1 2021-10-19 Jonathan Wakely <jwakely@redhat.com>
3 * doc/extend.texi (Basic PowerPC Built-in Functions): Fix typo.
5 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
9 * config/rs6000/emmintrin.h: Guard POWER8 intrinsics.
10 * config/rs6000/pmmintrin.h: Same.
11 * config/rs6000/smmintrin.h: Same.
12 * config/rs6000/tmmintrin.h: Same.
14 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
16 * config.gcc (extra_headers): Add nmmintrin.h.
18 2021-10-19 Richard Biener <rguenther@suse.de>
20 * tree-vectorizer.h (vect_supportable_dr_alignment): Add
21 misalignment parameter.
22 * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
23 Do not change DR_MISALIGNMENT in place, instead pass the
24 adjusted misalignment to vect_supportable_dr_alignment.
25 (vect_peeling_supportable): Likewise.
26 (vect_peeling_hash_get_lowest_cost): Adjust.
27 (vect_enhance_data_refs_alignment): Likewise.
28 (vect_vfa_access_size): Likewise.
29 (vect_supportable_dr_alignment): Add misalignment
30 parameter and simplify.
31 * tree-vect-stmts.c (get_negative_load_store_type): Adjust.
32 (get_group_load_store_type): Likewise.
33 (get_load_store_type): Likewise.
35 2021-10-19 Clément Chigot <clement.chigot@atos.net>
37 * config/rs6000/rs6000.c (rs6000_xcoff_file_end): Move
38 __tls_get_addr reference to .text csect.
40 2021-10-19 Martin Liska <mliska@suse.cz>
43 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
46 2021-10-19 Richard Biener <rguenther@suse.de>
48 * tree-vectorizer.h (vect_get_store_cost): Adjust signature.
49 (vect_get_load_cost): Likewise.
50 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
51 alignment support scheme and misalignment as arguments
53 (vect_get_peeling_costs_all_drs): Compute that info here
54 and note that we shouldn't need to.
55 * tree-vect-stmts.c (vect_model_store_cost): Get
56 alignment support scheme and misalignment as arguments.
57 (vect_get_store_cost): Likewise.
58 (vect_model_load_cost): Likewise.
59 (vect_get_load_cost): Likewise.
60 (vectorizable_store): Pass down alignment support scheme
61 and misalignment to costing.
62 (vectorizable_load): Likewise.
64 2021-10-19 Richard Biener <rguenther@suse.de>
66 * tree-vect-stmts.c (get_negative_load_store_type): Add
67 offset output parameter and initialize it.
68 (get_group_load_store_type): Likewise.
69 (get_load_store_type): Likewise.
70 (vectorizable_store): Use offset as computed by
72 (vectorizable_load): Likewise.
74 2021-10-19 Richard Biener <rguenther@suse.de>
76 PR tree-optimization/102827
77 * tree-if-conv.c (predicate_statements): Add pe parameter
78 and use that edge to insert invariant stmts on.
79 (combine_blocks): Pass through pe.
80 (tree_if_conversion): Compute the edge to insert invariant
81 stmts on and pass it along.
83 2021-10-19 Roger Sayle <roger@nextmovesoftware.com>
86 * config/bfin/bfin.md (addsubv2hi3, subaddv2hi3, ssaddsubv2hi3,
87 sssubaddv2hi3): Swap the order of operators in vec_concat.
89 2021-10-19 Xionghu Luo <luoxhu@linux.ibm.com>
91 * config/rs6000/altivec.md (*altivec_vmrghb_internal): Delete.
92 (altivec_vmrghb_direct): New.
93 (*altivec_vmrghh_internal): Delete.
94 (altivec_vmrghh_direct): New.
95 (*altivec_vmrghw_internal): Delete.
96 (altivec_vmrghw_direct_<mode>): New.
97 (altivec_vmrghw_direct): Delete.
98 (*altivec_vmrglb_internal): Delete.
99 (altivec_vmrglb_direct): New.
100 (*altivec_vmrglh_internal): Delete.
101 (altivec_vmrglh_direct): New.
102 (*altivec_vmrglw_internal): Delete.
103 (altivec_vmrglw_direct_<mode>): New.
104 (altivec_vmrglw_direct): Delete.
105 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
106 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):
108 * config/rs6000/vsx.md (vsx_xxmrghw_<mode>): Adjust.
109 (vsx_xxmrglw_<mode>): Adjust.
111 2021-10-19 Aldy Hernandez <aldyh@redhat.com>
113 * passes.def: Change threading comment before pass_ccp pass.
115 2021-10-19 Haochen Gui <guihaoc@gcc.gnu.org>
117 * config/rs6000/rs6000-call.c (altivec_expand_lxvr_builtin):
118 Modify the expansion for sign extension. All extensions are done
119 within VSX registers.
121 2021-10-19 Richard Biener <rguenther@suse.de>
123 * tree-vect-stmts.c (get_group_load_store_type): Add
124 misalignment output parameter and initialize it.
125 (get_group_load_store_type): Likewise.
126 (vectorizable_store): Remove now redundant queries.
127 (vectorizable_load): Likewise.
129 2021-10-19 Richard Biener <rguenther@suse.de>
131 * tree-vectorizer.h (vect_supportable_dr_alignment): Remove
132 check_aligned argument.
133 * tree-vect-data-refs.c (vect_supportable_dr_alignment):
135 (vect_peeling_hash_insert): Add supportable_if_not_aligned
136 argument and do not call vect_supportable_dr_alignment here.
137 (vect_peeling_supportable): Adjust.
138 (vect_enhance_data_refs_alignment): Compute whether the
139 access is supported with different alignment here and
140 pass that down to vect_peeling_hash_insert.
141 (vect_vfa_access_size): Adjust.
142 * tree-vect-stmts.c (vect_get_store_cost): Likewise.
143 (vect_get_load_cost): Likewise.
144 (get_negative_load_store_type): Likewise.
145 (get_group_load_store_type): Likewise.
146 (get_load_store_type): Likewise.
148 2021-10-19 Martin Liska <mliska@suse.cz>
151 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
152 * system.h (strip_whilespaces): New function.
154 2021-10-19 dianhong xu <dianhong.xu@intel.com>
156 * config/i386/avx512fp16intrin.h:
157 (_mm512_set1_pch): New intrinsic.
158 * config/i386/avx512fp16vlintrin.h:
159 (_mm256_set1_pch): New intrinsic.
160 (_mm_set1_pch): Ditto.
162 2021-10-18 Andrew MacLeod <amacleod@redhat.com>
164 PR tree-optimization/102796
165 * gimple-range.cc (gimple_ranger::range_on_edge): Process EH edges
166 normally. Return get_tree_range for non gimple_range_ssa_p names.
167 (gimple_ranger::range_of_stmt): Use get_tree_range for non
168 gimple_range_ssa_p names.
170 2021-10-18 Uroš Bizjak <ubizjak@gmail.com>
173 * config/i386/i386.c (ix86_print_operand_address):
174 Error out for non-address_operand asm operands.
176 2021-10-18 Richard Biener <rguenther@suse.de>
178 * tree-vect-data-refs.c (vect_peeling_hash_insert): Do
179 not auto-convert dr_alignment_support to bool.
180 (vect_peeling_supportable): Likewise.
181 (vect_enhance_data_refs_alignment): Likewise.
182 (vect_supportable_dr_alignment): Commonize read/write case.
183 * tree-vect-stmts.c (vect_get_store_cost): Use
184 dr_alignment_support, not int, for the vect_supportable_dr_alignment
186 (vect_get_load_cost): Likewise.
188 2021-10-18 Siddhesh Poyarekar <siddhesh@gotplt.org>
190 * tree-object-size.c (object_sizes_execute): Consolidate LHS
191 null check and do it early.
193 2021-10-18 Richard Biener <rguenther@suse.de>
195 * tree-vect-stmts.c (vectorizable_store): Use the
196 computed alignment scheme instead of querying
199 2021-10-18 Richard Biener <rguenther@suse.de>
201 * tree-vect-stmts.c (vectorizable_store): Do not recompute
202 alignment scheme already determined by get_load_store_type.
204 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
206 * tree-ssa-threadbackward.c (class pass_thread_jumps_full):
207 Clone corresponding pass.
209 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
211 * combine.c (recog_for_combine): For an unrecognized move/set of
212 a constant, try force_const_mem to place it in the constant pool.
213 * cse.c (constant_pool_entries_cost, constant_pool_entries_regcost):
214 Delete global variables (that are no longer assigned a cost value).
215 (cse_insn): Simplify logic for deciding whether to place a folded
216 constant in the constant pool using force_const_mem.
217 (cse_main): Remove zero initialization of constant_pool_entries_cost
218 and constant_pool_entries_regcost.
219 * config/i386/i386.c (ix86_rtx_costs): Make memory accesses
220 fractionally more expensive, when optimizing for speed.
222 2021-10-18 Martin Liska <mliska@suse.cz>
224 PR gcov-profile/102746
225 PR gcov-profile/102747
226 * gcov.c (main): Return return_code.
227 (output_gcov_file): Mark return_code when error happens.
228 (generate_results): Likewise.
229 (read_graph_file): Likewise.
230 (read_count_file): Likewise.
232 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
234 * config/bfin/bfin.md (define_constants): Remove UNSPEC_ONES.
235 (define_insn "ones"): Replace UNSPEC_ONES with a truncate of
236 a popcount, allowing compile-time evaluation/simplification.
237 (popcountsi2, popcounthi2): New expanders using a "ones" insn.
239 2021-10-18 Richard Biener <rguenther@suse.de>
241 PR tree-optimization/102788
242 * tree-vect-patterns.c (vect_init_pattern_stmt): Allow
244 (vect_pattern_recog_1): Likewise.
245 (vect_recog_bool_pattern): Continue matching the pattern
246 even if we do not have a vector type for a conversion
249 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
251 * simplify-rtx.c (simplify_const_unary_operation) [SS_NEG, SS_ABS]:
252 Evalute SS_NEG and SS_ABS of a constant argument.
254 2021-10-18 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
257 * gimple-match-head.c (try_conditional_simplification): Add case for single operand.
258 * internal-fn.def: Add entry for COND_NEG internal function.
259 * internal-fn.c (FOR_EACH_CODE_MAPPING): Add entry for
260 NEGATE_EXPR, COND_NEG mapping.
261 * optabs.def: Add entry for cond_neg_optab.
262 * match.pd (UNCOND_UNARY, COND_UNARY): New operator lists.
263 (vec_cond COND (foo A) B) -> (IFN_COND_FOO COND A B): New pattern.
264 (vec_cond COND B (foo A)) -> (IFN_COND_FOO ~COND A B): Likewise.
266 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
268 * tree-ssa-strlen.c (class strlen_pass): Rename from
270 (handle_builtin_stxncpy_strncat): Move to strlen_pass.
271 (handle_assign): Same.
272 (adjust_last_stmt): Same.
273 (maybe_warn_overflow): Same.
274 (handle_builtin_strlen): Same.
275 (handle_builtin_strchr): Same.
276 (handle_builtin_strcpy): Same.
277 (handle_builtin_strncat): Same.
278 (handle_builtin_stxncpy_strncat): Same.
279 (handle_builtin_memcpy): Same.
280 (handle_builtin_strcat): Same.
281 (handle_alloc_call): Same.
282 (handle_builtin_memset): Same.
283 (handle_builtin_memcmp): Same.
284 (get_len_or_size): Same.
285 (strxcmp_eqz_result): Same.
286 (handle_builtin_string_cmp): Same.
287 (handle_pointer_plus): Same.
288 (count_nonzero_bytes_addr): Same.
289 (count_nonzero_bytes): Same.
290 (handle_store): Same.
291 (strlen_check_and_optimize_call): Same.
292 (handle_integral_assign): Same.
293 (check_and_optimize_stmt): Same.
294 (printf_strlen_execute): Rename strlen_dom_walker to strlen_pass.
296 2021-10-18 Richard Biener <rguenther@suse.de>
298 PR tree-optimization/102798
299 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
300 Only copy points-to info to newly generated SSA names.
302 2021-10-18 Martin Liska <mliska@suse.cz>
304 * dbgcnt.c (dbg_cnt_process_opt): Remove unused but set variable.
305 * gcov.c (get_cycles_count): Likewise.
306 * lto-compress.c (lto_compression_zlib): Likewise.
307 (lto_uncompression_zlib): Likewise.
308 * targhooks.c (default_pch_valid_p): Likewise.
310 2021-10-17 Aldy Hernandez <aldyh@redhat.com>
312 * tree-pass.h (make_pass_thread_jumps_full): New.
313 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Inline.
314 (try_thread_blocks): Add resolve and speed arguments.
315 (pass_thread_jumps::execute): Inline.
316 (do_early_thread_jumps): New.
317 (do_thread_jumps): New.
318 (make_pass_thread_jumps): Move.
319 (pass_early_thread_jumps::gate): Inline.
320 (pass_early_thread_jumps::execute): Inline.
321 (class pass_thread_jumps_full): New.
323 2021-10-16 Piotr Kubaj <pkubaj@FreeBSD.org>
325 * configure.ac: Treat powerpc64*-*-freebsd* the same as
326 powerpc64-*-freebsd*.
327 * configure: Regenerate.
329 2021-10-16 H.J. Lu <hjl.tools@gmail.com>
331 * value-query.cc (get_ssa_name_ptr_info_nonnull): Change
332 set_ptr_nonull to set_ptr_nonnull in comments.
334 2021-10-16 Jan Hubicka <hubicka@ucw.cz>
336 PR tree-optimization/102720
337 * tree-ssa-structalias.c (compute_points_to_sets): Fix producing
338 of call used and clobbered sets.
340 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
342 * config/darwin.h (LINK_COMMAND_SPEC_A): Update 'r' handling to
343 skip gomp and itm when r or nodefaultlibs is given.
344 (DSYMUTIL_SPEC): Do not call dsymutil for '-r' link lines.
345 Update ordering of exclusions, remove duplicate 'v' addition
346 (collect2 will add this from the main command line).
348 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
350 * config/darwin-driver.c (darwin_driver_init): Revise comments, handle
351 filelist and framework options in specs instead of code.
352 * config/darwin.h (SUBTARGET_DRIVER_SELF_SPECS): Update to handle link
353 specs that are really driver ones.
354 (DARWIN_CC1_SPEC): Likewise.
355 (CPP_SPEC): Likewise.
356 (SYSROOT_SPEC): Append space.
357 (LINK_SYSROOT_SPEC): Remove most driver link specs.
358 (STANDARD_STARTFILE_PREFIX_2): Update link-related specs.
359 (STARTFILE_SPEC): Likewise.
360 (ASM_MMACOSX_VERSION_MIN_SPEC): Fix line wrap.
361 (ASM_SPEC): Update driver-related specs.
362 (ASM_FINAL_SPEC): Likewise.
363 * config/darwin.opt: Remove now unused option aliases.
364 * config/i386/darwin.h (EXTRA_ASM_OPTS): Ensure space after opt.
365 (ASM_SPEC): Update driver-related specs.
367 2021-10-15 Roger Sayle <roger@nextmovesoftware.com>
369 * config/i386/i386.c (ix86_hardreg_mov_ok): For vector modes,
370 allow standard_sse_constant_p immediate constants.
372 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
374 * config.gcc: Add tm-dwarf2.h to tm_d-file.
376 2021-10-15 Andrew MacLeod <amacleod@redhat.com>
378 * gimple-range-fold.h (gimple_range_ssa_p): Don't process names
379 that occur in abnormal phis.
380 * gimple-range.cc (gimple_ranger::range_on_edge): Return false for
381 abnormal and EH edges.
382 * gimple-ssa-evrp.c (rvrp_folder::value_of_expr): Ditto.
383 (rvrp_folder::value_on_edge): Ditto.
384 (rvrp_folder::value_of_stmt): Ditto.
385 (hybrid_folder::value_of_expr): Ditto for ranger queries.
386 (hybrid_folder::value_on_edge): Ditto.
387 (hybrid_folder::value_of_stmt): Ditto.
388 * value-query.cc (gimple_range_global): Always return a range if
389 the type is supported.
391 2021-10-15 John David Anglin <danglin@gcc.gnu.org>
393 * config/pa/pa.md: Consistently use "rG" constraint for copy
394 instruction in move patterns.
396 2021-10-15 Andrew Stubbs <ams@codesourcery.com>
398 * config/gcn/gcn-hsa.h (S_FIJI): Set unconditionally.
401 * config/gcn/gcn.c: Hard code SRAM ECC settings for old architectures.
402 * config/gcn/mkoffload.c (ELFABIVERSION_AMDGPU_HSA): Rename to ...
403 (ELFABIVERSION_AMDGPU_HSA_V3): ... this.
404 (ELFABIVERSION_AMDGPU_HSA_V4): New.
405 (SET_SRAM_ECC_UNSUPPORTED): New.
406 (copy_early_debug_info): Create elf flags to match the other objects.
407 (main): Just let the attribute flags pass through.
409 2021-10-15 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
411 * tree-loop-distribution.c (reduction_var_overflows_first):
412 Pass the type of reduction_var as first argument as it is also
413 done for the load type.
414 (loop_distribution::transform_reduction_loop): Add missing
415 TREE_TYPE while determining precission of reduction_var.
417 2021-10-15 Richard Biener <rguenther@suse.de>
419 * defaults.h (PREFERRED_DEBUGGING_TYPE): Choose DWARF2_DEBUG
421 * toplev.c (process_options): Warn when STABS debugging is
422 enabled but not the preferred format.
423 * config/pa/som.h (PREFERRED_DEBUGGING_TYPE): Define to
425 * config/pdp11/pdp11.h (PREFERRED_DEBUGGING_TYPE): Likewise.
427 2021-10-15 Richard Biener <rguenther@suse.de>
430 * tree-inline.c (copy_bb): Avoid underflowing nargs.
432 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
434 * config/i386/i386-expand.c (ix86_expand_vec_perm): Convert
435 HFmode input operand to HImode.
436 (ix86_vectorize_vec_perm_const): Likewise.
437 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1_hf):
439 (*avx512f_permvar_truncv8siv8hi_1_hf):
442 2021-10-15 Richard Biener <rguenther@suse.de>
445 * expmed.c (store_bit_field_1): Ensure a LHS subreg would
446 not create a paradoxical subreg.
448 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
450 * config/i386/i386-expand.c (ix86_expand_vector_init):
451 For half_vector concat for HFmode, handle them like HImode.
453 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
455 * tree-ssa-threadbackward.c (class back_threader): Add m_resolve.
456 (back_threader::back_threader): Same.
457 (back_threader::resolve_phi): Try to solve without looking back if
459 (back_threader::find_paths_to_names): Same.
460 (try_thread_blocks): Pass resolve argument to back threader.
461 (pass_early_thread_jumps::execute): Same.
463 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
465 * doc/invoke.texi: Remove max-fsm-thread-length,
466 max-fsm-thread-paths, and fsm-maximum-phi-arguments.
468 * tree-ssa-threadbackward.c (back_threader::back_threader): Remove
470 (back_threader_registry::back_threader_registry): Same.
471 (back_threader_profitability::profitable_path_p): Remove
472 param_max_fsm_thread-length.
473 (back_threader_registry::register_path): Remove
474 m_max_allowable_paths.
476 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
478 * tree-ssa-threadbackward.c (class back_threader): Make m_imports
480 (back_threader::~back_threader): Do not release m_path.
482 2021-10-14 Andrew MacLeod <amacleod@redhat.com>
484 PR tree-optimization/102738
485 * vr-values.c (simplify_using_ranges::simplify): Handle RSHIFT_EXPR.
487 2021-10-14 Kwok Cheung Yeung <kcy@codesourcery.com>
489 * omp-general.c (omp_check_context_selector): Move from c-omp.c.
490 (omp_mark_declare_variant): Move from c-omp.c.
491 (omp_context_name_list_prop): Update for Fortran strings.
492 * omp-general.h (omp_check_context_selector): New prototype.
493 (omp_mark_declare_variant): New prototype.
495 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
497 * config/rs6000/rs6000.c (rs6000_density_test): Move early
498 exit test further up the function.
500 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
502 * config/arm/arm.c (arm_add_stmt_cost): Delete.
503 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
505 2021-10-14 Martin Jambor <mjambor@suse.cz>
507 * doc/invoke.texi (Optimize Options): Add entry for
508 ipa-cp-recursive-freq-factor.
510 2021-10-14 Tamar Christina <tamar.christina@arm.com>
512 * match.pd: New rule.
514 2021-10-14 Jan Hubicka <hubicka@ucw.cz>
517 * ipa-modref-tree.h (modref_access_node::update2):
518 Also check that parm_offset is unchanged.
519 (modref_ref_node::insert_access): Fix updating of
522 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
524 * tree-ssa-threadbackward.c (back_threader::resolve_phi): Add
527 2021-10-14 Richard Biener <rguenther@suse.de>
529 PR tree-optimization/102659
530 * tree-if-conv.c (if_convertible_gimple_assign_stmt_p): Also
531 rewrite pointer typed undefined overflow operations.
532 (predicate_statements): Likewise. Make sure to emit invariant
533 conversions in the preheader.
534 * tree-vectorizer.c (vect_loop_vectorized_call): Look through
535 non-empty preheaders.
536 * tree-data-ref.c (dr_analyze_indices): Strip useless
537 conversions to the MEM_REF base type.
539 2021-10-14 Martin Liska <mliska@suse.cz>
541 * common.opt: Stop using AUTODETECT_VALUE
542 and use EnabledBy where possible.
543 * opts.c: Enable OPT_fvar_tracking with optimize >= 1.
544 * toplev.c (AUTODETECT_VALUE): Remove macro.
545 (process_options): Simplify by using EnabledBy and
546 OPT_fvar_tracking. Use OPTION_SET_P macro instead of
549 2021-10-14 Jonathan Wright <jonathan.wright@arm.com>
551 * config/aarch64/arm_neon.h (vld1_s8_x3): Use signed type for
553 (vld1_s32_x3): Likewise.
555 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
557 PR tree-optimization/102736
558 PR tree-optimization/102736
559 * gimple-range-path.cc (path_range_query::range_on_path_entry):
560 Assert that the requested range is defined outside the path.
561 (path_range_query::ssa_range_in_phi): Do not call
562 range_on_path_entry for SSA names that are defined within the
565 2021-10-14 Iain Sandoe <iain@sandoe.co.uk>
567 * config/darwin-driver.c (darwin_find_version_from_kernel):
568 Quote internal identifiers and avoid contractions in
570 (darwin_default_min_version): Likewise.
571 (darwin_driver_init): Likewise.
573 2021-10-14 Martin Jambor <mjambor@suse.cz>
575 * params.opt (ipa-cp-recursive-freq-factor): New.
576 * ipa-cp.c (ipcp_value): Switch to inline initialization. New members
577 scc_no, self_recursion_generated_level, same_scc and
578 self_recursion_generated_p.
579 (ipcp_lattice::add_value): Replaced parameter unlimited with
580 same_lat_gen_level, usit it determine limit of values and store it to
582 (ipcp_lattice<valtype>::print): Dump the new fileds.
583 (allocate_and_init_ipcp_value): Take same_lat_gen_level as a new
584 parameter and store it to the new value.
585 (self_recursively_generated_p): Removed.
586 (propagate_vals_across_arith_jfunc): Use self_recursion_generated_p
587 instead of self_recursively_generated_p, store self generation level
589 (value_topo_info<valtype>::add_val): Set scc_no.
590 (value_topo_info<valtype>::propagate_effects): Multiply frequencies of
591 recursively feeding values and self generated values by appropriate
594 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
596 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Remove
597 redundant test for flag_vect_cost_model.
599 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
601 * bitmap.c (debug): New overloaded function for auto_bitmaps.
602 * bitmap.h (debug): Same.
604 2021-10-14 Hongyu Wang <hongyu.wang@intel.com>
606 * config/i386/avx512fp16intrin.h (_mm512_mask_fcmadd_pch):
608 (_mm512_mask3_fcmadd_pch): Likewise.
609 (_mm512_mask_fmadd_pch): Likewise
610 (_mm512_mask3_fmadd_pch): Likewise
611 (_mm512_mask_fcmadd_round_pch): Likewise
612 (_mm512_mask3_fcmadd_round_pch): Likewise
613 (_mm512_mask_fmadd_round_pch): Likewise
614 (_mm512_mask3_fmadd_round_pch): Likewise
615 (_mm_mask_fcmadd_sch): Likewise
616 (_mm_mask3_fcmadd_sch): Likewise
617 (_mm_mask_fmadd_sch): Likewise
618 (_mm_mask3_fmadd_sch): Likewise
619 (_mm_mask_fcmadd_round_sch): Likewise
620 (_mm_mask3_fcmadd_round_sch): Likewise
621 (_mm_mask_fmadd_round_sch): Likewise
622 (_mm_mask3_fmadd_round_sch): Likewise
623 (_mm_fcmadd_round_sch): Likewise
624 * config/i386/avx512fp16vlintrin.h (_mm_mask_fmadd_pch):
626 (_mm_mask3_fmadd_pch): Likewise
627 (_mm256_mask_fmadd_pch): Likewise
628 (_mm256_mask3_fmadd_pch): Likewise
629 (_mm_mask_fcmadd_pch): Likewise
630 (_mm_mask3_fcmadd_pch): Likewise
631 (_mm256_mask_fcmadd_pch): Likewise
632 (_mm256_mask3_fcmadd_pch): Likewise
633 * config/i386/i386-builtin.def: Add mask3 builtin for complex
634 fma, and adjust mask_builtin to corresponding expander.
635 * config/i386/i386-expand.c (ix86_expand_round_builtin):
636 Skip eraseing embedded rounding for expanders that emits
638 * config/i386/sse.md (complexmove): New mode_attr.
639 (<avx512>_fmaddc_<mode>_mask1<round_expand_name>): New expander.
640 (<avx512>_fcmaddc_<mode>_mask1<round_expand_name>): Likewise.
641 (avx512fp16_fmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
642 (avx512fp16_fcmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
643 (avx512fp16_fcmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
644 (avx512fp16_fmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
645 * config/i386/subst.md (round_embedded_complex): New subst.
647 2021-10-14 John David Anglin <danglin@gcc.gnu.org>
649 * config/pa/pa.md (cbranchsf4): Disable if TARGET_SOFT_FLOAT.
650 (cbranchdf4): Likewise.
651 Add missing move patterns for TARGET_SOFT_FLOAT.
653 2021-10-13 Roger Sayle <roger@nextmovesoftware.com>
655 * config/i386/i386-expand.c (ix86_expand_vector_move): Use a
656 pseudo intermediate when moving a SUBREG into a hard register,
657 by checking ix86_hardreg_mov_ok.
658 (ix86_expand_vector_extract): Store zero-extended SImode
659 intermediate in a pseudo, then set target using a SUBREG_PROMOTED
661 * config/i386/sse.md (mov<VMOVE>_internal): Prevent CSE creating
662 complex (SUBREG) sets of (vector) hard registers before reload, by
663 checking ix86_hardreg_mov_ok.
665 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
667 * ctfc.h (enum ctf_dtu_d_union_enum): Remove redundant comma.
669 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
671 * dwarf2ctf.c (gen_ctf_array_type): Fix typo in comment.
673 2021-10-13 Martin Sebor <msebor@redhat.com>
676 * pointer-query.cc (compute_objsize_r): Handle named address spaces.
678 2021-10-13 Iain Sandoe <iain@sandoe.co.uk>
680 * collect2.c (is_lto_object_file): Release simple-object
681 resources, close files.
683 2021-10-13 Andre Vieira <andre.simoesdiasvieira@arm.com>
685 * config/arm/arm.c (thumb2_legitimate_address_p): Use VALID_MVE_MODE
686 when checking mve addressing modes.
687 (mve_vector_mem_operand): Fix the way we handle pre, post and offset
689 (arm_print_operand): Fix printing of POST_ and PRE_MODIFY.
690 * config/arm/mve.md: Use mve_memory_operand predicate everywhere where
691 there is a single Ux constraint.
693 2021-10-13 John David Anglin <danglin@gcc.gnu.org>
695 * config/pa/pa.md (muldi3): Add support for inlining 64-bit
696 multiplication on 32-bit PA 1.1 and 2.0 targets.
698 2021-10-13 Martin Liska <mliska@suse.cz>
700 PR gcov-profile/90364
701 * coverage.c (build_info): Emit checksum to the global variable.
702 (build_info_type): Add new field for checksum.
703 (coverage_obj_finish): Pass object_checksum.
704 (coverage_init): Use 0 as checksum for .gcno files.
705 * gcov-dump.c (dump_gcov_file): Dump also new checksum field.
706 * gcov.c (read_graph_file): Read also checksum.
707 * doc/invoke.texi: Document the behaviour change.
709 2021-10-13 Richard Biener <rguenther@suse.de>
711 * gimple-iterator.h (gsi_iterator_update): Add GSI_LAST_NEW_STMT,
712 start at integer value 2.
713 * gimple-iterator.c (gsi_insert_seq_nodes_before): Update
714 the iterator for GSI_LAST_NEW_STMT.
715 (gsi_insert_seq_nodes_after): Likewise.
716 * tree-if-conv.c (predicate_statements): Use GSI_LAST_NEW_STMT.
717 * tree-ssa.c (execute_update_addresses_taken): Correct bogus
718 arguments to gsi_replace.
720 2021-10-13 Martin Liska <mliska@suse.cz>
723 * common.opt: Use EnabledBy instead of detection in
724 finish_options and process_options.
725 * opts.c (finish_options): Remove handling of
726 x_flag_unroll_all_loops.
727 * toplev.c (process_options): Likewise for flag_web and
728 flag_rename_registers.
730 2021-10-13 Richard Biener <rguenther@suse.de>
732 PR tree-optimization/102659
733 * tree-if-conv.c (need_to_rewrite_undefined): New flag.
734 (if_convertible_gimple_assign_stmt_p): Mark the loop for
735 rewrite when stmts with undefined behavior on integer
737 (combine_blocks): Predicate also when we need to rewrite stmts.
738 (predicate_statements): Rewrite affected stmts to something
739 with well-defined behavior on overflow.
740 (tree_if_conversion): Initialize need_to_rewrite_undefined.
742 2021-10-13 Richard Biener <rguenther@suse.de>
745 * ipa-sra.c (ptr_parm_has_nonarg_uses): Fix volatileness
748 2021-10-13 Jose E. Marchesi <jose.marchesi@oracle.com>
750 * dwarf2ctf.c: Fix typo in comment.
752 2021-10-12 Bill Schmidt <wschmidt@linux.ibm.com>
755 * config/rs6000/altivec.h (vec_cpsgn): Swap operand order.
756 * config/rs6000/rs6000-overload.def (VEC_COPYSIGN): Use SKIP to
757 avoid generating an automatic #define of vec_cpsgn. Use the
758 correct built-in for V4SFmode that doesn't depend on VSX.
760 2021-10-12 Uroš Bizjak <ubizjak@gmail.com>
764 * config/i386/i386.md (*add<mode>_1_slp): Rewrite as
765 define_insn_and_split pattern. Add alternative 1 and split it
766 post reload to insert operand 1 into the low part of operand 0.
767 (*sub<mode>_1_slp): Ditto.
768 (*and<mode>_1_slp): Ditto.
769 (*<any_or:code><mode>_1_slp): Ditto.
770 (*ashl<mode>3_1_slp): Ditto.
771 (*<any_shiftrt:insn><mode>3_1_slp): Ditto.
772 (*<any_rotate:insn><mode>3_1_slp): Ditto.
773 (*neg<mode>_1_slp): New insn_and_split pattern.
774 (*one_cmpl<mode>_1_slp): Ditto.
776 2021-10-12 David Edelsohn <dje.gcc@gmail.com>
778 * doc/install.texi: Update MinGW and mingw-64 Binaries
781 2021-10-12 Daniel Le Duc Khoi Nguyen <greenrecyclebin@gmail.com>
783 * doc/extend.texi (Common Variable Attributes): Fix typos in
784 alloc_size documentation.
786 2021-10-12 Richard Biener <rguenther@suse.de>
788 PR tree-optimization/102696
789 * tree-vect-slp.c (vect_build_slp_tree_2): Properly mark
790 the tree fatally failed when we reject a BIT_FIELD_REF.
792 2021-10-12 Richard Biener <rguenther@suse.de>
794 PR tree-optimization/102572
795 * tree-vect-stmts.c (vect_build_gather_load_calls): When
796 gathering the vectorized defs for the mask pass in the
797 desired mask vector type so invariants will be handled
800 2021-10-12 Tamar Christina <tamar.christina@arm.com>
802 * config/aarch64/aarch64-sve.md (*fcm<cmp_op><mode>_bic_combine,
803 *fcm<cmp_op><mode>_nor_combine, *fcmuo<mode>_bic_combine,
804 *fcmuo<mode>_nor_combine): New.
806 2021-10-12 Eric Botcazou <ebotcazou@adacore.com>
809 * config/sparc/sparc-modes.def (OI): New integer mode.
811 2021-10-12 Jakub Jelinek <jakub@redhat.com>
813 * gimple-fold.h (clear_padding_type_may_have_padding_p): Declare.
814 * gimple-fold.c (clear_padding_type_may_have_padding_p): No longer
817 2021-10-12 Jakub Jelinek <jakub@redhat.com>
819 * tree-vectorizer.h (loop_cost_model): New function.
820 (unlimited_cost_model): Use it.
821 * tree-vect-loop.c (vect_analyze_loop_costing): Use loop_cost_model
822 call instead of flag_vect_cost_model.
823 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
824 (vect_prune_runtime_alias_test_list): Likewise. Also use it instead
825 of flag_simd_cost_model.
827 2021-10-12 liuhongt <hongtao.liu@intel.com>
830 * config/i386/i386-expand.c (emit_reduc_half): Handle
832 * config/i386/mmx.md (reduc_<code>_scal_v4qi): New expander.
833 (reduc_plus_scal_v4qi): Ditto.
835 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
837 * config/rs6000/smmintrin.h (_mm_cmpeq_epi64, _mm_cmpgt_epi64,
838 _mm_mullo_epi32, _mm_mul_epi32, _mm_packus_epi32): New.
839 * config/rs6000/nmmintrin.h: Copy from i386, tweak to suit.
841 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
843 * config/rs6000/smmintrin.h (_mm_cvtepi8_epi16, _mm_cvtepi8_epi32,
844 _mm_cvtepi8_epi64, _mm_cvtepi16_epi32, _mm_cvtepi16_epi64,
845 _mm_cvtepi32_epi64, _mm_cvtepu8_epi16, _mm_cvtepu8_epi32,
846 _mm_cvtepu8_epi64, _mm_cvtepu16_epi32, _mm_cvtepu16_epi64,
847 _mm_cvtepu32_epi64): New.
849 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
851 * config/rs6000/smmintrin.h (_mm_test_all_zeros,
852 _mm_test_all_ones, _mm_test_mix_ones_zeros): Rewrite as macro.
854 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
856 * config/rs6000/smmintrin.h (_mm_min_epi8, _mm_min_epu16,
857 _mm_min_epi32, _mm_min_epu32, _mm_max_epi8, _mm_max_epu16,
858 _mm_max_epi32, _mm_max_epu32): New.
860 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
862 * ipa-modref-tree.h (struct modref_access_node): Revert
864 (struct modref_ref_node): Likewise.
866 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
868 * ipa-modref-tree.h (modref_tree::global_access_p): New member
871 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
872 ignore_stores_eaf_flags): Move to ipa-modref.h
873 (remove_useless_eaf_flags): Remove early exit on NOCLOBBER.
874 (modref_summary::global_memory_read_p): New member function.
875 (modref_summary::global_memory_written_p): New member function.
876 * ipa-modref.h (modref_summary::global_memory_read_p,
877 modref_summary::global_memory_written_p): Declare.
878 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
879 ignore_stores_eaf_flags): move here.
880 * tree-ssa-structalias.c: Include ipa-modref-tree.h, ipa-modref.h
882 (handle_rhs_call): Rewrite.
883 (handle_call_arg): New function.
884 (determine_global_memory_access): New function.
885 (handle_const_call): Remove
886 (handle_pure_call): Remove
887 (find_func_aliases_for_call): Update use of handle_rhs_call.
888 (compute_points_to_sets): Handle global memory acccesses
891 2021-10-11 Diane Meirowitz <diane.meirowitz@oracle.com>
893 * doc/invoke.texi: Add link to UndefinedBehaviorSanitizer
894 documentation, mention UBSAN_OPTIONS, similar to what is done
895 for AddressSanitizer.
897 2021-10-11 Richard Biener <rguenther@suse.de>
900 * internal-fn.c (expand_DEFERRED_INIT): Check for mode
901 availability before building an integer type for storage
904 2021-10-11 Richard Biener <rguenther@suse.de>
907 * gimple.c (gimple_call_fnspec): Do not mark operator new/delete
910 2021-10-11 Martin Liska <mliska@suse.cz>
912 * common.opt: Remove Init(2) for some options.
913 * toplev.c (process_options): Do not use AUTODETECT_VALUE, but
914 use rather OPTION_SET_P.
916 2021-10-11 Martin Liska <mliska@suse.cz>
918 * common.opt: Remove usage of IRA_REGION_AUTODETECT.
919 * flag-types.h (enum ira_region): Likewise.
920 * toplev.c (process_options): Use OPTION_SET_P instead of
921 IRA_REGION_AUTODETECT.
923 2021-10-11 Jakub Jelinek <jakub@redhat.com>
925 * omp-low.c (omp_runtime_api_call): Handle omp_get_max_teams,
926 omp_[sg]et_teams_thread_limit and omp_set_num_teams.
928 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
930 * config/s390/s390-protos.h (s390_rawmemchr): Add prototype.
931 * config/s390/s390.c (s390_rawmemchr): New function.
932 * config/s390/s390.md (rawmemchr<SINT:mode>): New expander.
933 * config/s390/vector.md (@vec_vfees<mode>): Basically a copy of
934 the pattern vfees<mode> from vx-builtins.md.
935 * config/s390/vx-builtins.md (*vfees<mode>): Remove.
937 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
939 * builtins.c (get_memory_rtx): Change to external linkage.
940 * builtins.h (get_memory_rtx): Add function prototype.
941 * doc/md.texi (rawmemchr<mode>): Document.
942 * internal-fn.c (expand_RAWMEMCHR): Define.
943 * internal-fn.def (RAWMEMCHR): Add.
944 * optabs.def (rawmemchr_optab): Add.
945 * tree-loop-distribution.c (find_single_drs): Change return code
946 behaviour by also returning true if no single store was found
948 (loop_distribution::classify_partition): Respect the new return
949 code behaviour of function find_single_drs.
950 (loop_distribution::execute): Call new function
951 transform_reduction_loop in order to replace rawmemchr or strlen
952 like loops by calls into builtins.
953 (generate_reduction_builtin_1): New function.
954 (generate_rawmemchr_builtin): New function.
955 (generate_strlen_builtin_1): New function.
956 (generate_strlen_builtin): New function.
957 (generate_strlen_builtin_using_rawmemchr): New function.
958 (reduction_var_overflows_first): New function.
959 (determine_reduction_stmt_1): New function.
960 (determine_reduction_stmt): New function.
961 (loop_distribution::transform_reduction_loop): New function.
963 2021-10-11 Martin Liska <mliska@suse.cz>
965 * tree.c (cl_option_hasher::hash): Use cl_optimization_hash
966 and remove legacy hashing code.
968 2021-10-11 Kito Cheng <kito.cheng@sifive.com>
971 * builtins.c (maybe_emit_call_builtin___clear_cache): Allow
972 CONST_INT for BEGIN and END, and use gcc_assert rather than
975 2021-10-10 Jakub Jelinek <jakub@redhat.com>
978 * var-tracking.c (add_stores): For cselib_sp_derived_value_p values
979 use MO_VAL_SET if loc is not sp.
981 2021-10-10 Andrew Pinski <apinski@marvell.com>
983 PR tree-optimization/102622
984 * match.pd: Swap the order of a?pow2cst:0 and a?-1:0 transformations.
985 Swap the order of a?0:pow2cst and a?0:-1 transformations.
987 2021-10-09 liuhongt <hongtao.liu@intel.com>
990 * config/i386/i386-expand.c (ix86_valid_mask_cmp_mode): Handle
992 (ix86_use_mask_cmp_p): Ditto.
993 (ix86_expand_sse_movcc): Ditto.
994 * config/i386/i386.md (setcc_hf_mask): New define_insn.
996 (UNSPEC_MOVCC_MASK): New unspec.
997 * config/i386/sse.md (UNSPEC_PCMP): Move to i386.md.
999 2021-10-08 Vladimir N. Makarov <vmakarov@redhat.com>
1001 PR rtl-optimization/102627
1002 * lra-constraints.c (split_reg): Use at least natural mode of hard reg.
1004 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
1006 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Grow
1009 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
1011 * value-range.cc (irange::debug): New.
1012 * value-range.h (irange::debug): New.
1014 2021-10-08 Richard Sandiford <richard.sandiford@arm.com>
1016 PR tree-optimization/102385
1017 * predict.h (change_edge_frequency): Declare.
1018 * predict.c (change_edge_frequency): New function.
1019 * tree-ssa-loop-manip.h (tree_transform_and_unroll_loop): Remove
1021 (tree_unroll_loop): Likewise.
1022 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Update accordingly.
1023 * tree-predcom.c (pcom_worker::tree_predictive_commoning_loop):
1025 * tree-ssa-loop-prefetch.c (loop_prefetch_arrays): Likewise.
1026 * tree-ssa-loop-manip.c (tree_unroll_loop): Likewise.
1027 (tree_transform_and_unroll_loop): Likewise. Use single_dom_exit
1028 to retrieve the exit edges. Make all the old profile update code
1029 conditional on !single_loop_p -- the case it was written for --
1030 and use a different approach for the single-loop case.
1032 2021-10-08 Martin Liska <mliska@suse.cz>
1034 * config/alpha/alpha.c (alpha_option_override): Use new macro
1036 * config/arc/arc.c (arc_override_options): Likewise.
1037 * config/arm/arm.c (arm_option_override): Likewise.
1038 * config/bfin/bfin.c (bfin_load_pic_reg): Likewise.
1039 * config/c6x/c6x.c (c6x_option_override): Likewise.
1040 * config/csky/csky.c: Likewise.
1041 * config/darwin.c (darwin_override_options): Likewise.
1042 * config/frv/frv.c (frv_option_override): Likewise.
1043 * config/i386/djgpp.h: Likewise.
1044 * config/i386/i386.c (ix86_stack_protect_guard): Likewise.
1045 (ix86_max_noce_ifcvt_seq_cost): Likewise.
1046 * config/ia64/ia64.c (ia64_option_override): Likewise.
1047 (ia64_override_options_after_change): Likewise.
1048 * config/m32c/m32c.c (m32c_option_override): Likewise.
1049 * config/m32r/m32r.c (m32r_init): Likewise.
1050 * config/m68k/m68k.c (m68k_option_override): Likewise.
1051 * config/microblaze/microblaze.c (microblaze_option_override): Likewise.
1052 * config/mips/mips.c (mips_option_override): Likewise.
1053 * config/nios2/nios2.c (nios2_option_override): Likewise.
1054 * config/nvptx/nvptx.c (nvptx_option_override): Likewise.
1055 * config/pa/pa.c (pa_option_override): Likewise.
1056 * config/riscv/riscv.c (riscv_option_override): Likewise.
1057 * config/rs6000/aix71.h: Likewise.
1058 * config/rs6000/aix72.h: Likewise.
1059 * config/rs6000/aix73.h: Likewise.
1060 * config/rs6000/rs6000.c (darwin_rs6000_override_options): Likewise.
1061 (rs6000_override_options_after_change): Likewise.
1062 (rs6000_linux64_override_options): Likewise.
1063 (glibc_supports_ieee_128bit): Likewise.
1064 (rs6000_option_override_internal): Likewise.
1065 (rs6000_file_start): Likewise.
1066 (rs6000_darwin_file_start): Likewise.
1067 * config/rs6000/rtems.h: Likewise.
1068 * config/rs6000/sysv4.h: Likewise.
1069 * config/rs6000/vxworks.h (SUB3TARGET_OVERRIDE_OPTIONS): Likewise.
1070 * config/s390/s390.c (s390_option_override): Likewise.
1071 * config/sh/linux.h: Likewise.
1072 * config/sh/netbsd-elf.h (while): Likewise.
1073 * config/sh/sh.c (sh_option_override): Likewise.
1074 * config/sol2.c (solaris_override_options): Likewise.
1075 * config/sparc/sparc.c (sparc_option_override): Likewise.
1076 * config/tilegx/tilegx.c (tilegx_option_override): Likewise.
1077 * config/visium/visium.c (visium_option_override): Likewise.
1078 * config/vxworks.c (vxworks_override_options): Likewise.
1079 * lto-opts.c (lto_write_options): Likewise.
1080 * omp-expand.c (expand_omp_simd): Likewise.
1081 * omp-general.c (omp_max_vf): Likewise.
1082 * omp-offload.c (oacc_xform_loop): Likewise.
1083 * opts.h (OPTION_SET_P): Likewise.
1084 * targhooks.c (default_max_noce_ifcvt_seq_cost): Likewise.
1085 * toplev.c (process_options): Likewise.
1086 * tree-predcom.c: Likewise.
1087 * tree-sra.c (analyze_all_variable_accesses): Likewise.
1089 2021-10-08 liuhongt <hongtao.liu@intel.com>
1092 * config/i386/i386.c (ix86_optab_supported_p):
1093 Return true for HFmode.
1094 * match.pd: Simplify (_Float16) ceil ((double) x) to
1095 __builtin_ceilf16 (a) when a is _Float16 type and
1096 direct_internal_fn_supported_p.
1098 2021-10-08 liuhongt <hongtao.liu@intel.com>
1101 * config/i386/i386-expand.c (emit_reduc_half): Hanlde V4HImode.
1102 * config/i386/mmx.md (reduc_plus_scal_v4hi): New.
1103 (reduc_<code>_scal_v4hi): New.
1105 2021-10-08 liuhongt <hongtao.liu@intel.com>
1107 * common.opt (ftree-vectorize): Add Var(flag_tree_vectorize).
1108 * doc/invoke.texi (Options That Control Optimization): Update
1110 * opts.c (default_options_table): Enable auto-vectorization at
1111 O2 with very-cheap cost model.
1112 (finish_options): Use cheap cost model for
1113 explicit -ftree{,-loop}-vectorize.
1115 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
1117 * ctfc.c (ctfc_delete_container): Free hash table contents.
1119 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
1121 * toplev.c (process_options): Do not warn for GNU GIMPLE.
1123 2021-10-07 Siddhesh Poyarekar <siddhesh@gotplt.org>
1125 * tree-object-size.c (addr_object_size,
1126 compute_builtin_object_size): Drop PDECL and POFF arguments.
1127 (addr_object_size): Adjust calls.
1128 * tree-object-size.h (compute_builtin_object_size): Drop PDECL
1131 2021-10-07 Roger Sayle <roger@nextmovesoftware.com>
1133 * rtl.def (SMUL_HIGHPART, UMUL_HIGHPART): New RTX codes for
1134 representing signed and unsigned high-part multiplication resp.
1135 * simplify-rtx.c (simplify_binary_operation_1) [SMUL_HIGHPART,
1136 UMUL_HIGHPART]: Simplify high-part multiplications by zero.
1137 [SS_PLUS, US_PLUS, SS_MINUS, US_MINUS, SS_MULT, US_MULT,
1138 SS_DIV, US_DIV]: Similar simplifications for saturating
1140 (simplify_const_binary_operation) [SS_PLUS, US_PLUS, SS_MINUS,
1141 US_MINUS, SS_MULT, US_MULT, SMUL_HIGHPART, UMUL_HIGHPART]:
1142 Implement compile-time evaluation for constant operands.
1143 * dwarf2out.c (mem_loc_descriptor): Skip SMUL_HIGHPART and
1145 * doc/rtl.texi (smul_highpart, umul_highpart): Document RTX codes.
1146 * doc/md.texi (smul@var{m}3_highpart, umul@var{m3}_highpart):
1147 Mention the new smul_highpart and umul_highpart RTX codes.
1148 * doc/invoke.texi: Silence @xref "compilation" warnings.
1150 2021-10-07 Martin Jambor <mjambor@suse.cz>
1153 * ipa-prop.c (ipa_edge_args_sum_t::duplicate): Also handle the
1154 case when the source reference description corresponds to a
1155 referance taken in a function src->caller is inlined to.
1157 2021-10-07 Jan Hubicka <hubicka@ucw.cz>
1160 * ipa-modref-tree.h (modref_access_node::contains_p): Handle offsets
1162 (modref_access_node::try_merge_with): Add sanity check that there
1163 are no redundant entries in the list.
1165 2021-10-07 Richard Biener <rguenther@suse.de>
1167 PR tree-optimization/102608
1168 * tree-ssa-sccvn.c (visit_stmt): Drop .DEFERRED_INIT to
1171 2021-10-07 Martin Liska <mliska@suse.cz>
1173 * toplev.c (toplev::main): Make
1174 save_opt_decoded_options a pointer type
1175 * toplev.h: Likewise.
1177 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
1179 * config/gcn/gcn-valu.md (gather<mode>_insn_2offsets<exec>): Apply
1180 HAVE_GCN_ASM_GLOBAL_LOAD_FIXED.
1181 (scatter<mode>_insn_2offsets<exec_scatter>): Likewise.
1183 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
1185 * config/gcn/gcn-hsa.h (SRAMOPT): Include the whole option string.
1186 Adjust for new -msram-ecc=any behaviour.
1187 (ASM_SPEC): Adjust -mxnack and -msram-ecc usage.
1188 * config/gcn/gcn.c (output_file_start): Implement -msram-ecc=any.
1189 * config/gcn/mkoffload.c (EF_AMDGPU_XNACK): Rename to ...
1190 (EF_AMDGPU_XNACK_V3): ... this.
1191 (EF_AMDGPU_SRAM_ECC): Rename to ...
1192 (EF_AMDGPU_SRAM_ECC_V3): ... this.
1193 (EF_AMDGPU_FEATURE_XNACK_V4): New.
1194 (EF_AMDGPU_FEATURE_XNACK_UNSUPPORTED_V4): New.
1195 (EF_AMDGPU_FEATURE_XNACK_ANY_V4): New.
1196 (EF_AMDGPU_FEATURE_XNACK_OFF_V4): New.
1197 (EF_AMDGPU_FEATURE_XNACK_ON_V4): New.
1198 (EF_AMDGPU_FEATURE_SRAMECC_V4): New.
1199 (EF_AMDGPU_FEATURE_SRAMECC_UNSUPPORTED_V4): New.
1200 (EF_AMDGPU_FEATURE_SRAMECC_ANY_V4): New.
1201 (EF_AMDGPU_FEATURE_SRAMECC_OFF_V4): New.
1202 (EF_AMDGPU_FEATURE_SRAMECC_ON_V4): New.
1203 (SET_XNACK_ON): New.
1204 (SET_XNACK_OFF): New.
1206 (SET_SRAM_ECC_ON): New.
1207 (SET_SRAM_ECC_ANY): New.
1208 (SET_SRAM_ECC_OFF): New.
1209 (TEST_SRAM_ECC_ANY): New.
1210 (TEST_SRAM_ECC_ON): New.
1211 (main): Implement HSACOv4 and -msram-ecc=any.
1213 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
1215 * config.in: Regenerate.
1216 * config/gcn/gcn-hsa.h (X_FIJI): New macro.
1220 (A_FIJI): Rename to ...
1222 (A_900): Rename to ...
1224 (A_906): Rename to ...
1226 (A_908): Rename to ...
1228 (SRAMOPT): New macro.
1229 (ASM_SPEC): Adjust xnack option usage.
1230 * config/gcn/gcn.c (output_file_start): Adjust amdgcn_target usage.
1231 * configure: Regenerate.
1232 * configure.ac: Detect LLVM assembler dialect.
1234 2021-10-07 Richard Biener <rguenther@suse.de>
1236 * tree-pretty-print.c (dump_generic_node): Do not elide
1237 printing '&' when dumping with -gimple.
1239 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
1241 * gimple-range-cache.cc (non_null_ref::adjust_range): Call new
1243 * gimple-range-fold.cc (adjust_pointer_diff_expr): Ditto.
1244 (adjust_imagpart_expr): Ditto.
1245 * value-range.cc (irange::irange_intersect): Call new routine if
1246 RHS is a single pair.
1247 (irange::intersect): New wide_int version.
1248 * value-range.h (class irange): New prototype.
1250 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
1252 * gimple-range-edge.cc (gimple_outgoing_range::gimple_outgoing_range):
1253 Add parameter to limit size when recognizing switches.
1254 (gimple_outgoing_range::edge_range_p): Check size limit.
1255 * gimple-range-edge.h (gimple_outgoing_range): Add size field.
1256 * gimple-range-gori.cc (gori_map::calculate_gori): Ignore switches
1257 that exceed the size limit.
1258 (gori_compute::gori_compute): Add initializer.
1259 * params.opt (evrp-switch-limit): New.
1260 * doc/invoke.texi: Update docs.
1262 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
1264 * value-range.h (irange::set_varying): Use TYPE_MIN_VALUE and
1265 TYPE_MAX_VALUE instead of creating new trees when possible.
1267 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
1269 * gimple-range-cache.cc (non_null_ref::adjust_range): Check for
1270 zero and non-zero more efficently.
1272 2021-10-06 Richard Biener <rguenther@suse.de>
1275 * dumpfile.h (TDF_GIMPLE_VAL): New.
1276 (dump_flag): Re-order and adjust TDF_* flags. Make
1277 the enum uint32_t. Use std::underlying_type in the
1279 (optgroup_flag): Likewise for the operator overloads.
1280 * tree-pretty-print.c (dump_generic_node): Wrap ADDR_EXPR
1281 in _Literal if TDF_GIMPLE_VAL.
1282 * gimple-pretty-print.c (dump_gimple_assign): Add
1283 TDF_GIMPLE_VAL to flags when dumping operands where only
1284 is_gimple_val are allowed.
1285 (dump_gimple_cond): Likewise.
1287 2021-10-06 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
1289 * gimple-isel.cc (gimple_expand_vec_cond_expr): Remove redundant if
1292 2021-10-05 qing zhao <qing.zhao@oracle.com>
1294 PR middle-end/102359
1295 * gimplify.c (gimplify_decl_expr): Not add initialization for an
1296 auto variable when it has been initialized by frontend.
1298 2021-10-05 Aldy Hernandez <aldyh@redhat.com>
1300 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
1303 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
1305 * common/config/avr/avr-common.c (avr_handle_option): Mark
1306 argument as ATTRIBUTE_UNUSED.
1308 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
1310 * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC):
1311 Undefine before redefinition.
1313 2021-10-05 Richard Biener <rguenther@suse.de>
1315 * toplev.c (no_backend): Remove global var.
1316 (process_options): Pass in no_backend, move post_options
1317 langhook call to toplev::main.
1318 (do_compile): Pass in no_backend, move process_options call
1320 (toplev::run_self_tests): Check no_backend at the caller.
1321 (toplev::main): Call post_options and process_options
1322 split out from do_compile, do self-tests only if
1323 no_backend is initialized.
1325 2021-10-05 Richard Biener <rguenther@suse.de>
1327 * tree-cfg.c (dump_function_to_file): Dump the UID of the
1328 function as part of the name when requested.
1329 * tree-pretty-print.c (dump_function_name): Dump the UID when
1330 requested and the langhook produced the actual name.
1332 2021-10-05 Richard Biener <rguenther@suse.de>
1334 PR middle-end/102587
1335 PR middle-end/102285
1336 * internal-fn.c (expand_DEFERRED_INIT): Fall back to
1337 zero-initialization as last resort, use the constant
1338 size as given by the DEFERRED_INIT argument to build
1341 2021-10-04 Marek Polacek <polacek@redhat.com>
1344 * doc/invoke.texi: Document -Warray-compare.
1346 2021-10-04 Richard Biener <rguenther@suse.de>
1348 * gimplify.c (is_var_need_auto_init): DECL_HARD_REGISTER
1349 variables are not to be initialized.
1351 2021-10-04 Richard Biener <rguenther@suse.de>
1353 * expr.h (non_mem_decl_p): Declare.
1354 (mem_ref_refers_to_non_mem_p): Likewise.
1355 * expr.c (non_mem_decl_p): Export.
1356 (mem_ref_refers_to_non_mem_p): Likewise.
1357 * internal-fn.c (expand_DEFERRED_INIT): Do not expand the LHS
1358 but check the base with mem_ref_refers_to_non_mem_p
1361 2021-10-04 Richard Biener <rguenther@suse.de>
1363 PR tree-optimization/102570
1364 * tree-ssa-sccvn.h (vn_reference_op_struct): Document
1365 we are using clique for the internal function code.
1366 * tree-ssa-sccvn.c (vn_reference_op_eq): Compare the
1367 internal function code.
1368 (print_vn_reference_ops): Print the internal function code.
1369 (vn_reference_op_compute_hash): Hash it.
1370 (copy_reference_ops_from_call): Record it.
1371 (visit_stmt): Remove the restriction around internal function
1373 (fully_constant_vn_reference_p): Use fold_const_call and handle
1375 (vn_reference_eq): Compare call return types.
1376 * tree-ssa-pre.c (create_expression_by_pieces): Handle
1377 generating calls to internal functions.
1378 (compute_avail): Remove the restriction around internal function
1381 2021-10-04 Aldy Hernandez <aldyh@redhat.com>
1383 PR tree-optimization/102560
1384 * gimple-ssa-warn-alloca.c (alloca_call_type): Remove static
1385 marker for invalid_range.
1387 2021-10-04 Richard Biener <rguenther@suse.de>
1389 PR middle-end/102587
1390 * internal-fn.c (expand_DEFERRED_INIT): Guard register
1391 initialization path an avoid initializing VLA registers
1394 2021-10-04 Eric Botcazou <ebotcazou@adacore.com>
1396 * config/rs6000/vxworks.h (TARGET_INIT_LIBFUNCS): Delete.
1398 2021-10-03 Martin Liska <mliska@suse.cz>
1400 * toplev.c (toplev::main): Check opt_index if it is a part
1403 2021-10-02 Aldy Hernandez <aldyh@redhat.com>
1405 PR tree-optimization/102563
1406 * range-op.cc (operator_lshift::op1_range): Do not clobber
1409 2021-10-02 Martin Liska <mliska@suse.cz>
1411 * toplev.c (toplev::main): save_decoded_options[0] is program
1412 name and so it should be skipped.
1414 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
1416 PR tree-optimization/102546
1417 * range-op.cc (operator_lshift::op1_range): Teach range-ops that
1418 X << Y is non-zero implies X is also non-zero.
1420 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1422 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
1424 * config/aarch64/aarch64-tune.md: Regenerate.
1425 * doc/invoke.texi: Update docs.
1427 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1429 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
1431 * config/aarch64/aarch64-tune.md: Regenerate.
1432 * doc/invoke.texi: Update docs.
1434 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1436 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
1438 * config/aarch64/aarch64-tune.md: Regenerate.
1439 * doc/invoke.texi: Update docs.
1441 2021-10-01 Martin Sebor <msebor@redhat.com>
1444 * doc/invoke.texi (-Waddress): Update.
1445 * gengtype.c (write_types): Avoid -Waddress.
1446 * poly-int.h (POLY_SET_COEFF): Avoid using null.
1448 2021-10-01 John David Anglin <danglin@gcc.gnu.org>
1451 * config/pa/pa.c (pa_option_override): Default to dwarf version 4
1454 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1456 * config/aarch64/aarch64.h (AARCH64_FL_V9): Update value.
1458 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
1460 * gimple-range-path.cc (path_range_query::compute_ranges): Use
1462 * gimple-range-path.h (class path_range_query): Remove shadowed
1464 (path_range_query::get_path_oracle): New.
1466 2021-10-01 Jakub Jelinek <jakub@redhat.com>
1467 Richard Biener <rguenther@suse.de>
1470 * doc/invoke.texi (-fsanitize=integer-divide-by-zero): Remove
1471 INT_MIN / -1 division detection from here ...
1472 (-fsanitize=signed-integer-overflow): ... and add it here.
1474 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1476 * config/aarch64/aarch64-arches.def (AARCH64_ARCH): Added
1478 * config/aarch64/aarch64.h (AARCH64_FL_V9): New.
1479 (AARCH64_FL_FOR_ARCH9): New flags for Armv9-A.
1480 (AARCH64_ISA_V9): New ISA flag.
1481 * doc/invoke.texi: Update docs.
1483 2021-10-01 Martin Liska <mliska@suse.cz>
1485 * toplev.c (toplev::main): Save decoded optimization options.
1486 * toplev.h (save_opt_decoded_options): New.
1487 * doc/extend.texi: Be more clear about optimize and target
1490 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
1492 * explow.c: Include langhooks.h.
1493 (set_stack_check_libfunc): Build a proper function type.
1495 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
1498 * config/i386/i386.c (legitimate_pic_address_disp_p): For PE-COFF do
1499 not return true for external weak function symbols in medium model.
1501 2021-10-01 Jakub Jelinek <jakub@redhat.com>
1503 * tree.h (OMP_CLAUSE_ORDER_REPRODUCIBLE): Define.
1504 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_ORDER>: Print
1505 reproducible: for OMP_CLAUSE_ORDER_REPRODUCIBLE.
1506 * omp-general.c (omp_extract_for_data): If OMP_CLAUSE_ORDER is seen
1507 without OMP_CLAUSE_ORDER_UNCONSTRAINED, overwrite sched_kind to
1508 OMP_CLAUSE_SCHEDULE_STATIC.
1510 2021-10-01 Richard Biener <rguenther@suse.de>
1512 PR middle-end/102518
1513 * tree-inline.c (setup_one_parameter): Avoid substituting
1514 an invariant into contexts where a GIMPLE register is not valid.
1516 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1518 * config/arm/arm-cpus.in: Add Cortex-R52+ CPU.
1519 * config/arm/arm-tables.opt: Regenerate.
1520 * config/arm/arm-tune.md: Regenerate.
1521 * doc/invoke.texi: Update docs.
1523 2021-09-30 Uroš Bizjak <ubizjak@gmail.com>
1526 * config/i386/i386.md
1527 (sign_extend:WIDE (any_logic:NARROW (memory, immediate)) splitters):
1530 2021-09-30 Tobias Burnus <tobias@codesourcery.com>
1532 * omp-low.c (omp_runtime_api_call): Add omp_aligned_{,c}alloc and
1533 omp_{c,re}alloc, fix omp_alloc/omp_free.
1535 2021-09-30 Martin Liska <mliska@suse.cz>
1537 * defaults.h (ASM_OUTPUT_ASCII): Do not hide global variable
1538 asm_out_file and stream directly to MYFILE.
1540 2021-09-30 Richard Biener <rguenther@suse.de>
1542 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
1543 Restore and fix condition under which we apply npeel to
1544 the DRs misalignment value.
1546 2021-09-30 Richard Biener <rguenther@suse.de>
1548 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
1549 Fix npeel check for variable amount of peeling.
1551 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
1553 * lto-wrapper.c (run_gcc): Plug snprintf overflow.
1555 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
1557 * gimple-range.cc (gimple_ranger::debug): New.
1558 * gimple-range.h (class gimple_ranger): Add debug.
1560 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
1562 PR middle-end/102519
1563 * tree-vrp.c (hybrid_threader::~hybrid_threader): Free m_query.
1565 2021-09-29 Indu Bhagat <indu.bhagat@oracle.com>
1568 * btfout.c (GTY): Add GTY (()) albeit for cosmetic only purpose.
1569 (btf_finalize): Empty the hash_map btf_var_ids.
1571 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
1573 * tree-vrp.c (thread_through_all_blocks): Return bool.
1574 (execute_vrp_threader): Return TODO_* flags.
1575 (pass_data_vrp_threader): Set todo_flags_finish to 0.
1577 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
1579 * timevar.def (TV_TREE_VRP_THREADER): New.
1580 * tree-vrp.c: Use TV_TREE_VRP_THREADER for VRP threader pass.
1582 2021-09-29 David Faust <david.faust@oracle.com>
1584 * config.gcc (bpf-*-*): Do not overwrite extra_headers.
1586 2021-09-29 Jonathan Wright <jonathan.wright@arm.com>
1588 * config/aarch64/aarch64-builtins.c (TYPES_BINOP_PPU): Define
1589 new type qualifier enum.
1590 (TYPES_TERNOP_SSSU): Likewise.
1591 (TYPES_TERNOP_PPPU): Likewise.
1592 * config/aarch64/aarch64-simd-builtins.def: Define PPU, SSU,
1593 PPPU and SSSU builtin generator macros for qtbl1 and qtbx1
1595 * config/aarch64/arm_neon.h (vqtbl1_p8): Use type-qualified
1596 builtin and remove casts.
1597 (vqtbl1_s8): Likewise.
1598 (vqtbl1q_p8): Likewise.
1599 (vqtbl1q_s8): Likewise.
1600 (vqtbx1_s8): Likewise.
1601 (vqtbx1_p8): Likewise.
1602 (vqtbx1q_s8): Likewise.
1603 (vqtbx1q_p8): Likewise.
1604 (vtbl1_p8): Likewise.
1605 (vtbl2_p8): Likewise.
1606 (vtbx2_p8): Likewise.
1608 2021-09-29 Richard Biener <rguenther@suse.de>
1610 * tree-vect-data-refs.c (vect_dr_misalign_for_aligned_access):
1612 (vect_update_misalignment_for_peel): Use it to update
1613 misaligned to the value necessary for an aligned access.
1614 (vect_get_peeling_costs_all_drs): Likewise.
1615 (vect_enhance_data_refs_alignment): Likewise.
1617 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1619 * config/aarch64/aarch64.c (aarch64_expand_cpymem): Count number of
1620 emitted operations and adjust heuristic for code size.
1622 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1624 * config/aarch64/aarch64.c (aarch64_expand_setmem): Count number of
1625 emitted operations and adjust heuristic for code size.
1627 2021-09-29 Jakub Jelinek <jakub@redhat.com>
1629 PR middle-end/102504
1630 * gimplify.c (gimplify_scan_omp_clauses): Use omp_check_private even
1631 in OMP_SCOPE clauses, not just on worksharing construct clauses.
1633 2021-09-28 Geng Qi <gengqi@linux.alibaba.com>
1635 * config/riscv/riscv.md (mulv<mode>4): Call gen_smul<mode>3_highpart.
1636 (<u>mulditi3): Call <su>muldi3_highpart.
1637 (<u>muldi3_highpart): Rename to <su>muldi3_highpart.
1638 (<u>mulsidi3): Call <su>mulsi3_highpart.
1639 (<u>mulsi3_highpart): Rename to <su>mulsi3_highpart.
1641 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
1643 * config/darwin.h (DSYMUTIL_SPEC): Recognize D sources.
1645 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
1647 * config/rs6000/darwin.h (FIXED_R13): Add for PPC64.
1648 (FIRST_SAVED_GP_REGNO): Save from R13 even when it is one
1651 2021-09-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1653 * config/aarch64/aarch64.h (AARCH64_FL_LS64): Define
1654 (AARCH64_FL_V8_7): Likewise.
1655 (AARCH64_FL_FOR_ARCH8_7): Likewise.
1656 * config/aarch64/aarch64-arches.def (armv8.7-a): Define.
1657 * config/aarch64/aarch64-option-extensions.def (ls64): Define.
1658 * doc/invoke.texi: Document the above.
1660 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
1662 * dbgcnt.c (dbg_cnt_counter): New.
1663 * dbgcnt.h (dbg_cnt_counter): New.
1664 * dumpfile.c (dump_options): Add entry for TDF_THREADING.
1665 * dumpfile.h (enum dump_flag): Add TDF_THREADING.
1666 * gimple-range-path.cc (DEBUG_SOLVER): Use TDF_THREADING.
1667 * tree-ssa-threadupdate.c (dump_jump_thread_path): Dump out
1670 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
1672 * cfgcleanup.c (pass_jump::execute): Check
1673 flag_expensive_optimizations.
1674 (pass_jump_after_combine::gate): Same.
1675 * doc/invoke.texi (-fthread-jumps): Enable for -O1.
1676 * opts.c (default_options_table): Enable -fthread-jumps at -O1.
1677 * tree-ssa-threadupdate.c
1678 (fwd_jt_path_registry::remove_jump_threads_including): Bail unless
1681 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
1683 * tree-ssa-reassoc.c (biased_names): New global.
1684 (propagate_bias_p): New function.
1685 (loop_carried_phi): Remove.
1686 (propagate_rank): Propagate bias along single uses.
1687 (get_rank): Update biased_names when needed.
1689 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
1691 * passes.def (pass_reassoc): Rename parameter to early_p.
1692 * tree-ssa-reassoc.c (reassoc_bias_loop_carried_phi_ranks_p):
1694 (phi_rank): Don't bias loop-carried phi ranks
1695 before vectorization pass.
1696 (execute_reassoc): Add bias_loop_carried_phi_ranks_p parameter.
1697 (pass_reassoc::pass_reassoc): Add bias_loop_carried_phi_ranks_p
1699 (pass_reassoc::set_param): Set bias_loop_carried_phi_ranks_p
1701 (pass_reassoc::execute): Pass bias_loop_carried_phi_ranks_p to
1703 (pass_reassoc::bias_loop_carried_phi_ranks_p): New member.
1705 2021-09-28 Jakub Jelinek <jakub@redhat.com>
1708 * config/i386/i386.c (standard_80387_constant_p): Don't recognize
1709 special 80387 instruction XFmode constants if flag_rounding_math.
1711 2021-09-28 Richard Biener <rguenther@suse.de>
1713 PR tree-optimization/100112
1714 * tree-ssa-sccvn.c (visit_reference_op_load): Record the
1715 referece into the hashtable twice in case last_vuse is
1716 different from the original vuse on the stmt.
1718 2021-09-28 Jakub Jelinek <jakub@redhat.com>
1720 PR middle-end/102492
1721 * gimplify.c (gimplify_adjust_omp_clauses_1): Don't call the
1722 omp_finish_clause langhook on implicitly added OMP_CLAUSE_PRIVATE
1723 clauses on SIMD constructs.
1725 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
1727 PR tree-optimization/102511
1728 * gimple-range-path.cc (path_range_query::range_on_path_entry):
1729 Return VARYING when nothing found.
1731 2021-09-28 Hongyu Wang <hongyu.wang@intel.com>
1734 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): Add
1736 (VALID_SSE2_REG_VHF_MODE): Add V4HFmode and V2HFmode.
1737 (VALID_MMX_REG_MODE): Add V4HFmode.
1738 (SSE_REG_MODE_P): Replace VALID_AVX512FP16_REG_MODE with
1739 vector mode condition.
1740 * config/i386/i386.c (classify_argument): Parse V4HF/V2HF
1742 (function_arg_32): Add V4HFmode.
1743 (function_arg_advance_32): Likewise.
1744 * config/i386/i386.md (mode): Add V4HF/V2HF.
1745 (MODE_SIZE): Likewise.
1746 * config/i386/mmx.md (MMXMODE): Add V4HF mode.
1747 (V_32): Add V2HF mode.
1748 (VHF_32_64): New mode iterator.
1749 (*mov<mode>_internal): Adjust sse alternatives to support
1751 (*mov<mode>_internal): Adjust sse alternatives to support
1753 (<insn><mode>3): New define_insn for add/sub/mul/div.
1755 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
1757 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Check
1759 (pass_early_thread_jumps::gate): Same.
1760 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
1761 Return if !flag_thread_jumps.
1762 * tree-ssa-threadupdate.c
1763 (jt_path_registry::register_jump_thread): Assert that
1764 flag_thread_jumps is true.
1766 2021-09-28 liuhongt <hongtao.liu@intel.com>
1769 (simplify_context::simplify_binary_operation_1): Relax
1770 condition of simplifying (vec_concat:M (vec_select op0
1771 index0)(vec_select op1 index1)) to allow different modes
1772 between op0 and M, but have same inner mode.
1774 2021-09-28 liuhongt <hongtao.liu@intel.com>
1776 * config/i386/i386-expand.c (emit_reduc_half): Handle
1777 V8HF/V16HF/V32HFmode.
1778 * config/i386/sse.md (REDUC_SSE_PLUS_MODE): Add V8HF.
1779 (REDUC_SSE_SMINMAX_MODE): Ditto.
1780 (REDUC_PLUS_MODE): Add V16HF and V32HF.
1781 (REDUC_SMINMAX_MODE): Ditto.
1783 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
1785 * gimple-range-path.cc
1786 (path_range_query::precompute_ranges_in_block): Rename to...
1787 (path_range_query::compute_ranges_in_block): ...this.
1788 (path_range_query::precompute_ranges): Rename to...
1789 (path_range_query::compute_ranges): ...this.
1790 (path_range_query::precompute_relations): Rename to...
1791 (path_range_query::compute_relations): ...this.
1792 (path_range_query::precompute_phi_relations): Rename to...
1793 (path_range_query::compute_phi_relations): ...this.
1794 * gimple-range-path.h: Rename precompute* to compute*.
1795 * tree-ssa-threadbackward.c
1796 (back_threader::find_taken_edge_switch): Same.
1797 (back_threader::find_taken_edge_cond): Same.
1798 * tree-ssa-threadedge.c
1799 (hybrid_jt_simplifier::compute_ranges_from_state): Same.
1800 (hybrid_jt_state::register_equivs_stmt): Inline...
1801 * tree-ssa-threadedge.h: ...here.
1803 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
1805 * tree-vrp.c (lhs_of_dominating_assert): Remove.
1806 (class vrp_jt_state): Remove.
1807 (class vrp_jt_simplifier): Remove.
1808 (vrp_jt_simplifier::simplify): Remove.
1809 (class vrp_jump_threader): Remove.
1810 (vrp_jump_threader::vrp_jump_threader): Remove.
1811 (vrp_jump_threader::~vrp_jump_threader): Remove.
1812 (vrp_jump_threader::before_dom_children): Remove.
1813 (vrp_jump_threader::after_dom_children): Remove.
1815 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
1817 * passes.def (pass_vrp_threader): New.
1818 * tree-pass.h (make_pass_vrp_threader): Add make_pass_vrp_threader.
1819 * tree-ssa-threadedge.c (hybrid_jt_state::register_equivs_stmt): New.
1820 (hybrid_jt_simplifier::hybrid_jt_simplifier): New.
1821 (hybrid_jt_simplifier::simplify): New.
1822 (hybrid_jt_simplifier::compute_ranges_from_state): New.
1823 * tree-ssa-threadedge.h (class hybrid_jt_state): New.
1824 (class hybrid_jt_simplifier): New.
1825 * tree-vrp.c (execute_vrp): Remove ASSERT_EXPR based jump
1827 (class hybrid_threader): New.
1828 (hybrid_threader::hybrid_threader): New.
1829 (hybrid_threader::~hybrid_threader): New.
1830 (hybrid_threader::before_dom_children): New.
1831 (hybrid_threader::after_dom_children): New.
1832 (execute_vrp_threader): New.
1833 (class pass_vrp_threader): New.
1834 (make_pass_vrp_threader): New.
1836 2021-09-27 Martin Liska <mliska@suse.cz>
1838 * output.h (enum section_flag): New.
1839 (SECTION_FORGET): Remove.
1840 (SECTION_ENTSIZE): Make it (1UL << 8) - 1.
1841 (SECTION_STYLE_MASK): Define it based on other enum
1843 * varasm.c (switch_to_section): Remove unused handling of
1846 2021-09-27 Martin Liska <mliska@suse.cz>
1848 * common.opt: Add new variable flag_default_complex_method.
1849 * opts.c (finish_options): Handle flags related to
1850 x_flag_complex_method.
1851 * toplev.c (process_options): Remove option handling related
1852 to flag_complex_method.
1854 2021-09-27 Richard Biener <rguenther@suse.de>
1856 PR middle-end/102450
1857 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid using
1858 type_for_size, instead use int_mode_for_size.
1860 2021-09-27 Andrew Pinski <apinski@marvell.com>
1863 * gimplify.c (gimplify_save_expr): Return early
1864 if the type of val is error_mark_node.
1866 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
1868 * tree-ssanames.c (ssa_name_has_boolean_range): Use
1871 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
1873 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
1874 vrp_visit_cond_stmt.
1875 * tree-ssa-dom.c (cprop_operand): Convert to range_query API.
1876 (cprop_into_stmt): Same.
1877 (dom_opt_dom_walker::optimize_stmt): Same.
1879 2021-09-27 Richard Biener <rguenther@suse.de>
1881 PR tree-optimization/97351
1882 PR tree-optimization/97352
1883 PR tree-optimization/82426
1884 * tree-vectorizer.h (dr_misalignment): Add vector type
1886 (aligned_access_p): Likewise.
1887 (known_alignment_for_access_p): Likewise.
1888 (vect_supportable_dr_alignment): Likewise.
1889 (vect_known_alignment_in_bytes): Likewise. Refactor.
1890 (DR_MISALIGNMENT): Remove.
1891 (vect_update_shared_vectype): Likewise.
1892 * tree-vect-data-refs.c (dr_misalignment): Refactor, handle
1893 a vector type with larger alignment requirement and apply
1894 the negative step adjustment here.
1895 (vect_calculate_target_alignment): Remove.
1896 (vect_compute_data_ref_alignment): Get explicit vector type
1897 argument, do not apply a negative step alignment adjustment
1899 (vect_slp_analyze_node_alignment): Re-analyze alignment
1900 when we re-visit the DR with a bigger desired alignment but
1901 keep more precise results from smaller alignments.
1902 * tree-vect-slp.c (vect_update_shared_vectype): Remove.
1903 (vect_slp_analyze_node_operations_1): Do not update the
1904 shared vector type on stmts.
1905 * tree-vect-stmts.c (vect_analyze_stmt): Push/pop the
1906 vector type of an SLP node to the representative stmt-info.
1907 (vect_transform_stmt): Likewise.
1909 2021-09-27 liuhongt <hongtao.liu@intel.com>
1912 2021-09-09 liuhongt <hongtao.liu@intel.com>
1915 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
1916 (reduc_plus_scal_v4sf): .. this, New define_expand.
1917 (reduc_plus_scal_v2df): .. and this, New define_expand.
1919 2021-09-26 liuhongt <hongtao.liu@intel.com>
1921 * doc/extend.texi (Half-Precision): Remove storage only
1922 description for _Float16 w/o avx512fp16.
1924 2021-09-25 Dimitar Dimitrov <dimitar@dinux.eu>
1926 * config/pru/constraints.md (Rrio): New constraint.
1927 * config/pru/predicates.md (regio_operand): New predicate.
1928 * config/pru/pru-pragma.c (pru_register_pragmas): Register
1929 the __regio_symbol address space.
1930 * config/pru/pru-protos.h (pru_symref2ioregno): Declaration.
1931 * config/pru/pru.c (pru_symref2ioregno): New helper function.
1932 (pru_legitimate_address_p): Remove.
1933 (pru_addr_space_legitimate_address_p): Use the address space
1935 (pru_nongeneric_pointer_addrspace): New helper function.
1936 (pru_insert_attributes): New function to validate __regio_symbol
1938 (TARGET_INSERT_ATTRIBUTES): New macro.
1939 (TARGET_LEGITIMATE_ADDRESS_P): Remove.
1940 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): New macro.
1941 * config/pru/pru.h (enum reg_class): Add REGIO_REGS class.
1942 * config/pru/pru.md (*regio_readsi): New pattern to read I/O
1944 (*regio_nozext_writesi): New pattern to write to I/O registers.
1945 (*regio_zext_write_r30<EQS0:mode>): Ditto.
1946 * doc/extend.texi: Document the new PRU Named Address Space.
1948 2021-09-24 Patrick Palka <ppalka@redhat.com>
1952 * real.c (encode_ieee_double): Avoid unwanted sign extension.
1953 (encode_ieee_quad): Likewise.
1955 2021-09-24 Vladimir Makarov <vmakarov@redhat.com>
1957 PR rtl-optimization/102147
1958 * ira-build.c (ira_conflict_vector_profitable_p): Make
1959 profitability calculation independent of host compiler pointer and
1962 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
1964 * gimple-range-path.cc (path_range_query::path_range_query):
1965 Move debugging header...
1966 (path_range_query::precompute_ranges): ...here.
1967 (path_range_query::internal_range_of_expr): Do not call
1968 range_on_path_entry if NAME is defined in the current block.
1970 2021-09-24 Richard Biener <rguenther@suse.de>
1972 * cfghooks.c (verify_flow_info): Verify unallocated BB and
1973 edge flags are not set.
1975 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
1977 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
1979 (jt_path_registry::register_jump_thread): Call
1980 cancel_invalid_paths.
1981 * tree-ssa-threadupdate.h (class jt_path_registry): Add
1982 cancel_invalid_paths.
1984 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
1986 PR tree-optimization/102400
1987 * tree-ssa-sccvn.c (vn_reference_insert_pieces): Initialize
1988 result_vdef to zero value.
1990 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
1992 PR tree-optimization/102451
1993 * tree-ssa-dse.c (delete_dead_or_redundant_call): Record bb of stmt
1996 2021-09-24 Hongyu Wang <hongyu.wang@intel.com>
1998 * config/i386/sse.md (cond_<insn><mode>): Extend to support
2000 (cond_mul<mode>): Likewise.
2001 (cond_div<mode>): Likewise.
2002 (cond_<code><mode>): Likewise.
2003 (cond_fma<mode>): Likewise.
2004 (cond_fms<mode>): Likewise.
2005 (cond_fnma<mode>): Likewise.
2006 (cond_fnms<mode>): Likewise.
2008 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
2010 PR tree-optimization/102463
2011 * gimple-range-fold.cc (fold_using_range::relation_fold_and_or): If
2012 there is no range-ops handler, don't look for a relation.
2014 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
2016 * gimple-range-cache.cc (ranger_cache::ranger_cache): Take
2017 non-executable_edge flag as parameter.
2018 * gimple-range-cache.h (ranger_cache): Adjust prototype.
2019 * gimple-range-gori.cc (gori_compute::gori_compute): Take
2020 non-executable_edge flag as parameter.
2021 (gori_compute::outgoing_edge_range_p): Check new flag.
2022 * gimple-range-gori.h (gori_compute): Adjust prototype.
2023 * gimple-range.cc (gimple_ranger::gimple_ranger): Create new flag.
2024 (gimple_ranger::range_on_edge): Check new flag.
2025 * gimple-range.h (gimple_ranger::non_executable_edge_flag): New.
2026 * gimple-ssa-evrp.c (rvrp_folder): Pass ranger flag to simplifer.
2027 (hybrid_folder::hybrid_folder): Set ranger non-executable flag value.
2028 (hybrid_folder::fold_stmt): Set flag value in the simplifer.
2029 * vr-values.c (simplify_using_ranges::set_and_propagate_unexecutable):
2030 Use not_executable flag if provided inmstead of EDGE_EXECUTABLE.
2031 (simplify_using_ranges::simplify_switch_using_ranges): Clear
2032 EDGE_EXECUTABLE like it originally did.
2033 (simplify_using_ranges::cleanup_edges_and_switches): Clear any
2034 NON_EXECUTABLE flags.
2035 (simplify_using_ranges::simplify_using_ranges): Adjust.
2036 * vr-values.h (class simplify_using_ranges): Adjust.
2037 (simplify_using_ranges::set_range_query): Add non-executable flag param.
2039 2021-09-23 Bill Schmidt <wschmidt@linux.ibm.com>
2042 * config/rs6000/rs6000-call.c (rs6000_aggregate_candidate): Detect
2043 zero-width bit fields and return indicator.
2044 (rs6000_discover_homogeneous_aggregate): Diagnose when the
2045 presence of a zero-width bit field changes parameter passing in
2048 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
2050 * gimple-range-fold.cc (fold_using_range::range_of_phi):
2051 Remove dominator check.
2053 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
2055 * gimple-range-path.cc (path_range_query::precompute_relations):
2056 Hoist edge calculations before using EDGE_SUCC.
2058 2021-09-23 Jonathan Wakely <jwakely@redhat.com>
2060 * configure.ac: Fix --with-multilib-list description.
2061 * configure: Regenerate.
2063 2021-09-23 Richard Biener <rguenther@suse.de>
2065 PR tree-optimization/102448
2066 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
2067 Clear alignment info copied from DR_PTR_INFO.
2069 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
2071 * config/i386/i386-expand.c (ix86_use_mask_cmp_p): Enable
2073 * config/i386/sse.md (sseintvecmodelower): Add HF vector modes.
2074 (<avx512>_store<mode>_mask): Extend to support HF vector modes.
2075 (vec_cmp<mode><avx512fmaskmodelower>): Likewise.
2076 (vcond_mask_<mode><avx512fmaskmodelower>): Likewise.
2077 (vcond<mode><mode>): New expander.
2078 (vcond<mode><sseintvecmodelower>): Likewise.
2079 (vcond<sseintvecmodelower><mode>): Likewise.
2080 (vcondu<mode><sseintvecmodelower>): Likewise.
2082 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
2084 * config/i386/sse.md (extend<ssePHmodelower><mode>2):
2086 (extendv4hf<mode>2): Likewise.
2087 (extendv2hfv2df2): Likewise.
2088 (trunc<mode><ssePHmodelower>2): Likewise.
2089 (avx512fp16_vcvt<castmode>2ph_<mode>): Rename to ...
2090 (trunc<mode>v4hf2): ... this, and drop constraints.
2091 (avx512fp16_vcvtpd2ph_v2df): Rename to ...
2092 (truncv2dfv2hf2): ... this, and likewise.
2094 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
2096 * config/i386/sse.md (float<floatunssuffix><mode><ssePHmodelower>2):
2098 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>):
2100 (float<floatunssuffix><mode>v4hf2): ... this, and drop constraints.
2101 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Rename to ...
2102 (float<floatunssuffix>v2div2hf2): ... this, and likewise.
2104 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
2106 * config/i386/i386.md (fix<fixunssuffix>_trunchf<mode>2): New expander.
2107 (fixuns_trunchfhi2): Likewise.
2108 (*fixuns_trunchfsi2zext): New define_insn.
2109 * config/i386/sse.md (ssePHmodelower): New mode_attr.
2110 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
2111 New expander for same element vector fix_truncate.
2112 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
2113 Likewise for V4HF to V4SI/V4DI fix_truncate.
2114 (fix<fixunssuffix>_truncv2hfv2di2):
2115 Likeise for V2HF to V2DI fix_truncate.
2117 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
2119 * config/i386/i386.md (<code>hf3): New expander.
2121 2021-09-23 liuhongt <hongtao.liu@intel.com>
2123 * config/i386/sse.md (FMAMODEM): extend to handle FP16.
2124 (VFH_SF_AVX512VL): Extend to handle HFmode.
2125 (VF_SF_AVX512VL): Deleted.
2127 2021-09-23 liuhongt <hongtao.liu@intel.com>
2129 * config/i386/i386.md (rinthf2): New expander.
2130 (nearbyinthf2): New expander.
2132 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
2134 * tree-ssa-dom.c (class dom_jump_threader_simplifier): Rename...
2135 (class dom_jt_state): ...this and provide virtual overrides.
2136 (dom_jt_state::register_equiv): New.
2137 (class dom_jt_simplifier): Rename from
2138 dom_jump_threader_simplifier.
2139 (dom_jump_threader_simplifier::simplify): Rename...
2140 (dom_jt_simplifier::simplify): ...to this.
2141 (pass_dominator::execute): Use dom_jt_simplifier and
2143 * tree-ssa-threadedge.c (jump_threader::jump_threader):
2145 (jt_state::register_equivs_stmt): Abstract out...
2146 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
2148 (jump_threader::thread_around_empty_blocks): Update state.
2149 (jump_threader::thread_through_normal_block): Same.
2150 (jt_state::jt_state): Remove.
2151 (jt_state::push): Remove pass specific bits. Keep block vector
2153 (jt_state::append_path): New.
2154 (jt_state::pop): Remove pass specific bits.
2155 (jt_state::register_equiv): Same.
2156 (jt_state::record_ranges_from_stmt): Same.
2157 (jt_state::register_equivs_on_edge): Same. Rename...
2158 (jt_state::register_equivs_edge): ...to this.
2159 (jt_state::dump): New.
2160 (jt_state::debug): New.
2161 (jump_threader_simplifier::simplify): Remove.
2162 (jt_state::get_path): New.
2163 * tree-ssa-threadedge.h (class jt_simplifier): Make into a base
2164 class. Expose common functionality as virtual methods.
2165 (class jump_threader_simplifier): Same. Rename...
2166 (class jt_simplifier): ...to this.
2167 * tree-vrp.c (class vrp_jump_threader_simplifier): Rename...
2168 (class vrp_jt_simplifier): ...to this. Provide pass specific
2170 (class vrp_jt_state): New.
2171 (vrp_jump_threader_simplifier::simplify): Rename...
2172 (vrp_jt_simplifier::simplify): ...to this. Inline code from
2173 what used to be the base class.
2174 (vrp_jump_threader::vrp_jump_threader): Use vrp_jt_state and
2177 2021-09-22 Tobias Burnus <tobias@codesourcery.com>
2180 * doc/invoke.texi (-Wno-missing-include-dirs.): Document Fortran
2183 2021-09-22 Roger Sayle <roger@nextmovesoftware.com>
2184 Richard Biener <rguenther@suse.de>
2186 * match.pd (negation simplifications): Implement some negation
2187 folding transformations from fold-const.c's fold_negate_expr.
2188 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Add a SIMPLIFY
2189 argument, to control whether the op should be simplified prior
2190 to looking up/assigning a value number.
2191 (vn_nary_build_or_lookup): Update call to vn_nary_build_or_lookup_1.
2192 (vn_nary_simplify): Likewise.
2193 (visit_nary_op): Likewise, but when constructing a NEGATE_EXPR
2194 now call vn_nary_build_or_lookup_1 disabling simplification.
2196 2021-09-22 Jiufu Guo <guojiufu@linux.ibm.com>
2198 PR tree-optimization/102087
2199 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
2200 Update bound/cmp/control for niter.
2202 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
2204 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
2205 Move check for non-empty BB here.
2206 (fur_source::register_outgoing_edges): ...from here.
2208 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
2210 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
2211 Remove call to improve_range_with_equivs.
2212 (path_range_query::improve_range_with_equivs): Remove
2213 * gimple-range-path.h: Remove improve_range_with_equivs.
2215 2021-09-22 dianhong xu <dianhong.xu@intel.com>
2217 * config/i386/avx512fp16intrin.h:
2218 (_mm512_mask_blend_ph): New intrinsic.
2219 (_mm512_permutex2var_ph): Ditto.
2220 (_mm512_permutexvar_ph): Ditto.
2221 * config/i386/avx512fp16vlintrin.h:
2222 (_mm256_mask_blend_ph): New intrinsic.
2223 (_mm256_permutex2var_ph): Ditto.
2224 (_mm256_permutexvar_ph): Ditto.
2225 (_mm_mask_blend_ph): Ditto.
2226 (_mm_permutex2var_ph): Ditto.
2227 (_mm_permutexvar_ph): Ditto.
2229 2021-09-22 dianhong xu <dianhong.xu@intel.com>
2231 * config/i386/avx512fp16intrin.h: Add new intrinsics.
2232 (_mm512_conj_pch): New intrinsic.
2233 (_mm512_mask_conj_pch): Ditto.
2234 (_mm512_maskz_conj_pch): Ditto.
2235 * config/i386/avx512fp16vlintrin.h: Add new intrinsics.
2236 (_mm256_conj_pch): New intrinsic.
2237 (_mm256_mask_conj_pch): Ditto.
2238 (_mm256_maskz_conj_pch): Ditto.
2239 (_mm_conj_pch): Ditto.
2240 (_mm_mask_conj_pch): Ditto.
2241 (_mm_maskz_conj_pch): Ditto.
2243 2021-09-22 dianhong xu <dianhong.xu@intel.com>
2245 * config/i386/avx512fp16intrin.h (_MM512_REDUCE_OP): New macro
2246 (_mm512_reduce_add_ph): New intrinsic.
2247 (_mm512_reduce_mul_ph): Ditto.
2248 (_mm512_reduce_min_ph): Ditto.
2249 (_mm512_reduce_max_ph): Ditto.
2250 * config/i386/avx512fp16vlintrin.h
2251 (_MM256_REDUCE_OP/_MM_REDUCE_OP): New macro.
2252 (_mm256_reduce_add_ph): New intrinsic.
2253 (_mm256_reduce_mul_ph): Ditto.
2254 (_mm256_reduce_min_ph): Ditto.
2255 (_mm256_reduce_max_ph): Ditto.
2256 (_mm_reduce_add_ph): Ditto.
2257 (_mm_reduce_mul_ph): Ditto.
2258 (_mm_reduce_min_ph): Ditto.
2259 (_mm_reduce_max_ph): Ditto.
2261 2021-09-22 dianhong xu <dianhong.xu@intel.com>
2263 * config/i386/avx512fp16intrin.h (__m512h_u, __m256h_u,
2264 __m128h_u): New typedef.
2265 (_mm512_load_ph): New intrinsic.
2266 (_mm256_load_ph): Ditto.
2267 (_mm_load_ph): Ditto.
2268 (_mm512_loadu_ph): Ditto.
2269 (_mm256_loadu_ph): Ditto.
2270 (_mm_loadu_ph): Ditto.
2271 (_mm512_store_ph): Ditto.
2272 (_mm256_store_ph): Ditto.
2273 (_mm_store_ph): Ditto.
2274 (_mm512_storeu_ph): Ditto.
2275 (_mm256_storeu_ph): Ditto.
2276 (_mm_storeu_ph): Ditto.
2277 (_mm512_abs_ph): Ditto.
2278 * config/i386/avx512fp16vlintrin.h
2279 (_mm_abs_ph): Ditto.
2280 (_mm256_abs_ph): Ditto.
2282 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
2284 * config/s390/tpf.md (prologue_tpf, epilogue_tpf): Add cc clobber.
2286 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
2289 * config/s390/s390.c (s390_expand_insv): Emit a normal move if it
2290 is actually a full copy of the source operand into the target.
2291 Don't emit a strict low part move if source and target mode match.
2293 2021-09-22 Jakub Jelinek <jakub@redhat.com>
2295 PR middle-end/102415
2296 * omp-expand.c (expand_omp_single): If region->exit is NULL,
2297 assert region->entry is GIMPLE_OMP_SCOPE region and return.
2299 2021-09-22 Jakub Jelinek <jakub@redhat.com>
2301 * tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
2302 * tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
2303 arguments from 2 to 3.
2304 * tree-pretty-print.c (dump_omp_clause): Print allocator() around
2305 allocate clause allocator and print align if present.
2306 * omp-low.c (scan_sharing_clauses): Force allocate_map entry even
2307 for omp_default_mem_alloc if align modifier is present. If align
2308 modifier is present, use TREE_LIST to encode both allocator and
2310 (lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
2311 Handle align modifier on allocator clause if present.
2313 2021-09-22 liuhongt <hongtao.liu@intel.com>
2315 * config/i386/i386.md (define_attr "isa"): Add
2317 (define_attr "enabled"): Correspond fma_or_avx512vl to
2318 TARGET_FMA || TARGET_AVX512VL.
2319 * config/i386/mmx.md (fmav2sf4): Extend to AVX512 fma.
2324 2021-09-22 liuhongt <hongtao.liu@intel.com>
2326 * config/i386/i386.md (cstorehf3): New define_expand.
2328 2021-09-22 liuhongt <hongtao.liu@intel.com>
2330 * config/i386/i386.md (<rounding_insn>hf2): New expander.
2331 (sse4_1_round<mode>2): Extend from MODEF to MODEFH.
2332 * config/i386/sse.md (*sse4_1_round<ssescalarmodesuffix>):
2333 Extend from VF_128 to VFH_128.
2335 2021-09-22 liuhongt <hongtao.liu@intel.com>
2337 * config/i386/i386-features.c (i386-features.c): Handle
2339 * config/i386/i386.md (sqrthf2): New expander.
2340 (*sqrthf2): New define_insn.
2341 * config/i386/sse.md
2342 (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
2345 2021-09-22 liuhongt <hongtao.liu@intel.com>
2347 * config/i386/avx512fp16intrin.h (_mm_mask_fcmadd_sch):
2349 (_mm_mask3_fcmadd_sch): Likewise.
2350 (_mm_maskz_fcmadd_sch): Likewise.
2351 (_mm_fcmadd_sch): Likewise.
2352 (_mm_mask_fmadd_sch): Likewise.
2353 (_mm_mask3_fmadd_sch): Likewise.
2354 (_mm_maskz_fmadd_sch): Likewise.
2355 (_mm_fmadd_sch): Likewise.
2356 (_mm_mask_fcmadd_round_sch): Likewise.
2357 (_mm_mask3_fcmadd_round_sch): Likewise.
2358 (_mm_maskz_fcmadd_round_sch): Likewise.
2359 (_mm_fcmadd_round_sch): Likewise.
2360 (_mm_mask_fmadd_round_sch): Likewise.
2361 (_mm_mask3_fmadd_round_sch): Likewise.
2362 (_mm_maskz_fmadd_round_sch): Likewise.
2363 (_mm_fmadd_round_sch): Likewise.
2364 (_mm_fcmul_sch): Likewise.
2365 (_mm_mask_fcmul_sch): Likewise.
2366 (_mm_maskz_fcmul_sch): Likewise.
2367 (_mm_fmul_sch): Likewise.
2368 (_mm_mask_fmul_sch): Likewise.
2369 (_mm_maskz_fmul_sch): Likewise.
2370 (_mm_fcmul_round_sch): Likewise.
2371 (_mm_mask_fcmul_round_sch): Likewise.
2372 (_mm_maskz_fcmul_round_sch): Likewise.
2373 (_mm_fmul_round_sch): Likewise.
2374 (_mm_mask_fmul_round_sch): Likewise.
2375 (_mm_maskz_fmul_round_sch): Likewise.
2376 * config/i386/i386-builtin.def: Add corresponding new builtins.
2377 * config/i386/sse.md
2378 (avx512fp16_fmaddcsh_v8hf_maskz<round_expand_name>): New expander.
2379 (avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>): Ditto.
2380 (avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>):
2382 (avx512fp16_<complexopname>sh_v8hf_mask<round_name>): Ditto.
2383 (avx512fp16_<complexopname>sh_v8hf<mask_scalarc_name><round_scalarcz_name>):
2385 * config/i386/subst.md (mask_scalarcz_name): New.
2386 (mask_scalarc_name): Ditto.
2387 (mask_scalarc_operand3): Ditto.
2388 (mask_scalarcz_operand4): Ditto.
2389 (round_scalarcz_name): Ditto.
2390 (round_scalarc_mask_operand3): Ditto.
2391 (round_scalarcz_mask_operand4): Ditto.
2392 (round_scalarc_mask_op3): Ditto.
2393 (round_scalarcz_mask_op4): Ditto.
2394 (round_scalarcz_constraint): Ditto.
2395 (round_scalarcz_nimm_predicate): Ditto.
2396 (mask_scalarcz): Ditto.
2397 (mask_scalarc): Ditto.
2398 (round_scalarcz): Ditto.
2400 2021-09-22 liuhongt <hongtao.liu@intel.com>
2402 * config/i386/avx512fp16intrin.h (_mm512_fcmadd_pch):
2404 (_mm512_mask_fcmadd_pch): Likewise.
2405 (_mm512_mask3_fcmadd_pch): Likewise.
2406 (_mm512_maskz_fcmadd_pch): Likewise.
2407 (_mm512_fmadd_pch): Likewise.
2408 (_mm512_mask_fmadd_pch): Likewise.
2409 (_mm512_mask3_fmadd_pch): Likewise.
2410 (_mm512_maskz_fmadd_pch): Likewise.
2411 (_mm512_fcmadd_round_pch): Likewise.
2412 (_mm512_mask_fcmadd_round_pch): Likewise.
2413 (_mm512_mask3_fcmadd_round_pch): Likewise.
2414 (_mm512_maskz_fcmadd_round_pch): Likewise.
2415 (_mm512_fmadd_round_pch): Likewise.
2416 (_mm512_mask_fmadd_round_pch): Likewise.
2417 (_mm512_mask3_fmadd_round_pch): Likewise.
2418 (_mm512_maskz_fmadd_round_pch): Likewise.
2419 (_mm512_fcmul_pch): Likewise.
2420 (_mm512_mask_fcmul_pch): Likewise.
2421 (_mm512_maskz_fcmul_pch): Likewise.
2422 (_mm512_fmul_pch): Likewise.
2423 (_mm512_mask_fmul_pch): Likewise.
2424 (_mm512_maskz_fmul_pch): Likewise.
2425 (_mm512_fcmul_round_pch): Likewise.
2426 (_mm512_mask_fcmul_round_pch): Likewise.
2427 (_mm512_maskz_fcmul_round_pch): Likewise.
2428 (_mm512_fmul_round_pch): Likewise.
2429 (_mm512_mask_fmul_round_pch): Likewise.
2430 (_mm512_maskz_fmul_round_pch): Likewise.
2431 * config/i386/avx512fp16vlintrin.h (_mm_fmadd_pch):
2433 (_mm_mask_fmadd_pch): Likewise.
2434 (_mm_mask3_fmadd_pch): Likewise.
2435 (_mm_maskz_fmadd_pch): Likewise.
2436 (_mm256_fmadd_pch): Likewise.
2437 (_mm256_mask_fmadd_pch): Likewise.
2438 (_mm256_mask3_fmadd_pch): Likewise.
2439 (_mm256_maskz_fmadd_pch): Likewise.
2440 (_mm_fcmadd_pch): Likewise.
2441 (_mm_mask_fcmadd_pch): Likewise.
2442 (_mm_mask3_fcmadd_pch): Likewise.
2443 (_mm_maskz_fcmadd_pch): Likewise.
2444 (_mm256_fcmadd_pch): Likewise.
2445 (_mm256_mask_fcmadd_pch): Likewise.
2446 (_mm256_mask3_fcmadd_pch): Likewise.
2447 (_mm256_maskz_fcmadd_pch): Likewise.
2448 (_mm_fmul_pch): Likewise.
2449 (_mm_mask_fmul_pch): Likewise.
2450 (_mm_maskz_fmul_pch): Likewise.
2451 (_mm256_fmul_pch): Likewise.
2452 (_mm256_mask_fmul_pch): Likewise.
2453 (_mm256_maskz_fmul_pch): Likewise.
2454 (_mm_fcmul_pch): Likewise.
2455 (_mm_mask_fcmul_pch): Likewise.
2456 (_mm_maskz_fcmul_pch): Likewise.
2457 (_mm256_fcmul_pch): Likewise.
2458 (_mm256_mask_fcmul_pch): Likewise.
2459 (_mm256_maskz_fcmul_pch): Likewise.
2460 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF_V8HF_V8HF,
2461 V8HF_FTYPE_V16HF_V16HF_V16HF, V16HF_FTYPE_V16HF_V16HF_V16HF_UQI,
2462 V32HF_FTYPE_V32HF_V32HF_V32HF_INT,
2463 V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT): Add new builtin types.
2464 * config/i386/i386-builtin.def: Add new builtins.
2465 * config/i386/i386-expand.c: Handle new builtin types.
2466 * config/i386/subst.md (SUBST_CV): New.
2467 (maskc_name): Ditto.
2468 (maskc_operand3): Ditto.
2470 (sdc_maskz_name): Ditto.
2471 (sdc_mask_op4): Ditto.
2472 (sdc_mask_op5): Ditto.
2473 (sdc_mask_mode512bit_condition): Ditto.
2475 (round_maskc_operand3): Ditto.
2476 (round_sdc_mask_operand4): Ditto.
2477 (round_maskc_op3): Ditto.
2478 (round_sdc_mask_op4): Ditto.
2479 (round_saeonly_sdc_mask_operand5): Ditto.
2480 * config/i386/sse.md (unspec): Add complex fma unspecs.
2481 (avx512fmaskcmode): New.
2482 (UNSPEC_COMPLEX_F_C_MA): Ditto.
2483 (UNSPEC_COMPLEX_F_C_MUL): Ditto.
2484 (complexopname): Ditto.
2485 (<avx512>_fmaddc_<mode>_maskz<round_expand_name>): New expander.
2486 (<avx512>_fcmaddc_<mode>_maskz<round_expand_name>): Ditto.
2487 (fma_<complexopname>_<mode><sdc_maskz_name><round_name>): New
2489 (<avx512>_<complexopname>_<mode>_mask<round_name>): Ditto.
2490 (<avx512>_<complexopname>_<mode><maskc_name><round_name>): Ditto.
2492 2021-09-22 Kewen Lin <linkw@linux.ibm.com>
2494 * config/rs6000/rs6000.opt (rs6000-density-pct-threshold,
2495 rs6000-density-size-threshold, rs6000-density-penalty,
2496 rs6000-density-load-pct-threshold,
2497 rs6000-density-load-num-threshold): New parameter.
2498 * config/rs6000/rs6000.c (rs6000_density_test): Adjust with
2499 corresponding parameters.
2501 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2503 * gimple-range-path.cc (path_range_query::defined_outside_path):
2505 (path_range_query::range_on_path_entry): New.
2506 (path_range_query::internal_range_of_expr): Resolve unknowns
2508 (path_range_query::improve_range_with_equivs): New.
2509 (path_range_query::ssa_range_in_phi): Resolve unknowns with
2511 * gimple-range-path.h (class path_range_query): Add
2512 defined_outside_path, range_on_path_entry, and
2513 improve_range_with_equivs.
2515 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2517 * gimple-range-path.cc (path_range_query::add_to_imports): New.
2518 (path_range_query::add_copies_to_imports): New.
2519 (path_range_query::precompute_ranges): Call
2520 add_copies_to_imports.
2521 * gimple-range-path.h (class path_range_query): Add prototypes
2522 for add_copies_to_imports and add_to_imports.
2524 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2526 * gimple-range-path.cc (path_range_query::range_defined_in_block):
2527 Remove useless code.
2529 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2531 * gimple-range-fold.h (class fur_source): Make oracle protected.
2532 * gimple-range-path.cc (path_range_query::path_range_query): Add
2533 resolve argument. Initialize oracle.
2534 (path_range_query::~path_range_query): Delete oracle.
2535 (path_range_query::range_of_stmt): Adapt to use relations.
2536 (path_range_query::precompute_ranges): Pre-compute relations.
2537 (class jt_fur_source): New
2538 (jt_fur_source::jt_fur_source): New.
2539 (jt_fur_source::register_relation): New.
2540 (jt_fur_source::query_relation): New.
2541 (path_range_query::precompute_relations): New.
2542 (path_range_query::precompute_phi_relations): New.
2543 * gimple-range-path.h (path_range_query): Add resolve argument.
2544 Add oracle, precompute_relations, precompute_phi_relations.
2545 * tree-ssa-threadbackward.c (back_threader::back_threader): Pass
2546 resolve argument to solver.
2548 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2550 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
2551 Rename postfold_gcond_edges to register_outgoing_edges and
2553 (fold_using_range::postfold_gcond_edges): Rename...
2554 (fur_source::register_outgoing_edges): ...to this.
2555 * gimple-range-fold.h (postfold_gcond_edges): Rename to
2556 register_outgoing_edges and move to fur_source.
2558 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2560 * gimple-range-fold.cc (fold_using_range::range_of_phi): Check
2561 dom_info_available_p.
2563 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2565 * gimple-range-cache.cc (non_null_ref::non_null_ref): Use create
2566 and quick_grow_cleared instead of safe_grow_cleared.
2568 2021-09-21 Thomas Schwinge <thomas@codesourcery.com>
2571 * omp-oacc-neuter-broadcast.cc (oacc_do_neutering): Evaluate
2574 2021-09-21 Richard Earnshaw <rearnsha@arm.com>
2576 * configure.ac: Detect when the assembler supports new-style
2577 architecture extensions.
2578 * common/config/arm/arm-common.c (arm_rewrite_mcpu): Return
2579 the full CPU string if the assembler can grok it.
2580 (arm_rewrite_march): Likewise but for the architecture.
2581 * config.in: Regenerate.
2582 * configure: Regenerate.
2584 2021-09-21 Richard Biener <rguenther@suse.de>
2586 PR tree-optimization/102421
2587 * tree-vect-loop.c (vect_dissolve_slp_only_groups): Copy and
2588 adjust alignment info.
2590 2021-09-21 Kewen Lin <linkw@linux.ibm.com>
2592 * ipa-fnsummary.c (ipa_fn_summary_write): Remove inconsistent
2593 bitfield stream out.
2595 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
2597 * gimple-range-fold.cc (fold_using_range::range_of_phi): Ignore
2598 undefined edges, apply an equivalence if appropriate.
2599 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Return
2600 UNDEFINED if EDGE_EXECUTABLE is not set.
2601 * gimple-range.cc (gimple_ranger::gimple_ranger): Set all edges
2602 as EXECUTABLE upon startup.
2603 (gimple_ranger::range_on_edge): Return UNDEFINED for edges without
2604 EDGE_EXECUTABLE set.
2605 * vr-values.c (set_and_propagate_unexecutable): New.
2606 (simplify_using_ranges::fold_cond): Call set_and_propagate.
2607 (simplify_using_ranges::simplify_switch_using_ranges): Ditto.
2608 * vr-values.h: Add prototype.
2610 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
2612 * value-relation.cc (equiv_oracle::register_initial_def): New.
2613 (equiv_oracle::register_relation): Call register_initial_def.
2614 (equiv_oracle::add_equiv_to_block): New. Split register_relation.
2615 (relation_oracle::register_stmt): Check def block of PHI arguments.
2616 * value-relation.h (equiv_oracle): Add new prototypes.
2618 2021-09-20 Matthias Kretz <m.kretz@gsi.de>
2620 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
2621 Define __RECIPROCAL_MATH__, __NO_SIGNED_ZEROS__,
2622 __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__, and
2623 __ROUNDING_MATH__ according to their corresponding flags.
2624 * doc/cpp.texi: Document __RECIPROCAL_MATH__,
2625 __NO_SIGNED_ZEROS__, __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__,
2626 and __ROUNDING_MATH__.
2628 2021-09-20 Richard Biener <rguenther@suse.de>
2630 * tree-vect-stmts.c (vectorizable_load): Use the vectype
2633 2021-09-20 Richard Biener <rguenther@suse.de>
2635 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
2636 Do not compute alignment of the vectorized access here.
2638 2021-09-20 Richard Biener <rguenther@suse.de>
2640 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
2641 Store -1 for runtime alias peeling iterations.
2643 2021-09-20 Richard Biener <rguenther@suse.de>
2645 * config.gcc: Obsolete hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11*.
2647 2021-09-20 Thomas Schwinge <thomas@codesourcery.com>
2649 * input.c (string_concat_db::record_string_concatenation)
2650 (string_concat_db::get_string_concatenation): Skip for
2651 'RESERVED_LOCATION_P'.
2653 2021-09-20 Richard Biener <rguenther@suse.de>
2655 PR tree-optimization/65206
2656 * tree-data-ref.h (struct data_reference): Add alt_indices,
2658 * tree-data-ref.c (free_data_ref): Release alt_indices.
2659 (dr_analyze_indices): Work on struct indices and get DR_REF as tree.
2660 (create_data_ref): Adjust.
2661 (initialize_data_dependence_relation): Split into head
2662 and tail. When the base objects fail to match up try
2663 again with pointer-based analysis of indices.
2664 * tree-vectorizer.c (vec_info_shared::check_datarefs): Do
2665 not compare the lazily computed alternate set of indices.
2667 2021-09-20 Iain Sandoe <iain@sandoe.co.uk>
2669 * gcc.c: Test for execute OK when we find the
2670 programs for assembler linker and dsymutil and those
2671 were specified at configure-time.
2673 2021-09-19 Martin Sebor <msebor@redhat.com>
2675 PR middle-end/102403
2676 * gimple-predicate-analysis.cc (predicate::init_from_control_deps):
2677 Correct a function pre/postcondition.
2679 2021-09-19 Martin Sebor <msebor@redhat.com>
2681 PR middle-end/102243
2682 * tree-ssa-strlen.c (get_range): Handle null cfun.
2684 2021-09-19 Iain Sandoe <iain@sandoe.co.uk>
2686 * config/darwin.h (LINK_COMMAND_SPEC_A): Use Darwin10
2687 unwinder shim as a convenience library.
2689 2021-09-19 Andrew Pinski <apinski@marvell.com>
2691 * doc/install.texi: Add note about
2692 binutils 2.35 is required for LTO usage.
2694 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
2696 * tree-ssa-threadbackward.c
2697 (back_threader_registry::register_path): Use push_edge.
2698 * tree-ssa-threadedge.c
2699 (jump_threader::thread_around_empty_blocks): Same.
2700 (jump_threader::thread_through_normal_block): Same.
2701 (jump_threader::thread_across_edge): Same. Also, use auto_bitmap.
2703 * tree-ssa-threadupdate.c
2704 (jt_path_registry::allocate_thread_edge): Remove.
2705 (jt_path_registry::push_edge): New.
2706 (dump_jump_thread_path): Make static.
2707 * tree-ssa-threadupdate.h (allocate_thread_edge): Remove.
2710 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
2712 * gimple-range-path.cc (path_range_query::path_range_query): Add
2714 (path_range_query::dump): Remove extern declaration of dump_ranger.
2715 * gimple-range-trace.cc (dump_ranger): Add DEBUG_FUNCTION marker.
2716 * gimple-range-trace.h (dump_ranger): Add prototype.
2718 2021-09-19 John Ericson <git@JohnEricson.me>
2720 * gcc.c (find_a_program): New function, factored out of...
2721 (find_a_file): Here.
2722 (execute): Use find_a_program when looking for programs rather
2725 2021-09-19 Matwey V. Kornilov <matwey.kornilov@gmail.com>
2727 * config/avr/avr-mcus.def: Add atmega324pb.
2728 * doc/avr-mmcu.texi: Corresponding changes.
2730 2021-09-19 Roger Sayle <roger@nextmovesoftware.com>
2733 * match.pd (cmp @0 REAL_CST@1): When @0 is also REAL_CST, apply
2734 the same transformations as to @1. For comparisons against NaN,
2735 don't check HONOR_SNANS but confirm that neither operand is a
2738 2021-09-19 Benjamin Peterson <benjamin@locrian.net>
2740 * attribs.c (make_unique_name): Delete.
2741 * attribs.h (make_unique_name): Delete.
2743 2021-09-19 Andrew Pinski <apinski@marvell.com>
2745 * lra-constraints.c (check_and_process_move): Assert
2746 that dclass and sclass are greater than or equal to NO_REGS.
2748 2021-09-18 Jakub Jelinek <jakub@redhat.com>
2750 * tree.h (OMP_CLAUSE_ORDER_UNCONSTRAINED): Define.
2751 * tree-pretty-print.c (dump_omp_clause): Print unconstrained:
2752 for OMP_CLAUSE_ORDER_UNCONSTRAINED.
2754 2021-09-18 liuhongt <hongtao.liu@intel.com>
2756 * config/i386/i386-features.c (remove_partial_avx_dependency):
2757 Restrict TARGET_USE_VECTOR_FP_CONVERTS and
2758 TARGET_USE_VECTOR_CONVERTS to conversion instructions only.
2760 2021-09-18 Jakub Jelinek <jakub@redhat.com>
2762 * gimplify.c (omp_default_clause): For C/C++ default({,first}private),
2763 if file/namespace scope variable doesn't have predetermined sharing,
2764 treat it as if there was default(none).
2766 2021-09-18 liuhongt <hongtao.liu@intel.com>
2768 * config/i386/avx512fp16intrin.h (_mm_fmadd_sh):
2770 (_mm_mask_fmadd_sh): Likewise.
2771 (_mm_mask3_fmadd_sh): Likewise.
2772 (_mm_maskz_fmadd_sh): Likewise.
2773 (_mm_fmadd_round_sh): Likewise.
2774 (_mm_mask_fmadd_round_sh): Likewise.
2775 (_mm_mask3_fmadd_round_sh): Likewise.
2776 (_mm_maskz_fmadd_round_sh): Likewise.
2777 (_mm_fnmadd_sh): Likewise.
2778 (_mm_mask_fnmadd_sh): Likewise.
2779 (_mm_mask3_fnmadd_sh): Likewise.
2780 (_mm_maskz_fnmadd_sh): Likewise.
2781 (_mm_fnmadd_round_sh): Likewise.
2782 (_mm_mask_fnmadd_round_sh): Likewise.
2783 (_mm_mask3_fnmadd_round_sh): Likewise.
2784 (_mm_maskz_fnmadd_round_sh): Likewise.
2785 (_mm_fmsub_sh): Likewise.
2786 (_mm_mask_fmsub_sh): Likewise.
2787 (_mm_mask3_fmsub_sh): Likewise.
2788 (_mm_maskz_fmsub_sh): Likewise.
2789 (_mm_fmsub_round_sh): Likewise.
2790 (_mm_mask_fmsub_round_sh): Likewise.
2791 (_mm_mask3_fmsub_round_sh): Likewise.
2792 (_mm_maskz_fmsub_round_sh): Likewise.
2793 (_mm_fnmsub_sh): Likewise.
2794 (_mm_mask_fnmsub_sh): Likewise.
2795 (_mm_mask3_fnmsub_sh): Likewise.
2796 (_mm_maskz_fnmsub_sh): Likewise.
2797 (_mm_fnmsub_round_sh): Likewise.
2798 (_mm_mask_fnmsub_round_sh): Likewise.
2799 (_mm_mask3_fnmsub_round_sh): Likewise.
2800 (_mm_maskz_fnmsub_round_sh): Likewise.
2801 * config/i386/i386-builtin-types.def
2802 (V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT): New builtin type.
2803 * config/i386/i386-builtin.def: Add new builtins.
2804 * config/i386/i386-expand.c: Handle new builtin type.
2805 * config/i386/sse.md (fmai_vmfmadd_<mode><round_name>):
2806 Ajdust to support FP16.
2807 (fmai_vmfmsub_<mode><round_name>): Ditto.
2808 (fmai_vmfnmadd_<mode><round_name>): Ditto.
2809 (fmai_vmfnmsub_<mode><round_name>): Ditto.
2810 (*fmai_fmadd_<mode>): Ditto.
2811 (*fmai_fmsub_<mode>): Ditto.
2812 (*fmai_fnmadd_<mode><round_name>): Ditto.
2813 (*fmai_fnmsub_<mode><round_name>): Ditto.
2814 (avx512f_vmfmadd_<mode>_mask<round_name>): Ditto.
2815 (avx512f_vmfmadd_<mode>_mask3<round_name>): Ditto.
2816 (avx512f_vmfmadd_<mode>_maskz<round_expand_name>): Ditto.
2817 (avx512f_vmfmadd_<mode>_maskz_1<round_name>): Ditto.
2818 (*avx512f_vmfmsub_<mode>_mask<round_name>): Ditto.
2819 (avx512f_vmfmsub_<mode>_mask3<round_name>): Ditto.
2820 (*avx512f_vmfmsub_<mode>_maskz_1<round_name>): Ditto.
2821 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
2822 (*avx512f_vmfnmsub_<mode>_mask3<round_name>): Ditto.
2823 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
2824 (*avx512f_vmfnmadd_<mode>_mask<round_name>): Renamed to ...
2825 (avx512f_vmfnmadd_<mode>_mask<round_name>) ... this, and
2826 adjust to support FP16.
2827 (avx512f_vmfnmadd_<mode>_mask3<round_name>): Ditto.
2828 (avx512f_vmfnmadd_<mode>_maskz_1<round_name>): Ditto.
2829 (avx512f_vmfnmadd_<mode>_maskz<round_expand_name>): New
2832 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
2834 * config/i386/sse.md (avx512fmaskmodelower): Extend to support
2836 (maskload<mode><avx512fmaskmodelower>): Ditto.
2837 (maskstore<mode><avx512fmaskmodelower>): Ditto.
2839 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
2841 * config/i386/i386-expand.c (ix86_expand_fp_absneg_operator):
2843 (ix86_expand_copysign): Ditto.
2844 (ix86_expand_xorsign): Ditto.
2845 * config/i386/i386.c (ix86_build_const_vector): Handle HF vector
2847 (ix86_build_signbit_mask): Ditto.
2848 (ix86_can_change_mode_class): Ditto.
2849 * config/i386/i386.md
2850 (SSEMODEF): Add HFmode.
2851 (ssevecmodef): Ditto.
2852 (<code>hf2): New define_expand.
2853 (*<code>hf2_1): New define_insn_and_split.
2854 (copysign<mode>): Extend to support HFmode under AVX512FP16.
2855 (xorsign<mode>): Ditto.
2856 * config/i386/sse.md (VFB): New mode iterator.
2857 (VFB_128_256): Ditto.
2859 (sseintvecmode2): Support HF vector mode.
2860 (<code><mode>2): Use new mode iterator.
2861 (*<code><mode>2): Ditto.
2862 (copysign<mode>3): Ditto.
2863 (xorsign<mode>3): Ditto.
2864 (<code><mode>3<mask_name>): Ditto.
2865 (<code><mode>3<mask_name>): Ditto.
2866 (<sse>_andnot<mode>3<mask_name>): Adjust for HF vector mode.
2867 (<sse>_andnot<mode>3<mask_name>): Ditto.
2868 (*<code><mode>3<mask_name>): Ditto.
2869 (*<code><mode>3<mask_name>): Ditto.
2871 2021-09-18 liuhongt <hongtao.liu@intel.com>
2873 * config/i386/avx512fp16intrin.h (_mm512_mask_fmadd_ph):
2875 (_mm512_mask3_fmadd_ph): Likewise.
2876 (_mm512_maskz_fmadd_ph): Likewise.
2877 (_mm512_fmadd_round_ph): Likewise.
2878 (_mm512_mask_fmadd_round_ph): Likewise.
2879 (_mm512_mask3_fmadd_round_ph): Likewise.
2880 (_mm512_maskz_fmadd_round_ph): Likewise.
2881 (_mm512_fnmadd_ph): Likewise.
2882 (_mm512_mask_fnmadd_ph): Likewise.
2883 (_mm512_mask3_fnmadd_ph): Likewise.
2884 (_mm512_maskz_fnmadd_ph): Likewise.
2885 (_mm512_fnmadd_round_ph): Likewise.
2886 (_mm512_mask_fnmadd_round_ph): Likewise.
2887 (_mm512_mask3_fnmadd_round_ph): Likewise.
2888 (_mm512_maskz_fnmadd_round_ph): Likewise.
2889 (_mm512_fmsub_ph): Likewise.
2890 (_mm512_mask_fmsub_ph): Likewise.
2891 (_mm512_mask3_fmsub_ph): Likewise.
2892 (_mm512_maskz_fmsub_ph): Likewise.
2893 (_mm512_fmsub_round_ph): Likewise.
2894 (_mm512_mask_fmsub_round_ph): Likewise.
2895 (_mm512_mask3_fmsub_round_ph): Likewise.
2896 (_mm512_maskz_fmsub_round_ph): Likewise.
2897 (_mm512_fnmsub_ph): Likewise.
2898 (_mm512_mask_fnmsub_ph): Likewise.
2899 (_mm512_mask3_fnmsub_ph): Likewise.
2900 (_mm512_maskz_fnmsub_ph): Likewise.
2901 (_mm512_fnmsub_round_ph): Likewise.
2902 (_mm512_mask_fnmsub_round_ph): Likewise.
2903 (_mm512_mask3_fnmsub_round_ph): Likewise.
2904 (_mm512_maskz_fnmsub_round_ph): Likewise.
2905 * config/i386/avx512fp16vlintrin.h (_mm256_fmadd_ph):
2907 (_mm256_mask_fmadd_ph): Likewise.
2908 (_mm256_mask3_fmadd_ph): Likewise.
2909 (_mm256_maskz_fmadd_ph): Likewise.
2910 (_mm_fmadd_ph): Likewise.
2911 (_mm_mask_fmadd_ph): Likewise.
2912 (_mm_mask3_fmadd_ph): Likewise.
2913 (_mm_maskz_fmadd_ph): Likewise.
2914 (_mm256_fnmadd_ph): Likewise.
2915 (_mm256_mask_fnmadd_ph): Likewise.
2916 (_mm256_mask3_fnmadd_ph): Likewise.
2917 (_mm256_maskz_fnmadd_ph): Likewise.
2918 (_mm_fnmadd_ph): Likewise.
2919 (_mm_mask_fnmadd_ph): Likewise.
2920 (_mm_mask3_fnmadd_ph): Likewise.
2921 (_mm_maskz_fnmadd_ph): Likewise.
2922 (_mm256_fmsub_ph): Likewise.
2923 (_mm256_mask_fmsub_ph): Likewise.
2924 (_mm256_mask3_fmsub_ph): Likewise.
2925 (_mm256_maskz_fmsub_ph): Likewise.
2926 (_mm_fmsub_ph): Likewise.
2927 (_mm_mask_fmsub_ph): Likewise.
2928 (_mm_mask3_fmsub_ph): Likewise.
2929 (_mm_maskz_fmsub_ph): Likewise.
2930 (_mm256_fnmsub_ph): Likewise.
2931 (_mm256_mask_fnmsub_ph): Likewise.
2932 (_mm256_mask3_fnmsub_ph): Likewise.
2933 (_mm256_maskz_fnmsub_ph): Likewise.
2934 (_mm_fnmsub_ph): Likewise.
2935 (_mm_mask_fnmsub_ph): Likewise.
2936 (_mm_mask3_fnmsub_ph): Likewise.
2937 (_mm_maskz_fnmsub_ph): Likewise.
2938 * config/i386/i386-builtin.def: Add corresponding new builtins.
2939 * config/i386/sse.md
2940 (<avx512>_fmadd_<mode>_maskz<round_expand_name>): Adjust to
2941 support HF vector modes.
2942 (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
2944 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
2945 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
2946 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
2947 (<avx512>_fmadd_<mode>_mask<round_name>): Ditto.
2948 (<avx512>_fmadd_<mode>_mask3<round_name>): Ditto.
2949 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Ditto.
2950 (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>):
2952 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
2953 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
2954 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
2955 (<avx512>_fmsub_<mode>_mask<round_name>): Ditto.
2956 (<avx512>_fmsub_<mode>_mask3<round_name>): Ditto.
2957 (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>):
2959 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
2960 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
2961 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
2962 (<avx512>_fnmadd_<mode>_mask<round_name>): Ditto.
2963 (<avx512>_fnmadd_<mode>_mask3<round_name>): Ditto.
2964 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Ditto.
2965 (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>):
2967 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
2968 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
2969 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
2970 (<avx512>_fnmsub_<mode>_mask<round_name>): Ditto.
2971 (<avx512>_fnmsub_<mode>_mask3<round_name>): Ditto.
2973 2021-09-18 liuhongt <hongtao.liu@intel.com>
2975 * config/i386/avx512fp16intrin.h (_mm512_fmaddsub_ph):
2977 (_mm512_mask_fmaddsub_ph): Likewise.
2978 (_mm512_mask3_fmaddsub_ph): Likewise.
2979 (_mm512_maskz_fmaddsub_ph): Likewise.
2980 (_mm512_fmaddsub_round_ph): Likewise.
2981 (_mm512_mask_fmaddsub_round_ph): Likewise.
2982 (_mm512_mask3_fmaddsub_round_ph): Likewise.
2983 (_mm512_maskz_fmaddsub_round_ph): Likewise.
2984 (_mm512_mask_fmsubadd_ph): Likewise.
2985 (_mm512_mask3_fmsubadd_ph): Likewise.
2986 (_mm512_maskz_fmsubadd_ph): Likewise.
2987 (_mm512_fmsubadd_round_ph): Likewise.
2988 (_mm512_mask_fmsubadd_round_ph): Likewise.
2989 (_mm512_mask3_fmsubadd_round_ph): Likewise.
2990 (_mm512_maskz_fmsubadd_round_ph): Likewise.
2991 * config/i386/avx512fp16vlintrin.h (_mm256_fmaddsub_ph):
2993 (_mm256_mask_fmaddsub_ph): Likewise.
2994 (_mm256_mask3_fmaddsub_ph): Likewise.
2995 (_mm256_maskz_fmaddsub_ph): Likewise.
2996 (_mm_fmaddsub_ph): Likewise.
2997 (_mm_mask_fmaddsub_ph): Likewise.
2998 (_mm_mask3_fmaddsub_ph): Likewise.
2999 (_mm_maskz_fmaddsub_ph): Likewise.
3000 (_mm256_fmsubadd_ph): Likewise.
3001 (_mm256_mask_fmsubadd_ph): Likewise.
3002 (_mm256_mask3_fmsubadd_ph): Likewise.
3003 (_mm256_maskz_fmsubadd_ph): Likewise.
3004 (_mm_fmsubadd_ph): Likewise.
3005 (_mm_mask_fmsubadd_ph): Likewise.
3006 (_mm_mask3_fmsubadd_ph): Likewise.
3007 (_mm_maskz_fmsubadd_ph): Likewise.
3008 * config/i386/i386-builtin.def: Add corresponding new builtins.
3009 * config/i386/sse.md (VFH_SF_AVX512VL): New mode iterator.
3010 * (<avx512>_fmsubadd_<mode>_maskz<round_expand_name>): New expander.
3011 * (<avx512>_fmaddsub_<mode>_maskz<round_expand_name>): Use
3013 * (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
3015 * (<avx512>_fmaddsub_<mode>_mask<round_name>): Ditto.
3016 * (<avx512>_fmaddsub_<mode>_mask3<round_name>): Ditto.
3017 * (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
3019 * (<avx512>_fmsubadd_<mode>_mask<round_name>): Ditto.
3020 * (<avx512>_fmsubadd_<mode>_mask3<round_name>): Ditto.
3022 2021-09-18 liuhongt <hongtao.liu@intel.com>
3025 * config/i386/i386.c (ix86_print_operand): Handle
3026 V8HF/V16HF/V32HFmode.
3027 * config/i386/i386.h (VALID_BCST_MODE_P): Add HFmode.
3028 * config/i386/sse.md (avx512bcst): Remove.
3030 2021-09-17 Martin Sebor <msebor@redhat.com>
3032 * Makefile.in (OBJS): Add gimple-predicate-analysis.o.
3033 * tree-ssa-uninit.c (max_phi_args): Move to gimple-predicate-analysis.
3034 (MASK_SET_BIT, MASK_TEST_BIT, MASK_EMPTY): Same.
3035 (check_defs): Add comment.
3036 (can_skip_redundant_opnd): Update comment.
3037 (compute_uninit_opnds_pos): Adjust to namespace change.
3038 (find_pdom): Move to gimple-predicate-analysis.cc.
3040 (struct uninit_undef_val_t): New.
3041 (is_non_loop_exit_postdominating): Move to gimple-predicate-analysis.cc.
3042 (find_control_equiv_block): Same.
3043 (MAX_NUM_CHAINS, MAX_CHAIN_LEN, MAX_POSTDOM_CHECK): Same.
3044 (MAX_SWITCH_CASES): Same.
3045 (compute_control_dep_chain): Same.
3046 (find_uninit_use): Use predicate analyzer.
3047 (struct pred_info): Move to gimple-predicate-analysis.
3048 (convert_control_dep_chain_into_preds): Same.
3049 (find_predicates): Same.
3050 (collect_phi_def_edges): Same.
3051 (warn_uninitialized_phi): Use predicate analyzer.
3052 (find_def_preds): Move to gimple-predicate-analysis.
3053 (dump_pred_info): Same.
3054 (dump_pred_chain): Same.
3055 (dump_predicates): Same.
3056 (destroy_predicate_vecs): Remove.
3057 (execute_late_warn_uninitialized): New.
3058 (get_cmp_code): Move to gimple-predicate-analysis.
3059 (is_value_included_in): Same.
3060 (value_sat_pred_p): Same.
3061 (find_matching_predicate_in_rest_chains): Same.
3062 (is_use_properly_guarded): Same.
3063 (prune_uninit_phi_opnds): Same.
3064 (find_var_cmp_const): Same.
3065 (use_pred_not_overlap_with_undef_path_pred): Same.
3066 (pred_equal_p): Same.
3067 (is_neq_relop_p): Same.
3068 (is_neq_zero_form_p): Same.
3069 (pred_expr_equal_p): Same.
3070 (is_pred_expr_subset_of): Same.
3071 (is_pred_chain_subset_of): Same.
3072 (is_included_in): Same.
3073 (is_superset_of): Same.
3075 (simplify_pred): Same.
3076 (simplify_preds_2): Same.
3077 (simplify_preds_3): Same.
3078 (simplify_preds_4): Same.
3079 (simplify_preds): Same.
3081 (push_to_worklist): Same.
3082 (get_pred_info_from_cmp): Same.
3083 (is_degenerated_phi): Same.
3084 (normalize_one_pred_1): Same.
3085 (normalize_one_pred): Same.
3086 (normalize_one_pred_chain): Same.
3087 (normalize_preds): Same.
3088 (can_one_predicate_be_invalidated_p): Same.
3089 (can_chain_union_be_invalidated_p): Same.
3090 (uninit_uses_cannot_happen): Same.
3091 (pass_late_warn_uninitialized::execute): Define.
3092 * gimple-predicate-analysis.cc: New file.
3093 * gimple-predicate-analysis.h: New file.
3095 2021-09-17 Julian Brown <julian@codesourcery.com>
3097 * config/gcn/gcn.c (gimple.h): Include.
3098 (gcn_fork_join): Emit barrier for worker-level joins.
3099 * omp-oacc-neuter-broadcast.cc (find_local_vars_to_propagate): Add
3100 writes_gang_private bitmap parameter. Set bit for blocks
3101 containing gang-private variable writes.
3102 (worker_single_simple): Don't emit barrier after predicated block.
3103 (worker_single_copy): Don't emit barrier if we're not broadcasting
3104 anything and the block contains no gang-private writes.
3105 (neuter_worker_single): Don't predicate blocks that only contain
3106 NOPs or internal marker functions. Pass has_gang_private_write
3107 argument to worker_single_copy.
3108 (oacc_do_neutering): Add writes_gang_private bitmap handling.
3110 2021-09-17 Julian Brown <julian@codesourcery.com>
3112 * config/gcn/gcn-protos.h
3113 (gcn_goacc_create_worker_broadcast_record): Update prototype.
3114 * config/gcn/gcn-tree.c (gcn_goacc_get_worker_red_decl): Use
3115 preallocated block of LDS memory. Do not cache/share decls for
3116 reduction temporaries between invocations.
3117 (gcn_goacc_reduction_teardown): Unshare VAR on second use.
3118 (gcn_goacc_create_worker_broadcast_record): Add OFFSET parameter
3119 and return temporary LDS space at that offset. Return pointer in
3121 * config/gcn/gcn.c (acc_lds_size, gang_private_hwm, lds_allocs):
3123 (ACC_LDS_SIZE): Define as acc_lds_size.
3124 (gcn_init_machine_status): Don't initialise lds_allocated,
3125 lds_allocs, reduc_decls fields of machine function struct.
3126 (gcn_option_override): Handle default size for gang-private
3127 variables and -mgang-private-size option.
3128 (gcn_expand_prologue): Use LDS_SIZE instead of LDS_SIZE-1 when
3129 initialising M0_REG.
3130 (gcn_shared_mem_layout): New function.
3131 (gcn_print_lds_decl): Update comment. Use global lds_allocs map and
3132 gang_private_hwm variable.
3133 (TARGET_GOACC_SHARED_MEM_LAYOUT): Define target hook.
3134 * config/gcn/gcn.h (machine_function): Remove lds_allocated,
3135 lds_allocs, reduc_decls. Add reduction_base, reduction_limit.
3136 * config/gcn/gcn.opt (gang_private_size_opt): New global.
3137 (mgang-private-size=): New option.
3138 * doc/tm.texi.in (TARGET_GOACC_SHARED_MEM_LAYOUT): Place
3140 * doc/tm.texi: Regenerate.
3141 * omp-oacc-neuter-broadcast.cc (targhooks.h, diagnostic-core.h):
3143 (build_sender_ref): Handle sender_decl being pointer.
3144 (worker_single_copy): Add PLACEMENT and ISOLATE_BROADCASTS
3145 parameters. Pass placement argument to
3146 create_worker_broadcast_record hook invocations. Handle
3147 sender_decl being pointer and isolate_broadcasts inserting extra
3149 (blk_offset_map_t): Add typedef.
3150 (neuter_worker_single): Add BLK_OFFSET_MAP parameter. Pass
3151 preallocated range to worker_single_copy call.
3152 (dfs_broadcast_reachable_1): New function.
3153 (idx_decl_pair_t, used_range_vec_t): New typedefs.
3154 (sort_size_descending): New function.
3155 (addr_range): New class.
3156 (splay_tree_compare_addr_range, splay_tree_free_key)
3157 (first_fit_range, merge_ranges_1, merge_ranges): New functions.
3158 (execute_omp_oacc_neuter_broadcast): Rename to...
3159 (oacc_do_neutering): ... this. Add BOUNDS_LO, BOUNDS_HI
3160 parameters. Arrange layout of shared memory for broadcast
3162 (execute_omp_oacc_neuter_broadcast): New function.
3163 (pass_omp_oacc_neuter_broadcast::gate): Remove num_workers==1
3164 handling from here. Enable pass for all OpenACC routines in order
3165 to call shared memory-layout hook.
3166 * target.def (create_worker_broadcast_record): Add OFFSET
3168 (shared_mem_layout): New hook.
3170 2021-09-17 Julian Brown <julian@codesourcery.com>
3171 Thomas Schwinge <thomas@codesourcery.com>
3173 * omp-oacc-neuter-broadcast.cc
3174 (pass_omp_oacc_neuter_broadcast::gate): Disable if num_workers is
3176 (execute_omp_oacc_neuter_broadcast): Adjust.
3178 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
3180 * value-relation.cc (class equiv_chain): Move to header file.
3181 (path_oracle::path_oracle): New.
3182 (path_oracle::~path_oracle): New.
3183 (path_oracle::register_relation): New.
3184 (path_oracle::query_relation): New.
3185 (path_oracle::reset_path): New.
3186 (path_oracle::dump): New.
3187 * value-relation.h (class equiv_chain): Move to here.
3188 (class path_oracle): New.
3190 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
3192 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a DOM
3194 * gimple-range-fold.cc (fur_depend::register_relation): Use
3195 register_stmt/edge routines.
3196 * value-relation.cc (equiv_chain::find): Relocate from equiv_oracle.
3197 (equiv_oracle::equiv_oracle): Create self equivalence cache.
3198 (equiv_oracle::~equiv_oracle): Release same.
3199 (equiv_oracle::equiv_set): Return entry from self equiv cache if there
3200 are no equivalences.
3201 (equiv_oracle::find_equiv_block): Move list find to equiv_chain.
3202 (equiv_oracle::register_relation): Rename from register_equiv.
3203 (relation_chain_head::find_relation): Relocate from dom_oracle.
3204 (relation_oracle::register_stmt): New.
3205 (relation_oracle::register_edge): New.
3206 (dom_oracle::*): Rename from relation_oracle.
3207 (dom_oracle::register_relation): Adjust to call equiv_oracle.
3208 (dom_oracle::set_one_relation): Split from register_relation.
3209 (dom_oracle::register_transitives): Consolidate 2 methods.
3210 (dom_oracle::find_relation_block): Move core to relation_chain.
3211 (dom_oracle::query_relation): Rename from find_relation_dom and adjust.
3212 * value-relation.h (class relation_oracle): New pure virtual base.
3213 (class equiv_oracle): Inherit from relation_oracle and adjust.
3214 (class dom_oracle): Rename from old relation_oracle and adjust.
3216 2021-09-17 Martin Sebor <msebor@redhat.com>
3218 PR middle-end/102200
3219 * pointer-query.cc (access_ref::inform_access): Handle MIN/MAX_EXPR.
3220 (handle_min_max_size): Change argument. Store original SSA_NAME for
3221 operands to potentially distinct (sub)objects.
3222 (compute_objsize_r): Adjust call to the above.
3224 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
3226 * config/rs6000/rs6000.c (rs6000-builtins.h): New include.
3227 (rs6000_new_builtin_vectorized_function): New function.
3228 (rs6000_new_builtin_md_vectorized_function): Likewise.
3229 (rs6000_builtin_vectorized_function): Call
3230 rs6000_new_builtin_vectorized_function.
3231 (rs6000_builtin_md_vectorized_function): Call
3232 rs6000_new_builtin_md_vectorized_function.
3234 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
3236 * config/rs6000/rs6000-builtin-new.def (ASSEMBLE_ACC): Add mmaint flag.
3237 (ASSEMBLE_PAIR): Likewise.
3238 (BUILD_ACC): Likewise.
3239 (DISASSEMBLE_ACC): Likewise.
3240 (DISASSEMBLE_PAIR): Likewise.
3241 (PMXVBF16GER2): Likewise.
3242 (PMXVBF16GER2NN): Likewise.
3243 (PMXVBF16GER2NP): Likewise.
3244 (PMXVBF16GER2PN): Likewise.
3245 (PMXVBF16GER2PP): Likewise.
3246 (PMXVF16GER2): Likewise.
3247 (PMXVF16GER2NN): Likewise.
3248 (PMXVF16GER2NP): Likewise.
3249 (PMXVF16GER2PN): Likewise.
3250 (PMXVF16GER2PP): Likewise.
3251 (PMXVF32GER): Likewise.
3252 (PMXVF32GERNN): Likewise.
3253 (PMXVF32GERNP): Likewise.
3254 (PMXVF32GERPN): Likewise.
3255 (PMXVF32GERPP): Likewise.
3256 (PMXVF64GER): Likewise.
3257 (PMXVF64GERNN): Likewise.
3258 (PMXVF64GERNP): Likewise.
3259 (PMXVF64GERPN): Likewise.
3260 (PMXVF64GERPP): Likewise.
3261 (PMXVI16GER2): Likewise.
3262 (PMXVI16GER2PP): Likewise.
3263 (PMXVI16GER2S): Likewise.
3264 (PMXVI16GER2SPP): Likewise.
3265 (PMXVI4GER8): Likewise.
3266 (PMXVI4GER8PP): Likewise.
3267 (PMXVI8GER4): Likewise.
3268 (PMXVI8GER4PP): Likewise.
3269 (PMXVI8GER4SPP): Likewise.
3270 (XVBF16GER2): Likewise.
3271 (XVBF16GER2NN): Likewise.
3272 (XVBF16GER2NP): Likewise.
3273 (XVBF16GER2PN): Likewise.
3274 (XVBF16GER2PP): Likewise.
3275 (XVF16GER2): Likewise.
3276 (XVF16GER2NN): Likewise.
3277 (XVF16GER2NP): Likewise.
3278 (XVF16GER2PN): Likewise.
3279 (XVF16GER2PP): Likewise.
3280 (XVF32GER): Likewise.
3281 (XVF32GERNN): Likewise.
3282 (XVF32GERNP): Likewise.
3283 (XVF32GERPN): Likewise.
3284 (XVF32GERPP): Likewise.
3285 (XVF64GER): Likewise.
3286 (XVF64GERNN): Likewise.
3287 (XVF64GERNP): Likewise.
3288 (XVF64GERPN): Likewise.
3289 (XVF64GERPP): Likewise.
3290 (XVI16GER2): Likewise.
3291 (XVI16GER2PP): Likewise.
3292 (XVI16GER2S): Likewise.
3293 (XVI16GER2SPP): Likewise.
3294 (XVI4GER8): Likewise.
3295 (XVI4GER8PP): Likewise.
3296 (XVI8GER4): Likewise.
3297 (XVI8GER4PP): Likewise.
3298 (XVI8GER4SPP): Likewise.
3299 (XXMFACC): Likewise.
3300 (XXMTACC): Likewise.
3301 (XXSETACCZ): Likewise.
3302 (ASSEMBLE_PAIR_V): Likewise.
3303 (BUILD_PAIR): Likewise.
3304 (DISASSEMBLE_PAIR_V): Likewise.
3307 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_mma_builtin):
3308 Handle RS6000_BIF_LXVP and RS6000_BIF_STXVP.
3309 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add ismmaint.
3310 (parse_bif_attrs): Handle ismmaint.
3311 (write_decls): Add bif_mmaint_bit and bif_is_mmaint.
3312 (write_bif_static_init): Handle ismmaint.
3314 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
3316 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin): New
3318 (rs6000_gimple_fold_builtin): Call rs6000_gimple_fold_new_builtin.
3319 (rs6000_new_builtin_valid_without_lhs): New function.
3320 (rs6000_gimple_fold_new_mma_builtin): Likewise.
3321 (rs6000_gimple_fold_new_builtin): Likewise.
3323 2021-09-17 Thomas Schwinge <thomas@codesourcery.com>
3325 * hash-table.h (hash_table<Descriptor, Lazy, Allocator>::expand):
3326 Destruct stale Value objects.
3327 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
3330 2021-09-17 Roger Sayle <roger@nextmovesoftware.com>
3333 * match.pd (shift optimizations): Disable recent sign-changing
3334 optimization for shifts by zero, these will be folded later.
3336 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
3338 * config/rs6000/rs6000-builtin-new.def (__builtin_mffsl): Move from
3339 [power9] to [always].
3341 2021-09-17 Richard Biener <rguenther@suse.de>
3343 * tree-vect-stmts.c (vectorizable_load): Do not frob
3346 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
3348 * config/i386/i386-features.c (remove_partial_avx_dependency):
3349 Also check TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY and
3350 and TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY before generating
3352 * config/i386/i386.h (TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY):
3354 (TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
3355 * config/i386/i386.md (SSE FP to FP splitters): Replace
3356 TARGET_SSE_PARTIAL_REG_DEPENDENCY with
3357 TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY.
3358 (SSE INT to FP splitter): Replace TARGET_SSE_PARTIAL_REG_DEPENDENCY
3359 with TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY.
3360 * config/i386/x86-tune.def
3361 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): New.
3362 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
3364 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
3367 * config/i386/i386-features.c (remove_partial_avx_dependency):
3368 Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
3369 before generating vxorps.
3371 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
3373 * config/i386/i386-options.c (processor_cost_table): Use
3374 tremont_cost for Tremont.
3375 * config/i386/x86-tune-costs.h (tremont_memcpy): New.
3376 (tremont_memset): Likewise.
3377 (tremont_cost): Likewise.
3378 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
3381 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
3383 * common/config/i386/i386-common.c: Use Haswell scheduling model
3385 * config/i386/i386.c (ix86_sched_init_global): Prepare for Tremont
3387 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Tremont
3389 (ix86_adjust_cost): Handle Tremont.
3390 * config/i386/x86-tune.def (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY):
3392 (X86_TUNE_USE_LEAVE): Likewise.
3393 (X86_TUNE_PUSH_MEMORY): Likewise.
3394 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
3395 (X86_TUNE_USE_CLTD): Likewise.
3396 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
3397 (X86_TUNE_AVOID_MFENCE): Likewise.
3398 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
3399 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
3400 (X86_TUNE_ACCUMULATE_OUTGOING_ARGS): Disable for Tremont.
3401 (X86_TUNE_FOUR_JUMP_LIMIT): Likewise.
3402 (X86_TUNE_OPT_AGU): Likewise.
3403 (X86_TUNE_AVOID_LEA_FOR_ADDR): Likewise.
3404 (X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE): Likewise.
3405 (X86_TUNE_EXPAND_ABS): Likewise.
3406 (X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS): Likewise.
3407 (X86_TUNE_SLOW_PSHUFB): Likewise.
3409 2021-09-17 Eric Botcazou <ebotcazou@adacore.com>
3411 PR rtl-optimization/102306
3412 * combine.c (try_combine): Abort the combination if we are about to
3413 duplicate volatile references.
3415 2021-09-17 liuhongt <hongtao.liu@intel.com>
3417 * config/i386/avx512fp16intrin.h (_mm_undefined_ph):
3419 (_mm256_undefined_ph): Likewise.
3420 (_mm512_undefined_ph): Likewise.
3421 (_mm_cvtsh_h): Likewise.
3422 (_mm256_cvtsh_h): Likewise.
3423 (_mm512_cvtsh_h): Likewise.
3424 (_mm512_castph_ps): Likewise.
3425 (_mm512_castph_pd): Likewise.
3426 (_mm512_castph_si512): Likewise.
3427 (_mm512_castph512_ph128): Likewise.
3428 (_mm512_castph512_ph256): Likewise.
3429 (_mm512_castph128_ph512): Likewise.
3430 (_mm512_castph256_ph512): Likewise.
3431 (_mm512_zextph128_ph512): Likewise.
3432 (_mm512_zextph256_ph512): Likewise.
3433 (_mm512_castps_ph): Likewise.
3434 (_mm512_castpd_ph): Likewise.
3435 (_mm512_castsi512_ph): Likewise.
3436 * config/i386/avx512fp16vlintrin.h (_mm_castph_ps):
3438 (_mm256_castph_ps): Likewise.
3439 (_mm_castph_pd): Likewise.
3440 (_mm256_castph_pd): Likewise.
3441 (_mm_castph_si128): Likewise.
3442 (_mm256_castph_si256): Likewise.
3443 (_mm_castps_ph): Likewise.
3444 (_mm256_castps_ph): Likewise.
3445 (_mm_castpd_ph): Likewise.
3446 (_mm256_castpd_ph): Likewise.
3447 (_mm_castsi128_ph): Likewise.
3448 (_mm256_castsi256_ph): Likewise.
3449 (_mm256_castph256_ph128): Likewise.
3450 (_mm256_castph128_ph256): Likewise.
3451 (_mm256_zextph128_ph256): Likewise.
3453 2021-09-17 liuhongt <hongtao.liu@intel.com>
3455 * config/i386/avx512fp16intrin.h (_mm_cvtsh_ss):
3457 (_mm_mask_cvtsh_ss): Likewise.
3458 (_mm_maskz_cvtsh_ss): Likewise.
3459 (_mm_cvtsh_sd): Likewise.
3460 (_mm_mask_cvtsh_sd): Likewise.
3461 (_mm_maskz_cvtsh_sd): Likewise.
3462 (_mm_cvt_roundsh_ss): Likewise.
3463 (_mm_mask_cvt_roundsh_ss): Likewise.
3464 (_mm_maskz_cvt_roundsh_ss): Likewise.
3465 (_mm_cvt_roundsh_sd): Likewise.
3466 (_mm_mask_cvt_roundsh_sd): Likewise.
3467 (_mm_maskz_cvt_roundsh_sd): Likewise.
3468 (_mm_cvtss_sh): Likewise.
3469 (_mm_mask_cvtss_sh): Likewise.
3470 (_mm_maskz_cvtss_sh): Likewise.
3471 (_mm_cvtsd_sh): Likewise.
3472 (_mm_mask_cvtsd_sh): Likewise.
3473 (_mm_maskz_cvtsd_sh): Likewise.
3474 (_mm_cvt_roundss_sh): Likewise.
3475 (_mm_mask_cvt_roundss_sh): Likewise.
3476 (_mm_maskz_cvt_roundss_sh): Likewise.
3477 (_mm_cvt_roundsd_sh): Likewise.
3478 (_mm_mask_cvt_roundsd_sh): Likewise.
3479 (_mm_maskz_cvt_roundsd_sh): Likewise.
3480 * config/i386/i386-builtin-types.def
3481 (V8HF_FTYPE_V2DF_V8HF_V8HF_UQI_INT,
3482 V8HF_FTYPE_V4SF_V8HF_V8HF_UQI_INT,
3483 V2DF_FTYPE_V8HF_V2DF_V2DF_UQI_INT,
3484 V4SF_FTYPE_V8HF_V4SF_V4SF_UQI_INT): Add new builtin types.
3485 * config/i386/i386-builtin.def: Add corrresponding new builtins.
3486 * config/i386/i386-expand.c: Handle new builtin types.
3487 * config/i386/sse.md (VF48_128): New mode iterator.
3488 (avx512fp16_vcvtsh2<ssescalarmodesuffix><mask_scalar_name><round_saeonly_scalar_name>):
3490 (avx512fp16_vcvt<ssescalarmodesuffix>2sh<mask_scalar_name><round_scalar_name>):
3493 2021-09-17 liuhongt <hongtao.liu@intel.com>
3495 * config/i386/avx512fp16intrin.h (_mm512_cvtph_pd):
3497 (_mm512_mask_cvtph_pd): Likewise.
3498 (_mm512_maskz_cvtph_pd): Likewise.
3499 (_mm512_cvt_roundph_pd): Likewise.
3500 (_mm512_mask_cvt_roundph_pd): Likewise.
3501 (_mm512_maskz_cvt_roundph_pd): Likewise.
3502 (_mm512_cvtxph_ps): Likewise.
3503 (_mm512_mask_cvtxph_ps): Likewise.
3504 (_mm512_maskz_cvtxph_ps): Likewise.
3505 (_mm512_cvtx_roundph_ps): Likewise.
3506 (_mm512_mask_cvtx_roundph_ps): Likewise.
3507 (_mm512_maskz_cvtx_roundph_ps): Likewise.
3508 (_mm512_cvtxps_ph): Likewise.
3509 (_mm512_mask_cvtxps_ph): Likewise.
3510 (_mm512_maskz_cvtxps_ph): Likewise.
3511 (_mm512_cvtx_roundps_ph): Likewise.
3512 (_mm512_mask_cvtx_roundps_ph): Likewise.
3513 (_mm512_maskz_cvtx_roundps_ph): Likewise.
3514 (_mm512_cvtpd_ph): Likewise.
3515 (_mm512_mask_cvtpd_ph): Likewise.
3516 (_mm512_maskz_cvtpd_ph): Likewise.
3517 (_mm512_cvt_roundpd_ph): Likewise.
3518 (_mm512_mask_cvt_roundpd_ph): Likewise.
3519 (_mm512_maskz_cvt_roundpd_ph): Likewise.
3520 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_pd):
3522 (_mm_mask_cvtph_pd): Likewise.
3523 (_mm_maskz_cvtph_pd): Likewise.
3524 (_mm256_cvtph_pd): Likewise.
3525 (_mm256_mask_cvtph_pd): Likewise.
3526 (_mm256_maskz_cvtph_pd): Likewise.
3527 (_mm_cvtxph_ps): Likewise.
3528 (_mm_mask_cvtxph_ps): Likewise.
3529 (_mm_maskz_cvtxph_ps): Likewise.
3530 (_mm256_cvtxph_ps): Likewise.
3531 (_mm256_mask_cvtxph_ps): Likewise.
3532 (_mm256_maskz_cvtxph_ps): Likewise.
3533 (_mm_cvtxps_ph): Likewise.
3534 (_mm_mask_cvtxps_ph): Likewise.
3535 (_mm_maskz_cvtxps_ph): Likewise.
3536 (_mm256_cvtxps_ph): Likewise.
3537 (_mm256_mask_cvtxps_ph): Likewise.
3538 (_mm256_maskz_cvtxps_ph): Likewise.
3539 (_mm_cvtpd_ph): Likewise.
3540 (_mm_mask_cvtpd_ph): Likewise.
3541 (_mm_maskz_cvtpd_ph): Likewise.
3542 (_mm256_cvtpd_ph): Likewise.
3543 (_mm256_mask_cvtpd_ph): Likewise.
3544 (_mm256_maskz_cvtpd_ph): Likewise.
3545 * config/i386/i386-builtin.def: Add corresponding new builtins.
3546 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3547 * config/i386/i386-expand.c: Handle new builtin types.
3548 * config/i386/sse.md
3549 (VF4_128_8_256): New.
3550 (VF48H_AVX512VL): Ditto.
3551 (ssePHmode): Add HF vector modes.
3552 (castmode): Add new convertable modes.
3555 (avx512fp16_vcvt<castmode>2ph_<mode><mask_name><round_name>): Ditto.
3556 (avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
3557 (*avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
3558 (avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
3559 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
3560 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask_1): Ditto.
3561 (avx512fp16_float_extend_ph<mode>2<mask_name><round_saeonly_name>):
3563 (avx512fp16_float_extend_ph<mode>2<mask_name>): Ditto.
3564 (*avx512fp16_float_extend_ph<mode>2_load<mask_name>): Ditto.
3565 (avx512fp16_float_extend_phv2df2<mask_name>): Ditto.
3566 (*avx512fp16_float_extend_phv2df2_load<mask_name>): Ditto.
3568 2021-09-17 liuhongt <hongtao.liu@intel.com>
3570 * config/i386/avx512fp16intrin.h (_mm_cvttsh_i32):
3572 (_mm_cvttsh_u32): Likewise.
3573 (_mm_cvtt_roundsh_i32): Likewise.
3574 (_mm_cvtt_roundsh_u32): Likewise.
3575 (_mm_cvttsh_i64): Likewise.
3576 (_mm_cvttsh_u64): Likewise.
3577 (_mm_cvtt_roundsh_i64): Likewise.
3578 (_mm_cvtt_roundsh_u64): Likewise.
3579 * config/i386/i386-builtin.def: Add corresponding new builtins.
3580 * config/i386/sse.md
3581 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<round_saeonly_name>):
3584 2021-09-17 liuhongt <hongtao.liu@intel.com>
3586 * config/i386/avx512fp16intrin.h (_mm512_cvttph_epi32):
3588 (_mm512_mask_cvttph_epi32): Likewise.
3589 (_mm512_maskz_cvttph_epi32): Likewise.
3590 (_mm512_cvtt_roundph_epi32): Likewise.
3591 (_mm512_mask_cvtt_roundph_epi32): Likewise.
3592 (_mm512_maskz_cvtt_roundph_epi32): Likewise.
3593 (_mm512_cvttph_epu32): Likewise.
3594 (_mm512_mask_cvttph_epu32): Likewise.
3595 (_mm512_maskz_cvttph_epu32): Likewise.
3596 (_mm512_cvtt_roundph_epu32): Likewise.
3597 (_mm512_mask_cvtt_roundph_epu32): Likewise.
3598 (_mm512_maskz_cvtt_roundph_epu32): Likewise.
3599 (_mm512_cvttph_epi64): Likewise.
3600 (_mm512_mask_cvttph_epi64): Likewise.
3601 (_mm512_maskz_cvttph_epi64): Likewise.
3602 (_mm512_cvtt_roundph_epi64): Likewise.
3603 (_mm512_mask_cvtt_roundph_epi64): Likewise.
3604 (_mm512_maskz_cvtt_roundph_epi64): Likewise.
3605 (_mm512_cvttph_epu64): Likewise.
3606 (_mm512_mask_cvttph_epu64): Likewise.
3607 (_mm512_maskz_cvttph_epu64): Likewise.
3608 (_mm512_cvtt_roundph_epu64): Likewise.
3609 (_mm512_mask_cvtt_roundph_epu64): Likewise.
3610 (_mm512_maskz_cvtt_roundph_epu64): Likewise.
3611 (_mm512_cvttph_epi16): Likewise.
3612 (_mm512_mask_cvttph_epi16): Likewise.
3613 (_mm512_maskz_cvttph_epi16): Likewise.
3614 (_mm512_cvtt_roundph_epi16): Likewise.
3615 (_mm512_mask_cvtt_roundph_epi16): Likewise.
3616 (_mm512_maskz_cvtt_roundph_epi16): Likewise.
3617 (_mm512_cvttph_epu16): Likewise.
3618 (_mm512_mask_cvttph_epu16): Likewise.
3619 (_mm512_maskz_cvttph_epu16): Likewise.
3620 (_mm512_cvtt_roundph_epu16): Likewise.
3621 (_mm512_mask_cvtt_roundph_epu16): Likewise.
3622 (_mm512_maskz_cvtt_roundph_epu16): Likewise.
3623 * config/i386/avx512fp16vlintrin.h (_mm_cvttph_epi32):
3625 (_mm_mask_cvttph_epi32): Likewise.
3626 (_mm_maskz_cvttph_epi32): Likewise.
3627 (_mm256_cvttph_epi32): Likewise.
3628 (_mm256_mask_cvttph_epi32): Likewise.
3629 (_mm256_maskz_cvttph_epi32): Likewise.
3630 (_mm_cvttph_epu32): Likewise.
3631 (_mm_mask_cvttph_epu32): Likewise.
3632 (_mm_maskz_cvttph_epu32): Likewise.
3633 (_mm256_cvttph_epu32): Likewise.
3634 (_mm256_mask_cvttph_epu32): Likewise.
3635 (_mm256_maskz_cvttph_epu32): Likewise.
3636 (_mm_cvttph_epi64): Likewise.
3637 (_mm_mask_cvttph_epi64): Likewise.
3638 (_mm_maskz_cvttph_epi64): Likewise.
3639 (_mm256_cvttph_epi64): Likewise.
3640 (_mm256_mask_cvttph_epi64): Likewise.
3641 (_mm256_maskz_cvttph_epi64): Likewise.
3642 (_mm_cvttph_epu64): Likewise.
3643 (_mm_mask_cvttph_epu64): Likewise.
3644 (_mm_maskz_cvttph_epu64): Likewise.
3645 (_mm256_cvttph_epu64): Likewise.
3646 (_mm256_mask_cvttph_epu64): Likewise.
3647 (_mm256_maskz_cvttph_epu64): Likewise.
3648 (_mm_cvttph_epi16): Likewise.
3649 (_mm_mask_cvttph_epi16): Likewise.
3650 (_mm_maskz_cvttph_epi16): Likewise.
3651 (_mm256_cvttph_epi16): Likewise.
3652 (_mm256_mask_cvttph_epi16): Likewise.
3653 (_mm256_maskz_cvttph_epi16): Likewise.
3654 (_mm_cvttph_epu16): Likewise.
3655 (_mm_mask_cvttph_epu16): Likewise.
3656 (_mm_maskz_cvttph_epu16): Likewise.
3657 (_mm256_cvttph_epu16): Likewise.
3658 (_mm256_mask_cvttph_epu16): Likewise.
3659 (_mm256_maskz_cvttph_epu16): Likewise.
3660 * config/i386/i386-builtin.def: Add new builtins.
3661 * config/i386/sse.md
3662 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name><round_saeonly_name>):
3664 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name>): Ditto.
3665 (*avx512fp16_fix<fixunssuffix>_trunc<mode>2_load<mask_name>): Ditto.
3666 (avx512fp16_fix<fixunssuffix>_truncv2di2<mask_name>): Ditto.
3667 (avx512fp16_fix<fixunssuffix>_truncv2di2_load<mask_name>): Ditto.
3669 2021-09-17 liuhongt <hongtao.liu@intel.com>
3671 * config/i386/avx512fp16intrin.h (_mm_cvtsh_i32): New intrinsic.
3672 (_mm_cvtsh_u32): Likewise.
3673 (_mm_cvt_roundsh_i32): Likewise.
3674 (_mm_cvt_roundsh_u32): Likewise.
3675 (_mm_cvtsh_i64): Likewise.
3676 (_mm_cvtsh_u64): Likewise.
3677 (_mm_cvt_roundsh_i64): Likewise.
3678 (_mm_cvt_roundsh_u64): Likewise.
3679 (_mm_cvti32_sh): Likewise.
3680 (_mm_cvtu32_sh): Likewise.
3681 (_mm_cvt_roundi32_sh): Likewise.
3682 (_mm_cvt_roundu32_sh): Likewise.
3683 (_mm_cvti64_sh): Likewise.
3684 (_mm_cvtu64_sh): Likewise.
3685 (_mm_cvt_roundi64_sh): Likewise.
3686 (_mm_cvt_roundu64_sh): Likewise.
3687 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3688 * config/i386/i386-builtin.def: Add corresponding new builtins.
3689 * config/i386/i386-expand.c (ix86_expand_round_builtin):
3690 Handle new builtin types.
3691 * config/i386/sse.md
3692 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix><round_name>):
3694 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix>_2): Likewise.
3695 (avx512fp16_vcvt<floatsuffix>si2sh<rex64namesuffix><round_name>): Likewise.
3697 2021-09-16 Bill Schmidt <wschmidt@linux.ibm.com>
3699 * config/rs6000/rs6000-c.c (rs6000-builtins.h): New include.
3700 (altivec_resolve_new_overloaded_builtin): New forward decl.
3701 (rs6000_new_builtin_type_compatible): New function.
3702 (altivec_resolve_overloaded_builtin): Call
3703 altivec_resolve_new_overloaded_builtin.
3704 (altivec_build_new_resolved_builtin): New function.
3705 (altivec_resolve_new_overloaded_builtin): Likewise.
3706 * config/rs6000/rs6000-call.c (rs6000_new_builtin_is_supported):
3708 * config/rs6000/rs6000-gen-builtins.c (write_decls): Remove _p from
3709 name of rs6000_new_builtin_is_supported.
3711 2021-09-16 Uroš Bizjak <ubizjak@gmail.com>
3713 * config/i386/i386-protos.h (ix86_decompose_address):
3714 Change return type to bool.
3715 * config/i386/i386.c (ix86_decompose_address): Ditto.
3717 2021-09-16 Tobias Burnus <tobias@codesourcery.com>
3720 * config/rs6000/t-rs6000 (build/rs6000-gen-builtins.o, build/rbtree.o):
3721 Added 'build/' to target, use build/%.o rule.
3722 (build/rs6000-gen-builtins$(build_exeext)): Add 'build/' and
3723 '$(build_exeext)' to target and 'build/' for the *.o files.
3724 (rs6000-builtins.c): Update for those changes; run rs6000-gen-builtins
3727 2021-09-16 Martin Jambor <mjambor@suse.cz>
3729 * cgraph.c (cgraph_node::dump): Do not check caller count sums if
3730 the body has been removed. Remove trailing whitespace.
3732 2021-09-16 Richard Biener <rguenther@suse.de>
3734 PR middle-end/102360
3735 * internal-fn.c (expand_DEFERRED_INIT): Make pattern-init
3736 of non-memory more robust.
3738 2021-09-16 Daniel Cederman <cederman@gaisler.com>
3740 * config/sparc/sparc-opts.h (enum sparc_processor_type): Add LEON5
3741 * config/sparc/sparc.c (struct processor_costs): Add LEON5 costs
3742 (leon5_adjust_cost): Increase cost of store with data dependency
3743 on ALU instruction and FPU anti-dependencies.
3744 (sparc_option_override): Add LEON5 costs
3745 (sparc_adjust_cost): Add LEON5 cost adjustments
3746 * config/sparc/sparc.h: Add LEON5
3747 * config/sparc/sparc.md: Include LEON5 scheduling information
3748 * config/sparc/sparc.opt: Add LEON5
3749 * doc/invoke.texi: Add LEON5
3750 * config/sparc/leon5.md: New file.
3752 2021-09-16 Daniel Cederman <cederman@gaisler.com>
3754 * config/sparc/sparc.md (stack_protect_set32): Add NOP to prevent
3755 sensitive sequence for B2BST errata workaround.
3757 2021-09-16 Daniel Cederman <cederman@gaisler.com>
3759 * config/sparc/sparc.c (sparc_do_work_around_errata): Do not begin
3760 functions with atomic instruction in the UT700 errata workaround.
3762 2021-09-16 Daniel Cederman <cederman@gaisler.com>
3764 * config/sparc/sparc.c (next_active_non_empty_insn): New function
3765 that returns next active non empty assembly instruction.
3766 (sparc_do_work_around_errata): Use new function.
3768 2021-09-16 Daniel Cederman <cederman@gaisler.com>
3770 * config/sparc/sparc.c (store_insn_p): Add predicate for store
3772 (load_insn_p): Add predicate for load attributes.
3773 (sparc_do_work_around_errata): Use new predicates.
3775 2021-09-16 Andreas Larsson <andreas@gaisler.com>
3777 * config/sparc/sparc.c (dump_target_flag_bits): Print bit names for
3780 2021-09-16 Martin Liska <mliska@suse.cz>
3782 * config/mips/netbsd.h: Fix typo in name of a macro.
3784 2021-09-16 liuhongt <hongtao.liu@intel.com>
3786 PR middle-end/102080
3787 * match.pd: Check mask type when doing cond_op related gimple
3789 * tree.c (is_truth_type_for): New function.
3790 * tree.h (is_truth_type_for): New declaration.
3792 2021-09-16 liuhongt <hongtao.liu@intel.com>
3794 * config/i386/avx512fp16intrin.h (_mm512_cvtepi32_ph): New
3796 (_mm512_mask_cvtepi32_ph): Likewise.
3797 (_mm512_maskz_cvtepi32_ph): Likewise.
3798 (_mm512_cvt_roundepi32_ph): Likewise.
3799 (_mm512_mask_cvt_roundepi32_ph): Likewise.
3800 (_mm512_maskz_cvt_roundepi32_ph): Likewise.
3801 (_mm512_cvtepu32_ph): Likewise.
3802 (_mm512_mask_cvtepu32_ph): Likewise.
3803 (_mm512_maskz_cvtepu32_ph): Likewise.
3804 (_mm512_cvt_roundepu32_ph): Likewise.
3805 (_mm512_mask_cvt_roundepu32_ph): Likewise.
3806 (_mm512_maskz_cvt_roundepu32_ph): Likewise.
3807 (_mm512_cvtepi64_ph): Likewise.
3808 (_mm512_mask_cvtepi64_ph): Likewise.
3809 (_mm512_maskz_cvtepi64_ph): Likewise.
3810 (_mm512_cvt_roundepi64_ph): Likewise.
3811 (_mm512_mask_cvt_roundepi64_ph): Likewise.
3812 (_mm512_maskz_cvt_roundepi64_ph): Likewise.
3813 (_mm512_cvtepu64_ph): Likewise.
3814 (_mm512_mask_cvtepu64_ph): Likewise.
3815 (_mm512_maskz_cvtepu64_ph): Likewise.
3816 (_mm512_cvt_roundepu64_ph): Likewise.
3817 (_mm512_mask_cvt_roundepu64_ph): Likewise.
3818 (_mm512_maskz_cvt_roundepu64_ph): Likewise.
3819 (_mm512_cvtepi16_ph): Likewise.
3820 (_mm512_mask_cvtepi16_ph): Likewise.
3821 (_mm512_maskz_cvtepi16_ph): Likewise.
3822 (_mm512_cvt_roundepi16_ph): Likewise.
3823 (_mm512_mask_cvt_roundepi16_ph): Likewise.
3824 (_mm512_maskz_cvt_roundepi16_ph): Likewise.
3825 (_mm512_cvtepu16_ph): Likewise.
3826 (_mm512_mask_cvtepu16_ph): Likewise.
3827 (_mm512_maskz_cvtepu16_ph): Likewise.
3828 (_mm512_cvt_roundepu16_ph): Likewise.
3829 (_mm512_mask_cvt_roundepu16_ph): Likewise.
3830 (_mm512_maskz_cvt_roundepu16_ph): Likewise.
3831 * config/i386/avx512fp16vlintrin.h (_mm_cvtepi32_ph): New
3833 (_mm_mask_cvtepi32_ph): Likewise.
3834 (_mm_maskz_cvtepi32_ph): Likewise.
3835 (_mm256_cvtepi32_ph): Likewise.
3836 (_mm256_mask_cvtepi32_ph): Likewise.
3837 (_mm256_maskz_cvtepi32_ph): Likewise.
3838 (_mm_cvtepu32_ph): Likewise.
3839 (_mm_mask_cvtepu32_ph): Likewise.
3840 (_mm_maskz_cvtepu32_ph): Likewise.
3841 (_mm256_cvtepu32_ph): Likewise.
3842 (_mm256_mask_cvtepu32_ph): Likewise.
3843 (_mm256_maskz_cvtepu32_ph): Likewise.
3844 (_mm_cvtepi64_ph): Likewise.
3845 (_mm_mask_cvtepi64_ph): Likewise.
3846 (_mm_maskz_cvtepi64_ph): Likewise.
3847 (_mm256_cvtepi64_ph): Likewise.
3848 (_mm256_mask_cvtepi64_ph): Likewise.
3849 (_mm256_maskz_cvtepi64_ph): Likewise.
3850 (_mm_cvtepu64_ph): Likewise.
3851 (_mm_mask_cvtepu64_ph): Likewise.
3852 (_mm_maskz_cvtepu64_ph): Likewise.
3853 (_mm256_cvtepu64_ph): Likewise.
3854 (_mm256_mask_cvtepu64_ph): Likewise.
3855 (_mm256_maskz_cvtepu64_ph): Likewise.
3856 (_mm_cvtepi16_ph): Likewise.
3857 (_mm_mask_cvtepi16_ph): Likewise.
3858 (_mm_maskz_cvtepi16_ph): Likewise.
3859 (_mm256_cvtepi16_ph): Likewise.
3860 (_mm256_mask_cvtepi16_ph): Likewise.
3861 (_mm256_maskz_cvtepi16_ph): Likewise.
3862 (_mm_cvtepu16_ph): Likewise.
3863 (_mm_mask_cvtepu16_ph): Likewise.
3864 (_mm_maskz_cvtepu16_ph): Likewise.
3865 (_mm256_cvtepu16_ph): Likewise.
3866 (_mm256_mask_cvtepu16_ph): Likewise.
3867 (_mm256_maskz_cvtepu16_ph): Likewise.
3868 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3869 * config/i386/i386-builtin.def: Add corresponding new builtins.
3870 * config/i386/i386-expand.c
3871 (ix86_expand_args_builtin): Handle new builtin types.
3872 (ix86_expand_round_builtin): Ditto.
3873 * config/i386/i386-modes.def: Declare V2HF and V6HF.
3874 * config/i386/sse.md (VI2H_AVX512VL): New.
3876 (sseintvecmode): Add HF vector modes.
3877 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode><mask_name><round_name>):
3879 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
3880 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
3881 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
3882 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
3883 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask_1): Ditto.
3884 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
3885 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
3886 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
3887 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
3888 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask_1): Ditto.
3889 * config/i386/subst.md (round_qq2phsuff): New subst_attr.
3891 2021-09-16 liuhongt <hongtao.liu@intel.com>
3893 * config/i386/avx512fp16intrin.h (_mm512_cvtph_epi32):
3895 (_mm512_mask_cvtph_epi32): Likewise.
3896 (_mm512_maskz_cvtph_epi32): Likewise.
3897 (_mm512_cvt_roundph_epi32): Likewise.
3898 (_mm512_mask_cvt_roundph_epi32): Likewise.
3899 (_mm512_maskz_cvt_roundph_epi32): Likewise.
3900 (_mm512_cvtph_epu32): Likewise.
3901 (_mm512_mask_cvtph_epu32): Likewise.
3902 (_mm512_maskz_cvtph_epu32): Likewise.
3903 (_mm512_cvt_roundph_epu32): Likewise.
3904 (_mm512_mask_cvt_roundph_epu32): Likewise.
3905 (_mm512_maskz_cvt_roundph_epu32): Likewise.
3906 (_mm512_cvtph_epi64): Likewise.
3907 (_mm512_mask_cvtph_epi64): Likewise.
3908 (_mm512_maskz_cvtph_epi64): Likewise.
3909 (_mm512_cvt_roundph_epi64): Likewise.
3910 (_mm512_mask_cvt_roundph_epi64): Likewise.
3911 (_mm512_maskz_cvt_roundph_epi64): Likewise.
3912 (_mm512_cvtph_epu64): Likewise.
3913 (_mm512_mask_cvtph_epu64): Likewise.
3914 (_mm512_maskz_cvtph_epu64): Likewise.
3915 (_mm512_cvt_roundph_epu64): Likewise.
3916 (_mm512_mask_cvt_roundph_epu64): Likewise.
3917 (_mm512_maskz_cvt_roundph_epu64): Likewise.
3918 (_mm512_cvtph_epi16): Likewise.
3919 (_mm512_mask_cvtph_epi16): Likewise.
3920 (_mm512_maskz_cvtph_epi16): Likewise.
3921 (_mm512_cvt_roundph_epi16): Likewise.
3922 (_mm512_mask_cvt_roundph_epi16): Likewise.
3923 (_mm512_maskz_cvt_roundph_epi16): Likewise.
3924 (_mm512_cvtph_epu16): Likewise.
3925 (_mm512_mask_cvtph_epu16): Likewise.
3926 (_mm512_maskz_cvtph_epu16): Likewise.
3927 (_mm512_cvt_roundph_epu16): Likewise.
3928 (_mm512_mask_cvt_roundph_epu16): Likewise.
3929 (_mm512_maskz_cvt_roundph_epu16): Likewise.
3930 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_epi32):
3932 (_mm_mask_cvtph_epi32): Likewise.
3933 (_mm_maskz_cvtph_epi32): Likewise.
3934 (_mm256_cvtph_epi32): Likewise.
3935 (_mm256_mask_cvtph_epi32): Likewise.
3936 (_mm256_maskz_cvtph_epi32): Likewise.
3937 (_mm_cvtph_epu32): Likewise.
3938 (_mm_mask_cvtph_epu32): Likewise.
3939 (_mm_maskz_cvtph_epu32): Likewise.
3940 (_mm256_cvtph_epu32): Likewise.
3941 (_mm256_mask_cvtph_epu32): Likewise.
3942 (_mm256_maskz_cvtph_epu32): Likewise.
3943 (_mm_cvtph_epi64): Likewise.
3944 (_mm_mask_cvtph_epi64): Likewise.
3945 (_mm_maskz_cvtph_epi64): Likewise.
3946 (_mm256_cvtph_epi64): Likewise.
3947 (_mm256_mask_cvtph_epi64): Likewise.
3948 (_mm256_maskz_cvtph_epi64): Likewise.
3949 (_mm_cvtph_epu64): Likewise.
3950 (_mm_mask_cvtph_epu64): Likewise.
3951 (_mm_maskz_cvtph_epu64): Likewise.
3952 (_mm256_cvtph_epu64): Likewise.
3953 (_mm256_mask_cvtph_epu64): Likewise.
3954 (_mm256_maskz_cvtph_epu64): Likewise.
3955 (_mm_cvtph_epi16): Likewise.
3956 (_mm_mask_cvtph_epi16): Likewise.
3957 (_mm_maskz_cvtph_epi16): Likewise.
3958 (_mm256_cvtph_epi16): Likewise.
3959 (_mm256_mask_cvtph_epi16): Likewise.
3960 (_mm256_maskz_cvtph_epi16): Likewise.
3961 (_mm_cvtph_epu16): Likewise.
3962 (_mm_mask_cvtph_epu16): Likewise.
3963 (_mm_maskz_cvtph_epu16): Likewise.
3964 (_mm256_cvtph_epu16): Likewise.
3965 (_mm256_mask_cvtph_epu16): Likewise.
3966 (_mm256_maskz_cvtph_epu16): Likewise.
3967 * config/i386/i386-builtin-types.def: Add new builtin types.
3968 * config/i386/i386-builtin.def: Add new builtins.
3969 * config/i386/i386-expand.c
3970 (ix86_expand_args_builtin): Handle new builtin types.
3971 (ix86_expand_round_builtin): Ditto.
3972 * config/i386/sse.md (sseintconvert): New.
3974 (UNSPEC_US_FIX_NOTRUNC): Ditto.
3975 (sseintconvertsignprefix): Ditto.
3976 (avx512fp16_vcvtph2<sseintconvertsignprefix><sseintconvert>_<mode><mask_name><round_name>):
3979 2021-09-16 liuhongt <hongtao.liu@intel.com>
3981 * config/i386/avx512fp16intrin.h: (_mm_cvtsi16_si128):
3983 (_mm_cvtsi128_si16): Likewise.
3984 (_mm_mask_load_sh): Likewise.
3985 (_mm_maskz_load_sh): Likewise.
3986 (_mm_mask_store_sh): Likewise.
3987 (_mm_move_sh): Likewise.
3988 (_mm_mask_move_sh): Likewise.
3989 (_mm_maskz_move_sh): Likewise.
3990 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3991 * config/i386/i386-builtin.def: Add corresponding new builtins.
3992 * config/i386/i386-expand.c
3993 (ix86_expand_special_args_builtin): Handle new builtin types.
3994 (ix86_expand_vector_init_one_nonzero): Adjust for FP16 target.
3995 * config/i386/sse.md (VI2F): New mode iterator.
3996 (vec_set<mode>_0): Use new mode iterator.
3997 (avx512f_mov<ssescalarmodelower>_mask): Adjust for HF vector mode.
3998 (avx512f_store<mode>_mask): Ditto.
4000 2021-09-16 Kewen Lin <linkw@linux.ibm.com>
4002 * config/rs6000/rs6000.opt (-mtoc-fusion): Remove.
4004 2021-09-15 David Edelsohn <dje.gcc@gmail.com>
4006 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
4007 Proceed if no symbol summary or the symbol alias flag is false.
4009 2021-09-15 Jakub Jelinek <jakub@redhat.com>
4013 * varasm.c (output_constructor_regular_field): Instead of assertion
4014 that array_size_for_constructor result is equal to size of
4015 TREE_TYPE (local->val) in bytes, assert that the type size is greater
4016 or equal to array_size_for_constructor result and use type size as
4019 2021-09-15 Martin Liska <mliska@suse.cz>
4022 * config/i386/vxworks.h: Use new macro TARGET_CPU_P.
4024 2021-09-15 Martin Liska <mliska@suse.cz>
4027 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
4028 Check that we have a symbol summary for a symbol.
4030 2021-09-15 Richard Biener <rguenther@suse.de>
4033 * config/rs6000/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
4034 to inherit from elfos.h
4036 2021-09-15 liuhongt <hongtao.liu@intel.com>
4039 * config/i386/i386-expand.c
4040 (ix86_expand_vector_init_interleave): Use puncklwd to pack 2
4042 (ix86_expand_vector_set): Use blendw instead of pinsrw.
4043 * config/i386/i386.c (ix86_can_change_mode_class): Adjust for
4044 AVX512FP16 which supports 16bit vector load.
4045 * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>):
4047 (avx512bw_interleave_high<mode><mask_name>): .. this, and
4048 extend to V32HFmode.
4049 (avx2_interleave_highv16hi<mask_name>): Rename to ..
4050 (avx2_interleave_high<mode><mask_name>): .. this, and extend
4052 (vec_interleave_highv8hi<mask_name>): Rename to ..
4053 (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode.
4054 (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>):
4056 (<mask_codefor>avx512bw_interleave_low<mode><mask_name>):
4057 this, and extend to V32HFmode.
4058 (avx2_interleave_lowv16hi<mask_name>): Rename to ..
4059 (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode.
4060 (vec_interleave_lowv8hi<mask_name>): Rename to ..
4061 (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode.
4062 (sse4_1_pblendw): Rename to ..
4063 (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode.
4064 (avx2_pblendph): New define_expand.
4065 (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use
4066 sseintmodesuffix instead of ssemodesuffix.
4067 (blendsuf): New mode attr.
4069 2021-09-15 Richard Biener <rguenther@suse.de>
4071 * tree-vectorizer.h (dr_misalignment): Move out of line.
4072 (dr_target_alignment): New.
4073 (DR_TARGET_ALIGNMENT): Wrap dr_target_alignment.
4074 (set_dr_target_alignment): New.
4075 (SET_DR_TARGET_ALIGNMENT): Wrap set_dr_target_alignment.
4076 * tree-vect-data-refs.c (dr_misalignment): Compute and
4077 return the group members misalignment.
4078 (vect_compute_data_ref_alignment): Use SET_DR_TARGET_ALIGNMENT.
4079 (vect_analyze_data_refs_alignment): Compute alignment only
4080 for the first element of a DR group.
4081 (vect_slp_analyze_node_alignment): Likewise.
4083 2021-09-15 Hongyu Wang <hongyu.wang@intel.com>
4085 * config/i386/avx512fp16intrin.h: Adjust all builtin calls.
4086 * config/i386/avx512fp16vlintrin.h: Likewise.
4087 * config/i386/i386-builtin.def: Adjust builtin name and
4088 enumeration to match AVX512F style.
4090 2021-09-15 Richard Biener <rguenther@suse.de>
4092 PR tree-optimization/102318
4093 * tree-vect-loop.c (vect_transform_cycle_phi): Revert
4094 previous change and do the mode conversion separately from
4095 the sign conversion.
4097 2021-09-15 Hongtao Liu <hongtao.liu@intel.com>
4098 Peter Cordes <peter@cordes.ca>
4101 * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
4102 (*vec_extract<mode><ssescalarmodelower>_valign): Output
4103 vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
4106 2021-09-15 Richard Biener <rguenther@suse.de>
4108 * config.gcc: Remove vax-*-openbsd* configuration.
4110 2021-09-15 Richard Biener <rguenther@suse.de>
4112 * config.gcc: Remove m68k-openbsd.
4114 2021-09-15 Max Filippov <jcmvbkbc@gmail.com>
4117 * config/xtensa/t-xtensa (TM_H): Add include/xtensa-config.h.
4119 2021-09-14 Peter Bergner <bergner@linux.ibm.com>
4121 * config/rs6000/mma.md (unspec): Delete UNSPEC_MMA_XXSETACCZ.
4122 (unspecv): Add UNSPECV_MMA_XXSETACCZ.
4123 (*mma_xxsetaccz): Delete.
4124 (mma_xxsetaccz): Change to define_insn. Remove operand 1.
4125 Use UNSPECV_MMA_XXSETACCZ. Update comment.
4126 * config/rs6000/rs6000.c (rs6000_rtx_costs): Use UNSPECV_MMA_XXSETACCZ.
4128 2021-09-14 Iain Sandoe <iain@sandoe.co.uk>
4130 * Makefile.in: Remove variables related to applying no-PIE
4131 to the exes on $build.
4132 * configure: Regenerate.
4133 * configure.ac: Remove configuration related to applying
4134 no-PIE to the exes on $build.
4136 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
4138 * config/arc/arc.md (doloop_end): Add missing mode.
4139 (loop_end): Likewise.
4141 2021-09-14 Jakub Jelinek <jakub@redhat.com>
4143 * gimplify.c (goa_stabilize_expr): Add depth argument, propagate
4144 it to recursive calls, for depth above 7 just gimplify or return.
4145 Perform a test even for MODIFY_EXPR, ADDR_EXPR, COMPOUND_EXPR with
4146 __builtin_clear_padding and TARGET_EXPR.
4147 (gimplify_omp_atomic): Adjust goa_stabilize_expr callers.
4149 2021-09-14 liuhongt <hongtao.liu@intel.com>
4151 * config/i386/avx512fp16intrin.h (_mm_fpclass_sh_mask):
4153 (_mm_mask_fpclass_sh_mask): Likewise.
4154 (_mm512_mask_fpclass_ph_mask): Likewise.
4155 (_mm512_fpclass_ph_mask): Likewise.
4156 (_mm_getexp_sh): Likewise.
4157 (_mm_mask_getexp_sh): Likewise.
4158 (_mm_maskz_getexp_sh): Likewise.
4159 (_mm512_getexp_ph): Likewise.
4160 (_mm512_mask_getexp_ph): Likewise.
4161 (_mm512_maskz_getexp_ph): Likewise.
4162 (_mm_getexp_round_sh): Likewise.
4163 (_mm_mask_getexp_round_sh): Likewise.
4164 (_mm_maskz_getexp_round_sh): Likewise.
4165 (_mm512_getexp_round_ph): Likewise.
4166 (_mm512_mask_getexp_round_ph): Likewise.
4167 (_mm512_maskz_getexp_round_ph): Likewise.
4168 (_mm_getmant_sh): Likewise.
4169 (_mm_mask_getmant_sh): Likewise.
4170 (_mm_maskz_getmant_sh): Likewise.
4171 (_mm512_getmant_ph): Likewise.
4172 (_mm512_mask_getmant_ph): Likewise.
4173 (_mm512_maskz_getmant_ph): Likewise.
4174 (_mm_getmant_round_sh): Likewise.
4175 (_mm_mask_getmant_round_sh): Likewise.
4176 (_mm_maskz_getmant_round_sh): Likewise.
4177 (_mm512_getmant_round_ph): Likewise.
4178 (_mm512_mask_getmant_round_ph): Likewise.
4179 (_mm512_maskz_getmant_round_ph): Likewise.
4180 * config/i386/avx512fp16vlintrin.h (_mm_mask_fpclass_ph_mask):
4182 (_mm_fpclass_ph_mask): Likewise.
4183 (_mm256_mask_fpclass_ph_mask): Likewise.
4184 (_mm256_fpclass_ph_mask): Likewise.
4185 (_mm256_getexp_ph): Likewise.
4186 (_mm256_mask_getexp_ph): Likewise.
4187 (_mm256_maskz_getexp_ph): Likewise.
4188 (_mm_getexp_ph): Likewise.
4189 (_mm_mask_getexp_ph): Likewise.
4190 (_mm_maskz_getexp_ph): Likewise.
4191 (_mm256_getmant_ph): Likewise.
4192 (_mm256_mask_getmant_ph): Likewise.
4193 (_mm256_maskz_getmant_ph): Likewise.
4194 (_mm_getmant_ph): Likewise.
4195 (_mm_mask_getmant_ph): Likewise.
4196 (_mm_maskz_getmant_ph): Likewise.
4197 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
4198 * config/i386/i386-builtin.def: Add corresponding new builtins.
4199 * config/i386/i386-expand.c
4200 (ix86_expand_args_builtin): Handle new builtin types.
4201 (ix86_expand_round_builtin): Ditto.
4202 * config/i386/sse.md (vecmemsuffix): Add HF vector modes.
4203 (<avx512>_getexp<mode><mask_name><round_saeonly_name>): Adjust
4204 to support HF vector modes.
4205 (avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name):
4207 (avx512dq_fpclass<mode><mask_scalar_merge_name>): Ditto.
4208 (avx512dq_vmfpclass<mode><mask_scalar_merge_name>): Ditto.
4209 (<avx512>_getmant<mode><mask_name><round_saeonly_name>): Ditto.
4210 (avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4213 2021-09-14 liuhongt <hongtao.liu@intel.com>
4215 * config/i386/avx512fp16intrin.h (_mm512_reduce_ph):
4217 (_mm512_mask_reduce_ph): Likewise.
4218 (_mm512_maskz_reduce_ph): Likewise.
4219 (_mm512_reduce_round_ph): Likewise.
4220 (_mm512_mask_reduce_round_ph): Likewise.
4221 (_mm512_maskz_reduce_round_ph): Likewise.
4222 (_mm_reduce_sh): Likewise.
4223 (_mm_mask_reduce_sh): Likewise.
4224 (_mm_maskz_reduce_sh): Likewise.
4225 (_mm_reduce_round_sh): Likewise.
4226 (_mm_mask_reduce_round_sh): Likewise.
4227 (_mm_maskz_reduce_round_sh): Likewise.
4228 (_mm512_roundscale_ph): Likewise.
4229 (_mm512_mask_roundscale_ph): Likewise.
4230 (_mm512_maskz_roundscale_ph): Likewise.
4231 (_mm512_roundscale_round_ph): Likewise.
4232 (_mm512_mask_roundscale_round_ph): Likewise.
4233 (_mm512_maskz_roundscale_round_ph): Likewise.
4234 (_mm_roundscale_sh): Likewise.
4235 (_mm_mask_roundscale_sh): Likewise.
4236 (_mm_maskz_roundscale_sh): Likewise.
4237 (_mm_roundscale_round_sh): Likewise.
4238 (_mm_mask_roundscale_round_sh): Likewise.
4239 (_mm_maskz_roundscale_round_sh): Likewise.
4240 * config/i386/avx512fp16vlintrin.h: (_mm_reduce_ph):
4242 (_mm_mask_reduce_ph): Likewise.
4243 (_mm_maskz_reduce_ph): Likewise.
4244 (_mm256_reduce_ph): Likewise.
4245 (_mm256_mask_reduce_ph): Likewise.
4246 (_mm256_maskz_reduce_ph): Likewise.
4247 (_mm_roundscale_ph): Likewise.
4248 (_mm_mask_roundscale_ph): Likewise.
4249 (_mm_maskz_roundscale_ph): Likewise.
4250 (_mm256_roundscale_ph): Likewise.
4251 (_mm256_mask_roundscale_ph): Likewise.
4252 (_mm256_maskz_roundscale_ph): Likewise.
4253 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
4254 * config/i386/i386-builtin.def: Add corresponding new builtins.
4255 * config/i386/i386-expand.c
4256 (ix86_expand_args_builtin): Handle new builtin types.
4257 (ix86_expand_round_builtin): Ditto.
4258 * config/i386/sse.md (<mask_codefor>reducep<mode><mask_name>):
4260 (<mask_codefor>reducep<mode><mask_name><round_saeonly_name>):
4261 ... this, and adjust for round operands.
4262 (reduces<mode><mask_scalar_name>): Likewise, with ...
4263 (reduces<mode><mask_scalar_name><round_saeonly_scalar_name):
4265 (<avx512>_rndscale<mode><mask_name><round_saeonly_name>):
4266 Adjust for HF vector modes.
4267 (avx512f_rndscale<mode><mask_scalar_name><round_saeonly_scalar_name>):
4269 (*avx512f_rndscale<mode><round_saeonly_name>): Ditto.
4271 2021-09-14 liuhongt <hongtao.liu@intel.com>
4273 * config/i386/avx512fp16intrin.h: (_mm512_rcp_ph):
4275 (_mm512_mask_rcp_ph): Likewise.
4276 (_mm512_maskz_rcp_ph): Likewise.
4277 (_mm_rcp_sh): Likewise.
4278 (_mm_mask_rcp_sh): Likewise.
4279 (_mm_maskz_rcp_sh): Likewise.
4280 (_mm512_scalef_ph): Likewise.
4281 (_mm512_mask_scalef_ph): Likewise.
4282 (_mm512_maskz_scalef_ph): Likewise.
4283 (_mm512_scalef_round_ph): Likewise.
4284 (_mm512_mask_scalef_round_ph): Likewise.
4285 (_mm512_maskz_scalef_round_ph): Likewise.
4286 (_mm_scalef_sh): Likewise.
4287 (_mm_mask_scalef_sh): Likewise.
4288 (_mm_maskz_scalef_sh): Likewise.
4289 (_mm_scalef_round_sh): Likewise.
4290 (_mm_mask_scalef_round_sh): Likewise.
4291 (_mm_maskz_scalef_round_sh): Likewise.
4292 * config/i386/avx512fp16vlintrin.h (_mm_rcp_ph):
4294 (_mm256_rcp_ph): Likewise.
4295 (_mm_mask_rcp_ph): Likewise.
4296 (_mm256_mask_rcp_ph): Likewise.
4297 (_mm_maskz_rcp_ph): Likewise.
4298 (_mm256_maskz_rcp_ph): Likewise.
4299 (_mm_scalef_ph): Likewise.
4300 (_mm256_scalef_ph): Likewise.
4301 (_mm_mask_scalef_ph): Likewise.
4302 (_mm256_mask_scalef_ph): Likewise.
4303 (_mm_maskz_scalef_ph): Likewise.
4304 (_mm256_maskz_scalef_ph): Likewise.
4305 * config/i386/i386-builtin.def: Add new builtins.
4306 * config/i386/sse.md (VFH_AVX512VL): New.
4307 (avx512fp16_rcp<mode>2<mask_name>): Ditto.
4308 (avx512fp16_vmrcpv8hf2<mask_scalar_name>): Ditto.
4309 (avx512f_vmscalef<mode><mask_scalar_name><round_scalar_name>):
4310 Adjust to support HF vector modes.
4311 (<avx512>_scalef<mode><mask_name><round_name>): Ditto.
4313 2021-09-14 liuhongt <hongtao.liu@intel.com>
4315 * config/i386/avx512fp16intrin.h: (_mm512_sqrt_ph):
4317 (_mm512_mask_sqrt_ph): Likewise.
4318 (_mm512_maskz_sqrt_ph): Likewise.
4319 (_mm512_sqrt_round_ph): Likewise.
4320 (_mm512_mask_sqrt_round_ph): Likewise.
4321 (_mm512_maskz_sqrt_round_ph): Likewise.
4322 (_mm512_rsqrt_ph): Likewise.
4323 (_mm512_mask_rsqrt_ph): Likewise.
4324 (_mm512_maskz_rsqrt_ph): Likewise.
4325 (_mm_rsqrt_sh): Likewise.
4326 (_mm_mask_rsqrt_sh): Likewise.
4327 (_mm_maskz_rsqrt_sh): Likewise.
4328 (_mm_sqrt_sh): Likewise.
4329 (_mm_mask_sqrt_sh): Likewise.
4330 (_mm_maskz_sqrt_sh): Likewise.
4331 (_mm_sqrt_round_sh): Likewise.
4332 (_mm_mask_sqrt_round_sh): Likewise.
4333 (_mm_maskz_sqrt_round_sh): Likewise.
4334 * config/i386/avx512fp16vlintrin.h (_mm_sqrt_ph): New intrinsic.
4335 (_mm256_sqrt_ph): Likewise.
4336 (_mm_mask_sqrt_ph): Likewise.
4337 (_mm256_mask_sqrt_ph): Likewise.
4338 (_mm_maskz_sqrt_ph): Likewise.
4339 (_mm256_maskz_sqrt_ph): Likewise.
4340 (_mm_rsqrt_ph): Likewise.
4341 (_mm256_rsqrt_ph): Likewise.
4342 (_mm_mask_rsqrt_ph): Likewise.
4343 (_mm256_mask_rsqrt_ph): Likewise.
4344 (_mm_maskz_rsqrt_ph): Likewise.
4345 (_mm256_maskz_rsqrt_ph): Likewise.
4346 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
4347 * config/i386/i386-builtin.def: Add corresponding new builtins.
4348 * config/i386/i386-expand.c
4349 (ix86_expand_args_builtin): Handle new builtins.
4350 (ix86_expand_round_builtin): Ditto.
4351 * config/i386/sse.md (VF_AVX512FP16VL): New.
4352 (sqrt<mode>2): Adjust for HF vector modes.
4353 (<sse>_sqrt<mode>2<mask_name><round_name>): Likewise.
4354 (<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
4356 (<sse>_rsqrt<mode>2<mask_name>): New.
4357 (avx512fp16_vmrsqrtv8hf2<mask_scalar_name>): Likewise.
4359 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
4362 * diagnostic-spec.c (warning_suppressed_at, copy_warning): Handle
4363 'RESERVED_LOCATION_P' locations.
4364 * warning-control.cc (get_nowarn_spec, suppress_warning)
4365 (copy_warning): Likewise.
4367 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
4369 * diagnostic-spec.h (typedef xint_hash_t): Use 'location_t' instead of...
4370 (typedef key_type_t): ... this. Remove.
4371 (nowarn_map): Document.
4372 * diagnostic-spec.c (nowarn_map): Likewise.
4373 * warning-control.cc (convert_to_key): Evolve functions into...
4374 (get_location): ... these. Adjust all users.
4376 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
4378 * warning-control.cc (copy_warning): Remove 'nowarn_map' setup.
4380 2021-09-13 Jason Merrill <jason@redhat.com>
4382 * params.opt: Add destructive-interference-size and
4383 constructive-interference-size.
4384 * doc/invoke.texi: Document them.
4385 * config/aarch64/aarch64.c (aarch64_override_options_internal):
4387 * config/arm/arm.c (arm_option_override): Set them.
4388 * config/i386/i386-options.c (ix86_option_override_internal):
4391 2021-09-13 Martin Liska <mliska@suse.cz>
4392 H.J. Lu <hjl.tools@gmail.com>
4395 * common/config/i386/cpuinfo.h (cpu_indicator_init): Add support
4396 for x86-64 micro levels for __builtin_cpu_supports.
4397 * common/config/i386/i386-cpuinfo.h (enum feature_priority):
4398 Add priorities for the micro-arch levels.
4399 (enum processor_features): Add new features.
4400 * common/config/i386/i386-isas.h: Add micro-arch features.
4401 * config/i386/i386-builtins.c (get_builtin_code_for_version):
4402 Support the micro-arch levels by callsing
4403 __builtin_cpu_supports.
4404 * doc/extend.texi: Document that the levels are support by
4405 __builtin_cpu_supports.
4407 2021-09-13 Andrew Pinski <apinski@marvell.com>
4410 * config/aarch64/aarch64-builtins.c (aarch64_fold_builtin_lane_check):
4412 (aarch64_general_fold_builtin): Handle AARCH64_SIMD_BUILTIN_LANE_CHECK.
4413 (aarch64_general_gimple_fold_builtin): Likewise.
4415 2021-09-13 Andrew Pinski <apinski@marvell.com>
4417 * config.gcc: Add m32r-*-linux* and m32rle-*-linux*
4418 to the Unsupported targets list.
4419 Remove support for m32r-*-linux* and m32rle-*-linux*.
4420 * config/m32r/linux.h: Removed.
4421 * config/m32r/t-linux: Removed.
4423 2021-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4426 * config/aarch64/aarch64.c (aarch64_classify_address): Don't allow
4427 register index for SVE predicate modes.
4429 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
4431 * tree-ssa-threadbackward.c
4432 (back_threader_profitability::profitable_path_p): Remove FSM
4434 (back_threader_registry::register_path): Same.
4435 * tree-ssa-threadedge.c
4436 (jump_threader::simplify_control_stmt_condition): Same.
4437 * tree-ssa-threadupdate.c (jt_path_registry::jt_path_registry):
4438 Add backedge_threads argument.
4439 (fwd_jt_path_registry::fwd_jt_path_registry): Pass
4440 backedge_threads argument.
4441 (back_jt_path_registry::back_jt_path_registry): Same.
4442 (dump_jump_thread_path): Adjust for FSM removal.
4443 (back_jt_path_registry::rewire_first_differing_edge): Same.
4444 (back_jt_path_registry::adjust_paths_after_duplication): Same.
4445 (back_jt_path_registry::update_cfg): Same.
4446 (jt_path_registry::register_jump_thread): Same.
4447 * tree-ssa-threadupdate.h (enum jump_thread_edge_type): Remove
4449 (class back_jt_path_registry): Add backedge_threads to
4452 2021-09-13 Martin Liska <mliska@suse.cz>
4455 * asan.h (sanitize_coverage_p): Handle when fn == NULL.
4457 2021-09-13 H.J. Lu <hjl.tools@gmail.com>
4460 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): New.
4461 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
4462 (MOVE_MAX): Check TARGET_AVX256_MOVE_BY_PIECES and
4463 TARGET_AVX256_STORE_BY_PIECES instead of
4464 TARGET_AVX256_SPLIT_UNALIGNED_LOAD and
4465 TARGET_AVX256_SPLIT_UNALIGNED_STORE.
4466 (STORE_MAX_PIECES): Check TARGET_AVX256_STORE_BY_PIECES instead
4467 of TARGET_AVX256_SPLIT_UNALIGNED_STORE.
4468 * config/i386/x86-tune.def (X86_TUNE_AVX256_MOVE_BY_PIECES): New.
4469 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
4471 2021-09-13 liuhongt <hongtao.liu@intel.com>
4474 * expmed.c (extract_bit_field_using_extv): Use
4475 gen_lowpart_if_possible instead of gen_lowpart to avoid ICE.
4477 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
4479 * Makefile.in (OBJS): Add value-pointer-equiv.o.
4480 * gimple-ssa-evrp.c (class ssa_equiv_stack): Move to
4481 value-pointer-equiv.*.
4482 (ssa_equiv_stack::ssa_equiv_stack): Same.
4483 (ssa_equiv_stack::enter): Same.
4484 (ssa_equiv_stack::leave): Same.
4485 (ssa_equiv_stack::push_replacement): Same.
4486 (ssa_equiv_stack::get_replacement): Same.
4487 (is_pointer_ssa): Same.
4488 (class pointer_equiv_analyzer): Same.
4489 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
4490 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Same.
4491 (pointer_equiv_analyzer::set_global_equiv): Same.
4492 (pointer_equiv_analyzer::set_cond_equiv): Same.
4493 (pointer_equiv_analyzer::get_equiv): Same.
4494 (pointer_equiv_analyzer::enter): Same.
4495 (pointer_equiv_analyzer::leave): Same.
4496 (pointer_equiv_analyzer::get_equiv_expr): Same.
4497 (pta_valueize): Same.
4498 (pointer_equiv_analyzer::visit_stmt): Same.
4499 (pointer_equiv_analyzer::visit_edge): Same.
4500 (hybrid_folder::value_of_expr): Same.
4501 (hybrid_folder::value_on_edge): Same.
4502 * value-pointer-equiv.cc: New file.
4503 * value-pointer-equiv.h: New file.
4505 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
4508 * gimple-fold.c (gimple_fold_builtin_memory_op): Allow folding
4509 memcpy if the size is not more than MOVE_MAX * MOVE_RATIO.
4511 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
4514 * config/arm/arm.md (movmisaligndi): New define_expand.
4515 * config/arm/vec-common.md (movmisalign<mode>): Iterate over VDQ mode.
4517 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
4520 * emit-rtl.c (gen_highpart): Use adjust_address to handle
4521 MEM rather than calling simplify_gen_subreg.
4523 2021-09-13 Jan-Benedict Glaw <jbglaw@ług-owl.de>
4525 * config/alpha/vms.h (INIT_CUMULATIVE_ARGS): Wrap multi-statment
4526 define into a block.
4528 2021-09-13 Richard Biener <rguenther@suse.de>
4530 * config/darwin.h (DARWIN_PREFER_DWARF): Do not define.
4531 * config/i386/darwin.h (PREFERRED_DEBUGGING_TYPE): Do not
4532 change based on DARWIN_PREFER_DWARF not being defined.
4534 2021-09-13 Richard Biener <rguenther@suse.de>
4536 * config/i386/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
4537 to inherit from elfos.h
4539 2021-09-13 Richard Biener <rguenther@suse.de>
4541 * config.gcc: Add cr16-*-* to the list of obsoleted targets.
4543 2021-09-13 Richard Biener <rguenther@suse.de>
4545 * config/avr/elf.h (PREFERRED_DEBUGGING_TYPE): Remove
4546 override, pick up DWARF2_DEBUG define from elfos.h
4548 2021-09-13 Richard Biener <rguenther@suse.de>
4550 * config/rx/rx.h (PREFERRED_DEBUGGING_TYPE): Always define to
4553 2021-09-13 Richard Biener <rguenther@suse.de>
4555 * config/alpha/vms.h (PREFERRED_DEBUGGING_TYPE): Define to
4558 2021-09-13 Richard Biener <rguenther@suse.de>
4560 * config/i386/cygming.h: Always default to DWARF2 debugging.
4561 Do not define DBX_DEBUGGING_INFO, that's done via dbxcoff.h
4563 * doc/install.texi: Document binutils 2.16 as minimum
4564 requirement for mingw.
4566 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
4568 * config/rs6000/rs6000.c (struct rs6000_cost_data): New members
4569 nstmts, nloads and extra_ctor_cost.
4570 (rs6000_density_test): Add load density related heuristics. Do
4571 extra costing on vector construction statements if need.
4572 (rs6000_init_cost): Init new members.
4573 (rs6000_update_target_cost_per_stmt): New function.
4574 (rs6000_add_stmt_cost): Factor vect_nonmem hunk out to function
4575 rs6000_update_target_cost_per_stmt and call it.
4577 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
4579 * config/rs6000/rs6000.c (struct rs6000_cost_data): Remove typedef.
4580 (rs6000_init_cost): Adjust.
4582 2021-09-13 liuhongt <hongtao.liu@intel.com>
4584 * config/i386/i386.md: (UNSPEC_COPYSIGN): Remove.
4585 (UNSPEC_XORSIGN): Ditto.
4587 2021-09-12 Roger Sayle <roger@nextmovesoftware.com>
4589 * expr.c (convert_move): Preserve SUBREG_PROMOTED_VAR_P when
4590 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
4593 2021-09-11 Aldy Hernandez <aldyh@redhat.com>
4595 * tree-ssa-threadbackward.c (class back_threader_registry): Use
4596 back_jt_path_registry.
4597 * tree-ssa-threadedge.c (jump_threader::jump_threader): Use
4598 fwd_jt_path_registry.
4599 * tree-ssa-threadedge.h (class jump_threader): Same..
4600 * tree-ssa-threadupdate.c
4601 (jump_thread_path_registry::jump_thread_path_registry): Rename...
4602 (jt_path_registry::jt_path_registry): ...to this.
4603 (jump_thread_path_registry::~jump_thread_path_registry): Rename...
4604 (jt_path_registry::~jt_path_registry): ...this.
4605 (fwd_jt_path_registry::fwd_jt_path_registry): New.
4606 (fwd_jt_path_registry::~fwd_jt_path_registry): New.
4607 (jump_thread_path_registry::allocate_thread_edge): Rename...
4608 (jt_path_registry::allocate_thread_edge): ...to this.
4609 (jump_thread_path_registry::allocate_thread_path): Rename...
4610 (jt_path_registry::allocate_thread_path): ...to this.
4611 (jump_thread_path_registry::lookup_redirection_data): Rename...
4612 (fwd_jt_path_registry::lookup_redirection_data): ...to this.
4613 (jump_thread_path_registry::thread_block_1): Rename...
4614 (fwd_jt_path_registry::thread_block_1): ...to this.
4615 (jump_thread_path_registry::thread_block): Rename...
4616 (fwd_jt_path_registry::thread_block): ...to this.
4617 (jt_path_registry::thread_through_loop_header): Rename...
4618 (fwd_jt_path_registry::thread_through_loop_header): ...to this.
4619 (jump_thread_path_registry::mark_threaded_blocks): Rename...
4620 (fwd_jt_path_registry::mark_threaded_blocks): ...to this.
4621 (jump_thread_path_registry::debug_path): Rename...
4622 (jt_path_registry::debug_path): ...to this.
4623 (jump_thread_path_registry::dump): Rename...
4624 (jt_path_registry::debug): ...to this.
4625 (jump_thread_path_registry::rewire_first_differing_edge): Rename...
4626 (back_jt_path_registry::rewire_first_differing_edge): ...to this.
4627 (jump_thread_path_registry::adjust_paths_after_duplication): Rename...
4628 (back_jt_path_registry::adjust_paths_after_duplication): ...to this.
4629 (jump_thread_path_registry::duplicate_thread_path): Rename...
4630 (back_jt_path_registry::duplicate_thread_path): ...to this. Also,
4631 drop ill-formed candidates.
4632 (jump_thread_path_registry::remove_jump_threads_including): Rename...
4633 (fwd_jt_path_registry::remove_jump_threads_including): ...to this.
4634 (jt_path_registry::thread_through_all_blocks): New.
4635 (back_jt_path_registry::update_cfg): New.
4636 (fwd_jt_path_registry::update_cfg): New.
4637 (jump_thread_path_registry::register_jump_thread): Rename...
4638 (jt_path_registry::register_jump_thread): ...to this.
4639 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
4641 (class jt_path_registry): ...here.
4642 (class fwd_jt_path_registry): New.
4643 (class back_jt_path_registry): New.
4645 2021-09-10 liuhongt <hongtao.liu@intel.com>
4648 2021-09-01 liuhongt <hongtao.liu@intel.com>
4650 * emit-rtl.c (validate_subreg): Get rid of all float-int
4653 2021-09-10 Jakub Jelinek <jakub@redhat.com>
4655 * tree-core.h (enum omp_memory_order): Add OMP_MEMORY_ORDER_MASK,
4656 OMP_FAIL_MEMORY_ORDER_UNSPECIFIED, OMP_FAIL_MEMORY_ORDER_RELAXED,
4657 OMP_FAIL_MEMORY_ORDER_ACQUIRE, OMP_FAIL_MEMORY_ORDER_RELEASE,
4658 OMP_FAIL_MEMORY_ORDER_ACQ_REL, OMP_FAIL_MEMORY_ORDER_SEQ_CST and
4659 OMP_FAIL_MEMORY_ORDER_MASK enumerators.
4660 (OMP_FAIL_MEMORY_ORDER_SHIFT): Define.
4661 * gimple-pretty-print.c (dump_gimple_omp_atomic_load,
4662 dump_gimple_omp_atomic_store): Print [weak] for weak atomic
4664 * gimple.h (enum gf_mask): Change GF_OMP_ATOMIC_MEMORY_ORDER
4665 to 6-bit mask, adjust GF_OMP_ATOMIC_NEED_VALUE value and add
4667 (gimple_omp_atomic_weak_p, gimple_omp_atomic_set_weak): New inline
4669 * tree.h (OMP_ATOMIC_WEAK): Define.
4670 * tree-pretty-print.c (dump_omp_atomic_memory_order): Adjust for
4671 fail memory order being encoded in the same enum and also print
4672 fail clause if present.
4673 (dump_generic_node): Print weak clause if OMP_ATOMIC_WEAK.
4674 * gimplify.c (goa_stabilize_expr): Add target_expr and rhs arguments,
4675 handle pre_p == NULL case as a test mode that only returns value
4676 but doesn't change gimplify nor change anything otherwise, adjust
4677 recursive calls, add MODIFY_EXPR, ADDR_EXPR, COND_EXPR, TARGET_EXPR
4678 and CALL_EXPR handling, adjust COMPOUND_EXPR handling for
4679 __builtin_clear_padding calls, for !rhs gimplify as lvalue rather
4681 (gimplify_omp_atomic): Adjust goa_stabilize_expr caller. Handle
4682 COND_EXPR rhs. Set weak flag on gimple load/store for
4684 * omp-expand.c (omp_memory_order_to_fail_memmodel): New function.
4685 (omp_memory_order_to_memmodel): Adjust for fail clause encoded
4687 (expand_omp_atomic_cas): New function.
4688 (expand_omp_atomic_pipeline): Use omp_memory_order_to_fail_memmodel
4690 (expand_omp_atomic): Attempt to optimize atomic compare and exchange
4691 using expand_omp_atomic_cas.
4693 2021-09-10 Aldy Hernandez <aldyh@redhat.com>
4694 Michael Matz <matz@suse.de>
4696 * tree-pass.h (PROP_loop_opts_done): New.
4697 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
4698 Intersect with global range.
4699 * tree-ssa-loop.c (tree_ssa_loop_done): Set PROP_loop_opts_done.
4700 * tree-ssa-threadbackward.c
4701 (back_threader_profitability::profitable_path_p): Disable
4702 threading through latches until after loop optimizations have run.
4704 2021-09-10 David Faust <david.faust@oracle.com>
4706 * doc/invoke.texi: Document BPF -mcpu, -mjmpext, -mjmp32 and -malu32
4709 2021-09-10 David Faust <david.faust@oracle.com>
4711 * config/bpf/bpf-opts.h (bpf_isa_version): New enum.
4712 * config/bpf/bpf-protos.h (bpf_expand_cbranch): New.
4713 * config/bpf/bpf.c (bpf_option_override): Handle -mcpu option.
4714 (bpf_expand_cbranch): New function.
4715 * config/bpf/bpf.md (AM mode iterator): Conditionalize support for SI
4717 (zero_extendsidi2): Only use mov32 instruction if it is available.
4718 (SIM mode iterator): Conditionalize support for SI mode.
4719 (JM mode iterator): New.
4720 (cbranchdi4): Update name, use new JM iterator. Use bpf_expand_cbranch.
4721 (*branch_on_di): Update name, use new JM iterator.
4722 * config/bpf/bpf.opt: (mjmpext): New option.
4726 (bpf_isa): New enum.
4728 2021-09-10 David Faust <david.faust@oracle.com>
4730 * config/bpf/bpf.md (zero_extendhidi2): Add new output template
4731 for register-to-register extensions.
4732 (zero_extendqidi2): Likewise.
4734 2021-09-10 Richard Biener <rguenther@suse.de>
4736 PR middle-end/102273
4737 * internal-fn.c (expand_DEFERRED_INIT): Always expand non-SSA vars.
4739 2021-09-10 Richard Biener <rguenther@suse.de>
4741 PR middle-end/102269
4742 * gimplify.c (is_var_need_auto_init): Empty types do not need
4745 2021-09-10 Richard Biener <rguenther@suse.de>
4747 * configure.ac (--with-stabs): Remove.
4748 * configure: Regenerate.
4749 * doc/install.texi: Remove --with-stabs documentation.
4751 2021-09-10 liuhongt <hongtao.liu@intel.com>
4753 * config/i386/avx512fp16intrin.h: (_mm512_cmp_ph_mask):
4755 (_mm512_mask_cmp_ph_mask): Likewise.
4756 (_mm512_cmp_round_ph_mask): Likewise.
4757 (_mm512_mask_cmp_round_ph_mask): Likewise.
4758 (_mm_cmp_sh_mask): Likewise.
4759 (_mm_mask_cmp_sh_mask): Likewise.
4760 (_mm_cmp_round_sh_mask): Likewise.
4761 (_mm_mask_cmp_round_sh_mask): Likewise.
4762 (_mm_comieq_sh): Likewise.
4763 (_mm_comilt_sh): Likewise.
4764 (_mm_comile_sh): Likewise.
4765 (_mm_comigt_sh): Likewise.
4766 (_mm_comige_sh): Likewise.
4767 (_mm_comineq_sh): Likewise.
4768 (_mm_ucomieq_sh): Likewise.
4769 (_mm_ucomilt_sh): Likewise.
4770 (_mm_ucomile_sh): Likewise.
4771 (_mm_ucomigt_sh): Likewise.
4772 (_mm_ucomige_sh): Likewise.
4773 (_mm_ucomineq_sh): Likewise.
4774 (_mm_comi_round_sh): Likewise.
4775 (_mm_comi_sh): Likewise.
4776 * config/i386/avx512fp16vlintrin.h (_mm_cmp_ph_mask): New intrinsic.
4777 (_mm_mask_cmp_ph_mask): Likewise.
4778 (_mm256_cmp_ph_mask): Likewise.
4779 (_mm256_mask_cmp_ph_mask): Likewise.
4780 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
4781 * config/i386/i386-builtin.def: Add corresponding new builtins.
4782 * config/i386/i386-expand.c
4783 (ix86_expand_args_builtin): Handle new builtin types.
4784 (ix86_expand_round_builtin): Ditto.
4785 * config/i386/i386.md (ssevecmode): Add HF mode.
4786 (MODEFH): New mode iterator.
4787 * config/i386/sse.md
4788 (V48H_AVX512VL): New mode iterator to support HF vector modes.
4789 Ajdust corresponding description.
4790 (ssecmpintprefix): New.
4791 (VI12_AVX512VL): Adjust to support HF vector modes.
4792 (cmp_imm_predicate): Likewise.
4793 (<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
4795 (avx512f_vmcmp<mode>3<round_saeonly_name>): Likewise.
4796 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Likewise.
4797 (<sse>_<unord>comi<round_saeonly_name>): Likewise.
4799 2021-09-10 liuhongt <hongtao.liu@intel.com>
4801 * config/i386/avx512fp16intrin.h: (_mm512_max_ph): New intrinsic.
4802 (_mm512_mask_max_ph): Likewise.
4803 (_mm512_maskz_max_ph): Likewise.
4804 (_mm512_min_ph): Likewise.
4805 (_mm512_mask_min_ph): Likewise.
4806 (_mm512_maskz_min_ph): Likewise.
4807 (_mm512_max_round_ph): Likewise.
4808 (_mm512_mask_max_round_ph): Likewise.
4809 (_mm512_maskz_max_round_ph): Likewise.
4810 (_mm512_min_round_ph): Likewise.
4811 (_mm512_mask_min_round_ph): Likewise.
4812 (_mm512_maskz_min_round_ph): Likewise.
4813 (_mm_max_sh): Likewise.
4814 (_mm_mask_max_sh): Likewise.
4815 (_mm_maskz_max_sh): Likewise.
4816 (_mm_min_sh): Likewise.
4817 (_mm_mask_min_sh): Likewise.
4818 (_mm_maskz_min_sh): Likewise.
4819 (_mm_max_round_sh): Likewise.
4820 (_mm_mask_max_round_sh): Likewise.
4821 (_mm_maskz_max_round_sh): Likewise.
4822 (_mm_min_round_sh): Likewise.
4823 (_mm_mask_min_round_sh): Likewise.
4824 (_mm_maskz_min_round_sh): Likewise.
4825 * config/i386/avx512fp16vlintrin.h (_mm_max_ph): New intrinsic.
4826 (_mm256_max_ph): Likewise.
4827 (_mm_mask_max_ph): Likewise.
4828 (_mm256_mask_max_ph): Likewise.
4829 (_mm_maskz_max_ph): Likewise.
4830 (_mm256_maskz_max_ph): Likewise.
4831 (_mm_min_ph): Likewise.
4832 (_mm256_min_ph): Likewise.
4833 (_mm_mask_min_ph): Likewise.
4834 (_mm256_mask_min_ph): Likewise.
4835 (_mm_maskz_min_ph): Likewise.
4836 (_mm256_maskz_min_ph): Likewise.
4837 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
4838 * config/i386/i386-builtin.def: Add corresponding new builtins.
4839 * config/i386/i386-expand.c
4840 (ix86_expand_args_builtin): Handle new builtin types.
4841 * config/i386/sse.md
4842 (<code><mode>3<mask_name><round_saeonly_name>): Adjust to
4843 support HF vector modes.
4844 (*<code><mode>3<mask_name><round_saeonly_name>): Likewise.
4845 (ieee_<ieee_maxmin><mode>3<mask_name><round_saeonly_name>):
4847 (<sse>_vm<code><mode>3<mask_scalar_name><round_saeonly_scalar_name>):
4849 * config/i386/subst.md (round_saeonly_mode512bit_condition):
4850 Adjust for HF vector modes.
4852 2021-09-10 Liu, Hongtao <hongtao.liu@intel.com>
4854 * config/i386/avx512fp16intrin.h (_mm_add_sh): New intrinsic.
4855 (_mm_mask_add_sh): Likewise.
4856 (_mm_maskz_add_sh): Likewise.
4857 (_mm_sub_sh): Likewise.
4858 (_mm_mask_sub_sh): Likewise.
4859 (_mm_maskz_sub_sh): Likewise.
4860 (_mm_mul_sh): Likewise.
4861 (_mm_mask_mul_sh): Likewise.
4862 (_mm_maskz_mul_sh): Likewise.
4863 (_mm_div_sh): Likewise.
4864 (_mm_mask_div_sh): Likewise.
4865 (_mm_maskz_div_sh): Likewise.
4866 (_mm_add_round_sh): Likewise.
4867 (_mm_mask_add_round_sh): Likewise.
4868 (_mm_maskz_add_round_sh): Likewise.
4869 (_mm_sub_round_sh): Likewise.
4870 (_mm_mask_sub_round_sh): Likewise.
4871 (_mm_maskz_sub_round_sh): Likewise.
4872 (_mm_mul_round_sh): Likewise.
4873 (_mm_mask_mul_round_sh): Likewise.
4874 (_mm_maskz_mul_round_sh): Likewise.
4875 (_mm_div_round_sh): Likewise.
4876 (_mm_mask_div_round_sh): Likewise.
4877 (_mm_maskz_div_round_sh): Likewise.
4878 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
4879 * config/i386/i386-builtin.def: Add corresponding new builtins.
4880 * config/i386/i386-expand.c
4881 (ix86_expand_round_builtin): Handle new builtins.
4882 * config/i386/sse.md (VF_128): Change description.
4883 (<sse>_vm<plusminus_insn><mode>3<mask_scalar_name><round_scalar_name>):
4884 Adjust to support HF vector modes.
4885 (<sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name><round_scalar_name>):
4888 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
4890 * config/i386/i386-expand.c
4891 (ix86_avx256_split_vector_move_misalign): Handle V16HF mode.
4892 * config/i386/i386.c
4893 (ix86_preferred_simd_mode): Handle HF mode.
4894 * config/i386/sse.md (V_256H): New mode iterator.
4895 (avx_vextractf128<mode>): Use it.
4896 (VEC_INIT_MODE): Align vector HFmode condition to vector
4897 HImodes since there're no real HF instruction used.
4898 (VEC_INIT_HALF_MODE): Ditto.
4900 (VIHF_AVX512BW): Ditto.
4901 (*vec_extracthf): Ditto.
4902 (VEC_EXTRACT_MODE): Ditto.
4904 2021-09-10 Richard Biener <rguenther@suse.de>
4907 * config/dbx.h: Remove.
4908 * config/dbxcoff.h: Do not define PREFERRED_DEBUGGING_TYPE.
4909 * config/lynx.h: Likewise.
4911 2021-09-10 liuhongt <hongtao.liu@intel.com>
4913 * config/i386/i386-expand.c (ix86_expand_copysign): Expand
4914 right into ANDNOT + AND + IOR, using paradoxical subregs.
4915 (ix86_split_copysign_const): Remove.
4916 (ix86_split_copysign_var): Ditto.
4917 * config/i386/i386-protos.h (ix86_split_copysign_const): Dotto.
4918 (ix86_split_copysign_var): Ditto.
4919 * config/i386/i386.md (@copysign<mode>3_const): Ditto.
4920 (@copysign<mode>3_var): Ditto.
4922 2021-09-09 qing zhao <qing.zhao@oracle.com>
4924 * builtins.c (expand_builtin_memset): Make external visible.
4925 * builtins.h (expand_builtin_memset): Declare extern.
4926 * common.opt (ftrivial-auto-var-init=): New option.
4927 * doc/extend.texi: Document the uninitialized attribute.
4928 * doc/invoke.texi: Document -ftrivial-auto-var-init.
4929 * flag-types.h (enum auto_init_type): New enumerated type
4931 * gimple-fold.c (clear_padding_type): Add one new parameter.
4932 (clear_padding_union): Likewise.
4933 (clear_padding_emit_loop): Likewise.
4934 (clear_type_padding_in_mask): Likewise.
4935 (gimple_fold_builtin_clear_padding): Handle this new parameter.
4936 * gimplify.c (gimple_add_init_for_auto_var): New function.
4937 (gimple_add_padding_init_for_auto_var): New function.
4938 (is_var_need_auto_init): New function.
4939 (gimplify_decl_expr): Add initialization to automatic variables per
4941 (gimplify_call_expr): Add one new parameter for call to
4942 __builtin_clear_padding.
4943 (gimplify_init_constructor): Add padding initialization in the end.
4944 * internal-fn.c (INIT_PATTERN_VALUE): New macro.
4945 (expand_DEFERRED_INIT): New function.
4946 * internal-fn.def (DEFERRED_INIT): New internal function.
4947 * tree-cfg.c (verify_gimple_call): Verify calls to .DEFERRED_INIT.
4948 * tree-sra.c (generate_subtree_deferred_init): New function.
4949 (scan_function): Avoid setting cannot_scalarize_away_bitmap for
4950 calls to .DEFERRED_INIT.
4951 (sra_modify_deferred_init): New function.
4952 (sra_modify_function_body): Handle calls to DEFERRED_INIT specially.
4953 * tree-ssa-structalias.c (find_func_aliases_for_call): Likewise.
4954 * tree-ssa-uninit.c (warn_uninit): Handle calls to DEFERRED_INIT
4956 (check_defs): Likewise.
4957 (warn_uninitialized_vars): Likewise.
4958 * tree-ssa.c (ssa_undefined_value_p): Likewise.
4959 * tree.c (build_common_builtin_nodes): Build tree node for
4960 BUILT_IN_CLEAR_PADDING when needed.
4962 2021-09-09 Richard Biener <rguenther@suse.de>
4964 * tree-ssa-loop-im.c (fill_always_executed_in_1): Walk
4967 2021-09-09 Richard Biener <rguenther@suse.de>
4969 * tree-ssa-loop-im.c (fill_always_executed_in_1): Integrate
4970 DOM walk from get_loop_body_in_dom_order using a worklist
4973 2021-09-09 liuhongt <hongtao.liu@intel.com>
4975 * config.gcc: Add avx512fp16vlintrin.h.
4976 * config/i386/avx512fp16intrin.h: (_mm512_add_ph): New intrinsic.
4977 (_mm512_mask_add_ph): Likewise.
4978 (_mm512_maskz_add_ph): Likewise.
4979 (_mm512_sub_ph): Likewise.
4980 (_mm512_mask_sub_ph): Likewise.
4981 (_mm512_maskz_sub_ph): Likewise.
4982 (_mm512_mul_ph): Likewise.
4983 (_mm512_mask_mul_ph): Likewise.
4984 (_mm512_maskz_mul_ph): Likewise.
4985 (_mm512_div_ph): Likewise.
4986 (_mm512_mask_div_ph): Likewise.
4987 (_mm512_maskz_div_ph): Likewise.
4988 (_mm512_add_round_ph): Likewise.
4989 (_mm512_mask_add_round_ph): Likewise.
4990 (_mm512_maskz_add_round_ph): Likewise.
4991 (_mm512_sub_round_ph): Likewise.
4992 (_mm512_mask_sub_round_ph): Likewise.
4993 (_mm512_maskz_sub_round_ph): Likewise.
4994 (_mm512_mul_round_ph): Likewise.
4995 (_mm512_mask_mul_round_ph): Likewise.
4996 (_mm512_maskz_mul_round_ph): Likewise.
4997 (_mm512_div_round_ph): Likewise.
4998 (_mm512_mask_div_round_ph): Likewise.
4999 (_mm512_maskz_div_round_ph): Likewise.
5000 * config/i386/avx512fp16vlintrin.h: New header.
5001 * config/i386/i386-builtin-types.def (V16HF, V8HF, V32HF):
5002 Add new builtin types.
5003 * config/i386/i386-builtin.def: Add corresponding builtins.
5004 * config/i386/i386-expand.c
5005 (ix86_expand_args_builtin): Handle new builtin types.
5006 (ix86_expand_round_builtin): Likewise.
5007 * config/i386/immintrin.h: Include avx512fp16vlintrin.h
5008 * config/i386/sse.md (VFH): New mode_iterator.
5010 (avx512fmaskmode): Add HF vector modes.
5011 (avx512fmaskhalfmode): Likewise.
5012 (<plusminus_insn><mode>3<mask_name><round_name>): Adjust to for
5014 (*<plusminus_insn><mode>3<mask_name><round_name>): Likewise.
5015 (mul<mode>3<mask_name><round_name>): Likewise.
5016 (*mul<mode>3<mask_name><round_name>): Likewise.
5017 (div<mode>3): Likewise.
5018 (<sse>_div<mode>3<mask_name><round_name>): Likewise.
5019 * config/i386/subst.md (SUBST_V): Add HF vector modes.
5020 (SUBST_A): Likewise.
5021 (round_mode512bit_condition): Adjust for V32HFmode.
5023 2021-09-09 liuhongt <hongtao.liu@intel.com>
5026 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
5027 (reduc_plus_scal_v4sf): .. this, New define_expand.
5028 (reduc_plus_scal_v2df): .. and this, New define_expand.
5030 2021-09-09 liuhongt <hongtao.liu@intel.com>
5033 * config/i386/sse.md (*vec_extract<mode><ssescalarmodelower>_valign):
5036 2021-09-08 Jonathan Wakely <jwakely@redhat.com>
5039 * doc/trouble.texi (Copy Assignment): Fix description of
5040 behaviour and fix code in example.
5042 2021-09-08 Segher Boessenkool <segher@kernel.crashing.org>
5045 * config/rs6000/rs6000-logue.c (rs6000_emit_epilogue): For ELFv2 use
5046 r11 instead of r12 for restoring CR.
5048 2021-09-08 Jakub Jelinek <jakub@redhat.com>
5049 liuhongt <hongtao.liu@intel.com>
5052 * config/i386/i386.md (@xorsign<mode>3_1): Remove.
5053 * config/i386/i386-expand.c (ix86_expand_xorsign): Expand right away
5054 into AND with mask and XOR, using paradoxical subregs.
5055 (ix86_split_xorsign): Remove.
5056 * config/i386/i386-protos.h (ix86_split_xorsign): Remove.
5058 2021-09-08 Di Zhao <dizhao@os.amperecomputing.com>
5060 * tree-ssa-sccvn.c (vn_nary_op_insert_into): fix result compare
5062 2021-09-08 Jakub Jelinek <jakub@redhat.com>
5065 * config/i386/i386.md (xorsign<mode>3): If operands[1] is equal to
5066 operands[2], emit abs<mode>2 instead.
5067 (@xorsign<mode>3_1): Add early-clobbers for output operand, enable
5068 first alternative even for avx, add another alternative with
5069 =&Yv <- 0, Yv, Yvm constraints.
5070 * config/i386/i386-expand.c (ix86_split_xorsign): If op0 is equal
5071 to op1, emit vpandn instead.
5073 2021-09-08 liuhongt <hongtao.liu@intel.com>
5075 * config/i386/avx512fp16intrin.h (_mm_set_ph): New intrinsic.
5076 (_mm256_set_ph): Likewise.
5077 (_mm512_set_ph): Likewise.
5078 (_mm_setr_ph): Likewise.
5079 (_mm256_setr_ph): Likewise.
5080 (_mm512_setr_ph): Likewise.
5081 (_mm_set1_ph): Likewise.
5082 (_mm256_set1_ph): Likewise.
5083 (_mm512_set1_ph): Likewise.
5084 (_mm_setzero_ph): Likewise.
5085 (_mm256_setzero_ph): Likewise.
5086 (_mm512_setzero_ph): Likewise.
5087 (_mm_set_sh): Likewise.
5088 (_mm_load_sh): Likewise.
5089 (_mm_store_sh): Likewise.
5090 * config/i386/i386-builtin-types.def (V8HF): New type.
5091 (DEF_FUNCTION_TYPE (V8HF, V8HI)): New builtin function type
5092 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
5093 Support vector HFmodes.
5094 (ix86_expand_vector_init_one_nonzero): Likewise.
5095 (ix86_expand_vector_init_one_var): Likewise.
5096 (ix86_expand_vector_init_interleave): Likewise.
5097 (ix86_expand_vector_init_general): Likewise.
5098 (ix86_expand_vector_set): Likewise.
5099 (ix86_expand_vector_extract): Likewise.
5100 (ix86_expand_vector_init_concat): Likewise.
5101 (ix86_expand_sse_movcc): Handle vector HFmodes.
5102 (ix86_expand_vector_set_var): Ditto.
5103 * config/i386/i386-modes.def: Add HF vector modes in comment.
5104 * config/i386/i386.c (classify_argument): Add HF vector modes.
5105 (ix86_hard_regno_mode_ok): Allow HF vector modes for AVX512FP16.
5106 (ix86_vector_mode_supported_p): Likewise.
5107 (ix86_set_reg_reg_cost): Handle vector HFmode.
5108 (ix86_get_ssemov): Handle vector HFmode.
5109 (function_arg_advance_64): Pass unamed V16HFmode and V32HFmode
5111 (function_arg_advance_32): Pass V8HF/V16HF/V32HF by sse reg for 32bit
5113 (function_arg_advance_32): Ditto.
5114 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): New.
5115 (VALID_AVX256_REG_OR_OI_MODE): Rename to ..
5116 (VALID_AVX256_REG_OR_OI_VHF_MODE): .. this, and add V16HF.
5117 (VALID_SSE2_REG_VHF_MODE): New.
5118 (VALID_AVX512VL_128_REG_MODE): Add V8HF and TImode.
5119 (SSE_REG_MODE_P): Add vector HFmode.
5120 * config/i386/i386.md (mode): Add HF vector modes.
5121 (MODE_SIZE): Likewise.
5122 (ssemodesuffix): Add ph suffix for HF vector modes.
5123 * config/i386/sse.md (VFH_128): New mode iterator.
5124 (VMOVE): Adjust for HF vector modes.
5126 (V_256_512): Likewise.
5128 (avx512fmaskmode): Likewise.
5129 (shuffletype): Likewise.
5130 (sseinsnmode): Likewise.
5131 (ssedoublevecmode): Likewise.
5132 (ssehalfvecmode): Likewise.
5133 (ssehalfvecmodelower): Likewise.
5134 (ssePScmode): Likewise.
5135 (ssescalarmode): Likewise.
5136 (ssescalarmodelower): Likewise.
5137 (sseintprefix): Likewise.
5139 (bcstscalarsuff): Likewise.
5140 (xtg_mode): Likewise.
5141 (VI12HF_AVX512VL): New mode_iterator.
5142 (VF_AVX512FP16): Likewise.
5144 (VIHF_256): Likewise.
5145 (VIHF_AVX512BW): Likewise.
5146 (V16_256): Likewise.
5147 (V32_512): Likewise.
5148 (sseintmodesuffix): New mode_attr.
5149 (sse): Add scalar and vector HFmodes.
5150 (ssescalarmode): Add vector HFmode mapping.
5151 (ssescalarmodesuffix): Add sh suffix for HFmode.
5152 (*<sse>_vm<insn><mode>3): Use VFH_128.
5153 (*<sse>_vm<multdiv_mnemonic><mode>3): Likewise.
5154 (*ieee_<ieee_maxmin><mode>3): Likewise.
5155 (<avx512>_blendm<mode>): New define_insn.
5156 (vec_setv8hf): New define_expand.
5157 (vec_set<mode>_0): New define_insn for HF vector set.
5158 (*avx512fp16_movsh): Likewise.
5159 (avx512fp16_movsh): Likewise.
5160 (vec_extract_lo_v32hi): Rename to ...
5161 (vec_extract_lo_<mode>): ... this, and adjust to allow HF
5163 (vec_extract_hi_v32hi): Likewise.
5164 (vec_extract_hi_<mode>): Likewise.
5165 (vec_extract_lo_v16hi): Likewise.
5166 (vec_extract_lo_<mode>): Likewise.
5167 (vec_extract_hi_v16hi): Likewise.
5168 (vec_extract_hi_<mode>): Likewise.
5169 (vec_set_hi_v16hi): Likewise.
5170 (vec_set_hi_<mode>): Likewise.
5171 (vec_set_lo_v16hi): Likewise.
5172 (vec_set_lo_<mode>): Likewise.
5173 (*vec_extract<mode>_0): New define_insn_and_split for HF
5175 (*vec_extracthf): New define_insn.
5176 (VEC_EXTRACT_MODE): Add HF vector modes.
5177 (PINSR_MODE): Add V8HF.
5178 (sse2p4_1): Likewise.
5179 (pinsr_evex_isa): Likewise.
5180 (<sse2p4_1>_pinsr<ssemodesuffix>): Adjust to support
5181 insert for V8HFmode.
5182 (pbroadcast_evex_isa): Add HF vector modes.
5183 (AVX2_VEC_DUP_MODE): Likewise.
5184 (VEC_INIT_MODE): Likewise.
5185 (VEC_INIT_HALF_MODE): Likewise.
5186 (avx2_pbroadcast<mode>): Adjust to support HF vector mode
5188 (avx2_pbroadcast<mode>_1): Likewise.
5189 (<avx512>_vec_dup<mode>_1): Likewise.
5190 (<avx512>_vec_dup<mode><mask_name>): Likewise.
5191 (<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>):
5194 2021-09-08 Guo, Xuepeng <xuepeng.guo@intel.com>
5195 H.J. Lu <hongjiu.lu@intel.com>
5196 Liu Hongtao <hongtao.liu@intel.com>
5197 Wang Hongyu <hongyu.wang@intel.com>
5198 Xu Dianhong <dianhong.xu@intel.com>
5200 * common/config/i386/cpuinfo.h (get_available_features):
5201 Detect FEATURE_AVX512FP16.
5202 * common/config/i386/i386-common.c
5203 (OPTION_MASK_ISA_AVX512FP16_SET,
5204 OPTION_MASK_ISA_AVX512FP16_UNSET,
5205 OPTION_MASK_ISA2_AVX512FP16_SET,
5206 OPTION_MASK_ISA2_AVX512FP16_UNSET): New.
5207 (OPTION_MASK_ISA2_AVX512BW_UNSET,
5208 OPTION_MASK_ISA2_AVX512BF16_UNSET): Add AVX512FP16.
5209 (ix86_handle_option): Handle -mavx512fp16.
5210 * common/config/i386/i386-cpuinfo.h (enum processor_features):
5211 Add FEATURE_AVX512FP16.
5212 * common/config/i386/i386-isas.h: Add entry for AVX512FP16.
5213 * config.gcc: Add avx512fp16intrin.h.
5214 * config/i386/avx512fp16intrin.h: New intrinsic header.
5215 * config/i386/cpuid.h: Add bit_AVX512FP16.
5216 * config/i386/i386-builtin-types.def: (FLOAT16): New primitive type.
5217 * config/i386/i386-builtins.c: Support _Float16 type for i386
5219 (ix86_register_float16_builtin_type): New function.
5220 (ix86_float16_type_node): New.
5221 * config/i386/i386-c.c (ix86_target_macros_internal): Define
5223 * config/i386/i386-expand.c (ix86_expand_branch): Support
5225 (ix86_prepare_fp_compare_args): Adjust TARGET_SSE_MATH &&
5226 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
5227 (ix86_expand_fp_movcc): Ditto.
5228 * config/i386/i386-isa.def: Add PTA define for AVX512FP16.
5229 * config/i386/i386-options.c (isa2_opts): Add -mavx512fp16.
5230 (ix86_valid_target_attribute_inner_p): Add avx512fp16 attribute.
5231 * config/i386/i386.c (ix86_get_ssemov): Use
5232 vmovdqu16/vmovw/vmovsh for HFmode/HImode scalar or vector.
5233 (ix86_get_excess_precision): Use
5234 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when TARGET_AVX512FP16
5236 (sse_store_index): Use SFmode cost for HFmode cost.
5237 (inline_memory_move_cost): Add HFmode, and perfer SSE cost over
5238 GPR cost for HFmode.
5239 (ix86_hard_regno_mode_ok): Allow HImode in sse register.
5240 (ix86_mangle_type): Add manlging for _Float16 type.
5241 (inline_secondary_memory_needed): No memory is needed for
5242 16bit movement between gpr and sse reg under
5244 (ix86_multiplication_cost): Adjust TARGET_SSE_MATH &&
5245 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
5246 (ix86_division_cost): Ditto.
5247 (ix86_rtx_costs): Ditto.
5248 (ix86_add_stmt_cost): Ditto.
5249 (ix86_optab_supported_p): Ditto.
5250 * config/i386/i386.h (VALID_AVX512F_SCALAR_MODE): Add HFmode.
5251 (SSE_FLOAT_MODE_SSEMATH_OR_HF_P): Add HFmode.
5252 (PTA_SAPPHIRERAPIDS): Add PTA_AVX512FP16.
5253 * config/i386/i386.md (mode): Add HFmode.
5254 (MODE_SIZE): Add HFmode.
5255 (isa): Add avx512fp16.
5256 (enabled): Handle avx512fp16.
5257 (ssemodesuffix): Add sh suffix for HFmode.
5258 (comm): Add mult, div.
5259 (plusminusmultdiv): New code iterator.
5260 (insn): Add mult, div.
5261 (*movhf_internal): Adjust for avx512fp16 instruction.
5262 (*movhi_internal): Ditto.
5263 (*cmpi<unord>hf): New define_insn for HFmode.
5264 (*ieee_s<ieee_maxmin>hf3): Likewise.
5265 (extendhf<mode>2): Likewise.
5266 (trunc<mode>hf2): Likewise.
5267 (float<floatunssuffix><mode>hf2): Likewise.
5268 (*<insn>hf): Likewise.
5269 (cbranchhf4): New expander.
5270 (movhfcc): Likewise.
5271 (<insn>hf3): Likewise.
5274 * config/i386/i386.opt: Add mavx512fp16.
5275 * config/i386/immintrin.h: Include avx512fp16intrin.h.
5276 * doc/invoke.texi: Add mavx512fp16.
5277 * doc/extend.texi: Add avx512fp16 Usage Notes.
5279 2021-09-08 liuhongt <hongtao.liu@intel.com>
5281 * common.opt: Support -fexcess-precision=16.
5282 * config/aarch64/aarch64.c (aarch64_excess_precision): Return
5283 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when
5284 EXCESS_PRECISION_TYPE_FLOAT16.
5285 * config/arm/arm.c (arm_excess_precision): Ditto.
5286 * config/i386/i386.c (ix86_get_excess_precision): Ditto.
5287 * config/m68k/m68k.c (m68k_excess_precision): Issue an error
5288 when EXCESS_PRECISION_TYPE_FLOAT16.
5289 * config/s390/s390.c (s390_excess_precision): Ditto.
5290 * coretypes.h (enum excess_precision_type): Add
5291 EXCESS_PRECISION_TYPE_FLOAT16.
5292 * doc/tm.texi (TARGET_C_EXCESS_PRECISION): Update documents.
5293 * doc/tm.texi.in (TARGET_C_EXCESS_PRECISION): Ditto.
5294 * doc/extend.texi (Half-Precision): Document
5295 -fexcess-precision=16.
5296 * flag-types.h (enum excess_precision): Add
5297 EXCESS_PRECISION_FLOAT16.
5298 * target.def (excess_precision): Update document.
5299 * tree.c (excess_precision_type): Set excess_precision_type to
5300 EXCESS_PRECISION_FLOAT16 when -fexcess-precision=16.
5302 2021-09-08 liuhongt <hongtao.liu@intel.com>
5304 * doc/extend.texi: (@node Floating Types): Adjust the wording.
5305 (@node Half-Precision): Ditto.
5307 2021-09-07 Takayuki 'January June' Suwa <jjsuwa_sys3175@yahoo.co.jp>
5310 * config/xtensa/xtensa.c (xtensa_emit_move_sequence): Add
5311 'CONST_INT_P (src)' to the condition of the block that tries to
5312 eliminate literal when loading integer contant.
5314 2021-09-07 David Faust <david.faust@oracle.com>
5316 * doc/extend.texi (BPF Type Attributes) New node.
5317 Document new preserve_access_index attribute.
5318 Document new preserve_access_index builtin.
5319 * doc/invoke.texi: Document -mco-re and -mno-co-re options.
5321 2021-09-07 David Faust <david.faust@oracle.com>
5323 * config/bpf/bpf.c: Adjust includes.
5324 (bpf_handle_preserve_access_index_attribute): New function.
5325 (bpf_attribute_table): Use it here.
5326 (bpf_builtins): Add BPF_BUILTIN_PRESERVE_ACCESS_INDEX.
5327 (bpf_option_override): Handle "-mco-re" option.
5328 (bpf_asm_init_sections): New.
5329 (TARGET_ASM_INIT_SECTIONS): Redefine.
5330 (bpf_file_end): New.
5331 (TARGET_ASM_FILE_END): Redefine.
5332 (bpf_init_builtins): Add "__builtin_preserve_access_index".
5333 (bpf_core_compute, bpf_core_get_index): New.
5334 (is_attr_preserve_access): New.
5335 (bpf_expand_builtin): Handle new builtins.
5336 (bpf_core_newdecl, bpf_core_is_maybe_aggregate_access): New.
5337 (bpf_core_walk): New.
5338 (bpf_resolve_overloaded_builtin): New.
5339 (TARGET_RESOLVE_OVERLOADED_BUILTIN): Redefine.
5341 (pass_bpf_core_attr): New RTL pass.
5342 * config/bpf/bpf-passes.def: New file.
5343 * config/bpf/bpf-protos.h (make_pass_bpf_core_attr): New.
5344 * config/bpf/coreout.c: New file.
5345 * config/bpf/coreout.h: Likewise.
5346 * config/bpf/t-bpf (TM_H): Add $(srcdir)/config/bpf/coreout.h.
5347 (coreout.o): New rule.
5348 (PASSES_EXTRA): Add $(srcdir)/config/bpf/bpf-passes.def.
5349 * config.gcc (bpf): Add coreout.h to extra_headers.
5350 Add coreout.o to extra_objs.
5351 Add $(srcdir)/config/bpf/coreout.c to target_gtfiles.
5353 2021-09-07 David Faust <david.faust@oracle.com>
5355 * btfout.c (get_btf_id): Function is no longer static.
5356 * ctfc.h: Expose it here.
5358 2021-09-07 David Faust <david.faust@oracle.com>
5360 * ctfc.c (ctf_lookup_tree_type): New function.
5363 2021-09-07 David Faust <david.faust@oracle.com>
5365 * ctfc.c (ctf_dtd_lookup): Function is no longer static.
5366 * ctfc.h: Analogous change.
5368 2021-09-07 David Faust <david.faust@oracle.com>
5370 * dwarf2out.c (lookup_type_die): Function is no longer static.
5371 * dwarf2out.h: Expose it here.
5373 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
5375 * dwarf2ctf.c (ctf_debug_finalize): Make it static.
5376 (ctf_debug_early_finish): New definition.
5377 (ctf_debug_finish): Likewise.
5378 * dwarf2ctf.h (ctf_debug_finalize): Remove declaration.
5379 (ctf_debug_early_finish): New declaration.
5380 (ctf_debug_finish): Likewise.
5381 * dwarf2out.c (dwarf2out_finish): Invoke ctf_debug_finish.
5382 (dwarf2out_early_finish): Invoke ctf_debug_early_finish.
5384 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
5386 * config/bpf/bpf.c (bpf_option_override): For BPF backend, disable LTO
5387 support when compiling for CO-RE.
5388 * config/bpf/bpf.opt: Add new command line option -mco-re.
5390 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
5392 * flag-types.h (enum debug_info_type): Add new enum
5393 DINFO_TYPE_BTF_WITH_CORE.
5394 (BTF_WITH_CORE_DEBUG): New bitmask.
5395 * flags.h (btf_with_core_debuginfo_p): New declaration.
5396 * opts.c (btf_with_core_debuginfo_p): New definition.
5398 2021-09-07 Jason Merrill <jason@redhat.com>
5400 * tree.h (error_operand_p): Change to inline function.
5402 2021-09-07 Aldy Hernandez <aldyh@redhat.com>
5404 * tree-ssa-threadedge.c (forwarder_block_p): Rename to...
5405 (empty_block_with_phis_p): ...this.
5406 (potentially_threadable_block): Same.
5407 (jump_threader::thread_through_normal_block): Same.
5409 2021-09-07 Eric Botcazou <ebotcazou@adacore.com>
5412 * dwarf2out.c (mark_base_types): New overloaded function.
5413 (dwarf2out_early_finish): Invoke it on the COMDAT type list as well
5414 as the compilation unit, and call move_marked_base_types afterward.
5416 2021-09-07 H.J. Lu <hjl.tools@gmail.com>
5419 * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
5421 (ix86_expand_vector_convert_uns_vsivsf): Likewise.
5423 2021-09-07 Richard Biener <rguenther@suse.de>
5425 PR tree-optimization/102226
5426 * tree-vect-loop.c (vect_transform_cycle_phi): Record
5427 the converted value for the epilogue PHI use.
5429 2021-09-07 Martin Liska <mliska@suse.cz>
5431 PR gcov-profile/80223
5432 * ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
5433 options, do not inline when no_profile_instrument_function
5434 attributes are different in early inliner. It's fine to inline
5435 it after PGO instrumentation.
5437 2021-09-07 Richard Biener <rguenther@suse.de>
5439 PR tree-optimization/101555
5440 * tree-ssa-pre.c (translate_vuse_through_block): Do not
5441 perform an alias walk to determine the validity of the
5442 mem at the start of the block which is already guaranteed
5443 by means of prune_clobbered_mems.
5444 (phi_translate_1): Pass edge to translate_vuse_through_block.
5446 2021-09-07 Xionghu Luo <luoxhu@linux.ibm.com>
5449 * config/rs6000/rs6000.md (fmod<mode>3): New define_expand.
5450 (remainder<mode>3): Likewise.
5452 2021-09-07 YunQiang Su <yunqiang.su@cipunited.com>
5454 * config/mips/mips.c (mips_file_start): add .module for
5457 2021-09-06 Roger Sayle <roger@nextmovesoftware.com>
5459 * wide-int.cc (wi::clz): Reorder tests to ensure the result
5460 is zero for all negative values.
5462 2021-09-06 Tobias Burnus <tobias@codesourcery.com>
5464 * doc/invoke.texi (-foffload-options): Fix @opindex.
5466 2021-09-06 H.J. Lu <hjl.tools@gmail.com>
5469 * config/i386/i386-expand.c (ix86_split_xorsign): Use operands[2].
5470 * config/i386/i386.md (@xorsign<mode>3_1): Add non-destructive
5471 source alternative for AVX.
5473 2021-09-06 liuhongt <hongtao.liu@intel.com>
5475 PR middle-end/102182
5476 * optabs.c (expand_fix): Add from1 to avoid from being
5479 2021-09-06 Eric Botcazou <ebotcazou@adacore.com>
5481 * dwarf2out.c (modified_type_die): Deal with all array types earlier
5482 and use local variable consistently throughout the function.
5484 2021-09-06 Jakub Jelinek <jakub@redhat.com>
5486 PR tree-optimization/102207
5487 * match.pd: Don't demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
5488 were promoted from signed to wider unsigned type.
5490 2021-09-06 Andrew Pinski <apinski@marvell.com>
5492 PR tree-optimization/63184
5493 * match.pd: Add simplification of pointer_diff of two pointer_plus
5494 with addr_expr in the first operand of each pointer_plus.
5495 Add simplificatoin of ne/eq of two pointer_plus with addr_expr
5496 in the first operand of each pointer_plus.
5498 2021-09-06 Richard Biener <rguenther@suse.de>
5500 PR tree-optimization/102176
5501 * tree-vect-slp.c (vect_slp_gather_vectorized_scalar_stmts):
5503 (vect_bb_slp_scalar_cost): Use the computed set of
5504 vectorized scalar stmts instead of relying on the out-of-date
5505 and not accurate PURE_SLP_STMT.
5506 (vect_bb_vectorization_profitable_p): Compute the set
5507 of vectorized scalar stmts.
5509 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
5511 * gimple-range-path.cc (path_range_query::range_of_stmt): Remove
5512 GIMPLE_COND special casing.
5513 (path_range_query::range_defined_in_block): Use range_of_stmt
5514 instead of calling fold_range directly.
5516 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
5518 * gimple-range-path.cc (path_range_query::range_of_expr): Set
5519 m_undefined_path when appropriate.
5520 (path_range_query::internal_range_of_expr): Copy from range_of_expr.
5521 (path_range_query::unreachable_path_p): New.
5522 (path_range_query::precompute_ranges): Set m_undefined_path.
5523 * gimple-range-path.h (path_range_query::unreachable_path_p): New.
5524 (path_range_query::internal_range_of_expr): New.
5525 * tree-ssa-threadbackward.c (back_threader::find_taken_edge_cond):
5526 Use unreachable_path_p.
5528 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
5530 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
5531 Remove argument and call find_taken_edge.
5532 (back_threader::resolve_phi): Do not calculate taken edge before
5533 calling maybe_register_path.
5534 (back_threader::find_paths_to_names): Same.
5536 2021-09-05 Jeff Law <jlaw@localhost.localdomain>
5538 * config/h8300/h8300.md (QHSI2 mode iterator): New mode iterator.
5539 * config/h8300/testcompare.md (store_c): Update name, use new
5541 (store_neg_c, store_shifted_c): New patterns.
5543 2021-09-03 Segher Boessenkool <segher@kernel.crashing.org>
5546 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): On ELFv2 use r11
5547 instead of r12 for CR save, in all cases.
5549 2021-09-03 Andrew Pinski <apinski@marvell.com>
5551 * config/aarch64/aarch64-sve-builtins.cc (register_vector_type):
5552 Handle error_mark_node as the type of the type_decl.
5554 2021-09-03 Andrew Pinski <apinski@marvell.com>
5556 * config/aarch64/aarch64-builtins.c (struct aarch64_simd_type_info):
5558 (aarch64_simd_types): Likewise.
5559 (aarch64_simd_intOI_type_node): Likewise.
5560 (aarch64_simd_intCI_type_node): Likewise.
5561 (aarch64_simd_intXI_type_node): Likewise.
5562 * config/aarch64/aarch64.h (aarch64_fp16_type_node): Likewise.
5563 (aarch64_fp16_ptr_type_node): Likewise.
5564 (aarch64_bf16_type_node): Likewise.
5565 (aarch64_bf16_ptr_type_node): Likewise.
5567 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5569 * range-op.cc (operator_minus::op1_op2_relation_effect): Abstract
5571 (minus_op1_op2_relation_effect): ...here.
5572 (class operator_pointer_diff): New.
5573 (operator_pointer_diff::op1_op2_relation_effect): Call
5574 minus_op1_op2_relation_effect.
5575 (integral_table::integral_table): Add entry for POINTER_DIFF_EXPR.
5577 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5579 * tree-ssa-threadbackward.c (back_threader::thread_through_all_blocks):
5580 Add may_peel_loop_headers.
5581 (back_threader_registry::thread_through_all_blocks): Same.
5582 (try_thread_blocks): Pass may_peel_loop_headers argument.
5583 (pass_early_thread_jumps::execute): Same.
5585 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5587 * tree-ssa-threadedge.c (has_phis_p): New.
5588 (forwarder_block_p): New.
5589 (potentially_threadable_block): Call forwarder_block_p.
5590 (jump_threader::thread_around_empty_blocks): Call has_phis_p.
5591 (jump_threader::thread_through_normal_block): Call
5594 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5596 * tree-ssa-threadbackward.c (back_threader::dump): New.
5597 (back_threader::debug): New.
5598 (back_threader_profitability::profitable_path_p): Dump blocks
5599 even if we are bailing early.
5601 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5603 * tree-ssa-threadupdate.c (cancel_thread): New.
5604 (jump_thread_path_registry::thread_block_1): Use cancel_thread.
5605 (jump_thread_path_registry::mark_threaded_blocks): Same.
5606 (jump_thread_path_registry::register_jump_thread): Same.
5608 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5610 * tree-ssa-threadedge.c (jt_state::push): Only call methods for
5611 which objects are available.
5612 (jt_state::pop): Same.
5613 (jt_state::register_equiv): Same.
5614 (jt_state::register_equivs_on_edge): Same.
5616 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5618 * tree-ssa-threadedge.c (jump_threader::thread_across_edge):
5619 Move pop until after a thread is registered.
5621 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5623 * tree-ssa-threadupdate.c (debug): New.
5625 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5627 * gimple-range-trace.cc (push_dump_file::push_dump_file): New.
5628 (push_dump_file::~push_dump_file): New.
5629 (dump_ranger): Change dump_file temporarily while dumping
5631 * gimple-range-trace.h (class push_dump_file): New.
5633 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5635 * gimple-range-trace.cc (debug_seed_ranger): Remove static.
5636 (dump_ranger): Dump function name.
5638 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5640 * gimple-range-path.cc (path_range_query::range_defined_in_block):
5641 Adjust for non-null.
5642 (path_range_query::adjust_for_non_null_uses): New.
5643 (path_range_query::precompute_ranges): Call
5644 adjust_for_non_null_uses.
5645 * gimple-range-path.h: Add m_non_null and
5646 adjust_for_non_null_uses.
5648 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5650 * gimple-range-path.cc (path_range_query::dump): Dump path
5652 (path_range_query::precompute_ranges): Dump entire path.
5654 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5656 * value-relation.cc (relation_oracle::debug): New.
5657 * value-relation.h (relation_oracle::debug): New.
5659 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5661 * tree-ssa-loop-ch.c: Remove unnecessary include file.
5663 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5665 * gimple-range-fold.cc (fold_using_range::postfold_gcond_edges):
5666 Skip statements with no defining BB.
5667 * gimple-range-path.cc (path_range_query::range_defined_in_block):
5668 Do not get confused by statements with no defining BB.
5670 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5672 * gimple-range-fold.cc (adjust_imagpart_expr): Move from
5673 gimple_range_adjustment. Add support for constants.
5674 (adjust_realpart_expr): New.
5675 (gimple_range_adjustment): Move IMAGPART_EXPR code to
5676 adjust_imagpart_expr.
5677 * range-op.cc (integral_table::integral_table): Add entry for
5680 2021-09-03 Jakub Jelinek <jakub@redhat.com>
5682 * omp-expand.c (expand_omp_atomic_pipeline): Use
5683 IFN_ATOMIC_COMPARE_EXCHANGE instead of
5684 BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_? so that memory order
5687 2021-09-03 Jakub Jelinek <jakub@redhat.com>
5690 * tree.h (DECL_FIELD_ABI_IGNORED): Changed into rvalue only macro
5691 that is false if DECL_BIT_FIELD.
5692 (SET_DECL_FIELD_ABI_IGNORED, DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD,
5693 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD): Define.
5694 * tree-streamer-out.c (pack_ts_decl_common_value_fields): For
5695 DECL_BIT_FIELD stream DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead
5696 of DECL_FIELD_ABI_IGNORED.
5697 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
5698 SET_DECL_FIELD_ABI_IGNORED instead of writing to
5699 DECL_FIELD_ABI_IGNORED and for DECL_BIT_FIELD use
5700 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead.
5701 * lto-streamer-out.c (hash_tree): For DECL_BIT_FIELD hash
5702 DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead of DECL_FIELD_ABI_IGNORED.
5704 2021-09-03 liuhongt <hongtao.liu@intel.com>
5707 * config/i386/amxbf16intrin.h : Remove macro check for __AMX_BF16__.
5708 * config/i386/amxint8intrin.h : Remove macro check for __AMX_INT8__.
5709 * config/i386/amxtileintrin.h : Remove macro check for __AMX_TILE__.
5711 2021-09-02 Martin Sebor <msebor@redhat.com>
5713 PR tree-optimization/17506
5715 * tree-ssa-uninit.c (warn_uninit): Remove conditional guarding note.
5717 2021-09-02 Richard Biener <rguenther@suse.de>
5719 * tree-ssa-loop-im.c (fill_always_executed_in_1): Refine
5720 fix for PR78185 and continue processing when leaving
5723 2021-09-02 Jakub Jelinek <jakub@redhat.com>
5725 PR tree-optimization/99591
5726 * match.pd: Demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
5729 2021-09-02 Richard Biener <rguenther@suse.de>
5732 2021-09-02 Richard Biener <rguenther@suse.de>
5734 PR tree-optimization/102155
5735 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
5736 over a part of the RPO array and do not recurse here.
5737 Dump blocks marked as always executed.
5738 (fill_always_executed_in): Walk over the RPO array and
5739 process loops whose header we run into.
5740 (loop_invariant_motion_in_fun): Compute the first RPO
5741 using rev_post_order_and_mark_dfs_back_seme in iteration
5742 order and pass that to fill_always_executed_in.
5744 2021-09-02 liuhongt <hongtao.liu@intel.com>
5746 * config/i386/i386-modes.def (FLOAT_MODE): Define ieee HFmode.
5747 * config/i386/i386.c (enum x86_64_reg_class): Add
5749 (merge_classes): Handle X86_64_SSEHF_CLASS.
5750 (examine_argument): Ditto.
5751 (construct_container): Ditto.
5752 (classify_argument): Ditto, and set HFmode/HCmode to
5754 (function_value_32): Return _FLoat16/Complex Float16 by
5756 (function_value_64): Return _Float16/Complex Float16 by SSE
5758 (ix86_print_operand): Handle CONST_DOUBLE HFmode.
5759 (ix86_secondary_reload): Require gpr as intermediate register
5760 to store _Float16 from sse register when sse4 is not
5762 (ix86_libgcc_floating_mode_supported_p): Enable _FLoat16 under
5764 (ix86_scalar_mode_supported_p): Ditto.
5765 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Defined.
5766 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add HFmode.
5767 (VALID_INT_MODE_P): Add HFmode and HCmode.
5768 * config/i386/i386.md (*pushhf_rex64): New define_insn.
5770 (*movhf_internal): Ditto.
5771 * doc/extend.texi (Half-Precision Floating Point): Documemt
5774 2021-09-02 Richard Biener <rguenther@suse.de>
5776 PR tree-optimization/102155
5777 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
5778 over a part of the RPO array and do not recurse here.
5779 Dump blocks marked as always executed.
5780 (fill_always_executed_in): Walk over the RPO array and
5781 process loops whose header we run into.
5782 (loop_invariant_motion_in_fun): Compute the first RPO
5783 using rev_post_order_and_mark_dfs_back_seme in iteration
5784 order and pass that to fill_always_executed_in.
5786 2021-09-02 YunQiang Su <syq@debian.org>
5789 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
5791 * config/mips/mips.c (mips_module_isa_name): New.
5792 mips_file_start: add .module mipsREV to all asm output
5794 2021-09-01 Jeff Law <jlaw@localhost.localdomain>
5796 PR tree-optimization/102152
5797 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Reduce a vector
5798 comparison to a scalar comparison before calling
5799 update_stmt_if_modified.
5801 2021-09-01 Andrew Pinski <apinski@marvell.com>
5804 * config/aarch64/aarch64.c (aarch64_expand_setmem):
5805 Check STRICT_ALIGNMENT before creating an overlapping
5808 2021-09-01 Martin Sebor <msebor@redhat.com>
5810 * gimple-ssa-warn-access.cc (get_size_range): Add argument.
5811 (check_access): Pass additional argument.
5812 (check_memop_access): Remove template and make a member function.
5813 (maybe_check_dealloc_call): Make a pass_waccess member function.
5814 (class pass_waccess): Add, rename, and remove members.
5815 (pass_waccess::pass_waccess): Adjust to name change.
5816 (pass_waccess::~pass_waccess): Same.
5817 (check_alloca): Make a member function.
5818 (check_alloc_size_call): Same.
5819 (check_strcat): Same.
5820 (check_strncat): Same.
5821 (check_stxcpy): Same.
5822 (check_stxncpy): Same.
5823 (check_strncmp): Same.
5824 (maybe_warn_rdwr_sizes): Rename...
5825 (pass_waccess::maybe_check_access_sizes): ...to this.
5826 (pass_waccess::check_call): Adjust to name changes.
5827 (pass_waccess::maybe_check_dealloc_call): Make a pass_waccess member
5829 (pass_waccess::execute): Adjust to name changes.
5830 * gimple-ssa-warn-access.h (check_memop_access): Remove.
5831 * pointer-query.cc (access_ref::phi): Handle null pointer.
5832 (access_ref::inform_access): Same.
5833 (pointer_query::put_ref): Modify a cached value, not a copy of it.
5834 (pointer_query::dump): New function.
5835 (compute_objsize_r): Avoid overwriting access_ref::bndrng. Cache
5837 * pointer-query.h (pointer_query::dump): Declare.
5838 * tree-ssa-strlen.c (get_range): Simplify. Use function query.
5839 (dump_strlen_info): Use function query.
5840 (printf_strlen_execute): Factor code out into pointer_query::put_ref.
5842 2021-09-01 Thomas Schwinge <thomas@codesourcery.com>
5844 * tree.c (walk_tree_1) <OMP_CLAUSE>: Simplify.
5846 2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
5848 * doc/extend.texi: Document unavailable attribute.
5849 * print-tree.c (print_node): Handle unavailable attribute.
5850 * tree-core.h (struct tree_base): Add a bit to carry unavailability.
5851 * tree.c (error_unavailable_use): New.
5852 * tree.h (TREE_UNAVAILABLE): New.
5853 (error_unavailable_use): New.
5855 2021-09-01 Jakub Jelinek <jakub@redhat.com>
5857 PR tree-optimization/102124
5858 * tree-vect-patterns.c (vect_recog_widen_op_pattern): For ORIG_CODE
5859 MINUS_EXPR, if itype is unsigned with smaller precision than type,
5860 add an extra cast to signed variant of itype to ensure sign-extension.
5862 2021-09-01 Martin Liska <mliska@suse.cz>
5864 * graph.c (draw_cfg_node_succ_edges): Do not color fallthru
5865 edges and rather use colors for TRUE and FALSE edges.
5867 2021-09-01 Richard Biener <rguenther@suse.de>
5869 PR tree-optimization/93491
5870 * tree-ssa-pre.c (compute_avail): Set BB_MAY_NOTRETURN
5871 after processing the stmt itself. Do not consider
5872 pure functions possibly not returning. Properly avoid
5873 adding possibly trapping calls to EXP_GEN when there's
5874 a preceeding possibly not returning call.
5875 * tree-ssa-sccvn.c (vn_reference_may_trap): Conservatively
5878 2021-09-01 Richard Biener <rguenther@suse.de>
5880 PR tree-optimization/102139
5881 * tree-vectorizer.h (vec_base_alignments): Adjust hash-map
5882 type to record a std::pair of the stmt-info and the innermost
5884 (dr_vec_info::group): New member.
5885 * tree-vect-data-refs.c (vect_record_base_alignment): Adjust.
5886 (vect_compute_data_ref_alignment): Verify the recorded
5887 base alignment can be used.
5888 (data_ref_pair): Remove.
5889 (dr_group_sort_cmp): Adjust.
5890 (vect_analyze_data_ref_accesses): Store the group-ID in the
5891 dr_vec_info and operate on a vector of dr_vec_infos.
5893 2021-09-01 YunQiang Su <yunqiang.su@cipunited.com>
5895 * read-md.c (md_reader::handle_enum): support value assignation.
5896 * doc/md.texi: record define_c_enum value assignation support.
5898 2021-09-01 Jakub Jelinek <jakub@redhat.com>
5900 PR tree-optimization/102141
5901 * gimple-ssa-store-merging.c (bswap_view_convert): Add BEFORE
5902 argument. If false, emit stmts after gsi instead of before, and
5904 (bswap_replace): Adjust callers. When converting output of bswap,
5905 emit VIEW_CONVERT prepratation stmts after a copy of gsi instead
5908 2021-09-01 liuhongt <hongtao.liu@intel.com>
5910 * emit-rtl.c (validate_subreg): Get rid of all float-int
5913 2021-09-01 liuhongt <hongtao.liu@intel.com>
5916 2021-08-30 liuhongt <hongtao.liu@intel.com>
5918 * expmed.c (extract_bit_field_1): Make sure we're playing with
5919 integral modes before call extract_integral_bit_field.
5920 (extract_integral_bit_field): Add a parameter of type
5921 scalar_int_mode which corresponds to of tmode.
5922 And call extract_and_convert_fixed_bit_field instead of
5923 extract_fixed_bit_field and convert_extracted_bit_field.
5924 (extract_and_convert_fixed_bit_field): New function, it's a
5925 combination of extract_fixed_bit_field and
5926 convert_extracted_bit_field.
5928 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
5930 * tree.c (walk_tree_1) <OMP_CLAUSE_TILE>: Handle three operands.
5932 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
5934 * omp-general.h (omp_is_reference): Rename to...
5935 (omp_privatize_by_reference): ... this. Adjust all users...
5936 * omp-general.c: ... here, ...
5937 * gimplify.c: ... here, ...
5938 * omp-expand.c: ... here, ...
5939 * omp-low.c: ... here.
5941 2021-08-31 Martin Sebor <msebor@redhat.com>
5943 * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Test
5944 pointer element for equality to zero, not that of the cotaining
5947 2021-08-31 Martin Sebor <msebor@redhat.com>
5949 * gcc-rich-location.h (gcc_rich_location): Make ctor explicit.
5951 2021-08-31 Martin Sebor <msebor@redhat.com>
5953 * function.h (function): Add comments.
5954 (get_range_query): Same. Add attribute returns nonnull.
5956 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
5958 * expr.c (convert_modes): Don't use subreg_promoted_mode on a
5959 SUBREG if it can't be guaranteed to a SUBREG_PROMOTED_VAR_P set.
5960 Instead use the standard (safer) is_a <scalar_int_mode> idiom.
5962 2021-08-31 Jeff Law <jlaw@localhost.localdomain>
5964 * config.gcc (cris-*-elf, cris-*-none): Remove dbxelf.h from
5966 (m32r-*-elf, m32rle-*-elf, m32r-*-linux): Likewise.
5967 (mn10300-*-*, am33_2.0-*-linux*): Likewise.
5968 (xtensa*-*-elf, xtensa*-*-linux, xtensa*-*-uclinux): Likewise.
5969 (m32c-*-elf*, m32c-*-rtems*): Likewise.
5970 * config/cris/cris.h (DBX_NO_XREFS): Remove.
5971 (DBX_CONTIN_LENGTH, DBX_CONTIN_CHAR): Likewise.
5972 * config/m32r/m32r.h (DBXOUT_SOURCE_LINE): Likewise.
5973 (DBX_DEBUGGING_INFO, DBX_CONTIN_LENGTH): Likewise.
5974 * config/mn10300/mn10300.h (DEFAULT_GDB_EXTENSIONS): Likewise.
5975 * config/mn10300/linux.h (DBX_REGISTER_NAMES): Likewise.
5977 2021-08-31 Marcel Vollweiler <marcel@codesourcery.com>
5979 * gimplify.c (gimplify_scan_omp_clauses): Error handling. 'ancestor' only
5980 allowed on target constructs and only with particular other clauses.
5981 * omp-expand.c (expand_omp_target): Output of 'sorry, not supported' if
5983 * omp-low.c (check_omp_nesting_restrictions): Error handling. No nested OpenMP
5984 structs when 'ancestor' is used.
5985 (scan_omp_1_stmt): No usage of OpenMP runtime routines in a target region when
5987 * tree-pretty-print.c (dump_omp_clause): Append 'ancestor'.
5988 * tree.h (OMP_CLAUSE_DEVICE_ANCESTOR): Define macro.
5990 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
5992 * expr.c (convert_modes): Preserve SUBREG_PROMOTED_VAR_P when
5993 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
5995 * simplify-rtx.c (simplify_unary_operation_1) [SIGN_EXTEND]:
5996 Likewise, preserve SUBREG_PROMOTED_VAR_P when creating a (wider)
5997 partial subreg from a SUBREG_PROMOTED_VAR_P subreg. Generate
5998 SIGN_EXTEND of the SUBREG_REG when a subreg would be paradoxical.
5999 [ZERO_EXTEND]: Likewise, preserve SUBREG_PROMOTED_VAR_P when
6000 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
6001 subreg. Generate ZERO_EXTEND of the SUBREG_REG when a subreg
6002 would be paradoxical.
6004 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
6006 * combine.c (combine_simplify_rtx): Avoid converting an explicit
6007 TRUNCATE into a lowpart SUBREG on !TRULY_NOOP_TRUNCATION targets.
6008 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
6010 2021-08-31 Richard Biener <rguenther@suse.de>
6012 PR tree-optimization/102142
6013 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Fix
6014 condition under which to unset the visited flag.
6016 2021-08-31 Richard Biener <rguenther@suse.de>
6018 PR middle-end/102129
6019 * tree-ssa-ter.c (find_replaceable_in_bb): Do not move
6020 possibly trapping expressions across calls.
6022 2021-08-31 Jakub Jelinek <jakub@redhat.com>
6024 PR tree-optimization/102134
6025 * tree-ssa-ccp.c (bit_value_binop) <case RSHIFT_EXPR>: If sgn is
6026 UNSIGNED and r1val | r1mask has MSB set, ensure lzcount doesn't
6029 2021-08-31 Andrew Pinski <apinski@marvell.com>
6032 * collect-utils.c (setup_signals): New declaration.
6033 * collect-utils.h (setup_signals): New function.
6034 * collect2.c (handler): Delete.
6035 (main): Instead of manually setting up the signals,
6036 just call setup_signals.
6037 * lto-wrapper.c (main): Likewise.
6039 2021-08-31 Andrew Pinski <apinski@marvell.com>
6042 * config/i386/i386-protos.h (x86_output_aligned_bss):
6043 Change align argument to unsigned type.
6044 (x86_elf_aligned_decl_common): Likewise.
6045 * config/i386/i386.c (x86_elf_aligned_decl_common): Likewise.
6046 (x86_output_aligned_bss): Likewise.
6048 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
6050 * config/mips/mips.c (mips_module_isa_name): New.
6051 mips_file_start: add .module mipsREV to all asm output
6053 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
6055 * config/mips/mips.h (struct mips_cpu_info): define enum mips_isa;
6056 use enum instead of int for 'isa' member.
6057 * config.gcc, config/mips/mips.c, config/mips/mips-cpus.def,
6058 config/mips/netbsd.h: replace hardcoded numbers with enum.
6060 2021-08-31 liuhongt <hongtao.liu@intel.com>
6062 * config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
6063 define_split to define_insn_and_split.
6064 (*avx2_eq<mode>3): Removed.
6065 (<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
6066 (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
6067 (*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
6069 (*avx2_gt<mode>3): Removed.
6070 (<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
6071 define_insn to define_expand, and adjust pattern.
6072 (UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.
6074 2021-08-30 David Malcolm <dmalcolm@redhat.com>
6077 * Makefile.in (ANALYZER_OBJS): Add analyzer/call-info.o.
6079 2021-08-30 Jason Merrill <jason@redhat.com>
6081 * doc/invoke.texi: Document -Wmissing-requires.
6083 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
6085 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Remove
6086 TARGET_EXTRA_BUILTINS guard.
6088 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
6090 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Change
6091 initialization of V2DI_type_node and unsigned_V2DI_type_node.
6093 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
6095 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Use the new
6096 decl when new_builtins_are_live.
6097 * config/rs6000/rs6000-builtin-new.def (__builtin_cfstring): New
6100 2021-08-30 Pat Haugen <pthaugen@linux.ibm.com>
6102 * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add
6103 OPTION_MASK_P10_FUSION_2STORE.
6104 (POWERPC_MASKS): Likewise.
6105 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
6106 store fusion for Power10.
6107 (is_fusable_store): New.
6108 (power10_sched_reorder): Likewise.
6109 (rs6000_sched_reorder): Do Power10 specific reordering.
6110 (rs6000_sched_reorder2): Likewise.
6111 * config/rs6000/rs6000.opt: Add new option.
6113 2021-08-30 Richard Biener <rguenther@suse.de>
6115 PR tree-optimization/102128
6116 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
6117 Move scanning for if-converted scalar code to the caller
6118 and instead delay clearing the visited flag for profitable
6120 (vect_slp_region): Cost all subgraphs before scheduling.
6121 For if-converted BB vectorization scan for scalar COND_EXPRs
6122 and do not vectorize if any found and the cost model is
6125 2021-08-30 Richard Biener <rguenther@suse.de>
6127 * common.opt (fexceptions): Mark
6128 EnabledBy(fnon-call-exceptions).
6129 * doc/invoke.texi (fnon-call-exceptions): Document this
6130 enables -fexceptions.
6132 2021-08-30 Sebastian Huber <sebastian.huber@embedded-brains.de>
6134 * tsystem.h (abort): Define abort() if inhibit_libc is defined and it
6135 is not already defined.
6137 2021-08-30 liuhongt <hongtao.liu@intel.com>
6139 * expmed.c (extract_bit_field_1): Make sure we're playing with
6140 integral modes before call extract_integral_bit_field.
6141 (extract_integral_bit_field): Add a parameter of type
6142 scalar_int_mode which corresponds to of tmode.
6143 And call extract_and_convert_fixed_bit_field instead of
6144 extract_fixed_bit_field and convert_extracted_bit_field.
6145 (extract_and_convert_fixed_bit_field): New function, it's a
6146 combination of extract_fixed_bit_field and
6147 convert_extracted_bit_field.
6149 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
6151 * config/darwin.c (darwin_libc_has_function): Do not run
6152 the checks for x86 or modern Darwin. Make sure that there
6153 is a value set for darwin_macosx_version_min before testing.
6155 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
6157 * config/i386/darwin.h (CLEAR_INSN_CACHE): New.
6159 2021-08-28 Jan Hubicka <hubicka@ucw.cz>
6161 * ipa-modref-tree.h (modref_access_node::merge): Break out
6162 logic combining offsets and logic merging ranges to ...
6163 (modref_access_node::combined_offsets): ... here
6164 (modref_access_node::update2): ... here
6165 (modref_access_node::closer_pair_p): New member function.
6166 (modref_access_node::forced_merge): New member function.
6167 (modre_ref_node::insert): Do merging when table is full.
6169 2021-08-28 YunQiang Su <yunqiang.su@cipunited.com>
6172 * config.gcc: MIPS: use N64 ABI by default if the triple end
6173 with -gnuabi64, which is used by Debian since 2013.
6175 2021-08-28 Alexandre Oliva <oliva@adacore.com>
6177 * ipa-modref.c (analyze_function): Skip debug stmts.
6178 * tree-inline.c (estimate_num_insn): Consider builtins even
6179 without a cgraph_node.
6181 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
6183 * config/h8300/bitfield.md (cstore<mode>4): Remove expander.
6184 * config/h8300/h8300.c (h8300_expand_branch): Remove function.
6185 * config/h8300/h8300-protos.h (h8300_expadn_branch): Remove prototype.
6186 * config/h8300/h8300.md (eqne): New code iterator.
6187 (geultu, geultu_to_c): Similarly.
6188 * config/h8300/testcompare.md (cstore<mode>4): Dummy expander.
6189 (store_c_<mode>, store_c_i_<mode>): New define_insn_and_splits
6190 (cmp<mode>_c): New pattern
6192 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
6194 * tree-ssa-dom.c (reduce_vector_comparison_to_scalar_comparison): New
6196 (dom_opt_dom_walker::optimize_stmt): Use it.
6198 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
6200 * config/darwin.c (finalize_ctors): Add a section-start linker-
6202 (finalize_dtors): Likewise.
6203 * config/darwin.h (MIN_LD64_INIT_TERM_START_LABELS): New.
6205 2021-08-27 Bill Schmidt <wschmidt@linux.ibm.com>
6207 * config/rs6000/rs6000-call.c (rs6000-builtins.h): New #include.
6208 (rs6000_init_builtins): Call rs6000_init_generated_builtins. Skip the
6209 old initialization logic when new builtins are enabled.
6210 * config/rs6000/rs6000-gen-builtins.c (write_decls): Rename
6211 rs6000_autoinit_builtins to rs6000_init_generated_builtins.
6212 (write_init_file): Likewise.
6214 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
6216 * configure.ac (darwin2[[0-9]]* | darwin19*): Alter use of
6217 gcc_GAS_CHECK_FEATURE to remove an extraneous parameter.
6218 (amdgcn-* | gcn-*) Likewise.
6220 2021-08-27 Anthony Sharp <anthonysharp15@gmail.com>
6222 * symbol-summary.h: Added missing template keyword.
6224 2021-08-27 Richard Biener <rguenther@suse.de>
6226 PR tree-optimization/45178
6227 * tree-ssa-dce.c (find_obviously_necessary_stmts): For
6228 infinite loops without exit do not mark control dependent
6229 edges of the latch necessary.
6231 2021-08-27 konglin1 <lingling.kong@intel.com>
6234 * config/i386/sse.md: (<avx512>scattersi<mode>): Add mask operand to
6236 (<avx512>scattersi<mode>): Likewise.
6237 (*avx512f_scattersi<VI48F:mode>): Merge mask operand to set_dest.
6238 (*avx512f_scatterdi<VI48F:mode>): Likewise
6240 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
6242 * config/rs6000/rs6000.c (rs6000_builtin_md_vectorized_function): Add
6243 support for built-in functions MISC_BUILTIN_DIVWE, MISC_BUILTIN_DIVWEU,
6244 MISC_BUILTIN_DIVDE, MISC_BUILTIN_DIVDEU, P10_BUILTIN_CFUGED,
6245 P10_BUILTIN_CNTLZDM, P10_BUILTIN_CNTTZDM, P10_BUILTIN_PDEPD and
6246 P10_BUILTIN_PEXTD on Power10.
6248 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
6250 * config/rs6000/rs6000-call.c (builtin_function_type): Add unsigned
6251 signedness for some Power10 bifs.
6253 2021-08-27 David Edelsohn <dje.gcc@gmail.com>
6256 * config/rs6000/rs6000.c (rs6000_adjust_field_align): Use
6257 computed alignment if the entire struct has attribute packed.
6259 2021-08-27 liuhongt <hongtao.liu@intel.com>
6263 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
6264 IX86_BUILTIN_SHUFPD512, IX86_BUILTIN_SHUFPS512,
6265 IX86_BUILTIN_SHUFPD256, IX86_BUILTIN_SHUFPS,
6266 IX86_BUILTIN_SHUFPS256.
6267 (ix86_masked_all_ones): New function.
6269 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
6271 * config/i386/i386.md (*btr<mode>_1): Call force_reg unconditionally.
6272 (conditional moves with memory inputs splitters): Ditto.
6273 * config/i386/sse.md (one_cmpl<mode>2): Simplify.
6275 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
6277 * ipa-modref-tree.h (modref_access_node::try_merge_with): Restart
6278 search after merging.
6280 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
6282 * config/rs6000/rs6000-overload.def: Add remaining overloads.
6284 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
6286 * config/rs6000/rs6000-builtin-new.def: Add cell stanza.
6288 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
6290 * config/rs6000/rs6000-builtin-new.def: Add ieee128-hw, dfp,
6291 crypto, and htm stanzas.
6293 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
6295 * config/rs6000/rs6000-builtin-new.def: Add mma stanza.
6297 2021-08-26 Martin Sebor <msebor@redhat.com>
6299 * tree-ssa-uninit.c (warn_uninit): Refactor and simplify.
6300 (warn_uninit_phi_uses): Remove argument from calls to warn_uninit.
6301 (warn_uninitialized_vars): Same. Reduce visibility of locals.
6302 (warn_uninitialized_phi): Same.
6304 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
6306 * tree-ssa-ccp.c (get_individual_bits): Helper function to
6307 extract the individual bits from a widest_int constant (mask).
6308 (gray_code_bit_flips): New read-only table for effiently
6309 enumerating permutations/combinations of bits.
6310 (bit_value_binop) [LROTATE_EXPR, RROTATE_EXPR]: Handle rotates
6311 by unknown counts that are guaranteed less than the target
6312 precision and four or fewer unknown bits by enumeration.
6313 [LSHIFT_EXPR, RSHIFT_EXPR]: Likewise, also handle shifts by
6314 enumeration under the same conditions. Handle remaining
6315 shifts as a mask based upon the minimum possible shift value.
6317 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
6318 Richard Biener <rguenther@suse.de>
6320 * match.pd (shift transformations): Remove a redundant
6321 !POINTER_TYPE_P check.
6323 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
6326 * config/i386/i386.md (cmove reg-reg move elimination peephole2s):
6327 Set all_regs to true in the call to replace_rtx.
6329 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
6331 * ipa-modref-tree.c (test_insert_search_collapse): Update test.
6332 * ipa-modref-tree.h (modref_base_node::insert): Be smarter when
6333 hiting --param modref-max-refs limit.
6334 (modref_tree:insert_base): Be smarter when hitting
6335 --param modref-max-bases limit. Add new parameter REF.
6336 (modref_tree:insert): Update.
6337 (modref_tree:merge): Update.
6338 * ipa-modref.c (read_modref_records): Update.
6340 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
6342 * params.opt: (modref-max-adjustments): Add full stop.
6344 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
6346 * ipa-modref-tree.h (modref_ref_node::verify): New member
6348 (modref_ref_node::insert): Use it.
6349 (modref_ref_node::try_mere_with): Fix off by one error.
6351 2021-08-26 Martin Liska <mliska@suse.cz>
6352 Stefan Kneifel <stefan.kneifel@bluewin.ch>
6354 * cgraph.h (create_version_clone_with_body): Add new parameter.
6355 * cgraphclones.c: Likewise.
6356 * multiple_target.c (create_dispatcher_calls): Do not use
6358 (create_target_clone): Likewise here.
6360 2021-08-26 Jonathan Yong <10walls@gmail.com>
6362 * doc/extend.texi: Add note about reserved priorities
6363 to the constructor attribute.
6365 2021-08-25 Martin Sebor <msebor@redhat.com>
6367 * gimple-range-cache.cc (ssa_global_cache::dump): Avoid printing
6368 range table header alone.
6369 * gimple-range.cc (gimple_ranger::export_global_ranges): Same.
6371 2021-08-25 Jan Hubicka <hubicka@ucw.cz>
6373 * doc/invoke.texi: Document --param modref-max-adjustments.
6374 * ipa-modref-tree.c (test_insert_search_collapse): Update.
6375 (test_merge): Update.
6376 * ipa-modref-tree.h (struct modref_access_node): Add adjustments;
6377 (modref_access_node::operator==): Fix handling of access ranges.
6378 (modref_access_node::contains): Constify parameter; handle also
6379 mismatched parm offsets.
6380 (modref_access_node::update): New function.
6381 (modref_access_node::merge): New function.
6382 (unspecified_modref_access_node): Update constructor.
6383 (modref_ref_node::insert_access): Add record_adjustments parameter;
6385 (modref_ref_node::try_merge_with): New private function.
6386 (modref_tree::insert): New record_adjustments parameter.
6387 (modref_tree::merge): New record_adjustments parameter.
6388 (modref_tree::copy_from): Update.
6389 * ipa-modref.c (dump_access): Dump adjustments field.
6390 (get_access): Update constructor.
6391 (record_access): Update call of insert.
6392 (record_access_lto): Update call of insert.
6393 (merge_call_side_effects): Add record_adjustments parameter.
6394 (get_access_for_fnspec): Update.
6395 (process_fnspec): Update.
6396 (analyze_call): Update.
6397 (analyze_function): Update.
6398 (read_modref_records): Update.
6399 (ipa_merge_modref_summary_after_inlining): Update.
6400 (propagate_unknown_call): Update.
6401 (modref_propagate_in_scc): Update.
6402 * params.opt (param-max-modref-adjustments=): New.
6404 2021-08-25 Michael Meissner <meissner@linux.ibm.com>
6406 * config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
6408 (xxspltiw_v4si): Use vecperm type attribute.
6409 (xxspltiw_v4si_inst): Use vecperm type attribute.
6410 (xxspltiw_v4sf_inst): Likewise.
6411 (xxspltidp_v2df): Use vecperm type attribute. Use
6412 UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
6413 (xxspltidp_v2df_inst): Likewise.
6414 (xxsplti32dx_v4si): Use vecperm type attribute.
6415 (xxsplti32dx_v4si_inst): Likewise.
6416 (xxsplti32dx_v4sf_inst): Likewise.
6417 (xxblend_<mode>): Likewise.
6418 (xxpermx): Likewise.
6419 (xxpermx_inst): Likewise.
6422 2021-08-25 Lewis Hyatt <lhyatt@gmail.com>
6425 * coretypes.h (typedef diagnostic_input_charset_callback): Declare.
6426 * diagnostic.c (diagnostic_initialize_input_context): New function.
6427 * diagnostic.h (diagnostic_initialize_input_context): Declare.
6428 * input.c (default_charset_callback): New function.
6429 (file_cache::initialize_input_context): New function.
6430 (file_cache_slot::create): Added ability to convert the input
6431 according to the input context.
6432 (file_cache::file_cache): Initialize the new input context.
6433 (class file_cache_slot): Added new m_alloc_offset member.
6434 (file_cache_slot::file_cache_slot): Initialize the new member.
6435 (file_cache_slot::~file_cache_slot): Handle potentially offset buffer.
6436 (file_cache_slot::maybe_grow): Likewise.
6437 (file_cache_slot::needs_read_p): Handle NULL fp, which is now possible.
6438 (file_cache_slot::get_next_line): Likewise.
6439 * input.h (class file_cache): Added input context member.
6441 2021-08-25 Richard Biener <rguenther@suse.de>
6443 PR tree-optimization/102046
6444 * tree-vect-slp.c (vect_build_slp_tree_2): Conservatively
6445 update ->any_pattern when swapping operands.
6447 2021-08-25 Hongyu Wang <hongyu.wang@intel.com>
6450 * config/i386/i386.c (ix86_live_on_entry): Adjust comment.
6451 (ix86_decompose_address): Remove retval check for ASHIFT,
6452 allow non-canonical zero extend if AND mask covers ASHIFT
6454 (ix86_legitimate_address_p): Adjust condition for decompose.
6455 (ix86_rtx_costs): Adjust cost for lea with non-canonical
6457 Co-Authored by: Uros Bizjak <ubizjak@gmail.com>
6459 2021-08-25 Jiufu Guo <guojiufu@linux.ibm.com>
6461 PR tree-optimization/101145
6462 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
6464 (number_of_iterations_lt): Invoke above function.
6465 (adjust_cond_for_loop_until_wrap):
6466 Merge to number_of_iterations_until_wrap.
6467 (number_of_iterations_cond): Update invokes for
6468 adjust_cond_for_loop_until_wrap and number_of_iterations_lt.
6470 2021-08-25 konglin1 <lingling.kong@intel.com>
6473 * config/i386/avx512dqintrin.h (_mm512_fpclass_ps_mask): Fix
6475 (_mm512_mask_fpclass_ps_mask): Ditto.
6477 2021-08-25 Kewen Lin <linkw@linux.ibm.com>
6479 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Remove.
6480 (vec_unpacku_hi_v8hi): Likewise.
6481 (vec_unpacku_lo_v16qi): Likewise.
6482 (vec_unpacku_lo_v8hi): Likewise.
6483 (vec_unpacku_hi_<VP_small_lc>): New define_expand.
6484 (vec_unpacku_lo_<VP_small_lc>): Likewise.
6486 2021-08-24 David Edelsohn <dje.gcc@gmail.com>
6488 * config/rs6000/aix.h (SYSTEM_IMPLICIT_EXTERN_C): Delete.
6489 * config/rs6000/aix71.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
6490 * config/rs6000/aix72.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
6491 * config/rs6000/aix73.h (TARGET_AIX_VERSION): Increase to 73.
6493 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
6495 PR middle-end/102031
6496 * simplify-rtx.c (simplify_truncation): When comparing precisions
6497 use "subreg_prec" variable, not "subreg_mode".
6499 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
6501 * config/rs6000/rs6000-builtin-new.def: Add power10 and power10-64
6504 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
6506 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Initialize
6507 various pointer type nodes.
6508 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add enum
6509 values for various pointer types.
6510 (ptr_V16QI_type_node): New macro.
6511 (ptr_V1TI_type_node): New macro.
6512 (ptr_V2DI_type_node): New macro.
6513 (ptr_V2DF_type_node): New macro.
6514 (ptr_V4SI_type_node): New macro.
6515 (ptr_V4SF_type_node): New macro.
6516 (ptr_V8HI_type_node): New macro.
6517 (ptr_unsigned_V16QI_type_node): New macro.
6518 (ptr_unsigned_V1TI_type_node): New macro.
6519 (ptr_unsigned_V8HI_type_node): New macro.
6520 (ptr_unsigned_V4SI_type_node): New macro.
6521 (ptr_unsigned_V2DI_type_node): New macro.
6522 (ptr_bool_V16QI_type_node): New macro.
6523 (ptr_bool_V8HI_type_node): New macro.
6524 (ptr_bool_V4SI_type_node): New macro.
6525 (ptr_bool_V2DI_type_node): New macro.
6526 (ptr_bool_V1TI_type_node): New macro.
6527 (ptr_pixel_type_node): New macro.
6528 (ptr_intQI_type_node): New macro.
6529 (ptr_uintQI_type_node): New macro.
6530 (ptr_intHI_type_node): New macro.
6531 (ptr_uintHI_type_node): New macro.
6532 (ptr_intSI_type_node): New macro.
6533 (ptr_uintSI_type_node): New macro.
6534 (ptr_intDI_type_node): New macro.
6535 (ptr_uintDI_type_node): New macro.
6536 (ptr_intTI_type_node): New macro.
6537 (ptr_uintTI_type_node): New macro.
6538 (ptr_long_integer_type_node): New macro.
6539 (ptr_long_unsigned_type_node): New macro.
6540 (ptr_float_type_node): New macro.
6541 (ptr_double_type_node): New macro.
6542 (ptr_long_double_type_node): New macro.
6543 (ptr_dfloat64_type_node): New macro.
6544 (ptr_dfloat128_type_node): New macro.
6545 (ptr_ieee128_type_node): New macro.
6546 (ptr_ibm128_type_node): New macro.
6547 (ptr_vector_pair_type_node): New macro.
6548 (ptr_vector_quad_type_node): New macro.
6549 (ptr_long_long_integer_type_node): New macro.
6550 (ptr_long_long_unsigned_type_node): New macro.
6552 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
6554 * config/rs6000/rs6000-builtin-new.def: Add power9-vector, power9,
6555 and power9-64 stanzas.
6557 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
6558 Tom de Vries <tdevries@suse.de>
6560 * config.gcc (nvptx-*-*): Define {c,c++}_target_objs.
6561 * config/nvptx/nvptx-protos.h (nvptx_cpu_cpp_builtins): Prototype.
6562 * config/nvptx/nvptx.h (TARGET_CPU_CPP_BUILTINS): Implement with
6563 a call to the new nvptx_cpu_cpp_builtins function in nvptx-c.c.
6564 * config/nvptx/t-nvptx (nvptx-c.o): New rule.
6565 * config/nvptx/nvptx-c.c: New source file.
6566 (nvptx_cpu_cpp_builtins): Move implementation here.
6568 2021-08-24 Martin Sebor <msebor@redhat.com>
6570 PR middle-end/101600
6571 PR middle-end/101977
6572 * gimple-ssa-warn-access.cc (maybe_warn_for_bound): Tighten up
6573 the phrasing of a warning.
6574 (check_access): Use the remaining size after subtracting any offset
6575 rather than the whole object size.
6576 * pointer-query.cc (access_ref::get_ref): Clear BASE0 flag if it's
6577 clear for any nonnull PHI argument.
6578 (compute_objsize): Clear argument.
6580 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
6582 * config/rs6000/rs6000-builtin-new.def: Add power8-vector stanza.
6584 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
6586 * config/rs6000/rs6000-builtin-new.def: Add power7 and power7-64
6589 2021-08-24 Andrew MacLeod <amacleod@redhat.com>
6591 * value-relation.cc (rr_transitive_table): New.
6592 (relation_transitive): New.
6593 (value_relation::swap): Remove.
6594 (value_relation::apply_transitive): New.
6595 (relation_oracle::relation_oracle): Allocate a new tmp bitmap.
6596 (relation_oracle::register_relation): Call register_transitives.
6597 (relation_oracle::register_transitives): New.
6598 * value-relation.h (relation_oracle): Add new temporary bitmap and
6601 2021-08-24 H.J. Lu <hjl.tools@gmail.com>
6604 * config/i386/i386-expand.c (ix86_expand_vector_move): Broadcast
6605 from integer to a pseudo vector register.
6607 2021-08-24 Richard Biener <rguenther@suse.de>
6609 PR tree-optimization/100089
6610 * tree-vectorizer.h (vect_slp_bb): Rename to ...
6611 (vect_slp_if_converted_bb): ... this and get the original
6612 loop as new argument.
6613 * tree-vectorizer.c (try_vectorize_loop_1): Revert previous fix,
6614 pass original loop to vect_slp_if_converted_bb.
6615 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
6616 If orig_loop was passed scan the not vectorized stmts
6617 for COND_EXPRs and force not profitable if found.
6618 (vect_slp_region): Pass down all SLP instances to costing
6619 if orig_loop was specified.
6620 (vect_slp_bbs): Pass through orig_loop.
6621 (vect_slp_bb): Rename to ...
6622 (vect_slp_if_converted_bb): ... this and get the original
6623 loop as new argument.
6624 (vect_slp_function): Adjust.
6626 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
6629 * config/arm/arm.md (attribute arch): Add fix_vlldm.
6630 (arch_enabled): Use it.
6631 * config/arm/vfp.md (lazy_store_multiple_insn): Add alternative to
6632 use when erratum mitigation is needed.
6634 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
6637 * config/arm/arm.opt (mfix-cmse-cve-2021-35465): New option.
6638 * doc/invoke.texi (Arm Options): Document it.
6639 * config/arm/arm-cpus.in (quirk_vlldm): New feature bit.
6640 (ALL_QUIRKS): Add quirk_vlldm.
6641 (cortex-m33): Add quirk_vlldm.
6642 (cortex-m35p, cortex-m55): Likewise.
6643 * config/arm/arm.c (arm_option_override): Enable fix_vlldm if
6644 targetting an affected CPU and not explicitly controlled on
6647 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
6649 * config/arm/vfp.md (lazy_store_multiple_insn): Rewrite as valid RTL.
6650 (lazy_load_multiple_insn): Likewise.
6652 2021-08-24 liuhongt <hongtao.liu@intel.com>
6655 * config/i386/sse.md (<avx512>_vternlog<mode><sd_maskz_name>):
6656 Enable avx512 embedded broadcast.
6657 (*<avx512>_vternlog<mode>_all): Ditto.
6658 (<avx512>_vternlog<mode>_mask): Ditto.
6660 2021-08-24 liuhongt <hongtao.liu@intel.com>
6663 * config/i386/i386.c (ix86_rtx_costs): Define cost for
6665 * config/i386/i386.h (STRIP_UNARY): New macro.
6666 * config/i386/predicates.md (reg_or_notreg_operand): New
6668 * config/i386/sse.md (*<avx512>_vternlog<mode>_all): New define_insn.
6669 (*<avx512>_vternlog<mode>_1): New pre_reload
6670 define_insn_and_split.
6671 (*<avx512>_vternlog<mode>_2): Ditto.
6672 (*<avx512>_vternlog<mode>_3): Ditto.
6673 (any_logic1,any_logic2): New code iterator.
6674 (logic_op): New code attribute.
6675 (ternlogsuffix): Extend to VNxDF and VNxSF.
6677 2021-08-24 Richard Biener <rguenther@suse.de>
6679 * doc/invoke.texi (vect-inner-loop-cost-factor): Adjust.
6680 * params.opt (--param vect-inner-loop-cost-factor): Adjust
6682 * tree-vect-loop.c (vect_analyze_loop_form): Initialize
6683 inner_loop_cost_factor to the minimum of the estimated number
6684 of iterations of the inner loop and vect-inner-loop-cost-factor.
6686 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
6687 Richard Biener <rguenther@suse.de>
6689 * config/i386/i386-features.c (compute_convert_gain): Provide
6690 more accurate values for CONST_INT, when optimizing for size.
6691 * config/i386/i386.c (COSTS_N_BYTES): Move definition from here...
6692 * config/i386/i386.h (COSTS_N_BYTES): to here.
6694 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
6695 Jakub Jelinek <jakub@redhat.com>
6697 PR middle-end/102029
6698 * match.pd (shift transformations): Add an additional check for
6699 !POINTER_TYPE_P in the recently added left shift transformation.
6701 2021-08-24 liuhongt <hongtao.liu@intel.com>
6703 PR tree-optimization/100089
6704 * tree-vectorizer.c (try_vectorize_loop_1): Disable slp in
6705 loop vectorizer when cost model is very-cheap.
6707 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
6709 * config/rs6000/rs6000-gen-builtins.c (parse_bif_entry): Don't call
6710 asprintf, which is not available on AIX.
6712 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
6714 * config.gcc (target_gtfiles): Add ./rs6000-builtins.h.
6715 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Set.
6717 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
6719 * config.gcc (powerpc*-*-*): Add rs6000-builtins.o to extra_objs.
6720 * config/rs6000/rs6000-gen-builtins.c (main): Close init_file
6722 * config/rs6000/t-rs6000 (rs6000-gen-builtins.o): New target.
6723 (rbtree.o): Likewise.
6724 (rs6000-gen-builtins): Likewise.
6725 (rs6000-builtins.c): Likewise.
6726 (rs6000-builtins.h): Likewise.
6727 (rs6000.o): Add dependency.
6728 (EXTRA_HEADERS): Add rs6000-vecdefines.h.
6729 (rs6000-vecdefines.h): New target.
6730 (rs6000-builtins.o): Likewise.
6731 (rs6000-call.o): Add rs6000-builtins.h as a dependency.
6732 (rs6000-c.o): Likewise.
6734 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
6737 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace):
6738 Diagnose buffer overrun.
6739 (safe_inc_pos): Fix overrun detection.
6740 (match_identifier): Diagnose buffer overrun.
6741 (match_integer): Likewise.
6742 (match_to_right_bracket): Likewise.
6744 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
6746 * ipa-modref-tree.h (modref_access_node::range_info_useful_p):
6747 Improve range compare.
6748 (modref_access_node::contains): New member function.
6749 (modref_access_node::search): Remove.
6750 (modref_access_node::insert): Be smarter about subaccesses.
6752 2021-08-23 Thomas Schwinge <thomas@codesourcery.com>
6754 * config/i386/i386-options.c (ix86_omp_device_kind_arch_isa)
6755 <omp_device_arch> [ACCEL_COMPILER]: Match "intel_mic".
6756 * config/i386/t-omp-device (omp-device-properties-i386) <arch>:
6759 2021-08-23 Jeff Law <jlaw@localhost.localdomain>
6761 * config/h8300/h8300-protos.h (h8300_expand_epilogue): Add new
6763 * config/h8300/jumpcall.md (call, call_value): Restrict to
6764 !SIBLING_CALL_P cases.
6765 (subcall, sibcall_value): New patterns & expanders.
6766 * config/h8300/proepi.md (epilogue): Pass new argument to
6767 h8300_expand_epilogue.
6768 (sibcall_epilogue): New expander.
6769 * config/h8300/h8300.c (h8300_expand_epilogue): Handle sibcall
6771 (h8300_ok_for_sibcall_p): New function.
6772 (TARGET_FUNCTION_OK_FOR_SIBCALL): define.
6774 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
6776 * simplify-rtx.c (simplify_unary_operation_1): [TRUNCATE]:
6777 Handle case where the operand is already the desired mode.
6779 2021-08-23 Richard Biener <rguenther@suse.de>
6782 * tree-ssa-structalias.c (ipa_pta_execute): Check in_other_partition
6783 in addition to has_gimple_body.
6785 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
6787 PR middle-end/101949
6788 * ipa-modref.c (analyze_ssa_name_flags): Fix merging of
6791 2021-08-23 Martin Liska <mliska@suse.cz>
6793 * doc/invoke.texi: Put the option out of -mxl-mode-app-model
6796 2021-08-23 Richard Biener <rguenther@suse.de>
6798 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6799 Properly scale the inner loop cost only once.
6801 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
6803 * tree-ssa-ccp.c (bit_value_binop) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]:
6804 Provide bounds for unsigned (and signed with non-negative operands)
6805 division and modulus.
6807 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
6809 * simplify-rtx.c (simplify_truncation): Generalize simplification
6810 of (truncate:A (subreg:B X)).
6811 (simplify_unary_operation_1) [FLOAT_TRUNCATE, FLOAT_EXTEND,
6812 SIGN_EXTEND, ZERO_EXTEND]: Handle cases where the operand
6813 already has the desired machine mode.
6814 (test_scalar_int_ops): Add tests that useless extensions and
6815 truncations are optimized away.
6816 (test_scalar_int_ext_ops): New self-test function to confirm
6817 that truncations of extensions are correctly simplified.
6818 (test_scalar_int_ext_ops2): New self-test function to check
6819 truncations of truncations, extensions of extensions, and
6820 truncations of extensions.
6821 (test_scalar_ops): Call the above two functions with a
6822 representative sampling of integer machine modes.
6824 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
6826 * match.pd (shift transformations): Change the sign of an
6827 LSHIFT_EXPR if it reduces the number of explicit conversions.
6829 2021-08-23 Jakub Jelinek <jakub@redhat.com>
6831 PR tree-optimization/86723
6832 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
6833 cast64_to_32 argument, set *cast64_to_32 to false, unless n is
6834 non-memory permutation of 64-bit src which only has bytes of
6835 0 or [5..8] and n->range is 4.
6836 (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
6837 find_bswap_or_nop_finalize caller, support bswap with some bytes
6838 zeroed, as long as at least two bytes are not zeroed.
6839 (bswap_replace): Add mask argument and handle masking of bswap
6841 (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
6842 caller, punt if cast64_to_32 or mask is not all ones.
6843 (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
6844 caller, for now punt if cast64_to_32.
6846 2021-08-23 Richard Biener <rguenther@suse.de>
6848 PR tree-optimization/79334
6849 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Record
6850 a type also for COMPONENT_REFs.
6851 (vn_reference_may_trap): Check ARRAY_REF with constant index
6852 against the array domain.
6854 2021-08-23 liuhongt <hongtao.liu@intel.com>
6857 * config/i386/sse.md (*avx512f_pshufb_truncv8hiv8qi_1): Add
6858 TARGET_AVX512BW to condition.
6860 2021-08-23 Jakub Jelinek <jakub@redhat.com>
6863 * dwarf2out.c (gen_variable_die): Add DW_AT_location for global
6864 register variables already during early_dwarf if possible.
6866 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
6868 * config/arm/arm_mve.h: Fix __arm_vctp16q return type.
6870 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
6873 * config/arm/arm.opt: Fix typo.
6874 * config/arm/t-rmprofile: Fix typo.
6876 2021-08-23 Jakub Jelinek <jakub@redhat.com>
6878 * tree.h (OMP_CLAUSE_GRAINSIZE_STRICT): Define.
6879 (OMP_CLAUSE_NUM_TASKS_STRICT): Define.
6880 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_GRAINSIZE,
6881 case OMP_CLAUSE_NUM_TASKS>: Print strict: modifier.
6882 * omp-expand.c (expand_task_call): Use GOMP_TASK_FLAG_STRICT in iflags
6883 if either grainsize or num_tasks clause has the strict modifier.
6885 2021-08-23 Martin Liska <mliska@suse.cz>
6887 * dbgcnt.def (DEBUG_COUNTER): New counter.
6888 * gimple.c (gimple_call_arg_flags): Use it in IPA PTA.
6890 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
6892 * ipa-modref.c (analyze_ssa_name_flags): Improve handling of return slot.
6894 2021-08-23 Xi Ruoyao <xry111@mengyan1223.wang>
6897 * config/mips/mips-protos.h (mips_msa_output_shift_immediate):
6899 * config/mips/mips.c (mips_msa_output_shift_immediate): New
6901 * config/mips/mips-msa.md (vashl<mode>3, vashr<mode>3,
6902 vlshr<mode>3): Call it.
6904 2021-08-22 Jan Hubicka <hubicka@ucw.cz>
6905 Martin Liska <mliska@suse.cz>
6907 PR middle-end/101949
6908 * ipa-modref.c (analyze_ssa_name_flags): Indirect call implies
6911 2021-08-21 Dragan Mladjenovic <OT_Dragan.Mladjenovic@mediatek.com>
6913 * config/mips/mips.c (mips_function_rodata_section,
6914 TARGET_ASM_FUNCTION_RODATA_SECTION): Removed.
6916 2021-08-21 John David Anglin <danglin@gcc.gnu.org>
6918 * config/pa/pa.c (pa_asm_output_aligned_common): Remove warning.
6920 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
6922 * configure.ac (thread-local storage support): Remove tls_first_major
6923 and tls_first_minor. Use "$conftest_s" to check support.
6924 * configure: Regenerate.
6926 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
6928 * configure.ac: Fixup formatting.
6930 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
6932 * acinclude.m4 (gcc_GAS_CHECK_FEATURE): Remove third argument and ...
6933 * configure.ac: ... update all callers.
6935 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
6938 * acinclude.m4 (_gcc_COMPUTE_GAS_VERSION, _gcc_GAS_VERSION_GTE_IFELSE)
6939 (gcc_GAS_VERSION_GTE_IFELSE): Remove.
6940 (gcc_GAS_CHECK_FEATURE): Do not handle in-tree case specially.
6941 * configure.ac: Remove gcc_cv_gas_major_version, gcc_cv_gas_minor_version.
6942 Remove remaining checks for in-tree assembler.
6943 * configure: Regenerate.
6945 2021-08-20 Jeff Law <jlaw@localhost.localdomain>
6947 * config/h8300/h8300.c (shift_alg_hi): Improve arithmetic shift right
6948 by 15 bits for H8/300H and H8/S. Improve logical shifts by 12
6950 (shift_alg_si): Improve arithmetic right shift by 28-30 bits for
6951 H8/300H. Improve arithmetic shift right by 15 bits for H8/S.
6952 Improve logical shifts by 27 bits for H8/S.
6953 (get_shift_alg): Corresponding changes.
6954 (h8300_option_override): Revert to loops for -Os when profitable.
6956 2021-08-20 Richard Biener <rguenther@suse.de>
6958 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
6960 (vect_analyze_data_ref_accesses): Likewise. Assign the BB
6961 index as group_id when dataref_groups were not computed.
6962 * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
6963 we advace to the next BB.
6965 2021-08-20 Jakub Jelinek <jakub@redhat.com>
6967 * omp-builtins.def (BUILT_IN_GOMP_WARNING, BUILT_IN_GOMP_ERROR): New
6970 2021-08-20 Martin Liska <mliska@suse.cz>
6972 PR gcov-profile/89961
6973 * gcov.c (make_gcov_file_name): Rewrite using std::string.
6974 (mangle_name): Simplify, do not used the second argument.
6975 (strip_extention): New function.
6976 (get_md5sum): Likewise.
6977 (get_gcov_intermediate_filename): Handle properly -p and -x
6979 (output_gcov_file): Use string type.
6980 (generate_results): Likewise.
6981 (md5sum_to_hex): Remove.
6983 2021-08-20 Michael Meissner <meissner@linux.ibm.com>
6985 * config/rs6000/altivec.md (UNSPEC_XXEVAL): Move to vsx.md.
6986 (UNSPEC_XXSPLTIW): Move to vsx.md.
6987 (UNSPEC_XXSPLTID): Move to vsx.md.
6988 (UNSPEC_XXSPLTI32DX): Move to vsx.md.
6989 (UNSPEC_XXBLEND): Move to vsx.md.
6990 (UNSPEC_XXPERMX): Move to vsx.md.
6991 (VM3): Move to vsx.md.
6992 (VM3_char): Move to vsx.md.
6993 (xxspltiw_v4si): Move to vsx.md.
6994 (xxspltiw_v4sf): Move to vsx.md.
6995 (xxspltiw_v4sf_inst): Move to vsx.md.
6996 (xxspltidp_v2df): Move to vsx.md.
6997 (xxspltidp_v2df_inst): Move to vsx.md.
6998 (xxsplti32dx_v4si_inst): Move to vsx.md.
6999 (xxsplti32dx_v4sf): Move to vsx.md.
7000 (xxsplti32dx_v4sf_inst): Move to vsx.md.
7001 (xxblend_<mode>): Move to vsx.md.
7002 (xxpermx): Move to vsx.md.
7003 (xxpermx_inst): Move to vsx.md.
7004 * config/rs6000/vsx.md (UNSPEC_XXEVAL): Move from altivec.md.
7005 (UNSPEC_XXSPLTIW): Move from altivec.md.
7006 (UNSPEC_XXSPLTID): Move from altivec.md.
7007 (UNSPEC_XXSPLTI32DX): Move from altivec.md.
7008 (UNSPEC_XXBLEND): Move from altivec.md.
7009 (UNSPEC_XXPERMX): Move from altivec.md.
7010 (VM3): Move from altivec.md.
7011 (VM3_char): Move from altivec.md.
7012 (xxspltiw_v4si): Move from altivec.md.
7013 (xxspltiw_v4sf): Move from altivec.md.
7014 (xxspltiw_v4sf_inst): Move from altivec.md.
7015 (xxspltidp_v2df): Move from altivec.md.
7016 (xxspltidp_v2df_inst): Move from altivec.md.
7017 (xxsplti32dx_v4si_inst): Move from altivec.md.
7018 (xxsplti32dx_v4sf): Move from altivec.md.
7019 (xxsplti32dx_v4sf_inst): Move from altivec.md.
7020 (xxblend_<mode>): Move from altivec.md.
7021 (xxpermx): Move from altivec.md.
7022 (xxpermx_inst): Move from altivec.md.
7024 2021-08-19 Roger Sayle <roger@nextmovesoftware.com>
7026 * tree-vect-generic.c (expand_vector_operations_1): Use either
7027 gimplify_build1 or gimplify_build2 instead of gimple_build_assign
7028 when constructing scalar splat expressions.
7030 2021-08-19 Peter Bergner <bergner@linux.ibm.com>
7033 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Cast
7034 pointer to __vector_pair *.
7036 2021-08-19 Martin Sebor <msebor@redhat.com>
7038 * gimple-range.cc: Add comments.
7039 * gimple-range.h: Same.
7041 2021-08-19 Martin Sebor <msebor@redhat.com>
7043 PR middle-end/101984
7044 * gimple-ssa-warn-access.cc (pass_waccess::execute): Also call
7047 2021-08-19 Jeff Law <jlaw@localhost.localdomain>
7049 * config.gcc (h8300-*-elf*): Do not include dbxelf.h.
7050 (h8300-*-linux*, v850-*-rtems*, v850*-elf*): Likewise.
7051 * config/v850/v850.h (DEFAULT_GDB_EXTENSIONS): Remove.
7053 2021-08-19 Jakub Jelinek <jakub@redhat.com>
7055 PR middle-end/101950
7056 * optabs.c (expand_clrsb_using_clz): New function.
7057 (expand_unop): Use it as another clrsb expansion fallback.
7059 2021-08-19 liuhongt <hongtao.liu@intel.com>
7062 2021-07-28 liuhongt <hongtao.liu@intel.com>
7065 * config/i386/i386.h (processor_costs): Add new member
7067 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
7068 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
7069 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
7070 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
7071 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
7072 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
7073 generic_cost, core_cost): Initialize integer_to_sse same value
7075 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
7076 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7077 Use integer_to_sse instead of sse_op to calculate the cost of
7080 2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
7082 * config.gcc: Include rpath.opt for Darwin.
7083 * config/darwin.h (DRIVER_SELF_SPECS): Handle -rpath.
7085 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
7088 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
7091 2021-08-18 Jonathan Wright <jonathan.wright@arm.com>
7093 * config/aarch64/arm_neon.h (vld3_lane_f64): Use float RTL
7094 pattern and type cast.
7095 (vld4_lane_f32): Use float RTL pattern.
7096 (vld4q_lane_f64): Use float type cast.
7098 2021-08-18 Jan Hubicka <hubicka@ucw.cz>
7100 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Check also
7103 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
7105 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor): Extend.
7106 (test_map_of_type_with_ctor_and_dtor_expand): Add function.
7107 (hash_map_tests_c_tests): Call it.
7109 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
7111 * ggc.h (enum ggc_collect): New.
7112 (ggc_collect): Use it.
7113 * ggc-page.c: Adjust.
7114 * ggc-common.c: Likewise.
7115 * ggc-tests.c: Likewise.
7116 * read-rtl-function.c: Likewise.
7117 * selftest-run-tests.c: Likewise.
7118 * doc/gty.texi (Invoking the garbage collector): Likewise.
7120 2021-08-18 liuhongt <hongtao.liu@intel.com>
7123 * config/i386/i386.h (TARGET_V2DF_REDUCTION_PREFER_HADDPD):
7125 * config/i386/sse.md (*sse3_haddv2df3_low): Add
7126 TARGET_V2DF_REDUCTION_PREFER_HADDPD.
7127 (*sse3_hsubv2df3_low): Ditto.
7128 * config/i386/x86-tune.def
7129 (X86_TUNE_V2DF_REDUCTION_PREFER_HADDPD): New tune.
7131 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
7133 * gimple-range-gori.cc (gori_compute::gori_compute): Enable tracing.
7134 (gori_compute::compute_operand_range): Add tracing.
7135 (gori_compute::logical_combine): Ditto.
7136 (gori_compute::compute_logical_operands): Ditto.
7137 (gori_compute::compute_operand1_range): Ditto.
7138 (gori_compute::compute_operand2_range): Ditto.
7139 (gori_compute::outgoing_edge_range_p): Ditto.
7140 * gimple-range-gori.h (class gori_compute): Add range_tracer.
7142 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
7144 * flag-types.h (enum evrp_mode): Adjust evrp-mode values.
7145 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Relocate from.
7146 * gimple-range-trace.h (DEBUG_RANGE_CACHE): Here.
7147 * params.opt (--param=evrp-mode): Adjust options.
7149 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
7151 * Makefile.in (OBJS): Add gimple-range-trace.o.
7152 * gimple-range-cache.h (enable_new_values): Remove unused prototype.
7153 * gimple-range-fold.cc: Adjust headers.
7154 * gimple-range-trace.cc: New.
7155 * gimple-range-trace.h: New.
7156 * gimple-range.cc (gimple_ranger::gimple_ranger): Enable tracer.
7157 (gimple_ranger::range_of_expr): Add tracing.
7158 (gimple_ranger::range_on_entry): Ditto.
7159 (gimple_ranger::range_on_exit): Ditto.
7160 (gimple_ranger::range_on_edge): Ditto.
7161 (gimple_ranger::fold_range_internal): Ditto.
7162 (gimple_ranger::dump_bb): Do not calculate edge range twice.
7163 (trace_ranger::*): Remove.
7164 (enable_ranger): Never create a trace_ranger.
7165 (debug_seed_ranger): Move to gimple-range-trace.cc.
7166 (dump_ranger): Ditto.
7167 (debug_ranger): Ditto.
7168 * gimple-range.h: Include gimple-range-trace.h.
7169 (range_on_entry, range_on_exit): No longer virtual.
7170 (class trace_ranger): Remove.
7171 (DEBUG_RANGE_CACHE): Move to gimple-range-trace.h.
7173 2021-08-17 Martin Sebor <msebor@redhat.com>
7175 PR middle-end/101854
7176 * builtins.c (expand_builtin_alloca): Move warning code to check_alloca
7177 in gimple-ssa-warn-access.cc.
7178 * calls.c (alloc_max_size): Move code to check_alloca.
7179 (get_size_range): Move to pointer-query.cc.
7180 (maybe_warn_alloc_args_overflow): Move to gimple-ssa-warn-access.cc.
7181 (get_attr_nonstring_decl): Move to tree.c.
7182 (fntype_argno_type): Move to gimple-ssa-warn-access.cc.
7183 (append_attrname): Same.
7184 (maybe_warn_rdwr_sizes): Same.
7185 (initialize_argument_information): Move code to
7186 gimple-ssa-warn-access.cc.
7187 * calls.h (maybe_warn_alloc_args_overflow): Move to
7188 gimple-ssa-warn-access.h.
7189 (get_attr_nonstring_decl): Move to tree.h.
7190 (maybe_warn_nonstring_arg): Move to gimple-ssa-warn-access.h.
7191 (enum size_range_flags): Move to pointer-query.h.
7192 (get_size_range): Same.
7193 * gimple-ssa-warn-access.cc (has_location): Remove unused overload
7194 to avoid Clang -Wunused-function.
7195 (get_size_range): Declare static.
7196 (maybe_emit_free_warning): Rename...
7197 (maybe_check_dealloc_call): ...to this for consistency.
7198 (class pass_waccess): Add members.
7199 (pass_waccess::~pass_waccess): Defined.
7200 (alloc_max_size): Move here from calls.c.
7201 (maybe_warn_alloc_args_overflow): Same.
7202 (check_alloca): New function.
7203 (check_alloc_size_call): New function.
7204 (check_strncat): Handle another warning flag.
7205 (pass_waccess::check_builtin): Handle alloca.
7206 (fntype_argno_type): Move here from calls.c.
7207 (append_attrname): Same.
7208 (maybe_warn_rdwr_sizes): Same.
7209 (pass_waccess::check_call): Define.
7210 (check_nonstring_args): New function.
7211 (pass_waccess::check): Call new member functions.
7212 (pass_waccess::execute): Enable ranger.
7213 * gimple-ssa-warn-access.h (get_size_range): Move here from calls.h.
7214 (maybe_warn_nonstring_arg): Same.
7215 * gimple-ssa-warn-restrict.c: Remove #include.
7216 * pointer-query.cc (get_size_range): Move here from calls.c.
7217 * pointer-query.h (enum size_range_flags): Same.
7218 (get_size_range): Same.
7219 * tree.c (get_attr_nonstring_decl): Move here from calls.c.
7220 * tree.h (get_attr_nonstring_decl): Move here from calls.h.
7222 2021-08-17 Thomas Schwinge <thomas@codesourcery.com>
7224 * ggc.h (ggc_collect): Add 'force_collect' parameter.
7225 * ggc-page.c (ggc_collect): Use that one instead of global
7226 'ggc_force_collect'. Adjust all users.
7227 * doc/gty.texi (Invoking the garbage collector): Update.
7228 * ggc-internal.h (ggc_force_collect): Remove.
7229 * ggc-common.c (ggc_force_collect): Likewise.
7230 * selftest.h (forcibly_ggc_collect): Remove.
7231 * ggc-tests.c (selftest::forcibly_ggc_collect): Likewise.
7232 * read-rtl-function.c (test_loading_labels): Adjust.
7233 * selftest-run-tests.c (run_tests): Likewise.
7235 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
7237 * config/darwin.c (darwin_file_end): Reset and reclaim the
7238 section names table at the end of compile.
7240 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
7243 * config.in: Regenerate.
7244 * config/i386/darwin.h (EXTRA_ASM_OPTS): New
7245 (ASM_SPEC): Pass options to disable branch shortening where
7247 * configure: Regenerate.
7248 * configure.ac: Detect versions of 'as' that support the
7249 optimisation which has the bug.
7251 2021-08-17 Richard Biener <rguenther@suse.de>
7253 * optabs-query.c (supports_vec_gather_load_p): Also check
7255 (supports_vec_scatter_store_p): Likewise.
7256 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Fall
7257 back to masked variants if non-masked are not supported.
7258 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
7259 When we need to use masked gather/scatter but do not have
7260 a mask set up a constant true one.
7261 * tree-vect-stmts.c (vect_check_scalar_mask): Also allow
7264 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
7266 * tree-ssa-ccp.c (bit_value_binop) [MINUS_EXPR]: Use same
7267 algorithm as PLUS_EXPR to improve subtraction bit bounds.
7268 [POINTER_DIFF_EXPR]: Treat as synonymous with MINUS_EXPR.
7270 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
7272 * tree-ssa-ccp.c (bit_value_mult_const): New helper function to
7273 calculate the mask-value pair result of a multiplication by an
7275 (bit_value_binop) [MULT_EXPR]: Call it from here for
7276 multiplications by (sparse) non-negative constants.
7278 2021-08-17 Christophe Lyon <christophe.lyon@foss.st.com>
7281 * config.gcc (gcc_cv_initfini_array): Leave undefined for
7282 uclinuxfdpiceabi targets.
7284 2021-08-17 Alexandre Oliva <oliva@adacore.com>
7286 * tree-inline.c (maybe_move_debug_stmts_to_successors): Don't
7287 reverse debug stmts.
7289 2021-08-17 Alexandre Oliva <oliva@adacore.com>
7291 * tree-cfg.c (dump_function_to_file): Use fun, not cfun.
7293 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
7295 * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete.
7296 (__LD4Q_LANE_FUNC): Likewise.
7297 (vld4_lane_u8): Define without macro.
7298 (vld4_lane_u16): Likewise.
7299 (vld4_lane_u32): Likewise.
7300 (vld4_lane_u64): Likewise.
7301 (vld4_lane_s8): Likewise.
7302 (vld4_lane_s16): Likewise.
7303 (vld4_lane_s32): Likewise.
7304 (vld4_lane_s64): Likewise.
7305 (vld4_lane_f16): Likewise.
7306 (vld4_lane_f32): Likewise.
7307 (vld4_lane_f64): Likewise.
7308 (vld4_lane_p8): Likewise.
7309 (vld4_lane_p16): Likewise.
7310 (vld4_lane_p64): Likewise.
7311 (vld4q_lane_u8): Likewise.
7312 (vld4q_lane_u16): Likewise.
7313 (vld4q_lane_u32): Likewise.
7314 (vld4q_lane_u64): Likewise.
7315 (vld4q_lane_s8): Likewise.
7316 (vld4q_lane_s16): Likewise.
7317 (vld4q_lane_s32): Likewise.
7318 (vld4q_lane_s64): Likewise.
7319 (vld4q_lane_f16): Likewise.
7320 (vld4q_lane_f32): Likewise.
7321 (vld4q_lane_f64): Likewise.
7322 (vld4q_lane_p8): Likewise.
7323 (vld4q_lane_p16): Likewise.
7324 (vld4q_lane_p64): Likewise.
7325 (vld4_lane_bf16): Likewise.
7326 (vld4q_lane_bf16): Likewise.
7328 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
7330 * config/aarch64/arm_neon.h (__LD3_LANE_FUNC): Delete.
7331 (__LD3Q_LANE_FUNC): Delete.
7332 (vld3_lane_u8): Define without macro.
7333 (vld3_lane_u16): Likewise.
7334 (vld3_lane_u32): Likewise.
7335 (vld3_lane_u64): Likewise.
7336 (vld3_lane_s8): Likewise.
7337 (vld3_lane_s16): Likewise.
7338 (vld3_lane_s32): Likewise.
7339 (vld3_lane_s64): Likewise.
7340 (vld3_lane_f16): Likewise.
7341 (vld3_lane_f32): Likewise.
7342 (vld3_lane_f64): Likewise.
7343 (vld3_lane_p8): Likewise.
7344 (vld3_lane_p16): Likewise.
7345 (vld3_lane_p64): Likewise.
7346 (vld3q_lane_u8): Likewise.
7347 (vld3q_lane_u16): Likewise.
7348 (vld3q_lane_u32): Likewise.
7349 (vld3q_lane_u64): Likewise.
7350 (vld3q_lane_s8): Likewise.
7351 (vld3q_lane_s16): Likewise.
7352 (vld3q_lane_s32): Likewise.
7353 (vld3q_lane_s64): Likewise.
7354 (vld3q_lane_f16): Likewise.
7355 (vld3q_lane_f32): Likewise.
7356 (vld3q_lane_f64): Likewise.
7357 (vld3q_lane_p8): Likewise.
7358 (vld3q_lane_p16): Likewise.
7359 (vld3q_lane_p64): Likewise.
7360 (vld3_lane_bf16): Likewise.
7361 (vld3q_lane_bf16): Likewise.
7363 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
7365 * config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Delete.
7366 (__LD2Q_LANE_FUNC): Likewise.
7367 (vld2_lane_u8): Define without macro.
7368 (vld2_lane_u16): Likewise.
7369 (vld2_lane_u32): Likewise.
7370 (vld2_lane_u64): Likewise.
7371 (vld2_lane_s8): Likewise.
7372 (vld2_lane_s16): Likewise.
7373 (vld2_lane_s32): Likewise.
7374 (vld2_lane_s64): Likewise.
7375 (vld2_lane_f16): Likewise.
7376 (vld2_lane_f32): Likewise.
7377 (vld2_lane_f64): Likewise.
7378 (vld2_lane_p8): Likewise.
7379 (vld2_lane_p16): Likewise.
7380 (vld2_lane_p64): Likewise.
7381 (vld2q_lane_u8): Likewise.
7382 (vld2q_lane_u16): Likewise.
7383 (vld2q_lane_u32): Likewise.
7384 (vld2q_lane_u64): Likewise.
7385 (vld2q_lane_s8): Likewise.
7386 (vld2q_lane_s16): Likewise.
7387 (vld2q_lane_s32): Likewise.
7388 (vld2q_lane_s64): Likewise.
7389 (vld2q_lane_f16): Likewise.
7390 (vld2q_lane_f32): Likewise.
7391 (vld2q_lane_f64): Likewise.
7392 (vld2q_lane_p8): Likewise.
7393 (vld2q_lane_p16): Likewise.
7394 (vld2q_lane_p64): Likewise.
7395 (vld2_lane_bf16): Likewise.
7396 (vld2q_lane_bf16): Likewise.
7398 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
7400 * haifa-sched.c (advance_one_cycle): Output more context-synchronization
7403 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
7405 * haifa-sched.c (enum rfs_decision, rfs_str): Add RFS_AUTOPREF.
7406 (rank_for_schedule): Use it.
7408 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
7410 PR rtl-optimization/91598
7411 * haifa-sched.c (autopref_rank_for_schedule): Prioritize "irrelevant"
7412 insns after memory reads and before memory writes.
7414 2021-08-17 Alistair_Lee <alistair.lee@arm.com>
7416 * rtl.h (CONST_VECTOR_P): New macro.
7417 * config/aarch64/aarch64.c (aarch64_get_sve_pred_bits): Use RTL
7418 code testing macros.
7419 (aarch64_ptrue_all_mode): Likewise.
7420 (aarch64_expand_mov_immediate): Likewise.
7421 (aarch64_const_vec_all_in_range_p): Likewise.
7422 (aarch64_rtx_costs): Likewise.
7423 (aarch64_legitimate_constant_p): Likewise.
7424 (aarch64_simd_valid_immediate): Likewise.
7425 (aarch64_simd_make_constant): Likewise.
7426 (aarch64_convert_mult_to_shift): Likewise.
7427 (aarch64_expand_sve_vec_perm): Likewise.
7428 (aarch64_vec_fpconst_pow_of_2): Likewise.
7430 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
7432 PR tree-optimization/101938
7433 * range-op.cc (operator_abs::op1_range): Special case
7434 -TYPE_MIN_VALUE for flag_wrapv.
7436 2021-08-17 Kewen Lin <linkw@linux.ibm.com>
7438 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Add the cost for
7441 2021-08-17 Jakub Jelinek <jakub@redhat.com>
7443 * tree.def (OMP_SCOPE): New tree code.
7444 * tree.h (OMP_SCOPE_BODY, OMP_SCOPE_CLAUSES): Define.
7445 * tree-nested.c (convert_nonlocal_reference_stmt,
7446 convert_local_reference_stmt, convert_gimple_call): Handle
7448 * tree-pretty-print.c (dump_generic_node): Handle OMP_SCOPE.
7449 * gimple.def (GIMPLE_OMP_SCOPE): New gimple code.
7450 * gimple.c (gimple_build_omp_scope): New function.
7451 (gimple_copy): Handle GIMPLE_OMP_SCOPE.
7452 * gimple.h (gimple_build_omp_scope): Declare.
7453 (gimple_has_substatements): Handle GIMPLE_OMP_SCOPE.
7454 (gimple_omp_scope_clauses, gimple_omp_scope_clauses_ptr,
7455 gimple_omp_scope_set_clauses): New inline functions.
7456 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_SCOPE.
7457 * gimple-pretty-print.c (dump_gimple_omp_scope): New function.
7458 (pp_gimple_stmt_1): Handle GIMPLE_OMP_SCOPE.
7459 * gimple-walk.c (walk_gimple_stmt): Likewise.
7460 * gimple-low.c (lower_stmt): Likewise.
7461 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
7462 (gimplify_scan_omp_clauses): For task reductions, handle OMP_SCOPE
7463 like ORT_WORKSHARE constructs. Adjust diagnostics for %<scope%>
7464 allowing task reductions. Reject inscan reductions on scope.
7465 (omp_find_stores_stmt): Handle GIMPLE_OMP_SCOPE.
7466 (gimplify_omp_workshare, gimplify_expr): Handle OMP_SCOPE.
7467 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_SCOPE.
7468 (estimate_num_insns): Likewise.
7469 * omp-low.c (build_outer_var_ref): Look through GIMPLE_OMP_SCOPE
7470 contexts if var isn't privatized there.
7471 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_SCOPE.
7472 (scan_omp_1_stmt): Likewise.
7473 (maybe_add_implicit_barrier_cancel): Look through outer
7475 (lower_omp_scope): New function.
7476 (lower_omp_task_reductions): Handle OMP_SCOPE.
7477 (lower_omp_1): Handle GIMPLE_OMP_SCOPE.
7478 (diagnose_sb_1, diagnose_sb_2): Likewise.
7479 * omp-expand.c (expand_omp_single): Support also GIMPLE_OMP_SCOPE.
7480 (expand_omp): Handle GIMPLE_OMP_SCOPE.
7481 (omp_make_gimple_edges): Likewise.
7482 * omp-builtins.def (BUILT_IN_GOMP_SCOPE_START): New built-in.
7484 2021-08-17 Richard Biener <rguenther@suse.de>
7486 PR tree-optimization/101925
7487 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Set
7488 reverse on COMPONENT_REF and ARRAY_REF according to
7489 what reverse_storage_order_for_component_p does.
7490 (vn_reference_eq): Compare reversed on reference ops.
7491 (reverse_storage_order_for_component_p): New overload.
7492 (vn_reference_lookup_3): Check reverse_storage_order_for_component_p
7493 on the reference looked up.
7495 2021-08-17 Jeff Law <jlaw@localhost.localdomain>
7497 * config/h8300/h8300.c (shift_alg_si): Avoid loops for most SImode
7499 (h8300_option_override): Use loops on H8/S more often when optimizing
7501 (get_shift_alg): Handle new "special" cases on H8/S. Simplify
7502 accordingly. Handle various arithmetic right shifts with special
7503 sequences that we couldn't handle before.
7505 2021-08-16 Jeff Law <jlaw@localhost.localdomain>
7507 * config.gcc (rl78-*-elf*): Do not include dbxelf.h.
7509 2021-08-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
7511 * config/sparc/rtemself.h (SPARC_GCOV_TYPE_SIZE): Define.
7512 * config/sparc/sparc.c (sparc_gcov_type_size): New.
7513 (TARGET_GCOV_TYPE_SIZE): Redefine if SPARC_GCOV_TYPE_SIZE is defined.
7514 * coverage.c (get_gcov_type): Use targetm.gcov_type_size().
7515 * doc/tm.texi (TARGET_GCOV_TYPE_SIZE): Add hook under "Misc".
7516 * doc/tm.texi.in: Regenerate.
7517 * target.def (gcov_type_size): New target hook.
7518 * targhooks.c (default_gcov_type_size): New.
7519 * targhooks.h (default_gcov_type_size): Declare.
7520 * tree-profile.c (gimple_gen_edge_profiler): Use precision of
7522 (gimple_gen_time_profiler): Likewise.
7524 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
7526 * dwarf2out.c (add_scalar_info): Deal with DW_AT_data_bit_offset.
7528 2021-08-16 Tobias Burnus <tobias@codesourcery.com>
7530 PR middle-end/101931
7531 * omp-low.c (omp_runtime_api_call): Update for routines
7532 added in the meanwhile.
7534 2021-08-16 Martin Liska <mliska@suse.cz>
7536 PR tree-optimization/100393
7537 * tree-switch-conversion.c (group_cluster::dump): Use
7538 get_comparison_count.
7539 (jump_table_cluster::find_jump_tables): Pre-compute number of
7540 comparisons and then decrement it. Cache also max_ratio.
7541 (jump_table_cluster::can_be_handled): Change signature.
7542 * tree-switch-conversion.h (get_comparison_count): New.
7544 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
7546 * dwarf2out.c (add_data_member_location_attribute): Use GNAT
7547 encodings only when -fgnat-encodings=all is specified.
7548 (add_bound_info): Likewise.
7549 (add_byte_size_attribute): Likewise.
7550 (gen_member_die): Likewise.
7552 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
7554 * omp-oacc-neuter-broadcast.cc
7555 (execute_omp_oacc_neuter_broadcast): Plug 'par' memory leak.
7557 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
7559 * omp-oacc-neuter-broadcast.cc
7560 (execute_omp_oacc_neuter_broadcast): Clarify memory management for
7563 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
7565 * omp-oacc-neuter-broadcast.cc (field_map): Move variable into...
7566 (execute_omp_oacc_neuter_broadcast): ... here.
7567 (install_var_field, build_receiver_ref, build_sender_ref): Take
7568 'field_map_t *' parameter. Adjust all users.
7569 (worker_single_copy, neuter_worker_single): Take a
7570 'record_field_map_t *' parameter. Adjust all users.
7572 2021-08-16 liuhongt <hongtao.liu@intel.com>
7575 * config/i386/i386.md (ldexp<mode>3): Force operands[1] to
7578 2021-08-16 Martin Liska <mliska@suse.cz>
7581 * multiple_target.c (create_dispatcher_calls): Make default
7582 function local only if it is a definition.
7584 2021-08-16 Martin Liska <mliska@suse.cz>
7587 * ipa-icf-gimple.c (func_checker::compare_ssa_name): Do not
7588 consider equal SSA_NAMEs when one is a param.
7590 2021-08-16 liuhongt <hongtao.liu@intel.com>
7593 * config/i386/i386-expand.c (ix86_expand_vec_perm_vpermt2):
7594 Support vpermi2b for V32QI/V16QImode.
7595 (ix86_extract_perm_from_pool_constant): New function.
7596 (ix86_expand_vec_one_operand_perm_avx512): Support
7597 vpermw/vpermb under TARGET_AVX512BW/TARGET_AVX512VBMI.
7598 (expand_vec_perm_1): Adjust comments for upper.
7599 * config/i386/i386-protos.h (ix86_extract_perm_from_pool_constant):
7601 * config/i386/predicates.md (permvar_truncate_operand): New predicate.
7602 (pshufb_truncv4siv4hi_operand): Ditto.
7603 (pshufb_truncv8hiv8qi_operand): Ditto.
7604 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1):
7605 New pre_reload define_insn_and_split.
7606 (*avx512f_permvar_truncv8siv8hi_1): Ditto.
7607 (*avx512f_vpermvar_truncv8div8si_1): Ditto.
7608 (*avx512f_permvar_truncv32hiv32qi_1): Ditto.
7609 (*avx512f_permvar_truncv16hiv16qi_1): Ditto.
7610 (*avx512f_permvar_truncv4div4si_1): Ditto.
7611 (*avx512f_pshufb_truncv8hiv8qi_1): Ditto.
7612 (*avx512f_pshufb_truncv4siv4hi_1): Ditto.
7613 (*avx512f_pshufd_truncv2div2si_1): Ditto.
7615 2021-08-16 Kito Cheng <kito.cheng@sifive.com>
7617 * config/riscv/multilib-generator: Support code model option for
7619 * doc/install.texi: Add document of new option for
7620 --with-multilib-generator.
7622 2021-08-15 Clément Chigot <clement.chigot@atos.net>
7624 * config/rs6000/rs6000.c (xcoff_tls_exec_model_detected): New.
7625 (rs6000_legitimize_tls_address_aix): Use it.
7626 (rs6000_xcoff_file_end): Add ".ref __tls_get_addr" when
7627 xcoff_tls_exec_model_detected is true.
7629 2021-08-15 Jeff Law <jlaw@localhost.localdomain>
7631 * config/h8300/h8300.c (shift_alg_si): Retune H8/300H shifts
7632 to allow a bit more code growth, saving many dozens of cycles.
7633 (h8300_option_override): Adjus shift_alg_si if optimizing for
7635 (get_shift_alg): Use special + inline shifts for residuals
7638 2021-08-14 Stafford Horne <shorne@gmail.com>
7641 * config/or1k/or1k-opts.h: New file.
7642 * config/or1k/or1k.c (or1k_legitimize_address_1, print_reloc):
7643 Support generating gotha relocations if -mcmodel=large is
7645 * config/or1k/or1k.h (TARGET_CMODEL_SMALL, TARGET_CMODEL_LARGE):
7647 * config/or1k/or1k.opt (mcmodel=): New option.
7648 * doc/invoke.texi (OpenRISC Options): Document mcmodel.
7650 2021-08-14 Martin Sebor <msebor@redhat.com>
7652 PR middle-end/101791
7653 * gimple-ssa-warn-access.cc (new_delete_mismatch_p): Use new argument
7654 to valid_new_delete_pair_p.
7655 * tree.c (valid_new_delete_pair_p): Add argument.
7656 * tree.h (valid_new_delete_pair_p): Same.
7658 2021-08-14 Jakub Jelinek <jakub@redhat.com>
7661 * config/i386/i386-expand.c (expand_vec_perm_broadcast_1)
7662 <case E_V64QImode>: For this mode assert
7663 !TARGET_AVX512BW || d->perm[0] rather than !TARGET_AVX2 || d->perm[0].
7665 2021-08-13 Michael Meissner <meissner@linux.ibm.com>
7668 * config/rs6000/altivec.md (xxeval): Use register_predicate
7669 instead of altivec_register_predicate.
7671 2021-08-13 Martin Sebor <msebor@redhat.com>
7673 PR middle-end/101734
7674 * tree-ssa-uninit.c (maybe_warn_read_write_only): New function.
7675 (maybe_warn_operand): Call it.
7677 2021-08-13 Martin Liska <mliska@suse.cz>
7680 * attribs.c (decl_attributes): Make naked functions "noipa"
7683 2021-08-13 Martin Liska <mliska@suse.cz>
7686 * symtab.c (symtab_node::noninterposable_alias): Do not create
7687 local aliases for target_clone functions as the clonning pass
7690 2021-08-13 Martin Liska <mliska@suse.cz>
7692 * opts.c (LIVE_PATCHING_OPTION): Define.
7693 (control_options_for_live_patching): Use it in error messages.
7695 2021-08-13 Jan Hubicka <hubicka@ucw.cz>
7697 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOREAD.
7698 (implicit_const_eaf_flags, implicit_pure_eaf_flags,
7699 ignore_stores_eaf_flags): New constants.
7700 (remove_useless_eaf_flags): New function.
7701 (eaf_flags_useful_p): Use it.
7702 (deref_flags): Add EAF_NOT_RETURNED if flag is unused;
7704 (modref_lattice::init): Add EAF_NOREAD.
7705 (modref_lattice::add_escape_point): Do not reacord escape point if
7707 (modref_lattice::merge): EAF_NOESCAPE implies EAF_NODIRECTESCAPE;
7708 use remove_useless_eaf_flags.
7709 (modref_lattice::merge_deref): Use ignore_stores_eaf_flags.
7710 (modref_lattice::merge_direct_load): Add EAF_NOREAD
7711 (analyze_ssa_name_flags): Fix handling EAF_NOT_RETURNED
7712 (analyze_parms): Use remove_useless_eaf_flags.
7713 (ipa_merge_modref_summary_after_inlining): Use ignore_stores_eaf_flags.
7714 (modref_merge_call_site_flags): Add caller and ecf_flags parameter;
7715 use remove_useless_eaf_flags.
7716 (modref_propagate_flags_in_scc): Update.
7717 * ipa-modref.h: Turn eaf_flags_t back to char.
7718 * tree-core.h (EAF_NOT_RETURNED): Fix.
7719 (EAF_NOREAD): New constant
7720 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1): Check for
7722 * tree-ssa-structalias.c (handle_rhs_call): Handle new flags.
7723 (handle_pure_call): Likewise.
7725 2021-08-12 Jakub Jelinek <jakub@redhat.com>
7727 * tree.def (OMP_MASKED): New tree code.
7728 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_FILTER.
7729 * tree.h (OMP_MASKED_BODY, OMP_MASKED_CLAUSES, OMP_MASKED_COMBINED,
7730 OMP_CLAUSE_FILTER_EXPR): Define.
7731 * tree.c (omp_clause_num_ops): Add OMP_CLAUSE_FILTER entry.
7732 (omp_clause_code_name): Likewise.
7733 (walk_tree_1): Handle OMP_CLAUSE_FILTER.
7734 * tree-nested.c (convert_nonlocal_omp_clauses,
7735 convert_local_omp_clauses): Handle OMP_CLAUSE_FILTER.
7736 (convert_nonlocal_reference_stmt, convert_local_reference_stmt,
7737 convert_gimple_call): Handle GIMPLE_OMP_MASTER.
7738 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_FILTER.
7739 (dump_generic_node): Handle OMP_MASTER.
7740 * gimple.def (GIMPLE_OMP_MASKED): New gimple code.
7741 * gimple.c (gimple_build_omp_masked): New function.
7742 (gimple_copy): Handle GIMPLE_OMP_MASKED.
7743 * gimple.h (gimple_build_omp_masked): Declare.
7744 (gimple_has_substatements): Handle GIMPLE_OMP_MASKED.
7745 (gimple_omp_masked_clauses, gimple_omp_masked_clauses_ptr,
7746 gimple_omp_masked_set_clauses): New inline functions.
7747 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_MASKED.
7748 * gimple-pretty-print.c (dump_gimple_omp_masked): New function.
7749 (pp_gimple_stmt_1): Handle GIMPLE_OMP_MASKED.
7750 * gimple-walk.c (walk_gimple_stmt): Likewise.
7751 * gimple-low.c (lower_stmt): Likewise.
7752 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
7753 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_FILTER. For clauses
7754 that take one expression rather than decl or constant, force
7755 gimplification of that into a SSA_NAME or temporary unless min
7757 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_FILTER.
7758 (gimplify_expr): Handle OMP_MASKED.
7759 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_MASKED.
7760 (estimate_num_insns): Likewise.
7761 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_FILTER.
7762 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_MASKED. Adjust
7763 diagnostics for existence of masked construct.
7764 (scan_omp_1_stmt, lower_omp_master, lower_omp_1, diagnose_sb_1,
7765 diagnose_sb_2): Handle GIMPLE_OMP_MASKED.
7766 * omp-expand.c (expand_omp_synch, expand_omp, omp_make_gimple_edges):
7769 2021-08-12 Uroš Bizjak <ubizjak@gmail.com>
7772 * config/i386/i386.md (avx512f_scalef<mode>2): New insn pattern.
7773 (ldexp<mode>3): Use avx512f_scalef<mode>2.
7774 (UNSPEC_SCALEF): Move from ...
7775 * config/i386/sse.md (UNSPEC_SCALEF): ... here.
7777 2021-08-12 Jan Hubicka <hubicka@ucw.cz>
7779 * ipa-split.c (consider_split): Fix condition testing void functions.
7781 2021-08-12 Aldy Hernandez <aldyh@redhat.com>
7783 * doc/invoke.texi: Remove docs for threader-mode param.
7784 * flag-types.h (enum threader_mode): Remove.
7785 * params.opt: Remove threader-mode param.
7786 * tree-ssa-threadbackward.c (class back_threader): Remove
7787 path_is_unreachable_p.
7788 Make find_paths private.
7789 Add maybe_thread and thread_through_all_blocks.
7790 Remove reference marker for m_registry.
7791 Remove reference marker for m_profit.
7792 (back_threader::back_threader): Adjust for registry and profit not
7794 (dump_path): Move down.
7796 (class thread_jumps): Remove.
7797 (class back_threader_registry): Remove m_all_paths.
7799 (thread_jumps::thread_through_all_blocks): Move to back_threader
7801 (fsm_find_thread_path): Remove
7802 (back_threader::maybe_thread): New.
7803 (back_threader::thread_through_all_blocks): Move from
7805 (back_threader_registry::back_threader_registry): Remove
7807 (back_threader_registry::~back_threader_registry): Remove.
7808 (thread_jumps::find_taken_edge): Remove.
7809 (thread_jumps::check_subpath_and_update_thread_path): Remove.
7810 (thread_jumps::maybe_register_path): Remove.
7811 (thread_jumps::handle_phi): Remove.
7812 (handle_assignment_p): Remove.
7813 (thread_jumps::handle_assignment): Remove.
7814 (thread_jumps::fsm_find_control_statement_thread_paths): Remove.
7815 (thread_jumps::find_jump_threads_backwards): Remove.
7816 (thread_jumps::find_jump_threads_backwards_with_ranger): Remove.
7817 (try_thread_blocks): Rename find_jump_threads_backwards to
7819 (pass_early_thread_jumps::execute): Same.
7821 2021-08-12 Tobias Burnus <tobias@codesourcery.com>
7823 * tree-core.h (omp_clause_proc_bind_kind): Add
7824 OMP_CLAUSE_PROC_BIND_PRIMARY.
7825 * tree-pretty-print.c (dump_omp_clause): Add TODO comment to
7826 change 'master' to 'primary' in proc_bind for OpenMP 5.1.
7828 2021-08-12 Claudiu Zissulescu <claziss@synopsys.com>
7830 * common/config/arc/arc-common.c (arc_option_init_struct): Remove
7831 fno-common reference.
7832 * config/arc/arc.c (arc_override_options): Remove overriding of
7835 2021-08-12 Jakub Jelinek <jakub@redhat.com>
7838 * config/i386/i386-expand.c (ix86_expand_vec_one_operand_perm_avx512):
7839 If d->testing_p, return true after performing checks instead of
7840 actually expanding the insn.
7841 (expand_vec_perm_broadcast_1): Handle V32HImode - assert
7842 !TARGET_AVX512BW and return false.
7844 2021-08-12 Eric Botcazou <ebotcazou@gcc.gnu.org>
7846 * configure.ac (PE linker --disable-dynamicbase support): New check.
7847 * configure: Regenerate.
7848 * config.in: Likewise.
7849 * config/i386/mingw32.h (LINK_SPEC_DISABLE_DYNAMICBASE): New define.
7850 (LINK_SPEC): Use it.
7851 * config/i386/mingw-w64.h (LINK_SPEC_DISABLE_DYNAMICBASE): Likewise.
7852 (LINK_SPEC): Likewise.
7854 2021-08-12 liuhongt <hongtao.liu@intel.com>
7857 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_2): New
7858 post_reload define_insn_and_split.
7859 (*avx512bw_zero_extendv32qiv32hi2_2): Ditto.
7860 (*sse4_1_zero_extendv8qiv8hi2_4): Ditto.
7861 (*avx512f_zero_extendv16hiv16si2_2): Ditto.
7862 (*avx2_zero_extendv8hiv8si2_2): Ditto.
7863 (*sse4_1_zero_extendv4hiv4si2_4): Ditto.
7864 (*avx512f_zero_extendv8siv8di2_2): Ditto.
7865 (*avx2_zero_extendv4siv4di2_2): Ditto.
7866 (*sse4_1_zero_extendv2siv2di2_4): Ditto.
7867 (VI248_256, VI248_512, VI148_512, VI148_256, VI148_128): New
7870 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
7872 * config/rs6000/rs6000-builtin-new.def: Add always, power5, and
7875 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
7877 * config/rs6000/rs6000-builtin-new.def: Add vsx stanza.
7879 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
7881 * config/rs6000/rs6000-builtin-new.def: Finish altivec stanza.
7882 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
7883 initialization of pcvoid_type_node here...
7884 (altivec_init_builtins): ...from here.
7885 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add
7886 RS6000_BTI_const_ptr_void.
7887 (pcvoid_type_node): New macro.
7889 2021-08-11 Richard Biener <rguenther@suse.de>
7892 * tree-ssa-forwprop.c (pass_forwprop::execute): Do not decompose
7893 hard-register accesses.
7895 2021-08-11 Richard Biener <rguenther@suse.de>
7897 * tree-ssa-operands.c (operands_scanner::get_expr_operands):
7898 Do not look at COMPONENT_REF FIELD_DECLs TREE_THIS_VOLATILE
7899 to determine has_volatile_ops.
7901 2021-08-11 Eric Botcazou <ebotcazou@gcc.gnu.org>
7903 * cfgexpand.c (expand_used_vars): Reuse attribs local variable.
7905 2021-08-11 Jan Hubicka <hubicka@ucw.cz>
7906 Alexandre Oliva <oliva@adacore.com>
7908 * ipa-modref.c (modref_lattice::dump): Fix escape_point's min_flags
7910 (modref_lattice::merge_deref): Fix handling of indirect scape points.
7911 (update_escape_summary_1): Likewise.
7912 (update_escape_summary): Likewise.
7913 (ipa_merge_modref_summary_after_inlining): Likewise.
7915 2021-08-11 Richard Biener <rguenther@suse.de>
7917 PR middle-end/101858
7918 * fold-const.c (fold_binary_loc): Guard simplification
7919 of X < (cast) (1 << Y) to integer types.
7921 2021-08-11 Richard Biener <rguenther@suse.de>
7923 PR tree-optimization/101861
7924 * tree-vect-stmts.c (vectorizable_load): Fix error in
7925 previous change with regard to gather vectorization.
7927 2021-08-11 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
7930 * config/arm/arm_neon.h (vdup_n_s8): Replace call to builtin
7932 (vdup_n_s16): Likewise.
7933 (vdup_n_s32): Likewise.
7934 (vdup_n_s64): Likewise.
7935 (vdup_n_u8): Likewise.
7936 (vdup_n_u16): Likewise.
7937 (vdup_n_u32): Likewise.
7938 (vdup_n_u64): Likewise.
7939 (vdup_n_p8): Likewise.
7940 (vdup_n_p16): Likewise.
7941 (vdup_n_p64): Likewise.
7942 (vdup_n_f16): Likewise.
7943 (vdup_n_f32): Likewise.
7944 (vdupq_n_s8): Likewise.
7945 (vdupq_n_s16): Likewise.
7946 (vdupq_n_s32): Likewise.
7947 (vdupq_n_s64): Likewise.
7948 (vdupq_n_u8): Likewise.
7949 (vdupq_n_u16): Likewise.
7950 (vdupq_n_u32): Likewise.
7951 (vdupq_n_u64): Likewise.
7952 (vdupq_n_p8): Likewise.
7953 (vdupq_n_p16): Likewise.
7954 (vdupq_n_p64): Likewise.
7955 (vdupq_n_f16): Likewise.
7956 (vdupq_n_f32): Likewise.
7957 (vmov_n_s8): Replace call to builtin with call to corresponding
7959 (vmov_n_s16): Likewise.
7960 (vmov_n_s32): Likewise.
7961 (vmov_n_s64): Likewise.
7962 (vmov_n_u8): Likewise.
7963 (vmov_n_u16): Likewise.
7964 (vmov_n_u32): Likewise.
7965 (vmov_n_u64): Likewise.
7966 (vmov_n_p8): Likewise.
7967 (vmov_n_p16): Likewise.
7968 (vmov_n_f16): Likewise.
7969 (vmov_n_f32): Likewise.
7970 (vmovq_n_s8): Likewise.
7971 (vmovq_n_s16): Likewise.
7972 (vmovq_n_s32): Likewise.
7973 (vmovq_n_s64): Likewise.
7974 (vmovq_n_u8): Likewise.
7975 (vmovq_n_u16): Likewise.
7976 (vmovq_n_u32): Likewise.
7977 (vmovq_n_u64): Likewise.
7978 (vmovq_n_p8): Likewise.
7979 (vmovq_n_p16): Likewise.
7980 (vmovq_n_f16): Likewise.
7981 (vmovq_n_f32): Likewise.
7982 * config/arm/arm_neon_builtins.def: Remove entries for vdup_n.
7984 2021-08-11 liuhongt <hongtao.liu@intel.com>
7987 * config/i386/i386.md (ldexp<mode>3): Extend to vscalefs[sd]
7988 when TARGET_AVX512F and TARGET_SSE_MATH.
7990 2021-08-10 Jakub Jelinek <jakub@redhat.com>
7993 * config/i386/i386-expand.c (expand_vec_perm_even_odd): Return false
7994 for V32HImode if !TARGET_AVX512BW.
7995 (ix86_vectorize_vec_perm_const) <case E_V32HImode, case E_V64QImode>:
7996 If !TARGET_AVX512BW and TARGET_AVX512F and d.testing_p, don't fail
7997 early, but actually check the permutation.
7999 2021-08-10 Richard Biener <rguenther@suse.de>
8001 PR tree-optimization/101809
8002 * tree-vect-stmts.c (get_load_store_type): Allow emulated
8003 gathers with offset vector nunits being a constant multiple
8004 of the data vector nunits.
8005 (vect_get_gather_scatter_ops): Use the appropriate nunits
8006 for the offset vector defs.
8007 (vectorizable_store): Adjust call to
8008 vect_get_gather_scatter_ops.
8009 (vectorizable_load): Likewise. Handle the case of less
8010 offset vectors than data vectors.
8012 2021-08-10 Jakub Jelinek <jakub@redhat.com>
8015 * config/i386/sse.md (*avx512f_shuf_<shuffletype>64x2_1<mask_name>_1,
8016 *avx512f_shuf_<shuffletype>32x4_1<mask_name>_1): New define_insn
8019 2021-08-10 Richard Biener <rguenther@suse.de>
8021 PR tree-optimization/101801
8022 PR tree-optimization/101819
8023 * tree-vectorizer.h (vect_emulated_vector_p): Declare.
8024 * tree-vect-loop.c (vect_emulated_vector_p): New function.
8025 (vectorizable_reduction): Re-instantiate a check for emulated
8027 * tree-vect-stmts.c (vectorizable_shift): Likewise.
8028 (vectorizable_operation): Likewise. Cost emulated vector
8029 operations according to the scalar sequence synthesized by
8032 2021-08-10 Richard Biener <rguenther@suse.de>
8034 PR middle-end/101824
8035 * tree-nested.c (get_frame_field): Mark the COMPONENT_REF as
8036 volatile in case the variable was.
8038 2021-08-10 H.J. Lu <hjl.tools@gmail.com>
8041 * config/i386/constraints.md (BC): Document for integer SSE
8042 constant all bits set operand.
8043 (BF): New constraint for const floating-point all bits set
8045 * config/i386/i386.c (standard_sse_constant_p): Likewise.
8046 (standard_sse_constant_opcode): Likewise.
8047 * config/i386/sse.md (sseconstm1): New mode attribute.
8048 (mov<mode>_internal): Replace BC with <sseconstm1>.
8050 2021-08-10 liuhongt <hongtao.liu@intel.com>
8052 * config/i386/sse.md (cond_<insn><mode>): New expander.
8053 (VI248_AVX512VLBW): New mode iterator.
8054 * config/i386/predicates.md
8055 (nonimmediate_or_const_vec_dup_operand): New predicate.
8057 2021-08-09 Andrew MacLeod <amacleod@redhat.com>
8059 PR tree-optimization/101741
8060 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Check
8061 type of parameter for toupper/tolower.
8063 2021-08-09 Martin Jambor <mjambor@suse.cz>
8066 * ipa-prop.c (propagate_controlled_uses): Removed a spurious space.
8068 2021-08-09 Pat Haugen <pthaugen@linux.ibm.com>
8070 * config/rs6000/rs6000.c (is_load_insn1): Verify destination is a
8072 (is_store_insn1): Verify source is a register.
8074 2021-08-09 Uroš Bizjak <ubizjak@gmail.com>
8077 * config/i386/mmx.md (<any_logic:code>v2sf3):
8078 Rename from *mmx_<any_logic:code>v2sf3
8080 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
8082 * config/nvptx/nvptx.c: Cross-reference parts adapted in
8083 'gcc/omp-oacc-neuter-broadcast.cc'.
8084 * omp-low.c: Likewise.
8085 * omp-oacc-neuter-broadcast.cc: Cross-reference parts adapted from
8088 2021-08-09 Julian Brown <julian@codesourcery.com>
8089 Kwok Cheung Yeung <kcy@codesourcery.com>
8090 Thomas Schwinge <thomas@codesourcery.com>
8092 * config/gcn/gcn.c (gcn_init_builtins): Override decls for
8093 BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START,
8094 BUILT_IN_GOACC_SINGLE_COPY_END and BUILT_IN_GOACC_BARRIER.
8095 (gcn_goacc_validate_dims): Turn on worker partitioning unconditionally.
8096 (gcn_fork_join): Update comment.
8097 * config/gcn/gcn.opt (flag_worker_partitioning): Remove.
8098 (macc_experimental_workers): Remove unused option.
8100 2021-08-09 Julian Brown <julian@codesourcery.com>
8101 Nathan Sidwell <nathan@codesourcery.com> (via 'gcc/config/nvptx/nvptx.c' master)
8102 Kwok Cheung Yeung <kcy@codesourcery.com>
8103 Thomas Schwinge <thomas@codesourcery.com>
8105 * Makefile.in (OBJS): Add omp-oacc-neuter-broadcast.o.
8106 * doc/tm.texi.in (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD):
8107 Add documentation hook.
8108 * doc/tm.texi: Regenerate.
8109 * omp-oacc-neuter-broadcast.cc: New file.
8110 * omp-builtins.def (BUILT_IN_GOACC_BARRIER)
8111 (BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START)
8112 (BUILT_IN_GOACC_SINGLE_COPY_END): New builtins.
8113 * passes.def (pass_omp_oacc_neuter_broadcast): Add pass.
8114 * target.def (goacc.create_worker_broadcast_record): Add target
8116 * tree-pass.h (make_pass_omp_oacc_neuter_broadcast): Add
8118 * config/gcn/gcn-protos.h (gcn_goacc_adjust_propagation_record):
8119 Rename prototype to...
8120 (gcn_goacc_create_worker_broadcast_record): ... this.
8121 * config/gcn/gcn-tree.c (gcn_goacc_adjust_propagation_record): Rename
8123 (gcn_goacc_create_worker_broadcast_record): ... this.
8124 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_PROPAGATION_RECORD):
8126 (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD): ... this.
8128 2021-08-09 Tejas Belagod <tejas.belagod@arm.com>
8131 * config/aarch64/aarch64-simd.md (vlshr<mode>3, vashr<mode>3): Use
8134 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
8136 * Makefile.in (GTFILES): Remove '$(srcdir)/omp-offload.c'.
8138 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
8140 * builtins.def (DEF_GOACC_BUILTIN, DEF_GOMP_BUILTIN): Don't
8141 consider '-foffload-abi'.
8142 * common.opt (-foffload-abi): Remove 'Var', 'Init'.
8143 * opts.c (common_handle_option) <-foffload-abi> [ACCEL_COMPILER]:
8146 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
8148 * optc-gen.awk: Sanity check that 'Init' doesn't appear without
8151 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
8153 * omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE): Remove.
8155 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
8157 * doc/gty.texi (Files): Update.
8159 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
8161 * doc/gty.texi (Files): Fix GTY header file example.
8163 2021-08-09 Roger Sayle <roger@nextmovesoftware.com>
8165 * tree-ssa-ccp.c (value_mask_to_min_max): Helper function to
8166 determine the upper and lower bounds from a mask-value pair.
8167 (bit_value_unop) [ABS_EXPR, ABSU_EXPR]: Add support for
8168 absolute value and unsigned absolute value expressions.
8169 (bit_value_binop): Initialize *VAL's precision.
8170 [LT_EXPR, LE_EXPR]: Use value_mask_to_min_max to determine
8171 upper and lower bounds of operands. Add LE_EXPR/GE_EXPR
8172 support when the operands are unknown but potentially equal.
8173 [MIN_EXPR, MAX_EXPR]: Support minimum/maximum expressions.
8175 2021-08-09 Bin Cheng <bin.cheng@linux.alibaba.com>
8177 * config/aarch64/aarch64.md
8178 (*extend<SHORT:mode><GPI:mode>2_aarch64): Use %<GPI:w>0.
8180 2021-08-08 Sergei Trofimovich <siarheit@google.com>
8182 * lra-constraints.c: Fix s/otput/output/ typo.
8184 2021-08-06 Martin Sebor <msebor@redhat.com>
8186 * builtins.c (expand_builtin_memchr): Move to gimple-ssa-warn-access.cc.
8187 (expand_builtin_strcat): Same.
8188 (expand_builtin_stpncpy): Same.
8189 (expand_builtin_strncat): Same.
8190 (check_read_access): Same.
8191 (check_memop_access): Same.
8192 (expand_builtin_strlen): Move checks to gimple-ssa-warn-access.cc.
8193 (expand_builtin_strnlen): Same.
8194 (expand_builtin_memcpy): Same.
8195 (expand_builtin_memmove): Same.
8196 (expand_builtin_mempcpy): Same.
8197 (expand_builtin_strcpy): Same.
8198 (expand_builtin_strcpy_args): Same.
8199 (expand_builtin_stpcpy_1): Same.
8200 (expand_builtin_strncpy): Same.
8201 (expand_builtin_memset): Same.
8202 (expand_builtin_bzero): Same.
8203 (expand_builtin_strcmp): Same.
8204 (expand_builtin_strncmp): Same.
8205 (expand_builtin): Remove handlers.
8206 (fold_builtin_strlen): Add a comment.
8207 * builtins.h (check_access): Move to gimple-ssa-warn-access.cc.
8208 * calls.c (maybe_warn_nonstring_arg): Same.
8209 * diagnostic-spec.c (nowarn_spec_t::nowarn_spec_t): Add warning option.
8210 * gimple-fold.c (gimple_fold_builtin_strcpy): Pass argument to callee.
8211 (gimple_fold_builtin_stpcpy): Same.
8212 * gimple-ssa-warn-access.cc (has_location): New function.
8213 (get_location): Same.
8214 (get_callee_fndecl): Same.
8217 (warn_string_no_nul): Define.
8218 (unterminated_array): Same.
8219 (check_nul_terminated_array): Same.
8220 (maybe_warn_nonstring_arg): Same.
8221 (maybe_warn_for_bound): Same.
8222 (warn_for_access): Same.
8223 (check_access): Same.
8224 (check_memop_access): Same.
8225 (check_read_access): Same.
8226 (warn_dealloc_offset): Use helper functions.
8227 (maybe_emit_free_warning): Same.
8228 (class pass_waccess): Add members.
8229 (check_strcat): New function.
8230 (check_strncat): New function.
8231 (check_stxcpy): New function.
8232 (check_stxncpy): New function.
8233 (check_strncmp): New function.
8234 (pass_waccess::check_builtin): New function.
8235 (pass_waccess::check): Call it.
8236 * gimple-ssa-warn-access.h (warn_string_no_nul): Move here from
8238 (maybe_warn_for_bound): Same.
8239 (check_access): Same.
8240 (check_memop_access): Same.
8241 (check_read_access): Same.
8242 * pointer-query.h (struct access_data): Define a ctor overload.
8244 2021-08-06 Richard Biener <rguenther@suse.de>
8246 PR tree-optimization/101801
8247 * tree-vectorizer.h (vect_worthwhile_without_simd_p): Rename...
8248 (vect_can_vectorize_without_simd_p): ... to this.
8249 * tree-vect-loop.c (vect_worthwhile_without_simd_p): Rename...
8250 (vect_can_vectorize_without_simd_p): ... to this and fold
8251 in vect_min_worthwhile_factor.
8252 (vect_min_worthwhile_factor): Remove.
8253 (vectorizable_reduction): Adjust and remove the cost part.
8254 * tree-vect-stmts.c (vectorizable_shift): Likewise.
8255 (vectorizable_operation): Likewise.
8257 2021-08-06 Uroš Bizjak <ubizjak@gmail.com>
8260 * config/i386/i386.md (cmove reg-to-reg move elimination peephole2s):
8261 Add general_gr_operand predicate to operand 3.
8263 2021-08-06 Roger Sayle <roger@nextmovesoftware.com>
8265 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern): Use
8266 CFN_BUILT_IN_CLRSB* instead of BUILT_IN_CLRSB* for consistency.
8268 2021-08-06 Tamar Christina <tamar.christina@arm.com>
8270 * config/aarch64/aarch64-sve-builtins.cc (register_svpattern,
8271 register_svprfop): Pass vec<> by pointer.
8272 * langhooks-def.h (lhd_simulate_enum_decl): Likewise.
8273 * langhooks.c (lhd_simulate_enum_decl): Likewise.
8274 * langhooks.h (struct lang_hooks_for_types): Likewise.
8276 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
8278 * config/aarch64/arm_neon.h (vst1_bf16_x2): Use
8279 __builtin_memcpy instead of constructing an additional
8280 __builtin_aarch64_simd_oi one vector at a time.
8281 (vst1q_bf16_x2): Likewise.
8282 (vst1_bf16_x3): Use __builtin_memcpy instead of constructing
8283 an additional __builtin_aarch64_simd_ci one vector at a time.
8284 (vst1q_bf16_x3): Likewise.
8285 (vst1_bf16_x4): Use __builtin_memcpy instead of a union.
8286 (vst1q_bf16_x4): Likewise.
8287 (vst2_bf16): Use __builtin_memcpy instead of constructing an
8288 additional __builtin_aarch64_simd_oi one vector at a time.
8289 (vst2q_bf16): Likewise.
8290 (vst3_bf16): Use __builtin_memcpy instead of constructing an
8291 additional __builtin_aarch64_simd_ci mode one vector at a
8293 (vst3q_bf16): Likewise.
8294 (vst4_bf16): Use __builtin_memcpy instead of constructing an
8295 additional __builtin_aarch64_simd_xi one vector at a time.
8296 (vst4q_bf16): Likewise.
8298 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
8300 * config/aarch64/arm_neon.h (__ST2_LANE_FUNC): Delete.
8301 (__ST2Q_LANE_FUNC): Delete.
8302 (vst2_lane_f16): Use __builtin_memcpy to copy vector
8303 structure instead of constructing __builtin_aarch64_simd_oi
8304 one vector at a time.
8305 (vst2_lane_f32): Likewise.
8306 (vst2_lane_f64): Likewise.
8307 (vst2_lane_p8): Likewise.
8308 (vst2_lane_p16): Likewise.
8309 (vst2_lane_p64): Likewise.
8310 (vst2_lane_s8): Likewise.
8311 (vst2_lane_s16): Likewise.
8312 (vst2_lane_s32): Likewise.
8313 (vst2_lane_s64): Likewise.
8314 (vst2_lane_u8): Likewise.
8315 (vst2_lane_u16): Likewise.
8316 (vst2_lane_u32): Likewise.
8317 (vst2_lane_u64): Likewise.
8318 (vst2_lane_bf16): Likewise.
8319 (vst2q_lane_f16): Use __builtin_memcpy to copy vector
8320 structure instead of using a union.
8321 (vst2q_lane_f32): Likewise.
8322 (vst2q_lane_f64): Likewise.
8323 (vst2q_lane_p8): Likewise.
8324 (vst2q_lane_p16): Likewise.
8325 (vst2q_lane_p64): Likewise.
8326 (vst2q_lane_s8): Likewise.
8327 (vst2q_lane_s16): Likewise.
8328 (vst2q_lane_s32): Likewise.
8329 (vst2q_lane_s64): Likewise.
8330 (vst2q_lane_u8): Likewise.
8331 (vst2q_lane_u16): Likewise.
8332 (vst2q_lane_u32): Likewise.
8333 (vst2q_lane_u64): Likewise.
8334 (vst2q_lane_bf16): Likewise.
8336 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
8338 * config/aarch64/arm_neon.h (__ST3_LANE_FUNC): Delete.
8339 (__ST3Q_LANE_FUNC): Delete.
8340 (vst3_lane_f16): Use __builtin_memcpy to copy vector
8341 structure instead of constructing __builtin_aarch64_simd_ci
8342 one vector at a time.
8343 (vst3_lane_f32): Likewise.
8344 (vst3_lane_f64): Likewise.
8345 (vst3_lane_p8): Likewise.
8346 (vst3_lane_p16): Likewise.
8347 (vst3_lane_p64): Likewise.
8348 (vst3_lane_s8): Likewise.
8349 (vst3_lane_s16): Likewise.
8350 (vst3_lane_s32): Likewise.
8351 (vst3_lane_s64): Likewise.
8352 (vst3_lane_u8): Likewise.
8353 (vst3_lane_u16): Likewise.
8354 (vst3_lane_u32): Likewise.
8355 (vst3_lane_u64): Likewise.
8356 (vst3_lane_bf16): Likewise.
8357 (vst3q_lane_f16): Use __builtin_memcpy to copy vector
8358 structure instead of using a union.
8359 (vst3q_lane_f32): Likewise.
8360 (vst3q_lane_f64): Likewise.
8361 (vst3q_lane_p8): Likewise.
8362 (vst3q_lane_p16): Likewise.
8363 (vst3q_lane_p64): Likewise.
8364 (vst3q_lane_s8): Likewise.
8365 (vst3q_lane_s16): Likewise.
8366 (vst3q_lane_s32): Likewise.
8367 (vst3q_lane_s64): Likewise.
8368 (vst3q_lane_u8): Likewise.
8369 (vst3q_lane_u16): Likewise.
8370 (vst3q_lane_u32): Likewise.
8371 (vst3q_lane_u64): Likewise.
8372 (vst3q_lane_bf16): Likewise.
8374 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
8376 * config/aarch64/arm_neon.h (__ST4_LANE_FUNC): Delete.
8377 (__ST4Q_LANE_FUNC): Delete.
8378 (vst4_lane_f16): Use __builtin_memcpy to copy vector
8379 structure instead of constructing __builtin_aarch64_simd_xi
8380 one vector at a time.
8381 (vst4_lane_f32): Likewise.
8382 (vst4_lane_f64): Likewise.
8383 (vst4_lane_p8): Likewise.
8384 (vst4_lane_p16): Likewise.
8385 (vst4_lane_p64): Likewise.
8386 (vst4_lane_s8): Likewise.
8387 (vst4_lane_s16): Likewise.
8388 (vst4_lane_s32): Likewise.
8389 (vst4_lane_s64): Likewise.
8390 (vst4_lane_u8): Likewise.
8391 (vst4_lane_u16): Likewise.
8392 (vst4_lane_u32): Likewise.
8393 (vst4_lane_u64): Likewise.
8394 (vst4_lane_bf16): Likewise.
8395 (vst4q_lane_f16): Use __builtin_memcpy to copy vector
8396 structure instead of using a union.
8397 (vst4q_lane_f32): Likewise.
8398 (vst4q_lane_f64): Likewise.
8399 (vst4q_lane_p8): Likewise.
8400 (vst4q_lane_p16): Likewise.
8401 (vst4q_lane_p64): Likewise.
8402 (vst4q_lane_s8): Likewise.
8403 (vst4q_lane_s16): Likewise.
8404 (vst4q_lane_s32): Likewise.
8405 (vst4q_lane_s64): Likewise.
8406 (vst4q_lane_u8): Likewise.
8407 (vst4q_lane_u16): Likewise.
8408 (vst4q_lane_u32): Likewise.
8409 (vst4q_lane_u64): Likewise.
8410 (vst4q_lane_bf16): Likewise.
8412 2021-08-06 Martin Liska <mliska@suse.cz>
8414 * config/rs6000/rs6000.c (rs6000_option_override_internal): When
8415 a target option is restored, it can have
8416 rs6000_long_double_type_size set to FLOAT_PRECISION_TFmode
8417 and error should not be emitted.
8419 2021-08-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
8421 * gcov-io.h (gcov_write): Declare.
8422 * gcov-io.c (gcov_write): New.
8423 (gcov_write_counter): Remove.
8424 (gcov_write_tag_length): Likewise.
8425 (gcov_write_summary): Replace gcov_write_tag_length() with calls to
8426 gcov_write_unsigned().
8427 * doc/invoke.texi (fprofile-info-section): Mention
8428 __gcov_info_to_gdca().
8430 2021-08-06 Martin Sebor <msebor@redhat.com>
8432 * dominance.c (prune_bbs_to_update_dominators): Adjust by-value vec
8433 arguments to by-reference.
8434 (iterate_fix_dominators): Same.
8435 * dominance.h (iterate_fix_dominators): Same.
8436 * ipa-prop.h: Call auto_vec::to_vec_legacy.
8437 * tree-data-ref.c (dump_data_dependence_relation): Adjust by-value vec
8438 arguments to by-reference.
8439 (debug_data_dependence_relation): Same.
8440 (dump_data_dependence_relations): Same.
8441 * tree-data-ref.h (debug_data_dependence_relation): Same.
8442 (dump_data_dependence_relations): Same.
8443 * tree-predcom.c (dump_chains): Same.
8444 (initialize_root_vars_lm): Same.
8445 (determine_unroll_factor): Same.
8446 (replace_phis_by_defined_names): Same.
8447 (insert_init_seqs): Same.
8448 (pcom_worker::tree_predictive_commoning_loop): Call
8449 auto_vec::to_vec_legacy.
8450 * tree-ssa-pre.c (insert_into_preds_of_block): Adjust by-value vec
8451 arguments to by-reference.
8452 * tree-ssa-threadbackward.c (populate_worklist): Same.
8453 (back_threader::resolve_def): Same.
8454 * tree-vect-data-refs.c (vect_check_nonzero_value): Same.
8455 (vect_enhance_data_refs_alignment): Same.
8456 (vect_check_lower_bound): Same.
8457 (vect_prune_runtime_alias_test_list): Same.
8458 (vect_permute_store_chain): Same.
8459 * tree-vect-slp-patterns.c (vect_normalize_conj_loc): Same.
8460 * tree-vect-stmts.c (vect_create_vectorized_demotion_stmts): Same.
8461 * tree-vectorizer.h (vect_permute_store_chain): Same.
8462 * vec.c (test_init): New function.
8463 (vec_c_tests): Call new function.
8464 * vec.h (vec): Declare ctors, dtor, and assignment.
8465 (auto_vec::vec_to_legacy): New function.
8466 (vec::copy): Adjust initialization.
8468 2021-08-05 H.J. Lu <hjl.tools@gmail.com>
8471 * config/i386/i386.c (ix86_can_inline_p): Ignore MASK_80387 if
8472 callee only uses GPRs.
8473 * config/i386/ia32intrin.h: Revert commit 5463cee2770.
8474 * config/i386/serializeintrin.h: Revert commit 71958f740f1.
8475 * config/i386/x86gprintrin.h: Add
8476 #pragma GCC target("general-regs-only") and #pragma GCC pop_options
8477 to disable non-GPR ISAs.
8479 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
8481 PR middle-end/101787
8482 * doc/md.texi (cond_ashl, cond_ashr, cond_lshr): Document.
8484 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
8486 * tree-vectorizer.h (vect_is_store_elt_extraction, vect_is_reduction)
8487 (vect_reduc_type, vect_embedded_comparison_type, vect_comparison_type)
8488 (vect_is_extending_load, vect_is_integer_truncation): New functions,
8489 moved from aarch64.c but given different names.
8490 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction)
8491 (aarch64_is_reduction, aarch64_reduc_type)
8492 (aarch64_embedded_comparison_type, aarch64_comparison_type)
8493 (aarch64_extending_load_p, aarch64_integer_truncation_p): Delete
8494 in favor of the above. Update callers accordingly.
8496 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
8499 * config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
8500 writing .cpu directive in asm output.
8501 * config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
8502 (arm_last_printed_arch_string): Delete.
8503 (arm_last-printed_fpu_string): Delete.
8504 (arm_configure_build_target): If use of floating-point/SIMD is
8505 disabled, remove all fp/simd related features from the target ISA.
8506 (last_arm_targ_options): New variable.
8507 (arm_print_asm_arch_directives): Add new parameters. Change order
8508 of emitted directives and handle all cases here.
8509 (arm_file_start): Always call arm_print_asm_arch_directives, move
8510 all generation of .arch/.arch_extension here.
8511 (arm_file_end): Call arm_print_asm_arch.
8512 (arm_declare_function_name): Call arm_print_asm_arch_directives
8513 instead of printing .arch/.fpu directives directly.
8515 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
8517 * config/arm/arm.c (arm_configure_build_target): Don't call
8518 arm_option_reconfigure_globals.
8519 (arm_option_restore): Call arm_option_reconfigure_globals after
8520 reconfiguring the target.
8521 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
8523 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
8525 * config/arm/arm.c (arm_configure_build_target): Ensure the target's
8526 arch_name is always set.
8528 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
8530 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
8531 of vec_select high-half from being added into Neon subtract
8534 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
8536 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
8537 of vec_select high-half from being added into Neon add cost.
8539 2021-08-05 Kewen Lin <linkw@linux.ibm.com>
8541 * cfgloop.h (loops_list::loops_list): Add one optional argument
8542 root and adjust accordingly, update loop tree walking and factor
8544 * cfgloop.c (loops_list::walk_loop_tree): ... this. New function.
8546 2021-08-05 Eric Botcazou <ebotcazou@gcc.gnu.org>
8548 PR tree-optimization/101626
8549 * tree-sra.c (propagate_subaccesses_from_rhs): Do not set the
8550 reverse scalar storage order on a pointer or vector component.
8552 2021-08-05 liuhongt <hongtao.liu@intel.com>
8554 * config/i386/sse.md (cond_<code><mode>): New expander.
8556 2021-08-05 liuhongt <hongtao.liu@intel.com>
8558 * config/i386/sse.md (cond_<code><mode>): New expander.
8560 2021-08-05 liuhongt <hongtao.liu@intel.com>
8562 * config/i386/sse.md (cond_<code><mode>): New expander.
8564 2021-08-04 David Malcolm <dmalcolm@redhat.com>
8567 * Makefile.in (ANALYZER_OBJS): Add analyzer/region-model-asm.o.
8569 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
8572 * config/i386/i386.h (STORE_MAX_PIECES): Allow 16/32/64 bytes
8573 only if TARGET_INTER_UNIT_MOVES_TO_VEC is true.
8575 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
8578 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
8579 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
8580 data with SSE register from one memory location to another.
8582 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
8584 * config/s390/s390.c (expand_perm_with_vpdi): New function.
8585 (vectorize_vec_perm_const_1): Call expand_perm_with_vpdi.
8586 * config/s390/vector.md (*vpdi1<mode>, @vpdi1<mode>): Enable a
8587 parameterized expander.
8588 (*vpdi4<mode>, @vpdi4<mode>): Likewise.
8590 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
8592 * config/s390/s390.c (MAX_VECT_LEN): Define macro.
8593 (struct expand_vec_perm_d): Define struct.
8594 (expand_perm_with_merge): New function.
8595 (vectorize_vec_perm_const_1): New function.
8596 (s390_vectorize_vec_perm_const): New function.
8597 (TARGET_VECTORIZE_VEC_PERM_CONST): Define target macro.
8599 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
8601 * config/s390/vector.md (V_HW_64): Remove mode iterator.
8602 (*vec_load_pair<mode>): Use V_HW_2 instead of V_HW_64.
8603 * config/s390/vx-builtins.md
8604 (vec_scatter_element<V_HW_2:mode>_SI): Use V_HW_2 instead of
8607 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
8609 * config/s390/s390.md (UNSPEC_VEC_PERMI): Remove constant
8611 * config/s390/vector.md (*vpdi1<mode>, *vpdi4<mode>): New pattern
8613 * config/s390/vx-builtins.md (*vec_permi<mode>): Emit generic rtx
8614 instead of an unspec.
8616 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
8618 * config/s390/s390-modes.def: Add more vector modes to support
8619 concatenation of two vectors.
8620 * config/s390/s390-protos.h (s390_expand_merge_perm_const): Add
8622 (s390_expand_merge): Likewise.
8623 * config/s390/s390.c (s390_expand_merge_perm_const): New function.
8624 (s390_expand_merge): New function.
8625 * config/s390/s390.md (UNSPEC_VEC_MERGEH, UNSPEC_VEC_MERGEL):
8626 Remove constant definitions.
8627 * config/s390/vector.md (V_HW_2): Add mode iterators.
8628 (VI_HW_4, V_HW_4): Rename VI_HW_4 to V_HW_4.
8629 (vec_2x_nelts, vec_2x_wide): New mode attributes.
8630 (*vmrhb, *vmrlb, *vmrhh, *vmrlh, *vmrhf, *vmrlf, *vmrhg, *vmrlg):
8631 New pattern definitions.
8632 (vec_widen_umult_lo_<mode>, vec_widen_umult_hi_<mode>)
8633 (vec_widen_smult_lo_<mode>, vec_widen_smult_hi_<mode>)
8634 (vec_unpacks_lo_v4sf, vec_unpacks_hi_v4sf, vec_unpacks_lo_v2df)
8635 (vec_unpacks_hi_v2df): Adjust expanders to emit non-unspec RTX for
8637 * config/s390/vx-builtins.md (V_HW_4): Remove mode iterator. Now
8639 (vec_mergeh<mode>, vec_mergel<mode>): Use s390_expand_merge to
8640 emit vec merge pattern.
8642 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
8644 * config/aarch64/aarch64.c (aarch64_strip_extend_vec_half):
8646 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent cost of
8647 vec_select high-half from being added into Neon multiply
8649 * rtlanal.c (vec_series_highpart_p): Define.
8650 * rtlanal.h (vec_series_highpart_p): Declare.
8652 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
8654 * config/aarch64/aarch64.c (aarch64_strip_duplicate_vec_elt):
8656 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent
8657 vec_select cost from being added into Neon multiply cost.
8659 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
8661 * tree-vect-loop.c (vect_better_loop_vinfo_p): Detect cases in
8662 which old_loop_vinfo is an epilogue loop that handles a constant
8663 number of iterations.
8665 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
8667 * tree-vect-loop.c (vect_analyze_loop): Print a dump message
8668 when a reanalyzed loop fails to be cheaper than the current
8671 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
8673 * config/aarch64/aarch64.c: Fix a typo.
8675 2021-08-04 Vincent Lefèvre <vincent-gcc@vinc17.net>
8677 PR gcov-profile/101773
8678 * gcov-io.c (gcov_close): Check return code of a fclose.
8680 2021-08-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
8683 * dwarf2out.c (dwarf2out_assembly_start): Emit a dummy
8684 .file statement when needed.
8686 2021-08-04 Richard Biener <rguenther@suse.de>
8688 * tree-vect-data-refs.c (vect_check_gather_scatter):
8689 Include widening conversions only when the result is
8690 still handed by native gather or the current offset
8691 size not already matches the data size.
8692 Also succeed analysis in case there's no native support,
8693 noted by a IFN_LAST ifn and a NULL decl.
8694 (vect_analyze_data_refs): Always consider gathers.
8695 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
8696 Test for no IFN gather rather than decl gather.
8697 * tree-vect-stmts.c (vect_model_load_cost): Pass in the
8698 gather-scatter info and cost emulated gathers accordingly.
8699 (vect_truncate_gather_scatter_offset): Properly test for
8701 (vect_use_strided_gather_scatters_p): Likewise.
8702 (get_load_store_type): Handle emulated gathers and its
8704 (vectorizable_load): Likewise. Emulate them by extracting
8705 scalar offsets, doing scalar loads and a vector construct.
8707 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
8710 * expr.c (op_by_pieces_d::op_by_pieces_d): Add a max_pieces
8711 argument to set m_max_size.
8712 (move_by_pieces_d): Pass MOVE_MAX_PIECES to op_by_pieces_d.
8713 (store_by_pieces_d): Pass STORE_MAX_PIECES to op_by_pieces_d.
8714 (compare_by_pieces_d): Pass COMPARE_MAX_PIECES to op_by_pieces_d.
8716 2021-08-04 Roger Sayle <roger@nextmovesoftware.com>
8717 Marc Glisse <marc.glisse@inria.fr>
8719 * match.pd (bit_ior, bit_xor): Canonicalize (X*C1)|(X*C2) and
8720 (X*C1)^(X*C2) as X*(C1+C2), and related variants, using
8721 tree_nonzero_bits to ensure that operands are bit-wise disjoint.
8723 2021-08-04 Richard Biener <rguenther@suse.de>
8725 * tree-ssa-forwprop.c (pass_forwprop::execute): Split
8726 out code to decompose vector loads ...
8727 (optimize_vector_load): ... here. Generalize it to
8728 handle intermediate widening and TARGET_MEM_REF loads
8729 and apply it to loads with a supported vector mode as well.
8731 2021-08-04 Richard Biener <rguenther@suse.de>
8733 PR tree-optimization/101756
8734 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Make sure
8735 the result of the reduction epilogue is compatible to the original
8738 2021-08-04 liuhongt <hongtao.liu@intel.com>
8741 * config/i386/i386.md (peephole2): Refine predicate from
8742 register_operand to general_reg_operand.
8744 2021-08-04 Aldy Hernandez <aldyh@redhat.com>
8746 * gimple-range-path.h (path_range_query::dump): Mark override.
8748 2021-08-04 Richard Biener <rguenther@suse.de>
8750 PR tree-optimization/101769
8751 * tree-tailcall.c (eliminate_tail_call): Add the created loop
8752 for the first recursion and return it via the new output parameter.
8753 (optimize_tail_call): Pass through new output param.
8754 (tree_optimize_tail_calls_1): After creating all latches,
8755 add the created loop to the loop tree. Do not mark loops for fixup.
8757 2021-08-04 Martin Liska <mliska@suse.cz>
8759 * doc/invoke.texi: Document threader-mode param.
8761 2021-08-04 liuhongt <hongtao.liu@intel.com>
8763 * config/i386/sse.md (cond_fma<mode>): New expander.
8764 (cond_fms<mode>): Ditto.
8765 (cond_fnma<mode>): Ditto.
8766 (cond_fnms<mode>): Ditto.
8768 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
8770 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode>): Use && instead of &.
8772 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
8774 * config/rs6000/constraints.md: Remove "e" from the list of available
8775 constraint characters.
8777 2021-08-03 Eugene Rozenfeld <erozen@microsoft.com>
8779 PR gcov-profile/71672
8780 * auto-profile.c (afdo_indirect_call): Fix setup of the historgram value for indirect calls.
8782 2021-08-03 Paul A. Clarke <pc@us.ibm.com>
8784 * config/rs6000/smmintrin.h (_mm_minpos_epu16): New.
8786 2021-08-03 H.J. Lu <hjl.tools@gmail.com>
8788 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): In 64-bit mode,
8789 try XMM31 to avoid vzeroupper.
8791 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8793 * doc/invoke.texi: Document -mtune=neoverse-512tvb and
8794 -mcpu=neoverse-512tvb.
8795 * config/aarch64/aarch64-cores.def (neoverse-512tvb): New entry.
8796 * config/aarch64/aarch64-tune.md: Regenerate.
8797 * config/aarch64/aarch64.c (neoverse512tvb_sve_vector_cost)
8798 (neoverse512tvb_sve_issue_info, neoverse512tvb_vec_issue_info)
8799 (neoverse512tvb_vector_cost, neoverse512tvb_tunings): New structures.
8800 (aarch64_adjust_body_cost_sve): Handle -mtune=neoverse-512tvb.
8801 (aarch64_adjust_body_cost): Likewise.
8803 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8805 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Only
8806 record issue information for operations that occur in the
8809 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8811 * config/aarch64/aarch64.c (aarch64_multiply_add_p): Add a vec_flags
8812 parameter. Detect cases in which an Advanced SIMD MLA would almost
8813 certainly require a MOV.
8814 (aarch64_count_ops): Update accordingly.
8816 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8818 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction): New
8819 function, split out from...
8820 (aarch64_detect_vector_stmt_subtype): ...here.
8821 (aarch64_add_stmt_cost): Treat extracting element 0 as free.
8823 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8825 * config/aarch64/aarch64-protos.h (sve_vec_cost):
8826 Add gather_load_x32_cost and gather_load_x64_cost.
8827 * config/aarch64/aarch64.c (generic_sve_vector_cost)
8828 (a64fx_sve_vector_cost, neoversev1_sve_vector_cost): Update
8829 accordingly, using the values given by the scalar_load * number
8830 of elements calculation that we used previously.
8831 (aarch64_detect_vector_stmt_subtype): Use the new fields.
8833 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8835 * config/aarch64/aarch64.c (aarch64_adjust_body_cost_sve): New
8836 function, split out from...
8837 (aarch64_adjust_body_cost): ...here.
8839 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8841 * config/aarch64/fractional-cost.h: New file.
8842 * config/aarch64/aarch64.c: Include <algorithm> (indirectly)
8843 and cost_fraction.h.
8844 (vec_cost_fraction): New typedef.
8845 (aarch64_detect_scalar_stmt_subtype): Use it for statement costs.
8846 (aarch64_detect_vector_stmt_subtype): Likewise.
8847 (aarch64_sve_adjust_stmt_cost, aarch64_adjust_stmt_cost): Likewise.
8848 (aarch64_estimate_min_cycles_per_iter): Use vec_cost_fraction
8850 (aarch64_adjust_body_cost): Likewise.
8851 (aarch64_test_cost_fraction): New function.
8852 (aarch64_run_selftests): Call it.
8854 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8856 * config/aarch64/aarch64-protos.h (tune_params::sve_width): Turn
8858 * config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Update
8860 (aarch64_estimated_poly_value): Likewise. Use the least significant
8861 set bit for the minimum and likely values. Use the most significant
8862 set bit for the maximum value.
8864 2021-08-03 liuhongt <hongtao.liu@intel.com>
8866 * config/i386/sse.md (cond_<insn><mode>): New expander.
8867 (cond_mul<mode>): Ditto.
8869 2021-08-03 Kewen Lin <linkw@linux.ibm.com>
8871 * tree-cfg.c (move_sese_region_to_fn): Fix typos on dloop.
8873 2021-08-03 liuhongt <hongtao.liu@intel.com>
8875 * config/i386/sse.md (cond_<insn><mode>):New expander.
8876 (cond_mul<mode>): Ditto.
8877 (cond_div<mode>): Ditto.
8879 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
8881 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Also
8882 check stack_realign_needed for stack realignment.
8883 (ix86_legitimate_constant_p): Always allow CONST_WIDE_INT smaller
8884 than the largest integer supported by vector register.
8885 * config/i386/i386.h (MAX_MOVE_MAX): New. Set to 64.
8886 (MOVE_MAX): Set to bytes of the largest integer supported by
8888 (STORE_MAX_PIECES): New.
8890 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
8892 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
8893 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
8894 data from one memory location to another.
8896 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
8899 * config/i386/i386.c (TARGET_GEN_MEMSET_SCRATCH_RTX): New.
8901 2021-08-02 Aldy Hernandez <aldyh@redhat.com>
8903 PR tree-optimization/101724
8904 * params.opt: Remove --param=threader-iterative.
8905 * tree-ssa-threadbackward.c (pass_thread_jumps::execute): Remove
8908 2021-08-02 Tom de Vries <tdevries@suse.de>
8910 PR middle-end/101665
8911 * doc/extend.texi (nonnull attribute): Improve documentation.
8913 2021-08-02 Andrew Pinski <apinski@marvell.com>
8915 PR rtl-optimization/101683
8916 * rtlanal.c (may_trap_p_1): Handle UNSIGNED_FIX.
8918 2021-08-02 Roger Sayle <roger@nextmovesoftware.com>
8920 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern):
8921 Renamed from cond_removal_in_popcount_clz_ctz_pattern.
8922 Add support for BSWAP, FFS, PARITY and CLRSB builtins.
8923 (tree_ssa_phiop_worker): Update call to function above.
8925 2021-08-01 H.J. Lu <hjl.tools@gmail.com>
8928 * config/i386/i386.md (bsr_rex64_1_zext): New.
8929 (combine splitter for constant - clzll): Replace gen_bsr_rex64_1
8930 with gen_bsr_rex64_1_zext.
8932 2021-07-31 Jakub Jelinek <jakub@redhat.com>
8935 * config/i386/i386.md (bsr_rex64_1, bsr_1, bsr_zext_1): New
8936 define_insn patterns.
8937 (*bsr_rex64_2, *bsr_2): New define_insn_and_split patterns.
8938 Add combine splitters for constant - clz.
8939 (clz<mode>2): Use a temporary pseudo for bsr result.
8941 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
8943 * config/rs6000/smmintrin.h (_mm_floor_pd, _mm_floor_ps,
8944 _mm_floor_sd, _mm_floor_ss): New.
8946 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
8948 * config/rs6000/smmintrin.h (_mm_ceil_pd, _mm_ceil_ps,
8949 _mm_ceil_sd, _mm_ceil_ss): New.
8951 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
8953 * config/rs6000/smmintrin.h (_mm_blend_pd, _mm_blendv_pd,
8954 _mm_blend_ps, _mm_blendv_ps): New.
8956 2021-07-30 Roger Sayle <roger@nextmovesoftware.com>
8957 Uroš Bizjak <ubizjak@gmail.com>
8959 * config/i386/i386.md (*dec_cmov<mode>): New define_insn_and_split
8960 to generate a conditional move using the carry flag after sub $1.
8961 (peephole2): Eliminate a register-to-register move by inverting
8962 the condition of a conditional move.
8964 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
8966 * config/mmix/mmix.md ("call", "call_value", "*call_real")
8967 ("*call_value_real"): Don't generate rtx mentioning the generic
8968 operands 1 and 2 to "call", and similarly for "call_value".
8969 * config/mmix/mmix.c (mmix_print_operand_punct_valid_p)
8970 (mmix_print_operand): Use '!' instead of 'p'.
8972 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
8974 * doc/md.texi (call): Correct information about operand 2.
8975 * config/mmix/mmix.md ("call", "call_value"): Remove fixed FIXMEs.
8977 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
8979 * range-op.cc (operator_trunc_mod::wi_fold): Fold constants.
8981 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
8983 * range-op.cc (operator_div::wi_fold): Return UNDEFINED for [0, 0] divisor.
8985 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
8987 * gimple-range-cache.cc (*::set_bb_range): Change const basic_block to
8989 (*::get_bb_range): Ditto.
8990 (*::bb_range_p): Ditto.
8991 * gimple-range-cache.h: Change prototypes.
8993 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
8996 * builtins.c (builtin_memcpy_read_str): Change the mode argument
8997 from scalar_int_mode to fixed_size_mode.
8998 (builtin_strncpy_read_str): Likewise.
8999 (gen_memset_value_from_prev): New function.
9000 (builtin_memset_read_str): Change the mode argument from
9001 scalar_int_mode to fixed_size_mode. Use gen_memset_value_from_prev
9002 and support CONST_VECTOR.
9003 (builtin_memset_gen_str): Likewise.
9004 (try_store_by_multiple_pieces): Use by_pieces_constfn to declare
9006 * builtins.h (builtin_strncpy_read_str): Replace scalar_int_mode
9007 with fixed_size_mode.
9008 (builtin_memset_read_str): Likewise.
9009 * expr.c (widest_int_mode_for_size): Renamed to ...
9010 (widest_fixed_size_mode_for_size): Add a bool argument to
9011 indicate if QI vector mode can be used.
9012 (by_pieces_ninsns): Call widest_fixed_size_mode_for_size
9013 instead of widest_int_mode_for_size.
9014 (pieces_addr::adjust): Change the mode argument from
9015 scalar_int_mode to fixed_size_mode.
9016 (op_by_pieces_d): Make m_len read-only. Add a bool member,
9017 m_qi_vector_mode, to indicate that QI vector mode can be used.
9018 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
9019 initialize m_qi_vector_mode. Call widest_fixed_size_mode_for_size
9020 instead of widest_int_mode_for_size.
9021 (op_by_pieces_d::get_usable_mode): Change the mode argument from
9022 scalar_int_mode to fixed_size_mode. Call
9023 widest_fixed_size_mode_for_size instead of
9024 widest_int_mode_for_size.
9025 (op_by_pieces_d::smallest_fixed_size_mode_for_size): New member
9026 function to return the smallest integer or QI vector mode.
9027 (op_by_pieces_d::run): Call widest_fixed_size_mode_for_size
9028 instead of widest_int_mode_for_size. Call
9029 smallest_fixed_size_mode_for_size instead of
9030 smallest_int_mode_for_size.
9031 (store_by_pieces_d::store_by_pieces_d): Add a bool argument to
9032 indicate that QI vector mode can be used and pass it to
9033 op_by_pieces_d::op_by_pieces_d.
9034 (can_store_by_pieces): Call widest_fixed_size_mode_for_size
9035 instead of widest_int_mode_for_size. Pass memsetp to
9036 widest_fixed_size_mode_for_size to support QI vector mode.
9037 Allow all CONST_VECTORs for memset if vec_duplicate is supported.
9038 (store_by_pieces): Pass memsetp to
9039 store_by_pieces_d::store_by_pieces_d.
9040 (clear_by_pieces_1): Removed.
9041 (clear_by_pieces): Replace clear_by_pieces_1 with
9042 builtin_memset_read_str and pass true to store_by_pieces_d to
9043 support vector mode broadcast.
9044 (string_cst_read_str): Change the mode argument from
9045 scalar_int_mode to fixed_size_mode.
9046 * expr.h (by_pieces_constfn): Change scalar_int_mode to
9048 (by_pieces_prev): Likewise.
9049 * rtl.h (lowpart_subreg_regno): New.
9050 * rtlanal.c (lowpart_subreg_regno): New. A wrapper around
9051 simplify_subreg_regno.
9052 * target.def (gen_memset_scratch_rtx): New hook.
9053 * doc/tm.texi.in: Add TARGET_GEN_MEMSET_SCRATCH_RTX.
9054 * doc/tm.texi: Regenerated.
9056 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
9059 * config/mips/mips.c (mips_atomic_assign_expand_fenv): Use
9060 TARGET_EXPR instead of MODIFY_EXPR.
9062 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
9065 * config/mips/mips-protos.h (mips_expand_vec_cmp_expr): Declare.
9066 * config/mips/mips.c (mips_expand_vec_cmp_expr): New function.
9067 * config/mips/mips-msa.md (vec_cmp<MSA:mode><mode_i>): New
9069 (vec_cmpu<IMSA:mode><mode_i>): New expander.
9071 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
9074 * config/i386/i386-options.c (ix86_option_override_internal):
9075 Don't enable LZCNT/POPCNT if they have been disabled explicitly.
9077 2021-07-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
9080 * config/arm/arm_neon.h (vld1_p64): Replace call to builtin by
9081 explicitly dereferencing __a.
9082 (vld1_s64): Likewise.
9083 (vld1_u64): Likewise.
9084 * config/arm/arm_neon_builtins.def (vld1): Remove entry for di
9085 and change to VAR13.
9087 2021-07-30 Aldy Hernandez <aldyh@redhat.com>
9089 * gimple-loop-versioning.cc (lv_dom_walker::lv_dom_walker): Remove
9090 use of m_range_analyzer.
9091 (loop_versioning::lv_dom_walker::before_dom_children): Same.
9092 (loop_versioning::lv_dom_walker::after_dom_children): Remove.
9093 (loop_versioning::prune_loop_conditions): Replace vr_values use
9094 with range_query interface.
9095 (pass_loop_versioning::execute): Use ranger.
9097 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
9100 * ipa-devirt.c (ipa_odr_read_section): Compare the precision of
9101 enum values, and emit a warning if they mismatch.
9103 2021-07-30 Kewen Lin <linkw@linux.ibm.com>
9105 * cfgloop.h (as_const): New function.
9106 (class loop_iterator): Rename to ...
9107 (class loops_list): ... this.
9108 (loop_iterator::next): Rename to ...
9109 (loops_list::Iter::fill_curr_loop): ... this and adjust.
9110 (loop_iterator::loop_iterator): Rename to ...
9111 (loops_list::loops_list): ... this and adjust.
9112 (loops_list::Iter): New class.
9113 (loops_list::iterator): New type.
9114 (loops_list::const_iterator): New type.
9115 (loops_list::begin): New function.
9116 (loops_list::end): Likewise.
9117 (loops_list::begin const): Likewise.
9118 (loops_list::end const): Likewise.
9119 (FOR_EACH_LOOP): Remove.
9120 (FOR_EACH_LOOP_FN): Remove.
9121 * cfgloop.c (flow_loops_dump): Adjust FOR_EACH_LOOP* with range-based
9122 for loop with loops_list instance.
9123 (sort_sibling_loops): Likewise.
9124 (disambiguate_loops_with_multiple_latches): Likewise.
9125 (verify_loop_structure): Likewise.
9126 * cfgloopmanip.c (create_preheaders): Likewise.
9127 (force_single_succ_latches): Likewise.
9128 * config/aarch64/falkor-tag-collision-avoidance.c
9129 (execute_tag_collision_avoidance): Likewise.
9130 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Likewise.
9131 * config/s390/s390.c (s390_adjust_loops): Likewise.
9132 * doc/loop.texi: Likewise.
9133 * gimple-loop-interchange.cc (pass_linterchange::execute): Likewise.
9134 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
9135 * gimple-loop-versioning.cc (loop_versioning::analyze_blocks): Likewise.
9136 (loop_versioning::make_versioning_decisions): Likewise.
9137 * gimple-ssa-split-paths.c (split_paths): Likewise.
9138 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
9139 * graphite.c (canonicalize_loop_form): Likewise.
9140 (graphite_transform_loops): Likewise.
9141 * ipa-fnsummary.c (analyze_function_body): Likewise.
9142 * ipa-pure-const.c (analyze_function): Likewise.
9143 * loop-doloop.c (doloop_optimize_loops): Likewise.
9144 * loop-init.c (loop_optimizer_finalize): Likewise.
9145 (fix_loop_structure): Likewise.
9146 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
9147 (move_loop_invariants): Likewise.
9148 * loop-unroll.c (decide_unrolling): Likewise.
9149 (unroll_loops): Likewise.
9150 * modulo-sched.c (sms_schedule): Likewise.
9151 * predict.c (predict_loops): Likewise.
9152 (pass_profile::execute): Likewise.
9153 * profile.c (branch_prob): Likewise.
9154 * sel-sched-ir.c (sel_finish_pipelining): Likewise.
9155 (sel_find_rgns): Likewise.
9156 * tree-cfg.c (replace_loop_annotate): Likewise.
9157 (replace_uses_by): Likewise.
9158 (move_sese_region_to_fn): Likewise.
9159 * tree-if-conv.c (pass_if_conversion::execute): Likewise.
9160 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
9161 * tree-parloops.c (parallelize_loops): Likewise.
9162 * tree-predcom.c (tree_predictive_commoning): Likewise.
9163 * tree-scalar-evolution.c (scev_initialize): Likewise.
9164 (scev_reset): Likewise.
9165 * tree-ssa-dce.c (find_obviously_necessary_stmts): Likewise.
9166 * tree-ssa-live.c (remove_unused_locals): Likewise.
9167 * tree-ssa-loop-ch.c (ch_base::copy_headers): Likewise.
9168 * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
9169 (tree_ssa_lim_initialize): Likewise.
9170 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Likewise.
9171 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize): Likewise.
9172 * tree-ssa-loop-manip.c (get_loops_exits): Likewise.
9173 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations): Likewise.
9174 (free_numbers_of_iterations_estimates): Likewise.
9175 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Likewise.
9176 * tree-ssa-loop-split.c (tree_ssa_split_loops): Likewise.
9177 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Likewise.
9178 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
9179 (pass_scev_cprop::execute): Likewise.
9180 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
9181 * tree-ssa-sccvn.c (do_rpo_vn): Likewise.
9182 * tree-ssa-threadupdate.c
9183 (jump_thread_path_registry::thread_through_all_blocks): Likewise.
9184 * tree-vectorizer.c (vectorize_loops): Likewise.
9185 * tree-vrp.c (vrp_asserts::find_assert_locations): Likewise.
9187 2021-07-29 Hans-Peter Nilsson <hp@bitrange.com>
9189 * config/mmix/mmix.c (mmix_function_arg_1): Avoid
9190 generating a VOIDmode register for e.g the
9191 function_arg_info::end_marker.
9193 2021-07-29 Jeff Law <jeffreyalaw@gmail.com>
9195 * config/h8300/h8300-modes.def: Add CCZ, CCV and CCC, drop CCZNV.
9196 * config/h8300/h8300.md (H8cc mode iterator): Add CCZ.
9197 (cc mode_attr): Similarly.
9198 (ccz subst_attr): Similarly.
9199 * config/h8300/jumpcall.md: Add new patterns for branch-on-bit.
9200 * config/h8300/testcompare.md: Remove various cc0 based patterns
9201 that had been commented out. Add pattern to set CCZ from a bit
9204 2021-07-29 Thomas Schwinge <thomas@codesourcery.com>
9205 Julian Brown <julian@codesourcery.com>
9206 Kwok Cheung Yeung <kcy@codesourcery.com>
9208 * omp-offload.c (oacc_loop_xform_head_tail, oacc_loop_process):
9209 'update_stmt' after modification.
9210 (pass_oacc_loop_designation): New function, extracted out of...
9211 (pass_oacc_device_lower): ... this.
9212 (pass_data_oacc_loop_designation, pass_oacc_loop_designation)
9213 (make_pass_oacc_loop_designation): New
9214 * passes.def: Add it.
9215 * tree-parloops.c (create_parallel_loop): Adjust.
9216 * tree-pass.h (make_pass_oacc_loop_designation): New.
9218 2021-07-29 Aldy Hernandez <aldyh@redhat.com>
9220 * flag-types.h (enum threader_mode): New.
9221 * params.opt: Add entry for --param=threader-mode.
9222 * tree-ssa-threadbackward.c (THREADER_ITERATIVE_MODE): New.
9223 (class back_threader): New.
9224 (back_threader::back_threader): New.
9225 (back_threader::~back_threader): New.
9226 (back_threader::maybe_register_path): New.
9227 (back_threader::find_taken_edge): New.
9228 (back_threader::find_taken_edge_switch): New.
9229 (back_threader::find_taken_edge_cond): New.
9230 (back_threader::resolve_def): New.
9231 (back_threader::resolve_phi): New.
9232 (back_threader::find_paths_to_names): New.
9233 (back_threader::find_paths): New.
9236 (thread_jumps::find_jump_threads_backwards): Call ranger threader.
9237 (thread_jumps::find_jump_threads_backwards_with_ranger): New.
9238 (pass_thread_jumps::execute): Abstract out code...
9239 (try_thread_blocks): ...here.
9240 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
9241 Abstract out threading candidate code to...
9242 (single_succ_to_potentially_threadable_block): ...here.
9243 * tree-ssa-threadedge.h (single_succ_to_potentially_threadable_block):
9245 * tree-ssa-threadupdate.c (register_jump_thread): Return boolean.
9246 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
9247 Return bool from register_jump_thread.
9249 2021-07-29 Andreas Krebbel <krebbel@linux.ibm.com>
9251 * target.def: in0 and in1 do not need to be registers.
9252 * doc/tm.texi: Regenerate.
9254 2021-07-29 liuhongt <hongtao.liu@intel.com>
9257 * config/i386/i386.c (ix86_widen_mult_cost): New function.
9258 (ix86_add_stmt_cost): Use ix86_widen_mult_cost for
9261 2021-07-29 Jiufu Guo <guojiufu@linux.ibm.com>
9264 * config/rs6000/rs6000.c (TARGET_PREFERRED_DOLOOP_MODE): New hook.
9265 (rs6000_preferred_doloop_mode): New hook.
9266 * doc/tm.texi: Regenerate.
9267 * doc/tm.texi.in: Add hook preferred_doloop_mode.
9268 * target.def (preferred_doloop_mode): New hook.
9269 * targhooks.c (default_preferred_doloop_mode): New hook.
9270 * targhooks.h (default_preferred_doloop_mode): New hook.
9271 * tree-ssa-loop-ivopts.c (compute_doloop_base_on_mode): New function.
9272 (add_iv_candidate_for_doloop): Call targetm.preferred_doloop_mode
9273 and compute_doloop_base_on_mode.
9275 2021-07-28 Martin Sebor <msebor@redhat.com>
9277 PR middle-end/101494
9278 * tree-ssa-uninit.c (maybe_warn_operand): Correct object offset
9279 and size computation.
9281 2021-07-28 Martin Sebor <msebor@redhat.com>
9283 PR middle-end/101601
9284 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Remove
9286 Handle pointers to functions.
9288 2021-07-28 Martin Sebor <msebor@redhat.com>
9290 * Makefile.in (OBJS): Add gimple-ssa-warn-access.o and pointer-query.o.
9291 * attribs.h (fndecl_dealloc_argno): Move fndecl_dealloc_argno to tree.h.
9292 * builtins.c (compute_objsize_r): Move to pointer-query.cc.
9293 (access_ref::access_ref): Same.
9294 (access_ref::phi): Same.
9295 (access_ref::get_ref): Same.
9296 (access_ref::size_remaining): Same.
9297 (access_ref::offset_in_range): Same.
9298 (access_ref::add_offset): Same.
9299 (access_ref::inform_access): Same.
9300 (ssa_name_limit_t::visit_phi): Same.
9301 (ssa_name_limit_t::leave_phi): Same.
9302 (ssa_name_limit_t::next): Same.
9303 (ssa_name_limit_t::next_phi): Same.
9304 (ssa_name_limit_t::~ssa_name_limit_t): Same.
9305 (pointer_query::pointer_query): Same.
9306 (pointer_query::get_ref): Same.
9307 (pointer_query::put_ref): Same.
9308 (pointer_query::flush_cache): Same.
9309 (warn_string_no_nul): Move to gimple-ssa-warn-access.cc.
9310 (check_nul_terminated_array): Same.
9311 (unterminated_array): Same.
9312 (maybe_warn_for_bound): Same.
9313 (check_read_access): Same.
9314 (warn_for_access): Same.
9315 (get_size_range): Same.
9316 (check_access): Same.
9317 (gimple_call_alloc_size): Move to tree.c.
9318 (gimple_parm_array_size): Move to pointer-query.cc.
9319 (get_offset_range): Same.
9320 (gimple_call_return_array): Same.
9321 (handle_min_max_size): Same.
9322 (handle_array_ref): Same.
9323 (handle_mem_ref): Same.
9324 (compute_objsize): Same.
9325 (gimple_call_alloc_p): Move to gimple-ssa-warn-access.cc.
9326 (call_dealloc_argno): Same.
9327 (fndecl_dealloc_argno): Same.
9328 (new_delete_mismatch_p): Same.
9329 (matching_alloc_calls_p): Same.
9330 (warn_dealloc_offset): Same.
9331 (maybe_emit_free_warning): Same.
9332 * builtins.h (check_nul_terminated_array): Move to
9333 gimple-ssa-warn-access.h.
9334 (check_nul_terminated_array): Same.
9335 (warn_string_no_nul): Same.
9336 (unterminated_array): Same.
9337 (class ssa_name_limit_t): Same.
9338 (class pointer_query): Same.
9339 (struct access_ref): Same.
9340 (class range_query): Same.
9341 (struct access_data): Same.
9342 (gimple_call_alloc_size): Same.
9343 (gimple_parm_array_size): Same.
9344 (compute_objsize): Same.
9345 (class access_data): Same.
9346 (maybe_emit_free_warning): Same.
9347 * calls.c (initialize_argument_information): Remove call to
9348 maybe_emit_free_warning.
9349 * gimple-array-bounds.cc: Include new header..
9350 * gimple-fold.c: Same.
9351 * gimple-ssa-sprintf.c: Same.
9352 * gimple-ssa-warn-restrict.c: Same.
9353 * passes.def: Add pass_warn_access.
9354 * tree-pass.h (make_pass_warn_access): Declare.
9355 * tree-ssa-strlen.c: Include new headers.
9356 * tree.c (fndecl_dealloc_argno): Move here from builtins.c.
9357 * tree.h (fndecl_dealloc_argno): Move here from attribs.h.
9358 * gimple-ssa-warn-access.cc: New file.
9359 * gimple-ssa-warn-access.h: New file.
9360 * pointer-query.cc: New file.
9361 * pointer-query.h: New file.
9363 2021-07-28 Jakub Jelinek <jakub@redhat.com>
9365 PR middle-end/101624
9366 * ubsan.c (maybe_instrument_pointer_overflow,
9367 instrument_object_size): Only test DECL_REGISTER on VAR_DECLs,
9368 PARM_DECLs or RESULT_DECLs.
9369 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
9371 2021-07-28 Jakub Jelinek <jakub@redhat.com>
9373 PR middle-end/101642
9374 * match.pd (bswap16 (x) == bswap16 (y)): Cast both operands
9375 to type of bswap16 for comparison.
9376 (bswap16 (x) == cst): Cast bswap16 operand to type of cst.
9378 2021-07-28 Richard Biener <rguenther@suse.de>
9380 PR tree-optimization/101615
9381 * tree-vect-slp.c (vect_optimize_slp): Materialize permutes
9382 at CTOR SLP graph entries.
9384 2021-07-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9386 * config/aarch64/aarch64.md (*extend<SHORT:mode><GPI:mode>2_aarch64):
9387 Add "r,w" alternative.
9389 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
9392 * config/i386/i386.c (ix86_avx_u128_mode_needed): Don't set
9393 AVX_U128_DIRTY when all bits are zero.
9395 2021-07-28 Richard Biener <rguenther@suse.de>
9397 PR tree-optimization/101615
9398 * tree-vect-slp.c (vect_optimize_slp): Pre-existing vector
9399 external nodes cannot be permuted so make them perm_out 0.
9401 2021-07-28 Andrew Stubbs <ams@codesourcery.com>
9404 * config.in: Regenerate.
9405 * config/gcn/gcn-hsa.h (A_FIJI): New define.
9406 (A_900): New define.
9407 (A_906): New define.
9408 (A_908): New define.
9409 (ASM_SPEC): Use A_FIJI, A_900, A_906 and A_908.
9410 * config/gcn/gcn.c (output_file_start): Adjust attributes according
9411 to the assembler capabilities.
9412 * config/gcn/mkoffload.c (main): Likewise.
9413 * configure: Regenerate.
9414 * configure.ac: Add tests for LLVM assembler attribute features.
9416 2021-07-28 Andrew MacLeod <amacleod@redhat.com>
9418 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Check for
9419 cond_false and cond_true on branches.
9421 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
9423 * config/aarch64/aarch64.c (aarch64_gen_adjusted_ldpstp): use
9426 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
9428 * alias.c (init_alias_analysis): Don't skip prologue/epilogue.
9430 2021-07-28 Jakub Jelinek <jakub@redhat.com>
9433 * config/i386/sse.md (vashr<mode>3): Split into vashrv8di3 expander
9434 and vashrv4di3 expander, where the latter requires just TARGET_AVX2
9435 and has special !TARGET_AVX512VL expansion.
9436 (vashrv2di3<mask_name>): Rename to ...
9437 (vashrv2di3): ... this. Change condition to TARGET_XOP || TARGET_AVX2
9438 and add special !TARGET_XOP && !TARGET_AVX512VL expansion.
9440 2021-07-28 Martin Uecker <muecker@gwdg.de>
9442 * calls.c (maybe_warn_rdwr_sizes): Correct argument
9443 numbers in warning that were switched.
9445 2021-07-28 Kewen Lin <linkw@linux.ibm.com>
9447 PR tree-optimization/101596
9448 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Fix wrong check
9449 by using new_type's precision instead.
9451 2021-07-28 liuhongt <hongtao.liu@intel.com>
9454 * config/i386/i386.h (processor_costs): Add new member
9456 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
9457 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
9458 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
9459 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
9460 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
9461 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
9462 generic_cost, core_cost): Initialize integer_to_sse same value
9464 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
9465 * config/i386/i386.c (ix86_builtin_vectorization_cost):
9466 Use integer_to_sse instead of sse_op to calculate the cost of
9469 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
9471 * config/rs6000/rs6000-gen-builtins.c (write_ovld_static_init): New
9473 (write_init_file): Call write_ovld_static_init.
9475 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
9477 * config/rs6000/rs6000-gen-builtins.c (write_bif_static_init): New
9479 (write_init_file): Call write_bif_static_init.
9481 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
9483 * config/rs6000/rs6000-gen-builtins.c (typemap): New struct.
9484 (TYPE_MAP_SIZE): New macro.
9485 (type_map): New initialized variable.
9486 (typemap_cmp): New function.
9487 (write_type_node): Likewise.
9488 (write_fntype_init): Implement.
9490 2021-07-27 Martin Sebor <msebor@redhat.com>
9492 PR tree-optimization/101584
9493 * tree-ssa-uninit.c (builtin_call_nomodifying_p): New function.
9494 (check_defs): Call it.
9496 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
9498 * tree-ssa-dom.c (dom_jump_threader_simplifier):
9499 Put avail_exprs_stack in the class, instead of passing it to
9500 jump_threader_simplifier.
9501 (dom_jump_threader_simplifier::simplify): Add state argument.
9502 (dom_opt_dom_walker): Add state.
9503 (pass_dominator::execute): Pass state to threader.
9504 (dom_opt_dom_walker::before_dom_children): Use state.
9505 * tree-ssa-threadedge.c (jump_threader::jump_threader): Replace
9507 (jump_threader::record_temporary_equivalences_from_phis):
9508 Register equivalences through the state variable.
9509 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
9510 Record ranges in a statement through the state variable.
9511 (jump_threader::simplify_control_stmt_condition): Pass state to
9513 (jump_threader::simplify_control_stmt_condition_1): Same.
9514 (jump_threader::thread_around_empty_blocks): Remove obsolete
9516 (jump_threader::thread_through_normal_block): Record equivalences
9517 on edge through the state variable.
9518 (jump_threader::thread_across_edge): Abstract state pushing.
9519 (jt_state::jt_state): New.
9520 (jt_state::push): New.
9521 (jt_state::pop): New.
9522 (jt_state::register_equiv): New.
9523 (jt_state::record_ranges_from_stmt): New.
9524 (jt_state::register_equivs_on_edge): New.
9525 (jump_threader_simplifier::jump_threader_simplifier): Move from
9527 (jump_threader_simplifier::simplify): Add state argument.
9528 * tree-ssa-threadedge.h (class jt_state): New.
9529 (class jump_threader): Add state to constructor.
9530 (class jump_threader_simplifier): Add state to simplify. Remove
9531 avail_exprs_stack from class.
9532 * tree-vrp.c (vrp_jump_threader_simplifier::simplify): Add state
9534 (vrp_jump_threader::vrp_jump_threader): Add state.
9535 (vrp_jump_threader::~vrp_jump_threader): Cleanup state.
9537 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
9539 * Makefile.in (OBJS): Add gimple-range-path.o.
9540 * gimple-range-path.cc: New file.
9541 * gimple-range-path.h: New file.
9543 2021-07-27 Jonathan Wright <jonathan.wright@arm.com>
9545 * config/aarch64/aarch64-simd.md: Push sign/zero-extension
9546 inside vec_duplicate for all patterns.
9547 * simplify-rtx.c (simplify_context::simplify_unary_operation_1):
9548 Push sign/zero-extension inside vec_duplicate.
9550 2021-07-27 Richard Biener <rguenther@suse.de>
9552 PR tree-optimization/101573
9553 * tree-ssa-uninit.c (warn_uninit_phi_uses): New function
9554 looking at uninitialized PHI arg defs in some constrained cases.
9555 (warn_uninitialized_vars): Call it.
9556 (execute_early_warn_uninitialized): Calculate dominators.
9558 2021-07-27 Richard Biener <rguenther@suse.de>
9560 PR tree-optimization/39821
9561 * tree-vect-stmts.c (vect_model_promotion_demotion_cost): Use
9562 vector_stmt for widening arithmetic.
9563 (vectorizable_conversion): Adjust.
9565 2021-07-27 Martin Jambor <mjambor@suse.cz>
9567 * cgraph.h (ipa_replace_map): New field force_load_ref.
9568 * ipa-prop.h (ipa_param_descriptor): Reduce precision of move_cost,
9569 aded new flag load_dereferenced, adjusted comments.
9570 (ipa_get_param_dereferenced): New function.
9571 (ipa_set_param_dereferenced): Likewise.
9572 * cgraphclones.c (cgraph_node::create_virtual_clone): Follow it.
9573 * ipa-cp.c: Include gimple.h.
9574 (ipcp_discover_new_direct_edges): Take into account dereferenced flag.
9575 (get_replacement_map): New parameter force_load_ref, set the
9576 appropriate flag in ipa_replace_map if set.
9577 (struct symbol_and_index_together): New type.
9578 (adjust_refs_in_act_callers): New function.
9579 (adjust_references_in_caller): Likewise.
9580 (create_specialized_node): When appropriate, call
9581 adjust_references_in_caller and force only load references.
9582 * ipa-prop.c (load_from_dereferenced_name): New function.
9583 (ipa_analyze_controlled_uses): Also detect loads from a
9584 dereference, harden testing of call statements.
9585 (ipa_write_node_info): Stream the dereferenced flag.
9586 (ipa_read_node_info): Likewise.
9587 (ipa_set_jf_constant): Also create refdesc when jump function
9588 references a variable.
9589 (cgraph_node_for_jfunc): Rename to symtab_node_for_jfunc, work
9590 also on references of variables and return a symtab_node. Adjust
9592 (propagate_controlled_uses): Also remove references to VAR_DECLs.
9594 2021-07-27 Jakub Jelinek <jakub@redhat.com>
9596 PR middle-end/101586
9597 * gimple-fold.c (clear_padding_type): Ignore FIELD_DECLs with byte
9598 positions above or equal to sz except for diagnostics of flexible
9601 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
9603 PR tree-optimization/78888
9604 * gimple-range-fold.cc (get_letter_range): New.
9605 (fold_using_range::range_of_builtin_call): Call get_letter_range.
9607 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
9609 PR tree-optimization/78888
9610 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Add cases
9611 for CFN_BUILT_IN_TOUPPER and CFN_BUILT_IN_TOLOWER.
9613 2021-07-26 Roger Sayle <roger@nextmovesoftware.com>
9614 Marc Glisse <marc.glisse@inria.fr>
9616 * match.pd (rotate): Simplify equality/inequality of rotations.
9617 (bswap): Simplify equality/inequality tests of byte swapping.
9619 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
9621 * range-op.cc (operator_bitwise_xor::op1_op2_relation_effect):
9624 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
9626 * range-op.cc (operator_lshift::fold_range): Pass rel to
9627 base class fold_range.
9628 (operator_rshift::fold_range): Same.
9630 2021-07-26 Ashimida <ashimida@linux.alibaba.com>
9633 * toplev.h (min_align_loops_log): Remove declaration.
9634 (min_align_jumps_log, min_align_labels_log): Likewise.
9635 (min_align_functions_log): Likewise.
9637 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
9639 * tree-vrp.c (vrp_simplify_cond_using_ranges): Rename vr_values
9641 (execute_vrp): Abstract out simplification of conditionals...
9642 (simplify_casted_conds): ...here.
9644 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
9646 * gimple-array-bounds.cc (array_bounds_checker::get_value_range):
9647 Add gimple argument.
9648 (array_bounds_checker::check_array_ref): Same.
9649 (array_bounds_checker::check_addr_expr): Same.
9650 (array_bounds_checker::check_array_bounds): Pass statement to
9651 check_array_bounds and check_addr_expr.
9652 * gimple-array-bounds.h (check_array_bounds): Add gimple argument.
9653 (check_addr_expr): Same.
9654 (get_value_range): Same.
9656 2021-07-26 Tamar Christina <tamar.christina@arm.com>
9658 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): Rename to..
9659 (sdot_prod, udot_prod): ... This.
9660 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): Merged
9662 (<sur>dot_prod<vsi2qi>): ... this.
9663 (aarch64_<sur>dot_lane<vsi2qi>, aarch64_<sur>dot_laneq<vsi2qi>):
9664 Change operands order.
9665 (<sur>sadv16qi): Use new operands order.
9666 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32,
9667 vdotq_s32): Use new RTL ordering.
9669 2021-07-26 Tamar Christina <tamar.christina@arm.com>
9671 * config/aarch64/aarch64-builtins.c (TYPES_TERNOP_SUSS,
9672 aarch64_types_ternop_suss_qualifiers): New.
9673 * config/aarch64/aarch64-simd-builtins.def (usdot_prod): Use it.
9674 * config/aarch64/aarch64-simd.md (usdot_prod<vsi2qi>): Re-organize RTL.
9675 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Use it.
9677 2021-07-23 Jakub Jelinek <jakub@redhat.com>
9679 PR rtl-optimization/101562
9680 * expmed.c (store_integral_bit_field): Only use movstrict_optab
9681 if the operand isn't paradoxical.
9683 2021-07-23 Aldy Hernandez <aldyh@redhat.com>
9685 * gimple-array-bounds.h (class array_bounds_checker): Change
9686 ranges type to range_query.
9688 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9690 * config/aarch64/arm_neon.h (vst1_s64_x2): Use
9691 __builtin_memcpy instead of constructing
9692 __builtin_aarch64_simd_oi one vector at a time.
9693 (vst1_u64_x2): Likewise.
9694 (vst1_f64_x2): Likewise.
9695 (vst1_s8_x2): Likewise.
9696 (vst1_p8_x2): Likewise.
9697 (vst1_s16_x2): Likewise.
9698 (vst1_p16_x2): Likewise.
9699 (vst1_s32_x2): Likewise.
9700 (vst1_u8_x2): Likewise.
9701 (vst1_u16_x2): Likewise.
9702 (vst1_u32_x2): Likewise.
9703 (vst1_f16_x2): Likewise.
9704 (vst1_f32_x2): Likewise.
9705 (vst1_p64_x2): Likewise.
9706 (vst1q_s8_x2): Likewise.
9707 (vst1q_p8_x2): Likewise.
9708 (vst1q_s16_x2): Likewise.
9709 (vst1q_p16_x2): Likewise.
9710 (vst1q_s32_x2): Likewise.
9711 (vst1q_s64_x2): Likewise.
9712 (vst1q_u8_x2): Likewise.
9713 (vst1q_u16_x2): Likewise.
9714 (vst1q_u32_x2): Likewise.
9715 (vst1q_u64_x2): Likewise.
9716 (vst1q_f16_x2): Likewise.
9717 (vst1q_f32_x2): Likewise.
9718 (vst1q_f64_x2): Likewise.
9719 (vst1q_p64_x2): Likewise.
9721 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9723 * config/aarch64/arm_neon.h (vst1_s64_x3): Use
9724 __builtin_memcpy instead of constructing
9725 __builtin_aarch64_simd_ci one vector at a time.
9726 (vst1_u64_x3): Likewise.
9727 (vst1_f64_x3): Likewise.
9728 (vst1_s8_x3): Likewise.
9729 (vst1_p8_x3): Likewise.
9730 (vst1_s16_x3): Likewise.
9731 (vst1_p16_x3): Likewise.
9732 (vst1_s32_x3): Likewise.
9733 (vst1_u8_x3): Likewise.
9734 (vst1_u16_x3): Likewise.
9735 (vst1_u32_x3): Likewise.
9736 (vst1_f16_x3): Likewise.
9737 (vst1_f32_x3): Likewise.
9738 (vst1_p64_x3): Likewise.
9739 (vst1q_s8_x3): Likewise.
9740 (vst1q_p8_x3): Likewise.
9741 (vst1q_s16_x3): Likewise.
9742 (vst1q_p16_x3): Likewise.
9743 (vst1q_s32_x3): Likewise.
9744 (vst1q_s64_x3): Likewise.
9745 (vst1q_u8_x3): Likewise.
9746 (vst1q_u16_x3): Likewise.
9747 (vst1q_u32_x3): Likewise.
9748 (vst1q_u64_x3): Likewise.
9749 (vst1q_f16_x3): Likewise.
9750 (vst1q_f32_x3): Likewise.
9751 (vst1q_f64_x3): Likewise.
9752 (vst1q_p64_x3): Likewise.
9754 2021-07-23 H.J. Lu <hjl.tools@gmail.com>
9757 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): Don't return
9758 hard register when LRA is in progress.
9760 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9762 * config/aarch64/arm_neon.h (vst1_s8_x4): Use
9763 __builtin_memcpy instead of using a union.
9764 (vst1q_s8_x4): Likewise.
9765 (vst1_s16_x4): Likewise.
9766 (vst1q_s16_x4): Likewise.
9767 (vst1_s32_x4): Likewise.
9768 (vst1q_s32_x4): Likewise.
9769 (vst1_u8_x4): Likewise.
9770 (vst1q_u8_x4): Likewise.
9771 (vst1_u16_x4): Likewise.
9772 (vst1q_u16_x4): Likewise.
9773 (vst1_u32_x4): Likewise.
9774 (vst1q_u32_x4): Likewise.
9775 (vst1_f16_x4): Likewise.
9776 (vst1q_f16_x4): Likewise.
9777 (vst1_f32_x4): Likewise.
9778 (vst1q_f32_x4): Likewise.
9779 (vst1_p8_x4): Likewise.
9780 (vst1q_p8_x4): Likewise.
9781 (vst1_p16_x4): Likewise.
9782 (vst1q_p16_x4): Likewise.
9783 (vst1_s64_x4): Likewise.
9784 (vst1_u64_x4): Likewise.
9785 (vst1_p64_x4): Likewise.
9786 (vst1q_s64_x4): Likewise.
9787 (vst1q_u64_x4): Likewise.
9788 (vst1q_p64_x4): Likewise.
9789 (vst1_f64_x4): Likewise.
9790 (vst1q_f64_x4): Likewise.
9792 2021-07-23 Jonathan Wrightt <jonathan.wright@arm.com>
9794 * config/aarch64/arm_neon.h (vst2_s64): Use __builtin_memcpy
9795 instead of constructing __builtin_aarch64_simd_oi one vector
9797 (vst2_u64): Likewise.
9798 (vst2_f64): Likewise.
9799 (vst2_s8): Likewise.
9800 (vst2_p8): Likewise.
9801 (vst2_s16): Likewise.
9802 (vst2_p16): Likewise.
9803 (vst2_s32): Likewise.
9804 (vst2_u8): Likewise.
9805 (vst2_u16): Likewise.
9806 (vst2_u32): Likewise.
9807 (vst2_f16): Likewise.
9808 (vst2_f32): Likewise.
9809 (vst2_p64): Likewise.
9810 (vst2q_s8): Likewise.
9811 (vst2q_p8): Likewise.
9812 (vst2q_s16): Likewise.
9813 (vst2q_p16): Likewise.
9814 (vst2q_s32): Likewise.
9815 (vst2q_s64): Likewise.
9816 (vst2q_u8): Likewise.
9817 (vst2q_u16): Likewise.
9818 (vst2q_u32): Likewise.
9819 (vst2q_u64): Likewise.
9820 (vst2q_f16): Likewise.
9821 (vst2q_f32): Likewise.
9822 (vst2q_f64): Likewise.
9823 (vst2q_p64): Likewise.
9825 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9827 * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy
9828 instead of constructing __builtin_aarch64_simd_ci one vector
9830 (vst3_u64): Likewise.
9831 (vst3_f64): Likewise.
9832 (vst3_s8): Likewise.
9833 (vst3_p8): Likewise.
9834 (vst3_s16): Likewise.
9835 (vst3_p16): Likewise.
9836 (vst3_s32): Likewise.
9837 (vst3_u8): Likewise.
9838 (vst3_u16): Likewise.
9839 (vst3_u32): Likewise.
9840 (vst3_f16): Likewise.
9841 (vst3_f32): Likewise.
9842 (vst3_p64): Likewise.
9843 (vst3q_s8): Likewise.
9844 (vst3q_p8): Likewise.
9845 (vst3q_s16): Likewise.
9846 (vst3q_p16): Likewise.
9847 (vst3q_s32): Likewise.
9848 (vst3q_s64): Likewise.
9849 (vst3q_u8): Likewise.
9850 (vst3q_u16): Likewise.
9851 (vst3q_u32): Likewise.
9852 (vst3q_u64): Likewise.
9853 (vst3q_f16): Likewise.
9854 (vst3q_f32): Likewise.
9855 (vst3q_f64): Likewise.
9856 (vst3q_p64): Likewise.
9858 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9860 * config/aarch64/arm_neon.h (vst4_s64): Use __builtin_memcpy
9861 instead of constructing __builtin_aarch64_simd_xi one vector
9863 (vst4_u64): Likewise.
9864 (vst4_f64): Likewise.
9865 (vst4_s8): Likewise.
9866 (vst4_p8): Likewise.
9867 (vst4_s16): Likewise.
9868 (vst4_p16): Likewise.
9869 (vst4_s32): Likewise.
9870 (vst4_u8): Likewise.
9871 (vst4_u16): Likewise.
9872 (vst4_u32): Likewise.
9873 (vst4_f16): Likewise.
9874 (vst4_f32): Likewise.
9875 (vst4_p64): Likewise.
9876 (vst4q_s8): Likewise.
9877 (vst4q_p8): Likewise.
9878 (vst4q_s16): Likewise.
9879 (vst4q_p16): Likewise.
9880 (vst4q_s32): Likewise.
9881 (vst4q_s64): Likewise.
9882 (vst4q_u8): Likewise.
9883 (vst4q_u16): Likewise.
9884 (vst4q_u32): Likewise.
9885 (vst4q_u64): Likewise.
9886 (vst4q_f16): Likewise.
9887 (vst4q_f32): Likewise.
9888 (vst4q_f64): Likewise.
9889 (vst4q_p64): Likewise.
9891 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9893 * config/aarch64/arm_neon.h (vtbx4_s8): Use __builtin_memcpy
9894 instead of constructing __builtin_aarch64_simd_oi one vector
9896 (vtbx4_u8): Likewise.
9897 (vtbx4_p8): Likewise.
9899 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9901 * config/aarch64/arm_neon.h (vtbl3_s8): Use __builtin_memcpy
9902 instead of constructing __builtin_aarch64_simd_oi one vector
9904 (vtbl3_u8): Likewise.
9905 (vtbl3_p8): Likewise.
9906 (vtbl4_s8): Likewise.
9907 (vtbl4_u8): Likewise.
9908 (vtbl4_p8): Likewise.
9910 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9912 * config/aarch64/arm_neon.h (vqtbx2_s8): Use __builtin_memcpy
9913 instead of constructing __builtin_aarch64_simd_oi one vector
9915 (vqtbx2_u8): Likewise.
9916 (vqtbx2_p8): Likewise.
9917 (vqtbx2q_s8): Likewise.
9918 (vqtbx2q_u8): Likewise.
9919 (vqtbx2q_p8): Likewise.
9920 (vqtbx3_s8): Use __builtin_memcpy instead of constructing
9921 __builtin_aarch64_simd_ci one vector at a time.
9922 (vqtbx3_u8): Likewise.
9923 (vqtbx3_p8): Likewise.
9924 (vqtbx3q_s8): Likewise.
9925 (vqtbx3q_u8): Likewise.
9926 (vqtbx3q_p8): Likewise.
9927 (vqtbx4_s8): Use __builtin_memcpy instead of constructing
9928 __builtin_aarch64_simd_xi one vector at a time.
9929 (vqtbx4_u8): Likewise.
9930 (vqtbx4_p8): Likewise.
9931 (vqtbx4q_s8): Likewise.
9932 (vqtbx4q_u8): Likewise.
9933 (vqtbx4q_p8): Likewise.
9935 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9937 * config/aarch64/arm_neon.h (vqtbl2_s8): Use __builtin_memcpy
9938 instead of constructing __builtin_aarch64_simd_oi one vector
9940 (vqtbl2_u8): Likewise.
9941 (vqtbl2_p8): Likewise.
9942 (vqtbl2q_s8): Likewise.
9943 (vqtbl2q_u8): Likewise.
9944 (vqtbl2q_p8): Likewise.
9945 (vqtbl3_s8): Use __builtin_memcpy instead of constructing
9946 __builtin_aarch64_simd_ci one vector at a time.
9947 (vqtbl3_u8): Likewise.
9948 (vqtbl3_p8): Likewise.
9949 (vqtbl3q_s8): Likewise.
9950 (vqtbl3q_u8): Likewise.
9951 (vqtbl3q_p8): Likewise.
9952 (vqtbl4_s8): Use __builtin_memcpy instead of constructing
9953 __builtin_aarch64_simd_xi one vector at a time.
9954 (vqtbl4_u8): Likewise.
9955 (vqtbl4_p8): Likewise.
9956 (vqtbl4q_s8): Likewise.
9957 (vqtbl4q_u8): Likewise.
9958 (vqtbl4q_p8): Likewise.
9960 2021-07-23 Haochen Gui <guihaoc@gcc.gnu.org>
9963 * config/rs6000/rs6000.md (cstore<mode>4): Fix wrong fall through.
9965 2021-07-22 Andrew Pinski <apinski@marvell.com>
9967 PR tree-optimization/10153
9968 * tree-tailcall.c (create_tailcall_accumulator):
9969 Don't call fold_convert as the type should be correct already.
9970 (tree_optimize_tail_calls_1): Use build_{one,zero}_cst instead
9971 of integer_{one,zero}_node for the call of create_tailcall_accumulator.
9973 2021-07-22 Aldy Hernandez <aldyh@redhat.com>
9975 * gimple-range-cache.cc (non_null_ref::adjust_range): Replace
9976 varying_p check for null/non-null check.
9978 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
9980 PR tree-optimization/101511
9981 * value-relation.cc (relation_oracle::query_relation): Check if ssa1
9982 is in ssa2's equiv set, and don't trap if so.
9984 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
9986 PR tree-optimization/101497
9987 * gimple-range-fold.cc (fold_using_range::range_of_cond_expr): Check
9990 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
9992 PR tree-optimization/101496
9993 * vr-values.c (simplify_using_ranges::fold_cond): Call range_of_stmt
9994 first, then vrp_visit_cond_Stmt.
9996 2021-07-22 liuhongt <hongtao.liu@intel.com>
9998 * config/i386/i386-expand.c
9999 (ix86_broadcast_from_integer_constant): Rename to ..
10000 (ix86_broadcast_from_constant): .. this, and extend it to
10002 (ix86_expand_vector_move): Extend to float mode.
10003 * config/i386/i386-features.c
10004 (replace_constant_pool_with_broadcast): Remove.
10005 (remove_partial_avx_dependency_gate): Ditto.
10006 (constant_pool_broadcast): Ditto.
10007 (class pass_constant_pool_broadcast): Ditto.
10008 (make_pass_constant_pool_broadcast): Ditto.
10009 (remove_partial_avx_dependency): Adjust gate.
10010 * config/i386/i386-passes.def: Remove pass_constant_pool_broadcast.
10011 * config/i386/i386-protos.h
10012 (make_pass_constant_pool_broadcast): Remove.
10014 2021-07-22 liuhongt <hongtao.liu@intel.com>
10016 * config/i386/constraints.md (Wb): New constraint.
10018 * config/i386/i386.md (*ashlhi3_1): Extend to avx512 mask
10020 (*ashlqi3_1): Ditto.
10021 (*<insn><mode>3_1): Split to ..
10022 (*ashr<mode>3_1): this, ...
10023 (*lshr<mode>3_1): and this, also extend this pattern to avx512
10025 (*<insn><mode>3_1): Split to ..
10026 (*ashr<mode>3_1): this, ...
10027 (*lshrqi3_1): and this, also extend this pattern to avx512
10029 (*lshrhi3_1): And this, also extend this pattern to avx512
10031 * config/i386/sse.md (k<code><mode>): New define_split after
10032 it to convert generic shift pattern to mask shift ones.
10034 2021-07-21 Thomas Schwinge <thomas@codesourcery.com>
10035 Joseph Myers <joseph@codesourcery.com>
10036 Cesar Philippidis <cesar@codesourcery.com>
10038 * tree-core.h (omp_clause_code): Add 'OMP_CLAUSE_NOHOST'.
10039 * tree.c (omp_clause_num_ops, omp_clause_code_name, walk_tree_1):
10041 * tree-pretty-print.c (dump_omp_clause): Likewise.
10042 * omp-general.c (oacc_verify_routine_clauses): Likewise.
10043 * gimplify.c (gimplify_scan_omp_clauses)
10044 (gimplify_adjust_omp_clauses): Likewise.
10045 * tree-nested.c (convert_nonlocal_omp_clauses)
10046 (convert_local_omp_clauses): Likewise.
10047 * omp-low.c (scan_sharing_clauses): Likewise.
10048 * omp-offload.c (execute_oacc_device_lower): Update.
10050 2021-07-21 Martin Sebor <msebor@redhat.com>
10052 * tree-ssa-alias.c (walk_aliased_vdefs_1): Fix typos in a comment.
10054 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
10056 * config/rs6000/rs6000-gen-builtins.c (write_init_bif_table):
10059 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
10061 * config/rs6000/rs6000-gen-builtins.c (write_fntype): New
10063 (write_fntype_init): New stub function.
10064 (write_init_bif_table): Likewise.
10065 (write_init_ovld_table): New function.
10066 (write_init_file): Implement.
10068 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
10070 * config/rs6000/rs6000-gen-builtins.c
10071 (write_autogenerated_header): New function.
10072 (write_decls): Likewise.
10073 (write_extern_fntype): New callback function.
10074 (write_header_file): Implement.
10076 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
10078 * config/rs6000/rs6000-gen-builtins.c (write_defines_file):
10081 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
10083 * config/rs6000/rs6000-gen-builtins.c (complete_vector_type): New
10085 (complete_base_type): Likewise.
10086 (construct_fntype_id): Likewise.
10087 (parse_bif_entry): Call contruct_fntype_id.
10088 (parse_ovld_entry): Likewise.
10090 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
10092 * config/rs6000/rs6000-gen-builtins.c (ovld_stanza): New struct.
10093 (MAXOVLDSTANZAS): New macro.
10094 (ovld_stanzas): New variable.
10095 (curr_ovld_stanza): Likewise.
10096 (MAXOVLDS): New macro.
10097 (ovlddata): New struct.
10098 (ovlds): New variable.
10099 (curr_ovld): Likewise.
10100 (max_ovld_args): Likewise.
10101 (parse_ovld_entry): New function.
10102 (parse_ovld_stanza): Likewise.
10103 (parse_ovld): Implement.
10105 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
10107 * config/rs6000/rs6000-gen-builtins.c (parse_bif_attrs):
10110 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
10112 * config/rs6000/rs6000-gen-builtins.c (parse_args): New function.
10113 (parse_prototype): Implement.
10115 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
10117 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): New enum.
10118 (curr_bif_stanza): New variable.
10119 (stanza_entry): New struct.
10120 (stanza_map): New initialized variable.
10121 (enable_string): Likewise.
10122 (fnkinds): New enum.
10123 (typelist): New struct.
10124 (attrinfo): Likewise.
10125 (MAXRESTROPNDS): New macro.
10126 (prototype): New struct.
10127 (MAXBIFS): New macro.
10128 (bifdata): New struct.
10129 (bifs): New variable.
10130 (curr_bif): Likewise.
10131 (bif_order): Likewise.
10132 (bif_index): Likewise.
10133 (fatal): New function.
10134 (stanza_name_to_stanza): Likewise.
10135 (parse_bif_attrs): New stub function.
10136 (parse_prototype): Likewise.
10137 (parse_bif_entry): New function.
10138 (parse_bif_stanza): Likewise.
10139 (parse_bif): Implement.
10140 (set_bif_order): New function.
10141 (create_bif_order): Implement.
10143 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
10145 * config/rs6000/rs6000-gen-builtins.c (rbtree.h): New #include.
10146 (num_bifs): New variable.
10147 (num_ovld_stanzas): Likewise.
10148 (num_ovlds): Likewise.
10149 (parse_codes): New enum.
10150 (bif_rbt): New variable.
10151 (ovld_rbt): Likewise.
10152 (fntype_rbt): Likewise.
10153 (bifo_rbt): Likewise.
10154 (parse_bif): New stub function.
10155 (create_bif_order): Likewise.
10156 (parse_ovld): Likewise.
10157 (write_header_file): Likewise.
10158 (write_init_file): Likewise.
10159 (write_defines_file): Likewise.
10160 (delete_output_files): New function.
10163 2021-07-21 H.J. Lu <hjl.tools@gmail.com>
10166 * config/i386/i386-builtin.def: Remove OPTION_MASK_ISA_SSE4_2
10167 from CRC32 _builtin functions.
10169 2021-07-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
10171 * coverage.c (build_gcov_info_var_registration): Mark the object placed
10172 in the linker set as referenced so that it does not get optimized away.
10174 2021-07-21 Kito Cheng <kito.cheng@sifive.com>
10177 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
10179 * config.gcc (riscv*-*-*): Detect which python is available.
10181 2021-07-21 Jakub Jelinek <jakub@redhat.com>
10183 PR middle-end/101535
10184 * gimplify.c (omp_check_private): Properly skip ORT_TARGET_DATA
10185 contexts in which decl isn't privatized and for ORT_TARGET return
10186 false if decl is mapped.
10188 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
10190 * gimple-loop-jam.c: Include tree-ssa-sccvn.h.
10191 (tree_loop_unroll_and_jam): Run value-numbering on a loop that
10192 has been successfully unrolled.
10194 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
10196 * tree-ssa-loop-manip.c (determine_exit_conditions): Return a null
10197 exit condition if no tail loop is needed, and if the original exit
10198 condition should therefore be kept as-is.
10199 (tree_transform_and_unroll_loop): Handle that case here too.
10201 2021-07-21 Kewen Lin <linkw@linux.ibm.com>
10203 * tree-data-ref.c (free_dependence_relations): Adjust to pass vec
10205 (free_data_refs): Likewise.
10206 * tree-data-ref.h (free_dependence_relations): Likewise.
10207 (free_data_refs): Likewise.
10208 * tree-predcom.c (struct chain): Use auto_vec instead of vec for
10210 (struct component): Likewise.
10211 (pcom_worker::pcom_worker): Adjust for auto_vec and renaming changes.
10212 (pcom_worker::~pcom_worker): Likewise.
10213 (pcom_worker::release_chain): Adjust as auto_vec changes.
10214 (pcom_worker::loop): Rename to ...
10215 (pcom_worker::m_loop): ... this.
10216 (pcom_worker::datarefs): Rename to ...
10217 (pcom_worker::m_datarefs): ... this. Use auto_vec instead of vec.
10218 (pcom_worker::dependences): Rename to ...
10219 (pcom_worker::m_dependences): ... this. Use auto_vec instead of vec.
10220 (pcom_worker::chains): Rename to ...
10221 (pcom_worker::m_chains): ... this. Use auto_vec instead of vec.
10222 (pcom_worker::looparound_phis): Rename to ...
10223 (pcom_worker::m_looparound_phis): ... this. Use auto_vec instead of
10225 (pcom_worker::cache): Rename to ...
10226 (pcom_worker::m_cache): ... this. Use auto_vec instead of vec.
10227 (pcom_worker::release_chain): Adjust for auto_vec changes.
10228 (pcom_worker::release_chains): Adjust for auto_vec and renaming
10230 (release_component): Remove.
10231 (release_components): Adjust for release_component removal.
10232 (component_of): Adjust to use vec.
10233 (merge_comps): Likewise.
10234 (pcom_worker::aff_combination_dr_offset): Adjust for renaming changes.
10235 (pcom_worker::determine_offset): Likewise.
10236 (class comp_ptrs): Remove.
10237 (pcom_worker::split_data_refs_to_components): Adjust for renaming
10238 changes, for comp_ptrs removal with auto_vec.
10239 (pcom_worker::suitable_component_p): Adjust for renaming changes.
10240 (pcom_worker::filter_suitable_components): Adjust for release_component
10242 (pcom_worker::valid_initializer_p): Adjust for renaming changes.
10243 (pcom_worker::find_looparound_phi): Likewise.
10244 (pcom_worker::add_looparound_copies): Likewise.
10245 (pcom_worker::determine_roots_comp): Likewise.
10246 (pcom_worker::single_nonlooparound_use): Likewise.
10247 (pcom_worker::execute_pred_commoning_chain): Likewise.
10248 (pcom_worker::execute_pred_commoning): Likewise.
10249 (pcom_worker::try_combine_chains): Likewise.
10250 (pcom_worker::prepare_initializers_chain): Likewise.
10251 (pcom_worker::prepare_initializers): Likewise.
10252 (pcom_worker::prepare_finalizers_chain): Likewise.
10253 (pcom_worker::prepare_finalizers): Likewise.
10254 (pcom_worker::tree_predictive_commoning_loop): Likewise.
10256 2021-07-20 Martin Sebor <msebor@redhat.com>
10258 PR middle-end/101397
10259 * builtins.c (gimple_call_return_array): Add argument. Correct
10260 offsets for memchr, mempcpy, stpcpy, and stpncpy.
10261 (compute_objsize_r): Adjust offset computation for argument returning
10264 2021-07-20 Martin Sebor <msebor@redhat.com>
10266 PR middle-end/101300
10267 * tree-ssa-uninit.c (check_defs): Handle UBSAN built-ins.
10269 2021-07-20 Jeff Law <jlaw@localhost.localdomain>
10271 * function.c (assign_parm_setup_block): Use adjust_address instead
10272 of change_address to preserve MEM_EXPR and friends.
10274 2021-07-20 Martin Sebor <msebor@redhat.com>
10276 * cfgloop.h (single_likely_exit): Adjust by-value argument to
10277 by-const-reference.
10278 * cfgloopanal.c (single_likely_exit): Same.
10279 * cgraph.h (struct cgraph_node): Same.
10280 * cgraphclones.c (cgraph_node::create_virtual_clone): Same.
10281 * genautomata.c (merge_states): Same.
10282 * genextract.c (VEC_char_to_string): Same.
10283 * genmatch.c (dt_node::gen_kids_1): Same.
10284 (walk_captures): Adjust by-value argument to by-reference.
10285 * gimple-ssa-store-merging.c (check_no_overlap): Adjust by-value argument
10286 to by-const-reference.
10287 * gimple.c (gimple_build_call_vec): Same.
10288 (gimple_build_call_internal_vec): Same.
10289 (gimple_build_switch): Same.
10290 (sort_case_labels): Same.
10291 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
10293 * gimple.h (gimple_build_call_vec): Adjust by-value argument to
10294 by-const-reference.
10295 (gimple_build_call_internal_vec): Same.
10296 (gimple_build_switch): Same.
10297 (sort_case_labels): Same.
10298 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
10300 * haifa-sched.c (calc_priorities): Adjust by-value argument to
10301 by-const-reference.
10302 (sched_init_luids): Same.
10303 (haifa_init_h_i_d): Same.
10304 * ipa-cp.c (ipa_get_indirect_edge_target_1): Same.
10305 (adjust_callers_for_value_intersection): Adjust by-value argument to
10307 (find_more_scalar_values_for_callers_subset): Adjust by-value argument to
10308 by-const-reference.
10309 (find_more_contexts_for_caller_subset): Same.
10310 (find_aggregate_values_for_callers_subset): Same.
10311 (copy_useful_known_contexts): Same.
10312 * ipa-fnsummary.c (remap_edge_summaries): Same.
10313 (remap_freqcounting_predicate): Same.
10314 * ipa-inline.c (add_new_edges_to_heap): Adjust by-value argument to
10316 * ipa-predicate.c (predicate::remap_after_inlining): Adjust by-value argument
10317 to by-const-reference.
10318 * ipa-predicate.h (predicate::remap_after_inlining): Same.
10319 * ipa-prop.c (ipa_find_agg_cst_for_param): Same.
10320 * ipa-prop.h (ipa_find_agg_cst_for_param): Same.
10321 * ira-build.c (ira_loop_tree_body_rev_postorder): Same.
10322 * read-rtl.c (add_overload_instance): Same.
10323 * rtl.h (native_decode_rtx): Same.
10324 (native_decode_vector_rtx): Same.
10325 * sched-int.h (sched_init_luids): Same.
10326 (haifa_init_h_i_d): Same.
10327 * simplify-rtx.c (native_decode_vector_rtx): Same.
10328 (native_decode_rtx): Same.
10329 * tree-call-cdce.c (gen_shrink_wrap_conditions): Same.
10330 (shrink_wrap_one_built_in_call_with_conds): Same.
10331 (shrink_wrap_conditional_dead_built_in_calls): Same.
10332 * tree-data-ref.c (create_runtime_alias_checks): Same.
10333 (compute_all_dependences): Same.
10334 * tree-data-ref.h (compute_all_dependences): Same.
10335 (create_runtime_alias_checks): Same.
10336 (index_in_loop_nest): Same.
10337 * tree-if-conv.c (mask_exists): Same.
10338 * tree-loop-distribution.c (class loop_distribution): Same.
10339 (loop_distribution::create_rdg_vertices): Same.
10340 (dump_rdg_partitions): Same.
10341 (debug_rdg_partitions): Same.
10342 (partition_contains_all_rw): Same.
10343 (loop_distribution::distribute_loop): Same.
10344 * tree-parloops.c (oacc_entry_exit_ok_1): Same.
10345 (oacc_entry_exit_single_gang): Same.
10346 * tree-ssa-loop-im.c (hoist_memory_references): Same.
10347 (loop_suitable_for_sm): Same.
10348 * tree-ssa-loop-niter.c (bound_index): Same.
10349 * tree-ssa-reassoc.c (update_ops): Same.
10350 (swap_ops_for_binary_stmt): Same.
10351 (rewrite_expr_tree): Same.
10352 (rewrite_expr_tree_parallel): Same.
10353 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Same.
10354 * tree-ssa-sccvn.h (ao_ref_init_from_vn_reference): Same.
10355 * tree-ssa-structalias.c (process_all_all_constraints): Same.
10356 (make_constraints_to): Same.
10357 (handle_lhs_call): Same.
10358 (find_func_aliases_for_builtin_call): Same.
10359 (sort_fieldstack): Same.
10360 (check_for_overlaps): Same.
10361 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Same.
10362 (vect_create_cond_for_unequal_addrs): Same.
10363 (vect_create_cond_for_lower_bounds): Same.
10364 (vect_create_cond_for_alias_checks): Same.
10365 * tree-vect-slp-patterns.c (vect_validate_multiplication): Same.
10366 * tree-vect-slp.c (vect_analyze_slp_instance): Same.
10367 (vect_make_slp_decision): Same.
10368 (vect_slp_bbs): Same.
10369 (duplicate_and_interleave): Same.
10370 (vect_transform_slp_perm_load): Same.
10371 (vect_schedule_slp): Same.
10372 * tree-vectorizer.h (vect_transform_slp_perm_load): Same.
10373 (vect_schedule_slp): Same.
10374 (duplicate_and_interleave): Same.
10375 * tree.c (build_vector_from_ctor): Same.
10376 (build_vector): Same.
10377 (check_vector_cst): Same.
10378 (check_vector_cst_duplicate): Same.
10379 (check_vector_cst_fill): Same.
10380 (check_vector_cst_stepped): Same.
10381 * tree.h (build_vector_from_ctor): Same.
10383 2021-07-20 Jakub Jelinek <jakub@redhat.com>
10386 * config/rs6000/rs6000-protos.h (easy_altivec_constant): Change return
10387 type from bool to int.
10388 * config/rs6000/rs6000.c (vspltis_constant): Fix up handling the
10389 EASY_VECTOR_MSB case if either step or copies is not 1.
10390 (vspltis_shifted): Fix comment typo.
10391 (easy_altivec_constant): Change return type from bool to int, instead
10392 of returning true return byte size of the element mode that should be
10393 used to synthetize the constant.
10394 * config/rs6000/predicates.md (easy_vector_constant_msb): Require
10395 that vspltis_shifted is 0, handle the case where easy_altivec_constant
10396 assumes using different vector mode from CONST_VECTOR's mode.
10397 * config/rs6000/altivec.md (easy_vector_constant_msb splitter): Use
10398 easy_altivec_constant to determine mode in which -1 >> -1 should be
10399 performed, use rs6000_expand_vector_init instead of gen_vec_initv4sisi.
10401 2021-07-20 Richard Biener <rguenther@suse.de>
10404 * dwarf2out.h (dwarf_file_data): Add key member.
10405 * dwarf2out.c (dwarf_file_hasher::equal): Compare key.
10406 (dwarf_file_hasher::hash): Hash key.
10407 (lookup_filename): Remap the filename and store it in the
10408 filename member of dwarf_file_data when creating a new
10410 (file_name_acquire): Do not remap the filename again.
10411 (maybe_emit_file): Likewise.
10413 2021-07-20 Jonathan Wright <jonathan.wright@arm.com>
10415 * config/aarch64/aarch64-simd-builtins.def: Use two variant
10416 generators for all TBL/TBX intrinsics and rename to
10417 consistent forms: qtbl[1234] or qtbx[1234].
10418 * config/aarch64/aarch64-simd.md (aarch64_tbl1<mode>):
10420 (aarch64_qtbl1<mode>): This.
10421 (aarch64_tbx1<mode>): Rename to...
10422 (aarch64_qtbx1<mode>): This.
10423 (aarch64_tbl2v16qi): Delete.
10424 (aarch64_tbl3<mode>): Rename to...
10425 (aarch64_qtbl2<mode>): This.
10426 (aarch64_tbx4<mode>): Rename to...
10427 (aarch64_qtbx2<mode>): This.
10428 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_1): Use
10429 renamed qtbl1 and qtbl2 RTL patterns.
10430 * config/aarch64/arm_neon.h (vqtbl1_p8): Use renamed qtbl1
10432 (vqtbl1_s8): Likewise.
10433 (vqtbl1_u8): Likewise.
10434 (vqtbl1q_p8): Likewise.
10435 (vqtbl1q_s8): Likewise.
10436 (vqtbl1q_u8): Likewise.
10437 (vqtbx1_s8): Use renamed qtbx1 RTL pattern.
10438 (vqtbx1_u8): Likewise.
10439 (vqtbx1_p8): Likewise.
10440 (vqtbx1q_s8): Likewise.
10441 (vqtbx1q_u8): Likewise.
10442 (vqtbx1q_p8): Likewise.
10443 (vtbl1_s8): Use renamed qtbl1 RTL pattern.
10444 (vtbl1_u8): Likewise.
10445 (vtbl1_p8): Likewise.
10446 (vtbl2_s8): Likewise
10447 (vtbl2_u8): Likewise.
10448 (vtbl2_p8): Likewise.
10449 (vtbl3_s8): Use renamed qtbl2 RTL pattern.
10450 (vtbl3_u8): Likewise.
10451 (vtbl3_p8): Likewise.
10452 (vtbl4_s8): Likewise.
10453 (vtbl4_u8): Likewise.
10454 (vtbl4_p8): Likewise.
10455 (vtbx2_s8): Use renamed qtbx2 RTL pattern.
10456 (vtbx2_u8): Likewise.
10457 (vtbx2_p8): Likewise.
10458 (vqtbl2_s8): Use renamed qtbl2 RTL pattern.
10459 (vqtbl2_u8): Likewise.
10460 (vqtbl2_p8): Likewise.
10461 (vqtbl2q_s8): Likewise.
10462 (vqtbl2q_u8): Likewise.
10463 (vqtbl2q_p8): Likewise.
10464 (vqtbx2_s8): Use renamed qtbx2 RTL pattern.
10465 (vqtbx2_u8): Likewise.
10466 (vqtbx2_p8): Likewise.
10467 (vqtbx2q_s8): Likewise.
10468 (vqtbx2q_u8): Likewise.
10469 (vqtbx2q_p8): Likewise.
10470 (vtbx4_s8): Likewise.
10471 (vtbx4_u8): Likewise.
10472 (vtbx4_p8): Likewise.
10474 2021-07-20 Uroš Bizjak <ubizjak@gmail.com>
10477 * config/i386/sync.md (define_peephole2 atomic_storedi_fpu):
10479 (define_peephole2 atomic_loaddi_fpu): Ditto.
10481 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
10483 * config.gcc (riscv*-*-*): Detect which python is available.
10485 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
10487 * config/rs6000/vsx.md (mulhs_<mode>): Rename to...
10488 (smul<mode>3_highpart): ... this.
10489 (mulhu_<mode>): Rename to...
10490 (umul<mode>3_highpart): ... this.
10491 * config/rs6000/rs6000-builtin.def (MULHS_V2DI, MULHS_V4SI,
10492 MULHU_V2DI, MULHU_V4SI): Adjust.
10494 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
10496 PR tree-optimization/100696
10497 * internal-fn.c (first_commutative_argument): Add info for IFN_MULH.
10498 * internal-fn.def (IFN_MULH): New internal function.
10499 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Add support to
10500 recog normal multiply highpart as IFN_MULH.
10501 * config/i386/i386.c (ix86_add_stmt_cost): Adjust for combined
10504 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
10506 * config/elfos.h (CTF_DEBUGGING_INFO): New definition.
10507 (BTF_DEBUGGING_INFO): Likewise.
10508 * doc/tm.texi.in: Document the new macros.
10509 * doc/tm.texi: Regenerated.
10510 * toplev.c: Guard initialization of debug hooks.
10512 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
10514 * flags.h (ctf_debuginfo_p): New function declaration.
10515 * opts.c (ctf_debuginfo_p): New function definition.
10517 2021-07-19 Andrew Stubbs <ams@codesourcery.com>
10520 * config/gcn/gcn-hsa.h (DRIVER_SELF_SPECS): New.
10521 (ASM_SPEC): Set -mattr for xnack and sram-ecc.
10522 * config/gcn/gcn-opts.h (enum sram_ecc_type): New.
10523 * config/gcn/gcn-valu.md: Add a warning comment.
10524 * config/gcn/gcn.c (gcn_option_override): Add "sorry" for -mxnack.
10525 (output_file_start): Add xnack and sram-ecc state to ".amdgcn_target".
10526 * config/gcn/gcn.md: Add a warning comment.
10527 * config/gcn/gcn.opt: Add -mxnack and -msram-ecc.
10528 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): Remove
10530 (EF_AMDGPU_XNACK): New.
10531 (EF_AMDGPU_SRAM_ECC): New.
10533 (copy_early_debug_info): Use elf_flags.
10534 (main): Handle -mxnack and -msram-ecc options.
10535 * doc/invoke.texi: Document -mxnack and -msram-ecc.
10537 2021-07-19 Andrew Pinski <apinski@marvell.com>
10540 * config/aarch64/aarch64.md (csneg3_uxtw_insn): Rename to ...
10541 (*cs<neg_not_cs>3_uxtw_insn4): and extend to NEG_NOT.
10543 2021-07-19 Richard Biener <rguenther@suse.de>
10545 PR tree-optimization/101505
10546 * tree-vect-patterns.c (vect_determine_precisions): Walk
10547 PHIs also for loop vectorization.
10549 2021-07-19 Richard Biener <rguenther@suse.de>
10551 * gimple.h (gimple_expr_type): Remove.
10552 * doc/gimple.texi: Remove gimple_expr_type documentation.
10554 2021-07-19 Richard Biener <rguenther@suse.de>
10556 * tree-ssa-sccvn.c (vn_reference_eq): Handle NULL vr->type.
10557 (ao_ref_init_from_vn_reference): Likewise.
10558 (fully_constant_reference): Likewise.
10559 (vn_reference_lookup_call): Do not set vr->type to random
10561 * tree-ssa-pre.c (compute_avail): Do not try to PRE calls
10563 * tree-vect-generic.c (expand_vector_piecewise): Pass in
10564 whether we expanded parallel.
10565 (expand_vector_parallel): Adjust.
10566 (expand_vector_addition): Likewise.
10567 (expand_vector_comparison): Likewise.
10568 (expand_vector_operation): Likewise.
10569 (expand_vector_scalar_condition): Likewise.
10570 (expand_vector_conversion): Likewise.
10572 2021-07-19 Richard Biener <rguenther@suse.de>
10574 * tree-vrp.c (register_edge_assert_for_2): Use the
10576 (vrp_folder::fold_predicate_in): Likewise.
10577 * vr-values.c (gimple_assign_nonzero_p): Likewise.
10578 (vr_values::extract_range_from_comparison): Likewise.
10579 (vr_values::extract_range_from_ubsan_builtin): Use the
10580 type of the first operand.
10581 (vr_values::extract_range_basic): Push down type
10582 computation, use the appropriate LHS.
10583 (vr_values::extract_range_from_assignment): Use the
10586 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
10589 * common/config/i386/i386-common.c (ix86_handle_option): For
10590 -mgeneral-regs-only, enable the GPR only instructions which are
10591 enabled implicitly by SSE ISAs unless they have been disabled
10594 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
10597 * config/i386/i386.c (ix86_check_avx_upper_stores): Moved before
10598 ix86_avx_u128_mode_needed.
10599 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY if callee
10600 returns AVX register.
10602 2021-07-17 Jan Hubicka <hubicka@ucw.cz>
10604 * tree-ssa-structalias.c (handle_rhs_call): Support EAF_NOT_RETURNED.
10605 (handle_const_call): Liekise
10606 (handle_pure_call): Liekise
10608 2021-07-17 Andrew MacLeod <amacleod@redhat.com>
10610 PR tree-optimization/96542
10611 * range-op.cc (range_operator::wi_fold_in_parts): New.
10612 (range_operator::fold_range): Call wi_fold_in_parts.
10613 (operator_lshift::wi_fold): Fix broken lshift by [0,0].
10614 * range-op.h (wi_fold_in_parts): Add prototype.
10616 2021-07-16 David Malcolm <dmalcolm@redhat.com>
10618 * doc/analyzer.texi: Add __analyzer_dump_state.
10620 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10622 * config/rs6000/rbtree.c: New file.
10623 * config/rs6000/rbtree.h: New file.
10625 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10627 * config/rs6000/rs6000-gen-builtins.c (restriction): New enum.
10628 (typeinfo): Add restr field.
10629 (match_bracketed_pair): New function.
10630 (match_const_restriction): Implement.
10632 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10634 * config/rs6000/rs6000-gen-builtins.c (match_basetype): Implement.
10636 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10638 * config/rs6000/rs6000-gen-builtins.c (void_status): New enum.
10639 (basetype): Likewise.
10640 (typeinfo): Likewise.
10641 (handle_pointer): New function.
10642 (match_basetype): New stub function.
10643 (match_const_restriction): Likewise.
10644 (match_type): New function.
10646 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10648 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace): New
10650 (advance_line): Likewise.
10651 (safe_inc_pos): Likewise.
10652 (match_identifier): Likewise.
10653 (match_integer): Likewise.
10654 (match_to_right_bracket): Likewise.
10656 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10658 * config/rs6000/rs6000-gen-builtins.c (bif_file): New variable.
10659 (ovld_file): Likewise.
10660 (header_file): Likewise.
10661 (init_file): Likewise.
10662 (defines_file): Likewise.
10663 (pgm_path): Likewise.
10664 (bif_path): Likewise.
10665 (ovld_path): Likewise.
10666 (header_path): Likewise.
10667 (init_path): Likewise.
10668 (defines_path): Likewise.
10669 (LINELEN): New macro.
10670 (linebuf): New variable.
10674 (bif_diag): New function.
10675 (ovld_diag): Likewise.
10677 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10679 * config/rs6000/rs6000-builtin-new.def: New.
10680 * config/rs6000/rs6000-overload.def: New.
10682 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10684 * config/rs6000/rs6000-gen-builtins.c: New.
10686 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10688 * Makefile.in (EXTRA_GTYPE_DEPS): New variable.
10689 (s-gtype): Depend on EXTRA_GTYPE_DEPS.
10690 * gengtype-state.c (state_writer::write_state_file_list): Add a
10691 parameter to the fileslist expression for the number of build
10693 (read_state_files_list): Detect build headers and strip the
10694 initial "./" or ".\" from their names.
10695 * gengtype.c (build_headers): New global variable.
10696 (num_build_headers): Likewise.
10697 (open_base_files): Emit #include for each build header.
10698 (main): Detect and count build headers.
10699 * gengtype.h (build_headers): New extern variable.
10700 (num_build_headers): Likewise.
10702 2021-07-16 Richard Biener <rguenther@suse.de>
10704 * gimple-ssa-store-merging.c (verify_symbolic_number_p): Use
10705 the type of the LHS.
10706 (find_bswap_or_nop_1): Likewise.
10707 (find_bswap_or_nop): Likewise.
10708 * tree-vectorizer.h (vect_get_smallest_scalar_type): Adjust
10710 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
10711 Remove unused parameters, pass in the scalar type. Fix
10712 internal store function handling.
10713 * tree-vect-stmts.c (vect_analyze_stmt): Remove assert.
10714 (vect_get_vector_types_for_stmt): Move down check for
10715 existing vector stmt after we've determined a scalar type.
10716 Pass down the used scalar type to vect_get_smallest_scalar_type.
10717 * tree-vect-generic.c (expand_vector_condition): Use
10718 the type of the LHS.
10719 (expand_vector_scalar_condition): Likewise.
10720 (expand_vector_operations_1): Likewise.
10721 * tree-vect-patterns.c (vect_widened_op_tree): Likewise.
10722 (vect_recog_dot_prod_pattern): Likewise.
10723 (vect_recog_sad_pattern): Likewise.
10724 (vect_recog_widen_op_pattern): Likewise.
10725 (vect_recog_widen_sum_pattern): Likewise.
10726 (vect_recog_mixed_size_cond_pattern): Likewise.
10728 2021-07-16 Jan Hubicka <hubicka@ucw.cz>
10730 * ipa-modref.c (struct escape_entry): Use eaf_fleags_t.
10731 (dump_eaf_flags): Dump EAF_NOT_RETURNED
10732 (eaf_flags_useful_p): Use eaf_fleags_t; handle const functions
10733 and EAF_NOT_RETURNED.
10734 (modref_summary::useful_p): Likewise.
10735 (modref_summary_lto::useful_p): Likewise.
10736 (struct) modref_summary_lto: Use eaf_fleags_t.
10737 (deref_flags): Handle EAF_NOT_RETURNED.
10738 (struct escape_point): Use min_flags.
10739 (modref_lattice::init): Add EAF_NOT_RETURNED.
10740 (merge_call_lhs_flags): Ignore EAF_NOT_RETURNED functions
10741 (analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
10743 (analyze_parms): Also analyze const functions; update conition on
10745 (modref_write): Update streaming.
10746 (read_section): Update streaming.
10747 (remap_arg_flags): Use eaf_flags_t.
10748 (modref_merge_call_site_flags): Hanlde EAF_NOT_RETURNED.
10749 * ipa-modref.h: (eaf_flags_t): New typedef.
10750 (struct modref_summary): Use eaf_flags_t.
10751 * tree-core.h (EAF_NOT_RETURNED): New constant.
10753 2021-07-16 Richard Biener <rguenther@suse.de>
10755 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Use
10756 the type of the LHS.
10757 (gimple_assign_nonnegative_warnv_p): Likewise.
10758 (gimple_call_nonnegative_warnv_p): Likewise. Return false
10759 if the call has no LHS.
10760 * gimple.c (gimple_could_trap_p_1): Use the type of the LHS.
10761 * tree-eh.c (stmt_could_throw_1_p): Likewise.
10762 * tree-inline.c (insert_init_stmt): Likewise.
10763 * tree-ssa-loop-niter.c (get_val_for): Likewise.
10764 * tree-outof-ssa.c (ssa_is_replaceable_p): Use the type of
10766 * tree-ssa-sccvn.c (init_vn_nary_op_from_stmt): Take a
10767 gassign *. Use the type of the lhs.
10768 (vn_nary_op_lookup_stmt): Adjust.
10769 (vn_nary_op_insert_stmt): Likewise.
10771 2021-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
10773 * config/s390/predicates.md (bras_sym_operand): Accept all
10774 functions in 64-bit mode, use UNSPEC_PLT31.
10775 (larl_operand): Use UNSPEC_PLT31.
10776 * config/s390/s390.c (s390_loadrelative_operand_p): Likewise.
10777 (legitimize_pic_address): Likewise.
10778 (s390_emit_tls_call_insn): Mark __tls_get_offset as function,
10780 (s390_delegitimize_address): Use UNSPEC_PLT31.
10781 (s390_output_addr_const_extra): Likewise.
10782 (print_operand): Add @PLT to TLS calls, handle %K.
10783 (s390_function_profiler): Mark __fentry__/_mcount as function,
10784 use %K, use UNSPEC_PLT31.
10785 (s390_output_mi_thunk): Use only UNSPEC_GOT, use %K.
10786 (s390_emit_call): Use UNSPEC_PLT31.
10787 (s390_emit_tpf_eh_return): Mark __tpf_eh_return as function.
10788 * config/s390/s390.md (UNSPEC_PLT31): Rename from UNSPEC_PLT.
10789 (*movdi_64): Use %K.
10790 (reload_base_64): Likewise.
10791 (*sibcall_brc): Likewise.
10792 (*sibcall_brcl): Likewise.
10793 (*sibcall_value_brc): Likewise.
10794 (*sibcall_value_brcl): Likewise.
10796 (*brasl): Likewise.
10797 (*bras_r): Likewise.
10798 (*brasl_r): Likewise.
10799 (*bras_tls): Likewise.
10800 (*brasl_tls): Likewise.
10801 (main_base_64): Likewise.
10802 (reload_base_64): Likewise.
10803 (@split_stack_call<mode>): Likewise.
10805 2021-07-16 Richard Biener <rguenther@suse.de>
10807 PR tree-optimization/101467
10808 * tree-vect-stmts.c (vect_gen_while): Properly guard
10809 make_temp_ssa_name usage.
10811 2021-07-16 Cooper Qu <cooper.qu@linux.alibaba.com>
10813 * config.gcc: Don't use forked print-sysroot-suffix.sh and
10814 t-sysroot-suffix for C-SKY.
10815 * config/csky/print-sysroot-suffix.sh: Delete.
10816 * config/csky/t-csky-linux: Delete.
10817 * config/csky/t-sysroot-suffix: Define MULTILIB_DIRNAMES
10818 instead of CSKY_MULTILIB_DIRNAMES.
10820 2021-07-16 Richard Biener <rguenther@suse.de>
10822 * tree-vect-loop.c (vect_transform_cycle_phi): Correct sign
10823 conversion issues with the partial reduction of the reused
10824 vector accumulator.
10826 2021-07-16 Richard Biener <rguenther@suse.de>
10828 * config/i386/i386-options.c (ix86_option_override_internal): Set
10829 param_vect_partial_vector_usage to zero if not set.
10831 2021-07-15 Uroš Bizjak <ubizjak@gmail.com>
10834 * config/i386/i386.h (VALID_SSE_REG_MODE): Add TDmode.
10835 (VALID_INT_MODE_P): Add SDmode and DDmode.
10836 Add TDmode for TARGET_64BIT.
10837 (VALID_DFP_MODE_P): Remove.
10838 * config/i386/i386.c (ix86_hard_regno_mode_ok):
10839 Do not use VALID_DFP_MODE_P.
10841 2021-07-15 Andrew MacLeod <amacleod@redhat.com>
10843 * gimple-range-fold.cc (adjust_pointer_diff_expr): Use
10845 (fold_using_range::fold_stmt): Ditto.
10846 (fold_using_range::range_of_range_op): Ditto.
10847 (fold_using_range::range_of_phi): Ditto.
10848 (fold_using_range::range_of_call): Ditto.
10849 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
10850 (fold_using_range::range_of_builtin_call): Ditto.
10851 (fold_using_range::range_of_cond_expr): Ditto.
10852 * gimple-range-fold.h (gimple_range_type): New.
10854 2021-07-15 Martin Sebor <msebor@redhat.com>
10856 PR middle-end/97027
10857 * tree-ssa-strlen.c (handle_assign): New function.
10858 (maybe_warn_overflow): Add argument.
10859 (nonzero_bytes_for_type): New function.
10860 (count_nonzero_bytes): Handle more tree types. Call
10861 nonzero_bytes_for_tye.
10862 (count_nonzero_bytes): Handle types.
10863 (handle_store): Handle stores from function calls.
10864 (strlen_check_and_optimize_call): Move code to handle_assign. Call
10865 it for assignments from function calls.
10867 2021-07-15 David Malcolm <dmalcolm@redhat.com>
10872 * doc/invoke.texi: Add -Wanalyzer-use-of-uninitialized-value.
10874 2021-07-15 David Malcolm <dmalcolm@redhat.com>
10876 * doc/invoke.texi (-fdump-analyzer-exploded-paths): New.
10878 2021-07-15 Martin Sebor <msebor@redhat.com>
10882 * fold-const.c (operand_compare::operand_equal_p): Handle OEP_DECL_NAME.
10883 (operand_compare::verify_hash_value): Same.
10884 * tree-core.h (OEP_DECL_NAME): New.
10886 2021-07-15 Martin Jambor <mjambor@suse.cz>
10888 * profile-count.h (profile_count::value): Change the return type to
10890 * gimple-pretty-print.c (dump_gimple_bb_header): Adjust print
10892 * tree-cfg.c (dump_function_to_file): Likewise.
10894 2021-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
10897 * config/rs6000/rs6000-p8swap.c (has_part_mult): New.
10898 (rs6000_analyze_swaps): Insns containing a subreg of a mult are
10901 2021-07-15 Richard Biener <rguenther@suse.de>
10903 * tree-vectorizer.h (vect_gen_while): Match up with
10904 vect_gen_while_not.
10905 * tree-vect-stmts.c (vect_gen_while): Adjust API to that
10906 of vect_gen_while_not.
10907 (vect_gen_while_not): Adjust.
10908 * tree-vect-loop-manip.c (vect_set_loop_controls_directly): Likewise.
10910 2021-07-15 Aldy Hernandez <aldyh@redhat.com>
10912 * gimple-range-cache.cc (non_null_ref::adjust_range): New.
10913 (ranger_cache::range_of_def): Call adjust_range.
10914 (ranger_cache::entry_range): Same.
10915 * gimple-range-cache.h (non_null_ref::adjust_range): New.
10916 * gimple-range.cc (gimple_ranger::range_of_expr): Call
10918 (gimple_ranger::range_on_entry): Same.
10920 2021-07-15 Tamar Christina <tamar.christina@arm.com>
10923 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10925 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
10927 2021-07-15 Tamar Christina <tamar.christina@arm.com>
10930 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10932 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
10933 (sdot_prod, udot_prod): ...These.
10934 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
10935 (aarch64_<sur>dot<vsi2qi>): Rename to...
10936 (<sur>dot_prod<vsi2qi>): ...This.
10937 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
10940 2021-07-15 Jakub Jelinek <jakub@redhat.com>
10942 PR middle-end/101437
10943 * gimplify.c (gimplify_expr): Throw away volatile reads from empty
10944 types even if they have non-BLKmode TYPE_MODE.
10946 2021-07-15 Richard Biener <rguenther@suse.de>
10949 * gcc.c (process_command): Process -gtoggle like process_options
10950 would after parsing options.
10952 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
10954 * cfgexpand.c (expand_asm_loc): Adjust.
10955 (expand_asm_stmt): Likewise.
10956 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Likewise.
10957 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
10958 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
10959 * config/avr/avr.c (avr_md_asm_adjust): Likewise.
10960 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
10961 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
10962 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
10963 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
10964 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
10965 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
10966 * config/s390/s390.c (s390_md_asm_adjust): Likewise.
10967 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
10968 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
10969 * doc/tm.texi: Regenerate.
10970 * target.def: Add location argument to md_asm_adjust.
10972 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
10974 * tree-diagnostic.c (diagnostic_report_current_function): Use the
10975 diagnostic's location, not input_location.
10977 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
10979 * cfgexpand.c (tree_conflicts_with_clobbers_p): Pass location to
10981 (expand_asm_stmt): Likewise.
10983 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
10985 * config/rs6000/rs6000.c (adjacent_mem_locations): Return the lower
10986 addressed memory rtx, if any.
10987 (rs6000_split_multireg_move): Fix code formatting.
10988 Handle MMA build built-ins with operands in adjacent memory locations.
10990 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
10992 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Move to later
10995 2021-07-14 Jason Merrill <jason@redhat.com>
10997 * sel-sched-ir.h (get_all_loop_exits): Use auto_vec.
10999 2021-07-14 Jason Merrill <jason@redhat.com>
11001 * doc/invoke.texi: -fdelete-dead-exceptions is on by default for
11004 2021-07-14 Tamar Christina <tamar.christina@arm.com>
11006 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
11007 Remove erroneous line.
11009 2021-07-14 Andrew MacLeod <amacleod@redhat.com>
11011 * params.opt (param_evrp_mode): Change default.
11013 2021-07-14 Tamar Christina <tamar.christina@arm.com>
11015 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
11016 (sdot_prod, udot_prod): ...These.
11017 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
11018 (aarch64_<sur>dot<vsi2qi>): Rename to...
11019 (<sur>dot_prod<vsi2qi>): ...This.
11020 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
11023 2021-07-14 Tamar Christina <tamar.christina@arm.com>
11025 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
11027 2021-07-14 Tamar Christina <tamar.christina@arm.com>
11029 * doc/sourcebuild.texi (arm_v8_2a_i8mm_neon_hw): Document.
11031 2021-07-14 Tamar Christina <tamar.christina@arm.com>
11033 * config/arm/neon.md (usdot_prod<vsi2qi>): New.
11035 2021-07-14 Tamar Christina <tamar.christina@arm.com>
11037 * config/aarch64/aarch64-simd.md (aarch64_usdot<vsi2qi>): Rename to...
11038 (usdot_prod<vsi2qi>): ... This.
11039 * config/aarch64/aarch64-simd-builtins.def (usdot): Rename to...
11040 (usdot_prod): ...This.
11041 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Likewise.
11042 * config/aarch64/aarch64-sve.md (@aarch64_<sur>dot_prod<vsi2qi>):
11044 (@<sur>dot_prod<vsi2qi>): ...This.
11045 * config/aarch64/aarch64-sve-builtins-base.cc
11046 (svusdot_impl::expand): Use it.
11048 2021-07-14 Tamar Christina <tamar.christina@arm.com>
11050 * optabs.def (usdot_prod_optab): New.
11051 * doc/md.texi: Document it and clarify other dot prod optabs.
11052 * optabs-tree.h (enum optab_subtype): Add optab_vector_mixed_sign.
11053 * optabs-tree.c (optab_for_tree_code): Support usdot_prod_optab.
11054 * optabs.c (expand_widen_pattern_expr): Likewise.
11055 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
11056 * tree-vect-loop.c (vectorizable_reduction): Query dot-product kind.
11057 * tree-vect-patterns.c (vect_supportable_direct_optab_p): Take optional
11059 (vect_widened_op_tree): Optionally ignore
11061 (vect_recog_dot_prod_pattern): Support usdot_prod_optab.
11063 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
11066 * config/i386/driver-i386.c (host_detect_local_cpu): Check
11067 "arch [32|64]" and "tune [32|64]" for 32-bit and 64-bit codegen.
11068 Enable UINTR only for 64-bit codegen.
11069 * config/i386/i386-options.c
11070 (ix86_option_override_internal::DEF_PTA): Skip PTA_UINTR if not
11072 * config/i386/i386.h (ARCH_ARG): New.
11073 (CC1_CPU_SPEC): Pass "[arch|tune] 32" for 32-bit codegen and
11074 "[arch|tune] 64" for 64-bit codegen.
11076 2021-07-14 Richard Biener <rguenther@suse.de>
11078 PR tree-optimization/101445
11079 * tree-vect-stmts.c (vectorizable_load): Do the gap adjustment
11080 of the IV in the correct direction for negative stride
11083 2021-07-14 Jakub Jelinek <jakub@redhat.com>
11086 * godump.c (godump_str_hash): New type.
11087 (godump_container::pot_dummy_types): Use string_hash instead of
11088 ptr_hash in the hash_set.
11090 2021-07-14 Richard Biener <rguenther@suse.de>
11092 * tree-vect-loop.c (vect_find_reusable_accumulator): Handle
11093 vector types where the old vector type has a multiple of
11094 the new vector type elements.
11095 (vect_create_partial_epilog): New function, split out from...
11096 (vect_create_epilog_for_reduction): ... here.
11097 (vect_transform_cycle_phi): Reduce the re-used accumulator
11098 to the new vector type.
11100 2021-07-14 Alexandre Oliva <oliva@adacore.com>
11102 * tree-ssa-alias.c (attr_fnspec::verify): Fix index in
11103 non-'t'-sized arg check.
11105 2021-07-14 Alexandre Oliva <oliva@adacore.com>
11107 * tree-cfg.c (cleanup_dead_labels_eh): Update
11108 post_landing_pad label upon change of landing pad block's
11110 (cleanup_dead_labels): Check that a removed label is not that
11113 2021-07-13 Jonathan Wright <jonathan.wright@arm.com>
11115 * combine.c (combine_simplify_rtx): Add vec_select -> subreg
11117 * config/aarch64/aarch64.md (*zero_extend<SHORT:mode><GPI:mode>2_aarch64):
11118 Add Neon to general purpose register case for zero-extend
11120 * config/arm/vfp.md (*arm_movsi_vfp): Remove "*" from *t -> r
11121 case to prevent some cases opting to go through memory.
11122 * cse.c (fold_rtx): Add vec_select -> subreg simplification.
11123 * rtl.c (rtvec_series_p): Define predicate to determine
11124 whether a vector contains a linear series of integers.
11125 * rtl.h (rtvec_series_p): Define.
11126 * rtlanal.c (vec_series_lowpart_p): Define predicate to
11127 determine if a vector selection is equivalent to the low part
11129 * rtlanal.h (vec_series_lowpart_p): Define.
11130 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
11131 Add vec_select -> subreg simplification.
11133 2021-07-13 Paul A. Clarke <pc@us.ibm.com>
11135 * config/rs6000/smmintrin.h (_mm_testz_si128, _mm_testc_si128,
11136 _mm_testnzc_si128, _mm_test_all_ones, _mm_test_all_zeros,
11137 _mm_test_mix_ones_zeros): New.
11139 2021-07-13 Roger Sayle <roger@nextmovesoftware.com>
11140 Richard Biener <rguenther@suse.de>
11142 * gimple.c (gimple_could_trap_p_1): Make S argument a
11143 "const gimple*". Preserve constness in call to
11144 gimple_asm_volatile_p.
11145 (gimple_could_trap_p): Make S argument a "const gimple*".
11146 * gimple.h (gimple_could_trap_p_1, gimple_could_trap_p):
11147 Update function prototypes.
11149 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
11151 * tree-vectorizer.h (vect_reusable_accumulator): New structure.
11152 (_loop_vec_info::main_loop_edge): New field.
11153 (_loop_vec_info::skip_main_loop_edge): Likewise.
11154 (_loop_vec_info::skip_this_loop_edge): Likewise.
11155 (_loop_vec_info::reusable_accumulators): Likewise.
11156 (_stmt_vec_info::reduc_scalar_results): Likewise.
11157 (_stmt_vec_info::reused_accumulator): Likewise.
11158 (vect_get_main_loop_result): Declare.
11159 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize
11160 reduc_scalar_inputs.
11161 (vec_info::free_stmt_vec_info): Free reduc_scalar_inputs.
11162 * tree-vect-loop-manip.c (vect_get_main_loop_result): New function.
11163 (vect_do_peeling): Fill an epilogue loop's main_loop_edge,
11164 skip_main_loop_edge and skip_this_loop_edge fields.
11165 * tree-vect-loop.c (INCLUDE_ALGORITHM): Define.
11166 (vect_emit_reduction_init_stmts): New function.
11167 (get_initial_def_for_reduction): Use it.
11168 (get_initial_defs_for_reduction): Likewise. Change the vinfo
11169 parameter to a loop_vec_info.
11170 (vect_create_epilog_for_reduction): Store the scalar results
11171 in the reduc_info. If an epilogue loop is reusing an accumulator
11172 from the main loop, and if the epilogue loop can also be skipped,
11173 try to place the reduction code in the join block. Record
11174 accumulators that could potentially be reused by epilogue loops.
11175 (vect_transform_cycle_phi): When vectorizing epilogue loops,
11176 try to reuse accumulators from the main loop. Record the initial
11177 value in reduc_info for non-SLP reductions too.
11179 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
11181 * tree-vect-loop.c (get_initial_def_for_reduction): Remove
11182 adjustment handling. Take the neutral value as an argument,
11183 in place of the code argument.
11184 (vect_transform_cycle_phi): Update accordingly. Handle the
11185 initial values of cond reductions separately from code reductions.
11186 Choose the adjustment here rather than in
11187 get_initial_def_for_reduction. Sink the splat of vec_initial_def.
11189 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
11191 * tree-vect-loop.c (neutral_op_for_slp_reduction): Replace with...
11192 (neutral_op_for_reduction): ...this, providing a more general
11194 (vect_create_epilog_for_reduction): Update accordingly.
11195 (vectorizable_reduction): Likewise.
11196 (vect_transform_cycle_phi): Likewise.
11198 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
11200 * tree-vect-loop.c (get_initial_def_for_reduction): Take the
11201 reduc_info instead of the original stmt_vec_info.
11202 (vect_transform_cycle_phi): Update accordingly.
11204 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
11206 * tree-vect-loop.c (get_initial_defs_for_reduction): Take the
11207 reduc_info as an additional parameter.
11208 (vect_transform_cycle_phi): Update accordingly.
11210 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
11212 * tree-vectorizer.h: Include tree-ssa-operands.h.
11213 (vect_phi_initial_value): New function.
11214 * tree-vect-loop.c (neutral_op_for_slp_reduction): Use it.
11215 (get_initial_defs_for_reduction, info_for_reduction): Likewise.
11216 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
11217 (vect_transform_cycle_phi, vectorizable_induction): Likewise.
11219 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
11221 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
11222 the phi results to vectype after creating them. Remove later
11223 conversion code that thus becomes redundant.
11225 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
11227 * tree-vect-loop.c (vect_create_epilog_for_reduction): Replace
11228 the new_phis vector with a reduc_inputs vector. Combine handling
11229 of reduction chains and ncopies > 1.
11231 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
11233 * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
11234 scalar_results to group_size elements after reducing down from
11235 N*group_size elements. Construct an array_slice of the live-out
11236 stmts and assert that there is one stmt per scalar result.
11238 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
11240 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
11241 nested_in_vect_loop and use double_reduc everywhere. Remove dead
11242 assignment to "loop".
11244 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
11246 * internal-fn.c (vectorized_internal_fn_supported_p): Handle
11247 vector types first. For scalar types, consider both the preferred
11248 vector mode and the alternative vector modes.
11249 * optabs-query.c (can_vec_mask_load_store_p): Use the same
11250 structure as above, in particular using related_vector_mode
11251 for modes provided by autovectorize_vector_modes.
11253 2021-07-13 Jakub Jelinek <jakub@redhat.com>
11254 Richard Biener <rguenther@suse.de>
11256 PR tree-optimization/101419
11257 * tree-pass.h (PROP_objsz): Define.
11258 (make_pass_early_object_sizes): Declare.
11259 * passes.def (pass_all_early_optimizations): Rename pass_object_sizes
11260 there to pass_early_object_sizes, drop parameter.
11261 (pass_all_optimizations): Move pass_object_sizes right after pass_ccp,
11262 drop parameter, move pass_post_ipa_warn right after that.
11263 * tree-object-size.c (pass_object_sizes::execute): Rename to...
11264 (object_sizes_execute): ... this. Add insert_min_max_p argument.
11265 (pass_data_object_sizes): Move after object_sizes_execute.
11266 (pass_object_sizes): Likewise. In execute method call
11267 object_sizes_execute, drop set_pass_param method and insert_min_max_p
11268 non-static data member and its initializer in the ctor.
11269 (pass_data_early_object_sizes, pass_early_object_sizes,
11270 make_pass_early_object_sizes): New.
11271 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Use
11272 (cfun->curr_properties & PROP_objsz) instead of cfun->after_inlining.
11274 2021-07-13 Kito Cheng <kito.cheng@sifive.com>
11277 * config/riscv/constraints.md ("S"): Update description and remove
11279 * doc/md.texi (Machine Constraints): Document the 'S' constraints
11282 2021-07-13 Richard Biener <rguenther@suse.de>
11285 2021-07-12 Richard Biener <rguenther@suse.de>
11287 * tree-vect-slp.c (vect_slp_region): Show the number of
11288 SLP graph entries in the optimization message.
11290 2021-07-13 Michael Meissner <meissner@linux.ibm.com>
11292 * config/rs6000/altivec.md (xxspltiw_v4sf): Change local variable
11294 * config/rs6000/rs6000-protos.h (rs6000_const_f32_to_i32): Change
11295 return type to long.
11296 * config/rs6000/rs6000.c (rs6000_const_f32_to_i32): Change return
11299 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
11301 * gimple-range-fold.cc (fold_using_range::range_of_builtin_ubsan_call):
11302 Query relation between the 2 operands and use it.
11304 2021-07-12 Sergei Trofimovich <siarheit@google.com>
11306 * doc/cfg.texi: Fix s/ei_safe_safe/ei_safe_edge/ typo.
11308 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
11311 * config/i386/predicates.md (vec_setm_sse41_operand):
11312 Rename from vec_setm_operand.
11313 (vec_setm_avx2_operand): New predicate.
11314 * config/i386/sse.md (vec_set<V_128:mode>): Use V_128 mode iterator.
11315 Use vec_setm_sse41_operand as operand 2 predicate.
11316 (vec_set<V_256_512:mode): New expander.
11317 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_sse41_operand
11318 as operand 2 predicate.
11320 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
11322 PR tree-optimization/101335
11323 * range-op.cc (operator_cast::lhs_op1_relation): Delete.
11325 2021-07-12 Andrew Pinski <apinski@marvell.com>
11327 * tree-ssa-phiopt.c (match_simplify_replacement): Move
11328 insert of the sequence before the movement of the
11329 statement. Check if to see if the statement is used
11330 outside of the original phi to see if we should move it.
11332 2021-07-12 Richard Biener <rguenther@suse.de>
11334 * dump-context.h (debug_dump_context::debug_dump_context):
11335 Add FILE * parameter defaulted to stderr.
11336 * dumpfile.c (debug_dump_context::debug_dump_context): Adjust.
11337 * tree-vect-slp.c (dot_slp_tree): New functions.
11339 2021-07-12 Richard Biener <rguenther@suse.de>
11341 PR tree-optimization/101373
11342 * tree-ssa-pre.c (prune_clobbered_mems): Also prune trapping
11343 references when the BB may not return.
11344 (compute_avail): Pass in the function we're working on and
11345 replace cfun references with it. Externally throwing
11346 const calls also possibly terminate the function.
11347 (pass_pre::execute): Pass down the function we're working on.
11348 * gcse.c (compute_hash_table_work): Externally throwing
11349 const/pure calls also need record_last_mem_set_info.
11350 * postreload-gcse.c (record_opr_changes): Looping or externally
11351 throwing const/pure calls also need record_last_mem_set_info.
11353 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
11355 * recog.c (memory_address_addr_space_p): Change the type to bool.
11356 Return true/false instead of 1/0.
11357 (offsettable_memref_p): Ditto.
11358 (offsettable_nonstrict_memref_p): Ditto.
11359 (offsettable_address_addr_space_p): Ditto.
11360 Change the type of addressp indirect function to bool.
11361 * recog.h (memory_address_addr_space_p): Change the type to bool.
11362 (strict_memory_address_addr_space_p): Ditto.
11363 (offsettable_memref_p): Ditto.
11364 (offsettable_nonstrict_memref_p): Ditto.
11365 (offsettable_address_addr_space_p): Ditto.
11366 * reload.c (maybe_memory_address_addr_space_p): Ditto.
11367 (strict_memory_address_addr_space_p): Change the type to bool.
11368 Return true/false instead of 1/0.
11369 (maybe_memory_address_addr_space_p): Change the type to bool.
11371 2021-07-12 Richard Biener <rguenther@suse.de>
11373 * tree-vect-slp.c (vect_slp_region): Show the number of
11374 SLP graph entries in the optimization message.
11376 2021-07-12 Richard Biener <rguenther@suse.de>
11378 PR tree-optimization/101394
11379 * tree-ssa-pre.c (do_pre_regular_insertion): Avoid inserting
11380 copies from abnormals for a full redundancy.
11382 2021-07-12 Richard Biener <rguenther@suse.de>
11384 PR middle-end/101423
11385 * gimple.c (gimple_could_trap_p_1): Internal function calls
11387 * tree-eh.c (tree_could_trap_p): Likewise.
11389 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
11392 * config/arm/arm_neon.h (vmul_n_u32): Replace call to builtin with
11394 (vmulq_n_u32): Likewise.
11395 (vmul_n_f32): Gate __a * __b on __FAST_MATH__.
11396 (vmulq_n_f32): Likewise.
11397 (vmul_n_f16): Likewise.
11398 (vmulq_n_f16): Likewise.
11400 2021-07-12 Martin Liska <mliska@suse.cz>
11402 PR sanitizer/101425
11403 * gcc.c (check_offload_target_name): Call
11404 candidates_list_and_hint only if we have a candidate.
11406 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
11409 * config/arm/neon.md (vec_init): Move to ...
11410 * config/arm/vec-common.md (vec_init): ... here.
11411 Change the pattern's mode to VDQX and gate it on VALID_MVE_MODE.
11413 2021-07-12 Roger Sayle <roger@nextmovesoftware.com>
11415 PR tree-optimization/101403
11416 * match.pd ((T)bswap(X)>>C): Correctly handle cases where
11417 signedness of the shift is not the same as the signedness of
11418 the type extension.
11420 2021-07-09 Roger Sayle <roger@nextmovesoftware.com>
11421 Uroš Bizjak <ubizjak@gmail.com>
11423 * config/i386/i386.md (*divmodsi4_const): Optimize SImode
11424 divmod of a constant numerator with new define_insn_and_split.
11426 2021-07-09 Iain Sandoe <iain@sandoe.co.uk>
11429 * config/i386/i386-expand.c (ix86_expand_call): If a call is
11430 to a non-local-binding, or local but to a public symbol, then
11431 assume that it might be indirected via the lazy symbol binder.
11432 Mark R10 and R10 as clobbered in that case.
11434 2021-07-09 Eric Botcazou <ebotcazou@adacore.com>
11437 * gcc.c (ASM_DEBUG_DWARF_OPTION): Set again to --gdwarf2 in
11438 the case where HAVE_AS_WORKING_DWARF_N_FLAG is not defined
11439 and HAVE_LD_BROKEN_PE_DWARF5 is defined.
11441 2021-07-09 Uroš Bizjak <ubizjak@gmail.com>
11443 * config/i386/i386.md (*udivmodsi4_pow2_zext_1): Limit the
11444 log2 range of operands[3] to [1,31].
11445 (*udivmodsi4_pow2_zext_2): Ditto. Correct insn RTX pattern.
11447 2021-07-09 Sergei Trofimovich <siarheit@google.com>
11449 * doc/md.texi: Don't split @smallexample in multiple @groups.
11451 2021-07-09 Sergei Trofimovich <siarheit@google.com>
11453 * doc/md.texi: Add missing 'see' word.
11455 2021-07-09 Andrew Pinski <apinski@marvell.com>
11457 * tree-ssa-phiopt.c (phiopt_early_allow): Change arguments
11458 to take sequence and gimple_match_op. Accept the case where
11459 op is a SSA_NAME and one statement in the sequence.
11460 Also allow constants.
11461 (gimple_simplify_phiopt): Always pass a sequence to resimplify.
11462 Update call to phiopt_early_allow. Discard the sequence if not
11465 2021-07-09 Xi Ruoyao <xry111@mengyan1223.wang>
11470 * config/mips/mips.c (mips_const_insns): Use MSA_SUPPORTED_MODE_P
11471 instead of ISA_HAS_MSA.
11472 (mips_expand_vec_unpack): Likewise.
11473 (mips_expand_vector_init): Likewise.
11475 2021-07-09 Kewen Lin <linkw@linux.ibm.com>
11477 * config/rs6000/vsx.md (mods_<mode>): Rename to...
11478 (mod<mode>3): ... this.
11479 (modu_<mode>): Rename to...
11480 (umod<mode>3): ... this.
11481 * config/rs6000/rs6000-builtin.def (MODS_V2DI, MODS_V4SI, MODU_V2DI,
11482 MODU_V4SI): Adjust.
11484 2021-07-08 Jeff Law <jeffreyalaw@gmail.com>
11486 * config/h8300/shiftrotate.md (variable shifts): Expose condition
11487 code handling for the test before the loop.
11489 2021-07-08 Martin Jambor <mjambor@suse.cz>
11492 * ipa-sra.c (class isra_call_summary): New member
11493 m_before_any_store, initialize it in the constructor.
11494 (isra_call_summary::dump): Dump the new field.
11495 (ipa_sra_call_summaries::duplicate): Copy it.
11496 (process_scan_results): Set it.
11497 (isra_write_edge_summary): Stream it.
11498 (isra_read_edge_summary): Likewise.
11499 (param_splitting_across_edge): Only override
11500 safe_to_import_accesses if m_before_any_store is set.
11502 2021-07-08 Martin Sebor <msebor@redhat.com>
11504 PR bootstrap/101374
11505 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref):
11506 Use Object Size Type 0 instead of 1.
11508 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
11510 * tree-vect-loop.c (vectorizable_reduction): Remove always-true
11513 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
11515 * match.pd: Simplify an extend-operate-truncate sequence involving
11518 2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
11519 Richard Biener <rguenther@suse.de>
11521 PR tree-optimization/40210
11522 * match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
11523 (x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
11524 when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
11526 2021-07-08 Uroš Bizjak <ubizjak@gmail.com>
11529 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
11531 * config/i386/mmx.md (V_32): New mode iterator.
11532 (mov<V_32:mode>): Use V_32 mode iterator.
11533 (*mov<V_32:mode>_internal): Ditto.
11534 (*push<V_32:mode>2_rex64): Ditto.
11535 (*push<V_32:mode>2): Ditto.
11536 (movmisalign<V_32:mode>): Ditto.
11537 (mmx_<any_shiftrt:insn>v1si3): New insn pattern.
11538 (sse4_1_<any_extend:code>v2qiv2hi2): Ditto.
11539 (vec_unpacks_lo_v4qi): New expander.
11540 (vec_unpacks_hi_v4qi): Ditto.
11541 (vec_unpacku_lo_v4qi): Ditto.
11542 (vec_unpacku_hi_v4qi): Ditto.
11543 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add V1SImode.
11544 (VALID_INT_MODE_P): Ditto.
11546 2021-07-08 Michael Meissner <meissner@linux.ibm.com>
11549 * config/rs6000/rs6000.md (udivti3): New insn.
11550 (divti3): New insn.
11551 (umodti3): New insn.
11552 (modti3): New insn.
11554 2021-07-07 Martin Sebor <msebor@redhat.com>
11556 PR tree-optimization/100137
11557 PR tree-optimization/99121
11558 PR tree-optimization/97027
11559 * builtins.c (access_ref::access_ref): Also set offmax.
11560 (access_ref::offset_in_range): Define new function.
11561 (access_ref::add_offset): Set offmax.
11562 (access_ref::inform_access): Handle access_none.
11563 (handle_mem_ref): Clear ostype.
11564 (compute_objsize_r): Handle ASSERT_EXPR.
11565 * builtins.h (struct access_ref): Add offmax member.
11566 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Use
11567 compute_objsize() and simplify.
11569 2021-07-07 Peter Bergner <bergner@linux.ibm.com>
11571 * config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
11572 and VSX_BUILTIN_STXVP.
11574 2021-07-07 Martin Sebor <msebor@redhat.com>
11577 * config/aarch64/aarch64.c (aarch64_simd_lane_bounds): Remove
11578 a stray %K from error_at() missed in r12-2088.
11580 2021-07-07 Richard Biener <rguenther@suse.de>
11582 PR tree-optimization/99728
11583 * tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
11585 (mem_refs_may_alias_p): Add assert we handled aggregate
11587 (sm_seq_valid_bb): Give up when running into aggregate copies.
11588 (ref_indep_loop_p): Handle aggregate copies as never
11589 being invariant themselves but allow other refs to be
11590 disambiguated against them.
11591 (can_sm_ref_p): Do not try to apply store-motion to aggregate
11594 2021-07-06 Indu Bhagat <indu.bhagat@oracle.com>
11597 * dwarf2ctf.c (ctf_get_AT_data_member_location): Multiply by 8 to get
11600 2021-07-06 Martin Sebor <msebor@redhat.com>
11602 * gimple-pretty-print.c (percent_G_format): Remove.
11603 * tree-diagnostic.c (default_tree_printer): Remove calls.
11604 * tree-pretty-print.c (percent_K_format): Remove.
11605 * tree-pretty-print.h (percent_K_format): Remove.
11607 2021-07-06 Martin Sebor <msebor@redhat.com>
11609 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_builtin):
11610 Remove %K and use error_at.
11611 (aarch64_expand_fcmla_builtin): Same.
11612 (aarch64_expand_builtin_tme): Same.
11613 (aarch64_expand_builtin_memtag): Same.
11614 * config/arm/arm-builtins.c (arm_expand_acle_builtin): Same.
11615 (arm_expand_builtin): Same.
11616 * config/arm/arm.c (bounds_check): Same.
11618 2021-07-06 Martin Sebor <msebor@redhat.com>
11620 * builtins.c (warn_string_no_nul): Remove %G.
11621 (maybe_warn_for_bound): Same.
11622 (warn_for_access): Same.
11623 (check_access): Same.
11624 (check_strncat_sizes): Same.
11625 (expand_builtin_strncat): Same.
11626 (expand_builtin_strncmp): Same.
11627 (expand_builtin): Same.
11628 (expand_builtin_object_size): Same.
11629 (warn_dealloc_offset): Same.
11630 (maybe_emit_free_warning): Same.
11631 * calls.c (maybe_warn_alloc_args_overflow): Same.
11632 (maybe_warn_nonstring_arg): Same.
11633 (maybe_warn_rdwr_sizes): Same.
11634 * expr.c (expand_expr_real_1): Remove %K.
11635 * gimple-fold.c (gimple_fold_builtin_strncpy): Remove %G.
11636 (gimple_fold_builtin_strncat): Same.
11637 * gimple-ssa-sprintf.c (format_directive): Same.
11638 (handle_printf_call): Same.
11639 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Same.
11640 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
11641 (maybe_diag_access_bounds): Same. Call gimple_location.
11642 (check_bounds_or_overlap): Same.
11643 * trans-mem.c (ipa_tm_scan_irr_block): Remove %K. Simplify.
11644 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Remove %G.
11645 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
11646 (maybe_diag_stxncpy_trunc): Same.
11647 (handle_builtin_stxncpy_strncat): Same.
11648 (maybe_warn_pointless_strcmp): Same.
11649 * tree-ssa-uninit.c (maybe_warn_operand): Same.
11651 2021-07-06 Uroš Bizjak <ubizjak@gmail.com>
11654 * config/i386/predicates.md (vec_setm_operand): Enable
11655 register_operand for TARGET_SSE4_1.
11656 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_operand
11657 as operand 2 predicate. Call ix86_expand_vector_set_var
11658 for non-constant index operand.
11659 (vec_setv4qi): Use vec_setm_mmx_operand as operand 2 predicate.
11660 Call ix86_expand_vector_set_var for non-constant index operand.
11662 2021-07-06 Jeff Law <jeffreyalaw@gmail.com>
11664 * config/h8300/jumpcall.md (*branch): When possible, generate
11665 the comparison in CCZN mode.
11666 * config/h8300/predicates.md (simple_memory_operand): Reject all
11667 auto-increment addressing modes.
11669 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
11671 PR bootstrap/100246
11672 * config/i386/i386.h (struct stringop_algs): Define a CTOR for
11675 2021-07-06 Richard Biener <rguenther@suse.de>
11677 * doc/md.texi (vec_fmaddsub<mode>4): Document.
11678 (vec_fmsubadd<mode>4): Likewise.
11679 * optabs.def (vec_fmaddsub$a4): Add.
11680 (vec_fmsubadd$a4): Likewise.
11681 * internal-fn.def (IFN_VEC_FMADDSUB): Add.
11682 (IFN_VEC_FMSUBADD): Likewise.
11683 * tree-vect-slp-patterns.c (addsub_pattern::recognize):
11684 Refactor to handle IFN_VEC_FMADDSUB and IFN_VEC_FMSUBADD.
11685 (addsub_pattern::build): Likewise.
11686 * tree-vect-slp.c (vect_optimize_slp): CFN_VEC_FMADDSUB
11687 and CFN_VEC_FMSUBADD are not transparent for permutes.
11688 * config/i386/sse.md (vec_fmaddsub<mode>4): New expander.
11689 (vec_fmsubadd<mode>4): Likewise.
11691 2021-07-06 Richard Biener <rguenther@suse.de>
11693 * doc/invoke.texi (fmove-loop-stores): Document.
11694 * common.opt (fmove-loop-stores): New option.
11695 * opts.c (default_options_table): Enable -fmove-loop-stores
11696 at -O1 but not -Og.
11697 * tree-ssa-loop-im.c (pass_lim::execute): Pass
11698 flag_move_loop_stores instead of true to
11699 loop_invariant_motion_in_fun.
11701 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
11703 * doc/install.texi: Document --with-dsymutil.
11705 2021-07-06 Andrew Pinski <apinski@marvell.com>
11707 PR tree-optimization/101256
11708 * dbgcnt.def (phiopt_edge_range): New counter.
11709 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
11710 Check to make sure the new name is defined in the same
11711 bb as the conditional before duplicating range info.
11712 Also add debug counter.
11714 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
11716 PR rtl-optimization/100328
11717 * config/i386/i386-options.c (ix86_option_override_internal):
11718 Set param_ira_consider_dup_in_all_alts to 0.
11720 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
11722 PR rtl-optimization/100328
11723 * doc/invoke.texi (ira-consider-dup-in-all-alts): Document new
11725 * ira.c (ira_get_dup_out_num): Adjust as parameter
11726 param_ira_consider_dup_in_all_alts.
11727 * params.opt (ira-consider-dup-in-all-alts): New.
11728 * ira-conflicts.c (process_regs_for_copy): Add one parameter
11729 single_input_op_has_cstr_p.
11730 (get_freq_for_shuffle_copy): New function.
11731 (add_insn_allocno_copies): Adjust as single_input_op_has_cstr_p.
11732 * ira-int.h (ira_get_dup_out_num): Add one bool parameter.
11734 2021-07-05 Jeff Law <jeffreyalaw@gmail.com>
11736 * config/h8300/shiftrotate.md (shift-by-variable patterns): Update to
11737 generate condition code aware RTL directly.
11739 2021-07-05 Andrew Pinski <apinski@marvell.com>
11741 PR tree-optimization/101039
11742 * match.pd (A CMP 0 ? A : -A): New patterns.
11743 * tree-ssa-phiopt.c (abs_replacement): Delete function.
11744 (tree_ssa_phiopt_worker): Don't call abs_replacement.
11745 Update comment about abs_replacement.
11747 2021-07-05 Andrew Pinski <apinski@marvell.com>
11749 * tree-ssa-phiopt.c (gimple_simplify_phiopt):
11750 If "A ? B : C" fails to simplify, try "(!A) ? C : B".
11752 2021-07-05 Andrew Pinski <apinski@marvell.com>
11754 * tree-ssa-phiopt.c (match_simplify_replacement):
11755 Add early_p argument. Call gimple_simplify_phiopt
11756 instead of gimple_simplify.
11757 (tree_ssa_phiopt_worker): Update call to
11758 match_simplify_replacement and allow unconditionally.
11759 (phiopt_early_allow): New function.
11760 (gimple_simplify_phiopt): New function.
11762 2021-07-05 Andrew Pinski <apinski@marvell.com>
11764 PR middle-end/101237
11765 * fold-const.c (negate_expr_p): Remove call to element_mode
11766 and TREE_MODE/TREE_TYPE when calling HONOR_SIGNED_ZEROS,
11767 HONOR_SIGN_DEPENDENT_ROUNDING, and HONOR_SNANS.
11768 (fold_negate_expr_1): Likewise.
11769 (const_unop): Likewise.
11770 (fold_cond_expr_with_comparison): Likewise.
11771 (fold_binary_loc): Likewise.
11772 (fold_ternary_loc): Likewise.
11773 (tree_call_nonnegative_warnv_p): Likewise.
11774 * match.pd (-(A + B) -> (-B) - A): Likewise.
11776 2021-07-05 Iain Sandoe <iain@sandoe.co.uk>
11778 * configure.ac: Handle --with-dsymutil in the same way as we
11779 do for the assembler and linker. (DEFAULT_DSYMUTIL): New.
11780 Extract the type and version for the dsymutil configured or
11781 found by the default searches.
11782 * config.in: Regenerated.
11783 * configure: Regenerated.
11784 * collect2.c (do_dsymutil): Handle locating dsymutil in the
11785 same way as for the assembler and linker.
11786 * config/darwin.h (DSYMUTIL): Delete.
11787 * gcc.c: Report a configured dsymutil correctly.
11788 * exec-tool.in: Allow for dsymutil.
11790 2021-07-05 Uroš Bizjak <ubizjak@gmail.com>
11792 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
11793 Handle V4QI and V2HI modes.
11794 (expand_vec_perm_blend): Allow 4-byte vector modes with TARGET_SSE4_1.
11795 Handle V4QI mode. Emit mmx_pblendvb32 for 4-byte modes.
11796 (expand_vec_perm_pshufb): Rewrite to use switch statemets.
11797 Handle 4-byte dual operands with TARGET_XOP and single operands
11798 with TARGET_SSSE3. Emit mmx_ppermv32 for TARGET_XOP and
11799 mmx_pshufbv4qi3 for TARGET_SSSE3.
11800 (expand_vec_perm_pblendv): Allow 4-byte vector modes with TARGET_SSE4_1.
11801 (expand_vec_perm_interleave2): Allow 4-byte vector modes.
11802 (expand_vec_perm_pshufb2): Allow 4-byte vector modes with TARGET_SSSE3.
11803 (expand_vec_perm_even_odd_1): Handle V4QI mode.
11804 (expand_vec_perm_broadcast_1): Handle V4QI mode.
11805 (ix86_vectorize_vec_perm_const): Handle V4QI mode.
11806 * config/i386/mmx.md (mmx_ppermv32): New insn pattern.
11807 (mmx_pshufbv4qi3): Ditto.
11808 (*mmx_pblendw32): Ditto.
11809 (*mmx_pblendw64): Rename from *mmx_pblendw.
11810 (mmx_punpckhbw_low): New insn_and_split pattern.
11811 (mmx_punpcklbw_low): Ditto.
11813 2021-07-05 Richard Biener <rguenther@suse.de>
11815 * tree-vect-loop-manip.c (vect_loop_versioning): Do not
11816 set LOOP_C_INFINITE on the vectorized loop.
11818 2021-07-05 Richard Biener <rguenther@suse.de>
11820 PR middle-end/101291
11821 * cfgloopmanip.c (loop_version): Set the loop copy of the
11822 versioned loop to the new loop.
11824 2021-07-04 Iain Sandoe <iain@sandoe.co.uk>
11827 * config.gcc: Ensure that Darwin biarch definitions are
11828 added before i386.h.
11829 * config/i386/darwin.h (TARGET_64BIT): Remove.
11830 (PR80556_WORKAROUND): New.
11831 (REAL_LIBGCC_SPEC): Amend to use PR80556_WORKAROUND.
11832 (DARWIN_SUBARCH_SPEC): New.
11833 * config/i386/darwin32-biarch.h (TARGET_64BIT_DEFAULT,
11834 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
11835 (REAL_LIBGCC_SPEC): Remove.
11836 * config/i386/darwin64-biarch.h (TARGET_64BIT_DEFAULT,
11837 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
11838 (REAL_LIBGCC_SPEC): Remove.
11840 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
11842 PR middle-end/101294
11843 * expr.c (store_constructor): Don't use vec_duplicate on vector.
11845 2021-07-02 Martin Sebor <msebor@redhat.com>
11847 PR middle-end/98871
11848 PR middle-end/98512
11849 * diagnostic.c (get_any_inlining_info): New.
11850 (update_effective_level_from_pragmas): Handle inlining context.
11851 (diagnostic_enabled): Same.
11852 (diagnostic_report_diagnostic): Same.
11853 * diagnostic.h (struct diagnostic_info): Add ctor.
11854 (struct diagnostic_context): Add new member.
11855 * tree-diagnostic.c (set_inlining_locations): New.
11856 (tree_diagnostics_defaults): Set new callback pointer.
11858 2021-07-02 Peter Bergner <bergner@linux.ibm.com>
11860 * config/rs6000/rs6000-builtin.def (BU_MMA_PAIR_LD, BU_MMA_PAIR_ST):
11862 (__builtin_vsx_lxvp, __builtin_vsx_stxvp): New built-ins.
11863 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Expand
11864 lxvp and stxvp built-ins.
11865 (mma_init_builtins): Handle lxvp and stxvp built-ins.
11866 (builtin_function_type): Likewise.
11867 * doc/extend.texi (__builtin_vsx_lxvp, __builtin_mma_stxvp): Document.
11869 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
11871 * config/h8300/h8300-protos.h (compute_a_shift_cc): Accept
11872 additional argument for the code.
11873 * config/h8300/h8300.c (compute_a_shift_cc): Accept additional
11874 argument for the code. Just return if the ZN bits are useful or
11875 not rather than the old style CC_* enums.
11876 * config/h8300/shiftrotate.md (shiftqi_noscratch): Move before
11877 more generic shiftqi patterns.
11878 (shifthi_noscratch, shiftsi_noscratch): Similarly.
11879 (shiftqi_noscratch_set_flags): New pattern.
11880 (shifthi_noscratch_set_flags, shiftsi_noscratch_set_flags): Likewise.
11882 2021-07-02 Andrew MacLeod <amacleod@redhat.com>
11884 PR tree-optimization/101223
11885 * range-op.cc (build_lt): Add -1 for signed values.
11886 (built_gt): Subtract -1 for signed values.
11888 2021-07-02 David Faust <david.faust@oracle.com>
11890 * btfout.c (get_btf_kind): Support BTF_KIND_FLOAT.
11891 (btf_asm_type): Likewise.
11893 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
11895 * config/h8300/h8300-protos.h (output_a_shift): Make first argument
11896 an array of rtx rather than a pointer to rtx. Add code argument.
11897 (compute_a_shift_length): Similarly.
11898 * config/h8300/h8300.c (h8300_shift_costs): Adjust now that the
11899 shift itself isn't an operand. Create dummy operand[0] to carry
11900 a mode and pass a suitable rtx code to compute_a_shift_length.
11901 (get_shift_alg): Adjust operand number of clobber in output templates.
11902 (output_a_shift): Make first argument an array of rtx rather than
11903 a pointer to rtx. Add code argument for the type of shift.
11904 Adjust now that the shift itself is no longer an operand.
11905 (compute_a_shift_length): Similarly.
11906 * config/h8300/shiftrotate.md (shiftqi, shifthi, shiftsi): Use an
11907 iterator rather than nshift_operator.
11908 (shiftqi_noscratch, shifthi_noscratch, shiftsi_noscratch): Likewise.
11909 (shiftqi_clobber_flags): Adjust to API changes in output_a_shift
11910 and compute_a_shift_length.
11911 (shiftqi_noscratch_clobber_flags): Likewise.
11912 (shifthi_noscratch_clobber_flags): Likewise.
11913 (shiftsi_noscratch_clobber_flags): Likewise.
11915 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
11918 * config/darwin.h (DSYMUTIL_SPEC): Do not try to run
11919 dsymutil for BTF/CTF.
11921 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
11924 * config/darwin.h (CTF_INFO_SECTION_NAME): Update the
11925 segment to include BTF.
11926 (BTF_INFO_SECTION_NAME): New.
11928 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
11930 * config/m32r/m32r-protos.h (call_operand): Adjust return type.
11931 (small_data_operand, memreg_operand, small_insn_p): Likewise.
11932 * config/m32r/m32r.c (call_operand): Adjust return type.
11933 (small_data_operand, memreg_operand): Likewise.
11935 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
11937 * config/frv/frv-protos.h (integer_register_operand): Adjust return
11939 (frv_load_operand, gpr_or_fpr_operand, gpr_no_subreg_operand): Likewise.
11940 (fpr_or_int6_operand, gpr_or_int_operand); Likewise.
11941 (gpr_or_int12_operand, gpr_or_int10_operand); Likewise.
11942 (move_source_operand, move_destination_operand): Likewise.
11943 (condexec_source_operand, condexec_dest_operand): Likewise.
11944 (lr_operand, gpr_or_memory_operand, fpr_or_memory_operand): Likewise.
11945 (reg_or_0_operand, fcc_operand, icc_operand, cc_operand): Likewise.
11946 (fcr_operand, icr_operand, cr_operand, call_operand): Likewise.
11947 (fpr_operand, even_reg_operand, odd_reg_operand): Likewise.
11948 (even_gpr_operand, odd_gpr_operand, quad_fpr_operand): Likewise.
11949 (even_fpr_operand, odd_fpr_operand): Likewise.
11950 (dbl_memory_one_insn_operand, dbl_memory_two_insn_operand): Likewise.
11951 (int12_operand, int6_operand, int5_operand, uint5_operand): Likewise.
11952 (uint4_operand, uint1_operand, int_2word_operand): Likewise
11953 (upper_int16_operand, uint16_operand, symbolic_operand): Likewise.
11954 (relational_operator, float_relational_operator): Likewise.
11955 (ccr_eqne_operator, minmax_operator): Likewise.
11956 (condexec_si_binary_operator, condexec_si_media_operator): Likewise.
11957 (condexec_si_divide_operator, condexec_si_unary_operator): Likewise.
11958 (condexec_sf_conv_operator, condexec_sf_add_operator): Likewise.
11959 (intop_compare_operator, acc_operand, even_acc_operand): Likewise.
11960 (quad_acc_operand, accg_operand): Likewise.
11962 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
11964 * config/stormy16/stormy16-protos.h (xstormy16_below_100_symbol): Change
11965 return type to a bool.
11966 (nonimmediate_nonstack_operand): Likewise.
11967 (xstormy16_splittable_below100_operand): Likewise.
11968 * config/stormy16/stormy16.c (xstormy16_below_100_symbol): Fix
11970 (xstormy16_splittable_below100_operand): Likewise.
11972 2021-07-02 Richard Biener <rguenther@suse.de>
11974 PR tree-optimization/101293
11975 * tree-ssa-loop-im.c (mem_ref_hasher::equal): Compare MEM_REF bases
11976 with combined offsets.
11977 (gather_mem_refs_stmt): Hash MEM_REFs as if their offset were
11978 combined with the rest of the offset.
11980 2021-07-02 Eric Botcazou <ebotcazou@adacore.com>
11982 * config/i386/i386.c (asm_preferred_eh_data_format): Always use the
11983 PIC encodings for PE-COFF targets.
11985 2021-07-02 Jakub Jelinek <jakub@redhat.com>
11988 * config/i386/i386-expand.c (ix86_broadcast_from_integer_constant):
11989 Return nullptr for TImode inner mode.
11991 2021-07-02 Richard Biener <rguenther@suse.de>
11993 PR tree-optimization/101280
11994 PR tree-optimization/101173
11995 * gimple-loop-interchange.cc
11996 (tree_loop_interchange::valid_data_dependences): Properly
11997 guard all dependence checks with DDR_REVERSED_P or its
12000 2021-07-02 Hongyu Wang <hongyu.wang@intel.com>
12002 * config/i386/i386-expand.c (ix86_expand_builtin):
12003 Add branch to clear odata when ZF is set for asedecenc_expand
12004 and wideaesdecenc_expand.
12006 2021-07-02 Eugene Rozenfeld <erozen@microsoft.com>
12008 * config/i386/gcc-auto-profile: regenerate
12010 2021-07-02 liuhongt <hongtao.liu@intel.com>
12012 * config/i386/sse.md (trunc<mode><pmov_dst_4>2): Refined to ..
12013 (trunc<mode><pmov_dst_4_lower>2): this.
12015 2021-07-01 David Malcolm <dmalcolm@redhat.com>
12017 * diagnostic.h (diagnostic_context::m_file_cache): New field.
12018 * input.c (class fcache): Rename to...
12019 (class file_cache_slot): ...this, making most members private and
12020 prefixing fields with "m_".
12021 (file_cache_slot::get_file_path): New accessor.
12022 (file_cache_slot::get_use_count): New accessor.
12023 (file_cache_slot::missing_trailing_newline_p): New accessor.
12024 (file_cache_slot::inc_use_count): New.
12025 (fcache_buffer_size): Move to...
12026 (file_cache_slot::buffer_size): ...here.
12027 (fcache_line_record_size): Move to...
12028 (file_cache_slot::line_record_size): ...here.
12029 (fcache_tab): Delete, in favor of global_dc->m_file_cache.
12030 (fcache_tab_size): Move to file_cache::num_file_slots.
12031 (diagnostic_file_cache_init): Update for move of fcache_tab
12032 to global_dc->m_file_cache.
12033 (diagnostic_file_cache_fini): Likewise.
12034 (lookup_file_in_cache_tab): Convert to...
12035 (file_cache::lookup_file): ...this.
12036 (diagnostics_file_cache_forcibly_evict_file): Update for move of
12037 fcache_tab to global_dc->m_file_cache, moving most of
12038 implementation to...
12039 (file_cache::forcibly_evict_file): ...this new function and...
12040 (file_cache_slot::evict): ...this new function.
12041 (evicted_cache_tab_entry): Convert to...
12042 (file_cache::evicted_cache_tab_entry): ...this.
12043 (add_file_to_cache_tab): Convert to...
12044 (file_cache::add_file): ...this, moving bulk of implementation
12046 (file_cache_slot::create): ..this new function.
12047 (file_cache::file_cache): New.
12048 (file_cache::~file_cache): New.
12049 (lookup_or_add_file_to_cache_tab): Convert to...
12050 (file_cache::lookup_or_add_file): ..this new function.
12051 (fcache::fcache): Rename to...
12052 (file_cache_slot::file_cache_slot): ...this, adding "m_" prefixes
12054 (fcache::~fcache): Rename to...
12055 (file_cache_slot::~file_cache_slot): ...this, adding "m_" prefixes
12057 (needs_read): Convert to...
12058 (file_cache_slot::needs_read_p): ...this.
12059 (needs_grow): Convert to...
12060 (file_cache_slot::needs_grow_p): ...this.
12061 (maybe_grow): Convert to...
12062 (file_cache_slot::maybe_grow): ...this.
12063 (read_data): Convert to...
12064 (file_cache_slot::read_data): ...this.
12065 (maybe_read_data): Convert to...
12066 (file_cache_slot::maybe_read_data): ...this.
12067 (get_next_line): Convert to...
12068 (file_cache_slot::get_next_line): ...this.
12069 (goto_next_line): Convert to...
12070 (file_cache_slot::goto_next_line): ...this.
12071 (read_line_num): Convert to...
12072 (file_cache_slot::read_line_num): ...this.
12073 (location_get_source_line): Update for moving of globals to
12074 global_dc->m_file_cache.
12075 (location_missing_trailing_newline): Likewise.
12076 * input.h (class file_cache_slot): New forward decl.
12077 (class file_cache): New.
12079 2021-07-01 Michael Meissner <meissner@linux.ibm.com>
12081 * config/rs6000/rs6000.c (rs6000_maybe_emit_fp_cmove): Add IEEE
12082 128-bit floating point conditional move support.
12083 (have_compare_and_set_mask): Add IEEE 128-bit floating point
12085 * config/rs6000/rs6000.md (mov<mode>cc, IEEE128 iterator): New insn.
12086 (mov<mode>cc_p10, IEEE128 iterator): New insn.
12087 (mov<mode>cc_invert_p10, IEEE128 iterator): New insn.
12088 (fpmask<mode>, IEEE128 iterator): New insn.
12089 (xxsel<mode>, IEEE128 iterator): New insn.
12091 2021-07-01 Iain Sandoe <iain@sandoe.co.uk>
12094 * config/darwin.h (CTF_INFO_SECTION_NAME): New.
12096 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
12098 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
12100 * config/i386/i386-protos.h (ix86_expand_vector_init_duplicate):
12102 * config/i386/sse.md (INT_BROADCAST_MODE): New mode iterator.
12103 (vec_duplicate<mode>): New expander.
12105 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
12108 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
12110 (ix86_byte_broadcast): New function.
12111 (ix86_convert_const_wide_int_to_broadcast): Likewise.
12112 (ix86_expand_move): Convert CONST_WIDE_INT to broadcast if mode
12113 size is 16 bytes or bigger.
12114 (ix86_broadcast_from_integer_constant): New function.
12115 (ix86_expand_vector_move): Convert CONST_WIDE_INT and CONST_VECTOR
12116 to broadcast if mode size is 16 bytes or bigger.
12117 * config/i386/i386-protos.h (ix86_gen_scratch_sse_rtx): New
12119 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): New function.
12121 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
12123 * config/i386/predicates.md (ix86_endbr_immediate_operand):
12124 Return true/false instead of 1/0.
12125 (movq_parallel): Ditto.
12127 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
12129 * recog.c (general_operand): Return true/false instead of 1/0.
12130 (register_operand): Ditto.
12131 (immediate_operand): Ditto.
12132 (const_int_operand): Ditto.
12133 (const_scalar_int_operand): Ditto.
12134 (const_double_operand): Ditto.
12135 (push_operand): Ditto.
12136 (pop_operand): Ditto.
12137 (memory_operand): Ditto.
12138 (indirect_operand): Ditto.
12140 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
12142 * genpreds.c (write_predicate_subfunction):
12143 Change the type of written subfunction to bool.
12144 (write_one_predicate_function):
12145 Change the type of written function to bool.
12146 (write_tm_preds_h): Ditto.
12147 * recog.h (*insn_operand_predicate_fn): Change the type to bool.
12148 * recog.c (general_operand): Change the type to bool.
12149 (address_operand): Ditto.
12150 (register_operand): Ditto.
12151 (pmode_register_operand): Ditto.
12152 (scratch_operand): Ditto.
12153 (immediate_operand): Ditto.
12154 (const_int_operand): Ditto.
12155 (const_scalar_int_operand): Ditto.
12156 (const_double_operand): Ditto.
12157 (nonimmediate_operand): Ditto.
12158 (nonmemory_operand): Ditto.
12159 (push_operand): Ditto.
12160 (pop_operand): Ditto.
12161 (memory_operand): Ditto.
12162 (indirect_operand): Ditto.
12163 (ordered_comparison_operator): Ditto.
12164 (comparison_operator): Ditto.
12165 * config/i386/i386-expand.c (ix86_expand_sse_cmp):
12166 Change the type of indirect predicate function to bool.
12167 * config/rs6000/rs6000.c (easy_vector_constant):
12168 Change the type to bool.
12169 * config/mips/mips-protos.h (m16_based_address_p):
12170 Change the type of operand 3 to bool.
12172 2021-07-01 Richard Biener <rguenther@suse.de>
12174 PR tree-optimization/101280
12175 PR tree-optimization/101173
12176 * gimple-loop-interchange.cc
12177 (tree_loop_interchange::valid_data_dependences): Revert
12178 previous change and instead correctly handle DDR_REVERSED_P
12181 2021-07-01 Richard Biener <rguenther@suse.de>
12183 PR tree-optimization/101278
12184 * tree-ssa-dse.c (dse_classify_store): First check for
12185 uses, then ignore stmt for chaining purposes.
12187 2021-07-01 Richard Biener <rguenther@suse.de>
12189 PR tree-optimization/100778
12190 * tree-vect-slp.c (vect_schedule_slp_node): Do not place trapping
12191 vectorized ops ahead of their scalar BB.
12193 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
12196 * config/i386/i386.md (*nabs<dwi>2_doubleword):
12197 New insn_and_split pattern.
12198 (*nabs<dwi>2_1): Ditto.
12199 * config/i386/i386-features.c
12200 (general_scalar_chain::compute_convert_gain):
12201 Handle (NEG (ABS (...))) RTX. Rewrite src code
12202 scanner as switch statement.
12203 (general_scalar_chain::convert_insn):
12204 Handle (NEG (ABS (...))) RTX.
12205 (general_scalar_to_vector_candidate_p):
12206 Detect (NEG (ABS (...))) RTX. Reorder case statements
12207 for (AND (NOT (...) ...)) fallthrough.
12209 2021-07-01 Richard Biener <rguenther@suse.de>
12211 PR tree-optimization/101178
12212 * tree-vect-slp.c (slpg_vertex::materialize): Remove.
12213 (slpg::perm_in): Add.
12214 (slpg::get_perm_in): Remove.
12215 (slpg::get_perm_materialized): Add.
12216 (vect_optimize_slp): Handle VEC_PERM nodes more optimally
12217 during permute propagation and materialization.
12219 2021-07-01 Jakub Jelinek <jakub@redhat.com>
12222 * dwarf2out.c (loc_list_from_tree_1): Handle COMPOUND_LITERAL_EXPR.
12224 2021-07-01 Jakub Jelinek <jakub@redhat.com>
12226 PR middle-end/94366
12227 * omp-low.c (lower_rec_input_clauses): Rename is_fp_and_or to
12228 is_truth_op, set it for TRUTH_*IF_EXPR regardless of new_var's type,
12229 use boolean_type_node instead of integer_type_node as NE_EXPR type.
12230 (lower_reduction_clauses): Likewise.
12232 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
12234 * config/gcn/gcn.c: Include dwarf2.h.
12235 (gcn_addr_space_debug): New function.
12236 (TARGET_ADDR_SPACE_DEBUG): New hook.
12238 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
12240 * common/config/gcn/gcn-common.c
12241 (gcn_option_optimization_table): Change OPT_fomit_frame_pointer to -O3.
12242 * config/gcn/gcn.c (gcn_expand_prologue): Prefer the frame pointer
12244 (gcn_expand_prologue): Prefer the frame pointer when emitting CFI.
12245 (gcn_frame_pointer_rqd): New function.
12246 (TARGET_FRAME_POINTER_REQUIRED): New hook.
12248 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
12250 * config/gcn/gcn.c (move_callee_saved_registers): Emit CFI notes for
12251 prologue register saves.
12252 (gcn_debug_unwind_info): Use UI_DWARF2.
12253 (gcn_dwarf_register_number): Map DWARF_LINK_REGISTER to DWARF PC.
12254 (gcn_dwarf_register_span): DWARF_LINK_REGISTER doesn't span.
12255 * config/gcn/gcn.h: (DWARF_FRAME_RETURN_COLUMN): New define.
12256 (DWARF_LINK_REGISTER): New define.
12257 (FIRST_PSEUDO_REGISTER): Increment.
12258 (FIXED_REGISTERS): Add entry for DWARF_LINK_REGISTER.
12259 (CALL_USED_REGISTERS): Likewise.
12260 (REGISTER_NAMES): Likewise.
12262 2021-06-30 Richard Biener <rguenther@suse.de>
12264 PR tree-optimization/101267
12265 * tree-vect-stmts.c (vect_check_scalar_mask): Adjust
12266 API and use SLP compatible interface of vect_is_simple_use.
12267 Reject not vectorized SLP defs for callers that do not support
12269 (vect_check_store_rhs): Handle masked stores and pass down
12270 the appropriate operator index.
12271 (vectorizable_call): Adjust.
12272 (vectorizable_store): Likewise.
12273 (vectorizable_load): Likewise. Handle SLP pecularity of
12275 (vect_is_simple_use): Remove special-casing of masked stores.
12277 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
12279 * common.opt (foffload): Remove help as Driver only.
12280 * gcc.c (display_help): Add -foffload.
12282 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
12284 * gcc.c (close_at_file, execute): Replace alloca by XALLOCAVEC.
12285 (check_offload_target_name): Fix splitting OFFLOAD_TARGETS into
12286 a candidate list; better inform no offload target is configured
12287 and fix hint extraction when passed target is not '\0' at [len].
12288 * common.opt (foffload): Add tailing '.'.
12289 (foffload-options): Likewise; fix flag name in the help string.
12291 2021-06-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
12294 * config/arm/arm_neon.h: Move vabs intrinsics before vcage_f32.
12295 (vcage_f32): Gate comparison on __FAST_MATH__.
12296 (vcageq_f32): Likewise.
12297 (vcale_f32): Likewise.
12298 (vcaleq_f32): Likewise.
12299 (vcagt_f32): Likewise.
12300 (vcagtq_f32): Likewise.
12301 (vcalt_f32): Likewise.
12302 (vcaltq_f32): Likewise.
12303 (vcage_f16): Likewise.
12304 (vcageq_f16): Likewise.
12305 (vcale_f16): Likewise.
12306 (vcaleq_f16): Likewise.
12307 (vcagt_f16): Likewise.
12308 (vcagtq_f16): Likewise.
12309 (vcalt_f16): Likewise.
12310 (vcaltq_f16): Likewise.
12312 2021-06-30 Richard Biener <rguenther@suse.de>
12314 PR tree-optimization/101264
12315 * tree-vect-slp.c (vect_optimize_slp): Propagate the
12316 computed perm_in to all "any" permute successors
12317 we cannot de-duplicate immediately.
12319 2021-06-30 liuhongt <hongtao.liu@intel.com>
12322 * config/i386/sse.md
12323 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>):
12325 (avx512f_sfixupimm<mode><maskz_scalar_name><round_saeonly_name>):
12327 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>"): Refined.
12328 * config/i386/subst.md (maskz_scalar): New define_subst.
12329 (maskz_scalar_name): New subst_attr.
12330 (maskz_scalar_op5): Ditto.
12331 (round_saeonly_maskz_scalar_op5): Ditto.
12332 (round_saeonly_maskz_scalar_operand5): Ditto.
12334 2021-06-30 David Edelsohn <dje.gcc@gmail.com>
12336 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
12337 Increase code CSECT alignment to at least 32 bytes.
12338 * config/rs6000/xcoff.h (TEXT_SECTION_ASM_OP): Add 32 byte
12339 alignment designation.
12341 2021-06-29 Sergei Trofimovich <siarheit@google.com>
12343 * doc/generic.texi: Fix s/net yet/not yet/ typo.
12345 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
12347 PR tree-optimization/101254
12348 * range-op.cc (operator_minus::op1_op2_relation_effect): Check for
12349 wrapping/non-wrapping when setting the result range.
12351 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
12353 * value-query.cc (gimple_range_global): Allow phis.
12355 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
12357 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): Use stmt.
12358 (simplify_using_ranges::op_with_boolean_value_range_p): Add a
12359 statement for location context.
12360 (check_for_binary_op_overflow): Ditto.
12361 (simplify_using_ranges::get_vr_for_comparison): Ditto.
12362 (simplify_using_ranges::compare_name_with_value): Ditto.
12363 (simplify_using_ranges::compare_names): Ditto.
12364 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Ditto.
12365 (simplify_using_ranges::simplify_truth_ops_using_ranges): Ditto.
12366 (simplify_using_ranges::simplify_min_or_max_using_ranges): Ditto.
12367 (simplify_using_ranges::simplify_internal_call_using_ranges): Ditto.
12368 (simplify_using_ranges::two_valued_val_range_p): Ditto.
12369 (simplify_using_ranges::simplify): Ditto.
12370 * vr-values.h: Adjust prototypes.
12372 2021-06-29 Uroš Bizjak <ubizjak@gmail.com>
12375 * config/i386/mmx.md (vec_addsubv2sf3): New insn pattern.
12377 2021-06-29 Julian Brown <julian@codesourcery.com>
12379 * config/gcn/gcn.c (gcn_init_libfuncs): New function.
12380 (TARGET_INIT_LIBFUNCS): Define target hook using above function.
12381 * config/gcn/gcn.h (UNITS_PER_WORD): Define to 8 for IN_LIBGCC2, 4
12383 (LIBGCC2_UNITS_PER_WORD, BITS_PER_WORD): Remove definitions.
12384 (MAX_FIXED_MODE_SIZE): Change to 128.
12386 2021-06-29 Julian Brown <julian@codesourcery.com>
12388 * config/gcn/gcn.md (UNSPEC_FLBIT_INT): New unspec constant.
12389 (s_mnemonic): Add clrsb.
12390 (gcn_flbit<mode>_int): Add insn pattern for SImode/DImode.
12391 (clrsb<mode>2): Add expander for SImode/DImode.
12393 2021-06-29 Julian Brown <julian@codesourcery.com>
12395 * config/gcn/gcn.md (<su>mulsidi3, <su>mulsidi3_reg, <su>mulsidi3_imm,
12396 muldi3): Add patterns.
12398 2021-06-29 Julian Brown <julian@codesourcery.com>
12400 * config/gcn/gcn.md (<su>mulsi3_highpart): Change to expander.
12401 (<su>mulsi3_highpart_reg, <su>mulsi3_highpart_imm): New patterns.
12403 2021-06-29 Julian Brown <julian@codesourcery.com>
12405 * config/gcn/gcn.md (mulsi3): Make s_mulk_i32 variant clobber SCC.
12407 2021-06-29 Joseph Myers <joseph@codesourcery.com>
12409 * btfout.c, ctfout.c: Include "memmodel.h".
12411 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
12413 * gcc.c (check_offload_target_name): Cast len argument to
12414 %q.*s to 'int'; avoid -Wstringop-truncation warning.
12416 2021-06-29 Richard Biener <rguenther@suse.de>
12418 * tree-vect-slp.c (vect_optimize_slp): Forward propagate
12419 to "any" permute nodes and relax "any" permute proapgation
12420 during iterative backward propagation.
12422 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
12425 * common.opt (-foffload=): Update description.
12426 (-foffload-options=): New.
12427 * doc/invoke.texi (C Language Options): Document
12428 -foffload and -foffload-options.
12429 * gcc.c (check_offload_target_name): New, split off from
12430 handle_foffload_option.
12431 (check_foffload_target_names): New.
12432 (handle_foffload_option): Handle -foffload=default.
12433 (driver_handle_option): Update for -foffload-options.
12434 * lto-opts.c (lto_write_options): Use -foffload-options
12435 instead of -foffload.
12436 * lto-wrapper.c (merge_and_complain, append_offload_options):
12438 * opts.c (common_handle_option): Likewise.
12440 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
12442 * doc/invoke.texi (C Language Options): Sort options
12443 alphabetically in optlist and also the description itself.
12444 Remove leftover -fallow-single-precision from and add missing
12445 -fgnu-tm to the optlist.
12447 2021-06-29 Richard Biener <rguenther@suse.de>
12449 * tree-vect-slp.c (slpg_vertex::visited): Remove.
12450 (vect_slp_perms_eq): Handle -1 permutes.
12451 (vect_optimize_slp): Rewrite permute propagation.
12453 2021-06-29 Jakub Jelinek <jakub@redhat.com>
12456 * match.pd ((intptr_t)x eq/ne CST to x eq/ne (typeof x) CST): Don't
12457 perform the optimization in GENERIC when sanitizing and x has a
12460 2021-06-29 Richard Biener <rguenther@suse.de>
12462 PR tree-optimization/101242
12463 * tree-vect-slp.c (vect_slp_build_vertices): Force-add
12464 PHIs with not represented initial values as leafs.
12466 2021-06-29 Jan-Benedict Glaw <jbglaw@getslash.de>
12468 * config/pdp11/pdp11.h (ASM_OUTPUT_SKIP): Fix signedness warning.
12469 * config/pdp11/pdp11.c (pdp11_asm_print_operand_punct_valid_p): Remove
12470 "register" keyword.
12471 (pdp11_initial_elimination_offset) Remove unused variable.
12472 (pdp11_cmp_length) Ditto.
12473 (pdp11_insn_cost): Ditto, and fix signedness warning.
12475 2021-06-29 David Edelsohn <dje.gcc@gmail.com>
12477 * btfout.c: Include tm_p.h.
12480 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
12482 * config/bpf/bpf.c (bpf_expand_prologue): Do not mark insns as
12484 (bpf_expand_epilogue): Likewise.
12485 * config/bpf/bpf.h (DWARF2_FRAME_INFO): Define to 0.
12486 Do not define DBX_DEBUGGING_INFO.
12488 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
12490 * doc/invoke.texi: Document the CTF and BTF debug info options.
12492 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
12493 David Faust <david.faust@oracle.com>
12494 Jose E. Marchesi <jose.marchesi@oracle.com>
12495 Weimin Pan <weimin.pan@oracle.com>
12497 * Makefile.in: Add ctfc.*, ctfout.c and btfout.c files to
12498 GTFILES. Add new object files.
12499 * common.opt: Add CTF and BTF debug info options.
12500 * btfout.c: New file.
12501 * ctfc.c: Likewise.
12502 * ctfc.h: Likewise.
12503 * ctfout.c: Likewise.
12504 * dwarf2ctf.c: Likewise.
12505 * dwarf2ctf.h: Likewise.
12506 * dwarf2cfi.c (dwarf2out_do_frame): Acknowledge CTF_DEBUG and
12508 * dwarf2out.c (dwarf2out_source_line): Likewise.
12509 (dwarf2out_finish): Skip emitting DWARF if CTF or BTF are to
12511 (debug_format_do_cu): New function.
12512 (dwarf2out_early_finish): Traverse DIEs and emit CTF/BTF for
12514 Include dwarf2ctf.c.
12515 * final.c (dwarf2_debug_info_emitted_p): Acknowledge DWARF-based debug
12517 * flag-types.h (enum debug_info_type): Add CTF_DEBUG and BTF_DEBUG.
12518 (CTF_DEBUG): New bitmask.
12519 (BTF_DEBUG): Likewise.
12520 (enum ctf_debug_info_levels): New enum.
12521 * gengtype.c (open_base_files): Handle ctfc.h.
12522 (main): Handle uint32_t type.
12523 * flags.h (btf_debuginfo_p): New definition.
12524 (dwarf_based_debuginfo_p): Likewise.
12525 * opts.c (debug_type_names): Add entries for CTF and BTF.
12526 (btf_debuginfo_p): New function.
12527 (dwarf_based_debuginfo_p): Likewise.
12528 (common_handle_option): Handle -gctfN and -gbtf options.
12529 (set_debug_level): Set CTF_DEBUG, BTF_DEBUG whenever appropriate.
12530 * toplev.c (process_options): Inform the user and ignore -gctfLEVEL if
12533 2021-06-28 Jose E. Marchesi <jose.marchesi@oracle.com>
12535 * dwarf2out.c (AT_class): Function is no longer static.
12536 (AT_int): Likewise.
12537 (AT_unsigned): Likewise.
12538 (AT_loc): Likewise.
12539 (get_AT): Likewise.
12540 (get_AT_string): Likewise.
12541 (get_AT_flag): Likewise.
12542 (get_AT_unsigned): Likewise.
12543 (get_AT_ref): Likewise.
12544 (new_die_raw): Likewise.
12545 (lookup_decl_die): Likewise.
12546 (base_type_die): Likewise.
12547 (add_name_attribute): Likewise.
12548 (add_AT_int): Likewise.
12549 (add_AT_unsigned): Likewise.
12550 (add_AT_loc): Likewise.
12551 (dw_get_die_tag): New function.
12552 (dw_get_die_child): Likewise.
12553 (dw_get_die_sib): Likewise.
12554 (struct dwarf_file_data): Move from here to dwarf2out.h
12555 (struct dw_attr_struct): Likewise.
12556 * dwarf2out.h: Analogous changes.
12558 2021-06-28 Martin Jambor <mjambor@suse.cz>
12561 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
12562 members m_dead_stmts and m_dead_ssas.
12563 * ipa-param-manipulation.c
12564 (ipa_param_body_adjustments::mark_dead_statements): New function.
12565 (ipa_param_body_adjustments::common_initialization): Call it on
12566 all removed but not split parameters.
12567 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
12569 (ipa_param_body_adjustments::modify_call_stmt): Remove arguments that
12571 * tree-inline.c (remap_gimple_stmt): Do not copy dead statements, reset
12572 dead debug statements.
12573 (copy_phis_for_bb): Do not copy dead PHI nodes.
12575 2021-06-28 Martin Jambor <mjambor@suse.cz>
12578 * symtab-clones.h (clone_info): Removed member param_adjustments.
12579 * ipa-param-manipulation.h: Adjust initial comment to reflect how we
12580 deal with pass-through splits now.
12581 (ipa_param_performed_split): Removed.
12582 (ipa_param_adjustments::modify_call): Adjusted parameters.
12583 (class ipa_param_body_adjustments): Adjusted parameters of
12584 register_replacement, modify_gimple_stmt and modify_call_stmt.
12585 (ipa_verify_edge_has_no_modifications): Declare.
12586 (ipa_edge_modifications_finalize): Declare.
12587 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Remove
12588 performed_splits processing, pas only edge to padjs->modify_call,
12589 check that call arguments were not modified if they should not have
12591 * cgraphclones.c (cgraph_node::create_clone): Do not copy performed
12593 * ipa-param-manipulation.c (struct pass_through_split_map): New type.
12594 (ipa_edge_modification_info): Likewise.
12595 (ipa_edge_modification_sum): Likewise.
12596 (ipa_edge_modifications): New edge summary.
12597 (ipa_verify_edge_has_no_modifications): New function.
12598 (transitive_split_p): Removed.
12599 (transitive_split_map): Likewise.
12600 (init_transitive_splits): Likewise.
12601 (ipa_param_adjustments::modify_call): Adjusted to use the new edge
12602 summary instead of performed_splits.
12603 (ipa_param_body_adjustments::register_replacement): Drop dummy
12604 parameter, set base_index of the created ipa_param_body_replacement.
12605 (phi_arg_will_live_p): New function.
12606 (ipa_param_body_adjustments::common_initialization): Do not create
12607 IPA_SRA dummy decls.
12608 (simple_tree_swap_info): Removed.
12609 (remap_split_decl_to_dummy): Likewise.
12610 (record_argument_state_1): New function.
12611 (record_argument_state): Likewise.
12612 (ipa_param_body_adjustments::modify_call_stmt): New parameter
12613 orig_stmt. Do not work with dummy decls, save necessary info about
12614 changes to ipa_edge_modifications.
12615 (ipa_param_body_adjustments::modify_gimple_stmt): New parameter
12616 orig_stmt, pass it to modify_call_stmt.
12617 (ipa_param_body_adjustments::modify_cfun_body): Adjust call to
12618 modify_gimple_stmt.
12619 (ipa_edge_modifications_finalize): New function.
12620 * tree-inline.c (remap_gimple_stmt): Pass original statement to
12621 modify_gimple_stmt.
12622 (copy_phis_for_bb): Do not copy dead PHI nodes.
12623 (expand_call_inline): Do not remap performed_splits.
12624 (update_clone_info): Likewise.
12625 * toplev.c: Include ipa-param-manipulation.h.
12626 (toplev::finalize): Call ipa_edge_modifications_finalize.
12628 2021-06-28 Andrew Pinski <apinski@marvell.com>
12630 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Duplicate range
12631 info if we're the only things setting the target PHI.
12632 (value_replacement): Don't duplicate range here.
12633 (minmax_replacement): Likewise.
12635 2021-06-28 Richard Biener <rguenther@suse.de>
12637 PR tree-optimization/101229
12638 * gimple-walk.c (gimple_walk_op): Handle PHIs.
12640 2021-06-28 Martin Liska <mliska@suse.cz>
12642 * config/v850/v850.c (construct_dispose_instruction): Allocate
12644 (construct_prepare_instruction): Likewise.
12646 2021-06-28 Martin Liska <mliska@suse.cz>
12648 * config/v850/v850.c (v850_option_override): Build default
12650 (v850_can_inline_p): New. Allow MASK_PROLOG_FUNCTION to be
12651 ignored for inlining.
12652 (TARGET_CAN_INLINE_P): New.
12654 2021-06-28 Richard Biener <rguenther@suse.de>
12656 PR tree-optimization/101207
12657 * tree-vect-slp.c (vect_optimize_slp): Do BB reduction
12658 permute eliding for load permutations properly.
12660 2021-06-28 Richard Biener <rguenther@suse.de>
12662 PR tree-optimization/101173
12663 * gimple-loop-interchange.cc
12664 (tree_loop_interchange::valid_data_dependences): Disallow outer
12665 loop dependence distance of zero.
12667 2021-06-28 liuhongt <hongtao.liu@intel.com>
12670 * config/i386/sse.md (*avx_cmp<mode>3_lt): New
12671 define_insn_and_split.
12672 (*avx_cmp<mode>3_ltint): Ditto.
12673 (*avx2_pcmp<mode>3_3): Ditto.
12674 (*avx2_pcmp<mode>3_4): Ditto.
12675 (*avx2_pcmp<mode>3_5): Ditto.
12677 2021-06-28 liuhongt <hongtao.liu@intel.com>
12679 * config/i386/i386-builtin.def (IX86_BUILTIN_BLENDVPD256,
12680 IX86_BUILTIN_BLENDVPS256, IX86_BUILTIN_PBLENDVB256,
12681 IX86_BUILTIN_BLENDVPD, IX86_BUILTIN_BLENDVPS,
12682 IX86_BUILTIN_PBLENDVB128): Replace icode with
12684 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold blendv
12686 * config/i386/sse.md (*<sse4_1_avx2>_pblendvb_lt_subreg_not):
12687 New pre_reload splitter.
12689 2021-06-27 Andrew Pinski <apinski@marvell.com>
12691 PR middle-end/101230
12692 * fold-const.c (fold_ternary_loc): Check
12693 the return value of invert_tree_comparison.
12695 2021-06-27 David Edelsohn <dje.gcc@gmail.com>
12697 * config.gcc: Add SPDX License Identifier.
12698 (powerpc-ibm-aix789): Default to aix73.h.
12699 (powerpc-ibm-aix7.2.*.*): New stanza.
12700 * config/rs6000/aix72.h: Add SPDX License Identifier.
12701 * config/rs6000/aix73.h: New file.
12703 2021-06-26 Jason Merrill <jason@redhat.com>
12705 * except.c: #include "dwarf2.h" instead of "dwarf2out.h".
12707 2021-06-26 Andrew Pinski <apinski@marvell.com>
12709 * genmatch.c (lower_cond): Copy for_subst_vec
12710 for the simplify also.
12711 (lower): Swap the order for lower_for and lower_cond.
12713 2021-06-26 Andrew Pinski <apinski@marvell.com>
12715 * tree-ssa-phiopt.c (match_simplify_replacement): Reset
12716 flow senatitive info on the moved ssa set.
12718 2021-06-26 Andrew Pinski <apinski@marvell.com>
12720 * fold-const.c (fold_cond_expr_with_comparison):
12721 Exand arg0 into comp_code, arg00, and arg01.
12722 (fold_ternary_loc): Use invert_tree_comparison
12723 instead of fold_invert_truthvalue for the case
12724 where we have A CMP B ? C : A.
12726 2021-06-25 Martin Sebor <msebor@redhat.com>
12728 PR middle-end/101216
12729 * calls.c (maybe_warn_rdwr_sizes): Use the no_warning constant.
12731 2021-06-25 Jeff Law <jeffreyalaw@gmail.com>
12733 * config/h8300/h8300.c (select_cc_mode): Handle ASHIFTRT and LSHIFTRT.
12735 2021-06-25 Richard Biener <rguenther@suse.de>
12737 PR tree-optimization/101202
12738 * tree-vect-slp.c (vect_optimize_slp): Explicitely handle
12741 2021-06-25 Richard Biener <rguenther@suse.de>
12743 * tree-vect-slp-patterns.c (addsub_pattern::build): Copy
12744 STMT_VINFO_REDUC_DEF from the original representative.
12746 2021-06-25 Martin Sebor <msebor@redhat.com>
12748 * builtins.c (warn_string_no_nul): Replace uses of TREE_NO_WARNING,
12749 gimple_no_warning_p and gimple_set_no_warning with
12750 warning_suppressed_p, and suppress_warning.
12752 (maybe_warn_for_bound): Same.
12753 (warn_for_access): Same.
12754 (check_access): Same.
12755 (expand_builtin_strncmp): Same.
12756 (fold_builtin_varargs): Same.
12757 * calls.c (maybe_warn_nonstring_arg): Same.
12758 (maybe_warn_rdwr_sizes): Same.
12759 * cfgexpand.c (expand_call_stmt): Same.
12760 * cgraphunit.c (check_global_declaration): Same.
12761 * fold-const.c (fold_undefer_overflow_warnings): Same.
12762 (fold_truth_not_expr): Same.
12763 (fold_unary_loc): Same.
12764 (fold_checksum_tree): Same.
12765 * gimple-array-bounds.cc (array_bounds_checker::check_array_ref): Same.
12766 (array_bounds_checker::check_mem_ref): Same.
12767 (array_bounds_checker::check_addr_expr): Same.
12768 (array_bounds_checker::check_array_bounds): Same.
12769 * gimple-expr.c (copy_var_decl): Same.
12770 * gimple-fold.c (gimple_fold_builtin_strcpy): Same.
12771 (gimple_fold_builtin_strncat): Same.
12772 (gimple_fold_builtin_stxcpy_chk): Same.
12773 (gimple_fold_builtin_stpcpy): Same.
12774 (gimple_fold_builtin_sprintf): Same.
12775 (fold_stmt_1): Same.
12776 * gimple-ssa-isolate-paths.c (diag_returned_locals): Same.
12777 * gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Same.
12778 * gimple-ssa-sprintf.c (handle_printf_call): Same.
12779 * gimple-ssa-store-merging.c (imm_store_chain_info::output_merged_store): Same.
12780 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
12781 * gimple-ssa-warn-restrict.h: Adjust declarations.
12782 (maybe_diag_access_bounds): Replace uses of TREE_NO_WARNING,
12783 gimple_no_warning_p and gimple_set_no_warning with
12784 warning_suppressed_p, and suppress_warning.
12785 (check_call): Same.
12786 (check_bounds_or_overlap): Same.
12787 * gimple.c (gimple_build_call_from_tree): Same.
12788 * gimplify.c (gimplify_return_expr): Same.
12789 (gimplify_cond_expr): Same.
12790 (gimplify_modify_expr_complex_part): Same.
12791 (gimplify_modify_expr): Same.
12792 (gimple_push_cleanup): Same.
12793 (gimplify_expr): Same.
12794 * omp-expand.c (expand_omp_for_generic): Same.
12795 (expand_omp_taskloop_for_outer): Same.
12796 * omp-low.c (lower_rec_input_clauses): Same.
12797 (lower_lastprivate_clauses): Same.
12798 (lower_send_clauses): Same.
12799 (lower_omp_target): Same.
12800 * tree-cfg.c (pass_warn_function_return::execute): Same.
12801 * tree-complex.c (create_one_component_var): Same.
12802 * tree-inline.c (remap_gimple_op_r): Same.
12803 (copy_tree_body_r): Same.
12804 (declare_return_variable): Same.
12805 (expand_call_inline): Same.
12806 * tree-nested.c (lookup_field_for_decl): Same.
12807 * tree-sra.c (create_access_replacement): Same.
12808 (generate_subtree_copies): Same.
12809 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Same.
12810 * tree-ssa-forwprop.c (combine_cond_expr_cond): Same.
12811 * tree-ssa-loop-ch.c (ch_base::copy_headers): Same.
12812 * tree-ssa-loop-im.c (execute_sm): Same.
12813 * tree-ssa-phiopt.c (cond_store_replacement): Same.
12814 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
12815 (handle_builtin_strcpy): Same.
12816 (maybe_diag_stxncpy_trunc): Same.
12817 (handle_builtin_stxncpy_strncat): Same.
12818 (handle_builtin_strcat): Same.
12819 * tree-ssa-uninit.c (get_no_uninit_warning): Same.
12820 (set_no_uninit_warning): Same.
12821 (uninit_undefined_value_p): Same.
12822 (warn_uninit): Same.
12823 (maybe_warn_operand): Same.
12824 * tree-vrp.c (compare_values_warnv): Same.
12825 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr): Same.
12826 (test_for_singularity): Same.
12827 * gimple.h (warning_suppressed_p): New function.
12828 (suppress_warning): Same.
12829 (copy_no_warning): Same.
12830 (gimple_set_block): Call gimple_set_location.
12831 (gimple_set_location): Call copy_warning.
12833 2021-06-25 Martin Sebor <msebor@redhat.com>
12835 * tree.h (warning_suppressed_at, copy_warning,
12836 warning_suppressed_p, suppress_warning): New functions.
12838 2021-06-25 Martin Sebor <msebor@redhat.com>
12840 * Makefile.in (OBJS-libcommon): Add diagnostic-spec.o.
12841 * gengtype.c (open_base_files): Add diagnostic-spec.h.
12842 * diagnostic-spec.c: New file.
12843 * diagnostic-spec.h: New file.
12844 * tree.h (no_warning, all_warnings, suppress_warning_at): New
12846 * warning-control.cc: New file.
12848 2021-06-25 liuhongt <hongtao.liu@intel.com>
12851 * config/i386/i386.c (x86_order_regs_for_local_alloc):
12854 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
12856 PR tree-optimization/101189
12857 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Pass
12858 LHS range of condition to postfold routine.
12859 (fold_using_range::postfold_gcond_edges): Only process the TRUE or
12860 FALSE edge if the LHS range supports it being taken.
12861 * gimple-range-fold.h (postfold_gcond_edges): Add range parameter.
12863 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
12865 * value-relation.cc (equiv_oracle::dump): Do not dump NULL blocks.
12866 (relation_oracle::find_relation_block): Check correct bitmap.
12867 (relation_oracle::dump): Do not dump NULL blocks.
12869 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
12871 * gimple-range-cache.cc (ranger_cache::propagate_cache): Call
12872 range_on_edge instead of manually calculating.
12874 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
12876 * range-op.cc: Fix comment.
12878 2021-06-24 Uroš Bizjak <ubizjak@gmail.com>
12881 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
12882 Handle V8QI and V4HI modes.
12883 * config/i386/mmx.md (sse4_1_<any_extend:code>v4qiv4hi2):
12885 (sse4_1_<any_extend:code>v4qiv4hi2): Ditto.
12886 (mmxpackmode): New mode attribute.
12887 (vec_pack_trunc_<mmxpackmode:mode>): New expander.
12888 (mmxunpackmode): New mode attribute.
12889 (vec_unpacks_lo_<mmxunpackmode:mode>): New expander.
12890 (vec_unpacks_hi_<mmxunpackmode:mode>): Ditto.
12891 (vec_unpacku_lo_<mmxunpackmode:mode>): Ditto.
12892 (vec_unpacku_hi_<mmxunpackmode:mode>): Ditto.
12893 * config/i386/i386.md (extsuffix): Move from ...
12894 * config/i386/sse.md: ... here.
12896 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
12898 * dwarf2out.c (dwarf2out_assembly_start): Emit .file 0 marker here..
12899 (dwarf2out_finish): ...instead of here.
12901 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
12903 * configure.ac (--gdwarf-5 option): Use objdump instead of readelf.
12904 (working --gdwarf-4/--gdwarf-5 for all sources): Likewise.
12905 (--gdwarf-4 not refusing generated .debug_line): Adjust for Windows.
12906 * configure: Regenerate.
12908 2021-06-24 Richard Biener <rguenther@suse.de>
12910 * config/i386/sse.md (vec_addsubv4df3, vec_addsubv2df3,
12911 vec_addsubv8sf3, vec_addsubv4sf3): Merge into ...
12912 (vec_addsub<mode>3): ... using a new addsub_cst mode attribute.
12914 2021-06-24 Richard Biener <rguenther@suse.de>
12916 * config/i386/sse.md (avx_addsubv4df3): Rename to
12918 (avx_addsubv8sf3): Rename to vec_addsubv8sf3.
12919 (sse3_addsubv2df3): Rename to vec_addsubv2df3.
12920 (sse3_addsubv4sf3): Rename to vec_addsubv4sf3.
12921 * config/i386/i386-builtin.def: Adjust.
12922 * internal-fn.def (VEC_ADDSUB): New internal optab fn.
12923 * optabs.def (vec_addsub_optab): New optab.
12924 * tree-vect-slp-patterns.c (class addsub_pattern): New.
12925 (slp_patterns): Add addsub_pattern.
12926 * tree-vect-slp.c (vect_optimize_slp): Disable propagation
12927 across CFN_VEC_ADDSUB.
12928 * tree-vectorizer.h (vect_pattern::vect_pattern): Make
12930 * doc/md.texi (vec_addsub<mode>3): Document.
12932 2021-06-24 Jakub Jelinek <jakub@redhat.com>
12934 PR middle-end/101170
12935 * df-scan.c (df_ref_record): For paradoxical big-endian SUBREGs
12936 where regno + subreg_regno_offset wraps around use 0 as starting
12939 2021-06-24 Jakub Jelinek <jakub@redhat.com>
12941 PR middle-end/101172
12942 * stor-layout.c (finish_bitfield_representative): If nextf has
12943 error_mark_node type, set repr type to error_mark_node too.
12945 2021-06-24 Ilya Leoshkevich <iii@linux.ibm.com>
12947 * config/s390/s390.c (s390_function_profiler): Ignore labelno
12949 * config/s390/s390.h (NO_PROFILE_COUNTERS): Define.
12951 2021-06-24 Richard Biener <rguenther@suse.de>
12953 * tree-vect-slp.c (vect_optimize_slp): Do not propagate
12954 across operations that have different semantics on different
12957 2021-06-24 Jakub Jelinek <jakub@redhat.com>
12959 * tree.h (OMP_CLAUSE_MAP_IN_REDUCTION): Document meaning for OpenMP.
12960 * gimplify.c (gimplify_scan_omp_clauses): For OpenMP map clauses
12961 with OMP_CLAUSE_MAP_IN_REDUCTION flag partially defer gimplification
12962 of non-decl OMP_CLAUSE_DECL. For OMP_CLAUSE_IN_REDUCTION on
12963 OMP_TARGET user outer_ctx instead of ctx for placeholders and
12964 initializer/combiner gimplification.
12965 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_MAP_IN_REDUCTION
12966 on target constructs.
12967 (lower_rec_input_clauses): Likewise.
12968 (lower_omp_target): Likewise.
12969 * omp-expand.c (expand_omp_target): Temporarily ignore nowait clause
12970 on target if in_reduction is present.
12972 2021-06-24 Kewen Lin <linkw@linux.ibm.com>
12974 * tree-predcom.c (class pcom_worker): New class.
12975 (release_chain): Renamed to...
12976 (pcom_worker::release_chain): ...this.
12977 (release_chains): Renamed to...
12978 (pcom_worker::release_chains): ...this.
12979 (aff_combination_dr_offset): Renamed to...
12980 (pcom_worker::aff_combination_dr_offset): ...this.
12981 (determine_offset): Renamed to...
12982 (pcom_worker::determine_offset): ...this.
12983 (class comp_ptrs): New class.
12984 (split_data_refs_to_components): Renamed to...
12985 (pcom_worker::split_data_refs_to_components): ...this,
12986 and update with class comp_ptrs.
12987 (suitable_component_p): Renamed to...
12988 (pcom_worker::suitable_component_p): ...this.
12989 (filter_suitable_components): Renamed to...
12990 (pcom_worker::filter_suitable_components): ...this.
12991 (valid_initializer_p): Renamed to...
12992 (pcom_worker::valid_initializer_p): ...this.
12993 (find_looparound_phi): Renamed to...
12994 (pcom_worker::find_looparound_phi): ...this.
12995 (add_looparound_copies): Renamed to...
12996 (pcom_worker::add_looparound_copies): ...this.
12997 (determine_roots_comp): Renamed to...
12998 (pcom_worker::determine_roots_comp): ...this.
12999 (determine_roots): Renamed to...
13000 (pcom_worker::determine_roots): ...this.
13001 (single_nonlooparound_use): Renamed to...
13002 (pcom_worker::single_nonlooparound_use): ...this.
13003 (remove_stmt): Renamed to...
13004 (pcom_worker::remove_stmt): ...this.
13005 (execute_pred_commoning_chain): Renamed to...
13006 (pcom_worker::execute_pred_commoning_chain): ...this.
13007 (execute_pred_commoning): Renamed to...
13008 (pcom_worker::execute_pred_commoning): ...this.
13009 (struct epcc_data): New member worker.
13010 (execute_pred_commoning_cbck): Call execute_pred_commoning
13011 with pcom_worker pointer.
13012 (find_use_stmt): Renamed to...
13013 (pcom_worker::find_use_stmt): ...this.
13014 (find_associative_operation_root): Renamed to...
13015 (pcom_worker::find_associative_operation_root): ...this.
13016 (find_common_use_stmt): Renamed to...
13017 (pcom_worker::find_common_use_stmt): ...this.
13018 (combinable_refs_p): Renamed to...
13019 (pcom_worker::combinable_refs_p): ...this.
13020 (reassociate_to_the_same_stmt): Renamed to...
13021 (pcom_worker::reassociate_to_the_same_stmt): ...this.
13022 (stmt_combining_refs): Renamed to...
13023 (pcom_worker::stmt_combining_refs): ...this.
13024 (combine_chains): Renamed to...
13025 (pcom_worker::combine_chains): ...this.
13026 (try_combine_chains): Renamed to...
13027 (pcom_worker::try_combine_chains): ...this.
13028 (prepare_initializers_chain): Renamed to...
13029 (pcom_worker::prepare_initializers_chain): ...this.
13030 (prepare_initializers): Renamed to...
13031 (pcom_worker::prepare_initializers): ...this.
13032 (prepare_finalizers_chain): Renamed to...
13033 (pcom_worker::prepare_finalizers_chain): ...this.
13034 (prepare_finalizers): Renamed to...
13035 (pcom_worker::prepare_finalizers): ...this.
13036 (tree_predictive_commoning_loop): Renamed to...
13037 (pcom_worker::tree_predictive_commoning_loop): ...this, adjust
13038 some calls and remove some cleanup code.
13039 (tree_predictive_commoning): Adjusted to use pcom_worker instance.
13040 (static variable looparound_phis): Remove.
13041 (static variable name_expansions): Remove.
13043 2021-06-24 Richard Biener <rguenther@suse.de>
13045 * tree-vect-slp.c (slpg_vertex): New struct.
13046 (vect_slp_build_vertices): Adjust.
13047 (vect_optimize_slp): Likewise. Maintain an outgoing permute
13048 and a materialized one.
13050 2021-06-24 Richard Biener <rguenther@suse.de>
13052 PR tree-optimization/101105
13053 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
13054 Only ignore steps when they are equal or scalar order is preserved.
13056 2021-06-24 liuhongt <hongtao.liu@intel.com>
13059 * config/i386/i386-expand.c (ix86_expand_vec_interleave):
13060 Adjust comments for ix86_expand_vecop_qihi2.
13061 (ix86_expand_vecmul_qihi): Renamed to ..
13062 (ix86_expand_vecop_qihi2): Adjust function prototype to
13063 support shift operation, add static to definition.
13064 (ix86_expand_vec_shift_qihi_constant): Add static to definition.
13065 (ix86_expand_vecop_qihi): Call ix86_expand_vecop_qihi2 and
13066 ix86_expand_vec_shift_qihi_constant.
13067 * config/i386/i386-protos.h (ix86_expand_vecmul_qihi): Deleted.
13068 (ix86_expand_vec_shift_qihi_constant): Deleted.
13069 * config/i386/sse.md (VI12_256_512_AVX512VL): New mode
13071 (mulv8qi3): Call ix86_expand_vecop_qihi directly, add
13072 condition TARGET_64BIT.
13073 (mul<mode>3): Ditto.
13074 (<insn><mode>3): Ditto.
13075 (vlshr<mode>3): Extend to support avx512 vlshr.
13076 (v<insn><mode>3): New expander for
13078 (v<insn>v8qi3): Ditto.
13079 (vashrv8hi3<mask_name>): Renamed to ..
13080 (vashr<mode>3): And extend to support V16QImode for avx512.
13081 (vashrv16qi3): Deleted.
13082 (vashrv2di3<mask_name>): Extend expander to support avx512
13085 2021-06-23 Dimitar Dimitrov <dimitar@dinux.eu>
13087 * doc/lto.texi (Design Overview): Update that slim objects are
13090 2021-06-23 Aaron Sawdey <acsawdey@linux.ibm.com>
13092 * config/rs6000/rs6000-cpus.def: Take OPTION_MASK_PCREL_OPT out
13093 of OTHER_POWER10_MASKS so it will not be enabled by default.
13095 2021-06-23 Richard Biener <rguenther@suse.de>
13096 Martin Jambor <mjambor@suse.cz>
13098 * tree-inline.c (setup_one_parameter): Set TREE_READONLY of the
13099 param replacement unconditionally. Adjust comment.
13101 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
13103 * Makefile.in (OBJS): Add gimple-range-fold.o
13104 * gimple-range-fold.cc: New.
13105 * gimple-range-fold.h: New.
13106 * gimple-range-gori.cc (gimple_range_calc_op1): Move to here.
13107 (gimple_range_calc_op2): Ditto.
13108 * gimple-range-gori.h: Move prototypes to here.
13109 * gimple-range.cc: Adjust include files.
13110 (fur_source:fur_source): Relocate to gimple-range-fold.cc.
13111 (fur_source::get_operand): Ditto.
13112 (fur_source::get_phi_operand): Ditto.
13113 (fur_source::query_relation): Ditto.
13114 (fur_source::register_relation): Ditto.
13115 (class fur_edge): Ditto.
13116 (fur_edge::fur_edge): Ditto.
13117 (fur_edge::get_operand): Ditto.
13118 (fur_edge::get_phi_operand): Ditto.
13119 (fur_stmt::fur_stmt): Ditto.
13120 (fur_stmt::get_operand): Ditto.
13121 (fur_stmt::get_phi_operand): Ditto.
13122 (fur_stmt::query_relation): Ditto.
13123 (class fur_depend): Relocate to gimple-range-fold.h.
13124 (fur_depend::fur_depend): Relocate to gimple-range-fold.cc.
13125 (fur_depend::register_relation): Ditto.
13126 (fur_depend::register_relation): Ditto.
13127 (class fur_list): Ditto.
13128 (fur_list::fur_list): Ditto.
13129 (fur_list::get_operand): Ditto.
13130 (fur_list::get_phi_operand): Ditto.
13131 (fold_range): Ditto.
13132 (adjust_pointer_diff_expr): Ditto.
13133 (gimple_range_adjustment): Ditto.
13134 (gimple_range_base_of_assignment): Ditto.
13135 (gimple_range_operand1): Ditto.
13136 (gimple_range_operand2): Ditto.
13137 (gimple_range_calc_op1): Relocate to gimple-range-gori.cc.
13138 (gimple_range_calc_op2): Ditto.
13139 (fold_using_range::fold_stmt): Relocate to gimple-range-fold.cc.
13140 (fold_using_range::range_of_range_op): Ditto.
13141 (fold_using_range::range_of_address): Ditto.
13142 (fold_using_range::range_of_phi): Ditto.
13143 (fold_using_range::range_of_call): Ditto.
13144 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
13145 (fold_using_range::range_of_builtin_call): Ditto.
13146 (fold_using_range::range_of_cond_expr): Ditto.
13147 (fold_using_range::range_of_ssa_name_with_loop_info): Ditto.
13148 (fold_using_range::relation_fold_and_or): Ditto.
13149 (fold_using_range::postfold_gcond_edges): Ditto.
13150 * gimple-range.h: Add gimple-range-fold.h to include files. Change
13151 GIMPLE_RANGE_STMT_H to GIMPLE_RANGE_H.
13152 (gimple_range_handler): Relocate to gimple-range-fold.h.
13153 (gimple_range_ssa_p): Ditto.
13154 (range_compatible_p): Ditto.
13155 (class fur_source): Ditto.
13156 (class fur_stmt): Ditto.
13157 (class fold_using_range): Ditto.
13158 (gimple_range_calc_op1): Relocate to gimple-range-gori.h
13159 (gimple_range_calc_op2): Ditto.
13161 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
13163 PR tree-optimization/101148
13164 PR tree-optimization/101014
13165 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
13166 (ranger_cache::~ranger_cache): Adjust.
13167 (ranger_cache::block_range): Check if propagation disallowed.
13168 (ranger_cache::propagate_cache): Disallow propagation if new value
13169 can't be stored properly.
13170 * gimple-range-cache.h (ranger_cache::m_propfail): New member.
13172 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
13174 * gimple-range-cache.cc (class ssa_block_ranges): Adjust prototype.
13175 (sbr_vector::set_bb_range): Return true.
13176 (class sbr_sparse_bitmap): Adjust.
13177 (sbr_sparse_bitmap::set_bb_range): Return value.
13178 (block_range_cache::set_bb_range): Return value.
13179 (ranger_cache::propagate_cache): Use return value to print msg.
13180 * gimple-range-cache.h (class block_range_cache): Adjust.
13182 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
13184 * gimple-range.cc (dump_bb): Use range_on_edge from the cache.
13186 2021-06-23 Jeff Law <jeffreyalaw@gmail.com>
13188 * config/h8300/logical.md (<code><mode>3<ccnz>): Use <cczn>
13189 so this pattern can be used for test/compare removal. Pass
13190 current insn to compute_logical_op_length and output_logical_op.
13191 * config/h8300/h8300.c (compute_logical_op_cc): Remove.
13192 (h8300_and_costs): Add argument to compute_logical_op_length.
13193 (output_logical_op): Add new argument. Use it to determine if the
13194 condition codes are used and adjust the output accordingly.
13195 (compute_logical_op_length): Add new argument and update length
13196 computations when condition codes are used.
13197 * config/h8300/h8300-protos.h (compute_logical_op_length): Update
13199 (output_logical_op): Likewise.
13201 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
13204 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
13205 Handle 64bit modes for TARGET_XOP. Use indirect gen_* functions.
13206 * config/i386/mmx.md (mmx_ppermv64): New insn pattern.
13207 * config/i386/i386.md (unspec): Move UNSPEC_XOP_PERMUTE from ...
13208 * config/i386/sse.md (unspec): ... here.
13210 2021-06-23 Martin Liska <mliska@suse.cz>
13213 * optc-save-gen.awk: Put back arm_fp16_format to
13216 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
13219 * config/i386/i386.md (bsr_rex64): Add zero-flag setting RTX.
13222 (clz<mode>2): Update RTX pattern for additions.
13224 2021-06-23 Jakub Jelinek <jakub@redhat.com>
13226 PR middle-end/101167
13227 * omp-low.c (lower_omp_regimplify_p): Regimplify also PARM_DECLs
13228 and RESULT_DECLs that have DECL_HAS_VALUE_EXPR_P set.
13230 2021-06-22 Sergei Trofimovich <siarheit@google.com>
13232 * doc/rtl.texi: drop unbalanced parenthesis.
13234 2021-06-22 Richard Biener <rguenther@suse.de>
13236 PR middle-end/101156
13237 * gimplify.c (gimplify_expr): Remove premature incorrect
13240 2021-06-22 Jakub Jelinek <jakub@redhat.com>
13242 PR tree-optimization/101159
13243 * tree-vect-patterns.c (vect_recog_popcount_pattern): Fix some
13246 2021-06-22 Jakub Jelinek <jakub@redhat.com>
13248 PR middle-end/101160
13249 * function.c (assign_parms): For decl_result with TYPE_EMPTY_P type
13250 clear crtl->return_rtx instead of keeping it referencing a pseudo.
13252 2021-06-22 Jakub Jelinek <jakub@redhat.com>
13253 Andrew Pinski <apinski@marvell.com>
13255 PR tree-optimization/101162
13256 * fold-const.c (range_check_type): Handle OFFSET_TYPE like pointer
13259 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
13261 * range-op.cc (range_relational_tests): New.
13262 (range_op_tests): Call range_relational_tests.
13264 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
13266 * range-op.cc (operator_cast::lhs_op1_relation): New.
13267 (operator_identity::lhs_op1_relation): Mew.
13269 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
13271 * range-op.cc (operator_minus::op1_op2_relation_effect): New.
13273 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
13275 * range-op.cc (operator_plus::lhs_op1_relation): New.
13276 (operator_plus::lhs_op2_relation): New.
13278 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
13280 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a
13281 relation_oracle if dominators exist.
13282 (ranger_cache::~ranger_cache): Dispose of oracle.
13283 (ranger_cache::dump_bb): Dump oracle.
13284 * gimple-range.cc (fur_source::fur_source): New.
13285 (fur_source::get_operand): Use mmeber query.
13286 (fur_source::get_phi_operand): Use member_query.
13287 (fur_source::query_relation): New.
13288 (fur_source::register_dependency): Delete.
13289 (fur_source::register_relation): New.
13290 (fur_edge::fur_edge): Adjust.
13291 (fur_edge::get_phi_operand): Fix comment.
13292 (fur_edge::query): Delete.
13293 (fur_stmt::fur_stmt): Adjust.
13294 (fur_stmt::query): Delete.
13295 (fur_depend::fur_depend): Adjust.
13296 (fur_depend::register_relation): New.
13297 (fur_depend::register_relation): New.
13298 (fur_list::fur_list): Adjust.
13299 (fur_list::get_operand): Use member query.
13300 (fold_using_range::range_of_range_op): Process and query relations.
13301 (fold_using_range::range_of_address): Adjust dependency call.
13302 (fold_using_range::range_of_phi): Ditto.
13303 (gimple_ranger::gimple_ranger): New. Use ranger_ache oracle.
13304 (fold_using_range::relation_fold_and_or): New.
13305 (fold_using_range::postfold_gcond_edges): New.
13306 * gimple-range.h (class gimple_ranger): Adjust.
13307 (class fur_source): Adjust members.
13308 (class fur_stmt): Ditto.
13309 (class fold_using_range): Ditto.
13311 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
13313 * range-op.cc (range_operator::wi_fold): Apply relation effect.
13314 (range_operator::fold_range): Adjust and apply relation effect.
13315 (*::fold_range): Add relation parameters.
13316 (*::op1_range): Ditto.
13317 (*::op2_range): Ditto.
13318 (range_operator::lhs_op1_relation): New.
13319 (range_operator::lhs_op2_relation): New.
13320 (range_operator::op1_op2_relation): New.
13321 (range_operator::op1_op2_relation_effect): New.
13322 (relop_early_resolve): New.
13323 (operator_equal::op1_op2_relation): New.
13324 (operator_equal::fold_range): Call relop_early_resolve.
13325 (operator_not_equal::op1_op2_relation): New.
13326 (operator_not_equal::fold_range): Call relop_early_resolve.
13327 (operator_lt::op1_op2_relation): New.
13328 (operator_lt::fold_range): Call relop_early_resolve.
13329 (operator_le::op1_op2_relation): New.
13330 (operator_le::fold_range): Call relop_early_resolve.
13331 (operator_gt::op1_op2_relation): New.
13332 (operator_gt::fold_range): Call relop_early_resolve.
13333 (operator_ge::op1_op2_relation): New.
13334 (operator_ge::fold_range): Call relop_early_resolve.
13335 * range-op.h (class range_operator): Adjust parameters and methods.
13337 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
13339 * Makefile.in (OBJS): Add value-relation.o.
13340 * gimple-range.h: Adjust include files.
13341 * tree-data-ref.c: Adjust include file order.
13342 * value-query.cc (range_query::get_value_range): Default to no oracle.
13343 (range_query::query_relation): New.
13344 (range_query::query_relation): New.
13345 * value-query.h (class range_query): Adjust.
13346 * value-relation.cc: New.
13347 * value-relation.h: New.
13349 2021-06-22 Richard Biener <rguenther@suse.de>
13351 PR tree-optimization/101151
13352 * tree-ssa-sink.c (statement_sink_location): Expand irreducible
13355 2021-06-22 Jojo R <rjiejie@linux.alibaba.com>
13357 * config/riscv/riscv.c (thead_c906_tune_info): New.
13358 (riscv_tune_info_table): Use new tune.
13360 2021-06-22 Richard Biener <rguenther@suse.de>
13362 PR tree-optimization/101158
13363 * tree-vect-slp.c (vect_build_slp_tree_1): Move same operand
13364 checking after checking for matching operation.
13366 2021-06-22 Richard Biener <rguenther@suse.de>
13368 PR tree-optimization/101159
13369 * tree-vect-patterns.c (vect_recog_popcount_pattern): Add
13370 missing NULL vectype check.
13372 2021-06-22 Richard Biener <rguenther@suse.de>
13374 PR tree-optimization/101154
13375 * tree-vect-slp.c (vect_build_slp_tree_2): Fix out-of-bound access.
13377 2021-06-22 Jakub Jelinek <jakub@redhat.com>
13380 * config/i386/i386-protos.h (ix86_last_zero_store_uid): Declare.
13381 * config/i386/i386-expand.c (ix86_last_zero_store_uid): New variable.
13382 * config/i386/i386.c (ix86_expand_prologue): Clear it.
13383 * config/i386/i386.md (peephole2s for 1/2/4 stores of const0_rtx):
13384 Remove "" from match_operand. Emit new insns using emit_move_insn and
13385 set ix86_last_zero_store_uid to INSN_UID of the last store.
13386 Add peephole2s for 1/2/4 stores of const0_rtx following previous
13389 2021-06-22 Martin Liska <mliska@suse.cz>
13391 * auto-profile.c (AUTO_PROFILE_VERSION): Bump as string format
13394 2021-06-22 Martin Liska <mliska@suse.cz>
13396 * gcov-io.h: Remove padding entries.
13398 2021-06-22 liuhongt <hongtao.liu@intel.com>
13400 PR tree-optimization/97770
13401 * tree-vect-patterns.c (vect_recog_popcount_pattern):
13403 (vect_recog_func vect_vect_recog_func_ptrs): Add new pattern.
13405 2021-06-22 liuhongt <hongtao.liu@intel.com>
13408 * config/i386/i386-builtin.def (BDESC): Adjust builtin name.
13409 * config/i386/sse.md (<avx512>_expand<mode>_mask): Rename to ..
13410 (expand<mode>_mask): this ..
13411 (*expand<mode>_mask): New pre_reload splitter to transform
13412 v{,p}expand* to vmov* when mask is zero, all ones, or has all
13413 ones in it's lower part, otherwise still generate
13416 2021-06-22 liuhongt <hongtao.liu@intel.com>
13419 * config/i386/i386-expand.c
13420 (ix86_expand_special_args_builtin): Keep constm1_operand only
13421 if it satisfies insn's operand predicate.
13423 2021-06-21 Jason Merrill <jason@redhat.com>
13426 * df-scan.c (df_ref_record): Check that regno < endregno.
13427 * function.c (assign_parms, expand_function_end): Do nothing with a
13428 TYPE_EMPTY_P result.
13430 2021-06-21 Richard Biener <rguenther@suse.de>
13432 PR tree-optimization/101120
13433 * tree-vect-data-refs.c (bump_vector_ptr): Fold the
13435 * tree-vect-slp.c (vect_transform_slp_perm_load): Add
13436 DR chain DCE capability.
13437 * tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
13438 * tree-vect-stmts.c (vectorizable_load): Remove unused
13439 loads in the DR chain for SLP.
13441 2021-06-21 Jakub Jelinek <jakub@redhat.com>
13443 PR inline-asm/100785
13444 * gimplify.c (gimplify_asm_expr): Don't diagnose errors if
13445 output or input operands were already error_mark_node.
13446 * cfgexpand.c (expand_asm_stmt): If errors are emitted,
13447 remove all inputs, outputs and clobbers from the asm and
13448 set template to "".
13450 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
13452 * config/arm/arm_neon.h (vceq_s8): Replace builtin with __a == __b.
13453 (vceq_s16): Likewise.
13454 (vceq_s32): Likewise.
13455 (vceq_u8): Likewise.
13456 (vceq_u16): Likewise.
13457 (vceq_u32): Likewise.
13458 (vceq_p8): Likewise.
13459 (vceqq_s8): Likewise.
13460 (vceqq_s16): Likewise.
13461 (vceqq_s32): Likewise.
13462 (vceqq_u8): Likewise.
13463 (vceqq_u16): Likewise.
13464 (vceqq_u32): Likewise.
13465 (vceqq_p8): Likewise.
13466 (vceq_f32): Gate __a == __b on __FAST_MATH__.
13467 (vceqq_f32): Likewise.
13468 (vceq_f16): Likewise.
13469 (vceqq_f16): Likewise.
13471 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
13474 * config/arm/iterators.md (NEON_VACMP): Remove.
13475 * config/arm/neon.md (neon_vca<cmp_op><mode>): Use GLTE instead of GTGE
13477 (neon_vca<cmp_op><mode>_insn): Likewise.
13478 (neon_vca<cmp_op_unsp><mode>_insn_unspec): Use NEON_VAGLTE instead of
13481 2021-06-21 Richard Biener <rguenther@suse.de>
13483 PR tree-optimization/101121
13484 * tree-vect-slp.c (vect_build_slp_tree_2): To not fail fatally
13485 when we just lack a stmt with the desired op when doing permutation.
13486 (vect_build_slp_tree): When caching a failed SLP build attempt
13487 assert that at least one lane is marked as not matching.
13489 2021-06-21 liuhongt <hongtao.liu@intel.com>
13492 * config/i386/i386.md: (*anddi_1): Disparage slightly the mask
13493 register alternative.
13494 (*and<mode>_1): Ditto.
13496 (*andn<mode>_1): Ditto.
13497 (*<code><mode>_1): Ditto.
13498 (*<code>qi_1): Ditto.
13499 (*one_cmpl<mode>2_1): Ditto.
13500 (*one_cmplsi2_1_zext): Ditto.
13501 (*one_cmplqi2_1): Ditto.
13502 * config/i386/i386.c (x86_order_regs_for_local_alloc): Change
13503 the order of mask registers to be before general registers.
13505 2021-06-21 Roger Sayle <roger@nextmovesoftware.com>
13508 * config/i386/i386.md: New define_peephole2s to shrink writing
13509 1, 2 or 4 consecutive zeros to memory when optimizing for size.
13511 2021-06-18 Jeff Law <jeffreyalaw@gmail.com>
13513 * config/h8300/h8300.c (h8300_select_cc_mode): Handle SYMBOL_REF.
13514 * config/h8300/logical.md (<code><mode>3 logcial expander): Generate
13515 more efficient code when the source can be trivially simplified.
13517 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
13519 * gimple-range-cache.cc (ranger_cache::range_of_def): Calculate
13520 a range if global is not available.
13521 (ranger_cache::entry_range): Fallback to range_of_def.
13522 * gimple-range-cache.h (range_of_def): Adjust prototype.
13524 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
13526 PR tree-optimization/101014
13527 * gimple-range-cache.cc (ranger_cache::ranger_cache): Remove poor
13529 (ranger_cache::~ranger_cache): Ditto.
13530 (ranger_cache::enable_new_values): Delete.
13531 (ranger_cache::push_poor_value): Delete.
13532 (ranger_cache::range_of_def): Remove poor value processing.
13533 (ranger_cache::entry_range): Ditto.
13534 (ranger_cache::fill_block_cache): Ditto.
13535 * gimple-range-cache.h (class ranger_cache): Remove poor value members.
13536 * gimple-range.cc (gimple_ranger::range_of_expr): Remove call.
13537 * gimple-range.h (class gimple_ranger): Adjust.
13539 2021-06-18 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
13542 * common/config/arm/arm-common.c (arm_canon_arch_option_1): New function
13543 derived from arm_canon_arch.
13544 (arm_canon_arch_option): Call it.
13545 (arm_canon_arch_multilib_option): New function.
13546 * config/arm/arm-cpus.in (IGNORE_FOR_MULTILIB): New fgroup.
13547 * config/arm/arm.h (arm_canon_arch_multilib_option): New prototype.
13548 (CANON_ARCH_MULTILIB_SPEC_FUNCTION): New macro.
13549 (MULTILIB_ARCH_CANONICAL_SPECS): New macro.
13550 (DRIVER_SELF_SPECS): Add MULTILIB_ARCH_CANONICAL_SPECS.
13551 * config/arm/arm.opt (mlibarch): New option.
13552 * config/arm/t-rmprofile (MULTILIB_MATCHES): For armv8*-m, replace use
13553 of march on RHS with mlibarch.
13555 2021-06-18 Marcel Vollweiler <marcel@codesourcery.com>
13557 * config.in: Regenerate.
13558 * config/gcn/gcn.c (print_operand_address): Fix for global_load assembler
13560 * configure: Regenerate.
13561 * configure.ac: Fix for global_load assembler functions.
13563 2021-06-18 Richard Biener <rguenther@suse.de>
13565 PR tree-optimization/101112
13566 * tree-vect-slp.c (vect_slp_linearize_chain): Fix condition
13567 to lookup a pattern stmt def.
13569 2021-06-18 Jakub Jelinek <jakub@redhat.com>
13571 PR middle-end/101062
13572 * stor-layout.c (finish_bitfield_layout): Don't add bitfield
13573 representatives in QUAL_UNION_TYPE.
13575 2021-06-18 Andrew Pinski <apinski@marvell.com>
13577 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
13578 Add counting of how many times it is done.
13579 (factor_out_conditional_conversion): Likewise.
13580 (match_simplify_replacement): Likewise.
13581 (value_replacement): Likewise.
13582 (spaceship_replacement): Likewise.
13583 (cond_store_replacement): Likewise.
13584 (cond_if_else_store_replacement_1): Likewise.
13585 (hoist_adjacent_loads): Likewise.
13587 2021-06-18 Andrew Pinski <apinski@marvell.com>
13589 * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
13590 types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and CNONJ_EXPR.
13591 (verify_gimple_assign_binary): Reject point and offset types on
13592 MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
13593 FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
13594 FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.
13596 2021-06-18 Michael Meissner <meissner@linux.ibm.com>
13598 * config/rs6000/rs6000.c (rs6000_emit_minmax): Add support for ISA
13599 3.1 IEEE 128-bit floating point xsmaxcqp/xsmincqp instructions.
13600 * config/rs6000/rs6000.md (s<minmax><mode>3, IEEE128 iterator):
13603 2021-06-17 Aaron Sawdey <acsawdey@linux.ibm.com>
13605 * config/rs6000/genfusion.pl (gen_logical_addsubf): Add
13606 earlyclobber to alts 0/1.
13607 (gen_addadd): Add earlyclobber to alts 0/1.
13608 * config/rs6000/fusion.md: Regenerate file.
13610 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13612 * cfgloopanal.c (get_loop_hot_path): Make path an auto_vec.
13614 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
13616 * gimple-range-cache.cc: Comment cleanups.
13617 * gimple-range-gori.cc: Comment cleanups.
13618 * gimple-range.cc: Comment/spacing cleanups
13619 * value-range.h: Comment cleanups.
13621 2021-06-17 H.J. Lu <hjl.tools@gmail.com>
13624 * calls.c (expand_call): Replace PUSH_ARGS with
13625 targetm.calls.push_argument (0).
13626 (emit_library_call_value_1): Likewise.
13627 * defaults.h (PUSH_ARGS): Removed.
13628 (PUSH_ARGS_REVERSED): Replace PUSH_ARGS with
13629 targetm.calls.push_argument (0).
13630 * expr.c (block_move_libcall_safe_for_call_parm): Likewise.
13631 (emit_push_insn): Pass the number bytes to push to
13632 targetm.calls.push_argument and pass 0 if ARGS_ADDR is 0.
13633 * hooks.c (hook_bool_uint_true): New.
13634 * hooks.h (hook_bool_uint_true): Likewise.
13635 * rtlanal.c (nonzero_bits1): Replace PUSH_ARGS with
13636 targetm.calls.push_argument (0).
13637 * target.def (push_argument): Add a targetm.calls hook.
13638 * targhooks.c (default_push_argument): New.
13639 * targhooks.h (default_push_argument): Likewise.
13640 * config/bpf/bpf.h (PUSH_ARGS): Removed.
13641 * config/cr16/cr16.c (TARGET_PUSH_ARGUMENT): New.
13642 * config/cr16/cr16.h (PUSH_ARGS): Removed.
13643 * config/i386/i386.c (ix86_push_argument): New.
13644 (TARGET_PUSH_ARGUMENT): Likewise.
13645 * config/i386/i386.h (PUSH_ARGS): Removed.
13646 * config/m32c/m32c.c (TARGET_PUSH_ARGUMENT): New.
13647 * config/m32c/m32c.h (PUSH_ARGS): Removed.
13648 * config/nios2/nios2.h (PUSH_ARGS): Likewise.
13649 * config/pru/pru.h (PUSH_ARGS): Likewise.
13650 * doc/tm.texi.in: Remove PUSH_ARGS documentation. Add
13651 TARGET_PUSH_ARGUMENT hook.
13652 * doc/tm.texi: Regenerated.
13654 2021-06-17 Uroš Bizjak <ubizjak@gmail.com>
13657 * config/i386/i386-expand.c (expand_vector_set_var):
13658 Handle V2FS mode remapping. Pass TARGET_MMX_WITH_SSE to
13659 ix86_expand_vector_init_duplicate.
13660 (ix86_expand_vector_init_duplicate): Emit insv_1 for
13661 QImode for !TARGET_PARTIAL_REG_STALL.
13662 * config/i386/predicates.md (vec_setm_mmx_operand): New predicate.
13663 * config/i386/mmx.md (vec_setv2sf): Use vec_setm_mmx_operand
13664 as operand 2 predicate. Call ix86_expand_vector_set_var
13665 for non-constant index operand.
13666 (vec_setv2si): Ditto.
13667 (vec_setv4hi): Ditto.
13668 (vec_setv8qi): ditto.
13670 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
13672 PR tree-optimization/100790
13673 * gimple-range.cc (range_of_builtin_call): Cleanup clz and ctz
13676 2021-06-17 Martin Liska <mliska@suse.cz>
13678 * doc/invoke.texi: Use consistently -O1 instead of -O.
13680 2021-06-17 Martin Liska <mliska@suse.cz>
13682 * gcov-io.h: Update documentation entry about string format.
13684 2021-06-17 Marius Hillenbrand <mhillen@linux.ibm.com>
13687 * config/s390/vecintrin.h (vec_doublee): Fix to use
13688 __builtin_s390_vflls.
13689 (vec_floate): Fix to use __builtin_s390_vflrd.
13691 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13693 * dominance.c (get_dominated_to_depth): Return auto_vec<basic_block>.
13694 * dominance.h (get_dominated_to_depth): Likewise.
13695 (get_all_dominated_blocks): Likewise.
13696 * cfgcleanup.c (delete_unreachable_blocks): Adjust.
13697 * gcse.c (hoist_code): Likewise.
13698 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
13699 * tree-parloops.c (oacc_entry_exit_ok): Likewise.
13700 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Likewise.
13701 * tree-ssa-phiprop.c (pass_phiprop::execute): Likewise.
13703 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13705 * dominance.c (get_dominated_by_region): Return auto_vec<basic_block>.
13706 * dominance.h (get_dominated_by_region): Likewise.
13707 * tree-cfg.c (gimple_duplicate_sese_region): Adjust.
13708 (gimple_duplicate_sese_tail): Likewise.
13709 (move_sese_region_to_fn): Likewise.
13711 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13713 * dominance.c (get_dominated_by): Return auto_vec<basic_block>.
13714 * dominance.h (get_dominated_by): Likewise.
13715 * auto-profile.c (afdo_find_equiv_class): Adjust.
13716 * cfgloopmanip.c (duplicate_loop_to_header_edge): Likewise.
13717 * loop-unroll.c (unroll_loop_runtime_iterations): Likewise.
13718 * tree-cfg.c (test_linear_chain): Likewise.
13719 (test_diamond): Likewise.
13721 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13723 * cfgloop.h (get_loop_hot_path): Return auto_vec<basic_block>.
13724 * cfgloopanal.c (get_loop_hot_path): Likewise.
13725 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
13727 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13729 * cgraph.c (cgraph_node::collect_callers): Return
13730 auto_vec<cgraph_edge *>.
13731 * cgraph.h (cgraph_node::collect_callers): Likewise.
13732 * ipa-cp.c (create_specialized_node): Adjust.
13733 (decide_about_value): Likewise.
13734 (decide_whether_version_node): Likewise.
13735 * ipa-sra.c (process_isra_node_results): Likewise.
13737 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13739 * vec.h (vl_ptr>::using_auto_storage): Handle null m_vec.
13740 (auto_vec<T, 0>::auto_vec): Define move constructor, and delete copy
13742 (auto_vec<T, 0>::operator=): Define move assignment and delete copy
13745 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
13747 * gimple-range.cc (debug_seed_ranger): New.
13748 (dump_ranger): New.
13749 (debug_ranger): New.
13751 2021-06-17 Richard Biener <rguenther@suse.de>
13753 PR tree-optimization/54400
13754 * tree-vectorizer.h (enum slp_instance_kind): Add
13755 slp_inst_kind_bb_reduc.
13756 (reduction_fn_for_scalar_code): Declare.
13757 * tree-vect-data-refs.c (vect_slp_analyze_instance_dependence):
13758 Check SLP_INSTANCE_KIND instead of looking at the
13760 (vect_slp_analyze_instance_alignment): Likewise.
13761 * tree-vect-loop.c (reduction_fn_for_scalar_code): Export.
13762 * tree-vect-slp.c (vect_slp_linearize_chain): Split out
13763 chain linearization from vect_build_slp_tree_2 and generalize
13764 for the use of BB reduction vectorization.
13765 (vect_build_slp_tree_2): Adjust accordingly.
13766 (vect_optimize_slp): Elide permutes at the root of BB reduction
13768 (vectorizable_bb_reduc_epilogue): New function.
13769 (vect_slp_prune_covered_roots): Likewise.
13770 (vect_slp_analyze_operations): Use them.
13771 (vect_slp_check_for_constructors): Recognize associatable
13772 chains for BB reduction vectorization.
13773 (vectorize_slp_instance_root_stmt): Generate code for the
13774 BB reduction epilogue.
13776 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
13778 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Check with
13780 (gori_compute::may_recompute_p): New.
13781 (gori_compute::outgoing_edge_range_p): Perform recomputations.
13782 * gimple-range-gori.h (class gori_compute): Add prototype.
13784 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
13786 * gimple-range-cache.cc (ranger_cache::range_on_edge): Always return
13787 true when a range can be calculated.
13788 * gimple-range.cc (gimple_ranger::dump_bb): Check has_edge_range_p.
13790 2021-06-16 Martin Sebor <msebor@redhat.com>
13792 * doc/invoke.texi (-Wmismatched-dealloc, -Wmismatched-new-delete):
13793 Correct documented defaults.
13795 2021-06-16 Andrew MacLeod <amacleod@redhat.com>
13797 * gimple-range-cache.cc (ranger_cache::ranger_cache): Initialize
13798 m_new_value_p directly.
13800 2021-06-16 Uroš Bizjak <ubizjak@gmail.com>
13803 * config/i386/i386-expand.c (expand_vec_perm_2perm_pblendv):
13804 Handle 64bit modes for TARGET_SSE4_1.
13805 (expand_vec_perm_pshufb2): Handle 64bit modes for TARGET_SSSE3.
13806 (expand_vec_perm_even_odd_pack): Handle V4HI mode.
13807 (expand_vec_perm_even_odd_1) <case E_V4HImode>: Expand via
13808 expand_vec_perm_pshufb2 for TARGET_SSSE3 and via
13809 expand_vec_perm_even_odd_pack for TARGET_SSE4_1.
13810 * config/i386/mmx.md (mmx_packusdw): New insn pattern.
13812 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
13814 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn<mode>):
13815 Change to an expander that emits the correct instruction
13816 depending on endianness.
13817 (aarch64_<sur><addsub>hn<mode>_insn_le): Define.
13818 (aarch64_<sur><addsub>hn<mode>_insn_be): Define.
13820 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
13822 * config/aarch64/aarch64-simd-builtins.def: Split generator
13823 for aarch64_<su>qmovn builtins into scalar and vector
13825 * config/aarch64/aarch64-simd.md (aarch64_<su>qmovn<mode>_insn_le):
13827 (aarch64_<su>qmovn<mode>_insn_be): Define.
13828 (aarch64_<su>qmovn<mode>): Split into scalar and vector
13829 variants. Change vector variant to an expander that emits the
13830 correct instruction depending on endianness.
13832 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
13834 * config/aarch64/aarch64-simd-builtins.def: Split generator
13835 for aarch64_sqmovun builtins into scalar and vector variants.
13836 * config/aarch64/aarch64-simd.md (aarch64_sqmovun<mode>):
13837 Split into scalar and vector variants. Change vector variant
13838 to an expander that emits the correct instruction depending
13840 (aarch64_sqmovun<mode>_insn_le): Define.
13841 (aarch64_sqmovun<mode>_insn_be): Define.
13843 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
13845 * config/aarch64/aarch64-simd.md (aarch64_xtn<mode>_insn_le):
13846 Define - modeling zero-high-half semantics.
13847 (aarch64_xtn<mode>): Change to an expander that emits the
13848 appropriate instruction depending on endianness.
13849 (aarch64_xtn<mode>_insn_be): Define - modeling zero-high-half
13851 (aarch64_xtn2<mode>_le): Rename to...
13852 (aarch64_xtn2<mode>_insn_le): This.
13853 (aarch64_xtn2<mode>_be): Rename to...
13854 (aarch64_xtn2<mode>_insn_be): This.
13855 (vec_pack_trunc_<mode>): Emit truncation instruction instead
13857 * config/aarch64/iterators.md (Vnarrowd): Add Vnarrowd mode
13858 attribute iterator.
13860 2021-06-16 Martin Jambor <mjambor@suse.cz>
13862 PR tree-optimization/100453
13863 * tree-sra.c (create_access): Disqualify any const candidates
13864 which are written to.
13865 (sra_modify_expr): Do not store sub-replacements back to a const base.
13866 (handle_unscalarized_data_in_subtree): Likewise.
13867 (sra_modify_assign): Likewise. Earlier, use TREE_READONLy test
13868 instead of constant_decl_p.
13870 2021-06-16 Jakub Jelinek <jakub@redhat.com>
13872 PR middle-end/101062
13873 * stor-layout.c (finish_bitfield_representative): For fields in unions
13874 assume nextf is always NULL.
13875 (finish_bitfield_layout): Compute bit field representatives also in
13876 unions, but handle it as if each bitfield was the only field in the
13879 2021-06-16 Richard Biener <rguenther@suse.de>
13881 PR tree-optimization/101088
13882 * tree-ssa-loop-im.c (sm_seq_valid_bb): Only look for
13883 supported refs on edges. Do not assert same ref but
13884 different kind stores are unsuported but mark them so.
13885 (hoist_memory_references): Only look for supported refs
13888 2021-06-16 Roger Sayle <roger@nextmovesoftware.com>
13890 PR rtl-optimization/46235
13891 * config/i386/i386.md: New define_split for bt followed by cmov.
13892 (*bt<mode>_setcqi): New define_insn_and_split for bt followed by setc.
13893 (*bt<mode>_setncqi): New define_insn_and_split for bt then setnc.
13894 (*bt<mode>_setnc<mode>): New define_insn_and_split for bt followed
13895 by setnc with zero extension.
13897 2021-06-16 Richard Biener <rguenther@suse.de>
13899 PR tree-optimization/101083
13900 * tree-vect-slp.c (vect_slp_build_two_operator_nodes): Get
13901 vectype as argument.
13902 (vect_build_slp_tree_2): Adjust.
13904 2021-06-15 Martin Sebor <msebor@redhat.com>
13906 PR middle-end/100876
13907 * builtins.c: (gimple_call_return_array): Account for size_t
13908 mangling as either unsigned int or unsigned long
13910 2021-06-15 Jeff Law <jeffreyalaw@gmail.com>
13912 * compare-elim.c (try_eliminate_compare): Run DCE to clean things
13913 up before eliminating comparisons.
13915 2021-06-15 Aldy Hernandez <aldyh@redhat.com>
13917 * range-op.cc (operator_bitwise_or::wi_fold): Make sure
13918 nonzero|X is nonzero.
13919 (range_op_bitwise_and_tests): Add tests for above.
13921 2021-06-15 Carl Love <cel@us.ibm.com>
13924 * config/rs6000/rs6000-builtin.def (VCMPEQUT): Fix the ICODE for the
13926 (VRLQ, VSLQ, VSRQ, VSRAQ): Remove unused BU_P10_OVERLOAD_2
13929 2021-06-15 Tobias Burnus <tobias@codesourcery.com>
13932 * gimplify.c (enum gimplify_defaultmap_kind): Add GDMK_SCALAR_TARGET.
13933 (struct gimplify_omp_ctx): Extend defaultmap array by one.
13934 (new_omp_context): Init defaultmap[GDMK_SCALAR_TARGET].
13935 (omp_notice_variable): Update type classification for Fortran.
13936 (gimplify_scan_omp_clauses): Update calls for new argument; handle
13937 GDMK_SCALAR_TARGET; for Fortran, GDMK_POINTER avoid GOVD_MAP_0LEN_ARRAY.
13938 * langhooks-def.h (lhd_omp_scalar_p): Add 'ptr_ok' argument.
13939 * langhooks.c (lhd_omp_scalar_p): Likewise.
13940 (LANG_HOOKS_OMP_ALLOCATABLE_P, LANG_HOOKS_OMP_SCALAR_TARGET_P): New.
13941 (LANG_HOOKS_DECLS): Add them.
13942 * langhooks.h (struct lang_hooks_for_decls): Add new hooks, update
13943 omp_scalar_p pointer type to include the new bool argument.
13945 2021-06-15 David Malcolm <dmalcolm@redhat.com>
13947 * doc/analyzer.texi
13948 (Special Functions for Debugging the Analyzer): Add
13949 __analyzer_dump_capacity.
13951 2021-06-15 Jakub Jelinek <jakub@redhat.com>
13954 * expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
13955 case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.
13957 2021-06-15 Richard Biener <rguenther@suse.de>
13959 * cfgloopanal.c (mark_irreducible_loops): Use a dominance
13960 check to identify loop latches.
13961 * cfgloop.c (verify_loop_structure): Likewise.
13962 * loop-init.c (apply_loop_flags): Allow marked irreducible
13963 regions even with multiple latches.
13964 * predict.c (rebuild_frequencies): Simplify.
13966 2021-06-15 Richard Biener <rguenther@suse.de>
13968 * tree-ssa-threadupdate.c
13969 (jump_thread_path_registry::mark_threaded_blocks): Assert we
13970 have marked irreducible regions.
13972 2021-06-14 Martin Sebor <msebor@redhat.com>
13975 * builtins.c (gimple_call_return_array): Check for attribute fn spec.
13976 Handle calls to placement new.
13977 (ndecl_dealloc_argno): Avoid placement delete.
13979 2021-06-14 Peter Bergner <bergner@linux.ibm.com>
13982 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
13983 create_tmp_reg_or_ssa_name().
13985 2021-06-14 Andrew MacLeod <amacleod@redhat.com>
13987 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
13988 (ranger_cache::enable_new_values): Set to specified value and
13989 return the old value.
13990 (ranger_cache::disable_new_values): Delete.
13991 (ranger_cache::fill_block_cache): Disable non 1st order derived
13993 * gimple-range-cache.h (ranger_cache): Adjust prototypes.
13994 * gimple-range.cc (gimple_ranger::range_of_expr): Adjust.
13996 2021-06-14 Uroš Bizjak <ubizjak@gmail.com>
13999 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
14000 Return true early when testing with V2HImode.
14001 * config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.
14003 2021-06-14 Christophe Lyon <christophe.lyon@linaro.org>
14005 * config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
14006 (mve_vec_unpack<US>_hi_<mode>): New pattern.
14007 (@mve_vec_pack_trunc_lo_<mode>): New pattern.
14008 (mve_vmovntq_<supf><mode>): Prefix with '@'.
14009 * config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
14011 (vec_unpack<US>_lo_<mode>): Likewise.
14012 (vec_pack_trunc_<mode>): Rename to
14013 neon_quad_vec_pack_trunc_<mode>.
14014 * config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
14016 (vec_unpack<US>_lo_<mode>): New.
14017 (vec_pack_trunc_<mode>): New.
14019 2021-06-14 Richard Biener <rguenther@suse.de>
14021 PR tree-optimization/100934
14022 * tree-ssa-dom.c (pass_dominator::execute): Properly
14023 mark irreducible regions.
14025 2021-06-14 Martin Liska <mliska@suse.cz>
14027 * doc/invoke.texi: Put r{...} on the same line as @item.
14029 2021-06-14 Martin Liska <mliska@suse.cz>
14031 * doc/invoke.texi: Add missing newline.
14033 2021-06-14 Martin Liska <mliska@suse.cz>
14035 * doc/invoke.texi: Remove '+' charasters.
14037 2021-06-14 Claudiu Zissulescu <claziss@synopsys.com>
14039 * config.gcc (arc): Add support for with_cpu option.
14040 * config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.
14042 2021-06-14 Richard Biener <rguenther@suse.de>
14044 PR tree-optimization/101031
14045 * tree-ssa-strlen.c (maybe_invalidate): Increment max_size
14046 instead of size when accounting for a possibly string
14049 2021-06-14 Martin Liska <mliska@suse.cz>
14051 * gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].
14053 2021-06-14 Aldy Hernandez <aldyh@redhat.com>
14055 * value-query.cc (gimple_range_global): Call get_range_global
14056 if called after inlining.
14058 2021-06-13 Uroš Bizjak <ubizjak@gmail.com>
14061 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
14062 Emit constant permutation insn directly from here.
14064 2021-06-13 Trevor Saunders <tbsaunde@tbsaunde.org>
14066 * attribs.c (find_attribute_namespace): Iterate over vec<> with
14068 * auto-profile.c (afdo_find_equiv_class): Likewise.
14069 * gcc.c (do_specs_vec): Likewise.
14070 (do_spec_1): Likewise.
14071 (driver::set_up_specs): Likewise.
14072 * gimple-loop-jam.c (any_access_function_variant_p): Likewise.
14073 * gimple-ssa-store-merging.c (compatible_load_p): Likewise.
14074 (imm_store_chain_info::try_coalesce_bswap): Likewise.
14075 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
14076 (get_location_for_stmts): Likewise.
14077 * graphite-poly.c (print_iteration_domains): Likewise.
14078 (free_poly_bb): Likewise.
14079 (remove_gbbs_in_scop): Likewise.
14080 (free_scop): Likewise.
14081 (dump_gbb_cases): Likewise.
14082 (dump_gbb_conditions): Likewise.
14083 (print_pdrs): Likewise.
14084 (print_scop): Likewise.
14085 * ifcvt.c (cond_move_process_if_block): Likewise.
14086 * lower-subreg.c (decompose_multiword_subregs): Likewise.
14087 * regcprop.c (pass_cprop_hardreg::execute): Likewise.
14088 * sanopt.c (sanitize_rewrite_addressable_params): Likewise.
14089 * sel-sched-dump.c (dump_insn_vector): Likewise.
14090 * store-motion.c (store_ops_ok): Likewise.
14091 (store_killed_in_insn): Likewise.
14092 * timevar.c (timer::named_items::print): Likewise.
14093 * tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
14094 (cleanup_tree_cfg_noloop): Likewise.
14095 * tree-data-ref.c (dump_data_references): Likewise.
14096 (print_dir_vectors): Likewise.
14097 (print_dist_vectors): Likewise.
14098 (dump_data_dependence_relations): Likewise.
14099 (dump_dist_dir_vectors): Likewise.
14100 (dump_ddrs): Likewise.
14101 (create_runtime_alias_checks): Likewise.
14102 (free_subscripts): Likewise.
14103 (save_dist_v): Likewise.
14104 (save_dir_v): Likewise.
14105 (invariant_access_functions): Likewise.
14106 (same_access_functions): Likewise.
14107 (access_functions_are_affine_or_constant_p): Likewise.
14108 (find_data_references_in_stmt): Likewise.
14109 (graphite_find_data_references_in_stmt): Likewise.
14110 (free_dependence_relations): Likewise.
14111 (free_data_refs): Likewise.
14112 * tree-inline.c (copy_debug_stmts): Likewise.
14113 * tree-into-ssa.c (dump_currdefs): Likewise.
14114 (rewrite_update_phi_arguments): Likewise.
14115 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
14116 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
14118 (vect_slp_analyze_node_dependences): Likewise.
14119 (vect_slp_analyze_instance_dependence): Likewise.
14120 (vect_record_base_alignments): Likewise.
14121 (vect_get_peeling_costs_all_drs): Likewise.
14122 (vect_peeling_supportable): Likewise.
14123 * tree-vectorizer.c (vec_info::~vec_info): Likewise.
14124 (vec_info::free_stmt_vec_infos): Likewise.
14126 2021-06-13 Jeff Law <jeffreyalaw@gmail.com>
14128 * config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
14129 (andqi3_1<cczn>): Removed.
14130 (<ors>qi3_1): Do not split for IOR/XOR a single bit.
14131 (H8/SX bit logicals): Split out from other patterns.
14132 * config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
14133 mulqihi3_const_clobber_flags.
14134 (mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly
14136 2021-06-13 H.J. Lu <hjl.tools@gmail.com>
14139 * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
14140 to true if red zone is used.
14141 (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
14142 ix86_red_zone_used.
14143 * config/i386/i386.h (machine_function): Add red_zone_used.
14144 (ix86_red_zone_size): Removed.
14145 (ix86_red_zone_used): New.
14146 * config/i386/i386.md (peephole2 patterns): Replace
14147 ix86_red_zone_size with ix86_red_zone_used.
14149 2021-06-12 Jason Merrill <jason@redhat.com>
14151 * doc/extend.texi (unused variable attribute): Applies to
14152 structure fields as well.
14154 2021-06-12 Eugene Rozenfeld <erozen@microsoft.com>
14156 * auto-profile.c (read_profile): fix a typo in an error string
14158 2021-06-11 Thomas Schwinge <thomas@codesourcery.com>
14160 * tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
14162 * tree-pretty-print.c (dump_omp_clauses): Update.
14163 (dump_generic_node) <OMP_CLAUSE>: Use it.
14165 2021-06-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
14168 * config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
14169 int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
14170 the polymorphic variants matching code.
14171 (__arm_vld1q_z): Likewise.
14172 (__arm_vld2q): Likewise.
14173 (__arm_vld4q): Likewise.
14174 (__arm_vldrbq_gather_offset): Likewise.
14175 (__arm_vldrbq_gather_offset_z): Likewise.
14177 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
14179 PR tree-optimization/96392
14180 * fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.
14182 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
14184 PR tree-optimization/96392
14185 * fold-const.c (fold_real_zero_addition_p): Take both arguments
14186 of the addition or subtraction, not just the zero. Use this
14187 other argument in tests for signaling NaNs and signed zeros.
14188 (tree_expr_maybe_real_minus_zero_p): New predicate.
14189 * fold-const.h (fold_real_zero_addition_p): Update prototype.
14190 (tree_expr_maybe_real_minus_zero_p): New function prototype.
14191 * match.pd: Update calls to fold_real_zero_addition_p.
14192 Replace HONOR_NANS with tree_expr_maybe_nan_p.
14193 Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
14194 Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
14195 * tree-ssa-reassoc.c (eliminate_using_constants): Update
14196 call to fold_real_zero_addition_p.
14198 2021-06-11 Richard Biener <rguenther@suse.de>
14200 PR tree-optimization/101025
14201 * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
14202 all refs that require dependence checking.
14204 2021-06-11 Richard Biener <rguenther@suse.de>
14206 PR tree-optimization/101028
14207 * tree-vect-slp.c (vect_build_slp_tree_2): When SLP
14208 reassoc discovery fails fatally, mark appropriate lanes
14211 2021-06-11 Richard Biener <rguenther@suse.de>
14213 PR tree-optimization/101026
14214 * tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
14215 have a representative for the associated chain nodes.
14217 2021-06-11 Jakub Jelinek <jakub@redhat.com>
14219 PR rtl-optimization/101008
14220 * simplify-rtx.c (relational_result): New function.
14221 (simplify_logical_relational_operation,
14222 simplify_relational_operation): Use it.
14224 2021-06-11 Jakub Jelinek <jakub@redhat.com>
14227 * config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.
14229 2021-06-11 Uroš Bizjak <ubizjak@gmail.com>
14232 * config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
14233 false if the permutation can be implemented with constant
14234 permutation instruction in wider mode.
14235 (canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
14236 Handle V8QImode and V4HImode.
14238 2021-06-11 Martin Liska <mliska@suse.cz>
14240 PR gcov-profile/100788
14241 * common.opt: Add new option.
14242 * coverage.c (coverage_begin_function): Emit warning instead on
14243 the internal compiler error.
14244 * doc/invoke.texi: Document the option.
14245 * toplev.c (process_options): Enable it by default.
14247 2021-06-11 Richard Biener <rguenther@suse.de>
14249 PR middle-end/101009
14250 * tree-data-ref.c (build_classic_dist_vector_1): Make sure
14251 to set *init_b to true when we encounter a constant equal
14253 (compute_affine_dependence): Also dump the actual DR_REF.
14255 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
14257 PR tree-optimization/100984
14258 * gimple-ssa-evrp.c (ssa_equiv_stack): Use auto_vec for
14259 replacements table.
14260 (ssa_equiv_stack::~ssa_equiv_stack): Remove.
14262 2021-06-11 Kewen Lin <linkw@linux.ibm.com>
14264 * config/rs6000/rs6000.md
14265 (floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
14266 define_insn_and_split.
14268 2021-06-11 Richard Biener <rguenther@suse.de>
14270 * tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
14271 to sort operands of the associative chain.
14273 2021-06-11 Richard Biener <rguenther@suse.de>
14275 * system.h (gcc_stablesort_r): Declare.
14276 * sort.cc (gcc_sort_r): Support stable sort.
14277 (gcc_stablesort_r): Define.
14278 * vec.h (vec<>::stablesort): Add.
14280 2021-06-10 Uroš Bizjak <ubizjak@gmail.com>
14283 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
14284 Handle V2SF mode. Emit SHUFPS to fixup unpack-high for V2SF mode.
14285 (expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
14286 (expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
14287 (expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
14288 (expand_vec_perm_interleave2): Handle 64bit modes.
14289 (expand_vec_perm_even_odd_pack): Handle V8QI mode.
14290 (expand_vec_perm_even_odd_1): Ditto.
14291 (ix86_vectorize_vec_perm_const): Ditto.
14292 * config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
14293 * config/i386/sse.md: ... here.
14294 * config/i386/mmx.md (*vec_interleave_lowv2sf):
14295 New insn_and_split pattern.
14296 (*vec_interleave_highv2sf): Ditto.
14297 (mmx_pshufbv8qi3): New insn pattern.
14298 (*mmx_pblendw): Ditto.
14300 2021-06-10 Peter Bergner <bergner@linux.ibm.com>
14302 * config/rs6000/rs6000-builtin.def (build_pair): New built-in.
14303 (build_acc): Likewise.
14304 * config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
14305 source operands in little-endian mode.
14306 (rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
14307 (mma_init_builtins): Likewise.
14308 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
14309 ordering for the MMA assemble and build source operands.
14310 * doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
14312 (__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
14315 2021-06-10 Jeff Law <jeffreyalaw@gmail.com>
14317 * config/h8300/h8300.c (select_cc_mode): Handle MEM. Use
14319 * config/h8300/extensions.md: Replace _clobber_flags patterns
14322 2021-06-10 Robin Dapp <rdapp@linux.ibm.com>
14324 * config/s390/vector.md (vcond_mask_<mode><mode>): Change to
14325 (vcond_mask_<mode><tointvec>): this.
14327 2021-06-10 Andrew Stubbs <ams@codesourcery.com>
14328 Thomas Schwinge <thomas@codesourcery.com>
14330 * omp-builtins.def (BUILT_IN_GOACC_ENTER_EXIT_DATA): Split into...
14331 (BUILT_IN_GOACC_ENTER_DATA, BUILT_IN_GOACC_EXIT_DATA): ... these.
14332 * gimple.h (enum gf_mask): Split
14333 'GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA' into
14334 'GF_OMP_TARGET_KIND_OACC_ENTER_DATA' and
14335 'GF_OMP_TARGET_KIND_OACC_EXIT_DATA'.
14336 (is_gimple_omp_oacc): Update.
14337 * gimple-pretty-print.c (dump_gimple_omp_target): Likewise.
14338 * gimplify.c (gimplify_omp_target_update): Likewise.
14339 * omp-expand.c (expand_omp_target, build_omp_regions_1)
14340 (omp_make_gimple_edges): Likewise.
14341 * omp-low.c (check_omp_nesting_restrictions, lower_omp_target):
14344 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
14346 * value-query.cc (value_query::value_on_edge): Rename name to
14348 (range_query::range_on_edge): Same.
14349 (range_query::value_of_expr): Same.
14350 (range_query::value_on_edge): Same.
14351 * value-query.h (class value_query): Same.
14352 (class range_query): Same.
14354 2021-06-10 Richard Biener <rguenther@suse.de>
14356 PR tree-optimization/101003
14357 * tree-vect-slp.c (vect_build_slp_tree_2): Appropriately
14358 use the pattern stmt defs when linearizing a chain.
14360 2021-06-10 Jakub Jelinek <jakub@redhat.com>
14363 * ifcvt.c (noce_get_alt_condition, noce_try_abs): Use
14364 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
14366 2021-06-10 Clement Chigot <clement.chigot@atos.net>
14368 * config/rs6000/aix71.h (ASM_CPU_SPEC): Add Power10 directive.
14369 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
14371 2021-06-09 Andrew Pinski <apinski@marvell.com>
14373 PR tree-optimization/100925
14374 * match.pd (a ? CST1 : CST2): Limit transformations
14375 that would produce a negative to integeral types only.
14376 Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.
14378 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
14381 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
14383 * doc/tm.texi: Correctly update.
14385 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
14387 * doc/tm.texi: Correctly update.
14389 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
14392 * doc/tm.texi.in (Trampolines): Add a missing blank line.
14394 2021-06-09 Paul Eggert <eggert@cs.ucla.edu>
14397 * doc/invoke.texi (Code Gen Options); Document that -fno-trampolines
14398 and -ftrampolines work only with Ada.
14399 * doc/tm.texi.in (Trampolines): Likewise.
14400 * doc/tm.texi: Regenerated.
14402 2021-06-09 Carl Love <cel@us.ibm.com>
14404 * config/rs6000/altivec.h (vec_signextll, vec_signexti, vec_signextq):
14405 Add define for new builtins.
14406 * config/rs6000/altivec.md(altivec_vreveti2): Add define_expand.
14407 * config/rs6000/rs6000-builtin.def (VSIGNEXTI, VSIGNEXTLL): Add
14408 overloaded builtin definitions.
14409 (VSIGNEXTSB2W, VSIGNEXTSH2W, VSIGNEXTSB2D, VSIGNEXTSH2D,VSIGNEXTSW2D,
14410 VSIGNEXTSD2Q): Add builtin expansions.
14411 (SIGNEXT): Add P10 overload definition.
14412 * config/rs6000/rs6000-call.c (P9V_BUILTIN_VEC_VSIGNEXTI, P9V_BUILTIN_VEC_VSIGNEXTLL,
14413 P10_BUILTIN_VEC_SIGNEXT): Add overloaded argument definitions.
14414 * config/rs6000/vsx.md (vsx_sign_extend_v2di_v1ti): Add define_insn.
14415 (vsignextend_v2di_v1ti, vsignextend_qi_<mode>, vsignextend_hi_<mode>,
14416 vsignextend_si_v2di)[VIlong]: Add define_expand.
14417 Make define_insn vsx_sign_extend_si_v2di visible.
14418 * doc/extend.texi: Add documentation for the vec_signexti,
14419 vec_signextll builtins and vec_signextq.
14421 2021-06-09 Carl Love <cel@us.ibm.com>
14423 * config/rs6000/rs6000.c (__fixkfti, __fixunskfti, __floattikf,
14424 __floatuntikf): Names changed to __fixkfti_sw, __fixunskfti_sw,
14425 __floattikf_sw, __floatuntikf_sw respectively.
14426 * config/rs6000/rs6000.md (floatti<mode>2, floatunsti<mode>2,
14427 fix_trunc<mode>ti2, fixuns_trunc<mode>ti2): Add
14428 define_insn for mode IEEE 128.
14430 2021-06-09 Carl Love <cel@us.ibm.com>
14432 * config/rs6000/altivec.md (altivec_vslq, altivec_vsrq):
14433 Rename to altivec_vslq_<mode>, altivec_vsrq_<mode>, mode VEC_TI.
14434 * config/rs6000/vector.md (VEC_TI): Was named VSX_TI in vsx.md.
14435 (vashlv1ti3): Change to vashl<mode>3, mode VEC_TI.
14436 (vlshrv1ti3): Change to vlshr<mode>3, mode VEC_TI.
14437 * config/rs6000/vsx.md (VSX_TI): Remove define_mode_iterator. Update
14438 uses of VSX_TI to VEC_TI.
14440 2021-06-09 Carl Love <cel@us.ibm.com>
14442 * config/rs6000/dfp.md (floattitd2, fixtdti2): New define_insns.
14444 2021-06-09 Carl Love <cel@us.ibm.com>
14446 * config/rs6000/altivec.h (vec_dive, vec_mod): Add define for new
14448 * config/rs6000/altivec.md (UNSPEC_VMULEUD, UNSPEC_VMULESD,
14449 UNSPEC_VMULOUD, UNSPEC_VMULOSD): New unspecs.
14450 (altivec_eqv1ti, altivec_gtv1ti, altivec_gtuv1ti, altivec_vmuleud,
14451 altivec_vmuloud, altivec_vmulesd, altivec_vmulosd, altivec_vrlq,
14452 altivec_vrlqmi, altivec_vrlqmi_inst, altivec_vrlqnm,
14453 altivec_vrlqnm_inst, altivec_vslq, altivec_vsrq, altivec_vsraq,
14454 altivec_vcmpequt_p, altivec_vcmpgtst_p, altivec_vcmpgtut_p): New
14456 (vec_widen_umult_even_v2di, vec_widen_smult_even_v2di,
14457 vec_widen_umult_odd_v2di, vec_widen_smult_odd_v2di, altivec_vrlqmi,
14458 altivec_vrlqnm): New define_expands.
14459 * config/rs6000/rs6000-builtin.def (VCMPEQUT_P, VCMPGTST_P,
14460 VCMPGTUT_P): Add macro expansions.
14461 (BU_P10V_AV_P): Add builtin predicate definition.
14462 (VCMPGTUT, VCMPGTST, VCMPEQUT, CMPNET, CMPGE_1TI,
14463 CMPGE_U1TI, CMPLE_1TI, CMPLE_U1TI, VNOR_V1TI_UNS, VNOR_V1TI, VCMPNET_P,
14464 VCMPAET_P, VMULEUD, VMULESD, VMULOUD, VMULOSD, VRLQ,
14465 VSLQ, VSRQ, VSRAQ, VRLQNM, DIV_V1TI, UDIV_V1TI, DIVES_V1TI, DIVEU_V1TI,
14466 MODS_V1TI, MODU_V1TI, VRLQMI): New macro expansions.
14467 (VRLQ, VSLQ, VSRQ, VSRAQ, DIVE, MOD): New overload expansions.
14468 * config/rs6000/rs6000-call.c (P10_BUILTIN_VCMPEQUT,
14469 P10V_BUILTIN_CMPGE_1TI, P10V_BUILTIN_CMPGE_U1TI,
14470 P10V_BUILTIN_VCMPGTUT, P10V_BUILTIN_VCMPGTST,
14471 P10V_BUILTIN_CMPLE_1TI, P10V_BUILTIN_VCMPLE_U1TI,
14472 P10V_BUILTIN_DIV_V1TI, P10V_BUILTIN_UDIV_V1TI,
14473 P10V_BUILTIN_VMULESD, P10V_BUILTIN_VMULEUD,
14474 P10V_BUILTIN_VMULOSD, P10V_BUILTIN_VMULOUD,
14475 P10V_BUILTIN_VNOR_V1TI, P10V_BUILTIN_VNOR_V1TI_UNS,
14476 P10V_BUILTIN_VRLQ, P10V_BUILTIN_VRLQMI,
14477 P10V_BUILTIN_VRLQNM, P10V_BUILTIN_VSLQ,
14478 P10V_BUILTIN_VSRQ, P10V_BUILTIN_VSRAQ,
14479 P10V_BUILTIN_VCMPGTUT_P, P10V_BUILTIN_VCMPGTST_P,
14480 P10V_BUILTIN_VCMPEQUT_P, P10V_BUILTIN_VCMPGTUT_P,
14481 P10V_BUILTIN_VCMPGTST_P, P10V_BUILTIN_CMPNET,
14482 P10V_BUILTIN_VCMPNET_P, P10V_BUILTIN_VCMPAET_P,
14483 P10V_BUILTIN_DIVES_V1TI, P10V_BUILTIN_MODS_V1TI,
14484 P10V_BUILTIN_MODU_V1TI):
14485 New overloaded definitions.
14486 (rs6000_gimple_fold_builtin) [P10V_BUILTIN_VCMPEQUT,
14487 P10V_BUILTIN_CMPNET, P10V_BUILTIN_CMPGE_1TI,
14488 P10V_BUILTIN_CMPGE_U1TI, P10V_BUILTIN_VCMPGTUT,
14489 P10V_BUILTIN_VCMPGTST, P10V_BUILTIN_CMPLE_1TI,
14490 P10V_BUILTIN_CMPLE_U1TI]: New case statements.
14491 (rs6000_init_builtins) [bool_V1TI_type_node, int_ftype_int_v1ti_v1ti]:
14493 (altivec_init_builtins): New E_V1TImode case statement.
14494 (builtin_function_type)[P10_BUILTIN_128BIT_VMULEUD,
14495 P10_BUILTIN_128BIT_VMULOUD, P10_BUILTIN_128BIT_DIVEU_V1TI,
14496 P10_BUILTIN_128BIT_MODU_V1TI, P10_BUILTIN_CMPGE_U1TI,
14497 P10_BUILTIN_VCMPGTUT, P10_BUILTIN_VCMPEQUT]: New case statements.
14498 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute) [E_TImode,
14499 E_V1TImode]: New case statements.
14500 * config/rs6000/rs6000.h (rs6000_builtin_type_index): New enum
14501 value RS6000_BTI_bool_V1TI.
14502 * config/rs6000/vector.md (vector_gtv1ti,vector_nltv1ti,
14503 vector_gtuv1ti, vector_nltuv1ti, vector_ngtv1ti, vector_ngtuv1ti,
14504 vector_eq_v1ti_p, vector_ne_v1ti_p, vector_ae_v1ti_p,
14505 vector_gt_v1ti_p, vector_gtu_v1ti_p, vrotlv1ti3, vashlv1ti3,
14506 vlshrv1ti3, vashrv1ti3): New define_expands.
14507 * config/rs6000/vsx.md (UNSPEC_VSX_DIVSQ, UNSPEC_VSX_DIVUQ,
14508 UNSPEC_VSX_DIVESQ, UNSPEC_VSX_DIVEUQ, UNSPEC_VSX_MODSQ,
14509 UNSPEC_VSX_MODUQ): New unspecs.
14510 (mulv2di3, vsx_div_v1ti, vsx_udiv_v1ti, vsx_dives_v1ti,
14511 vsx_diveu_v1ti, vsx_mods_v1ti, vsx_modu_v1ti, xxswapd_v1ti): New
14513 (vcmpnet): New define_expand.
14514 * doc/extend.texi: Add documentation for the new builtins vec_rl,
14515 vec_rlmi, vec_rlnm, vec_sl, vec_sr, vec_sra, vec_mule, vec_mulo,
14516 vec_div, vec_dive, vec_mod, vec_cmpeq, vec_cmpne, vec_cmpgt, vec_cmplt,
14517 vec_cmpge, vec_cmple, vec_all_eq, vec_all_ne, vec_all_gt, vec_all_lt,
14518 vec_all_ge, vec_all_le, vec_any_eq, vec_any_ne, vec_any_gt, vec_any_lt,
14519 vec_any_ge, vec_any_le.
14521 2021-06-09 Carl Love <cel@us.ibm.com>
14523 * config/rs6000/altivec.md (altivec_vrl<VI_char>mi): Fix
14524 bug in argument generation.
14526 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
14528 * config/arm/iterators.md (<supf>): Remove VCLZQ_U, VCLZQ_S.
14530 * config/arm/mve.md (mve_vclzq_<supf><mode>): Add '@' prefix,
14531 remove <supf> iterator.
14532 (mve_vclzq_u<mode>): New.
14533 * config/arm/neon.md (clz<mode>2): Rename to neon_vclz<mode>.
14534 (neon_vclz<mode): Move to ...
14535 * config/arm/unspecs.md (VCLZQ_U, VCLZQ_S): Remove.
14536 * config/arm/vec-common.md: ... here. Add support for MVE.
14538 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
14540 * config/arm/mve.md (mve_vhaddq_<supf><mode>): Prefix with '@'.
14541 (@mve_vrhaddq_<supf><mode): Likewise.
14542 * config/arm/neon.md (neon_v<r>hadd<sup><mode>): Likewise.
14543 * config/arm/vec-common.md (avg<mode>3_floor, uavg<mode>3_floor)
14544 (avg<mode>3_ceil", uavg<mode>3_ceil): New patterns.
14546 2021-06-09 imba-tjd <109224573@qq.com>
14548 * doc/invoke.texi: Fix typo.
14550 2021-06-09 Roger Sayle <roger@nextmovesoftware.com>
14552 PR middle-end/53267
14553 * fold-const-call.c (fold_const_call_sss) [CASE_CFN_FMOD]:
14554 Support evaluation of fmod/fmodf/fmodl at compile-time.
14556 2021-06-09 Richard Biener <rguenther@suse.de>
14558 PR tree-optimization/100981
14559 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
14560 gimple_get_lhs to also handle calls.
14561 * tree-vect-slp-patterns.c (complex_pattern::build): Transfer
14564 2021-06-09 Richard Biener <rguenther@suse.de>
14566 PR tree-optimization/97832
14567 * tree-vectorizer.h (_slp_tree::failed): New.
14568 * tree-vect-slp.c (_slp_tree::_slp_tree): Initialize
14570 (_slp_tree::~_slp_tree): Free failed.
14571 (vect_build_slp_tree): Retain failed nodes and record
14572 matches in them, copying that back out when running
14573 into a cached fail. Dump start and end of discovery.
14574 (dt_sort_cmp): New.
14575 (vect_build_slp_tree_2): Handle associatable chains
14576 together doing more aggressive operand swapping.
14578 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
14581 * config.gcc (gcc_cv_initfini_array): Set to yes for Linux and
14583 * doc/install.texi: Require glibc 2.1 and binutils 2.12 for
14584 Linux and GNU targets.
14586 2021-06-09 Richard Biener <rguenther@suse.de>
14588 * tree-vect-stmts.c (vect_is_simple_use): Always get dt
14591 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
14593 * config/arc/arc.md (loop_end): Change it to
14594 define_insn_and_split.
14596 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
14598 * config/arc/arc.md (maddhisi4): Use VMAC2H instruction.
14599 (machi): New pattern.
14600 (umaddhisi4): Use VMAC2HU instruction.
14601 (umachi): New pattern.
14603 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
14605 * config/arc/arc-protos.h (arc_split_move_p): New prototype.
14606 * config/arc/arc.c (arc_split_move_p): New function.
14607 (arc_split_move): Clean up.
14608 * config/arc/arc.md (movdi_insn): Clean up, use arc_split_move_p.
14609 (movdf_insn): Likewise.
14610 * config/arc/simdext.md (mov<VWH>_insn): Likewise.
14612 2021-06-09 Uroš Bizjak <ubizjak@gmail.com>
14615 * config/i386/i386.c (print_operand_address_as): Rename "no_rip"
14616 argument to "raw". Do not emit segment overrides when "raw" is true.
14618 2021-06-09 Martin Liska <mliska@suse.cz>
14620 * doc/gcov.texi: Create a proper JSON files.
14621 * doc/invoke.texi: Remove dots in order to make it a valid
14624 2021-06-09 Xionghu Luo <luoxhu@linux.ibm.com>
14626 * config/rs6000/rs6000-p8swap.c (pattern_is_rotate64): New.
14627 (insn_is_load_p): Use pattern_is_rotate64.
14628 (insn_is_swap_p): Likewise.
14629 (quad_aligned_load_p): Likewise.
14630 (const_load_sequence_p): Likewise.
14631 (replace_swapped_aligned_load): Likewise.
14632 (recombine_lvx_pattern): Likewise.
14633 (recombine_stvx_pattern): Likewise.
14635 2021-06-09 Andrew MacLeod <amacleod@redhat.com>
14637 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Use a
14638 fur_stmt source record.
14639 * gimple-range.cc (fur_source::get_operand): Generic range query.
14640 (fur_source::get_phi_operand): New.
14641 (fur_source::register_dependency): New.
14642 (fur_source::query): New.
14643 (class fur_edge): New. Edge source for operands.
14644 (fur_edge::fur_edge): New.
14645 (fur_edge::get_operand): New.
14646 (fur_edge::get_phi_operand): New.
14647 (fur_edge::query): New.
14648 (fur_stmt::fur_stmt): New.
14649 (fur_stmt::get_operand): New.
14650 (fur_stmt::get_phi_operand): New.
14651 (fur_stmt::query): New.
14652 (class fur_depend): New. Statement source and process dependencies.
14653 (fur_depend::fur_depend): New.
14654 (fur_depend::register_dependency): New.
14655 (class fur_list): New. List source for operands.
14656 (fur_list::fur_list): New.
14657 (fur_list::get_operand): New.
14658 (fur_list::get_phi_operand): New.
14659 (fold_range): New. Instantiate appropriate fur_source class and fold.
14660 (fold_using_range::range_of_range_op): Use new API.
14661 (fold_using_range::range_of_address): Ditto.
14662 (fold_using_range::range_of_phi): Ditto.
14663 (imple_ranger::fold_range_internal): Use fur_depend class.
14664 (fold_using_range::range_of_ssa_name_with_loop_info): Use new API.
14665 * gimple-range.h (class fur_source): Now a base class.
14666 (class fur_stmt): New.
14667 (fold_range): New prototypes.
14668 (fur_source::fur_source): Delete.
14670 2021-06-08 Andrew Pinski <apinski@marvell.com>
14672 PR tree-optimization/25290
14673 * tree-ssa-phiopt.c (xor_replacement): Delete.
14674 (tree_ssa_phiopt_worker): Delete use of xor_replacement.
14675 (match_simplify_replacement): Allow one cheap preparation
14676 statement that can be moved to before the if.
14678 2021-06-08 Pat Haugen <pthaugen@linux.ibm.com>
14680 * config/rs6000/power10.md (power10-fused-load, power10-fused-store,
14681 power10-fused_alu, power10-fused-vec, power10-fused-branch): New.
14683 2021-06-08 Jeff Law <jeffreyalaw@gmail.com>
14685 * config/h8300/logical.md (andqi3_1): Move BCLR case into define_insn_and_split.
14686 Create length attribute on define_insn_and_split. Only split for cases which we
14688 (andqi3_1<cczn>): Renamed from andqi3_1_clobber_flags. Only handle AND here and
14689 fix length computation.
14690 (b<code><mode>msx): Combine QImode and HImode H8/SX patterns using iterator.
14692 2021-06-08 Richard Biener <rguenther@suse.de>
14694 PR tree-optimization/100923
14695 * tree-ssa-sccvn.c (valueize_refs_1): Take a pointer to
14696 the operand vector to be valueized.
14697 (valueize_refs): Likewise.
14698 (valueize_shared_reference_ops_from_ref): Adjust.
14699 (valueize_shared_reference_ops_from_call): Likewise.
14700 (vn_reference_lookup_3): Likewise.
14701 (vn_reference_lookup_pieces): Likewise. Re-valueize
14702 with honoring availability when we are about to create
14703 the ao_ref and valueized before.
14704 (vn_reference_lookup): Likewise.
14705 (vn_reference_insert_pieces): Adjust.
14707 2021-06-08 Richard Biener <rguenther@suse.de>
14709 * tree-vectorizer.h (_slp_instance::root_stmt): Change to...
14710 (_slp_instance::root_stmts): ... a vector.
14711 (SLP_INSTANCE_ROOT_STMT): Rename to ...
14712 (SLP_INSTANCE_ROOT_STMTS): ... this.
14713 (slp_root::root): Change to...
14714 (slp_root::roots): ... a vector.
14715 (slp_root::slp_root): Adjust.
14716 * tree-vect-slp.c (_slp_instance::location): Adjust.
14717 (vect_free_slp_instance): Release the root stmt vector.
14718 (vect_build_slp_instance): Adjust.
14719 (vect_analyze_slp): Likewise.
14720 (_bb_vec_info::~_bb_vec_info): Likewise.
14721 (vect_slp_analyze_operations): Likewise.
14722 (vect_bb_vectorization_profitable_p): Likewise. Adjust
14723 costs for the root stmt.
14724 (vect_slp_check_for_constructors): Gather all BIT_INSERT_EXPRs
14726 (vect_slp_analyze_bb_1): Simplify by marking all root stmts
14728 (vectorize_slp_instance_root_stmt): Adjust.
14729 (vect_schedule_slp): Likewise.
14731 2021-06-08 Aldy Hernandez <aldyh@redhat.com>
14733 * gimple-ssa-evrp.c (class ssa_equiv_stack): New.
14734 (ssa_equiv_stack::ssa_equiv_stack): New.
14735 (ssa_equiv_stack::~ssa_equiv_stack): New.
14736 (ssa_equiv_stack::enter): New.
14737 (ssa_equiv_stack::leave): New.
14738 (ssa_equiv_stack::push_replacement): New.
14739 (ssa_equiv_stack::get_replacement): New.
14740 (is_pointer_ssa): New.
14741 (class pointer_equiv_analyzer): New.
14742 (pointer_equiv_analyzer::pointer_equiv_analyzer): New.
14743 (pointer_equiv_analyzer::~pointer_equiv_analyzer): New.
14744 (pointer_equiv_analyzer::set_global_equiv): New.
14745 (pointer_equiv_analyzer::set_cond_equiv): New.
14746 (pointer_equiv_analyzer::get_equiv): New.
14747 (pointer_equiv_analyzer::enter): New.
14748 (pointer_equiv_analyzer::leave): New.
14749 (pointer_equiv_analyzer::get_equiv_expr): New.
14750 (pta_valueize): New.
14751 (pointer_equiv_analyzer::visit_stmt): New.
14752 (pointer_equiv_analyzer::visit_edge): New.
14753 (hybrid_folder::value_of_expr): Call PTA.
14754 (hybrid_folder::value_on_edge): Same.
14755 (hybrid_folder::pre_fold_bb): New.
14756 (hybrid_folder::post_fold_bb): New.
14757 (hybrid_folder::pre_fold_stmt): New.
14758 (rvrp_folder::pre_fold_bb): New.
14759 (rvrp_folder::post_fold_bb): New.
14760 (rvrp_folder::pre_fold_stmt): New.
14761 (rvrp_folder::value_of_expr): Call PTA.
14762 (rvrp_folder::value_on_edge): Same.
14764 2021-06-08 Jakub Jelinek <jakub@redhat.com>
14767 * tree-inline.c (copy_tree_body_r): For OMP_CLAUSE_DEPEND don't
14768 check TREE_CODE if OMP_CLAUSE_DECL is NULL.
14770 2021-06-08 Richard Biener <rguenther@suse.de>
14772 PR middle-end/100951
14773 * tree-vect-generic.c (expand_vector_piecewise): Build a
14774 VECTOR_CST if all elements are constant.
14775 (expand_vector_condition): Likewise.
14776 (lower_vec_perm): Likewise.
14777 (expand_vector_conversion): Likewise.
14779 2021-06-08 Martin Liska <mliska@suse.cz>
14781 * doc/invoke.texi: Document new param evrp-sparse-threshold.
14783 2021-06-08 Martin Liska <mliska@suse.cz>
14785 * genautomata.c (create_automata): Fix typo.
14787 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
14789 PR tree-optimization/100794
14790 * tree-predcom.c (tree_predictive_commoning_loop): Add parameter
14791 allow_unroll_p and only allow unrolling when it's true.
14792 (tree_predictive_commoning): Add parameter allow_unroll_p and
14794 (run_tree_predictive_commoning): Likewise.
14795 (pass_predcom::gate): Check flag_tree_loop_vectorize and
14796 global_options_set.x_flag_predictive_commoning.
14797 (pass_predcom::execute): Adjust for allow_unroll_p.
14799 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
14801 * tree-predcom.c (execute_pred_commoning): Remove update_ssa call.
14802 (tree_predictive_commoning_loop): Factor some cleanup stuffs into
14803 lambda function cleanup, remove scev_reset call, and adjust return
14805 (tree_predictive_commoning): Adjust for different changed values,
14806 only set flag TODO_update_ssa_only_virtuals if changed.
14807 (pass_data pass_data_predcom): Remove TODO_update_ssa_only_virtuals
14808 from todo_flags_finish.
14810 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
14812 * gimple-range-cache.cc (class sbr_sparse_bitmap): New.
14813 (sbr_sparse_bitmap::sbr_sparse_bitmap): New.
14814 (sbr_sparse_bitmap::bitmap_set_quad): New.
14815 (sbr_sparse_bitmap::bitmap_get_quad): New.
14816 (sbr_sparse_bitmap::set_bb_range): New.
14817 (sbr_sparse_bitmap::get_bb_range): New.
14818 (sbr_sparse_bitmap::bb_range_p): New.
14819 (block_range_cache::block_range_cache): initialize bitmap obstack.
14820 (block_range_cache::~block_range_cache): Destruct obstack.
14821 (block_range_cache::set_bb_range): Decide when to utilze the
14822 sparse on entry cache.
14823 * gimple-range-cache.h (block_range_cache): Add bitmap obstack.
14824 * params.opt (-param=evrp-sparse-threshold): New.
14826 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
14828 * bitmap.c (bitmap_set_aligned_chunk): New.
14829 (bitmap_get_aligned_chunk): New.
14830 (test_aligned_chunk): New.
14831 (bitmap_c_tests): Call test_aligned_chunk.
14832 * bitmap.h (bitmap_set_aligned_chunk, bitmap_get_aligned_chunk): New.
14834 2021-06-07 Uroš Bizjak <ubizjak@gmail.com>
14837 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
14839 (ix86_expand_vector_init_one_nonzero): Ditto.
14840 (ix86_expand_vector_init_one_var): Ditto.
14841 (ix86_expand_vector_init_general): Ditto.
14842 * config/i386/mmx.md (vec_initv4qiqi): New expander.
14844 2021-06-07 Jeff Law <jeffreyalaw@gmail.com>
14846 * config/h8300/movepush.md: Change most _clobber_flags
14847 patterns to instead use <cczn> subst.
14848 (movsi_cczn): New pattern with usable CC cases split out.
14849 (movsi_h8sx_cczn): Likewise.
14851 2021-06-07 Martin Liska <mliska@suse.cz>
14853 * common/common-target.def: Split long lines and replace them
14855 * target.def: Likewise.
14856 * doc/tm.texi: Re-generated.
14858 2021-06-07 Jakub Jelinek <jakub@redhat.com>
14861 * fold-const.c (fold_read_from_vector): Return NULL if trying to
14862 read from a CONSTRUCTOR with vector type elements.
14864 2021-06-07 Jakub Jelinek <jakub@redhat.com>
14866 PR middle-end/100898
14867 * tree-inline.c (copy_bb): Only use gimple_call_arg_ptr if memcpy
14868 should copy any arguments. Don't call gimple_call_num_args
14869 on id->call_stmt or call_stmt more than once.
14871 2021-06-07 liuhongt <hongtao.liu@intel.com>
14874 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): Refine
14876 (<insn>v4siv4di2): Delete constraints for define_expand.
14878 2021-06-07 liuhongt <hongtao.liu@intel.com>
14881 * config/i386/i386-expand.c (ix86_expand_builtin): Remove
14882 assignment of cfun->machine->has_explicit_vzeroupper.
14883 * config/i386/i386-features.c
14884 (ix86_add_reg_usage_to_vzerouppers): Delete.
14885 (ix86_add_reg_usage_to_vzeroupper): Ditto.
14886 (rest_of_handle_insert_vzeroupper): Remove
14887 ix86_add_reg_usage_to_vzerouppers, add df_analyze at the end
14889 (gate): Remove cfun->machine->has_explicit_vzeroupper.
14890 * config/i386/i386-protos.h (ix86_expand_avx_vzeroupper):
14892 * config/i386/i386.c (ix86_insn_callee_abi): New function.
14893 (ix86_initialize_callee_abi): Ditto.
14894 (ix86_expand_avx_vzeroupper): Ditto.
14895 (ix86_hard_regno_call_part_clobbered): Adjust for vzeroupper
14897 (TARGET_INSN_CALLEE_ABI): Define as ix86_insn_callee_abi.
14898 (ix86_emit_mode_set): Call ix86_expand_avx_vzeroupper
14900 * config/i386/i386.h (struct GTY(()) machine_function): Delete
14901 has_explicit_vzeroupper.
14902 * config/i386/i386.md (enum unspec): New member
14904 (ABI_DEFAULT,ABI_VZEROUPPER,ABI_UNKNOWN): New
14905 define_constants for insn callee abi index.
14906 * config/i386/predicates.md (vzeroupper_pattern): Adjust.
14907 * config/i386/sse.md (UNSPECV_VZEROUPPER): Deleted.
14908 (avx_vzeroupper): Call ix86_expand_avx_vzeroupper.
14909 (*avx_vzeroupper): Rename to ..
14910 (avx_vzeroupper_callee_abi): .. this, and adjust pattern as
14911 call_insn which has a special vzeroupper ABI.
14912 (*avx_vzeroupper_1): Deleted.
14914 2021-06-07 liuhongt <hongtao.liu@intel.com>
14917 * df-scan.c (df_get_call_refs): When call_insn is a fake call,
14918 it won't use stack pointer reg.
14919 * final.c (leaf_function_p): When call_insn is a fake call, it
14920 won't affect caller as a leaf function.
14921 * reg-stack.c (callee_clobbers_any_stack_reg): New.
14922 (subst_stack_regs): When call_insn doesn't clobber any stack
14923 reg, don't clear the arguments.
14924 * rtl.c (shallow_copy_rtx): Don't clear flag used when orig is
14926 * shrink-wrap.c (requires_stack_frame_p): No need for stack
14927 frame for a fake call.
14928 * rtl.h (FAKE_CALL_P): New macro.
14930 2021-06-06 Eric Botcazou <ebotcazou@adacore.com>
14932 * config/sparc/sparc-protos.h (order_regs_for_local_alloc): Rename
14934 (sparc_order_regs_for_local_alloc): ...this.
14935 (sparc_leaf_reg_remap): Declare.
14936 * config/sparc/sparc.h (ADJUST_REG_ALLOC_ORDER): Adjust.
14937 (LEAF_REG_REMAP): Reimplement as call to sparc_leaf_reg_remap.
14938 * config/sparc/sparc.c (leaf_reg_remap): Delete.
14939 (order_regs_for_local_alloc): Rename to...
14940 (sparc_order_regs_for_local_alloc): ...this.
14941 (sparc_leaf_reg_remap): New function.
14942 (sparc_conditional_register_usage): Do not modify leaf_reg_remap.
14944 2021-06-06 David Edelsohn <dje.gcc@gmail.com>
14946 * config/rs6000/rs6000.c (rs6000_xcoff_asm_output_aligned_decl_common):
14947 Use assemble_name to output BSS section name.
14949 2021-06-06 Uroš Bizjak <ubizjak@gmail.com>
14951 * config/i386/constraints.md (Bs):
14952 Remove boolean operators from match_test RTX.
14955 (M): Use "mode" variable instead of GET_MODE (op) in match_test RTX.
14958 2021-06-06 Martin Liska <mliska@suse.cz>
14960 * doc/extend.texi: Add missing @headitem.
14961 * doc/invoke.texi: Likewise.
14962 * doc/objc.texi: Likewise.
14964 2021-06-06 Martin Liska <mliska@suse.cz>
14966 * genhooks.c (emit_findices): Remove unused function.
14967 (emit_documentation): Do not call emit_findices
14968 and do not search for @Fcode directives.
14970 2021-06-06 Martin Liska <mliska@suse.cz>
14972 * doc/invoke.texi: Remove extra character.
14974 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
14976 * config/sh/sh.md (doloop_end_split): Fix empty split condition.
14978 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
14980 * config/sparc/sparc.md (*snedi<W:mode>_zero_vis3,
14981 *neg_snedi<W:mode>_zero_subxc, *plus_snedi<W:mode>_zero,
14982 *plus_plus_snedi<W:mode>_zero, *minus_snedi<W:mode>_zero,
14983 *minus_minus_snedi<W:mode>_zero): Fix empty split condition.
14985 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
14987 * config/or1k/or1k.md (*movdi): Fix empty split condition.
14989 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
14991 * config/mips/mips.md (<anonymous>, bswapsi2, bswapdi2): Fix empty
14994 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
14996 * config/m68k/m68k.md (*zero_extend_inc, *zero_extend_dec,
14997 *zero_extendsidi2): Fix empty split condition.
14999 2021-06-05 Jeff Law <jeffreyalaw@gmail.com>
15001 * config/h8300/addsub.md: Fix split condition in define_insn_and_split
15003 * config/h8300/bitfield.md: Likewise.
15004 * config/h8300/combiner.md: Likewise.
15005 * config/h8300/divmod.md: Likewise.
15006 * config/h8300/extensions.md: Likewise.
15007 * config/h8300/jumpcall.md: Likewise.
15008 * config/h8300/movepush.md: Likewise.
15009 * config/h8300/multiply.md: Likewise.
15010 * config/h8300/other.md: Likewise.
15011 * config/h8300/shiftrotate.md: Likewise.
15012 * config/h8300/logical.md: Likewise. Fix split pattern to use
15013 code iterator that somehow slipped through.
15015 2021-06-04 Tobias Burnus <tobias@codesourcery.com>
15017 PR middle-end/100905
15018 * tree-nested.c (convert_nonlocal_omp_clauses,
15019 convert_local_omp_clauses): Handle OMP_CLAUSE_BIND.
15021 2021-06-04 Martin Sebor <msebor@redhat.com>
15023 PR middle-end/100732
15024 * gimple-fold.c (gimple_fold_builtin_sprintf): Avoid folding calls
15025 with either source or destination argument of invalid type.
15026 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Avoid checking
15027 calls with arguments of invalid type.
15029 2021-06-04 Martin Sebor <msebor@redhat.com>
15031 * attribs.c (init_attr_rdwr_indices): Use VLA bounds in the expected
15033 (attr_access::vla_bounds): Also handle VLA bounds.
15035 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
15037 * config/i386/predicates.md (GOT_memory_operand):
15038 Implement using match_code RTXes.
15039 (GOT32_symbol_operand): Ditto.
15041 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
15044 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
15046 (ix86_expand_vector_init_general): Ditto.
15047 Use SImode instead of word_mode for logic operations
15048 when GET_MODE_SIZE (mode) < UNITS_PER_WORD.
15049 (expand_vec_perm_even_odd_1): Assert that V2HI mode should be
15050 implemented by expand_vec_perm_1.
15051 (expand_vec_perm_broadcast_1): Assert that V2HI and V4HI modes
15052 should be implemented using standard shuffle patterns.
15053 (ix86_vectorize_vec_perm_const): Handle V2HImode. Add V4HI and
15054 V2HI modes to modes, implementable with shuffle for one operand.
15055 * config/i386/mmx.md (*punpckwd): New insn_and_split pattern.
15056 (*pshufw_1): New insn pattern.
15057 (*vec_dupv2hi): Ditto.
15058 (vec_initv2hihi): New expander.
15060 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
15062 * config/arm/vfp.md (no_literal_pool_df_immediate,
15063 no_literal_pool_sf_immediate): Fix empty split condition.
15065 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
15067 * config/i386/i386.md (*load_tp_x32_zext, *add_tp_x32_zext,
15068 *tls_dynamic_gnu2_combine_32): Fix empty split condition.
15069 * config/i386/sse.md (*<sse2_avx2>_pmovmskb_lt,
15070 *<sse2_avx2>_pmovmskb_zext_lt, *sse2_pmovmskb_ext_lt,
15071 *<sse4_1_avx2>_pblendvb_lt): Likewise.
15073 2021-06-04 Jakub Jelinek <jakub@redhat.com>
15076 * config/i386/i386-expand.c (ix86_expand_vector_init): Handle
15077 concatenation from half-sized modes with TImode elements.
15079 2021-06-04 Claudiu Zissulescu <claziss@synopsys.com>
15081 * config/arc/arc.c (arc_override_options): Disable millicode
15082 thunks when RF16 is on.
15084 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
15086 * config/rs6000/rs6000.h (PROMOTE_MODE): Remove.
15088 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
15090 * config/rs6000/rs6000-call.c (rs6000_promote_function_mode):
15091 Replace PROMOTE_MODE marco with its content.
15093 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
15095 * config/cris/cris.md (*addi_reload): Fix empty split condition.
15097 2021-06-03 Jim Wilson <jimw@sifive.com>
15099 * config.gcc (riscv*-*-*): If --with-riscv-attribute not used,
15100 turn it on for all riscv targets.
15102 2021-06-03 Uroš Bizjak <ubizjak@gmail.com>
15105 * config/i386/i386-expand.c (ix86_expand_vector_set):
15106 Handle V2HI and V4QI modes.
15107 (ix86_expand_vector_extract): Ditto.
15108 * config/i386/mmx.md (*pinsrw): New insn pattern.
15111 (*pextrw_zext): Ditto.
15113 (*pextrb_zext): Ditto.
15114 (vec_setv2hi): New expander.
15115 (vec_extractv2hihi): Ditto.
15116 (vec_setv4qi): Ditto.
15117 (vec_extractv4qiqi): Ditto.
15118 (vec_setv8qi): Enable only for TARGET_SSE4_1.
15119 (vec_extractv8qiqi): Ditto.
15121 2021-06-03 Aaron Sawdey <acsawdey@linux.ibm.com>
15123 * config/rs6000/genfusion.pl (gen_logical_addsubf): Fix input
15124 order to subf instruction.
15125 * config/rs6000/fusion.md: Regenerate.
15127 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
15129 * calls.c (get_size_range): Use range_of_expr instead of
15130 determine_value_range.
15131 * tree-affine.c (expr_to_aff_combination): Same.
15132 * tree-data-ref.c (split_constant_offset): Same.
15133 * tree-vrp.c (determine_value_range_1): Remove.
15134 (determine_value_range): Remove.
15135 * tree-vrp.h (determine_value_range): Remove.
15137 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
15139 * function-tests.c (test_ranges): Call gimple_range_tests.
15140 * gimple-range-cache.cc (ranger_cache::range_of_expr): Pass stmt
15142 * gimple-range.cc (fur_source::get_operand): Do not call
15143 get_tree_range or gimple_range_global.
15145 (get_tree_range): Move to value-query.cc.
15146 Call get_arith_expr_range.
15147 (gimple_ranger::range_of_expr): Add argument to get_tree_range.
15148 Include gimple-range-tests.cc.
15149 * gimple-range.h (fold_range): Add argument.
15150 (get_tree_range): Remove.
15151 * selftest.h (gimple_range_tests): New.
15152 * value-query.cc (global_range_query::range_of_expr): Add
15154 (range_query::get_tree_range): Move from gimple-range.cc.
15155 * value-query.h (class range_query): Add get_tree_range and
15156 get_arith_expr_range. Make fur_source a friend.
15157 * vr-values.c (vr_values::range_of_expr): Pass stmt to
15159 * gimple-range-tests.cc: New file.
15161 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
15163 * gimple-range.cc (gimple_ranger::export_global_ranges): Call
15164 update_global_range.
15165 * value-query.cc (update_global_range): New.
15166 * value-query.h (update_global_range): New.
15168 2021-06-03 David Malcolm <dmalcolm@redhat.com>
15170 * diagnostic-show-locus.c (diagnostic_show_locus): Don't reject
15171 printing the same location twice if there are fix-it hints,
15172 multiple locations, or a label.
15174 2021-06-03 Andre Vieira <andre.simoesdiasvieira@arm.com>
15176 * tree-vect-loop.c (vect_transform_loop): Use main loop's various'
15177 thresholds to narrow the upper bound on epilogue iterations.
15179 2021-06-03 Christophe Lyon <christophe.lyon@linaro.org>
15181 * config/arm/mve.md (mve_vabsq_f<mode>): Use 'abs' instead of unspec.
15182 (mve_vabsq_s<mode>): Likewise.
15183 * config/arm/neon.md (abs<mode>2): Rename to neon_abs<mode>2.
15184 * config/arm/unspecs.md (VABSQ_F, VABSQ_S): Delete.
15185 * config/arm/vec-common.md (neg<mode>2): Rename to
15186 <absneg_str><mode>2.
15188 2021-06-03 Claudiu Zissulescu <claziss@synopsys.com>
15190 * common/config/arc/arc-common.c (arc_option_optimization_table):
15191 Remove malign-call.
15192 * config/arc/arc.c (arc_unalign_branch_p): Remove unused function.
15193 * config/arc/arc.h (TARGET_MIXED_CODE): Remove macro.
15194 (INDEX_REG_CLASS): Only refer to GENERAL_REGS.
15195 * config/arc/arc.md (abssi2_mixed): Remove pattern.
15196 * config/arc/arc.opt (munalign-prob-threshold): Mark it obsolete.
15197 (malign-call): Likewise.
15198 (mmixed-code): Likewise.
15199 * doc/invoke.texi (ARC): Update doc.
15201 2021-06-03 Martin Liska <mliska@suse.cz>
15203 * common.opt: Use proper Enum values.
15204 * opts.c (COVERAGE_SANITIZER_OPT): Remove.
15205 (parse_sanitizer_options): Handle only sanitizer_opts.
15206 (common_handle_option): Just assign value.
15208 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
15211 * tree-inline.c (inline_forbidden_p): Remove test on return type.
15213 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
15215 * dwarf2out.c (loc_list_from_tree_1) <FUNCTION_DECL>: Also generate
15216 DW_OP_GNU_variable_value referencing an existing DIE at file scope.
15217 (type_byte_size): Inline into...
15218 (add_byte_size_attribute): ...this and call add_scalar_info.
15220 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
15222 * dwarf2out.c (mem_loc_descriptor) <UDIV>: Fix typo.
15223 (typed_binop_from_tree): New function.
15224 (loc_list_from_tree_1) <EXACT_DIV_EXPR>: For an unsigned type,
15225 turn a divide by a power of 2 into a shift.
15226 <CEIL_DIV_EXPR>: For an unsigned type, use a signed divide if the
15227 size of the mode is lower than DWARF2_ADDR_SIZE; otherwise, do a
15228 typed divide by calling typed_binop_from_tree.
15230 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
15232 * dwarf2out.c (scompare_loc_descriptor): Fix head comment.
15233 (is_handled_procedure_type): Likewise.
15234 (struct loc_descr_context): Add strict_signedness field.
15235 (resolve_args_picking_1): Deal with DW_OP_[GNU_]deref_type,
15236 DW_OP_[GNU_]convert and DW_OP_[GNU_]reinterpret.
15237 (resolve_args_picking): Minor tweak.
15238 (function_to_dwarf_procedure): Initialize strict_signedness field.
15239 (type_byte_size): Likewise.
15240 (field_byte_offset): Likewise.
15241 (gen_descr_array_type_die): Likewise.
15242 (gen_variant_part): Likewise.
15243 (loc_list_from_tree_1) <CALL_EXPR>: Tidy up and set strict_signedness
15244 to true when a context is present before evaluating the arguments.
15245 <COND_EXPR>: Do not generate a useless comparison with zero.
15246 When dereferencing an address, if strict_signedness is true and the
15247 type is small and signed, use DW_OP_deref_type to do the dereference
15248 and then DW_OP_convert to convert back to the generic type.
15250 2021-06-03 Jakub Jelinek <jakub@redhat.com>
15253 * tree-inline.c (copy_tree_body_r): Handle iterators on
15254 OMP_CLAUSE_AFFINITY or OMP_CLAUSE_DEPEND.
15256 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
15258 * config/arc/arc.md (*bbit_di): Remove.
15260 2021-06-02 Christoph Muellner <cmuellner@gcc.gnu.org>
15262 PR rtl-optimization/100264
15263 * ree.c (get_sub_rtx): Ignore SET expressions without register
15264 destinations and remove assertion, as it is not valid anymore
15265 with this new behaviour.
15266 (merge_def_and_ext): Eliminate destination check for register
15267 as such SET expressions can't occur anymore.
15268 (combine_reaching_defs): Likewise.
15270 2021-06-02 Jakub Jelinek <jakub@redhat.com>
15273 * config/xtensa/xtensa.h (LEAF_REG_REMAP): Cast REGNO to int to avoid
15274 -Wtype-limits warnings.
15275 (DWARF_FRAME_REGISTER): Rewrite into ternary operator with addition
15276 in operands to avoid -Wsign-compare warnings.
15278 2021-06-02 Pat Haugen <pthaugen@linux.ibm.com>
15280 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): Use
15283 2021-06-02 Vineet Gupta <vgupta@synopsys.com>
15285 * config/arc/arc.h (TARGET_CPU_DEFAULT): Change to hs38_linux.
15287 2021-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
15289 * config/s390/s390.md(*ashrdi3_31<setcc><cconly>): Use a single
15291 * config/s390/subst.md(cconly_subst): Use a single constraint
15292 in (match_scratch).
15294 2021-06-02 Martin Liska <mliska@suse.cz>
15296 * ipa-icf.h: Use auto_vec for memory_access_types.
15298 2021-06-02 Jeff Law <jeffreyalaw@gmail.com>
15300 * config/h8300/h8300-protos.h (compute_a_shift_length): Drop unused
15301 argument from prototype.
15302 (output_logical_op): Add rtx_code argument.
15303 (compute_logical_op_length): Likewise.
15304 * config/h8300/h8300.c (h8300_and_costs): Pass additional argument
15305 to compute_a_shift_length.
15306 (output_logical_op); New argument with the rtx code rather than
15307 extracting it from an operand. Handle QImode too.
15308 (compute_logical_op_length): Similary.
15309 (compute_a_shift_length): Drop unused argument.
15310 * config/h8300/h8300.md (logicals): New code iterator.
15311 * config/h8300/logical.md (<code><mode>3 expander): Combine
15312 the "and" expander with the "ior"/"xor" expander.
15313 (bclr<mode>msx): Combine the QI/HI mode patterns.
15314 (<logical><mode>3 insns): Use code iterator rather than match_operator.
15315 Handle QImode as well. Update call to output_logical_op and
15316 compute_logical_op_length to pass in rtx_code
15317 Fix split condition on all define_insn_and_split patterns.
15318 (one_cmpl<mode>2<cczn>): Use <cczn> to support both clobbering
15319 the flags and setting ZN via existing define_subst.
15320 * config/h8300/shiftrotate.md: Drop unused argument from
15321 calls to compute_a_shift_length.
15322 Signed-off-by: Jeff Law <jeffreyalaw@gmail.com>
15324 2021-06-01 Andrew Pinski <apinski@marvell.com>
15326 PR tree-optimization/25290
15327 * tree-ssa-phiopt.c (match_simplify_replacement):
15329 (tree_ssa_phiopt_worker): Use match_simplify_replacement.
15330 (two_value_replacement): Change the comment about
15331 conditional_replacement.
15332 (conditional_replacement): Delete.
15334 2021-06-01 Andrew Pinski <apinski@marvell.com>
15336 PR tree-optimization/95481
15337 * tree-tailcall.c (find_tail_calls): Handle empty typed
15340 2021-06-01 Andrew Pinski <apinski@marvell.com>
15342 * gimplify.c (zero_sized_field_decl): Delete
15343 (zero_sized_type): Delete
15344 (gimplify_init_ctor_eval): Use is_empty_type instead
15345 of zero_sized_field_decl.
15346 (gimplify_modify_expr): Use is_empty_type instead of
15349 2021-06-01 Jason Merrill <jason@redhat.com>
15352 * tree.h (CALL_FROM_NEW_OR_DELETE_P): Adjust comment.
15354 2021-06-01 Jason Merrill <jason@redhat.com>
15357 * diagnostic.h (warning_enabled_at): Declare.
15358 * diagnostic.c (diagnostic_enabled): Factor out from...
15359 (diagnostic_report_diagnostic): ...here.
15360 (warning_enabled_at): New.
15362 2021-06-01 Aldy Hernandez <aldyh@redhat.com>
15364 * gimple-ssa-evrp.c: Enable exporting of global ranges.
15366 2021-06-01 Martin Liska <mliska@suse.cz>
15369 * doc/invoke.texi: Mention that -fgcse-after-reload
15370 is enabled with -O3.
15372 2021-06-01 liuhongt <hongtao.liu@intel.com>
15374 PR tree-optimization/98365
15375 * tree-if-conv.c (strip_nop_cond_scalar_reduction): New function.
15376 (is_cond_scalar_reduction): Handle nop_expr in cond scalar reduction.
15377 (convert_scalar_cond_reduction): Ditto.
15378 (predicate_scalar_phi): Ditto.
15380 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
15382 PR tree-optimization/100781
15383 * gimple-range-cache.cc (ranger_cache::ranger_cache): Enable new
15384 value calculation by default.
15385 (ranger_cache::enable_new_values): New.
15386 (ranger_cache::disable_new_values): New.
15387 (ranger_cache::push_poor_value): Check if new values are allowed.
15388 * gimple-range-cache.h (class ranger_cache): New member/methods.
15389 * gimple-range.cc (gimple_ranger::range_of_expr): Check for debug
15390 statement, and disable/renable new value calculation.
15392 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
15394 * gimple-range-cache.cc (ranger_cache::ssa_range_in_bb): Delete.
15395 (ranger_cache::range_of_def): New.
15396 (ranger_cache::entry_range): New.
15397 (ranger_cache::exit_range): New.
15398 (ranger_cache::range_of_expr): Adjust.
15399 (ranger_cache::range_on_edge): Adjust.
15400 (ranger_cache::propagate_cache): Call exit_range directly.
15401 * gimple-range-cache.h (class ranger_cache): Adjust.
15403 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
15405 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust for
15406 gori_compute being a member rather than base class.
15407 dervied call to member call.
15408 (ranger_cache::dump): No longer dump gori_map.
15409 (ranger_cache::dump_bb): New.
15410 (ranger_cache::get_non_stale_global_range): Adjust for gori_compute
15411 being a member rather than base class.
15412 (ranger_cache::set_global_range): Ditto.
15413 (ranger_cache::ssa_range_in_bb): Ditto.
15414 (ranger_cache::range_of_expr): New.
15415 (ranger_cache::range_on_edge): New.
15416 (ranger_cache::block_range): Adjust for gori_computes. Debug changes.
15417 (ranger_cache::propagate_cache): Adjust debugging output.
15418 (ranger_cache::fill_block_cache): Adjust for gori_computes. Debug
15420 * gimple-range-cache.h (class ranger_cache): Make gori_compute a
15421 member, and inherit from range_query instead.
15422 (ranger_cache::dump_bb): New. split from dump.
15423 * gimple-range-gori.cc (gori_compute::ssa_range_in_bb): Delete.
15424 (gori_compute::expr_range_at_stmt): Delete.
15425 (gori_compute::compute_name_range_op): Delete.
15426 (gori_compute::compute_operand_range_switch): Add fur_source.
15427 (gori_compute::compute_operand_range): Add fur_source param, inline
15428 old compute_name_range_op and optimize_logical_operands.
15429 (struct tf_range): Delete.
15430 (gori_compute::logical_combine): Adjust
15431 (gori_compute::optimize_logical_operands): Delete.
15432 (gori_compute::compute_logical_operands_in_chain): Delete.
15433 (gori_compute::compute_logical_operands): Adjust.
15434 (gori_compute::compute_operand1_range): Adjust to fur_source.
15435 (gori_compute::compute_operand2_range): Ditto.
15436 (gori_compute::compute_operand1_and_operand2_range): Ditto.
15437 (gori_compute::outgoing_edge_range_p): Add range_query parameter,
15438 and adjust to fur_source.
15439 * gimple-range-gori.h (class gori_compute): Simplify and adjust to
15440 range_query and fur_source.
15441 * gimple-range.cc (gimple_ranger::range_on_edge): Query range_on_edge
15442 from the ranger_cache..
15443 (gimple_ranger::fold_range_internal): Adjust to base class change of
15445 (gimple_ranger::dump_bb): Adjust dump.
15446 * gimple-range.h (gimple_ranger):export gori computes object.
15448 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
15450 PR tree-optimization/100774
15451 * gimple-range-cache.cc (ranger_cache::get_non_stale_global_range):
15452 Constant values are also not stale.
15453 (ranger_cache::set_global_range): Range invariant values should also
15454 have the correct timestamp.
15456 2021-05-31 Martin Liska <mliska@suse.cz>
15458 * tree-streamer-in.c (unpack_ts_function_decl_value_fields):
15459 Unpack FUNCTION_DECL_DECL_TYPE.
15460 * tree-streamer-out.c (pack_ts_function_decl_value_fields):
15461 Stream FUNCTION_DECL_DECL_TYPE instead of
15462 DECL_IS_OPERATOR_NEW_P.
15463 * tree.h (set_function_decl_type): Use FUNCTION_DECL_DECL_TYPE
15465 (DECL_IS_OPERATOR_NEW_P): Likewise.
15466 (DECL_IS_OPERATOR_DELETE_P): Likewise.
15467 (DECL_LAMBDA_FUNCTION_P): Likewise.
15469 2021-05-31 Richard Biener <rguenther@suse.de>
15472 * internal-fn.c (expand_SHUFFLEVECTOR): Define.
15473 * internal-fn.def (SHUFFLEVECTOR): New.
15474 * internal-fn.h (expand_SHUFFLEVECTOR): Declare.
15475 * doc/extend.texi: Document __builtin_shufflevector.
15477 2021-05-31 Peter Bergner <bergner@linux.ibm.com>
15480 * config/rs6000/predicates.md(mma_assemble_input_operand): Allow
15481 indexed form addresses.
15483 2021-05-29 Jeff Law <jlaw@tachyum.com>
15485 * config/h8300/h8300.c (h8300_emit_stack_adjustment): Drop unused
15486 parameter. Call callers fixed.
15488 (output_plussi): Add FALLTHRU markers.
15489 (h8300_shift_needs_scratch_p): Add gcc_unreachable marker.
15491 2021-05-29 Jakub Jelinek <jakub@redhat.com>
15493 PR middle-end/99928
15494 * gimplify.c (gimplify_scan_omp_clauses): For taskloop simd
15495 combined with parallel, make sure to add shared clause to
15496 parallel for explicit linear clause.
15498 2021-05-29 Aldy Hernandez <aldyh@redhat.com>
15500 PR tree-optimization/100787
15501 * gimple-ssa-evrp.c: Disable exporting of global ranges.
15503 2021-05-28 Jason Merrill <jason@redhat.com>
15505 * tree-iterator.h (struct tree_stmt_iterator): Add operator++,
15506 operator--, operator*, operator==, and operator!=.
15507 (class tsi_range): New.
15509 2021-05-28 Richard Biener <rguenther@suse.de>
15511 PR tree-optimization/100778
15512 * tree-vect-slp.c (vect_build_slp_tree_1): Prevent possibly
15513 trapping ops in different BBs.
15515 2021-05-28 Richard Biener <rguenther@suse.de>
15518 * tree-inline.c (copy_bb): When processing __builtin_va_arg_pack
15519 copy fntype from original call.
15521 2021-05-28 Martin Liska <mliska@suse.cz>
15523 PR gcov-profile/100751
15524 * doc/gcov.texi: Revert partially a hunk that was wrong.
15526 2021-05-28 Cooper Qu <cooper.qu@linux.alibaba.com>
15528 * config/csky/csky-linux-elf.h (HAVE_sync_compare_and_swapqi):
15530 (HAVE_sync_compare_and_swaphi): Likewise.
15531 (HAVE_sync_compare_and_swapsi): Likewise.
15533 2021-05-28 Jakub Jelinek <jakub@redhat.com>
15535 PR middle-end/99928
15536 * tree.h (OMP_CLAUSE_MAP_IMPLICIT): Define.
15538 2021-05-28 Tobias Burnus <tobias@codesourcery.com>
15540 * gimplify.c (gimplify_omp_affinity): New.
15541 (gimplify_scan_omp_clauses): Call it; remove affinity clause afterwards.
15542 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_AFFINITY.
15543 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_AFFINITY.
15544 * tree.c (omp_clause_num_ops, omp_clause_code_name): Add clause.
15545 (walk_tree_1): Handle OMP_CLAUSE_AFFINITY.
15547 2021-05-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
15548 Richard Biener <rguenther@suse.de>
15550 * match.pd <popcount & / + pattern matching>:
15551 When generating popcount directly fails, try doing it in two halves.
15553 2021-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
15555 * Makefile.in (generated_files): Add gimple-match.c and
15558 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
15560 * gensupport.c (alter_predicate_for_insn): Handle MATCH_DUP.
15562 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
15564 * gensupport.c (alter_constraints): Add MATCH_SCRATCH case.
15566 2021-05-28 Kewen Lin <linkw@linux.ibm.com>
15568 PR tree-optimization/99398
15569 * tree-ssa-forwprop.c (simplify_permutation): Optimize some cases
15570 where the fed operands are CTOR/CST and propagated through
15571 VIEW_CONVERT_EXPR. Call vec_perm_indices::new_shrunk_vector.
15572 * vec-perm-indices.c (vec_perm_indices::new_shrunk_vector): New
15574 * vec-perm-indices.h (vec_perm_indices::new_shrunk_vector): New
15577 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
15579 * config/i386/mmx.md (addv2sf3): Do not call
15580 ix86_fixup_binary_operands_no_copy.
15583 (<smaxmin:code>v2sf3): Ditto.
15584 (<plusminus:insn><MMXMODEI:mode>3): Ditto.
15585 (<plusminus:insn><VI_32:mode>3): Remove expander.
15586 (<plusminus:insn><VI_32:mode>3): Rename from
15587 "*<plusminus:insn><VI_32:mode>3".
15588 (mulv4hi): Do not call ix86_fixup_binary_operands_no_copy.
15589 (mulv2hi3): Remove expander.
15590 (mulv2hi3): Rename from *mulv2hi3.
15591 (<s>mulv2hi3_highpart): Remove expander.
15592 (<s>mulv2hi3_highpart): Rename from *<s>mulv2hi3_highpart.
15593 (<smaxmin:code><MMXMODE14:mode>3): Rename from
15594 "*mmx_<smaxmin:code><MMXMODE14:mode>3".
15595 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Remove expander.
15596 (SMAXMIN_MMXMODEI): Remove mode iterator.
15597 (<smaxmin:code>v4hi3): New expander.
15598 (<smaxmin:code>v4qi3): Rename from *<smaxmin:code>v4qi3.
15599 (<smaxmin:code>v2hi3): Rename from *<smaxmin:code>v2hi3.
15600 (<smaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
15601 (SMAXMIN_VI_32): Remove mode iterator.
15602 (<umaxmin:code><MMXMODE24:mode>3): Rename from
15603 "*mmx_<umaxmin:code><MMXMODE24:mode>3".
15604 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Remove expander.
15605 (UMAXMIN_MMXMODEI): Remove mode iterator.
15606 (<umaxmin:code>v8qi3): New expander.
15607 (<umaxmin:code>v4qi3): Rename from *<umaxmin:code>v4qi3.
15608 (<umaxmin:code>v2hi3): Rename from *<umaxmin:code>v2hi3.
15609 (<umaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
15610 (UMAXMIN_VI_32): Remove mode iterator.
15611 (<any_shift:insn>v2hi3): Remove expander.
15612 (<any_shift:insn>v2hi3): Rename from *<any_shift:insn>v2hi3.
15613 (<any_logic:code><MMXMODEI:mode>3): Do not call
15614 ix86_fixup_binary_operands_no_copy.
15615 (<any_logic:code><VI_32:mode>3): Remove expander.
15616 (<any_logic:code><VI_32:mode>3): Rename from
15617 "*<any_logic:code><VI_32:mode>3".
15618 (uavg<mode>3_ceil): Do not call ix86_fixup_binary_operands_no_copy.
15619 * config/i386/sse.md (div<VF2:mode>3): Do not call
15620 ix86_fixup_binary_operands_no_copy.
15621 (div<VF1:mode>3): Ditto.
15622 (<maxmin:code><VI8_AVX2_AVX512F:mode>3): Ditto.
15623 (smulhrsv4hi3): Ditto.
15624 (smulhrsv2hi3): Ditto.
15626 2021-05-27 Martin Sebor <msebor@redhat.com>
15628 * ggc.h (gt_ggc_mx): Add overloads for all integers.
15630 * hash-map.h (class hash_map): Add pch_nx_helper overloads for all
15632 (hash_map::operator==): New function.
15634 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
15637 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
15638 For TARGET_XOP bypass SSE comparisons for all supported vector modes.
15639 * config/i386/mmx.md (*xop_maskcmp<MMXMODEI:mode>3): New insn pattern.
15640 (*xop_maskcmp<VI_32:mode>3): Ditto.
15641 (*xop_maskcmp_uns<MMXMODEI:mode>3): Ditto.
15642 (*xop_maskcmp_uns<VI_32:mode>3): Ditto.
15644 2021-05-27 Richard Earnshaw <rearnsha@arm.com>
15647 * config/arm/arm.c (arm_configure_build_target): Remove parameter
15648 opts_set, directly check opts parameters for being non-null.
15649 (arm_option_restore): Update call to arm_configure_build_target.
15650 (arm_option_override): Likewise.
15651 (arm_can_inline_p): Likewise.
15652 (arm_valid_target_attribute_tree): Likewise.
15653 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
15654 * config/arm/arm-protos.h (arm_configure_build_target): Adjust
15657 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
15659 * vr-values.c (simplify_conversion_using_ranges): Use
15660 get_range_query instead of get_global_range_query.
15662 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
15664 * gimple-range.cc (get_range_global): Move to value-query.cc.
15665 (gimple_range_global): Same.
15666 (get_global_range_query): Same.
15667 (global_range_query::range_of_expr): Same.
15668 * gimple-range.h (class global_range_query): Move to
15670 (gimple_range_global): Same.
15671 * tree-ssanames.c (get_range_info): Move to value-query.cc.
15672 (get_ptr_nonnull): Same.
15673 * tree-ssanames.h (get_range_info): Remove.
15674 (get_ptr_nonnull): Remove.
15675 * value-query.cc (get_ssa_name_range_info): Move from
15677 (get_ssa_name_ptr_info_nonnull): Same.
15678 (get_range_global): Move from gimple-range.cc.
15679 (gimple_range_global): Same.
15680 (get_global_range_query): Same.
15681 (global_range_query::range_of_expr): Same.
15682 * value-query.h (class global_range_query): Move from
15684 (gimple_range_global): Same.
15686 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
15689 * config/i386/mmx.md (uavgv4qi3_ceil): New insn pattern.
15690 (uavgv2hi3_ceil): Ditto.
15692 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
15695 * doc/extend.texi (scalar_storage_order): Rephrase slightly.
15697 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
15699 * tree-ssanames.c (get_range_info): Merge both copies of
15700 get_range_info into one that works with irange.
15701 * tree-ssanames.h (get_range_info): Remove version that works on
15704 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
15706 * builtins.c (check_nul_terminated_array): Convert to get_range_query.
15707 (expand_builtin_strnlen): Same.
15708 (determine_block_size): Same.
15709 * fold-const.c (expr_not_equal_to): Same.
15710 * gimple-fold.c (size_must_be_zero_p): Same.
15711 * gimple-match-head.c: Include gimple-range.h.
15712 * gimple-pretty-print.c (dump_ssaname_info): Convert to get_range_query.
15713 * gimple-ssa-warn-restrict.c
15714 (builtin_memref::extend_offset_range): Same.
15715 * graphite-sese-to-poly.c (add_param_constraints): Same.
15716 * internal-fn.c (get_min_precision): Same.
15717 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): Same.
15718 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Same.
15720 * tree-data-ref.c (split_constant_offset): Same.
15721 (dr_step_indicator): Same.
15722 * tree-dfa.c (get_ref_base_and_extent): Same.
15723 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
15724 * tree-ssa-loop-niter.c (refine_value_range_using_guard): Same.
15725 (determine_value_range): Same.
15726 (record_nonwrapping_iv): Same.
15727 (infer_loop_bounds_from_signedness): Same.
15728 (scev_var_range_cant_overflow): Same.
15729 * tree-ssa-phiopt.c (two_value_replacement): Same.
15730 * tree-ssa-pre.c (insert_into_preds_of_block): Same.
15731 * tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Same.
15732 * tree-ssa-strlen.c (handle_builtin_stxncpy_strncat): Same.
15734 (dump_strlen_info): Same.
15735 (set_strlen_range): Same.
15736 (maybe_diag_stxncpy_trunc): Same.
15737 (get_len_or_size): Same.
15738 (handle_integral_assign): Same.
15739 * tree-ssa-structalias.c (find_what_p_points_to): Same.
15740 * tree-ssa-uninit.c (find_var_cmp_const): Same.
15741 * tree-switch-conversion.c (bit_test_cluster::emit): Same.
15742 * tree-vect-patterns.c (vect_get_range_info): Same.
15743 (vect_recog_divmod_pattern): Same.
15744 * tree-vrp.c (intersect_range_with_nonzero_bits): Same.
15745 (register_edge_assert_for_2): Same.
15746 (determine_value_range_1): Same.
15747 * tree.c (get_range_pos_neg): Same.
15748 * vr-values.c (vr_values::get_lattice_entry): Same.
15749 (vr_values::update_value_range): Same.
15750 (simplify_conversion_using_ranges): Same.
15752 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
15754 * gimple-ssa-warn-alloca.c (alloca_call_type): Use
15755 get_range_query instead of query argument.
15756 (pass_walloca::execute): Enable and disable global ranger.
15758 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
15760 * gimple-ssa-evrp.c (rvrp_folder::rvrp_folder): Call
15762 (rvrp_folder::~rvrp_folder): Call disable_ranger.
15763 (hybrid_folder::hybrid_folder): Call enable_ranger.
15764 (hybrid_folder::~hybrid_folder): Call disable_ranger.
15766 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
15768 * function.c (allocate_struct_function): Set cfun->x_range_query.
15769 * function.h (struct function): Declare x_range_query.
15770 (get_range_query): New.
15771 (get_global_range_query): New.
15772 * gimple-range-cache.cc (ssa_global_cache::ssa_global_cache):
15773 Remove call to safe_grow_cleared.
15774 * gimple-range.cc (get_range_global): New.
15775 (gimple_range_global): Move from gimple-range.h.
15776 (get_global_range_query): New.
15777 (global_range_query::range_of_expr): New.
15778 (enable_ranger): New.
15779 (disable_ranger): New.
15780 * gimple-range.h (gimple_range_global): Move to gimple-range.cc.
15781 (class global_range_query): New.
15782 (enable_ranger): New.
15783 (disable_ranger): New.
15784 * gimple-ssa-evrp.c (evrp_folder::~evrp_folder): Rename
15785 dump_all_value_ranges to dump.
15786 * tree-vrp.c (vrp_prop::finalize): Same.
15787 * value-query.cc (range_query::dump): New.
15788 * value-query.h (range_query::dump): New.
15789 * vr-values.c (vr_values::dump_all_value_ranges): Rename to...
15790 (vr_values::dump): ...this.
15791 * vr-values.h (class vr_values): Rename dump_all_value_ranges to
15792 dump and make virtual.
15794 2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
15796 * config/i386/i386.c (ix86_autovectorize_vector_modes):
15797 Add V4QImode and V16QImode for TARGET_SSE2.
15798 * doc/sourcebuild.texi (Vector-specific attributes):
15799 Add vect64 and vect32 description.
15801 2021-05-26 Bernd Edlinger <bernd.edlinger@hotmail.de>
15803 * gimple-range-gori.cc (range_def_chain::register_dependency):
15804 Resize m_def_chain when needed.
15806 2021-05-26 Christophe Lyon <christophe.lyon@linaro.org>
15808 * config/arm/mve.md (mve_vaddvq_<supf><mode>): Prefix with '@'.
15809 * config/arm/neon.md (reduc_plus_scal_<mode>): Move to ..
15810 * config/arm/vec-common.md: .. here. Add support for MVE.
15812 2021-05-26 Jakub Jelinek <jakub@redhat.com>
15814 * config/epiphany/epiphany.c (epiphany_print_operand_address): Remove
15816 * config/microblaze/microblaze.c (microblaze_legitimize_address,
15818 microblaze_option_override, print_operand): Likewise.
15819 * config/microblaze/microblaze.md (call_internal_plt,
15820 call_value_intern_plt, call_value_intern): Likewise.
15821 * config/arm/aout.h (ASM_OUTPUT_ALIGN): Likewise.
15822 * config/iq2000/iq2000.md (call_internal1, call_value_internal1,
15823 call_value_multiple_internal1): Likewise.
15824 * config/bfin/bfin.c (symbolic_reference_mentioned_p): Likewise.
15826 2021-05-26 Jan-Benedict Glaw <jbglaw@lug-owl.de>
15828 * config/arc/arc.c (arc_address_cost, arc_print_operand_address,
15829 arc_ccfsm_advance, symbolic_reference_mentioned_p,
15830 arc_raw_symbolic_reference_mentioned_p): Remove register
15833 2021-05-26 Jakub Jelinek <jakub@redhat.com>
15836 * omp-low.c: Include omp-offload.h.
15837 (create_omp_child_function): If current_function_decl has
15838 "omp declare target" attribute and is_gimple_omp_offloaded,
15839 remove that attribute from the copy of attribute list and
15840 add "omp target entrypoint" attribute instead.
15841 (lower_omp_target): Mark .omp_data_sizes.* and .omp_data_kinds.*
15842 variables for offloading if in omp_maybe_offloaded_ctx.
15843 * omp-offload.c (pass_omp_target_link::execute): Nullify second
15844 argument to GOMP_target_data_ext in offloaded code.
15846 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
15848 * config/csky/csky.c (csky_can_change_mode_class): Delete.
15849 For csky, HF/SF mode use the low bits of VREGS.
15851 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
15853 * gimplify.c (gimplify_decl_expr): Do not clear TREE_READONLY on a
15854 DECL which is a reference for OMP.
15856 2021-05-26 Martin Liska <mliska@suse.cz>
15858 PR gcov-profile/100751
15859 * doc/gcov.texi: Document that __gcov_dump can be called just
15860 once and that __gcov_reset resets run-time counters.
15862 2021-05-26 Martin Liska <mliska@suse.cz>
15864 * doc/install.texi: Port relevant part from install-old.texi
15865 and re-generate list of CPUs and systems.
15867 2021-05-26 Martin Liska <mliska@suse.cz>
15869 * Makefile.in: Remove it.
15870 * doc/include/fdl.texi: Update next/previous chapters.
15871 * doc/install.texi: Likewise.
15872 * doc/install-old.texi: Removed.
15874 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
15876 * config/csky/csky.c (ck810_legitimate_index_p): Support
15877 "base + index" with DF mode.
15878 * config/csky/constraints.md ("Y"): New constraint for memory operands
15879 without index register.
15880 * config/csky/csky_insn_fpuv2.md (fpuv3_movdf): Use "Y" instead of "m"
15881 when mov between memory and general registers, and lower their priority.
15882 * config/csky/csky_insn_fpuv3.md (fpuv2_movdf): Likewise.
15884 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
15886 * config/csky/csky.c (TARGET_PROMOTE_PROTOTYPES): Delete.
15888 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
15890 * config/csky/csky.md (untyped_call): Emit clobber for return
15891 registers to mark them used.
15893 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
15895 * config/csky/csky.md (cskyv2_sextend_ldbs): New.
15897 2021-05-26 Andrew Pinski <apinski@marvell.com>
15899 * match.pd (x < 0 ? ~y : y): New patterns.
15901 2021-05-26 Andrew Pinski <apinski@marvell.com>
15903 * match.pd (A?CST1:CST2): Add simplifcations for A?0:+-1, A?+-1:0,
15904 A?POW2:0 and A?0:POW2.
15906 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15908 * gimple-range-gori.cc (class logical_stmt_cache): Delete
15909 (logical_stmt_cache::logical_stmt_cache ): Delete.
15910 (logical_stmt_cache::~logical_stmt_cache): Delete.
15911 (logical_stmt_cache::cache_entry::dump): Delete.
15912 (logical_stmt_cache::get_range): Delete.
15913 (logical_stmt_cache::cached_name ): Delete.
15914 (logical_stmt_cache::same_cached_name): Delete.
15915 (logical_stmt_cache::cacheable_p): Delete.
15916 (logical_stmt_cache::slot_diagnostics ): Delete.
15917 (logical_stmt_cache::dump): Delete.
15918 (gori_compute_cache::gori_compute_cache): Delete.
15919 (gori_compute_cache::~gori_compute_cache): Delete.
15920 (gori_compute_cache::compute_operand_range): Delete.
15921 (gori_compute_cache::cache_stmt): Delete.
15922 * gimple-range-gori.h (gori_compute::compute_operand_range): Remove
15924 (class gori_compute_cache): Delete.
15926 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15928 * gimple-range.cc (fold_using_range::range_of_range_op): Use m_gori
15930 (fold_using_range::range_of_address): Adjust.
15931 (fold_using_range::range_of_phi): Adjust.
15932 * gimple-range.h (class fur_source): Adjust.
15933 (fur_source::fur_source): Adjust.
15935 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15937 * gimple-range-gori.cc (gori_compute::expr_range_at_stmt): Rename
15938 from expr_range_in_bb and adjust.
15939 (gori_compute::compute_name_range_op): Adjust.
15940 (gori_compute::optimize_logical_operands): Adjust.
15941 (gori_compute::compute_logical_operands_in_chain): Adjust.
15942 (gori_compute::compute_operand1_range): Adjust.
15943 (gori_compute::compute_operand2_range): Adjust.
15944 (ori_compute_cache::cache_stmt): Adjust.
15945 * gimple-range-gori.h (gori_compute): Rename prototype.
15947 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15949 * gimple-range.cc (gimple_ranger::range_of_expr): Non-null should be
15950 checked only after range_of_stmt, not range_on_entry.
15951 (gimple_ranger::range_on_entry): Check for non-null in any
15952 predecessor block, if it is not already non-null.
15953 (gimple_ranger::range_on_exit): DOnt check for non-null after
15954 range on entry call.
15955 (gimple_ranger::dump_bb): New. Split from dump.
15956 (gimple_ranger::dump): Adjust.
15957 * gimple-range.h (class gimple_ranger): Adjust.
15959 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15961 * gimple-range-cache.cc (struct range_timestamp): Delete.
15962 (class temporal_cache): Adjust.
15963 (temporal_cache::get_timestamp): Delete.
15964 (temporal_cache::set_dependency): Delete.
15965 (temporal_cache::temporal_value): Adjust.
15966 (temporal_cache::current_p): Take dependencies as params.
15967 (temporal_cache::set_timestamp): Adjust.
15968 (temporal_cache::set_always_current): Adjust.
15969 (ranger_cache::get_non_stale_global_range): Adjust.
15970 (ranger_cache::register_dependency): Delete.
15971 * gimple-range-cache.h (class range_cache): Adjust.
15973 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15975 * gimple-range-gori.cc (range_def_chain::range_def_chain): init
15977 (range_def_chain::~range_def_chain): Dispose of obstack rather than
15978 each individual bitmap.
15979 (range_def_chain::set_import): New.
15980 (range_def_chain::get_imports): New.
15981 (range_def_chain::chain_import_p): New.
15982 (range_def_chain::register_dependency): Rename from build_def_chain
15984 (range_def_chain::def_chain_in_bitmap_p): New.
15985 (range_def_chain::add_def_chain_to_bitmap): New.
15986 (range_def_chain::has_def_chain): Just check first depenedence.
15987 (range_def_chain::get_def_chain): Process imports, use generic
15988 register_dependency routine.
15989 (range_def_chain::dump): New.
15990 (gori_map::gori_map): Allocate import list.
15991 (gori_map::~gori_map): Release imports.
15992 (gori_map::exports): Check for past allocated block size.
15993 (gori_map::imports): New.
15994 (gori_map::def_chain_in_export_p): Delete.
15995 (gori_map::is_import_p): New.
15996 (gori_map::maybe_add_gori): Handle imports.
15997 (gori_map::dump): Adjust output, add imports.
15998 (gori_compute::has_edge_range_p): Remove def_chain_in_export call.
15999 (gori_export_iterator::gori_export_iterator): New.
16000 (gori_export_iterator::next): New.
16001 (gori_export_iterator::get_name): New.
16002 * gimple-range-gori.h (range_def_chain): Add imports and direct
16003 dependecies via struct rdc.
16004 (range_def_chain::depend1): New.
16005 (range_def_chain::depend2): New.
16006 (class gori_map): Adjust.
16007 (FOR_EACH_GORI_IMPORT_NAME): New.
16008 (FOR_EACH_GORI_EXPORT_NAME): New.
16009 (class gori_export_iterator): New.
16011 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
16013 * gimple-range-cache.cc (ranger_cache::ranger_cache): Move initial
16014 export cache filling to here.
16015 * gimple-range-gori.cc (gori_compute::gori_compute) : From Here.
16017 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
16019 * gimple-range-gori.cc (range_def_chain): Move to gimple-range-gori.h.
16020 (gori_map): Move to gimple-range-gori.h.
16021 (gori_compute::gori_compute): Adjust.
16022 (gori_compute::~gori_compute): Delete.
16023 (gori_compute::compute_operand_range_switch): Adjust.
16024 (gori_compute::compute_operand_range): Adjust.
16025 (gori_compute::compute_logical_operands): Adjust.
16026 (gori_compute::has_edge_range_p ): Adjust.
16027 (gori_compute::set_range_invariant): Delete.
16028 (gori_compute::dump): Adjust.
16029 (gori_compute::outgoing_edge_range_p): Adjust.
16030 * gimple-range-gori.h (class range_def_chain): Relocate here.
16031 (class gori_map): Relocate here.
16032 (class gori_compute): Inherit from gori_map, and adjust.
16034 2021-05-25 Aldy Hernandez <aldyh@redhat.com>
16036 * value-range.cc (range_tests_legacy): Use
16037 build_nonstandard_integer_type instead of int and short.
16039 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
16041 * gimplify.c (gimplify_decl_expr): Clear TREE_READONLY on the DECL
16042 when really creating an initialization statement for it.
16044 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
16046 * tree-inline.c (setup_one_parameter): Fix thinko in new condition.
16048 2021-05-25 Kito Cheng <kito.cheng@sifive.com>
16050 * config/riscv/riscv.h (ASM_SPEC): Pass -mno-relax.
16052 2021-05-25 Martin Liska <mliska@suse.cz>
16054 PR tree-optimization/92860
16056 * optc-save-gen.awk: Remove exceptions.
16058 2021-05-25 Martin Liska <mliska@suse.cz>
16060 * asan.h (sanitize_coverage_p): New function.
16061 * doc/extend.texi: Document it.
16062 * fold-const.c (fold_range_test): Use sanitize_flags_p
16063 instead of flag_sanitize_coverage.
16064 (fold_truth_andor): Likewise.
16065 * sancov.c: Likewise.
16066 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
16067 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Handle
16068 -fsanitize-coverage when inlining.
16070 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
16072 * config/csky/csky-modes.def : Fix copyright.
16074 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
16076 * config/csky/csky-modes.def : Amend copyright.
16077 * config/csky/csky_insn_fpuv2.md : Likewise.
16078 * config/csky/csky_insn_fpuv3.md : Likewise.
16080 2021-05-25 Richard Biener <rguenther@suse.de>
16082 PR middle-end/100727
16083 * calls.c (initialize_argument_information): Explicitely test
16084 for WITH_SIZE_EXPR.
16085 * gimple-expr.c (mark_addressable): Skip outer WITH_SIZE_EXPR.
16087 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
16089 * config/csky/csky.h (FRAME_POINTER_REGNUM): Use
16090 HARD_FRAME_POINTER_REGNUM and FRAME_POINTER_REGNUM instead of
16091 the signle definition. The signle definition may not work well
16092 at simplify_subreg_regno().
16093 (HARD_FRAME_POINTER_REGNUM): New.
16094 (ELIMINABLE_REGS): Add for HARD_FRAME_POINTER_REGNUM.
16095 * config/csky/csky.c (get_csky_live_regs, csky_can_eliminate,
16096 csky_initial_elimination_offset, csky_expand_prologue,
16097 csky_expand_epilogue): Add for HARD_FRAME_POINTER_REGNUM.
16099 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
16101 * config/csky/csky.c (csky_option_override):
16102 Init csky_arch_isa_features[] in advance, so TARGET_DSP
16103 and TARGET_DIV can be set well.
16105 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
16107 * config/csky/constraints.md ("l", "h"): Delete.
16108 * config/csky/csky.h (reg_class, REG_CLASS_NAMES,
16109 REG_CLASS_CONTENTS): Delete LO_REGS and HI_REGS.
16110 * config/csky/csky.c (regno_reg_classm,
16111 csky_secondary_reload, csky_register_move_cost):
16112 Use HILO_REGS instead of LO_REGS and HI_REGS.
16114 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
16116 * config/csky/constraints.md ("W"): New constriant for mem operand
16117 with base reg, index register.
16118 ("Q"): Renamed and modified "csky_valid_fpuv2_mem_operand" to
16119 "csky_valid_mem_constraint_operand" to deal with both "Q" and "W"
16121 ("Dv"): New constraint for const double value that can be used at
16123 * config/csky/csky-modes.def (HFmode): New mode.
16124 * config/csky/csky-protos.h (csky_valid_fpuv2_mem_operand): Rename
16125 to "csky_valid_mem_constraint_operand" and support new constraint
16127 (csky_get_movedouble_length): New.
16128 (fpuv3_output_move): New.
16129 (fpuv3_const_double): New.
16130 * config/csky/csky.c (csky_option_override): New arch CK860 with fpv3.
16131 (decompose_csky_address): Refine.
16132 (csky_print_operand): New "CONST_DOUBLE" operand.
16133 (csky_output_move): Support fpv3 instructions.
16134 (csky_get_movedouble_length): New.
16135 (fpuv3_output_move): New.
16136 (fpuv3_const_double): New.
16137 (csky_emit_compare): Cover float comparsion.
16138 (csky_emit_compare_float): Refine.
16139 (csky_vaild_fpuv2_mem_operand): Rename to
16140 "csky_valid_mem_constraint_operand" and support new constraint "W".
16141 (ck860_rtx_costs): New.
16142 (csky_rtx_costs): Add the cost calculation of CK860.
16143 (regno_reg_class): New vregs for fpuv3.
16144 (csky_dbx_regno): Likewise.
16145 (csky_cpu_cpp_builtins): New builtin macro for fpuv3.
16146 (csky_conditional_register_usage): Suporrot fpuv3.
16147 (csky_dwarf_register_span): Suporrot fpuv3.
16148 (csky_init_builtins, csky_mangle_type): Support "__fp16" type.
16149 (ck810_legitimate_index_p): Support fp16.
16150 * config/csky/csky.h (TARGET_TLS): ADD CK860.
16151 (CSKY_VREG_P, CSKY_VREG_LO_P, CSKY_VREG_HI_P): Support fpuv3.
16152 (TARGET_SINGLE_FPU): Support fpuv3.
16153 (TARGET_SUPPORT_FPV3): New.
16154 (FIRST_PSEUDO_REGISTER): Change to 202 to hold the new fpuv3 registers.
16155 (FIXED_REGISTERS, CALL_REALLY_USED_REGISTERS, REGISTER_NAMES,
16156 REG_CLASS_CONTENTS): Support fpuv3.
16157 * config/csky/csky.md (movsf): Move to cksy_insn_fpu.md and refine.
16158 (csky_movsf_fpv2): Likewise.
16159 (ck801_movsf): Likewise.
16160 (csky_movsf): Likewise.
16162 (csky_movdf_fpv2): Likewise.
16163 (ck801_movdf): Likewise.
16164 (csky_movdf): Likewise.
16165 (movsicc): Refine. Use "comparison_operatior" instead of
16166 "ordered_comparison_operatior".
16167 (addsicc): Likewise.
16168 (CSKY_FIRST_VFP3_REGNUM, CSKY_LAST_VFP3_REGNUM): New constant.
16169 (call_value_internal_vh): New.
16170 * config/csky/csky_cores.def (CK860): New arch and cpu.
16175 * config/csky/csky_insn_fpu.md: Refactor. Separate all float patterns
16176 into emit-patterns and match-patterns, remain the emit-patterns here,
16177 and move the match-patterns to csky_insn_fpuv2.md or
16178 csky_insn_fpuv3.md.
16179 * config/csky/csky_insn_fpuv2.md: New file for fpuv2 instructions.
16180 * config/csky/csky_insn_fpuv3.md: New file and new patterns for fpuv3
16182 * config/csky/csky_isa.def (fcr): New.
16187 (CK860): New definition for ck860.
16188 * config/csky/csky_tables.opt (ck860): New processors ck860,
16189 ck860f. And new arch ck860.
16194 * config/csky/predicates.md (csky_float_comparsion_operator): Delete
16195 "geu", "gtu", "leu", "ltu", which will never appear at float comparison.
16196 * config/csky/t-csky-elf: Support 860.
16197 * config/csky/t-csky-linux: Likewise.
16198 * doc/md.texi: Add "Q" and "W" constraints for C-SKY.
16200 2021-05-24 Aaron Sawdey <acsawdey@linux.ibm.com>
16202 * config/rs6000/genfusion.pl (gen_logical_addsubf): Refactor to
16203 add generation of logical-add and add-logical fusion pairs.
16204 * config/rs6000/rs6000-cpus.def: Add new fusion to ISA 3.1 mask
16206 * config/rs6000/rs6000.c (rs6000_option_override_internal): Turn on
16207 logical-add and add-logical fusion by default.
16208 * config/rs6000/rs6000.opt: Add -mpower10-fusion-logical-add and
16209 -mpower10-fusion-add-logical options.
16210 * config/rs6000/fusion.md: Regenerate file.
16212 2021-05-24 Aldy Hernandez <aldyh@redhat.com>
16214 * value-range.cc (irange::legacy_equal_p): Check type when
16215 comparing VR_VARYING types.
16216 (range_tests_legacy): Test comparing VARYING ranges of different
16219 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
16221 * config/aarch64/aarch64.c (neoversen1_tunings):
16222 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
16224 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
16226 * config/aarch64/aarch64.c (aarch64_classify_symbol): Use GOT for
16227 extern weak symbols. Limit symbol offsets for non-GOT symbols with
16230 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
16232 * config/arm/neon.md (vec_load_lanesxi<mode>)
16233 (vec_store_lanexoi<mode>): Move ...
16234 * config/arm/vec-common.md: here.
16236 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
16238 * config/arm/neon.md (vec_load_lanesoi<mode>)
16239 (vec_store_lanesoi<mode>): Move ...
16240 * config/arm/vec-common.md: here.
16242 2021-05-24 liuhongt <hongtao.liu@intel.com>
16245 * config/i386/i386.c (ix86_gimple_fold_builtin): Replacing
16246 stmt with GIMPLE_NOP when lhs doesn't exist.
16248 2021-05-23 Uroš Bizjak <ubizjak@gmail.com>
16251 * config/i386/mmx.md (*push<VI_32:mode>2_rex64):
16252 New instruction pattern.
16253 (*push<VI_32:mode>2): Ditto.
16254 (push splitter for SSE registers): New splitter.
16256 2021-05-23 Andrew Pinski <apinski@marvell.com>
16258 * match.pd ((A & C) != 0 ? D : 0): Limit to non pointer types.
16260 2021-05-22 Aaron Sawdey <acsawdey@linux.ibm.com>
16262 * config/rs6000/genfusion.pl (gen_addadd): Fix incorrect attr types.
16263 * config/rs6000/fusion.md: Regenerate file.
16265 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
16267 * config/rs6000/genfusion.pl (gen_addadd): New function.
16268 * config/rs6000/fusion.md: Regenerate file.
16269 * config/rs6000/rs6000-cpus.def: Add
16270 OPTION_MASK_P10_FUSION_2ADD to masks.
16271 * config/rs6000/rs6000.c (rs6000_option_override_internal):
16272 Handle default value of OPTION_MASK_P10_FUSION_2ADD.
16273 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2add.
16275 2021-05-21 Jakub Jelinek <jakub@redhat.com>
16277 PR middle-end/99928
16278 * tree.h (OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET): Define.
16279 * gimplify.c (enum gimplify_omp_var_data): Fix up
16280 GOVD_MAP_HAS_ATTACHMENTS value, add GOVD_FIRSTPRIVATE_IMPLICIT.
16281 (omp_lastprivate_for_combined_outer_constructs): If combined target
16282 has GOVD_FIRSTPRIVATE_IMPLICIT set for the decl, change it to
16283 GOVD_MAP | GOVD_SEEN.
16284 (gimplify_scan_omp_clauses): Set GOVD_FIRSTPRIVATE_IMPLICIT for
16285 firstprivate clauses with OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT.
16286 (gimplify_adjust_omp_clauses): For firstprivate clauses with
16287 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT either clear that bit and
16288 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET too, or remove it and
16289 let it be replaced by implicit map clause.
16291 2021-05-21 Jakub Jelinek <jakub@redhat.com>
16293 PR middle-end/99928
16294 * gimplify.c (omp_lastprivate_for_combined_outer_constructs): New
16296 (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LASTPRIVATE>: Use it.
16297 (gimplify_omp_for): Likewise.
16299 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16301 PR middle-end/90115
16302 * omp-low.c (oacc_privatization_candidate_p): Reject 'static',
16303 'external' in blocks.
16305 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16307 PR middle-end/90115
16308 * flag-types.h (enum openacc_privatization): New.
16309 * params.opt (-param=openacc-privatization): New.
16310 * doc/invoke.texi (openacc-privatization): Document it.
16311 * omp-general.h (get_openacc_privatization_dump_flags): New
16313 * omp-low.c (oacc_privatization_candidate_p): Add diagnostics.
16314 * omp-offload.c (execute_oacc_device_lower)
16315 <IFN_UNIQUE_OACC_PRIVATE>: Re-work diagnostics.
16316 * target.def (goacc.adjust_private_decl): Add 'location_t'
16318 * doc/tm.texi: Regenerate.
16319 * config/gcn/gcn-protos.h (gcn_goacc_adjust_private_decl): Adjust.
16320 * config/gcn/gcn-tree.c (gcn_goacc_adjust_private_decl): Likewise.
16321 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl):
16322 Likewise. Preserve it for...
16323 (nvptx_goacc_expand_var_decl): ... use here.
16325 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16327 * doc/sourcebuild.texi (Other attributes): Document '__OPTIMIZE__'
16330 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16332 PR middle-end/90115
16333 * omp-low.c (oacc_privatization_candidate_p): New function.
16334 (oacc_privatization_scan_clause_chain)
16335 (oacc_privatization_scan_decl_chain): Use it. Also
16336 'gcc_checking_assert' that we're not seeing duplicates.
16338 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16340 PR middle-end/90115
16341 * omp-offload.c (execute_oacc_device_lower): Skip processing if no
16344 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16346 PR middle-end/90115
16347 * omp-offload.c (execute_oacc_device_lower): Explain.
16349 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16351 PR middle-end/90115
16352 * omp-offload.c (execute_oacc_device_lower)
16353 <IFN_UNIQUE_OACC_PRIVATE>: Diagnose and handle for 'level == -1'
16355 * internal-fn.c (expand_UNIQUE): Don't expect
16356 'IFN_UNIQUE_OACC_PRIVATE'.
16358 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16360 PR middle-end/90115
16361 * omp-low.c (lower_omp_for): Don't evaluate OpenMP 'for' clauses.
16363 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16365 PR middle-end/90115
16366 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl)
16367 (nvptx_goacc_expand_var_decl): Tighten.
16369 2021-05-21 Julian Brown <julian@codesourcery.com>
16370 Chung-Lin Tang <cltang@codesourcery.com>
16371 Thomas Schwinge <thomas@codesourcery.com>
16373 PR middle-end/90115
16374 * doc/tm.texi.in (TARGET_GOACC_EXPAND_VAR_DECL)
16375 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Add documentation hooks.
16376 * doc/tm.texi: Regenerate.
16377 * expr.c (expand_expr_real_1): Expand decls using the
16378 expand_var_decl OpenACC hook if defined.
16379 * internal-fn.c (expand_UNIQUE): Handle IFN_UNIQUE_OACC_PRIVATE.
16380 * internal-fn.h (IFN_UNIQUE_CODES): Add OACC_PRIVATE.
16381 * omp-low.c (omp_context): Add oacc_privatization_candidates
16383 (lower_oacc_reductions): Add PRIVATE_MARKER parameter. Insert
16385 (lower_oacc_head_tail): Add PRIVATE_MARKER parameter. Modify
16386 private marker's gimple call arguments, and pass it to
16387 lower_oacc_reductions.
16388 (oacc_privatization_scan_clause_chain)
16389 (oacc_privatization_scan_decl_chain, lower_oacc_private_marker):
16391 (lower_omp_for, lower_omp_target, lower_omp_1): Use these.
16392 * omp-offload.c (convert.h): Include.
16393 (oacc_loop_xform_head_tail): Treat private-variable markers like
16394 fork/join when transforming head/tail sequences.
16395 (struct var_decl_rewrite_info): Add struct.
16396 (oacc_rewrite_var_decl, is_sync_builtin_call): New functions.
16397 (execute_oacc_device_lower): Support rewriting gang-private
16398 variables using target hook, and fix up addr_expr and var_decl
16400 * target.def (adjust_private_decl, expand_var_decl): New hooks.
16401 * config/gcn/gcn-protos.h (gcn_goacc_adjust_gangprivate_decl):
16403 (gcn_goacc_adjust_private_decl): ...this.
16404 * config/gcn/gcn-tree.c (gcn_goacc_adjust_gangprivate_decl):
16406 (gcn_goacc_adjust_private_decl): ...this. Add LEVEL parameter.
16407 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_GANGPRIVATE_DECL): Rename
16408 definition using gcn_goacc_adjust_gangprivate_decl...
16409 (TARGET_GOACC_ADJUST_PRIVATE_DECL): ...to this, using
16410 gcn_goacc_adjust_private_decl.
16411 * config/nvptx/nvptx.c (tree-pretty-print.h): Include.
16412 (gang_private_shared_size): New global variable.
16413 (gang_private_shared_align): Likewise.
16414 (gang_private_shared_sym): Likewise.
16415 (gang_private_shared_hmap): Likewise.
16416 (nvptx_option_override): Initialize these.
16417 (nvptx_file_end): Output gang_private_shared_sym.
16418 (nvptx_goacc_adjust_private_decl, nvptx_goacc_expand_var_decl):
16420 (nvptx_set_current_function): Clear gang_private_shared_hmap.
16421 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Define hook.
16422 (TARGET_GOACC_EXPAND_VAR_DECL): Likewise.
16424 2021-05-21 H.J. Lu <hjl.tools@gmail.com>
16426 * config/i386/i386-modes.def (MAX_BITSIZE_MODE_ANY_INT): Removed.
16428 2021-05-21 Richard Biener <rguenther@suse.de>
16429 H.J. Lu <hjl.tools@gmail.com>
16431 PR middle-end/90773
16432 * expr.c (expand_constructor): Elide expand_constructor if
16433 move by pieces is preferred.
16435 2021-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16437 * config/aarch64/aarch64-builtins.c (aarch64_call_properties):
16438 Take a flag and mode value as arguments.
16439 (aarch64_modifies_global_state_p): Likewise.
16440 (aarch64_reads_global_state_p): Likewise.
16441 (aarch64_could_trap_p): Likewise.
16442 (aarch64_get_attributes): Likewise.
16443 (aarch64_init_simd_builtins): Adjust callsite of above.
16444 (aarch64_init_fcmla_laneq_builtins): Use aarch64_get_attributes to get
16445 function attributes to apply to builtins.
16446 (aarch64_init_crc32_builtins): Likewise.
16447 (aarch64_init_builtin_rsqrt): Likewise.
16449 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
16451 * config/rs6000/rs6000.md (define_attr "type"): Add types for fusion.
16452 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10): Use new fusion types.
16453 (gen_2logical): Use new fusion types.
16454 * config/rs6000/fusion.md: Regenerate.
16456 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
16459 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
16460 Handle V4QI and V2HI modes.
16461 (ix86_expand_sse_movcc): Ditto.
16462 * config/i386/mmx.md (*<sat_plusminus:insn><VI_32:mode>3):
16463 New instruction pattern.
16464 (*eq<VI_32:mode>3): Ditto.
16465 (*gt<VI_32:mode>3): Ditto.
16466 (*xop_pcmov_<VI_32:mode>): Ditto.
16467 (mmx_pblendvb32): Ditto.
16468 (mmx_pblendvb64): Rename from mmx_pblendvb.
16469 (vec_cmp<VI_32:mode><VI_32:mode>): New expander.
16470 (vec_cmpu<VI_32:mode><VI_32:mode>): Ditto.
16471 (vcond<VI_32:mode><VI_32:mode>): Ditto.
16472 (vcondu<VI_32:mode><VI_32:mode>): Ditto.
16473 (vcond_mask_<VI_32:mode><VI_32:mode>): Ditto.
16475 2021-05-21 Jakub Jelinek <jakub@redhat.com>
16477 PR tree-optimization/94589
16478 * tree-ssa-phiopt.c (spaceship_replacement): For integral rhs1 and
16479 rhs2, treat x <= 4 equivalently to x < 5 etc. In cmp1 and cmp2 (if
16480 not the same as cmp3) treat <= the same as < and >= the same as >.
16481 Don't require that cond2_phi_edge is true edge, instead take
16482 false/true edges into account based on cmp1/cmp2 comparison kinds.
16484 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
16487 * config/i386/mmx.md (SMAXMIN_MMXMODEI): New mode iterator.
16488 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Macroize expander
16489 from <smaxmin:code>v4hi3> and <smaxmin:code><MMXMODE14:mode>3
16490 using SMAXMIN_MMXMODEI mode iterator.
16491 (*<smaxmin:code>v4qi3): New insn pattern.
16492 (*<smaxmin:code>v2hi3): Ditto.
16493 (SMAXMIN_VI_32): New mode iterator.
16494 (<smaxmin:code><SMAXMIN_VI_32>mode3): New expander.
16495 (UMAXMIN_MMXMODEI): New mode iterator.
16496 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Macroize expander
16497 from <umaxmin:code>v8qi3> and <umaxmin:code><MMXMODE24:mode>3
16498 using UMAXMIN_MMXMODEI mode iterator.
16499 (*<umaxmin:code>v4qi3): New insn pattern.
16500 (*<umaxmin:code>v2hi3): Ditto.
16501 (UMAXMIN_VI_32): New mode iterator.
16502 (<umaxmin:code><UMAXMIN_VI_32>mode3): New expander.
16503 (abs<VI_32:mode>2): New insn pattern.
16504 (ssse3_abs<MMXMODEI:mode>2, abs<MMXMODEI:mode>2): Move from ...
16505 * config/i386/sse.md: ... here.
16507 2021-05-20 Clement Chigot <clement.chigot@atos.net>
16508 David Edelsohn <dje.gcc@gmail.com>
16510 * collect2.c (scan_prog_file): Issue non-fatal warning for
16513 2021-05-20 Jonathan Wakely <jwakely@redhat.com>
16515 * doc/invoke.texi (-Wno-c++11-extensions)
16516 (-Wno-c++14-extensions, -Wno-c++17-extensions)
16517 (-Wno-c++20-extensions, -Wno-c++23-extensions): Document
16520 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
16522 * config/c6x/c6x.c (c6x_output_file_unwind): Use dwarf_debuginfo_p.
16523 * config/darwin.c (darwin_override_options): Likewise.
16524 * config/i386/cygming.h (DBX_REGISTER_NUMBER): Likewise.
16525 * config/i386/darwin.h (DBX_REGISTER_NUMBER): Likewise.
16526 (DWARF2_FRAME_REG_OUT): Likewise.
16527 * config/mips/mips.c (mips_output_filename): Likewise.
16528 * config/rs6000/rs6000.c (rs6000_xcoff_declare_function_name):
16530 (rs6000_dbx_register_number): Likewise.
16531 * dbxout.c: Include flags.h.
16532 * dwarf2cfi.c (cfi_label_required_p): Likewise.
16533 (dwarf2out_do_frame): Likewise.
16534 * except.c: Include flags.h.
16535 * final.c (dwarf2_debug_info_emitted_p): Likewise.
16536 (final_scan_insn_1): Likewise.
16537 * flags.h (dwarf_debuginfo_p): New function declaration.
16538 * opts.c (dwarf_debuginfo_p): New function definition.
16539 * targhooks.c (default_debug_unwind_info): Use dwarf_debuginfo_p.
16540 * toplev.c (process_options): Likewise.
16542 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
16544 * common.opt: Change type to support bitmasks.
16545 * flag-types.h (enum debug_info_type): Rename enumerator constants.
16546 (NO_DEBUG): New bitmask.
16547 (DBX_DEBUG): Likewise.
16548 (DWARF2_DEBUG): Likewise.
16549 (XCOFF_DEBUG): Likewise.
16550 (VMS_DEBUG): Likewise.
16551 (VMS_AND_DWARF2_DEBUG): Likewise.
16552 * flags.h (debug_set_to_format): New function declaration.
16553 (debug_set_count): Likewise.
16554 (debug_set_names): Likewise.
16555 * opts.c (debug_type_masks): Array of bitmasks for debug formats.
16556 (debug_set_to_format): New function definition.
16557 (debug_set_count): Likewise.
16558 (debug_set_names): Likewise.
16559 (set_debug_level): Update access to debug_type_names.
16560 * toplev.c: Likewise.
16562 2021-05-20 Martin Sebor <msebor@redhat.com>
16564 PR middle-end/100684
16565 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Handle C++ lambda.
16567 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
16570 * config/i386/i386.md (isa): Remove x64_bmi.
16571 (enabled): Remove x64_bmi.
16572 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
16573 Remove general register alternative.
16574 (*andnot<VI_32:mode>3): Ditto.
16575 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
16576 (*<any_logic:code><VI_32:mode>3): Ditto.
16578 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
16580 * config/arm/arm.c: Include head files tree-vectorizer.h and
16583 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
16586 * config/i386/mmx.md (Yv_Yw): Revert adding V4QI and V2HI modes.
16587 (*<plusminus:insn><VI_32:mode>3): Use Yw instad of <Yv_Yw> constrint.
16588 (<s>mulv4hi3_highpart): New expander.
16589 (*<s>mulv2hi3_highpart): New insn pattern.
16590 (<s>mulv2hi3_higpart): New expander.
16591 (*<any_shift:insn>v2hi3): New insn pattern.
16592 (<any_shift:insn>v2hi3): New expander.
16593 * config/i386/sse.md (smulhrsv2hi3): New expander.
16594 (*smulhrsv2hi3): New insn pattern.
16596 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
16598 * doc/invoke.texi (vect-inner-loop-cost-factor): Document new
16600 * params.opt (vect-inner-loop-cost-factor): New.
16601 * targhooks.c (default_add_stmt_cost): Replace hardcoded factor
16602 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR, include head file
16603 tree-vectorizer.h and its required ones.
16604 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Replace
16605 hardcoded factor 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR.
16606 * config/arm/arm.c (arm_add_stmt_cost): Likewise.
16607 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
16608 * config/rs6000/rs6000.c (rs6000_add_stmt_cost): Likewise.
16609 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
16611 (_loop_vec_info::_loop_vec_info): Init inner_loop_cost_factor.
16612 * tree-vectorizer.h (_loop_vec_info): Add inner_loop_cost_factor.
16613 (LOOP_VINFO_INNER_LOOP_COST_FACTOR): New macro.
16615 2021-05-20 Christophe Lyon <christophe.lyon@linaro.org>
16616 Torbjörn Svensson <torbjorn.svensson@st.com>
16619 * doc/cpp.texi (Common Predefined Macros): Document __FILE_NAME__.
16621 2021-05-20 Jakub Jelinek <jakub@redhat.com>
16623 PR middle-end/99928
16624 * gimplify.c (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LINEAR>: For
16625 explicit linear clause when combined with target, make it map(tofrom:)
16626 instead of no clause or firstprivate.
16628 2021-05-20 Jakub Jelinek <jakub@redhat.com>
16630 PR tree-optimization/94589
16631 * match.pd ((X & Y) == X -> (X & ~Y) == 0): Simplify even in presence
16632 of integral conversions.
16634 2021-05-19 Andrew MacLeod <amacleod@redhat.com>
16636 * gimple-range.cc (fur_source::get_operand): New.
16637 (gimple_range_fold): Delete.
16638 (fold_using_range::fold_stmt): Move from gimple_ranger::calc_stmt.
16639 (fold_using_range::range_of_range_op): Move from gimple_ranger.
16640 (fold_using_range::range_of_address): Ditto.
16641 (fold_using_range::range_of_phi): Ditto.
16642 (fold_using_range::range_of_call): Ditto.
16643 (fold_using_range::range_of_builtin_ubsan_call): Move from
16644 range_of_builtin_ubsan_call.
16645 (fold_using_range::range_of_builtin_call): Move from
16646 range_of_builtin_call.
16647 (gimple_ranger::range_of_builtin_call): Delete.
16648 (fold_using_range::range_of_cond_expr): Move from gimple_ranger.
16649 (gimple_ranger::fold_range_internal): New.
16650 (gimple_ranger::range_of_stmt): Use new fold_using_range API.
16651 (fold_using_range::range_of_ssa_name_with_loop_info): Move from
16652 gimple_ranger. Improve ranges of SSA_NAMES when possible.
16653 * gimple-range.h (gimple_ranger): Remove various range_of routines.
16654 (class fur_source): New.
16655 (class fold_using_range): New.
16656 (fur_source::fur_source): New.
16658 * vr-values.c (vr_values::extract_range_basic): Use fold_using_range
16659 instead of range_of_builtin_call.
16661 2021-05-19 Jonathan Wakely <jwakely@redhat.com>
16663 * doc/cpp.texi (Common Predefined Macros): Update documentation
16664 for the __GXX_EXPERIMENTAL_CXX0X__ macro.
16666 2021-05-19 Alex Coplan <alex.coplan@arm.com>
16669 * config/arm/arm.md (nonsecure_call_internal): Always ensure
16670 callee's address is in a register.
16672 2021-05-19 Geng Qi <gengqi@linux.alibaba.com>
16674 * common/config/riscv/riscv-common.c
16675 (riscv_subset_list::parsing_subset_version): Properly parse the letter
16677 (riscv_subset_list::parse_std_ext,
16678 riscv_subset_list::parse_multiletter_ext): To handle errors generated
16679 in riscv_subset_list::parsing_subset_version.
16681 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
16683 * config/aarch64/aarch64-simd.md: Use "neon_move_narrow_q"
16684 type attribute in patterns generating XTN(2).
16686 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
16688 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_pack_trunc_<mode>):
16689 Remove as duplicate of...
16690 (aarch64_xtn<mode>): This.
16691 (aarch64_xtn2<mode>_le): Move position in file.
16692 (aarch64_xtn2<mode>_be): Move position in file.
16693 (aarch64_xtn2<mode>): Move position in file.
16694 (vec_pack_trunc_<mode>): Define as an expander.
16696 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
16698 * config/aarch64/aarch64-simd-builtins.def: Split builtin
16699 generation for aarch64_<sur>q<r>shr<u>n_n<mode> pattern into
16700 separate scalar and vector generators.
16701 * config/aarch64/aarch64-simd.md
16702 (aarch64_<sur>q<r>shr<u>n_n<mode>): Define as an expander and
16704 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_le): This and...
16705 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_be): This.
16706 * config/aarch64/iterators.md: Define SD_HSDI iterator.
16708 2021-05-19 Jonathn Wright <jonathan.wright@arm.com>
16710 * config/aarch64/aarch64-simd.md: Use UNSPEC_SQXTUN instead
16712 * config/aarch64/iterators.md: Remove UNSPEC_SQXTUN2.
16714 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
16716 * config/aarch64/aarch64-simd.md (aarch64_<sur>q<r>shr<u>n2_n<mode>):
16717 Implement as an expand emitting a big/little endian
16718 instruction pattern.
16719 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_le): Define.
16720 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_be): Define.
16722 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
16724 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn2<mode>):
16725 Implement as an expand emitting a big/little endian
16726 instruction pattern.
16727 (aarch64_<sur><addsub>hn2<mode>_insn_le): Define.
16728 (aarch64_<sur><addsub>hn2<mode>_insn_be): Define.
16729 * config/aarch64/iterators.md: Remove UNSPEC_[R]ADDHN2 and
16730 UNSPEC_[R]SUBHN2 unspecs and ADDSUBHN2 iterator.
16732 2021-05-19 Richard Biener <rguenther@suse.de>
16734 PR middle-end/100672
16735 * fold-const.c (fold_negate_expr_1): Use element_precision.
16736 (negate_expr_p): Likewise.
16738 2021-05-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
16740 * config/aarch64/iterators.md (SVE_PRED_LOAD): New iterator.
16741 (pred_load): New int attribute.
16742 * config/aarch64/aarch64-sve.md
16743 (aarch64_load_<ANY_EXTEND:optab><SVE_HSDI:mode><SVE_PARTIAL_I:mode>): Use
16744 SVE_PRED_LOAD enum iterator and corresponding pred_load attribute.
16745 * config/aarch64/aarch64-sve-builtins-base.cc (expand): Update call to
16746 code_for_aarch64_load.
16748 2021-05-19 Richard Biener <rguenther@suse.de>
16750 * cfgexpand.c (discover_nonconstant_array_refs_r): Make
16751 sure TARGET_MEM_REF bases are expanded as memory.
16752 * tree-ssa-operands.c (operands_scanner::get_tmr_operands):
16753 Do not mark TARGET_MEM_REF bases addressable.
16754 * tree-ssa.c (non_rewritable_mem_ref_base): Handle
16755 TARGET_MEM_REF bases as never rewritable.
16756 * gimple-walk.c (walk_stmt_load_store_addr_ops): Do not
16757 walk TARGET_MEM_REF bases as address-takens.
16758 * tree-ssa-dce.c (ref_may_be_aliased): Handle TARGET_MEM_REF.
16760 2021-05-19 Richard Biener <rguenther@suse.de>
16762 * builtins.c (get_object_alignment_1): Strip outer
16764 * tree-dfa.c (get_ref_base_and_extent): Handle outer
16765 WITH_SIZE_EXPR for size processing and process the
16767 * tree-ssa-alias.c (ao_ref_base_alias_set): Strip
16768 outer WITH_SIZE_EXPR.
16769 (ao_ref_base_alias_ptr_type): Likewise.
16770 (refs_may_alias_p_2): Allow WITH_SIZE_EXPR in ref->ref
16771 and handle that accordingly, stripping it for the
16772 core alias workers.
16773 * tree.c (get_base_address): Handle WITH_SIZE_EXPR by
16774 looking through it instead of returning NULL.
16776 2021-05-19 Jakub Jelinek <jakub@redhat.com>
16778 PR middle-end/100576
16779 * builtins.c (check_read_access): Convert bound to size_type_node if
16782 2021-05-19 Richard Biener <rguenther@suse.de>
16784 * tree-cfg.c (verify_types_in_gimple_min_lval): Inline...
16785 (verify_types_in_gimple_reference): ... here. Sanitize.
16786 (verify_gimple_call): Verify references in LHS and arguments.
16787 (verify_gimple_assign_single): Reject WITH_SIZE_EXPR.
16789 2021-05-19 Uroš Bizjak <ubizjak@gmail.com>
16791 * config/i386/i386.h (VALID_INT_MODE_P):
16792 Add V8QI, V4HI and V2SI modes for TARGET_64BIT.
16793 * config/i386/i386.md (isa): Add x64_bmi.
16794 (enabled): Handle x64_bmi.
16795 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
16796 Add alternative using 64bit general registers.
16797 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
16799 2021-05-19 Jakub Jelinek <jakub@redhat.com>
16801 PR middle-end/99928
16802 * tree.h (OMP_MASTER_COMBINED): Define.
16803 * gimplify.c (gimplify_scan_omp_clauses): Rewrite lastprivate
16804 handling for outer combined/composite constructs to a loop.
16805 Handle lastprivate on combined target.
16806 (gimplify_expr): Formatting fix.
16808 2021-05-19 Xionghu Luo <luoxhu@linux.ibm.com>
16810 * passes.def: Add sink_code pass before store_merging.
16811 * tree-ssa-sink.c (pass_sink_code:clone): New.
16813 2021-05-18 Bill Schmidt <wschmidt@linux.ibm.com>
16815 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Remove call to
16816 rs6000_special_adjust_field_align_p.
16817 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
16818 * config/rs6000/rs6000-call.c (rs6000_function_arg_boundary):
16819 Remove ABI warning.
16820 (rs6000_function_arg): Likewise.
16821 * config/rs6000/rs6000-protos.h
16822 (rs6000_special_adjust_field_align_p): Remove prototype.
16823 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
16825 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Remove call to
16826 rs6000_special_adjust_field_align_p.
16828 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
16831 * config/i386/i386.h (VALID_SSE2_REG_MODE):
16832 Add V4QI and V2HI modes.
16833 (VALID_INT_MODE_P): Ditto.
16834 * config/i386/mmx.md (VI_32): New mode iterator.
16835 (mmxvecsize): Handle V4QI and V2HI.
16837 (mov<VI_32:mode>): New expander.
16838 (*mov<mode>_internal): New insn pattern.
16839 (movmisalign<VI_32:mode>): New expander.
16840 (neg<VI_32:mode>): New expander.
16841 (<plusminus:insn><VI_32:mode>3): New expander.
16842 (*<plusminus:insn><VI_32:mode>3): New insn pattern.
16843 (mulv2hi3): New expander.
16844 (*mulv2hi3): New insn pattern.
16845 (one_cmpl<VI_32:mode>2): New expander.
16846 (*andnot<VI_32:mode>3): New insn pattern.
16847 (<any_logic:code><VI_32:mode>3): New expander.
16848 (*<any_logic:code><VI_32:mode>3): New insn pattern.
16850 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
16852 * config/i386/sse.md (<any_extend:insn>v4qiv4di2):
16853 Fix a mode mismatch with operand 1.
16855 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
16858 * config/i386/i386-expand.c (split_double_mode): Return
16859 temporary register when simplify_gen_subreg fails with
16860 the high half od the paradoxical subreg.
16862 2021-05-18 Richard Biener <rguenther@suse.de>
16864 * cfgexpand.c (expand_one_var): Pass in forced_stack_var
16865 and honor it when expanding.
16866 (expand_used_vars_for_block): Pass through forced_stack_var.
16867 (expand_used_vars): Likewise.
16868 (discover_nonconstant_array_refs_r): Set bits in
16869 forced_stack_vars instead of marking vars TREE_ADDRESSABLE.
16870 (avoid_type_punning_on_regs): Likewise.
16871 (discover_nonconstant_array_refs): Likewise.
16872 (pass_expand::execute): Create and pass down forced_stack_var
16873 bitmap. For parameters and returns temporarily set
16874 TREE_ADDRESSABLE when expand_function_start.
16876 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
16878 * doc/sourcebuild.texi: Document 'dg-note'.
16880 2021-05-18 Tobias Burnus <tobias@codesourcery.com>
16883 * configure: Regenerate.
16884 * configure.ac (BUILD_CFLAG, BUILD_CXXFLAGS): Add $(CFLAGS-$@).
16886 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
16888 * gimple.h (is_gimple_omp_oacc): Tighten.
16889 * omp-low.c (check_omp_nesting_restrictions): Adjust.
16891 2021-05-18 Richard Biener <rguenther@suse.de>
16893 * tree-ssa-operands.c (mark_address_taken): Simplify.
16895 2021-05-18 Martin Liska <mliska@suse.cz>
16897 * config/gcn/mkoffload.c (STR): Redefine.
16898 * config/i386/intelmic-mkoffload.c (STR): Likewise.
16899 * config/nvptx/mkoffload.c (STR): Likewise.
16901 2021-05-18 Martin Liska <mliska@suse.cz>
16903 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
16904 Use startswith function instead of strncmp.
16905 * common/config/bfin/bfin-common.c (bfin_handle_option): Likewise.
16906 * common/config/riscv/riscv-common.c (riscv_subset_list::parse): Likewise.
16907 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): Likewise.
16908 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Likewise.
16909 * config/alpha/alpha.c (alpha_elf_section_type_flags): Likewise.
16910 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
16911 * config/arm/arm.c (arm_file_start): Likewise.
16912 (arm_valid_target_attribute_rec): Likewise.
16913 (thumb1_md_asm_adjust): Likewise.
16914 * config/arm/driver-arm.c (host_detect_local_cpu): Likewise.
16915 * config/avr/avr.c (STR_PREFIX_P): Likewise.
16916 (avr_set_current_function): Likewise.
16917 (avr_handle_addr_attribute): Likewise.
16918 (avr_asm_output_aligned_decl_common): Likewise.
16919 (avr_asm_named_section): Likewise.
16920 (avr_section_type_flags): Likewise.
16921 (avr_asm_select_section): Likewise.
16922 * config/c6x/c6x.c (c6x_in_small_data_p): Likewise.
16923 (c6x_section_type_flags): Likewise.
16924 * config/darwin-c.c (darwin_cfstring_ref_p): Likewise.
16925 (darwin_objc_declare_unresolved_class_reference): Likewise.
16926 (darwin_objc_declare_class_definition): Likewise.
16927 * config/darwin.c (indirect_data): Likewise.
16928 (darwin_encode_section_info): Likewise.
16929 (darwin_objc2_section): Likewise.
16930 (darwin_objc1_section): Likewise.
16931 (machopic_select_section): Likewise.
16932 (darwin_globalize_label): Likewise.
16933 (darwin_label_is_anonymous_local_objc_name): Likewise.
16934 (darwin_asm_named_section): Likewise.
16935 (darwin_asm_output_dwarf_offset): Likewise.
16936 * config/frv/frv.c (frv_string_begins_with): Likewise.
16937 (frv_in_small_data_p): Likewise.
16938 * config/gcn/mkoffload.c (STR): Likewise.
16940 * config/i386/i386-builtins.c (get_builtin_code_for_version): Likewise.
16941 * config/i386/i386-options.c (ix86_option_override_internal): Likewise.
16942 * config/i386/i386.c (x86_64_elf_section_type_flags): Likewise.
16943 (ix86_md_asm_adjust): Likewise.
16944 * config/i386/intelmic-mkoffload.c (STR): Likewise.
16945 * config/i386/winnt.c (i386_pe_asm_named_section): Likewise.
16946 (i386_pe_file_end): Likewise.
16947 * config/ia64/ia64.c (ia64_in_small_data_p): Likewise.
16948 (ia64_section_type_flags): Likewise.
16949 * config/mips/driver-native.c (host_detect_local_cpu): Likewise.
16950 * config/mips/mips.c (mips_handle_interrupt_attr): Likewise.
16951 (mips16_stub_function_p): Likewise.
16952 (mips_function_rodata_section): Likewise.
16953 * config/msp430/msp430.c (msp430_mcu_name): Likewise.
16954 (msp430_function_section): Likewise.
16955 (msp430_section_type_flags): Likewise.
16956 (msp430_expand_helper): Likewise.
16957 * config/nios2/nios2.c (nios2_small_section_name_p): Likewise.
16958 (nios2_valid_target_attribute_rec): Likewise.
16959 * config/nvptx/mkoffload.c (process): Likewise.
16961 * config/pa/som.h: Likewise.
16962 * config/pdp11/pdp11.c (pdp11_output_ident): Likewise.
16963 * config/riscv/riscv.c (riscv_elf_select_rtx_section): Likewise.
16964 * config/rs6000/rs6000.c (VTABLE_NAME_P): Likewise.
16965 (rs6000_inner_target_options): Likewise.
16966 * config/s390/driver-native.c (s390_host_detect_local_cpu): Likewise.
16967 * config/sparc/driver-sparc.c (host_detect_local_cpu): Likewise.
16968 * config/vax/vax.c (vax_output_int_move): Likewise.
16969 * config/vms/vms-ld.c (startswith): Likewise.
16970 (process_args): Likewise.
16972 * config/vms/vms.c: Likewise.
16974 2021-05-18 Jakub Jelinek <jakub@redhat.com>
16976 PR rtl-optimization/100590
16977 * regcprop.c (copyprop_hardreg_forward_1): Only DCE dead sets if
16978 they are NONJUMP_INSN_P.
16980 2021-05-18 Jakub Jelinek <jakub@redhat.com>
16983 * function.c (push_dummy_function): Set DECL_ARTIFICIAL and
16984 DECL_ASSEMBLER_NAME on the fn_decl.
16986 2021-05-18 Jakub Jelinek <jakub@redhat.com>
16988 PR tree-optimization/94589
16989 * tree-ssa-phiopt.c (spaceship_replacement): Pattern match
16990 phi result used in (res & ~1) == 0 comparison as res >= 0 as
16991 res == 2 would be UB with -ffinite-math-only.
16993 2021-05-18 Martin Liska <mliska@suse.cz>
16995 * Makefile.in: genversion.o should depend on DATESTAMP.
16997 2021-05-18 Claudiu Zissulescu <claziss@synopsys.com>
16999 * config/arc/simdext.md (negv2si2): Remove round bracket.
17001 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
17003 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Define
17004 _Bool as macro expanding to _Bool.
17006 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
17009 * tree.c (build_reference_type_for_mode)
17010 (build_pointer_type_for_mode): Pick pointer mode if MODE argument
17012 (build_reference_type, build_pointer_type): Invoke
17013 build_*_type_for_mode with VOIDmode.
17015 2021-05-17 Andrew MacLeod <amacleod@redhat.com>
17017 PR tree-optimization/100512
17018 * gimple-range-cache.cc (ranger_cache::set_global_range): Mark const
17019 and non-zero pointer ranges as invariant.
17020 * gimple-range.cc (gimple_ranger::range_of_stmt): Remove pointer
17021 processing from here.
17023 2021-05-17 Tom de Vries <tdevries@suse.de>
17026 * config/nvptx/nvptx-protos.h (nvptx_output_atomic_insn): Declare
17027 * config/nvptx/nvptx.c (nvptx_output_barrier)
17028 (nvptx_output_atomic_insn): New function.
17029 (nvptx_print_operand): Add support for 'B'.
17030 * config/nvptx/nvptx.md: Use nvptx_output_atomic_insn for atomic
17033 2021-05-17 Aldy Hernandez <aldyh@redhat.com>
17035 PR tree-optimization/100349
17036 * vr-values.c (bounds_of_var_in_loop): Bail if scev returns
17039 2021-05-17 Tamar Christina <tamar.christina@arm.com>
17041 * config/aarch64/driver-aarch64.c (DEFAULT_ARCH): New.
17042 (host_detect_local_cpu): Use it.
17044 2021-05-17 Martin Liska <mliska@suse.cz>
17046 * doc/invoke.texi: Add 2 missing dots.
17048 2021-05-17 Marius Hillenbrand <mhillen@linux.ibm.com>
17050 PR bootstrap/100552
17051 * configure.ac: Replace pattern substitution with call to sed.
17052 * configure: Regenerate.
17054 2021-05-17 Richard Biener <rguenther@suse.de>
17056 PR middle-end/100582
17057 * tree.c (array_at_struct_end_p): Get to the base of the
17058 reference before looking for the underlying decl.
17060 2021-05-17 Joern Rennecke <joern.rennecke@embecosm.com>
17062 * genoutput.c (validate_insn_alternatives) Make "wrong number of
17063 alternatives" message more specific, and remove assumption on where
17066 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
17068 * config/arm/iterators.md (V16): New iterator.
17069 (VH_cvtto): New iterator.
17070 (v_cmp_result): Added V4HF and V8HF support.
17071 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result>): Use VDQWH.
17072 (vcond<mode><mode>): Likewise.
17073 (vcond_mask_<mode><v_cmp_result>): Likewise.
17074 (vcond<VH_cvtto><mode>): New expander.
17076 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
17078 * config/arm/arm-protos.h (arm_expand_vector_compare): Update
17080 * config/arm/arm.c (arm_expand_vector_compare): Add support for
17082 (arm_expand_vcond): Likewise.
17083 * config/arm/iterators.md (supf): Remove VCMPNEQ_S, VCMPEQQ_S,
17084 VCMPEQQ_N_S, VCMPNEQ_N_S.
17085 (VCMPNEQ, VCMPEQQ, VCMPEQQ_N, VCMPNEQ_N): Remove.
17086 * config/arm/mve.md (@mve_vcmp<mve_cmp_op>q_<mode>): Add '@' prefix.
17087 (@mve_vcmp<mve_cmp_op>q_f<mode>): Likewise.
17088 (@mve_vcmp<mve_cmp_op>q_n_f<mode>): Likewise.
17089 (@mve_vpselq_<supf><mode>): Likewise.
17090 (@mve_vpselq_f<mode>"): Likewise.
17091 * config/arm/neon.md (vec_cmp<mode><v_cmp_result): Enable for MVE
17092 and move to vec-common.md.
17093 (vec_cmpu<mode><mode>): Likewise.
17094 (vcond<mode><mode>): Likewise.
17095 (vcond<V_cvtto><mode>): Likewise.
17096 (vcondu<mode><v_cmp_result>): Likewise.
17097 (vcond_mask_<mode><v_cmp_result>): Likewise.
17098 * config/arm/unspecs.md (VCMPNEQ_U, VCMPNEQ_S, VCMPEQQ_S)
17099 (VCMPEQQ_N_S, VCMPNEQ_N_S, VCMPEQQ_U, CMPEQQ_N_U, VCMPNEQ_N_U)
17100 (VCMPGEQ_N_S, VCMPGEQ_S, VCMPGTQ_N_S, VCMPGTQ_S, VCMPLEQ_N_S)
17101 (VCMPLEQ_S, VCMPLTQ_N_S, VCMPLTQ_S, VCMPCSQ_N_U, VCMPCSQ_U)
17102 (VCMPHIQ_N_U, VCMPHIQ_U): Remove.
17103 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result): Moved
17105 (vec_cmpu<mode><mode>): Likewise.
17106 (vcond<mode><mode>): Likewise.
17107 (vcond<V_cvtto><mode>): Likewise.
17108 (vcondu<mode><v_cmp_result>): Likewise.
17109 (vcond_mask_<mode><v_cmp_result>): Likewise. Added unsafe math
17112 2021-05-17 liuhongt <hongtao.liu@intel.com>
17115 * config/i386/i386.c (ix86_gimple_fold_builtin): Use
17116 gsi_insert_seq_before instead.
17118 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
17120 * doc/sourcebuild.texi (arm_qbit_ok): Rename into...
17121 (arm_sat_ok): ...this.
17123 2021-05-17 Martin Liska <mliska@suse.cz>
17125 * lto-wrapper.c (merge_flto_options): Factor out a new function.
17126 (merge_and_complain): Use it.
17127 (run_gcc): Merge also linker command line -flto=foo argument
17130 2021-05-16 Christophe Lyon <christophe.lyon@linaro.org>
17132 * config/arm/arm.h (CPP_SPEC): Remove error message about
17133 -mlittle-endian/-mbig-endian conflict.
17135 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
17137 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
17138 __ROP_PROTECT__ if -mrop-protect is selected.
17140 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
17142 * config/rs6000/rs6000-internal.h (rs6000_stack): Add
17143 rop_hash_save_offset and rop_hash_size.
17144 * config/rs6000/rs6000-logue.c (rs6000_stack_info): Compute
17145 rop_hash_size and rop_hash_save_offset.
17146 (debug_stack_info): Dump rop_hash_save_offset and rop_hash_size.
17147 (rs6000_emit_prologue): Emit hashst[p] in prologue.
17148 (rs6000_emit_epilogue): Emit hashchk[p] in epilogue.
17149 * config/rs6000/rs6000.md (unspec): Add UNSPEC_HASHST and
17151 (hashst): New define_insn.
17152 (hashchk): Likewise.
17154 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
17156 * config/rs6000/rs6000.c (rs6000_option_override_internal):
17157 Disable shrink wrap when inserting ROP-protect instructions.
17158 * config/rs6000/rs6000.opt (mrop-protect): New option.
17159 (mprivileged): Likewise.
17160 * doc/invoke.texi: Document mrop-protect and mprivileged.
17162 2021-05-15 Hans-Peter Nilsson <hp@axis.com>
17164 * reorg.c (fill_slots_from_thread): Reinstate code typoed out in
17167 2021-05-15 Martin Jambor <mjambor@suse.cz>
17170 2021-05-13 Martin Jambor <mjambor@suse.cz>
17172 PR tree-optimization/100453
17173 * tree-sra.c (sra_modify_assign): All const base accesses do not
17174 need refreshing, not just those from decl_pool.
17175 (sra_modify_assign): Do not refresh into a const base decl.
17177 2021-05-15 Jakub Jelinek <jakub@redhat.com>
17179 PR rtl-optimization/100342
17180 * regcprop.c (copy_value): When copying a source reg in a wider
17181 mode than it has recorded for the value, adjust recorded destination
17182 mode too or punt if !REG_CAN_CHANGE_MODE_P.
17184 2021-05-14 Jason Merrill <jason@redhat.com>
17186 * intl.h: Add comments.
17188 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17190 * config/aarch64/aarch64-simd.md
17191 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Split into...
17192 (aarch64_sqdmlsl2_lane<mode>_internal): ... This...
17193 (aarch64_sqdmlal2_lane<mode>_internal): ... And this.
17194 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Split into ...
17195 (aarch64_sqdmlsl2_laneq<mode>_internal): ... This...
17196 (aarch64_sqdmlal2_laneq<mode>_internal): ... And this.
17197 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>_internal): Split into...
17198 (aarch64_sqdmlsl2_n<mode>_internal): ... This...
17199 (aarch64_sqdmlal2_n<mode>_internal): ... And this.
17201 2021-05-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17204 * config/arm/arm_neon.h (vtst_s8): Replace call to vtst builtin with it's
17205 boolean logic equivalent.
17206 (vtst_s16): Likewise.
17207 (vtst_s32): Likewise.
17208 (vtst_u8): Likewise.
17209 (vtst_u16): Likewise.
17210 (vtst_u32): Likewise.
17211 (vtst_p8): Likewise.
17212 (vtst_p16): Likewise.
17213 (vtstq_s8): Likewise.
17214 (vtstq_s16): Likewise.
17215 (vtstq_s32): Likewise.
17216 (vtstq_u8): Likewise.
17217 (vtstq_u16): Likewise.
17218 (vtstq_u32): Likewise.
17219 (vtstq_p8): Likewise.
17220 (vtstq_p16): Likewise.
17221 * config/arm/arm_neon_builtins.def: Remove entry for vtst.
17222 * config/arm/neon.md (neon_vtst<mode>): Remove pattern.
17224 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17226 * config/aarch64/aarch64-simd.md (aarch64_sqdmlal2<mode>): Merge into...
17227 (aarch64_sqdml<SBINQOPS:as>l2<mode>): ... This.
17228 (aarch64_sqdmlsl2<mode>): Delete.
17229 (aarch64_sqdmlal2_lane<mode>): Merge this...
17230 (aarch64_sqdmlsl2_lane<mode>): ... And this...
17231 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>): ... Into this.
17232 (aarch64_sqdmlal2_laneq<mode>): Merge this...
17233 (aarch64_sqdmlsl2_laneq<mode>): ... And this...
17234 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>): ... Into this.
17235 (aarch64_sqdmlal2_n<mode>): Merge this...
17236 (aarch64_sqdmlsl2_n<mode>): ... And this...
17237 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>): ... Into this.
17239 2021-05-13 Martin Sebor <msebor@redhat.com>
17241 PR middle-end/100574
17242 * builtins.c (access_ref::get_ref): Improve detection of PHIs with
17243 all null arguments.
17245 2021-05-13 Martin Sebor <msebor@redhat.com>
17247 PR tree-optimization/93100
17248 PR middle-end/98583
17249 * tree-ssa-uninit.c (check_defs): Exclude intrinsic functions that
17250 don't modify referenced objects.
17252 2021-05-13 Martin Jambor <mjambor@suse.cz>
17254 PR tree-optimization/100453
17255 * tree-sra.c (sra_modify_assign): All const base accesses do not
17256 need refreshing, not just those from decl_pool.
17257 (sra_modify_assign): Do not refresh into a const base decl.
17259 2021-05-13 Martin Liska <mliska@suse.cz>
17261 * tree-ssa-dom.c: Remove m_simplifier.
17263 2021-05-13 Richard Earnshaw <rearnsha@arm.com>
17266 * config/arm/arm.c (arm_canonicalize_comparison): Correctly
17267 canonicalize DImode inequality comparisons against the
17268 maximum integral value.
17270 2021-05-13 Jakub Jelinek <jakub@redhat.com>
17272 PR tree-optimization/98856
17273 * config/i386/i386.c (ix86_shift_rotate_cost): Add CODE argument.
17274 Expect V2DI and V4DI arithmetic right shifts to be emulated.
17275 (ix86_rtx_costs, ix86_add_stmt_cost): Adjust ix86_shift_rotate_cost
17277 * config/i386/i386-expand.c (expand_vec_perm_2perm_interleave,
17278 expand_vec_perm_2perm_pblendv): New functions.
17279 (ix86_expand_vec_perm_const_1): Use them.
17280 * config/i386/sse.md (ashr<mode>3<mask_name>): Rename to ...
17281 (<mask_codefor>ashr<mode>3<mask_name>): ... this.
17282 (ashr<mode>3): New define_expand with VI248_AVX512BW iterator.
17283 (ashrv4di3): New define_expand.
17284 (ashrv2di3): Change condition to TARGET_SSE2, handle !TARGET_XOP
17285 and !TARGET_AVX512VL expansion.
17287 2021-05-13 Uroš Bizjak <ubizjak@gmail.com>
17290 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Force mode
17291 sizes < 16 to a register when constructing vpcmov pattern.
17292 * config/i386/mmx.md (*xop_pcmov_<mode>): Use MMXMODE124 mode.
17294 2021-05-13 Martin Liska <mliska@suse.cz>
17296 * gcov-io.c (gcov_write_block): Remove.
17297 (gcov_write_words): Likewise.
17298 (gcov_read_words): Re-implement using gcov_read_bytes.
17299 (gcov_allocate): Remove.
17300 (GCOV_BLOCK_SIZE): Likewise.
17301 (struct gcov_var): Remove most of the fields.
17302 (gcov_position): Implement with ftell.
17303 (gcov_rewrite): Remove setting of start and offset fields.
17304 (from_file): Re-format.
17305 (gcov_open): Remove setbuf call. It should not be needed.
17306 (gcov_close): Remove internal buffer handling.
17307 (gcov_magic): Use __builtin_bswap32.
17308 (gcov_write_counter): Use directly gcov_write_unsigned.
17309 (gcov_write_string): Use direct fwrite and do not round
17311 (gcov_seek): Use directly fseek.
17312 (gcov_write_tag): Use gcov_write_unsigned directly.
17313 (gcov_write_length): Likewise.
17314 (gcov_write_tag_length): Likewise.
17315 (gcov_read_bytes): Use directly fread.
17316 (gcov_read_unsigned): Use gcov_read_words.
17317 (gcov_read_counter): Likewise.
17318 (gcov_read_string): Use gcov_read_bytes.
17319 * gcov-io.h (GCOV_WORD_SIZE): Adjust to reflect
17320 that size is not in bytes, but words (4B).
17321 (GCOV_TAG_FUNCTION_LENGTH): Likewise.
17322 (GCOV_TAG_ARCS_LENGTH): Likewise.
17323 (GCOV_TAG_ARCS_NUM): Likewise.
17324 (GCOV_TAG_COUNTER_LENGTH): Likewise.
17325 (GCOV_TAG_COUNTER_NUM): Likewise.
17326 (GCOV_TAG_SUMMARY_LENGTH): Likewise.
17328 2021-05-13 liuhongt <hongtao.liu@intel.com>
17331 * config/i386/sse.md (ssedoublevecmode): Add attribute for
17332 V64QI/V32HI/V16SI/V4DI.
17333 (ssehalfvecmode): Add attribute for V2DI/V2DF.
17334 (*vec_concatv4si_0): Extend to VI124_128.
17335 (*vec_concat<mode>_0): New pre-reload splitter.
17336 * config/i386/predicates.md (movq_parallel): New predicate.
17338 2021-05-13 Alexandre Oliva <oliva@adacore.com>
17340 * targhooks.c (default_zero_call_used_regs): Retry using
17341 successfully-zeroed registers as sources.
17343 2021-05-12 Tobias Burnus <tobias@codesourcery.com>
17345 * omp-low.c (finish_taskreg_scan): Use the proper detach decl.
17347 2021-05-12 Aldy Hernandez <aldyh@redhat.com>
17350 * gimple-range.cc (range_of_builtin_call): Skip out on
17351 processing __builtin_clz when varying.
17353 2021-05-12 Tom de Vries <tdevries@suse.de>
17356 * config/nvptx/nvptx-opts.h (enum ptx_version): New enum.
17357 * config/nvptx/nvptx.c (nvptx_file_start): Print .version according
17358 to ptx_version_option.
17359 * config/nvptx/nvptx.h (TARGET_PTX_6_3): Define.
17360 * config/nvptx/nvptx.md (define_insn "nvptx_shuffle<mode>")
17361 (define_insn "nvptx_vote_ballot"): Use sync variant for
17363 * config/nvptx/nvptx.opt (ptx_version): Add enum.
17364 (mptx): Add option.
17365 * doc/invoke.texi (Nvidia PTX Options): Add mptx item.
17367 2021-05-12 Richard Biener <rguenther@suse.de>
17369 PR tree-optimization/100566
17370 * tree-ssa-sccvn.c (dominated_by_p_w_unex): Properly handle
17371 allow_back for all edge queries.
17373 2021-05-12 liuhongt <hongtao.liu@intel.com>
17376 * config/i386/sse.md (<sse4_1_avx2>_pblendvb): Add
17377 splitters for pblendvb of NOT mask register.
17379 2021-05-12 Richard Biener <rguenther@suse.de>
17381 PR tree-optimization/100519
17382 * tree-ssa-reassoc.c (can_associate_p): Split into...
17383 (can_associate_op_p): ... this
17384 (can_associate_type_p): ... and this.
17385 (is_reassociable_op): Call can_associate_op_p.
17386 (break_up_subtract_bb): Call the appropriate predicates.
17387 (reassociate_bb): Likewise.
17389 2021-05-12 Martin Liska <mliska@suse.cz>
17391 * lto-wrapper.c (merge_and_complain): Merge -flto=arg options.
17392 (run_gcc): Use -flto argument detection for merged
17395 2021-05-12 Martin Liska <mliska@suse.cz>
17397 * lto-wrapper.c (print_lto_docs_link): New function.
17398 (run_gcc): Print warning about missing job server detection
17399 after we know NR of partitions. Do the same for -flto{,=1}.
17400 * opts.c (get_option_html_page): Support -flto option.
17402 2021-05-12 Martin Liska <mliska@suse.cz>
17404 * lto-wrapper.c (get_options_from_collect_gcc_options): Change
17406 (append_option): Remove.
17407 (find_option): Rework to use the vector type.
17408 (remove_option): Remove.
17409 (merge_and_complain): Use vectors for cl_decoded_option data
17411 (append_compiler_options): Likewise.
17412 (append_diag_options): Likewise.
17413 (append_linker_options): Likewise.
17414 (append_offload_options): Likewise.
17415 (compile_offload_image): Likewise.
17416 (compile_images_for_offload_targets): Likewise.
17417 (find_and_merge_options): Likewise.
17418 (run_gcc): Likewise.
17420 2021-05-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
17423 * dwarf2out.c (dwarf2out_finish): Set
17424 have_multiple_function_sections with multi-range text_section.
17426 2021-05-12 Martin Liska <mliska@suse.cz>
17428 PR bootstrap/100560
17429 * Makefile.in: Remove version.h from linker command line.
17431 2021-05-12 Richard Biener <rguenther@suse.de>
17433 PR middle-end/100547
17434 * rtl.h (rtvec_alloc): Make argument size_t.
17435 * rtl.c (rtvec_alloc): Verify the count is less than INT_MAX.
17437 2021-05-12 Jakub Jelinek <jakub@redhat.com>
17439 PR middle-end/100508
17440 * cfgexpand.c (expand_debug_expr): For DEBUG_EXPR_DECL with vector
17441 type, don't reuse DECL_RTL if it has different mode, instead force
17442 creation of a new DEBUG_EXPR.
17444 2021-05-12 Jakub Jelinek <jakub@redhat.com>
17445 Marc Glisse <marc.glisse@inria.fr>
17447 PR tree-optimization/94589
17448 * match.pd ((X & Y) == X -> (X & ~Y) == 0,
17449 (X | Y) == Y -> (X & ~Y) == 0): New GIMPLE simplifications.
17451 2021-05-12 Uroš Bizjak <ubizjak@gmail.com>
17454 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Handle V2SF mode.
17455 * config/i386/mmx.md (MMXMODE124): New mode iterator.
17457 (mmxintvecmode): New mode attribute.
17458 (mmxintvecmodelower): Ditto.
17459 (*mmx_maskcmpv2sf3_comm): New insn pattern.
17460 (*mmx_maskcmpv2sf3): Ditto.
17461 (vec_cmpv2sfv2si): New expander.
17462 (vcond<V2FI:mode>v2si): Ditto.
17463 (mmx_vlendvps): New insn pattern.
17464 (vcond<MMXMODE124:mode><MMXMODEI:mode>): Also handle V2SFmode.
17465 (vcondu<MMXMODE124:mode><MMXMODEI:mode>): Ditto.
17466 (vcond_mask_<mode><mmxintvecmodelower>): Ditto.
17468 2021-05-11 Martin Sebor <msebor@redhat.com>
17470 PR middle-end/21433
17471 * expr.c (expand_expr_real_1): Replace unreachable code with an assert.
17473 2021-05-11 Richard Biener <rguenther@suse.de>
17475 * gimple-fold.c (gimple_fold_call): Do not call
17476 maybe_fold_reference on call arguments or the static chain.
17477 (fold_stmt_1): Do not call maybe_fold_reference on GIMPLE_ASM
17480 2021-05-11 Martin Liska <mliska@suse.cz>
17482 * builtins.def (DEF_HSAIL_BUILTIN): Remove.
17483 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
17484 (DEF_HSAIL_SAT_BUILTIN): Likewise.
17485 (DEF_HSAIL_INTR_BUILTIN): Likewise.
17486 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
17487 * doc/frontends.texi: Remove BRIG.
17488 * doc/install.texi: Likewise.
17489 * doc/invoke.texi: Likewise.
17490 * doc/standards.texi: Likewise.
17491 * brig-builtins.def: Removed.
17492 * brig/ChangeLog: Removed.
17493 * brig/Make-lang.in: Removed.
17494 * brig/brig-builtins.h: Removed.
17495 * brig/brig-c.h: Removed.
17496 * brig/brig-lang.c: Removed.
17497 * brig/brigfrontend/brig-arg-block-handler.cc: Removed.
17498 * brig/brigfrontend/brig-atomic-inst-handler.cc: Removed.
17499 * brig/brigfrontend/brig-basic-inst-handler.cc: Removed.
17500 * brig/brigfrontend/brig-branch-inst-handler.cc: Removed.
17501 * brig/brigfrontend/brig-cmp-inst-handler.cc: Removed.
17502 * brig/brigfrontend/brig-code-entry-handler.cc: Removed.
17503 * brig/brigfrontend/brig-code-entry-handler.h: Removed.
17504 * brig/brigfrontend/brig-comment-handler.cc: Removed.
17505 * brig/brigfrontend/brig-control-handler.cc: Removed.
17506 * brig/brigfrontend/brig-copy-move-inst-handler.cc: Removed.
17507 * brig/brigfrontend/brig-cvt-inst-handler.cc: Removed.
17508 * brig/brigfrontend/brig-fbarrier-handler.cc: Removed.
17509 * brig/brigfrontend/brig-function-handler.cc: Removed.
17510 * brig/brigfrontend/brig-function.cc: Removed.
17511 * brig/brigfrontend/brig-function.h: Removed.
17512 * brig/brigfrontend/brig-inst-mod-handler.cc: Removed.
17513 * brig/brigfrontend/brig-label-handler.cc: Removed.
17514 * brig/brigfrontend/brig-lane-inst-handler.cc: Removed.
17515 * brig/brigfrontend/brig-machine.c: Removed.
17516 * brig/brigfrontend/brig-machine.h: Removed.
17517 * brig/brigfrontend/brig-mem-inst-handler.cc: Removed.
17518 * brig/brigfrontend/brig-module-handler.cc: Removed.
17519 * brig/brigfrontend/brig-queue-inst-handler.cc: Removed.
17520 * brig/brigfrontend/brig-seg-inst-handler.cc: Removed.
17521 * brig/brigfrontend/brig-signal-inst-handler.cc: Removed.
17522 * brig/brigfrontend/brig-to-generic.cc: Removed.
17523 * brig/brigfrontend/brig-to-generic.h: Removed.
17524 * brig/brigfrontend/brig-util.cc: Removed.
17525 * brig/brigfrontend/brig-util.h: Removed.
17526 * brig/brigfrontend/brig-variable-handler.cc: Removed.
17527 * brig/brigfrontend/hsa-brig-format.h: Removed.
17528 * brig/brigfrontend/phsa.h: Removed.
17529 * brig/brigspec.c: Removed.
17530 * brig/config-lang.in: Removed.
17531 * brig/gccbrig.texi: Removed.
17532 * brig/lang-specs.h: Removed.
17533 * brig/lang.opt: Removed.
17535 2021-05-11 Richard Biener <rguenther@suse.de>
17538 * ipa-param-manipulation.c
17539 (ipa_param_body_adjustments::modify_call_stmt): Avoid
17540 altering SSA_NAME_DEF_STMT by adjusting the calls LHS
17541 via gimple_call_lhs_ptr.
17543 2021-05-11 Alex Coplan <alex.coplan@arm.com>
17546 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear):
17547 Avoid emitting CFA adjusts on the sp if we have the fp.
17549 2021-05-11 Richard Sandiford <richard.sandiford@arm.com>
17551 * config/aarch64/iterators.md (VMUL_CHANGE_NLANES): Delete.
17552 (VMULD): New iterator.
17553 (VCOND): Handle V4HF and V8HF.
17554 (VCONQ): Fix entry for V2SF.
17555 * config/aarch64/aarch64-simd.md (mul_lane<mode>3): Use VMULD
17556 instead of VMUL. Use a 64-bit vector mode for the indexed operand.
17557 (*aarch64_mul3_elt_<vswap_width_name><mode>): Merge with...
17558 (mul_laneq<mode>3): ...this define_insn. Use VMUL instead of VDQSF.
17559 Use a 128-bit vector mode for the indexed operand. Use stype for
17560 the scheduling type.
17562 2021-05-11 Richard Biener <rguenther@suse.de>
17564 * gimple-fold.c (maybe_fold_reference): Only return
17565 is_gimple_min_invariant values.
17567 2021-05-11 Richard Biener <rguenther@suse.de>
17569 PR middle-end/100509
17570 * gimple-fold.c (fold_gimple_assign): Only call
17571 get_symbol_constant_value on register type symbols.
17573 2021-05-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
17574 Joe Ramsay <joe.ramsay@arm.com>
17577 * config/arm/arm_mve.h (__arm_vstrwq_scatter_offset): Fix wrong arguments.
17578 (__arm_vcmpneq): Remove duplicate definition.
17579 (__arm_vstrwq_scatter_offset_p): Likewise.
17580 (__arm_vmaxq_x): Likewise.
17581 (__arm_vmlsdavaq): Likewise.
17582 (__arm_vmlsdavaxq): Likewise.
17583 (__arm_vmlsdavq_p): Likewise.
17584 (__arm_vmlsdavxq_p): Likewise.
17585 (__arm_vrmlaldavhaq): Likewise.
17586 (__arm_vstrbq_p): Likewise.
17587 (__arm_vstrbq_scatter_offset): Likewise.
17588 (__arm_vstrbq_scatter_offset_p): Likewise.
17589 (__arm_vstrdq_scatter_offset): Likewise.
17590 (__arm_vstrdq_scatter_offset_p): Likewise.
17591 (__arm_vstrdq_scatter_shifted_offset): Likewise.
17592 (__arm_vstrdq_scatter_shifted_offset_p): Likewise.
17594 2021-05-11 Jakub Jelinek <jakub@redhat.com>
17596 PR middle-end/100471
17597 * omp-low.c (lower_omp_task_reductions): For OMP_TASKLOOP, if data
17598 is 0, bypass the reduction loop including
17599 GOMP_taskgroup_reduction_unregister call.
17601 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
17603 * config/rs6000/rs6000.c (struct rs6000_cost_data): New member
17604 costing_for_scalar.
17605 (rs6000_density_test): Early return if costing_for_scalar is true.
17606 (rs6000_init_cost): Init costing_for_scalar of rs6000_cost_data.
17608 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
17610 * doc/tm.texi: Regenerated.
17611 * target.def (init_cost): Add new parameter costing_for_scalar.
17612 * targhooks.c (default_init_cost): Adjust for new parameter.
17613 * targhooks.h (default_init_cost): Likewise.
17614 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
17615 (vect_compute_single_scalar_iteration_cost): Likewise.
17616 (vect_analyze_loop_2): Likewise.
17617 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
17618 (vect_bb_vectorization_profitable_p): Likewise.
17619 * tree-vectorizer.h (init_cost): Likewise.
17620 * config/aarch64/aarch64.c (aarch64_init_cost): Likewise.
17621 * config/i386/i386.c (ix86_init_cost): Likewise.
17622 * config/rs6000/rs6000.c (rs6000_init_cost): Likewise.
17624 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
17626 * config/rs6000/rs6000.c (rs6000_vect_nonmem): Renamed to
17627 vect_nonmem and moved into...
17628 (struct rs6000_cost_data): ...here.
17629 (rs6000_init_cost): Use vect_nonmem of cost_data instead.
17630 (rs6000_add_stmt_cost): Likewise.
17631 (rs6000_finish_cost): Likewise.
17633 2021-05-10 Eric Botcazou <ebotcazou@adacore.com>
17635 * range-op.cc (get_bool_state): Adjust head comment.
17636 (operator_not_equal::op1_range): Fix comment.
17637 (operator_bitwise_xor::op1_range): Remove call to gcc_unreachable.
17639 2021-05-10 Martin Sebor <msebor@redhat.com>
17641 PR middle-end/100425
17642 PR middle-end/100510
17643 * gimple-ssa-warn-alloca.c (pass_walloca::firast_time_p): Rename...
17644 (pass_walloca::xlimit_certain_p): ...to this.
17645 (pass_walloca::gate): Execute for any kind of handled warning.
17646 (pass_walloca::execute): Avoid issuing "maybe" and "unbounded"
17647 warnings when xlimit_certain_p is set.
17649 2021-05-10 Pat Haugen <pthaugen@linux.ibm.com>
17651 * config/rs6000/rs6000.c (rs6000_ira_change_pseudo_allocno_class):
17652 Return ALTIVEC_REGS if that is best_class.
17653 (rs6000_compute_pressure_classes): Add ALTIVEC_REGS.
17655 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17657 * config/arm/arm.h (CPP_SPEC): Remove error message about
17660 2021-05-10 Martin Jambor <mjambor@suse.cz>
17662 * ipa-prop.h (IPA_NODE_REF): Removed.
17663 (IPA_NODE_REF_GET_CREATE): Likewise.
17664 (IPA_EDGE_REF): Likewise.
17665 (IPA_EDGE_REF_GET_CREATE): Likewise.
17666 (IS_VALID_JUMP_FUNC_INDEX): Likewise.
17667 * ipa-cp.c (print_all_lattices): Replaced IPA_NODE_REF with a direct
17668 use of ipa_node_params_sum.
17669 (ipcp_versionable_function_p): Likewise.
17670 (push_node_to_stack): Likewise.
17671 (pop_node_from_stack): Likewise.
17672 (set_single_call_flag): Replaced two IPA_NODE_REF with one single
17673 direct use of ipa_node_params_sum.
17674 (initialize_node_lattices): Replaced IPA_NODE_REF with a direct use of
17675 ipa_node_params_sum.
17676 (ipa_context_from_jfunc): Replaced IPA_EDGE_REF with a direct use of
17678 (ipcp_verify_propagated_values): Replaced IPA_NODE_REF with a direct
17679 use of ipa_node_params_sum.
17680 (self_recursively_generated_p): Likewise.
17681 (propagate_scalar_across_jump_function): Likewise.
17682 (propagate_context_across_jump_function): Replaced IPA_EDGE_REF with a
17683 direct use of ipa_edge_args_sum, moved the lookup after the early
17684 exit. Replaced IPA_NODE_REF with a direct use of ipa_node_params_sum.
17685 (propagate_bits_across_jump_function): Replaced IPA_NODE_REF with
17686 direct uses of ipa_node_params_sum.
17687 (propagate_vr_across_jump_function): Likewise.
17688 (propagate_aggregate_lattice): Likewise.
17689 (propagate_aggs_across_jump_function): Likewise.
17690 (propagate_constants_across_call): Likewise, also replaced
17691 IPA_EDGE_REF with a direct use of ipa_edge_args_sum.
17692 (good_cloning_opportunity_p): Replaced IPA_NODE_REF with a direct use
17693 of ipa_node_params_sum.
17694 (estimate_local_effects): Likewise.
17695 (add_all_node_vals_to_toposort): Likewise.
17696 (propagate_constants_topo): Likewise.
17697 (ipcp_propagate_stage): Likewise.
17698 (ipcp_discover_new_direct_edges): Likewise.
17699 (calls_same_node_or_its_all_contexts_clone_p): Likewise.
17700 (cgraph_edge_brings_value_p): Likewise (in both overloaded functions).
17701 (get_info_about_necessary_edges): Likewise.
17702 (want_remove_some_param_p): Likewise.
17703 (create_specialized_node): Likewise.
17704 (self_recursive_pass_through_p): Likewise.
17705 (self_recursive_agg_pass_through_p): Likewise.
17706 (find_more_scalar_values_for_callers_subset): Likewise and also
17707 replaced IPA_EDGE_REF with direct uses of ipa_edge_args_sum, in one
17708 case replacing two of those with a single query.
17709 (find_more_contexts_for_caller_subset): Likewise for the
17710 ipa_polymorphic_call_context overload.
17711 (intersect_aggregates_with_edge): Replaced IPA_EDGE_REF with a direct
17712 use of ipa_edge_args_sum. Replaced IPA_NODE_REF with direct uses of
17713 ipa_node_params_sum.
17714 (find_aggregate_values_for_callers_subset): Likewise, also reusing
17715 results of ipa_edge_args_sum->get.
17716 (cgraph_edge_brings_all_scalars_for_node): Replaced IPA_NODE_REF with
17717 direct uses of ipa_node_params_sum, replaced IPA_EDGE_REF with a
17718 direct use of ipa_edge_args_sum.
17719 (cgraph_edge_brings_all_agg_vals_for_node): Likewise, moved node
17720 summary query after the early exit and reused the result later.
17721 (decide_about_value): Replaced IPA_NODE_REF with a direct use of
17722 ipa_node_params_sum.
17723 (decide_whether_version_node): Likewise. Removed re-querying for
17724 summaries after cloning.
17725 (spread_undeadness): Replaced IPA_NODE_REF with a direct use of
17726 ipa_node_params_sum.
17727 (has_undead_caller_from_outside_scc_p): Likewise, reusing results of
17729 (identify_dead_nodes): Likewise.
17730 (ipcp_store_bits_results): Replaced IPA_NODE_REF with direct uses of
17731 ipa_node_params_sum.
17732 (ipcp_store_vr_results): Likewise.
17733 * ipa-fnsummary.c (evaluate_properties_for_edge): Likewise.
17734 (ipa_fn_summary_t::duplicate): Likewise.
17735 (analyze_function_body): Likewise.
17736 (estimate_calls_size_and_time): Likewise.
17737 (ipa_cached_call_context::duplicate_from): Likewise.
17738 (ipa_call_context::equal_to): Likewise.
17739 (remap_edge_params): Likewise.
17740 (ipa_merge_fn_summary_after_inlining): Likewise.
17741 (inline_read_section): Likewise.
17742 * ipa-icf.c (sem_function::param_used_p): Likewise.
17743 * ipa-modref.c (compute_parm_map): Likewise.
17744 (compute_parm_map): Replaced IPA_EDGE_REF with a direct use of
17746 (get_access_for_fnspec): Replaced IPA_NODE_REF with a direct use of
17747 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
17749 * ipa-profile.c (check_argument_count): Likewise.
17750 * ipa-prop.c (ipa_alloc_node_params): Replaced IPA_NODE_REF_GET_CREATE
17751 with a direct use of ipa_node_params_sum.
17752 (ipa_initialize_node_params): Likewise.
17753 (ipa_print_node_jump_functions_for_edge): Replaced IPA_EDGE_REF with a
17754 direct use of ipa_edge_args_sum and reused the query result.
17755 (ipa_compute_jump_functions_for_edge): Replaced IPA_NODE_REF with a
17756 direct use of ipa_node_params_sum and replaced IPA_EDGE_REF with a
17757 direct use of ipa_edge_args_sum.
17758 (ipa_note_param_call): Replaced IPA_NODE_REF with a direct use of
17759 ipa_node_params_sum and reused the result of the query.
17760 (ipa_analyze_node): Likewise.
17761 (ipa_analyze_controlled_uses): Replaced IPA_NODE_REF with a direct use
17762 of ipa_node_params_sum.
17763 (update_jump_functions_after_inlining): Replaced IPA_EDGE_REF with
17764 direct uses of ipa_edge_args_sum.
17765 (update_indirect_edges_after_inlining): Replaced IPA_NODE_REF with
17766 direct uses of ipa_node_params_sum and replaced IPA_EDGE_REF with a
17767 direct use of ipa_edge_args_sum. Removed superficial re-querying the
17769 (propagate_controlled_uses): Replaced IPA_NODE_REF with direct uses of
17770 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
17772 (ipa_propagate_indirect_call_infos): Replaced IPA_EDGE_REF with a
17773 direct use of ipa_edge_args_sum.
17774 (ipa_edge_args_sum_t::duplicate): Replaced IPA_NODE_REF with a direct
17775 use of ipa_node_params_sum.
17776 (ipa_print_node_params): Likewise.
17777 (ipa_write_node_info): Likewise and also replaced IPA_EDGE_REF with
17778 direct uses of ipa_edge_args_sum.
17779 (ipa_read_edge_info): Replaced IPA_EDGE_REF with a direct use of
17781 (ipa_read_node_info): Replaced IPA_NODE_REF with a direct use of
17782 ipa_node_params_sum.
17783 (ipa_prop_write_jump_functions): Likewise. Move variable node to the
17784 scopes where it is used.
17786 2021-05-10 Uroš Bizjak <ubizjak@gmail.com>
17788 * config/i386/i386-expand.c (ix86_expand_sse_movcc)
17789 <case E_V2SImode>: Force op_true to register.
17791 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17793 * config/arm/iterators.md (MVE_FP_COMPARISONS): New.
17794 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_f<mode>)
17795 (mve_vcmp<mve_cmp_op>q_n_f<mode>): New, merge all vcmp_*f*
17797 (mve_vcmpeqq_f<mode>, mve_vcmpeqq_n_f<mode>, mve_vcmpgeq_f<mode>)
17798 (mve_vcmpgeq_n_f<mode>, mve_vcmpgtq_f<mode>)
17799 (mve_vcmpgtq_n_f<mode>, mve_vcmpleq_f<mode>)
17800 (mve_vcmpleq_n_f<mode>, mve_vcmpltq_f<mode>)
17801 (mve_vcmpltq_n_f<mode>, mve_vcmpneq_f<mode>)
17802 (mve_vcmpneq_n_f<mode>): Remove.
17803 * config/arm/unspecs.md (VCMPEQQ_F, VCMPEQQ_N_F, VCMPGEQ_F)
17804 (VCMPGEQ_N_F, VCMPGTQ_F, VCMPGTQ_N_F, VCMPLEQ_F, VCMPLEQ_N_F)
17805 (VCMPLTQ_F, VCMPLTQ_N_F, VCMPNEQ_F, VCMPNEQ_N_F): Remove.
17807 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17809 * config/arm/iterators.md (MVE_COMPARISONS): New.
17811 (mve_cmp_type): New.
17812 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_<mode>): New, merge all
17814 (mve_vcmpneq_<mode>, mve_vcmpcsq_n_<mode>, mve_vcmpcsq_<mode>)
17815 (mve_vcmpeqq_n_<mode>, mve_vcmpeqq_<mode>, mve_vcmpgeq_n_<mode>)
17816 (mve_vcmpgeq_<mode>, mve_vcmpgtq_n_<mode>, mve_vcmpgtq_<mode>)
17817 (mve_vcmphiq_n_<mode>, mve_vcmphiq_<mode>, mve_vcmpleq_n_<mode>)
17818 (mve_vcmpleq_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
17819 (mve_vcmpneq_n_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
17820 (mve_vcmpneq_n_<mode>): Remove.
17822 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17824 * config/arm/arm_mve.h (__arm_vcmp*): Remove 's' suffix.
17825 * config/arm/arm_mve_builtins.def (vcmp*): Remove 's' suffix.
17826 * config/arm/mve.md (mve_vcmp*): Remove 's' suffix in pattern
17829 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17831 * config/arm/arm_mve_builtins.def (vcmpneq_u): Remove.
17832 (vcmpneq_n_u): Likewise.
17833 (vcmpeqq_u,): Likewise.
17834 (vcmpeqq_n_u): Likewise.
17835 * config/arm/iterators.md (supf): Remove VCMPNEQ_U, VCMPEQQ_U,
17836 VCMPEQQ_N_U and VCMPNEQ_N_U.
17837 * config/arm/mve.md (mve_vcmpneq): Remove <supf> iteration.
17838 (mve_vcmpeqq_n): Likewise.
17839 (mve_vcmpeqq): Likewise.
17840 (mve_vcmpneq_n): Likewise.
17842 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17844 * config/arm/arm_mve.h (__arm_vcmpeq*u*, __arm_vcmpne*u*): Call
17845 the 's' version of the builtin.
17847 2021-05-10 Richard Biener <rguenther@suse.de>
17849 PR tree-optimization/100492
17850 * tree-loop-distribution.c (find_seed_stmts_for_distribution):
17851 Find nothing when the loop contains an irreducible region.
17853 2021-05-10 Richard Biener <rguenther@suse.de>
17855 PR middle-end/100464
17857 * gimple-fold.c (canonicalize_constructor_val): Do not set
17860 2021-05-10 Richard Biener <rguenther@suse.de>
17862 PR tree-optimization/100434
17863 * tree-ssa-dse.c (initialize_ao_ref_for_dse): Handle
17865 (dse_optimize_stmt): Handle call LHS by dropping the
17866 LHS or the whole call if it doesn't have other
17868 (pass_dse::execute): Adjust.
17870 2021-05-10 Martin Liska <mliska@suse.cz>
17872 * Makefile.in: Add missing genversion rule.
17874 2021-05-10 Alex Coplan <alex.coplan@arm.com>
17877 * config/arm/mve.md (*mve_mov<mode>): Simplify output code. Use
17878 vldrw.u32 and vstrw.32 for V2D[IF]mode loads and stores.
17880 2021-05-10 Martin Liska <mliska@suse.cz>
17882 * builtins.c (is_builtin_name): Use startswith
17883 function instead of strncmp.
17884 * collect2.c (main): Likewise.
17885 (has_lto_section): Likewise.
17886 (scan_libraries): Likewise.
17887 * coverage.c (coverage_checksum_string): Likewise.
17888 (coverage_init): Likewise.
17889 * dwarf2out.c (is_cxx): Likewise.
17890 (gen_compile_unit_die): Likewise.
17891 * gcc-ar.c (main): Likewise.
17892 * gcc.c (init_spec): Likewise.
17893 (read_specs): Likewise.
17894 (execute): Likewise.
17895 (check_live_switch): Likewise.
17896 * genattrtab.c (write_attr_case): Likewise.
17897 (IS_ATTR_GROUP): Likewise.
17898 * gencfn-macros.c (main): Likewise.
17899 * gengtype.c (type_for_name): Likewise.
17900 (gen_rtx_next): Likewise.
17901 (get_file_langdir): Likewise.
17902 (write_local): Likewise.
17903 * genmatch.c (get_operator): Likewise.
17904 (get_operand_type): Likewise.
17905 (expr::gen_transform): Likewise.
17906 * genoutput.c (validate_optab_operands): Likewise.
17907 * incpath.c (add_sysroot_to_chain): Likewise.
17908 * langhooks.c (lang_GNU_C): Likewise.
17909 (lang_GNU_CXX): Likewise.
17910 (lang_GNU_Fortran): Likewise.
17911 (lang_GNU_OBJC): Likewise.
17912 * lto-wrapper.c (run_gcc): Likewise.
17913 * omp-general.c (omp_max_simt_vf): Likewise.
17914 * omp-low.c (omp_runtime_api_call): Likewise.
17915 * opts-common.c (parse_options_from_collect_gcc_options): Likewise.
17916 * read-rtl-function.c (function_reader::read_rtx_operand_r): Likewise.
17917 * real.c (real_from_string): Likewise.
17918 * selftest.c (assert_str_startswith): Likewise.
17919 * timevar.c (timer::validate_phases): Likewise.
17920 * tree.c (get_file_function_name): Likewise.
17921 * ubsan.c (ubsan_use_new_style_p): Likewise.
17922 * varasm.c (default_function_rodata_section): Likewise.
17923 (incorporeal_function_p): Likewise.
17924 (default_section_type_flags): Likewise.
17925 * system.h (startswith): Define startswith.
17927 2021-05-10 Martin Liska <mliska@suse.cz>
17929 * bitmap.h (class auto_bitmap): Remove
17930 __cplusplus >= 201103.
17931 * config/aarch64/aarch64.c: Likewise.
17932 * gimple-ssa-store-merging.c (store_immediate_info::store_immediate_info):
17934 * sbitmap.h: Likewise.
17936 2021-05-10 Martin Liska <mliska@suse.cz>
17938 * Makefile.in: Rename gcov-iov to genversion and depend
17939 on version.h (instead of gcov-iov.h).
17940 * gcov-io.h: Include version.h instread of gcov-iov.h.
17941 * gengtype-state.c (read_state_version): Likewise.
17942 * gcov-iov.c: Moved to...
17943 * genversion.c: ...here.
17944 * lto-streamer.h (LTO_major_version): Define it with
17946 * version.c: Removed.
17947 * version.h: Removed.
17949 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17951 * config/arc/arc.md (UNSPEC_ARC_DMPYWH): Define.
17952 * config/arc/simdext.md (VCT): Add predicates for iterator
17955 (voptab): Likewise.
17956 (vec_widen_<V_US>mult_hi_v4hi): Change pattern predicate.
17957 (<voptab>v2si3): New patterns.
17959 (reduc_plus_scal_v4hi): Likewise.
17960 (reduc_plus_scal_v2si): Likewise.
17961 (vec_duplicatev2si): Likewise.
17962 (vec_duplicatev4hi): Likewise.
17964 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17966 * config/arc/simdext.md: Format and cleanup file.
17968 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17970 * config/arc/simdext.md (movmisalignv2hi): Allow misaligned access
17971 only when munaligned-access option is on.
17972 (movmisalign<mode>): Likewise.
17974 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17976 * common/config/arc/arc-common.c (arc_handle_option): Remove dot
17978 * config/arc/arc.c (arc_reorg): Remove underscore from string.
17980 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17982 * config/arc/arc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
17983 (CTZ_DEFINED_VALUE_AT_ZERO): Likewise.
17984 * config/arc/arc.md (clrsbsi2): Cleanup pattern.
17985 (norm_f): Likewise.
17988 (clzsi2): Use fls instruction when available.
17989 (arc_clzsi2): Likewise.
17991 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17993 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add r26 and r27.
17995 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17997 * doc/extend.texi (__builtin_arc_sr): Swap arguments.
17999 2021-05-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
18001 PR middle-end/100467
18002 * toplev.c (compile_file): Call insn_locations_init before
18003 targetm.asm_out.code_end.
18005 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
18008 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
18010 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
18012 2021-05-07 Jakub Jelinek <jakub@redhat.com>
18013 Andrew Stubbs <amd@codesourcery.com>
18016 * builtins.c (try_store_by_multiple_pieces): Use force_operand for
18017 emit_move_insn operands.
18019 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
18021 * cfgexpand.c (expand_gimple_basic_block): Do not inherit a current
18022 location for the outgoing edges of an empty block.
18023 * dwarf2out.c (add_subscript_info): Retrieve the bounds and index
18024 type by means of the get_array_descr_info langhook, if it is set and
18025 returns true. Remove obsolete code dealing with unnamed subtypes.
18027 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
18029 * gimple-range-cache.cc (ssa_block_ranges): Virtualize.
18030 (sbr_vector): Renamed from ssa_block_cache.
18031 (sbr_vector::sbr_vector): Allocate from obstack abd initialize.
18032 (ssa_block_ranges::~ssa_block_ranges): Remove.
18033 (sbr_vector::set_bb_range): Use varying and undefined cached values.
18034 (ssa_block_ranges::set_bb_varying): Remove.
18035 (sbr_vector::get_bb_range): Adjust assert.
18036 (sbr_vector::bb_range_p): Adjust assert.
18037 (~block_range_cache): No freeing loop required.
18038 (block_range_cache::get_block_ranges): Remove.
18039 (block_range_cache::set_bb_range): Inline get_block_ranges.
18040 (block_range_cache::set_bb_varying): Remove.
18041 * gimple-range-cache.h (set_bb_varying): Remove prototype.
18042 * value-range.h (irange_allocator::get_memory): New.
18044 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
18046 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Search
18047 dominator tree is available and requested.
18048 (ranger_cache::ssa_range_in_bb): Don't search dom tree here.
18049 (ranger_cache::fill_block_cache): Don't search dom tree here either.
18050 * gimple-range-cache.h (non_null_deref_p): Add dom_search param.
18052 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
18054 * gimple-range.cc (gimple_ranger::range_on_exit): Handle block with
18055 only PHI nodes better.
18057 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
18059 * gimple-range-edge.h (gimple_outgoing_range): Rename from
18061 (gcond_edge_range): Export prototype.
18062 * gimple-range-edge.cc (gcond_edge_range): New.
18063 (gimple_outgoing_range::edge_range_p): Use gcond_edge_range.
18064 * gimple-range-gori.h (gori_compute): Use gimple_outgoing_range.
18066 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
18068 * gimple-range-edge.cc (outgoing_range::calc_switch_ranges): Compute
18069 default range into a temp and allocate only what is needed.
18071 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
18073 * range-op.cc (operator_trunc_mod::wi_fold): x % 0 is UNDEFINED.
18075 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
18077 * gimple-range.h (gimple_range_global): Pick up parameter initial
18078 values, and use-before defined locals are UNDEFINED.
18080 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
18082 * doc/extend.texi (scalar_storage_order): Mention effect on pointer
18084 * tree.h (reverse_storage_order_for_component_p): Return false if
18085 the type is a pointer.
18087 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
18089 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
18091 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
18094 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
18095 Handle V8QI, V4HI and V2SI modes.
18096 * config/i386/mmx.md (mmx_pblendvb): New insn pattern.
18097 * config/i386/sse.md (unspec): Move UNSPEC_BLENDV ...
18098 * config/i386/i386.md (unspec): ... here.
18100 2021-05-07 Tobias Burnus <tobias@codesourcery.com>
18101 Tom de Vries <tdevries@suse.de>
18103 * omp-low.c (lower_rec_simd_input_clauses): Set max_vf = 1 if
18104 a truth_value_p reduction variable is nonintegral.
18106 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
18109 * config/i386/i386-expand.c (ix86_use_mask_cmp_p):
18110 Return false for mode sizes < 16.
18112 2021-05-07 Jakub Jelinek <jakub@redhat.com>
18115 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn.
18117 2021-05-06 Martin Jambor <mjambor@suse.cz>
18119 * ipa-sra.c (ipa_sra_dump_all_summaries): Dump edge summaries even
18120 when there is no function summary.
18121 (ipa_sra_summarize_function): produce edge summaries even when
18124 2021-05-06 Tom Tromey <tom@tromey.com>
18126 * godump.c (string_hash_eq): Remove.
18127 (go_finish): Use htab_eq_string.
18129 2021-05-06 Tom Tromey <tom@tromey.com>
18131 * gengtype-state.c (read_state): Use htab_eq_string.
18132 (string_eq): Remove.
18134 2021-05-06 Tom Tromey <tom@tromey.com>
18136 * gensupport.c (htab_eq_string): Remove.
18138 2021-05-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
18141 * debug.h (gcc_debug_hooks): Add set_ignored_loc function pointer.
18142 * dwarf2out.h (dw_fde_node::ignored_debug): New data item.
18143 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Add dummy
18144 set_ignored_loc callbacks.
18145 * debug.c (do_nothing_debug_hooks): Likewise.
18146 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
18147 * dwarf2out.c (text_section_used, cold_text_section_used): Remove.
18148 (in_text_section_p, last_text_label, last_cold_label,
18149 switch_text_ranges, switch_cold_ranges): New data items.
18150 (dwarf2out_note_section_used): Remove.
18151 (dwarf2out_begin_prologue): Set fde->ignored_debug and
18153 (mark_ignored_debug_section): New helper function.
18154 (dwarf2out_end_epilogue, dwarf2out_switch_text_section): Call
18155 mark_ignored_debug_section.
18156 (dwarf2_debug_hooks): Use dwarf2out_set_ignored_loc.
18157 (dwarf2_lineno_debug_hooks): Use dummy for set_ignored_loc.
18158 (size_of_aranges): Adjust formula for multi-part text ranges size.
18159 (output_aranges): Output multi-part text ranges.
18160 (dwarf2out_set_ignored_loc): New callback function.
18161 (dwarf2out_finish): Output multi-part text ranges.
18162 (dwarf2out_c_finalize): Clear new data items.
18163 * final.c (final_start_function_1): Call set_ignored_loc callback.
18164 (final_scan_insn_1): Likewise.
18165 * ggc-page.c (gt_ggc_mx): New helper function.
18166 * stringpool.c (gt_pch_nx): Likewise.
18168 2021-05-06 Richard Biener <rguenther@suse.de>
18170 * timevar.def (TV_TREE_INSERT_PHI_NODES): Remove.
18171 (TV_TREE_SSA_REWRITE_BLOCKS): Likewise.
18172 (TV_TREE_INTO_SSA): New.
18173 * tree-into-ssa.c (insert_phi_nodes): Do not account separately.
18174 (rewrite_blocks): Likewise.
18175 (pass_data_build_ssa): Account to TV_TREE_INTO_SSA.
18177 2021-05-06 Jakub Jelinek <jakub@redhat.com>
18179 * tree-ssa-phiopt.c (value_replacement, minmax_replacement,
18180 abs_replacement, xor_replacement,
18181 cond_removal_in_popcount_clz_ctz_pattern,
18182 replace_phi_edge_with_variable): Change type of phi argument from
18183 gimple * to gphi *.
18185 2021-05-06 Richard Biener <rguenther@suse.de>
18187 * tree-ssa-loop-split.c (split_loop): Delay updating SSA form.
18188 Output an opt-info message.
18189 (do_split_loop_on_cond): Likewise.
18190 (tree_ssa_split_loops): Update SSA form here.
18192 2021-05-06 Richard Biener <rguenther@suse.de>
18194 * tree-inline.c (tree_function_versioning): Fix DECL_BY_REFERENCE
18195 return variable removal.
18197 2021-05-06 Marius Hillenbrand <mhillen@linux.ibm.com>
18199 * config/s390/s390-builtins.def (O_M5, O1_M5, ...): Remove unused macros.
18200 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
18201 (s390_vec_permi_dbl, s390_vpdi): Use the O3_U2 type for the immediate
18203 * config/s390/s390.c (s390_const_operand_ok): Remove unused
18206 2021-05-06 Jakub Jelinek <jakub@redhat.com>
18208 PR tree-optimization/94589
18209 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Call
18210 spaceship_replacement.
18211 (cond_only_block_p, spaceship_replacement): New functions.
18213 2021-05-06 Richard Biener <rguenther@suse.de>
18216 * tree-emutls.c (gen_emutls_addr): Pass in whether we're
18217 dealing with a debug use and only query existing addresses
18219 (lower_emutls_1): Avoid splitting out addresses for debug
18220 stmts, reset the debug stmt when we fail to find existing
18222 (lower_emutls_phi_arg): Set wi.stmt.
18224 2021-05-06 Christoph Muellner <cmuellner@gcc.gnu.org>
18227 * config/riscv/riscv.c (riscv_block_move_loop): Use cbranch helper.
18228 * config/riscv/riscv.md (cbranch<mode>4): Generate helpers.
18229 (stack_protect_test): Use cbranch helper.
18231 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
18234 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
18235 always return the establisher frame for __builtin_frame_address (0).
18237 2021-05-05 Ivan Sorokin <vanyacpp@gmail.com>
18240 * config/i386/i386-builtins.c (ix86_cpu_model_type_node): New.
18241 (ix86_cpu_model_var): Likewise.
18242 (ix86_cpu_features2_type_node): Likewise.
18243 (ix86_cpu_features2_var): Likewise.
18244 (fold_builtin_cpu): Cache __cpu_model and __cpu_features2 with
18247 2021-05-05 Martin Sebor <msebor@redhat.com>
18249 * passes.def (pass_warn_printf): Run after SSA.
18251 2021-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18253 * config/arm/neon.md (neon_vtst_combine<mode>): New pattern.
18254 * config/arm/predicates.md (minus_one_operand): New predicate.
18256 2021-05-05 Jeff Law <jlaw@tachyum.com>
18258 * config/avr/avr.md: Remove references to CC_STATUS_INIT.
18260 2021-05-05 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
18262 PR rtl-optimization/100263
18263 * postreload.c (move2add_valid_value_p): Ensure register can
18266 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
18268 PR rtl-optimization/100411
18269 * cfgcleanup.c (try_crossjump_to_edge): Also skip end of prologue
18270 and beginning of function markers.
18272 2021-05-05 Jeff Law <jlaw@tachyum.com>
18274 * config/cr16/cr16.h (NOTICE_UPDATE_CC): Remove.
18275 * config/cr16/cr16.c (notice_update_cc): Remove.
18276 * config/cr16/cr16-protos.h (notice_update_cc): Remove.
18278 2021-05-05 Uroš Bizjak <ubizjak@gmail.com>
18281 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
18282 Handle V8QI, V4HI and V2SI modes.
18283 * config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
18284 (ix86_build_signbit_mask): Ditto.
18285 * config/i386/mmx.md (MMXMODE14): New mode iterator.
18286 (<smaxmin:code><MMXMODE14:mode>3): New expander.
18287 (*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
18288 (<umaxmin:code><MMXMODE24:mode>3): New expander.
18289 (*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
18290 (vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
18291 (vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
18292 (vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
18293 (vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
18294 (vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
18296 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
18298 * dwarf2out.c (loc_list_from_tree_1) <DECL>: During early DWARF, do
18299 not expand the VALUE_EXPR of variables put in the non-local frame.
18300 * gimplify.c (gimplify_type_sizes) <RECORD_TYPE>: If the type is not
18301 to be ignored for debug info, ensure its variable offsets are not.
18303 2021-05-05 Richard Biener <rguenther@suse.de>
18305 PR tree-optimization/79333
18306 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
18307 Fold stmt following SSA edges.
18309 2021-05-05 Richard Biener <rguenther@suse.de>
18311 PR middle-end/100394
18312 * calls.c (expand_call): Preserve possibly throwing calls.
18313 * cfgexpand.c (expand_call_stmt): When a call can throw signal
18314 RTL expansion there are side-effects.
18315 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Simplify,
18316 mark all possibly throwing stmts necessary unless we can elide
18318 * tree-ssa-dse.c (pass_dse::execute): Preserve exceptions unless
18319 -fdelete-dead-exceptions.
18320 * tree.h (DECL_PURE_P): Add note about exceptions.
18322 2021-05-05 Alexandre Oliva <oliva@adacore.com>
18324 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Make it
18327 2021-05-04 David Edelsohn <dje.gcc@gmail.com>
18329 * config/rs6000/rs6000-call.c (rs6000_output_mi_thunk): Use
18330 get_fnname_from_decl for name of thunk.
18331 * config/rs6000/rs6000.c (rs6000_declare_alias): Use assemble_name
18332 and ASM_OUTPUT_LABEL.
18333 (rs6000_xcoff_declare_function_name): Use assemble_name and
18335 (rs6000_xcoff_declare_object_name): Use ASM_OUTPUT_LABEL.
18336 (rs6000_xcoff_encode_section_info): Don't add mapping class
18337 for aliases. Always add [DS] mapping class to primary
18339 (rs6000_asm_weaken_decl): Don't explicitly add [DS].
18341 2021-05-04 Martin Sebor <msebor@redhat.com>
18343 PR middle-end/100307
18344 * builtins.c (compute_objsize_r): Clear base0 for pointers.
18346 2021-05-04 Jeff Law <jlaw@tachyum.com>
18348 * config/bfin/bfin.h (NOTICE_UPDATE_CC): Remove.
18350 2021-05-04 Segher Boessenkool <segher@kernel.crashing.org>
18352 * caller-save.c: Remove CC0.
18353 * cfgcleanup.c: Remove CC0.
18354 * cfgrtl.c: Remove CC0.
18355 * combine.c: Remove CC0.
18356 * compare-elim.c: Remove CC0.
18357 * conditions.h: Remove CC0.
18358 * config/h8300/h8300.h: Remove CC0.
18359 * config/h8300/h8300-protos.h: Remove CC0.
18360 * config/h8300/peepholes.md: Remove CC0.
18361 * config/i386/x86-tune-sched.c: Remove CC0.
18362 * config/m68k/m68k.c: Remove CC0.
18363 * config/rl78/rl78.c: Remove CC0.
18364 * config/sparc/sparc.c: Remove CC0.
18365 * config/xtensa/xtensa.c: Remove CC0.
18366 (gen_conditional_move): Use pc_rtx instead of cc0_rtx in a piece of
18367 RTL where that is used as a placeholder only.
18368 * cprop.c: Remove CC0.
18369 * cse.c: Remove CC0.
18370 * cselib.c: Remove CC0.
18371 * df-problems.c: Remove CC0.
18372 * df-scan.c: Remove CC0.
18373 * doc/md.texi: Remove CC0. Adjust an example.
18374 * doc/rtl.texi: Remove CC0. Adjust an example.
18375 * doc/tm.texi: Regenerate.
18376 * doc/tm.texi.in: Remove CC0.
18377 * emit-rtl.c: Remove CC0.
18378 * final.c: Remove CC0.
18379 * fwprop.c: Remove CC0.
18380 * gcse-common.c: Remove CC0.
18381 * gcse.c: Remove CC0.
18382 * genattrtab.c: Remove CC0.
18383 * genconfig.c: Remove CC0.
18384 * genemit.c: Remove CC0.
18385 * genextract.c: Remove CC0.
18386 * gengenrtl.c: Remove CC0.
18387 * genrecog.c: Remove CC0.
18388 * haifa-sched.c: Remove CC0.
18389 * ifcvt.c: Remove CC0.
18390 * ira-costs.c: Remove CC0.
18391 * ira.c: Remove CC0.
18392 * jump.c: Remove CC0.
18393 * loop-invariant.c: Remove CC0.
18394 * lra-constraints.c: Remove CC0.
18395 * lra-eliminations.c: Remove CC0.
18396 * optabs.c: Remove CC0.
18397 * postreload-gcse.c: Remove CC0.
18398 * postreload.c: Remove CC0.
18399 * print-rtl.c: Remove CC0.
18400 * read-rtl-function.c: Remove CC0.
18401 * reg-notes.def: Remove CC0.
18402 * reg-stack.c: Remove CC0.
18403 * reginfo.c: Remove CC0.
18404 * regrename.c: Remove CC0.
18405 * reload.c: Remove CC0.
18406 * reload1.c: Remove CC0.
18407 * reorg.c: Remove CC0.
18408 * resource.c: Remove CC0.
18409 * rtl.c: Remove CC0.
18410 * rtl.def: Remove CC0.
18411 * rtl.h: Remove CC0.
18412 * rtlanal.c: Remove CC0.
18413 * sched-deps.c: Remove CC0.
18414 * sched-rgn.c: Remove CC0.
18415 * shrink-wrap.c: Remove CC0.
18416 * simplify-rtx.c: Remove CC0.
18417 * system.h: Remove CC0. Poison NOTICE_UPDATE_CC, CC_STATUS_MDEP_INIT,
18418 CC_STATUS_MDEP, and CC_STATUS.
18419 * target.def: Remove CC0.
18420 * valtrack.c: Remove CC0.
18421 * var-tracking.c: Remove CC0.
18423 2021-05-04 Richard Biener <rguenther@suse.de>
18425 PR tree-optimization/100414
18426 * tree-ssa-phiopt.c (get_non_trapping): Do not compute dominance
18428 (tree_ssa_phiopt_worker): But unconditionally here.
18430 2021-05-04 Tobias Burnus <tobias@codesourcery.com>
18432 * omp-low.c (lower_rec_input_clauses, lower_reduction_clauses): Handle
18433 && and || with floating-point and complex arguments.
18435 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
18437 * tree-inline.c (insert_debug_decl_map): Delete.
18438 (copy_debug_stmt): Minor tweak.
18439 (setup_one_parameter): Do not use a variable if the value is either
18440 a read-only DECL or a non-addressable local variable in the caller.
18441 In this case, insert the debug-only variable in the map manually.
18442 (expand_call_inline): Do not generate a CLOBBER for these values.
18443 * tree-inline.h (debug_map): Minor tweak.
18445 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
18447 * builtins.c (builtin_with_linkage_p): Return true for stp[n]cpy.
18448 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Tidy up.
18450 2021-05-04 Richard Biener <rguenther@suse.de>
18452 PR tree-optimization/100329
18453 * tree-ssa-reassoc.c (can_reassociate_p): Do not reassociate
18455 (insert_stmt_after): Assert we're not running into asm goto.
18457 2021-05-04 Richard Biener <rguenther@suse.de>
18459 PR tree-optimization/100398
18460 * tree-ssa-dse.c (pass_dse::execute): Preserve control
18463 2021-05-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18465 * builtins.c (try_store_by_multiple_pieces): Fix constfun's prototype.
18467 2021-05-04 Alexandre Oliva <oliva@adacore.com>
18469 * builtins.c (try_store_by_multiple_pieces): New.
18470 (expand_builtin_memset_args): Use it. If target_char_cast
18471 fails, proceed as for non-constant val. Pass len's ctz to...
18472 * expr.c (clear_storage_hints): ... this. Try store by
18473 multiple pieces after setmem.
18474 (clear_storage): Adjust.
18475 * expr.h (clear_storage_hints): Likewise.
18476 (try_store_by_multiple_pieces): Declare.
18477 * passes.def: Replace the last copy_prop with ccp.
18479 2021-05-03 Tom de Vries <tdevries@suse.de>
18482 * omp-low.c (lower_rec_input_clauses): Disable SIMT for user-defined
18485 2021-05-03 Richard Biener <rguenther@suse.de>
18487 * tree-ssa-dse.c (dse_classify_store): Track two PHI defs.
18489 2021-05-03 Richard Biener <rguenther@suse.de>
18491 * tree-ssa-dse.c: Do not include domwalk.h but cfganal.h.
18492 (dse_dom_walker): Remove.
18493 (dse_dom_walker::dse_optimize_stmt): Rename...
18494 (dse_optimize_stmt): ... to this, pass in live_bytes sbitmap.
18495 (dse_dom_walker::before_dom_children): Inline ...
18496 (pass_dse::execute): ... here. Perform a reverse program
18499 2021-05-03 H.J. Lu <hjl.tools@gmail.com>
18502 * configure: Regenerated.
18504 2021-05-03 Ilya Leoshkevich <iii@linux.ibm.com>
18507 * config/s390/s390.c (s390_hard_fp_reg_p): New function.
18508 (s390_md_asm_adjust): Handle hard registers.
18510 2021-05-03 Jakub Jelinek <jakub@redhat.com>
18512 PR tree-optimization/100382
18513 * tree-ssa-dse.c: Include tree-eh.h.
18514 (dse_dom_walker::before_dom_children): Don't remove stmts if
18515 stmt_unremovable_because_of_non_call_eh_p is true.
18517 2021-05-02 David Edelsohn <dje.gcc@gmail.com>
18519 * varasm.c (compute_reloc_for_var): Split out from...
18520 (get_variable_section): Use it.
18521 * output.h (compute_reloc_for_var): Declare.
18522 * config/rs6000/rs6000-protos.h
18523 (rs6000_xcoff_asm_output_aligned_decl_common): Change alignment to
18525 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address_aix):
18526 Don't append storage mapping class to symbol.
18527 (rs6000_xcoff_asm_named_section): Add BS and UL mapping classes.
18528 Don't convert TLS BSS to common.
18529 (rs6000_xcoff_unique_section): Don't fall back to select_secton.
18530 (rs6000_xcoff_section_type_flags): Add SECTION_BSS if DECL is
18532 (rs6000_xcoff_asm_globalize_decl_name): Don't strip storage
18534 (rs6000_xcoff_asm_output_aligned_decl_common): Align is unsigned int.
18535 If align is 0 from TLS class, use the same rules as varasm.c
18536 If not common, switch to BSS section manually.
18537 If common, emit appropriate comm or lcomm directive.
18538 (rs6000_xcoff_encode_section_info): Add logic to append all
18539 storage mapping classes.
18540 (rs6000_asm_weaken_decl): Adjust for qualname symbols.
18541 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
18542 rs6000_xcoff_asm_output_aligned_decl_common.
18543 (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
18544 rs6000_xcoff_asm_output_aligned_decl_common.
18545 (ASM_OUTPUT_TLS_COMMON): Use
18546 rs6000_xcoff_asm_output_aligned_decl_common.
18548 2021-05-02 Jakub Jelinek <jakub@redhat.com>
18551 * config/nvptx/nvptx.c (nvptx_sese_pseudo): Use nullptr instead of 0
18552 as first argument of pseudo_node_t constructors.
18554 2021-05-02 Jakub Jelinek <jakub@redhat.com>
18557 * config/i386/t-i386 (TM_H): Add $(srcdir)/config/i386/i386-isa.def.
18559 2021-05-01 Aldy Hernandez <aldyh@redhat.com>
18561 * value-range.cc (DEFINE_INT_RANGE_GC_STUBS): Remove.
18562 (gt_pch_nx (int_range<1> *&)): New.
18563 (gt_ggc_mx (int_range<1> *&)): New.
18564 * value-range.h (class irange): Add GTY support for
18567 2021-05-01 Geng Qi <gengqi@linux.alibaba.com>
18569 * doc/options.texi (Negative): Change either or to both and.
18571 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
18573 * config/aarch64/aarch64-simd-builtins.def: Add
18574 float_ml[as][q]_laneq builtin generator macros.
18575 * config/aarch64/aarch64-simd.md (mul_laneq<mode>3): Define.
18576 (aarch64_float_mla_laneq<mode>): Define.
18577 (aarch64_float_mls_laneq<mode>): Define.
18578 * config/aarch64/arm_neon.h (vmla_laneq_f32): Use RTL builtin
18579 instead of GCC vector extensions.
18580 (vmlaq_laneq_f32): Likewise.
18581 (vmls_laneq_f32): Likewise.
18582 (vmlsq_laneq_f32): Likewise.
18584 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
18586 * config/aarch64/aarch64-simd-builtins.def: Add
18587 float_ml[as]_lane builtin generator macros.
18588 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt<mode>):
18590 (mul_lane<mode>3): This, and re-order arguments.
18591 (aarch64_float_mla_lane<mode>): Define.
18592 (aarch64_float_mls_lane<mode>): Define.
18593 * config/aarch64/arm_neon.h (vmla_lane_f32): Use RTL builtin
18594 instead of GCC vector extensions.
18595 (vmlaq_lane_f32): Likewise.
18596 (vmls_lane_f32): Likewise.
18597 (vmlsq_lane_f32): Likewise.
18599 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
18601 * config/aarch64/aarch64-simd-builtins.def: Add float_ml[as]
18602 builtin generator macros.
18603 * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>):
18605 (aarch64_float_mls<mode>): Define.
18606 * config/aarch64/arm_neon.h (vmla_f32): Use RTL builtin
18607 instead of relying on GCC vector extensions.
18608 (vmla_f64): Likewise.
18609 (vmlaq_f32): Likewise.
18610 (vmlaq_f64): Likewise.
18611 (vmls_f32): Likewise.
18612 (vmls_f64): Likewise.
18613 (vmlsq_f32): Likewise.
18614 (vmlsq_f64): Likewise.
18615 * config/aarch64/iterators.md: Define VDQF_DF mode iterator.
18617 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
18619 * config/aarch64/aarch64-simd-builtins.def: Add
18620 float_ml[as]_n_builtin generator macros.
18621 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt_from_dup<mode>):
18623 (mul_n<mode>3): This, and re-order arguments.
18624 (aarch64_float_mla_n<mode>): Define.
18625 (aarch64_float_mls_n<mode>): Define.
18626 * config/aarch64/arm_neon.h (vmla_n_f32): Use RTL builtin
18627 instead of inline asm.
18628 (vmlaq_n_f32): Likewise.
18629 (vmls_n_f32): Likewise.
18630 (vmlsq_n_f32): Likewise.
18632 2021-04-30 Jonathan Wright <joanthan.wright@arm.com>
18634 * config/aarch64/aarch64-simd-builtins.def: Add pmull[2]
18635 builtin generator macros.
18636 * config/aarch64/aarch64-simd.md (aarch64_pmullv8qi): Define.
18637 (aarch64_pmull_hiv16qi_insn): Define.
18638 (aarch64_pmull_hiv16qi): Define.
18639 * config/aarch64/arm_neon.h (vmull_high_p8): Use RTL builtin
18640 instead of inline asm.
18641 (vmull_p8): Likewise.
18643 2021-04-30 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
18645 * config/avr/avr.md: Adjust peepholes to match and
18646 generate parallels with clobber of REG_CC.
18647 (mov<mode>_insn): Rename to mov<mode>_insn_split.
18648 (*mov<mode>_insn): Rename to mov<mode>_insn.
18650 2021-04-30 David Edelsohn <dje.gcc@gmail.com>
18652 * varasm.c (use_blocks_for_decl_p): Don't use section anchors
18653 for VAR_DECLs if -fdata-sections enabled.
18655 2021-04-30 Michael Meissner <meissner@linux.ibm.com>
18657 PR bootstrap/100327
18658 * config/rs6000/rs6000.c
18659 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Define.
18660 (rs6000_libgcc_floating_mode_supported_p): New target hook.
18662 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
18664 * tree-ssa-threadbackward.c (class thread_jumps): Split out code
18666 (class back_threader_registry): ...to here...
18667 (class back_threader_profitability): ...and here...
18668 (thread_jumps::thread_through_all_blocks): Remove argument.
18669 (back_threader_registry::back_threader_registry): New.
18670 (back_threader_registry::~back_threader_registry): New.
18671 (back_threader_registry::thread_through_all_blocks): New.
18672 (thread_jumps::profitable_jump_thread_path): Move from here...
18673 (back_threader_profitability::profitable_path_p): ...to here.
18674 (thread_jumps::find_taken_edge): New.
18675 (thread_jumps::convert_and_register_current_path): Move...
18676 (back_threader_registry::register_path): ...to here.
18677 (thread_jumps::register_jump_thread_path_if_profitable): Move...
18678 (thread_jumps::maybe_register_path): ...to here.
18679 (thread_jumps::handle_phi): Call find_taken_edge and
18680 maybe_register_path.
18681 (thread_jumps::handle_assignment): Same.
18682 (thread_jumps::fsm_find_control_statement_thread_paths): Remove
18683 tree argument to handle_phi and handle_assignment.
18684 (thread_jumps::find_jump_threads_backwards): Set m_name. Remove
18685 set of m_speed_p and m_max_threaded_paths.
18686 (pass_thread_jumps::execute): Remove second argument from
18687 find_jump_threads_backwards.
18688 (pass_early_thread_jumps::execute): Same.
18690 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
18692 * tree-ssa-dom.c (class dom_jump_threader_simplifier): New.
18693 (class dom_opt_dom_walker): Initialize some class variables.
18694 (pass_dominator::execute): Pass evrp_range_analyzer and
18695 dom_jump_threader_simplifier to dom_opt_dom_walker.
18696 Adjust for some functions moving into classes.
18697 (simplify_stmt_for_jump_threading): Adjust and move to...
18698 (jump_threader_simplifier::simplify): ...here.
18699 (dom_opt_dom_walker::before_dom_children): Adjust for
18700 m_evrp_range_analyzer.
18701 (dom_opt_dom_walker::after_dom_children): Remove x_vr_values hack.
18702 (test_for_singularity): Place in dom_opt_dom_walker class.
18703 (dom_opt_dom_walker::optimize_stmt): The argument
18704 evrp_range_analyzer is now a class field.
18705 * tree-ssa-threadbackward.c (class thread_jumps): Add m_registry.
18706 (thread_jumps::thread_through_all_blocks): New.
18707 (thread_jumps::convert_and_register_current_path): Use m_registry.
18708 (pass_thread_jumps::execute): Adjust for thread_through_all_blocks
18709 being in the threader class.
18710 (pass_early_thread_jumps::execute): Same.
18711 * tree-ssa-threadedge.c (threadedge_initialize_values): Move...
18712 (jump_threader::jump_threader): ...here.
18713 (threadedge_finalize_values): Move...
18714 (jump_threader::~jump_threader): ...here.
18715 (jump_threader::remove_jump_threads_including): New.
18716 (jump_threader::thread_through_all_blocks): New.
18717 (record_temporary_equivalences_from_phis): Move...
18718 (jump_threader::record_temporary_equivalences_from_phis): ...here.
18719 (record_temporary_equivalences_from_stmts_at_dest): Move...
18720 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
18722 (simplify_control_stmt_condition_1): Move to jump_threader class.
18723 (simplify_control_stmt_condition): Move...
18724 (jump_threader::simplify_control_stmt_condition): ...here.
18725 (thread_around_empty_blocks): Move...
18726 (jump_threader::thread_around_empty_blocks): ...here.
18727 (thread_through_normal_block): Move...
18728 (jump_threader::thread_through_normal_block): ...here.
18729 (thread_across_edge): Move...
18730 (jump_threader::thread_across_edge): ...here.
18731 (thread_outgoing_edges): Move...
18732 (jump_threader::thread_outgoing_edges): ...here.
18733 * tree-ssa-threadedge.h: Move externally facing functings...
18734 (class jump_threader): ...here...
18735 (class jump_threader_simplifier): ...and here.
18736 * tree-ssa-threadupdate.c (struct redirection_data): Remove comment.
18737 (jump_thread_path_allocator::jump_thread_path_allocator): New.
18738 (jump_thread_path_allocator::~jump_thread_path_allocator): New.
18739 (jump_thread_path_allocator::allocate_thread_edge): New.
18740 (jump_thread_path_allocator::allocate_thread_path): New.
18741 (jump_thread_path_registry::jump_thread_path_registry): New.
18742 (jump_thread_path_registry::~jump_thread_path_registry): New.
18743 (jump_thread_path_registry::allocate_thread_edge): New.
18744 (jump_thread_path_registry::allocate_thread_path): New.
18745 (dump_jump_thread_path): Make extern.
18746 (debug (const vec<jump_thread_edge *> &path)): New.
18747 (struct removed_edges): Move to tree-ssa-threadupdate.h.
18748 (struct thread_stats_d): Remove.
18749 (remove_ctrl_stmt_and_useless_edges): Make static.
18750 (lookup_redirection_data): Move...
18751 (jump_thread_path_registry::lookup_redirection_data): ...here.
18752 (ssa_redirect_edges): Make static.
18753 (thread_block_1): Move...
18754 (jump_thread_path_registry::thread_block_1): ...here.
18755 (thread_block): Move...
18756 (jump_thread_path_registry::thread_block): ...here.
18757 (thread_through_loop_header): Move...
18758 (jump_thread_path_registry::thread_through_loop_header): ...here.
18759 (mark_threaded_blocks): Move...
18760 (jump_thread_path_registry::mark_threaded_blocks): ...here.
18761 (debug_path): Move...
18762 (jump_thread_path_registry::debug_path): ...here.
18763 (debug_all_paths): Move...
18764 (jump_thread_path_registry::dump): ..here.
18765 (rewire_first_differing_edge): Move...
18766 (jump_thread_path_registry::rewire_first_differing_edge): ...here.
18767 (adjust_paths_after_duplication): Move...
18768 (jump_thread_path_registry::adjust_paths_after_duplication): ...here.
18769 (duplicate_thread_path): Move...
18770 (jump_thread_path_registry::duplicate_thread_path): ..here.
18771 (remove_jump_threads_including): Move...
18772 (jump_thread_path_registry::remove_jump_threads_including): ...here.
18773 (thread_through_all_blocks): Move to...
18774 (jump_thread_path_registry::thread_through_all_blocks): ...here.
18775 (delete_jump_thread_path): Remove.
18776 (register_jump_thread): Move...
18777 (jump_thread_path_registry::register_jump_thread): ...here.
18778 * tree-ssa-threadupdate.h: Move externally facing functions...
18779 (class jump_thread_path_allocator): ...here...
18780 (class jump_thread_path_registry): ...and here.
18781 (thread_through_all_blocks): Remove.
18782 (struct removed_edges): New.
18783 (register_jump_thread): Remove.
18784 (remove_jump_threads_including): Remove.
18785 (delete_jump_thread_path): Remove.
18786 (remove_ctrl_stmt_and_useless_edges): Remove.
18787 (free_dom_edge_info): New prototype.
18788 * tree-vrp.c: Remove x_vr_values hack.
18789 (class vrp_jump_threader_simplifier): New.
18790 (vrp_jump_threader_simplifier::simplify): New.
18791 (vrp_jump_threader::vrp_jump_threader): Adjust method signature.
18792 Remove m_dummy_cond.
18793 Instantiate m_simplifier and m_threader.
18794 (vrp_jump_threader::thread_through_all_blocks): New.
18795 (vrp_jump_threader::simplify_stmt): Remove.
18796 (vrp_jump_threader::after_dom_children): Do not set m_dummy_cond.
18797 Remove x_vr_values hack.
18798 (execute_vrp): Adjust for thread_through_all_blocks being in a
18801 2021-04-30 Christophe Lyon <christophe.lyon@linaro.org>
18803 * genflags.c (gen_insn): Print failed expansion string.
18805 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
18807 * expr.c (alignment_for_piecewise_move): Call mode_for_size
18808 without limit to MAX_FIXED_MODE_SIZE.
18810 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
18812 PR middle-end/90773
18813 * builtins.c (builtin_memset_gen_str): Don't use return from
18814 simplify_gen_subreg.
18816 2021-04-30 Uroš Bizjak <ubizjak@gmail.com>
18819 * config/i386/i386.md (*add<mode>3_carry_0r): New insn pattern.
18820 (*addsi3_carry_zext_0r): Ditto.
18821 (*sub<mode>3_carry_0): Ditto.
18822 (*subsi3_carry_zext_0r): Ditto.
18823 * config/i386/predicates.md (ix86_carry_flag_unset_operator):
18825 * config/i386/i386.c (ix86_rtx_costs) <case PLUS, case MINUS>:
18826 Also consider ix86_carry_flag_unset_operator to calculate
18827 the cost of adc/sbb insn.
18829 2021-04-30 Roman Zhuykov <zhroma@ispras.ru>
18831 PR rtl-optimization/100225
18832 PR rtl-optimization/84878
18833 * modulo-sched.c (sms_schedule): Use note_stores to skip loops
18834 where we have an instruction which touches (writes) any hard
18835 register from df->regular_block_artificial_uses set.
18836 Allow not-single-set instruction only right before basic block
18839 2021-04-30 Geng Qi <gengqi@linux.alibaba.com>
18841 * config/riscv/riscv.opt (march=,mabi=): Negative itself.
18843 2021-04-30 LevyHsu <admin@levyhsu.com>
18845 * config/riscv/riscv.c (riscv_min_arithmetic_precision): New.
18846 * config/riscv/riscv.h (TARGET_MIN_ARITHMETIC_PRECISION): New.
18847 * config/riscv/riscv.md (addv<mode>4, uaddv<mode>4): New.
18848 (subv<mode>4, usubv<mode>4, mulv<mode>4, umulv<mode>4): New.
18850 2021-04-29 Alexandre Oliva <oliva@adacore.com>
18852 * config.gcc: Merged x86 and x86_64 cpu_type-setting cases.
18854 2021-04-29 Alexandre Oliva <oliva@adacore.com>
18856 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Rename to...
18857 (ASM_OUTPUT_MAX_SKIP_ALIGN): ... this. Enclose in do/while(0).
18858 * config/i386/i386.c: Adjust.
18859 * config/i386/i386.md: Adjust.
18860 * config/i386/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Drop.
18861 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18862 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18863 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18864 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18865 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18866 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18867 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18868 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18869 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18870 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
18872 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
18874 * config/i386/i386-expand.c (ix86_expand_int_compare):
18875 Swap operands of GTU and LEU comparison to emit carry flag comparison.
18876 * config/i386/i386.md (*add<mode>3_carry_0): Change insn
18877 predicate to allow more combine opportunities with memory operands.
18878 (*sub<mode>3_carry_0): Ditto.
18880 2021-04-29 Richard Sandiford <richard.sandiford@arm.com>
18882 PR rtl-optimization/100303
18883 * rtl-ssa/accesses.cc (function_info::make_use_available): Take a
18884 boolean that indicates whether the use will only be used in
18885 debug instructions. Treat it in the same way that existing
18886 cross-EBB debug references would be handled if so.
18887 (function_info::make_uses_available): Likewise.
18888 * rtl-ssa/functions.h (function_info::make_uses_available): Update
18889 prototype accordingly.
18890 (function_info::make_uses_available): Likewise.
18891 * fwprop.c (try_fwprop_subst): Update call accordingly.
18893 2021-04-29 Jeff Law <jlaw@tachyum.com>
18895 * config/nios2/nios2-protos.h (nios2_fpu_insn_enabled): Move outside
18898 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
18899 Richard Biener <rguenther@suse.de>
18902 * config/i386/i386-builtin.def (IX86_BUILTIN_MASKLOADPD)
18903 (IX86_BUILTIN_MASKLOADPS, IX86_BUILTIN_MASKLOADPD256)
18904 (IX86_BUILTIN_MASKLOADPS256, IX86_BUILTIN_MASKLOADD)
18905 (IX86_BUILTIN_MASKLOADQ, IX86_BUILTIN_MASKLOADD256)
18906 (IX86_BUILTIN_MASKLOADQ256): Move from SPECIAL_ARGS
18907 to PURE_ARGS category.
18908 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
18909 Handle PURE_ARGS category.
18910 * config/i386/i386-expand.c (ix86_expand_builtin): Ditto.
18912 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
18914 * configure.ac: Check for the presence of sys/locking.h header and
18915 for whether _LK_LOCK is supported by _locking.
18916 * configure: Regenerate.
18917 * config.in: Likewise.
18918 * gcov-io.h: Define GCOV_LOCKED_WITH_LOCKING if HOST_HAS_LK_LOCK.
18919 * gcov-io.c (gcov_open): Add support for GCOV_LOCKED_WITH_LOCKING.
18920 * system.h: Include <sys/locking.h> if HAVE_SYS_LOCKING_H.
18922 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
18924 * config/i386/predicates.md (fcmov_comparison_operator):
18925 Do not check for trivial FP comparison operator.
18926 <case GEU, case LTU>: Allow CCGZmode.
18927 <case GTU, case LEU>: Do not allow CCCmode.
18928 (ix86_comparison_operator) <case GTU, case LEU>: Allow only CCmode.
18929 (ix86_carry_flag_operator): Match only LTU and UNLT code.
18930 Do not check for trivial FP comparison operator. Allow CCGZmode.
18932 2021-04-29 Tom de Vries <tdevries@suse.de>
18934 * omp-expand.c (expand_omp_simd): Add step_orig, and replace uses of
18935 fd->loop.step by either step or orig_step.
18937 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
18939 * config/sparc/sparc.c (gen_load_pcrel_sym): Delete.
18940 (load_got_register): Do the PIC dance here.
18941 (sparc_legitimize_tls_address): Simplify.
18942 (sparc_emit_probe_stack_range): Likewise.
18943 (sparc32_initialize_trampoline): Likewise.
18944 (sparc64_initialize_trampoline): Likewise.
18945 * config/sparc/sparc.md (load_pcrel_sym<P:mode>): Add @ marker.
18946 (probe_stack_range<P:mode>): Likewise.
18947 (flush<P:mode>): Likewise.
18948 (tgd_hi22<P:mode>): Likewise.
18949 (tgd_lo10<P:mode>): Likewise.
18950 (tgd_add<P:mode>): Likewise.
18951 (tgd_call<P:mode>): Likewise.
18952 (tldm_hi22<P:mode>): Likewise.
18953 (tldm_lo10<P:mode>): Likewise.
18954 (tldm_add<P:mode>): Likewise.
18955 (tldm_call<P:mode>): Likewise.
18956 (tldo_hix22<P:mode>): Likewise.
18957 (tldo_lox10<P:mode>): Likewise.
18958 (tldo_add<P:mode>): Likewise.
18959 (tie_hi22<P:mode>): Likewise.
18960 (tie_lo10<P:mode>): Likewise.
18961 (tie_add<P:mode>): Likewise.
18962 (tle_hix22<P:mode>): Likewise.
18963 (tle_lox10<P:mode>): Likewise.
18964 (stack_protect_setsi): Rename to...
18965 (stack_protect_set32): ...this.
18966 (stack_protect_setdi): Rename to...
18967 (stack_protect_set64): ...this.
18968 (stack_protect_set): Adjust calls to above.
18969 (stack_protect_testsi): Rename to...
18970 (stack_protect_test32): ...this.
18971 (stack_protect_testdi): Rename to...
18972 (stack_protect_test64): ...this.
18973 (stack_protect_test): Adjust calls to above.
18975 2021-04-29 H.J. Lu <hjl.tools@gmail.com>
18977 PR middle-end/90773
18978 * builtins.c (builtin_memcpy_read_str): Add a dummy argument.
18979 (builtin_strncpy_read_str): Likewise.
18980 (builtin_memset_read_str): Add an argument for the previous RTL
18981 information and generate the new RTL from the previous RTL info.
18982 (builtin_memset_gen_str): Likewise.
18983 * builtins.h (builtin_strncpy_read_str): Update the prototype.
18984 (builtin_memset_read_str): Likewise.
18985 * expr.c (by_pieces_ninsns): If targetm.overlap_op_by_pieces_p()
18986 returns true, round up size and alignment to the widest integer
18987 mode for maximum size.
18988 (pieces_addr::adjust): Add a pointer to by_pieces_prev argument
18989 and pass it to m_constfn.
18990 (op_by_pieces_d): Add m_push and m_overlap_op_by_pieces.
18991 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
18992 initialize m_push. Initialize m_overlap_op_by_pieces with
18993 targetm.overlap_op_by_pieces_p ().
18994 (op_by_pieces_d::run): Pass the previous RTL information to
18995 pieces_addr::adjust and generate overlapping operations if
18996 m_overlap_op_by_pieces is true.
18998 (move_by_pieces_d::move_by_pieces_d): Updated for op_by_pieces_d
19000 (store_by_pieces_d::store_by_pieces_d): Updated for op_by_pieces_d
19002 (can_store_by_pieces): Use by_pieces_constfn on constfun.
19003 (store_by_pieces): Use by_pieces_constfn on constfun. Updated
19004 for op_by_pieces_d change.
19005 (clear_by_pieces_1): Add a dummy argument.
19006 (clear_by_pieces): Updated for op_by_pieces_d change.
19007 (compare_by_pieces_d::compare_by_pieces_d): Likewise.
19008 (string_cst_read_str): Add a dummy argument.
19009 * expr.h (by_pieces_constfn): Add a dummy argument.
19010 (by_pieces_prev): New.
19011 * target.def (overlap_op_by_pieces_p): New target hook.
19012 * config/i386/i386.c (TARGET_OVERLAP_OP_BY_PIECES_P): New.
19013 * doc/tm.texi.in: Add TARGET_OVERLAP_OP_BY_PIECES_P.
19014 * doc/tm.texi: Regenerated.
19016 2021-04-29 Richard Biener <rguenther@suse.de>
19018 PR tree-optimization/100253
19019 * tree-vect-stmts.c (vectorizable_load): Do not assume
19020 element alignment when DR_MISALIGNMENT is -1.
19021 (vectorizable_store): Likewise.
19023 2021-04-29 Jakub Jelinek <jakub@redhat.com>
19026 * config/aarch64/aarch64.c (aarch64_add_offset_1_temporaries): Use
19027 absu_hwi instead of abs_hwi.
19029 2021-04-29 Richard Biener <rguenther@suse.de>
19031 PR middle-end/38474
19032 * tree-ssa-structalias.c (add_graph_edge): Avoid direct
19033 forwarding when indirect forwarding through ESCAPED
19036 2021-04-29 Tom de Vries <tdevries@suse.de>
19039 * internal-fn.c (expand_GOMP_SIMT_ENTER_ALLOC)
19040 (expand_GOMP_SIMT_LAST_LANE, expand_GOMP_SIMT_ORDERED_PRED)
19041 (expand_GOMP_SIMT_VOTE_ANY, expand_GOMP_SIMT_XCHG_BFLY)
19042 (expand_GOMP_SIMT_XCHG_IDX): Ensure target is assigned to.
19044 2021-04-29 Richard Biener <rguenther@suse.de>
19046 PR tree-optimization/99912
19047 * tree-ssa-dse.c (dse_dom_walker::m_need_cfg_cleanup): New.
19048 (dse_dom_walker::todo): Likewise.
19049 (dse_dom_walker::dse_optimize_stmt): Move VDEF check to the
19051 (dse_dom_walker::before_dom_children): Remove trivially
19052 dead SSA defs and schedule CFG cleanup if we removed all
19054 (pass_dse::execute): Get TODO as computed by the DOM walker
19055 and return it. Wipe dominator info earlier.
19057 2021-04-29 Richard Biener <rguenther@suse.de>
19060 * ipa-prop.c (ipcp_modif_dom_walker::before_dom_children):
19061 Track blocks to cleanup EH in new m_need_eh_cleanup.
19062 (ipcp_modif_dom_walker::cleanup_eh): New.
19063 (ipcp_transform_function): Release dominator info before
19066 2021-04-29 Martin Sebor <msebor@redhat.com>
19068 PR middle-end/100250
19069 * attribs.c (attr_access::array_as_string): Avoid dereferencing
19070 a pointer when it's null.
19072 2021-04-29 Martin Sebor <msebor@redhat.com>
19074 * Makefile.in (OBJS): Add ipa-free-lang-data.o.
19075 * ipa-free-lang-data.cc: New file.
19076 * tree.c: Move pass free_lang_data to file above.
19077 (build_array_type_1): Declare extern.
19078 * tree.h (build_array_type_1): Declare.
19080 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
19082 * config/aarch64/aarch64-simd-builtins.def: Modify comment to
19083 make consistent with updated RTL pattern.
19084 * config/aarch64/aarch64-simd.md (aarch64_<sur>qmovn<mode>):
19085 Implement using ss_truncate and us_truncate rather than
19087 * config/aarch64/iterators.md: Remove redundant unspecs and
19088 iterator: UNSPEC_[SU]QXTN and SUQMOVN respectively.
19090 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
19092 * config/aarch64/arm_acle.h (__attribute__): Make intrinsic
19093 attributes consistent with those defined in arm_neon.h.
19095 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
19097 * config/aarch64/arm_fp16.h (__attribute__): Make intrinsic
19098 attributes consistent with those defined in arm_neon.h.
19100 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
19102 * config/aarch64/aarch64-simd-builtins.def: Add
19103 float_trunc_rodd builtin generator macros.
19104 * config/aarch64/aarch64-simd.md (aarch64_float_trunc_rodd_df):
19106 (aarch64_float_trunc_rodd_lo_v2sf): Define.
19107 (aarch64_float_trunc_rodd_hi_v4sf_le): Define.
19108 (aarch64_float_trunc_rodd_hi_v4sf_be): Define.
19109 (aarch64_float_trunc_rodd_hi_v4sf): Define.
19110 * config/aarch64/arm_neon.h (vcvtx_f32_f64): Use RTL builtin
19111 instead of inline asm.
19112 (vcvtx_high_f32_f64): Likewise.
19113 (vcvtxd_f32_f64): Likewise.
19114 * config/aarch64/iterators.md: Add FCVTXN unspec.
19116 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
19118 * config/aarch64/aarch64-simd-builtins.def: Add tbx1 builtin
19120 * config/aarch64/aarch64-simd.md (aarch64_tbx1<mode>):
19122 * config/aarch64/arm_neon.h (vqtbx1_s8): USE RTL builtin
19123 instead of inline asm.
19124 (vqtbx1_u8): Likewise.
19125 (vqtbx1_p8): Likewise.
19126 (vqtbx1q_s8): Likewise.
19127 (vqtbx1q_u8): Likewise.
19128 (vqtbx1q_p8): Likewise.
19129 (vtbx2_s8): Likewise.
19130 (vtbx2_u8): Likewise.
19131 (vtbx2_p8): Likewise.
19133 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
19135 * config/aarch64/aarch64-simd-builtins.def: Add tbl1 builtin
19137 * config/aarch64/arm_neon.h (vqtbl1_p8): Use RTL builtin
19138 instead of inline asm.
19139 (vqtbl1_s8): Likewise.
19140 (vqtbl1_u8): Likewise.
19141 (vqtbl1q_p8): Likewise.
19142 (vqtbl1q_s8): Likewise.
19143 (vqtbl1q_u8): Likewise.
19144 (vtbl1_s8): Likewise.
19145 (vtbl1_u8): Likewise.
19146 (vtbl1_p8): Likewise.
19147 (vtbl2_s8): Likewise.
19148 (vtbl2_u8): Likewise.
19149 (vtbl2_p8): Likewise.
19151 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
19153 * config/aarch64/aarch64-simd-builtins.def: Add polynomial
19154 ssri_n buitin generator macro.
19155 * config/aarch64/arm_neon.h (vsri_n_p8): Use RTL builtin
19156 instead of inline asm.
19157 (vsri_n_p16): Likewise.
19158 (vsri_n_p64): Likewise.
19159 (vsriq_n_p8): Likewise.
19160 (vsriq_n_p16): Likewise.
19161 (vsriq_n_p64): Likewise.
19163 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
19165 * config/aarch64/aarch64-simd-builtins.def: Use VALLP mode
19166 iterator for polynomial ssli_n builtin generator macro.
19167 * config/aarch64/arm_neon.h (vsli_n_p8): Use RTL builtin
19168 instead of inline asm.
19169 (vsli_n_p16): Likewise.
19170 (vsliq_n_p8): Likewise.
19171 (vsliq_n_p16): Likewise.
19172 * config/aarch64/iterators.md: Define VALLP mode iterator.
19174 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
19176 * config/aarch64/aarch64-simd-builtins.def: Use VDQV_L
19177 iterator to generate [su]adalp RTL builtins.
19178 * config/aarch64/aarch64-simd.md: Use VDQV_L iterator in
19179 [su]adalp RTL pattern.
19180 * config/aarch64/arm_neon.h (vpadal_s32): Use RTL builtin
19181 instead of inline asm.
19182 (vpadal_u32): Likewise.
19184 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
19186 * config/aarch64/aarch64-simd-builtins.def: Add [su]addlp
19187 builtin generator macros.
19188 * config/aarch64/aarch64-simd.md (aarch64_<su>addlp<mode>):
19190 * config/aarch64/arm_neon.h (vpaddl_s8): Use RTL builtin
19191 instead of inline asm.
19192 (vpaddl_s16): Likewise.
19193 (vpaddl_s32): Likewise.
19194 (vpaddl_u8): Likewise.
19195 (vpaddl_u16): Likewise.
19196 (vpaddl_u32): Likewise.
19197 (vpaddlq_s8): Likewise.
19198 (vpaddlq_s16): Likewise.
19199 (vpaddlq_s32): Likewise.
19200 (vpaddlq_u8): Likewise.
19201 (vpaddlq_u16): Likewise.
19202 (vpaddlq_u32): Liwewise.
19203 * config/aarch64/iterators.md: Define [SU]ADDLP unspecs with
19204 appropriate attributes.
19206 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
19208 * config/aarch64/aarch64-simd-builtins.def: Use VDQ_I iterator
19209 for aarch64_addp<mode> builtin macro generator.
19210 * config/aarch64/aarch64-simd.md: Use VDQ_I iterator in
19211 aarch64_addp<mode> RTL pattern.
19212 * config/aarch64/arm_neon.h (vpaddq_s8): Use RTL builtin
19213 instead of inline asm.
19214 (vpaddq_s16): Likewise.
19215 (vpaddq_s32): Likewise.
19216 (vpaddq_s64): Likewise.
19217 (vpaddq_u8): Likewise.
19218 (vpaddq_u16): Likewise.
19219 (vpaddq_u32): Likewise.
19220 (vpaddq_u64): Likewise.
19222 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
19224 * config/aarch64/aarch64-simd-builtins.def: Add sq[r]dmulh_n
19225 builtin generator macros.
19226 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_n<mode>):
19228 * config/aarch64/arm_neon.h (vqdmulh_n_s16): Use RTL builtin
19229 instead of inline asm.
19230 (vqdmulh_n_s32): Likewise.
19231 (vqdmulhq_n_s16): Likewise.
19232 (vqdmulhq_n_s32): Likewise.
19233 (vqrdmulh_n_s16): Likewise.
19234 (vqrdmulh_n_s32): Likewise.
19235 (vqrdmulhq_n_s16): Likewise.
19236 (vqrdmulhq_n_s32): Likewise.
19238 2021-04-28 Tobias Burnus <tobias@codesourcery.com>
19240 * doc/install.texi (--enable-offload-defaulted): Document.
19242 2021-04-28 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
19244 * config/avr/avr-dimode.md: Turn existing patterns into
19245 define_insn_and_split style patterns where the splitter
19246 adds a clobber of the condition code register. Drop "cc"
19247 attribute. Add new patterns to match output of
19249 * config/avr/avr-fixed.md: Likewise.
19250 * config/avr/avr.c (cc_reg_rtx): New.
19251 (avr_parallel_insn_from_insns): Adjust insn count
19252 for removal of set of cc0.
19253 (avr_is_casesi_sequence): Likewise.
19254 (avr_casei_sequence_check_operands): Likewise.
19255 (avr_optimize_casesi): Likewise. Also insert
19256 new insns after jump_insn.
19257 (avr_pass_casesi::avr_rest_of_handle_casesi): Adjust
19258 for removal of set of cc0.
19259 (avr_init_expanders): Initialize cc_reg_rtx.
19260 (avr_regno_reg_class): Handle REG_CC.
19261 (cond_string): Remove usage of CC_OVERFLOW_UNUSABLE.
19262 (avr_notice_update_cc): Remove function.
19263 (ret_cond_branch): Remove usage of CC_OVERFLOW_UNUSABLE.
19264 (compare_condition): Adjust for PARALLEL with
19266 (out_shift_with_cnt): Likewise.
19267 (ashlhi3_out): Likewise.
19268 (ashrhi3_out): Likewise.
19269 (lshrhi3_out): Likewise.
19270 (avr_class_max_nregs): Return single reg for REG_CC.
19271 (avr_compare_pattern): Check for REG_CC instead
19273 (avr_reorg_remove_redundant_compare): Likewise.
19274 (avr_reorg):Adjust for PARALLEL with REG_CC clobber.
19275 (avr_hard_regno_nregs): Return single reg for REG_CC.
19276 (avr_hard_regno_mode_ok): Allow only CCmode for REG_CC.
19277 (avr_md_asm_adjust): Clobber REG_CC.
19278 (TARGET_HARD_REGNO_NREGS): Define.
19279 (TARGET_CLASS_MAX_NREGS): Define.
19280 (TARGET_MD_ASM_ADJUST): Define.
19281 * config/avr/avr.h (FIRST_PSEUDO_REGISTER): Adjust
19283 (enum reg_class): Add CC_REG class.
19284 (NOTICE_UPDATE_CC): Remove.
19285 (CC_OVERFLOW_UNUSABLE): Remove.
19286 (CC_NO_CARRY): Remove.
19287 * config/avr/avr.md: Turn existing patterns into
19288 define_insn_and_split style patterns where the splitter
19289 adds a clobber of the condition code register. Drop "cc"
19290 attribute. Add new patterns to match output of
19292 (sez): Remove unused pattern.
19294 2021-04-28 Richard Earnshaw <rearnsha@arm.com>
19297 * config/arm/arm.c (arm_hard_regno_mode_ok): Only allow VPR to be
19300 2021-04-28 Richard Sandiford <richard.sandiford@arm.com>
19303 * config/aarch64/constraints.md (Utq): Require the address to
19304 be valid for both the element mode and for V2DImode.
19306 2021-04-28 Jakub Jelinek <jakub@redhat.com>
19307 Tobias Burnus <tobias@codesourcery.com>
19309 * configure.ac (OFFLOAD_DEFAULTED): AC_DEFINE if offload-defaulted.
19310 * gcc.c (process_command): New variable.
19311 (driver::maybe_putenv_OFFLOAD_TARGETS): If OFFLOAD_DEFAULTED,
19312 set it if -foffload is defaulted.
19313 * lto-wrapper.c (OFFLOAD_TARGET_DEFAULT_ENV): Define.
19314 (compile_offload_image): If OFFLOAD_DEFAULTED and
19315 OFFLOAD_TARGET_DEFAULT is in the environment, don't fail
19316 if corresponding mkoffload can't be found.
19317 (compile_images_for_offload_targets): Likewise. Free and clear
19318 offload_names if no valid offload is found.
19319 * config.in: Regenerate.
19320 * configure: Regenerate.
19322 2021-04-28 Richard Biener <rguenther@suse.de>
19324 PR tree-optimization/100292
19325 * tree-vect-generic.c (expand_vector_condition): Do not fold
19328 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
19330 * config/rs6000/aix.h (SUBTARGET_DRIVER_SELF_SPECS): New.
19331 * config/rs6000/aix64.opt (m64): New.
19334 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
19336 * config/vax/vax.c (print_operand_address, vax_address_cost_1)
19337 (index_term_p): Handle ASHIFT too.
19339 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
19341 * config/vax/builtins.md (jbb<ccss>i<mode>): Remove operand #3.
19342 (sync_lock_test_and_set<mode>): Adjust accordingly.
19343 (sync_lock_release<mode>): Likewise.
19345 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
19347 * config/vax/vax-protos.h (adjacent_operands_p): Remove
19349 * config/vax/vax.c (adjacent_operands_p): Remove.
19351 2021-04-27 Maciej W. Rozycki <macro@linux-mips.org>
19353 * ifcvt.c (dead_or_predicable) [!IFCVT_MODIFY_TESTS]: Fall
19354 through to the non-conditional execution case if getting the
19355 condition for conditional execution has failed.
19357 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
19359 PR middle-end/100284
19360 * gimple.c (gimple_could_trap_p_1): Remove VEC_COND_EXPR test.
19361 * tree-eh.c (operation_could_trap_p): Handle VEC_COND_EXPR rather
19362 than asserting on it.
19364 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
19366 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): Protect
19367 with TARGET_AIX_OS.
19369 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
19372 * calls.c (precompute_register_parameters): Additionally test
19373 targetm.precompute_tls_p to pre-compute argument.
19374 * config/rs6000/aix.h (TARGET_PRECOMPUTE_TLS_P): Define.
19375 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): New.
19376 * target.def (precompute_tls_p): New.
19377 * doc/tm.texi.in (TARGET_PRECOMPUTE_TLS_P): Add hook documentation.
19378 * doc/tm.texi: Regenerated.
19380 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19383 * config/aarch64/aarch64.c (aarch64_print_operand): Cast -UINTVAL
19384 back to HOST_WIDE_INT.
19386 2021-04-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
19389 * simplify-rtx.c (simplify_context::simplify_subreg): Check the
19390 memory alignment for the outer mode.
19392 2021-04-27 H.J. Lu <hjl.tools@gmail.com>
19394 PR middle-end/90773
19395 * expr.c (op_by_pieces_d::get_usable_mode): New member function.
19396 (op_by_pieces_d::run): Cange a while loop to a do-while loop.
19398 2021-04-27 Alex Coplan <alex.coplan@arm.com>
19401 * config/arm/arm.c (arm_split_compare_and_swap): Fix up codegen
19402 with negative immediates: ensure we expand cbranchsi4_scratch
19403 correctly and ensure we satisfy its constraints.
19404 * config/arm/sync.md
19405 (@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): Don't
19406 attempt to tie two output operands together with constraints;
19407 collapse two alternatives.
19408 (@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
19409 * config/arm/thumb1.md (cbranchsi4_neg_late): New.
19411 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19414 * config/aarch64/predicates.md (aarch64_sub_immediate,
19415 aarch64_plus_immediate): Use -UINTVAL instead of -INTVAL.
19416 * config/aarch64/aarch64.md (casesi, rotl<mode>3): Likewise.
19417 * config/aarch64/aarch64.c (aarch64_print_operand,
19418 aarch64_split_atomic_op, aarch64_expand_subvti): Likewise.
19420 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19422 PR tree-optimization/100239
19423 * tree-vect-generic.c (lower_vec_perm): Don't accept constant
19424 permutations with all indices from the first zero element as vec_shl.
19426 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19428 PR rtl-optimization/100254
19429 * cfgcleanup.c (outgoing_edges_match): Check REG_EH_REGION on
19430 last1 and last2 insns rather than BB_END (bb1) and BB_END (bb2) insns.
19432 2021-04-27 Richard Biener <rguenther@suse.de>
19434 PR tree-optimization/99912
19435 * passes.def: Add comment about new TODO_remove_unused_locals.
19436 * tree-stdarg.c (pass_data_stdarg): Run TODO_remove_unused_locals
19439 2021-04-27 Richard Biener <rguenther@suse.de>
19441 PR tree-optimization/99912
19442 * passes.def (pass_all_optimizations): Add pass_dse before
19443 the first pass_dce, move the first pass_dse before the
19444 pass_dce following pass_pre.
19446 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19448 PR tree-optimization/95527
19449 * generic-match-head.c: Include tm.h.
19450 * gimple-match-head.c: Include tm.h.
19451 * match.pd (CLZ == INTEGER_CST): Don't use
19452 #ifdef CLZ_DEFINED_VALUE_AT_ZERO, only test CLZ_DEFINED_VALUE_AT_ZERO
19453 if clz == CFN_CLZ. Add missing val declaration.
19454 (CTZ cmp CST): New simplifications.
19456 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19458 PR tree-optimization/96696
19459 * expr.c (expand_expr_divmod): New function.
19460 (expand_expr_real_2) <case TRUNC_DIV_EXPR>: Use it for truncations and
19461 divisions. Formatting fixes.
19462 <case MULT_EXPR>: Optimize x / y * y as x - x % y if the latter is
19465 2021-04-27 Martin Jambor <mjambor@suse.cz>
19468 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
19469 If removing a call statement LHS SSA name, release it.
19471 2021-04-27 Richard Earnshaw <rearnsha@arm.com>
19474 * config/arm/arm.c (THUMB2_WORK_REGS): Check PIC_OFFSET_TABLE_REGNUM
19475 is valid before including it in the mask.
19477 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
19480 * config/aarch64/aarch64.c (aarch64_comp_type_attributes): Handle
19483 2021-04-27 Richard Biener <rguenther@suse.de>
19485 PR tree-optimization/100051
19486 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Add
19487 disambiguator based on access size vs. decl size.
19489 2021-04-27 Richard Biener <rguenther@suse.de>
19491 PR tree-optimization/100278
19492 * tree-ssa-pre.c (compute_avail): Give up when we cannot
19493 adjust TBAA beacuse of mismatching bases.
19495 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19498 * config/i386/i386.md (*<insn><mode>3_mask, *<insn><mode>3_mask_1):
19499 For any_rotate define_insn_split and following splitters, use
19500 SWI iterator instead of SWI48.
19502 2021-04-27 Richard Biener <rguenther@suse.de>
19504 PR tree-optimization/99776
19505 * match.pd (bit_field_ref (ctor)): Relax element extract
19506 type compatibility checks.
19508 2021-04-27 Cui,Lili <lili.cui@intel.com>
19510 * common/config/i386/i386-common.c (processor_names):
19511 Sync processor_names with processor_type.
19512 * config/i386/i386-options.c (processor_cost_table):
19513 Sync processor_cost_table with processor_type.
19515 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
19517 * value-range.cc (irange::irange_set_1bit_anti_range): Add assert.
19518 (irange::set): Call irange_set_1bit_anti_range for handling all
19519 1-bit ranges. Fall through on ~[MIN,MAX].
19521 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
19523 * value-range.cc (irange::legacy_num_pairs): Remove.
19524 (irange::invert): Change gcc_assert to gcc_checking_assert.
19525 * value-range.h (irange::num_pairs): Adjust for a cached
19526 num_pairs(). Also, rename all gcc_assert's to
19527 gcc_checking_assert's.
19529 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
19531 * value-range.cc (irange::operator=): Set m_kind.
19532 (irange::copy_to_legacy): Handle varying and undefined sources
19533 as a legacy copy since they can be easily copied.
19534 (irange::irange_set): Set m_kind.
19535 (irange::irange_set_anti_range): Same.
19536 (irange::set): Rename normalize_min_max to normalize_kind.
19537 (irange::verify_range): Adjust for multi-ranges having the
19539 (irange::irange_union): Set m_kind.
19540 (irange::irange_intersect): Same.
19541 (irange::invert): Same.
19542 * value-range.h (irange::kind): Always return m_kind.
19543 (irange::varying_p): Rename to...
19544 (irange::varying_comptaible_p): ...this.
19545 (irange::undefined_p): Only look at m_kind.
19546 (irange::irange): Always set VR_UNDEFINED if applicable.
19547 (irange::set_undefined): Always set VR_UNDEFINED.
19548 (irange::set_varying): Always set m_kind to VR_VARYING.
19549 (irange::normalize_min_max): Rename to...
19550 (irange::normalize_kind): ...this.
19552 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
19554 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
19555 Adjust for constant_p including varying_p.
19556 * tree-vrp.c (vrp_prop::finalize): Same.
19557 (determine_value_range): Same.
19558 * vr-values.c (vr_values::range_of_expr): Same.
19559 * value-range.cc (irange::symbolic_p): Do not check varying_p.
19560 (irange::constant_p): Same.
19562 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
19564 * value-range.cc (irange::legacy_lower_bound): Replace
19565 !undefined_p check with num_ranges > 0.
19566 (irange::legacy_upper_bound): Same.
19567 * value-range.h (irange::type): Same.
19568 (irange::lower_bound): Same.
19569 (irange::upper_bound): Same.
19571 2021-04-26 Richard Biener <rguenther@suse.de>
19573 PR tree-optimization/99956
19574 * gimple-loop-interchange.cc (compute_access_stride):
19575 Try instantiating the access in a shallower loop nest
19576 if instantiating failed.
19577 (compute_access_strides): Pass adjustable loop_nest
19578 to compute_access_stride.
19580 2021-04-26 Christophe Lyon <christophe.lyon@linaro.org>
19582 * doc/sourcebuild.texi (arm_cmse_hw): Document.
19584 2021-04-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19586 * config/aarch64/iterators.md (vwcore): Handle V4BF, V8BF.
19588 2021-04-26 Thomas Schwinge <thomas@codesourcery.com>
19589 Nathan Sidwell <nathan@codesourcery.com>
19590 Tom de Vries <vries@codesourcery.com>
19591 Julian Brown <julian@codesourcery.com>
19592 Kwok Cheung Yeung <kcy@codesourcery.com>
19594 * omp-offload.c (oacc_validate_dims): Implement
19595 '-Wopenacc-parallelism'.
19596 * doc/invoke.texi (-Wopenacc-parallelism): Document.
19598 2021-04-26 Richard Biener <rguenther@suse.de>
19600 * tree-cfg.h (gimplify_build1): Remove.
19601 (gimplify_build2): Likewise.
19602 (gimplify_build3): Likewise.
19603 * tree-cfg.c (gimplify_build1): Move to tree-vect-generic.c.
19604 (gimplify_build2): Likewise.
19605 (gimplify_build3): Likewise.
19606 * tree-vect-generic.c (gimplify_build1): Move from tree-cfg.c.
19608 (gimplify_build2): Likewise.
19609 (gimplify_build3): Likewise.
19610 (tree_vec_extract): Use resimplify with following SSA edges.
19611 (expand_vector_parallel): Avoid passing NULL size/bitpos
19612 to tree_vec_extract.
19613 * expr.c (store_constructor): Deal with zero-element CTORs.
19614 * match.pd (bit_field_ref <vector CTOR>): Make sure to
19615 produce vector constants when possible.
19617 2021-04-26 Richard Biener <rguenther@suse.de>
19619 * tree-complex.c: Include gimple-fold.h.
19620 (expand_complex_addition): Use gimple_build.
19621 (expand_complex_multiplication_components): Likewise.
19622 (expand_complex_multiplication): Likewise.
19623 (expand_complex_div_straight): Likewise.
19624 (expand_complex_div_wide): Likewise.
19625 (expand_complex_division): Likewise.
19626 (expand_complex_conjugate): Likewise.
19627 (expand_complex_comparison): Likewise.
19629 2021-04-26 Richard Biener <rguenther@suse.de>
19631 * tree-ssa-phiopt.c (two_value_replacement): Remove use
19632 of legacy gimplify_buildN API.
19634 2021-04-26 Richard Biener <rguenther@suse.de>
19636 PR tree-optimization/99473
19637 * tree-ssa-phiopt.c (cond_store_replacement): Handle all
19640 2021-04-26 Richard Biener <rguenther@suse.de>
19642 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin):
19643 Use replace_call_with_value.
19645 2021-04-26 Richard Biener <rguenther@suse.de>
19647 * tree-ssa-propagate.h (valid_gimple_rhs_p): Remove.
19648 (update_gimple_call): Likewise.
19649 (update_call_from_tree): Likewise.
19650 * tree-ssa-propagate.c (valid_gimple_rhs_p): Remove.
19651 (valid_gimple_call_p): Likewise.
19652 (move_ssa_defining_stmt_for_defs): Likewise.
19653 (finish_update_gimple_call): Likewise.
19654 (update_gimple_call): Likewise.
19655 (update_call_from_tree): Likewise.
19656 (propagate_tree_value_into_stmt): Use replace_call_with_value.
19657 * gimple-fold.h (update_gimple_call): Declare.
19658 * gimple-fold.c (valid_gimple_rhs_p): Move here from
19659 tree-ssa-propagate.c.
19660 (update_gimple_call): Likewise.
19661 (valid_gimple_call_p): Likewise.
19662 (finish_update_gimple_call): Likewise, and simplify.
19663 (gimplify_and_update_call_from_tree): Implement
19664 update_call_from_tree functionality, avoid excessive
19665 push/pop_gimplify_context.
19666 (gimple_fold_builtin): Use only gimplify_and_update_call_from_tree.
19667 (gimple_fold_call): Likewise.
19668 * gimple-ssa-sprintf.c (try_substitute_return_value): Likewise.
19669 * tree-ssa-ccp.c (ccp_folder::fold_stmt): Likewise.
19670 (pass_fold_builtins::execute): Likewise.
19671 (optimize_stack_restore): Use replace_call_with_value.
19672 * tree-cfg.c (fold_loop_internal_call): Likewise.
19673 * tree-ssa-dce.c (maybe_optimize_arith_overflow): Use
19674 only gimplify_and_update_call_from_tree.
19675 * tree-ssa-strlen.c (handle_builtin_strlen): Likewise.
19676 (handle_builtin_strchr): Likewise.
19677 * tsan.c: Include gimple-fold.h instead of tree-ssa-propagate.h.
19679 2021-04-26 Jakub Jelinek <jakub@redhat.com>
19682 * vmsdbgout.c (ASM_OUTPUT_DEBUG_STRING, vmsdbgout_begin_block,
19683 vmsdbgout_end_block, lookup_filename, vmsdbgout_source_line): Remove
19686 2021-04-25 liuhongt <hongtao.liu@intel.com>
19689 * config/i386/i386-builtin.def (BDESC): Change the icode of
19690 the following builtins to CODE_FOR_nothing.
19691 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
19692 IX86_BUILTIN_PCMPEQB128, IX86_BUILTIN_PCMPEQW128,
19693 IX86_BUILTIN_PCMPEQD128, IX86_BUILTIN_PCMPEQQ,
19694 IX86_BUILTIN_PCMPEQB256, IX86_BUILTIN_PCMPEQW256,
19695 IX86_BUILTIN_PCMPEQD256, IX86_BUILTIN_PCMPEQQ256,
19696 IX86_BUILTIN_PCMPGTB128, IX86_BUILTIN_PCMPGTW128,
19697 IX86_BUILTIN_PCMPGTD128, IX86_BUILTIN_PCMPGTQ,
19698 IX86_BUILTIN_PCMPGTB256, IX86_BUILTIN_PCMPGTW256,
19699 IX86_BUILTIN_PCMPGTD256, IX86_BUILTIN_PCMPGTQ256.
19700 * config/i386/sse.md (avx2_eq<mode>3): Deleted.
19701 (sse2_eq<mode>3): Ditto.
19702 (sse4_1_eqv2di3): Ditto.
19703 (sse2_gt<mode>3): Rename to ..
19704 (*sse2_gt<mode>3): .. this.
19706 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
19709 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
19712 * config/darwin.c (darwin_binds_local_p): Assume that any
19713 public symbol might be interposed for PIC code. Update function
19714 header comment to reflect current Darwin capability.
19716 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
19719 * config/darwin.c (darwin_binds_local_p): Assume that any
19720 public symbol might be interposed for PIC code. Update function
19721 header comment to reflect current Darwin capability.
19723 2021-04-24 Richard Sandiford <richard.sandiford@arm.com>
19725 * doc/sourcebuild.texi: Document no-opts and any-opts target
19728 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
19730 * config/bpf/bpf.h (ASM_OUTPUT_ALIGNED_BSS): Use .type and .lcomm.
19732 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
19734 * config/bpf/bpf.h (FUNCTION_BOUNDARY): Set to 64.
19736 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
19739 * config/i386/i386-options.c (ix86_option_override_internal):
19740 Error out when -m96bit-long-double is used with 64bit targets.
19741 * config/i386/i386.md (*pushxf_rounded): Remove pattern.
19743 2021-04-23 Martin Liska <mliska@suse.cz>
19745 * lto-wrapper.c: Remove FIXME about usage of
19746 hardware_concurrency. The function is not on par with
19749 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
19752 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
19753 Copy operand 3 to operand 4. Use sse_reg_operand
19754 as operand 3 predicate.
19755 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): Ditto.
19756 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
19757 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
19758 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2):
19759 Copy operand 1 to operand 0.
19760 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): Ditto.
19761 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
19762 (LDX_ATOMIC/LDX_ATOMIC FP store peephole2 with mem blockage): Ditto.
19764 2021-04-23 Alex Coplan <alex.coplan@arm.com>
19766 PR rtl-optimization/100230
19767 * early-remat.c (early_remat::sort_candidates): Use delete[]
19768 instead of delete for array allocated with new[].
19770 2021-04-23 Richard Biener <rguenther@suse.de>
19772 * genmatch.c (lower_cond): Remove VEC_COND_EXPR special-casing.
19773 (capture_info::capture_info): Likewise.
19774 (capture_info::walk_match): Likewise.
19775 (expr::gen_transform): Likewise.
19776 (dt_simplify::gen_1): Likewise.
19777 * gimple-match-head.c (maybe_resimplify_conditional_op):
19778 Remove VEC_COND_EXPR special-casing.
19779 (gimple_simplify): Likewise.
19780 * gimple.c (gimple_could_trap_p_1): Adjust.
19781 * tree-ssa-pre.c (compute_avail): Allow VEC_COND_EXPR
19782 to participate in PRE.
19784 2021-04-23 Richard Biener <rguenther@suse.de>
19786 * cfganal.c (connect_infinite_loops_to_exit): First call
19787 add_noreturn_fake_exit_edges.
19788 * ipa-sra.c (process_scan_results): Do not call the now redundant
19789 add_noreturn_fake_exit_edges.
19790 * predict.c (tree_estimate_probability): Likewise.
19791 (rebuild_frequencies): Likewise.
19792 * store-motion.c (one_store_motion_pass): Likewise.
19794 2021-04-23 Richard Biener <rguenther@suse.de>
19796 PR tree-optimization/100222
19797 * predict.c (pass_profile::execute): Remove redundant call to
19798 mark_irreducible_loops.
19799 (report_predictor_hitrates): Likewise.
19801 2021-04-23 Richard Biener <rguenther@suse.de>
19803 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Avoid
19804 valid_gimple_rhs_p by instead gimplifying to one.
19806 2021-04-23 Richard Biener <rguenther@suse.de>
19808 PR tree-optimization/99971
19809 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
19810 Always use TBAA for loads.
19812 2021-04-23 liuhongt <hongtao.liu@intel.com>
19815 * config/i386/i386-options.c (ix86_option_override_internal):
19816 Clear MASK_AVX256_SPLIT_UNALIGNED_LOAD/STORE in x_target_flags
19817 when X86_TUNE_AVX256_UNALIGNED_LOAD/STORE_OPTIMAL is enabled
19818 by target attribute.
19820 2021-04-23 David Edelsohn <dje.gcc@gmail.com>
19822 * config/rs6000/aix71.h (PREFERRED_DEBUGGING_TYPE): Change to
19824 * config/rs6000/aix72.h (PREFERRED_DEBUGGING_TYPE): Same.
19826 2021-04-22 David Edelsohn <dje.gcc@gmail.com>
19828 * config.gcc (powerpc-ibm-aix6.*): Remove.
19829 * config/rs6000/aix61.h: Delete.
19831 2021-04-22 Martin Liska <mliska@suse.cz>
19833 PR testsuite/100159
19834 PR testsuite/100192
19835 * builtins.c (expand_builtin): Fix typos and missing comments.
19836 * dwarf2out.c (gen_subprogram_die): Likewise.
19837 (gen_struct_or_union_type_die): Likewise.
19839 2021-04-22 Uroš Bizjak <ubizjak@gmail.com>
19842 * config/i386/i386-expand.c (ix86_expand_convert_uns_sidf_sse):
19843 Remove the sign with FE_DOWNWARD, where x - x = -0.0.
19845 2021-04-21 Iain Sandoe <iain@sandoe.co.uk>
19847 * config/i386/darwin.h (TARGET_64BIT): Remove definition
19848 based on TARGET_ISA_64BIT.
19849 (TARGET_64BIT_P): Remove definition based on
19850 TARGET_ISA_64BIT_P().
19852 2021-04-21 Martin Liska <mliska@suse.cz>
19855 2021-04-21 Martin Liska <mliska@suse.cz>
19857 * lto-wrapper.c (cpuset_popcount): Remove.
19858 (init_num_threads): Remove and use hardware_concurrency.
19860 2021-04-21 Martin Liska <mliska@suse.cz>
19863 * main.c (main): Call toplev::finalize in CHECKING_P mode.
19864 * ipa-modref.c (ipa_modref_c_finalize): summaries are NULL
19865 when incremental LTO linking happens.
19867 2021-04-21 Martin Liska <mliska@suse.cz>
19869 * lto-wrapper.c (run_gcc): When -flto=jobserver is used, but the
19870 makeserver cannot be detected, then use -flto=N fallback.
19872 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
19874 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): When cross-compiling,
19875 default to yes for aarch64-linux-gnu.
19876 * configure: Regenerate.
19878 2021-04-21 Martin Liska <mliska@suse.cz>
19880 * lto-wrapper.c (cpuset_popcount): Remove.
19881 (init_num_threads): Remove and use hardware_concurrency.
19883 2021-04-21 Martin Liska <mliska@suse.cz>
19885 * config/i386/i386.c: Remove superfluous || TARGET_MACHO
19886 which remains to be '(... || 0)' and clang complains about it.
19887 * dwarf2out.c (AT_vms_delta): Declare conditionally.
19888 (add_AT_vms_delta): Likewise.
19889 * tree.c (fld_simplified_type): Use rather more common pattern
19890 for disabling of something (#if 0).
19891 (get_tree_code_name): Likewise.
19892 (verify_type_variant): Likewise.
19894 2021-04-21 Martin Liska <mliska@suse.cz>
19896 * config/i386/i386-expand.c (decide_alignment): Use newly named
19897 macro TARGET_CPU_P.
19898 * config/i386/i386.c (ix86_decompose_address): Likewise.
19899 (ix86_address_cost): Likewise.
19900 (ix86_lea_outperforms): Likewise.
19901 (ix86_avoid_lea_for_addr): Likewise.
19902 (ix86_add_stmt_cost): Likewise.
19903 * config/i386/i386.h (TARGET_*): Remove.
19904 (TARGET_CPU_P): New macro.
19905 * config/i386/i386.md: Use newly named macro TARGET_CPU_P.
19906 * config/i386/x86-tune-sched-atom.c (do_reorder_for_imul): Likewise.
19907 (swap_top_of_ready_list): Likewise.
19908 (ix86_atom_sched_reorder): Likewise.
19909 * config/i386/x86-tune-sched-bd.c (ix86_bd_has_dispatch): Likewise.
19910 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Likewise.
19912 2021-04-21 Martin Liska <mliska@suse.cz>
19914 * config/i386/i386-options.c (TARGET_EXPLICIT_NO_SAHF_P):
19916 (SET_TARGET_NO_SAHF): Likewise.
19917 (TARGET_EXPLICIT_PREFETCH_SSE_P): Likewise.
19918 (SET_TARGET_PREFETCH_SSE): Likewise.
19919 (TARGET_EXPLICIT_NO_TUNE_P): Likewise.
19920 (SET_TARGET_NO_TUNE): Likewise.
19921 (TARGET_EXPLICIT_NO_80387_P): Likewise.
19922 (SET_TARGET_NO_80387): Likewise.
19924 * config/i386/i386.h (TARGET_*): Remove.
19925 * opth-gen.awk: Generate new used macros.
19927 2021-04-21 Martin Liska <mliska@suse.cz>
19929 * config/i386/i386.h (PTA_*): Remove.
19930 (enum pta_flag): New.
19931 (DEF_PTA): Generate PTA_* values from i386-isa.def.
19932 * config/i386/i386-isa.def: New file.
19934 2021-04-21 Alex Coplan <alex.coplan@arm.com>
19937 * config/aarch64/aarch64-bti-insert.c (aarch64_bti_j_insn_p): New.
19938 (rest_of_insert_bti): Avoid inserting duplicate bti j insns for
19939 jump table targets.
19941 2021-04-21 H.J. Lu <hjl.tools@gmail.com>
19943 * config.gcc: Install mwaitintrin.h for i[34567]86-*-* and
19944 x86_64-*-* targets.
19945 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_MWAIT_SET):
19947 (OPTION_MASK_ISA2_MWAIT_UNSET): Likewise.
19948 (ix86_handle_option): Handle -mmwait.
19949 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
19950 Replace OPTION_MASK_ISA_SSE3 with OPTION_MASK_ISA2_MWAIT on
19951 __builtin_ia32_monitor and __builtin_ia32_mwait.
19952 * config/i386/i386-options.c (isa2_opts): Add -mmwait.
19953 (ix86_valid_target_attribute_inner_p): Likewise.
19954 (ix86_option_override_internal): Enable mwait/monitor
19955 instructions for -msse3.
19956 * config/i386/i386.h (TARGET_MWAIT): New.
19957 (TARGET_MWAIT_P): Likewise.
19958 * config/i386/i386.opt: Add -mmwait.
19959 * config/i386/mwaitintrin.h: New file.
19960 * config/i386/pmmintrin.h: Include <mwaitintrin.h>.
19961 * config/i386/sse.md (sse3_mwait): Replace TARGET_SSE3 with
19963 (@sse3_monitor_<mode>): Likewise.
19964 * config/i386/x86gprintrin.h: Include <mwaitintrin.h>.
19965 * doc/extend.texi: Document mwait target attribute.
19966 * doc/invoke.texi: Document -mmwait.
19968 2021-04-21 Martin Liska <mliska@suse.cz>
19970 * config/i386/i386-options.c (DEF_ENUM): Remove it.
19971 * config/i386/i386-opts.h (DEF_ENUM): Likewise.
19972 * config/i386/stringop.def (DEF_ENUM): Likewise.
19974 2021-04-21 Martin Liska <mliska@suse.cz>
19976 * tree-cfg.c (gimple_verify_flow_info): Use qD instead
19977 of print_generic_expr.
19979 2021-04-21 Jakub Jelinek <jakub@redhat.com>
19981 PR rtl-optimization/100148
19982 * cprop.c (constprop_register): Use next_nondebug_insn instead of
19985 2021-04-21 Martin Liska <mliska@suse.cz>
19988 * cgraphunit.c (cgraph_node::analyze): Remove duplicate
19989 free_dominance_info calls.
19991 2021-04-21 Richard Biener <rguenther@suse.de>
19993 * gimple-fold.c (maybe_fold_reference): Remove is_lhs
19994 parameter (and assume it to be false).
19995 (fold_gimple_assign): Adjust, remove all callers of
19996 maybe_fold_reference calling it with is_lhs true.
19997 (gimple_fold_call): Likewise.
19998 (fold_stmt_1): Likewise.
20000 2021-04-21 Richard Biener <rguenther@suse.de>
20002 * fold-const.c (pedantic_non_lvalue_loc): Remove.
20003 (fold_binary_loc): Adjust.
20004 (fold_ternary_loc): Likewise.
20006 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
20008 PR middle-end/100130
20009 * varasm.c (get_block_for_decl): Make sure that any use of the
20010 retain attribute matches the section's retain flag.
20011 (switch_to_section): Check for retain mismatches even when
20012 changing sections, but do not warn if the given decl is the
20013 section's named.decl.
20014 (output_object_block): Pass the first decl in the block (if any)
20015 to switch_to_section.
20017 2021-04-20 H.J. Lu <hjl.tools@gmail.com>
20019 * config/i386/i386-c.c (ix86_target_macros_internal): Define
20020 __CRC32__ for -mcrc32.
20021 * config/i386/i386-options.c (ix86_option_override_internal):
20022 Enable crc32 instruction for -msse4.2.
20023 * config/i386/i386.md (sse4_2_crc32<mode>): Remove TARGET_SSE4_2
20025 (sse4_2_crc32di): Likewise.
20026 * config/i386/ia32intrin.h: Use crc32 target option for CRC32
20029 2021-04-20 Segher Boessenkool <segher@kernel.crashing.org>
20032 * config/rs6000/rs6000.c (rs6000_machine_from_flags): Do not consider
20035 2021-04-20 Martin Liska <mliska@suse.cz>
20037 * doc/invoke.texi: Fix typo.
20038 * params.opt: Likewise.
20040 2021-04-20 Martin Liska <mliska@suse.cz>
20042 * doc/invoke.texi: Document new param.
20044 2021-04-19 Andrew MacLeod <amacleod@redhat.com>
20046 PR tree-optimization/100081
20047 * gimple-range-cache.h (ranger_cache): Inherit from gori_compute
20048 rather than gori_compute_cache.
20049 * gimple-range-gori.cc (is_gimple_logical_p): Move to top of file.
20050 (range_def_chain::m_logical_depth): New member.
20051 (range_def_chain::range_def_chain): Initialize m_logical_depth.
20052 (range_def_chain::get_def_chain): Don't build defchains through more
20053 than LOGICAL_LIMIT logical expressions.
20054 * params.opt (param_ranger_logical_depth): New.
20056 2021-04-19 Richard Earnshaw <rearnsha@arm.com>
20059 * config/arm/arm.c (arm_configure_build_target): Do not strip
20060 extended FPU/SIMD feature bits from the target ISA when -mfpu
20061 is specified (partial revert of r11-8168).
20063 2021-04-19 Thomas Schwinge <thomas@codesourcery.com>
20065 * params.opt (-param=openacc-kernels=): Add.
20066 * omp-oacc-kernels-decompose.cc
20067 (pass_omp_oacc_kernels_decompose::gate): Use it.
20068 * doc/invoke.texi (-fopenacc-kernels=@var{mode}): Move...
20069 (--param): ... here, 'openacc-kernels'.
20071 2021-04-19 Martin Liska <mliska@suse.cz>
20074 * gengtype.c (finish_root_table): Align function arguments
20075 in between declaration and definition.
20077 2021-04-19 Eric Botcazou <ebotcazou@adacore.com>
20079 * config/i386/winnt.c (i386_pe_seh_cold_init): Properly deal with
20080 frames larger than the SEH maximum frame size.
20082 2021-04-18 Segher Boessenkool <segher@kernel.crashing.org>
20084 PR rtl-optimization/99927
20085 * combine.c (distribute_notes) [REG_UNUSED]: If the register already
20086 is dead, just drop it.
20088 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
20091 * config/i386/winnt-d.c (TARGET_D_TEMPLATES_ALWAYS_COMDAT): Define.
20092 * doc/tm.texi: Regenerate.
20093 * doc/tm.texi.in (D language and ABI): Add @hook for
20094 TARGET_D_TEMPLATES_ALWAYS_COMDAT.
20096 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
20098 * config/darwin-d.c (darwin_d_handle_target_object_format): New
20100 (darwin_d_register_target_info): New function.
20101 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
20102 * config/dragonfly-d.c (dragonfly_d_handle_target_object_format): New
20104 (dragonfly_d_register_target_info): New function.
20105 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
20106 * config/freebsd-d.c (freebsd_d_handle_target_object_format): New
20108 (freebsd_d_register_target_info): New function.
20109 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
20110 * config/glibc-d.c (glibc_d_handle_target_object_format): New
20112 (glibc_d_register_target_info): New function.
20113 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
20114 * config/i386/i386-d.c (ix86_d_handle_target_object_format): New
20116 (ix86_d_register_target_info): Add ix86_d_handle_target_object_format
20117 as handler for objectFormat key.
20118 * config/i386/winnt-d.c (winnt_d_handle_target_object_format): New
20120 (winnt_d_register_target_info): New function.
20121 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
20122 * config/netbsd-d.c (netbsd_d_handle_target_object_format): New
20124 (netbsd_d_register_target_info): New function.
20125 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
20126 * config/openbsd-d.c (openbsd_d_handle_target_object_format): New
20128 (openbsd_d_register_target_info): New function.
20129 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
20130 * config/pa/pa-d.c (pa_d_handle_target_object_format): New function.
20131 (pa_d_register_target_info): Add pa_d_handle_target_object_format as
20132 handler for objectFormat key.
20133 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_object_format): New
20135 (rs6000_d_register_target_info): Add
20136 rs6000_d_handle_target_object_format as handler for objectFormat key.
20137 * config/sol2-d.c (solaris_d_handle_target_object_format): New
20139 (solaris_d_register_target_info): New function.
20140 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
20142 2021-04-16 Jakub Jelinek <jakub@redhat.com>
20145 * config/aarch64/aarch64.c (aarch64_function_arg_alignment): Change
20146 abi_break argument from bool * to unsigned *, store there the pre-GCC 9
20148 (aarch64_layout_arg, aarch64_gimplify_va_arg_expr): Adjust callers.
20149 (aarch64_function_arg_regno_p): Likewise. Only emit -Wpsabi note if
20150 the old and new alignment after applying MIN/MAX to it is different.
20152 2021-04-16 Tamar Christina <tamar.christina@arm.com>
20155 * config/aarch64/aarch64-sve.md (@aarch64_sve_trn1_conv<mode>): New.
20156 * config/aarch64/aarch64.c (aarch64_expand_sve_const_pred_trn): Use new
20158 * config/aarch64/iterators.md (UNSPEC_TRN1_CONV): New.
20160 2021-04-16 Bill Schmidt <wschmidt@linux.ibm.com>
20162 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Revise
20163 this section and its subsections.
20165 2021-04-16 Jakub Jelinek <jakub@redhat.com>
20168 * config/aarch64/aarch64.md (*neg_asr_si2_extr, *extrsi5_insn_di): New
20169 define_insn patterns.
20171 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
20173 PR rtl-optimization/98689
20174 * reg-notes.def (UNTYPED_CALL): New note.
20175 * combine.c (distribute_notes): Handle it.
20176 * emit-rtl.c (try_split): Likewise.
20177 * rtlanal.c (rtx_properties::try_to_add_insn): Likewise. Assume
20178 that calls with the note implicitly set all return value registers.
20179 * builtins.c (expand_builtin_apply): Add a REG_UNTYPED_CALL
20182 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
20184 PR rtl-optimization/99596
20185 * rtlanal.c (rtx_properties::try_to_add_insn): Don't add global
20186 register accesses for const calls. Assume that pure functions
20187 can only read from global registers. Ignore cases in which
20188 the stack pointer has been marked global.
20190 2021-04-16 Jakub Jelinek <jakub@redhat.com>
20193 * tree-vect-loop.c (vect_transform_loop): Don't remove just
20194 dead scalar .MASK_LOAD calls, but also dead .COND_* calls - replace
20195 them by their last argument.
20197 2021-04-15 Martin Liska <mliska@suse.cz>
20199 * doc/invoke.texi: Other params don't use it, remove it.
20201 2021-04-15 Richard Biener <rguenther@suse.de>
20203 * gimple-builder.h: Add deprecation note.
20205 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
20208 * attribs.h (restrict_type_identity_attributes_to): Declare.
20209 * attribs.c (restrict_type_identity_attributes_to): New function.
20211 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
20214 * attribs.h (affects_type_identity_attributes): Declare.
20215 * attribs.c (remove_attributes_matching): New function.
20216 (affects_type_identity_attributes): Likewise.
20218 2021-04-15 Jakub Jelinek <jakub@redhat.com>
20221 * config/aarch64/aarch64.md (*<LOGICAL:optab>_<SHIFT:optab><mode>3):
20222 Add combine splitters for *<LOGICAL:optab>_ashl<mode>3 with
20223 ZERO_EXTEND, SIGN_EXTEND or AND.
20225 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
20227 PR rtl-optimization/99929
20228 * rtl.h (same_vector_encodings_p): New function.
20229 * cse.c (exp_equiv_p): Check that CONST_VECTORs have the same encoding.
20230 * cselib.c (rtx_equal_for_cselib_1): Likewise.
20231 * jump.c (rtx_renumbered_equal_p): Likewise.
20232 * lra-constraints.c (operands_match_p): Likewise.
20233 * reload.c (operands_match_p): Likewise.
20234 * rtl.c (rtx_equal_p_cb, rtx_equal_p): Likewise.
20236 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
20238 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Print
20239 more information about variable-length CONST_VECTORs.
20241 2021-04-14 Vladimir N. Makarov <vmakarov@redhat.com>
20243 PR rtl-optimization/100066
20244 * lra-constraints.c (split_reg): Check paradoxical_subreg_p for
20245 ordered modes when choosing splitting mode for hard reg.
20247 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
20250 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector_sel):
20252 (aarch64_expand_sve_const_vector): Use it for nelts_per_pattern==2.
20254 2021-04-14 Andreas Krebbel <krebbel@linux.ibm.com>
20256 * config/s390/s390-builtins.def (O_M5, O_M12, ...): Add new macros
20257 for mask operand types.
20258 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
20259 (s390_vec_permi_dbl, s390_vpdi): Use the M5 type for the immediate
20261 (s390_vec_msum_u128, s390_vmslg): Use the M12 type for the
20263 * config/s390/s390.c (s390_const_operand_ok): Check the new
20264 operand types and generate a list of valid values.
20266 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
20268 * doc/tm.texi: Regenerate.
20269 * doc/tm.texi.in (D language and ABI): Add @hook for
20270 TARGET_D_REGISTER_OS_TARGET_INFO.
20272 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
20274 * config/aarch64/aarch64-d.c (aarch64_d_handle_target_float_abi): New
20276 (aarch64_d_register_target_info): New function.
20277 * config/aarch64/aarch64-protos.h (aarch64_d_register_target_info):
20279 * config/aarch64/aarch64.h (TARGET_D_REGISTER_CPU_TARGET_INFO):
20281 * config/arm/arm-d.c (arm_d_handle_target_float_abi): New function.
20282 (arm_d_register_target_info): New function.
20283 * config/arm/arm-protos.h (arm_d_register_target_info): Declare.
20284 * config/arm/arm.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
20285 * config/i386/i386-d.c (ix86_d_handle_target_float_abi): New function.
20286 (ix86_d_register_target_info): New function.
20287 * config/i386/i386-protos.h (ix86_d_register_target_info): Declare.
20288 * config/i386/i386.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
20289 * config/mips/mips-d.c (mips_d_handle_target_float_abi): New function.
20290 (mips_d_register_target_info): New function.
20291 * config/mips/mips-protos.h (mips_d_register_target_info): Declare.
20292 * config/mips/mips.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
20293 * config/pa/pa-d.c (pa_d_handle_target_float_abi): New function.
20294 (pa_d_register_target_info): New function.
20295 * config/pa/pa-protos.h (pa_d_register_target_info): Declare.
20296 * config/pa/pa.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
20297 * config/riscv/riscv-d.c (riscv_d_handle_target_float_abi): New
20299 (riscv_d_register_target_info): New function.
20300 * config/riscv/riscv-protos.h (riscv_d_register_target_info): Declare.
20301 * config/riscv/riscv.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
20302 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_float_abi): New
20304 (rs6000_d_register_target_info): New function.
20305 * config/rs6000/rs6000-protos.h (rs6000_d_register_target_info):
20307 * config/rs6000/rs6000.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
20308 * config/s390/s390-d.c (s390_d_handle_target_float_abi): New function.
20309 (s390_d_register_target_info): New function.
20310 * config/s390/s390-protos.h (s390_d_register_target_info): Declare.
20311 * config/s390/s390.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
20312 * config/sparc/sparc-d.c (sparc_d_handle_target_float_abi): New
20314 (sparc_d_register_target_info): New function.
20315 * config/sparc/sparc-protos.h (sparc_d_register_target_info): Declare.
20316 * config/sparc/sparc.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
20317 * doc/tm.texi: Regenerate.
20318 * doc/tm.texi.in (D language and ABI): Add @hook for
20319 TARGET_D_REGISTER_CPU_TARGET_INFO.
20321 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
20323 * config/i386/i386-d.c (ix86_d_has_stdcall_convention): New function.
20324 * config/i386/i386-protos.h (ix86_d_has_stdcall_convention): Declare.
20325 * config/i386/i386.h (TARGET_D_HAS_STDCALL_CONVENTION): Define.
20326 * doc/tm.texi: Regenerate.
20327 * doc/tm.texi.in (D language and ABI): Add @hook for
20328 TARGET_D_HAS_STDCALL_CONVENTION.
20330 2021-04-14 Richard Biener <rguenther@suse.de>
20332 * tree-cfg.c (verify_gimple_assign_ternary): Verify that
20333 VEC_COND_EXPRs have a gimple_val condition.
20334 * tree-ssa-propagate.c (valid_gimple_rhs_p): VEC_COND_EXPR
20335 can no longer have a GENERIC condition.
20337 2021-04-14 Richard Earnshaw <rearnsha@arm.com>
20340 * config/arm/arm.c (arm_configure_build_target): Strip isa_all_fpbits
20341 from the isa_delta when -mfpu has been used.
20342 (arm_options_perform_arch_sanity_checks): It's the architecture that
20343 lacks an FPU not the processor.
20345 2021-04-13 Richard Biener <rguenther@suse.de>
20347 PR tree-optimization/100053
20348 * tree-ssa-sccvn.c (vn_nary_op_get_predicated_value): Do
20349 not use optimistic dominance queries for backedges to validate
20351 (dominated_by_p_w_unex): Add parameter to ignore executable
20352 state on backedges.
20353 (rpo_elim::eliminate_avail): Adjust.
20355 2021-04-13 Jakub Jelinek <jakub@redhat.com>
20358 * config/aarch64/aarch64.md (*aarch64_bfxil<mode>_extr,
20359 *aarch64_bfxilsi_extrdi): New define_insn patterns.
20361 2021-04-13 Jakub Jelinek <jakub@redhat.com>
20364 * simplify-rtx.c (simplify_immed_subreg): For MODE_COMPOSITE_P
20365 outermode, return NULL if the result doesn't encode back to the
20366 original byte sequence.
20367 (simplify_gen_subreg): Don't create SUBREGs from constants to
20368 MODE_COMPOSITE_P outermode.
20370 2021-04-12 Jakub Jelinek <jakub@redhat.com>
20372 PR rtl-optimization/99905
20373 * combine.c (expand_compound_operation): If pos + len > modewidth,
20374 perform the right shift by pos in inner_mode and then convert to mode,
20375 instead of trying to simplify a shift of rtx with inner_mode by pos
20376 as if it was a shift in mode.
20378 2021-04-12 Jakub Jelinek <jakub@redhat.com>
20381 * combine.c (simplify_and_const_int_1): Don't optimize varop
20382 away if it has side-effects.
20384 2021-04-12 Martin Liska <mliska@suse.cz>
20386 * doc/extend.texi: Escape @smallexample content.
20388 2021-04-12 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
20390 * config/s390/s390.md ("*movdi_31", "*movdi_64"): Add
20391 alternative in order to load a DFP zero.
20393 2021-04-12 Martin Liska <mliska@suse.cz>
20395 * doc/extend.texi: Be more precise in documentation
20396 of symver attribute.
20398 2021-04-12 Martin Liska <mliska@suse.cz>
20401 * gimplify.c (gimplify_expr): Right now, we unpoison all
20402 variables before a goto <dest>. We should not do it if we are
20405 2021-04-12 Cui,Lili <lili.cui@intel.com>
20407 * common/config/i386/cpuinfo.h (get_intel_cpu): Handle
20409 * common/config/i386/i386-common.c (processor_names): Add
20411 (processor_alias_table): Add rocketlake.
20412 * common/config/i386/i386-cpuinfo.h (processor_subtypes): Add
20413 INTEL_COREI7_ROCKETLAKE.
20414 * config.gcc: Add -march=rocketlake.
20415 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20417 * config/i386/i386-options.c (m_ROCKETLAKE) : Define.
20418 (processor_cost_table): Add rocketlake cost.
20419 * config/i386/i386.h (ix86_size_cost) : Define
20421 (processor_type) : Add PROCESSOR_ROCKETLAKE.
20422 (PTA_ROCKETLAKE): Ditto.
20423 * doc/extend.texi: Add rocketlake.
20424 * doc/invoke.texi: Add rocketlake.
20426 2021-04-12 Cui,Lili <lili.cui@intel.com>
20428 * config/i386/i386.h (PTA_ALDERLAKE): Change alderlake ISA list.
20429 * config/i386/i386-options.c (m_CORE_AVX2): Add m_ALDERLAKE.
20430 * common/config/i386/cpuinfo.h (get_intel_cpu): Add AlderLake model.
20431 * doc/invoke.texi: Change alderlake ISA list.
20433 2021-04-11 Hafiz Abid Qadeer <abidh@codesourcery.com>
20435 PR middle-end/98088
20436 * omp-expand.c (expand_oacc_collapse_init): Update condition in
20439 2021-04-10 H.J. Lu <hjl.tools@gmail.com>
20442 * config/i386/serializeintrin.h (_serialize): Defined as macro.
20444 2021-04-10 Jakub Jelinek <jakub@redhat.com>
20447 * expr.c (expand_expr_addr_expr_1): Test is_global_var rather than
20448 just TREE_STATIC on COMPOUND_LITERAL_EXPR_DECLs.
20450 2021-04-10 Jakub Jelinek <jakub@redhat.com>
20452 PR middle-end/99989
20453 * gimple-ssa-warn-alloca.c
20454 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit to
20455 0 with integer precision unconditionally.
20457 2021-04-10 Jakub Jelinek <jakub@redhat.com>
20459 PR rtl-optimization/98601
20460 * rtlanal.c (rtx_addr_can_trap_p_1): Allow in assert unknown size
20461 not just for BLKmode, but also for VOIDmode. For STRICT_ALIGNMENT
20462 unaligned_mems handle VOIDmode like BLKmode.
20464 2021-04-10 Jan Hubicka <hubicka@ucw.cz>
20467 * tree.c (free_lang_data_in_decl): Do not release body of
20468 declare_variant_alt.
20470 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
20472 * config/aarch64/aarch64.c (aarch64_option_restore): If the
20473 architecture was specified explicitly and the tuning wasn't,
20474 tune for the architecture rather than the configured default CPU.
20476 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
20478 * config/aarch64/aarch64.md (tlsdesc_small_sve_<mode>): Use X30
20479 as the temporary register.
20481 2021-04-09 Martin Liska <mliska@suse.cz>
20483 * doc/extend.texi: Move non-target attributes on the top level.
20485 2021-04-09 Martin Liska <mliska@suse.cz>
20487 * doc/invoke.texi: Document minimum and maximum value of the
20488 argument for both supported compression algorithms.
20490 2021-04-08 David Edelsohn <dje.gcc@gmail.com>
20492 * config/rs6000/rs6000.c (rs6000_xcoff_select_section): Select
20493 TLS BSS before TLS data.
20494 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .comm.
20496 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
20498 * doc/sourcebuild.texi (stdint_types_mbig_endian): Document.
20500 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
20502 * match.pd: Extend vec_cond folds to handle shifts.
20504 2021-04-08 Maciej W. Rozycki <macro@orcam.me.uk>
20506 * config/vax/vax.md: Fix comment for `*bit<mode>' pattern's
20509 2021-04-08 Alex Coplan <alex.coplan@arm.com>
20512 * config/arm/iterators.md (MVE_vecs): New.
20513 (V_elem): Also handle V2DF.
20514 * config/arm/mve.md (*mve_mov<mode>): Rename to ...
20515 (*mve_vdup<mode>): ... this. Remove second alternative since
20516 vec_duplicate of const_int is not canonical RTL, and we don't
20517 want to match symbol_refs.
20518 (*mve_vec_duplicate<mode>): Delete (pattern is redundant).
20520 2021-04-08 Xionghu Luo <luoxhu@linux.ibm.com>
20522 * fold-const.c (fold_single_bit_test): Fix typo.
20523 * print-rtl.c (print_rtx_insn_vec): Call print_rtl_single
20526 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
20528 PR tree-optimization/97513
20529 * tree-vect-slp.c (vect_add_slp_permutation): New function,
20531 (vectorizable_slp_permutation): ...here. Detect cases in which
20532 all VEC_PERM_EXPRs are guaranteed to have the same stepped
20533 permute vector and only generate one permute vector for that case.
20534 Extend that case to handle variable-length vectors.
20536 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
20538 PR tree-optimization/99873
20539 * tree-vect-slp.c (vect_slp_prefer_store_lanes_p): New function.
20540 (vect_build_slp_instance): Don't split store groups that could
20541 use IFN_STORE_LANES.
20543 2021-04-07 Jakub Jelinek <jakub@redhat.com>
20546 * varasm.c (output_constant_pool_contents): Don't strip name encoding
20547 from XSTR (desc->sym, 0) or from label before passing those to
20550 2021-04-07 Richard Biener <rguenther@suse.de>
20552 PR tree-optimization/99954
20553 * tree-loop-distribution.c: Include tree-affine.h.
20554 (generate_memcpy_builtin): Try using tree-affine to prove
20556 (loop_distribution::classify_builtin_ldst): Always classify
20559 2021-04-07 Richard Biener <rguenther@suse.de>
20561 PR tree-optimization/99947
20562 * tree-vect-loop.c (vectorizable_induction): Pre-allocate
20563 steps vector to avoid pushing elements from the reallocated
20566 2021-04-07 Richard Biener <rguenther@suse.de>
20568 * tree-ssa-sccvn.h (print_vn_reference_ops): Declare.
20569 * tree-ssa-pre.c (print_pre_expr): Factor out VN reference operand
20571 * tree-ssa-sccvn.c (print_vn_reference_ops): ... into this new
20573 (debug_vn_reference_ops): New.
20575 2021-04-07 Bin Cheng <bin.cheng@linux.alibaba.com>
20577 PR tree-optimization/98736
20578 * tree-loop-distribution.c
20579 * (loop_distribution::bb_top_order_init):
20580 Compute RPO with programing order preserved by calling function
20581 rev_post_order_and_mark_dfs_back_seme.
20583 2021-04-06 Vladimir N. Makarov <vmakarov@redhat.com>
20586 * lra-constraints.c (split_reg): Don't check paradoxical_subreg_p.
20587 * lra-lives.c (clear_sparseset_regnos, regnos_in_sparseset_p): New
20589 (process_bb_lives): Don't update biggest mode of hard reg for
20590 implicit in multi-register group. Use the new functions for
20591 updating dead_set and unused_set by register notes.
20593 2021-04-06 Xianmiao Qu <xianmiao_qu@c-sky.com>
20595 * config/csky/csky_pipeline_ck802.md : Use insn reservation name
20598 2021-04-06 H.J. Lu <hjl.tools@gmail.com>
20600 * config/i386/x86-tune-costs.h (skylake_memcpy): Updated.
20601 (skylake_memset): Likewise.
20602 (skylake_cost): Change CLEAR_RATIO to 17.
20603 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
20604 Replace m_CANNONLAKE, m_ICELAKE_CLIENT, m_ICELAKE_SERVER,
20605 m_TIGERLAKE and m_SAPPHIRERAPIDS with m_SKYLAKE and m_CORE_AVX512.
20607 2021-04-06 Richard Biener <rguenther@suse.de>
20609 PR tree-optimization/99880
20610 * tree-vect-loop.c (maybe_set_vectorized_backedge_value): Only
20611 set vectorized defs of relevant PHIs.
20613 2021-04-06 Richard Biener <rguenther@suse.de>
20615 PR tree-optimization/99924
20616 * tree-vect-slp.c (vect_bb_partition_graph_r): Do not mark
20617 nodes w/o scalar stmts as visited.
20619 2021-04-06 Alex Coplan <alex.coplan@arm.com>
20622 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Also use base
20623 PCS for [su]fix_optab.
20625 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
20627 * config/darwin.c (machopic_legitimize_pic_address): Check
20628 that the current pic register is one of the hard reg set
20629 before setting liveness.
20631 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
20633 * config/darwin.c (machopic_legitimize_pic_address): Fix
20634 whitespace, remove unused code.
20636 2021-04-03 Jakub Jelinek <jakub@redhat.com>
20638 PR tree-optimization/99882
20639 * gimple-ssa-store-merging.c (bswap_view_convert): Handle val with
20642 2021-04-03 Jakub Jelinek <jakub@redhat.com>
20644 PR rtl-optimization/99863
20645 * dse.c (replace_read): Drop regs_live argument. Instead of
20646 regs_live, use store_insn->fixed_regs_live if non-NULL,
20647 otherwise punt if insns sequence clobbers or sets any hard
20650 2021-04-03 Jakub Jelinek <jakub@redhat.com>
20653 * targhooks.h (default_print_patchable_function_entry_1): Declare.
20654 * targhooks.c (default_print_patchable_function_entry_1): New function,
20655 copied from default_print_patchable_function_entry with an added flags
20657 (default_print_patchable_function_entry): Rewritten into a small
20658 wrapper around default_print_patchable_function_entry_1.
20659 * config/rs6000/rs6000.c (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
20661 (rs6000_print_patchable_function_entry): New function.
20663 2021-04-02 Eric Botcazou <ebotcazou@adacore.com>
20665 * doc/invoke.texi (fdelete-dead-exceptions): Minor tweak.
20667 2021-04-01 Jason Merrill <jason@redhat.com>
20670 * common.opt: Document v15 and v16.
20672 2021-04-01 Richard Biener <rguenther@suse.de>
20674 PR tree-optimization/99863
20675 * gimplify.c (gimplify_init_constructor): Recompute vector
20678 2021-04-01 Jakub Jelinek <jakub@redhat.com>
20680 * doc/extend.texi (symver attribute): Fix up syntax errors
20683 2021-04-01 Jakub Jelinek <jakub@redhat.com>
20685 PR tree-optimization/96573
20686 * gimple-ssa-store-merging.c (init_symbolic_number): Handle
20687 also pointer types.
20689 2021-04-01 Richard Biener <rguenther@suse.de>
20691 PR tree-optimization/99856
20692 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Promote
20693 precision to vector element precision.
20695 2021-04-01 Martin Jambor <mjambor@suse.cz>
20697 PR tree-optimization/97009
20698 * tree-sra.c (access_or_its_child_written): New function.
20699 (propagate_subaccesses_from_rhs): Use it instead of a simple grp_write
20702 2021-03-31 Jan Hubicka <hubicka@ucw.cz>
20705 * cif-code.def (USES_COMDAT_LOCAL): Make CIF_FINAL_NORMAL.
20707 2021-03-31 Pat Haugen <pthaugen@linux.ibm.com>
20710 * config/rs6000/altivec.md (xxspltiw_v4si, xxspltiw_v4sf_inst,
20711 xxspltidp_v2df_inst, xxsplti32dx_v4si_inst, xxsplti32dx_v4sf_inst,
20712 xxblend_<mode>, xxpermx_inst, xxeval): Mark prefixed.
20713 * config/rs6000/mma.md (mma_<vvi4i4i8>, mma_<avvi4i4i8>,
20714 mma_<vvi4i4i2>, mma_<avvi4i4i2>, mma_<vvi4i4>, mma_<avvi4i4>,
20715 mma_<pvi4i2>, mma_<apvi4i2>, mma_<vvi4i4i4>, mma_<avvi4i4i4>):
20717 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Adjust test.
20718 * config/rs6000/rs6000.md (define_attr "maybe_prefixed"): New.
20719 (define_attr "prefixed"): Update initializer.
20721 2021-03-31 Jakub Jelinek <jakub@redhat.com>
20724 * dwarf2out.c (debug_ranges_dwo_section): New variable.
20725 (DW_RANGES_IDX_SKELETON): Define.
20726 (struct dw_ranges): Add begin_entry and end_entry members.
20727 (DEBUG_DWO_RNGLISTS_SECTION): Define.
20728 (add_ranges_num): Adjust r initializer for addition of *_entry
20730 (add_ranges_by_labels): For -gsplit-dwarf and force_direct,
20731 set idx to DW_RANGES_IDX_SKELETON.
20732 (use_distinct_base_address_for_range): New function.
20733 (index_rnglists): Don't set r->idx if it is equal to
20734 DW_RANGES_IDX_SKELETON. Initialize r->begin_entry and
20735 r->end_entry for -gsplit-dwarf if those will be needed by
20737 (output_rnglists): Add DWO argument. If true, switch to
20738 debug_ranges_dwo_section rather than debug_ranges_section.
20739 Adjust l1/l2 label indexes. Only output the offset table when
20740 dwo is true and don't include in there the skeleton range
20741 entry if present. For -gsplit-dwarf, skip ranges that belong
20742 to the other rnglists section. Change return type from void
20743 to bool and return true if there are any range entries for
20744 the other section. For dwarf_split_debug_info use
20745 DW_RLE_startx_endx, DW_RLE_startx_length and DW_RLE_base_addressx
20746 entries instead of DW_RLE_start_end, DW_RLE_start_length and
20747 DW_RLE_base_address. Use use_distinct_base_address_for_range.
20748 (init_sections_and_labels): Initialize debug_ranges_dwo_section
20749 if -gsplit-dwarf and DWARF >= 5. Adjust ranges_section_label
20750 and range_base_label indexes.
20751 (dwarf2out_finish): Call index_rnglists earlier before finalizing
20752 .debug_addr. Never emit DW_AT_rnglists_base attribute. For
20753 -gsplit-dwarf and DWARF >= 5 call output_rnglists up to twice
20754 with different dwo arguments.
20755 (dwarf2out_c_finalize): Clear debug_ranges_dwo_section.
20757 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
20759 PR tree-optimization/98268
20760 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Call
20761 recompute_tree_invariant_for_addr_expr after successfully
20762 folding a TARGET_MEM_REF that occurs inside an ADDR_EXPR.
20764 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
20766 PR tree-optimization/99726
20767 * tree-data-ref.c (create_intersect_range_checks_index): Bail
20768 out if there is more than one access function SCEV for the loop
20771 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
20773 PR rtl-optimization/97141
20774 PR rtl-optimization/98726
20775 * emit-rtl.c (valid_for_const_vector_p): Return true for
20777 * rtx-vector-builder.h (rtx_vector_builder::step): Return a
20778 poly_wide_int instead of a wide_int.
20779 (rtx_vector_builder::apply_set): Take a poly_wide_int instead
20781 * rtx-vector-builder.c (rtx_vector_builder::apply_set): Likewise.
20782 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Return
20783 false for CONST_VECTORs that cannot be forced to memory.
20784 * config/aarch64/aarch64-simd.md (mov<mode>): If a CONST_VECTOR
20785 is too complex to force to memory, build it up from individual
20788 2021-03-31 Jan Hubicka <jh@suse.cz>
20791 * cgraph.c (cgraph_node::release_body): Fix overactive check.
20793 2021-03-31 Christophe Lyon <christophe.lyon@linaro.org>
20796 * config/arm/vec-common.md (mul<mode>3): Disable on iwMMXT, expect
20799 2021-03-31 H.J. Lu <hjl.tools@gmail.com>
20801 * config/i386/i386-expand.c (expand_set_or_cpymem_via_rep):
20802 For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, don't convert QImode
20804 (decide_alg): For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, use
20805 "rep movsb/stosb" only for known sizes.
20806 * config/i386/i386-options.c (processor_cost_table): Use Ice
20807 Lake cost for Cannon Lake, Ice Lake, Tiger Lake, Sapphire
20808 Rapids and Alder Lake.
20809 * config/i386/i386.h (TARGET_PREFER_KNOWN_REP_MOVSB_STOSB): New.
20810 * config/i386/x86-tune-costs.h (icelake_memcpy): New.
20811 (icelake_memset): Likewise.
20812 (icelake_cost): Likewise.
20813 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
20816 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
20819 * config/aarch64/aarch64.c
20820 (aarch64_vectorize_preferred_vector_alignment): Query the size
20821 of the provided SVE vector; do not assume that all SVE vectors
20822 have the same size.
20824 2021-03-31 Jan Hubicka <jh@suse.cz>
20827 * cgraph.c (cgraph_node::release_body): Remove all callers and
20829 * cgraphclones.c (cgraph_node::materialize_clone): Do not do it here.
20830 * cgraphunit.c (cgraph_node::expand): And here.
20832 2021-03-31 Martin Liska <mliska@suse.cz>
20834 * ipa-modref.c (analyze_ssa_name_flags): Fix coding style
20835 and one negated condition.
20837 2021-03-31 Jakub Jelinek <jakub@redhat.com>
20838 Richard Sandiford <richard.sandiford@arm.com>
20841 * config/aarch64/aarch64.md (*add<mode>3_poly_1): Swap Uai and Uav
20842 constraints on operands[2] and similarly 0 and rk constraints
20843 on operands[1] corresponding to that.
20845 2021-03-31 Jakub Jelinek <jakub@redhat.com>
20848 * configure.ac (HAVE_LD_BROKEN_PE_DWARF5): New AC_DEFINE if PECOFF
20849 linker doesn't support DWARF sections new in DWARF5.
20850 * config/i386/i386-options.c (ix86_option_override_internal): Default
20851 to dwarf_version 4 if HAVE_LD_BROKEN_PE_DWARF5 for TARGET_PECOFF
20853 * config.in: Regenerated.
20854 * configure: Regenerated.
20856 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20859 * config/aarch64/aarch64.c (aarch64_analyze_loop_vinfo): Check for
20860 available issue_info before using it.
20862 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20865 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): Do not allow zero
20868 2021-03-30 Xionghu Luo <luoxhu@linux.ibm.com>
20871 * config/rs6000/altivec.md (altivec_lvsl_reg): Change to ...
20872 (altivec_lvsl_reg_<mode>): ... this.
20873 (altivec_lvsr_reg): Change to ...
20874 (altivec_lvsr_reg_<mode>): ... this.
20875 * config/rs6000/predicates.md (vec_set_index_operand): New.
20876 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20877 Enable 32bit variable vec_insert for all TARGET_VSX.
20878 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
20879 Enable 32bit variable vec_insert for p9 and above.
20880 (rs6000_expand_vector_set_var_p8): Rename to ...
20881 (rs6000_expand_vector_set_var_p7): ... this.
20882 (rs6000_expand_vector_set): Use TARGET_VSX and adjust assert
20884 * config/rs6000/vector.md (vec_set<mode>): Use vec_set_index_operand.
20885 * config/rs6000/vsx.md (xl_len_r): Use gen_altivec_lvsl_reg_di and
20886 gen_altivec_lvsr_reg_di.
20888 2021-03-30 H.J. Lu <hjl.tools@gmail.com>
20891 * config/i386/ia32intrin.h (__rdtsc): Defined as macro.
20892 (__rdtscp): Likewise.
20894 2021-03-30 Tamar Christina <tamar.christina@arm.com>
20896 PR tree-optimization/99825
20897 * tree-vect-slp-patterns.c (vect_check_evenodd_blend):
20898 Reject non-mult 2 lanes.
20900 2021-03-30 Richard Earnshaw <rearnsha@arm.com>
20903 * config/arm/arm.c (arm_file_start): Fix emission of
20904 Tag_ABI_VFP_args attribute.
20906 2021-03-30 Richard Biener <rguenther@suse.de>
20908 PR tree-optimization/99824
20909 * stor-layout.c (set_min_and_max_values_for_integral_type):
20910 Assert the precision is within the bounds of
20911 WIDE_INT_MAX_PRECISION.
20912 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
20913 the outermost component ref only to lower the access size
20914 and initialize that from the access type.
20916 2021-03-30 Richard Sandiford <richard.sandiford@arm.com>
20919 * config/aarch64/aarch64.md (mov<mode>): Pass multi-instruction
20920 CONST_INTs to aarch64_expand_mov_immediate when called after RA.
20922 2021-03-30 Mihailo Stojanovic <mihailo.stojanovic@typhoon-hil.com>
20924 * config/aarch64/aarch64.md
20925 (<optab>_trunc<fcvt_target><GPI:mode>2): Set the "arch"
20926 attribute to disambiguate between SIMD and FP variants of the
20929 2021-03-29 Jan Hubicka <hubicka@ucw.cz>
20931 * ipa-modref.c (merge_call_lhs_flags): Correct handling of deref.
20932 (analyze_ssa_name_flags): Fix typo in comment.
20934 2021-03-29 Alex Coplan <alex.coplan@arm.com>
20937 * config/aarch64/aarch64-sve-builtins.cc
20938 (function_builder::add_function): Add placeholder_p argument, use
20939 placeholder decls if this is set.
20940 (function_builder::add_unique_function): Instead of conditionally adding
20941 direct overloads, unconditionally add either a direct overload or a
20943 (function_builder::add_overloaded_function): Set placeholder_p if we're
20944 using C++ overloads. Use the obstack for string storage instead
20945 of relying on the tree nodes.
20946 (function_builder::add_overloaded_functions): Don't return early for
20947 m_direct_overloads: we need to add placeholders.
20948 * config/aarch64/aarch64-sve-builtins.h
20949 (function_builder::add_function): Add placeholder_p argument.
20951 2021-03-29 Richard Biener <rguenther@suse.de>
20953 PR tree-optimization/99807
20954 * tree-vect-slp.c (vect_slp_analyze_node_operations_1): Move
20955 assert below VEC_PERM handling.
20957 2021-03-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20960 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>): Use
20961 aarch64_simd_or_scalar_imm_zero to match zeroes. Remove pattern
20962 matching const_int 0.
20963 (move_lo_quad_internal_be_<mode>): Likewise.
20964 (move_lo_quad_<mode>): Update for the above.
20965 * config/aarch64/iterators.md (VQ_2E): Delete.
20967 2021-03-29 Jakub Jelinek <jakub@redhat.com>
20969 PR tree-optimization/99777
20970 * fold-const.c (extract_muldiv_1): For conversions, punt on casts from
20971 types other than scalar integral types.
20973 2021-03-28 David Edelsohn <dje.gcc@gmail.com>
20975 * config/rs6000/rs6000.c (rs6000_output_dwarf_dtprel): Do not add
20976 XCOFF TLS reloc decorations.
20978 2021-03-28 Gerald Pfeifer <gerald@pfeifer.com>
20980 * doc/analyzer.texi (Analyzer Internals): Update link to
20981 "A Memory Model for Static Analysis of C Programs".
20983 2021-03-26 David Edelsohn <dje.gcc@gmail.com>
20985 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Call function.
20986 * config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align):
20988 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align): New.
20989 (rs6000_special_round_type_align): Recursively check innermost first
20992 2021-03-26 Jakub Jelinek <jakub@redhat.com>
20995 * dwarf2out.h (struct dw_fde_node): Add rule18 member.
20996 * dwarf2cfi.c (dwarf2out_frame_debug_expr): When handling (set hfp sp)
20997 assignment with drap_reg active, queue reg save for hfp with offset 0
20998 and flush queued reg saves. When handling a push with rule18,
20999 defer queueing reg save for hfp and just assert the offset is 0.
21000 (scan_trace): Assert that fde->rule18 is false.
21002 2021-03-26 Vladimir Makarov <vmakarov@redhat.com>
21005 * ira-costs.c (record_reg_classes): Put case with
21006 CT_RELAXED_MEMORY adjacent to one with CT_MEMORY.
21007 * ira.c (ira_setup_alts): Ditto.
21008 * lra-constraints.c (process_alt_operands): Ditto.
21009 * recog.c (asm_operand_ok): Ditto.
21010 * reload.c (find_reloads): Ditto.
21012 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21014 * config/aarch64/aarch64-protos.h
21015 (cpu_addrcost_table::post_modify_ld3_st3): New member variable.
21016 (cpu_addrcost_table::post_modify_ld4_st4): Likewise.
21017 * config/aarch64/aarch64.c (generic_addrcost_table): Update
21018 accordingly, using the same costs as for post_modify.
21019 (exynosm1_addrcost_table, xgene1_addrcost_table): Likewise.
21020 (thunderx2t99_addrcost_table, thunderx3t110_addrcost_table):
21021 (tsv110_addrcost_table, qdf24xx_addrcost_table): Likewise.
21022 (a64fx_addrcost_table): Likewise.
21023 (neoversev1_addrcost_table): New.
21024 (neoversev1_tunings): Use neoversev1_addrcost_table.
21025 (aarch64_address_cost): Use the new post_modify costs for CImode
21028 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21030 * config/aarch64/aarch64.opt
21031 (-param=aarch64-loop-vect-issue-rate-niters=): New parameter.
21032 * doc/invoke.texi: Document it.
21033 * config/aarch64/aarch64-protos.h (aarch64_base_vec_issue_info)
21034 (aarch64_scalar_vec_issue_info, aarch64_simd_vec_issue_info)
21035 (aarch64_advsimd_vec_issue_info, aarch64_sve_vec_issue_info)
21036 (aarch64_vec_issue_info): New structures.
21037 (cpu_vector_cost): Write comments above the variables rather
21039 (cpu_vector_cost::issue_info): New member variable.
21040 * config/aarch64/aarch64.c: Include gimple-pretty-print.h
21041 and tree-ssa-loop-niter.h.
21042 (generic_vector_cost, a64fx_vector_cost, qdf24xx_vector_cost)
21043 (thunderx_vector_cost, tsv110_vector_cost, cortexa57_vector_cost)
21044 (exynosm1_vector_cost, xgene1_vector_cost, thunderx2t99_vector_cost)
21045 (thunderx3t110_vector_cost): Initialize issue_info to null.
21046 (neoversev1_scalar_issue_info, neoversev1_advsimd_issue_info)
21047 (neoversev1_sve_issue_info, neoversev1_vec_issue_info): New structures.
21048 (neoversev1_vector_cost): Use them.
21049 (aarch64_vec_op_count, aarch64_sve_op_count): New structures.
21050 (aarch64_vector_costs::saw_sve_only_op): New member variable.
21051 (aarch64_vector_costs::num_vector_iterations): Likewise.
21052 (aarch64_vector_costs::scalar_ops): Likewise.
21053 (aarch64_vector_costs::advsimd_ops): Likewise.
21054 (aarch64_vector_costs::sve_ops): Likewise.
21055 (aarch64_vector_costs::seen_loads): Likewise.
21056 (aarch64_simd_vec_costs_for_flags): New function.
21057 (aarch64_analyze_loop_vinfo): Initialize num_vector_iterations.
21058 Count the number of predicate operations required by SVE WHILE
21060 (aarch64_comparison_type, aarch64_multiply_add_p): New functions.
21061 (aarch64_sve_only_stmt_p, aarch64_in_loop_reduction_latency): Likewise.
21062 (aarch64_count_ops): Likewise.
21063 (aarch64_add_stmt_cost): Record whether see an SVE operation
21064 that cannot currently be implementing using Advanced SIMD.
21065 Record issue information about the scalar, Advanced SIMD
21066 and (where relevant) SVE versions of a loop.
21067 (aarch64_vec_op_count::dump): New function.
21068 (aarch64_sve_op_count::dump): Likewise.
21069 (aarch64_estimate_min_cycles_per_iter): Likewise.
21070 (aarch64_adjust_body_cost): If issue information is available,
21071 try to compare the issue rates of the various loop implementations
21072 and increase or decrease the vector body cost accordingly.
21074 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21076 * config/aarch64/aarch64.c (aarch64_detect_vector_stmt_subtype):
21077 Assume a zero cost for induction phis.
21079 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21081 * config/aarch64/aarch64.c (aarch64_embedded_comparison_type): New
21083 (aarch64_adjust_stmt_cost): Add the costs of embedded scalar and
21084 vector comparisons.
21086 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21088 * config/aarch64/aarch64.c (aarch64_detect_scalar_stmt_subtype):
21090 (aarch64_add_stmt_cost): Call it.
21092 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21094 * config/aarch64/aarch64-tuning-flags.def (matched_vector_throughput):
21095 New tuning parameter.
21096 * config/aarch64/aarch64.c (neoversev1_tunings): Use it.
21097 (aarch64_estimated_sve_vq): New function.
21098 (aarch64_vector_costs::analyzed_vinfo): New member variable.
21099 (aarch64_vector_costs::is_loop): Likewise.
21100 (aarch64_vector_costs::unrolled_advsimd_niters): Likewise.
21101 (aarch64_vector_costs::unrolled_advsimd_stmts): Likewise.
21102 (aarch64_record_potential_advsimd_unrolling): New function.
21103 (aarch64_analyze_loop_vinfo, aarch64_analyze_bb_vinfo): Likewise.
21104 (aarch64_add_stmt_cost): Call aarch64_analyze_loop_vinfo or
21105 aarch64_analyze_bb_vinfo on the first use of a costs structure.
21106 Detect whether we're vectorizing a loop for SVE that might be
21107 completely unrolled if it used Advanced SIMD instead.
21108 (aarch64_adjust_body_cost_for_latency): New function.
21109 (aarch64_finish_cost): Call it.
21111 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21113 * config/aarch64/aarch64.c (aarch64_vector_costs): New structure.
21114 (aarch64_init_cost): New function.
21115 (aarch64_add_stmt_cost): Use aarch64_vector_costs instead of
21116 the default unsigned[3].
21117 (aarch64_finish_cost, aarch64_destroy_cost_data): New functions.
21118 (TARGET_VECTORIZE_INIT_COST): Override.
21119 (TARGET_VECTORIZE_FINISH_COST): Likewise.
21120 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
21122 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21124 * config/aarch64/aarch64.c (neoversev1_advsimd_vector_cost)
21125 (neoversev1_sve_vector_cost): New cost structures.
21126 (neoversev1_vector_cost): Likewise.
21127 (neoversev1_tunings): Use them. Enable use_new_vector_costs.
21129 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21131 * config/aarch64/aarch64-protos.h
21132 (sve_vec_cost::scatter_store_elt_cost): New member variable.
21133 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
21134 accordingly, taking the cost from the cost of a scalar_store.
21135 (a64fx_sve_vector_cost): Likewise.
21136 (aarch64_detect_vector_stmt_subtype): Detect scatter stores.
21138 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21140 * config/aarch64/aarch64-protos.h
21141 (simd_vec_cost::store_elt_extra_cost): New member variable.
21142 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
21143 accordingly, using the vec_to_scalar cost for the new field.
21144 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
21145 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
21146 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
21147 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
21148 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
21149 (thunderx3t110_advsimd_vector_cost): Likewise.
21150 (aarch64_detect_vector_stmt_subtype): Detect single-element stores.
21152 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21154 * config/aarch64/aarch64-protos.h (simd_vec_cost::ld2_st2_permute_cost)
21155 (simd_vec_cost::ld3_st3_permute_cost): New member variables.
21156 (simd_vec_cost::ld4_st4_permute_cost): Likewise.
21157 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
21158 accordingly, using zero for the new costs.
21159 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
21160 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
21161 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
21162 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
21163 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
21164 (thunderx3t110_advsimd_vector_cost): Likewise.
21165 (aarch64_ld234_st234_vectors): New function.
21166 (aarch64_adjust_stmt_cost): Likewise.
21167 (aarch64_add_stmt_cost): Call aarch64_adjust_stmt_cost if using
21168 the new vector costs.
21170 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21172 * config/aarch64/aarch64-protos.h (sve_vec_cost): Turn into a
21173 derived class of simd_vec_cost. Add information about CLAST[AB]
21174 and FADDA instructions.
21175 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
21176 accordingly, using the vec_to_scalar costs for the new fields.
21177 (a64fx_sve_vector_cost): Likewise.
21178 (aarch64_reduc_type): New function.
21179 (aarch64_sve_in_loop_reduction_latency): Likewise.
21180 (aarch64_detect_vector_stmt_subtype): Take a vinfo parameter.
21181 Use aarch64_sve_in_loop_reduction_latency to handle SVE reductions
21182 that occur in the loop body.
21183 (aarch64_add_stmt_cost): Update call accordingly.
21185 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
21187 * config/aarch64/aarch64-tuning-flags.def (use_new_vector_costs):
21189 * config/aarch64/aarch64-protos.h (simd_vec_cost): Put comments
21190 above the fields rather than to the right.
21191 (simd_vec_cost::reduc_i8_cost): New member variable.
21192 (simd_vec_cost::reduc_i16_cost): Likewise.
21193 (simd_vec_cost::reduc_i32_cost): Likewise.
21194 (simd_vec_cost::reduc_i64_cost): Likewise.
21195 (simd_vec_cost::reduc_f16_cost): Likewise.
21196 (simd_vec_cost::reduc_f32_cost): Likewise.
21197 (simd_vec_cost::reduc_f64_cost): Likewise.
21198 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
21199 accordingly, using the vec_to_scalar_cost for the new fields.
21200 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
21201 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
21202 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
21203 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
21204 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
21205 (thunderx3t110_advsimd_vector_cost): Likewise.
21206 (aarch64_use_new_vector_costs_p): New function.
21207 (aarch64_simd_vec_costs): New function, split out from...
21208 (aarch64_builtin_vectorization_cost): ...here.
21209 (aarch64_is_reduction): New function.
21210 (aarch64_detect_vector_stmt_subtype): Likewise.
21211 (aarch64_add_stmt_cost): Call aarch64_detect_vector_stmt_subtype if
21212 using the new vector costs.
21214 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
21217 * tree-emutls.c (get_emutls_init_templ_addr): Mark initializer of weak
21218 TLS declarations as public.
21220 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
21222 * config/aarch64/aarch64-d.c (IN_TARGET_CODE): Define.
21223 * config/arm/arm-d.c (IN_TARGET_CODE): Likewise.
21224 * config/i386/i386-d.c (IN_TARGET_CODE): Likewise.
21225 * config/mips/mips-d.c (IN_TARGET_CODE): Likewise.
21226 * config/pa/pa-d.c (IN_TARGET_CODE): Likewise.
21227 * config/riscv/riscv-d.c (IN_TARGET_CODE): Likewise.
21228 * config/rs6000/rs6000-d.c (IN_TARGET_CODE): Likewise.
21229 * config/s390/s390-d.c (IN_TARGET_CODE): Likewise.
21230 * config/sparc/sparc-d.c (IN_TARGET_CODE): Likewise.
21232 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
21235 * config.gcc (*-*-cygwin*): Add winnt-d.o
21236 (*-*-mingw*): Likewise.
21237 * config/i386/cygwin.h (EXTRA_TARGET_D_OS_VERSIONS): New macro.
21238 * config/i386/mingw32.h (EXTRA_TARGET_D_OS_VERSIONS): Likewise.
21239 * config/i386/t-cygming: Add winnt-d.o.
21240 * config/i386/winnt-d.c: New file.
21242 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
21244 * config/freebsd-d.c: Include memmodel.h.
21246 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
21249 * config.gcc (*-*-openbsd*): Add openbsd-d.o.
21250 * config/t-openbsd: Add openbsd-d.o.
21251 * config/openbsd-d.c: New file.
21253 2021-03-25 Stam Markianos-Wright <stam.markianos-wright@arm.com>
21255 PR tree-optimization/96974
21256 * tree-vect-stmts.c (vect_get_vector_types_for_stmt): Replace assert
21257 with graceful exit.
21259 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
21262 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
21266 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
21267 always_inline in system headers.
21269 2021-03-25 Kewen Lin <linkw@linux.ibm.com>
21271 * tree-vect-loop.c (vect_model_reduction_cost): Init inside_cost.
21273 2021-03-25 Jakub Jelinek <jakub@redhat.com>
21276 * tree-core.h (enum operand_equal_flag): Add OEP_ADDRESS_OF_SAME_FIELD.
21277 * fold-const.c (operand_compare::operand_equal_p): Don't compare
21278 field offsets if OEP_ADDRESS_OF_SAME_FIELD.
21280 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
21284 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
21285 always_inline in system headers.
21287 2021-03-25 Richard Biener <rguenther@suse.de>
21289 PR tree-optimization/99746
21290 * tree-vect-slp-patterns.c (complex_pattern::build): Do not mark
21291 the scalar stmt as patterned. Instead set up required things
21294 2021-03-25 Xionghu Luo <luoxhu@linux.ibm.com>
21296 * config/rs6000/rs6000.c (power8_costs): Change l2 cache
21299 2021-03-24 Martin Liska <mliska@suse.cz>
21302 * common/config/i386/i386-common.c (ARRAY_SIZE): Fix off-by-one
21304 * config/i386/i386-options.c (ix86_option_override_internal):
21305 Add run-time assert.
21307 2021-03-24 Martin Jambor <mjambor@suse.cz>
21310 * ipa-cp.c (initialize_node_lattices): Mark as bottom all
21311 parameters with unknown type.
21312 (ipacp_value_safe_for_type): New function.
21313 (propagate_vals_across_arith_jfunc): Verify that the constant type
21314 can be used for a type of the formal parameter.
21315 (propagate_vals_across_ancestor): Likewise.
21316 (propagate_scalar_across_jump_function): Likewise. Pass the type
21317 also to propagate_vals_across_ancestor.
21319 2021-03-24 Christophe Lyon <christophe.lyon@linaro.org>
21322 * config/arm/mve.md (movmisalign<mode>_mve_store): Use Ux
21324 (movmisalign<mode>_mve_load): Likewise.
21326 2021-03-24 Jakub Jelinek <jakub@redhat.com>
21329 * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
21330 movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.
21332 2021-03-24 Alexandre Oliva <oliva@adacore.com>
21334 * doc/sourcebuild.texi (sysconf): New effective target.
21336 2021-03-24 Alexandre Oliva <oliva@adacore.com>
21338 * config/i386/predicates.md (reg_or_const_vec_operand): New.
21339 * config/i386/sse.md (ssse3_pshufbv8qi3): Add an expander for
21340 the now *-prefixed insn_and_split, turn the splitter const vec
21341 into an input for the insn, making it an ignored immediate for
21342 non-split cases, and loaded into the scratch register
21345 2021-03-23 Vladimir N. Makarov <vmakarov@redhat.com>
21348 * config/aarch64/constraints.md (Utq, UOb, UOh, UOw, UOd, UOty):
21349 Use define_relaxed_memory_constraint for them.
21351 2021-03-23 Iain Sandoe <iain@sandoe.co.uk>
21354 * config/host-darwin.c (darwin_gt_pch_use_address): Add a
21355 colon to the diagnostic message.
21357 2021-03-23 Ilya Leoshkevich <iii@linux.ibm.com>
21359 * fwprop.c (fwprop_propagation::fwprop_propagation): Look at
21361 (try_fwprop_subst_note): Use set_info instead of insn_info.
21362 (try_fwprop_subst_pattern): Likewise.
21363 (try_fwprop_subst_notes): Likewise.
21364 (try_fwprop_subst): Likewise.
21365 (forward_propagate_subreg): Likewise.
21366 (forward_propagate_and_simplify): Likewise.
21367 (forward_propagate_into): Likewise.
21368 * rtl-ssa/accesses.h (set_info::single_nondebug_use) New
21370 (set_info::single_nondebug_insn_use): Likewise.
21371 (set_info::single_phi_use): Likewise.
21372 * rtl-ssa/member-fns.inl (set_info::single_nondebug_use) New
21374 (set_info::single_nondebug_insn_use): Likewise.
21375 (set_info::single_phi_use): Likewise.
21377 2021-03-23 Christophe Lyon <christophe.lyon@linaro.org>
21379 * doc/sourcebuild.texi (arm_dsp_ok, arm_dsp): Document.
21381 2021-03-23 Jakub Jelinek <jakub@redhat.com>
21384 * config/aarch64/aarch64.c (aarch64_add_offset): Tell
21385 expand_mult to perform an unsigned rather than a signed
21388 2021-03-23 H.J. Lu <hjl.tools@gmail.com>
21391 * config/i386/cpuid.h (__cpuid): Add __volatile__.
21392 (__cpuid_count): Likewise.
21394 2021-03-23 Richard Biener <rguenther@suse.de>
21396 PR tree-optimization/99721
21397 * tree-vect-slp.c (vect_slp_analyze_node_operations):
21398 Make sure we can schedule the node.
21400 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
21402 * config/riscv/riscv.c (riscv_subword): Take endianness into
21403 account when calculating the byte offset.
21405 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
21407 * config/riscv/predicates.md (subreg_lowpart_operator): New predicate
21408 * config/riscv/riscv.md (*addsi3_extended2, *subsi3_extended2)
21409 (*negsi2_extended2, *mulsi3_extended2, *<optab>si3_mask)
21410 (*<optab>si3_mask_1, *<optab>di3_mask, *<optab>di3_mask_1)
21411 (*<optab>si3_extend_mask, *<optab>si3_extend_mask_1): Use
21412 new predicate "subreg_lowpart_operator"
21414 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
21416 * config/riscv/riscv.c (riscv_swap_instruction): New function
21417 to byteswap an SImode rtx containing an instruction.
21418 (riscv_trampoline_init): Byteswap the generated instructions
21421 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
21423 * common/config/riscv/riscv-common.c
21424 (TARGET_DEFAULT_TARGET_FLAGS): Set default endianness.
21425 * config.gcc (riscv32be-*, riscv64be-*): Set
21426 TARGET_BIG_ENDIAN_DEFAULT to 1.
21427 * config/riscv/elf.h (LINK_SPEC): Change -melf* value
21428 depending on default endianness.
21429 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
21430 * config/riscv/linux.h (LINK_SPEC): Likewise.
21431 * config/riscv/riscv.c (TARGET_DEFAULT_TARGET_FLAGS): Set
21432 default endianness.
21433 * config/riscv/riscv.h (DEFAULT_ENDIAN_SPEC): New macro.
21435 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
21437 * config/riscv/elf.h (LINK_SPEC): Pass linker endianness flag.
21438 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
21439 * config/riscv/linux.h (LINK_SPEC): Likewise.
21440 * config/riscv/riscv.h (ASM_SPEC): Pass -mbig-endian and
21442 (BYTES_BIG_ENDIAN): Handle big endian.
21443 (WORDS_BIG_ENDIAN): Define to BYTES_BIG_ENDIAN.
21444 * config/riscv/riscv.opt (-mbig-endian, -mlittle-endian): New
21446 * doc/invoke.texi (-mbig-endian, -mlittle-endian): Document.
21448 2021-03-23 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
21450 * regcprop.c (find_oldest_value_reg): Ask target whether
21451 different mode is fine for replacement register.
21453 2021-03-23 Aldy Hernandez <aldyh@redhat.com>
21455 PR tree-optimization/99296
21456 * value-range.cc (irange::irange_set_1bit_anti_range): New.
21457 (irange::irange_set_anti_range): Call irange_set_1bit_anti_range
21458 * value-range.h (irange::irange_set_1bit_anti_range): New.
21460 2021-03-22 Vladimir N. Makarov <vmakarov@redhat.com>
21463 * config/aarch64/constraints.md (UtQ): Use
21464 define_relaxed_memory_constraint for it.
21465 * doc/md.texi (define_relaxed_memory_constraint): Describe it.
21466 * genoutput.c (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
21467 * genpreds.c (constraint_data): Add bitfield is_relaxed_memory.
21468 (have_relaxed_memory_constraints): New static var.
21469 (relaxed_memory_start, relaxed_memory_end): Ditto.
21470 (add_constraint): Add arg is_relaxed_memory. Check name for
21471 relaxed memory. Set up is_relaxed_memory in constraint_data and
21472 have_relaxed_memory_constraints. Adjust calls.
21473 (choose_enum_order): Process relaxed memory.
21474 (write_tm_preds_h): Ditto.
21475 (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
21476 * gensupport.c (process_rtx): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
21477 * ira-costs.c (record_reg_classes): Process CT_RELAXED_MEMORY.
21478 * ira-lives.c (single_reg_class): Use
21479 insn_extra_relaxed_memory_constraint.
21480 * ira.c (ira_setup_alts): CT_RELAXED_MEMORY.
21481 * lra-constraints.c (valid_address_p): Use
21482 insn_extra_relaxed_memory_constraint instead of other memory
21484 (process_alt_operands): Process CT_RELAXED_MEMORY.
21485 (curr_insn_transform): Use insn_extra_relaxed_memory_constraint.
21486 * recog.c (asm_operand_ok, preprocess_constraints): Process
21488 * reload.c (find_reloads): Ditto.
21489 * rtl.def (DEFINE_RELAXED_MEMORY_CONSTRAINT): New.
21490 * stmt.c (parse_input_constraint): Use
21491 insn_extra_relaxed_memory_constraint.
21493 2021-03-22 Segher Boessenkool <segher@kernel.crashing.org>
21496 * ubsan.c (ubsan_instrument_float_cast): Don't test for unordered if
21499 2021-03-22 Alex Coplan <alex.coplan@arm.com>
21502 * config/arm/arm-protos.h (neon_make_constant): Add generate
21503 argument to guard emitting insns, default to true.
21504 * config/arm/arm.c (arm_legitimate_constant_p_1): Reject
21505 CONST_VECTORs which neon_make_constant can't handle.
21506 (neon_vdup_constant): Add generate argument, avoid emitting
21507 insns if it's not set.
21508 (neon_make_constant): Plumb new generate argument through.
21509 * config/arm/constraints.md (Ui): New. Use it...
21510 * config/arm/mve.md (*mve_mov<mode>): ... here.
21511 * config/arm/vec-common.md (movv8hf): Use neon_make_constant to
21512 synthesize constants.
21514 2021-03-22 Richard Biener <rguenther@suse.de>
21516 * debug.h: Add deprecation warning.
21518 2021-03-22 Richard Biener <rguenther@suse.de>
21520 PR tree-optimization/99694
21521 * tree-ssa-sccvn.c (visit_phi): Ignore edges with the
21524 2021-03-22 Kito Cheng <kito.cheng@sifive.com>
21527 * config/riscv/riscv.c (riscv_expand_block_move): Get RTL value
21528 after type checking.
21530 2021-03-22 Jakub Jelinek <jakub@redhat.com>
21534 * dwarf2out.c (get_full_len): Use get_precision rather than
21536 (add_const_value_attribute): Make sure add_AT_wide argument has
21537 precision prec rather than some very wide one.
21539 2021-03-22 Kewen Lin <linkw@linux.ibm.com>
21541 * config/rs6000/rs6000.md (*rotldi3_insert_sf,
21542 *mov<SFDF:mode><SFDF2:mode>cc_p9, floatsi<mode>2_lfiwax,
21543 floatsi<mode>2_lfiwax_mem, floatunssi<mode>2_lfiwzx,
21544 floatunssi<mode>2_lfiwzx_mem, *floatsidf2_internal,
21545 *floatunssidf2_internal, fix_trunc<mode>si2_stfiwx,
21546 fix_trunc<mode>si2_internal, fixuns_trunc<mode>si2_stfiwx,
21547 *round32<mode>2_fprs, *roundu32<mode>2_fprs,
21548 *fix_trunc<mode>si2_internal): Fix empty split condition.
21549 * config/rs6000/vsx.md (*vsx_le_undo_permute_<mode>,
21550 vsx_reduc_<VEC_reduc_name>_v2df, vsx_reduc_<VEC_reduc_name>_v4sf,
21551 *vsx_reduc_<VEC_reduc_name>_v2df_scalar,
21552 *vsx_reduc_<VEC_reduc_name>_v4sf_scalar): Likewise.
21554 2021-03-22 Xionghu Luo <luoxhu@linux.ibm.com>
21557 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
21558 Convert idx to DImode.
21559 (rs6000_expand_vector_set_var_p8): Likewise.
21561 2021-03-21 Jakub Jelinek <jakub@redhat.com>
21564 * dwarf2out.c (insert_float): Change return type from void to
21565 unsigned, handle GET_MODE_SIZE (mode) == 2 and return element size.
21566 (mem_loc_descriptor, loc_descriptor, add_const_value_attribute):
21569 2021-03-20 H.J. Lu <hjl.tools@gmail.com>
21572 * config/i386/i386.c (construct_container): Check cfun != NULL
21573 before accessing silent_p.
21575 2021-03-20 Ahamed Husni <ahamedhusni73@gmail.com>
21577 * asan.c: Fix typos in comments.
21579 2021-03-20 Vladimir N. Makarov <vmakarov@redhat.com>
21581 PR rtl-optimization/99680
21582 * lra-constraints.c (skip_contraint_modifiers): Rename to skip_constraint_modifiers.
21583 (process_address_1): Check empty constraint before using
21586 2021-03-19 Pat Haugen <pthaugen@linux.ibm.com>
21588 * config/rs6000/rs6000.c (power10_cost): New.
21589 (rs6000_option_override_internal): Set Power10 costs.
21590 (rs6000_issue_rate): Set Power10 issue rate.
21591 * config/rs6000/power10.md: Rewrite for Power10.
21593 2021-03-19 Vladimir N. Makarov <vmakarov@redhat.com>
21596 * lra-constraints.c (process_address_1): Don't use unknown
21597 constraint for address constraint.
21599 2021-03-19 Iain Sandoe <iain@sandoe.co.uk>
21602 * config.gcc (powerpc-*-darwin8): Delete the reference to
21603 the now removed darwin8.h.
21605 2021-03-19 Olivier Hainque <hainque@adacore.com>
21608 * config/vxworksae.h (VX_CPU_PREFIX): Define.
21610 2021-03-19 John David Anglin <danglin@gcc.gnu.org>
21612 * config/pa/pa.c (import_milli): Use memcpy instead of strncpy.
21614 2021-03-19 Tamar Christina <tamar.christina@arm.com>
21616 PR tree-optimization/99656
21617 * tree-vect-slp-patterns.c (linear_loads_p,
21618 complex_add_pattern::matches, is_eq_or_top,
21619 vect_validate_multiplication, complex_mul_pattern::matches,
21620 complex_fms_pattern::matches): Remove complex_perm_kinds_t.
21621 * tree-vectorizer.h: (complex_load_perm_t): Removed.
21622 (slp_tree_to_load_perm_map_t): Use complex_perm_kinds_t instead of
21623 complex_load_perm_t.
21625 2021-03-19 H.J. Lu <hjl.tools@gmail.com>
21628 * config/i386/i386-options.c (ix86_init_machine_status): Set
21630 * config/i386/i386.c (init_cumulative_args): Set silent_p to
21632 (construct_container): Return early for return and argument
21633 errors if silent_p is true.
21634 * config/i386/i386.h (machine_function): Add silent_p.
21636 2021-03-19 Jakub Jelinek <jakub@redhat.com>
21639 * config/arm/constraints.md (Ds): New constraint.
21640 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): Use w,Ds
21641 constraint instead of w,Dm.
21643 2021-03-19 Andrew Stubbs <ams@codesourcery.com>
21645 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Fix quotes
21648 2021-03-19 Eric Botcazou <ebotcazou@adacore.com>
21650 PR middle-end/99641
21651 * fold-const.c (native_encode_initializer) <CONSTRUCTOR>: For an
21652 array type, do the computation of the current position in sizetype.
21654 2021-03-18 Vladimir N. Makarov <vmakarov@redhat.com>
21657 * lra-constraints.c (process_address_1): Use lookup_constraint
21658 only for a single constraint.
21660 2021-03-18 Martin Sebor <msebor@redhat.com>
21662 PR middle-end/99502
21663 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): Rename...
21664 (inbounds_memaccess_p): ...to this. Check the ending offset of
21665 the accessed member.
21667 2021-03-18 Andrew Stubbs <ams@codesourcery.com>
21669 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Add %< and
21670 %> quote markers to error messages.
21671 (gcn_goacc_validate_dims): Likewise.
21672 (gcn_conditional_register_usage): Remove exclaimation mark from error
21674 (gcn_vectorize_vec_perm_const): Ensure perm is fully uninitialized.
21676 2021-03-18 Jan Hubicka <hubicka@ucw.cz>
21678 * config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
21681 2021-03-18 Sinan Lin <sinan@isrc.iscas.ac.cn>
21682 Kito Cheng <kito.cheng@sifive.com>
21684 * config/riscv/riscv.c (riscv_block_move_straight): Change type
21685 to unsigned HOST_WIDE_INT for parameter and local variable with
21686 HOST_WIDE_INT type.
21687 (riscv_adjust_block_mem): Ditto.
21688 (riscv_block_move_loop): Ditto.
21689 (riscv_expand_block_move): Ditto.
21691 2021-03-18 Nick Clifton <nickc@redhat.com>
21693 * config/v850/v850.c (construct_restore_jr): Increase static
21695 (construct_save_jarl): Likewise.
21696 * config/v850/v850.h (DWARF2_DEBUGGING_INFO): Define.
21698 2021-03-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21700 * config/aarch64/aarch64.c (aarch64_adjust_generic_arch_tuning): Define.
21701 (aarch64_override_options_internal): Use it.
21702 (generic_tunings): Add AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS to
21705 2021-03-17 Sandra Loosemore <sandra@codesourcery.com>
21707 * config/nios2/nios2.c (nios2_custom_check_insns): Clean up
21708 error message format issues.
21709 (nios2_option_override): Likewise.
21710 (nios2_expand_fpu_builtin): Likewise.
21711 (nios2_init_custom_builtins): Adjust to avoid bogus strncpy
21712 truncation warning.
21713 (nios2_expand_custom_builtin): More error message format fixes.
21714 (nios2_expand_rdwrctl_builtin): Likewise.
21715 (nios2_expand_rdprs_builtin): Likewise.
21716 (nios2_expand_eni_builtin): Likewise.
21717 (nios2_expand_builtin): Likewise.
21718 (nios2_register_custom_code): Likewise.
21719 (nios2_valid_target_attribute_rec): Likewise.
21720 (nios2_add_insn_asm): Fix uninitialized variable warning.
21722 2021-03-17 Jan Hubicka <jh@suse.cz>
21724 * config/i386/x86-tune-costs.h (struct processor_costs): Update costs
21725 of gather to match reality.
21726 * config/i386/x86-tune.def (X86_TUNE_USE_GATHER): Enable for znver3.
21728 2021-03-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21730 * config/aarch64/aarch64-builtins.c (aarch64_expand_rng_builtin): Use EQ
21731 to compare against CC_REG rather than NE.
21733 2021-03-17 H.J. Lu <hjl.tools@gmail.com>
21736 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support
21737 inline assembly statements.
21738 (ix86_print_operand): Update 'P' handling for -fno-plt.
21740 2021-03-17 Tamar Christina <tamar.christina@arm.com>
21743 * config/aarch64/aarch64.c
21744 (aarch64_simd_clone_compute_vecsize_and_simdlen): Remove unused var.
21746 2021-03-16 Segher Boessenkool <segher@kernel.crashing.org>
21749 * config/rs6000/predicates.md (branch_comparison_operator): Allow
21750 ordered and unordered for CCFPmode, if flag_finite_math_only.
21752 2021-03-16 Jakub Jelinek <jakub@redhat.com>
21755 * config/i386/i386-expand.c (ix86_split_lea_for_addr): Emit a MULT
21756 rather than ASHIFT.
21757 * config/i386/i386.md (mult by 1248 into ashift): New splitter.
21759 2021-03-16 Martin Liska <mliska@suse.cz>
21762 * optc-save-gen.awk: Add flag_ipa_ra to exceptions for
21763 cl_optimization_compare function.
21765 2021-03-16 Ilya Leoshkevich <iii@linux.ibm.com>
21767 * config/s390/s390.c (f_constraint_p): Treat "fv" constraints
21770 2021-03-16 Jakub Jelinek <jakub@redhat.com>
21773 * config/i386/i386.h (struct machine_function): Add
21774 has_explicit_vzeroupper bitfield.
21775 * config/i386/i386-expand.c (ix86_expand_builtin): Set
21776 cfun->machine->has_explicit_vzeroupper when expanding
21777 IX86_BUILTIN_VZEROUPPER.
21778 * config/i386/i386-features.c (rest_of_handle_insert_vzeroupper):
21779 Do the mode switching only when TARGET_VZEROUPPER, expensive
21780 optimizations turned on and not optimizing for size.
21781 (pass_insert_vzeroupper::gate): Enable even when
21782 cfun->machine->has_explicit_vzeroupper is set.
21784 2021-03-16 Jakub Jelinek <jakub@redhat.com>
21787 * config/aarch64/aarch64.c
21788 (aarch64_simd_clone_compute_vecsize_and_simdlen): If not a function
21789 definition, walk TYPE_ARG_TYPES list if non-NULL for argument types
21790 instead of DECL_ARGUMENTS. Ignore types for uniform arguments.
21792 2021-03-15 Richard Biener <rguenther@suse.de>
21794 PR tree-optimization/98834
21795 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle missing
21796 subsetting by truncating the access size.
21798 2021-03-15 Jan Hubicka <hubicka@ucw.cz>
21800 * config/i386/i386-options.c (processor_cost_table): Add znver3_cost.
21801 * config/i386/x86-tune-costs.h (znver3_cost): New gobal variable; copy
21804 2021-03-15 Martin Liska <mliska@suse.cz>
21806 * spellcheck.c: Add missing comma in initialization.
21808 2021-03-14 Uroš Bizjak <ubizjak@gmail.com>
21810 * config/i386/sse.md (*vec_extract<mode>): Merge alternative 0 with
21811 alternative 2 and alternative 1 with alternative 3 using
21812 YW register constraint.
21813 (*vec_extract<PEXTR_MODE12:mode>_zext): Merge alternatives
21814 using YW register constraint.
21815 (*vec_extractv16qi_zext): Ditto.
21816 (*vec_extractv4si): Merge alternatives 4 and 5
21817 using Yw register constraint.
21818 (*ssse3_palignr<mode>_perm): Use Yw instead of v for alternative 3.
21820 2021-03-13 Martin Sebor <msebor@redhat.com>
21822 PR tree-optimization/99489
21823 * builtins.c (gimple_call_alloc_size): Fail gracefully when argument
21824 is not a call statement.
21826 2021-03-13 Jakub Jelinek <jakub@redhat.com>
21828 PR tree-optimization/99544
21829 * match.pd (X + (X << C) -> X * (1 + (1 << C))): Don't simplify
21830 if for vector types multiplication can't be done in type's mode.
21832 2021-03-12 Eric Botcazou <ebotcazou@adacore.com>
21835 * config/sparc/constraints.md (w): Rename to...
21836 (W): ... this and ditch previous implementation.
21837 * config/sparc/sparc.md (*movdi_insn_sp64): Replace W with m.
21838 (*movdf_insn_sp64): Likewise.
21839 (*mov<VM64:mode>_insn_sp64): Likewise.
21840 * config/sparc/sync.md (*atomic_compare_and_swap<mode>_1): Replace
21842 (atomic_compare_and_swap_leon3_1): Likewise.
21843 (*atomic_compare_and_swapdi_v8plus): Likewise.
21844 * config/sparc/sparc.c (memory_ok_for_ldd): Remove useless test on
21845 architecture and add missing address validity check during LRA.
21847 2021-03-12 Tobias Burnus <tobias@codesourcery.com>
21850 * gimplify.c (omp_add_variable): Handle NULL_TREE as size
21851 occuring for assumed-size arrays in use_device_{ptr,addr}.
21853 2021-03-12 Jakub Jelinek <jakub@redhat.com>
21856 * config/i386/constraints.md (YW): New internal constraint.
21857 * config/i386/sse.md (v_Yw): Add V4TI, V2TI, V1TI and TI cases.
21858 (*<sse2_avx2>_<insn><mode>3<mask_name>,
21859 *<sse2_avx2>_uavg<mode>3<mask_name>, *abs<mode>2,
21860 *<s>mul<mode>3_highpart<mask_name>): Use <v_Yw> instead of v in
21862 (<sse2_avx2>_psadbw): Use YW instead of v in constraints.
21863 (*avx2_pmaddwd, *sse2_pmaddwd, *<code>v8hi3, *<code>v16qi3,
21864 avx2_pmaddubsw256, ssse3_pmaddubsw128): Merge last two alternatives
21865 into one, use Yw instead of former x,v.
21866 (ashr<mode>3, <insn><mode>3): Use <v_Yw> instead of x in constraints of
21867 the last alternative.
21868 (<sse2_avx2>_packsswb<mask_name>, <sse2_avx2>_packssdw<mask_name>,
21869 <sse2_avx2>_packuswb<mask_name>, <sse4_1_avx2>_packusdw<mask_name>,
21870 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, <ssse3_avx2>_palignr<mode>,
21871 <ssse3_avx2>_pshufb<mode>3<mask_name>): Merge last two alternatives
21872 into one, use <v_Yw> instead of former x,v.
21873 (avx2_interleave_highv32qi<mask_name>,
21874 vec_interleave_highv16qi<mask_name>): Use Yw instead of v in
21875 constraints. Add && <mask_avx512bw_condition> to condition.
21876 (avx2_interleave_lowv32qi<mask_name>,
21877 vec_interleave_lowv16qi<mask_name>,
21878 avx2_interleave_highv16hi<mask_name>,
21879 vec_interleave_highv8hi<mask_name>,
21880 avx2_interleave_lowv16hi<mask_name>, vec_interleave_lowv8hi<mask_name>,
21881 avx2_pshuflw_1<mask_name>, sse2_pshuflw_1<mask_name>,
21882 avx2_pshufhw_1<mask_name>, sse2_pshufhw_1<mask_name>,
21883 avx2_<code>v16qiv16hi2<mask_name>, sse4_1_<code>v8qiv8hi2<mask_name>,
21884 *sse4_1_<code>v8qiv8hi2<mask_name>_1, <sse2_avx2>_<insn><mode>3): Use
21885 Yw instead of v in constraints.
21886 * config/i386/mmx.md (Yv_Yw): New define_mode_attr.
21887 (*mmx_<insn><mode>3, mmx_ashr<mode>3, mmx_<insn><mode>3): Use <Yv_Yw>
21888 instead of Yv in constraints.
21889 (*mmx_<insn><mode>3, *mmx_mulv4hi3, *mmx_smulv4hi3_highpart,
21890 *mmx_umulv4hi3_highpart, *mmx_pmaddwd, *mmx_<code>v4hi3,
21891 *mmx_<code>v8qi3, mmx_pack<s_trunsuffix>swb, mmx_packssdw,
21892 mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
21893 *mmx_uavgv8qi3, *mmx_uavgv4hi3, mmx_psadbw): Use Yw instead of Yv in
21895 (*mmx_pinsrw, *mmx_pinsrb, *mmx_pextrw, *mmx_pextrw_zext, *mmx_pextrb,
21896 *mmx_pextrb_zext): Use YW instead of Yv in constraints.
21897 (*mmx_eq<mode>3, mmx_gt<mode>3): Use x instead of Yv in constraints.
21898 (mmx_andnot<mode>3, *mmx_<code><mode>3): Split last alternative into
21899 two, one with just x, another isa avx512vl with v.
21901 2021-03-12 Martin Liska <mliska@suse.cz>
21903 * doc/invoke.texi: Add missing param documentation.
21905 2021-03-11 David Malcolm <dmalcolm@redhat.com>
21908 * Makefile.in (ANALYZER_OBJS): Add analyzer/feasible-graph.o and
21909 analyzer/trimmed-graph.o.
21910 * doc/analyzer.texi (Analyzer Paths): Rewrite description of
21911 feasibility checking to reflect new implementation.
21912 * doc/invoke.texi (-fdump-analyzer-feasibility): Document new
21914 * shortest-paths.h (shortest_paths::get_shortest_distance): New.
21916 2021-03-11 David Malcolm <dmalcolm@redhat.com>
21918 * digraph.cc (selftest::test_shortest_paths): Update
21919 shortest_paths init for new param. Add test of
21920 SPS_TO_GIVEN_TARGET.
21921 * shortest-paths.h (enum shortest_path_sense): New.
21922 (shortest_paths::shortest_paths): Add "sense" param.
21923 Update for renamings. Generalize to use "sense" param.
21924 (shortest_paths::get_shortest_path): Rename param.
21925 (shortest_paths::m_sense): New field.
21926 (shortest_paths::m_prev): Rename...
21927 (shortest_paths::m_best_edge): ...to this.
21928 (shortest_paths::get_shortest_path): Update for renamings.
21929 Conditionalize flipping of path on sense of traversal.
21931 2021-03-11 David Malcolm <dmalcolm@redhat.com>
21933 * digraph.cc (selftest::test_shortest_paths): Add test coverage
21934 for paths from B and C.
21935 * shortest-paths.h (shortest_paths::shortest_paths): Handle
21936 unreachable nodes, rather than asserting.
21938 2021-03-11 David Edelsohn <dje.gcc@gmail.com>
21941 * config/rs6000/rs6000.c (rs6000_xcoff_file_start): Don't create
21942 xcoff_tbss_section_name.
21943 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .lcomm.
21944 * xcoffout.c (xcoff_tbss_section_name): Delete.
21945 * xcoffout.h (xcoff_tbss_section_name): Delete.
21947 2021-03-11 Richard Biener <rguenther@suse.de>
21949 PR tree-optimization/99523
21950 * tree-cfg.c (dump_function_to_file): Dump SSA names
21951 w/o identifier to the decls section as well, not only those
21952 without a VAR_DECL.
21954 2021-03-11 Jakub Jelinek <jakub@redhat.com>
21957 * ipa-icf-gimple.c (func_checker::compare_gimple_call): For internal
21958 function calls with lhs fail if the lhs don't have compatible types.
21960 2021-03-11 Hans-Peter Nilsson <hp@axis.com>
21962 * config/cris/cris.h (HARD_FRAME_POINTER_REGNUM): Define.
21963 Change FRAME_POINTER_REGNUM to correspond to a new faked
21964 register faked_fp, part of GENNONACR_REGS like faked_ap.
21965 (CRIS_FAKED_REGS_CONTENTS): New helper macro.
21966 (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS, CALL_USED_REGISTERS):
21967 (REG_ALLOC_ORDER, REG_CLASS_CONTENTS, REGNO_OK_FOR_BASE_P)
21968 (ELIMINABLE_REGS, REGISTER_NAMES): Adjust accordingly.
21969 * config/cris/cris.md (CRIS_FP_REGNUM): Renumber to new faked
21971 (CRIS_REAL_FP_REGNUM): New constant.
21972 * config/cris/cris.c (cris_reg_saved_in_regsave_area): Check
21973 for HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM.
21974 (cris_initial_elimination_offset): Handle elimination changes
21975 to HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM
21976 and add one from FRAME_POINTER_REGNUM to
21977 HARD_FRAME_POINTER_REGNUM.
21978 (cris_expand_prologue, cris_expand_epilogue): Emit code for
21979 hard_frame_pointer_rtx instead of frame_pointer_rtx.
21981 2021-03-10 David Edelsohn <dje.gcc@gmail.com>
21984 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Add check for DCmode.
21985 * config/rs6000/rs6000.c (rs6000_special_round_type_align): Same.
21987 2021-03-10 Vladimir N. Makarov <vmakarov@redhat.com>
21990 * lra-constraints.c (process_address_1): Don't check unknown
21991 constraint, use X for empty constraint.
21993 2021-03-10 Alex Coplan <alex.coplan@arm.com>
21995 * config/aarch64/aarch64.c (aarch64_vfp_is_call_or_return_candidate):
21996 Fix typo in comment describing "is_ha" argument.
21998 2021-03-10 John David Anglin <danglin@gcc.gnu.org>
22000 * doc/sourcebuild.texi: Document LRA target selector.
22002 2021-03-10 David Malcolm <dmalcolm@redhat.com>
22004 * doc/ux.texi: Add subsection contrasting interactive versus
22005 batch usage of GCC.
22007 2021-03-10 Joel Hutton <joel.hutton@arm.com>
22010 * tree-vect-stmts.c (vectorizable_store): Fix scatter store mask
22012 (vectorizable_load): Fix gather load mask check condition.
22014 2021-03-10 Richard Biener <rguenther@suse.de>
22016 PR tree-optimization/99510
22017 * tree.c (check_aligned_type): Check that the candidate
22018 has TYPE_USER_ALIGN set instead of matching with the
22021 2021-03-10 Eric Botcazou <ebotcazou@adacore.com>
22023 * config/sparc/sparc.c (sparc_regmode_natural_size): Return 4 for
22024 float and vector integer modes only if the mode is not larger.
22026 2021-03-10 Hans-Peter Nilsson <hp@axis.com>
22028 * config/cris/cris.h (DWARF_FRAME_REGISTERS): Define.
22030 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
22032 * ira.c (ira_setup_alts, ira_get_dup_out_num): Process digital
22034 * ira-lives.c (single_reg_class): Ditto.
22036 2021-03-09 Sebastian Huber <sebastian.huber@embedded-brains.de>
22038 * config.gcc (aarch64-*-rtems*): Include general rtems.h after
22039 the architecture-specific rtems.h.
22040 (aarch64-*-rtems*): Likewise.
22041 (arm*-*-rtems*): Likewise.
22042 (epiphany-*-rtems*): Likewise.
22043 (riscv*-*-rtems*): Likewise.
22045 2021-03-09 Jakub Jelinek <jakub@redhat.com>
22047 PR tree-optimization/99305
22048 * tree-ssa-phiopt.c (conditional_replacement): Test integer_pow2p
22049 before integer_all_onesp instead of vice versa.
22051 2021-03-09 Richard Earnshaw <rearnsha@arm.com>
22053 * common/config/arm/arm-common.c (arm_config_default): Change type
22054 of 'i' to unsigned.
22056 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
22059 * lra-constraints.c (process_address_1): Process constraint 'g'
22060 separately and digital constraints containing more one digit.
22062 2021-03-09 Nick Clifton <nickc@redhat.com>
22064 * config/rx/rx.h (DBX_DEBUGGING_INFO): Define.
22065 (DWARF"_DEBUGGING_INFO): Define.
22067 2021-03-09 Eric Botcazou <ebotcazou@adacore.com>
22070 * calls.c (initialize_argument_information): When the argument
22071 is passed by reference, do not make a copy in a thunk only if
22072 the argument is already in memory. Remove redundant test for
22073 the case of callee copy.
22075 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
22078 * lra-constraints.c (process_address_1): Process 0..9 constraints
22079 in process_address_1.
22081 2021-03-09 Andreas Krebbel <krebbel@linux.ibm.com>
22083 * config/s390/s390.c (struct s390_processor processor_table):
22084 Binutils name string must not be empty.
22086 2021-03-09 Claudiu Zissulescu <claziss@synopsys.com>
22088 * config/arc/arc.c (arc_attr_type): Remove function.
22090 2021-03-09 Martin Liska <mliska@suse.cz>
22093 * config/i386/i386-options.c (ix86_option_override_internal):
22094 Set isa_flags for OPTS argument and not for the global
22097 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
22099 * config/rs6000/predicates.md (ds_form_mem_operand): Check
22102 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
22105 * config/rs6000/predicates.md (ds_form_mem_operand) New
22107 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10) Use
22108 ds_form_mem_operand in ld/lwa patterns.
22109 * config/rs6000/fusion.md: Regenerate file.
22111 2021-03-08 Martin Sebor <msebor@redhat.com>
22113 PR middle-end/98266
22114 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): New function.
22115 (array_bounds_checker::check_array_bounds): Call it.
22117 2021-03-08 Martin Sebor <msebor@redhat.com>
22119 PR middle-end/97631
22120 * tree-ssa-strlen.c (maybe_warn_overflow): Test rawmem.
22121 (handle_builtin_stxncpy_strncat): Rename locals. Determine
22122 destination size from allocation calls. Issue a more appropriate
22124 (handle_builtin_memcpy): Pass true as rawmem to maybe_warn_overflow.
22125 (handle_builtin_memset): Same.
22127 2021-03-08 Peter Bergner <bergner@linux.ibm.com>
22130 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_permute): Add an assert
22131 to ensure we do not have an Altivec style address.
22132 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): Disable if passed
22133 an Altivec style address.
22134 (*vsx_le_perm_store_<mode>): Likewise.
22135 (splitters after *vsx_le_perm_store_<mode>): Likewise.
22136 (vsx_load_<mode>): Disable special expander if passed an Altivec
22138 (vsx_store_<mode>): Likewise.
22140 2021-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22143 * config/aarch64/predicates.md (aarch64_simd_shift_imm_vec_qi): Define.
22144 (aarch64_simd_shift_imm_vec_hi): Likewise.
22145 (aarch64_simd_shift_imm_vec_si): Likewise.
22146 (aarch64_simd_shift_imm_vec_di): Likewise.
22147 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le): Use
22148 predicate from above.
22149 (aarch64_shrn<mode>_insn_be): Likewise.
22150 (aarch64_rshrn<mode>_insn_le): Likewise.
22151 (aarch64_rshrn<mode>_insn_be): Likewise.
22152 (aarch64_shrn2<mode>_insn_le): Likewise.
22153 (aarch64_shrn2<mode>_insn_be): Likewise.
22154 (aarch64_rshrn2<mode>_insn_le): Likewise.
22155 (aarch64_rshrn2<mode>_insn_be): Likewise.
22157 2021-03-08 Vladimir N. Makarov <vmakarov@redhat.com>
22160 * lra-constraints.c (skip_contraint_modifiers): New function.
22161 (process_address_1): Use it before lookup_constraint call.
22163 2021-03-08 Martin Liska <mliska@suse.cz>
22166 * config/i386/i386-options.c (ix86_option_override_internal):
22167 Enable UINTR and HRESET for -march that supports it.
22169 2021-03-08 Ilya Leoshkevich <iii@linux.ibm.com>
22171 * config/s390/s390.c (f_constraint_p): New function.
22172 (s390_md_asm_adjust): Implement TARGET_MD_ASM_ADJUST.
22173 (TARGET_MD_ASM_ADJUST): Likewise.
22175 2021-03-08 Tobias Burnus <tobias@codesourcery.com>
22178 * tree-nested.c (convert_local_reference_stmt): Avoid calling
22179 lookup_field_for_decl for Fortran module (= namespace context).
22181 2021-03-08 Andreas Krebbel <krebbel@linux.ibm.com>
22183 * config/s390/s390.c (s390_expand_vec_compare): Implement <0
22184 comparison with arithmetic right shift.
22185 (s390_expand_vcond): No need for a force_reg anymore.
22186 s390_vec_compare will do it.
22187 * config/s390/vector.md ("vec_cmp<mode><tointvec>"): Accept also
22188 immediate operands.
22190 2021-03-07 Jakub Jelinek <jakub@redhat.com>
22193 * config/i386/constraints.md (Yw): Use SSE_REGS if TARGET_SSE
22194 but TARGET_AVX512BW or TARGET_AVX512VL is not set. Adjust description
22196 * config/i386/sse.md (v_Yw): New define_mode_attr.
22197 (*<insn><mode>3, *mul<mode>3<mask_name>, *avx2_<code><mode>3,
22198 *sse4_1_<code><mode>3<mask_name>): Use <v_Yw> instead of v
22200 * config/i386/mmx.md (mmx_pshufw_1, *vec_dupv4hi): Use Yw instead of
22201 xYw in constraints.
22203 2021-03-06 Julian Brown <julian@codesourcery.com>
22205 * tree-pretty-print.c (dump_generic_node): Emit non-generic
22206 address space info for aggregates.
22208 2021-03-06 Hans-Peter Nilsson <hp@axis.com>
22210 * config/cris/cris.h (MAX_FIXED_MODE_SIZE): Don't define.
22212 2021-03-05 Jakub Jelinek <jakub@redhat.com>
22214 PR middle-end/99322
22215 * tree-cfg.c (bb_to_omp_idx): New variable.
22216 (execute_build_cfg): Release the bb_to_omp_idx vector after
22217 cleanup_tree_cfg returns.
22218 (handle_abnormal_edges): Remove bb_to_omp_idx argument, adjust
22219 for bb_to_omp_idx being a vec<int> instead of pointer to array
22221 (make_edges): Remove bb_to_omp_idx local variable, don't pass
22222 it to handle_abnormal_edges, adjust for bb_to_omp_idx being a
22223 vec<int> instead of pointer to array of ints and don't free/release
22225 (remove_bb): When removing a bb and placing forced label somewhere
22226 else, ensure it is put into the same OpenMP region during cfg
22227 pass if possible or to entry successor as fallback. Unregister
22228 bb from bb_to_omp_idx.
22230 2021-03-05 Vladimir N. Makarov <vmakarov@redhat.com>
22233 * lra-constraints.c (process_address_1): Skip decomposing address
22234 for asm insn operand with unknown constraint.
22236 2021-03-05 Martin Jambor <mjambor@suse.cz>
22239 * cgraph.c (cgraph_edge::set_call_stmt): Do not update all
22240 corresponding speculative edges if we are about to resolve
22241 sepculation. Make edge direct (and so resolve speculations) before
22242 removing it from call_site_hash.
22243 (cgraph_edge::make_direct): Relax the initial assert to allow calling
22244 the function on speculative direct edges.
22246 2021-03-05 Eric Botcazou <ebotcazou@adacore.com>
22248 PR rtl-optimization/99376
22249 * rtlanal.c (nonzero_bits1) <arithmetic operators>: If the number
22250 of low-order zero bits is too large, set the result to 0 directly.
22252 2021-03-04 Jakub Jelinek <jakub@redhat.com>
22254 PR middle-end/93235
22255 * expmed.c (store_bit_field_using_insv): Return false of xop0 is a
22256 SUBREG and a SUBREG to op_mode can't be created.
22258 2021-03-04 Alex Coplan <alex.coplan@arm.com>
22261 * config/aarch64/aarch64-sve-builtins.cc
22262 (function_resolver::require_vector_type): Handle error_mark_node.
22264 2021-03-04 Ilya Leoshkevich <iii@linux.ibm.com>
22266 * cfgexpand.c (expand_asm_loc): Pass new parameter.
22267 (expand_asm_stmt): Likewise.
22268 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Add new
22270 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
22271 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
22272 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
22273 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
22274 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
22275 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
22276 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
22277 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
22278 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
22279 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
22280 * doc/tm.texi (md_asm_adjust): Likewise.
22281 * target.def (md_asm_adjust): Likewise.
22283 2021-03-04 Richard Biener <rguenther@suse.de>
22285 PR middle-end/97855
22286 * tree-pretty-print.c: Poison pp_printf.
22287 (dump_decl_name): Avoid use of pp_printf.
22288 (dump_block_node): Likewise.
22289 (dump_generic_node): Likewise.
22291 2021-03-04 Martin Sebor <msebor@redhat.com>
22293 PR middle-end/96963
22294 PR middle-end/94655
22295 * builtins.c (handle_array_ref): New helper.
22296 (handle_mem_ref): New helper.
22297 (compute_objsize_r): Factor out ARRAY_REF and MEM_REF handling
22298 into new helper functions. Correct a workaround for vectorized
22301 2021-03-03 Pat Haugen <pthaugen@linux.ibm.com>
22303 * config/rs6000/dfp.md (extendddtd2, trunctddd2, *cmp<mode>_internal1,
22304 floatditd2, ftrunc<mode>2, fix<mode>di2, dfp_ddedpd_<mode>,
22305 dfp_denbcd_<mode>, dfp_dxex_<mode>, dfp_diex_<mode>,
22306 *dfp_sgnfcnc_<mode>, dfp_dscli_<mode>, dfp_dscri_<mode>): Update size
22307 attribute for Power10.
22308 * config/rs6000/mma.md (*movoo): Likewise.
22309 * config/rs6000/rs6000.md (define_attr "size"): Add 256.
22310 (define_mode_attr bits): Add DD/TD modes.
22311 * config/rs6000/sync.md (load_quadpti, store_quadpti, load_lockedpti,
22312 store_conditionalpti): Update size attribute for Power10.
22314 2021-03-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22317 * config/sparc/t-sparc (tree-ssanames.o-warn): Don't error for
22318 -Wuninitialized, -Wmaybe-uninitialized.
22319 (wide-int.o-warn): Likewise.
22321 2021-03-03 Richard Earnshaw <rearnsha@arm.com>
22323 * common/config/arm/arm-common.c: Include configargs.h.
22324 (arm_config_default): New function.
22325 (arm_target_mode): Renamed from arm_target_thumb_only. Handle
22326 processors that do not support Thumb. Take into account the
22327 --with-mode configuration setting for selecting the default.
22328 * config/arm/arm.h (OPTION_DEFAULT_SPECS): Remove entry for 'mode'.
22329 (TARGET_MODE_SPEC_FUNCTIONS): Update for function name change.
22331 2021-03-03 Martin Liska <mliska@suse.cz>
22333 PR gcov-profile/97461
22334 * gcov-io.h (GCOV_PREALLOCATED_KVP): Remove.
22336 2021-03-03 Eric Botcazou <ebotcazou@adacore.com>
22339 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
22340 point back the hard frame pointer to its default location when the
22341 frame is larger than SEH_MAX_FRAME_SIZE.
22343 2021-03-03 Jakub Jelinek <jakub@redhat.com>
22346 * config/i386/predicates.md (logic_operator): New define_predicate.
22347 * config/i386/i386.md (mov + mem using comm arith peephole2):
22348 Punt if operands[1] is EXT_REX_SSE_REGNO_P, AVX512BW is not enabled
22349 and the inner mode is [QH]Imode.
22351 2021-03-03 Jakub Jelinek <jakub@redhat.com>
22354 * dwarf2out.c (dw_loc_list_struct): Add end_entry member.
22355 (new_loc_list): Clear end_entry.
22356 (output_loc_list): Only use DW_LLE_startx_length for -gsplit-dwarf
22357 if HAVE_AS_LEB128, otherwise use DW_LLE_startx_endx. Fix comment
22359 (index_location_lists): For dwarf_version >= 5 without HAVE_AS_LEB128,
22360 initialize also end_entry.
22362 2021-03-03 Jakub Jelinek <jakub@redhat.com>
22365 * cfgrtl.c (fixup_partitions): When changing some bbs from hot to cold
22366 partitions, if in non-layout mode after reorder_blocks also move
22367 affected blocks to ensure a single partition transition.
22369 2021-03-03 Jason Merrill <jason@redhat.com>
22372 * cgraphunit.c (process_function_and_variable_attributes): Don't
22373 warn about flatten on an alias if the target also has it.
22374 * cgraph.h (symtab_node::get_alias_target_tree): New.
22376 2021-03-02 David Edelsohn <dje.gcc@gmail.com>
22378 * config/rs6000/rs6000.md (tls_get_tpointer_internal): Prepend
22379 period to symbol name.
22380 (tls_get_addr_internal<mode>): Same.
22382 2021-03-02 David Malcolm <dmalcolm@redhat.com>
22385 * diagnostic-show-locus.c
22386 (selftest::test_one_liner_many_fixits_2): Fix accidental usage of
22389 2021-03-02 Martin Sebor <msebor@redhat.com>
22391 PR middle-end/99276
22392 * builtins.c (warn_for_access): Remove stray warning text.
22394 2021-03-02 Martin Sebor <msebor@redhat.com>
22396 PR middle-end/99295
22397 * doc/extend.texi (attribute malloc): Reword and clarify nonaliasing
22400 2021-03-02 Jakub Jelinek <jakub@redhat.com>
22403 * dwarf2out.c (output_macinfo_op): Use DW_MACRO_*_str* even with
22404 -gdwarf-5 -gstrict-dwarf. For -gsplit-dwarf -gdwarf-5 use
22405 DW_MACRO_*_strx instead of DW_MACRO_*_strp. Handle
22406 DW_MACRO_define_strx and DW_MACRO_undef_strx.
22407 (save_macinfo_strings): Use DW_MACRO_*_str* even with
22408 -gdwarf-5 -gstrict-dwarf. Handle DW_MACRO_define_strx and
22409 DW_MACRO_undef_strx.
22411 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
22413 * config/s390/s390-builtin-types.def (BT_FN_V4SF_V8HI_UINT): New
22415 (BT_FN_V8HI_V8HI_UINT): Likewise.
22416 (BT_FN_V8HI_V4SF_V4SF_UINT): Likewise.
22417 * config/s390/s390-builtins.def (B_NNPA): New macro definition.
22418 (s390_vclfnhs, s390_vclfnls, s390_vcrnfs, s390_vcfn, s390_vcnf):
22419 New builtin definitions.
22420 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Bump
22421 vector extension version.
22422 * config/s390/s390.c (s390_expand_builtin): Check if builtins are
22423 available with current -march level.
22424 * config/s390/s390.md (UNSPEC_NNPA_VCLFNHS_V8HI)
22425 (UNSPEC_NNPA_VCLFNLS_V8HI, UNSPEC_NNPA_VCRNFS_V8HI)
22426 (UNSPEC_NNPA_VCFN_V8HI, UNSPEC_NNPA_VCNF_V8HI): New constants.
22427 * config/s390/vecintrin.h (vec_extend_to_fp32_hi): New macro.
22428 (vec_extend_to_fp32_lo): Likewise.
22429 (vec_round_from_fp32): Likewise.
22430 (vec_convert_to_fp16): Likewise.
22431 (vec_convert_from_fp16): Likewise.
22432 * config/s390/vx-builtins.md (vclfnhs_v8hi): New insn pattern.
22433 (vclfnls_v8hi): Likewise.
22434 (vcrnfs_v8hi): Likewise.
22435 (vcfn_v8hi): Likewise.
22436 (vcnf_v8hi): Likewise.
22438 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
22440 * common/config/s390/s390-common.c (processor_flags_table): New entry.
22441 * config.gcc: Enable arch14 for --with-arch and --with-tune.
22442 * config/s390/driver-native.c (s390_host_detect_local_cpu): Pick
22443 arch14 for unknown CPU models.
22444 * config/s390/s390-opts.h (enum processor_type): Add PROCESSOR_ARCH14.
22445 * config/s390/s390.c (s390_issue_rate): Add case for PROCESSOR_ARCH14.
22446 (s390_get_sched_attrmask): Likewise.
22447 (s390_get_unit_mask): Likewise.
22448 * config/s390/s390.h (enum processor_flags): Add PF_NNPA and PF_ARCH14.
22449 (TARGET_CPU_ARCH14, TARGET_CPU_ARCH14_P, TARGET_CPU_NNPA)
22450 (TARGET_CPU_NNPA_P, TARGET_ARCH14, TARGET_ARCH14_P, TARGET_NNPA)
22451 (TARGET_NNPA_P): New macro definitions.
22452 * config/s390/s390.md ("cpu_facility", "enabled"): Add arch14 and nnpa.
22453 * config/s390/s390.opt: Add PROCESSOR_ARCH14.
22455 2021-03-02 Jakub Jelinek <jakub@redhat.com>
22457 PR middle-end/95757
22458 * tree-vrp.c (register_edge_assert_for): Remove superfluous ()s around
22459 condition. Call register_edge_assert_for_1 for == 0, != 0, == 1 and
22460 != 1 comparisons if name is lhs of a comparison.
22462 2021-03-01 Iain Sandoe <iain@sandoe.co.uk>
22466 * config/darwin-protos.h (darwin_should_restore_cfa_state): New.
22467 * config/darwin.c (darwin_should_restore_cfa_state): New.
22468 * config/darwin.h (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New.
22469 * doc/tm.texi: Regenerated.
22470 * doc/tm.texi.in: Document TARGET_ASM_SHOULD_RESTORE_CFA_STATE.
22471 * dwarf2cfi.c (connect_traces): If the target requests, restore
22472 the CFA expression after a DW_CFA_restore.
22473 * target.def (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New hook.
22475 2021-03-01 Martin Liska <mliska@suse.cz>
22478 * optc-save-gen.awk: Add 4 more exceptions.
22480 2021-03-01 Nathan Sidwell <nathan@acm.org>
22483 * tree.h (TYPE_ALIGN_RAW): New accessor.
22484 (TYPE_ALIGN): Use it.
22486 2021-03-01 Jan Hubicka <jh@suse.cz>
22489 * ipa-fnsummary.c (compute_fn_summary): Fix sanity check.
22491 2021-03-01 Eric Botcazou <ebotcazou@adacore.com>
22494 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
22495 point the hard frame pointer to the SSE register save area instead
22496 of the general register save area. Perform only minimal adjustment
22497 for small frames if it is initially not correctly aligned.
22498 (ix86_expand_prologue): Remove early saves for a SEH target.
22499 * config/i386/winnt.c (struct seh_frame_state): Document constraint.
22501 2021-02-28 Jakub Jelinek <jakub@redhat.com>
22504 * ipa.c (symbol_table::remove_unreachable_nodes): Fix a comment
22505 typo - referneced -> referenced.
22506 * tree.c (component_ref_size): Fix comment typo -
22507 refernce -> reference.
22508 * tree-ssa-alias.c (access_path_may_continue_p): Fix comment typo -
22509 traling -> trailing.
22510 (aliasing_component_refs_p): Fix comment typos -
22511 refernce -> reference and refernece -> reference and
22512 traling -> trailing.
22513 (nonoverlapping_refs_since_match_p): Fix comment typo -
22514 referneces -> references.
22515 * doc/invoke.texi (--param modref-max-bases): Fix a typo -
22516 referneces -> references.
22518 2021-02-27 Iain Sandoe <iain@sandoe.co.uk>
22520 * config/host-darwin.c (darwin_gt_pch_use_address): Modify
22521 diagnostic message to avoid use of a contraction and format
22524 2021-02-27 Jakub Jelinek <jakub@redhat.com>
22527 * gcse.c (gcse_or_cprop_is_too_expensive): Use %wu instead of
22528 HOST_WIDE_INT_PRINT_UNSIGNED in warning format string.
22529 * ipa-devirt.c (ipa_odr_read_section): Use %wd instead of
22530 HOST_WIDE_INT_PRINT_DEC in inform format string. Fix comment
22533 2021-02-26 Richard Biener <rguenther@suse.de>
22535 PR middle-end/99281
22536 * expr.c (store_field): For calls with return-slot optimization
22537 and addressable return type expand the store directly.
22539 2021-02-26 Richard Biener <rguenther@suse.de>
22542 * builtins.c (warn_string_no_nul): Fix diagnostic formatting.
22544 2021-02-26 Peter Bergner <bergner@linux.ibm.com>
22547 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Replace assert
22550 2021-02-26 Aaron Sawdey <acsawdey@linux.ibm.com>
22552 * config.gcc: Add rs6000-pcrel-opt.o.
22553 * config/rs6000/rs6000-pcrel-opt.c: New file.
22554 * config/rs6000/pcrel-opt.md: New file.
22555 * config/rs6000/predicates.md: Add d_form_memory predicate.
22556 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_PCREL_OPT.
22557 * config/rs6000/rs6000-passes.def: Add pass_pcrel_opt.
22558 * config/rs6000/rs6000-protos.h: Add reg_to_non_prefixed(),
22559 pcrel_opt_valid_mem_p(), output_pcrel_opt_reloc(),
22560 and make_pass_pcrel_opt().
22561 * config/rs6000/rs6000.c (reg_to_non_prefixed): Make global.
22562 (rs6000_option_override_internal): Add pcrel-opt.
22563 (rs6000_delegitimize_address): Support pcrel-opt.
22564 (rs6000_opt_masks): Add pcrel-opt.
22565 (pcrel_opt_valid_mem_p): New function.
22566 (reg_to_non_prefixed): Make global.
22567 (rs6000_asm_output_opcode): Reset prepend_p_to_next_insn.
22568 (output_pcrel_opt_reloc): New function.
22569 * config/rs6000/rs6000.md (loads_extern_addr): New attr.
22570 (pcrel_extern_addr): Set loads_extern_addr.
22571 Add include for pcrel-opt.md.
22572 * config/rs6000/rs6000.opt: Add -mpcrel-opt.
22573 * config/rs6000/t-rs6000: Add rules for pcrel-opt.c and
22576 2021-02-26 YunQiang Su <yunqiang.su@cipunited.com>
22579 * config/mips/mips.c (mips_expand_ext_as_unaligned_load):
22580 If TARGET_64BIT and dest is SUBREG, we check the width, if it
22581 equal to SImode, we use SImode operation, just like what we are
22584 2021-02-26 Marek Polacek <polacek@redhat.com>
22586 * builtins.c (warn_for_access): Fix typos.
22588 2021-02-25 Iain Sandoe <iain@sandoe.co.uk>
22590 * config/aarch64/aarch64.md (<optab>_rol<mode>3): Add a '#'
22591 mark in front of the immediate quantity.
22592 (<optab>_rolsi3_uxtw): Likewise.
22594 2021-02-25 Richard Earnshaw <rearnsha@arm.com>
22597 * config/arm/thumb2.md (nonsecure_call_reg_thumb2_fpcxt): New pattern.
22598 (nonsecure_call_value_reg_thumb2_fpcxt): Likewise.
22599 (nonsecure_call_reg_thumb2): Restrict to using r4 for the callee
22600 address and disable when the FPCXT is not available.
22601 (nonsecure_call_value_reg_thumb2): Likewise.
22603 2021-02-25 Nathan Sidwell <nathan@acm.org>
22606 * doc/invoke.texi (flang-info-module-cmi): Renamed option.
22608 2021-02-25 Tamar Christina <tamar.christina@arm.com>
22610 * tree-vect-slp.c (optimize_load_redistribution_1): Abort on NULL nodes.
22612 2021-02-25 Richard Biener <rguenther@suse.de>
22614 PR tree-optimization/99253
22615 * tree-vect-loop.c (check_reduction_path): First compute
22616 code, then verify out-of-loop uses.
22618 2021-02-25 Jakub Jelinek <jakub@redhat.com>
22621 * match.pd ((T)(A) + CST -> (T)(A + CST)): Add :s to convert.
22623 2021-02-25 Jakub Jelinek <jakub@redhat.com>
22625 PR tree-optimization/80635
22626 * tree-vrp.c (vrp_simplify_cond_using_ranges): Also handle
22627 VIEW_CONVERT_EXPR if modes are the same, innerop is integral and
22628 has mode precision.
22630 2021-02-25 Richard Biener <rguenther@suse.de>
22632 * tree-vect-slp.c (optimize_load_redistribution_1): Delay
22633 load_map population.
22634 (vect_match_slp_patterns_2): Revert part of last change.
22635 (vect_analyze_slp): Do not interleave optimize_load_redistribution
22636 with pattern detection but do it afterwards. Dump the
22637 whole SLP graph after pattern recognition and load
22638 redistribution optimization finished.
22640 2021-02-24 Jakub Jelinek <jakub@redhat.com>
22643 * omp-low.c (struct omp_context): Add teams_nested_p and
22644 nonteams_nested_p members.
22645 (scan_omp_target): Diagnose teams nested inside of target with other
22646 directives strictly nested inside of the same target.
22647 (check_omp_nesting_restrictions): Set ctx->teams_nested_p or
22648 ctx->nonteams_nested_p as needed.
22650 2021-02-24 Vladimir N. Makarov <vmakarov@redhat.com>
22652 PR inline-asm/99123
22653 * lra-constraints.c (uses_hard_regs_p): Don't use decompose_mem_address.
22655 2021-02-24 Hans-Peter Nilsson <hp@axis.com>
22657 * config/cris/cris.c (cris_expand_prologue): Set
22658 current_function_static_stack_size, if flag_stack_usage_info.
22660 2021-02-24 Pat Haugen <pthaugen@linux.ibm.com>
22662 * config/rs6000/rs6000.c (next_insn_prefixed_p): Rename.
22663 (rs6000_final_prescan_insn): Adjust.
22664 (rs6000_asm_output_opcode): Likewise.
22666 2021-02-24 Martin Sebor <msebor@redhat.com>
22668 PR middle-end/97172
22669 * attribs.c (attr_access::free_lang_data): Clear attribute arg spec
22670 from function arguments.
22672 2021-02-24 Tamar Christina <tamar.christina@arm.com>
22674 PR tree-optimization/99220
22675 * tree-vect-slp.c (optimize_load_redistribution_1): Remove
22676 node from cache when it's about to be deleted.
22678 2021-02-24 Jakub Jelinek <jakub@redhat.com>
22680 PR tree-optimization/99225
22681 * fold-const.c (fold_binary_loc) <case NE_EXPR>: In (x & (1 << y)) != 0
22682 to ((x >> y) & 1) != 0 simplifications use build_one_cst instead of
22683 build_int_cst (..., 1). Formatting fixes.
22685 2021-02-24 Tamar Christina <tamar.christina@arm.com>
22687 PR tree-optimization/99149
22688 * tree-vect-slp-patterns.c (vect_detect_pair_op): Don't recreate the
22690 (vect_slp_reset_pattern): Remove.
22691 (complex_fma_pattern::matches): Remove call to vect_slp_reset_pattern.
22692 (complex_mul_pattern::build, complex_fma_pattern::build,
22693 complex_fms_pattern::build): Fix ref counts.
22694 * tree-vect-slp.c (vect_free_slp_tree): Undo SLP only pattern relevancy
22695 when node is being deleted.
22696 (vect_match_slp_patterns_2): Correct result of cache hit on patterns.
22697 (vect_schedule_slp): Invalidate SLP_TREE_REPRESENTATIVE of removed
22699 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize value.
22701 2021-02-24 Matthias Klose <doko@ubuntu.com>
22704 2020-12-07 Matthias Klose <doko@ubuntu.com>
22706 * genextract.c (print_header): Undefine ENABLE_RTL_CHECKING
22707 and ENABLE_RTL_FLAG_CHECKING.
22709 2021-02-24 Richard Biener <rguenther@suse.de>
22712 * builtins.c (fold_builtin_next_arg): Avoid NULL arg.
22714 2021-02-23 Peter Bergner <bergner@linux.ibm.com>
22716 * config/rs6000/mma.md (mma_assemble_pair): Rename from this...
22717 (vsx_assemble_pair): ...to this.
22718 (*mma_assemble_pair): Rename from this...
22719 (*vsx_assemble_pair): ...to this.
22720 (mma_disassemble_pair): Rename from this...
22721 (vsx_disassemble_pair): ...to this.
22722 (*mma_disassemble_pair): Rename from this...
22723 (*vsx_disassemble_pair): ...to this.
22724 * config/rs6000/rs6000-builtin.def (BU_MMA_V2, BU_MMA_V3,
22725 BU_COMPAT): New macros.
22726 (mma_assemble_pair): Rename from this...
22727 (vsx_assemble_pair): ...to this.
22728 (mma_disassemble_pair): Rename from this...
22729 (vsx_disassemble_pair): ...to this.
22730 (mma_assemble_pair): New compatibility built-in.
22731 (mma_disassemble_pair): Likewise.
22732 * config/rs6000/rs6000-call.c (struct builtin_compatibility): New.
22733 (RS6000_BUILTIN_COMPAT): Define.
22734 (bdesc_compat): New.
22735 (mma_expand_builtin): Use VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
22736 (rs6000_gimple_fold_mma_builtin): Use MMA_BUILTIN_DISASSEMBLE_PAIR
22737 and VSX_BUILTIN_ASSEMBLE_PAIR.
22738 (rs6000_init_builtins): Register compatibility built-ins.
22739 (mma_init_builtins): Use VSX_BUILTIN_ASSEMBLE_PAIR,
22740 VSX_BUILTIN_ASSEMBLE_PAIR_INTERNAL, VSX_BUILTIN_DISASSEMBLE_PAIR and
22741 VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
22742 * doc/extend.texi (__builtin_mma_assemble_pair): Rename from this...
22743 (__builtin_vsx_assemble_pair): ...to this.
22744 (__builtin_mma_disassemble_pair): Rename from this...
22745 (__builtin_vsx_disassemble_pair): ...to this.
22747 2021-02-23 Martin Liska <mliska@suse.cz>
22750 * ipa-icf.c (sem_variable::merge): Do not merge 2 variables
22751 with different alignment. That leads to an invalid red zone
22752 size allocated in runtime.
22754 2021-02-23 Jakub Jelinek <jakub@redhat.com>
22756 PR tree-optimization/99204
22757 * fold-const.c (fold_read_from_constant_string): Check that
22758 tree_fits_uhwi_p (index) rather than just that index is INTEGER_CST.
22760 2021-02-23 Segher Boessenkool <segher@kernel.crashing.org>
22761 Kewen Lin <linkw@gcc.gnu.org>
22763 * config/rs6000/rs6000.md (*rotl<mode>3_insert_3): Renamed to...
22764 (rotl<mode>3_insert_3): ...this.
22765 (plus_ior_xor): New code_iterator.
22766 (define_split for GPR rl*imi): New splitter.
22767 * config/rs6000/vsx.md (vsx_init_v4si): Use gen_rotldi3_insert_3
22768 for integer merging.
22770 2021-02-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22772 * config/aarch64/aarch64-tuning-flags.def (cse_sve_vl_constants):
22774 * config/aarch64/aarch64.md (add<mode>3): Force CONST_POLY_INT immediates
22775 into a register when the above is enabled.
22776 * config/aarch64/aarch64.c (neoversev1_tunings):
22777 AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
22778 (aarch64_rtx_costs): Use AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
22780 2021-02-22 Hans-Peter Nilsson <hp@axis.com>
22782 * config/cris/cris.c (cris_print_operand) <'T'>: Change
22783 valid operand from is now an addi mult-value to shift-value.
22784 * config/cris/cris.md (*addi): Change expression of scaled
22785 operand from mult to ashift.
22786 * config/cris/cris.md (*addi_reload): New insn_and_split.
22788 2021-02-22 John David Anglin <danglin@gcc.gnu.org>
22791 * config/pa/pa.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Define as
22792 hook_bool_const_tree_hwi_hwi_const_tree_true.
22793 (pa_asm_output_mi_thunk): Add support for nonzero vcall_offset.
22795 2021-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
22797 PR rtl-optimization/98791
22798 * ira-conflicts.c (process_regs_for_copy): Don't create allocno copies
22799 for unordered modes.
22801 2021-02-22 Martin Liska <mliska@suse.cz>
22803 * tree-inline.c (inline_forbidden_p): Set
22804 inline_forbidden_reason.
22806 2021-02-22 Richard Biener <rguenther@suse.de>
22808 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Dump
22811 2021-02-22 Richard Biener <rguenther@suse.de>
22813 PR tree-optimization/99165
22814 * gimple-ssa-store-merging.c (pass_store_merging::process_store):
22815 Accumulate changed to ret.
22817 2021-02-21 Uros Bizjak <ubizjak@gmail.com>
22820 2020-12-09 Uroš Bizjak <ubizjak@gmail.com>
22822 * config/i386/i386.h (REG_ALLOC_ORDER): Remove
22824 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
22827 * config/s390/vector.md (trunctf<DFP_ALL:mode>2_vr): New
22829 (trunctf<DFP_ALL:mode>2): Likewise.
22830 (trunctdtf2_vr): Likewise.
22831 (trunctdtf2): Likewise.
22832 (extend<DFP_ALL:mode>tf2_vr): Likewise.
22833 (extend<DFP_ALL:mode>tf2): Likewise.
22834 (extendtftd2_vr): Likewise.
22835 (extendtftd2): Likewise.
22837 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
22839 * config/s390/vector.md (*fprx2_to_tf): Rename to fprx2_to_tf,
22840 add memory alternative.
22841 (tf_to_fprx2): New pattern.
22843 2021-02-19 Martin Sebor <msebor@redhat.com>
22846 * attribs.c (init_attr_rdwr_indices): Guard vblist use.
22847 (attr_access::free_lang_data): Remove a spurious test.
22849 2021-02-19 Nathan Sidwell <nathan@acm.org>
22851 * doc/invoke.texi (flang-info-module-read): Document.
22853 2021-02-19 Martin Liska <mliska@suse.cz>
22855 PR translation/99167
22856 * params.opt: Fix typo.
22858 2021-02-19 Richard Biener <rguenther@suse.de>
22860 PR middle-end/99122
22861 * tree-inline.c (inline_forbidden_p): Do not inline functions
22862 with VLA arguments or return value.
22864 2021-02-19 Jakub Jelinek <jakub@redhat.com>
22867 * config/arm/arm.md (*stack_protect_combined_set_insn,
22868 *stack_protect_combined_test_insn): If force_const_mem result
22869 is not valid general operand, force its address into the destination
22872 2021-02-19 Jakub Jelinek <jakub@redhat.com>
22875 * tree-cfg.c (gimple_merge_blocks): If bb a starts with eh landing
22876 pad or non-local label, put FORCED_LABELs from bb b after that label
22877 rather than before it.
22879 2021-02-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
22882 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3): Use
22883 expand_vector_broadcast' to emit the vec_duplicate operand.
22885 2021-02-18 Vladimir N. Makarov <vmakarov@redhat.com>
22887 PR rtl-optimization/96264
22888 * lra-remat.c (reg_overlap_for_remat_p): Check also output insn
22891 2021-02-18 H.J. Lu <hjl.tools@gmail.com>
22894 * varasm.c (get_section): Replace SUPPORTS_SHF_GNU_RETAIN with
22895 looking up the retain attribute.
22896 (resolve_unique_section): Likewise.
22897 (get_variable_section): Likewise.
22898 (switch_to_section): Likewise. Warn when a symbol without the
22899 retain attribute and a symbol with the retain attribute are
22900 placed in the section with the same name, instead of the used
22902 * doc/extend.texi: Document the "retain" attribute.
22904 2021-02-18 Nathan Sidwell <nathan@acm.org>
22907 * doc/invoke.texi (flang-info-include-translate): Document header
22910 2021-02-18 Richard Biener <rguenther@suse.de>
22912 PR middle-end/99122
22913 * ipa-fnsummary.c (analyze_function_body): Set
22914 CIF_FUNCTION_NOT_INLINABLE for VLA parameter calls.
22915 * tree-inline.c (insert_init_debug_bind): Pass NULL for
22916 error_mark_node values.
22917 (force_value_to_type): Do not build V_C_Es for WITH_SIZE_EXPR
22919 (setup_one_parameter): Delay force_value_to_type until when
22922 2021-02-18 Hans-Peter Nilsson <hp@axis.com>
22924 PR tree-optimization/99142
22925 * match.pd (clz cmp 0): Gate replacement on single_use of clz result.
22927 2021-02-18 Jakub Jelinek <jakub@redhat.com>
22929 * wide-int-bitmask.h (wide_int_bitmask::wide_int_bitmask (),
22930 wide_int_bitmask::wide_int_bitmask (uint64_t),
22931 wide_int_bitmask::wide_int_bitmask (uint64_t, uint64_t),
22932 wide_int_bitmask::operator ~ () const,
22933 wide_int_bitmask::operator | (wide_int_bitmask) const,
22934 wide_int_bitmask::operator & (wide_int_bitmask) const): Use constexpr
22936 * config/i386/i386.h (PTA_3DNOW, PTA_3DNOW_A, PTA_64BIT, PTA_ABM,
22937 PTA_AES, PTA_AVX, PTA_BMI, PTA_CX16, PTA_F16C, PTA_FMA, PTA_FMA4,
22938 PTA_FSGSBASE, PTA_LWP, PTA_LZCNT, PTA_MMX, PTA_MOVBE, PTA_NO_SAHF,
22939 PTA_PCLMUL, PTA_POPCNT, PTA_PREFETCH_SSE, PTA_RDRND, PTA_SSE, PTA_SSE2,
22940 PTA_SSE3, PTA_SSE4_1, PTA_SSE4_2, PTA_SSE4A, PTA_SSSE3, PTA_TBM,
22941 PTA_XOP, PTA_AVX2, PTA_BMI2, PTA_RTM, PTA_HLE, PTA_PRFCHW, PTA_RDSEED,
22942 PTA_ADX, PTA_FXSR, PTA_XSAVE, PTA_XSAVEOPT, PTA_AVX512F, PTA_AVX512ER,
22943 PTA_AVX512PF, PTA_AVX512CD, PTA_NO_TUNE, PTA_SHA, PTA_PREFETCHWT1,
22944 PTA_CLFLUSHOPT, PTA_XSAVEC, PTA_XSAVES, PTA_AVX512DQ, PTA_AVX512BW,
22945 PTA_AVX512VL, PTA_AVX512IFMA, PTA_AVX512VBMI, PTA_CLWB, PTA_MWAITX,
22946 PTA_CLZERO, PTA_NO_80387, PTA_PKU, PTA_AVX5124VNNIW, PTA_AVX5124FMAPS,
22947 PTA_AVX512VPOPCNTDQ, PTA_SGX, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES,
22948 PTA_AVX512VBMI2, PTA_VPCLMULQDQ, PTA_AVX512BITALG, PTA_RDPID,
22949 PTA_PCONFIG, PTA_WBNOINVD, PTA_AVX512VP2INTERSECT, PTA_PTWRITE,
22950 PTA_AVX512BF16, PTA_WAITPKG, PTA_MOVDIRI, PTA_MOVDIR64B, PTA_ENQCMD,
22951 PTA_CLDEMOTE, PTA_SERIALIZE, PTA_TSXLDTRK, PTA_AMX_TILE, PTA_AMX_INT8,
22952 PTA_AMX_BF16, PTA_UINTR, PTA_HRESET, PTA_KL, PTA_WIDEKL, PTA_AVXVNNI,
22953 PTA_X86_64_BASELINE, PTA_X86_64_V2, PTA_X86_64_V3, PTA_X86_64_V4,
22954 PTA_CORE2, PTA_NEHALEM, PTA_WESTMERE, PTA_SANDYBRIDGE, PTA_IVYBRIDGE,
22955 PTA_HASWELL, PTA_BROADWELL, PTA_SKYLAKE, PTA_SKYLAKE_AVX512,
22956 PTA_CASCADELAKE, PTA_COOPERLAKE, PTA_CANNONLAKE, PTA_ICELAKE_CLIENT,
22957 PTA_ICELAKE_SERVER, PTA_TIGERLAKE, PTA_SAPPHIRERAPIDS, PTA_ALDERLAKE,
22958 PTA_KNL, PTA_BONNELL, PTA_SILVERMONT, PTA_GOLDMONT, PTA_GOLDMONT_PLUS,
22959 PTA_TREMONT, PTA_KNM): Use constexpr instead of const.
22961 2021-02-18 Jakub Jelinek <jakub@redhat.com>
22963 PR middle-end/99109
22964 * gimple-array-bounds.cc (build_zero_elt_array_type): Rename to ...
22965 (build_printable_array_type): ... this. Add nelts argument. For
22966 overaligned eltype, use TYPE_MAIN_VARIANT (eltype) instead. If
22967 nelts, call build_array_type_nelts.
22968 (array_bounds_checker::check_mem_ref): Use build_printable_array_type
22969 instead of build_zero_elt_array_type and build_array_type_nelts.
22971 2021-02-18 Jakub Jelinek <jakub@redhat.com>
22974 * config/i386/i386.c (distance_non_agu_define): Don't call
22975 extract_insn_cached here.
22976 (ix86_lea_outperforms): Save and restore recog_data around call
22977 to distance_non_agu_define and distance_agu_use.
22978 (ix86_ok_to_clobber_flags): Remove.
22979 (ix86_avoid_lea_for_add): Don't call ix86_ok_to_clobber_flags.
22980 (ix86_avoid_lea_for_addr): Likewise. Adjust function comment.
22981 * config/i386/i386.md (*lea<mode>): Change from define_insn_and_split
22982 into define_insn. Move the splitting to define_peephole2 and
22983 check there using peep2_regno_dead_p if FLAGS_REG is dead.
22985 2021-02-17 Julian Brown <julian@codesourcery.com>
22987 * gimplify.c (gimplify_scan_omp_clauses): Handle ATTACH_DETACH
22990 2021-02-17 Xi Ruoyao <xry111@mengyan1223.wang>
22993 * config/mips/mips.c (mips_symbol_insns): Do not use
22994 MSA_SUPPORTED_MODE_P if mode is MAX_MACHINE_MODE.
22996 2021-02-16 Vladimir N. Makarov <vmakarov@redhat.com>
22998 PR inline-asm/98096
22999 * stmt.c (resolve_operand_name_1): Take inout operands into account
23000 for access to labels by names.
23001 * doc/extend.texi: Describe counting operands for accessing labels.
23003 2021-02-16 Richard Biener <rguenther@suse.de>
23005 PR tree-optimization/38474
23006 * tree-ssa-structalias.c (variable_info::address_taken): New.
23007 (new_var_info): Initialize address_taken.
23008 (process_constraint): Set address_taken.
23009 (solve_constraints): Use the new address_taken flag rather
23010 than is_reg_var for sorting variables.
23011 (dump_constraint): Dump the variable number if the name
23014 2021-02-16 Jakub Jelinek <jakub@redhat.com>
23017 * tree-vect-stmts.c (vectorizable_simd_clone_call): For num_calls != 1
23018 multiply by 4096 and for inbranch by 8192.
23019 * config/i386/i386.c (ix86_simd_clone_usable): For TARGET_AVX512F,
23020 return 3, 2 or 1 for mangle letters 'b', 'c' or 'd'.
23022 2021-02-15 Maya Rashish <coypu@sdf.org>
23024 * config/aarch64/aarch64.c (aarch64_init_builtins):
23025 Call SUBTARGET_INIT_BUILTINS.
23027 2021-02-15 Peter Bergner <bergner@linux.ibm.com>
23029 PR rtl-optimization/98872
23030 * init-regs.c (initialize_uninitialized_regs): Skip initialization
23031 if CONST0_RTX is NULL.
23033 2021-02-15 Richard Sandiford <richard.sandiford@arm.com>
23035 PR rtl-optimization/98863
23036 * rtl-ssa/functions.h (function_info::bb_live_out_info): Delete.
23037 (function_info::build_info): Turn into a declaration, moving the
23038 definition to internals.h.
23039 (function_info::bb_walker): Declare.
23040 (function_info::create_reg_use): Likewise.
23041 (function_info::calculate_potential_phi_regs): Take a build_info
23043 (function_info::place_phis, function_info::create_ebbs): Declare.
23044 (function_info::calculate_ebb_live_in_for_debug): Likewise.
23045 (function_info::populate_backedge_phis): Delete.
23046 (function_info::start_block, function_info::end_block): Declare.
23047 (function_info::populate_phi_inputs): Delete.
23048 (function_info::m_potential_phi_regs): Move information to build_info.
23049 * rtl-ssa/internals.h: New file.
23050 (function_info::bb_phi_info): New class.
23051 (function_info::build_info): Moved from functions.h.
23052 Add a constructor and destructor.
23053 (function_info::build_info::ebb_use): Delete.
23054 (function_info::build_info::ebb_def): Likewise.
23055 (function_info::build_info::bb_live_out): Likewise.
23056 (function_info::build_info::tmp_ebb_live_in_for_debug): New variable.
23057 (function_info::build_info::potential_phi_regs): Likewise.
23058 (function_info::build_info::potential_phi_regs_for_debug): Likewise.
23059 (function_info::build_info::ebb_def_regs): Likewise.
23060 (function_info::build_info::bb_phis): Likewise.
23061 (function_info::build_info::bb_mem_live_out): Likewise.
23062 (function_info::build_info::bb_to_rpo): Likewise.
23063 (function_info::build_info::def_stack): Likewise.
23064 (function_info::build_info::old_def_stack_limit): Likewise.
23065 * rtl-ssa/internals.inl (function_info::build_info::record_reg_def):
23066 Remove the regno argument. Push the previous definition onto the
23067 definition stack where necessary.
23068 * rtl-ssa/accesses.cc: Include internals.h.
23069 * rtl-ssa/changes.cc: Likewise.
23070 * rtl-ssa/blocks.cc: Likewise.
23071 (function_info::build_info::build_info): Define.
23072 (function_info::build_info::~build_info): Likewise.
23073 (function_info::bb_walker): New class.
23074 (function_info::bb_walker::bb_walker): Define.
23075 (function_info::add_live_out_use): Convert a logarithmic-complexity
23076 test into a linear one. Allow the same definition to be passed
23078 (function_info::calculate_potential_phi_regs): Moved from
23079 functions.cc. Take a build_info parameter and store the
23080 information there instead.
23081 (function_info::place_phis): New function.
23082 (function_info::add_entry_block_defs): Update call to record_reg_def.
23083 (function_info::calculate_ebb_live_in_for_debug): New function.
23084 (function_info::add_phi_nodes): Use bb_phis to decide which
23085 registers need phi nodes and initialize ebb_def_regs accordingly.
23086 Do not add degenerate phis here.
23087 (function_info::add_artificial_accesses): Use create_reg_use.
23088 Assert that all definitions are listed in the DF LR sets.
23089 Update call to record_reg_def.
23090 (function_info::record_block_live_out): Record live-out register
23091 values in the phis of successor blocks. Use the live-out set
23092 when processing the last block in an EBB, instead of always
23093 using the live-in sets of successor blocks. AND the live sets
23094 with the set of registers that have been defined in the EBB,
23095 rather than with all potential phi registers. Cope correctly
23096 with branches back to the start of the current EBB.
23097 (function_info::start_block): New function.
23098 (function_info::end_block): Likewise.
23099 (function_info::populate_phi_inputs): Likewise.
23100 (function_info::create_ebbs): Likewise.
23101 (function_info::process_all_blocks): Rewrite into a multi-phase
23103 * rtl-ssa/functions.cc: Include internals.h.
23104 (function_info::calculate_potential_phi_regs): Move to blocks.cc.
23105 (function_info::init_function_data): Remove caller.
23106 * rtl-ssa/insns.cc: Include internals.h
23107 (function_info::create_reg_use): New function. Lazily any
23108 degenerate phis needed by the linear RPO view.
23109 (function_info::record_use): Use create_reg_use. When processing
23110 debug uses, use potential_phi_regs and test it before checking
23111 whether the register is live on entry to the current EBB. Lazily
23112 calculate ebb_live_in_for_debug.
23113 (function_info::record_call_clobbers): Update call to record_reg_def.
23114 (function_info::record_def): Likewise.
23116 2021-02-15 Martin Liska <mliska@suse.cz>
23118 * toplev.c (init_asm_output): Free output of
23119 gen_command_line_string function.
23120 (process_options): Likewise.
23122 2021-02-15 Martin Liska <mliska@suse.cz>
23124 * params.opt: Add 2 missing Param keywords.
23126 2021-02-15 Eric Botcazou <ebotcazou@adacore.com>
23128 * df-core.c (df_worklist_dataflow_doublequeue): Use proper cast.
23130 2021-02-15 Jakub Jelinek <jakub@redhat.com>
23132 PR tree-optimization/99079
23133 * match.pd (A % (pow2pcst << N) -> A & ((pow2pcst << N) - 1)): Remove
23134 useless tree_nop_conversion_p (type, TREE_TYPE (@3)) check. Instead
23135 require both type and TREE_TYPE (@1) to be integral types and either
23136 type having smaller or equal precision, or TREE_TYPE (@1) being
23137 unsigned type, or type being signed type. If TREE_TYPE (@1)
23138 doesn't have wrapping overflow, perform the subtraction of one in
23141 2021-02-14 Jan Hubicka <hubicka@ucw.cz>
23142 Richard Biener <rguether@suse.de>
23145 * ipa-reference.c (ipa_init): Only conditinally initialize
23146 reference_vars_to_consider.
23147 (propagate): Conditionally deninitialize reference_vars_to_consider.
23148 (ipa_reference_write_optimization_summary): Sanity check that
23149 reference_vars_to_consider is not allocated.
23151 2021-02-13 Levy Hsu <admin@levyhsu.com>
23154 * config/riscv/riscv-shorten-memrefs.c (pass_shorten_memrefs): Add
23155 extend parameter to get_si_mem_base_reg declaration.
23156 (get_si_mem_base_reg): Add extend parameter. Set it.
23157 (analyze): Pass extend arg to get_si_mem_base_reg.
23158 (transform): Likewise. Use it when rewriting mems.
23159 * config/riscv/riscv.c (riscv_legitimize_move): Check for subword
23160 loads and emit sign/zero extending load followed by subreg move.
23162 2021-02-13 Jim Wilson <jimw@sifive.com>
23165 * config/riscv/riscv.c (riscv_compressed_lw_address_p): Drop early
23166 exit when !reload_completed. Only perform check for compressed reg
23167 if reload_completed.
23168 (riscv_rtx_costs): In MEM case, when optimizing for size and
23169 shorten memrefs, if not compressible, then increase cost.
23171 2021-02-13 Jakub Jelinek <jakub@redhat.com>
23173 PR rtl-optimization/98439
23174 * recog.c (pass_split_before_regstack::gate): Enable even when
23175 pass_split_before_sched2 is enabled if -fselective-scheduling2 is
23178 2021-02-13 Jakub Jelinek <jakub@redhat.com>
23181 * config/i386/mmx.md (*mmx_pshufd_1): Add a combine splitter for
23182 swap of V2SImode elements in memory into DImode memory rotate by 32.
23184 2021-02-12 Martin Sebor <msebor@redhat.com>
23186 * tree-pretty-print.c (print_generic_expr_to_str): Update comment.
23188 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
23190 * rtl-ssa/accesses.cc (function_info::make_use_available): Use
23191 m_temp_obstack rather than m_obstack to allocate the temporary use.
23193 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
23195 * df-problems.c (df_lr_bb_local_compute): Treat partial definitions
23196 as read-modify operations.
23198 2021-02-12 Richard Biener <rguenther@suse.de>
23200 PR middle-end/38474
23201 * ipa-fnsummary.c (unmodified_parm_1): Only walk when
23202 fbi->aa_walk_budget is bigger than zero. Update
23203 fbi->aa_walk_budget.
23204 (param_change_prob): Likewise.
23205 * ipa-prop.c (detect_type_change_from_memory_writes):
23206 Properly account walk_aliased_vdefs.
23207 (parm_preserved_before_stmt_p): Canonicalize updates.
23208 (parm_ref_data_preserved_p): Likewise.
23209 (parm_ref_data_pass_through_p): Likewise.
23210 (determine_known_aggregate_parts): Account own alias queries.
23212 2021-02-12 Martin Liska <mliska@suse.cz>
23214 * opts-common.c (decode_cmdline_option): Release werror_arg.
23215 * opts.c (gen_producer_string): Release output of
23216 gen_command_line_string.
23218 2021-02-12 Richard Biener <rguenther@suse.de>
23220 PR tree-optimization/38474
23221 * params.opt (-param=max-store-chains-to-track=): New param.
23222 (-param=max-stores-to-track=): Likewise.
23223 * doc/invoke.texi (max-store-chains-to-track): Document.
23224 (max-stores-to-track): Likewise.
23225 * gimple-ssa-store-merging.c (pass_store_merging::m_n_chains):
23227 (pass_store_merging::m_n_stores): Likewise.
23228 (pass_store_merging::terminate_and_process_chain): Update
23229 m_n_stores and m_n_chains.
23230 (pass_store_merging::process_store): Likewise. Terminate
23231 oldest chains if the number of stores or chains get too large.
23232 (imm_store_chain_info::terminate_and_process_chain): Dump
23235 2021-02-11 Eric Botcazou <ebotcazou@adacore.com>
23237 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
23238 the cold section, emit a nop before the directive if the previous
23239 active instruction can throw.
23241 2021-02-11 Peter Bergner <bergner@linux.ibm.com>
23244 * config/rs6000/predicates.md (mma_assemble_input_operand): Restrict
23245 memory addresses that are legal for quad word accesses.
23247 2021-02-11 Andrea Corallo <andrea.corallo@arm.com>
23250 * config/arm/thumb2.md (*doloop_end_internal): Generate
23251 alternative sequence to handle long range branches.
23253 2021-02-11 Joel Hutton <joel.hutton@arm.com>
23255 PR tree-optimization/98772
23256 * optabs-tree.c (supportable_half_widening_operation): New function
23257 to check for supportable V8QI->V8HI widening patterns.
23258 * optabs-tree.h (supportable_half_widening_operation): New function.
23259 * tree-vect-stmts.c (vect_create_half_widening_stmts): New function
23260 to create promotion stmts for V8QI->V8HI widening patterns.
23261 (vectorizable_conversion): Add case for V8QI->V8HI.
23263 2021-02-11 Richard Biener <rguenther@suse.de>
23265 * sparseset.h (SPARSESET_ELT_BITS): Remove.
23266 (SPARSESET_ELT_TYPE): Use unsigned int.
23267 * fwprop.c: Do not include sparseset.h.
23269 2021-02-10 Jakub Jelinek <jakub@redhat.com>
23272 * varasm.c (declare_weak): For -fsyntax-only, allow even
23273 TREE_ASM_WRITTEN function decls.
23275 2021-02-10 Jakub Jelinek <jakub@redhat.com>
23278 * config/i386/sse.md (fix<fixunssuffix>_truncv2sfv2di2,
23279 <insn>v8qiv8hi2, <insn>v8qiv8si2, <insn>v4qiv4si2, <insn>v4hiv4si2,
23280 <insn>v8qiv8di2, <insn>v4qiv4di2, <insn>v2qiv2di2, <insn>v4hiv4di2,
23281 <insn>v2hiv2di2, <insn>v2siv2di2): Force operands[1] into REG before
23282 calling simplify_gen_subreg on it.
23284 2021-02-10 Martin Liska <mliska@suse.cz>
23286 * config/nvptx/nvptx.c (nvptx_option_override): Use
23287 flag_patchable_function_entry instead of the removed
23288 function_entry_patch_area_size.
23290 2021-02-10 Martin Liska <mliska@suse.cz>
23292 PR tree-optimization/99002
23293 PR tree-optimization/99026
23294 * gimple-if-to-switch.cc (if_chain::is_beneficial): Fix memory
23295 leak when adjacent cases are merged.
23296 * tree-switch-conversion.c (switch_decision_tree::analyze_switch_statement): Use
23298 (make_pass_lower_switch): Remove trailing whitespace.
23299 * tree-switch-conversion.h (release_clusters): New.
23301 2021-02-10 Richard Biener <rguenther@suse.de>
23303 PR rtl-optimization/99054
23304 * cfgrtl.c (rtl-optimization/99054): Return an auto_vec.
23305 (fixup_partitions): Adjust.
23306 (rtl_verify_edges): Likewise.
23308 2021-02-10 Jakub Jelinek <jakub@redhat.com>
23310 PR middle-end/99007
23311 * gimplify.c (gimplify_scan_omp_clauses): For MEM_REF on reductions,
23312 temporarily disable gimplify_ctxp->into_ssa around gimplify_expr
23315 2021-02-10 Richard Biener <rguenther@suse.de>
23318 * ipa-pure-const.c (propagate_malloc): Use an auto_vec<>
23321 2021-02-10 Richard Biener <rguenther@suse.de>
23323 PR tree-optimization/99024
23324 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Only
23325 clear loop->aux if it is associated with the destroyed loop_vinfo.
23327 2021-02-10 Martin Liska <mliska@suse.cz>
23329 PR tree-optimization/99002
23330 * gimple-if-to-switch.cc (find_conditions): Fix memory leak
23333 2021-02-10 Martin Liska <mliska@suse.cz>
23336 * ipa-icf.c (sem_item::add_reference): Fix memory leak when
23337 a reference exists.
23339 2021-02-10 Jakub Jelinek <jakub@redhat.com>
23342 * dwarf2out.c (prune_unused_types_walk): Mark DW_TAG_variable DIEs
23343 at class scope for DWARF5+.
23345 2021-02-09 Eric Botcazou <ebotcazou@adacore.com>
23347 PR rtl-optimization/96015
23348 * reorg.c (skip_consecutive_labels): Minor comment tweaks.
23349 (relax_delay_slots): When deleting a jump to the next active
23350 instruction over a barrier, first delete the barrier if the
23351 jump is the only way to reach the target label.
23353 2021-02-09 Andre Vieira <andre.simoesdiasvieira@arm.com>
23355 * config/aarch64/aarch64-cost-tables.h: Add entries for vect.mul.
23356 * config/aarch64/aarch64.c (aarch64_rtx_mult_cost): Use vect.mul for
23357 vector multiplies and vect.alu for SSRA.
23358 * config/arm/aarch-common-protos.h (struct vector_cost_table): Define
23359 vect.mul cost field.
23360 * config/arm/aarch-cost-tables.h: Add entries for vect.mul.
23361 * config/arm/arm.c: Likewise.
23363 2021-02-09 Richard Biener <rguenther@suse.de>
23365 PR tree-optimization/98863
23366 * tree-ssa-sccvn.h (vn_avail::next_undo): Add.
23367 * tree-ssa-sccvn.c (last_pushed_avail): New global.
23368 (rpo_elim::eliminate_push_avail): Chain pushed avails.
23369 (unwind_state::avail_top): Add.
23370 (do_unwind): Rewrite unwinding of avail entries.
23371 (do_rpo_vn): Initialize last_pushed_avail and
23372 avail_top of the undo state.
23374 2021-02-09 Jakub Jelinek <jakub@redhat.com>
23376 PR middle-end/99004
23377 * calls.c (maybe_warn_rdwr_sizes): Change s0 and s1 type from
23378 const char * to char * and free those pointers after use.
23380 2021-02-09 Richard Biener <rguenther@suse.de>
23382 PR tree-optimization/99017
23383 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Allow
23384 zero vector cost entries.
23386 2021-02-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
23388 PR middle-end/98974
23389 * tree-vect-stmts.c (vectorizable_condition): Remove shadow vec_num
23390 parameter in vectorizable_condition.
23392 2021-02-08 Richard Biener <rguenther@suse.de>
23395 * tree.c (walk_tree_1): Walk VECTOR_CST elements.
23397 2021-02-08 Martin Liska <mliska@suse.cz>
23400 * cfgexpand.c (pass_expand::execute): Parse per-function option
23401 flag_patchable_function_entry and use it.
23402 * common.opt: Remove function_entry_patch_area_size and
23403 function_entry_patch_area_start global variables.
23404 * opts.c (parse_and_check_patch_area): New function.
23405 (common_handle_option): Use it.
23406 * opts.h (parse_and_check_patch_area): New function.
23407 * toplev.c (process_options): Parse and use
23408 function_entry_patch_area_size.
23410 2021-02-08 Martin Sebor <msebor@redhat.com>
23412 * doc/extend.texi (attribute malloc): Correct typos.
23414 2021-02-05 Nathan Sidwell <nathan@acm.org>
23417 * gcc.c (driver::maybe_run_linker): Check for input file
23418 accessibility if not linking.
23420 2021-02-05 Richard Biener <rguenther@suse.de>
23422 PR tree-optimization/98855
23423 * tree-vectorizer.h (add_stmt_cost): New overload.
23424 * tree-vect-slp.c (li_cost_vec_cmp): New.
23425 (vect_bb_slp_scalar_cost): Cost individual loop regions
23426 separately. Account for the scalar instance root stmt.
23428 2021-02-05 Tom de Vries <tdevries@suse.de>
23431 * tree-switch-conversion.c (jump_table_cluster::emit): Add loc
23433 (bit_test_cluster::emit): Reuse location_t for newly created
23435 (switch_decision_tree::try_switch_expansion): Preserve
23437 * tree-switch-conversion.h: Change function signatures.
23439 2021-02-05 Jakub Jelinek <jakub@redhat.com>
23442 * config/i386/i386-options.c (m_NONE, m_ALL): Define.
23443 * config/i386/x86-tune.def (X86_TUNE_BRANCH_PREDICTION_HINTS,
23444 X86_TUNE_PROMOTE_QI_REGS): Use m_NONE instead of 0U.
23445 (X86_TUNE_QIMODE_MATH): Use m_ALL instead of ~0U.
23447 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23449 * config/aarch64/aarch64-simd-builtins.def (get_high): Define builtin.
23450 * config/aarch64/aarch64-simd.md (aarch64_get_high<mode>): Define.
23451 * config/aarch64/arm_neon.h (__GET_HIGH): Delete.
23452 (vget_high_f16): Reimplement using new builtin.
23453 (vget_high_f32): Likewise.
23454 (vget_high_f64): Likewise.
23455 (vget_high_p8): Likewise.
23456 (vget_high_p16): Likewise.
23457 (vget_high_p64): Likewise.
23458 (vget_high_s8): Likewise.
23459 (vget_high_s16): Likewise.
23460 (vget_high_s32): Likewise.
23461 (vget_high_s64): Likewise.
23462 (vget_high_u8): Likewise.
23463 (vget_high_u16): Likewise.
23464 (vget_high_u32): Likewise.
23465 (vget_high_u64): Likewise.
23467 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23469 * config/aarch64/aarch64-simd-builtins.def (get_low): Define builtin.
23470 * config/aarch64/aarch64-simd.md (aarch64_get_low<mode>): Define.
23471 * config/aarch64/arm_neon.h (__GET_LOW): Delete.
23472 (vget_low_f16): Reimplement using new builtin.
23473 (vget_low_f32): Likewise.
23474 (vget_low_f64): Likewise.
23475 (vget_low_p8): Likewise.
23476 (vget_low_p16): Likewise.
23477 (vget_low_p64): Likewise.
23478 (vget_low_s8): Likewise.
23479 (vget_low_s16): Likewise.
23480 (vget_low_s32): Likewise.
23481 (vget_low_s64): Likewise.
23482 (vget_low_u8): Likewise.
23483 (vget_low_u16): Likewise.
23484 (vget_low_u32): Likewise.
23485 (vget_low_u64): Likewise.
23487 2021-02-05 Kito Cheng <kito.cheng@sifive.com>
23489 * gcc.c (print_multilib_info): Check all required argument is provided
23492 2021-02-05 liuhongt <hongtao.liu@intel.com>
23495 * config/i386/i386-expand.c (ix86_expand_sse_cmp): Don't
23496 generate integer mask comparison for 128/256-bits vector when
23497 op_true/op_false is NULL_RTX or CONSTM1_RTX/CONST0_RTX. Also
23498 delete redundant !maskcmp condition.
23499 (ix86_expand_int_vec_cmp): Ditto but no redundant deletion
23501 (ix86_expand_sse_movcc): Delete definition of maskcmp, add the
23502 condition directly to if (maskcmp), add extra check for
23503 cmpmode, it should be MODE_INT.
23504 (ix86_expand_fp_vec_cmp): Pass NULL to ix86_expand_sse_cmp's
23505 parameters op_true/op_false.
23506 (ix86_use_mask_cmp_p): New.
23508 2021-02-05 liuhongt <hongtao.liu@intel.com>
23511 * config/i386/x86-tune.def (X86_TUNE_AVX256_UNALIGNED_LOAD_OPTIMAL):
23512 Remove m_GENERIC from ~list.
23513 (X86_TUNE_AVX256_UNALIGNED_STORE_OPTIMAL): Ditto.
23515 2021-02-04 David Malcolm <dmalcolm@redhat.com>
23518 * diagnostic-show-locus.c (compatible_locations_p): Require
23519 locations in the same macro map to be either both from the
23520 macro definition, or both from the macro arguments.
23522 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
23524 * config/aarch64/aarch64-simd-builtins.def: Add
23525 [su]mull_hi_lane[q] builtin generator macros.
23526 * config/aarch64/aarch64-simd.md
23527 (aarch64_<su>mull_hi_lane<mode>_insn): Define.
23528 (aarch64_<su>mull_hi_lane<mode>): Define.
23529 (aarch64_<su>mull_hi_laneq<mode>_insn): Define.
23530 (aarch64_<su>mull_hi_laneq<mode>): Define.
23531 * config/aarch64/arm_neon.h (vmull_high_lane_s16): Use RTL
23532 builtin instead of inline asm.
23533 (vmull_high_lane_s32): Likewise.
23534 (vmull_high_lane_u16): Likewise.
23535 (vmull_high_lane_u32): Likewise.
23536 (vmull_high_laneq_s16): Likewise.
23537 (vmull_high_laneq_s32): Likewise.
23538 (vmull_high_laneq_u16): Likewise.
23539 (vmull_high_laneq_u32): Liekwise.
23541 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
23543 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_hi_n
23544 builtin generator macros.
23545 * config/aarch64/aarch64-simd.md
23546 (aarch64_<su>mull_hi_n<mode>_insn): Define.
23547 (aarch64_<su>mull_hi_n<mode>): Define.
23548 * config/aarch64/arm_neon.h (vmull_high_n_s16): Use RTL builtin
23549 instead of inline asm.
23550 (vmull_high_n_s32): Likewise.
23551 (vmull_high_n_u16): Likewise.
23552 (vmull_high_n_u32): Likewise.
23554 2021-02-04 Richard Biener <rguenther@suse.de>
23556 PR tree-optimization/98855
23557 * tree-vect-loop.c (vectorizable_phi): Do not cost
23558 single-argument PHIs.
23559 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Likewise.
23560 * tree-vect-stmts.c (vectorizable_bswap): Also perform
23561 costing for SLP operation.
23563 2021-02-04 Martin Liska <mliska@suse.cz>
23565 * doc/extend.texi: Mention -mprefer-vector-width in target
23568 2021-02-03 Martin Sebor <msebor@redhat.com>
23570 PR tree-optimization/98937
23571 * tree-ssa-strlen.c (strlen_dom_walker::~strlen_dom_walker): Define.
23572 Flush pointer_query cache.
23574 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
23576 * config/rs6000/genfusion.pl (gen_2logical): Add missing
23577 fixes based on patch review.
23578 * config/rs6000/fusion.md: Regenerate file.
23580 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
23582 * config/rs6000/t-rs6000: Comment out auto generation of
23585 2021-02-03 Andrew Stubbs <ams@codesourcery.com>
23587 * config/gcn/gcn-opts.h (enum processor_type): Add PROCESSOR_GFX908.
23588 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Add gfx908.
23589 (output_file_start): Add gfx908.
23590 * config/gcn/gcn.opt (gpu_type): Add gfx908.
23591 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Add march=gfx908.
23592 (MULTILIB_DIRNAMES): Add gfx908.
23593 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): New define.
23594 (main): Recognize gfx908.
23595 * config/gcn/t-omp-device: Add gfx908.
23597 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
23599 * config/aarch64/aarch64-simd-builtins.def: Add
23600 [su]mlsl_hi_lane[q] builtin macro generators.
23601 * config/aarch64/aarch64-simd.md
23602 (aarch64_<su>mlsl_hi_lane<mode>_insn): Define.
23603 (aarch64_<su>mlsl_hi_lane<mode>): Define.
23604 (aarch64_<su>mlsl_hi_laneq<mode>_insn): Define.
23605 (aarch64_<su>mlsl_hi_laneq<mode>): Define.
23606 * config/aarch64/arm_neon.h (vmlsl_high_lane_s16): Use RTL
23607 builtin instead of inline asm.
23608 (vmlsl_high_lane_s32): Likewise.
23609 (vmlsl_high_lane_u16): Likewise.
23610 (vmlsl_high_lane_u32): Likewise.
23611 (vmlsl_high_laneq_s16): Likewise.
23612 (vmlsl_high_laneq_s32): Likewise.
23613 (vmlsl_high_laneq_u16): Likewise.
23614 (vmlsl_high_laneq_u32): Likewise.
23615 (vmlal_high_laneq_u32): Likewise.
23617 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
23619 * config/aarch64/aarch64-simd-builtins.def: Add
23620 [su]mlal_hi_lane[q] builtin generator macros.
23621 * config/aarch64/aarch64-simd.md
23622 (aarch64_<su>mlal_hi_lane<mode>_insn): Define.
23623 (aarch64_<su>mlal_hi_lane<mode>): Define.
23624 (aarch64_<su>mlal_hi_laneq<mode>_insn): Define.
23625 (aarch64_<su>mlal_hi_laneq<mode>): Define.
23626 * config/aarch64/arm_neon.h (vmlal_high_lane_s16): Use RTL
23627 builtin instead of inline asm.
23628 (vmlal_high_lane_s32): Likewise.
23629 (vmlal_high_lane_u16): Likewise.
23630 (vmlal_high_lane_u32): Likewise.
23631 (vmlal_high_laneq_s16): Likewise.
23632 (vmlal_high_laneq_s32): Likewise.
23633 (vmlal_high_laneq_u16): Likewise.
23634 (vmlal_high_laneq_u32): Likewise.
23636 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
23638 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_hi_n
23639 builtin generator macros.
23640 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_hi_n<mode>_insn):
23642 (aarch64_<su>mlsl_hi_n<mode>): Define.
23643 * config/aarch64/arm_neon.h (vmlsl_high_n_s16): Use RTL builtin
23644 instead of inline asm.
23645 (vmlsl_high_n_s32): Likewise.
23646 (vmlsl_high_n_u16): Likewise.
23647 (vmlsl_high_n_u32): Likewise.
23649 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
23651 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_hi_n
23652 builtin generator macros.
23653 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_hi_n<mode>_insn):
23655 (aarch64_<su>mlal_hi_n<mode>): Define.
23656 * config/aarch64/arm_neon.h (vmlal_high_n_s16): Use RTL builtin
23657 instead of inline asm.
23658 (vmlal_high_n_s32): Likewise.
23659 (vmlal_high_n_u16): Likewise.
23660 (vmlal_high_n_u32): Likewise.
23662 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
23664 * config/aarch64/aarch64-simd-builtins.def: Add RTL builtin
23666 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal_hi<mode>):
23668 (aarch64_<su>mlal_hi<mode>_insn): This.
23669 (aarch64_<su>mlal_hi<mode>): Define.
23670 * config/aarch64/arm_neon.h (vmlal_high_s8): Use RTL builtin
23671 instead of inline asm.
23672 (vmlal_high_s16): Likewise.
23673 (vmlal_high_s32): Likewise.
23674 (vmlal_high_u8): Likewise.
23675 (vmlal_high_u16): Likewise.
23676 (vmlal_high_u32): Likewise.
23678 2021-02-03 Ilya Leoshkevich <iii@linux.ibm.com>
23680 * lra-spills.c (remove_pseudos): Call lra_update_insn_recog_data()
23681 after calling alter_subreg() on a (mem).
23683 2021-02-03 Martin Liska <mliska@suse.cz>
23686 * lto-streamer-out.c (produce_lto_section): Fill up missing
23688 * lto-streamer.h (struct lto_section): Add _padding field.
23690 2021-02-03 Richard Biener <rguenther@suse.de>
23692 * lto-streamer.c (lto_get_section_name): Free temporary
23694 * tree-loop-distribution.c
23695 (loop_distribution::merge_dep_scc_partitions): Free edge data.
23697 2021-02-03 Jakub Jelinek <jakub@redhat.com>
23699 PR middle-end/97487
23700 * ifcvt.c (noce_can_force_operand): New function.
23701 (noce_emit_move_insn): Use it.
23702 (noce_try_sign_mask): Likewise. Formatting fix.
23704 2021-02-03 Jakub Jelinek <jakub@redhat.com>
23706 PR middle-end/97971
23707 * lra-constraints.c (process_alt_operands): For inline asm, don't call
23708 fatal_insn, but instead return false.
23710 2021-02-03 Jakub Jelinek <jakub@redhat.com>
23712 PR tree-optimization/98287
23713 * config/i386/mmx.md (<insn><mode>3): For shifts don't enable expander
23716 2021-02-03 Tamar Christina <tamar.christina@arm.com>
23718 PR tree-optimization/98928
23719 * tree-vect-loop.c (vect_analyze_loop_2): Change
23720 STMT_VINFO_SLP_VECT_ONLY to STMT_VINFO_SLP_VECT_ONLY_PATTERN.
23721 * tree-vect-slp-patterns.c (complex_pattern::build): Likewise.
23722 * tree-vectorizer.h (STMT_VINFO_SLP_VECT_ONLY_PATTERN): New.
23723 (class _stmt_vec_info): Add slp_vect_pattern_only_p.
23725 2021-02-02 Richard Biener <rguenther@suse.de>
23727 * gimple-loop-interchange.cc (prepare_data_references):
23729 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
23730 * tree-ssa-loop-im.c (hoist_memory_references): Likewise.
23731 * tree-vect-stmts.c (vectorizable_condition): Do not
23733 (vectorizable_comparison): Likewise.
23735 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23737 * config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
23738 * config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
23739 * config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
23740 (vrsqrteq_u32): Likewise.
23742 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23744 * config/aarch64/aarch64-simd-builtins.def (sqxtun2): Define builtin.
23745 * config/aarch64/aarch64-simd.md (aarch64_sqxtun2<mode>_le): Define.
23746 (aarch64_sqxtun2<mode>_be): Likewise.
23747 (aarch64_sqxtun2<mode>): Likewise.
23748 * config/aarch64/arm_neon.h (vqmovun_high_s16): Reimplement using builtin.
23749 (vqmovun_high_s32): Likewise.
23750 (vqmovun_high_s64): Likewise.
23751 * config/aarch64/iterators.md (UNSPEC_SQXTUN2): Define.
23753 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23755 * config/aarch64/aarch64-simd-builtins.def (bfdot_lane, bfdot_laneq): Use
23757 (bfmlalb_lane, bfmlalt_lane, bfmlalb_lane_q, bfmlalt_lane_q): Use FP flags.
23759 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23761 * config/aarch64/aarch64-simd-builtins.def (fcmla_lane0, fcmla_lane90,
23762 fcmla_lane180, fcmla_lane270, fcmlaq_lane0, fcmlaq_lane90, fcmlaq_lane180,
23763 fcmlaq_lane270, scvtf, ucvtf, fcvtzs, fcvtzu, scvtfsi, scvtfdi, ucvtfsi,
23764 ucvtfdi, fcvtzshf, fcvtzuhf, fmlal_lane_low, fmlsl_lane_low,
23765 fmlal_laneq_low, fmlsl_laneq_low, fmlalq_lane_low, fmlslq_lane_low,
23766 fmlalq_laneq_low, fmlslq_laneq_low, fmlal_lane_high, fmlsl_lane_high,
23767 fmlal_laneq_high, fmlsl_laneq_high, fmlalq_lane_high, fmlslq_lane_high,
23768 fmlalq_laneq_high, fmlslq_laneq_high): Use FP flags.
23770 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23772 * config/aarch64/aarch64-builtins.c (FLAG_LOAD): Define.
23773 * config/aarch64/aarch64-simd-builtins.def (ld1x2, ld2, ld3, ld4, ld2r,
23774 ld3r, ld4r, ld1, ld1x3, ld1x4): Use LOAD flags.
23776 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23778 * config/aarch64/aarch64-simd-builtins.def (combine, zip1, zip2,
23779 uzp1, uzp2, trn1, trn2, simd_bsl): Use AUTO_FP flags.
23781 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23783 * config/aarch64/aarch64-simd-builtins.def (clrsb, clz, ctz, popcount,
23784 vec_smult_lane_, vec_smlal_lane_, vec_smult_laneq_, vec_smlal_laneq_,
23785 vec_umult_lane_, vec_umlal_lane_, vec_umult_laneq_, vec_umlal_laneq_,
23786 ashl, sshl, ushl, srshl, urshl, sdot_lane, udot_lane, sdot_laneq,
23787 udot_laneq, usdot_lane, usdot_laneq, sudot_lane, sudot_laneq, ashr,
23788 ashr_simd, lshr, lshr_simd, srshr_n, urshr_n, ssra_n, usra_n, srsra_n,
23789 ursra_n, sshll_n, ushll_n, sshll2_n, ushll2_n, ssri_n, usri_n, ssli_n,
23790 ssli_n, usli_n, bswap, rbit, simd_bsl, eor3q, rax1q, xarq, bcaxq): Use
23791 NONE builtin flags.
23793 2021-02-02 Jakub Jelinek <jakub@redhat.com>
23795 PR tree-optimization/98848
23796 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
23797 STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
23799 2021-02-02 Kito Cheng <kito.cheng@sifive.com>
23802 * expr.c: Check mode before calling store_expr.
23804 2021-02-02 Christophe Lyon <christophe.lyon@linaro.org>
23806 * config/arm/iterators.md (supf): Remove VORNQ_S and VORNQ_U.
23808 * config/arm/mve.md (mve_vornq_s<mode>): New entry for vorn
23809 instruction using expression ior.
23810 (mve_vornq_u<mode>): New expander.
23811 (mve_vornq_f<mode>): Use ior code instead of unspec.
23812 * config/arm/unspecs.md (VORNQ_S, VORNQ_U, VORNQ_F): Remove.
23814 2021-02-02 Alexandre Oliva <oliva@adacore.com>
23816 * tree-nested.c (convert_nonlocal_reference_op): Move
23817 current_function_decl restore after re-gimplification.
23818 (convert_local_reference_op): Likewise.
23820 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23822 * config/aarch64/aarch64-simd-builtins.def (rshrn, rshrn2):
23824 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>_insn_le):
23826 (aarch64_rshrn<mode>_insn_be): Likewise.
23827 (aarch64_rshrn<mode>): Likewise.
23828 (aarch64_rshrn2<mode>_insn_le): Likewise.
23829 (aarch64_rshrn2<mode>_insn_be): Likewise.
23830 (aarch64_rshrn2<mode>): Likewise.
23831 * config/aarch64/aarch64.md (unspec): Add UNSPEC_RSHRN.
23832 * config/aarch64/arm_neon.h (vrshrn_high_n_s16): Reimplement
23834 (vrshrn_high_n_s32): Likewise.
23835 (vrshrn_high_n_s64): Likewise.
23836 (vrshrn_high_n_u16): Likewise.
23837 (vrshrn_high_n_u32): Likewise.
23838 (vrshrn_high_n_u64): Likewise.
23839 (vrshrn_n_s16): Likewise.
23840 (vrshrn_n_s32): Likewise.
23841 (vrshrn_n_s64): Likewise.
23842 (vrshrn_n_u16): Likewise.
23843 (vrshrn_n_u32): Likewise.
23844 (vrshrn_n_u64): Likewise.
23846 2021-02-01 Sergei Trofimovich <siarheit@google.com>
23848 PR tree-optimization/98499
23849 * ipa-modref.c (analyze_ssa_name_flags): treat RVO
23850 conservatively and assume all possible side-effects.
23852 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23854 * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi,
23855 vec_unpacku_hi_): Define builtins.
23856 * config/aarch64/arm_neon.h (vmovl_high_s8): Reimplement using
23858 (vmovl_high_s16): Likewise.
23859 (vmovl_high_s32): Likewise.
23860 (vmovl_high_u8): Likewise.
23861 (vmovl_high_u16): Likewise.
23862 (vmovl_high_u32): Likewise.
23864 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23866 * config/aarch64/aarch64-simd-builtins.def (sabdl, uabdl):
23868 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl<mode>): New
23870 * config/aarch64/aarch64.md (unspec): Define UNSPEC_SABDL,
23872 * config/aarch64/arm_neon.h (vabdl_s8): Reimplemet using
23874 (vabdl_s16): Likewise.
23875 (vabdl_s32): Likewise.
23876 (vabdl_u8): Likewise.
23877 (vabdl_u16): Likewise.
23878 (vabdl_u32): Likewise.
23879 * config/aarch64/iterators.md (ABDL): New int iterator.
23880 (sur): Handle UNSPEC_SABDL, UNSPEC_UABDL.
23882 2021-02-01 Martin Sebor <msebor@redhat.com>
23884 * tree.h (BLOCK_VARS): Add comment.
23885 (BLOCK_SUBBLOCKS): Same.
23886 (BLOCK_SUPERCONTEXT): Same.
23887 (BLOCK_ABSTRACT_ORIGIN): Same.
23888 (inlined_function_outer_scope_p): Same.
23890 2021-02-01 Martin Sebor <msebor@redhat.com>
23892 PR middle-end/97172
23893 * attribs.c (attr_access::free_lang_data): Define new function.
23894 * attribs.h (attr_access::free_lang_data): Declare new function.
23896 2021-02-01 Richard Biener <rguenther@suse.de>
23898 * vec.h (auto_vec::auto_vec): Add memory stat parameters
23900 * bitmap.h (auto_bitmap::auto_bitmap): Likewise.
23902 2021-02-01 Tamar Christina <tamar.christina@arm.com>
23904 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>,
23905 aarch64_<su>mlsl<mode>, aarch64_<su>mlsl_n<mode>): Flip mult operands.
23907 2021-02-01 Richard Biener <rguenther@suse.de>
23909 PR rtl-optimization/98863
23910 * config/i386/i386-features.c (convert_scalars_to_vector):
23911 Set DF_RD_PRUNE_DEAD_DEFS.
23913 2021-01-31 Eric Botcazou <ebotcazou@adacore.com>
23915 * system.h (SIZE_MAX): Define if not already defined.
23917 2021-01-30 Aaron Sawdey <acsawdey@linux.ibm.com>
23919 * config/rs6000/genfusion.pl (gen_2logical): New function to
23920 generate patterns for logical-logical fusion.
23921 * config/rs6000/fusion.md: Regenerated patterns.
23922 * config/rs6000/rs6000-cpus.def: Add
23923 OPTION_MASK_P10_FUSION_2LOGICAL.
23924 * config/rs6000/rs6000.c (rs6000_option_override_internal):
23925 Enable logical-logical fusion for p10.
23926 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2logical.
23928 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
23930 * config/rs6000/rs6000.opt: Add periods to new AIX options.
23932 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
23934 * config/rs6000/rs6000.opt (mabi=vec-extabi): New.
23935 (mabi=vec-default): New.
23936 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
23937 __EXTABI__ for AIX Vector extended ABI.
23938 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print AIX Vector
23940 (conditional_register_usage): If AIX vec_extabi enabled, vs20-vs31
23942 * doc/invoke.texi (PowerPC mabi): Add AIX vec-extabi and vec-default.
23944 2021-01-30 Jakub Jelinek <jakub@redhat.com>
23946 * config/i386/i386-features.c (remove_partial_avx_dependency): Clear
23947 DF_DEFER_INSN_RESCAN after calling df_process_deferred_rescans.
23949 2021-01-29 Vladimir N. Makarov <vmakarov@redhat.com>
23952 * lra-constraints.c (in_class_p): Don't narrow class only for REG
23955 2021-01-29 Will Schmidt <will_schmidt@vnet.ibm.com>
23957 * config/rs6000/rs6000-call.c (rs6000_expand_binup_builtin): Add
23958 clauses for CODE_FOR_vsx_xvcvuxddp_scale and
23959 CODE_FOR_vsx_xvcvsxddp_scale to the parameter checking code.
23961 2021-01-29 Andrew MacLeod <amacleod@redhat.com>
23963 PR tree-optimization/98866
23964 * gimple-range-gori.h (gori_compute:set_range_invariant): New.
23965 * gimple-range-gori.cc (gori_map::set_range_invariant): New.
23966 (gori_map::m_maybe_invariant): Rename from all_outgoing.
23967 (gori_map::gori_map): Rename all_outgoing to m_maybe_invariant.
23968 (gori_map::is_export_p): Ditto.
23969 (gori_map::calculate_gori): Ditto.
23970 (gori_compute::set_range_invariant): New.
23971 * gimple-range.cc (gimple_ranger::range_of_stmt): Set range
23972 invariant for pointers evaluating to [1, +INF].
23974 2021-01-29 Richard Biener <rguenther@suse.de>
23976 PR rtl-optimization/98863
23977 * config/i386/i386-features.c (remove_partial_avx_dependency):
23978 Do not perform DF analysis.
23979 (pass_data_remove_partial_avx_dependency): Remove
23982 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
23984 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_n
23985 builtin generator macros.
23986 * config/aarch64/aarch64-simd.md (aarch64_<su>mull_n<mode>):
23988 * config/aarch64/arm_neon.h (vmull_n_s16): Use RTL builtin
23989 instead of inline asm.
23990 (vmull_n_s32): Likewise.
23991 (vmull_n_u16): Likewise.
23992 (vmull_n_u32): Likewise.
23994 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23996 * config/aarch64/aarch64-simd-builtins.def (sabdl2, uabdl2):
23998 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
24000 (aarch64_<sur>abdl2<mode>): ... This.
24001 (<sur>sadv16qi): Adjust use of above.
24002 * config/aarch64/arm_neon.h (vabdl_high_s8): Reimplement using
24004 (vabdl_high_s16): Likewise.
24005 (vabdl_high_s32): Likewise.
24006 (vabdl_high_u8): Likewise.
24007 (vabdl_high_u16): Likewise.
24008 (vabdl_high_u32): Likewise.
24010 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24012 * config/aarch64/aarch64-simd-builtins.def (sabal2): Define
24014 (uabal2): Likewise.
24015 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal2<mode>): New
24017 * config/aarch64/aarch64.md (unspec): Add UNSPEC_SABAL2 and
24019 * config/aarch64/arm_neon.h (vabal_high_s8): Reimplement using
24021 (vabal_high_s16): Likewise.
24022 (vabal_high_s32): Likewise.
24023 (vabal_high_u8): Likewise.
24024 (vabal_high_u16): Likewise.
24025 (vabal_high_u32): Likewise.
24026 * config/aarch64/iterators.md (ABAL2): New mode iterator.
24027 (sur): Handle UNSPEC_SABAL2, UNSPEC_UABAL2.
24029 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24031 * config/aarch64/aarch64-simd-builtins.def (sabal): Define
24034 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal<mode>_4):
24036 (aarch64_<sur>abal<mode>): ... This
24037 (<sur>sadv16qi): Adust use of the above.
24038 * config/aarch64/arm_neon.h (vabal_s8): Reimplement using
24040 (vabal_s16): Likewise.
24041 (vabal_s32): Likewise.
24042 (vabal_u8): Likewise.
24043 (vabal_u16): Likewise.
24044 (vabal_u32): Likewise.
24046 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24048 * config/aarch64/aarch64-simd-builtins.def (saddlv, uaddlv):
24050 * config/aarch64/aarch64-simd.md (aarch64_<su>addlv<mode>):
24052 * config/aarch64/arm_neon.h (vaddlv_s8): Reimplement using
24054 (vaddlv_s16): Likewise.
24055 (vaddlv_u8): Likewise.
24056 (vaddlv_u16): Likewise.
24057 (vaddlvq_s8): Likewise.
24058 (vaddlvq_s16): Likewise.
24059 (vaddlvq_s32): Likewise.
24060 (vaddlvq_u8): Likewise.
24061 (vaddlvq_u16): Likewise.
24062 (vaddlvq_u32): Likewise.
24063 (vaddlv_s32): Likewise.
24064 (vaddlv_u32): Likewise.
24065 * config/aarch64/iterators.md (VDQV_L): New mode iterator.
24066 (unspec): Add UNSPEC_SADDLV, UNSPEC_UADDLV.
24067 (Vwstype): New mode attribute.
24069 (VWIDE_S): Likewise.
24070 (USADDLV): New int iterator.
24071 (su): Handle UNSPEC_SADDLV, UNSPEC_UADDLV.
24073 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
24075 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_lane[q]
24076 builtin generator macros.
24077 * config/aarch64/aarch64-simd.md (aarch64_vec_<su>mlsl_lane<Qlane>):
24079 * config/aarch64/arm_neon.h (vmlsl_lane_s16): Use RTL builtin
24080 instead of inline asm.
24081 (vmlsl_lane_s32): Likewise.
24082 (vmlsl_lane_u16): Likewise.
24083 (vmlsl_lane_u32): Likewise.
24084 (vmlsl_laneq_s16): Likewise.
24085 (vmlsl_laneq_s32): Likewise.
24086 (vmlsl_laneq_u16): Likewise.
24087 (vmlsl_laneq_u32): Likewise.
24089 2021-01-29 Richard Biener <rguenther@suse.de>
24091 * doc/invoke.texi (--param max-gcse-memory): Document unit
24093 * gcse.c (gcse_or_cprop_is_too_expensive): Adjust.
24094 * params.opt (--param max-gcse-memory): Adjust default and
24095 document unit of size.
24097 2021-01-29 Richard Biener <rguenther@suse.de>
24099 PR rtl-optimization/98863
24100 * gcse.c (gcse_or_cprop_is_too_expensive): Use unsigned
24101 HOST_WIDE_INT for the memory estimate.
24103 2021-01-29 Bin Cheng <bin.cheng@linux.alibaba.com>
24104 Richard Biener <rguenther@suse.de>
24106 PR tree-optimization/97627
24107 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
24108 Do not analyze fake edges.
24110 2021-01-29 Richard Biener <rguenther@suse.de>
24112 PR rtl-optimization/98144
24113 * df.h (df_mir_bb_info): Add con_visited member.
24114 * df-problems.c (df_mir_alloc): Initialize con_visited,
24115 do not fully populate IN and OUT.
24116 (df_mir_reset): Likewise.
24117 (df_mir_confluence_0): Set con_visited.
24118 (df_mir_confluence_n): Properly handle implicitely
24119 fully populated IN and OUT as designated by con_visited
24120 and update con_visited accordingly.
24122 2021-01-29 Jakub Jelinek <jakub@redhat.com>
24125 * config/arm/vec-common.md (mve_vshlq_<supf><mode>,
24126 vashl<mode>3, vashr<mode>3, vlshr<mode>3): Add
24127 && !TARGET_REALLY_IWMMXT to conditions.
24129 2021-01-29 Jakub Jelinek <jakub@redhat.com>
24132 * cfgbuild.c (find_bb_boundaries): Reset debug_insn when seeing
24135 2021-01-28 Marek Polacek <polacek@redhat.com>
24138 * stor-layout.c (finalize_type_size): If we reset TYPE_USER_ALIGN in
24139 the main variant, maybe reset it in its variants too.
24140 * tree.c (check_base_type): Return true only if TYPE_USER_ALIGN match.
24141 (check_aligned_type): Check if TYPE_USER_ALIGN match.
24143 2021-01-28 Christophe Lyon <christophe.lyon@linaro.org>
24146 * config/arm/arm.c (arm_rtx_costs_internal): Adjust cost of vector
24147 of constant zero for comparisons.
24149 2021-01-28 Michael Meissner <meissner@linux.ibm.com>
24151 * config/rs6000/rs6000.c (rs6000_mangle_decl_assembler_name): Add
24152 support for mapping built-in function names for long double
24153 built-in functions if long double is IEEE 128-bit.
24155 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
24157 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_n
24158 builtin generator macros.
24159 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_n<mode>):
24161 * config/aarch64/arm_neon.h (vmlsl_n_s16): Use RTL builtin
24162 instead of inline asm.
24163 (vmlsl_n_s32): Likewise.
24164 (vmlsl_n_u16): Likewise.
24165 (vmlsl_n_u32): Likewise.
24167 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
24169 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_n
24170 builtin generator macros.
24171 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>):
24173 * config/aarch64/arm_neon.h (vmlal_n_s16): Use RTL builtin
24174 instead of inline asm.
24175 (vmlal_n_s32): Likewise.
24176 (vmlal_n_u16): Likewise.
24177 (vmlal_n_u32): Likewise.
24179 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24181 * config/aarch64/aarch64-simd-builtins.def (shrn2): Define
24183 * config/aarch64/aarch64-simd.md (aarch64_shrn2<mode>_insn_le):
24185 (aarch64_shrn2<mode>_insn_be): Likewise.
24186 (aarch64_shrn2<mode>): Likewise.
24187 * config/aarch64/arm_neon.h (vshrn_high_n_s16): Reimlplement
24189 (vshrn_high_n_s32): Likewise.
24190 (vshrn_high_n_s64): Likewise.
24191 (vshrn_high_n_u16): Likewise.
24192 (vshrn_high_n_u32): Likewise.
24193 (vshrn_high_n_u64): Likewise.
24195 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24197 * config/aarch64/aarch64-simd-builtins.def (shrn): Define
24199 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le):
24201 (aarch64_shrn<mode>_insn_be): Likewise.
24202 (aarch64_shrn<mode>): Likewise.
24203 * config/aarch64/arm_neon.h (vshrn_n_s16): Reimplement using
24205 (vshrn_n_s32): Likewise.
24206 (vshrn_n_s64): Likewise.
24207 (vshrn_n_u16): Likewise.
24208 (vshrn_n_u32): Likewise.
24209 (vshrn_n_u64): Likewise.
24210 * config/aarch64/iterators.md (vn_mode): New mode attribute.
24212 2021-01-28 Richard Biener <rguenther@suse.de>
24214 PR rtl-optimization/80960
24215 * dse.c (check_mem_read_rtx): Call get_addr on the
24218 2021-01-28 Xionghu Luo <luoxhu@linux.ibm.com>
24219 David Edelsohn <dje.gcc@gmail.com>
24222 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
24223 Don't generate VIEW_CONVERT_EXPR for fcode ALTIVEC_BUILTIN_VEC_INSERT
24225 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
24227 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Remove the
24228 wrapper call rs6000_expand_vector_set_var for cleanup. Call
24229 rs6000_expand_vector_set_var_p9 and rs6000_expand_vector_set_var_p8
24231 (rs6000_expand_vector_set_var): Delete.
24232 (rs6000_expand_vector_set_var_p9): Make static.
24233 (rs6000_expand_vector_set_var_p8): Make static.
24235 2021-01-28 Xing GUO <higuoxing@gmail.com>
24237 * common/config/riscv/riscv-common.c
24238 (riscv_subset_list::parsing_subset_version): Fix -march option parsing
24239 when `p` extension exists.
24241 2021-01-27 Vladimir N. Makarov <vmakarov@redhat.com>
24243 PR rtl-optimization/97684
24244 * ira.c (ira): Call ira_set_pseudo_classes before
24245 update_equiv_regs when it is necessary.
24247 2021-01-27 Jakub Jelinek <jakub@redhat.com>
24250 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Use
24251 %w0, %w1 and %2 instead of %0, %1 and %2.
24253 2021-01-27 Aaron Sawdey <acsawdey@linux.ibm.com>
24255 * config/rs6000/genfusion.pl: New script to generate
24256 define_insn_and_split patterns so combine can arrange fused
24257 instructions next to each other.
24258 * config/rs6000/fusion.md: New file, generated fused instruction
24259 patterns for combine.
24260 * config/rs6000/predicates.md (const_m1_to_1_operand): New predicate.
24261 (non_update_memory_operand): New predicate.
24262 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_P10_FUSION and
24263 OPTION_MASK_P10_FUSION_LD_CMPI to ISA_3_1_MASKS_SERVER and
24265 * config/rs6000/rs6000-protos.h (address_is_non_pfx_d_or_x): Add
24267 * config/rs6000/rs6000.c (rs6000_option_override_internal):
24268 Automatically set OPTION_MASK_P10_FUSION and
24269 OPTION_MASK_P10_FUSION_LD_CMPI if target is power10.
24270 (rs600_opt_masks): Allow -mpower10-fusion
24271 in function attributes.
24272 (address_is_non_pfx_d_or_x): New function.
24273 * config/rs6000/rs6000.h: Add MASK_P10_FUSION.
24274 * config/rs6000/rs6000.md: Include fusion.md.
24275 * config/rs6000/rs6000.opt: Add -mpower10-fusion
24276 and -mpower10-fusion-ld-cmpi.
24277 * config/rs6000/t-rs6000: Add dependencies involving fusion.md.
24279 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
24281 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal
24282 builtin generator macros.
24283 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal<mode>):
24285 (aarch64_<su>mlal<mode>): This.
24286 * config/aarch64/arm_neon.h (vmlal_s8): Use RTL builtin
24287 instead of inline asm.
24288 (vmlal_s16): Likewise.
24289 (vmlal_s32): Likewise.
24290 (vmlal_u8): Likewise.
24291 (vmlal_u16): Likewise.
24292 (vmlal_u32): Likewise.
24294 2021-01-27 Richard Biener <rguenther@suse.de>
24296 PR tree-optimization/98854
24297 * tree-vect-slp.c (vect_build_slp_tree_2): Also build
24298 PHIs from scalars when the number of CTORs matches the
24299 number of children.
24301 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
24303 * config/aarch64/aarch64-simd-builtins.def: Add mls_n builtin
24305 * config/aarch64/aarch64-simd.md (*aarch64_mls_elt_merge<mode>):
24307 (aarch64_mls_n<mode>): This.
24308 * config/aarch64/arm_neon.h (vmls_n_s16): Use RTL builtin
24310 (vmls_n_s32): Likewise.
24311 (vmls_n_u16): Likewise.
24312 (vmls_n_u32): Likewise.
24313 (vmlsq_n_s16): Likewise.
24314 (vmlsq_n_s32): Likewise.
24315 (vmlsq_n_u16): Likewise.
24316 (vmlsq_n_u32): Likewise.
24318 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
24320 * config/aarch64/aarch64-simd-builtins.def: Add mls builtin
24322 * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather
24324 (vmls_s16): Likewise.
24325 (vmls_s32): Likewise.
24326 (vmls_u8): Likewise.
24327 (vmls_u16): Likewise.
24328 (vmls_u32): Likewise.
24329 (vmlsq_s8): Likewise.
24330 (vmlsq_s16): Likewise.
24331 (vmlsq_s32): Likewise.
24332 (vmlsq_u8): Likewise.
24333 (vmlsq_u16): Likewise.
24334 (vmlsq_u32): Likewise.
24336 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
24338 * config/aarch64/aarch64-simd-builtins.def: Add mla_n builtin
24340 * config/aarch64/aarch64-simd.md (*aarch64_mla_elt_merge<mode>):
24342 (aarch64_mla_n<mode>): This.
24343 * config/aarch64/arm_neon.h (vmla_n_s16): Use RTL builtin
24345 (vmla_n_s32): Likewise.
24346 (vmla_n_u16): Likewise.
24347 (vmla_n_u32): Likewise.
24348 (vmlaq_n_s16): Likewise.
24349 (vmlaq_n_s32): Likewise.
24350 (vmlaq_n_u16): Likewise.
24351 (vmlaq_n_u32): Likewise.
24353 2021-01-27 liuhongt <hongtao.liu@intel.com>
24356 * config/i386/sse.md (sse2_gt<mode>3): Drop !TARGET_XOP in condition.
24357 (*sse2_eq<mode>3): Ditto.
24359 2021-01-27 Jakub Jelinek <jakub@redhat.com>
24361 * tree-pass.h (PROP_trees): Rename to ...
24362 (PROP_gimple): ... this.
24363 * cfgexpand.c (pass_data_expand): Replace PROP_trees with PROP_gimple.
24364 * passes.c (execute_function_dump, execute_function_todo,
24365 execute_one_ipa_transform_pass, execute_one_pass): Likewise.
24366 * varpool.c (ctor_for_folding): Likewise.
24368 2021-01-27 Jakub Jelinek <jakub@redhat.com>
24370 PR tree-optimization/97260
24371 * varpool.c: Include tree-pass.h.
24372 (ctor_for_folding): In GENERIC return DECL_INITIAL for TREE_READONLY
24373 non-TREE_SIDE_EFFECTS automatic variables.
24375 2021-01-26 Paul Fee <paul.f.fee@gmail.com>
24377 * doc/cpp.texi (__cplusplus): Document value for -std=c++23
24379 * doc/invoke.texi: Document -std=c++23 and -std=gnu++23.
24380 * dwarf2out.c (highest_c_language): Recognise C++20 and C++23.
24381 (gen_compile_unit_die): Recognise C++23.
24383 2021-01-26 Jakub Jelinek <jakub@redhat.com>
24386 * dwarf2asm.c (dw2_assemble_integer): Cast DWARF2_ADDR_SIZE to int
24389 2021-01-26 Jakub Jelinek <jakub@redhat.com>
24392 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
24393 Use UINTVAL (shft_amnt) and UINTVAL (mask) instead of INTVAL (shft_amnt)
24394 and INTVAL (mask). Add && INTVAL (mask) > 0 condition.
24396 2021-01-26 Richard Biener <rguenther@suse.de>
24398 * gimple-pretty-print.c (dump_binary_rhs): Handle
24399 VEC_WIDEN_{PLUS,MINUS}_{LO,HI}_EXPR.
24401 2021-01-26 Richard Biener <rguenther@suse.de>
24403 PR middle-end/98726
24404 * tree.h (vector_cst_int_elt): Remove.
24405 * tree.c (vector_cst_int_elt): Use poly_wide_int for computations,
24408 2021-01-26 Andrew Stubbs <ams@codesourcery.com>
24410 * config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
24411 for V64DFmode min/max reductions.
24413 2021-01-26 Jakub Jelinek <jakub@redhat.com>
24415 * dwarf2asm.c (dw2_assemble_integer): Handle size twice as large
24416 as DWARF2_ADDR_SIZE if x is not a scalar int by emitting it as
24417 two halves, one with x and the other with const0_rtx, ordered
24418 depending on endianity.
24420 2021-01-26 Alexandre Oliva <oliva@adacore.com>
24422 * gimplify.c (gimplify_decl_expr): Skip asan marking calls for
24423 temporaries not seen in binding block, and not about to be
24424 added as gimple variables.
24426 2021-01-25 Martin Sebor <msebor@redhat.com>
24429 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust warning text.
24431 2021-01-25 Martin Liska <mliska@suse.cz>
24433 * value-prof.c (get_nth_most_common_value): Use %s instead
24436 2021-01-25 Jakub Jelinek <jakub@redhat.com>
24439 * configure.ac (HAVE_AS_GDWARF_5_DEBUG_FLAG): Only define if
24440 readelf -wi is able to read the emitted .debug_info back.
24441 * configure: Regenerated.
24443 2021-01-25 Martin Liska <mliska@suse.cz>
24445 PR gcov-profile/98739
24446 * common.opt: Add missing sign symbol.
24447 * value-prof.c (get_nth_most_common_value): Restore handling
24448 of PROFILE_REPRODUCIBILITY_PARALLEL_RUNS and
24449 PROFILE_REPRODUCIBILITY_MULTITHREADED.
24451 2021-01-25 Richard Biener <rguenther@suse.de>
24453 PR middle-end/98807
24454 * tree.c (vector_element_bits): Always use precision of
24455 the element type for boolean vectors.
24457 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
24459 * config/rtems.h (STARTFILE_SPEC): Remove qnolinkcmds.
24460 (ENDFILE_SPEC): Evaluate qnolinkcmds.
24462 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
24464 * config/rtems.h (STARTFILE_SPEC): Remove nostdlib and
24465 nostartfiles handling since this is already done by
24466 LINK_COMMAND_SPEC. Evaluate qnolinkcmds.
24467 (ENDFILE_SPEC): Remove nostdlib and nostartfiles handling since this
24468 is already done by LINK_COMMAND_SPEC.
24469 (LIB_SPECS): Remove nostdlib and nodefaultlibs handling since
24470 this is already done by LINK_COMMAND_SPEC. Remove qnolinkcmds
24473 2021-01-25 Jakub Jelinek <jakub@redhat.com>
24476 * fold-const-call.c (host_size_t_cst_p): Renamed to ...
24477 (size_t_cst_p): ... this. Check and store unsigned HOST_WIDE_INT
24478 value rather than host size_t.
24479 (fold_const_call): Change type of s2 from size_t to
24480 unsigned HOST_WIDE_INT. Use size_t_cst_p instead of
24481 host_size_t_cst_p. For strncmp calls, pass MIN (s2, SIZE_MAX)
24482 instead of s2 as last argument.
24484 2021-01-25 Tamar Christina <tamar.christina@arm.com>
24486 * config/arm/iterators.md (rotsplit1, rotsplit2, conj_op, fcmac1,
24487 VCMLA_OP, VCMUL_OP): New.
24488 * config/arm/mve.md (mve_vcmlaq<mve_rot><mode>): Support vec_dup 0.
24489 * config/arm/neon.md (cmul<conj_op><mode>3): New.
24490 * config/arm/unspecs.md (UNSPEC_VCMLA_CONJ, UNSPEC_VCMLA180_CONJ,
24491 UNSPEC_VCMUL_CONJ): New.
24492 * config/arm/vec-common.md (cmul<conj_op><mode>3, arm_vcmla<rot><mode>,
24493 cml<fcmac1><conj_op><mode>4): New.
24495 2021-01-23 Jakub Jelinek <jakub@redhat.com>
24498 * config/rs6000/mmintrin.h (__m64): Add __may_alias__ attribute.
24500 2021-01-22 Jonathan Wright <jonathan.wright@arm.com>
24502 * config/aarch64/aarch64-simd-builtins.def: Add mla builtin
24504 * config/aarch64/arm_neon.h (vmla_s8): Use RTL builtin rather
24506 (vmla_s16): Likewise.
24507 (vmla_s32): Likewise.
24508 (vmla_u8): Likewise.
24509 (vmla_u16): Likewise.
24510 (vmla_u32): Likewise.
24511 (vmlaq_s8): Likewise.
24512 (vmlaq_s16): Likewise.
24513 (vmlaq_s32): Likewise.
24514 (vmlaq_u8): Likewise.
24515 (vmlaq_u16): Likewise.
24516 (vmlaq_u32): Likewise.
24518 2021-01-22 David Malcolm <dmalcolm@redhat.com>
24520 * doc/invoke.texi (GCC_EXTRA_DIAGNOSTIC_OUTPUT): Add @findex
24523 2021-01-22 Jakub Jelinek <jakub@redhat.com>
24526 * dwarf2out.c (output_file_names): For -gdwarf-5, if there are no
24527 filenames to emit, still emit the required 0 index directory and
24528 filename entries that match DW_AT_comp_dir and DW_AT_name of the
24531 2021-01-22 Marek Polacek <polacek@redhat.com>
24534 * doc/invoke.texi: Update C++ ABI Version 15 description.
24536 2021-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24538 PR tree-optimization/98766
24539 * tree-ssa-math-opts.c (convert_mult_to_fma): Use maybe_le when
24540 comparing against type size with param_avoid_fma_max_bits.
24542 2021-01-22 Richard Biener <rguenther@suse.de>
24544 PR middle-end/98793
24545 * tree.c (vector_element_bits): Key single-bit bool vector on
24546 integer mode rather than not vector mode.
24548 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
24551 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
24552 Generate ARRAY_REF(VIEW_CONVERT_EXPR) for P8 and later
24554 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): Update
24555 to call different path for P8 and P9.
24556 (rs6000_expand_vector_set_var_p9): New function.
24557 (rs6000_expand_vector_set_var_p8): New function.
24559 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
24563 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
24564 Ajdust variable index vec_insert from address dereference to
24565 ARRAY_REF(VIEW_CONVERT_EXPR) tree expression.
24566 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
24568 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): New function.
24570 2021-01-22 Martin Liska <mliska@suse.cz>
24572 PR gcov-profile/98739
24573 * profile.c (compute_value_histograms): Drop time profile for
24574 -fprofile-reproducible=multithreaded.
24576 2021-01-22 Nathan Sidwell <nathan@acm.org>
24578 * gcc.c (process_command): Don't check OPT_SPECIAL_input_file
24581 2021-01-22 Richard Biener <rguenther@suse.de>
24583 PR middle-end/98773
24584 * tree-data-ref.c (initalize_matrix_A): Revert previous
24585 change, retaining failing on HOST_WIDE_INT_MIN CHREC_RIGHT.
24587 2021-01-22 Jakub Jelinek <jakub@redhat.com>
24589 PR tree-optimization/90248
24590 * match.pd (X cmp 0.0 ? 1.0 : -1.0 -> copysign(1, +-X),
24591 X cmp 0.0 ? -1.0 : +1.0 -> copysign(1, -+X)): Remove
24593 (X * (X cmp 0.0 ? 1.0 : -1.0) -> +-abs(X),
24594 X * (X cmp 0.0 ? -1.0 : 1.0) -> +-abs(X)): New simplifications.
24596 2021-01-22 Jakub Jelinek <jakub@redhat.com>
24598 PR tree-optimization/98255
24599 * tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
24600 extend index - low_bound from sizetype's precision rather than index
24602 (get_addr_base_and_unit_offset_1): Likewise.
24603 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
24604 * gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
24606 2021-01-22 Richard Biener <rguenther@suse.de>
24608 PR tree-optimization/98786
24609 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Avoid
24610 adding new uses of abnormals. Verify we deal with a conditional
24613 2021-01-22 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24616 * optc-save-gen.awk: Add arm_fp16_format to checked_options.
24618 2021-01-22 liuhongt <hongtao.liu@intel.com>
24622 * config/i386/sse.md (VI_128_256): New mode iterator.
24623 (*avx_cmp<mode>3_1, *avx_cmp<mode>3_2, *avx_cmp<mode>3_3,
24624 *avx_cmp<mode>3_4, *avx2_eq<mode>3, *avx2_pcmp<mode>3_1,
24625 *avx2_pcmp<mode>3_2, *avx2_gt<mode>3): New
24626 define_insn_and_split to lower avx512 vector comparison to avx
24627 version when dest is vector.
24628 (*<avx512>_cmp<mode>3,*<avx512>_cmp<mode>3,*<avx512>_ucmp<mode>3):
24629 define_insn_and_split for negating the comparison result.
24630 * config/i386/predicates.md (float_vector_all_ones_operand):
24632 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Use
24633 general NOT operator without UNSPEC_MASKOP.
24635 2021-01-21 Vladimir N. Makarov <vmakarov@redhat.com>
24637 PR rtl-optimization/98777
24638 * lra-int.h (lra_pmode_pseudo): New extern.
24639 * lra.c (lra_pmode_pseudo): New global.
24641 * lra-eliminations.c (eliminate_regs_in_insn): Use it.
24643 2021-01-21 Ilya Leoshkevich <iii@linux.ibm.com>
24645 * fwprop.c (fwprop_propagation::classify_result): Allow
24646 (subreg (mem)) simplifications.
24648 2021-01-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24650 * config/aarch64/aarch64-simd.md (aarch64_sqdml<SBINQOPS:as>l<mode>):
24652 (aarch64_sqdmlal<mode>): ... This...
24653 (aarch64_sqdmlsl<mode>): ... And this.
24654 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Split into...
24655 (aarch64_sqdmlal_lane<mode>): ... This...
24656 (aarch64_sqdmlsl_lane<mode>): ... And this.
24657 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Split into...
24658 (aarch64_sqdmlsl_laneq<mode>): ... This...
24659 (aarch64_sqdmlal_laneq<mode>): ... And this.
24660 (aarch64_sqdml<SBINQOPS:as>l_n<mode>): Split into...
24661 (aarch64_sqdmlsl_n<mode>): ... This...
24662 (aarch64_sqdmlal_n<mode>): ... And this.
24663 (aarch64_sqdml<SBINQOPS:as>l2<mode>_internal): Split into...
24664 (aarch64_sqdmlal2<mode>_internal): ... This...
24665 (aarch64_sqdmlsl2<mode>_internal): ... And this.
24667 2021-01-21 Christophe Lyon <christophe.lyon@linaro.org>
24669 * config/arm/arm_mve.h (__arm_vcmpneq_s8): Fix return type.
24671 2021-01-21 Andrea Corallo <andrea.corallo@arm.com>
24674 * doc/sourcebuild.texi (arm_thumb2_no_arm_v8_1_lob): Document.
24676 2021-01-21 liuhongt <hongtao.liu@intel.com>
24678 PR rtl-optimization/98694
24679 * regcprop.c (copy_value): If SRC had been assigned a mode
24680 narrower than the copy, we can't link DEST into the chain even
24681 they have same hard_regno_nregs(i.e. HImode/SImode in i386
24684 2021-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24686 * config/aarch64/aarch64-simd.md (aarch64_get_lane<mode>):
24687 Convert to define_insn_and_split. Split into simple move when moving
24690 2021-01-20 Segher Boessenkool <segher@kernel.crashing.org>
24692 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Change assert.
24693 Adjust comment. Simplify code.
24695 2021-01-20 Jakub Jelinek <jakub@redhat.com>
24698 * dwarf2out.c (reset_indirect_string): Also reset indirect strings
24699 with DW_FORM_line_strp form.
24700 (prune_unused_types_update_strings): Don't add into debug_str_hash
24701 indirect strings with DW_FORM_line_strp form.
24702 (adjust_name_comp_dir): New function.
24703 (dwarf2out_finish): Call it on CU DIEs after resetting
24704 debug_line_str_hash.
24706 2021-01-20 Vladimir N. Makarov <vmakarov@redhat.com>
24708 PR rtl-optimization/98722
24709 * lra-eliminations.c (eliminate_regs_in_insn): Check that target
24710 has no 3-op add insn to transform insns containing two pluses.
24712 2021-01-20 Richard Biener <rguenther@suse.de>
24714 * hwint.h (add_hwi): New function.
24715 (mul_hwi): Likewise.
24716 * tree-data-ref.c (initialize_matrix_A): Properly translate
24717 tree constants and avoid HOST_WIDE_INT_MIN.
24718 (lambda_matrix_row_add): Avoid undefined integer overflow
24719 and return true on such overflow.
24720 (lambda_matrix_right_hermite): Handle overflow from
24721 lambda_matrix_row_add gracefully. Simplify previous fix.
24722 (analyze_subscript_affine_affine): Likewise.
24724 2021-01-20 Eugene Rozenfeld <erozen@microsoft.com>
24726 PR tree-optimization/96674
24727 * match.pd: New patterns: x < y || y == XXX_MIN --> x <= y - 1
24728 x >= y && y != XXX_MIN --> x > y - 1
24730 2021-01-20 Richard Sandiford <richard.sandiford@arm.com>
24732 PR tree-optimization/98535
24733 * tree-vect-slp.c (duplicate_and_interleave): Use quick_grow_cleared.
24734 If the high and low permutes are the same, remove the high permutes
24735 from the working set and only continue with the low ones.
24737 2021-01-20 Jakub Jelinek <jakub@redhat.com>
24739 PR tree-optimization/98721
24740 * builtins.c (access_ref::inform_access): Don't assume
24741 SSA_NAME_IDENTIFIER must be non-NULL. Print messages about
24742 object whenever allocfn is NULL, rather than only when DECL_P
24743 is true. Use %qE instead of %qD for that. Formatting fixes.
24745 2021-01-20 Richard Biener <rguenther@suse.de>
24747 PR tree-optimization/98758
24748 * tree-data-ref.c (int_divides_p): Use lambda_int arguments.
24749 (lambda_matrix_right_hermite): Avoid undefinedness with
24750 signed integer abs and multiplication.
24751 (analyze_subscript_affine_affine): Use lambda_int.
24753 2021-01-20 David Malcolm <dmalcolm@redhat.com>
24756 * dwarf2out.c (output_line_info): Rename static variable
24757 "generation", moving it out of the function to...
24758 (output_line_info_generation): New.
24759 (init_sections_and_labels): Likewise, renaming the variable to...
24760 (init_sections_and_labels_generation): New.
24761 (dwarf2out_c_finalize): Reset the new variables.
24763 2021-01-19 Martin Sebor <msebor@redhat.com>
24765 PR middle-end/98664
24766 * tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
24767 all functions, even if they're not declared artificial or inline.
24768 * tree.c (tree_inlined_location): Use macro expansion location
24769 only if scope traversal fails to expose one.
24771 2021-01-19 Richard Sandiford <richard.sandiford@arm.com>
24773 PR rtl-optimization/92294
24774 * alias.c (compare_base_symbol_refs): Take an extra parameter
24775 and add the distance between two symbols to it. Enshrine in
24776 comments that -1 means "either 0 or 1, but we can't tell
24777 which at compile time".
24778 (memrefs_conflict_p): Update call accordingly.
24779 (rtx_equal_for_memref_p): Likewise. Take the distance between symbols
24782 2021-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24784 * config/aarch64/aarch64-simd-builtins.def (sqshl, uqshl,
24785 sqrshl, uqrshl, sqadd, uqadd, sqsub, uqsub, suqadd, usqadd, sqmovn,
24786 uqmovn, sqxtn2, uqxtn2, sqabs, sqneg, sqdmlal, sqdmlsl, sqdmlal_lane,
24787 sqdmlsl_lane, sqdmlal_laneq, sqdmlsl_laneq, sqdmlal_n, sqdmlsl_n,
24788 sqdmlal2, sqdmlsl2, sqdmlal2_lane, sqdmlsl2_lane, sqdmlal2_laneq,
24789 sqdmlsl2_laneq, sqdmlal2_n, sqdmlsl2_n, sqdmull, sqdmull_lane,
24790 sqdmull_laneq, sqdmull_n, sqdmull2, sqdmull2_lane, sqdmull2_laneq,
24791 sqdmull2_n, sqdmulh, sqrdmulh, sqdmulh_lane, sqdmulh_laneq,
24792 sqrdmulh_lane, sqrdmulh_laneq, sqshrun_n, sqrshrun_n, sqshrn_n,
24793 uqshrn_n, sqrshrn_n, uqrshrn_n, sqshlu_n, sqshl_n, uqshl_n, sqrdmlah,
24794 sqrdmlsh, sqrdmlah_lane, sqrdmlsh_lane, sqrdmlah_laneq, sqrdmlsh_laneq,
24795 sqmovun): Use NONE flags.
24797 2021-01-19 Richard Biener <rguenther@suse.de>
24800 * ipa-modref.c (analyze_stmt): Only record a summary for a
24803 2021-01-19 Richard Biener <rguenther@suse.de>
24805 PR middle-end/98638
24806 * tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
24808 2021-01-19 Daniel Hellstrom <daniel@gaisler.com>
24810 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
24811 built-in define __FIX_LEON3FT_TN0018.
24813 2021-01-19 Richard Biener <rguenther@suse.de>
24816 * tree-inline.c (tree_function_versioning): Set input_location
24817 to UNKNOWN_LOCATION throughout the function.
24819 2021-01-19 Tobias Burnus <tobias@codesourcery.com>
24822 * omp-low.c (lower_omp_target): Handle nonpointer is_device_ptr.
24824 2021-01-19 Martin Jambor <mjambor@suse.cz>
24827 * ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
24828 whether non-call exceptions allow removal of a statement.
24829 (isra_analyze_call): Pass the appropriate function to
24830 ssa_name_only_returned_p.
24832 2021-01-19 Geng Qi <gengqi@linux.alibaba.com>
24834 * config/riscv/arch-canonicalize (longext_sort): New function for
24835 sorting 'multi-letter'.
24836 * config/riscv/multilib-generator: Adjusting the loop of 'alt' in
24837 'alts'. The 'arch' may not be the first of 'alts'.
24838 (_expand_combination): Add underline for the 'ext' without '*'.
24839 This is because, a single-letter extension can always be treated well
24840 with a '_' prefix, but it cannot be separated out if it is appended
24843 2021-01-18 Vladimir N. Makarov <vmakarov@redhat.com>
24846 * ira.c (ira): Skip abnormal critical edge splitting.
24848 2021-01-18 Jakub Jelinek <jakub@redhat.com>
24850 PR tree-optimization/98727
24851 * tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
24852 second .MUL_OVERFLOW operand for signed multiplication with overflow
24853 checking if the second operand of multiplication is not constant.
24855 2021-01-18 David Edelsohn <dje.gcc@gmail.com>
24857 * doc/invoke.texi (-gdwarf): TPF defaults to version 2 and AIX
24858 defaults to version 4.
24860 2021-01-18 David Malcolm <dmalcolm@redhat.com>
24862 * attribs.h (fndecl_dealloc_argno): New decl.
24863 * builtins.c (call_dealloc_argno): Split out second half of
24865 (fndecl_dealloc_argno): New.
24866 * doc/extend.texi (Common Function Attributes): Document the
24867 interaction between the analyzer and the malloc attribute.
24868 * doc/invoke.texi (Static Analyzer Options): Likewise.
24870 2021-01-17 David Edelsohn <dje.gcc@gmail.com>
24872 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Override
24873 dwarf_version to 4.
24874 * config/rs6000/aix72.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
24876 2021-01-17 Martin Jambor <mjambor@suse.cz>
24879 * cgraph.c (clone_of_p): Check also former_clone_of as we climb
24882 2021-01-17 Mark Wielaard <mark@klomp.org>
24884 * common.opt (gdwarf-): Init(5).
24885 * doc/invoke.texi (-gdwarf): Document default to 5.
24887 2021-01-16 Kwok Cheung Yeung <kcy@codesourcery.com>
24889 * builtin-types.def
24890 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
24892 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
24893 ...this. Add extra argument.
24894 * gimplify.c (omp_default_clause): Ensure that event handle is
24895 firstprivate in a task region.
24896 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_DETACH.
24897 (gimplify_adjust_omp_clauses): Likewise.
24898 * omp-builtins.def (BUILT_IN_GOMP_TASK): Change function type to
24899 BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR.
24900 * omp-expand.c (expand_task_call): Add GOMP_TASK_FLAG_DETACH to flags
24901 if detach clause specified. Add detach argument when generating
24903 * omp-low.c (scan_sharing_clauses): Setup data environment for detach
24905 (finish_taskreg_scan): Move field for variable containing the event
24906 handle to the front of the struct.
24907 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_DETACH. Fix
24909 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
24910 OMP_CLAUSE_DETACH clause.
24911 (convert_local_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
24912 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_DETACH.
24913 * tree.c (omp_clause_num_ops): Add entry for OMP_CLAUSE_DETACH.
24915 (omp_clause_code_name): Add entry for OMP_CLAUSE_DETACH. Fix
24917 (walk_tree_1): Handle OMP_CLAUSE_DETACH.
24919 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
24921 * config/nios2/t-rtems: Reset all MULTILIB_* variables. Shorten
24922 multilib directory names. Use MULTILIB_REQUIRED instead of
24923 MULTILIB_EXCEPTIONS. Add -mhw-mul -mhw-mulx -mhw-div
24924 -mcustom-fpu-cfg=fph2 multilib.
24926 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
24928 * config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
24929 (nios2_init_fpu_configs): Provide register values for new
24930 -mcustom-fpu-cfg=fph2 option variant.
24931 * doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
24934 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
24936 * config/nios2/nios2.c (nios2_custom_check_insns): Remove
24937 custom instruction warnings.
24939 2021-01-16 Jakub Jelinek <jakub@redhat.com>
24941 PR tree-optimization/96669
24942 * match.pd ((CST << x) & 1 -> x == 0): New simplification.
24944 2021-01-16 Jakub Jelinek <jakub@redhat.com>
24946 PR tree-optimization/96271
24947 * passes.def: Pass false argument to first two pass_cd_dce
24948 instances and true to last instance. Add comment that
24949 last instance rewrites no longer addressed locals.
24950 * tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
24952 (pass_cd_dce::set_pass_param): New method.
24953 (pass_cd_dce::execute): Return TODO_update_address_taken from
24954 last cd_dce instance.
24956 2021-01-15 Carl Love <cel@us.ibm.com>
24958 * config/rs6000/altivec.h (vec_mulh, vec_div, vec_dive, vec_mod):
24960 * config/rs6000/altivec.md (VIlong): Move define to file vsx.md.
24961 * config/rs6000/rs6000-builtin.def (DIVES_V4SI, DIVES_V2DI,
24962 DIVEU_V4SI, DIVEU_V2DI, DIVS_V4SI, DIVS_V2DI, DIVU_V4SI,
24963 DIVU_V2DI, MODS_V2DI, MODS_V4SI, MODU_V2DI, MODU_V4SI,
24964 MULHS_V2DI, MULHS_V4SI, MULHU_V2DI, MULHU_V4SI, MULLD_V2DI):
24965 Add builtin define.
24966 (MULH, DIVE, MOD): Add new BU_P10_OVERLOAD_2 definitions.
24967 * config/rs6000/rs6000-call.c (VSX_BUILTIN_VEC_DIV,
24968 VSX_BUILTIN_VEC_DIVE, P10_BUILTIN_VEC_MOD, P10_BUILTIN_VEC_MULH):
24969 New overloaded definitions.
24970 (builtin_function_type) [P10V_BUILTIN_DIVEU_V4SI,
24971 P10V_BUILTIN_DIVEU_V2DI, P10V_BUILTIN_DIVU_V4SI,
24972 P10V_BUILTIN_DIVU_V2DI, P10V_BUILTIN_MODU_V2DI,
24973 P10V_BUILTIN_MODU_V4SI, P10V_BUILTIN_MULHU_V2DI,
24974 P10V_BUILTIN_MULHU_V4SI]: Add case
24975 statement for builtins.
24976 * config/rs6000/rs6000.md (bits): Add new attribute sizes V4SI, V2DI.
24977 * config/rs6000/vsx.md (VIlong): Moved from config/rs6000/altivec.md.
24978 (UNSPEC_VDIVES, UNSPEC_VDIVEU): New unspec definitions.
24979 (vsx_mul_v2di): Add if TARGET_POWER10 statement.
24980 (vsx_udiv_v2di): Add if TARGET_POWER10 statement.
24981 (dives_<mode>, diveu_<mode>, div<mode>3, uvdiv<mode>3,
24982 mods_<mode>, modu_<mode>, mulhs_<mode>, mulhu_<mode>, mulv2di3):
24983 Add define_insn, mode is VIlong.
24984 * doc/extend.texi (vec_mulh, vec_mul, vec_div, vec_dive, vec_mod):
24985 Add builtin descriptions.
24987 2021-01-15 Eric Botcazou <ebotcazou@adacore.com>
24989 * final.c (final_start_function_1): Reset force_source_line.
24991 2021-01-15 Jakub Jelinek <jakub@redhat.com>
24993 PR tree-optimization/96669
24994 * match.pd (((1 << A) & 1) != 0 -> A == 0,
24995 ((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
24996 possibly different power of two constants and to right shift too.
24998 2021-01-15 Jakub Jelinek <jakub@redhat.com>
25000 PR tree-optimization/96681
25001 * match.pd ((x < 0) ^ (y < 0) to (x ^ y) < 0): New simplification.
25002 ((x >= 0) ^ (y >= 0) to (x ^ y) < 0): Likewise.
25003 ((x < 0) ^ (y >= 0) to (x ^ y) >= 0): Likewise.
25004 ((x >= 0) ^ (y < 0) to (x ^ y) >= 0): Likewise.
25006 2021-01-15 Alexandre Oliva <oliva@adacore.com>
25008 * opts.c (gen_command_line_string): Exclude -dumpbase-ext.
25010 2021-01-15 Tamar Christina <tamar.christina@arm.com>
25012 * config/aarch64/aarch64-simd.md (cml<fcmac1><conj_op><mode>4,
25013 cmul<conj_op><mode>3): New.
25014 * config/aarch64/iterators.md (UNSPEC_FCMUL,
25015 UNSPEC_FCMUL180, UNSPEC_FCMLA_CONJ, UNSPEC_FCMLA180_CONJ,
25016 UNSPEC_CMLA_CONJ, UNSPEC_CMLA180_CONJ, UNSPEC_CMUL, UNSPEC_CMUL180,
25017 FCMLA_OP, FCMUL_OP, conj_op, rotsplit1, rotsplit2, fcmac1, sve_rot1,
25018 sve_rot2, SVE2_INT_CMLA_OP, SVE2_INT_CMUL_OP, SVE2_INT_CADD_OP): New.
25019 (rot): Add UNSPEC_FCMUL, UNSPEC_FCMUL180.
25020 (rot_op): Renamed to conj_op.
25021 * config/aarch64/aarch64-sve.md (cml<fcmac1><conj_op><mode>4,
25022 cmul<conj_op><mode>3): New.
25023 * config/aarch64/aarch64-sve2.md (cml<fcmac1><conj_op><mode>4,
25024 cmul<conj_op><mode>3): New.
25026 2021-01-15 David Malcolm <dmalcolm@redhat.com>
25030 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
25031 Escape the tempfile name when constructing the expected output.
25033 2021-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25035 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlsl_hi<mode>):
25037 (aarch64_<su>mlsl_hi<mode>): ... This.
25038 (aarch64_<su>mlsl_hi<mode>): Define.
25039 (*aarch64_<su>mlsl<mode): Rename to...
25040 (aarch64_<su>mlsl<mode): ... This.
25041 * config/aarch64/aarch64-simd-builtins.def (smlsl, umlsl,
25042 smlsl_hi, umlsl_hi): Define builtins.
25043 * config/aarch64/arm_neon.h (vmlsl_high_s8, vmlsl_high_s16,
25044 vmlsl_high_s32, vmlsl_high_u8, vmlsl_high_u16, vmlsl_high_u32,
25045 vmlsl_s8, vmlsl_s16, vmlsl_s32, vmlsl_u8,
25046 vmlsl_u16, vmlsl_u32): Reimplement with builtins.
25048 2021-01-15 Uroš Bizjak <ubizjak@gmail.com>
25050 * config/i386/i386-c.c (ix86_target_macros):
25051 Use cpp_define_formatted for __SIZEOF_FLOAT80__ definition.
25053 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
25056 * config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
25057 * Makefile.in (RTL_SSA_H): New variable.
25058 * config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
25059 * config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
25060 * config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
25062 * config/aarch64/aarch64-cc-fusion.cc: New file.
25064 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
25066 * recog.h (insn_change_watermark::~insn_change_watermark): Avoid
25067 calling cancel_changes for changes that no longer exist.
25069 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
25071 * rtl-ssa/functions.h (function_info::ref_defs): Rename to...
25072 (function_info::reg_defs): ...this.
25073 * rtl-ssa/member-fns.inl (function_info::ref_defs): Rename to...
25074 (function_info::reg_defs): ...this.
25076 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
25079 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
25081 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
25084 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
25087 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
25089 2021-01-15 Richard Biener <rguenther@suse.de>
25091 PR tree-optimization/96376
25092 * tree-vect-stmts.c (get_load_store_type): Disregard alignment
25093 for VMAT_INVARIANT.
25095 2021-01-15 Martin Liska <mliska@suse.cz>
25097 * doc/install.texi: Document that some tests need pytest module.
25098 * doc/sourcebuild.texi: Likewise.
25100 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
25103 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
25105 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
25107 * config/arm/mve.md (mve_vshrq_n_s<mode>_imm): New entry.
25108 (mve_vshrq_n_u<mode>_imm): Likewise.
25109 * config/arm/neon.md (vashr<mode>3, vlshr<mode>3): Move to ...
25110 * config/arm/vec-common.md: ... here.
25112 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
25114 * config/arm/mve.md (mve_vshlq_<supf><mode>): Move to
25116 * config/arm/neon.md (vashl<mode>3): Delete.
25117 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): New.
25118 (vasl<mode>3): New expander.
25120 2021-01-15 Richard Biener <rguenther@suse.de>
25122 PR tree-optimization/98685
25123 * tree-vect-slp.c (vect_schedule_slp_node): Refactor handling
25124 of vector extern defs.
25126 2021-01-14 David Malcolm <dmalcolm@redhat.com>
25129 * diagnostic.c (diagnostic_kind_text): Break out this array
25131 (diagnostic_build_prefix): ...here.
25132 (fancy_abort): Detect when diagnostic_initialize has not yet been
25133 called and fall back to a minimal implementation of printing the
25134 ICE, rather than segfaulting in internal_error.
25136 2021-01-14 David Malcolm <dmalcolm@redhat.com>
25138 * diagnostic.c (diagnostic_initialize): Eliminate
25139 parseable_fixits_p in favor of initializing extra_output_kind from
25140 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
25141 (convert_column_unit): New function, split out from...
25142 (diagnostic_converted_column): ...this.
25143 (print_parseable_fixits): Add "column_unit" and "tabstop" params.
25144 Use them to call convert_column_unit on the column values.
25145 (diagnostic_report_diagnostic): Eliminate conditional on
25146 parseable_fixits_p in favor of a switch statement on
25147 extra_output_kind, passing the appropriate values to the new
25148 params of print_parseable_fixits.
25149 (selftest::test_print_parseable_fixits_none): Update for new
25150 params of print_parseable_fixits.
25151 (selftest::test_print_parseable_fixits_insert): Likewise.
25152 (selftest::test_print_parseable_fixits_remove): Likewise.
25153 (selftest::test_print_parseable_fixits_replace): Likewise.
25154 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
25156 (selftest::diagnostic_c_tests): Call it.
25157 * diagnostic.h (enum diagnostics_extra_output_kind): New.
25158 (diagnostic_context::parseable_fixits_p): Delete field in favor
25160 (diagnostic_context::extra_output_kind): ...this new field.
25161 * doc/invoke.texi (Environment Variables): Add
25162 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
25163 * opts.c (common_handle_option): Update handling of
25164 OPT_fdiagnostics_parseable_fixits for change to diagnostic_context
25167 2021-01-14 Tamar Christina <tamar.christina@arm.com>
25169 * tree-vect-slp-patterns.c (class complex_operations_pattern,
25170 complex_operations_pattern::matches,
25171 complex_operations_pattern::recognize,
25172 complex_operations_pattern::build): New.
25173 (slp_patterns): Use it.
25175 2021-01-14 Tamar Christina <tamar.christina@arm.com>
25177 * internal-fn.def (COMPLEX_FMS, COMPLEX_FMS_CONJ): New.
25178 * optabs.def (cmls_optab, cmls_conj_optab): New.
25179 * doc/md.texi: Document them.
25180 * tree-vect-slp-patterns.c (class complex_fms_pattern,
25181 complex_fms_pattern::matches, complex_fms_pattern::recognize,
25182 complex_fms_pattern::build): New.
25184 2021-01-14 Tamar Christina <tamar.christina@arm.com>
25186 * internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
25187 * optabs.def (cmla_optab, cmla_conj_optab): New.
25188 * doc/md.texi: Document them.
25189 * tree-vect-slp-patterns.c (vect_match_call_p,
25190 class complex_fma_pattern, vect_slp_reset_pattern,
25191 complex_fma_pattern::matches, complex_fma_pattern::recognize,
25192 complex_fma_pattern::build): New.
25194 2021-01-14 Tamar Christina <tamar.christina@arm.com>
25196 * internal-fn.def (COMPLEX_MUL, COMPLEX_MUL_CONJ): New.
25197 * optabs.def (cmul_optab, cmul_conj_optab): New.
25198 * doc/md.texi: Document them.
25199 * tree-vect-slp-patterns.c (vect_match_call_complex_mla,
25200 vect_normalize_conj_loc, is_eq_or_top, vect_validate_multiplication,
25201 vect_build_combine_node, class complex_mul_pattern,
25202 complex_mul_pattern::matches, complex_mul_pattern::recognize,
25203 complex_mul_pattern::build): New.
25205 2021-01-14 Tamar Christina <tamar.christina@arm.com>
25207 * tree-vect-slp.c (optimize_load_redistribution_1): New.
25208 (optimize_load_redistribution, vect_is_slp_load_node): New.
25209 (vect_match_slp_patterns): Use it.
25211 2021-01-14 Tamar Christina <tamar.christina@arm.com>
25213 * tree-vect-slp-patterns.c (complex_add_pattern::build):
25216 2021-01-14 Thomas Schwinge <thomas@codesourcery.com>
25218 * config/gcn/mkoffload.c (main): Create an offload image only in
25219 64-bit configurations.
25221 2021-01-14 H.J. Lu <hjl.tools@gmail.com>
25224 * config/i386/i386-options.c (ix86_option_override_internal):
25225 Issue an error for -fcf-protection with CF_BRANCH when compiling
25226 for 32-bit non-TARGET_CMOV targets.
25228 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
25231 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p):
25232 Remove declaration and initialization of shadow variable "ret".
25233 (ix86_option_override_internal): Remove delcaration of
25234 shadow variable "i". Redeclare shadowed variable to unsigned.
25235 * common/config/i386/i386-common.c (pta_size): Redeclare to unsigned.
25236 * config/i386/i386-builtins.c (get_builtin_code_for_version):
25237 Update for redeclaration.
25238 * config/i386/i386.h (pta_size): Ditto.
25240 2021-01-14 Richard Biener <rguenther@suse.de>
25242 PR tree-optimization/98674
25243 * tree-data-ref.c (base_supports_access_fn_components_p): New.
25244 (initialize_data_dependence_relation): For two bases without
25245 possible access fns resort to type size equality when determining
25246 shape compatibility.
25248 2021-01-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
25251 * config/arm/arm_neon.h: Replace calls to __builtin_vcge* by
25252 <=, >= operators in vcle and vcge intrinsics respectively.
25253 * config/arm/arm_neon_builtins.def: Remove entry for
25256 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
25259 * config/i386/i386-options.c (ix86_function_specific_save):
25260 Remove redundant assignment to opts->x_ix86_branch_cost.
25261 * config/i386/i386.c (ix86_prefetch_sse):
25262 Rename from x86_prefetch_sse. Update all uses.
25263 * config/i386/i386.h: Update for rename.
25264 * config/i386/i386-options.h: Ditto.
25266 2021-01-14 Jakub Jelinek <jakub@redhat.com>
25269 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3,
25270 *sse4_1_zero_extendv4hiv4si2_3, *sse4_1_zero_extendv2siv2di2_3):
25271 Use Bm instead of m for non-avx. Add isa attribute.
25273 2021-01-14 Jakub Jelinek <jakub@redhat.com>
25275 PR tree-optimization/96688
25276 * match.pd (~(X >> Y) -> ~X >> Y): New simplification if
25277 ~X can be simplified.
25279 2021-01-14 Richard Sandiford <richard.sandiford@arm.com>
25281 * tree-vect-stmts.c (vect_model_load_cost): Account for unused
25282 IFN_LOAD_LANES results.
25284 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25286 * config/aarch64/aarch64-simd.md (aarch64_<su>xtl<mode>):
25288 (aarch64_xtn<mode>): Likewise.
25289 * config/aarch64/aarch64-simd-builtins.def (sxtl, uxtl, xtn):
25292 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
25294 (vmovl_s16): Likewise.
25295 (vmovl_s32): Likewise.
25296 (vmovl_u8): Likewise.
25297 (vmovl_u16): Likewise.
25298 (vmovl_u32): Likewise.
25299 (vmovn_s16): Likewise.
25300 (vmovn_s32): Likewise.
25301 (vmovn_s64): Likewise.
25302 (vmovn_u16): Likewise.
25303 (vmovn_u32): Likewise.
25304 (vmovn_u64): Likewise.
25306 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25308 * config/aarch64/aarch64-simd.md (aarch64_<su>qxtn2<mode>_le):
25310 (aarch64_<su>qxtn2<mode>_be): Likewise.
25311 (aarch64_<su>qxtn2<mode>): Likewise.
25312 * config/aarch64/aarch64-simd-builtins.def (sqxtn2, uqxtn2):
25314 * config/aarch64/iterators.md (SAT_TRUNC): Define code_iterator.
25315 (su): Handle ss_truncate and us_truncate.
25316 * config/aarch64/arm_neon.h (vqmovn_high_s16): Reimplement using
25318 (vqmovn_high_s32): Likewise.
25319 (vqmovn_high_s64): Likewise.
25320 (vqmovn_high_u16): Likewise.
25321 (vqmovn_high_u32): Likewise.
25322 (vqmovn_high_u64): Likewise.
25324 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25326 * config/aarch64/aarch64-simd.md (aarch64_xtn2<mode>_le):
25328 (aarch64_xtn2<mode>_be): Likewise.
25329 (aarch64_xtn2<mode>): Likewise.
25330 * config/aarch64/aarch64-simd-builtins.def (xtn2): Define
25332 * config/aarch64/arm_neon.h (vmovn_high_s16): Reimplement using
25334 (vmovn_high_s32): Likewise.
25335 (vmovn_high_s64): Likewise.
25336 (vmovn_high_u16): Likewise.
25337 (vmovn_high_u32): Likewise.
25338 (vmovn_high_u64): Likewise.
25340 2021-01-13 Stafford Horne <shorne@gmail.com>
25342 * config/or1k/or1k.h (ASM_PREFERRED_EH_DATA_FORMAT): New macro.
25344 2021-01-13 Stafford Horne <shorne@gmail.com>
25346 * config/or1k/linux.h (TARGET_ASM_FILE_END): Define macro.
25348 2021-01-13 Stafford Horne <shorne@gmail.com>
25350 * config/or1k/or1k.h (TARGET_CPU_CPP_BUILTINS): Add builtin
25351 define for __or1k_hard_float__.
25353 2021-01-13 Stafford Horne <shorne@gmail.com>
25355 * config/or1k/or1k.h (NO_PROFILE_COUNTERS): Define as 1.
25356 (PROFILE_HOOK): Define to call _mcount.
25357 (FUNCTION_PROFILER): Change from abort to no-op.
25359 2021-01-13 Jakub Jelinek <jakub@redhat.com>
25361 PR tree-optimization/96691
25362 * match.pd ((~X | C) ^ D -> (X | C) ^ (~D ^ C),
25363 (~X & C) ^ D -> (X & C) ^ (D ^ C)): New simplifications if
25364 (~D ^ C) or (D ^ C) can be simplified.
25366 2021-01-13 Richard Biener <rguenther@suse.de>
25368 PR tree-optimization/92645
25369 * match.pd (BIT_FIELD_REF to conversion): Delay canonicalization
25370 until after vector lowering.
25372 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
25374 * config/aarch64/aarch64-sve.md (fnma<mode>4): Extend from SVE_FULL_I
25376 (@aarch64_pred_fnma<mode>, cond_fnma<mode>, *cond_fnma<mode>_2)
25377 (*cond_fnma<mode>_4, *cond_fnma<mode>_any): Likewise.
25379 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
25381 * config/aarch64/aarch64-sve.md (fma<mode>4): Extend from SVE_FULL_I
25383 (@aarch64_pred_fma<mode>, cond_fma<mode>, *cond_fma<mode>_2)
25384 (*cond_fma<mode>_4, *cond_fma<mode>_any): Likewise.
25386 2021-01-13 Richard Biener <rguenther@suse.de>
25388 PR tree-optimization/92645
25389 * tree-vect-slp.c (vect_build_slp_tree_1): Relax supported
25390 BIT_FIELD_REF argument.
25391 (vect_build_slp_tree_2): Record the desired vector type
25392 on the external vector def.
25393 (vectorizable_slp_permutation): Handle required punning
25394 of existing vector defs.
25396 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
25398 * rtl-ssa/accesses.h (def_lookup): Fix order of comparison results.
25400 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
25402 * config/sh/sh.md (movsf_ie): Remove operands[2] test.
25404 2021-01-13 Samuel Thibault <samuel.thibault@ens-lyon.org>
25406 * config.gcc [$target == *-*-gnu*]: Enable
25407 'default_gnu_indirect_function'.
25409 2021-01-13 Jakub Jelinek <jakub@redhat.com>
25412 * optabs.c (expand_vec_perm_const): Don't force v0 and v1 into
25413 registers before calling targetm.vectorize.vec_perm_const, only after
25415 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const): Handle
25416 two argument permutation when one operand is zero vector and only
25417 after that force operands into registers.
25418 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_1): New
25419 define_insn_and_split pattern.
25420 (*avx512bw_zero_extendv32qiv32hi2_1): Likewise.
25421 (*avx512f_zero_extendv16hiv16si2_1): Likewise.
25422 (*avx2_zero_extendv8hiv8si2_1): Likewise.
25423 (*avx512f_zero_extendv8siv8di2_1): Likewise.
25424 (*avx2_zero_extendv4siv4di2_1): Likewise.
25425 * config/mips/mips.c (mips_vectorize_vec_perm_const): Force operands
25427 * config/arm/arm.c (arm_vectorize_vec_perm_const): Likewise.
25428 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Likewise.
25429 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const): Likewise.
25430 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const): Likewise.
25431 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const): Likewise.
25432 * config/gcn/gcn.c (gcn_vectorize_vec_perm_const): Likewise. Use std::swap.
25434 2021-01-13 Martin Liska <mliska@suse.cz>
25436 PR tree-optimization/98455
25437 * gimple-if-to-switch.cc (condition_info::record_phi_mapping):
25438 Record also virtual PHIs.
25439 (pass_if_to_switch::execute): Return TODO_cleanup_cfg only
25442 2021-01-13 Jonathan Wakely <jwakely@redhat.com>
25444 * doc/invoke.texi (C++ Modules): Fix typos.
25446 2021-01-13 Richard Biener <rguenther@suse.de>
25448 PR tree-optimization/98640
25449 * tree-ssa-sccvn.c (visit_nary_op): Do not try to
25450 handle plus or minus from a truncated operand to be
25453 2021-01-13 Jakub Jelinek <jakub@redhat.com>
25456 * config/i386/i386.md (*btr<mode>_1, *btr<mode>_2): New
25457 define_insn_and_split patterns.
25458 (splitter after *btr<mode>_2): New splitter.
25460 2021-01-13 Martin Liska <mliska@suse.cz>
25463 * cgraphunit.c (analyze_functions): Remove dead code.
25465 2021-01-13 Qian Jianhua <qianjh@cn.fujitsu.com>
25467 * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
25468 * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
25469 (a64fx_regmove_cost, a64fx_vector_cost): New.
25470 (a64fx_tunings): Use the new added cost tables.
25472 2021-01-13 Jakub Jelinek <jakub@redhat.com>
25475 * config/i386/predicates.md (pmovzx_parallel): New predicate.
25476 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): New
25477 define_insn_and_split pattern.
25478 (*sse4_1_zero_extendv4hiv4si2_3): Likewise.
25479 (*sse4_1_zero_extendv2siv2di2_3): Likewise.
25481 2021-01-13 Julian Brown <julian@codesourcery.com>
25483 * config/gcn/gcn.c (gcn_conditional_register_usage): Remove dead code
25484 to fix v0 register.
25486 2021-01-13 Julian Brown <julian@codesourcery.com>
25488 * config/gcn/gcn.c (gcn_md_reorg): Fix case where EXEC reg is live
25491 2021-01-13 Julian Brown <julian@codesourcery.com>
25493 * config/gcn/gcn-valu.md (recip<mode>2<exec>, recip<mode>2): Use unspec
25494 for reciprocal-approximation instructions.
25495 (div<mode>3): Use fused multiply-accumulate operations for reciprocal
25496 refinement and division result.
25497 * config/gcn/gcn.md (UNSPEC_RCP): New unspec constant.
25499 2021-01-13 Julian Brown <julian@codesourcery.com>
25501 * config/gcn/gcn-valu.md (subdf): Rename to...
25504 2021-01-12 Martin Liska <mliska@suse.cz>
25506 * gcov.c (source_info::debug): Fix printf format for 32-bit hosts.
25508 2021-01-12 Andrea Corallo <andrea.corallo@arm.com>
25510 * function-abi.h: Fix typo.
25512 2021-01-12 Christophe Lyon <christophe.lyon@linaro.org>
25516 * config/arm/arm.h (ARM_HAVE_NEON_V8QI_LDST): New macro.
25517 (ARM_HAVE_NEON_V16QI_LDST, ARM_HAVE_NEON_V4HI_LDST): Likewise.
25518 (ARM_HAVE_NEON_V8HI_LDST, ARM_HAVE_NEON_V2SI_LDST): Likewise.
25519 (ARM_HAVE_NEON_V4SI_LDST, ARM_HAVE_NEON_V4HF_LDST): Likewise.
25520 (ARM_HAVE_NEON_V8HF_LDST, ARM_HAVE_NEON_V4BF_LDST): Likewise.
25521 (ARM_HAVE_NEON_V8BF_LDST, ARM_HAVE_NEON_V2SF_LDST): Likewise.
25522 (ARM_HAVE_NEON_V4SF_LDST, ARM_HAVE_NEON_DI_LDST): Likewise.
25523 (ARM_HAVE_NEON_V2DI_LDST): Likewise.
25524 (ARM_HAVE_V8QI_LDST, ARM_HAVE_V16QI_LDST): Likewise.
25525 (ARM_HAVE_V4HI_LDST, ARM_HAVE_V8HI_LDST): Likewise.
25526 (ARM_HAVE_V2SI_LDST, ARM_HAVE_V4SI_LDST, ARM_HAVE_V4HF_LDST): Likewise.
25527 (ARM_HAVE_V8HF_LDST, ARM_HAVE_V4BF_LDST, ARM_HAVE_V8BF_LDST): Likewise.
25528 (ARM_HAVE_V2SF_LDST, ARM_HAVE_V4SF_LDST, ARM_HAVE_DI_LDST): Likewise.
25529 (ARM_HAVE_V2DI_LDST): Likewise.
25530 * config/arm/mve.md (*movmisalign<mode>_mve_store): New pattern.
25531 (*movmisalign<mode>_mve_load): New pattern.
25532 * config/arm/neon.md (movmisalign<mode>): Move to ...
25533 * config/arm/vec-common.md: ... here.
25535 2021-01-12 Vladimir N. Makarov <vmakarov@redhat.com>
25538 * lra-eliminations.c (eliminate_regs_in_insn): Add transformation
25539 of pattern 'plus (plus (hard reg, const), pseudo)'.
25541 2021-01-12 Richard Biener <rguenther@suse.de>
25543 PR tree-optimization/98550
25544 * tree-vect-slp.c (vect_record_max_nunits): Check whether
25545 the group size is a multiple of the vector element count.
25546 (vect_build_slp_tree_1): When we need to fail because
25547 the vector type choosen causes unrolling do so lazily
25548 without affecting matches only at the end to guide group splitting.
25550 2021-01-12 Martin Liska <mliska@suse.cz>
25553 * optc-save-gen.awk: Compare also n_target_save vars with
25556 2021-01-12 Martin Liska <mliska@suse.cz>
25558 * gcov.c (source_info::debug): New.
25559 (print_usage): Add --debug (-D) option.
25560 (process_args): Likewise.
25561 (generate_results): Call src->debug after
25562 accumulate_line_counts.
25563 (read_graph_file): Properly assign id for EXIT_BLOCK.
25564 * profile.c (branch_prob): Dump function body before it is
25567 2021-01-12 Jakub Jelinek <jakub@redhat.com>
25569 PR tree-optimization/98629
25570 * tree-ssa-math-opts.c (arith_overflow_check_p): Don't update use_stmt
25571 unless returning non-zero.
25573 2021-01-12 Jakub Jelinek <jakub@redhat.com>
25575 PR tree-optimization/95731
25576 * tree-ssa-reassoc.c (optimize_range_tests_cmp_bitwise): Also optimize
25577 x < 0 && y < 0 && z < 0 into (x | y | z) < 0 for signed x, y, z.
25578 (optimize_range_tests): Call optimize_range_tests_cmp_bitwise
25579 only after optimize_range_tests_var_bound.
25581 2021-01-12 Jakub Jelinek <jakub@redhat.com>
25583 * configure.ac: Ensure c/Make-lang.in comes first in @all_lang_makefrags@.
25584 * configure: Regenerated.
25586 2021-01-12 liuhongt <hongtao.liu@intel.com>
25589 * config/i386/i386-builtins.h (BUILTIN_DESC_SWAP_OPERANDS):
25591 * config/i386/i386-expand.c (ix86_expand_sse_comi): Delete
25594 2021-01-12 Alexandre Oliva <oliva@adacore.com>
25596 * ssa-iterators.h (end_imm_use_stmt_traverse): Forward
25598 (auto_end_imm_use_stmt_traverse): New struct.
25599 (FOR_EACH_IMM_USE_STMT): Use it.
25600 (BREAK_FROM_IMM_USE_STMT, RETURN_FROM_IMM_USE_STMT): Remove,
25602 * gimple-ssa-strength-reduction.c: ... here, ...
25603 * graphite-scop-detection.c: ... here, ...
25604 * ipa-modref.c, ipa-pure-const.c, ipa-sra.c: ... here, ...
25605 * tree-predcom.c, tree-ssa-ccp.c: ... here, ...
25606 * tree-ssa-dce.c, tree-ssa-dse.c: ... here, ...
25607 * tree-ssa-loop-ivopts.c, tree-ssa-math-opts.c: ... here, ...
25608 * tree-ssa-phiprop.c, tree-ssa.c: ... here, ...
25609 * tree-vect-slp.c: ... and here, ...
25610 * doc/tree-ssa.texi: ... and the example here.
25612 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25614 * config/aarch64/aarch64-sve.md (sdiv_pow2<mode>3): Extend from
25615 SVE_FULL_I to SVE_I. Generate an UNSPEC_PRED_X.
25616 (*sdiv_pow2<mode>3): New pattern.
25617 (@cond_<sve_int_op><mode>): Extend from SVE_FULL_I to SVE_I.
25618 Wrap the ASRD in an UNSPEC_PRED_X.
25619 (*cond_<sve_int_op><mode>_2): Likewise. Replace the UNSPEC_PRED_X
25620 predicate with a constant PTRUE, if it isn't already.
25621 (*cond_<sve_int_op><mode>_z): Replace with...
25622 (*cond_<sve_int_op><mode>_any): ...this new pattern.
25624 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25626 * config/aarch64/aarch64-sve.md (*cond_bic<mode>_2): Extend from
25627 SVE_FULL_I to SVE_I.
25628 (*cond_bic<mode>_any): Likewise.
25630 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25632 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart)
25633 (@aarch64_pred_<MUL_HIGHPART:optab><mode>): Extend from SVE_FULL_I
25636 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25638 * config/aarch64/aarch64-sve.md (<su>abd<mode>_3): Extend from
25639 SVE_FULL_I to SVE_I.
25640 (*aarch64_cond_<su>abd<mode>_2): Likewise.
25641 (*aarch64_cond_<su>abd<mode>_any): Likewise.
25642 (@aarch64_pred_<su>abd<mode>): Likewise. Use UNSPEC_PRED_X
25643 for the max and min but not for the minus.
25644 (*aarch64_cond_<su>abd<mode>_3): New pattern.
25646 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25648 * config/aarch64/iterators.md (SVE_24I): New iterator.
25649 * config/aarch64/aarch64-sve.md (*aarch64_adr<mode>_shift): Extend from
25650 SVE_FULL_SDI to SVE_24I. Use containers rather than elements.
25652 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25654 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_BINARY:optab><mode>)
25655 (*cond_<SVE_INT_BINARY:optab><mode>_2): Extend from SVE_FULL_I
25657 (*cond_<SVE_INT_BINARY:optab><mode>_3): Likewise.
25658 (*cond_<SVE_INT_BINARY:optab><mode>_any): Likewise.
25659 (*cond_<SVE_INT_BINARY:optab><mode>_2_const): Likewise.
25660 (*cond_<SVE_INT_BINARY:optab><mode>_any_const): Likewise.
25662 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25664 * config/aarch64/aarch64-sve.md (<SVE_INT_BINARY_IMM:optab><mode>3)
25665 (@aarch64_pred_<SVE_INT_BINARY_IMM:optab><mode>)
25666 (*post_ra_<SVE_INT_BINARY_IMM:optab><mode>3): Extend from SVE_FULL_I
25669 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25671 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3)
25672 (v<ASHIFT:optab><mode>3, @aarch64_pred_<optab><mode>)
25673 (*post_ra_v<ASHIFT:optab><mode>3): Extend from SVE_FULL_I to SVE_I.
25675 2021-01-11 Martin Liska <mliska@suse.cz>
25678 * symtab-clones.h (clone_info::release): Release
25679 symtab::m_clones with ggc_delete as it's a GGC memory.
25681 2021-01-11 Matthias Klose <doko@ubuntu.com>
25683 * Makefile.in (LINK_PROGRESS): Show the link target.
25685 2021-01-11 Richard Biener <rguenther@suse.de>
25687 PR tree-optimization/91403
25688 * tree-vect-data-refs.c (vect_analyze_group_access_1): Cap
25689 single-element interleaving group size at 4096 elements.
25691 2021-01-11 Richard Biener <rguenther@suse.de>
25693 PR tree-optimization/98526
25694 * tree-vect-loop.c (vect_model_reduction_cost): Remove costing
25695 of the actual reduction op for the regular case.
25696 (vectorizable_reduction): Cost the stmts
25697 vect_transform_reduction produces here.
25699 2021-01-11 Andreas Krebbel <krebbel@linux.ibm.com>
25701 * tree-ssa-forwprop.c (simplify_vector_constructor): For
25702 big-endian, use UNPACK[_FLOAT]_HI.
25704 2021-01-11 Tamar Christina <tamar.christina@arm.com>
25706 * tree-vect-slp-patterns.c (class complex_pattern,
25707 class complex_add_pattern): Add parameters to matches.
25708 (complex_add_pattern::build): Free memory.
25709 (complex_add_pattern::matches): Move validation end of match.
25710 (complex_add_pattern::recognize): Likewise.
25712 2021-01-11 Tamar Christina <tamar.christina@arm.com>
25714 * tree-vect-slp-patterns.c (linear_loads_p): Fix externals.
25716 2021-01-11 Tamar Christina <tamar.christina@arm.com>
25718 * tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
25720 2021-01-11 Jakub Jelinek <jakub@redhat.com>
25722 PR tree-optimization/95867
25723 * tree-ssa-math-opts.h: New header.
25724 * tree-ssa-math-opts.c: Include tree-ssa-math-opts.h.
25725 (powi_as_mults): No longer static. Use build_one_cst instead of
25726 build_real. Formatting fix.
25727 * tree-ssa-reassoc.c: Include tree-ssa-math-opts.h.
25728 (attempt_builtin_powi): Handle multiplication reassociation without
25729 powi_fndecl using powi_as_mults.
25730 (reassociate_bb): For integral types don't require
25731 -funsafe-math-optimizations to call attempt_builtin_powi.
25733 2021-01-11 Jakub Jelinek <jakub@redhat.com>
25735 PR tree-optimization/95852
25736 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): Change
25737 mul_stmts parameter type to vec<gimple *> &. Before cond_stmt
25738 allow in the bb any of the stmts in that vector, div_stmt and
25739 up to 3 cast stmts.
25740 (arith_cast_equal_p): New function.
25741 (arith_overflow_check_p): Add cast_stmt argument, handle signed
25742 multiply overflow checks.
25743 (match_arith_overflow): Adjust caller. Handle signed multiply
25746 2021-01-11 Jakub Jelinek <jakub@redhat.com>
25748 PR tree-optimization/95852
25749 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): New function.
25750 (uaddsub_overflow_check_p): Renamed to ...
25751 (arith_overflow_check_p): ... this. Handle also multiplication
25752 with overflow check.
25753 (match_uaddsub_overflow): Renamed to ...
25754 (match_arith_overflow): ... this. Add cfg_changed argument. Handle
25755 also multiplication with overflow check. Adjust function comment.
25756 (math_opts_dom_walker::after_dom_children): Adjust callers. Call
25757 match_arith_overflow also for MULT_EXPR.
25759 2021-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25761 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
25762 __builtin_convertvector.
25763 (vmovl_s16): Likewise.
25764 (vmovl_s32): Likewise.
25765 (vmovl_u8): Likewise.
25766 (vmovl_u16): Likewise.
25767 (vmovl_u32): Likewise.
25768 (vmovn_s16): Likewise.
25769 (vmovn_s32): Likewise.
25770 (vmovn_s64): Likewise.
25771 (vmovn_u16): Likewise.
25772 (vmovn_u32): Likewise.
25773 (vmovn_u64): Likewise.
25775 2021-01-11 Martin Liska <mliska@suse.cz>
25777 * gimple-if-to-switch.cc (struct condition_info): Use auto_var.
25778 (if_chain::is_beneficial): Delete clusters
25779 (find_conditions): Make second argument of conditions_in_bbs a
25780 pointer so that we control over it's lifetime.
25781 (pass_if_to_switch::execute): Delete them.
25783 2021-01-11 Kewen Lin <linkw@linux.ibm.com>
25785 * ira.c (move_unallocated_pseudos): Check other_reg and skip if
25788 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
25790 * config/vax/vax.md (cc): Remove mode attribute.
25791 (subst_<cc>, subst_f<cc>): Rename to...
25792 (subst_<mode>, subst_f<VAXccnz:mode>): ... these respectively.
25793 (*cbranch<VAXint:mode>4_<VAXcc:mode>): Update for `cc' removal.
25794 (*cbranch<VAXfp:mode>4_<VAXccnz:mode>): Likewise.
25795 (*branch_<mode>, *branch_<mode>_reversed): Likewise.
25797 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
25799 * config/vax/vax.md (subst_f<cc>): Add mode to operands and
25800 `const_double_zero'.
25802 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
25804 * config/pdp11/pdp11.md (PDPfp): New mode iterator.
25805 (fcc_cc, fcc_ccnz): Use it. Add mode to `const_double_zero' and
25808 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
25810 * genemit.c (gen_exp) <CONST_DOUBLE>: Handle `const_double_zero'
25812 * read-rtl.c (rtx_reader::read_rtx_code): Handle machine mode
25813 with `const_double_zero'.
25814 * doc/rtl.texi (Constant Expression Types): Document it.
25816 2021-01-09 Jakub Jelinek <jakub@redhat.com>
25819 * tree-cfg.c (verify_gimple_assign_binary): Allow lhs of
25820 POINTER_DIFF_EXPR to be any integral type.
25822 2021-01-09 Jakub Jelinek <jakub@redhat.com>
25824 PR rtl-optimization/98603
25825 * function.c (instantiate_virtual_regs_in_insn): For asm goto
25826 with impossible constraints, drop all SETs, CLOBBERs, drop PARALLEL
25827 if any, set ASM_OPERANDS mode to VOIDmode and change
25828 ASM_OPERANDS_OUTPUT_CONSTRAINT and ASM_OPERANDS_OUTPUT_IDX.
25830 2021-01-09 Alexandre Oliva <oliva@gnu.org>
25833 * final.c (notice_source_line): Narrow down the condition to
25834 skip a line-0 marker.
25836 2021-01-08 Sergei Trofimovich <siarheit@google.com>
25838 * ipa-modref.c (merge_call_side_effects): Fix
25839 linebreak split by reordering two print calls.
25841 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
25843 * config/s390/vector.md (*tf_to_fprx2_0): Rename from
25844 "*mov_tf_to_fprx2_0" for consistency, fix constraint.
25845 (*tf_to_fprx2_1): Rename from "*mov_tf_to_fprx2_1" for
25846 consistency, fix constraint.
25848 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
25850 * config/s390/s390-c.c (s390_def_or_undef_macro): Accept
25851 callables instead of mask values.
25852 (struct target_flag_set_p): New predicate.
25853 (s390_cpu_cpp_builtins_internal): Define or undefine
25854 __LONG_DOUBLE_VX__ macro.
25856 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
25859 * config/i386/i386.c (x86_function_profiler): Use R10 and R11
25860 to call mcount in large model with PIC for NO_PROFILE_COUNTERS
25863 2021-01-08 Richard Biener <rguenther@suse.de>
25865 * tree-ssa-sccvn.c (pass_fre::execute): Reset the SCEV hash table.
25867 2021-01-08 Richard Biener <rguenther@suse.de>
25869 * tree-vect-slp.c (scalar_stmts_to_slp_tree_map_t): Fix.
25870 (vect_build_slp_tree): On cache hit release the matched
25871 scalar stmts vector.
25872 * tree-vect-stmts.c (vectorizable_store): Properly free
25873 vec_oprnds before possibly gathering them again.
25875 2021-01-08 Richard Biener <rguenther@suse.de>
25877 PR tree-optimization/98544
25878 * tree-vect-slp.c (vect_optimize_slp): Always materialize
25879 permutes at a permute node.
25881 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
25884 * config/i386/i386.c (x86_function_profiler): Use R10 to call
25885 mcount in large model. Sorry for large model with PIC.
25887 2021-01-08 Jakub Jelinek <jakub@redhat.com>
25890 * config/i386/i386.opt (ix86_cmodel, ix86_incoming_stack_boundary_arg,
25891 ix86_pmode, ix86_preferred_stack_boundary_arg, ix86_regparm,
25892 ix86_veclibabi_type): Remove x_ prefix, use TargetVariable instead of
25893 TargetSave and initialize for variables with enum types.
25894 (mfentry, mstack-protector-guard-reg=, mstack-protector-guard-offset=,
25895 mstack-protector-guard-symbol=): Add Save.
25896 * config/i386/i386-options.c (ix86_function_specific_save,
25897 ix86_function_specific_restore): Don't save or restore x_ix86_cmodel,
25898 x_ix86_incoming_stack_boundary_arg, x_ix86_pmode,
25899 x_ix86_preferred_stack_boundary_arg, x_ix86_regparm,
25900 x_ix86_veclibabi_type.
25902 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
25904 * config/aarch64/aarch64-sve.md (*cnot<mode>): Extend from
25905 SVE_FULL_I to SVE_I.
25906 (*cond_cnot<mode>_2, *cond_cnot<mode>_any): Likewise.
25908 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
25910 * config/aarch64/aarch64-sve.md (*cond_uxt<mode>_2): Extend from
25911 SVE_FULL_I to SVE_I.
25912 (*cond_uxt<mode>_any): Likewise.
25914 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25916 * config/aarch64/iterators.md (Vwhalf): New iterator.
25917 * config/aarch64/aarch64-simd.md (aarch64_<sur>adalp<mode>_3):
25919 (aarch64_<sur>adalp<mode>): ... This. Make more
25921 (<sur>sadv16qi): Adjust callsite of the above.
25922 * config/aarch64/aarch64-simd-builtins.def (sadalp, uadalp): New
25924 * config/aarch64/arm_neon.h (vpadal_s8): Reimplement using
25926 (vpadal_s16): Likewise.
25927 (vpadal_u8): Likewise.
25928 (vpadal_u16): Likewise.
25929 (vpadalq_s8): Likewise.
25930 (vpadalq_s16): Likewise.
25931 (vpadalq_s32): Likewise.
25932 (vpadalq_u8): Likewise.
25933 (vpadalq_u16): Likewise.
25934 (vpadalq_u32): Likewise.
25936 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25938 * config/aarch64/aarch64-simd.md (aarch64_<su>abd<mode>_3):
25940 (aarch64_<su>abd<mode>): ... This.
25941 (<sur>sadv16qi): Adjust callsite of the above.
25942 * config/aarch64/aarch64-simd-builtins.def (sabd, uabd): Define
25944 * config/aarch64/arm_neon.h (vabd_s8): Reimplement using
25946 (vabd_s16): Likewise.
25947 (vabd_s32): Likewise.
25948 (vabd_u8): Likewise.
25949 (vabd_u16): Likewise.
25950 (vabd_u32): Likewise.
25951 (vabdq_s8): Likewise.
25952 (vabdq_s16): Likewise.
25953 (vabdq_s32): Likewise.
25954 (vabdq_u8): Likewise.
25955 (vabdq_u16): Likewise.
25956 (vabdq_u32): Likewise.
25958 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25960 * config/aarch64/aarch64-simd-builtins.def (saba, uaba): Define
25962 * config/aarch64/arm_neon.h (vaba_s8): Implement using builtin.
25963 (vaba_s16): Likewise.
25964 (vaba_s32): Likewise.
25965 (vaba_u8): Likewise.
25966 (vaba_u16): Likewise.
25967 (vaba_u32): Likewise.
25968 (vabaq_s8): Likewise.
25969 (vabaq_s16): Likewise.
25970 (vabaq_s32): Likewise.
25971 (vabaq_u8): Likewise.
25972 (vabaq_u16): Likewise.
25973 (vabaq_u32): Likewise.
25975 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25977 * config/aarch64/aarch64-simd.md (aba<mode>_3): Rename to...
25978 (aarch64_<su>aba<mode>): ... This. Handle uaba as well.
25979 Change RTL pattern to match.
25981 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
25983 * common/config/riscv/riscv-common.c (riscv_current_subset_list): New.
25984 * config/riscv/riscv-c.c (riscv-subset.h): New.
25985 (INCLUDE_STRING): Define.
25986 (riscv_cpu_cpp_builtins): Add new style architecture extension
25988 * config/riscv/riscv-subset.h (riscv_subset_list::begin): New.
25989 (riscv_subset_list::end): New.
25990 (riscv_current_subset_list): New.
25992 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
25994 * common/config/riscv/riscv-common.c (RISCV_DONT_CARE_VERSION):
25995 Move to riscv-subset.h.
25996 (struct riscv_subset_t): Ditto.
25997 (class riscv_subset_list): Ditto.
25998 * config/riscv/riscv-subset.h (RISCV_DONT_CARE_VERSION): Move
25999 from riscv-common.c.
26000 (struct riscv_subset_t): Ditto.
26001 (class riscv_subset_list): Ditto.
26002 * config/riscv/t-riscv ($(common_out_file)): Add file
26005 2021-01-07 Jakub Jelinek <jakub@redhat.com>
26008 * config/i386/i386.md (*bmi_blsi_<mode>_cmp, *bmi_blsi_<mode>_ccno):
26009 New define_insn patterns.
26011 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
26013 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_UNARY:optab><mode>)
26014 (*cond_<SVE_INT_UNARY:optab><mode>_2): Extend from SVE_FULL_I to SVE_I.
26015 (*cond_<SVE_INT_UNARY:optab><mode>_any): Likewise.
26017 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
26019 PR tree-optimization/98560
26020 * internal-fn.def (IFN_VCONDU, IFN_VCONDEQ): Use type vec_cond.
26021 * internal-fn.c (vec_cond_mask_direct): Get the data mode from
26023 (vec_cond_direct): Likewise argument 2.
26024 (vec_condu_direct, vec_condeq_direct): Delete.
26025 (expand_vect_cond_optab_fn): Rename to...
26026 (expand_vec_cond_optab_fn): ...this, replacing old macro.
26027 (expand_vec_condu_optab_fn, expand_vec_condeq_optab_fn): Delete.
26028 (expand_vect_cond_mask_optab_fn): Rename to...
26029 (expand_vec_cond_mask_optab_fn): ...this, replacing old macro.
26030 (direct_vec_cond_mask_optab_supported_p): Treat the optab as a
26032 (direct_vec_cond_optab_supported_p): Likewise.
26033 (direct_vec_condu_optab_supported_p): Delete.
26034 (direct_vec_condeq_optab_supported_p): Delete.
26035 * gimple-isel.cc: Include internal-fn.h.
26036 (gimple_expand_vec_cond_expr): Check that IFN_VCONDEQ is supported
26039 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
26041 PR tree-optimization/98560
26042 * gimple-isel.cc (gimple_expand_vec_cond_expr): If we fail to use
26043 IFN_VCOND{,U,EQ}, fall back on IFN_VCOND_MASK.
26045 2021-01-07 Uroš Bizjak <ubizjak@gmail.com>
26047 * config/i386/i386.md (insn): Merge from plusminus_insn, shift_insn,
26048 rotate_insn and optab code attributes.
26049 Update all uses to merged code attribute.
26050 * config/i386/sse.md: Update all uses to merged code attribute.
26051 * config/i386/mmx.md: Update all uses to merged code attribute.
26053 2021-01-07 Jakub Jelinek <jakub@redhat.com>
26055 PR tree-optimization/98568
26056 * gimple-ssa-store-merging.c (bswap_view_convert): New function.
26057 (bswap_replace): Use it.
26059 2021-01-06 Vladimir N. Makarov <vmakarov@redhat.com>
26061 PR rtl-optimization/97978
26062 * lra-int.h (lra_hard_reg_split_p): New external.
26063 * lra.c (lra_hard_reg_split_p): New global.
26064 (lra): Set up lra_hard_reg_split_p after splitting a hard reg.
26065 * lra-assigns.c (lra_assign): Don't check allocation correctness
26066 after hard reg splitting.
26068 2021-01-06 Martin Sebor <msebor@redhat.com>
26071 * builtins.c (new_delete_mismatch_p): New overload.
26072 (new_delete_mismatch_p (tree, tree)): Call it.
26074 2021-01-06 Alexandre Oliva <oliva@adacore.com>
26076 * Makefile.in (T_GLIMITS_H): New.
26077 (stmp-int-hdrs): Depend on it, use it.
26078 * config/t-vxworks (T_GLIMITS_H): Override it.
26079 (vxw-glimits.h): New.
26081 2021-01-06 Richard Biener <rguenther@suse.de>
26083 PR tree-optimization/98513
26084 * value-range.cc (intersect_ranges): Compare the upper bounds
26085 for the expected relation.
26087 2021-01-06 Gerald Pfeifer <gerald@pfeifer.com>
26090 2020-12-28 Gerald Pfeifer <gerald@pfeifer.com>
26092 * doc/standards.texi (HSAIL): Remove section.
26094 2021-01-05 Samuel Thibault <samuel.thibault@ens-lyon.org>
26096 * configure: Re-generate.
26098 2021-01-05 Jakub Jelinek <jakub@redhat.com>
26100 * doc/invoke.texi (-std=c++20): Adjust for the publication of
26101 ISO 14882:2020 standard.
26102 * doc/standards.texi: Likewise.
26104 2021-01-05 Jakub Jelinek <jakub@redhat.com>
26106 PR tree-optimization/94802
26107 * expr.h (maybe_optimize_sub_cmp_0): Declare.
26108 * expr.c: Include tree-pretty-print.h and flags.h.
26109 (maybe_optimize_sub_cmp_0): New function.
26110 (do_store_flag): Use it.
26111 * cfgexpand.c (expand_gimple_cond): Likewise.
26113 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
26115 * mux-utils.h (pointer_mux::m_ptr): Tweak description of contents.
26116 * rtlanal.c (simple_regno_set): Tweak description to clarify the
26119 2021-01-05 Richard Biener <rguenther@suse.de>
26121 PR tree-optimization/98516
26122 * tree-vect-slp.c (vect_optimize_slp): Permute the incoming
26123 lanes when materializing on a VEC_PERM node.
26124 (vectorizable_slp_permutation): Dump the permute properly.
26126 2021-01-05 Richard Biener <rguenther@suse.de>
26128 * tree-vect-slp.c (vect_slp_region): Move debug counter
26129 to cover individual subgraphs.
26131 2021-01-05 Richard Biener <rguenther@suse.de>
26133 PR tree-optimization/98428
26134 * tree-vect-slp.c (vect_build_slp_tree_1): Properly reject
26135 vector lane extracts for loop vectorization.
26137 2021-01-05 Jakub Jelinek <jakub@redhat.com>
26139 PR tree-optimization/98514
26140 * tree-ssa-reassoc.c (bb_rank): Change type from long * to
26142 (operand_rank): Change type from hash_map<tree, long> to
26143 hash_map<tree, int64_t>.
26144 (phi_rank): Change return type from long to int64_t.
26145 (loop_carried_phi): Change block_rank variable type from long to
26147 (propagate_rank): Change return type, rank parameter type and
26148 op_rank variable type from long to int64_t.
26149 (find_operand_rank): Change return type from long to int64_t
26150 and change slot variable type from long * to int64_t *.
26151 (insert_operand_rank): Change rank parameter type from long to
26153 (get_rank): Change return type and rank variable type from long to
26154 int64_t. Use PRId64 instead of ld to print the rank.
26155 (init_reassoc): Change rank variable type from long to int64_t
26156 and adjust correspondingly bb_rank and operand_rank initialization.
26158 2021-01-05 Jakub Jelinek <jakub@redhat.com>
26160 PR tree-optimization/96928
26161 * tree-ssa-phiopt.c (xor_replacement): New function.
26162 (tree_ssa_phiopt_worker): Call it.
26164 2021-01-05 Jakub Jelinek <jakub@redhat.com>
26166 PR tree-optimization/96930
26167 * match.pd ((A / (1 << B)) -> (A >> B)): If A is extended
26168 from narrower value which has the same type as 1 << B, perform
26169 the right shift on the narrower value followed by extension.
26171 2021-01-05 Jakub Jelinek <jakub@redhat.com>
26173 PR tree-optimization/96239
26174 * gimple-ssa-store-merging.c (maybe_optimize_vector_constructor): New
26176 (get_status_for_store_merging): Don't return BB_INVALID for blocks
26177 with potential bswap optimizable CONSTRUCTORs.
26178 (pass_store_merging::execute): Optimize vector CONSTRUCTORs with bswap
26181 2021-01-05 Richard Biener <rguenther@suse.de>
26183 PR tree-optimization/98381
26184 * tree.c (vector_element_bits): Properly compute bool vector
26186 * tree-vect-loop.c (vectorizable_live_operation): Properly
26187 compute the last lane bit offset.
26189 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
26192 * config/i386/sse.md (sse_cvtps2pi): Redefine as define_insn_and_split.
26193 Clear the top 64 bytes of the input XMM register.
26194 (sse_cvttps2pi): Ditto.
26196 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
26199 * config/i386/xopintrin.h (_mm256_cmov_si256): New.
26201 2021-01-05 H.J. Lu <hjl.tools@gmail.com>
26204 * config/i386/xmmintrin.h (_mm_extract_pi16): Cast to unsigned
26207 2021-01-05 Claudiu Zissulescu <claziss@synopsys.com>
26209 * config/arc/arc.md (maddsidi4_split): Use ACC_REG_FIRST.
26210 (umaddsidi4_split): Likewise.
26212 2021-01-05 liuhongt <hongtao.liu@intel.com>
26215 * config/i386/sse.md (*sse2_pmovskb_zexthisi): New
26216 define_insn_and_split for zero_extend of subreg HI of pmovskb
26218 (*sse2_pmovskb_zexthisi): Add new combine splitters for
26219 zero_extend of not of subreg HI of pmovskb result.
26221 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
26224 * explow.c (convert_memory_address_addr_space_1): Handle UNSPECs
26226 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate): Use
26227 convert_memory_address to convert symbolic immediates to ptr_mode
26228 before forcing them to memory.
26230 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
26232 PR rtl-optimization/97144
26233 * recog.c (constrain_operands): Initialize matching_operand
26234 for each alternative, rather than only doing it once.
26236 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
26238 PR rtl-optimization/98403
26239 * rtl-ssa/changes.cc (function_info::finalize_new_accesses): Explain
26240 why we don't remove call clobbers.
26241 (function_info::apply_changes_to_insn): Don't attempt to add
26242 call clobbers here.
26244 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
26246 PR tree-optimization/98371
26247 * tree-vect-loop.c (vect_reanalyze_as_main_loop): New function.
26248 (vect_analyze_loop): If an epilogue loop appears to be cheaper
26249 than the main loop, re-analyze it as a main loop before adopting
26252 2021-01-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
26255 * configure.ac (NETLIBS): Determine using AX_LIB_SOCKET_NSL.
26256 * aclocal.m4, configure: Regenerate.
26257 * Makefile.in (NETLIBS): Define.
26258 (BACKEND): Remove $(CODYLIB).
26260 2021-01-05 Jakub Jelinek <jakub@redhat.com>
26262 PR rtl-optimization/98334
26263 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
26264 Optimize (X - 1) * Y + Y to X * Y or (X + 1) * Y - Y to X * Y.
26266 2021-01-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
26268 * tree-inline.c (expand_call_inline): Restore input_location.
26269 Return result from recursive call.
26271 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
26273 PR tree-optimization/95401
26274 * config/aarch64/aarch64-sve-builtins.cc
26275 (gimple_folder::load_store_cookie): Use bits rather than bytes
26276 for the alignment argument to IFN_MASK_LOAD and IFN_MASK_STORE.
26277 * gimple-fold.c (gimple_fold_mask_load_store_mem_ref): Likewise.
26278 * tree-vect-stmts.c (vectorizable_store): Likewise.
26279 (vectorizable_load): Likewise.
26281 2021-01-04 Richard Biener <rguenther@suse.de>
26283 PR tree-optimization/98308
26284 * tree-vect-stmts.c (vectorizable_load): Set invariant mask
26287 2021-01-04 Jakub Jelinek <jakub@redhat.com>
26289 PR tree-optimization/95771
26290 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Handle types
26291 with precision smaller than int's precision and types with precision
26292 twice as large as long long. Formatting fixes.
26294 2021-01-04 Richard Biener <rguenther@suse.de>
26296 PR tree-optimization/98464
26297 * tree-ssa-sccvn.c (vn_valueize_for_srt): Rename from ...
26298 (vn_valueize_wrapper): ... this. Temporarily adjust vn_context_bb.
26299 (process_bb): Adjust.
26301 2021-01-04 Matthew Malcomson <matthew.malcomson@arm.com>
26304 * doc/invoke.texi (-fsanitize=address): Fix wording describing
26305 clash with -fsanitize=hwaddress.
26307 2021-01-04 Richard Biener <rguenther@suse.de>
26309 PR tree-optimization/98282
26310 * tree-ssa-sccvn.c (vn_get_stmt_kind): Classify tcc_reference on
26311 invariants as VN_NARY.
26313 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
26316 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Accept
26317 aarch64_simd_reg_or_zero for operand 2. Use the combinez patterns
26318 to handle zero operands.
26320 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
26322 * config/aarch64/aarch64.c (offset_6bit_signed_scaled_p): New function.
26323 (offset_6bit_unsigned_scaled_p): Fix typo in comment.
26324 (aarch64_sve_prefetch_operand_p): Accept MUL VLs in the range
26327 2021-01-04 Richard Biener <rguenther@suse.de>
26329 PR tree-optimization/98393
26330 * tree-vect-slp.c (vect_build_slp_tree): Properly zero matches
26331 when hitting the limit.
26333 2021-01-04 Richard Biener <rguenther@suse.de>
26335 PR tree-optimization/98291
26336 * tree-vect-loop.c (vectorizable_reduction): Bypass
26337 associativity check for SLP reductions with VF 1.
26339 2021-01-04 Jakub Jelinek <jakub@redhat.com>
26341 PR tree-optimization/96782
26342 * match.pd (x == ~x -> false, x != ~x -> true): New simplifications.
26344 2021-01-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
26346 * collect-utils.c (collect_execute): Check dumppfx.
26347 * collect2.c (maybe_run_lto_and_relink, do_link): Pass atsuffix
26348 to collect_execute.
26349 (do_link): Add new parameter atsuffix.
26350 (main): Handle -dumpdir option. Skip one argument for
26351 -o, -isystem and -B options.
26352 * gcc.c (make_at_file): New helper function.
26353 (close_at_file): Use it.
26355 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26357 * config/darwin.h (MIN_LD64_NO_COAL_SECTS): Adjust.
26358 Amend handling for LD64_VERSION fallback defaults.
26360 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26362 * config.gcc: Compute default version information
26363 from the configured target. Likewise defaults for
26365 * config/darwin10.h: Removed.
26366 * config/darwin12.h: Removed.
26367 * config/darwin9.h: Removed.
26368 * config/rs6000/darwin8.h: Removed.
26370 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26372 * config/darwin9.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
26374 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26376 * config/darwin9.h (STACK_CHECK_STATIC_BUILTIN): Move from here..
26377 * config/darwin.h (STACK_CHECK_STATIC_BUILTIN): .. to here.
26379 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26381 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move from
26383 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): ... to here.
26385 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26387 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move the spec
26388 for the Darwin10 unwinder stub from here ...
26389 * config/darwin.h (LINK_COMMAND_SPEC_A): ... to here.
26391 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26393 * config/darwin.h (DSYMUTIL_SPEC): Default to DWARF
26394 (ASM_DEBUG_SPEC):Only define if the assembler supports
26396 (PREFERRED_DEBUGGING_TYPE): Default to DWARF.
26397 (DARWIN_PREFER_DWARF): Define.
26398 * config/darwin9.h (PREFERRED_DEBUGGING_TYPE): Remove.
26399 (DARWIN_PREFER_DWARF): Likewise
26400 (DSYMUTIL_SPEC): Likewise.
26401 (COLLECT_RUN_DSYMUTIL): Likewise.
26402 (ASM_DEBUG_SPEC): Likewise.
26403 (ASM_DEBUG_OPTION_SPEC): Likewise.
26405 2021-01-02 Jan Hubicka <jh@suse.cz>
26407 * cfg.c (free_block): ggc_free bb.
26409 2021-01-01 Jakub Jelinek <jakub@redhat.com>
26411 * gcc.c (process_command): Update copyright notice dates.
26412 * gcov-dump.c (print_version): Ditto.
26413 * gcov.c (print_version): Ditto.
26414 * gcov-tool.c (print_version): Ditto.
26415 * gengtype.c (create_file): Ditto.
26416 * doc/cpp.texi: Bump @copying's copyright year.
26417 * doc/cppinternals.texi: Ditto.
26418 * doc/gcc.texi: Ditto.
26419 * doc/gccint.texi: Ditto.
26420 * doc/gcov.texi: Ditto.
26421 * doc/install.texi: Ditto.
26422 * doc/invoke.texi: Ditto.
26424 2021-01-01 Jakub Jelinek <jakub@redhat.com>
26426 * ChangeLog-2020: Rotate ChangeLog. New file.
26429 Copyright (C) 2021 Free Software Foundation, Inc.
26431 Copying and distribution of this file, with or without modification,
26432 are permitted in any medium without royalty provided the copyright
26433 notice and this notice are preserved.