UDIV_V1TI,
14671 P10V_BUILTIN_VMULESD, P10V_BUILTIN_VMULEUD,
14672 P10V_BUILTIN_VMULOSD, P10V_BUILTIN_VMULOUD,
14673 P10V_BUILTIN_VNOR_V1TI, P10V_BUILTIN_VNOR_V1TI_UNS,
14674 P10V_BUILTIN_VRLQ, P10V_BUILTIN_VRLQMI,
14675 P10V_BUILTIN_VRLQNM, P10V_BUILTIN_VSLQ,
14676 P10V_BUILTIN_VSRQ, P10V_BUILTIN_VSRAQ,
14677 P10V_BUILTIN_VCMPGTUT_P, P10V_BUILTIN_VCMPGTST_P,
14678 P10V_BUILTIN_VCMPEQUT_P, P10V_BUILTIN_VCMPGTUT_P,
14679 P10V_BUILTIN_VCMPGTST_P, P10V_BUILTIN_CMPNET,
14680 P10V_BUILTIN_VCMPNET_P, P10V_BUILTIN_VCMPAET_P,
14681 P10V_BUILTIN_DIVES_V1TI, P10V_BUILTIN_MODS_V1TI,
14682 P10V_BUILTIN_MODU_V1TI):
14683 New overloaded definitions.
14684 (rs6000_gimple_fold_builtin) [P10V_BUILTIN_VCMPEQUT,
14685 P10V_BUILTIN_CMPNET, P10V_BUILTIN_CMPGE_1TI,
14686 P10V_BUILTIN_CMPGE_U1TI, P10V_BUILTIN_VCMPGTUT,
14687 P10V_BUILTIN_VCMPGTST, P10V_BUILTIN_CMPLE_1TI,
14688 P10V_BUILTIN_CMPLE_U1TI]: New case statements.
14689 (rs6000_init_builtins) [bool_V1TI_type_node, int_ftype_int_v1ti_v1ti]:
14691 (altivec_init_builtins): New E_V1TImode case statement.
14692 (builtin_function_type)[P10_BUILTIN_128BIT_VMULEUD,
14693 P10_BUILTIN_128BIT_VMULOUD, P10_BUILTIN_128BIT_DIVEU_V1TI,
14694 P10_BUILTIN_128BIT_MODU_V1TI, P10_BUILTIN_CMPGE_U1TI,
14695 P10_BUILTIN_VCMPGTUT, P10_BUILTIN_VCMPEQUT]: New case statements.
14696 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute) [E_TImode,
14697 E_V1TImode]: New case statements.
14698 * config/rs6000/rs6000.h (rs6000_builtin_type_index): New enum
14699 value RS6000_BTI_bool_V1TI.
14700 * config/rs6000/vector.md (vector_gtv1ti,vector_nltv1ti,
14701 vector_gtuv1ti, vector_nltuv1ti, vector_ngtv1ti, vector_ngtuv1ti,
14702 vector_eq_v1ti_p, vector_ne_v1ti_p, vector_ae_v1ti_p,
14703 vector_gt_v1ti_p, vector_gtu_v1ti_p, vrotlv1ti3, vashlv1ti3,
14704 vlshrv1ti3, vashrv1ti3): New define_expands.
14705 * config/rs6000/vsx.md (UNSPEC_VSX_DIVSQ, UNSPEC_VSX_DIVUQ,
14706 UNSPEC_VSX_DIVESQ, UNSPEC_VSX_DIVEUQ, UNSPEC_VSX_MODSQ,
14707 UNSPEC_VSX_MODUQ): New unspecs.
14708 (mulv2di3, vsx_div_v1ti, vsx_udiv_v1ti, vsx_dives_v1ti,
14709 vsx_diveu_v1ti, vsx_mods_v1ti, vsx_modu_v1ti, xxswapd_v1ti): New
14711 (vcmpnet): New define_expand.
14712 * doc/extend.texi: Add documentation for the new builtins vec_rl,
14713 vec_rlmi, vec_rlnm, vec_sl, vec_sr, vec_sra, vec_mule, vec_mulo,
14714 vec_div, vec_dive, vec_mod, vec_cmpeq, vec_cmpne, vec_cmpgt, vec_cmplt,
14715 vec_cmpge, vec_cmple, vec_all_eq, vec_all_ne, vec_all_gt, vec_all_lt,
14716 vec_all_ge, vec_all_le, vec_any_eq, vec_any_ne, vec_any_gt, vec_any_lt,
14717 vec_any_ge, vec_any_le.
14719 2021-06-09 Carl Love <cel@us.ibm.com>
14721 * config/rs6000/altivec.md (altivec_vrl<VI_char>mi): Fix
14722 bug in argument generation.
14724 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
14726 * config/arm/iterators.md (<supf>): Remove VCLZQ_U, VCLZQ_S.
14728 * config/arm/mve.md (mve_vclzq_<supf><mode>): Add '@' prefix,
14729 remove <supf> iterator.
14730 (mve_vclzq_u<mode>): New.
14731 * config/arm/neon.md (clz<mode>2): Rename to neon_vclz<mode>.
14732 (neon_vclz<mode): Move to ...
14733 * config/arm/unspecs.md (VCLZQ_U, VCLZQ_S): Remove.
14734 * config/arm/vec-common.md: ... here. Add support for MVE.
14736 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
14738 * config/arm/mve.md (mve_vhaddq_<supf><mode>): Prefix with '@'.
14739 (@mve_vrhaddq_<supf><mode): Likewise.
14740 * config/arm/neon.md (neon_v<r>hadd<sup><mode>): Likewise.
14741 * config/arm/vec-common.md (avg<mode>3_floor, uavg<mode>3_floor)
14742 (avg<mode>3_ceil", uavg<mode>3_ceil): New patterns.
14744 2021-06-09 imba-tjd <109224573@qq.com>
14746 * doc/invoke.texi: Fix typo.
14748 2021-06-09 Roger Sayle <roger@nextmovesoftware.com>
14750 PR middle-end/53267
14751 * fold-const-call.c (fold_const_call_sss) [CASE_CFN_FMOD]:
14752 Support evaluation of fmod/fmodf/fmodl at compile-time.
14754 2021-06-09 Richard Biener <rguenther@suse.de>
14756 PR tree-optimization/100981
14757 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
14758 gimple_get_lhs to also handle calls.
14759 * tree-vect-slp-patterns.c (complex_pattern::build): Transfer
14762 2021-06-09 Richard Biener <rguenther@suse.de>
14764 PR tree-optimization/97832
14765 * tree-vectorizer.h (_slp_tree::failed): New.
14766 * tree-vect-slp.c (_slp_tree::_slp_tree): Initialize
14768 (_slp_tree::~_slp_tree): Free failed.
14769 (vect_build_slp_tree): Retain failed nodes and record
14770 matches in them, copying that back out when running
14771 into a cached fail. Dump start and end of discovery.
14772 (dt_sort_cmp): New.
14773 (vect_build_slp_tree_2): Handle associatable chains
14774 together doing more aggressive operand swapping.
14776 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
14779 * config.gcc (gcc_cv_initfini_array): Set to yes for Linux and
14781 * doc/install.texi: Require glibc 2.1 and binutils 2.12 for
14782 Linux and GNU targets.
14784 2021-06-09 Richard Biener <rguenther@suse.de>
14786 * tree-vect-stmts.c (vect_is_simple_use): Always get dt
14789 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
14791 * config/arc/arc.md (loop_end): Change it to
14792 define_insn_and_split.
14794 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
14796 * config/arc/arc.md (maddhisi4): Use VMAC2H instruction.
14797 (machi): New pattern.
14798 (umaddhisi4): Use VMAC2HU instruction.
14799 (umachi): New pattern.
14801 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
14803 * config/arc/arc-protos.h (arc_split_move_p): New prototype.
14804 * config/arc/arc.c (arc_split_move_p): New function.
14805 (arc_split_move): Clean up.
14806 * config/arc/arc.md (movdi_insn): Clean up, use arc_split_move_p.
14807 (movdf_insn): Likewise.
14808 * config/arc/simdext.md (mov<VWH>_insn): Likewise.
14810 2021-06-09 Uroš Bizjak <ubizjak@gmail.com>
14813 * config/i386/i386.c (print_operand_address_as): Rename "no_rip"
14814 argument to "raw". Do not emit segment overrides when "raw" is true.
14816 2021-06-09 Martin Liska <mliska@suse.cz>
14818 * doc/gcov.texi: Create a proper JSON files.
14819 * doc/invoke.texi: Remove dots in order to make it a valid
14822 2021-06-09 Xionghu Luo <luoxhu@linux.ibm.com>
14824 * config/rs6000/rs6000-p8swap.c (pattern_is_rotate64): New.
14825 (insn_is_load_p): Use pattern_is_rotate64.
14826 (insn_is_swap_p): Likewise.
14827 (quad_aligned_load_p): Likewise.
14828 (const_load_sequence_p): Likewise.
14829 (replace_swapped_aligned_load): Likewise.
14830 (recombine_lvx_pattern): Likewise.
14831 (recombine_stvx_pattern): Likewise.
14833 2021-06-09 Andrew MacLeod <amacleod@redhat.com>
14835 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Use a
14836 fur_stmt source record.
14837 * gimple-range.cc (fur_source::get_operand): Generic range query.
14838 (fur_source::get_phi_operand): New.
14839 (fur_source::register_dependency): New.
14840 (fur_source::query): New.
14841 (class fur_edge): New. Edge source for operands.
14842 (fur_edge::fur_edge): New.
14843 (fur_edge::get_operand): New.
14844 (fur_edge::get_phi_operand): New.
14845 (fur_edge::query): New.
14846 (fur_stmt::fur_stmt): New.
14847 (fur_stmt::get_operand): New.
14848 (fur_stmt::get_phi_operand): New.
14849 (fur_stmt::query): New.
14850 (class fur_depend): New. Statement source and process dependencies.
14851 (fur_depend::fur_depend): New.
14852 (fur_depend::register_dependency): New.
14853 (class fur_list): New. List source for operands.
14854 (fur_list::fur_list): New.
14855 (fur_list::get_operand): New.
14856 (fur_list::get_phi_operand): New.
14857 (fold_range): New. Instantiate appropriate fur_source class and fold.
14858 (fold_using_range::range_of_range_op): Use new API.
14859 (fold_using_range::range_of_address): Ditto.
14860 (fold_using_range::range_of_phi): Ditto.
14861 (imple_ranger::fold_range_internal): Use fur_depend class.
14862 (fold_using_range::range_of_ssa_name_with_loop_info): Use new API.
14863 * gimple-range.h (class fur_source): Now a base class.
14864 (class fur_stmt): New.
14865 (fold_range): New prototypes.
14866 (fur_source::fur_source): Delete.
14868 2021-06-08 Andrew Pinski <apinski@marvell.com>
14870 PR tree-optimization/25290
14871 * tree-ssa-phiopt.c (xor_replacement): Delete.
14872 (tree_ssa_phiopt_worker): Delete use of xor_replacement.
14873 (match_simplify_replacement): Allow one cheap preparation
14874 statement that can be moved to before the if.
14876 2021-06-08 Pat Haugen <pthaugen@linux.ibm.com>
14878 * config/rs6000/power10.md (power10-fused-load, power10-fused-store,
14879 power10-fused_alu, power10-fused-vec, power10-fused-branch): New.
14881 2021-06-08 Jeff Law <jeffreyalaw@gmail.com>
14883 * config/h8300/logical.md (andqi3_1): Move BCLR case into define_insn_and_split.
14884 Create length attribute on define_insn_and_split. Only split for cases which we
14886 (andqi3_1<cczn>): Renamed from andqi3_1_clobber_flags. Only handle AND here and
14887 fix length computation.
14888 (b<code><mode>msx): Combine QImode and HImode H8/SX patterns using iterator.
14890 2021-06-08 Richard Biener <rguenther@suse.de>
14892 PR tree-optimization/100923
14893 * tree-ssa-sccvn.c (valueize_refs_1): Take a pointer to
14894 the operand vector to be valueized.
14895 (valueize_refs): Likewise.
14896 (valueize_shared_reference_ops_from_ref): Adjust.
14897 (valueize_shared_reference_ops_from_call): Likewise.
14898 (vn_reference_lookup_3): Likewise.
14899 (vn_reference_lookup_pieces): Likewise. Re-valueize
14900 with honoring availability when we are about to create
14901 the ao_ref and valueized before.
14902 (vn_reference_lookup): Likewise.
14903 (vn_reference_insert_pieces): Adjust.
14905 2021-06-08 Richard Biener <rguenther@suse.de>
14907 * tree-vectorizer.h (_slp_instance::root_stmt): Change to...
14908 (_slp_instance::root_stmts): ... a vector.
14909 (SLP_INSTANCE_ROOT_STMT): Rename to ...
14910 (SLP_INSTANCE_ROOT_STMTS): ... this.
14911 (slp_root::root): Change to...
14912 (slp_root::roots): ... a vector.
14913 (slp_root::slp_root): Adjust.
14914 * tree-vect-slp.c (_slp_instance::location): Adjust.
14915 (vect_free_slp_instance): Release the root stmt vector.
14916 (vect_build_slp_instance): Adjust.
14917 (vect_analyze_slp): Likewise.
14918 (_bb_vec_info::~_bb_vec_info): Likewise.
14919 (vect_slp_analyze_operations): Likewise.
14920 (vect_bb_vectorization_profitable_p): Likewise. Adjust
14921 costs for the root stmt.
14922 (vect_slp_check_for_constructors): Gather all BIT_INSERT_EXPRs
14924 (vect_slp_analyze_bb_1): Simplify by marking all root stmts
14926 (vectorize_slp_instance_root_stmt): Adjust.
14927 (vect_schedule_slp): Likewise.
14929 2021-06-08 Aldy Hernandez <aldyh@redhat.com>
14931 * gimple-ssa-evrp.c (class ssa_equiv_stack): New.
14932 (ssa_equiv_stack::ssa_equiv_stack): New.
14933 (ssa_equiv_stack::~ssa_equiv_stack): New.
14934 (ssa_equiv_stack::enter): New.
14935 (ssa_equiv_stack::leave): New.
14936 (ssa_equiv_stack::push_replacement): New.
14937 (ssa_equiv_stack::get_replacement): New.
14938 (is_pointer_ssa): New.
14939 (class pointer_equiv_analyzer): New.
14940 (pointer_equiv_analyzer::pointer_equiv_analyzer): New.
14941 (pointer_equiv_analyzer::~pointer_equiv_analyzer): New.
14942 (pointer_equiv_analyzer::set_global_equiv): New.
14943 (pointer_equiv_analyzer::set_cond_equiv): New.
14944 (pointer_equiv_analyzer::get_equiv): New.
14945 (pointer_equiv_analyzer::enter): New.
14946 (pointer_equiv_analyzer::leave): New.
14947 (pointer_equiv_analyzer::get_equiv_expr): New.
14948 (pta_valueize): New.
14949 (pointer_equiv_analyzer::visit_stmt): New.
14950 (pointer_equiv_analyzer::visit_edge): New.
14951 (hybrid_folder::value_of_expr): Call PTA.
14952 (hybrid_folder::value_on_edge): Same.
14953 (hybrid_folder::pre_fold_bb): New.
14954 (hybrid_folder::post_fold_bb): New.
14955 (hybrid_folder::pre_fold_stmt): New.
14956 (rvrp_folder::pre_fold_bb): New.
14957 (rvrp_folder::post_fold_bb): New.
14958 (rvrp_folder::pre_fold_stmt): New.
14959 (rvrp_folder::value_of_expr): Call PTA.
14960 (rvrp_folder::value_on_edge): Same.
14962 2021-06-08 Jakub Jelinek <jakub@redhat.com>
14965 * tree-inline.c (copy_tree_body_r): For OMP_CLAUSE_DEPEND don't
14966 check TREE_CODE if OMP_CLAUSE_DECL is NULL.
14968 2021-06-08 Richard Biener <rguenther@suse.de>
14970 PR middle-end/100951
14971 * tree-vect-generic.c (expand_vector_piecewise): Build a
14972 VECTOR_CST if all elements are constant.
14973 (expand_vector_condition): Likewise.
14974 (lower_vec_perm): Likewise.
14975 (expand_vector_conversion): Likewise.
14977 2021-06-08 Martin Liska <mliska@suse.cz>
14979 * doc/invoke.texi: Document new param evrp-sparse-threshold.
14981 2021-06-08 Martin Liska <mliska@suse.cz>
14983 * genautomata.c (create_automata): Fix typo.
14985 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
14987 PR tree-optimization/100794
14988 * tree-predcom.c (tree_predictive_commoning_loop): Add parameter
14989 allow_unroll_p and only allow unrolling when it's true.
14990 (tree_predictive_commoning): Add parameter allow_unroll_p and
14992 (run_tree_predictive_commoning): Likewise.
14993 (pass_predcom::gate): Check flag_tree_loop_vectorize and
14994 global_options_set.x_flag_predictive_commoning.
14995 (pass_predcom::execute): Adjust for allow_unroll_p.
14997 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
14999 * tree-predcom.c (execute_pred_commoning): Remove update_ssa call.
15000 (tree_predictive_commoning_loop): Factor some cleanup stuffs into
15001 lambda function cleanup, remove scev_reset call, and adjust return
15003 (tree_predictive_commoning): Adjust for different changed values,
15004 only set flag TODO_update_ssa_only_virtuals if changed.
15005 (pass_data pass_data_predcom): Remove TODO_update_ssa_only_virtuals
15006 from todo_flags_finish.
15008 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
15010 * gimple-range-cache.cc (class sbr_sparse_bitmap): New.
15011 (sbr_sparse_bitmap::sbr_sparse_bitmap): New.
15012 (sbr_sparse_bitmap::bitmap_set_quad): New.
15013 (sbr_sparse_bitmap::bitmap_get_quad): New.
15014 (sbr_sparse_bitmap::set_bb_range): New.
15015 (sbr_sparse_bitmap::get_bb_range): New.
15016 (sbr_sparse_bitmap::bb_range_p): New.
15017 (block_range_cache::block_range_cache): initialize bitmap obstack.
15018 (block_range_cache::~block_range_cache): Destruct obstack.
15019 (block_range_cache::set_bb_range): Decide when to utilze the
15020 sparse on entry cache.
15021 * gimple-range-cache.h (block_range_cache): Add bitmap obstack.
15022 * params.opt (-param=evrp-sparse-threshold): New.
15024 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
15026 * bitmap.c (bitmap_set_aligned_chunk): New.
15027 (bitmap_get_aligned_chunk): New.
15028 (test_aligned_chunk): New.
15029 (bitmap_c_tests): Call test_aligned_chunk.
15030 * bitmap.h (bitmap_set_aligned_chunk, bitmap_get_aligned_chunk): New.
15032 2021-06-07 Uroš Bizjak <ubizjak@gmail.com>
15035 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
15037 (ix86_expand_vector_init_one_nonzero): Ditto.
15038 (ix86_expand_vector_init_one_var): Ditto.
15039 (ix86_expand_vector_init_general): Ditto.
15040 * config/i386/mmx.md (vec_initv4qiqi): New expander.
15042 2021-06-07 Jeff Law <jeffreyalaw@gmail.com>
15044 * config/h8300/movepush.md: Change most _clobber_flags
15045 patterns to instead use <cczn> subst.
15046 (movsi_cczn): New pattern with usable CC cases split out.
15047 (movsi_h8sx_cczn): Likewise.
15049 2021-06-07 Martin Liska <mliska@suse.cz>
15051 * common/common-target.def: Split long lines and replace them
15053 * target.def: Likewise.
15054 * doc/tm.texi: Re-generated.
15056 2021-06-07 Jakub Jelinek <jakub@redhat.com>
15059 * fold-const.c (fold_read_from_vector): Return NULL if trying to
15060 read from a CONSTRUCTOR with vector type elements.
15062 2021-06-07 Jakub Jelinek <jakub@redhat.com>
15064 PR middle-end/100898
15065 * tree-inline.c (copy_bb): Only use gimple_call_arg_ptr if memcpy
15066 should copy any arguments. Don't call gimple_call_num_args
15067 on id->call_stmt or call_stmt more than once.
15069 2021-06-07 liuhongt <hongtao.liu@intel.com>
15072 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): Refine
15074 (<insn>v4siv4di2): Delete constraints for define_expand.
15076 2021-06-07 liuhongt <hongtao.liu@intel.com>
15079 * config/i386/i386-expand.c (ix86_expand_builtin): Remove
15080 assignment of cfun->machine->has_explicit_vzeroupper.
15081 * config/i386/i386-features.c
15082 (ix86_add_reg_usage_to_vzerouppers): Delete.
15083 (ix86_add_reg_usage_to_vzeroupper): Ditto.
15084 (rest_of_handle_insert_vzeroupper): Remove
15085 ix86_add_reg_usage_to_vzerouppers, add df_analyze at the end
15087 (gate): Remove cfun->machine->has_explicit_vzeroupper.
15088 * config/i386/i386-protos.h (ix86_expand_avx_vzeroupper):
15090 * config/i386/i386.c (ix86_insn_callee_abi): New function.
15091 (ix86_initialize_callee_abi): Ditto.
15092 (ix86_expand_avx_vzeroupper): Ditto.
15093 (ix86_hard_regno_call_part_clobbered): Adjust for vzeroupper
15095 (TARGET_INSN_CALLEE_ABI): Define as ix86_insn_callee_abi.
15096 (ix86_emit_mode_set): Call ix86_expand_avx_vzeroupper
15098 * config/i386/i386.h (struct GTY(()) machine_function): Delete
15099 has_explicit_vzeroupper.
15100 * config/i386/i386.md (enum unspec): New member
15102 (ABI_DEFAULT,ABI_VZEROUPPER,ABI_UNKNOWN): New
15103 define_constants for insn callee abi index.
15104 * config/i386/predicates.md (vzeroupper_pattern): Adjust.
15105 * config/i386/sse.md (UNSPECV_VZEROUPPER): Deleted.
15106 (avx_vzeroupper): Call ix86_expand_avx_vzeroupper.
15107 (*avx_vzeroupper): Rename to ..
15108 (avx_vzeroupper_callee_abi): .. this, and adjust pattern as
15109 call_insn which has a special vzeroupper ABI.
15110 (*avx_vzeroupper_1): Deleted.
15112 2021-06-07 liuhongt <hongtao.liu@intel.com>
15115 * df-scan.c (df_get_call_refs): When call_insn is a fake call,
15116 it won't use stack pointer reg.
15117 * final.c (leaf_function_p): When call_insn is a fake call, it
15118 won't affect caller as a leaf function.
15119 * reg-stack.c (callee_clobbers_any_stack_reg): New.
15120 (subst_stack_regs): When call_insn doesn't clobber any stack
15121 reg, don't clear the arguments.
15122 * rtl.c (shallow_copy_rtx): Don't clear flag used when orig is
15124 * shrink-wrap.c (requires_stack_frame_p): No need for stack
15125 frame for a fake call.
15126 * rtl.h (FAKE_CALL_P): New macro.
15128 2021-06-06 Eric Botcazou <ebotcazou@adacore.com>
15130 * config/sparc/sparc-protos.h (order_regs_for_local_alloc): Rename
15132 (sparc_order_regs_for_local_alloc): ...this.
15133 (sparc_leaf_reg_remap): Declare.
15134 * config/sparc/sparc.h (ADJUST_REG_ALLOC_ORDER): Adjust.
15135 (LEAF_REG_REMAP): Reimplement as call to sparc_leaf_reg_remap.
15136 * config/sparc/sparc.c (leaf_reg_remap): Delete.
15137 (order_regs_for_local_alloc): Rename to...
15138 (sparc_order_regs_for_local_alloc): ...this.
15139 (sparc_leaf_reg_remap): New function.
15140 (sparc_conditional_register_usage): Do not modify leaf_reg_remap.
15142 2021-06-06 David Edelsohn <dje.gcc@gmail.com>
15144 * config/rs6000/rs6000.c (rs6000_xcoff_asm_output_aligned_decl_common):
15145 Use assemble_name to output BSS section name.
15147 2021-06-06 Uroš Bizjak <ubizjak@gmail.com>
15149 * config/i386/constraints.md (Bs):
15150 Remove boolean operators from match_test RTX.
15153 (M): Use "mode" variable instead of GET_MODE (op) in match_test RTX.
15156 2021-06-06 Martin Liska <mliska@suse.cz>
15158 * doc/extend.texi: Add missing @headitem.
15159 * doc/invoke.texi: Likewise.
15160 * doc/objc.texi: Likewise.
15162 2021-06-06 Martin Liska <mliska@suse.cz>
15164 * genhooks.c (emit_findices): Remove unused function.
15165 (emit_documentation): Do not call emit_findices
15166 and do not search for @Fcode directives.
15168 2021-06-06 Martin Liska <mliska@suse.cz>
15170 * doc/invoke.texi: Remove extra character.
15172 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
15174 * config/sh/sh.md (doloop_end_split): Fix empty split condition.
15176 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
15178 * config/sparc/sparc.md (*snedi<W:mode>_zero_vis3,
15179 *neg_snedi<W:mode>_zero_subxc, *plus_snedi<W:mode>_zero,
15180 *plus_plus_snedi<W:mode>_zero, *minus_snedi<W:mode>_zero,
15181 *minus_minus_snedi<W:mode>_zero): Fix empty split condition.
15183 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
15185 * config/or1k/or1k.md (*movdi): Fix empty split condition.
15187 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
15189 * config/mips/mips.md (<anonymous>, bswapsi2, bswapdi2): Fix empty
15192 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
15194 * config/m68k/m68k.md (*zero_extend_inc, *zero_extend_dec,
15195 *zero_extendsidi2): Fix empty split condition.
15197 2021-06-05 Jeff Law <jeffreyalaw@gmail.com>
15199 * config/h8300/addsub.md: Fix split condition in define_insn_and_split
15201 * config/h8300/bitfield.md: Likewise.
15202 * config/h8300/combiner.md: Likewise.
15203 * config/h8300/divmod.md: Likewise.
15204 * config/h8300/extensions.md: Likewise.
15205 * config/h8300/jumpcall.md: Likewise.
15206 * config/h8300/movepush.md: Likewise.
15207 * config/h8300/multiply.md: Likewise.
15208 * config/h8300/other.md: Likewise.
15209 * config/h8300/shiftrotate.md: Likewise.
15210 * config/h8300/logical.md: Likewise. Fix split pattern to use
15211 code iterator that somehow slipped through.
15213 2021-06-04 Tobias Burnus <tobias@codesourcery.com>
15215 PR middle-end/100905
15216 * tree-nested.c (convert_nonlocal_omp_clauses,
15217 convert_local_omp_clauses): Handle OMP_CLAUSE_BIND.
15219 2021-06-04 Martin Sebor <msebor@redhat.com>
15221 PR middle-end/100732
15222 * gimple-fold.c (gimple_fold_builtin_sprintf): Avoid folding calls
15223 with either source or destination argument of invalid type.
15224 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Avoid checking
15225 calls with arguments of invalid type.
15227 2021-06-04 Martin Sebor <msebor@redhat.com>
15229 * attribs.c (init_attr_rdwr_indices): Use VLA bounds in the expected
15231 (attr_access::vla_bounds): Also handle VLA bounds.
15233 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
15235 * config/i386/predicates.md (GOT_memory_operand):
15236 Implement using match_code RTXes.
15237 (GOT32_symbol_operand): Ditto.
15239 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
15242 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
15244 (ix86_expand_vector_init_general): Ditto.
15245 Use SImode instead of word_mode for logic operations
15246 when GET_MODE_SIZE (mode) < UNITS_PER_WORD.
15247 (expand_vec_perm_even_odd_1): Assert that V2HI mode should be
15248 implemented by expand_vec_perm_1.
15249 (expand_vec_perm_broadcast_1): Assert that V2HI and V4HI modes
15250 should be implemented using standard shuffle patterns.
15251 (ix86_vectorize_vec_perm_const): Handle V2HImode. Add V4HI and
15252 V2HI modes to modes, implementable with shuffle for one operand.
15253 * config/i386/mmx.md (*punpckwd): New insn_and_split pattern.
15254 (*pshufw_1): New insn pattern.
15255 (*vec_dupv2hi): Ditto.
15256 (vec_initv2hihi): New expander.
15258 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
15260 * config/arm/vfp.md (no_literal_pool_df_immediate,
15261 no_literal_pool_sf_immediate): Fix empty split condition.
15263 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
15265 * config/i386/i386.md (*load_tp_x32_zext, *add_tp_x32_zext,
15266 *tls_dynamic_gnu2_combine_32): Fix empty split condition.
15267 * config/i386/sse.md (*<sse2_avx2>_pmovmskb_lt,
15268 *<sse2_avx2>_pmovmskb_zext_lt, *sse2_pmovmskb_ext_lt,
15269 *<sse4_1_avx2>_pblendvb_lt): Likewise.
15271 2021-06-04 Jakub Jelinek <jakub@redhat.com>
15274 * config/i386/i386-expand.c (ix86_expand_vector_init): Handle
15275 concatenation from half-sized modes with TImode elements.
15277 2021-06-04 Claudiu Zissulescu <claziss@synopsys.com>
15279 * config/arc/arc.c (arc_override_options): Disable millicode
15280 thunks when RF16 is on.
15282 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
15284 * config/rs6000/rs6000.h (PROMOTE_MODE): Remove.
15286 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
15288 * config/rs6000/rs6000-call.c (rs6000_promote_function_mode):
15289 Replace PROMOTE_MODE marco with its content.
15291 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
15293 * config/cris/cris.md (*addi_reload): Fix empty split condition.
15295 2021-06-03 Jim Wilson <jimw@sifive.com>
15297 * config.gcc (riscv*-*-*): If --with-riscv-attribute not used,
15298 turn it on for all riscv targets.
15300 2021-06-03 Uroš Bizjak <ubizjak@gmail.com>
15303 * config/i386/i386-expand.c (ix86_expand_vector_set):
15304 Handle V2HI and V4QI modes.
15305 (ix86_expand_vector_extract): Ditto.
15306 * config/i386/mmx.md (*pinsrw): New insn pattern.
15309 (*pextrw_zext): Ditto.
15311 (*pextrb_zext): Ditto.
15312 (vec_setv2hi): New expander.
15313 (vec_extractv2hihi): Ditto.
15314 (vec_setv4qi): Ditto.
15315 (vec_extractv4qiqi): Ditto.
15316 (vec_setv8qi): Enable only for TARGET_SSE4_1.
15317 (vec_extractv8qiqi): Ditto.
15319 2021-06-03 Aaron Sawdey <acsawdey@linux.ibm.com>
15321 * config/rs6000/genfusion.pl (gen_logical_addsubf): Fix input
15322 order to subf instruction.
15323 * config/rs6000/fusion.md: Regenerate.
15325 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
15327 * calls.c (get_size_range): Use range_of_expr instead of
15328 determine_value_range.
15329 * tree-affine.c (expr_to_aff_combination): Same.
15330 * tree-data-ref.c (split_constant_offset): Same.
15331 * tree-vrp.c (determine_value_range_1): Remove.
15332 (determine_value_range): Remove.
15333 * tree-vrp.h (determine_value_range): Remove.
15335 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
15337 * function-tests.c (test_ranges): Call gimple_range_tests.
15338 * gimple-range-cache.cc (ranger_cache::range_of_expr): Pass stmt
15340 * gimple-range.cc (fur_source::get_operand): Do not call
15341 get_tree_range or gimple_range_global.
15343 (get_tree_range): Move to value-query.cc.
15344 Call get_arith_expr_range.
15345 (gimple_ranger::range_of_expr): Add argument to get_tree_range.
15346 Include gimple-range-tests.cc.
15347 * gimple-range.h (fold_range): Add argument.
15348 (get_tree_range): Remove.
15349 * selftest.h (gimple_range_tests): New.
15350 * value-query.cc (global_range_query::range_of_expr): Add
15352 (range_query::get_tree_range): Move from gimple-range.cc.
15353 * value-query.h (class range_query): Add get_tree_range and
15354 get_arith_expr_range. Make fur_source a friend.
15355 * vr-values.c (vr_values::range_of_expr): Pass stmt to
15357 * gimple-range-tests.cc: New file.
15359 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
15361 * gimple-range.cc (gimple_ranger::export_global_ranges): Call
15362 update_global_range.
15363 * value-query.cc (update_global_range): New.
15364 * value-query.h (update_global_range): New.
15366 2021-06-03 David Malcolm <dmalcolm@redhat.com>
15368 * diagnostic-show-locus.c (diagnostic_show_locus): Don't reject
15369 printing the same location twice if there are fix-it hints,
15370 multiple locations, or a label.
15372 2021-06-03 Andre Vieira <andre.simoesdiasvieira@arm.com>
15374 * tree-vect-loop.c (vect_transform_loop): Use main loop's various'
15375 thresholds to narrow the upper bound on epilogue iterations.
15377 2021-06-03 Christophe Lyon <christophe.lyon@linaro.org>
15379 * config/arm/mve.md (mve_vabsq_f<mode>): Use 'abs' instead of unspec.
15380 (mve_vabsq_s<mode>): Likewise.
15381 * config/arm/neon.md (abs<mode>2): Rename to neon_abs<mode>2.
15382 * config/arm/unspecs.md (VABSQ_F, VABSQ_S): Delete.
15383 * config/arm/vec-common.md (neg<mode>2): Rename to
15384 <absneg_str><mode>2.
15386 2021-06-03 Claudiu Zissulescu <claziss@synopsys.com>
15388 * common/config/arc/arc-common.c (arc_option_optimization_table):
15389 Remove malign-call.
15390 * config/arc/arc.c (arc_unalign_branch_p): Remove unused function.
15391 * config/arc/arc.h (TARGET_MIXED_CODE): Remove macro.
15392 (INDEX_REG_CLASS): Only refer to GENERAL_REGS.
15393 * config/arc/arc.md (abssi2_mixed): Remove pattern.
15394 * config/arc/arc.opt (munalign-prob-threshold): Mark it obsolete.
15395 (malign-call): Likewise.
15396 (mmixed-code): Likewise.
15397 * doc/invoke.texi (ARC): Update doc.
15399 2021-06-03 Martin Liska <mliska@suse.cz>
15401 * common.opt: Use proper Enum values.
15402 * opts.c (COVERAGE_SANITIZER_OPT): Remove.
15403 (parse_sanitizer_options): Handle only sanitizer_opts.
15404 (common_handle_option): Just assign value.
15406 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
15409 * tree-inline.c (inline_forbidden_p): Remove test on return type.
15411 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
15413 * dwarf2out.c (loc_list_from_tree_1) <FUNCTION_DECL>: Also generate
15414 DW_OP_GNU_variable_value referencing an existing DIE at file scope.
15415 (type_byte_size): Inline into...
15416 (add_byte_size_attribute): ...this and call add_scalar_info.
15418 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
15420 * dwarf2out.c (mem_loc_descriptor) <UDIV>: Fix typo.
15421 (typed_binop_from_tree): New function.
15422 (loc_list_from_tree_1) <EXACT_DIV_EXPR>: For an unsigned type,
15423 turn a divide by a power of 2 into a shift.
15424 <CEIL_DIV_EXPR>: For an unsigned type, use a signed divide if the
15425 size of the mode is lower than DWARF2_ADDR_SIZE; otherwise, do a
15426 typed divide by calling typed_binop_from_tree.
15428 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
15430 * dwarf2out.c (scompare_loc_descriptor): Fix head comment.
15431 (is_handled_procedure_type): Likewise.
15432 (struct loc_descr_context): Add strict_signedness field.
15433 (resolve_args_picking_1): Deal with DW_OP_[GNU_]deref_type,
15434 DW_OP_[GNU_]convert and DW_OP_[GNU_]reinterpret.
15435 (resolve_args_picking): Minor tweak.
15436 (function_to_dwarf_procedure): Initialize strict_signedness field.
15437 (type_byte_size): Likewise.
15438 (field_byte_offset): Likewise.
15439 (gen_descr_array_type_die): Likewise.
15440 (gen_variant_part): Likewise.
15441 (loc_list_from_tree_1) <CALL_EXPR>: Tidy up and set strict_signedness
15442 to true when a context is present before evaluating the arguments.
15443 <COND_EXPR>: Do not generate a useless comparison with zero.
15444 When dereferencing an address, if strict_signedness is true and the
15445 type is small and signed, use DW_OP_deref_type to do the dereference
15446 and then DW_OP_convert to convert back to the generic type.
15448 2021-06-03 Jakub Jelinek <jakub@redhat.com>
15451 * tree-inline.c (copy_tree_body_r): Handle iterators on
15452 OMP_CLAUSE_AFFINITY or OMP_CLAUSE_DEPEND.
15454 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
15456 * config/arc/arc.md (*bbit_di): Remove.
15458 2021-06-02 Christoph Muellner <cmuellner@gcc.gnu.org>
15460 PR rtl-optimization/100264
15461 * ree.c (get_sub_rtx): Ignore SET expressions without register
15462 destinations and remove assertion, as it is not valid anymore
15463 with this new behaviour.
15464 (merge_def_and_ext): Eliminate destination check for register
15465 as such SET expressions can't occur anymore.
15466 (combine_reaching_defs): Likewise.
15468 2021-06-02 Jakub Jelinek <jakub@redhat.com>
15471 * config/xtensa/xtensa.h (LEAF_REG_REMAP): Cast REGNO to int to avoid
15472 -Wtype-limits warnings.
15473 (DWARF_FRAME_REGISTER): Rewrite into ternary operator with addition
15474 in operands to avoid -Wsign-compare warnings.
15476 2021-06-02 Pat Haugen <pthaugen@linux.ibm.com>
15478 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): Use
15481 2021-06-02 Vineet Gupta <vgupta@synopsys.com>
15483 * config/arc/arc.h (TARGET_CPU_DEFAULT): Change to hs38_linux.
15485 2021-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
15487 * config/s390/s390.md(*ashrdi3_31<setcc><cconly>): Use a single
15489 * config/s390/subst.md(cconly_subst): Use a single constraint
15490 in (match_scratch).
15492 2021-06-02 Martin Liska <mliska@suse.cz>
15494 * ipa-icf.h: Use auto_vec for memory_access_types.
15496 2021-06-02 Jeff Law <jeffreyalaw@gmail.com>
15498 * config/h8300/h8300-protos.h (compute_a_shift_length): Drop unused
15499 argument from prototype.
15500 (output_logical_op): Add rtx_code argument.
15501 (compute_logical_op_length): Likewise.
15502 * config/h8300/h8300.c (h8300_and_costs): Pass additional argument
15503 to compute_a_shift_length.
15504 (output_logical_op); New argument with the rtx code rather than
15505 extracting it from an operand. Handle QImode too.
15506 (compute_logical_op_length): Similary.
15507 (compute_a_shift_length): Drop unused argument.
15508 * config/h8300/h8300.md (logicals): New code iterator.
15509 * config/h8300/logical.md (<code><mode>3 expander): Combine
15510 the "and" expander with the "ior"/"xor" expander.
15511 (bclr<mode>msx): Combine the QI/HI mode patterns.
15512 (<logical><mode>3 insns): Use code iterator rather than match_operator.
15513 Handle QImode as well. Update call to output_logical_op and
15514 compute_logical_op_length to pass in rtx_code
15515 Fix split condition on all define_insn_and_split patterns.
15516 (one_cmpl<mode>2<cczn>): Use <cczn> to support both clobbering
15517 the flags and setting ZN via existing define_subst.
15518 * config/h8300/shiftrotate.md: Drop unused argument from
15519 calls to compute_a_shift_length.
15520 Signed-off-by: Jeff Law <jeffreyalaw@gmail.com>
15522 2021-06-01 Andrew Pinski <apinski@marvell.com>
15524 PR tree-optimization/25290
15525 * tree-ssa-phiopt.c (match_simplify_replacement):
15527 (tree_ssa_phiopt_worker): Use match_simplify_replacement.
15528 (two_value_replacement): Change the comment about
15529 conditional_replacement.
15530 (conditional_replacement): Delete.
15532 2021-06-01 Andrew Pinski <apinski@marvell.com>
15534 PR tree-optimization/95481
15535 * tree-tailcall.c (find_tail_calls): Handle empty typed
15538 2021-06-01 Andrew Pinski <apinski@marvell.com>
15540 * gimplify.c (zero_sized_field_decl): Delete
15541 (zero_sized_type): Delete
15542 (gimplify_init_ctor_eval): Use is_empty_type instead
15543 of zero_sized_field_decl.
15544 (gimplify_modify_expr): Use is_empty_type instead of
15547 2021-06-01 Jason Merrill <jason@redhat.com>
15550 * tree.h (CALL_FROM_NEW_OR_DELETE_P): Adjust comment.
15552 2021-06-01 Jason Merrill <jason@redhat.com>
15555 * diagnostic.h (warning_enabled_at): Declare.
15556 * diagnostic.c (diagnostic_enabled): Factor out from...
15557 (diagnostic_report_diagnostic): ...here.
15558 (warning_enabled_at): New.
15560 2021-06-01 Aldy Hernandez <aldyh@redhat.com>
15562 * gimple-ssa-evrp.c: Enable exporting of global ranges.
15564 2021-06-01 Martin Liska <mliska@suse.cz>
15567 * doc/invoke.texi: Mention that -fgcse-after-reload
15568 is enabled with -O3.
15570 2021-06-01 liuhongt <hongtao.liu@intel.com>
15572 PR tree-optimization/98365
15573 * tree-if-conv.c (strip_nop_cond_scalar_reduction): New function.
15574 (is_cond_scalar_reduction): Handle nop_expr in cond scalar reduction.
15575 (convert_scalar_cond_reduction): Ditto.
15576 (predicate_scalar_phi): Ditto.
15578 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
15580 PR tree-optimization/100781
15581 * gimple-range-cache.cc (ranger_cache::ranger_cache): Enable new
15582 value calculation by default.
15583 (ranger_cache::enable_new_values): New.
15584 (ranger_cache::disable_new_values): New.
15585 (ranger_cache::push_poor_value): Check if new values are allowed.
15586 * gimple-range-cache.h (class ranger_cache): New member/methods.
15587 * gimple-range.cc (gimple_ranger::range_of_expr): Check for debug
15588 statement, and disable/renable new value calculation.
15590 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
15592 * gimple-range-cache.cc (ranger_cache::ssa_range_in_bb): Delete.
15593 (ranger_cache::range_of_def): New.
15594 (ranger_cache::entry_range): New.
15595 (ranger_cache::exit_range): New.
15596 (ranger_cache::range_of_expr): Adjust.
15597 (ranger_cache::range_on_edge): Adjust.
15598 (ranger_cache::propagate_cache): Call exit_range directly.
15599 * gimple-range-cache.h (class ranger_cache): Adjust.
15601 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
15603 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust for
15604 gori_compute being a member rather than base class.
15605 dervied call to member call.
15606 (ranger_cache::dump): No longer dump gori_map.
15607 (ranger_cache::dump_bb): New.
15608 (ranger_cache::get_non_stale_global_range): Adjust for gori_compute
15609 being a member rather than base class.
15610 (ranger_cache::set_global_range): Ditto.
15611 (ranger_cache::ssa_range_in_bb): Ditto.
15612 (ranger_cache::range_of_expr): New.
15613 (ranger_cache::range_on_edge): New.
15614 (ranger_cache::block_range): Adjust for gori_computes. Debug changes.
15615 (ranger_cache::propagate_cache): Adjust debugging output.
15616 (ranger_cache::fill_block_cache): Adjust for gori_computes. Debug
15618 * gimple-range-cache.h (class ranger_cache): Make gori_compute a
15619 member, and inherit from range_query instead.
15620 (ranger_cache::dump_bb): New. split from dump.
15621 * gimple-range-gori.cc (gori_compute::ssa_range_in_bb): Delete.
15622 (gori_compute::expr_range_at_stmt): Delete.
15623 (gori_compute::compute_name_range_op): Delete.
15624 (gori_compute::compute_operand_range_switch): Add fur_source.
15625 (gori_compute::compute_operand_range): Add fur_source param, inline
15626 old compute_name_range_op and optimize_logical_operands.
15627 (struct tf_range): Delete.
15628 (gori_compute::logical_combine): Adjust
15629 (gori_compute::optimize_logical_operands): Delete.
15630 (gori_compute::compute_logical_operands_in_chain): Delete.
15631 (gori_compute::compute_logical_operands): Adjust.
15632 (gori_compute::compute_operand1_range): Adjust to fur_source.
15633 (gori_compute::compute_operand2_range): Ditto.
15634 (gori_compute::compute_operand1_and_operand2_range): Ditto.
15635 (gori_compute::outgoing_edge_range_p): Add range_query parameter,
15636 and adjust to fur_source.
15637 * gimple-range-gori.h (class gori_compute): Simplify and adjust to
15638 range_query and fur_source.
15639 * gimple-range.cc (gimple_ranger::range_on_edge): Query range_on_edge
15640 from the ranger_cache..
15641 (gimple_ranger::fold_range_internal): Adjust to base class change of
15643 (gimple_ranger::dump_bb): Adjust dump.
15644 * gimple-range.h (gimple_ranger):export gori computes object.
15646 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
15648 PR tree-optimization/100774
15649 * gimple-range-cache.cc (ranger_cache::get_non_stale_global_range):
15650 Constant values are also not stale.
15651 (ranger_cache::set_global_range): Range invariant values should also
15652 have the correct timestamp.
15654 2021-05-31 Martin Liska <mliska@suse.cz>
15656 * tree-streamer-in.c (unpack_ts_function_decl_value_fields):
15657 Unpack FUNCTION_DECL_DECL_TYPE.
15658 * tree-streamer-out.c (pack_ts_function_decl_value_fields):
15659 Stream FUNCTION_DECL_DECL_TYPE instead of
15660 DECL_IS_OPERATOR_NEW_P.
15661 * tree.h (set_function_decl_type): Use FUNCTION_DECL_DECL_TYPE
15663 (DECL_IS_OPERATOR_NEW_P): Likewise.
15664 (DECL_IS_OPERATOR_DELETE_P): Likewise.
15665 (DECL_LAMBDA_FUNCTION_P): Likewise.
15667 2021-05-31 Richard Biener <rguenther@suse.de>
15670 * internal-fn.c (expand_SHUFFLEVECTOR): Define.
15671 * internal-fn.def (SHUFFLEVECTOR): New.
15672 * internal-fn.h (expand_SHUFFLEVECTOR): Declare.
15673 * doc/extend.texi: Document __builtin_shufflevector.
15675 2021-05-31 Peter Bergner <bergner@linux.ibm.com>
15678 * config/rs6000/predicates.md(mma_assemble_input_operand): Allow
15679 indexed form addresses.
15681 2021-05-29 Jeff Law <jlaw@tachyum.com>
15683 * config/h8300/h8300.c (h8300_emit_stack_adjustment): Drop unused
15684 parameter. Call callers fixed.
15686 (output_plussi): Add FALLTHRU markers.
15687 (h8300_shift_needs_scratch_p): Add gcc_unreachable marker.
15689 2021-05-29 Jakub Jelinek <jakub@redhat.com>
15691 PR middle-end/99928
15692 * gimplify.c (gimplify_scan_omp_clauses): For taskloop simd
15693 combined with parallel, make sure to add shared clause to
15694 parallel for explicit linear clause.
15696 2021-05-29 Aldy Hernandez <aldyh@redhat.com>
15698 PR tree-optimization/100787
15699 * gimple-ssa-evrp.c: Disable exporting of global ranges.
15701 2021-05-28 Jason Merrill <jason@redhat.com>
15703 * tree-iterator.h (struct tree_stmt_iterator): Add operator++,
15704 operator--, operator*, operator==, and operator!=.
15705 (class tsi_range): New.
15707 2021-05-28 Richard Biener <rguenther@suse.de>
15709 PR tree-optimization/100778
15710 * tree-vect-slp.c (vect_build_slp_tree_1): Prevent possibly
15711 trapping ops in different BBs.
15713 2021-05-28 Richard Biener <rguenther@suse.de>
15716 * tree-inline.c (copy_bb): When processing __builtin_va_arg_pack
15717 copy fntype from original call.
15719 2021-05-28 Martin Liska <mliska@suse.cz>
15721 PR gcov-profile/100751
15722 * doc/gcov.texi: Revert partially a hunk that was wrong.
15724 2021-05-28 Cooper Qu <cooper.qu@linux.alibaba.com>
15726 * config/csky/csky-linux-elf.h (HAVE_sync_compare_and_swapqi):
15728 (HAVE_sync_compare_and_swaphi): Likewise.
15729 (HAVE_sync_compare_and_swapsi): Likewise.
15731 2021-05-28 Jakub Jelinek <jakub@redhat.com>
15733 PR middle-end/99928
15734 * tree.h (OMP_CLAUSE_MAP_IMPLICIT): Define.
15736 2021-05-28 Tobias Burnus <tobias@codesourcery.com>
15738 * gimplify.c (gimplify_omp_affinity): New.
15739 (gimplify_scan_omp_clauses): Call it; remove affinity clause afterwards.
15740 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_AFFINITY.
15741 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_AFFINITY.
15742 * tree.c (omp_clause_num_ops, omp_clause_code_name): Add clause.
15743 (walk_tree_1): Handle OMP_CLAUSE_AFFINITY.
15745 2021-05-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
15746 Richard Biener <rguenther@suse.de>
15748 * match.pd <popcount & / + pattern matching>:
15749 When generating popcount directly fails, try doing it in two halves.
15751 2021-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
15753 * Makefile.in (generated_files): Add gimple-match.c and
15756 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
15758 * gensupport.c (alter_predicate_for_insn): Handle MATCH_DUP.
15760 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
15762 * gensupport.c (alter_constraints): Add MATCH_SCRATCH case.
15764 2021-05-28 Kewen Lin <linkw@linux.ibm.com>
15766 PR tree-optimization/99398
15767 * tree-ssa-forwprop.c (simplify_permutation): Optimize some cases
15768 where the fed operands are CTOR/CST and propagated through
15769 VIEW_CONVERT_EXPR. Call vec_perm_indices::new_shrunk_vector.
15770 * vec-perm-indices.c (vec_perm_indices::new_shrunk_vector): New
15772 * vec-perm-indices.h (vec_perm_indices::new_shrunk_vector): New
15775 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
15777 * config/i386/mmx.md (addv2sf3): Do not call
15778 ix86_fixup_binary_operands_no_copy.
15781 (<smaxmin:code>v2sf3): Ditto.
15782 (<plusminus:insn><MMXMODEI:mode>3): Ditto.
15783 (<plusminus:insn><VI_32:mode>3): Remove expander.
15784 (<plusminus:insn><VI_32:mode>3): Rename from
15785 "*<plusminus:insn><VI_32:mode>3".
15786 (mulv4hi): Do not call ix86_fixup_binary_operands_no_copy.
15787 (mulv2hi3): Remove expander.
15788 (mulv2hi3): Rename from *mulv2hi3.
15789 (<s>mulv2hi3_highpart): Remove expander.
15790 (<s>mulv2hi3_highpart): Rename from *<s>mulv2hi3_highpart.
15791 (<smaxmin:code><MMXMODE14:mode>3): Rename from
15792 "*mmx_<smaxmin:code><MMXMODE14:mode>3".
15793 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Remove expander.
15794 (SMAXMIN_MMXMODEI): Remove mode iterator.
15795 (<smaxmin:code>v4hi3): New expander.
15796 (<smaxmin:code>v4qi3): Rename from *<smaxmin:code>v4qi3.
15797 (<smaxmin:code>v2hi3): Rename from *<smaxmin:code>v2hi3.
15798 (<smaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
15799 (SMAXMIN_VI_32): Remove mode iterator.
15800 (<umaxmin:code><MMXMODE24:mode>3): Rename from
15801 "*mmx_<umaxmin:code><MMXMODE24:mode>3".
15802 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Remove expander.
15803 (UMAXMIN_MMXMODEI): Remove mode iterator.
15804 (<umaxmin:code>v8qi3): New expander.
15805 (<umaxmin:code>v4qi3): Rename from *<umaxmin:code>v4qi3.
15806 (<umaxmin:code>v2hi3): Rename from *<umaxmin:code>v2hi3.
15807 (<umaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
15808 (UMAXMIN_VI_32): Remove mode iterator.
15809 (<any_shift:insn>v2hi3): Remove expander.
15810 (<any_shift:insn>v2hi3): Rename from *<any_shift:insn>v2hi3.
15811 (<any_logic:code><MMXMODEI:mode>3): Do not call
15812 ix86_fixup_binary_operands_no_copy.
15813 (<any_logic:code><VI_32:mode>3): Remove expander.
15814 (<any_logic:code><VI_32:mode>3): Rename from
15815 "*<any_logic:code><VI_32:mode>3".
15816 (uavg<mode>3_ceil): Do not call ix86_fixup_binary_operands_no_copy.
15817 * config/i386/sse.md (div<VF2:mode>3): Do not call
15818 ix86_fixup_binary_operands_no_copy.
15819 (div<VF1:mode>3): Ditto.
15820 (<maxmin:code><VI8_AVX2_AVX512F:mode>3): Ditto.
15821 (smulhrsv4hi3): Ditto.
15822 (smulhrsv2hi3): Ditto.
15824 2021-05-27 Martin Sebor <msebor@redhat.com>
15826 * ggc.h (gt_ggc_mx): Add overloads for all integers.
15828 * hash-map.h (class hash_map): Add pch_nx_helper overloads for all
15830 (hash_map::operator==): New function.
15832 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
15835 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
15836 For TARGET_XOP bypass SSE comparisons for all supported vector modes.
15837 * config/i386/mmx.md (*xop_maskcmp<MMXMODEI:mode>3): New insn pattern.
15838 (*xop_maskcmp<VI_32:mode>3): Ditto.
15839 (*xop_maskcmp_uns<MMXMODEI:mode>3): Ditto.
15840 (*xop_maskcmp_uns<VI_32:mode>3): Ditto.
15842 2021-05-27 Richard Earnshaw <rearnsha@arm.com>
15845 * config/arm/arm.c (arm_configure_build_target): Remove parameter
15846 opts_set, directly check opts parameters for being non-null.
15847 (arm_option_restore): Update call to arm_configure_build_target.
15848 (arm_option_override): Likewise.
15849 (arm_can_inline_p): Likewise.
15850 (arm_valid_target_attribute_tree): Likewise.
15851 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
15852 * config/arm/arm-protos.h (arm_configure_build_target): Adjust
15855 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
15857 * vr-values.c (simplify_conversion_using_ranges): Use
15858 get_range_query instead of get_global_range_query.
15860 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
15862 * gimple-range.cc (get_range_global): Move to value-query.cc.
15863 (gimple_range_global): Same.
15864 (get_global_range_query): Same.
15865 (global_range_query::range_of_expr): Same.
15866 * gimple-range.h (class global_range_query): Move to
15868 (gimple_range_global): Same.
15869 * tree-ssanames.c (get_range_info): Move to value-query.cc.
15870 (get_ptr_nonnull): Same.
15871 * tree-ssanames.h (get_range_info): Remove.
15872 (get_ptr_nonnull): Remove.
15873 * value-query.cc (get_ssa_name_range_info): Move from
15875 (get_ssa_name_ptr_info_nonnull): Same.
15876 (get_range_global): Move from gimple-range.cc.
15877 (gimple_range_global): Same.
15878 (get_global_range_query): Same.
15879 (global_range_query::range_of_expr): Same.
15880 * value-query.h (class global_range_query): Move from
15882 (gimple_range_global): Same.
15884 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
15887 * config/i386/mmx.md (uavgv4qi3_ceil): New insn pattern.
15888 (uavgv2hi3_ceil): Ditto.
15890 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
15893 * doc/extend.texi (scalar_storage_order): Rephrase slightly.
15895 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
15897 * tree-ssanames.c (get_range_info): Merge both copies of
15898 get_range_info into one that works with irange.
15899 * tree-ssanames.h (get_range_info): Remove version that works on
15902 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
15904 * builtins.c (check_nul_terminated_array): Convert to get_range_query.
15905 (expand_builtin_strnlen): Same.
15906 (determine_block_size): Same.
15907 * fold-const.c (expr_not_equal_to): Same.
15908 * gimple-fold.c (size_must_be_zero_p): Same.
15909 * gimple-match-head.c: Include gimple-range.h.
15910 * gimple-pretty-print.c (dump_ssaname_info): Convert to get_range_query.
15911 * gimple-ssa-warn-restrict.c
15912 (builtin_memref::extend_offset_range): Same.
15913 * graphite-sese-to-poly.c (add_param_constraints): Same.
15914 * internal-fn.c (get_min_precision): Same.
15915 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): Same.
15916 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Same.
15918 * tree-data-ref.c (split_constant_offset): Same.
15919 (dr_step_indicator): Same.
15920 * tree-dfa.c (get_ref_base_and_extent): Same.
15921 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
15922 * tree-ssa-loop-niter.c (refine_value_range_using_guard): Same.
15923 (determine_value_range): Same.
15924 (record_nonwrapping_iv): Same.
15925 (infer_loop_bounds_from_signedness): Same.
15926 (scev_var_range_cant_overflow): Same.
15927 * tree-ssa-phiopt.c (two_value_replacement): Same.
15928 * tree-ssa-pre.c (insert_into_preds_of_block): Same.
15929 * tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Same.
15930 * tree-ssa-strlen.c (handle_builtin_stxncpy_strncat): Same.
15932 (dump_strlen_info): Same.
15933 (set_strlen_range): Same.
15934 (maybe_diag_stxncpy_trunc): Same.
15935 (get_len_or_size): Same.
15936 (handle_integral_assign): Same.
15937 * tree-ssa-structalias.c (find_what_p_points_to): Same.
15938 * tree-ssa-uninit.c (find_var_cmp_const): Same.
15939 * tree-switch-conversion.c (bit_test_cluster::emit): Same.
15940 * tree-vect-patterns.c (vect_get_range_info): Same.
15941 (vect_recog_divmod_pattern): Same.
15942 * tree-vrp.c (intersect_range_with_nonzero_bits): Same.
15943 (register_edge_assert_for_2): Same.
15944 (determine_value_range_1): Same.
15945 * tree.c (get_range_pos_neg): Same.
15946 * vr-values.c (vr_values::get_lattice_entry): Same.
15947 (vr_values::update_value_range): Same.
15948 (simplify_conversion_using_ranges): Same.
15950 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
15952 * gimple-ssa-warn-alloca.c (alloca_call_type): Use
15953 get_range_query instead of query argument.
15954 (pass_walloca::execute): Enable and disable global ranger.
15956 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
15958 * gimple-ssa-evrp.c (rvrp_folder::rvrp_folder): Call
15960 (rvrp_folder::~rvrp_folder): Call disable_ranger.
15961 (hybrid_folder::hybrid_folder): Call enable_ranger.
15962 (hybrid_folder::~hybrid_folder): Call disable_ranger.
15964 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
15966 * function.c (allocate_struct_function): Set cfun->x_range_query.
15967 * function.h (struct function): Declare x_range_query.
15968 (get_range_query): New.
15969 (get_global_range_query): New.
15970 * gimple-range-cache.cc (ssa_global_cache::ssa_global_cache):
15971 Remove call to safe_grow_cleared.
15972 * gimple-range.cc (get_range_global): New.
15973 (gimple_range_global): Move from gimple-range.h.
15974 (get_global_range_query): New.
15975 (global_range_query::range_of_expr): New.
15976 (enable_ranger): New.
15977 (disable_ranger): New.
15978 * gimple-range.h (gimple_range_global): Move to gimple-range.cc.
15979 (class global_range_query): New.
15980 (enable_ranger): New.
15981 (disable_ranger): New.
15982 * gimple-ssa-evrp.c (evrp_folder::~evrp_folder): Rename
15983 dump_all_value_ranges to dump.
15984 * tree-vrp.c (vrp_prop::finalize): Same.
15985 * value-query.cc (range_query::dump): New.
15986 * value-query.h (range_query::dump): New.
15987 * vr-values.c (vr_values::dump_all_value_ranges): Rename to...
15988 (vr_values::dump): ...this.
15989 * vr-values.h (class vr_values): Rename dump_all_value_ranges to
15990 dump and make virtual.
15992 2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
15994 * config/i386/i386.c (ix86_autovectorize_vector_modes):
15995 Add V4QImode and V16QImode for TARGET_SSE2.
15996 * doc/sourcebuild.texi (Vector-specific attributes):
15997 Add vect64 and vect32 description.
15999 2021-05-26 Bernd Edlinger <bernd.edlinger@hotmail.de>
16001 * gimple-range-gori.cc (range_def_chain::register_dependency):
16002 Resize m_def_chain when needed.
16004 2021-05-26 Christophe Lyon <christophe.lyon@linaro.org>
16006 * config/arm/mve.md (mve_vaddvq_<supf><mode>): Prefix with '@'.
16007 * config/arm/neon.md (reduc_plus_scal_<mode>): Move to ..
16008 * config/arm/vec-common.md: .. here. Add support for MVE.
16010 2021-05-26 Jakub Jelinek <jakub@redhat.com>
16012 * config/epiphany/epiphany.c (epiphany_print_operand_address): Remove
16014 * config/microblaze/microblaze.c (microblaze_legitimize_address,
16016 microblaze_option_override, print_operand): Likewise.
16017 * config/microblaze/microblaze.md (call_internal_plt,
16018 call_value_intern_plt, call_value_intern): Likewise.
16019 * config/arm/aout.h (ASM_OUTPUT_ALIGN): Likewise.
16020 * config/iq2000/iq2000.md (call_internal1, call_value_internal1,
16021 call_value_multiple_internal1): Likewise.
16022 * config/bfin/bfin.c (symbolic_reference_mentioned_p): Likewise.
16024 2021-05-26 Jan-Benedict Glaw <jbglaw@lug-owl.de>
16026 * config/arc/arc.c (arc_address_cost, arc_print_operand_address,
16027 arc_ccfsm_advance, symbolic_reference_mentioned_p,
16028 arc_raw_symbolic_reference_mentioned_p): Remove register
16031 2021-05-26 Jakub Jelinek <jakub@redhat.com>
16034 * omp-low.c: Include omp-offload.h.
16035 (create_omp_child_function): If current_function_decl has
16036 "omp declare target" attribute and is_gimple_omp_offloaded,
16037 remove that attribute from the copy of attribute list and
16038 add "omp target entrypoint" attribute instead.
16039 (lower_omp_target): Mark .omp_data_sizes.* and .omp_data_kinds.*
16040 variables for offloading if in omp_maybe_offloaded_ctx.
16041 * omp-offload.c (pass_omp_target_link::execute): Nullify second
16042 argument to GOMP_target_data_ext in offloaded code.
16044 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
16046 * config/csky/csky.c (csky_can_change_mode_class): Delete.
16047 For csky, HF/SF mode use the low bits of VREGS.
16049 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
16051 * gimplify.c (gimplify_decl_expr): Do not clear TREE_READONLY on a
16052 DECL which is a reference for OMP.
16054 2021-05-26 Martin Liska <mliska@suse.cz>
16056 PR gcov-profile/100751
16057 * doc/gcov.texi: Document that __gcov_dump can be called just
16058 once and that __gcov_reset resets run-time counters.
16060 2021-05-26 Martin Liska <mliska@suse.cz>
16062 * doc/install.texi: Port relevant part from install-old.texi
16063 and re-generate list of CPUs and systems.
16065 2021-05-26 Martin Liska <mliska@suse.cz>
16067 * Makefile.in: Remove it.
16068 * doc/include/fdl.texi: Update next/previous chapters.
16069 * doc/install.texi: Likewise.
16070 * doc/install-old.texi: Removed.
16072 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
16074 * config/csky/csky.c (ck810_legitimate_index_p): Support
16075 "base + index" with DF mode.
16076 * config/csky/constraints.md ("Y"): New constraint for memory operands
16077 without index register.
16078 * config/csky/csky_insn_fpuv2.md (fpuv3_movdf): Use "Y" instead of "m"
16079 when mov between memory and general registers, and lower their priority.
16080 * config/csky/csky_insn_fpuv3.md (fpuv2_movdf): Likewise.
16082 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
16084 * config/csky/csky.c (TARGET_PROMOTE_PROTOTYPES): Delete.
16086 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
16088 * config/csky/csky.md (untyped_call): Emit clobber for return
16089 registers to mark them used.
16091 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
16093 * config/csky/csky.md (cskyv2_sextend_ldbs): New.
16095 2021-05-26 Andrew Pinski <apinski@marvell.com>
16097 * match.pd (x < 0 ? ~y : y): New patterns.
16099 2021-05-26 Andrew Pinski <apinski@marvell.com>
16101 * match.pd (A?CST1:CST2): Add simplifcations for A?0:+-1, A?+-1:0,
16102 A?POW2:0 and A?0:POW2.
16104 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
16106 * gimple-range-gori.cc (class logical_stmt_cache): Delete
16107 (logical_stmt_cache::logical_stmt_cache ): Delete.
16108 (logical_stmt_cache::~logical_stmt_cache): Delete.
16109 (logical_stmt_cache::cache_entry::dump): Delete.
16110 (logical_stmt_cache::get_range): Delete.
16111 (logical_stmt_cache::cached_name ): Delete.
16112 (logical_stmt_cache::same_cached_name): Delete.
16113 (logical_stmt_cache::cacheable_p): Delete.
16114 (logical_stmt_cache::slot_diagnostics ): Delete.
16115 (logical_stmt_cache::dump): Delete.
16116 (gori_compute_cache::gori_compute_cache): Delete.
16117 (gori_compute_cache::~gori_compute_cache): Delete.
16118 (gori_compute_cache::compute_operand_range): Delete.
16119 (gori_compute_cache::cache_stmt): Delete.
16120 * gimple-range-gori.h (gori_compute::compute_operand_range): Remove
16122 (class gori_compute_cache): Delete.
16124 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
16126 * gimple-range.cc (fold_using_range::range_of_range_op): Use m_gori
16128 (fold_using_range::range_of_address): Adjust.
16129 (fold_using_range::range_of_phi): Adjust.
16130 * gimple-range.h (class fur_source): Adjust.
16131 (fur_source::fur_source): Adjust.
16133 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
16135 * gimple-range-gori.cc (gori_compute::expr_range_at_stmt): Rename
16136 from expr_range_in_bb and adjust.
16137 (gori_compute::compute_name_range_op): Adjust.
16138 (gori_compute::optimize_logical_operands): Adjust.
16139 (gori_compute::compute_logical_operands_in_chain): Adjust.
16140 (gori_compute::compute_operand1_range): Adjust.
16141 (gori_compute::compute_operand2_range): Adjust.
16142 (ori_compute_cache::cache_stmt): Adjust.
16143 * gimple-range-gori.h (gori_compute): Rename prototype.
16145 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
16147 * gimple-range.cc (gimple_ranger::range_of_expr): Non-null should be
16148 checked only after range_of_stmt, not range_on_entry.
16149 (gimple_ranger::range_on_entry): Check for non-null in any
16150 predecessor block, if it is not already non-null.
16151 (gimple_ranger::range_on_exit): DOnt check for non-null after
16152 range on entry call.
16153 (gimple_ranger::dump_bb): New. Split from dump.
16154 (gimple_ranger::dump): Adjust.
16155 * gimple-range.h (class gimple_ranger): Adjust.
16157 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
16159 * gimple-range-cache.cc (struct range_timestamp): Delete.
16160 (class temporal_cache): Adjust.
16161 (temporal_cache::get_timestamp): Delete.
16162 (temporal_cache::set_dependency): Delete.
16163 (temporal_cache::temporal_value): Adjust.
16164 (temporal_cache::current_p): Take dependencies as params.
16165 (temporal_cache::set_timestamp): Adjust.
16166 (temporal_cache::set_always_current): Adjust.
16167 (ranger_cache::get_non_stale_global_range): Adjust.
16168 (ranger_cache::register_dependency): Delete.
16169 * gimple-range-cache.h (class range_cache): Adjust.
16171 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
16173 * gimple-range-gori.cc (range_def_chain::range_def_chain): init
16175 (range_def_chain::~range_def_chain): Dispose of obstack rather than
16176 each individual bitmap.
16177 (range_def_chain::set_import): New.
16178 (range_def_chain::get_imports): New.
16179 (range_def_chain::chain_import_p): New.
16180 (range_def_chain::register_dependency): Rename from build_def_chain
16182 (range_def_chain::def_chain_in_bitmap_p): New.
16183 (range_def_chain::add_def_chain_to_bitmap): New.
16184 (range_def_chain::has_def_chain): Just check first depenedence.
16185 (range_def_chain::get_def_chain): Process imports, use generic
16186 register_dependency routine.
16187 (range_def_chain::dump): New.
16188 (gori_map::gori_map): Allocate import list.
16189 (gori_map::~gori_map): Release imports.
16190 (gori_map::exports): Check for past allocated block size.
16191 (gori_map::imports): New.
16192 (gori_map::def_chain_in_export_p): Delete.
16193 (gori_map::is_import_p): New.
16194 (gori_map::maybe_add_gori): Handle imports.
16195 (gori_map::dump): Adjust output, add imports.
16196 (gori_compute::has_edge_range_p): Remove def_chain_in_export call.
16197 (gori_export_iterator::gori_export_iterator): New.
16198 (gori_export_iterator::next): New.
16199 (gori_export_iterator::get_name): New.
16200 * gimple-range-gori.h (range_def_chain): Add imports and direct
16201 dependecies via struct rdc.
16202 (range_def_chain::depend1): New.
16203 (range_def_chain::depend2): New.
16204 (class gori_map): Adjust.
16205 (FOR_EACH_GORI_IMPORT_NAME): New.
16206 (FOR_EACH_GORI_EXPORT_NAME): New.
16207 (class gori_export_iterator): New.
16209 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
16211 * gimple-range-cache.cc (ranger_cache::ranger_cache): Move initial
16212 export cache filling to here.
16213 * gimple-range-gori.cc (gori_compute::gori_compute) : From Here.
16215 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
16217 * gimple-range-gori.cc (range_def_chain): Move to gimple-range-gori.h.
16218 (gori_map): Move to gimple-range-gori.h.
16219 (gori_compute::gori_compute): Adjust.
16220 (gori_compute::~gori_compute): Delete.
16221 (gori_compute::compute_operand_range_switch): Adjust.
16222 (gori_compute::compute_operand_range): Adjust.
16223 (gori_compute::compute_logical_operands): Adjust.
16224 (gori_compute::has_edge_range_p ): Adjust.
16225 (gori_compute::set_range_invariant): Delete.
16226 (gori_compute::dump): Adjust.
16227 (gori_compute::outgoing_edge_range_p): Adjust.
16228 * gimple-range-gori.h (class range_def_chain): Relocate here.
16229 (class gori_map): Relocate here.
16230 (class gori_compute): Inherit from gori_map, and adjust.
16232 2021-05-25 Aldy Hernandez <aldyh@redhat.com>
16234 * value-range.cc (range_tests_legacy): Use
16235 build_nonstandard_integer_type instead of int and short.
16237 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
16239 * gimplify.c (gimplify_decl_expr): Clear TREE_READONLY on the DECL
16240 when really creating an initialization statement for it.
16242 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
16244 * tree-inline.c (setup_one_parameter): Fix thinko in new condition.
16246 2021-05-25 Kito Cheng <kito.cheng@sifive.com>
16248 * config/riscv/riscv.h (ASM_SPEC): Pass -mno-relax.
16250 2021-05-25 Martin Liska <mliska@suse.cz>
16252 PR tree-optimization/92860
16254 * optc-save-gen.awk: Remove exceptions.
16256 2021-05-25 Martin Liska <mliska@suse.cz>
16258 * asan.h (sanitize_coverage_p): New function.
16259 * doc/extend.texi: Document it.
16260 * fold-const.c (fold_range_test): Use sanitize_flags_p
16261 instead of flag_sanitize_coverage.
16262 (fold_truth_andor): Likewise.
16263 * sancov.c: Likewise.
16264 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
16265 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Handle
16266 -fsanitize-coverage when inlining.
16268 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
16270 * config/csky/csky-modes.def : Fix copyright.
16272 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
16274 * config/csky/csky-modes.def : Amend copyright.
16275 * config/csky/csky_insn_fpuv2.md : Likewise.
16276 * config/csky/csky_insn_fpuv3.md : Likewise.
16278 2021-05-25 Richard Biener <rguenther@suse.de>
16280 PR middle-end/100727
16281 * calls.c (initialize_argument_information): Explicitely test
16282 for WITH_SIZE_EXPR.
16283 * gimple-expr.c (mark_addressable): Skip outer WITH_SIZE_EXPR.
16285 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
16287 * config/csky/csky.h (FRAME_POINTER_REGNUM): Use
16288 HARD_FRAME_POINTER_REGNUM and FRAME_POINTER_REGNUM instead of
16289 the signle definition. The signle definition may not work well
16290 at simplify_subreg_regno().
16291 (HARD_FRAME_POINTER_REGNUM): New.
16292 (ELIMINABLE_REGS): Add for HARD_FRAME_POINTER_REGNUM.
16293 * config/csky/csky.c (get_csky_live_regs, csky_can_eliminate,
16294 csky_initial_elimination_offset, csky_expand_prologue,
16295 csky_expand_epilogue): Add for HARD_FRAME_POINTER_REGNUM.
16297 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
16299 * config/csky/csky.c (csky_option_override):
16300 Init csky_arch_isa_features[] in advance, so TARGET_DSP
16301 and TARGET_DIV can be set well.
16303 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
16305 * config/csky/constraints.md ("l", "h"): Delete.
16306 * config/csky/csky.h (reg_class, REG_CLASS_NAMES,
16307 REG_CLASS_CONTENTS): Delete LO_REGS and HI_REGS.
16308 * config/csky/csky.c (regno_reg_classm,
16309 csky_secondary_reload, csky_register_move_cost):
16310 Use HILO_REGS instead of LO_REGS and HI_REGS.
16312 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
16314 * config/csky/constraints.md ("W"): New constriant for mem operand
16315 with base reg, index register.
16316 ("Q"): Renamed and modified "csky_valid_fpuv2_mem_operand" to
16317 "csky_valid_mem_constraint_operand" to deal with both "Q" and "W"
16319 ("Dv"): New constraint for const double value that can be used at
16321 * config/csky/csky-modes.def (HFmode): New mode.
16322 * config/csky/csky-protos.h (csky_valid_fpuv2_mem_operand): Rename
16323 to "csky_valid_mem_constraint_operand" and support new constraint
16325 (csky_get_movedouble_length): New.
16326 (fpuv3_output_move): New.
16327 (fpuv3_const_double): New.
16328 * config/csky/csky.c (csky_option_override): New arch CK860 with fpv3.
16329 (decompose_csky_address): Refine.
16330 (csky_print_operand): New "CONST_DOUBLE" operand.
16331 (csky_output_move): Support fpv3 instructions.
16332 (csky_get_movedouble_length): New.
16333 (fpuv3_output_move): New.
16334 (fpuv3_const_double): New.
16335 (csky_emit_compare): Cover float comparsion.
16336 (csky_emit_compare_float): Refine.
16337 (csky_vaild_fpuv2_mem_operand): Rename to
16338 "csky_valid_mem_constraint_operand" and support new constraint "W".
16339 (ck860_rtx_costs): New.
16340 (csky_rtx_costs): Add the cost calculation of CK860.
16341 (regno_reg_class): New vregs for fpuv3.
16342 (csky_dbx_regno): Likewise.
16343 (csky_cpu_cpp_builtins): New builtin macro for fpuv3.
16344 (csky_conditional_register_usage): Suporrot fpuv3.
16345 (csky_dwarf_register_span): Suporrot fpuv3.
16346 (csky_init_builtins, csky_mangle_type): Support "__fp16" type.
16347 (ck810_legitimate_index_p): Support fp16.
16348 * config/csky/csky.h (TARGET_TLS): ADD CK860.
16349 (CSKY_VREG_P, CSKY_VREG_LO_P, CSKY_VREG_HI_P): Support fpuv3.
16350 (TARGET_SINGLE_FPU): Support fpuv3.
16351 (TARGET_SUPPORT_FPV3): New.
16352 (FIRST_PSEUDO_REGISTER): Change to 202 to hold the new fpuv3 registers.
16353 (FIXED_REGISTERS, CALL_REALLY_USED_REGISTERS, REGISTER_NAMES,
16354 REG_CLASS_CONTENTS): Support fpuv3.
16355 * config/csky/csky.md (movsf): Move to cksy_insn_fpu.md and refine.
16356 (csky_movsf_fpv2): Likewise.
16357 (ck801_movsf): Likewise.
16358 (csky_movsf): Likewise.
16360 (csky_movdf_fpv2): Likewise.
16361 (ck801_movdf): Likewise.
16362 (csky_movdf): Likewise.
16363 (movsicc): Refine. Use "comparison_operatior" instead of
16364 "ordered_comparison_operatior".
16365 (addsicc): Likewise.
16366 (CSKY_FIRST_VFP3_REGNUM, CSKY_LAST_VFP3_REGNUM): New constant.
16367 (call_value_internal_vh): New.
16368 * config/csky/csky_cores.def (CK860): New arch and cpu.
16373 * config/csky/csky_insn_fpu.md: Refactor. Separate all float patterns
16374 into emit-patterns and match-patterns, remain the emit-patterns here,
16375 and move the match-patterns to csky_insn_fpuv2.md or
16376 csky_insn_fpuv3.md.
16377 * config/csky/csky_insn_fpuv2.md: New file for fpuv2 instructions.
16378 * config/csky/csky_insn_fpuv3.md: New file and new patterns for fpuv3
16380 * config/csky/csky_isa.def (fcr): New.
16385 (CK860): New definition for ck860.
16386 * config/csky/csky_tables.opt (ck860): New processors ck860,
16387 ck860f. And new arch ck860.
16392 * config/csky/predicates.md (csky_float_comparsion_operator): Delete
16393 "geu", "gtu", "leu", "ltu", which will never appear at float comparison.
16394 * config/csky/t-csky-elf: Support 860.
16395 * config/csky/t-csky-linux: Likewise.
16396 * doc/md.texi: Add "Q" and "W" constraints for C-SKY.
16398 2021-05-24 Aaron Sawdey <acsawdey@linux.ibm.com>
16400 * config/rs6000/genfusion.pl (gen_logical_addsubf): Refactor to
16401 add generation of logical-add and add-logical fusion pairs.
16402 * config/rs6000/rs6000-cpus.def: Add new fusion to ISA 3.1 mask
16404 * config/rs6000/rs6000.c (rs6000_option_override_internal): Turn on
16405 logical-add and add-logical fusion by default.
16406 * config/rs6000/rs6000.opt: Add -mpower10-fusion-logical-add and
16407 -mpower10-fusion-add-logical options.
16408 * config/rs6000/fusion.md: Regenerate file.
16410 2021-05-24 Aldy Hernandez <aldyh@redhat.com>
16412 * value-range.cc (irange::legacy_equal_p): Check type when
16413 comparing VR_VARYING types.
16414 (range_tests_legacy): Test comparing VARYING ranges of different
16417 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
16419 * config/aarch64/aarch64.c (neoversen1_tunings):
16420 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
16422 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
16424 * config/aarch64/aarch64.c (aarch64_classify_symbol): Use GOT for
16425 extern weak symbols. Limit symbol offsets for non-GOT symbols with
16428 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
16430 * config/arm/neon.md (vec_load_lanesxi<mode>)
16431 (vec_store_lanexoi<mode>): Move ...
16432 * config/arm/vec-common.md: here.
16434 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
16436 * config/arm/neon.md (vec_load_lanesoi<mode>)
16437 (vec_store_lanesoi<mode>): Move ...
16438 * config/arm/vec-common.md: here.
16440 2021-05-24 liuhongt <hongtao.liu@intel.com>
16443 * config/i386/i386.c (ix86_gimple_fold_builtin): Replacing
16444 stmt with GIMPLE_NOP when lhs doesn't exist.
16446 2021-05-23 Uroš Bizjak <ubizjak@gmail.com>
16449 * config/i386/mmx.md (*push<VI_32:mode>2_rex64):
16450 New instruction pattern.
16451 (*push<VI_32:mode>2): Ditto.
16452 (push splitter for SSE registers): New splitter.
16454 2021-05-23 Andrew Pinski <apinski@marvell.com>
16456 * match.pd ((A & C) != 0 ? D : 0): Limit to non pointer types.
16458 2021-05-22 Aaron Sawdey <acsawdey@linux.ibm.com>
16460 * config/rs6000/genfusion.pl (gen_addadd): Fix incorrect attr types.
16461 * config/rs6000/fusion.md: Regenerate file.
16463 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
16465 * config/rs6000/genfusion.pl (gen_addadd): New function.
16466 * config/rs6000/fusion.md: Regenerate file.
16467 * config/rs6000/rs6000-cpus.def: Add
16468 OPTION_MASK_P10_FUSION_2ADD to masks.
16469 * config/rs6000/rs6000.c (rs6000_option_override_internal):
16470 Handle default value of OPTION_MASK_P10_FUSION_2ADD.
16471 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2add.
16473 2021-05-21 Jakub Jelinek <jakub@redhat.com>
16475 PR middle-end/99928
16476 * tree.h (OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET): Define.
16477 * gimplify.c (enum gimplify_omp_var_data): Fix up
16478 GOVD_MAP_HAS_ATTACHMENTS value, add GOVD_FIRSTPRIVATE_IMPLICIT.
16479 (omp_lastprivate_for_combined_outer_constructs): If combined target
16480 has GOVD_FIRSTPRIVATE_IMPLICIT set for the decl, change it to
16481 GOVD_MAP | GOVD_SEEN.
16482 (gimplify_scan_omp_clauses): Set GOVD_FIRSTPRIVATE_IMPLICIT for
16483 firstprivate clauses with OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT.
16484 (gimplify_adjust_omp_clauses): For firstprivate clauses with
16485 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT either clear that bit and
16486 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET too, or remove it and
16487 let it be replaced by implicit map clause.
16489 2021-05-21 Jakub Jelinek <jakub@redhat.com>
16491 PR middle-end/99928
16492 * gimplify.c (omp_lastprivate_for_combined_outer_constructs): New
16494 (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LASTPRIVATE>: Use it.
16495 (gimplify_omp_for): Likewise.
16497 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16499 PR middle-end/90115
16500 * omp-low.c (oacc_privatization_candidate_p): Reject 'static',
16501 'external' in blocks.
16503 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16505 PR middle-end/90115
16506 * flag-types.h (enum openacc_privatization): New.
16507 * params.opt (-param=openacc-privatization): New.
16508 * doc/invoke.texi (openacc-privatization): Document it.
16509 * omp-general.h (get_openacc_privatization_dump_flags): New
16511 * omp-low.c (oacc_privatization_candidate_p): Add diagnostics.
16512 * omp-offload.c (execute_oacc_device_lower)
16513 <IFN_UNIQUE_OACC_PRIVATE>: Re-work diagnostics.
16514 * target.def (goacc.adjust_private_decl): Add 'location_t'
16516 * doc/tm.texi: Regenerate.
16517 * config/gcn/gcn-protos.h (gcn_goacc_adjust_private_decl): Adjust.
16518 * config/gcn/gcn-tree.c (gcn_goacc_adjust_private_decl): Likewise.
16519 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl):
16520 Likewise. Preserve it for...
16521 (nvptx_goacc_expand_var_decl): ... use here.
16523 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16525 * doc/sourcebuild.texi (Other attributes): Document '__OPTIMIZE__'
16528 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16530 PR middle-end/90115
16531 * omp-low.c (oacc_privatization_candidate_p): New function.
16532 (oacc_privatization_scan_clause_chain)
16533 (oacc_privatization_scan_decl_chain): Use it. Also
16534 'gcc_checking_assert' that we're not seeing duplicates.
16536 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16538 PR middle-end/90115
16539 * omp-offload.c (execute_oacc_device_lower): Skip processing if no
16542 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16544 PR middle-end/90115
16545 * omp-offload.c (execute_oacc_device_lower): Explain.
16547 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16549 PR middle-end/90115
16550 * omp-offload.c (execute_oacc_device_lower)
16551 <IFN_UNIQUE_OACC_PRIVATE>: Diagnose and handle for 'level == -1'
16553 * internal-fn.c (expand_UNIQUE): Don't expect
16554 'IFN_UNIQUE_OACC_PRIVATE'.
16556 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16558 PR middle-end/90115
16559 * omp-low.c (lower_omp_for): Don't evaluate OpenMP 'for' clauses.
16561 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16563 PR middle-end/90115
16564 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl)
16565 (nvptx_goacc_expand_var_decl): Tighten.
16567 2021-05-21 Julian Brown <julian@codesourcery.com>
16568 Chung-Lin Tang <cltang@codesourcery.com>
16569 Thomas Schwinge <thomas@codesourcery.com>
16571 PR middle-end/90115
16572 * doc/tm.texi.in (TARGET_GOACC_EXPAND_VAR_DECL)
16573 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Add documentation hooks.
16574 * doc/tm.texi: Regenerate.
16575 * expr.c (expand_expr_real_1): Expand decls using the
16576 expand_var_decl OpenACC hook if defined.
16577 * internal-fn.c (expand_UNIQUE): Handle IFN_UNIQUE_OACC_PRIVATE.
16578 * internal-fn.h (IFN_UNIQUE_CODES): Add OACC_PRIVATE.
16579 * omp-low.c (omp_context): Add oacc_privatization_candidates
16581 (lower_oacc_reductions): Add PRIVATE_MARKER parameter. Insert
16583 (lower_oacc_head_tail): Add PRIVATE_MARKER parameter. Modify
16584 private marker's gimple call arguments, and pass it to
16585 lower_oacc_reductions.
16586 (oacc_privatization_scan_clause_chain)
16587 (oacc_privatization_scan_decl_chain, lower_oacc_private_marker):
16589 (lower_omp_for, lower_omp_target, lower_omp_1): Use these.
16590 * omp-offload.c (convert.h): Include.
16591 (oacc_loop_xform_head_tail): Treat private-variable markers like
16592 fork/join when transforming head/tail sequences.
16593 (struct var_decl_rewrite_info): Add struct.
16594 (oacc_rewrite_var_decl, is_sync_builtin_call): New functions.
16595 (execute_oacc_device_lower): Support rewriting gang-private
16596 variables using target hook, and fix up addr_expr and var_decl
16598 * target.def (adjust_private_decl, expand_var_decl): New hooks.
16599 * config/gcn/gcn-protos.h (gcn_goacc_adjust_gangprivate_decl):
16601 (gcn_goacc_adjust_private_decl): ...this.
16602 * config/gcn/gcn-tree.c (gcn_goacc_adjust_gangprivate_decl):
16604 (gcn_goacc_adjust_private_decl): ...this. Add LEVEL parameter.
16605 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_GANGPRIVATE_DECL): Rename
16606 definition using gcn_goacc_adjust_gangprivate_decl...
16607 (TARGET_GOACC_ADJUST_PRIVATE_DECL): ...to this, using
16608 gcn_goacc_adjust_private_decl.
16609 * config/nvptx/nvptx.c (tree-pretty-print.h): Include.
16610 (gang_private_shared_size): New global variable.
16611 (gang_private_shared_align): Likewise.
16612 (gang_private_shared_sym): Likewise.
16613 (gang_private_shared_hmap): Likewise.
16614 (nvptx_option_override): Initialize these.
16615 (nvptx_file_end): Output gang_private_shared_sym.
16616 (nvptx_goacc_adjust_private_decl, nvptx_goacc_expand_var_decl):
16618 (nvptx_set_current_function): Clear gang_private_shared_hmap.
16619 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Define hook.
16620 (TARGET_GOACC_EXPAND_VAR_DECL): Likewise.
16622 2021-05-21 H.J. Lu <hjl.tools@gmail.com>
16624 * config/i386/i386-modes.def (MAX_BITSIZE_MODE_ANY_INT): Removed.
16626 2021-05-21 Richard Biener <rguenther@suse.de>
16627 H.J. Lu <hjl.tools@gmail.com>
16629 PR middle-end/90773
16630 * expr.c (expand_constructor): Elide expand_constructor if
16631 move by pieces is preferred.
16633 2021-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16635 * config/aarch64/aarch64-builtins.c (aarch64_call_properties):
16636 Take a flag and mode value as arguments.
16637 (aarch64_modifies_global_state_p): Likewise.
16638 (aarch64_reads_global_state_p): Likewise.
16639 (aarch64_could_trap_p): Likewise.
16640 (aarch64_get_attributes): Likewise.
16641 (aarch64_init_simd_builtins): Adjust callsite of above.
16642 (aarch64_init_fcmla_laneq_builtins): Use aarch64_get_attributes to get
16643 function attributes to apply to builtins.
16644 (aarch64_init_crc32_builtins): Likewise.
16645 (aarch64_init_builtin_rsqrt): Likewise.
16647 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
16649 * config/rs6000/rs6000.md (define_attr "type"): Add types for fusion.
16650 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10): Use new fusion types.
16651 (gen_2logical): Use new fusion types.
16652 * config/rs6000/fusion.md: Regenerate.
16654 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
16657 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
16658 Handle V4QI and V2HI modes.
16659 (ix86_expand_sse_movcc): Ditto.
16660 * config/i386/mmx.md (*<sat_plusminus:insn><VI_32:mode>3):
16661 New instruction pattern.
16662 (*eq<VI_32:mode>3): Ditto.
16663 (*gt<VI_32:mode>3): Ditto.
16664 (*xop_pcmov_<VI_32:mode>): Ditto.
16665 (mmx_pblendvb32): Ditto.
16666 (mmx_pblendvb64): Rename from mmx_pblendvb.
16667 (vec_cmp<VI_32:mode><VI_32:mode>): New expander.
16668 (vec_cmpu<VI_32:mode><VI_32:mode>): Ditto.
16669 (vcond<VI_32:mode><VI_32:mode>): Ditto.
16670 (vcondu<VI_32:mode><VI_32:mode>): Ditto.
16671 (vcond_mask_<VI_32:mode><VI_32:mode>): Ditto.
16673 2021-05-21 Jakub Jelinek <jakub@redhat.com>
16675 PR tree-optimization/94589
16676 * tree-ssa-phiopt.c (spaceship_replacement): For integral rhs1 and
16677 rhs2, treat x <= 4 equivalently to x < 5 etc. In cmp1 and cmp2 (if
16678 not the same as cmp3) treat <= the same as < and >= the same as >.
16679 Don't require that cond2_phi_edge is true edge, instead take
16680 false/true edges into account based on cmp1/cmp2 comparison kinds.
16682 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
16685 * config/i386/mmx.md (SMAXMIN_MMXMODEI): New mode iterator.
16686 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Macroize expander
16687 from <smaxmin:code>v4hi3> and <smaxmin:code><MMXMODE14:mode>3
16688 using SMAXMIN_MMXMODEI mode iterator.
16689 (*<smaxmin:code>v4qi3): New insn pattern.
16690 (*<smaxmin:code>v2hi3): Ditto.
16691 (SMAXMIN_VI_32): New mode iterator.
16692 (<smaxmin:code><SMAXMIN_VI_32>mode3): New expander.
16693 (UMAXMIN_MMXMODEI): New mode iterator.
16694 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Macroize expander
16695 from <umaxmin:code>v8qi3> and <umaxmin:code><MMXMODE24:mode>3
16696 using UMAXMIN_MMXMODEI mode iterator.
16697 (*<umaxmin:code>v4qi3): New insn pattern.
16698 (*<umaxmin:code>v2hi3): Ditto.
16699 (UMAXMIN_VI_32): New mode iterator.
16700 (<umaxmin:code><UMAXMIN_VI_32>mode3): New expander.
16701 (abs<VI_32:mode>2): New insn pattern.
16702 (ssse3_abs<MMXMODEI:mode>2, abs<MMXMODEI:mode>2): Move from ...
16703 * config/i386/sse.md: ... here.
16705 2021-05-20 Clement Chigot <clement.chigot@atos.net>
16706 David Edelsohn <dje.gcc@gmail.com>
16708 * collect2.c (scan_prog_file): Issue non-fatal warning for
16711 2021-05-20 Jonathan Wakely <jwakely@redhat.com>
16713 * doc/invoke.texi (-Wno-c++11-extensions)
16714 (-Wno-c++14-extensions, -Wno-c++17-extensions)
16715 (-Wno-c++20-extensions, -Wno-c++23-extensions): Document
16718 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
16720 * config/c6x/c6x.c (c6x_output_file_unwind): Use dwarf_debuginfo_p.
16721 * config/darwin.c (darwin_override_options): Likewise.
16722 * config/i386/cygming.h (DBX_REGISTER_NUMBER): Likewise.
16723 * config/i386/darwin.h (DBX_REGISTER_NUMBER): Likewise.
16724 (DWARF2_FRAME_REG_OUT): Likewise.
16725 * config/mips/mips.c (mips_output_filename): Likewise.
16726 * config/rs6000/rs6000.c (rs6000_xcoff_declare_function_name):
16728 (rs6000_dbx_register_number): Likewise.
16729 * dbxout.c: Include flags.h.
16730 * dwarf2cfi.c (cfi_label_required_p): Likewise.
16731 (dwarf2out_do_frame): Likewise.
16732 * except.c: Include flags.h.
16733 * final.c (dwarf2_debug_info_emitted_p): Likewise.
16734 (final_scan_insn_1): Likewise.
16735 * flags.h (dwarf_debuginfo_p): New function declaration.
16736 * opts.c (dwarf_debuginfo_p): New function definition.
16737 * targhooks.c (default_debug_unwind_info): Use dwarf_debuginfo_p.
16738 * toplev.c (process_options): Likewise.
16740 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
16742 * common.opt: Change type to support bitmasks.
16743 * flag-types.h (enum debug_info_type): Rename enumerator constants.
16744 (NO_DEBUG): New bitmask.
16745 (DBX_DEBUG): Likewise.
16746 (DWARF2_DEBUG): Likewise.
16747 (XCOFF_DEBUG): Likewise.
16748 (VMS_DEBUG): Likewise.
16749 (VMS_AND_DWARF2_DEBUG): Likewise.
16750 * flags.h (debug_set_to_format): New function declaration.
16751 (debug_set_count): Likewise.
16752 (debug_set_names): Likewise.
16753 * opts.c (debug_type_masks): Array of bitmasks for debug formats.
16754 (debug_set_to_format): New function definition.
16755 (debug_set_count): Likewise.
16756 (debug_set_names): Likewise.
16757 (set_debug_level): Update access to debug_type_names.
16758 * toplev.c: Likewise.
16760 2021-05-20 Martin Sebor <msebor@redhat.com>
16762 PR middle-end/100684
16763 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Handle C++ lambda.
16765 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
16768 * config/i386/i386.md (isa): Remove x64_bmi.
16769 (enabled): Remove x64_bmi.
16770 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
16771 Remove general register alternative.
16772 (*andnot<VI_32:mode>3): Ditto.
16773 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
16774 (*<any_logic:code><VI_32:mode>3): Ditto.
16776 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
16778 * config/arm/arm.c: Include head files tree-vectorizer.h and
16781 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
16784 * config/i386/mmx.md (Yv_Yw): Revert adding V4QI and V2HI modes.
16785 (*<plusminus:insn><VI_32:mode>3): Use Yw instad of <Yv_Yw> constrint.
16786 (<s>mulv4hi3_highpart): New expander.
16787 (*<s>mulv2hi3_highpart): New insn pattern.
16788 (<s>mulv2hi3_higpart): New expander.
16789 (*<any_shift:insn>v2hi3): New insn pattern.
16790 (<any_shift:insn>v2hi3): New expander.
16791 * config/i386/sse.md (smulhrsv2hi3): New expander.
16792 (*smulhrsv2hi3): New insn pattern.
16794 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
16796 * doc/invoke.texi (vect-inner-loop-cost-factor): Document new
16798 * params.opt (vect-inner-loop-cost-factor): New.
16799 * targhooks.c (default_add_stmt_cost): Replace hardcoded factor
16800 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR, include head file
16801 tree-vectorizer.h and its required ones.
16802 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Replace
16803 hardcoded factor 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR.
16804 * config/arm/arm.c (arm_add_stmt_cost): Likewise.
16805 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
16806 * config/rs6000/rs6000.c (rs6000_add_stmt_cost): Likewise.
16807 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
16809 (_loop_vec_info::_loop_vec_info): Init inner_loop_cost_factor.
16810 * tree-vectorizer.h (_loop_vec_info): Add inner_loop_cost_factor.
16811 (LOOP_VINFO_INNER_LOOP_COST_FACTOR): New macro.
16813 2021-05-20 Christophe Lyon <christophe.lyon@linaro.org>
16814 Torbjörn Svensson <torbjorn.svensson@st.com>
16817 * doc/cpp.texi (Common Predefined Macros): Document __FILE_NAME__.
16819 2021-05-20 Jakub Jelinek <jakub@redhat.com>
16821 PR middle-end/99928
16822 * gimplify.c (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LINEAR>: For
16823 explicit linear clause when combined with target, make it map(tofrom:)
16824 instead of no clause or firstprivate.
16826 2021-05-20 Jakub Jelinek <jakub@redhat.com>
16828 PR tree-optimization/94589
16829 * match.pd ((X & Y) == X -> (X & ~Y) == 0): Simplify even in presence
16830 of integral conversions.
16832 2021-05-19 Andrew MacLeod <amacleod@redhat.com>
16834 * gimple-range.cc (fur_source::get_operand): New.
16835 (gimple_range_fold): Delete.
16836 (fold_using_range::fold_stmt): Move from gimple_ranger::calc_stmt.
16837 (fold_using_range::range_of_range_op): Move from gimple_ranger.
16838 (fold_using_range::range_of_address): Ditto.
16839 (fold_using_range::range_of_phi): Ditto.
16840 (fold_using_range::range_of_call): Ditto.
16841 (fold_using_range::range_of_builtin_ubsan_call): Move from
16842 range_of_builtin_ubsan_call.
16843 (fold_using_range::range_of_builtin_call): Move from
16844 range_of_builtin_call.
16845 (gimple_ranger::range_of_builtin_call): Delete.
16846 (fold_using_range::range_of_cond_expr): Move from gimple_ranger.
16847 (gimple_ranger::fold_range_internal): New.
16848 (gimple_ranger::range_of_stmt): Use new fold_using_range API.
16849 (fold_using_range::range_of_ssa_name_with_loop_info): Move from
16850 gimple_ranger. Improve ranges of SSA_NAMES when possible.
16851 * gimple-range.h (gimple_ranger): Remove various range_of routines.
16852 (class fur_source): New.
16853 (class fold_using_range): New.
16854 (fur_source::fur_source): New.
16856 * vr-values.c (vr_values::extract_range_basic): Use fold_using_range
16857 instead of range_of_builtin_call.
16859 2021-05-19 Jonathan Wakely <jwakely@redhat.com>
16861 * doc/cpp.texi (Common Predefined Macros): Update documentation
16862 for the __GXX_EXPERIMENTAL_CXX0X__ macro.
16864 2021-05-19 Alex Coplan <alex.coplan@arm.com>
16867 * config/arm/arm.md (nonsecure_call_internal): Always ensure
16868 callee's address is in a register.
16870 2021-05-19 Geng Qi <gengqi@linux.alibaba.com>
16872 * common/config/riscv/riscv-common.c
16873 (riscv_subset_list::parsing_subset_version): Properly parse the letter
16875 (riscv_subset_list::parse_std_ext,
16876 riscv_subset_list::parse_multiletter_ext): To handle errors generated
16877 in riscv_subset_list::parsing_subset_version.
16879 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
16881 * config/aarch64/aarch64-simd.md: Use "neon_move_narrow_q"
16882 type attribute in patterns generating XTN(2).
16884 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
16886 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_pack_trunc_<mode>):
16887 Remove as duplicate of...
16888 (aarch64_xtn<mode>): This.
16889 (aarch64_xtn2<mode>_le): Move position in file.
16890 (aarch64_xtn2<mode>_be): Move position in file.
16891 (aarch64_xtn2<mode>): Move position in file.
16892 (vec_pack_trunc_<mode>): Define as an expander.
16894 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
16896 * config/aarch64/aarch64-simd-builtins.def: Split builtin
16897 generation for aarch64_<sur>q<r>shr<u>n_n<mode> pattern into
16898 separate scalar and vector generators.
16899 * config/aarch64/aarch64-simd.md
16900 (aarch64_<sur>q<r>shr<u>n_n<mode>): Define as an expander and
16902 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_le): This and...
16903 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_be): This.
16904 * config/aarch64/iterators.md: Define SD_HSDI iterator.
16906 2021-05-19 Jonathn Wright <jonathan.wright@arm.com>
16908 * config/aarch64/aarch64-simd.md: Use UNSPEC_SQXTUN instead
16910 * config/aarch64/iterators.md: Remove UNSPEC_SQXTUN2.
16912 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
16914 * config/aarch64/aarch64-simd.md (aarch64_<sur>q<r>shr<u>n2_n<mode>):
16915 Implement as an expand emitting a big/little endian
16916 instruction pattern.
16917 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_le): Define.
16918 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_be): Define.
16920 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
16922 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn2<mode>):
16923 Implement as an expand emitting a big/little endian
16924 instruction pattern.
16925 (aarch64_<sur><addsub>hn2<mode>_insn_le): Define.
16926 (aarch64_<sur><addsub>hn2<mode>_insn_be): Define.
16927 * config/aarch64/iterators.md: Remove UNSPEC_[R]ADDHN2 and
16928 UNSPEC_[R]SUBHN2 unspecs and ADDSUBHN2 iterator.
16930 2021-05-19 Richard Biener <rguenther@suse.de>
16932 PR middle-end/100672
16933 * fold-const.c (fold_negate_expr_1): Use element_precision.
16934 (negate_expr_p): Likewise.
16936 2021-05-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
16938 * config/aarch64/iterators.md (SVE_PRED_LOAD): New iterator.
16939 (pred_load): New int attribute.
16940 * config/aarch64/aarch64-sve.md
16941 (aarch64_load_<ANY_EXTEND:optab><SVE_HSDI:mode><SVE_PARTIAL_I:mode>): Use
16942 SVE_PRED_LOAD enum iterator and corresponding pred_load attribute.
16943 * config/aarch64/aarch64-sve-builtins-base.cc (expand): Update call to
16944 code_for_aarch64_load.
16946 2021-05-19 Richard Biener <rguenther@suse.de>
16948 * cfgexpand.c (discover_nonconstant_array_refs_r): Make
16949 sure TARGET_MEM_REF bases are expanded as memory.
16950 * tree-ssa-operands.c (operands_scanner::get_tmr_operands):
16951 Do not mark TARGET_MEM_REF bases addressable.
16952 * tree-ssa.c (non_rewritable_mem_ref_base): Handle
16953 TARGET_MEM_REF bases as never rewritable.
16954 * gimple-walk.c (walk_stmt_load_store_addr_ops): Do not
16955 walk TARGET_MEM_REF bases as address-takens.
16956 * tree-ssa-dce.c (ref_may_be_aliased): Handle TARGET_MEM_REF.
16958 2021-05-19 Richard Biener <rguenther@suse.de>
16960 * builtins.c (get_object_alignment_1): Strip outer
16962 * tree-dfa.c (get_ref_base_and_extent): Handle outer
16963 WITH_SIZE_EXPR for size processing and process the
16965 * tree-ssa-alias.c (ao_ref_base_alias_set): Strip
16966 outer WITH_SIZE_EXPR.
16967 (ao_ref_base_alias_ptr_type): Likewise.
16968 (refs_may_alias_p_2): Allow WITH_SIZE_EXPR in ref->ref
16969 and handle that accordingly, stripping it for the
16970 core alias workers.
16971 * tree.c (get_base_address): Handle WITH_SIZE_EXPR by
16972 looking through it instead of returning NULL.
16974 2021-05-19 Jakub Jelinek <jakub@redhat.com>
16976 PR middle-end/100576
16977 * builtins.c (check_read_access): Convert bound to size_type_node if
16980 2021-05-19 Richard Biener <rguenther@suse.de>
16982 * tree-cfg.c (verify_types_in_gimple_min_lval): Inline...
16983 (verify_types_in_gimple_reference): ... here. Sanitize.
16984 (verify_gimple_call): Verify references in LHS and arguments.
16985 (verify_gimple_assign_single): Reject WITH_SIZE_EXPR.
16987 2021-05-19 Uroš Bizjak <ubizjak@gmail.com>
16989 * config/i386/i386.h (VALID_INT_MODE_P):
16990 Add V8QI, V4HI and V2SI modes for TARGET_64BIT.
16991 * config/i386/i386.md (isa): Add x64_bmi.
16992 (enabled): Handle x64_bmi.
16993 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
16994 Add alternative using 64bit general registers.
16995 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
16997 2021-05-19 Jakub Jelinek <jakub@redhat.com>
16999 PR middle-end/99928
17000 * tree.h (OMP_MASTER_COMBINED): Define.
17001 * gimplify.c (gimplify_scan_omp_clauses): Rewrite lastprivate
17002 handling for outer combined/composite constructs to a loop.
17003 Handle lastprivate on combined target.
17004 (gimplify_expr): Formatting fix.
17006 2021-05-19 Xionghu Luo <luoxhu@linux.ibm.com>
17008 * passes.def: Add sink_code pass before store_merging.
17009 * tree-ssa-sink.c (pass_sink_code:clone): New.
17011 2021-05-18 Bill Schmidt <wschmidt@linux.ibm.com>
17013 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Remove call to
17014 rs6000_special_adjust_field_align_p.
17015 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
17016 * config/rs6000/rs6000-call.c (rs6000_function_arg_boundary):
17017 Remove ABI warning.
17018 (rs6000_function_arg): Likewise.
17019 * config/rs6000/rs6000-protos.h
17020 (rs6000_special_adjust_field_align_p): Remove prototype.
17021 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
17023 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Remove call to
17024 rs6000_special_adjust_field_align_p.
17026 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
17029 * config/i386/i386.h (VALID_SSE2_REG_MODE):
17030 Add V4QI and V2HI modes.
17031 (VALID_INT_MODE_P): Ditto.
17032 * config/i386/mmx.md (VI_32): New mode iterator.
17033 (mmxvecsize): Handle V4QI and V2HI.
17035 (mov<VI_32:mode>): New expander.
17036 (*mov<mode>_internal): New insn pattern.
17037 (movmisalign<VI_32:mode>): New expander.
17038 (neg<VI_32:mode>): New expander.
17039 (<plusminus:insn><VI_32:mode>3): New expander.
17040 (*<plusminus:insn><VI_32:mode>3): New insn pattern.
17041 (mulv2hi3): New expander.
17042 (*mulv2hi3): New insn pattern.
17043 (one_cmpl<VI_32:mode>2): New expander.
17044 (*andnot<VI_32:mode>3): New insn pattern.
17045 (<any_logic:code><VI_32:mode>3): New expander.
17046 (*<any_logic:code><VI_32:mode>3): New insn pattern.
17048 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
17050 * config/i386/sse.md (<any_extend:insn>v4qiv4di2):
17051 Fix a mode mismatch with operand 1.
17053 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
17056 * config/i386/i386-expand.c (split_double_mode): Return
17057 temporary register when simplify_gen_subreg fails with
17058 the high half od the paradoxical subreg.
17060 2021-05-18 Richard Biener <rguenther@suse.de>
17062 * cfgexpand.c (expand_one_var): Pass in forced_stack_var
17063 and honor it when expanding.
17064 (expand_used_vars_for_block): Pass through forced_stack_var.
17065 (expand_used_vars): Likewise.
17066 (discover_nonconstant_array_refs_r): Set bits in
17067 forced_stack_vars instead of marking vars TREE_ADDRESSABLE.
17068 (avoid_type_punning_on_regs): Likewise.
17069 (discover_nonconstant_array_refs): Likewise.
17070 (pass_expand::execute): Create and pass down forced_stack_var
17071 bitmap. For parameters and returns temporarily set
17072 TREE_ADDRESSABLE when expand_function_start.
17074 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
17076 * doc/sourcebuild.texi: Document 'dg-note'.
17078 2021-05-18 Tobias Burnus <tobias@codesourcery.com>
17081 * configure: Regenerate.
17082 * configure.ac (BUILD_CFLAG, BUILD_CXXFLAGS): Add $(CFLAGS-$@).
17084 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
17086 * gimple.h (is_gimple_omp_oacc): Tighten.
17087 * omp-low.c (check_omp_nesting_restrictions): Adjust.
17089 2021-05-18 Richard Biener <rguenther@suse.de>
17091 * tree-ssa-operands.c (mark_address_taken): Simplify.
17093 2021-05-18 Martin Liska <mliska@suse.cz>
17095 * config/gcn/mkoffload.c (STR): Redefine.
17096 * config/i386/intelmic-mkoffload.c (STR): Likewise.
17097 * config/nvptx/mkoffload.c (STR): Likewise.
17099 2021-05-18 Martin Liska <mliska@suse.cz>
17101 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
17102 Use startswith function instead of strncmp.
17103 * common/config/bfin/bfin-common.c (bfin_handle_option): Likewise.
17104 * common/config/riscv/riscv-common.c (riscv_subset_list::parse): Likewise.
17105 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): Likewise.
17106 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Likewise.
17107 * config/alpha/alpha.c (alpha_elf_section_type_flags): Likewise.
17108 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
17109 * config/arm/arm.c (arm_file_start): Likewise.
17110 (arm_valid_target_attribute_rec): Likewise.
17111 (thumb1_md_asm_adjust): Likewise.
17112 * config/arm/driver-arm.c (host_detect_local_cpu): Likewise.
17113 * config/avr/avr.c (STR_PREFIX_P): Likewise.
17114 (avr_set_current_function): Likewise.
17115 (avr_handle_addr_attribute): Likewise.
17116 (avr_asm_output_aligned_decl_common): Likewise.
17117 (avr_asm_named_section): Likewise.
17118 (avr_section_type_flags): Likewise.
17119 (avr_asm_select_section): Likewise.
17120 * config/c6x/c6x.c (c6x_in_small_data_p): Likewise.
17121 (c6x_section_type_flags): Likewise.
17122 * config/darwin-c.c (darwin_cfstring_ref_p): Likewise.
17123 (darwin_objc_declare_unresolved_class_reference): Likewise.
17124 (darwin_objc_declare_class_definition): Likewise.
17125 * config/darwin.c (indirect_data): Likewise.
17126 (darwin_encode_section_info): Likewise.
17127 (darwin_objc2_section): Likewise.
17128 (darwin_objc1_section): Likewise.
17129 (machopic_select_section): Likewise.
17130 (darwin_globalize_label): Likewise.
17131 (darwin_label_is_anonymous_local_objc_name): Likewise.
17132 (darwin_asm_named_section): Likewise.
17133 (darwin_asm_output_dwarf_offset): Likewise.
17134 * config/frv/frv.c (frv_string_begins_with): Likewise.
17135 (frv_in_small_data_p): Likewise.
17136 * config/gcn/mkoffload.c (STR): Likewise.
17138 * config/i386/i386-builtins.c (get_builtin_code_for_version): Likewise.
17139 * config/i386/i386-options.c (ix86_option_override_internal): Likewise.
17140 * config/i386/i386.c (x86_64_elf_section_type_flags): Likewise.
17141 (ix86_md_asm_adjust): Likewise.
17142 * config/i386/intelmic-mkoffload.c (STR): Likewise.
17143 * config/i386/winnt.c (i386_pe_asm_named_section): Likewise.
17144 (i386_pe_file_end): Likewise.
17145 * config/ia64/ia64.c (ia64_in_small_data_p): Likewise.
17146 (ia64_section_type_flags): Likewise.
17147 * config/mips/driver-native.c (host_detect_local_cpu): Likewise.
17148 * config/mips/mips.c (mips_handle_interrupt_attr): Likewise.
17149 (mips16_stub_function_p): Likewise.
17150 (mips_function_rodata_section): Likewise.
17151 * config/msp430/msp430.c (msp430_mcu_name): Likewise.
17152 (msp430_function_section): Likewise.
17153 (msp430_section_type_flags): Likewise.
17154 (msp430_expand_helper): Likewise.
17155 * config/nios2/nios2.c (nios2_small_section_name_p): Likewise.
17156 (nios2_valid_target_attribute_rec): Likewise.
17157 * config/nvptx/mkoffload.c (process): Likewise.
17159 * config/pa/som.h: Likewise.
17160 * config/pdp11/pdp11.c (pdp11_output_ident): Likewise.
17161 * config/riscv/riscv.c (riscv_elf_select_rtx_section): Likewise.
17162 * config/rs6000/rs6000.c (VTABLE_NAME_P): Likewise.
17163 (rs6000_inner_target_options): Likewise.
17164 * config/s390/driver-native.c (s390_host_detect_local_cpu): Likewise.
17165 * config/sparc/driver-sparc.c (host_detect_local_cpu): Likewise.
17166 * config/vax/vax.c (vax_output_int_move): Likewise.
17167 * config/vms/vms-ld.c (startswith): Likewise.
17168 (process_args): Likewise.
17170 * config/vms/vms.c: Likewise.
17172 2021-05-18 Jakub Jelinek <jakub@redhat.com>
17174 PR rtl-optimization/100590
17175 * regcprop.c (copyprop_hardreg_forward_1): Only DCE dead sets if
17176 they are NONJUMP_INSN_P.
17178 2021-05-18 Jakub Jelinek <jakub@redhat.com>
17181 * function.c (push_dummy_function): Set DECL_ARTIFICIAL and
17182 DECL_ASSEMBLER_NAME on the fn_decl.
17184 2021-05-18 Jakub Jelinek <jakub@redhat.com>
17186 PR tree-optimization/94589
17187 * tree-ssa-phiopt.c (spaceship_replacement): Pattern match
17188 phi result used in (res & ~1) == 0 comparison as res >= 0 as
17189 res == 2 would be UB with -ffinite-math-only.
17191 2021-05-18 Martin Liska <mliska@suse.cz>
17193 * Makefile.in: genversion.o should depend on DATESTAMP.
17195 2021-05-18 Claudiu Zissulescu <claziss@synopsys.com>
17197 * config/arc/simdext.md (negv2si2): Remove round bracket.
17199 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
17201 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Define
17202 _Bool as macro expanding to _Bool.
17204 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
17207 * tree.c (build_reference_type_for_mode)
17208 (build_pointer_type_for_mode): Pick pointer mode if MODE argument
17210 (build_reference_type, build_pointer_type): Invoke
17211 build_*_type_for_mode with VOIDmode.
17213 2021-05-17 Andrew MacLeod <amacleod@redhat.com>
17215 PR tree-optimization/100512
17216 * gimple-range-cache.cc (ranger_cache::set_global_range): Mark const
17217 and non-zero pointer ranges as invariant.
17218 * gimple-range.cc (gimple_ranger::range_of_stmt): Remove pointer
17219 processing from here.
17221 2021-05-17 Tom de Vries <tdevries@suse.de>
17224 * config/nvptx/nvptx-protos.h (nvptx_output_atomic_insn): Declare
17225 * config/nvptx/nvptx.c (nvptx_output_barrier)
17226 (nvptx_output_atomic_insn): New function.
17227 (nvptx_print_operand): Add support for 'B'.
17228 * config/nvptx/nvptx.md: Use nvptx_output_atomic_insn for atomic
17231 2021-05-17 Aldy Hernandez <aldyh@redhat.com>
17233 PR tree-optimization/100349
17234 * vr-values.c (bounds_of_var_in_loop): Bail if scev returns
17237 2021-05-17 Tamar Christina <tamar.christina@arm.com>
17239 * config/aarch64/driver-aarch64.c (DEFAULT_ARCH): New.
17240 (host_detect_local_cpu): Use it.
17242 2021-05-17 Martin Liska <mliska@suse.cz>
17244 * doc/invoke.texi: Add 2 missing dots.
17246 2021-05-17 Marius Hillenbrand <mhillen@linux.ibm.com>
17248 PR bootstrap/100552
17249 * configure.ac: Replace pattern substitution with call to sed.
17250 * configure: Regenerate.
17252 2021-05-17 Richard Biener <rguenther@suse.de>
17254 PR middle-end/100582
17255 * tree.c (array_at_struct_end_p): Get to the base of the
17256 reference before looking for the underlying decl.
17258 2021-05-17 Joern Rennecke <joern.rennecke@embecosm.com>
17260 * genoutput.c (validate_insn_alternatives) Make "wrong number of
17261 alternatives" message more specific, and remove assumption on where
17264 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
17266 * config/arm/iterators.md (V16): New iterator.
17267 (VH_cvtto): New iterator.
17268 (v_cmp_result): Added V4HF and V8HF support.
17269 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result>): Use VDQWH.
17270 (vcond<mode><mode>): Likewise.
17271 (vcond_mask_<mode><v_cmp_result>): Likewise.
17272 (vcond<VH_cvtto><mode>): New expander.
17274 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
17276 * config/arm/arm-protos.h (arm_expand_vector_compare): Update
17278 * config/arm/arm.c (arm_expand_vector_compare): Add support for
17280 (arm_expand_vcond): Likewise.
17281 * config/arm/iterators.md (supf): Remove VCMPNEQ_S, VCMPEQQ_S,
17282 VCMPEQQ_N_S, VCMPNEQ_N_S.
17283 (VCMPNEQ, VCMPEQQ, VCMPEQQ_N, VCMPNEQ_N): Remove.
17284 * config/arm/mve.md (@mve_vcmp<mve_cmp_op>q_<mode>): Add '@' prefix.
17285 (@mve_vcmp<mve_cmp_op>q_f<mode>): Likewise.
17286 (@mve_vcmp<mve_cmp_op>q_n_f<mode>): Likewise.
17287 (@mve_vpselq_<supf><mode>): Likewise.
17288 (@mve_vpselq_f<mode>"): Likewise.
17289 * config/arm/neon.md (vec_cmp<mode><v_cmp_result): Enable for MVE
17290 and move to vec-common.md.
17291 (vec_cmpu<mode><mode>): Likewise.
17292 (vcond<mode><mode>): Likewise.
17293 (vcond<V_cvtto><mode>): Likewise.
17294 (vcondu<mode><v_cmp_result>): Likewise.
17295 (vcond_mask_<mode><v_cmp_result>): Likewise.
17296 * config/arm/unspecs.md (VCMPNEQ_U, VCMPNEQ_S, VCMPEQQ_S)
17297 (VCMPEQQ_N_S, VCMPNEQ_N_S, VCMPEQQ_U, CMPEQQ_N_U, VCMPNEQ_N_U)
17298 (VCMPGEQ_N_S, VCMPGEQ_S, VCMPGTQ_N_S, VCMPGTQ_S, VCMPLEQ_N_S)
17299 (VCMPLEQ_S, VCMPLTQ_N_S, VCMPLTQ_S, VCMPCSQ_N_U, VCMPCSQ_U)
17300 (VCMPHIQ_N_U, VCMPHIQ_U): Remove.
17301 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result): Moved
17303 (vec_cmpu<mode><mode>): Likewise.
17304 (vcond<mode><mode>): Likewise.
17305 (vcond<V_cvtto><mode>): Likewise.
17306 (vcondu<mode><v_cmp_result>): Likewise.
17307 (vcond_mask_<mode><v_cmp_result>): Likewise. Added unsafe math
17310 2021-05-17 liuhongt <hongtao.liu@intel.com>
17313 * config/i386/i386.c (ix86_gimple_fold_builtin): Use
17314 gsi_insert_seq_before instead.
17316 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
17318 * doc/sourcebuild.texi (arm_qbit_ok): Rename into...
17319 (arm_sat_ok): ...this.
17321 2021-05-17 Martin Liska <mliska@suse.cz>
17323 * lto-wrapper.c (merge_flto_options): Factor out a new function.
17324 (merge_and_complain): Use it.
17325 (run_gcc): Merge also linker command line -flto=foo argument
17328 2021-05-16 Christophe Lyon <christophe.lyon@linaro.org>
17330 * config/arm/arm.h (CPP_SPEC): Remove error message about
17331 -mlittle-endian/-mbig-endian conflict.
17333 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
17335 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
17336 __ROP_PROTECT__ if -mrop-protect is selected.
17338 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
17340 * config/rs6000/rs6000-internal.h (rs6000_stack): Add
17341 rop_hash_save_offset and rop_hash_size.
17342 * config/rs6000/rs6000-logue.c (rs6000_stack_info): Compute
17343 rop_hash_size and rop_hash_save_offset.
17344 (debug_stack_info): Dump rop_hash_save_offset and rop_hash_size.
17345 (rs6000_emit_prologue): Emit hashst[p] in prologue.
17346 (rs6000_emit_epilogue): Emit hashchk[p] in epilogue.
17347 * config/rs6000/rs6000.md (unspec): Add UNSPEC_HASHST and
17349 (hashst): New define_insn.
17350 (hashchk): Likewise.
17352 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
17354 * config/rs6000/rs6000.c (rs6000_option_override_internal):
17355 Disable shrink wrap when inserting ROP-protect instructions.
17356 * config/rs6000/rs6000.opt (mrop-protect): New option.
17357 (mprivileged): Likewise.
17358 * doc/invoke.texi: Document mrop-protect and mprivileged.
17360 2021-05-15 Hans-Peter Nilsson <hp@axis.com>
17362 * reorg.c (fill_slots_from_thread): Reinstate code typoed out in
17365 2021-05-15 Martin Jambor <mjambor@suse.cz>
17368 2021-05-13 Martin Jambor <mjambor@suse.cz>
17370 PR tree-optimization/100453
17371 * tree-sra.c (sra_modify_assign): All const base accesses do not
17372 need refreshing, not just those from decl_pool.
17373 (sra_modify_assign): Do not refresh into a const base decl.
17375 2021-05-15 Jakub Jelinek <jakub@redhat.com>
17377 PR rtl-optimization/100342
17378 * regcprop.c (copy_value): When copying a source reg in a wider
17379 mode than it has recorded for the value, adjust recorded destination
17380 mode too or punt if !REG_CAN_CHANGE_MODE_P.
17382 2021-05-14 Jason Merrill <jason@redhat.com>
17384 * intl.h: Add comments.
17386 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17388 * config/aarch64/aarch64-simd.md
17389 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Split into...
17390 (aarch64_sqdmlsl2_lane<mode>_internal): ... This...
17391 (aarch64_sqdmlal2_lane<mode>_internal): ... And this.
17392 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Split into ...
17393 (aarch64_sqdmlsl2_laneq<mode>_internal): ... This...
17394 (aarch64_sqdmlal2_laneq<mode>_internal): ... And this.
17395 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>_internal): Split into...
17396 (aarch64_sqdmlsl2_n<mode>_internal): ... This...
17397 (aarch64_sqdmlal2_n<mode>_internal): ... And this.
17399 2021-05-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17402 * config/arm/arm_neon.h (vtst_s8): Replace call to vtst builtin with it's
17403 boolean logic equivalent.
17404 (vtst_s16): Likewise.
17405 (vtst_s32): Likewise.
17406 (vtst_u8): Likewise.
17407 (vtst_u16): Likewise.
17408 (vtst_u32): Likewise.
17409 (vtst_p8): Likewise.
17410 (vtst_p16): Likewise.
17411 (vtstq_s8): Likewise.
17412 (vtstq_s16): Likewise.
17413 (vtstq_s32): Likewise.
17414 (vtstq_u8): Likewise.
17415 (vtstq_u16): Likewise.
17416 (vtstq_u32): Likewise.
17417 (vtstq_p8): Likewise.
17418 (vtstq_p16): Likewise.
17419 * config/arm/arm_neon_builtins.def: Remove entry for vtst.
17420 * config/arm/neon.md (neon_vtst<mode>): Remove pattern.
17422 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17424 * config/aarch64/aarch64-simd.md (aarch64_sqdmlal2<mode>): Merge into...
17425 (aarch64_sqdml<SBINQOPS:as>l2<mode>): ... This.
17426 (aarch64_sqdmlsl2<mode>): Delete.
17427 (aarch64_sqdmlal2_lane<mode>): Merge this...
17428 (aarch64_sqdmlsl2_lane<mode>): ... And this...
17429 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>): ... Into this.
17430 (aarch64_sqdmlal2_laneq<mode>): Merge this...
17431 (aarch64_sqdmlsl2_laneq<mode>): ... And this...
17432 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>): ... Into this.
17433 (aarch64_sqdmlal2_n<mode>): Merge this...
17434 (aarch64_sqdmlsl2_n<mode>): ... And this...
17435 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>): ... Into this.
17437 2021-05-13 Martin Sebor <msebor@redhat.com>
17439 PR middle-end/100574
17440 * builtins.c (access_ref::get_ref): Improve detection of PHIs with
17441 all null arguments.
17443 2021-05-13 Martin Sebor <msebor@redhat.com>
17445 PR tree-optimization/93100
17446 PR middle-end/98583
17447 * tree-ssa-uninit.c (check_defs): Exclude intrinsic functions that
17448 don't modify referenced objects.
17450 2021-05-13 Martin Jambor <mjambor@suse.cz>
17452 PR tree-optimization/100453
17453 * tree-sra.c (sra_modify_assign): All const base accesses do not
17454 need refreshing, not just those from decl_pool.
17455 (sra_modify_assign): Do not refresh into a const base decl.
17457 2021-05-13 Martin Liska <mliska@suse.cz>
17459 * tree-ssa-dom.c: Remove m_simplifier.
17461 2021-05-13 Richard Earnshaw <rearnsha@arm.com>
17464 * config/arm/arm.c (arm_canonicalize_comparison): Correctly
17465 canonicalize DImode inequality comparisons against the
17466 maximum integral value.
17468 2021-05-13 Jakub Jelinek <jakub@redhat.com>
17470 PR tree-optimization/98856
17471 * config/i386/i386.c (ix86_shift_rotate_cost): Add CODE argument.
17472 Expect V2DI and V4DI arithmetic right shifts to be emulated.
17473 (ix86_rtx_costs, ix86_add_stmt_cost): Adjust ix86_shift_rotate_cost
17475 * config/i386/i386-expand.c (expand_vec_perm_2perm_interleave,
17476 expand_vec_perm_2perm_pblendv): New functions.
17477 (ix86_expand_vec_perm_const_1): Use them.
17478 * config/i386/sse.md (ashr<mode>3<mask_name>): Rename to ...
17479 (<mask_codefor>ashr<mode>3<mask_name>): ... this.
17480 (ashr<mode>3): New define_expand with VI248_AVX512BW iterator.
17481 (ashrv4di3): New define_expand.
17482 (ashrv2di3): Change condition to TARGET_SSE2, handle !TARGET_XOP
17483 and !TARGET_AVX512VL expansion.
17485 2021-05-13 Uroš Bizjak <ubizjak@gmail.com>
17488 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Force mode
17489 sizes < 16 to a register when constructing vpcmov pattern.
17490 * config/i386/mmx.md (*xop_pcmov_<mode>): Use MMXMODE124 mode.
17492 2021-05-13 Martin Liska <mliska@suse.cz>
17494 * gcov-io.c (gcov_write_block): Remove.
17495 (gcov_write_words): Likewise.
17496 (gcov_read_words): Re-implement using gcov_read_bytes.
17497 (gcov_allocate): Remove.
17498 (GCOV_BLOCK_SIZE): Likewise.
17499 (struct gcov_var): Remove most of the fields.
17500 (gcov_position): Implement with ftell.
17501 (gcov_rewrite): Remove setting of start and offset fields.
17502 (from_file): Re-format.
17503 (gcov_open): Remove setbuf call. It should not be needed.
17504 (gcov_close): Remove internal buffer handling.
17505 (gcov_magic): Use __builtin_bswap32.
17506 (gcov_write_counter): Use directly gcov_write_unsigned.
17507 (gcov_write_string): Use direct fwrite and do not round
17509 (gcov_seek): Use directly fseek.
17510 (gcov_write_tag): Use gcov_write_unsigned directly.
17511 (gcov_write_length): Likewise.
17512 (gcov_write_tag_length): Likewise.
17513 (gcov_read_bytes): Use directly fread.
17514 (gcov_read_unsigned): Use gcov_read_words.
17515 (gcov_read_counter): Likewise.
17516 (gcov_read_string): Use gcov_read_bytes.
17517 * gcov-io.h (GCOV_WORD_SIZE): Adjust to reflect
17518 that size is not in bytes, but words (4B).
17519 (GCOV_TAG_FUNCTION_LENGTH): Likewise.
17520 (GCOV_TAG_ARCS_LENGTH): Likewise.
17521 (GCOV_TAG_ARCS_NUM): Likewise.
17522 (GCOV_TAG_COUNTER_LENGTH): Likewise.
17523 (GCOV_TAG_COUNTER_NUM): Likewise.
17524 (GCOV_TAG_SUMMARY_LENGTH): Likewise.
17526 2021-05-13 liuhongt <hongtao.liu@intel.com>
17529 * config/i386/sse.md (ssedoublevecmode): Add attribute for
17530 V64QI/V32HI/V16SI/V4DI.
17531 (ssehalfvecmode): Add attribute for V2DI/V2DF.
17532 (*vec_concatv4si_0): Extend to VI124_128.
17533 (*vec_concat<mode>_0): New pre-reload splitter.
17534 * config/i386/predicates.md (movq_parallel): New predicate.
17536 2021-05-13 Alexandre Oliva <oliva@adacore.com>
17538 * targhooks.c (default_zero_call_used_regs): Retry using
17539 successfully-zeroed registers as sources.
17541 2021-05-12 Tobias Burnus <tobias@codesourcery.com>
17543 * omp-low.c (finish_taskreg_scan): Use the proper detach decl.
17545 2021-05-12 Aldy Hernandez <aldyh@redhat.com>
17548 * gimple-range.cc (range_of_builtin_call): Skip out on
17549 processing __builtin_clz when varying.
17551 2021-05-12 Tom de Vries <tdevries@suse.de>
17554 * config/nvptx/nvptx-opts.h (enum ptx_version): New enum.
17555 * config/nvptx/nvptx.c (nvptx_file_start): Print .version according
17556 to ptx_version_option.
17557 * config/nvptx/nvptx.h (TARGET_PTX_6_3): Define.
17558 * config/nvptx/nvptx.md (define_insn "nvptx_shuffle<mode>")
17559 (define_insn "nvptx_vote_ballot"): Use sync variant for
17561 * config/nvptx/nvptx.opt (ptx_version): Add enum.
17562 (mptx): Add option.
17563 * doc/invoke.texi (Nvidia PTX Options): Add mptx item.
17565 2021-05-12 Richard Biener <rguenther@suse.de>
17567 PR tree-optimization/100566
17568 * tree-ssa-sccvn.c (dominated_by_p_w_unex): Properly handle
17569 allow_back for all edge queries.
17571 2021-05-12 liuhongt <hongtao.liu@intel.com>
17574 * config/i386/sse.md (<sse4_1_avx2>_pblendvb): Add
17575 splitters for pblendvb of NOT mask register.
17577 2021-05-12 Richard Biener <rguenther@suse.de>
17579 PR tree-optimization/100519
17580 * tree-ssa-reassoc.c (can_associate_p): Split into...
17581 (can_associate_op_p): ... this
17582 (can_associate_type_p): ... and this.
17583 (is_reassociable_op): Call can_associate_op_p.
17584 (break_up_subtract_bb): Call the appropriate predicates.
17585 (reassociate_bb): Likewise.
17587 2021-05-12 Martin Liska <mliska@suse.cz>
17589 * lto-wrapper.c (merge_and_complain): Merge -flto=arg options.
17590 (run_gcc): Use -flto argument detection for merged
17593 2021-05-12 Martin Liska <mliska@suse.cz>
17595 * lto-wrapper.c (print_lto_docs_link): New function.
17596 (run_gcc): Print warning about missing job server detection
17597 after we know NR of partitions. Do the same for -flto{,=1}.
17598 * opts.c (get_option_html_page): Support -flto option.
17600 2021-05-12 Martin Liska <mliska@suse.cz>
17602 * lto-wrapper.c (get_options_from_collect_gcc_options): Change
17604 (append_option): Remove.
17605 (find_option): Rework to use the vector type.
17606 (remove_option): Remove.
17607 (merge_and_complain): Use vectors for cl_decoded_option data
17609 (append_compiler_options): Likewise.
17610 (append_diag_options): Likewise.
17611 (append_linker_options): Likewise.
17612 (append_offload_options): Likewise.
17613 (compile_offload_image): Likewise.
17614 (compile_images_for_offload_targets): Likewise.
17615 (find_and_merge_options): Likewise.
17616 (run_gcc): Likewise.
17618 2021-05-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
17621 * dwarf2out.c (dwarf2out_finish): Set
17622 have_multiple_function_sections with multi-range text_section.
17624 2021-05-12 Martin Liska <mliska@suse.cz>
17626 PR bootstrap/100560
17627 * Makefile.in: Remove version.h from linker command line.
17629 2021-05-12 Richard Biener <rguenther@suse.de>
17631 PR middle-end/100547
17632 * rtl.h (rtvec_alloc): Make argument size_t.
17633 * rtl.c (rtvec_alloc): Verify the count is less than INT_MAX.
17635 2021-05-12 Jakub Jelinek <jakub@redhat.com>
17637 PR middle-end/100508
17638 * cfgexpand.c (expand_debug_expr): For DEBUG_EXPR_DECL with vector
17639 type, don't reuse DECL_RTL if it has different mode, instead force
17640 creation of a new DEBUG_EXPR.
17642 2021-05-12 Jakub Jelinek <jakub@redhat.com>
17643 Marc Glisse <marc.glisse@inria.fr>
17645 PR tree-optimization/94589
17646 * match.pd ((X & Y) == X -> (X & ~Y) == 0,
17647 (X | Y) == Y -> (X & ~Y) == 0): New GIMPLE simplifications.
17649 2021-05-12 Uroš Bizjak <ubizjak@gmail.com>
17652 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Handle V2SF mode.
17653 * config/i386/mmx.md (MMXMODE124): New mode iterator.
17655 (mmxintvecmode): New mode attribute.
17656 (mmxintvecmodelower): Ditto.
17657 (*mmx_maskcmpv2sf3_comm): New insn pattern.
17658 (*mmx_maskcmpv2sf3): Ditto.
17659 (vec_cmpv2sfv2si): New expander.
17660 (vcond<V2FI:mode>v2si): Ditto.
17661 (mmx_vlendvps): New insn pattern.
17662 (vcond<MMXMODE124:mode><MMXMODEI:mode>): Also handle V2SFmode.
17663 (vcondu<MMXMODE124:mode><MMXMODEI:mode>): Ditto.
17664 (vcond_mask_<mode><mmxintvecmodelower>): Ditto.
17666 2021-05-11 Martin Sebor <msebor@redhat.com>
17668 PR middle-end/21433
17669 * expr.c (expand_expr_real_1): Replace unreachable code with an assert.
17671 2021-05-11 Richard Biener <rguenther@suse.de>
17673 * gimple-fold.c (gimple_fold_call): Do not call
17674 maybe_fold_reference on call arguments or the static chain.
17675 (fold_stmt_1): Do not call maybe_fold_reference on GIMPLE_ASM
17678 2021-05-11 Martin Liska <mliska@suse.cz>
17680 * builtins.def (DEF_HSAIL_BUILTIN): Remove.
17681 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
17682 (DEF_HSAIL_SAT_BUILTIN): Likewise.
17683 (DEF_HSAIL_INTR_BUILTIN): Likewise.
17684 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
17685 * doc/frontends.texi: Remove BRIG.
17686 * doc/install.texi: Likewise.
17687 * doc/invoke.texi: Likewise.
17688 * doc/standards.texi: Likewise.
17689 * brig-builtins.def: Removed.
17690 * brig/ChangeLog: Removed.
17691 * brig/Make-lang.in: Removed.
17692 * brig/brig-builtins.h: Removed.
17693 * brig/brig-c.h: Removed.
17694 * brig/brig-lang.c: Removed.
17695 * brig/brigfrontend/brig-arg-block-handler.cc: Removed.
17696 * brig/brigfrontend/brig-atomic-inst-handler.cc: Removed.
17697 * brig/brigfrontend/brig-basic-inst-handler.cc: Removed.
17698 * brig/brigfrontend/brig-branch-inst-handler.cc: Removed.
17699 * brig/brigfrontend/brig-cmp-inst-handler.cc: Removed.
17700 * brig/brigfrontend/brig-code-entry-handler.cc: Removed.
17701 * brig/brigfrontend/brig-code-entry-handler.h: Removed.
17702 * brig/brigfrontend/brig-comment-handler.cc: Removed.
17703 * brig/brigfrontend/brig-control-handler.cc: Removed.
17704 * brig/brigfrontend/brig-copy-move-inst-handler.cc: Removed.
17705 * brig/brigfrontend/brig-cvt-inst-handler.cc: Removed.
17706 * brig/brigfrontend/brig-fbarrier-handler.cc: Removed.
17707 * brig/brigfrontend/brig-function-handler.cc: Removed.
17708 * brig/brigfrontend/brig-function.cc: Removed.
17709 * brig/brigfrontend/brig-function.h: Removed.
17710 * brig/brigfrontend/brig-inst-mod-handler.cc: Removed.
17711 * brig/brigfrontend/brig-label-handler.cc: Removed.
17712 * brig/brigfrontend/brig-lane-inst-handler.cc: Removed.
17713 * brig/brigfrontend/brig-machine.c: Removed.
17714 * brig/brigfrontend/brig-machine.h: Removed.
17715 * brig/brigfrontend/brig-mem-inst-handler.cc: Removed.
17716 * brig/brigfrontend/brig-module-handler.cc: Removed.
17717 * brig/brigfrontend/brig-queue-inst-handler.cc: Removed.
17718 * brig/brigfrontend/brig-seg-inst-handler.cc: Removed.
17719 * brig/brigfrontend/brig-signal-inst-handler.cc: Removed.
17720 * brig/brigfrontend/brig-to-generic.cc: Removed.
17721 * brig/brigfrontend/brig-to-generic.h: Removed.
17722 * brig/brigfrontend/brig-util.cc: Removed.
17723 * brig/brigfrontend/brig-util.h: Removed.
17724 * brig/brigfrontend/brig-variable-handler.cc: Removed.
17725 * brig/brigfrontend/hsa-brig-format.h: Removed.
17726 * brig/brigfrontend/phsa.h: Removed.
17727 * brig/brigspec.c: Removed.
17728 * brig/config-lang.in: Removed.
17729 * brig/gccbrig.texi: Removed.
17730 * brig/lang-specs.h: Removed.
17731 * brig/lang.opt: Removed.
17733 2021-05-11 Richard Biener <rguenther@suse.de>
17736 * ipa-param-manipulation.c
17737 (ipa_param_body_adjustments::modify_call_stmt): Avoid
17738 altering SSA_NAME_DEF_STMT by adjusting the calls LHS
17739 via gimple_call_lhs_ptr.
17741 2021-05-11 Alex Coplan <alex.coplan@arm.com>
17744 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear):
17745 Avoid emitting CFA adjusts on the sp if we have the fp.
17747 2021-05-11 Richard Sandiford <richard.sandiford@arm.com>
17749 * config/aarch64/iterators.md (VMUL_CHANGE_NLANES): Delete.
17750 (VMULD): New iterator.
17751 (VCOND): Handle V4HF and V8HF.
17752 (VCONQ): Fix entry for V2SF.
17753 * config/aarch64/aarch64-simd.md (mul_lane<mode>3): Use VMULD
17754 instead of VMUL. Use a 64-bit vector mode for the indexed operand.
17755 (*aarch64_mul3_elt_<vswap_width_name><mode>): Merge with...
17756 (mul_laneq<mode>3): ...this define_insn. Use VMUL instead of VDQSF.
17757 Use a 128-bit vector mode for the indexed operand. Use stype for
17758 the scheduling type.
17760 2021-05-11 Richard Biener <rguenther@suse.de>
17762 * gimple-fold.c (maybe_fold_reference): Only return
17763 is_gimple_min_invariant values.
17765 2021-05-11 Richard Biener <rguenther@suse.de>
17767 PR middle-end/100509
17768 * gimple-fold.c (fold_gimple_assign): Only call
17769 get_symbol_constant_value on register type symbols.
17771 2021-05-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
17772 Joe Ramsay <joe.ramsay@arm.com>
17775 * config/arm/arm_mve.h (__arm_vstrwq_scatter_offset): Fix wrong arguments.
17776 (__arm_vcmpneq): Remove duplicate definition.
17777 (__arm_vstrwq_scatter_offset_p): Likewise.
17778 (__arm_vmaxq_x): Likewise.
17779 (__arm_vmlsdavaq): Likewise.
17780 (__arm_vmlsdavaxq): Likewise.
17781 (__arm_vmlsdavq_p): Likewise.
17782 (__arm_vmlsdavxq_p): Likewise.
17783 (__arm_vrmlaldavhaq): Likewise.
17784 (__arm_vstrbq_p): Likewise.
17785 (__arm_vstrbq_scatter_offset): Likewise.
17786 (__arm_vstrbq_scatter_offset_p): Likewise.
17787 (__arm_vstrdq_scatter_offset): Likewise.
17788 (__arm_vstrdq_scatter_offset_p): Likewise.
17789 (__arm_vstrdq_scatter_shifted_offset): Likewise.
17790 (__arm_vstrdq_scatter_shifted_offset_p): Likewise.
17792 2021-05-11 Jakub Jelinek <jakub@redhat.com>
17794 PR middle-end/100471
17795 * omp-low.c (lower_omp_task_reductions): For OMP_TASKLOOP, if data
17796 is 0, bypass the reduction loop including
17797 GOMP_taskgroup_reduction_unregister call.
17799 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
17801 * config/rs6000/rs6000.c (struct rs6000_cost_data): New member
17802 costing_for_scalar.
17803 (rs6000_density_test): Early return if costing_for_scalar is true.
17804 (rs6000_init_cost): Init costing_for_scalar of rs6000_cost_data.
17806 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
17808 * doc/tm.texi: Regenerated.
17809 * target.def (init_cost): Add new parameter costing_for_scalar.
17810 * targhooks.c (default_init_cost): Adjust for new parameter.
17811 * targhooks.h (default_init_cost): Likewise.
17812 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
17813 (vect_compute_single_scalar_iteration_cost): Likewise.
17814 (vect_analyze_loop_2): Likewise.
17815 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
17816 (vect_bb_vectorization_profitable_p): Likewise.
17817 * tree-vectorizer.h (init_cost): Likewise.
17818 * config/aarch64/aarch64.c (aarch64_init_cost): Likewise.
17819 * config/i386/i386.c (ix86_init_cost): Likewise.
17820 * config/rs6000/rs6000.c (rs6000_init_cost): Likewise.
17822 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
17824 * config/rs6000/rs6000.c (rs6000_vect_nonmem): Renamed to
17825 vect_nonmem and moved into...
17826 (struct rs6000_cost_data): ...here.
17827 (rs6000_init_cost): Use vect_nonmem of cost_data instead.
17828 (rs6000_add_stmt_cost): Likewise.
17829 (rs6000_finish_cost): Likewise.
17831 2021-05-10 Eric Botcazou <ebotcazou@adacore.com>
17833 * range-op.cc (get_bool_state): Adjust head comment.
17834 (operator_not_equal::op1_range): Fix comment.
17835 (operator_bitwise_xor::op1_range): Remove call to gcc_unreachable.
17837 2021-05-10 Martin Sebor <msebor@redhat.com>
17839 PR middle-end/100425
17840 PR middle-end/100510
17841 * gimple-ssa-warn-alloca.c (pass_walloca::firast_time_p): Rename...
17842 (pass_walloca::xlimit_certain_p): ...to this.
17843 (pass_walloca::gate): Execute for any kind of handled warning.
17844 (pass_walloca::execute): Avoid issuing "maybe" and "unbounded"
17845 warnings when xlimit_certain_p is set.
17847 2021-05-10 Pat Haugen <pthaugen@linux.ibm.com>
17849 * config/rs6000/rs6000.c (rs6000_ira_change_pseudo_allocno_class):
17850 Return ALTIVEC_REGS if that is best_class.
17851 (rs6000_compute_pressure_classes): Add ALTIVEC_REGS.
17853 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17855 * config/arm/arm.h (CPP_SPEC): Remove error message about
17858 2021-05-10 Martin Jambor <mjambor@suse.cz>
17860 * ipa-prop.h (IPA_NODE_REF): Removed.
17861 (IPA_NODE_REF_GET_CREATE): Likewise.
17862 (IPA_EDGE_REF): Likewise.
17863 (IPA_EDGE_REF_GET_CREATE): Likewise.
17864 (IS_VALID_JUMP_FUNC_INDEX): Likewise.
17865 * ipa-cp.c (print_all_lattices): Replaced IPA_NODE_REF with a direct
17866 use of ipa_node_params_sum.
17867 (ipcp_versionable_function_p): Likewise.
17868 (push_node_to_stack): Likewise.
17869 (pop_node_from_stack): Likewise.
17870 (set_single_call_flag): Replaced two IPA_NODE_REF with one single
17871 direct use of ipa_node_params_sum.
17872 (initialize_node_lattices): Replaced IPA_NODE_REF with a direct use of
17873 ipa_node_params_sum.
17874 (ipa_context_from_jfunc): Replaced IPA_EDGE_REF with a direct use of
17876 (ipcp_verify_propagated_values): Replaced IPA_NODE_REF with a direct
17877 use of ipa_node_params_sum.
17878 (self_recursively_generated_p): Likewise.
17879 (propagate_scalar_across_jump_function): Likewise.
17880 (propagate_context_across_jump_function): Replaced IPA_EDGE_REF with a
17881 direct use of ipa_edge_args_sum, moved the lookup after the early
17882 exit. Replaced IPA_NODE_REF with a direct use of ipa_node_params_sum.
17883 (propagate_bits_across_jump_function): Replaced IPA_NODE_REF with
17884 direct uses of ipa_node_params_sum.
17885 (propagate_vr_across_jump_function): Likewise.
17886 (propagate_aggregate_lattice): Likewise.
17887 (propagate_aggs_across_jump_function): Likewise.
17888 (propagate_constants_across_call): Likewise, also replaced
17889 IPA_EDGE_REF with a direct use of ipa_edge_args_sum.
17890 (good_cloning_opportunity_p): Replaced IPA_NODE_REF with a direct use
17891 of ipa_node_params_sum.
17892 (estimate_local_effects): Likewise.
17893 (add_all_node_vals_to_toposort): Likewise.
17894 (propagate_constants_topo): Likewise.
17895 (ipcp_propagate_stage): Likewise.
17896 (ipcp_discover_new_direct_edges): Likewise.
17897 (calls_same_node_or_its_all_contexts_clone_p): Likewise.
17898 (cgraph_edge_brings_value_p): Likewise (in both overloaded functions).
17899 (get_info_about_necessary_edges): Likewise.
17900 (want_remove_some_param_p): Likewise.
17901 (create_specialized_node): Likewise.
17902 (self_recursive_pass_through_p): Likewise.
17903 (self_recursive_agg_pass_through_p): Likewise.
17904 (find_more_scalar_values_for_callers_subset): Likewise and also
17905 replaced IPA_EDGE_REF with direct uses of ipa_edge_args_sum, in one
17906 case replacing two of those with a single query.
17907 (find_more_contexts_for_caller_subset): Likewise for the
17908 ipa_polymorphic_call_context overload.
17909 (intersect_aggregates_with_edge): Replaced IPA_EDGE_REF with a direct
17910 use of ipa_edge_args_sum. Replaced IPA_NODE_REF with direct uses of
17911 ipa_node_params_sum.
17912 (find_aggregate_values_for_callers_subset): Likewise, also reusing
17913 results of ipa_edge_args_sum->get.
17914 (cgraph_edge_brings_all_scalars_for_node): Replaced IPA_NODE_REF with
17915 direct uses of ipa_node_params_sum, replaced IPA_EDGE_REF with a
17916 direct use of ipa_edge_args_sum.
17917 (cgraph_edge_brings_all_agg_vals_for_node): Likewise, moved node
17918 summary query after the early exit and reused the result later.
17919 (decide_about_value): Replaced IPA_NODE_REF with a direct use of
17920 ipa_node_params_sum.
17921 (decide_whether_version_node): Likewise. Removed re-querying for
17922 summaries after cloning.
17923 (spread_undeadness): Replaced IPA_NODE_REF with a direct use of
17924 ipa_node_params_sum.
17925 (has_undead_caller_from_outside_scc_p): Likewise, reusing results of
17927 (identify_dead_nodes): Likewise.
17928 (ipcp_store_bits_results): Replaced IPA_NODE_REF with direct uses of
17929 ipa_node_params_sum.
17930 (ipcp_store_vr_results): Likewise.
17931 * ipa-fnsummary.c (evaluate_properties_for_edge): Likewise.
17932 (ipa_fn_summary_t::duplicate): Likewise.
17933 (analyze_function_body): Likewise.
17934 (estimate_calls_size_and_time): Likewise.
17935 (ipa_cached_call_context::duplicate_from): Likewise.
17936 (ipa_call_context::equal_to): Likewise.
17937 (remap_edge_params): Likewise.
17938 (ipa_merge_fn_summary_after_inlining): Likewise.
17939 (inline_read_section): Likewise.
17940 * ipa-icf.c (sem_function::param_used_p): Likewise.
17941 * ipa-modref.c (compute_parm_map): Likewise.
17942 (compute_parm_map): Replaced IPA_EDGE_REF with a direct use of
17944 (get_access_for_fnspec): Replaced IPA_NODE_REF with a direct use of
17945 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
17947 * ipa-profile.c (check_argument_count): Likewise.
17948 * ipa-prop.c (ipa_alloc_node_params): Replaced IPA_NODE_REF_GET_CREATE
17949 with a direct use of ipa_node_params_sum.
17950 (ipa_initialize_node_params): Likewise.
17951 (ipa_print_node_jump_functions_for_edge): Replaced IPA_EDGE_REF with a
17952 direct use of ipa_edge_args_sum and reused the query result.
17953 (ipa_compute_jump_functions_for_edge): Replaced IPA_NODE_REF with a
17954 direct use of ipa_node_params_sum and replaced IPA_EDGE_REF with a
17955 direct use of ipa_edge_args_sum.
17956 (ipa_note_param_call): Replaced IPA_NODE_REF with a direct use of
17957 ipa_node_params_sum and reused the result of the query.
17958 (ipa_analyze_node): Likewise.
17959 (ipa_analyze_controlled_uses): Replaced IPA_NODE_REF with a direct use
17960 of ipa_node_params_sum.
17961 (update_jump_functions_after_inlining): Replaced IPA_EDGE_REF with
17962 direct uses of ipa_edge_args_sum.
17963 (update_indirect_edges_after_inlining): Replaced IPA_NODE_REF with
17964 direct uses of ipa_node_params_sum and replaced IPA_EDGE_REF with a
17965 direct use of ipa_edge_args_sum. Removed superficial re-querying the
17967 (propagate_controlled_uses): Replaced IPA_NODE_REF with direct uses of
17968 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
17970 (ipa_propagate_indirect_call_infos): Replaced IPA_EDGE_REF with a
17971 direct use of ipa_edge_args_sum.
17972 (ipa_edge_args_sum_t::duplicate): Replaced IPA_NODE_REF with a direct
17973 use of ipa_node_params_sum.
17974 (ipa_print_node_params): Likewise.
17975 (ipa_write_node_info): Likewise and also replaced IPA_EDGE_REF with
17976 direct uses of ipa_edge_args_sum.
17977 (ipa_read_edge_info): Replaced IPA_EDGE_REF with a direct use of
17979 (ipa_read_node_info): Replaced IPA_NODE_REF with a direct use of
17980 ipa_node_params_sum.
17981 (ipa_prop_write_jump_functions): Likewise. Move variable node to the
17982 scopes where it is used.
17984 2021-05-10 Uroš Bizjak <ubizjak@gmail.com>
17986 * config/i386/i386-expand.c (ix86_expand_sse_movcc)
17987 <case E_V2SImode>: Force op_true to register.
17989 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17991 * config/arm/iterators.md (MVE_FP_COMPARISONS): New.
17992 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_f<mode>)
17993 (mve_vcmp<mve_cmp_op>q_n_f<mode>): New, merge all vcmp_*f*
17995 (mve_vcmpeqq_f<mode>, mve_vcmpeqq_n_f<mode>, mve_vcmpgeq_f<mode>)
17996 (mve_vcmpgeq_n_f<mode>, mve_vcmpgtq_f<mode>)
17997 (mve_vcmpgtq_n_f<mode>, mve_vcmpleq_f<mode>)
17998 (mve_vcmpleq_n_f<mode>, mve_vcmpltq_f<mode>)
17999 (mve_vcmpltq_n_f<mode>, mve_vcmpneq_f<mode>)
18000 (mve_vcmpneq_n_f<mode>): Remove.
18001 * config/arm/unspecs.md (VCMPEQQ_F, VCMPEQQ_N_F, VCMPGEQ_F)
18002 (VCMPGEQ_N_F, VCMPGTQ_F, VCMPGTQ_N_F, VCMPLEQ_F, VCMPLEQ_N_F)
18003 (VCMPLTQ_F, VCMPLTQ_N_F, VCMPNEQ_F, VCMPNEQ_N_F): Remove.
18005 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
18007 * config/arm/iterators.md (MVE_COMPARISONS): New.
18009 (mve_cmp_type): New.
18010 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_<mode>): New, merge all
18012 (mve_vcmpneq_<mode>, mve_vcmpcsq_n_<mode>, mve_vcmpcsq_<mode>)
18013 (mve_vcmpeqq_n_<mode>, mve_vcmpeqq_<mode>, mve_vcmpgeq_n_<mode>)
18014 (mve_vcmpgeq_<mode>, mve_vcmpgtq_n_<mode>, mve_vcmpgtq_<mode>)
18015 (mve_vcmphiq_n_<mode>, mve_vcmphiq_<mode>, mve_vcmpleq_n_<mode>)
18016 (mve_vcmpleq_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
18017 (mve_vcmpneq_n_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
18018 (mve_vcmpneq_n_<mode>): Remove.
18020 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
18022 * config/arm/arm_mve.h (__arm_vcmp*): Remove 's' suffix.
18023 * config/arm/arm_mve_builtins.def (vcmp*): Remove 's' suffix.
18024 * config/arm/mve.md (mve_vcmp*): Remove 's' suffix in pattern
18027 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
18029 * config/arm/arm_mve_builtins.def (vcmpneq_u): Remove.
18030 (vcmpneq_n_u): Likewise.
18031 (vcmpeqq_u,): Likewise.
18032 (vcmpeqq_n_u): Likewise.
18033 * config/arm/iterators.md (supf): Remove VCMPNEQ_U, VCMPEQQ_U,
18034 VCMPEQQ_N_U and VCMPNEQ_N_U.
18035 * config/arm/mve.md (mve_vcmpneq): Remove <supf> iteration.
18036 (mve_vcmpeqq_n): Likewise.
18037 (mve_vcmpeqq): Likewise.
18038 (mve_vcmpneq_n): Likewise.
18040 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
18042 * config/arm/arm_mve.h (__arm_vcmpeq*u*, __arm_vcmpne*u*): Call
18043 the 's' version of the builtin.
18045 2021-05-10 Richard Biener <rguenther@suse.de>
18047 PR tree-optimization/100492
18048 * tree-loop-distribution.c (find_seed_stmts_for_distribution):
18049 Find nothing when the loop contains an irreducible region.
18051 2021-05-10 Richard Biener <rguenther@suse.de>
18053 PR middle-end/100464
18055 * gimple-fold.c (canonicalize_constructor_val): Do not set
18058 2021-05-10 Richard Biener <rguenther@suse.de>
18060 PR tree-optimization/100434
18061 * tree-ssa-dse.c (initialize_ao_ref_for_dse): Handle
18063 (dse_optimize_stmt): Handle call LHS by dropping the
18064 LHS or the whole call if it doesn't have other
18066 (pass_dse::execute): Adjust.
18068 2021-05-10 Martin Liska <mliska@suse.cz>
18070 * Makefile.in: Add missing genversion rule.
18072 2021-05-10 Alex Coplan <alex.coplan@arm.com>
18075 * config/arm/mve.md (*mve_mov<mode>): Simplify output code. Use
18076 vldrw.u32 and vstrw.32 for V2D[IF]mode loads and stores.
18078 2021-05-10 Martin Liska <mliska@suse.cz>
18080 * builtins.c (is_builtin_name): Use startswith
18081 function instead of strncmp.
18082 * collect2.c (main): Likewise.
18083 (has_lto_section): Likewise.
18084 (scan_libraries): Likewise.
18085 * coverage.c (coverage_checksum_string): Likewise.
18086 (coverage_init): Likewise.
18087 * dwarf2out.c (is_cxx): Likewise.
18088 (gen_compile_unit_die): Likewise.
18089 * gcc-ar.c (main): Likewise.
18090 * gcc.c (init_spec): Likewise.
18091 (read_specs): Likewise.
18092 (execute): Likewise.
18093 (check_live_switch): Likewise.
18094 * genattrtab.c (write_attr_case): Likewise.
18095 (IS_ATTR_GROUP): Likewise.
18096 * gencfn-macros.c (main): Likewise.
18097 * gengtype.c (type_for_name): Likewise.
18098 (gen_rtx_next): Likewise.
18099 (get_file_langdir): Likewise.
18100 (write_local): Likewise.
18101 * genmatch.c (get_operator): Likewise.
18102 (get_operand_type): Likewise.
18103 (expr::gen_transform): Likewise.
18104 * genoutput.c (validate_optab_operands): Likewise.
18105 * incpath.c (add_sysroot_to_chain): Likewise.
18106 * langhooks.c (lang_GNU_C): Likewise.
18107 (lang_GNU_CXX): Likewise.
18108 (lang_GNU_Fortran): Likewise.
18109 (lang_GNU_OBJC): Likewise.
18110 * lto-wrapper.c (run_gcc): Likewise.
18111 * omp-general.c (omp_max_simt_vf): Likewise.
18112 * omp-low.c (omp_runtime_api_call): Likewise.
18113 * opts-common.c (parse_options_from_collect_gcc_options): Likewise.
18114 * read-rtl-function.c (function_reader::read_rtx_operand_r): Likewise.
18115 * real.c (real_from_string): Likewise.
18116 * selftest.c (assert_str_startswith): Likewise.
18117 * timevar.c (timer::validate_phases): Likewise.
18118 * tree.c (get_file_function_name): Likewise.
18119 * ubsan.c (ubsan_use_new_style_p): Likewise.
18120 * varasm.c (default_function_rodata_section): Likewise.
18121 (incorporeal_function_p): Likewise.
18122 (default_section_type_flags): Likewise.
18123 * system.h (startswith): Define startswith.
18125 2021-05-10 Martin Liska <mliska@suse.cz>
18127 * bitmap.h (class auto_bitmap): Remove
18128 __cplusplus >= 201103.
18129 * config/aarch64/aarch64.c: Likewise.
18130 * gimple-ssa-store-merging.c (store_immediate_info::store_immediate_info):
18132 * sbitmap.h: Likewise.
18134 2021-05-10 Martin Liska <mliska@suse.cz>
18136 * Makefile.in: Rename gcov-iov to genversion and depend
18137 on version.h (instead of gcov-iov.h).
18138 * gcov-io.h: Include version.h instread of gcov-iov.h.
18139 * gengtype-state.c (read_state_version): Likewise.
18140 * gcov-iov.c: Moved to...
18141 * genversion.c: ...here.
18142 * lto-streamer.h (LTO_major_version): Define it with
18144 * version.c: Removed.
18145 * version.h: Removed.
18147 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
18149 * config/arc/arc.md (UNSPEC_ARC_DMPYWH): Define.
18150 * config/arc/simdext.md (VCT): Add predicates for iterator
18153 (voptab): Likewise.
18154 (vec_widen_<V_US>mult_hi_v4hi): Change pattern predicate.
18155 (<voptab>v2si3): New patterns.
18157 (reduc_plus_scal_v4hi): Likewise.
18158 (reduc_plus_scal_v2si): Likewise.
18159 (vec_duplicatev2si): Likewise.
18160 (vec_duplicatev4hi): Likewise.
18162 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
18164 * config/arc/simdext.md: Format and cleanup file.
18166 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
18168 * config/arc/simdext.md (movmisalignv2hi): Allow misaligned access
18169 only when munaligned-access option is on.
18170 (movmisalign<mode>): Likewise.
18172 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
18174 * common/config/arc/arc-common.c (arc_handle_option): Remove dot
18176 * config/arc/arc.c (arc_reorg): Remove underscore from string.
18178 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
18180 * config/arc/arc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
18181 (CTZ_DEFINED_VALUE_AT_ZERO): Likewise.
18182 * config/arc/arc.md (clrsbsi2): Cleanup pattern.
18183 (norm_f): Likewise.
18186 (clzsi2): Use fls instruction when available.
18187 (arc_clzsi2): Likewise.
18189 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
18191 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add r26 and r27.
18193 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
18195 * doc/extend.texi (__builtin_arc_sr): Swap arguments.
18197 2021-05-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
18199 PR middle-end/100467
18200 * toplev.c (compile_file): Call insn_locations_init before
18201 targetm.asm_out.code_end.
18203 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
18206 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
18208 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
18210 2021-05-07 Jakub Jelinek <jakub@redhat.com>
18211 Andrew Stubbs <amd@codesourcery.com>
18214 * builtins.c (try_store_by_multiple_pieces): Use force_operand for
18215 emit_move_insn operands.
18217 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
18219 * cfgexpand.c (expand_gimple_basic_block): Do not inherit a current
18220 location for the outgoing edges of an empty block.
18221 * dwarf2out.c (add_subscript_info): Retrieve the bounds and index
18222 type by means of the get_array_descr_info langhook, if it is set and
18223 returns true. Remove obsolete code dealing with unnamed subtypes.
18225 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
18227 * gimple-range-cache.cc (ssa_block_ranges): Virtualize.
18228 (sbr_vector): Renamed from ssa_block_cache.
18229 (sbr_vector::sbr_vector): Allocate from obstack abd initialize.
18230 (ssa_block_ranges::~ssa_block_ranges): Remove.
18231 (sbr_vector::set_bb_range): Use varying and undefined cached values.
18232 (ssa_block_ranges::set_bb_varying): Remove.
18233 (sbr_vector::get_bb_range): Adjust assert.
18234 (sbr_vector::bb_range_p): Adjust assert.
18235 (~block_range_cache): No freeing loop required.
18236 (block_range_cache::get_block_ranges): Remove.
18237 (block_range_cache::set_bb_range): Inline get_block_ranges.
18238 (block_range_cache::set_bb_varying): Remove.
18239 * gimple-range-cache.h (set_bb_varying): Remove prototype.
18240 * value-range.h (irange_allocator::get_memory): New.
18242 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
18244 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Search
18245 dominator tree is available and requested.
18246 (ranger_cache::ssa_range_in_bb): Don't search dom tree here.
18247 (ranger_cache::fill_block_cache): Don't search dom tree here either.
18248 * gimple-range-cache.h (non_null_deref_p): Add dom_search param.
18250 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
18252 * gimple-range.cc (gimple_ranger::range_on_exit): Handle block with
18253 only PHI nodes better.
18255 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
18257 * gimple-range-edge.h (gimple_outgoing_range): Rename from
18259 (gcond_edge_range): Export prototype.
18260 * gimple-range-edge.cc (gcond_edge_range): New.
18261 (gimple_outgoing_range::edge_range_p): Use gcond_edge_range.
18262 * gimple-range-gori.h (gori_compute): Use gimple_outgoing_range.
18264 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
18266 * gimple-range-edge.cc (outgoing_range::calc_switch_ranges): Compute
18267 default range into a temp and allocate only what is needed.
18269 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
18271 * range-op.cc (operator_trunc_mod::wi_fold): x % 0 is UNDEFINED.
18273 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
18275 * gimple-range.h (gimple_range_global): Pick up parameter initial
18276 values, and use-before defined locals are UNDEFINED.
18278 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
18280 * doc/extend.texi (scalar_storage_order): Mention effect on pointer
18282 * tree.h (reverse_storage_order_for_component_p): Return false if
18283 the type is a pointer.
18285 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
18287 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
18289 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
18292 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
18293 Handle V8QI, V4HI and V2SI modes.
18294 * config/i386/mmx.md (mmx_pblendvb): New insn pattern.
18295 * config/i386/sse.md (unspec): Move UNSPEC_BLENDV ...
18296 * config/i386/i386.md (unspec): ... here.
18298 2021-05-07 Tobias Burnus <tobias@codesourcery.com>
18299 Tom de Vries <tdevries@suse.de>
18301 * omp-low.c (lower_rec_simd_input_clauses): Set max_vf = 1 if
18302 a truth_value_p reduction variable is nonintegral.
18304 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
18307 * config/i386/i386-expand.c (ix86_use_mask_cmp_p):
18308 Return false for mode sizes < 16.
18310 2021-05-07 Jakub Jelinek <jakub@redhat.com>
18313 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn.
18315 2021-05-06 Martin Jambor <mjambor@suse.cz>
18317 * ipa-sra.c (ipa_sra_dump_all_summaries): Dump edge summaries even
18318 when there is no function summary.
18319 (ipa_sra_summarize_function): produce edge summaries even when
18322 2021-05-06 Tom Tromey <tom@tromey.com>
18324 * godump.c (string_hash_eq): Remove.
18325 (go_finish): Use htab_eq_string.
18327 2021-05-06 Tom Tromey <tom@tromey.com>
18329 * gengtype-state.c (read_state): Use htab_eq_string.
18330 (string_eq): Remove.
18332 2021-05-06 Tom Tromey <tom@tromey.com>
18334 * gensupport.c (htab_eq_string): Remove.
18336 2021-05-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
18339 * debug.h (gcc_debug_hooks): Add set_ignored_loc function pointer.
18340 * dwarf2out.h (dw_fde_node::ignored_debug): New data item.
18341 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Add dummy
18342 set_ignored_loc callbacks.
18343 * debug.c (do_nothing_debug_hooks): Likewise.
18344 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
18345 * dwarf2out.c (text_section_used, cold_text_section_used): Remove.
18346 (in_text_section_p, last_text_label, last_cold_label,
18347 switch_text_ranges, switch_cold_ranges): New data items.
18348 (dwarf2out_note_section_used): Remove.
18349 (dwarf2out_begin_prologue): Set fde->ignored_debug and
18351 (mark_ignored_debug_section): New helper function.
18352 (dwarf2out_end_epilogue, dwarf2out_switch_text_section): Call
18353 mark_ignored_debug_section.
18354 (dwarf2_debug_hooks): Use dwarf2out_set_ignored_loc.
18355 (dwarf2_lineno_debug_hooks): Use dummy for set_ignored_loc.
18356 (size_of_aranges): Adjust formula for multi-part text ranges size.
18357 (output_aranges): Output multi-part text ranges.
18358 (dwarf2out_set_ignored_loc): New callback function.
18359 (dwarf2out_finish): Output multi-part text ranges.
18360 (dwarf2out_c_finalize): Clear new data items.
18361 * final.c (final_start_function_1): Call set_ignored_loc callback.
18362 (final_scan_insn_1): Likewise.
18363 * ggc-page.c (gt_ggc_mx): New helper function.
18364 * stringpool.c (gt_pch_nx): Likewise.
18366 2021-05-06 Richard Biener <rguenther@suse.de>
18368 * timevar.def (TV_TREE_INSERT_PHI_NODES): Remove.
18369 (TV_TREE_SSA_REWRITE_BLOCKS): Likewise.
18370 (TV_TREE_INTO_SSA): New.
18371 * tree-into-ssa.c (insert_phi_nodes): Do not account separately.
18372 (rewrite_blocks): Likewise.
18373 (pass_data_build_ssa): Account to TV_TREE_INTO_SSA.
18375 2021-05-06 Jakub Jelinek <jakub@redhat.com>
18377 * tree-ssa-phiopt.c (value_replacement, minmax_replacement,
18378 abs_replacement, xor_replacement,
18379 cond_removal_in_popcount_clz_ctz_pattern,
18380 replace_phi_edge_with_variable): Change type of phi argument from
18381 gimple * to gphi *.
18383 2021-05-06 Richard Biener <rguenther@suse.de>
18385 * tree-ssa-loop-split.c (split_loop): Delay updating SSA form.
18386 Output an opt-info message.
18387 (do_split_loop_on_cond): Likewise.
18388 (tree_ssa_split_loops): Update SSA form here.
18390 2021-05-06 Richard Biener <rguenther@suse.de>
18392 * tree-inline.c (tree_function_versioning): Fix DECL_BY_REFERENCE
18393 return variable removal.
18395 2021-05-06 Marius Hillenbrand <mhillen@linux.ibm.com>
18397 * config/s390/s390-builtins.def (O_M5, O1_M5, ...): Remove unused macros.
18398 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
18399 (s390_vec_permi_dbl, s390_vpdi): Use the O3_U2 type for the immediate
18401 * config/s390/s390.c (s390_const_operand_ok): Remove unused
18404 2021-05-06 Jakub Jelinek <jakub@redhat.com>
18406 PR tree-optimization/94589
18407 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Call
18408 spaceship_replacement.
18409 (cond_only_block_p, spaceship_replacement): New functions.
18411 2021-05-06 Richard Biener <rguenther@suse.de>
18414 * tree-emutls.c (gen_emutls_addr): Pass in whether we're
18415 dealing with a debug use and only query existing addresses
18417 (lower_emutls_1): Avoid splitting out addresses for debug
18418 stmts, reset the debug stmt when we fail to find existing
18420 (lower_emutls_phi_arg): Set wi.stmt.
18422 2021-05-06 Christoph Muellner <cmuellner@gcc.gnu.org>
18425 * config/riscv/riscv.c (riscv_block_move_loop): Use cbranch helper.
18426 * config/riscv/riscv.md (cbranch<mode>4): Generate helpers.
18427 (stack_protect_test): Use cbranch helper.
18429 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
18432 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
18433 always return the establisher frame for __builtin_frame_address (0).
18435 2021-05-05 Ivan Sorokin <vanyacpp@gmail.com>
18438 * config/i386/i386-builtins.c (ix86_cpu_model_type_node): New.
18439 (ix86_cpu_model_var): Likewise.
18440 (ix86_cpu_features2_type_node): Likewise.
18441 (ix86_cpu_features2_var): Likewise.
18442 (fold_builtin_cpu): Cache __cpu_model and __cpu_features2 with
18445 2021-05-05 Martin Sebor <msebor@redhat.com>
18447 * passes.def (pass_warn_printf): Run after SSA.
18449 2021-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18451 * config/arm/neon.md (neon_vtst_combine<mode>): New pattern.
18452 * config/arm/predicates.md (minus_one_operand): New predicate.
18454 2021-05-05 Jeff Law <jlaw@tachyum.com>
18456 * config/avr/avr.md: Remove references to CC_STATUS_INIT.
18458 2021-05-05 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
18460 PR rtl-optimization/100263
18461 * postreload.c (move2add_valid_value_p): Ensure register can
18464 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
18466 PR rtl-optimization/100411
18467 * cfgcleanup.c (try_crossjump_to_edge): Also skip end of prologue
18468 and beginning of function markers.
18470 2021-05-05 Jeff Law <jlaw@tachyum.com>
18472 * config/cr16/cr16.h (NOTICE_UPDATE_CC): Remove.
18473 * config/cr16/cr16.c (notice_update_cc): Remove.
18474 * config/cr16/cr16-protos.h (notice_update_cc): Remove.
18476 2021-05-05 Uroš Bizjak <ubizjak@gmail.com>
18479 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
18480 Handle V8QI, V4HI and V2SI modes.
18481 * config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
18482 (ix86_build_signbit_mask): Ditto.
18483 * config/i386/mmx.md (MMXMODE14): New mode iterator.
18484 (<smaxmin:code><MMXMODE14:mode>3): New expander.
18485 (*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
18486 (<umaxmin:code><MMXMODE24:mode>3): New expander.
18487 (*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
18488 (vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
18489 (vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
18490 (vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
18491 (vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
18492 (vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
18494 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
18496 * dwarf2out.c (loc_list_from_tree_1) <DECL>: During early DWARF, do
18497 not expand the VALUE_EXPR of variables put in the non-local frame.
18498 * gimplify.c (gimplify_type_sizes) <RECORD_TYPE>: If the type is not
18499 to be ignored for debug info, ensure its variable offsets are not.
18501 2021-05-05 Richard Biener <rguenther@suse.de>
18503 PR tree-optimization/79333
18504 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
18505 Fold stmt following SSA edges.
18507 2021-05-05 Richard Biener <rguenther@suse.de>
18509 PR middle-end/100394
18510 * calls.c (expand_call): Preserve possibly throwing calls.
18511 * cfgexpand.c (expand_call_stmt): When a call can throw signal
18512 RTL expansion there are side-effects.
18513 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Simplify,
18514 mark all possibly throwing stmts necessary unless we can elide
18516 * tree-ssa-dse.c (pass_dse::execute): Preserve exceptions unless
18517 -fdelete-dead-exceptions.
18518 * tree.h (DECL_PURE_P): Add note about exceptions.
18520 2021-05-05 Alexandre Oliva <oliva@adacore.com>
18522 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Make it
18525 2021-05-04 David Edelsohn <dje.gcc@gmail.com>
18527 * config/rs6000/rs6000-call.c (rs6000_output_mi_thunk): Use
18528 get_fnname_from_decl for name of thunk.
18529 * config/rs6000/rs6000.c (rs6000_declare_alias): Use assemble_name
18530 and ASM_OUTPUT_LABEL.
18531 (rs6000_xcoff_declare_function_name): Use assemble_name and
18533 (rs6000_xcoff_declare_object_name): Use ASM_OUTPUT_LABEL.
18534 (rs6000_xcoff_encode_section_info): Don't add mapping class
18535 for aliases. Always add [DS] mapping class to primary
18537 (rs6000_asm_weaken_decl): Don't explicitly add [DS].
18539 2021-05-04 Martin Sebor <msebor@redhat.com>
18541 PR middle-end/100307
18542 * builtins.c (compute_objsize_r): Clear base0 for pointers.
18544 2021-05-04 Jeff Law <jlaw@tachyum.com>
18546 * config/bfin/bfin.h (NOTICE_UPDATE_CC): Remove.
18548 2021-05-04 Segher Boessenkool <segher@kernel.crashing.org>
18550 * caller-save.c: Remove CC0.
18551 * cfgcleanup.c: Remove CC0.
18552 * cfgrtl.c: Remove CC0.
18553 * combine.c: Remove CC0.
18554 * compare-elim.c: Remove CC0.
18555 * conditions.h: Remove CC0.
18556 * config/h8300/h8300.h: Remove CC0.
18557 * config/h8300/h8300-protos.h: Remove CC0.
18558 * config/h8300/peepholes.md: Remove CC0.
18559 * config/i386/x86-tune-sched.c: Remove CC0.
18560 * config/m68k/m68k.c: Remove CC0.
18561 * config/rl78/rl78.c: Remove CC0.
18562 * config/sparc/sparc.c: Remove CC0.
18563 * config/xtensa/xtensa.c: Remove CC0.
18564 (gen_conditional_move): Use pc_rtx instead of cc0_rtx in a piece of
18565 RTL where that is used as a placeholder only.
18566 * cprop.c: Remove CC0.
18567 * cse.c: Remove CC0.
18568 * cselib.c: Remove CC0.
18569 * df-problems.c: Remove CC0.
18570 * df-scan.c: Remove CC0.
18571 * doc/md.texi: Remove CC0. Adjust an example.
18572 * doc/rtl.texi: Remove CC0. Adjust an example.
18573 * doc/tm.texi: Regenerate.
18574 * doc/tm.texi.in: Remove CC0.
18575 * emit-rtl.c: Remove CC0.
18576 * final.c: Remove CC0.
18577 * fwprop.c: Remove CC0.
18578 * gcse-common.c: Remove CC0.
18579 * gcse.c: Remove CC0.
18580 * genattrtab.c: Remove CC0.
18581 * genconfig.c: Remove CC0.
18582 * genemit.c: Remove CC0.
18583 * genextract.c: Remove CC0.
18584 * gengenrtl.c: Remove CC0.
18585 * genrecog.c: Remove CC0.
18586 * haifa-sched.c: Remove CC0.
18587 * ifcvt.c: Remove CC0.
18588 * ira-costs.c: Remove CC0.
18589 * ira.c: Remove CC0.
18590 * jump.c: Remove CC0.
18591 * loop-invariant.c: Remove CC0.
18592 * lra-constraints.c: Remove CC0.
18593 * lra-eliminations.c: Remove CC0.
18594 * optabs.c: Remove CC0.
18595 * postreload-gcse.c: Remove CC0.
18596 * postreload.c: Remove CC0.
18597 * print-rtl.c: Remove CC0.
18598 * read-rtl-function.c: Remove CC0.
18599 * reg-notes.def: Remove CC0.
18600 * reg-stack.c: Remove CC0.
18601 * reginfo.c: Remove CC0.
18602 * regrename.c: Remove CC0.
18603 * reload.c: Remove CC0.
18604 * reload1.c: Remove CC0.
18605 * reorg.c: Remove CC0.
18606 * resource.c: Remove CC0.
18607 * rtl.c: Remove CC0.
18608 * rtl.def: Remove CC0.
18609 * rtl.h: Remove CC0.
18610 * rtlanal.c: Remove CC0.
18611 * sched-deps.c: Remove CC0.
18612 * sched-rgn.c: Remove CC0.
18613 * shrink-wrap.c: Remove CC0.
18614 * simplify-rtx.c: Remove CC0.
18615 * system.h: Remove CC0. Poison NOTICE_UPDATE_CC, CC_STATUS_MDEP_INIT,
18616 CC_STATUS_MDEP, and CC_STATUS.
18617 * target.def: Remove CC0.
18618 * valtrack.c: Remove CC0.
18619 * var-tracking.c: Remove CC0.
18621 2021-05-04 Richard Biener <rguenther@suse.de>
18623 PR tree-optimization/100414
18624 * tree-ssa-phiopt.c (get_non_trapping): Do not compute dominance
18626 (tree_ssa_phiopt_worker): But unconditionally here.
18628 2021-05-04 Tobias Burnus <tobias@codesourcery.com>
18630 * omp-low.c (lower_rec_input_clauses, lower_reduction_clauses): Handle
18631 && and || with floating-point and complex arguments.
18633 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
18635 * tree-inline.c (insert_debug_decl_map): Delete.
18636 (copy_debug_stmt): Minor tweak.
18637 (setup_one_parameter): Do not use a variable if the value is either
18638 a read-only DECL or a non-addressable local variable in the caller.
18639 In this case, insert the debug-only variable in the map manually.
18640 (expand_call_inline): Do not generate a CLOBBER for these values.
18641 * tree-inline.h (debug_map): Minor tweak.
18643 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
18645 * builtins.c (builtin_with_linkage_p): Return true for stp[n]cpy.
18646 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Tidy up.
18648 2021-05-04 Richard Biener <rguenther@suse.de>
18650 PR tree-optimization/100329
18651 * tree-ssa-reassoc.c (can_reassociate_p): Do not reassociate
18653 (insert_stmt_after): Assert we're not running into asm goto.
18655 2021-05-04 Richard Biener <rguenther@suse.de>
18657 PR tree-optimization/100398
18658 * tree-ssa-dse.c (pass_dse::execute): Preserve control
18661 2021-05-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18663 * builtins.c (try_store_by_multiple_pieces): Fix constfun's prototype.
18665 2021-05-04 Alexandre Oliva <oliva@adacore.com>
18667 * builtins.c (try_store_by_multiple_pieces): New.
18668 (expand_builtin_memset_args): Use it. If target_char_cast
18669 fails, proceed as for non-constant val. Pass len's ctz to...
18670 * expr.c (clear_storage_hints): ... this. Try store by
18671 multiple pieces after setmem.
18672 (clear_storage): Adjust.
18673 * expr.h (clear_storage_hints): Likewise.
18674 (try_store_by_multiple_pieces): Declare.
18675 * passes.def: Replace the last copy_prop with ccp.
18677 2021-05-03 Tom de Vries <tdevries@suse.de>
18680 * omp-low.c (lower_rec_input_clauses): Disable SIMT for user-defined
18683 2021-05-03 Richard Biener <rguenther@suse.de>
18685 * tree-ssa-dse.c (dse_classify_store): Track two PHI defs.
18687 2021-05-03 Richard Biener <rguenther@suse.de>
18689 * tree-ssa-dse.c: Do not include domwalk.h but cfganal.h.
18690 (dse_dom_walker): Remove.
18691 (dse_dom_walker::dse_optimize_stmt): Rename...
18692 (dse_optimize_stmt): ... to this, pass in live_bytes sbitmap.
18693 (dse_dom_walker::before_dom_children): Inline ...
18694 (pass_dse::execute): ... here. Perform a reverse program
18697 2021-05-03 H.J. Lu <hjl.tools@gmail.com>
18700 * configure: Regenerated.
18702 2021-05-03 Ilya Leoshkevich <iii@linux.ibm.com>
18705 * config/s390/s390.c (s390_hard_fp_reg_p): New function.
18706 (s390_md_asm_adjust): Handle hard registers.
18708 2021-05-03 Jakub Jelinek <jakub@redhat.com>
18710 PR tree-optimization/100382
18711 * tree-ssa-dse.c: Include tree-eh.h.
18712 (dse_dom_walker::before_dom_children): Don't remove stmts if
18713 stmt_unremovable_because_of_non_call_eh_p is true.
18715 2021-05-02 David Edelsohn <dje.gcc@gmail.com>
18717 * varasm.c (compute_reloc_for_var): Split out from...
18718 (get_variable_section): Use it.
18719 * output.h (compute_reloc_for_var): Declare.
18720 * config/rs6000/rs6000-protos.h
18721 (rs6000_xcoff_asm_output_aligned_decl_common): Change alignment to
18723 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address_aix):
18724 Don't append storage mapping class to symbol.
18725 (rs6000_xcoff_asm_named_section): Add BS and UL mapping classes.
18726 Don't convert TLS BSS to common.
18727 (rs6000_xcoff_unique_section): Don't fall back to select_secton.
18728 (rs6000_xcoff_section_type_flags): Add SECTION_BSS if DECL is
18730 (rs6000_xcoff_asm_globalize_decl_name): Don't strip storage
18732 (rs6000_xcoff_asm_output_aligned_decl_common): Align is unsigned int.
18733 If align is 0 from TLS class, use the same rules as varasm.c
18734 If not common, switch to BSS section manually.
18735 If common, emit appropriate comm or lcomm directive.
18736 (rs6000_xcoff_encode_section_info): Add logic to append all
18737 storage mapping classes.
18738 (rs6000_asm_weaken_decl): Adjust for qualname symbols.
18739 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
18740 rs6000_xcoff_asm_output_aligned_decl_common.
18741 (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
18742 rs6000_xcoff_asm_output_aligned_decl_common.
18743 (ASM_OUTPUT_TLS_COMMON): Use
18744 rs6000_xcoff_asm_output_aligned_decl_common.
18746 2021-05-02 Jakub Jelinek <jakub@redhat.com>
18749 * config/nvptx/nvptx.c (nvptx_sese_pseudo): Use nullptr instead of 0
18750 as first argument of pseudo_node_t constructors.
18752 2021-05-02 Jakub Jelinek <jakub@redhat.com>
18755 * config/i386/t-i386 (TM_H): Add $(srcdir)/config/i386/i386-isa.def.
18757 2021-05-01 Aldy Hernandez <aldyh@redhat.com>
18759 * value-range.cc (DEFINE_INT_RANGE_GC_STUBS): Remove.
18760 (gt_pch_nx (int_range<1> *&)): New.
18761 (gt_ggc_mx (int_range<1> *&)): New.
18762 * value-range.h (class irange): Add GTY support for
18765 2021-05-01 Geng Qi <gengqi@linux.alibaba.com>
18767 * doc/options.texi (Negative): Change either or to both and.
18769 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
18771 * config/aarch64/aarch64-simd-builtins.def: Add
18772 float_ml[as][q]_laneq builtin generator macros.
18773 * config/aarch64/aarch64-simd.md (mul_laneq<mode>3): Define.
18774 (aarch64_float_mla_laneq<mode>): Define.
18775 (aarch64_float_mls_laneq<mode>): Define.
18776 * config/aarch64/arm_neon.h (vmla_laneq_f32): Use RTL builtin
18777 instead of GCC vector extensions.
18778 (vmlaq_laneq_f32): Likewise.
18779 (vmls_laneq_f32): Likewise.
18780 (vmlsq_laneq_f32): Likewise.
18782 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
18784 * config/aarch64/aarch64-simd-builtins.def: Add
18785 float_ml[as]_lane builtin generator macros.
18786 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt<mode>):
18788 (mul_lane<mode>3): This, and re-order arguments.
18789 (aarch64_float_mla_lane<mode>): Define.
18790 (aarch64_float_mls_lane<mode>): Define.
18791 * config/aarch64/arm_neon.h (vmla_lane_f32): Use RTL builtin
18792 instead of GCC vector extensions.
18793 (vmlaq_lane_f32): Likewise.
18794 (vmls_lane_f32): Likewise.
18795 (vmlsq_lane_f32): Likewise.
18797 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
18799 * config/aarch64/aarch64-simd-builtins.def: Add float_ml[as]
18800 builtin generator macros.
18801 * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>):
18803 (aarch64_float_mls<mode>): Define.
18804 * config/aarch64/arm_neon.h (vmla_f32): Use RTL builtin
18805 instead of relying on GCC vector extensions.
18806 (vmla_f64): Likewise.
18807 (vmlaq_f32): Likewise.
18808 (vmlaq_f64): Likewise.
18809 (vmls_f32): Likewise.
18810 (vmls_f64): Likewise.
18811 (vmlsq_f32): Likewise.
18812 (vmlsq_f64): Likewise.
18813 * config/aarch64/iterators.md: Define VDQF_DF mode iterator.
18815 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
18817 * config/aarch64/aarch64-simd-builtins.def: Add
18818 float_ml[as]_n_builtin generator macros.
18819 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt_from_dup<mode>):
18821 (mul_n<mode>3): This, and re-order arguments.
18822 (aarch64_float_mla_n<mode>): Define.
18823 (aarch64_float_mls_n<mode>): Define.
18824 * config/aarch64/arm_neon.h (vmla_n_f32): Use RTL builtin
18825 instead of inline asm.
18826 (vmlaq_n_f32): Likewise.
18827 (vmls_n_f32): Likewise.
18828 (vmlsq_n_f32): Likewise.
18830 2021-04-30 Jonathan Wright <joanthan.wright@arm.com>
18832 * config/aarch64/aarch64-simd-builtins.def: Add pmull[2]
18833 builtin generator macros.
18834 * config/aarch64/aarch64-simd.md (aarch64_pmullv8qi): Define.
18835 (aarch64_pmull_hiv16qi_insn): Define.
18836 (aarch64_pmull_hiv16qi): Define.
18837 * config/aarch64/arm_neon.h (vmull_high_p8): Use RTL builtin
18838 instead of inline asm.
18839 (vmull_p8): Likewise.
18841 2021-04-30 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
18843 * config/avr/avr.md: Adjust peepholes to match and
18844 generate parallels with clobber of REG_CC.
18845 (mov<mode>_insn): Rename to mov<mode>_insn_split.
18846 (*mov<mode>_insn): Rename to mov<mode>_insn.
18848 2021-04-30 David Edelsohn <dje.gcc@gmail.com>
18850 * varasm.c (use_blocks_for_decl_p): Don't use section anchors
18851 for VAR_DECLs if -fdata-sections enabled.
18853 2021-04-30 Michael Meissner <meissner@linux.ibm.com>
18855 PR bootstrap/100327
18856 * config/rs6000/rs6000.c
18857 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Define.
18858 (rs6000_libgcc_floating_mode_supported_p): New target hook.
18860 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
18862 * tree-ssa-threadbackward.c (class thread_jumps): Split out code
18864 (class back_threader_registry): ...to here...
18865 (class back_threader_profitability): ...and here...
18866 (thread_jumps::thread_through_all_blocks): Remove argument.
18867 (back_threader_registry::back_threader_registry): New.
18868 (back_threader_registry::~back_threader_registry): New.
18869 (back_threader_registry::thread_through_all_blocks): New.
18870 (thread_jumps::profitable_jump_thread_path): Move from here...
18871 (back_threader_profitability::profitable_path_p): ...to here.
18872 (thread_jumps::find_taken_edge): New.
18873 (thread_jumps::convert_and_register_current_path): Move...
18874 (back_threader_registry::register_path): ...to here.
18875 (thread_jumps::register_jump_thread_path_if_profitable): Move...
18876 (thread_jumps::maybe_register_path): ...to here.
18877 (thread_jumps::handle_phi): Call find_taken_edge and
18878 maybe_register_path.
18879 (thread_jumps::handle_assignment): Same.
18880 (thread_jumps::fsm_find_control_statement_thread_paths): Remove
18881 tree argument to handle_phi and handle_assignment.
18882 (thread_jumps::find_jump_threads_backwards): Set m_name. Remove
18883 set of m_speed_p and m_max_threaded_paths.
18884 (pass_thread_jumps::execute): Remove second argument from
18885 find_jump_threads_backwards.
18886 (pass_early_thread_jumps::execute): Same.
18888 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
18890 * tree-ssa-dom.c (class dom_jump_threader_simplifier): New.
18891 (class dom_opt_dom_walker): Initialize some class variables.
18892 (pass_dominator::execute): Pass evrp_range_analyzer and
18893 dom_jump_threader_simplifier to dom_opt_dom_walker.
18894 Adjust for some functions moving into classes.
18895 (simplify_stmt_for_jump_threading): Adjust and move to...
18896 (jump_threader_simplifier::simplify): ...here.
18897 (dom_opt_dom_walker::before_dom_children): Adjust for
18898 m_evrp_range_analyzer.
18899 (dom_opt_dom_walker::after_dom_children): Remove x_vr_values hack.
18900 (test_for_singularity): Place in dom_opt_dom_walker class.
18901 (dom_opt_dom_walker::optimize_stmt): The argument
18902 evrp_range_analyzer is now a class field.
18903 * tree-ssa-threadbackward.c (class thread_jumps): Add m_registry.
18904 (thread_jumps::thread_through_all_blocks): New.
18905 (thread_jumps::convert_and_register_current_path): Use m_registry.
18906 (pass_thread_jumps::execute): Adjust for thread_through_all_blocks
18907 being in the threader class.
18908 (pass_early_thread_jumps::execute): Same.
18909 * tree-ssa-threadedge.c (threadedge_initialize_values): Move...
18910 (jump_threader::jump_threader): ...here.
18911 (threadedge_finalize_values): Move...
18912 (jump_threader::~jump_threader): ...here.
18913 (jump_threader::remove_jump_threads_including): New.
18914 (jump_threader::thread_through_all_blocks): New.
18915 (record_temporary_equivalences_from_phis): Move...
18916 (jump_threader::record_temporary_equivalences_from_phis): ...here.
18917 (record_temporary_equivalences_from_stmts_at_dest): Move...
18918 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
18920 (simplify_control_stmt_condition_1): Move to jump_threader class.
18921 (simplify_control_stmt_condition): Move...
18922 (jump_threader::simplify_control_stmt_condition): ...here.
18923 (thread_around_empty_blocks): Move...
18924 (jump_threader::thread_around_empty_blocks): ...here.
18925 (thread_through_normal_block): Move...
18926 (jump_threader::thread_through_normal_block): ...here.
18927 (thread_across_edge): Move...
18928 (jump_threader::thread_across_edge): ...here.
18929 (thread_outgoing_edges): Move...
18930 (jump_threader::thread_outgoing_edges): ...here.
18931 * tree-ssa-threadedge.h: Move externally facing functings...
18932 (class jump_threader): ...here...
18933 (class jump_threader_simplifier): ...and here.
18934 * tree-ssa-threadupdate.c (struct redirection_data): Remove comment.
18935 (jump_thread_path_allocator::jump_thread_path_allocator): New.
18936 (jump_thread_path_allocator::~jump_thread_path_allocator): New.
18937 (jump_thread_path_allocator::allocate_thread_edge): New.
18938 (jump_thread_path_allocator::allocate_thread_path): New.
18939 (jump_thread_path_registry::jump_thread_path_registry): New.
18940 (jump_thread_path_registry::~jump_thread_path_registry): New.
18941 (jump_thread_path_registry::allocate_thread_edge): New.
18942 (jump_thread_path_registry::allocate_thread_path): New.
18943 (dump_jump_thread_path): Make extern.
18944 (debug (const vec<jump_thread_edge *> &path)): New.
18945 (struct removed_edges): Move to tree-ssa-threadupdate.h.
18946 (struct thread_stats_d): Remove.
18947 (remove_ctrl_stmt_and_useless_edges): Make static.
18948 (lookup_redirection_data): Move...
18949 (jump_thread_path_registry::lookup_redirection_data): ...here.
18950 (ssa_redirect_edges): Make static.
18951 (thread_block_1): Move...
18952 (jump_thread_path_registry::thread_block_1): ...here.
18953 (thread_block): Move...
18954 (jump_thread_path_registry::thread_block): ...here.
18955 (thread_through_loop_header): Move...
18956 (jump_thread_path_registry::thread_through_loop_header): ...here.
18957 (mark_threaded_blocks): Move...
18958 (jump_thread_path_registry::mark_threaded_blocks): ...here.
18959 (debug_path): Move...
18960 (jump_thread_path_registry::debug_path): ...here.
18961 (debug_all_paths): Move...
18962 (jump_thread_path_registry::dump): ..here.
18963 (rewire_first_differing_edge): Move...
18964 (jump_thread_path_registry::rewire_first_differing_edge): ...here.
18965 (adjust_paths_after_duplication): Move...
18966 (jump_thread_path_registry::adjust_paths_after_duplication): ...here.
18967 (duplicate_thread_path): Move...
18968 (jump_thread_path_registry::duplicate_thread_path): ..here.
18969 (remove_jump_threads_including): Move...
18970 (jump_thread_path_registry::remove_jump_threads_including): ...here.
18971 (thread_through_all_blocks): Move to...
18972 (jump_thread_path_registry::thread_through_all_blocks): ...here.
18973 (delete_jump_thread_path): Remove.
18974 (register_jump_thread): Move...
18975 (jump_thread_path_registry::register_jump_thread): ...here.
18976 * tree-ssa-threadupdate.h: Move externally facing functions...
18977 (class jump_thread_path_allocator): ...here...
18978 (class jump_thread_path_registry): ...and here.
18979 (thread_through_all_blocks): Remove.
18980 (struct removed_edges): New.
18981 (register_jump_thread): Remove.
18982 (remove_jump_threads_including): Remove.
18983 (delete_jump_thread_path): Remove.
18984 (remove_ctrl_stmt_and_useless_edges): Remove.
18985 (free_dom_edge_info): New prototype.
18986 * tree-vrp.c: Remove x_vr_values hack.
18987 (class vrp_jump_threader_simplifier): New.
18988 (vrp_jump_threader_simplifier::simplify): New.
18989 (vrp_jump_threader::vrp_jump_threader): Adjust method signature.
18990 Remove m_dummy_cond.
18991 Instantiate m_simplifier and m_threader.
18992 (vrp_jump_threader::thread_through_all_blocks): New.
18993 (vrp_jump_threader::simplify_stmt): Remove.
18994 (vrp_jump_threader::after_dom_children): Do not set m_dummy_cond.
18995 Remove x_vr_values hack.
18996 (execute_vrp): Adjust for thread_through_all_blocks being in a
18999 2021-04-30 Christophe Lyon <christophe.lyon@linaro.org>
19001 * genflags.c (gen_insn): Print failed expansion string.
19003 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
19005 * expr.c (alignment_for_piecewise_move): Call mode_for_size
19006 without limit to MAX_FIXED_MODE_SIZE.
19008 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
19010 PR middle-end/90773
19011 * builtins.c (builtin_memset_gen_str): Don't use return from
19012 simplify_gen_subreg.
19014 2021-04-30 Uroš Bizjak <ubizjak@gmail.com>
19017 * config/i386/i386.md (*add<mode>3_carry_0r): New insn pattern.
19018 (*addsi3_carry_zext_0r): Ditto.
19019 (*sub<mode>3_carry_0): Ditto.
19020 (*subsi3_carry_zext_0r): Ditto.
19021 * config/i386/predicates.md (ix86_carry_flag_unset_operator):
19023 * config/i386/i386.c (ix86_rtx_costs) <case PLUS, case MINUS>:
19024 Also consider ix86_carry_flag_unset_operator to calculate
19025 the cost of adc/sbb insn.
19027 2021-04-30 Roman Zhuykov <zhroma@ispras.ru>
19029 PR rtl-optimization/100225
19030 PR rtl-optimization/84878
19031 * modulo-sched.c (sms_schedule): Use note_stores to skip loops
19032 where we have an instruction which touches (writes) any hard
19033 register from df->regular_block_artificial_uses set.
19034 Allow not-single-set instruction only right before basic block
19037 2021-04-30 Geng Qi <gengqi@linux.alibaba.com>
19039 * config/riscv/riscv.opt (march=,mabi=): Negative itself.
19041 2021-04-30 LevyHsu <admin@levyhsu.com>
19043 * config/riscv/riscv.c (riscv_min_arithmetic_precision): New.
19044 * config/riscv/riscv.h (TARGET_MIN_ARITHMETIC_PRECISION): New.
19045 * config/riscv/riscv.md (addv<mode>4, uaddv<mode>4): New.
19046 (subv<mode>4, usubv<mode>4, mulv<mode>4, umulv<mode>4): New.
19048 2021-04-29 Alexandre Oliva <oliva@adacore.com>
19050 * config.gcc: Merged x86 and x86_64 cpu_type-setting cases.
19052 2021-04-29 Alexandre Oliva <oliva@adacore.com>
19054 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Rename to...
19055 (ASM_OUTPUT_MAX_SKIP_ALIGN): ... this. Enclose in do/while(0).
19056 * config/i386/i386.c: Adjust.
19057 * config/i386/i386.md: Adjust.
19058 * config/i386/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Drop.
19059 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
19060 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
19061 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
19062 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
19063 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
19064 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
19065 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
19066 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
19067 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
19068 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
19070 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
19072 * config/i386/i386-expand.c (ix86_expand_int_compare):
19073 Swap operands of GTU and LEU comparison to emit carry flag comparison.
19074 * config/i386/i386.md (*add<mode>3_carry_0): Change insn
19075 predicate to allow more combine opportunities with memory operands.
19076 (*sub<mode>3_carry_0): Ditto.
19078 2021-04-29 Richard Sandiford <richard.sandiford@arm.com>
19080 PR rtl-optimization/100303
19081 * rtl-ssa/accesses.cc (function_info::make_use_available): Take a
19082 boolean that indicates whether the use will only be used in
19083 debug instructions. Treat it in the same way that existing
19084 cross-EBB debug references would be handled if so.
19085 (function_info::make_uses_available): Likewise.
19086 * rtl-ssa/functions.h (function_info::make_uses_available): Update
19087 prototype accordingly.
19088 (function_info::make_uses_available): Likewise.
19089 * fwprop.c (try_fwprop_subst): Update call accordingly.
19091 2021-04-29 Jeff Law <jlaw@tachyum.com>
19093 * config/nios2/nios2-protos.h (nios2_fpu_insn_enabled): Move outside
19096 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
19097 Richard Biener <rguenther@suse.de>
19100 * config/i386/i386-builtin.def (IX86_BUILTIN_MASKLOADPD)
19101 (IX86_BUILTIN_MASKLOADPS, IX86_BUILTIN_MASKLOADPD256)
19102 (IX86_BUILTIN_MASKLOADPS256, IX86_BUILTIN_MASKLOADD)
19103 (IX86_BUILTIN_MASKLOADQ, IX86_BUILTIN_MASKLOADD256)
19104 (IX86_BUILTIN_MASKLOADQ256): Move from SPECIAL_ARGS
19105 to PURE_ARGS category.
19106 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
19107 Handle PURE_ARGS category.
19108 * config/i386/i386-expand.c (ix86_expand_builtin): Ditto.
19110 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
19112 * configure.ac: Check for the presence of sys/locking.h header and
19113 for whether _LK_LOCK is supported by _locking.
19114 * configure: Regenerate.
19115 * config.in: Likewise.
19116 * gcov-io.h: Define GCOV_LOCKED_WITH_LOCKING if HOST_HAS_LK_LOCK.
19117 * gcov-io.c (gcov_open): Add support for GCOV_LOCKED_WITH_LOCKING.
19118 * system.h: Include <sys/locking.h> if HAVE_SYS_LOCKING_H.
19120 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
19122 * config/i386/predicates.md (fcmov_comparison_operator):
19123 Do not check for trivial FP comparison operator.
19124 <case GEU, case LTU>: Allow CCGZmode.
19125 <case GTU, case LEU>: Do not allow CCCmode.
19126 (ix86_comparison_operator) <case GTU, case LEU>: Allow only CCmode.
19127 (ix86_carry_flag_operator): Match only LTU and UNLT code.
19128 Do not check for trivial FP comparison operator. Allow CCGZmode.
19130 2021-04-29 Tom de Vries <tdevries@suse.de>
19132 * omp-expand.c (expand_omp_simd): Add step_orig, and replace uses of
19133 fd->loop.step by either step or orig_step.
19135 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
19137 * config/sparc/sparc.c (gen_load_pcrel_sym): Delete.
19138 (load_got_register): Do the PIC dance here.
19139 (sparc_legitimize_tls_address): Simplify.
19140 (sparc_emit_probe_stack_range): Likewise.
19141 (sparc32_initialize_trampoline): Likewise.
19142 (sparc64_initialize_trampoline): Likewise.
19143 * config/sparc/sparc.md (load_pcrel_sym<P:mode>): Add @ marker.
19144 (probe_stack_range<P:mode>): Likewise.
19145 (flush<P:mode>): Likewise.
19146 (tgd_hi22<P:mode>): Likewise.
19147 (tgd_lo10<P:mode>): Likewise.
19148 (tgd_add<P:mode>): Likewise.
19149 (tgd_call<P:mode>): Likewise.
19150 (tldm_hi22<P:mode>): Likewise.
19151 (tldm_lo10<P:mode>): Likewise.
19152 (tldm_add<P:mode>): Likewise.
19153 (tldm_call<P:mode>): Likewise.
19154 (tldo_hix22<P:mode>): Likewise.
19155 (tldo_lox10<P:mode>): Likewise.
19156 (tldo_add<P:mode>): Likewise.
19157 (tie_hi22<P:mode>): Likewise.
19158 (tie_lo10<P:mode>): Likewise.
19159 (tie_add<P:mode>): Likewise.
19160 (tle_hix22<P:mode>): Likewise.
19161 (tle_lox10<P:mode>): Likewise.
19162 (stack_protect_setsi): Rename to...
19163 (stack_protect_set32): ...this.
19164 (stack_protect_setdi): Rename to...
19165 (stack_protect_set64): ...this.
19166 (stack_protect_set): Adjust calls to above.
19167 (stack_protect_testsi): Rename to...
19168 (stack_protect_test32): ...this.
19169 (stack_protect_testdi): Rename to...
19170 (stack_protect_test64): ...this.
19171 (stack_protect_test): Adjust calls to above.
19173 2021-04-29 H.J. Lu <hjl.tools@gmail.com>
19175 PR middle-end/90773
19176 * builtins.c (builtin_memcpy_read_str): Add a dummy argument.
19177 (builtin_strncpy_read_str): Likewise.
19178 (builtin_memset_read_str): Add an argument for the previous RTL
19179 information and generate the new RTL from the previous RTL info.
19180 (builtin_memset_gen_str): Likewise.
19181 * builtins.h (builtin_strncpy_read_str): Update the prototype.
19182 (builtin_memset_read_str): Likewise.
19183 * expr.c (by_pieces_ninsns): If targetm.overlap_op_by_pieces_p()
19184 returns true, round up size and alignment to the widest integer
19185 mode for maximum size.
19186 (pieces_addr::adjust): Add a pointer to by_pieces_prev argument
19187 and pass it to m_constfn.
19188 (op_by_pieces_d): Add m_push and m_overlap_op_by_pieces.
19189 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
19190 initialize m_push. Initialize m_overlap_op_by_pieces with
19191 targetm.overlap_op_by_pieces_p ().
19192 (op_by_pieces_d::run): Pass the previous RTL information to
19193 pieces_addr::adjust and generate overlapping operations if
19194 m_overlap_op_by_pieces is true.
19196 (move_by_pieces_d::move_by_pieces_d): Updated for op_by_pieces_d
19198 (store_by_pieces_d::store_by_pieces_d): Updated for op_by_pieces_d
19200 (can_store_by_pieces): Use by_pieces_constfn on constfun.
19201 (store_by_pieces): Use by_pieces_constfn on constfun. Updated
19202 for op_by_pieces_d change.
19203 (clear_by_pieces_1): Add a dummy argument.
19204 (clear_by_pieces): Updated for op_by_pieces_d change.
19205 (compare_by_pieces_d::compare_by_pieces_d): Likewise.
19206 (string_cst_read_str): Add a dummy argument.
19207 * expr.h (by_pieces_constfn): Add a dummy argument.
19208 (by_pieces_prev): New.
19209 * target.def (overlap_op_by_pieces_p): New target hook.
19210 * config/i386/i386.c (TARGET_OVERLAP_OP_BY_PIECES_P): New.
19211 * doc/tm.texi.in: Add TARGET_OVERLAP_OP_BY_PIECES_P.
19212 * doc/tm.texi: Regenerated.
19214 2021-04-29 Richard Biener <rguenther@suse.de>
19216 PR tree-optimization/100253
19217 * tree-vect-stmts.c (vectorizable_load): Do not assume
19218 element alignment when DR_MISALIGNMENT is -1.
19219 (vectorizable_store): Likewise.
19221 2021-04-29 Jakub Jelinek <jakub@redhat.com>
19224 * config/aarch64/aarch64.c (aarch64_add_offset_1_temporaries): Use
19225 absu_hwi instead of abs_hwi.
19227 2021-04-29 Richard Biener <rguenther@suse.de>
19229 PR middle-end/38474
19230 * tree-ssa-structalias.c (add_graph_edge): Avoid direct
19231 forwarding when indirect forwarding through ESCAPED
19234 2021-04-29 Tom de Vries <tdevries@suse.de>
19237 * internal-fn.c (expand_GOMP_SIMT_ENTER_ALLOC)
19238 (expand_GOMP_SIMT_LAST_LANE, expand_GOMP_SIMT_ORDERED_PRED)
19239 (expand_GOMP_SIMT_VOTE_ANY, expand_GOMP_SIMT_XCHG_BFLY)
19240 (expand_GOMP_SIMT_XCHG_IDX): Ensure target is assigned to.
19242 2021-04-29 Richard Biener <rguenther@suse.de>
19244 PR tree-optimization/99912
19245 * tree-ssa-dse.c (dse_dom_walker::m_need_cfg_cleanup): New.
19246 (dse_dom_walker::todo): Likewise.
19247 (dse_dom_walker::dse_optimize_stmt): Move VDEF check to the
19249 (dse_dom_walker::before_dom_children): Remove trivially
19250 dead SSA defs and schedule CFG cleanup if we removed all
19252 (pass_dse::execute): Get TODO as computed by the DOM walker
19253 and return it. Wipe dominator info earlier.
19255 2021-04-29 Richard Biener <rguenther@suse.de>
19258 * ipa-prop.c (ipcp_modif_dom_walker::before_dom_children):
19259 Track blocks to cleanup EH in new m_need_eh_cleanup.
19260 (ipcp_modif_dom_walker::cleanup_eh): New.
19261 (ipcp_transform_function): Release dominator info before
19264 2021-04-29 Martin Sebor <msebor@redhat.com>
19266 PR middle-end/100250
19267 * attribs.c (attr_access::array_as_string): Avoid dereferencing
19268 a pointer when it's null.
19270 2021-04-29 Martin Sebor <msebor@redhat.com>
19272 * Makefile.in (OBJS): Add ipa-free-lang-data.o.
19273 * ipa-free-lang-data.cc: New file.
19274 * tree.c: Move pass free_lang_data to file above.
19275 (build_array_type_1): Declare extern.
19276 * tree.h (build_array_type_1): Declare.
19278 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
19280 * config/aarch64/aarch64-simd-builtins.def: Modify comment to
19281 make consistent with updated RTL pattern.
19282 * config/aarch64/aarch64-simd.md (aarch64_<sur>qmovn<mode>):
19283 Implement using ss_truncate and us_truncate rather than
19285 * config/aarch64/iterators.md: Remove redundant unspecs and
19286 iterator: UNSPEC_[SU]QXTN and SUQMOVN respectively.
19288 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
19290 * config/aarch64/arm_acle.h (__attribute__): Make intrinsic
19291 attributes consistent with those defined in arm_neon.h.
19293 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
19295 * config/aarch64/arm_fp16.h (__attribute__): Make intrinsic
19296 attributes consistent with those defined in arm_neon.h.
19298 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
19300 * config/aarch64/aarch64-simd-builtins.def: Add
19301 float_trunc_rodd builtin generator macros.
19302 * config/aarch64/aarch64-simd.md (aarch64_float_trunc_rodd_df):
19304 (aarch64_float_trunc_rodd_lo_v2sf): Define.
19305 (aarch64_float_trunc_rodd_hi_v4sf_le): Define.
19306 (aarch64_float_trunc_rodd_hi_v4sf_be): Define.
19307 (aarch64_float_trunc_rodd_hi_v4sf): Define.
19308 * config/aarch64/arm_neon.h (vcvtx_f32_f64): Use RTL builtin
19309 instead of inline asm.
19310 (vcvtx_high_f32_f64): Likewise.
19311 (vcvtxd_f32_f64): Likewise.
19312 * config/aarch64/iterators.md: Add FCVTXN unspec.
19314 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
19316 * config/aarch64/aarch64-simd-builtins.def: Add tbx1 builtin
19318 * config/aarch64/aarch64-simd.md (aarch64_tbx1<mode>):
19320 * config/aarch64/arm_neon.h (vqtbx1_s8): USE RTL builtin
19321 instead of inline asm.
19322 (vqtbx1_u8): Likewise.
19323 (vqtbx1_p8): Likewise.
19324 (vqtbx1q_s8): Likewise.
19325 (vqtbx1q_u8): Likewise.
19326 (vqtbx1q_p8): Likewise.
19327 (vtbx2_s8): Likewise.
19328 (vtbx2_u8): Likewise.
19329 (vtbx2_p8): Likewise.
19331 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
19333 * config/aarch64/aarch64-simd-builtins.def: Add tbl1 builtin
19335 * config/aarch64/arm_neon.h (vqtbl1_p8): Use RTL builtin
19336 instead of inline asm.
19337 (vqtbl1_s8): Likewise.
19338 (vqtbl1_u8): Likewise.
19339 (vqtbl1q_p8): Likewise.
19340 (vqtbl1q_s8): Likewise.
19341 (vqtbl1q_u8): Likewise.
19342 (vtbl1_s8): Likewise.
19343 (vtbl1_u8): Likewise.
19344 (vtbl1_p8): Likewise.
19345 (vtbl2_s8): Likewise.
19346 (vtbl2_u8): Likewise.
19347 (vtbl2_p8): Likewise.
19349 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
19351 * config/aarch64/aarch64-simd-builtins.def: Add polynomial
19352 ssri_n buitin generator macro.
19353 * config/aarch64/arm_neon.h (vsri_n_p8): Use RTL builtin
19354 instead of inline asm.
19355 (vsri_n_p16): Likewise.
19356 (vsri_n_p64): Likewise.
19357 (vsriq_n_p8): Likewise.
19358 (vsriq_n_p16): Likewise.
19359 (vsriq_n_p64): Likewise.
19361 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
19363 * config/aarch64/aarch64-simd-builtins.def: Use VALLP mode
19364 iterator for polynomial ssli_n builtin generator macro.
19365 * config/aarch64/arm_neon.h (vsli_n_p8): Use RTL builtin
19366 instead of inline asm.
19367 (vsli_n_p16): Likewise.
19368 (vsliq_n_p8): Likewise.
19369 (vsliq_n_p16): Likewise.
19370 * config/aarch64/iterators.md: Define VALLP mode iterator.
19372 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
19374 * config/aarch64/aarch64-simd-builtins.def: Use VDQV_L
19375 iterator to generate [su]adalp RTL builtins.
19376 * config/aarch64/aarch64-simd.md: Use VDQV_L iterator in
19377 [su]adalp RTL pattern.
19378 * config/aarch64/arm_neon.h (vpadal_s32): Use RTL builtin
19379 instead of inline asm.
19380 (vpadal_u32): Likewise.
19382 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
19384 * config/aarch64/aarch64-simd-builtins.def: Add [su]addlp
19385 builtin generator macros.
19386 * config/aarch64/aarch64-simd.md (aarch64_<su>addlp<mode>):
19388 * config/aarch64/arm_neon.h (vpaddl_s8): Use RTL builtin
19389 instead of inline asm.
19390 (vpaddl_s16): Likewise.
19391 (vpaddl_s32): Likewise.
19392 (vpaddl_u8): Likewise.
19393 (vpaddl_u16): Likewise.
19394 (vpaddl_u32): Likewise.
19395 (vpaddlq_s8): Likewise.
19396 (vpaddlq_s16): Likewise.
19397 (vpaddlq_s32): Likewise.
19398 (vpaddlq_u8): Likewise.
19399 (vpaddlq_u16): Likewise.
19400 (vpaddlq_u32): Liwewise.
19401 * config/aarch64/iterators.md: Define [SU]ADDLP unspecs with
19402 appropriate attributes.
19404 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
19406 * config/aarch64/aarch64-simd-builtins.def: Use VDQ_I iterator
19407 for aarch64_addp<mode> builtin macro generator.
19408 * config/aarch64/aarch64-simd.md: Use VDQ_I iterator in
19409 aarch64_addp<mode> RTL pattern.
19410 * config/aarch64/arm_neon.h (vpaddq_s8): Use RTL builtin
19411 instead of inline asm.
19412 (vpaddq_s16): Likewise.
19413 (vpaddq_s32): Likewise.
19414 (vpaddq_s64): Likewise.
19415 (vpaddq_u8): Likewise.
19416 (vpaddq_u16): Likewise.
19417 (vpaddq_u32): Likewise.
19418 (vpaddq_u64): Likewise.
19420 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
19422 * config/aarch64/aarch64-simd-builtins.def: Add sq[r]dmulh_n
19423 builtin generator macros.
19424 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_n<mode>):
19426 * config/aarch64/arm_neon.h (vqdmulh_n_s16): Use RTL builtin
19427 instead of inline asm.
19428 (vqdmulh_n_s32): Likewise.
19429 (vqdmulhq_n_s16): Likewise.
19430 (vqdmulhq_n_s32): Likewise.
19431 (vqrdmulh_n_s16): Likewise.
19432 (vqrdmulh_n_s32): Likewise.
19433 (vqrdmulhq_n_s16): Likewise.
19434 (vqrdmulhq_n_s32): Likewise.
19436 2021-04-28 Tobias Burnus <tobias@codesourcery.com>
19438 * doc/install.texi (--enable-offload-defaulted): Document.
19440 2021-04-28 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
19442 * config/avr/avr-dimode.md: Turn existing patterns into
19443 define_insn_and_split style patterns where the splitter
19444 adds a clobber of the condition code register. Drop "cc"
19445 attribute. Add new patterns to match output of
19447 * config/avr/avr-fixed.md: Likewise.
19448 * config/avr/avr.c (cc_reg_rtx): New.
19449 (avr_parallel_insn_from_insns): Adjust insn count
19450 for removal of set of cc0.
19451 (avr_is_casesi_sequence): Likewise.
19452 (avr_casei_sequence_check_operands): Likewise.
19453 (avr_optimize_casesi): Likewise. Also insert
19454 new insns after jump_insn.
19455 (avr_pass_casesi::avr_rest_of_handle_casesi): Adjust
19456 for removal of set of cc0.
19457 (avr_init_expanders): Initialize cc_reg_rtx.
19458 (avr_regno_reg_class): Handle REG_CC.
19459 (cond_string): Remove usage of CC_OVERFLOW_UNUSABLE.
19460 (avr_notice_update_cc): Remove function.
19461 (ret_cond_branch): Remove usage of CC_OVERFLOW_UNUSABLE.
19462 (compare_condition): Adjust for PARALLEL with
19464 (out_shift_with_cnt): Likewise.
19465 (ashlhi3_out): Likewise.
19466 (ashrhi3_out): Likewise.
19467 (lshrhi3_out): Likewise.
19468 (avr_class_max_nregs): Return single reg for REG_CC.
19469 (avr_compare_pattern): Check for REG_CC instead
19471 (avr_reorg_remove_redundant_compare): Likewise.
19472 (avr_reorg):Adjust for PARALLEL with REG_CC clobber.
19473 (avr_hard_regno_nregs): Return single reg for REG_CC.
19474 (avr_hard_regno_mode_ok): Allow only CCmode for REG_CC.
19475 (avr_md_asm_adjust): Clobber REG_CC.
19476 (TARGET_HARD_REGNO_NREGS): Define.
19477 (TARGET_CLASS_MAX_NREGS): Define.
19478 (TARGET_MD_ASM_ADJUST): Define.
19479 * config/avr/avr.h (FIRST_PSEUDO_REGISTER): Adjust
19481 (enum reg_class): Add CC_REG class.
19482 (NOTICE_UPDATE_CC): Remove.
19483 (CC_OVERFLOW_UNUSABLE): Remove.
19484 (CC_NO_CARRY): Remove.
19485 * config/avr/avr.md: Turn existing patterns into
19486 define_insn_and_split style patterns where the splitter
19487 adds a clobber of the condition code register. Drop "cc"
19488 attribute. Add new patterns to match output of
19490 (sez): Remove unused pattern.
19492 2021-04-28 Richard Earnshaw <rearnsha@arm.com>
19495 * config/arm/arm.c (arm_hard_regno_mode_ok): Only allow VPR to be
19498 2021-04-28 Richard Sandiford <richard.sandiford@arm.com>
19501 * config/aarch64/constraints.md (Utq): Require the address to
19502 be valid for both the element mode and for V2DImode.
19504 2021-04-28 Jakub Jelinek <jakub@redhat.com>
19505 Tobias Burnus <tobias@codesourcery.com>
19507 * configure.ac (OFFLOAD_DEFAULTED): AC_DEFINE if offload-defaulted.
19508 * gcc.c (process_command): New variable.
19509 (driver::maybe_putenv_OFFLOAD_TARGETS): If OFFLOAD_DEFAULTED,
19510 set it if -foffload is defaulted.
19511 * lto-wrapper.c (OFFLOAD_TARGET_DEFAULT_ENV): Define.
19512 (compile_offload_image): If OFFLOAD_DEFAULTED and
19513 OFFLOAD_TARGET_DEFAULT is in the environment, don't fail
19514 if corresponding mkoffload can't be found.
19515 (compile_images_for_offload_targets): Likewise. Free and clear
19516 offload_names if no valid offload is found.
19517 * config.in: Regenerate.
19518 * configure: Regenerate.
19520 2021-04-28 Richard Biener <rguenther@suse.de>
19522 PR tree-optimization/100292
19523 * tree-vect-generic.c (expand_vector_condition): Do not fold
19526 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
19528 * config/rs6000/aix.h (SUBTARGET_DRIVER_SELF_SPECS): New.
19529 * config/rs6000/aix64.opt (m64): New.
19532 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
19534 * config/vax/vax.c (print_operand_address, vax_address_cost_1)
19535 (index_term_p): Handle ASHIFT too.
19537 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
19539 * config/vax/builtins.md (jbb<ccss>i<mode>): Remove operand #3.
19540 (sync_lock_test_and_set<mode>): Adjust accordingly.
19541 (sync_lock_release<mode>): Likewise.
19543 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
19545 * config/vax/vax-protos.h (adjacent_operands_p): Remove
19547 * config/vax/vax.c (adjacent_operands_p): Remove.
19549 2021-04-27 Maciej W. Rozycki <macro@linux-mips.org>
19551 * ifcvt.c (dead_or_predicable) [!IFCVT_MODIFY_TESTS]: Fall
19552 through to the non-conditional execution case if getting the
19553 condition for conditional execution has failed.
19555 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
19557 PR middle-end/100284
19558 * gimple.c (gimple_could_trap_p_1): Remove VEC_COND_EXPR test.
19559 * tree-eh.c (operation_could_trap_p): Handle VEC_COND_EXPR rather
19560 than asserting on it.
19562 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
19564 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): Protect
19565 with TARGET_AIX_OS.
19567 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
19570 * calls.c (precompute_register_parameters): Additionally test
19571 targetm.precompute_tls_p to pre-compute argument.
19572 * config/rs6000/aix.h (TARGET_PRECOMPUTE_TLS_P): Define.
19573 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): New.
19574 * target.def (precompute_tls_p): New.
19575 * doc/tm.texi.in (TARGET_PRECOMPUTE_TLS_P): Add hook documentation.
19576 * doc/tm.texi: Regenerated.
19578 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19581 * config/aarch64/aarch64.c (aarch64_print_operand): Cast -UINTVAL
19582 back to HOST_WIDE_INT.
19584 2021-04-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
19587 * simplify-rtx.c (simplify_context::simplify_subreg): Check the
19588 memory alignment for the outer mode.
19590 2021-04-27 H.J. Lu <hjl.tools@gmail.com>
19592 PR middle-end/90773
19593 * expr.c (op_by_pieces_d::get_usable_mode): New member function.
19594 (op_by_pieces_d::run): Cange a while loop to a do-while loop.
19596 2021-04-27 Alex Coplan <alex.coplan@arm.com>
19599 * config/arm/arm.c (arm_split_compare_and_swap): Fix up codegen
19600 with negative immediates: ensure we expand cbranchsi4_scratch
19601 correctly and ensure we satisfy its constraints.
19602 * config/arm/sync.md
19603 (@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): Don't
19604 attempt to tie two output operands together with constraints;
19605 collapse two alternatives.
19606 (@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
19607 * config/arm/thumb1.md (cbranchsi4_neg_late): New.
19609 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19612 * config/aarch64/predicates.md (aarch64_sub_immediate,
19613 aarch64_plus_immediate): Use -UINTVAL instead of -INTVAL.
19614 * config/aarch64/aarch64.md (casesi, rotl<mode>3): Likewise.
19615 * config/aarch64/aarch64.c (aarch64_print_operand,
19616 aarch64_split_atomic_op, aarch64_expand_subvti): Likewise.
19618 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19620 PR tree-optimization/100239
19621 * tree-vect-generic.c (lower_vec_perm): Don't accept constant
19622 permutations with all indices from the first zero element as vec_shl.
19624 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19626 PR rtl-optimization/100254
19627 * cfgcleanup.c (outgoing_edges_match): Check REG_EH_REGION on
19628 last1 and last2 insns rather than BB_END (bb1) and BB_END (bb2) insns.
19630 2021-04-27 Richard Biener <rguenther@suse.de>
19632 PR tree-optimization/99912
19633 * passes.def: Add comment about new TODO_remove_unused_locals.
19634 * tree-stdarg.c (pass_data_stdarg): Run TODO_remove_unused_locals
19637 2021-04-27 Richard Biener <rguenther@suse.de>
19639 PR tree-optimization/99912
19640 * passes.def (pass_all_optimizations): Add pass_dse before
19641 the first pass_dce, move the first pass_dse before the
19642 pass_dce following pass_pre.
19644 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19646 PR tree-optimization/95527
19647 * generic-match-head.c: Include tm.h.
19648 * gimple-match-head.c: Include tm.h.
19649 * match.pd (CLZ == INTEGER_CST): Don't use
19650 #ifdef CLZ_DEFINED_VALUE_AT_ZERO, only test CLZ_DEFINED_VALUE_AT_ZERO
19651 if clz == CFN_CLZ. Add missing val declaration.
19652 (CTZ cmp CST): New simplifications.
19654 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19656 PR tree-optimization/96696
19657 * expr.c (expand_expr_divmod): New function.
19658 (expand_expr_real_2) <case TRUNC_DIV_EXPR>: Use it for truncations and
19659 divisions. Formatting fixes.
19660 <case MULT_EXPR>: Optimize x / y * y as x - x % y if the latter is
19663 2021-04-27 Martin Jambor <mjambor@suse.cz>
19666 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
19667 If removing a call statement LHS SSA name, release it.
19669 2021-04-27 Richard Earnshaw <rearnsha@arm.com>
19672 * config/arm/arm.c (THUMB2_WORK_REGS): Check PIC_OFFSET_TABLE_REGNUM
19673 is valid before including it in the mask.
19675 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
19678 * config/aarch64/aarch64.c (aarch64_comp_type_attributes): Handle
19681 2021-04-27 Richard Biener <rguenther@suse.de>
19683 PR tree-optimization/100051
19684 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Add
19685 disambiguator based on access size vs. decl size.
19687 2021-04-27 Richard Biener <rguenther@suse.de>
19689 PR tree-optimization/100278
19690 * tree-ssa-pre.c (compute_avail): Give up when we cannot
19691 adjust TBAA beacuse of mismatching bases.
19693 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19696 * config/i386/i386.md (*<insn><mode>3_mask, *<insn><mode>3_mask_1):
19697 For any_rotate define_insn_split and following splitters, use
19698 SWI iterator instead of SWI48.
19700 2021-04-27 Richard Biener <rguenther@suse.de>
19702 PR tree-optimization/99776
19703 * match.pd (bit_field_ref (ctor)): Relax element extract
19704 type compatibility checks.
19706 2021-04-27 Cui,Lili <lili.cui@intel.com>
19708 * common/config/i386/i386-common.c (processor_names):
19709 Sync processor_names with processor_type.
19710 * config/i386/i386-options.c (processor_cost_table):
19711 Sync processor_cost_table with processor_type.
19713 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
19715 * value-range.cc (irange::irange_set_1bit_anti_range): Add assert.
19716 (irange::set): Call irange_set_1bit_anti_range for handling all
19717 1-bit ranges. Fall through on ~[MIN,MAX].
19719 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
19721 * value-range.cc (irange::legacy_num_pairs): Remove.
19722 (irange::invert): Change gcc_assert to gcc_checking_assert.
19723 * value-range.h (irange::num_pairs): Adjust for a cached
19724 num_pairs(). Also, rename all gcc_assert's to
19725 gcc_checking_assert's.
19727 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
19729 * value-range.cc (irange::operator=): Set m_kind.
19730 (irange::copy_to_legacy): Handle varying and undefined sources
19731 as a legacy copy since they can be easily copied.
19732 (irange::irange_set): Set m_kind.
19733 (irange::irange_set_anti_range): Same.
19734 (irange::set): Rename normalize_min_max to normalize_kind.
19735 (irange::verify_range): Adjust for multi-ranges having the
19737 (irange::irange_union): Set m_kind.
19738 (irange::irange_intersect): Same.
19739 (irange::invert): Same.
19740 * value-range.h (irange::kind): Always return m_kind.
19741 (irange::varying_p): Rename to...
19742 (irange::varying_comptaible_p): ...this.
19743 (irange::undefined_p): Only look at m_kind.
19744 (irange::irange): Always set VR_UNDEFINED if applicable.
19745 (irange::set_undefined): Always set VR_UNDEFINED.
19746 (irange::set_varying): Always set m_kind to VR_VARYING.
19747 (irange::normalize_min_max): Rename to...
19748 (irange::normalize_kind): ...this.
19750 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
19752 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
19753 Adjust for constant_p including varying_p.
19754 * tree-vrp.c (vrp_prop::finalize): Same.
19755 (determine_value_range): Same.
19756 * vr-values.c (vr_values::range_of_expr): Same.
19757 * value-range.cc (irange::symbolic_p): Do not check varying_p.
19758 (irange::constant_p): Same.
19760 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
19762 * value-range.cc (irange::legacy_lower_bound): Replace
19763 !undefined_p check with num_ranges > 0.
19764 (irange::legacy_upper_bound): Same.
19765 * value-range.h (irange::type): Same.
19766 (irange::lower_bound): Same.
19767 (irange::upper_bound): Same.
19769 2021-04-26 Richard Biener <rguenther@suse.de>
19771 PR tree-optimization/99956
19772 * gimple-loop-interchange.cc (compute_access_stride):
19773 Try instantiating the access in a shallower loop nest
19774 if instantiating failed.
19775 (compute_access_strides): Pass adjustable loop_nest
19776 to compute_access_stride.
19778 2021-04-26 Christophe Lyon <christophe.lyon@linaro.org>
19780 * doc/sourcebuild.texi (arm_cmse_hw): Document.
19782 2021-04-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19784 * config/aarch64/iterators.md (vwcore): Handle V4BF, V8BF.
19786 2021-04-26 Thomas Schwinge <thomas@codesourcery.com>
19787 Nathan Sidwell <nathan@codesourcery.com>
19788 Tom de Vries <vries@codesourcery.com>
19789 Julian Brown <julian@codesourcery.com>
19790 Kwok Cheung Yeung <kcy@codesourcery.com>
19792 * omp-offload.c (oacc_validate_dims): Implement
19793 '-Wopenacc-parallelism'.
19794 * doc/invoke.texi (-Wopenacc-parallelism): Document.
19796 2021-04-26 Richard Biener <rguenther@suse.de>
19798 * tree-cfg.h (gimplify_build1): Remove.
19799 (gimplify_build2): Likewise.
19800 (gimplify_build3): Likewise.
19801 * tree-cfg.c (gimplify_build1): Move to tree-vect-generic.c.
19802 (gimplify_build2): Likewise.
19803 (gimplify_build3): Likewise.
19804 * tree-vect-generic.c (gimplify_build1): Move from tree-cfg.c.
19806 (gimplify_build2): Likewise.
19807 (gimplify_build3): Likewise.
19808 (tree_vec_extract): Use resimplify with following SSA edges.
19809 (expand_vector_parallel): Avoid passing NULL size/bitpos
19810 to tree_vec_extract.
19811 * expr.c (store_constructor): Deal with zero-element CTORs.
19812 * match.pd (bit_field_ref <vector CTOR>): Make sure to
19813 produce vector constants when possible.
19815 2021-04-26 Richard Biener <rguenther@suse.de>
19817 * tree-complex.c: Include gimple-fold.h.
19818 (expand_complex_addition): Use gimple_build.
19819 (expand_complex_multiplication_components): Likewise.
19820 (expand_complex_multiplication): Likewise.
19821 (expand_complex_div_straight): Likewise.
19822 (expand_complex_div_wide): Likewise.
19823 (expand_complex_division): Likewise.
19824 (expand_complex_conjugate): Likewise.
19825 (expand_complex_comparison): Likewise.
19827 2021-04-26 Richard Biener <rguenther@suse.de>
19829 * tree-ssa-phiopt.c (two_value_replacement): Remove use
19830 of legacy gimplify_buildN API.
19832 2021-04-26 Richard Biener <rguenther@suse.de>
19834 PR tree-optimization/99473
19835 * tree-ssa-phiopt.c (cond_store_replacement): Handle all
19838 2021-04-26 Richard Biener <rguenther@suse.de>
19840 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin):
19841 Use replace_call_with_value.
19843 2021-04-26 Richard Biener <rguenther@suse.de>
19845 * tree-ssa-propagate.h (valid_gimple_rhs_p): Remove.
19846 (update_gimple_call): Likewise.
19847 (update_call_from_tree): Likewise.
19848 * tree-ssa-propagate.c (valid_gimple_rhs_p): Remove.
19849 (valid_gimple_call_p): Likewise.
19850 (move_ssa_defining_stmt_for_defs): Likewise.
19851 (finish_update_gimple_call): Likewise.
19852 (update_gimple_call): Likewise.
19853 (update_call_from_tree): Likewise.
19854 (propagate_tree_value_into_stmt): Use replace_call_with_value.
19855 * gimple-fold.h (update_gimple_call): Declare.
19856 * gimple-fold.c (valid_gimple_rhs_p): Move here from
19857 tree-ssa-propagate.c.
19858 (update_gimple_call): Likewise.
19859 (valid_gimple_call_p): Likewise.
19860 (finish_update_gimple_call): Likewise, and simplify.
19861 (gimplify_and_update_call_from_tree): Implement
19862 update_call_from_tree functionality, avoid excessive
19863 push/pop_gimplify_context.
19864 (gimple_fold_builtin): Use only gimplify_and_update_call_from_tree.
19865 (gimple_fold_call): Likewise.
19866 * gimple-ssa-sprintf.c (try_substitute_return_value): Likewise.
19867 * tree-ssa-ccp.c (ccp_folder::fold_stmt): Likewise.
19868 (pass_fold_builtins::execute): Likewise.
19869 (optimize_stack_restore): Use replace_call_with_value.
19870 * tree-cfg.c (fold_loop_internal_call): Likewise.
19871 * tree-ssa-dce.c (maybe_optimize_arith_overflow): Use
19872 only gimplify_and_update_call_from_tree.
19873 * tree-ssa-strlen.c (handle_builtin_strlen): Likewise.
19874 (handle_builtin_strchr): Likewise.
19875 * tsan.c: Include gimple-fold.h instead of tree-ssa-propagate.h.
19877 2021-04-26 Jakub Jelinek <jakub@redhat.com>
19880 * vmsdbgout.c (ASM_OUTPUT_DEBUG_STRING, vmsdbgout_begin_block,
19881 vmsdbgout_end_block, lookup_filename, vmsdbgout_source_line): Remove
19884 2021-04-25 liuhongt <hongtao.liu@intel.com>
19887 * config/i386/i386-builtin.def (BDESC): Change the icode of
19888 the following builtins to CODE_FOR_nothing.
19889 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
19890 IX86_BUILTIN_PCMPEQB128, IX86_BUILTIN_PCMPEQW128,
19891 IX86_BUILTIN_PCMPEQD128, IX86_BUILTIN_PCMPEQQ,
19892 IX86_BUILTIN_PCMPEQB256, IX86_BUILTIN_PCMPEQW256,
19893 IX86_BUILTIN_PCMPEQD256, IX86_BUILTIN_PCMPEQQ256,
19894 IX86_BUILTIN_PCMPGTB128, IX86_BUILTIN_PCMPGTW128,
19895 IX86_BUILTIN_PCMPGTD128, IX86_BUILTIN_PCMPGTQ,
19896 IX86_BUILTIN_PCMPGTB256, IX86_BUILTIN_PCMPGTW256,
19897 IX86_BUILTIN_PCMPGTD256, IX86_BUILTIN_PCMPGTQ256.
19898 * config/i386/sse.md (avx2_eq<mode>3): Deleted.
19899 (sse2_eq<mode>3): Ditto.
19900 (sse4_1_eqv2di3): Ditto.
19901 (sse2_gt<mode>3): Rename to ..
19902 (*sse2_gt<mode>3): .. this.
19904 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
19907 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
19910 * config/darwin.c (darwin_binds_local_p): Assume that any
19911 public symbol might be interposed for PIC code. Update function
19912 header comment to reflect current Darwin capability.
19914 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
19917 * config/darwin.c (darwin_binds_local_p): Assume that any
19918 public symbol might be interposed for PIC code. Update function
19919 header comment to reflect current Darwin capability.
19921 2021-04-24 Richard Sandiford <richard.sandiford@arm.com>
19923 * doc/sourcebuild.texi: Document no-opts and any-opts target
19926 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
19928 * config/bpf/bpf.h (ASM_OUTPUT_ALIGNED_BSS): Use .type and .lcomm.
19930 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
19932 * config/bpf/bpf.h (FUNCTION_BOUNDARY): Set to 64.
19934 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
19937 * config/i386/i386-options.c (ix86_option_override_internal):
19938 Error out when -m96bit-long-double is used with 64bit targets.
19939 * config/i386/i386.md (*pushxf_rounded): Remove pattern.
19941 2021-04-23 Martin Liska <mliska@suse.cz>
19943 * lto-wrapper.c: Remove FIXME about usage of
19944 hardware_concurrency. The function is not on par with
19947 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
19950 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
19951 Copy operand 3 to operand 4. Use sse_reg_operand
19952 as operand 3 predicate.
19953 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): Ditto.
19954 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
19955 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
19956 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2):
19957 Copy operand 1 to operand 0.
19958 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): Ditto.
19959 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
19960 (LDX_ATOMIC/LDX_ATOMIC FP store peephole2 with mem blockage): Ditto.
19962 2021-04-23 Alex Coplan <alex.coplan@arm.com>
19964 PR rtl-optimization/100230
19965 * early-remat.c (early_remat::sort_candidates): Use delete[]
19966 instead of delete for array allocated with new[].
19968 2021-04-23 Richard Biener <rguenther@suse.de>
19970 * genmatch.c (lower_cond): Remove VEC_COND_EXPR special-casing.
19971 (capture_info::capture_info): Likewise.
19972 (capture_info::walk_match): Likewise.
19973 (expr::gen_transform): Likewise.
19974 (dt_simplify::gen_1): Likewise.
19975 * gimple-match-head.c (maybe_resimplify_conditional_op):
19976 Remove VEC_COND_EXPR special-casing.
19977 (gimple_simplify): Likewise.
19978 * gimple.c (gimple_could_trap_p_1): Adjust.
19979 * tree-ssa-pre.c (compute_avail): Allow VEC_COND_EXPR
19980 to participate in PRE.
19982 2021-04-23 Richard Biener <rguenther@suse.de>
19984 * cfganal.c (connect_infinite_loops_to_exit): First call
19985 add_noreturn_fake_exit_edges.
19986 * ipa-sra.c (process_scan_results): Do not call the now redundant
19987 add_noreturn_fake_exit_edges.
19988 * predict.c (tree_estimate_probability): Likewise.
19989 (rebuild_frequencies): Likewise.
19990 * store-motion.c (one_store_motion_pass): Likewise.
19992 2021-04-23 Richard Biener <rguenther@suse.de>
19994 PR tree-optimization/100222
19995 * predict.c (pass_profile::execute): Remove redundant call to
19996 mark_irreducible_loops.
19997 (report_predictor_hitrates): Likewise.
19999 2021-04-23 Richard Biener <rguenther@suse.de>
20001 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Avoid
20002 valid_gimple_rhs_p by instead gimplifying to one.
20004 2021-04-23 Richard Biener <rguenther@suse.de>
20006 PR tree-optimization/99971
20007 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
20008 Always use TBAA for loads.
20010 2021-04-23 liuhongt <hongtao.liu@intel.com>
20013 * config/i386/i386-options.c (ix86_option_override_internal):
20014 Clear MASK_AVX256_SPLIT_UNALIGNED_LOAD/STORE in x_target_flags
20015 when X86_TUNE_AVX256_UNALIGNED_LOAD/STORE_OPTIMAL is enabled
20016 by target attribute.
20018 2021-04-23 David Edelsohn <dje.gcc@gmail.com>
20020 * config/rs6000/aix71.h (PREFERRED_DEBUGGING_TYPE): Change to
20022 * config/rs6000/aix72.h (PREFERRED_DEBUGGING_TYPE): Same.
20024 2021-04-22 David Edelsohn <dje.gcc@gmail.com>
20026 * config.gcc (powerpc-ibm-aix6.*): Remove.
20027 * config/rs6000/aix61.h: Delete.
20029 2021-04-22 Martin Liska <mliska@suse.cz>
20031 PR testsuite/100159
20032 PR testsuite/100192
20033 * builtins.c (expand_builtin): Fix typos and missing comments.
20034 * dwarf2out.c (gen_subprogram_die): Likewise.
20035 (gen_struct_or_union_type_die): Likewise.
20037 2021-04-22 Uroš Bizjak <ubizjak@gmail.com>
20040 * config/i386/i386-expand.c (ix86_expand_convert_uns_sidf_sse):
20041 Remove the sign with FE_DOWNWARD, where x - x = -0.0.
20043 2021-04-21 Iain Sandoe <iain@sandoe.co.uk>
20045 * config/i386/darwin.h (TARGET_64BIT): Remove definition
20046 based on TARGET_ISA_64BIT.
20047 (TARGET_64BIT_P): Remove definition based on
20048 TARGET_ISA_64BIT_P().
20050 2021-04-21 Martin Liska <mliska@suse.cz>
20053 2021-04-21 Martin Liska <mliska@suse.cz>
20055 * lto-wrapper.c (cpuset_popcount): Remove.
20056 (init_num_threads): Remove and use hardware_concurrency.
20058 2021-04-21 Martin Liska <mliska@suse.cz>
20061 * main.c (main): Call toplev::finalize in CHECKING_P mode.
20062 * ipa-modref.c (ipa_modref_c_finalize): summaries are NULL
20063 when incremental LTO linking happens.
20065 2021-04-21 Martin Liska <mliska@suse.cz>
20067 * lto-wrapper.c (run_gcc): When -flto=jobserver is used, but the
20068 makeserver cannot be detected, then use -flto=N fallback.
20070 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
20072 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): When cross-compiling,
20073 default to yes for aarch64-linux-gnu.
20074 * configure: Regenerate.
20076 2021-04-21 Martin Liska <mliska@suse.cz>
20078 * lto-wrapper.c (cpuset_popcount): Remove.
20079 (init_num_threads): Remove and use hardware_concurrency.
20081 2021-04-21 Martin Liska <mliska@suse.cz>
20083 * config/i386/i386.c: Remove superfluous || TARGET_MACHO
20084 which remains to be '(... || 0)' and clang complains about it.
20085 * dwarf2out.c (AT_vms_delta): Declare conditionally.
20086 (add_AT_vms_delta): Likewise.
20087 * tree.c (fld_simplified_type): Use rather more common pattern
20088 for disabling of something (#if 0).
20089 (get_tree_code_name): Likewise.
20090 (verify_type_variant): Likewise.
20092 2021-04-21 Martin Liska <mliska@suse.cz>
20094 * config/i386/i386-expand.c (decide_alignment): Use newly named
20095 macro TARGET_CPU_P.
20096 * config/i386/i386.c (ix86_decompose_address): Likewise.
20097 (ix86_address_cost): Likewise.
20098 (ix86_lea_outperforms): Likewise.
20099 (ix86_avoid_lea_for_addr): Likewise.
20100 (ix86_add_stmt_cost): Likewise.
20101 * config/i386/i386.h (TARGET_*): Remove.
20102 (TARGET_CPU_P): New macro.
20103 * config/i386/i386.md: Use newly named macro TARGET_CPU_P.
20104 * config/i386/x86-tune-sched-atom.c (do_reorder_for_imul): Likewise.
20105 (swap_top_of_ready_list): Likewise.
20106 (ix86_atom_sched_reorder): Likewise.
20107 * config/i386/x86-tune-sched-bd.c (ix86_bd_has_dispatch): Likewise.
20108 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Likewise.
20110 2021-04-21 Martin Liska <mliska@suse.cz>
20112 * config/i386/i386-options.c (TARGET_EXPLICIT_NO_SAHF_P):
20114 (SET_TARGET_NO_SAHF): Likewise.
20115 (TARGET_EXPLICIT_PREFETCH_SSE_P): Likewise.
20116 (SET_TARGET_PREFETCH_SSE): Likewise.
20117 (TARGET_EXPLICIT_NO_TUNE_P): Likewise.
20118 (SET_TARGET_NO_TUNE): Likewise.
20119 (TARGET_EXPLICIT_NO_80387_P): Likewise.
20120 (SET_TARGET_NO_80387): Likewise.
20122 * config/i386/i386.h (TARGET_*): Remove.
20123 * opth-gen.awk: Generate new used macros.
20125 2021-04-21 Martin Liska <mliska@suse.cz>
20127 * config/i386/i386.h (PTA_*): Remove.
20128 (enum pta_flag): New.
20129 (DEF_PTA): Generate PTA_* values from i386-isa.def.
20130 * config/i386/i386-isa.def: New file.
20132 2021-04-21 Alex Coplan <alex.coplan@arm.com>
20135 * config/aarch64/aarch64-bti-insert.c (aarch64_bti_j_insn_p): New.
20136 (rest_of_insert_bti): Avoid inserting duplicate bti j insns for
20137 jump table targets.
20139 2021-04-21 H.J. Lu <hjl.tools@gmail.com>
20141 * config.gcc: Install mwaitintrin.h for i[34567]86-*-* and
20142 x86_64-*-* targets.
20143 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_MWAIT_SET):
20145 (OPTION_MASK_ISA2_MWAIT_UNSET): Likewise.
20146 (ix86_handle_option): Handle -mmwait.
20147 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
20148 Replace OPTION_MASK_ISA_SSE3 with OPTION_MASK_ISA2_MWAIT on
20149 __builtin_ia32_monitor and __builtin_ia32_mwait.
20150 * config/i386/i386-options.c (isa2_opts): Add -mmwait.
20151 (ix86_valid_target_attribute_inner_p): Likewise.
20152 (ix86_option_override_internal): Enable mwait/monitor
20153 instructions for -msse3.
20154 * config/i386/i386.h (TARGET_MWAIT): New.
20155 (TARGET_MWAIT_P): Likewise.
20156 * config/i386/i386.opt: Add -mmwait.
20157 * config/i386/mwaitintrin.h: New file.
20158 * config/i386/pmmintrin.h: Include <mwaitintrin.h>.
20159 * config/i386/sse.md (sse3_mwait): Replace TARGET_SSE3 with
20161 (@sse3_monitor_<mode>): Likewise.
20162 * config/i386/x86gprintrin.h: Include <mwaitintrin.h>.
20163 * doc/extend.texi: Document mwait target attribute.
20164 * doc/invoke.texi: Document -mmwait.
20166 2021-04-21 Martin Liska <mliska@suse.cz>
20168 * config/i386/i386-options.c (DEF_ENUM): Remove it.
20169 * config/i386/i386-opts.h (DEF_ENUM): Likewise.
20170 * config/i386/stringop.def (DEF_ENUM): Likewise.
20172 2021-04-21 Martin Liska <mliska@suse.cz>
20174 * tree-cfg.c (gimple_verify_flow_info): Use qD instead
20175 of print_generic_expr.
20177 2021-04-21 Jakub Jelinek <jakub@redhat.com>
20179 PR rtl-optimization/100148
20180 * cprop.c (constprop_register): Use next_nondebug_insn instead of
20183 2021-04-21 Martin Liska <mliska@suse.cz>
20186 * cgraphunit.c (cgraph_node::analyze): Remove duplicate
20187 free_dominance_info calls.
20189 2021-04-21 Richard Biener <rguenther@suse.de>
20191 * gimple-fold.c (maybe_fold_reference): Remove is_lhs
20192 parameter (and assume it to be false).
20193 (fold_gimple_assign): Adjust, remove all callers of
20194 maybe_fold_reference calling it with is_lhs true.
20195 (gimple_fold_call): Likewise.
20196 (fold_stmt_1): Likewise.
20198 2021-04-21 Richard Biener <rguenther@suse.de>
20200 * fold-const.c (pedantic_non_lvalue_loc): Remove.
20201 (fold_binary_loc): Adjust.
20202 (fold_ternary_loc): Likewise.
20204 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
20206 PR middle-end/100130
20207 * varasm.c (get_block_for_decl): Make sure that any use of the
20208 retain attribute matches the section's retain flag.
20209 (switch_to_section): Check for retain mismatches even when
20210 changing sections, but do not warn if the given decl is the
20211 section's named.decl.
20212 (output_object_block): Pass the first decl in the block (if any)
20213 to switch_to_section.
20215 2021-04-20 H.J. Lu <hjl.tools@gmail.com>
20217 * config/i386/i386-c.c (ix86_target_macros_internal): Define
20218 __CRC32__ for -mcrc32.
20219 * config/i386/i386-options.c (ix86_option_override_internal):
20220 Enable crc32 instruction for -msse4.2.
20221 * config/i386/i386.md (sse4_2_crc32<mode>): Remove TARGET_SSE4_2
20223 (sse4_2_crc32di): Likewise.
20224 * config/i386/ia32intrin.h: Use crc32 target option for CRC32
20227 2021-04-20 Segher Boessenkool <segher@kernel.crashing.org>
20230 * config/rs6000/rs6000.c (rs6000_machine_from_flags): Do not consider
20233 2021-04-20 Martin Liska <mliska@suse.cz>
20235 * doc/invoke.texi: Fix typo.
20236 * params.opt: Likewise.
20238 2021-04-20 Martin Liska <mliska@suse.cz>
20240 * doc/invoke.texi: Document new param.
20242 2021-04-19 Andrew MacLeod <amacleod@redhat.com>
20244 PR tree-optimization/100081
20245 * gimple-range-cache.h (ranger_cache): Inherit from gori_compute
20246 rather than gori_compute_cache.
20247 * gimple-range-gori.cc (is_gimple_logical_p): Move to top of file.
20248 (range_def_chain::m_logical_depth): New member.
20249 (range_def_chain::range_def_chain): Initialize m_logical_depth.
20250 (range_def_chain::get_def_chain): Don't build defchains through more
20251 than LOGICAL_LIMIT logical expressions.
20252 * params.opt (param_ranger_logical_depth): New.
20254 2021-04-19 Richard Earnshaw <rearnsha@arm.com>
20257 * config/arm/arm.c (arm_configure_build_target): Do not strip
20258 extended FPU/SIMD feature bits from the target ISA when -mfpu
20259 is specified (partial revert of r11-8168).
20261 2021-04-19 Thomas Schwinge <thomas@codesourcery.com>
20263 * params.opt (-param=openacc-kernels=): Add.
20264 * omp-oacc-kernels-decompose.cc
20265 (pass_omp_oacc_kernels_decompose::gate): Use it.
20266 * doc/invoke.texi (-fopenacc-kernels=@var{mode}): Move...
20267 (--param): ... here, 'openacc-kernels'.
20269 2021-04-19 Martin Liska <mliska@suse.cz>
20272 * gengtype.c (finish_root_table): Align function arguments
20273 in between declaration and definition.
20275 2021-04-19 Eric Botcazou <ebotcazou@adacore.com>
20277 * config/i386/winnt.c (i386_pe_seh_cold_init): Properly deal with
20278 frames larger than the SEH maximum frame size.
20280 2021-04-18 Segher Boessenkool <segher@kernel.crashing.org>
20282 PR rtl-optimization/99927
20283 * combine.c (distribute_notes) [REG_UNUSED]: If the register already
20284 is dead, just drop it.
20286 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
20289 * config/i386/winnt-d.c (TARGET_D_TEMPLATES_ALWAYS_COMDAT): Define.
20290 * doc/tm.texi: Regenerate.
20291 * doc/tm.texi.in (D language and ABI): Add @hook for
20292 TARGET_D_TEMPLATES_ALWAYS_COMDAT.
20294 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
20296 * config/darwin-d.c (darwin_d_handle_target_object_format): New
20298 (darwin_d_register_target_info): New function.
20299 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
20300 * config/dragonfly-d.c (dragonfly_d_handle_target_object_format): New
20302 (dragonfly_d_register_target_info): New function.
20303 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
20304 * config/freebsd-d.c (freebsd_d_handle_target_object_format): New
20306 (freebsd_d_register_target_info): New function.
20307 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
20308 * config/glibc-d.c (glibc_d_handle_target_object_format): New
20310 (glibc_d_register_target_info): New function.
20311 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
20312 * config/i386/i386-d.c (ix86_d_handle_target_object_format): New
20314 (ix86_d_register_target_info): Add ix86_d_handle_target_object_format
20315 as handler for objectFormat key.
20316 * config/i386/winnt-d.c (winnt_d_handle_target_object_format): New
20318 (winnt_d_register_target_info): New function.
20319 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
20320 * config/netbsd-d.c (netbsd_d_handle_target_object_format): New
20322 (netbsd_d_register_target_info): New function.
20323 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
20324 * config/openbsd-d.c (openbsd_d_handle_target_object_format): New
20326 (openbsd_d_register_target_info): New function.
20327 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
20328 * config/pa/pa-d.c (pa_d_handle_target_object_format): New function.
20329 (pa_d_register_target_info): Add pa_d_handle_target_object_format as
20330 handler for objectFormat key.
20331 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_object_format): New
20333 (rs6000_d_register_target_info): Add
20334 rs6000_d_handle_target_object_format as handler for objectFormat key.
20335 * config/sol2-d.c (solaris_d_handle_target_object_format): New
20337 (solaris_d_register_target_info): New function.
20338 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
20340 2021-04-16 Jakub Jelinek <jakub@redhat.com>
20343 * config/aarch64/aarch64.c (aarch64_function_arg_alignment): Change
20344 abi_break argument from bool * to unsigned *, store there the pre-GCC 9
20346 (aarch64_layout_arg, aarch64_gimplify_va_arg_expr): Adjust callers.
20347 (aarch64_function_arg_regno_p): Likewise. Only emit -Wpsabi note if
20348 the old and new alignment after applying MIN/MAX to it is different.
20350 2021-04-16 Tamar Christina <tamar.christina@arm.com>
20353 * config/aarch64/aarch64-sve.md (@aarch64_sve_trn1_conv<mode>): New.
20354 * config/aarch64/aarch64.c (aarch64_expand_sve_const_pred_trn): Use new
20356 * config/aarch64/iterators.md (UNSPEC_TRN1_CONV): New.
20358 2021-04-16 Bill Schmidt <wschmidt@linux.ibm.com>
20360 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Revise
20361 this section and its subsections.
20363 2021-04-16 Jakub Jelinek <jakub@redhat.com>
20366 * config/aarch64/aarch64.md (*neg_asr_si2_extr, *extrsi5_insn_di): New
20367 define_insn patterns.
20369 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
20371 PR rtl-optimization/98689
20372 * reg-notes.def (UNTYPED_CALL): New note.
20373 * combine.c (distribute_notes): Handle it.
20374 * emit-rtl.c (try_split): Likewise.
20375 * rtlanal.c (rtx_properties::try_to_add_insn): Likewise. Assume
20376 that calls with the note implicitly set all return value registers.
20377 * builtins.c (expand_builtin_apply): Add a REG_UNTYPED_CALL
20380 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
20382 PR rtl-optimization/99596
20383 * rtlanal.c (rtx_properties::try_to_add_insn): Don't add global
20384 register accesses for const calls. Assume that pure functions
20385 can only read from global registers. Ignore cases in which
20386 the stack pointer has been marked global.
20388 2021-04-16 Jakub Jelinek <jakub@redhat.com>
20391 * tree-vect-loop.c (vect_transform_loop): Don't remove just
20392 dead scalar .MASK_LOAD calls, but also dead .COND_* calls - replace
20393 them by their last argument.
20395 2021-04-15 Martin Liska <mliska@suse.cz>
20397 * doc/invoke.texi: Other params don't use it, remove it.
20399 2021-04-15 Richard Biener <rguenther@suse.de>
20401 * gimple-builder.h: Add deprecation note.
20403 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
20406 * attribs.h (restrict_type_identity_attributes_to): Declare.
20407 * attribs.c (restrict_type_identity_attributes_to): New function.
20409 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
20412 * attribs.h (affects_type_identity_attributes): Declare.
20413 * attribs.c (remove_attributes_matching): New function.
20414 (affects_type_identity_attributes): Likewise.
20416 2021-04-15 Jakub Jelinek <jakub@redhat.com>
20419 * config/aarch64/aarch64.md (*<LOGICAL:optab>_<SHIFT:optab><mode>3):
20420 Add combine splitters for *<LOGICAL:optab>_ashl<mode>3 with
20421 ZERO_EXTEND, SIGN_EXTEND or AND.
20423 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
20425 PR rtl-optimization/99929
20426 * rtl.h (same_vector_encodings_p): New function.
20427 * cse.c (exp_equiv_p): Check that CONST_VECTORs have the same encoding.
20428 * cselib.c (rtx_equal_for_cselib_1): Likewise.
20429 * jump.c (rtx_renumbered_equal_p): Likewise.
20430 * lra-constraints.c (operands_match_p): Likewise.
20431 * reload.c (operands_match_p): Likewise.
20432 * rtl.c (rtx_equal_p_cb, rtx_equal_p): Likewise.
20434 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
20436 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Print
20437 more information about variable-length CONST_VECTORs.
20439 2021-04-14 Vladimir N. Makarov <vmakarov@redhat.com>
20441 PR rtl-optimization/100066
20442 * lra-constraints.c (split_reg): Check paradoxical_subreg_p for
20443 ordered modes when choosing splitting mode for hard reg.
20445 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
20448 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector_sel):
20450 (aarch64_expand_sve_const_vector): Use it for nelts_per_pattern==2.
20452 2021-04-14 Andreas Krebbel <krebbel@linux.ibm.com>
20454 * config/s390/s390-builtins.def (O_M5, O_M12, ...): Add new macros
20455 for mask operand types.
20456 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
20457 (s390_vec_permi_dbl, s390_vpdi): Use the M5 type for the immediate
20459 (s390_vec_msum_u128, s390_vmslg): Use the M12 type for the
20461 * config/s390/s390.c (s390_const_operand_ok): Check the new
20462 operand types and generate a list of valid values.
20464 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
20466 * doc/tm.texi: Regenerate.
20467 * doc/tm.texi.in (D language and ABI): Add @hook for
20468 TARGET_D_REGISTER_OS_TARGET_INFO.
20470 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
20472 * config/aarch64/aarch64-d.c (aarch64_d_handle_target_float_abi): New
20474 (aarch64_d_register_target_info): New function.
20475 * config/aarch64/aarch64-protos.h (aarch64_d_register_target_info):
20477 * config/aarch64/aarch64.h (TARGET_D_REGISTER_CPU_TARGET_INFO):
20479 * config/arm/arm-d.c (arm_d_handle_target_float_abi): New function.
20480 (arm_d_register_target_info): New function.
20481 * config/arm/arm-protos.h (arm_d_register_target_info): Declare.
20482 * config/arm/arm.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
20483 * config/i386/i386-d.c (ix86_d_handle_target_float_abi): New function.
20484 (ix86_d_register_target_info): New function.
20485 * config/i386/i386-protos.h (ix86_d_register_target_info): Declare.
20486 * config/i386/i386.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
20487 * config/mips/mips-d.c (mips_d_handle_target_float_abi): New function.
20488 (mips_d_register_target_info): New function.
20489 * config/mips/mips-protos.h (mips_d_register_target_info): Declare.
20490 * config/mips/mips.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
20491 * config/pa/pa-d.c (pa_d_handle_target_float_abi): New function.
20492 (pa_d_register_target_info): New function.
20493 * config/pa/pa-protos.h (pa_d_register_target_info): Declare.
20494 * config/pa/pa.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
20495 * config/riscv/riscv-d.c (riscv_d_handle_target_float_abi): New
20497 (riscv_d_register_target_info): New function.
20498 * config/riscv/riscv-protos.h (riscv_d_register_target_info): Declare.
20499 * config/riscv/riscv.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
20500 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_float_abi): New
20502 (rs6000_d_register_target_info): New function.
20503 * config/rs6000/rs6000-protos.h (rs6000_d_register_target_info):
20505 * config/rs6000/rs6000.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
20506 * config/s390/s390-d.c (s390_d_handle_target_float_abi): New function.
20507 (s390_d_register_target_info): New function.
20508 * config/s390/s390-protos.h (s390_d_register_target_info): Declare.
20509 * config/s390/s390.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
20510 * config/sparc/sparc-d.c (sparc_d_handle_target_float_abi): New
20512 (sparc_d_register_target_info): New function.
20513 * config/sparc/sparc-protos.h (sparc_d_register_target_info): Declare.
20514 * config/sparc/sparc.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
20515 * doc/tm.texi: Regenerate.
20516 * doc/tm.texi.in (D language and ABI): Add @hook for
20517 TARGET_D_REGISTER_CPU_TARGET_INFO.
20519 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
20521 * config/i386/i386-d.c (ix86_d_has_stdcall_convention): New function.
20522 * config/i386/i386-protos.h (ix86_d_has_stdcall_convention): Declare.
20523 * config/i386/i386.h (TARGET_D_HAS_STDCALL_CONVENTION): Define.
20524 * doc/tm.texi: Regenerate.
20525 * doc/tm.texi.in (D language and ABI): Add @hook for
20526 TARGET_D_HAS_STDCALL_CONVENTION.
20528 2021-04-14 Richard Biener <rguenther@suse.de>
20530 * tree-cfg.c (verify_gimple_assign_ternary): Verify that
20531 VEC_COND_EXPRs have a gimple_val condition.
20532 * tree-ssa-propagate.c (valid_gimple_rhs_p): VEC_COND_EXPR
20533 can no longer have a GENERIC condition.
20535 2021-04-14 Richard Earnshaw <rearnsha@arm.com>
20538 * config/arm/arm.c (arm_configure_build_target): Strip isa_all_fpbits
20539 from the isa_delta when -mfpu has been used.
20540 (arm_options_perform_arch_sanity_checks): It's the architecture that
20541 lacks an FPU not the processor.
20543 2021-04-13 Richard Biener <rguenther@suse.de>
20545 PR tree-optimization/100053
20546 * tree-ssa-sccvn.c (vn_nary_op_get_predicated_value): Do
20547 not use optimistic dominance queries for backedges to validate
20549 (dominated_by_p_w_unex): Add parameter to ignore executable
20550 state on backedges.
20551 (rpo_elim::eliminate_avail): Adjust.
20553 2021-04-13 Jakub Jelinek <jakub@redhat.com>
20556 * config/aarch64/aarch64.md (*aarch64_bfxil<mode>_extr,
20557 *aarch64_bfxilsi_extrdi): New define_insn patterns.
20559 2021-04-13 Jakub Jelinek <jakub@redhat.com>
20562 * simplify-rtx.c (simplify_immed_subreg): For MODE_COMPOSITE_P
20563 outermode, return NULL if the result doesn't encode back to the
20564 original byte sequence.
20565 (simplify_gen_subreg): Don't create SUBREGs from constants to
20566 MODE_COMPOSITE_P outermode.
20568 2021-04-12 Jakub Jelinek <jakub@redhat.com>
20570 PR rtl-optimization/99905
20571 * combine.c (expand_compound_operation): If pos + len > modewidth,
20572 perform the right shift by pos in inner_mode and then convert to mode,
20573 instead of trying to simplify a shift of rtx with inner_mode by pos
20574 as if it was a shift in mode.
20576 2021-04-12 Jakub Jelinek <jakub@redhat.com>
20579 * combine.c (simplify_and_const_int_1): Don't optimize varop
20580 away if it has side-effects.
20582 2021-04-12 Martin Liska <mliska@suse.cz>
20584 * doc/extend.texi: Escape @smallexample content.
20586 2021-04-12 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
20588 * config/s390/s390.md ("*movdi_31", "*movdi_64"): Add
20589 alternative in order to load a DFP zero.
20591 2021-04-12 Martin Liska <mliska@suse.cz>
20593 * doc/extend.texi: Be more precise in documentation
20594 of symver attribute.
20596 2021-04-12 Martin Liska <mliska@suse.cz>
20599 * gimplify.c (gimplify_expr): Right now, we unpoison all
20600 variables before a goto <dest>. We should not do it if we are
20603 2021-04-12 Cui,Lili <lili.cui@intel.com>
20605 * common/config/i386/cpuinfo.h (get_intel_cpu): Handle
20607 * common/config/i386/i386-common.c (processor_names): Add
20609 (processor_alias_table): Add rocketlake.
20610 * common/config/i386/i386-cpuinfo.h (processor_subtypes): Add
20611 INTEL_COREI7_ROCKETLAKE.
20612 * config.gcc: Add -march=rocketlake.
20613 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20615 * config/i386/i386-options.c (m_ROCKETLAKE) : Define.
20616 (processor_cost_table): Add rocketlake cost.
20617 * config/i386/i386.h (ix86_size_cost) : Define
20619 (processor_type) : Add PROCESSOR_ROCKETLAKE.
20620 (PTA_ROCKETLAKE): Ditto.
20621 * doc/extend.texi: Add rocketlake.
20622 * doc/invoke.texi: Add rocketlake.
20624 2021-04-12 Cui,Lili <lili.cui@intel.com>
20626 * config/i386/i386.h (PTA_ALDERLAKE): Change alderlake ISA list.
20627 * config/i386/i386-options.c (m_CORE_AVX2): Add m_ALDERLAKE.
20628 * common/config/i386/cpuinfo.h (get_intel_cpu): Add AlderLake model.
20629 * doc/invoke.texi: Change alderlake ISA list.
20631 2021-04-11 Hafiz Abid Qadeer <abidh@codesourcery.com>
20633 PR middle-end/98088
20634 * omp-expand.c (expand_oacc_collapse_init): Update condition in
20637 2021-04-10 H.J. Lu <hjl.tools@gmail.com>
20640 * config/i386/serializeintrin.h (_serialize): Defined as macro.
20642 2021-04-10 Jakub Jelinek <jakub@redhat.com>
20645 * expr.c (expand_expr_addr_expr_1): Test is_global_var rather than
20646 just TREE_STATIC on COMPOUND_LITERAL_EXPR_DECLs.
20648 2021-04-10 Jakub Jelinek <jakub@redhat.com>
20650 PR middle-end/99989
20651 * gimple-ssa-warn-alloca.c
20652 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit to
20653 0 with integer precision unconditionally.
20655 2021-04-10 Jakub Jelinek <jakub@redhat.com>
20657 PR rtl-optimization/98601
20658 * rtlanal.c (rtx_addr_can_trap_p_1): Allow in assert unknown size
20659 not just for BLKmode, but also for VOIDmode. For STRICT_ALIGNMENT
20660 unaligned_mems handle VOIDmode like BLKmode.
20662 2021-04-10 Jan Hubicka <hubicka@ucw.cz>
20665 * tree.c (free_lang_data_in_decl): Do not release body of
20666 declare_variant_alt.
20668 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
20670 * config/aarch64/aarch64.c (aarch64_option_restore): If the
20671 architecture was specified explicitly and the tuning wasn't,
20672 tune for the architecture rather than the configured default CPU.
20674 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
20676 * config/aarch64/aarch64.md (tlsdesc_small_sve_<mode>): Use X30
20677 as the temporary register.
20679 2021-04-09 Martin Liska <mliska@suse.cz>
20681 * doc/extend.texi: Move non-target attributes on the top level.
20683 2021-04-09 Martin Liska <mliska@suse.cz>
20685 * doc/invoke.texi: Document minimum and maximum value of the
20686 argument for both supported compression algorithms.
20688 2021-04-08 David Edelsohn <dje.gcc@gmail.com>
20690 * config/rs6000/rs6000.c (rs6000_xcoff_select_section): Select
20691 TLS BSS before TLS data.
20692 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .comm.
20694 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
20696 * doc/sourcebuild.texi (stdint_types_mbig_endian): Document.
20698 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
20700 * match.pd: Extend vec_cond folds to handle shifts.
20702 2021-04-08 Maciej W. Rozycki <macro@orcam.me.uk>
20704 * config/vax/vax.md: Fix comment for `*bit<mode>' pattern's
20707 2021-04-08 Alex Coplan <alex.coplan@arm.com>
20710 * config/arm/iterators.md (MVE_vecs): New.
20711 (V_elem): Also handle V2DF.
20712 * config/arm/mve.md (*mve_mov<mode>): Rename to ...
20713 (*mve_vdup<mode>): ... this. Remove second alternative since
20714 vec_duplicate of const_int is not canonical RTL, and we don't
20715 want to match symbol_refs.
20716 (*mve_vec_duplicate<mode>): Delete (pattern is redundant).
20718 2021-04-08 Xionghu Luo <luoxhu@linux.ibm.com>
20720 * fold-const.c (fold_single_bit_test): Fix typo.
20721 * print-rtl.c (print_rtx_insn_vec): Call print_rtl_single
20724 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
20726 PR tree-optimization/97513
20727 * tree-vect-slp.c (vect_add_slp_permutation): New function,
20729 (vectorizable_slp_permutation): ...here. Detect cases in which
20730 all VEC_PERM_EXPRs are guaranteed to have the same stepped
20731 permute vector and only generate one permute vector for that case.
20732 Extend that case to handle variable-length vectors.
20734 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
20736 PR tree-optimization/99873
20737 * tree-vect-slp.c (vect_slp_prefer_store_lanes_p): New function.
20738 (vect_build_slp_instance): Don't split store groups that could
20739 use IFN_STORE_LANES.
20741 2021-04-07 Jakub Jelinek <jakub@redhat.com>
20744 * varasm.c (output_constant_pool_contents): Don't strip name encoding
20745 from XSTR (desc->sym, 0) or from label before passing those to
20748 2021-04-07 Richard Biener <rguenther@suse.de>
20750 PR tree-optimization/99954
20751 * tree-loop-distribution.c: Include tree-affine.h.
20752 (generate_memcpy_builtin): Try using tree-affine to prove
20754 (loop_distribution::classify_builtin_ldst): Always classify
20757 2021-04-07 Richard Biener <rguenther@suse.de>
20759 PR tree-optimization/99947
20760 * tree-vect-loop.c (vectorizable_induction): Pre-allocate
20761 steps vector to avoid pushing elements from the reallocated
20764 2021-04-07 Richard Biener <rguenther@suse.de>
20766 * tree-ssa-sccvn.h (print_vn_reference_ops): Declare.
20767 * tree-ssa-pre.c (print_pre_expr): Factor out VN reference operand
20769 * tree-ssa-sccvn.c (print_vn_reference_ops): ... into this new
20771 (debug_vn_reference_ops): New.
20773 2021-04-07 Bin Cheng <bin.cheng@linux.alibaba.com>
20775 PR tree-optimization/98736
20776 * tree-loop-distribution.c
20777 * (loop_distribution::bb_top_order_init):
20778 Compute RPO with programing order preserved by calling function
20779 rev_post_order_and_mark_dfs_back_seme.
20781 2021-04-06 Vladimir N. Makarov <vmakarov@redhat.com>
20784 * lra-constraints.c (split_reg): Don't check paradoxical_subreg_p.
20785 * lra-lives.c (clear_sparseset_regnos, regnos_in_sparseset_p): New
20787 (process_bb_lives): Don't update biggest mode of hard reg for
20788 implicit in multi-register group. Use the new functions for
20789 updating dead_set and unused_set by register notes.
20791 2021-04-06 Xianmiao Qu <xianmiao_qu@c-sky.com>
20793 * config/csky/csky_pipeline_ck802.md : Use insn reservation name
20796 2021-04-06 H.J. Lu <hjl.tools@gmail.com>
20798 * config/i386/x86-tune-costs.h (skylake_memcpy): Updated.
20799 (skylake_memset): Likewise.
20800 (skylake_cost): Change CLEAR_RATIO to 17.
20801 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
20802 Replace m_CANNONLAKE, m_ICELAKE_CLIENT, m_ICELAKE_SERVER,
20803 m_TIGERLAKE and m_SAPPHIRERAPIDS with m_SKYLAKE and m_CORE_AVX512.
20805 2021-04-06 Richard Biener <rguenther@suse.de>
20807 PR tree-optimization/99880
20808 * tree-vect-loop.c (maybe_set_vectorized_backedge_value): Only
20809 set vectorized defs of relevant PHIs.
20811 2021-04-06 Richard Biener <rguenther@suse.de>
20813 PR tree-optimization/99924
20814 * tree-vect-slp.c (vect_bb_partition_graph_r): Do not mark
20815 nodes w/o scalar stmts as visited.
20817 2021-04-06 Alex Coplan <alex.coplan@arm.com>
20820 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Also use base
20821 PCS for [su]fix_optab.
20823 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
20825 * config/darwin.c (machopic_legitimize_pic_address): Check
20826 that the current pic register is one of the hard reg set
20827 before setting liveness.
20829 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
20831 * config/darwin.c (machopic_legitimize_pic_address): Fix
20832 whitespace, remove unused code.
20834 2021-04-03 Jakub Jelinek <jakub@redhat.com>
20836 PR tree-optimization/99882
20837 * gimple-ssa-store-merging.c (bswap_view_convert): Handle val with
20840 2021-04-03 Jakub Jelinek <jakub@redhat.com>
20842 PR rtl-optimization/99863
20843 * dse.c (replace_read): Drop regs_live argument. Instead of
20844 regs_live, use store_insn->fixed_regs_live if non-NULL,
20845 otherwise punt if insns sequence clobbers or sets any hard
20848 2021-04-03 Jakub Jelinek <jakub@redhat.com>
20851 * targhooks.h (default_print_patchable_function_entry_1): Declare.
20852 * targhooks.c (default_print_patchable_function_entry_1): New function,
20853 copied from default_print_patchable_function_entry with an added flags
20855 (default_print_patchable_function_entry): Rewritten into a small
20856 wrapper around default_print_patchable_function_entry_1.
20857 * config/rs6000/rs6000.c (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
20859 (rs6000_print_patchable_function_entry): New function.
20861 2021-04-02 Eric Botcazou <ebotcazou@adacore.com>
20863 * doc/invoke.texi (fdelete-dead-exceptions): Minor tweak.
20865 2021-04-01 Jason Merrill <jason@redhat.com>
20868 * common.opt: Document v15 and v16.
20870 2021-04-01 Richard Biener <rguenther@suse.de>
20872 PR tree-optimization/99863
20873 * gimplify.c (gimplify_init_constructor): Recompute vector
20876 2021-04-01 Jakub Jelinek <jakub@redhat.com>
20878 * doc/extend.texi (symver attribute): Fix up syntax errors
20881 2021-04-01 Jakub Jelinek <jakub@redhat.com>
20883 PR tree-optimization/96573
20884 * gimple-ssa-store-merging.c (init_symbolic_number): Handle
20885 also pointer types.
20887 2021-04-01 Richard Biener <rguenther@suse.de>
20889 PR tree-optimization/99856
20890 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Promote
20891 precision to vector element precision.
20893 2021-04-01 Martin Jambor <mjambor@suse.cz>
20895 PR tree-optimization/97009
20896 * tree-sra.c (access_or_its_child_written): New function.
20897 (propagate_subaccesses_from_rhs): Use it instead of a simple grp_write
20900 2021-03-31 Jan Hubicka <hubicka@ucw.cz>
20903 * cif-code.def (USES_COMDAT_LOCAL): Make CIF_FINAL_NORMAL.
20905 2021-03-31 Pat Haugen <pthaugen@linux.ibm.com>
20908 * config/rs6000/altivec.md (xxspltiw_v4si, xxspltiw_v4sf_inst,
20909 xxspltidp_v2df_inst, xxsplti32dx_v4si_inst, xxsplti32dx_v4sf_inst,
20910 xxblend_<mode>, xxpermx_inst, xxeval): Mark prefixed.
20911 * config/rs6000/mma.md (mma_<vvi4i4i8>, mma_<avvi4i4i8>,
20912 mma_<vvi4i4i2>, mma_<avvi4i4i2>, mma_<vvi4i4>, mma_<avvi4i4>,
20913 mma_<pvi4i2>, mma_<apvi4i2>, mma_<vvi4i4i4>, mma_<avvi4i4i4>):
20915 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Adjust test.
20916 * config/rs6000/rs6000.md (define_attr "maybe_prefixed"): New.
20917 (define_attr "prefixed"): Update initializer.
20919 2021-03-31 Jakub Jelinek <jakub@redhat.com>
20922 * dwarf2out.c (debug_ranges_dwo_section): New variable.
20923 (DW_RANGES_IDX_SKELETON): Define.
20924 (struct dw_ranges): Add begin_entry and end_entry members.
20925 (DEBUG_DWO_RNGLISTS_SECTION): Define.
20926 (add_ranges_num): Adjust r initializer for addition of *_entry
20928 (add_ranges_by_labels): For -gsplit-dwarf and force_direct,
20929 set idx to DW_RANGES_IDX_SKELETON.
20930 (use_distinct_base_address_for_range): New function.
20931 (index_rnglists): Don't set r->idx if it is equal to
20932 DW_RANGES_IDX_SKELETON. Initialize r->begin_entry and
20933 r->end_entry for -gsplit-dwarf if those will be needed by
20935 (output_rnglists): Add DWO argument. If true, switch to
20936 debug_ranges_dwo_section rather than debug_ranges_section.
20937 Adjust l1/l2 label indexes. Only output the offset table when
20938 dwo is true and don't include in there the skeleton range
20939 entry if present. For -gsplit-dwarf, skip ranges that belong
20940 to the other rnglists section. Change return type from void
20941 to bool and return true if there are any range entries for
20942 the other section. For dwarf_split_debug_info use
20943 DW_RLE_startx_endx, DW_RLE_startx_length and DW_RLE_base_addressx
20944 entries instead of DW_RLE_start_end, DW_RLE_start_length and
20945 DW_RLE_base_address. Use use_distinct_base_address_for_range.
20946 (init_sections_and_labels): Initialize debug_ranges_dwo_section
20947 if -gsplit-dwarf and DWARF >= 5. Adjust ranges_section_label
20948 and range_base_label indexes.
20949 (dwarf2out_finish): Call index_rnglists earlier before finalizing
20950 .debug_addr. Never emit DW_AT_rnglists_base attribute. For
20951 -gsplit-dwarf and DWARF >= 5 call output_rnglists up to twice
20952 with different dwo arguments.
20953 (dwarf2out_c_finalize): Clear debug_ranges_dwo_section.
20955 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
20957 PR tree-optimization/98268
20958 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Call
20959 recompute_tree_invariant_for_addr_expr after successfully
20960 folding a TARGET_MEM_REF that occurs inside an ADDR_EXPR.
20962 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
20964 PR tree-optimization/99726
20965 * tree-data-ref.c (create_intersect_range_checks_index): Bail
20966 out if there is more than one access function SCEV for the loop
20969 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
20971 PR rtl-optimization/97141
20972 PR rtl-optimization/98726
20973 * emit-rtl.c (valid_for_const_vector_p): Return true for
20975 * rtx-vector-builder.h (rtx_vector_builder::step): Return a
20976 poly_wide_int instead of a wide_int.
20977 (rtx_vector_builder::apply_set): Take a poly_wide_int instead
20979 * rtx-vector-builder.c (rtx_vector_builder::apply_set): Likewise.
20980 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Return
20981 false for CONST_VECTORs that cannot be forced to memory.
20982 * config/aarch64/aarch64-simd.md (mov<mode>): If a CONST_VECTOR
20983 is too complex to force to memory, build it up from individual
20986 2021-03-31 Jan Hubicka <jh@suse.cz>
20989 * cgraph.c (cgraph_node::release_body): Fix overactive check.
20991 2021-03-31 Christophe Lyon <christophe.lyon@linaro.org>
20994 * config/arm/vec-common.md (mul<mode>3): Disable on iwMMXT, expect
20997 2021-03-31 H.J. Lu <hjl.tools@gmail.com>
20999 * config/i386/i386-expand.c (expand_set_or_cpymem_via_rep):
21000 For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, don't convert QImode
21002 (decide_alg): For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, use
21003 "rep movsb/stosb" only for known sizes.
21004 * config/i386/i386-options.c (processor_cost_table): Use Ice
21005 Lake cost for Cannon Lake, Ice Lake, Tiger Lake, Sapphire
21006 Rapids and Alder Lake.
21007 * config/i386/i386.h (TARGET_PREFER_KNOWN_REP_MOVSB_STOSB): New.
21008 * config/i386/x86-tune-costs.h (icelake_memcpy): New.
21009 (icelake_memset): Likewise.
21010 (icelake_cost): Likewise.
21011 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
21014 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
21017 * config/aarch64/aarch64.c
21018 (aarch64_vectorize_preferred_vector_alignment): Query the size
21019 of the provided SVE vector; do not assume that all SVE vectors
21020 have the same size.
21022 2021-03-31 Jan Hubicka <jh@suse.cz>
21025 * cgraph.c (cgraph_node::release_body): Remove all callers and
21027 * cgraphclones.c (cgraph_node::materialize_clone): Do not do it here.
21028 * cgraphunit.c (cgraph_node::expand): And here.
21030 2021-03-31 Martin Liska <mliska@suse.cz>
21032 * ipa-modref.c (analyze_ssa_name_flags): Fix coding style
21033 and one negated condition.
21035 2021-03-31 Jakub Jelinek <jakub@redhat.com>
21036 Richard Sandiford <richard.sandiford@arm.com>
21039 * config/aarch64/aarch64.md (*add<mode>3_poly_1): Swap Uai and Uav
21040 constraints on operands[2] and similarly 0 and rk constraints
21041 on operands[1] corresponding to that.
21043 2021-03-31 Jakub Jelinek <jakub@redhat.com>
21046 * configure.ac (HAVE_LD_BROKEN_PE_DWARF5): New AC_DEFINE if PECOFF
21047 linker doesn't support DWARF sections new in DWARF5.
21048 * config/i386/i386-options.c (ix86_option_override_internal): Default
21049 to dwarf_version 4 if HAVE_LD_BROKEN_PE_DWARF5 for TARGET_PECOFF
21051 * config.in: Regenerated.
21052 * configure: Regenerated.
21054 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21057 * config/aarch64/aarch64.c (aarch64_analyze_loop_vinfo): Check for
21058 available issue_info before using it.
21060 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21063 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): Do not allow zero
21066 2021-03-30 Xionghu Luo <luoxhu@linux.ibm.com>
21069 * config/rs6000/altivec.md (altivec_lvsl_reg): Change to ...
21070 (altivec_lvsl_reg_<mode>): ... this.
21071 (altivec_lvsr_reg): Change to ...
21072 (altivec_lvsr_reg_<mode>): ... this.
21073 * config/rs6000/predicates.md (vec_set_index_operand): New.
21074 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21075 Enable 32bit variable vec_insert for all TARGET_VSX.
21076 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
21077 Enable 32bit variable vec_insert for p9 and above.
21078 (rs6000_expand_vector_set_var_p8): Rename to ...
21079 (rs6000_expand_vector_set_var_p7): ... this.
21080 (rs6000_expand_vector_set): Use TARGET_VSX and adjust assert
21082 * config/rs6000/vector.md (vec_set<mode>): Use vec_set_index_operand.
21083 * config/rs6000/vsx.md (xl_len_r): Use gen_altivec_lvsl_reg_di and
21084 gen_altivec_lvsr_reg_di.
21086 2021-03-30 H.J. Lu <hjl.tools@gmail.com>
21089 * config/i386/ia32intrin.h (__rdtsc): Defined as macro.
21090 (__rdtscp): Likewise.
21092 2021-03-30 Tamar Christina <tamar.christina@arm.com>
21094 PR tree-optimization/99825
21095 * tree-vect-slp-patterns.c (vect_check_evenodd_blend):
21096 Reject non-mult 2 lanes.
21098 2021-03-30 Richard Earnshaw <rearnsha@arm.com>
21101 * config/arm/arm.c (arm_file_start): Fix emission of
21102 Tag_ABI_VFP_args attribute.
21104 2021-03-30 Richard Biener <rguenther@suse.de>
21106 PR tree-optimization/99824
21107 * stor-layout.c (set_min_and_max_values_for_integral_type):
21108 Assert the precision is within the bounds of
21109 WIDE_INT_MAX_PRECISION.
21110 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
21111 the outermost component ref only to lower the access size
21112 and initialize that from the access type.
21114 2021-03-30 Richard Sandiford <richard.sandiford@arm.com>
21117 * config/aarch64/aarch64.md (mov<mode>): Pass multi-instruction
21118 CONST_INTs to aarch64_expand_mov_immediate when called after RA.
21120 2021-03-30 Mihailo Stojanovic <mihailo.stojanovic@typhoon-hil.com>
21122 * config/aarch64/aarch64.md
21123 (<optab>_trunc<fcvt_target><GPI:mode>2): Set the "arch"
21124 attribute to disambiguate between SIMD and FP variants of the
21127 2021-03-29 Jan Hubicka <hubicka@ucw.cz>
21129 * ipa-modref.c (merge_call_lhs_flags): Correct handling of deref.
21130 (analyze_ssa_name_flags): Fix typo in comment.
21132 2021-03-29 Alex Coplan <alex.coplan@arm.com>
21135 * config/aarch64/aarch64-sve-builtins.cc
21136 (function_builder::add_function): Add placeholder_p argument, use
21137 placeholder decls if this is set.
21138 (function_builder::add_unique_function): Instead of conditionally adding
21139 direct overloads, unconditionally add either a direct overload or a
21141 (function_builder::add_overloaded_function): Set placeholder_p if we're
21142 using C++ overloads. Use the obstack for string storage instead
21143 of relying on the tree nodes.
21144 (function_builder::add_overloaded_functions): Don't return early for
21145 m_direct_overloads: we need to add placeholders.
21146 * config/aarch64/aarch64-sve-builtins.h
21147 (function_builder::add_function): Add placeholder_p argument.
21149 2021-03-29 Richard Biener <rguenther@suse.de>
21151 PR tree-optimization/99807
21152 * tree-vect-slp.c (vect_slp_analyze_node_operations_1): Move
21153 assert below VEC_PERM handling.
21155 2021-03-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21158 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>): Use
21159 aarch64_simd_or_scalar_imm_zero to match zeroes. Remove pattern
21160 matching const_int 0.
21161 (move_lo_quad_internal_be_<mode>): Likewise.
21162 (move_lo_quad_<mode>): Update for the above.
21163 * config/aarch64/iterators.md (VQ_2E): Delete.
21165 2021-03-29 Jakub Jelinek <jakub@redhat.com>
21167 PR tree-optimization/99777
21168 * fold-const.c (extract_muldiv_1): For conversions, punt on casts from
21169 types other than scalar integral types.
21171 2021-03-28 David Edelsohn <dje.gcc@gmail.com>
21173 * config/rs6000/rs6000.c (rs6000_output_dwarf_dtprel): Do not add
21174 XCOFF TLS reloc decorations.
21176 2021-03-28 Gerald Pfeifer <gerald@pfeifer.com>
21178 * doc/analyzer.texi (Analyzer Internals): Update link to
21179 "A Memory Model for Static Analysis of C Programs".
21181 2021-03-26 David Edelsohn <dje.gcc@gmail.com>
21183 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Call function.
21184 * config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align):
21186 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align): New.
21187 (rs6000_special_round_type_align): Recursively check innermost first
21190 2021-03-26 Jakub Jelinek <jakub@redhat.com>
21193 * dwarf2out.h (struct dw_fde_node): Add rule18 member.
21194 * dwarf2cfi.c (dwarf2out_frame_debug_expr): When handling (set hfp sp)
21195 assignment with drap_reg active, queue reg save for hfp with offset 0
21196 and flush queued reg saves. When handling a push with rule18,
21197 defer queueing reg save for hfp and just assert the offset is 0.
21198 (scan_trace): Assert that fde->rule18 is false.
21200 2021-03-26 Vladimir Makarov <vmakarov@redhat.com>
21203 * ira-costs.c (record_reg_classes): Put case with
21204 CT_RELAXED_MEMORY adjacent to one with CT_MEMORY.
21205 * ira.c (ira_setup_alts): Ditto.
21206 * lra-constraints.c (process_alt_operands): Ditto.
21207 * recog.c (asm_operand_ok): Ditto.
21208 * reload.c (find_reloads): Ditto.
21210 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21212 * config/aarch64/aarch64-protos.h
21213 (cpu_addrcost_table::post_modify_ld3_st3): New member variable.
21214 (cpu_addrcost_table::post_modify_ld4_st4): Likewise.
21215 * config/aarch64/aarch64.c (generic_addrcost_table): Update
21216 accordingly, using the same costs as for post_modify.
21217 (exynosm1_addrcost_table, xgene1_addrcost_table): Likewise.
21218 (thunderx2t99_addrcost_table, thunderx3t110_addrcost_table):
21219 (tsv110_addrcost_table, qdf24xx_addrcost_table): Likewise.
21220 (a64fx_addrcost_table): Likewise.
21221 (neoversev1_addrcost_table): New.
21222 (neoversev1_tunings): Use neoversev1_addrcost_table.
21223 (aarch64_address_cost): Use the new post_modify costs for CImode
21226 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21228 * config/aarch64/aarch64.opt
21229 (-param=aarch64-loop-vect-issue-rate-niters=): New parameter.
21230 * doc/invoke.texi: Document it.
21231 * config/aarch64/aarch64-protos.h (aarch64_base_vec_issue_info)
21232 (aarch64_scalar_vec_issue_info, aarch64_simd_vec_issue_info)
21233 (aarch64_advsimd_vec_issue_info, aarch64_sve_vec_issue_info)
21234 (aarch64_vec_issue_info): New structures.
21235 (cpu_vector_cost): Write comments above the variables rather
21237 (cpu_vector_cost::issue_info): New member variable.
21238 * config/aarch64/aarch64.c: Include gimple-pretty-print.h
21239 and tree-ssa-loop-niter.h.
21240 (generic_vector_cost, a64fx_vector_cost, qdf24xx_vector_cost)
21241 (thunderx_vector_cost, tsv110_vector_cost, cortexa57_vector_cost)
21242 (exynosm1_vector_cost, xgene1_vector_cost, thunderx2t99_vector_cost)
21243 (thunderx3t110_vector_cost): Initialize issue_info to null.
21244 (neoversev1_scalar_issue_info, neoversev1_advsimd_issue_info)
21245 (neoversev1_sve_issue_info, neoversev1_vec_issue_info): New structures.
21246 (neoversev1_vector_cost): Use them.
21247 (aarch64_vec_op_count, aarch64_sve_op_count): New structures.
21248 (aarch64_vector_costs::saw_sve_only_op): New member variable.
21249 (aarch64_vector_costs::num_vector_iterations): Likewise.
21250 (aarch64_vector_costs::scalar_ops): Likewise.
21251 (aarch64_vector_costs::advsimd_ops): Likewise.
21252 (aarch64_vector_costs::sve_ops): Likewise.
21253 (aarch64_vector_costs::seen_loads): Likewise.
21254 (aarch64_simd_vec_costs_for_flags): New function.
21255 (aarch64_analyze_loop_vinfo): Initialize num_vector_iterations.
21256 Count the number of predicate operations required by SVE WHILE
21258 (aarch64_comparison_type, aarch64_multiply_add_p): New functions.
21259 (aarch64_sve_only_stmt_p, aarch64_in_loop_reduction_latency): Likewise.
21260 (aarch64_count_ops): Likewise.
21261 (aarch64_add_stmt_cost): Record whether see an SVE operation
21262 that cannot currently be implementing using Advanced SIMD.
21263 Record issue information about the scalar, Advanced SIMD
21264 and (where relevant) SVE versions of a loop.
21265 (aarch64_vec_op_count::dump): New function.
21266 (aarch64_sve_op_count::dump): Likewise.
21267 (aarch64_estimate_min_cycles_per_iter): Likewise.
21268 (aarch64_adjust_body_cost): If issue information is available,
21269 try to compare the issue rates of the various loop implementations
21270 and increase or decrease the vector body cost accordingly.
21272 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21274 * config/aarch64/aarch64.c (aarch64_detect_vector_stmt_subtype):
21275 Assume a zero cost for induction phis.
21277 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21279 * config/aarch64/aarch64.c (aarch64_embedded_comparison_type): New
21281 (aarch64_adjust_stmt_cost): Add the costs of embedded scalar and
21282 vector comparisons.
21284 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21286 * config/aarch64/aarch64.c (aarch64_detect_scalar_stmt_subtype):
21288 (aarch64_add_stmt_cost): Call it.
21290 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21292 * config/aarch64/aarch64-tuning-flags.def (matched_vector_throughput):
21293 New tuning parameter.
21294 * config/aarch64/aarch64.c (neoversev1_tunings): Use it.
21295 (aarch64_estimated_sve_vq): New function.
21296 (aarch64_vector_costs::analyzed_vinfo): New member variable.
21297 (aarch64_vector_costs::is_loop): Likewise.
21298 (aarch64_vector_costs::unrolled_advsimd_niters): Likewise.
21299 (aarch64_vector_costs::unrolled_advsimd_stmts): Likewise.
21300 (aarch64_record_potential_advsimd_unrolling): New function.
21301 (aarch64_analyze_loop_vinfo, aarch64_analyze_bb_vinfo): Likewise.
21302 (aarch64_add_stmt_cost): Call aarch64_analyze_loop_vinfo or
21303 aarch64_analyze_bb_vinfo on the first use of a costs structure.
21304 Detect whether we're vectorizing a loop for SVE that might be
21305 completely unrolled if it used Advanced SIMD instead.
21306 (aarch64_adjust_body_cost_for_latency): New function.
21307 (aarch64_finish_cost): Call it.
21309 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21311 * config/aarch64/aarch64.c (aarch64_vector_costs): New structure.
21312 (aarch64_init_cost): New function.
21313 (aarch64_add_stmt_cost): Use aarch64_vector_costs instead of
21314 the default unsigned[3].
21315 (aarch64_finish_cost, aarch64_destroy_cost_data): New functions.
21316 (TARGET_VECTORIZE_INIT_COST): Override.
21317 (TARGET_VECTORIZE_FINISH_COST): Likewise.
21318 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
21320 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21322 * config/aarch64/aarch64.c (neoversev1_advsimd_vector_cost)
21323 (neoversev1_sve_vector_cost): New cost structures.
21324 (neoversev1_vector_cost): Likewise.
21325 (neoversev1_tunings): Use them. Enable use_new_vector_costs.
21327 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21329 * config/aarch64/aarch64-protos.h
21330 (sve_vec_cost::scatter_store_elt_cost): New member variable.
21331 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
21332 accordingly, taking the cost from the cost of a scalar_store.
21333 (a64fx_sve_vector_cost): Likewise.
21334 (aarch64_detect_vector_stmt_subtype): Detect scatter stores.
21336 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21338 * config/aarch64/aarch64-protos.h
21339 (simd_vec_cost::store_elt_extra_cost): New member variable.
21340 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
21341 accordingly, using the vec_to_scalar cost for the new field.
21342 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
21343 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
21344 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
21345 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
21346 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
21347 (thunderx3t110_advsimd_vector_cost): Likewise.
21348 (aarch64_detect_vector_stmt_subtype): Detect single-element stores.
21350 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21352 * config/aarch64/aarch64-protos.h (simd_vec_cost::ld2_st2_permute_cost)
21353 (simd_vec_cost::ld3_st3_permute_cost): New member variables.
21354 (simd_vec_cost::ld4_st4_permute_cost): Likewise.
21355 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
21356 accordingly, using zero for the new costs.
21357 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
21358 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
21359 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
21360 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
21361 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
21362 (thunderx3t110_advsimd_vector_cost): Likewise.
21363 (aarch64_ld234_st234_vectors): New function.
21364 (aarch64_adjust_stmt_cost): Likewise.
21365 (aarch64_add_stmt_cost): Call aarch64_adjust_stmt_cost if using
21366 the new vector costs.
21368 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21370 * config/aarch64/aarch64-protos.h (sve_vec_cost): Turn into a
21371 derived class of simd_vec_cost. Add information about CLAST[AB]
21372 and FADDA instructions.
21373 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
21374 accordingly, using the vec_to_scalar costs for the new fields.
21375 (a64fx_sve_vector_cost): Likewise.
21376 (aarch64_reduc_type): New function.
21377 (aarch64_sve_in_loop_reduction_latency): Likewise.
21378 (aarch64_detect_vector_stmt_subtype): Take a vinfo parameter.
21379 Use aarch64_sve_in_loop_reduction_latency to handle SVE reductions
21380 that occur in the loop body.
21381 (aarch64_add_stmt_cost): Update call accordingly.
21383 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21385 * config/aarch64/aarch64-tuning-flags.def (use_new_vector_costs):
21387 * config/aarch64/aarch64-protos.h (simd_vec_cost): Put comments
21388 above the fields rather than to the right.
21389 (simd_vec_cost::reduc_i8_cost): New member variable.
21390 (simd_vec_cost::reduc_i16_cost): Likewise.
21391 (simd_vec_cost::reduc_i32_cost): Likewise.
21392 (simd_vec_cost::reduc_i64_cost): Likewise.
21393 (simd_vec_cost::reduc_f16_cost): Likewise.
21394 (simd_vec_cost::reduc_f32_cost): Likewise.
21395 (simd_vec_cost::reduc_f64_cost): Likewise.
21396 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
21397 accordingly, using the vec_to_scalar_cost for the new fields.
21398 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
21399 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
21400 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
21401 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
21402 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
21403 (thunderx3t110_advsimd_vector_cost): Likewise.
21404 (aarch64_use_new_vector_costs_p): New function.
21405 (aarch64_simd_vec_costs): New function, split out from...
21406 (aarch64_builtin_vectorization_cost): ...here.
21407 (aarch64_is_reduction): New function.
21408 (aarch64_detect_vector_stmt_subtype): Likewise.
21409 (aarch64_add_stmt_cost): Call aarch64_detect_vector_stmt_subtype if
21410 using the new vector costs.
21412 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
21415 * tree-emutls.c (get_emutls_init_templ_addr): Mark initializer of weak
21416 TLS declarations as public.
21418 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
21420 * config/aarch64/aarch64-d.c (IN_TARGET_CODE): Define.
21421 * config/arm/arm-d.c (IN_TARGET_CODE): Likewise.
21422 * config/i386/i386-d.c (IN_TARGET_CODE): Likewise.
21423 * config/mips/mips-d.c (IN_TARGET_CODE): Likewise.
21424 * config/pa/pa-d.c (IN_TARGET_CODE): Likewise.
21425 * config/riscv/riscv-d.c (IN_TARGET_CODE): Likewise.
21426 * config/rs6000/rs6000-d.c (IN_TARGET_CODE): Likewise.
21427 * config/s390/s390-d.c (IN_TARGET_CODE): Likewise.
21428 * config/sparc/sparc-d.c (IN_TARGET_CODE): Likewise.
21430 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
21433 * config.gcc (*-*-cygwin*): Add winnt-d.o
21434 (*-*-mingw*): Likewise.
21435 * config/i386/cygwin.h (EXTRA_TARGET_D_OS_VERSIONS): New macro.
21436 * config/i386/mingw32.h (EXTRA_TARGET_D_OS_VERSIONS): Likewise.
21437 * config/i386/t-cygming: Add winnt-d.o.
21438 * config/i386/winnt-d.c: New file.
21440 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
21442 * config/freebsd-d.c: Include memmodel.h.
21444 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
21447 * config.gcc (*-*-openbsd*): Add openbsd-d.o.
21448 * config/t-openbsd: Add openbsd-d.o.
21449 * config/openbsd-d.c: New file.
21451 2021-03-25 Stam Markianos-Wright <stam.markianos-wright@arm.com>
21453 PR tree-optimization/96974
21454 * tree-vect-stmts.c (vect_get_vector_types_for_stmt): Replace assert
21455 with graceful exit.
21457 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
21460 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
21464 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
21465 always_inline in system headers.
21467 2021-03-25 Kewen Lin <linkw@linux.ibm.com>
21469 * tree-vect-loop.c (vect_model_reduction_cost): Init inside_cost.
21471 2021-03-25 Jakub Jelinek <jakub@redhat.com>
21474 * tree-core.h (enum operand_equal_flag): Add OEP_ADDRESS_OF_SAME_FIELD.
21475 * fold-const.c (operand_compare::operand_equal_p): Don't compare
21476 field offsets if OEP_ADDRESS_OF_SAME_FIELD.
21478 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
21482 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
21483 always_inline in system headers.
21485 2021-03-25 Richard Biener <rguenther@suse.de>
21487 PR tree-optimization/99746
21488 * tree-vect-slp-patterns.c (complex_pattern::build): Do not mark
21489 the scalar stmt as patterned. Instead set up required things
21492 2021-03-25 Xionghu Luo <luoxhu@linux.ibm.com>
21494 * config/rs6000/rs6000.c (power8_costs): Change l2 cache
21497 2021-03-24 Martin Liska <mliska@suse.cz>
21500 * common/config/i386/i386-common.c (ARRAY_SIZE): Fix off-by-one
21502 * config/i386/i386-options.c (ix86_option_override_internal):
21503 Add run-time assert.
21505 2021-03-24 Martin Jambor <mjambor@suse.cz>
21508 * ipa-cp.c (initialize_node_lattices): Mark as bottom all
21509 parameters with unknown type.
21510 (ipacp_value_safe_for_type): New function.
21511 (propagate_vals_across_arith_jfunc): Verify that the constant type
21512 can be used for a type of the formal parameter.
21513 (propagate_vals_across_ancestor): Likewise.
21514 (propagate_scalar_across_jump_function): Likewise. Pass the type
21515 also to propagate_vals_across_ancestor.
21517 2021-03-24 Christophe Lyon <christophe.lyon@linaro.org>
21520 * config/arm/mve.md (movmisalign<mode>_mve_store): Use Ux
21522 (movmisalign<mode>_mve_load): Likewise.
21524 2021-03-24 Jakub Jelinek <jakub@redhat.com>
21527 * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
21528 movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.
21530 2021-03-24 Alexandre Oliva <oliva@adacore.com>
21532 * doc/sourcebuild.texi (sysconf): New effective target.
21534 2021-03-24 Alexandre Oliva <oliva@adacore.com>
21536 * config/i386/predicates.md (reg_or_const_vec_operand): New.
21537 * config/i386/sse.md (ssse3_pshufbv8qi3): Add an expander for
21538 the now *-prefixed insn_and_split, turn the splitter const vec
21539 into an input for the insn, making it an ignored immediate for
21540 non-split cases, and loaded into the scratch register
21543 2021-03-23 Vladimir N. Makarov <vmakarov@redhat.com>
21546 * config/aarch64/constraints.md (Utq, UOb, UOh, UOw, UOd, UOty):
21547 Use define_relaxed_memory_constraint for them.
21549 2021-03-23 Iain Sandoe <iain@sandoe.co.uk>
21552 * config/host-darwin.c (darwin_gt_pch_use_address): Add a
21553 colon to the diagnostic message.
21555 2021-03-23 Ilya Leoshkevich <iii@linux.ibm.com>
21557 * fwprop.c (fwprop_propagation::fwprop_propagation): Look at
21559 (try_fwprop_subst_note): Use set_info instead of insn_info.
21560 (try_fwprop_subst_pattern): Likewise.
21561 (try_fwprop_subst_notes): Likewise.
21562 (try_fwprop_subst): Likewise.
21563 (forward_propagate_subreg): Likewise.
21564 (forward_propagate_and_simplify): Likewise.
21565 (forward_propagate_into): Likewise.
21566 * rtl-ssa/accesses.h (set_info::single_nondebug_use) New
21568 (set_info::single_nondebug_insn_use): Likewise.
21569 (set_info::single_phi_use): Likewise.
21570 * rtl-ssa/member-fns.inl (set_info::single_nondebug_use) New
21572 (set_info::single_nondebug_insn_use): Likewise.
21573 (set_info::single_phi_use): Likewise.
21575 2021-03-23 Christophe Lyon <christophe.lyon@linaro.org>
21577 * doc/sourcebuild.texi (arm_dsp_ok, arm_dsp): Document.
21579 2021-03-23 Jakub Jelinek <jakub@redhat.com>
21582 * config/aarch64/aarch64.c (aarch64_add_offset): Tell
21583 expand_mult to perform an unsigned rather than a signed
21586 2021-03-23 H.J. Lu <hjl.tools@gmail.com>
21589 * config/i386/cpuid.h (__cpuid): Add __volatile__.
21590 (__cpuid_count): Likewise.
21592 2021-03-23 Richard Biener <rguenther@suse.de>
21594 PR tree-optimization/99721
21595 * tree-vect-slp.c (vect_slp_analyze_node_operations):
21596 Make sure we can schedule the node.
21598 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
21600 * config/riscv/riscv.c (riscv_subword): Take endianness into
21601 account when calculating the byte offset.
21603 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
21605 * config/riscv/predicates.md (subreg_lowpart_operator): New predicate
21606 * config/riscv/riscv.md (*addsi3_extended2, *subsi3_extended2)
21607 (*negsi2_extended2, *mulsi3_extended2, *<optab>si3_mask)
21608 (*<optab>si3_mask_1, *<optab>di3_mask, *<optab>di3_mask_1)
21609 (*<optab>si3_extend_mask, *<optab>si3_extend_mask_1): Use
21610 new predicate "subreg_lowpart_operator"
21612 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
21614 * config/riscv/riscv.c (riscv_swap_instruction): New function
21615 to byteswap an SImode rtx containing an instruction.
21616 (riscv_trampoline_init): Byteswap the generated instructions
21619 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
21621 * common/config/riscv/riscv-common.c
21622 (TARGET_DEFAULT_TARGET_FLAGS): Set default endianness.
21623 * config.gcc (riscv32be-*, riscv64be-*): Set
21624 TARGET_BIG_ENDIAN_DEFAULT to 1.
21625 * config/riscv/elf.h (LINK_SPEC): Change -melf* value
21626 depending on default endianness.
21627 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
21628 * config/riscv/linux.h (LINK_SPEC): Likewise.
21629 * config/riscv/riscv.c (TARGET_DEFAULT_TARGET_FLAGS): Set
21630 default endianness.
21631 * config/riscv/riscv.h (DEFAULT_ENDIAN_SPEC): New macro.
21633 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
21635 * config/riscv/elf.h (LINK_SPEC): Pass linker endianness flag.
21636 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
21637 * config/riscv/linux.h (LINK_SPEC): Likewise.
21638 * config/riscv/riscv.h (ASM_SPEC): Pass -mbig-endian and
21640 (BYTES_BIG_ENDIAN): Handle big endian.
21641 (WORDS_BIG_ENDIAN): Define to BYTES_BIG_ENDIAN.
21642 * config/riscv/riscv.opt (-mbig-endian, -mlittle-endian): New
21644 * doc/invoke.texi (-mbig-endian, -mlittle-endian): Document.
21646 2021-03-23 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
21648 * regcprop.c (find_oldest_value_reg): Ask target whether
21649 different mode is fine for replacement register.
21651 2021-03-23 Aldy Hernandez <aldyh@redhat.com>
21653 PR tree-optimization/99296
21654 * value-range.cc (irange::irange_set_1bit_anti_range): New.
21655 (irange::irange_set_anti_range): Call irange_set_1bit_anti_range
21656 * value-range.h (irange::irange_set_1bit_anti_range): New.
21658 2021-03-22 Vladimir N. Makarov <vmakarov@redhat.com>
21661 * config/aarch64/constraints.md (UtQ): Use
21662 define_relaxed_memory_constraint for it.
21663 * doc/md.texi (define_relaxed_memory_constraint): Describe it.
21664 * genoutput.c (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
21665 * genpreds.c (constraint_data): Add bitfield is_relaxed_memory.
21666 (have_relaxed_memory_constraints): New static var.
21667 (relaxed_memory_start, relaxed_memory_end): Ditto.
21668 (add_constraint): Add arg is_relaxed_memory. Check name for
21669 relaxed memory. Set up is_relaxed_memory in constraint_data and
21670 have_relaxed_memory_constraints. Adjust calls.
21671 (choose_enum_order): Process relaxed memory.
21672 (write_tm_preds_h): Ditto.
21673 (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
21674 * gensupport.c (process_rtx): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
21675 * ira-costs.c (record_reg_classes): Process CT_RELAXED_MEMORY.
21676 * ira-lives.c (single_reg_class): Use
21677 insn_extra_relaxed_memory_constraint.
21678 * ira.c (ira_setup_alts): CT_RELAXED_MEMORY.
21679 * lra-constraints.c (valid_address_p): Use
21680 insn_extra_relaxed_memory_constraint instead of other memory
21682 (process_alt_operands): Process CT_RELAXED_MEMORY.
21683 (curr_insn_transform): Use insn_extra_relaxed_memory_constraint.
21684 * recog.c (asm_operand_ok, preprocess_constraints): Process
21686 * reload.c (find_reloads): Ditto.
21687 * rtl.def (DEFINE_RELAXED_MEMORY_CONSTRAINT): New.
21688 * stmt.c (parse_input_constraint): Use
21689 insn_extra_relaxed_memory_constraint.
21691 2021-03-22 Segher Boessenkool <segher@kernel.crashing.org>
21694 * ubsan.c (ubsan_instrument_float_cast): Don't test for unordered if
21697 2021-03-22 Alex Coplan <alex.coplan@arm.com>
21700 * config/arm/arm-protos.h (neon_make_constant): Add generate
21701 argument to guard emitting insns, default to true.
21702 * config/arm/arm.c (arm_legitimate_constant_p_1): Reject
21703 CONST_VECTORs which neon_make_constant can't handle.
21704 (neon_vdup_constant): Add generate argument, avoid emitting
21705 insns if it's not set.
21706 (neon_make_constant): Plumb new generate argument through.
21707 * config/arm/constraints.md (Ui): New. Use it...
21708 * config/arm/mve.md (*mve_mov<mode>): ... here.
21709 * config/arm/vec-common.md (movv8hf): Use neon_make_constant to
21710 synthesize constants.
21712 2021-03-22 Richard Biener <rguenther@suse.de>
21714 * debug.h: Add deprecation warning.
21716 2021-03-22 Richard Biener <rguenther@suse.de>
21718 PR tree-optimization/99694
21719 * tree-ssa-sccvn.c (visit_phi): Ignore edges with the
21722 2021-03-22 Kito Cheng <kito.cheng@sifive.com>
21725 * config/riscv/riscv.c (riscv_expand_block_move): Get RTL value
21726 after type checking.
21728 2021-03-22 Jakub Jelinek <jakub@redhat.com>
21732 * dwarf2out.c (get_full_len): Use get_precision rather than
21734 (add_const_value_attribute): Make sure add_AT_wide argument has
21735 precision prec rather than some very wide one.
21737 2021-03-22 Kewen Lin <linkw@linux.ibm.com>
21739 * config/rs6000/rs6000.md (*rotldi3_insert_sf,
21740 *mov<SFDF:mode><SFDF2:mode>cc_p9, floatsi<mode>2_lfiwax,
21741 floatsi<mode>2_lfiwax_mem, floatunssi<mode>2_lfiwzx,
21742 floatunssi<mode>2_lfiwzx_mem, *floatsidf2_internal,
21743 *floatunssidf2_internal, fix_trunc<mode>si2_stfiwx,
21744 fix_trunc<mode>si2_internal, fixuns_trunc<mode>si2_stfiwx,
21745 *round32<mode>2_fprs, *roundu32<mode>2_fprs,
21746 *fix_trunc<mode>si2_internal): Fix empty split condition.
21747 * config/rs6000/vsx.md (*vsx_le_undo_permute_<mode>,
21748 vsx_reduc_<VEC_reduc_name>_v2df, vsx_reduc_<VEC_reduc_name>_v4sf,
21749 *vsx_reduc_<VEC_reduc_name>_v2df_scalar,
21750 *vsx_reduc_<VEC_reduc_name>_v4sf_scalar): Likewise.
21752 2021-03-22 Xionghu Luo <luoxhu@linux.ibm.com>
21755 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
21756 Convert idx to DImode.
21757 (rs6000_expand_vector_set_var_p8): Likewise.
21759 2021-03-21 Jakub Jelinek <jakub@redhat.com>
21762 * dwarf2out.c (insert_float): Change return type from void to
21763 unsigned, handle GET_MODE_SIZE (mode) == 2 and return element size.
21764 (mem_loc_descriptor, loc_descriptor, add_const_value_attribute):
21767 2021-03-20 H.J. Lu <hjl.tools@gmail.com>
21770 * config/i386/i386.c (construct_container): Check cfun != NULL
21771 before accessing silent_p.
21773 2021-03-20 Ahamed Husni <ahamedhusni73@gmail.com>
21775 * asan.c: Fix typos in comments.
21777 2021-03-20 Vladimir N. Makarov <vmakarov@redhat.com>
21779 PR rtl-optimization/99680
21780 * lra-constraints.c (skip_contraint_modifiers): Rename to skip_constraint_modifiers.
21781 (process_address_1): Check empty constraint before using
21784 2021-03-19 Pat Haugen <pthaugen@linux.ibm.com>
21786 * config/rs6000/rs6000.c (power10_cost): New.
21787 (rs6000_option_override_internal): Set Power10 costs.
21788 (rs6000_issue_rate): Set Power10 issue rate.
21789 * config/rs6000/power10.md: Rewrite for Power10.
21791 2021-03-19 Vladimir N. Makarov <vmakarov@redhat.com>
21794 * lra-constraints.c (process_address_1): Don't use unknown
21795 constraint for address constraint.
21797 2021-03-19 Iain Sandoe <iain@sandoe.co.uk>
21800 * config.gcc (powerpc-*-darwin8): Delete the reference to
21801 the now removed darwin8.h.
21803 2021-03-19 Olivier Hainque <hainque@adacore.com>
21806 * config/vxworksae.h (VX_CPU_PREFIX): Define.
21808 2021-03-19 John David Anglin <danglin@gcc.gnu.org>
21810 * config/pa/pa.c (import_milli): Use memcpy instead of strncpy.
21812 2021-03-19 Tamar Christina <tamar.christina@arm.com>
21814 PR tree-optimization/99656
21815 * tree-vect-slp-patterns.c (linear_loads_p,
21816 complex_add_pattern::matches, is_eq_or_top,
21817 vect_validate_multiplication, complex_mul_pattern::matches,
21818 complex_fms_pattern::matches): Remove complex_perm_kinds_t.
21819 * tree-vectorizer.h: (complex_load_perm_t): Removed.
21820 (slp_tree_to_load_perm_map_t): Use complex_perm_kinds_t instead of
21821 complex_load_perm_t.
21823 2021-03-19 H.J. Lu <hjl.tools@gmail.com>
21826 * config/i386/i386-options.c (ix86_init_machine_status): Set
21828 * config/i386/i386.c (init_cumulative_args): Set silent_p to
21830 (construct_container): Return early for return and argument
21831 errors if silent_p is true.
21832 * config/i386/i386.h (machine_function): Add silent_p.
21834 2021-03-19 Jakub Jelinek <jakub@redhat.com>
21837 * config/arm/constraints.md (Ds): New constraint.
21838 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): Use w,Ds
21839 constraint instead of w,Dm.
21841 2021-03-19 Andrew Stubbs <ams@codesourcery.com>
21843 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Fix quotes
21846 2021-03-19 Eric Botcazou <ebotcazou@adacore.com>
21848 PR middle-end/99641
21849 * fold-const.c (native_encode_initializer) <CONSTRUCTOR>: For an
21850 array type, do the computation of the current position in sizetype.
21852 2021-03-18 Vladimir N. Makarov <vmakarov@redhat.com>
21855 * lra-constraints.c (process_address_1): Use lookup_constraint
21856 only for a single constraint.
21858 2021-03-18 Martin Sebor <msebor@redhat.com>
21860 PR middle-end/99502
21861 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): Rename...
21862 (inbounds_memaccess_p): ...to this. Check the ending offset of
21863 the accessed member.
21865 2021-03-18 Andrew Stubbs <ams@codesourcery.com>
21867 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Add %< and
21868 %> quote markers to error messages.
21869 (gcn_goacc_validate_dims): Likewise.
21870 (gcn_conditional_register_usage): Remove exclaimation mark from error
21872 (gcn_vectorize_vec_perm_const): Ensure perm is fully uninitialized.
21874 2021-03-18 Jan Hubicka <hubicka@ucw.cz>
21876 * config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
21879 2021-03-18 Sinan Lin <sinan@isrc.iscas.ac.cn>
21880 Kito Cheng <kito.cheng@sifive.com>
21882 * config/riscv/riscv.c (riscv_block_move_straight): Change type
21883 to unsigned HOST_WIDE_INT for parameter and local variable with
21884 HOST_WIDE_INT type.
21885 (riscv_adjust_block_mem): Ditto.
21886 (riscv_block_move_loop): Ditto.
21887 (riscv_expand_block_move): Ditto.
21889 2021-03-18 Nick Clifton <nickc@redhat.com>
21891 * config/v850/v850.c (construct_restore_jr): Increase static
21893 (construct_save_jarl): Likewise.
21894 * config/v850/v850.h (DWARF2_DEBUGGING_INFO): Define.
21896 2021-03-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21898 * config/aarch64/aarch64.c (aarch64_adjust_generic_arch_tuning): Define.
21899 (aarch64_override_options_internal): Use it.
21900 (generic_tunings): Add AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS to
21903 2021-03-17 Sandra Loosemore <sandra@codesourcery.com>
21905 * config/nios2/nios2.c (nios2_custom_check_insns): Clean up
21906 error message format issues.
21907 (nios2_option_override): Likewise.
21908 (nios2_expand_fpu_builtin): Likewise.
21909 (nios2_init_custom_builtins): Adjust to avoid bogus strncpy
21910 truncation warning.
21911 (nios2_expand_custom_builtin): More error message format fixes.
21912 (nios2_expand_rdwrctl_builtin): Likewise.
21913 (nios2_expand_rdprs_builtin): Likewise.
21914 (nios2_expand_eni_builtin): Likewise.
21915 (nios2_expand_builtin): Likewise.
21916 (nios2_register_custom_code): Likewise.
21917 (nios2_valid_target_attribute_rec): Likewise.
21918 (nios2_add_insn_asm): Fix uninitialized variable warning.
21920 2021-03-17 Jan Hubicka <jh@suse.cz>
21922 * config/i386/x86-tune-costs.h (struct processor_costs): Update costs
21923 of gather to match reality.
21924 * config/i386/x86-tune.def (X86_TUNE_USE_GATHER): Enable for znver3.
21926 2021-03-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21928 * config/aarch64/aarch64-builtins.c (aarch64_expand_rng_builtin): Use EQ
21929 to compare against CC_REG rather than NE.
21931 2021-03-17 H.J. Lu <hjl.tools@gmail.com>
21934 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support
21935 inline assembly statements.
21936 (ix86_print_operand): Update 'P' handling for -fno-plt.
21938 2021-03-17 Tamar Christina <tamar.christina@arm.com>
21941 * config/aarch64/aarch64.c
21942 (aarch64_simd_clone_compute_vecsize_and_simdlen): Remove unused var.
21944 2021-03-16 Segher Boessenkool <segher@kernel.crashing.org>
21947 * config/rs6000/predicates.md (branch_comparison_operator): Allow
21948 ordered and unordered for CCFPmode, if flag_finite_math_only.
21950 2021-03-16 Jakub Jelinek <jakub@redhat.com>
21953 * config/i386/i386-expand.c (ix86_split_lea_for_addr): Emit a MULT
21954 rather than ASHIFT.
21955 * config/i386/i386.md (mult by 1248 into ashift): New splitter.
21957 2021-03-16 Martin Liska <mliska@suse.cz>
21960 * optc-save-gen.awk: Add flag_ipa_ra to exceptions for
21961 cl_optimization_compare function.
21963 2021-03-16 Ilya Leoshkevich <iii@linux.ibm.com>
21965 * config/s390/s390.c (f_constraint_p): Treat "fv" constraints
21968 2021-03-16 Jakub Jelinek <jakub@redhat.com>
21971 * config/i386/i386.h (struct machine_function): Add
21972 has_explicit_vzeroupper bitfield.
21973 * config/i386/i386-expand.c (ix86_expand_builtin): Set
21974 cfun->machine->has_explicit_vzeroupper when expanding
21975 IX86_BUILTIN_VZEROUPPER.
21976 * config/i386/i386-features.c (rest_of_handle_insert_vzeroupper):
21977 Do the mode switching only when TARGET_VZEROUPPER, expensive
21978 optimizations turned on and not optimizing for size.
21979 (pass_insert_vzeroupper::gate): Enable even when
21980 cfun->machine->has_explicit_vzeroupper is set.
21982 2021-03-16 Jakub Jelinek <jakub@redhat.com>
21985 * config/aarch64/aarch64.c
21986 (aarch64_simd_clone_compute_vecsize_and_simdlen): If not a function
21987 definition, walk TYPE_ARG_TYPES list if non-NULL for argument types
21988 instead of DECL_ARGUMENTS. Ignore types for uniform arguments.
21990 2021-03-15 Richard Biener <rguenther@suse.de>
21992 PR tree-optimization/98834
21993 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle missing
21994 subsetting by truncating the access size.
21996 2021-03-15 Jan Hubicka <hubicka@ucw.cz>
21998 * config/i386/i386-options.c (processor_cost_table): Add znver3_cost.
21999 * config/i386/x86-tune-costs.h (znver3_cost): New gobal variable; copy
22002 2021-03-15 Martin Liska <mliska@suse.cz>
22004 * spellcheck.c: Add missing comma in initialization.
22006 2021-03-14 Uroš Bizjak <ubizjak@gmail.com>
22008 * config/i386/sse.md (*vec_extract<mode>): Merge alternative 0 with
22009 alternative 2 and alternative 1 with alternative 3 using
22010 YW register constraint.
22011 (*vec_extract<PEXTR_MODE12:mode>_zext): Merge alternatives
22012 using YW register constraint.
22013 (*vec_extractv16qi_zext): Ditto.
22014 (*vec_extractv4si): Merge alternatives 4 and 5
22015 using Yw register constraint.
22016 (*ssse3_palignr<mode>_perm): Use Yw instead of v for alternative 3.
22018 2021-03-13 Martin Sebor <msebor@redhat.com>
22020 PR tree-optimization/99489
22021 * builtins.c (gimple_call_alloc_size): Fail gracefully when argument
22022 is not a call statement.
22024 2021-03-13 Jakub Jelinek <jakub@redhat.com>
22026 PR tree-optimization/99544
22027 * match.pd (X + (X << C) -> X * (1 + (1 << C))): Don't simplify
22028 if for vector types multiplication can't be done in type's mode.
22030 2021-03-12 Eric Botcazou <ebotcazou@adacore.com>
22033 * config/sparc/constraints.md (w): Rename to...
22034 (W): ... this and ditch previous implementation.
22035 * config/sparc/sparc.md (*movdi_insn_sp64): Replace W with m.
22036 (*movdf_insn_sp64): Likewise.
22037 (*mov<VM64:mode>_insn_sp64): Likewise.
22038 * config/sparc/sync.md (*atomic_compare_and_swap<mode>_1): Replace
22040 (atomic_compare_and_swap_leon3_1): Likewise.
22041 (*atomic_compare_and_swapdi_v8plus): Likewise.
22042 * config/sparc/sparc.c (memory_ok_for_ldd): Remove useless test on
22043 architecture and add missing address validity check during LRA.
22045 2021-03-12 Tobias Burnus <tobias@codesourcery.com>
22048 * gimplify.c (omp_add_variable): Handle NULL_TREE as size
22049 occuring for assumed-size arrays in use_device_{ptr,addr}.
22051 2021-03-12 Jakub Jelinek <jakub@redhat.com>
22054 * config/i386/constraints.md (YW): New internal constraint.
22055 * config/i386/sse.md (v_Yw): Add V4TI, V2TI, V1TI and TI cases.
22056 (*<sse2_avx2>_<insn><mode>3<mask_name>,
22057 *<sse2_avx2>_uavg<mode>3<mask_name>, *abs<mode>2,
22058 *<s>mul<mode>3_highpart<mask_name>): Use <v_Yw> instead of v in
22060 (<sse2_avx2>_psadbw): Use YW instead of v in constraints.
22061 (*avx2_pmaddwd, *sse2_pmaddwd, *<code>v8hi3, *<code>v16qi3,
22062 avx2_pmaddubsw256, ssse3_pmaddubsw128): Merge last two alternatives
22063 into one, use Yw instead of former x,v.
22064 (ashr<mode>3, <insn><mode>3): Use <v_Yw> instead of x in constraints of
22065 the last alternative.
22066 (<sse2_avx2>_packsswb<mask_name>, <sse2_avx2>_packssdw<mask_name>,
22067 <sse2_avx2>_packuswb<mask_name>, <sse4_1_avx2>_packusdw<mask_name>,
22068 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, <ssse3_avx2>_palignr<mode>,
22069 <ssse3_avx2>_pshufb<mode>3<mask_name>): Merge last two alternatives
22070 into one, use <v_Yw> instead of former x,v.
22071 (avx2_interleave_highv32qi<mask_name>,
22072 vec_interleave_highv16qi<mask_name>): Use Yw instead of v in
22073 constraints. Add && <mask_avx512bw_condition> to condition.
22074 (avx2_interleave_lowv32qi<mask_name>,
22075 vec_interleave_lowv16qi<mask_name>,
22076 avx2_interleave_highv16hi<mask_name>,
22077 vec_interleave_highv8hi<mask_name>,
22078 avx2_interleave_lowv16hi<mask_name>, vec_interleave_lowv8hi<mask_name>,
22079 avx2_pshuflw_1<mask_name>, sse2_pshuflw_1<mask_name>,
22080 avx2_pshufhw_1<mask_name>, sse2_pshufhw_1<mask_name>,
22081 avx2_<code>v16qiv16hi2<mask_name>, sse4_1_<code>v8qiv8hi2<mask_name>,
22082 *sse4_1_<code>v8qiv8hi2<mask_name>_1, <sse2_avx2>_<insn><mode>3): Use
22083 Yw instead of v in constraints.
22084 * config/i386/mmx.md (Yv_Yw): New define_mode_attr.
22085 (*mmx_<insn><mode>3, mmx_ashr<mode>3, mmx_<insn><mode>3): Use <Yv_Yw>
22086 instead of Yv in constraints.
22087 (*mmx_<insn><mode>3, *mmx_mulv4hi3, *mmx_smulv4hi3_highpart,
22088 *mmx_umulv4hi3_highpart, *mmx_pmaddwd, *mmx_<code>v4hi3,
22089 *mmx_<code>v8qi3, mmx_pack<s_trunsuffix>swb, mmx_packssdw,
22090 mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
22091 *mmx_uavgv8qi3, *mmx_uavgv4hi3, mmx_psadbw): Use Yw instead of Yv in
22093 (*mmx_pinsrw, *mmx_pinsrb, *mmx_pextrw, *mmx_pextrw_zext, *mmx_pextrb,
22094 *mmx_pextrb_zext): Use YW instead of Yv in constraints.
22095 (*mmx_eq<mode>3, mmx_gt<mode>3): Use x instead of Yv in constraints.
22096 (mmx_andnot<mode>3, *mmx_<code><mode>3): Split last alternative into
22097 two, one with just x, another isa avx512vl with v.
22099 2021-03-12 Martin Liska <mliska@suse.cz>
22101 * doc/invoke.texi: Add missing param documentation.
22103 2021-03-11 David Malcolm <dmalcolm@redhat.com>
22106 * Makefile.in (ANALYZER_OBJS): Add analyzer/feasible-graph.o and
22107 analyzer/trimmed-graph.o.
22108 * doc/analyzer.texi (Analyzer Paths): Rewrite description of
22109 feasibility checking to reflect new implementation.
22110 * doc/invoke.texi (-fdump-analyzer-feasibility): Document new
22112 * shortest-paths.h (shortest_paths::get_shortest_distance): New.
22114 2021-03-11 David Malcolm <dmalcolm@redhat.com>
22116 * digraph.cc (selftest::test_shortest_paths): Update
22117 shortest_paths init for new param. Add test of
22118 SPS_TO_GIVEN_TARGET.
22119 * shortest-paths.h (enum shortest_path_sense): New.
22120 (shortest_paths::shortest_paths): Add "sense" param.
22121 Update for renamings. Generalize to use "sense" param.
22122 (shortest_paths::get_shortest_path): Rename param.
22123 (shortest_paths::m_sense): New field.
22124 (shortest_paths::m_prev): Rename...
22125 (shortest_paths::m_best_edge): ...to this.
22126 (shortest_paths::get_shortest_path): Update for renamings.
22127 Conditionalize flipping of path on sense of traversal.
22129 2021-03-11 David Malcolm <dmalcolm@redhat.com>
22131 * digraph.cc (selftest::test_shortest_paths): Add test coverage
22132 for paths from B and C.
22133 * shortest-paths.h (shortest_paths::shortest_paths): Handle
22134 unreachable nodes, rather than asserting.
22136 2021-03-11 David Edelsohn <dje.gcc@gmail.com>
22139 * config/rs6000/rs6000.c (rs6000_xcoff_file_start): Don't create
22140 xcoff_tbss_section_name.
22141 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .lcomm.
22142 * xcoffout.c (xcoff_tbss_section_name): Delete.
22143 * xcoffout.h (xcoff_tbss_section_name): Delete.
22145 2021-03-11 Richard Biener <rguenther@suse.de>
22147 PR tree-optimization/99523
22148 * tree-cfg.c (dump_function_to_file): Dump SSA names
22149 w/o identifier to the decls section as well, not only those
22150 without a VAR_DECL.
22152 2021-03-11 Jakub Jelinek <jakub@redhat.com>
22155 * ipa-icf-gimple.c (func_checker::compare_gimple_call): For internal
22156 function calls with lhs fail if the lhs don't have compatible types.
22158 2021-03-11 Hans-Peter Nilsson <hp@axis.com>
22160 * config/cris/cris.h (HARD_FRAME_POINTER_REGNUM): Define.
22161 Change FRAME_POINTER_REGNUM to correspond to a new faked
22162 register faked_fp, part of GENNONACR_REGS like faked_ap.
22163 (CRIS_FAKED_REGS_CONTENTS): New helper macro.
22164 (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS, CALL_USED_REGISTERS):
22165 (REG_ALLOC_ORDER, REG_CLASS_CONTENTS, REGNO_OK_FOR_BASE_P)
22166 (ELIMINABLE_REGS, REGISTER_NAMES): Adjust accordingly.
22167 * config/cris/cris.md (CRIS_FP_REGNUM): Renumber to new faked
22169 (CRIS_REAL_FP_REGNUM): New constant.
22170 * config/cris/cris.c (cris_reg_saved_in_regsave_area): Check
22171 for HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM.
22172 (cris_initial_elimination_offset): Handle elimination changes
22173 to HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM
22174 and add one from FRAME_POINTER_REGNUM to
22175 HARD_FRAME_POINTER_REGNUM.
22176 (cris_expand_prologue, cris_expand_epilogue): Emit code for
22177 hard_frame_pointer_rtx instead of frame_pointer_rtx.
22179 2021-03-10 David Edelsohn <dje.gcc@gmail.com>
22182 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Add check for DCmode.
22183 * config/rs6000/rs6000.c (rs6000_special_round_type_align): Same.
22185 2021-03-10 Vladimir N. Makarov <vmakarov@redhat.com>
22188 * lra-constraints.c (process_address_1): Don't check unknown
22189 constraint, use X for empty constraint.
22191 2021-03-10 Alex Coplan <alex.coplan@arm.com>
22193 * config/aarch64/aarch64.c (aarch64_vfp_is_call_or_return_candidate):
22194 Fix typo in comment describing "is_ha" argument.
22196 2021-03-10 John David Anglin <danglin@gcc.gnu.org>
22198 * doc/sourcebuild.texi: Document LRA target selector.
22200 2021-03-10 David Malcolm <dmalcolm@redhat.com>
22202 * doc/ux.texi: Add subsection contrasting interactive versus
22203 batch usage of GCC.
22205 2021-03-10 Joel Hutton <joel.hutton@arm.com>
22208 * tree-vect-stmts.c (vectorizable_store): Fix scatter store mask
22210 (vectorizable_load): Fix gather load mask check condition.
22212 2021-03-10 Richard Biener <rguenther@suse.de>
22214 PR tree-optimization/99510
22215 * tree.c (check_aligned_type): Check that the candidate
22216 has TYPE_USER_ALIGN set instead of matching with the
22219 2021-03-10 Eric Botcazou <ebotcazou@adacore.com>
22221 * config/sparc/sparc.c (sparc_regmode_natural_size): Return 4 for
22222 float and vector integer modes only if the mode is not larger.
22224 2021-03-10 Hans-Peter Nilsson <hp@axis.com>
22226 * config/cris/cris.h (DWARF_FRAME_REGISTERS): Define.
22228 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
22230 * ira.c (ira_setup_alts, ira_get_dup_out_num): Process digital
22232 * ira-lives.c (single_reg_class): Ditto.
22234 2021-03-09 Sebastian Huber <sebastian.huber@embedded-brains.de>
22236 * config.gcc (aarch64-*-rtems*): Include general rtems.h after
22237 the architecture-specific rtems.h.
22238 (aarch64-*-rtems*): Likewise.
22239 (arm*-*-rtems*): Likewise.
22240 (epiphany-*-rtems*): Likewise.
22241 (riscv*-*-rtems*): Likewise.
22243 2021-03-09 Jakub Jelinek <jakub@redhat.com>
22245 PR tree-optimization/99305
22246 * tree-ssa-phiopt.c (conditional_replacement): Test integer_pow2p
22247 before integer_all_onesp instead of vice versa.
22249 2021-03-09 Richard Earnshaw <rearnsha@arm.com>
22251 * common/config/arm/arm-common.c (arm_config_default): Change type
22252 of 'i' to unsigned.
22254 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
22257 * lra-constraints.c (process_address_1): Process constraint 'g'
22258 separately and digital constraints containing more one digit.
22260 2021-03-09 Nick Clifton <nickc@redhat.com>
22262 * config/rx/rx.h (DBX_DEBUGGING_INFO): Define.
22263 (DWARF"_DEBUGGING_INFO): Define.
22265 2021-03-09 Eric Botcazou <ebotcazou@adacore.com>
22268 * calls.c (initialize_argument_information): When the argument
22269 is passed by reference, do not make a copy in a thunk only if
22270 the argument is already in memory. Remove redundant test for
22271 the case of callee copy.
22273 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
22276 * lra-constraints.c (process_address_1): Process 0..9 constraints
22277 in process_address_1.
22279 2021-03-09 Andreas Krebbel <krebbel@linux.ibm.com>
22281 * config/s390/s390.c (struct s390_processor processor_table):
22282 Binutils name string must not be empty.
22284 2021-03-09 Claudiu Zissulescu <claziss@synopsys.com>
22286 * config/arc/arc.c (arc_attr_type): Remove function.
22288 2021-03-09 Martin Liska <mliska@suse.cz>
22291 * config/i386/i386-options.c (ix86_option_override_internal):
22292 Set isa_flags for OPTS argument and not for the global
22295 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
22297 * config/rs6000/predicates.md (ds_form_mem_operand): Check
22300 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
22303 * config/rs6000/predicates.md (ds_form_mem_operand) New
22305 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10) Use
22306 ds_form_mem_operand in ld/lwa patterns.
22307 * config/rs6000/fusion.md: Regenerate file.
22309 2021-03-08 Martin Sebor <msebor@redhat.com>
22311 PR middle-end/98266
22312 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): New function.
22313 (array_bounds_checker::check_array_bounds): Call it.
22315 2021-03-08 Martin Sebor <msebor@redhat.com>
22317 PR middle-end/97631
22318 * tree-ssa-strlen.c (maybe_warn_overflow): Test rawmem.
22319 (handle_builtin_stxncpy_strncat): Rename locals. Determine
22320 destination size from allocation calls. Issue a more appropriate
22322 (handle_builtin_memcpy): Pass true as rawmem to maybe_warn_overflow.
22323 (handle_builtin_memset): Same.
22325 2021-03-08 Peter Bergner <bergner@linux.ibm.com>
22328 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_permute): Add an assert
22329 to ensure we do not have an Altivec style address.
22330 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): Disable if passed
22331 an Altivec style address.
22332 (*vsx_le_perm_store_<mode>): Likewise.
22333 (splitters after *vsx_le_perm_store_<mode>): Likewise.
22334 (vsx_load_<mode>): Disable special expander if passed an Altivec
22336 (vsx_store_<mode>): Likewise.
22338 2021-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22341 * config/aarch64/predicates.md (aarch64_simd_shift_imm_vec_qi): Define.
22342 (aarch64_simd_shift_imm_vec_hi): Likewise.
22343 (aarch64_simd_shift_imm_vec_si): Likewise.
22344 (aarch64_simd_shift_imm_vec_di): Likewise.
22345 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le): Use
22346 predicate from above.
22347 (aarch64_shrn<mode>_insn_be): Likewise.
22348 (aarch64_rshrn<mode>_insn_le): Likewise.
22349 (aarch64_rshrn<mode>_insn_be): Likewise.
22350 (aarch64_shrn2<mode>_insn_le): Likewise.
22351 (aarch64_shrn2<mode>_insn_be): Likewise.
22352 (aarch64_rshrn2<mode>_insn_le): Likewise.
22353 (aarch64_rshrn2<mode>_insn_be): Likewise.
22355 2021-03-08 Vladimir N. Makarov <vmakarov@redhat.com>
22358 * lra-constraints.c (skip_contraint_modifiers): New function.
22359 (process_address_1): Use it before lookup_constraint call.
22361 2021-03-08 Martin Liska <mliska@suse.cz>
22364 * config/i386/i386-options.c (ix86_option_override_internal):
22365 Enable UINTR and HRESET for -march that supports it.
22367 2021-03-08 Ilya Leoshkevich <iii@linux.ibm.com>
22369 * config/s390/s390.c (f_constraint_p): New function.
22370 (s390_md_asm_adjust): Implement TARGET_MD_ASM_ADJUST.
22371 (TARGET_MD_ASM_ADJUST): Likewise.
22373 2021-03-08 Tobias Burnus <tobias@codesourcery.com>
22376 * tree-nested.c (convert_local_reference_stmt): Avoid calling
22377 lookup_field_for_decl for Fortran module (= namespace context).
22379 2021-03-08 Andreas Krebbel <krebbel@linux.ibm.com>
22381 * config/s390/s390.c (s390_expand_vec_compare): Implement <0
22382 comparison with arithmetic right shift.
22383 (s390_expand_vcond): No need for a force_reg anymore.
22384 s390_vec_compare will do it.
22385 * config/s390/vector.md ("vec_cmp<mode><tointvec>"): Accept also
22386 immediate operands.
22388 2021-03-07 Jakub Jelinek <jakub@redhat.com>
22391 * config/i386/constraints.md (Yw): Use SSE_REGS if TARGET_SSE
22392 but TARGET_AVX512BW or TARGET_AVX512VL is not set. Adjust description
22394 * config/i386/sse.md (v_Yw): New define_mode_attr.
22395 (*<insn><mode>3, *mul<mode>3<mask_name>, *avx2_<code><mode>3,
22396 *sse4_1_<code><mode>3<mask_name>): Use <v_Yw> instead of v
22398 * config/i386/mmx.md (mmx_pshufw_1, *vec_dupv4hi): Use Yw instead of
22399 xYw in constraints.
22401 2021-03-06 Julian Brown <julian@codesourcery.com>
22403 * tree-pretty-print.c (dump_generic_node): Emit non-generic
22404 address space info for aggregates.
22406 2021-03-06 Hans-Peter Nilsson <hp@axis.com>
22408 * config/cris/cris.h (MAX_FIXED_MODE_SIZE): Don't define.
22410 2021-03-05 Jakub Jelinek <jakub@redhat.com>
22412 PR middle-end/99322
22413 * tree-cfg.c (bb_to_omp_idx): New variable.
22414 (execute_build_cfg): Release the bb_to_omp_idx vector after
22415 cleanup_tree_cfg returns.
22416 (handle_abnormal_edges): Remove bb_to_omp_idx argument, adjust
22417 for bb_to_omp_idx being a vec<int> instead of pointer to array
22419 (make_edges): Remove bb_to_omp_idx local variable, don't pass
22420 it to handle_abnormal_edges, adjust for bb_to_omp_idx being a
22421 vec<int> instead of pointer to array of ints and don't free/release
22423 (remove_bb): When removing a bb and placing forced label somewhere
22424 else, ensure it is put into the same OpenMP region during cfg
22425 pass if possible or to entry successor as fallback. Unregister
22426 bb from bb_to_omp_idx.
22428 2021-03-05 Vladimir N. Makarov <vmakarov@redhat.com>
22431 * lra-constraints.c (process_address_1): Skip decomposing address
22432 for asm insn operand with unknown constraint.
22434 2021-03-05 Martin Jambor <mjambor@suse.cz>
22437 * cgraph.c (cgraph_edge::set_call_stmt): Do not update all
22438 corresponding speculative edges if we are about to resolve
22439 sepculation. Make edge direct (and so resolve speculations) before
22440 removing it from call_site_hash.
22441 (cgraph_edge::make_direct): Relax the initial assert to allow calling
22442 the function on speculative direct edges.
22444 2021-03-05 Eric Botcazou <ebotcazou@adacore.com>
22446 PR rtl-optimization/99376
22447 * rtlanal.c (nonzero_bits1) <arithmetic operators>: If the number
22448 of low-order zero bits is too large, set the result to 0 directly.
22450 2021-03-04 Jakub Jelinek <jakub@redhat.com>
22452 PR middle-end/93235
22453 * expmed.c (store_bit_field_using_insv): Return false of xop0 is a
22454 SUBREG and a SUBREG to op_mode can't be created.
22456 2021-03-04 Alex Coplan <alex.coplan@arm.com>
22459 * config/aarch64/aarch64-sve-builtins.cc
22460 (function_resolver::require_vector_type): Handle error_mark_node.
22462 2021-03-04 Ilya Leoshkevich <iii@linux.ibm.com>
22464 * cfgexpand.c (expand_asm_loc): Pass new parameter.
22465 (expand_asm_stmt): Likewise.
22466 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Add new
22468 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
22469 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
22470 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
22471 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
22472 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
22473 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
22474 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
22475 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
22476 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
22477 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
22478 * doc/tm.texi (md_asm_adjust): Likewise.
22479 * target.def (md_asm_adjust): Likewise.
22481 2021-03-04 Richard Biener <rguenther@suse.de>
22483 PR middle-end/97855
22484 * tree-pretty-print.c: Poison pp_printf.
22485 (dump_decl_name): Avoid use of pp_printf.
22486 (dump_block_node): Likewise.
22487 (dump_generic_node): Likewise.
22489 2021-03-04 Martin Sebor <msebor@redhat.com>
22491 PR middle-end/96963
22492 PR middle-end/94655
22493 * builtins.c (handle_array_ref): New helper.
22494 (handle_mem_ref): New helper.
22495 (compute_objsize_r): Factor out ARRAY_REF and MEM_REF handling
22496 into new helper functions. Correct a workaround for vectorized
22499 2021-03-03 Pat Haugen <pthaugen@linux.ibm.com>
22501 * config/rs6000/dfp.md (extendddtd2, trunctddd2, *cmp<mode>_internal1,
22502 floatditd2, ftrunc<mode>2, fix<mode>di2, dfp_ddedpd_<mode>,
22503 dfp_denbcd_<mode>, dfp_dxex_<mode>, dfp_diex_<mode>,
22504 *dfp_sgnfcnc_<mode>, dfp_dscli_<mode>, dfp_dscri_<mode>): Update size
22505 attribute for Power10.
22506 * config/rs6000/mma.md (*movoo): Likewise.
22507 * config/rs6000/rs6000.md (define_attr "size"): Add 256.
22508 (define_mode_attr bits): Add DD/TD modes.
22509 * config/rs6000/sync.md (load_quadpti, store_quadpti, load_lockedpti,
22510 store_conditionalpti): Update size attribute for Power10.
22512 2021-03-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22515 * config/sparc/t-sparc (tree-ssanames.o-warn): Don't error for
22516 -Wuninitialized, -Wmaybe-uninitialized.
22517 (wide-int.o-warn): Likewise.
22519 2021-03-03 Richard Earnshaw <rearnsha@arm.com>
22521 * common/config/arm/arm-common.c: Include configargs.h.
22522 (arm_config_default): New function.
22523 (arm_target_mode): Renamed from arm_target_thumb_only. Handle
22524 processors that do not support Thumb. Take into account the
22525 --with-mode configuration setting for selecting the default.
22526 * config/arm/arm.h (OPTION_DEFAULT_SPECS): Remove entry for 'mode'.
22527 (TARGET_MODE_SPEC_FUNCTIONS): Update for function name change.
22529 2021-03-03 Martin Liska <mliska@suse.cz>
22531 PR gcov-profile/97461
22532 * gcov-io.h (GCOV_PREALLOCATED_KVP): Remove.
22534 2021-03-03 Eric Botcazou <ebotcazou@adacore.com>
22537 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
22538 point back the hard frame pointer to its default location when the
22539 frame is larger than SEH_MAX_FRAME_SIZE.
22541 2021-03-03 Jakub Jelinek <jakub@redhat.com>
22544 * config/i386/predicates.md (logic_operator): New define_predicate.
22545 * config/i386/i386.md (mov + mem using comm arith peephole2):
22546 Punt if operands[1] is EXT_REX_SSE_REGNO_P, AVX512BW is not enabled
22547 and the inner mode is [QH]Imode.
22549 2021-03-03 Jakub Jelinek <jakub@redhat.com>
22552 * dwarf2out.c (dw_loc_list_struct): Add end_entry member.
22553 (new_loc_list): Clear end_entry.
22554 (output_loc_list): Only use DW_LLE_startx_length for -gsplit-dwarf
22555 if HAVE_AS_LEB128, otherwise use DW_LLE_startx_endx. Fix comment
22557 (index_location_lists): For dwarf_version >= 5 without HAVE_AS_LEB128,
22558 initialize also end_entry.
22560 2021-03-03 Jakub Jelinek <jakub@redhat.com>
22563 * cfgrtl.c (fixup_partitions): When changing some bbs from hot to cold
22564 partitions, if in non-layout mode after reorder_blocks also move
22565 affected blocks to ensure a single partition transition.
22567 2021-03-03 Jason Merrill <jason@redhat.com>
22570 * cgraphunit.c (process_function_and_variable_attributes): Don't
22571 warn about flatten on an alias if the target also has it.
22572 * cgraph.h (symtab_node::get_alias_target_tree): New.
22574 2021-03-02 David Edelsohn <dje.gcc@gmail.com>
22576 * config/rs6000/rs6000.md (tls_get_tpointer_internal): Prepend
22577 period to symbol name.
22578 (tls_get_addr_internal<mode>): Same.
22580 2021-03-02 David Malcolm <dmalcolm@redhat.com>
22583 * diagnostic-show-locus.c
22584 (selftest::test_one_liner_many_fixits_2): Fix accidental usage of
22587 2021-03-02 Martin Sebor <msebor@redhat.com>
22589 PR middle-end/99276
22590 * builtins.c (warn_for_access): Remove stray warning text.
22592 2021-03-02 Martin Sebor <msebor@redhat.com>
22594 PR middle-end/99295
22595 * doc/extend.texi (attribute malloc): Reword and clarify nonaliasing
22598 2021-03-02 Jakub Jelinek <jakub@redhat.com>
22601 * dwarf2out.c (output_macinfo_op): Use DW_MACRO_*_str* even with
22602 -gdwarf-5 -gstrict-dwarf. For -gsplit-dwarf -gdwarf-5 use
22603 DW_MACRO_*_strx instead of DW_MACRO_*_strp. Handle
22604 DW_MACRO_define_strx and DW_MACRO_undef_strx.
22605 (save_macinfo_strings): Use DW_MACRO_*_str* even with
22606 -gdwarf-5 -gstrict-dwarf. Handle DW_MACRO_define_strx and
22607 DW_MACRO_undef_strx.
22609 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
22611 * config/s390/s390-builtin-types.def (BT_FN_V4SF_V8HI_UINT): New
22613 (BT_FN_V8HI_V8HI_UINT): Likewise.
22614 (BT_FN_V8HI_V4SF_V4SF_UINT): Likewise.
22615 * config/s390/s390-builtins.def (B_NNPA): New macro definition.
22616 (s390_vclfnhs, s390_vclfnls, s390_vcrnfs, s390_vcfn, s390_vcnf):
22617 New builtin definitions.
22618 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Bump
22619 vector extension version.
22620 * config/s390/s390.c (s390_expand_builtin): Check if builtins are
22621 available with current -march level.
22622 * config/s390/s390.md (UNSPEC_NNPA_VCLFNHS_V8HI)
22623 (UNSPEC_NNPA_VCLFNLS_V8HI, UNSPEC_NNPA_VCRNFS_V8HI)
22624 (UNSPEC_NNPA_VCFN_V8HI, UNSPEC_NNPA_VCNF_V8HI): New constants.
22625 * config/s390/vecintrin.h (vec_extend_to_fp32_hi): New macro.
22626 (vec_extend_to_fp32_lo): Likewise.
22627 (vec_round_from_fp32): Likewise.
22628 (vec_convert_to_fp16): Likewise.
22629 (vec_convert_from_fp16): Likewise.
22630 * config/s390/vx-builtins.md (vclfnhs_v8hi): New insn pattern.
22631 (vclfnls_v8hi): Likewise.
22632 (vcrnfs_v8hi): Likewise.
22633 (vcfn_v8hi): Likewise.
22634 (vcnf_v8hi): Likewise.
22636 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
22638 * common/config/s390/s390-common.c (processor_flags_table): New entry.
22639 * config.gcc: Enable arch14 for --with-arch and --with-tune.
22640 * config/s390/driver-native.c (s390_host_detect_local_cpu): Pick
22641 arch14 for unknown CPU models.
22642 * config/s390/s390-opts.h (enum processor_type): Add PROCESSOR_ARCH14.
22643 * config/s390/s390.c (s390_issue_rate): Add case for PROCESSOR_ARCH14.
22644 (s390_get_sched_attrmask): Likewise.
22645 (s390_get_unit_mask): Likewise.
22646 * config/s390/s390.h (enum processor_flags): Add PF_NNPA and PF_ARCH14.
22647 (TARGET_CPU_ARCH14, TARGET_CPU_ARCH14_P, TARGET_CPU_NNPA)
22648 (TARGET_CPU_NNPA_P, TARGET_ARCH14, TARGET_ARCH14_P, TARGET_NNPA)
22649 (TARGET_NNPA_P): New macro definitions.
22650 * config/s390/s390.md ("cpu_facility", "enabled"): Add arch14 and nnpa.
22651 * config/s390/s390.opt: Add PROCESSOR_ARCH14.
22653 2021-03-02 Jakub Jelinek <jakub@redhat.com>
22655 PR middle-end/95757
22656 * tree-vrp.c (register_edge_assert_for): Remove superfluous ()s around
22657 condition. Call register_edge_assert_for_1 for == 0, != 0, == 1 and
22658 != 1 comparisons if name is lhs of a comparison.
22660 2021-03-01 Iain Sandoe <iain@sandoe.co.uk>
22664 * config/darwin-protos.h (darwin_should_restore_cfa_state): New.
22665 * config/darwin.c (darwin_should_restore_cfa_state): New.
22666 * config/darwin.h (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New.
22667 * doc/tm.texi: Regenerated.
22668 * doc/tm.texi.in: Document TARGET_ASM_SHOULD_RESTORE_CFA_STATE.
22669 * dwarf2cfi.c (connect_traces): If the target requests, restore
22670 the CFA expression after a DW_CFA_restore.
22671 * target.def (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New hook.
22673 2021-03-01 Martin Liska <mliska@suse.cz>
22676 * optc-save-gen.awk: Add 4 more exceptions.
22678 2021-03-01 Nathan Sidwell <nathan@acm.org>
22681 * tree.h (TYPE_ALIGN_RAW): New accessor.
22682 (TYPE_ALIGN): Use it.
22684 2021-03-01 Jan Hubicka <jh@suse.cz>
22687 * ipa-fnsummary.c (compute_fn_summary): Fix sanity check.
22689 2021-03-01 Eric Botcazou <ebotcazou@adacore.com>
22692 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
22693 point the hard frame pointer to the SSE register save area instead
22694 of the general register save area. Perform only minimal adjustment
22695 for small frames if it is initially not correctly aligned.
22696 (ix86_expand_prologue): Remove early saves for a SEH target.
22697 * config/i386/winnt.c (struct seh_frame_state): Document constraint.
22699 2021-02-28 Jakub Jelinek <jakub@redhat.com>
22702 * ipa.c (symbol_table::remove_unreachable_nodes): Fix a comment
22703 typo - referneced -> referenced.
22704 * tree.c (component_ref_size): Fix comment typo -
22705 refernce -> reference.
22706 * tree-ssa-alias.c (access_path_may_continue_p): Fix comment typo -
22707 traling -> trailing.
22708 (aliasing_component_refs_p): Fix comment typos -
22709 refernce -> reference and refernece -> reference and
22710 traling -> trailing.
22711 (nonoverlapping_refs_since_match_p): Fix comment typo -
22712 referneces -> references.
22713 * doc/invoke.texi (--param modref-max-bases): Fix a typo -
22714 referneces -> references.
22716 2021-02-27 Iain Sandoe <iain@sandoe.co.uk>
22718 * config/host-darwin.c (darwin_gt_pch_use_address): Modify
22719 diagnostic message to avoid use of a contraction and format
22722 2021-02-27 Jakub Jelinek <jakub@redhat.com>
22725 * gcse.c (gcse_or_cprop_is_too_expensive): Use %wu instead of
22726 HOST_WIDE_INT_PRINT_UNSIGNED in warning format string.
22727 * ipa-devirt.c (ipa_odr_read_section): Use %wd instead of
22728 HOST_WIDE_INT_PRINT_DEC in inform format string. Fix comment
22731 2021-02-26 Richard Biener <rguenther@suse.de>
22733 PR middle-end/99281
22734 * expr.c (store_field): For calls with return-slot optimization
22735 and addressable return type expand the store directly.
22737 2021-02-26 Richard Biener <rguenther@suse.de>
22740 * builtins.c (warn_string_no_nul): Fix diagnostic formatting.
22742 2021-02-26 Peter Bergner <bergner@linux.ibm.com>
22745 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Replace assert
22748 2021-02-26 Aaron Sawdey <acsawdey@linux.ibm.com>
22750 * config.gcc: Add rs6000-pcrel-opt.o.
22751 * config/rs6000/rs6000-pcrel-opt.c: New file.
22752 * config/rs6000/pcrel-opt.md: New file.
22753 * config/rs6000/predicates.md: Add d_form_memory predicate.
22754 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_PCREL_OPT.
22755 * config/rs6000/rs6000-passes.def: Add pass_pcrel_opt.
22756 * config/rs6000/rs6000-protos.h: Add reg_to_non_prefixed(),
22757 pcrel_opt_valid_mem_p(), output_pcrel_opt_reloc(),
22758 and make_pass_pcrel_opt().
22759 * config/rs6000/rs6000.c (reg_to_non_prefixed): Make global.
22760 (rs6000_option_override_internal): Add pcrel-opt.
22761 (rs6000_delegitimize_address): Support pcrel-opt.
22762 (rs6000_opt_masks): Add pcrel-opt.
22763 (pcrel_opt_valid_mem_p): New function.
22764 (reg_to_non_prefixed): Make global.
22765 (rs6000_asm_output_opcode): Reset prepend_p_to_next_insn.
22766 (output_pcrel_opt_reloc): New function.
22767 * config/rs6000/rs6000.md (loads_extern_addr): New attr.
22768 (pcrel_extern_addr): Set loads_extern_addr.
22769 Add include for pcrel-opt.md.
22770 * config/rs6000/rs6000.opt: Add -mpcrel-opt.
22771 * config/rs6000/t-rs6000: Add rules for pcrel-opt.c and
22774 2021-02-26 YunQiang Su <yunqiang.su@cipunited.com>
22777 * config/mips/mips.c (mips_expand_ext_as_unaligned_load):
22778 If TARGET_64BIT and dest is SUBREG, we check the width, if it
22779 equal to SImode, we use SImode operation, just like what we are
22782 2021-02-26 Marek Polacek <polacek@redhat.com>
22784 * builtins.c (warn_for_access): Fix typos.
22786 2021-02-25 Iain Sandoe <iain@sandoe.co.uk>
22788 * config/aarch64/aarch64.md (<optab>_rol<mode>3): Add a '#'
22789 mark in front of the immediate quantity.
22790 (<optab>_rolsi3_uxtw): Likewise.
22792 2021-02-25 Richard Earnshaw <rearnsha@arm.com>
22795 * config/arm/thumb2.md (nonsecure_call_reg_thumb2_fpcxt): New pattern.
22796 (nonsecure_call_value_reg_thumb2_fpcxt): Likewise.
22797 (nonsecure_call_reg_thumb2): Restrict to using r4 for the callee
22798 address and disable when the FPCXT is not available.
22799 (nonsecure_call_value_reg_thumb2): Likewise.
22801 2021-02-25 Nathan Sidwell <nathan@acm.org>
22804 * doc/invoke.texi (flang-info-module-cmi): Renamed option.
22806 2021-02-25 Tamar Christina <tamar.christina@arm.com>
22808 * tree-vect-slp.c (optimize_load_redistribution_1): Abort on NULL nodes.
22810 2021-02-25 Richard Biener <rguenther@suse.de>
22812 PR tree-optimization/99253
22813 * tree-vect-loop.c (check_reduction_path): First compute
22814 code, then verify out-of-loop uses.
22816 2021-02-25 Jakub Jelinek <jakub@redhat.com>
22819 * match.pd ((T)(A) + CST -> (T)(A + CST)): Add :s to convert.
22821 2021-02-25 Jakub Jelinek <jakub@redhat.com>
22823 PR tree-optimization/80635
22824 * tree-vrp.c (vrp_simplify_cond_using_ranges): Also handle
22825 VIEW_CONVERT_EXPR if modes are the same, innerop is integral and
22826 has mode precision.
22828 2021-02-25 Richard Biener <rguenther@suse.de>
22830 * tree-vect-slp.c (optimize_load_redistribution_1): Delay
22831 load_map population.
22832 (vect_match_slp_patterns_2): Revert part of last change.
22833 (vect_analyze_slp): Do not interleave optimize_load_redistribution
22834 with pattern detection but do it afterwards. Dump the
22835 whole SLP graph after pattern recognition and load
22836 redistribution optimization finished.
22838 2021-02-24 Jakub Jelinek <jakub@redhat.com>
22841 * omp-low.c (struct omp_context): Add teams_nested_p and
22842 nonteams_nested_p members.
22843 (scan_omp_target): Diagnose teams nested inside of target with other
22844 directives strictly nested inside of the same target.
22845 (check_omp_nesting_restrictions): Set ctx->teams_nested_p or
22846 ctx->nonteams_nested_p as needed.
22848 2021-02-24 Vladimir N. Makarov <vmakarov@redhat.com>
22850 PR inline-asm/99123
22851 * lra-constraints.c (uses_hard_regs_p): Don't use decompose_mem_address.
22853 2021-02-24 Hans-Peter Nilsson <hp@axis.com>
22855 * config/cris/cris.c (cris_expand_prologue): Set
22856 current_function_static_stack_size, if flag_stack_usage_info.
22858 2021-02-24 Pat Haugen <pthaugen@linux.ibm.com>
22860 * config/rs6000/rs6000.c (next_insn_prefixed_p): Rename.
22861 (rs6000_final_prescan_insn): Adjust.
22862 (rs6000_asm_output_opcode): Likewise.
22864 2021-02-24 Martin Sebor <msebor@redhat.com>
22866 PR middle-end/97172
22867 * attribs.c (attr_access::free_lang_data): Clear attribute arg spec
22868 from function arguments.
22870 2021-02-24 Tamar Christina <tamar.christina@arm.com>
22872 PR tree-optimization/99220
22873 * tree-vect-slp.c (optimize_load_redistribution_1): Remove
22874 node from cache when it's about to be deleted.
22876 2021-02-24 Jakub Jelinek <jakub@redhat.com>
22878 PR tree-optimization/99225
22879 * fold-const.c (fold_binary_loc) <case NE_EXPR>: In (x & (1 << y)) != 0
22880 to ((x >> y) & 1) != 0 simplifications use build_one_cst instead of
22881 build_int_cst (..., 1). Formatting fixes.
22883 2021-02-24 Tamar Christina <tamar.christina@arm.com>
22885 PR tree-optimization/99149
22886 * tree-vect-slp-patterns.c (vect_detect_pair_op): Don't recreate the
22888 (vect_slp_reset_pattern): Remove.
22889 (complex_fma_pattern::matches): Remove call to vect_slp_reset_pattern.
22890 (complex_mul_pattern::build, complex_fma_pattern::build,
22891 complex_fms_pattern::build): Fix ref counts.
22892 * tree-vect-slp.c (vect_free_slp_tree): Undo SLP only pattern relevancy
22893 when node is being deleted.
22894 (vect_match_slp_patterns_2): Correct result of cache hit on patterns.
22895 (vect_schedule_slp): Invalidate SLP_TREE_REPRESENTATIVE of removed
22897 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize value.
22899 2021-02-24 Matthias Klose <doko@ubuntu.com>
22902 2020-12-07 Matthias Klose <doko@ubuntu.com>
22904 * genextract.c (print_header): Undefine ENABLE_RTL_CHECKING
22905 and ENABLE_RTL_FLAG_CHECKING.
22907 2021-02-24 Richard Biener <rguenther@suse.de>
22910 * builtins.c (fold_builtin_next_arg): Avoid NULL arg.
22912 2021-02-23 Peter Bergner <bergner@linux.ibm.com>
22914 * config/rs6000/mma.md (mma_assemble_pair): Rename from this...
22915 (vsx_assemble_pair): ...to this.
22916 (*mma_assemble_pair): Rename from this...
22917 (*vsx_assemble_pair): ...to this.
22918 (mma_disassemble_pair): Rename from this...
22919 (vsx_disassemble_pair): ...to this.
22920 (*mma_disassemble_pair): Rename from this...
22921 (*vsx_disassemble_pair): ...to this.
22922 * config/rs6000/rs6000-builtin.def (BU_MMA_V2, BU_MMA_V3,
22923 BU_COMPAT): New macros.
22924 (mma_assemble_pair): Rename from this...
22925 (vsx_assemble_pair): ...to this.
22926 (mma_disassemble_pair): Rename from this...
22927 (vsx_disassemble_pair): ...to this.
22928 (mma_assemble_pair): New compatibility built-in.
22929 (mma_disassemble_pair): Likewise.
22930 * config/rs6000/rs6000-call.c (struct builtin_compatibility): New.
22931 (RS6000_BUILTIN_COMPAT): Define.
22932 (bdesc_compat): New.
22933 (mma_expand_builtin): Use VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
22934 (rs6000_gimple_fold_mma_builtin): Use MMA_BUILTIN_DISASSEMBLE_PAIR
22935 and VSX_BUILTIN_ASSEMBLE_PAIR.
22936 (rs6000_init_builtins): Register compatibility built-ins.
22937 (mma_init_builtins): Use VSX_BUILTIN_ASSEMBLE_PAIR,
22938 VSX_BUILTIN_ASSEMBLE_PAIR_INTERNAL, VSX_BUILTIN_DISASSEMBLE_PAIR and
22939 VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
22940 * doc/extend.texi (__builtin_mma_assemble_pair): Rename from this...
22941 (__builtin_vsx_assemble_pair): ...to this.
22942 (__builtin_mma_disassemble_pair): Rename from this...
22943 (__builtin_vsx_disassemble_pair): ...to this.
22945 2021-02-23 Martin Liska <mliska@suse.cz>
22948 * ipa-icf.c (sem_variable::merge): Do not merge 2 variables
22949 with different alignment. That leads to an invalid red zone
22950 size allocated in runtime.
22952 2021-02-23 Jakub Jelinek <jakub@redhat.com>
22954 PR tree-optimization/99204
22955 * fold-const.c (fold_read_from_constant_string): Check that
22956 tree_fits_uhwi_p (index) rather than just that index is INTEGER_CST.
22958 2021-02-23 Segher Boessenkool <segher@kernel.crashing.org>
22959 Kewen Lin <linkw@gcc.gnu.org>
22961 * config/rs6000/rs6000.md (*rotl<mode>3_insert_3): Renamed to...
22962 (rotl<mode>3_insert_3): ...this.
22963 (plus_ior_xor): New code_iterator.
22964 (define_split for GPR rl*imi): New splitter.
22965 * config/rs6000/vsx.md (vsx_init_v4si): Use gen_rotldi3_insert_3
22966 for integer merging.
22968 2021-02-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22970 * config/aarch64/aarch64-tuning-flags.def (cse_sve_vl_constants):
22972 * config/aarch64/aarch64.md (add<mode>3): Force CONST_POLY_INT immediates
22973 into a register when the above is enabled.
22974 * config/aarch64/aarch64.c (neoversev1_tunings):
22975 AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
22976 (aarch64_rtx_costs): Use AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
22978 2021-02-22 Hans-Peter Nilsson <hp@axis.com>
22980 * config/cris/cris.c (cris_print_operand) <'T'>: Change
22981 valid operand from is now an addi mult-value to shift-value.
22982 * config/cris/cris.md (*addi): Change expression of scaled
22983 operand from mult to ashift.
22984 * config/cris/cris.md (*addi_reload): New insn_and_split.
22986 2021-02-22 John David Anglin <danglin@gcc.gnu.org>
22989 * config/pa/pa.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Define as
22990 hook_bool_const_tree_hwi_hwi_const_tree_true.
22991 (pa_asm_output_mi_thunk): Add support for nonzero vcall_offset.
22993 2021-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
22995 PR rtl-optimization/98791
22996 * ira-conflicts.c (process_regs_for_copy): Don't create allocno copies
22997 for unordered modes.
22999 2021-02-22 Martin Liska <mliska@suse.cz>
23001 * tree-inline.c (inline_forbidden_p): Set
23002 inline_forbidden_reason.
23004 2021-02-22 Richard Biener <rguenther@suse.de>
23006 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Dump
23009 2021-02-22 Richard Biener <rguenther@suse.de>
23011 PR tree-optimization/99165
23012 * gimple-ssa-store-merging.c (pass_store_merging::process_store):
23013 Accumulate changed to ret.
23015 2021-02-21 Uros Bizjak <ubizjak@gmail.com>
23018 2020-12-09 Uroš Bizjak <ubizjak@gmail.com>
23020 * config/i386/i386.h (REG_ALLOC_ORDER): Remove
23022 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
23025 * config/s390/vector.md (trunctf<DFP_ALL:mode>2_vr): New
23027 (trunctf<DFP_ALL:mode>2): Likewise.
23028 (trunctdtf2_vr): Likewise.
23029 (trunctdtf2): Likewise.
23030 (extend<DFP_ALL:mode>tf2_vr): Likewise.
23031 (extend<DFP_ALL:mode>tf2): Likewise.
23032 (extendtftd2_vr): Likewise.
23033 (extendtftd2): Likewise.
23035 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
23037 * config/s390/vector.md (*fprx2_to_tf): Rename to fprx2_to_tf,
23038 add memory alternative.
23039 (tf_to_fprx2): New pattern.
23041 2021-02-19 Martin Sebor <msebor@redhat.com>
23044 * attribs.c (init_attr_rdwr_indices): Guard vblist use.
23045 (attr_access::free_lang_data): Remove a spurious test.
23047 2021-02-19 Nathan Sidwell <nathan@acm.org>
23049 * doc/invoke.texi (flang-info-module-read): Document.
23051 2021-02-19 Martin Liska <mliska@suse.cz>
23053 PR translation/99167
23054 * params.opt: Fix typo.
23056 2021-02-19 Richard Biener <rguenther@suse.de>
23058 PR middle-end/99122
23059 * tree-inline.c (inline_forbidden_p): Do not inline functions
23060 with VLA arguments or return value.
23062 2021-02-19 Jakub Jelinek <jakub@redhat.com>
23065 * config/arm/arm.md (*stack_protect_combined_set_insn,
23066 *stack_protect_combined_test_insn): If force_const_mem result
23067 is not valid general operand, force its address into the destination
23070 2021-02-19 Jakub Jelinek <jakub@redhat.com>
23073 * tree-cfg.c (gimple_merge_blocks): If bb a starts with eh landing
23074 pad or non-local label, put FORCED_LABELs from bb b after that label
23075 rather than before it.
23077 2021-02-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
23080 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3): Use
23081 expand_vector_broadcast' to emit the vec_duplicate operand.
23083 2021-02-18 Vladimir N. Makarov <vmakarov@redhat.com>
23085 PR rtl-optimization/96264
23086 * lra-remat.c (reg_overlap_for_remat_p): Check also output insn
23089 2021-02-18 H.J. Lu <hjl.tools@gmail.com>
23092 * varasm.c (get_section): Replace SUPPORTS_SHF_GNU_RETAIN with
23093 looking up the retain attribute.
23094 (resolve_unique_section): Likewise.
23095 (get_variable_section): Likewise.
23096 (switch_to_section): Likewise. Warn when a symbol without the
23097 retain attribute and a symbol with the retain attribute are
23098 placed in the section with the same name, instead of the used
23100 * doc/extend.texi: Document the "retain" attribute.
23102 2021-02-18 Nathan Sidwell <nathan@acm.org>
23105 * doc/invoke.texi (flang-info-include-translate): Document header
23108 2021-02-18 Richard Biener <rguenther@suse.de>
23110 PR middle-end/99122
23111 * ipa-fnsummary.c (analyze_function_body): Set
23112 CIF_FUNCTION_NOT_INLINABLE for VLA parameter calls.
23113 * tree-inline.c (insert_init_debug_bind): Pass NULL for
23114 error_mark_node values.
23115 (force_value_to_type): Do not build V_C_Es for WITH_SIZE_EXPR
23117 (setup_one_parameter): Delay force_value_to_type until when
23120 2021-02-18 Hans-Peter Nilsson <hp@axis.com>
23122 PR tree-optimization/99142
23123 * match.pd (clz cmp 0): Gate replacement on single_use of clz result.
23125 2021-02-18 Jakub Jelinek <jakub@redhat.com>
23127 * wide-int-bitmask.h (wide_int_bitmask::wide_int_bitmask (),
23128 wide_int_bitmask::wide_int_bitmask (uint64_t),
23129 wide_int_bitmask::wide_int_bitmask (uint64_t, uint64_t),
23130 wide_int_bitmask::operator ~ () const,
23131 wide_int_bitmask::operator | (wide_int_bitmask) const,
23132 wide_int_bitmask::operator & (wide_int_bitmask) const): Use constexpr
23134 * config/i386/i386.h (PTA_3DNOW, PTA_3DNOW_A, PTA_64BIT, PTA_ABM,
23135 PTA_AES, PTA_AVX, PTA_BMI, PTA_CX16, PTA_F16C, PTA_FMA, PTA_FMA4,
23136 PTA_FSGSBASE, PTA_LWP, PTA_LZCNT, PTA_MMX, PTA_MOVBE, PTA_NO_SAHF,
23137 PTA_PCLMUL, PTA_POPCNT, PTA_PREFETCH_SSE, PTA_RDRND, PTA_SSE, PTA_SSE2,
23138 PTA_SSE3, PTA_SSE4_1, PTA_SSE4_2, PTA_SSE4A, PTA_SSSE3, PTA_TBM,
23139 PTA_XOP, PTA_AVX2, PTA_BMI2, PTA_RTM, PTA_HLE, PTA_PRFCHW, PTA_RDSEED,
23140 PTA_ADX, PTA_FXSR, PTA_XSAVE, PTA_XSAVEOPT, PTA_AVX512F, PTA_AVX512ER,
23141 PTA_AVX512PF, PTA_AVX512CD, PTA_NO_TUNE, PTA_SHA, PTA_PREFETCHWT1,
23142 PTA_CLFLUSHOPT, PTA_XSAVEC, PTA_XSAVES, PTA_AVX512DQ, PTA_AVX512BW,
23143 PTA_AVX512VL, PTA_AVX512IFMA, PTA_AVX512VBMI, PTA_CLWB, PTA_MWAITX,
23144 PTA_CLZERO, PTA_NO_80387, PTA_PKU, PTA_AVX5124VNNIW, PTA_AVX5124FMAPS,
23145 PTA_AVX512VPOPCNTDQ, PTA_SGX, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES,
23146 PTA_AVX512VBMI2, PTA_VPCLMULQDQ, PTA_AVX512BITALG, PTA_RDPID,
23147 PTA_PCONFIG, PTA_WBNOINVD, PTA_AVX512VP2INTERSECT, PTA_PTWRITE,
23148 PTA_AVX512BF16, PTA_WAITPKG, PTA_MOVDIRI, PTA_MOVDIR64B, PTA_ENQCMD,
23149 PTA_CLDEMOTE, PTA_SERIALIZE, PTA_TSXLDTRK, PTA_AMX_TILE, PTA_AMX_INT8,
23150 PTA_AMX_BF16, PTA_UINTR, PTA_HRESET, PTA_KL, PTA_WIDEKL, PTA_AVXVNNI,
23151 PTA_X86_64_BASELINE, PTA_X86_64_V2, PTA_X86_64_V3, PTA_X86_64_V4,
23152 PTA_CORE2, PTA_NEHALEM, PTA_WESTMERE, PTA_SANDYBRIDGE, PTA_IVYBRIDGE,
23153 PTA_HASWELL, PTA_BROADWELL, PTA_SKYLAKE, PTA_SKYLAKE_AVX512,
23154 PTA_CASCADELAKE, PTA_COOPERLAKE, PTA_CANNONLAKE, PTA_ICELAKE_CLIENT,
23155 PTA_ICELAKE_SERVER, PTA_TIGERLAKE, PTA_SAPPHIRERAPIDS, PTA_ALDERLAKE,
23156 PTA_KNL, PTA_BONNELL, PTA_SILVERMONT, PTA_GOLDMONT, PTA_GOLDMONT_PLUS,
23157 PTA_TREMONT, PTA_KNM): Use constexpr instead of const.
23159 2021-02-18 Jakub Jelinek <jakub@redhat.com>
23161 PR middle-end/99109
23162 * gimple-array-bounds.cc (build_zero_elt_array_type): Rename to ...
23163 (build_printable_array_type): ... this. Add nelts argument. For
23164 overaligned eltype, use TYPE_MAIN_VARIANT (eltype) instead. If
23165 nelts, call build_array_type_nelts.
23166 (array_bounds_checker::check_mem_ref): Use build_printable_array_type
23167 instead of build_zero_elt_array_type and build_array_type_nelts.
23169 2021-02-18 Jakub Jelinek <jakub@redhat.com>
23172 * config/i386/i386.c (distance_non_agu_define): Don't call
23173 extract_insn_cached here.
23174 (ix86_lea_outperforms): Save and restore recog_data around call
23175 to distance_non_agu_define and distance_agu_use.
23176 (ix86_ok_to_clobber_flags): Remove.
23177 (ix86_avoid_lea_for_add): Don't call ix86_ok_to_clobber_flags.
23178 (ix86_avoid_lea_for_addr): Likewise. Adjust function comment.
23179 * config/i386/i386.md (*lea<mode>): Change from define_insn_and_split
23180 into define_insn. Move the splitting to define_peephole2 and
23181 check there using peep2_regno_dead_p if FLAGS_REG is dead.
23183 2021-02-17 Julian Brown <julian@codesourcery.com>
23185 * gimplify.c (gimplify_scan_omp_clauses): Handle ATTACH_DETACH
23188 2021-02-17 Xi Ruoyao <xry111@mengyan1223.wang>
23191 * config/mips/mips.c (mips_symbol_insns): Do not use
23192 MSA_SUPPORTED_MODE_P if mode is MAX_MACHINE_MODE.
23194 2021-02-16 Vladimir N. Makarov <vmakarov@redhat.com>
23196 PR inline-asm/98096
23197 * stmt.c (resolve_operand_name_1): Take inout operands into account
23198 for access to labels by names.
23199 * doc/extend.texi: Describe counting operands for accessing labels.
23201 2021-02-16 Richard Biener <rguenther@suse.de>
23203 PR tree-optimization/38474
23204 * tree-ssa-structalias.c (variable_info::address_taken): New.
23205 (new_var_info): Initialize address_taken.
23206 (process_constraint): Set address_taken.
23207 (solve_constraints): Use the new address_taken flag rather
23208 than is_reg_var for sorting variables.
23209 (dump_constraint): Dump the variable number if the name
23212 2021-02-16 Jakub Jelinek <jakub@redhat.com>
23215 * tree-vect-stmts.c (vectorizable_simd_clone_call): For num_calls != 1
23216 multiply by 4096 and for inbranch by 8192.
23217 * config/i386/i386.c (ix86_simd_clone_usable): For TARGET_AVX512F,
23218 return 3, 2 or 1 for mangle letters 'b', 'c' or 'd'.
23220 2021-02-15 Maya Rashish <coypu@sdf.org>
23222 * config/aarch64/aarch64.c (aarch64_init_builtins):
23223 Call SUBTARGET_INIT_BUILTINS.
23225 2021-02-15 Peter Bergner <bergner@linux.ibm.com>
23227 PR rtl-optimization/98872
23228 * init-regs.c (initialize_uninitialized_regs): Skip initialization
23229 if CONST0_RTX is NULL.
23231 2021-02-15 Richard Sandiford <richard.sandiford@arm.com>
23233 PR rtl-optimization/98863
23234 * rtl-ssa/functions.h (function_info::bb_live_out_info): Delete.
23235 (function_info::build_info): Turn into a declaration, moving the
23236 definition to internals.h.
23237 (function_info::bb_walker): Declare.
23238 (function_info::create_reg_use): Likewise.
23239 (function_info::calculate_potential_phi_regs): Take a build_info
23241 (function_info::place_phis, function_info::create_ebbs): Declare.
23242 (function_info::calculate_ebb_live_in_for_debug): Likewise.
23243 (function_info::populate_backedge_phis): Delete.
23244 (function_info::start_block, function_info::end_block): Declare.
23245 (function_info::populate_phi_inputs): Delete.
23246 (function_info::m_potential_phi_regs): Move information to build_info.
23247 * rtl-ssa/internals.h: New file.
23248 (function_info::bb_phi_info): New class.
23249 (function_info::build_info): Moved from functions.h.
23250 Add a constructor and destructor.
23251 (function_info::build_info::ebb_use): Delete.
23252 (function_info::build_info::ebb_def): Likewise.
23253 (function_info::build_info::bb_live_out): Likewise.
23254 (function_info::build_info::tmp_ebb_live_in_for_debug): New variable.
23255 (function_info::build_info::potential_phi_regs): Likewise.
23256 (function_info::build_info::potential_phi_regs_for_debug): Likewise.
23257 (function_info::build_info::ebb_def_regs): Likewise.
23258 (function_info::build_info::bb_phis): Likewise.
23259 (function_info::build_info::bb_mem_live_out): Likewise.
23260 (function_info::build_info::bb_to_rpo): Likewise.
23261 (function_info::build_info::def_stack): Likewise.
23262 (function_info::build_info::old_def_stack_limit): Likewise.
23263 * rtl-ssa/internals.inl (function_info::build_info::record_reg_def):
23264 Remove the regno argument. Push the previous definition onto the
23265 definition stack where necessary.
23266 * rtl-ssa/accesses.cc: Include internals.h.
23267 * rtl-ssa/changes.cc: Likewise.
23268 * rtl-ssa/blocks.cc: Likewise.
23269 (function_info::build_info::build_info): Define.
23270 (function_info::build_info::~build_info): Likewise.
23271 (function_info::bb_walker): New class.
23272 (function_info::bb_walker::bb_walker): Define.
23273 (function_info::add_live_out_use): Convert a logarithmic-complexity
23274 test into a linear one. Allow the same definition to be passed
23276 (function_info::calculate_potential_phi_regs): Moved from
23277 functions.cc. Take a build_info parameter and store the
23278 information there instead.
23279 (function_info::place_phis): New function.
23280 (function_info::add_entry_block_defs): Update call to record_reg_def.
23281 (function_info::calculate_ebb_live_in_for_debug): New function.
23282 (function_info::add_phi_nodes): Use bb_phis to decide which
23283 registers need phi nodes and initialize ebb_def_regs accordingly.
23284 Do not add degenerate phis here.
23285 (function_info::add_artificial_accesses): Use create_reg_use.
23286 Assert that all definitions are listed in the DF LR sets.
23287 Update call to record_reg_def.
23288 (function_info::record_block_live_out): Record live-out register
23289 values in the phis of successor blocks. Use the live-out set
23290 when processing the last block in an EBB, instead of always
23291 using the live-in sets of successor blocks. AND the live sets
23292 with the set of registers that have been defined in the EBB,
23293 rather than with all potential phi registers. Cope correctly
23294 with branches back to the start of the current EBB.
23295 (function_info::start_block): New function.
23296 (function_info::end_block): Likewise.
23297 (function_info::populate_phi_inputs): Likewise.
23298 (function_info::create_ebbs): Likewise.
23299 (function_info::process_all_blocks): Rewrite into a multi-phase
23301 * rtl-ssa/functions.cc: Include internals.h.
23302 (function_info::calculate_potential_phi_regs): Move to blocks.cc.
23303 (function_info::init_function_data): Remove caller.
23304 * rtl-ssa/insns.cc: Include internals.h
23305 (function_info::create_reg_use): New function. Lazily any
23306 degenerate phis needed by the linear RPO view.
23307 (function_info::record_use): Use create_reg_use. When processing
23308 debug uses, use potential_phi_regs and test it before checking
23309 whether the register is live on entry to the current EBB. Lazily
23310 calculate ebb_live_in_for_debug.
23311 (function_info::record_call_clobbers): Update call to record_reg_def.
23312 (function_info::record_def): Likewise.
23314 2021-02-15 Martin Liska <mliska@suse.cz>
23316 * toplev.c (init_asm_output): Free output of
23317 gen_command_line_string function.
23318 (process_options): Likewise.
23320 2021-02-15 Martin Liska <mliska@suse.cz>
23322 * params.opt: Add 2 missing Param keywords.
23324 2021-02-15 Eric Botcazou <ebotcazou@adacore.com>
23326 * df-core.c (df_worklist_dataflow_doublequeue): Use proper cast.
23328 2021-02-15 Jakub Jelinek <jakub@redhat.com>
23330 PR tree-optimization/99079
23331 * match.pd (A % (pow2pcst << N) -> A & ((pow2pcst << N) - 1)): Remove
23332 useless tree_nop_conversion_p (type, TREE_TYPE (@3)) check. Instead
23333 require both type and TREE_TYPE (@1) to be integral types and either
23334 type having smaller or equal precision, or TREE_TYPE (@1) being
23335 unsigned type, or type being signed type. If TREE_TYPE (@1)
23336 doesn't have wrapping overflow, perform the subtraction of one in
23339 2021-02-14 Jan Hubicka <hubicka@ucw.cz>
23340 Richard Biener <rguether@suse.de>
23343 * ipa-reference.c (ipa_init): Only conditinally initialize
23344 reference_vars_to_consider.
23345 (propagate): Conditionally deninitialize reference_vars_to_consider.
23346 (ipa_reference_write_optimization_summary): Sanity check that
23347 reference_vars_to_consider is not allocated.
23349 2021-02-13 Levy Hsu <admin@levyhsu.com>
23352 * config/riscv/riscv-shorten-memrefs.c (pass_shorten_memrefs): Add
23353 extend parameter to get_si_mem_base_reg declaration.
23354 (get_si_mem_base_reg): Add extend parameter. Set it.
23355 (analyze): Pass extend arg to get_si_mem_base_reg.
23356 (transform): Likewise. Use it when rewriting mems.
23357 * config/riscv/riscv.c (riscv_legitimize_move): Check for subword
23358 loads and emit sign/zero extending load followed by subreg move.
23360 2021-02-13 Jim Wilson <jimw@sifive.com>
23363 * config/riscv/riscv.c (riscv_compressed_lw_address_p): Drop early
23364 exit when !reload_completed. Only perform check for compressed reg
23365 if reload_completed.
23366 (riscv_rtx_costs): In MEM case, when optimizing for size and
23367 shorten memrefs, if not compressible, then increase cost.
23369 2021-02-13 Jakub Jelinek <jakub@redhat.com>
23371 PR rtl-optimization/98439
23372 * recog.c (pass_split_before_regstack::gate): Enable even when
23373 pass_split_before_sched2 is enabled if -fselective-scheduling2 is
23376 2021-02-13 Jakub Jelinek <jakub@redhat.com>
23379 * config/i386/mmx.md (*mmx_pshufd_1): Add a combine splitter for
23380 swap of V2SImode elements in memory into DImode memory rotate by 32.
23382 2021-02-12 Martin Sebor <msebor@redhat.com>
23384 * tree-pretty-print.c (print_generic_expr_to_str): Update comment.
23386 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
23388 * rtl-ssa/accesses.cc (function_info::make_use_available): Use
23389 m_temp_obstack rather than m_obstack to allocate the temporary use.
23391 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
23393 * df-problems.c (df_lr_bb_local_compute): Treat partial definitions
23394 as read-modify operations.
23396 2021-02-12 Richard Biener <rguenther@suse.de>
23398 PR middle-end/38474
23399 * ipa-fnsummary.c (unmodified_parm_1): Only walk when
23400 fbi->aa_walk_budget is bigger than zero. Update
23401 fbi->aa_walk_budget.
23402 (param_change_prob): Likewise.
23403 * ipa-prop.c (detect_type_change_from_memory_writes):
23404 Properly account walk_aliased_vdefs.
23405 (parm_preserved_before_stmt_p): Canonicalize updates.
23406 (parm_ref_data_preserved_p): Likewise.
23407 (parm_ref_data_pass_through_p): Likewise.
23408 (determine_known_aggregate_parts): Account own alias queries.
23410 2021-02-12 Martin Liska <mliska@suse.cz>
23412 * opts-common.c (decode_cmdline_option): Release werror_arg.
23413 * opts.c (gen_producer_string): Release output of
23414 gen_command_line_string.
23416 2021-02-12 Richard Biener <rguenther@suse.de>
23418 PR tree-optimization/38474
23419 * params.opt (-param=max-store-chains-to-track=): New param.
23420 (-param=max-stores-to-track=): Likewise.
23421 * doc/invoke.texi (max-store-chains-to-track): Document.
23422 (max-stores-to-track): Likewise.
23423 * gimple-ssa-store-merging.c (pass_store_merging::m_n_chains):
23425 (pass_store_merging::m_n_stores): Likewise.
23426 (pass_store_merging::terminate_and_process_chain): Update
23427 m_n_stores and m_n_chains.
23428 (pass_store_merging::process_store): Likewise. Terminate
23429 oldest chains if the number of stores or chains get too large.
23430 (imm_store_chain_info::terminate_and_process_chain): Dump
23433 2021-02-11 Eric Botcazou <ebotcazou@adacore.com>
23435 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
23436 the cold section, emit a nop before the directive if the previous
23437 active instruction can throw.
23439 2021-02-11 Peter Bergner <bergner@linux.ibm.com>
23442 * config/rs6000/predicates.md (mma_assemble_input_operand): Restrict
23443 memory addresses that are legal for quad word accesses.
23445 2021-02-11 Andrea Corallo <andrea.corallo@arm.com>
23448 * config/arm/thumb2.md (*doloop_end_internal): Generate
23449 alternative sequence to handle long range branches.
23451 2021-02-11 Joel Hutton <joel.hutton@arm.com>
23453 PR tree-optimization/98772
23454 * optabs-tree.c (supportable_half_widening_operation): New function
23455 to check for supportable V8QI->V8HI widening patterns.
23456 * optabs-tree.h (supportable_half_widening_operation): New function.
23457 * tree-vect-stmts.c (vect_create_half_widening_stmts): New function
23458 to create promotion stmts for V8QI->V8HI widening patterns.
23459 (vectorizable_conversion): Add case for V8QI->V8HI.
23461 2021-02-11 Richard Biener <rguenther@suse.de>
23463 * sparseset.h (SPARSESET_ELT_BITS): Remove.
23464 (SPARSESET_ELT_TYPE): Use unsigned int.
23465 * fwprop.c: Do not include sparseset.h.
23467 2021-02-10 Jakub Jelinek <jakub@redhat.com>
23470 * varasm.c (declare_weak): For -fsyntax-only, allow even
23471 TREE_ASM_WRITTEN function decls.
23473 2021-02-10 Jakub Jelinek <jakub@redhat.com>
23476 * config/i386/sse.md (fix<fixunssuffix>_truncv2sfv2di2,
23477 <insn>v8qiv8hi2, <insn>v8qiv8si2, <insn>v4qiv4si2, <insn>v4hiv4si2,
23478 <insn>v8qiv8di2, <insn>v4qiv4di2, <insn>v2qiv2di2, <insn>v4hiv4di2,
23479 <insn>v2hiv2di2, <insn>v2siv2di2): Force operands[1] into REG before
23480 calling simplify_gen_subreg on it.
23482 2021-02-10 Martin Liska <mliska@suse.cz>
23484 * config/nvptx/nvptx.c (nvptx_option_override): Use
23485 flag_patchable_function_entry instead of the removed
23486 function_entry_patch_area_size.
23488 2021-02-10 Martin Liska <mliska@suse.cz>
23490 PR tree-optimization/99002
23491 PR tree-optimization/99026
23492 * gimple-if-to-switch.cc (if_chain::is_beneficial): Fix memory
23493 leak when adjacent cases are merged.
23494 * tree-switch-conversion.c (switch_decision_tree::analyze_switch_statement): Use
23496 (make_pass_lower_switch): Remove trailing whitespace.
23497 * tree-switch-conversion.h (release_clusters): New.
23499 2021-02-10 Richard Biener <rguenther@suse.de>
23501 PR rtl-optimization/99054
23502 * cfgrtl.c (rtl-optimization/99054): Return an auto_vec.
23503 (fixup_partitions): Adjust.
23504 (rtl_verify_edges): Likewise.
23506 2021-02-10 Jakub Jelinek <jakub@redhat.com>
23508 PR middle-end/99007
23509 * gimplify.c (gimplify_scan_omp_clauses): For MEM_REF on reductions,
23510 temporarily disable gimplify_ctxp->into_ssa around gimplify_expr
23513 2021-02-10 Richard Biener <rguenther@suse.de>
23516 * ipa-pure-const.c (propagate_malloc): Use an auto_vec<>
23519 2021-02-10 Richard Biener <rguenther@suse.de>
23521 PR tree-optimization/99024
23522 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Only
23523 clear loop->aux if it is associated with the destroyed loop_vinfo.
23525 2021-02-10 Martin Liska <mliska@suse.cz>
23527 PR tree-optimization/99002
23528 * gimple-if-to-switch.cc (find_conditions): Fix memory leak
23531 2021-02-10 Martin Liska <mliska@suse.cz>
23534 * ipa-icf.c (sem_item::add_reference): Fix memory leak when
23535 a reference exists.
23537 2021-02-10 Jakub Jelinek <jakub@redhat.com>
23540 * dwarf2out.c (prune_unused_types_walk): Mark DW_TAG_variable DIEs
23541 at class scope for DWARF5+.
23543 2021-02-09 Eric Botcazou <ebotcazou@adacore.com>
23545 PR rtl-optimization/96015
23546 * reorg.c (skip_consecutive_labels): Minor comment tweaks.
23547 (relax_delay_slots): When deleting a jump to the next active
23548 instruction over a barrier, first delete the barrier if the
23549 jump is the only way to reach the target label.
23551 2021-02-09 Andre Vieira <andre.simoesdiasvieira@arm.com>
23553 * config/aarch64/aarch64-cost-tables.h: Add entries for vect.mul.
23554 * config/aarch64/aarch64.c (aarch64_rtx_mult_cost): Use vect.mul for
23555 vector multiplies and vect.alu for SSRA.
23556 * config/arm/aarch-common-protos.h (struct vector_cost_table): Define
23557 vect.mul cost field.
23558 * config/arm/aarch-cost-tables.h: Add entries for vect.mul.
23559 * config/arm/arm.c: Likewise.
23561 2021-02-09 Richard Biener <rguenther@suse.de>
23563 PR tree-optimization/98863
23564 * tree-ssa-sccvn.h (vn_avail::next_undo): Add.
23565 * tree-ssa-sccvn.c (last_pushed_avail): New global.
23566 (rpo_elim::eliminate_push_avail): Chain pushed avails.
23567 (unwind_state::avail_top): Add.
23568 (do_unwind): Rewrite unwinding of avail entries.
23569 (do_rpo_vn): Initialize last_pushed_avail and
23570 avail_top of the undo state.
23572 2021-02-09 Jakub Jelinek <jakub@redhat.com>
23574 PR middle-end/99004
23575 * calls.c (maybe_warn_rdwr_sizes): Change s0 and s1 type from
23576 const char * to char * and free those pointers after use.
23578 2021-02-09 Richard Biener <rguenther@suse.de>
23580 PR tree-optimization/99017
23581 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Allow
23582 zero vector cost entries.
23584 2021-02-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
23586 PR middle-end/98974
23587 * tree-vect-stmts.c (vectorizable_condition): Remove shadow vec_num
23588 parameter in vectorizable_condition.
23590 2021-02-08 Richard Biener <rguenther@suse.de>
23593 * tree.c (walk_tree_1): Walk VECTOR_CST elements.
23595 2021-02-08 Martin Liska <mliska@suse.cz>
23598 * cfgexpand.c (pass_expand::execute): Parse per-function option
23599 flag_patchable_function_entry and use it.
23600 * common.opt: Remove function_entry_patch_area_size and
23601 function_entry_patch_area_start global variables.
23602 * opts.c (parse_and_check_patch_area): New function.
23603 (common_handle_option): Use it.
23604 * opts.h (parse_and_check_patch_area): New function.
23605 * toplev.c (process_options): Parse and use
23606 function_entry_patch_area_size.
23608 2021-02-08 Martin Sebor <msebor@redhat.com>
23610 * doc/extend.texi (attribute malloc): Correct typos.
23612 2021-02-05 Nathan Sidwell <nathan@acm.org>
23615 * gcc.c (driver::maybe_run_linker): Check for input file
23616 accessibility if not linking.
23618 2021-02-05 Richard Biener <rguenther@suse.de>
23620 PR tree-optimization/98855
23621 * tree-vectorizer.h (add_stmt_cost): New overload.
23622 * tree-vect-slp.c (li_cost_vec_cmp): New.
23623 (vect_bb_slp_scalar_cost): Cost individual loop regions
23624 separately. Account for the scalar instance root stmt.
23626 2021-02-05 Tom de Vries <tdevries@suse.de>
23629 * tree-switch-conversion.c (jump_table_cluster::emit): Add loc
23631 (bit_test_cluster::emit): Reuse location_t for newly created
23633 (switch_decision_tree::try_switch_expansion): Preserve
23635 * tree-switch-conversion.h: Change function signatures.
23637 2021-02-05 Jakub Jelinek <jakub@redhat.com>
23640 * config/i386/i386-options.c (m_NONE, m_ALL): Define.
23641 * config/i386/x86-tune.def (X86_TUNE_BRANCH_PREDICTION_HINTS,
23642 X86_TUNE_PROMOTE_QI_REGS): Use m_NONE instead of 0U.
23643 (X86_TUNE_QIMODE_MATH): Use m_ALL instead of ~0U.
23645 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23647 * config/aarch64/aarch64-simd-builtins.def (get_high): Define builtin.
23648 * config/aarch64/aarch64-simd.md (aarch64_get_high<mode>): Define.
23649 * config/aarch64/arm_neon.h (__GET_HIGH): Delete.
23650 (vget_high_f16): Reimplement using new builtin.
23651 (vget_high_f32): Likewise.
23652 (vget_high_f64): Likewise.
23653 (vget_high_p8): Likewise.
23654 (vget_high_p16): Likewise.
23655 (vget_high_p64): Likewise.
23656 (vget_high_s8): Likewise.
23657 (vget_high_s16): Likewise.
23658 (vget_high_s32): Likewise.
23659 (vget_high_s64): Likewise.
23660 (vget_high_u8): Likewise.
23661 (vget_high_u16): Likewise.
23662 (vget_high_u32): Likewise.
23663 (vget_high_u64): Likewise.
23665 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23667 * config/aarch64/aarch64-simd-builtins.def (get_low): Define builtin.
23668 * config/aarch64/aarch64-simd.md (aarch64_get_low<mode>): Define.
23669 * config/aarch64/arm_neon.h (__GET_LOW): Delete.
23670 (vget_low_f16): Reimplement using new builtin.
23671 (vget_low_f32): Likewise.
23672 (vget_low_f64): Likewise.
23673 (vget_low_p8): Likewise.
23674 (vget_low_p16): Likewise.
23675 (vget_low_p64): Likewise.
23676 (vget_low_s8): Likewise.
23677 (vget_low_s16): Likewise.
23678 (vget_low_s32): Likewise.
23679 (vget_low_s64): Likewise.
23680 (vget_low_u8): Likewise.
23681 (vget_low_u16): Likewise.
23682 (vget_low_u32): Likewise.
23683 (vget_low_u64): Likewise.
23685 2021-02-05 Kito Cheng <kito.cheng@sifive.com>
23687 * gcc.c (print_multilib_info): Check all required argument is provided
23690 2021-02-05 liuhongt <hongtao.liu@intel.com>
23693 * config/i386/i386-expand.c (ix86_expand_sse_cmp): Don't
23694 generate integer mask comparison for 128/256-bits vector when
23695 op_true/op_false is NULL_RTX or CONSTM1_RTX/CONST0_RTX. Also
23696 delete redundant !maskcmp condition.
23697 (ix86_expand_int_vec_cmp): Ditto but no redundant deletion
23699 (ix86_expand_sse_movcc): Delete definition of maskcmp, add the
23700 condition directly to if (maskcmp), add extra check for
23701 cmpmode, it should be MODE_INT.
23702 (ix86_expand_fp_vec_cmp): Pass NULL to ix86_expand_sse_cmp's
23703 parameters op_true/op_false.
23704 (ix86_use_mask_cmp_p): New.
23706 2021-02-05 liuhongt <hongtao.liu@intel.com>
23709 * config/i386/x86-tune.def (X86_TUNE_AVX256_UNALIGNED_LOAD_OPTIMAL):
23710 Remove m_GENERIC from ~list.
23711 (X86_TUNE_AVX256_UNALIGNED_STORE_OPTIMAL): Ditto.
23713 2021-02-04 David Malcolm <dmalcolm@redhat.com>
23716 * diagnostic-show-locus.c (compatible_locations_p): Require
23717 locations in the same macro map to be either both from the
23718 macro definition, or both from the macro arguments.
23720 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
23722 * config/aarch64/aarch64-simd-builtins.def: Add
23723 [su]mull_hi_lane[q] builtin generator macros.
23724 * config/aarch64/aarch64-simd.md
23725 (aarch64_<su>mull_hi_lane<mode>_insn): Define.
23726 (aarch64_<su>mull_hi_lane<mode>): Define.
23727 (aarch64_<su>mull_hi_laneq<mode>_insn): Define.
23728 (aarch64_<su>mull_hi_laneq<mode>): Define.
23729 * config/aarch64/arm_neon.h (vmull_high_lane_s16): Use RTL
23730 builtin instead of inline asm.
23731 (vmull_high_lane_s32): Likewise.
23732 (vmull_high_lane_u16): Likewise.
23733 (vmull_high_lane_u32): Likewise.
23734 (vmull_high_laneq_s16): Likewise.
23735 (vmull_high_laneq_s32): Likewise.
23736 (vmull_high_laneq_u16): Likewise.
23737 (vmull_high_laneq_u32): Liekwise.
23739 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
23741 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_hi_n
23742 builtin generator macros.
23743 * config/aarch64/aarch64-simd.md
23744 (aarch64_<su>mull_hi_n<mode>_insn): Define.
23745 (aarch64_<su>mull_hi_n<mode>): Define.
23746 * config/aarch64/arm_neon.h (vmull_high_n_s16): Use RTL builtin
23747 instead of inline asm.
23748 (vmull_high_n_s32): Likewise.
23749 (vmull_high_n_u16): Likewise.
23750 (vmull_high_n_u32): Likewise.
23752 2021-02-04 Richard Biener <rguenther@suse.de>
23754 PR tree-optimization/98855
23755 * tree-vect-loop.c (vectorizable_phi): Do not cost
23756 single-argument PHIs.
23757 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Likewise.
23758 * tree-vect-stmts.c (vectorizable_bswap): Also perform
23759 costing for SLP operation.
23761 2021-02-04 Martin Liska <mliska@suse.cz>
23763 * doc/extend.texi: Mention -mprefer-vector-width in target
23766 2021-02-03 Martin Sebor <msebor@redhat.com>
23768 PR tree-optimization/98937
23769 * tree-ssa-strlen.c (strlen_dom_walker::~strlen_dom_walker): Define.
23770 Flush pointer_query cache.
23772 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
23774 * config/rs6000/genfusion.pl (gen_2logical): Add missing
23775 fixes based on patch review.
23776 * config/rs6000/fusion.md: Regenerate file.
23778 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
23780 * config/rs6000/t-rs6000: Comment out auto generation of
23783 2021-02-03 Andrew Stubbs <ams@codesourcery.com>
23785 * config/gcn/gcn-opts.h (enum processor_type): Add PROCESSOR_GFX908.
23786 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Add gfx908.
23787 (output_file_start): Add gfx908.
23788 * config/gcn/gcn.opt (gpu_type): Add gfx908.
23789 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Add march=gfx908.
23790 (MULTILIB_DIRNAMES): Add gfx908.
23791 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): New define.
23792 (main): Recognize gfx908.
23793 * config/gcn/t-omp-device: Add gfx908.
23795 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
23797 * config/aarch64/aarch64-simd-builtins.def: Add
23798 [su]mlsl_hi_lane[q] builtin macro generators.
23799 * config/aarch64/aarch64-simd.md
23800 (aarch64_<su>mlsl_hi_lane<mode>_insn): Define.
23801 (aarch64_<su>mlsl_hi_lane<mode>): Define.
23802 (aarch64_<su>mlsl_hi_laneq<mode>_insn): Define.
23803 (aarch64_<su>mlsl_hi_laneq<mode>): Define.
23804 * config/aarch64/arm_neon.h (vmlsl_high_lane_s16): Use RTL
23805 builtin instead of inline asm.
23806 (vmlsl_high_lane_s32): Likewise.
23807 (vmlsl_high_lane_u16): Likewise.
23808 (vmlsl_high_lane_u32): Likewise.
23809 (vmlsl_high_laneq_s16): Likewise.
23810 (vmlsl_high_laneq_s32): Likewise.
23811 (vmlsl_high_laneq_u16): Likewise.
23812 (vmlsl_high_laneq_u32): Likewise.
23813 (vmlal_high_laneq_u32): Likewise.
23815 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
23817 * config/aarch64/aarch64-simd-builtins.def: Add
23818 [su]mlal_hi_lane[q] builtin generator macros.
23819 * config/aarch64/aarch64-simd.md
23820 (aarch64_<su>mlal_hi_lane<mode>_insn): Define.
23821 (aarch64_<su>mlal_hi_lane<mode>): Define.
23822 (aarch64_<su>mlal_hi_laneq<mode>_insn): Define.
23823 (aarch64_<su>mlal_hi_laneq<mode>): Define.
23824 * config/aarch64/arm_neon.h (vmlal_high_lane_s16): Use RTL
23825 builtin instead of inline asm.
23826 (vmlal_high_lane_s32): Likewise.
23827 (vmlal_high_lane_u16): Likewise.
23828 (vmlal_high_lane_u32): Likewise.
23829 (vmlal_high_laneq_s16): Likewise.
23830 (vmlal_high_laneq_s32): Likewise.
23831 (vmlal_high_laneq_u16): Likewise.
23832 (vmlal_high_laneq_u32): Likewise.
23834 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
23836 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_hi_n
23837 builtin generator macros.
23838 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_hi_n<mode>_insn):
23840 (aarch64_<su>mlsl_hi_n<mode>): Define.
23841 * config/aarch64/arm_neon.h (vmlsl_high_n_s16): Use RTL builtin
23842 instead of inline asm.
23843 (vmlsl_high_n_s32): Likewise.
23844 (vmlsl_high_n_u16): Likewise.
23845 (vmlsl_high_n_u32): Likewise.
23847 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
23849 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_hi_n
23850 builtin generator macros.
23851 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_hi_n<mode>_insn):
23853 (aarch64_<su>mlal_hi_n<mode>): Define.
23854 * config/aarch64/arm_neon.h (vmlal_high_n_s16): Use RTL builtin
23855 instead of inline asm.
23856 (vmlal_high_n_s32): Likewise.
23857 (vmlal_high_n_u16): Likewise.
23858 (vmlal_high_n_u32): Likewise.
23860 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
23862 * config/aarch64/aarch64-simd-builtins.def: Add RTL builtin
23864 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal_hi<mode>):
23866 (aarch64_<su>mlal_hi<mode>_insn): This.
23867 (aarch64_<su>mlal_hi<mode>): Define.
23868 * config/aarch64/arm_neon.h (vmlal_high_s8): Use RTL builtin
23869 instead of inline asm.
23870 (vmlal_high_s16): Likewise.
23871 (vmlal_high_s32): Likewise.
23872 (vmlal_high_u8): Likewise.
23873 (vmlal_high_u16): Likewise.
23874 (vmlal_high_u32): Likewise.
23876 2021-02-03 Ilya Leoshkevich <iii@linux.ibm.com>
23878 * lra-spills.c (remove_pseudos): Call lra_update_insn_recog_data()
23879 after calling alter_subreg() on a (mem).
23881 2021-02-03 Martin Liska <mliska@suse.cz>
23884 * lto-streamer-out.c (produce_lto_section): Fill up missing
23886 * lto-streamer.h (struct lto_section): Add _padding field.
23888 2021-02-03 Richard Biener <rguenther@suse.de>
23890 * lto-streamer.c (lto_get_section_name): Free temporary
23892 * tree-loop-distribution.c
23893 (loop_distribution::merge_dep_scc_partitions): Free edge data.
23895 2021-02-03 Jakub Jelinek <jakub@redhat.com>
23897 PR middle-end/97487
23898 * ifcvt.c (noce_can_force_operand): New function.
23899 (noce_emit_move_insn): Use it.
23900 (noce_try_sign_mask): Likewise. Formatting fix.
23902 2021-02-03 Jakub Jelinek <jakub@redhat.com>
23904 PR middle-end/97971
23905 * lra-constraints.c (process_alt_operands): For inline asm, don't call
23906 fatal_insn, but instead return false.
23908 2021-02-03 Jakub Jelinek <jakub@redhat.com>
23910 PR tree-optimization/98287
23911 * config/i386/mmx.md (<insn><mode>3): For shifts don't enable expander
23914 2021-02-03 Tamar Christina <tamar.christina@arm.com>
23916 PR tree-optimization/98928
23917 * tree-vect-loop.c (vect_analyze_loop_2): Change
23918 STMT_VINFO_SLP_VECT_ONLY to STMT_VINFO_SLP_VECT_ONLY_PATTERN.
23919 * tree-vect-slp-patterns.c (complex_pattern::build): Likewise.
23920 * tree-vectorizer.h (STMT_VINFO_SLP_VECT_ONLY_PATTERN): New.
23921 (class _stmt_vec_info): Add slp_vect_pattern_only_p.
23923 2021-02-02 Richard Biener <rguenther@suse.de>
23925 * gimple-loop-interchange.cc (prepare_data_references):
23927 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
23928 * tree-ssa-loop-im.c (hoist_memory_references): Likewise.
23929 * tree-vect-stmts.c (vectorizable_condition): Do not
23931 (vectorizable_comparison): Likewise.
23933 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23935 * config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
23936 * config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
23937 * config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
23938 (vrsqrteq_u32): Likewise.
23940 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23942 * config/aarch64/aarch64-simd-builtins.def (sqxtun2): Define builtin.
23943 * config/aarch64/aarch64-simd.md (aarch64_sqxtun2<mode>_le): Define.
23944 (aarch64_sqxtun2<mode>_be): Likewise.
23945 (aarch64_sqxtun2<mode>): Likewise.
23946 * config/aarch64/arm_neon.h (vqmovun_high_s16): Reimplement using builtin.
23947 (vqmovun_high_s32): Likewise.
23948 (vqmovun_high_s64): Likewise.
23949 * config/aarch64/iterators.md (UNSPEC_SQXTUN2): Define.
23951 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23953 * config/aarch64/aarch64-simd-builtins.def (bfdot_lane, bfdot_laneq): Use
23955 (bfmlalb_lane, bfmlalt_lane, bfmlalb_lane_q, bfmlalt_lane_q): Use FP flags.
23957 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23959 * config/aarch64/aarch64-simd-builtins.def (fcmla_lane0, fcmla_lane90,
23960 fcmla_lane180, fcmla_lane270, fcmlaq_lane0, fcmlaq_lane90, fcmlaq_lane180,
23961 fcmlaq_lane270, scvtf, ucvtf, fcvtzs, fcvtzu, scvtfsi, scvtfdi, ucvtfsi,
23962 ucvtfdi, fcvtzshf, fcvtzuhf, fmlal_lane_low, fmlsl_lane_low,
23963 fmlal_laneq_low, fmlsl_laneq_low, fmlalq_lane_low, fmlslq_lane_low,
23964 fmlalq_laneq_low, fmlslq_laneq_low, fmlal_lane_high, fmlsl_lane_high,
23965 fmlal_laneq_high, fmlsl_laneq_high, fmlalq_lane_high, fmlslq_lane_high,
23966 fmlalq_laneq_high, fmlslq_laneq_high): Use FP flags.
23968 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23970 * config/aarch64/aarch64-builtins.c (FLAG_LOAD): Define.
23971 * config/aarch64/aarch64-simd-builtins.def (ld1x2, ld2, ld3, ld4, ld2r,
23972 ld3r, ld4r, ld1, ld1x3, ld1x4): Use LOAD flags.
23974 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23976 * config/aarch64/aarch64-simd-builtins.def (combine, zip1, zip2,
23977 uzp1, uzp2, trn1, trn2, simd_bsl): Use AUTO_FP flags.
23979 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23981 * config/aarch64/aarch64-simd-builtins.def (clrsb, clz, ctz, popcount,
23982 vec_smult_lane_, vec_smlal_lane_, vec_smult_laneq_, vec_smlal_laneq_,
23983 vec_umult_lane_, vec_umlal_lane_, vec_umult_laneq_, vec_umlal_laneq_,
23984 ashl, sshl, ushl, srshl, urshl, sdot_lane, udot_lane, sdot_laneq,
23985 udot_laneq, usdot_lane, usdot_laneq, sudot_lane, sudot_laneq, ashr,
23986 ashr_simd, lshr, lshr_simd, srshr_n, urshr_n, ssra_n, usra_n, srsra_n,
23987 ursra_n, sshll_n, ushll_n, sshll2_n, ushll2_n, ssri_n, usri_n, ssli_n,
23988 ssli_n, usli_n, bswap, rbit, simd_bsl, eor3q, rax1q, xarq, bcaxq): Use
23989 NONE builtin flags.
23991 2021-02-02 Jakub Jelinek <jakub@redhat.com>
23993 PR tree-optimization/98848
23994 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
23995 STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
23997 2021-02-02 Kito Cheng <kito.cheng@sifive.com>
24000 * expr.c: Check mode before calling store_expr.
24002 2021-02-02 Christophe Lyon <christophe.lyon@linaro.org>
24004 * config/arm/iterators.md (supf): Remove VORNQ_S and VORNQ_U.
24006 * config/arm/mve.md (mve_vornq_s<mode>): New entry for vorn
24007 instruction using expression ior.
24008 (mve_vornq_u<mode>): New expander.
24009 (mve_vornq_f<mode>): Use ior code instead of unspec.
24010 * config/arm/unspecs.md (VORNQ_S, VORNQ_U, VORNQ_F): Remove.
24012 2021-02-02 Alexandre Oliva <oliva@adacore.com>
24014 * tree-nested.c (convert_nonlocal_reference_op): Move
24015 current_function_decl restore after re-gimplification.
24016 (convert_local_reference_op): Likewise.
24018 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24020 * config/aarch64/aarch64-simd-builtins.def (rshrn, rshrn2):
24022 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>_insn_le):
24024 (aarch64_rshrn<mode>_insn_be): Likewise.
24025 (aarch64_rshrn<mode>): Likewise.
24026 (aarch64_rshrn2<mode>_insn_le): Likewise.
24027 (aarch64_rshrn2<mode>_insn_be): Likewise.
24028 (aarch64_rshrn2<mode>): Likewise.
24029 * config/aarch64/aarch64.md (unspec): Add UNSPEC_RSHRN.
24030 * config/aarch64/arm_neon.h (vrshrn_high_n_s16): Reimplement
24032 (vrshrn_high_n_s32): Likewise.
24033 (vrshrn_high_n_s64): Likewise.
24034 (vrshrn_high_n_u16): Likewise.
24035 (vrshrn_high_n_u32): Likewise.
24036 (vrshrn_high_n_u64): Likewise.
24037 (vrshrn_n_s16): Likewise.
24038 (vrshrn_n_s32): Likewise.
24039 (vrshrn_n_s64): Likewise.
24040 (vrshrn_n_u16): Likewise.
24041 (vrshrn_n_u32): Likewise.
24042 (vrshrn_n_u64): Likewise.
24044 2021-02-01 Sergei Trofimovich <siarheit@google.com>
24046 PR tree-optimization/98499
24047 * ipa-modref.c (analyze_ssa_name_flags): treat RVO
24048 conservatively and assume all possible side-effects.
24050 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24052 * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi,
24053 vec_unpacku_hi_): Define builtins.
24054 * config/aarch64/arm_neon.h (vmovl_high_s8): Reimplement using
24056 (vmovl_high_s16): Likewise.
24057 (vmovl_high_s32): Likewise.
24058 (vmovl_high_u8): Likewise.
24059 (vmovl_high_u16): Likewise.
24060 (vmovl_high_u32): Likewise.
24062 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24064 * config/aarch64/aarch64-simd-builtins.def (sabdl, uabdl):
24066 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl<mode>): New
24068 * config/aarch64/aarch64.md (unspec): Define UNSPEC_SABDL,
24070 * config/aarch64/arm_neon.h (vabdl_s8): Reimplemet using
24072 (vabdl_s16): Likewise.
24073 (vabdl_s32): Likewise.
24074 (vabdl_u8): Likewise.
24075 (vabdl_u16): Likewise.
24076 (vabdl_u32): Likewise.
24077 * config/aarch64/iterators.md (ABDL): New int iterator.
24078 (sur): Handle UNSPEC_SABDL, UNSPEC_UABDL.
24080 2021-02-01 Martin Sebor <msebor@redhat.com>
24082 * tree.h (BLOCK_VARS): Add comment.
24083 (BLOCK_SUBBLOCKS): Same.
24084 (BLOCK_SUPERCONTEXT): Same.
24085 (BLOCK_ABSTRACT_ORIGIN): Same.
24086 (inlined_function_outer_scope_p): Same.
24088 2021-02-01 Martin Sebor <msebor@redhat.com>
24090 PR middle-end/97172
24091 * attribs.c (attr_access::free_lang_data): Define new function.
24092 * attribs.h (attr_access::free_lang_data): Declare new function.
24094 2021-02-01 Richard Biener <rguenther@suse.de>
24096 * vec.h (auto_vec::auto_vec): Add memory stat parameters
24098 * bitmap.h (auto_bitmap::auto_bitmap): Likewise.
24100 2021-02-01 Tamar Christina <tamar.christina@arm.com>
24102 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>,
24103 aarch64_<su>mlsl<mode>, aarch64_<su>mlsl_n<mode>): Flip mult operands.
24105 2021-02-01 Richard Biener <rguenther@suse.de>
24107 PR rtl-optimization/98863
24108 * config/i386/i386-features.c (convert_scalars_to_vector):
24109 Set DF_RD_PRUNE_DEAD_DEFS.
24111 2021-01-31 Eric Botcazou <ebotcazou@adacore.com>
24113 * system.h (SIZE_MAX): Define if not already defined.
24115 2021-01-30 Aaron Sawdey <acsawdey@linux.ibm.com>
24117 * config/rs6000/genfusion.pl (gen_2logical): New function to
24118 generate patterns for logical-logical fusion.
24119 * config/rs6000/fusion.md: Regenerated patterns.
24120 * config/rs6000/rs6000-cpus.def: Add
24121 OPTION_MASK_P10_FUSION_2LOGICAL.
24122 * config/rs6000/rs6000.c (rs6000_option_override_internal):
24123 Enable logical-logical fusion for p10.
24124 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2logical.
24126 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
24128 * config/rs6000/rs6000.opt: Add periods to new AIX options.
24130 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
24132 * config/rs6000/rs6000.opt (mabi=vec-extabi): New.
24133 (mabi=vec-default): New.
24134 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
24135 __EXTABI__ for AIX Vector extended ABI.
24136 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print AIX Vector
24138 (conditional_register_usage): If AIX vec_extabi enabled, vs20-vs31
24140 * doc/invoke.texi (PowerPC mabi): Add AIX vec-extabi and vec-default.
24142 2021-01-30 Jakub Jelinek <jakub@redhat.com>
24144 * config/i386/i386-features.c (remove_partial_avx_dependency): Clear
24145 DF_DEFER_INSN_RESCAN after calling df_process_deferred_rescans.
24147 2021-01-29 Vladimir N. Makarov <vmakarov@redhat.com>
24150 * lra-constraints.c (in_class_p): Don't narrow class only for REG
24153 2021-01-29 Will Schmidt <will_schmidt@vnet.ibm.com>
24155 * config/rs6000/rs6000-call.c (rs6000_expand_binup_builtin): Add
24156 clauses for CODE_FOR_vsx_xvcvuxddp_scale and
24157 CODE_FOR_vsx_xvcvsxddp_scale to the parameter checking code.
24159 2021-01-29 Andrew MacLeod <amacleod@redhat.com>
24161 PR tree-optimization/98866
24162 * gimple-range-gori.h (gori_compute:set_range_invariant): New.
24163 * gimple-range-gori.cc (gori_map::set_range_invariant): New.
24164 (gori_map::m_maybe_invariant): Rename from all_outgoing.
24165 (gori_map::gori_map): Rename all_outgoing to m_maybe_invariant.
24166 (gori_map::is_export_p): Ditto.
24167 (gori_map::calculate_gori): Ditto.
24168 (gori_compute::set_range_invariant): New.
24169 * gimple-range.cc (gimple_ranger::range_of_stmt): Set range
24170 invariant for pointers evaluating to [1, +INF].
24172 2021-01-29 Richard Biener <rguenther@suse.de>
24174 PR rtl-optimization/98863
24175 * config/i386/i386-features.c (remove_partial_avx_dependency):
24176 Do not perform DF analysis.
24177 (pass_data_remove_partial_avx_dependency): Remove
24180 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
24182 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_n
24183 builtin generator macros.
24184 * config/aarch64/aarch64-simd.md (aarch64_<su>mull_n<mode>):
24186 * config/aarch64/arm_neon.h (vmull_n_s16): Use RTL builtin
24187 instead of inline asm.
24188 (vmull_n_s32): Likewise.
24189 (vmull_n_u16): Likewise.
24190 (vmull_n_u32): Likewise.
24192 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24194 * config/aarch64/aarch64-simd-builtins.def (sabdl2, uabdl2):
24196 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
24198 (aarch64_<sur>abdl2<mode>): ... This.
24199 (<sur>sadv16qi): Adjust use of above.
24200 * config/aarch64/arm_neon.h (vabdl_high_s8): Reimplement using
24202 (vabdl_high_s16): Likewise.
24203 (vabdl_high_s32): Likewise.
24204 (vabdl_high_u8): Likewise.
24205 (vabdl_high_u16): Likewise.
24206 (vabdl_high_u32): Likewise.
24208 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24210 * config/aarch64/aarch64-simd-builtins.def (sabal2): Define
24212 (uabal2): Likewise.
24213 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal2<mode>): New
24215 * config/aarch64/aarch64.md (unspec): Add UNSPEC_SABAL2 and
24217 * config/aarch64/arm_neon.h (vabal_high_s8): Reimplement using
24219 (vabal_high_s16): Likewise.
24220 (vabal_high_s32): Likewise.
24221 (vabal_high_u8): Likewise.
24222 (vabal_high_u16): Likewise.
24223 (vabal_high_u32): Likewise.
24224 * config/aarch64/iterators.md (ABAL2): New mode iterator.
24225 (sur): Handle UNSPEC_SABAL2, UNSPEC_UABAL2.
24227 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24229 * config/aarch64/aarch64-simd-builtins.def (sabal): Define
24232 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal<mode>_4):
24234 (aarch64_<sur>abal<mode>): ... This
24235 (<sur>sadv16qi): Adust use of the above.
24236 * config/aarch64/arm_neon.h (vabal_s8): Reimplement using
24238 (vabal_s16): Likewise.
24239 (vabal_s32): Likewise.
24240 (vabal_u8): Likewise.
24241 (vabal_u16): Likewise.
24242 (vabal_u32): Likewise.
24244 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24246 * config/aarch64/aarch64-simd-builtins.def (saddlv, uaddlv):
24248 * config/aarch64/aarch64-simd.md (aarch64_<su>addlv<mode>):
24250 * config/aarch64/arm_neon.h (vaddlv_s8): Reimplement using
24252 (vaddlv_s16): Likewise.
24253 (vaddlv_u8): Likewise.
24254 (vaddlv_u16): Likewise.
24255 (vaddlvq_s8): Likewise.
24256 (vaddlvq_s16): Likewise.
24257 (vaddlvq_s32): Likewise.
24258 (vaddlvq_u8): Likewise.
24259 (vaddlvq_u16): Likewise.
24260 (vaddlvq_u32): Likewise.
24261 (vaddlv_s32): Likewise.
24262 (vaddlv_u32): Likewise.
24263 * config/aarch64/iterators.md (VDQV_L): New mode iterator.
24264 (unspec): Add UNSPEC_SADDLV, UNSPEC_UADDLV.
24265 (Vwstype): New mode attribute.
24267 (VWIDE_S): Likewise.
24268 (USADDLV): New int iterator.
24269 (su): Handle UNSPEC_SADDLV, UNSPEC_UADDLV.
24271 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
24273 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_lane[q]
24274 builtin generator macros.
24275 * config/aarch64/aarch64-simd.md (aarch64_vec_<su>mlsl_lane<Qlane>):
24277 * config/aarch64/arm_neon.h (vmlsl_lane_s16): Use RTL builtin
24278 instead of inline asm.
24279 (vmlsl_lane_s32): Likewise.
24280 (vmlsl_lane_u16): Likewise.
24281 (vmlsl_lane_u32): Likewise.
24282 (vmlsl_laneq_s16): Likewise.
24283 (vmlsl_laneq_s32): Likewise.
24284 (vmlsl_laneq_u16): Likewise.
24285 (vmlsl_laneq_u32): Likewise.
24287 2021-01-29 Richard Biener <rguenther@suse.de>
24289 * doc/invoke.texi (--param max-gcse-memory): Document unit
24291 * gcse.c (gcse_or_cprop_is_too_expensive): Adjust.
24292 * params.opt (--param max-gcse-memory): Adjust default and
24293 document unit of size.
24295 2021-01-29 Richard Biener <rguenther@suse.de>
24297 PR rtl-optimization/98863
24298 * gcse.c (gcse_or_cprop_is_too_expensive): Use unsigned
24299 HOST_WIDE_INT for the memory estimate.
24301 2021-01-29 Bin Cheng <bin.cheng@linux.alibaba.com>
24302 Richard Biener <rguenther@suse.de>
24304 PR tree-optimization/97627
24305 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
24306 Do not analyze fake edges.
24308 2021-01-29 Richard Biener <rguenther@suse.de>
24310 PR rtl-optimization/98144
24311 * df.h (df_mir_bb_info): Add con_visited member.
24312 * df-problems.c (df_mir_alloc): Initialize con_visited,
24313 do not fully populate IN and OUT.
24314 (df_mir_reset): Likewise.
24315 (df_mir_confluence_0): Set con_visited.
24316 (df_mir_confluence_n): Properly handle implicitely
24317 fully populated IN and OUT as designated by con_visited
24318 and update con_visited accordingly.
24320 2021-01-29 Jakub Jelinek <jakub@redhat.com>
24323 * config/arm/vec-common.md (mve_vshlq_<supf><mode>,
24324 vashl<mode>3, vashr<mode>3, vlshr<mode>3): Add
24325 && !TARGET_REALLY_IWMMXT to conditions.
24327 2021-01-29 Jakub Jelinek <jakub@redhat.com>
24330 * cfgbuild.c (find_bb_boundaries): Reset debug_insn when seeing
24333 2021-01-28 Marek Polacek <polacek@redhat.com>
24336 * stor-layout.c (finalize_type_size): If we reset TYPE_USER_ALIGN in
24337 the main variant, maybe reset it in its variants too.
24338 * tree.c (check_base_type): Return true only if TYPE_USER_ALIGN match.
24339 (check_aligned_type): Check if TYPE_USER_ALIGN match.
24341 2021-01-28 Christophe Lyon <christophe.lyon@linaro.org>
24344 * config/arm/arm.c (arm_rtx_costs_internal): Adjust cost of vector
24345 of constant zero for comparisons.
24347 2021-01-28 Michael Meissner <meissner@linux.ibm.com>
24349 * config/rs6000/rs6000.c (rs6000_mangle_decl_assembler_name): Add
24350 support for mapping built-in function names for long double
24351 built-in functions if long double is IEEE 128-bit.
24353 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
24355 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_n
24356 builtin generator macros.
24357 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_n<mode>):
24359 * config/aarch64/arm_neon.h (vmlsl_n_s16): Use RTL builtin
24360 instead of inline asm.
24361 (vmlsl_n_s32): Likewise.
24362 (vmlsl_n_u16): Likewise.
24363 (vmlsl_n_u32): Likewise.
24365 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
24367 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_n
24368 builtin generator macros.
24369 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>):
24371 * config/aarch64/arm_neon.h (vmlal_n_s16): Use RTL builtin
24372 instead of inline asm.
24373 (vmlal_n_s32): Likewise.
24374 (vmlal_n_u16): Likewise.
24375 (vmlal_n_u32): Likewise.
24377 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24379 * config/aarch64/aarch64-simd-builtins.def (shrn2): Define
24381 * config/aarch64/aarch64-simd.md (aarch64_shrn2<mode>_insn_le):
24383 (aarch64_shrn2<mode>_insn_be): Likewise.
24384 (aarch64_shrn2<mode>): Likewise.
24385 * config/aarch64/arm_neon.h (vshrn_high_n_s16): Reimlplement
24387 (vshrn_high_n_s32): Likewise.
24388 (vshrn_high_n_s64): Likewise.
24389 (vshrn_high_n_u16): Likewise.
24390 (vshrn_high_n_u32): Likewise.
24391 (vshrn_high_n_u64): Likewise.
24393 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24395 * config/aarch64/aarch64-simd-builtins.def (shrn): Define
24397 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le):
24399 (aarch64_shrn<mode>_insn_be): Likewise.
24400 (aarch64_shrn<mode>): Likewise.
24401 * config/aarch64/arm_neon.h (vshrn_n_s16): Reimplement using
24403 (vshrn_n_s32): Likewise.
24404 (vshrn_n_s64): Likewise.
24405 (vshrn_n_u16): Likewise.
24406 (vshrn_n_u32): Likewise.
24407 (vshrn_n_u64): Likewise.
24408 * config/aarch64/iterators.md (vn_mode): New mode attribute.
24410 2021-01-28 Richard Biener <rguenther@suse.de>
24412 PR rtl-optimization/80960
24413 * dse.c (check_mem_read_rtx): Call get_addr on the
24416 2021-01-28 Xionghu Luo <luoxhu@linux.ibm.com>
24417 David Edelsohn <dje.gcc@gmail.com>
24420 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
24421 Don't generate VIEW_CONVERT_EXPR for fcode ALTIVEC_BUILTIN_VEC_INSERT
24423 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
24425 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Remove the
24426 wrapper call rs6000_expand_vector_set_var for cleanup. Call
24427 rs6000_expand_vector_set_var_p9 and rs6000_expand_vector_set_var_p8
24429 (rs6000_expand_vector_set_var): Delete.
24430 (rs6000_expand_vector_set_var_p9): Make static.
24431 (rs6000_expand_vector_set_var_p8): Make static.
24433 2021-01-28 Xing GUO <higuoxing@gmail.com>
24435 * common/config/riscv/riscv-common.c
24436 (riscv_subset_list::parsing_subset_version): Fix -march option parsing
24437 when `p` extension exists.
24439 2021-01-27 Vladimir N. Makarov <vmakarov@redhat.com>
24441 PR rtl-optimization/97684
24442 * ira.c (ira): Call ira_set_pseudo_classes before
24443 update_equiv_regs when it is necessary.
24445 2021-01-27 Jakub Jelinek <jakub@redhat.com>
24448 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Use
24449 %w0, %w1 and %2 instead of %0, %1 and %2.
24451 2021-01-27 Aaron Sawdey <acsawdey@linux.ibm.com>
24453 * config/rs6000/genfusion.pl: New script to generate
24454 define_insn_and_split patterns so combine can arrange fused
24455 instructions next to each other.
24456 * config/rs6000/fusion.md: New file, generated fused instruction
24457 patterns for combine.
24458 * config/rs6000/predicates.md (const_m1_to_1_operand): New predicate.
24459 (non_update_memory_operand): New predicate.
24460 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_P10_FUSION and
24461 OPTION_MASK_P10_FUSION_LD_CMPI to ISA_3_1_MASKS_SERVER and
24463 * config/rs6000/rs6000-protos.h (address_is_non_pfx_d_or_x): Add
24465 * config/rs6000/rs6000.c (rs6000_option_override_internal):
24466 Automatically set OPTION_MASK_P10_FUSION and
24467 OPTION_MASK_P10_FUSION_LD_CMPI if target is power10.
24468 (rs600_opt_masks): Allow -mpower10-fusion
24469 in function attributes.
24470 (address_is_non_pfx_d_or_x): New function.
24471 * config/rs6000/rs6000.h: Add MASK_P10_FUSION.
24472 * config/rs6000/rs6000.md: Include fusion.md.
24473 * config/rs6000/rs6000.opt: Add -mpower10-fusion
24474 and -mpower10-fusion-ld-cmpi.
24475 * config/rs6000/t-rs6000: Add dependencies involving fusion.md.
24477 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
24479 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal
24480 builtin generator macros.
24481 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal<mode>):
24483 (aarch64_<su>mlal<mode>): This.
24484 * config/aarch64/arm_neon.h (vmlal_s8): Use RTL builtin
24485 instead of inline asm.
24486 (vmlal_s16): Likewise.
24487 (vmlal_s32): Likewise.
24488 (vmlal_u8): Likewise.
24489 (vmlal_u16): Likewise.
24490 (vmlal_u32): Likewise.
24492 2021-01-27 Richard Biener <rguenther@suse.de>
24494 PR tree-optimization/98854
24495 * tree-vect-slp.c (vect_build_slp_tree_2): Also build
24496 PHIs from scalars when the number of CTORs matches the
24497 number of children.
24499 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
24501 * config/aarch64/aarch64-simd-builtins.def: Add mls_n builtin
24503 * config/aarch64/aarch64-simd.md (*aarch64_mls_elt_merge<mode>):
24505 (aarch64_mls_n<mode>): This.
24506 * config/aarch64/arm_neon.h (vmls_n_s16): Use RTL builtin
24508 (vmls_n_s32): Likewise.
24509 (vmls_n_u16): Likewise.
24510 (vmls_n_u32): Likewise.
24511 (vmlsq_n_s16): Likewise.
24512 (vmlsq_n_s32): Likewise.
24513 (vmlsq_n_u16): Likewise.
24514 (vmlsq_n_u32): Likewise.
24516 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
24518 * config/aarch64/aarch64-simd-builtins.def: Add mls builtin
24520 * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather
24522 (vmls_s16): Likewise.
24523 (vmls_s32): Likewise.
24524 (vmls_u8): Likewise.
24525 (vmls_u16): Likewise.
24526 (vmls_u32): Likewise.
24527 (vmlsq_s8): Likewise.
24528 (vmlsq_s16): Likewise.
24529 (vmlsq_s32): Likewise.
24530 (vmlsq_u8): Likewise.
24531 (vmlsq_u16): Likewise.
24532 (vmlsq_u32): Likewise.
24534 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
24536 * config/aarch64/aarch64-simd-builtins.def: Add mla_n builtin
24538 * config/aarch64/aarch64-simd.md (*aarch64_mla_elt_merge<mode>):
24540 (aarch64_mla_n<mode>): This.
24541 * config/aarch64/arm_neon.h (vmla_n_s16): Use RTL builtin
24543 (vmla_n_s32): Likewise.
24544 (vmla_n_u16): Likewise.
24545 (vmla_n_u32): Likewise.
24546 (vmlaq_n_s16): Likewise.
24547 (vmlaq_n_s32): Likewise.
24548 (vmlaq_n_u16): Likewise.
24549 (vmlaq_n_u32): Likewise.
24551 2021-01-27 liuhongt <hongtao.liu@intel.com>
24554 * config/i386/sse.md (sse2_gt<mode>3): Drop !TARGET_XOP in condition.
24555 (*sse2_eq<mode>3): Ditto.
24557 2021-01-27 Jakub Jelinek <jakub@redhat.com>
24559 * tree-pass.h (PROP_trees): Rename to ...
24560 (PROP_gimple): ... this.
24561 * cfgexpand.c (pass_data_expand): Replace PROP_trees with PROP_gimple.
24562 * passes.c (execute_function_dump, execute_function_todo,
24563 execute_one_ipa_transform_pass, execute_one_pass): Likewise.
24564 * varpool.c (ctor_for_folding): Likewise.
24566 2021-01-27 Jakub Jelinek <jakub@redhat.com>
24568 PR tree-optimization/97260
24569 * varpool.c: Include tree-pass.h.
24570 (ctor_for_folding): In GENERIC return DECL_INITIAL for TREE_READONLY
24571 non-TREE_SIDE_EFFECTS automatic variables.
24573 2021-01-26 Paul Fee <paul.f.fee@gmail.com>
24575 * doc/cpp.texi (__cplusplus): Document value for -std=c++23
24577 * doc/invoke.texi: Document -std=c++23 and -std=gnu++23.
24578 * dwarf2out.c (highest_c_language): Recognise C++20 and C++23.
24579 (gen_compile_unit_die): Recognise C++23.
24581 2021-01-26 Jakub Jelinek <jakub@redhat.com>
24584 * dwarf2asm.c (dw2_assemble_integer): Cast DWARF2_ADDR_SIZE to int
24587 2021-01-26 Jakub Jelinek <jakub@redhat.com>
24590 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
24591 Use UINTVAL (shft_amnt) and UINTVAL (mask) instead of INTVAL (shft_amnt)
24592 and INTVAL (mask). Add && INTVAL (mask) > 0 condition.
24594 2021-01-26 Richard Biener <rguenther@suse.de>
24596 * gimple-pretty-print.c (dump_binary_rhs): Handle
24597 VEC_WIDEN_{PLUS,MINUS}_{LO,HI}_EXPR.
24599 2021-01-26 Richard Biener <rguenther@suse.de>
24601 PR middle-end/98726
24602 * tree.h (vector_cst_int_elt): Remove.
24603 * tree.c (vector_cst_int_elt): Use poly_wide_int for computations,
24606 2021-01-26 Andrew Stubbs <ams@codesourcery.com>
24608 * config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
24609 for V64DFmode min/max reductions.
24611 2021-01-26 Jakub Jelinek <jakub@redhat.com>
24613 * dwarf2asm.c (dw2_assemble_integer): Handle size twice as large
24614 as DWARF2_ADDR_SIZE if x is not a scalar int by emitting it as
24615 two halves, one with x and the other with const0_rtx, ordered
24616 depending on endianity.
24618 2021-01-26 Alexandre Oliva <oliva@adacore.com>
24620 * gimplify.c (gimplify_decl_expr): Skip asan marking calls for
24621 temporaries not seen in binding block, and not about to be
24622 added as gimple variables.
24624 2021-01-25 Martin Sebor <msebor@redhat.com>
24627 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust warning text.
24629 2021-01-25 Martin Liska <mliska@suse.cz>
24631 * value-prof.c (get_nth_most_common_value): Use %s instead
24634 2021-01-25 Jakub Jelinek <jakub@redhat.com>
24637 * configure.ac (HAVE_AS_GDWARF_5_DEBUG_FLAG): Only define if
24638 readelf -wi is able to read the emitted .debug_info back.
24639 * configure: Regenerated.
24641 2021-01-25 Martin Liska <mliska@suse.cz>
24643 PR gcov-profile/98739
24644 * common.opt: Add missing sign symbol.
24645 * value-prof.c (get_nth_most_common_value): Restore handling
24646 of PROFILE_REPRODUCIBILITY_PARALLEL_RUNS and
24647 PROFILE_REPRODUCIBILITY_MULTITHREADED.
24649 2021-01-25 Richard Biener <rguenther@suse.de>
24651 PR middle-end/98807
24652 * tree.c (vector_element_bits): Always use precision of
24653 the element type for boolean vectors.
24655 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
24657 * config/rtems.h (STARTFILE_SPEC): Remove qnolinkcmds.
24658 (ENDFILE_SPEC): Evaluate qnolinkcmds.
24660 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
24662 * config/rtems.h (STARTFILE_SPEC): Remove nostdlib and
24663 nostartfiles handling since this is already done by
24664 LINK_COMMAND_SPEC. Evaluate qnolinkcmds.
24665 (ENDFILE_SPEC): Remove nostdlib and nostartfiles handling since this
24666 is already done by LINK_COMMAND_SPEC.
24667 (LIB_SPECS): Remove nostdlib and nodefaultlibs handling since
24668 this is already done by LINK_COMMAND_SPEC. Remove qnolinkcmds
24671 2021-01-25 Jakub Jelinek <jakub@redhat.com>
24674 * fold-const-call.c (host_size_t_cst_p): Renamed to ...
24675 (size_t_cst_p): ... this. Check and store unsigned HOST_WIDE_INT
24676 value rather than host size_t.
24677 (fold_const_call): Change type of s2 from size_t to
24678 unsigned HOST_WIDE_INT. Use size_t_cst_p instead of
24679 host_size_t_cst_p. For strncmp calls, pass MIN (s2, SIZE_MAX)
24680 instead of s2 as last argument.
24682 2021-01-25 Tamar Christina <tamar.christina@arm.com>
24684 * config/arm/iterators.md (rotsplit1, rotsplit2, conj_op, fcmac1,
24685 VCMLA_OP, VCMUL_OP): New.
24686 * config/arm/mve.md (mve_vcmlaq<mve_rot><mode>): Support vec_dup 0.
24687 * config/arm/neon.md (cmul<conj_op><mode>3): New.
24688 * config/arm/unspecs.md (UNSPEC_VCMLA_CONJ, UNSPEC_VCMLA180_CONJ,
24689 UNSPEC_VCMUL_CONJ): New.
24690 * config/arm/vec-common.md (cmul<conj_op><mode>3, arm_vcmla<rot><mode>,
24691 cml<fcmac1><conj_op><mode>4): New.
24693 2021-01-23 Jakub Jelinek <jakub@redhat.com>
24696 * config/rs6000/mmintrin.h (__m64): Add __may_alias__ attribute.
24698 2021-01-22 Jonathan Wright <jonathan.wright@arm.com>
24700 * config/aarch64/aarch64-simd-builtins.def: Add mla builtin
24702 * config/aarch64/arm_neon.h (vmla_s8): Use RTL builtin rather
24704 (vmla_s16): Likewise.
24705 (vmla_s32): Likewise.
24706 (vmla_u8): Likewise.
24707 (vmla_u16): Likewise.
24708 (vmla_u32): Likewise.
24709 (vmlaq_s8): Likewise.
24710 (vmlaq_s16): Likewise.
24711 (vmlaq_s32): Likewise.
24712 (vmlaq_u8): Likewise.
24713 (vmlaq_u16): Likewise.
24714 (vmlaq_u32): Likewise.
24716 2021-01-22 David Malcolm <dmalcolm@redhat.com>
24718 * doc/invoke.texi (GCC_EXTRA_DIAGNOSTIC_OUTPUT): Add @findex
24721 2021-01-22 Jakub Jelinek <jakub@redhat.com>
24724 * dwarf2out.c (output_file_names): For -gdwarf-5, if there are no
24725 filenames to emit, still emit the required 0 index directory and
24726 filename entries that match DW_AT_comp_dir and DW_AT_name of the
24729 2021-01-22 Marek Polacek <polacek@redhat.com>
24732 * doc/invoke.texi: Update C++ ABI Version 15 description.
24734 2021-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24736 PR tree-optimization/98766
24737 * tree-ssa-math-opts.c (convert_mult_to_fma): Use maybe_le when
24738 comparing against type size with param_avoid_fma_max_bits.
24740 2021-01-22 Richard Biener <rguenther@suse.de>
24742 PR middle-end/98793
24743 * tree.c (vector_element_bits): Key single-bit bool vector on
24744 integer mode rather than not vector mode.
24746 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
24749 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
24750 Generate ARRAY_REF(VIEW_CONVERT_EXPR) for P8 and later
24752 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): Update
24753 to call different path for P8 and P9.
24754 (rs6000_expand_vector_set_var_p9): New function.
24755 (rs6000_expand_vector_set_var_p8): New function.
24757 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
24761 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
24762 Ajdust variable index vec_insert from address dereference to
24763 ARRAY_REF(VIEW_CONVERT_EXPR) tree expression.
24764 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
24766 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): New function.
24768 2021-01-22 Martin Liska <mliska@suse.cz>
24770 PR gcov-profile/98739
24771 * profile.c (compute_value_histograms): Drop time profile for
24772 -fprofile-reproducible=multithreaded.
24774 2021-01-22 Nathan Sidwell <nathan@acm.org>
24776 * gcc.c (process_command): Don't check OPT_SPECIAL_input_file
24779 2021-01-22 Richard Biener <rguenther@suse.de>
24781 PR middle-end/98773
24782 * tree-data-ref.c (initalize_matrix_A): Revert previous
24783 change, retaining failing on HOST_WIDE_INT_MIN CHREC_RIGHT.
24785 2021-01-22 Jakub Jelinek <jakub@redhat.com>
24787 PR tree-optimization/90248
24788 * match.pd (X cmp 0.0 ? 1.0 : -1.0 -> copysign(1, +-X),
24789 X cmp 0.0 ? -1.0 : +1.0 -> copysign(1, -+X)): Remove
24791 (X * (X cmp 0.0 ? 1.0 : -1.0) -> +-abs(X),
24792 X * (X cmp 0.0 ? -1.0 : 1.0) -> +-abs(X)): New simplifications.
24794 2021-01-22 Jakub Jelinek <jakub@redhat.com>
24796 PR tree-optimization/98255
24797 * tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
24798 extend index - low_bound from sizetype's precision rather than index
24800 (get_addr_base_and_unit_offset_1): Likewise.
24801 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
24802 * gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
24804 2021-01-22 Richard Biener <rguenther@suse.de>
24806 PR tree-optimization/98786
24807 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Avoid
24808 adding new uses of abnormals. Verify we deal with a conditional
24811 2021-01-22 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24814 * optc-save-gen.awk: Add arm_fp16_format to checked_options.
24816 2021-01-22 liuhongt <hongtao.liu@intel.com>
24820 * config/i386/sse.md (VI_128_256): New mode iterator.
24821 (*avx_cmp<mode>3_1, *avx_cmp<mode>3_2, *avx_cmp<mode>3_3,
24822 *avx_cmp<mode>3_4, *avx2_eq<mode>3, *avx2_pcmp<mode>3_1,
24823 *avx2_pcmp<mode>3_2, *avx2_gt<mode>3): New
24824 define_insn_and_split to lower avx512 vector comparison to avx
24825 version when dest is vector.
24826 (*<avx512>_cmp<mode>3,*<avx512>_cmp<mode>3,*<avx512>_ucmp<mode>3):
24827 define_insn_and_split for negating the comparison result.
24828 * config/i386/predicates.md (float_vector_all_ones_operand):
24830 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Use
24831 general NOT operator without UNSPEC_MASKOP.
24833 2021-01-21 Vladimir N. Makarov <vmakarov@redhat.com>
24835 PR rtl-optimization/98777
24836 * lra-int.h (lra_pmode_pseudo): New extern.
24837 * lra.c (lra_pmode_pseudo): New global.
24839 * lra-eliminations.c (eliminate_regs_in_insn): Use it.
24841 2021-01-21 Ilya Leoshkevich <iii@linux.ibm.com>
24843 * fwprop.c (fwprop_propagation::classify_result): Allow
24844 (subreg (mem)) simplifications.
24846 2021-01-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24848 * config/aarch64/aarch64-simd.md (aarch64_sqdml<SBINQOPS:as>l<mode>):
24850 (aarch64_sqdmlal<mode>): ... This...
24851 (aarch64_sqdmlsl<mode>): ... And this.
24852 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Split into...
24853 (aarch64_sqdmlal_lane<mode>): ... This...
24854 (aarch64_sqdmlsl_lane<mode>): ... And this.
24855 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Split into...
24856 (aarch64_sqdmlsl_laneq<mode>): ... This...
24857 (aarch64_sqdmlal_laneq<mode>): ... And this.
24858 (aarch64_sqdml<SBINQOPS:as>l_n<mode>): Split into...
24859 (aarch64_sqdmlsl_n<mode>): ... This...
24860 (aarch64_sqdmlal_n<mode>): ... And this.
24861 (aarch64_sqdml<SBINQOPS:as>l2<mode>_internal): Split into...
24862 (aarch64_sqdmlal2<mode>_internal): ... This...
24863 (aarch64_sqdmlsl2<mode>_internal): ... And this.
24865 2021-01-21 Christophe Lyon <christophe.lyon@linaro.org>
24867 * config/arm/arm_mve.h (__arm_vcmpneq_s8): Fix return type.
24869 2021-01-21 Andrea Corallo <andrea.corallo@arm.com>
24872 * doc/sourcebuild.texi (arm_thumb2_no_arm_v8_1_lob): Document.
24874 2021-01-21 liuhongt <hongtao.liu@intel.com>
24876 PR rtl-optimization/98694
24877 * regcprop.c (copy_value): If SRC had been assigned a mode
24878 narrower than the copy, we can't link DEST into the chain even
24879 they have same hard_regno_nregs(i.e. HImode/SImode in i386
24882 2021-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24884 * config/aarch64/aarch64-simd.md (aarch64_get_lane<mode>):
24885 Convert to define_insn_and_split. Split into simple move when moving
24888 2021-01-20 Segher Boessenkool <segher@kernel.crashing.org>
24890 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Change assert.
24891 Adjust comment. Simplify code.
24893 2021-01-20 Jakub Jelinek <jakub@redhat.com>
24896 * dwarf2out.c (reset_indirect_string): Also reset indirect strings
24897 with DW_FORM_line_strp form.
24898 (prune_unused_types_update_strings): Don't add into debug_str_hash
24899 indirect strings with DW_FORM_line_strp form.
24900 (adjust_name_comp_dir): New function.
24901 (dwarf2out_finish): Call it on CU DIEs after resetting
24902 debug_line_str_hash.
24904 2021-01-20 Vladimir N. Makarov <vmakarov@redhat.com>
24906 PR rtl-optimization/98722
24907 * lra-eliminations.c (eliminate_regs_in_insn): Check that target
24908 has no 3-op add insn to transform insns containing two pluses.
24910 2021-01-20 Richard Biener <rguenther@suse.de>
24912 * hwint.h (add_hwi): New function.
24913 (mul_hwi): Likewise.
24914 * tree-data-ref.c (initialize_matrix_A): Properly translate
24915 tree constants and avoid HOST_WIDE_INT_MIN.
24916 (lambda_matrix_row_add): Avoid undefined integer overflow
24917 and return true on such overflow.
24918 (lambda_matrix_right_hermite): Handle overflow from
24919 lambda_matrix_row_add gracefully. Simplify previous fix.
24920 (analyze_subscript_affine_affine): Likewise.
24922 2021-01-20 Eugene Rozenfeld <erozen@microsoft.com>
24924 PR tree-optimization/96674
24925 * match.pd: New patterns: x < y || y == XXX_MIN --> x <= y - 1
24926 x >= y && y != XXX_MIN --> x > y - 1
24928 2021-01-20 Richard Sandiford <richard.sandiford@arm.com>
24930 PR tree-optimization/98535
24931 * tree-vect-slp.c (duplicate_and_interleave): Use quick_grow_cleared.
24932 If the high and low permutes are the same, remove the high permutes
24933 from the working set and only continue with the low ones.
24935 2021-01-20 Jakub Jelinek <jakub@redhat.com>
24937 PR tree-optimization/98721
24938 * builtins.c (access_ref::inform_access): Don't assume
24939 SSA_NAME_IDENTIFIER must be non-NULL. Print messages about
24940 object whenever allocfn is NULL, rather than only when DECL_P
24941 is true. Use %qE instead of %qD for that. Formatting fixes.
24943 2021-01-20 Richard Biener <rguenther@suse.de>
24945 PR tree-optimization/98758
24946 * tree-data-ref.c (int_divides_p): Use lambda_int arguments.
24947 (lambda_matrix_right_hermite): Avoid undefinedness with
24948 signed integer abs and multiplication.
24949 (analyze_subscript_affine_affine): Use lambda_int.
24951 2021-01-20 David Malcolm <dmalcolm@redhat.com>
24954 * dwarf2out.c (output_line_info): Rename static variable
24955 "generation", moving it out of the function to...
24956 (output_line_info_generation): New.
24957 (init_sections_and_labels): Likewise, renaming the variable to...
24958 (init_sections_and_labels_generation): New.
24959 (dwarf2out_c_finalize): Reset the new variables.
24961 2021-01-19 Martin Sebor <msebor@redhat.com>
24963 PR middle-end/98664
24964 * tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
24965 all functions, even if they're not declared artificial or inline.
24966 * tree.c (tree_inlined_location): Use macro expansion location
24967 only if scope traversal fails to expose one.
24969 2021-01-19 Richard Sandiford <richard.sandiford@arm.com>
24971 PR rtl-optimization/92294
24972 * alias.c (compare_base_symbol_refs): Take an extra parameter
24973 and add the distance between two symbols to it. Enshrine in
24974 comments that -1 means "either 0 or 1, but we can't tell
24975 which at compile time".
24976 (memrefs_conflict_p): Update call accordingly.
24977 (rtx_equal_for_memref_p): Likewise. Take the distance between symbols
24980 2021-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24982 * config/aarch64/aarch64-simd-builtins.def (sqshl, uqshl,
24983 sqrshl, uqrshl, sqadd, uqadd, sqsub, uqsub, suqadd, usqadd, sqmovn,
24984 uqmovn, sqxtn2, uqxtn2, sqabs, sqneg, sqdmlal, sqdmlsl, sqdmlal_lane,
24985 sqdmlsl_lane, sqdmlal_laneq, sqdmlsl_laneq, sqdmlal_n, sqdmlsl_n,
24986 sqdmlal2, sqdmlsl2, sqdmlal2_lane, sqdmlsl2_lane, sqdmlal2_laneq,
24987 sqdmlsl2_laneq, sqdmlal2_n, sqdmlsl2_n, sqdmull, sqdmull_lane,
24988 sqdmull_laneq, sqdmull_n, sqdmull2, sqdmull2_lane, sqdmull2_laneq,
24989 sqdmull2_n, sqdmulh, sqrdmulh, sqdmulh_lane, sqdmulh_laneq,
24990 sqrdmulh_lane, sqrdmulh_laneq, sqshrun_n, sqrshrun_n, sqshrn_n,
24991 uqshrn_n, sqrshrn_n, uqrshrn_n, sqshlu_n, sqshl_n, uqshl_n, sqrdmlah,
24992 sqrdmlsh, sqrdmlah_lane, sqrdmlsh_lane, sqrdmlah_laneq, sqrdmlsh_laneq,
24993 sqmovun): Use NONE flags.
24995 2021-01-19 Richard Biener <rguenther@suse.de>
24998 * ipa-modref.c (analyze_stmt): Only record a summary for a
25001 2021-01-19 Richard Biener <rguenther@suse.de>
25003 PR middle-end/98638
25004 * tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
25006 2021-01-19 Daniel Hellstrom <daniel@gaisler.com>
25008 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
25009 built-in define __FIX_LEON3FT_TN0018.
25011 2021-01-19 Richard Biener <rguenther@suse.de>
25014 * tree-inline.c (tree_function_versioning): Set input_location
25015 to UNKNOWN_LOCATION throughout the function.
25017 2021-01-19 Tobias Burnus <tobias@codesourcery.com>
25020 * omp-low.c (lower_omp_target): Handle nonpointer is_device_ptr.
25022 2021-01-19 Martin Jambor <mjambor@suse.cz>
25025 * ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
25026 whether non-call exceptions allow removal of a statement.
25027 (isra_analyze_call): Pass the appropriate function to
25028 ssa_name_only_returned_p.
25030 2021-01-19 Geng Qi <gengqi@linux.alibaba.com>
25032 * config/riscv/arch-canonicalize (longext_sort): New function for
25033 sorting 'multi-letter'.
25034 * config/riscv/multilib-generator: Adjusting the loop of 'alt' in
25035 'alts'. The 'arch' may not be the first of 'alts'.
25036 (_expand_combination): Add underline for the 'ext' without '*'.
25037 This is because, a single-letter extension can always be treated well
25038 with a '_' prefix, but it cannot be separated out if it is appended
25041 2021-01-18 Vladimir N. Makarov <vmakarov@redhat.com>
25044 * ira.c (ira): Skip abnormal critical edge splitting.
25046 2021-01-18 Jakub Jelinek <jakub@redhat.com>
25048 PR tree-optimization/98727
25049 * tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
25050 second .MUL_OVERFLOW operand for signed multiplication with overflow
25051 checking if the second operand of multiplication is not constant.
25053 2021-01-18 David Edelsohn <dje.gcc@gmail.com>
25055 * doc/invoke.texi (-gdwarf): TPF defaults to version 2 and AIX
25056 defaults to version 4.
25058 2021-01-18 David Malcolm <dmalcolm@redhat.com>
25060 * attribs.h (fndecl_dealloc_argno): New decl.
25061 * builtins.c (call_dealloc_argno): Split out second half of
25063 (fndecl_dealloc_argno): New.
25064 * doc/extend.texi (Common Function Attributes): Document the
25065 interaction between the analyzer and the malloc attribute.
25066 * doc/invoke.texi (Static Analyzer Options): Likewise.
25068 2021-01-17 David Edelsohn <dje.gcc@gmail.com>
25070 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Override
25071 dwarf_version to 4.
25072 * config/rs6000/aix72.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
25074 2021-01-17 Martin Jambor <mjambor@suse.cz>
25077 * cgraph.c (clone_of_p): Check also former_clone_of as we climb
25080 2021-01-17 Mark Wielaard <mark@klomp.org>
25082 * common.opt (gdwarf-): Init(5).
25083 * doc/invoke.texi (-gdwarf): Document default to 5.
25085 2021-01-16 Kwok Cheung Yeung <kcy@codesourcery.com>
25087 * builtin-types.def
25088 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
25090 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
25091 ...this. Add extra argument.
25092 * gimplify.c (omp_default_clause): Ensure that event handle is
25093 firstprivate in a task region.
25094 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_DETACH.
25095 (gimplify_adjust_omp_clauses): Likewise.
25096 * omp-builtins.def (BUILT_IN_GOMP_TASK): Change function type to
25097 BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR.
25098 * omp-expand.c (expand_task_call): Add GOMP_TASK_FLAG_DETACH to flags
25099 if detach clause specified. Add detach argument when generating
25101 * omp-low.c (scan_sharing_clauses): Setup data environment for detach
25103 (finish_taskreg_scan): Move field for variable containing the event
25104 handle to the front of the struct.
25105 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_DETACH. Fix
25107 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
25108 OMP_CLAUSE_DETACH clause.
25109 (convert_local_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
25110 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_DETACH.
25111 * tree.c (omp_clause_num_ops): Add entry for OMP_CLAUSE_DETACH.
25113 (omp_clause_code_name): Add entry for OMP_CLAUSE_DETACH. Fix
25115 (walk_tree_1): Handle OMP_CLAUSE_DETACH.
25117 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
25119 * config/nios2/t-rtems: Reset all MULTILIB_* variables. Shorten
25120 multilib directory names. Use MULTILIB_REQUIRED instead of
25121 MULTILIB_EXCEPTIONS. Add -mhw-mul -mhw-mulx -mhw-div
25122 -mcustom-fpu-cfg=fph2 multilib.
25124 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
25126 * config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
25127 (nios2_init_fpu_configs): Provide register values for new
25128 -mcustom-fpu-cfg=fph2 option variant.
25129 * doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
25132 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
25134 * config/nios2/nios2.c (nios2_custom_check_insns): Remove
25135 custom instruction warnings.
25137 2021-01-16 Jakub Jelinek <jakub@redhat.com>
25139 PR tree-optimization/96669
25140 * match.pd ((CST << x) & 1 -> x == 0): New simplification.
25142 2021-01-16 Jakub Jelinek <jakub@redhat.com>
25144 PR tree-optimization/96271
25145 * passes.def: Pass false argument to first two pass_cd_dce
25146 instances and true to last instance. Add comment that
25147 last instance rewrites no longer addressed locals.
25148 * tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
25150 (pass_cd_dce::set_pass_param): New method.
25151 (pass_cd_dce::execute): Return TODO_update_address_taken from
25152 last cd_dce instance.
25154 2021-01-15 Carl Love <cel@us.ibm.com>
25156 * config/rs6000/altivec.h (vec_mulh, vec_div, vec_dive, vec_mod):
25158 * config/rs6000/altivec.md (VIlong): Move define to file vsx.md.
25159 * config/rs6000/rs6000-builtin.def (DIVES_V4SI, DIVES_V2DI,
25160 DIVEU_V4SI, DIVEU_V2DI, DIVS_V4SI, DIVS_V2DI, DIVU_V4SI,
25161 DIVU_V2DI, MODS_V2DI, MODS_V4SI, MODU_V2DI, MODU_V4SI,
25162 MULHS_V2DI, MULHS_V4SI, MULHU_V2DI, MULHU_V4SI, MULLD_V2DI):
25163 Add builtin define.
25164 (MULH, DIVE, MOD): Add new BU_P10_OVERLOAD_2 definitions.
25165 * config/rs6000/rs6000-call.c (VSX_BUILTIN_VEC_DIV,
25166 VSX_BUILTIN_VEC_DIVE, P10_BUILTIN_VEC_MOD, P10_BUILTIN_VEC_MULH):
25167 New overloaded definitions.
25168 (builtin_function_type) [P10V_BUILTIN_DIVEU_V4SI,
25169 P10V_BUILTIN_DIVEU_V2DI, P10V_BUILTIN_DIVU_V4SI,
25170 P10V_BUILTIN_DIVU_V2DI, P10V_BUILTIN_MODU_V2DI,
25171 P10V_BUILTIN_MODU_V4SI, P10V_BUILTIN_MULHU_V2DI,
25172 P10V_BUILTIN_MULHU_V4SI]: Add case
25173 statement for builtins.
25174 * config/rs6000/rs6000.md (bits): Add new attribute sizes V4SI, V2DI.
25175 * config/rs6000/vsx.md (VIlong): Moved from config/rs6000/altivec.md.
25176 (UNSPEC_VDIVES, UNSPEC_VDIVEU): New unspec definitions.
25177 (vsx_mul_v2di): Add if TARGET_POWER10 statement.
25178 (vsx_udiv_v2di): Add if TARGET_POWER10 statement.
25179 (dives_<mode>, diveu_<mode>, div<mode>3, uvdiv<mode>3,
25180 mods_<mode>, modu_<mode>, mulhs_<mode>, mulhu_<mode>, mulv2di3):
25181 Add define_insn, mode is VIlong.
25182 * doc/extend.texi (vec_mulh, vec_mul, vec_div, vec_dive, vec_mod):
25183 Add builtin descriptions.
25185 2021-01-15 Eric Botcazou <ebotcazou@adacore.com>
25187 * final.c (final_start_function_1): Reset force_source_line.
25189 2021-01-15 Jakub Jelinek <jakub@redhat.com>
25191 PR tree-optimization/96669
25192 * match.pd (((1 << A) & 1) != 0 -> A == 0,
25193 ((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
25194 possibly different power of two constants and to right shift too.
25196 2021-01-15 Jakub Jelinek <jakub@redhat.com>
25198 PR tree-optimization/96681
25199 * match.pd ((x < 0) ^ (y < 0) to (x ^ y) < 0): New simplification.
25200 ((x >= 0) ^ (y >= 0) to (x ^ y) < 0): Likewise.
25201 ((x < 0) ^ (y >= 0) to (x ^ y) >= 0): Likewise.
25202 ((x >= 0) ^ (y < 0) to (x ^ y) >= 0): Likewise.
25204 2021-01-15 Alexandre Oliva <oliva@adacore.com>
25206 * opts.c (gen_command_line_string): Exclude -dumpbase-ext.
25208 2021-01-15 Tamar Christina <tamar.christina@arm.com>
25210 * config/aarch64/aarch64-simd.md (cml<fcmac1><conj_op><mode>4,
25211 cmul<conj_op><mode>3): New.
25212 * config/aarch64/iterators.md (UNSPEC_FCMUL,
25213 UNSPEC_FCMUL180, UNSPEC_FCMLA_CONJ, UNSPEC_FCMLA180_CONJ,
25214 UNSPEC_CMLA_CONJ, UNSPEC_CMLA180_CONJ, UNSPEC_CMUL, UNSPEC_CMUL180,
25215 FCMLA_OP, FCMUL_OP, conj_op, rotsplit1, rotsplit2, fcmac1, sve_rot1,
25216 sve_rot2, SVE2_INT_CMLA_OP, SVE2_INT_CMUL_OP, SVE2_INT_CADD_OP): New.
25217 (rot): Add UNSPEC_FCMUL, UNSPEC_FCMUL180.
25218 (rot_op): Renamed to conj_op.
25219 * config/aarch64/aarch64-sve.md (cml<fcmac1><conj_op><mode>4,
25220 cmul<conj_op><mode>3): New.
25221 * config/aarch64/aarch64-sve2.md (cml<fcmac1><conj_op><mode>4,
25222 cmul<conj_op><mode>3): New.
25224 2021-01-15 David Malcolm <dmalcolm@redhat.com>
25228 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
25229 Escape the tempfile name when constructing the expected output.
25231 2021-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25233 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlsl_hi<mode>):
25235 (aarch64_<su>mlsl_hi<mode>): ... This.
25236 (aarch64_<su>mlsl_hi<mode>): Define.
25237 (*aarch64_<su>mlsl<mode): Rename to...
25238 (aarch64_<su>mlsl<mode): ... This.
25239 * config/aarch64/aarch64-simd-builtins.def (smlsl, umlsl,
25240 smlsl_hi, umlsl_hi): Define builtins.
25241 * config/aarch64/arm_neon.h (vmlsl_high_s8, vmlsl_high_s16,
25242 vmlsl_high_s32, vmlsl_high_u8, vmlsl_high_u16, vmlsl_high_u32,
25243 vmlsl_s8, vmlsl_s16, vmlsl_s32, vmlsl_u8,
25244 vmlsl_u16, vmlsl_u32): Reimplement with builtins.
25246 2021-01-15 Uroš Bizjak <ubizjak@gmail.com>
25248 * config/i386/i386-c.c (ix86_target_macros):
25249 Use cpp_define_formatted for __SIZEOF_FLOAT80__ definition.
25251 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
25254 * config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
25255 * Makefile.in (RTL_SSA_H): New variable.
25256 * config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
25257 * config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
25258 * config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
25260 * config/aarch64/aarch64-cc-fusion.cc: New file.
25262 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
25264 * recog.h (insn_change_watermark::~insn_change_watermark): Avoid
25265 calling cancel_changes for changes that no longer exist.
25267 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
25269 * rtl-ssa/functions.h (function_info::ref_defs): Rename to...
25270 (function_info::reg_defs): ...this.
25271 * rtl-ssa/member-fns.inl (function_info::ref_defs): Rename to...
25272 (function_info::reg_defs): ...this.
25274 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
25277 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
25279 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
25282 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
25285 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
25287 2021-01-15 Richard Biener <rguenther@suse.de>
25289 PR tree-optimization/96376
25290 * tree-vect-stmts.c (get_load_store_type): Disregard alignment
25291 for VMAT_INVARIANT.
25293 2021-01-15 Martin Liska <mliska@suse.cz>
25295 * doc/install.texi: Document that some tests need pytest module.
25296 * doc/sourcebuild.texi: Likewise.
25298 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
25301 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
25303 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
25305 * config/arm/mve.md (mve_vshrq_n_s<mode>_imm): New entry.
25306 (mve_vshrq_n_u<mode>_imm): Likewise.
25307 * config/arm/neon.md (vashr<mode>3, vlshr<mode>3): Move to ...
25308 * config/arm/vec-common.md: ... here.
25310 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
25312 * config/arm/mve.md (mve_vshlq_<supf><mode>): Move to
25314 * config/arm/neon.md (vashl<mode>3): Delete.
25315 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): New.
25316 (vasl<mode>3): New expander.
25318 2021-01-15 Richard Biener <rguenther@suse.de>
25320 PR tree-optimization/98685
25321 * tree-vect-slp.c (vect_schedule_slp_node): Refactor handling
25322 of vector extern defs.
25324 2021-01-14 David Malcolm <dmalcolm@redhat.com>
25327 * diagnostic.c (diagnostic_kind_text): Break out this array
25329 (diagnostic_build_prefix): ...here.
25330 (fancy_abort): Detect when diagnostic_initialize has not yet been
25331 called and fall back to a minimal implementation of printing the
25332 ICE, rather than segfaulting in internal_error.
25334 2021-01-14 David Malcolm <dmalcolm@redhat.com>
25336 * diagnostic.c (diagnostic_initialize): Eliminate
25337 parseable_fixits_p in favor of initializing extra_output_kind from
25338 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
25339 (convert_column_unit): New function, split out from...
25340 (diagnostic_converted_column): ...this.
25341 (print_parseable_fixits): Add "column_unit" and "tabstop" params.
25342 Use them to call convert_column_unit on the column values.
25343 (diagnostic_report_diagnostic): Eliminate conditional on
25344 parseable_fixits_p in favor of a switch statement on
25345 extra_output_kind, passing the appropriate values to the new
25346 params of print_parseable_fixits.
25347 (selftest::test_print_parseable_fixits_none): Update for new
25348 params of print_parseable_fixits.
25349 (selftest::test_print_parseable_fixits_insert): Likewise.
25350 (selftest::test_print_parseable_fixits_remove): Likewise.
25351 (selftest::test_print_parseable_fixits_replace): Likewise.
25352 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
25354 (selftest::diagnostic_c_tests): Call it.
25355 * diagnostic.h (enum diagnostics_extra_output_kind): New.
25356 (diagnostic_context::parseable_fixits_p): Delete field in favor
25358 (diagnostic_context::extra_output_kind): ...this new field.
25359 * doc/invoke.texi (Environment Variables): Add
25360 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
25361 * opts.c (common_handle_option): Update handling of
25362 OPT_fdiagnostics_parseable_fixits for change to diagnostic_context
25365 2021-01-14 Tamar Christina <tamar.christina@arm.com>
25367 * tree-vect-slp-patterns.c (class complex_operations_pattern,
25368 complex_operations_pattern::matches,
25369 complex_operations_pattern::recognize,
25370 complex_operations_pattern::build): New.
25371 (slp_patterns): Use it.
25373 2021-01-14 Tamar Christina <tamar.christina@arm.com>
25375 * internal-fn.def (COMPLEX_FMS, COMPLEX_FMS_CONJ): New.
25376 * optabs.def (cmls_optab, cmls_conj_optab): New.
25377 * doc/md.texi: Document them.
25378 * tree-vect-slp-patterns.c (class complex_fms_pattern,
25379 complex_fms_pattern::matches, complex_fms_pattern::recognize,
25380 complex_fms_pattern::build): New.
25382 2021-01-14 Tamar Christina <tamar.christina@arm.com>
25384 * internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
25385 * optabs.def (cmla_optab, cmla_conj_optab): New.
25386 * doc/md.texi: Document them.
25387 * tree-vect-slp-patterns.c (vect_match_call_p,
25388 class complex_fma_pattern, vect_slp_reset_pattern,
25389 complex_fma_pattern::matches, complex_fma_pattern::recognize,
25390 complex_fma_pattern::build): New.
25392 2021-01-14 Tamar Christina <tamar.christina@arm.com>
25394 * internal-fn.def (COMPLEX_MUL, COMPLEX_MUL_CONJ): New.
25395 * optabs.def (cmul_optab, cmul_conj_optab): New.
25396 * doc/md.texi: Document them.
25397 * tree-vect-slp-patterns.c (vect_match_call_complex_mla,
25398 vect_normalize_conj_loc, is_eq_or_top, vect_validate_multiplication,
25399 vect_build_combine_node, class complex_mul_pattern,
25400 complex_mul_pattern::matches, complex_mul_pattern::recognize,
25401 complex_mul_pattern::build): New.
25403 2021-01-14 Tamar Christina <tamar.christina@arm.com>
25405 * tree-vect-slp.c (optimize_load_redistribution_1): New.
25406 (optimize_load_redistribution, vect_is_slp_load_node): New.
25407 (vect_match_slp_patterns): Use it.
25409 2021-01-14 Tamar Christina <tamar.christina@arm.com>
25411 * tree-vect-slp-patterns.c (complex_add_pattern::build):
25414 2021-01-14 Thomas Schwinge <thomas@codesourcery.com>
25416 * config/gcn/mkoffload.c (main): Create an offload image only in
25417 64-bit configurations.
25419 2021-01-14 H.J. Lu <hjl.tools@gmail.com>
25422 * config/i386/i386-options.c (ix86_option_override_internal):
25423 Issue an error for -fcf-protection with CF_BRANCH when compiling
25424 for 32-bit non-TARGET_CMOV targets.
25426 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
25429 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p):
25430 Remove declaration and initialization of shadow variable "ret".
25431 (ix86_option_override_internal): Remove delcaration of
25432 shadow variable "i". Redeclare shadowed variable to unsigned.
25433 * common/config/i386/i386-common.c (pta_size): Redeclare to unsigned.
25434 * config/i386/i386-builtins.c (get_builtin_code_for_version):
25435 Update for redeclaration.
25436 * config/i386/i386.h (pta_size): Ditto.
25438 2021-01-14 Richard Biener <rguenther@suse.de>
25440 PR tree-optimization/98674
25441 * tree-data-ref.c (base_supports_access_fn_components_p): New.
25442 (initialize_data_dependence_relation): For two bases without
25443 possible access fns resort to type size equality when determining
25444 shape compatibility.
25446 2021-01-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
25449 * config/arm/arm_neon.h: Replace calls to __builtin_vcge* by
25450 <=, >= operators in vcle and vcge intrinsics respectively.
25451 * config/arm/arm_neon_builtins.def: Remove entry for
25454 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
25457 * config/i386/i386-options.c (ix86_function_specific_save):
25458 Remove redundant assignment to opts->x_ix86_branch_cost.
25459 * config/i386/i386.c (ix86_prefetch_sse):
25460 Rename from x86_prefetch_sse. Update all uses.
25461 * config/i386/i386.h: Update for rename.
25462 * config/i386/i386-options.h: Ditto.
25464 2021-01-14 Jakub Jelinek <jakub@redhat.com>
25467 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3,
25468 *sse4_1_zero_extendv4hiv4si2_3, *sse4_1_zero_extendv2siv2di2_3):
25469 Use Bm instead of m for non-avx. Add isa attribute.
25471 2021-01-14 Jakub Jelinek <jakub@redhat.com>
25473 PR tree-optimization/96688
25474 * match.pd (~(X >> Y) -> ~X >> Y): New simplification if
25475 ~X can be simplified.
25477 2021-01-14 Richard Sandiford <richard.sandiford@arm.com>
25479 * tree-vect-stmts.c (vect_model_load_cost): Account for unused
25480 IFN_LOAD_LANES results.
25482 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25484 * config/aarch64/aarch64-simd.md (aarch64_<su>xtl<mode>):
25486 (aarch64_xtn<mode>): Likewise.
25487 * config/aarch64/aarch64-simd-builtins.def (sxtl, uxtl, xtn):
25490 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
25492 (vmovl_s16): Likewise.
25493 (vmovl_s32): Likewise.
25494 (vmovl_u8): Likewise.
25495 (vmovl_u16): Likewise.
25496 (vmovl_u32): Likewise.
25497 (vmovn_s16): Likewise.
25498 (vmovn_s32): Likewise.
25499 (vmovn_s64): Likewise.
25500 (vmovn_u16): Likewise.
25501 (vmovn_u32): Likewise.
25502 (vmovn_u64): Likewise.
25504 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25506 * config/aarch64/aarch64-simd.md (aarch64_<su>qxtn2<mode>_le):
25508 (aarch64_<su>qxtn2<mode>_be): Likewise.
25509 (aarch64_<su>qxtn2<mode>): Likewise.
25510 * config/aarch64/aarch64-simd-builtins.def (sqxtn2, uqxtn2):
25512 * config/aarch64/iterators.md (SAT_TRUNC): Define code_iterator.
25513 (su): Handle ss_truncate and us_truncate.
25514 * config/aarch64/arm_neon.h (vqmovn_high_s16): Reimplement using
25516 (vqmovn_high_s32): Likewise.
25517 (vqmovn_high_s64): Likewise.
25518 (vqmovn_high_u16): Likewise.
25519 (vqmovn_high_u32): Likewise.
25520 (vqmovn_high_u64): Likewise.
25522 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25524 * config/aarch64/aarch64-simd.md (aarch64_xtn2<mode>_le):
25526 (aarch64_xtn2<mode>_be): Likewise.
25527 (aarch64_xtn2<mode>): Likewise.
25528 * config/aarch64/aarch64-simd-builtins.def (xtn2): Define
25530 * config/aarch64/arm_neon.h (vmovn_high_s16): Reimplement using
25532 (vmovn_high_s32): Likewise.
25533 (vmovn_high_s64): Likewise.
25534 (vmovn_high_u16): Likewise.
25535 (vmovn_high_u32): Likewise.
25536 (vmovn_high_u64): Likewise.
25538 2021-01-13 Stafford Horne <shorne@gmail.com>
25540 * config/or1k/or1k.h (ASM_PREFERRED_EH_DATA_FORMAT): New macro.
25542 2021-01-13 Stafford Horne <shorne@gmail.com>
25544 * config/or1k/linux.h (TARGET_ASM_FILE_END): Define macro.
25546 2021-01-13 Stafford Horne <shorne@gmail.com>
25548 * config/or1k/or1k.h (TARGET_CPU_CPP_BUILTINS): Add builtin
25549 define for __or1k_hard_float__.
25551 2021-01-13 Stafford Horne <shorne@gmail.com>
25553 * config/or1k/or1k.h (NO_PROFILE_COUNTERS): Define as 1.
25554 (PROFILE_HOOK): Define to call _mcount.
25555 (FUNCTION_PROFILER): Change from abort to no-op.
25557 2021-01-13 Jakub Jelinek <jakub@redhat.com>
25559 PR tree-optimization/96691
25560 * match.pd ((~X | C) ^ D -> (X | C) ^ (~D ^ C),
25561 (~X & C) ^ D -> (X & C) ^ (D ^ C)): New simplifications if
25562 (~D ^ C) or (D ^ C) can be simplified.
25564 2021-01-13 Richard Biener <rguenther@suse.de>
25566 PR tree-optimization/92645
25567 * match.pd (BIT_FIELD_REF to conversion): Delay canonicalization
25568 until after vector lowering.
25570 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
25572 * config/aarch64/aarch64-sve.md (fnma<mode>4): Extend from SVE_FULL_I
25574 (@aarch64_pred_fnma<mode>, cond_fnma<mode>, *cond_fnma<mode>_2)
25575 (*cond_fnma<mode>_4, *cond_fnma<mode>_any): Likewise.
25577 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
25579 * config/aarch64/aarch64-sve.md (fma<mode>4): Extend from SVE_FULL_I
25581 (@aarch64_pred_fma<mode>, cond_fma<mode>, *cond_fma<mode>_2)
25582 (*cond_fma<mode>_4, *cond_fma<mode>_any): Likewise.
25584 2021-01-13 Richard Biener <rguenther@suse.de>
25586 PR tree-optimization/92645
25587 * tree-vect-slp.c (vect_build_slp_tree_1): Relax supported
25588 BIT_FIELD_REF argument.
25589 (vect_build_slp_tree_2): Record the desired vector type
25590 on the external vector def.
25591 (vectorizable_slp_permutation): Handle required punning
25592 of existing vector defs.
25594 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
25596 * rtl-ssa/accesses.h (def_lookup): Fix order of comparison results.
25598 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
25600 * config/sh/sh.md (movsf_ie): Remove operands[2] test.
25602 2021-01-13 Samuel Thibault <samuel.thibault@ens-lyon.org>
25604 * config.gcc [$target == *-*-gnu*]: Enable
25605 'default_gnu_indirect_function'.
25607 2021-01-13 Jakub Jelinek <jakub@redhat.com>
25610 * optabs.c (expand_vec_perm_const): Don't force v0 and v1 into
25611 registers before calling targetm.vectorize.vec_perm_const, only after
25613 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const): Handle
25614 two argument permutation when one operand is zero vector and only
25615 after that force operands into registers.
25616 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_1): New
25617 define_insn_and_split pattern.
25618 (*avx512bw_zero_extendv32qiv32hi2_1): Likewise.
25619 (*avx512f_zero_extendv16hiv16si2_1): Likewise.
25620 (*avx2_zero_extendv8hiv8si2_1): Likewise.
25621 (*avx512f_zero_extendv8siv8di2_1): Likewise.
25622 (*avx2_zero_extendv4siv4di2_1): Likewise.
25623 * config/mips/mips.c (mips_vectorize_vec_perm_const): Force operands
25625 * config/arm/arm.c (arm_vectorize_vec_perm_const): Likewise.
25626 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Likewise.
25627 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const): Likewise.
25628 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const): Likewise.
25629 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const): Likewise.
25630 * config/gcn/gcn.c (gcn_vectorize_vec_perm_const): Likewise. Use std::swap.
25632 2021-01-13 Martin Liska <mliska@suse.cz>
25634 PR tree-optimization/98455
25635 * gimple-if-to-switch.cc (condition_info::record_phi_mapping):
25636 Record also virtual PHIs.
25637 (pass_if_to_switch::execute): Return TODO_cleanup_cfg only
25640 2021-01-13 Jonathan Wakely <jwakely@redhat.com>
25642 * doc/invoke.texi (C++ Modules): Fix typos.
25644 2021-01-13 Richard Biener <rguenther@suse.de>
25646 PR tree-optimization/98640
25647 * tree-ssa-sccvn.c (visit_nary_op): Do not try to
25648 handle plus or minus from a truncated operand to be
25651 2021-01-13 Jakub Jelinek <jakub@redhat.com>
25654 * config/i386/i386.md (*btr<mode>_1, *btr<mode>_2): New
25655 define_insn_and_split patterns.
25656 (splitter after *btr<mode>_2): New splitter.
25658 2021-01-13 Martin Liska <mliska@suse.cz>
25661 * cgraphunit.c (analyze_functions): Remove dead code.
25663 2021-01-13 Qian Jianhua <qianjh@cn.fujitsu.com>
25665 * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
25666 * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
25667 (a64fx_regmove_cost, a64fx_vector_cost): New.
25668 (a64fx_tunings): Use the new added cost tables.
25670 2021-01-13 Jakub Jelinek <jakub@redhat.com>
25673 * config/i386/predicates.md (pmovzx_parallel): New predicate.
25674 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): New
25675 define_insn_and_split pattern.
25676 (*sse4_1_zero_extendv4hiv4si2_3): Likewise.
25677 (*sse4_1_zero_extendv2siv2di2_3): Likewise.
25679 2021-01-13 Julian Brown <julian@codesourcery.com>
25681 * config/gcn/gcn.c (gcn_conditional_register_usage): Remove dead code
25682 to fix v0 register.
25684 2021-01-13 Julian Brown <julian@codesourcery.com>
25686 * config/gcn/gcn.c (gcn_md_reorg): Fix case where EXEC reg is live
25689 2021-01-13 Julian Brown <julian@codesourcery.com>
25691 * config/gcn/gcn-valu.md (recip<mode>2<exec>, recip<mode>2): Use unspec
25692 for reciprocal-approximation instructions.
25693 (div<mode>3): Use fused multiply-accumulate operations for reciprocal
25694 refinement and division result.
25695 * config/gcn/gcn.md (UNSPEC_RCP): New unspec constant.
25697 2021-01-13 Julian Brown <julian@codesourcery.com>
25699 * config/gcn/gcn-valu.md (subdf): Rename to...
25702 2021-01-12 Martin Liska <mliska@suse.cz>
25704 * gcov.c (source_info::debug): Fix printf format for 32-bit hosts.
25706 2021-01-12 Andrea Corallo <andrea.corallo@arm.com>
25708 * function-abi.h: Fix typo.
25710 2021-01-12 Christophe Lyon <christophe.lyon@linaro.org>
25714 * config/arm/arm.h (ARM_HAVE_NEON_V8QI_LDST): New macro.
25715 (ARM_HAVE_NEON_V16QI_LDST, ARM_HAVE_NEON_V4HI_LDST): Likewise.
25716 (ARM_HAVE_NEON_V8HI_LDST, ARM_HAVE_NEON_V2SI_LDST): Likewise.
25717 (ARM_HAVE_NEON_V4SI_LDST, ARM_HAVE_NEON_V4HF_LDST): Likewise.
25718 (ARM_HAVE_NEON_V8HF_LDST, ARM_HAVE_NEON_V4BF_LDST): Likewise.
25719 (ARM_HAVE_NEON_V8BF_LDST, ARM_HAVE_NEON_V2SF_LDST): Likewise.
25720 (ARM_HAVE_NEON_V4SF_LDST, ARM_HAVE_NEON_DI_LDST): Likewise.
25721 (ARM_HAVE_NEON_V2DI_LDST): Likewise.
25722 (ARM_HAVE_V8QI_LDST, ARM_HAVE_V16QI_LDST): Likewise.
25723 (ARM_HAVE_V4HI_LDST, ARM_HAVE_V8HI_LDST): Likewise.
25724 (ARM_HAVE_V2SI_LDST, ARM_HAVE_V4SI_LDST, ARM_HAVE_V4HF_LDST): Likewise.
25725 (ARM_HAVE_V8HF_LDST, ARM_HAVE_V4BF_LDST, ARM_HAVE_V8BF_LDST): Likewise.
25726 (ARM_HAVE_V2SF_LDST, ARM_HAVE_V4SF_LDST, ARM_HAVE_DI_LDST): Likewise.
25727 (ARM_HAVE_V2DI_LDST): Likewise.
25728 * config/arm/mve.md (*movmisalign<mode>_mve_store): New pattern.
25729 (*movmisalign<mode>_mve_load): New pattern.
25730 * config/arm/neon.md (movmisalign<mode>): Move to ...
25731 * config/arm/vec-common.md: ... here.
25733 2021-01-12 Vladimir N. Makarov <vmakarov@redhat.com>
25736 * lra-eliminations.c (eliminate_regs_in_insn): Add transformation
25737 of pattern 'plus (plus (hard reg, const), pseudo)'.
25739 2021-01-12 Richard Biener <rguenther@suse.de>
25741 PR tree-optimization/98550
25742 * tree-vect-slp.c (vect_record_max_nunits): Check whether
25743 the group size is a multiple of the vector element count.
25744 (vect_build_slp_tree_1): When we need to fail because
25745 the vector type choosen causes unrolling do so lazily
25746 without affecting matches only at the end to guide group splitting.
25748 2021-01-12 Martin Liska <mliska@suse.cz>
25751 * optc-save-gen.awk: Compare also n_target_save vars with
25754 2021-01-12 Martin Liska <mliska@suse.cz>
25756 * gcov.c (source_info::debug): New.
25757 (print_usage): Add --debug (-D) option.
25758 (process_args): Likewise.
25759 (generate_results): Call src->debug after
25760 accumulate_line_counts.
25761 (read_graph_file): Properly assign id for EXIT_BLOCK.
25762 * profile.c (branch_prob): Dump function body before it is
25765 2021-01-12 Jakub Jelinek <jakub@redhat.com>
25767 PR tree-optimization/98629
25768 * tree-ssa-math-opts.c (arith_overflow_check_p): Don't update use_stmt
25769 unless returning non-zero.
25771 2021-01-12 Jakub Jelinek <jakub@redhat.com>
25773 PR tree-optimization/95731
25774 * tree-ssa-reassoc.c (optimize_range_tests_cmp_bitwise): Also optimize
25775 x < 0 && y < 0 && z < 0 into (x | y | z) < 0 for signed x, y, z.
25776 (optimize_range_tests): Call optimize_range_tests_cmp_bitwise
25777 only after optimize_range_tests_var_bound.
25779 2021-01-12 Jakub Jelinek <jakub@redhat.com>
25781 * configure.ac: Ensure c/Make-lang.in comes first in @all_lang_makefrags@.
25782 * configure: Regenerated.
25784 2021-01-12 liuhongt <hongtao.liu@intel.com>
25787 * config/i386/i386-builtins.h (BUILTIN_DESC_SWAP_OPERANDS):
25789 * config/i386/i386-expand.c (ix86_expand_sse_comi): Delete
25792 2021-01-12 Alexandre Oliva <oliva@adacore.com>
25794 * ssa-iterators.h (end_imm_use_stmt_traverse): Forward
25796 (auto_end_imm_use_stmt_traverse): New struct.
25797 (FOR_EACH_IMM_USE_STMT): Use it.
25798 (BREAK_FROM_IMM_USE_STMT, RETURN_FROM_IMM_USE_STMT): Remove,
25800 * gimple-ssa-strength-reduction.c: ... here, ...
25801 * graphite-scop-detection.c: ... here, ...
25802 * ipa-modref.c, ipa-pure-const.c, ipa-sra.c: ... here, ...
25803 * tree-predcom.c, tree-ssa-ccp.c: ... here, ...
25804 * tree-ssa-dce.c, tree-ssa-dse.c: ... here, ...
25805 * tree-ssa-loop-ivopts.c, tree-ssa-math-opts.c: ... here, ...
25806 * tree-ssa-phiprop.c, tree-ssa.c: ... here, ...
25807 * tree-vect-slp.c: ... and here, ...
25808 * doc/tree-ssa.texi: ... and the example here.
25810 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25812 * config/aarch64/aarch64-sve.md (sdiv_pow2<mode>3): Extend from
25813 SVE_FULL_I to SVE_I. Generate an UNSPEC_PRED_X.
25814 (*sdiv_pow2<mode>3): New pattern.
25815 (@cond_<sve_int_op><mode>): Extend from SVE_FULL_I to SVE_I.
25816 Wrap the ASRD in an UNSPEC_PRED_X.
25817 (*cond_<sve_int_op><mode>_2): Likewise. Replace the UNSPEC_PRED_X
25818 predicate with a constant PTRUE, if it isn't already.
25819 (*cond_<sve_int_op><mode>_z): Replace with...
25820 (*cond_<sve_int_op><mode>_any): ...this new pattern.
25822 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25824 * config/aarch64/aarch64-sve.md (*cond_bic<mode>_2): Extend from
25825 SVE_FULL_I to SVE_I.
25826 (*cond_bic<mode>_any): Likewise.
25828 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25830 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart)
25831 (@aarch64_pred_<MUL_HIGHPART:optab><mode>): Extend from SVE_FULL_I
25834 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25836 * config/aarch64/aarch64-sve.md (<su>abd<mode>_3): Extend from
25837 SVE_FULL_I to SVE_I.
25838 (*aarch64_cond_<su>abd<mode>_2): Likewise.
25839 (*aarch64_cond_<su>abd<mode>_any): Likewise.
25840 (@aarch64_pred_<su>abd<mode>): Likewise. Use UNSPEC_PRED_X
25841 for the max and min but not for the minus.
25842 (*aarch64_cond_<su>abd<mode>_3): New pattern.
25844 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25846 * config/aarch64/iterators.md (SVE_24I): New iterator.
25847 * config/aarch64/aarch64-sve.md (*aarch64_adr<mode>_shift): Extend from
25848 SVE_FULL_SDI to SVE_24I. Use containers rather than elements.
25850 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25852 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_BINARY:optab><mode>)
25853 (*cond_<SVE_INT_BINARY:optab><mode>_2): Extend from SVE_FULL_I
25855 (*cond_<SVE_INT_BINARY:optab><mode>_3): Likewise.
25856 (*cond_<SVE_INT_BINARY:optab><mode>_any): Likewise.
25857 (*cond_<SVE_INT_BINARY:optab><mode>_2_const): Likewise.
25858 (*cond_<SVE_INT_BINARY:optab><mode>_any_const): Likewise.
25860 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25862 * config/aarch64/aarch64-sve.md (<SVE_INT_BINARY_IMM:optab><mode>3)
25863 (@aarch64_pred_<SVE_INT_BINARY_IMM:optab><mode>)
25864 (*post_ra_<SVE_INT_BINARY_IMM:optab><mode>3): Extend from SVE_FULL_I
25867 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25869 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3)
25870 (v<ASHIFT:optab><mode>3, @aarch64_pred_<optab><mode>)
25871 (*post_ra_v<ASHIFT:optab><mode>3): Extend from SVE_FULL_I to SVE_I.
25873 2021-01-11 Martin Liska <mliska@suse.cz>
25876 * symtab-clones.h (clone_info::release): Release
25877 symtab::m_clones with ggc_delete as it's a GGC memory.
25879 2021-01-11 Matthias Klose <doko@ubuntu.com>
25881 * Makefile.in (LINK_PROGRESS): Show the link target.
25883 2021-01-11 Richard Biener <rguenther@suse.de>
25885 PR tree-optimization/91403
25886 * tree-vect-data-refs.c (vect_analyze_group_access_1): Cap
25887 single-element interleaving group size at 4096 elements.
25889 2021-01-11 Richard Biener <rguenther@suse.de>
25891 PR tree-optimization/98526
25892 * tree-vect-loop.c (vect_model_reduction_cost): Remove costing
25893 of the actual reduction op for the regular case.
25894 (vectorizable_reduction): Cost the stmts
25895 vect_transform_reduction produces here.
25897 2021-01-11 Andreas Krebbel <krebbel@linux.ibm.com>
25899 * tree-ssa-forwprop.c (simplify_vector_constructor): For
25900 big-endian, use UNPACK[_FLOAT]_HI.
25902 2021-01-11 Tamar Christina <tamar.christina@arm.com>
25904 * tree-vect-slp-patterns.c (class complex_pattern,
25905 class complex_add_pattern): Add parameters to matches.
25906 (complex_add_pattern::build): Free memory.
25907 (complex_add_pattern::matches): Move validation end of match.
25908 (complex_add_pattern::recognize): Likewise.
25910 2021-01-11 Tamar Christina <tamar.christina@arm.com>
25912 * tree-vect-slp-patterns.c (linear_loads_p): Fix externals.
25914 2021-01-11 Tamar Christina <tamar.christina@arm.com>
25916 * tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
25918 2021-01-11 Jakub Jelinek <jakub@redhat.com>
25920 PR tree-optimization/95867
25921 * tree-ssa-math-opts.h: New header.
25922 * tree-ssa-math-opts.c: Include tree-ssa-math-opts.h.
25923 (powi_as_mults): No longer static. Use build_one_cst instead of
25924 build_real. Formatting fix.
25925 * tree-ssa-reassoc.c: Include tree-ssa-math-opts.h.
25926 (attempt_builtin_powi): Handle multiplication reassociation without
25927 powi_fndecl using powi_as_mults.
25928 (reassociate_bb): For integral types don't require
25929 -funsafe-math-optimizations to call attempt_builtin_powi.
25931 2021-01-11 Jakub Jelinek <jakub@redhat.com>
25933 PR tree-optimization/95852
25934 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): Change
25935 mul_stmts parameter type to vec<gimple *> &. Before cond_stmt
25936 allow in the bb any of the stmts in that vector, div_stmt and
25937 up to 3 cast stmts.
25938 (arith_cast_equal_p): New function.
25939 (arith_overflow_check_p): Add cast_stmt argument, handle signed
25940 multiply overflow checks.
25941 (match_arith_overflow): Adjust caller. Handle signed multiply
25944 2021-01-11 Jakub Jelinek <jakub@redhat.com>
25946 PR tree-optimization/95852
25947 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): New function.
25948 (uaddsub_overflow_check_p): Renamed to ...
25949 (arith_overflow_check_p): ... this. Handle also multiplication
25950 with overflow check.
25951 (match_uaddsub_overflow): Renamed to ...
25952 (match_arith_overflow): ... this. Add cfg_changed argument. Handle
25953 also multiplication with overflow check. Adjust function comment.
25954 (math_opts_dom_walker::after_dom_children): Adjust callers. Call
25955 match_arith_overflow also for MULT_EXPR.
25957 2021-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25959 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
25960 __builtin_convertvector.
25961 (vmovl_s16): Likewise.
25962 (vmovl_s32): Likewise.
25963 (vmovl_u8): Likewise.
25964 (vmovl_u16): Likewise.
25965 (vmovl_u32): Likewise.
25966 (vmovn_s16): Likewise.
25967 (vmovn_s32): Likewise.
25968 (vmovn_s64): Likewise.
25969 (vmovn_u16): Likewise.
25970 (vmovn_u32): Likewise.
25971 (vmovn_u64): Likewise.
25973 2021-01-11 Martin Liska <mliska@suse.cz>
25975 * gimple-if-to-switch.cc (struct condition_info): Use auto_var.
25976 (if_chain::is_beneficial): Delete clusters
25977 (find_conditions): Make second argument of conditions_in_bbs a
25978 pointer so that we control over it's lifetime.
25979 (pass_if_to_switch::execute): Delete them.
25981 2021-01-11 Kewen Lin <linkw@linux.ibm.com>
25983 * ira.c (move_unallocated_pseudos): Check other_reg and skip if
25986 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
25988 * config/vax/vax.md (cc): Remove mode attribute.
25989 (subst_<cc>, subst_f<cc>): Rename to...
25990 (subst_<mode>, subst_f<VAXccnz:mode>): ... these respectively.
25991 (*cbranch<VAXint:mode>4_<VAXcc:mode>): Update for `cc' removal.
25992 (*cbranch<VAXfp:mode>4_<VAXccnz:mode>): Likewise.
25993 (*branch_<mode>, *branch_<mode>_reversed): Likewise.
25995 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
25997 * config/vax/vax.md (subst_f<cc>): Add mode to operands and
25998 `const_double_zero'.
26000 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
26002 * config/pdp11/pdp11.md (PDPfp): New mode iterator.
26003 (fcc_cc, fcc_ccnz): Use it. Add mode to `const_double_zero' and
26006 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
26008 * genemit.c (gen_exp) <CONST_DOUBLE>: Handle `const_double_zero'
26010 * read-rtl.c (rtx_reader::read_rtx_code): Handle machine mode
26011 with `const_double_zero'.
26012 * doc/rtl.texi (Constant Expression Types): Document it.
26014 2021-01-09 Jakub Jelinek <jakub@redhat.com>
26017 * tree-cfg.c (verify_gimple_assign_binary): Allow lhs of
26018 POINTER_DIFF_EXPR to be any integral type.
26020 2021-01-09 Jakub Jelinek <jakub@redhat.com>
26022 PR rtl-optimization/98603
26023 * function.c (instantiate_virtual_regs_in_insn): For asm goto
26024 with impossible constraints, drop all SETs, CLOBBERs, drop PARALLEL
26025 if any, set ASM_OPERANDS mode to VOIDmode and change
26026 ASM_OPERANDS_OUTPUT_CONSTRAINT and ASM_OPERANDS_OUTPUT_IDX.
26028 2021-01-09 Alexandre Oliva <oliva@gnu.org>
26031 * final.c (notice_source_line): Narrow down the condition to
26032 skip a line-0 marker.
26034 2021-01-08 Sergei Trofimovich <siarheit@google.com>
26036 * ipa-modref.c (merge_call_side_effects): Fix
26037 linebreak split by reordering two print calls.
26039 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
26041 * config/s390/vector.md (*tf_to_fprx2_0): Rename from
26042 "*mov_tf_to_fprx2_0" for consistency, fix constraint.
26043 (*tf_to_fprx2_1): Rename from "*mov_tf_to_fprx2_1" for
26044 consistency, fix constraint.
26046 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
26048 * config/s390/s390-c.c (s390_def_or_undef_macro): Accept
26049 callables instead of mask values.
26050 (struct target_flag_set_p): New predicate.
26051 (s390_cpu_cpp_builtins_internal): Define or undefine
26052 __LONG_DOUBLE_VX__ macro.
26054 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
26057 * config/i386/i386.c (x86_function_profiler): Use R10 and R11
26058 to call mcount in large model with PIC for NO_PROFILE_COUNTERS
26061 2021-01-08 Richard Biener <rguenther@suse.de>
26063 * tree-ssa-sccvn.c (pass_fre::execute): Reset the SCEV hash table.
26065 2021-01-08 Richard Biener <rguenther@suse.de>
26067 * tree-vect-slp.c (scalar_stmts_to_slp_tree_map_t): Fix.
26068 (vect_build_slp_tree): On cache hit release the matched
26069 scalar stmts vector.
26070 * tree-vect-stmts.c (vectorizable_store): Properly free
26071 vec_oprnds before possibly gathering them again.
26073 2021-01-08 Richard Biener <rguenther@suse.de>
26075 PR tree-optimization/98544
26076 * tree-vect-slp.c (vect_optimize_slp): Always materialize
26077 permutes at a permute node.
26079 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
26082 * config/i386/i386.c (x86_function_profiler): Use R10 to call
26083 mcount in large model. Sorry for large model with PIC.
26085 2021-01-08 Jakub Jelinek <jakub@redhat.com>
26088 * config/i386/i386.opt (ix86_cmodel, ix86_incoming_stack_boundary_arg,
26089 ix86_pmode, ix86_preferred_stack_boundary_arg, ix86_regparm,
26090 ix86_veclibabi_type): Remove x_ prefix, use TargetVariable instead of
26091 TargetSave and initialize for variables with enum types.
26092 (mfentry, mstack-protector-guard-reg=, mstack-protector-guard-offset=,
26093 mstack-protector-guard-symbol=): Add Save.
26094 * config/i386/i386-options.c (ix86_function_specific_save,
26095 ix86_function_specific_restore): Don't save or restore x_ix86_cmodel,
26096 x_ix86_incoming_stack_boundary_arg, x_ix86_pmode,
26097 x_ix86_preferred_stack_boundary_arg, x_ix86_regparm,
26098 x_ix86_veclibabi_type.
26100 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
26102 * config/aarch64/aarch64-sve.md (*cnot<mode>): Extend from
26103 SVE_FULL_I to SVE_I.
26104 (*cond_cnot<mode>_2, *cond_cnot<mode>_any): Likewise.
26106 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
26108 * config/aarch64/aarch64-sve.md (*cond_uxt<mode>_2): Extend from
26109 SVE_FULL_I to SVE_I.
26110 (*cond_uxt<mode>_any): Likewise.
26112 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26114 * config/aarch64/iterators.md (Vwhalf): New iterator.
26115 * config/aarch64/aarch64-simd.md (aarch64_<sur>adalp<mode>_3):
26117 (aarch64_<sur>adalp<mode>): ... This. Make more
26119 (<sur>sadv16qi): Adjust callsite of the above.
26120 * config/aarch64/aarch64-simd-builtins.def (sadalp, uadalp): New
26122 * config/aarch64/arm_neon.h (vpadal_s8): Reimplement using
26124 (vpadal_s16): Likewise.
26125 (vpadal_u8): Likewise.
26126 (vpadal_u16): Likewise.
26127 (vpadalq_s8): Likewise.
26128 (vpadalq_s16): Likewise.
26129 (vpadalq_s32): Likewise.
26130 (vpadalq_u8): Likewise.
26131 (vpadalq_u16): Likewise.
26132 (vpadalq_u32): Likewise.
26134 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26136 * config/aarch64/aarch64-simd.md (aarch64_<su>abd<mode>_3):
26138 (aarch64_<su>abd<mode>): ... This.
26139 (<sur>sadv16qi): Adjust callsite of the above.
26140 * config/aarch64/aarch64-simd-builtins.def (sabd, uabd): Define
26142 * config/aarch64/arm_neon.h (vabd_s8): Reimplement using
26144 (vabd_s16): Likewise.
26145 (vabd_s32): Likewise.
26146 (vabd_u8): Likewise.
26147 (vabd_u16): Likewise.
26148 (vabd_u32): Likewise.
26149 (vabdq_s8): Likewise.
26150 (vabdq_s16): Likewise.
26151 (vabdq_s32): Likewise.
26152 (vabdq_u8): Likewise.
26153 (vabdq_u16): Likewise.
26154 (vabdq_u32): Likewise.
26156 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26158 * config/aarch64/aarch64-simd-builtins.def (saba, uaba): Define
26160 * config/aarch64/arm_neon.h (vaba_s8): Implement using builtin.
26161 (vaba_s16): Likewise.
26162 (vaba_s32): Likewise.
26163 (vaba_u8): Likewise.
26164 (vaba_u16): Likewise.
26165 (vaba_u32): Likewise.
26166 (vabaq_s8): Likewise.
26167 (vabaq_s16): Likewise.
26168 (vabaq_s32): Likewise.
26169 (vabaq_u8): Likewise.
26170 (vabaq_u16): Likewise.
26171 (vabaq_u32): Likewise.
26173 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26175 * config/aarch64/aarch64-simd.md (aba<mode>_3): Rename to...
26176 (aarch64_<su>aba<mode>): ... This. Handle uaba as well.
26177 Change RTL pattern to match.
26179 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
26181 * common/config/riscv/riscv-common.c (riscv_current_subset_list): New.
26182 * config/riscv/riscv-c.c (riscv-subset.h): New.
26183 (INCLUDE_STRING): Define.
26184 (riscv_cpu_cpp_builtins): Add new style architecture extension
26186 * config/riscv/riscv-subset.h (riscv_subset_list::begin): New.
26187 (riscv_subset_list::end): New.
26188 (riscv_current_subset_list): New.
26190 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
26192 * common/config/riscv/riscv-common.c (RISCV_DONT_CARE_VERSION):
26193 Move to riscv-subset.h.
26194 (struct riscv_subset_t): Ditto.
26195 (class riscv_subset_list): Ditto.
26196 * config/riscv/riscv-subset.h (RISCV_DONT_CARE_VERSION): Move
26197 from riscv-common.c.
26198 (struct riscv_subset_t): Ditto.
26199 (class riscv_subset_list): Ditto.
26200 * config/riscv/t-riscv ($(common_out_file)): Add file
26203 2021-01-07 Jakub Jelinek <jakub@redhat.com>
26206 * config/i386/i386.md (*bmi_blsi_<mode>_cmp, *bmi_blsi_<mode>_ccno):
26207 New define_insn patterns.
26209 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
26211 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_UNARY:optab><mode>)
26212 (*cond_<SVE_INT_UNARY:optab><mode>_2): Extend from SVE_FULL_I to SVE_I.
26213 (*cond_<SVE_INT_UNARY:optab><mode>_any): Likewise.
26215 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
26217 PR tree-optimization/98560
26218 * internal-fn.def (IFN_VCONDU, IFN_VCONDEQ): Use type vec_cond.
26219 * internal-fn.c (vec_cond_mask_direct): Get the data mode from
26221 (vec_cond_direct): Likewise argument 2.
26222 (vec_condu_direct, vec_condeq_direct): Delete.
26223 (expand_vect_cond_optab_fn): Rename to...
26224 (expand_vec_cond_optab_fn): ...this, replacing old macro.
26225 (expand_vec_condu_optab_fn, expand_vec_condeq_optab_fn): Delete.
26226 (expand_vect_cond_mask_optab_fn): Rename to...
26227 (expand_vec_cond_mask_optab_fn): ...this, replacing old macro.
26228 (direct_vec_cond_mask_optab_supported_p): Treat the optab as a
26230 (direct_vec_cond_optab_supported_p): Likewise.
26231 (direct_vec_condu_optab_supported_p): Delete.
26232 (direct_vec_condeq_optab_supported_p): Delete.
26233 * gimple-isel.cc: Include internal-fn.h.
26234 (gimple_expand_vec_cond_expr): Check that IFN_VCONDEQ is supported
26237 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
26239 PR tree-optimization/98560
26240 * gimple-isel.cc (gimple_expand_vec_cond_expr): If we fail to use
26241 IFN_VCOND{,U,EQ}, fall back on IFN_VCOND_MASK.
26243 2021-01-07 Uroš Bizjak <ubizjak@gmail.com>
26245 * config/i386/i386.md (insn): Merge from plusminus_insn, shift_insn,
26246 rotate_insn and optab code attributes.
26247 Update all uses to merged code attribute.
26248 * config/i386/sse.md: Update all uses to merged code attribute.
26249 * config/i386/mmx.md: Update all uses to merged code attribute.
26251 2021-01-07 Jakub Jelinek <jakub@redhat.com>
26253 PR tree-optimization/98568
26254 * gimple-ssa-store-merging.c (bswap_view_convert): New function.
26255 (bswap_replace): Use it.
26257 2021-01-06 Vladimir N. Makarov <vmakarov@redhat.com>
26259 PR rtl-optimization/97978
26260 * lra-int.h (lra_hard_reg_split_p): New external.
26261 * lra.c (lra_hard_reg_split_p): New global.
26262 (lra): Set up lra_hard_reg_split_p after splitting a hard reg.
26263 * lra-assigns.c (lra_assign): Don't check allocation correctness
26264 after hard reg splitting.
26266 2021-01-06 Martin Sebor <msebor@redhat.com>
26269 * builtins.c (new_delete_mismatch_p): New overload.
26270 (new_delete_mismatch_p (tree, tree)): Call it.
26272 2021-01-06 Alexandre Oliva <oliva@adacore.com>
26274 * Makefile.in (T_GLIMITS_H): New.
26275 (stmp-int-hdrs): Depend on it, use it.
26276 * config/t-vxworks (T_GLIMITS_H): Override it.
26277 (vxw-glimits.h): New.
26279 2021-01-06 Richard Biener <rguenther@suse.de>
26281 PR tree-optimization/98513
26282 * value-range.cc (intersect_ranges): Compare the upper bounds
26283 for the expected relation.
26285 2021-01-06 Gerald Pfeifer <gerald@pfeifer.com>
26288 2020-12-28 Gerald Pfeifer <gerald@pfeifer.com>
26290 * doc/standards.texi (HSAIL): Remove section.
26292 2021-01-05 Samuel Thibault <samuel.thibault@ens-lyon.org>
26294 * configure: Re-generate.
26296 2021-01-05 Jakub Jelinek <jakub@redhat.com>
26298 * doc/invoke.texi (-std=c++20): Adjust for the publication of
26299 ISO 14882:2020 standard.
26300 * doc/standards.texi: Likewise.
26302 2021-01-05 Jakub Jelinek <jakub@redhat.com>
26304 PR tree-optimization/94802
26305 * expr.h (maybe_optimize_sub_cmp_0): Declare.
26306 * expr.c: Include tree-pretty-print.h and flags.h.
26307 (maybe_optimize_sub_cmp_0): New function.
26308 (do_store_flag): Use it.
26309 * cfgexpand.c (expand_gimple_cond): Likewise.
26311 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
26313 * mux-utils.h (pointer_mux::m_ptr): Tweak description of contents.
26314 * rtlanal.c (simple_regno_set): Tweak description to clarify the
26317 2021-01-05 Richard Biener <rguenther@suse.de>
26319 PR tree-optimization/98516
26320 * tree-vect-slp.c (vect_optimize_slp): Permute the incoming
26321 lanes when materializing on a VEC_PERM node.
26322 (vectorizable_slp_permutation): Dump the permute properly.
26324 2021-01-05 Richard Biener <rguenther@suse.de>
26326 * tree-vect-slp.c (vect_slp_region): Move debug counter
26327 to cover individual subgraphs.
26329 2021-01-05 Richard Biener <rguenther@suse.de>
26331 PR tree-optimization/98428
26332 * tree-vect-slp.c (vect_build_slp_tree_1): Properly reject
26333 vector lane extracts for loop vectorization.
26335 2021-01-05 Jakub Jelinek <jakub@redhat.com>
26337 PR tree-optimization/98514
26338 * tree-ssa-reassoc.c (bb_rank): Change type from long * to
26340 (operand_rank): Change type from hash_map<tree, long> to
26341 hash_map<tree, int64_t>.
26342 (phi_rank): Change return type from long to int64_t.
26343 (loop_carried_phi): Change block_rank variable type from long to
26345 (propagate_rank): Change return type, rank parameter type and
26346 op_rank variable type from long to int64_t.
26347 (find_operand_rank): Change return type from long to int64_t
26348 and change slot variable type from long * to int64_t *.
26349 (insert_operand_rank): Change rank parameter type from long to
26351 (get_rank): Change return type and rank variable type from long to
26352 int64_t. Use PRId64 instead of ld to print the rank.
26353 (init_reassoc): Change rank variable type from long to int64_t
26354 and adjust correspondingly bb_rank and operand_rank initialization.
26356 2021-01-05 Jakub Jelinek <jakub@redhat.com>
26358 PR tree-optimization/96928
26359 * tree-ssa-phiopt.c (xor_replacement): New function.
26360 (tree_ssa_phiopt_worker): Call it.
26362 2021-01-05 Jakub Jelinek <jakub@redhat.com>
26364 PR tree-optimization/96930
26365 * match.pd ((A / (1 << B)) -> (A >> B)): If A is extended
26366 from narrower value which has the same type as 1 << B, perform
26367 the right shift on the narrower value followed by extension.
26369 2021-01-05 Jakub Jelinek <jakub@redhat.com>
26371 PR tree-optimization/96239
26372 * gimple-ssa-store-merging.c (maybe_optimize_vector_constructor): New
26374 (get_status_for_store_merging): Don't return BB_INVALID for blocks
26375 with potential bswap optimizable CONSTRUCTORs.
26376 (pass_store_merging::execute): Optimize vector CONSTRUCTORs with bswap
26379 2021-01-05 Richard Biener <rguenther@suse.de>
26381 PR tree-optimization/98381
26382 * tree.c (vector_element_bits): Properly compute bool vector
26384 * tree-vect-loop.c (vectorizable_live_operation): Properly
26385 compute the last lane bit offset.
26387 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
26390 * config/i386/sse.md (sse_cvtps2pi): Redefine as define_insn_and_split.
26391 Clear the top 64 bytes of the input XMM register.
26392 (sse_cvttps2pi): Ditto.
26394 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
26397 * config/i386/xopintrin.h (_mm256_cmov_si256): New.
26399 2021-01-05 H.J. Lu <hjl.tools@gmail.com>
26402 * config/i386/xmmintrin.h (_mm_extract_pi16): Cast to unsigned
26405 2021-01-05 Claudiu Zissulescu <claziss@synopsys.com>
26407 * config/arc/arc.md (maddsidi4_split): Use ACC_REG_FIRST.
26408 (umaddsidi4_split): Likewise.
26410 2021-01-05 liuhongt <hongtao.liu@intel.com>
26413 * config/i386/sse.md (*sse2_pmovskb_zexthisi): New
26414 define_insn_and_split for zero_extend of subreg HI of pmovskb
26416 (*sse2_pmovskb_zexthisi): Add new combine splitters for
26417 zero_extend of not of subreg HI of pmovskb result.
26419 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
26422 * explow.c (convert_memory_address_addr_space_1): Handle UNSPECs
26424 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate): Use
26425 convert_memory_address to convert symbolic immediates to ptr_mode
26426 before forcing them to memory.
26428 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
26430 PR rtl-optimization/97144
26431 * recog.c (constrain_operands): Initialize matching_operand
26432 for each alternative, rather than only doing it once.
26434 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
26436 PR rtl-optimization/98403
26437 * rtl-ssa/changes.cc (function_info::finalize_new_accesses): Explain
26438 why we don't remove call clobbers.
26439 (function_info::apply_changes_to_insn): Don't attempt to add
26440 call clobbers here.
26442 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
26444 PR tree-optimization/98371
26445 * tree-vect-loop.c (vect_reanalyze_as_main_loop): New function.
26446 (vect_analyze_loop): If an epilogue loop appears to be cheaper
26447 than the main loop, re-analyze it as a main loop before adopting
26450 2021-01-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
26453 * configure.ac (NETLIBS): Determine using AX_LIB_SOCKET_NSL.
26454 * aclocal.m4, configure: Regenerate.
26455 * Makefile.in (NETLIBS): Define.
26456 (BACKEND): Remove $(CODYLIB).
26458 2021-01-05 Jakub Jelinek <jakub@redhat.com>
26460 PR rtl-optimization/98334
26461 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
26462 Optimize (X - 1) * Y + Y to X * Y or (X + 1) * Y - Y to X * Y.
26464 2021-01-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
26466 * tree-inline.c (expand_call_inline): Restore input_location.
26467 Return result from recursive call.
26469 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
26471 PR tree-optimization/95401
26472 * config/aarch64/aarch64-sve-builtins.cc
26473 (gimple_folder::load_store_cookie): Use bits rather than bytes
26474 for the alignment argument to IFN_MASK_LOAD and IFN_MASK_STORE.
26475 * gimple-fold.c (gimple_fold_mask_load_store_mem_ref): Likewise.
26476 * tree-vect-stmts.c (vectorizable_store): Likewise.
26477 (vectorizable_load): Likewise.
26479 2021-01-04 Richard Biener <rguenther@suse.de>
26481 PR tree-optimization/98308
26482 * tree-vect-stmts.c (vectorizable_load): Set invariant mask
26485 2021-01-04 Jakub Jelinek <jakub@redhat.com>
26487 PR tree-optimization/95771
26488 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Handle types
26489 with precision smaller than int's precision and types with precision
26490 twice as large as long long. Formatting fixes.
26492 2021-01-04 Richard Biener <rguenther@suse.de>
26494 PR tree-optimization/98464
26495 * tree-ssa-sccvn.c (vn_valueize_for_srt): Rename from ...
26496 (vn_valueize_wrapper): ... this. Temporarily adjust vn_context_bb.
26497 (process_bb): Adjust.
26499 2021-01-04 Matthew Malcomson <matthew.malcomson@arm.com>
26502 * doc/invoke.texi (-fsanitize=address): Fix wording describing
26503 clash with -fsanitize=hwaddress.
26505 2021-01-04 Richard Biener <rguenther@suse.de>
26507 PR tree-optimization/98282
26508 * tree-ssa-sccvn.c (vn_get_stmt_kind): Classify tcc_reference on
26509 invariants as VN_NARY.
26511 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
26514 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Accept
26515 aarch64_simd_reg_or_zero for operand 2. Use the combinez patterns
26516 to handle zero operands.
26518 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
26520 * config/aarch64/aarch64.c (offset_6bit_signed_scaled_p): New function.
26521 (offset_6bit_unsigned_scaled_p): Fix typo in comment.
26522 (aarch64_sve_prefetch_operand_p): Accept MUL VLs in the range
26525 2021-01-04 Richard Biener <rguenther@suse.de>
26527 PR tree-optimization/98393
26528 * tree-vect-slp.c (vect_build_slp_tree): Properly zero matches
26529 when hitting the limit.
26531 2021-01-04 Richard Biener <rguenther@suse.de>
26533 PR tree-optimization/98291
26534 * tree-vect-loop.c (vectorizable_reduction): Bypass
26535 associativity check for SLP reductions with VF 1.
26537 2021-01-04 Jakub Jelinek <jakub@redhat.com>
26539 PR tree-optimization/96782
26540 * match.pd (x == ~x -> false, x != ~x -> true): New simplifications.
26542 2021-01-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
26544 * collect-utils.c (collect_execute): Check dumppfx.
26545 * collect2.c (maybe_run_lto_and_relink, do_link): Pass atsuffix
26546 to collect_execute.
26547 (do_link): Add new parameter atsuffix.
26548 (main): Handle -dumpdir option. Skip one argument for
26549 -o, -isystem and -B options.
26550 * gcc.c (make_at_file): New helper function.
26551 (close_at_file): Use it.
26553 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26555 * config/darwin.h (MIN_LD64_NO_COAL_SECTS): Adjust.
26556 Amend handling for LD64_VERSION fallback defaults.
26558 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26560 * config.gcc: Compute default version information
26561 from the configured target. Likewise defaults for
26563 * config/darwin10.h: Removed.
26564 * config/darwin12.h: Removed.
26565 * config/darwin9.h: Removed.
26566 * config/rs6000/darwin8.h: Removed.
26568 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26570 * config/darwin9.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
26572 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26574 * config/darwin9.h (STACK_CHECK_STATIC_BUILTIN): Move from here..
26575 * config/darwin.h (STACK_CHECK_STATIC_BUILTIN): .. to here.
26577 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26579 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move from
26581 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): ... to here.
26583 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26585 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move the spec
26586 for the Darwin10 unwinder stub from here ...
26587 * config/darwin.h (LINK_COMMAND_SPEC_A): ... to here.
26589 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26591 * config/darwin.h (DSYMUTIL_SPEC): Default to DWARF
26592 (ASM_DEBUG_SPEC):Only define if the assembler supports
26594 (PREFERRED_DEBUGGING_TYPE): Default to DWARF.
26595 (DARWIN_PREFER_DWARF): Define.
26596 * config/darwin9.h (PREFERRED_DEBUGGING_TYPE): Remove.
26597 (DARWIN_PREFER_DWARF): Likewise
26598 (DSYMUTIL_SPEC): Likewise.
26599 (COLLECT_RUN_DSYMUTIL): Likewise.
26600 (ASM_DEBUG_SPEC): Likewise.
26601 (ASM_DEBUG_OPTION_SPEC): Likewise.
26603 2021-01-02 Jan Hubicka <jh@suse.cz>
26605 * cfg.c (free_block): ggc_free bb.
26607 2021-01-01 Jakub Jelinek <jakub@redhat.com>
26609 * gcc.c (process_command): Update copyright notice dates.
26610 * gcov-dump.c (print_version): Ditto.
26611 * gcov.c (print_version): Ditto.
26612 * gcov-tool.c (print_version): Ditto.
26613 * gengtype.c (create_file): Ditto.
26614 * doc/cpp.texi: Bump @copying's copyright year.
26615 * doc/cppinternals.texi: Ditto.
26616 * doc/gcc.texi: Ditto.
26617 * doc/gccint.texi: Ditto.
26618 * doc/gcov.texi: Ditto.
26619 * doc/install.texi: Ditto.
26620 * doc/invoke.texi: Ditto.
26622 2021-01-01 Jakub Jelinek <jakub@redhat.com>
26624 * ChangeLog-2020: Rotate ChangeLog. New file.
26627 Copyright (C) 2021 Free Software Foundation, Inc.
26629 Copying and distribution of this file, with or without modification,
26630 are permitted in any medium without royalty provided the copyright
26631 notice and this notice are preserved.