1 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
3 * config/darwin.h (LINK_COMMAND_SPEC_A): Update 'r' handling to
4 skip gomp and itm when r or nodefaultlibs is given.
5 (DSYMUTIL_SPEC): Do not call dsymutil for '-r' link lines.
6 Update ordering of exclusions, remove duplicate 'v' addition
7 (collect2 will add this from the main command line).
9 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
11 * config/darwin-driver.c (darwin_driver_init): Revise comments, handle
12 filelist and framework options in specs instead of code.
13 * config/darwin.h (SUBTARGET_DRIVER_SELF_SPECS): Update to handle link
14 specs that are really driver ones.
15 (DARWIN_CC1_SPEC): Likewise.
17 (SYSROOT_SPEC): Append space.
18 (LINK_SYSROOT_SPEC): Remove most driver link specs.
19 (STANDARD_STARTFILE_PREFIX_2): Update link-related specs.
20 (STARTFILE_SPEC): Likewise.
21 (ASM_MMACOSX_VERSION_MIN_SPEC): Fix line wrap.
22 (ASM_SPEC): Update driver-related specs.
23 (ASM_FINAL_SPEC): Likewise.
24 * config/darwin.opt: Remove now unused option aliases.
25 * config/i386/darwin.h (EXTRA_ASM_OPTS): Ensure space after opt.
26 (ASM_SPEC): Update driver-related specs.
28 2021-10-15 Roger Sayle <roger@nextmovesoftware.com>
30 * config/i386/i386.c (ix86_hardreg_mov_ok): For vector modes,
31 allow standard_sse_constant_p immediate constants.
33 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
35 * config.gcc: Add tm-dwarf2.h to tm_d-file.
37 2021-10-15 Andrew MacLeod <amacleod@redhat.com>
39 * gimple-range-fold.h (gimple_range_ssa_p): Don't process names
40 that occur in abnormal phis.
41 * gimple-range.cc (gimple_ranger::range_on_edge): Return false for
42 abnormal and EH edges.
43 * gimple-ssa-evrp.c (rvrp_folder::value_of_expr): Ditto.
44 (rvrp_folder::value_on_edge): Ditto.
45 (rvrp_folder::value_of_stmt): Ditto.
46 (hybrid_folder::value_of_expr): Ditto for ranger queries.
47 (hybrid_folder::value_on_edge): Ditto.
48 (hybrid_folder::value_of_stmt): Ditto.
49 * value-query.cc (gimple_range_global): Always return a range if
50 the type is supported.
52 2021-10-15 John David Anglin <danglin@gcc.gnu.org>
54 * config/pa/pa.md: Consistently use "rG" constraint for copy
55 instruction in move patterns.
57 2021-10-15 Andrew Stubbs <ams@codesourcery.com>
59 * config/gcn/gcn-hsa.h (S_FIJI): Set unconditionally.
62 * config/gcn/gcn.c: Hard code SRAM ECC settings for old architectures.
63 * config/gcn/mkoffload.c (ELFABIVERSION_AMDGPU_HSA): Rename to ...
64 (ELFABIVERSION_AMDGPU_HSA_V3): ... this.
65 (ELFABIVERSION_AMDGPU_HSA_V4): New.
66 (SET_SRAM_ECC_UNSUPPORTED): New.
67 (copy_early_debug_info): Create elf flags to match the other objects.
68 (main): Just let the attribute flags pass through.
70 2021-10-15 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
72 * tree-loop-distribution.c (reduction_var_overflows_first):
73 Pass the type of reduction_var as first argument as it is also
74 done for the load type.
75 (loop_distribution::transform_reduction_loop): Add missing
76 TREE_TYPE while determining precission of reduction_var.
78 2021-10-15 Richard Biener <rguenther@suse.de>
80 * defaults.h (PREFERRED_DEBUGGING_TYPE): Choose DWARF2_DEBUG
82 * toplev.c (process_options): Warn when STABS debugging is
83 enabled but not the preferred format.
84 * config/pa/som.h (PREFERRED_DEBUGGING_TYPE): Define to
86 * config/pdp11/pdp11.h (PREFERRED_DEBUGGING_TYPE): Likewise.
88 2021-10-15 Richard Biener <rguenther@suse.de>
91 * tree-inline.c (copy_bb): Avoid underflowing nargs.
93 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
95 * config/i386/i386-expand.c (ix86_expand_vec_perm): Convert
96 HFmode input operand to HImode.
97 (ix86_vectorize_vec_perm_const): Likewise.
98 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1_hf):
100 (*avx512f_permvar_truncv8siv8hi_1_hf):
103 2021-10-15 Richard Biener <rguenther@suse.de>
106 * expmed.c (store_bit_field_1): Ensure a LHS subreg would
107 not create a paradoxical subreg.
109 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
111 * config/i386/i386-expand.c (ix86_expand_vector_init):
112 For half_vector concat for HFmode, handle them like HImode.
114 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
116 * tree-ssa-threadbackward.c (class back_threader): Add m_resolve.
117 (back_threader::back_threader): Same.
118 (back_threader::resolve_phi): Try to solve without looking back if
120 (back_threader::find_paths_to_names): Same.
121 (try_thread_blocks): Pass resolve argument to back threader.
122 (pass_early_thread_jumps::execute): Same.
124 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
126 * doc/invoke.texi: Remove max-fsm-thread-length,
127 max-fsm-thread-paths, and fsm-maximum-phi-arguments.
129 * tree-ssa-threadbackward.c (back_threader::back_threader): Remove
131 (back_threader_registry::back_threader_registry): Same.
132 (back_threader_profitability::profitable_path_p): Remove
133 param_max_fsm_thread-length.
134 (back_threader_registry::register_path): Remove
135 m_max_allowable_paths.
137 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
139 * tree-ssa-threadbackward.c (class back_threader): Make m_imports
141 (back_threader::~back_threader): Do not release m_path.
143 2021-10-14 Andrew MacLeod <amacleod@redhat.com>
145 PR tree-optimization/102738
146 * vr-values.c (simplify_using_ranges::simplify): Handle RSHIFT_EXPR.
148 2021-10-14 Kwok Cheung Yeung <kcy@codesourcery.com>
150 * omp-general.c (omp_check_context_selector): Move from c-omp.c.
151 (omp_mark_declare_variant): Move from c-omp.c.
152 (omp_context_name_list_prop): Update for Fortran strings.
153 * omp-general.h (omp_check_context_selector): New prototype.
154 (omp_mark_declare_variant): New prototype.
156 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
158 * config/rs6000/rs6000.c (rs6000_density_test): Move early
159 exit test further up the function.
161 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
163 * config/arm/arm.c (arm_add_stmt_cost): Delete.
164 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
166 2021-10-14 Martin Jambor <mjambor@suse.cz>
168 * doc/invoke.texi (Optimize Options): Add entry for
169 ipa-cp-recursive-freq-factor.
171 2021-10-14 Tamar Christina <tamar.christina@arm.com>
173 * match.pd: New rule.
175 2021-10-14 Jan Hubicka <hubicka@ucw.cz>
178 * ipa-modref-tree.h (modref_access_node::update2):
179 Also check that parm_offset is unchanged.
180 (modref_ref_node::insert_access): Fix updating of
183 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
185 * tree-ssa-threadbackward.c (back_threader::resolve_phi): Add
188 2021-10-14 Richard Biener <rguenther@suse.de>
190 PR tree-optimization/102659
191 * tree-if-conv.c (if_convertible_gimple_assign_stmt_p): Also
192 rewrite pointer typed undefined overflow operations.
193 (predicate_statements): Likewise. Make sure to emit invariant
194 conversions in the preheader.
195 * tree-vectorizer.c (vect_loop_vectorized_call): Look through
196 non-empty preheaders.
197 * tree-data-ref.c (dr_analyze_indices): Strip useless
198 conversions to the MEM_REF base type.
200 2021-10-14 Martin Liska <mliska@suse.cz>
202 * common.opt: Stop using AUTODETECT_VALUE
203 and use EnabledBy where possible.
204 * opts.c: Enable OPT_fvar_tracking with optimize >= 1.
205 * toplev.c (AUTODETECT_VALUE): Remove macro.
206 (process_options): Simplify by using EnabledBy and
207 OPT_fvar_tracking. Use OPTION_SET_P macro instead of
210 2021-10-14 Jonathan Wright <jonathan.wright@arm.com>
212 * config/aarch64/arm_neon.h (vld1_s8_x3): Use signed type for
214 (vld1_s32_x3): Likewise.
216 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
218 PR tree-optimization/102736
219 PR tree-optimization/102736
220 * gimple-range-path.cc (path_range_query::range_on_path_entry):
221 Assert that the requested range is defined outside the path.
222 (path_range_query::ssa_range_in_phi): Do not call
223 range_on_path_entry for SSA names that are defined within the
226 2021-10-14 Iain Sandoe <iain@sandoe.co.uk>
228 * config/darwin-driver.c (darwin_find_version_from_kernel):
229 Quote internal identifiers and avoid contractions in
231 (darwin_default_min_version): Likewise.
232 (darwin_driver_init): Likewise.
234 2021-10-14 Martin Jambor <mjambor@suse.cz>
236 * params.opt (ipa-cp-recursive-freq-factor): New.
237 * ipa-cp.c (ipcp_value): Switch to inline initialization. New members
238 scc_no, self_recursion_generated_level, same_scc and
239 self_recursion_generated_p.
240 (ipcp_lattice::add_value): Replaced parameter unlimited with
241 same_lat_gen_level, usit it determine limit of values and store it to
243 (ipcp_lattice<valtype>::print): Dump the new fileds.
244 (allocate_and_init_ipcp_value): Take same_lat_gen_level as a new
245 parameter and store it to the new value.
246 (self_recursively_generated_p): Removed.
247 (propagate_vals_across_arith_jfunc): Use self_recursion_generated_p
248 instead of self_recursively_generated_p, store self generation level
250 (value_topo_info<valtype>::add_val): Set scc_no.
251 (value_topo_info<valtype>::propagate_effects): Multiply frequencies of
252 recursively feeding values and self generated values by appropriate
255 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
257 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Remove
258 redundant test for flag_vect_cost_model.
260 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
262 * bitmap.c (debug): New overloaded function for auto_bitmaps.
263 * bitmap.h (debug): Same.
265 2021-10-14 Hongyu Wang <hongyu.wang@intel.com>
267 * config/i386/avx512fp16intrin.h (_mm512_mask_fcmadd_pch):
269 (_mm512_mask3_fcmadd_pch): Likewise.
270 (_mm512_mask_fmadd_pch): Likewise
271 (_mm512_mask3_fmadd_pch): Likewise
272 (_mm512_mask_fcmadd_round_pch): Likewise
273 (_mm512_mask3_fcmadd_round_pch): Likewise
274 (_mm512_mask_fmadd_round_pch): Likewise
275 (_mm512_mask3_fmadd_round_pch): Likewise
276 (_mm_mask_fcmadd_sch): Likewise
277 (_mm_mask3_fcmadd_sch): Likewise
278 (_mm_mask_fmadd_sch): Likewise
279 (_mm_mask3_fmadd_sch): Likewise
280 (_mm_mask_fcmadd_round_sch): Likewise
281 (_mm_mask3_fcmadd_round_sch): Likewise
282 (_mm_mask_fmadd_round_sch): Likewise
283 (_mm_mask3_fmadd_round_sch): Likewise
284 (_mm_fcmadd_round_sch): Likewise
285 * config/i386/avx512fp16vlintrin.h (_mm_mask_fmadd_pch):
287 (_mm_mask3_fmadd_pch): Likewise
288 (_mm256_mask_fmadd_pch): Likewise
289 (_mm256_mask3_fmadd_pch): Likewise
290 (_mm_mask_fcmadd_pch): Likewise
291 (_mm_mask3_fcmadd_pch): Likewise
292 (_mm256_mask_fcmadd_pch): Likewise
293 (_mm256_mask3_fcmadd_pch): Likewise
294 * config/i386/i386-builtin.def: Add mask3 builtin for complex
295 fma, and adjust mask_builtin to corresponding expander.
296 * config/i386/i386-expand.c (ix86_expand_round_builtin):
297 Skip eraseing embedded rounding for expanders that emits
299 * config/i386/sse.md (complexmove): New mode_attr.
300 (<avx512>_fmaddc_<mode>_mask1<round_expand_name>): New expander.
301 (<avx512>_fcmaddc_<mode>_mask1<round_expand_name>): Likewise.
302 (avx512fp16_fmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
303 (avx512fp16_fcmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
304 (avx512fp16_fcmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
305 (avx512fp16_fmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
306 * config/i386/subst.md (round_embedded_complex): New subst.
308 2021-10-14 John David Anglin <danglin@gcc.gnu.org>
310 * config/pa/pa.md (cbranchsf4): Disable if TARGET_SOFT_FLOAT.
311 (cbranchdf4): Likewise.
312 Add missing move patterns for TARGET_SOFT_FLOAT.
314 2021-10-13 Roger Sayle <roger@nextmovesoftware.com>
316 * config/i386/i386-expand.c (ix86_expand_vector_move): Use a
317 pseudo intermediate when moving a SUBREG into a hard register,
318 by checking ix86_hardreg_mov_ok.
319 (ix86_expand_vector_extract): Store zero-extended SImode
320 intermediate in a pseudo, then set target using a SUBREG_PROMOTED
322 * config/i386/sse.md (mov<VMOVE>_internal): Prevent CSE creating
323 complex (SUBREG) sets of (vector) hard registers before reload, by
324 checking ix86_hardreg_mov_ok.
326 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
328 * ctfc.h (enum ctf_dtu_d_union_enum): Remove redundant comma.
330 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
332 * dwarf2ctf.c (gen_ctf_array_type): Fix typo in comment.
334 2021-10-13 Martin Sebor <msebor@redhat.com>
337 * pointer-query.cc (compute_objsize_r): Handle named address spaces.
339 2021-10-13 Iain Sandoe <iain@sandoe.co.uk>
341 * collect2.c (is_lto_object_file): Release simple-object
342 resources, close files.
344 2021-10-13 Andre Vieira <andre.simoesdiasvieira@arm.com>
346 * config/arm/arm.c (thumb2_legitimate_address_p): Use VALID_MVE_MODE
347 when checking mve addressing modes.
348 (mve_vector_mem_operand): Fix the way we handle pre, post and offset
350 (arm_print_operand): Fix printing of POST_ and PRE_MODIFY.
351 * config/arm/mve.md: Use mve_memory_operand predicate everywhere where
352 there is a single Ux constraint.
354 2021-10-13 John David Anglin <danglin@gcc.gnu.org>
356 * config/pa/pa.md (muldi3): Add support for inlining 64-bit
357 multiplication on 32-bit PA 1.1 and 2.0 targets.
359 2021-10-13 Martin Liska <mliska@suse.cz>
361 PR gcov-profile/90364
362 * coverage.c (build_info): Emit checksum to the global variable.
363 (build_info_type): Add new field for checksum.
364 (coverage_obj_finish): Pass object_checksum.
365 (coverage_init): Use 0 as checksum for .gcno files.
366 * gcov-dump.c (dump_gcov_file): Dump also new checksum field.
367 * gcov.c (read_graph_file): Read also checksum.
368 * doc/invoke.texi: Document the behaviour change.
370 2021-10-13 Richard Biener <rguenther@suse.de>
372 * gimple-iterator.h (gsi_iterator_update): Add GSI_LAST_NEW_STMT,
373 start at integer value 2.
374 * gimple-iterator.c (gsi_insert_seq_nodes_before): Update
375 the iterator for GSI_LAST_NEW_STMT.
376 (gsi_insert_seq_nodes_after): Likewise.
377 * tree-if-conv.c (predicate_statements): Use GSI_LAST_NEW_STMT.
378 * tree-ssa.c (execute_update_addresses_taken): Correct bogus
379 arguments to gsi_replace.
381 2021-10-13 Martin Liska <mliska@suse.cz>
384 * common.opt: Use EnabledBy instead of detection in
385 finish_options and process_options.
386 * opts.c (finish_options): Remove handling of
387 x_flag_unroll_all_loops.
388 * toplev.c (process_options): Likewise for flag_web and
389 flag_rename_registers.
391 2021-10-13 Richard Biener <rguenther@suse.de>
393 PR tree-optimization/102659
394 * tree-if-conv.c (need_to_rewrite_undefined): New flag.
395 (if_convertible_gimple_assign_stmt_p): Mark the loop for
396 rewrite when stmts with undefined behavior on integer
398 (combine_blocks): Predicate also when we need to rewrite stmts.
399 (predicate_statements): Rewrite affected stmts to something
400 with well-defined behavior on overflow.
401 (tree_if_conversion): Initialize need_to_rewrite_undefined.
403 2021-10-13 Richard Biener <rguenther@suse.de>
406 * ipa-sra.c (ptr_parm_has_nonarg_uses): Fix volatileness
409 2021-10-13 Jose E. Marchesi <jose.marchesi@oracle.com>
411 * dwarf2ctf.c: Fix typo in comment.
413 2021-10-12 Bill Schmidt <wschmidt@linux.ibm.com>
416 * config/rs6000/altivec.h (vec_cpsgn): Swap operand order.
417 * config/rs6000/rs6000-overload.def (VEC_COPYSIGN): Use SKIP to
418 avoid generating an automatic #define of vec_cpsgn. Use the
419 correct built-in for V4SFmode that doesn't depend on VSX.
421 2021-10-12 Uroš Bizjak <ubizjak@gmail.com>
425 * config/i386/i386.md (*add<mode>_1_slp): Rewrite as
426 define_insn_and_split pattern. Add alternative 1 and split it
427 post reload to insert operand 1 into the low part of operand 0.
428 (*sub<mode>_1_slp): Ditto.
429 (*and<mode>_1_slp): Ditto.
430 (*<any_or:code><mode>_1_slp): Ditto.
431 (*ashl<mode>3_1_slp): Ditto.
432 (*<any_shiftrt:insn><mode>3_1_slp): Ditto.
433 (*<any_rotate:insn><mode>3_1_slp): Ditto.
434 (*neg<mode>_1_slp): New insn_and_split pattern.
435 (*one_cmpl<mode>_1_slp): Ditto.
437 2021-10-12 David Edelsohn <dje.gcc@gmail.com>
439 * doc/install.texi: Update MinGW and mingw-64 Binaries
442 2021-10-12 Daniel Le Duc Khoi Nguyen <greenrecyclebin@gmail.com>
444 * doc/extend.texi (Common Variable Attributes): Fix typos in
445 alloc_size documentation.
447 2021-10-12 Richard Biener <rguenther@suse.de>
449 PR tree-optimization/102696
450 * tree-vect-slp.c (vect_build_slp_tree_2): Properly mark
451 the tree fatally failed when we reject a BIT_FIELD_REF.
453 2021-10-12 Richard Biener <rguenther@suse.de>
455 PR tree-optimization/102572
456 * tree-vect-stmts.c (vect_build_gather_load_calls): When
457 gathering the vectorized defs for the mask pass in the
458 desired mask vector type so invariants will be handled
461 2021-10-12 Tamar Christina <tamar.christina@arm.com>
463 * config/aarch64/aarch64-sve.md (*fcm<cmp_op><mode>_bic_combine,
464 *fcm<cmp_op><mode>_nor_combine, *fcmuo<mode>_bic_combine,
465 *fcmuo<mode>_nor_combine): New.
467 2021-10-12 Eric Botcazou <ebotcazou@adacore.com>
470 * config/sparc/sparc-modes.def (OI): New integer mode.
472 2021-10-12 Jakub Jelinek <jakub@redhat.com>
474 * gimple-fold.h (clear_padding_type_may_have_padding_p): Declare.
475 * gimple-fold.c (clear_padding_type_may_have_padding_p): No longer
478 2021-10-12 Jakub Jelinek <jakub@redhat.com>
480 * tree-vectorizer.h (loop_cost_model): New function.
481 (unlimited_cost_model): Use it.
482 * tree-vect-loop.c (vect_analyze_loop_costing): Use loop_cost_model
483 call instead of flag_vect_cost_model.
484 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
485 (vect_prune_runtime_alias_test_list): Likewise. Also use it instead
486 of flag_simd_cost_model.
488 2021-10-12 liuhongt <hongtao.liu@intel.com>
491 * config/i386/i386-expand.c (emit_reduc_half): Handle
493 * config/i386/mmx.md (reduc_<code>_scal_v4qi): New expander.
494 (reduc_plus_scal_v4qi): Ditto.
496 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
498 * config/rs6000/smmintrin.h (_mm_cmpeq_epi64, _mm_cmpgt_epi64,
499 _mm_mullo_epi32, _mm_mul_epi32, _mm_packus_epi32): New.
500 * config/rs6000/nmmintrin.h: Copy from i386, tweak to suit.
502 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
504 * config/rs6000/smmintrin.h (_mm_cvtepi8_epi16, _mm_cvtepi8_epi32,
505 _mm_cvtepi8_epi64, _mm_cvtepi16_epi32, _mm_cvtepi16_epi64,
506 _mm_cvtepi32_epi64, _mm_cvtepu8_epi16, _mm_cvtepu8_epi32,
507 _mm_cvtepu8_epi64, _mm_cvtepu16_epi32, _mm_cvtepu16_epi64,
508 _mm_cvtepu32_epi64): New.
510 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
512 * config/rs6000/smmintrin.h (_mm_test_all_zeros,
513 _mm_test_all_ones, _mm_test_mix_ones_zeros): Rewrite as macro.
515 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
517 * config/rs6000/smmintrin.h (_mm_min_epi8, _mm_min_epu16,
518 _mm_min_epi32, _mm_min_epu32, _mm_max_epi8, _mm_max_epu16,
519 _mm_max_epi32, _mm_max_epu32): New.
521 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
523 * ipa-modref-tree.h (struct modref_access_node): Revert
525 (struct modref_ref_node): Likewise.
527 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
529 * ipa-modref-tree.h (modref_tree::global_access_p): New member
532 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
533 ignore_stores_eaf_flags): Move to ipa-modref.h
534 (remove_useless_eaf_flags): Remove early exit on NOCLOBBER.
535 (modref_summary::global_memory_read_p): New member function.
536 (modref_summary::global_memory_written_p): New member function.
537 * ipa-modref.h (modref_summary::global_memory_read_p,
538 modref_summary::global_memory_written_p): Declare.
539 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
540 ignore_stores_eaf_flags): move here.
541 * tree-ssa-structalias.c: Include ipa-modref-tree.h, ipa-modref.h
543 (handle_rhs_call): Rewrite.
544 (handle_call_arg): New function.
545 (determine_global_memory_access): New function.
546 (handle_const_call): Remove
547 (handle_pure_call): Remove
548 (find_func_aliases_for_call): Update use of handle_rhs_call.
549 (compute_points_to_sets): Handle global memory acccesses
552 2021-10-11 Diane Meirowitz <diane.meirowitz@oracle.com>
554 * doc/invoke.texi: Add link to UndefinedBehaviorSanitizer
555 documentation, mention UBSAN_OPTIONS, similar to what is done
556 for AddressSanitizer.
558 2021-10-11 Richard Biener <rguenther@suse.de>
561 * internal-fn.c (expand_DEFERRED_INIT): Check for mode
562 availability before building an integer type for storage
565 2021-10-11 Richard Biener <rguenther@suse.de>
568 * gimple.c (gimple_call_fnspec): Do not mark operator new/delete
571 2021-10-11 Martin Liska <mliska@suse.cz>
573 * common.opt: Remove Init(2) for some options.
574 * toplev.c (process_options): Do not use AUTODETECT_VALUE, but
575 use rather OPTION_SET_P.
577 2021-10-11 Martin Liska <mliska@suse.cz>
579 * common.opt: Remove usage of IRA_REGION_AUTODETECT.
580 * flag-types.h (enum ira_region): Likewise.
581 * toplev.c (process_options): Use OPTION_SET_P instead of
582 IRA_REGION_AUTODETECT.
584 2021-10-11 Jakub Jelinek <jakub@redhat.com>
586 * omp-low.c (omp_runtime_api_call): Handle omp_get_max_teams,
587 omp_[sg]et_teams_thread_limit and omp_set_num_teams.
589 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
591 * config/s390/s390-protos.h (s390_rawmemchr): Add prototype.
592 * config/s390/s390.c (s390_rawmemchr): New function.
593 * config/s390/s390.md (rawmemchr<SINT:mode>): New expander.
594 * config/s390/vector.md (@vec_vfees<mode>): Basically a copy of
595 the pattern vfees<mode> from vx-builtins.md.
596 * config/s390/vx-builtins.md (*vfees<mode>): Remove.
598 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
600 * builtins.c (get_memory_rtx): Change to external linkage.
601 * builtins.h (get_memory_rtx): Add function prototype.
602 * doc/md.texi (rawmemchr<mode>): Document.
603 * internal-fn.c (expand_RAWMEMCHR): Define.
604 * internal-fn.def (RAWMEMCHR): Add.
605 * optabs.def (rawmemchr_optab): Add.
606 * tree-loop-distribution.c (find_single_drs): Change return code
607 behaviour by also returning true if no single store was found
609 (loop_distribution::classify_partition): Respect the new return
610 code behaviour of function find_single_drs.
611 (loop_distribution::execute): Call new function
612 transform_reduction_loop in order to replace rawmemchr or strlen
613 like loops by calls into builtins.
614 (generate_reduction_builtin_1): New function.
615 (generate_rawmemchr_builtin): New function.
616 (generate_strlen_builtin_1): New function.
617 (generate_strlen_builtin): New function.
618 (generate_strlen_builtin_using_rawmemchr): New function.
619 (reduction_var_overflows_first): New function.
620 (determine_reduction_stmt_1): New function.
621 (determine_reduction_stmt): New function.
622 (loop_distribution::transform_reduction_loop): New function.
624 2021-10-11 Martin Liska <mliska@suse.cz>
626 * tree.c (cl_option_hasher::hash): Use cl_optimization_hash
627 and remove legacy hashing code.
629 2021-10-11 Kito Cheng <kito.cheng@sifive.com>
632 * builtins.c (maybe_emit_call_builtin___clear_cache): Allow
633 CONST_INT for BEGIN and END, and use gcc_assert rather than
636 2021-10-10 Jakub Jelinek <jakub@redhat.com>
639 * var-tracking.c (add_stores): For cselib_sp_derived_value_p values
640 use MO_VAL_SET if loc is not sp.
642 2021-10-10 Andrew Pinski <apinski@marvell.com>
644 PR tree-optimization/102622
645 * match.pd: Swap the order of a?pow2cst:0 and a?-1:0 transformations.
646 Swap the order of a?0:pow2cst and a?0:-1 transformations.
648 2021-10-09 liuhongt <hongtao.liu@intel.com>
651 * config/i386/i386-expand.c (ix86_valid_mask_cmp_mode): Handle
653 (ix86_use_mask_cmp_p): Ditto.
654 (ix86_expand_sse_movcc): Ditto.
655 * config/i386/i386.md (setcc_hf_mask): New define_insn.
657 (UNSPEC_MOVCC_MASK): New unspec.
658 * config/i386/sse.md (UNSPEC_PCMP): Move to i386.md.
660 2021-10-08 Vladimir N. Makarov <vmakarov@redhat.com>
662 PR rtl-optimization/102627
663 * lra-constraints.c (split_reg): Use at least natural mode of hard reg.
665 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
667 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Grow
670 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
672 * value-range.cc (irange::debug): New.
673 * value-range.h (irange::debug): New.
675 2021-10-08 Richard Sandiford <richard.sandiford@arm.com>
677 PR tree-optimization/102385
678 * predict.h (change_edge_frequency): Declare.
679 * predict.c (change_edge_frequency): New function.
680 * tree-ssa-loop-manip.h (tree_transform_and_unroll_loop): Remove
682 (tree_unroll_loop): Likewise.
683 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Update accordingly.
684 * tree-predcom.c (pcom_worker::tree_predictive_commoning_loop):
686 * tree-ssa-loop-prefetch.c (loop_prefetch_arrays): Likewise.
687 * tree-ssa-loop-manip.c (tree_unroll_loop): Likewise.
688 (tree_transform_and_unroll_loop): Likewise. Use single_dom_exit
689 to retrieve the exit edges. Make all the old profile update code
690 conditional on !single_loop_p -- the case it was written for --
691 and use a different approach for the single-loop case.
693 2021-10-08 Martin Liska <mliska@suse.cz>
695 * config/alpha/alpha.c (alpha_option_override): Use new macro
697 * config/arc/arc.c (arc_override_options): Likewise.
698 * config/arm/arm.c (arm_option_override): Likewise.
699 * config/bfin/bfin.c (bfin_load_pic_reg): Likewise.
700 * config/c6x/c6x.c (c6x_option_override): Likewise.
701 * config/csky/csky.c: Likewise.
702 * config/darwin.c (darwin_override_options): Likewise.
703 * config/frv/frv.c (frv_option_override): Likewise.
704 * config/i386/djgpp.h: Likewise.
705 * config/i386/i386.c (ix86_stack_protect_guard): Likewise.
706 (ix86_max_noce_ifcvt_seq_cost): Likewise.
707 * config/ia64/ia64.c (ia64_option_override): Likewise.
708 (ia64_override_options_after_change): Likewise.
709 * config/m32c/m32c.c (m32c_option_override): Likewise.
710 * config/m32r/m32r.c (m32r_init): Likewise.
711 * config/m68k/m68k.c (m68k_option_override): Likewise.
712 * config/microblaze/microblaze.c (microblaze_option_override): Likewise.
713 * config/mips/mips.c (mips_option_override): Likewise.
714 * config/nios2/nios2.c (nios2_option_override): Likewise.
715 * config/nvptx/nvptx.c (nvptx_option_override): Likewise.
716 * config/pa/pa.c (pa_option_override): Likewise.
717 * config/riscv/riscv.c (riscv_option_override): Likewise.
718 * config/rs6000/aix71.h: Likewise.
719 * config/rs6000/aix72.h: Likewise.
720 * config/rs6000/aix73.h: Likewise.
721 * config/rs6000/rs6000.c (darwin_rs6000_override_options): Likewise.
722 (rs6000_override_options_after_change): Likewise.
723 (rs6000_linux64_override_options): Likewise.
724 (glibc_supports_ieee_128bit): Likewise.
725 (rs6000_option_override_internal): Likewise.
726 (rs6000_file_start): Likewise.
727 (rs6000_darwin_file_start): Likewise.
728 * config/rs6000/rtems.h: Likewise.
729 * config/rs6000/sysv4.h: Likewise.
730 * config/rs6000/vxworks.h (SUB3TARGET_OVERRIDE_OPTIONS): Likewise.
731 * config/s390/s390.c (s390_option_override): Likewise.
732 * config/sh/linux.h: Likewise.
733 * config/sh/netbsd-elf.h (while): Likewise.
734 * config/sh/sh.c (sh_option_override): Likewise.
735 * config/sol2.c (solaris_override_options): Likewise.
736 * config/sparc/sparc.c (sparc_option_override): Likewise.
737 * config/tilegx/tilegx.c (tilegx_option_override): Likewise.
738 * config/visium/visium.c (visium_option_override): Likewise.
739 * config/vxworks.c (vxworks_override_options): Likewise.
740 * lto-opts.c (lto_write_options): Likewise.
741 * omp-expand.c (expand_omp_simd): Likewise.
742 * omp-general.c (omp_max_vf): Likewise.
743 * omp-offload.c (oacc_xform_loop): Likewise.
744 * opts.h (OPTION_SET_P): Likewise.
745 * targhooks.c (default_max_noce_ifcvt_seq_cost): Likewise.
746 * toplev.c (process_options): Likewise.
747 * tree-predcom.c: Likewise.
748 * tree-sra.c (analyze_all_variable_accesses): Likewise.
750 2021-10-08 liuhongt <hongtao.liu@intel.com>
753 * config/i386/i386.c (ix86_optab_supported_p):
754 Return true for HFmode.
755 * match.pd: Simplify (_Float16) ceil ((double) x) to
756 __builtin_ceilf16 (a) when a is _Float16 type and
757 direct_internal_fn_supported_p.
759 2021-10-08 liuhongt <hongtao.liu@intel.com>
762 * config/i386/i386-expand.c (emit_reduc_half): Hanlde V4HImode.
763 * config/i386/mmx.md (reduc_plus_scal_v4hi): New.
764 (reduc_<code>_scal_v4hi): New.
766 2021-10-08 liuhongt <hongtao.liu@intel.com>
768 * common.opt (ftree-vectorize): Add Var(flag_tree_vectorize).
769 * doc/invoke.texi (Options That Control Optimization): Update
771 * opts.c (default_options_table): Enable auto-vectorization at
772 O2 with very-cheap cost model.
773 (finish_options): Use cheap cost model for
774 explicit -ftree{,-loop}-vectorize.
776 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
778 * ctfc.c (ctfc_delete_container): Free hash table contents.
780 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
782 * toplev.c (process_options): Do not warn for GNU GIMPLE.
784 2021-10-07 Siddhesh Poyarekar <siddhesh@gotplt.org>
786 * tree-object-size.c (addr_object_size,
787 compute_builtin_object_size): Drop PDECL and POFF arguments.
788 (addr_object_size): Adjust calls.
789 * tree-object-size.h (compute_builtin_object_size): Drop PDECL
792 2021-10-07 Roger Sayle <roger@nextmovesoftware.com>
794 * rtl.def (SMUL_HIGHPART, UMUL_HIGHPART): New RTX codes for
795 representing signed and unsigned high-part multiplication resp.
796 * simplify-rtx.c (simplify_binary_operation_1) [SMUL_HIGHPART,
797 UMUL_HIGHPART]: Simplify high-part multiplications by zero.
798 [SS_PLUS, US_PLUS, SS_MINUS, US_MINUS, SS_MULT, US_MULT,
799 SS_DIV, US_DIV]: Similar simplifications for saturating
801 (simplify_const_binary_operation) [SS_PLUS, US_PLUS, SS_MINUS,
802 US_MINUS, SS_MULT, US_MULT, SMUL_HIGHPART, UMUL_HIGHPART]:
803 Implement compile-time evaluation for constant operands.
804 * dwarf2out.c (mem_loc_descriptor): Skip SMUL_HIGHPART and
806 * doc/rtl.texi (smul_highpart, umul_highpart): Document RTX codes.
807 * doc/md.texi (smul@var{m}3_highpart, umul@var{m3}_highpart):
808 Mention the new smul_highpart and umul_highpart RTX codes.
809 * doc/invoke.texi: Silence @xref "compilation" warnings.
811 2021-10-07 Martin Jambor <mjambor@suse.cz>
814 * ipa-prop.c (ipa_edge_args_sum_t::duplicate): Also handle the
815 case when the source reference description corresponds to a
816 referance taken in a function src->caller is inlined to.
818 2021-10-07 Jan Hubicka <hubicka@ucw.cz>
821 * ipa-modref-tree.h (modref_access_node::contains_p): Handle offsets
823 (modref_access_node::try_merge_with): Add sanity check that there
824 are no redundant entries in the list.
826 2021-10-07 Richard Biener <rguenther@suse.de>
828 PR tree-optimization/102608
829 * tree-ssa-sccvn.c (visit_stmt): Drop .DEFERRED_INIT to
832 2021-10-07 Martin Liska <mliska@suse.cz>
834 * toplev.c (toplev::main): Make
835 save_opt_decoded_options a pointer type
836 * toplev.h: Likewise.
838 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
840 * config/gcn/gcn-valu.md (gather<mode>_insn_2offsets<exec>): Apply
841 HAVE_GCN_ASM_GLOBAL_LOAD_FIXED.
842 (scatter<mode>_insn_2offsets<exec_scatter>): Likewise.
844 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
846 * config/gcn/gcn-hsa.h (SRAMOPT): Include the whole option string.
847 Adjust for new -msram-ecc=any behaviour.
848 (ASM_SPEC): Adjust -mxnack and -msram-ecc usage.
849 * config/gcn/gcn.c (output_file_start): Implement -msram-ecc=any.
850 * config/gcn/mkoffload.c (EF_AMDGPU_XNACK): Rename to ...
851 (EF_AMDGPU_XNACK_V3): ... this.
852 (EF_AMDGPU_SRAM_ECC): Rename to ...
853 (EF_AMDGPU_SRAM_ECC_V3): ... this.
854 (EF_AMDGPU_FEATURE_XNACK_V4): New.
855 (EF_AMDGPU_FEATURE_XNACK_UNSUPPORTED_V4): New.
856 (EF_AMDGPU_FEATURE_XNACK_ANY_V4): New.
857 (EF_AMDGPU_FEATURE_XNACK_OFF_V4): New.
858 (EF_AMDGPU_FEATURE_XNACK_ON_V4): New.
859 (EF_AMDGPU_FEATURE_SRAMECC_V4): New.
860 (EF_AMDGPU_FEATURE_SRAMECC_UNSUPPORTED_V4): New.
861 (EF_AMDGPU_FEATURE_SRAMECC_ANY_V4): New.
862 (EF_AMDGPU_FEATURE_SRAMECC_OFF_V4): New.
863 (EF_AMDGPU_FEATURE_SRAMECC_ON_V4): New.
865 (SET_XNACK_OFF): New.
867 (SET_SRAM_ECC_ON): New.
868 (SET_SRAM_ECC_ANY): New.
869 (SET_SRAM_ECC_OFF): New.
870 (TEST_SRAM_ECC_ANY): New.
871 (TEST_SRAM_ECC_ON): New.
872 (main): Implement HSACOv4 and -msram-ecc=any.
874 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
876 * config.in: Regenerate.
877 * config/gcn/gcn-hsa.h (X_FIJI): New macro.
881 (A_FIJI): Rename to ...
883 (A_900): Rename to ...
885 (A_906): Rename to ...
887 (A_908): Rename to ...
889 (SRAMOPT): New macro.
890 (ASM_SPEC): Adjust xnack option usage.
891 * config/gcn/gcn.c (output_file_start): Adjust amdgcn_target usage.
892 * configure: Regenerate.
893 * configure.ac: Detect LLVM assembler dialect.
895 2021-10-07 Richard Biener <rguenther@suse.de>
897 * tree-pretty-print.c (dump_generic_node): Do not elide
898 printing '&' when dumping with -gimple.
900 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
902 * gimple-range-cache.cc (non_null_ref::adjust_range): Call new
904 * gimple-range-fold.cc (adjust_pointer_diff_expr): Ditto.
905 (adjust_imagpart_expr): Ditto.
906 * value-range.cc (irange::irange_intersect): Call new routine if
907 RHS is a single pair.
908 (irange::intersect): New wide_int version.
909 * value-range.h (class irange): New prototype.
911 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
913 * gimple-range-edge.cc (gimple_outgoing_range::gimple_outgoing_range):
914 Add parameter to limit size when recognizing switches.
915 (gimple_outgoing_range::edge_range_p): Check size limit.
916 * gimple-range-edge.h (gimple_outgoing_range): Add size field.
917 * gimple-range-gori.cc (gori_map::calculate_gori): Ignore switches
918 that exceed the size limit.
919 (gori_compute::gori_compute): Add initializer.
920 * params.opt (evrp-switch-limit): New.
921 * doc/invoke.texi: Update docs.
923 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
925 * value-range.h (irange::set_varying): Use TYPE_MIN_VALUE and
926 TYPE_MAX_VALUE instead of creating new trees when possible.
928 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
930 * gimple-range-cache.cc (non_null_ref::adjust_range): Check for
931 zero and non-zero more efficently.
933 2021-10-06 Richard Biener <rguenther@suse.de>
936 * dumpfile.h (TDF_GIMPLE_VAL): New.
937 (dump_flag): Re-order and adjust TDF_* flags. Make
938 the enum uint32_t. Use std::underlying_type in the
940 (optgroup_flag): Likewise for the operator overloads.
941 * tree-pretty-print.c (dump_generic_node): Wrap ADDR_EXPR
942 in _Literal if TDF_GIMPLE_VAL.
943 * gimple-pretty-print.c (dump_gimple_assign): Add
944 TDF_GIMPLE_VAL to flags when dumping operands where only
945 is_gimple_val are allowed.
946 (dump_gimple_cond): Likewise.
948 2021-10-06 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
950 * gimple-isel.cc (gimple_expand_vec_cond_expr): Remove redundant if
953 2021-10-05 qing zhao <qing.zhao@oracle.com>
956 * gimplify.c (gimplify_decl_expr): Not add initialization for an
957 auto variable when it has been initialized by frontend.
959 2021-10-05 Aldy Hernandez <aldyh@redhat.com>
961 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
964 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
966 * common/config/avr/avr-common.c (avr_handle_option): Mark
967 argument as ATTRIBUTE_UNUSED.
969 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
971 * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC):
972 Undefine before redefinition.
974 2021-10-05 Richard Biener <rguenther@suse.de>
976 * toplev.c (no_backend): Remove global var.
977 (process_options): Pass in no_backend, move post_options
978 langhook call to toplev::main.
979 (do_compile): Pass in no_backend, move process_options call
981 (toplev::run_self_tests): Check no_backend at the caller.
982 (toplev::main): Call post_options and process_options
983 split out from do_compile, do self-tests only if
984 no_backend is initialized.
986 2021-10-05 Richard Biener <rguenther@suse.de>
988 * tree-cfg.c (dump_function_to_file): Dump the UID of the
989 function as part of the name when requested.
990 * tree-pretty-print.c (dump_function_name): Dump the UID when
991 requested and the langhook produced the actual name.
993 2021-10-05 Richard Biener <rguenther@suse.de>
997 * internal-fn.c (expand_DEFERRED_INIT): Fall back to
998 zero-initialization as last resort, use the constant
999 size as given by the DEFERRED_INIT argument to build
1002 2021-10-04 Marek Polacek <polacek@redhat.com>
1005 * doc/invoke.texi: Document -Warray-compare.
1007 2021-10-04 Richard Biener <rguenther@suse.de>
1009 * gimplify.c (is_var_need_auto_init): DECL_HARD_REGISTER
1010 variables are not to be initialized.
1012 2021-10-04 Richard Biener <rguenther@suse.de>
1014 * expr.h (non_mem_decl_p): Declare.
1015 (mem_ref_refers_to_non_mem_p): Likewise.
1016 * expr.c (non_mem_decl_p): Export.
1017 (mem_ref_refers_to_non_mem_p): Likewise.
1018 * internal-fn.c (expand_DEFERRED_INIT): Do not expand the LHS
1019 but check the base with mem_ref_refers_to_non_mem_p
1022 2021-10-04 Richard Biener <rguenther@suse.de>
1024 PR tree-optimization/102570
1025 * tree-ssa-sccvn.h (vn_reference_op_struct): Document
1026 we are using clique for the internal function code.
1027 * tree-ssa-sccvn.c (vn_reference_op_eq): Compare the
1028 internal function code.
1029 (print_vn_reference_ops): Print the internal function code.
1030 (vn_reference_op_compute_hash): Hash it.
1031 (copy_reference_ops_from_call): Record it.
1032 (visit_stmt): Remove the restriction around internal function
1034 (fully_constant_vn_reference_p): Use fold_const_call and handle
1036 (vn_reference_eq): Compare call return types.
1037 * tree-ssa-pre.c (create_expression_by_pieces): Handle
1038 generating calls to internal functions.
1039 (compute_avail): Remove the restriction around internal function
1042 2021-10-04 Aldy Hernandez <aldyh@redhat.com>
1044 PR tree-optimization/102560
1045 * gimple-ssa-warn-alloca.c (alloca_call_type): Remove static
1046 marker for invalid_range.
1048 2021-10-04 Richard Biener <rguenther@suse.de>
1050 PR middle-end/102587
1051 * internal-fn.c (expand_DEFERRED_INIT): Guard register
1052 initialization path an avoid initializing VLA registers
1055 2021-10-04 Eric Botcazou <ebotcazou@adacore.com>
1057 * config/rs6000/vxworks.h (TARGET_INIT_LIBFUNCS): Delete.
1059 2021-10-03 Martin Liska <mliska@suse.cz>
1061 * toplev.c (toplev::main): Check opt_index if it is a part
1064 2021-10-02 Aldy Hernandez <aldyh@redhat.com>
1066 PR tree-optimization/102563
1067 * range-op.cc (operator_lshift::op1_range): Do not clobber
1070 2021-10-02 Martin Liska <mliska@suse.cz>
1072 * toplev.c (toplev::main): save_decoded_options[0] is program
1073 name and so it should be skipped.
1075 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
1077 PR tree-optimization/102546
1078 * range-op.cc (operator_lshift::op1_range): Teach range-ops that
1079 X << Y is non-zero implies X is also non-zero.
1081 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1083 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
1085 * config/aarch64/aarch64-tune.md: Regenerate.
1086 * doc/invoke.texi: Update docs.
1088 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1090 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
1092 * config/aarch64/aarch64-tune.md: Regenerate.
1093 * doc/invoke.texi: Update docs.
1095 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1097 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
1099 * config/aarch64/aarch64-tune.md: Regenerate.
1100 * doc/invoke.texi: Update docs.
1102 2021-10-01 Martin Sebor <msebor@redhat.com>
1105 * doc/invoke.texi (-Waddress): Update.
1106 * gengtype.c (write_types): Avoid -Waddress.
1107 * poly-int.h (POLY_SET_COEFF): Avoid using null.
1109 2021-10-01 John David Anglin <danglin@gcc.gnu.org>
1112 * config/pa/pa.c (pa_option_override): Default to dwarf version 4
1115 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1117 * config/aarch64/aarch64.h (AARCH64_FL_V9): Update value.
1119 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
1121 * gimple-range-path.cc (path_range_query::compute_ranges): Use
1123 * gimple-range-path.h (class path_range_query): Remove shadowed
1125 (path_range_query::get_path_oracle): New.
1127 2021-10-01 Jakub Jelinek <jakub@redhat.com>
1128 Richard Biener <rguenther@suse.de>
1131 * doc/invoke.texi (-fsanitize=integer-divide-by-zero): Remove
1132 INT_MIN / -1 division detection from here ...
1133 (-fsanitize=signed-integer-overflow): ... and add it here.
1135 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1137 * config/aarch64/aarch64-arches.def (AARCH64_ARCH): Added
1139 * config/aarch64/aarch64.h (AARCH64_FL_V9): New.
1140 (AARCH64_FL_FOR_ARCH9): New flags for Armv9-A.
1141 (AARCH64_ISA_V9): New ISA flag.
1142 * doc/invoke.texi: Update docs.
1144 2021-10-01 Martin Liska <mliska@suse.cz>
1146 * toplev.c (toplev::main): Save decoded optimization options.
1147 * toplev.h (save_opt_decoded_options): New.
1148 * doc/extend.texi: Be more clear about optimize and target
1151 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
1153 * explow.c: Include langhooks.h.
1154 (set_stack_check_libfunc): Build a proper function type.
1156 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
1159 * config/i386/i386.c (legitimate_pic_address_disp_p): For PE-COFF do
1160 not return true for external weak function symbols in medium model.
1162 2021-10-01 Jakub Jelinek <jakub@redhat.com>
1164 * tree.h (OMP_CLAUSE_ORDER_REPRODUCIBLE): Define.
1165 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_ORDER>: Print
1166 reproducible: for OMP_CLAUSE_ORDER_REPRODUCIBLE.
1167 * omp-general.c (omp_extract_for_data): If OMP_CLAUSE_ORDER is seen
1168 without OMP_CLAUSE_ORDER_UNCONSTRAINED, overwrite sched_kind to
1169 OMP_CLAUSE_SCHEDULE_STATIC.
1171 2021-10-01 Richard Biener <rguenther@suse.de>
1173 PR middle-end/102518
1174 * tree-inline.c (setup_one_parameter): Avoid substituting
1175 an invariant into contexts where a GIMPLE register is not valid.
1177 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1179 * config/arm/arm-cpus.in: Add Cortex-R52+ CPU.
1180 * config/arm/arm-tables.opt: Regenerate.
1181 * config/arm/arm-tune.md: Regenerate.
1182 * doc/invoke.texi: Update docs.
1184 2021-09-30 Uroš Bizjak <ubizjak@gmail.com>
1187 * config/i386/i386.md
1188 (sign_extend:WIDE (any_logic:NARROW (memory, immediate)) splitters):
1191 2021-09-30 Tobias Burnus <tobias@codesourcery.com>
1193 * omp-low.c (omp_runtime_api_call): Add omp_aligned_{,c}alloc and
1194 omp_{c,re}alloc, fix omp_alloc/omp_free.
1196 2021-09-30 Martin Liska <mliska@suse.cz>
1198 * defaults.h (ASM_OUTPUT_ASCII): Do not hide global variable
1199 asm_out_file and stream directly to MYFILE.
1201 2021-09-30 Richard Biener <rguenther@suse.de>
1203 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
1204 Restore and fix condition under which we apply npeel to
1205 the DRs misalignment value.
1207 2021-09-30 Richard Biener <rguenther@suse.de>
1209 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
1210 Fix npeel check for variable amount of peeling.
1212 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
1214 * lto-wrapper.c (run_gcc): Plug snprintf overflow.
1216 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
1218 * gimple-range.cc (gimple_ranger::debug): New.
1219 * gimple-range.h (class gimple_ranger): Add debug.
1221 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
1223 PR middle-end/102519
1224 * tree-vrp.c (hybrid_threader::~hybrid_threader): Free m_query.
1226 2021-09-29 Indu Bhagat <indu.bhagat@oracle.com>
1229 * btfout.c (GTY): Add GTY (()) albeit for cosmetic only purpose.
1230 (btf_finalize): Empty the hash_map btf_var_ids.
1232 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
1234 * tree-vrp.c (thread_through_all_blocks): Return bool.
1235 (execute_vrp_threader): Return TODO_* flags.
1236 (pass_data_vrp_threader): Set todo_flags_finish to 0.
1238 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
1240 * timevar.def (TV_TREE_VRP_THREADER): New.
1241 * tree-vrp.c: Use TV_TREE_VRP_THREADER for VRP threader pass.
1243 2021-09-29 David Faust <david.faust@oracle.com>
1245 * config.gcc (bpf-*-*): Do not overwrite extra_headers.
1247 2021-09-29 Jonathan Wright <jonathan.wright@arm.com>
1249 * config/aarch64/aarch64-builtins.c (TYPES_BINOP_PPU): Define
1250 new type qualifier enum.
1251 (TYPES_TERNOP_SSSU): Likewise.
1252 (TYPES_TERNOP_PPPU): Likewise.
1253 * config/aarch64/aarch64-simd-builtins.def: Define PPU, SSU,
1254 PPPU and SSSU builtin generator macros for qtbl1 and qtbx1
1256 * config/aarch64/arm_neon.h (vqtbl1_p8): Use type-qualified
1257 builtin and remove casts.
1258 (vqtbl1_s8): Likewise.
1259 (vqtbl1q_p8): Likewise.
1260 (vqtbl1q_s8): Likewise.
1261 (vqtbx1_s8): Likewise.
1262 (vqtbx1_p8): Likewise.
1263 (vqtbx1q_s8): Likewise.
1264 (vqtbx1q_p8): Likewise.
1265 (vtbl1_p8): Likewise.
1266 (vtbl2_p8): Likewise.
1267 (vtbx2_p8): Likewise.
1269 2021-09-29 Richard Biener <rguenther@suse.de>
1271 * tree-vect-data-refs.c (vect_dr_misalign_for_aligned_access):
1273 (vect_update_misalignment_for_peel): Use it to update
1274 misaligned to the value necessary for an aligned access.
1275 (vect_get_peeling_costs_all_drs): Likewise.
1276 (vect_enhance_data_refs_alignment): Likewise.
1278 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1280 * config/aarch64/aarch64.c (aarch64_expand_cpymem): Count number of
1281 emitted operations and adjust heuristic for code size.
1283 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1285 * config/aarch64/aarch64.c (aarch64_expand_setmem): Count number of
1286 emitted operations and adjust heuristic for code size.
1288 2021-09-29 Jakub Jelinek <jakub@redhat.com>
1290 PR middle-end/102504
1291 * gimplify.c (gimplify_scan_omp_clauses): Use omp_check_private even
1292 in OMP_SCOPE clauses, not just on worksharing construct clauses.
1294 2021-09-28 Geng Qi <gengqi@linux.alibaba.com>
1296 * config/riscv/riscv.md (mulv<mode>4): Call gen_smul<mode>3_highpart.
1297 (<u>mulditi3): Call <su>muldi3_highpart.
1298 (<u>muldi3_highpart): Rename to <su>muldi3_highpart.
1299 (<u>mulsidi3): Call <su>mulsi3_highpart.
1300 (<u>mulsi3_highpart): Rename to <su>mulsi3_highpart.
1302 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
1304 * config/darwin.h (DSYMUTIL_SPEC): Recognize D sources.
1306 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
1308 * config/rs6000/darwin.h (FIXED_R13): Add for PPC64.
1309 (FIRST_SAVED_GP_REGNO): Save from R13 even when it is one
1312 2021-09-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1314 * config/aarch64/aarch64.h (AARCH64_FL_LS64): Define
1315 (AARCH64_FL_V8_7): Likewise.
1316 (AARCH64_FL_FOR_ARCH8_7): Likewise.
1317 * config/aarch64/aarch64-arches.def (armv8.7-a): Define.
1318 * config/aarch64/aarch64-option-extensions.def (ls64): Define.
1319 * doc/invoke.texi: Document the above.
1321 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
1323 * dbgcnt.c (dbg_cnt_counter): New.
1324 * dbgcnt.h (dbg_cnt_counter): New.
1325 * dumpfile.c (dump_options): Add entry for TDF_THREADING.
1326 * dumpfile.h (enum dump_flag): Add TDF_THREADING.
1327 * gimple-range-path.cc (DEBUG_SOLVER): Use TDF_THREADING.
1328 * tree-ssa-threadupdate.c (dump_jump_thread_path): Dump out
1331 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
1333 * cfgcleanup.c (pass_jump::execute): Check
1334 flag_expensive_optimizations.
1335 (pass_jump_after_combine::gate): Same.
1336 * doc/invoke.texi (-fthread-jumps): Enable for -O1.
1337 * opts.c (default_options_table): Enable -fthread-jumps at -O1.
1338 * tree-ssa-threadupdate.c
1339 (fwd_jt_path_registry::remove_jump_threads_including): Bail unless
1342 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
1344 * tree-ssa-reassoc.c (biased_names): New global.
1345 (propagate_bias_p): New function.
1346 (loop_carried_phi): Remove.
1347 (propagate_rank): Propagate bias along single uses.
1348 (get_rank): Update biased_names when needed.
1350 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
1352 * passes.def (pass_reassoc): Rename parameter to early_p.
1353 * tree-ssa-reassoc.c (reassoc_bias_loop_carried_phi_ranks_p):
1355 (phi_rank): Don't bias loop-carried phi ranks
1356 before vectorization pass.
1357 (execute_reassoc): Add bias_loop_carried_phi_ranks_p parameter.
1358 (pass_reassoc::pass_reassoc): Add bias_loop_carried_phi_ranks_p
1360 (pass_reassoc::set_param): Set bias_loop_carried_phi_ranks_p
1362 (pass_reassoc::execute): Pass bias_loop_carried_phi_ranks_p to
1364 (pass_reassoc::bias_loop_carried_phi_ranks_p): New member.
1366 2021-09-28 Jakub Jelinek <jakub@redhat.com>
1369 * config/i386/i386.c (standard_80387_constant_p): Don't recognize
1370 special 80387 instruction XFmode constants if flag_rounding_math.
1372 2021-09-28 Richard Biener <rguenther@suse.de>
1374 PR tree-optimization/100112
1375 * tree-ssa-sccvn.c (visit_reference_op_load): Record the
1376 referece into the hashtable twice in case last_vuse is
1377 different from the original vuse on the stmt.
1379 2021-09-28 Jakub Jelinek <jakub@redhat.com>
1381 PR middle-end/102492
1382 * gimplify.c (gimplify_adjust_omp_clauses_1): Don't call the
1383 omp_finish_clause langhook on implicitly added OMP_CLAUSE_PRIVATE
1384 clauses on SIMD constructs.
1386 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
1388 PR tree-optimization/102511
1389 * gimple-range-path.cc (path_range_query::range_on_path_entry):
1390 Return VARYING when nothing found.
1392 2021-09-28 Hongyu Wang <hongyu.wang@intel.com>
1395 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): Add
1397 (VALID_SSE2_REG_VHF_MODE): Add V4HFmode and V2HFmode.
1398 (VALID_MMX_REG_MODE): Add V4HFmode.
1399 (SSE_REG_MODE_P): Replace VALID_AVX512FP16_REG_MODE with
1400 vector mode condition.
1401 * config/i386/i386.c (classify_argument): Parse V4HF/V2HF
1403 (function_arg_32): Add V4HFmode.
1404 (function_arg_advance_32): Likewise.
1405 * config/i386/i386.md (mode): Add V4HF/V2HF.
1406 (MODE_SIZE): Likewise.
1407 * config/i386/mmx.md (MMXMODE): Add V4HF mode.
1408 (V_32): Add V2HF mode.
1409 (VHF_32_64): New mode iterator.
1410 (*mov<mode>_internal): Adjust sse alternatives to support
1412 (*mov<mode>_internal): Adjust sse alternatives to support
1414 (<insn><mode>3): New define_insn for add/sub/mul/div.
1416 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
1418 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Check
1420 (pass_early_thread_jumps::gate): Same.
1421 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
1422 Return if !flag_thread_jumps.
1423 * tree-ssa-threadupdate.c
1424 (jt_path_registry::register_jump_thread): Assert that
1425 flag_thread_jumps is true.
1427 2021-09-28 liuhongt <hongtao.liu@intel.com>
1430 (simplify_context::simplify_binary_operation_1): Relax
1431 condition of simplifying (vec_concat:M (vec_select op0
1432 index0)(vec_select op1 index1)) to allow different modes
1433 between op0 and M, but have same inner mode.
1435 2021-09-28 liuhongt <hongtao.liu@intel.com>
1437 * config/i386/i386-expand.c (emit_reduc_half): Handle
1438 V8HF/V16HF/V32HFmode.
1439 * config/i386/sse.md (REDUC_SSE_PLUS_MODE): Add V8HF.
1440 (REDUC_SSE_SMINMAX_MODE): Ditto.
1441 (REDUC_PLUS_MODE): Add V16HF and V32HF.
1442 (REDUC_SMINMAX_MODE): Ditto.
1444 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
1446 * gimple-range-path.cc
1447 (path_range_query::precompute_ranges_in_block): Rename to...
1448 (path_range_query::compute_ranges_in_block): ...this.
1449 (path_range_query::precompute_ranges): Rename to...
1450 (path_range_query::compute_ranges): ...this.
1451 (path_range_query::precompute_relations): Rename to...
1452 (path_range_query::compute_relations): ...this.
1453 (path_range_query::precompute_phi_relations): Rename to...
1454 (path_range_query::compute_phi_relations): ...this.
1455 * gimple-range-path.h: Rename precompute* to compute*.
1456 * tree-ssa-threadbackward.c
1457 (back_threader::find_taken_edge_switch): Same.
1458 (back_threader::find_taken_edge_cond): Same.
1459 * tree-ssa-threadedge.c
1460 (hybrid_jt_simplifier::compute_ranges_from_state): Same.
1461 (hybrid_jt_state::register_equivs_stmt): Inline...
1462 * tree-ssa-threadedge.h: ...here.
1464 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
1466 * tree-vrp.c (lhs_of_dominating_assert): Remove.
1467 (class vrp_jt_state): Remove.
1468 (class vrp_jt_simplifier): Remove.
1469 (vrp_jt_simplifier::simplify): Remove.
1470 (class vrp_jump_threader): Remove.
1471 (vrp_jump_threader::vrp_jump_threader): Remove.
1472 (vrp_jump_threader::~vrp_jump_threader): Remove.
1473 (vrp_jump_threader::before_dom_children): Remove.
1474 (vrp_jump_threader::after_dom_children): Remove.
1476 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
1478 * passes.def (pass_vrp_threader): New.
1479 * tree-pass.h (make_pass_vrp_threader): Add make_pass_vrp_threader.
1480 * tree-ssa-threadedge.c (hybrid_jt_state::register_equivs_stmt): New.
1481 (hybrid_jt_simplifier::hybrid_jt_simplifier): New.
1482 (hybrid_jt_simplifier::simplify): New.
1483 (hybrid_jt_simplifier::compute_ranges_from_state): New.
1484 * tree-ssa-threadedge.h (class hybrid_jt_state): New.
1485 (class hybrid_jt_simplifier): New.
1486 * tree-vrp.c (execute_vrp): Remove ASSERT_EXPR based jump
1488 (class hybrid_threader): New.
1489 (hybrid_threader::hybrid_threader): New.
1490 (hybrid_threader::~hybrid_threader): New.
1491 (hybrid_threader::before_dom_children): New.
1492 (hybrid_threader::after_dom_children): New.
1493 (execute_vrp_threader): New.
1494 (class pass_vrp_threader): New.
1495 (make_pass_vrp_threader): New.
1497 2021-09-27 Martin Liska <mliska@suse.cz>
1499 * output.h (enum section_flag): New.
1500 (SECTION_FORGET): Remove.
1501 (SECTION_ENTSIZE): Make it (1UL << 8) - 1.
1502 (SECTION_STYLE_MASK): Define it based on other enum
1504 * varasm.c (switch_to_section): Remove unused handling of
1507 2021-09-27 Martin Liska <mliska@suse.cz>
1509 * common.opt: Add new variable flag_default_complex_method.
1510 * opts.c (finish_options): Handle flags related to
1511 x_flag_complex_method.
1512 * toplev.c (process_options): Remove option handling related
1513 to flag_complex_method.
1515 2021-09-27 Richard Biener <rguenther@suse.de>
1517 PR middle-end/102450
1518 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid using
1519 type_for_size, instead use int_mode_for_size.
1521 2021-09-27 Andrew Pinski <apinski@marvell.com>
1524 * gimplify.c (gimplify_save_expr): Return early
1525 if the type of val is error_mark_node.
1527 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
1529 * tree-ssanames.c (ssa_name_has_boolean_range): Use
1532 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
1534 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
1535 vrp_visit_cond_stmt.
1536 * tree-ssa-dom.c (cprop_operand): Convert to range_query API.
1537 (cprop_into_stmt): Same.
1538 (dom_opt_dom_walker::optimize_stmt): Same.
1540 2021-09-27 Richard Biener <rguenther@suse.de>
1542 PR tree-optimization/97351
1543 PR tree-optimization/97352
1544 PR tree-optimization/82426
1545 * tree-vectorizer.h (dr_misalignment): Add vector type
1547 (aligned_access_p): Likewise.
1548 (known_alignment_for_access_p): Likewise.
1549 (vect_supportable_dr_alignment): Likewise.
1550 (vect_known_alignment_in_bytes): Likewise. Refactor.
1551 (DR_MISALIGNMENT): Remove.
1552 (vect_update_shared_vectype): Likewise.
1553 * tree-vect-data-refs.c (dr_misalignment): Refactor, handle
1554 a vector type with larger alignment requirement and apply
1555 the negative step adjustment here.
1556 (vect_calculate_target_alignment): Remove.
1557 (vect_compute_data_ref_alignment): Get explicit vector type
1558 argument, do not apply a negative step alignment adjustment
1560 (vect_slp_analyze_node_alignment): Re-analyze alignment
1561 when we re-visit the DR with a bigger desired alignment but
1562 keep more precise results from smaller alignments.
1563 * tree-vect-slp.c (vect_update_shared_vectype): Remove.
1564 (vect_slp_analyze_node_operations_1): Do not update the
1565 shared vector type on stmts.
1566 * tree-vect-stmts.c (vect_analyze_stmt): Push/pop the
1567 vector type of an SLP node to the representative stmt-info.
1568 (vect_transform_stmt): Likewise.
1570 2021-09-27 liuhongt <hongtao.liu@intel.com>
1573 2021-09-09 liuhongt <hongtao.liu@intel.com>
1576 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
1577 (reduc_plus_scal_v4sf): .. this, New define_expand.
1578 (reduc_plus_scal_v2df): .. and this, New define_expand.
1580 2021-09-26 liuhongt <hongtao.liu@intel.com>
1582 * doc/extend.texi (Half-Precision): Remove storage only
1583 description for _Float16 w/o avx512fp16.
1585 2021-09-25 Dimitar Dimitrov <dimitar@dinux.eu>
1587 * config/pru/constraints.md (Rrio): New constraint.
1588 * config/pru/predicates.md (regio_operand): New predicate.
1589 * config/pru/pru-pragma.c (pru_register_pragmas): Register
1590 the __regio_symbol address space.
1591 * config/pru/pru-protos.h (pru_symref2ioregno): Declaration.
1592 * config/pru/pru.c (pru_symref2ioregno): New helper function.
1593 (pru_legitimate_address_p): Remove.
1594 (pru_addr_space_legitimate_address_p): Use the address space
1596 (pru_nongeneric_pointer_addrspace): New helper function.
1597 (pru_insert_attributes): New function to validate __regio_symbol
1599 (TARGET_INSERT_ATTRIBUTES): New macro.
1600 (TARGET_LEGITIMATE_ADDRESS_P): Remove.
1601 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): New macro.
1602 * config/pru/pru.h (enum reg_class): Add REGIO_REGS class.
1603 * config/pru/pru.md (*regio_readsi): New pattern to read I/O
1605 (*regio_nozext_writesi): New pattern to write to I/O registers.
1606 (*regio_zext_write_r30<EQS0:mode>): Ditto.
1607 * doc/extend.texi: Document the new PRU Named Address Space.
1609 2021-09-24 Patrick Palka <ppalka@redhat.com>
1613 * real.c (encode_ieee_double): Avoid unwanted sign extension.
1614 (encode_ieee_quad): Likewise.
1616 2021-09-24 Vladimir Makarov <vmakarov@redhat.com>
1618 PR rtl-optimization/102147
1619 * ira-build.c (ira_conflict_vector_profitable_p): Make
1620 profitability calculation independent of host compiler pointer and
1623 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
1625 * gimple-range-path.cc (path_range_query::path_range_query):
1626 Move debugging header...
1627 (path_range_query::precompute_ranges): ...here.
1628 (path_range_query::internal_range_of_expr): Do not call
1629 range_on_path_entry if NAME is defined in the current block.
1631 2021-09-24 Richard Biener <rguenther@suse.de>
1633 * cfghooks.c (verify_flow_info): Verify unallocated BB and
1634 edge flags are not set.
1636 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
1638 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
1640 (jt_path_registry::register_jump_thread): Call
1641 cancel_invalid_paths.
1642 * tree-ssa-threadupdate.h (class jt_path_registry): Add
1643 cancel_invalid_paths.
1645 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
1647 PR tree-optimization/102400
1648 * tree-ssa-sccvn.c (vn_reference_insert_pieces): Initialize
1649 result_vdef to zero value.
1651 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
1653 PR tree-optimization/102451
1654 * tree-ssa-dse.c (delete_dead_or_redundant_call): Record bb of stmt
1657 2021-09-24 Hongyu Wang <hongyu.wang@intel.com>
1659 * config/i386/sse.md (cond_<insn><mode>): Extend to support
1661 (cond_mul<mode>): Likewise.
1662 (cond_div<mode>): Likewise.
1663 (cond_<code><mode>): Likewise.
1664 (cond_fma<mode>): Likewise.
1665 (cond_fms<mode>): Likewise.
1666 (cond_fnma<mode>): Likewise.
1667 (cond_fnms<mode>): Likewise.
1669 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
1671 PR tree-optimization/102463
1672 * gimple-range-fold.cc (fold_using_range::relation_fold_and_or): If
1673 there is no range-ops handler, don't look for a relation.
1675 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
1677 * gimple-range-cache.cc (ranger_cache::ranger_cache): Take
1678 non-executable_edge flag as parameter.
1679 * gimple-range-cache.h (ranger_cache): Adjust prototype.
1680 * gimple-range-gori.cc (gori_compute::gori_compute): Take
1681 non-executable_edge flag as parameter.
1682 (gori_compute::outgoing_edge_range_p): Check new flag.
1683 * gimple-range-gori.h (gori_compute): Adjust prototype.
1684 * gimple-range.cc (gimple_ranger::gimple_ranger): Create new flag.
1685 (gimple_ranger::range_on_edge): Check new flag.
1686 * gimple-range.h (gimple_ranger::non_executable_edge_flag): New.
1687 * gimple-ssa-evrp.c (rvrp_folder): Pass ranger flag to simplifer.
1688 (hybrid_folder::hybrid_folder): Set ranger non-executable flag value.
1689 (hybrid_folder::fold_stmt): Set flag value in the simplifer.
1690 * vr-values.c (simplify_using_ranges::set_and_propagate_unexecutable):
1691 Use not_executable flag if provided inmstead of EDGE_EXECUTABLE.
1692 (simplify_using_ranges::simplify_switch_using_ranges): Clear
1693 EDGE_EXECUTABLE like it originally did.
1694 (simplify_using_ranges::cleanup_edges_and_switches): Clear any
1695 NON_EXECUTABLE flags.
1696 (simplify_using_ranges::simplify_using_ranges): Adjust.
1697 * vr-values.h (class simplify_using_ranges): Adjust.
1698 (simplify_using_ranges::set_range_query): Add non-executable flag param.
1700 2021-09-23 Bill Schmidt <wschmidt@linux.ibm.com>
1703 * config/rs6000/rs6000-call.c (rs6000_aggregate_candidate): Detect
1704 zero-width bit fields and return indicator.
1705 (rs6000_discover_homogeneous_aggregate): Diagnose when the
1706 presence of a zero-width bit field changes parameter passing in
1709 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
1711 * gimple-range-fold.cc (fold_using_range::range_of_phi):
1712 Remove dominator check.
1714 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
1716 * gimple-range-path.cc (path_range_query::precompute_relations):
1717 Hoist edge calculations before using EDGE_SUCC.
1719 2021-09-23 Jonathan Wakely <jwakely@redhat.com>
1721 * configure.ac: Fix --with-multilib-list description.
1722 * configure: Regenerate.
1724 2021-09-23 Richard Biener <rguenther@suse.de>
1726 PR tree-optimization/102448
1727 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
1728 Clear alignment info copied from DR_PTR_INFO.
1730 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
1732 * config/i386/i386-expand.c (ix86_use_mask_cmp_p): Enable
1734 * config/i386/sse.md (sseintvecmodelower): Add HF vector modes.
1735 (<avx512>_store<mode>_mask): Extend to support HF vector modes.
1736 (vec_cmp<mode><avx512fmaskmodelower>): Likewise.
1737 (vcond_mask_<mode><avx512fmaskmodelower>): Likewise.
1738 (vcond<mode><mode>): New expander.
1739 (vcond<mode><sseintvecmodelower>): Likewise.
1740 (vcond<sseintvecmodelower><mode>): Likewise.
1741 (vcondu<mode><sseintvecmodelower>): Likewise.
1743 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
1745 * config/i386/sse.md (extend<ssePHmodelower><mode>2):
1747 (extendv4hf<mode>2): Likewise.
1748 (extendv2hfv2df2): Likewise.
1749 (trunc<mode><ssePHmodelower>2): Likewise.
1750 (avx512fp16_vcvt<castmode>2ph_<mode>): Rename to ...
1751 (trunc<mode>v4hf2): ... this, and drop constraints.
1752 (avx512fp16_vcvtpd2ph_v2df): Rename to ...
1753 (truncv2dfv2hf2): ... this, and likewise.
1755 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
1757 * config/i386/sse.md (float<floatunssuffix><mode><ssePHmodelower>2):
1759 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>):
1761 (float<floatunssuffix><mode>v4hf2): ... this, and drop constraints.
1762 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Rename to ...
1763 (float<floatunssuffix>v2div2hf2): ... this, and likewise.
1765 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
1767 * config/i386/i386.md (fix<fixunssuffix>_trunchf<mode>2): New expander.
1768 (fixuns_trunchfhi2): Likewise.
1769 (*fixuns_trunchfsi2zext): New define_insn.
1770 * config/i386/sse.md (ssePHmodelower): New mode_attr.
1771 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
1772 New expander for same element vector fix_truncate.
1773 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
1774 Likewise for V4HF to V4SI/V4DI fix_truncate.
1775 (fix<fixunssuffix>_truncv2hfv2di2):
1776 Likeise for V2HF to V2DI fix_truncate.
1778 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
1780 * config/i386/i386.md (<code>hf3): New expander.
1782 2021-09-23 liuhongt <hongtao.liu@intel.com>
1784 * config/i386/sse.md (FMAMODEM): extend to handle FP16.
1785 (VFH_SF_AVX512VL): Extend to handle HFmode.
1786 (VF_SF_AVX512VL): Deleted.
1788 2021-09-23 liuhongt <hongtao.liu@intel.com>
1790 * config/i386/i386.md (rinthf2): New expander.
1791 (nearbyinthf2): New expander.
1793 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
1795 * tree-ssa-dom.c (class dom_jump_threader_simplifier): Rename...
1796 (class dom_jt_state): ...this and provide virtual overrides.
1797 (dom_jt_state::register_equiv): New.
1798 (class dom_jt_simplifier): Rename from
1799 dom_jump_threader_simplifier.
1800 (dom_jump_threader_simplifier::simplify): Rename...
1801 (dom_jt_simplifier::simplify): ...to this.
1802 (pass_dominator::execute): Use dom_jt_simplifier and
1804 * tree-ssa-threadedge.c (jump_threader::jump_threader):
1806 (jt_state::register_equivs_stmt): Abstract out...
1807 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
1809 (jump_threader::thread_around_empty_blocks): Update state.
1810 (jump_threader::thread_through_normal_block): Same.
1811 (jt_state::jt_state): Remove.
1812 (jt_state::push): Remove pass specific bits. Keep block vector
1814 (jt_state::append_path): New.
1815 (jt_state::pop): Remove pass specific bits.
1816 (jt_state::register_equiv): Same.
1817 (jt_state::record_ranges_from_stmt): Same.
1818 (jt_state::register_equivs_on_edge): Same. Rename...
1819 (jt_state::register_equivs_edge): ...to this.
1820 (jt_state::dump): New.
1821 (jt_state::debug): New.
1822 (jump_threader_simplifier::simplify): Remove.
1823 (jt_state::get_path): New.
1824 * tree-ssa-threadedge.h (class jt_simplifier): Make into a base
1825 class. Expose common functionality as virtual methods.
1826 (class jump_threader_simplifier): Same. Rename...
1827 (class jt_simplifier): ...to this.
1828 * tree-vrp.c (class vrp_jump_threader_simplifier): Rename...
1829 (class vrp_jt_simplifier): ...to this. Provide pass specific
1831 (class vrp_jt_state): New.
1832 (vrp_jump_threader_simplifier::simplify): Rename...
1833 (vrp_jt_simplifier::simplify): ...to this. Inline code from
1834 what used to be the base class.
1835 (vrp_jump_threader::vrp_jump_threader): Use vrp_jt_state and
1838 2021-09-22 Tobias Burnus <tobias@codesourcery.com>
1841 * doc/invoke.texi (-Wno-missing-include-dirs.): Document Fortran
1844 2021-09-22 Roger Sayle <roger@nextmovesoftware.com>
1845 Richard Biener <rguenther@suse.de>
1847 * match.pd (negation simplifications): Implement some negation
1848 folding transformations from fold-const.c's fold_negate_expr.
1849 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Add a SIMPLIFY
1850 argument, to control whether the op should be simplified prior
1851 to looking up/assigning a value number.
1852 (vn_nary_build_or_lookup): Update call to vn_nary_build_or_lookup_1.
1853 (vn_nary_simplify): Likewise.
1854 (visit_nary_op): Likewise, but when constructing a NEGATE_EXPR
1855 now call vn_nary_build_or_lookup_1 disabling simplification.
1857 2021-09-22 Jiufu Guo <guojiufu@linux.ibm.com>
1859 PR tree-optimization/102087
1860 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
1861 Update bound/cmp/control for niter.
1863 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
1865 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
1866 Move check for non-empty BB here.
1867 (fur_source::register_outgoing_edges): ...from here.
1869 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
1871 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
1872 Remove call to improve_range_with_equivs.
1873 (path_range_query::improve_range_with_equivs): Remove
1874 * gimple-range-path.h: Remove improve_range_with_equivs.
1876 2021-09-22 dianhong xu <dianhong.xu@intel.com>
1878 * config/i386/avx512fp16intrin.h:
1879 (_mm512_mask_blend_ph): New intrinsic.
1880 (_mm512_permutex2var_ph): Ditto.
1881 (_mm512_permutexvar_ph): Ditto.
1882 * config/i386/avx512fp16vlintrin.h:
1883 (_mm256_mask_blend_ph): New intrinsic.
1884 (_mm256_permutex2var_ph): Ditto.
1885 (_mm256_permutexvar_ph): Ditto.
1886 (_mm_mask_blend_ph): Ditto.
1887 (_mm_permutex2var_ph): Ditto.
1888 (_mm_permutexvar_ph): Ditto.
1890 2021-09-22 dianhong xu <dianhong.xu@intel.com>
1892 * config/i386/avx512fp16intrin.h: Add new intrinsics.
1893 (_mm512_conj_pch): New intrinsic.
1894 (_mm512_mask_conj_pch): Ditto.
1895 (_mm512_maskz_conj_pch): Ditto.
1896 * config/i386/avx512fp16vlintrin.h: Add new intrinsics.
1897 (_mm256_conj_pch): New intrinsic.
1898 (_mm256_mask_conj_pch): Ditto.
1899 (_mm256_maskz_conj_pch): Ditto.
1900 (_mm_conj_pch): Ditto.
1901 (_mm_mask_conj_pch): Ditto.
1902 (_mm_maskz_conj_pch): Ditto.
1904 2021-09-22 dianhong xu <dianhong.xu@intel.com>
1906 * config/i386/avx512fp16intrin.h (_MM512_REDUCE_OP): New macro
1907 (_mm512_reduce_add_ph): New intrinsic.
1908 (_mm512_reduce_mul_ph): Ditto.
1909 (_mm512_reduce_min_ph): Ditto.
1910 (_mm512_reduce_max_ph): Ditto.
1911 * config/i386/avx512fp16vlintrin.h
1912 (_MM256_REDUCE_OP/_MM_REDUCE_OP): New macro.
1913 (_mm256_reduce_add_ph): New intrinsic.
1914 (_mm256_reduce_mul_ph): Ditto.
1915 (_mm256_reduce_min_ph): Ditto.
1916 (_mm256_reduce_max_ph): Ditto.
1917 (_mm_reduce_add_ph): Ditto.
1918 (_mm_reduce_mul_ph): Ditto.
1919 (_mm_reduce_min_ph): Ditto.
1920 (_mm_reduce_max_ph): Ditto.
1922 2021-09-22 dianhong xu <dianhong.xu@intel.com>
1924 * config/i386/avx512fp16intrin.h (__m512h_u, __m256h_u,
1925 __m128h_u): New typedef.
1926 (_mm512_load_ph): New intrinsic.
1927 (_mm256_load_ph): Ditto.
1928 (_mm_load_ph): Ditto.
1929 (_mm512_loadu_ph): Ditto.
1930 (_mm256_loadu_ph): Ditto.
1931 (_mm_loadu_ph): Ditto.
1932 (_mm512_store_ph): Ditto.
1933 (_mm256_store_ph): Ditto.
1934 (_mm_store_ph): Ditto.
1935 (_mm512_storeu_ph): Ditto.
1936 (_mm256_storeu_ph): Ditto.
1937 (_mm_storeu_ph): Ditto.
1938 (_mm512_abs_ph): Ditto.
1939 * config/i386/avx512fp16vlintrin.h
1940 (_mm_abs_ph): Ditto.
1941 (_mm256_abs_ph): Ditto.
1943 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
1945 * config/s390/tpf.md (prologue_tpf, epilogue_tpf): Add cc clobber.
1947 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
1950 * config/s390/s390.c (s390_expand_insv): Emit a normal move if it
1951 is actually a full copy of the source operand into the target.
1952 Don't emit a strict low part move if source and target mode match.
1954 2021-09-22 Jakub Jelinek <jakub@redhat.com>
1956 PR middle-end/102415
1957 * omp-expand.c (expand_omp_single): If region->exit is NULL,
1958 assert region->entry is GIMPLE_OMP_SCOPE region and return.
1960 2021-09-22 Jakub Jelinek <jakub@redhat.com>
1962 * tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
1963 * tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
1964 arguments from 2 to 3.
1965 * tree-pretty-print.c (dump_omp_clause): Print allocator() around
1966 allocate clause allocator and print align if present.
1967 * omp-low.c (scan_sharing_clauses): Force allocate_map entry even
1968 for omp_default_mem_alloc if align modifier is present. If align
1969 modifier is present, use TREE_LIST to encode both allocator and
1971 (lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
1972 Handle align modifier on allocator clause if present.
1974 2021-09-22 liuhongt <hongtao.liu@intel.com>
1976 * config/i386/i386.md (define_attr "isa"): Add
1978 (define_attr "enabled"): Correspond fma_or_avx512vl to
1979 TARGET_FMA || TARGET_AVX512VL.
1980 * config/i386/mmx.md (fmav2sf4): Extend to AVX512 fma.
1985 2021-09-22 liuhongt <hongtao.liu@intel.com>
1987 * config/i386/i386.md (cstorehf3): New define_expand.
1989 2021-09-22 liuhongt <hongtao.liu@intel.com>
1991 * config/i386/i386.md (<rounding_insn>hf2): New expander.
1992 (sse4_1_round<mode>2): Extend from MODEF to MODEFH.
1993 * config/i386/sse.md (*sse4_1_round<ssescalarmodesuffix>):
1994 Extend from VF_128 to VFH_128.
1996 2021-09-22 liuhongt <hongtao.liu@intel.com>
1998 * config/i386/i386-features.c (i386-features.c): Handle
2000 * config/i386/i386.md (sqrthf2): New expander.
2001 (*sqrthf2): New define_insn.
2002 * config/i386/sse.md
2003 (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
2006 2021-09-22 liuhongt <hongtao.liu@intel.com>
2008 * config/i386/avx512fp16intrin.h (_mm_mask_fcmadd_sch):
2010 (_mm_mask3_fcmadd_sch): Likewise.
2011 (_mm_maskz_fcmadd_sch): Likewise.
2012 (_mm_fcmadd_sch): Likewise.
2013 (_mm_mask_fmadd_sch): Likewise.
2014 (_mm_mask3_fmadd_sch): Likewise.
2015 (_mm_maskz_fmadd_sch): Likewise.
2016 (_mm_fmadd_sch): Likewise.
2017 (_mm_mask_fcmadd_round_sch): Likewise.
2018 (_mm_mask3_fcmadd_round_sch): Likewise.
2019 (_mm_maskz_fcmadd_round_sch): Likewise.
2020 (_mm_fcmadd_round_sch): Likewise.
2021 (_mm_mask_fmadd_round_sch): Likewise.
2022 (_mm_mask3_fmadd_round_sch): Likewise.
2023 (_mm_maskz_fmadd_round_sch): Likewise.
2024 (_mm_fmadd_round_sch): Likewise.
2025 (_mm_fcmul_sch): Likewise.
2026 (_mm_mask_fcmul_sch): Likewise.
2027 (_mm_maskz_fcmul_sch): Likewise.
2028 (_mm_fmul_sch): Likewise.
2029 (_mm_mask_fmul_sch): Likewise.
2030 (_mm_maskz_fmul_sch): Likewise.
2031 (_mm_fcmul_round_sch): Likewise.
2032 (_mm_mask_fcmul_round_sch): Likewise.
2033 (_mm_maskz_fcmul_round_sch): Likewise.
2034 (_mm_fmul_round_sch): Likewise.
2035 (_mm_mask_fmul_round_sch): Likewise.
2036 (_mm_maskz_fmul_round_sch): Likewise.
2037 * config/i386/i386-builtin.def: Add corresponding new builtins.
2038 * config/i386/sse.md
2039 (avx512fp16_fmaddcsh_v8hf_maskz<round_expand_name>): New expander.
2040 (avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>): Ditto.
2041 (avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>):
2043 (avx512fp16_<complexopname>sh_v8hf_mask<round_name>): Ditto.
2044 (avx512fp16_<complexopname>sh_v8hf<mask_scalarc_name><round_scalarcz_name>):
2046 * config/i386/subst.md (mask_scalarcz_name): New.
2047 (mask_scalarc_name): Ditto.
2048 (mask_scalarc_operand3): Ditto.
2049 (mask_scalarcz_operand4): Ditto.
2050 (round_scalarcz_name): Ditto.
2051 (round_scalarc_mask_operand3): Ditto.
2052 (round_scalarcz_mask_operand4): Ditto.
2053 (round_scalarc_mask_op3): Ditto.
2054 (round_scalarcz_mask_op4): Ditto.
2055 (round_scalarcz_constraint): Ditto.
2056 (round_scalarcz_nimm_predicate): Ditto.
2057 (mask_scalarcz): Ditto.
2058 (mask_scalarc): Ditto.
2059 (round_scalarcz): Ditto.
2061 2021-09-22 liuhongt <hongtao.liu@intel.com>
2063 * config/i386/avx512fp16intrin.h (_mm512_fcmadd_pch):
2065 (_mm512_mask_fcmadd_pch): Likewise.
2066 (_mm512_mask3_fcmadd_pch): Likewise.
2067 (_mm512_maskz_fcmadd_pch): Likewise.
2068 (_mm512_fmadd_pch): Likewise.
2069 (_mm512_mask_fmadd_pch): Likewise.
2070 (_mm512_mask3_fmadd_pch): Likewise.
2071 (_mm512_maskz_fmadd_pch): Likewise.
2072 (_mm512_fcmadd_round_pch): Likewise.
2073 (_mm512_mask_fcmadd_round_pch): Likewise.
2074 (_mm512_mask3_fcmadd_round_pch): Likewise.
2075 (_mm512_maskz_fcmadd_round_pch): Likewise.
2076 (_mm512_fmadd_round_pch): Likewise.
2077 (_mm512_mask_fmadd_round_pch): Likewise.
2078 (_mm512_mask3_fmadd_round_pch): Likewise.
2079 (_mm512_maskz_fmadd_round_pch): Likewise.
2080 (_mm512_fcmul_pch): Likewise.
2081 (_mm512_mask_fcmul_pch): Likewise.
2082 (_mm512_maskz_fcmul_pch): Likewise.
2083 (_mm512_fmul_pch): Likewise.
2084 (_mm512_mask_fmul_pch): Likewise.
2085 (_mm512_maskz_fmul_pch): Likewise.
2086 (_mm512_fcmul_round_pch): Likewise.
2087 (_mm512_mask_fcmul_round_pch): Likewise.
2088 (_mm512_maskz_fcmul_round_pch): Likewise.
2089 (_mm512_fmul_round_pch): Likewise.
2090 (_mm512_mask_fmul_round_pch): Likewise.
2091 (_mm512_maskz_fmul_round_pch): Likewise.
2092 * config/i386/avx512fp16vlintrin.h (_mm_fmadd_pch):
2094 (_mm_mask_fmadd_pch): Likewise.
2095 (_mm_mask3_fmadd_pch): Likewise.
2096 (_mm_maskz_fmadd_pch): Likewise.
2097 (_mm256_fmadd_pch): Likewise.
2098 (_mm256_mask_fmadd_pch): Likewise.
2099 (_mm256_mask3_fmadd_pch): Likewise.
2100 (_mm256_maskz_fmadd_pch): Likewise.
2101 (_mm_fcmadd_pch): Likewise.
2102 (_mm_mask_fcmadd_pch): Likewise.
2103 (_mm_mask3_fcmadd_pch): Likewise.
2104 (_mm_maskz_fcmadd_pch): Likewise.
2105 (_mm256_fcmadd_pch): Likewise.
2106 (_mm256_mask_fcmadd_pch): Likewise.
2107 (_mm256_mask3_fcmadd_pch): Likewise.
2108 (_mm256_maskz_fcmadd_pch): Likewise.
2109 (_mm_fmul_pch): Likewise.
2110 (_mm_mask_fmul_pch): Likewise.
2111 (_mm_maskz_fmul_pch): Likewise.
2112 (_mm256_fmul_pch): Likewise.
2113 (_mm256_mask_fmul_pch): Likewise.
2114 (_mm256_maskz_fmul_pch): Likewise.
2115 (_mm_fcmul_pch): Likewise.
2116 (_mm_mask_fcmul_pch): Likewise.
2117 (_mm_maskz_fcmul_pch): Likewise.
2118 (_mm256_fcmul_pch): Likewise.
2119 (_mm256_mask_fcmul_pch): Likewise.
2120 (_mm256_maskz_fcmul_pch): Likewise.
2121 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF_V8HF_V8HF,
2122 V8HF_FTYPE_V16HF_V16HF_V16HF, V16HF_FTYPE_V16HF_V16HF_V16HF_UQI,
2123 V32HF_FTYPE_V32HF_V32HF_V32HF_INT,
2124 V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT): Add new builtin types.
2125 * config/i386/i386-builtin.def: Add new builtins.
2126 * config/i386/i386-expand.c: Handle new builtin types.
2127 * config/i386/subst.md (SUBST_CV): New.
2128 (maskc_name): Ditto.
2129 (maskc_operand3): Ditto.
2131 (sdc_maskz_name): Ditto.
2132 (sdc_mask_op4): Ditto.
2133 (sdc_mask_op5): Ditto.
2134 (sdc_mask_mode512bit_condition): Ditto.
2136 (round_maskc_operand3): Ditto.
2137 (round_sdc_mask_operand4): Ditto.
2138 (round_maskc_op3): Ditto.
2139 (round_sdc_mask_op4): Ditto.
2140 (round_saeonly_sdc_mask_operand5): Ditto.
2141 * config/i386/sse.md (unspec): Add complex fma unspecs.
2142 (avx512fmaskcmode): New.
2143 (UNSPEC_COMPLEX_F_C_MA): Ditto.
2144 (UNSPEC_COMPLEX_F_C_MUL): Ditto.
2145 (complexopname): Ditto.
2146 (<avx512>_fmaddc_<mode>_maskz<round_expand_name>): New expander.
2147 (<avx512>_fcmaddc_<mode>_maskz<round_expand_name>): Ditto.
2148 (fma_<complexopname>_<mode><sdc_maskz_name><round_name>): New
2150 (<avx512>_<complexopname>_<mode>_mask<round_name>): Ditto.
2151 (<avx512>_<complexopname>_<mode><maskc_name><round_name>): Ditto.
2153 2021-09-22 Kewen Lin <linkw@linux.ibm.com>
2155 * config/rs6000/rs6000.opt (rs6000-density-pct-threshold,
2156 rs6000-density-size-threshold, rs6000-density-penalty,
2157 rs6000-density-load-pct-threshold,
2158 rs6000-density-load-num-threshold): New parameter.
2159 * config/rs6000/rs6000.c (rs6000_density_test): Adjust with
2160 corresponding parameters.
2162 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2164 * gimple-range-path.cc (path_range_query::defined_outside_path):
2166 (path_range_query::range_on_path_entry): New.
2167 (path_range_query::internal_range_of_expr): Resolve unknowns
2169 (path_range_query::improve_range_with_equivs): New.
2170 (path_range_query::ssa_range_in_phi): Resolve unknowns with
2172 * gimple-range-path.h (class path_range_query): Add
2173 defined_outside_path, range_on_path_entry, and
2174 improve_range_with_equivs.
2176 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2178 * gimple-range-path.cc (path_range_query::add_to_imports): New.
2179 (path_range_query::add_copies_to_imports): New.
2180 (path_range_query::precompute_ranges): Call
2181 add_copies_to_imports.
2182 * gimple-range-path.h (class path_range_query): Add prototypes
2183 for add_copies_to_imports and add_to_imports.
2185 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2187 * gimple-range-path.cc (path_range_query::range_defined_in_block):
2188 Remove useless code.
2190 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2192 * gimple-range-fold.h (class fur_source): Make oracle protected.
2193 * gimple-range-path.cc (path_range_query::path_range_query): Add
2194 resolve argument. Initialize oracle.
2195 (path_range_query::~path_range_query): Delete oracle.
2196 (path_range_query::range_of_stmt): Adapt to use relations.
2197 (path_range_query::precompute_ranges): Pre-compute relations.
2198 (class jt_fur_source): New
2199 (jt_fur_source::jt_fur_source): New.
2200 (jt_fur_source::register_relation): New.
2201 (jt_fur_source::query_relation): New.
2202 (path_range_query::precompute_relations): New.
2203 (path_range_query::precompute_phi_relations): New.
2204 * gimple-range-path.h (path_range_query): Add resolve argument.
2205 Add oracle, precompute_relations, precompute_phi_relations.
2206 * tree-ssa-threadbackward.c (back_threader::back_threader): Pass
2207 resolve argument to solver.
2209 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2211 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
2212 Rename postfold_gcond_edges to register_outgoing_edges and
2214 (fold_using_range::postfold_gcond_edges): Rename...
2215 (fur_source::register_outgoing_edges): ...to this.
2216 * gimple-range-fold.h (postfold_gcond_edges): Rename to
2217 register_outgoing_edges and move to fur_source.
2219 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2221 * gimple-range-fold.cc (fold_using_range::range_of_phi): Check
2222 dom_info_available_p.
2224 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2226 * gimple-range-cache.cc (non_null_ref::non_null_ref): Use create
2227 and quick_grow_cleared instead of safe_grow_cleared.
2229 2021-09-21 Thomas Schwinge <thomas@codesourcery.com>
2232 * omp-oacc-neuter-broadcast.cc (oacc_do_neutering): Evaluate
2235 2021-09-21 Richard Earnshaw <rearnsha@arm.com>
2237 * configure.ac: Detect when the assembler supports new-style
2238 architecture extensions.
2239 * common/config/arm/arm-common.c (arm_rewrite_mcpu): Return
2240 the full CPU string if the assembler can grok it.
2241 (arm_rewrite_march): Likewise but for the architecture.
2242 * config.in: Regenerate.
2243 * configure: Regenerate.
2245 2021-09-21 Richard Biener <rguenther@suse.de>
2247 PR tree-optimization/102421
2248 * tree-vect-loop.c (vect_dissolve_slp_only_groups): Copy and
2249 adjust alignment info.
2251 2021-09-21 Kewen Lin <linkw@linux.ibm.com>
2253 * ipa-fnsummary.c (ipa_fn_summary_write): Remove inconsistent
2254 bitfield stream out.
2256 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
2258 * gimple-range-fold.cc (fold_using_range::range_of_phi): Ignore
2259 undefined edges, apply an equivalence if appropriate.
2260 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Return
2261 UNDEFINED if EDGE_EXECUTABLE is not set.
2262 * gimple-range.cc (gimple_ranger::gimple_ranger): Set all edges
2263 as EXECUTABLE upon startup.
2264 (gimple_ranger::range_on_edge): Return UNDEFINED for edges without
2265 EDGE_EXECUTABLE set.
2266 * vr-values.c (set_and_propagate_unexecutable): New.
2267 (simplify_using_ranges::fold_cond): Call set_and_propagate.
2268 (simplify_using_ranges::simplify_switch_using_ranges): Ditto.
2269 * vr-values.h: Add prototype.
2271 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
2273 * value-relation.cc (equiv_oracle::register_initial_def): New.
2274 (equiv_oracle::register_relation): Call register_initial_def.
2275 (equiv_oracle::add_equiv_to_block): New. Split register_relation.
2276 (relation_oracle::register_stmt): Check def block of PHI arguments.
2277 * value-relation.h (equiv_oracle): Add new prototypes.
2279 2021-09-20 Matthias Kretz <m.kretz@gsi.de>
2281 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
2282 Define __RECIPROCAL_MATH__, __NO_SIGNED_ZEROS__,
2283 __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__, and
2284 __ROUNDING_MATH__ according to their corresponding flags.
2285 * doc/cpp.texi: Document __RECIPROCAL_MATH__,
2286 __NO_SIGNED_ZEROS__, __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__,
2287 and __ROUNDING_MATH__.
2289 2021-09-20 Richard Biener <rguenther@suse.de>
2291 * tree-vect-stmts.c (vectorizable_load): Use the vectype
2294 2021-09-20 Richard Biener <rguenther@suse.de>
2296 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
2297 Do not compute alignment of the vectorized access here.
2299 2021-09-20 Richard Biener <rguenther@suse.de>
2301 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
2302 Store -1 for runtime alias peeling iterations.
2304 2021-09-20 Richard Biener <rguenther@suse.de>
2306 * config.gcc: Obsolete hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11*.
2308 2021-09-20 Thomas Schwinge <thomas@codesourcery.com>
2310 * input.c (string_concat_db::record_string_concatenation)
2311 (string_concat_db::get_string_concatenation): Skip for
2312 'RESERVED_LOCATION_P'.
2314 2021-09-20 Richard Biener <rguenther@suse.de>
2316 PR tree-optimization/65206
2317 * tree-data-ref.h (struct data_reference): Add alt_indices,
2319 * tree-data-ref.c (free_data_ref): Release alt_indices.
2320 (dr_analyze_indices): Work on struct indices and get DR_REF as tree.
2321 (create_data_ref): Adjust.
2322 (initialize_data_dependence_relation): Split into head
2323 and tail. When the base objects fail to match up try
2324 again with pointer-based analysis of indices.
2325 * tree-vectorizer.c (vec_info_shared::check_datarefs): Do
2326 not compare the lazily computed alternate set of indices.
2328 2021-09-20 Iain Sandoe <iain@sandoe.co.uk>
2330 * gcc.c: Test for execute OK when we find the
2331 programs for assembler linker and dsymutil and those
2332 were specified at configure-time.
2334 2021-09-19 Martin Sebor <msebor@redhat.com>
2336 PR middle-end/102403
2337 * gimple-predicate-analysis.cc (predicate::init_from_control_deps):
2338 Correct a function pre/postcondition.
2340 2021-09-19 Martin Sebor <msebor@redhat.com>
2342 PR middle-end/102243
2343 * tree-ssa-strlen.c (get_range): Handle null cfun.
2345 2021-09-19 Iain Sandoe <iain@sandoe.co.uk>
2347 * config/darwin.h (LINK_COMMAND_SPEC_A): Use Darwin10
2348 unwinder shim as a convenience library.
2350 2021-09-19 Andrew Pinski <apinski@marvell.com>
2352 * doc/install.texi: Add note about
2353 binutils 2.35 is required for LTO usage.
2355 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
2357 * tree-ssa-threadbackward.c
2358 (back_threader_registry::register_path): Use push_edge.
2359 * tree-ssa-threadedge.c
2360 (jump_threader::thread_around_empty_blocks): Same.
2361 (jump_threader::thread_through_normal_block): Same.
2362 (jump_threader::thread_across_edge): Same. Also, use auto_bitmap.
2364 * tree-ssa-threadupdate.c
2365 (jt_path_registry::allocate_thread_edge): Remove.
2366 (jt_path_registry::push_edge): New.
2367 (dump_jump_thread_path): Make static.
2368 * tree-ssa-threadupdate.h (allocate_thread_edge): Remove.
2371 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
2373 * gimple-range-path.cc (path_range_query::path_range_query): Add
2375 (path_range_query::dump): Remove extern declaration of dump_ranger.
2376 * gimple-range-trace.cc (dump_ranger): Add DEBUG_FUNCTION marker.
2377 * gimple-range-trace.h (dump_ranger): Add prototype.
2379 2021-09-19 John Ericson <git@JohnEricson.me>
2381 * gcc.c (find_a_program): New function, factored out of...
2382 (find_a_file): Here.
2383 (execute): Use find_a_program when looking for programs rather
2386 2021-09-19 Matwey V. Kornilov <matwey.kornilov@gmail.com>
2388 * config/avr/avr-mcus.def: Add atmega324pb.
2389 * doc/avr-mmcu.texi: Corresponding changes.
2391 2021-09-19 Roger Sayle <roger@nextmovesoftware.com>
2394 * match.pd (cmp @0 REAL_CST@1): When @0 is also REAL_CST, apply
2395 the same transformations as to @1. For comparisons against NaN,
2396 don't check HONOR_SNANS but confirm that neither operand is a
2399 2021-09-19 Benjamin Peterson <benjamin@locrian.net>
2401 * attribs.c (make_unique_name): Delete.
2402 * attribs.h (make_unique_name): Delete.
2404 2021-09-19 Andrew Pinski <apinski@marvell.com>
2406 * lra-constraints.c (check_and_process_move): Assert
2407 that dclass and sclass are greater than or equal to NO_REGS.
2409 2021-09-18 Jakub Jelinek <jakub@redhat.com>
2411 * tree.h (OMP_CLAUSE_ORDER_UNCONSTRAINED): Define.
2412 * tree-pretty-print.c (dump_omp_clause): Print unconstrained:
2413 for OMP_CLAUSE_ORDER_UNCONSTRAINED.
2415 2021-09-18 liuhongt <hongtao.liu@intel.com>
2417 * config/i386/i386-features.c (remove_partial_avx_dependency):
2418 Restrict TARGET_USE_VECTOR_FP_CONVERTS and
2419 TARGET_USE_VECTOR_CONVERTS to conversion instructions only.
2421 2021-09-18 Jakub Jelinek <jakub@redhat.com>
2423 * gimplify.c (omp_default_clause): For C/C++ default({,first}private),
2424 if file/namespace scope variable doesn't have predetermined sharing,
2425 treat it as if there was default(none).
2427 2021-09-18 liuhongt <hongtao.liu@intel.com>
2429 * config/i386/avx512fp16intrin.h (_mm_fmadd_sh):
2431 (_mm_mask_fmadd_sh): Likewise.
2432 (_mm_mask3_fmadd_sh): Likewise.
2433 (_mm_maskz_fmadd_sh): Likewise.
2434 (_mm_fmadd_round_sh): Likewise.
2435 (_mm_mask_fmadd_round_sh): Likewise.
2436 (_mm_mask3_fmadd_round_sh): Likewise.
2437 (_mm_maskz_fmadd_round_sh): Likewise.
2438 (_mm_fnmadd_sh): Likewise.
2439 (_mm_mask_fnmadd_sh): Likewise.
2440 (_mm_mask3_fnmadd_sh): Likewise.
2441 (_mm_maskz_fnmadd_sh): Likewise.
2442 (_mm_fnmadd_round_sh): Likewise.
2443 (_mm_mask_fnmadd_round_sh): Likewise.
2444 (_mm_mask3_fnmadd_round_sh): Likewise.
2445 (_mm_maskz_fnmadd_round_sh): Likewise.
2446 (_mm_fmsub_sh): Likewise.
2447 (_mm_mask_fmsub_sh): Likewise.
2448 (_mm_mask3_fmsub_sh): Likewise.
2449 (_mm_maskz_fmsub_sh): Likewise.
2450 (_mm_fmsub_round_sh): Likewise.
2451 (_mm_mask_fmsub_round_sh): Likewise.
2452 (_mm_mask3_fmsub_round_sh): Likewise.
2453 (_mm_maskz_fmsub_round_sh): Likewise.
2454 (_mm_fnmsub_sh): Likewise.
2455 (_mm_mask_fnmsub_sh): Likewise.
2456 (_mm_mask3_fnmsub_sh): Likewise.
2457 (_mm_maskz_fnmsub_sh): Likewise.
2458 (_mm_fnmsub_round_sh): Likewise.
2459 (_mm_mask_fnmsub_round_sh): Likewise.
2460 (_mm_mask3_fnmsub_round_sh): Likewise.
2461 (_mm_maskz_fnmsub_round_sh): Likewise.
2462 * config/i386/i386-builtin-types.def
2463 (V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT): New builtin type.
2464 * config/i386/i386-builtin.def: Add new builtins.
2465 * config/i386/i386-expand.c: Handle new builtin type.
2466 * config/i386/sse.md (fmai_vmfmadd_<mode><round_name>):
2467 Ajdust to support FP16.
2468 (fmai_vmfmsub_<mode><round_name>): Ditto.
2469 (fmai_vmfnmadd_<mode><round_name>): Ditto.
2470 (fmai_vmfnmsub_<mode><round_name>): Ditto.
2471 (*fmai_fmadd_<mode>): Ditto.
2472 (*fmai_fmsub_<mode>): Ditto.
2473 (*fmai_fnmadd_<mode><round_name>): Ditto.
2474 (*fmai_fnmsub_<mode><round_name>): Ditto.
2475 (avx512f_vmfmadd_<mode>_mask<round_name>): Ditto.
2476 (avx512f_vmfmadd_<mode>_mask3<round_name>): Ditto.
2477 (avx512f_vmfmadd_<mode>_maskz<round_expand_name>): Ditto.
2478 (avx512f_vmfmadd_<mode>_maskz_1<round_name>): Ditto.
2479 (*avx512f_vmfmsub_<mode>_mask<round_name>): Ditto.
2480 (avx512f_vmfmsub_<mode>_mask3<round_name>): Ditto.
2481 (*avx512f_vmfmsub_<mode>_maskz_1<round_name>): Ditto.
2482 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
2483 (*avx512f_vmfnmsub_<mode>_mask3<round_name>): Ditto.
2484 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
2485 (*avx512f_vmfnmadd_<mode>_mask<round_name>): Renamed to ...
2486 (avx512f_vmfnmadd_<mode>_mask<round_name>) ... this, and
2487 adjust to support FP16.
2488 (avx512f_vmfnmadd_<mode>_mask3<round_name>): Ditto.
2489 (avx512f_vmfnmadd_<mode>_maskz_1<round_name>): Ditto.
2490 (avx512f_vmfnmadd_<mode>_maskz<round_expand_name>): New
2493 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
2495 * config/i386/sse.md (avx512fmaskmodelower): Extend to support
2497 (maskload<mode><avx512fmaskmodelower>): Ditto.
2498 (maskstore<mode><avx512fmaskmodelower>): Ditto.
2500 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
2502 * config/i386/i386-expand.c (ix86_expand_fp_absneg_operator):
2504 (ix86_expand_copysign): Ditto.
2505 (ix86_expand_xorsign): Ditto.
2506 * config/i386/i386.c (ix86_build_const_vector): Handle HF vector
2508 (ix86_build_signbit_mask): Ditto.
2509 (ix86_can_change_mode_class): Ditto.
2510 * config/i386/i386.md
2511 (SSEMODEF): Add HFmode.
2512 (ssevecmodef): Ditto.
2513 (<code>hf2): New define_expand.
2514 (*<code>hf2_1): New define_insn_and_split.
2515 (copysign<mode>): Extend to support HFmode under AVX512FP16.
2516 (xorsign<mode>): Ditto.
2517 * config/i386/sse.md (VFB): New mode iterator.
2518 (VFB_128_256): Ditto.
2520 (sseintvecmode2): Support HF vector mode.
2521 (<code><mode>2): Use new mode iterator.
2522 (*<code><mode>2): Ditto.
2523 (copysign<mode>3): Ditto.
2524 (xorsign<mode>3): Ditto.
2525 (<code><mode>3<mask_name>): Ditto.
2526 (<code><mode>3<mask_name>): Ditto.
2527 (<sse>_andnot<mode>3<mask_name>): Adjust for HF vector mode.
2528 (<sse>_andnot<mode>3<mask_name>): Ditto.
2529 (*<code><mode>3<mask_name>): Ditto.
2530 (*<code><mode>3<mask_name>): Ditto.
2532 2021-09-18 liuhongt <hongtao.liu@intel.com>
2534 * config/i386/avx512fp16intrin.h (_mm512_mask_fmadd_ph):
2536 (_mm512_mask3_fmadd_ph): Likewise.
2537 (_mm512_maskz_fmadd_ph): Likewise.
2538 (_mm512_fmadd_round_ph): Likewise.
2539 (_mm512_mask_fmadd_round_ph): Likewise.
2540 (_mm512_mask3_fmadd_round_ph): Likewise.
2541 (_mm512_maskz_fmadd_round_ph): Likewise.
2542 (_mm512_fnmadd_ph): Likewise.
2543 (_mm512_mask_fnmadd_ph): Likewise.
2544 (_mm512_mask3_fnmadd_ph): Likewise.
2545 (_mm512_maskz_fnmadd_ph): Likewise.
2546 (_mm512_fnmadd_round_ph): Likewise.
2547 (_mm512_mask_fnmadd_round_ph): Likewise.
2548 (_mm512_mask3_fnmadd_round_ph): Likewise.
2549 (_mm512_maskz_fnmadd_round_ph): Likewise.
2550 (_mm512_fmsub_ph): Likewise.
2551 (_mm512_mask_fmsub_ph): Likewise.
2552 (_mm512_mask3_fmsub_ph): Likewise.
2553 (_mm512_maskz_fmsub_ph): Likewise.
2554 (_mm512_fmsub_round_ph): Likewise.
2555 (_mm512_mask_fmsub_round_ph): Likewise.
2556 (_mm512_mask3_fmsub_round_ph): Likewise.
2557 (_mm512_maskz_fmsub_round_ph): Likewise.
2558 (_mm512_fnmsub_ph): Likewise.
2559 (_mm512_mask_fnmsub_ph): Likewise.
2560 (_mm512_mask3_fnmsub_ph): Likewise.
2561 (_mm512_maskz_fnmsub_ph): Likewise.
2562 (_mm512_fnmsub_round_ph): Likewise.
2563 (_mm512_mask_fnmsub_round_ph): Likewise.
2564 (_mm512_mask3_fnmsub_round_ph): Likewise.
2565 (_mm512_maskz_fnmsub_round_ph): Likewise.
2566 * config/i386/avx512fp16vlintrin.h (_mm256_fmadd_ph):
2568 (_mm256_mask_fmadd_ph): Likewise.
2569 (_mm256_mask3_fmadd_ph): Likewise.
2570 (_mm256_maskz_fmadd_ph): Likewise.
2571 (_mm_fmadd_ph): Likewise.
2572 (_mm_mask_fmadd_ph): Likewise.
2573 (_mm_mask3_fmadd_ph): Likewise.
2574 (_mm_maskz_fmadd_ph): Likewise.
2575 (_mm256_fnmadd_ph): Likewise.
2576 (_mm256_mask_fnmadd_ph): Likewise.
2577 (_mm256_mask3_fnmadd_ph): Likewise.
2578 (_mm256_maskz_fnmadd_ph): Likewise.
2579 (_mm_fnmadd_ph): Likewise.
2580 (_mm_mask_fnmadd_ph): Likewise.
2581 (_mm_mask3_fnmadd_ph): Likewise.
2582 (_mm_maskz_fnmadd_ph): Likewise.
2583 (_mm256_fmsub_ph): Likewise.
2584 (_mm256_mask_fmsub_ph): Likewise.
2585 (_mm256_mask3_fmsub_ph): Likewise.
2586 (_mm256_maskz_fmsub_ph): Likewise.
2587 (_mm_fmsub_ph): Likewise.
2588 (_mm_mask_fmsub_ph): Likewise.
2589 (_mm_mask3_fmsub_ph): Likewise.
2590 (_mm_maskz_fmsub_ph): Likewise.
2591 (_mm256_fnmsub_ph): Likewise.
2592 (_mm256_mask_fnmsub_ph): Likewise.
2593 (_mm256_mask3_fnmsub_ph): Likewise.
2594 (_mm256_maskz_fnmsub_ph): Likewise.
2595 (_mm_fnmsub_ph): Likewise.
2596 (_mm_mask_fnmsub_ph): Likewise.
2597 (_mm_mask3_fnmsub_ph): Likewise.
2598 (_mm_maskz_fnmsub_ph): Likewise.
2599 * config/i386/i386-builtin.def: Add corresponding new builtins.
2600 * config/i386/sse.md
2601 (<avx512>_fmadd_<mode>_maskz<round_expand_name>): Adjust to
2602 support HF vector modes.
2603 (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
2605 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
2606 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
2607 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
2608 (<avx512>_fmadd_<mode>_mask<round_name>): Ditto.
2609 (<avx512>_fmadd_<mode>_mask3<round_name>): Ditto.
2610 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Ditto.
2611 (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>):
2613 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
2614 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
2615 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
2616 (<avx512>_fmsub_<mode>_mask<round_name>): Ditto.
2617 (<avx512>_fmsub_<mode>_mask3<round_name>): Ditto.
2618 (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>):
2620 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
2621 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
2622 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
2623 (<avx512>_fnmadd_<mode>_mask<round_name>): Ditto.
2624 (<avx512>_fnmadd_<mode>_mask3<round_name>): Ditto.
2625 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Ditto.
2626 (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>):
2628 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
2629 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
2630 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
2631 (<avx512>_fnmsub_<mode>_mask<round_name>): Ditto.
2632 (<avx512>_fnmsub_<mode>_mask3<round_name>): Ditto.
2634 2021-09-18 liuhongt <hongtao.liu@intel.com>
2636 * config/i386/avx512fp16intrin.h (_mm512_fmaddsub_ph):
2638 (_mm512_mask_fmaddsub_ph): Likewise.
2639 (_mm512_mask3_fmaddsub_ph): Likewise.
2640 (_mm512_maskz_fmaddsub_ph): Likewise.
2641 (_mm512_fmaddsub_round_ph): Likewise.
2642 (_mm512_mask_fmaddsub_round_ph): Likewise.
2643 (_mm512_mask3_fmaddsub_round_ph): Likewise.
2644 (_mm512_maskz_fmaddsub_round_ph): Likewise.
2645 (_mm512_mask_fmsubadd_ph): Likewise.
2646 (_mm512_mask3_fmsubadd_ph): Likewise.
2647 (_mm512_maskz_fmsubadd_ph): Likewise.
2648 (_mm512_fmsubadd_round_ph): Likewise.
2649 (_mm512_mask_fmsubadd_round_ph): Likewise.
2650 (_mm512_mask3_fmsubadd_round_ph): Likewise.
2651 (_mm512_maskz_fmsubadd_round_ph): Likewise.
2652 * config/i386/avx512fp16vlintrin.h (_mm256_fmaddsub_ph):
2654 (_mm256_mask_fmaddsub_ph): Likewise.
2655 (_mm256_mask3_fmaddsub_ph): Likewise.
2656 (_mm256_maskz_fmaddsub_ph): Likewise.
2657 (_mm_fmaddsub_ph): Likewise.
2658 (_mm_mask_fmaddsub_ph): Likewise.
2659 (_mm_mask3_fmaddsub_ph): Likewise.
2660 (_mm_maskz_fmaddsub_ph): Likewise.
2661 (_mm256_fmsubadd_ph): Likewise.
2662 (_mm256_mask_fmsubadd_ph): Likewise.
2663 (_mm256_mask3_fmsubadd_ph): Likewise.
2664 (_mm256_maskz_fmsubadd_ph): Likewise.
2665 (_mm_fmsubadd_ph): Likewise.
2666 (_mm_mask_fmsubadd_ph): Likewise.
2667 (_mm_mask3_fmsubadd_ph): Likewise.
2668 (_mm_maskz_fmsubadd_ph): Likewise.
2669 * config/i386/i386-builtin.def: Add corresponding new builtins.
2670 * config/i386/sse.md (VFH_SF_AVX512VL): New mode iterator.
2671 * (<avx512>_fmsubadd_<mode>_maskz<round_expand_name>): New expander.
2672 * (<avx512>_fmaddsub_<mode>_maskz<round_expand_name>): Use
2674 * (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
2676 * (<avx512>_fmaddsub_<mode>_mask<round_name>): Ditto.
2677 * (<avx512>_fmaddsub_<mode>_mask3<round_name>): Ditto.
2678 * (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
2680 * (<avx512>_fmsubadd_<mode>_mask<round_name>): Ditto.
2681 * (<avx512>_fmsubadd_<mode>_mask3<round_name>): Ditto.
2683 2021-09-18 liuhongt <hongtao.liu@intel.com>
2686 * config/i386/i386.c (ix86_print_operand): Handle
2687 V8HF/V16HF/V32HFmode.
2688 * config/i386/i386.h (VALID_BCST_MODE_P): Add HFmode.
2689 * config/i386/sse.md (avx512bcst): Remove.
2691 2021-09-17 Martin Sebor <msebor@redhat.com>
2693 * Makefile.in (OBJS): Add gimple-predicate-analysis.o.
2694 * tree-ssa-uninit.c (max_phi_args): Move to gimple-predicate-analysis.
2695 (MASK_SET_BIT, MASK_TEST_BIT, MASK_EMPTY): Same.
2696 (check_defs): Add comment.
2697 (can_skip_redundant_opnd): Update comment.
2698 (compute_uninit_opnds_pos): Adjust to namespace change.
2699 (find_pdom): Move to gimple-predicate-analysis.cc.
2701 (struct uninit_undef_val_t): New.
2702 (is_non_loop_exit_postdominating): Move to gimple-predicate-analysis.cc.
2703 (find_control_equiv_block): Same.
2704 (MAX_NUM_CHAINS, MAX_CHAIN_LEN, MAX_POSTDOM_CHECK): Same.
2705 (MAX_SWITCH_CASES): Same.
2706 (compute_control_dep_chain): Same.
2707 (find_uninit_use): Use predicate analyzer.
2708 (struct pred_info): Move to gimple-predicate-analysis.
2709 (convert_control_dep_chain_into_preds): Same.
2710 (find_predicates): Same.
2711 (collect_phi_def_edges): Same.
2712 (warn_uninitialized_phi): Use predicate analyzer.
2713 (find_def_preds): Move to gimple-predicate-analysis.
2714 (dump_pred_info): Same.
2715 (dump_pred_chain): Same.
2716 (dump_predicates): Same.
2717 (destroy_predicate_vecs): Remove.
2718 (execute_late_warn_uninitialized): New.
2719 (get_cmp_code): Move to gimple-predicate-analysis.
2720 (is_value_included_in): Same.
2721 (value_sat_pred_p): Same.
2722 (find_matching_predicate_in_rest_chains): Same.
2723 (is_use_properly_guarded): Same.
2724 (prune_uninit_phi_opnds): Same.
2725 (find_var_cmp_const): Same.
2726 (use_pred_not_overlap_with_undef_path_pred): Same.
2727 (pred_equal_p): Same.
2728 (is_neq_relop_p): Same.
2729 (is_neq_zero_form_p): Same.
2730 (pred_expr_equal_p): Same.
2731 (is_pred_expr_subset_of): Same.
2732 (is_pred_chain_subset_of): Same.
2733 (is_included_in): Same.
2734 (is_superset_of): Same.
2736 (simplify_pred): Same.
2737 (simplify_preds_2): Same.
2738 (simplify_preds_3): Same.
2739 (simplify_preds_4): Same.
2740 (simplify_preds): Same.
2742 (push_to_worklist): Same.
2743 (get_pred_info_from_cmp): Same.
2744 (is_degenerated_phi): Same.
2745 (normalize_one_pred_1): Same.
2746 (normalize_one_pred): Same.
2747 (normalize_one_pred_chain): Same.
2748 (normalize_preds): Same.
2749 (can_one_predicate_be_invalidated_p): Same.
2750 (can_chain_union_be_invalidated_p): Same.
2751 (uninit_uses_cannot_happen): Same.
2752 (pass_late_warn_uninitialized::execute): Define.
2753 * gimple-predicate-analysis.cc: New file.
2754 * gimple-predicate-analysis.h: New file.
2756 2021-09-17 Julian Brown <julian@codesourcery.com>
2758 * config/gcn/gcn.c (gimple.h): Include.
2759 (gcn_fork_join): Emit barrier for worker-level joins.
2760 * omp-oacc-neuter-broadcast.cc (find_local_vars_to_propagate): Add
2761 writes_gang_private bitmap parameter. Set bit for blocks
2762 containing gang-private variable writes.
2763 (worker_single_simple): Don't emit barrier after predicated block.
2764 (worker_single_copy): Don't emit barrier if we're not broadcasting
2765 anything and the block contains no gang-private writes.
2766 (neuter_worker_single): Don't predicate blocks that only contain
2767 NOPs or internal marker functions. Pass has_gang_private_write
2768 argument to worker_single_copy.
2769 (oacc_do_neutering): Add writes_gang_private bitmap handling.
2771 2021-09-17 Julian Brown <julian@codesourcery.com>
2773 * config/gcn/gcn-protos.h
2774 (gcn_goacc_create_worker_broadcast_record): Update prototype.
2775 * config/gcn/gcn-tree.c (gcn_goacc_get_worker_red_decl): Use
2776 preallocated block of LDS memory. Do not cache/share decls for
2777 reduction temporaries between invocations.
2778 (gcn_goacc_reduction_teardown): Unshare VAR on second use.
2779 (gcn_goacc_create_worker_broadcast_record): Add OFFSET parameter
2780 and return temporary LDS space at that offset. Return pointer in
2782 * config/gcn/gcn.c (acc_lds_size, gang_private_hwm, lds_allocs):
2784 (ACC_LDS_SIZE): Define as acc_lds_size.
2785 (gcn_init_machine_status): Don't initialise lds_allocated,
2786 lds_allocs, reduc_decls fields of machine function struct.
2787 (gcn_option_override): Handle default size for gang-private
2788 variables and -mgang-private-size option.
2789 (gcn_expand_prologue): Use LDS_SIZE instead of LDS_SIZE-1 when
2790 initialising M0_REG.
2791 (gcn_shared_mem_layout): New function.
2792 (gcn_print_lds_decl): Update comment. Use global lds_allocs map and
2793 gang_private_hwm variable.
2794 (TARGET_GOACC_SHARED_MEM_LAYOUT): Define target hook.
2795 * config/gcn/gcn.h (machine_function): Remove lds_allocated,
2796 lds_allocs, reduc_decls. Add reduction_base, reduction_limit.
2797 * config/gcn/gcn.opt (gang_private_size_opt): New global.
2798 (mgang-private-size=): New option.
2799 * doc/tm.texi.in (TARGET_GOACC_SHARED_MEM_LAYOUT): Place
2801 * doc/tm.texi: Regenerate.
2802 * omp-oacc-neuter-broadcast.cc (targhooks.h, diagnostic-core.h):
2804 (build_sender_ref): Handle sender_decl being pointer.
2805 (worker_single_copy): Add PLACEMENT and ISOLATE_BROADCASTS
2806 parameters. Pass placement argument to
2807 create_worker_broadcast_record hook invocations. Handle
2808 sender_decl being pointer and isolate_broadcasts inserting extra
2810 (blk_offset_map_t): Add typedef.
2811 (neuter_worker_single): Add BLK_OFFSET_MAP parameter. Pass
2812 preallocated range to worker_single_copy call.
2813 (dfs_broadcast_reachable_1): New function.
2814 (idx_decl_pair_t, used_range_vec_t): New typedefs.
2815 (sort_size_descending): New function.
2816 (addr_range): New class.
2817 (splay_tree_compare_addr_range, splay_tree_free_key)
2818 (first_fit_range, merge_ranges_1, merge_ranges): New functions.
2819 (execute_omp_oacc_neuter_broadcast): Rename to...
2820 (oacc_do_neutering): ... this. Add BOUNDS_LO, BOUNDS_HI
2821 parameters. Arrange layout of shared memory for broadcast
2823 (execute_omp_oacc_neuter_broadcast): New function.
2824 (pass_omp_oacc_neuter_broadcast::gate): Remove num_workers==1
2825 handling from here. Enable pass for all OpenACC routines in order
2826 to call shared memory-layout hook.
2827 * target.def (create_worker_broadcast_record): Add OFFSET
2829 (shared_mem_layout): New hook.
2831 2021-09-17 Julian Brown <julian@codesourcery.com>
2832 Thomas Schwinge <thomas@codesourcery.com>
2834 * omp-oacc-neuter-broadcast.cc
2835 (pass_omp_oacc_neuter_broadcast::gate): Disable if num_workers is
2837 (execute_omp_oacc_neuter_broadcast): Adjust.
2839 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
2841 * value-relation.cc (class equiv_chain): Move to header file.
2842 (path_oracle::path_oracle): New.
2843 (path_oracle::~path_oracle): New.
2844 (path_oracle::register_relation): New.
2845 (path_oracle::query_relation): New.
2846 (path_oracle::reset_path): New.
2847 (path_oracle::dump): New.
2848 * value-relation.h (class equiv_chain): Move to here.
2849 (class path_oracle): New.
2851 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
2853 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a DOM
2855 * gimple-range-fold.cc (fur_depend::register_relation): Use
2856 register_stmt/edge routines.
2857 * value-relation.cc (equiv_chain::find): Relocate from equiv_oracle.
2858 (equiv_oracle::equiv_oracle): Create self equivalence cache.
2859 (equiv_oracle::~equiv_oracle): Release same.
2860 (equiv_oracle::equiv_set): Return entry from self equiv cache if there
2861 are no equivalences.
2862 (equiv_oracle::find_equiv_block): Move list find to equiv_chain.
2863 (equiv_oracle::register_relation): Rename from register_equiv.
2864 (relation_chain_head::find_relation): Relocate from dom_oracle.
2865 (relation_oracle::register_stmt): New.
2866 (relation_oracle::register_edge): New.
2867 (dom_oracle::*): Rename from relation_oracle.
2868 (dom_oracle::register_relation): Adjust to call equiv_oracle.
2869 (dom_oracle::set_one_relation): Split from register_relation.
2870 (dom_oracle::register_transitives): Consolidate 2 methods.
2871 (dom_oracle::find_relation_block): Move core to relation_chain.
2872 (dom_oracle::query_relation): Rename from find_relation_dom and adjust.
2873 * value-relation.h (class relation_oracle): New pure virtual base.
2874 (class equiv_oracle): Inherit from relation_oracle and adjust.
2875 (class dom_oracle): Rename from old relation_oracle and adjust.
2877 2021-09-17 Martin Sebor <msebor@redhat.com>
2879 PR middle-end/102200
2880 * pointer-query.cc (access_ref::inform_access): Handle MIN/MAX_EXPR.
2881 (handle_min_max_size): Change argument. Store original SSA_NAME for
2882 operands to potentially distinct (sub)objects.
2883 (compute_objsize_r): Adjust call to the above.
2885 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
2887 * config/rs6000/rs6000.c (rs6000-builtins.h): New include.
2888 (rs6000_new_builtin_vectorized_function): New function.
2889 (rs6000_new_builtin_md_vectorized_function): Likewise.
2890 (rs6000_builtin_vectorized_function): Call
2891 rs6000_new_builtin_vectorized_function.
2892 (rs6000_builtin_md_vectorized_function): Call
2893 rs6000_new_builtin_md_vectorized_function.
2895 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
2897 * config/rs6000/rs6000-builtin-new.def (ASSEMBLE_ACC): Add mmaint flag.
2898 (ASSEMBLE_PAIR): Likewise.
2899 (BUILD_ACC): Likewise.
2900 (DISASSEMBLE_ACC): Likewise.
2901 (DISASSEMBLE_PAIR): Likewise.
2902 (PMXVBF16GER2): Likewise.
2903 (PMXVBF16GER2NN): Likewise.
2904 (PMXVBF16GER2NP): Likewise.
2905 (PMXVBF16GER2PN): Likewise.
2906 (PMXVBF16GER2PP): Likewise.
2907 (PMXVF16GER2): Likewise.
2908 (PMXVF16GER2NN): Likewise.
2909 (PMXVF16GER2NP): Likewise.
2910 (PMXVF16GER2PN): Likewise.
2911 (PMXVF16GER2PP): Likewise.
2912 (PMXVF32GER): Likewise.
2913 (PMXVF32GERNN): Likewise.
2914 (PMXVF32GERNP): Likewise.
2915 (PMXVF32GERPN): Likewise.
2916 (PMXVF32GERPP): Likewise.
2917 (PMXVF64GER): Likewise.
2918 (PMXVF64GERNN): Likewise.
2919 (PMXVF64GERNP): Likewise.
2920 (PMXVF64GERPN): Likewise.
2921 (PMXVF64GERPP): Likewise.
2922 (PMXVI16GER2): Likewise.
2923 (PMXVI16GER2PP): Likewise.
2924 (PMXVI16GER2S): Likewise.
2925 (PMXVI16GER2SPP): Likewise.
2926 (PMXVI4GER8): Likewise.
2927 (PMXVI4GER8PP): Likewise.
2928 (PMXVI8GER4): Likewise.
2929 (PMXVI8GER4PP): Likewise.
2930 (PMXVI8GER4SPP): Likewise.
2931 (XVBF16GER2): Likewise.
2932 (XVBF16GER2NN): Likewise.
2933 (XVBF16GER2NP): Likewise.
2934 (XVBF16GER2PN): Likewise.
2935 (XVBF16GER2PP): Likewise.
2936 (XVF16GER2): Likewise.
2937 (XVF16GER2NN): Likewise.
2938 (XVF16GER2NP): Likewise.
2939 (XVF16GER2PN): Likewise.
2940 (XVF16GER2PP): Likewise.
2941 (XVF32GER): Likewise.
2942 (XVF32GERNN): Likewise.
2943 (XVF32GERNP): Likewise.
2944 (XVF32GERPN): Likewise.
2945 (XVF32GERPP): Likewise.
2946 (XVF64GER): Likewise.
2947 (XVF64GERNN): Likewise.
2948 (XVF64GERNP): Likewise.
2949 (XVF64GERPN): Likewise.
2950 (XVF64GERPP): Likewise.
2951 (XVI16GER2): Likewise.
2952 (XVI16GER2PP): Likewise.
2953 (XVI16GER2S): Likewise.
2954 (XVI16GER2SPP): Likewise.
2955 (XVI4GER8): Likewise.
2956 (XVI4GER8PP): Likewise.
2957 (XVI8GER4): Likewise.
2958 (XVI8GER4PP): Likewise.
2959 (XVI8GER4SPP): Likewise.
2960 (XXMFACC): Likewise.
2961 (XXMTACC): Likewise.
2962 (XXSETACCZ): Likewise.
2963 (ASSEMBLE_PAIR_V): Likewise.
2964 (BUILD_PAIR): Likewise.
2965 (DISASSEMBLE_PAIR_V): Likewise.
2968 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_mma_builtin):
2969 Handle RS6000_BIF_LXVP and RS6000_BIF_STXVP.
2970 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add ismmaint.
2971 (parse_bif_attrs): Handle ismmaint.
2972 (write_decls): Add bif_mmaint_bit and bif_is_mmaint.
2973 (write_bif_static_init): Handle ismmaint.
2975 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
2977 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin): New
2979 (rs6000_gimple_fold_builtin): Call rs6000_gimple_fold_new_builtin.
2980 (rs6000_new_builtin_valid_without_lhs): New function.
2981 (rs6000_gimple_fold_new_mma_builtin): Likewise.
2982 (rs6000_gimple_fold_new_builtin): Likewise.
2984 2021-09-17 Thomas Schwinge <thomas@codesourcery.com>
2986 * hash-table.h (hash_table<Descriptor, Lazy, Allocator>::expand):
2987 Destruct stale Value objects.
2988 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
2991 2021-09-17 Roger Sayle <roger@nextmovesoftware.com>
2994 * match.pd (shift optimizations): Disable recent sign-changing
2995 optimization for shifts by zero, these will be folded later.
2997 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
2999 * config/rs6000/rs6000-builtin-new.def (__builtin_mffsl): Move from
3000 [power9] to [always].
3002 2021-09-17 Richard Biener <rguenther@suse.de>
3004 * tree-vect-stmts.c (vectorizable_load): Do not frob
3007 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
3009 * config/i386/i386-features.c (remove_partial_avx_dependency):
3010 Also check TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY and
3011 and TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY before generating
3013 * config/i386/i386.h (TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY):
3015 (TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
3016 * config/i386/i386.md (SSE FP to FP splitters): Replace
3017 TARGET_SSE_PARTIAL_REG_DEPENDENCY with
3018 TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY.
3019 (SSE INT to FP splitter): Replace TARGET_SSE_PARTIAL_REG_DEPENDENCY
3020 with TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY.
3021 * config/i386/x86-tune.def
3022 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): New.
3023 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
3025 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
3028 * config/i386/i386-features.c (remove_partial_avx_dependency):
3029 Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
3030 before generating vxorps.
3032 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
3034 * config/i386/i386-options.c (processor_cost_table): Use
3035 tremont_cost for Tremont.
3036 * config/i386/x86-tune-costs.h (tremont_memcpy): New.
3037 (tremont_memset): Likewise.
3038 (tremont_cost): Likewise.
3039 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
3042 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
3044 * common/config/i386/i386-common.c: Use Haswell scheduling model
3046 * config/i386/i386.c (ix86_sched_init_global): Prepare for Tremont
3048 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Tremont
3050 (ix86_adjust_cost): Handle Tremont.
3051 * config/i386/x86-tune.def (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY):
3053 (X86_TUNE_USE_LEAVE): Likewise.
3054 (X86_TUNE_PUSH_MEMORY): Likewise.
3055 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
3056 (X86_TUNE_USE_CLTD): Likewise.
3057 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
3058 (X86_TUNE_AVOID_MFENCE): Likewise.
3059 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
3060 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
3061 (X86_TUNE_ACCUMULATE_OUTGOING_ARGS): Disable for Tremont.
3062 (X86_TUNE_FOUR_JUMP_LIMIT): Likewise.
3063 (X86_TUNE_OPT_AGU): Likewise.
3064 (X86_TUNE_AVOID_LEA_FOR_ADDR): Likewise.
3065 (X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE): Likewise.
3066 (X86_TUNE_EXPAND_ABS): Likewise.
3067 (X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS): Likewise.
3068 (X86_TUNE_SLOW_PSHUFB): Likewise.
3070 2021-09-17 Eric Botcazou <ebotcazou@adacore.com>
3072 PR rtl-optimization/102306
3073 * combine.c (try_combine): Abort the combination if we are about to
3074 duplicate volatile references.
3076 2021-09-17 liuhongt <hongtao.liu@intel.com>
3078 * config/i386/avx512fp16intrin.h (_mm_undefined_ph):
3080 (_mm256_undefined_ph): Likewise.
3081 (_mm512_undefined_ph): Likewise.
3082 (_mm_cvtsh_h): Likewise.
3083 (_mm256_cvtsh_h): Likewise.
3084 (_mm512_cvtsh_h): Likewise.
3085 (_mm512_castph_ps): Likewise.
3086 (_mm512_castph_pd): Likewise.
3087 (_mm512_castph_si512): Likewise.
3088 (_mm512_castph512_ph128): Likewise.
3089 (_mm512_castph512_ph256): Likewise.
3090 (_mm512_castph128_ph512): Likewise.
3091 (_mm512_castph256_ph512): Likewise.
3092 (_mm512_zextph128_ph512): Likewise.
3093 (_mm512_zextph256_ph512): Likewise.
3094 (_mm512_castps_ph): Likewise.
3095 (_mm512_castpd_ph): Likewise.
3096 (_mm512_castsi512_ph): Likewise.
3097 * config/i386/avx512fp16vlintrin.h (_mm_castph_ps):
3099 (_mm256_castph_ps): Likewise.
3100 (_mm_castph_pd): Likewise.
3101 (_mm256_castph_pd): Likewise.
3102 (_mm_castph_si128): Likewise.
3103 (_mm256_castph_si256): Likewise.
3104 (_mm_castps_ph): Likewise.
3105 (_mm256_castps_ph): Likewise.
3106 (_mm_castpd_ph): Likewise.
3107 (_mm256_castpd_ph): Likewise.
3108 (_mm_castsi128_ph): Likewise.
3109 (_mm256_castsi256_ph): Likewise.
3110 (_mm256_castph256_ph128): Likewise.
3111 (_mm256_castph128_ph256): Likewise.
3112 (_mm256_zextph128_ph256): Likewise.
3114 2021-09-17 liuhongt <hongtao.liu@intel.com>
3116 * config/i386/avx512fp16intrin.h (_mm_cvtsh_ss):
3118 (_mm_mask_cvtsh_ss): Likewise.
3119 (_mm_maskz_cvtsh_ss): Likewise.
3120 (_mm_cvtsh_sd): Likewise.
3121 (_mm_mask_cvtsh_sd): Likewise.
3122 (_mm_maskz_cvtsh_sd): Likewise.
3123 (_mm_cvt_roundsh_ss): Likewise.
3124 (_mm_mask_cvt_roundsh_ss): Likewise.
3125 (_mm_maskz_cvt_roundsh_ss): Likewise.
3126 (_mm_cvt_roundsh_sd): Likewise.
3127 (_mm_mask_cvt_roundsh_sd): Likewise.
3128 (_mm_maskz_cvt_roundsh_sd): Likewise.
3129 (_mm_cvtss_sh): Likewise.
3130 (_mm_mask_cvtss_sh): Likewise.
3131 (_mm_maskz_cvtss_sh): Likewise.
3132 (_mm_cvtsd_sh): Likewise.
3133 (_mm_mask_cvtsd_sh): Likewise.
3134 (_mm_maskz_cvtsd_sh): Likewise.
3135 (_mm_cvt_roundss_sh): Likewise.
3136 (_mm_mask_cvt_roundss_sh): Likewise.
3137 (_mm_maskz_cvt_roundss_sh): Likewise.
3138 (_mm_cvt_roundsd_sh): Likewise.
3139 (_mm_mask_cvt_roundsd_sh): Likewise.
3140 (_mm_maskz_cvt_roundsd_sh): Likewise.
3141 * config/i386/i386-builtin-types.def
3142 (V8HF_FTYPE_V2DF_V8HF_V8HF_UQI_INT,
3143 V8HF_FTYPE_V4SF_V8HF_V8HF_UQI_INT,
3144 V2DF_FTYPE_V8HF_V2DF_V2DF_UQI_INT,
3145 V4SF_FTYPE_V8HF_V4SF_V4SF_UQI_INT): Add new builtin types.
3146 * config/i386/i386-builtin.def: Add corrresponding new builtins.
3147 * config/i386/i386-expand.c: Handle new builtin types.
3148 * config/i386/sse.md (VF48_128): New mode iterator.
3149 (avx512fp16_vcvtsh2<ssescalarmodesuffix><mask_scalar_name><round_saeonly_scalar_name>):
3151 (avx512fp16_vcvt<ssescalarmodesuffix>2sh<mask_scalar_name><round_scalar_name>):
3154 2021-09-17 liuhongt <hongtao.liu@intel.com>
3156 * config/i386/avx512fp16intrin.h (_mm512_cvtph_pd):
3158 (_mm512_mask_cvtph_pd): Likewise.
3159 (_mm512_maskz_cvtph_pd): Likewise.
3160 (_mm512_cvt_roundph_pd): Likewise.
3161 (_mm512_mask_cvt_roundph_pd): Likewise.
3162 (_mm512_maskz_cvt_roundph_pd): Likewise.
3163 (_mm512_cvtxph_ps): Likewise.
3164 (_mm512_mask_cvtxph_ps): Likewise.
3165 (_mm512_maskz_cvtxph_ps): Likewise.
3166 (_mm512_cvtx_roundph_ps): Likewise.
3167 (_mm512_mask_cvtx_roundph_ps): Likewise.
3168 (_mm512_maskz_cvtx_roundph_ps): Likewise.
3169 (_mm512_cvtxps_ph): Likewise.
3170 (_mm512_mask_cvtxps_ph): Likewise.
3171 (_mm512_maskz_cvtxps_ph): Likewise.
3172 (_mm512_cvtx_roundps_ph): Likewise.
3173 (_mm512_mask_cvtx_roundps_ph): Likewise.
3174 (_mm512_maskz_cvtx_roundps_ph): Likewise.
3175 (_mm512_cvtpd_ph): Likewise.
3176 (_mm512_mask_cvtpd_ph): Likewise.
3177 (_mm512_maskz_cvtpd_ph): Likewise.
3178 (_mm512_cvt_roundpd_ph): Likewise.
3179 (_mm512_mask_cvt_roundpd_ph): Likewise.
3180 (_mm512_maskz_cvt_roundpd_ph): Likewise.
3181 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_pd):
3183 (_mm_mask_cvtph_pd): Likewise.
3184 (_mm_maskz_cvtph_pd): Likewise.
3185 (_mm256_cvtph_pd): Likewise.
3186 (_mm256_mask_cvtph_pd): Likewise.
3187 (_mm256_maskz_cvtph_pd): Likewise.
3188 (_mm_cvtxph_ps): Likewise.
3189 (_mm_mask_cvtxph_ps): Likewise.
3190 (_mm_maskz_cvtxph_ps): Likewise.
3191 (_mm256_cvtxph_ps): Likewise.
3192 (_mm256_mask_cvtxph_ps): Likewise.
3193 (_mm256_maskz_cvtxph_ps): Likewise.
3194 (_mm_cvtxps_ph): Likewise.
3195 (_mm_mask_cvtxps_ph): Likewise.
3196 (_mm_maskz_cvtxps_ph): Likewise.
3197 (_mm256_cvtxps_ph): Likewise.
3198 (_mm256_mask_cvtxps_ph): Likewise.
3199 (_mm256_maskz_cvtxps_ph): Likewise.
3200 (_mm_cvtpd_ph): Likewise.
3201 (_mm_mask_cvtpd_ph): Likewise.
3202 (_mm_maskz_cvtpd_ph): Likewise.
3203 (_mm256_cvtpd_ph): Likewise.
3204 (_mm256_mask_cvtpd_ph): Likewise.
3205 (_mm256_maskz_cvtpd_ph): Likewise.
3206 * config/i386/i386-builtin.def: Add corresponding new builtins.
3207 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3208 * config/i386/i386-expand.c: Handle new builtin types.
3209 * config/i386/sse.md
3210 (VF4_128_8_256): New.
3211 (VF48H_AVX512VL): Ditto.
3212 (ssePHmode): Add HF vector modes.
3213 (castmode): Add new convertable modes.
3216 (avx512fp16_vcvt<castmode>2ph_<mode><mask_name><round_name>): Ditto.
3217 (avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
3218 (*avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
3219 (avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
3220 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
3221 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask_1): Ditto.
3222 (avx512fp16_float_extend_ph<mode>2<mask_name><round_saeonly_name>):
3224 (avx512fp16_float_extend_ph<mode>2<mask_name>): Ditto.
3225 (*avx512fp16_float_extend_ph<mode>2_load<mask_name>): Ditto.
3226 (avx512fp16_float_extend_phv2df2<mask_name>): Ditto.
3227 (*avx512fp16_float_extend_phv2df2_load<mask_name>): Ditto.
3229 2021-09-17 liuhongt <hongtao.liu@intel.com>
3231 * config/i386/avx512fp16intrin.h (_mm_cvttsh_i32):
3233 (_mm_cvttsh_u32): Likewise.
3234 (_mm_cvtt_roundsh_i32): Likewise.
3235 (_mm_cvtt_roundsh_u32): Likewise.
3236 (_mm_cvttsh_i64): Likewise.
3237 (_mm_cvttsh_u64): Likewise.
3238 (_mm_cvtt_roundsh_i64): Likewise.
3239 (_mm_cvtt_roundsh_u64): Likewise.
3240 * config/i386/i386-builtin.def: Add corresponding new builtins.
3241 * config/i386/sse.md
3242 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<round_saeonly_name>):
3245 2021-09-17 liuhongt <hongtao.liu@intel.com>
3247 * config/i386/avx512fp16intrin.h (_mm512_cvttph_epi32):
3249 (_mm512_mask_cvttph_epi32): Likewise.
3250 (_mm512_maskz_cvttph_epi32): Likewise.
3251 (_mm512_cvtt_roundph_epi32): Likewise.
3252 (_mm512_mask_cvtt_roundph_epi32): Likewise.
3253 (_mm512_maskz_cvtt_roundph_epi32): Likewise.
3254 (_mm512_cvttph_epu32): Likewise.
3255 (_mm512_mask_cvttph_epu32): Likewise.
3256 (_mm512_maskz_cvttph_epu32): Likewise.
3257 (_mm512_cvtt_roundph_epu32): Likewise.
3258 (_mm512_mask_cvtt_roundph_epu32): Likewise.
3259 (_mm512_maskz_cvtt_roundph_epu32): Likewise.
3260 (_mm512_cvttph_epi64): Likewise.
3261 (_mm512_mask_cvttph_epi64): Likewise.
3262 (_mm512_maskz_cvttph_epi64): Likewise.
3263 (_mm512_cvtt_roundph_epi64): Likewise.
3264 (_mm512_mask_cvtt_roundph_epi64): Likewise.
3265 (_mm512_maskz_cvtt_roundph_epi64): Likewise.
3266 (_mm512_cvttph_epu64): Likewise.
3267 (_mm512_mask_cvttph_epu64): Likewise.
3268 (_mm512_maskz_cvttph_epu64): Likewise.
3269 (_mm512_cvtt_roundph_epu64): Likewise.
3270 (_mm512_mask_cvtt_roundph_epu64): Likewise.
3271 (_mm512_maskz_cvtt_roundph_epu64): Likewise.
3272 (_mm512_cvttph_epi16): Likewise.
3273 (_mm512_mask_cvttph_epi16): Likewise.
3274 (_mm512_maskz_cvttph_epi16): Likewise.
3275 (_mm512_cvtt_roundph_epi16): Likewise.
3276 (_mm512_mask_cvtt_roundph_epi16): Likewise.
3277 (_mm512_maskz_cvtt_roundph_epi16): Likewise.
3278 (_mm512_cvttph_epu16): Likewise.
3279 (_mm512_mask_cvttph_epu16): Likewise.
3280 (_mm512_maskz_cvttph_epu16): Likewise.
3281 (_mm512_cvtt_roundph_epu16): Likewise.
3282 (_mm512_mask_cvtt_roundph_epu16): Likewise.
3283 (_mm512_maskz_cvtt_roundph_epu16): Likewise.
3284 * config/i386/avx512fp16vlintrin.h (_mm_cvttph_epi32):
3286 (_mm_mask_cvttph_epi32): Likewise.
3287 (_mm_maskz_cvttph_epi32): Likewise.
3288 (_mm256_cvttph_epi32): Likewise.
3289 (_mm256_mask_cvttph_epi32): Likewise.
3290 (_mm256_maskz_cvttph_epi32): Likewise.
3291 (_mm_cvttph_epu32): Likewise.
3292 (_mm_mask_cvttph_epu32): Likewise.
3293 (_mm_maskz_cvttph_epu32): Likewise.
3294 (_mm256_cvttph_epu32): Likewise.
3295 (_mm256_mask_cvttph_epu32): Likewise.
3296 (_mm256_maskz_cvttph_epu32): Likewise.
3297 (_mm_cvttph_epi64): Likewise.
3298 (_mm_mask_cvttph_epi64): Likewise.
3299 (_mm_maskz_cvttph_epi64): Likewise.
3300 (_mm256_cvttph_epi64): Likewise.
3301 (_mm256_mask_cvttph_epi64): Likewise.
3302 (_mm256_maskz_cvttph_epi64): Likewise.
3303 (_mm_cvttph_epu64): Likewise.
3304 (_mm_mask_cvttph_epu64): Likewise.
3305 (_mm_maskz_cvttph_epu64): Likewise.
3306 (_mm256_cvttph_epu64): Likewise.
3307 (_mm256_mask_cvttph_epu64): Likewise.
3308 (_mm256_maskz_cvttph_epu64): Likewise.
3309 (_mm_cvttph_epi16): Likewise.
3310 (_mm_mask_cvttph_epi16): Likewise.
3311 (_mm_maskz_cvttph_epi16): Likewise.
3312 (_mm256_cvttph_epi16): Likewise.
3313 (_mm256_mask_cvttph_epi16): Likewise.
3314 (_mm256_maskz_cvttph_epi16): Likewise.
3315 (_mm_cvttph_epu16): Likewise.
3316 (_mm_mask_cvttph_epu16): Likewise.
3317 (_mm_maskz_cvttph_epu16): Likewise.
3318 (_mm256_cvttph_epu16): Likewise.
3319 (_mm256_mask_cvttph_epu16): Likewise.
3320 (_mm256_maskz_cvttph_epu16): Likewise.
3321 * config/i386/i386-builtin.def: Add new builtins.
3322 * config/i386/sse.md
3323 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name><round_saeonly_name>):
3325 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name>): Ditto.
3326 (*avx512fp16_fix<fixunssuffix>_trunc<mode>2_load<mask_name>): Ditto.
3327 (avx512fp16_fix<fixunssuffix>_truncv2di2<mask_name>): Ditto.
3328 (avx512fp16_fix<fixunssuffix>_truncv2di2_load<mask_name>): Ditto.
3330 2021-09-17 liuhongt <hongtao.liu@intel.com>
3332 * config/i386/avx512fp16intrin.h (_mm_cvtsh_i32): New intrinsic.
3333 (_mm_cvtsh_u32): Likewise.
3334 (_mm_cvt_roundsh_i32): Likewise.
3335 (_mm_cvt_roundsh_u32): Likewise.
3336 (_mm_cvtsh_i64): Likewise.
3337 (_mm_cvtsh_u64): Likewise.
3338 (_mm_cvt_roundsh_i64): Likewise.
3339 (_mm_cvt_roundsh_u64): Likewise.
3340 (_mm_cvti32_sh): Likewise.
3341 (_mm_cvtu32_sh): Likewise.
3342 (_mm_cvt_roundi32_sh): Likewise.
3343 (_mm_cvt_roundu32_sh): Likewise.
3344 (_mm_cvti64_sh): Likewise.
3345 (_mm_cvtu64_sh): Likewise.
3346 (_mm_cvt_roundi64_sh): Likewise.
3347 (_mm_cvt_roundu64_sh): Likewise.
3348 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3349 * config/i386/i386-builtin.def: Add corresponding new builtins.
3350 * config/i386/i386-expand.c (ix86_expand_round_builtin):
3351 Handle new builtin types.
3352 * config/i386/sse.md
3353 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix><round_name>):
3355 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix>_2): Likewise.
3356 (avx512fp16_vcvt<floatsuffix>si2sh<rex64namesuffix><round_name>): Likewise.
3358 2021-09-16 Bill Schmidt <wschmidt@linux.ibm.com>
3360 * config/rs6000/rs6000-c.c (rs6000-builtins.h): New include.
3361 (altivec_resolve_new_overloaded_builtin): New forward decl.
3362 (rs6000_new_builtin_type_compatible): New function.
3363 (altivec_resolve_overloaded_builtin): Call
3364 altivec_resolve_new_overloaded_builtin.
3365 (altivec_build_new_resolved_builtin): New function.
3366 (altivec_resolve_new_overloaded_builtin): Likewise.
3367 * config/rs6000/rs6000-call.c (rs6000_new_builtin_is_supported):
3369 * config/rs6000/rs6000-gen-builtins.c (write_decls): Remove _p from
3370 name of rs6000_new_builtin_is_supported.
3372 2021-09-16 Uroš Bizjak <ubizjak@gmail.com>
3374 * config/i386/i386-protos.h (ix86_decompose_address):
3375 Change return type to bool.
3376 * config/i386/i386.c (ix86_decompose_address): Ditto.
3378 2021-09-16 Tobias Burnus <tobias@codesourcery.com>
3381 * config/rs6000/t-rs6000 (build/rs6000-gen-builtins.o, build/rbtree.o):
3382 Added 'build/' to target, use build/%.o rule.
3383 (build/rs6000-gen-builtins$(build_exeext)): Add 'build/' and
3384 '$(build_exeext)' to target and 'build/' for the *.o files.
3385 (rs6000-builtins.c): Update for those changes; run rs6000-gen-builtins
3388 2021-09-16 Martin Jambor <mjambor@suse.cz>
3390 * cgraph.c (cgraph_node::dump): Do not check caller count sums if
3391 the body has been removed. Remove trailing whitespace.
3393 2021-09-16 Richard Biener <rguenther@suse.de>
3395 PR middle-end/102360
3396 * internal-fn.c (expand_DEFERRED_INIT): Make pattern-init
3397 of non-memory more robust.
3399 2021-09-16 Daniel Cederman <cederman@gaisler.com>
3401 * config/sparc/sparc-opts.h (enum sparc_processor_type): Add LEON5
3402 * config/sparc/sparc.c (struct processor_costs): Add LEON5 costs
3403 (leon5_adjust_cost): Increase cost of store with data dependency
3404 on ALU instruction and FPU anti-dependencies.
3405 (sparc_option_override): Add LEON5 costs
3406 (sparc_adjust_cost): Add LEON5 cost adjustments
3407 * config/sparc/sparc.h: Add LEON5
3408 * config/sparc/sparc.md: Include LEON5 scheduling information
3409 * config/sparc/sparc.opt: Add LEON5
3410 * doc/invoke.texi: Add LEON5
3411 * config/sparc/leon5.md: New file.
3413 2021-09-16 Daniel Cederman <cederman@gaisler.com>
3415 * config/sparc/sparc.md (stack_protect_set32): Add NOP to prevent
3416 sensitive sequence for B2BST errata workaround.
3418 2021-09-16 Daniel Cederman <cederman@gaisler.com>
3420 * config/sparc/sparc.c (sparc_do_work_around_errata): Do not begin
3421 functions with atomic instruction in the UT700 errata workaround.
3423 2021-09-16 Daniel Cederman <cederman@gaisler.com>
3425 * config/sparc/sparc.c (next_active_non_empty_insn): New function
3426 that returns next active non empty assembly instruction.
3427 (sparc_do_work_around_errata): Use new function.
3429 2021-09-16 Daniel Cederman <cederman@gaisler.com>
3431 * config/sparc/sparc.c (store_insn_p): Add predicate for store
3433 (load_insn_p): Add predicate for load attributes.
3434 (sparc_do_work_around_errata): Use new predicates.
3436 2021-09-16 Andreas Larsson <andreas@gaisler.com>
3438 * config/sparc/sparc.c (dump_target_flag_bits): Print bit names for
3441 2021-09-16 Martin Liska <mliska@suse.cz>
3443 * config/mips/netbsd.h: Fix typo in name of a macro.
3445 2021-09-16 liuhongt <hongtao.liu@intel.com>
3447 PR middle-end/102080
3448 * match.pd: Check mask type when doing cond_op related gimple
3450 * tree.c (is_truth_type_for): New function.
3451 * tree.h (is_truth_type_for): New declaration.
3453 2021-09-16 liuhongt <hongtao.liu@intel.com>
3455 * config/i386/avx512fp16intrin.h (_mm512_cvtepi32_ph): New
3457 (_mm512_mask_cvtepi32_ph): Likewise.
3458 (_mm512_maskz_cvtepi32_ph): Likewise.
3459 (_mm512_cvt_roundepi32_ph): Likewise.
3460 (_mm512_mask_cvt_roundepi32_ph): Likewise.
3461 (_mm512_maskz_cvt_roundepi32_ph): Likewise.
3462 (_mm512_cvtepu32_ph): Likewise.
3463 (_mm512_mask_cvtepu32_ph): Likewise.
3464 (_mm512_maskz_cvtepu32_ph): Likewise.
3465 (_mm512_cvt_roundepu32_ph): Likewise.
3466 (_mm512_mask_cvt_roundepu32_ph): Likewise.
3467 (_mm512_maskz_cvt_roundepu32_ph): Likewise.
3468 (_mm512_cvtepi64_ph): Likewise.
3469 (_mm512_mask_cvtepi64_ph): Likewise.
3470 (_mm512_maskz_cvtepi64_ph): Likewise.
3471 (_mm512_cvt_roundepi64_ph): Likewise.
3472 (_mm512_mask_cvt_roundepi64_ph): Likewise.
3473 (_mm512_maskz_cvt_roundepi64_ph): Likewise.
3474 (_mm512_cvtepu64_ph): Likewise.
3475 (_mm512_mask_cvtepu64_ph): Likewise.
3476 (_mm512_maskz_cvtepu64_ph): Likewise.
3477 (_mm512_cvt_roundepu64_ph): Likewise.
3478 (_mm512_mask_cvt_roundepu64_ph): Likewise.
3479 (_mm512_maskz_cvt_roundepu64_ph): Likewise.
3480 (_mm512_cvtepi16_ph): Likewise.
3481 (_mm512_mask_cvtepi16_ph): Likewise.
3482 (_mm512_maskz_cvtepi16_ph): Likewise.
3483 (_mm512_cvt_roundepi16_ph): Likewise.
3484 (_mm512_mask_cvt_roundepi16_ph): Likewise.
3485 (_mm512_maskz_cvt_roundepi16_ph): Likewise.
3486 (_mm512_cvtepu16_ph): Likewise.
3487 (_mm512_mask_cvtepu16_ph): Likewise.
3488 (_mm512_maskz_cvtepu16_ph): Likewise.
3489 (_mm512_cvt_roundepu16_ph): Likewise.
3490 (_mm512_mask_cvt_roundepu16_ph): Likewise.
3491 (_mm512_maskz_cvt_roundepu16_ph): Likewise.
3492 * config/i386/avx512fp16vlintrin.h (_mm_cvtepi32_ph): New
3494 (_mm_mask_cvtepi32_ph): Likewise.
3495 (_mm_maskz_cvtepi32_ph): Likewise.
3496 (_mm256_cvtepi32_ph): Likewise.
3497 (_mm256_mask_cvtepi32_ph): Likewise.
3498 (_mm256_maskz_cvtepi32_ph): Likewise.
3499 (_mm_cvtepu32_ph): Likewise.
3500 (_mm_mask_cvtepu32_ph): Likewise.
3501 (_mm_maskz_cvtepu32_ph): Likewise.
3502 (_mm256_cvtepu32_ph): Likewise.
3503 (_mm256_mask_cvtepu32_ph): Likewise.
3504 (_mm256_maskz_cvtepu32_ph): Likewise.
3505 (_mm_cvtepi64_ph): Likewise.
3506 (_mm_mask_cvtepi64_ph): Likewise.
3507 (_mm_maskz_cvtepi64_ph): Likewise.
3508 (_mm256_cvtepi64_ph): Likewise.
3509 (_mm256_mask_cvtepi64_ph): Likewise.
3510 (_mm256_maskz_cvtepi64_ph): Likewise.
3511 (_mm_cvtepu64_ph): Likewise.
3512 (_mm_mask_cvtepu64_ph): Likewise.
3513 (_mm_maskz_cvtepu64_ph): Likewise.
3514 (_mm256_cvtepu64_ph): Likewise.
3515 (_mm256_mask_cvtepu64_ph): Likewise.
3516 (_mm256_maskz_cvtepu64_ph): Likewise.
3517 (_mm_cvtepi16_ph): Likewise.
3518 (_mm_mask_cvtepi16_ph): Likewise.
3519 (_mm_maskz_cvtepi16_ph): Likewise.
3520 (_mm256_cvtepi16_ph): Likewise.
3521 (_mm256_mask_cvtepi16_ph): Likewise.
3522 (_mm256_maskz_cvtepi16_ph): Likewise.
3523 (_mm_cvtepu16_ph): Likewise.
3524 (_mm_mask_cvtepu16_ph): Likewise.
3525 (_mm_maskz_cvtepu16_ph): Likewise.
3526 (_mm256_cvtepu16_ph): Likewise.
3527 (_mm256_mask_cvtepu16_ph): Likewise.
3528 (_mm256_maskz_cvtepu16_ph): Likewise.
3529 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3530 * config/i386/i386-builtin.def: Add corresponding new builtins.
3531 * config/i386/i386-expand.c
3532 (ix86_expand_args_builtin): Handle new builtin types.
3533 (ix86_expand_round_builtin): Ditto.
3534 * config/i386/i386-modes.def: Declare V2HF and V6HF.
3535 * config/i386/sse.md (VI2H_AVX512VL): New.
3537 (sseintvecmode): Add HF vector modes.
3538 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode><mask_name><round_name>):
3540 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
3541 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
3542 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
3543 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
3544 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask_1): Ditto.
3545 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
3546 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
3547 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
3548 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
3549 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask_1): Ditto.
3550 * config/i386/subst.md (round_qq2phsuff): New subst_attr.
3552 2021-09-16 liuhongt <hongtao.liu@intel.com>
3554 * config/i386/avx512fp16intrin.h (_mm512_cvtph_epi32):
3556 (_mm512_mask_cvtph_epi32): Likewise.
3557 (_mm512_maskz_cvtph_epi32): Likewise.
3558 (_mm512_cvt_roundph_epi32): Likewise.
3559 (_mm512_mask_cvt_roundph_epi32): Likewise.
3560 (_mm512_maskz_cvt_roundph_epi32): Likewise.
3561 (_mm512_cvtph_epu32): Likewise.
3562 (_mm512_mask_cvtph_epu32): Likewise.
3563 (_mm512_maskz_cvtph_epu32): Likewise.
3564 (_mm512_cvt_roundph_epu32): Likewise.
3565 (_mm512_mask_cvt_roundph_epu32): Likewise.
3566 (_mm512_maskz_cvt_roundph_epu32): Likewise.
3567 (_mm512_cvtph_epi64): Likewise.
3568 (_mm512_mask_cvtph_epi64): Likewise.
3569 (_mm512_maskz_cvtph_epi64): Likewise.
3570 (_mm512_cvt_roundph_epi64): Likewise.
3571 (_mm512_mask_cvt_roundph_epi64): Likewise.
3572 (_mm512_maskz_cvt_roundph_epi64): Likewise.
3573 (_mm512_cvtph_epu64): Likewise.
3574 (_mm512_mask_cvtph_epu64): Likewise.
3575 (_mm512_maskz_cvtph_epu64): Likewise.
3576 (_mm512_cvt_roundph_epu64): Likewise.
3577 (_mm512_mask_cvt_roundph_epu64): Likewise.
3578 (_mm512_maskz_cvt_roundph_epu64): Likewise.
3579 (_mm512_cvtph_epi16): Likewise.
3580 (_mm512_mask_cvtph_epi16): Likewise.
3581 (_mm512_maskz_cvtph_epi16): Likewise.
3582 (_mm512_cvt_roundph_epi16): Likewise.
3583 (_mm512_mask_cvt_roundph_epi16): Likewise.
3584 (_mm512_maskz_cvt_roundph_epi16): Likewise.
3585 (_mm512_cvtph_epu16): Likewise.
3586 (_mm512_mask_cvtph_epu16): Likewise.
3587 (_mm512_maskz_cvtph_epu16): Likewise.
3588 (_mm512_cvt_roundph_epu16): Likewise.
3589 (_mm512_mask_cvt_roundph_epu16): Likewise.
3590 (_mm512_maskz_cvt_roundph_epu16): Likewise.
3591 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_epi32):
3593 (_mm_mask_cvtph_epi32): Likewise.
3594 (_mm_maskz_cvtph_epi32): Likewise.
3595 (_mm256_cvtph_epi32): Likewise.
3596 (_mm256_mask_cvtph_epi32): Likewise.
3597 (_mm256_maskz_cvtph_epi32): Likewise.
3598 (_mm_cvtph_epu32): Likewise.
3599 (_mm_mask_cvtph_epu32): Likewise.
3600 (_mm_maskz_cvtph_epu32): Likewise.
3601 (_mm256_cvtph_epu32): Likewise.
3602 (_mm256_mask_cvtph_epu32): Likewise.
3603 (_mm256_maskz_cvtph_epu32): Likewise.
3604 (_mm_cvtph_epi64): Likewise.
3605 (_mm_mask_cvtph_epi64): Likewise.
3606 (_mm_maskz_cvtph_epi64): Likewise.
3607 (_mm256_cvtph_epi64): Likewise.
3608 (_mm256_mask_cvtph_epi64): Likewise.
3609 (_mm256_maskz_cvtph_epi64): Likewise.
3610 (_mm_cvtph_epu64): Likewise.
3611 (_mm_mask_cvtph_epu64): Likewise.
3612 (_mm_maskz_cvtph_epu64): Likewise.
3613 (_mm256_cvtph_epu64): Likewise.
3614 (_mm256_mask_cvtph_epu64): Likewise.
3615 (_mm256_maskz_cvtph_epu64): Likewise.
3616 (_mm_cvtph_epi16): Likewise.
3617 (_mm_mask_cvtph_epi16): Likewise.
3618 (_mm_maskz_cvtph_epi16): Likewise.
3619 (_mm256_cvtph_epi16): Likewise.
3620 (_mm256_mask_cvtph_epi16): Likewise.
3621 (_mm256_maskz_cvtph_epi16): Likewise.
3622 (_mm_cvtph_epu16): Likewise.
3623 (_mm_mask_cvtph_epu16): Likewise.
3624 (_mm_maskz_cvtph_epu16): Likewise.
3625 (_mm256_cvtph_epu16): Likewise.
3626 (_mm256_mask_cvtph_epu16): Likewise.
3627 (_mm256_maskz_cvtph_epu16): Likewise.
3628 * config/i386/i386-builtin-types.def: Add new builtin types.
3629 * config/i386/i386-builtin.def: Add new builtins.
3630 * config/i386/i386-expand.c
3631 (ix86_expand_args_builtin): Handle new builtin types.
3632 (ix86_expand_round_builtin): Ditto.
3633 * config/i386/sse.md (sseintconvert): New.
3635 (UNSPEC_US_FIX_NOTRUNC): Ditto.
3636 (sseintconvertsignprefix): Ditto.
3637 (avx512fp16_vcvtph2<sseintconvertsignprefix><sseintconvert>_<mode><mask_name><round_name>):
3640 2021-09-16 liuhongt <hongtao.liu@intel.com>
3642 * config/i386/avx512fp16intrin.h: (_mm_cvtsi16_si128):
3644 (_mm_cvtsi128_si16): Likewise.
3645 (_mm_mask_load_sh): Likewise.
3646 (_mm_maskz_load_sh): Likewise.
3647 (_mm_mask_store_sh): Likewise.
3648 (_mm_move_sh): Likewise.
3649 (_mm_mask_move_sh): Likewise.
3650 (_mm_maskz_move_sh): Likewise.
3651 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3652 * config/i386/i386-builtin.def: Add corresponding new builtins.
3653 * config/i386/i386-expand.c
3654 (ix86_expand_special_args_builtin): Handle new builtin types.
3655 (ix86_expand_vector_init_one_nonzero): Adjust for FP16 target.
3656 * config/i386/sse.md (VI2F): New mode iterator.
3657 (vec_set<mode>_0): Use new mode iterator.
3658 (avx512f_mov<ssescalarmodelower>_mask): Adjust for HF vector mode.
3659 (avx512f_store<mode>_mask): Ditto.
3661 2021-09-16 Kewen Lin <linkw@linux.ibm.com>
3663 * config/rs6000/rs6000.opt (-mtoc-fusion): Remove.
3665 2021-09-15 David Edelsohn <dje.gcc@gmail.com>
3667 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
3668 Proceed if no symbol summary or the symbol alias flag is false.
3670 2021-09-15 Jakub Jelinek <jakub@redhat.com>
3674 * varasm.c (output_constructor_regular_field): Instead of assertion
3675 that array_size_for_constructor result is equal to size of
3676 TREE_TYPE (local->val) in bytes, assert that the type size is greater
3677 or equal to array_size_for_constructor result and use type size as
3680 2021-09-15 Martin Liska <mliska@suse.cz>
3683 * config/i386/vxworks.h: Use new macro TARGET_CPU_P.
3685 2021-09-15 Martin Liska <mliska@suse.cz>
3688 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
3689 Check that we have a symbol summary for a symbol.
3691 2021-09-15 Richard Biener <rguenther@suse.de>
3694 * config/rs6000/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
3695 to inherit from elfos.h
3697 2021-09-15 liuhongt <hongtao.liu@intel.com>
3700 * config/i386/i386-expand.c
3701 (ix86_expand_vector_init_interleave): Use puncklwd to pack 2
3703 (ix86_expand_vector_set): Use blendw instead of pinsrw.
3704 * config/i386/i386.c (ix86_can_change_mode_class): Adjust for
3705 AVX512FP16 which supports 16bit vector load.
3706 * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>):
3708 (avx512bw_interleave_high<mode><mask_name>): .. this, and
3709 extend to V32HFmode.
3710 (avx2_interleave_highv16hi<mask_name>): Rename to ..
3711 (avx2_interleave_high<mode><mask_name>): .. this, and extend
3713 (vec_interleave_highv8hi<mask_name>): Rename to ..
3714 (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode.
3715 (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>):
3717 (<mask_codefor>avx512bw_interleave_low<mode><mask_name>):
3718 this, and extend to V32HFmode.
3719 (avx2_interleave_lowv16hi<mask_name>): Rename to ..
3720 (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode.
3721 (vec_interleave_lowv8hi<mask_name>): Rename to ..
3722 (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode.
3723 (sse4_1_pblendw): Rename to ..
3724 (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode.
3725 (avx2_pblendph): New define_expand.
3726 (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use
3727 sseintmodesuffix instead of ssemodesuffix.
3728 (blendsuf): New mode attr.
3730 2021-09-15 Richard Biener <rguenther@suse.de>
3732 * tree-vectorizer.h (dr_misalignment): Move out of line.
3733 (dr_target_alignment): New.
3734 (DR_TARGET_ALIGNMENT): Wrap dr_target_alignment.
3735 (set_dr_target_alignment): New.
3736 (SET_DR_TARGET_ALIGNMENT): Wrap set_dr_target_alignment.
3737 * tree-vect-data-refs.c (dr_misalignment): Compute and
3738 return the group members misalignment.
3739 (vect_compute_data_ref_alignment): Use SET_DR_TARGET_ALIGNMENT.
3740 (vect_analyze_data_refs_alignment): Compute alignment only
3741 for the first element of a DR group.
3742 (vect_slp_analyze_node_alignment): Likewise.
3744 2021-09-15 Hongyu Wang <hongyu.wang@intel.com>
3746 * config/i386/avx512fp16intrin.h: Adjust all builtin calls.
3747 * config/i386/avx512fp16vlintrin.h: Likewise.
3748 * config/i386/i386-builtin.def: Adjust builtin name and
3749 enumeration to match AVX512F style.
3751 2021-09-15 Richard Biener <rguenther@suse.de>
3753 PR tree-optimization/102318
3754 * tree-vect-loop.c (vect_transform_cycle_phi): Revert
3755 previous change and do the mode conversion separately from
3756 the sign conversion.
3758 2021-09-15 Hongtao Liu <hongtao.liu@intel.com>
3759 Peter Cordes <peter@cordes.ca>
3762 * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
3763 (*vec_extract<mode><ssescalarmodelower>_valign): Output
3764 vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
3767 2021-09-15 Richard Biener <rguenther@suse.de>
3769 * config.gcc: Remove vax-*-openbsd* configuration.
3771 2021-09-15 Richard Biener <rguenther@suse.de>
3773 * config.gcc: Remove m68k-openbsd.
3775 2021-09-15 Max Filippov <jcmvbkbc@gmail.com>
3778 * config/xtensa/t-xtensa (TM_H): Add include/xtensa-config.h.
3780 2021-09-14 Peter Bergner <bergner@linux.ibm.com>
3782 * config/rs6000/mma.md (unspec): Delete UNSPEC_MMA_XXSETACCZ.
3783 (unspecv): Add UNSPECV_MMA_XXSETACCZ.
3784 (*mma_xxsetaccz): Delete.
3785 (mma_xxsetaccz): Change to define_insn. Remove operand 1.
3786 Use UNSPECV_MMA_XXSETACCZ. Update comment.
3787 * config/rs6000/rs6000.c (rs6000_rtx_costs): Use UNSPECV_MMA_XXSETACCZ.
3789 2021-09-14 Iain Sandoe <iain@sandoe.co.uk>
3791 * Makefile.in: Remove variables related to applying no-PIE
3792 to the exes on $build.
3793 * configure: Regenerate.
3794 * configure.ac: Remove configuration related to applying
3795 no-PIE to the exes on $build.
3797 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
3799 * config/arc/arc.md (doloop_end): Add missing mode.
3800 (loop_end): Likewise.
3802 2021-09-14 Jakub Jelinek <jakub@redhat.com>
3804 * gimplify.c (goa_stabilize_expr): Add depth argument, propagate
3805 it to recursive calls, for depth above 7 just gimplify or return.
3806 Perform a test even for MODIFY_EXPR, ADDR_EXPR, COMPOUND_EXPR with
3807 __builtin_clear_padding and TARGET_EXPR.
3808 (gimplify_omp_atomic): Adjust goa_stabilize_expr callers.
3810 2021-09-14 liuhongt <hongtao.liu@intel.com>
3812 * config/i386/avx512fp16intrin.h (_mm_fpclass_sh_mask):
3814 (_mm_mask_fpclass_sh_mask): Likewise.
3815 (_mm512_mask_fpclass_ph_mask): Likewise.
3816 (_mm512_fpclass_ph_mask): Likewise.
3817 (_mm_getexp_sh): Likewise.
3818 (_mm_mask_getexp_sh): Likewise.
3819 (_mm_maskz_getexp_sh): Likewise.
3820 (_mm512_getexp_ph): Likewise.
3821 (_mm512_mask_getexp_ph): Likewise.
3822 (_mm512_maskz_getexp_ph): Likewise.
3823 (_mm_getexp_round_sh): Likewise.
3824 (_mm_mask_getexp_round_sh): Likewise.
3825 (_mm_maskz_getexp_round_sh): Likewise.
3826 (_mm512_getexp_round_ph): Likewise.
3827 (_mm512_mask_getexp_round_ph): Likewise.
3828 (_mm512_maskz_getexp_round_ph): Likewise.
3829 (_mm_getmant_sh): Likewise.
3830 (_mm_mask_getmant_sh): Likewise.
3831 (_mm_maskz_getmant_sh): Likewise.
3832 (_mm512_getmant_ph): Likewise.
3833 (_mm512_mask_getmant_ph): Likewise.
3834 (_mm512_maskz_getmant_ph): Likewise.
3835 (_mm_getmant_round_sh): Likewise.
3836 (_mm_mask_getmant_round_sh): Likewise.
3837 (_mm_maskz_getmant_round_sh): Likewise.
3838 (_mm512_getmant_round_ph): Likewise.
3839 (_mm512_mask_getmant_round_ph): Likewise.
3840 (_mm512_maskz_getmant_round_ph): Likewise.
3841 * config/i386/avx512fp16vlintrin.h (_mm_mask_fpclass_ph_mask):
3843 (_mm_fpclass_ph_mask): Likewise.
3844 (_mm256_mask_fpclass_ph_mask): Likewise.
3845 (_mm256_fpclass_ph_mask): Likewise.
3846 (_mm256_getexp_ph): Likewise.
3847 (_mm256_mask_getexp_ph): Likewise.
3848 (_mm256_maskz_getexp_ph): Likewise.
3849 (_mm_getexp_ph): Likewise.
3850 (_mm_mask_getexp_ph): Likewise.
3851 (_mm_maskz_getexp_ph): Likewise.
3852 (_mm256_getmant_ph): Likewise.
3853 (_mm256_mask_getmant_ph): Likewise.
3854 (_mm256_maskz_getmant_ph): Likewise.
3855 (_mm_getmant_ph): Likewise.
3856 (_mm_mask_getmant_ph): Likewise.
3857 (_mm_maskz_getmant_ph): Likewise.
3858 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3859 * config/i386/i386-builtin.def: Add corresponding new builtins.
3860 * config/i386/i386-expand.c
3861 (ix86_expand_args_builtin): Handle new builtin types.
3862 (ix86_expand_round_builtin): Ditto.
3863 * config/i386/sse.md (vecmemsuffix): Add HF vector modes.
3864 (<avx512>_getexp<mode><mask_name><round_saeonly_name>): Adjust
3865 to support HF vector modes.
3866 (avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name):
3868 (avx512dq_fpclass<mode><mask_scalar_merge_name>): Ditto.
3869 (avx512dq_vmfpclass<mode><mask_scalar_merge_name>): Ditto.
3870 (<avx512>_getmant<mode><mask_name><round_saeonly_name>): Ditto.
3871 (avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3874 2021-09-14 liuhongt <hongtao.liu@intel.com>
3876 * config/i386/avx512fp16intrin.h (_mm512_reduce_ph):
3878 (_mm512_mask_reduce_ph): Likewise.
3879 (_mm512_maskz_reduce_ph): Likewise.
3880 (_mm512_reduce_round_ph): Likewise.
3881 (_mm512_mask_reduce_round_ph): Likewise.
3882 (_mm512_maskz_reduce_round_ph): Likewise.
3883 (_mm_reduce_sh): Likewise.
3884 (_mm_mask_reduce_sh): Likewise.
3885 (_mm_maskz_reduce_sh): Likewise.
3886 (_mm_reduce_round_sh): Likewise.
3887 (_mm_mask_reduce_round_sh): Likewise.
3888 (_mm_maskz_reduce_round_sh): Likewise.
3889 (_mm512_roundscale_ph): Likewise.
3890 (_mm512_mask_roundscale_ph): Likewise.
3891 (_mm512_maskz_roundscale_ph): Likewise.
3892 (_mm512_roundscale_round_ph): Likewise.
3893 (_mm512_mask_roundscale_round_ph): Likewise.
3894 (_mm512_maskz_roundscale_round_ph): Likewise.
3895 (_mm_roundscale_sh): Likewise.
3896 (_mm_mask_roundscale_sh): Likewise.
3897 (_mm_maskz_roundscale_sh): Likewise.
3898 (_mm_roundscale_round_sh): Likewise.
3899 (_mm_mask_roundscale_round_sh): Likewise.
3900 (_mm_maskz_roundscale_round_sh): Likewise.
3901 * config/i386/avx512fp16vlintrin.h: (_mm_reduce_ph):
3903 (_mm_mask_reduce_ph): Likewise.
3904 (_mm_maskz_reduce_ph): Likewise.
3905 (_mm256_reduce_ph): Likewise.
3906 (_mm256_mask_reduce_ph): Likewise.
3907 (_mm256_maskz_reduce_ph): Likewise.
3908 (_mm_roundscale_ph): Likewise.
3909 (_mm_mask_roundscale_ph): Likewise.
3910 (_mm_maskz_roundscale_ph): Likewise.
3911 (_mm256_roundscale_ph): Likewise.
3912 (_mm256_mask_roundscale_ph): Likewise.
3913 (_mm256_maskz_roundscale_ph): Likewise.
3914 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3915 * config/i386/i386-builtin.def: Add corresponding new builtins.
3916 * config/i386/i386-expand.c
3917 (ix86_expand_args_builtin): Handle new builtin types.
3918 (ix86_expand_round_builtin): Ditto.
3919 * config/i386/sse.md (<mask_codefor>reducep<mode><mask_name>):
3921 (<mask_codefor>reducep<mode><mask_name><round_saeonly_name>):
3922 ... this, and adjust for round operands.
3923 (reduces<mode><mask_scalar_name>): Likewise, with ...
3924 (reduces<mode><mask_scalar_name><round_saeonly_scalar_name):
3926 (<avx512>_rndscale<mode><mask_name><round_saeonly_name>):
3927 Adjust for HF vector modes.
3928 (avx512f_rndscale<mode><mask_scalar_name><round_saeonly_scalar_name>):
3930 (*avx512f_rndscale<mode><round_saeonly_name>): Ditto.
3932 2021-09-14 liuhongt <hongtao.liu@intel.com>
3934 * config/i386/avx512fp16intrin.h: (_mm512_rcp_ph):
3936 (_mm512_mask_rcp_ph): Likewise.
3937 (_mm512_maskz_rcp_ph): Likewise.
3938 (_mm_rcp_sh): Likewise.
3939 (_mm_mask_rcp_sh): Likewise.
3940 (_mm_maskz_rcp_sh): Likewise.
3941 (_mm512_scalef_ph): Likewise.
3942 (_mm512_mask_scalef_ph): Likewise.
3943 (_mm512_maskz_scalef_ph): Likewise.
3944 (_mm512_scalef_round_ph): Likewise.
3945 (_mm512_mask_scalef_round_ph): Likewise.
3946 (_mm512_maskz_scalef_round_ph): Likewise.
3947 (_mm_scalef_sh): Likewise.
3948 (_mm_mask_scalef_sh): Likewise.
3949 (_mm_maskz_scalef_sh): Likewise.
3950 (_mm_scalef_round_sh): Likewise.
3951 (_mm_mask_scalef_round_sh): Likewise.
3952 (_mm_maskz_scalef_round_sh): Likewise.
3953 * config/i386/avx512fp16vlintrin.h (_mm_rcp_ph):
3955 (_mm256_rcp_ph): Likewise.
3956 (_mm_mask_rcp_ph): Likewise.
3957 (_mm256_mask_rcp_ph): Likewise.
3958 (_mm_maskz_rcp_ph): Likewise.
3959 (_mm256_maskz_rcp_ph): Likewise.
3960 (_mm_scalef_ph): Likewise.
3961 (_mm256_scalef_ph): Likewise.
3962 (_mm_mask_scalef_ph): Likewise.
3963 (_mm256_mask_scalef_ph): Likewise.
3964 (_mm_maskz_scalef_ph): Likewise.
3965 (_mm256_maskz_scalef_ph): Likewise.
3966 * config/i386/i386-builtin.def: Add new builtins.
3967 * config/i386/sse.md (VFH_AVX512VL): New.
3968 (avx512fp16_rcp<mode>2<mask_name>): Ditto.
3969 (avx512fp16_vmrcpv8hf2<mask_scalar_name>): Ditto.
3970 (avx512f_vmscalef<mode><mask_scalar_name><round_scalar_name>):
3971 Adjust to support HF vector modes.
3972 (<avx512>_scalef<mode><mask_name><round_name>): Ditto.
3974 2021-09-14 liuhongt <hongtao.liu@intel.com>
3976 * config/i386/avx512fp16intrin.h: (_mm512_sqrt_ph):
3978 (_mm512_mask_sqrt_ph): Likewise.
3979 (_mm512_maskz_sqrt_ph): Likewise.
3980 (_mm512_sqrt_round_ph): Likewise.
3981 (_mm512_mask_sqrt_round_ph): Likewise.
3982 (_mm512_maskz_sqrt_round_ph): Likewise.
3983 (_mm512_rsqrt_ph): Likewise.
3984 (_mm512_mask_rsqrt_ph): Likewise.
3985 (_mm512_maskz_rsqrt_ph): Likewise.
3986 (_mm_rsqrt_sh): Likewise.
3987 (_mm_mask_rsqrt_sh): Likewise.
3988 (_mm_maskz_rsqrt_sh): Likewise.
3989 (_mm_sqrt_sh): Likewise.
3990 (_mm_mask_sqrt_sh): Likewise.
3991 (_mm_maskz_sqrt_sh): Likewise.
3992 (_mm_sqrt_round_sh): Likewise.
3993 (_mm_mask_sqrt_round_sh): Likewise.
3994 (_mm_maskz_sqrt_round_sh): Likewise.
3995 * config/i386/avx512fp16vlintrin.h (_mm_sqrt_ph): New intrinsic.
3996 (_mm256_sqrt_ph): Likewise.
3997 (_mm_mask_sqrt_ph): Likewise.
3998 (_mm256_mask_sqrt_ph): Likewise.
3999 (_mm_maskz_sqrt_ph): Likewise.
4000 (_mm256_maskz_sqrt_ph): Likewise.
4001 (_mm_rsqrt_ph): Likewise.
4002 (_mm256_rsqrt_ph): Likewise.
4003 (_mm_mask_rsqrt_ph): Likewise.
4004 (_mm256_mask_rsqrt_ph): Likewise.
4005 (_mm_maskz_rsqrt_ph): Likewise.
4006 (_mm256_maskz_rsqrt_ph): Likewise.
4007 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
4008 * config/i386/i386-builtin.def: Add corresponding new builtins.
4009 * config/i386/i386-expand.c
4010 (ix86_expand_args_builtin): Handle new builtins.
4011 (ix86_expand_round_builtin): Ditto.
4012 * config/i386/sse.md (VF_AVX512FP16VL): New.
4013 (sqrt<mode>2): Adjust for HF vector modes.
4014 (<sse>_sqrt<mode>2<mask_name><round_name>): Likewise.
4015 (<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
4017 (<sse>_rsqrt<mode>2<mask_name>): New.
4018 (avx512fp16_vmrsqrtv8hf2<mask_scalar_name>): Likewise.
4020 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
4023 * diagnostic-spec.c (warning_suppressed_at, copy_warning): Handle
4024 'RESERVED_LOCATION_P' locations.
4025 * warning-control.cc (get_nowarn_spec, suppress_warning)
4026 (copy_warning): Likewise.
4028 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
4030 * diagnostic-spec.h (typedef xint_hash_t): Use 'location_t' instead of...
4031 (typedef key_type_t): ... this. Remove.
4032 (nowarn_map): Document.
4033 * diagnostic-spec.c (nowarn_map): Likewise.
4034 * warning-control.cc (convert_to_key): Evolve functions into...
4035 (get_location): ... these. Adjust all users.
4037 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
4039 * warning-control.cc (copy_warning): Remove 'nowarn_map' setup.
4041 2021-09-13 Jason Merrill <jason@redhat.com>
4043 * params.opt: Add destructive-interference-size and
4044 constructive-interference-size.
4045 * doc/invoke.texi: Document them.
4046 * config/aarch64/aarch64.c (aarch64_override_options_internal):
4048 * config/arm/arm.c (arm_option_override): Set them.
4049 * config/i386/i386-options.c (ix86_option_override_internal):
4052 2021-09-13 Martin Liska <mliska@suse.cz>
4053 H.J. Lu <hjl.tools@gmail.com>
4056 * common/config/i386/cpuinfo.h (cpu_indicator_init): Add support
4057 for x86-64 micro levels for __builtin_cpu_supports.
4058 * common/config/i386/i386-cpuinfo.h (enum feature_priority):
4059 Add priorities for the micro-arch levels.
4060 (enum processor_features): Add new features.
4061 * common/config/i386/i386-isas.h: Add micro-arch features.
4062 * config/i386/i386-builtins.c (get_builtin_code_for_version):
4063 Support the micro-arch levels by callsing
4064 __builtin_cpu_supports.
4065 * doc/extend.texi: Document that the levels are support by
4066 __builtin_cpu_supports.
4068 2021-09-13 Andrew Pinski <apinski@marvell.com>
4071 * config/aarch64/aarch64-builtins.c (aarch64_fold_builtin_lane_check):
4073 (aarch64_general_fold_builtin): Handle AARCH64_SIMD_BUILTIN_LANE_CHECK.
4074 (aarch64_general_gimple_fold_builtin): Likewise.
4076 2021-09-13 Andrew Pinski <apinski@marvell.com>
4078 * config.gcc: Add m32r-*-linux* and m32rle-*-linux*
4079 to the Unsupported targets list.
4080 Remove support for m32r-*-linux* and m32rle-*-linux*.
4081 * config/m32r/linux.h: Removed.
4082 * config/m32r/t-linux: Removed.
4084 2021-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4087 * config/aarch64/aarch64.c (aarch64_classify_address): Don't allow
4088 register index for SVE predicate modes.
4090 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
4092 * tree-ssa-threadbackward.c
4093 (back_threader_profitability::profitable_path_p): Remove FSM
4095 (back_threader_registry::register_path): Same.
4096 * tree-ssa-threadedge.c
4097 (jump_threader::simplify_control_stmt_condition): Same.
4098 * tree-ssa-threadupdate.c (jt_path_registry::jt_path_registry):
4099 Add backedge_threads argument.
4100 (fwd_jt_path_registry::fwd_jt_path_registry): Pass
4101 backedge_threads argument.
4102 (back_jt_path_registry::back_jt_path_registry): Same.
4103 (dump_jump_thread_path): Adjust for FSM removal.
4104 (back_jt_path_registry::rewire_first_differing_edge): Same.
4105 (back_jt_path_registry::adjust_paths_after_duplication): Same.
4106 (back_jt_path_registry::update_cfg): Same.
4107 (jt_path_registry::register_jump_thread): Same.
4108 * tree-ssa-threadupdate.h (enum jump_thread_edge_type): Remove
4110 (class back_jt_path_registry): Add backedge_threads to
4113 2021-09-13 Martin Liska <mliska@suse.cz>
4116 * asan.h (sanitize_coverage_p): Handle when fn == NULL.
4118 2021-09-13 H.J. Lu <hjl.tools@gmail.com>
4121 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): New.
4122 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
4123 (MOVE_MAX): Check TARGET_AVX256_MOVE_BY_PIECES and
4124 TARGET_AVX256_STORE_BY_PIECES instead of
4125 TARGET_AVX256_SPLIT_UNALIGNED_LOAD and
4126 TARGET_AVX256_SPLIT_UNALIGNED_STORE.
4127 (STORE_MAX_PIECES): Check TARGET_AVX256_STORE_BY_PIECES instead
4128 of TARGET_AVX256_SPLIT_UNALIGNED_STORE.
4129 * config/i386/x86-tune.def (X86_TUNE_AVX256_MOVE_BY_PIECES): New.
4130 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
4132 2021-09-13 liuhongt <hongtao.liu@intel.com>
4135 * expmed.c (extract_bit_field_using_extv): Use
4136 gen_lowpart_if_possible instead of gen_lowpart to avoid ICE.
4138 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
4140 * Makefile.in (OBJS): Add value-pointer-equiv.o.
4141 * gimple-ssa-evrp.c (class ssa_equiv_stack): Move to
4142 value-pointer-equiv.*.
4143 (ssa_equiv_stack::ssa_equiv_stack): Same.
4144 (ssa_equiv_stack::enter): Same.
4145 (ssa_equiv_stack::leave): Same.
4146 (ssa_equiv_stack::push_replacement): Same.
4147 (ssa_equiv_stack::get_replacement): Same.
4148 (is_pointer_ssa): Same.
4149 (class pointer_equiv_analyzer): Same.
4150 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
4151 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Same.
4152 (pointer_equiv_analyzer::set_global_equiv): Same.
4153 (pointer_equiv_analyzer::set_cond_equiv): Same.
4154 (pointer_equiv_analyzer::get_equiv): Same.
4155 (pointer_equiv_analyzer::enter): Same.
4156 (pointer_equiv_analyzer::leave): Same.
4157 (pointer_equiv_analyzer::get_equiv_expr): Same.
4158 (pta_valueize): Same.
4159 (pointer_equiv_analyzer::visit_stmt): Same.
4160 (pointer_equiv_analyzer::visit_edge): Same.
4161 (hybrid_folder::value_of_expr): Same.
4162 (hybrid_folder::value_on_edge): Same.
4163 * value-pointer-equiv.cc: New file.
4164 * value-pointer-equiv.h: New file.
4166 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
4169 * gimple-fold.c (gimple_fold_builtin_memory_op): Allow folding
4170 memcpy if the size is not more than MOVE_MAX * MOVE_RATIO.
4172 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
4175 * config/arm/arm.md (movmisaligndi): New define_expand.
4176 * config/arm/vec-common.md (movmisalign<mode>): Iterate over VDQ mode.
4178 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
4181 * emit-rtl.c (gen_highpart): Use adjust_address to handle
4182 MEM rather than calling simplify_gen_subreg.
4184 2021-09-13 Jan-Benedict Glaw <jbglaw@ług-owl.de>
4186 * config/alpha/vms.h (INIT_CUMULATIVE_ARGS): Wrap multi-statment
4187 define into a block.
4189 2021-09-13 Richard Biener <rguenther@suse.de>
4191 * config/darwin.h (DARWIN_PREFER_DWARF): Do not define.
4192 * config/i386/darwin.h (PREFERRED_DEBUGGING_TYPE): Do not
4193 change based on DARWIN_PREFER_DWARF not being defined.
4195 2021-09-13 Richard Biener <rguenther@suse.de>
4197 * config/i386/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
4198 to inherit from elfos.h
4200 2021-09-13 Richard Biener <rguenther@suse.de>
4202 * config.gcc: Add cr16-*-* to the list of obsoleted targets.
4204 2021-09-13 Richard Biener <rguenther@suse.de>
4206 * config/avr/elf.h (PREFERRED_DEBUGGING_TYPE): Remove
4207 override, pick up DWARF2_DEBUG define from elfos.h
4209 2021-09-13 Richard Biener <rguenther@suse.de>
4211 * config/rx/rx.h (PREFERRED_DEBUGGING_TYPE): Always define to
4214 2021-09-13 Richard Biener <rguenther@suse.de>
4216 * config/alpha/vms.h (PREFERRED_DEBUGGING_TYPE): Define to
4219 2021-09-13 Richard Biener <rguenther@suse.de>
4221 * config/i386/cygming.h: Always default to DWARF2 debugging.
4222 Do not define DBX_DEBUGGING_INFO, that's done via dbxcoff.h
4224 * doc/install.texi: Document binutils 2.16 as minimum
4225 requirement for mingw.
4227 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
4229 * config/rs6000/rs6000.c (struct rs6000_cost_data): New members
4230 nstmts, nloads and extra_ctor_cost.
4231 (rs6000_density_test): Add load density related heuristics. Do
4232 extra costing on vector construction statements if need.
4233 (rs6000_init_cost): Init new members.
4234 (rs6000_update_target_cost_per_stmt): New function.
4235 (rs6000_add_stmt_cost): Factor vect_nonmem hunk out to function
4236 rs6000_update_target_cost_per_stmt and call it.
4238 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
4240 * config/rs6000/rs6000.c (struct rs6000_cost_data): Remove typedef.
4241 (rs6000_init_cost): Adjust.
4243 2021-09-13 liuhongt <hongtao.liu@intel.com>
4245 * config/i386/i386.md: (UNSPEC_COPYSIGN): Remove.
4246 (UNSPEC_XORSIGN): Ditto.
4248 2021-09-12 Roger Sayle <roger@nextmovesoftware.com>
4250 * expr.c (convert_move): Preserve SUBREG_PROMOTED_VAR_P when
4251 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
4254 2021-09-11 Aldy Hernandez <aldyh@redhat.com>
4256 * tree-ssa-threadbackward.c (class back_threader_registry): Use
4257 back_jt_path_registry.
4258 * tree-ssa-threadedge.c (jump_threader::jump_threader): Use
4259 fwd_jt_path_registry.
4260 * tree-ssa-threadedge.h (class jump_threader): Same..
4261 * tree-ssa-threadupdate.c
4262 (jump_thread_path_registry::jump_thread_path_registry): Rename...
4263 (jt_path_registry::jt_path_registry): ...to this.
4264 (jump_thread_path_registry::~jump_thread_path_registry): Rename...
4265 (jt_path_registry::~jt_path_registry): ...this.
4266 (fwd_jt_path_registry::fwd_jt_path_registry): New.
4267 (fwd_jt_path_registry::~fwd_jt_path_registry): New.
4268 (jump_thread_path_registry::allocate_thread_edge): Rename...
4269 (jt_path_registry::allocate_thread_edge): ...to this.
4270 (jump_thread_path_registry::allocate_thread_path): Rename...
4271 (jt_path_registry::allocate_thread_path): ...to this.
4272 (jump_thread_path_registry::lookup_redirection_data): Rename...
4273 (fwd_jt_path_registry::lookup_redirection_data): ...to this.
4274 (jump_thread_path_registry::thread_block_1): Rename...
4275 (fwd_jt_path_registry::thread_block_1): ...to this.
4276 (jump_thread_path_registry::thread_block): Rename...
4277 (fwd_jt_path_registry::thread_block): ...to this.
4278 (jt_path_registry::thread_through_loop_header): Rename...
4279 (fwd_jt_path_registry::thread_through_loop_header): ...to this.
4280 (jump_thread_path_registry::mark_threaded_blocks): Rename...
4281 (fwd_jt_path_registry::mark_threaded_blocks): ...to this.
4282 (jump_thread_path_registry::debug_path): Rename...
4283 (jt_path_registry::debug_path): ...to this.
4284 (jump_thread_path_registry::dump): Rename...
4285 (jt_path_registry::debug): ...to this.
4286 (jump_thread_path_registry::rewire_first_differing_edge): Rename...
4287 (back_jt_path_registry::rewire_first_differing_edge): ...to this.
4288 (jump_thread_path_registry::adjust_paths_after_duplication): Rename...
4289 (back_jt_path_registry::adjust_paths_after_duplication): ...to this.
4290 (jump_thread_path_registry::duplicate_thread_path): Rename...
4291 (back_jt_path_registry::duplicate_thread_path): ...to this. Also,
4292 drop ill-formed candidates.
4293 (jump_thread_path_registry::remove_jump_threads_including): Rename...
4294 (fwd_jt_path_registry::remove_jump_threads_including): ...to this.
4295 (jt_path_registry::thread_through_all_blocks): New.
4296 (back_jt_path_registry::update_cfg): New.
4297 (fwd_jt_path_registry::update_cfg): New.
4298 (jump_thread_path_registry::register_jump_thread): Rename...
4299 (jt_path_registry::register_jump_thread): ...to this.
4300 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
4302 (class jt_path_registry): ...here.
4303 (class fwd_jt_path_registry): New.
4304 (class back_jt_path_registry): New.
4306 2021-09-10 liuhongt <hongtao.liu@intel.com>
4309 2021-09-01 liuhongt <hongtao.liu@intel.com>
4311 * emit-rtl.c (validate_subreg): Get rid of all float-int
4314 2021-09-10 Jakub Jelinek <jakub@redhat.com>
4316 * tree-core.h (enum omp_memory_order): Add OMP_MEMORY_ORDER_MASK,
4317 OMP_FAIL_MEMORY_ORDER_UNSPECIFIED, OMP_FAIL_MEMORY_ORDER_RELAXED,
4318 OMP_FAIL_MEMORY_ORDER_ACQUIRE, OMP_FAIL_MEMORY_ORDER_RELEASE,
4319 OMP_FAIL_MEMORY_ORDER_ACQ_REL, OMP_FAIL_MEMORY_ORDER_SEQ_CST and
4320 OMP_FAIL_MEMORY_ORDER_MASK enumerators.
4321 (OMP_FAIL_MEMORY_ORDER_SHIFT): Define.
4322 * gimple-pretty-print.c (dump_gimple_omp_atomic_load,
4323 dump_gimple_omp_atomic_store): Print [weak] for weak atomic
4325 * gimple.h (enum gf_mask): Change GF_OMP_ATOMIC_MEMORY_ORDER
4326 to 6-bit mask, adjust GF_OMP_ATOMIC_NEED_VALUE value and add
4328 (gimple_omp_atomic_weak_p, gimple_omp_atomic_set_weak): New inline
4330 * tree.h (OMP_ATOMIC_WEAK): Define.
4331 * tree-pretty-print.c (dump_omp_atomic_memory_order): Adjust for
4332 fail memory order being encoded in the same enum and also print
4333 fail clause if present.
4334 (dump_generic_node): Print weak clause if OMP_ATOMIC_WEAK.
4335 * gimplify.c (goa_stabilize_expr): Add target_expr and rhs arguments,
4336 handle pre_p == NULL case as a test mode that only returns value
4337 but doesn't change gimplify nor change anything otherwise, adjust
4338 recursive calls, add MODIFY_EXPR, ADDR_EXPR, COND_EXPR, TARGET_EXPR
4339 and CALL_EXPR handling, adjust COMPOUND_EXPR handling for
4340 __builtin_clear_padding calls, for !rhs gimplify as lvalue rather
4342 (gimplify_omp_atomic): Adjust goa_stabilize_expr caller. Handle
4343 COND_EXPR rhs. Set weak flag on gimple load/store for
4345 * omp-expand.c (omp_memory_order_to_fail_memmodel): New function.
4346 (omp_memory_order_to_memmodel): Adjust for fail clause encoded
4348 (expand_omp_atomic_cas): New function.
4349 (expand_omp_atomic_pipeline): Use omp_memory_order_to_fail_memmodel
4351 (expand_omp_atomic): Attempt to optimize atomic compare and exchange
4352 using expand_omp_atomic_cas.
4354 2021-09-10 Aldy Hernandez <aldyh@redhat.com>
4355 Michael Matz <matz@suse.de>
4357 * tree-pass.h (PROP_loop_opts_done): New.
4358 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
4359 Intersect with global range.
4360 * tree-ssa-loop.c (tree_ssa_loop_done): Set PROP_loop_opts_done.
4361 * tree-ssa-threadbackward.c
4362 (back_threader_profitability::profitable_path_p): Disable
4363 threading through latches until after loop optimizations have run.
4365 2021-09-10 David Faust <david.faust@oracle.com>
4367 * doc/invoke.texi: Document BPF -mcpu, -mjmpext, -mjmp32 and -malu32
4370 2021-09-10 David Faust <david.faust@oracle.com>
4372 * config/bpf/bpf-opts.h (bpf_isa_version): New enum.
4373 * config/bpf/bpf-protos.h (bpf_expand_cbranch): New.
4374 * config/bpf/bpf.c (bpf_option_override): Handle -mcpu option.
4375 (bpf_expand_cbranch): New function.
4376 * config/bpf/bpf.md (AM mode iterator): Conditionalize support for SI
4378 (zero_extendsidi2): Only use mov32 instruction if it is available.
4379 (SIM mode iterator): Conditionalize support for SI mode.
4380 (JM mode iterator): New.
4381 (cbranchdi4): Update name, use new JM iterator. Use bpf_expand_cbranch.
4382 (*branch_on_di): Update name, use new JM iterator.
4383 * config/bpf/bpf.opt: (mjmpext): New option.
4387 (bpf_isa): New enum.
4389 2021-09-10 David Faust <david.faust@oracle.com>
4391 * config/bpf/bpf.md (zero_extendhidi2): Add new output template
4392 for register-to-register extensions.
4393 (zero_extendqidi2): Likewise.
4395 2021-09-10 Richard Biener <rguenther@suse.de>
4397 PR middle-end/102273
4398 * internal-fn.c (expand_DEFERRED_INIT): Always expand non-SSA vars.
4400 2021-09-10 Richard Biener <rguenther@suse.de>
4402 PR middle-end/102269
4403 * gimplify.c (is_var_need_auto_init): Empty types do not need
4406 2021-09-10 Richard Biener <rguenther@suse.de>
4408 * configure.ac (--with-stabs): Remove.
4409 * configure: Regenerate.
4410 * doc/install.texi: Remove --with-stabs documentation.
4412 2021-09-10 liuhongt <hongtao.liu@intel.com>
4414 * config/i386/avx512fp16intrin.h: (_mm512_cmp_ph_mask):
4416 (_mm512_mask_cmp_ph_mask): Likewise.
4417 (_mm512_cmp_round_ph_mask): Likewise.
4418 (_mm512_mask_cmp_round_ph_mask): Likewise.
4419 (_mm_cmp_sh_mask): Likewise.
4420 (_mm_mask_cmp_sh_mask): Likewise.
4421 (_mm_cmp_round_sh_mask): Likewise.
4422 (_mm_mask_cmp_round_sh_mask): Likewise.
4423 (_mm_comieq_sh): Likewise.
4424 (_mm_comilt_sh): Likewise.
4425 (_mm_comile_sh): Likewise.
4426 (_mm_comigt_sh): Likewise.
4427 (_mm_comige_sh): Likewise.
4428 (_mm_comineq_sh): Likewise.
4429 (_mm_ucomieq_sh): Likewise.
4430 (_mm_ucomilt_sh): Likewise.
4431 (_mm_ucomile_sh): Likewise.
4432 (_mm_ucomigt_sh): Likewise.
4433 (_mm_ucomige_sh): Likewise.
4434 (_mm_ucomineq_sh): Likewise.
4435 (_mm_comi_round_sh): Likewise.
4436 (_mm_comi_sh): Likewise.
4437 * config/i386/avx512fp16vlintrin.h (_mm_cmp_ph_mask): New intrinsic.
4438 (_mm_mask_cmp_ph_mask): Likewise.
4439 (_mm256_cmp_ph_mask): Likewise.
4440 (_mm256_mask_cmp_ph_mask): Likewise.
4441 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
4442 * config/i386/i386-builtin.def: Add corresponding new builtins.
4443 * config/i386/i386-expand.c
4444 (ix86_expand_args_builtin): Handle new builtin types.
4445 (ix86_expand_round_builtin): Ditto.
4446 * config/i386/i386.md (ssevecmode): Add HF mode.
4447 (MODEFH): New mode iterator.
4448 * config/i386/sse.md
4449 (V48H_AVX512VL): New mode iterator to support HF vector modes.
4450 Ajdust corresponding description.
4451 (ssecmpintprefix): New.
4452 (VI12_AVX512VL): Adjust to support HF vector modes.
4453 (cmp_imm_predicate): Likewise.
4454 (<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
4456 (avx512f_vmcmp<mode>3<round_saeonly_name>): Likewise.
4457 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Likewise.
4458 (<sse>_<unord>comi<round_saeonly_name>): Likewise.
4460 2021-09-10 liuhongt <hongtao.liu@intel.com>
4462 * config/i386/avx512fp16intrin.h: (_mm512_max_ph): New intrinsic.
4463 (_mm512_mask_max_ph): Likewise.
4464 (_mm512_maskz_max_ph): Likewise.
4465 (_mm512_min_ph): Likewise.
4466 (_mm512_mask_min_ph): Likewise.
4467 (_mm512_maskz_min_ph): Likewise.
4468 (_mm512_max_round_ph): Likewise.
4469 (_mm512_mask_max_round_ph): Likewise.
4470 (_mm512_maskz_max_round_ph): Likewise.
4471 (_mm512_min_round_ph): Likewise.
4472 (_mm512_mask_min_round_ph): Likewise.
4473 (_mm512_maskz_min_round_ph): Likewise.
4474 (_mm_max_sh): Likewise.
4475 (_mm_mask_max_sh): Likewise.
4476 (_mm_maskz_max_sh): Likewise.
4477 (_mm_min_sh): Likewise.
4478 (_mm_mask_min_sh): Likewise.
4479 (_mm_maskz_min_sh): Likewise.
4480 (_mm_max_round_sh): Likewise.
4481 (_mm_mask_max_round_sh): Likewise.
4482 (_mm_maskz_max_round_sh): Likewise.
4483 (_mm_min_round_sh): Likewise.
4484 (_mm_mask_min_round_sh): Likewise.
4485 (_mm_maskz_min_round_sh): Likewise.
4486 * config/i386/avx512fp16vlintrin.h (_mm_max_ph): New intrinsic.
4487 (_mm256_max_ph): Likewise.
4488 (_mm_mask_max_ph): Likewise.
4489 (_mm256_mask_max_ph): Likewise.
4490 (_mm_maskz_max_ph): Likewise.
4491 (_mm256_maskz_max_ph): Likewise.
4492 (_mm_min_ph): Likewise.
4493 (_mm256_min_ph): Likewise.
4494 (_mm_mask_min_ph): Likewise.
4495 (_mm256_mask_min_ph): Likewise.
4496 (_mm_maskz_min_ph): Likewise.
4497 (_mm256_maskz_min_ph): Likewise.
4498 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
4499 * config/i386/i386-builtin.def: Add corresponding new builtins.
4500 * config/i386/i386-expand.c
4501 (ix86_expand_args_builtin): Handle new builtin types.
4502 * config/i386/sse.md
4503 (<code><mode>3<mask_name><round_saeonly_name>): Adjust to
4504 support HF vector modes.
4505 (*<code><mode>3<mask_name><round_saeonly_name>): Likewise.
4506 (ieee_<ieee_maxmin><mode>3<mask_name><round_saeonly_name>):
4508 (<sse>_vm<code><mode>3<mask_scalar_name><round_saeonly_scalar_name>):
4510 * config/i386/subst.md (round_saeonly_mode512bit_condition):
4511 Adjust for HF vector modes.
4513 2021-09-10 Liu, Hongtao <hongtao.liu@intel.com>
4515 * config/i386/avx512fp16intrin.h (_mm_add_sh): New intrinsic.
4516 (_mm_mask_add_sh): Likewise.
4517 (_mm_maskz_add_sh): Likewise.
4518 (_mm_sub_sh): Likewise.
4519 (_mm_mask_sub_sh): Likewise.
4520 (_mm_maskz_sub_sh): Likewise.
4521 (_mm_mul_sh): Likewise.
4522 (_mm_mask_mul_sh): Likewise.
4523 (_mm_maskz_mul_sh): Likewise.
4524 (_mm_div_sh): Likewise.
4525 (_mm_mask_div_sh): Likewise.
4526 (_mm_maskz_div_sh): Likewise.
4527 (_mm_add_round_sh): Likewise.
4528 (_mm_mask_add_round_sh): Likewise.
4529 (_mm_maskz_add_round_sh): Likewise.
4530 (_mm_sub_round_sh): Likewise.
4531 (_mm_mask_sub_round_sh): Likewise.
4532 (_mm_maskz_sub_round_sh): Likewise.
4533 (_mm_mul_round_sh): Likewise.
4534 (_mm_mask_mul_round_sh): Likewise.
4535 (_mm_maskz_mul_round_sh): Likewise.
4536 (_mm_div_round_sh): Likewise.
4537 (_mm_mask_div_round_sh): Likewise.
4538 (_mm_maskz_div_round_sh): Likewise.
4539 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
4540 * config/i386/i386-builtin.def: Add corresponding new builtins.
4541 * config/i386/i386-expand.c
4542 (ix86_expand_round_builtin): Handle new builtins.
4543 * config/i386/sse.md (VF_128): Change description.
4544 (<sse>_vm<plusminus_insn><mode>3<mask_scalar_name><round_scalar_name>):
4545 Adjust to support HF vector modes.
4546 (<sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name><round_scalar_name>):
4549 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
4551 * config/i386/i386-expand.c
4552 (ix86_avx256_split_vector_move_misalign): Handle V16HF mode.
4553 * config/i386/i386.c
4554 (ix86_preferred_simd_mode): Handle HF mode.
4555 * config/i386/sse.md (V_256H): New mode iterator.
4556 (avx_vextractf128<mode>): Use it.
4557 (VEC_INIT_MODE): Align vector HFmode condition to vector
4558 HImodes since there're no real HF instruction used.
4559 (VEC_INIT_HALF_MODE): Ditto.
4561 (VIHF_AVX512BW): Ditto.
4562 (*vec_extracthf): Ditto.
4563 (VEC_EXTRACT_MODE): Ditto.
4565 2021-09-10 Richard Biener <rguenther@suse.de>
4568 * config/dbx.h: Remove.
4569 * config/dbxcoff.h: Do not define PREFERRED_DEBUGGING_TYPE.
4570 * config/lynx.h: Likewise.
4572 2021-09-10 liuhongt <hongtao.liu@intel.com>
4574 * config/i386/i386-expand.c (ix86_expand_copysign): Expand
4575 right into ANDNOT + AND + IOR, using paradoxical subregs.
4576 (ix86_split_copysign_const): Remove.
4577 (ix86_split_copysign_var): Ditto.
4578 * config/i386/i386-protos.h (ix86_split_copysign_const): Dotto.
4579 (ix86_split_copysign_var): Ditto.
4580 * config/i386/i386.md (@copysign<mode>3_const): Ditto.
4581 (@copysign<mode>3_var): Ditto.
4583 2021-09-09 qing zhao <qing.zhao@oracle.com>
4585 * builtins.c (expand_builtin_memset): Make external visible.
4586 * builtins.h (expand_builtin_memset): Declare extern.
4587 * common.opt (ftrivial-auto-var-init=): New option.
4588 * doc/extend.texi: Document the uninitialized attribute.
4589 * doc/invoke.texi: Document -ftrivial-auto-var-init.
4590 * flag-types.h (enum auto_init_type): New enumerated type
4592 * gimple-fold.c (clear_padding_type): Add one new parameter.
4593 (clear_padding_union): Likewise.
4594 (clear_padding_emit_loop): Likewise.
4595 (clear_type_padding_in_mask): Likewise.
4596 (gimple_fold_builtin_clear_padding): Handle this new parameter.
4597 * gimplify.c (gimple_add_init_for_auto_var): New function.
4598 (gimple_add_padding_init_for_auto_var): New function.
4599 (is_var_need_auto_init): New function.
4600 (gimplify_decl_expr): Add initialization to automatic variables per
4602 (gimplify_call_expr): Add one new parameter for call to
4603 __builtin_clear_padding.
4604 (gimplify_init_constructor): Add padding initialization in the end.
4605 * internal-fn.c (INIT_PATTERN_VALUE): New macro.
4606 (expand_DEFERRED_INIT): New function.
4607 * internal-fn.def (DEFERRED_INIT): New internal function.
4608 * tree-cfg.c (verify_gimple_call): Verify calls to .DEFERRED_INIT.
4609 * tree-sra.c (generate_subtree_deferred_init): New function.
4610 (scan_function): Avoid setting cannot_scalarize_away_bitmap for
4611 calls to .DEFERRED_INIT.
4612 (sra_modify_deferred_init): New function.
4613 (sra_modify_function_body): Handle calls to DEFERRED_INIT specially.
4614 * tree-ssa-structalias.c (find_func_aliases_for_call): Likewise.
4615 * tree-ssa-uninit.c (warn_uninit): Handle calls to DEFERRED_INIT
4617 (check_defs): Likewise.
4618 (warn_uninitialized_vars): Likewise.
4619 * tree-ssa.c (ssa_undefined_value_p): Likewise.
4620 * tree.c (build_common_builtin_nodes): Build tree node for
4621 BUILT_IN_CLEAR_PADDING when needed.
4623 2021-09-09 Richard Biener <rguenther@suse.de>
4625 * tree-ssa-loop-im.c (fill_always_executed_in_1): Walk
4628 2021-09-09 Richard Biener <rguenther@suse.de>
4630 * tree-ssa-loop-im.c (fill_always_executed_in_1): Integrate
4631 DOM walk from get_loop_body_in_dom_order using a worklist
4634 2021-09-09 liuhongt <hongtao.liu@intel.com>
4636 * config.gcc: Add avx512fp16vlintrin.h.
4637 * config/i386/avx512fp16intrin.h: (_mm512_add_ph): New intrinsic.
4638 (_mm512_mask_add_ph): Likewise.
4639 (_mm512_maskz_add_ph): Likewise.
4640 (_mm512_sub_ph): Likewise.
4641 (_mm512_mask_sub_ph): Likewise.
4642 (_mm512_maskz_sub_ph): Likewise.
4643 (_mm512_mul_ph): Likewise.
4644 (_mm512_mask_mul_ph): Likewise.
4645 (_mm512_maskz_mul_ph): Likewise.
4646 (_mm512_div_ph): Likewise.
4647 (_mm512_mask_div_ph): Likewise.
4648 (_mm512_maskz_div_ph): Likewise.
4649 (_mm512_add_round_ph): Likewise.
4650 (_mm512_mask_add_round_ph): Likewise.
4651 (_mm512_maskz_add_round_ph): Likewise.
4652 (_mm512_sub_round_ph): Likewise.
4653 (_mm512_mask_sub_round_ph): Likewise.
4654 (_mm512_maskz_sub_round_ph): Likewise.
4655 (_mm512_mul_round_ph): Likewise.
4656 (_mm512_mask_mul_round_ph): Likewise.
4657 (_mm512_maskz_mul_round_ph): Likewise.
4658 (_mm512_div_round_ph): Likewise.
4659 (_mm512_mask_div_round_ph): Likewise.
4660 (_mm512_maskz_div_round_ph): Likewise.
4661 * config/i386/avx512fp16vlintrin.h: New header.
4662 * config/i386/i386-builtin-types.def (V16HF, V8HF, V32HF):
4663 Add new builtin types.
4664 * config/i386/i386-builtin.def: Add corresponding builtins.
4665 * config/i386/i386-expand.c
4666 (ix86_expand_args_builtin): Handle new builtin types.
4667 (ix86_expand_round_builtin): Likewise.
4668 * config/i386/immintrin.h: Include avx512fp16vlintrin.h
4669 * config/i386/sse.md (VFH): New mode_iterator.
4671 (avx512fmaskmode): Add HF vector modes.
4672 (avx512fmaskhalfmode): Likewise.
4673 (<plusminus_insn><mode>3<mask_name><round_name>): Adjust to for
4675 (*<plusminus_insn><mode>3<mask_name><round_name>): Likewise.
4676 (mul<mode>3<mask_name><round_name>): Likewise.
4677 (*mul<mode>3<mask_name><round_name>): Likewise.
4678 (div<mode>3): Likewise.
4679 (<sse>_div<mode>3<mask_name><round_name>): Likewise.
4680 * config/i386/subst.md (SUBST_V): Add HF vector modes.
4681 (SUBST_A): Likewise.
4682 (round_mode512bit_condition): Adjust for V32HFmode.
4684 2021-09-09 liuhongt <hongtao.liu@intel.com>
4687 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
4688 (reduc_plus_scal_v4sf): .. this, New define_expand.
4689 (reduc_plus_scal_v2df): .. and this, New define_expand.
4691 2021-09-09 liuhongt <hongtao.liu@intel.com>
4694 * config/i386/sse.md (*vec_extract<mode><ssescalarmodelower>_valign):
4697 2021-09-08 Jonathan Wakely <jwakely@redhat.com>
4700 * doc/trouble.texi (Copy Assignment): Fix description of
4701 behaviour and fix code in example.
4703 2021-09-08 Segher Boessenkool <segher@kernel.crashing.org>
4706 * config/rs6000/rs6000-logue.c (rs6000_emit_epilogue): For ELFv2 use
4707 r11 instead of r12 for restoring CR.
4709 2021-09-08 Jakub Jelinek <jakub@redhat.com>
4710 liuhongt <hongtao.liu@intel.com>
4713 * config/i386/i386.md (@xorsign<mode>3_1): Remove.
4714 * config/i386/i386-expand.c (ix86_expand_xorsign): Expand right away
4715 into AND with mask and XOR, using paradoxical subregs.
4716 (ix86_split_xorsign): Remove.
4717 * config/i386/i386-protos.h (ix86_split_xorsign): Remove.
4719 2021-09-08 Di Zhao <dizhao@os.amperecomputing.com>
4721 * tree-ssa-sccvn.c (vn_nary_op_insert_into): fix result compare
4723 2021-09-08 Jakub Jelinek <jakub@redhat.com>
4726 * config/i386/i386.md (xorsign<mode>3): If operands[1] is equal to
4727 operands[2], emit abs<mode>2 instead.
4728 (@xorsign<mode>3_1): Add early-clobbers for output operand, enable
4729 first alternative even for avx, add another alternative with
4730 =&Yv <- 0, Yv, Yvm constraints.
4731 * config/i386/i386-expand.c (ix86_split_xorsign): If op0 is equal
4732 to op1, emit vpandn instead.
4734 2021-09-08 liuhongt <hongtao.liu@intel.com>
4736 * config/i386/avx512fp16intrin.h (_mm_set_ph): New intrinsic.
4737 (_mm256_set_ph): Likewise.
4738 (_mm512_set_ph): Likewise.
4739 (_mm_setr_ph): Likewise.
4740 (_mm256_setr_ph): Likewise.
4741 (_mm512_setr_ph): Likewise.
4742 (_mm_set1_ph): Likewise.
4743 (_mm256_set1_ph): Likewise.
4744 (_mm512_set1_ph): Likewise.
4745 (_mm_setzero_ph): Likewise.
4746 (_mm256_setzero_ph): Likewise.
4747 (_mm512_setzero_ph): Likewise.
4748 (_mm_set_sh): Likewise.
4749 (_mm_load_sh): Likewise.
4750 (_mm_store_sh): Likewise.
4751 * config/i386/i386-builtin-types.def (V8HF): New type.
4752 (DEF_FUNCTION_TYPE (V8HF, V8HI)): New builtin function type
4753 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
4754 Support vector HFmodes.
4755 (ix86_expand_vector_init_one_nonzero): Likewise.
4756 (ix86_expand_vector_init_one_var): Likewise.
4757 (ix86_expand_vector_init_interleave): Likewise.
4758 (ix86_expand_vector_init_general): Likewise.
4759 (ix86_expand_vector_set): Likewise.
4760 (ix86_expand_vector_extract): Likewise.
4761 (ix86_expand_vector_init_concat): Likewise.
4762 (ix86_expand_sse_movcc): Handle vector HFmodes.
4763 (ix86_expand_vector_set_var): Ditto.
4764 * config/i386/i386-modes.def: Add HF vector modes in comment.
4765 * config/i386/i386.c (classify_argument): Add HF vector modes.
4766 (ix86_hard_regno_mode_ok): Allow HF vector modes for AVX512FP16.
4767 (ix86_vector_mode_supported_p): Likewise.
4768 (ix86_set_reg_reg_cost): Handle vector HFmode.
4769 (ix86_get_ssemov): Handle vector HFmode.
4770 (function_arg_advance_64): Pass unamed V16HFmode and V32HFmode
4772 (function_arg_advance_32): Pass V8HF/V16HF/V32HF by sse reg for 32bit
4774 (function_arg_advance_32): Ditto.
4775 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): New.
4776 (VALID_AVX256_REG_OR_OI_MODE): Rename to ..
4777 (VALID_AVX256_REG_OR_OI_VHF_MODE): .. this, and add V16HF.
4778 (VALID_SSE2_REG_VHF_MODE): New.
4779 (VALID_AVX512VL_128_REG_MODE): Add V8HF and TImode.
4780 (SSE_REG_MODE_P): Add vector HFmode.
4781 * config/i386/i386.md (mode): Add HF vector modes.
4782 (MODE_SIZE): Likewise.
4783 (ssemodesuffix): Add ph suffix for HF vector modes.
4784 * config/i386/sse.md (VFH_128): New mode iterator.
4785 (VMOVE): Adjust for HF vector modes.
4787 (V_256_512): Likewise.
4789 (avx512fmaskmode): Likewise.
4790 (shuffletype): Likewise.
4791 (sseinsnmode): Likewise.
4792 (ssedoublevecmode): Likewise.
4793 (ssehalfvecmode): Likewise.
4794 (ssehalfvecmodelower): Likewise.
4795 (ssePScmode): Likewise.
4796 (ssescalarmode): Likewise.
4797 (ssescalarmodelower): Likewise.
4798 (sseintprefix): Likewise.
4800 (bcstscalarsuff): Likewise.
4801 (xtg_mode): Likewise.
4802 (VI12HF_AVX512VL): New mode_iterator.
4803 (VF_AVX512FP16): Likewise.
4805 (VIHF_256): Likewise.
4806 (VIHF_AVX512BW): Likewise.
4807 (V16_256): Likewise.
4808 (V32_512): Likewise.
4809 (sseintmodesuffix): New mode_attr.
4810 (sse): Add scalar and vector HFmodes.
4811 (ssescalarmode): Add vector HFmode mapping.
4812 (ssescalarmodesuffix): Add sh suffix for HFmode.
4813 (*<sse>_vm<insn><mode>3): Use VFH_128.
4814 (*<sse>_vm<multdiv_mnemonic><mode>3): Likewise.
4815 (*ieee_<ieee_maxmin><mode>3): Likewise.
4816 (<avx512>_blendm<mode>): New define_insn.
4817 (vec_setv8hf): New define_expand.
4818 (vec_set<mode>_0): New define_insn for HF vector set.
4819 (*avx512fp16_movsh): Likewise.
4820 (avx512fp16_movsh): Likewise.
4821 (vec_extract_lo_v32hi): Rename to ...
4822 (vec_extract_lo_<mode>): ... this, and adjust to allow HF
4824 (vec_extract_hi_v32hi): Likewise.
4825 (vec_extract_hi_<mode>): Likewise.
4826 (vec_extract_lo_v16hi): Likewise.
4827 (vec_extract_lo_<mode>): Likewise.
4828 (vec_extract_hi_v16hi): Likewise.
4829 (vec_extract_hi_<mode>): Likewise.
4830 (vec_set_hi_v16hi): Likewise.
4831 (vec_set_hi_<mode>): Likewise.
4832 (vec_set_lo_v16hi): Likewise.
4833 (vec_set_lo_<mode>): Likewise.
4834 (*vec_extract<mode>_0): New define_insn_and_split for HF
4836 (*vec_extracthf): New define_insn.
4837 (VEC_EXTRACT_MODE): Add HF vector modes.
4838 (PINSR_MODE): Add V8HF.
4839 (sse2p4_1): Likewise.
4840 (pinsr_evex_isa): Likewise.
4841 (<sse2p4_1>_pinsr<ssemodesuffix>): Adjust to support
4842 insert for V8HFmode.
4843 (pbroadcast_evex_isa): Add HF vector modes.
4844 (AVX2_VEC_DUP_MODE): Likewise.
4845 (VEC_INIT_MODE): Likewise.
4846 (VEC_INIT_HALF_MODE): Likewise.
4847 (avx2_pbroadcast<mode>): Adjust to support HF vector mode
4849 (avx2_pbroadcast<mode>_1): Likewise.
4850 (<avx512>_vec_dup<mode>_1): Likewise.
4851 (<avx512>_vec_dup<mode><mask_name>): Likewise.
4852 (<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>):
4855 2021-09-08 Guo, Xuepeng <xuepeng.guo@intel.com>
4856 H.J. Lu <hongjiu.lu@intel.com>
4857 Liu Hongtao <hongtao.liu@intel.com>
4858 Wang Hongyu <hongyu.wang@intel.com>
4859 Xu Dianhong <dianhong.xu@intel.com>
4861 * common/config/i386/cpuinfo.h (get_available_features):
4862 Detect FEATURE_AVX512FP16.
4863 * common/config/i386/i386-common.c
4864 (OPTION_MASK_ISA_AVX512FP16_SET,
4865 OPTION_MASK_ISA_AVX512FP16_UNSET,
4866 OPTION_MASK_ISA2_AVX512FP16_SET,
4867 OPTION_MASK_ISA2_AVX512FP16_UNSET): New.
4868 (OPTION_MASK_ISA2_AVX512BW_UNSET,
4869 OPTION_MASK_ISA2_AVX512BF16_UNSET): Add AVX512FP16.
4870 (ix86_handle_option): Handle -mavx512fp16.
4871 * common/config/i386/i386-cpuinfo.h (enum processor_features):
4872 Add FEATURE_AVX512FP16.
4873 * common/config/i386/i386-isas.h: Add entry for AVX512FP16.
4874 * config.gcc: Add avx512fp16intrin.h.
4875 * config/i386/avx512fp16intrin.h: New intrinsic header.
4876 * config/i386/cpuid.h: Add bit_AVX512FP16.
4877 * config/i386/i386-builtin-types.def: (FLOAT16): New primitive type.
4878 * config/i386/i386-builtins.c: Support _Float16 type for i386
4880 (ix86_register_float16_builtin_type): New function.
4881 (ix86_float16_type_node): New.
4882 * config/i386/i386-c.c (ix86_target_macros_internal): Define
4884 * config/i386/i386-expand.c (ix86_expand_branch): Support
4886 (ix86_prepare_fp_compare_args): Adjust TARGET_SSE_MATH &&
4887 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
4888 (ix86_expand_fp_movcc): Ditto.
4889 * config/i386/i386-isa.def: Add PTA define for AVX512FP16.
4890 * config/i386/i386-options.c (isa2_opts): Add -mavx512fp16.
4891 (ix86_valid_target_attribute_inner_p): Add avx512fp16 attribute.
4892 * config/i386/i386.c (ix86_get_ssemov): Use
4893 vmovdqu16/vmovw/vmovsh for HFmode/HImode scalar or vector.
4894 (ix86_get_excess_precision): Use
4895 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when TARGET_AVX512FP16
4897 (sse_store_index): Use SFmode cost for HFmode cost.
4898 (inline_memory_move_cost): Add HFmode, and perfer SSE cost over
4899 GPR cost for HFmode.
4900 (ix86_hard_regno_mode_ok): Allow HImode in sse register.
4901 (ix86_mangle_type): Add manlging for _Float16 type.
4902 (inline_secondary_memory_needed): No memory is needed for
4903 16bit movement between gpr and sse reg under
4905 (ix86_multiplication_cost): Adjust TARGET_SSE_MATH &&
4906 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
4907 (ix86_division_cost): Ditto.
4908 (ix86_rtx_costs): Ditto.
4909 (ix86_add_stmt_cost): Ditto.
4910 (ix86_optab_supported_p): Ditto.
4911 * config/i386/i386.h (VALID_AVX512F_SCALAR_MODE): Add HFmode.
4912 (SSE_FLOAT_MODE_SSEMATH_OR_HF_P): Add HFmode.
4913 (PTA_SAPPHIRERAPIDS): Add PTA_AVX512FP16.
4914 * config/i386/i386.md (mode): Add HFmode.
4915 (MODE_SIZE): Add HFmode.
4916 (isa): Add avx512fp16.
4917 (enabled): Handle avx512fp16.
4918 (ssemodesuffix): Add sh suffix for HFmode.
4919 (comm): Add mult, div.
4920 (plusminusmultdiv): New code iterator.
4921 (insn): Add mult, div.
4922 (*movhf_internal): Adjust for avx512fp16 instruction.
4923 (*movhi_internal): Ditto.
4924 (*cmpi<unord>hf): New define_insn for HFmode.
4925 (*ieee_s<ieee_maxmin>hf3): Likewise.
4926 (extendhf<mode>2): Likewise.
4927 (trunc<mode>hf2): Likewise.
4928 (float<floatunssuffix><mode>hf2): Likewise.
4929 (*<insn>hf): Likewise.
4930 (cbranchhf4): New expander.
4931 (movhfcc): Likewise.
4932 (<insn>hf3): Likewise.
4935 * config/i386/i386.opt: Add mavx512fp16.
4936 * config/i386/immintrin.h: Include avx512fp16intrin.h.
4937 * doc/invoke.texi: Add mavx512fp16.
4938 * doc/extend.texi: Add avx512fp16 Usage Notes.
4940 2021-09-08 liuhongt <hongtao.liu@intel.com>
4942 * common.opt: Support -fexcess-precision=16.
4943 * config/aarch64/aarch64.c (aarch64_excess_precision): Return
4944 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when
4945 EXCESS_PRECISION_TYPE_FLOAT16.
4946 * config/arm/arm.c (arm_excess_precision): Ditto.
4947 * config/i386/i386.c (ix86_get_excess_precision): Ditto.
4948 * config/m68k/m68k.c (m68k_excess_precision): Issue an error
4949 when EXCESS_PRECISION_TYPE_FLOAT16.
4950 * config/s390/s390.c (s390_excess_precision): Ditto.
4951 * coretypes.h (enum excess_precision_type): Add
4952 EXCESS_PRECISION_TYPE_FLOAT16.
4953 * doc/tm.texi (TARGET_C_EXCESS_PRECISION): Update documents.
4954 * doc/tm.texi.in (TARGET_C_EXCESS_PRECISION): Ditto.
4955 * doc/extend.texi (Half-Precision): Document
4956 -fexcess-precision=16.
4957 * flag-types.h (enum excess_precision): Add
4958 EXCESS_PRECISION_FLOAT16.
4959 * target.def (excess_precision): Update document.
4960 * tree.c (excess_precision_type): Set excess_precision_type to
4961 EXCESS_PRECISION_FLOAT16 when -fexcess-precision=16.
4963 2021-09-08 liuhongt <hongtao.liu@intel.com>
4965 * doc/extend.texi: (@node Floating Types): Adjust the wording.
4966 (@node Half-Precision): Ditto.
4968 2021-09-07 Takayuki 'January June' Suwa <jjsuwa_sys3175@yahoo.co.jp>
4971 * config/xtensa/xtensa.c (xtensa_emit_move_sequence): Add
4972 'CONST_INT_P (src)' to the condition of the block that tries to
4973 eliminate literal when loading integer contant.
4975 2021-09-07 David Faust <david.faust@oracle.com>
4977 * doc/extend.texi (BPF Type Attributes) New node.
4978 Document new preserve_access_index attribute.
4979 Document new preserve_access_index builtin.
4980 * doc/invoke.texi: Document -mco-re and -mno-co-re options.
4982 2021-09-07 David Faust <david.faust@oracle.com>
4984 * config/bpf/bpf.c: Adjust includes.
4985 (bpf_handle_preserve_access_index_attribute): New function.
4986 (bpf_attribute_table): Use it here.
4987 (bpf_builtins): Add BPF_BUILTIN_PRESERVE_ACCESS_INDEX.
4988 (bpf_option_override): Handle "-mco-re" option.
4989 (bpf_asm_init_sections): New.
4990 (TARGET_ASM_INIT_SECTIONS): Redefine.
4991 (bpf_file_end): New.
4992 (TARGET_ASM_FILE_END): Redefine.
4993 (bpf_init_builtins): Add "__builtin_preserve_access_index".
4994 (bpf_core_compute, bpf_core_get_index): New.
4995 (is_attr_preserve_access): New.
4996 (bpf_expand_builtin): Handle new builtins.
4997 (bpf_core_newdecl, bpf_core_is_maybe_aggregate_access): New.
4998 (bpf_core_walk): New.
4999 (bpf_resolve_overloaded_builtin): New.
5000 (TARGET_RESOLVE_OVERLOADED_BUILTIN): Redefine.
5002 (pass_bpf_core_attr): New RTL pass.
5003 * config/bpf/bpf-passes.def: New file.
5004 * config/bpf/bpf-protos.h (make_pass_bpf_core_attr): New.
5005 * config/bpf/coreout.c: New file.
5006 * config/bpf/coreout.h: Likewise.
5007 * config/bpf/t-bpf (TM_H): Add $(srcdir)/config/bpf/coreout.h.
5008 (coreout.o): New rule.
5009 (PASSES_EXTRA): Add $(srcdir)/config/bpf/bpf-passes.def.
5010 * config.gcc (bpf): Add coreout.h to extra_headers.
5011 Add coreout.o to extra_objs.
5012 Add $(srcdir)/config/bpf/coreout.c to target_gtfiles.
5014 2021-09-07 David Faust <david.faust@oracle.com>
5016 * btfout.c (get_btf_id): Function is no longer static.
5017 * ctfc.h: Expose it here.
5019 2021-09-07 David Faust <david.faust@oracle.com>
5021 * ctfc.c (ctf_lookup_tree_type): New function.
5024 2021-09-07 David Faust <david.faust@oracle.com>
5026 * ctfc.c (ctf_dtd_lookup): Function is no longer static.
5027 * ctfc.h: Analogous change.
5029 2021-09-07 David Faust <david.faust@oracle.com>
5031 * dwarf2out.c (lookup_type_die): Function is no longer static.
5032 * dwarf2out.h: Expose it here.
5034 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
5036 * dwarf2ctf.c (ctf_debug_finalize): Make it static.
5037 (ctf_debug_early_finish): New definition.
5038 (ctf_debug_finish): Likewise.
5039 * dwarf2ctf.h (ctf_debug_finalize): Remove declaration.
5040 (ctf_debug_early_finish): New declaration.
5041 (ctf_debug_finish): Likewise.
5042 * dwarf2out.c (dwarf2out_finish): Invoke ctf_debug_finish.
5043 (dwarf2out_early_finish): Invoke ctf_debug_early_finish.
5045 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
5047 * config/bpf/bpf.c (bpf_option_override): For BPF backend, disable LTO
5048 support when compiling for CO-RE.
5049 * config/bpf/bpf.opt: Add new command line option -mco-re.
5051 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
5053 * flag-types.h (enum debug_info_type): Add new enum
5054 DINFO_TYPE_BTF_WITH_CORE.
5055 (BTF_WITH_CORE_DEBUG): New bitmask.
5056 * flags.h (btf_with_core_debuginfo_p): New declaration.
5057 * opts.c (btf_with_core_debuginfo_p): New definition.
5059 2021-09-07 Jason Merrill <jason@redhat.com>
5061 * tree.h (error_operand_p): Change to inline function.
5063 2021-09-07 Aldy Hernandez <aldyh@redhat.com>
5065 * tree-ssa-threadedge.c (forwarder_block_p): Rename to...
5066 (empty_block_with_phis_p): ...this.
5067 (potentially_threadable_block): Same.
5068 (jump_threader::thread_through_normal_block): Same.
5070 2021-09-07 Eric Botcazou <ebotcazou@adacore.com>
5073 * dwarf2out.c (mark_base_types): New overloaded function.
5074 (dwarf2out_early_finish): Invoke it on the COMDAT type list as well
5075 as the compilation unit, and call move_marked_base_types afterward.
5077 2021-09-07 H.J. Lu <hjl.tools@gmail.com>
5080 * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
5082 (ix86_expand_vector_convert_uns_vsivsf): Likewise.
5084 2021-09-07 Richard Biener <rguenther@suse.de>
5086 PR tree-optimization/102226
5087 * tree-vect-loop.c (vect_transform_cycle_phi): Record
5088 the converted value for the epilogue PHI use.
5090 2021-09-07 Martin Liska <mliska@suse.cz>
5092 PR gcov-profile/80223
5093 * ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
5094 options, do not inline when no_profile_instrument_function
5095 attributes are different in early inliner. It's fine to inline
5096 it after PGO instrumentation.
5098 2021-09-07 Richard Biener <rguenther@suse.de>
5100 PR tree-optimization/101555
5101 * tree-ssa-pre.c (translate_vuse_through_block): Do not
5102 perform an alias walk to determine the validity of the
5103 mem at the start of the block which is already guaranteed
5104 by means of prune_clobbered_mems.
5105 (phi_translate_1): Pass edge to translate_vuse_through_block.
5107 2021-09-07 Xionghu Luo <luoxhu@linux.ibm.com>
5110 * config/rs6000/rs6000.md (fmod<mode>3): New define_expand.
5111 (remainder<mode>3): Likewise.
5113 2021-09-07 YunQiang Su <yunqiang.su@cipunited.com>
5115 * config/mips/mips.c (mips_file_start): add .module for
5118 2021-09-06 Roger Sayle <roger@nextmovesoftware.com>
5120 * wide-int.cc (wi::clz): Reorder tests to ensure the result
5121 is zero for all negative values.
5123 2021-09-06 Tobias Burnus <tobias@codesourcery.com>
5125 * doc/invoke.texi (-foffload-options): Fix @opindex.
5127 2021-09-06 H.J. Lu <hjl.tools@gmail.com>
5130 * config/i386/i386-expand.c (ix86_split_xorsign): Use operands[2].
5131 * config/i386/i386.md (@xorsign<mode>3_1): Add non-destructive
5132 source alternative for AVX.
5134 2021-09-06 liuhongt <hongtao.liu@intel.com>
5136 PR middle-end/102182
5137 * optabs.c (expand_fix): Add from1 to avoid from being
5140 2021-09-06 Eric Botcazou <ebotcazou@adacore.com>
5142 * dwarf2out.c (modified_type_die): Deal with all array types earlier
5143 and use local variable consistently throughout the function.
5145 2021-09-06 Jakub Jelinek <jakub@redhat.com>
5147 PR tree-optimization/102207
5148 * match.pd: Don't demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
5149 were promoted from signed to wider unsigned type.
5151 2021-09-06 Andrew Pinski <apinski@marvell.com>
5153 PR tree-optimization/63184
5154 * match.pd: Add simplification of pointer_diff of two pointer_plus
5155 with addr_expr in the first operand of each pointer_plus.
5156 Add simplificatoin of ne/eq of two pointer_plus with addr_expr
5157 in the first operand of each pointer_plus.
5159 2021-09-06 Richard Biener <rguenther@suse.de>
5161 PR tree-optimization/102176
5162 * tree-vect-slp.c (vect_slp_gather_vectorized_scalar_stmts):
5164 (vect_bb_slp_scalar_cost): Use the computed set of
5165 vectorized scalar stmts instead of relying on the out-of-date
5166 and not accurate PURE_SLP_STMT.
5167 (vect_bb_vectorization_profitable_p): Compute the set
5168 of vectorized scalar stmts.
5170 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
5172 * gimple-range-path.cc (path_range_query::range_of_stmt): Remove
5173 GIMPLE_COND special casing.
5174 (path_range_query::range_defined_in_block): Use range_of_stmt
5175 instead of calling fold_range directly.
5177 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
5179 * gimple-range-path.cc (path_range_query::range_of_expr): Set
5180 m_undefined_path when appropriate.
5181 (path_range_query::internal_range_of_expr): Copy from range_of_expr.
5182 (path_range_query::unreachable_path_p): New.
5183 (path_range_query::precompute_ranges): Set m_undefined_path.
5184 * gimple-range-path.h (path_range_query::unreachable_path_p): New.
5185 (path_range_query::internal_range_of_expr): New.
5186 * tree-ssa-threadbackward.c (back_threader::find_taken_edge_cond):
5187 Use unreachable_path_p.
5189 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
5191 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
5192 Remove argument and call find_taken_edge.
5193 (back_threader::resolve_phi): Do not calculate taken edge before
5194 calling maybe_register_path.
5195 (back_threader::find_paths_to_names): Same.
5197 2021-09-05 Jeff Law <jlaw@localhost.localdomain>
5199 * config/h8300/h8300.md (QHSI2 mode iterator): New mode iterator.
5200 * config/h8300/testcompare.md (store_c): Update name, use new
5202 (store_neg_c, store_shifted_c): New patterns.
5204 2021-09-03 Segher Boessenkool <segher@kernel.crashing.org>
5207 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): On ELFv2 use r11
5208 instead of r12 for CR save, in all cases.
5210 2021-09-03 Andrew Pinski <apinski@marvell.com>
5212 * config/aarch64/aarch64-sve-builtins.cc (register_vector_type):
5213 Handle error_mark_node as the type of the type_decl.
5215 2021-09-03 Andrew Pinski <apinski@marvell.com>
5217 * config/aarch64/aarch64-builtins.c (struct aarch64_simd_type_info):
5219 (aarch64_simd_types): Likewise.
5220 (aarch64_simd_intOI_type_node): Likewise.
5221 (aarch64_simd_intCI_type_node): Likewise.
5222 (aarch64_simd_intXI_type_node): Likewise.
5223 * config/aarch64/aarch64.h (aarch64_fp16_type_node): Likewise.
5224 (aarch64_fp16_ptr_type_node): Likewise.
5225 (aarch64_bf16_type_node): Likewise.
5226 (aarch64_bf16_ptr_type_node): Likewise.
5228 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5230 * range-op.cc (operator_minus::op1_op2_relation_effect): Abstract
5232 (minus_op1_op2_relation_effect): ...here.
5233 (class operator_pointer_diff): New.
5234 (operator_pointer_diff::op1_op2_relation_effect): Call
5235 minus_op1_op2_relation_effect.
5236 (integral_table::integral_table): Add entry for POINTER_DIFF_EXPR.
5238 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5240 * tree-ssa-threadbackward.c (back_threader::thread_through_all_blocks):
5241 Add may_peel_loop_headers.
5242 (back_threader_registry::thread_through_all_blocks): Same.
5243 (try_thread_blocks): Pass may_peel_loop_headers argument.
5244 (pass_early_thread_jumps::execute): Same.
5246 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5248 * tree-ssa-threadedge.c (has_phis_p): New.
5249 (forwarder_block_p): New.
5250 (potentially_threadable_block): Call forwarder_block_p.
5251 (jump_threader::thread_around_empty_blocks): Call has_phis_p.
5252 (jump_threader::thread_through_normal_block): Call
5255 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5257 * tree-ssa-threadbackward.c (back_threader::dump): New.
5258 (back_threader::debug): New.
5259 (back_threader_profitability::profitable_path_p): Dump blocks
5260 even if we are bailing early.
5262 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5264 * tree-ssa-threadupdate.c (cancel_thread): New.
5265 (jump_thread_path_registry::thread_block_1): Use cancel_thread.
5266 (jump_thread_path_registry::mark_threaded_blocks): Same.
5267 (jump_thread_path_registry::register_jump_thread): Same.
5269 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5271 * tree-ssa-threadedge.c (jt_state::push): Only call methods for
5272 which objects are available.
5273 (jt_state::pop): Same.
5274 (jt_state::register_equiv): Same.
5275 (jt_state::register_equivs_on_edge): Same.
5277 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5279 * tree-ssa-threadedge.c (jump_threader::thread_across_edge):
5280 Move pop until after a thread is registered.
5282 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5284 * tree-ssa-threadupdate.c (debug): New.
5286 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5288 * gimple-range-trace.cc (push_dump_file::push_dump_file): New.
5289 (push_dump_file::~push_dump_file): New.
5290 (dump_ranger): Change dump_file temporarily while dumping
5292 * gimple-range-trace.h (class push_dump_file): New.
5294 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5296 * gimple-range-trace.cc (debug_seed_ranger): Remove static.
5297 (dump_ranger): Dump function name.
5299 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5301 * gimple-range-path.cc (path_range_query::range_defined_in_block):
5302 Adjust for non-null.
5303 (path_range_query::adjust_for_non_null_uses): New.
5304 (path_range_query::precompute_ranges): Call
5305 adjust_for_non_null_uses.
5306 * gimple-range-path.h: Add m_non_null and
5307 adjust_for_non_null_uses.
5309 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5311 * gimple-range-path.cc (path_range_query::dump): Dump path
5313 (path_range_query::precompute_ranges): Dump entire path.
5315 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5317 * value-relation.cc (relation_oracle::debug): New.
5318 * value-relation.h (relation_oracle::debug): New.
5320 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5322 * tree-ssa-loop-ch.c: Remove unnecessary include file.
5324 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5326 * gimple-range-fold.cc (fold_using_range::postfold_gcond_edges):
5327 Skip statements with no defining BB.
5328 * gimple-range-path.cc (path_range_query::range_defined_in_block):
5329 Do not get confused by statements with no defining BB.
5331 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5333 * gimple-range-fold.cc (adjust_imagpart_expr): Move from
5334 gimple_range_adjustment. Add support for constants.
5335 (adjust_realpart_expr): New.
5336 (gimple_range_adjustment): Move IMAGPART_EXPR code to
5337 adjust_imagpart_expr.
5338 * range-op.cc (integral_table::integral_table): Add entry for
5341 2021-09-03 Jakub Jelinek <jakub@redhat.com>
5343 * omp-expand.c (expand_omp_atomic_pipeline): Use
5344 IFN_ATOMIC_COMPARE_EXCHANGE instead of
5345 BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_? so that memory order
5348 2021-09-03 Jakub Jelinek <jakub@redhat.com>
5351 * tree.h (DECL_FIELD_ABI_IGNORED): Changed into rvalue only macro
5352 that is false if DECL_BIT_FIELD.
5353 (SET_DECL_FIELD_ABI_IGNORED, DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD,
5354 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD): Define.
5355 * tree-streamer-out.c (pack_ts_decl_common_value_fields): For
5356 DECL_BIT_FIELD stream DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead
5357 of DECL_FIELD_ABI_IGNORED.
5358 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
5359 SET_DECL_FIELD_ABI_IGNORED instead of writing to
5360 DECL_FIELD_ABI_IGNORED and for DECL_BIT_FIELD use
5361 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead.
5362 * lto-streamer-out.c (hash_tree): For DECL_BIT_FIELD hash
5363 DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead of DECL_FIELD_ABI_IGNORED.
5365 2021-09-03 liuhongt <hongtao.liu@intel.com>
5368 * config/i386/amxbf16intrin.h : Remove macro check for __AMX_BF16__.
5369 * config/i386/amxint8intrin.h : Remove macro check for __AMX_INT8__.
5370 * config/i386/amxtileintrin.h : Remove macro check for __AMX_TILE__.
5372 2021-09-02 Martin Sebor <msebor@redhat.com>
5374 PR tree-optimization/17506
5376 * tree-ssa-uninit.c (warn_uninit): Remove conditional guarding note.
5378 2021-09-02 Richard Biener <rguenther@suse.de>
5380 * tree-ssa-loop-im.c (fill_always_executed_in_1): Refine
5381 fix for PR78185 and continue processing when leaving
5384 2021-09-02 Jakub Jelinek <jakub@redhat.com>
5386 PR tree-optimization/99591
5387 * match.pd: Demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
5390 2021-09-02 Richard Biener <rguenther@suse.de>
5393 2021-09-02 Richard Biener <rguenther@suse.de>
5395 PR tree-optimization/102155
5396 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
5397 over a part of the RPO array and do not recurse here.
5398 Dump blocks marked as always executed.
5399 (fill_always_executed_in): Walk over the RPO array and
5400 process loops whose header we run into.
5401 (loop_invariant_motion_in_fun): Compute the first RPO
5402 using rev_post_order_and_mark_dfs_back_seme in iteration
5403 order and pass that to fill_always_executed_in.
5405 2021-09-02 liuhongt <hongtao.liu@intel.com>
5407 * config/i386/i386-modes.def (FLOAT_MODE): Define ieee HFmode.
5408 * config/i386/i386.c (enum x86_64_reg_class): Add
5410 (merge_classes): Handle X86_64_SSEHF_CLASS.
5411 (examine_argument): Ditto.
5412 (construct_container): Ditto.
5413 (classify_argument): Ditto, and set HFmode/HCmode to
5415 (function_value_32): Return _FLoat16/Complex Float16 by
5417 (function_value_64): Return _Float16/Complex Float16 by SSE
5419 (ix86_print_operand): Handle CONST_DOUBLE HFmode.
5420 (ix86_secondary_reload): Require gpr as intermediate register
5421 to store _Float16 from sse register when sse4 is not
5423 (ix86_libgcc_floating_mode_supported_p): Enable _FLoat16 under
5425 (ix86_scalar_mode_supported_p): Ditto.
5426 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Defined.
5427 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add HFmode.
5428 (VALID_INT_MODE_P): Add HFmode and HCmode.
5429 * config/i386/i386.md (*pushhf_rex64): New define_insn.
5431 (*movhf_internal): Ditto.
5432 * doc/extend.texi (Half-Precision Floating Point): Documemt
5435 2021-09-02 Richard Biener <rguenther@suse.de>
5437 PR tree-optimization/102155
5438 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
5439 over a part of the RPO array and do not recurse here.
5440 Dump blocks marked as always executed.
5441 (fill_always_executed_in): Walk over the RPO array and
5442 process loops whose header we run into.
5443 (loop_invariant_motion_in_fun): Compute the first RPO
5444 using rev_post_order_and_mark_dfs_back_seme in iteration
5445 order and pass that to fill_always_executed_in.
5447 2021-09-02 YunQiang Su <syq@debian.org>
5450 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
5452 * config/mips/mips.c (mips_module_isa_name): New.
5453 mips_file_start: add .module mipsREV to all asm output
5455 2021-09-01 Jeff Law <jlaw@localhost.localdomain>
5457 PR tree-optimization/102152
5458 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Reduce a vector
5459 comparison to a scalar comparison before calling
5460 update_stmt_if_modified.
5462 2021-09-01 Andrew Pinski <apinski@marvell.com>
5465 * config/aarch64/aarch64.c (aarch64_expand_setmem):
5466 Check STRICT_ALIGNMENT before creating an overlapping
5469 2021-09-01 Martin Sebor <msebor@redhat.com>
5471 * gimple-ssa-warn-access.cc (get_size_range): Add argument.
5472 (check_access): Pass additional argument.
5473 (check_memop_access): Remove template and make a member function.
5474 (maybe_check_dealloc_call): Make a pass_waccess member function.
5475 (class pass_waccess): Add, rename, and remove members.
5476 (pass_waccess::pass_waccess): Adjust to name change.
5477 (pass_waccess::~pass_waccess): Same.
5478 (check_alloca): Make a member function.
5479 (check_alloc_size_call): Same.
5480 (check_strcat): Same.
5481 (check_strncat): Same.
5482 (check_stxcpy): Same.
5483 (check_stxncpy): Same.
5484 (check_strncmp): Same.
5485 (maybe_warn_rdwr_sizes): Rename...
5486 (pass_waccess::maybe_check_access_sizes): ...to this.
5487 (pass_waccess::check_call): Adjust to name changes.
5488 (pass_waccess::maybe_check_dealloc_call): Make a pass_waccess member
5490 (pass_waccess::execute): Adjust to name changes.
5491 * gimple-ssa-warn-access.h (check_memop_access): Remove.
5492 * pointer-query.cc (access_ref::phi): Handle null pointer.
5493 (access_ref::inform_access): Same.
5494 (pointer_query::put_ref): Modify a cached value, not a copy of it.
5495 (pointer_query::dump): New function.
5496 (compute_objsize_r): Avoid overwriting access_ref::bndrng. Cache
5498 * pointer-query.h (pointer_query::dump): Declare.
5499 * tree-ssa-strlen.c (get_range): Simplify. Use function query.
5500 (dump_strlen_info): Use function query.
5501 (printf_strlen_execute): Factor code out into pointer_query::put_ref.
5503 2021-09-01 Thomas Schwinge <thomas@codesourcery.com>
5505 * tree.c (walk_tree_1) <OMP_CLAUSE>: Simplify.
5507 2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
5509 * doc/extend.texi: Document unavailable attribute.
5510 * print-tree.c (print_node): Handle unavailable attribute.
5511 * tree-core.h (struct tree_base): Add a bit to carry unavailability.
5512 * tree.c (error_unavailable_use): New.
5513 * tree.h (TREE_UNAVAILABLE): New.
5514 (error_unavailable_use): New.
5516 2021-09-01 Jakub Jelinek <jakub@redhat.com>
5518 PR tree-optimization/102124
5519 * tree-vect-patterns.c (vect_recog_widen_op_pattern): For ORIG_CODE
5520 MINUS_EXPR, if itype is unsigned with smaller precision than type,
5521 add an extra cast to signed variant of itype to ensure sign-extension.
5523 2021-09-01 Martin Liska <mliska@suse.cz>
5525 * graph.c (draw_cfg_node_succ_edges): Do not color fallthru
5526 edges and rather use colors for TRUE and FALSE edges.
5528 2021-09-01 Richard Biener <rguenther@suse.de>
5530 PR tree-optimization/93491
5531 * tree-ssa-pre.c (compute_avail): Set BB_MAY_NOTRETURN
5532 after processing the stmt itself. Do not consider
5533 pure functions possibly not returning. Properly avoid
5534 adding possibly trapping calls to EXP_GEN when there's
5535 a preceeding possibly not returning call.
5536 * tree-ssa-sccvn.c (vn_reference_may_trap): Conservatively
5539 2021-09-01 Richard Biener <rguenther@suse.de>
5541 PR tree-optimization/102139
5542 * tree-vectorizer.h (vec_base_alignments): Adjust hash-map
5543 type to record a std::pair of the stmt-info and the innermost
5545 (dr_vec_info::group): New member.
5546 * tree-vect-data-refs.c (vect_record_base_alignment): Adjust.
5547 (vect_compute_data_ref_alignment): Verify the recorded
5548 base alignment can be used.
5549 (data_ref_pair): Remove.
5550 (dr_group_sort_cmp): Adjust.
5551 (vect_analyze_data_ref_accesses): Store the group-ID in the
5552 dr_vec_info and operate on a vector of dr_vec_infos.
5554 2021-09-01 YunQiang Su <yunqiang.su@cipunited.com>
5556 * read-md.c (md_reader::handle_enum): support value assignation.
5557 * doc/md.texi: record define_c_enum value assignation support.
5559 2021-09-01 Jakub Jelinek <jakub@redhat.com>
5561 PR tree-optimization/102141
5562 * gimple-ssa-store-merging.c (bswap_view_convert): Add BEFORE
5563 argument. If false, emit stmts after gsi instead of before, and
5565 (bswap_replace): Adjust callers. When converting output of bswap,
5566 emit VIEW_CONVERT prepratation stmts after a copy of gsi instead
5569 2021-09-01 liuhongt <hongtao.liu@intel.com>
5571 * emit-rtl.c (validate_subreg): Get rid of all float-int
5574 2021-09-01 liuhongt <hongtao.liu@intel.com>
5577 2021-08-30 liuhongt <hongtao.liu@intel.com>
5579 * expmed.c (extract_bit_field_1): Make sure we're playing with
5580 integral modes before call extract_integral_bit_field.
5581 (extract_integral_bit_field): Add a parameter of type
5582 scalar_int_mode which corresponds to of tmode.
5583 And call extract_and_convert_fixed_bit_field instead of
5584 extract_fixed_bit_field and convert_extracted_bit_field.
5585 (extract_and_convert_fixed_bit_field): New function, it's a
5586 combination of extract_fixed_bit_field and
5587 convert_extracted_bit_field.
5589 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
5591 * tree.c (walk_tree_1) <OMP_CLAUSE_TILE>: Handle three operands.
5593 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
5595 * omp-general.h (omp_is_reference): Rename to...
5596 (omp_privatize_by_reference): ... this. Adjust all users...
5597 * omp-general.c: ... here, ...
5598 * gimplify.c: ... here, ...
5599 * omp-expand.c: ... here, ...
5600 * omp-low.c: ... here.
5602 2021-08-31 Martin Sebor <msebor@redhat.com>
5604 * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Test
5605 pointer element for equality to zero, not that of the cotaining
5608 2021-08-31 Martin Sebor <msebor@redhat.com>
5610 * gcc-rich-location.h (gcc_rich_location): Make ctor explicit.
5612 2021-08-31 Martin Sebor <msebor@redhat.com>
5614 * function.h (function): Add comments.
5615 (get_range_query): Same. Add attribute returns nonnull.
5617 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
5619 * expr.c (convert_modes): Don't use subreg_promoted_mode on a
5620 SUBREG if it can't be guaranteed to a SUBREG_PROMOTED_VAR_P set.
5621 Instead use the standard (safer) is_a <scalar_int_mode> idiom.
5623 2021-08-31 Jeff Law <jlaw@localhost.localdomain>
5625 * config.gcc (cris-*-elf, cris-*-none): Remove dbxelf.h from
5627 (m32r-*-elf, m32rle-*-elf, m32r-*-linux): Likewise.
5628 (mn10300-*-*, am33_2.0-*-linux*): Likewise.
5629 (xtensa*-*-elf, xtensa*-*-linux, xtensa*-*-uclinux): Likewise.
5630 (m32c-*-elf*, m32c-*-rtems*): Likewise.
5631 * config/cris/cris.h (DBX_NO_XREFS): Remove.
5632 (DBX_CONTIN_LENGTH, DBX_CONTIN_CHAR): Likewise.
5633 * config/m32r/m32r.h (DBXOUT_SOURCE_LINE): Likewise.
5634 (DBX_DEBUGGING_INFO, DBX_CONTIN_LENGTH): Likewise.
5635 * config/mn10300/mn10300.h (DEFAULT_GDB_EXTENSIONS): Likewise.
5636 * config/mn10300/linux.h (DBX_REGISTER_NAMES): Likewise.
5638 2021-08-31 Marcel Vollweiler <marcel@codesourcery.com>
5640 * gimplify.c (gimplify_scan_omp_clauses): Error handling. 'ancestor' only
5641 allowed on target constructs and only with particular other clauses.
5642 * omp-expand.c (expand_omp_target): Output of 'sorry, not supported' if
5644 * omp-low.c (check_omp_nesting_restrictions): Error handling. No nested OpenMP
5645 structs when 'ancestor' is used.
5646 (scan_omp_1_stmt): No usage of OpenMP runtime routines in a target region when
5648 * tree-pretty-print.c (dump_omp_clause): Append 'ancestor'.
5649 * tree.h (OMP_CLAUSE_DEVICE_ANCESTOR): Define macro.
5651 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
5653 * expr.c (convert_modes): Preserve SUBREG_PROMOTED_VAR_P when
5654 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
5656 * simplify-rtx.c (simplify_unary_operation_1) [SIGN_EXTEND]:
5657 Likewise, preserve SUBREG_PROMOTED_VAR_P when creating a (wider)
5658 partial subreg from a SUBREG_PROMOTED_VAR_P subreg. Generate
5659 SIGN_EXTEND of the SUBREG_REG when a subreg would be paradoxical.
5660 [ZERO_EXTEND]: Likewise, preserve SUBREG_PROMOTED_VAR_P when
5661 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
5662 subreg. Generate ZERO_EXTEND of the SUBREG_REG when a subreg
5663 would be paradoxical.
5665 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
5667 * combine.c (combine_simplify_rtx): Avoid converting an explicit
5668 TRUNCATE into a lowpart SUBREG on !TRULY_NOOP_TRUNCATION targets.
5669 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
5671 2021-08-31 Richard Biener <rguenther@suse.de>
5673 PR tree-optimization/102142
5674 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Fix
5675 condition under which to unset the visited flag.
5677 2021-08-31 Richard Biener <rguenther@suse.de>
5679 PR middle-end/102129
5680 * tree-ssa-ter.c (find_replaceable_in_bb): Do not move
5681 possibly trapping expressions across calls.
5683 2021-08-31 Jakub Jelinek <jakub@redhat.com>
5685 PR tree-optimization/102134
5686 * tree-ssa-ccp.c (bit_value_binop) <case RSHIFT_EXPR>: If sgn is
5687 UNSIGNED and r1val | r1mask has MSB set, ensure lzcount doesn't
5690 2021-08-31 Andrew Pinski <apinski@marvell.com>
5693 * collect-utils.c (setup_signals): New declaration.
5694 * collect-utils.h (setup_signals): New function.
5695 * collect2.c (handler): Delete.
5696 (main): Instead of manually setting up the signals,
5697 just call setup_signals.
5698 * lto-wrapper.c (main): Likewise.
5700 2021-08-31 Andrew Pinski <apinski@marvell.com>
5703 * config/i386/i386-protos.h (x86_output_aligned_bss):
5704 Change align argument to unsigned type.
5705 (x86_elf_aligned_decl_common): Likewise.
5706 * config/i386/i386.c (x86_elf_aligned_decl_common): Likewise.
5707 (x86_output_aligned_bss): Likewise.
5709 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
5711 * config/mips/mips.c (mips_module_isa_name): New.
5712 mips_file_start: add .module mipsREV to all asm output
5714 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
5716 * config/mips/mips.h (struct mips_cpu_info): define enum mips_isa;
5717 use enum instead of int for 'isa' member.
5718 * config.gcc, config/mips/mips.c, config/mips/mips-cpus.def,
5719 config/mips/netbsd.h: replace hardcoded numbers with enum.
5721 2021-08-31 liuhongt <hongtao.liu@intel.com>
5723 * config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
5724 define_split to define_insn_and_split.
5725 (*avx2_eq<mode>3): Removed.
5726 (<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
5727 (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
5728 (*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
5730 (*avx2_gt<mode>3): Removed.
5731 (<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
5732 define_insn to define_expand, and adjust pattern.
5733 (UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.
5735 2021-08-30 David Malcolm <dmalcolm@redhat.com>
5738 * Makefile.in (ANALYZER_OBJS): Add analyzer/call-info.o.
5740 2021-08-30 Jason Merrill <jason@redhat.com>
5742 * doc/invoke.texi: Document -Wmissing-requires.
5744 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
5746 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Remove
5747 TARGET_EXTRA_BUILTINS guard.
5749 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
5751 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Change
5752 initialization of V2DI_type_node and unsigned_V2DI_type_node.
5754 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
5756 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Use the new
5757 decl when new_builtins_are_live.
5758 * config/rs6000/rs6000-builtin-new.def (__builtin_cfstring): New
5761 2021-08-30 Pat Haugen <pthaugen@linux.ibm.com>
5763 * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add
5764 OPTION_MASK_P10_FUSION_2STORE.
5765 (POWERPC_MASKS): Likewise.
5766 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
5767 store fusion for Power10.
5768 (is_fusable_store): New.
5769 (power10_sched_reorder): Likewise.
5770 (rs6000_sched_reorder): Do Power10 specific reordering.
5771 (rs6000_sched_reorder2): Likewise.
5772 * config/rs6000/rs6000.opt: Add new option.
5774 2021-08-30 Richard Biener <rguenther@suse.de>
5776 PR tree-optimization/102128
5777 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
5778 Move scanning for if-converted scalar code to the caller
5779 and instead delay clearing the visited flag for profitable
5781 (vect_slp_region): Cost all subgraphs before scheduling.
5782 For if-converted BB vectorization scan for scalar COND_EXPRs
5783 and do not vectorize if any found and the cost model is
5786 2021-08-30 Richard Biener <rguenther@suse.de>
5788 * common.opt (fexceptions): Mark
5789 EnabledBy(fnon-call-exceptions).
5790 * doc/invoke.texi (fnon-call-exceptions): Document this
5791 enables -fexceptions.
5793 2021-08-30 Sebastian Huber <sebastian.huber@embedded-brains.de>
5795 * tsystem.h (abort): Define abort() if inhibit_libc is defined and it
5796 is not already defined.
5798 2021-08-30 liuhongt <hongtao.liu@intel.com>
5800 * expmed.c (extract_bit_field_1): Make sure we're playing with
5801 integral modes before call extract_integral_bit_field.
5802 (extract_integral_bit_field): Add a parameter of type
5803 scalar_int_mode which corresponds to of tmode.
5804 And call extract_and_convert_fixed_bit_field instead of
5805 extract_fixed_bit_field and convert_extracted_bit_field.
5806 (extract_and_convert_fixed_bit_field): New function, it's a
5807 combination of extract_fixed_bit_field and
5808 convert_extracted_bit_field.
5810 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
5812 * config/darwin.c (darwin_libc_has_function): Do not run
5813 the checks for x86 or modern Darwin. Make sure that there
5814 is a value set for darwin_macosx_version_min before testing.
5816 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
5818 * config/i386/darwin.h (CLEAR_INSN_CACHE): New.
5820 2021-08-28 Jan Hubicka <hubicka@ucw.cz>
5822 * ipa-modref-tree.h (modref_access_node::merge): Break out
5823 logic combining offsets and logic merging ranges to ...
5824 (modref_access_node::combined_offsets): ... here
5825 (modref_access_node::update2): ... here
5826 (modref_access_node::closer_pair_p): New member function.
5827 (modref_access_node::forced_merge): New member function.
5828 (modre_ref_node::insert): Do merging when table is full.
5830 2021-08-28 YunQiang Su <yunqiang.su@cipunited.com>
5833 * config.gcc: MIPS: use N64 ABI by default if the triple end
5834 with -gnuabi64, which is used by Debian since 2013.
5836 2021-08-28 Alexandre Oliva <oliva@adacore.com>
5838 * ipa-modref.c (analyze_function): Skip debug stmts.
5839 * tree-inline.c (estimate_num_insn): Consider builtins even
5840 without a cgraph_node.
5842 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
5844 * config/h8300/bitfield.md (cstore<mode>4): Remove expander.
5845 * config/h8300/h8300.c (h8300_expand_branch): Remove function.
5846 * config/h8300/h8300-protos.h (h8300_expadn_branch): Remove prototype.
5847 * config/h8300/h8300.md (eqne): New code iterator.
5848 (geultu, geultu_to_c): Similarly.
5849 * config/h8300/testcompare.md (cstore<mode>4): Dummy expander.
5850 (store_c_<mode>, store_c_i_<mode>): New define_insn_and_splits
5851 (cmp<mode>_c): New pattern
5853 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
5855 * tree-ssa-dom.c (reduce_vector_comparison_to_scalar_comparison): New
5857 (dom_opt_dom_walker::optimize_stmt): Use it.
5859 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
5861 * config/darwin.c (finalize_ctors): Add a section-start linker-
5863 (finalize_dtors): Likewise.
5864 * config/darwin.h (MIN_LD64_INIT_TERM_START_LABELS): New.
5866 2021-08-27 Bill Schmidt <wschmidt@linux.ibm.com>
5868 * config/rs6000/rs6000-call.c (rs6000-builtins.h): New #include.
5869 (rs6000_init_builtins): Call rs6000_init_generated_builtins. Skip the
5870 old initialization logic when new builtins are enabled.
5871 * config/rs6000/rs6000-gen-builtins.c (write_decls): Rename
5872 rs6000_autoinit_builtins to rs6000_init_generated_builtins.
5873 (write_init_file): Likewise.
5875 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
5877 * configure.ac (darwin2[[0-9]]* | darwin19*): Alter use of
5878 gcc_GAS_CHECK_FEATURE to remove an extraneous parameter.
5879 (amdgcn-* | gcn-*) Likewise.
5881 2021-08-27 Anthony Sharp <anthonysharp15@gmail.com>
5883 * symbol-summary.h: Added missing template keyword.
5885 2021-08-27 Richard Biener <rguenther@suse.de>
5887 PR tree-optimization/45178
5888 * tree-ssa-dce.c (find_obviously_necessary_stmts): For
5889 infinite loops without exit do not mark control dependent
5890 edges of the latch necessary.
5892 2021-08-27 konglin1 <lingling.kong@intel.com>
5895 * config/i386/sse.md: (<avx512>scattersi<mode>): Add mask operand to
5897 (<avx512>scattersi<mode>): Likewise.
5898 (*avx512f_scattersi<VI48F:mode>): Merge mask operand to set_dest.
5899 (*avx512f_scatterdi<VI48F:mode>): Likewise
5901 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
5903 * config/rs6000/rs6000.c (rs6000_builtin_md_vectorized_function): Add
5904 support for built-in functions MISC_BUILTIN_DIVWE, MISC_BUILTIN_DIVWEU,
5905 MISC_BUILTIN_DIVDE, MISC_BUILTIN_DIVDEU, P10_BUILTIN_CFUGED,
5906 P10_BUILTIN_CNTLZDM, P10_BUILTIN_CNTTZDM, P10_BUILTIN_PDEPD and
5907 P10_BUILTIN_PEXTD on Power10.
5909 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
5911 * config/rs6000/rs6000-call.c (builtin_function_type): Add unsigned
5912 signedness for some Power10 bifs.
5914 2021-08-27 David Edelsohn <dje.gcc@gmail.com>
5917 * config/rs6000/rs6000.c (rs6000_adjust_field_align): Use
5918 computed alignment if the entire struct has attribute packed.
5920 2021-08-27 liuhongt <hongtao.liu@intel.com>
5924 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
5925 IX86_BUILTIN_SHUFPD512, IX86_BUILTIN_SHUFPS512,
5926 IX86_BUILTIN_SHUFPD256, IX86_BUILTIN_SHUFPS,
5927 IX86_BUILTIN_SHUFPS256.
5928 (ix86_masked_all_ones): New function.
5930 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
5932 * config/i386/i386.md (*btr<mode>_1): Call force_reg unconditionally.
5933 (conditional moves with memory inputs splitters): Ditto.
5934 * config/i386/sse.md (one_cmpl<mode>2): Simplify.
5936 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
5938 * ipa-modref-tree.h (modref_access_node::try_merge_with): Restart
5939 search after merging.
5941 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
5943 * config/rs6000/rs6000-overload.def: Add remaining overloads.
5945 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
5947 * config/rs6000/rs6000-builtin-new.def: Add cell stanza.
5949 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
5951 * config/rs6000/rs6000-builtin-new.def: Add ieee128-hw, dfp,
5952 crypto, and htm stanzas.
5954 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
5956 * config/rs6000/rs6000-builtin-new.def: Add mma stanza.
5958 2021-08-26 Martin Sebor <msebor@redhat.com>
5960 * tree-ssa-uninit.c (warn_uninit): Refactor and simplify.
5961 (warn_uninit_phi_uses): Remove argument from calls to warn_uninit.
5962 (warn_uninitialized_vars): Same. Reduce visibility of locals.
5963 (warn_uninitialized_phi): Same.
5965 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
5967 * tree-ssa-ccp.c (get_individual_bits): Helper function to
5968 extract the individual bits from a widest_int constant (mask).
5969 (gray_code_bit_flips): New read-only table for effiently
5970 enumerating permutations/combinations of bits.
5971 (bit_value_binop) [LROTATE_EXPR, RROTATE_EXPR]: Handle rotates
5972 by unknown counts that are guaranteed less than the target
5973 precision and four or fewer unknown bits by enumeration.
5974 [LSHIFT_EXPR, RSHIFT_EXPR]: Likewise, also handle shifts by
5975 enumeration under the same conditions. Handle remaining
5976 shifts as a mask based upon the minimum possible shift value.
5978 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
5979 Richard Biener <rguenther@suse.de>
5981 * match.pd (shift transformations): Remove a redundant
5982 !POINTER_TYPE_P check.
5984 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
5987 * config/i386/i386.md (cmove reg-reg move elimination peephole2s):
5988 Set all_regs to true in the call to replace_rtx.
5990 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
5992 * ipa-modref-tree.c (test_insert_search_collapse): Update test.
5993 * ipa-modref-tree.h (modref_base_node::insert): Be smarter when
5994 hiting --param modref-max-refs limit.
5995 (modref_tree:insert_base): Be smarter when hitting
5996 --param modref-max-bases limit. Add new parameter REF.
5997 (modref_tree:insert): Update.
5998 (modref_tree:merge): Update.
5999 * ipa-modref.c (read_modref_records): Update.
6001 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
6003 * params.opt: (modref-max-adjustments): Add full stop.
6005 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
6007 * ipa-modref-tree.h (modref_ref_node::verify): New member
6009 (modref_ref_node::insert): Use it.
6010 (modref_ref_node::try_mere_with): Fix off by one error.
6012 2021-08-26 Martin Liska <mliska@suse.cz>
6013 Stefan Kneifel <stefan.kneifel@bluewin.ch>
6015 * cgraph.h (create_version_clone_with_body): Add new parameter.
6016 * cgraphclones.c: Likewise.
6017 * multiple_target.c (create_dispatcher_calls): Do not use
6019 (create_target_clone): Likewise here.
6021 2021-08-26 Jonathan Yong <10walls@gmail.com>
6023 * doc/extend.texi: Add note about reserved priorities
6024 to the constructor attribute.
6026 2021-08-25 Martin Sebor <msebor@redhat.com>
6028 * gimple-range-cache.cc (ssa_global_cache::dump): Avoid printing
6029 range table header alone.
6030 * gimple-range.cc (gimple_ranger::export_global_ranges): Same.
6032 2021-08-25 Jan Hubicka <hubicka@ucw.cz>
6034 * doc/invoke.texi: Document --param modref-max-adjustments.
6035 * ipa-modref-tree.c (test_insert_search_collapse): Update.
6036 (test_merge): Update.
6037 * ipa-modref-tree.h (struct modref_access_node): Add adjustments;
6038 (modref_access_node::operator==): Fix handling of access ranges.
6039 (modref_access_node::contains): Constify parameter; handle also
6040 mismatched parm offsets.
6041 (modref_access_node::update): New function.
6042 (modref_access_node::merge): New function.
6043 (unspecified_modref_access_node): Update constructor.
6044 (modref_ref_node::insert_access): Add record_adjustments parameter;
6046 (modref_ref_node::try_merge_with): New private function.
6047 (modref_tree::insert): New record_adjustments parameter.
6048 (modref_tree::merge): New record_adjustments parameter.
6049 (modref_tree::copy_from): Update.
6050 * ipa-modref.c (dump_access): Dump adjustments field.
6051 (get_access): Update constructor.
6052 (record_access): Update call of insert.
6053 (record_access_lto): Update call of insert.
6054 (merge_call_side_effects): Add record_adjustments parameter.
6055 (get_access_for_fnspec): Update.
6056 (process_fnspec): Update.
6057 (analyze_call): Update.
6058 (analyze_function): Update.
6059 (read_modref_records): Update.
6060 (ipa_merge_modref_summary_after_inlining): Update.
6061 (propagate_unknown_call): Update.
6062 (modref_propagate_in_scc): Update.
6063 * params.opt (param-max-modref-adjustments=): New.
6065 2021-08-25 Michael Meissner <meissner@linux.ibm.com>
6067 * config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
6069 (xxspltiw_v4si): Use vecperm type attribute.
6070 (xxspltiw_v4si_inst): Use vecperm type attribute.
6071 (xxspltiw_v4sf_inst): Likewise.
6072 (xxspltidp_v2df): Use vecperm type attribute. Use
6073 UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
6074 (xxspltidp_v2df_inst): Likewise.
6075 (xxsplti32dx_v4si): Use vecperm type attribute.
6076 (xxsplti32dx_v4si_inst): Likewise.
6077 (xxsplti32dx_v4sf_inst): Likewise.
6078 (xxblend_<mode>): Likewise.
6079 (xxpermx): Likewise.
6080 (xxpermx_inst): Likewise.
6083 2021-08-25 Lewis Hyatt <lhyatt@gmail.com>
6086 * coretypes.h (typedef diagnostic_input_charset_callback): Declare.
6087 * diagnostic.c (diagnostic_initialize_input_context): New function.
6088 * diagnostic.h (diagnostic_initialize_input_context): Declare.
6089 * input.c (default_charset_callback): New function.
6090 (file_cache::initialize_input_context): New function.
6091 (file_cache_slot::create): Added ability to convert the input
6092 according to the input context.
6093 (file_cache::file_cache): Initialize the new input context.
6094 (class file_cache_slot): Added new m_alloc_offset member.
6095 (file_cache_slot::file_cache_slot): Initialize the new member.
6096 (file_cache_slot::~file_cache_slot): Handle potentially offset buffer.
6097 (file_cache_slot::maybe_grow): Likewise.
6098 (file_cache_slot::needs_read_p): Handle NULL fp, which is now possible.
6099 (file_cache_slot::get_next_line): Likewise.
6100 * input.h (class file_cache): Added input context member.
6102 2021-08-25 Richard Biener <rguenther@suse.de>
6104 PR tree-optimization/102046
6105 * tree-vect-slp.c (vect_build_slp_tree_2): Conservatively
6106 update ->any_pattern when swapping operands.
6108 2021-08-25 Hongyu Wang <hongyu.wang@intel.com>
6111 * config/i386/i386.c (ix86_live_on_entry): Adjust comment.
6112 (ix86_decompose_address): Remove retval check for ASHIFT,
6113 allow non-canonical zero extend if AND mask covers ASHIFT
6115 (ix86_legitimate_address_p): Adjust condition for decompose.
6116 (ix86_rtx_costs): Adjust cost for lea with non-canonical
6118 Co-Authored by: Uros Bizjak <ubizjak@gmail.com>
6120 2021-08-25 Jiufu Guo <guojiufu@linux.ibm.com>
6122 PR tree-optimization/101145
6123 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
6125 (number_of_iterations_lt): Invoke above function.
6126 (adjust_cond_for_loop_until_wrap):
6127 Merge to number_of_iterations_until_wrap.
6128 (number_of_iterations_cond): Update invokes for
6129 adjust_cond_for_loop_until_wrap and number_of_iterations_lt.
6131 2021-08-25 konglin1 <lingling.kong@intel.com>
6134 * config/i386/avx512dqintrin.h (_mm512_fpclass_ps_mask): Fix
6136 (_mm512_mask_fpclass_ps_mask): Ditto.
6138 2021-08-25 Kewen Lin <linkw@linux.ibm.com>
6140 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Remove.
6141 (vec_unpacku_hi_v8hi): Likewise.
6142 (vec_unpacku_lo_v16qi): Likewise.
6143 (vec_unpacku_lo_v8hi): Likewise.
6144 (vec_unpacku_hi_<VP_small_lc>): New define_expand.
6145 (vec_unpacku_lo_<VP_small_lc>): Likewise.
6147 2021-08-24 David Edelsohn <dje.gcc@gmail.com>
6149 * config/rs6000/aix.h (SYSTEM_IMPLICIT_EXTERN_C): Delete.
6150 * config/rs6000/aix71.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
6151 * config/rs6000/aix72.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
6152 * config/rs6000/aix73.h (TARGET_AIX_VERSION): Increase to 73.
6154 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
6156 PR middle-end/102031
6157 * simplify-rtx.c (simplify_truncation): When comparing precisions
6158 use "subreg_prec" variable, not "subreg_mode".
6160 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
6162 * config/rs6000/rs6000-builtin-new.def: Add power10 and power10-64
6165 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
6167 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Initialize
6168 various pointer type nodes.
6169 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add enum
6170 values for various pointer types.
6171 (ptr_V16QI_type_node): New macro.
6172 (ptr_V1TI_type_node): New macro.
6173 (ptr_V2DI_type_node): New macro.
6174 (ptr_V2DF_type_node): New macro.
6175 (ptr_V4SI_type_node): New macro.
6176 (ptr_V4SF_type_node): New macro.
6177 (ptr_V8HI_type_node): New macro.
6178 (ptr_unsigned_V16QI_type_node): New macro.
6179 (ptr_unsigned_V1TI_type_node): New macro.
6180 (ptr_unsigned_V8HI_type_node): New macro.
6181 (ptr_unsigned_V4SI_type_node): New macro.
6182 (ptr_unsigned_V2DI_type_node): New macro.
6183 (ptr_bool_V16QI_type_node): New macro.
6184 (ptr_bool_V8HI_type_node): New macro.
6185 (ptr_bool_V4SI_type_node): New macro.
6186 (ptr_bool_V2DI_type_node): New macro.
6187 (ptr_bool_V1TI_type_node): New macro.
6188 (ptr_pixel_type_node): New macro.
6189 (ptr_intQI_type_node): New macro.
6190 (ptr_uintQI_type_node): New macro.
6191 (ptr_intHI_type_node): New macro.
6192 (ptr_uintHI_type_node): New macro.
6193 (ptr_intSI_type_node): New macro.
6194 (ptr_uintSI_type_node): New macro.
6195 (ptr_intDI_type_node): New macro.
6196 (ptr_uintDI_type_node): New macro.
6197 (ptr_intTI_type_node): New macro.
6198 (ptr_uintTI_type_node): New macro.
6199 (ptr_long_integer_type_node): New macro.
6200 (ptr_long_unsigned_type_node): New macro.
6201 (ptr_float_type_node): New macro.
6202 (ptr_double_type_node): New macro.
6203 (ptr_long_double_type_node): New macro.
6204 (ptr_dfloat64_type_node): New macro.
6205 (ptr_dfloat128_type_node): New macro.
6206 (ptr_ieee128_type_node): New macro.
6207 (ptr_ibm128_type_node): New macro.
6208 (ptr_vector_pair_type_node): New macro.
6209 (ptr_vector_quad_type_node): New macro.
6210 (ptr_long_long_integer_type_node): New macro.
6211 (ptr_long_long_unsigned_type_node): New macro.
6213 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
6215 * config/rs6000/rs6000-builtin-new.def: Add power9-vector, power9,
6216 and power9-64 stanzas.
6218 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
6219 Tom de Vries <tdevries@suse.de>
6221 * config.gcc (nvptx-*-*): Define {c,c++}_target_objs.
6222 * config/nvptx/nvptx-protos.h (nvptx_cpu_cpp_builtins): Prototype.
6223 * config/nvptx/nvptx.h (TARGET_CPU_CPP_BUILTINS): Implement with
6224 a call to the new nvptx_cpu_cpp_builtins function in nvptx-c.c.
6225 * config/nvptx/t-nvptx (nvptx-c.o): New rule.
6226 * config/nvptx/nvptx-c.c: New source file.
6227 (nvptx_cpu_cpp_builtins): Move implementation here.
6229 2021-08-24 Martin Sebor <msebor@redhat.com>
6231 PR middle-end/101600
6232 PR middle-end/101977
6233 * gimple-ssa-warn-access.cc (maybe_warn_for_bound): Tighten up
6234 the phrasing of a warning.
6235 (check_access): Use the remaining size after subtracting any offset
6236 rather than the whole object size.
6237 * pointer-query.cc (access_ref::get_ref): Clear BASE0 flag if it's
6238 clear for any nonnull PHI argument.
6239 (compute_objsize): Clear argument.
6241 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
6243 * config/rs6000/rs6000-builtin-new.def: Add power8-vector stanza.
6245 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
6247 * config/rs6000/rs6000-builtin-new.def: Add power7 and power7-64
6250 2021-08-24 Andrew MacLeod <amacleod@redhat.com>
6252 * value-relation.cc (rr_transitive_table): New.
6253 (relation_transitive): New.
6254 (value_relation::swap): Remove.
6255 (value_relation::apply_transitive): New.
6256 (relation_oracle::relation_oracle): Allocate a new tmp bitmap.
6257 (relation_oracle::register_relation): Call register_transitives.
6258 (relation_oracle::register_transitives): New.
6259 * value-relation.h (relation_oracle): Add new temporary bitmap and
6262 2021-08-24 H.J. Lu <hjl.tools@gmail.com>
6265 * config/i386/i386-expand.c (ix86_expand_vector_move): Broadcast
6266 from integer to a pseudo vector register.
6268 2021-08-24 Richard Biener <rguenther@suse.de>
6270 PR tree-optimization/100089
6271 * tree-vectorizer.h (vect_slp_bb): Rename to ...
6272 (vect_slp_if_converted_bb): ... this and get the original
6273 loop as new argument.
6274 * tree-vectorizer.c (try_vectorize_loop_1): Revert previous fix,
6275 pass original loop to vect_slp_if_converted_bb.
6276 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
6277 If orig_loop was passed scan the not vectorized stmts
6278 for COND_EXPRs and force not profitable if found.
6279 (vect_slp_region): Pass down all SLP instances to costing
6280 if orig_loop was specified.
6281 (vect_slp_bbs): Pass through orig_loop.
6282 (vect_slp_bb): Rename to ...
6283 (vect_slp_if_converted_bb): ... this and get the original
6284 loop as new argument.
6285 (vect_slp_function): Adjust.
6287 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
6290 * config/arm/arm.md (attribute arch): Add fix_vlldm.
6291 (arch_enabled): Use it.
6292 * config/arm/vfp.md (lazy_store_multiple_insn): Add alternative to
6293 use when erratum mitigation is needed.
6295 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
6298 * config/arm/arm.opt (mfix-cmse-cve-2021-35465): New option.
6299 * doc/invoke.texi (Arm Options): Document it.
6300 * config/arm/arm-cpus.in (quirk_vlldm): New feature bit.
6301 (ALL_QUIRKS): Add quirk_vlldm.
6302 (cortex-m33): Add quirk_vlldm.
6303 (cortex-m35p, cortex-m55): Likewise.
6304 * config/arm/arm.c (arm_option_override): Enable fix_vlldm if
6305 targetting an affected CPU and not explicitly controlled on
6308 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
6310 * config/arm/vfp.md (lazy_store_multiple_insn): Rewrite as valid RTL.
6311 (lazy_load_multiple_insn): Likewise.
6313 2021-08-24 liuhongt <hongtao.liu@intel.com>
6316 * config/i386/sse.md (<avx512>_vternlog<mode><sd_maskz_name>):
6317 Enable avx512 embedded broadcast.
6318 (*<avx512>_vternlog<mode>_all): Ditto.
6319 (<avx512>_vternlog<mode>_mask): Ditto.
6321 2021-08-24 liuhongt <hongtao.liu@intel.com>
6324 * config/i386/i386.c (ix86_rtx_costs): Define cost for
6326 * config/i386/i386.h (STRIP_UNARY): New macro.
6327 * config/i386/predicates.md (reg_or_notreg_operand): New
6329 * config/i386/sse.md (*<avx512>_vternlog<mode>_all): New define_insn.
6330 (*<avx512>_vternlog<mode>_1): New pre_reload
6331 define_insn_and_split.
6332 (*<avx512>_vternlog<mode>_2): Ditto.
6333 (*<avx512>_vternlog<mode>_3): Ditto.
6334 (any_logic1,any_logic2): New code iterator.
6335 (logic_op): New code attribute.
6336 (ternlogsuffix): Extend to VNxDF and VNxSF.
6338 2021-08-24 Richard Biener <rguenther@suse.de>
6340 * doc/invoke.texi (vect-inner-loop-cost-factor): Adjust.
6341 * params.opt (--param vect-inner-loop-cost-factor): Adjust
6343 * tree-vect-loop.c (vect_analyze_loop_form): Initialize
6344 inner_loop_cost_factor to the minimum of the estimated number
6345 of iterations of the inner loop and vect-inner-loop-cost-factor.
6347 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
6348 Richard Biener <rguenther@suse.de>
6350 * config/i386/i386-features.c (compute_convert_gain): Provide
6351 more accurate values for CONST_INT, when optimizing for size.
6352 * config/i386/i386.c (COSTS_N_BYTES): Move definition from here...
6353 * config/i386/i386.h (COSTS_N_BYTES): to here.
6355 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
6356 Jakub Jelinek <jakub@redhat.com>
6358 PR middle-end/102029
6359 * match.pd (shift transformations): Add an additional check for
6360 !POINTER_TYPE_P in the recently added left shift transformation.
6362 2021-08-24 liuhongt <hongtao.liu@intel.com>
6364 PR tree-optimization/100089
6365 * tree-vectorizer.c (try_vectorize_loop_1): Disable slp in
6366 loop vectorizer when cost model is very-cheap.
6368 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
6370 * config/rs6000/rs6000-gen-builtins.c (parse_bif_entry): Don't call
6371 asprintf, which is not available on AIX.
6373 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
6375 * config.gcc (target_gtfiles): Add ./rs6000-builtins.h.
6376 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Set.
6378 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
6380 * config.gcc (powerpc*-*-*): Add rs6000-builtins.o to extra_objs.
6381 * config/rs6000/rs6000-gen-builtins.c (main): Close init_file
6383 * config/rs6000/t-rs6000 (rs6000-gen-builtins.o): New target.
6384 (rbtree.o): Likewise.
6385 (rs6000-gen-builtins): Likewise.
6386 (rs6000-builtins.c): Likewise.
6387 (rs6000-builtins.h): Likewise.
6388 (rs6000.o): Add dependency.
6389 (EXTRA_HEADERS): Add rs6000-vecdefines.h.
6390 (rs6000-vecdefines.h): New target.
6391 (rs6000-builtins.o): Likewise.
6392 (rs6000-call.o): Add rs6000-builtins.h as a dependency.
6393 (rs6000-c.o): Likewise.
6395 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
6398 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace):
6399 Diagnose buffer overrun.
6400 (safe_inc_pos): Fix overrun detection.
6401 (match_identifier): Diagnose buffer overrun.
6402 (match_integer): Likewise.
6403 (match_to_right_bracket): Likewise.
6405 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
6407 * ipa-modref-tree.h (modref_access_node::range_info_useful_p):
6408 Improve range compare.
6409 (modref_access_node::contains): New member function.
6410 (modref_access_node::search): Remove.
6411 (modref_access_node::insert): Be smarter about subaccesses.
6413 2021-08-23 Thomas Schwinge <thomas@codesourcery.com>
6415 * config/i386/i386-options.c (ix86_omp_device_kind_arch_isa)
6416 <omp_device_arch> [ACCEL_COMPILER]: Match "intel_mic".
6417 * config/i386/t-omp-device (omp-device-properties-i386) <arch>:
6420 2021-08-23 Jeff Law <jlaw@localhost.localdomain>
6422 * config/h8300/h8300-protos.h (h8300_expand_epilogue): Add new
6424 * config/h8300/jumpcall.md (call, call_value): Restrict to
6425 !SIBLING_CALL_P cases.
6426 (subcall, sibcall_value): New patterns & expanders.
6427 * config/h8300/proepi.md (epilogue): Pass new argument to
6428 h8300_expand_epilogue.
6429 (sibcall_epilogue): New expander.
6430 * config/h8300/h8300.c (h8300_expand_epilogue): Handle sibcall
6432 (h8300_ok_for_sibcall_p): New function.
6433 (TARGET_FUNCTION_OK_FOR_SIBCALL): define.
6435 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
6437 * simplify-rtx.c (simplify_unary_operation_1): [TRUNCATE]:
6438 Handle case where the operand is already the desired mode.
6440 2021-08-23 Richard Biener <rguenther@suse.de>
6443 * tree-ssa-structalias.c (ipa_pta_execute): Check in_other_partition
6444 in addition to has_gimple_body.
6446 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
6448 PR middle-end/101949
6449 * ipa-modref.c (analyze_ssa_name_flags): Fix merging of
6452 2021-08-23 Martin Liska <mliska@suse.cz>
6454 * doc/invoke.texi: Put the option out of -mxl-mode-app-model
6457 2021-08-23 Richard Biener <rguenther@suse.de>
6459 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6460 Properly scale the inner loop cost only once.
6462 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
6464 * tree-ssa-ccp.c (bit_value_binop) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]:
6465 Provide bounds for unsigned (and signed with non-negative operands)
6466 division and modulus.
6468 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
6470 * simplify-rtx.c (simplify_truncation): Generalize simplification
6471 of (truncate:A (subreg:B X)).
6472 (simplify_unary_operation_1) [FLOAT_TRUNCATE, FLOAT_EXTEND,
6473 SIGN_EXTEND, ZERO_EXTEND]: Handle cases where the operand
6474 already has the desired machine mode.
6475 (test_scalar_int_ops): Add tests that useless extensions and
6476 truncations are optimized away.
6477 (test_scalar_int_ext_ops): New self-test function to confirm
6478 that truncations of extensions are correctly simplified.
6479 (test_scalar_int_ext_ops2): New self-test function to check
6480 truncations of truncations, extensions of extensions, and
6481 truncations of extensions.
6482 (test_scalar_ops): Call the above two functions with a
6483 representative sampling of integer machine modes.
6485 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
6487 * match.pd (shift transformations): Change the sign of an
6488 LSHIFT_EXPR if it reduces the number of explicit conversions.
6490 2021-08-23 Jakub Jelinek <jakub@redhat.com>
6492 PR tree-optimization/86723
6493 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
6494 cast64_to_32 argument, set *cast64_to_32 to false, unless n is
6495 non-memory permutation of 64-bit src which only has bytes of
6496 0 or [5..8] and n->range is 4.
6497 (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
6498 find_bswap_or_nop_finalize caller, support bswap with some bytes
6499 zeroed, as long as at least two bytes are not zeroed.
6500 (bswap_replace): Add mask argument and handle masking of bswap
6502 (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
6503 caller, punt if cast64_to_32 or mask is not all ones.
6504 (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
6505 caller, for now punt if cast64_to_32.
6507 2021-08-23 Richard Biener <rguenther@suse.de>
6509 PR tree-optimization/79334
6510 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Record
6511 a type also for COMPONENT_REFs.
6512 (vn_reference_may_trap): Check ARRAY_REF with constant index
6513 against the array domain.
6515 2021-08-23 liuhongt <hongtao.liu@intel.com>
6518 * config/i386/sse.md (*avx512f_pshufb_truncv8hiv8qi_1): Add
6519 TARGET_AVX512BW to condition.
6521 2021-08-23 Jakub Jelinek <jakub@redhat.com>
6524 * dwarf2out.c (gen_variable_die): Add DW_AT_location for global
6525 register variables already during early_dwarf if possible.
6527 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
6529 * config/arm/arm_mve.h: Fix __arm_vctp16q return type.
6531 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
6534 * config/arm/arm.opt: Fix typo.
6535 * config/arm/t-rmprofile: Fix typo.
6537 2021-08-23 Jakub Jelinek <jakub@redhat.com>
6539 * tree.h (OMP_CLAUSE_GRAINSIZE_STRICT): Define.
6540 (OMP_CLAUSE_NUM_TASKS_STRICT): Define.
6541 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_GRAINSIZE,
6542 case OMP_CLAUSE_NUM_TASKS>: Print strict: modifier.
6543 * omp-expand.c (expand_task_call): Use GOMP_TASK_FLAG_STRICT in iflags
6544 if either grainsize or num_tasks clause has the strict modifier.
6546 2021-08-23 Martin Liska <mliska@suse.cz>
6548 * dbgcnt.def (DEBUG_COUNTER): New counter.
6549 * gimple.c (gimple_call_arg_flags): Use it in IPA PTA.
6551 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
6553 * ipa-modref.c (analyze_ssa_name_flags): Improve handling of return slot.
6555 2021-08-23 Xi Ruoyao <xry111@mengyan1223.wang>
6558 * config/mips/mips-protos.h (mips_msa_output_shift_immediate):
6560 * config/mips/mips.c (mips_msa_output_shift_immediate): New
6562 * config/mips/mips-msa.md (vashl<mode>3, vashr<mode>3,
6563 vlshr<mode>3): Call it.
6565 2021-08-22 Jan Hubicka <hubicka@ucw.cz>
6566 Martin Liska <mliska@suse.cz>
6568 PR middle-end/101949
6569 * ipa-modref.c (analyze_ssa_name_flags): Indirect call implies
6572 2021-08-21 Dragan Mladjenovic <OT_Dragan.Mladjenovic@mediatek.com>
6574 * config/mips/mips.c (mips_function_rodata_section,
6575 TARGET_ASM_FUNCTION_RODATA_SECTION): Removed.
6577 2021-08-21 John David Anglin <danglin@gcc.gnu.org>
6579 * config/pa/pa.c (pa_asm_output_aligned_common): Remove warning.
6581 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
6583 * configure.ac (thread-local storage support): Remove tls_first_major
6584 and tls_first_minor. Use "$conftest_s" to check support.
6585 * configure: Regenerate.
6587 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
6589 * configure.ac: Fixup formatting.
6591 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
6593 * acinclude.m4 (gcc_GAS_CHECK_FEATURE): Remove third argument and ...
6594 * configure.ac: ... update all callers.
6596 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
6599 * acinclude.m4 (_gcc_COMPUTE_GAS_VERSION, _gcc_GAS_VERSION_GTE_IFELSE)
6600 (gcc_GAS_VERSION_GTE_IFELSE): Remove.
6601 (gcc_GAS_CHECK_FEATURE): Do not handle in-tree case specially.
6602 * configure.ac: Remove gcc_cv_gas_major_version, gcc_cv_gas_minor_version.
6603 Remove remaining checks for in-tree assembler.
6604 * configure: Regenerate.
6606 2021-08-20 Jeff Law <jlaw@localhost.localdomain>
6608 * config/h8300/h8300.c (shift_alg_hi): Improve arithmetic shift right
6609 by 15 bits for H8/300H and H8/S. Improve logical shifts by 12
6611 (shift_alg_si): Improve arithmetic right shift by 28-30 bits for
6612 H8/300H. Improve arithmetic shift right by 15 bits for H8/S.
6613 Improve logical shifts by 27 bits for H8/S.
6614 (get_shift_alg): Corresponding changes.
6615 (h8300_option_override): Revert to loops for -Os when profitable.
6617 2021-08-20 Richard Biener <rguenther@suse.de>
6619 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
6621 (vect_analyze_data_ref_accesses): Likewise. Assign the BB
6622 index as group_id when dataref_groups were not computed.
6623 * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
6624 we advace to the next BB.
6626 2021-08-20 Jakub Jelinek <jakub@redhat.com>
6628 * omp-builtins.def (BUILT_IN_GOMP_WARNING, BUILT_IN_GOMP_ERROR): New
6631 2021-08-20 Martin Liska <mliska@suse.cz>
6633 PR gcov-profile/89961
6634 * gcov.c (make_gcov_file_name): Rewrite using std::string.
6635 (mangle_name): Simplify, do not used the second argument.
6636 (strip_extention): New function.
6637 (get_md5sum): Likewise.
6638 (get_gcov_intermediate_filename): Handle properly -p and -x
6640 (output_gcov_file): Use string type.
6641 (generate_results): Likewise.
6642 (md5sum_to_hex): Remove.
6644 2021-08-20 Michael Meissner <meissner@linux.ibm.com>
6646 * config/rs6000/altivec.md (UNSPEC_XXEVAL): Move to vsx.md.
6647 (UNSPEC_XXSPLTIW): Move to vsx.md.
6648 (UNSPEC_XXSPLTID): Move to vsx.md.
6649 (UNSPEC_XXSPLTI32DX): Move to vsx.md.
6650 (UNSPEC_XXBLEND): Move to vsx.md.
6651 (UNSPEC_XXPERMX): Move to vsx.md.
6652 (VM3): Move to vsx.md.
6653 (VM3_char): Move to vsx.md.
6654 (xxspltiw_v4si): Move to vsx.md.
6655 (xxspltiw_v4sf): Move to vsx.md.
6656 (xxspltiw_v4sf_inst): Move to vsx.md.
6657 (xxspltidp_v2df): Move to vsx.md.
6658 (xxspltidp_v2df_inst): Move to vsx.md.
6659 (xxsplti32dx_v4si_inst): Move to vsx.md.
6660 (xxsplti32dx_v4sf): Move to vsx.md.
6661 (xxsplti32dx_v4sf_inst): Move to vsx.md.
6662 (xxblend_<mode>): Move to vsx.md.
6663 (xxpermx): Move to vsx.md.
6664 (xxpermx_inst): Move to vsx.md.
6665 * config/rs6000/vsx.md (UNSPEC_XXEVAL): Move from altivec.md.
6666 (UNSPEC_XXSPLTIW): Move from altivec.md.
6667 (UNSPEC_XXSPLTID): Move from altivec.md.
6668 (UNSPEC_XXSPLTI32DX): Move from altivec.md.
6669 (UNSPEC_XXBLEND): Move from altivec.md.
6670 (UNSPEC_XXPERMX): Move from altivec.md.
6671 (VM3): Move from altivec.md.
6672 (VM3_char): Move from altivec.md.
6673 (xxspltiw_v4si): Move from altivec.md.
6674 (xxspltiw_v4sf): Move from altivec.md.
6675 (xxspltiw_v4sf_inst): Move from altivec.md.
6676 (xxspltidp_v2df): Move from altivec.md.
6677 (xxspltidp_v2df_inst): Move from altivec.md.
6678 (xxsplti32dx_v4si_inst): Move from altivec.md.
6679 (xxsplti32dx_v4sf): Move from altivec.md.
6680 (xxsplti32dx_v4sf_inst): Move from altivec.md.
6681 (xxblend_<mode>): Move from altivec.md.
6682 (xxpermx): Move from altivec.md.
6683 (xxpermx_inst): Move from altivec.md.
6685 2021-08-19 Roger Sayle <roger@nextmovesoftware.com>
6687 * tree-vect-generic.c (expand_vector_operations_1): Use either
6688 gimplify_build1 or gimplify_build2 instead of gimple_build_assign
6689 when constructing scalar splat expressions.
6691 2021-08-19 Peter Bergner <bergner@linux.ibm.com>
6694 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Cast
6695 pointer to __vector_pair *.
6697 2021-08-19 Martin Sebor <msebor@redhat.com>
6699 * gimple-range.cc: Add comments.
6700 * gimple-range.h: Same.
6702 2021-08-19 Martin Sebor <msebor@redhat.com>
6704 PR middle-end/101984
6705 * gimple-ssa-warn-access.cc (pass_waccess::execute): Also call
6708 2021-08-19 Jeff Law <jlaw@localhost.localdomain>
6710 * config.gcc (h8300-*-elf*): Do not include dbxelf.h.
6711 (h8300-*-linux*, v850-*-rtems*, v850*-elf*): Likewise.
6712 * config/v850/v850.h (DEFAULT_GDB_EXTENSIONS): Remove.
6714 2021-08-19 Jakub Jelinek <jakub@redhat.com>
6716 PR middle-end/101950
6717 * optabs.c (expand_clrsb_using_clz): New function.
6718 (expand_unop): Use it as another clrsb expansion fallback.
6720 2021-08-19 liuhongt <hongtao.liu@intel.com>
6723 2021-07-28 liuhongt <hongtao.liu@intel.com>
6726 * config/i386/i386.h (processor_costs): Add new member
6728 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
6729 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
6730 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
6731 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
6732 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
6733 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
6734 generic_cost, core_cost): Initialize integer_to_sse same value
6736 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
6737 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6738 Use integer_to_sse instead of sse_op to calculate the cost of
6741 2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
6743 * config.gcc: Include rpath.opt for Darwin.
6744 * config/darwin.h (DRIVER_SELF_SPECS): Handle -rpath.
6746 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
6749 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
6752 2021-08-18 Jonathan Wright <jonathan.wright@arm.com>
6754 * config/aarch64/arm_neon.h (vld3_lane_f64): Use float RTL
6755 pattern and type cast.
6756 (vld4_lane_f32): Use float RTL pattern.
6757 (vld4q_lane_f64): Use float type cast.
6759 2021-08-18 Jan Hubicka <hubicka@ucw.cz>
6761 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Check also
6764 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
6766 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor): Extend.
6767 (test_map_of_type_with_ctor_and_dtor_expand): Add function.
6768 (hash_map_tests_c_tests): Call it.
6770 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
6772 * ggc.h (enum ggc_collect): New.
6773 (ggc_collect): Use it.
6774 * ggc-page.c: Adjust.
6775 * ggc-common.c: Likewise.
6776 * ggc-tests.c: Likewise.
6777 * read-rtl-function.c: Likewise.
6778 * selftest-run-tests.c: Likewise.
6779 * doc/gty.texi (Invoking the garbage collector): Likewise.
6781 2021-08-18 liuhongt <hongtao.liu@intel.com>
6784 * config/i386/i386.h (TARGET_V2DF_REDUCTION_PREFER_HADDPD):
6786 * config/i386/sse.md (*sse3_haddv2df3_low): Add
6787 TARGET_V2DF_REDUCTION_PREFER_HADDPD.
6788 (*sse3_hsubv2df3_low): Ditto.
6789 * config/i386/x86-tune.def
6790 (X86_TUNE_V2DF_REDUCTION_PREFER_HADDPD): New tune.
6792 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
6794 * gimple-range-gori.cc (gori_compute::gori_compute): Enable tracing.
6795 (gori_compute::compute_operand_range): Add tracing.
6796 (gori_compute::logical_combine): Ditto.
6797 (gori_compute::compute_logical_operands): Ditto.
6798 (gori_compute::compute_operand1_range): Ditto.
6799 (gori_compute::compute_operand2_range): Ditto.
6800 (gori_compute::outgoing_edge_range_p): Ditto.
6801 * gimple-range-gori.h (class gori_compute): Add range_tracer.
6803 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
6805 * flag-types.h (enum evrp_mode): Adjust evrp-mode values.
6806 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Relocate from.
6807 * gimple-range-trace.h (DEBUG_RANGE_CACHE): Here.
6808 * params.opt (--param=evrp-mode): Adjust options.
6810 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
6812 * Makefile.in (OBJS): Add gimple-range-trace.o.
6813 * gimple-range-cache.h (enable_new_values): Remove unused prototype.
6814 * gimple-range-fold.cc: Adjust headers.
6815 * gimple-range-trace.cc: New.
6816 * gimple-range-trace.h: New.
6817 * gimple-range.cc (gimple_ranger::gimple_ranger): Enable tracer.
6818 (gimple_ranger::range_of_expr): Add tracing.
6819 (gimple_ranger::range_on_entry): Ditto.
6820 (gimple_ranger::range_on_exit): Ditto.
6821 (gimple_ranger::range_on_edge): Ditto.
6822 (gimple_ranger::fold_range_internal): Ditto.
6823 (gimple_ranger::dump_bb): Do not calculate edge range twice.
6824 (trace_ranger::*): Remove.
6825 (enable_ranger): Never create a trace_ranger.
6826 (debug_seed_ranger): Move to gimple-range-trace.cc.
6827 (dump_ranger): Ditto.
6828 (debug_ranger): Ditto.
6829 * gimple-range.h: Include gimple-range-trace.h.
6830 (range_on_entry, range_on_exit): No longer virtual.
6831 (class trace_ranger): Remove.
6832 (DEBUG_RANGE_CACHE): Move to gimple-range-trace.h.
6834 2021-08-17 Martin Sebor <msebor@redhat.com>
6836 PR middle-end/101854
6837 * builtins.c (expand_builtin_alloca): Move warning code to check_alloca
6838 in gimple-ssa-warn-access.cc.
6839 * calls.c (alloc_max_size): Move code to check_alloca.
6840 (get_size_range): Move to pointer-query.cc.
6841 (maybe_warn_alloc_args_overflow): Move to gimple-ssa-warn-access.cc.
6842 (get_attr_nonstring_decl): Move to tree.c.
6843 (fntype_argno_type): Move to gimple-ssa-warn-access.cc.
6844 (append_attrname): Same.
6845 (maybe_warn_rdwr_sizes): Same.
6846 (initialize_argument_information): Move code to
6847 gimple-ssa-warn-access.cc.
6848 * calls.h (maybe_warn_alloc_args_overflow): Move to
6849 gimple-ssa-warn-access.h.
6850 (get_attr_nonstring_decl): Move to tree.h.
6851 (maybe_warn_nonstring_arg): Move to gimple-ssa-warn-access.h.
6852 (enum size_range_flags): Move to pointer-query.h.
6853 (get_size_range): Same.
6854 * gimple-ssa-warn-access.cc (has_location): Remove unused overload
6855 to avoid Clang -Wunused-function.
6856 (get_size_range): Declare static.
6857 (maybe_emit_free_warning): Rename...
6858 (maybe_check_dealloc_call): ...to this for consistency.
6859 (class pass_waccess): Add members.
6860 (pass_waccess::~pass_waccess): Defined.
6861 (alloc_max_size): Move here from calls.c.
6862 (maybe_warn_alloc_args_overflow): Same.
6863 (check_alloca): New function.
6864 (check_alloc_size_call): New function.
6865 (check_strncat): Handle another warning flag.
6866 (pass_waccess::check_builtin): Handle alloca.
6867 (fntype_argno_type): Move here from calls.c.
6868 (append_attrname): Same.
6869 (maybe_warn_rdwr_sizes): Same.
6870 (pass_waccess::check_call): Define.
6871 (check_nonstring_args): New function.
6872 (pass_waccess::check): Call new member functions.
6873 (pass_waccess::execute): Enable ranger.
6874 * gimple-ssa-warn-access.h (get_size_range): Move here from calls.h.
6875 (maybe_warn_nonstring_arg): Same.
6876 * gimple-ssa-warn-restrict.c: Remove #include.
6877 * pointer-query.cc (get_size_range): Move here from calls.c.
6878 * pointer-query.h (enum size_range_flags): Same.
6879 (get_size_range): Same.
6880 * tree.c (get_attr_nonstring_decl): Move here from calls.c.
6881 * tree.h (get_attr_nonstring_decl): Move here from calls.h.
6883 2021-08-17 Thomas Schwinge <thomas@codesourcery.com>
6885 * ggc.h (ggc_collect): Add 'force_collect' parameter.
6886 * ggc-page.c (ggc_collect): Use that one instead of global
6887 'ggc_force_collect'. Adjust all users.
6888 * doc/gty.texi (Invoking the garbage collector): Update.
6889 * ggc-internal.h (ggc_force_collect): Remove.
6890 * ggc-common.c (ggc_force_collect): Likewise.
6891 * selftest.h (forcibly_ggc_collect): Remove.
6892 * ggc-tests.c (selftest::forcibly_ggc_collect): Likewise.
6893 * read-rtl-function.c (test_loading_labels): Adjust.
6894 * selftest-run-tests.c (run_tests): Likewise.
6896 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
6898 * config/darwin.c (darwin_file_end): Reset and reclaim the
6899 section names table at the end of compile.
6901 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
6904 * config.in: Regenerate.
6905 * config/i386/darwin.h (EXTRA_ASM_OPTS): New
6906 (ASM_SPEC): Pass options to disable branch shortening where
6908 * configure: Regenerate.
6909 * configure.ac: Detect versions of 'as' that support the
6910 optimisation which has the bug.
6912 2021-08-17 Richard Biener <rguenther@suse.de>
6914 * optabs-query.c (supports_vec_gather_load_p): Also check
6916 (supports_vec_scatter_store_p): Likewise.
6917 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Fall
6918 back to masked variants if non-masked are not supported.
6919 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
6920 When we need to use masked gather/scatter but do not have
6921 a mask set up a constant true one.
6922 * tree-vect-stmts.c (vect_check_scalar_mask): Also allow
6925 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
6927 * tree-ssa-ccp.c (bit_value_binop) [MINUS_EXPR]: Use same
6928 algorithm as PLUS_EXPR to improve subtraction bit bounds.
6929 [POINTER_DIFF_EXPR]: Treat as synonymous with MINUS_EXPR.
6931 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
6933 * tree-ssa-ccp.c (bit_value_mult_const): New helper function to
6934 calculate the mask-value pair result of a multiplication by an
6936 (bit_value_binop) [MULT_EXPR]: Call it from here for
6937 multiplications by (sparse) non-negative constants.
6939 2021-08-17 Christophe Lyon <christophe.lyon@foss.st.com>
6942 * config.gcc (gcc_cv_initfini_array): Leave undefined for
6943 uclinuxfdpiceabi targets.
6945 2021-08-17 Alexandre Oliva <oliva@adacore.com>
6947 * tree-inline.c (maybe_move_debug_stmts_to_successors): Don't
6948 reverse debug stmts.
6950 2021-08-17 Alexandre Oliva <oliva@adacore.com>
6952 * tree-cfg.c (dump_function_to_file): Use fun, not cfun.
6954 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
6956 * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete.
6957 (__LD4Q_LANE_FUNC): Likewise.
6958 (vld4_lane_u8): Define without macro.
6959 (vld4_lane_u16): Likewise.
6960 (vld4_lane_u32): Likewise.
6961 (vld4_lane_u64): Likewise.
6962 (vld4_lane_s8): Likewise.
6963 (vld4_lane_s16): Likewise.
6964 (vld4_lane_s32): Likewise.
6965 (vld4_lane_s64): Likewise.
6966 (vld4_lane_f16): Likewise.
6967 (vld4_lane_f32): Likewise.
6968 (vld4_lane_f64): Likewise.
6969 (vld4_lane_p8): Likewise.
6970 (vld4_lane_p16): Likewise.
6971 (vld4_lane_p64): Likewise.
6972 (vld4q_lane_u8): Likewise.
6973 (vld4q_lane_u16): Likewise.
6974 (vld4q_lane_u32): Likewise.
6975 (vld4q_lane_u64): Likewise.
6976 (vld4q_lane_s8): Likewise.
6977 (vld4q_lane_s16): Likewise.
6978 (vld4q_lane_s32): Likewise.
6979 (vld4q_lane_s64): Likewise.
6980 (vld4q_lane_f16): Likewise.
6981 (vld4q_lane_f32): Likewise.
6982 (vld4q_lane_f64): Likewise.
6983 (vld4q_lane_p8): Likewise.
6984 (vld4q_lane_p16): Likewise.
6985 (vld4q_lane_p64): Likewise.
6986 (vld4_lane_bf16): Likewise.
6987 (vld4q_lane_bf16): Likewise.
6989 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
6991 * config/aarch64/arm_neon.h (__LD3_LANE_FUNC): Delete.
6992 (__LD3Q_LANE_FUNC): Delete.
6993 (vld3_lane_u8): Define without macro.
6994 (vld3_lane_u16): Likewise.
6995 (vld3_lane_u32): Likewise.
6996 (vld3_lane_u64): Likewise.
6997 (vld3_lane_s8): Likewise.
6998 (vld3_lane_s16): Likewise.
6999 (vld3_lane_s32): Likewise.
7000 (vld3_lane_s64): Likewise.
7001 (vld3_lane_f16): Likewise.
7002 (vld3_lane_f32): Likewise.
7003 (vld3_lane_f64): Likewise.
7004 (vld3_lane_p8): Likewise.
7005 (vld3_lane_p16): Likewise.
7006 (vld3_lane_p64): Likewise.
7007 (vld3q_lane_u8): Likewise.
7008 (vld3q_lane_u16): Likewise.
7009 (vld3q_lane_u32): Likewise.
7010 (vld3q_lane_u64): Likewise.
7011 (vld3q_lane_s8): Likewise.
7012 (vld3q_lane_s16): Likewise.
7013 (vld3q_lane_s32): Likewise.
7014 (vld3q_lane_s64): Likewise.
7015 (vld3q_lane_f16): Likewise.
7016 (vld3q_lane_f32): Likewise.
7017 (vld3q_lane_f64): Likewise.
7018 (vld3q_lane_p8): Likewise.
7019 (vld3q_lane_p16): Likewise.
7020 (vld3q_lane_p64): Likewise.
7021 (vld3_lane_bf16): Likewise.
7022 (vld3q_lane_bf16): Likewise.
7024 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
7026 * config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Delete.
7027 (__LD2Q_LANE_FUNC): Likewise.
7028 (vld2_lane_u8): Define without macro.
7029 (vld2_lane_u16): Likewise.
7030 (vld2_lane_u32): Likewise.
7031 (vld2_lane_u64): Likewise.
7032 (vld2_lane_s8): Likewise.
7033 (vld2_lane_s16): Likewise.
7034 (vld2_lane_s32): Likewise.
7035 (vld2_lane_s64): Likewise.
7036 (vld2_lane_f16): Likewise.
7037 (vld2_lane_f32): Likewise.
7038 (vld2_lane_f64): Likewise.
7039 (vld2_lane_p8): Likewise.
7040 (vld2_lane_p16): Likewise.
7041 (vld2_lane_p64): Likewise.
7042 (vld2q_lane_u8): Likewise.
7043 (vld2q_lane_u16): Likewise.
7044 (vld2q_lane_u32): Likewise.
7045 (vld2q_lane_u64): Likewise.
7046 (vld2q_lane_s8): Likewise.
7047 (vld2q_lane_s16): Likewise.
7048 (vld2q_lane_s32): Likewise.
7049 (vld2q_lane_s64): Likewise.
7050 (vld2q_lane_f16): Likewise.
7051 (vld2q_lane_f32): Likewise.
7052 (vld2q_lane_f64): Likewise.
7053 (vld2q_lane_p8): Likewise.
7054 (vld2q_lane_p16): Likewise.
7055 (vld2q_lane_p64): Likewise.
7056 (vld2_lane_bf16): Likewise.
7057 (vld2q_lane_bf16): Likewise.
7059 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
7061 * haifa-sched.c (advance_one_cycle): Output more context-synchronization
7064 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
7066 * haifa-sched.c (enum rfs_decision, rfs_str): Add RFS_AUTOPREF.
7067 (rank_for_schedule): Use it.
7069 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
7071 PR rtl-optimization/91598
7072 * haifa-sched.c (autopref_rank_for_schedule): Prioritize "irrelevant"
7073 insns after memory reads and before memory writes.
7075 2021-08-17 Alistair_Lee <alistair.lee@arm.com>
7077 * rtl.h (CONST_VECTOR_P): New macro.
7078 * config/aarch64/aarch64.c (aarch64_get_sve_pred_bits): Use RTL
7079 code testing macros.
7080 (aarch64_ptrue_all_mode): Likewise.
7081 (aarch64_expand_mov_immediate): Likewise.
7082 (aarch64_const_vec_all_in_range_p): Likewise.
7083 (aarch64_rtx_costs): Likewise.
7084 (aarch64_legitimate_constant_p): Likewise.
7085 (aarch64_simd_valid_immediate): Likewise.
7086 (aarch64_simd_make_constant): Likewise.
7087 (aarch64_convert_mult_to_shift): Likewise.
7088 (aarch64_expand_sve_vec_perm): Likewise.
7089 (aarch64_vec_fpconst_pow_of_2): Likewise.
7091 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
7093 PR tree-optimization/101938
7094 * range-op.cc (operator_abs::op1_range): Special case
7095 -TYPE_MIN_VALUE for flag_wrapv.
7097 2021-08-17 Kewen Lin <linkw@linux.ibm.com>
7099 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Add the cost for
7102 2021-08-17 Jakub Jelinek <jakub@redhat.com>
7104 * tree.def (OMP_SCOPE): New tree code.
7105 * tree.h (OMP_SCOPE_BODY, OMP_SCOPE_CLAUSES): Define.
7106 * tree-nested.c (convert_nonlocal_reference_stmt,
7107 convert_local_reference_stmt, convert_gimple_call): Handle
7109 * tree-pretty-print.c (dump_generic_node): Handle OMP_SCOPE.
7110 * gimple.def (GIMPLE_OMP_SCOPE): New gimple code.
7111 * gimple.c (gimple_build_omp_scope): New function.
7112 (gimple_copy): Handle GIMPLE_OMP_SCOPE.
7113 * gimple.h (gimple_build_omp_scope): Declare.
7114 (gimple_has_substatements): Handle GIMPLE_OMP_SCOPE.
7115 (gimple_omp_scope_clauses, gimple_omp_scope_clauses_ptr,
7116 gimple_omp_scope_set_clauses): New inline functions.
7117 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_SCOPE.
7118 * gimple-pretty-print.c (dump_gimple_omp_scope): New function.
7119 (pp_gimple_stmt_1): Handle GIMPLE_OMP_SCOPE.
7120 * gimple-walk.c (walk_gimple_stmt): Likewise.
7121 * gimple-low.c (lower_stmt): Likewise.
7122 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
7123 (gimplify_scan_omp_clauses): For task reductions, handle OMP_SCOPE
7124 like ORT_WORKSHARE constructs. Adjust diagnostics for %<scope%>
7125 allowing task reductions. Reject inscan reductions on scope.
7126 (omp_find_stores_stmt): Handle GIMPLE_OMP_SCOPE.
7127 (gimplify_omp_workshare, gimplify_expr): Handle OMP_SCOPE.
7128 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_SCOPE.
7129 (estimate_num_insns): Likewise.
7130 * omp-low.c (build_outer_var_ref): Look through GIMPLE_OMP_SCOPE
7131 contexts if var isn't privatized there.
7132 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_SCOPE.
7133 (scan_omp_1_stmt): Likewise.
7134 (maybe_add_implicit_barrier_cancel): Look through outer
7136 (lower_omp_scope): New function.
7137 (lower_omp_task_reductions): Handle OMP_SCOPE.
7138 (lower_omp_1): Handle GIMPLE_OMP_SCOPE.
7139 (diagnose_sb_1, diagnose_sb_2): Likewise.
7140 * omp-expand.c (expand_omp_single): Support also GIMPLE_OMP_SCOPE.
7141 (expand_omp): Handle GIMPLE_OMP_SCOPE.
7142 (omp_make_gimple_edges): Likewise.
7143 * omp-builtins.def (BUILT_IN_GOMP_SCOPE_START): New built-in.
7145 2021-08-17 Richard Biener <rguenther@suse.de>
7147 PR tree-optimization/101925
7148 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Set
7149 reverse on COMPONENT_REF and ARRAY_REF according to
7150 what reverse_storage_order_for_component_p does.
7151 (vn_reference_eq): Compare reversed on reference ops.
7152 (reverse_storage_order_for_component_p): New overload.
7153 (vn_reference_lookup_3): Check reverse_storage_order_for_component_p
7154 on the reference looked up.
7156 2021-08-17 Jeff Law <jlaw@localhost.localdomain>
7158 * config/h8300/h8300.c (shift_alg_si): Avoid loops for most SImode
7160 (h8300_option_override): Use loops on H8/S more often when optimizing
7162 (get_shift_alg): Handle new "special" cases on H8/S. Simplify
7163 accordingly. Handle various arithmetic right shifts with special
7164 sequences that we couldn't handle before.
7166 2021-08-16 Jeff Law <jlaw@localhost.localdomain>
7168 * config.gcc (rl78-*-elf*): Do not include dbxelf.h.
7170 2021-08-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
7172 * config/sparc/rtemself.h (SPARC_GCOV_TYPE_SIZE): Define.
7173 * config/sparc/sparc.c (sparc_gcov_type_size): New.
7174 (TARGET_GCOV_TYPE_SIZE): Redefine if SPARC_GCOV_TYPE_SIZE is defined.
7175 * coverage.c (get_gcov_type): Use targetm.gcov_type_size().
7176 * doc/tm.texi (TARGET_GCOV_TYPE_SIZE): Add hook under "Misc".
7177 * doc/tm.texi.in: Regenerate.
7178 * target.def (gcov_type_size): New target hook.
7179 * targhooks.c (default_gcov_type_size): New.
7180 * targhooks.h (default_gcov_type_size): Declare.
7181 * tree-profile.c (gimple_gen_edge_profiler): Use precision of
7183 (gimple_gen_time_profiler): Likewise.
7185 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
7187 * dwarf2out.c (add_scalar_info): Deal with DW_AT_data_bit_offset.
7189 2021-08-16 Tobias Burnus <tobias@codesourcery.com>
7191 PR middle-end/101931
7192 * omp-low.c (omp_runtime_api_call): Update for routines
7193 added in the meanwhile.
7195 2021-08-16 Martin Liska <mliska@suse.cz>
7197 PR tree-optimization/100393
7198 * tree-switch-conversion.c (group_cluster::dump): Use
7199 get_comparison_count.
7200 (jump_table_cluster::find_jump_tables): Pre-compute number of
7201 comparisons and then decrement it. Cache also max_ratio.
7202 (jump_table_cluster::can_be_handled): Change signature.
7203 * tree-switch-conversion.h (get_comparison_count): New.
7205 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
7207 * dwarf2out.c (add_data_member_location_attribute): Use GNAT
7208 encodings only when -fgnat-encodings=all is specified.
7209 (add_bound_info): Likewise.
7210 (add_byte_size_attribute): Likewise.
7211 (gen_member_die): Likewise.
7213 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
7215 * omp-oacc-neuter-broadcast.cc
7216 (execute_omp_oacc_neuter_broadcast): Plug 'par' memory leak.
7218 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
7220 * omp-oacc-neuter-broadcast.cc
7221 (execute_omp_oacc_neuter_broadcast): Clarify memory management for
7224 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
7226 * omp-oacc-neuter-broadcast.cc (field_map): Move variable into...
7227 (execute_omp_oacc_neuter_broadcast): ... here.
7228 (install_var_field, build_receiver_ref, build_sender_ref): Take
7229 'field_map_t *' parameter. Adjust all users.
7230 (worker_single_copy, neuter_worker_single): Take a
7231 'record_field_map_t *' parameter. Adjust all users.
7233 2021-08-16 liuhongt <hongtao.liu@intel.com>
7236 * config/i386/i386.md (ldexp<mode>3): Force operands[1] to
7239 2021-08-16 Martin Liska <mliska@suse.cz>
7242 * multiple_target.c (create_dispatcher_calls): Make default
7243 function local only if it is a definition.
7245 2021-08-16 Martin Liska <mliska@suse.cz>
7248 * ipa-icf-gimple.c (func_checker::compare_ssa_name): Do not
7249 consider equal SSA_NAMEs when one is a param.
7251 2021-08-16 liuhongt <hongtao.liu@intel.com>
7254 * config/i386/i386-expand.c (ix86_expand_vec_perm_vpermt2):
7255 Support vpermi2b for V32QI/V16QImode.
7256 (ix86_extract_perm_from_pool_constant): New function.
7257 (ix86_expand_vec_one_operand_perm_avx512): Support
7258 vpermw/vpermb under TARGET_AVX512BW/TARGET_AVX512VBMI.
7259 (expand_vec_perm_1): Adjust comments for upper.
7260 * config/i386/i386-protos.h (ix86_extract_perm_from_pool_constant):
7262 * config/i386/predicates.md (permvar_truncate_operand): New predicate.
7263 (pshufb_truncv4siv4hi_operand): Ditto.
7264 (pshufb_truncv8hiv8qi_operand): Ditto.
7265 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1):
7266 New pre_reload define_insn_and_split.
7267 (*avx512f_permvar_truncv8siv8hi_1): Ditto.
7268 (*avx512f_vpermvar_truncv8div8si_1): Ditto.
7269 (*avx512f_permvar_truncv32hiv32qi_1): Ditto.
7270 (*avx512f_permvar_truncv16hiv16qi_1): Ditto.
7271 (*avx512f_permvar_truncv4div4si_1): Ditto.
7272 (*avx512f_pshufb_truncv8hiv8qi_1): Ditto.
7273 (*avx512f_pshufb_truncv4siv4hi_1): Ditto.
7274 (*avx512f_pshufd_truncv2div2si_1): Ditto.
7276 2021-08-16 Kito Cheng <kito.cheng@sifive.com>
7278 * config/riscv/multilib-generator: Support code model option for
7280 * doc/install.texi: Add document of new option for
7281 --with-multilib-generator.
7283 2021-08-15 Clément Chigot <clement.chigot@atos.net>
7285 * config/rs6000/rs6000.c (xcoff_tls_exec_model_detected): New.
7286 (rs6000_legitimize_tls_address_aix): Use it.
7287 (rs6000_xcoff_file_end): Add ".ref __tls_get_addr" when
7288 xcoff_tls_exec_model_detected is true.
7290 2021-08-15 Jeff Law <jlaw@localhost.localdomain>
7292 * config/h8300/h8300.c (shift_alg_si): Retune H8/300H shifts
7293 to allow a bit more code growth, saving many dozens of cycles.
7294 (h8300_option_override): Adjus shift_alg_si if optimizing for
7296 (get_shift_alg): Use special + inline shifts for residuals
7299 2021-08-14 Stafford Horne <shorne@gmail.com>
7302 * config/or1k/or1k-opts.h: New file.
7303 * config/or1k/or1k.c (or1k_legitimize_address_1, print_reloc):
7304 Support generating gotha relocations if -mcmodel=large is
7306 * config/or1k/or1k.h (TARGET_CMODEL_SMALL, TARGET_CMODEL_LARGE):
7308 * config/or1k/or1k.opt (mcmodel=): New option.
7309 * doc/invoke.texi (OpenRISC Options): Document mcmodel.
7311 2021-08-14 Martin Sebor <msebor@redhat.com>
7313 PR middle-end/101791
7314 * gimple-ssa-warn-access.cc (new_delete_mismatch_p): Use new argument
7315 to valid_new_delete_pair_p.
7316 * tree.c (valid_new_delete_pair_p): Add argument.
7317 * tree.h (valid_new_delete_pair_p): Same.
7319 2021-08-14 Jakub Jelinek <jakub@redhat.com>
7322 * config/i386/i386-expand.c (expand_vec_perm_broadcast_1)
7323 <case E_V64QImode>: For this mode assert
7324 !TARGET_AVX512BW || d->perm[0] rather than !TARGET_AVX2 || d->perm[0].
7326 2021-08-13 Michael Meissner <meissner@linux.ibm.com>
7329 * config/rs6000/altivec.md (xxeval): Use register_predicate
7330 instead of altivec_register_predicate.
7332 2021-08-13 Martin Sebor <msebor@redhat.com>
7334 PR middle-end/101734
7335 * tree-ssa-uninit.c (maybe_warn_read_write_only): New function.
7336 (maybe_warn_operand): Call it.
7338 2021-08-13 Martin Liska <mliska@suse.cz>
7341 * attribs.c (decl_attributes): Make naked functions "noipa"
7344 2021-08-13 Martin Liska <mliska@suse.cz>
7347 * symtab.c (symtab_node::noninterposable_alias): Do not create
7348 local aliases for target_clone functions as the clonning pass
7351 2021-08-13 Martin Liska <mliska@suse.cz>
7353 * opts.c (LIVE_PATCHING_OPTION): Define.
7354 (control_options_for_live_patching): Use it in error messages.
7356 2021-08-13 Jan Hubicka <hubicka@ucw.cz>
7358 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOREAD.
7359 (implicit_const_eaf_flags, implicit_pure_eaf_flags,
7360 ignore_stores_eaf_flags): New constants.
7361 (remove_useless_eaf_flags): New function.
7362 (eaf_flags_useful_p): Use it.
7363 (deref_flags): Add EAF_NOT_RETURNED if flag is unused;
7365 (modref_lattice::init): Add EAF_NOREAD.
7366 (modref_lattice::add_escape_point): Do not reacord escape point if
7368 (modref_lattice::merge): EAF_NOESCAPE implies EAF_NODIRECTESCAPE;
7369 use remove_useless_eaf_flags.
7370 (modref_lattice::merge_deref): Use ignore_stores_eaf_flags.
7371 (modref_lattice::merge_direct_load): Add EAF_NOREAD
7372 (analyze_ssa_name_flags): Fix handling EAF_NOT_RETURNED
7373 (analyze_parms): Use remove_useless_eaf_flags.
7374 (ipa_merge_modref_summary_after_inlining): Use ignore_stores_eaf_flags.
7375 (modref_merge_call_site_flags): Add caller and ecf_flags parameter;
7376 use remove_useless_eaf_flags.
7377 (modref_propagate_flags_in_scc): Update.
7378 * ipa-modref.h: Turn eaf_flags_t back to char.
7379 * tree-core.h (EAF_NOT_RETURNED): Fix.
7380 (EAF_NOREAD): New constant
7381 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1): Check for
7383 * tree-ssa-structalias.c (handle_rhs_call): Handle new flags.
7384 (handle_pure_call): Likewise.
7386 2021-08-12 Jakub Jelinek <jakub@redhat.com>
7388 * tree.def (OMP_MASKED): New tree code.
7389 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_FILTER.
7390 * tree.h (OMP_MASKED_BODY, OMP_MASKED_CLAUSES, OMP_MASKED_COMBINED,
7391 OMP_CLAUSE_FILTER_EXPR): Define.
7392 * tree.c (omp_clause_num_ops): Add OMP_CLAUSE_FILTER entry.
7393 (omp_clause_code_name): Likewise.
7394 (walk_tree_1): Handle OMP_CLAUSE_FILTER.
7395 * tree-nested.c (convert_nonlocal_omp_clauses,
7396 convert_local_omp_clauses): Handle OMP_CLAUSE_FILTER.
7397 (convert_nonlocal_reference_stmt, convert_local_reference_stmt,
7398 convert_gimple_call): Handle GIMPLE_OMP_MASTER.
7399 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_FILTER.
7400 (dump_generic_node): Handle OMP_MASTER.
7401 * gimple.def (GIMPLE_OMP_MASKED): New gimple code.
7402 * gimple.c (gimple_build_omp_masked): New function.
7403 (gimple_copy): Handle GIMPLE_OMP_MASKED.
7404 * gimple.h (gimple_build_omp_masked): Declare.
7405 (gimple_has_substatements): Handle GIMPLE_OMP_MASKED.
7406 (gimple_omp_masked_clauses, gimple_omp_masked_clauses_ptr,
7407 gimple_omp_masked_set_clauses): New inline functions.
7408 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_MASKED.
7409 * gimple-pretty-print.c (dump_gimple_omp_masked): New function.
7410 (pp_gimple_stmt_1): Handle GIMPLE_OMP_MASKED.
7411 * gimple-walk.c (walk_gimple_stmt): Likewise.
7412 * gimple-low.c (lower_stmt): Likewise.
7413 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
7414 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_FILTER. For clauses
7415 that take one expression rather than decl or constant, force
7416 gimplification of that into a SSA_NAME or temporary unless min
7418 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_FILTER.
7419 (gimplify_expr): Handle OMP_MASKED.
7420 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_MASKED.
7421 (estimate_num_insns): Likewise.
7422 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_FILTER.
7423 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_MASKED. Adjust
7424 diagnostics for existence of masked construct.
7425 (scan_omp_1_stmt, lower_omp_master, lower_omp_1, diagnose_sb_1,
7426 diagnose_sb_2): Handle GIMPLE_OMP_MASKED.
7427 * omp-expand.c (expand_omp_synch, expand_omp, omp_make_gimple_edges):
7430 2021-08-12 Uroš Bizjak <ubizjak@gmail.com>
7433 * config/i386/i386.md (avx512f_scalef<mode>2): New insn pattern.
7434 (ldexp<mode>3): Use avx512f_scalef<mode>2.
7435 (UNSPEC_SCALEF): Move from ...
7436 * config/i386/sse.md (UNSPEC_SCALEF): ... here.
7438 2021-08-12 Jan Hubicka <hubicka@ucw.cz>
7440 * ipa-split.c (consider_split): Fix condition testing void functions.
7442 2021-08-12 Aldy Hernandez <aldyh@redhat.com>
7444 * doc/invoke.texi: Remove docs for threader-mode param.
7445 * flag-types.h (enum threader_mode): Remove.
7446 * params.opt: Remove threader-mode param.
7447 * tree-ssa-threadbackward.c (class back_threader): Remove
7448 path_is_unreachable_p.
7449 Make find_paths private.
7450 Add maybe_thread and thread_through_all_blocks.
7451 Remove reference marker for m_registry.
7452 Remove reference marker for m_profit.
7453 (back_threader::back_threader): Adjust for registry and profit not
7455 (dump_path): Move down.
7457 (class thread_jumps): Remove.
7458 (class back_threader_registry): Remove m_all_paths.
7460 (thread_jumps::thread_through_all_blocks): Move to back_threader
7462 (fsm_find_thread_path): Remove
7463 (back_threader::maybe_thread): New.
7464 (back_threader::thread_through_all_blocks): Move from
7466 (back_threader_registry::back_threader_registry): Remove
7468 (back_threader_registry::~back_threader_registry): Remove.
7469 (thread_jumps::find_taken_edge): Remove.
7470 (thread_jumps::check_subpath_and_update_thread_path): Remove.
7471 (thread_jumps::maybe_register_path): Remove.
7472 (thread_jumps::handle_phi): Remove.
7473 (handle_assignment_p): Remove.
7474 (thread_jumps::handle_assignment): Remove.
7475 (thread_jumps::fsm_find_control_statement_thread_paths): Remove.
7476 (thread_jumps::find_jump_threads_backwards): Remove.
7477 (thread_jumps::find_jump_threads_backwards_with_ranger): Remove.
7478 (try_thread_blocks): Rename find_jump_threads_backwards to
7480 (pass_early_thread_jumps::execute): Same.
7482 2021-08-12 Tobias Burnus <tobias@codesourcery.com>
7484 * tree-core.h (omp_clause_proc_bind_kind): Add
7485 OMP_CLAUSE_PROC_BIND_PRIMARY.
7486 * tree-pretty-print.c (dump_omp_clause): Add TODO comment to
7487 change 'master' to 'primary' in proc_bind for OpenMP 5.1.
7489 2021-08-12 Claudiu Zissulescu <claziss@synopsys.com>
7491 * common/config/arc/arc-common.c (arc_option_init_struct): Remove
7492 fno-common reference.
7493 * config/arc/arc.c (arc_override_options): Remove overriding of
7496 2021-08-12 Jakub Jelinek <jakub@redhat.com>
7499 * config/i386/i386-expand.c (ix86_expand_vec_one_operand_perm_avx512):
7500 If d->testing_p, return true after performing checks instead of
7501 actually expanding the insn.
7502 (expand_vec_perm_broadcast_1): Handle V32HImode - assert
7503 !TARGET_AVX512BW and return false.
7505 2021-08-12 Eric Botcazou <ebotcazou@gcc.gnu.org>
7507 * configure.ac (PE linker --disable-dynamicbase support): New check.
7508 * configure: Regenerate.
7509 * config.in: Likewise.
7510 * config/i386/mingw32.h (LINK_SPEC_DISABLE_DYNAMICBASE): New define.
7511 (LINK_SPEC): Use it.
7512 * config/i386/mingw-w64.h (LINK_SPEC_DISABLE_DYNAMICBASE): Likewise.
7513 (LINK_SPEC): Likewise.
7515 2021-08-12 liuhongt <hongtao.liu@intel.com>
7518 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_2): New
7519 post_reload define_insn_and_split.
7520 (*avx512bw_zero_extendv32qiv32hi2_2): Ditto.
7521 (*sse4_1_zero_extendv8qiv8hi2_4): Ditto.
7522 (*avx512f_zero_extendv16hiv16si2_2): Ditto.
7523 (*avx2_zero_extendv8hiv8si2_2): Ditto.
7524 (*sse4_1_zero_extendv4hiv4si2_4): Ditto.
7525 (*avx512f_zero_extendv8siv8di2_2): Ditto.
7526 (*avx2_zero_extendv4siv4di2_2): Ditto.
7527 (*sse4_1_zero_extendv2siv2di2_4): Ditto.
7528 (VI248_256, VI248_512, VI148_512, VI148_256, VI148_128): New
7531 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
7533 * config/rs6000/rs6000-builtin-new.def: Add always, power5, and
7536 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
7538 * config/rs6000/rs6000-builtin-new.def: Add vsx stanza.
7540 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
7542 * config/rs6000/rs6000-builtin-new.def: Finish altivec stanza.
7543 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
7544 initialization of pcvoid_type_node here...
7545 (altivec_init_builtins): ...from here.
7546 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add
7547 RS6000_BTI_const_ptr_void.
7548 (pcvoid_type_node): New macro.
7550 2021-08-11 Richard Biener <rguenther@suse.de>
7553 * tree-ssa-forwprop.c (pass_forwprop::execute): Do not decompose
7554 hard-register accesses.
7556 2021-08-11 Richard Biener <rguenther@suse.de>
7558 * tree-ssa-operands.c (operands_scanner::get_expr_operands):
7559 Do not look at COMPONENT_REF FIELD_DECLs TREE_THIS_VOLATILE
7560 to determine has_volatile_ops.
7562 2021-08-11 Eric Botcazou <ebotcazou@gcc.gnu.org>
7564 * cfgexpand.c (expand_used_vars): Reuse attribs local variable.
7566 2021-08-11 Jan Hubicka <hubicka@ucw.cz>
7567 Alexandre Oliva <oliva@adacore.com>
7569 * ipa-modref.c (modref_lattice::dump): Fix escape_point's min_flags
7571 (modref_lattice::merge_deref): Fix handling of indirect scape points.
7572 (update_escape_summary_1): Likewise.
7573 (update_escape_summary): Likewise.
7574 (ipa_merge_modref_summary_after_inlining): Likewise.
7576 2021-08-11 Richard Biener <rguenther@suse.de>
7578 PR middle-end/101858
7579 * fold-const.c (fold_binary_loc): Guard simplification
7580 of X < (cast) (1 << Y) to integer types.
7582 2021-08-11 Richard Biener <rguenther@suse.de>
7584 PR tree-optimization/101861
7585 * tree-vect-stmts.c (vectorizable_load): Fix error in
7586 previous change with regard to gather vectorization.
7588 2021-08-11 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
7591 * config/arm/arm_neon.h (vdup_n_s8): Replace call to builtin
7593 (vdup_n_s16): Likewise.
7594 (vdup_n_s32): Likewise.
7595 (vdup_n_s64): Likewise.
7596 (vdup_n_u8): Likewise.
7597 (vdup_n_u16): Likewise.
7598 (vdup_n_u32): Likewise.
7599 (vdup_n_u64): Likewise.
7600 (vdup_n_p8): Likewise.
7601 (vdup_n_p16): Likewise.
7602 (vdup_n_p64): Likewise.
7603 (vdup_n_f16): Likewise.
7604 (vdup_n_f32): Likewise.
7605 (vdupq_n_s8): Likewise.
7606 (vdupq_n_s16): Likewise.
7607 (vdupq_n_s32): Likewise.
7608 (vdupq_n_s64): Likewise.
7609 (vdupq_n_u8): Likewise.
7610 (vdupq_n_u16): Likewise.
7611 (vdupq_n_u32): Likewise.
7612 (vdupq_n_u64): Likewise.
7613 (vdupq_n_p8): Likewise.
7614 (vdupq_n_p16): Likewise.
7615 (vdupq_n_p64): Likewise.
7616 (vdupq_n_f16): Likewise.
7617 (vdupq_n_f32): Likewise.
7618 (vmov_n_s8): Replace call to builtin with call to corresponding
7620 (vmov_n_s16): Likewise.
7621 (vmov_n_s32): Likewise.
7622 (vmov_n_s64): Likewise.
7623 (vmov_n_u8): Likewise.
7624 (vmov_n_u16): Likewise.
7625 (vmov_n_u32): Likewise.
7626 (vmov_n_u64): Likewise.
7627 (vmov_n_p8): Likewise.
7628 (vmov_n_p16): Likewise.
7629 (vmov_n_f16): Likewise.
7630 (vmov_n_f32): Likewise.
7631 (vmovq_n_s8): Likewise.
7632 (vmovq_n_s16): Likewise.
7633 (vmovq_n_s32): Likewise.
7634 (vmovq_n_s64): Likewise.
7635 (vmovq_n_u8): Likewise.
7636 (vmovq_n_u16): Likewise.
7637 (vmovq_n_u32): Likewise.
7638 (vmovq_n_u64): Likewise.
7639 (vmovq_n_p8): Likewise.
7640 (vmovq_n_p16): Likewise.
7641 (vmovq_n_f16): Likewise.
7642 (vmovq_n_f32): Likewise.
7643 * config/arm/arm_neon_builtins.def: Remove entries for vdup_n.
7645 2021-08-11 liuhongt <hongtao.liu@intel.com>
7648 * config/i386/i386.md (ldexp<mode>3): Extend to vscalefs[sd]
7649 when TARGET_AVX512F and TARGET_SSE_MATH.
7651 2021-08-10 Jakub Jelinek <jakub@redhat.com>
7654 * config/i386/i386-expand.c (expand_vec_perm_even_odd): Return false
7655 for V32HImode if !TARGET_AVX512BW.
7656 (ix86_vectorize_vec_perm_const) <case E_V32HImode, case E_V64QImode>:
7657 If !TARGET_AVX512BW and TARGET_AVX512F and d.testing_p, don't fail
7658 early, but actually check the permutation.
7660 2021-08-10 Richard Biener <rguenther@suse.de>
7662 PR tree-optimization/101809
7663 * tree-vect-stmts.c (get_load_store_type): Allow emulated
7664 gathers with offset vector nunits being a constant multiple
7665 of the data vector nunits.
7666 (vect_get_gather_scatter_ops): Use the appropriate nunits
7667 for the offset vector defs.
7668 (vectorizable_store): Adjust call to
7669 vect_get_gather_scatter_ops.
7670 (vectorizable_load): Likewise. Handle the case of less
7671 offset vectors than data vectors.
7673 2021-08-10 Jakub Jelinek <jakub@redhat.com>
7676 * config/i386/sse.md (*avx512f_shuf_<shuffletype>64x2_1<mask_name>_1,
7677 *avx512f_shuf_<shuffletype>32x4_1<mask_name>_1): New define_insn
7680 2021-08-10 Richard Biener <rguenther@suse.de>
7682 PR tree-optimization/101801
7683 PR tree-optimization/101819
7684 * tree-vectorizer.h (vect_emulated_vector_p): Declare.
7685 * tree-vect-loop.c (vect_emulated_vector_p): New function.
7686 (vectorizable_reduction): Re-instantiate a check for emulated
7688 * tree-vect-stmts.c (vectorizable_shift): Likewise.
7689 (vectorizable_operation): Likewise. Cost emulated vector
7690 operations according to the scalar sequence synthesized by
7693 2021-08-10 Richard Biener <rguenther@suse.de>
7695 PR middle-end/101824
7696 * tree-nested.c (get_frame_field): Mark the COMPONENT_REF as
7697 volatile in case the variable was.
7699 2021-08-10 H.J. Lu <hjl.tools@gmail.com>
7702 * config/i386/constraints.md (BC): Document for integer SSE
7703 constant all bits set operand.
7704 (BF): New constraint for const floating-point all bits set
7706 * config/i386/i386.c (standard_sse_constant_p): Likewise.
7707 (standard_sse_constant_opcode): Likewise.
7708 * config/i386/sse.md (sseconstm1): New mode attribute.
7709 (mov<mode>_internal): Replace BC with <sseconstm1>.
7711 2021-08-10 liuhongt <hongtao.liu@intel.com>
7713 * config/i386/sse.md (cond_<insn><mode>): New expander.
7714 (VI248_AVX512VLBW): New mode iterator.
7715 * config/i386/predicates.md
7716 (nonimmediate_or_const_vec_dup_operand): New predicate.
7718 2021-08-09 Andrew MacLeod <amacleod@redhat.com>
7720 PR tree-optimization/101741
7721 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Check
7722 type of parameter for toupper/tolower.
7724 2021-08-09 Martin Jambor <mjambor@suse.cz>
7727 * ipa-prop.c (propagate_controlled_uses): Removed a spurious space.
7729 2021-08-09 Pat Haugen <pthaugen@linux.ibm.com>
7731 * config/rs6000/rs6000.c (is_load_insn1): Verify destination is a
7733 (is_store_insn1): Verify source is a register.
7735 2021-08-09 Uroš Bizjak <ubizjak@gmail.com>
7738 * config/i386/mmx.md (<any_logic:code>v2sf3):
7739 Rename from *mmx_<any_logic:code>v2sf3
7741 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7743 * config/nvptx/nvptx.c: Cross-reference parts adapted in
7744 'gcc/omp-oacc-neuter-broadcast.cc'.
7745 * omp-low.c: Likewise.
7746 * omp-oacc-neuter-broadcast.cc: Cross-reference parts adapted from
7749 2021-08-09 Julian Brown <julian@codesourcery.com>
7750 Kwok Cheung Yeung <kcy@codesourcery.com>
7751 Thomas Schwinge <thomas@codesourcery.com>
7753 * config/gcn/gcn.c (gcn_init_builtins): Override decls for
7754 BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START,
7755 BUILT_IN_GOACC_SINGLE_COPY_END and BUILT_IN_GOACC_BARRIER.
7756 (gcn_goacc_validate_dims): Turn on worker partitioning unconditionally.
7757 (gcn_fork_join): Update comment.
7758 * config/gcn/gcn.opt (flag_worker_partitioning): Remove.
7759 (macc_experimental_workers): Remove unused option.
7761 2021-08-09 Julian Brown <julian@codesourcery.com>
7762 Nathan Sidwell <nathan@codesourcery.com> (via 'gcc/config/nvptx/nvptx.c' master)
7763 Kwok Cheung Yeung <kcy@codesourcery.com>
7764 Thomas Schwinge <thomas@codesourcery.com>
7766 * Makefile.in (OBJS): Add omp-oacc-neuter-broadcast.o.
7767 * doc/tm.texi.in (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD):
7768 Add documentation hook.
7769 * doc/tm.texi: Regenerate.
7770 * omp-oacc-neuter-broadcast.cc: New file.
7771 * omp-builtins.def (BUILT_IN_GOACC_BARRIER)
7772 (BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START)
7773 (BUILT_IN_GOACC_SINGLE_COPY_END): New builtins.
7774 * passes.def (pass_omp_oacc_neuter_broadcast): Add pass.
7775 * target.def (goacc.create_worker_broadcast_record): Add target
7777 * tree-pass.h (make_pass_omp_oacc_neuter_broadcast): Add
7779 * config/gcn/gcn-protos.h (gcn_goacc_adjust_propagation_record):
7780 Rename prototype to...
7781 (gcn_goacc_create_worker_broadcast_record): ... this.
7782 * config/gcn/gcn-tree.c (gcn_goacc_adjust_propagation_record): Rename
7784 (gcn_goacc_create_worker_broadcast_record): ... this.
7785 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_PROPAGATION_RECORD):
7787 (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD): ... this.
7789 2021-08-09 Tejas Belagod <tejas.belagod@arm.com>
7792 * config/aarch64/aarch64-simd.md (vlshr<mode>3, vashr<mode>3): Use
7795 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7797 * Makefile.in (GTFILES): Remove '$(srcdir)/omp-offload.c'.
7799 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7801 * builtins.def (DEF_GOACC_BUILTIN, DEF_GOMP_BUILTIN): Don't
7802 consider '-foffload-abi'.
7803 * common.opt (-foffload-abi): Remove 'Var', 'Init'.
7804 * opts.c (common_handle_option) <-foffload-abi> [ACCEL_COMPILER]:
7807 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7809 * optc-gen.awk: Sanity check that 'Init' doesn't appear without
7812 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7814 * omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE): Remove.
7816 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7818 * doc/gty.texi (Files): Update.
7820 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7822 * doc/gty.texi (Files): Fix GTY header file example.
7824 2021-08-09 Roger Sayle <roger@nextmovesoftware.com>
7826 * tree-ssa-ccp.c (value_mask_to_min_max): Helper function to
7827 determine the upper and lower bounds from a mask-value pair.
7828 (bit_value_unop) [ABS_EXPR, ABSU_EXPR]: Add support for
7829 absolute value and unsigned absolute value expressions.
7830 (bit_value_binop): Initialize *VAL's precision.
7831 [LT_EXPR, LE_EXPR]: Use value_mask_to_min_max to determine
7832 upper and lower bounds of operands. Add LE_EXPR/GE_EXPR
7833 support when the operands are unknown but potentially equal.
7834 [MIN_EXPR, MAX_EXPR]: Support minimum/maximum expressions.
7836 2021-08-09 Bin Cheng <bin.cheng@linux.alibaba.com>
7838 * config/aarch64/aarch64.md
7839 (*extend<SHORT:mode><GPI:mode>2_aarch64): Use %<GPI:w>0.
7841 2021-08-08 Sergei Trofimovich <siarheit@google.com>
7843 * lra-constraints.c: Fix s/otput/output/ typo.
7845 2021-08-06 Martin Sebor <msebor@redhat.com>
7847 * builtins.c (expand_builtin_memchr): Move to gimple-ssa-warn-access.cc.
7848 (expand_builtin_strcat): Same.
7849 (expand_builtin_stpncpy): Same.
7850 (expand_builtin_strncat): Same.
7851 (check_read_access): Same.
7852 (check_memop_access): Same.
7853 (expand_builtin_strlen): Move checks to gimple-ssa-warn-access.cc.
7854 (expand_builtin_strnlen): Same.
7855 (expand_builtin_memcpy): Same.
7856 (expand_builtin_memmove): Same.
7857 (expand_builtin_mempcpy): Same.
7858 (expand_builtin_strcpy): Same.
7859 (expand_builtin_strcpy_args): Same.
7860 (expand_builtin_stpcpy_1): Same.
7861 (expand_builtin_strncpy): Same.
7862 (expand_builtin_memset): Same.
7863 (expand_builtin_bzero): Same.
7864 (expand_builtin_strcmp): Same.
7865 (expand_builtin_strncmp): Same.
7866 (expand_builtin): Remove handlers.
7867 (fold_builtin_strlen): Add a comment.
7868 * builtins.h (check_access): Move to gimple-ssa-warn-access.cc.
7869 * calls.c (maybe_warn_nonstring_arg): Same.
7870 * diagnostic-spec.c (nowarn_spec_t::nowarn_spec_t): Add warning option.
7871 * gimple-fold.c (gimple_fold_builtin_strcpy): Pass argument to callee.
7872 (gimple_fold_builtin_stpcpy): Same.
7873 * gimple-ssa-warn-access.cc (has_location): New function.
7874 (get_location): Same.
7875 (get_callee_fndecl): Same.
7878 (warn_string_no_nul): Define.
7879 (unterminated_array): Same.
7880 (check_nul_terminated_array): Same.
7881 (maybe_warn_nonstring_arg): Same.
7882 (maybe_warn_for_bound): Same.
7883 (warn_for_access): Same.
7884 (check_access): Same.
7885 (check_memop_access): Same.
7886 (check_read_access): Same.
7887 (warn_dealloc_offset): Use helper functions.
7888 (maybe_emit_free_warning): Same.
7889 (class pass_waccess): Add members.
7890 (check_strcat): New function.
7891 (check_strncat): New function.
7892 (check_stxcpy): New function.
7893 (check_stxncpy): New function.
7894 (check_strncmp): New function.
7895 (pass_waccess::check_builtin): New function.
7896 (pass_waccess::check): Call it.
7897 * gimple-ssa-warn-access.h (warn_string_no_nul): Move here from
7899 (maybe_warn_for_bound): Same.
7900 (check_access): Same.
7901 (check_memop_access): Same.
7902 (check_read_access): Same.
7903 * pointer-query.h (struct access_data): Define a ctor overload.
7905 2021-08-06 Richard Biener <rguenther@suse.de>
7907 PR tree-optimization/101801
7908 * tree-vectorizer.h (vect_worthwhile_without_simd_p): Rename...
7909 (vect_can_vectorize_without_simd_p): ... to this.
7910 * tree-vect-loop.c (vect_worthwhile_without_simd_p): Rename...
7911 (vect_can_vectorize_without_simd_p): ... to this and fold
7912 in vect_min_worthwhile_factor.
7913 (vect_min_worthwhile_factor): Remove.
7914 (vectorizable_reduction): Adjust and remove the cost part.
7915 * tree-vect-stmts.c (vectorizable_shift): Likewise.
7916 (vectorizable_operation): Likewise.
7918 2021-08-06 Uroš Bizjak <ubizjak@gmail.com>
7921 * config/i386/i386.md (cmove reg-to-reg move elimination peephole2s):
7922 Add general_gr_operand predicate to operand 3.
7924 2021-08-06 Roger Sayle <roger@nextmovesoftware.com>
7926 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern): Use
7927 CFN_BUILT_IN_CLRSB* instead of BUILT_IN_CLRSB* for consistency.
7929 2021-08-06 Tamar Christina <tamar.christina@arm.com>
7931 * config/aarch64/aarch64-sve-builtins.cc (register_svpattern,
7932 register_svprfop): Pass vec<> by pointer.
7933 * langhooks-def.h (lhd_simulate_enum_decl): Likewise.
7934 * langhooks.c (lhd_simulate_enum_decl): Likewise.
7935 * langhooks.h (struct lang_hooks_for_types): Likewise.
7937 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
7939 * config/aarch64/arm_neon.h (vst1_bf16_x2): Use
7940 __builtin_memcpy instead of constructing an additional
7941 __builtin_aarch64_simd_oi one vector at a time.
7942 (vst1q_bf16_x2): Likewise.
7943 (vst1_bf16_x3): Use __builtin_memcpy instead of constructing
7944 an additional __builtin_aarch64_simd_ci one vector at a time.
7945 (vst1q_bf16_x3): Likewise.
7946 (vst1_bf16_x4): Use __builtin_memcpy instead of a union.
7947 (vst1q_bf16_x4): Likewise.
7948 (vst2_bf16): Use __builtin_memcpy instead of constructing an
7949 additional __builtin_aarch64_simd_oi one vector at a time.
7950 (vst2q_bf16): Likewise.
7951 (vst3_bf16): Use __builtin_memcpy instead of constructing an
7952 additional __builtin_aarch64_simd_ci mode one vector at a
7954 (vst3q_bf16): Likewise.
7955 (vst4_bf16): Use __builtin_memcpy instead of constructing an
7956 additional __builtin_aarch64_simd_xi one vector at a time.
7957 (vst4q_bf16): Likewise.
7959 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
7961 * config/aarch64/arm_neon.h (__ST2_LANE_FUNC): Delete.
7962 (__ST2Q_LANE_FUNC): Delete.
7963 (vst2_lane_f16): Use __builtin_memcpy to copy vector
7964 structure instead of constructing __builtin_aarch64_simd_oi
7965 one vector at a time.
7966 (vst2_lane_f32): Likewise.
7967 (vst2_lane_f64): Likewise.
7968 (vst2_lane_p8): Likewise.
7969 (vst2_lane_p16): Likewise.
7970 (vst2_lane_p64): Likewise.
7971 (vst2_lane_s8): Likewise.
7972 (vst2_lane_s16): Likewise.
7973 (vst2_lane_s32): Likewise.
7974 (vst2_lane_s64): Likewise.
7975 (vst2_lane_u8): Likewise.
7976 (vst2_lane_u16): Likewise.
7977 (vst2_lane_u32): Likewise.
7978 (vst2_lane_u64): Likewise.
7979 (vst2_lane_bf16): Likewise.
7980 (vst2q_lane_f16): Use __builtin_memcpy to copy vector
7981 structure instead of using a union.
7982 (vst2q_lane_f32): Likewise.
7983 (vst2q_lane_f64): Likewise.
7984 (vst2q_lane_p8): Likewise.
7985 (vst2q_lane_p16): Likewise.
7986 (vst2q_lane_p64): Likewise.
7987 (vst2q_lane_s8): Likewise.
7988 (vst2q_lane_s16): Likewise.
7989 (vst2q_lane_s32): Likewise.
7990 (vst2q_lane_s64): Likewise.
7991 (vst2q_lane_u8): Likewise.
7992 (vst2q_lane_u16): Likewise.
7993 (vst2q_lane_u32): Likewise.
7994 (vst2q_lane_u64): Likewise.
7995 (vst2q_lane_bf16): Likewise.
7997 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
7999 * config/aarch64/arm_neon.h (__ST3_LANE_FUNC): Delete.
8000 (__ST3Q_LANE_FUNC): Delete.
8001 (vst3_lane_f16): Use __builtin_memcpy to copy vector
8002 structure instead of constructing __builtin_aarch64_simd_ci
8003 one vector at a time.
8004 (vst3_lane_f32): Likewise.
8005 (vst3_lane_f64): Likewise.
8006 (vst3_lane_p8): Likewise.
8007 (vst3_lane_p16): Likewise.
8008 (vst3_lane_p64): Likewise.
8009 (vst3_lane_s8): Likewise.
8010 (vst3_lane_s16): Likewise.
8011 (vst3_lane_s32): Likewise.
8012 (vst3_lane_s64): Likewise.
8013 (vst3_lane_u8): Likewise.
8014 (vst3_lane_u16): Likewise.
8015 (vst3_lane_u32): Likewise.
8016 (vst3_lane_u64): Likewise.
8017 (vst3_lane_bf16): Likewise.
8018 (vst3q_lane_f16): Use __builtin_memcpy to copy vector
8019 structure instead of using a union.
8020 (vst3q_lane_f32): Likewise.
8021 (vst3q_lane_f64): Likewise.
8022 (vst3q_lane_p8): Likewise.
8023 (vst3q_lane_p16): Likewise.
8024 (vst3q_lane_p64): Likewise.
8025 (vst3q_lane_s8): Likewise.
8026 (vst3q_lane_s16): Likewise.
8027 (vst3q_lane_s32): Likewise.
8028 (vst3q_lane_s64): Likewise.
8029 (vst3q_lane_u8): Likewise.
8030 (vst3q_lane_u16): Likewise.
8031 (vst3q_lane_u32): Likewise.
8032 (vst3q_lane_u64): Likewise.
8033 (vst3q_lane_bf16): Likewise.
8035 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
8037 * config/aarch64/arm_neon.h (__ST4_LANE_FUNC): Delete.
8038 (__ST4Q_LANE_FUNC): Delete.
8039 (vst4_lane_f16): Use __builtin_memcpy to copy vector
8040 structure instead of constructing __builtin_aarch64_simd_xi
8041 one vector at a time.
8042 (vst4_lane_f32): Likewise.
8043 (vst4_lane_f64): Likewise.
8044 (vst4_lane_p8): Likewise.
8045 (vst4_lane_p16): Likewise.
8046 (vst4_lane_p64): Likewise.
8047 (vst4_lane_s8): Likewise.
8048 (vst4_lane_s16): Likewise.
8049 (vst4_lane_s32): Likewise.
8050 (vst4_lane_s64): Likewise.
8051 (vst4_lane_u8): Likewise.
8052 (vst4_lane_u16): Likewise.
8053 (vst4_lane_u32): Likewise.
8054 (vst4_lane_u64): Likewise.
8055 (vst4_lane_bf16): Likewise.
8056 (vst4q_lane_f16): Use __builtin_memcpy to copy vector
8057 structure instead of using a union.
8058 (vst4q_lane_f32): Likewise.
8059 (vst4q_lane_f64): Likewise.
8060 (vst4q_lane_p8): Likewise.
8061 (vst4q_lane_p16): Likewise.
8062 (vst4q_lane_p64): Likewise.
8063 (vst4q_lane_s8): Likewise.
8064 (vst4q_lane_s16): Likewise.
8065 (vst4q_lane_s32): Likewise.
8066 (vst4q_lane_s64): Likewise.
8067 (vst4q_lane_u8): Likewise.
8068 (vst4q_lane_u16): Likewise.
8069 (vst4q_lane_u32): Likewise.
8070 (vst4q_lane_u64): Likewise.
8071 (vst4q_lane_bf16): Likewise.
8073 2021-08-06 Martin Liska <mliska@suse.cz>
8075 * config/rs6000/rs6000.c (rs6000_option_override_internal): When
8076 a target option is restored, it can have
8077 rs6000_long_double_type_size set to FLOAT_PRECISION_TFmode
8078 and error should not be emitted.
8080 2021-08-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
8082 * gcov-io.h (gcov_write): Declare.
8083 * gcov-io.c (gcov_write): New.
8084 (gcov_write_counter): Remove.
8085 (gcov_write_tag_length): Likewise.
8086 (gcov_write_summary): Replace gcov_write_tag_length() with calls to
8087 gcov_write_unsigned().
8088 * doc/invoke.texi (fprofile-info-section): Mention
8089 __gcov_info_to_gdca().
8091 2021-08-06 Martin Sebor <msebor@redhat.com>
8093 * dominance.c (prune_bbs_to_update_dominators): Adjust by-value vec
8094 arguments to by-reference.
8095 (iterate_fix_dominators): Same.
8096 * dominance.h (iterate_fix_dominators): Same.
8097 * ipa-prop.h: Call auto_vec::to_vec_legacy.
8098 * tree-data-ref.c (dump_data_dependence_relation): Adjust by-value vec
8099 arguments to by-reference.
8100 (debug_data_dependence_relation): Same.
8101 (dump_data_dependence_relations): Same.
8102 * tree-data-ref.h (debug_data_dependence_relation): Same.
8103 (dump_data_dependence_relations): Same.
8104 * tree-predcom.c (dump_chains): Same.
8105 (initialize_root_vars_lm): Same.
8106 (determine_unroll_factor): Same.
8107 (replace_phis_by_defined_names): Same.
8108 (insert_init_seqs): Same.
8109 (pcom_worker::tree_predictive_commoning_loop): Call
8110 auto_vec::to_vec_legacy.
8111 * tree-ssa-pre.c (insert_into_preds_of_block): Adjust by-value vec
8112 arguments to by-reference.
8113 * tree-ssa-threadbackward.c (populate_worklist): Same.
8114 (back_threader::resolve_def): Same.
8115 * tree-vect-data-refs.c (vect_check_nonzero_value): Same.
8116 (vect_enhance_data_refs_alignment): Same.
8117 (vect_check_lower_bound): Same.
8118 (vect_prune_runtime_alias_test_list): Same.
8119 (vect_permute_store_chain): Same.
8120 * tree-vect-slp-patterns.c (vect_normalize_conj_loc): Same.
8121 * tree-vect-stmts.c (vect_create_vectorized_demotion_stmts): Same.
8122 * tree-vectorizer.h (vect_permute_store_chain): Same.
8123 * vec.c (test_init): New function.
8124 (vec_c_tests): Call new function.
8125 * vec.h (vec): Declare ctors, dtor, and assignment.
8126 (auto_vec::vec_to_legacy): New function.
8127 (vec::copy): Adjust initialization.
8129 2021-08-05 H.J. Lu <hjl.tools@gmail.com>
8132 * config/i386/i386.c (ix86_can_inline_p): Ignore MASK_80387 if
8133 callee only uses GPRs.
8134 * config/i386/ia32intrin.h: Revert commit 5463cee2770.
8135 * config/i386/serializeintrin.h: Revert commit 71958f740f1.
8136 * config/i386/x86gprintrin.h: Add
8137 #pragma GCC target("general-regs-only") and #pragma GCC pop_options
8138 to disable non-GPR ISAs.
8140 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
8142 PR middle-end/101787
8143 * doc/md.texi (cond_ashl, cond_ashr, cond_lshr): Document.
8145 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
8147 * tree-vectorizer.h (vect_is_store_elt_extraction, vect_is_reduction)
8148 (vect_reduc_type, vect_embedded_comparison_type, vect_comparison_type)
8149 (vect_is_extending_load, vect_is_integer_truncation): New functions,
8150 moved from aarch64.c but given different names.
8151 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction)
8152 (aarch64_is_reduction, aarch64_reduc_type)
8153 (aarch64_embedded_comparison_type, aarch64_comparison_type)
8154 (aarch64_extending_load_p, aarch64_integer_truncation_p): Delete
8155 in favor of the above. Update callers accordingly.
8157 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
8160 * config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
8161 writing .cpu directive in asm output.
8162 * config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
8163 (arm_last_printed_arch_string): Delete.
8164 (arm_last-printed_fpu_string): Delete.
8165 (arm_configure_build_target): If use of floating-point/SIMD is
8166 disabled, remove all fp/simd related features from the target ISA.
8167 (last_arm_targ_options): New variable.
8168 (arm_print_asm_arch_directives): Add new parameters. Change order
8169 of emitted directives and handle all cases here.
8170 (arm_file_start): Always call arm_print_asm_arch_directives, move
8171 all generation of .arch/.arch_extension here.
8172 (arm_file_end): Call arm_print_asm_arch.
8173 (arm_declare_function_name): Call arm_print_asm_arch_directives
8174 instead of printing .arch/.fpu directives directly.
8176 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
8178 * config/arm/arm.c (arm_configure_build_target): Don't call
8179 arm_option_reconfigure_globals.
8180 (arm_option_restore): Call arm_option_reconfigure_globals after
8181 reconfiguring the target.
8182 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
8184 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
8186 * config/arm/arm.c (arm_configure_build_target): Ensure the target's
8187 arch_name is always set.
8189 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
8191 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
8192 of vec_select high-half from being added into Neon subtract
8195 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
8197 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
8198 of vec_select high-half from being added into Neon add cost.
8200 2021-08-05 Kewen Lin <linkw@linux.ibm.com>
8202 * cfgloop.h (loops_list::loops_list): Add one optional argument
8203 root and adjust accordingly, update loop tree walking and factor
8205 * cfgloop.c (loops_list::walk_loop_tree): ... this. New function.
8207 2021-08-05 Eric Botcazou <ebotcazou@gcc.gnu.org>
8209 PR tree-optimization/101626
8210 * tree-sra.c (propagate_subaccesses_from_rhs): Do not set the
8211 reverse scalar storage order on a pointer or vector component.
8213 2021-08-05 liuhongt <hongtao.liu@intel.com>
8215 * config/i386/sse.md (cond_<code><mode>): New expander.
8217 2021-08-05 liuhongt <hongtao.liu@intel.com>
8219 * config/i386/sse.md (cond_<code><mode>): New expander.
8221 2021-08-05 liuhongt <hongtao.liu@intel.com>
8223 * config/i386/sse.md (cond_<code><mode>): New expander.
8225 2021-08-04 David Malcolm <dmalcolm@redhat.com>
8228 * Makefile.in (ANALYZER_OBJS): Add analyzer/region-model-asm.o.
8230 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
8233 * config/i386/i386.h (STORE_MAX_PIECES): Allow 16/32/64 bytes
8234 only if TARGET_INTER_UNIT_MOVES_TO_VEC is true.
8236 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
8239 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
8240 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
8241 data with SSE register from one memory location to another.
8243 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
8245 * config/s390/s390.c (expand_perm_with_vpdi): New function.
8246 (vectorize_vec_perm_const_1): Call expand_perm_with_vpdi.
8247 * config/s390/vector.md (*vpdi1<mode>, @vpdi1<mode>): Enable a
8248 parameterized expander.
8249 (*vpdi4<mode>, @vpdi4<mode>): Likewise.
8251 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
8253 * config/s390/s390.c (MAX_VECT_LEN): Define macro.
8254 (struct expand_vec_perm_d): Define struct.
8255 (expand_perm_with_merge): New function.
8256 (vectorize_vec_perm_const_1): New function.
8257 (s390_vectorize_vec_perm_const): New function.
8258 (TARGET_VECTORIZE_VEC_PERM_CONST): Define target macro.
8260 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
8262 * config/s390/vector.md (V_HW_64): Remove mode iterator.
8263 (*vec_load_pair<mode>): Use V_HW_2 instead of V_HW_64.
8264 * config/s390/vx-builtins.md
8265 (vec_scatter_element<V_HW_2:mode>_SI): Use V_HW_2 instead of
8268 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
8270 * config/s390/s390.md (UNSPEC_VEC_PERMI): Remove constant
8272 * config/s390/vector.md (*vpdi1<mode>, *vpdi4<mode>): New pattern
8274 * config/s390/vx-builtins.md (*vec_permi<mode>): Emit generic rtx
8275 instead of an unspec.
8277 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
8279 * config/s390/s390-modes.def: Add more vector modes to support
8280 concatenation of two vectors.
8281 * config/s390/s390-protos.h (s390_expand_merge_perm_const): Add
8283 (s390_expand_merge): Likewise.
8284 * config/s390/s390.c (s390_expand_merge_perm_const): New function.
8285 (s390_expand_merge): New function.
8286 * config/s390/s390.md (UNSPEC_VEC_MERGEH, UNSPEC_VEC_MERGEL):
8287 Remove constant definitions.
8288 * config/s390/vector.md (V_HW_2): Add mode iterators.
8289 (VI_HW_4, V_HW_4): Rename VI_HW_4 to V_HW_4.
8290 (vec_2x_nelts, vec_2x_wide): New mode attributes.
8291 (*vmrhb, *vmrlb, *vmrhh, *vmrlh, *vmrhf, *vmrlf, *vmrhg, *vmrlg):
8292 New pattern definitions.
8293 (vec_widen_umult_lo_<mode>, vec_widen_umult_hi_<mode>)
8294 (vec_widen_smult_lo_<mode>, vec_widen_smult_hi_<mode>)
8295 (vec_unpacks_lo_v4sf, vec_unpacks_hi_v4sf, vec_unpacks_lo_v2df)
8296 (vec_unpacks_hi_v2df): Adjust expanders to emit non-unspec RTX for
8298 * config/s390/vx-builtins.md (V_HW_4): Remove mode iterator. Now
8300 (vec_mergeh<mode>, vec_mergel<mode>): Use s390_expand_merge to
8301 emit vec merge pattern.
8303 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
8305 * config/aarch64/aarch64.c (aarch64_strip_extend_vec_half):
8307 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent cost of
8308 vec_select high-half from being added into Neon multiply
8310 * rtlanal.c (vec_series_highpart_p): Define.
8311 * rtlanal.h (vec_series_highpart_p): Declare.
8313 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
8315 * config/aarch64/aarch64.c (aarch64_strip_duplicate_vec_elt):
8317 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent
8318 vec_select cost from being added into Neon multiply cost.
8320 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
8322 * tree-vect-loop.c (vect_better_loop_vinfo_p): Detect cases in
8323 which old_loop_vinfo is an epilogue loop that handles a constant
8324 number of iterations.
8326 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
8328 * tree-vect-loop.c (vect_analyze_loop): Print a dump message
8329 when a reanalyzed loop fails to be cheaper than the current
8332 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
8334 * config/aarch64/aarch64.c: Fix a typo.
8336 2021-08-04 Vincent Lefèvre <vincent-gcc@vinc17.net>
8338 PR gcov-profile/101773
8339 * gcov-io.c (gcov_close): Check return code of a fclose.
8341 2021-08-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
8344 * dwarf2out.c (dwarf2out_assembly_start): Emit a dummy
8345 .file statement when needed.
8347 2021-08-04 Richard Biener <rguenther@suse.de>
8349 * tree-vect-data-refs.c (vect_check_gather_scatter):
8350 Include widening conversions only when the result is
8351 still handed by native gather or the current offset
8352 size not already matches the data size.
8353 Also succeed analysis in case there's no native support,
8354 noted by a IFN_LAST ifn and a NULL decl.
8355 (vect_analyze_data_refs): Always consider gathers.
8356 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
8357 Test for no IFN gather rather than decl gather.
8358 * tree-vect-stmts.c (vect_model_load_cost): Pass in the
8359 gather-scatter info and cost emulated gathers accordingly.
8360 (vect_truncate_gather_scatter_offset): Properly test for
8362 (vect_use_strided_gather_scatters_p): Likewise.
8363 (get_load_store_type): Handle emulated gathers and its
8365 (vectorizable_load): Likewise. Emulate them by extracting
8366 scalar offsets, doing scalar loads and a vector construct.
8368 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
8371 * expr.c (op_by_pieces_d::op_by_pieces_d): Add a max_pieces
8372 argument to set m_max_size.
8373 (move_by_pieces_d): Pass MOVE_MAX_PIECES to op_by_pieces_d.
8374 (store_by_pieces_d): Pass STORE_MAX_PIECES to op_by_pieces_d.
8375 (compare_by_pieces_d): Pass COMPARE_MAX_PIECES to op_by_pieces_d.
8377 2021-08-04 Roger Sayle <roger@nextmovesoftware.com>
8378 Marc Glisse <marc.glisse@inria.fr>
8380 * match.pd (bit_ior, bit_xor): Canonicalize (X*C1)|(X*C2) and
8381 (X*C1)^(X*C2) as X*(C1+C2), and related variants, using
8382 tree_nonzero_bits to ensure that operands are bit-wise disjoint.
8384 2021-08-04 Richard Biener <rguenther@suse.de>
8386 * tree-ssa-forwprop.c (pass_forwprop::execute): Split
8387 out code to decompose vector loads ...
8388 (optimize_vector_load): ... here. Generalize it to
8389 handle intermediate widening and TARGET_MEM_REF loads
8390 and apply it to loads with a supported vector mode as well.
8392 2021-08-04 Richard Biener <rguenther@suse.de>
8394 PR tree-optimization/101756
8395 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Make sure
8396 the result of the reduction epilogue is compatible to the original
8399 2021-08-04 liuhongt <hongtao.liu@intel.com>
8402 * config/i386/i386.md (peephole2): Refine predicate from
8403 register_operand to general_reg_operand.
8405 2021-08-04 Aldy Hernandez <aldyh@redhat.com>
8407 * gimple-range-path.h (path_range_query::dump): Mark override.
8409 2021-08-04 Richard Biener <rguenther@suse.de>
8411 PR tree-optimization/101769
8412 * tree-tailcall.c (eliminate_tail_call): Add the created loop
8413 for the first recursion and return it via the new output parameter.
8414 (optimize_tail_call): Pass through new output param.
8415 (tree_optimize_tail_calls_1): After creating all latches,
8416 add the created loop to the loop tree. Do not mark loops for fixup.
8418 2021-08-04 Martin Liska <mliska@suse.cz>
8420 * doc/invoke.texi: Document threader-mode param.
8422 2021-08-04 liuhongt <hongtao.liu@intel.com>
8424 * config/i386/sse.md (cond_fma<mode>): New expander.
8425 (cond_fms<mode>): Ditto.
8426 (cond_fnma<mode>): Ditto.
8427 (cond_fnms<mode>): Ditto.
8429 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
8431 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode>): Use && instead of &.
8433 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
8435 * config/rs6000/constraints.md: Remove "e" from the list of available
8436 constraint characters.
8438 2021-08-03 Eugene Rozenfeld <erozen@microsoft.com>
8440 PR gcov-profile/71672
8441 * auto-profile.c (afdo_indirect_call): Fix setup of the historgram value for indirect calls.
8443 2021-08-03 Paul A. Clarke <pc@us.ibm.com>
8445 * config/rs6000/smmintrin.h (_mm_minpos_epu16): New.
8447 2021-08-03 H.J. Lu <hjl.tools@gmail.com>
8449 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): In 64-bit mode,
8450 try XMM31 to avoid vzeroupper.
8452 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8454 * doc/invoke.texi: Document -mtune=neoverse-512tvb and
8455 -mcpu=neoverse-512tvb.
8456 * config/aarch64/aarch64-cores.def (neoverse-512tvb): New entry.
8457 * config/aarch64/aarch64-tune.md: Regenerate.
8458 * config/aarch64/aarch64.c (neoverse512tvb_sve_vector_cost)
8459 (neoverse512tvb_sve_issue_info, neoverse512tvb_vec_issue_info)
8460 (neoverse512tvb_vector_cost, neoverse512tvb_tunings): New structures.
8461 (aarch64_adjust_body_cost_sve): Handle -mtune=neoverse-512tvb.
8462 (aarch64_adjust_body_cost): Likewise.
8464 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8466 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Only
8467 record issue information for operations that occur in the
8470 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8472 * config/aarch64/aarch64.c (aarch64_multiply_add_p): Add a vec_flags
8473 parameter. Detect cases in which an Advanced SIMD MLA would almost
8474 certainly require a MOV.
8475 (aarch64_count_ops): Update accordingly.
8477 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8479 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction): New
8480 function, split out from...
8481 (aarch64_detect_vector_stmt_subtype): ...here.
8482 (aarch64_add_stmt_cost): Treat extracting element 0 as free.
8484 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8486 * config/aarch64/aarch64-protos.h (sve_vec_cost):
8487 Add gather_load_x32_cost and gather_load_x64_cost.
8488 * config/aarch64/aarch64.c (generic_sve_vector_cost)
8489 (a64fx_sve_vector_cost, neoversev1_sve_vector_cost): Update
8490 accordingly, using the values given by the scalar_load * number
8491 of elements calculation that we used previously.
8492 (aarch64_detect_vector_stmt_subtype): Use the new fields.
8494 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8496 * config/aarch64/aarch64.c (aarch64_adjust_body_cost_sve): New
8497 function, split out from...
8498 (aarch64_adjust_body_cost): ...here.
8500 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8502 * config/aarch64/fractional-cost.h: New file.
8503 * config/aarch64/aarch64.c: Include <algorithm> (indirectly)
8504 and cost_fraction.h.
8505 (vec_cost_fraction): New typedef.
8506 (aarch64_detect_scalar_stmt_subtype): Use it for statement costs.
8507 (aarch64_detect_vector_stmt_subtype): Likewise.
8508 (aarch64_sve_adjust_stmt_cost, aarch64_adjust_stmt_cost): Likewise.
8509 (aarch64_estimate_min_cycles_per_iter): Use vec_cost_fraction
8511 (aarch64_adjust_body_cost): Likewise.
8512 (aarch64_test_cost_fraction): New function.
8513 (aarch64_run_selftests): Call it.
8515 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8517 * config/aarch64/aarch64-protos.h (tune_params::sve_width): Turn
8519 * config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Update
8521 (aarch64_estimated_poly_value): Likewise. Use the least significant
8522 set bit for the minimum and likely values. Use the most significant
8523 set bit for the maximum value.
8525 2021-08-03 liuhongt <hongtao.liu@intel.com>
8527 * config/i386/sse.md (cond_<insn><mode>): New expander.
8528 (cond_mul<mode>): Ditto.
8530 2021-08-03 Kewen Lin <linkw@linux.ibm.com>
8532 * tree-cfg.c (move_sese_region_to_fn): Fix typos on dloop.
8534 2021-08-03 liuhongt <hongtao.liu@intel.com>
8536 * config/i386/sse.md (cond_<insn><mode>):New expander.
8537 (cond_mul<mode>): Ditto.
8538 (cond_div<mode>): Ditto.
8540 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
8542 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Also
8543 check stack_realign_needed for stack realignment.
8544 (ix86_legitimate_constant_p): Always allow CONST_WIDE_INT smaller
8545 than the largest integer supported by vector register.
8546 * config/i386/i386.h (MAX_MOVE_MAX): New. Set to 64.
8547 (MOVE_MAX): Set to bytes of the largest integer supported by
8549 (STORE_MAX_PIECES): New.
8551 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
8553 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
8554 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
8555 data from one memory location to another.
8557 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
8560 * config/i386/i386.c (TARGET_GEN_MEMSET_SCRATCH_RTX): New.
8562 2021-08-02 Aldy Hernandez <aldyh@redhat.com>
8564 PR tree-optimization/101724
8565 * params.opt: Remove --param=threader-iterative.
8566 * tree-ssa-threadbackward.c (pass_thread_jumps::execute): Remove
8569 2021-08-02 Tom de Vries <tdevries@suse.de>
8571 PR middle-end/101665
8572 * doc/extend.texi (nonnull attribute): Improve documentation.
8574 2021-08-02 Andrew Pinski <apinski@marvell.com>
8576 PR rtl-optimization/101683
8577 * rtlanal.c (may_trap_p_1): Handle UNSIGNED_FIX.
8579 2021-08-02 Roger Sayle <roger@nextmovesoftware.com>
8581 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern):
8582 Renamed from cond_removal_in_popcount_clz_ctz_pattern.
8583 Add support for BSWAP, FFS, PARITY and CLRSB builtins.
8584 (tree_ssa_phiop_worker): Update call to function above.
8586 2021-08-01 H.J. Lu <hjl.tools@gmail.com>
8589 * config/i386/i386.md (bsr_rex64_1_zext): New.
8590 (combine splitter for constant - clzll): Replace gen_bsr_rex64_1
8591 with gen_bsr_rex64_1_zext.
8593 2021-07-31 Jakub Jelinek <jakub@redhat.com>
8596 * config/i386/i386.md (bsr_rex64_1, bsr_1, bsr_zext_1): New
8597 define_insn patterns.
8598 (*bsr_rex64_2, *bsr_2): New define_insn_and_split patterns.
8599 Add combine splitters for constant - clz.
8600 (clz<mode>2): Use a temporary pseudo for bsr result.
8602 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
8604 * config/rs6000/smmintrin.h (_mm_floor_pd, _mm_floor_ps,
8605 _mm_floor_sd, _mm_floor_ss): New.
8607 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
8609 * config/rs6000/smmintrin.h (_mm_ceil_pd, _mm_ceil_ps,
8610 _mm_ceil_sd, _mm_ceil_ss): New.
8612 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
8614 * config/rs6000/smmintrin.h (_mm_blend_pd, _mm_blendv_pd,
8615 _mm_blend_ps, _mm_blendv_ps): New.
8617 2021-07-30 Roger Sayle <roger@nextmovesoftware.com>
8618 Uroš Bizjak <ubizjak@gmail.com>
8620 * config/i386/i386.md (*dec_cmov<mode>): New define_insn_and_split
8621 to generate a conditional move using the carry flag after sub $1.
8622 (peephole2): Eliminate a register-to-register move by inverting
8623 the condition of a conditional move.
8625 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
8627 * config/mmix/mmix.md ("call", "call_value", "*call_real")
8628 ("*call_value_real"): Don't generate rtx mentioning the generic
8629 operands 1 and 2 to "call", and similarly for "call_value".
8630 * config/mmix/mmix.c (mmix_print_operand_punct_valid_p)
8631 (mmix_print_operand): Use '!' instead of 'p'.
8633 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
8635 * doc/md.texi (call): Correct information about operand 2.
8636 * config/mmix/mmix.md ("call", "call_value"): Remove fixed FIXMEs.
8638 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
8640 * range-op.cc (operator_trunc_mod::wi_fold): Fold constants.
8642 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
8644 * range-op.cc (operator_div::wi_fold): Return UNDEFINED for [0, 0] divisor.
8646 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
8648 * gimple-range-cache.cc (*::set_bb_range): Change const basic_block to
8650 (*::get_bb_range): Ditto.
8651 (*::bb_range_p): Ditto.
8652 * gimple-range-cache.h: Change prototypes.
8654 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
8657 * builtins.c (builtin_memcpy_read_str): Change the mode argument
8658 from scalar_int_mode to fixed_size_mode.
8659 (builtin_strncpy_read_str): Likewise.
8660 (gen_memset_value_from_prev): New function.
8661 (builtin_memset_read_str): Change the mode argument from
8662 scalar_int_mode to fixed_size_mode. Use gen_memset_value_from_prev
8663 and support CONST_VECTOR.
8664 (builtin_memset_gen_str): Likewise.
8665 (try_store_by_multiple_pieces): Use by_pieces_constfn to declare
8667 * builtins.h (builtin_strncpy_read_str): Replace scalar_int_mode
8668 with fixed_size_mode.
8669 (builtin_memset_read_str): Likewise.
8670 * expr.c (widest_int_mode_for_size): Renamed to ...
8671 (widest_fixed_size_mode_for_size): Add a bool argument to
8672 indicate if QI vector mode can be used.
8673 (by_pieces_ninsns): Call widest_fixed_size_mode_for_size
8674 instead of widest_int_mode_for_size.
8675 (pieces_addr::adjust): Change the mode argument from
8676 scalar_int_mode to fixed_size_mode.
8677 (op_by_pieces_d): Make m_len read-only. Add a bool member,
8678 m_qi_vector_mode, to indicate that QI vector mode can be used.
8679 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
8680 initialize m_qi_vector_mode. Call widest_fixed_size_mode_for_size
8681 instead of widest_int_mode_for_size.
8682 (op_by_pieces_d::get_usable_mode): Change the mode argument from
8683 scalar_int_mode to fixed_size_mode. Call
8684 widest_fixed_size_mode_for_size instead of
8685 widest_int_mode_for_size.
8686 (op_by_pieces_d::smallest_fixed_size_mode_for_size): New member
8687 function to return the smallest integer or QI vector mode.
8688 (op_by_pieces_d::run): Call widest_fixed_size_mode_for_size
8689 instead of widest_int_mode_for_size. Call
8690 smallest_fixed_size_mode_for_size instead of
8691 smallest_int_mode_for_size.
8692 (store_by_pieces_d::store_by_pieces_d): Add a bool argument to
8693 indicate that QI vector mode can be used and pass it to
8694 op_by_pieces_d::op_by_pieces_d.
8695 (can_store_by_pieces): Call widest_fixed_size_mode_for_size
8696 instead of widest_int_mode_for_size. Pass memsetp to
8697 widest_fixed_size_mode_for_size to support QI vector mode.
8698 Allow all CONST_VECTORs for memset if vec_duplicate is supported.
8699 (store_by_pieces): Pass memsetp to
8700 store_by_pieces_d::store_by_pieces_d.
8701 (clear_by_pieces_1): Removed.
8702 (clear_by_pieces): Replace clear_by_pieces_1 with
8703 builtin_memset_read_str and pass true to store_by_pieces_d to
8704 support vector mode broadcast.
8705 (string_cst_read_str): Change the mode argument from
8706 scalar_int_mode to fixed_size_mode.
8707 * expr.h (by_pieces_constfn): Change scalar_int_mode to
8709 (by_pieces_prev): Likewise.
8710 * rtl.h (lowpart_subreg_regno): New.
8711 * rtlanal.c (lowpart_subreg_regno): New. A wrapper around
8712 simplify_subreg_regno.
8713 * target.def (gen_memset_scratch_rtx): New hook.
8714 * doc/tm.texi.in: Add TARGET_GEN_MEMSET_SCRATCH_RTX.
8715 * doc/tm.texi: Regenerated.
8717 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
8720 * config/mips/mips.c (mips_atomic_assign_expand_fenv): Use
8721 TARGET_EXPR instead of MODIFY_EXPR.
8723 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
8726 * config/mips/mips-protos.h (mips_expand_vec_cmp_expr): Declare.
8727 * config/mips/mips.c (mips_expand_vec_cmp_expr): New function.
8728 * config/mips/mips-msa.md (vec_cmp<MSA:mode><mode_i>): New
8730 (vec_cmpu<IMSA:mode><mode_i>): New expander.
8732 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
8735 * config/i386/i386-options.c (ix86_option_override_internal):
8736 Don't enable LZCNT/POPCNT if they have been disabled explicitly.
8738 2021-07-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
8741 * config/arm/arm_neon.h (vld1_p64): Replace call to builtin by
8742 explicitly dereferencing __a.
8743 (vld1_s64): Likewise.
8744 (vld1_u64): Likewise.
8745 * config/arm/arm_neon_builtins.def (vld1): Remove entry for di
8746 and change to VAR13.
8748 2021-07-30 Aldy Hernandez <aldyh@redhat.com>
8750 * gimple-loop-versioning.cc (lv_dom_walker::lv_dom_walker): Remove
8751 use of m_range_analyzer.
8752 (loop_versioning::lv_dom_walker::before_dom_children): Same.
8753 (loop_versioning::lv_dom_walker::after_dom_children): Remove.
8754 (loop_versioning::prune_loop_conditions): Replace vr_values use
8755 with range_query interface.
8756 (pass_loop_versioning::execute): Use ranger.
8758 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
8761 * ipa-devirt.c (ipa_odr_read_section): Compare the precision of
8762 enum values, and emit a warning if they mismatch.
8764 2021-07-30 Kewen Lin <linkw@linux.ibm.com>
8766 * cfgloop.h (as_const): New function.
8767 (class loop_iterator): Rename to ...
8768 (class loops_list): ... this.
8769 (loop_iterator::next): Rename to ...
8770 (loops_list::Iter::fill_curr_loop): ... this and adjust.
8771 (loop_iterator::loop_iterator): Rename to ...
8772 (loops_list::loops_list): ... this and adjust.
8773 (loops_list::Iter): New class.
8774 (loops_list::iterator): New type.
8775 (loops_list::const_iterator): New type.
8776 (loops_list::begin): New function.
8777 (loops_list::end): Likewise.
8778 (loops_list::begin const): Likewise.
8779 (loops_list::end const): Likewise.
8780 (FOR_EACH_LOOP): Remove.
8781 (FOR_EACH_LOOP_FN): Remove.
8782 * cfgloop.c (flow_loops_dump): Adjust FOR_EACH_LOOP* with range-based
8783 for loop with loops_list instance.
8784 (sort_sibling_loops): Likewise.
8785 (disambiguate_loops_with_multiple_latches): Likewise.
8786 (verify_loop_structure): Likewise.
8787 * cfgloopmanip.c (create_preheaders): Likewise.
8788 (force_single_succ_latches): Likewise.
8789 * config/aarch64/falkor-tag-collision-avoidance.c
8790 (execute_tag_collision_avoidance): Likewise.
8791 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Likewise.
8792 * config/s390/s390.c (s390_adjust_loops): Likewise.
8793 * doc/loop.texi: Likewise.
8794 * gimple-loop-interchange.cc (pass_linterchange::execute): Likewise.
8795 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
8796 * gimple-loop-versioning.cc (loop_versioning::analyze_blocks): Likewise.
8797 (loop_versioning::make_versioning_decisions): Likewise.
8798 * gimple-ssa-split-paths.c (split_paths): Likewise.
8799 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
8800 * graphite.c (canonicalize_loop_form): Likewise.
8801 (graphite_transform_loops): Likewise.
8802 * ipa-fnsummary.c (analyze_function_body): Likewise.
8803 * ipa-pure-const.c (analyze_function): Likewise.
8804 * loop-doloop.c (doloop_optimize_loops): Likewise.
8805 * loop-init.c (loop_optimizer_finalize): Likewise.
8806 (fix_loop_structure): Likewise.
8807 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
8808 (move_loop_invariants): Likewise.
8809 * loop-unroll.c (decide_unrolling): Likewise.
8810 (unroll_loops): Likewise.
8811 * modulo-sched.c (sms_schedule): Likewise.
8812 * predict.c (predict_loops): Likewise.
8813 (pass_profile::execute): Likewise.
8814 * profile.c (branch_prob): Likewise.
8815 * sel-sched-ir.c (sel_finish_pipelining): Likewise.
8816 (sel_find_rgns): Likewise.
8817 * tree-cfg.c (replace_loop_annotate): Likewise.
8818 (replace_uses_by): Likewise.
8819 (move_sese_region_to_fn): Likewise.
8820 * tree-if-conv.c (pass_if_conversion::execute): Likewise.
8821 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
8822 * tree-parloops.c (parallelize_loops): Likewise.
8823 * tree-predcom.c (tree_predictive_commoning): Likewise.
8824 * tree-scalar-evolution.c (scev_initialize): Likewise.
8825 (scev_reset): Likewise.
8826 * tree-ssa-dce.c (find_obviously_necessary_stmts): Likewise.
8827 * tree-ssa-live.c (remove_unused_locals): Likewise.
8828 * tree-ssa-loop-ch.c (ch_base::copy_headers): Likewise.
8829 * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
8830 (tree_ssa_lim_initialize): Likewise.
8831 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Likewise.
8832 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize): Likewise.
8833 * tree-ssa-loop-manip.c (get_loops_exits): Likewise.
8834 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations): Likewise.
8835 (free_numbers_of_iterations_estimates): Likewise.
8836 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Likewise.
8837 * tree-ssa-loop-split.c (tree_ssa_split_loops): Likewise.
8838 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Likewise.
8839 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
8840 (pass_scev_cprop::execute): Likewise.
8841 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
8842 * tree-ssa-sccvn.c (do_rpo_vn): Likewise.
8843 * tree-ssa-threadupdate.c
8844 (jump_thread_path_registry::thread_through_all_blocks): Likewise.
8845 * tree-vectorizer.c (vectorize_loops): Likewise.
8846 * tree-vrp.c (vrp_asserts::find_assert_locations): Likewise.
8848 2021-07-29 Hans-Peter Nilsson <hp@bitrange.com>
8850 * config/mmix/mmix.c (mmix_function_arg_1): Avoid
8851 generating a VOIDmode register for e.g the
8852 function_arg_info::end_marker.
8854 2021-07-29 Jeff Law <jeffreyalaw@gmail.com>
8856 * config/h8300/h8300-modes.def: Add CCZ, CCV and CCC, drop CCZNV.
8857 * config/h8300/h8300.md (H8cc mode iterator): Add CCZ.
8858 (cc mode_attr): Similarly.
8859 (ccz subst_attr): Similarly.
8860 * config/h8300/jumpcall.md: Add new patterns for branch-on-bit.
8861 * config/h8300/testcompare.md: Remove various cc0 based patterns
8862 that had been commented out. Add pattern to set CCZ from a bit
8865 2021-07-29 Thomas Schwinge <thomas@codesourcery.com>
8866 Julian Brown <julian@codesourcery.com>
8867 Kwok Cheung Yeung <kcy@codesourcery.com>
8869 * omp-offload.c (oacc_loop_xform_head_tail, oacc_loop_process):
8870 'update_stmt' after modification.
8871 (pass_oacc_loop_designation): New function, extracted out of...
8872 (pass_oacc_device_lower): ... this.
8873 (pass_data_oacc_loop_designation, pass_oacc_loop_designation)
8874 (make_pass_oacc_loop_designation): New
8875 * passes.def: Add it.
8876 * tree-parloops.c (create_parallel_loop): Adjust.
8877 * tree-pass.h (make_pass_oacc_loop_designation): New.
8879 2021-07-29 Aldy Hernandez <aldyh@redhat.com>
8881 * flag-types.h (enum threader_mode): New.
8882 * params.opt: Add entry for --param=threader-mode.
8883 * tree-ssa-threadbackward.c (THREADER_ITERATIVE_MODE): New.
8884 (class back_threader): New.
8885 (back_threader::back_threader): New.
8886 (back_threader::~back_threader): New.
8887 (back_threader::maybe_register_path): New.
8888 (back_threader::find_taken_edge): New.
8889 (back_threader::find_taken_edge_switch): New.
8890 (back_threader::find_taken_edge_cond): New.
8891 (back_threader::resolve_def): New.
8892 (back_threader::resolve_phi): New.
8893 (back_threader::find_paths_to_names): New.
8894 (back_threader::find_paths): New.
8897 (thread_jumps::find_jump_threads_backwards): Call ranger threader.
8898 (thread_jumps::find_jump_threads_backwards_with_ranger): New.
8899 (pass_thread_jumps::execute): Abstract out code...
8900 (try_thread_blocks): ...here.
8901 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
8902 Abstract out threading candidate code to...
8903 (single_succ_to_potentially_threadable_block): ...here.
8904 * tree-ssa-threadedge.h (single_succ_to_potentially_threadable_block):
8906 * tree-ssa-threadupdate.c (register_jump_thread): Return boolean.
8907 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
8908 Return bool from register_jump_thread.
8910 2021-07-29 Andreas Krebbel <krebbel@linux.ibm.com>
8912 * target.def: in0 and in1 do not need to be registers.
8913 * doc/tm.texi: Regenerate.
8915 2021-07-29 liuhongt <hongtao.liu@intel.com>
8918 * config/i386/i386.c (ix86_widen_mult_cost): New function.
8919 (ix86_add_stmt_cost): Use ix86_widen_mult_cost for
8922 2021-07-29 Jiufu Guo <guojiufu@linux.ibm.com>
8925 * config/rs6000/rs6000.c (TARGET_PREFERRED_DOLOOP_MODE): New hook.
8926 (rs6000_preferred_doloop_mode): New hook.
8927 * doc/tm.texi: Regenerate.
8928 * doc/tm.texi.in: Add hook preferred_doloop_mode.
8929 * target.def (preferred_doloop_mode): New hook.
8930 * targhooks.c (default_preferred_doloop_mode): New hook.
8931 * targhooks.h (default_preferred_doloop_mode): New hook.
8932 * tree-ssa-loop-ivopts.c (compute_doloop_base_on_mode): New function.
8933 (add_iv_candidate_for_doloop): Call targetm.preferred_doloop_mode
8934 and compute_doloop_base_on_mode.
8936 2021-07-28 Martin Sebor <msebor@redhat.com>
8938 PR middle-end/101494
8939 * tree-ssa-uninit.c (maybe_warn_operand): Correct object offset
8940 and size computation.
8942 2021-07-28 Martin Sebor <msebor@redhat.com>
8944 PR middle-end/101601
8945 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Remove
8947 Handle pointers to functions.
8949 2021-07-28 Martin Sebor <msebor@redhat.com>
8951 * Makefile.in (OBJS): Add gimple-ssa-warn-access.o and pointer-query.o.
8952 * attribs.h (fndecl_dealloc_argno): Move fndecl_dealloc_argno to tree.h.
8953 * builtins.c (compute_objsize_r): Move to pointer-query.cc.
8954 (access_ref::access_ref): Same.
8955 (access_ref::phi): Same.
8956 (access_ref::get_ref): Same.
8957 (access_ref::size_remaining): Same.
8958 (access_ref::offset_in_range): Same.
8959 (access_ref::add_offset): Same.
8960 (access_ref::inform_access): Same.
8961 (ssa_name_limit_t::visit_phi): Same.
8962 (ssa_name_limit_t::leave_phi): Same.
8963 (ssa_name_limit_t::next): Same.
8964 (ssa_name_limit_t::next_phi): Same.
8965 (ssa_name_limit_t::~ssa_name_limit_t): Same.
8966 (pointer_query::pointer_query): Same.
8967 (pointer_query::get_ref): Same.
8968 (pointer_query::put_ref): Same.
8969 (pointer_query::flush_cache): Same.
8970 (warn_string_no_nul): Move to gimple-ssa-warn-access.cc.
8971 (check_nul_terminated_array): Same.
8972 (unterminated_array): Same.
8973 (maybe_warn_for_bound): Same.
8974 (check_read_access): Same.
8975 (warn_for_access): Same.
8976 (get_size_range): Same.
8977 (check_access): Same.
8978 (gimple_call_alloc_size): Move to tree.c.
8979 (gimple_parm_array_size): Move to pointer-query.cc.
8980 (get_offset_range): Same.
8981 (gimple_call_return_array): Same.
8982 (handle_min_max_size): Same.
8983 (handle_array_ref): Same.
8984 (handle_mem_ref): Same.
8985 (compute_objsize): Same.
8986 (gimple_call_alloc_p): Move to gimple-ssa-warn-access.cc.
8987 (call_dealloc_argno): Same.
8988 (fndecl_dealloc_argno): Same.
8989 (new_delete_mismatch_p): Same.
8990 (matching_alloc_calls_p): Same.
8991 (warn_dealloc_offset): Same.
8992 (maybe_emit_free_warning): Same.
8993 * builtins.h (check_nul_terminated_array): Move to
8994 gimple-ssa-warn-access.h.
8995 (check_nul_terminated_array): Same.
8996 (warn_string_no_nul): Same.
8997 (unterminated_array): Same.
8998 (class ssa_name_limit_t): Same.
8999 (class pointer_query): Same.
9000 (struct access_ref): Same.
9001 (class range_query): Same.
9002 (struct access_data): Same.
9003 (gimple_call_alloc_size): Same.
9004 (gimple_parm_array_size): Same.
9005 (compute_objsize): Same.
9006 (class access_data): Same.
9007 (maybe_emit_free_warning): Same.
9008 * calls.c (initialize_argument_information): Remove call to
9009 maybe_emit_free_warning.
9010 * gimple-array-bounds.cc: Include new header..
9011 * gimple-fold.c: Same.
9012 * gimple-ssa-sprintf.c: Same.
9013 * gimple-ssa-warn-restrict.c: Same.
9014 * passes.def: Add pass_warn_access.
9015 * tree-pass.h (make_pass_warn_access): Declare.
9016 * tree-ssa-strlen.c: Include new headers.
9017 * tree.c (fndecl_dealloc_argno): Move here from builtins.c.
9018 * tree.h (fndecl_dealloc_argno): Move here from attribs.h.
9019 * gimple-ssa-warn-access.cc: New file.
9020 * gimple-ssa-warn-access.h: New file.
9021 * pointer-query.cc: New file.
9022 * pointer-query.h: New file.
9024 2021-07-28 Jakub Jelinek <jakub@redhat.com>
9026 PR middle-end/101624
9027 * ubsan.c (maybe_instrument_pointer_overflow,
9028 instrument_object_size): Only test DECL_REGISTER on VAR_DECLs,
9029 PARM_DECLs or RESULT_DECLs.
9030 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
9032 2021-07-28 Jakub Jelinek <jakub@redhat.com>
9034 PR middle-end/101642
9035 * match.pd (bswap16 (x) == bswap16 (y)): Cast both operands
9036 to type of bswap16 for comparison.
9037 (bswap16 (x) == cst): Cast bswap16 operand to type of cst.
9039 2021-07-28 Richard Biener <rguenther@suse.de>
9041 PR tree-optimization/101615
9042 * tree-vect-slp.c (vect_optimize_slp): Materialize permutes
9043 at CTOR SLP graph entries.
9045 2021-07-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9047 * config/aarch64/aarch64.md (*extend<SHORT:mode><GPI:mode>2_aarch64):
9048 Add "r,w" alternative.
9050 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
9053 * config/i386/i386.c (ix86_avx_u128_mode_needed): Don't set
9054 AVX_U128_DIRTY when all bits are zero.
9056 2021-07-28 Richard Biener <rguenther@suse.de>
9058 PR tree-optimization/101615
9059 * tree-vect-slp.c (vect_optimize_slp): Pre-existing vector
9060 external nodes cannot be permuted so make them perm_out 0.
9062 2021-07-28 Andrew Stubbs <ams@codesourcery.com>
9065 * config.in: Regenerate.
9066 * config/gcn/gcn-hsa.h (A_FIJI): New define.
9067 (A_900): New define.
9068 (A_906): New define.
9069 (A_908): New define.
9070 (ASM_SPEC): Use A_FIJI, A_900, A_906 and A_908.
9071 * config/gcn/gcn.c (output_file_start): Adjust attributes according
9072 to the assembler capabilities.
9073 * config/gcn/mkoffload.c (main): Likewise.
9074 * configure: Regenerate.
9075 * configure.ac: Add tests for LLVM assembler attribute features.
9077 2021-07-28 Andrew MacLeod <amacleod@redhat.com>
9079 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Check for
9080 cond_false and cond_true on branches.
9082 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
9084 * config/aarch64/aarch64.c (aarch64_gen_adjusted_ldpstp): use
9087 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
9089 * alias.c (init_alias_analysis): Don't skip prologue/epilogue.
9091 2021-07-28 Jakub Jelinek <jakub@redhat.com>
9094 * config/i386/sse.md (vashr<mode>3): Split into vashrv8di3 expander
9095 and vashrv4di3 expander, where the latter requires just TARGET_AVX2
9096 and has special !TARGET_AVX512VL expansion.
9097 (vashrv2di3<mask_name>): Rename to ...
9098 (vashrv2di3): ... this. Change condition to TARGET_XOP || TARGET_AVX2
9099 and add special !TARGET_XOP && !TARGET_AVX512VL expansion.
9101 2021-07-28 Martin Uecker <muecker@gwdg.de>
9103 * calls.c (maybe_warn_rdwr_sizes): Correct argument
9104 numbers in warning that were switched.
9106 2021-07-28 Kewen Lin <linkw@linux.ibm.com>
9108 PR tree-optimization/101596
9109 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Fix wrong check
9110 by using new_type's precision instead.
9112 2021-07-28 liuhongt <hongtao.liu@intel.com>
9115 * config/i386/i386.h (processor_costs): Add new member
9117 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
9118 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
9119 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
9120 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
9121 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
9122 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
9123 generic_cost, core_cost): Initialize integer_to_sse same value
9125 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
9126 * config/i386/i386.c (ix86_builtin_vectorization_cost):
9127 Use integer_to_sse instead of sse_op to calculate the cost of
9130 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
9132 * config/rs6000/rs6000-gen-builtins.c (write_ovld_static_init): New
9134 (write_init_file): Call write_ovld_static_init.
9136 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
9138 * config/rs6000/rs6000-gen-builtins.c (write_bif_static_init): New
9140 (write_init_file): Call write_bif_static_init.
9142 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
9144 * config/rs6000/rs6000-gen-builtins.c (typemap): New struct.
9145 (TYPE_MAP_SIZE): New macro.
9146 (type_map): New initialized variable.
9147 (typemap_cmp): New function.
9148 (write_type_node): Likewise.
9149 (write_fntype_init): Implement.
9151 2021-07-27 Martin Sebor <msebor@redhat.com>
9153 PR tree-optimization/101584
9154 * tree-ssa-uninit.c (builtin_call_nomodifying_p): New function.
9155 (check_defs): Call it.
9157 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
9159 * tree-ssa-dom.c (dom_jump_threader_simplifier):
9160 Put avail_exprs_stack in the class, instead of passing it to
9161 jump_threader_simplifier.
9162 (dom_jump_threader_simplifier::simplify): Add state argument.
9163 (dom_opt_dom_walker): Add state.
9164 (pass_dominator::execute): Pass state to threader.
9165 (dom_opt_dom_walker::before_dom_children): Use state.
9166 * tree-ssa-threadedge.c (jump_threader::jump_threader): Replace
9168 (jump_threader::record_temporary_equivalences_from_phis):
9169 Register equivalences through the state variable.
9170 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
9171 Record ranges in a statement through the state variable.
9172 (jump_threader::simplify_control_stmt_condition): Pass state to
9174 (jump_threader::simplify_control_stmt_condition_1): Same.
9175 (jump_threader::thread_around_empty_blocks): Remove obsolete
9177 (jump_threader::thread_through_normal_block): Record equivalences
9178 on edge through the state variable.
9179 (jump_threader::thread_across_edge): Abstract state pushing.
9180 (jt_state::jt_state): New.
9181 (jt_state::push): New.
9182 (jt_state::pop): New.
9183 (jt_state::register_equiv): New.
9184 (jt_state::record_ranges_from_stmt): New.
9185 (jt_state::register_equivs_on_edge): New.
9186 (jump_threader_simplifier::jump_threader_simplifier): Move from
9188 (jump_threader_simplifier::simplify): Add state argument.
9189 * tree-ssa-threadedge.h (class jt_state): New.
9190 (class jump_threader): Add state to constructor.
9191 (class jump_threader_simplifier): Add state to simplify. Remove
9192 avail_exprs_stack from class.
9193 * tree-vrp.c (vrp_jump_threader_simplifier::simplify): Add state
9195 (vrp_jump_threader::vrp_jump_threader): Add state.
9196 (vrp_jump_threader::~vrp_jump_threader): Cleanup state.
9198 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
9200 * Makefile.in (OBJS): Add gimple-range-path.o.
9201 * gimple-range-path.cc: New file.
9202 * gimple-range-path.h: New file.
9204 2021-07-27 Jonathan Wright <jonathan.wright@arm.com>
9206 * config/aarch64/aarch64-simd.md: Push sign/zero-extension
9207 inside vec_duplicate for all patterns.
9208 * simplify-rtx.c (simplify_context::simplify_unary_operation_1):
9209 Push sign/zero-extension inside vec_duplicate.
9211 2021-07-27 Richard Biener <rguenther@suse.de>
9213 PR tree-optimization/101573
9214 * tree-ssa-uninit.c (warn_uninit_phi_uses): New function
9215 looking at uninitialized PHI arg defs in some constrained cases.
9216 (warn_uninitialized_vars): Call it.
9217 (execute_early_warn_uninitialized): Calculate dominators.
9219 2021-07-27 Richard Biener <rguenther@suse.de>
9221 PR tree-optimization/39821
9222 * tree-vect-stmts.c (vect_model_promotion_demotion_cost): Use
9223 vector_stmt for widening arithmetic.
9224 (vectorizable_conversion): Adjust.
9226 2021-07-27 Martin Jambor <mjambor@suse.cz>
9228 * cgraph.h (ipa_replace_map): New field force_load_ref.
9229 * ipa-prop.h (ipa_param_descriptor): Reduce precision of move_cost,
9230 aded new flag load_dereferenced, adjusted comments.
9231 (ipa_get_param_dereferenced): New function.
9232 (ipa_set_param_dereferenced): Likewise.
9233 * cgraphclones.c (cgraph_node::create_virtual_clone): Follow it.
9234 * ipa-cp.c: Include gimple.h.
9235 (ipcp_discover_new_direct_edges): Take into account dereferenced flag.
9236 (get_replacement_map): New parameter force_load_ref, set the
9237 appropriate flag in ipa_replace_map if set.
9238 (struct symbol_and_index_together): New type.
9239 (adjust_refs_in_act_callers): New function.
9240 (adjust_references_in_caller): Likewise.
9241 (create_specialized_node): When appropriate, call
9242 adjust_references_in_caller and force only load references.
9243 * ipa-prop.c (load_from_dereferenced_name): New function.
9244 (ipa_analyze_controlled_uses): Also detect loads from a
9245 dereference, harden testing of call statements.
9246 (ipa_write_node_info): Stream the dereferenced flag.
9247 (ipa_read_node_info): Likewise.
9248 (ipa_set_jf_constant): Also create refdesc when jump function
9249 references a variable.
9250 (cgraph_node_for_jfunc): Rename to symtab_node_for_jfunc, work
9251 also on references of variables and return a symtab_node. Adjust
9253 (propagate_controlled_uses): Also remove references to VAR_DECLs.
9255 2021-07-27 Jakub Jelinek <jakub@redhat.com>
9257 PR middle-end/101586
9258 * gimple-fold.c (clear_padding_type): Ignore FIELD_DECLs with byte
9259 positions above or equal to sz except for diagnostics of flexible
9262 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
9264 PR tree-optimization/78888
9265 * gimple-range-fold.cc (get_letter_range): New.
9266 (fold_using_range::range_of_builtin_call): Call get_letter_range.
9268 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
9270 PR tree-optimization/78888
9271 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Add cases
9272 for CFN_BUILT_IN_TOUPPER and CFN_BUILT_IN_TOLOWER.
9274 2021-07-26 Roger Sayle <roger@nextmovesoftware.com>
9275 Marc Glisse <marc.glisse@inria.fr>
9277 * match.pd (rotate): Simplify equality/inequality of rotations.
9278 (bswap): Simplify equality/inequality tests of byte swapping.
9280 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
9282 * range-op.cc (operator_bitwise_xor::op1_op2_relation_effect):
9285 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
9287 * range-op.cc (operator_lshift::fold_range): Pass rel to
9288 base class fold_range.
9289 (operator_rshift::fold_range): Same.
9291 2021-07-26 Ashimida <ashimida@linux.alibaba.com>
9294 * toplev.h (min_align_loops_log): Remove declaration.
9295 (min_align_jumps_log, min_align_labels_log): Likewise.
9296 (min_align_functions_log): Likewise.
9298 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
9300 * tree-vrp.c (vrp_simplify_cond_using_ranges): Rename vr_values
9302 (execute_vrp): Abstract out simplification of conditionals...
9303 (simplify_casted_conds): ...here.
9305 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
9307 * gimple-array-bounds.cc (array_bounds_checker::get_value_range):
9308 Add gimple argument.
9309 (array_bounds_checker::check_array_ref): Same.
9310 (array_bounds_checker::check_addr_expr): Same.
9311 (array_bounds_checker::check_array_bounds): Pass statement to
9312 check_array_bounds and check_addr_expr.
9313 * gimple-array-bounds.h (check_array_bounds): Add gimple argument.
9314 (check_addr_expr): Same.
9315 (get_value_range): Same.
9317 2021-07-26 Tamar Christina <tamar.christina@arm.com>
9319 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): Rename to..
9320 (sdot_prod, udot_prod): ... This.
9321 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): Merged
9323 (<sur>dot_prod<vsi2qi>): ... this.
9324 (aarch64_<sur>dot_lane<vsi2qi>, aarch64_<sur>dot_laneq<vsi2qi>):
9325 Change operands order.
9326 (<sur>sadv16qi): Use new operands order.
9327 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32,
9328 vdotq_s32): Use new RTL ordering.
9330 2021-07-26 Tamar Christina <tamar.christina@arm.com>
9332 * config/aarch64/aarch64-builtins.c (TYPES_TERNOP_SUSS,
9333 aarch64_types_ternop_suss_qualifiers): New.
9334 * config/aarch64/aarch64-simd-builtins.def (usdot_prod): Use it.
9335 * config/aarch64/aarch64-simd.md (usdot_prod<vsi2qi>): Re-organize RTL.
9336 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Use it.
9338 2021-07-23 Jakub Jelinek <jakub@redhat.com>
9340 PR rtl-optimization/101562
9341 * expmed.c (store_integral_bit_field): Only use movstrict_optab
9342 if the operand isn't paradoxical.
9344 2021-07-23 Aldy Hernandez <aldyh@redhat.com>
9346 * gimple-array-bounds.h (class array_bounds_checker): Change
9347 ranges type to range_query.
9349 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9351 * config/aarch64/arm_neon.h (vst1_s64_x2): Use
9352 __builtin_memcpy instead of constructing
9353 __builtin_aarch64_simd_oi one vector at a time.
9354 (vst1_u64_x2): Likewise.
9355 (vst1_f64_x2): Likewise.
9356 (vst1_s8_x2): Likewise.
9357 (vst1_p8_x2): Likewise.
9358 (vst1_s16_x2): Likewise.
9359 (vst1_p16_x2): Likewise.
9360 (vst1_s32_x2): Likewise.
9361 (vst1_u8_x2): Likewise.
9362 (vst1_u16_x2): Likewise.
9363 (vst1_u32_x2): Likewise.
9364 (vst1_f16_x2): Likewise.
9365 (vst1_f32_x2): Likewise.
9366 (vst1_p64_x2): Likewise.
9367 (vst1q_s8_x2): Likewise.
9368 (vst1q_p8_x2): Likewise.
9369 (vst1q_s16_x2): Likewise.
9370 (vst1q_p16_x2): Likewise.
9371 (vst1q_s32_x2): Likewise.
9372 (vst1q_s64_x2): Likewise.
9373 (vst1q_u8_x2): Likewise.
9374 (vst1q_u16_x2): Likewise.
9375 (vst1q_u32_x2): Likewise.
9376 (vst1q_u64_x2): Likewise.
9377 (vst1q_f16_x2): Likewise.
9378 (vst1q_f32_x2): Likewise.
9379 (vst1q_f64_x2): Likewise.
9380 (vst1q_p64_x2): Likewise.
9382 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9384 * config/aarch64/arm_neon.h (vst1_s64_x3): Use
9385 __builtin_memcpy instead of constructing
9386 __builtin_aarch64_simd_ci one vector at a time.
9387 (vst1_u64_x3): Likewise.
9388 (vst1_f64_x3): Likewise.
9389 (vst1_s8_x3): Likewise.
9390 (vst1_p8_x3): Likewise.
9391 (vst1_s16_x3): Likewise.
9392 (vst1_p16_x3): Likewise.
9393 (vst1_s32_x3): Likewise.
9394 (vst1_u8_x3): Likewise.
9395 (vst1_u16_x3): Likewise.
9396 (vst1_u32_x3): Likewise.
9397 (vst1_f16_x3): Likewise.
9398 (vst1_f32_x3): Likewise.
9399 (vst1_p64_x3): Likewise.
9400 (vst1q_s8_x3): Likewise.
9401 (vst1q_p8_x3): Likewise.
9402 (vst1q_s16_x3): Likewise.
9403 (vst1q_p16_x3): Likewise.
9404 (vst1q_s32_x3): Likewise.
9405 (vst1q_s64_x3): Likewise.
9406 (vst1q_u8_x3): Likewise.
9407 (vst1q_u16_x3): Likewise.
9408 (vst1q_u32_x3): Likewise.
9409 (vst1q_u64_x3): Likewise.
9410 (vst1q_f16_x3): Likewise.
9411 (vst1q_f32_x3): Likewise.
9412 (vst1q_f64_x3): Likewise.
9413 (vst1q_p64_x3): Likewise.
9415 2021-07-23 H.J. Lu <hjl.tools@gmail.com>
9418 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): Don't return
9419 hard register when LRA is in progress.
9421 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9423 * config/aarch64/arm_neon.h (vst1_s8_x4): Use
9424 __builtin_memcpy instead of using a union.
9425 (vst1q_s8_x4): Likewise.
9426 (vst1_s16_x4): Likewise.
9427 (vst1q_s16_x4): Likewise.
9428 (vst1_s32_x4): Likewise.
9429 (vst1q_s32_x4): Likewise.
9430 (vst1_u8_x4): Likewise.
9431 (vst1q_u8_x4): Likewise.
9432 (vst1_u16_x4): Likewise.
9433 (vst1q_u16_x4): Likewise.
9434 (vst1_u32_x4): Likewise.
9435 (vst1q_u32_x4): Likewise.
9436 (vst1_f16_x4): Likewise.
9437 (vst1q_f16_x4): Likewise.
9438 (vst1_f32_x4): Likewise.
9439 (vst1q_f32_x4): Likewise.
9440 (vst1_p8_x4): Likewise.
9441 (vst1q_p8_x4): Likewise.
9442 (vst1_p16_x4): Likewise.
9443 (vst1q_p16_x4): Likewise.
9444 (vst1_s64_x4): Likewise.
9445 (vst1_u64_x4): Likewise.
9446 (vst1_p64_x4): Likewise.
9447 (vst1q_s64_x4): Likewise.
9448 (vst1q_u64_x4): Likewise.
9449 (vst1q_p64_x4): Likewise.
9450 (vst1_f64_x4): Likewise.
9451 (vst1q_f64_x4): Likewise.
9453 2021-07-23 Jonathan Wrightt <jonathan.wright@arm.com>
9455 * config/aarch64/arm_neon.h (vst2_s64): Use __builtin_memcpy
9456 instead of constructing __builtin_aarch64_simd_oi one vector
9458 (vst2_u64): Likewise.
9459 (vst2_f64): Likewise.
9460 (vst2_s8): Likewise.
9461 (vst2_p8): Likewise.
9462 (vst2_s16): Likewise.
9463 (vst2_p16): Likewise.
9464 (vst2_s32): Likewise.
9465 (vst2_u8): Likewise.
9466 (vst2_u16): Likewise.
9467 (vst2_u32): Likewise.
9468 (vst2_f16): Likewise.
9469 (vst2_f32): Likewise.
9470 (vst2_p64): Likewise.
9471 (vst2q_s8): Likewise.
9472 (vst2q_p8): Likewise.
9473 (vst2q_s16): Likewise.
9474 (vst2q_p16): Likewise.
9475 (vst2q_s32): Likewise.
9476 (vst2q_s64): Likewise.
9477 (vst2q_u8): Likewise.
9478 (vst2q_u16): Likewise.
9479 (vst2q_u32): Likewise.
9480 (vst2q_u64): Likewise.
9481 (vst2q_f16): Likewise.
9482 (vst2q_f32): Likewise.
9483 (vst2q_f64): Likewise.
9484 (vst2q_p64): Likewise.
9486 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9488 * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy
9489 instead of constructing __builtin_aarch64_simd_ci one vector
9491 (vst3_u64): Likewise.
9492 (vst3_f64): Likewise.
9493 (vst3_s8): Likewise.
9494 (vst3_p8): Likewise.
9495 (vst3_s16): Likewise.
9496 (vst3_p16): Likewise.
9497 (vst3_s32): Likewise.
9498 (vst3_u8): Likewise.
9499 (vst3_u16): Likewise.
9500 (vst3_u32): Likewise.
9501 (vst3_f16): Likewise.
9502 (vst3_f32): Likewise.
9503 (vst3_p64): Likewise.
9504 (vst3q_s8): Likewise.
9505 (vst3q_p8): Likewise.
9506 (vst3q_s16): Likewise.
9507 (vst3q_p16): Likewise.
9508 (vst3q_s32): Likewise.
9509 (vst3q_s64): Likewise.
9510 (vst3q_u8): Likewise.
9511 (vst3q_u16): Likewise.
9512 (vst3q_u32): Likewise.
9513 (vst3q_u64): Likewise.
9514 (vst3q_f16): Likewise.
9515 (vst3q_f32): Likewise.
9516 (vst3q_f64): Likewise.
9517 (vst3q_p64): Likewise.
9519 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9521 * config/aarch64/arm_neon.h (vst4_s64): Use __builtin_memcpy
9522 instead of constructing __builtin_aarch64_simd_xi one vector
9524 (vst4_u64): Likewise.
9525 (vst4_f64): Likewise.
9526 (vst4_s8): Likewise.
9527 (vst4_p8): Likewise.
9528 (vst4_s16): Likewise.
9529 (vst4_p16): Likewise.
9530 (vst4_s32): Likewise.
9531 (vst4_u8): Likewise.
9532 (vst4_u16): Likewise.
9533 (vst4_u32): Likewise.
9534 (vst4_f16): Likewise.
9535 (vst4_f32): Likewise.
9536 (vst4_p64): Likewise.
9537 (vst4q_s8): Likewise.
9538 (vst4q_p8): Likewise.
9539 (vst4q_s16): Likewise.
9540 (vst4q_p16): Likewise.
9541 (vst4q_s32): Likewise.
9542 (vst4q_s64): Likewise.
9543 (vst4q_u8): Likewise.
9544 (vst4q_u16): Likewise.
9545 (vst4q_u32): Likewise.
9546 (vst4q_u64): Likewise.
9547 (vst4q_f16): Likewise.
9548 (vst4q_f32): Likewise.
9549 (vst4q_f64): Likewise.
9550 (vst4q_p64): Likewise.
9552 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9554 * config/aarch64/arm_neon.h (vtbx4_s8): Use __builtin_memcpy
9555 instead of constructing __builtin_aarch64_simd_oi one vector
9557 (vtbx4_u8): Likewise.
9558 (vtbx4_p8): Likewise.
9560 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9562 * config/aarch64/arm_neon.h (vtbl3_s8): Use __builtin_memcpy
9563 instead of constructing __builtin_aarch64_simd_oi one vector
9565 (vtbl3_u8): Likewise.
9566 (vtbl3_p8): Likewise.
9567 (vtbl4_s8): Likewise.
9568 (vtbl4_u8): Likewise.
9569 (vtbl4_p8): Likewise.
9571 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9573 * config/aarch64/arm_neon.h (vqtbx2_s8): Use __builtin_memcpy
9574 instead of constructing __builtin_aarch64_simd_oi one vector
9576 (vqtbx2_u8): Likewise.
9577 (vqtbx2_p8): Likewise.
9578 (vqtbx2q_s8): Likewise.
9579 (vqtbx2q_u8): Likewise.
9580 (vqtbx2q_p8): Likewise.
9581 (vqtbx3_s8): Use __builtin_memcpy instead of constructing
9582 __builtin_aarch64_simd_ci one vector at a time.
9583 (vqtbx3_u8): Likewise.
9584 (vqtbx3_p8): Likewise.
9585 (vqtbx3q_s8): Likewise.
9586 (vqtbx3q_u8): Likewise.
9587 (vqtbx3q_p8): Likewise.
9588 (vqtbx4_s8): Use __builtin_memcpy instead of constructing
9589 __builtin_aarch64_simd_xi one vector at a time.
9590 (vqtbx4_u8): Likewise.
9591 (vqtbx4_p8): Likewise.
9592 (vqtbx4q_s8): Likewise.
9593 (vqtbx4q_u8): Likewise.
9594 (vqtbx4q_p8): Likewise.
9596 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9598 * config/aarch64/arm_neon.h (vqtbl2_s8): Use __builtin_memcpy
9599 instead of constructing __builtin_aarch64_simd_oi one vector
9601 (vqtbl2_u8): Likewise.
9602 (vqtbl2_p8): Likewise.
9603 (vqtbl2q_s8): Likewise.
9604 (vqtbl2q_u8): Likewise.
9605 (vqtbl2q_p8): Likewise.
9606 (vqtbl3_s8): Use __builtin_memcpy instead of constructing
9607 __builtin_aarch64_simd_ci one vector at a time.
9608 (vqtbl3_u8): Likewise.
9609 (vqtbl3_p8): Likewise.
9610 (vqtbl3q_s8): Likewise.
9611 (vqtbl3q_u8): Likewise.
9612 (vqtbl3q_p8): Likewise.
9613 (vqtbl4_s8): Use __builtin_memcpy instead of constructing
9614 __builtin_aarch64_simd_xi one vector at a time.
9615 (vqtbl4_u8): Likewise.
9616 (vqtbl4_p8): Likewise.
9617 (vqtbl4q_s8): Likewise.
9618 (vqtbl4q_u8): Likewise.
9619 (vqtbl4q_p8): Likewise.
9621 2021-07-23 Haochen Gui <guihaoc@gcc.gnu.org>
9624 * config/rs6000/rs6000.md (cstore<mode>4): Fix wrong fall through.
9626 2021-07-22 Andrew Pinski <apinski@marvell.com>
9628 PR tree-optimization/10153
9629 * tree-tailcall.c (create_tailcall_accumulator):
9630 Don't call fold_convert as the type should be correct already.
9631 (tree_optimize_tail_calls_1): Use build_{one,zero}_cst instead
9632 of integer_{one,zero}_node for the call of create_tailcall_accumulator.
9634 2021-07-22 Aldy Hernandez <aldyh@redhat.com>
9636 * gimple-range-cache.cc (non_null_ref::adjust_range): Replace
9637 varying_p check for null/non-null check.
9639 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
9641 PR tree-optimization/101511
9642 * value-relation.cc (relation_oracle::query_relation): Check if ssa1
9643 is in ssa2's equiv set, and don't trap if so.
9645 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
9647 PR tree-optimization/101497
9648 * gimple-range-fold.cc (fold_using_range::range_of_cond_expr): Check
9651 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
9653 PR tree-optimization/101496
9654 * vr-values.c (simplify_using_ranges::fold_cond): Call range_of_stmt
9655 first, then vrp_visit_cond_Stmt.
9657 2021-07-22 liuhongt <hongtao.liu@intel.com>
9659 * config/i386/i386-expand.c
9660 (ix86_broadcast_from_integer_constant): Rename to ..
9661 (ix86_broadcast_from_constant): .. this, and extend it to
9663 (ix86_expand_vector_move): Extend to float mode.
9664 * config/i386/i386-features.c
9665 (replace_constant_pool_with_broadcast): Remove.
9666 (remove_partial_avx_dependency_gate): Ditto.
9667 (constant_pool_broadcast): Ditto.
9668 (class pass_constant_pool_broadcast): Ditto.
9669 (make_pass_constant_pool_broadcast): Ditto.
9670 (remove_partial_avx_dependency): Adjust gate.
9671 * config/i386/i386-passes.def: Remove pass_constant_pool_broadcast.
9672 * config/i386/i386-protos.h
9673 (make_pass_constant_pool_broadcast): Remove.
9675 2021-07-22 liuhongt <hongtao.liu@intel.com>
9677 * config/i386/constraints.md (Wb): New constraint.
9679 * config/i386/i386.md (*ashlhi3_1): Extend to avx512 mask
9681 (*ashlqi3_1): Ditto.
9682 (*<insn><mode>3_1): Split to ..
9683 (*ashr<mode>3_1): this, ...
9684 (*lshr<mode>3_1): and this, also extend this pattern to avx512
9686 (*<insn><mode>3_1): Split to ..
9687 (*ashr<mode>3_1): this, ...
9688 (*lshrqi3_1): and this, also extend this pattern to avx512
9690 (*lshrhi3_1): And this, also extend this pattern to avx512
9692 * config/i386/sse.md (k<code><mode>): New define_split after
9693 it to convert generic shift pattern to mask shift ones.
9695 2021-07-21 Thomas Schwinge <thomas@codesourcery.com>
9696 Joseph Myers <joseph@codesourcery.com>
9697 Cesar Philippidis <cesar@codesourcery.com>
9699 * tree-core.h (omp_clause_code): Add 'OMP_CLAUSE_NOHOST'.
9700 * tree.c (omp_clause_num_ops, omp_clause_code_name, walk_tree_1):
9702 * tree-pretty-print.c (dump_omp_clause): Likewise.
9703 * omp-general.c (oacc_verify_routine_clauses): Likewise.
9704 * gimplify.c (gimplify_scan_omp_clauses)
9705 (gimplify_adjust_omp_clauses): Likewise.
9706 * tree-nested.c (convert_nonlocal_omp_clauses)
9707 (convert_local_omp_clauses): Likewise.
9708 * omp-low.c (scan_sharing_clauses): Likewise.
9709 * omp-offload.c (execute_oacc_device_lower): Update.
9711 2021-07-21 Martin Sebor <msebor@redhat.com>
9713 * tree-ssa-alias.c (walk_aliased_vdefs_1): Fix typos in a comment.
9715 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9717 * config/rs6000/rs6000-gen-builtins.c (write_init_bif_table):
9720 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9722 * config/rs6000/rs6000-gen-builtins.c (write_fntype): New
9724 (write_fntype_init): New stub function.
9725 (write_init_bif_table): Likewise.
9726 (write_init_ovld_table): New function.
9727 (write_init_file): Implement.
9729 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9731 * config/rs6000/rs6000-gen-builtins.c
9732 (write_autogenerated_header): New function.
9733 (write_decls): Likewise.
9734 (write_extern_fntype): New callback function.
9735 (write_header_file): Implement.
9737 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9739 * config/rs6000/rs6000-gen-builtins.c (write_defines_file):
9742 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9744 * config/rs6000/rs6000-gen-builtins.c (complete_vector_type): New
9746 (complete_base_type): Likewise.
9747 (construct_fntype_id): Likewise.
9748 (parse_bif_entry): Call contruct_fntype_id.
9749 (parse_ovld_entry): Likewise.
9751 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9753 * config/rs6000/rs6000-gen-builtins.c (ovld_stanza): New struct.
9754 (MAXOVLDSTANZAS): New macro.
9755 (ovld_stanzas): New variable.
9756 (curr_ovld_stanza): Likewise.
9757 (MAXOVLDS): New macro.
9758 (ovlddata): New struct.
9759 (ovlds): New variable.
9760 (curr_ovld): Likewise.
9761 (max_ovld_args): Likewise.
9762 (parse_ovld_entry): New function.
9763 (parse_ovld_stanza): Likewise.
9764 (parse_ovld): Implement.
9766 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9768 * config/rs6000/rs6000-gen-builtins.c (parse_bif_attrs):
9771 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9773 * config/rs6000/rs6000-gen-builtins.c (parse_args): New function.
9774 (parse_prototype): Implement.
9776 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9778 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): New enum.
9779 (curr_bif_stanza): New variable.
9780 (stanza_entry): New struct.
9781 (stanza_map): New initialized variable.
9782 (enable_string): Likewise.
9783 (fnkinds): New enum.
9784 (typelist): New struct.
9785 (attrinfo): Likewise.
9786 (MAXRESTROPNDS): New macro.
9787 (prototype): New struct.
9788 (MAXBIFS): New macro.
9789 (bifdata): New struct.
9790 (bifs): New variable.
9791 (curr_bif): Likewise.
9792 (bif_order): Likewise.
9793 (bif_index): Likewise.
9794 (fatal): New function.
9795 (stanza_name_to_stanza): Likewise.
9796 (parse_bif_attrs): New stub function.
9797 (parse_prototype): Likewise.
9798 (parse_bif_entry): New function.
9799 (parse_bif_stanza): Likewise.
9800 (parse_bif): Implement.
9801 (set_bif_order): New function.
9802 (create_bif_order): Implement.
9804 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9806 * config/rs6000/rs6000-gen-builtins.c (rbtree.h): New #include.
9807 (num_bifs): New variable.
9808 (num_ovld_stanzas): Likewise.
9809 (num_ovlds): Likewise.
9810 (parse_codes): New enum.
9811 (bif_rbt): New variable.
9812 (ovld_rbt): Likewise.
9813 (fntype_rbt): Likewise.
9814 (bifo_rbt): Likewise.
9815 (parse_bif): New stub function.
9816 (create_bif_order): Likewise.
9817 (parse_ovld): Likewise.
9818 (write_header_file): Likewise.
9819 (write_init_file): Likewise.
9820 (write_defines_file): Likewise.
9821 (delete_output_files): New function.
9824 2021-07-21 H.J. Lu <hjl.tools@gmail.com>
9827 * config/i386/i386-builtin.def: Remove OPTION_MASK_ISA_SSE4_2
9828 from CRC32 _builtin functions.
9830 2021-07-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
9832 * coverage.c (build_gcov_info_var_registration): Mark the object placed
9833 in the linker set as referenced so that it does not get optimized away.
9835 2021-07-21 Kito Cheng <kito.cheng@sifive.com>
9838 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
9840 * config.gcc (riscv*-*-*): Detect which python is available.
9842 2021-07-21 Jakub Jelinek <jakub@redhat.com>
9844 PR middle-end/101535
9845 * gimplify.c (omp_check_private): Properly skip ORT_TARGET_DATA
9846 contexts in which decl isn't privatized and for ORT_TARGET return
9847 false if decl is mapped.
9849 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
9851 * gimple-loop-jam.c: Include tree-ssa-sccvn.h.
9852 (tree_loop_unroll_and_jam): Run value-numbering on a loop that
9853 has been successfully unrolled.
9855 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
9857 * tree-ssa-loop-manip.c (determine_exit_conditions): Return a null
9858 exit condition if no tail loop is needed, and if the original exit
9859 condition should therefore be kept as-is.
9860 (tree_transform_and_unroll_loop): Handle that case here too.
9862 2021-07-21 Kewen Lin <linkw@linux.ibm.com>
9864 * tree-data-ref.c (free_dependence_relations): Adjust to pass vec
9866 (free_data_refs): Likewise.
9867 * tree-data-ref.h (free_dependence_relations): Likewise.
9868 (free_data_refs): Likewise.
9869 * tree-predcom.c (struct chain): Use auto_vec instead of vec for
9871 (struct component): Likewise.
9872 (pcom_worker::pcom_worker): Adjust for auto_vec and renaming changes.
9873 (pcom_worker::~pcom_worker): Likewise.
9874 (pcom_worker::release_chain): Adjust as auto_vec changes.
9875 (pcom_worker::loop): Rename to ...
9876 (pcom_worker::m_loop): ... this.
9877 (pcom_worker::datarefs): Rename to ...
9878 (pcom_worker::m_datarefs): ... this. Use auto_vec instead of vec.
9879 (pcom_worker::dependences): Rename to ...
9880 (pcom_worker::m_dependences): ... this. Use auto_vec instead of vec.
9881 (pcom_worker::chains): Rename to ...
9882 (pcom_worker::m_chains): ... this. Use auto_vec instead of vec.
9883 (pcom_worker::looparound_phis): Rename to ...
9884 (pcom_worker::m_looparound_phis): ... this. Use auto_vec instead of
9886 (pcom_worker::cache): Rename to ...
9887 (pcom_worker::m_cache): ... this. Use auto_vec instead of vec.
9888 (pcom_worker::release_chain): Adjust for auto_vec changes.
9889 (pcom_worker::release_chains): Adjust for auto_vec and renaming
9891 (release_component): Remove.
9892 (release_components): Adjust for release_component removal.
9893 (component_of): Adjust to use vec.
9894 (merge_comps): Likewise.
9895 (pcom_worker::aff_combination_dr_offset): Adjust for renaming changes.
9896 (pcom_worker::determine_offset): Likewise.
9897 (class comp_ptrs): Remove.
9898 (pcom_worker::split_data_refs_to_components): Adjust for renaming
9899 changes, for comp_ptrs removal with auto_vec.
9900 (pcom_worker::suitable_component_p): Adjust for renaming changes.
9901 (pcom_worker::filter_suitable_components): Adjust for release_component
9903 (pcom_worker::valid_initializer_p): Adjust for renaming changes.
9904 (pcom_worker::find_looparound_phi): Likewise.
9905 (pcom_worker::add_looparound_copies): Likewise.
9906 (pcom_worker::determine_roots_comp): Likewise.
9907 (pcom_worker::single_nonlooparound_use): Likewise.
9908 (pcom_worker::execute_pred_commoning_chain): Likewise.
9909 (pcom_worker::execute_pred_commoning): Likewise.
9910 (pcom_worker::try_combine_chains): Likewise.
9911 (pcom_worker::prepare_initializers_chain): Likewise.
9912 (pcom_worker::prepare_initializers): Likewise.
9913 (pcom_worker::prepare_finalizers_chain): Likewise.
9914 (pcom_worker::prepare_finalizers): Likewise.
9915 (pcom_worker::tree_predictive_commoning_loop): Likewise.
9917 2021-07-20 Martin Sebor <msebor@redhat.com>
9919 PR middle-end/101397
9920 * builtins.c (gimple_call_return_array): Add argument. Correct
9921 offsets for memchr, mempcpy, stpcpy, and stpncpy.
9922 (compute_objsize_r): Adjust offset computation for argument returning
9925 2021-07-20 Martin Sebor <msebor@redhat.com>
9927 PR middle-end/101300
9928 * tree-ssa-uninit.c (check_defs): Handle UBSAN built-ins.
9930 2021-07-20 Jeff Law <jlaw@localhost.localdomain>
9932 * function.c (assign_parm_setup_block): Use adjust_address instead
9933 of change_address to preserve MEM_EXPR and friends.
9935 2021-07-20 Martin Sebor <msebor@redhat.com>
9937 * cfgloop.h (single_likely_exit): Adjust by-value argument to
9939 * cfgloopanal.c (single_likely_exit): Same.
9940 * cgraph.h (struct cgraph_node): Same.
9941 * cgraphclones.c (cgraph_node::create_virtual_clone): Same.
9942 * genautomata.c (merge_states): Same.
9943 * genextract.c (VEC_char_to_string): Same.
9944 * genmatch.c (dt_node::gen_kids_1): Same.
9945 (walk_captures): Adjust by-value argument to by-reference.
9946 * gimple-ssa-store-merging.c (check_no_overlap): Adjust by-value argument
9947 to by-const-reference.
9948 * gimple.c (gimple_build_call_vec): Same.
9949 (gimple_build_call_internal_vec): Same.
9950 (gimple_build_switch): Same.
9951 (sort_case_labels): Same.
9952 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
9954 * gimple.h (gimple_build_call_vec): Adjust by-value argument to
9956 (gimple_build_call_internal_vec): Same.
9957 (gimple_build_switch): Same.
9958 (sort_case_labels): Same.
9959 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
9961 * haifa-sched.c (calc_priorities): Adjust by-value argument to
9963 (sched_init_luids): Same.
9964 (haifa_init_h_i_d): Same.
9965 * ipa-cp.c (ipa_get_indirect_edge_target_1): Same.
9966 (adjust_callers_for_value_intersection): Adjust by-value argument to
9968 (find_more_scalar_values_for_callers_subset): Adjust by-value argument to
9970 (find_more_contexts_for_caller_subset): Same.
9971 (find_aggregate_values_for_callers_subset): Same.
9972 (copy_useful_known_contexts): Same.
9973 * ipa-fnsummary.c (remap_edge_summaries): Same.
9974 (remap_freqcounting_predicate): Same.
9975 * ipa-inline.c (add_new_edges_to_heap): Adjust by-value argument to
9977 * ipa-predicate.c (predicate::remap_after_inlining): Adjust by-value argument
9978 to by-const-reference.
9979 * ipa-predicate.h (predicate::remap_after_inlining): Same.
9980 * ipa-prop.c (ipa_find_agg_cst_for_param): Same.
9981 * ipa-prop.h (ipa_find_agg_cst_for_param): Same.
9982 * ira-build.c (ira_loop_tree_body_rev_postorder): Same.
9983 * read-rtl.c (add_overload_instance): Same.
9984 * rtl.h (native_decode_rtx): Same.
9985 (native_decode_vector_rtx): Same.
9986 * sched-int.h (sched_init_luids): Same.
9987 (haifa_init_h_i_d): Same.
9988 * simplify-rtx.c (native_decode_vector_rtx): Same.
9989 (native_decode_rtx): Same.
9990 * tree-call-cdce.c (gen_shrink_wrap_conditions): Same.
9991 (shrink_wrap_one_built_in_call_with_conds): Same.
9992 (shrink_wrap_conditional_dead_built_in_calls): Same.
9993 * tree-data-ref.c (create_runtime_alias_checks): Same.
9994 (compute_all_dependences): Same.
9995 * tree-data-ref.h (compute_all_dependences): Same.
9996 (create_runtime_alias_checks): Same.
9997 (index_in_loop_nest): Same.
9998 * tree-if-conv.c (mask_exists): Same.
9999 * tree-loop-distribution.c (class loop_distribution): Same.
10000 (loop_distribution::create_rdg_vertices): Same.
10001 (dump_rdg_partitions): Same.
10002 (debug_rdg_partitions): Same.
10003 (partition_contains_all_rw): Same.
10004 (loop_distribution::distribute_loop): Same.
10005 * tree-parloops.c (oacc_entry_exit_ok_1): Same.
10006 (oacc_entry_exit_single_gang): Same.
10007 * tree-ssa-loop-im.c (hoist_memory_references): Same.
10008 (loop_suitable_for_sm): Same.
10009 * tree-ssa-loop-niter.c (bound_index): Same.
10010 * tree-ssa-reassoc.c (update_ops): Same.
10011 (swap_ops_for_binary_stmt): Same.
10012 (rewrite_expr_tree): Same.
10013 (rewrite_expr_tree_parallel): Same.
10014 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Same.
10015 * tree-ssa-sccvn.h (ao_ref_init_from_vn_reference): Same.
10016 * tree-ssa-structalias.c (process_all_all_constraints): Same.
10017 (make_constraints_to): Same.
10018 (handle_lhs_call): Same.
10019 (find_func_aliases_for_builtin_call): Same.
10020 (sort_fieldstack): Same.
10021 (check_for_overlaps): Same.
10022 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Same.
10023 (vect_create_cond_for_unequal_addrs): Same.
10024 (vect_create_cond_for_lower_bounds): Same.
10025 (vect_create_cond_for_alias_checks): Same.
10026 * tree-vect-slp-patterns.c (vect_validate_multiplication): Same.
10027 * tree-vect-slp.c (vect_analyze_slp_instance): Same.
10028 (vect_make_slp_decision): Same.
10029 (vect_slp_bbs): Same.
10030 (duplicate_and_interleave): Same.
10031 (vect_transform_slp_perm_load): Same.
10032 (vect_schedule_slp): Same.
10033 * tree-vectorizer.h (vect_transform_slp_perm_load): Same.
10034 (vect_schedule_slp): Same.
10035 (duplicate_and_interleave): Same.
10036 * tree.c (build_vector_from_ctor): Same.
10037 (build_vector): Same.
10038 (check_vector_cst): Same.
10039 (check_vector_cst_duplicate): Same.
10040 (check_vector_cst_fill): Same.
10041 (check_vector_cst_stepped): Same.
10042 * tree.h (build_vector_from_ctor): Same.
10044 2021-07-20 Jakub Jelinek <jakub@redhat.com>
10047 * config/rs6000/rs6000-protos.h (easy_altivec_constant): Change return
10048 type from bool to int.
10049 * config/rs6000/rs6000.c (vspltis_constant): Fix up handling the
10050 EASY_VECTOR_MSB case if either step or copies is not 1.
10051 (vspltis_shifted): Fix comment typo.
10052 (easy_altivec_constant): Change return type from bool to int, instead
10053 of returning true return byte size of the element mode that should be
10054 used to synthetize the constant.
10055 * config/rs6000/predicates.md (easy_vector_constant_msb): Require
10056 that vspltis_shifted is 0, handle the case where easy_altivec_constant
10057 assumes using different vector mode from CONST_VECTOR's mode.
10058 * config/rs6000/altivec.md (easy_vector_constant_msb splitter): Use
10059 easy_altivec_constant to determine mode in which -1 >> -1 should be
10060 performed, use rs6000_expand_vector_init instead of gen_vec_initv4sisi.
10062 2021-07-20 Richard Biener <rguenther@suse.de>
10065 * dwarf2out.h (dwarf_file_data): Add key member.
10066 * dwarf2out.c (dwarf_file_hasher::equal): Compare key.
10067 (dwarf_file_hasher::hash): Hash key.
10068 (lookup_filename): Remap the filename and store it in the
10069 filename member of dwarf_file_data when creating a new
10071 (file_name_acquire): Do not remap the filename again.
10072 (maybe_emit_file): Likewise.
10074 2021-07-20 Jonathan Wright <jonathan.wright@arm.com>
10076 * config/aarch64/aarch64-simd-builtins.def: Use two variant
10077 generators for all TBL/TBX intrinsics and rename to
10078 consistent forms: qtbl[1234] or qtbx[1234].
10079 * config/aarch64/aarch64-simd.md (aarch64_tbl1<mode>):
10081 (aarch64_qtbl1<mode>): This.
10082 (aarch64_tbx1<mode>): Rename to...
10083 (aarch64_qtbx1<mode>): This.
10084 (aarch64_tbl2v16qi): Delete.
10085 (aarch64_tbl3<mode>): Rename to...
10086 (aarch64_qtbl2<mode>): This.
10087 (aarch64_tbx4<mode>): Rename to...
10088 (aarch64_qtbx2<mode>): This.
10089 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_1): Use
10090 renamed qtbl1 and qtbl2 RTL patterns.
10091 * config/aarch64/arm_neon.h (vqtbl1_p8): Use renamed qtbl1
10093 (vqtbl1_s8): Likewise.
10094 (vqtbl1_u8): Likewise.
10095 (vqtbl1q_p8): Likewise.
10096 (vqtbl1q_s8): Likewise.
10097 (vqtbl1q_u8): Likewise.
10098 (vqtbx1_s8): Use renamed qtbx1 RTL pattern.
10099 (vqtbx1_u8): Likewise.
10100 (vqtbx1_p8): Likewise.
10101 (vqtbx1q_s8): Likewise.
10102 (vqtbx1q_u8): Likewise.
10103 (vqtbx1q_p8): Likewise.
10104 (vtbl1_s8): Use renamed qtbl1 RTL pattern.
10105 (vtbl1_u8): Likewise.
10106 (vtbl1_p8): Likewise.
10107 (vtbl2_s8): Likewise
10108 (vtbl2_u8): Likewise.
10109 (vtbl2_p8): Likewise.
10110 (vtbl3_s8): Use renamed qtbl2 RTL pattern.
10111 (vtbl3_u8): Likewise.
10112 (vtbl3_p8): Likewise.
10113 (vtbl4_s8): Likewise.
10114 (vtbl4_u8): Likewise.
10115 (vtbl4_p8): Likewise.
10116 (vtbx2_s8): Use renamed qtbx2 RTL pattern.
10117 (vtbx2_u8): Likewise.
10118 (vtbx2_p8): Likewise.
10119 (vqtbl2_s8): Use renamed qtbl2 RTL pattern.
10120 (vqtbl2_u8): Likewise.
10121 (vqtbl2_p8): Likewise.
10122 (vqtbl2q_s8): Likewise.
10123 (vqtbl2q_u8): Likewise.
10124 (vqtbl2q_p8): Likewise.
10125 (vqtbx2_s8): Use renamed qtbx2 RTL pattern.
10126 (vqtbx2_u8): Likewise.
10127 (vqtbx2_p8): Likewise.
10128 (vqtbx2q_s8): Likewise.
10129 (vqtbx2q_u8): Likewise.
10130 (vqtbx2q_p8): Likewise.
10131 (vtbx4_s8): Likewise.
10132 (vtbx4_u8): Likewise.
10133 (vtbx4_p8): Likewise.
10135 2021-07-20 Uroš Bizjak <ubizjak@gmail.com>
10138 * config/i386/sync.md (define_peephole2 atomic_storedi_fpu):
10140 (define_peephole2 atomic_loaddi_fpu): Ditto.
10142 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
10144 * config.gcc (riscv*-*-*): Detect which python is available.
10146 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
10148 * config/rs6000/vsx.md (mulhs_<mode>): Rename to...
10149 (smul<mode>3_highpart): ... this.
10150 (mulhu_<mode>): Rename to...
10151 (umul<mode>3_highpart): ... this.
10152 * config/rs6000/rs6000-builtin.def (MULHS_V2DI, MULHS_V4SI,
10153 MULHU_V2DI, MULHU_V4SI): Adjust.
10155 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
10157 PR tree-optimization/100696
10158 * internal-fn.c (first_commutative_argument): Add info for IFN_MULH.
10159 * internal-fn.def (IFN_MULH): New internal function.
10160 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Add support to
10161 recog normal multiply highpart as IFN_MULH.
10162 * config/i386/i386.c (ix86_add_stmt_cost): Adjust for combined
10165 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
10167 * config/elfos.h (CTF_DEBUGGING_INFO): New definition.
10168 (BTF_DEBUGGING_INFO): Likewise.
10169 * doc/tm.texi.in: Document the new macros.
10170 * doc/tm.texi: Regenerated.
10171 * toplev.c: Guard initialization of debug hooks.
10173 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
10175 * flags.h (ctf_debuginfo_p): New function declaration.
10176 * opts.c (ctf_debuginfo_p): New function definition.
10178 2021-07-19 Andrew Stubbs <ams@codesourcery.com>
10181 * config/gcn/gcn-hsa.h (DRIVER_SELF_SPECS): New.
10182 (ASM_SPEC): Set -mattr for xnack and sram-ecc.
10183 * config/gcn/gcn-opts.h (enum sram_ecc_type): New.
10184 * config/gcn/gcn-valu.md: Add a warning comment.
10185 * config/gcn/gcn.c (gcn_option_override): Add "sorry" for -mxnack.
10186 (output_file_start): Add xnack and sram-ecc state to ".amdgcn_target".
10187 * config/gcn/gcn.md: Add a warning comment.
10188 * config/gcn/gcn.opt: Add -mxnack and -msram-ecc.
10189 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): Remove
10191 (EF_AMDGPU_XNACK): New.
10192 (EF_AMDGPU_SRAM_ECC): New.
10194 (copy_early_debug_info): Use elf_flags.
10195 (main): Handle -mxnack and -msram-ecc options.
10196 * doc/invoke.texi: Document -mxnack and -msram-ecc.
10198 2021-07-19 Andrew Pinski <apinski@marvell.com>
10201 * config/aarch64/aarch64.md (csneg3_uxtw_insn): Rename to ...
10202 (*cs<neg_not_cs>3_uxtw_insn4): and extend to NEG_NOT.
10204 2021-07-19 Richard Biener <rguenther@suse.de>
10206 PR tree-optimization/101505
10207 * tree-vect-patterns.c (vect_determine_precisions): Walk
10208 PHIs also for loop vectorization.
10210 2021-07-19 Richard Biener <rguenther@suse.de>
10212 * gimple.h (gimple_expr_type): Remove.
10213 * doc/gimple.texi: Remove gimple_expr_type documentation.
10215 2021-07-19 Richard Biener <rguenther@suse.de>
10217 * tree-ssa-sccvn.c (vn_reference_eq): Handle NULL vr->type.
10218 (ao_ref_init_from_vn_reference): Likewise.
10219 (fully_constant_reference): Likewise.
10220 (vn_reference_lookup_call): Do not set vr->type to random
10222 * tree-ssa-pre.c (compute_avail): Do not try to PRE calls
10224 * tree-vect-generic.c (expand_vector_piecewise): Pass in
10225 whether we expanded parallel.
10226 (expand_vector_parallel): Adjust.
10227 (expand_vector_addition): Likewise.
10228 (expand_vector_comparison): Likewise.
10229 (expand_vector_operation): Likewise.
10230 (expand_vector_scalar_condition): Likewise.
10231 (expand_vector_conversion): Likewise.
10233 2021-07-19 Richard Biener <rguenther@suse.de>
10235 * tree-vrp.c (register_edge_assert_for_2): Use the
10237 (vrp_folder::fold_predicate_in): Likewise.
10238 * vr-values.c (gimple_assign_nonzero_p): Likewise.
10239 (vr_values::extract_range_from_comparison): Likewise.
10240 (vr_values::extract_range_from_ubsan_builtin): Use the
10241 type of the first operand.
10242 (vr_values::extract_range_basic): Push down type
10243 computation, use the appropriate LHS.
10244 (vr_values::extract_range_from_assignment): Use the
10247 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
10250 * common/config/i386/i386-common.c (ix86_handle_option): For
10251 -mgeneral-regs-only, enable the GPR only instructions which are
10252 enabled implicitly by SSE ISAs unless they have been disabled
10255 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
10258 * config/i386/i386.c (ix86_check_avx_upper_stores): Moved before
10259 ix86_avx_u128_mode_needed.
10260 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY if callee
10261 returns AVX register.
10263 2021-07-17 Jan Hubicka <hubicka@ucw.cz>
10265 * tree-ssa-structalias.c (handle_rhs_call): Support EAF_NOT_RETURNED.
10266 (handle_const_call): Liekise
10267 (handle_pure_call): Liekise
10269 2021-07-17 Andrew MacLeod <amacleod@redhat.com>
10271 PR tree-optimization/96542
10272 * range-op.cc (range_operator::wi_fold_in_parts): New.
10273 (range_operator::fold_range): Call wi_fold_in_parts.
10274 (operator_lshift::wi_fold): Fix broken lshift by [0,0].
10275 * range-op.h (wi_fold_in_parts): Add prototype.
10277 2021-07-16 David Malcolm <dmalcolm@redhat.com>
10279 * doc/analyzer.texi: Add __analyzer_dump_state.
10281 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10283 * config/rs6000/rbtree.c: New file.
10284 * config/rs6000/rbtree.h: New file.
10286 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10288 * config/rs6000/rs6000-gen-builtins.c (restriction): New enum.
10289 (typeinfo): Add restr field.
10290 (match_bracketed_pair): New function.
10291 (match_const_restriction): Implement.
10293 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10295 * config/rs6000/rs6000-gen-builtins.c (match_basetype): Implement.
10297 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10299 * config/rs6000/rs6000-gen-builtins.c (void_status): New enum.
10300 (basetype): Likewise.
10301 (typeinfo): Likewise.
10302 (handle_pointer): New function.
10303 (match_basetype): New stub function.
10304 (match_const_restriction): Likewise.
10305 (match_type): New function.
10307 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10309 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace): New
10311 (advance_line): Likewise.
10312 (safe_inc_pos): Likewise.
10313 (match_identifier): Likewise.
10314 (match_integer): Likewise.
10315 (match_to_right_bracket): Likewise.
10317 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10319 * config/rs6000/rs6000-gen-builtins.c (bif_file): New variable.
10320 (ovld_file): Likewise.
10321 (header_file): Likewise.
10322 (init_file): Likewise.
10323 (defines_file): Likewise.
10324 (pgm_path): Likewise.
10325 (bif_path): Likewise.
10326 (ovld_path): Likewise.
10327 (header_path): Likewise.
10328 (init_path): Likewise.
10329 (defines_path): Likewise.
10330 (LINELEN): New macro.
10331 (linebuf): New variable.
10335 (bif_diag): New function.
10336 (ovld_diag): Likewise.
10338 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10340 * config/rs6000/rs6000-builtin-new.def: New.
10341 * config/rs6000/rs6000-overload.def: New.
10343 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10345 * config/rs6000/rs6000-gen-builtins.c: New.
10347 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10349 * Makefile.in (EXTRA_GTYPE_DEPS): New variable.
10350 (s-gtype): Depend on EXTRA_GTYPE_DEPS.
10351 * gengtype-state.c (state_writer::write_state_file_list): Add a
10352 parameter to the fileslist expression for the number of build
10354 (read_state_files_list): Detect build headers and strip the
10355 initial "./" or ".\" from their names.
10356 * gengtype.c (build_headers): New global variable.
10357 (num_build_headers): Likewise.
10358 (open_base_files): Emit #include for each build header.
10359 (main): Detect and count build headers.
10360 * gengtype.h (build_headers): New extern variable.
10361 (num_build_headers): Likewise.
10363 2021-07-16 Richard Biener <rguenther@suse.de>
10365 * gimple-ssa-store-merging.c (verify_symbolic_number_p): Use
10366 the type of the LHS.
10367 (find_bswap_or_nop_1): Likewise.
10368 (find_bswap_or_nop): Likewise.
10369 * tree-vectorizer.h (vect_get_smallest_scalar_type): Adjust
10371 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
10372 Remove unused parameters, pass in the scalar type. Fix
10373 internal store function handling.
10374 * tree-vect-stmts.c (vect_analyze_stmt): Remove assert.
10375 (vect_get_vector_types_for_stmt): Move down check for
10376 existing vector stmt after we've determined a scalar type.
10377 Pass down the used scalar type to vect_get_smallest_scalar_type.
10378 * tree-vect-generic.c (expand_vector_condition): Use
10379 the type of the LHS.
10380 (expand_vector_scalar_condition): Likewise.
10381 (expand_vector_operations_1): Likewise.
10382 * tree-vect-patterns.c (vect_widened_op_tree): Likewise.
10383 (vect_recog_dot_prod_pattern): Likewise.
10384 (vect_recog_sad_pattern): Likewise.
10385 (vect_recog_widen_op_pattern): Likewise.
10386 (vect_recog_widen_sum_pattern): Likewise.
10387 (vect_recog_mixed_size_cond_pattern): Likewise.
10389 2021-07-16 Jan Hubicka <hubicka@ucw.cz>
10391 * ipa-modref.c (struct escape_entry): Use eaf_fleags_t.
10392 (dump_eaf_flags): Dump EAF_NOT_RETURNED
10393 (eaf_flags_useful_p): Use eaf_fleags_t; handle const functions
10394 and EAF_NOT_RETURNED.
10395 (modref_summary::useful_p): Likewise.
10396 (modref_summary_lto::useful_p): Likewise.
10397 (struct) modref_summary_lto: Use eaf_fleags_t.
10398 (deref_flags): Handle EAF_NOT_RETURNED.
10399 (struct escape_point): Use min_flags.
10400 (modref_lattice::init): Add EAF_NOT_RETURNED.
10401 (merge_call_lhs_flags): Ignore EAF_NOT_RETURNED functions
10402 (analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
10404 (analyze_parms): Also analyze const functions; update conition on
10406 (modref_write): Update streaming.
10407 (read_section): Update streaming.
10408 (remap_arg_flags): Use eaf_flags_t.
10409 (modref_merge_call_site_flags): Hanlde EAF_NOT_RETURNED.
10410 * ipa-modref.h: (eaf_flags_t): New typedef.
10411 (struct modref_summary): Use eaf_flags_t.
10412 * tree-core.h (EAF_NOT_RETURNED): New constant.
10414 2021-07-16 Richard Biener <rguenther@suse.de>
10416 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Use
10417 the type of the LHS.
10418 (gimple_assign_nonnegative_warnv_p): Likewise.
10419 (gimple_call_nonnegative_warnv_p): Likewise. Return false
10420 if the call has no LHS.
10421 * gimple.c (gimple_could_trap_p_1): Use the type of the LHS.
10422 * tree-eh.c (stmt_could_throw_1_p): Likewise.
10423 * tree-inline.c (insert_init_stmt): Likewise.
10424 * tree-ssa-loop-niter.c (get_val_for): Likewise.
10425 * tree-outof-ssa.c (ssa_is_replaceable_p): Use the type of
10427 * tree-ssa-sccvn.c (init_vn_nary_op_from_stmt): Take a
10428 gassign *. Use the type of the lhs.
10429 (vn_nary_op_lookup_stmt): Adjust.
10430 (vn_nary_op_insert_stmt): Likewise.
10432 2021-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
10434 * config/s390/predicates.md (bras_sym_operand): Accept all
10435 functions in 64-bit mode, use UNSPEC_PLT31.
10436 (larl_operand): Use UNSPEC_PLT31.
10437 * config/s390/s390.c (s390_loadrelative_operand_p): Likewise.
10438 (legitimize_pic_address): Likewise.
10439 (s390_emit_tls_call_insn): Mark __tls_get_offset as function,
10441 (s390_delegitimize_address): Use UNSPEC_PLT31.
10442 (s390_output_addr_const_extra): Likewise.
10443 (print_operand): Add @PLT to TLS calls, handle %K.
10444 (s390_function_profiler): Mark __fentry__/_mcount as function,
10445 use %K, use UNSPEC_PLT31.
10446 (s390_output_mi_thunk): Use only UNSPEC_GOT, use %K.
10447 (s390_emit_call): Use UNSPEC_PLT31.
10448 (s390_emit_tpf_eh_return): Mark __tpf_eh_return as function.
10449 * config/s390/s390.md (UNSPEC_PLT31): Rename from UNSPEC_PLT.
10450 (*movdi_64): Use %K.
10451 (reload_base_64): Likewise.
10452 (*sibcall_brc): Likewise.
10453 (*sibcall_brcl): Likewise.
10454 (*sibcall_value_brc): Likewise.
10455 (*sibcall_value_brcl): Likewise.
10457 (*brasl): Likewise.
10458 (*bras_r): Likewise.
10459 (*brasl_r): Likewise.
10460 (*bras_tls): Likewise.
10461 (*brasl_tls): Likewise.
10462 (main_base_64): Likewise.
10463 (reload_base_64): Likewise.
10464 (@split_stack_call<mode>): Likewise.
10466 2021-07-16 Richard Biener <rguenther@suse.de>
10468 PR tree-optimization/101467
10469 * tree-vect-stmts.c (vect_gen_while): Properly guard
10470 make_temp_ssa_name usage.
10472 2021-07-16 Cooper Qu <cooper.qu@linux.alibaba.com>
10474 * config.gcc: Don't use forked print-sysroot-suffix.sh and
10475 t-sysroot-suffix for C-SKY.
10476 * config/csky/print-sysroot-suffix.sh: Delete.
10477 * config/csky/t-csky-linux: Delete.
10478 * config/csky/t-sysroot-suffix: Define MULTILIB_DIRNAMES
10479 instead of CSKY_MULTILIB_DIRNAMES.
10481 2021-07-16 Richard Biener <rguenther@suse.de>
10483 * tree-vect-loop.c (vect_transform_cycle_phi): Correct sign
10484 conversion issues with the partial reduction of the reused
10485 vector accumulator.
10487 2021-07-16 Richard Biener <rguenther@suse.de>
10489 * config/i386/i386-options.c (ix86_option_override_internal): Set
10490 param_vect_partial_vector_usage to zero if not set.
10492 2021-07-15 Uroš Bizjak <ubizjak@gmail.com>
10495 * config/i386/i386.h (VALID_SSE_REG_MODE): Add TDmode.
10496 (VALID_INT_MODE_P): Add SDmode and DDmode.
10497 Add TDmode for TARGET_64BIT.
10498 (VALID_DFP_MODE_P): Remove.
10499 * config/i386/i386.c (ix86_hard_regno_mode_ok):
10500 Do not use VALID_DFP_MODE_P.
10502 2021-07-15 Andrew MacLeod <amacleod@redhat.com>
10504 * gimple-range-fold.cc (adjust_pointer_diff_expr): Use
10506 (fold_using_range::fold_stmt): Ditto.
10507 (fold_using_range::range_of_range_op): Ditto.
10508 (fold_using_range::range_of_phi): Ditto.
10509 (fold_using_range::range_of_call): Ditto.
10510 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
10511 (fold_using_range::range_of_builtin_call): Ditto.
10512 (fold_using_range::range_of_cond_expr): Ditto.
10513 * gimple-range-fold.h (gimple_range_type): New.
10515 2021-07-15 Martin Sebor <msebor@redhat.com>
10517 PR middle-end/97027
10518 * tree-ssa-strlen.c (handle_assign): New function.
10519 (maybe_warn_overflow): Add argument.
10520 (nonzero_bytes_for_type): New function.
10521 (count_nonzero_bytes): Handle more tree types. Call
10522 nonzero_bytes_for_tye.
10523 (count_nonzero_bytes): Handle types.
10524 (handle_store): Handle stores from function calls.
10525 (strlen_check_and_optimize_call): Move code to handle_assign. Call
10526 it for assignments from function calls.
10528 2021-07-15 David Malcolm <dmalcolm@redhat.com>
10533 * doc/invoke.texi: Add -Wanalyzer-use-of-uninitialized-value.
10535 2021-07-15 David Malcolm <dmalcolm@redhat.com>
10537 * doc/invoke.texi (-fdump-analyzer-exploded-paths): New.
10539 2021-07-15 Martin Sebor <msebor@redhat.com>
10543 * fold-const.c (operand_compare::operand_equal_p): Handle OEP_DECL_NAME.
10544 (operand_compare::verify_hash_value): Same.
10545 * tree-core.h (OEP_DECL_NAME): New.
10547 2021-07-15 Martin Jambor <mjambor@suse.cz>
10549 * profile-count.h (profile_count::value): Change the return type to
10551 * gimple-pretty-print.c (dump_gimple_bb_header): Adjust print
10553 * tree-cfg.c (dump_function_to_file): Likewise.
10555 2021-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
10558 * config/rs6000/rs6000-p8swap.c (has_part_mult): New.
10559 (rs6000_analyze_swaps): Insns containing a subreg of a mult are
10562 2021-07-15 Richard Biener <rguenther@suse.de>
10564 * tree-vectorizer.h (vect_gen_while): Match up with
10565 vect_gen_while_not.
10566 * tree-vect-stmts.c (vect_gen_while): Adjust API to that
10567 of vect_gen_while_not.
10568 (vect_gen_while_not): Adjust.
10569 * tree-vect-loop-manip.c (vect_set_loop_controls_directly): Likewise.
10571 2021-07-15 Aldy Hernandez <aldyh@redhat.com>
10573 * gimple-range-cache.cc (non_null_ref::adjust_range): New.
10574 (ranger_cache::range_of_def): Call adjust_range.
10575 (ranger_cache::entry_range): Same.
10576 * gimple-range-cache.h (non_null_ref::adjust_range): New.
10577 * gimple-range.cc (gimple_ranger::range_of_expr): Call
10579 (gimple_ranger::range_on_entry): Same.
10581 2021-07-15 Tamar Christina <tamar.christina@arm.com>
10584 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10586 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
10588 2021-07-15 Tamar Christina <tamar.christina@arm.com>
10591 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10593 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
10594 (sdot_prod, udot_prod): ...These.
10595 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
10596 (aarch64_<sur>dot<vsi2qi>): Rename to...
10597 (<sur>dot_prod<vsi2qi>): ...This.
10598 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
10601 2021-07-15 Jakub Jelinek <jakub@redhat.com>
10603 PR middle-end/101437
10604 * gimplify.c (gimplify_expr): Throw away volatile reads from empty
10605 types even if they have non-BLKmode TYPE_MODE.
10607 2021-07-15 Richard Biener <rguenther@suse.de>
10610 * gcc.c (process_command): Process -gtoggle like process_options
10611 would after parsing options.
10613 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
10615 * cfgexpand.c (expand_asm_loc): Adjust.
10616 (expand_asm_stmt): Likewise.
10617 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Likewise.
10618 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
10619 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
10620 * config/avr/avr.c (avr_md_asm_adjust): Likewise.
10621 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
10622 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
10623 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
10624 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
10625 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
10626 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
10627 * config/s390/s390.c (s390_md_asm_adjust): Likewise.
10628 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
10629 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
10630 * doc/tm.texi: Regenerate.
10631 * target.def: Add location argument to md_asm_adjust.
10633 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
10635 * tree-diagnostic.c (diagnostic_report_current_function): Use the
10636 diagnostic's location, not input_location.
10638 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
10640 * cfgexpand.c (tree_conflicts_with_clobbers_p): Pass location to
10642 (expand_asm_stmt): Likewise.
10644 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
10646 * config/rs6000/rs6000.c (adjacent_mem_locations): Return the lower
10647 addressed memory rtx, if any.
10648 (rs6000_split_multireg_move): Fix code formatting.
10649 Handle MMA build built-ins with operands in adjacent memory locations.
10651 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
10653 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Move to later
10656 2021-07-14 Jason Merrill <jason@redhat.com>
10658 * sel-sched-ir.h (get_all_loop_exits): Use auto_vec.
10660 2021-07-14 Jason Merrill <jason@redhat.com>
10662 * doc/invoke.texi: -fdelete-dead-exceptions is on by default for
10665 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10667 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
10668 Remove erroneous line.
10670 2021-07-14 Andrew MacLeod <amacleod@redhat.com>
10672 * params.opt (param_evrp_mode): Change default.
10674 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10676 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
10677 (sdot_prod, udot_prod): ...These.
10678 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
10679 (aarch64_<sur>dot<vsi2qi>): Rename to...
10680 (<sur>dot_prod<vsi2qi>): ...This.
10681 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
10684 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10686 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
10688 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10690 * doc/sourcebuild.texi (arm_v8_2a_i8mm_neon_hw): Document.
10692 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10694 * config/arm/neon.md (usdot_prod<vsi2qi>): New.
10696 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10698 * config/aarch64/aarch64-simd.md (aarch64_usdot<vsi2qi>): Rename to...
10699 (usdot_prod<vsi2qi>): ... This.
10700 * config/aarch64/aarch64-simd-builtins.def (usdot): Rename to...
10701 (usdot_prod): ...This.
10702 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Likewise.
10703 * config/aarch64/aarch64-sve.md (@aarch64_<sur>dot_prod<vsi2qi>):
10705 (@<sur>dot_prod<vsi2qi>): ...This.
10706 * config/aarch64/aarch64-sve-builtins-base.cc
10707 (svusdot_impl::expand): Use it.
10709 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10711 * optabs.def (usdot_prod_optab): New.
10712 * doc/md.texi: Document it and clarify other dot prod optabs.
10713 * optabs-tree.h (enum optab_subtype): Add optab_vector_mixed_sign.
10714 * optabs-tree.c (optab_for_tree_code): Support usdot_prod_optab.
10715 * optabs.c (expand_widen_pattern_expr): Likewise.
10716 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
10717 * tree-vect-loop.c (vectorizable_reduction): Query dot-product kind.
10718 * tree-vect-patterns.c (vect_supportable_direct_optab_p): Take optional
10720 (vect_widened_op_tree): Optionally ignore
10722 (vect_recog_dot_prod_pattern): Support usdot_prod_optab.
10724 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
10727 * config/i386/driver-i386.c (host_detect_local_cpu): Check
10728 "arch [32|64]" and "tune [32|64]" for 32-bit and 64-bit codegen.
10729 Enable UINTR only for 64-bit codegen.
10730 * config/i386/i386-options.c
10731 (ix86_option_override_internal::DEF_PTA): Skip PTA_UINTR if not
10733 * config/i386/i386.h (ARCH_ARG): New.
10734 (CC1_CPU_SPEC): Pass "[arch|tune] 32" for 32-bit codegen and
10735 "[arch|tune] 64" for 64-bit codegen.
10737 2021-07-14 Richard Biener <rguenther@suse.de>
10739 PR tree-optimization/101445
10740 * tree-vect-stmts.c (vectorizable_load): Do the gap adjustment
10741 of the IV in the correct direction for negative stride
10744 2021-07-14 Jakub Jelinek <jakub@redhat.com>
10747 * godump.c (godump_str_hash): New type.
10748 (godump_container::pot_dummy_types): Use string_hash instead of
10749 ptr_hash in the hash_set.
10751 2021-07-14 Richard Biener <rguenther@suse.de>
10753 * tree-vect-loop.c (vect_find_reusable_accumulator): Handle
10754 vector types where the old vector type has a multiple of
10755 the new vector type elements.
10756 (vect_create_partial_epilog): New function, split out from...
10757 (vect_create_epilog_for_reduction): ... here.
10758 (vect_transform_cycle_phi): Reduce the re-used accumulator
10759 to the new vector type.
10761 2021-07-14 Alexandre Oliva <oliva@adacore.com>
10763 * tree-ssa-alias.c (attr_fnspec::verify): Fix index in
10764 non-'t'-sized arg check.
10766 2021-07-14 Alexandre Oliva <oliva@adacore.com>
10768 * tree-cfg.c (cleanup_dead_labels_eh): Update
10769 post_landing_pad label upon change of landing pad block's
10771 (cleanup_dead_labels): Check that a removed label is not that
10774 2021-07-13 Jonathan Wright <jonathan.wright@arm.com>
10776 * combine.c (combine_simplify_rtx): Add vec_select -> subreg
10778 * config/aarch64/aarch64.md (*zero_extend<SHORT:mode><GPI:mode>2_aarch64):
10779 Add Neon to general purpose register case for zero-extend
10781 * config/arm/vfp.md (*arm_movsi_vfp): Remove "*" from *t -> r
10782 case to prevent some cases opting to go through memory.
10783 * cse.c (fold_rtx): Add vec_select -> subreg simplification.
10784 * rtl.c (rtvec_series_p): Define predicate to determine
10785 whether a vector contains a linear series of integers.
10786 * rtl.h (rtvec_series_p): Define.
10787 * rtlanal.c (vec_series_lowpart_p): Define predicate to
10788 determine if a vector selection is equivalent to the low part
10790 * rtlanal.h (vec_series_lowpart_p): Define.
10791 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
10792 Add vec_select -> subreg simplification.
10794 2021-07-13 Paul A. Clarke <pc@us.ibm.com>
10796 * config/rs6000/smmintrin.h (_mm_testz_si128, _mm_testc_si128,
10797 _mm_testnzc_si128, _mm_test_all_ones, _mm_test_all_zeros,
10798 _mm_test_mix_ones_zeros): New.
10800 2021-07-13 Roger Sayle <roger@nextmovesoftware.com>
10801 Richard Biener <rguenther@suse.de>
10803 * gimple.c (gimple_could_trap_p_1): Make S argument a
10804 "const gimple*". Preserve constness in call to
10805 gimple_asm_volatile_p.
10806 (gimple_could_trap_p): Make S argument a "const gimple*".
10807 * gimple.h (gimple_could_trap_p_1, gimple_could_trap_p):
10808 Update function prototypes.
10810 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10812 * tree-vectorizer.h (vect_reusable_accumulator): New structure.
10813 (_loop_vec_info::main_loop_edge): New field.
10814 (_loop_vec_info::skip_main_loop_edge): Likewise.
10815 (_loop_vec_info::skip_this_loop_edge): Likewise.
10816 (_loop_vec_info::reusable_accumulators): Likewise.
10817 (_stmt_vec_info::reduc_scalar_results): Likewise.
10818 (_stmt_vec_info::reused_accumulator): Likewise.
10819 (vect_get_main_loop_result): Declare.
10820 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize
10821 reduc_scalar_inputs.
10822 (vec_info::free_stmt_vec_info): Free reduc_scalar_inputs.
10823 * tree-vect-loop-manip.c (vect_get_main_loop_result): New function.
10824 (vect_do_peeling): Fill an epilogue loop's main_loop_edge,
10825 skip_main_loop_edge and skip_this_loop_edge fields.
10826 * tree-vect-loop.c (INCLUDE_ALGORITHM): Define.
10827 (vect_emit_reduction_init_stmts): New function.
10828 (get_initial_def_for_reduction): Use it.
10829 (get_initial_defs_for_reduction): Likewise. Change the vinfo
10830 parameter to a loop_vec_info.
10831 (vect_create_epilog_for_reduction): Store the scalar results
10832 in the reduc_info. If an epilogue loop is reusing an accumulator
10833 from the main loop, and if the epilogue loop can also be skipped,
10834 try to place the reduction code in the join block. Record
10835 accumulators that could potentially be reused by epilogue loops.
10836 (vect_transform_cycle_phi): When vectorizing epilogue loops,
10837 try to reuse accumulators from the main loop. Record the initial
10838 value in reduc_info for non-SLP reductions too.
10840 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10842 * tree-vect-loop.c (get_initial_def_for_reduction): Remove
10843 adjustment handling. Take the neutral value as an argument,
10844 in place of the code argument.
10845 (vect_transform_cycle_phi): Update accordingly. Handle the
10846 initial values of cond reductions separately from code reductions.
10847 Choose the adjustment here rather than in
10848 get_initial_def_for_reduction. Sink the splat of vec_initial_def.
10850 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10852 * tree-vect-loop.c (neutral_op_for_slp_reduction): Replace with...
10853 (neutral_op_for_reduction): ...this, providing a more general
10855 (vect_create_epilog_for_reduction): Update accordingly.
10856 (vectorizable_reduction): Likewise.
10857 (vect_transform_cycle_phi): Likewise.
10859 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10861 * tree-vect-loop.c (get_initial_def_for_reduction): Take the
10862 reduc_info instead of the original stmt_vec_info.
10863 (vect_transform_cycle_phi): Update accordingly.
10865 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10867 * tree-vect-loop.c (get_initial_defs_for_reduction): Take the
10868 reduc_info as an additional parameter.
10869 (vect_transform_cycle_phi): Update accordingly.
10871 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10873 * tree-vectorizer.h: Include tree-ssa-operands.h.
10874 (vect_phi_initial_value): New function.
10875 * tree-vect-loop.c (neutral_op_for_slp_reduction): Use it.
10876 (get_initial_defs_for_reduction, info_for_reduction): Likewise.
10877 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
10878 (vect_transform_cycle_phi, vectorizable_induction): Likewise.
10880 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10882 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
10883 the phi results to vectype after creating them. Remove later
10884 conversion code that thus becomes redundant.
10886 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10888 * tree-vect-loop.c (vect_create_epilog_for_reduction): Replace
10889 the new_phis vector with a reduc_inputs vector. Combine handling
10890 of reduction chains and ncopies > 1.
10892 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10894 * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
10895 scalar_results to group_size elements after reducing down from
10896 N*group_size elements. Construct an array_slice of the live-out
10897 stmts and assert that there is one stmt per scalar result.
10899 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10901 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
10902 nested_in_vect_loop and use double_reduc everywhere. Remove dead
10903 assignment to "loop".
10905 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10907 * internal-fn.c (vectorized_internal_fn_supported_p): Handle
10908 vector types first. For scalar types, consider both the preferred
10909 vector mode and the alternative vector modes.
10910 * optabs-query.c (can_vec_mask_load_store_p): Use the same
10911 structure as above, in particular using related_vector_mode
10912 for modes provided by autovectorize_vector_modes.
10914 2021-07-13 Jakub Jelinek <jakub@redhat.com>
10915 Richard Biener <rguenther@suse.de>
10917 PR tree-optimization/101419
10918 * tree-pass.h (PROP_objsz): Define.
10919 (make_pass_early_object_sizes): Declare.
10920 * passes.def (pass_all_early_optimizations): Rename pass_object_sizes
10921 there to pass_early_object_sizes, drop parameter.
10922 (pass_all_optimizations): Move pass_object_sizes right after pass_ccp,
10923 drop parameter, move pass_post_ipa_warn right after that.
10924 * tree-object-size.c (pass_object_sizes::execute): Rename to...
10925 (object_sizes_execute): ... this. Add insert_min_max_p argument.
10926 (pass_data_object_sizes): Move after object_sizes_execute.
10927 (pass_object_sizes): Likewise. In execute method call
10928 object_sizes_execute, drop set_pass_param method and insert_min_max_p
10929 non-static data member and its initializer in the ctor.
10930 (pass_data_early_object_sizes, pass_early_object_sizes,
10931 make_pass_early_object_sizes): New.
10932 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Use
10933 (cfun->curr_properties & PROP_objsz) instead of cfun->after_inlining.
10935 2021-07-13 Kito Cheng <kito.cheng@sifive.com>
10938 * config/riscv/constraints.md ("S"): Update description and remove
10940 * doc/md.texi (Machine Constraints): Document the 'S' constraints
10943 2021-07-13 Richard Biener <rguenther@suse.de>
10946 2021-07-12 Richard Biener <rguenther@suse.de>
10948 * tree-vect-slp.c (vect_slp_region): Show the number of
10949 SLP graph entries in the optimization message.
10951 2021-07-13 Michael Meissner <meissner@linux.ibm.com>
10953 * config/rs6000/altivec.md (xxspltiw_v4sf): Change local variable
10955 * config/rs6000/rs6000-protos.h (rs6000_const_f32_to_i32): Change
10956 return type to long.
10957 * config/rs6000/rs6000.c (rs6000_const_f32_to_i32): Change return
10960 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
10962 * gimple-range-fold.cc (fold_using_range::range_of_builtin_ubsan_call):
10963 Query relation between the 2 operands and use it.
10965 2021-07-12 Sergei Trofimovich <siarheit@google.com>
10967 * doc/cfg.texi: Fix s/ei_safe_safe/ei_safe_edge/ typo.
10969 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
10972 * config/i386/predicates.md (vec_setm_sse41_operand):
10973 Rename from vec_setm_operand.
10974 (vec_setm_avx2_operand): New predicate.
10975 * config/i386/sse.md (vec_set<V_128:mode>): Use V_128 mode iterator.
10976 Use vec_setm_sse41_operand as operand 2 predicate.
10977 (vec_set<V_256_512:mode): New expander.
10978 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_sse41_operand
10979 as operand 2 predicate.
10981 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
10983 PR tree-optimization/101335
10984 * range-op.cc (operator_cast::lhs_op1_relation): Delete.
10986 2021-07-12 Andrew Pinski <apinski@marvell.com>
10988 * tree-ssa-phiopt.c (match_simplify_replacement): Move
10989 insert of the sequence before the movement of the
10990 statement. Check if to see if the statement is used
10991 outside of the original phi to see if we should move it.
10993 2021-07-12 Richard Biener <rguenther@suse.de>
10995 * dump-context.h (debug_dump_context::debug_dump_context):
10996 Add FILE * parameter defaulted to stderr.
10997 * dumpfile.c (debug_dump_context::debug_dump_context): Adjust.
10998 * tree-vect-slp.c (dot_slp_tree): New functions.
11000 2021-07-12 Richard Biener <rguenther@suse.de>
11002 PR tree-optimization/101373
11003 * tree-ssa-pre.c (prune_clobbered_mems): Also prune trapping
11004 references when the BB may not return.
11005 (compute_avail): Pass in the function we're working on and
11006 replace cfun references with it. Externally throwing
11007 const calls also possibly terminate the function.
11008 (pass_pre::execute): Pass down the function we're working on.
11009 * gcse.c (compute_hash_table_work): Externally throwing
11010 const/pure calls also need record_last_mem_set_info.
11011 * postreload-gcse.c (record_opr_changes): Looping or externally
11012 throwing const/pure calls also need record_last_mem_set_info.
11014 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
11016 * recog.c (memory_address_addr_space_p): Change the type to bool.
11017 Return true/false instead of 1/0.
11018 (offsettable_memref_p): Ditto.
11019 (offsettable_nonstrict_memref_p): Ditto.
11020 (offsettable_address_addr_space_p): Ditto.
11021 Change the type of addressp indirect function to bool.
11022 * recog.h (memory_address_addr_space_p): Change the type to bool.
11023 (strict_memory_address_addr_space_p): Ditto.
11024 (offsettable_memref_p): Ditto.
11025 (offsettable_nonstrict_memref_p): Ditto.
11026 (offsettable_address_addr_space_p): Ditto.
11027 * reload.c (maybe_memory_address_addr_space_p): Ditto.
11028 (strict_memory_address_addr_space_p): Change the type to bool.
11029 Return true/false instead of 1/0.
11030 (maybe_memory_address_addr_space_p): Change the type to bool.
11032 2021-07-12 Richard Biener <rguenther@suse.de>
11034 * tree-vect-slp.c (vect_slp_region): Show the number of
11035 SLP graph entries in the optimization message.
11037 2021-07-12 Richard Biener <rguenther@suse.de>
11039 PR tree-optimization/101394
11040 * tree-ssa-pre.c (do_pre_regular_insertion): Avoid inserting
11041 copies from abnormals for a full redundancy.
11043 2021-07-12 Richard Biener <rguenther@suse.de>
11045 PR middle-end/101423
11046 * gimple.c (gimple_could_trap_p_1): Internal function calls
11048 * tree-eh.c (tree_could_trap_p): Likewise.
11050 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
11053 * config/arm/arm_neon.h (vmul_n_u32): Replace call to builtin with
11055 (vmulq_n_u32): Likewise.
11056 (vmul_n_f32): Gate __a * __b on __FAST_MATH__.
11057 (vmulq_n_f32): Likewise.
11058 (vmul_n_f16): Likewise.
11059 (vmulq_n_f16): Likewise.
11061 2021-07-12 Martin Liska <mliska@suse.cz>
11063 PR sanitizer/101425
11064 * gcc.c (check_offload_target_name): Call
11065 candidates_list_and_hint only if we have a candidate.
11067 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
11070 * config/arm/neon.md (vec_init): Move to ...
11071 * config/arm/vec-common.md (vec_init): ... here.
11072 Change the pattern's mode to VDQX and gate it on VALID_MVE_MODE.
11074 2021-07-12 Roger Sayle <roger@nextmovesoftware.com>
11076 PR tree-optimization/101403
11077 * match.pd ((T)bswap(X)>>C): Correctly handle cases where
11078 signedness of the shift is not the same as the signedness of
11079 the type extension.
11081 2021-07-09 Roger Sayle <roger@nextmovesoftware.com>
11082 Uroš Bizjak <ubizjak@gmail.com>
11084 * config/i386/i386.md (*divmodsi4_const): Optimize SImode
11085 divmod of a constant numerator with new define_insn_and_split.
11087 2021-07-09 Iain Sandoe <iain@sandoe.co.uk>
11090 * config/i386/i386-expand.c (ix86_expand_call): If a call is
11091 to a non-local-binding, or local but to a public symbol, then
11092 assume that it might be indirected via the lazy symbol binder.
11093 Mark R10 and R10 as clobbered in that case.
11095 2021-07-09 Eric Botcazou <ebotcazou@adacore.com>
11098 * gcc.c (ASM_DEBUG_DWARF_OPTION): Set again to --gdwarf2 in
11099 the case where HAVE_AS_WORKING_DWARF_N_FLAG is not defined
11100 and HAVE_LD_BROKEN_PE_DWARF5 is defined.
11102 2021-07-09 Uroš Bizjak <ubizjak@gmail.com>
11104 * config/i386/i386.md (*udivmodsi4_pow2_zext_1): Limit the
11105 log2 range of operands[3] to [1,31].
11106 (*udivmodsi4_pow2_zext_2): Ditto. Correct insn RTX pattern.
11108 2021-07-09 Sergei Trofimovich <siarheit@google.com>
11110 * doc/md.texi: Don't split @smallexample in multiple @groups.
11112 2021-07-09 Sergei Trofimovich <siarheit@google.com>
11114 * doc/md.texi: Add missing 'see' word.
11116 2021-07-09 Andrew Pinski <apinski@marvell.com>
11118 * tree-ssa-phiopt.c (phiopt_early_allow): Change arguments
11119 to take sequence and gimple_match_op. Accept the case where
11120 op is a SSA_NAME and one statement in the sequence.
11121 Also allow constants.
11122 (gimple_simplify_phiopt): Always pass a sequence to resimplify.
11123 Update call to phiopt_early_allow. Discard the sequence if not
11126 2021-07-09 Xi Ruoyao <xry111@mengyan1223.wang>
11131 * config/mips/mips.c (mips_const_insns): Use MSA_SUPPORTED_MODE_P
11132 instead of ISA_HAS_MSA.
11133 (mips_expand_vec_unpack): Likewise.
11134 (mips_expand_vector_init): Likewise.
11136 2021-07-09 Kewen Lin <linkw@linux.ibm.com>
11138 * config/rs6000/vsx.md (mods_<mode>): Rename to...
11139 (mod<mode>3): ... this.
11140 (modu_<mode>): Rename to...
11141 (umod<mode>3): ... this.
11142 * config/rs6000/rs6000-builtin.def (MODS_V2DI, MODS_V4SI, MODU_V2DI,
11143 MODU_V4SI): Adjust.
11145 2021-07-08 Jeff Law <jeffreyalaw@gmail.com>
11147 * config/h8300/shiftrotate.md (variable shifts): Expose condition
11148 code handling for the test before the loop.
11150 2021-07-08 Martin Jambor <mjambor@suse.cz>
11153 * ipa-sra.c (class isra_call_summary): New member
11154 m_before_any_store, initialize it in the constructor.
11155 (isra_call_summary::dump): Dump the new field.
11156 (ipa_sra_call_summaries::duplicate): Copy it.
11157 (process_scan_results): Set it.
11158 (isra_write_edge_summary): Stream it.
11159 (isra_read_edge_summary): Likewise.
11160 (param_splitting_across_edge): Only override
11161 safe_to_import_accesses if m_before_any_store is set.
11163 2021-07-08 Martin Sebor <msebor@redhat.com>
11165 PR bootstrap/101374
11166 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref):
11167 Use Object Size Type 0 instead of 1.
11169 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
11171 * tree-vect-loop.c (vectorizable_reduction): Remove always-true
11174 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
11176 * match.pd: Simplify an extend-operate-truncate sequence involving
11179 2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
11180 Richard Biener <rguenther@suse.de>
11182 PR tree-optimization/40210
11183 * match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
11184 (x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
11185 when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
11187 2021-07-08 Uroš Bizjak <ubizjak@gmail.com>
11190 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
11192 * config/i386/mmx.md (V_32): New mode iterator.
11193 (mov<V_32:mode>): Use V_32 mode iterator.
11194 (*mov<V_32:mode>_internal): Ditto.
11195 (*push<V_32:mode>2_rex64): Ditto.
11196 (*push<V_32:mode>2): Ditto.
11197 (movmisalign<V_32:mode>): Ditto.
11198 (mmx_<any_shiftrt:insn>v1si3): New insn pattern.
11199 (sse4_1_<any_extend:code>v2qiv2hi2): Ditto.
11200 (vec_unpacks_lo_v4qi): New expander.
11201 (vec_unpacks_hi_v4qi): Ditto.
11202 (vec_unpacku_lo_v4qi): Ditto.
11203 (vec_unpacku_hi_v4qi): Ditto.
11204 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add V1SImode.
11205 (VALID_INT_MODE_P): Ditto.
11207 2021-07-08 Michael Meissner <meissner@linux.ibm.com>
11210 * config/rs6000/rs6000.md (udivti3): New insn.
11211 (divti3): New insn.
11212 (umodti3): New insn.
11213 (modti3): New insn.
11215 2021-07-07 Martin Sebor <msebor@redhat.com>
11217 PR tree-optimization/100137
11218 PR tree-optimization/99121
11219 PR tree-optimization/97027
11220 * builtins.c (access_ref::access_ref): Also set offmax.
11221 (access_ref::offset_in_range): Define new function.
11222 (access_ref::add_offset): Set offmax.
11223 (access_ref::inform_access): Handle access_none.
11224 (handle_mem_ref): Clear ostype.
11225 (compute_objsize_r): Handle ASSERT_EXPR.
11226 * builtins.h (struct access_ref): Add offmax member.
11227 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Use
11228 compute_objsize() and simplify.
11230 2021-07-07 Peter Bergner <bergner@linux.ibm.com>
11232 * config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
11233 and VSX_BUILTIN_STXVP.
11235 2021-07-07 Martin Sebor <msebor@redhat.com>
11238 * config/aarch64/aarch64.c (aarch64_simd_lane_bounds): Remove
11239 a stray %K from error_at() missed in r12-2088.
11241 2021-07-07 Richard Biener <rguenther@suse.de>
11243 PR tree-optimization/99728
11244 * tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
11246 (mem_refs_may_alias_p): Add assert we handled aggregate
11248 (sm_seq_valid_bb): Give up when running into aggregate copies.
11249 (ref_indep_loop_p): Handle aggregate copies as never
11250 being invariant themselves but allow other refs to be
11251 disambiguated against them.
11252 (can_sm_ref_p): Do not try to apply store-motion to aggregate
11255 2021-07-06 Indu Bhagat <indu.bhagat@oracle.com>
11258 * dwarf2ctf.c (ctf_get_AT_data_member_location): Multiply by 8 to get
11261 2021-07-06 Martin Sebor <msebor@redhat.com>
11263 * gimple-pretty-print.c (percent_G_format): Remove.
11264 * tree-diagnostic.c (default_tree_printer): Remove calls.
11265 * tree-pretty-print.c (percent_K_format): Remove.
11266 * tree-pretty-print.h (percent_K_format): Remove.
11268 2021-07-06 Martin Sebor <msebor@redhat.com>
11270 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_builtin):
11271 Remove %K and use error_at.
11272 (aarch64_expand_fcmla_builtin): Same.
11273 (aarch64_expand_builtin_tme): Same.
11274 (aarch64_expand_builtin_memtag): Same.
11275 * config/arm/arm-builtins.c (arm_expand_acle_builtin): Same.
11276 (arm_expand_builtin): Same.
11277 * config/arm/arm.c (bounds_check): Same.
11279 2021-07-06 Martin Sebor <msebor@redhat.com>
11281 * builtins.c (warn_string_no_nul): Remove %G.
11282 (maybe_warn_for_bound): Same.
11283 (warn_for_access): Same.
11284 (check_access): Same.
11285 (check_strncat_sizes): Same.
11286 (expand_builtin_strncat): Same.
11287 (expand_builtin_strncmp): Same.
11288 (expand_builtin): Same.
11289 (expand_builtin_object_size): Same.
11290 (warn_dealloc_offset): Same.
11291 (maybe_emit_free_warning): Same.
11292 * calls.c (maybe_warn_alloc_args_overflow): Same.
11293 (maybe_warn_nonstring_arg): Same.
11294 (maybe_warn_rdwr_sizes): Same.
11295 * expr.c (expand_expr_real_1): Remove %K.
11296 * gimple-fold.c (gimple_fold_builtin_strncpy): Remove %G.
11297 (gimple_fold_builtin_strncat): Same.
11298 * gimple-ssa-sprintf.c (format_directive): Same.
11299 (handle_printf_call): Same.
11300 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Same.
11301 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
11302 (maybe_diag_access_bounds): Same. Call gimple_location.
11303 (check_bounds_or_overlap): Same.
11304 * trans-mem.c (ipa_tm_scan_irr_block): Remove %K. Simplify.
11305 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Remove %G.
11306 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
11307 (maybe_diag_stxncpy_trunc): Same.
11308 (handle_builtin_stxncpy_strncat): Same.
11309 (maybe_warn_pointless_strcmp): Same.
11310 * tree-ssa-uninit.c (maybe_warn_operand): Same.
11312 2021-07-06 Uroš Bizjak <ubizjak@gmail.com>
11315 * config/i386/predicates.md (vec_setm_operand): Enable
11316 register_operand for TARGET_SSE4_1.
11317 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_operand
11318 as operand 2 predicate. Call ix86_expand_vector_set_var
11319 for non-constant index operand.
11320 (vec_setv4qi): Use vec_setm_mmx_operand as operand 2 predicate.
11321 Call ix86_expand_vector_set_var for non-constant index operand.
11323 2021-07-06 Jeff Law <jeffreyalaw@gmail.com>
11325 * config/h8300/jumpcall.md (*branch): When possible, generate
11326 the comparison in CCZN mode.
11327 * config/h8300/predicates.md (simple_memory_operand): Reject all
11328 auto-increment addressing modes.
11330 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
11332 PR bootstrap/100246
11333 * config/i386/i386.h (struct stringop_algs): Define a CTOR for
11336 2021-07-06 Richard Biener <rguenther@suse.de>
11338 * doc/md.texi (vec_fmaddsub<mode>4): Document.
11339 (vec_fmsubadd<mode>4): Likewise.
11340 * optabs.def (vec_fmaddsub$a4): Add.
11341 (vec_fmsubadd$a4): Likewise.
11342 * internal-fn.def (IFN_VEC_FMADDSUB): Add.
11343 (IFN_VEC_FMSUBADD): Likewise.
11344 * tree-vect-slp-patterns.c (addsub_pattern::recognize):
11345 Refactor to handle IFN_VEC_FMADDSUB and IFN_VEC_FMSUBADD.
11346 (addsub_pattern::build): Likewise.
11347 * tree-vect-slp.c (vect_optimize_slp): CFN_VEC_FMADDSUB
11348 and CFN_VEC_FMSUBADD are not transparent for permutes.
11349 * config/i386/sse.md (vec_fmaddsub<mode>4): New expander.
11350 (vec_fmsubadd<mode>4): Likewise.
11352 2021-07-06 Richard Biener <rguenther@suse.de>
11354 * doc/invoke.texi (fmove-loop-stores): Document.
11355 * common.opt (fmove-loop-stores): New option.
11356 * opts.c (default_options_table): Enable -fmove-loop-stores
11357 at -O1 but not -Og.
11358 * tree-ssa-loop-im.c (pass_lim::execute): Pass
11359 flag_move_loop_stores instead of true to
11360 loop_invariant_motion_in_fun.
11362 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
11364 * doc/install.texi: Document --with-dsymutil.
11366 2021-07-06 Andrew Pinski <apinski@marvell.com>
11368 PR tree-optimization/101256
11369 * dbgcnt.def (phiopt_edge_range): New counter.
11370 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
11371 Check to make sure the new name is defined in the same
11372 bb as the conditional before duplicating range info.
11373 Also add debug counter.
11375 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
11377 PR rtl-optimization/100328
11378 * config/i386/i386-options.c (ix86_option_override_internal):
11379 Set param_ira_consider_dup_in_all_alts to 0.
11381 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
11383 PR rtl-optimization/100328
11384 * doc/invoke.texi (ira-consider-dup-in-all-alts): Document new
11386 * ira.c (ira_get_dup_out_num): Adjust as parameter
11387 param_ira_consider_dup_in_all_alts.
11388 * params.opt (ira-consider-dup-in-all-alts): New.
11389 * ira-conflicts.c (process_regs_for_copy): Add one parameter
11390 single_input_op_has_cstr_p.
11391 (get_freq_for_shuffle_copy): New function.
11392 (add_insn_allocno_copies): Adjust as single_input_op_has_cstr_p.
11393 * ira-int.h (ira_get_dup_out_num): Add one bool parameter.
11395 2021-07-05 Jeff Law <jeffreyalaw@gmail.com>
11397 * config/h8300/shiftrotate.md (shift-by-variable patterns): Update to
11398 generate condition code aware RTL directly.
11400 2021-07-05 Andrew Pinski <apinski@marvell.com>
11402 PR tree-optimization/101039
11403 * match.pd (A CMP 0 ? A : -A): New patterns.
11404 * tree-ssa-phiopt.c (abs_replacement): Delete function.
11405 (tree_ssa_phiopt_worker): Don't call abs_replacement.
11406 Update comment about abs_replacement.
11408 2021-07-05 Andrew Pinski <apinski@marvell.com>
11410 * tree-ssa-phiopt.c (gimple_simplify_phiopt):
11411 If "A ? B : C" fails to simplify, try "(!A) ? C : B".
11413 2021-07-05 Andrew Pinski <apinski@marvell.com>
11415 * tree-ssa-phiopt.c (match_simplify_replacement):
11416 Add early_p argument. Call gimple_simplify_phiopt
11417 instead of gimple_simplify.
11418 (tree_ssa_phiopt_worker): Update call to
11419 match_simplify_replacement and allow unconditionally.
11420 (phiopt_early_allow): New function.
11421 (gimple_simplify_phiopt): New function.
11423 2021-07-05 Andrew Pinski <apinski@marvell.com>
11425 PR middle-end/101237
11426 * fold-const.c (negate_expr_p): Remove call to element_mode
11427 and TREE_MODE/TREE_TYPE when calling HONOR_SIGNED_ZEROS,
11428 HONOR_SIGN_DEPENDENT_ROUNDING, and HONOR_SNANS.
11429 (fold_negate_expr_1): Likewise.
11430 (const_unop): Likewise.
11431 (fold_cond_expr_with_comparison): Likewise.
11432 (fold_binary_loc): Likewise.
11433 (fold_ternary_loc): Likewise.
11434 (tree_call_nonnegative_warnv_p): Likewise.
11435 * match.pd (-(A + B) -> (-B) - A): Likewise.
11437 2021-07-05 Iain Sandoe <iain@sandoe.co.uk>
11439 * configure.ac: Handle --with-dsymutil in the same way as we
11440 do for the assembler and linker. (DEFAULT_DSYMUTIL): New.
11441 Extract the type and version for the dsymutil configured or
11442 found by the default searches.
11443 * config.in: Regenerated.
11444 * configure: Regenerated.
11445 * collect2.c (do_dsymutil): Handle locating dsymutil in the
11446 same way as for the assembler and linker.
11447 * config/darwin.h (DSYMUTIL): Delete.
11448 * gcc.c: Report a configured dsymutil correctly.
11449 * exec-tool.in: Allow for dsymutil.
11451 2021-07-05 Uroš Bizjak <ubizjak@gmail.com>
11453 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
11454 Handle V4QI and V2HI modes.
11455 (expand_vec_perm_blend): Allow 4-byte vector modes with TARGET_SSE4_1.
11456 Handle V4QI mode. Emit mmx_pblendvb32 for 4-byte modes.
11457 (expand_vec_perm_pshufb): Rewrite to use switch statemets.
11458 Handle 4-byte dual operands with TARGET_XOP and single operands
11459 with TARGET_SSSE3. Emit mmx_ppermv32 for TARGET_XOP and
11460 mmx_pshufbv4qi3 for TARGET_SSSE3.
11461 (expand_vec_perm_pblendv): Allow 4-byte vector modes with TARGET_SSE4_1.
11462 (expand_vec_perm_interleave2): Allow 4-byte vector modes.
11463 (expand_vec_perm_pshufb2): Allow 4-byte vector modes with TARGET_SSSE3.
11464 (expand_vec_perm_even_odd_1): Handle V4QI mode.
11465 (expand_vec_perm_broadcast_1): Handle V4QI mode.
11466 (ix86_vectorize_vec_perm_const): Handle V4QI mode.
11467 * config/i386/mmx.md (mmx_ppermv32): New insn pattern.
11468 (mmx_pshufbv4qi3): Ditto.
11469 (*mmx_pblendw32): Ditto.
11470 (*mmx_pblendw64): Rename from *mmx_pblendw.
11471 (mmx_punpckhbw_low): New insn_and_split pattern.
11472 (mmx_punpcklbw_low): Ditto.
11474 2021-07-05 Richard Biener <rguenther@suse.de>
11476 * tree-vect-loop-manip.c (vect_loop_versioning): Do not
11477 set LOOP_C_INFINITE on the vectorized loop.
11479 2021-07-05 Richard Biener <rguenther@suse.de>
11481 PR middle-end/101291
11482 * cfgloopmanip.c (loop_version): Set the loop copy of the
11483 versioned loop to the new loop.
11485 2021-07-04 Iain Sandoe <iain@sandoe.co.uk>
11488 * config.gcc: Ensure that Darwin biarch definitions are
11489 added before i386.h.
11490 * config/i386/darwin.h (TARGET_64BIT): Remove.
11491 (PR80556_WORKAROUND): New.
11492 (REAL_LIBGCC_SPEC): Amend to use PR80556_WORKAROUND.
11493 (DARWIN_SUBARCH_SPEC): New.
11494 * config/i386/darwin32-biarch.h (TARGET_64BIT_DEFAULT,
11495 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
11496 (REAL_LIBGCC_SPEC): Remove.
11497 * config/i386/darwin64-biarch.h (TARGET_64BIT_DEFAULT,
11498 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
11499 (REAL_LIBGCC_SPEC): Remove.
11501 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
11503 PR middle-end/101294
11504 * expr.c (store_constructor): Don't use vec_duplicate on vector.
11506 2021-07-02 Martin Sebor <msebor@redhat.com>
11508 PR middle-end/98871
11509 PR middle-end/98512
11510 * diagnostic.c (get_any_inlining_info): New.
11511 (update_effective_level_from_pragmas): Handle inlining context.
11512 (diagnostic_enabled): Same.
11513 (diagnostic_report_diagnostic): Same.
11514 * diagnostic.h (struct diagnostic_info): Add ctor.
11515 (struct diagnostic_context): Add new member.
11516 * tree-diagnostic.c (set_inlining_locations): New.
11517 (tree_diagnostics_defaults): Set new callback pointer.
11519 2021-07-02 Peter Bergner <bergner@linux.ibm.com>
11521 * config/rs6000/rs6000-builtin.def (BU_MMA_PAIR_LD, BU_MMA_PAIR_ST):
11523 (__builtin_vsx_lxvp, __builtin_vsx_stxvp): New built-ins.
11524 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Expand
11525 lxvp and stxvp built-ins.
11526 (mma_init_builtins): Handle lxvp and stxvp built-ins.
11527 (builtin_function_type): Likewise.
11528 * doc/extend.texi (__builtin_vsx_lxvp, __builtin_mma_stxvp): Document.
11530 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
11532 * config/h8300/h8300-protos.h (compute_a_shift_cc): Accept
11533 additional argument for the code.
11534 * config/h8300/h8300.c (compute_a_shift_cc): Accept additional
11535 argument for the code. Just return if the ZN bits are useful or
11536 not rather than the old style CC_* enums.
11537 * config/h8300/shiftrotate.md (shiftqi_noscratch): Move before
11538 more generic shiftqi patterns.
11539 (shifthi_noscratch, shiftsi_noscratch): Similarly.
11540 (shiftqi_noscratch_set_flags): New pattern.
11541 (shifthi_noscratch_set_flags, shiftsi_noscratch_set_flags): Likewise.
11543 2021-07-02 Andrew MacLeod <amacleod@redhat.com>
11545 PR tree-optimization/101223
11546 * range-op.cc (build_lt): Add -1 for signed values.
11547 (built_gt): Subtract -1 for signed values.
11549 2021-07-02 David Faust <david.faust@oracle.com>
11551 * btfout.c (get_btf_kind): Support BTF_KIND_FLOAT.
11552 (btf_asm_type): Likewise.
11554 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
11556 * config/h8300/h8300-protos.h (output_a_shift): Make first argument
11557 an array of rtx rather than a pointer to rtx. Add code argument.
11558 (compute_a_shift_length): Similarly.
11559 * config/h8300/h8300.c (h8300_shift_costs): Adjust now that the
11560 shift itself isn't an operand. Create dummy operand[0] to carry
11561 a mode and pass a suitable rtx code to compute_a_shift_length.
11562 (get_shift_alg): Adjust operand number of clobber in output templates.
11563 (output_a_shift): Make first argument an array of rtx rather than
11564 a pointer to rtx. Add code argument for the type of shift.
11565 Adjust now that the shift itself is no longer an operand.
11566 (compute_a_shift_length): Similarly.
11567 * config/h8300/shiftrotate.md (shiftqi, shifthi, shiftsi): Use an
11568 iterator rather than nshift_operator.
11569 (shiftqi_noscratch, shifthi_noscratch, shiftsi_noscratch): Likewise.
11570 (shiftqi_clobber_flags): Adjust to API changes in output_a_shift
11571 and compute_a_shift_length.
11572 (shiftqi_noscratch_clobber_flags): Likewise.
11573 (shifthi_noscratch_clobber_flags): Likewise.
11574 (shiftsi_noscratch_clobber_flags): Likewise.
11576 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
11579 * config/darwin.h (DSYMUTIL_SPEC): Do not try to run
11580 dsymutil for BTF/CTF.
11582 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
11585 * config/darwin.h (CTF_INFO_SECTION_NAME): Update the
11586 segment to include BTF.
11587 (BTF_INFO_SECTION_NAME): New.
11589 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
11591 * config/m32r/m32r-protos.h (call_operand): Adjust return type.
11592 (small_data_operand, memreg_operand, small_insn_p): Likewise.
11593 * config/m32r/m32r.c (call_operand): Adjust return type.
11594 (small_data_operand, memreg_operand): Likewise.
11596 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
11598 * config/frv/frv-protos.h (integer_register_operand): Adjust return
11600 (frv_load_operand, gpr_or_fpr_operand, gpr_no_subreg_operand): Likewise.
11601 (fpr_or_int6_operand, gpr_or_int_operand); Likewise.
11602 (gpr_or_int12_operand, gpr_or_int10_operand); Likewise.
11603 (move_source_operand, move_destination_operand): Likewise.
11604 (condexec_source_operand, condexec_dest_operand): Likewise.
11605 (lr_operand, gpr_or_memory_operand, fpr_or_memory_operand): Likewise.
11606 (reg_or_0_operand, fcc_operand, icc_operand, cc_operand): Likewise.
11607 (fcr_operand, icr_operand, cr_operand, call_operand): Likewise.
11608 (fpr_operand, even_reg_operand, odd_reg_operand): Likewise.
11609 (even_gpr_operand, odd_gpr_operand, quad_fpr_operand): Likewise.
11610 (even_fpr_operand, odd_fpr_operand): Likewise.
11611 (dbl_memory_one_insn_operand, dbl_memory_two_insn_operand): Likewise.
11612 (int12_operand, int6_operand, int5_operand, uint5_operand): Likewise.
11613 (uint4_operand, uint1_operand, int_2word_operand): Likewise
11614 (upper_int16_operand, uint16_operand, symbolic_operand): Likewise.
11615 (relational_operator, float_relational_operator): Likewise.
11616 (ccr_eqne_operator, minmax_operator): Likewise.
11617 (condexec_si_binary_operator, condexec_si_media_operator): Likewise.
11618 (condexec_si_divide_operator, condexec_si_unary_operator): Likewise.
11619 (condexec_sf_conv_operator, condexec_sf_add_operator): Likewise.
11620 (intop_compare_operator, acc_operand, even_acc_operand): Likewise.
11621 (quad_acc_operand, accg_operand): Likewise.
11623 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
11625 * config/stormy16/stormy16-protos.h (xstormy16_below_100_symbol): Change
11626 return type to a bool.
11627 (nonimmediate_nonstack_operand): Likewise.
11628 (xstormy16_splittable_below100_operand): Likewise.
11629 * config/stormy16/stormy16.c (xstormy16_below_100_symbol): Fix
11631 (xstormy16_splittable_below100_operand): Likewise.
11633 2021-07-02 Richard Biener <rguenther@suse.de>
11635 PR tree-optimization/101293
11636 * tree-ssa-loop-im.c (mem_ref_hasher::equal): Compare MEM_REF bases
11637 with combined offsets.
11638 (gather_mem_refs_stmt): Hash MEM_REFs as if their offset were
11639 combined with the rest of the offset.
11641 2021-07-02 Eric Botcazou <ebotcazou@adacore.com>
11643 * config/i386/i386.c (asm_preferred_eh_data_format): Always use the
11644 PIC encodings for PE-COFF targets.
11646 2021-07-02 Jakub Jelinek <jakub@redhat.com>
11649 * config/i386/i386-expand.c (ix86_broadcast_from_integer_constant):
11650 Return nullptr for TImode inner mode.
11652 2021-07-02 Richard Biener <rguenther@suse.de>
11654 PR tree-optimization/101280
11655 PR tree-optimization/101173
11656 * gimple-loop-interchange.cc
11657 (tree_loop_interchange::valid_data_dependences): Properly
11658 guard all dependence checks with DDR_REVERSED_P or its
11661 2021-07-02 Hongyu Wang <hongyu.wang@intel.com>
11663 * config/i386/i386-expand.c (ix86_expand_builtin):
11664 Add branch to clear odata when ZF is set for asedecenc_expand
11665 and wideaesdecenc_expand.
11667 2021-07-02 Eugene Rozenfeld <erozen@microsoft.com>
11669 * config/i386/gcc-auto-profile: regenerate
11671 2021-07-02 liuhongt <hongtao.liu@intel.com>
11673 * config/i386/sse.md (trunc<mode><pmov_dst_4>2): Refined to ..
11674 (trunc<mode><pmov_dst_4_lower>2): this.
11676 2021-07-01 David Malcolm <dmalcolm@redhat.com>
11678 * diagnostic.h (diagnostic_context::m_file_cache): New field.
11679 * input.c (class fcache): Rename to...
11680 (class file_cache_slot): ...this, making most members private and
11681 prefixing fields with "m_".
11682 (file_cache_slot::get_file_path): New accessor.
11683 (file_cache_slot::get_use_count): New accessor.
11684 (file_cache_slot::missing_trailing_newline_p): New accessor.
11685 (file_cache_slot::inc_use_count): New.
11686 (fcache_buffer_size): Move to...
11687 (file_cache_slot::buffer_size): ...here.
11688 (fcache_line_record_size): Move to...
11689 (file_cache_slot::line_record_size): ...here.
11690 (fcache_tab): Delete, in favor of global_dc->m_file_cache.
11691 (fcache_tab_size): Move to file_cache::num_file_slots.
11692 (diagnostic_file_cache_init): Update for move of fcache_tab
11693 to global_dc->m_file_cache.
11694 (diagnostic_file_cache_fini): Likewise.
11695 (lookup_file_in_cache_tab): Convert to...
11696 (file_cache::lookup_file): ...this.
11697 (diagnostics_file_cache_forcibly_evict_file): Update for move of
11698 fcache_tab to global_dc->m_file_cache, moving most of
11699 implementation to...
11700 (file_cache::forcibly_evict_file): ...this new function and...
11701 (file_cache_slot::evict): ...this new function.
11702 (evicted_cache_tab_entry): Convert to...
11703 (file_cache::evicted_cache_tab_entry): ...this.
11704 (add_file_to_cache_tab): Convert to...
11705 (file_cache::add_file): ...this, moving bulk of implementation
11707 (file_cache_slot::create): ..this new function.
11708 (file_cache::file_cache): New.
11709 (file_cache::~file_cache): New.
11710 (lookup_or_add_file_to_cache_tab): Convert to...
11711 (file_cache::lookup_or_add_file): ..this new function.
11712 (fcache::fcache): Rename to...
11713 (file_cache_slot::file_cache_slot): ...this, adding "m_" prefixes
11715 (fcache::~fcache): Rename to...
11716 (file_cache_slot::~file_cache_slot): ...this, adding "m_" prefixes
11718 (needs_read): Convert to...
11719 (file_cache_slot::needs_read_p): ...this.
11720 (needs_grow): Convert to...
11721 (file_cache_slot::needs_grow_p): ...this.
11722 (maybe_grow): Convert to...
11723 (file_cache_slot::maybe_grow): ...this.
11724 (read_data): Convert to...
11725 (file_cache_slot::read_data): ...this.
11726 (maybe_read_data): Convert to...
11727 (file_cache_slot::maybe_read_data): ...this.
11728 (get_next_line): Convert to...
11729 (file_cache_slot::get_next_line): ...this.
11730 (goto_next_line): Convert to...
11731 (file_cache_slot::goto_next_line): ...this.
11732 (read_line_num): Convert to...
11733 (file_cache_slot::read_line_num): ...this.
11734 (location_get_source_line): Update for moving of globals to
11735 global_dc->m_file_cache.
11736 (location_missing_trailing_newline): Likewise.
11737 * input.h (class file_cache_slot): New forward decl.
11738 (class file_cache): New.
11740 2021-07-01 Michael Meissner <meissner@linux.ibm.com>
11742 * config/rs6000/rs6000.c (rs6000_maybe_emit_fp_cmove): Add IEEE
11743 128-bit floating point conditional move support.
11744 (have_compare_and_set_mask): Add IEEE 128-bit floating point
11746 * config/rs6000/rs6000.md (mov<mode>cc, IEEE128 iterator): New insn.
11747 (mov<mode>cc_p10, IEEE128 iterator): New insn.
11748 (mov<mode>cc_invert_p10, IEEE128 iterator): New insn.
11749 (fpmask<mode>, IEEE128 iterator): New insn.
11750 (xxsel<mode>, IEEE128 iterator): New insn.
11752 2021-07-01 Iain Sandoe <iain@sandoe.co.uk>
11755 * config/darwin.h (CTF_INFO_SECTION_NAME): New.
11757 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
11759 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
11761 * config/i386/i386-protos.h (ix86_expand_vector_init_duplicate):
11763 * config/i386/sse.md (INT_BROADCAST_MODE): New mode iterator.
11764 (vec_duplicate<mode>): New expander.
11766 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
11769 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
11771 (ix86_byte_broadcast): New function.
11772 (ix86_convert_const_wide_int_to_broadcast): Likewise.
11773 (ix86_expand_move): Convert CONST_WIDE_INT to broadcast if mode
11774 size is 16 bytes or bigger.
11775 (ix86_broadcast_from_integer_constant): New function.
11776 (ix86_expand_vector_move): Convert CONST_WIDE_INT and CONST_VECTOR
11777 to broadcast if mode size is 16 bytes or bigger.
11778 * config/i386/i386-protos.h (ix86_gen_scratch_sse_rtx): New
11780 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): New function.
11782 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
11784 * config/i386/predicates.md (ix86_endbr_immediate_operand):
11785 Return true/false instead of 1/0.
11786 (movq_parallel): Ditto.
11788 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
11790 * recog.c (general_operand): Return true/false instead of 1/0.
11791 (register_operand): Ditto.
11792 (immediate_operand): Ditto.
11793 (const_int_operand): Ditto.
11794 (const_scalar_int_operand): Ditto.
11795 (const_double_operand): Ditto.
11796 (push_operand): Ditto.
11797 (pop_operand): Ditto.
11798 (memory_operand): Ditto.
11799 (indirect_operand): Ditto.
11801 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
11803 * genpreds.c (write_predicate_subfunction):
11804 Change the type of written subfunction to bool.
11805 (write_one_predicate_function):
11806 Change the type of written function to bool.
11807 (write_tm_preds_h): Ditto.
11808 * recog.h (*insn_operand_predicate_fn): Change the type to bool.
11809 * recog.c (general_operand): Change the type to bool.
11810 (address_operand): Ditto.
11811 (register_operand): Ditto.
11812 (pmode_register_operand): Ditto.
11813 (scratch_operand): Ditto.
11814 (immediate_operand): Ditto.
11815 (const_int_operand): Ditto.
11816 (const_scalar_int_operand): Ditto.
11817 (const_double_operand): Ditto.
11818 (nonimmediate_operand): Ditto.
11819 (nonmemory_operand): Ditto.
11820 (push_operand): Ditto.
11821 (pop_operand): Ditto.
11822 (memory_operand): Ditto.
11823 (indirect_operand): Ditto.
11824 (ordered_comparison_operator): Ditto.
11825 (comparison_operator): Ditto.
11826 * config/i386/i386-expand.c (ix86_expand_sse_cmp):
11827 Change the type of indirect predicate function to bool.
11828 * config/rs6000/rs6000.c (easy_vector_constant):
11829 Change the type to bool.
11830 * config/mips/mips-protos.h (m16_based_address_p):
11831 Change the type of operand 3 to bool.
11833 2021-07-01 Richard Biener <rguenther@suse.de>
11835 PR tree-optimization/101280
11836 PR tree-optimization/101173
11837 * gimple-loop-interchange.cc
11838 (tree_loop_interchange::valid_data_dependences): Revert
11839 previous change and instead correctly handle DDR_REVERSED_P
11842 2021-07-01 Richard Biener <rguenther@suse.de>
11844 PR tree-optimization/101278
11845 * tree-ssa-dse.c (dse_classify_store): First check for
11846 uses, then ignore stmt for chaining purposes.
11848 2021-07-01 Richard Biener <rguenther@suse.de>
11850 PR tree-optimization/100778
11851 * tree-vect-slp.c (vect_schedule_slp_node): Do not place trapping
11852 vectorized ops ahead of their scalar BB.
11854 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
11857 * config/i386/i386.md (*nabs<dwi>2_doubleword):
11858 New insn_and_split pattern.
11859 (*nabs<dwi>2_1): Ditto.
11860 * config/i386/i386-features.c
11861 (general_scalar_chain::compute_convert_gain):
11862 Handle (NEG (ABS (...))) RTX. Rewrite src code
11863 scanner as switch statement.
11864 (general_scalar_chain::convert_insn):
11865 Handle (NEG (ABS (...))) RTX.
11866 (general_scalar_to_vector_candidate_p):
11867 Detect (NEG (ABS (...))) RTX. Reorder case statements
11868 for (AND (NOT (...) ...)) fallthrough.
11870 2021-07-01 Richard Biener <rguenther@suse.de>
11872 PR tree-optimization/101178
11873 * tree-vect-slp.c (slpg_vertex::materialize): Remove.
11874 (slpg::perm_in): Add.
11875 (slpg::get_perm_in): Remove.
11876 (slpg::get_perm_materialized): Add.
11877 (vect_optimize_slp): Handle VEC_PERM nodes more optimally
11878 during permute propagation and materialization.
11880 2021-07-01 Jakub Jelinek <jakub@redhat.com>
11883 * dwarf2out.c (loc_list_from_tree_1): Handle COMPOUND_LITERAL_EXPR.
11885 2021-07-01 Jakub Jelinek <jakub@redhat.com>
11887 PR middle-end/94366
11888 * omp-low.c (lower_rec_input_clauses): Rename is_fp_and_or to
11889 is_truth_op, set it for TRUTH_*IF_EXPR regardless of new_var's type,
11890 use boolean_type_node instead of integer_type_node as NE_EXPR type.
11891 (lower_reduction_clauses): Likewise.
11893 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
11895 * config/gcn/gcn.c: Include dwarf2.h.
11896 (gcn_addr_space_debug): New function.
11897 (TARGET_ADDR_SPACE_DEBUG): New hook.
11899 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
11901 * common/config/gcn/gcn-common.c
11902 (gcn_option_optimization_table): Change OPT_fomit_frame_pointer to -O3.
11903 * config/gcn/gcn.c (gcn_expand_prologue): Prefer the frame pointer
11905 (gcn_expand_prologue): Prefer the frame pointer when emitting CFI.
11906 (gcn_frame_pointer_rqd): New function.
11907 (TARGET_FRAME_POINTER_REQUIRED): New hook.
11909 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
11911 * config/gcn/gcn.c (move_callee_saved_registers): Emit CFI notes for
11912 prologue register saves.
11913 (gcn_debug_unwind_info): Use UI_DWARF2.
11914 (gcn_dwarf_register_number): Map DWARF_LINK_REGISTER to DWARF PC.
11915 (gcn_dwarf_register_span): DWARF_LINK_REGISTER doesn't span.
11916 * config/gcn/gcn.h: (DWARF_FRAME_RETURN_COLUMN): New define.
11917 (DWARF_LINK_REGISTER): New define.
11918 (FIRST_PSEUDO_REGISTER): Increment.
11919 (FIXED_REGISTERS): Add entry for DWARF_LINK_REGISTER.
11920 (CALL_USED_REGISTERS): Likewise.
11921 (REGISTER_NAMES): Likewise.
11923 2021-06-30 Richard Biener <rguenther@suse.de>
11925 PR tree-optimization/101267
11926 * tree-vect-stmts.c (vect_check_scalar_mask): Adjust
11927 API and use SLP compatible interface of vect_is_simple_use.
11928 Reject not vectorized SLP defs for callers that do not support
11930 (vect_check_store_rhs): Handle masked stores and pass down
11931 the appropriate operator index.
11932 (vectorizable_call): Adjust.
11933 (vectorizable_store): Likewise.
11934 (vectorizable_load): Likewise. Handle SLP pecularity of
11936 (vect_is_simple_use): Remove special-casing of masked stores.
11938 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
11940 * common.opt (foffload): Remove help as Driver only.
11941 * gcc.c (display_help): Add -foffload.
11943 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
11945 * gcc.c (close_at_file, execute): Replace alloca by XALLOCAVEC.
11946 (check_offload_target_name): Fix splitting OFFLOAD_TARGETS into
11947 a candidate list; better inform no offload target is configured
11948 and fix hint extraction when passed target is not '\0' at [len].
11949 * common.opt (foffload): Add tailing '.'.
11950 (foffload-options): Likewise; fix flag name in the help string.
11952 2021-06-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
11955 * config/arm/arm_neon.h: Move vabs intrinsics before vcage_f32.
11956 (vcage_f32): Gate comparison on __FAST_MATH__.
11957 (vcageq_f32): Likewise.
11958 (vcale_f32): Likewise.
11959 (vcaleq_f32): Likewise.
11960 (vcagt_f32): Likewise.
11961 (vcagtq_f32): Likewise.
11962 (vcalt_f32): Likewise.
11963 (vcaltq_f32): Likewise.
11964 (vcage_f16): Likewise.
11965 (vcageq_f16): Likewise.
11966 (vcale_f16): Likewise.
11967 (vcaleq_f16): Likewise.
11968 (vcagt_f16): Likewise.
11969 (vcagtq_f16): Likewise.
11970 (vcalt_f16): Likewise.
11971 (vcaltq_f16): Likewise.
11973 2021-06-30 Richard Biener <rguenther@suse.de>
11975 PR tree-optimization/101264
11976 * tree-vect-slp.c (vect_optimize_slp): Propagate the
11977 computed perm_in to all "any" permute successors
11978 we cannot de-duplicate immediately.
11980 2021-06-30 liuhongt <hongtao.liu@intel.com>
11983 * config/i386/sse.md
11984 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>):
11986 (avx512f_sfixupimm<mode><maskz_scalar_name><round_saeonly_name>):
11988 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>"): Refined.
11989 * config/i386/subst.md (maskz_scalar): New define_subst.
11990 (maskz_scalar_name): New subst_attr.
11991 (maskz_scalar_op5): Ditto.
11992 (round_saeonly_maskz_scalar_op5): Ditto.
11993 (round_saeonly_maskz_scalar_operand5): Ditto.
11995 2021-06-30 David Edelsohn <dje.gcc@gmail.com>
11997 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
11998 Increase code CSECT alignment to at least 32 bytes.
11999 * config/rs6000/xcoff.h (TEXT_SECTION_ASM_OP): Add 32 byte
12000 alignment designation.
12002 2021-06-29 Sergei Trofimovich <siarheit@google.com>
12004 * doc/generic.texi: Fix s/net yet/not yet/ typo.
12006 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
12008 PR tree-optimization/101254
12009 * range-op.cc (operator_minus::op1_op2_relation_effect): Check for
12010 wrapping/non-wrapping when setting the result range.
12012 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
12014 * value-query.cc (gimple_range_global): Allow phis.
12016 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
12018 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): Use stmt.
12019 (simplify_using_ranges::op_with_boolean_value_range_p): Add a
12020 statement for location context.
12021 (check_for_binary_op_overflow): Ditto.
12022 (simplify_using_ranges::get_vr_for_comparison): Ditto.
12023 (simplify_using_ranges::compare_name_with_value): Ditto.
12024 (simplify_using_ranges::compare_names): Ditto.
12025 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Ditto.
12026 (simplify_using_ranges::simplify_truth_ops_using_ranges): Ditto.
12027 (simplify_using_ranges::simplify_min_or_max_using_ranges): Ditto.
12028 (simplify_using_ranges::simplify_internal_call_using_ranges): Ditto.
12029 (simplify_using_ranges::two_valued_val_range_p): Ditto.
12030 (simplify_using_ranges::simplify): Ditto.
12031 * vr-values.h: Adjust prototypes.
12033 2021-06-29 Uroš Bizjak <ubizjak@gmail.com>
12036 * config/i386/mmx.md (vec_addsubv2sf3): New insn pattern.
12038 2021-06-29 Julian Brown <julian@codesourcery.com>
12040 * config/gcn/gcn.c (gcn_init_libfuncs): New function.
12041 (TARGET_INIT_LIBFUNCS): Define target hook using above function.
12042 * config/gcn/gcn.h (UNITS_PER_WORD): Define to 8 for IN_LIBGCC2, 4
12044 (LIBGCC2_UNITS_PER_WORD, BITS_PER_WORD): Remove definitions.
12045 (MAX_FIXED_MODE_SIZE): Change to 128.
12047 2021-06-29 Julian Brown <julian@codesourcery.com>
12049 * config/gcn/gcn.md (UNSPEC_FLBIT_INT): New unspec constant.
12050 (s_mnemonic): Add clrsb.
12051 (gcn_flbit<mode>_int): Add insn pattern for SImode/DImode.
12052 (clrsb<mode>2): Add expander for SImode/DImode.
12054 2021-06-29 Julian Brown <julian@codesourcery.com>
12056 * config/gcn/gcn.md (<su>mulsidi3, <su>mulsidi3_reg, <su>mulsidi3_imm,
12057 muldi3): Add patterns.
12059 2021-06-29 Julian Brown <julian@codesourcery.com>
12061 * config/gcn/gcn.md (<su>mulsi3_highpart): Change to expander.
12062 (<su>mulsi3_highpart_reg, <su>mulsi3_highpart_imm): New patterns.
12064 2021-06-29 Julian Brown <julian@codesourcery.com>
12066 * config/gcn/gcn.md (mulsi3): Make s_mulk_i32 variant clobber SCC.
12068 2021-06-29 Joseph Myers <joseph@codesourcery.com>
12070 * btfout.c, ctfout.c: Include "memmodel.h".
12072 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
12074 * gcc.c (check_offload_target_name): Cast len argument to
12075 %q.*s to 'int'; avoid -Wstringop-truncation warning.
12077 2021-06-29 Richard Biener <rguenther@suse.de>
12079 * tree-vect-slp.c (vect_optimize_slp): Forward propagate
12080 to "any" permute nodes and relax "any" permute proapgation
12081 during iterative backward propagation.
12083 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
12086 * common.opt (-foffload=): Update description.
12087 (-foffload-options=): New.
12088 * doc/invoke.texi (C Language Options): Document
12089 -foffload and -foffload-options.
12090 * gcc.c (check_offload_target_name): New, split off from
12091 handle_foffload_option.
12092 (check_foffload_target_names): New.
12093 (handle_foffload_option): Handle -foffload=default.
12094 (driver_handle_option): Update for -foffload-options.
12095 * lto-opts.c (lto_write_options): Use -foffload-options
12096 instead of -foffload.
12097 * lto-wrapper.c (merge_and_complain, append_offload_options):
12099 * opts.c (common_handle_option): Likewise.
12101 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
12103 * doc/invoke.texi (C Language Options): Sort options
12104 alphabetically in optlist and also the description itself.
12105 Remove leftover -fallow-single-precision from and add missing
12106 -fgnu-tm to the optlist.
12108 2021-06-29 Richard Biener <rguenther@suse.de>
12110 * tree-vect-slp.c (slpg_vertex::visited): Remove.
12111 (vect_slp_perms_eq): Handle -1 permutes.
12112 (vect_optimize_slp): Rewrite permute propagation.
12114 2021-06-29 Jakub Jelinek <jakub@redhat.com>
12117 * match.pd ((intptr_t)x eq/ne CST to x eq/ne (typeof x) CST): Don't
12118 perform the optimization in GENERIC when sanitizing and x has a
12121 2021-06-29 Richard Biener <rguenther@suse.de>
12123 PR tree-optimization/101242
12124 * tree-vect-slp.c (vect_slp_build_vertices): Force-add
12125 PHIs with not represented initial values as leafs.
12127 2021-06-29 Jan-Benedict Glaw <jbglaw@getslash.de>
12129 * config/pdp11/pdp11.h (ASM_OUTPUT_SKIP): Fix signedness warning.
12130 * config/pdp11/pdp11.c (pdp11_asm_print_operand_punct_valid_p): Remove
12131 "register" keyword.
12132 (pdp11_initial_elimination_offset) Remove unused variable.
12133 (pdp11_cmp_length) Ditto.
12134 (pdp11_insn_cost): Ditto, and fix signedness warning.
12136 2021-06-29 David Edelsohn <dje.gcc@gmail.com>
12138 * btfout.c: Include tm_p.h.
12141 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
12143 * config/bpf/bpf.c (bpf_expand_prologue): Do not mark insns as
12145 (bpf_expand_epilogue): Likewise.
12146 * config/bpf/bpf.h (DWARF2_FRAME_INFO): Define to 0.
12147 Do not define DBX_DEBUGGING_INFO.
12149 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
12151 * doc/invoke.texi: Document the CTF and BTF debug info options.
12153 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
12154 David Faust <david.faust@oracle.com>
12155 Jose E. Marchesi <jose.marchesi@oracle.com>
12156 Weimin Pan <weimin.pan@oracle.com>
12158 * Makefile.in: Add ctfc.*, ctfout.c and btfout.c files to
12159 GTFILES. Add new object files.
12160 * common.opt: Add CTF and BTF debug info options.
12161 * btfout.c: New file.
12162 * ctfc.c: Likewise.
12163 * ctfc.h: Likewise.
12164 * ctfout.c: Likewise.
12165 * dwarf2ctf.c: Likewise.
12166 * dwarf2ctf.h: Likewise.
12167 * dwarf2cfi.c (dwarf2out_do_frame): Acknowledge CTF_DEBUG and
12169 * dwarf2out.c (dwarf2out_source_line): Likewise.
12170 (dwarf2out_finish): Skip emitting DWARF if CTF or BTF are to
12172 (debug_format_do_cu): New function.
12173 (dwarf2out_early_finish): Traverse DIEs and emit CTF/BTF for
12175 Include dwarf2ctf.c.
12176 * final.c (dwarf2_debug_info_emitted_p): Acknowledge DWARF-based debug
12178 * flag-types.h (enum debug_info_type): Add CTF_DEBUG and BTF_DEBUG.
12179 (CTF_DEBUG): New bitmask.
12180 (BTF_DEBUG): Likewise.
12181 (enum ctf_debug_info_levels): New enum.
12182 * gengtype.c (open_base_files): Handle ctfc.h.
12183 (main): Handle uint32_t type.
12184 * flags.h (btf_debuginfo_p): New definition.
12185 (dwarf_based_debuginfo_p): Likewise.
12186 * opts.c (debug_type_names): Add entries for CTF and BTF.
12187 (btf_debuginfo_p): New function.
12188 (dwarf_based_debuginfo_p): Likewise.
12189 (common_handle_option): Handle -gctfN and -gbtf options.
12190 (set_debug_level): Set CTF_DEBUG, BTF_DEBUG whenever appropriate.
12191 * toplev.c (process_options): Inform the user and ignore -gctfLEVEL if
12194 2021-06-28 Jose E. Marchesi <jose.marchesi@oracle.com>
12196 * dwarf2out.c (AT_class): Function is no longer static.
12197 (AT_int): Likewise.
12198 (AT_unsigned): Likewise.
12199 (AT_loc): Likewise.
12200 (get_AT): Likewise.
12201 (get_AT_string): Likewise.
12202 (get_AT_flag): Likewise.
12203 (get_AT_unsigned): Likewise.
12204 (get_AT_ref): Likewise.
12205 (new_die_raw): Likewise.
12206 (lookup_decl_die): Likewise.
12207 (base_type_die): Likewise.
12208 (add_name_attribute): Likewise.
12209 (add_AT_int): Likewise.
12210 (add_AT_unsigned): Likewise.
12211 (add_AT_loc): Likewise.
12212 (dw_get_die_tag): New function.
12213 (dw_get_die_child): Likewise.
12214 (dw_get_die_sib): Likewise.
12215 (struct dwarf_file_data): Move from here to dwarf2out.h
12216 (struct dw_attr_struct): Likewise.
12217 * dwarf2out.h: Analogous changes.
12219 2021-06-28 Martin Jambor <mjambor@suse.cz>
12222 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
12223 members m_dead_stmts and m_dead_ssas.
12224 * ipa-param-manipulation.c
12225 (ipa_param_body_adjustments::mark_dead_statements): New function.
12226 (ipa_param_body_adjustments::common_initialization): Call it on
12227 all removed but not split parameters.
12228 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
12230 (ipa_param_body_adjustments::modify_call_stmt): Remove arguments that
12232 * tree-inline.c (remap_gimple_stmt): Do not copy dead statements, reset
12233 dead debug statements.
12234 (copy_phis_for_bb): Do not copy dead PHI nodes.
12236 2021-06-28 Martin Jambor <mjambor@suse.cz>
12239 * symtab-clones.h (clone_info): Removed member param_adjustments.
12240 * ipa-param-manipulation.h: Adjust initial comment to reflect how we
12241 deal with pass-through splits now.
12242 (ipa_param_performed_split): Removed.
12243 (ipa_param_adjustments::modify_call): Adjusted parameters.
12244 (class ipa_param_body_adjustments): Adjusted parameters of
12245 register_replacement, modify_gimple_stmt and modify_call_stmt.
12246 (ipa_verify_edge_has_no_modifications): Declare.
12247 (ipa_edge_modifications_finalize): Declare.
12248 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Remove
12249 performed_splits processing, pas only edge to padjs->modify_call,
12250 check that call arguments were not modified if they should not have
12252 * cgraphclones.c (cgraph_node::create_clone): Do not copy performed
12254 * ipa-param-manipulation.c (struct pass_through_split_map): New type.
12255 (ipa_edge_modification_info): Likewise.
12256 (ipa_edge_modification_sum): Likewise.
12257 (ipa_edge_modifications): New edge summary.
12258 (ipa_verify_edge_has_no_modifications): New function.
12259 (transitive_split_p): Removed.
12260 (transitive_split_map): Likewise.
12261 (init_transitive_splits): Likewise.
12262 (ipa_param_adjustments::modify_call): Adjusted to use the new edge
12263 summary instead of performed_splits.
12264 (ipa_param_body_adjustments::register_replacement): Drop dummy
12265 parameter, set base_index of the created ipa_param_body_replacement.
12266 (phi_arg_will_live_p): New function.
12267 (ipa_param_body_adjustments::common_initialization): Do not create
12268 IPA_SRA dummy decls.
12269 (simple_tree_swap_info): Removed.
12270 (remap_split_decl_to_dummy): Likewise.
12271 (record_argument_state_1): New function.
12272 (record_argument_state): Likewise.
12273 (ipa_param_body_adjustments::modify_call_stmt): New parameter
12274 orig_stmt. Do not work with dummy decls, save necessary info about
12275 changes to ipa_edge_modifications.
12276 (ipa_param_body_adjustments::modify_gimple_stmt): New parameter
12277 orig_stmt, pass it to modify_call_stmt.
12278 (ipa_param_body_adjustments::modify_cfun_body): Adjust call to
12279 modify_gimple_stmt.
12280 (ipa_edge_modifications_finalize): New function.
12281 * tree-inline.c (remap_gimple_stmt): Pass original statement to
12282 modify_gimple_stmt.
12283 (copy_phis_for_bb): Do not copy dead PHI nodes.
12284 (expand_call_inline): Do not remap performed_splits.
12285 (update_clone_info): Likewise.
12286 * toplev.c: Include ipa-param-manipulation.h.
12287 (toplev::finalize): Call ipa_edge_modifications_finalize.
12289 2021-06-28 Andrew Pinski <apinski@marvell.com>
12291 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Duplicate range
12292 info if we're the only things setting the target PHI.
12293 (value_replacement): Don't duplicate range here.
12294 (minmax_replacement): Likewise.
12296 2021-06-28 Richard Biener <rguenther@suse.de>
12298 PR tree-optimization/101229
12299 * gimple-walk.c (gimple_walk_op): Handle PHIs.
12301 2021-06-28 Martin Liska <mliska@suse.cz>
12303 * config/v850/v850.c (construct_dispose_instruction): Allocate
12305 (construct_prepare_instruction): Likewise.
12307 2021-06-28 Martin Liska <mliska@suse.cz>
12309 * config/v850/v850.c (v850_option_override): Build default
12311 (v850_can_inline_p): New. Allow MASK_PROLOG_FUNCTION to be
12312 ignored for inlining.
12313 (TARGET_CAN_INLINE_P): New.
12315 2021-06-28 Richard Biener <rguenther@suse.de>
12317 PR tree-optimization/101207
12318 * tree-vect-slp.c (vect_optimize_slp): Do BB reduction
12319 permute eliding for load permutations properly.
12321 2021-06-28 Richard Biener <rguenther@suse.de>
12323 PR tree-optimization/101173
12324 * gimple-loop-interchange.cc
12325 (tree_loop_interchange::valid_data_dependences): Disallow outer
12326 loop dependence distance of zero.
12328 2021-06-28 liuhongt <hongtao.liu@intel.com>
12331 * config/i386/sse.md (*avx_cmp<mode>3_lt): New
12332 define_insn_and_split.
12333 (*avx_cmp<mode>3_ltint): Ditto.
12334 (*avx2_pcmp<mode>3_3): Ditto.
12335 (*avx2_pcmp<mode>3_4): Ditto.
12336 (*avx2_pcmp<mode>3_5): Ditto.
12338 2021-06-28 liuhongt <hongtao.liu@intel.com>
12340 * config/i386/i386-builtin.def (IX86_BUILTIN_BLENDVPD256,
12341 IX86_BUILTIN_BLENDVPS256, IX86_BUILTIN_PBLENDVB256,
12342 IX86_BUILTIN_BLENDVPD, IX86_BUILTIN_BLENDVPS,
12343 IX86_BUILTIN_PBLENDVB128): Replace icode with
12345 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold blendv
12347 * config/i386/sse.md (*<sse4_1_avx2>_pblendvb_lt_subreg_not):
12348 New pre_reload splitter.
12350 2021-06-27 Andrew Pinski <apinski@marvell.com>
12352 PR middle-end/101230
12353 * fold-const.c (fold_ternary_loc): Check
12354 the return value of invert_tree_comparison.
12356 2021-06-27 David Edelsohn <dje.gcc@gmail.com>
12358 * config.gcc: Add SPDX License Identifier.
12359 (powerpc-ibm-aix789): Default to aix73.h.
12360 (powerpc-ibm-aix7.2.*.*): New stanza.
12361 * config/rs6000/aix72.h: Add SPDX License Identifier.
12362 * config/rs6000/aix73.h: New file.
12364 2021-06-26 Jason Merrill <jason@redhat.com>
12366 * except.c: #include "dwarf2.h" instead of "dwarf2out.h".
12368 2021-06-26 Andrew Pinski <apinski@marvell.com>
12370 * genmatch.c (lower_cond): Copy for_subst_vec
12371 for the simplify also.
12372 (lower): Swap the order for lower_for and lower_cond.
12374 2021-06-26 Andrew Pinski <apinski@marvell.com>
12376 * tree-ssa-phiopt.c (match_simplify_replacement): Reset
12377 flow senatitive info on the moved ssa set.
12379 2021-06-26 Andrew Pinski <apinski@marvell.com>
12381 * fold-const.c (fold_cond_expr_with_comparison):
12382 Exand arg0 into comp_code, arg00, and arg01.
12383 (fold_ternary_loc): Use invert_tree_comparison
12384 instead of fold_invert_truthvalue for the case
12385 where we have A CMP B ? C : A.
12387 2021-06-25 Martin Sebor <msebor@redhat.com>
12389 PR middle-end/101216
12390 * calls.c (maybe_warn_rdwr_sizes): Use the no_warning constant.
12392 2021-06-25 Jeff Law <jeffreyalaw@gmail.com>
12394 * config/h8300/h8300.c (select_cc_mode): Handle ASHIFTRT and LSHIFTRT.
12396 2021-06-25 Richard Biener <rguenther@suse.de>
12398 PR tree-optimization/101202
12399 * tree-vect-slp.c (vect_optimize_slp): Explicitely handle
12402 2021-06-25 Richard Biener <rguenther@suse.de>
12404 * tree-vect-slp-patterns.c (addsub_pattern::build): Copy
12405 STMT_VINFO_REDUC_DEF from the original representative.
12407 2021-06-25 Martin Sebor <msebor@redhat.com>
12409 * builtins.c (warn_string_no_nul): Replace uses of TREE_NO_WARNING,
12410 gimple_no_warning_p and gimple_set_no_warning with
12411 warning_suppressed_p, and suppress_warning.
12413 (maybe_warn_for_bound): Same.
12414 (warn_for_access): Same.
12415 (check_access): Same.
12416 (expand_builtin_strncmp): Same.
12417 (fold_builtin_varargs): Same.
12418 * calls.c (maybe_warn_nonstring_arg): Same.
12419 (maybe_warn_rdwr_sizes): Same.
12420 * cfgexpand.c (expand_call_stmt): Same.
12421 * cgraphunit.c (check_global_declaration): Same.
12422 * fold-const.c (fold_undefer_overflow_warnings): Same.
12423 (fold_truth_not_expr): Same.
12424 (fold_unary_loc): Same.
12425 (fold_checksum_tree): Same.
12426 * gimple-array-bounds.cc (array_bounds_checker::check_array_ref): Same.
12427 (array_bounds_checker::check_mem_ref): Same.
12428 (array_bounds_checker::check_addr_expr): Same.
12429 (array_bounds_checker::check_array_bounds): Same.
12430 * gimple-expr.c (copy_var_decl): Same.
12431 * gimple-fold.c (gimple_fold_builtin_strcpy): Same.
12432 (gimple_fold_builtin_strncat): Same.
12433 (gimple_fold_builtin_stxcpy_chk): Same.
12434 (gimple_fold_builtin_stpcpy): Same.
12435 (gimple_fold_builtin_sprintf): Same.
12436 (fold_stmt_1): Same.
12437 * gimple-ssa-isolate-paths.c (diag_returned_locals): Same.
12438 * gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Same.
12439 * gimple-ssa-sprintf.c (handle_printf_call): Same.
12440 * gimple-ssa-store-merging.c (imm_store_chain_info::output_merged_store): Same.
12441 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
12442 * gimple-ssa-warn-restrict.h: Adjust declarations.
12443 (maybe_diag_access_bounds): Replace uses of TREE_NO_WARNING,
12444 gimple_no_warning_p and gimple_set_no_warning with
12445 warning_suppressed_p, and suppress_warning.
12446 (check_call): Same.
12447 (check_bounds_or_overlap): Same.
12448 * gimple.c (gimple_build_call_from_tree): Same.
12449 * gimplify.c (gimplify_return_expr): Same.
12450 (gimplify_cond_expr): Same.
12451 (gimplify_modify_expr_complex_part): Same.
12452 (gimplify_modify_expr): Same.
12453 (gimple_push_cleanup): Same.
12454 (gimplify_expr): Same.
12455 * omp-expand.c (expand_omp_for_generic): Same.
12456 (expand_omp_taskloop_for_outer): Same.
12457 * omp-low.c (lower_rec_input_clauses): Same.
12458 (lower_lastprivate_clauses): Same.
12459 (lower_send_clauses): Same.
12460 (lower_omp_target): Same.
12461 * tree-cfg.c (pass_warn_function_return::execute): Same.
12462 * tree-complex.c (create_one_component_var): Same.
12463 * tree-inline.c (remap_gimple_op_r): Same.
12464 (copy_tree_body_r): Same.
12465 (declare_return_variable): Same.
12466 (expand_call_inline): Same.
12467 * tree-nested.c (lookup_field_for_decl): Same.
12468 * tree-sra.c (create_access_replacement): Same.
12469 (generate_subtree_copies): Same.
12470 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Same.
12471 * tree-ssa-forwprop.c (combine_cond_expr_cond): Same.
12472 * tree-ssa-loop-ch.c (ch_base::copy_headers): Same.
12473 * tree-ssa-loop-im.c (execute_sm): Same.
12474 * tree-ssa-phiopt.c (cond_store_replacement): Same.
12475 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
12476 (handle_builtin_strcpy): Same.
12477 (maybe_diag_stxncpy_trunc): Same.
12478 (handle_builtin_stxncpy_strncat): Same.
12479 (handle_builtin_strcat): Same.
12480 * tree-ssa-uninit.c (get_no_uninit_warning): Same.
12481 (set_no_uninit_warning): Same.
12482 (uninit_undefined_value_p): Same.
12483 (warn_uninit): Same.
12484 (maybe_warn_operand): Same.
12485 * tree-vrp.c (compare_values_warnv): Same.
12486 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr): Same.
12487 (test_for_singularity): Same.
12488 * gimple.h (warning_suppressed_p): New function.
12489 (suppress_warning): Same.
12490 (copy_no_warning): Same.
12491 (gimple_set_block): Call gimple_set_location.
12492 (gimple_set_location): Call copy_warning.
12494 2021-06-25 Martin Sebor <msebor@redhat.com>
12496 * tree.h (warning_suppressed_at, copy_warning,
12497 warning_suppressed_p, suppress_warning): New functions.
12499 2021-06-25 Martin Sebor <msebor@redhat.com>
12501 * Makefile.in (OBJS-libcommon): Add diagnostic-spec.o.
12502 * gengtype.c (open_base_files): Add diagnostic-spec.h.
12503 * diagnostic-spec.c: New file.
12504 * diagnostic-spec.h: New file.
12505 * tree.h (no_warning, all_warnings, suppress_warning_at): New
12507 * warning-control.cc: New file.
12509 2021-06-25 liuhongt <hongtao.liu@intel.com>
12512 * config/i386/i386.c (x86_order_regs_for_local_alloc):
12515 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
12517 PR tree-optimization/101189
12518 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Pass
12519 LHS range of condition to postfold routine.
12520 (fold_using_range::postfold_gcond_edges): Only process the TRUE or
12521 FALSE edge if the LHS range supports it being taken.
12522 * gimple-range-fold.h (postfold_gcond_edges): Add range parameter.
12524 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
12526 * value-relation.cc (equiv_oracle::dump): Do not dump NULL blocks.
12527 (relation_oracle::find_relation_block): Check correct bitmap.
12528 (relation_oracle::dump): Do not dump NULL blocks.
12530 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
12532 * gimple-range-cache.cc (ranger_cache::propagate_cache): Call
12533 range_on_edge instead of manually calculating.
12535 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
12537 * range-op.cc: Fix comment.
12539 2021-06-24 Uroš Bizjak <ubizjak@gmail.com>
12542 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
12543 Handle V8QI and V4HI modes.
12544 * config/i386/mmx.md (sse4_1_<any_extend:code>v4qiv4hi2):
12546 (sse4_1_<any_extend:code>v4qiv4hi2): Ditto.
12547 (mmxpackmode): New mode attribute.
12548 (vec_pack_trunc_<mmxpackmode:mode>): New expander.
12549 (mmxunpackmode): New mode attribute.
12550 (vec_unpacks_lo_<mmxunpackmode:mode>): New expander.
12551 (vec_unpacks_hi_<mmxunpackmode:mode>): Ditto.
12552 (vec_unpacku_lo_<mmxunpackmode:mode>): Ditto.
12553 (vec_unpacku_hi_<mmxunpackmode:mode>): Ditto.
12554 * config/i386/i386.md (extsuffix): Move from ...
12555 * config/i386/sse.md: ... here.
12557 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
12559 * dwarf2out.c (dwarf2out_assembly_start): Emit .file 0 marker here..
12560 (dwarf2out_finish): ...instead of here.
12562 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
12564 * configure.ac (--gdwarf-5 option): Use objdump instead of readelf.
12565 (working --gdwarf-4/--gdwarf-5 for all sources): Likewise.
12566 (--gdwarf-4 not refusing generated .debug_line): Adjust for Windows.
12567 * configure: Regenerate.
12569 2021-06-24 Richard Biener <rguenther@suse.de>
12571 * config/i386/sse.md (vec_addsubv4df3, vec_addsubv2df3,
12572 vec_addsubv8sf3, vec_addsubv4sf3): Merge into ...
12573 (vec_addsub<mode>3): ... using a new addsub_cst mode attribute.
12575 2021-06-24 Richard Biener <rguenther@suse.de>
12577 * config/i386/sse.md (avx_addsubv4df3): Rename to
12579 (avx_addsubv8sf3): Rename to vec_addsubv8sf3.
12580 (sse3_addsubv2df3): Rename to vec_addsubv2df3.
12581 (sse3_addsubv4sf3): Rename to vec_addsubv4sf3.
12582 * config/i386/i386-builtin.def: Adjust.
12583 * internal-fn.def (VEC_ADDSUB): New internal optab fn.
12584 * optabs.def (vec_addsub_optab): New optab.
12585 * tree-vect-slp-patterns.c (class addsub_pattern): New.
12586 (slp_patterns): Add addsub_pattern.
12587 * tree-vect-slp.c (vect_optimize_slp): Disable propagation
12588 across CFN_VEC_ADDSUB.
12589 * tree-vectorizer.h (vect_pattern::vect_pattern): Make
12591 * doc/md.texi (vec_addsub<mode>3): Document.
12593 2021-06-24 Jakub Jelinek <jakub@redhat.com>
12595 PR middle-end/101170
12596 * df-scan.c (df_ref_record): For paradoxical big-endian SUBREGs
12597 where regno + subreg_regno_offset wraps around use 0 as starting
12600 2021-06-24 Jakub Jelinek <jakub@redhat.com>
12602 PR middle-end/101172
12603 * stor-layout.c (finish_bitfield_representative): If nextf has
12604 error_mark_node type, set repr type to error_mark_node too.
12606 2021-06-24 Ilya Leoshkevich <iii@linux.ibm.com>
12608 * config/s390/s390.c (s390_function_profiler): Ignore labelno
12610 * config/s390/s390.h (NO_PROFILE_COUNTERS): Define.
12612 2021-06-24 Richard Biener <rguenther@suse.de>
12614 * tree-vect-slp.c (vect_optimize_slp): Do not propagate
12615 across operations that have different semantics on different
12618 2021-06-24 Jakub Jelinek <jakub@redhat.com>
12620 * tree.h (OMP_CLAUSE_MAP_IN_REDUCTION): Document meaning for OpenMP.
12621 * gimplify.c (gimplify_scan_omp_clauses): For OpenMP map clauses
12622 with OMP_CLAUSE_MAP_IN_REDUCTION flag partially defer gimplification
12623 of non-decl OMP_CLAUSE_DECL. For OMP_CLAUSE_IN_REDUCTION on
12624 OMP_TARGET user outer_ctx instead of ctx for placeholders and
12625 initializer/combiner gimplification.
12626 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_MAP_IN_REDUCTION
12627 on target constructs.
12628 (lower_rec_input_clauses): Likewise.
12629 (lower_omp_target): Likewise.
12630 * omp-expand.c (expand_omp_target): Temporarily ignore nowait clause
12631 on target if in_reduction is present.
12633 2021-06-24 Kewen Lin <linkw@linux.ibm.com>
12635 * tree-predcom.c (class pcom_worker): New class.
12636 (release_chain): Renamed to...
12637 (pcom_worker::release_chain): ...this.
12638 (release_chains): Renamed to...
12639 (pcom_worker::release_chains): ...this.
12640 (aff_combination_dr_offset): Renamed to...
12641 (pcom_worker::aff_combination_dr_offset): ...this.
12642 (determine_offset): Renamed to...
12643 (pcom_worker::determine_offset): ...this.
12644 (class comp_ptrs): New class.
12645 (split_data_refs_to_components): Renamed to...
12646 (pcom_worker::split_data_refs_to_components): ...this,
12647 and update with class comp_ptrs.
12648 (suitable_component_p): Renamed to...
12649 (pcom_worker::suitable_component_p): ...this.
12650 (filter_suitable_components): Renamed to...
12651 (pcom_worker::filter_suitable_components): ...this.
12652 (valid_initializer_p): Renamed to...
12653 (pcom_worker::valid_initializer_p): ...this.
12654 (find_looparound_phi): Renamed to...
12655 (pcom_worker::find_looparound_phi): ...this.
12656 (add_looparound_copies): Renamed to...
12657 (pcom_worker::add_looparound_copies): ...this.
12658 (determine_roots_comp): Renamed to...
12659 (pcom_worker::determine_roots_comp): ...this.
12660 (determine_roots): Renamed to...
12661 (pcom_worker::determine_roots): ...this.
12662 (single_nonlooparound_use): Renamed to...
12663 (pcom_worker::single_nonlooparound_use): ...this.
12664 (remove_stmt): Renamed to...
12665 (pcom_worker::remove_stmt): ...this.
12666 (execute_pred_commoning_chain): Renamed to...
12667 (pcom_worker::execute_pred_commoning_chain): ...this.
12668 (execute_pred_commoning): Renamed to...
12669 (pcom_worker::execute_pred_commoning): ...this.
12670 (struct epcc_data): New member worker.
12671 (execute_pred_commoning_cbck): Call execute_pred_commoning
12672 with pcom_worker pointer.
12673 (find_use_stmt): Renamed to...
12674 (pcom_worker::find_use_stmt): ...this.
12675 (find_associative_operation_root): Renamed to...
12676 (pcom_worker::find_associative_operation_root): ...this.
12677 (find_common_use_stmt): Renamed to...
12678 (pcom_worker::find_common_use_stmt): ...this.
12679 (combinable_refs_p): Renamed to...
12680 (pcom_worker::combinable_refs_p): ...this.
12681 (reassociate_to_the_same_stmt): Renamed to...
12682 (pcom_worker::reassociate_to_the_same_stmt): ...this.
12683 (stmt_combining_refs): Renamed to...
12684 (pcom_worker::stmt_combining_refs): ...this.
12685 (combine_chains): Renamed to...
12686 (pcom_worker::combine_chains): ...this.
12687 (try_combine_chains): Renamed to...
12688 (pcom_worker::try_combine_chains): ...this.
12689 (prepare_initializers_chain): Renamed to...
12690 (pcom_worker::prepare_initializers_chain): ...this.
12691 (prepare_initializers): Renamed to...
12692 (pcom_worker::prepare_initializers): ...this.
12693 (prepare_finalizers_chain): Renamed to...
12694 (pcom_worker::prepare_finalizers_chain): ...this.
12695 (prepare_finalizers): Renamed to...
12696 (pcom_worker::prepare_finalizers): ...this.
12697 (tree_predictive_commoning_loop): Renamed to...
12698 (pcom_worker::tree_predictive_commoning_loop): ...this, adjust
12699 some calls and remove some cleanup code.
12700 (tree_predictive_commoning): Adjusted to use pcom_worker instance.
12701 (static variable looparound_phis): Remove.
12702 (static variable name_expansions): Remove.
12704 2021-06-24 Richard Biener <rguenther@suse.de>
12706 * tree-vect-slp.c (slpg_vertex): New struct.
12707 (vect_slp_build_vertices): Adjust.
12708 (vect_optimize_slp): Likewise. Maintain an outgoing permute
12709 and a materialized one.
12711 2021-06-24 Richard Biener <rguenther@suse.de>
12713 PR tree-optimization/101105
12714 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
12715 Only ignore steps when they are equal or scalar order is preserved.
12717 2021-06-24 liuhongt <hongtao.liu@intel.com>
12720 * config/i386/i386-expand.c (ix86_expand_vec_interleave):
12721 Adjust comments for ix86_expand_vecop_qihi2.
12722 (ix86_expand_vecmul_qihi): Renamed to ..
12723 (ix86_expand_vecop_qihi2): Adjust function prototype to
12724 support shift operation, add static to definition.
12725 (ix86_expand_vec_shift_qihi_constant): Add static to definition.
12726 (ix86_expand_vecop_qihi): Call ix86_expand_vecop_qihi2 and
12727 ix86_expand_vec_shift_qihi_constant.
12728 * config/i386/i386-protos.h (ix86_expand_vecmul_qihi): Deleted.
12729 (ix86_expand_vec_shift_qihi_constant): Deleted.
12730 * config/i386/sse.md (VI12_256_512_AVX512VL): New mode
12732 (mulv8qi3): Call ix86_expand_vecop_qihi directly, add
12733 condition TARGET_64BIT.
12734 (mul<mode>3): Ditto.
12735 (<insn><mode>3): Ditto.
12736 (vlshr<mode>3): Extend to support avx512 vlshr.
12737 (v<insn><mode>3): New expander for
12739 (v<insn>v8qi3): Ditto.
12740 (vashrv8hi3<mask_name>): Renamed to ..
12741 (vashr<mode>3): And extend to support V16QImode for avx512.
12742 (vashrv16qi3): Deleted.
12743 (vashrv2di3<mask_name>): Extend expander to support avx512
12746 2021-06-23 Dimitar Dimitrov <dimitar@dinux.eu>
12748 * doc/lto.texi (Design Overview): Update that slim objects are
12751 2021-06-23 Aaron Sawdey <acsawdey@linux.ibm.com>
12753 * config/rs6000/rs6000-cpus.def: Take OPTION_MASK_PCREL_OPT out
12754 of OTHER_POWER10_MASKS so it will not be enabled by default.
12756 2021-06-23 Richard Biener <rguenther@suse.de>
12757 Martin Jambor <mjambor@suse.cz>
12759 * tree-inline.c (setup_one_parameter): Set TREE_READONLY of the
12760 param replacement unconditionally. Adjust comment.
12762 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
12764 * Makefile.in (OBJS): Add gimple-range-fold.o
12765 * gimple-range-fold.cc: New.
12766 * gimple-range-fold.h: New.
12767 * gimple-range-gori.cc (gimple_range_calc_op1): Move to here.
12768 (gimple_range_calc_op2): Ditto.
12769 * gimple-range-gori.h: Move prototypes to here.
12770 * gimple-range.cc: Adjust include files.
12771 (fur_source:fur_source): Relocate to gimple-range-fold.cc.
12772 (fur_source::get_operand): Ditto.
12773 (fur_source::get_phi_operand): Ditto.
12774 (fur_source::query_relation): Ditto.
12775 (fur_source::register_relation): Ditto.
12776 (class fur_edge): Ditto.
12777 (fur_edge::fur_edge): Ditto.
12778 (fur_edge::get_operand): Ditto.
12779 (fur_edge::get_phi_operand): Ditto.
12780 (fur_stmt::fur_stmt): Ditto.
12781 (fur_stmt::get_operand): Ditto.
12782 (fur_stmt::get_phi_operand): Ditto.
12783 (fur_stmt::query_relation): Ditto.
12784 (class fur_depend): Relocate to gimple-range-fold.h.
12785 (fur_depend::fur_depend): Relocate to gimple-range-fold.cc.
12786 (fur_depend::register_relation): Ditto.
12787 (fur_depend::register_relation): Ditto.
12788 (class fur_list): Ditto.
12789 (fur_list::fur_list): Ditto.
12790 (fur_list::get_operand): Ditto.
12791 (fur_list::get_phi_operand): Ditto.
12792 (fold_range): Ditto.
12793 (adjust_pointer_diff_expr): Ditto.
12794 (gimple_range_adjustment): Ditto.
12795 (gimple_range_base_of_assignment): Ditto.
12796 (gimple_range_operand1): Ditto.
12797 (gimple_range_operand2): Ditto.
12798 (gimple_range_calc_op1): Relocate to gimple-range-gori.cc.
12799 (gimple_range_calc_op2): Ditto.
12800 (fold_using_range::fold_stmt): Relocate to gimple-range-fold.cc.
12801 (fold_using_range::range_of_range_op): Ditto.
12802 (fold_using_range::range_of_address): Ditto.
12803 (fold_using_range::range_of_phi): Ditto.
12804 (fold_using_range::range_of_call): Ditto.
12805 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
12806 (fold_using_range::range_of_builtin_call): Ditto.
12807 (fold_using_range::range_of_cond_expr): Ditto.
12808 (fold_using_range::range_of_ssa_name_with_loop_info): Ditto.
12809 (fold_using_range::relation_fold_and_or): Ditto.
12810 (fold_using_range::postfold_gcond_edges): Ditto.
12811 * gimple-range.h: Add gimple-range-fold.h to include files. Change
12812 GIMPLE_RANGE_STMT_H to GIMPLE_RANGE_H.
12813 (gimple_range_handler): Relocate to gimple-range-fold.h.
12814 (gimple_range_ssa_p): Ditto.
12815 (range_compatible_p): Ditto.
12816 (class fur_source): Ditto.
12817 (class fur_stmt): Ditto.
12818 (class fold_using_range): Ditto.
12819 (gimple_range_calc_op1): Relocate to gimple-range-gori.h
12820 (gimple_range_calc_op2): Ditto.
12822 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
12824 PR tree-optimization/101148
12825 PR tree-optimization/101014
12826 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
12827 (ranger_cache::~ranger_cache): Adjust.
12828 (ranger_cache::block_range): Check if propagation disallowed.
12829 (ranger_cache::propagate_cache): Disallow propagation if new value
12830 can't be stored properly.
12831 * gimple-range-cache.h (ranger_cache::m_propfail): New member.
12833 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
12835 * gimple-range-cache.cc (class ssa_block_ranges): Adjust prototype.
12836 (sbr_vector::set_bb_range): Return true.
12837 (class sbr_sparse_bitmap): Adjust.
12838 (sbr_sparse_bitmap::set_bb_range): Return value.
12839 (block_range_cache::set_bb_range): Return value.
12840 (ranger_cache::propagate_cache): Use return value to print msg.
12841 * gimple-range-cache.h (class block_range_cache): Adjust.
12843 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
12845 * gimple-range.cc (dump_bb): Use range_on_edge from the cache.
12847 2021-06-23 Jeff Law <jeffreyalaw@gmail.com>
12849 * config/h8300/logical.md (<code><mode>3<ccnz>): Use <cczn>
12850 so this pattern can be used for test/compare removal. Pass
12851 current insn to compute_logical_op_length and output_logical_op.
12852 * config/h8300/h8300.c (compute_logical_op_cc): Remove.
12853 (h8300_and_costs): Add argument to compute_logical_op_length.
12854 (output_logical_op): Add new argument. Use it to determine if the
12855 condition codes are used and adjust the output accordingly.
12856 (compute_logical_op_length): Add new argument and update length
12857 computations when condition codes are used.
12858 * config/h8300/h8300-protos.h (compute_logical_op_length): Update
12860 (output_logical_op): Likewise.
12862 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
12865 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
12866 Handle 64bit modes for TARGET_XOP. Use indirect gen_* functions.
12867 * config/i386/mmx.md (mmx_ppermv64): New insn pattern.
12868 * config/i386/i386.md (unspec): Move UNSPEC_XOP_PERMUTE from ...
12869 * config/i386/sse.md (unspec): ... here.
12871 2021-06-23 Martin Liska <mliska@suse.cz>
12874 * optc-save-gen.awk: Put back arm_fp16_format to
12877 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
12880 * config/i386/i386.md (bsr_rex64): Add zero-flag setting RTX.
12883 (clz<mode>2): Update RTX pattern for additions.
12885 2021-06-23 Jakub Jelinek <jakub@redhat.com>
12887 PR middle-end/101167
12888 * omp-low.c (lower_omp_regimplify_p): Regimplify also PARM_DECLs
12889 and RESULT_DECLs that have DECL_HAS_VALUE_EXPR_P set.
12891 2021-06-22 Sergei Trofimovich <siarheit@google.com>
12893 * doc/rtl.texi: drop unbalanced parenthesis.
12895 2021-06-22 Richard Biener <rguenther@suse.de>
12897 PR middle-end/101156
12898 * gimplify.c (gimplify_expr): Remove premature incorrect
12901 2021-06-22 Jakub Jelinek <jakub@redhat.com>
12903 PR tree-optimization/101159
12904 * tree-vect-patterns.c (vect_recog_popcount_pattern): Fix some
12907 2021-06-22 Jakub Jelinek <jakub@redhat.com>
12909 PR middle-end/101160
12910 * function.c (assign_parms): For decl_result with TYPE_EMPTY_P type
12911 clear crtl->return_rtx instead of keeping it referencing a pseudo.
12913 2021-06-22 Jakub Jelinek <jakub@redhat.com>
12914 Andrew Pinski <apinski@marvell.com>
12916 PR tree-optimization/101162
12917 * fold-const.c (range_check_type): Handle OFFSET_TYPE like pointer
12920 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12922 * range-op.cc (range_relational_tests): New.
12923 (range_op_tests): Call range_relational_tests.
12925 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12927 * range-op.cc (operator_cast::lhs_op1_relation): New.
12928 (operator_identity::lhs_op1_relation): Mew.
12930 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12932 * range-op.cc (operator_minus::op1_op2_relation_effect): New.
12934 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12936 * range-op.cc (operator_plus::lhs_op1_relation): New.
12937 (operator_plus::lhs_op2_relation): New.
12939 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12941 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a
12942 relation_oracle if dominators exist.
12943 (ranger_cache::~ranger_cache): Dispose of oracle.
12944 (ranger_cache::dump_bb): Dump oracle.
12945 * gimple-range.cc (fur_source::fur_source): New.
12946 (fur_source::get_operand): Use mmeber query.
12947 (fur_source::get_phi_operand): Use member_query.
12948 (fur_source::query_relation): New.
12949 (fur_source::register_dependency): Delete.
12950 (fur_source::register_relation): New.
12951 (fur_edge::fur_edge): Adjust.
12952 (fur_edge::get_phi_operand): Fix comment.
12953 (fur_edge::query): Delete.
12954 (fur_stmt::fur_stmt): Adjust.
12955 (fur_stmt::query): Delete.
12956 (fur_depend::fur_depend): Adjust.
12957 (fur_depend::register_relation): New.
12958 (fur_depend::register_relation): New.
12959 (fur_list::fur_list): Adjust.
12960 (fur_list::get_operand): Use member query.
12961 (fold_using_range::range_of_range_op): Process and query relations.
12962 (fold_using_range::range_of_address): Adjust dependency call.
12963 (fold_using_range::range_of_phi): Ditto.
12964 (gimple_ranger::gimple_ranger): New. Use ranger_ache oracle.
12965 (fold_using_range::relation_fold_and_or): New.
12966 (fold_using_range::postfold_gcond_edges): New.
12967 * gimple-range.h (class gimple_ranger): Adjust.
12968 (class fur_source): Adjust members.
12969 (class fur_stmt): Ditto.
12970 (class fold_using_range): Ditto.
12972 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12974 * range-op.cc (range_operator::wi_fold): Apply relation effect.
12975 (range_operator::fold_range): Adjust and apply relation effect.
12976 (*::fold_range): Add relation parameters.
12977 (*::op1_range): Ditto.
12978 (*::op2_range): Ditto.
12979 (range_operator::lhs_op1_relation): New.
12980 (range_operator::lhs_op2_relation): New.
12981 (range_operator::op1_op2_relation): New.
12982 (range_operator::op1_op2_relation_effect): New.
12983 (relop_early_resolve): New.
12984 (operator_equal::op1_op2_relation): New.
12985 (operator_equal::fold_range): Call relop_early_resolve.
12986 (operator_not_equal::op1_op2_relation): New.
12987 (operator_not_equal::fold_range): Call relop_early_resolve.
12988 (operator_lt::op1_op2_relation): New.
12989 (operator_lt::fold_range): Call relop_early_resolve.
12990 (operator_le::op1_op2_relation): New.
12991 (operator_le::fold_range): Call relop_early_resolve.
12992 (operator_gt::op1_op2_relation): New.
12993 (operator_gt::fold_range): Call relop_early_resolve.
12994 (operator_ge::op1_op2_relation): New.
12995 (operator_ge::fold_range): Call relop_early_resolve.
12996 * range-op.h (class range_operator): Adjust parameters and methods.
12998 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
13000 * Makefile.in (OBJS): Add value-relation.o.
13001 * gimple-range.h: Adjust include files.
13002 * tree-data-ref.c: Adjust include file order.
13003 * value-query.cc (range_query::get_value_range): Default to no oracle.
13004 (range_query::query_relation): New.
13005 (range_query::query_relation): New.
13006 * value-query.h (class range_query): Adjust.
13007 * value-relation.cc: New.
13008 * value-relation.h: New.
13010 2021-06-22 Richard Biener <rguenther@suse.de>
13012 PR tree-optimization/101151
13013 * tree-ssa-sink.c (statement_sink_location): Expand irreducible
13016 2021-06-22 Jojo R <rjiejie@linux.alibaba.com>
13018 * config/riscv/riscv.c (thead_c906_tune_info): New.
13019 (riscv_tune_info_table): Use new tune.
13021 2021-06-22 Richard Biener <rguenther@suse.de>
13023 PR tree-optimization/101158
13024 * tree-vect-slp.c (vect_build_slp_tree_1): Move same operand
13025 checking after checking for matching operation.
13027 2021-06-22 Richard Biener <rguenther@suse.de>
13029 PR tree-optimization/101159
13030 * tree-vect-patterns.c (vect_recog_popcount_pattern): Add
13031 missing NULL vectype check.
13033 2021-06-22 Richard Biener <rguenther@suse.de>
13035 PR tree-optimization/101154
13036 * tree-vect-slp.c (vect_build_slp_tree_2): Fix out-of-bound access.
13038 2021-06-22 Jakub Jelinek <jakub@redhat.com>
13041 * config/i386/i386-protos.h (ix86_last_zero_store_uid): Declare.
13042 * config/i386/i386-expand.c (ix86_last_zero_store_uid): New variable.
13043 * config/i386/i386.c (ix86_expand_prologue): Clear it.
13044 * config/i386/i386.md (peephole2s for 1/2/4 stores of const0_rtx):
13045 Remove "" from match_operand. Emit new insns using emit_move_insn and
13046 set ix86_last_zero_store_uid to INSN_UID of the last store.
13047 Add peephole2s for 1/2/4 stores of const0_rtx following previous
13050 2021-06-22 Martin Liska <mliska@suse.cz>
13052 * auto-profile.c (AUTO_PROFILE_VERSION): Bump as string format
13055 2021-06-22 Martin Liska <mliska@suse.cz>
13057 * gcov-io.h: Remove padding entries.
13059 2021-06-22 liuhongt <hongtao.liu@intel.com>
13061 PR tree-optimization/97770
13062 * tree-vect-patterns.c (vect_recog_popcount_pattern):
13064 (vect_recog_func vect_vect_recog_func_ptrs): Add new pattern.
13066 2021-06-22 liuhongt <hongtao.liu@intel.com>
13069 * config/i386/i386-builtin.def (BDESC): Adjust builtin name.
13070 * config/i386/sse.md (<avx512>_expand<mode>_mask): Rename to ..
13071 (expand<mode>_mask): this ..
13072 (*expand<mode>_mask): New pre_reload splitter to transform
13073 v{,p}expand* to vmov* when mask is zero, all ones, or has all
13074 ones in it's lower part, otherwise still generate
13077 2021-06-22 liuhongt <hongtao.liu@intel.com>
13080 * config/i386/i386-expand.c
13081 (ix86_expand_special_args_builtin): Keep constm1_operand only
13082 if it satisfies insn's operand predicate.
13084 2021-06-21 Jason Merrill <jason@redhat.com>
13087 * df-scan.c (df_ref_record): Check that regno < endregno.
13088 * function.c (assign_parms, expand_function_end): Do nothing with a
13089 TYPE_EMPTY_P result.
13091 2021-06-21 Richard Biener <rguenther@suse.de>
13093 PR tree-optimization/101120
13094 * tree-vect-data-refs.c (bump_vector_ptr): Fold the
13096 * tree-vect-slp.c (vect_transform_slp_perm_load): Add
13097 DR chain DCE capability.
13098 * tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
13099 * tree-vect-stmts.c (vectorizable_load): Remove unused
13100 loads in the DR chain for SLP.
13102 2021-06-21 Jakub Jelinek <jakub@redhat.com>
13104 PR inline-asm/100785
13105 * gimplify.c (gimplify_asm_expr): Don't diagnose errors if
13106 output or input operands were already error_mark_node.
13107 * cfgexpand.c (expand_asm_stmt): If errors are emitted,
13108 remove all inputs, outputs and clobbers from the asm and
13109 set template to "".
13111 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
13113 * config/arm/arm_neon.h (vceq_s8): Replace builtin with __a == __b.
13114 (vceq_s16): Likewise.
13115 (vceq_s32): Likewise.
13116 (vceq_u8): Likewise.
13117 (vceq_u16): Likewise.
13118 (vceq_u32): Likewise.
13119 (vceq_p8): Likewise.
13120 (vceqq_s8): Likewise.
13121 (vceqq_s16): Likewise.
13122 (vceqq_s32): Likewise.
13123 (vceqq_u8): Likewise.
13124 (vceqq_u16): Likewise.
13125 (vceqq_u32): Likewise.
13126 (vceqq_p8): Likewise.
13127 (vceq_f32): Gate __a == __b on __FAST_MATH__.
13128 (vceqq_f32): Likewise.
13129 (vceq_f16): Likewise.
13130 (vceqq_f16): Likewise.
13132 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
13135 * config/arm/iterators.md (NEON_VACMP): Remove.
13136 * config/arm/neon.md (neon_vca<cmp_op><mode>): Use GLTE instead of GTGE
13138 (neon_vca<cmp_op><mode>_insn): Likewise.
13139 (neon_vca<cmp_op_unsp><mode>_insn_unspec): Use NEON_VAGLTE instead of
13142 2021-06-21 Richard Biener <rguenther@suse.de>
13144 PR tree-optimization/101121
13145 * tree-vect-slp.c (vect_build_slp_tree_2): To not fail fatally
13146 when we just lack a stmt with the desired op when doing permutation.
13147 (vect_build_slp_tree): When caching a failed SLP build attempt
13148 assert that at least one lane is marked as not matching.
13150 2021-06-21 liuhongt <hongtao.liu@intel.com>
13153 * config/i386/i386.md: (*anddi_1): Disparage slightly the mask
13154 register alternative.
13155 (*and<mode>_1): Ditto.
13157 (*andn<mode>_1): Ditto.
13158 (*<code><mode>_1): Ditto.
13159 (*<code>qi_1): Ditto.
13160 (*one_cmpl<mode>2_1): Ditto.
13161 (*one_cmplsi2_1_zext): Ditto.
13162 (*one_cmplqi2_1): Ditto.
13163 * config/i386/i386.c (x86_order_regs_for_local_alloc): Change
13164 the order of mask registers to be before general registers.
13166 2021-06-21 Roger Sayle <roger@nextmovesoftware.com>
13169 * config/i386/i386.md: New define_peephole2s to shrink writing
13170 1, 2 or 4 consecutive zeros to memory when optimizing for size.
13172 2021-06-18 Jeff Law <jeffreyalaw@gmail.com>
13174 * config/h8300/h8300.c (h8300_select_cc_mode): Handle SYMBOL_REF.
13175 * config/h8300/logical.md (<code><mode>3 logcial expander): Generate
13176 more efficient code when the source can be trivially simplified.
13178 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
13180 * gimple-range-cache.cc (ranger_cache::range_of_def): Calculate
13181 a range if global is not available.
13182 (ranger_cache::entry_range): Fallback to range_of_def.
13183 * gimple-range-cache.h (range_of_def): Adjust prototype.
13185 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
13187 PR tree-optimization/101014
13188 * gimple-range-cache.cc (ranger_cache::ranger_cache): Remove poor
13190 (ranger_cache::~ranger_cache): Ditto.
13191 (ranger_cache::enable_new_values): Delete.
13192 (ranger_cache::push_poor_value): Delete.
13193 (ranger_cache::range_of_def): Remove poor value processing.
13194 (ranger_cache::entry_range): Ditto.
13195 (ranger_cache::fill_block_cache): Ditto.
13196 * gimple-range-cache.h (class ranger_cache): Remove poor value members.
13197 * gimple-range.cc (gimple_ranger::range_of_expr): Remove call.
13198 * gimple-range.h (class gimple_ranger): Adjust.
13200 2021-06-18 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
13203 * common/config/arm/arm-common.c (arm_canon_arch_option_1): New function
13204 derived from arm_canon_arch.
13205 (arm_canon_arch_option): Call it.
13206 (arm_canon_arch_multilib_option): New function.
13207 * config/arm/arm-cpus.in (IGNORE_FOR_MULTILIB): New fgroup.
13208 * config/arm/arm.h (arm_canon_arch_multilib_option): New prototype.
13209 (CANON_ARCH_MULTILIB_SPEC_FUNCTION): New macro.
13210 (MULTILIB_ARCH_CANONICAL_SPECS): New macro.
13211 (DRIVER_SELF_SPECS): Add MULTILIB_ARCH_CANONICAL_SPECS.
13212 * config/arm/arm.opt (mlibarch): New option.
13213 * config/arm/t-rmprofile (MULTILIB_MATCHES): For armv8*-m, replace use
13214 of march on RHS with mlibarch.
13216 2021-06-18 Marcel Vollweiler <marcel@codesourcery.com>
13218 * config.in: Regenerate.
13219 * config/gcn/gcn.c (print_operand_address): Fix for global_load assembler
13221 * configure: Regenerate.
13222 * configure.ac: Fix for global_load assembler functions.
13224 2021-06-18 Richard Biener <rguenther@suse.de>
13226 PR tree-optimization/101112
13227 * tree-vect-slp.c (vect_slp_linearize_chain): Fix condition
13228 to lookup a pattern stmt def.
13230 2021-06-18 Jakub Jelinek <jakub@redhat.com>
13232 PR middle-end/101062
13233 * stor-layout.c (finish_bitfield_layout): Don't add bitfield
13234 representatives in QUAL_UNION_TYPE.
13236 2021-06-18 Andrew Pinski <apinski@marvell.com>
13238 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
13239 Add counting of how many times it is done.
13240 (factor_out_conditional_conversion): Likewise.
13241 (match_simplify_replacement): Likewise.
13242 (value_replacement): Likewise.
13243 (spaceship_replacement): Likewise.
13244 (cond_store_replacement): Likewise.
13245 (cond_if_else_store_replacement_1): Likewise.
13246 (hoist_adjacent_loads): Likewise.
13248 2021-06-18 Andrew Pinski <apinski@marvell.com>
13250 * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
13251 types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and CNONJ_EXPR.
13252 (verify_gimple_assign_binary): Reject point and offset types on
13253 MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
13254 FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
13255 FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.
13257 2021-06-18 Michael Meissner <meissner@linux.ibm.com>
13259 * config/rs6000/rs6000.c (rs6000_emit_minmax): Add support for ISA
13260 3.1 IEEE 128-bit floating point xsmaxcqp/xsmincqp instructions.
13261 * config/rs6000/rs6000.md (s<minmax><mode>3, IEEE128 iterator):
13264 2021-06-17 Aaron Sawdey <acsawdey@linux.ibm.com>
13266 * config/rs6000/genfusion.pl (gen_logical_addsubf): Add
13267 earlyclobber to alts 0/1.
13268 (gen_addadd): Add earlyclobber to alts 0/1.
13269 * config/rs6000/fusion.md: Regenerate file.
13271 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13273 * cfgloopanal.c (get_loop_hot_path): Make path an auto_vec.
13275 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
13277 * gimple-range-cache.cc: Comment cleanups.
13278 * gimple-range-gori.cc: Comment cleanups.
13279 * gimple-range.cc: Comment/spacing cleanups
13280 * value-range.h: Comment cleanups.
13282 2021-06-17 H.J. Lu <hjl.tools@gmail.com>
13285 * calls.c (expand_call): Replace PUSH_ARGS with
13286 targetm.calls.push_argument (0).
13287 (emit_library_call_value_1): Likewise.
13288 * defaults.h (PUSH_ARGS): Removed.
13289 (PUSH_ARGS_REVERSED): Replace PUSH_ARGS with
13290 targetm.calls.push_argument (0).
13291 * expr.c (block_move_libcall_safe_for_call_parm): Likewise.
13292 (emit_push_insn): Pass the number bytes to push to
13293 targetm.calls.push_argument and pass 0 if ARGS_ADDR is 0.
13294 * hooks.c (hook_bool_uint_true): New.
13295 * hooks.h (hook_bool_uint_true): Likewise.
13296 * rtlanal.c (nonzero_bits1): Replace PUSH_ARGS with
13297 targetm.calls.push_argument (0).
13298 * target.def (push_argument): Add a targetm.calls hook.
13299 * targhooks.c (default_push_argument): New.
13300 * targhooks.h (default_push_argument): Likewise.
13301 * config/bpf/bpf.h (PUSH_ARGS): Removed.
13302 * config/cr16/cr16.c (TARGET_PUSH_ARGUMENT): New.
13303 * config/cr16/cr16.h (PUSH_ARGS): Removed.
13304 * config/i386/i386.c (ix86_push_argument): New.
13305 (TARGET_PUSH_ARGUMENT): Likewise.
13306 * config/i386/i386.h (PUSH_ARGS): Removed.
13307 * config/m32c/m32c.c (TARGET_PUSH_ARGUMENT): New.
13308 * config/m32c/m32c.h (PUSH_ARGS): Removed.
13309 * config/nios2/nios2.h (PUSH_ARGS): Likewise.
13310 * config/pru/pru.h (PUSH_ARGS): Likewise.
13311 * doc/tm.texi.in: Remove PUSH_ARGS documentation. Add
13312 TARGET_PUSH_ARGUMENT hook.
13313 * doc/tm.texi: Regenerated.
13315 2021-06-17 Uroš Bizjak <ubizjak@gmail.com>
13318 * config/i386/i386-expand.c (expand_vector_set_var):
13319 Handle V2FS mode remapping. Pass TARGET_MMX_WITH_SSE to
13320 ix86_expand_vector_init_duplicate.
13321 (ix86_expand_vector_init_duplicate): Emit insv_1 for
13322 QImode for !TARGET_PARTIAL_REG_STALL.
13323 * config/i386/predicates.md (vec_setm_mmx_operand): New predicate.
13324 * config/i386/mmx.md (vec_setv2sf): Use vec_setm_mmx_operand
13325 as operand 2 predicate. Call ix86_expand_vector_set_var
13326 for non-constant index operand.
13327 (vec_setv2si): Ditto.
13328 (vec_setv4hi): Ditto.
13329 (vec_setv8qi): ditto.
13331 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
13333 PR tree-optimization/100790
13334 * gimple-range.cc (range_of_builtin_call): Cleanup clz and ctz
13337 2021-06-17 Martin Liska <mliska@suse.cz>
13339 * doc/invoke.texi: Use consistently -O1 instead of -O.
13341 2021-06-17 Martin Liska <mliska@suse.cz>
13343 * gcov-io.h: Update documentation entry about string format.
13345 2021-06-17 Marius Hillenbrand <mhillen@linux.ibm.com>
13348 * config/s390/vecintrin.h (vec_doublee): Fix to use
13349 __builtin_s390_vflls.
13350 (vec_floate): Fix to use __builtin_s390_vflrd.
13352 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13354 * dominance.c (get_dominated_to_depth): Return auto_vec<basic_block>.
13355 * dominance.h (get_dominated_to_depth): Likewise.
13356 (get_all_dominated_blocks): Likewise.
13357 * cfgcleanup.c (delete_unreachable_blocks): Adjust.
13358 * gcse.c (hoist_code): Likewise.
13359 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
13360 * tree-parloops.c (oacc_entry_exit_ok): Likewise.
13361 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Likewise.
13362 * tree-ssa-phiprop.c (pass_phiprop::execute): Likewise.
13364 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13366 * dominance.c (get_dominated_by_region): Return auto_vec<basic_block>.
13367 * dominance.h (get_dominated_by_region): Likewise.
13368 * tree-cfg.c (gimple_duplicate_sese_region): Adjust.
13369 (gimple_duplicate_sese_tail): Likewise.
13370 (move_sese_region_to_fn): Likewise.
13372 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13374 * dominance.c (get_dominated_by): Return auto_vec<basic_block>.
13375 * dominance.h (get_dominated_by): Likewise.
13376 * auto-profile.c (afdo_find_equiv_class): Adjust.
13377 * cfgloopmanip.c (duplicate_loop_to_header_edge): Likewise.
13378 * loop-unroll.c (unroll_loop_runtime_iterations): Likewise.
13379 * tree-cfg.c (test_linear_chain): Likewise.
13380 (test_diamond): Likewise.
13382 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13384 * cfgloop.h (get_loop_hot_path): Return auto_vec<basic_block>.
13385 * cfgloopanal.c (get_loop_hot_path): Likewise.
13386 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
13388 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13390 * cgraph.c (cgraph_node::collect_callers): Return
13391 auto_vec<cgraph_edge *>.
13392 * cgraph.h (cgraph_node::collect_callers): Likewise.
13393 * ipa-cp.c (create_specialized_node): Adjust.
13394 (decide_about_value): Likewise.
13395 (decide_whether_version_node): Likewise.
13396 * ipa-sra.c (process_isra_node_results): Likewise.
13398 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13400 * vec.h (vl_ptr>::using_auto_storage): Handle null m_vec.
13401 (auto_vec<T, 0>::auto_vec): Define move constructor, and delete copy
13403 (auto_vec<T, 0>::operator=): Define move assignment and delete copy
13406 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
13408 * gimple-range.cc (debug_seed_ranger): New.
13409 (dump_ranger): New.
13410 (debug_ranger): New.
13412 2021-06-17 Richard Biener <rguenther@suse.de>
13414 PR tree-optimization/54400
13415 * tree-vectorizer.h (enum slp_instance_kind): Add
13416 slp_inst_kind_bb_reduc.
13417 (reduction_fn_for_scalar_code): Declare.
13418 * tree-vect-data-refs.c (vect_slp_analyze_instance_dependence):
13419 Check SLP_INSTANCE_KIND instead of looking at the
13421 (vect_slp_analyze_instance_alignment): Likewise.
13422 * tree-vect-loop.c (reduction_fn_for_scalar_code): Export.
13423 * tree-vect-slp.c (vect_slp_linearize_chain): Split out
13424 chain linearization from vect_build_slp_tree_2 and generalize
13425 for the use of BB reduction vectorization.
13426 (vect_build_slp_tree_2): Adjust accordingly.
13427 (vect_optimize_slp): Elide permutes at the root of BB reduction
13429 (vectorizable_bb_reduc_epilogue): New function.
13430 (vect_slp_prune_covered_roots): Likewise.
13431 (vect_slp_analyze_operations): Use them.
13432 (vect_slp_check_for_constructors): Recognize associatable
13433 chains for BB reduction vectorization.
13434 (vectorize_slp_instance_root_stmt): Generate code for the
13435 BB reduction epilogue.
13437 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
13439 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Check with
13441 (gori_compute::may_recompute_p): New.
13442 (gori_compute::outgoing_edge_range_p): Perform recomputations.
13443 * gimple-range-gori.h (class gori_compute): Add prototype.
13445 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
13447 * gimple-range-cache.cc (ranger_cache::range_on_edge): Always return
13448 true when a range can be calculated.
13449 * gimple-range.cc (gimple_ranger::dump_bb): Check has_edge_range_p.
13451 2021-06-16 Martin Sebor <msebor@redhat.com>
13453 * doc/invoke.texi (-Wmismatched-dealloc, -Wmismatched-new-delete):
13454 Correct documented defaults.
13456 2021-06-16 Andrew MacLeod <amacleod@redhat.com>
13458 * gimple-range-cache.cc (ranger_cache::ranger_cache): Initialize
13459 m_new_value_p directly.
13461 2021-06-16 Uroš Bizjak <ubizjak@gmail.com>
13464 * config/i386/i386-expand.c (expand_vec_perm_2perm_pblendv):
13465 Handle 64bit modes for TARGET_SSE4_1.
13466 (expand_vec_perm_pshufb2): Handle 64bit modes for TARGET_SSSE3.
13467 (expand_vec_perm_even_odd_pack): Handle V4HI mode.
13468 (expand_vec_perm_even_odd_1) <case E_V4HImode>: Expand via
13469 expand_vec_perm_pshufb2 for TARGET_SSSE3 and via
13470 expand_vec_perm_even_odd_pack for TARGET_SSE4_1.
13471 * config/i386/mmx.md (mmx_packusdw): New insn pattern.
13473 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
13475 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn<mode>):
13476 Change to an expander that emits the correct instruction
13477 depending on endianness.
13478 (aarch64_<sur><addsub>hn<mode>_insn_le): Define.
13479 (aarch64_<sur><addsub>hn<mode>_insn_be): Define.
13481 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
13483 * config/aarch64/aarch64-simd-builtins.def: Split generator
13484 for aarch64_<su>qmovn builtins into scalar and vector
13486 * config/aarch64/aarch64-simd.md (aarch64_<su>qmovn<mode>_insn_le):
13488 (aarch64_<su>qmovn<mode>_insn_be): Define.
13489 (aarch64_<su>qmovn<mode>): Split into scalar and vector
13490 variants. Change vector variant to an expander that emits the
13491 correct instruction depending on endianness.
13493 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
13495 * config/aarch64/aarch64-simd-builtins.def: Split generator
13496 for aarch64_sqmovun builtins into scalar and vector variants.
13497 * config/aarch64/aarch64-simd.md (aarch64_sqmovun<mode>):
13498 Split into scalar and vector variants. Change vector variant
13499 to an expander that emits the correct instruction depending
13501 (aarch64_sqmovun<mode>_insn_le): Define.
13502 (aarch64_sqmovun<mode>_insn_be): Define.
13504 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
13506 * config/aarch64/aarch64-simd.md (aarch64_xtn<mode>_insn_le):
13507 Define - modeling zero-high-half semantics.
13508 (aarch64_xtn<mode>): Change to an expander that emits the
13509 appropriate instruction depending on endianness.
13510 (aarch64_xtn<mode>_insn_be): Define - modeling zero-high-half
13512 (aarch64_xtn2<mode>_le): Rename to...
13513 (aarch64_xtn2<mode>_insn_le): This.
13514 (aarch64_xtn2<mode>_be): Rename to...
13515 (aarch64_xtn2<mode>_insn_be): This.
13516 (vec_pack_trunc_<mode>): Emit truncation instruction instead
13518 * config/aarch64/iterators.md (Vnarrowd): Add Vnarrowd mode
13519 attribute iterator.
13521 2021-06-16 Martin Jambor <mjambor@suse.cz>
13523 PR tree-optimization/100453
13524 * tree-sra.c (create_access): Disqualify any const candidates
13525 which are written to.
13526 (sra_modify_expr): Do not store sub-replacements back to a const base.
13527 (handle_unscalarized_data_in_subtree): Likewise.
13528 (sra_modify_assign): Likewise. Earlier, use TREE_READONLy test
13529 instead of constant_decl_p.
13531 2021-06-16 Jakub Jelinek <jakub@redhat.com>
13533 PR middle-end/101062
13534 * stor-layout.c (finish_bitfield_representative): For fields in unions
13535 assume nextf is always NULL.
13536 (finish_bitfield_layout): Compute bit field representatives also in
13537 unions, but handle it as if each bitfield was the only field in the
13540 2021-06-16 Richard Biener <rguenther@suse.de>
13542 PR tree-optimization/101088
13543 * tree-ssa-loop-im.c (sm_seq_valid_bb): Only look for
13544 supported refs on edges. Do not assert same ref but
13545 different kind stores are unsuported but mark them so.
13546 (hoist_memory_references): Only look for supported refs
13549 2021-06-16 Roger Sayle <roger@nextmovesoftware.com>
13551 PR rtl-optimization/46235
13552 * config/i386/i386.md: New define_split for bt followed by cmov.
13553 (*bt<mode>_setcqi): New define_insn_and_split for bt followed by setc.
13554 (*bt<mode>_setncqi): New define_insn_and_split for bt then setnc.
13555 (*bt<mode>_setnc<mode>): New define_insn_and_split for bt followed
13556 by setnc with zero extension.
13558 2021-06-16 Richard Biener <rguenther@suse.de>
13560 PR tree-optimization/101083
13561 * tree-vect-slp.c (vect_slp_build_two_operator_nodes): Get
13562 vectype as argument.
13563 (vect_build_slp_tree_2): Adjust.
13565 2021-06-15 Martin Sebor <msebor@redhat.com>
13567 PR middle-end/100876
13568 * builtins.c: (gimple_call_return_array): Account for size_t
13569 mangling as either unsigned int or unsigned long
13571 2021-06-15 Jeff Law <jeffreyalaw@gmail.com>
13573 * compare-elim.c (try_eliminate_compare): Run DCE to clean things
13574 up before eliminating comparisons.
13576 2021-06-15 Aldy Hernandez <aldyh@redhat.com>
13578 * range-op.cc (operator_bitwise_or::wi_fold): Make sure
13579 nonzero|X is nonzero.
13580 (range_op_bitwise_and_tests): Add tests for above.
13582 2021-06-15 Carl Love <cel@us.ibm.com>
13585 * config/rs6000/rs6000-builtin.def (VCMPEQUT): Fix the ICODE for the
13587 (VRLQ, VSLQ, VSRQ, VSRAQ): Remove unused BU_P10_OVERLOAD_2
13590 2021-06-15 Tobias Burnus <tobias@codesourcery.com>
13593 * gimplify.c (enum gimplify_defaultmap_kind): Add GDMK_SCALAR_TARGET.
13594 (struct gimplify_omp_ctx): Extend defaultmap array by one.
13595 (new_omp_context): Init defaultmap[GDMK_SCALAR_TARGET].
13596 (omp_notice_variable): Update type classification for Fortran.
13597 (gimplify_scan_omp_clauses): Update calls for new argument; handle
13598 GDMK_SCALAR_TARGET; for Fortran, GDMK_POINTER avoid GOVD_MAP_0LEN_ARRAY.
13599 * langhooks-def.h (lhd_omp_scalar_p): Add 'ptr_ok' argument.
13600 * langhooks.c (lhd_omp_scalar_p): Likewise.
13601 (LANG_HOOKS_OMP_ALLOCATABLE_P, LANG_HOOKS_OMP_SCALAR_TARGET_P): New.
13602 (LANG_HOOKS_DECLS): Add them.
13603 * langhooks.h (struct lang_hooks_for_decls): Add new hooks, update
13604 omp_scalar_p pointer type to include the new bool argument.
13606 2021-06-15 David Malcolm <dmalcolm@redhat.com>
13608 * doc/analyzer.texi
13609 (Special Functions for Debugging the Analyzer): Add
13610 __analyzer_dump_capacity.
13612 2021-06-15 Jakub Jelinek <jakub@redhat.com>
13615 * expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
13616 case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.
13618 2021-06-15 Richard Biener <rguenther@suse.de>
13620 * cfgloopanal.c (mark_irreducible_loops): Use a dominance
13621 check to identify loop latches.
13622 * cfgloop.c (verify_loop_structure): Likewise.
13623 * loop-init.c (apply_loop_flags): Allow marked irreducible
13624 regions even with multiple latches.
13625 * predict.c (rebuild_frequencies): Simplify.
13627 2021-06-15 Richard Biener <rguenther@suse.de>
13629 * tree-ssa-threadupdate.c
13630 (jump_thread_path_registry::mark_threaded_blocks): Assert we
13631 have marked irreducible regions.
13633 2021-06-14 Martin Sebor <msebor@redhat.com>
13636 * builtins.c (gimple_call_return_array): Check for attribute fn spec.
13637 Handle calls to placement new.
13638 (ndecl_dealloc_argno): Avoid placement delete.
13640 2021-06-14 Peter Bergner <bergner@linux.ibm.com>
13643 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
13644 create_tmp_reg_or_ssa_name().
13646 2021-06-14 Andrew MacLeod <amacleod@redhat.com>
13648 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
13649 (ranger_cache::enable_new_values): Set to specified value and
13650 return the old value.
13651 (ranger_cache::disable_new_values): Delete.
13652 (ranger_cache::fill_block_cache): Disable non 1st order derived
13654 * gimple-range-cache.h (ranger_cache): Adjust prototypes.
13655 * gimple-range.cc (gimple_ranger::range_of_expr): Adjust.
13657 2021-06-14 Uroš Bizjak <ubizjak@gmail.com>
13660 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
13661 Return true early when testing with V2HImode.
13662 * config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.
13664 2021-06-14 Christophe Lyon <christophe.lyon@linaro.org>
13666 * config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
13667 (mve_vec_unpack<US>_hi_<mode>): New pattern.
13668 (@mve_vec_pack_trunc_lo_<mode>): New pattern.
13669 (mve_vmovntq_<supf><mode>): Prefix with '@'.
13670 * config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
13672 (vec_unpack<US>_lo_<mode>): Likewise.
13673 (vec_pack_trunc_<mode>): Rename to
13674 neon_quad_vec_pack_trunc_<mode>.
13675 * config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
13677 (vec_unpack<US>_lo_<mode>): New.
13678 (vec_pack_trunc_<mode>): New.
13680 2021-06-14 Richard Biener <rguenther@suse.de>
13682 PR tree-optimization/100934
13683 * tree-ssa-dom.c (pass_dominator::execute): Properly
13684 mark irreducible regions.
13686 2021-06-14 Martin Liska <mliska@suse.cz>
13688 * doc/invoke.texi: Put r{...} on the same line as @item.
13690 2021-06-14 Martin Liska <mliska@suse.cz>
13692 * doc/invoke.texi: Add missing newline.
13694 2021-06-14 Martin Liska <mliska@suse.cz>
13696 * doc/invoke.texi: Remove '+' charasters.
13698 2021-06-14 Claudiu Zissulescu <claziss@synopsys.com>
13700 * config.gcc (arc): Add support for with_cpu option.
13701 * config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.
13703 2021-06-14 Richard Biener <rguenther@suse.de>
13705 PR tree-optimization/101031
13706 * tree-ssa-strlen.c (maybe_invalidate): Increment max_size
13707 instead of size when accounting for a possibly string
13710 2021-06-14 Martin Liska <mliska@suse.cz>
13712 * gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].
13714 2021-06-14 Aldy Hernandez <aldyh@redhat.com>
13716 * value-query.cc (gimple_range_global): Call get_range_global
13717 if called after inlining.
13719 2021-06-13 Uroš Bizjak <ubizjak@gmail.com>
13722 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
13723 Emit constant permutation insn directly from here.
13725 2021-06-13 Trevor Saunders <tbsaunde@tbsaunde.org>
13727 * attribs.c (find_attribute_namespace): Iterate over vec<> with
13729 * auto-profile.c (afdo_find_equiv_class): Likewise.
13730 * gcc.c (do_specs_vec): Likewise.
13731 (do_spec_1): Likewise.
13732 (driver::set_up_specs): Likewise.
13733 * gimple-loop-jam.c (any_access_function_variant_p): Likewise.
13734 * gimple-ssa-store-merging.c (compatible_load_p): Likewise.
13735 (imm_store_chain_info::try_coalesce_bswap): Likewise.
13736 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
13737 (get_location_for_stmts): Likewise.
13738 * graphite-poly.c (print_iteration_domains): Likewise.
13739 (free_poly_bb): Likewise.
13740 (remove_gbbs_in_scop): Likewise.
13741 (free_scop): Likewise.
13742 (dump_gbb_cases): Likewise.
13743 (dump_gbb_conditions): Likewise.
13744 (print_pdrs): Likewise.
13745 (print_scop): Likewise.
13746 * ifcvt.c (cond_move_process_if_block): Likewise.
13747 * lower-subreg.c (decompose_multiword_subregs): Likewise.
13748 * regcprop.c (pass_cprop_hardreg::execute): Likewise.
13749 * sanopt.c (sanitize_rewrite_addressable_params): Likewise.
13750 * sel-sched-dump.c (dump_insn_vector): Likewise.
13751 * store-motion.c (store_ops_ok): Likewise.
13752 (store_killed_in_insn): Likewise.
13753 * timevar.c (timer::named_items::print): Likewise.
13754 * tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
13755 (cleanup_tree_cfg_noloop): Likewise.
13756 * tree-data-ref.c (dump_data_references): Likewise.
13757 (print_dir_vectors): Likewise.
13758 (print_dist_vectors): Likewise.
13759 (dump_data_dependence_relations): Likewise.
13760 (dump_dist_dir_vectors): Likewise.
13761 (dump_ddrs): Likewise.
13762 (create_runtime_alias_checks): Likewise.
13763 (free_subscripts): Likewise.
13764 (save_dist_v): Likewise.
13765 (save_dir_v): Likewise.
13766 (invariant_access_functions): Likewise.
13767 (same_access_functions): Likewise.
13768 (access_functions_are_affine_or_constant_p): Likewise.
13769 (find_data_references_in_stmt): Likewise.
13770 (graphite_find_data_references_in_stmt): Likewise.
13771 (free_dependence_relations): Likewise.
13772 (free_data_refs): Likewise.
13773 * tree-inline.c (copy_debug_stmts): Likewise.
13774 * tree-into-ssa.c (dump_currdefs): Likewise.
13775 (rewrite_update_phi_arguments): Likewise.
13776 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
13777 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
13779 (vect_slp_analyze_node_dependences): Likewise.
13780 (vect_slp_analyze_instance_dependence): Likewise.
13781 (vect_record_base_alignments): Likewise.
13782 (vect_get_peeling_costs_all_drs): Likewise.
13783 (vect_peeling_supportable): Likewise.
13784 * tree-vectorizer.c (vec_info::~vec_info): Likewise.
13785 (vec_info::free_stmt_vec_infos): Likewise.
13787 2021-06-13 Jeff Law <jeffreyalaw@gmail.com>
13789 * config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
13790 (andqi3_1<cczn>): Removed.
13791 (<ors>qi3_1): Do not split for IOR/XOR a single bit.
13792 (H8/SX bit logicals): Split out from other patterns.
13793 * config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
13794 mulqihi3_const_clobber_flags.
13795 (mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly
13797 2021-06-13 H.J. Lu <hjl.tools@gmail.com>
13800 * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
13801 to true if red zone is used.
13802 (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
13803 ix86_red_zone_used.
13804 * config/i386/i386.h (machine_function): Add red_zone_used.
13805 (ix86_red_zone_size): Removed.
13806 (ix86_red_zone_used): New.
13807 * config/i386/i386.md (peephole2 patterns): Replace
13808 ix86_red_zone_size with ix86_red_zone_used.
13810 2021-06-12 Jason Merrill <jason@redhat.com>
13812 * doc/extend.texi (unused variable attribute): Applies to
13813 structure fields as well.
13815 2021-06-12 Eugene Rozenfeld <erozen@microsoft.com>
13817 * auto-profile.c (read_profile): fix a typo in an error string
13819 2021-06-11 Thomas Schwinge <thomas@codesourcery.com>
13821 * tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
13823 * tree-pretty-print.c (dump_omp_clauses): Update.
13824 (dump_generic_node) <OMP_CLAUSE>: Use it.
13826 2021-06-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
13829 * config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
13830 int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
13831 the polymorphic variants matching code.
13832 (__arm_vld1q_z): Likewise.
13833 (__arm_vld2q): Likewise.
13834 (__arm_vld4q): Likewise.
13835 (__arm_vldrbq_gather_offset): Likewise.
13836 (__arm_vldrbq_gather_offset_z): Likewise.
13838 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
13840 PR tree-optimization/96392
13841 * fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.
13843 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
13845 PR tree-optimization/96392
13846 * fold-const.c (fold_real_zero_addition_p): Take both arguments
13847 of the addition or subtraction, not just the zero. Use this
13848 other argument in tests for signaling NaNs and signed zeros.
13849 (tree_expr_maybe_real_minus_zero_p): New predicate.
13850 * fold-const.h (fold_real_zero_addition_p): Update prototype.
13851 (tree_expr_maybe_real_minus_zero_p): New function prototype.
13852 * match.pd: Update calls to fold_real_zero_addition_p.
13853 Replace HONOR_NANS with tree_expr_maybe_nan_p.
13854 Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
13855 Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
13856 * tree-ssa-reassoc.c (eliminate_using_constants): Update
13857 call to fold_real_zero_addition_p.
13859 2021-06-11 Richard Biener <rguenther@suse.de>
13861 PR tree-optimization/101025
13862 * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
13863 all refs that require dependence checking.
13865 2021-06-11 Richard Biener <rguenther@suse.de>
13867 PR tree-optimization/101028
13868 * tree-vect-slp.c (vect_build_slp_tree_2): When SLP
13869 reassoc discovery fails fatally, mark appropriate lanes
13872 2021-06-11 Richard Biener <rguenther@suse.de>
13874 PR tree-optimization/101026
13875 * tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
13876 have a representative for the associated chain nodes.
13878 2021-06-11 Jakub Jelinek <jakub@redhat.com>
13880 PR rtl-optimization/101008
13881 * simplify-rtx.c (relational_result): New function.
13882 (simplify_logical_relational_operation,
13883 simplify_relational_operation): Use it.
13885 2021-06-11 Jakub Jelinek <jakub@redhat.com>
13888 * config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.
13890 2021-06-11 Uroš Bizjak <ubizjak@gmail.com>
13893 * config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
13894 false if the permutation can be implemented with constant
13895 permutation instruction in wider mode.
13896 (canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
13897 Handle V8QImode and V4HImode.
13899 2021-06-11 Martin Liska <mliska@suse.cz>
13901 PR gcov-profile/100788
13902 * common.opt: Add new option.
13903 * coverage.c (coverage_begin_function): Emit warning instead on
13904 the internal compiler error.
13905 * doc/invoke.texi: Document the option.
13906 * toplev.c (process_options): Enable it by default.
13908 2021-06-11 Richard Biener <rguenther@suse.de>
13910 PR middle-end/101009
13911 * tree-data-ref.c (build_classic_dist_vector_1): Make sure
13912 to set *init_b to true when we encounter a constant equal
13914 (compute_affine_dependence): Also dump the actual DR_REF.
13916 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
13918 PR tree-optimization/100984
13919 * gimple-ssa-evrp.c (ssa_equiv_stack): Use auto_vec for
13920 replacements table.
13921 (ssa_equiv_stack::~ssa_equiv_stack): Remove.
13923 2021-06-11 Kewen Lin <linkw@linux.ibm.com>
13925 * config/rs6000/rs6000.md
13926 (floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
13927 define_insn_and_split.
13929 2021-06-11 Richard Biener <rguenther@suse.de>
13931 * tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
13932 to sort operands of the associative chain.
13934 2021-06-11 Richard Biener <rguenther@suse.de>
13936 * system.h (gcc_stablesort_r): Declare.
13937 * sort.cc (gcc_sort_r): Support stable sort.
13938 (gcc_stablesort_r): Define.
13939 * vec.h (vec<>::stablesort): Add.
13941 2021-06-10 Uroš Bizjak <ubizjak@gmail.com>
13944 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
13945 Handle V2SF mode. Emit SHUFPS to fixup unpack-high for V2SF mode.
13946 (expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
13947 (expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
13948 (expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
13949 (expand_vec_perm_interleave2): Handle 64bit modes.
13950 (expand_vec_perm_even_odd_pack): Handle V8QI mode.
13951 (expand_vec_perm_even_odd_1): Ditto.
13952 (ix86_vectorize_vec_perm_const): Ditto.
13953 * config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
13954 * config/i386/sse.md: ... here.
13955 * config/i386/mmx.md (*vec_interleave_lowv2sf):
13956 New insn_and_split pattern.
13957 (*vec_interleave_highv2sf): Ditto.
13958 (mmx_pshufbv8qi3): New insn pattern.
13959 (*mmx_pblendw): Ditto.
13961 2021-06-10 Peter Bergner <bergner@linux.ibm.com>
13963 * config/rs6000/rs6000-builtin.def (build_pair): New built-in.
13964 (build_acc): Likewise.
13965 * config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
13966 source operands in little-endian mode.
13967 (rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
13968 (mma_init_builtins): Likewise.
13969 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
13970 ordering for the MMA assemble and build source operands.
13971 * doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
13973 (__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
13976 2021-06-10 Jeff Law <jeffreyalaw@gmail.com>
13978 * config/h8300/h8300.c (select_cc_mode): Handle MEM. Use
13980 * config/h8300/extensions.md: Replace _clobber_flags patterns
13983 2021-06-10 Robin Dapp <rdapp@linux.ibm.com>
13985 * config/s390/vector.md (vcond_mask_<mode><mode>): Change to
13986 (vcond_mask_<mode><tointvec>): this.
13988 2021-06-10 Andrew Stubbs <ams@codesourcery.com>
13989 Thomas Schwinge <thomas@codesourcery.com>
13991 * omp-builtins.def (BUILT_IN_GOACC_ENTER_EXIT_DATA): Split into...
13992 (BUILT_IN_GOACC_ENTER_DATA, BUILT_IN_GOACC_EXIT_DATA): ... these.
13993 * gimple.h (enum gf_mask): Split
13994 'GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA' into
13995 'GF_OMP_TARGET_KIND_OACC_ENTER_DATA' and
13996 'GF_OMP_TARGET_KIND_OACC_EXIT_DATA'.
13997 (is_gimple_omp_oacc): Update.
13998 * gimple-pretty-print.c (dump_gimple_omp_target): Likewise.
13999 * gimplify.c (gimplify_omp_target_update): Likewise.
14000 * omp-expand.c (expand_omp_target, build_omp_regions_1)
14001 (omp_make_gimple_edges): Likewise.
14002 * omp-low.c (check_omp_nesting_restrictions, lower_omp_target):
14005 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
14007 * value-query.cc (value_query::value_on_edge): Rename name to
14009 (range_query::range_on_edge): Same.
14010 (range_query::value_of_expr): Same.
14011 (range_query::value_on_edge): Same.
14012 * value-query.h (class value_query): Same.
14013 (class range_query): Same.
14015 2021-06-10 Richard Biener <rguenther@suse.de>
14017 PR tree-optimization/101003
14018 * tree-vect-slp.c (vect_build_slp_tree_2): Appropriately
14019 use the pattern stmt defs when linearizing a chain.
14021 2021-06-10 Jakub Jelinek <jakub@redhat.com>
14024 * ifcvt.c (noce_get_alt_condition, noce_try_abs): Use
14025 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
14027 2021-06-10 Clement Chigot <clement.chigot@atos.net>
14029 * config/rs6000/aix71.h (ASM_CPU_SPEC): Add Power10 directive.
14030 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
14032 2021-06-09 Andrew Pinski <apinski@marvell.com>
14034 PR tree-optimization/100925
14035 * match.pd (a ? CST1 : CST2): Limit transformations
14036 that would produce a negative to integeral types only.
14037 Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.
14039 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
14042 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
14044 * doc/tm.texi: Correctly update.
14046 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
14048 * doc/tm.texi: Correctly update.
14050 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
14053 * doc/tm.texi.in (Trampolines): Add a missing blank line.
14055 2021-06-09 Paul Eggert <eggert@cs.ucla.edu>
14058 * doc/invoke.texi (Code Gen Options); Document that -fno-trampolines
14059 and -ftrampolines work only with Ada.
14060 * doc/tm.texi.in (Trampolines): Likewise.
14061 * doc/tm.texi: Regenerated.
14063 2021-06-09 Carl Love <cel@us.ibm.com>
14065 * config/rs6000/altivec.h (vec_signextll, vec_signexti, vec_signextq):
14066 Add define for new builtins.
14067 * config/rs6000/altivec.md(altivec_vreveti2): Add define_expand.
14068 * config/rs6000/rs6000-builtin.def (VSIGNEXTI, VSIGNEXTLL): Add
14069 overloaded builtin definitions.
14070 (VSIGNEXTSB2W, VSIGNEXTSH2W, VSIGNEXTSB2D, VSIGNEXTSH2D,VSIGNEXTSW2D,
14071 VSIGNEXTSD2Q): Add builtin expansions.
14072 (SIGNEXT): Add P10 overload definition.
14073 * config/rs6000/rs6000-call.c (P9V_BUILTIN_VEC_VSIGNEXTI, P9V_BUILTIN_VEC_VSIGNEXTLL,
14074 P10_BUILTIN_VEC_SIGNEXT): Add overloaded argument definitions.
14075 * config/rs6000/vsx.md (vsx_sign_extend_v2di_v1ti): Add define_insn.
14076 (vsignextend_v2di_v1ti, vsignextend_qi_<mode>, vsignextend_hi_<mode>,
14077 vsignextend_si_v2di)[VIlong]: Add define_expand.
14078 Make define_insn vsx_sign_extend_si_v2di visible.
14079 * doc/extend.texi: Add documentation for the vec_signexti,
14080 vec_signextll builtins and vec_signextq.
14082 2021-06-09 Carl Love <cel@us.ibm.com>
14084 * config/rs6000/rs6000.c (__fixkfti, __fixunskfti, __floattikf,
14085 __floatuntikf): Names changed to __fixkfti_sw, __fixunskfti_sw,
14086 __floattikf_sw, __floatuntikf_sw respectively.
14087 * config/rs6000/rs6000.md (floatti<mode>2, floatunsti<mode>2,
14088 fix_trunc<mode>ti2, fixuns_trunc<mode>ti2): Add
14089 define_insn for mode IEEE 128.
14091 2021-06-09 Carl Love <cel@us.ibm.com>
14093 * config/rs6000/altivec.md (altivec_vslq, altivec_vsrq):
14094 Rename to altivec_vslq_<mode>, altivec_vsrq_<mode>, mode VEC_TI.
14095 * config/rs6000/vector.md (VEC_TI): Was named VSX_TI in vsx.md.
14096 (vashlv1ti3): Change to vashl<mode>3, mode VEC_TI.
14097 (vlshrv1ti3): Change to vlshr<mode>3, mode VEC_TI.
14098 * config/rs6000/vsx.md (VSX_TI): Remove define_mode_iterator. Update
14099 uses of VSX_TI to VEC_TI.
14101 2021-06-09 Carl Love <cel@us.ibm.com>
14103 * config/rs6000/dfp.md (floattitd2, fixtdti2): New define_insns.
14105 2021-06-09 Carl Love <cel@us.ibm.com>
14107 * config/rs6000/altivec.h (vec_dive, vec_mod): Add define for new
14109 * config/rs6000/altivec.md (UNSPEC_VMULEUD, UNSPEC_VMULESD,
14110 UNSPEC_VMULOUD, UNSPEC_VMULOSD): New unspecs.
14111 (altivec_eqv1ti, altivec_gtv1ti, altivec_gtuv1ti, altivec_vmuleud,
14112 altivec_vmuloud, altivec_vmulesd, altivec_vmulosd, altivec_vrlq,
14113 altivec_vrlqmi, altivec_vrlqmi_inst, altivec_vrlqnm,
14114 altivec_vrlqnm_inst, altivec_vslq, altivec_vsrq, altivec_vsraq,
14115 altivec_vcmpequt_p, altivec_vcmpgtst_p, altivec_vcmpgtut_p): New
14117 (vec_widen_umult_even_v2di, vec_widen_smult_even_v2di,
14118 vec_widen_umult_odd_v2di, vec_widen_smult_odd_v2di, altivec_vrlqmi,
14119 altivec_vrlqnm): New define_expands.
14120 * config/rs6000/rs6000-builtin.def (VCMPEQUT_P, VCMPGTST_P,
14121 VCMPGTUT_P): Add macro expansions.
14122 (BU_P10V_AV_P): Add builtin predicate definition.
14123 (VCMPGTUT, VCMPGTST, VCMPEQUT, CMPNET, CMPGE_1TI,
14124 CMPGE_U1TI, CMPLE_1TI, CMPLE_U1TI, VNOR_V1TI_UNS, VNOR_V1TI, VCMPNET_P,
14125 VCMPAET_P, VMULEUD, VMULESD, VMULOUD, VMULOSD, VRLQ,
14126 VSLQ, VSRQ, VSRAQ, VRLQNM, DIV_V1TI, UDIV_V1TI, DIVES_V1TI, DIVEU_V1TI,
14127 MODS_V1TI, MODU_V1TI, VRLQMI): New macro expansions.
14128 (VRLQ, VSLQ, VSRQ, VSRAQ, DIVE, MOD): New overload expansions.
14129 * config/rs6000/rs6000-call.c (P10_BUILTIN_VCMPEQUT,
14130 P10V_BUILTIN_CMPGE_1TI, P10V_BUILTIN_CMPGE_U1TI,
14131 P10V_BUILTIN_VCMPGTUT, P10V_BUILTIN_VCMPGTST,
14132 P10V_BUILTIN_CMPLE_1TI, P10V_BUILTIN_VCMPLE_U1TI,
14133 P10V_BUILTIN_DIV_V1TI, P10V_BUILTIN_UDIV_V1TI,
14134 P10V_BUILTIN_VMULESD, P10V_BUILTIN_VMULEUD,
14135 P10V_BUILTIN_VMULOSD, P10V_BUILTIN_VMULOUD,
14136 P10V_BUILTIN_VNOR_V1TI, P10V_BUILTIN_VNOR_V1TI_UNS,
14137 P10V_BUILTIN_VRLQ, P10V_BUILTIN_VRLQMI,
14138 P10V_BUILTIN_VRLQNM, P10V_BUILTIN_VSLQ,
14139 P10V_BUILTIN_VSRQ, P10V_BUILTIN_VSRAQ,
14140 P10V_BUILTIN_VCMPGTUT_P, P10V_BUILTIN_VCMPGTST_P,
14141 P10V_BUILTIN_VCMPEQUT_P, P10V_BUILTIN_VCMPGTUT_P,
14142 P10V_BUILTIN_VCMPGTST_P, P10V_BUILTIN_CMPNET,
14143 P10V_BUILTIN_VCMPNET_P, P10V_BUILTIN_VCMPAET_P,
14144 P10V_BUILTIN_DIVES_V1TI, P10V_BUILTIN_MODS_V1TI,
14145 P10V_BUILTIN_MODU_V1TI):
14146 New overloaded definitions.
14147 (rs6000_gimple_fold_builtin) [P10V_BUILTIN_VCMPEQUT,
14148 P10V_BUILTIN_CMPNET, P10V_BUILTIN_CMPGE_1TI,
14149 P10V_BUILTIN_CMPGE_U1TI, P10V_BUILTIN_VCMPGTUT,
14150 P10V_BUILTIN_VCMPGTST, P10V_BUILTIN_CMPLE_1TI,
14151 P10V_BUILTIN_CMPLE_U1TI]: New case statements.
14152 (rs6000_init_builtins) [bool_V1TI_type_node, int_ftype_int_v1ti_v1ti]:
14154 (altivec_init_builtins): New E_V1TImode case statement.
14155 (builtin_function_type)[P10_BUILTIN_128BIT_VMULEUD,
14156 P10_BUILTIN_128BIT_VMULOUD, P10_BUILTIN_128BIT_DIVEU_V1TI,
14157 P10_BUILTIN_128BIT_MODU_V1TI, P10_BUILTIN_CMPGE_U1TI,
14158 P10_BUILTIN_VCMPGTUT, P10_BUILTIN_VCMPEQUT]: New case statements.
14159 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute) [E_TImode,
14160 E_V1TImode]: New case statements.
14161 * config/rs6000/rs6000.h (rs6000_builtin_type_index): New enum
14162 value RS6000_BTI_bool_V1TI.
14163 * config/rs6000/vector.md (vector_gtv1ti,vector_nltv1ti,
14164 vector_gtuv1ti, vector_nltuv1ti, vector_ngtv1ti, vector_ngtuv1ti,
14165 vector_eq_v1ti_p, vector_ne_v1ti_p, vector_ae_v1ti_p,
14166 vector_gt_v1ti_p, vector_gtu_v1ti_p, vrotlv1ti3, vashlv1ti3,
14167 vlshrv1ti3, vashrv1ti3): New define_expands.
14168 * config/rs6000/vsx.md (UNSPEC_VSX_DIVSQ, UNSPEC_VSX_DIVUQ,
14169 UNSPEC_VSX_DIVESQ, UNSPEC_VSX_DIVEUQ, UNSPEC_VSX_MODSQ,
14170 UNSPEC_VSX_MODUQ): New unspecs.
14171 (mulv2di3, vsx_div_v1ti, vsx_udiv_v1ti, vsx_dives_v1ti,
14172 vsx_diveu_v1ti, vsx_mods_v1ti, vsx_modu_v1ti, xxswapd_v1ti): New
14174 (vcmpnet): New define_expand.
14175 * doc/extend.texi: Add documentation for the new builtins vec_rl,
14176 vec_rlmi, vec_rlnm, vec_sl, vec_sr, vec_sra, vec_mule, vec_mulo,
14177 vec_div, vec_dive, vec_mod, vec_cmpeq, vec_cmpne, vec_cmpgt, vec_cmplt,
14178 vec_cmpge, vec_cmple, vec_all_eq, vec_all_ne, vec_all_gt, vec_all_lt,
14179 vec_all_ge, vec_all_le, vec_any_eq, vec_any_ne, vec_any_gt, vec_any_lt,
14180 vec_any_ge, vec_any_le.
14182 2021-06-09 Carl Love <cel@us.ibm.com>
14184 * config/rs6000/altivec.md (altivec_vrl<VI_char>mi): Fix
14185 bug in argument generation.
14187 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
14189 * config/arm/iterators.md (<supf>): Remove VCLZQ_U, VCLZQ_S.
14191 * config/arm/mve.md (mve_vclzq_<supf><mode>): Add '@' prefix,
14192 remove <supf> iterator.
14193 (mve_vclzq_u<mode>): New.
14194 * config/arm/neon.md (clz<mode>2): Rename to neon_vclz<mode>.
14195 (neon_vclz<mode): Move to ...
14196 * config/arm/unspecs.md (VCLZQ_U, VCLZQ_S): Remove.
14197 * config/arm/vec-common.md: ... here. Add support for MVE.
14199 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
14201 * config/arm/mve.md (mve_vhaddq_<supf><mode>): Prefix with '@'.
14202 (@mve_vrhaddq_<supf><mode): Likewise.
14203 * config/arm/neon.md (neon_v<r>hadd<sup><mode>): Likewise.
14204 * config/arm/vec-common.md (avg<mode>3_floor, uavg<mode>3_floor)
14205 (avg<mode>3_ceil", uavg<mode>3_ceil): New patterns.
14207 2021-06-09 imba-tjd <109224573@qq.com>
14209 * doc/invoke.texi: Fix typo.
14211 2021-06-09 Roger Sayle <roger@nextmovesoftware.com>
14213 PR middle-end/53267
14214 * fold-const-call.c (fold_const_call_sss) [CASE_CFN_FMOD]:
14215 Support evaluation of fmod/fmodf/fmodl at compile-time.
14217 2021-06-09 Richard Biener <rguenther@suse.de>
14219 PR tree-optimization/100981
14220 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
14221 gimple_get_lhs to also handle calls.
14222 * tree-vect-slp-patterns.c (complex_pattern::build): Transfer
14225 2021-06-09 Richard Biener <rguenther@suse.de>
14227 PR tree-optimization/97832
14228 * tree-vectorizer.h (_slp_tree::failed): New.
14229 * tree-vect-slp.c (_slp_tree::_slp_tree): Initialize
14231 (_slp_tree::~_slp_tree): Free failed.
14232 (vect_build_slp_tree): Retain failed nodes and record
14233 matches in them, copying that back out when running
14234 into a cached fail. Dump start and end of discovery.
14235 (dt_sort_cmp): New.
14236 (vect_build_slp_tree_2): Handle associatable chains
14237 together doing more aggressive operand swapping.
14239 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
14242 * config.gcc (gcc_cv_initfini_array): Set to yes for Linux and
14244 * doc/install.texi: Require glibc 2.1 and binutils 2.12 for
14245 Linux and GNU targets.
14247 2021-06-09 Richard Biener <rguenther@suse.de>
14249 * tree-vect-stmts.c (vect_is_simple_use): Always get dt
14252 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
14254 * config/arc/arc.md (loop_end): Change it to
14255 define_insn_and_split.
14257 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
14259 * config/arc/arc.md (maddhisi4): Use VMAC2H instruction.
14260 (machi): New pattern.
14261 (umaddhisi4): Use VMAC2HU instruction.
14262 (umachi): New pattern.
14264 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
14266 * config/arc/arc-protos.h (arc_split_move_p): New prototype.
14267 * config/arc/arc.c (arc_split_move_p): New function.
14268 (arc_split_move): Clean up.
14269 * config/arc/arc.md (movdi_insn): Clean up, use arc_split_move_p.
14270 (movdf_insn): Likewise.
14271 * config/arc/simdext.md (mov<VWH>_insn): Likewise.
14273 2021-06-09 Uroš Bizjak <ubizjak@gmail.com>
14276 * config/i386/i386.c (print_operand_address_as): Rename "no_rip"
14277 argument to "raw". Do not emit segment overrides when "raw" is true.
14279 2021-06-09 Martin Liska <mliska@suse.cz>
14281 * doc/gcov.texi: Create a proper JSON files.
14282 * doc/invoke.texi: Remove dots in order to make it a valid
14285 2021-06-09 Xionghu Luo <luoxhu@linux.ibm.com>
14287 * config/rs6000/rs6000-p8swap.c (pattern_is_rotate64): New.
14288 (insn_is_load_p): Use pattern_is_rotate64.
14289 (insn_is_swap_p): Likewise.
14290 (quad_aligned_load_p): Likewise.
14291 (const_load_sequence_p): Likewise.
14292 (replace_swapped_aligned_load): Likewise.
14293 (recombine_lvx_pattern): Likewise.
14294 (recombine_stvx_pattern): Likewise.
14296 2021-06-09 Andrew MacLeod <amacleod@redhat.com>
14298 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Use a
14299 fur_stmt source record.
14300 * gimple-range.cc (fur_source::get_operand): Generic range query.
14301 (fur_source::get_phi_operand): New.
14302 (fur_source::register_dependency): New.
14303 (fur_source::query): New.
14304 (class fur_edge): New. Edge source for operands.
14305 (fur_edge::fur_edge): New.
14306 (fur_edge::get_operand): New.
14307 (fur_edge::get_phi_operand): New.
14308 (fur_edge::query): New.
14309 (fur_stmt::fur_stmt): New.
14310 (fur_stmt::get_operand): New.
14311 (fur_stmt::get_phi_operand): New.
14312 (fur_stmt::query): New.
14313 (class fur_depend): New. Statement source and process dependencies.
14314 (fur_depend::fur_depend): New.
14315 (fur_depend::register_dependency): New.
14316 (class fur_list): New. List source for operands.
14317 (fur_list::fur_list): New.
14318 (fur_list::get_operand): New.
14319 (fur_list::get_phi_operand): New.
14320 (fold_range): New. Instantiate appropriate fur_source class and fold.
14321 (fold_using_range::range_of_range_op): Use new API.
14322 (fold_using_range::range_of_address): Ditto.
14323 (fold_using_range::range_of_phi): Ditto.
14324 (imple_ranger::fold_range_internal): Use fur_depend class.
14325 (fold_using_range::range_of_ssa_name_with_loop_info): Use new API.
14326 * gimple-range.h (class fur_source): Now a base class.
14327 (class fur_stmt): New.
14328 (fold_range): New prototypes.
14329 (fur_source::fur_source): Delete.
14331 2021-06-08 Andrew Pinski <apinski@marvell.com>
14333 PR tree-optimization/25290
14334 * tree-ssa-phiopt.c (xor_replacement): Delete.
14335 (tree_ssa_phiopt_worker): Delete use of xor_replacement.
14336 (match_simplify_replacement): Allow one cheap preparation
14337 statement that can be moved to before the if.
14339 2021-06-08 Pat Haugen <pthaugen@linux.ibm.com>
14341 * config/rs6000/power10.md (power10-fused-load, power10-fused-store,
14342 power10-fused_alu, power10-fused-vec, power10-fused-branch): New.
14344 2021-06-08 Jeff Law <jeffreyalaw@gmail.com>
14346 * config/h8300/logical.md (andqi3_1): Move BCLR case into define_insn_and_split.
14347 Create length attribute on define_insn_and_split. Only split for cases which we
14349 (andqi3_1<cczn>): Renamed from andqi3_1_clobber_flags. Only handle AND here and
14350 fix length computation.
14351 (b<code><mode>msx): Combine QImode and HImode H8/SX patterns using iterator.
14353 2021-06-08 Richard Biener <rguenther@suse.de>
14355 PR tree-optimization/100923
14356 * tree-ssa-sccvn.c (valueize_refs_1): Take a pointer to
14357 the operand vector to be valueized.
14358 (valueize_refs): Likewise.
14359 (valueize_shared_reference_ops_from_ref): Adjust.
14360 (valueize_shared_reference_ops_from_call): Likewise.
14361 (vn_reference_lookup_3): Likewise.
14362 (vn_reference_lookup_pieces): Likewise. Re-valueize
14363 with honoring availability when we are about to create
14364 the ao_ref and valueized before.
14365 (vn_reference_lookup): Likewise.
14366 (vn_reference_insert_pieces): Adjust.
14368 2021-06-08 Richard Biener <rguenther@suse.de>
14370 * tree-vectorizer.h (_slp_instance::root_stmt): Change to...
14371 (_slp_instance::root_stmts): ... a vector.
14372 (SLP_INSTANCE_ROOT_STMT): Rename to ...
14373 (SLP_INSTANCE_ROOT_STMTS): ... this.
14374 (slp_root::root): Change to...
14375 (slp_root::roots): ... a vector.
14376 (slp_root::slp_root): Adjust.
14377 * tree-vect-slp.c (_slp_instance::location): Adjust.
14378 (vect_free_slp_instance): Release the root stmt vector.
14379 (vect_build_slp_instance): Adjust.
14380 (vect_analyze_slp): Likewise.
14381 (_bb_vec_info::~_bb_vec_info): Likewise.
14382 (vect_slp_analyze_operations): Likewise.
14383 (vect_bb_vectorization_profitable_p): Likewise. Adjust
14384 costs for the root stmt.
14385 (vect_slp_check_for_constructors): Gather all BIT_INSERT_EXPRs
14387 (vect_slp_analyze_bb_1): Simplify by marking all root stmts
14389 (vectorize_slp_instance_root_stmt): Adjust.
14390 (vect_schedule_slp): Likewise.
14392 2021-06-08 Aldy Hernandez <aldyh@redhat.com>
14394 * gimple-ssa-evrp.c (class ssa_equiv_stack): New.
14395 (ssa_equiv_stack::ssa_equiv_stack): New.
14396 (ssa_equiv_stack::~ssa_equiv_stack): New.
14397 (ssa_equiv_stack::enter): New.
14398 (ssa_equiv_stack::leave): New.
14399 (ssa_equiv_stack::push_replacement): New.
14400 (ssa_equiv_stack::get_replacement): New.
14401 (is_pointer_ssa): New.
14402 (class pointer_equiv_analyzer): New.
14403 (pointer_equiv_analyzer::pointer_equiv_analyzer): New.
14404 (pointer_equiv_analyzer::~pointer_equiv_analyzer): New.
14405 (pointer_equiv_analyzer::set_global_equiv): New.
14406 (pointer_equiv_analyzer::set_cond_equiv): New.
14407 (pointer_equiv_analyzer::get_equiv): New.
14408 (pointer_equiv_analyzer::enter): New.
14409 (pointer_equiv_analyzer::leave): New.
14410 (pointer_equiv_analyzer::get_equiv_expr): New.
14411 (pta_valueize): New.
14412 (pointer_equiv_analyzer::visit_stmt): New.
14413 (pointer_equiv_analyzer::visit_edge): New.
14414 (hybrid_folder::value_of_expr): Call PTA.
14415 (hybrid_folder::value_on_edge): Same.
14416 (hybrid_folder::pre_fold_bb): New.
14417 (hybrid_folder::post_fold_bb): New.
14418 (hybrid_folder::pre_fold_stmt): New.
14419 (rvrp_folder::pre_fold_bb): New.
14420 (rvrp_folder::post_fold_bb): New.
14421 (rvrp_folder::pre_fold_stmt): New.
14422 (rvrp_folder::value_of_expr): Call PTA.
14423 (rvrp_folder::value_on_edge): Same.
14425 2021-06-08 Jakub Jelinek <jakub@redhat.com>
14428 * tree-inline.c (copy_tree_body_r): For OMP_CLAUSE_DEPEND don't
14429 check TREE_CODE if OMP_CLAUSE_DECL is NULL.
14431 2021-06-08 Richard Biener <rguenther@suse.de>
14433 PR middle-end/100951
14434 * tree-vect-generic.c (expand_vector_piecewise): Build a
14435 VECTOR_CST if all elements are constant.
14436 (expand_vector_condition): Likewise.
14437 (lower_vec_perm): Likewise.
14438 (expand_vector_conversion): Likewise.
14440 2021-06-08 Martin Liska <mliska@suse.cz>
14442 * doc/invoke.texi: Document new param evrp-sparse-threshold.
14444 2021-06-08 Martin Liska <mliska@suse.cz>
14446 * genautomata.c (create_automata): Fix typo.
14448 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
14450 PR tree-optimization/100794
14451 * tree-predcom.c (tree_predictive_commoning_loop): Add parameter
14452 allow_unroll_p and only allow unrolling when it's true.
14453 (tree_predictive_commoning): Add parameter allow_unroll_p and
14455 (run_tree_predictive_commoning): Likewise.
14456 (pass_predcom::gate): Check flag_tree_loop_vectorize and
14457 global_options_set.x_flag_predictive_commoning.
14458 (pass_predcom::execute): Adjust for allow_unroll_p.
14460 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
14462 * tree-predcom.c (execute_pred_commoning): Remove update_ssa call.
14463 (tree_predictive_commoning_loop): Factor some cleanup stuffs into
14464 lambda function cleanup, remove scev_reset call, and adjust return
14466 (tree_predictive_commoning): Adjust for different changed values,
14467 only set flag TODO_update_ssa_only_virtuals if changed.
14468 (pass_data pass_data_predcom): Remove TODO_update_ssa_only_virtuals
14469 from todo_flags_finish.
14471 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
14473 * gimple-range-cache.cc (class sbr_sparse_bitmap): New.
14474 (sbr_sparse_bitmap::sbr_sparse_bitmap): New.
14475 (sbr_sparse_bitmap::bitmap_set_quad): New.
14476 (sbr_sparse_bitmap::bitmap_get_quad): New.
14477 (sbr_sparse_bitmap::set_bb_range): New.
14478 (sbr_sparse_bitmap::get_bb_range): New.
14479 (sbr_sparse_bitmap::bb_range_p): New.
14480 (block_range_cache::block_range_cache): initialize bitmap obstack.
14481 (block_range_cache::~block_range_cache): Destruct obstack.
14482 (block_range_cache::set_bb_range): Decide when to utilze the
14483 sparse on entry cache.
14484 * gimple-range-cache.h (block_range_cache): Add bitmap obstack.
14485 * params.opt (-param=evrp-sparse-threshold): New.
14487 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
14489 * bitmap.c (bitmap_set_aligned_chunk): New.
14490 (bitmap_get_aligned_chunk): New.
14491 (test_aligned_chunk): New.
14492 (bitmap_c_tests): Call test_aligned_chunk.
14493 * bitmap.h (bitmap_set_aligned_chunk, bitmap_get_aligned_chunk): New.
14495 2021-06-07 Uroš Bizjak <ubizjak@gmail.com>
14498 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
14500 (ix86_expand_vector_init_one_nonzero): Ditto.
14501 (ix86_expand_vector_init_one_var): Ditto.
14502 (ix86_expand_vector_init_general): Ditto.
14503 * config/i386/mmx.md (vec_initv4qiqi): New expander.
14505 2021-06-07 Jeff Law <jeffreyalaw@gmail.com>
14507 * config/h8300/movepush.md: Change most _clobber_flags
14508 patterns to instead use <cczn> subst.
14509 (movsi_cczn): New pattern with usable CC cases split out.
14510 (movsi_h8sx_cczn): Likewise.
14512 2021-06-07 Martin Liska <mliska@suse.cz>
14514 * common/common-target.def: Split long lines and replace them
14516 * target.def: Likewise.
14517 * doc/tm.texi: Re-generated.
14519 2021-06-07 Jakub Jelinek <jakub@redhat.com>
14522 * fold-const.c (fold_read_from_vector): Return NULL if trying to
14523 read from a CONSTRUCTOR with vector type elements.
14525 2021-06-07 Jakub Jelinek <jakub@redhat.com>
14527 PR middle-end/100898
14528 * tree-inline.c (copy_bb): Only use gimple_call_arg_ptr if memcpy
14529 should copy any arguments. Don't call gimple_call_num_args
14530 on id->call_stmt or call_stmt more than once.
14532 2021-06-07 liuhongt <hongtao.liu@intel.com>
14535 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): Refine
14537 (<insn>v4siv4di2): Delete constraints for define_expand.
14539 2021-06-07 liuhongt <hongtao.liu@intel.com>
14542 * config/i386/i386-expand.c (ix86_expand_builtin): Remove
14543 assignment of cfun->machine->has_explicit_vzeroupper.
14544 * config/i386/i386-features.c
14545 (ix86_add_reg_usage_to_vzerouppers): Delete.
14546 (ix86_add_reg_usage_to_vzeroupper): Ditto.
14547 (rest_of_handle_insert_vzeroupper): Remove
14548 ix86_add_reg_usage_to_vzerouppers, add df_analyze at the end
14550 (gate): Remove cfun->machine->has_explicit_vzeroupper.
14551 * config/i386/i386-protos.h (ix86_expand_avx_vzeroupper):
14553 * config/i386/i386.c (ix86_insn_callee_abi): New function.
14554 (ix86_initialize_callee_abi): Ditto.
14555 (ix86_expand_avx_vzeroupper): Ditto.
14556 (ix86_hard_regno_call_part_clobbered): Adjust for vzeroupper
14558 (TARGET_INSN_CALLEE_ABI): Define as ix86_insn_callee_abi.
14559 (ix86_emit_mode_set): Call ix86_expand_avx_vzeroupper
14561 * config/i386/i386.h (struct GTY(()) machine_function): Delete
14562 has_explicit_vzeroupper.
14563 * config/i386/i386.md (enum unspec): New member
14565 (ABI_DEFAULT,ABI_VZEROUPPER,ABI_UNKNOWN): New
14566 define_constants for insn callee abi index.
14567 * config/i386/predicates.md (vzeroupper_pattern): Adjust.
14568 * config/i386/sse.md (UNSPECV_VZEROUPPER): Deleted.
14569 (avx_vzeroupper): Call ix86_expand_avx_vzeroupper.
14570 (*avx_vzeroupper): Rename to ..
14571 (avx_vzeroupper_callee_abi): .. this, and adjust pattern as
14572 call_insn which has a special vzeroupper ABI.
14573 (*avx_vzeroupper_1): Deleted.
14575 2021-06-07 liuhongt <hongtao.liu@intel.com>
14578 * df-scan.c (df_get_call_refs): When call_insn is a fake call,
14579 it won't use stack pointer reg.
14580 * final.c (leaf_function_p): When call_insn is a fake call, it
14581 won't affect caller as a leaf function.
14582 * reg-stack.c (callee_clobbers_any_stack_reg): New.
14583 (subst_stack_regs): When call_insn doesn't clobber any stack
14584 reg, don't clear the arguments.
14585 * rtl.c (shallow_copy_rtx): Don't clear flag used when orig is
14587 * shrink-wrap.c (requires_stack_frame_p): No need for stack
14588 frame for a fake call.
14589 * rtl.h (FAKE_CALL_P): New macro.
14591 2021-06-06 Eric Botcazou <ebotcazou@adacore.com>
14593 * config/sparc/sparc-protos.h (order_regs_for_local_alloc): Rename
14595 (sparc_order_regs_for_local_alloc): ...this.
14596 (sparc_leaf_reg_remap): Declare.
14597 * config/sparc/sparc.h (ADJUST_REG_ALLOC_ORDER): Adjust.
14598 (LEAF_REG_REMAP): Reimplement as call to sparc_leaf_reg_remap.
14599 * config/sparc/sparc.c (leaf_reg_remap): Delete.
14600 (order_regs_for_local_alloc): Rename to...
14601 (sparc_order_regs_for_local_alloc): ...this.
14602 (sparc_leaf_reg_remap): New function.
14603 (sparc_conditional_register_usage): Do not modify leaf_reg_remap.
14605 2021-06-06 David Edelsohn <dje.gcc@gmail.com>
14607 * config/rs6000/rs6000.c (rs6000_xcoff_asm_output_aligned_decl_common):
14608 Use assemble_name to output BSS section name.
14610 2021-06-06 Uroš Bizjak <ubizjak@gmail.com>
14612 * config/i386/constraints.md (Bs):
14613 Remove boolean operators from match_test RTX.
14616 (M): Use "mode" variable instead of GET_MODE (op) in match_test RTX.
14619 2021-06-06 Martin Liska <mliska@suse.cz>
14621 * doc/extend.texi: Add missing @headitem.
14622 * doc/invoke.texi: Likewise.
14623 * doc/objc.texi: Likewise.
14625 2021-06-06 Martin Liska <mliska@suse.cz>
14627 * genhooks.c (emit_findices): Remove unused function.
14628 (emit_documentation): Do not call emit_findices
14629 and do not search for @Fcode directives.
14631 2021-06-06 Martin Liska <mliska@suse.cz>
14633 * doc/invoke.texi: Remove extra character.
14635 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
14637 * config/sh/sh.md (doloop_end_split): Fix empty split condition.
14639 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
14641 * config/sparc/sparc.md (*snedi<W:mode>_zero_vis3,
14642 *neg_snedi<W:mode>_zero_subxc, *plus_snedi<W:mode>_zero,
14643 *plus_plus_snedi<W:mode>_zero, *minus_snedi<W:mode>_zero,
14644 *minus_minus_snedi<W:mode>_zero): Fix empty split condition.
14646 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
14648 * config/or1k/or1k.md (*movdi): Fix empty split condition.
14650 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
14652 * config/mips/mips.md (<anonymous>, bswapsi2, bswapdi2): Fix empty
14655 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
14657 * config/m68k/m68k.md (*zero_extend_inc, *zero_extend_dec,
14658 *zero_extendsidi2): Fix empty split condition.
14660 2021-06-05 Jeff Law <jeffreyalaw@gmail.com>
14662 * config/h8300/addsub.md: Fix split condition in define_insn_and_split
14664 * config/h8300/bitfield.md: Likewise.
14665 * config/h8300/combiner.md: Likewise.
14666 * config/h8300/divmod.md: Likewise.
14667 * config/h8300/extensions.md: Likewise.
14668 * config/h8300/jumpcall.md: Likewise.
14669 * config/h8300/movepush.md: Likewise.
14670 * config/h8300/multiply.md: Likewise.
14671 * config/h8300/other.md: Likewise.
14672 * config/h8300/shiftrotate.md: Likewise.
14673 * config/h8300/logical.md: Likewise. Fix split pattern to use
14674 code iterator that somehow slipped through.
14676 2021-06-04 Tobias Burnus <tobias@codesourcery.com>
14678 PR middle-end/100905
14679 * tree-nested.c (convert_nonlocal_omp_clauses,
14680 convert_local_omp_clauses): Handle OMP_CLAUSE_BIND.
14682 2021-06-04 Martin Sebor <msebor@redhat.com>
14684 PR middle-end/100732
14685 * gimple-fold.c (gimple_fold_builtin_sprintf): Avoid folding calls
14686 with either source or destination argument of invalid type.
14687 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Avoid checking
14688 calls with arguments of invalid type.
14690 2021-06-04 Martin Sebor <msebor@redhat.com>
14692 * attribs.c (init_attr_rdwr_indices): Use VLA bounds in the expected
14694 (attr_access::vla_bounds): Also handle VLA bounds.
14696 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
14698 * config/i386/predicates.md (GOT_memory_operand):
14699 Implement using match_code RTXes.
14700 (GOT32_symbol_operand): Ditto.
14702 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
14705 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
14707 (ix86_expand_vector_init_general): Ditto.
14708 Use SImode instead of word_mode for logic operations
14709 when GET_MODE_SIZE (mode) < UNITS_PER_WORD.
14710 (expand_vec_perm_even_odd_1): Assert that V2HI mode should be
14711 implemented by expand_vec_perm_1.
14712 (expand_vec_perm_broadcast_1): Assert that V2HI and V4HI modes
14713 should be implemented using standard shuffle patterns.
14714 (ix86_vectorize_vec_perm_const): Handle V2HImode. Add V4HI and
14715 V2HI modes to modes, implementable with shuffle for one operand.
14716 * config/i386/mmx.md (*punpckwd): New insn_and_split pattern.
14717 (*pshufw_1): New insn pattern.
14718 (*vec_dupv2hi): Ditto.
14719 (vec_initv2hihi): New expander.
14721 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
14723 * config/arm/vfp.md (no_literal_pool_df_immediate,
14724 no_literal_pool_sf_immediate): Fix empty split condition.
14726 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
14728 * config/i386/i386.md (*load_tp_x32_zext, *add_tp_x32_zext,
14729 *tls_dynamic_gnu2_combine_32): Fix empty split condition.
14730 * config/i386/sse.md (*<sse2_avx2>_pmovmskb_lt,
14731 *<sse2_avx2>_pmovmskb_zext_lt, *sse2_pmovmskb_ext_lt,
14732 *<sse4_1_avx2>_pblendvb_lt): Likewise.
14734 2021-06-04 Jakub Jelinek <jakub@redhat.com>
14737 * config/i386/i386-expand.c (ix86_expand_vector_init): Handle
14738 concatenation from half-sized modes with TImode elements.
14740 2021-06-04 Claudiu Zissulescu <claziss@synopsys.com>
14742 * config/arc/arc.c (arc_override_options): Disable millicode
14743 thunks when RF16 is on.
14745 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
14747 * config/rs6000/rs6000.h (PROMOTE_MODE): Remove.
14749 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
14751 * config/rs6000/rs6000-call.c (rs6000_promote_function_mode):
14752 Replace PROMOTE_MODE marco with its content.
14754 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
14756 * config/cris/cris.md (*addi_reload): Fix empty split condition.
14758 2021-06-03 Jim Wilson <jimw@sifive.com>
14760 * config.gcc (riscv*-*-*): If --with-riscv-attribute not used,
14761 turn it on for all riscv targets.
14763 2021-06-03 Uroš Bizjak <ubizjak@gmail.com>
14766 * config/i386/i386-expand.c (ix86_expand_vector_set):
14767 Handle V2HI and V4QI modes.
14768 (ix86_expand_vector_extract): Ditto.
14769 * config/i386/mmx.md (*pinsrw): New insn pattern.
14772 (*pextrw_zext): Ditto.
14774 (*pextrb_zext): Ditto.
14775 (vec_setv2hi): New expander.
14776 (vec_extractv2hihi): Ditto.
14777 (vec_setv4qi): Ditto.
14778 (vec_extractv4qiqi): Ditto.
14779 (vec_setv8qi): Enable only for TARGET_SSE4_1.
14780 (vec_extractv8qiqi): Ditto.
14782 2021-06-03 Aaron Sawdey <acsawdey@linux.ibm.com>
14784 * config/rs6000/genfusion.pl (gen_logical_addsubf): Fix input
14785 order to subf instruction.
14786 * config/rs6000/fusion.md: Regenerate.
14788 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
14790 * calls.c (get_size_range): Use range_of_expr instead of
14791 determine_value_range.
14792 * tree-affine.c (expr_to_aff_combination): Same.
14793 * tree-data-ref.c (split_constant_offset): Same.
14794 * tree-vrp.c (determine_value_range_1): Remove.
14795 (determine_value_range): Remove.
14796 * tree-vrp.h (determine_value_range): Remove.
14798 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
14800 * function-tests.c (test_ranges): Call gimple_range_tests.
14801 * gimple-range-cache.cc (ranger_cache::range_of_expr): Pass stmt
14803 * gimple-range.cc (fur_source::get_operand): Do not call
14804 get_tree_range or gimple_range_global.
14806 (get_tree_range): Move to value-query.cc.
14807 Call get_arith_expr_range.
14808 (gimple_ranger::range_of_expr): Add argument to get_tree_range.
14809 Include gimple-range-tests.cc.
14810 * gimple-range.h (fold_range): Add argument.
14811 (get_tree_range): Remove.
14812 * selftest.h (gimple_range_tests): New.
14813 * value-query.cc (global_range_query::range_of_expr): Add
14815 (range_query::get_tree_range): Move from gimple-range.cc.
14816 * value-query.h (class range_query): Add get_tree_range and
14817 get_arith_expr_range. Make fur_source a friend.
14818 * vr-values.c (vr_values::range_of_expr): Pass stmt to
14820 * gimple-range-tests.cc: New file.
14822 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
14824 * gimple-range.cc (gimple_ranger::export_global_ranges): Call
14825 update_global_range.
14826 * value-query.cc (update_global_range): New.
14827 * value-query.h (update_global_range): New.
14829 2021-06-03 David Malcolm <dmalcolm@redhat.com>
14831 * diagnostic-show-locus.c (diagnostic_show_locus): Don't reject
14832 printing the same location twice if there are fix-it hints,
14833 multiple locations, or a label.
14835 2021-06-03 Andre Vieira <andre.simoesdiasvieira@arm.com>
14837 * tree-vect-loop.c (vect_transform_loop): Use main loop's various'
14838 thresholds to narrow the upper bound on epilogue iterations.
14840 2021-06-03 Christophe Lyon <christophe.lyon@linaro.org>
14842 * config/arm/mve.md (mve_vabsq_f<mode>): Use 'abs' instead of unspec.
14843 (mve_vabsq_s<mode>): Likewise.
14844 * config/arm/neon.md (abs<mode>2): Rename to neon_abs<mode>2.
14845 * config/arm/unspecs.md (VABSQ_F, VABSQ_S): Delete.
14846 * config/arm/vec-common.md (neg<mode>2): Rename to
14847 <absneg_str><mode>2.
14849 2021-06-03 Claudiu Zissulescu <claziss@synopsys.com>
14851 * common/config/arc/arc-common.c (arc_option_optimization_table):
14852 Remove malign-call.
14853 * config/arc/arc.c (arc_unalign_branch_p): Remove unused function.
14854 * config/arc/arc.h (TARGET_MIXED_CODE): Remove macro.
14855 (INDEX_REG_CLASS): Only refer to GENERAL_REGS.
14856 * config/arc/arc.md (abssi2_mixed): Remove pattern.
14857 * config/arc/arc.opt (munalign-prob-threshold): Mark it obsolete.
14858 (malign-call): Likewise.
14859 (mmixed-code): Likewise.
14860 * doc/invoke.texi (ARC): Update doc.
14862 2021-06-03 Martin Liska <mliska@suse.cz>
14864 * common.opt: Use proper Enum values.
14865 * opts.c (COVERAGE_SANITIZER_OPT): Remove.
14866 (parse_sanitizer_options): Handle only sanitizer_opts.
14867 (common_handle_option): Just assign value.
14869 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
14872 * tree-inline.c (inline_forbidden_p): Remove test on return type.
14874 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
14876 * dwarf2out.c (loc_list_from_tree_1) <FUNCTION_DECL>: Also generate
14877 DW_OP_GNU_variable_value referencing an existing DIE at file scope.
14878 (type_byte_size): Inline into...
14879 (add_byte_size_attribute): ...this and call add_scalar_info.
14881 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
14883 * dwarf2out.c (mem_loc_descriptor) <UDIV>: Fix typo.
14884 (typed_binop_from_tree): New function.
14885 (loc_list_from_tree_1) <EXACT_DIV_EXPR>: For an unsigned type,
14886 turn a divide by a power of 2 into a shift.
14887 <CEIL_DIV_EXPR>: For an unsigned type, use a signed divide if the
14888 size of the mode is lower than DWARF2_ADDR_SIZE; otherwise, do a
14889 typed divide by calling typed_binop_from_tree.
14891 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
14893 * dwarf2out.c (scompare_loc_descriptor): Fix head comment.
14894 (is_handled_procedure_type): Likewise.
14895 (struct loc_descr_context): Add strict_signedness field.
14896 (resolve_args_picking_1): Deal with DW_OP_[GNU_]deref_type,
14897 DW_OP_[GNU_]convert and DW_OP_[GNU_]reinterpret.
14898 (resolve_args_picking): Minor tweak.
14899 (function_to_dwarf_procedure): Initialize strict_signedness field.
14900 (type_byte_size): Likewise.
14901 (field_byte_offset): Likewise.
14902 (gen_descr_array_type_die): Likewise.
14903 (gen_variant_part): Likewise.
14904 (loc_list_from_tree_1) <CALL_EXPR>: Tidy up and set strict_signedness
14905 to true when a context is present before evaluating the arguments.
14906 <COND_EXPR>: Do not generate a useless comparison with zero.
14907 When dereferencing an address, if strict_signedness is true and the
14908 type is small and signed, use DW_OP_deref_type to do the dereference
14909 and then DW_OP_convert to convert back to the generic type.
14911 2021-06-03 Jakub Jelinek <jakub@redhat.com>
14914 * tree-inline.c (copy_tree_body_r): Handle iterators on
14915 OMP_CLAUSE_AFFINITY or OMP_CLAUSE_DEPEND.
14917 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
14919 * config/arc/arc.md (*bbit_di): Remove.
14921 2021-06-02 Christoph Muellner <cmuellner@gcc.gnu.org>
14923 PR rtl-optimization/100264
14924 * ree.c (get_sub_rtx): Ignore SET expressions without register
14925 destinations and remove assertion, as it is not valid anymore
14926 with this new behaviour.
14927 (merge_def_and_ext): Eliminate destination check for register
14928 as such SET expressions can't occur anymore.
14929 (combine_reaching_defs): Likewise.
14931 2021-06-02 Jakub Jelinek <jakub@redhat.com>
14934 * config/xtensa/xtensa.h (LEAF_REG_REMAP): Cast REGNO to int to avoid
14935 -Wtype-limits warnings.
14936 (DWARF_FRAME_REGISTER): Rewrite into ternary operator with addition
14937 in operands to avoid -Wsign-compare warnings.
14939 2021-06-02 Pat Haugen <pthaugen@linux.ibm.com>
14941 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): Use
14944 2021-06-02 Vineet Gupta <vgupta@synopsys.com>
14946 * config/arc/arc.h (TARGET_CPU_DEFAULT): Change to hs38_linux.
14948 2021-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
14950 * config/s390/s390.md(*ashrdi3_31<setcc><cconly>): Use a single
14952 * config/s390/subst.md(cconly_subst): Use a single constraint
14953 in (match_scratch).
14955 2021-06-02 Martin Liska <mliska@suse.cz>
14957 * ipa-icf.h: Use auto_vec for memory_access_types.
14959 2021-06-02 Jeff Law <jeffreyalaw@gmail.com>
14961 * config/h8300/h8300-protos.h (compute_a_shift_length): Drop unused
14962 argument from prototype.
14963 (output_logical_op): Add rtx_code argument.
14964 (compute_logical_op_length): Likewise.
14965 * config/h8300/h8300.c (h8300_and_costs): Pass additional argument
14966 to compute_a_shift_length.
14967 (output_logical_op); New argument with the rtx code rather than
14968 extracting it from an operand. Handle QImode too.
14969 (compute_logical_op_length): Similary.
14970 (compute_a_shift_length): Drop unused argument.
14971 * config/h8300/h8300.md (logicals): New code iterator.
14972 * config/h8300/logical.md (<code><mode>3 expander): Combine
14973 the "and" expander with the "ior"/"xor" expander.
14974 (bclr<mode>msx): Combine the QI/HI mode patterns.
14975 (<logical><mode>3 insns): Use code iterator rather than match_operator.
14976 Handle QImode as well. Update call to output_logical_op and
14977 compute_logical_op_length to pass in rtx_code
14978 Fix split condition on all define_insn_and_split patterns.
14979 (one_cmpl<mode>2<cczn>): Use <cczn> to support both clobbering
14980 the flags and setting ZN via existing define_subst.
14981 * config/h8300/shiftrotate.md: Drop unused argument from
14982 calls to compute_a_shift_length.
14983 Signed-off-by: Jeff Law <jeffreyalaw@gmail.com>
14985 2021-06-01 Andrew Pinski <apinski@marvell.com>
14987 PR tree-optimization/25290
14988 * tree-ssa-phiopt.c (match_simplify_replacement):
14990 (tree_ssa_phiopt_worker): Use match_simplify_replacement.
14991 (two_value_replacement): Change the comment about
14992 conditional_replacement.
14993 (conditional_replacement): Delete.
14995 2021-06-01 Andrew Pinski <apinski@marvell.com>
14997 PR tree-optimization/95481
14998 * tree-tailcall.c (find_tail_calls): Handle empty typed
15001 2021-06-01 Andrew Pinski <apinski@marvell.com>
15003 * gimplify.c (zero_sized_field_decl): Delete
15004 (zero_sized_type): Delete
15005 (gimplify_init_ctor_eval): Use is_empty_type instead
15006 of zero_sized_field_decl.
15007 (gimplify_modify_expr): Use is_empty_type instead of
15010 2021-06-01 Jason Merrill <jason@redhat.com>
15013 * tree.h (CALL_FROM_NEW_OR_DELETE_P): Adjust comment.
15015 2021-06-01 Jason Merrill <jason@redhat.com>
15018 * diagnostic.h (warning_enabled_at): Declare.
15019 * diagnostic.c (diagnostic_enabled): Factor out from...
15020 (diagnostic_report_diagnostic): ...here.
15021 (warning_enabled_at): New.
15023 2021-06-01 Aldy Hernandez <aldyh@redhat.com>
15025 * gimple-ssa-evrp.c: Enable exporting of global ranges.
15027 2021-06-01 Martin Liska <mliska@suse.cz>
15030 * doc/invoke.texi: Mention that -fgcse-after-reload
15031 is enabled with -O3.
15033 2021-06-01 liuhongt <hongtao.liu@intel.com>
15035 PR tree-optimization/98365
15036 * tree-if-conv.c (strip_nop_cond_scalar_reduction): New function.
15037 (is_cond_scalar_reduction): Handle nop_expr in cond scalar reduction.
15038 (convert_scalar_cond_reduction): Ditto.
15039 (predicate_scalar_phi): Ditto.
15041 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
15043 PR tree-optimization/100781
15044 * gimple-range-cache.cc (ranger_cache::ranger_cache): Enable new
15045 value calculation by default.
15046 (ranger_cache::enable_new_values): New.
15047 (ranger_cache::disable_new_values): New.
15048 (ranger_cache::push_poor_value): Check if new values are allowed.
15049 * gimple-range-cache.h (class ranger_cache): New member/methods.
15050 * gimple-range.cc (gimple_ranger::range_of_expr): Check for debug
15051 statement, and disable/renable new value calculation.
15053 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
15055 * gimple-range-cache.cc (ranger_cache::ssa_range_in_bb): Delete.
15056 (ranger_cache::range_of_def): New.
15057 (ranger_cache::entry_range): New.
15058 (ranger_cache::exit_range): New.
15059 (ranger_cache::range_of_expr): Adjust.
15060 (ranger_cache::range_on_edge): Adjust.
15061 (ranger_cache::propagate_cache): Call exit_range directly.
15062 * gimple-range-cache.h (class ranger_cache): Adjust.
15064 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
15066 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust for
15067 gori_compute being a member rather than base class.
15068 dervied call to member call.
15069 (ranger_cache::dump): No longer dump gori_map.
15070 (ranger_cache::dump_bb): New.
15071 (ranger_cache::get_non_stale_global_range): Adjust for gori_compute
15072 being a member rather than base class.
15073 (ranger_cache::set_global_range): Ditto.
15074 (ranger_cache::ssa_range_in_bb): Ditto.
15075 (ranger_cache::range_of_expr): New.
15076 (ranger_cache::range_on_edge): New.
15077 (ranger_cache::block_range): Adjust for gori_computes. Debug changes.
15078 (ranger_cache::propagate_cache): Adjust debugging output.
15079 (ranger_cache::fill_block_cache): Adjust for gori_computes. Debug
15081 * gimple-range-cache.h (class ranger_cache): Make gori_compute a
15082 member, and inherit from range_query instead.
15083 (ranger_cache::dump_bb): New. split from dump.
15084 * gimple-range-gori.cc (gori_compute::ssa_range_in_bb): Delete.
15085 (gori_compute::expr_range_at_stmt): Delete.
15086 (gori_compute::compute_name_range_op): Delete.
15087 (gori_compute::compute_operand_range_switch): Add fur_source.
15088 (gori_compute::compute_operand_range): Add fur_source param, inline
15089 old compute_name_range_op and optimize_logical_operands.
15090 (struct tf_range): Delete.
15091 (gori_compute::logical_combine): Adjust
15092 (gori_compute::optimize_logical_operands): Delete.
15093 (gori_compute::compute_logical_operands_in_chain): Delete.
15094 (gori_compute::compute_logical_operands): Adjust.
15095 (gori_compute::compute_operand1_range): Adjust to fur_source.
15096 (gori_compute::compute_operand2_range): Ditto.
15097 (gori_compute::compute_operand1_and_operand2_range): Ditto.
15098 (gori_compute::outgoing_edge_range_p): Add range_query parameter,
15099 and adjust to fur_source.
15100 * gimple-range-gori.h (class gori_compute): Simplify and adjust to
15101 range_query and fur_source.
15102 * gimple-range.cc (gimple_ranger::range_on_edge): Query range_on_edge
15103 from the ranger_cache..
15104 (gimple_ranger::fold_range_internal): Adjust to base class change of
15106 (gimple_ranger::dump_bb): Adjust dump.
15107 * gimple-range.h (gimple_ranger):export gori computes object.
15109 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
15111 PR tree-optimization/100774
15112 * gimple-range-cache.cc (ranger_cache::get_non_stale_global_range):
15113 Constant values are also not stale.
15114 (ranger_cache::set_global_range): Range invariant values should also
15115 have the correct timestamp.
15117 2021-05-31 Martin Liska <mliska@suse.cz>
15119 * tree-streamer-in.c (unpack_ts_function_decl_value_fields):
15120 Unpack FUNCTION_DECL_DECL_TYPE.
15121 * tree-streamer-out.c (pack_ts_function_decl_value_fields):
15122 Stream FUNCTION_DECL_DECL_TYPE instead of
15123 DECL_IS_OPERATOR_NEW_P.
15124 * tree.h (set_function_decl_type): Use FUNCTION_DECL_DECL_TYPE
15126 (DECL_IS_OPERATOR_NEW_P): Likewise.
15127 (DECL_IS_OPERATOR_DELETE_P): Likewise.
15128 (DECL_LAMBDA_FUNCTION_P): Likewise.
15130 2021-05-31 Richard Biener <rguenther@suse.de>
15133 * internal-fn.c (expand_SHUFFLEVECTOR): Define.
15134 * internal-fn.def (SHUFFLEVECTOR): New.
15135 * internal-fn.h (expand_SHUFFLEVECTOR): Declare.
15136 * doc/extend.texi: Document __builtin_shufflevector.
15138 2021-05-31 Peter Bergner <bergner@linux.ibm.com>
15141 * config/rs6000/predicates.md(mma_assemble_input_operand): Allow
15142 indexed form addresses.
15144 2021-05-29 Jeff Law <jlaw@tachyum.com>
15146 * config/h8300/h8300.c (h8300_emit_stack_adjustment): Drop unused
15147 parameter. Call callers fixed.
15149 (output_plussi): Add FALLTHRU markers.
15150 (h8300_shift_needs_scratch_p): Add gcc_unreachable marker.
15152 2021-05-29 Jakub Jelinek <jakub@redhat.com>
15154 PR middle-end/99928
15155 * gimplify.c (gimplify_scan_omp_clauses): For taskloop simd
15156 combined with parallel, make sure to add shared clause to
15157 parallel for explicit linear clause.
15159 2021-05-29 Aldy Hernandez <aldyh@redhat.com>
15161 PR tree-optimization/100787
15162 * gimple-ssa-evrp.c: Disable exporting of global ranges.
15164 2021-05-28 Jason Merrill <jason@redhat.com>
15166 * tree-iterator.h (struct tree_stmt_iterator): Add operator++,
15167 operator--, operator*, operator==, and operator!=.
15168 (class tsi_range): New.
15170 2021-05-28 Richard Biener <rguenther@suse.de>
15172 PR tree-optimization/100778
15173 * tree-vect-slp.c (vect_build_slp_tree_1): Prevent possibly
15174 trapping ops in different BBs.
15176 2021-05-28 Richard Biener <rguenther@suse.de>
15179 * tree-inline.c (copy_bb): When processing __builtin_va_arg_pack
15180 copy fntype from original call.
15182 2021-05-28 Martin Liska <mliska@suse.cz>
15184 PR gcov-profile/100751
15185 * doc/gcov.texi: Revert partially a hunk that was wrong.
15187 2021-05-28 Cooper Qu <cooper.qu@linux.alibaba.com>
15189 * config/csky/csky-linux-elf.h (HAVE_sync_compare_and_swapqi):
15191 (HAVE_sync_compare_and_swaphi): Likewise.
15192 (HAVE_sync_compare_and_swapsi): Likewise.
15194 2021-05-28 Jakub Jelinek <jakub@redhat.com>
15196 PR middle-end/99928
15197 * tree.h (OMP_CLAUSE_MAP_IMPLICIT): Define.
15199 2021-05-28 Tobias Burnus <tobias@codesourcery.com>
15201 * gimplify.c (gimplify_omp_affinity): New.
15202 (gimplify_scan_omp_clauses): Call it; remove affinity clause afterwards.
15203 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_AFFINITY.
15204 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_AFFINITY.
15205 * tree.c (omp_clause_num_ops, omp_clause_code_name): Add clause.
15206 (walk_tree_1): Handle OMP_CLAUSE_AFFINITY.
15208 2021-05-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
15209 Richard Biener <rguenther@suse.de>
15211 * match.pd <popcount & / + pattern matching>:
15212 When generating popcount directly fails, try doing it in two halves.
15214 2021-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
15216 * Makefile.in (generated_files): Add gimple-match.c and
15219 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
15221 * gensupport.c (alter_predicate_for_insn): Handle MATCH_DUP.
15223 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
15225 * gensupport.c (alter_constraints): Add MATCH_SCRATCH case.
15227 2021-05-28 Kewen Lin <linkw@linux.ibm.com>
15229 PR tree-optimization/99398
15230 * tree-ssa-forwprop.c (simplify_permutation): Optimize some cases
15231 where the fed operands are CTOR/CST and propagated through
15232 VIEW_CONVERT_EXPR. Call vec_perm_indices::new_shrunk_vector.
15233 * vec-perm-indices.c (vec_perm_indices::new_shrunk_vector): New
15235 * vec-perm-indices.h (vec_perm_indices::new_shrunk_vector): New
15238 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
15240 * config/i386/mmx.md (addv2sf3): Do not call
15241 ix86_fixup_binary_operands_no_copy.
15244 (<smaxmin:code>v2sf3): Ditto.
15245 (<plusminus:insn><MMXMODEI:mode>3): Ditto.
15246 (<plusminus:insn><VI_32:mode>3): Remove expander.
15247 (<plusminus:insn><VI_32:mode>3): Rename from
15248 "*<plusminus:insn><VI_32:mode>3".
15249 (mulv4hi): Do not call ix86_fixup_binary_operands_no_copy.
15250 (mulv2hi3): Remove expander.
15251 (mulv2hi3): Rename from *mulv2hi3.
15252 (<s>mulv2hi3_highpart): Remove expander.
15253 (<s>mulv2hi3_highpart): Rename from *<s>mulv2hi3_highpart.
15254 (<smaxmin:code><MMXMODE14:mode>3): Rename from
15255 "*mmx_<smaxmin:code><MMXMODE14:mode>3".
15256 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Remove expander.
15257 (SMAXMIN_MMXMODEI): Remove mode iterator.
15258 (<smaxmin:code>v4hi3): New expander.
15259 (<smaxmin:code>v4qi3): Rename from *<smaxmin:code>v4qi3.
15260 (<smaxmin:code>v2hi3): Rename from *<smaxmin:code>v2hi3.
15261 (<smaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
15262 (SMAXMIN_VI_32): Remove mode iterator.
15263 (<umaxmin:code><MMXMODE24:mode>3): Rename from
15264 "*mmx_<umaxmin:code><MMXMODE24:mode>3".
15265 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Remove expander.
15266 (UMAXMIN_MMXMODEI): Remove mode iterator.
15267 (<umaxmin:code>v8qi3): New expander.
15268 (<umaxmin:code>v4qi3): Rename from *<umaxmin:code>v4qi3.
15269 (<umaxmin:code>v2hi3): Rename from *<umaxmin:code>v2hi3.
15270 (<umaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
15271 (UMAXMIN_VI_32): Remove mode iterator.
15272 (<any_shift:insn>v2hi3): Remove expander.
15273 (<any_shift:insn>v2hi3): Rename from *<any_shift:insn>v2hi3.
15274 (<any_logic:code><MMXMODEI:mode>3): Do not call
15275 ix86_fixup_binary_operands_no_copy.
15276 (<any_logic:code><VI_32:mode>3): Remove expander.
15277 (<any_logic:code><VI_32:mode>3): Rename from
15278 "*<any_logic:code><VI_32:mode>3".
15279 (uavg<mode>3_ceil): Do not call ix86_fixup_binary_operands_no_copy.
15280 * config/i386/sse.md (div<VF2:mode>3): Do not call
15281 ix86_fixup_binary_operands_no_copy.
15282 (div<VF1:mode>3): Ditto.
15283 (<maxmin:code><VI8_AVX2_AVX512F:mode>3): Ditto.
15284 (smulhrsv4hi3): Ditto.
15285 (smulhrsv2hi3): Ditto.
15287 2021-05-27 Martin Sebor <msebor@redhat.com>
15289 * ggc.h (gt_ggc_mx): Add overloads for all integers.
15291 * hash-map.h (class hash_map): Add pch_nx_helper overloads for all
15293 (hash_map::operator==): New function.
15295 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
15298 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
15299 For TARGET_XOP bypass SSE comparisons for all supported vector modes.
15300 * config/i386/mmx.md (*xop_maskcmp<MMXMODEI:mode>3): New insn pattern.
15301 (*xop_maskcmp<VI_32:mode>3): Ditto.
15302 (*xop_maskcmp_uns<MMXMODEI:mode>3): Ditto.
15303 (*xop_maskcmp_uns<VI_32:mode>3): Ditto.
15305 2021-05-27 Richard Earnshaw <rearnsha@arm.com>
15308 * config/arm/arm.c (arm_configure_build_target): Remove parameter
15309 opts_set, directly check opts parameters for being non-null.
15310 (arm_option_restore): Update call to arm_configure_build_target.
15311 (arm_option_override): Likewise.
15312 (arm_can_inline_p): Likewise.
15313 (arm_valid_target_attribute_tree): Likewise.
15314 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
15315 * config/arm/arm-protos.h (arm_configure_build_target): Adjust
15318 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
15320 * vr-values.c (simplify_conversion_using_ranges): Use
15321 get_range_query instead of get_global_range_query.
15323 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
15325 * gimple-range.cc (get_range_global): Move to value-query.cc.
15326 (gimple_range_global): Same.
15327 (get_global_range_query): Same.
15328 (global_range_query::range_of_expr): Same.
15329 * gimple-range.h (class global_range_query): Move to
15331 (gimple_range_global): Same.
15332 * tree-ssanames.c (get_range_info): Move to value-query.cc.
15333 (get_ptr_nonnull): Same.
15334 * tree-ssanames.h (get_range_info): Remove.
15335 (get_ptr_nonnull): Remove.
15336 * value-query.cc (get_ssa_name_range_info): Move from
15338 (get_ssa_name_ptr_info_nonnull): Same.
15339 (get_range_global): Move from gimple-range.cc.
15340 (gimple_range_global): Same.
15341 (get_global_range_query): Same.
15342 (global_range_query::range_of_expr): Same.
15343 * value-query.h (class global_range_query): Move from
15345 (gimple_range_global): Same.
15347 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
15350 * config/i386/mmx.md (uavgv4qi3_ceil): New insn pattern.
15351 (uavgv2hi3_ceil): Ditto.
15353 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
15356 * doc/extend.texi (scalar_storage_order): Rephrase slightly.
15358 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
15360 * tree-ssanames.c (get_range_info): Merge both copies of
15361 get_range_info into one that works with irange.
15362 * tree-ssanames.h (get_range_info): Remove version that works on
15365 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
15367 * builtins.c (check_nul_terminated_array): Convert to get_range_query.
15368 (expand_builtin_strnlen): Same.
15369 (determine_block_size): Same.
15370 * fold-const.c (expr_not_equal_to): Same.
15371 * gimple-fold.c (size_must_be_zero_p): Same.
15372 * gimple-match-head.c: Include gimple-range.h.
15373 * gimple-pretty-print.c (dump_ssaname_info): Convert to get_range_query.
15374 * gimple-ssa-warn-restrict.c
15375 (builtin_memref::extend_offset_range): Same.
15376 * graphite-sese-to-poly.c (add_param_constraints): Same.
15377 * internal-fn.c (get_min_precision): Same.
15378 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): Same.
15379 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Same.
15381 * tree-data-ref.c (split_constant_offset): Same.
15382 (dr_step_indicator): Same.
15383 * tree-dfa.c (get_ref_base_and_extent): Same.
15384 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
15385 * tree-ssa-loop-niter.c (refine_value_range_using_guard): Same.
15386 (determine_value_range): Same.
15387 (record_nonwrapping_iv): Same.
15388 (infer_loop_bounds_from_signedness): Same.
15389 (scev_var_range_cant_overflow): Same.
15390 * tree-ssa-phiopt.c (two_value_replacement): Same.
15391 * tree-ssa-pre.c (insert_into_preds_of_block): Same.
15392 * tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Same.
15393 * tree-ssa-strlen.c (handle_builtin_stxncpy_strncat): Same.
15395 (dump_strlen_info): Same.
15396 (set_strlen_range): Same.
15397 (maybe_diag_stxncpy_trunc): Same.
15398 (get_len_or_size): Same.
15399 (handle_integral_assign): Same.
15400 * tree-ssa-structalias.c (find_what_p_points_to): Same.
15401 * tree-ssa-uninit.c (find_var_cmp_const): Same.
15402 * tree-switch-conversion.c (bit_test_cluster::emit): Same.
15403 * tree-vect-patterns.c (vect_get_range_info): Same.
15404 (vect_recog_divmod_pattern): Same.
15405 * tree-vrp.c (intersect_range_with_nonzero_bits): Same.
15406 (register_edge_assert_for_2): Same.
15407 (determine_value_range_1): Same.
15408 * tree.c (get_range_pos_neg): Same.
15409 * vr-values.c (vr_values::get_lattice_entry): Same.
15410 (vr_values::update_value_range): Same.
15411 (simplify_conversion_using_ranges): Same.
15413 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
15415 * gimple-ssa-warn-alloca.c (alloca_call_type): Use
15416 get_range_query instead of query argument.
15417 (pass_walloca::execute): Enable and disable global ranger.
15419 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
15421 * gimple-ssa-evrp.c (rvrp_folder::rvrp_folder): Call
15423 (rvrp_folder::~rvrp_folder): Call disable_ranger.
15424 (hybrid_folder::hybrid_folder): Call enable_ranger.
15425 (hybrid_folder::~hybrid_folder): Call disable_ranger.
15427 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
15429 * function.c (allocate_struct_function): Set cfun->x_range_query.
15430 * function.h (struct function): Declare x_range_query.
15431 (get_range_query): New.
15432 (get_global_range_query): New.
15433 * gimple-range-cache.cc (ssa_global_cache::ssa_global_cache):
15434 Remove call to safe_grow_cleared.
15435 * gimple-range.cc (get_range_global): New.
15436 (gimple_range_global): Move from gimple-range.h.
15437 (get_global_range_query): New.
15438 (global_range_query::range_of_expr): New.
15439 (enable_ranger): New.
15440 (disable_ranger): New.
15441 * gimple-range.h (gimple_range_global): Move to gimple-range.cc.
15442 (class global_range_query): New.
15443 (enable_ranger): New.
15444 (disable_ranger): New.
15445 * gimple-ssa-evrp.c (evrp_folder::~evrp_folder): Rename
15446 dump_all_value_ranges to dump.
15447 * tree-vrp.c (vrp_prop::finalize): Same.
15448 * value-query.cc (range_query::dump): New.
15449 * value-query.h (range_query::dump): New.
15450 * vr-values.c (vr_values::dump_all_value_ranges): Rename to...
15451 (vr_values::dump): ...this.
15452 * vr-values.h (class vr_values): Rename dump_all_value_ranges to
15453 dump and make virtual.
15455 2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
15457 * config/i386/i386.c (ix86_autovectorize_vector_modes):
15458 Add V4QImode and V16QImode for TARGET_SSE2.
15459 * doc/sourcebuild.texi (Vector-specific attributes):
15460 Add vect64 and vect32 description.
15462 2021-05-26 Bernd Edlinger <bernd.edlinger@hotmail.de>
15464 * gimple-range-gori.cc (range_def_chain::register_dependency):
15465 Resize m_def_chain when needed.
15467 2021-05-26 Christophe Lyon <christophe.lyon@linaro.org>
15469 * config/arm/mve.md (mve_vaddvq_<supf><mode>): Prefix with '@'.
15470 * config/arm/neon.md (reduc_plus_scal_<mode>): Move to ..
15471 * config/arm/vec-common.md: .. here. Add support for MVE.
15473 2021-05-26 Jakub Jelinek <jakub@redhat.com>
15475 * config/epiphany/epiphany.c (epiphany_print_operand_address): Remove
15477 * config/microblaze/microblaze.c (microblaze_legitimize_address,
15479 microblaze_option_override, print_operand): Likewise.
15480 * config/microblaze/microblaze.md (call_internal_plt,
15481 call_value_intern_plt, call_value_intern): Likewise.
15482 * config/arm/aout.h (ASM_OUTPUT_ALIGN): Likewise.
15483 * config/iq2000/iq2000.md (call_internal1, call_value_internal1,
15484 call_value_multiple_internal1): Likewise.
15485 * config/bfin/bfin.c (symbolic_reference_mentioned_p): Likewise.
15487 2021-05-26 Jan-Benedict Glaw <jbglaw@lug-owl.de>
15489 * config/arc/arc.c (arc_address_cost, arc_print_operand_address,
15490 arc_ccfsm_advance, symbolic_reference_mentioned_p,
15491 arc_raw_symbolic_reference_mentioned_p): Remove register
15494 2021-05-26 Jakub Jelinek <jakub@redhat.com>
15497 * omp-low.c: Include omp-offload.h.
15498 (create_omp_child_function): If current_function_decl has
15499 "omp declare target" attribute and is_gimple_omp_offloaded,
15500 remove that attribute from the copy of attribute list and
15501 add "omp target entrypoint" attribute instead.
15502 (lower_omp_target): Mark .omp_data_sizes.* and .omp_data_kinds.*
15503 variables for offloading if in omp_maybe_offloaded_ctx.
15504 * omp-offload.c (pass_omp_target_link::execute): Nullify second
15505 argument to GOMP_target_data_ext in offloaded code.
15507 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
15509 * config/csky/csky.c (csky_can_change_mode_class): Delete.
15510 For csky, HF/SF mode use the low bits of VREGS.
15512 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
15514 * gimplify.c (gimplify_decl_expr): Do not clear TREE_READONLY on a
15515 DECL which is a reference for OMP.
15517 2021-05-26 Martin Liska <mliska@suse.cz>
15519 PR gcov-profile/100751
15520 * doc/gcov.texi: Document that __gcov_dump can be called just
15521 once and that __gcov_reset resets run-time counters.
15523 2021-05-26 Martin Liska <mliska@suse.cz>
15525 * doc/install.texi: Port relevant part from install-old.texi
15526 and re-generate list of CPUs and systems.
15528 2021-05-26 Martin Liska <mliska@suse.cz>
15530 * Makefile.in: Remove it.
15531 * doc/include/fdl.texi: Update next/previous chapters.
15532 * doc/install.texi: Likewise.
15533 * doc/install-old.texi: Removed.
15535 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
15537 * config/csky/csky.c (ck810_legitimate_index_p): Support
15538 "base + index" with DF mode.
15539 * config/csky/constraints.md ("Y"): New constraint for memory operands
15540 without index register.
15541 * config/csky/csky_insn_fpuv2.md (fpuv3_movdf): Use "Y" instead of "m"
15542 when mov between memory and general registers, and lower their priority.
15543 * config/csky/csky_insn_fpuv3.md (fpuv2_movdf): Likewise.
15545 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
15547 * config/csky/csky.c (TARGET_PROMOTE_PROTOTYPES): Delete.
15549 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
15551 * config/csky/csky.md (untyped_call): Emit clobber for return
15552 registers to mark them used.
15554 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
15556 * config/csky/csky.md (cskyv2_sextend_ldbs): New.
15558 2021-05-26 Andrew Pinski <apinski@marvell.com>
15560 * match.pd (x < 0 ? ~y : y): New patterns.
15562 2021-05-26 Andrew Pinski <apinski@marvell.com>
15564 * match.pd (A?CST1:CST2): Add simplifcations for A?0:+-1, A?+-1:0,
15565 A?POW2:0 and A?0:POW2.
15567 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15569 * gimple-range-gori.cc (class logical_stmt_cache): Delete
15570 (logical_stmt_cache::logical_stmt_cache ): Delete.
15571 (logical_stmt_cache::~logical_stmt_cache): Delete.
15572 (logical_stmt_cache::cache_entry::dump): Delete.
15573 (logical_stmt_cache::get_range): Delete.
15574 (logical_stmt_cache::cached_name ): Delete.
15575 (logical_stmt_cache::same_cached_name): Delete.
15576 (logical_stmt_cache::cacheable_p): Delete.
15577 (logical_stmt_cache::slot_diagnostics ): Delete.
15578 (logical_stmt_cache::dump): Delete.
15579 (gori_compute_cache::gori_compute_cache): Delete.
15580 (gori_compute_cache::~gori_compute_cache): Delete.
15581 (gori_compute_cache::compute_operand_range): Delete.
15582 (gori_compute_cache::cache_stmt): Delete.
15583 * gimple-range-gori.h (gori_compute::compute_operand_range): Remove
15585 (class gori_compute_cache): Delete.
15587 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15589 * gimple-range.cc (fold_using_range::range_of_range_op): Use m_gori
15591 (fold_using_range::range_of_address): Adjust.
15592 (fold_using_range::range_of_phi): Adjust.
15593 * gimple-range.h (class fur_source): Adjust.
15594 (fur_source::fur_source): Adjust.
15596 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15598 * gimple-range-gori.cc (gori_compute::expr_range_at_stmt): Rename
15599 from expr_range_in_bb and adjust.
15600 (gori_compute::compute_name_range_op): Adjust.
15601 (gori_compute::optimize_logical_operands): Adjust.
15602 (gori_compute::compute_logical_operands_in_chain): Adjust.
15603 (gori_compute::compute_operand1_range): Adjust.
15604 (gori_compute::compute_operand2_range): Adjust.
15605 (ori_compute_cache::cache_stmt): Adjust.
15606 * gimple-range-gori.h (gori_compute): Rename prototype.
15608 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15610 * gimple-range.cc (gimple_ranger::range_of_expr): Non-null should be
15611 checked only after range_of_stmt, not range_on_entry.
15612 (gimple_ranger::range_on_entry): Check for non-null in any
15613 predecessor block, if it is not already non-null.
15614 (gimple_ranger::range_on_exit): DOnt check for non-null after
15615 range on entry call.
15616 (gimple_ranger::dump_bb): New. Split from dump.
15617 (gimple_ranger::dump): Adjust.
15618 * gimple-range.h (class gimple_ranger): Adjust.
15620 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15622 * gimple-range-cache.cc (struct range_timestamp): Delete.
15623 (class temporal_cache): Adjust.
15624 (temporal_cache::get_timestamp): Delete.
15625 (temporal_cache::set_dependency): Delete.
15626 (temporal_cache::temporal_value): Adjust.
15627 (temporal_cache::current_p): Take dependencies as params.
15628 (temporal_cache::set_timestamp): Adjust.
15629 (temporal_cache::set_always_current): Adjust.
15630 (ranger_cache::get_non_stale_global_range): Adjust.
15631 (ranger_cache::register_dependency): Delete.
15632 * gimple-range-cache.h (class range_cache): Adjust.
15634 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15636 * gimple-range-gori.cc (range_def_chain::range_def_chain): init
15638 (range_def_chain::~range_def_chain): Dispose of obstack rather than
15639 each individual bitmap.
15640 (range_def_chain::set_import): New.
15641 (range_def_chain::get_imports): New.
15642 (range_def_chain::chain_import_p): New.
15643 (range_def_chain::register_dependency): Rename from build_def_chain
15645 (range_def_chain::def_chain_in_bitmap_p): New.
15646 (range_def_chain::add_def_chain_to_bitmap): New.
15647 (range_def_chain::has_def_chain): Just check first depenedence.
15648 (range_def_chain::get_def_chain): Process imports, use generic
15649 register_dependency routine.
15650 (range_def_chain::dump): New.
15651 (gori_map::gori_map): Allocate import list.
15652 (gori_map::~gori_map): Release imports.
15653 (gori_map::exports): Check for past allocated block size.
15654 (gori_map::imports): New.
15655 (gori_map::def_chain_in_export_p): Delete.
15656 (gori_map::is_import_p): New.
15657 (gori_map::maybe_add_gori): Handle imports.
15658 (gori_map::dump): Adjust output, add imports.
15659 (gori_compute::has_edge_range_p): Remove def_chain_in_export call.
15660 (gori_export_iterator::gori_export_iterator): New.
15661 (gori_export_iterator::next): New.
15662 (gori_export_iterator::get_name): New.
15663 * gimple-range-gori.h (range_def_chain): Add imports and direct
15664 dependecies via struct rdc.
15665 (range_def_chain::depend1): New.
15666 (range_def_chain::depend2): New.
15667 (class gori_map): Adjust.
15668 (FOR_EACH_GORI_IMPORT_NAME): New.
15669 (FOR_EACH_GORI_EXPORT_NAME): New.
15670 (class gori_export_iterator): New.
15672 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15674 * gimple-range-cache.cc (ranger_cache::ranger_cache): Move initial
15675 export cache filling to here.
15676 * gimple-range-gori.cc (gori_compute::gori_compute) : From Here.
15678 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15680 * gimple-range-gori.cc (range_def_chain): Move to gimple-range-gori.h.
15681 (gori_map): Move to gimple-range-gori.h.
15682 (gori_compute::gori_compute): Adjust.
15683 (gori_compute::~gori_compute): Delete.
15684 (gori_compute::compute_operand_range_switch): Adjust.
15685 (gori_compute::compute_operand_range): Adjust.
15686 (gori_compute::compute_logical_operands): Adjust.
15687 (gori_compute::has_edge_range_p ): Adjust.
15688 (gori_compute::set_range_invariant): Delete.
15689 (gori_compute::dump): Adjust.
15690 (gori_compute::outgoing_edge_range_p): Adjust.
15691 * gimple-range-gori.h (class range_def_chain): Relocate here.
15692 (class gori_map): Relocate here.
15693 (class gori_compute): Inherit from gori_map, and adjust.
15695 2021-05-25 Aldy Hernandez <aldyh@redhat.com>
15697 * value-range.cc (range_tests_legacy): Use
15698 build_nonstandard_integer_type instead of int and short.
15700 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
15702 * gimplify.c (gimplify_decl_expr): Clear TREE_READONLY on the DECL
15703 when really creating an initialization statement for it.
15705 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
15707 * tree-inline.c (setup_one_parameter): Fix thinko in new condition.
15709 2021-05-25 Kito Cheng <kito.cheng@sifive.com>
15711 * config/riscv/riscv.h (ASM_SPEC): Pass -mno-relax.
15713 2021-05-25 Martin Liska <mliska@suse.cz>
15715 PR tree-optimization/92860
15717 * optc-save-gen.awk: Remove exceptions.
15719 2021-05-25 Martin Liska <mliska@suse.cz>
15721 * asan.h (sanitize_coverage_p): New function.
15722 * doc/extend.texi: Document it.
15723 * fold-const.c (fold_range_test): Use sanitize_flags_p
15724 instead of flag_sanitize_coverage.
15725 (fold_truth_andor): Likewise.
15726 * sancov.c: Likewise.
15727 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
15728 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Handle
15729 -fsanitize-coverage when inlining.
15731 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
15733 * config/csky/csky-modes.def : Fix copyright.
15735 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
15737 * config/csky/csky-modes.def : Amend copyright.
15738 * config/csky/csky_insn_fpuv2.md : Likewise.
15739 * config/csky/csky_insn_fpuv3.md : Likewise.
15741 2021-05-25 Richard Biener <rguenther@suse.de>
15743 PR middle-end/100727
15744 * calls.c (initialize_argument_information): Explicitely test
15745 for WITH_SIZE_EXPR.
15746 * gimple-expr.c (mark_addressable): Skip outer WITH_SIZE_EXPR.
15748 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
15750 * config/csky/csky.h (FRAME_POINTER_REGNUM): Use
15751 HARD_FRAME_POINTER_REGNUM and FRAME_POINTER_REGNUM instead of
15752 the signle definition. The signle definition may not work well
15753 at simplify_subreg_regno().
15754 (HARD_FRAME_POINTER_REGNUM): New.
15755 (ELIMINABLE_REGS): Add for HARD_FRAME_POINTER_REGNUM.
15756 * config/csky/csky.c (get_csky_live_regs, csky_can_eliminate,
15757 csky_initial_elimination_offset, csky_expand_prologue,
15758 csky_expand_epilogue): Add for HARD_FRAME_POINTER_REGNUM.
15760 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
15762 * config/csky/csky.c (csky_option_override):
15763 Init csky_arch_isa_features[] in advance, so TARGET_DSP
15764 and TARGET_DIV can be set well.
15766 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
15768 * config/csky/constraints.md ("l", "h"): Delete.
15769 * config/csky/csky.h (reg_class, REG_CLASS_NAMES,
15770 REG_CLASS_CONTENTS): Delete LO_REGS and HI_REGS.
15771 * config/csky/csky.c (regno_reg_classm,
15772 csky_secondary_reload, csky_register_move_cost):
15773 Use HILO_REGS instead of LO_REGS and HI_REGS.
15775 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
15777 * config/csky/constraints.md ("W"): New constriant for mem operand
15778 with base reg, index register.
15779 ("Q"): Renamed and modified "csky_valid_fpuv2_mem_operand" to
15780 "csky_valid_mem_constraint_operand" to deal with both "Q" and "W"
15782 ("Dv"): New constraint for const double value that can be used at
15784 * config/csky/csky-modes.def (HFmode): New mode.
15785 * config/csky/csky-protos.h (csky_valid_fpuv2_mem_operand): Rename
15786 to "csky_valid_mem_constraint_operand" and support new constraint
15788 (csky_get_movedouble_length): New.
15789 (fpuv3_output_move): New.
15790 (fpuv3_const_double): New.
15791 * config/csky/csky.c (csky_option_override): New arch CK860 with fpv3.
15792 (decompose_csky_address): Refine.
15793 (csky_print_operand): New "CONST_DOUBLE" operand.
15794 (csky_output_move): Support fpv3 instructions.
15795 (csky_get_movedouble_length): New.
15796 (fpuv3_output_move): New.
15797 (fpuv3_const_double): New.
15798 (csky_emit_compare): Cover float comparsion.
15799 (csky_emit_compare_float): Refine.
15800 (csky_vaild_fpuv2_mem_operand): Rename to
15801 "csky_valid_mem_constraint_operand" and support new constraint "W".
15802 (ck860_rtx_costs): New.
15803 (csky_rtx_costs): Add the cost calculation of CK860.
15804 (regno_reg_class): New vregs for fpuv3.
15805 (csky_dbx_regno): Likewise.
15806 (csky_cpu_cpp_builtins): New builtin macro for fpuv3.
15807 (csky_conditional_register_usage): Suporrot fpuv3.
15808 (csky_dwarf_register_span): Suporrot fpuv3.
15809 (csky_init_builtins, csky_mangle_type): Support "__fp16" type.
15810 (ck810_legitimate_index_p): Support fp16.
15811 * config/csky/csky.h (TARGET_TLS): ADD CK860.
15812 (CSKY_VREG_P, CSKY_VREG_LO_P, CSKY_VREG_HI_P): Support fpuv3.
15813 (TARGET_SINGLE_FPU): Support fpuv3.
15814 (TARGET_SUPPORT_FPV3): New.
15815 (FIRST_PSEUDO_REGISTER): Change to 202 to hold the new fpuv3 registers.
15816 (FIXED_REGISTERS, CALL_REALLY_USED_REGISTERS, REGISTER_NAMES,
15817 REG_CLASS_CONTENTS): Support fpuv3.
15818 * config/csky/csky.md (movsf): Move to cksy_insn_fpu.md and refine.
15819 (csky_movsf_fpv2): Likewise.
15820 (ck801_movsf): Likewise.
15821 (csky_movsf): Likewise.
15823 (csky_movdf_fpv2): Likewise.
15824 (ck801_movdf): Likewise.
15825 (csky_movdf): Likewise.
15826 (movsicc): Refine. Use "comparison_operatior" instead of
15827 "ordered_comparison_operatior".
15828 (addsicc): Likewise.
15829 (CSKY_FIRST_VFP3_REGNUM, CSKY_LAST_VFP3_REGNUM): New constant.
15830 (call_value_internal_vh): New.
15831 * config/csky/csky_cores.def (CK860): New arch and cpu.
15836 * config/csky/csky_insn_fpu.md: Refactor. Separate all float patterns
15837 into emit-patterns and match-patterns, remain the emit-patterns here,
15838 and move the match-patterns to csky_insn_fpuv2.md or
15839 csky_insn_fpuv3.md.
15840 * config/csky/csky_insn_fpuv2.md: New file for fpuv2 instructions.
15841 * config/csky/csky_insn_fpuv3.md: New file and new patterns for fpuv3
15843 * config/csky/csky_isa.def (fcr): New.
15848 (CK860): New definition for ck860.
15849 * config/csky/csky_tables.opt (ck860): New processors ck860,
15850 ck860f. And new arch ck860.
15855 * config/csky/predicates.md (csky_float_comparsion_operator): Delete
15856 "geu", "gtu", "leu", "ltu", which will never appear at float comparison.
15857 * config/csky/t-csky-elf: Support 860.
15858 * config/csky/t-csky-linux: Likewise.
15859 * doc/md.texi: Add "Q" and "W" constraints for C-SKY.
15861 2021-05-24 Aaron Sawdey <acsawdey@linux.ibm.com>
15863 * config/rs6000/genfusion.pl (gen_logical_addsubf): Refactor to
15864 add generation of logical-add and add-logical fusion pairs.
15865 * config/rs6000/rs6000-cpus.def: Add new fusion to ISA 3.1 mask
15867 * config/rs6000/rs6000.c (rs6000_option_override_internal): Turn on
15868 logical-add and add-logical fusion by default.
15869 * config/rs6000/rs6000.opt: Add -mpower10-fusion-logical-add and
15870 -mpower10-fusion-add-logical options.
15871 * config/rs6000/fusion.md: Regenerate file.
15873 2021-05-24 Aldy Hernandez <aldyh@redhat.com>
15875 * value-range.cc (irange::legacy_equal_p): Check type when
15876 comparing VR_VARYING types.
15877 (range_tests_legacy): Test comparing VARYING ranges of different
15880 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
15882 * config/aarch64/aarch64.c (neoversen1_tunings):
15883 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
15885 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
15887 * config/aarch64/aarch64.c (aarch64_classify_symbol): Use GOT for
15888 extern weak symbols. Limit symbol offsets for non-GOT symbols with
15891 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
15893 * config/arm/neon.md (vec_load_lanesxi<mode>)
15894 (vec_store_lanexoi<mode>): Move ...
15895 * config/arm/vec-common.md: here.
15897 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
15899 * config/arm/neon.md (vec_load_lanesoi<mode>)
15900 (vec_store_lanesoi<mode>): Move ...
15901 * config/arm/vec-common.md: here.
15903 2021-05-24 liuhongt <hongtao.liu@intel.com>
15906 * config/i386/i386.c (ix86_gimple_fold_builtin): Replacing
15907 stmt with GIMPLE_NOP when lhs doesn't exist.
15909 2021-05-23 Uroš Bizjak <ubizjak@gmail.com>
15912 * config/i386/mmx.md (*push<VI_32:mode>2_rex64):
15913 New instruction pattern.
15914 (*push<VI_32:mode>2): Ditto.
15915 (push splitter for SSE registers): New splitter.
15917 2021-05-23 Andrew Pinski <apinski@marvell.com>
15919 * match.pd ((A & C) != 0 ? D : 0): Limit to non pointer types.
15921 2021-05-22 Aaron Sawdey <acsawdey@linux.ibm.com>
15923 * config/rs6000/genfusion.pl (gen_addadd): Fix incorrect attr types.
15924 * config/rs6000/fusion.md: Regenerate file.
15926 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
15928 * config/rs6000/genfusion.pl (gen_addadd): New function.
15929 * config/rs6000/fusion.md: Regenerate file.
15930 * config/rs6000/rs6000-cpus.def: Add
15931 OPTION_MASK_P10_FUSION_2ADD to masks.
15932 * config/rs6000/rs6000.c (rs6000_option_override_internal):
15933 Handle default value of OPTION_MASK_P10_FUSION_2ADD.
15934 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2add.
15936 2021-05-21 Jakub Jelinek <jakub@redhat.com>
15938 PR middle-end/99928
15939 * tree.h (OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET): Define.
15940 * gimplify.c (enum gimplify_omp_var_data): Fix up
15941 GOVD_MAP_HAS_ATTACHMENTS value, add GOVD_FIRSTPRIVATE_IMPLICIT.
15942 (omp_lastprivate_for_combined_outer_constructs): If combined target
15943 has GOVD_FIRSTPRIVATE_IMPLICIT set for the decl, change it to
15944 GOVD_MAP | GOVD_SEEN.
15945 (gimplify_scan_omp_clauses): Set GOVD_FIRSTPRIVATE_IMPLICIT for
15946 firstprivate clauses with OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT.
15947 (gimplify_adjust_omp_clauses): For firstprivate clauses with
15948 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT either clear that bit and
15949 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET too, or remove it and
15950 let it be replaced by implicit map clause.
15952 2021-05-21 Jakub Jelinek <jakub@redhat.com>
15954 PR middle-end/99928
15955 * gimplify.c (omp_lastprivate_for_combined_outer_constructs): New
15957 (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LASTPRIVATE>: Use it.
15958 (gimplify_omp_for): Likewise.
15960 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15962 PR middle-end/90115
15963 * omp-low.c (oacc_privatization_candidate_p): Reject 'static',
15964 'external' in blocks.
15966 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15968 PR middle-end/90115
15969 * flag-types.h (enum openacc_privatization): New.
15970 * params.opt (-param=openacc-privatization): New.
15971 * doc/invoke.texi (openacc-privatization): Document it.
15972 * omp-general.h (get_openacc_privatization_dump_flags): New
15974 * omp-low.c (oacc_privatization_candidate_p): Add diagnostics.
15975 * omp-offload.c (execute_oacc_device_lower)
15976 <IFN_UNIQUE_OACC_PRIVATE>: Re-work diagnostics.
15977 * target.def (goacc.adjust_private_decl): Add 'location_t'
15979 * doc/tm.texi: Regenerate.
15980 * config/gcn/gcn-protos.h (gcn_goacc_adjust_private_decl): Adjust.
15981 * config/gcn/gcn-tree.c (gcn_goacc_adjust_private_decl): Likewise.
15982 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl):
15983 Likewise. Preserve it for...
15984 (nvptx_goacc_expand_var_decl): ... use here.
15986 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15988 * doc/sourcebuild.texi (Other attributes): Document '__OPTIMIZE__'
15991 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15993 PR middle-end/90115
15994 * omp-low.c (oacc_privatization_candidate_p): New function.
15995 (oacc_privatization_scan_clause_chain)
15996 (oacc_privatization_scan_decl_chain): Use it. Also
15997 'gcc_checking_assert' that we're not seeing duplicates.
15999 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16001 PR middle-end/90115
16002 * omp-offload.c (execute_oacc_device_lower): Skip processing if no
16005 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16007 PR middle-end/90115
16008 * omp-offload.c (execute_oacc_device_lower): Explain.
16010 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16012 PR middle-end/90115
16013 * omp-offload.c (execute_oacc_device_lower)
16014 <IFN_UNIQUE_OACC_PRIVATE>: Diagnose and handle for 'level == -1'
16016 * internal-fn.c (expand_UNIQUE): Don't expect
16017 'IFN_UNIQUE_OACC_PRIVATE'.
16019 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16021 PR middle-end/90115
16022 * omp-low.c (lower_omp_for): Don't evaluate OpenMP 'for' clauses.
16024 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16026 PR middle-end/90115
16027 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl)
16028 (nvptx_goacc_expand_var_decl): Tighten.
16030 2021-05-21 Julian Brown <julian@codesourcery.com>
16031 Chung-Lin Tang <cltang@codesourcery.com>
16032 Thomas Schwinge <thomas@codesourcery.com>
16034 PR middle-end/90115
16035 * doc/tm.texi.in (TARGET_GOACC_EXPAND_VAR_DECL)
16036 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Add documentation hooks.
16037 * doc/tm.texi: Regenerate.
16038 * expr.c (expand_expr_real_1): Expand decls using the
16039 expand_var_decl OpenACC hook if defined.
16040 * internal-fn.c (expand_UNIQUE): Handle IFN_UNIQUE_OACC_PRIVATE.
16041 * internal-fn.h (IFN_UNIQUE_CODES): Add OACC_PRIVATE.
16042 * omp-low.c (omp_context): Add oacc_privatization_candidates
16044 (lower_oacc_reductions): Add PRIVATE_MARKER parameter. Insert
16046 (lower_oacc_head_tail): Add PRIVATE_MARKER parameter. Modify
16047 private marker's gimple call arguments, and pass it to
16048 lower_oacc_reductions.
16049 (oacc_privatization_scan_clause_chain)
16050 (oacc_privatization_scan_decl_chain, lower_oacc_private_marker):
16052 (lower_omp_for, lower_omp_target, lower_omp_1): Use these.
16053 * omp-offload.c (convert.h): Include.
16054 (oacc_loop_xform_head_tail): Treat private-variable markers like
16055 fork/join when transforming head/tail sequences.
16056 (struct var_decl_rewrite_info): Add struct.
16057 (oacc_rewrite_var_decl, is_sync_builtin_call): New functions.
16058 (execute_oacc_device_lower): Support rewriting gang-private
16059 variables using target hook, and fix up addr_expr and var_decl
16061 * target.def (adjust_private_decl, expand_var_decl): New hooks.
16062 * config/gcn/gcn-protos.h (gcn_goacc_adjust_gangprivate_decl):
16064 (gcn_goacc_adjust_private_decl): ...this.
16065 * config/gcn/gcn-tree.c (gcn_goacc_adjust_gangprivate_decl):
16067 (gcn_goacc_adjust_private_decl): ...this. Add LEVEL parameter.
16068 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_GANGPRIVATE_DECL): Rename
16069 definition using gcn_goacc_adjust_gangprivate_decl...
16070 (TARGET_GOACC_ADJUST_PRIVATE_DECL): ...to this, using
16071 gcn_goacc_adjust_private_decl.
16072 * config/nvptx/nvptx.c (tree-pretty-print.h): Include.
16073 (gang_private_shared_size): New global variable.
16074 (gang_private_shared_align): Likewise.
16075 (gang_private_shared_sym): Likewise.
16076 (gang_private_shared_hmap): Likewise.
16077 (nvptx_option_override): Initialize these.
16078 (nvptx_file_end): Output gang_private_shared_sym.
16079 (nvptx_goacc_adjust_private_decl, nvptx_goacc_expand_var_decl):
16081 (nvptx_set_current_function): Clear gang_private_shared_hmap.
16082 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Define hook.
16083 (TARGET_GOACC_EXPAND_VAR_DECL): Likewise.
16085 2021-05-21 H.J. Lu <hjl.tools@gmail.com>
16087 * config/i386/i386-modes.def (MAX_BITSIZE_MODE_ANY_INT): Removed.
16089 2021-05-21 Richard Biener <rguenther@suse.de>
16090 H.J. Lu <hjl.tools@gmail.com>
16092 PR middle-end/90773
16093 * expr.c (expand_constructor): Elide expand_constructor if
16094 move by pieces is preferred.
16096 2021-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16098 * config/aarch64/aarch64-builtins.c (aarch64_call_properties):
16099 Take a flag and mode value as arguments.
16100 (aarch64_modifies_global_state_p): Likewise.
16101 (aarch64_reads_global_state_p): Likewise.
16102 (aarch64_could_trap_p): Likewise.
16103 (aarch64_get_attributes): Likewise.
16104 (aarch64_init_simd_builtins): Adjust callsite of above.
16105 (aarch64_init_fcmla_laneq_builtins): Use aarch64_get_attributes to get
16106 function attributes to apply to builtins.
16107 (aarch64_init_crc32_builtins): Likewise.
16108 (aarch64_init_builtin_rsqrt): Likewise.
16110 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
16112 * config/rs6000/rs6000.md (define_attr "type"): Add types for fusion.
16113 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10): Use new fusion types.
16114 (gen_2logical): Use new fusion types.
16115 * config/rs6000/fusion.md: Regenerate.
16117 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
16120 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
16121 Handle V4QI and V2HI modes.
16122 (ix86_expand_sse_movcc): Ditto.
16123 * config/i386/mmx.md (*<sat_plusminus:insn><VI_32:mode>3):
16124 New instruction pattern.
16125 (*eq<VI_32:mode>3): Ditto.
16126 (*gt<VI_32:mode>3): Ditto.
16127 (*xop_pcmov_<VI_32:mode>): Ditto.
16128 (mmx_pblendvb32): Ditto.
16129 (mmx_pblendvb64): Rename from mmx_pblendvb.
16130 (vec_cmp<VI_32:mode><VI_32:mode>): New expander.
16131 (vec_cmpu<VI_32:mode><VI_32:mode>): Ditto.
16132 (vcond<VI_32:mode><VI_32:mode>): Ditto.
16133 (vcondu<VI_32:mode><VI_32:mode>): Ditto.
16134 (vcond_mask_<VI_32:mode><VI_32:mode>): Ditto.
16136 2021-05-21 Jakub Jelinek <jakub@redhat.com>
16138 PR tree-optimization/94589
16139 * tree-ssa-phiopt.c (spaceship_replacement): For integral rhs1 and
16140 rhs2, treat x <= 4 equivalently to x < 5 etc. In cmp1 and cmp2 (if
16141 not the same as cmp3) treat <= the same as < and >= the same as >.
16142 Don't require that cond2_phi_edge is true edge, instead take
16143 false/true edges into account based on cmp1/cmp2 comparison kinds.
16145 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
16148 * config/i386/mmx.md (SMAXMIN_MMXMODEI): New mode iterator.
16149 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Macroize expander
16150 from <smaxmin:code>v4hi3> and <smaxmin:code><MMXMODE14:mode>3
16151 using SMAXMIN_MMXMODEI mode iterator.
16152 (*<smaxmin:code>v4qi3): New insn pattern.
16153 (*<smaxmin:code>v2hi3): Ditto.
16154 (SMAXMIN_VI_32): New mode iterator.
16155 (<smaxmin:code><SMAXMIN_VI_32>mode3): New expander.
16156 (UMAXMIN_MMXMODEI): New mode iterator.
16157 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Macroize expander
16158 from <umaxmin:code>v8qi3> and <umaxmin:code><MMXMODE24:mode>3
16159 using UMAXMIN_MMXMODEI mode iterator.
16160 (*<umaxmin:code>v4qi3): New insn pattern.
16161 (*<umaxmin:code>v2hi3): Ditto.
16162 (UMAXMIN_VI_32): New mode iterator.
16163 (<umaxmin:code><UMAXMIN_VI_32>mode3): New expander.
16164 (abs<VI_32:mode>2): New insn pattern.
16165 (ssse3_abs<MMXMODEI:mode>2, abs<MMXMODEI:mode>2): Move from ...
16166 * config/i386/sse.md: ... here.
16168 2021-05-20 Clement Chigot <clement.chigot@atos.net>
16169 David Edelsohn <dje.gcc@gmail.com>
16171 * collect2.c (scan_prog_file): Issue non-fatal warning for
16174 2021-05-20 Jonathan Wakely <jwakely@redhat.com>
16176 * doc/invoke.texi (-Wno-c++11-extensions)
16177 (-Wno-c++14-extensions, -Wno-c++17-extensions)
16178 (-Wno-c++20-extensions, -Wno-c++23-extensions): Document
16181 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
16183 * config/c6x/c6x.c (c6x_output_file_unwind): Use dwarf_debuginfo_p.
16184 * config/darwin.c (darwin_override_options): Likewise.
16185 * config/i386/cygming.h (DBX_REGISTER_NUMBER): Likewise.
16186 * config/i386/darwin.h (DBX_REGISTER_NUMBER): Likewise.
16187 (DWARF2_FRAME_REG_OUT): Likewise.
16188 * config/mips/mips.c (mips_output_filename): Likewise.
16189 * config/rs6000/rs6000.c (rs6000_xcoff_declare_function_name):
16191 (rs6000_dbx_register_number): Likewise.
16192 * dbxout.c: Include flags.h.
16193 * dwarf2cfi.c (cfi_label_required_p): Likewise.
16194 (dwarf2out_do_frame): Likewise.
16195 * except.c: Include flags.h.
16196 * final.c (dwarf2_debug_info_emitted_p): Likewise.
16197 (final_scan_insn_1): Likewise.
16198 * flags.h (dwarf_debuginfo_p): New function declaration.
16199 * opts.c (dwarf_debuginfo_p): New function definition.
16200 * targhooks.c (default_debug_unwind_info): Use dwarf_debuginfo_p.
16201 * toplev.c (process_options): Likewise.
16203 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
16205 * common.opt: Change type to support bitmasks.
16206 * flag-types.h (enum debug_info_type): Rename enumerator constants.
16207 (NO_DEBUG): New bitmask.
16208 (DBX_DEBUG): Likewise.
16209 (DWARF2_DEBUG): Likewise.
16210 (XCOFF_DEBUG): Likewise.
16211 (VMS_DEBUG): Likewise.
16212 (VMS_AND_DWARF2_DEBUG): Likewise.
16213 * flags.h (debug_set_to_format): New function declaration.
16214 (debug_set_count): Likewise.
16215 (debug_set_names): Likewise.
16216 * opts.c (debug_type_masks): Array of bitmasks for debug formats.
16217 (debug_set_to_format): New function definition.
16218 (debug_set_count): Likewise.
16219 (debug_set_names): Likewise.
16220 (set_debug_level): Update access to debug_type_names.
16221 * toplev.c: Likewise.
16223 2021-05-20 Martin Sebor <msebor@redhat.com>
16225 PR middle-end/100684
16226 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Handle C++ lambda.
16228 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
16231 * config/i386/i386.md (isa): Remove x64_bmi.
16232 (enabled): Remove x64_bmi.
16233 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
16234 Remove general register alternative.
16235 (*andnot<VI_32:mode>3): Ditto.
16236 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
16237 (*<any_logic:code><VI_32:mode>3): Ditto.
16239 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
16241 * config/arm/arm.c: Include head files tree-vectorizer.h and
16244 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
16247 * config/i386/mmx.md (Yv_Yw): Revert adding V4QI and V2HI modes.
16248 (*<plusminus:insn><VI_32:mode>3): Use Yw instad of <Yv_Yw> constrint.
16249 (<s>mulv4hi3_highpart): New expander.
16250 (*<s>mulv2hi3_highpart): New insn pattern.
16251 (<s>mulv2hi3_higpart): New expander.
16252 (*<any_shift:insn>v2hi3): New insn pattern.
16253 (<any_shift:insn>v2hi3): New expander.
16254 * config/i386/sse.md (smulhrsv2hi3): New expander.
16255 (*smulhrsv2hi3): New insn pattern.
16257 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
16259 * doc/invoke.texi (vect-inner-loop-cost-factor): Document new
16261 * params.opt (vect-inner-loop-cost-factor): New.
16262 * targhooks.c (default_add_stmt_cost): Replace hardcoded factor
16263 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR, include head file
16264 tree-vectorizer.h and its required ones.
16265 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Replace
16266 hardcoded factor 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR.
16267 * config/arm/arm.c (arm_add_stmt_cost): Likewise.
16268 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
16269 * config/rs6000/rs6000.c (rs6000_add_stmt_cost): Likewise.
16270 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
16272 (_loop_vec_info::_loop_vec_info): Init inner_loop_cost_factor.
16273 * tree-vectorizer.h (_loop_vec_info): Add inner_loop_cost_factor.
16274 (LOOP_VINFO_INNER_LOOP_COST_FACTOR): New macro.
16276 2021-05-20 Christophe Lyon <christophe.lyon@linaro.org>
16277 Torbjörn Svensson <torbjorn.svensson@st.com>
16280 * doc/cpp.texi (Common Predefined Macros): Document __FILE_NAME__.
16282 2021-05-20 Jakub Jelinek <jakub@redhat.com>
16284 PR middle-end/99928
16285 * gimplify.c (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LINEAR>: For
16286 explicit linear clause when combined with target, make it map(tofrom:)
16287 instead of no clause or firstprivate.
16289 2021-05-20 Jakub Jelinek <jakub@redhat.com>
16291 PR tree-optimization/94589
16292 * match.pd ((X & Y) == X -> (X & ~Y) == 0): Simplify even in presence
16293 of integral conversions.
16295 2021-05-19 Andrew MacLeod <amacleod@redhat.com>
16297 * gimple-range.cc (fur_source::get_operand): New.
16298 (gimple_range_fold): Delete.
16299 (fold_using_range::fold_stmt): Move from gimple_ranger::calc_stmt.
16300 (fold_using_range::range_of_range_op): Move from gimple_ranger.
16301 (fold_using_range::range_of_address): Ditto.
16302 (fold_using_range::range_of_phi): Ditto.
16303 (fold_using_range::range_of_call): Ditto.
16304 (fold_using_range::range_of_builtin_ubsan_call): Move from
16305 range_of_builtin_ubsan_call.
16306 (fold_using_range::range_of_builtin_call): Move from
16307 range_of_builtin_call.
16308 (gimple_ranger::range_of_builtin_call): Delete.
16309 (fold_using_range::range_of_cond_expr): Move from gimple_ranger.
16310 (gimple_ranger::fold_range_internal): New.
16311 (gimple_ranger::range_of_stmt): Use new fold_using_range API.
16312 (fold_using_range::range_of_ssa_name_with_loop_info): Move from
16313 gimple_ranger. Improve ranges of SSA_NAMES when possible.
16314 * gimple-range.h (gimple_ranger): Remove various range_of routines.
16315 (class fur_source): New.
16316 (class fold_using_range): New.
16317 (fur_source::fur_source): New.
16319 * vr-values.c (vr_values::extract_range_basic): Use fold_using_range
16320 instead of range_of_builtin_call.
16322 2021-05-19 Jonathan Wakely <jwakely@redhat.com>
16324 * doc/cpp.texi (Common Predefined Macros): Update documentation
16325 for the __GXX_EXPERIMENTAL_CXX0X__ macro.
16327 2021-05-19 Alex Coplan <alex.coplan@arm.com>
16330 * config/arm/arm.md (nonsecure_call_internal): Always ensure
16331 callee's address is in a register.
16333 2021-05-19 Geng Qi <gengqi@linux.alibaba.com>
16335 * common/config/riscv/riscv-common.c
16336 (riscv_subset_list::parsing_subset_version): Properly parse the letter
16338 (riscv_subset_list::parse_std_ext,
16339 riscv_subset_list::parse_multiletter_ext): To handle errors generated
16340 in riscv_subset_list::parsing_subset_version.
16342 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
16344 * config/aarch64/aarch64-simd.md: Use "neon_move_narrow_q"
16345 type attribute in patterns generating XTN(2).
16347 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
16349 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_pack_trunc_<mode>):
16350 Remove as duplicate of...
16351 (aarch64_xtn<mode>): This.
16352 (aarch64_xtn2<mode>_le): Move position in file.
16353 (aarch64_xtn2<mode>_be): Move position in file.
16354 (aarch64_xtn2<mode>): Move position in file.
16355 (vec_pack_trunc_<mode>): Define as an expander.
16357 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
16359 * config/aarch64/aarch64-simd-builtins.def: Split builtin
16360 generation for aarch64_<sur>q<r>shr<u>n_n<mode> pattern into
16361 separate scalar and vector generators.
16362 * config/aarch64/aarch64-simd.md
16363 (aarch64_<sur>q<r>shr<u>n_n<mode>): Define as an expander and
16365 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_le): This and...
16366 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_be): This.
16367 * config/aarch64/iterators.md: Define SD_HSDI iterator.
16369 2021-05-19 Jonathn Wright <jonathan.wright@arm.com>
16371 * config/aarch64/aarch64-simd.md: Use UNSPEC_SQXTUN instead
16373 * config/aarch64/iterators.md: Remove UNSPEC_SQXTUN2.
16375 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
16377 * config/aarch64/aarch64-simd.md (aarch64_<sur>q<r>shr<u>n2_n<mode>):
16378 Implement as an expand emitting a big/little endian
16379 instruction pattern.
16380 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_le): Define.
16381 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_be): Define.
16383 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
16385 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn2<mode>):
16386 Implement as an expand emitting a big/little endian
16387 instruction pattern.
16388 (aarch64_<sur><addsub>hn2<mode>_insn_le): Define.
16389 (aarch64_<sur><addsub>hn2<mode>_insn_be): Define.
16390 * config/aarch64/iterators.md: Remove UNSPEC_[R]ADDHN2 and
16391 UNSPEC_[R]SUBHN2 unspecs and ADDSUBHN2 iterator.
16393 2021-05-19 Richard Biener <rguenther@suse.de>
16395 PR middle-end/100672
16396 * fold-const.c (fold_negate_expr_1): Use element_precision.
16397 (negate_expr_p): Likewise.
16399 2021-05-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
16401 * config/aarch64/iterators.md (SVE_PRED_LOAD): New iterator.
16402 (pred_load): New int attribute.
16403 * config/aarch64/aarch64-sve.md
16404 (aarch64_load_<ANY_EXTEND:optab><SVE_HSDI:mode><SVE_PARTIAL_I:mode>): Use
16405 SVE_PRED_LOAD enum iterator and corresponding pred_load attribute.
16406 * config/aarch64/aarch64-sve-builtins-base.cc (expand): Update call to
16407 code_for_aarch64_load.
16409 2021-05-19 Richard Biener <rguenther@suse.de>
16411 * cfgexpand.c (discover_nonconstant_array_refs_r): Make
16412 sure TARGET_MEM_REF bases are expanded as memory.
16413 * tree-ssa-operands.c (operands_scanner::get_tmr_operands):
16414 Do not mark TARGET_MEM_REF bases addressable.
16415 * tree-ssa.c (non_rewritable_mem_ref_base): Handle
16416 TARGET_MEM_REF bases as never rewritable.
16417 * gimple-walk.c (walk_stmt_load_store_addr_ops): Do not
16418 walk TARGET_MEM_REF bases as address-takens.
16419 * tree-ssa-dce.c (ref_may_be_aliased): Handle TARGET_MEM_REF.
16421 2021-05-19 Richard Biener <rguenther@suse.de>
16423 * builtins.c (get_object_alignment_1): Strip outer
16425 * tree-dfa.c (get_ref_base_and_extent): Handle outer
16426 WITH_SIZE_EXPR for size processing and process the
16428 * tree-ssa-alias.c (ao_ref_base_alias_set): Strip
16429 outer WITH_SIZE_EXPR.
16430 (ao_ref_base_alias_ptr_type): Likewise.
16431 (refs_may_alias_p_2): Allow WITH_SIZE_EXPR in ref->ref
16432 and handle that accordingly, stripping it for the
16433 core alias workers.
16434 * tree.c (get_base_address): Handle WITH_SIZE_EXPR by
16435 looking through it instead of returning NULL.
16437 2021-05-19 Jakub Jelinek <jakub@redhat.com>
16439 PR middle-end/100576
16440 * builtins.c (check_read_access): Convert bound to size_type_node if
16443 2021-05-19 Richard Biener <rguenther@suse.de>
16445 * tree-cfg.c (verify_types_in_gimple_min_lval): Inline...
16446 (verify_types_in_gimple_reference): ... here. Sanitize.
16447 (verify_gimple_call): Verify references in LHS and arguments.
16448 (verify_gimple_assign_single): Reject WITH_SIZE_EXPR.
16450 2021-05-19 Uroš Bizjak <ubizjak@gmail.com>
16452 * config/i386/i386.h (VALID_INT_MODE_P):
16453 Add V8QI, V4HI and V2SI modes for TARGET_64BIT.
16454 * config/i386/i386.md (isa): Add x64_bmi.
16455 (enabled): Handle x64_bmi.
16456 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
16457 Add alternative using 64bit general registers.
16458 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
16460 2021-05-19 Jakub Jelinek <jakub@redhat.com>
16462 PR middle-end/99928
16463 * tree.h (OMP_MASTER_COMBINED): Define.
16464 * gimplify.c (gimplify_scan_omp_clauses): Rewrite lastprivate
16465 handling for outer combined/composite constructs to a loop.
16466 Handle lastprivate on combined target.
16467 (gimplify_expr): Formatting fix.
16469 2021-05-19 Xionghu Luo <luoxhu@linux.ibm.com>
16471 * passes.def: Add sink_code pass before store_merging.
16472 * tree-ssa-sink.c (pass_sink_code:clone): New.
16474 2021-05-18 Bill Schmidt <wschmidt@linux.ibm.com>
16476 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Remove call to
16477 rs6000_special_adjust_field_align_p.
16478 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
16479 * config/rs6000/rs6000-call.c (rs6000_function_arg_boundary):
16480 Remove ABI warning.
16481 (rs6000_function_arg): Likewise.
16482 * config/rs6000/rs6000-protos.h
16483 (rs6000_special_adjust_field_align_p): Remove prototype.
16484 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
16486 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Remove call to
16487 rs6000_special_adjust_field_align_p.
16489 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
16492 * config/i386/i386.h (VALID_SSE2_REG_MODE):
16493 Add V4QI and V2HI modes.
16494 (VALID_INT_MODE_P): Ditto.
16495 * config/i386/mmx.md (VI_32): New mode iterator.
16496 (mmxvecsize): Handle V4QI and V2HI.
16498 (mov<VI_32:mode>): New expander.
16499 (*mov<mode>_internal): New insn pattern.
16500 (movmisalign<VI_32:mode>): New expander.
16501 (neg<VI_32:mode>): New expander.
16502 (<plusminus:insn><VI_32:mode>3): New expander.
16503 (*<plusminus:insn><VI_32:mode>3): New insn pattern.
16504 (mulv2hi3): New expander.
16505 (*mulv2hi3): New insn pattern.
16506 (one_cmpl<VI_32:mode>2): New expander.
16507 (*andnot<VI_32:mode>3): New insn pattern.
16508 (<any_logic:code><VI_32:mode>3): New expander.
16509 (*<any_logic:code><VI_32:mode>3): New insn pattern.
16511 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
16513 * config/i386/sse.md (<any_extend:insn>v4qiv4di2):
16514 Fix a mode mismatch with operand 1.
16516 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
16519 * config/i386/i386-expand.c (split_double_mode): Return
16520 temporary register when simplify_gen_subreg fails with
16521 the high half od the paradoxical subreg.
16523 2021-05-18 Richard Biener <rguenther@suse.de>
16525 * cfgexpand.c (expand_one_var): Pass in forced_stack_var
16526 and honor it when expanding.
16527 (expand_used_vars_for_block): Pass through forced_stack_var.
16528 (expand_used_vars): Likewise.
16529 (discover_nonconstant_array_refs_r): Set bits in
16530 forced_stack_vars instead of marking vars TREE_ADDRESSABLE.
16531 (avoid_type_punning_on_regs): Likewise.
16532 (discover_nonconstant_array_refs): Likewise.
16533 (pass_expand::execute): Create and pass down forced_stack_var
16534 bitmap. For parameters and returns temporarily set
16535 TREE_ADDRESSABLE when expand_function_start.
16537 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
16539 * doc/sourcebuild.texi: Document 'dg-note'.
16541 2021-05-18 Tobias Burnus <tobias@codesourcery.com>
16544 * configure: Regenerate.
16545 * configure.ac (BUILD_CFLAG, BUILD_CXXFLAGS): Add $(CFLAGS-$@).
16547 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
16549 * gimple.h (is_gimple_omp_oacc): Tighten.
16550 * omp-low.c (check_omp_nesting_restrictions): Adjust.
16552 2021-05-18 Richard Biener <rguenther@suse.de>
16554 * tree-ssa-operands.c (mark_address_taken): Simplify.
16556 2021-05-18 Martin Liska <mliska@suse.cz>
16558 * config/gcn/mkoffload.c (STR): Redefine.
16559 * config/i386/intelmic-mkoffload.c (STR): Likewise.
16560 * config/nvptx/mkoffload.c (STR): Likewise.
16562 2021-05-18 Martin Liska <mliska@suse.cz>
16564 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
16565 Use startswith function instead of strncmp.
16566 * common/config/bfin/bfin-common.c (bfin_handle_option): Likewise.
16567 * common/config/riscv/riscv-common.c (riscv_subset_list::parse): Likewise.
16568 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): Likewise.
16569 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Likewise.
16570 * config/alpha/alpha.c (alpha_elf_section_type_flags): Likewise.
16571 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
16572 * config/arm/arm.c (arm_file_start): Likewise.
16573 (arm_valid_target_attribute_rec): Likewise.
16574 (thumb1_md_asm_adjust): Likewise.
16575 * config/arm/driver-arm.c (host_detect_local_cpu): Likewise.
16576 * config/avr/avr.c (STR_PREFIX_P): Likewise.
16577 (avr_set_current_function): Likewise.
16578 (avr_handle_addr_attribute): Likewise.
16579 (avr_asm_output_aligned_decl_common): Likewise.
16580 (avr_asm_named_section): Likewise.
16581 (avr_section_type_flags): Likewise.
16582 (avr_asm_select_section): Likewise.
16583 * config/c6x/c6x.c (c6x_in_small_data_p): Likewise.
16584 (c6x_section_type_flags): Likewise.
16585 * config/darwin-c.c (darwin_cfstring_ref_p): Likewise.
16586 (darwin_objc_declare_unresolved_class_reference): Likewise.
16587 (darwin_objc_declare_class_definition): Likewise.
16588 * config/darwin.c (indirect_data): Likewise.
16589 (darwin_encode_section_info): Likewise.
16590 (darwin_objc2_section): Likewise.
16591 (darwin_objc1_section): Likewise.
16592 (machopic_select_section): Likewise.
16593 (darwin_globalize_label): Likewise.
16594 (darwin_label_is_anonymous_local_objc_name): Likewise.
16595 (darwin_asm_named_section): Likewise.
16596 (darwin_asm_output_dwarf_offset): Likewise.
16597 * config/frv/frv.c (frv_string_begins_with): Likewise.
16598 (frv_in_small_data_p): Likewise.
16599 * config/gcn/mkoffload.c (STR): Likewise.
16601 * config/i386/i386-builtins.c (get_builtin_code_for_version): Likewise.
16602 * config/i386/i386-options.c (ix86_option_override_internal): Likewise.
16603 * config/i386/i386.c (x86_64_elf_section_type_flags): Likewise.
16604 (ix86_md_asm_adjust): Likewise.
16605 * config/i386/intelmic-mkoffload.c (STR): Likewise.
16606 * config/i386/winnt.c (i386_pe_asm_named_section): Likewise.
16607 (i386_pe_file_end): Likewise.
16608 * config/ia64/ia64.c (ia64_in_small_data_p): Likewise.
16609 (ia64_section_type_flags): Likewise.
16610 * config/mips/driver-native.c (host_detect_local_cpu): Likewise.
16611 * config/mips/mips.c (mips_handle_interrupt_attr): Likewise.
16612 (mips16_stub_function_p): Likewise.
16613 (mips_function_rodata_section): Likewise.
16614 * config/msp430/msp430.c (msp430_mcu_name): Likewise.
16615 (msp430_function_section): Likewise.
16616 (msp430_section_type_flags): Likewise.
16617 (msp430_expand_helper): Likewise.
16618 * config/nios2/nios2.c (nios2_small_section_name_p): Likewise.
16619 (nios2_valid_target_attribute_rec): Likewise.
16620 * config/nvptx/mkoffload.c (process): Likewise.
16622 * config/pa/som.h: Likewise.
16623 * config/pdp11/pdp11.c (pdp11_output_ident): Likewise.
16624 * config/riscv/riscv.c (riscv_elf_select_rtx_section): Likewise.
16625 * config/rs6000/rs6000.c (VTABLE_NAME_P): Likewise.
16626 (rs6000_inner_target_options): Likewise.
16627 * config/s390/driver-native.c (s390_host_detect_local_cpu): Likewise.
16628 * config/sparc/driver-sparc.c (host_detect_local_cpu): Likewise.
16629 * config/vax/vax.c (vax_output_int_move): Likewise.
16630 * config/vms/vms-ld.c (startswith): Likewise.
16631 (process_args): Likewise.
16633 * config/vms/vms.c: Likewise.
16635 2021-05-18 Jakub Jelinek <jakub@redhat.com>
16637 PR rtl-optimization/100590
16638 * regcprop.c (copyprop_hardreg_forward_1): Only DCE dead sets if
16639 they are NONJUMP_INSN_P.
16641 2021-05-18 Jakub Jelinek <jakub@redhat.com>
16644 * function.c (push_dummy_function): Set DECL_ARTIFICIAL and
16645 DECL_ASSEMBLER_NAME on the fn_decl.
16647 2021-05-18 Jakub Jelinek <jakub@redhat.com>
16649 PR tree-optimization/94589
16650 * tree-ssa-phiopt.c (spaceship_replacement): Pattern match
16651 phi result used in (res & ~1) == 0 comparison as res >= 0 as
16652 res == 2 would be UB with -ffinite-math-only.
16654 2021-05-18 Martin Liska <mliska@suse.cz>
16656 * Makefile.in: genversion.o should depend on DATESTAMP.
16658 2021-05-18 Claudiu Zissulescu <claziss@synopsys.com>
16660 * config/arc/simdext.md (negv2si2): Remove round bracket.
16662 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
16664 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Define
16665 _Bool as macro expanding to _Bool.
16667 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
16670 * tree.c (build_reference_type_for_mode)
16671 (build_pointer_type_for_mode): Pick pointer mode if MODE argument
16673 (build_reference_type, build_pointer_type): Invoke
16674 build_*_type_for_mode with VOIDmode.
16676 2021-05-17 Andrew MacLeod <amacleod@redhat.com>
16678 PR tree-optimization/100512
16679 * gimple-range-cache.cc (ranger_cache::set_global_range): Mark const
16680 and non-zero pointer ranges as invariant.
16681 * gimple-range.cc (gimple_ranger::range_of_stmt): Remove pointer
16682 processing from here.
16684 2021-05-17 Tom de Vries <tdevries@suse.de>
16687 * config/nvptx/nvptx-protos.h (nvptx_output_atomic_insn): Declare
16688 * config/nvptx/nvptx.c (nvptx_output_barrier)
16689 (nvptx_output_atomic_insn): New function.
16690 (nvptx_print_operand): Add support for 'B'.
16691 * config/nvptx/nvptx.md: Use nvptx_output_atomic_insn for atomic
16694 2021-05-17 Aldy Hernandez <aldyh@redhat.com>
16696 PR tree-optimization/100349
16697 * vr-values.c (bounds_of_var_in_loop): Bail if scev returns
16700 2021-05-17 Tamar Christina <tamar.christina@arm.com>
16702 * config/aarch64/driver-aarch64.c (DEFAULT_ARCH): New.
16703 (host_detect_local_cpu): Use it.
16705 2021-05-17 Martin Liska <mliska@suse.cz>
16707 * doc/invoke.texi: Add 2 missing dots.
16709 2021-05-17 Marius Hillenbrand <mhillen@linux.ibm.com>
16711 PR bootstrap/100552
16712 * configure.ac: Replace pattern substitution with call to sed.
16713 * configure: Regenerate.
16715 2021-05-17 Richard Biener <rguenther@suse.de>
16717 PR middle-end/100582
16718 * tree.c (array_at_struct_end_p): Get to the base of the
16719 reference before looking for the underlying decl.
16721 2021-05-17 Joern Rennecke <joern.rennecke@embecosm.com>
16723 * genoutput.c (validate_insn_alternatives) Make "wrong number of
16724 alternatives" message more specific, and remove assumption on where
16727 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
16729 * config/arm/iterators.md (V16): New iterator.
16730 (VH_cvtto): New iterator.
16731 (v_cmp_result): Added V4HF and V8HF support.
16732 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result>): Use VDQWH.
16733 (vcond<mode><mode>): Likewise.
16734 (vcond_mask_<mode><v_cmp_result>): Likewise.
16735 (vcond<VH_cvtto><mode>): New expander.
16737 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
16739 * config/arm/arm-protos.h (arm_expand_vector_compare): Update
16741 * config/arm/arm.c (arm_expand_vector_compare): Add support for
16743 (arm_expand_vcond): Likewise.
16744 * config/arm/iterators.md (supf): Remove VCMPNEQ_S, VCMPEQQ_S,
16745 VCMPEQQ_N_S, VCMPNEQ_N_S.
16746 (VCMPNEQ, VCMPEQQ, VCMPEQQ_N, VCMPNEQ_N): Remove.
16747 * config/arm/mve.md (@mve_vcmp<mve_cmp_op>q_<mode>): Add '@' prefix.
16748 (@mve_vcmp<mve_cmp_op>q_f<mode>): Likewise.
16749 (@mve_vcmp<mve_cmp_op>q_n_f<mode>): Likewise.
16750 (@mve_vpselq_<supf><mode>): Likewise.
16751 (@mve_vpselq_f<mode>"): Likewise.
16752 * config/arm/neon.md (vec_cmp<mode><v_cmp_result): Enable for MVE
16753 and move to vec-common.md.
16754 (vec_cmpu<mode><mode>): Likewise.
16755 (vcond<mode><mode>): Likewise.
16756 (vcond<V_cvtto><mode>): Likewise.
16757 (vcondu<mode><v_cmp_result>): Likewise.
16758 (vcond_mask_<mode><v_cmp_result>): Likewise.
16759 * config/arm/unspecs.md (VCMPNEQ_U, VCMPNEQ_S, VCMPEQQ_S)
16760 (VCMPEQQ_N_S, VCMPNEQ_N_S, VCMPEQQ_U, CMPEQQ_N_U, VCMPNEQ_N_U)
16761 (VCMPGEQ_N_S, VCMPGEQ_S, VCMPGTQ_N_S, VCMPGTQ_S, VCMPLEQ_N_S)
16762 (VCMPLEQ_S, VCMPLTQ_N_S, VCMPLTQ_S, VCMPCSQ_N_U, VCMPCSQ_U)
16763 (VCMPHIQ_N_U, VCMPHIQ_U): Remove.
16764 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result): Moved
16766 (vec_cmpu<mode><mode>): Likewise.
16767 (vcond<mode><mode>): Likewise.
16768 (vcond<V_cvtto><mode>): Likewise.
16769 (vcondu<mode><v_cmp_result>): Likewise.
16770 (vcond_mask_<mode><v_cmp_result>): Likewise. Added unsafe math
16773 2021-05-17 liuhongt <hongtao.liu@intel.com>
16776 * config/i386/i386.c (ix86_gimple_fold_builtin): Use
16777 gsi_insert_seq_before instead.
16779 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
16781 * doc/sourcebuild.texi (arm_qbit_ok): Rename into...
16782 (arm_sat_ok): ...this.
16784 2021-05-17 Martin Liska <mliska@suse.cz>
16786 * lto-wrapper.c (merge_flto_options): Factor out a new function.
16787 (merge_and_complain): Use it.
16788 (run_gcc): Merge also linker command line -flto=foo argument
16791 2021-05-16 Christophe Lyon <christophe.lyon@linaro.org>
16793 * config/arm/arm.h (CPP_SPEC): Remove error message about
16794 -mlittle-endian/-mbig-endian conflict.
16796 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
16798 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
16799 __ROP_PROTECT__ if -mrop-protect is selected.
16801 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
16803 * config/rs6000/rs6000-internal.h (rs6000_stack): Add
16804 rop_hash_save_offset and rop_hash_size.
16805 * config/rs6000/rs6000-logue.c (rs6000_stack_info): Compute
16806 rop_hash_size and rop_hash_save_offset.
16807 (debug_stack_info): Dump rop_hash_save_offset and rop_hash_size.
16808 (rs6000_emit_prologue): Emit hashst[p] in prologue.
16809 (rs6000_emit_epilogue): Emit hashchk[p] in epilogue.
16810 * config/rs6000/rs6000.md (unspec): Add UNSPEC_HASHST and
16812 (hashst): New define_insn.
16813 (hashchk): Likewise.
16815 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
16817 * config/rs6000/rs6000.c (rs6000_option_override_internal):
16818 Disable shrink wrap when inserting ROP-protect instructions.
16819 * config/rs6000/rs6000.opt (mrop-protect): New option.
16820 (mprivileged): Likewise.
16821 * doc/invoke.texi: Document mrop-protect and mprivileged.
16823 2021-05-15 Hans-Peter Nilsson <hp@axis.com>
16825 * reorg.c (fill_slots_from_thread): Reinstate code typoed out in
16828 2021-05-15 Martin Jambor <mjambor@suse.cz>
16831 2021-05-13 Martin Jambor <mjambor@suse.cz>
16833 PR tree-optimization/100453
16834 * tree-sra.c (sra_modify_assign): All const base accesses do not
16835 need refreshing, not just those from decl_pool.
16836 (sra_modify_assign): Do not refresh into a const base decl.
16838 2021-05-15 Jakub Jelinek <jakub@redhat.com>
16840 PR rtl-optimization/100342
16841 * regcprop.c (copy_value): When copying a source reg in a wider
16842 mode than it has recorded for the value, adjust recorded destination
16843 mode too or punt if !REG_CAN_CHANGE_MODE_P.
16845 2021-05-14 Jason Merrill <jason@redhat.com>
16847 * intl.h: Add comments.
16849 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16851 * config/aarch64/aarch64-simd.md
16852 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Split into...
16853 (aarch64_sqdmlsl2_lane<mode>_internal): ... This...
16854 (aarch64_sqdmlal2_lane<mode>_internal): ... And this.
16855 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Split into ...
16856 (aarch64_sqdmlsl2_laneq<mode>_internal): ... This...
16857 (aarch64_sqdmlal2_laneq<mode>_internal): ... And this.
16858 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>_internal): Split into...
16859 (aarch64_sqdmlsl2_n<mode>_internal): ... This...
16860 (aarch64_sqdmlal2_n<mode>_internal): ... And this.
16862 2021-05-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16865 * config/arm/arm_neon.h (vtst_s8): Replace call to vtst builtin with it's
16866 boolean logic equivalent.
16867 (vtst_s16): Likewise.
16868 (vtst_s32): Likewise.
16869 (vtst_u8): Likewise.
16870 (vtst_u16): Likewise.
16871 (vtst_u32): Likewise.
16872 (vtst_p8): Likewise.
16873 (vtst_p16): Likewise.
16874 (vtstq_s8): Likewise.
16875 (vtstq_s16): Likewise.
16876 (vtstq_s32): Likewise.
16877 (vtstq_u8): Likewise.
16878 (vtstq_u16): Likewise.
16879 (vtstq_u32): Likewise.
16880 (vtstq_p8): Likewise.
16881 (vtstq_p16): Likewise.
16882 * config/arm/arm_neon_builtins.def: Remove entry for vtst.
16883 * config/arm/neon.md (neon_vtst<mode>): Remove pattern.
16885 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16887 * config/aarch64/aarch64-simd.md (aarch64_sqdmlal2<mode>): Merge into...
16888 (aarch64_sqdml<SBINQOPS:as>l2<mode>): ... This.
16889 (aarch64_sqdmlsl2<mode>): Delete.
16890 (aarch64_sqdmlal2_lane<mode>): Merge this...
16891 (aarch64_sqdmlsl2_lane<mode>): ... And this...
16892 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>): ... Into this.
16893 (aarch64_sqdmlal2_laneq<mode>): Merge this...
16894 (aarch64_sqdmlsl2_laneq<mode>): ... And this...
16895 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>): ... Into this.
16896 (aarch64_sqdmlal2_n<mode>): Merge this...
16897 (aarch64_sqdmlsl2_n<mode>): ... And this...
16898 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>): ... Into this.
16900 2021-05-13 Martin Sebor <msebor@redhat.com>
16902 PR middle-end/100574
16903 * builtins.c (access_ref::get_ref): Improve detection of PHIs with
16904 all null arguments.
16906 2021-05-13 Martin Sebor <msebor@redhat.com>
16908 PR tree-optimization/93100
16909 PR middle-end/98583
16910 * tree-ssa-uninit.c (check_defs): Exclude intrinsic functions that
16911 don't modify referenced objects.
16913 2021-05-13 Martin Jambor <mjambor@suse.cz>
16915 PR tree-optimization/100453
16916 * tree-sra.c (sra_modify_assign): All const base accesses do not
16917 need refreshing, not just those from decl_pool.
16918 (sra_modify_assign): Do not refresh into a const base decl.
16920 2021-05-13 Martin Liska <mliska@suse.cz>
16922 * tree-ssa-dom.c: Remove m_simplifier.
16924 2021-05-13 Richard Earnshaw <rearnsha@arm.com>
16927 * config/arm/arm.c (arm_canonicalize_comparison): Correctly
16928 canonicalize DImode inequality comparisons against the
16929 maximum integral value.
16931 2021-05-13 Jakub Jelinek <jakub@redhat.com>
16933 PR tree-optimization/98856
16934 * config/i386/i386.c (ix86_shift_rotate_cost): Add CODE argument.
16935 Expect V2DI and V4DI arithmetic right shifts to be emulated.
16936 (ix86_rtx_costs, ix86_add_stmt_cost): Adjust ix86_shift_rotate_cost
16938 * config/i386/i386-expand.c (expand_vec_perm_2perm_interleave,
16939 expand_vec_perm_2perm_pblendv): New functions.
16940 (ix86_expand_vec_perm_const_1): Use them.
16941 * config/i386/sse.md (ashr<mode>3<mask_name>): Rename to ...
16942 (<mask_codefor>ashr<mode>3<mask_name>): ... this.
16943 (ashr<mode>3): New define_expand with VI248_AVX512BW iterator.
16944 (ashrv4di3): New define_expand.
16945 (ashrv2di3): Change condition to TARGET_SSE2, handle !TARGET_XOP
16946 and !TARGET_AVX512VL expansion.
16948 2021-05-13 Uroš Bizjak <ubizjak@gmail.com>
16951 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Force mode
16952 sizes < 16 to a register when constructing vpcmov pattern.
16953 * config/i386/mmx.md (*xop_pcmov_<mode>): Use MMXMODE124 mode.
16955 2021-05-13 Martin Liska <mliska@suse.cz>
16957 * gcov-io.c (gcov_write_block): Remove.
16958 (gcov_write_words): Likewise.
16959 (gcov_read_words): Re-implement using gcov_read_bytes.
16960 (gcov_allocate): Remove.
16961 (GCOV_BLOCK_SIZE): Likewise.
16962 (struct gcov_var): Remove most of the fields.
16963 (gcov_position): Implement with ftell.
16964 (gcov_rewrite): Remove setting of start and offset fields.
16965 (from_file): Re-format.
16966 (gcov_open): Remove setbuf call. It should not be needed.
16967 (gcov_close): Remove internal buffer handling.
16968 (gcov_magic): Use __builtin_bswap32.
16969 (gcov_write_counter): Use directly gcov_write_unsigned.
16970 (gcov_write_string): Use direct fwrite and do not round
16972 (gcov_seek): Use directly fseek.
16973 (gcov_write_tag): Use gcov_write_unsigned directly.
16974 (gcov_write_length): Likewise.
16975 (gcov_write_tag_length): Likewise.
16976 (gcov_read_bytes): Use directly fread.
16977 (gcov_read_unsigned): Use gcov_read_words.
16978 (gcov_read_counter): Likewise.
16979 (gcov_read_string): Use gcov_read_bytes.
16980 * gcov-io.h (GCOV_WORD_SIZE): Adjust to reflect
16981 that size is not in bytes, but words (4B).
16982 (GCOV_TAG_FUNCTION_LENGTH): Likewise.
16983 (GCOV_TAG_ARCS_LENGTH): Likewise.
16984 (GCOV_TAG_ARCS_NUM): Likewise.
16985 (GCOV_TAG_COUNTER_LENGTH): Likewise.
16986 (GCOV_TAG_COUNTER_NUM): Likewise.
16987 (GCOV_TAG_SUMMARY_LENGTH): Likewise.
16989 2021-05-13 liuhongt <hongtao.liu@intel.com>
16992 * config/i386/sse.md (ssedoublevecmode): Add attribute for
16993 V64QI/V32HI/V16SI/V4DI.
16994 (ssehalfvecmode): Add attribute for V2DI/V2DF.
16995 (*vec_concatv4si_0): Extend to VI124_128.
16996 (*vec_concat<mode>_0): New pre-reload splitter.
16997 * config/i386/predicates.md (movq_parallel): New predicate.
16999 2021-05-13 Alexandre Oliva <oliva@adacore.com>
17001 * targhooks.c (default_zero_call_used_regs): Retry using
17002 successfully-zeroed registers as sources.
17004 2021-05-12 Tobias Burnus <tobias@codesourcery.com>
17006 * omp-low.c (finish_taskreg_scan): Use the proper detach decl.
17008 2021-05-12 Aldy Hernandez <aldyh@redhat.com>
17011 * gimple-range.cc (range_of_builtin_call): Skip out on
17012 processing __builtin_clz when varying.
17014 2021-05-12 Tom de Vries <tdevries@suse.de>
17017 * config/nvptx/nvptx-opts.h (enum ptx_version): New enum.
17018 * config/nvptx/nvptx.c (nvptx_file_start): Print .version according
17019 to ptx_version_option.
17020 * config/nvptx/nvptx.h (TARGET_PTX_6_3): Define.
17021 * config/nvptx/nvptx.md (define_insn "nvptx_shuffle<mode>")
17022 (define_insn "nvptx_vote_ballot"): Use sync variant for
17024 * config/nvptx/nvptx.opt (ptx_version): Add enum.
17025 (mptx): Add option.
17026 * doc/invoke.texi (Nvidia PTX Options): Add mptx item.
17028 2021-05-12 Richard Biener <rguenther@suse.de>
17030 PR tree-optimization/100566
17031 * tree-ssa-sccvn.c (dominated_by_p_w_unex): Properly handle
17032 allow_back for all edge queries.
17034 2021-05-12 liuhongt <hongtao.liu@intel.com>
17037 * config/i386/sse.md (<sse4_1_avx2>_pblendvb): Add
17038 splitters for pblendvb of NOT mask register.
17040 2021-05-12 Richard Biener <rguenther@suse.de>
17042 PR tree-optimization/100519
17043 * tree-ssa-reassoc.c (can_associate_p): Split into...
17044 (can_associate_op_p): ... this
17045 (can_associate_type_p): ... and this.
17046 (is_reassociable_op): Call can_associate_op_p.
17047 (break_up_subtract_bb): Call the appropriate predicates.
17048 (reassociate_bb): Likewise.
17050 2021-05-12 Martin Liska <mliska@suse.cz>
17052 * lto-wrapper.c (merge_and_complain): Merge -flto=arg options.
17053 (run_gcc): Use -flto argument detection for merged
17056 2021-05-12 Martin Liska <mliska@suse.cz>
17058 * lto-wrapper.c (print_lto_docs_link): New function.
17059 (run_gcc): Print warning about missing job server detection
17060 after we know NR of partitions. Do the same for -flto{,=1}.
17061 * opts.c (get_option_html_page): Support -flto option.
17063 2021-05-12 Martin Liska <mliska@suse.cz>
17065 * lto-wrapper.c (get_options_from_collect_gcc_options): Change
17067 (append_option): Remove.
17068 (find_option): Rework to use the vector type.
17069 (remove_option): Remove.
17070 (merge_and_complain): Use vectors for cl_decoded_option data
17072 (append_compiler_options): Likewise.
17073 (append_diag_options): Likewise.
17074 (append_linker_options): Likewise.
17075 (append_offload_options): Likewise.
17076 (compile_offload_image): Likewise.
17077 (compile_images_for_offload_targets): Likewise.
17078 (find_and_merge_options): Likewise.
17079 (run_gcc): Likewise.
17081 2021-05-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
17084 * dwarf2out.c (dwarf2out_finish): Set
17085 have_multiple_function_sections with multi-range text_section.
17087 2021-05-12 Martin Liska <mliska@suse.cz>
17089 PR bootstrap/100560
17090 * Makefile.in: Remove version.h from linker command line.
17092 2021-05-12 Richard Biener <rguenther@suse.de>
17094 PR middle-end/100547
17095 * rtl.h (rtvec_alloc): Make argument size_t.
17096 * rtl.c (rtvec_alloc): Verify the count is less than INT_MAX.
17098 2021-05-12 Jakub Jelinek <jakub@redhat.com>
17100 PR middle-end/100508
17101 * cfgexpand.c (expand_debug_expr): For DEBUG_EXPR_DECL with vector
17102 type, don't reuse DECL_RTL if it has different mode, instead force
17103 creation of a new DEBUG_EXPR.
17105 2021-05-12 Jakub Jelinek <jakub@redhat.com>
17106 Marc Glisse <marc.glisse@inria.fr>
17108 PR tree-optimization/94589
17109 * match.pd ((X & Y) == X -> (X & ~Y) == 0,
17110 (X | Y) == Y -> (X & ~Y) == 0): New GIMPLE simplifications.
17112 2021-05-12 Uroš Bizjak <ubizjak@gmail.com>
17115 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Handle V2SF mode.
17116 * config/i386/mmx.md (MMXMODE124): New mode iterator.
17118 (mmxintvecmode): New mode attribute.
17119 (mmxintvecmodelower): Ditto.
17120 (*mmx_maskcmpv2sf3_comm): New insn pattern.
17121 (*mmx_maskcmpv2sf3): Ditto.
17122 (vec_cmpv2sfv2si): New expander.
17123 (vcond<V2FI:mode>v2si): Ditto.
17124 (mmx_vlendvps): New insn pattern.
17125 (vcond<MMXMODE124:mode><MMXMODEI:mode>): Also handle V2SFmode.
17126 (vcondu<MMXMODE124:mode><MMXMODEI:mode>): Ditto.
17127 (vcond_mask_<mode><mmxintvecmodelower>): Ditto.
17129 2021-05-11 Martin Sebor <msebor@redhat.com>
17131 PR middle-end/21433
17132 * expr.c (expand_expr_real_1): Replace unreachable code with an assert.
17134 2021-05-11 Richard Biener <rguenther@suse.de>
17136 * gimple-fold.c (gimple_fold_call): Do not call
17137 maybe_fold_reference on call arguments or the static chain.
17138 (fold_stmt_1): Do not call maybe_fold_reference on GIMPLE_ASM
17141 2021-05-11 Martin Liska <mliska@suse.cz>
17143 * builtins.def (DEF_HSAIL_BUILTIN): Remove.
17144 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
17145 (DEF_HSAIL_SAT_BUILTIN): Likewise.
17146 (DEF_HSAIL_INTR_BUILTIN): Likewise.
17147 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
17148 * doc/frontends.texi: Remove BRIG.
17149 * doc/install.texi: Likewise.
17150 * doc/invoke.texi: Likewise.
17151 * doc/standards.texi: Likewise.
17152 * brig-builtins.def: Removed.
17153 * brig/ChangeLog: Removed.
17154 * brig/Make-lang.in: Removed.
17155 * brig/brig-builtins.h: Removed.
17156 * brig/brig-c.h: Removed.
17157 * brig/brig-lang.c: Removed.
17158 * brig/brigfrontend/brig-arg-block-handler.cc: Removed.
17159 * brig/brigfrontend/brig-atomic-inst-handler.cc: Removed.
17160 * brig/brigfrontend/brig-basic-inst-handler.cc: Removed.
17161 * brig/brigfrontend/brig-branch-inst-handler.cc: Removed.
17162 * brig/brigfrontend/brig-cmp-inst-handler.cc: Removed.
17163 * brig/brigfrontend/brig-code-entry-handler.cc: Removed.
17164 * brig/brigfrontend/brig-code-entry-handler.h: Removed.
17165 * brig/brigfrontend/brig-comment-handler.cc: Removed.
17166 * brig/brigfrontend/brig-control-handler.cc: Removed.
17167 * brig/brigfrontend/brig-copy-move-inst-handler.cc: Removed.
17168 * brig/brigfrontend/brig-cvt-inst-handler.cc: Removed.
17169 * brig/brigfrontend/brig-fbarrier-handler.cc: Removed.
17170 * brig/brigfrontend/brig-function-handler.cc: Removed.
17171 * brig/brigfrontend/brig-function.cc: Removed.
17172 * brig/brigfrontend/brig-function.h: Removed.
17173 * brig/brigfrontend/brig-inst-mod-handler.cc: Removed.
17174 * brig/brigfrontend/brig-label-handler.cc: Removed.
17175 * brig/brigfrontend/brig-lane-inst-handler.cc: Removed.
17176 * brig/brigfrontend/brig-machine.c: Removed.
17177 * brig/brigfrontend/brig-machine.h: Removed.
17178 * brig/brigfrontend/brig-mem-inst-handler.cc: Removed.
17179 * brig/brigfrontend/brig-module-handler.cc: Removed.
17180 * brig/brigfrontend/brig-queue-inst-handler.cc: Removed.
17181 * brig/brigfrontend/brig-seg-inst-handler.cc: Removed.
17182 * brig/brigfrontend/brig-signal-inst-handler.cc: Removed.
17183 * brig/brigfrontend/brig-to-generic.cc: Removed.
17184 * brig/brigfrontend/brig-to-generic.h: Removed.
17185 * brig/brigfrontend/brig-util.cc: Removed.
17186 * brig/brigfrontend/brig-util.h: Removed.
17187 * brig/brigfrontend/brig-variable-handler.cc: Removed.
17188 * brig/brigfrontend/hsa-brig-format.h: Removed.
17189 * brig/brigfrontend/phsa.h: Removed.
17190 * brig/brigspec.c: Removed.
17191 * brig/config-lang.in: Removed.
17192 * brig/gccbrig.texi: Removed.
17193 * brig/lang-specs.h: Removed.
17194 * brig/lang.opt: Removed.
17196 2021-05-11 Richard Biener <rguenther@suse.de>
17199 * ipa-param-manipulation.c
17200 (ipa_param_body_adjustments::modify_call_stmt): Avoid
17201 altering SSA_NAME_DEF_STMT by adjusting the calls LHS
17202 via gimple_call_lhs_ptr.
17204 2021-05-11 Alex Coplan <alex.coplan@arm.com>
17207 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear):
17208 Avoid emitting CFA adjusts on the sp if we have the fp.
17210 2021-05-11 Richard Sandiford <richard.sandiford@arm.com>
17212 * config/aarch64/iterators.md (VMUL_CHANGE_NLANES): Delete.
17213 (VMULD): New iterator.
17214 (VCOND): Handle V4HF and V8HF.
17215 (VCONQ): Fix entry for V2SF.
17216 * config/aarch64/aarch64-simd.md (mul_lane<mode>3): Use VMULD
17217 instead of VMUL. Use a 64-bit vector mode for the indexed operand.
17218 (*aarch64_mul3_elt_<vswap_width_name><mode>): Merge with...
17219 (mul_laneq<mode>3): ...this define_insn. Use VMUL instead of VDQSF.
17220 Use a 128-bit vector mode for the indexed operand. Use stype for
17221 the scheduling type.
17223 2021-05-11 Richard Biener <rguenther@suse.de>
17225 * gimple-fold.c (maybe_fold_reference): Only return
17226 is_gimple_min_invariant values.
17228 2021-05-11 Richard Biener <rguenther@suse.de>
17230 PR middle-end/100509
17231 * gimple-fold.c (fold_gimple_assign): Only call
17232 get_symbol_constant_value on register type symbols.
17234 2021-05-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
17235 Joe Ramsay <joe.ramsay@arm.com>
17238 * config/arm/arm_mve.h (__arm_vstrwq_scatter_offset): Fix wrong arguments.
17239 (__arm_vcmpneq): Remove duplicate definition.
17240 (__arm_vstrwq_scatter_offset_p): Likewise.
17241 (__arm_vmaxq_x): Likewise.
17242 (__arm_vmlsdavaq): Likewise.
17243 (__arm_vmlsdavaxq): Likewise.
17244 (__arm_vmlsdavq_p): Likewise.
17245 (__arm_vmlsdavxq_p): Likewise.
17246 (__arm_vrmlaldavhaq): Likewise.
17247 (__arm_vstrbq_p): Likewise.
17248 (__arm_vstrbq_scatter_offset): Likewise.
17249 (__arm_vstrbq_scatter_offset_p): Likewise.
17250 (__arm_vstrdq_scatter_offset): Likewise.
17251 (__arm_vstrdq_scatter_offset_p): Likewise.
17252 (__arm_vstrdq_scatter_shifted_offset): Likewise.
17253 (__arm_vstrdq_scatter_shifted_offset_p): Likewise.
17255 2021-05-11 Jakub Jelinek <jakub@redhat.com>
17257 PR middle-end/100471
17258 * omp-low.c (lower_omp_task_reductions): For OMP_TASKLOOP, if data
17259 is 0, bypass the reduction loop including
17260 GOMP_taskgroup_reduction_unregister call.
17262 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
17264 * config/rs6000/rs6000.c (struct rs6000_cost_data): New member
17265 costing_for_scalar.
17266 (rs6000_density_test): Early return if costing_for_scalar is true.
17267 (rs6000_init_cost): Init costing_for_scalar of rs6000_cost_data.
17269 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
17271 * doc/tm.texi: Regenerated.
17272 * target.def (init_cost): Add new parameter costing_for_scalar.
17273 * targhooks.c (default_init_cost): Adjust for new parameter.
17274 * targhooks.h (default_init_cost): Likewise.
17275 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
17276 (vect_compute_single_scalar_iteration_cost): Likewise.
17277 (vect_analyze_loop_2): Likewise.
17278 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
17279 (vect_bb_vectorization_profitable_p): Likewise.
17280 * tree-vectorizer.h (init_cost): Likewise.
17281 * config/aarch64/aarch64.c (aarch64_init_cost): Likewise.
17282 * config/i386/i386.c (ix86_init_cost): Likewise.
17283 * config/rs6000/rs6000.c (rs6000_init_cost): Likewise.
17285 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
17287 * config/rs6000/rs6000.c (rs6000_vect_nonmem): Renamed to
17288 vect_nonmem and moved into...
17289 (struct rs6000_cost_data): ...here.
17290 (rs6000_init_cost): Use vect_nonmem of cost_data instead.
17291 (rs6000_add_stmt_cost): Likewise.
17292 (rs6000_finish_cost): Likewise.
17294 2021-05-10 Eric Botcazou <ebotcazou@adacore.com>
17296 * range-op.cc (get_bool_state): Adjust head comment.
17297 (operator_not_equal::op1_range): Fix comment.
17298 (operator_bitwise_xor::op1_range): Remove call to gcc_unreachable.
17300 2021-05-10 Martin Sebor <msebor@redhat.com>
17302 PR middle-end/100425
17303 PR middle-end/100510
17304 * gimple-ssa-warn-alloca.c (pass_walloca::firast_time_p): Rename...
17305 (pass_walloca::xlimit_certain_p): ...to this.
17306 (pass_walloca::gate): Execute for any kind of handled warning.
17307 (pass_walloca::execute): Avoid issuing "maybe" and "unbounded"
17308 warnings when xlimit_certain_p is set.
17310 2021-05-10 Pat Haugen <pthaugen@linux.ibm.com>
17312 * config/rs6000/rs6000.c (rs6000_ira_change_pseudo_allocno_class):
17313 Return ALTIVEC_REGS if that is best_class.
17314 (rs6000_compute_pressure_classes): Add ALTIVEC_REGS.
17316 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17318 * config/arm/arm.h (CPP_SPEC): Remove error message about
17321 2021-05-10 Martin Jambor <mjambor@suse.cz>
17323 * ipa-prop.h (IPA_NODE_REF): Removed.
17324 (IPA_NODE_REF_GET_CREATE): Likewise.
17325 (IPA_EDGE_REF): Likewise.
17326 (IPA_EDGE_REF_GET_CREATE): Likewise.
17327 (IS_VALID_JUMP_FUNC_INDEX): Likewise.
17328 * ipa-cp.c (print_all_lattices): Replaced IPA_NODE_REF with a direct
17329 use of ipa_node_params_sum.
17330 (ipcp_versionable_function_p): Likewise.
17331 (push_node_to_stack): Likewise.
17332 (pop_node_from_stack): Likewise.
17333 (set_single_call_flag): Replaced two IPA_NODE_REF with one single
17334 direct use of ipa_node_params_sum.
17335 (initialize_node_lattices): Replaced IPA_NODE_REF with a direct use of
17336 ipa_node_params_sum.
17337 (ipa_context_from_jfunc): Replaced IPA_EDGE_REF with a direct use of
17339 (ipcp_verify_propagated_values): Replaced IPA_NODE_REF with a direct
17340 use of ipa_node_params_sum.
17341 (self_recursively_generated_p): Likewise.
17342 (propagate_scalar_across_jump_function): Likewise.
17343 (propagate_context_across_jump_function): Replaced IPA_EDGE_REF with a
17344 direct use of ipa_edge_args_sum, moved the lookup after the early
17345 exit. Replaced IPA_NODE_REF with a direct use of ipa_node_params_sum.
17346 (propagate_bits_across_jump_function): Replaced IPA_NODE_REF with
17347 direct uses of ipa_node_params_sum.
17348 (propagate_vr_across_jump_function): Likewise.
17349 (propagate_aggregate_lattice): Likewise.
17350 (propagate_aggs_across_jump_function): Likewise.
17351 (propagate_constants_across_call): Likewise, also replaced
17352 IPA_EDGE_REF with a direct use of ipa_edge_args_sum.
17353 (good_cloning_opportunity_p): Replaced IPA_NODE_REF with a direct use
17354 of ipa_node_params_sum.
17355 (estimate_local_effects): Likewise.
17356 (add_all_node_vals_to_toposort): Likewise.
17357 (propagate_constants_topo): Likewise.
17358 (ipcp_propagate_stage): Likewise.
17359 (ipcp_discover_new_direct_edges): Likewise.
17360 (calls_same_node_or_its_all_contexts_clone_p): Likewise.
17361 (cgraph_edge_brings_value_p): Likewise (in both overloaded functions).
17362 (get_info_about_necessary_edges): Likewise.
17363 (want_remove_some_param_p): Likewise.
17364 (create_specialized_node): Likewise.
17365 (self_recursive_pass_through_p): Likewise.
17366 (self_recursive_agg_pass_through_p): Likewise.
17367 (find_more_scalar_values_for_callers_subset): Likewise and also
17368 replaced IPA_EDGE_REF with direct uses of ipa_edge_args_sum, in one
17369 case replacing two of those with a single query.
17370 (find_more_contexts_for_caller_subset): Likewise for the
17371 ipa_polymorphic_call_context overload.
17372 (intersect_aggregates_with_edge): Replaced IPA_EDGE_REF with a direct
17373 use of ipa_edge_args_sum. Replaced IPA_NODE_REF with direct uses of
17374 ipa_node_params_sum.
17375 (find_aggregate_values_for_callers_subset): Likewise, also reusing
17376 results of ipa_edge_args_sum->get.
17377 (cgraph_edge_brings_all_scalars_for_node): Replaced IPA_NODE_REF with
17378 direct uses of ipa_node_params_sum, replaced IPA_EDGE_REF with a
17379 direct use of ipa_edge_args_sum.
17380 (cgraph_edge_brings_all_agg_vals_for_node): Likewise, moved node
17381 summary query after the early exit and reused the result later.
17382 (decide_about_value): Replaced IPA_NODE_REF with a direct use of
17383 ipa_node_params_sum.
17384 (decide_whether_version_node): Likewise. Removed re-querying for
17385 summaries after cloning.
17386 (spread_undeadness): Replaced IPA_NODE_REF with a direct use of
17387 ipa_node_params_sum.
17388 (has_undead_caller_from_outside_scc_p): Likewise, reusing results of
17390 (identify_dead_nodes): Likewise.
17391 (ipcp_store_bits_results): Replaced IPA_NODE_REF with direct uses of
17392 ipa_node_params_sum.
17393 (ipcp_store_vr_results): Likewise.
17394 * ipa-fnsummary.c (evaluate_properties_for_edge): Likewise.
17395 (ipa_fn_summary_t::duplicate): Likewise.
17396 (analyze_function_body): Likewise.
17397 (estimate_calls_size_and_time): Likewise.
17398 (ipa_cached_call_context::duplicate_from): Likewise.
17399 (ipa_call_context::equal_to): Likewise.
17400 (remap_edge_params): Likewise.
17401 (ipa_merge_fn_summary_after_inlining): Likewise.
17402 (inline_read_section): Likewise.
17403 * ipa-icf.c (sem_function::param_used_p): Likewise.
17404 * ipa-modref.c (compute_parm_map): Likewise.
17405 (compute_parm_map): Replaced IPA_EDGE_REF with a direct use of
17407 (get_access_for_fnspec): Replaced IPA_NODE_REF with a direct use of
17408 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
17410 * ipa-profile.c (check_argument_count): Likewise.
17411 * ipa-prop.c (ipa_alloc_node_params): Replaced IPA_NODE_REF_GET_CREATE
17412 with a direct use of ipa_node_params_sum.
17413 (ipa_initialize_node_params): Likewise.
17414 (ipa_print_node_jump_functions_for_edge): Replaced IPA_EDGE_REF with a
17415 direct use of ipa_edge_args_sum and reused the query result.
17416 (ipa_compute_jump_functions_for_edge): Replaced IPA_NODE_REF with a
17417 direct use of ipa_node_params_sum and replaced IPA_EDGE_REF with a
17418 direct use of ipa_edge_args_sum.
17419 (ipa_note_param_call): Replaced IPA_NODE_REF with a direct use of
17420 ipa_node_params_sum and reused the result of the query.
17421 (ipa_analyze_node): Likewise.
17422 (ipa_analyze_controlled_uses): Replaced IPA_NODE_REF with a direct use
17423 of ipa_node_params_sum.
17424 (update_jump_functions_after_inlining): Replaced IPA_EDGE_REF with
17425 direct uses of ipa_edge_args_sum.
17426 (update_indirect_edges_after_inlining): Replaced IPA_NODE_REF with
17427 direct uses of ipa_node_params_sum and replaced IPA_EDGE_REF with a
17428 direct use of ipa_edge_args_sum. Removed superficial re-querying the
17430 (propagate_controlled_uses): Replaced IPA_NODE_REF with direct uses of
17431 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
17433 (ipa_propagate_indirect_call_infos): Replaced IPA_EDGE_REF with a
17434 direct use of ipa_edge_args_sum.
17435 (ipa_edge_args_sum_t::duplicate): Replaced IPA_NODE_REF with a direct
17436 use of ipa_node_params_sum.
17437 (ipa_print_node_params): Likewise.
17438 (ipa_write_node_info): Likewise and also replaced IPA_EDGE_REF with
17439 direct uses of ipa_edge_args_sum.
17440 (ipa_read_edge_info): Replaced IPA_EDGE_REF with a direct use of
17442 (ipa_read_node_info): Replaced IPA_NODE_REF with a direct use of
17443 ipa_node_params_sum.
17444 (ipa_prop_write_jump_functions): Likewise. Move variable node to the
17445 scopes where it is used.
17447 2021-05-10 Uroš Bizjak <ubizjak@gmail.com>
17449 * config/i386/i386-expand.c (ix86_expand_sse_movcc)
17450 <case E_V2SImode>: Force op_true to register.
17452 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17454 * config/arm/iterators.md (MVE_FP_COMPARISONS): New.
17455 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_f<mode>)
17456 (mve_vcmp<mve_cmp_op>q_n_f<mode>): New, merge all vcmp_*f*
17458 (mve_vcmpeqq_f<mode>, mve_vcmpeqq_n_f<mode>, mve_vcmpgeq_f<mode>)
17459 (mve_vcmpgeq_n_f<mode>, mve_vcmpgtq_f<mode>)
17460 (mve_vcmpgtq_n_f<mode>, mve_vcmpleq_f<mode>)
17461 (mve_vcmpleq_n_f<mode>, mve_vcmpltq_f<mode>)
17462 (mve_vcmpltq_n_f<mode>, mve_vcmpneq_f<mode>)
17463 (mve_vcmpneq_n_f<mode>): Remove.
17464 * config/arm/unspecs.md (VCMPEQQ_F, VCMPEQQ_N_F, VCMPGEQ_F)
17465 (VCMPGEQ_N_F, VCMPGTQ_F, VCMPGTQ_N_F, VCMPLEQ_F, VCMPLEQ_N_F)
17466 (VCMPLTQ_F, VCMPLTQ_N_F, VCMPNEQ_F, VCMPNEQ_N_F): Remove.
17468 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17470 * config/arm/iterators.md (MVE_COMPARISONS): New.
17472 (mve_cmp_type): New.
17473 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_<mode>): New, merge all
17475 (mve_vcmpneq_<mode>, mve_vcmpcsq_n_<mode>, mve_vcmpcsq_<mode>)
17476 (mve_vcmpeqq_n_<mode>, mve_vcmpeqq_<mode>, mve_vcmpgeq_n_<mode>)
17477 (mve_vcmpgeq_<mode>, mve_vcmpgtq_n_<mode>, mve_vcmpgtq_<mode>)
17478 (mve_vcmphiq_n_<mode>, mve_vcmphiq_<mode>, mve_vcmpleq_n_<mode>)
17479 (mve_vcmpleq_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
17480 (mve_vcmpneq_n_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
17481 (mve_vcmpneq_n_<mode>): Remove.
17483 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17485 * config/arm/arm_mve.h (__arm_vcmp*): Remove 's' suffix.
17486 * config/arm/arm_mve_builtins.def (vcmp*): Remove 's' suffix.
17487 * config/arm/mve.md (mve_vcmp*): Remove 's' suffix in pattern
17490 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17492 * config/arm/arm_mve_builtins.def (vcmpneq_u): Remove.
17493 (vcmpneq_n_u): Likewise.
17494 (vcmpeqq_u,): Likewise.
17495 (vcmpeqq_n_u): Likewise.
17496 * config/arm/iterators.md (supf): Remove VCMPNEQ_U, VCMPEQQ_U,
17497 VCMPEQQ_N_U and VCMPNEQ_N_U.
17498 * config/arm/mve.md (mve_vcmpneq): Remove <supf> iteration.
17499 (mve_vcmpeqq_n): Likewise.
17500 (mve_vcmpeqq): Likewise.
17501 (mve_vcmpneq_n): Likewise.
17503 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17505 * config/arm/arm_mve.h (__arm_vcmpeq*u*, __arm_vcmpne*u*): Call
17506 the 's' version of the builtin.
17508 2021-05-10 Richard Biener <rguenther@suse.de>
17510 PR tree-optimization/100492
17511 * tree-loop-distribution.c (find_seed_stmts_for_distribution):
17512 Find nothing when the loop contains an irreducible region.
17514 2021-05-10 Richard Biener <rguenther@suse.de>
17516 PR middle-end/100464
17518 * gimple-fold.c (canonicalize_constructor_val): Do not set
17521 2021-05-10 Richard Biener <rguenther@suse.de>
17523 PR tree-optimization/100434
17524 * tree-ssa-dse.c (initialize_ao_ref_for_dse): Handle
17526 (dse_optimize_stmt): Handle call LHS by dropping the
17527 LHS or the whole call if it doesn't have other
17529 (pass_dse::execute): Adjust.
17531 2021-05-10 Martin Liska <mliska@suse.cz>
17533 * Makefile.in: Add missing genversion rule.
17535 2021-05-10 Alex Coplan <alex.coplan@arm.com>
17538 * config/arm/mve.md (*mve_mov<mode>): Simplify output code. Use
17539 vldrw.u32 and vstrw.32 for V2D[IF]mode loads and stores.
17541 2021-05-10 Martin Liska <mliska@suse.cz>
17543 * builtins.c (is_builtin_name): Use startswith
17544 function instead of strncmp.
17545 * collect2.c (main): Likewise.
17546 (has_lto_section): Likewise.
17547 (scan_libraries): Likewise.
17548 * coverage.c (coverage_checksum_string): Likewise.
17549 (coverage_init): Likewise.
17550 * dwarf2out.c (is_cxx): Likewise.
17551 (gen_compile_unit_die): Likewise.
17552 * gcc-ar.c (main): Likewise.
17553 * gcc.c (init_spec): Likewise.
17554 (read_specs): Likewise.
17555 (execute): Likewise.
17556 (check_live_switch): Likewise.
17557 * genattrtab.c (write_attr_case): Likewise.
17558 (IS_ATTR_GROUP): Likewise.
17559 * gencfn-macros.c (main): Likewise.
17560 * gengtype.c (type_for_name): Likewise.
17561 (gen_rtx_next): Likewise.
17562 (get_file_langdir): Likewise.
17563 (write_local): Likewise.
17564 * genmatch.c (get_operator): Likewise.
17565 (get_operand_type): Likewise.
17566 (expr::gen_transform): Likewise.
17567 * genoutput.c (validate_optab_operands): Likewise.
17568 * incpath.c (add_sysroot_to_chain): Likewise.
17569 * langhooks.c (lang_GNU_C): Likewise.
17570 (lang_GNU_CXX): Likewise.
17571 (lang_GNU_Fortran): Likewise.
17572 (lang_GNU_OBJC): Likewise.
17573 * lto-wrapper.c (run_gcc): Likewise.
17574 * omp-general.c (omp_max_simt_vf): Likewise.
17575 * omp-low.c (omp_runtime_api_call): Likewise.
17576 * opts-common.c (parse_options_from_collect_gcc_options): Likewise.
17577 * read-rtl-function.c (function_reader::read_rtx_operand_r): Likewise.
17578 * real.c (real_from_string): Likewise.
17579 * selftest.c (assert_str_startswith): Likewise.
17580 * timevar.c (timer::validate_phases): Likewise.
17581 * tree.c (get_file_function_name): Likewise.
17582 * ubsan.c (ubsan_use_new_style_p): Likewise.
17583 * varasm.c (default_function_rodata_section): Likewise.
17584 (incorporeal_function_p): Likewise.
17585 (default_section_type_flags): Likewise.
17586 * system.h (startswith): Define startswith.
17588 2021-05-10 Martin Liska <mliska@suse.cz>
17590 * bitmap.h (class auto_bitmap): Remove
17591 __cplusplus >= 201103.
17592 * config/aarch64/aarch64.c: Likewise.
17593 * gimple-ssa-store-merging.c (store_immediate_info::store_immediate_info):
17595 * sbitmap.h: Likewise.
17597 2021-05-10 Martin Liska <mliska@suse.cz>
17599 * Makefile.in: Rename gcov-iov to genversion and depend
17600 on version.h (instead of gcov-iov.h).
17601 * gcov-io.h: Include version.h instread of gcov-iov.h.
17602 * gengtype-state.c (read_state_version): Likewise.
17603 * gcov-iov.c: Moved to...
17604 * genversion.c: ...here.
17605 * lto-streamer.h (LTO_major_version): Define it with
17607 * version.c: Removed.
17608 * version.h: Removed.
17610 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17612 * config/arc/arc.md (UNSPEC_ARC_DMPYWH): Define.
17613 * config/arc/simdext.md (VCT): Add predicates for iterator
17616 (voptab): Likewise.
17617 (vec_widen_<V_US>mult_hi_v4hi): Change pattern predicate.
17618 (<voptab>v2si3): New patterns.
17620 (reduc_plus_scal_v4hi): Likewise.
17621 (reduc_plus_scal_v2si): Likewise.
17622 (vec_duplicatev2si): Likewise.
17623 (vec_duplicatev4hi): Likewise.
17625 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17627 * config/arc/simdext.md: Format and cleanup file.
17629 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17631 * config/arc/simdext.md (movmisalignv2hi): Allow misaligned access
17632 only when munaligned-access option is on.
17633 (movmisalign<mode>): Likewise.
17635 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17637 * common/config/arc/arc-common.c (arc_handle_option): Remove dot
17639 * config/arc/arc.c (arc_reorg): Remove underscore from string.
17641 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17643 * config/arc/arc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
17644 (CTZ_DEFINED_VALUE_AT_ZERO): Likewise.
17645 * config/arc/arc.md (clrsbsi2): Cleanup pattern.
17646 (norm_f): Likewise.
17649 (clzsi2): Use fls instruction when available.
17650 (arc_clzsi2): Likewise.
17652 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17654 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add r26 and r27.
17656 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17658 * doc/extend.texi (__builtin_arc_sr): Swap arguments.
17660 2021-05-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
17662 PR middle-end/100467
17663 * toplev.c (compile_file): Call insn_locations_init before
17664 targetm.asm_out.code_end.
17666 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
17669 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
17671 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
17673 2021-05-07 Jakub Jelinek <jakub@redhat.com>
17674 Andrew Stubbs <amd@codesourcery.com>
17677 * builtins.c (try_store_by_multiple_pieces): Use force_operand for
17678 emit_move_insn operands.
17680 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
17682 * cfgexpand.c (expand_gimple_basic_block): Do not inherit a current
17683 location for the outgoing edges of an empty block.
17684 * dwarf2out.c (add_subscript_info): Retrieve the bounds and index
17685 type by means of the get_array_descr_info langhook, if it is set and
17686 returns true. Remove obsolete code dealing with unnamed subtypes.
17688 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
17690 * gimple-range-cache.cc (ssa_block_ranges): Virtualize.
17691 (sbr_vector): Renamed from ssa_block_cache.
17692 (sbr_vector::sbr_vector): Allocate from obstack abd initialize.
17693 (ssa_block_ranges::~ssa_block_ranges): Remove.
17694 (sbr_vector::set_bb_range): Use varying and undefined cached values.
17695 (ssa_block_ranges::set_bb_varying): Remove.
17696 (sbr_vector::get_bb_range): Adjust assert.
17697 (sbr_vector::bb_range_p): Adjust assert.
17698 (~block_range_cache): No freeing loop required.
17699 (block_range_cache::get_block_ranges): Remove.
17700 (block_range_cache::set_bb_range): Inline get_block_ranges.
17701 (block_range_cache::set_bb_varying): Remove.
17702 * gimple-range-cache.h (set_bb_varying): Remove prototype.
17703 * value-range.h (irange_allocator::get_memory): New.
17705 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
17707 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Search
17708 dominator tree is available and requested.
17709 (ranger_cache::ssa_range_in_bb): Don't search dom tree here.
17710 (ranger_cache::fill_block_cache): Don't search dom tree here either.
17711 * gimple-range-cache.h (non_null_deref_p): Add dom_search param.
17713 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
17715 * gimple-range.cc (gimple_ranger::range_on_exit): Handle block with
17716 only PHI nodes better.
17718 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
17720 * gimple-range-edge.h (gimple_outgoing_range): Rename from
17722 (gcond_edge_range): Export prototype.
17723 * gimple-range-edge.cc (gcond_edge_range): New.
17724 (gimple_outgoing_range::edge_range_p): Use gcond_edge_range.
17725 * gimple-range-gori.h (gori_compute): Use gimple_outgoing_range.
17727 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
17729 * gimple-range-edge.cc (outgoing_range::calc_switch_ranges): Compute
17730 default range into a temp and allocate only what is needed.
17732 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
17734 * range-op.cc (operator_trunc_mod::wi_fold): x % 0 is UNDEFINED.
17736 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
17738 * gimple-range.h (gimple_range_global): Pick up parameter initial
17739 values, and use-before defined locals are UNDEFINED.
17741 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
17743 * doc/extend.texi (scalar_storage_order): Mention effect on pointer
17745 * tree.h (reverse_storage_order_for_component_p): Return false if
17746 the type is a pointer.
17748 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
17750 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
17752 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
17755 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
17756 Handle V8QI, V4HI and V2SI modes.
17757 * config/i386/mmx.md (mmx_pblendvb): New insn pattern.
17758 * config/i386/sse.md (unspec): Move UNSPEC_BLENDV ...
17759 * config/i386/i386.md (unspec): ... here.
17761 2021-05-07 Tobias Burnus <tobias@codesourcery.com>
17762 Tom de Vries <tdevries@suse.de>
17764 * omp-low.c (lower_rec_simd_input_clauses): Set max_vf = 1 if
17765 a truth_value_p reduction variable is nonintegral.
17767 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
17770 * config/i386/i386-expand.c (ix86_use_mask_cmp_p):
17771 Return false for mode sizes < 16.
17773 2021-05-07 Jakub Jelinek <jakub@redhat.com>
17776 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn.
17778 2021-05-06 Martin Jambor <mjambor@suse.cz>
17780 * ipa-sra.c (ipa_sra_dump_all_summaries): Dump edge summaries even
17781 when there is no function summary.
17782 (ipa_sra_summarize_function): produce edge summaries even when
17785 2021-05-06 Tom Tromey <tom@tromey.com>
17787 * godump.c (string_hash_eq): Remove.
17788 (go_finish): Use htab_eq_string.
17790 2021-05-06 Tom Tromey <tom@tromey.com>
17792 * gengtype-state.c (read_state): Use htab_eq_string.
17793 (string_eq): Remove.
17795 2021-05-06 Tom Tromey <tom@tromey.com>
17797 * gensupport.c (htab_eq_string): Remove.
17799 2021-05-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
17802 * debug.h (gcc_debug_hooks): Add set_ignored_loc function pointer.
17803 * dwarf2out.h (dw_fde_node::ignored_debug): New data item.
17804 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Add dummy
17805 set_ignored_loc callbacks.
17806 * debug.c (do_nothing_debug_hooks): Likewise.
17807 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
17808 * dwarf2out.c (text_section_used, cold_text_section_used): Remove.
17809 (in_text_section_p, last_text_label, last_cold_label,
17810 switch_text_ranges, switch_cold_ranges): New data items.
17811 (dwarf2out_note_section_used): Remove.
17812 (dwarf2out_begin_prologue): Set fde->ignored_debug and
17814 (mark_ignored_debug_section): New helper function.
17815 (dwarf2out_end_epilogue, dwarf2out_switch_text_section): Call
17816 mark_ignored_debug_section.
17817 (dwarf2_debug_hooks): Use dwarf2out_set_ignored_loc.
17818 (dwarf2_lineno_debug_hooks): Use dummy for set_ignored_loc.
17819 (size_of_aranges): Adjust formula for multi-part text ranges size.
17820 (output_aranges): Output multi-part text ranges.
17821 (dwarf2out_set_ignored_loc): New callback function.
17822 (dwarf2out_finish): Output multi-part text ranges.
17823 (dwarf2out_c_finalize): Clear new data items.
17824 * final.c (final_start_function_1): Call set_ignored_loc callback.
17825 (final_scan_insn_1): Likewise.
17826 * ggc-page.c (gt_ggc_mx): New helper function.
17827 * stringpool.c (gt_pch_nx): Likewise.
17829 2021-05-06 Richard Biener <rguenther@suse.de>
17831 * timevar.def (TV_TREE_INSERT_PHI_NODES): Remove.
17832 (TV_TREE_SSA_REWRITE_BLOCKS): Likewise.
17833 (TV_TREE_INTO_SSA): New.
17834 * tree-into-ssa.c (insert_phi_nodes): Do not account separately.
17835 (rewrite_blocks): Likewise.
17836 (pass_data_build_ssa): Account to TV_TREE_INTO_SSA.
17838 2021-05-06 Jakub Jelinek <jakub@redhat.com>
17840 * tree-ssa-phiopt.c (value_replacement, minmax_replacement,
17841 abs_replacement, xor_replacement,
17842 cond_removal_in_popcount_clz_ctz_pattern,
17843 replace_phi_edge_with_variable): Change type of phi argument from
17844 gimple * to gphi *.
17846 2021-05-06 Richard Biener <rguenther@suse.de>
17848 * tree-ssa-loop-split.c (split_loop): Delay updating SSA form.
17849 Output an opt-info message.
17850 (do_split_loop_on_cond): Likewise.
17851 (tree_ssa_split_loops): Update SSA form here.
17853 2021-05-06 Richard Biener <rguenther@suse.de>
17855 * tree-inline.c (tree_function_versioning): Fix DECL_BY_REFERENCE
17856 return variable removal.
17858 2021-05-06 Marius Hillenbrand <mhillen@linux.ibm.com>
17860 * config/s390/s390-builtins.def (O_M5, O1_M5, ...): Remove unused macros.
17861 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
17862 (s390_vec_permi_dbl, s390_vpdi): Use the O3_U2 type for the immediate
17864 * config/s390/s390.c (s390_const_operand_ok): Remove unused
17867 2021-05-06 Jakub Jelinek <jakub@redhat.com>
17869 PR tree-optimization/94589
17870 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Call
17871 spaceship_replacement.
17872 (cond_only_block_p, spaceship_replacement): New functions.
17874 2021-05-06 Richard Biener <rguenther@suse.de>
17877 * tree-emutls.c (gen_emutls_addr): Pass in whether we're
17878 dealing with a debug use and only query existing addresses
17880 (lower_emutls_1): Avoid splitting out addresses for debug
17881 stmts, reset the debug stmt when we fail to find existing
17883 (lower_emutls_phi_arg): Set wi.stmt.
17885 2021-05-06 Christoph Muellner <cmuellner@gcc.gnu.org>
17888 * config/riscv/riscv.c (riscv_block_move_loop): Use cbranch helper.
17889 * config/riscv/riscv.md (cbranch<mode>4): Generate helpers.
17890 (stack_protect_test): Use cbranch helper.
17892 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
17895 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
17896 always return the establisher frame for __builtin_frame_address (0).
17898 2021-05-05 Ivan Sorokin <vanyacpp@gmail.com>
17901 * config/i386/i386-builtins.c (ix86_cpu_model_type_node): New.
17902 (ix86_cpu_model_var): Likewise.
17903 (ix86_cpu_features2_type_node): Likewise.
17904 (ix86_cpu_features2_var): Likewise.
17905 (fold_builtin_cpu): Cache __cpu_model and __cpu_features2 with
17908 2021-05-05 Martin Sebor <msebor@redhat.com>
17910 * passes.def (pass_warn_printf): Run after SSA.
17912 2021-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17914 * config/arm/neon.md (neon_vtst_combine<mode>): New pattern.
17915 * config/arm/predicates.md (minus_one_operand): New predicate.
17917 2021-05-05 Jeff Law <jlaw@tachyum.com>
17919 * config/avr/avr.md: Remove references to CC_STATUS_INIT.
17921 2021-05-05 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
17923 PR rtl-optimization/100263
17924 * postreload.c (move2add_valid_value_p): Ensure register can
17927 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
17929 PR rtl-optimization/100411
17930 * cfgcleanup.c (try_crossjump_to_edge): Also skip end of prologue
17931 and beginning of function markers.
17933 2021-05-05 Jeff Law <jlaw@tachyum.com>
17935 * config/cr16/cr16.h (NOTICE_UPDATE_CC): Remove.
17936 * config/cr16/cr16.c (notice_update_cc): Remove.
17937 * config/cr16/cr16-protos.h (notice_update_cc): Remove.
17939 2021-05-05 Uroš Bizjak <ubizjak@gmail.com>
17942 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
17943 Handle V8QI, V4HI and V2SI modes.
17944 * config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
17945 (ix86_build_signbit_mask): Ditto.
17946 * config/i386/mmx.md (MMXMODE14): New mode iterator.
17947 (<smaxmin:code><MMXMODE14:mode>3): New expander.
17948 (*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
17949 (<umaxmin:code><MMXMODE24:mode>3): New expander.
17950 (*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
17951 (vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
17952 (vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
17953 (vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
17954 (vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
17955 (vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
17957 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
17959 * dwarf2out.c (loc_list_from_tree_1) <DECL>: During early DWARF, do
17960 not expand the VALUE_EXPR of variables put in the non-local frame.
17961 * gimplify.c (gimplify_type_sizes) <RECORD_TYPE>: If the type is not
17962 to be ignored for debug info, ensure its variable offsets are not.
17964 2021-05-05 Richard Biener <rguenther@suse.de>
17966 PR tree-optimization/79333
17967 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
17968 Fold stmt following SSA edges.
17970 2021-05-05 Richard Biener <rguenther@suse.de>
17972 PR middle-end/100394
17973 * calls.c (expand_call): Preserve possibly throwing calls.
17974 * cfgexpand.c (expand_call_stmt): When a call can throw signal
17975 RTL expansion there are side-effects.
17976 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Simplify,
17977 mark all possibly throwing stmts necessary unless we can elide
17979 * tree-ssa-dse.c (pass_dse::execute): Preserve exceptions unless
17980 -fdelete-dead-exceptions.
17981 * tree.h (DECL_PURE_P): Add note about exceptions.
17983 2021-05-05 Alexandre Oliva <oliva@adacore.com>
17985 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Make it
17988 2021-05-04 David Edelsohn <dje.gcc@gmail.com>
17990 * config/rs6000/rs6000-call.c (rs6000_output_mi_thunk): Use
17991 get_fnname_from_decl for name of thunk.
17992 * config/rs6000/rs6000.c (rs6000_declare_alias): Use assemble_name
17993 and ASM_OUTPUT_LABEL.
17994 (rs6000_xcoff_declare_function_name): Use assemble_name and
17996 (rs6000_xcoff_declare_object_name): Use ASM_OUTPUT_LABEL.
17997 (rs6000_xcoff_encode_section_info): Don't add mapping class
17998 for aliases. Always add [DS] mapping class to primary
18000 (rs6000_asm_weaken_decl): Don't explicitly add [DS].
18002 2021-05-04 Martin Sebor <msebor@redhat.com>
18004 PR middle-end/100307
18005 * builtins.c (compute_objsize_r): Clear base0 for pointers.
18007 2021-05-04 Jeff Law <jlaw@tachyum.com>
18009 * config/bfin/bfin.h (NOTICE_UPDATE_CC): Remove.
18011 2021-05-04 Segher Boessenkool <segher@kernel.crashing.org>
18013 * caller-save.c: Remove CC0.
18014 * cfgcleanup.c: Remove CC0.
18015 * cfgrtl.c: Remove CC0.
18016 * combine.c: Remove CC0.
18017 * compare-elim.c: Remove CC0.
18018 * conditions.h: Remove CC0.
18019 * config/h8300/h8300.h: Remove CC0.
18020 * config/h8300/h8300-protos.h: Remove CC0.
18021 * config/h8300/peepholes.md: Remove CC0.
18022 * config/i386/x86-tune-sched.c: Remove CC0.
18023 * config/m68k/m68k.c: Remove CC0.
18024 * config/rl78/rl78.c: Remove CC0.
18025 * config/sparc/sparc.c: Remove CC0.
18026 * config/xtensa/xtensa.c: Remove CC0.
18027 (gen_conditional_move): Use pc_rtx instead of cc0_rtx in a piece of
18028 RTL where that is used as a placeholder only.
18029 * cprop.c: Remove CC0.
18030 * cse.c: Remove CC0.
18031 * cselib.c: Remove CC0.
18032 * df-problems.c: Remove CC0.
18033 * df-scan.c: Remove CC0.
18034 * doc/md.texi: Remove CC0. Adjust an example.
18035 * doc/rtl.texi: Remove CC0. Adjust an example.
18036 * doc/tm.texi: Regenerate.
18037 * doc/tm.texi.in: Remove CC0.
18038 * emit-rtl.c: Remove CC0.
18039 * final.c: Remove CC0.
18040 * fwprop.c: Remove CC0.
18041 * gcse-common.c: Remove CC0.
18042 * gcse.c: Remove CC0.
18043 * genattrtab.c: Remove CC0.
18044 * genconfig.c: Remove CC0.
18045 * genemit.c: Remove CC0.
18046 * genextract.c: Remove CC0.
18047 * gengenrtl.c: Remove CC0.
18048 * genrecog.c: Remove CC0.
18049 * haifa-sched.c: Remove CC0.
18050 * ifcvt.c: Remove CC0.
18051 * ira-costs.c: Remove CC0.
18052 * ira.c: Remove CC0.
18053 * jump.c: Remove CC0.
18054 * loop-invariant.c: Remove CC0.
18055 * lra-constraints.c: Remove CC0.
18056 * lra-eliminations.c: Remove CC0.
18057 * optabs.c: Remove CC0.
18058 * postreload-gcse.c: Remove CC0.
18059 * postreload.c: Remove CC0.
18060 * print-rtl.c: Remove CC0.
18061 * read-rtl-function.c: Remove CC0.
18062 * reg-notes.def: Remove CC0.
18063 * reg-stack.c: Remove CC0.
18064 * reginfo.c: Remove CC0.
18065 * regrename.c: Remove CC0.
18066 * reload.c: Remove CC0.
18067 * reload1.c: Remove CC0.
18068 * reorg.c: Remove CC0.
18069 * resource.c: Remove CC0.
18070 * rtl.c: Remove CC0.
18071 * rtl.def: Remove CC0.
18072 * rtl.h: Remove CC0.
18073 * rtlanal.c: Remove CC0.
18074 * sched-deps.c: Remove CC0.
18075 * sched-rgn.c: Remove CC0.
18076 * shrink-wrap.c: Remove CC0.
18077 * simplify-rtx.c: Remove CC0.
18078 * system.h: Remove CC0. Poison NOTICE_UPDATE_CC, CC_STATUS_MDEP_INIT,
18079 CC_STATUS_MDEP, and CC_STATUS.
18080 * target.def: Remove CC0.
18081 * valtrack.c: Remove CC0.
18082 * var-tracking.c: Remove CC0.
18084 2021-05-04 Richard Biener <rguenther@suse.de>
18086 PR tree-optimization/100414
18087 * tree-ssa-phiopt.c (get_non_trapping): Do not compute dominance
18089 (tree_ssa_phiopt_worker): But unconditionally here.
18091 2021-05-04 Tobias Burnus <tobias@codesourcery.com>
18093 * omp-low.c (lower_rec_input_clauses, lower_reduction_clauses): Handle
18094 && and || with floating-point and complex arguments.
18096 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
18098 * tree-inline.c (insert_debug_decl_map): Delete.
18099 (copy_debug_stmt): Minor tweak.
18100 (setup_one_parameter): Do not use a variable if the value is either
18101 a read-only DECL or a non-addressable local variable in the caller.
18102 In this case, insert the debug-only variable in the map manually.
18103 (expand_call_inline): Do not generate a CLOBBER for these values.
18104 * tree-inline.h (debug_map): Minor tweak.
18106 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
18108 * builtins.c (builtin_with_linkage_p): Return true for stp[n]cpy.
18109 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Tidy up.
18111 2021-05-04 Richard Biener <rguenther@suse.de>
18113 PR tree-optimization/100329
18114 * tree-ssa-reassoc.c (can_reassociate_p): Do not reassociate
18116 (insert_stmt_after): Assert we're not running into asm goto.
18118 2021-05-04 Richard Biener <rguenther@suse.de>
18120 PR tree-optimization/100398
18121 * tree-ssa-dse.c (pass_dse::execute): Preserve control
18124 2021-05-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18126 * builtins.c (try_store_by_multiple_pieces): Fix constfun's prototype.
18128 2021-05-04 Alexandre Oliva <oliva@adacore.com>
18130 * builtins.c (try_store_by_multiple_pieces): New.
18131 (expand_builtin_memset_args): Use it. If target_char_cast
18132 fails, proceed as for non-constant val. Pass len's ctz to...
18133 * expr.c (clear_storage_hints): ... this. Try store by
18134 multiple pieces after setmem.
18135 (clear_storage): Adjust.
18136 * expr.h (clear_storage_hints): Likewise.
18137 (try_store_by_multiple_pieces): Declare.
18138 * passes.def: Replace the last copy_prop with ccp.
18140 2021-05-03 Tom de Vries <tdevries@suse.de>
18143 * omp-low.c (lower_rec_input_clauses): Disable SIMT for user-defined
18146 2021-05-03 Richard Biener <rguenther@suse.de>
18148 * tree-ssa-dse.c (dse_classify_store): Track two PHI defs.
18150 2021-05-03 Richard Biener <rguenther@suse.de>
18152 * tree-ssa-dse.c: Do not include domwalk.h but cfganal.h.
18153 (dse_dom_walker): Remove.
18154 (dse_dom_walker::dse_optimize_stmt): Rename...
18155 (dse_optimize_stmt): ... to this, pass in live_bytes sbitmap.
18156 (dse_dom_walker::before_dom_children): Inline ...
18157 (pass_dse::execute): ... here. Perform a reverse program
18160 2021-05-03 H.J. Lu <hjl.tools@gmail.com>
18163 * configure: Regenerated.
18165 2021-05-03 Ilya Leoshkevich <iii@linux.ibm.com>
18168 * config/s390/s390.c (s390_hard_fp_reg_p): New function.
18169 (s390_md_asm_adjust): Handle hard registers.
18171 2021-05-03 Jakub Jelinek <jakub@redhat.com>
18173 PR tree-optimization/100382
18174 * tree-ssa-dse.c: Include tree-eh.h.
18175 (dse_dom_walker::before_dom_children): Don't remove stmts if
18176 stmt_unremovable_because_of_non_call_eh_p is true.
18178 2021-05-02 David Edelsohn <dje.gcc@gmail.com>
18180 * varasm.c (compute_reloc_for_var): Split out from...
18181 (get_variable_section): Use it.
18182 * output.h (compute_reloc_for_var): Declare.
18183 * config/rs6000/rs6000-protos.h
18184 (rs6000_xcoff_asm_output_aligned_decl_common): Change alignment to
18186 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address_aix):
18187 Don't append storage mapping class to symbol.
18188 (rs6000_xcoff_asm_named_section): Add BS and UL mapping classes.
18189 Don't convert TLS BSS to common.
18190 (rs6000_xcoff_unique_section): Don't fall back to select_secton.
18191 (rs6000_xcoff_section_type_flags): Add SECTION_BSS if DECL is
18193 (rs6000_xcoff_asm_globalize_decl_name): Don't strip storage
18195 (rs6000_xcoff_asm_output_aligned_decl_common): Align is unsigned int.
18196 If align is 0 from TLS class, use the same rules as varasm.c
18197 If not common, switch to BSS section manually.
18198 If common, emit appropriate comm or lcomm directive.
18199 (rs6000_xcoff_encode_section_info): Add logic to append all
18200 storage mapping classes.
18201 (rs6000_asm_weaken_decl): Adjust for qualname symbols.
18202 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
18203 rs6000_xcoff_asm_output_aligned_decl_common.
18204 (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
18205 rs6000_xcoff_asm_output_aligned_decl_common.
18206 (ASM_OUTPUT_TLS_COMMON): Use
18207 rs6000_xcoff_asm_output_aligned_decl_common.
18209 2021-05-02 Jakub Jelinek <jakub@redhat.com>
18212 * config/nvptx/nvptx.c (nvptx_sese_pseudo): Use nullptr instead of 0
18213 as first argument of pseudo_node_t constructors.
18215 2021-05-02 Jakub Jelinek <jakub@redhat.com>
18218 * config/i386/t-i386 (TM_H): Add $(srcdir)/config/i386/i386-isa.def.
18220 2021-05-01 Aldy Hernandez <aldyh@redhat.com>
18222 * value-range.cc (DEFINE_INT_RANGE_GC_STUBS): Remove.
18223 (gt_pch_nx (int_range<1> *&)): New.
18224 (gt_ggc_mx (int_range<1> *&)): New.
18225 * value-range.h (class irange): Add GTY support for
18228 2021-05-01 Geng Qi <gengqi@linux.alibaba.com>
18230 * doc/options.texi (Negative): Change either or to both and.
18232 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
18234 * config/aarch64/aarch64-simd-builtins.def: Add
18235 float_ml[as][q]_laneq builtin generator macros.
18236 * config/aarch64/aarch64-simd.md (mul_laneq<mode>3): Define.
18237 (aarch64_float_mla_laneq<mode>): Define.
18238 (aarch64_float_mls_laneq<mode>): Define.
18239 * config/aarch64/arm_neon.h (vmla_laneq_f32): Use RTL builtin
18240 instead of GCC vector extensions.
18241 (vmlaq_laneq_f32): Likewise.
18242 (vmls_laneq_f32): Likewise.
18243 (vmlsq_laneq_f32): Likewise.
18245 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
18247 * config/aarch64/aarch64-simd-builtins.def: Add
18248 float_ml[as]_lane builtin generator macros.
18249 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt<mode>):
18251 (mul_lane<mode>3): This, and re-order arguments.
18252 (aarch64_float_mla_lane<mode>): Define.
18253 (aarch64_float_mls_lane<mode>): Define.
18254 * config/aarch64/arm_neon.h (vmla_lane_f32): Use RTL builtin
18255 instead of GCC vector extensions.
18256 (vmlaq_lane_f32): Likewise.
18257 (vmls_lane_f32): Likewise.
18258 (vmlsq_lane_f32): Likewise.
18260 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
18262 * config/aarch64/aarch64-simd-builtins.def: Add float_ml[as]
18263 builtin generator macros.
18264 * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>):
18266 (aarch64_float_mls<mode>): Define.
18267 * config/aarch64/arm_neon.h (vmla_f32): Use RTL builtin
18268 instead of relying on GCC vector extensions.
18269 (vmla_f64): Likewise.
18270 (vmlaq_f32): Likewise.
18271 (vmlaq_f64): Likewise.
18272 (vmls_f32): Likewise.
18273 (vmls_f64): Likewise.
18274 (vmlsq_f32): Likewise.
18275 (vmlsq_f64): Likewise.
18276 * config/aarch64/iterators.md: Define VDQF_DF mode iterator.
18278 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
18280 * config/aarch64/aarch64-simd-builtins.def: Add
18281 float_ml[as]_n_builtin generator macros.
18282 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt_from_dup<mode>):
18284 (mul_n<mode>3): This, and re-order arguments.
18285 (aarch64_float_mla_n<mode>): Define.
18286 (aarch64_float_mls_n<mode>): Define.
18287 * config/aarch64/arm_neon.h (vmla_n_f32): Use RTL builtin
18288 instead of inline asm.
18289 (vmlaq_n_f32): Likewise.
18290 (vmls_n_f32): Likewise.
18291 (vmlsq_n_f32): Likewise.
18293 2021-04-30 Jonathan Wright <joanthan.wright@arm.com>
18295 * config/aarch64/aarch64-simd-builtins.def: Add pmull[2]
18296 builtin generator macros.
18297 * config/aarch64/aarch64-simd.md (aarch64_pmullv8qi): Define.
18298 (aarch64_pmull_hiv16qi_insn): Define.
18299 (aarch64_pmull_hiv16qi): Define.
18300 * config/aarch64/arm_neon.h (vmull_high_p8): Use RTL builtin
18301 instead of inline asm.
18302 (vmull_p8): Likewise.
18304 2021-04-30 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
18306 * config/avr/avr.md: Adjust peepholes to match and
18307 generate parallels with clobber of REG_CC.
18308 (mov<mode>_insn): Rename to mov<mode>_insn_split.
18309 (*mov<mode>_insn): Rename to mov<mode>_insn.
18311 2021-04-30 David Edelsohn <dje.gcc@gmail.com>
18313 * varasm.c (use_blocks_for_decl_p): Don't use section anchors
18314 for VAR_DECLs if -fdata-sections enabled.
18316 2021-04-30 Michael Meissner <meissner@linux.ibm.com>
18318 PR bootstrap/100327
18319 * config/rs6000/rs6000.c
18320 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Define.
18321 (rs6000_libgcc_floating_mode_supported_p): New target hook.
18323 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
18325 * tree-ssa-threadbackward.c (class thread_jumps): Split out code
18327 (class back_threader_registry): ...to here...
18328 (class back_threader_profitability): ...and here...
18329 (thread_jumps::thread_through_all_blocks): Remove argument.
18330 (back_threader_registry::back_threader_registry): New.
18331 (back_threader_registry::~back_threader_registry): New.
18332 (back_threader_registry::thread_through_all_blocks): New.
18333 (thread_jumps::profitable_jump_thread_path): Move from here...
18334 (back_threader_profitability::profitable_path_p): ...to here.
18335 (thread_jumps::find_taken_edge): New.
18336 (thread_jumps::convert_and_register_current_path): Move...
18337 (back_threader_registry::register_path): ...to here.
18338 (thread_jumps::register_jump_thread_path_if_profitable): Move...
18339 (thread_jumps::maybe_register_path): ...to here.
18340 (thread_jumps::handle_phi): Call find_taken_edge and
18341 maybe_register_path.
18342 (thread_jumps::handle_assignment): Same.
18343 (thread_jumps::fsm_find_control_statement_thread_paths): Remove
18344 tree argument to handle_phi and handle_assignment.
18345 (thread_jumps::find_jump_threads_backwards): Set m_name. Remove
18346 set of m_speed_p and m_max_threaded_paths.
18347 (pass_thread_jumps::execute): Remove second argument from
18348 find_jump_threads_backwards.
18349 (pass_early_thread_jumps::execute): Same.
18351 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
18353 * tree-ssa-dom.c (class dom_jump_threader_simplifier): New.
18354 (class dom_opt_dom_walker): Initialize some class variables.
18355 (pass_dominator::execute): Pass evrp_range_analyzer and
18356 dom_jump_threader_simplifier to dom_opt_dom_walker.
18357 Adjust for some functions moving into classes.
18358 (simplify_stmt_for_jump_threading): Adjust and move to...
18359 (jump_threader_simplifier::simplify): ...here.
18360 (dom_opt_dom_walker::before_dom_children): Adjust for
18361 m_evrp_range_analyzer.
18362 (dom_opt_dom_walker::after_dom_children): Remove x_vr_values hack.
18363 (test_for_singularity): Place in dom_opt_dom_walker class.
18364 (dom_opt_dom_walker::optimize_stmt): The argument
18365 evrp_range_analyzer is now a class field.
18366 * tree-ssa-threadbackward.c (class thread_jumps): Add m_registry.
18367 (thread_jumps::thread_through_all_blocks): New.
18368 (thread_jumps::convert_and_register_current_path): Use m_registry.
18369 (pass_thread_jumps::execute): Adjust for thread_through_all_blocks
18370 being in the threader class.
18371 (pass_early_thread_jumps::execute): Same.
18372 * tree-ssa-threadedge.c (threadedge_initialize_values): Move...
18373 (jump_threader::jump_threader): ...here.
18374 (threadedge_finalize_values): Move...
18375 (jump_threader::~jump_threader): ...here.
18376 (jump_threader::remove_jump_threads_including): New.
18377 (jump_threader::thread_through_all_blocks): New.
18378 (record_temporary_equivalences_from_phis): Move...
18379 (jump_threader::record_temporary_equivalences_from_phis): ...here.
18380 (record_temporary_equivalences_from_stmts_at_dest): Move...
18381 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
18383 (simplify_control_stmt_condition_1): Move to jump_threader class.
18384 (simplify_control_stmt_condition): Move...
18385 (jump_threader::simplify_control_stmt_condition): ...here.
18386 (thread_around_empty_blocks): Move...
18387 (jump_threader::thread_around_empty_blocks): ...here.
18388 (thread_through_normal_block): Move...
18389 (jump_threader::thread_through_normal_block): ...here.
18390 (thread_across_edge): Move...
18391 (jump_threader::thread_across_edge): ...here.
18392 (thread_outgoing_edges): Move...
18393 (jump_threader::thread_outgoing_edges): ...here.
18394 * tree-ssa-threadedge.h: Move externally facing functings...
18395 (class jump_threader): ...here...
18396 (class jump_threader_simplifier): ...and here.
18397 * tree-ssa-threadupdate.c (struct redirection_data): Remove comment.
18398 (jump_thread_path_allocator::jump_thread_path_allocator): New.
18399 (jump_thread_path_allocator::~jump_thread_path_allocator): New.
18400 (jump_thread_path_allocator::allocate_thread_edge): New.
18401 (jump_thread_path_allocator::allocate_thread_path): New.
18402 (jump_thread_path_registry::jump_thread_path_registry): New.
18403 (jump_thread_path_registry::~jump_thread_path_registry): New.
18404 (jump_thread_path_registry::allocate_thread_edge): New.
18405 (jump_thread_path_registry::allocate_thread_path): New.
18406 (dump_jump_thread_path): Make extern.
18407 (debug (const vec<jump_thread_edge *> &path)): New.
18408 (struct removed_edges): Move to tree-ssa-threadupdate.h.
18409 (struct thread_stats_d): Remove.
18410 (remove_ctrl_stmt_and_useless_edges): Make static.
18411 (lookup_redirection_data): Move...
18412 (jump_thread_path_registry::lookup_redirection_data): ...here.
18413 (ssa_redirect_edges): Make static.
18414 (thread_block_1): Move...
18415 (jump_thread_path_registry::thread_block_1): ...here.
18416 (thread_block): Move...
18417 (jump_thread_path_registry::thread_block): ...here.
18418 (thread_through_loop_header): Move...
18419 (jump_thread_path_registry::thread_through_loop_header): ...here.
18420 (mark_threaded_blocks): Move...
18421 (jump_thread_path_registry::mark_threaded_blocks): ...here.
18422 (debug_path): Move...
18423 (jump_thread_path_registry::debug_path): ...here.
18424 (debug_all_paths): Move...
18425 (jump_thread_path_registry::dump): ..here.
18426 (rewire_first_differing_edge): Move...
18427 (jump_thread_path_registry::rewire_first_differing_edge): ...here.
18428 (adjust_paths_after_duplication): Move...
18429 (jump_thread_path_registry::adjust_paths_after_duplication): ...here.
18430 (duplicate_thread_path): Move...
18431 (jump_thread_path_registry::duplicate_thread_path): ..here.
18432 (remove_jump_threads_including): Move...
18433 (jump_thread_path_registry::remove_jump_threads_including): ...here.
18434 (thread_through_all_blocks): Move to...
18435 (jump_thread_path_registry::thread_through_all_blocks): ...here.
18436 (delete_jump_thread_path): Remove.
18437 (register_jump_thread): Move...
18438 (jump_thread_path_registry::register_jump_thread): ...here.
18439 * tree-ssa-threadupdate.h: Move externally facing functions...
18440 (class jump_thread_path_allocator): ...here...
18441 (class jump_thread_path_registry): ...and here.
18442 (thread_through_all_blocks): Remove.
18443 (struct removed_edges): New.
18444 (register_jump_thread): Remove.
18445 (remove_jump_threads_including): Remove.
18446 (delete_jump_thread_path): Remove.
18447 (remove_ctrl_stmt_and_useless_edges): Remove.
18448 (free_dom_edge_info): New prototype.
18449 * tree-vrp.c: Remove x_vr_values hack.
18450 (class vrp_jump_threader_simplifier): New.
18451 (vrp_jump_threader_simplifier::simplify): New.
18452 (vrp_jump_threader::vrp_jump_threader): Adjust method signature.
18453 Remove m_dummy_cond.
18454 Instantiate m_simplifier and m_threader.
18455 (vrp_jump_threader::thread_through_all_blocks): New.
18456 (vrp_jump_threader::simplify_stmt): Remove.
18457 (vrp_jump_threader::after_dom_children): Do not set m_dummy_cond.
18458 Remove x_vr_values hack.
18459 (execute_vrp): Adjust for thread_through_all_blocks being in a
18462 2021-04-30 Christophe Lyon <christophe.lyon@linaro.org>
18464 * genflags.c (gen_insn): Print failed expansion string.
18466 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
18468 * expr.c (alignment_for_piecewise_move): Call mode_for_size
18469 without limit to MAX_FIXED_MODE_SIZE.
18471 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
18473 PR middle-end/90773
18474 * builtins.c (builtin_memset_gen_str): Don't use return from
18475 simplify_gen_subreg.
18477 2021-04-30 Uroš Bizjak <ubizjak@gmail.com>
18480 * config/i386/i386.md (*add<mode>3_carry_0r): New insn pattern.
18481 (*addsi3_carry_zext_0r): Ditto.
18482 (*sub<mode>3_carry_0): Ditto.
18483 (*subsi3_carry_zext_0r): Ditto.
18484 * config/i386/predicates.md (ix86_carry_flag_unset_operator):
18486 * config/i386/i386.c (ix86_rtx_costs) <case PLUS, case MINUS>:
18487 Also consider ix86_carry_flag_unset_operator to calculate
18488 the cost of adc/sbb insn.
18490 2021-04-30 Roman Zhuykov <zhroma@ispras.ru>
18492 PR rtl-optimization/100225
18493 PR rtl-optimization/84878
18494 * modulo-sched.c (sms_schedule): Use note_stores to skip loops
18495 where we have an instruction which touches (writes) any hard
18496 register from df->regular_block_artificial_uses set.
18497 Allow not-single-set instruction only right before basic block
18500 2021-04-30 Geng Qi <gengqi@linux.alibaba.com>
18502 * config/riscv/riscv.opt (march=,mabi=): Negative itself.
18504 2021-04-30 LevyHsu <admin@levyhsu.com>
18506 * config/riscv/riscv.c (riscv_min_arithmetic_precision): New.
18507 * config/riscv/riscv.h (TARGET_MIN_ARITHMETIC_PRECISION): New.
18508 * config/riscv/riscv.md (addv<mode>4, uaddv<mode>4): New.
18509 (subv<mode>4, usubv<mode>4, mulv<mode>4, umulv<mode>4): New.
18511 2021-04-29 Alexandre Oliva <oliva@adacore.com>
18513 * config.gcc: Merged x86 and x86_64 cpu_type-setting cases.
18515 2021-04-29 Alexandre Oliva <oliva@adacore.com>
18517 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Rename to...
18518 (ASM_OUTPUT_MAX_SKIP_ALIGN): ... this. Enclose in do/while(0).
18519 * config/i386/i386.c: Adjust.
18520 * config/i386/i386.md: Adjust.
18521 * config/i386/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Drop.
18522 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18523 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18524 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18525 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18526 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18527 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18528 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18529 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18530 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18531 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
18533 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
18535 * config/i386/i386-expand.c (ix86_expand_int_compare):
18536 Swap operands of GTU and LEU comparison to emit carry flag comparison.
18537 * config/i386/i386.md (*add<mode>3_carry_0): Change insn
18538 predicate to allow more combine opportunities with memory operands.
18539 (*sub<mode>3_carry_0): Ditto.
18541 2021-04-29 Richard Sandiford <richard.sandiford@arm.com>
18543 PR rtl-optimization/100303
18544 * rtl-ssa/accesses.cc (function_info::make_use_available): Take a
18545 boolean that indicates whether the use will only be used in
18546 debug instructions. Treat it in the same way that existing
18547 cross-EBB debug references would be handled if so.
18548 (function_info::make_uses_available): Likewise.
18549 * rtl-ssa/functions.h (function_info::make_uses_available): Update
18550 prototype accordingly.
18551 (function_info::make_uses_available): Likewise.
18552 * fwprop.c (try_fwprop_subst): Update call accordingly.
18554 2021-04-29 Jeff Law <jlaw@tachyum.com>
18556 * config/nios2/nios2-protos.h (nios2_fpu_insn_enabled): Move outside
18559 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
18560 Richard Biener <rguenther@suse.de>
18563 * config/i386/i386-builtin.def (IX86_BUILTIN_MASKLOADPD)
18564 (IX86_BUILTIN_MASKLOADPS, IX86_BUILTIN_MASKLOADPD256)
18565 (IX86_BUILTIN_MASKLOADPS256, IX86_BUILTIN_MASKLOADD)
18566 (IX86_BUILTIN_MASKLOADQ, IX86_BUILTIN_MASKLOADD256)
18567 (IX86_BUILTIN_MASKLOADQ256): Move from SPECIAL_ARGS
18568 to PURE_ARGS category.
18569 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
18570 Handle PURE_ARGS category.
18571 * config/i386/i386-expand.c (ix86_expand_builtin): Ditto.
18573 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
18575 * configure.ac: Check for the presence of sys/locking.h header and
18576 for whether _LK_LOCK is supported by _locking.
18577 * configure: Regenerate.
18578 * config.in: Likewise.
18579 * gcov-io.h: Define GCOV_LOCKED_WITH_LOCKING if HOST_HAS_LK_LOCK.
18580 * gcov-io.c (gcov_open): Add support for GCOV_LOCKED_WITH_LOCKING.
18581 * system.h: Include <sys/locking.h> if HAVE_SYS_LOCKING_H.
18583 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
18585 * config/i386/predicates.md (fcmov_comparison_operator):
18586 Do not check for trivial FP comparison operator.
18587 <case GEU, case LTU>: Allow CCGZmode.
18588 <case GTU, case LEU>: Do not allow CCCmode.
18589 (ix86_comparison_operator) <case GTU, case LEU>: Allow only CCmode.
18590 (ix86_carry_flag_operator): Match only LTU and UNLT code.
18591 Do not check for trivial FP comparison operator. Allow CCGZmode.
18593 2021-04-29 Tom de Vries <tdevries@suse.de>
18595 * omp-expand.c (expand_omp_simd): Add step_orig, and replace uses of
18596 fd->loop.step by either step or orig_step.
18598 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
18600 * config/sparc/sparc.c (gen_load_pcrel_sym): Delete.
18601 (load_got_register): Do the PIC dance here.
18602 (sparc_legitimize_tls_address): Simplify.
18603 (sparc_emit_probe_stack_range): Likewise.
18604 (sparc32_initialize_trampoline): Likewise.
18605 (sparc64_initialize_trampoline): Likewise.
18606 * config/sparc/sparc.md (load_pcrel_sym<P:mode>): Add @ marker.
18607 (probe_stack_range<P:mode>): Likewise.
18608 (flush<P:mode>): Likewise.
18609 (tgd_hi22<P:mode>): Likewise.
18610 (tgd_lo10<P:mode>): Likewise.
18611 (tgd_add<P:mode>): Likewise.
18612 (tgd_call<P:mode>): Likewise.
18613 (tldm_hi22<P:mode>): Likewise.
18614 (tldm_lo10<P:mode>): Likewise.
18615 (tldm_add<P:mode>): Likewise.
18616 (tldm_call<P:mode>): Likewise.
18617 (tldo_hix22<P:mode>): Likewise.
18618 (tldo_lox10<P:mode>): Likewise.
18619 (tldo_add<P:mode>): Likewise.
18620 (tie_hi22<P:mode>): Likewise.
18621 (tie_lo10<P:mode>): Likewise.
18622 (tie_add<P:mode>): Likewise.
18623 (tle_hix22<P:mode>): Likewise.
18624 (tle_lox10<P:mode>): Likewise.
18625 (stack_protect_setsi): Rename to...
18626 (stack_protect_set32): ...this.
18627 (stack_protect_setdi): Rename to...
18628 (stack_protect_set64): ...this.
18629 (stack_protect_set): Adjust calls to above.
18630 (stack_protect_testsi): Rename to...
18631 (stack_protect_test32): ...this.
18632 (stack_protect_testdi): Rename to...
18633 (stack_protect_test64): ...this.
18634 (stack_protect_test): Adjust calls to above.
18636 2021-04-29 H.J. Lu <hjl.tools@gmail.com>
18638 PR middle-end/90773
18639 * builtins.c (builtin_memcpy_read_str): Add a dummy argument.
18640 (builtin_strncpy_read_str): Likewise.
18641 (builtin_memset_read_str): Add an argument for the previous RTL
18642 information and generate the new RTL from the previous RTL info.
18643 (builtin_memset_gen_str): Likewise.
18644 * builtins.h (builtin_strncpy_read_str): Update the prototype.
18645 (builtin_memset_read_str): Likewise.
18646 * expr.c (by_pieces_ninsns): If targetm.overlap_op_by_pieces_p()
18647 returns true, round up size and alignment to the widest integer
18648 mode for maximum size.
18649 (pieces_addr::adjust): Add a pointer to by_pieces_prev argument
18650 and pass it to m_constfn.
18651 (op_by_pieces_d): Add m_push and m_overlap_op_by_pieces.
18652 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
18653 initialize m_push. Initialize m_overlap_op_by_pieces with
18654 targetm.overlap_op_by_pieces_p ().
18655 (op_by_pieces_d::run): Pass the previous RTL information to
18656 pieces_addr::adjust and generate overlapping operations if
18657 m_overlap_op_by_pieces is true.
18659 (move_by_pieces_d::move_by_pieces_d): Updated for op_by_pieces_d
18661 (store_by_pieces_d::store_by_pieces_d): Updated for op_by_pieces_d
18663 (can_store_by_pieces): Use by_pieces_constfn on constfun.
18664 (store_by_pieces): Use by_pieces_constfn on constfun. Updated
18665 for op_by_pieces_d change.
18666 (clear_by_pieces_1): Add a dummy argument.
18667 (clear_by_pieces): Updated for op_by_pieces_d change.
18668 (compare_by_pieces_d::compare_by_pieces_d): Likewise.
18669 (string_cst_read_str): Add a dummy argument.
18670 * expr.h (by_pieces_constfn): Add a dummy argument.
18671 (by_pieces_prev): New.
18672 * target.def (overlap_op_by_pieces_p): New target hook.
18673 * config/i386/i386.c (TARGET_OVERLAP_OP_BY_PIECES_P): New.
18674 * doc/tm.texi.in: Add TARGET_OVERLAP_OP_BY_PIECES_P.
18675 * doc/tm.texi: Regenerated.
18677 2021-04-29 Richard Biener <rguenther@suse.de>
18679 PR tree-optimization/100253
18680 * tree-vect-stmts.c (vectorizable_load): Do not assume
18681 element alignment when DR_MISALIGNMENT is -1.
18682 (vectorizable_store): Likewise.
18684 2021-04-29 Jakub Jelinek <jakub@redhat.com>
18687 * config/aarch64/aarch64.c (aarch64_add_offset_1_temporaries): Use
18688 absu_hwi instead of abs_hwi.
18690 2021-04-29 Richard Biener <rguenther@suse.de>
18692 PR middle-end/38474
18693 * tree-ssa-structalias.c (add_graph_edge): Avoid direct
18694 forwarding when indirect forwarding through ESCAPED
18697 2021-04-29 Tom de Vries <tdevries@suse.de>
18700 * internal-fn.c (expand_GOMP_SIMT_ENTER_ALLOC)
18701 (expand_GOMP_SIMT_LAST_LANE, expand_GOMP_SIMT_ORDERED_PRED)
18702 (expand_GOMP_SIMT_VOTE_ANY, expand_GOMP_SIMT_XCHG_BFLY)
18703 (expand_GOMP_SIMT_XCHG_IDX): Ensure target is assigned to.
18705 2021-04-29 Richard Biener <rguenther@suse.de>
18707 PR tree-optimization/99912
18708 * tree-ssa-dse.c (dse_dom_walker::m_need_cfg_cleanup): New.
18709 (dse_dom_walker::todo): Likewise.
18710 (dse_dom_walker::dse_optimize_stmt): Move VDEF check to the
18712 (dse_dom_walker::before_dom_children): Remove trivially
18713 dead SSA defs and schedule CFG cleanup if we removed all
18715 (pass_dse::execute): Get TODO as computed by the DOM walker
18716 and return it. Wipe dominator info earlier.
18718 2021-04-29 Richard Biener <rguenther@suse.de>
18721 * ipa-prop.c (ipcp_modif_dom_walker::before_dom_children):
18722 Track blocks to cleanup EH in new m_need_eh_cleanup.
18723 (ipcp_modif_dom_walker::cleanup_eh): New.
18724 (ipcp_transform_function): Release dominator info before
18727 2021-04-29 Martin Sebor <msebor@redhat.com>
18729 PR middle-end/100250
18730 * attribs.c (attr_access::array_as_string): Avoid dereferencing
18731 a pointer when it's null.
18733 2021-04-29 Martin Sebor <msebor@redhat.com>
18735 * Makefile.in (OBJS): Add ipa-free-lang-data.o.
18736 * ipa-free-lang-data.cc: New file.
18737 * tree.c: Move pass free_lang_data to file above.
18738 (build_array_type_1): Declare extern.
18739 * tree.h (build_array_type_1): Declare.
18741 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18743 * config/aarch64/aarch64-simd-builtins.def: Modify comment to
18744 make consistent with updated RTL pattern.
18745 * config/aarch64/aarch64-simd.md (aarch64_<sur>qmovn<mode>):
18746 Implement using ss_truncate and us_truncate rather than
18748 * config/aarch64/iterators.md: Remove redundant unspecs and
18749 iterator: UNSPEC_[SU]QXTN and SUQMOVN respectively.
18751 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18753 * config/aarch64/arm_acle.h (__attribute__): Make intrinsic
18754 attributes consistent with those defined in arm_neon.h.
18756 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18758 * config/aarch64/arm_fp16.h (__attribute__): Make intrinsic
18759 attributes consistent with those defined in arm_neon.h.
18761 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18763 * config/aarch64/aarch64-simd-builtins.def: Add
18764 float_trunc_rodd builtin generator macros.
18765 * config/aarch64/aarch64-simd.md (aarch64_float_trunc_rodd_df):
18767 (aarch64_float_trunc_rodd_lo_v2sf): Define.
18768 (aarch64_float_trunc_rodd_hi_v4sf_le): Define.
18769 (aarch64_float_trunc_rodd_hi_v4sf_be): Define.
18770 (aarch64_float_trunc_rodd_hi_v4sf): Define.
18771 * config/aarch64/arm_neon.h (vcvtx_f32_f64): Use RTL builtin
18772 instead of inline asm.
18773 (vcvtx_high_f32_f64): Likewise.
18774 (vcvtxd_f32_f64): Likewise.
18775 * config/aarch64/iterators.md: Add FCVTXN unspec.
18777 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18779 * config/aarch64/aarch64-simd-builtins.def: Add tbx1 builtin
18781 * config/aarch64/aarch64-simd.md (aarch64_tbx1<mode>):
18783 * config/aarch64/arm_neon.h (vqtbx1_s8): USE RTL builtin
18784 instead of inline asm.
18785 (vqtbx1_u8): Likewise.
18786 (vqtbx1_p8): Likewise.
18787 (vqtbx1q_s8): Likewise.
18788 (vqtbx1q_u8): Likewise.
18789 (vqtbx1q_p8): Likewise.
18790 (vtbx2_s8): Likewise.
18791 (vtbx2_u8): Likewise.
18792 (vtbx2_p8): Likewise.
18794 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18796 * config/aarch64/aarch64-simd-builtins.def: Add tbl1 builtin
18798 * config/aarch64/arm_neon.h (vqtbl1_p8): Use RTL builtin
18799 instead of inline asm.
18800 (vqtbl1_s8): Likewise.
18801 (vqtbl1_u8): Likewise.
18802 (vqtbl1q_p8): Likewise.
18803 (vqtbl1q_s8): Likewise.
18804 (vqtbl1q_u8): Likewise.
18805 (vtbl1_s8): Likewise.
18806 (vtbl1_u8): Likewise.
18807 (vtbl1_p8): Likewise.
18808 (vtbl2_s8): Likewise.
18809 (vtbl2_u8): Likewise.
18810 (vtbl2_p8): Likewise.
18812 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18814 * config/aarch64/aarch64-simd-builtins.def: Add polynomial
18815 ssri_n buitin generator macro.
18816 * config/aarch64/arm_neon.h (vsri_n_p8): Use RTL builtin
18817 instead of inline asm.
18818 (vsri_n_p16): Likewise.
18819 (vsri_n_p64): Likewise.
18820 (vsriq_n_p8): Likewise.
18821 (vsriq_n_p16): Likewise.
18822 (vsriq_n_p64): Likewise.
18824 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18826 * config/aarch64/aarch64-simd-builtins.def: Use VALLP mode
18827 iterator for polynomial ssli_n builtin generator macro.
18828 * config/aarch64/arm_neon.h (vsli_n_p8): Use RTL builtin
18829 instead of inline asm.
18830 (vsli_n_p16): Likewise.
18831 (vsliq_n_p8): Likewise.
18832 (vsliq_n_p16): Likewise.
18833 * config/aarch64/iterators.md: Define VALLP mode iterator.
18835 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18837 * config/aarch64/aarch64-simd-builtins.def: Use VDQV_L
18838 iterator to generate [su]adalp RTL builtins.
18839 * config/aarch64/aarch64-simd.md: Use VDQV_L iterator in
18840 [su]adalp RTL pattern.
18841 * config/aarch64/arm_neon.h (vpadal_s32): Use RTL builtin
18842 instead of inline asm.
18843 (vpadal_u32): Likewise.
18845 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18847 * config/aarch64/aarch64-simd-builtins.def: Add [su]addlp
18848 builtin generator macros.
18849 * config/aarch64/aarch64-simd.md (aarch64_<su>addlp<mode>):
18851 * config/aarch64/arm_neon.h (vpaddl_s8): Use RTL builtin
18852 instead of inline asm.
18853 (vpaddl_s16): Likewise.
18854 (vpaddl_s32): Likewise.
18855 (vpaddl_u8): Likewise.
18856 (vpaddl_u16): Likewise.
18857 (vpaddl_u32): Likewise.
18858 (vpaddlq_s8): Likewise.
18859 (vpaddlq_s16): Likewise.
18860 (vpaddlq_s32): Likewise.
18861 (vpaddlq_u8): Likewise.
18862 (vpaddlq_u16): Likewise.
18863 (vpaddlq_u32): Liwewise.
18864 * config/aarch64/iterators.md: Define [SU]ADDLP unspecs with
18865 appropriate attributes.
18867 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18869 * config/aarch64/aarch64-simd-builtins.def: Use VDQ_I iterator
18870 for aarch64_addp<mode> builtin macro generator.
18871 * config/aarch64/aarch64-simd.md: Use VDQ_I iterator in
18872 aarch64_addp<mode> RTL pattern.
18873 * config/aarch64/arm_neon.h (vpaddq_s8): Use RTL builtin
18874 instead of inline asm.
18875 (vpaddq_s16): Likewise.
18876 (vpaddq_s32): Likewise.
18877 (vpaddq_s64): Likewise.
18878 (vpaddq_u8): Likewise.
18879 (vpaddq_u16): Likewise.
18880 (vpaddq_u32): Likewise.
18881 (vpaddq_u64): Likewise.
18883 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18885 * config/aarch64/aarch64-simd-builtins.def: Add sq[r]dmulh_n
18886 builtin generator macros.
18887 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_n<mode>):
18889 * config/aarch64/arm_neon.h (vqdmulh_n_s16): Use RTL builtin
18890 instead of inline asm.
18891 (vqdmulh_n_s32): Likewise.
18892 (vqdmulhq_n_s16): Likewise.
18893 (vqdmulhq_n_s32): Likewise.
18894 (vqrdmulh_n_s16): Likewise.
18895 (vqrdmulh_n_s32): Likewise.
18896 (vqrdmulhq_n_s16): Likewise.
18897 (vqrdmulhq_n_s32): Likewise.
18899 2021-04-28 Tobias Burnus <tobias@codesourcery.com>
18901 * doc/install.texi (--enable-offload-defaulted): Document.
18903 2021-04-28 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
18905 * config/avr/avr-dimode.md: Turn existing patterns into
18906 define_insn_and_split style patterns where the splitter
18907 adds a clobber of the condition code register. Drop "cc"
18908 attribute. Add new patterns to match output of
18910 * config/avr/avr-fixed.md: Likewise.
18911 * config/avr/avr.c (cc_reg_rtx): New.
18912 (avr_parallel_insn_from_insns): Adjust insn count
18913 for removal of set of cc0.
18914 (avr_is_casesi_sequence): Likewise.
18915 (avr_casei_sequence_check_operands): Likewise.
18916 (avr_optimize_casesi): Likewise. Also insert
18917 new insns after jump_insn.
18918 (avr_pass_casesi::avr_rest_of_handle_casesi): Adjust
18919 for removal of set of cc0.
18920 (avr_init_expanders): Initialize cc_reg_rtx.
18921 (avr_regno_reg_class): Handle REG_CC.
18922 (cond_string): Remove usage of CC_OVERFLOW_UNUSABLE.
18923 (avr_notice_update_cc): Remove function.
18924 (ret_cond_branch): Remove usage of CC_OVERFLOW_UNUSABLE.
18925 (compare_condition): Adjust for PARALLEL with
18927 (out_shift_with_cnt): Likewise.
18928 (ashlhi3_out): Likewise.
18929 (ashrhi3_out): Likewise.
18930 (lshrhi3_out): Likewise.
18931 (avr_class_max_nregs): Return single reg for REG_CC.
18932 (avr_compare_pattern): Check for REG_CC instead
18934 (avr_reorg_remove_redundant_compare): Likewise.
18935 (avr_reorg):Adjust for PARALLEL with REG_CC clobber.
18936 (avr_hard_regno_nregs): Return single reg for REG_CC.
18937 (avr_hard_regno_mode_ok): Allow only CCmode for REG_CC.
18938 (avr_md_asm_adjust): Clobber REG_CC.
18939 (TARGET_HARD_REGNO_NREGS): Define.
18940 (TARGET_CLASS_MAX_NREGS): Define.
18941 (TARGET_MD_ASM_ADJUST): Define.
18942 * config/avr/avr.h (FIRST_PSEUDO_REGISTER): Adjust
18944 (enum reg_class): Add CC_REG class.
18945 (NOTICE_UPDATE_CC): Remove.
18946 (CC_OVERFLOW_UNUSABLE): Remove.
18947 (CC_NO_CARRY): Remove.
18948 * config/avr/avr.md: Turn existing patterns into
18949 define_insn_and_split style patterns where the splitter
18950 adds a clobber of the condition code register. Drop "cc"
18951 attribute. Add new patterns to match output of
18953 (sez): Remove unused pattern.
18955 2021-04-28 Richard Earnshaw <rearnsha@arm.com>
18958 * config/arm/arm.c (arm_hard_regno_mode_ok): Only allow VPR to be
18961 2021-04-28 Richard Sandiford <richard.sandiford@arm.com>
18964 * config/aarch64/constraints.md (Utq): Require the address to
18965 be valid for both the element mode and for V2DImode.
18967 2021-04-28 Jakub Jelinek <jakub@redhat.com>
18968 Tobias Burnus <tobias@codesourcery.com>
18970 * configure.ac (OFFLOAD_DEFAULTED): AC_DEFINE if offload-defaulted.
18971 * gcc.c (process_command): New variable.
18972 (driver::maybe_putenv_OFFLOAD_TARGETS): If OFFLOAD_DEFAULTED,
18973 set it if -foffload is defaulted.
18974 * lto-wrapper.c (OFFLOAD_TARGET_DEFAULT_ENV): Define.
18975 (compile_offload_image): If OFFLOAD_DEFAULTED and
18976 OFFLOAD_TARGET_DEFAULT is in the environment, don't fail
18977 if corresponding mkoffload can't be found.
18978 (compile_images_for_offload_targets): Likewise. Free and clear
18979 offload_names if no valid offload is found.
18980 * config.in: Regenerate.
18981 * configure: Regenerate.
18983 2021-04-28 Richard Biener <rguenther@suse.de>
18985 PR tree-optimization/100292
18986 * tree-vect-generic.c (expand_vector_condition): Do not fold
18989 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
18991 * config/rs6000/aix.h (SUBTARGET_DRIVER_SELF_SPECS): New.
18992 * config/rs6000/aix64.opt (m64): New.
18995 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
18997 * config/vax/vax.c (print_operand_address, vax_address_cost_1)
18998 (index_term_p): Handle ASHIFT too.
19000 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
19002 * config/vax/builtins.md (jbb<ccss>i<mode>): Remove operand #3.
19003 (sync_lock_test_and_set<mode>): Adjust accordingly.
19004 (sync_lock_release<mode>): Likewise.
19006 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
19008 * config/vax/vax-protos.h (adjacent_operands_p): Remove
19010 * config/vax/vax.c (adjacent_operands_p): Remove.
19012 2021-04-27 Maciej W. Rozycki <macro@linux-mips.org>
19014 * ifcvt.c (dead_or_predicable) [!IFCVT_MODIFY_TESTS]: Fall
19015 through to the non-conditional execution case if getting the
19016 condition for conditional execution has failed.
19018 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
19020 PR middle-end/100284
19021 * gimple.c (gimple_could_trap_p_1): Remove VEC_COND_EXPR test.
19022 * tree-eh.c (operation_could_trap_p): Handle VEC_COND_EXPR rather
19023 than asserting on it.
19025 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
19027 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): Protect
19028 with TARGET_AIX_OS.
19030 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
19033 * calls.c (precompute_register_parameters): Additionally test
19034 targetm.precompute_tls_p to pre-compute argument.
19035 * config/rs6000/aix.h (TARGET_PRECOMPUTE_TLS_P): Define.
19036 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): New.
19037 * target.def (precompute_tls_p): New.
19038 * doc/tm.texi.in (TARGET_PRECOMPUTE_TLS_P): Add hook documentation.
19039 * doc/tm.texi: Regenerated.
19041 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19044 * config/aarch64/aarch64.c (aarch64_print_operand): Cast -UINTVAL
19045 back to HOST_WIDE_INT.
19047 2021-04-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
19050 * simplify-rtx.c (simplify_context::simplify_subreg): Check the
19051 memory alignment for the outer mode.
19053 2021-04-27 H.J. Lu <hjl.tools@gmail.com>
19055 PR middle-end/90773
19056 * expr.c (op_by_pieces_d::get_usable_mode): New member function.
19057 (op_by_pieces_d::run): Cange a while loop to a do-while loop.
19059 2021-04-27 Alex Coplan <alex.coplan@arm.com>
19062 * config/arm/arm.c (arm_split_compare_and_swap): Fix up codegen
19063 with negative immediates: ensure we expand cbranchsi4_scratch
19064 correctly and ensure we satisfy its constraints.
19065 * config/arm/sync.md
19066 (@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): Don't
19067 attempt to tie two output operands together with constraints;
19068 collapse two alternatives.
19069 (@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
19070 * config/arm/thumb1.md (cbranchsi4_neg_late): New.
19072 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19075 * config/aarch64/predicates.md (aarch64_sub_immediate,
19076 aarch64_plus_immediate): Use -UINTVAL instead of -INTVAL.
19077 * config/aarch64/aarch64.md (casesi, rotl<mode>3): Likewise.
19078 * config/aarch64/aarch64.c (aarch64_print_operand,
19079 aarch64_split_atomic_op, aarch64_expand_subvti): Likewise.
19081 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19083 PR tree-optimization/100239
19084 * tree-vect-generic.c (lower_vec_perm): Don't accept constant
19085 permutations with all indices from the first zero element as vec_shl.
19087 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19089 PR rtl-optimization/100254
19090 * cfgcleanup.c (outgoing_edges_match): Check REG_EH_REGION on
19091 last1 and last2 insns rather than BB_END (bb1) and BB_END (bb2) insns.
19093 2021-04-27 Richard Biener <rguenther@suse.de>
19095 PR tree-optimization/99912
19096 * passes.def: Add comment about new TODO_remove_unused_locals.
19097 * tree-stdarg.c (pass_data_stdarg): Run TODO_remove_unused_locals
19100 2021-04-27 Richard Biener <rguenther@suse.de>
19102 PR tree-optimization/99912
19103 * passes.def (pass_all_optimizations): Add pass_dse before
19104 the first pass_dce, move the first pass_dse before the
19105 pass_dce following pass_pre.
19107 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19109 PR tree-optimization/95527
19110 * generic-match-head.c: Include tm.h.
19111 * gimple-match-head.c: Include tm.h.
19112 * match.pd (CLZ == INTEGER_CST): Don't use
19113 #ifdef CLZ_DEFINED_VALUE_AT_ZERO, only test CLZ_DEFINED_VALUE_AT_ZERO
19114 if clz == CFN_CLZ. Add missing val declaration.
19115 (CTZ cmp CST): New simplifications.
19117 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19119 PR tree-optimization/96696
19120 * expr.c (expand_expr_divmod): New function.
19121 (expand_expr_real_2) <case TRUNC_DIV_EXPR>: Use it for truncations and
19122 divisions. Formatting fixes.
19123 <case MULT_EXPR>: Optimize x / y * y as x - x % y if the latter is
19126 2021-04-27 Martin Jambor <mjambor@suse.cz>
19129 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
19130 If removing a call statement LHS SSA name, release it.
19132 2021-04-27 Richard Earnshaw <rearnsha@arm.com>
19135 * config/arm/arm.c (THUMB2_WORK_REGS): Check PIC_OFFSET_TABLE_REGNUM
19136 is valid before including it in the mask.
19138 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
19141 * config/aarch64/aarch64.c (aarch64_comp_type_attributes): Handle
19144 2021-04-27 Richard Biener <rguenther@suse.de>
19146 PR tree-optimization/100051
19147 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Add
19148 disambiguator based on access size vs. decl size.
19150 2021-04-27 Richard Biener <rguenther@suse.de>
19152 PR tree-optimization/100278
19153 * tree-ssa-pre.c (compute_avail): Give up when we cannot
19154 adjust TBAA beacuse of mismatching bases.
19156 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19159 * config/i386/i386.md (*<insn><mode>3_mask, *<insn><mode>3_mask_1):
19160 For any_rotate define_insn_split and following splitters, use
19161 SWI iterator instead of SWI48.
19163 2021-04-27 Richard Biener <rguenther@suse.de>
19165 PR tree-optimization/99776
19166 * match.pd (bit_field_ref (ctor)): Relax element extract
19167 type compatibility checks.
19169 2021-04-27 Cui,Lili <lili.cui@intel.com>
19171 * common/config/i386/i386-common.c (processor_names):
19172 Sync processor_names with processor_type.
19173 * config/i386/i386-options.c (processor_cost_table):
19174 Sync processor_cost_table with processor_type.
19176 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
19178 * value-range.cc (irange::irange_set_1bit_anti_range): Add assert.
19179 (irange::set): Call irange_set_1bit_anti_range for handling all
19180 1-bit ranges. Fall through on ~[MIN,MAX].
19182 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
19184 * value-range.cc (irange::legacy_num_pairs): Remove.
19185 (irange::invert): Change gcc_assert to gcc_checking_assert.
19186 * value-range.h (irange::num_pairs): Adjust for a cached
19187 num_pairs(). Also, rename all gcc_assert's to
19188 gcc_checking_assert's.
19190 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
19192 * value-range.cc (irange::operator=): Set m_kind.
19193 (irange::copy_to_legacy): Handle varying and undefined sources
19194 as a legacy copy since they can be easily copied.
19195 (irange::irange_set): Set m_kind.
19196 (irange::irange_set_anti_range): Same.
19197 (irange::set): Rename normalize_min_max to normalize_kind.
19198 (irange::verify_range): Adjust for multi-ranges having the
19200 (irange::irange_union): Set m_kind.
19201 (irange::irange_intersect): Same.
19202 (irange::invert): Same.
19203 * value-range.h (irange::kind): Always return m_kind.
19204 (irange::varying_p): Rename to...
19205 (irange::varying_comptaible_p): ...this.
19206 (irange::undefined_p): Only look at m_kind.
19207 (irange::irange): Always set VR_UNDEFINED if applicable.
19208 (irange::set_undefined): Always set VR_UNDEFINED.
19209 (irange::set_varying): Always set m_kind to VR_VARYING.
19210 (irange::normalize_min_max): Rename to...
19211 (irange::normalize_kind): ...this.
19213 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
19215 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
19216 Adjust for constant_p including varying_p.
19217 * tree-vrp.c (vrp_prop::finalize): Same.
19218 (determine_value_range): Same.
19219 * vr-values.c (vr_values::range_of_expr): Same.
19220 * value-range.cc (irange::symbolic_p): Do not check varying_p.
19221 (irange::constant_p): Same.
19223 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
19225 * value-range.cc (irange::legacy_lower_bound): Replace
19226 !undefined_p check with num_ranges > 0.
19227 (irange::legacy_upper_bound): Same.
19228 * value-range.h (irange::type): Same.
19229 (irange::lower_bound): Same.
19230 (irange::upper_bound): Same.
19232 2021-04-26 Richard Biener <rguenther@suse.de>
19234 PR tree-optimization/99956
19235 * gimple-loop-interchange.cc (compute_access_stride):
19236 Try instantiating the access in a shallower loop nest
19237 if instantiating failed.
19238 (compute_access_strides): Pass adjustable loop_nest
19239 to compute_access_stride.
19241 2021-04-26 Christophe Lyon <christophe.lyon@linaro.org>
19243 * doc/sourcebuild.texi (arm_cmse_hw): Document.
19245 2021-04-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19247 * config/aarch64/iterators.md (vwcore): Handle V4BF, V8BF.
19249 2021-04-26 Thomas Schwinge <thomas@codesourcery.com>
19250 Nathan Sidwell <nathan@codesourcery.com>
19251 Tom de Vries <vries@codesourcery.com>
19252 Julian Brown <julian@codesourcery.com>
19253 Kwok Cheung Yeung <kcy@codesourcery.com>
19255 * omp-offload.c (oacc_validate_dims): Implement
19256 '-Wopenacc-parallelism'.
19257 * doc/invoke.texi (-Wopenacc-parallelism): Document.
19259 2021-04-26 Richard Biener <rguenther@suse.de>
19261 * tree-cfg.h (gimplify_build1): Remove.
19262 (gimplify_build2): Likewise.
19263 (gimplify_build3): Likewise.
19264 * tree-cfg.c (gimplify_build1): Move to tree-vect-generic.c.
19265 (gimplify_build2): Likewise.
19266 (gimplify_build3): Likewise.
19267 * tree-vect-generic.c (gimplify_build1): Move from tree-cfg.c.
19269 (gimplify_build2): Likewise.
19270 (gimplify_build3): Likewise.
19271 (tree_vec_extract): Use resimplify with following SSA edges.
19272 (expand_vector_parallel): Avoid passing NULL size/bitpos
19273 to tree_vec_extract.
19274 * expr.c (store_constructor): Deal with zero-element CTORs.
19275 * match.pd (bit_field_ref <vector CTOR>): Make sure to
19276 produce vector constants when possible.
19278 2021-04-26 Richard Biener <rguenther@suse.de>
19280 * tree-complex.c: Include gimple-fold.h.
19281 (expand_complex_addition): Use gimple_build.
19282 (expand_complex_multiplication_components): Likewise.
19283 (expand_complex_multiplication): Likewise.
19284 (expand_complex_div_straight): Likewise.
19285 (expand_complex_div_wide): Likewise.
19286 (expand_complex_division): Likewise.
19287 (expand_complex_conjugate): Likewise.
19288 (expand_complex_comparison): Likewise.
19290 2021-04-26 Richard Biener <rguenther@suse.de>
19292 * tree-ssa-phiopt.c (two_value_replacement): Remove use
19293 of legacy gimplify_buildN API.
19295 2021-04-26 Richard Biener <rguenther@suse.de>
19297 PR tree-optimization/99473
19298 * tree-ssa-phiopt.c (cond_store_replacement): Handle all
19301 2021-04-26 Richard Biener <rguenther@suse.de>
19303 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin):
19304 Use replace_call_with_value.
19306 2021-04-26 Richard Biener <rguenther@suse.de>
19308 * tree-ssa-propagate.h (valid_gimple_rhs_p): Remove.
19309 (update_gimple_call): Likewise.
19310 (update_call_from_tree): Likewise.
19311 * tree-ssa-propagate.c (valid_gimple_rhs_p): Remove.
19312 (valid_gimple_call_p): Likewise.
19313 (move_ssa_defining_stmt_for_defs): Likewise.
19314 (finish_update_gimple_call): Likewise.
19315 (update_gimple_call): Likewise.
19316 (update_call_from_tree): Likewise.
19317 (propagate_tree_value_into_stmt): Use replace_call_with_value.
19318 * gimple-fold.h (update_gimple_call): Declare.
19319 * gimple-fold.c (valid_gimple_rhs_p): Move here from
19320 tree-ssa-propagate.c.
19321 (update_gimple_call): Likewise.
19322 (valid_gimple_call_p): Likewise.
19323 (finish_update_gimple_call): Likewise, and simplify.
19324 (gimplify_and_update_call_from_tree): Implement
19325 update_call_from_tree functionality, avoid excessive
19326 push/pop_gimplify_context.
19327 (gimple_fold_builtin): Use only gimplify_and_update_call_from_tree.
19328 (gimple_fold_call): Likewise.
19329 * gimple-ssa-sprintf.c (try_substitute_return_value): Likewise.
19330 * tree-ssa-ccp.c (ccp_folder::fold_stmt): Likewise.
19331 (pass_fold_builtins::execute): Likewise.
19332 (optimize_stack_restore): Use replace_call_with_value.
19333 * tree-cfg.c (fold_loop_internal_call): Likewise.
19334 * tree-ssa-dce.c (maybe_optimize_arith_overflow): Use
19335 only gimplify_and_update_call_from_tree.
19336 * tree-ssa-strlen.c (handle_builtin_strlen): Likewise.
19337 (handle_builtin_strchr): Likewise.
19338 * tsan.c: Include gimple-fold.h instead of tree-ssa-propagate.h.
19340 2021-04-26 Jakub Jelinek <jakub@redhat.com>
19343 * vmsdbgout.c (ASM_OUTPUT_DEBUG_STRING, vmsdbgout_begin_block,
19344 vmsdbgout_end_block, lookup_filename, vmsdbgout_source_line): Remove
19347 2021-04-25 liuhongt <hongtao.liu@intel.com>
19350 * config/i386/i386-builtin.def (BDESC): Change the icode of
19351 the following builtins to CODE_FOR_nothing.
19352 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
19353 IX86_BUILTIN_PCMPEQB128, IX86_BUILTIN_PCMPEQW128,
19354 IX86_BUILTIN_PCMPEQD128, IX86_BUILTIN_PCMPEQQ,
19355 IX86_BUILTIN_PCMPEQB256, IX86_BUILTIN_PCMPEQW256,
19356 IX86_BUILTIN_PCMPEQD256, IX86_BUILTIN_PCMPEQQ256,
19357 IX86_BUILTIN_PCMPGTB128, IX86_BUILTIN_PCMPGTW128,
19358 IX86_BUILTIN_PCMPGTD128, IX86_BUILTIN_PCMPGTQ,
19359 IX86_BUILTIN_PCMPGTB256, IX86_BUILTIN_PCMPGTW256,
19360 IX86_BUILTIN_PCMPGTD256, IX86_BUILTIN_PCMPGTQ256.
19361 * config/i386/sse.md (avx2_eq<mode>3): Deleted.
19362 (sse2_eq<mode>3): Ditto.
19363 (sse4_1_eqv2di3): Ditto.
19364 (sse2_gt<mode>3): Rename to ..
19365 (*sse2_gt<mode>3): .. this.
19367 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
19370 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
19373 * config/darwin.c (darwin_binds_local_p): Assume that any
19374 public symbol might be interposed for PIC code. Update function
19375 header comment to reflect current Darwin capability.
19377 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
19380 * config/darwin.c (darwin_binds_local_p): Assume that any
19381 public symbol might be interposed for PIC code. Update function
19382 header comment to reflect current Darwin capability.
19384 2021-04-24 Richard Sandiford <richard.sandiford@arm.com>
19386 * doc/sourcebuild.texi: Document no-opts and any-opts target
19389 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
19391 * config/bpf/bpf.h (ASM_OUTPUT_ALIGNED_BSS): Use .type and .lcomm.
19393 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
19395 * config/bpf/bpf.h (FUNCTION_BOUNDARY): Set to 64.
19397 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
19400 * config/i386/i386-options.c (ix86_option_override_internal):
19401 Error out when -m96bit-long-double is used with 64bit targets.
19402 * config/i386/i386.md (*pushxf_rounded): Remove pattern.
19404 2021-04-23 Martin Liska <mliska@suse.cz>
19406 * lto-wrapper.c: Remove FIXME about usage of
19407 hardware_concurrency. The function is not on par with
19410 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
19413 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
19414 Copy operand 3 to operand 4. Use sse_reg_operand
19415 as operand 3 predicate.
19416 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): Ditto.
19417 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
19418 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
19419 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2):
19420 Copy operand 1 to operand 0.
19421 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): Ditto.
19422 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
19423 (LDX_ATOMIC/LDX_ATOMIC FP store peephole2 with mem blockage): Ditto.
19425 2021-04-23 Alex Coplan <alex.coplan@arm.com>
19427 PR rtl-optimization/100230
19428 * early-remat.c (early_remat::sort_candidates): Use delete[]
19429 instead of delete for array allocated with new[].
19431 2021-04-23 Richard Biener <rguenther@suse.de>
19433 * genmatch.c (lower_cond): Remove VEC_COND_EXPR special-casing.
19434 (capture_info::capture_info): Likewise.
19435 (capture_info::walk_match): Likewise.
19436 (expr::gen_transform): Likewise.
19437 (dt_simplify::gen_1): Likewise.
19438 * gimple-match-head.c (maybe_resimplify_conditional_op):
19439 Remove VEC_COND_EXPR special-casing.
19440 (gimple_simplify): Likewise.
19441 * gimple.c (gimple_could_trap_p_1): Adjust.
19442 * tree-ssa-pre.c (compute_avail): Allow VEC_COND_EXPR
19443 to participate in PRE.
19445 2021-04-23 Richard Biener <rguenther@suse.de>
19447 * cfganal.c (connect_infinite_loops_to_exit): First call
19448 add_noreturn_fake_exit_edges.
19449 * ipa-sra.c (process_scan_results): Do not call the now redundant
19450 add_noreturn_fake_exit_edges.
19451 * predict.c (tree_estimate_probability): Likewise.
19452 (rebuild_frequencies): Likewise.
19453 * store-motion.c (one_store_motion_pass): Likewise.
19455 2021-04-23 Richard Biener <rguenther@suse.de>
19457 PR tree-optimization/100222
19458 * predict.c (pass_profile::execute): Remove redundant call to
19459 mark_irreducible_loops.
19460 (report_predictor_hitrates): Likewise.
19462 2021-04-23 Richard Biener <rguenther@suse.de>
19464 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Avoid
19465 valid_gimple_rhs_p by instead gimplifying to one.
19467 2021-04-23 Richard Biener <rguenther@suse.de>
19469 PR tree-optimization/99971
19470 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
19471 Always use TBAA for loads.
19473 2021-04-23 liuhongt <hongtao.liu@intel.com>
19476 * config/i386/i386-options.c (ix86_option_override_internal):
19477 Clear MASK_AVX256_SPLIT_UNALIGNED_LOAD/STORE in x_target_flags
19478 when X86_TUNE_AVX256_UNALIGNED_LOAD/STORE_OPTIMAL is enabled
19479 by target attribute.
19481 2021-04-23 David Edelsohn <dje.gcc@gmail.com>
19483 * config/rs6000/aix71.h (PREFERRED_DEBUGGING_TYPE): Change to
19485 * config/rs6000/aix72.h (PREFERRED_DEBUGGING_TYPE): Same.
19487 2021-04-22 David Edelsohn <dje.gcc@gmail.com>
19489 * config.gcc (powerpc-ibm-aix6.*): Remove.
19490 * config/rs6000/aix61.h: Delete.
19492 2021-04-22 Martin Liska <mliska@suse.cz>
19494 PR testsuite/100159
19495 PR testsuite/100192
19496 * builtins.c (expand_builtin): Fix typos and missing comments.
19497 * dwarf2out.c (gen_subprogram_die): Likewise.
19498 (gen_struct_or_union_type_die): Likewise.
19500 2021-04-22 Uroš Bizjak <ubizjak@gmail.com>
19503 * config/i386/i386-expand.c (ix86_expand_convert_uns_sidf_sse):
19504 Remove the sign with FE_DOWNWARD, where x - x = -0.0.
19506 2021-04-21 Iain Sandoe <iain@sandoe.co.uk>
19508 * config/i386/darwin.h (TARGET_64BIT): Remove definition
19509 based on TARGET_ISA_64BIT.
19510 (TARGET_64BIT_P): Remove definition based on
19511 TARGET_ISA_64BIT_P().
19513 2021-04-21 Martin Liska <mliska@suse.cz>
19516 2021-04-21 Martin Liska <mliska@suse.cz>
19518 * lto-wrapper.c (cpuset_popcount): Remove.
19519 (init_num_threads): Remove and use hardware_concurrency.
19521 2021-04-21 Martin Liska <mliska@suse.cz>
19524 * main.c (main): Call toplev::finalize in CHECKING_P mode.
19525 * ipa-modref.c (ipa_modref_c_finalize): summaries are NULL
19526 when incremental LTO linking happens.
19528 2021-04-21 Martin Liska <mliska@suse.cz>
19530 * lto-wrapper.c (run_gcc): When -flto=jobserver is used, but the
19531 makeserver cannot be detected, then use -flto=N fallback.
19533 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
19535 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): When cross-compiling,
19536 default to yes for aarch64-linux-gnu.
19537 * configure: Regenerate.
19539 2021-04-21 Martin Liska <mliska@suse.cz>
19541 * lto-wrapper.c (cpuset_popcount): Remove.
19542 (init_num_threads): Remove and use hardware_concurrency.
19544 2021-04-21 Martin Liska <mliska@suse.cz>
19546 * config/i386/i386.c: Remove superfluous || TARGET_MACHO
19547 which remains to be '(... || 0)' and clang complains about it.
19548 * dwarf2out.c (AT_vms_delta): Declare conditionally.
19549 (add_AT_vms_delta): Likewise.
19550 * tree.c (fld_simplified_type): Use rather more common pattern
19551 for disabling of something (#if 0).
19552 (get_tree_code_name): Likewise.
19553 (verify_type_variant): Likewise.
19555 2021-04-21 Martin Liska <mliska@suse.cz>
19557 * config/i386/i386-expand.c (decide_alignment): Use newly named
19558 macro TARGET_CPU_P.
19559 * config/i386/i386.c (ix86_decompose_address): Likewise.
19560 (ix86_address_cost): Likewise.
19561 (ix86_lea_outperforms): Likewise.
19562 (ix86_avoid_lea_for_addr): Likewise.
19563 (ix86_add_stmt_cost): Likewise.
19564 * config/i386/i386.h (TARGET_*): Remove.
19565 (TARGET_CPU_P): New macro.
19566 * config/i386/i386.md: Use newly named macro TARGET_CPU_P.
19567 * config/i386/x86-tune-sched-atom.c (do_reorder_for_imul): Likewise.
19568 (swap_top_of_ready_list): Likewise.
19569 (ix86_atom_sched_reorder): Likewise.
19570 * config/i386/x86-tune-sched-bd.c (ix86_bd_has_dispatch): Likewise.
19571 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Likewise.
19573 2021-04-21 Martin Liska <mliska@suse.cz>
19575 * config/i386/i386-options.c (TARGET_EXPLICIT_NO_SAHF_P):
19577 (SET_TARGET_NO_SAHF): Likewise.
19578 (TARGET_EXPLICIT_PREFETCH_SSE_P): Likewise.
19579 (SET_TARGET_PREFETCH_SSE): Likewise.
19580 (TARGET_EXPLICIT_NO_TUNE_P): Likewise.
19581 (SET_TARGET_NO_TUNE): Likewise.
19582 (TARGET_EXPLICIT_NO_80387_P): Likewise.
19583 (SET_TARGET_NO_80387): Likewise.
19585 * config/i386/i386.h (TARGET_*): Remove.
19586 * opth-gen.awk: Generate new used macros.
19588 2021-04-21 Martin Liska <mliska@suse.cz>
19590 * config/i386/i386.h (PTA_*): Remove.
19591 (enum pta_flag): New.
19592 (DEF_PTA): Generate PTA_* values from i386-isa.def.
19593 * config/i386/i386-isa.def: New file.
19595 2021-04-21 Alex Coplan <alex.coplan@arm.com>
19598 * config/aarch64/aarch64-bti-insert.c (aarch64_bti_j_insn_p): New.
19599 (rest_of_insert_bti): Avoid inserting duplicate bti j insns for
19600 jump table targets.
19602 2021-04-21 H.J. Lu <hjl.tools@gmail.com>
19604 * config.gcc: Install mwaitintrin.h for i[34567]86-*-* and
19605 x86_64-*-* targets.
19606 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_MWAIT_SET):
19608 (OPTION_MASK_ISA2_MWAIT_UNSET): Likewise.
19609 (ix86_handle_option): Handle -mmwait.
19610 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
19611 Replace OPTION_MASK_ISA_SSE3 with OPTION_MASK_ISA2_MWAIT on
19612 __builtin_ia32_monitor and __builtin_ia32_mwait.
19613 * config/i386/i386-options.c (isa2_opts): Add -mmwait.
19614 (ix86_valid_target_attribute_inner_p): Likewise.
19615 (ix86_option_override_internal): Enable mwait/monitor
19616 instructions for -msse3.
19617 * config/i386/i386.h (TARGET_MWAIT): New.
19618 (TARGET_MWAIT_P): Likewise.
19619 * config/i386/i386.opt: Add -mmwait.
19620 * config/i386/mwaitintrin.h: New file.
19621 * config/i386/pmmintrin.h: Include <mwaitintrin.h>.
19622 * config/i386/sse.md (sse3_mwait): Replace TARGET_SSE3 with
19624 (@sse3_monitor_<mode>): Likewise.
19625 * config/i386/x86gprintrin.h: Include <mwaitintrin.h>.
19626 * doc/extend.texi: Document mwait target attribute.
19627 * doc/invoke.texi: Document -mmwait.
19629 2021-04-21 Martin Liska <mliska@suse.cz>
19631 * config/i386/i386-options.c (DEF_ENUM): Remove it.
19632 * config/i386/i386-opts.h (DEF_ENUM): Likewise.
19633 * config/i386/stringop.def (DEF_ENUM): Likewise.
19635 2021-04-21 Martin Liska <mliska@suse.cz>
19637 * tree-cfg.c (gimple_verify_flow_info): Use qD instead
19638 of print_generic_expr.
19640 2021-04-21 Jakub Jelinek <jakub@redhat.com>
19642 PR rtl-optimization/100148
19643 * cprop.c (constprop_register): Use next_nondebug_insn instead of
19646 2021-04-21 Martin Liska <mliska@suse.cz>
19649 * cgraphunit.c (cgraph_node::analyze): Remove duplicate
19650 free_dominance_info calls.
19652 2021-04-21 Richard Biener <rguenther@suse.de>
19654 * gimple-fold.c (maybe_fold_reference): Remove is_lhs
19655 parameter (and assume it to be false).
19656 (fold_gimple_assign): Adjust, remove all callers of
19657 maybe_fold_reference calling it with is_lhs true.
19658 (gimple_fold_call): Likewise.
19659 (fold_stmt_1): Likewise.
19661 2021-04-21 Richard Biener <rguenther@suse.de>
19663 * fold-const.c (pedantic_non_lvalue_loc): Remove.
19664 (fold_binary_loc): Adjust.
19665 (fold_ternary_loc): Likewise.
19667 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
19669 PR middle-end/100130
19670 * varasm.c (get_block_for_decl): Make sure that any use of the
19671 retain attribute matches the section's retain flag.
19672 (switch_to_section): Check for retain mismatches even when
19673 changing sections, but do not warn if the given decl is the
19674 section's named.decl.
19675 (output_object_block): Pass the first decl in the block (if any)
19676 to switch_to_section.
19678 2021-04-20 H.J. Lu <hjl.tools@gmail.com>
19680 * config/i386/i386-c.c (ix86_target_macros_internal): Define
19681 __CRC32__ for -mcrc32.
19682 * config/i386/i386-options.c (ix86_option_override_internal):
19683 Enable crc32 instruction for -msse4.2.
19684 * config/i386/i386.md (sse4_2_crc32<mode>): Remove TARGET_SSE4_2
19686 (sse4_2_crc32di): Likewise.
19687 * config/i386/ia32intrin.h: Use crc32 target option for CRC32
19690 2021-04-20 Segher Boessenkool <segher@kernel.crashing.org>
19693 * config/rs6000/rs6000.c (rs6000_machine_from_flags): Do not consider
19696 2021-04-20 Martin Liska <mliska@suse.cz>
19698 * doc/invoke.texi: Fix typo.
19699 * params.opt: Likewise.
19701 2021-04-20 Martin Liska <mliska@suse.cz>
19703 * doc/invoke.texi: Document new param.
19705 2021-04-19 Andrew MacLeod <amacleod@redhat.com>
19707 PR tree-optimization/100081
19708 * gimple-range-cache.h (ranger_cache): Inherit from gori_compute
19709 rather than gori_compute_cache.
19710 * gimple-range-gori.cc (is_gimple_logical_p): Move to top of file.
19711 (range_def_chain::m_logical_depth): New member.
19712 (range_def_chain::range_def_chain): Initialize m_logical_depth.
19713 (range_def_chain::get_def_chain): Don't build defchains through more
19714 than LOGICAL_LIMIT logical expressions.
19715 * params.opt (param_ranger_logical_depth): New.
19717 2021-04-19 Richard Earnshaw <rearnsha@arm.com>
19720 * config/arm/arm.c (arm_configure_build_target): Do not strip
19721 extended FPU/SIMD feature bits from the target ISA when -mfpu
19722 is specified (partial revert of r11-8168).
19724 2021-04-19 Thomas Schwinge <thomas@codesourcery.com>
19726 * params.opt (-param=openacc-kernels=): Add.
19727 * omp-oacc-kernels-decompose.cc
19728 (pass_omp_oacc_kernels_decompose::gate): Use it.
19729 * doc/invoke.texi (-fopenacc-kernels=@var{mode}): Move...
19730 (--param): ... here, 'openacc-kernels'.
19732 2021-04-19 Martin Liska <mliska@suse.cz>
19735 * gengtype.c (finish_root_table): Align function arguments
19736 in between declaration and definition.
19738 2021-04-19 Eric Botcazou <ebotcazou@adacore.com>
19740 * config/i386/winnt.c (i386_pe_seh_cold_init): Properly deal with
19741 frames larger than the SEH maximum frame size.
19743 2021-04-18 Segher Boessenkool <segher@kernel.crashing.org>
19745 PR rtl-optimization/99927
19746 * combine.c (distribute_notes) [REG_UNUSED]: If the register already
19747 is dead, just drop it.
19749 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
19752 * config/i386/winnt-d.c (TARGET_D_TEMPLATES_ALWAYS_COMDAT): Define.
19753 * doc/tm.texi: Regenerate.
19754 * doc/tm.texi.in (D language and ABI): Add @hook for
19755 TARGET_D_TEMPLATES_ALWAYS_COMDAT.
19757 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
19759 * config/darwin-d.c (darwin_d_handle_target_object_format): New
19761 (darwin_d_register_target_info): New function.
19762 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19763 * config/dragonfly-d.c (dragonfly_d_handle_target_object_format): New
19765 (dragonfly_d_register_target_info): New function.
19766 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19767 * config/freebsd-d.c (freebsd_d_handle_target_object_format): New
19769 (freebsd_d_register_target_info): New function.
19770 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19771 * config/glibc-d.c (glibc_d_handle_target_object_format): New
19773 (glibc_d_register_target_info): New function.
19774 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19775 * config/i386/i386-d.c (ix86_d_handle_target_object_format): New
19777 (ix86_d_register_target_info): Add ix86_d_handle_target_object_format
19778 as handler for objectFormat key.
19779 * config/i386/winnt-d.c (winnt_d_handle_target_object_format): New
19781 (winnt_d_register_target_info): New function.
19782 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19783 * config/netbsd-d.c (netbsd_d_handle_target_object_format): New
19785 (netbsd_d_register_target_info): New function.
19786 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19787 * config/openbsd-d.c (openbsd_d_handle_target_object_format): New
19789 (openbsd_d_register_target_info): New function.
19790 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19791 * config/pa/pa-d.c (pa_d_handle_target_object_format): New function.
19792 (pa_d_register_target_info): Add pa_d_handle_target_object_format as
19793 handler for objectFormat key.
19794 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_object_format): New
19796 (rs6000_d_register_target_info): Add
19797 rs6000_d_handle_target_object_format as handler for objectFormat key.
19798 * config/sol2-d.c (solaris_d_handle_target_object_format): New
19800 (solaris_d_register_target_info): New function.
19801 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19803 2021-04-16 Jakub Jelinek <jakub@redhat.com>
19806 * config/aarch64/aarch64.c (aarch64_function_arg_alignment): Change
19807 abi_break argument from bool * to unsigned *, store there the pre-GCC 9
19809 (aarch64_layout_arg, aarch64_gimplify_va_arg_expr): Adjust callers.
19810 (aarch64_function_arg_regno_p): Likewise. Only emit -Wpsabi note if
19811 the old and new alignment after applying MIN/MAX to it is different.
19813 2021-04-16 Tamar Christina <tamar.christina@arm.com>
19816 * config/aarch64/aarch64-sve.md (@aarch64_sve_trn1_conv<mode>): New.
19817 * config/aarch64/aarch64.c (aarch64_expand_sve_const_pred_trn): Use new
19819 * config/aarch64/iterators.md (UNSPEC_TRN1_CONV): New.
19821 2021-04-16 Bill Schmidt <wschmidt@linux.ibm.com>
19823 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Revise
19824 this section and its subsections.
19826 2021-04-16 Jakub Jelinek <jakub@redhat.com>
19829 * config/aarch64/aarch64.md (*neg_asr_si2_extr, *extrsi5_insn_di): New
19830 define_insn patterns.
19832 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
19834 PR rtl-optimization/98689
19835 * reg-notes.def (UNTYPED_CALL): New note.
19836 * combine.c (distribute_notes): Handle it.
19837 * emit-rtl.c (try_split): Likewise.
19838 * rtlanal.c (rtx_properties::try_to_add_insn): Likewise. Assume
19839 that calls with the note implicitly set all return value registers.
19840 * builtins.c (expand_builtin_apply): Add a REG_UNTYPED_CALL
19843 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
19845 PR rtl-optimization/99596
19846 * rtlanal.c (rtx_properties::try_to_add_insn): Don't add global
19847 register accesses for const calls. Assume that pure functions
19848 can only read from global registers. Ignore cases in which
19849 the stack pointer has been marked global.
19851 2021-04-16 Jakub Jelinek <jakub@redhat.com>
19854 * tree-vect-loop.c (vect_transform_loop): Don't remove just
19855 dead scalar .MASK_LOAD calls, but also dead .COND_* calls - replace
19856 them by their last argument.
19858 2021-04-15 Martin Liska <mliska@suse.cz>
19860 * doc/invoke.texi: Other params don't use it, remove it.
19862 2021-04-15 Richard Biener <rguenther@suse.de>
19864 * gimple-builder.h: Add deprecation note.
19866 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
19869 * attribs.h (restrict_type_identity_attributes_to): Declare.
19870 * attribs.c (restrict_type_identity_attributes_to): New function.
19872 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
19875 * attribs.h (affects_type_identity_attributes): Declare.
19876 * attribs.c (remove_attributes_matching): New function.
19877 (affects_type_identity_attributes): Likewise.
19879 2021-04-15 Jakub Jelinek <jakub@redhat.com>
19882 * config/aarch64/aarch64.md (*<LOGICAL:optab>_<SHIFT:optab><mode>3):
19883 Add combine splitters for *<LOGICAL:optab>_ashl<mode>3 with
19884 ZERO_EXTEND, SIGN_EXTEND or AND.
19886 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
19888 PR rtl-optimization/99929
19889 * rtl.h (same_vector_encodings_p): New function.
19890 * cse.c (exp_equiv_p): Check that CONST_VECTORs have the same encoding.
19891 * cselib.c (rtx_equal_for_cselib_1): Likewise.
19892 * jump.c (rtx_renumbered_equal_p): Likewise.
19893 * lra-constraints.c (operands_match_p): Likewise.
19894 * reload.c (operands_match_p): Likewise.
19895 * rtl.c (rtx_equal_p_cb, rtx_equal_p): Likewise.
19897 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
19899 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Print
19900 more information about variable-length CONST_VECTORs.
19902 2021-04-14 Vladimir N. Makarov <vmakarov@redhat.com>
19904 PR rtl-optimization/100066
19905 * lra-constraints.c (split_reg): Check paradoxical_subreg_p for
19906 ordered modes when choosing splitting mode for hard reg.
19908 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
19911 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector_sel):
19913 (aarch64_expand_sve_const_vector): Use it for nelts_per_pattern==2.
19915 2021-04-14 Andreas Krebbel <krebbel@linux.ibm.com>
19917 * config/s390/s390-builtins.def (O_M5, O_M12, ...): Add new macros
19918 for mask operand types.
19919 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
19920 (s390_vec_permi_dbl, s390_vpdi): Use the M5 type for the immediate
19922 (s390_vec_msum_u128, s390_vmslg): Use the M12 type for the
19924 * config/s390/s390.c (s390_const_operand_ok): Check the new
19925 operand types and generate a list of valid values.
19927 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
19929 * doc/tm.texi: Regenerate.
19930 * doc/tm.texi.in (D language and ABI): Add @hook for
19931 TARGET_D_REGISTER_OS_TARGET_INFO.
19933 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
19935 * config/aarch64/aarch64-d.c (aarch64_d_handle_target_float_abi): New
19937 (aarch64_d_register_target_info): New function.
19938 * config/aarch64/aarch64-protos.h (aarch64_d_register_target_info):
19940 * config/aarch64/aarch64.h (TARGET_D_REGISTER_CPU_TARGET_INFO):
19942 * config/arm/arm-d.c (arm_d_handle_target_float_abi): New function.
19943 (arm_d_register_target_info): New function.
19944 * config/arm/arm-protos.h (arm_d_register_target_info): Declare.
19945 * config/arm/arm.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19946 * config/i386/i386-d.c (ix86_d_handle_target_float_abi): New function.
19947 (ix86_d_register_target_info): New function.
19948 * config/i386/i386-protos.h (ix86_d_register_target_info): Declare.
19949 * config/i386/i386.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19950 * config/mips/mips-d.c (mips_d_handle_target_float_abi): New function.
19951 (mips_d_register_target_info): New function.
19952 * config/mips/mips-protos.h (mips_d_register_target_info): Declare.
19953 * config/mips/mips.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19954 * config/pa/pa-d.c (pa_d_handle_target_float_abi): New function.
19955 (pa_d_register_target_info): New function.
19956 * config/pa/pa-protos.h (pa_d_register_target_info): Declare.
19957 * config/pa/pa.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19958 * config/riscv/riscv-d.c (riscv_d_handle_target_float_abi): New
19960 (riscv_d_register_target_info): New function.
19961 * config/riscv/riscv-protos.h (riscv_d_register_target_info): Declare.
19962 * config/riscv/riscv.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19963 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_float_abi): New
19965 (rs6000_d_register_target_info): New function.
19966 * config/rs6000/rs6000-protos.h (rs6000_d_register_target_info):
19968 * config/rs6000/rs6000.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19969 * config/s390/s390-d.c (s390_d_handle_target_float_abi): New function.
19970 (s390_d_register_target_info): New function.
19971 * config/s390/s390-protos.h (s390_d_register_target_info): Declare.
19972 * config/s390/s390.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19973 * config/sparc/sparc-d.c (sparc_d_handle_target_float_abi): New
19975 (sparc_d_register_target_info): New function.
19976 * config/sparc/sparc-protos.h (sparc_d_register_target_info): Declare.
19977 * config/sparc/sparc.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19978 * doc/tm.texi: Regenerate.
19979 * doc/tm.texi.in (D language and ABI): Add @hook for
19980 TARGET_D_REGISTER_CPU_TARGET_INFO.
19982 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
19984 * config/i386/i386-d.c (ix86_d_has_stdcall_convention): New function.
19985 * config/i386/i386-protos.h (ix86_d_has_stdcall_convention): Declare.
19986 * config/i386/i386.h (TARGET_D_HAS_STDCALL_CONVENTION): Define.
19987 * doc/tm.texi: Regenerate.
19988 * doc/tm.texi.in (D language and ABI): Add @hook for
19989 TARGET_D_HAS_STDCALL_CONVENTION.
19991 2021-04-14 Richard Biener <rguenther@suse.de>
19993 * tree-cfg.c (verify_gimple_assign_ternary): Verify that
19994 VEC_COND_EXPRs have a gimple_val condition.
19995 * tree-ssa-propagate.c (valid_gimple_rhs_p): VEC_COND_EXPR
19996 can no longer have a GENERIC condition.
19998 2021-04-14 Richard Earnshaw <rearnsha@arm.com>
20001 * config/arm/arm.c (arm_configure_build_target): Strip isa_all_fpbits
20002 from the isa_delta when -mfpu has been used.
20003 (arm_options_perform_arch_sanity_checks): It's the architecture that
20004 lacks an FPU not the processor.
20006 2021-04-13 Richard Biener <rguenther@suse.de>
20008 PR tree-optimization/100053
20009 * tree-ssa-sccvn.c (vn_nary_op_get_predicated_value): Do
20010 not use optimistic dominance queries for backedges to validate
20012 (dominated_by_p_w_unex): Add parameter to ignore executable
20013 state on backedges.
20014 (rpo_elim::eliminate_avail): Adjust.
20016 2021-04-13 Jakub Jelinek <jakub@redhat.com>
20019 * config/aarch64/aarch64.md (*aarch64_bfxil<mode>_extr,
20020 *aarch64_bfxilsi_extrdi): New define_insn patterns.
20022 2021-04-13 Jakub Jelinek <jakub@redhat.com>
20025 * simplify-rtx.c (simplify_immed_subreg): For MODE_COMPOSITE_P
20026 outermode, return NULL if the result doesn't encode back to the
20027 original byte sequence.
20028 (simplify_gen_subreg): Don't create SUBREGs from constants to
20029 MODE_COMPOSITE_P outermode.
20031 2021-04-12 Jakub Jelinek <jakub@redhat.com>
20033 PR rtl-optimization/99905
20034 * combine.c (expand_compound_operation): If pos + len > modewidth,
20035 perform the right shift by pos in inner_mode and then convert to mode,
20036 instead of trying to simplify a shift of rtx with inner_mode by pos
20037 as if it was a shift in mode.
20039 2021-04-12 Jakub Jelinek <jakub@redhat.com>
20042 * combine.c (simplify_and_const_int_1): Don't optimize varop
20043 away if it has side-effects.
20045 2021-04-12 Martin Liska <mliska@suse.cz>
20047 * doc/extend.texi: Escape @smallexample content.
20049 2021-04-12 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
20051 * config/s390/s390.md ("*movdi_31", "*movdi_64"): Add
20052 alternative in order to load a DFP zero.
20054 2021-04-12 Martin Liska <mliska@suse.cz>
20056 * doc/extend.texi: Be more precise in documentation
20057 of symver attribute.
20059 2021-04-12 Martin Liska <mliska@suse.cz>
20062 * gimplify.c (gimplify_expr): Right now, we unpoison all
20063 variables before a goto <dest>. We should not do it if we are
20066 2021-04-12 Cui,Lili <lili.cui@intel.com>
20068 * common/config/i386/cpuinfo.h (get_intel_cpu): Handle
20070 * common/config/i386/i386-common.c (processor_names): Add
20072 (processor_alias_table): Add rocketlake.
20073 * common/config/i386/i386-cpuinfo.h (processor_subtypes): Add
20074 INTEL_COREI7_ROCKETLAKE.
20075 * config.gcc: Add -march=rocketlake.
20076 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20078 * config/i386/i386-options.c (m_ROCKETLAKE) : Define.
20079 (processor_cost_table): Add rocketlake cost.
20080 * config/i386/i386.h (ix86_size_cost) : Define
20082 (processor_type) : Add PROCESSOR_ROCKETLAKE.
20083 (PTA_ROCKETLAKE): Ditto.
20084 * doc/extend.texi: Add rocketlake.
20085 * doc/invoke.texi: Add rocketlake.
20087 2021-04-12 Cui,Lili <lili.cui@intel.com>
20089 * config/i386/i386.h (PTA_ALDERLAKE): Change alderlake ISA list.
20090 * config/i386/i386-options.c (m_CORE_AVX2): Add m_ALDERLAKE.
20091 * common/config/i386/cpuinfo.h (get_intel_cpu): Add AlderLake model.
20092 * doc/invoke.texi: Change alderlake ISA list.
20094 2021-04-11 Hafiz Abid Qadeer <abidh@codesourcery.com>
20096 PR middle-end/98088
20097 * omp-expand.c (expand_oacc_collapse_init): Update condition in
20100 2021-04-10 H.J. Lu <hjl.tools@gmail.com>
20103 * config/i386/serializeintrin.h (_serialize): Defined as macro.
20105 2021-04-10 Jakub Jelinek <jakub@redhat.com>
20108 * expr.c (expand_expr_addr_expr_1): Test is_global_var rather than
20109 just TREE_STATIC on COMPOUND_LITERAL_EXPR_DECLs.
20111 2021-04-10 Jakub Jelinek <jakub@redhat.com>
20113 PR middle-end/99989
20114 * gimple-ssa-warn-alloca.c
20115 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit to
20116 0 with integer precision unconditionally.
20118 2021-04-10 Jakub Jelinek <jakub@redhat.com>
20120 PR rtl-optimization/98601
20121 * rtlanal.c (rtx_addr_can_trap_p_1): Allow in assert unknown size
20122 not just for BLKmode, but also for VOIDmode. For STRICT_ALIGNMENT
20123 unaligned_mems handle VOIDmode like BLKmode.
20125 2021-04-10 Jan Hubicka <hubicka@ucw.cz>
20128 * tree.c (free_lang_data_in_decl): Do not release body of
20129 declare_variant_alt.
20131 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
20133 * config/aarch64/aarch64.c (aarch64_option_restore): If the
20134 architecture was specified explicitly and the tuning wasn't,
20135 tune for the architecture rather than the configured default CPU.
20137 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
20139 * config/aarch64/aarch64.md (tlsdesc_small_sve_<mode>): Use X30
20140 as the temporary register.
20142 2021-04-09 Martin Liska <mliska@suse.cz>
20144 * doc/extend.texi: Move non-target attributes on the top level.
20146 2021-04-09 Martin Liska <mliska@suse.cz>
20148 * doc/invoke.texi: Document minimum and maximum value of the
20149 argument for both supported compression algorithms.
20151 2021-04-08 David Edelsohn <dje.gcc@gmail.com>
20153 * config/rs6000/rs6000.c (rs6000_xcoff_select_section): Select
20154 TLS BSS before TLS data.
20155 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .comm.
20157 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
20159 * doc/sourcebuild.texi (stdint_types_mbig_endian): Document.
20161 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
20163 * match.pd: Extend vec_cond folds to handle shifts.
20165 2021-04-08 Maciej W. Rozycki <macro@orcam.me.uk>
20167 * config/vax/vax.md: Fix comment for `*bit<mode>' pattern's
20170 2021-04-08 Alex Coplan <alex.coplan@arm.com>
20173 * config/arm/iterators.md (MVE_vecs): New.
20174 (V_elem): Also handle V2DF.
20175 * config/arm/mve.md (*mve_mov<mode>): Rename to ...
20176 (*mve_vdup<mode>): ... this. Remove second alternative since
20177 vec_duplicate of const_int is not canonical RTL, and we don't
20178 want to match symbol_refs.
20179 (*mve_vec_duplicate<mode>): Delete (pattern is redundant).
20181 2021-04-08 Xionghu Luo <luoxhu@linux.ibm.com>
20183 * fold-const.c (fold_single_bit_test): Fix typo.
20184 * print-rtl.c (print_rtx_insn_vec): Call print_rtl_single
20187 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
20189 PR tree-optimization/97513
20190 * tree-vect-slp.c (vect_add_slp_permutation): New function,
20192 (vectorizable_slp_permutation): ...here. Detect cases in which
20193 all VEC_PERM_EXPRs are guaranteed to have the same stepped
20194 permute vector and only generate one permute vector for that case.
20195 Extend that case to handle variable-length vectors.
20197 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
20199 PR tree-optimization/99873
20200 * tree-vect-slp.c (vect_slp_prefer_store_lanes_p): New function.
20201 (vect_build_slp_instance): Don't split store groups that could
20202 use IFN_STORE_LANES.
20204 2021-04-07 Jakub Jelinek <jakub@redhat.com>
20207 * varasm.c (output_constant_pool_contents): Don't strip name encoding
20208 from XSTR (desc->sym, 0) or from label before passing those to
20211 2021-04-07 Richard Biener <rguenther@suse.de>
20213 PR tree-optimization/99954
20214 * tree-loop-distribution.c: Include tree-affine.h.
20215 (generate_memcpy_builtin): Try using tree-affine to prove
20217 (loop_distribution::classify_builtin_ldst): Always classify
20220 2021-04-07 Richard Biener <rguenther@suse.de>
20222 PR tree-optimization/99947
20223 * tree-vect-loop.c (vectorizable_induction): Pre-allocate
20224 steps vector to avoid pushing elements from the reallocated
20227 2021-04-07 Richard Biener <rguenther@suse.de>
20229 * tree-ssa-sccvn.h (print_vn_reference_ops): Declare.
20230 * tree-ssa-pre.c (print_pre_expr): Factor out VN reference operand
20232 * tree-ssa-sccvn.c (print_vn_reference_ops): ... into this new
20234 (debug_vn_reference_ops): New.
20236 2021-04-07 Bin Cheng <bin.cheng@linux.alibaba.com>
20238 PR tree-optimization/98736
20239 * tree-loop-distribution.c
20240 * (loop_distribution::bb_top_order_init):
20241 Compute RPO with programing order preserved by calling function
20242 rev_post_order_and_mark_dfs_back_seme.
20244 2021-04-06 Vladimir N. Makarov <vmakarov@redhat.com>
20247 * lra-constraints.c (split_reg): Don't check paradoxical_subreg_p.
20248 * lra-lives.c (clear_sparseset_regnos, regnos_in_sparseset_p): New
20250 (process_bb_lives): Don't update biggest mode of hard reg for
20251 implicit in multi-register group. Use the new functions for
20252 updating dead_set and unused_set by register notes.
20254 2021-04-06 Xianmiao Qu <xianmiao_qu@c-sky.com>
20256 * config/csky/csky_pipeline_ck802.md : Use insn reservation name
20259 2021-04-06 H.J. Lu <hjl.tools@gmail.com>
20261 * config/i386/x86-tune-costs.h (skylake_memcpy): Updated.
20262 (skylake_memset): Likewise.
20263 (skylake_cost): Change CLEAR_RATIO to 17.
20264 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
20265 Replace m_CANNONLAKE, m_ICELAKE_CLIENT, m_ICELAKE_SERVER,
20266 m_TIGERLAKE and m_SAPPHIRERAPIDS with m_SKYLAKE and m_CORE_AVX512.
20268 2021-04-06 Richard Biener <rguenther@suse.de>
20270 PR tree-optimization/99880
20271 * tree-vect-loop.c (maybe_set_vectorized_backedge_value): Only
20272 set vectorized defs of relevant PHIs.
20274 2021-04-06 Richard Biener <rguenther@suse.de>
20276 PR tree-optimization/99924
20277 * tree-vect-slp.c (vect_bb_partition_graph_r): Do not mark
20278 nodes w/o scalar stmts as visited.
20280 2021-04-06 Alex Coplan <alex.coplan@arm.com>
20283 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Also use base
20284 PCS for [su]fix_optab.
20286 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
20288 * config/darwin.c (machopic_legitimize_pic_address): Check
20289 that the current pic register is one of the hard reg set
20290 before setting liveness.
20292 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
20294 * config/darwin.c (machopic_legitimize_pic_address): Fix
20295 whitespace, remove unused code.
20297 2021-04-03 Jakub Jelinek <jakub@redhat.com>
20299 PR tree-optimization/99882
20300 * gimple-ssa-store-merging.c (bswap_view_convert): Handle val with
20303 2021-04-03 Jakub Jelinek <jakub@redhat.com>
20305 PR rtl-optimization/99863
20306 * dse.c (replace_read): Drop regs_live argument. Instead of
20307 regs_live, use store_insn->fixed_regs_live if non-NULL,
20308 otherwise punt if insns sequence clobbers or sets any hard
20311 2021-04-03 Jakub Jelinek <jakub@redhat.com>
20314 * targhooks.h (default_print_patchable_function_entry_1): Declare.
20315 * targhooks.c (default_print_patchable_function_entry_1): New function,
20316 copied from default_print_patchable_function_entry with an added flags
20318 (default_print_patchable_function_entry): Rewritten into a small
20319 wrapper around default_print_patchable_function_entry_1.
20320 * config/rs6000/rs6000.c (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
20322 (rs6000_print_patchable_function_entry): New function.
20324 2021-04-02 Eric Botcazou <ebotcazou@adacore.com>
20326 * doc/invoke.texi (fdelete-dead-exceptions): Minor tweak.
20328 2021-04-01 Jason Merrill <jason@redhat.com>
20331 * common.opt: Document v15 and v16.
20333 2021-04-01 Richard Biener <rguenther@suse.de>
20335 PR tree-optimization/99863
20336 * gimplify.c (gimplify_init_constructor): Recompute vector
20339 2021-04-01 Jakub Jelinek <jakub@redhat.com>
20341 * doc/extend.texi (symver attribute): Fix up syntax errors
20344 2021-04-01 Jakub Jelinek <jakub@redhat.com>
20346 PR tree-optimization/96573
20347 * gimple-ssa-store-merging.c (init_symbolic_number): Handle
20348 also pointer types.
20350 2021-04-01 Richard Biener <rguenther@suse.de>
20352 PR tree-optimization/99856
20353 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Promote
20354 precision to vector element precision.
20356 2021-04-01 Martin Jambor <mjambor@suse.cz>
20358 PR tree-optimization/97009
20359 * tree-sra.c (access_or_its_child_written): New function.
20360 (propagate_subaccesses_from_rhs): Use it instead of a simple grp_write
20363 2021-03-31 Jan Hubicka <hubicka@ucw.cz>
20366 * cif-code.def (USES_COMDAT_LOCAL): Make CIF_FINAL_NORMAL.
20368 2021-03-31 Pat Haugen <pthaugen@linux.ibm.com>
20371 * config/rs6000/altivec.md (xxspltiw_v4si, xxspltiw_v4sf_inst,
20372 xxspltidp_v2df_inst, xxsplti32dx_v4si_inst, xxsplti32dx_v4sf_inst,
20373 xxblend_<mode>, xxpermx_inst, xxeval): Mark prefixed.
20374 * config/rs6000/mma.md (mma_<vvi4i4i8>, mma_<avvi4i4i8>,
20375 mma_<vvi4i4i2>, mma_<avvi4i4i2>, mma_<vvi4i4>, mma_<avvi4i4>,
20376 mma_<pvi4i2>, mma_<apvi4i2>, mma_<vvi4i4i4>, mma_<avvi4i4i4>):
20378 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Adjust test.
20379 * config/rs6000/rs6000.md (define_attr "maybe_prefixed"): New.
20380 (define_attr "prefixed"): Update initializer.
20382 2021-03-31 Jakub Jelinek <jakub@redhat.com>
20385 * dwarf2out.c (debug_ranges_dwo_section): New variable.
20386 (DW_RANGES_IDX_SKELETON): Define.
20387 (struct dw_ranges): Add begin_entry and end_entry members.
20388 (DEBUG_DWO_RNGLISTS_SECTION): Define.
20389 (add_ranges_num): Adjust r initializer for addition of *_entry
20391 (add_ranges_by_labels): For -gsplit-dwarf and force_direct,
20392 set idx to DW_RANGES_IDX_SKELETON.
20393 (use_distinct_base_address_for_range): New function.
20394 (index_rnglists): Don't set r->idx if it is equal to
20395 DW_RANGES_IDX_SKELETON. Initialize r->begin_entry and
20396 r->end_entry for -gsplit-dwarf if those will be needed by
20398 (output_rnglists): Add DWO argument. If true, switch to
20399 debug_ranges_dwo_section rather than debug_ranges_section.
20400 Adjust l1/l2 label indexes. Only output the offset table when
20401 dwo is true and don't include in there the skeleton range
20402 entry if present. For -gsplit-dwarf, skip ranges that belong
20403 to the other rnglists section. Change return type from void
20404 to bool and return true if there are any range entries for
20405 the other section. For dwarf_split_debug_info use
20406 DW_RLE_startx_endx, DW_RLE_startx_length and DW_RLE_base_addressx
20407 entries instead of DW_RLE_start_end, DW_RLE_start_length and
20408 DW_RLE_base_address. Use use_distinct_base_address_for_range.
20409 (init_sections_and_labels): Initialize debug_ranges_dwo_section
20410 if -gsplit-dwarf and DWARF >= 5. Adjust ranges_section_label
20411 and range_base_label indexes.
20412 (dwarf2out_finish): Call index_rnglists earlier before finalizing
20413 .debug_addr. Never emit DW_AT_rnglists_base attribute. For
20414 -gsplit-dwarf and DWARF >= 5 call output_rnglists up to twice
20415 with different dwo arguments.
20416 (dwarf2out_c_finalize): Clear debug_ranges_dwo_section.
20418 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
20420 PR tree-optimization/98268
20421 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Call
20422 recompute_tree_invariant_for_addr_expr after successfully
20423 folding a TARGET_MEM_REF that occurs inside an ADDR_EXPR.
20425 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
20427 PR tree-optimization/99726
20428 * tree-data-ref.c (create_intersect_range_checks_index): Bail
20429 out if there is more than one access function SCEV for the loop
20432 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
20434 PR rtl-optimization/97141
20435 PR rtl-optimization/98726
20436 * emit-rtl.c (valid_for_const_vector_p): Return true for
20438 * rtx-vector-builder.h (rtx_vector_builder::step): Return a
20439 poly_wide_int instead of a wide_int.
20440 (rtx_vector_builder::apply_set): Take a poly_wide_int instead
20442 * rtx-vector-builder.c (rtx_vector_builder::apply_set): Likewise.
20443 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Return
20444 false for CONST_VECTORs that cannot be forced to memory.
20445 * config/aarch64/aarch64-simd.md (mov<mode>): If a CONST_VECTOR
20446 is too complex to force to memory, build it up from individual
20449 2021-03-31 Jan Hubicka <jh@suse.cz>
20452 * cgraph.c (cgraph_node::release_body): Fix overactive check.
20454 2021-03-31 Christophe Lyon <christophe.lyon@linaro.org>
20457 * config/arm/vec-common.md (mul<mode>3): Disable on iwMMXT, expect
20460 2021-03-31 H.J. Lu <hjl.tools@gmail.com>
20462 * config/i386/i386-expand.c (expand_set_or_cpymem_via_rep):
20463 For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, don't convert QImode
20465 (decide_alg): For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, use
20466 "rep movsb/stosb" only for known sizes.
20467 * config/i386/i386-options.c (processor_cost_table): Use Ice
20468 Lake cost for Cannon Lake, Ice Lake, Tiger Lake, Sapphire
20469 Rapids and Alder Lake.
20470 * config/i386/i386.h (TARGET_PREFER_KNOWN_REP_MOVSB_STOSB): New.
20471 * config/i386/x86-tune-costs.h (icelake_memcpy): New.
20472 (icelake_memset): Likewise.
20473 (icelake_cost): Likewise.
20474 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
20477 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
20480 * config/aarch64/aarch64.c
20481 (aarch64_vectorize_preferred_vector_alignment): Query the size
20482 of the provided SVE vector; do not assume that all SVE vectors
20483 have the same size.
20485 2021-03-31 Jan Hubicka <jh@suse.cz>
20488 * cgraph.c (cgraph_node::release_body): Remove all callers and
20490 * cgraphclones.c (cgraph_node::materialize_clone): Do not do it here.
20491 * cgraphunit.c (cgraph_node::expand): And here.
20493 2021-03-31 Martin Liska <mliska@suse.cz>
20495 * ipa-modref.c (analyze_ssa_name_flags): Fix coding style
20496 and one negated condition.
20498 2021-03-31 Jakub Jelinek <jakub@redhat.com>
20499 Richard Sandiford <richard.sandiford@arm.com>
20502 * config/aarch64/aarch64.md (*add<mode>3_poly_1): Swap Uai and Uav
20503 constraints on operands[2] and similarly 0 and rk constraints
20504 on operands[1] corresponding to that.
20506 2021-03-31 Jakub Jelinek <jakub@redhat.com>
20509 * configure.ac (HAVE_LD_BROKEN_PE_DWARF5): New AC_DEFINE if PECOFF
20510 linker doesn't support DWARF sections new in DWARF5.
20511 * config/i386/i386-options.c (ix86_option_override_internal): Default
20512 to dwarf_version 4 if HAVE_LD_BROKEN_PE_DWARF5 for TARGET_PECOFF
20514 * config.in: Regenerated.
20515 * configure: Regenerated.
20517 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20520 * config/aarch64/aarch64.c (aarch64_analyze_loop_vinfo): Check for
20521 available issue_info before using it.
20523 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20526 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): Do not allow zero
20529 2021-03-30 Xionghu Luo <luoxhu@linux.ibm.com>
20532 * config/rs6000/altivec.md (altivec_lvsl_reg): Change to ...
20533 (altivec_lvsl_reg_<mode>): ... this.
20534 (altivec_lvsr_reg): Change to ...
20535 (altivec_lvsr_reg_<mode>): ... this.
20536 * config/rs6000/predicates.md (vec_set_index_operand): New.
20537 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20538 Enable 32bit variable vec_insert for all TARGET_VSX.
20539 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
20540 Enable 32bit variable vec_insert for p9 and above.
20541 (rs6000_expand_vector_set_var_p8): Rename to ...
20542 (rs6000_expand_vector_set_var_p7): ... this.
20543 (rs6000_expand_vector_set): Use TARGET_VSX and adjust assert
20545 * config/rs6000/vector.md (vec_set<mode>): Use vec_set_index_operand.
20546 * config/rs6000/vsx.md (xl_len_r): Use gen_altivec_lvsl_reg_di and
20547 gen_altivec_lvsr_reg_di.
20549 2021-03-30 H.J. Lu <hjl.tools@gmail.com>
20552 * config/i386/ia32intrin.h (__rdtsc): Defined as macro.
20553 (__rdtscp): Likewise.
20555 2021-03-30 Tamar Christina <tamar.christina@arm.com>
20557 PR tree-optimization/99825
20558 * tree-vect-slp-patterns.c (vect_check_evenodd_blend):
20559 Reject non-mult 2 lanes.
20561 2021-03-30 Richard Earnshaw <rearnsha@arm.com>
20564 * config/arm/arm.c (arm_file_start): Fix emission of
20565 Tag_ABI_VFP_args attribute.
20567 2021-03-30 Richard Biener <rguenther@suse.de>
20569 PR tree-optimization/99824
20570 * stor-layout.c (set_min_and_max_values_for_integral_type):
20571 Assert the precision is within the bounds of
20572 WIDE_INT_MAX_PRECISION.
20573 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
20574 the outermost component ref only to lower the access size
20575 and initialize that from the access type.
20577 2021-03-30 Richard Sandiford <richard.sandiford@arm.com>
20580 * config/aarch64/aarch64.md (mov<mode>): Pass multi-instruction
20581 CONST_INTs to aarch64_expand_mov_immediate when called after RA.
20583 2021-03-30 Mihailo Stojanovic <mihailo.stojanovic@typhoon-hil.com>
20585 * config/aarch64/aarch64.md
20586 (<optab>_trunc<fcvt_target><GPI:mode>2): Set the "arch"
20587 attribute to disambiguate between SIMD and FP variants of the
20590 2021-03-29 Jan Hubicka <hubicka@ucw.cz>
20592 * ipa-modref.c (merge_call_lhs_flags): Correct handling of deref.
20593 (analyze_ssa_name_flags): Fix typo in comment.
20595 2021-03-29 Alex Coplan <alex.coplan@arm.com>
20598 * config/aarch64/aarch64-sve-builtins.cc
20599 (function_builder::add_function): Add placeholder_p argument, use
20600 placeholder decls if this is set.
20601 (function_builder::add_unique_function): Instead of conditionally adding
20602 direct overloads, unconditionally add either a direct overload or a
20604 (function_builder::add_overloaded_function): Set placeholder_p if we're
20605 using C++ overloads. Use the obstack for string storage instead
20606 of relying on the tree nodes.
20607 (function_builder::add_overloaded_functions): Don't return early for
20608 m_direct_overloads: we need to add placeholders.
20609 * config/aarch64/aarch64-sve-builtins.h
20610 (function_builder::add_function): Add placeholder_p argument.
20612 2021-03-29 Richard Biener <rguenther@suse.de>
20614 PR tree-optimization/99807
20615 * tree-vect-slp.c (vect_slp_analyze_node_operations_1): Move
20616 assert below VEC_PERM handling.
20618 2021-03-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20621 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>): Use
20622 aarch64_simd_or_scalar_imm_zero to match zeroes. Remove pattern
20623 matching const_int 0.
20624 (move_lo_quad_internal_be_<mode>): Likewise.
20625 (move_lo_quad_<mode>): Update for the above.
20626 * config/aarch64/iterators.md (VQ_2E): Delete.
20628 2021-03-29 Jakub Jelinek <jakub@redhat.com>
20630 PR tree-optimization/99777
20631 * fold-const.c (extract_muldiv_1): For conversions, punt on casts from
20632 types other than scalar integral types.
20634 2021-03-28 David Edelsohn <dje.gcc@gmail.com>
20636 * config/rs6000/rs6000.c (rs6000_output_dwarf_dtprel): Do not add
20637 XCOFF TLS reloc decorations.
20639 2021-03-28 Gerald Pfeifer <gerald@pfeifer.com>
20641 * doc/analyzer.texi (Analyzer Internals): Update link to
20642 "A Memory Model for Static Analysis of C Programs".
20644 2021-03-26 David Edelsohn <dje.gcc@gmail.com>
20646 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Call function.
20647 * config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align):
20649 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align): New.
20650 (rs6000_special_round_type_align): Recursively check innermost first
20653 2021-03-26 Jakub Jelinek <jakub@redhat.com>
20656 * dwarf2out.h (struct dw_fde_node): Add rule18 member.
20657 * dwarf2cfi.c (dwarf2out_frame_debug_expr): When handling (set hfp sp)
20658 assignment with drap_reg active, queue reg save for hfp with offset 0
20659 and flush queued reg saves. When handling a push with rule18,
20660 defer queueing reg save for hfp and just assert the offset is 0.
20661 (scan_trace): Assert that fde->rule18 is false.
20663 2021-03-26 Vladimir Makarov <vmakarov@redhat.com>
20666 * ira-costs.c (record_reg_classes): Put case with
20667 CT_RELAXED_MEMORY adjacent to one with CT_MEMORY.
20668 * ira.c (ira_setup_alts): Ditto.
20669 * lra-constraints.c (process_alt_operands): Ditto.
20670 * recog.c (asm_operand_ok): Ditto.
20671 * reload.c (find_reloads): Ditto.
20673 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20675 * config/aarch64/aarch64-protos.h
20676 (cpu_addrcost_table::post_modify_ld3_st3): New member variable.
20677 (cpu_addrcost_table::post_modify_ld4_st4): Likewise.
20678 * config/aarch64/aarch64.c (generic_addrcost_table): Update
20679 accordingly, using the same costs as for post_modify.
20680 (exynosm1_addrcost_table, xgene1_addrcost_table): Likewise.
20681 (thunderx2t99_addrcost_table, thunderx3t110_addrcost_table):
20682 (tsv110_addrcost_table, qdf24xx_addrcost_table): Likewise.
20683 (a64fx_addrcost_table): Likewise.
20684 (neoversev1_addrcost_table): New.
20685 (neoversev1_tunings): Use neoversev1_addrcost_table.
20686 (aarch64_address_cost): Use the new post_modify costs for CImode
20689 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20691 * config/aarch64/aarch64.opt
20692 (-param=aarch64-loop-vect-issue-rate-niters=): New parameter.
20693 * doc/invoke.texi: Document it.
20694 * config/aarch64/aarch64-protos.h (aarch64_base_vec_issue_info)
20695 (aarch64_scalar_vec_issue_info, aarch64_simd_vec_issue_info)
20696 (aarch64_advsimd_vec_issue_info, aarch64_sve_vec_issue_info)
20697 (aarch64_vec_issue_info): New structures.
20698 (cpu_vector_cost): Write comments above the variables rather
20700 (cpu_vector_cost::issue_info): New member variable.
20701 * config/aarch64/aarch64.c: Include gimple-pretty-print.h
20702 and tree-ssa-loop-niter.h.
20703 (generic_vector_cost, a64fx_vector_cost, qdf24xx_vector_cost)
20704 (thunderx_vector_cost, tsv110_vector_cost, cortexa57_vector_cost)
20705 (exynosm1_vector_cost, xgene1_vector_cost, thunderx2t99_vector_cost)
20706 (thunderx3t110_vector_cost): Initialize issue_info to null.
20707 (neoversev1_scalar_issue_info, neoversev1_advsimd_issue_info)
20708 (neoversev1_sve_issue_info, neoversev1_vec_issue_info): New structures.
20709 (neoversev1_vector_cost): Use them.
20710 (aarch64_vec_op_count, aarch64_sve_op_count): New structures.
20711 (aarch64_vector_costs::saw_sve_only_op): New member variable.
20712 (aarch64_vector_costs::num_vector_iterations): Likewise.
20713 (aarch64_vector_costs::scalar_ops): Likewise.
20714 (aarch64_vector_costs::advsimd_ops): Likewise.
20715 (aarch64_vector_costs::sve_ops): Likewise.
20716 (aarch64_vector_costs::seen_loads): Likewise.
20717 (aarch64_simd_vec_costs_for_flags): New function.
20718 (aarch64_analyze_loop_vinfo): Initialize num_vector_iterations.
20719 Count the number of predicate operations required by SVE WHILE
20721 (aarch64_comparison_type, aarch64_multiply_add_p): New functions.
20722 (aarch64_sve_only_stmt_p, aarch64_in_loop_reduction_latency): Likewise.
20723 (aarch64_count_ops): Likewise.
20724 (aarch64_add_stmt_cost): Record whether see an SVE operation
20725 that cannot currently be implementing using Advanced SIMD.
20726 Record issue information about the scalar, Advanced SIMD
20727 and (where relevant) SVE versions of a loop.
20728 (aarch64_vec_op_count::dump): New function.
20729 (aarch64_sve_op_count::dump): Likewise.
20730 (aarch64_estimate_min_cycles_per_iter): Likewise.
20731 (aarch64_adjust_body_cost): If issue information is available,
20732 try to compare the issue rates of the various loop implementations
20733 and increase or decrease the vector body cost accordingly.
20735 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20737 * config/aarch64/aarch64.c (aarch64_detect_vector_stmt_subtype):
20738 Assume a zero cost for induction phis.
20740 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20742 * config/aarch64/aarch64.c (aarch64_embedded_comparison_type): New
20744 (aarch64_adjust_stmt_cost): Add the costs of embedded scalar and
20745 vector comparisons.
20747 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20749 * config/aarch64/aarch64.c (aarch64_detect_scalar_stmt_subtype):
20751 (aarch64_add_stmt_cost): Call it.
20753 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20755 * config/aarch64/aarch64-tuning-flags.def (matched_vector_throughput):
20756 New tuning parameter.
20757 * config/aarch64/aarch64.c (neoversev1_tunings): Use it.
20758 (aarch64_estimated_sve_vq): New function.
20759 (aarch64_vector_costs::analyzed_vinfo): New member variable.
20760 (aarch64_vector_costs::is_loop): Likewise.
20761 (aarch64_vector_costs::unrolled_advsimd_niters): Likewise.
20762 (aarch64_vector_costs::unrolled_advsimd_stmts): Likewise.
20763 (aarch64_record_potential_advsimd_unrolling): New function.
20764 (aarch64_analyze_loop_vinfo, aarch64_analyze_bb_vinfo): Likewise.
20765 (aarch64_add_stmt_cost): Call aarch64_analyze_loop_vinfo or
20766 aarch64_analyze_bb_vinfo on the first use of a costs structure.
20767 Detect whether we're vectorizing a loop for SVE that might be
20768 completely unrolled if it used Advanced SIMD instead.
20769 (aarch64_adjust_body_cost_for_latency): New function.
20770 (aarch64_finish_cost): Call it.
20772 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20774 * config/aarch64/aarch64.c (aarch64_vector_costs): New structure.
20775 (aarch64_init_cost): New function.
20776 (aarch64_add_stmt_cost): Use aarch64_vector_costs instead of
20777 the default unsigned[3].
20778 (aarch64_finish_cost, aarch64_destroy_cost_data): New functions.
20779 (TARGET_VECTORIZE_INIT_COST): Override.
20780 (TARGET_VECTORIZE_FINISH_COST): Likewise.
20781 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
20783 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20785 * config/aarch64/aarch64.c (neoversev1_advsimd_vector_cost)
20786 (neoversev1_sve_vector_cost): New cost structures.
20787 (neoversev1_vector_cost): Likewise.
20788 (neoversev1_tunings): Use them. Enable use_new_vector_costs.
20790 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20792 * config/aarch64/aarch64-protos.h
20793 (sve_vec_cost::scatter_store_elt_cost): New member variable.
20794 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
20795 accordingly, taking the cost from the cost of a scalar_store.
20796 (a64fx_sve_vector_cost): Likewise.
20797 (aarch64_detect_vector_stmt_subtype): Detect scatter stores.
20799 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20801 * config/aarch64/aarch64-protos.h
20802 (simd_vec_cost::store_elt_extra_cost): New member variable.
20803 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
20804 accordingly, using the vec_to_scalar cost for the new field.
20805 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
20806 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
20807 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
20808 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
20809 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
20810 (thunderx3t110_advsimd_vector_cost): Likewise.
20811 (aarch64_detect_vector_stmt_subtype): Detect single-element stores.
20813 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20815 * config/aarch64/aarch64-protos.h (simd_vec_cost::ld2_st2_permute_cost)
20816 (simd_vec_cost::ld3_st3_permute_cost): New member variables.
20817 (simd_vec_cost::ld4_st4_permute_cost): Likewise.
20818 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
20819 accordingly, using zero for the new costs.
20820 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
20821 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
20822 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
20823 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
20824 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
20825 (thunderx3t110_advsimd_vector_cost): Likewise.
20826 (aarch64_ld234_st234_vectors): New function.
20827 (aarch64_adjust_stmt_cost): Likewise.
20828 (aarch64_add_stmt_cost): Call aarch64_adjust_stmt_cost if using
20829 the new vector costs.
20831 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20833 * config/aarch64/aarch64-protos.h (sve_vec_cost): Turn into a
20834 derived class of simd_vec_cost. Add information about CLAST[AB]
20835 and FADDA instructions.
20836 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
20837 accordingly, using the vec_to_scalar costs for the new fields.
20838 (a64fx_sve_vector_cost): Likewise.
20839 (aarch64_reduc_type): New function.
20840 (aarch64_sve_in_loop_reduction_latency): Likewise.
20841 (aarch64_detect_vector_stmt_subtype): Take a vinfo parameter.
20842 Use aarch64_sve_in_loop_reduction_latency to handle SVE reductions
20843 that occur in the loop body.
20844 (aarch64_add_stmt_cost): Update call accordingly.
20846 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20848 * config/aarch64/aarch64-tuning-flags.def (use_new_vector_costs):
20850 * config/aarch64/aarch64-protos.h (simd_vec_cost): Put comments
20851 above the fields rather than to the right.
20852 (simd_vec_cost::reduc_i8_cost): New member variable.
20853 (simd_vec_cost::reduc_i16_cost): Likewise.
20854 (simd_vec_cost::reduc_i32_cost): Likewise.
20855 (simd_vec_cost::reduc_i64_cost): Likewise.
20856 (simd_vec_cost::reduc_f16_cost): Likewise.
20857 (simd_vec_cost::reduc_f32_cost): Likewise.
20858 (simd_vec_cost::reduc_f64_cost): Likewise.
20859 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
20860 accordingly, using the vec_to_scalar_cost for the new fields.
20861 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
20862 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
20863 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
20864 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
20865 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
20866 (thunderx3t110_advsimd_vector_cost): Likewise.
20867 (aarch64_use_new_vector_costs_p): New function.
20868 (aarch64_simd_vec_costs): New function, split out from...
20869 (aarch64_builtin_vectorization_cost): ...here.
20870 (aarch64_is_reduction): New function.
20871 (aarch64_detect_vector_stmt_subtype): Likewise.
20872 (aarch64_add_stmt_cost): Call aarch64_detect_vector_stmt_subtype if
20873 using the new vector costs.
20875 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
20878 * tree-emutls.c (get_emutls_init_templ_addr): Mark initializer of weak
20879 TLS declarations as public.
20881 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
20883 * config/aarch64/aarch64-d.c (IN_TARGET_CODE): Define.
20884 * config/arm/arm-d.c (IN_TARGET_CODE): Likewise.
20885 * config/i386/i386-d.c (IN_TARGET_CODE): Likewise.
20886 * config/mips/mips-d.c (IN_TARGET_CODE): Likewise.
20887 * config/pa/pa-d.c (IN_TARGET_CODE): Likewise.
20888 * config/riscv/riscv-d.c (IN_TARGET_CODE): Likewise.
20889 * config/rs6000/rs6000-d.c (IN_TARGET_CODE): Likewise.
20890 * config/s390/s390-d.c (IN_TARGET_CODE): Likewise.
20891 * config/sparc/sparc-d.c (IN_TARGET_CODE): Likewise.
20893 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
20896 * config.gcc (*-*-cygwin*): Add winnt-d.o
20897 (*-*-mingw*): Likewise.
20898 * config/i386/cygwin.h (EXTRA_TARGET_D_OS_VERSIONS): New macro.
20899 * config/i386/mingw32.h (EXTRA_TARGET_D_OS_VERSIONS): Likewise.
20900 * config/i386/t-cygming: Add winnt-d.o.
20901 * config/i386/winnt-d.c: New file.
20903 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
20905 * config/freebsd-d.c: Include memmodel.h.
20907 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
20910 * config.gcc (*-*-openbsd*): Add openbsd-d.o.
20911 * config/t-openbsd: Add openbsd-d.o.
20912 * config/openbsd-d.c: New file.
20914 2021-03-25 Stam Markianos-Wright <stam.markianos-wright@arm.com>
20916 PR tree-optimization/96974
20917 * tree-vect-stmts.c (vect_get_vector_types_for_stmt): Replace assert
20918 with graceful exit.
20920 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
20923 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
20927 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
20928 always_inline in system headers.
20930 2021-03-25 Kewen Lin <linkw@linux.ibm.com>
20932 * tree-vect-loop.c (vect_model_reduction_cost): Init inside_cost.
20934 2021-03-25 Jakub Jelinek <jakub@redhat.com>
20937 * tree-core.h (enum operand_equal_flag): Add OEP_ADDRESS_OF_SAME_FIELD.
20938 * fold-const.c (operand_compare::operand_equal_p): Don't compare
20939 field offsets if OEP_ADDRESS_OF_SAME_FIELD.
20941 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
20945 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
20946 always_inline in system headers.
20948 2021-03-25 Richard Biener <rguenther@suse.de>
20950 PR tree-optimization/99746
20951 * tree-vect-slp-patterns.c (complex_pattern::build): Do not mark
20952 the scalar stmt as patterned. Instead set up required things
20955 2021-03-25 Xionghu Luo <luoxhu@linux.ibm.com>
20957 * config/rs6000/rs6000.c (power8_costs): Change l2 cache
20960 2021-03-24 Martin Liska <mliska@suse.cz>
20963 * common/config/i386/i386-common.c (ARRAY_SIZE): Fix off-by-one
20965 * config/i386/i386-options.c (ix86_option_override_internal):
20966 Add run-time assert.
20968 2021-03-24 Martin Jambor <mjambor@suse.cz>
20971 * ipa-cp.c (initialize_node_lattices): Mark as bottom all
20972 parameters with unknown type.
20973 (ipacp_value_safe_for_type): New function.
20974 (propagate_vals_across_arith_jfunc): Verify that the constant type
20975 can be used for a type of the formal parameter.
20976 (propagate_vals_across_ancestor): Likewise.
20977 (propagate_scalar_across_jump_function): Likewise. Pass the type
20978 also to propagate_vals_across_ancestor.
20980 2021-03-24 Christophe Lyon <christophe.lyon@linaro.org>
20983 * config/arm/mve.md (movmisalign<mode>_mve_store): Use Ux
20985 (movmisalign<mode>_mve_load): Likewise.
20987 2021-03-24 Jakub Jelinek <jakub@redhat.com>
20990 * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
20991 movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.
20993 2021-03-24 Alexandre Oliva <oliva@adacore.com>
20995 * doc/sourcebuild.texi (sysconf): New effective target.
20997 2021-03-24 Alexandre Oliva <oliva@adacore.com>
20999 * config/i386/predicates.md (reg_or_const_vec_operand): New.
21000 * config/i386/sse.md (ssse3_pshufbv8qi3): Add an expander for
21001 the now *-prefixed insn_and_split, turn the splitter const vec
21002 into an input for the insn, making it an ignored immediate for
21003 non-split cases, and loaded into the scratch register
21006 2021-03-23 Vladimir N. Makarov <vmakarov@redhat.com>
21009 * config/aarch64/constraints.md (Utq, UOb, UOh, UOw, UOd, UOty):
21010 Use define_relaxed_memory_constraint for them.
21012 2021-03-23 Iain Sandoe <iain@sandoe.co.uk>
21015 * config/host-darwin.c (darwin_gt_pch_use_address): Add a
21016 colon to the diagnostic message.
21018 2021-03-23 Ilya Leoshkevich <iii@linux.ibm.com>
21020 * fwprop.c (fwprop_propagation::fwprop_propagation): Look at
21022 (try_fwprop_subst_note): Use set_info instead of insn_info.
21023 (try_fwprop_subst_pattern): Likewise.
21024 (try_fwprop_subst_notes): Likewise.
21025 (try_fwprop_subst): Likewise.
21026 (forward_propagate_subreg): Likewise.
21027 (forward_propagate_and_simplify): Likewise.
21028 (forward_propagate_into): Likewise.
21029 * rtl-ssa/accesses.h (set_info::single_nondebug_use) New
21031 (set_info::single_nondebug_insn_use): Likewise.
21032 (set_info::single_phi_use): Likewise.
21033 * rtl-ssa/member-fns.inl (set_info::single_nondebug_use) New
21035 (set_info::single_nondebug_insn_use): Likewise.
21036 (set_info::single_phi_use): Likewise.
21038 2021-03-23 Christophe Lyon <christophe.lyon@linaro.org>
21040 * doc/sourcebuild.texi (arm_dsp_ok, arm_dsp): Document.
21042 2021-03-23 Jakub Jelinek <jakub@redhat.com>
21045 * config/aarch64/aarch64.c (aarch64_add_offset): Tell
21046 expand_mult to perform an unsigned rather than a signed
21049 2021-03-23 H.J. Lu <hjl.tools@gmail.com>
21052 * config/i386/cpuid.h (__cpuid): Add __volatile__.
21053 (__cpuid_count): Likewise.
21055 2021-03-23 Richard Biener <rguenther@suse.de>
21057 PR tree-optimization/99721
21058 * tree-vect-slp.c (vect_slp_analyze_node_operations):
21059 Make sure we can schedule the node.
21061 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
21063 * config/riscv/riscv.c (riscv_subword): Take endianness into
21064 account when calculating the byte offset.
21066 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
21068 * config/riscv/predicates.md (subreg_lowpart_operator): New predicate
21069 * config/riscv/riscv.md (*addsi3_extended2, *subsi3_extended2)
21070 (*negsi2_extended2, *mulsi3_extended2, *<optab>si3_mask)
21071 (*<optab>si3_mask_1, *<optab>di3_mask, *<optab>di3_mask_1)
21072 (*<optab>si3_extend_mask, *<optab>si3_extend_mask_1): Use
21073 new predicate "subreg_lowpart_operator"
21075 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
21077 * config/riscv/riscv.c (riscv_swap_instruction): New function
21078 to byteswap an SImode rtx containing an instruction.
21079 (riscv_trampoline_init): Byteswap the generated instructions
21082 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
21084 * common/config/riscv/riscv-common.c
21085 (TARGET_DEFAULT_TARGET_FLAGS): Set default endianness.
21086 * config.gcc (riscv32be-*, riscv64be-*): Set
21087 TARGET_BIG_ENDIAN_DEFAULT to 1.
21088 * config/riscv/elf.h (LINK_SPEC): Change -melf* value
21089 depending on default endianness.
21090 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
21091 * config/riscv/linux.h (LINK_SPEC): Likewise.
21092 * config/riscv/riscv.c (TARGET_DEFAULT_TARGET_FLAGS): Set
21093 default endianness.
21094 * config/riscv/riscv.h (DEFAULT_ENDIAN_SPEC): New macro.
21096 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
21098 * config/riscv/elf.h (LINK_SPEC): Pass linker endianness flag.
21099 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
21100 * config/riscv/linux.h (LINK_SPEC): Likewise.
21101 * config/riscv/riscv.h (ASM_SPEC): Pass -mbig-endian and
21103 (BYTES_BIG_ENDIAN): Handle big endian.
21104 (WORDS_BIG_ENDIAN): Define to BYTES_BIG_ENDIAN.
21105 * config/riscv/riscv.opt (-mbig-endian, -mlittle-endian): New
21107 * doc/invoke.texi (-mbig-endian, -mlittle-endian): Document.
21109 2021-03-23 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
21111 * regcprop.c (find_oldest_value_reg): Ask target whether
21112 different mode is fine for replacement register.
21114 2021-03-23 Aldy Hernandez <aldyh@redhat.com>
21116 PR tree-optimization/99296
21117 * value-range.cc (irange::irange_set_1bit_anti_range): New.
21118 (irange::irange_set_anti_range): Call irange_set_1bit_anti_range
21119 * value-range.h (irange::irange_set_1bit_anti_range): New.
21121 2021-03-22 Vladimir N. Makarov <vmakarov@redhat.com>
21124 * config/aarch64/constraints.md (UtQ): Use
21125 define_relaxed_memory_constraint for it.
21126 * doc/md.texi (define_relaxed_memory_constraint): Describe it.
21127 * genoutput.c (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
21128 * genpreds.c (constraint_data): Add bitfield is_relaxed_memory.
21129 (have_relaxed_memory_constraints): New static var.
21130 (relaxed_memory_start, relaxed_memory_end): Ditto.
21131 (add_constraint): Add arg is_relaxed_memory. Check name for
21132 relaxed memory. Set up is_relaxed_memory in constraint_data and
21133 have_relaxed_memory_constraints. Adjust calls.
21134 (choose_enum_order): Process relaxed memory.
21135 (write_tm_preds_h): Ditto.
21136 (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
21137 * gensupport.c (process_rtx): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
21138 * ira-costs.c (record_reg_classes): Process CT_RELAXED_MEMORY.
21139 * ira-lives.c (single_reg_class): Use
21140 insn_extra_relaxed_memory_constraint.
21141 * ira.c (ira_setup_alts): CT_RELAXED_MEMORY.
21142 * lra-constraints.c (valid_address_p): Use
21143 insn_extra_relaxed_memory_constraint instead of other memory
21145 (process_alt_operands): Process CT_RELAXED_MEMORY.
21146 (curr_insn_transform): Use insn_extra_relaxed_memory_constraint.
21147 * recog.c (asm_operand_ok, preprocess_constraints): Process
21149 * reload.c (find_reloads): Ditto.
21150 * rtl.def (DEFINE_RELAXED_MEMORY_CONSTRAINT): New.
21151 * stmt.c (parse_input_constraint): Use
21152 insn_extra_relaxed_memory_constraint.
21154 2021-03-22 Segher Boessenkool <segher@kernel.crashing.org>
21157 * ubsan.c (ubsan_instrument_float_cast): Don't test for unordered if
21160 2021-03-22 Alex Coplan <alex.coplan@arm.com>
21163 * config/arm/arm-protos.h (neon_make_constant): Add generate
21164 argument to guard emitting insns, default to true.
21165 * config/arm/arm.c (arm_legitimate_constant_p_1): Reject
21166 CONST_VECTORs which neon_make_constant can't handle.
21167 (neon_vdup_constant): Add generate argument, avoid emitting
21168 insns if it's not set.
21169 (neon_make_constant): Plumb new generate argument through.
21170 * config/arm/constraints.md (Ui): New. Use it...
21171 * config/arm/mve.md (*mve_mov<mode>): ... here.
21172 * config/arm/vec-common.md (movv8hf): Use neon_make_constant to
21173 synthesize constants.
21175 2021-03-22 Richard Biener <rguenther@suse.de>
21177 * debug.h: Add deprecation warning.
21179 2021-03-22 Richard Biener <rguenther@suse.de>
21181 PR tree-optimization/99694
21182 * tree-ssa-sccvn.c (visit_phi): Ignore edges with the
21185 2021-03-22 Kito Cheng <kito.cheng@sifive.com>
21188 * config/riscv/riscv.c (riscv_expand_block_move): Get RTL value
21189 after type checking.
21191 2021-03-22 Jakub Jelinek <jakub@redhat.com>
21195 * dwarf2out.c (get_full_len): Use get_precision rather than
21197 (add_const_value_attribute): Make sure add_AT_wide argument has
21198 precision prec rather than some very wide one.
21200 2021-03-22 Kewen Lin <linkw@linux.ibm.com>
21202 * config/rs6000/rs6000.md (*rotldi3_insert_sf,
21203 *mov<SFDF:mode><SFDF2:mode>cc_p9, floatsi<mode>2_lfiwax,
21204 floatsi<mode>2_lfiwax_mem, floatunssi<mode>2_lfiwzx,
21205 floatunssi<mode>2_lfiwzx_mem, *floatsidf2_internal,
21206 *floatunssidf2_internal, fix_trunc<mode>si2_stfiwx,
21207 fix_trunc<mode>si2_internal, fixuns_trunc<mode>si2_stfiwx,
21208 *round32<mode>2_fprs, *roundu32<mode>2_fprs,
21209 *fix_trunc<mode>si2_internal): Fix empty split condition.
21210 * config/rs6000/vsx.md (*vsx_le_undo_permute_<mode>,
21211 vsx_reduc_<VEC_reduc_name>_v2df, vsx_reduc_<VEC_reduc_name>_v4sf,
21212 *vsx_reduc_<VEC_reduc_name>_v2df_scalar,
21213 *vsx_reduc_<VEC_reduc_name>_v4sf_scalar): Likewise.
21215 2021-03-22 Xionghu Luo <luoxhu@linux.ibm.com>
21218 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
21219 Convert idx to DImode.
21220 (rs6000_expand_vector_set_var_p8): Likewise.
21222 2021-03-21 Jakub Jelinek <jakub@redhat.com>
21225 * dwarf2out.c (insert_float): Change return type from void to
21226 unsigned, handle GET_MODE_SIZE (mode) == 2 and return element size.
21227 (mem_loc_descriptor, loc_descriptor, add_const_value_attribute):
21230 2021-03-20 H.J. Lu <hjl.tools@gmail.com>
21233 * config/i386/i386.c (construct_container): Check cfun != NULL
21234 before accessing silent_p.
21236 2021-03-20 Ahamed Husni <ahamedhusni73@gmail.com>
21238 * asan.c: Fix typos in comments.
21240 2021-03-20 Vladimir N. Makarov <vmakarov@redhat.com>
21242 PR rtl-optimization/99680
21243 * lra-constraints.c (skip_contraint_modifiers): Rename to skip_constraint_modifiers.
21244 (process_address_1): Check empty constraint before using
21247 2021-03-19 Pat Haugen <pthaugen@linux.ibm.com>
21249 * config/rs6000/rs6000.c (power10_cost): New.
21250 (rs6000_option_override_internal): Set Power10 costs.
21251 (rs6000_issue_rate): Set Power10 issue rate.
21252 * config/rs6000/power10.md: Rewrite for Power10.
21254 2021-03-19 Vladimir N. Makarov <vmakarov@redhat.com>
21257 * lra-constraints.c (process_address_1): Don't use unknown
21258 constraint for address constraint.
21260 2021-03-19 Iain Sandoe <iain@sandoe.co.uk>
21263 * config.gcc (powerpc-*-darwin8): Delete the reference to
21264 the now removed darwin8.h.
21266 2021-03-19 Olivier Hainque <hainque@adacore.com>
21269 * config/vxworksae.h (VX_CPU_PREFIX): Define.
21271 2021-03-19 John David Anglin <danglin@gcc.gnu.org>
21273 * config/pa/pa.c (import_milli): Use memcpy instead of strncpy.
21275 2021-03-19 Tamar Christina <tamar.christina@arm.com>
21277 PR tree-optimization/99656
21278 * tree-vect-slp-patterns.c (linear_loads_p,
21279 complex_add_pattern::matches, is_eq_or_top,
21280 vect_validate_multiplication, complex_mul_pattern::matches,
21281 complex_fms_pattern::matches): Remove complex_perm_kinds_t.
21282 * tree-vectorizer.h: (complex_load_perm_t): Removed.
21283 (slp_tree_to_load_perm_map_t): Use complex_perm_kinds_t instead of
21284 complex_load_perm_t.
21286 2021-03-19 H.J. Lu <hjl.tools@gmail.com>
21289 * config/i386/i386-options.c (ix86_init_machine_status): Set
21291 * config/i386/i386.c (init_cumulative_args): Set silent_p to
21293 (construct_container): Return early for return and argument
21294 errors if silent_p is true.
21295 * config/i386/i386.h (machine_function): Add silent_p.
21297 2021-03-19 Jakub Jelinek <jakub@redhat.com>
21300 * config/arm/constraints.md (Ds): New constraint.
21301 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): Use w,Ds
21302 constraint instead of w,Dm.
21304 2021-03-19 Andrew Stubbs <ams@codesourcery.com>
21306 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Fix quotes
21309 2021-03-19 Eric Botcazou <ebotcazou@adacore.com>
21311 PR middle-end/99641
21312 * fold-const.c (native_encode_initializer) <CONSTRUCTOR>: For an
21313 array type, do the computation of the current position in sizetype.
21315 2021-03-18 Vladimir N. Makarov <vmakarov@redhat.com>
21318 * lra-constraints.c (process_address_1): Use lookup_constraint
21319 only for a single constraint.
21321 2021-03-18 Martin Sebor <msebor@redhat.com>
21323 PR middle-end/99502
21324 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): Rename...
21325 (inbounds_memaccess_p): ...to this. Check the ending offset of
21326 the accessed member.
21328 2021-03-18 Andrew Stubbs <ams@codesourcery.com>
21330 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Add %< and
21331 %> quote markers to error messages.
21332 (gcn_goacc_validate_dims): Likewise.
21333 (gcn_conditional_register_usage): Remove exclaimation mark from error
21335 (gcn_vectorize_vec_perm_const): Ensure perm is fully uninitialized.
21337 2021-03-18 Jan Hubicka <hubicka@ucw.cz>
21339 * config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
21342 2021-03-18 Sinan Lin <sinan@isrc.iscas.ac.cn>
21343 Kito Cheng <kito.cheng@sifive.com>
21345 * config/riscv/riscv.c (riscv_block_move_straight): Change type
21346 to unsigned HOST_WIDE_INT for parameter and local variable with
21347 HOST_WIDE_INT type.
21348 (riscv_adjust_block_mem): Ditto.
21349 (riscv_block_move_loop): Ditto.
21350 (riscv_expand_block_move): Ditto.
21352 2021-03-18 Nick Clifton <nickc@redhat.com>
21354 * config/v850/v850.c (construct_restore_jr): Increase static
21356 (construct_save_jarl): Likewise.
21357 * config/v850/v850.h (DWARF2_DEBUGGING_INFO): Define.
21359 2021-03-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21361 * config/aarch64/aarch64.c (aarch64_adjust_generic_arch_tuning): Define.
21362 (aarch64_override_options_internal): Use it.
21363 (generic_tunings): Add AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS to
21366 2021-03-17 Sandra Loosemore <sandra@codesourcery.com>
21368 * config/nios2/nios2.c (nios2_custom_check_insns): Clean up
21369 error message format issues.
21370 (nios2_option_override): Likewise.
21371 (nios2_expand_fpu_builtin): Likewise.
21372 (nios2_init_custom_builtins): Adjust to avoid bogus strncpy
21373 truncation warning.
21374 (nios2_expand_custom_builtin): More error message format fixes.
21375 (nios2_expand_rdwrctl_builtin): Likewise.
21376 (nios2_expand_rdprs_builtin): Likewise.
21377 (nios2_expand_eni_builtin): Likewise.
21378 (nios2_expand_builtin): Likewise.
21379 (nios2_register_custom_code): Likewise.
21380 (nios2_valid_target_attribute_rec): Likewise.
21381 (nios2_add_insn_asm): Fix uninitialized variable warning.
21383 2021-03-17 Jan Hubicka <jh@suse.cz>
21385 * config/i386/x86-tune-costs.h (struct processor_costs): Update costs
21386 of gather to match reality.
21387 * config/i386/x86-tune.def (X86_TUNE_USE_GATHER): Enable for znver3.
21389 2021-03-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21391 * config/aarch64/aarch64-builtins.c (aarch64_expand_rng_builtin): Use EQ
21392 to compare against CC_REG rather than NE.
21394 2021-03-17 H.J. Lu <hjl.tools@gmail.com>
21397 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support
21398 inline assembly statements.
21399 (ix86_print_operand): Update 'P' handling for -fno-plt.
21401 2021-03-17 Tamar Christina <tamar.christina@arm.com>
21404 * config/aarch64/aarch64.c
21405 (aarch64_simd_clone_compute_vecsize_and_simdlen): Remove unused var.
21407 2021-03-16 Segher Boessenkool <segher@kernel.crashing.org>
21410 * config/rs6000/predicates.md (branch_comparison_operator): Allow
21411 ordered and unordered for CCFPmode, if flag_finite_math_only.
21413 2021-03-16 Jakub Jelinek <jakub@redhat.com>
21416 * config/i386/i386-expand.c (ix86_split_lea_for_addr): Emit a MULT
21417 rather than ASHIFT.
21418 * config/i386/i386.md (mult by 1248 into ashift): New splitter.
21420 2021-03-16 Martin Liska <mliska@suse.cz>
21423 * optc-save-gen.awk: Add flag_ipa_ra to exceptions for
21424 cl_optimization_compare function.
21426 2021-03-16 Ilya Leoshkevich <iii@linux.ibm.com>
21428 * config/s390/s390.c (f_constraint_p): Treat "fv" constraints
21431 2021-03-16 Jakub Jelinek <jakub@redhat.com>
21434 * config/i386/i386.h (struct machine_function): Add
21435 has_explicit_vzeroupper bitfield.
21436 * config/i386/i386-expand.c (ix86_expand_builtin): Set
21437 cfun->machine->has_explicit_vzeroupper when expanding
21438 IX86_BUILTIN_VZEROUPPER.
21439 * config/i386/i386-features.c (rest_of_handle_insert_vzeroupper):
21440 Do the mode switching only when TARGET_VZEROUPPER, expensive
21441 optimizations turned on and not optimizing for size.
21442 (pass_insert_vzeroupper::gate): Enable even when
21443 cfun->machine->has_explicit_vzeroupper is set.
21445 2021-03-16 Jakub Jelinek <jakub@redhat.com>
21448 * config/aarch64/aarch64.c
21449 (aarch64_simd_clone_compute_vecsize_and_simdlen): If not a function
21450 definition, walk TYPE_ARG_TYPES list if non-NULL for argument types
21451 instead of DECL_ARGUMENTS. Ignore types for uniform arguments.
21453 2021-03-15 Richard Biener <rguenther@suse.de>
21455 PR tree-optimization/98834
21456 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle missing
21457 subsetting by truncating the access size.
21459 2021-03-15 Jan Hubicka <hubicka@ucw.cz>
21461 * config/i386/i386-options.c (processor_cost_table): Add znver3_cost.
21462 * config/i386/x86-tune-costs.h (znver3_cost): New gobal variable; copy
21465 2021-03-15 Martin Liska <mliska@suse.cz>
21467 * spellcheck.c: Add missing comma in initialization.
21469 2021-03-14 Uroš Bizjak <ubizjak@gmail.com>
21471 * config/i386/sse.md (*vec_extract<mode>): Merge alternative 0 with
21472 alternative 2 and alternative 1 with alternative 3 using
21473 YW register constraint.
21474 (*vec_extract<PEXTR_MODE12:mode>_zext): Merge alternatives
21475 using YW register constraint.
21476 (*vec_extractv16qi_zext): Ditto.
21477 (*vec_extractv4si): Merge alternatives 4 and 5
21478 using Yw register constraint.
21479 (*ssse3_palignr<mode>_perm): Use Yw instead of v for alternative 3.
21481 2021-03-13 Martin Sebor <msebor@redhat.com>
21483 PR tree-optimization/99489
21484 * builtins.c (gimple_call_alloc_size): Fail gracefully when argument
21485 is not a call statement.
21487 2021-03-13 Jakub Jelinek <jakub@redhat.com>
21489 PR tree-optimization/99544
21490 * match.pd (X + (X << C) -> X * (1 + (1 << C))): Don't simplify
21491 if for vector types multiplication can't be done in type's mode.
21493 2021-03-12 Eric Botcazou <ebotcazou@adacore.com>
21496 * config/sparc/constraints.md (w): Rename to...
21497 (W): ... this and ditch previous implementation.
21498 * config/sparc/sparc.md (*movdi_insn_sp64): Replace W with m.
21499 (*movdf_insn_sp64): Likewise.
21500 (*mov<VM64:mode>_insn_sp64): Likewise.
21501 * config/sparc/sync.md (*atomic_compare_and_swap<mode>_1): Replace
21503 (atomic_compare_and_swap_leon3_1): Likewise.
21504 (*atomic_compare_and_swapdi_v8plus): Likewise.
21505 * config/sparc/sparc.c (memory_ok_for_ldd): Remove useless test on
21506 architecture and add missing address validity check during LRA.
21508 2021-03-12 Tobias Burnus <tobias@codesourcery.com>
21511 * gimplify.c (omp_add_variable): Handle NULL_TREE as size
21512 occuring for assumed-size arrays in use_device_{ptr,addr}.
21514 2021-03-12 Jakub Jelinek <jakub@redhat.com>
21517 * config/i386/constraints.md (YW): New internal constraint.
21518 * config/i386/sse.md (v_Yw): Add V4TI, V2TI, V1TI and TI cases.
21519 (*<sse2_avx2>_<insn><mode>3<mask_name>,
21520 *<sse2_avx2>_uavg<mode>3<mask_name>, *abs<mode>2,
21521 *<s>mul<mode>3_highpart<mask_name>): Use <v_Yw> instead of v in
21523 (<sse2_avx2>_psadbw): Use YW instead of v in constraints.
21524 (*avx2_pmaddwd, *sse2_pmaddwd, *<code>v8hi3, *<code>v16qi3,
21525 avx2_pmaddubsw256, ssse3_pmaddubsw128): Merge last two alternatives
21526 into one, use Yw instead of former x,v.
21527 (ashr<mode>3, <insn><mode>3): Use <v_Yw> instead of x in constraints of
21528 the last alternative.
21529 (<sse2_avx2>_packsswb<mask_name>, <sse2_avx2>_packssdw<mask_name>,
21530 <sse2_avx2>_packuswb<mask_name>, <sse4_1_avx2>_packusdw<mask_name>,
21531 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, <ssse3_avx2>_palignr<mode>,
21532 <ssse3_avx2>_pshufb<mode>3<mask_name>): Merge last two alternatives
21533 into one, use <v_Yw> instead of former x,v.
21534 (avx2_interleave_highv32qi<mask_name>,
21535 vec_interleave_highv16qi<mask_name>): Use Yw instead of v in
21536 constraints. Add && <mask_avx512bw_condition> to condition.
21537 (avx2_interleave_lowv32qi<mask_name>,
21538 vec_interleave_lowv16qi<mask_name>,
21539 avx2_interleave_highv16hi<mask_name>,
21540 vec_interleave_highv8hi<mask_name>,
21541 avx2_interleave_lowv16hi<mask_name>, vec_interleave_lowv8hi<mask_name>,
21542 avx2_pshuflw_1<mask_name>, sse2_pshuflw_1<mask_name>,
21543 avx2_pshufhw_1<mask_name>, sse2_pshufhw_1<mask_name>,
21544 avx2_<code>v16qiv16hi2<mask_name>, sse4_1_<code>v8qiv8hi2<mask_name>,
21545 *sse4_1_<code>v8qiv8hi2<mask_name>_1, <sse2_avx2>_<insn><mode>3): Use
21546 Yw instead of v in constraints.
21547 * config/i386/mmx.md (Yv_Yw): New define_mode_attr.
21548 (*mmx_<insn><mode>3, mmx_ashr<mode>3, mmx_<insn><mode>3): Use <Yv_Yw>
21549 instead of Yv in constraints.
21550 (*mmx_<insn><mode>3, *mmx_mulv4hi3, *mmx_smulv4hi3_highpart,
21551 *mmx_umulv4hi3_highpart, *mmx_pmaddwd, *mmx_<code>v4hi3,
21552 *mmx_<code>v8qi3, mmx_pack<s_trunsuffix>swb, mmx_packssdw,
21553 mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
21554 *mmx_uavgv8qi3, *mmx_uavgv4hi3, mmx_psadbw): Use Yw instead of Yv in
21556 (*mmx_pinsrw, *mmx_pinsrb, *mmx_pextrw, *mmx_pextrw_zext, *mmx_pextrb,
21557 *mmx_pextrb_zext): Use YW instead of Yv in constraints.
21558 (*mmx_eq<mode>3, mmx_gt<mode>3): Use x instead of Yv in constraints.
21559 (mmx_andnot<mode>3, *mmx_<code><mode>3): Split last alternative into
21560 two, one with just x, another isa avx512vl with v.
21562 2021-03-12 Martin Liska <mliska@suse.cz>
21564 * doc/invoke.texi: Add missing param documentation.
21566 2021-03-11 David Malcolm <dmalcolm@redhat.com>
21569 * Makefile.in (ANALYZER_OBJS): Add analyzer/feasible-graph.o and
21570 analyzer/trimmed-graph.o.
21571 * doc/analyzer.texi (Analyzer Paths): Rewrite description of
21572 feasibility checking to reflect new implementation.
21573 * doc/invoke.texi (-fdump-analyzer-feasibility): Document new
21575 * shortest-paths.h (shortest_paths::get_shortest_distance): New.
21577 2021-03-11 David Malcolm <dmalcolm@redhat.com>
21579 * digraph.cc (selftest::test_shortest_paths): Update
21580 shortest_paths init for new param. Add test of
21581 SPS_TO_GIVEN_TARGET.
21582 * shortest-paths.h (enum shortest_path_sense): New.
21583 (shortest_paths::shortest_paths): Add "sense" param.
21584 Update for renamings. Generalize to use "sense" param.
21585 (shortest_paths::get_shortest_path): Rename param.
21586 (shortest_paths::m_sense): New field.
21587 (shortest_paths::m_prev): Rename...
21588 (shortest_paths::m_best_edge): ...to this.
21589 (shortest_paths::get_shortest_path): Update for renamings.
21590 Conditionalize flipping of path on sense of traversal.
21592 2021-03-11 David Malcolm <dmalcolm@redhat.com>
21594 * digraph.cc (selftest::test_shortest_paths): Add test coverage
21595 for paths from B and C.
21596 * shortest-paths.h (shortest_paths::shortest_paths): Handle
21597 unreachable nodes, rather than asserting.
21599 2021-03-11 David Edelsohn <dje.gcc@gmail.com>
21602 * config/rs6000/rs6000.c (rs6000_xcoff_file_start): Don't create
21603 xcoff_tbss_section_name.
21604 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .lcomm.
21605 * xcoffout.c (xcoff_tbss_section_name): Delete.
21606 * xcoffout.h (xcoff_tbss_section_name): Delete.
21608 2021-03-11 Richard Biener <rguenther@suse.de>
21610 PR tree-optimization/99523
21611 * tree-cfg.c (dump_function_to_file): Dump SSA names
21612 w/o identifier to the decls section as well, not only those
21613 without a VAR_DECL.
21615 2021-03-11 Jakub Jelinek <jakub@redhat.com>
21618 * ipa-icf-gimple.c (func_checker::compare_gimple_call): For internal
21619 function calls with lhs fail if the lhs don't have compatible types.
21621 2021-03-11 Hans-Peter Nilsson <hp@axis.com>
21623 * config/cris/cris.h (HARD_FRAME_POINTER_REGNUM): Define.
21624 Change FRAME_POINTER_REGNUM to correspond to a new faked
21625 register faked_fp, part of GENNONACR_REGS like faked_ap.
21626 (CRIS_FAKED_REGS_CONTENTS): New helper macro.
21627 (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS, CALL_USED_REGISTERS):
21628 (REG_ALLOC_ORDER, REG_CLASS_CONTENTS, REGNO_OK_FOR_BASE_P)
21629 (ELIMINABLE_REGS, REGISTER_NAMES): Adjust accordingly.
21630 * config/cris/cris.md (CRIS_FP_REGNUM): Renumber to new faked
21632 (CRIS_REAL_FP_REGNUM): New constant.
21633 * config/cris/cris.c (cris_reg_saved_in_regsave_area): Check
21634 for HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM.
21635 (cris_initial_elimination_offset): Handle elimination changes
21636 to HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM
21637 and add one from FRAME_POINTER_REGNUM to
21638 HARD_FRAME_POINTER_REGNUM.
21639 (cris_expand_prologue, cris_expand_epilogue): Emit code for
21640 hard_frame_pointer_rtx instead of frame_pointer_rtx.
21642 2021-03-10 David Edelsohn <dje.gcc@gmail.com>
21645 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Add check for DCmode.
21646 * config/rs6000/rs6000.c (rs6000_special_round_type_align): Same.
21648 2021-03-10 Vladimir N. Makarov <vmakarov@redhat.com>
21651 * lra-constraints.c (process_address_1): Don't check unknown
21652 constraint, use X for empty constraint.
21654 2021-03-10 Alex Coplan <alex.coplan@arm.com>
21656 * config/aarch64/aarch64.c (aarch64_vfp_is_call_or_return_candidate):
21657 Fix typo in comment describing "is_ha" argument.
21659 2021-03-10 John David Anglin <danglin@gcc.gnu.org>
21661 * doc/sourcebuild.texi: Document LRA target selector.
21663 2021-03-10 David Malcolm <dmalcolm@redhat.com>
21665 * doc/ux.texi: Add subsection contrasting interactive versus
21666 batch usage of GCC.
21668 2021-03-10 Joel Hutton <joel.hutton@arm.com>
21671 * tree-vect-stmts.c (vectorizable_store): Fix scatter store mask
21673 (vectorizable_load): Fix gather load mask check condition.
21675 2021-03-10 Richard Biener <rguenther@suse.de>
21677 PR tree-optimization/99510
21678 * tree.c (check_aligned_type): Check that the candidate
21679 has TYPE_USER_ALIGN set instead of matching with the
21682 2021-03-10 Eric Botcazou <ebotcazou@adacore.com>
21684 * config/sparc/sparc.c (sparc_regmode_natural_size): Return 4 for
21685 float and vector integer modes only if the mode is not larger.
21687 2021-03-10 Hans-Peter Nilsson <hp@axis.com>
21689 * config/cris/cris.h (DWARF_FRAME_REGISTERS): Define.
21691 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
21693 * ira.c (ira_setup_alts, ira_get_dup_out_num): Process digital
21695 * ira-lives.c (single_reg_class): Ditto.
21697 2021-03-09 Sebastian Huber <sebastian.huber@embedded-brains.de>
21699 * config.gcc (aarch64-*-rtems*): Include general rtems.h after
21700 the architecture-specific rtems.h.
21701 (aarch64-*-rtems*): Likewise.
21702 (arm*-*-rtems*): Likewise.
21703 (epiphany-*-rtems*): Likewise.
21704 (riscv*-*-rtems*): Likewise.
21706 2021-03-09 Jakub Jelinek <jakub@redhat.com>
21708 PR tree-optimization/99305
21709 * tree-ssa-phiopt.c (conditional_replacement): Test integer_pow2p
21710 before integer_all_onesp instead of vice versa.
21712 2021-03-09 Richard Earnshaw <rearnsha@arm.com>
21714 * common/config/arm/arm-common.c (arm_config_default): Change type
21715 of 'i' to unsigned.
21717 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
21720 * lra-constraints.c (process_address_1): Process constraint 'g'
21721 separately and digital constraints containing more one digit.
21723 2021-03-09 Nick Clifton <nickc@redhat.com>
21725 * config/rx/rx.h (DBX_DEBUGGING_INFO): Define.
21726 (DWARF"_DEBUGGING_INFO): Define.
21728 2021-03-09 Eric Botcazou <ebotcazou@adacore.com>
21731 * calls.c (initialize_argument_information): When the argument
21732 is passed by reference, do not make a copy in a thunk only if
21733 the argument is already in memory. Remove redundant test for
21734 the case of callee copy.
21736 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
21739 * lra-constraints.c (process_address_1): Process 0..9 constraints
21740 in process_address_1.
21742 2021-03-09 Andreas Krebbel <krebbel@linux.ibm.com>
21744 * config/s390/s390.c (struct s390_processor processor_table):
21745 Binutils name string must not be empty.
21747 2021-03-09 Claudiu Zissulescu <claziss@synopsys.com>
21749 * config/arc/arc.c (arc_attr_type): Remove function.
21751 2021-03-09 Martin Liska <mliska@suse.cz>
21754 * config/i386/i386-options.c (ix86_option_override_internal):
21755 Set isa_flags for OPTS argument and not for the global
21758 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
21760 * config/rs6000/predicates.md (ds_form_mem_operand): Check
21763 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
21766 * config/rs6000/predicates.md (ds_form_mem_operand) New
21768 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10) Use
21769 ds_form_mem_operand in ld/lwa patterns.
21770 * config/rs6000/fusion.md: Regenerate file.
21772 2021-03-08 Martin Sebor <msebor@redhat.com>
21774 PR middle-end/98266
21775 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): New function.
21776 (array_bounds_checker::check_array_bounds): Call it.
21778 2021-03-08 Martin Sebor <msebor@redhat.com>
21780 PR middle-end/97631
21781 * tree-ssa-strlen.c (maybe_warn_overflow): Test rawmem.
21782 (handle_builtin_stxncpy_strncat): Rename locals. Determine
21783 destination size from allocation calls. Issue a more appropriate
21785 (handle_builtin_memcpy): Pass true as rawmem to maybe_warn_overflow.
21786 (handle_builtin_memset): Same.
21788 2021-03-08 Peter Bergner <bergner@linux.ibm.com>
21791 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_permute): Add an assert
21792 to ensure we do not have an Altivec style address.
21793 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): Disable if passed
21794 an Altivec style address.
21795 (*vsx_le_perm_store_<mode>): Likewise.
21796 (splitters after *vsx_le_perm_store_<mode>): Likewise.
21797 (vsx_load_<mode>): Disable special expander if passed an Altivec
21799 (vsx_store_<mode>): Likewise.
21801 2021-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21804 * config/aarch64/predicates.md (aarch64_simd_shift_imm_vec_qi): Define.
21805 (aarch64_simd_shift_imm_vec_hi): Likewise.
21806 (aarch64_simd_shift_imm_vec_si): Likewise.
21807 (aarch64_simd_shift_imm_vec_di): Likewise.
21808 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le): Use
21809 predicate from above.
21810 (aarch64_shrn<mode>_insn_be): Likewise.
21811 (aarch64_rshrn<mode>_insn_le): Likewise.
21812 (aarch64_rshrn<mode>_insn_be): Likewise.
21813 (aarch64_shrn2<mode>_insn_le): Likewise.
21814 (aarch64_shrn2<mode>_insn_be): Likewise.
21815 (aarch64_rshrn2<mode>_insn_le): Likewise.
21816 (aarch64_rshrn2<mode>_insn_be): Likewise.
21818 2021-03-08 Vladimir N. Makarov <vmakarov@redhat.com>
21821 * lra-constraints.c (skip_contraint_modifiers): New function.
21822 (process_address_1): Use it before lookup_constraint call.
21824 2021-03-08 Martin Liska <mliska@suse.cz>
21827 * config/i386/i386-options.c (ix86_option_override_internal):
21828 Enable UINTR and HRESET for -march that supports it.
21830 2021-03-08 Ilya Leoshkevich <iii@linux.ibm.com>
21832 * config/s390/s390.c (f_constraint_p): New function.
21833 (s390_md_asm_adjust): Implement TARGET_MD_ASM_ADJUST.
21834 (TARGET_MD_ASM_ADJUST): Likewise.
21836 2021-03-08 Tobias Burnus <tobias@codesourcery.com>
21839 * tree-nested.c (convert_local_reference_stmt): Avoid calling
21840 lookup_field_for_decl for Fortran module (= namespace context).
21842 2021-03-08 Andreas Krebbel <krebbel@linux.ibm.com>
21844 * config/s390/s390.c (s390_expand_vec_compare): Implement <0
21845 comparison with arithmetic right shift.
21846 (s390_expand_vcond): No need for a force_reg anymore.
21847 s390_vec_compare will do it.
21848 * config/s390/vector.md ("vec_cmp<mode><tointvec>"): Accept also
21849 immediate operands.
21851 2021-03-07 Jakub Jelinek <jakub@redhat.com>
21854 * config/i386/constraints.md (Yw): Use SSE_REGS if TARGET_SSE
21855 but TARGET_AVX512BW or TARGET_AVX512VL is not set. Adjust description
21857 * config/i386/sse.md (v_Yw): New define_mode_attr.
21858 (*<insn><mode>3, *mul<mode>3<mask_name>, *avx2_<code><mode>3,
21859 *sse4_1_<code><mode>3<mask_name>): Use <v_Yw> instead of v
21861 * config/i386/mmx.md (mmx_pshufw_1, *vec_dupv4hi): Use Yw instead of
21862 xYw in constraints.
21864 2021-03-06 Julian Brown <julian@codesourcery.com>
21866 * tree-pretty-print.c (dump_generic_node): Emit non-generic
21867 address space info for aggregates.
21869 2021-03-06 Hans-Peter Nilsson <hp@axis.com>
21871 * config/cris/cris.h (MAX_FIXED_MODE_SIZE): Don't define.
21873 2021-03-05 Jakub Jelinek <jakub@redhat.com>
21875 PR middle-end/99322
21876 * tree-cfg.c (bb_to_omp_idx): New variable.
21877 (execute_build_cfg): Release the bb_to_omp_idx vector after
21878 cleanup_tree_cfg returns.
21879 (handle_abnormal_edges): Remove bb_to_omp_idx argument, adjust
21880 for bb_to_omp_idx being a vec<int> instead of pointer to array
21882 (make_edges): Remove bb_to_omp_idx local variable, don't pass
21883 it to handle_abnormal_edges, adjust for bb_to_omp_idx being a
21884 vec<int> instead of pointer to array of ints and don't free/release
21886 (remove_bb): When removing a bb and placing forced label somewhere
21887 else, ensure it is put into the same OpenMP region during cfg
21888 pass if possible or to entry successor as fallback. Unregister
21889 bb from bb_to_omp_idx.
21891 2021-03-05 Vladimir N. Makarov <vmakarov@redhat.com>
21894 * lra-constraints.c (process_address_1): Skip decomposing address
21895 for asm insn operand with unknown constraint.
21897 2021-03-05 Martin Jambor <mjambor@suse.cz>
21900 * cgraph.c (cgraph_edge::set_call_stmt): Do not update all
21901 corresponding speculative edges if we are about to resolve
21902 sepculation. Make edge direct (and so resolve speculations) before
21903 removing it from call_site_hash.
21904 (cgraph_edge::make_direct): Relax the initial assert to allow calling
21905 the function on speculative direct edges.
21907 2021-03-05 Eric Botcazou <ebotcazou@adacore.com>
21909 PR rtl-optimization/99376
21910 * rtlanal.c (nonzero_bits1) <arithmetic operators>: If the number
21911 of low-order zero bits is too large, set the result to 0 directly.
21913 2021-03-04 Jakub Jelinek <jakub@redhat.com>
21915 PR middle-end/93235
21916 * expmed.c (store_bit_field_using_insv): Return false of xop0 is a
21917 SUBREG and a SUBREG to op_mode can't be created.
21919 2021-03-04 Alex Coplan <alex.coplan@arm.com>
21922 * config/aarch64/aarch64-sve-builtins.cc
21923 (function_resolver::require_vector_type): Handle error_mark_node.
21925 2021-03-04 Ilya Leoshkevich <iii@linux.ibm.com>
21927 * cfgexpand.c (expand_asm_loc): Pass new parameter.
21928 (expand_asm_stmt): Likewise.
21929 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Add new
21931 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
21932 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
21933 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
21934 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
21935 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
21936 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
21937 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
21938 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
21939 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
21940 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
21941 * doc/tm.texi (md_asm_adjust): Likewise.
21942 * target.def (md_asm_adjust): Likewise.
21944 2021-03-04 Richard Biener <rguenther@suse.de>
21946 PR middle-end/97855
21947 * tree-pretty-print.c: Poison pp_printf.
21948 (dump_decl_name): Avoid use of pp_printf.
21949 (dump_block_node): Likewise.
21950 (dump_generic_node): Likewise.
21952 2021-03-04 Martin Sebor <msebor@redhat.com>
21954 PR middle-end/96963
21955 PR middle-end/94655
21956 * builtins.c (handle_array_ref): New helper.
21957 (handle_mem_ref): New helper.
21958 (compute_objsize_r): Factor out ARRAY_REF and MEM_REF handling
21959 into new helper functions. Correct a workaround for vectorized
21962 2021-03-03 Pat Haugen <pthaugen@linux.ibm.com>
21964 * config/rs6000/dfp.md (extendddtd2, trunctddd2, *cmp<mode>_internal1,
21965 floatditd2, ftrunc<mode>2, fix<mode>di2, dfp_ddedpd_<mode>,
21966 dfp_denbcd_<mode>, dfp_dxex_<mode>, dfp_diex_<mode>,
21967 *dfp_sgnfcnc_<mode>, dfp_dscli_<mode>, dfp_dscri_<mode>): Update size
21968 attribute for Power10.
21969 * config/rs6000/mma.md (*movoo): Likewise.
21970 * config/rs6000/rs6000.md (define_attr "size"): Add 256.
21971 (define_mode_attr bits): Add DD/TD modes.
21972 * config/rs6000/sync.md (load_quadpti, store_quadpti, load_lockedpti,
21973 store_conditionalpti): Update size attribute for Power10.
21975 2021-03-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21978 * config/sparc/t-sparc (tree-ssanames.o-warn): Don't error for
21979 -Wuninitialized, -Wmaybe-uninitialized.
21980 (wide-int.o-warn): Likewise.
21982 2021-03-03 Richard Earnshaw <rearnsha@arm.com>
21984 * common/config/arm/arm-common.c: Include configargs.h.
21985 (arm_config_default): New function.
21986 (arm_target_mode): Renamed from arm_target_thumb_only. Handle
21987 processors that do not support Thumb. Take into account the
21988 --with-mode configuration setting for selecting the default.
21989 * config/arm/arm.h (OPTION_DEFAULT_SPECS): Remove entry for 'mode'.
21990 (TARGET_MODE_SPEC_FUNCTIONS): Update for function name change.
21992 2021-03-03 Martin Liska <mliska@suse.cz>
21994 PR gcov-profile/97461
21995 * gcov-io.h (GCOV_PREALLOCATED_KVP): Remove.
21997 2021-03-03 Eric Botcazou <ebotcazou@adacore.com>
22000 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
22001 point back the hard frame pointer to its default location when the
22002 frame is larger than SEH_MAX_FRAME_SIZE.
22004 2021-03-03 Jakub Jelinek <jakub@redhat.com>
22007 * config/i386/predicates.md (logic_operator): New define_predicate.
22008 * config/i386/i386.md (mov + mem using comm arith peephole2):
22009 Punt if operands[1] is EXT_REX_SSE_REGNO_P, AVX512BW is not enabled
22010 and the inner mode is [QH]Imode.
22012 2021-03-03 Jakub Jelinek <jakub@redhat.com>
22015 * dwarf2out.c (dw_loc_list_struct): Add end_entry member.
22016 (new_loc_list): Clear end_entry.
22017 (output_loc_list): Only use DW_LLE_startx_length for -gsplit-dwarf
22018 if HAVE_AS_LEB128, otherwise use DW_LLE_startx_endx. Fix comment
22020 (index_location_lists): For dwarf_version >= 5 without HAVE_AS_LEB128,
22021 initialize also end_entry.
22023 2021-03-03 Jakub Jelinek <jakub@redhat.com>
22026 * cfgrtl.c (fixup_partitions): When changing some bbs from hot to cold
22027 partitions, if in non-layout mode after reorder_blocks also move
22028 affected blocks to ensure a single partition transition.
22030 2021-03-03 Jason Merrill <jason@redhat.com>
22033 * cgraphunit.c (process_function_and_variable_attributes): Don't
22034 warn about flatten on an alias if the target also has it.
22035 * cgraph.h (symtab_node::get_alias_target_tree): New.
22037 2021-03-02 David Edelsohn <dje.gcc@gmail.com>
22039 * config/rs6000/rs6000.md (tls_get_tpointer_internal): Prepend
22040 period to symbol name.
22041 (tls_get_addr_internal<mode>): Same.
22043 2021-03-02 David Malcolm <dmalcolm@redhat.com>
22046 * diagnostic-show-locus.c
22047 (selftest::test_one_liner_many_fixits_2): Fix accidental usage of
22050 2021-03-02 Martin Sebor <msebor@redhat.com>
22052 PR middle-end/99276
22053 * builtins.c (warn_for_access): Remove stray warning text.
22055 2021-03-02 Martin Sebor <msebor@redhat.com>
22057 PR middle-end/99295
22058 * doc/extend.texi (attribute malloc): Reword and clarify nonaliasing
22061 2021-03-02 Jakub Jelinek <jakub@redhat.com>
22064 * dwarf2out.c (output_macinfo_op): Use DW_MACRO_*_str* even with
22065 -gdwarf-5 -gstrict-dwarf. For -gsplit-dwarf -gdwarf-5 use
22066 DW_MACRO_*_strx instead of DW_MACRO_*_strp. Handle
22067 DW_MACRO_define_strx and DW_MACRO_undef_strx.
22068 (save_macinfo_strings): Use DW_MACRO_*_str* even with
22069 -gdwarf-5 -gstrict-dwarf. Handle DW_MACRO_define_strx and
22070 DW_MACRO_undef_strx.
22072 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
22074 * config/s390/s390-builtin-types.def (BT_FN_V4SF_V8HI_UINT): New
22076 (BT_FN_V8HI_V8HI_UINT): Likewise.
22077 (BT_FN_V8HI_V4SF_V4SF_UINT): Likewise.
22078 * config/s390/s390-builtins.def (B_NNPA): New macro definition.
22079 (s390_vclfnhs, s390_vclfnls, s390_vcrnfs, s390_vcfn, s390_vcnf):
22080 New builtin definitions.
22081 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Bump
22082 vector extension version.
22083 * config/s390/s390.c (s390_expand_builtin): Check if builtins are
22084 available with current -march level.
22085 * config/s390/s390.md (UNSPEC_NNPA_VCLFNHS_V8HI)
22086 (UNSPEC_NNPA_VCLFNLS_V8HI, UNSPEC_NNPA_VCRNFS_V8HI)
22087 (UNSPEC_NNPA_VCFN_V8HI, UNSPEC_NNPA_VCNF_V8HI): New constants.
22088 * config/s390/vecintrin.h (vec_extend_to_fp32_hi): New macro.
22089 (vec_extend_to_fp32_lo): Likewise.
22090 (vec_round_from_fp32): Likewise.
22091 (vec_convert_to_fp16): Likewise.
22092 (vec_convert_from_fp16): Likewise.
22093 * config/s390/vx-builtins.md (vclfnhs_v8hi): New insn pattern.
22094 (vclfnls_v8hi): Likewise.
22095 (vcrnfs_v8hi): Likewise.
22096 (vcfn_v8hi): Likewise.
22097 (vcnf_v8hi): Likewise.
22099 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
22101 * common/config/s390/s390-common.c (processor_flags_table): New entry.
22102 * config.gcc: Enable arch14 for --with-arch and --with-tune.
22103 * config/s390/driver-native.c (s390_host_detect_local_cpu): Pick
22104 arch14 for unknown CPU models.
22105 * config/s390/s390-opts.h (enum processor_type): Add PROCESSOR_ARCH14.
22106 * config/s390/s390.c (s390_issue_rate): Add case for PROCESSOR_ARCH14.
22107 (s390_get_sched_attrmask): Likewise.
22108 (s390_get_unit_mask): Likewise.
22109 * config/s390/s390.h (enum processor_flags): Add PF_NNPA and PF_ARCH14.
22110 (TARGET_CPU_ARCH14, TARGET_CPU_ARCH14_P, TARGET_CPU_NNPA)
22111 (TARGET_CPU_NNPA_P, TARGET_ARCH14, TARGET_ARCH14_P, TARGET_NNPA)
22112 (TARGET_NNPA_P): New macro definitions.
22113 * config/s390/s390.md ("cpu_facility", "enabled"): Add arch14 and nnpa.
22114 * config/s390/s390.opt: Add PROCESSOR_ARCH14.
22116 2021-03-02 Jakub Jelinek <jakub@redhat.com>
22118 PR middle-end/95757
22119 * tree-vrp.c (register_edge_assert_for): Remove superfluous ()s around
22120 condition. Call register_edge_assert_for_1 for == 0, != 0, == 1 and
22121 != 1 comparisons if name is lhs of a comparison.
22123 2021-03-01 Iain Sandoe <iain@sandoe.co.uk>
22127 * config/darwin-protos.h (darwin_should_restore_cfa_state): New.
22128 * config/darwin.c (darwin_should_restore_cfa_state): New.
22129 * config/darwin.h (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New.
22130 * doc/tm.texi: Regenerated.
22131 * doc/tm.texi.in: Document TARGET_ASM_SHOULD_RESTORE_CFA_STATE.
22132 * dwarf2cfi.c (connect_traces): If the target requests, restore
22133 the CFA expression after a DW_CFA_restore.
22134 * target.def (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New hook.
22136 2021-03-01 Martin Liska <mliska@suse.cz>
22139 * optc-save-gen.awk: Add 4 more exceptions.
22141 2021-03-01 Nathan Sidwell <nathan@acm.org>
22144 * tree.h (TYPE_ALIGN_RAW): New accessor.
22145 (TYPE_ALIGN): Use it.
22147 2021-03-01 Jan Hubicka <jh@suse.cz>
22150 * ipa-fnsummary.c (compute_fn_summary): Fix sanity check.
22152 2021-03-01 Eric Botcazou <ebotcazou@adacore.com>
22155 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
22156 point the hard frame pointer to the SSE register save area instead
22157 of the general register save area. Perform only minimal adjustment
22158 for small frames if it is initially not correctly aligned.
22159 (ix86_expand_prologue): Remove early saves for a SEH target.
22160 * config/i386/winnt.c (struct seh_frame_state): Document constraint.
22162 2021-02-28 Jakub Jelinek <jakub@redhat.com>
22165 * ipa.c (symbol_table::remove_unreachable_nodes): Fix a comment
22166 typo - referneced -> referenced.
22167 * tree.c (component_ref_size): Fix comment typo -
22168 refernce -> reference.
22169 * tree-ssa-alias.c (access_path_may_continue_p): Fix comment typo -
22170 traling -> trailing.
22171 (aliasing_component_refs_p): Fix comment typos -
22172 refernce -> reference and refernece -> reference and
22173 traling -> trailing.
22174 (nonoverlapping_refs_since_match_p): Fix comment typo -
22175 referneces -> references.
22176 * doc/invoke.texi (--param modref-max-bases): Fix a typo -
22177 referneces -> references.
22179 2021-02-27 Iain Sandoe <iain@sandoe.co.uk>
22181 * config/host-darwin.c (darwin_gt_pch_use_address): Modify
22182 diagnostic message to avoid use of a contraction and format
22185 2021-02-27 Jakub Jelinek <jakub@redhat.com>
22188 * gcse.c (gcse_or_cprop_is_too_expensive): Use %wu instead of
22189 HOST_WIDE_INT_PRINT_UNSIGNED in warning format string.
22190 * ipa-devirt.c (ipa_odr_read_section): Use %wd instead of
22191 HOST_WIDE_INT_PRINT_DEC in inform format string. Fix comment
22194 2021-02-26 Richard Biener <rguenther@suse.de>
22196 PR middle-end/99281
22197 * expr.c (store_field): For calls with return-slot optimization
22198 and addressable return type expand the store directly.
22200 2021-02-26 Richard Biener <rguenther@suse.de>
22203 * builtins.c (warn_string_no_nul): Fix diagnostic formatting.
22205 2021-02-26 Peter Bergner <bergner@linux.ibm.com>
22208 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Replace assert
22211 2021-02-26 Aaron Sawdey <acsawdey@linux.ibm.com>
22213 * config.gcc: Add rs6000-pcrel-opt.o.
22214 * config/rs6000/rs6000-pcrel-opt.c: New file.
22215 * config/rs6000/pcrel-opt.md: New file.
22216 * config/rs6000/predicates.md: Add d_form_memory predicate.
22217 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_PCREL_OPT.
22218 * config/rs6000/rs6000-passes.def: Add pass_pcrel_opt.
22219 * config/rs6000/rs6000-protos.h: Add reg_to_non_prefixed(),
22220 pcrel_opt_valid_mem_p(), output_pcrel_opt_reloc(),
22221 and make_pass_pcrel_opt().
22222 * config/rs6000/rs6000.c (reg_to_non_prefixed): Make global.
22223 (rs6000_option_override_internal): Add pcrel-opt.
22224 (rs6000_delegitimize_address): Support pcrel-opt.
22225 (rs6000_opt_masks): Add pcrel-opt.
22226 (pcrel_opt_valid_mem_p): New function.
22227 (reg_to_non_prefixed): Make global.
22228 (rs6000_asm_output_opcode): Reset prepend_p_to_next_insn.
22229 (output_pcrel_opt_reloc): New function.
22230 * config/rs6000/rs6000.md (loads_extern_addr): New attr.
22231 (pcrel_extern_addr): Set loads_extern_addr.
22232 Add include for pcrel-opt.md.
22233 * config/rs6000/rs6000.opt: Add -mpcrel-opt.
22234 * config/rs6000/t-rs6000: Add rules for pcrel-opt.c and
22237 2021-02-26 YunQiang Su <yunqiang.su@cipunited.com>
22240 * config/mips/mips.c (mips_expand_ext_as_unaligned_load):
22241 If TARGET_64BIT and dest is SUBREG, we check the width, if it
22242 equal to SImode, we use SImode operation, just like what we are
22245 2021-02-26 Marek Polacek <polacek@redhat.com>
22247 * builtins.c (warn_for_access): Fix typos.
22249 2021-02-25 Iain Sandoe <iain@sandoe.co.uk>
22251 * config/aarch64/aarch64.md (<optab>_rol<mode>3): Add a '#'
22252 mark in front of the immediate quantity.
22253 (<optab>_rolsi3_uxtw): Likewise.
22255 2021-02-25 Richard Earnshaw <rearnsha@arm.com>
22258 * config/arm/thumb2.md (nonsecure_call_reg_thumb2_fpcxt): New pattern.
22259 (nonsecure_call_value_reg_thumb2_fpcxt): Likewise.
22260 (nonsecure_call_reg_thumb2): Restrict to using r4 for the callee
22261 address and disable when the FPCXT is not available.
22262 (nonsecure_call_value_reg_thumb2): Likewise.
22264 2021-02-25 Nathan Sidwell <nathan@acm.org>
22267 * doc/invoke.texi (flang-info-module-cmi): Renamed option.
22269 2021-02-25 Tamar Christina <tamar.christina@arm.com>
22271 * tree-vect-slp.c (optimize_load_redistribution_1): Abort on NULL nodes.
22273 2021-02-25 Richard Biener <rguenther@suse.de>
22275 PR tree-optimization/99253
22276 * tree-vect-loop.c (check_reduction_path): First compute
22277 code, then verify out-of-loop uses.
22279 2021-02-25 Jakub Jelinek <jakub@redhat.com>
22282 * match.pd ((T)(A) + CST -> (T)(A + CST)): Add :s to convert.
22284 2021-02-25 Jakub Jelinek <jakub@redhat.com>
22286 PR tree-optimization/80635
22287 * tree-vrp.c (vrp_simplify_cond_using_ranges): Also handle
22288 VIEW_CONVERT_EXPR if modes are the same, innerop is integral and
22289 has mode precision.
22291 2021-02-25 Richard Biener <rguenther@suse.de>
22293 * tree-vect-slp.c (optimize_load_redistribution_1): Delay
22294 load_map population.
22295 (vect_match_slp_patterns_2): Revert part of last change.
22296 (vect_analyze_slp): Do not interleave optimize_load_redistribution
22297 with pattern detection but do it afterwards. Dump the
22298 whole SLP graph after pattern recognition and load
22299 redistribution optimization finished.
22301 2021-02-24 Jakub Jelinek <jakub@redhat.com>
22304 * omp-low.c (struct omp_context): Add teams_nested_p and
22305 nonteams_nested_p members.
22306 (scan_omp_target): Diagnose teams nested inside of target with other
22307 directives strictly nested inside of the same target.
22308 (check_omp_nesting_restrictions): Set ctx->teams_nested_p or
22309 ctx->nonteams_nested_p as needed.
22311 2021-02-24 Vladimir N. Makarov <vmakarov@redhat.com>
22313 PR inline-asm/99123
22314 * lra-constraints.c (uses_hard_regs_p): Don't use decompose_mem_address.
22316 2021-02-24 Hans-Peter Nilsson <hp@axis.com>
22318 * config/cris/cris.c (cris_expand_prologue): Set
22319 current_function_static_stack_size, if flag_stack_usage_info.
22321 2021-02-24 Pat Haugen <pthaugen@linux.ibm.com>
22323 * config/rs6000/rs6000.c (next_insn_prefixed_p): Rename.
22324 (rs6000_final_prescan_insn): Adjust.
22325 (rs6000_asm_output_opcode): Likewise.
22327 2021-02-24 Martin Sebor <msebor@redhat.com>
22329 PR middle-end/97172
22330 * attribs.c (attr_access::free_lang_data): Clear attribute arg spec
22331 from function arguments.
22333 2021-02-24 Tamar Christina <tamar.christina@arm.com>
22335 PR tree-optimization/99220
22336 * tree-vect-slp.c (optimize_load_redistribution_1): Remove
22337 node from cache when it's about to be deleted.
22339 2021-02-24 Jakub Jelinek <jakub@redhat.com>
22341 PR tree-optimization/99225
22342 * fold-const.c (fold_binary_loc) <case NE_EXPR>: In (x & (1 << y)) != 0
22343 to ((x >> y) & 1) != 0 simplifications use build_one_cst instead of
22344 build_int_cst (..., 1). Formatting fixes.
22346 2021-02-24 Tamar Christina <tamar.christina@arm.com>
22348 PR tree-optimization/99149
22349 * tree-vect-slp-patterns.c (vect_detect_pair_op): Don't recreate the
22351 (vect_slp_reset_pattern): Remove.
22352 (complex_fma_pattern::matches): Remove call to vect_slp_reset_pattern.
22353 (complex_mul_pattern::build, complex_fma_pattern::build,
22354 complex_fms_pattern::build): Fix ref counts.
22355 * tree-vect-slp.c (vect_free_slp_tree): Undo SLP only pattern relevancy
22356 when node is being deleted.
22357 (vect_match_slp_patterns_2): Correct result of cache hit on patterns.
22358 (vect_schedule_slp): Invalidate SLP_TREE_REPRESENTATIVE of removed
22360 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize value.
22362 2021-02-24 Matthias Klose <doko@ubuntu.com>
22365 2020-12-07 Matthias Klose <doko@ubuntu.com>
22367 * genextract.c (print_header): Undefine ENABLE_RTL_CHECKING
22368 and ENABLE_RTL_FLAG_CHECKING.
22370 2021-02-24 Richard Biener <rguenther@suse.de>
22373 * builtins.c (fold_builtin_next_arg): Avoid NULL arg.
22375 2021-02-23 Peter Bergner <bergner@linux.ibm.com>
22377 * config/rs6000/mma.md (mma_assemble_pair): Rename from this...
22378 (vsx_assemble_pair): ...to this.
22379 (*mma_assemble_pair): Rename from this...
22380 (*vsx_assemble_pair): ...to this.
22381 (mma_disassemble_pair): Rename from this...
22382 (vsx_disassemble_pair): ...to this.
22383 (*mma_disassemble_pair): Rename from this...
22384 (*vsx_disassemble_pair): ...to this.
22385 * config/rs6000/rs6000-builtin.def (BU_MMA_V2, BU_MMA_V3,
22386 BU_COMPAT): New macros.
22387 (mma_assemble_pair): Rename from this...
22388 (vsx_assemble_pair): ...to this.
22389 (mma_disassemble_pair): Rename from this...
22390 (vsx_disassemble_pair): ...to this.
22391 (mma_assemble_pair): New compatibility built-in.
22392 (mma_disassemble_pair): Likewise.
22393 * config/rs6000/rs6000-call.c (struct builtin_compatibility): New.
22394 (RS6000_BUILTIN_COMPAT): Define.
22395 (bdesc_compat): New.
22396 (mma_expand_builtin): Use VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
22397 (rs6000_gimple_fold_mma_builtin): Use MMA_BUILTIN_DISASSEMBLE_PAIR
22398 and VSX_BUILTIN_ASSEMBLE_PAIR.
22399 (rs6000_init_builtins): Register compatibility built-ins.
22400 (mma_init_builtins): Use VSX_BUILTIN_ASSEMBLE_PAIR,
22401 VSX_BUILTIN_ASSEMBLE_PAIR_INTERNAL, VSX_BUILTIN_DISASSEMBLE_PAIR and
22402 VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
22403 * doc/extend.texi (__builtin_mma_assemble_pair): Rename from this...
22404 (__builtin_vsx_assemble_pair): ...to this.
22405 (__builtin_mma_disassemble_pair): Rename from this...
22406 (__builtin_vsx_disassemble_pair): ...to this.
22408 2021-02-23 Martin Liska <mliska@suse.cz>
22411 * ipa-icf.c (sem_variable::merge): Do not merge 2 variables
22412 with different alignment. That leads to an invalid red zone
22413 size allocated in runtime.
22415 2021-02-23 Jakub Jelinek <jakub@redhat.com>
22417 PR tree-optimization/99204
22418 * fold-const.c (fold_read_from_constant_string): Check that
22419 tree_fits_uhwi_p (index) rather than just that index is INTEGER_CST.
22421 2021-02-23 Segher Boessenkool <segher@kernel.crashing.org>
22422 Kewen Lin <linkw@gcc.gnu.org>
22424 * config/rs6000/rs6000.md (*rotl<mode>3_insert_3): Renamed to...
22425 (rotl<mode>3_insert_3): ...this.
22426 (plus_ior_xor): New code_iterator.
22427 (define_split for GPR rl*imi): New splitter.
22428 * config/rs6000/vsx.md (vsx_init_v4si): Use gen_rotldi3_insert_3
22429 for integer merging.
22431 2021-02-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22433 * config/aarch64/aarch64-tuning-flags.def (cse_sve_vl_constants):
22435 * config/aarch64/aarch64.md (add<mode>3): Force CONST_POLY_INT immediates
22436 into a register when the above is enabled.
22437 * config/aarch64/aarch64.c (neoversev1_tunings):
22438 AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
22439 (aarch64_rtx_costs): Use AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
22441 2021-02-22 Hans-Peter Nilsson <hp@axis.com>
22443 * config/cris/cris.c (cris_print_operand) <'T'>: Change
22444 valid operand from is now an addi mult-value to shift-value.
22445 * config/cris/cris.md (*addi): Change expression of scaled
22446 operand from mult to ashift.
22447 * config/cris/cris.md (*addi_reload): New insn_and_split.
22449 2021-02-22 John David Anglin <danglin@gcc.gnu.org>
22452 * config/pa/pa.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Define as
22453 hook_bool_const_tree_hwi_hwi_const_tree_true.
22454 (pa_asm_output_mi_thunk): Add support for nonzero vcall_offset.
22456 2021-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
22458 PR rtl-optimization/98791
22459 * ira-conflicts.c (process_regs_for_copy): Don't create allocno copies
22460 for unordered modes.
22462 2021-02-22 Martin Liska <mliska@suse.cz>
22464 * tree-inline.c (inline_forbidden_p): Set
22465 inline_forbidden_reason.
22467 2021-02-22 Richard Biener <rguenther@suse.de>
22469 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Dump
22472 2021-02-22 Richard Biener <rguenther@suse.de>
22474 PR tree-optimization/99165
22475 * gimple-ssa-store-merging.c (pass_store_merging::process_store):
22476 Accumulate changed to ret.
22478 2021-02-21 Uros Bizjak <ubizjak@gmail.com>
22481 2020-12-09 Uroš Bizjak <ubizjak@gmail.com>
22483 * config/i386/i386.h (REG_ALLOC_ORDER): Remove
22485 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
22488 * config/s390/vector.md (trunctf<DFP_ALL:mode>2_vr): New
22490 (trunctf<DFP_ALL:mode>2): Likewise.
22491 (trunctdtf2_vr): Likewise.
22492 (trunctdtf2): Likewise.
22493 (extend<DFP_ALL:mode>tf2_vr): Likewise.
22494 (extend<DFP_ALL:mode>tf2): Likewise.
22495 (extendtftd2_vr): Likewise.
22496 (extendtftd2): Likewise.
22498 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
22500 * config/s390/vector.md (*fprx2_to_tf): Rename to fprx2_to_tf,
22501 add memory alternative.
22502 (tf_to_fprx2): New pattern.
22504 2021-02-19 Martin Sebor <msebor@redhat.com>
22507 * attribs.c (init_attr_rdwr_indices): Guard vblist use.
22508 (attr_access::free_lang_data): Remove a spurious test.
22510 2021-02-19 Nathan Sidwell <nathan@acm.org>
22512 * doc/invoke.texi (flang-info-module-read): Document.
22514 2021-02-19 Martin Liska <mliska@suse.cz>
22516 PR translation/99167
22517 * params.opt: Fix typo.
22519 2021-02-19 Richard Biener <rguenther@suse.de>
22521 PR middle-end/99122
22522 * tree-inline.c (inline_forbidden_p): Do not inline functions
22523 with VLA arguments or return value.
22525 2021-02-19 Jakub Jelinek <jakub@redhat.com>
22528 * config/arm/arm.md (*stack_protect_combined_set_insn,
22529 *stack_protect_combined_test_insn): If force_const_mem result
22530 is not valid general operand, force its address into the destination
22533 2021-02-19 Jakub Jelinek <jakub@redhat.com>
22536 * tree-cfg.c (gimple_merge_blocks): If bb a starts with eh landing
22537 pad or non-local label, put FORCED_LABELs from bb b after that label
22538 rather than before it.
22540 2021-02-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
22543 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3): Use
22544 expand_vector_broadcast' to emit the vec_duplicate operand.
22546 2021-02-18 Vladimir N. Makarov <vmakarov@redhat.com>
22548 PR rtl-optimization/96264
22549 * lra-remat.c (reg_overlap_for_remat_p): Check also output insn
22552 2021-02-18 H.J. Lu <hjl.tools@gmail.com>
22555 * varasm.c (get_section): Replace SUPPORTS_SHF_GNU_RETAIN with
22556 looking up the retain attribute.
22557 (resolve_unique_section): Likewise.
22558 (get_variable_section): Likewise.
22559 (switch_to_section): Likewise. Warn when a symbol without the
22560 retain attribute and a symbol with the retain attribute are
22561 placed in the section with the same name, instead of the used
22563 * doc/extend.texi: Document the "retain" attribute.
22565 2021-02-18 Nathan Sidwell <nathan@acm.org>
22568 * doc/invoke.texi (flang-info-include-translate): Document header
22571 2021-02-18 Richard Biener <rguenther@suse.de>
22573 PR middle-end/99122
22574 * ipa-fnsummary.c (analyze_function_body): Set
22575 CIF_FUNCTION_NOT_INLINABLE for VLA parameter calls.
22576 * tree-inline.c (insert_init_debug_bind): Pass NULL for
22577 error_mark_node values.
22578 (force_value_to_type): Do not build V_C_Es for WITH_SIZE_EXPR
22580 (setup_one_parameter): Delay force_value_to_type until when
22583 2021-02-18 Hans-Peter Nilsson <hp@axis.com>
22585 PR tree-optimization/99142
22586 * match.pd (clz cmp 0): Gate replacement on single_use of clz result.
22588 2021-02-18 Jakub Jelinek <jakub@redhat.com>
22590 * wide-int-bitmask.h (wide_int_bitmask::wide_int_bitmask (),
22591 wide_int_bitmask::wide_int_bitmask (uint64_t),
22592 wide_int_bitmask::wide_int_bitmask (uint64_t, uint64_t),
22593 wide_int_bitmask::operator ~ () const,
22594 wide_int_bitmask::operator | (wide_int_bitmask) const,
22595 wide_int_bitmask::operator & (wide_int_bitmask) const): Use constexpr
22597 * config/i386/i386.h (PTA_3DNOW, PTA_3DNOW_A, PTA_64BIT, PTA_ABM,
22598 PTA_AES, PTA_AVX, PTA_BMI, PTA_CX16, PTA_F16C, PTA_FMA, PTA_FMA4,
22599 PTA_FSGSBASE, PTA_LWP, PTA_LZCNT, PTA_MMX, PTA_MOVBE, PTA_NO_SAHF,
22600 PTA_PCLMUL, PTA_POPCNT, PTA_PREFETCH_SSE, PTA_RDRND, PTA_SSE, PTA_SSE2,
22601 PTA_SSE3, PTA_SSE4_1, PTA_SSE4_2, PTA_SSE4A, PTA_SSSE3, PTA_TBM,
22602 PTA_XOP, PTA_AVX2, PTA_BMI2, PTA_RTM, PTA_HLE, PTA_PRFCHW, PTA_RDSEED,
22603 PTA_ADX, PTA_FXSR, PTA_XSAVE, PTA_XSAVEOPT, PTA_AVX512F, PTA_AVX512ER,
22604 PTA_AVX512PF, PTA_AVX512CD, PTA_NO_TUNE, PTA_SHA, PTA_PREFETCHWT1,
22605 PTA_CLFLUSHOPT, PTA_XSAVEC, PTA_XSAVES, PTA_AVX512DQ, PTA_AVX512BW,
22606 PTA_AVX512VL, PTA_AVX512IFMA, PTA_AVX512VBMI, PTA_CLWB, PTA_MWAITX,
22607 PTA_CLZERO, PTA_NO_80387, PTA_PKU, PTA_AVX5124VNNIW, PTA_AVX5124FMAPS,
22608 PTA_AVX512VPOPCNTDQ, PTA_SGX, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES,
22609 PTA_AVX512VBMI2, PTA_VPCLMULQDQ, PTA_AVX512BITALG, PTA_RDPID,
22610 PTA_PCONFIG, PTA_WBNOINVD, PTA_AVX512VP2INTERSECT, PTA_PTWRITE,
22611 PTA_AVX512BF16, PTA_WAITPKG, PTA_MOVDIRI, PTA_MOVDIR64B, PTA_ENQCMD,
22612 PTA_CLDEMOTE, PTA_SERIALIZE, PTA_TSXLDTRK, PTA_AMX_TILE, PTA_AMX_INT8,
22613 PTA_AMX_BF16, PTA_UINTR, PTA_HRESET, PTA_KL, PTA_WIDEKL, PTA_AVXVNNI,
22614 PTA_X86_64_BASELINE, PTA_X86_64_V2, PTA_X86_64_V3, PTA_X86_64_V4,
22615 PTA_CORE2, PTA_NEHALEM, PTA_WESTMERE, PTA_SANDYBRIDGE, PTA_IVYBRIDGE,
22616 PTA_HASWELL, PTA_BROADWELL, PTA_SKYLAKE, PTA_SKYLAKE_AVX512,
22617 PTA_CASCADELAKE, PTA_COOPERLAKE, PTA_CANNONLAKE, PTA_ICELAKE_CLIENT,
22618 PTA_ICELAKE_SERVER, PTA_TIGERLAKE, PTA_SAPPHIRERAPIDS, PTA_ALDERLAKE,
22619 PTA_KNL, PTA_BONNELL, PTA_SILVERMONT, PTA_GOLDMONT, PTA_GOLDMONT_PLUS,
22620 PTA_TREMONT, PTA_KNM): Use constexpr instead of const.
22622 2021-02-18 Jakub Jelinek <jakub@redhat.com>
22624 PR middle-end/99109
22625 * gimple-array-bounds.cc (build_zero_elt_array_type): Rename to ...
22626 (build_printable_array_type): ... this. Add nelts argument. For
22627 overaligned eltype, use TYPE_MAIN_VARIANT (eltype) instead. If
22628 nelts, call build_array_type_nelts.
22629 (array_bounds_checker::check_mem_ref): Use build_printable_array_type
22630 instead of build_zero_elt_array_type and build_array_type_nelts.
22632 2021-02-18 Jakub Jelinek <jakub@redhat.com>
22635 * config/i386/i386.c (distance_non_agu_define): Don't call
22636 extract_insn_cached here.
22637 (ix86_lea_outperforms): Save and restore recog_data around call
22638 to distance_non_agu_define and distance_agu_use.
22639 (ix86_ok_to_clobber_flags): Remove.
22640 (ix86_avoid_lea_for_add): Don't call ix86_ok_to_clobber_flags.
22641 (ix86_avoid_lea_for_addr): Likewise. Adjust function comment.
22642 * config/i386/i386.md (*lea<mode>): Change from define_insn_and_split
22643 into define_insn. Move the splitting to define_peephole2 and
22644 check there using peep2_regno_dead_p if FLAGS_REG is dead.
22646 2021-02-17 Julian Brown <julian@codesourcery.com>
22648 * gimplify.c (gimplify_scan_omp_clauses): Handle ATTACH_DETACH
22651 2021-02-17 Xi Ruoyao <xry111@mengyan1223.wang>
22654 * config/mips/mips.c (mips_symbol_insns): Do not use
22655 MSA_SUPPORTED_MODE_P if mode is MAX_MACHINE_MODE.
22657 2021-02-16 Vladimir N. Makarov <vmakarov@redhat.com>
22659 PR inline-asm/98096
22660 * stmt.c (resolve_operand_name_1): Take inout operands into account
22661 for access to labels by names.
22662 * doc/extend.texi: Describe counting operands for accessing labels.
22664 2021-02-16 Richard Biener <rguenther@suse.de>
22666 PR tree-optimization/38474
22667 * tree-ssa-structalias.c (variable_info::address_taken): New.
22668 (new_var_info): Initialize address_taken.
22669 (process_constraint): Set address_taken.
22670 (solve_constraints): Use the new address_taken flag rather
22671 than is_reg_var for sorting variables.
22672 (dump_constraint): Dump the variable number if the name
22675 2021-02-16 Jakub Jelinek <jakub@redhat.com>
22678 * tree-vect-stmts.c (vectorizable_simd_clone_call): For num_calls != 1
22679 multiply by 4096 and for inbranch by 8192.
22680 * config/i386/i386.c (ix86_simd_clone_usable): For TARGET_AVX512F,
22681 return 3, 2 or 1 for mangle letters 'b', 'c' or 'd'.
22683 2021-02-15 Maya Rashish <coypu@sdf.org>
22685 * config/aarch64/aarch64.c (aarch64_init_builtins):
22686 Call SUBTARGET_INIT_BUILTINS.
22688 2021-02-15 Peter Bergner <bergner@linux.ibm.com>
22690 PR rtl-optimization/98872
22691 * init-regs.c (initialize_uninitialized_regs): Skip initialization
22692 if CONST0_RTX is NULL.
22694 2021-02-15 Richard Sandiford <richard.sandiford@arm.com>
22696 PR rtl-optimization/98863
22697 * rtl-ssa/functions.h (function_info::bb_live_out_info): Delete.
22698 (function_info::build_info): Turn into a declaration, moving the
22699 definition to internals.h.
22700 (function_info::bb_walker): Declare.
22701 (function_info::create_reg_use): Likewise.
22702 (function_info::calculate_potential_phi_regs): Take a build_info
22704 (function_info::place_phis, function_info::create_ebbs): Declare.
22705 (function_info::calculate_ebb_live_in_for_debug): Likewise.
22706 (function_info::populate_backedge_phis): Delete.
22707 (function_info::start_block, function_info::end_block): Declare.
22708 (function_info::populate_phi_inputs): Delete.
22709 (function_info::m_potential_phi_regs): Move information to build_info.
22710 * rtl-ssa/internals.h: New file.
22711 (function_info::bb_phi_info): New class.
22712 (function_info::build_info): Moved from functions.h.
22713 Add a constructor and destructor.
22714 (function_info::build_info::ebb_use): Delete.
22715 (function_info::build_info::ebb_def): Likewise.
22716 (function_info::build_info::bb_live_out): Likewise.
22717 (function_info::build_info::tmp_ebb_live_in_for_debug): New variable.
22718 (function_info::build_info::potential_phi_regs): Likewise.
22719 (function_info::build_info::potential_phi_regs_for_debug): Likewise.
22720 (function_info::build_info::ebb_def_regs): Likewise.
22721 (function_info::build_info::bb_phis): Likewise.
22722 (function_info::build_info::bb_mem_live_out): Likewise.
22723 (function_info::build_info::bb_to_rpo): Likewise.
22724 (function_info::build_info::def_stack): Likewise.
22725 (function_info::build_info::old_def_stack_limit): Likewise.
22726 * rtl-ssa/internals.inl (function_info::build_info::record_reg_def):
22727 Remove the regno argument. Push the previous definition onto the
22728 definition stack where necessary.
22729 * rtl-ssa/accesses.cc: Include internals.h.
22730 * rtl-ssa/changes.cc: Likewise.
22731 * rtl-ssa/blocks.cc: Likewise.
22732 (function_info::build_info::build_info): Define.
22733 (function_info::build_info::~build_info): Likewise.
22734 (function_info::bb_walker): New class.
22735 (function_info::bb_walker::bb_walker): Define.
22736 (function_info::add_live_out_use): Convert a logarithmic-complexity
22737 test into a linear one. Allow the same definition to be passed
22739 (function_info::calculate_potential_phi_regs): Moved from
22740 functions.cc. Take a build_info parameter and store the
22741 information there instead.
22742 (function_info::place_phis): New function.
22743 (function_info::add_entry_block_defs): Update call to record_reg_def.
22744 (function_info::calculate_ebb_live_in_for_debug): New function.
22745 (function_info::add_phi_nodes): Use bb_phis to decide which
22746 registers need phi nodes and initialize ebb_def_regs accordingly.
22747 Do not add degenerate phis here.
22748 (function_info::add_artificial_accesses): Use create_reg_use.
22749 Assert that all definitions are listed in the DF LR sets.
22750 Update call to record_reg_def.
22751 (function_info::record_block_live_out): Record live-out register
22752 values in the phis of successor blocks. Use the live-out set
22753 when processing the last block in an EBB, instead of always
22754 using the live-in sets of successor blocks. AND the live sets
22755 with the set of registers that have been defined in the EBB,
22756 rather than with all potential phi registers. Cope correctly
22757 with branches back to the start of the current EBB.
22758 (function_info::start_block): New function.
22759 (function_info::end_block): Likewise.
22760 (function_info::populate_phi_inputs): Likewise.
22761 (function_info::create_ebbs): Likewise.
22762 (function_info::process_all_blocks): Rewrite into a multi-phase
22764 * rtl-ssa/functions.cc: Include internals.h.
22765 (function_info::calculate_potential_phi_regs): Move to blocks.cc.
22766 (function_info::init_function_data): Remove caller.
22767 * rtl-ssa/insns.cc: Include internals.h
22768 (function_info::create_reg_use): New function. Lazily any
22769 degenerate phis needed by the linear RPO view.
22770 (function_info::record_use): Use create_reg_use. When processing
22771 debug uses, use potential_phi_regs and test it before checking
22772 whether the register is live on entry to the current EBB. Lazily
22773 calculate ebb_live_in_for_debug.
22774 (function_info::record_call_clobbers): Update call to record_reg_def.
22775 (function_info::record_def): Likewise.
22777 2021-02-15 Martin Liska <mliska@suse.cz>
22779 * toplev.c (init_asm_output): Free output of
22780 gen_command_line_string function.
22781 (process_options): Likewise.
22783 2021-02-15 Martin Liska <mliska@suse.cz>
22785 * params.opt: Add 2 missing Param keywords.
22787 2021-02-15 Eric Botcazou <ebotcazou@adacore.com>
22789 * df-core.c (df_worklist_dataflow_doublequeue): Use proper cast.
22791 2021-02-15 Jakub Jelinek <jakub@redhat.com>
22793 PR tree-optimization/99079
22794 * match.pd (A % (pow2pcst << N) -> A & ((pow2pcst << N) - 1)): Remove
22795 useless tree_nop_conversion_p (type, TREE_TYPE (@3)) check. Instead
22796 require both type and TREE_TYPE (@1) to be integral types and either
22797 type having smaller or equal precision, or TREE_TYPE (@1) being
22798 unsigned type, or type being signed type. If TREE_TYPE (@1)
22799 doesn't have wrapping overflow, perform the subtraction of one in
22802 2021-02-14 Jan Hubicka <hubicka@ucw.cz>
22803 Richard Biener <rguether@suse.de>
22806 * ipa-reference.c (ipa_init): Only conditinally initialize
22807 reference_vars_to_consider.
22808 (propagate): Conditionally deninitialize reference_vars_to_consider.
22809 (ipa_reference_write_optimization_summary): Sanity check that
22810 reference_vars_to_consider is not allocated.
22812 2021-02-13 Levy Hsu <admin@levyhsu.com>
22815 * config/riscv/riscv-shorten-memrefs.c (pass_shorten_memrefs): Add
22816 extend parameter to get_si_mem_base_reg declaration.
22817 (get_si_mem_base_reg): Add extend parameter. Set it.
22818 (analyze): Pass extend arg to get_si_mem_base_reg.
22819 (transform): Likewise. Use it when rewriting mems.
22820 * config/riscv/riscv.c (riscv_legitimize_move): Check for subword
22821 loads and emit sign/zero extending load followed by subreg move.
22823 2021-02-13 Jim Wilson <jimw@sifive.com>
22826 * config/riscv/riscv.c (riscv_compressed_lw_address_p): Drop early
22827 exit when !reload_completed. Only perform check for compressed reg
22828 if reload_completed.
22829 (riscv_rtx_costs): In MEM case, when optimizing for size and
22830 shorten memrefs, if not compressible, then increase cost.
22832 2021-02-13 Jakub Jelinek <jakub@redhat.com>
22834 PR rtl-optimization/98439
22835 * recog.c (pass_split_before_regstack::gate): Enable even when
22836 pass_split_before_sched2 is enabled if -fselective-scheduling2 is
22839 2021-02-13 Jakub Jelinek <jakub@redhat.com>
22842 * config/i386/mmx.md (*mmx_pshufd_1): Add a combine splitter for
22843 swap of V2SImode elements in memory into DImode memory rotate by 32.
22845 2021-02-12 Martin Sebor <msebor@redhat.com>
22847 * tree-pretty-print.c (print_generic_expr_to_str): Update comment.
22849 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
22851 * rtl-ssa/accesses.cc (function_info::make_use_available): Use
22852 m_temp_obstack rather than m_obstack to allocate the temporary use.
22854 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
22856 * df-problems.c (df_lr_bb_local_compute): Treat partial definitions
22857 as read-modify operations.
22859 2021-02-12 Richard Biener <rguenther@suse.de>
22861 PR middle-end/38474
22862 * ipa-fnsummary.c (unmodified_parm_1): Only walk when
22863 fbi->aa_walk_budget is bigger than zero. Update
22864 fbi->aa_walk_budget.
22865 (param_change_prob): Likewise.
22866 * ipa-prop.c (detect_type_change_from_memory_writes):
22867 Properly account walk_aliased_vdefs.
22868 (parm_preserved_before_stmt_p): Canonicalize updates.
22869 (parm_ref_data_preserved_p): Likewise.
22870 (parm_ref_data_pass_through_p): Likewise.
22871 (determine_known_aggregate_parts): Account own alias queries.
22873 2021-02-12 Martin Liska <mliska@suse.cz>
22875 * opts-common.c (decode_cmdline_option): Release werror_arg.
22876 * opts.c (gen_producer_string): Release output of
22877 gen_command_line_string.
22879 2021-02-12 Richard Biener <rguenther@suse.de>
22881 PR tree-optimization/38474
22882 * params.opt (-param=max-store-chains-to-track=): New param.
22883 (-param=max-stores-to-track=): Likewise.
22884 * doc/invoke.texi (max-store-chains-to-track): Document.
22885 (max-stores-to-track): Likewise.
22886 * gimple-ssa-store-merging.c (pass_store_merging::m_n_chains):
22888 (pass_store_merging::m_n_stores): Likewise.
22889 (pass_store_merging::terminate_and_process_chain): Update
22890 m_n_stores and m_n_chains.
22891 (pass_store_merging::process_store): Likewise. Terminate
22892 oldest chains if the number of stores or chains get too large.
22893 (imm_store_chain_info::terminate_and_process_chain): Dump
22896 2021-02-11 Eric Botcazou <ebotcazou@adacore.com>
22898 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
22899 the cold section, emit a nop before the directive if the previous
22900 active instruction can throw.
22902 2021-02-11 Peter Bergner <bergner@linux.ibm.com>
22905 * config/rs6000/predicates.md (mma_assemble_input_operand): Restrict
22906 memory addresses that are legal for quad word accesses.
22908 2021-02-11 Andrea Corallo <andrea.corallo@arm.com>
22911 * config/arm/thumb2.md (*doloop_end_internal): Generate
22912 alternative sequence to handle long range branches.
22914 2021-02-11 Joel Hutton <joel.hutton@arm.com>
22916 PR tree-optimization/98772
22917 * optabs-tree.c (supportable_half_widening_operation): New function
22918 to check for supportable V8QI->V8HI widening patterns.
22919 * optabs-tree.h (supportable_half_widening_operation): New function.
22920 * tree-vect-stmts.c (vect_create_half_widening_stmts): New function
22921 to create promotion stmts for V8QI->V8HI widening patterns.
22922 (vectorizable_conversion): Add case for V8QI->V8HI.
22924 2021-02-11 Richard Biener <rguenther@suse.de>
22926 * sparseset.h (SPARSESET_ELT_BITS): Remove.
22927 (SPARSESET_ELT_TYPE): Use unsigned int.
22928 * fwprop.c: Do not include sparseset.h.
22930 2021-02-10 Jakub Jelinek <jakub@redhat.com>
22933 * varasm.c (declare_weak): For -fsyntax-only, allow even
22934 TREE_ASM_WRITTEN function decls.
22936 2021-02-10 Jakub Jelinek <jakub@redhat.com>
22939 * config/i386/sse.md (fix<fixunssuffix>_truncv2sfv2di2,
22940 <insn>v8qiv8hi2, <insn>v8qiv8si2, <insn>v4qiv4si2, <insn>v4hiv4si2,
22941 <insn>v8qiv8di2, <insn>v4qiv4di2, <insn>v2qiv2di2, <insn>v4hiv4di2,
22942 <insn>v2hiv2di2, <insn>v2siv2di2): Force operands[1] into REG before
22943 calling simplify_gen_subreg on it.
22945 2021-02-10 Martin Liska <mliska@suse.cz>
22947 * config/nvptx/nvptx.c (nvptx_option_override): Use
22948 flag_patchable_function_entry instead of the removed
22949 function_entry_patch_area_size.
22951 2021-02-10 Martin Liska <mliska@suse.cz>
22953 PR tree-optimization/99002
22954 PR tree-optimization/99026
22955 * gimple-if-to-switch.cc (if_chain::is_beneficial): Fix memory
22956 leak when adjacent cases are merged.
22957 * tree-switch-conversion.c (switch_decision_tree::analyze_switch_statement): Use
22959 (make_pass_lower_switch): Remove trailing whitespace.
22960 * tree-switch-conversion.h (release_clusters): New.
22962 2021-02-10 Richard Biener <rguenther@suse.de>
22964 PR rtl-optimization/99054
22965 * cfgrtl.c (rtl-optimization/99054): Return an auto_vec.
22966 (fixup_partitions): Adjust.
22967 (rtl_verify_edges): Likewise.
22969 2021-02-10 Jakub Jelinek <jakub@redhat.com>
22971 PR middle-end/99007
22972 * gimplify.c (gimplify_scan_omp_clauses): For MEM_REF on reductions,
22973 temporarily disable gimplify_ctxp->into_ssa around gimplify_expr
22976 2021-02-10 Richard Biener <rguenther@suse.de>
22979 * ipa-pure-const.c (propagate_malloc): Use an auto_vec<>
22982 2021-02-10 Richard Biener <rguenther@suse.de>
22984 PR tree-optimization/99024
22985 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Only
22986 clear loop->aux if it is associated with the destroyed loop_vinfo.
22988 2021-02-10 Martin Liska <mliska@suse.cz>
22990 PR tree-optimization/99002
22991 * gimple-if-to-switch.cc (find_conditions): Fix memory leak
22994 2021-02-10 Martin Liska <mliska@suse.cz>
22997 * ipa-icf.c (sem_item::add_reference): Fix memory leak when
22998 a reference exists.
23000 2021-02-10 Jakub Jelinek <jakub@redhat.com>
23003 * dwarf2out.c (prune_unused_types_walk): Mark DW_TAG_variable DIEs
23004 at class scope for DWARF5+.
23006 2021-02-09 Eric Botcazou <ebotcazou@adacore.com>
23008 PR rtl-optimization/96015
23009 * reorg.c (skip_consecutive_labels): Minor comment tweaks.
23010 (relax_delay_slots): When deleting a jump to the next active
23011 instruction over a barrier, first delete the barrier if the
23012 jump is the only way to reach the target label.
23014 2021-02-09 Andre Vieira <andre.simoesdiasvieira@arm.com>
23016 * config/aarch64/aarch64-cost-tables.h: Add entries for vect.mul.
23017 * config/aarch64/aarch64.c (aarch64_rtx_mult_cost): Use vect.mul for
23018 vector multiplies and vect.alu for SSRA.
23019 * config/arm/aarch-common-protos.h (struct vector_cost_table): Define
23020 vect.mul cost field.
23021 * config/arm/aarch-cost-tables.h: Add entries for vect.mul.
23022 * config/arm/arm.c: Likewise.
23024 2021-02-09 Richard Biener <rguenther@suse.de>
23026 PR tree-optimization/98863
23027 * tree-ssa-sccvn.h (vn_avail::next_undo): Add.
23028 * tree-ssa-sccvn.c (last_pushed_avail): New global.
23029 (rpo_elim::eliminate_push_avail): Chain pushed avails.
23030 (unwind_state::avail_top): Add.
23031 (do_unwind): Rewrite unwinding of avail entries.
23032 (do_rpo_vn): Initialize last_pushed_avail and
23033 avail_top of the undo state.
23035 2021-02-09 Jakub Jelinek <jakub@redhat.com>
23037 PR middle-end/99004
23038 * calls.c (maybe_warn_rdwr_sizes): Change s0 and s1 type from
23039 const char * to char * and free those pointers after use.
23041 2021-02-09 Richard Biener <rguenther@suse.de>
23043 PR tree-optimization/99017
23044 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Allow
23045 zero vector cost entries.
23047 2021-02-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
23049 PR middle-end/98974
23050 * tree-vect-stmts.c (vectorizable_condition): Remove shadow vec_num
23051 parameter in vectorizable_condition.
23053 2021-02-08 Richard Biener <rguenther@suse.de>
23056 * tree.c (walk_tree_1): Walk VECTOR_CST elements.
23058 2021-02-08 Martin Liska <mliska@suse.cz>
23061 * cfgexpand.c (pass_expand::execute): Parse per-function option
23062 flag_patchable_function_entry and use it.
23063 * common.opt: Remove function_entry_patch_area_size and
23064 function_entry_patch_area_start global variables.
23065 * opts.c (parse_and_check_patch_area): New function.
23066 (common_handle_option): Use it.
23067 * opts.h (parse_and_check_patch_area): New function.
23068 * toplev.c (process_options): Parse and use
23069 function_entry_patch_area_size.
23071 2021-02-08 Martin Sebor <msebor@redhat.com>
23073 * doc/extend.texi (attribute malloc): Correct typos.
23075 2021-02-05 Nathan Sidwell <nathan@acm.org>
23078 * gcc.c (driver::maybe_run_linker): Check for input file
23079 accessibility if not linking.
23081 2021-02-05 Richard Biener <rguenther@suse.de>
23083 PR tree-optimization/98855
23084 * tree-vectorizer.h (add_stmt_cost): New overload.
23085 * tree-vect-slp.c (li_cost_vec_cmp): New.
23086 (vect_bb_slp_scalar_cost): Cost individual loop regions
23087 separately. Account for the scalar instance root stmt.
23089 2021-02-05 Tom de Vries <tdevries@suse.de>
23092 * tree-switch-conversion.c (jump_table_cluster::emit): Add loc
23094 (bit_test_cluster::emit): Reuse location_t for newly created
23096 (switch_decision_tree::try_switch_expansion): Preserve
23098 * tree-switch-conversion.h: Change function signatures.
23100 2021-02-05 Jakub Jelinek <jakub@redhat.com>
23103 * config/i386/i386-options.c (m_NONE, m_ALL): Define.
23104 * config/i386/x86-tune.def (X86_TUNE_BRANCH_PREDICTION_HINTS,
23105 X86_TUNE_PROMOTE_QI_REGS): Use m_NONE instead of 0U.
23106 (X86_TUNE_QIMODE_MATH): Use m_ALL instead of ~0U.
23108 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23110 * config/aarch64/aarch64-simd-builtins.def (get_high): Define builtin.
23111 * config/aarch64/aarch64-simd.md (aarch64_get_high<mode>): Define.
23112 * config/aarch64/arm_neon.h (__GET_HIGH): Delete.
23113 (vget_high_f16): Reimplement using new builtin.
23114 (vget_high_f32): Likewise.
23115 (vget_high_f64): Likewise.
23116 (vget_high_p8): Likewise.
23117 (vget_high_p16): Likewise.
23118 (vget_high_p64): Likewise.
23119 (vget_high_s8): Likewise.
23120 (vget_high_s16): Likewise.
23121 (vget_high_s32): Likewise.
23122 (vget_high_s64): Likewise.
23123 (vget_high_u8): Likewise.
23124 (vget_high_u16): Likewise.
23125 (vget_high_u32): Likewise.
23126 (vget_high_u64): Likewise.
23128 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23130 * config/aarch64/aarch64-simd-builtins.def (get_low): Define builtin.
23131 * config/aarch64/aarch64-simd.md (aarch64_get_low<mode>): Define.
23132 * config/aarch64/arm_neon.h (__GET_LOW): Delete.
23133 (vget_low_f16): Reimplement using new builtin.
23134 (vget_low_f32): Likewise.
23135 (vget_low_f64): Likewise.
23136 (vget_low_p8): Likewise.
23137 (vget_low_p16): Likewise.
23138 (vget_low_p64): Likewise.
23139 (vget_low_s8): Likewise.
23140 (vget_low_s16): Likewise.
23141 (vget_low_s32): Likewise.
23142 (vget_low_s64): Likewise.
23143 (vget_low_u8): Likewise.
23144 (vget_low_u16): Likewise.
23145 (vget_low_u32): Likewise.
23146 (vget_low_u64): Likewise.
23148 2021-02-05 Kito Cheng <kito.cheng@sifive.com>
23150 * gcc.c (print_multilib_info): Check all required argument is provided
23153 2021-02-05 liuhongt <hongtao.liu@intel.com>
23156 * config/i386/i386-expand.c (ix86_expand_sse_cmp): Don't
23157 generate integer mask comparison for 128/256-bits vector when
23158 op_true/op_false is NULL_RTX or CONSTM1_RTX/CONST0_RTX. Also
23159 delete redundant !maskcmp condition.
23160 (ix86_expand_int_vec_cmp): Ditto but no redundant deletion
23162 (ix86_expand_sse_movcc): Delete definition of maskcmp, add the
23163 condition directly to if (maskcmp), add extra check for
23164 cmpmode, it should be MODE_INT.
23165 (ix86_expand_fp_vec_cmp): Pass NULL to ix86_expand_sse_cmp's
23166 parameters op_true/op_false.
23167 (ix86_use_mask_cmp_p): New.
23169 2021-02-05 liuhongt <hongtao.liu@intel.com>
23172 * config/i386/x86-tune.def (X86_TUNE_AVX256_UNALIGNED_LOAD_OPTIMAL):
23173 Remove m_GENERIC from ~list.
23174 (X86_TUNE_AVX256_UNALIGNED_STORE_OPTIMAL): Ditto.
23176 2021-02-04 David Malcolm <dmalcolm@redhat.com>
23179 * diagnostic-show-locus.c (compatible_locations_p): Require
23180 locations in the same macro map to be either both from the
23181 macro definition, or both from the macro arguments.
23183 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
23185 * config/aarch64/aarch64-simd-builtins.def: Add
23186 [su]mull_hi_lane[q] builtin generator macros.
23187 * config/aarch64/aarch64-simd.md
23188 (aarch64_<su>mull_hi_lane<mode>_insn): Define.
23189 (aarch64_<su>mull_hi_lane<mode>): Define.
23190 (aarch64_<su>mull_hi_laneq<mode>_insn): Define.
23191 (aarch64_<su>mull_hi_laneq<mode>): Define.
23192 * config/aarch64/arm_neon.h (vmull_high_lane_s16): Use RTL
23193 builtin instead of inline asm.
23194 (vmull_high_lane_s32): Likewise.
23195 (vmull_high_lane_u16): Likewise.
23196 (vmull_high_lane_u32): Likewise.
23197 (vmull_high_laneq_s16): Likewise.
23198 (vmull_high_laneq_s32): Likewise.
23199 (vmull_high_laneq_u16): Likewise.
23200 (vmull_high_laneq_u32): Liekwise.
23202 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
23204 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_hi_n
23205 builtin generator macros.
23206 * config/aarch64/aarch64-simd.md
23207 (aarch64_<su>mull_hi_n<mode>_insn): Define.
23208 (aarch64_<su>mull_hi_n<mode>): Define.
23209 * config/aarch64/arm_neon.h (vmull_high_n_s16): Use RTL builtin
23210 instead of inline asm.
23211 (vmull_high_n_s32): Likewise.
23212 (vmull_high_n_u16): Likewise.
23213 (vmull_high_n_u32): Likewise.
23215 2021-02-04 Richard Biener <rguenther@suse.de>
23217 PR tree-optimization/98855
23218 * tree-vect-loop.c (vectorizable_phi): Do not cost
23219 single-argument PHIs.
23220 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Likewise.
23221 * tree-vect-stmts.c (vectorizable_bswap): Also perform
23222 costing for SLP operation.
23224 2021-02-04 Martin Liska <mliska@suse.cz>
23226 * doc/extend.texi: Mention -mprefer-vector-width in target
23229 2021-02-03 Martin Sebor <msebor@redhat.com>
23231 PR tree-optimization/98937
23232 * tree-ssa-strlen.c (strlen_dom_walker::~strlen_dom_walker): Define.
23233 Flush pointer_query cache.
23235 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
23237 * config/rs6000/genfusion.pl (gen_2logical): Add missing
23238 fixes based on patch review.
23239 * config/rs6000/fusion.md: Regenerate file.
23241 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
23243 * config/rs6000/t-rs6000: Comment out auto generation of
23246 2021-02-03 Andrew Stubbs <ams@codesourcery.com>
23248 * config/gcn/gcn-opts.h (enum processor_type): Add PROCESSOR_GFX908.
23249 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Add gfx908.
23250 (output_file_start): Add gfx908.
23251 * config/gcn/gcn.opt (gpu_type): Add gfx908.
23252 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Add march=gfx908.
23253 (MULTILIB_DIRNAMES): Add gfx908.
23254 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): New define.
23255 (main): Recognize gfx908.
23256 * config/gcn/t-omp-device: Add gfx908.
23258 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
23260 * config/aarch64/aarch64-simd-builtins.def: Add
23261 [su]mlsl_hi_lane[q] builtin macro generators.
23262 * config/aarch64/aarch64-simd.md
23263 (aarch64_<su>mlsl_hi_lane<mode>_insn): Define.
23264 (aarch64_<su>mlsl_hi_lane<mode>): Define.
23265 (aarch64_<su>mlsl_hi_laneq<mode>_insn): Define.
23266 (aarch64_<su>mlsl_hi_laneq<mode>): Define.
23267 * config/aarch64/arm_neon.h (vmlsl_high_lane_s16): Use RTL
23268 builtin instead of inline asm.
23269 (vmlsl_high_lane_s32): Likewise.
23270 (vmlsl_high_lane_u16): Likewise.
23271 (vmlsl_high_lane_u32): Likewise.
23272 (vmlsl_high_laneq_s16): Likewise.
23273 (vmlsl_high_laneq_s32): Likewise.
23274 (vmlsl_high_laneq_u16): Likewise.
23275 (vmlsl_high_laneq_u32): Likewise.
23276 (vmlal_high_laneq_u32): Likewise.
23278 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
23280 * config/aarch64/aarch64-simd-builtins.def: Add
23281 [su]mlal_hi_lane[q] builtin generator macros.
23282 * config/aarch64/aarch64-simd.md
23283 (aarch64_<su>mlal_hi_lane<mode>_insn): Define.
23284 (aarch64_<su>mlal_hi_lane<mode>): Define.
23285 (aarch64_<su>mlal_hi_laneq<mode>_insn): Define.
23286 (aarch64_<su>mlal_hi_laneq<mode>): Define.
23287 * config/aarch64/arm_neon.h (vmlal_high_lane_s16): Use RTL
23288 builtin instead of inline asm.
23289 (vmlal_high_lane_s32): Likewise.
23290 (vmlal_high_lane_u16): Likewise.
23291 (vmlal_high_lane_u32): Likewise.
23292 (vmlal_high_laneq_s16): Likewise.
23293 (vmlal_high_laneq_s32): Likewise.
23294 (vmlal_high_laneq_u16): Likewise.
23295 (vmlal_high_laneq_u32): Likewise.
23297 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
23299 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_hi_n
23300 builtin generator macros.
23301 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_hi_n<mode>_insn):
23303 (aarch64_<su>mlsl_hi_n<mode>): Define.
23304 * config/aarch64/arm_neon.h (vmlsl_high_n_s16): Use RTL builtin
23305 instead of inline asm.
23306 (vmlsl_high_n_s32): Likewise.
23307 (vmlsl_high_n_u16): Likewise.
23308 (vmlsl_high_n_u32): Likewise.
23310 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
23312 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_hi_n
23313 builtin generator macros.
23314 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_hi_n<mode>_insn):
23316 (aarch64_<su>mlal_hi_n<mode>): Define.
23317 * config/aarch64/arm_neon.h (vmlal_high_n_s16): Use RTL builtin
23318 instead of inline asm.
23319 (vmlal_high_n_s32): Likewise.
23320 (vmlal_high_n_u16): Likewise.
23321 (vmlal_high_n_u32): Likewise.
23323 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
23325 * config/aarch64/aarch64-simd-builtins.def: Add RTL builtin
23327 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal_hi<mode>):
23329 (aarch64_<su>mlal_hi<mode>_insn): This.
23330 (aarch64_<su>mlal_hi<mode>): Define.
23331 * config/aarch64/arm_neon.h (vmlal_high_s8): Use RTL builtin
23332 instead of inline asm.
23333 (vmlal_high_s16): Likewise.
23334 (vmlal_high_s32): Likewise.
23335 (vmlal_high_u8): Likewise.
23336 (vmlal_high_u16): Likewise.
23337 (vmlal_high_u32): Likewise.
23339 2021-02-03 Ilya Leoshkevich <iii@linux.ibm.com>
23341 * lra-spills.c (remove_pseudos): Call lra_update_insn_recog_data()
23342 after calling alter_subreg() on a (mem).
23344 2021-02-03 Martin Liska <mliska@suse.cz>
23347 * lto-streamer-out.c (produce_lto_section): Fill up missing
23349 * lto-streamer.h (struct lto_section): Add _padding field.
23351 2021-02-03 Richard Biener <rguenther@suse.de>
23353 * lto-streamer.c (lto_get_section_name): Free temporary
23355 * tree-loop-distribution.c
23356 (loop_distribution::merge_dep_scc_partitions): Free edge data.
23358 2021-02-03 Jakub Jelinek <jakub@redhat.com>
23360 PR middle-end/97487
23361 * ifcvt.c (noce_can_force_operand): New function.
23362 (noce_emit_move_insn): Use it.
23363 (noce_try_sign_mask): Likewise. Formatting fix.
23365 2021-02-03 Jakub Jelinek <jakub@redhat.com>
23367 PR middle-end/97971
23368 * lra-constraints.c (process_alt_operands): For inline asm, don't call
23369 fatal_insn, but instead return false.
23371 2021-02-03 Jakub Jelinek <jakub@redhat.com>
23373 PR tree-optimization/98287
23374 * config/i386/mmx.md (<insn><mode>3): For shifts don't enable expander
23377 2021-02-03 Tamar Christina <tamar.christina@arm.com>
23379 PR tree-optimization/98928
23380 * tree-vect-loop.c (vect_analyze_loop_2): Change
23381 STMT_VINFO_SLP_VECT_ONLY to STMT_VINFO_SLP_VECT_ONLY_PATTERN.
23382 * tree-vect-slp-patterns.c (complex_pattern::build): Likewise.
23383 * tree-vectorizer.h (STMT_VINFO_SLP_VECT_ONLY_PATTERN): New.
23384 (class _stmt_vec_info): Add slp_vect_pattern_only_p.
23386 2021-02-02 Richard Biener <rguenther@suse.de>
23388 * gimple-loop-interchange.cc (prepare_data_references):
23390 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
23391 * tree-ssa-loop-im.c (hoist_memory_references): Likewise.
23392 * tree-vect-stmts.c (vectorizable_condition): Do not
23394 (vectorizable_comparison): Likewise.
23396 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23398 * config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
23399 * config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
23400 * config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
23401 (vrsqrteq_u32): Likewise.
23403 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23405 * config/aarch64/aarch64-simd-builtins.def (sqxtun2): Define builtin.
23406 * config/aarch64/aarch64-simd.md (aarch64_sqxtun2<mode>_le): Define.
23407 (aarch64_sqxtun2<mode>_be): Likewise.
23408 (aarch64_sqxtun2<mode>): Likewise.
23409 * config/aarch64/arm_neon.h (vqmovun_high_s16): Reimplement using builtin.
23410 (vqmovun_high_s32): Likewise.
23411 (vqmovun_high_s64): Likewise.
23412 * config/aarch64/iterators.md (UNSPEC_SQXTUN2): Define.
23414 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23416 * config/aarch64/aarch64-simd-builtins.def (bfdot_lane, bfdot_laneq): Use
23418 (bfmlalb_lane, bfmlalt_lane, bfmlalb_lane_q, bfmlalt_lane_q): Use FP flags.
23420 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23422 * config/aarch64/aarch64-simd-builtins.def (fcmla_lane0, fcmla_lane90,
23423 fcmla_lane180, fcmla_lane270, fcmlaq_lane0, fcmlaq_lane90, fcmlaq_lane180,
23424 fcmlaq_lane270, scvtf, ucvtf, fcvtzs, fcvtzu, scvtfsi, scvtfdi, ucvtfsi,
23425 ucvtfdi, fcvtzshf, fcvtzuhf, fmlal_lane_low, fmlsl_lane_low,
23426 fmlal_laneq_low, fmlsl_laneq_low, fmlalq_lane_low, fmlslq_lane_low,
23427 fmlalq_laneq_low, fmlslq_laneq_low, fmlal_lane_high, fmlsl_lane_high,
23428 fmlal_laneq_high, fmlsl_laneq_high, fmlalq_lane_high, fmlslq_lane_high,
23429 fmlalq_laneq_high, fmlslq_laneq_high): Use FP flags.
23431 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23433 * config/aarch64/aarch64-builtins.c (FLAG_LOAD): Define.
23434 * config/aarch64/aarch64-simd-builtins.def (ld1x2, ld2, ld3, ld4, ld2r,
23435 ld3r, ld4r, ld1, ld1x3, ld1x4): Use LOAD flags.
23437 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23439 * config/aarch64/aarch64-simd-builtins.def (combine, zip1, zip2,
23440 uzp1, uzp2, trn1, trn2, simd_bsl): Use AUTO_FP flags.
23442 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23444 * config/aarch64/aarch64-simd-builtins.def (clrsb, clz, ctz, popcount,
23445 vec_smult_lane_, vec_smlal_lane_, vec_smult_laneq_, vec_smlal_laneq_,
23446 vec_umult_lane_, vec_umlal_lane_, vec_umult_laneq_, vec_umlal_laneq_,
23447 ashl, sshl, ushl, srshl, urshl, sdot_lane, udot_lane, sdot_laneq,
23448 udot_laneq, usdot_lane, usdot_laneq, sudot_lane, sudot_laneq, ashr,
23449 ashr_simd, lshr, lshr_simd, srshr_n, urshr_n, ssra_n, usra_n, srsra_n,
23450 ursra_n, sshll_n, ushll_n, sshll2_n, ushll2_n, ssri_n, usri_n, ssli_n,
23451 ssli_n, usli_n, bswap, rbit, simd_bsl, eor3q, rax1q, xarq, bcaxq): Use
23452 NONE builtin flags.
23454 2021-02-02 Jakub Jelinek <jakub@redhat.com>
23456 PR tree-optimization/98848
23457 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
23458 STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
23460 2021-02-02 Kito Cheng <kito.cheng@sifive.com>
23463 * expr.c: Check mode before calling store_expr.
23465 2021-02-02 Christophe Lyon <christophe.lyon@linaro.org>
23467 * config/arm/iterators.md (supf): Remove VORNQ_S and VORNQ_U.
23469 * config/arm/mve.md (mve_vornq_s<mode>): New entry for vorn
23470 instruction using expression ior.
23471 (mve_vornq_u<mode>): New expander.
23472 (mve_vornq_f<mode>): Use ior code instead of unspec.
23473 * config/arm/unspecs.md (VORNQ_S, VORNQ_U, VORNQ_F): Remove.
23475 2021-02-02 Alexandre Oliva <oliva@adacore.com>
23477 * tree-nested.c (convert_nonlocal_reference_op): Move
23478 current_function_decl restore after re-gimplification.
23479 (convert_local_reference_op): Likewise.
23481 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23483 * config/aarch64/aarch64-simd-builtins.def (rshrn, rshrn2):
23485 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>_insn_le):
23487 (aarch64_rshrn<mode>_insn_be): Likewise.
23488 (aarch64_rshrn<mode>): Likewise.
23489 (aarch64_rshrn2<mode>_insn_le): Likewise.
23490 (aarch64_rshrn2<mode>_insn_be): Likewise.
23491 (aarch64_rshrn2<mode>): Likewise.
23492 * config/aarch64/aarch64.md (unspec): Add UNSPEC_RSHRN.
23493 * config/aarch64/arm_neon.h (vrshrn_high_n_s16): Reimplement
23495 (vrshrn_high_n_s32): Likewise.
23496 (vrshrn_high_n_s64): Likewise.
23497 (vrshrn_high_n_u16): Likewise.
23498 (vrshrn_high_n_u32): Likewise.
23499 (vrshrn_high_n_u64): Likewise.
23500 (vrshrn_n_s16): Likewise.
23501 (vrshrn_n_s32): Likewise.
23502 (vrshrn_n_s64): Likewise.
23503 (vrshrn_n_u16): Likewise.
23504 (vrshrn_n_u32): Likewise.
23505 (vrshrn_n_u64): Likewise.
23507 2021-02-01 Sergei Trofimovich <siarheit@google.com>
23509 PR tree-optimization/98499
23510 * ipa-modref.c (analyze_ssa_name_flags): treat RVO
23511 conservatively and assume all possible side-effects.
23513 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23515 * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi,
23516 vec_unpacku_hi_): Define builtins.
23517 * config/aarch64/arm_neon.h (vmovl_high_s8): Reimplement using
23519 (vmovl_high_s16): Likewise.
23520 (vmovl_high_s32): Likewise.
23521 (vmovl_high_u8): Likewise.
23522 (vmovl_high_u16): Likewise.
23523 (vmovl_high_u32): Likewise.
23525 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23527 * config/aarch64/aarch64-simd-builtins.def (sabdl, uabdl):
23529 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl<mode>): New
23531 * config/aarch64/aarch64.md (unspec): Define UNSPEC_SABDL,
23533 * config/aarch64/arm_neon.h (vabdl_s8): Reimplemet using
23535 (vabdl_s16): Likewise.
23536 (vabdl_s32): Likewise.
23537 (vabdl_u8): Likewise.
23538 (vabdl_u16): Likewise.
23539 (vabdl_u32): Likewise.
23540 * config/aarch64/iterators.md (ABDL): New int iterator.
23541 (sur): Handle UNSPEC_SABDL, UNSPEC_UABDL.
23543 2021-02-01 Martin Sebor <msebor@redhat.com>
23545 * tree.h (BLOCK_VARS): Add comment.
23546 (BLOCK_SUBBLOCKS): Same.
23547 (BLOCK_SUPERCONTEXT): Same.
23548 (BLOCK_ABSTRACT_ORIGIN): Same.
23549 (inlined_function_outer_scope_p): Same.
23551 2021-02-01 Martin Sebor <msebor@redhat.com>
23553 PR middle-end/97172
23554 * attribs.c (attr_access::free_lang_data): Define new function.
23555 * attribs.h (attr_access::free_lang_data): Declare new function.
23557 2021-02-01 Richard Biener <rguenther@suse.de>
23559 * vec.h (auto_vec::auto_vec): Add memory stat parameters
23561 * bitmap.h (auto_bitmap::auto_bitmap): Likewise.
23563 2021-02-01 Tamar Christina <tamar.christina@arm.com>
23565 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>,
23566 aarch64_<su>mlsl<mode>, aarch64_<su>mlsl_n<mode>): Flip mult operands.
23568 2021-02-01 Richard Biener <rguenther@suse.de>
23570 PR rtl-optimization/98863
23571 * config/i386/i386-features.c (convert_scalars_to_vector):
23572 Set DF_RD_PRUNE_DEAD_DEFS.
23574 2021-01-31 Eric Botcazou <ebotcazou@adacore.com>
23576 * system.h (SIZE_MAX): Define if not already defined.
23578 2021-01-30 Aaron Sawdey <acsawdey@linux.ibm.com>
23580 * config/rs6000/genfusion.pl (gen_2logical): New function to
23581 generate patterns for logical-logical fusion.
23582 * config/rs6000/fusion.md: Regenerated patterns.
23583 * config/rs6000/rs6000-cpus.def: Add
23584 OPTION_MASK_P10_FUSION_2LOGICAL.
23585 * config/rs6000/rs6000.c (rs6000_option_override_internal):
23586 Enable logical-logical fusion for p10.
23587 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2logical.
23589 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
23591 * config/rs6000/rs6000.opt: Add periods to new AIX options.
23593 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
23595 * config/rs6000/rs6000.opt (mabi=vec-extabi): New.
23596 (mabi=vec-default): New.
23597 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
23598 __EXTABI__ for AIX Vector extended ABI.
23599 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print AIX Vector
23601 (conditional_register_usage): If AIX vec_extabi enabled, vs20-vs31
23603 * doc/invoke.texi (PowerPC mabi): Add AIX vec-extabi and vec-default.
23605 2021-01-30 Jakub Jelinek <jakub@redhat.com>
23607 * config/i386/i386-features.c (remove_partial_avx_dependency): Clear
23608 DF_DEFER_INSN_RESCAN after calling df_process_deferred_rescans.
23610 2021-01-29 Vladimir N. Makarov <vmakarov@redhat.com>
23613 * lra-constraints.c (in_class_p): Don't narrow class only for REG
23616 2021-01-29 Will Schmidt <will_schmidt@vnet.ibm.com>
23618 * config/rs6000/rs6000-call.c (rs6000_expand_binup_builtin): Add
23619 clauses for CODE_FOR_vsx_xvcvuxddp_scale and
23620 CODE_FOR_vsx_xvcvsxddp_scale to the parameter checking code.
23622 2021-01-29 Andrew MacLeod <amacleod@redhat.com>
23624 PR tree-optimization/98866
23625 * gimple-range-gori.h (gori_compute:set_range_invariant): New.
23626 * gimple-range-gori.cc (gori_map::set_range_invariant): New.
23627 (gori_map::m_maybe_invariant): Rename from all_outgoing.
23628 (gori_map::gori_map): Rename all_outgoing to m_maybe_invariant.
23629 (gori_map::is_export_p): Ditto.
23630 (gori_map::calculate_gori): Ditto.
23631 (gori_compute::set_range_invariant): New.
23632 * gimple-range.cc (gimple_ranger::range_of_stmt): Set range
23633 invariant for pointers evaluating to [1, +INF].
23635 2021-01-29 Richard Biener <rguenther@suse.de>
23637 PR rtl-optimization/98863
23638 * config/i386/i386-features.c (remove_partial_avx_dependency):
23639 Do not perform DF analysis.
23640 (pass_data_remove_partial_avx_dependency): Remove
23643 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
23645 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_n
23646 builtin generator macros.
23647 * config/aarch64/aarch64-simd.md (aarch64_<su>mull_n<mode>):
23649 * config/aarch64/arm_neon.h (vmull_n_s16): Use RTL builtin
23650 instead of inline asm.
23651 (vmull_n_s32): Likewise.
23652 (vmull_n_u16): Likewise.
23653 (vmull_n_u32): Likewise.
23655 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23657 * config/aarch64/aarch64-simd-builtins.def (sabdl2, uabdl2):
23659 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
23661 (aarch64_<sur>abdl2<mode>): ... This.
23662 (<sur>sadv16qi): Adjust use of above.
23663 * config/aarch64/arm_neon.h (vabdl_high_s8): Reimplement using
23665 (vabdl_high_s16): Likewise.
23666 (vabdl_high_s32): Likewise.
23667 (vabdl_high_u8): Likewise.
23668 (vabdl_high_u16): Likewise.
23669 (vabdl_high_u32): Likewise.
23671 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23673 * config/aarch64/aarch64-simd-builtins.def (sabal2): Define
23675 (uabal2): Likewise.
23676 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal2<mode>): New
23678 * config/aarch64/aarch64.md (unspec): Add UNSPEC_SABAL2 and
23680 * config/aarch64/arm_neon.h (vabal_high_s8): Reimplement using
23682 (vabal_high_s16): Likewise.
23683 (vabal_high_s32): Likewise.
23684 (vabal_high_u8): Likewise.
23685 (vabal_high_u16): Likewise.
23686 (vabal_high_u32): Likewise.
23687 * config/aarch64/iterators.md (ABAL2): New mode iterator.
23688 (sur): Handle UNSPEC_SABAL2, UNSPEC_UABAL2.
23690 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23692 * config/aarch64/aarch64-simd-builtins.def (sabal): Define
23695 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal<mode>_4):
23697 (aarch64_<sur>abal<mode>): ... This
23698 (<sur>sadv16qi): Adust use of the above.
23699 * config/aarch64/arm_neon.h (vabal_s8): Reimplement using
23701 (vabal_s16): Likewise.
23702 (vabal_s32): Likewise.
23703 (vabal_u8): Likewise.
23704 (vabal_u16): Likewise.
23705 (vabal_u32): Likewise.
23707 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23709 * config/aarch64/aarch64-simd-builtins.def (saddlv, uaddlv):
23711 * config/aarch64/aarch64-simd.md (aarch64_<su>addlv<mode>):
23713 * config/aarch64/arm_neon.h (vaddlv_s8): Reimplement using
23715 (vaddlv_s16): Likewise.
23716 (vaddlv_u8): Likewise.
23717 (vaddlv_u16): Likewise.
23718 (vaddlvq_s8): Likewise.
23719 (vaddlvq_s16): Likewise.
23720 (vaddlvq_s32): Likewise.
23721 (vaddlvq_u8): Likewise.
23722 (vaddlvq_u16): Likewise.
23723 (vaddlvq_u32): Likewise.
23724 (vaddlv_s32): Likewise.
23725 (vaddlv_u32): Likewise.
23726 * config/aarch64/iterators.md (VDQV_L): New mode iterator.
23727 (unspec): Add UNSPEC_SADDLV, UNSPEC_UADDLV.
23728 (Vwstype): New mode attribute.
23730 (VWIDE_S): Likewise.
23731 (USADDLV): New int iterator.
23732 (su): Handle UNSPEC_SADDLV, UNSPEC_UADDLV.
23734 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
23736 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_lane[q]
23737 builtin generator macros.
23738 * config/aarch64/aarch64-simd.md (aarch64_vec_<su>mlsl_lane<Qlane>):
23740 * config/aarch64/arm_neon.h (vmlsl_lane_s16): Use RTL builtin
23741 instead of inline asm.
23742 (vmlsl_lane_s32): Likewise.
23743 (vmlsl_lane_u16): Likewise.
23744 (vmlsl_lane_u32): Likewise.
23745 (vmlsl_laneq_s16): Likewise.
23746 (vmlsl_laneq_s32): Likewise.
23747 (vmlsl_laneq_u16): Likewise.
23748 (vmlsl_laneq_u32): Likewise.
23750 2021-01-29 Richard Biener <rguenther@suse.de>
23752 * doc/invoke.texi (--param max-gcse-memory): Document unit
23754 * gcse.c (gcse_or_cprop_is_too_expensive): Adjust.
23755 * params.opt (--param max-gcse-memory): Adjust default and
23756 document unit of size.
23758 2021-01-29 Richard Biener <rguenther@suse.de>
23760 PR rtl-optimization/98863
23761 * gcse.c (gcse_or_cprop_is_too_expensive): Use unsigned
23762 HOST_WIDE_INT for the memory estimate.
23764 2021-01-29 Bin Cheng <bin.cheng@linux.alibaba.com>
23765 Richard Biener <rguenther@suse.de>
23767 PR tree-optimization/97627
23768 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
23769 Do not analyze fake edges.
23771 2021-01-29 Richard Biener <rguenther@suse.de>
23773 PR rtl-optimization/98144
23774 * df.h (df_mir_bb_info): Add con_visited member.
23775 * df-problems.c (df_mir_alloc): Initialize con_visited,
23776 do not fully populate IN and OUT.
23777 (df_mir_reset): Likewise.
23778 (df_mir_confluence_0): Set con_visited.
23779 (df_mir_confluence_n): Properly handle implicitely
23780 fully populated IN and OUT as designated by con_visited
23781 and update con_visited accordingly.
23783 2021-01-29 Jakub Jelinek <jakub@redhat.com>
23786 * config/arm/vec-common.md (mve_vshlq_<supf><mode>,
23787 vashl<mode>3, vashr<mode>3, vlshr<mode>3): Add
23788 && !TARGET_REALLY_IWMMXT to conditions.
23790 2021-01-29 Jakub Jelinek <jakub@redhat.com>
23793 * cfgbuild.c (find_bb_boundaries): Reset debug_insn when seeing
23796 2021-01-28 Marek Polacek <polacek@redhat.com>
23799 * stor-layout.c (finalize_type_size): If we reset TYPE_USER_ALIGN in
23800 the main variant, maybe reset it in its variants too.
23801 * tree.c (check_base_type): Return true only if TYPE_USER_ALIGN match.
23802 (check_aligned_type): Check if TYPE_USER_ALIGN match.
23804 2021-01-28 Christophe Lyon <christophe.lyon@linaro.org>
23807 * config/arm/arm.c (arm_rtx_costs_internal): Adjust cost of vector
23808 of constant zero for comparisons.
23810 2021-01-28 Michael Meissner <meissner@linux.ibm.com>
23812 * config/rs6000/rs6000.c (rs6000_mangle_decl_assembler_name): Add
23813 support for mapping built-in function names for long double
23814 built-in functions if long double is IEEE 128-bit.
23816 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
23818 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_n
23819 builtin generator macros.
23820 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_n<mode>):
23822 * config/aarch64/arm_neon.h (vmlsl_n_s16): Use RTL builtin
23823 instead of inline asm.
23824 (vmlsl_n_s32): Likewise.
23825 (vmlsl_n_u16): Likewise.
23826 (vmlsl_n_u32): Likewise.
23828 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
23830 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_n
23831 builtin generator macros.
23832 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>):
23834 * config/aarch64/arm_neon.h (vmlal_n_s16): Use RTL builtin
23835 instead of inline asm.
23836 (vmlal_n_s32): Likewise.
23837 (vmlal_n_u16): Likewise.
23838 (vmlal_n_u32): Likewise.
23840 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23842 * config/aarch64/aarch64-simd-builtins.def (shrn2): Define
23844 * config/aarch64/aarch64-simd.md (aarch64_shrn2<mode>_insn_le):
23846 (aarch64_shrn2<mode>_insn_be): Likewise.
23847 (aarch64_shrn2<mode>): Likewise.
23848 * config/aarch64/arm_neon.h (vshrn_high_n_s16): Reimlplement
23850 (vshrn_high_n_s32): Likewise.
23851 (vshrn_high_n_s64): Likewise.
23852 (vshrn_high_n_u16): Likewise.
23853 (vshrn_high_n_u32): Likewise.
23854 (vshrn_high_n_u64): Likewise.
23856 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23858 * config/aarch64/aarch64-simd-builtins.def (shrn): Define
23860 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le):
23862 (aarch64_shrn<mode>_insn_be): Likewise.
23863 (aarch64_shrn<mode>): Likewise.
23864 * config/aarch64/arm_neon.h (vshrn_n_s16): Reimplement using
23866 (vshrn_n_s32): Likewise.
23867 (vshrn_n_s64): Likewise.
23868 (vshrn_n_u16): Likewise.
23869 (vshrn_n_u32): Likewise.
23870 (vshrn_n_u64): Likewise.
23871 * config/aarch64/iterators.md (vn_mode): New mode attribute.
23873 2021-01-28 Richard Biener <rguenther@suse.de>
23875 PR rtl-optimization/80960
23876 * dse.c (check_mem_read_rtx): Call get_addr on the
23879 2021-01-28 Xionghu Luo <luoxhu@linux.ibm.com>
23880 David Edelsohn <dje.gcc@gmail.com>
23883 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
23884 Don't generate VIEW_CONVERT_EXPR for fcode ALTIVEC_BUILTIN_VEC_INSERT
23886 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
23888 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Remove the
23889 wrapper call rs6000_expand_vector_set_var for cleanup. Call
23890 rs6000_expand_vector_set_var_p9 and rs6000_expand_vector_set_var_p8
23892 (rs6000_expand_vector_set_var): Delete.
23893 (rs6000_expand_vector_set_var_p9): Make static.
23894 (rs6000_expand_vector_set_var_p8): Make static.
23896 2021-01-28 Xing GUO <higuoxing@gmail.com>
23898 * common/config/riscv/riscv-common.c
23899 (riscv_subset_list::parsing_subset_version): Fix -march option parsing
23900 when `p` extension exists.
23902 2021-01-27 Vladimir N. Makarov <vmakarov@redhat.com>
23904 PR rtl-optimization/97684
23905 * ira.c (ira): Call ira_set_pseudo_classes before
23906 update_equiv_regs when it is necessary.
23908 2021-01-27 Jakub Jelinek <jakub@redhat.com>
23911 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Use
23912 %w0, %w1 and %2 instead of %0, %1 and %2.
23914 2021-01-27 Aaron Sawdey <acsawdey@linux.ibm.com>
23916 * config/rs6000/genfusion.pl: New script to generate
23917 define_insn_and_split patterns so combine can arrange fused
23918 instructions next to each other.
23919 * config/rs6000/fusion.md: New file, generated fused instruction
23920 patterns for combine.
23921 * config/rs6000/predicates.md (const_m1_to_1_operand): New predicate.
23922 (non_update_memory_operand): New predicate.
23923 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_P10_FUSION and
23924 OPTION_MASK_P10_FUSION_LD_CMPI to ISA_3_1_MASKS_SERVER and
23926 * config/rs6000/rs6000-protos.h (address_is_non_pfx_d_or_x): Add
23928 * config/rs6000/rs6000.c (rs6000_option_override_internal):
23929 Automatically set OPTION_MASK_P10_FUSION and
23930 OPTION_MASK_P10_FUSION_LD_CMPI if target is power10.
23931 (rs600_opt_masks): Allow -mpower10-fusion
23932 in function attributes.
23933 (address_is_non_pfx_d_or_x): New function.
23934 * config/rs6000/rs6000.h: Add MASK_P10_FUSION.
23935 * config/rs6000/rs6000.md: Include fusion.md.
23936 * config/rs6000/rs6000.opt: Add -mpower10-fusion
23937 and -mpower10-fusion-ld-cmpi.
23938 * config/rs6000/t-rs6000: Add dependencies involving fusion.md.
23940 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
23942 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal
23943 builtin generator macros.
23944 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal<mode>):
23946 (aarch64_<su>mlal<mode>): This.
23947 * config/aarch64/arm_neon.h (vmlal_s8): Use RTL builtin
23948 instead of inline asm.
23949 (vmlal_s16): Likewise.
23950 (vmlal_s32): Likewise.
23951 (vmlal_u8): Likewise.
23952 (vmlal_u16): Likewise.
23953 (vmlal_u32): Likewise.
23955 2021-01-27 Richard Biener <rguenther@suse.de>
23957 PR tree-optimization/98854
23958 * tree-vect-slp.c (vect_build_slp_tree_2): Also build
23959 PHIs from scalars when the number of CTORs matches the
23960 number of children.
23962 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
23964 * config/aarch64/aarch64-simd-builtins.def: Add mls_n builtin
23966 * config/aarch64/aarch64-simd.md (*aarch64_mls_elt_merge<mode>):
23968 (aarch64_mls_n<mode>): This.
23969 * config/aarch64/arm_neon.h (vmls_n_s16): Use RTL builtin
23971 (vmls_n_s32): Likewise.
23972 (vmls_n_u16): Likewise.
23973 (vmls_n_u32): Likewise.
23974 (vmlsq_n_s16): Likewise.
23975 (vmlsq_n_s32): Likewise.
23976 (vmlsq_n_u16): Likewise.
23977 (vmlsq_n_u32): Likewise.
23979 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
23981 * config/aarch64/aarch64-simd-builtins.def: Add mls builtin
23983 * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather
23985 (vmls_s16): Likewise.
23986 (vmls_s32): Likewise.
23987 (vmls_u8): Likewise.
23988 (vmls_u16): Likewise.
23989 (vmls_u32): Likewise.
23990 (vmlsq_s8): Likewise.
23991 (vmlsq_s16): Likewise.
23992 (vmlsq_s32): Likewise.
23993 (vmlsq_u8): Likewise.
23994 (vmlsq_u16): Likewise.
23995 (vmlsq_u32): Likewise.
23997 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
23999 * config/aarch64/aarch64-simd-builtins.def: Add mla_n builtin
24001 * config/aarch64/aarch64-simd.md (*aarch64_mla_elt_merge<mode>):
24003 (aarch64_mla_n<mode>): This.
24004 * config/aarch64/arm_neon.h (vmla_n_s16): Use RTL builtin
24006 (vmla_n_s32): Likewise.
24007 (vmla_n_u16): Likewise.
24008 (vmla_n_u32): Likewise.
24009 (vmlaq_n_s16): Likewise.
24010 (vmlaq_n_s32): Likewise.
24011 (vmlaq_n_u16): Likewise.
24012 (vmlaq_n_u32): Likewise.
24014 2021-01-27 liuhongt <hongtao.liu@intel.com>
24017 * config/i386/sse.md (sse2_gt<mode>3): Drop !TARGET_XOP in condition.
24018 (*sse2_eq<mode>3): Ditto.
24020 2021-01-27 Jakub Jelinek <jakub@redhat.com>
24022 * tree-pass.h (PROP_trees): Rename to ...
24023 (PROP_gimple): ... this.
24024 * cfgexpand.c (pass_data_expand): Replace PROP_trees with PROP_gimple.
24025 * passes.c (execute_function_dump, execute_function_todo,
24026 execute_one_ipa_transform_pass, execute_one_pass): Likewise.
24027 * varpool.c (ctor_for_folding): Likewise.
24029 2021-01-27 Jakub Jelinek <jakub@redhat.com>
24031 PR tree-optimization/97260
24032 * varpool.c: Include tree-pass.h.
24033 (ctor_for_folding): In GENERIC return DECL_INITIAL for TREE_READONLY
24034 non-TREE_SIDE_EFFECTS automatic variables.
24036 2021-01-26 Paul Fee <paul.f.fee@gmail.com>
24038 * doc/cpp.texi (__cplusplus): Document value for -std=c++23
24040 * doc/invoke.texi: Document -std=c++23 and -std=gnu++23.
24041 * dwarf2out.c (highest_c_language): Recognise C++20 and C++23.
24042 (gen_compile_unit_die): Recognise C++23.
24044 2021-01-26 Jakub Jelinek <jakub@redhat.com>
24047 * dwarf2asm.c (dw2_assemble_integer): Cast DWARF2_ADDR_SIZE to int
24050 2021-01-26 Jakub Jelinek <jakub@redhat.com>
24053 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
24054 Use UINTVAL (shft_amnt) and UINTVAL (mask) instead of INTVAL (shft_amnt)
24055 and INTVAL (mask). Add && INTVAL (mask) > 0 condition.
24057 2021-01-26 Richard Biener <rguenther@suse.de>
24059 * gimple-pretty-print.c (dump_binary_rhs): Handle
24060 VEC_WIDEN_{PLUS,MINUS}_{LO,HI}_EXPR.
24062 2021-01-26 Richard Biener <rguenther@suse.de>
24064 PR middle-end/98726
24065 * tree.h (vector_cst_int_elt): Remove.
24066 * tree.c (vector_cst_int_elt): Use poly_wide_int for computations,
24069 2021-01-26 Andrew Stubbs <ams@codesourcery.com>
24071 * config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
24072 for V64DFmode min/max reductions.
24074 2021-01-26 Jakub Jelinek <jakub@redhat.com>
24076 * dwarf2asm.c (dw2_assemble_integer): Handle size twice as large
24077 as DWARF2_ADDR_SIZE if x is not a scalar int by emitting it as
24078 two halves, one with x and the other with const0_rtx, ordered
24079 depending on endianity.
24081 2021-01-26 Alexandre Oliva <oliva@adacore.com>
24083 * gimplify.c (gimplify_decl_expr): Skip asan marking calls for
24084 temporaries not seen in binding block, and not about to be
24085 added as gimple variables.
24087 2021-01-25 Martin Sebor <msebor@redhat.com>
24090 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust warning text.
24092 2021-01-25 Martin Liska <mliska@suse.cz>
24094 * value-prof.c (get_nth_most_common_value): Use %s instead
24097 2021-01-25 Jakub Jelinek <jakub@redhat.com>
24100 * configure.ac (HAVE_AS_GDWARF_5_DEBUG_FLAG): Only define if
24101 readelf -wi is able to read the emitted .debug_info back.
24102 * configure: Regenerated.
24104 2021-01-25 Martin Liska <mliska@suse.cz>
24106 PR gcov-profile/98739
24107 * common.opt: Add missing sign symbol.
24108 * value-prof.c (get_nth_most_common_value): Restore handling
24109 of PROFILE_REPRODUCIBILITY_PARALLEL_RUNS and
24110 PROFILE_REPRODUCIBILITY_MULTITHREADED.
24112 2021-01-25 Richard Biener <rguenther@suse.de>
24114 PR middle-end/98807
24115 * tree.c (vector_element_bits): Always use precision of
24116 the element type for boolean vectors.
24118 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
24120 * config/rtems.h (STARTFILE_SPEC): Remove qnolinkcmds.
24121 (ENDFILE_SPEC): Evaluate qnolinkcmds.
24123 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
24125 * config/rtems.h (STARTFILE_SPEC): Remove nostdlib and
24126 nostartfiles handling since this is already done by
24127 LINK_COMMAND_SPEC. Evaluate qnolinkcmds.
24128 (ENDFILE_SPEC): Remove nostdlib and nostartfiles handling since this
24129 is already done by LINK_COMMAND_SPEC.
24130 (LIB_SPECS): Remove nostdlib and nodefaultlibs handling since
24131 this is already done by LINK_COMMAND_SPEC. Remove qnolinkcmds
24134 2021-01-25 Jakub Jelinek <jakub@redhat.com>
24137 * fold-const-call.c (host_size_t_cst_p): Renamed to ...
24138 (size_t_cst_p): ... this. Check and store unsigned HOST_WIDE_INT
24139 value rather than host size_t.
24140 (fold_const_call): Change type of s2 from size_t to
24141 unsigned HOST_WIDE_INT. Use size_t_cst_p instead of
24142 host_size_t_cst_p. For strncmp calls, pass MIN (s2, SIZE_MAX)
24143 instead of s2 as last argument.
24145 2021-01-25 Tamar Christina <tamar.christina@arm.com>
24147 * config/arm/iterators.md (rotsplit1, rotsplit2, conj_op, fcmac1,
24148 VCMLA_OP, VCMUL_OP): New.
24149 * config/arm/mve.md (mve_vcmlaq<mve_rot><mode>): Support vec_dup 0.
24150 * config/arm/neon.md (cmul<conj_op><mode>3): New.
24151 * config/arm/unspecs.md (UNSPEC_VCMLA_CONJ, UNSPEC_VCMLA180_CONJ,
24152 UNSPEC_VCMUL_CONJ): New.
24153 * config/arm/vec-common.md (cmul<conj_op><mode>3, arm_vcmla<rot><mode>,
24154 cml<fcmac1><conj_op><mode>4): New.
24156 2021-01-23 Jakub Jelinek <jakub@redhat.com>
24159 * config/rs6000/mmintrin.h (__m64): Add __may_alias__ attribute.
24161 2021-01-22 Jonathan Wright <jonathan.wright@arm.com>
24163 * config/aarch64/aarch64-simd-builtins.def: Add mla builtin
24165 * config/aarch64/arm_neon.h (vmla_s8): Use RTL builtin rather
24167 (vmla_s16): Likewise.
24168 (vmla_s32): Likewise.
24169 (vmla_u8): Likewise.
24170 (vmla_u16): Likewise.
24171 (vmla_u32): Likewise.
24172 (vmlaq_s8): Likewise.
24173 (vmlaq_s16): Likewise.
24174 (vmlaq_s32): Likewise.
24175 (vmlaq_u8): Likewise.
24176 (vmlaq_u16): Likewise.
24177 (vmlaq_u32): Likewise.
24179 2021-01-22 David Malcolm <dmalcolm@redhat.com>
24181 * doc/invoke.texi (GCC_EXTRA_DIAGNOSTIC_OUTPUT): Add @findex
24184 2021-01-22 Jakub Jelinek <jakub@redhat.com>
24187 * dwarf2out.c (output_file_names): For -gdwarf-5, if there are no
24188 filenames to emit, still emit the required 0 index directory and
24189 filename entries that match DW_AT_comp_dir and DW_AT_name of the
24192 2021-01-22 Marek Polacek <polacek@redhat.com>
24195 * doc/invoke.texi: Update C++ ABI Version 15 description.
24197 2021-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24199 PR tree-optimization/98766
24200 * tree-ssa-math-opts.c (convert_mult_to_fma): Use maybe_le when
24201 comparing against type size with param_avoid_fma_max_bits.
24203 2021-01-22 Richard Biener <rguenther@suse.de>
24205 PR middle-end/98793
24206 * tree.c (vector_element_bits): Key single-bit bool vector on
24207 integer mode rather than not vector mode.
24209 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
24212 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
24213 Generate ARRAY_REF(VIEW_CONVERT_EXPR) for P8 and later
24215 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): Update
24216 to call different path for P8 and P9.
24217 (rs6000_expand_vector_set_var_p9): New function.
24218 (rs6000_expand_vector_set_var_p8): New function.
24220 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
24224 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
24225 Ajdust variable index vec_insert from address dereference to
24226 ARRAY_REF(VIEW_CONVERT_EXPR) tree expression.
24227 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
24229 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): New function.
24231 2021-01-22 Martin Liska <mliska@suse.cz>
24233 PR gcov-profile/98739
24234 * profile.c (compute_value_histograms): Drop time profile for
24235 -fprofile-reproducible=multithreaded.
24237 2021-01-22 Nathan Sidwell <nathan@acm.org>
24239 * gcc.c (process_command): Don't check OPT_SPECIAL_input_file
24242 2021-01-22 Richard Biener <rguenther@suse.de>
24244 PR middle-end/98773
24245 * tree-data-ref.c (initalize_matrix_A): Revert previous
24246 change, retaining failing on HOST_WIDE_INT_MIN CHREC_RIGHT.
24248 2021-01-22 Jakub Jelinek <jakub@redhat.com>
24250 PR tree-optimization/90248
24251 * match.pd (X cmp 0.0 ? 1.0 : -1.0 -> copysign(1, +-X),
24252 X cmp 0.0 ? -1.0 : +1.0 -> copysign(1, -+X)): Remove
24254 (X * (X cmp 0.0 ? 1.0 : -1.0) -> +-abs(X),
24255 X * (X cmp 0.0 ? -1.0 : 1.0) -> +-abs(X)): New simplifications.
24257 2021-01-22 Jakub Jelinek <jakub@redhat.com>
24259 PR tree-optimization/98255
24260 * tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
24261 extend index - low_bound from sizetype's precision rather than index
24263 (get_addr_base_and_unit_offset_1): Likewise.
24264 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
24265 * gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
24267 2021-01-22 Richard Biener <rguenther@suse.de>
24269 PR tree-optimization/98786
24270 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Avoid
24271 adding new uses of abnormals. Verify we deal with a conditional
24274 2021-01-22 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24277 * optc-save-gen.awk: Add arm_fp16_format to checked_options.
24279 2021-01-22 liuhongt <hongtao.liu@intel.com>
24283 * config/i386/sse.md (VI_128_256): New mode iterator.
24284 (*avx_cmp<mode>3_1, *avx_cmp<mode>3_2, *avx_cmp<mode>3_3,
24285 *avx_cmp<mode>3_4, *avx2_eq<mode>3, *avx2_pcmp<mode>3_1,
24286 *avx2_pcmp<mode>3_2, *avx2_gt<mode>3): New
24287 define_insn_and_split to lower avx512 vector comparison to avx
24288 version when dest is vector.
24289 (*<avx512>_cmp<mode>3,*<avx512>_cmp<mode>3,*<avx512>_ucmp<mode>3):
24290 define_insn_and_split for negating the comparison result.
24291 * config/i386/predicates.md (float_vector_all_ones_operand):
24293 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Use
24294 general NOT operator without UNSPEC_MASKOP.
24296 2021-01-21 Vladimir N. Makarov <vmakarov@redhat.com>
24298 PR rtl-optimization/98777
24299 * lra-int.h (lra_pmode_pseudo): New extern.
24300 * lra.c (lra_pmode_pseudo): New global.
24302 * lra-eliminations.c (eliminate_regs_in_insn): Use it.
24304 2021-01-21 Ilya Leoshkevich <iii@linux.ibm.com>
24306 * fwprop.c (fwprop_propagation::classify_result): Allow
24307 (subreg (mem)) simplifications.
24309 2021-01-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24311 * config/aarch64/aarch64-simd.md (aarch64_sqdml<SBINQOPS:as>l<mode>):
24313 (aarch64_sqdmlal<mode>): ... This...
24314 (aarch64_sqdmlsl<mode>): ... And this.
24315 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Split into...
24316 (aarch64_sqdmlal_lane<mode>): ... This...
24317 (aarch64_sqdmlsl_lane<mode>): ... And this.
24318 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Split into...
24319 (aarch64_sqdmlsl_laneq<mode>): ... This...
24320 (aarch64_sqdmlal_laneq<mode>): ... And this.
24321 (aarch64_sqdml<SBINQOPS:as>l_n<mode>): Split into...
24322 (aarch64_sqdmlsl_n<mode>): ... This...
24323 (aarch64_sqdmlal_n<mode>): ... And this.
24324 (aarch64_sqdml<SBINQOPS:as>l2<mode>_internal): Split into...
24325 (aarch64_sqdmlal2<mode>_internal): ... This...
24326 (aarch64_sqdmlsl2<mode>_internal): ... And this.
24328 2021-01-21 Christophe Lyon <christophe.lyon@linaro.org>
24330 * config/arm/arm_mve.h (__arm_vcmpneq_s8): Fix return type.
24332 2021-01-21 Andrea Corallo <andrea.corallo@arm.com>
24335 * doc/sourcebuild.texi (arm_thumb2_no_arm_v8_1_lob): Document.
24337 2021-01-21 liuhongt <hongtao.liu@intel.com>
24339 PR rtl-optimization/98694
24340 * regcprop.c (copy_value): If SRC had been assigned a mode
24341 narrower than the copy, we can't link DEST into the chain even
24342 they have same hard_regno_nregs(i.e. HImode/SImode in i386
24345 2021-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24347 * config/aarch64/aarch64-simd.md (aarch64_get_lane<mode>):
24348 Convert to define_insn_and_split. Split into simple move when moving
24351 2021-01-20 Segher Boessenkool <segher@kernel.crashing.org>
24353 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Change assert.
24354 Adjust comment. Simplify code.
24356 2021-01-20 Jakub Jelinek <jakub@redhat.com>
24359 * dwarf2out.c (reset_indirect_string): Also reset indirect strings
24360 with DW_FORM_line_strp form.
24361 (prune_unused_types_update_strings): Don't add into debug_str_hash
24362 indirect strings with DW_FORM_line_strp form.
24363 (adjust_name_comp_dir): New function.
24364 (dwarf2out_finish): Call it on CU DIEs after resetting
24365 debug_line_str_hash.
24367 2021-01-20 Vladimir N. Makarov <vmakarov@redhat.com>
24369 PR rtl-optimization/98722
24370 * lra-eliminations.c (eliminate_regs_in_insn): Check that target
24371 has no 3-op add insn to transform insns containing two pluses.
24373 2021-01-20 Richard Biener <rguenther@suse.de>
24375 * hwint.h (add_hwi): New function.
24376 (mul_hwi): Likewise.
24377 * tree-data-ref.c (initialize_matrix_A): Properly translate
24378 tree constants and avoid HOST_WIDE_INT_MIN.
24379 (lambda_matrix_row_add): Avoid undefined integer overflow
24380 and return true on such overflow.
24381 (lambda_matrix_right_hermite): Handle overflow from
24382 lambda_matrix_row_add gracefully. Simplify previous fix.
24383 (analyze_subscript_affine_affine): Likewise.
24385 2021-01-20 Eugene Rozenfeld <erozen@microsoft.com>
24387 PR tree-optimization/96674
24388 * match.pd: New patterns: x < y || y == XXX_MIN --> x <= y - 1
24389 x >= y && y != XXX_MIN --> x > y - 1
24391 2021-01-20 Richard Sandiford <richard.sandiford@arm.com>
24393 PR tree-optimization/98535
24394 * tree-vect-slp.c (duplicate_and_interleave): Use quick_grow_cleared.
24395 If the high and low permutes are the same, remove the high permutes
24396 from the working set and only continue with the low ones.
24398 2021-01-20 Jakub Jelinek <jakub@redhat.com>
24400 PR tree-optimization/98721
24401 * builtins.c (access_ref::inform_access): Don't assume
24402 SSA_NAME_IDENTIFIER must be non-NULL. Print messages about
24403 object whenever allocfn is NULL, rather than only when DECL_P
24404 is true. Use %qE instead of %qD for that. Formatting fixes.
24406 2021-01-20 Richard Biener <rguenther@suse.de>
24408 PR tree-optimization/98758
24409 * tree-data-ref.c (int_divides_p): Use lambda_int arguments.
24410 (lambda_matrix_right_hermite): Avoid undefinedness with
24411 signed integer abs and multiplication.
24412 (analyze_subscript_affine_affine): Use lambda_int.
24414 2021-01-20 David Malcolm <dmalcolm@redhat.com>
24417 * dwarf2out.c (output_line_info): Rename static variable
24418 "generation", moving it out of the function to...
24419 (output_line_info_generation): New.
24420 (init_sections_and_labels): Likewise, renaming the variable to...
24421 (init_sections_and_labels_generation): New.
24422 (dwarf2out_c_finalize): Reset the new variables.
24424 2021-01-19 Martin Sebor <msebor@redhat.com>
24426 PR middle-end/98664
24427 * tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
24428 all functions, even if they're not declared artificial or inline.
24429 * tree.c (tree_inlined_location): Use macro expansion location
24430 only if scope traversal fails to expose one.
24432 2021-01-19 Richard Sandiford <richard.sandiford@arm.com>
24434 PR rtl-optimization/92294
24435 * alias.c (compare_base_symbol_refs): Take an extra parameter
24436 and add the distance between two symbols to it. Enshrine in
24437 comments that -1 means "either 0 or 1, but we can't tell
24438 which at compile time".
24439 (memrefs_conflict_p): Update call accordingly.
24440 (rtx_equal_for_memref_p): Likewise. Take the distance between symbols
24443 2021-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24445 * config/aarch64/aarch64-simd-builtins.def (sqshl, uqshl,
24446 sqrshl, uqrshl, sqadd, uqadd, sqsub, uqsub, suqadd, usqadd, sqmovn,
24447 uqmovn, sqxtn2, uqxtn2, sqabs, sqneg, sqdmlal, sqdmlsl, sqdmlal_lane,
24448 sqdmlsl_lane, sqdmlal_laneq, sqdmlsl_laneq, sqdmlal_n, sqdmlsl_n,
24449 sqdmlal2, sqdmlsl2, sqdmlal2_lane, sqdmlsl2_lane, sqdmlal2_laneq,
24450 sqdmlsl2_laneq, sqdmlal2_n, sqdmlsl2_n, sqdmull, sqdmull_lane,
24451 sqdmull_laneq, sqdmull_n, sqdmull2, sqdmull2_lane, sqdmull2_laneq,
24452 sqdmull2_n, sqdmulh, sqrdmulh, sqdmulh_lane, sqdmulh_laneq,
24453 sqrdmulh_lane, sqrdmulh_laneq, sqshrun_n, sqrshrun_n, sqshrn_n,
24454 uqshrn_n, sqrshrn_n, uqrshrn_n, sqshlu_n, sqshl_n, uqshl_n, sqrdmlah,
24455 sqrdmlsh, sqrdmlah_lane, sqrdmlsh_lane, sqrdmlah_laneq, sqrdmlsh_laneq,
24456 sqmovun): Use NONE flags.
24458 2021-01-19 Richard Biener <rguenther@suse.de>
24461 * ipa-modref.c (analyze_stmt): Only record a summary for a
24464 2021-01-19 Richard Biener <rguenther@suse.de>
24466 PR middle-end/98638
24467 * tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
24469 2021-01-19 Daniel Hellstrom <daniel@gaisler.com>
24471 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
24472 built-in define __FIX_LEON3FT_TN0018.
24474 2021-01-19 Richard Biener <rguenther@suse.de>
24477 * tree-inline.c (tree_function_versioning): Set input_location
24478 to UNKNOWN_LOCATION throughout the function.
24480 2021-01-19 Tobias Burnus <tobias@codesourcery.com>
24483 * omp-low.c (lower_omp_target): Handle nonpointer is_device_ptr.
24485 2021-01-19 Martin Jambor <mjambor@suse.cz>
24488 * ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
24489 whether non-call exceptions allow removal of a statement.
24490 (isra_analyze_call): Pass the appropriate function to
24491 ssa_name_only_returned_p.
24493 2021-01-19 Geng Qi <gengqi@linux.alibaba.com>
24495 * config/riscv/arch-canonicalize (longext_sort): New function for
24496 sorting 'multi-letter'.
24497 * config/riscv/multilib-generator: Adjusting the loop of 'alt' in
24498 'alts'. The 'arch' may not be the first of 'alts'.
24499 (_expand_combination): Add underline for the 'ext' without '*'.
24500 This is because, a single-letter extension can always be treated well
24501 with a '_' prefix, but it cannot be separated out if it is appended
24504 2021-01-18 Vladimir N. Makarov <vmakarov@redhat.com>
24507 * ira.c (ira): Skip abnormal critical edge splitting.
24509 2021-01-18 Jakub Jelinek <jakub@redhat.com>
24511 PR tree-optimization/98727
24512 * tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
24513 second .MUL_OVERFLOW operand for signed multiplication with overflow
24514 checking if the second operand of multiplication is not constant.
24516 2021-01-18 David Edelsohn <dje.gcc@gmail.com>
24518 * doc/invoke.texi (-gdwarf): TPF defaults to version 2 and AIX
24519 defaults to version 4.
24521 2021-01-18 David Malcolm <dmalcolm@redhat.com>
24523 * attribs.h (fndecl_dealloc_argno): New decl.
24524 * builtins.c (call_dealloc_argno): Split out second half of
24526 (fndecl_dealloc_argno): New.
24527 * doc/extend.texi (Common Function Attributes): Document the
24528 interaction between the analyzer and the malloc attribute.
24529 * doc/invoke.texi (Static Analyzer Options): Likewise.
24531 2021-01-17 David Edelsohn <dje.gcc@gmail.com>
24533 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Override
24534 dwarf_version to 4.
24535 * config/rs6000/aix72.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
24537 2021-01-17 Martin Jambor <mjambor@suse.cz>
24540 * cgraph.c (clone_of_p): Check also former_clone_of as we climb
24543 2021-01-17 Mark Wielaard <mark@klomp.org>
24545 * common.opt (gdwarf-): Init(5).
24546 * doc/invoke.texi (-gdwarf): Document default to 5.
24548 2021-01-16 Kwok Cheung Yeung <kcy@codesourcery.com>
24550 * builtin-types.def
24551 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
24553 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
24554 ...this. Add extra argument.
24555 * gimplify.c (omp_default_clause): Ensure that event handle is
24556 firstprivate in a task region.
24557 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_DETACH.
24558 (gimplify_adjust_omp_clauses): Likewise.
24559 * omp-builtins.def (BUILT_IN_GOMP_TASK): Change function type to
24560 BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR.
24561 * omp-expand.c (expand_task_call): Add GOMP_TASK_FLAG_DETACH to flags
24562 if detach clause specified. Add detach argument when generating
24564 * omp-low.c (scan_sharing_clauses): Setup data environment for detach
24566 (finish_taskreg_scan): Move field for variable containing the event
24567 handle to the front of the struct.
24568 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_DETACH. Fix
24570 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
24571 OMP_CLAUSE_DETACH clause.
24572 (convert_local_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
24573 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_DETACH.
24574 * tree.c (omp_clause_num_ops): Add entry for OMP_CLAUSE_DETACH.
24576 (omp_clause_code_name): Add entry for OMP_CLAUSE_DETACH. Fix
24578 (walk_tree_1): Handle OMP_CLAUSE_DETACH.
24580 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
24582 * config/nios2/t-rtems: Reset all MULTILIB_* variables. Shorten
24583 multilib directory names. Use MULTILIB_REQUIRED instead of
24584 MULTILIB_EXCEPTIONS. Add -mhw-mul -mhw-mulx -mhw-div
24585 -mcustom-fpu-cfg=fph2 multilib.
24587 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
24589 * config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
24590 (nios2_init_fpu_configs): Provide register values for new
24591 -mcustom-fpu-cfg=fph2 option variant.
24592 * doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
24595 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
24597 * config/nios2/nios2.c (nios2_custom_check_insns): Remove
24598 custom instruction warnings.
24600 2021-01-16 Jakub Jelinek <jakub@redhat.com>
24602 PR tree-optimization/96669
24603 * match.pd ((CST << x) & 1 -> x == 0): New simplification.
24605 2021-01-16 Jakub Jelinek <jakub@redhat.com>
24607 PR tree-optimization/96271
24608 * passes.def: Pass false argument to first two pass_cd_dce
24609 instances and true to last instance. Add comment that
24610 last instance rewrites no longer addressed locals.
24611 * tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
24613 (pass_cd_dce::set_pass_param): New method.
24614 (pass_cd_dce::execute): Return TODO_update_address_taken from
24615 last cd_dce instance.
24617 2021-01-15 Carl Love <cel@us.ibm.com>
24619 * config/rs6000/altivec.h (vec_mulh, vec_div, vec_dive, vec_mod):
24621 * config/rs6000/altivec.md (VIlong): Move define to file vsx.md.
24622 * config/rs6000/rs6000-builtin.def (DIVES_V4SI, DIVES_V2DI,
24623 DIVEU_V4SI, DIVEU_V2DI, DIVS_V4SI, DIVS_V2DI, DIVU_V4SI,
24624 DIVU_V2DI, MODS_V2DI, MODS_V4SI, MODU_V2DI, MODU_V4SI,
24625 MULHS_V2DI, MULHS_V4SI, MULHU_V2DI, MULHU_V4SI, MULLD_V2DI):
24626 Add builtin define.
24627 (MULH, DIVE, MOD): Add new BU_P10_OVERLOAD_2 definitions.
24628 * config/rs6000/rs6000-call.c (VSX_BUILTIN_VEC_DIV,
24629 VSX_BUILTIN_VEC_DIVE, P10_BUILTIN_VEC_MOD, P10_BUILTIN_VEC_MULH):
24630 New overloaded definitions.
24631 (builtin_function_type) [P10V_BUILTIN_DIVEU_V4SI,
24632 P10V_BUILTIN_DIVEU_V2DI, P10V_BUILTIN_DIVU_V4SI,
24633 P10V_BUILTIN_DIVU_V2DI, P10V_BUILTIN_MODU_V2DI,
24634 P10V_BUILTIN_MODU_V4SI, P10V_BUILTIN_MULHU_V2DI,
24635 P10V_BUILTIN_MULHU_V4SI]: Add case
24636 statement for builtins.
24637 * config/rs6000/rs6000.md (bits): Add new attribute sizes V4SI, V2DI.
24638 * config/rs6000/vsx.md (VIlong): Moved from config/rs6000/altivec.md.
24639 (UNSPEC_VDIVES, UNSPEC_VDIVEU): New unspec definitions.
24640 (vsx_mul_v2di): Add if TARGET_POWER10 statement.
24641 (vsx_udiv_v2di): Add if TARGET_POWER10 statement.
24642 (dives_<mode>, diveu_<mode>, div<mode>3, uvdiv<mode>3,
24643 mods_<mode>, modu_<mode>, mulhs_<mode>, mulhu_<mode>, mulv2di3):
24644 Add define_insn, mode is VIlong.
24645 * doc/extend.texi (vec_mulh, vec_mul, vec_div, vec_dive, vec_mod):
24646 Add builtin descriptions.
24648 2021-01-15 Eric Botcazou <ebotcazou@adacore.com>
24650 * final.c (final_start_function_1): Reset force_source_line.
24652 2021-01-15 Jakub Jelinek <jakub@redhat.com>
24654 PR tree-optimization/96669
24655 * match.pd (((1 << A) & 1) != 0 -> A == 0,
24656 ((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
24657 possibly different power of two constants and to right shift too.
24659 2021-01-15 Jakub Jelinek <jakub@redhat.com>
24661 PR tree-optimization/96681
24662 * match.pd ((x < 0) ^ (y < 0) to (x ^ y) < 0): New simplification.
24663 ((x >= 0) ^ (y >= 0) to (x ^ y) < 0): Likewise.
24664 ((x < 0) ^ (y >= 0) to (x ^ y) >= 0): Likewise.
24665 ((x >= 0) ^ (y < 0) to (x ^ y) >= 0): Likewise.
24667 2021-01-15 Alexandre Oliva <oliva@adacore.com>
24669 * opts.c (gen_command_line_string): Exclude -dumpbase-ext.
24671 2021-01-15 Tamar Christina <tamar.christina@arm.com>
24673 * config/aarch64/aarch64-simd.md (cml<fcmac1><conj_op><mode>4,
24674 cmul<conj_op><mode>3): New.
24675 * config/aarch64/iterators.md (UNSPEC_FCMUL,
24676 UNSPEC_FCMUL180, UNSPEC_FCMLA_CONJ, UNSPEC_FCMLA180_CONJ,
24677 UNSPEC_CMLA_CONJ, UNSPEC_CMLA180_CONJ, UNSPEC_CMUL, UNSPEC_CMUL180,
24678 FCMLA_OP, FCMUL_OP, conj_op, rotsplit1, rotsplit2, fcmac1, sve_rot1,
24679 sve_rot2, SVE2_INT_CMLA_OP, SVE2_INT_CMUL_OP, SVE2_INT_CADD_OP): New.
24680 (rot): Add UNSPEC_FCMUL, UNSPEC_FCMUL180.
24681 (rot_op): Renamed to conj_op.
24682 * config/aarch64/aarch64-sve.md (cml<fcmac1><conj_op><mode>4,
24683 cmul<conj_op><mode>3): New.
24684 * config/aarch64/aarch64-sve2.md (cml<fcmac1><conj_op><mode>4,
24685 cmul<conj_op><mode>3): New.
24687 2021-01-15 David Malcolm <dmalcolm@redhat.com>
24691 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
24692 Escape the tempfile name when constructing the expected output.
24694 2021-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24696 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlsl_hi<mode>):
24698 (aarch64_<su>mlsl_hi<mode>): ... This.
24699 (aarch64_<su>mlsl_hi<mode>): Define.
24700 (*aarch64_<su>mlsl<mode): Rename to...
24701 (aarch64_<su>mlsl<mode): ... This.
24702 * config/aarch64/aarch64-simd-builtins.def (smlsl, umlsl,
24703 smlsl_hi, umlsl_hi): Define builtins.
24704 * config/aarch64/arm_neon.h (vmlsl_high_s8, vmlsl_high_s16,
24705 vmlsl_high_s32, vmlsl_high_u8, vmlsl_high_u16, vmlsl_high_u32,
24706 vmlsl_s8, vmlsl_s16, vmlsl_s32, vmlsl_u8,
24707 vmlsl_u16, vmlsl_u32): Reimplement with builtins.
24709 2021-01-15 Uroš Bizjak <ubizjak@gmail.com>
24711 * config/i386/i386-c.c (ix86_target_macros):
24712 Use cpp_define_formatted for __SIZEOF_FLOAT80__ definition.
24714 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
24717 * config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
24718 * Makefile.in (RTL_SSA_H): New variable.
24719 * config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
24720 * config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
24721 * config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
24723 * config/aarch64/aarch64-cc-fusion.cc: New file.
24725 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
24727 * recog.h (insn_change_watermark::~insn_change_watermark): Avoid
24728 calling cancel_changes for changes that no longer exist.
24730 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
24732 * rtl-ssa/functions.h (function_info::ref_defs): Rename to...
24733 (function_info::reg_defs): ...this.
24734 * rtl-ssa/member-fns.inl (function_info::ref_defs): Rename to...
24735 (function_info::reg_defs): ...this.
24737 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
24740 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
24742 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
24745 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
24748 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
24750 2021-01-15 Richard Biener <rguenther@suse.de>
24752 PR tree-optimization/96376
24753 * tree-vect-stmts.c (get_load_store_type): Disregard alignment
24754 for VMAT_INVARIANT.
24756 2021-01-15 Martin Liska <mliska@suse.cz>
24758 * doc/install.texi: Document that some tests need pytest module.
24759 * doc/sourcebuild.texi: Likewise.
24761 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
24764 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
24766 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
24768 * config/arm/mve.md (mve_vshrq_n_s<mode>_imm): New entry.
24769 (mve_vshrq_n_u<mode>_imm): Likewise.
24770 * config/arm/neon.md (vashr<mode>3, vlshr<mode>3): Move to ...
24771 * config/arm/vec-common.md: ... here.
24773 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
24775 * config/arm/mve.md (mve_vshlq_<supf><mode>): Move to
24777 * config/arm/neon.md (vashl<mode>3): Delete.
24778 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): New.
24779 (vasl<mode>3): New expander.
24781 2021-01-15 Richard Biener <rguenther@suse.de>
24783 PR tree-optimization/98685
24784 * tree-vect-slp.c (vect_schedule_slp_node): Refactor handling
24785 of vector extern defs.
24787 2021-01-14 David Malcolm <dmalcolm@redhat.com>
24790 * diagnostic.c (diagnostic_kind_text): Break out this array
24792 (diagnostic_build_prefix): ...here.
24793 (fancy_abort): Detect when diagnostic_initialize has not yet been
24794 called and fall back to a minimal implementation of printing the
24795 ICE, rather than segfaulting in internal_error.
24797 2021-01-14 David Malcolm <dmalcolm@redhat.com>
24799 * diagnostic.c (diagnostic_initialize): Eliminate
24800 parseable_fixits_p in favor of initializing extra_output_kind from
24801 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
24802 (convert_column_unit): New function, split out from...
24803 (diagnostic_converted_column): ...this.
24804 (print_parseable_fixits): Add "column_unit" and "tabstop" params.
24805 Use them to call convert_column_unit on the column values.
24806 (diagnostic_report_diagnostic): Eliminate conditional on
24807 parseable_fixits_p in favor of a switch statement on
24808 extra_output_kind, passing the appropriate values to the new
24809 params of print_parseable_fixits.
24810 (selftest::test_print_parseable_fixits_none): Update for new
24811 params of print_parseable_fixits.
24812 (selftest::test_print_parseable_fixits_insert): Likewise.
24813 (selftest::test_print_parseable_fixits_remove): Likewise.
24814 (selftest::test_print_parseable_fixits_replace): Likewise.
24815 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
24817 (selftest::diagnostic_c_tests): Call it.
24818 * diagnostic.h (enum diagnostics_extra_output_kind): New.
24819 (diagnostic_context::parseable_fixits_p): Delete field in favor
24821 (diagnostic_context::extra_output_kind): ...this new field.
24822 * doc/invoke.texi (Environment Variables): Add
24823 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
24824 * opts.c (common_handle_option): Update handling of
24825 OPT_fdiagnostics_parseable_fixits for change to diagnostic_context
24828 2021-01-14 Tamar Christina <tamar.christina@arm.com>
24830 * tree-vect-slp-patterns.c (class complex_operations_pattern,
24831 complex_operations_pattern::matches,
24832 complex_operations_pattern::recognize,
24833 complex_operations_pattern::build): New.
24834 (slp_patterns): Use it.
24836 2021-01-14 Tamar Christina <tamar.christina@arm.com>
24838 * internal-fn.def (COMPLEX_FMS, COMPLEX_FMS_CONJ): New.
24839 * optabs.def (cmls_optab, cmls_conj_optab): New.
24840 * doc/md.texi: Document them.
24841 * tree-vect-slp-patterns.c (class complex_fms_pattern,
24842 complex_fms_pattern::matches, complex_fms_pattern::recognize,
24843 complex_fms_pattern::build): New.
24845 2021-01-14 Tamar Christina <tamar.christina@arm.com>
24847 * internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
24848 * optabs.def (cmla_optab, cmla_conj_optab): New.
24849 * doc/md.texi: Document them.
24850 * tree-vect-slp-patterns.c (vect_match_call_p,
24851 class complex_fma_pattern, vect_slp_reset_pattern,
24852 complex_fma_pattern::matches, complex_fma_pattern::recognize,
24853 complex_fma_pattern::build): New.
24855 2021-01-14 Tamar Christina <tamar.christina@arm.com>
24857 * internal-fn.def (COMPLEX_MUL, COMPLEX_MUL_CONJ): New.
24858 * optabs.def (cmul_optab, cmul_conj_optab): New.
24859 * doc/md.texi: Document them.
24860 * tree-vect-slp-patterns.c (vect_match_call_complex_mla,
24861 vect_normalize_conj_loc, is_eq_or_top, vect_validate_multiplication,
24862 vect_build_combine_node, class complex_mul_pattern,
24863 complex_mul_pattern::matches, complex_mul_pattern::recognize,
24864 complex_mul_pattern::build): New.
24866 2021-01-14 Tamar Christina <tamar.christina@arm.com>
24868 * tree-vect-slp.c (optimize_load_redistribution_1): New.
24869 (optimize_load_redistribution, vect_is_slp_load_node): New.
24870 (vect_match_slp_patterns): Use it.
24872 2021-01-14 Tamar Christina <tamar.christina@arm.com>
24874 * tree-vect-slp-patterns.c (complex_add_pattern::build):
24877 2021-01-14 Thomas Schwinge <thomas@codesourcery.com>
24879 * config/gcn/mkoffload.c (main): Create an offload image only in
24880 64-bit configurations.
24882 2021-01-14 H.J. Lu <hjl.tools@gmail.com>
24885 * config/i386/i386-options.c (ix86_option_override_internal):
24886 Issue an error for -fcf-protection with CF_BRANCH when compiling
24887 for 32-bit non-TARGET_CMOV targets.
24889 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
24892 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p):
24893 Remove declaration and initialization of shadow variable "ret".
24894 (ix86_option_override_internal): Remove delcaration of
24895 shadow variable "i". Redeclare shadowed variable to unsigned.
24896 * common/config/i386/i386-common.c (pta_size): Redeclare to unsigned.
24897 * config/i386/i386-builtins.c (get_builtin_code_for_version):
24898 Update for redeclaration.
24899 * config/i386/i386.h (pta_size): Ditto.
24901 2021-01-14 Richard Biener <rguenther@suse.de>
24903 PR tree-optimization/98674
24904 * tree-data-ref.c (base_supports_access_fn_components_p): New.
24905 (initialize_data_dependence_relation): For two bases without
24906 possible access fns resort to type size equality when determining
24907 shape compatibility.
24909 2021-01-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24912 * config/arm/arm_neon.h: Replace calls to __builtin_vcge* by
24913 <=, >= operators in vcle and vcge intrinsics respectively.
24914 * config/arm/arm_neon_builtins.def: Remove entry for
24917 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
24920 * config/i386/i386-options.c (ix86_function_specific_save):
24921 Remove redundant assignment to opts->x_ix86_branch_cost.
24922 * config/i386/i386.c (ix86_prefetch_sse):
24923 Rename from x86_prefetch_sse. Update all uses.
24924 * config/i386/i386.h: Update for rename.
24925 * config/i386/i386-options.h: Ditto.
24927 2021-01-14 Jakub Jelinek <jakub@redhat.com>
24930 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3,
24931 *sse4_1_zero_extendv4hiv4si2_3, *sse4_1_zero_extendv2siv2di2_3):
24932 Use Bm instead of m for non-avx. Add isa attribute.
24934 2021-01-14 Jakub Jelinek <jakub@redhat.com>
24936 PR tree-optimization/96688
24937 * match.pd (~(X >> Y) -> ~X >> Y): New simplification if
24938 ~X can be simplified.
24940 2021-01-14 Richard Sandiford <richard.sandiford@arm.com>
24942 * tree-vect-stmts.c (vect_model_load_cost): Account for unused
24943 IFN_LOAD_LANES results.
24945 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24947 * config/aarch64/aarch64-simd.md (aarch64_<su>xtl<mode>):
24949 (aarch64_xtn<mode>): Likewise.
24950 * config/aarch64/aarch64-simd-builtins.def (sxtl, uxtl, xtn):
24953 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
24955 (vmovl_s16): Likewise.
24956 (vmovl_s32): Likewise.
24957 (vmovl_u8): Likewise.
24958 (vmovl_u16): Likewise.
24959 (vmovl_u32): Likewise.
24960 (vmovn_s16): Likewise.
24961 (vmovn_s32): Likewise.
24962 (vmovn_s64): Likewise.
24963 (vmovn_u16): Likewise.
24964 (vmovn_u32): Likewise.
24965 (vmovn_u64): Likewise.
24967 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24969 * config/aarch64/aarch64-simd.md (aarch64_<su>qxtn2<mode>_le):
24971 (aarch64_<su>qxtn2<mode>_be): Likewise.
24972 (aarch64_<su>qxtn2<mode>): Likewise.
24973 * config/aarch64/aarch64-simd-builtins.def (sqxtn2, uqxtn2):
24975 * config/aarch64/iterators.md (SAT_TRUNC): Define code_iterator.
24976 (su): Handle ss_truncate and us_truncate.
24977 * config/aarch64/arm_neon.h (vqmovn_high_s16): Reimplement using
24979 (vqmovn_high_s32): Likewise.
24980 (vqmovn_high_s64): Likewise.
24981 (vqmovn_high_u16): Likewise.
24982 (vqmovn_high_u32): Likewise.
24983 (vqmovn_high_u64): Likewise.
24985 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24987 * config/aarch64/aarch64-simd.md (aarch64_xtn2<mode>_le):
24989 (aarch64_xtn2<mode>_be): Likewise.
24990 (aarch64_xtn2<mode>): Likewise.
24991 * config/aarch64/aarch64-simd-builtins.def (xtn2): Define
24993 * config/aarch64/arm_neon.h (vmovn_high_s16): Reimplement using
24995 (vmovn_high_s32): Likewise.
24996 (vmovn_high_s64): Likewise.
24997 (vmovn_high_u16): Likewise.
24998 (vmovn_high_u32): Likewise.
24999 (vmovn_high_u64): Likewise.
25001 2021-01-13 Stafford Horne <shorne@gmail.com>
25003 * config/or1k/or1k.h (ASM_PREFERRED_EH_DATA_FORMAT): New macro.
25005 2021-01-13 Stafford Horne <shorne@gmail.com>
25007 * config/or1k/linux.h (TARGET_ASM_FILE_END): Define macro.
25009 2021-01-13 Stafford Horne <shorne@gmail.com>
25011 * config/or1k/or1k.h (TARGET_CPU_CPP_BUILTINS): Add builtin
25012 define for __or1k_hard_float__.
25014 2021-01-13 Stafford Horne <shorne@gmail.com>
25016 * config/or1k/or1k.h (NO_PROFILE_COUNTERS): Define as 1.
25017 (PROFILE_HOOK): Define to call _mcount.
25018 (FUNCTION_PROFILER): Change from abort to no-op.
25020 2021-01-13 Jakub Jelinek <jakub@redhat.com>
25022 PR tree-optimization/96691
25023 * match.pd ((~X | C) ^ D -> (X | C) ^ (~D ^ C),
25024 (~X & C) ^ D -> (X & C) ^ (D ^ C)): New simplifications if
25025 (~D ^ C) or (D ^ C) can be simplified.
25027 2021-01-13 Richard Biener <rguenther@suse.de>
25029 PR tree-optimization/92645
25030 * match.pd (BIT_FIELD_REF to conversion): Delay canonicalization
25031 until after vector lowering.
25033 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
25035 * config/aarch64/aarch64-sve.md (fnma<mode>4): Extend from SVE_FULL_I
25037 (@aarch64_pred_fnma<mode>, cond_fnma<mode>, *cond_fnma<mode>_2)
25038 (*cond_fnma<mode>_4, *cond_fnma<mode>_any): Likewise.
25040 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
25042 * config/aarch64/aarch64-sve.md (fma<mode>4): Extend from SVE_FULL_I
25044 (@aarch64_pred_fma<mode>, cond_fma<mode>, *cond_fma<mode>_2)
25045 (*cond_fma<mode>_4, *cond_fma<mode>_any): Likewise.
25047 2021-01-13 Richard Biener <rguenther@suse.de>
25049 PR tree-optimization/92645
25050 * tree-vect-slp.c (vect_build_slp_tree_1): Relax supported
25051 BIT_FIELD_REF argument.
25052 (vect_build_slp_tree_2): Record the desired vector type
25053 on the external vector def.
25054 (vectorizable_slp_permutation): Handle required punning
25055 of existing vector defs.
25057 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
25059 * rtl-ssa/accesses.h (def_lookup): Fix order of comparison results.
25061 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
25063 * config/sh/sh.md (movsf_ie): Remove operands[2] test.
25065 2021-01-13 Samuel Thibault <samuel.thibault@ens-lyon.org>
25067 * config.gcc [$target == *-*-gnu*]: Enable
25068 'default_gnu_indirect_function'.
25070 2021-01-13 Jakub Jelinek <jakub@redhat.com>
25073 * optabs.c (expand_vec_perm_const): Don't force v0 and v1 into
25074 registers before calling targetm.vectorize.vec_perm_const, only after
25076 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const): Handle
25077 two argument permutation when one operand is zero vector and only
25078 after that force operands into registers.
25079 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_1): New
25080 define_insn_and_split pattern.
25081 (*avx512bw_zero_extendv32qiv32hi2_1): Likewise.
25082 (*avx512f_zero_extendv16hiv16si2_1): Likewise.
25083 (*avx2_zero_extendv8hiv8si2_1): Likewise.
25084 (*avx512f_zero_extendv8siv8di2_1): Likewise.
25085 (*avx2_zero_extendv4siv4di2_1): Likewise.
25086 * config/mips/mips.c (mips_vectorize_vec_perm_const): Force operands
25088 * config/arm/arm.c (arm_vectorize_vec_perm_const): Likewise.
25089 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Likewise.
25090 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const): Likewise.
25091 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const): Likewise.
25092 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const): Likewise.
25093 * config/gcn/gcn.c (gcn_vectorize_vec_perm_const): Likewise. Use std::swap.
25095 2021-01-13 Martin Liska <mliska@suse.cz>
25097 PR tree-optimization/98455
25098 * gimple-if-to-switch.cc (condition_info::record_phi_mapping):
25099 Record also virtual PHIs.
25100 (pass_if_to_switch::execute): Return TODO_cleanup_cfg only
25103 2021-01-13 Jonathan Wakely <jwakely@redhat.com>
25105 * doc/invoke.texi (C++ Modules): Fix typos.
25107 2021-01-13 Richard Biener <rguenther@suse.de>
25109 PR tree-optimization/98640
25110 * tree-ssa-sccvn.c (visit_nary_op): Do not try to
25111 handle plus or minus from a truncated operand to be
25114 2021-01-13 Jakub Jelinek <jakub@redhat.com>
25117 * config/i386/i386.md (*btr<mode>_1, *btr<mode>_2): New
25118 define_insn_and_split patterns.
25119 (splitter after *btr<mode>_2): New splitter.
25121 2021-01-13 Martin Liska <mliska@suse.cz>
25124 * cgraphunit.c (analyze_functions): Remove dead code.
25126 2021-01-13 Qian Jianhua <qianjh@cn.fujitsu.com>
25128 * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
25129 * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
25130 (a64fx_regmove_cost, a64fx_vector_cost): New.
25131 (a64fx_tunings): Use the new added cost tables.
25133 2021-01-13 Jakub Jelinek <jakub@redhat.com>
25136 * config/i386/predicates.md (pmovzx_parallel): New predicate.
25137 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): New
25138 define_insn_and_split pattern.
25139 (*sse4_1_zero_extendv4hiv4si2_3): Likewise.
25140 (*sse4_1_zero_extendv2siv2di2_3): Likewise.
25142 2021-01-13 Julian Brown <julian@codesourcery.com>
25144 * config/gcn/gcn.c (gcn_conditional_register_usage): Remove dead code
25145 to fix v0 register.
25147 2021-01-13 Julian Brown <julian@codesourcery.com>
25149 * config/gcn/gcn.c (gcn_md_reorg): Fix case where EXEC reg is live
25152 2021-01-13 Julian Brown <julian@codesourcery.com>
25154 * config/gcn/gcn-valu.md (recip<mode>2<exec>, recip<mode>2): Use unspec
25155 for reciprocal-approximation instructions.
25156 (div<mode>3): Use fused multiply-accumulate operations for reciprocal
25157 refinement and division result.
25158 * config/gcn/gcn.md (UNSPEC_RCP): New unspec constant.
25160 2021-01-13 Julian Brown <julian@codesourcery.com>
25162 * config/gcn/gcn-valu.md (subdf): Rename to...
25165 2021-01-12 Martin Liska <mliska@suse.cz>
25167 * gcov.c (source_info::debug): Fix printf format for 32-bit hosts.
25169 2021-01-12 Andrea Corallo <andrea.corallo@arm.com>
25171 * function-abi.h: Fix typo.
25173 2021-01-12 Christophe Lyon <christophe.lyon@linaro.org>
25177 * config/arm/arm.h (ARM_HAVE_NEON_V8QI_LDST): New macro.
25178 (ARM_HAVE_NEON_V16QI_LDST, ARM_HAVE_NEON_V4HI_LDST): Likewise.
25179 (ARM_HAVE_NEON_V8HI_LDST, ARM_HAVE_NEON_V2SI_LDST): Likewise.
25180 (ARM_HAVE_NEON_V4SI_LDST, ARM_HAVE_NEON_V4HF_LDST): Likewise.
25181 (ARM_HAVE_NEON_V8HF_LDST, ARM_HAVE_NEON_V4BF_LDST): Likewise.
25182 (ARM_HAVE_NEON_V8BF_LDST, ARM_HAVE_NEON_V2SF_LDST): Likewise.
25183 (ARM_HAVE_NEON_V4SF_LDST, ARM_HAVE_NEON_DI_LDST): Likewise.
25184 (ARM_HAVE_NEON_V2DI_LDST): Likewise.
25185 (ARM_HAVE_V8QI_LDST, ARM_HAVE_V16QI_LDST): Likewise.
25186 (ARM_HAVE_V4HI_LDST, ARM_HAVE_V8HI_LDST): Likewise.
25187 (ARM_HAVE_V2SI_LDST, ARM_HAVE_V4SI_LDST, ARM_HAVE_V4HF_LDST): Likewise.
25188 (ARM_HAVE_V8HF_LDST, ARM_HAVE_V4BF_LDST, ARM_HAVE_V8BF_LDST): Likewise.
25189 (ARM_HAVE_V2SF_LDST, ARM_HAVE_V4SF_LDST, ARM_HAVE_DI_LDST): Likewise.
25190 (ARM_HAVE_V2DI_LDST): Likewise.
25191 * config/arm/mve.md (*movmisalign<mode>_mve_store): New pattern.
25192 (*movmisalign<mode>_mve_load): New pattern.
25193 * config/arm/neon.md (movmisalign<mode>): Move to ...
25194 * config/arm/vec-common.md: ... here.
25196 2021-01-12 Vladimir N. Makarov <vmakarov@redhat.com>
25199 * lra-eliminations.c (eliminate_regs_in_insn): Add transformation
25200 of pattern 'plus (plus (hard reg, const), pseudo)'.
25202 2021-01-12 Richard Biener <rguenther@suse.de>
25204 PR tree-optimization/98550
25205 * tree-vect-slp.c (vect_record_max_nunits): Check whether
25206 the group size is a multiple of the vector element count.
25207 (vect_build_slp_tree_1): When we need to fail because
25208 the vector type choosen causes unrolling do so lazily
25209 without affecting matches only at the end to guide group splitting.
25211 2021-01-12 Martin Liska <mliska@suse.cz>
25214 * optc-save-gen.awk: Compare also n_target_save vars with
25217 2021-01-12 Martin Liska <mliska@suse.cz>
25219 * gcov.c (source_info::debug): New.
25220 (print_usage): Add --debug (-D) option.
25221 (process_args): Likewise.
25222 (generate_results): Call src->debug after
25223 accumulate_line_counts.
25224 (read_graph_file): Properly assign id for EXIT_BLOCK.
25225 * profile.c (branch_prob): Dump function body before it is
25228 2021-01-12 Jakub Jelinek <jakub@redhat.com>
25230 PR tree-optimization/98629
25231 * tree-ssa-math-opts.c (arith_overflow_check_p): Don't update use_stmt
25232 unless returning non-zero.
25234 2021-01-12 Jakub Jelinek <jakub@redhat.com>
25236 PR tree-optimization/95731
25237 * tree-ssa-reassoc.c (optimize_range_tests_cmp_bitwise): Also optimize
25238 x < 0 && y < 0 && z < 0 into (x | y | z) < 0 for signed x, y, z.
25239 (optimize_range_tests): Call optimize_range_tests_cmp_bitwise
25240 only after optimize_range_tests_var_bound.
25242 2021-01-12 Jakub Jelinek <jakub@redhat.com>
25244 * configure.ac: Ensure c/Make-lang.in comes first in @all_lang_makefrags@.
25245 * configure: Regenerated.
25247 2021-01-12 liuhongt <hongtao.liu@intel.com>
25250 * config/i386/i386-builtins.h (BUILTIN_DESC_SWAP_OPERANDS):
25252 * config/i386/i386-expand.c (ix86_expand_sse_comi): Delete
25255 2021-01-12 Alexandre Oliva <oliva@adacore.com>
25257 * ssa-iterators.h (end_imm_use_stmt_traverse): Forward
25259 (auto_end_imm_use_stmt_traverse): New struct.
25260 (FOR_EACH_IMM_USE_STMT): Use it.
25261 (BREAK_FROM_IMM_USE_STMT, RETURN_FROM_IMM_USE_STMT): Remove,
25263 * gimple-ssa-strength-reduction.c: ... here, ...
25264 * graphite-scop-detection.c: ... here, ...
25265 * ipa-modref.c, ipa-pure-const.c, ipa-sra.c: ... here, ...
25266 * tree-predcom.c, tree-ssa-ccp.c: ... here, ...
25267 * tree-ssa-dce.c, tree-ssa-dse.c: ... here, ...
25268 * tree-ssa-loop-ivopts.c, tree-ssa-math-opts.c: ... here, ...
25269 * tree-ssa-phiprop.c, tree-ssa.c: ... here, ...
25270 * tree-vect-slp.c: ... and here, ...
25271 * doc/tree-ssa.texi: ... and the example here.
25273 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25275 * config/aarch64/aarch64-sve.md (sdiv_pow2<mode>3): Extend from
25276 SVE_FULL_I to SVE_I. Generate an UNSPEC_PRED_X.
25277 (*sdiv_pow2<mode>3): New pattern.
25278 (@cond_<sve_int_op><mode>): Extend from SVE_FULL_I to SVE_I.
25279 Wrap the ASRD in an UNSPEC_PRED_X.
25280 (*cond_<sve_int_op><mode>_2): Likewise. Replace the UNSPEC_PRED_X
25281 predicate with a constant PTRUE, if it isn't already.
25282 (*cond_<sve_int_op><mode>_z): Replace with...
25283 (*cond_<sve_int_op><mode>_any): ...this new pattern.
25285 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25287 * config/aarch64/aarch64-sve.md (*cond_bic<mode>_2): Extend from
25288 SVE_FULL_I to SVE_I.
25289 (*cond_bic<mode>_any): Likewise.
25291 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25293 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart)
25294 (@aarch64_pred_<MUL_HIGHPART:optab><mode>): Extend from SVE_FULL_I
25297 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25299 * config/aarch64/aarch64-sve.md (<su>abd<mode>_3): Extend from
25300 SVE_FULL_I to SVE_I.
25301 (*aarch64_cond_<su>abd<mode>_2): Likewise.
25302 (*aarch64_cond_<su>abd<mode>_any): Likewise.
25303 (@aarch64_pred_<su>abd<mode>): Likewise. Use UNSPEC_PRED_X
25304 for the max and min but not for the minus.
25305 (*aarch64_cond_<su>abd<mode>_3): New pattern.
25307 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25309 * config/aarch64/iterators.md (SVE_24I): New iterator.
25310 * config/aarch64/aarch64-sve.md (*aarch64_adr<mode>_shift): Extend from
25311 SVE_FULL_SDI to SVE_24I. Use containers rather than elements.
25313 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25315 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_BINARY:optab><mode>)
25316 (*cond_<SVE_INT_BINARY:optab><mode>_2): Extend from SVE_FULL_I
25318 (*cond_<SVE_INT_BINARY:optab><mode>_3): Likewise.
25319 (*cond_<SVE_INT_BINARY:optab><mode>_any): Likewise.
25320 (*cond_<SVE_INT_BINARY:optab><mode>_2_const): Likewise.
25321 (*cond_<SVE_INT_BINARY:optab><mode>_any_const): Likewise.
25323 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25325 * config/aarch64/aarch64-sve.md (<SVE_INT_BINARY_IMM:optab><mode>3)
25326 (@aarch64_pred_<SVE_INT_BINARY_IMM:optab><mode>)
25327 (*post_ra_<SVE_INT_BINARY_IMM:optab><mode>3): Extend from SVE_FULL_I
25330 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25332 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3)
25333 (v<ASHIFT:optab><mode>3, @aarch64_pred_<optab><mode>)
25334 (*post_ra_v<ASHIFT:optab><mode>3): Extend from SVE_FULL_I to SVE_I.
25336 2021-01-11 Martin Liska <mliska@suse.cz>
25339 * symtab-clones.h (clone_info::release): Release
25340 symtab::m_clones with ggc_delete as it's a GGC memory.
25342 2021-01-11 Matthias Klose <doko@ubuntu.com>
25344 * Makefile.in (LINK_PROGRESS): Show the link target.
25346 2021-01-11 Richard Biener <rguenther@suse.de>
25348 PR tree-optimization/91403
25349 * tree-vect-data-refs.c (vect_analyze_group_access_1): Cap
25350 single-element interleaving group size at 4096 elements.
25352 2021-01-11 Richard Biener <rguenther@suse.de>
25354 PR tree-optimization/98526
25355 * tree-vect-loop.c (vect_model_reduction_cost): Remove costing
25356 of the actual reduction op for the regular case.
25357 (vectorizable_reduction): Cost the stmts
25358 vect_transform_reduction produces here.
25360 2021-01-11 Andreas Krebbel <krebbel@linux.ibm.com>
25362 * tree-ssa-forwprop.c (simplify_vector_constructor): For
25363 big-endian, use UNPACK[_FLOAT]_HI.
25365 2021-01-11 Tamar Christina <tamar.christina@arm.com>
25367 * tree-vect-slp-patterns.c (class complex_pattern,
25368 class complex_add_pattern): Add parameters to matches.
25369 (complex_add_pattern::build): Free memory.
25370 (complex_add_pattern::matches): Move validation end of match.
25371 (complex_add_pattern::recognize): Likewise.
25373 2021-01-11 Tamar Christina <tamar.christina@arm.com>
25375 * tree-vect-slp-patterns.c (linear_loads_p): Fix externals.
25377 2021-01-11 Tamar Christina <tamar.christina@arm.com>
25379 * tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
25381 2021-01-11 Jakub Jelinek <jakub@redhat.com>
25383 PR tree-optimization/95867
25384 * tree-ssa-math-opts.h: New header.
25385 * tree-ssa-math-opts.c: Include tree-ssa-math-opts.h.
25386 (powi_as_mults): No longer static. Use build_one_cst instead of
25387 build_real. Formatting fix.
25388 * tree-ssa-reassoc.c: Include tree-ssa-math-opts.h.
25389 (attempt_builtin_powi): Handle multiplication reassociation without
25390 powi_fndecl using powi_as_mults.
25391 (reassociate_bb): For integral types don't require
25392 -funsafe-math-optimizations to call attempt_builtin_powi.
25394 2021-01-11 Jakub Jelinek <jakub@redhat.com>
25396 PR tree-optimization/95852
25397 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): Change
25398 mul_stmts parameter type to vec<gimple *> &. Before cond_stmt
25399 allow in the bb any of the stmts in that vector, div_stmt and
25400 up to 3 cast stmts.
25401 (arith_cast_equal_p): New function.
25402 (arith_overflow_check_p): Add cast_stmt argument, handle signed
25403 multiply overflow checks.
25404 (match_arith_overflow): Adjust caller. Handle signed multiply
25407 2021-01-11 Jakub Jelinek <jakub@redhat.com>
25409 PR tree-optimization/95852
25410 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): New function.
25411 (uaddsub_overflow_check_p): Renamed to ...
25412 (arith_overflow_check_p): ... this. Handle also multiplication
25413 with overflow check.
25414 (match_uaddsub_overflow): Renamed to ...
25415 (match_arith_overflow): ... this. Add cfg_changed argument. Handle
25416 also multiplication with overflow check. Adjust function comment.
25417 (math_opts_dom_walker::after_dom_children): Adjust callers. Call
25418 match_arith_overflow also for MULT_EXPR.
25420 2021-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25422 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
25423 __builtin_convertvector.
25424 (vmovl_s16): Likewise.
25425 (vmovl_s32): Likewise.
25426 (vmovl_u8): Likewise.
25427 (vmovl_u16): Likewise.
25428 (vmovl_u32): Likewise.
25429 (vmovn_s16): Likewise.
25430 (vmovn_s32): Likewise.
25431 (vmovn_s64): Likewise.
25432 (vmovn_u16): Likewise.
25433 (vmovn_u32): Likewise.
25434 (vmovn_u64): Likewise.
25436 2021-01-11 Martin Liska <mliska@suse.cz>
25438 * gimple-if-to-switch.cc (struct condition_info): Use auto_var.
25439 (if_chain::is_beneficial): Delete clusters
25440 (find_conditions): Make second argument of conditions_in_bbs a
25441 pointer so that we control over it's lifetime.
25442 (pass_if_to_switch::execute): Delete them.
25444 2021-01-11 Kewen Lin <linkw@linux.ibm.com>
25446 * ira.c (move_unallocated_pseudos): Check other_reg and skip if
25449 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
25451 * config/vax/vax.md (cc): Remove mode attribute.
25452 (subst_<cc>, subst_f<cc>): Rename to...
25453 (subst_<mode>, subst_f<VAXccnz:mode>): ... these respectively.
25454 (*cbranch<VAXint:mode>4_<VAXcc:mode>): Update for `cc' removal.
25455 (*cbranch<VAXfp:mode>4_<VAXccnz:mode>): Likewise.
25456 (*branch_<mode>, *branch_<mode>_reversed): Likewise.
25458 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
25460 * config/vax/vax.md (subst_f<cc>): Add mode to operands and
25461 `const_double_zero'.
25463 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
25465 * config/pdp11/pdp11.md (PDPfp): New mode iterator.
25466 (fcc_cc, fcc_ccnz): Use it. Add mode to `const_double_zero' and
25469 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
25471 * genemit.c (gen_exp) <CONST_DOUBLE>: Handle `const_double_zero'
25473 * read-rtl.c (rtx_reader::read_rtx_code): Handle machine mode
25474 with `const_double_zero'.
25475 * doc/rtl.texi (Constant Expression Types): Document it.
25477 2021-01-09 Jakub Jelinek <jakub@redhat.com>
25480 * tree-cfg.c (verify_gimple_assign_binary): Allow lhs of
25481 POINTER_DIFF_EXPR to be any integral type.
25483 2021-01-09 Jakub Jelinek <jakub@redhat.com>
25485 PR rtl-optimization/98603
25486 * function.c (instantiate_virtual_regs_in_insn): For asm goto
25487 with impossible constraints, drop all SETs, CLOBBERs, drop PARALLEL
25488 if any, set ASM_OPERANDS mode to VOIDmode and change
25489 ASM_OPERANDS_OUTPUT_CONSTRAINT and ASM_OPERANDS_OUTPUT_IDX.
25491 2021-01-09 Alexandre Oliva <oliva@gnu.org>
25494 * final.c (notice_source_line): Narrow down the condition to
25495 skip a line-0 marker.
25497 2021-01-08 Sergei Trofimovich <siarheit@google.com>
25499 * ipa-modref.c (merge_call_side_effects): Fix
25500 linebreak split by reordering two print calls.
25502 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
25504 * config/s390/vector.md (*tf_to_fprx2_0): Rename from
25505 "*mov_tf_to_fprx2_0" for consistency, fix constraint.
25506 (*tf_to_fprx2_1): Rename from "*mov_tf_to_fprx2_1" for
25507 consistency, fix constraint.
25509 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
25511 * config/s390/s390-c.c (s390_def_or_undef_macro): Accept
25512 callables instead of mask values.
25513 (struct target_flag_set_p): New predicate.
25514 (s390_cpu_cpp_builtins_internal): Define or undefine
25515 __LONG_DOUBLE_VX__ macro.
25517 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
25520 * config/i386/i386.c (x86_function_profiler): Use R10 and R11
25521 to call mcount in large model with PIC for NO_PROFILE_COUNTERS
25524 2021-01-08 Richard Biener <rguenther@suse.de>
25526 * tree-ssa-sccvn.c (pass_fre::execute): Reset the SCEV hash table.
25528 2021-01-08 Richard Biener <rguenther@suse.de>
25530 * tree-vect-slp.c (scalar_stmts_to_slp_tree_map_t): Fix.
25531 (vect_build_slp_tree): On cache hit release the matched
25532 scalar stmts vector.
25533 * tree-vect-stmts.c (vectorizable_store): Properly free
25534 vec_oprnds before possibly gathering them again.
25536 2021-01-08 Richard Biener <rguenther@suse.de>
25538 PR tree-optimization/98544
25539 * tree-vect-slp.c (vect_optimize_slp): Always materialize
25540 permutes at a permute node.
25542 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
25545 * config/i386/i386.c (x86_function_profiler): Use R10 to call
25546 mcount in large model. Sorry for large model with PIC.
25548 2021-01-08 Jakub Jelinek <jakub@redhat.com>
25551 * config/i386/i386.opt (ix86_cmodel, ix86_incoming_stack_boundary_arg,
25552 ix86_pmode, ix86_preferred_stack_boundary_arg, ix86_regparm,
25553 ix86_veclibabi_type): Remove x_ prefix, use TargetVariable instead of
25554 TargetSave and initialize for variables with enum types.
25555 (mfentry, mstack-protector-guard-reg=, mstack-protector-guard-offset=,
25556 mstack-protector-guard-symbol=): Add Save.
25557 * config/i386/i386-options.c (ix86_function_specific_save,
25558 ix86_function_specific_restore): Don't save or restore x_ix86_cmodel,
25559 x_ix86_incoming_stack_boundary_arg, x_ix86_pmode,
25560 x_ix86_preferred_stack_boundary_arg, x_ix86_regparm,
25561 x_ix86_veclibabi_type.
25563 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
25565 * config/aarch64/aarch64-sve.md (*cnot<mode>): Extend from
25566 SVE_FULL_I to SVE_I.
25567 (*cond_cnot<mode>_2, *cond_cnot<mode>_any): Likewise.
25569 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
25571 * config/aarch64/aarch64-sve.md (*cond_uxt<mode>_2): Extend from
25572 SVE_FULL_I to SVE_I.
25573 (*cond_uxt<mode>_any): Likewise.
25575 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25577 * config/aarch64/iterators.md (Vwhalf): New iterator.
25578 * config/aarch64/aarch64-simd.md (aarch64_<sur>adalp<mode>_3):
25580 (aarch64_<sur>adalp<mode>): ... This. Make more
25582 (<sur>sadv16qi): Adjust callsite of the above.
25583 * config/aarch64/aarch64-simd-builtins.def (sadalp, uadalp): New
25585 * config/aarch64/arm_neon.h (vpadal_s8): Reimplement using
25587 (vpadal_s16): Likewise.
25588 (vpadal_u8): Likewise.
25589 (vpadal_u16): Likewise.
25590 (vpadalq_s8): Likewise.
25591 (vpadalq_s16): Likewise.
25592 (vpadalq_s32): Likewise.
25593 (vpadalq_u8): Likewise.
25594 (vpadalq_u16): Likewise.
25595 (vpadalq_u32): Likewise.
25597 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25599 * config/aarch64/aarch64-simd.md (aarch64_<su>abd<mode>_3):
25601 (aarch64_<su>abd<mode>): ... This.
25602 (<sur>sadv16qi): Adjust callsite of the above.
25603 * config/aarch64/aarch64-simd-builtins.def (sabd, uabd): Define
25605 * config/aarch64/arm_neon.h (vabd_s8): Reimplement using
25607 (vabd_s16): Likewise.
25608 (vabd_s32): Likewise.
25609 (vabd_u8): Likewise.
25610 (vabd_u16): Likewise.
25611 (vabd_u32): Likewise.
25612 (vabdq_s8): Likewise.
25613 (vabdq_s16): Likewise.
25614 (vabdq_s32): Likewise.
25615 (vabdq_u8): Likewise.
25616 (vabdq_u16): Likewise.
25617 (vabdq_u32): Likewise.
25619 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25621 * config/aarch64/aarch64-simd-builtins.def (saba, uaba): Define
25623 * config/aarch64/arm_neon.h (vaba_s8): Implement using builtin.
25624 (vaba_s16): Likewise.
25625 (vaba_s32): Likewise.
25626 (vaba_u8): Likewise.
25627 (vaba_u16): Likewise.
25628 (vaba_u32): Likewise.
25629 (vabaq_s8): Likewise.
25630 (vabaq_s16): Likewise.
25631 (vabaq_s32): Likewise.
25632 (vabaq_u8): Likewise.
25633 (vabaq_u16): Likewise.
25634 (vabaq_u32): Likewise.
25636 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25638 * config/aarch64/aarch64-simd.md (aba<mode>_3): Rename to...
25639 (aarch64_<su>aba<mode>): ... This. Handle uaba as well.
25640 Change RTL pattern to match.
25642 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
25644 * common/config/riscv/riscv-common.c (riscv_current_subset_list): New.
25645 * config/riscv/riscv-c.c (riscv-subset.h): New.
25646 (INCLUDE_STRING): Define.
25647 (riscv_cpu_cpp_builtins): Add new style architecture extension
25649 * config/riscv/riscv-subset.h (riscv_subset_list::begin): New.
25650 (riscv_subset_list::end): New.
25651 (riscv_current_subset_list): New.
25653 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
25655 * common/config/riscv/riscv-common.c (RISCV_DONT_CARE_VERSION):
25656 Move to riscv-subset.h.
25657 (struct riscv_subset_t): Ditto.
25658 (class riscv_subset_list): Ditto.
25659 * config/riscv/riscv-subset.h (RISCV_DONT_CARE_VERSION): Move
25660 from riscv-common.c.
25661 (struct riscv_subset_t): Ditto.
25662 (class riscv_subset_list): Ditto.
25663 * config/riscv/t-riscv ($(common_out_file)): Add file
25666 2021-01-07 Jakub Jelinek <jakub@redhat.com>
25669 * config/i386/i386.md (*bmi_blsi_<mode>_cmp, *bmi_blsi_<mode>_ccno):
25670 New define_insn patterns.
25672 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
25674 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_UNARY:optab><mode>)
25675 (*cond_<SVE_INT_UNARY:optab><mode>_2): Extend from SVE_FULL_I to SVE_I.
25676 (*cond_<SVE_INT_UNARY:optab><mode>_any): Likewise.
25678 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
25680 PR tree-optimization/98560
25681 * internal-fn.def (IFN_VCONDU, IFN_VCONDEQ): Use type vec_cond.
25682 * internal-fn.c (vec_cond_mask_direct): Get the data mode from
25684 (vec_cond_direct): Likewise argument 2.
25685 (vec_condu_direct, vec_condeq_direct): Delete.
25686 (expand_vect_cond_optab_fn): Rename to...
25687 (expand_vec_cond_optab_fn): ...this, replacing old macro.
25688 (expand_vec_condu_optab_fn, expand_vec_condeq_optab_fn): Delete.
25689 (expand_vect_cond_mask_optab_fn): Rename to...
25690 (expand_vec_cond_mask_optab_fn): ...this, replacing old macro.
25691 (direct_vec_cond_mask_optab_supported_p): Treat the optab as a
25693 (direct_vec_cond_optab_supported_p): Likewise.
25694 (direct_vec_condu_optab_supported_p): Delete.
25695 (direct_vec_condeq_optab_supported_p): Delete.
25696 * gimple-isel.cc: Include internal-fn.h.
25697 (gimple_expand_vec_cond_expr): Check that IFN_VCONDEQ is supported
25700 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
25702 PR tree-optimization/98560
25703 * gimple-isel.cc (gimple_expand_vec_cond_expr): If we fail to use
25704 IFN_VCOND{,U,EQ}, fall back on IFN_VCOND_MASK.
25706 2021-01-07 Uroš Bizjak <ubizjak@gmail.com>
25708 * config/i386/i386.md (insn): Merge from plusminus_insn, shift_insn,
25709 rotate_insn and optab code attributes.
25710 Update all uses to merged code attribute.
25711 * config/i386/sse.md: Update all uses to merged code attribute.
25712 * config/i386/mmx.md: Update all uses to merged code attribute.
25714 2021-01-07 Jakub Jelinek <jakub@redhat.com>
25716 PR tree-optimization/98568
25717 * gimple-ssa-store-merging.c (bswap_view_convert): New function.
25718 (bswap_replace): Use it.
25720 2021-01-06 Vladimir N. Makarov <vmakarov@redhat.com>
25722 PR rtl-optimization/97978
25723 * lra-int.h (lra_hard_reg_split_p): New external.
25724 * lra.c (lra_hard_reg_split_p): New global.
25725 (lra): Set up lra_hard_reg_split_p after splitting a hard reg.
25726 * lra-assigns.c (lra_assign): Don't check allocation correctness
25727 after hard reg splitting.
25729 2021-01-06 Martin Sebor <msebor@redhat.com>
25732 * builtins.c (new_delete_mismatch_p): New overload.
25733 (new_delete_mismatch_p (tree, tree)): Call it.
25735 2021-01-06 Alexandre Oliva <oliva@adacore.com>
25737 * Makefile.in (T_GLIMITS_H): New.
25738 (stmp-int-hdrs): Depend on it, use it.
25739 * config/t-vxworks (T_GLIMITS_H): Override it.
25740 (vxw-glimits.h): New.
25742 2021-01-06 Richard Biener <rguenther@suse.de>
25744 PR tree-optimization/98513
25745 * value-range.cc (intersect_ranges): Compare the upper bounds
25746 for the expected relation.
25748 2021-01-06 Gerald Pfeifer <gerald@pfeifer.com>
25751 2020-12-28 Gerald Pfeifer <gerald@pfeifer.com>
25753 * doc/standards.texi (HSAIL): Remove section.
25755 2021-01-05 Samuel Thibault <samuel.thibault@ens-lyon.org>
25757 * configure: Re-generate.
25759 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25761 * doc/invoke.texi (-std=c++20): Adjust for the publication of
25762 ISO 14882:2020 standard.
25763 * doc/standards.texi: Likewise.
25765 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25767 PR tree-optimization/94802
25768 * expr.h (maybe_optimize_sub_cmp_0): Declare.
25769 * expr.c: Include tree-pretty-print.h and flags.h.
25770 (maybe_optimize_sub_cmp_0): New function.
25771 (do_store_flag): Use it.
25772 * cfgexpand.c (expand_gimple_cond): Likewise.
25774 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
25776 * mux-utils.h (pointer_mux::m_ptr): Tweak description of contents.
25777 * rtlanal.c (simple_regno_set): Tweak description to clarify the
25780 2021-01-05 Richard Biener <rguenther@suse.de>
25782 PR tree-optimization/98516
25783 * tree-vect-slp.c (vect_optimize_slp): Permute the incoming
25784 lanes when materializing on a VEC_PERM node.
25785 (vectorizable_slp_permutation): Dump the permute properly.
25787 2021-01-05 Richard Biener <rguenther@suse.de>
25789 * tree-vect-slp.c (vect_slp_region): Move debug counter
25790 to cover individual subgraphs.
25792 2021-01-05 Richard Biener <rguenther@suse.de>
25794 PR tree-optimization/98428
25795 * tree-vect-slp.c (vect_build_slp_tree_1): Properly reject
25796 vector lane extracts for loop vectorization.
25798 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25800 PR tree-optimization/98514
25801 * tree-ssa-reassoc.c (bb_rank): Change type from long * to
25803 (operand_rank): Change type from hash_map<tree, long> to
25804 hash_map<tree, int64_t>.
25805 (phi_rank): Change return type from long to int64_t.
25806 (loop_carried_phi): Change block_rank variable type from long to
25808 (propagate_rank): Change return type, rank parameter type and
25809 op_rank variable type from long to int64_t.
25810 (find_operand_rank): Change return type from long to int64_t
25811 and change slot variable type from long * to int64_t *.
25812 (insert_operand_rank): Change rank parameter type from long to
25814 (get_rank): Change return type and rank variable type from long to
25815 int64_t. Use PRId64 instead of ld to print the rank.
25816 (init_reassoc): Change rank variable type from long to int64_t
25817 and adjust correspondingly bb_rank and operand_rank initialization.
25819 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25821 PR tree-optimization/96928
25822 * tree-ssa-phiopt.c (xor_replacement): New function.
25823 (tree_ssa_phiopt_worker): Call it.
25825 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25827 PR tree-optimization/96930
25828 * match.pd ((A / (1 << B)) -> (A >> B)): If A is extended
25829 from narrower value which has the same type as 1 << B, perform
25830 the right shift on the narrower value followed by extension.
25832 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25834 PR tree-optimization/96239
25835 * gimple-ssa-store-merging.c (maybe_optimize_vector_constructor): New
25837 (get_status_for_store_merging): Don't return BB_INVALID for blocks
25838 with potential bswap optimizable CONSTRUCTORs.
25839 (pass_store_merging::execute): Optimize vector CONSTRUCTORs with bswap
25842 2021-01-05 Richard Biener <rguenther@suse.de>
25844 PR tree-optimization/98381
25845 * tree.c (vector_element_bits): Properly compute bool vector
25847 * tree-vect-loop.c (vectorizable_live_operation): Properly
25848 compute the last lane bit offset.
25850 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
25853 * config/i386/sse.md (sse_cvtps2pi): Redefine as define_insn_and_split.
25854 Clear the top 64 bytes of the input XMM register.
25855 (sse_cvttps2pi): Ditto.
25857 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
25860 * config/i386/xopintrin.h (_mm256_cmov_si256): New.
25862 2021-01-05 H.J. Lu <hjl.tools@gmail.com>
25865 * config/i386/xmmintrin.h (_mm_extract_pi16): Cast to unsigned
25868 2021-01-05 Claudiu Zissulescu <claziss@synopsys.com>
25870 * config/arc/arc.md (maddsidi4_split): Use ACC_REG_FIRST.
25871 (umaddsidi4_split): Likewise.
25873 2021-01-05 liuhongt <hongtao.liu@intel.com>
25876 * config/i386/sse.md (*sse2_pmovskb_zexthisi): New
25877 define_insn_and_split for zero_extend of subreg HI of pmovskb
25879 (*sse2_pmovskb_zexthisi): Add new combine splitters for
25880 zero_extend of not of subreg HI of pmovskb result.
25882 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
25885 * explow.c (convert_memory_address_addr_space_1): Handle UNSPECs
25887 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate): Use
25888 convert_memory_address to convert symbolic immediates to ptr_mode
25889 before forcing them to memory.
25891 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
25893 PR rtl-optimization/97144
25894 * recog.c (constrain_operands): Initialize matching_operand
25895 for each alternative, rather than only doing it once.
25897 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
25899 PR rtl-optimization/98403
25900 * rtl-ssa/changes.cc (function_info::finalize_new_accesses): Explain
25901 why we don't remove call clobbers.
25902 (function_info::apply_changes_to_insn): Don't attempt to add
25903 call clobbers here.
25905 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
25907 PR tree-optimization/98371
25908 * tree-vect-loop.c (vect_reanalyze_as_main_loop): New function.
25909 (vect_analyze_loop): If an epilogue loop appears to be cheaper
25910 than the main loop, re-analyze it as a main loop before adopting
25913 2021-01-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
25916 * configure.ac (NETLIBS): Determine using AX_LIB_SOCKET_NSL.
25917 * aclocal.m4, configure: Regenerate.
25918 * Makefile.in (NETLIBS): Define.
25919 (BACKEND): Remove $(CODYLIB).
25921 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25923 PR rtl-optimization/98334
25924 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
25925 Optimize (X - 1) * Y + Y to X * Y or (X + 1) * Y - Y to X * Y.
25927 2021-01-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
25929 * tree-inline.c (expand_call_inline): Restore input_location.
25930 Return result from recursive call.
25932 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
25934 PR tree-optimization/95401
25935 * config/aarch64/aarch64-sve-builtins.cc
25936 (gimple_folder::load_store_cookie): Use bits rather than bytes
25937 for the alignment argument to IFN_MASK_LOAD and IFN_MASK_STORE.
25938 * gimple-fold.c (gimple_fold_mask_load_store_mem_ref): Likewise.
25939 * tree-vect-stmts.c (vectorizable_store): Likewise.
25940 (vectorizable_load): Likewise.
25942 2021-01-04 Richard Biener <rguenther@suse.de>
25944 PR tree-optimization/98308
25945 * tree-vect-stmts.c (vectorizable_load): Set invariant mask
25948 2021-01-04 Jakub Jelinek <jakub@redhat.com>
25950 PR tree-optimization/95771
25951 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Handle types
25952 with precision smaller than int's precision and types with precision
25953 twice as large as long long. Formatting fixes.
25955 2021-01-04 Richard Biener <rguenther@suse.de>
25957 PR tree-optimization/98464
25958 * tree-ssa-sccvn.c (vn_valueize_for_srt): Rename from ...
25959 (vn_valueize_wrapper): ... this. Temporarily adjust vn_context_bb.
25960 (process_bb): Adjust.
25962 2021-01-04 Matthew Malcomson <matthew.malcomson@arm.com>
25965 * doc/invoke.texi (-fsanitize=address): Fix wording describing
25966 clash with -fsanitize=hwaddress.
25968 2021-01-04 Richard Biener <rguenther@suse.de>
25970 PR tree-optimization/98282
25971 * tree-ssa-sccvn.c (vn_get_stmt_kind): Classify tcc_reference on
25972 invariants as VN_NARY.
25974 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
25977 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Accept
25978 aarch64_simd_reg_or_zero for operand 2. Use the combinez patterns
25979 to handle zero operands.
25981 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
25983 * config/aarch64/aarch64.c (offset_6bit_signed_scaled_p): New function.
25984 (offset_6bit_unsigned_scaled_p): Fix typo in comment.
25985 (aarch64_sve_prefetch_operand_p): Accept MUL VLs in the range
25988 2021-01-04 Richard Biener <rguenther@suse.de>
25990 PR tree-optimization/98393
25991 * tree-vect-slp.c (vect_build_slp_tree): Properly zero matches
25992 when hitting the limit.
25994 2021-01-04 Richard Biener <rguenther@suse.de>
25996 PR tree-optimization/98291
25997 * tree-vect-loop.c (vectorizable_reduction): Bypass
25998 associativity check for SLP reductions with VF 1.
26000 2021-01-04 Jakub Jelinek <jakub@redhat.com>
26002 PR tree-optimization/96782
26003 * match.pd (x == ~x -> false, x != ~x -> true): New simplifications.
26005 2021-01-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
26007 * collect-utils.c (collect_execute): Check dumppfx.
26008 * collect2.c (maybe_run_lto_and_relink, do_link): Pass atsuffix
26009 to collect_execute.
26010 (do_link): Add new parameter atsuffix.
26011 (main): Handle -dumpdir option. Skip one argument for
26012 -o, -isystem and -B options.
26013 * gcc.c (make_at_file): New helper function.
26014 (close_at_file): Use it.
26016 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26018 * config/darwin.h (MIN_LD64_NO_COAL_SECTS): Adjust.
26019 Amend handling for LD64_VERSION fallback defaults.
26021 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26023 * config.gcc: Compute default version information
26024 from the configured target. Likewise defaults for
26026 * config/darwin10.h: Removed.
26027 * config/darwin12.h: Removed.
26028 * config/darwin9.h: Removed.
26029 * config/rs6000/darwin8.h: Removed.
26031 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26033 * config/darwin9.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
26035 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26037 * config/darwin9.h (STACK_CHECK_STATIC_BUILTIN): Move from here..
26038 * config/darwin.h (STACK_CHECK_STATIC_BUILTIN): .. to here.
26040 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26042 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move from
26044 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): ... to here.
26046 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26048 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move the spec
26049 for the Darwin10 unwinder stub from here ...
26050 * config/darwin.h (LINK_COMMAND_SPEC_A): ... to here.
26052 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26054 * config/darwin.h (DSYMUTIL_SPEC): Default to DWARF
26055 (ASM_DEBUG_SPEC):Only define if the assembler supports
26057 (PREFERRED_DEBUGGING_TYPE): Default to DWARF.
26058 (DARWIN_PREFER_DWARF): Define.
26059 * config/darwin9.h (PREFERRED_DEBUGGING_TYPE): Remove.
26060 (DARWIN_PREFER_DWARF): Likewise
26061 (DSYMUTIL_SPEC): Likewise.
26062 (COLLECT_RUN_DSYMUTIL): Likewise.
26063 (ASM_DEBUG_SPEC): Likewise.
26064 (ASM_DEBUG_OPTION_SPEC): Likewise.
26066 2021-01-02 Jan Hubicka <jh@suse.cz>
26068 * cfg.c (free_block): ggc_free bb.
26070 2021-01-01 Jakub Jelinek <jakub@redhat.com>
26072 * gcc.c (process_command): Update copyright notice dates.
26073 * gcov-dump.c (print_version): Ditto.
26074 * gcov.c (print_version): Ditto.
26075 * gcov-tool.c (print_version): Ditto.
26076 * gengtype.c (create_file): Ditto.
26077 * doc/cpp.texi: Bump @copying's copyright year.
26078 * doc/cppinternals.texi: Ditto.
26079 * doc/gcc.texi: Ditto.
26080 * doc/gccint.texi: Ditto.
26081 * doc/gcov.texi: Ditto.
26082 * doc/install.texi: Ditto.
26083 * doc/invoke.texi: Ditto.
26085 2021-01-01 Jakub Jelinek <jakub@redhat.com>
26087 * ChangeLog-2020: Rotate ChangeLog. New file.
26090 Copyright (C) 2021 Free Software Foundation, Inc.
26092 Copying and distribution of this file, with or without modification,
26093 are permitted in any medium without royalty provided the copyright
26094 notice and this notice are preserved.