1 2021-10-17 Aldy Hernandez <aldyh@redhat.com>
3 * tree-pass.h (make_pass_thread_jumps_full): New.
4 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Inline.
5 (try_thread_blocks): Add resolve and speed arguments.
6 (pass_thread_jumps::execute): Inline.
7 (do_early_thread_jumps): New.
8 (do_thread_jumps): New.
9 (make_pass_thread_jumps): Move.
10 (pass_early_thread_jumps::gate): Inline.
11 (pass_early_thread_jumps::execute): Inline.
12 (class pass_thread_jumps_full): New.
14 2021-10-16 Piotr Kubaj <pkubaj@FreeBSD.org>
16 * configure.ac: Treat powerpc64*-*-freebsd* the same as
18 * configure: Regenerate.
20 2021-10-16 H.J. Lu <hjl.tools@gmail.com>
22 * value-query.cc (get_ssa_name_ptr_info_nonnull): Change
23 set_ptr_nonull to set_ptr_nonnull in comments.
25 2021-10-16 Jan Hubicka <hubicka@ucw.cz>
27 PR tree-optimization/102720
28 * tree-ssa-structalias.c (compute_points_to_sets): Fix producing
29 of call used and clobbered sets.
31 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
33 * config/darwin.h (LINK_COMMAND_SPEC_A): Update 'r' handling to
34 skip gomp and itm when r or nodefaultlibs is given.
35 (DSYMUTIL_SPEC): Do not call dsymutil for '-r' link lines.
36 Update ordering of exclusions, remove duplicate 'v' addition
37 (collect2 will add this from the main command line).
39 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
41 * config/darwin-driver.c (darwin_driver_init): Revise comments, handle
42 filelist and framework options in specs instead of code.
43 * config/darwin.h (SUBTARGET_DRIVER_SELF_SPECS): Update to handle link
44 specs that are really driver ones.
45 (DARWIN_CC1_SPEC): Likewise.
47 (SYSROOT_SPEC): Append space.
48 (LINK_SYSROOT_SPEC): Remove most driver link specs.
49 (STANDARD_STARTFILE_PREFIX_2): Update link-related specs.
50 (STARTFILE_SPEC): Likewise.
51 (ASM_MMACOSX_VERSION_MIN_SPEC): Fix line wrap.
52 (ASM_SPEC): Update driver-related specs.
53 (ASM_FINAL_SPEC): Likewise.
54 * config/darwin.opt: Remove now unused option aliases.
55 * config/i386/darwin.h (EXTRA_ASM_OPTS): Ensure space after opt.
56 (ASM_SPEC): Update driver-related specs.
58 2021-10-15 Roger Sayle <roger@nextmovesoftware.com>
60 * config/i386/i386.c (ix86_hardreg_mov_ok): For vector modes,
61 allow standard_sse_constant_p immediate constants.
63 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
65 * config.gcc: Add tm-dwarf2.h to tm_d-file.
67 2021-10-15 Andrew MacLeod <amacleod@redhat.com>
69 * gimple-range-fold.h (gimple_range_ssa_p): Don't process names
70 that occur in abnormal phis.
71 * gimple-range.cc (gimple_ranger::range_on_edge): Return false for
72 abnormal and EH edges.
73 * gimple-ssa-evrp.c (rvrp_folder::value_of_expr): Ditto.
74 (rvrp_folder::value_on_edge): Ditto.
75 (rvrp_folder::value_of_stmt): Ditto.
76 (hybrid_folder::value_of_expr): Ditto for ranger queries.
77 (hybrid_folder::value_on_edge): Ditto.
78 (hybrid_folder::value_of_stmt): Ditto.
79 * value-query.cc (gimple_range_global): Always return a range if
80 the type is supported.
82 2021-10-15 John David Anglin <danglin@gcc.gnu.org>
84 * config/pa/pa.md: Consistently use "rG" constraint for copy
85 instruction in move patterns.
87 2021-10-15 Andrew Stubbs <ams@codesourcery.com>
89 * config/gcn/gcn-hsa.h (S_FIJI): Set unconditionally.
92 * config/gcn/gcn.c: Hard code SRAM ECC settings for old architectures.
93 * config/gcn/mkoffload.c (ELFABIVERSION_AMDGPU_HSA): Rename to ...
94 (ELFABIVERSION_AMDGPU_HSA_V3): ... this.
95 (ELFABIVERSION_AMDGPU_HSA_V4): New.
96 (SET_SRAM_ECC_UNSUPPORTED): New.
97 (copy_early_debug_info): Create elf flags to match the other objects.
98 (main): Just let the attribute flags pass through.
100 2021-10-15 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
102 * tree-loop-distribution.c (reduction_var_overflows_first):
103 Pass the type of reduction_var as first argument as it is also
104 done for the load type.
105 (loop_distribution::transform_reduction_loop): Add missing
106 TREE_TYPE while determining precission of reduction_var.
108 2021-10-15 Richard Biener <rguenther@suse.de>
110 * defaults.h (PREFERRED_DEBUGGING_TYPE): Choose DWARF2_DEBUG
112 * toplev.c (process_options): Warn when STABS debugging is
113 enabled but not the preferred format.
114 * config/pa/som.h (PREFERRED_DEBUGGING_TYPE): Define to
116 * config/pdp11/pdp11.h (PREFERRED_DEBUGGING_TYPE): Likewise.
118 2021-10-15 Richard Biener <rguenther@suse.de>
121 * tree-inline.c (copy_bb): Avoid underflowing nargs.
123 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
125 * config/i386/i386-expand.c (ix86_expand_vec_perm): Convert
126 HFmode input operand to HImode.
127 (ix86_vectorize_vec_perm_const): Likewise.
128 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1_hf):
130 (*avx512f_permvar_truncv8siv8hi_1_hf):
133 2021-10-15 Richard Biener <rguenther@suse.de>
136 * expmed.c (store_bit_field_1): Ensure a LHS subreg would
137 not create a paradoxical subreg.
139 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
141 * config/i386/i386-expand.c (ix86_expand_vector_init):
142 For half_vector concat for HFmode, handle them like HImode.
144 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
146 * tree-ssa-threadbackward.c (class back_threader): Add m_resolve.
147 (back_threader::back_threader): Same.
148 (back_threader::resolve_phi): Try to solve without looking back if
150 (back_threader::find_paths_to_names): Same.
151 (try_thread_blocks): Pass resolve argument to back threader.
152 (pass_early_thread_jumps::execute): Same.
154 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
156 * doc/invoke.texi: Remove max-fsm-thread-length,
157 max-fsm-thread-paths, and fsm-maximum-phi-arguments.
159 * tree-ssa-threadbackward.c (back_threader::back_threader): Remove
161 (back_threader_registry::back_threader_registry): Same.
162 (back_threader_profitability::profitable_path_p): Remove
163 param_max_fsm_thread-length.
164 (back_threader_registry::register_path): Remove
165 m_max_allowable_paths.
167 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
169 * tree-ssa-threadbackward.c (class back_threader): Make m_imports
171 (back_threader::~back_threader): Do not release m_path.
173 2021-10-14 Andrew MacLeod <amacleod@redhat.com>
175 PR tree-optimization/102738
176 * vr-values.c (simplify_using_ranges::simplify): Handle RSHIFT_EXPR.
178 2021-10-14 Kwok Cheung Yeung <kcy@codesourcery.com>
180 * omp-general.c (omp_check_context_selector): Move from c-omp.c.
181 (omp_mark_declare_variant): Move from c-omp.c.
182 (omp_context_name_list_prop): Update for Fortran strings.
183 * omp-general.h (omp_check_context_selector): New prototype.
184 (omp_mark_declare_variant): New prototype.
186 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
188 * config/rs6000/rs6000.c (rs6000_density_test): Move early
189 exit test further up the function.
191 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
193 * config/arm/arm.c (arm_add_stmt_cost): Delete.
194 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
196 2021-10-14 Martin Jambor <mjambor@suse.cz>
198 * doc/invoke.texi (Optimize Options): Add entry for
199 ipa-cp-recursive-freq-factor.
201 2021-10-14 Tamar Christina <tamar.christina@arm.com>
203 * match.pd: New rule.
205 2021-10-14 Jan Hubicka <hubicka@ucw.cz>
208 * ipa-modref-tree.h (modref_access_node::update2):
209 Also check that parm_offset is unchanged.
210 (modref_ref_node::insert_access): Fix updating of
213 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
215 * tree-ssa-threadbackward.c (back_threader::resolve_phi): Add
218 2021-10-14 Richard Biener <rguenther@suse.de>
220 PR tree-optimization/102659
221 * tree-if-conv.c (if_convertible_gimple_assign_stmt_p): Also
222 rewrite pointer typed undefined overflow operations.
223 (predicate_statements): Likewise. Make sure to emit invariant
224 conversions in the preheader.
225 * tree-vectorizer.c (vect_loop_vectorized_call): Look through
226 non-empty preheaders.
227 * tree-data-ref.c (dr_analyze_indices): Strip useless
228 conversions to the MEM_REF base type.
230 2021-10-14 Martin Liska <mliska@suse.cz>
232 * common.opt: Stop using AUTODETECT_VALUE
233 and use EnabledBy where possible.
234 * opts.c: Enable OPT_fvar_tracking with optimize >= 1.
235 * toplev.c (AUTODETECT_VALUE): Remove macro.
236 (process_options): Simplify by using EnabledBy and
237 OPT_fvar_tracking. Use OPTION_SET_P macro instead of
240 2021-10-14 Jonathan Wright <jonathan.wright@arm.com>
242 * config/aarch64/arm_neon.h (vld1_s8_x3): Use signed type for
244 (vld1_s32_x3): Likewise.
246 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
248 PR tree-optimization/102736
249 PR tree-optimization/102736
250 * gimple-range-path.cc (path_range_query::range_on_path_entry):
251 Assert that the requested range is defined outside the path.
252 (path_range_query::ssa_range_in_phi): Do not call
253 range_on_path_entry for SSA names that are defined within the
256 2021-10-14 Iain Sandoe <iain@sandoe.co.uk>
258 * config/darwin-driver.c (darwin_find_version_from_kernel):
259 Quote internal identifiers and avoid contractions in
261 (darwin_default_min_version): Likewise.
262 (darwin_driver_init): Likewise.
264 2021-10-14 Martin Jambor <mjambor@suse.cz>
266 * params.opt (ipa-cp-recursive-freq-factor): New.
267 * ipa-cp.c (ipcp_value): Switch to inline initialization. New members
268 scc_no, self_recursion_generated_level, same_scc and
269 self_recursion_generated_p.
270 (ipcp_lattice::add_value): Replaced parameter unlimited with
271 same_lat_gen_level, usit it determine limit of values and store it to
273 (ipcp_lattice<valtype>::print): Dump the new fileds.
274 (allocate_and_init_ipcp_value): Take same_lat_gen_level as a new
275 parameter and store it to the new value.
276 (self_recursively_generated_p): Removed.
277 (propagate_vals_across_arith_jfunc): Use self_recursion_generated_p
278 instead of self_recursively_generated_p, store self generation level
280 (value_topo_info<valtype>::add_val): Set scc_no.
281 (value_topo_info<valtype>::propagate_effects): Multiply frequencies of
282 recursively feeding values and self generated values by appropriate
285 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
287 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Remove
288 redundant test for flag_vect_cost_model.
290 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
292 * bitmap.c (debug): New overloaded function for auto_bitmaps.
293 * bitmap.h (debug): Same.
295 2021-10-14 Hongyu Wang <hongyu.wang@intel.com>
297 * config/i386/avx512fp16intrin.h (_mm512_mask_fcmadd_pch):
299 (_mm512_mask3_fcmadd_pch): Likewise.
300 (_mm512_mask_fmadd_pch): Likewise
301 (_mm512_mask3_fmadd_pch): Likewise
302 (_mm512_mask_fcmadd_round_pch): Likewise
303 (_mm512_mask3_fcmadd_round_pch): Likewise
304 (_mm512_mask_fmadd_round_pch): Likewise
305 (_mm512_mask3_fmadd_round_pch): Likewise
306 (_mm_mask_fcmadd_sch): Likewise
307 (_mm_mask3_fcmadd_sch): Likewise
308 (_mm_mask_fmadd_sch): Likewise
309 (_mm_mask3_fmadd_sch): Likewise
310 (_mm_mask_fcmadd_round_sch): Likewise
311 (_mm_mask3_fcmadd_round_sch): Likewise
312 (_mm_mask_fmadd_round_sch): Likewise
313 (_mm_mask3_fmadd_round_sch): Likewise
314 (_mm_fcmadd_round_sch): Likewise
315 * config/i386/avx512fp16vlintrin.h (_mm_mask_fmadd_pch):
317 (_mm_mask3_fmadd_pch): Likewise
318 (_mm256_mask_fmadd_pch): Likewise
319 (_mm256_mask3_fmadd_pch): Likewise
320 (_mm_mask_fcmadd_pch): Likewise
321 (_mm_mask3_fcmadd_pch): Likewise
322 (_mm256_mask_fcmadd_pch): Likewise
323 (_mm256_mask3_fcmadd_pch): Likewise
324 * config/i386/i386-builtin.def: Add mask3 builtin for complex
325 fma, and adjust mask_builtin to corresponding expander.
326 * config/i386/i386-expand.c (ix86_expand_round_builtin):
327 Skip eraseing embedded rounding for expanders that emits
329 * config/i386/sse.md (complexmove): New mode_attr.
330 (<avx512>_fmaddc_<mode>_mask1<round_expand_name>): New expander.
331 (<avx512>_fcmaddc_<mode>_mask1<round_expand_name>): Likewise.
332 (avx512fp16_fmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
333 (avx512fp16_fcmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
334 (avx512fp16_fcmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
335 (avx512fp16_fmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
336 * config/i386/subst.md (round_embedded_complex): New subst.
338 2021-10-14 John David Anglin <danglin@gcc.gnu.org>
340 * config/pa/pa.md (cbranchsf4): Disable if TARGET_SOFT_FLOAT.
341 (cbranchdf4): Likewise.
342 Add missing move patterns for TARGET_SOFT_FLOAT.
344 2021-10-13 Roger Sayle <roger@nextmovesoftware.com>
346 * config/i386/i386-expand.c (ix86_expand_vector_move): Use a
347 pseudo intermediate when moving a SUBREG into a hard register,
348 by checking ix86_hardreg_mov_ok.
349 (ix86_expand_vector_extract): Store zero-extended SImode
350 intermediate in a pseudo, then set target using a SUBREG_PROMOTED
352 * config/i386/sse.md (mov<VMOVE>_internal): Prevent CSE creating
353 complex (SUBREG) sets of (vector) hard registers before reload, by
354 checking ix86_hardreg_mov_ok.
356 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
358 * ctfc.h (enum ctf_dtu_d_union_enum): Remove redundant comma.
360 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
362 * dwarf2ctf.c (gen_ctf_array_type): Fix typo in comment.
364 2021-10-13 Martin Sebor <msebor@redhat.com>
367 * pointer-query.cc (compute_objsize_r): Handle named address spaces.
369 2021-10-13 Iain Sandoe <iain@sandoe.co.uk>
371 * collect2.c (is_lto_object_file): Release simple-object
372 resources, close files.
374 2021-10-13 Andre Vieira <andre.simoesdiasvieira@arm.com>
376 * config/arm/arm.c (thumb2_legitimate_address_p): Use VALID_MVE_MODE
377 when checking mve addressing modes.
378 (mve_vector_mem_operand): Fix the way we handle pre, post and offset
380 (arm_print_operand): Fix printing of POST_ and PRE_MODIFY.
381 * config/arm/mve.md: Use mve_memory_operand predicate everywhere where
382 there is a single Ux constraint.
384 2021-10-13 John David Anglin <danglin@gcc.gnu.org>
386 * config/pa/pa.md (muldi3): Add support for inlining 64-bit
387 multiplication on 32-bit PA 1.1 and 2.0 targets.
389 2021-10-13 Martin Liska <mliska@suse.cz>
391 PR gcov-profile/90364
392 * coverage.c (build_info): Emit checksum to the global variable.
393 (build_info_type): Add new field for checksum.
394 (coverage_obj_finish): Pass object_checksum.
395 (coverage_init): Use 0 as checksum for .gcno files.
396 * gcov-dump.c (dump_gcov_file): Dump also new checksum field.
397 * gcov.c (read_graph_file): Read also checksum.
398 * doc/invoke.texi: Document the behaviour change.
400 2021-10-13 Richard Biener <rguenther@suse.de>
402 * gimple-iterator.h (gsi_iterator_update): Add GSI_LAST_NEW_STMT,
403 start at integer value 2.
404 * gimple-iterator.c (gsi_insert_seq_nodes_before): Update
405 the iterator for GSI_LAST_NEW_STMT.
406 (gsi_insert_seq_nodes_after): Likewise.
407 * tree-if-conv.c (predicate_statements): Use GSI_LAST_NEW_STMT.
408 * tree-ssa.c (execute_update_addresses_taken): Correct bogus
409 arguments to gsi_replace.
411 2021-10-13 Martin Liska <mliska@suse.cz>
414 * common.opt: Use EnabledBy instead of detection in
415 finish_options and process_options.
416 * opts.c (finish_options): Remove handling of
417 x_flag_unroll_all_loops.
418 * toplev.c (process_options): Likewise for flag_web and
419 flag_rename_registers.
421 2021-10-13 Richard Biener <rguenther@suse.de>
423 PR tree-optimization/102659
424 * tree-if-conv.c (need_to_rewrite_undefined): New flag.
425 (if_convertible_gimple_assign_stmt_p): Mark the loop for
426 rewrite when stmts with undefined behavior on integer
428 (combine_blocks): Predicate also when we need to rewrite stmts.
429 (predicate_statements): Rewrite affected stmts to something
430 with well-defined behavior on overflow.
431 (tree_if_conversion): Initialize need_to_rewrite_undefined.
433 2021-10-13 Richard Biener <rguenther@suse.de>
436 * ipa-sra.c (ptr_parm_has_nonarg_uses): Fix volatileness
439 2021-10-13 Jose E. Marchesi <jose.marchesi@oracle.com>
441 * dwarf2ctf.c: Fix typo in comment.
443 2021-10-12 Bill Schmidt <wschmidt@linux.ibm.com>
446 * config/rs6000/altivec.h (vec_cpsgn): Swap operand order.
447 * config/rs6000/rs6000-overload.def (VEC_COPYSIGN): Use SKIP to
448 avoid generating an automatic #define of vec_cpsgn. Use the
449 correct built-in for V4SFmode that doesn't depend on VSX.
451 2021-10-12 Uroš Bizjak <ubizjak@gmail.com>
455 * config/i386/i386.md (*add<mode>_1_slp): Rewrite as
456 define_insn_and_split pattern. Add alternative 1 and split it
457 post reload to insert operand 1 into the low part of operand 0.
458 (*sub<mode>_1_slp): Ditto.
459 (*and<mode>_1_slp): Ditto.
460 (*<any_or:code><mode>_1_slp): Ditto.
461 (*ashl<mode>3_1_slp): Ditto.
462 (*<any_shiftrt:insn><mode>3_1_slp): Ditto.
463 (*<any_rotate:insn><mode>3_1_slp): Ditto.
464 (*neg<mode>_1_slp): New insn_and_split pattern.
465 (*one_cmpl<mode>_1_slp): Ditto.
467 2021-10-12 David Edelsohn <dje.gcc@gmail.com>
469 * doc/install.texi: Update MinGW and mingw-64 Binaries
472 2021-10-12 Daniel Le Duc Khoi Nguyen <greenrecyclebin@gmail.com>
474 * doc/extend.texi (Common Variable Attributes): Fix typos in
475 alloc_size documentation.
477 2021-10-12 Richard Biener <rguenther@suse.de>
479 PR tree-optimization/102696
480 * tree-vect-slp.c (vect_build_slp_tree_2): Properly mark
481 the tree fatally failed when we reject a BIT_FIELD_REF.
483 2021-10-12 Richard Biener <rguenther@suse.de>
485 PR tree-optimization/102572
486 * tree-vect-stmts.c (vect_build_gather_load_calls): When
487 gathering the vectorized defs for the mask pass in the
488 desired mask vector type so invariants will be handled
491 2021-10-12 Tamar Christina <tamar.christina@arm.com>
493 * config/aarch64/aarch64-sve.md (*fcm<cmp_op><mode>_bic_combine,
494 *fcm<cmp_op><mode>_nor_combine, *fcmuo<mode>_bic_combine,
495 *fcmuo<mode>_nor_combine): New.
497 2021-10-12 Eric Botcazou <ebotcazou@adacore.com>
500 * config/sparc/sparc-modes.def (OI): New integer mode.
502 2021-10-12 Jakub Jelinek <jakub@redhat.com>
504 * gimple-fold.h (clear_padding_type_may_have_padding_p): Declare.
505 * gimple-fold.c (clear_padding_type_may_have_padding_p): No longer
508 2021-10-12 Jakub Jelinek <jakub@redhat.com>
510 * tree-vectorizer.h (loop_cost_model): New function.
511 (unlimited_cost_model): Use it.
512 * tree-vect-loop.c (vect_analyze_loop_costing): Use loop_cost_model
513 call instead of flag_vect_cost_model.
514 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
515 (vect_prune_runtime_alias_test_list): Likewise. Also use it instead
516 of flag_simd_cost_model.
518 2021-10-12 liuhongt <hongtao.liu@intel.com>
521 * config/i386/i386-expand.c (emit_reduc_half): Handle
523 * config/i386/mmx.md (reduc_<code>_scal_v4qi): New expander.
524 (reduc_plus_scal_v4qi): Ditto.
526 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
528 * config/rs6000/smmintrin.h (_mm_cmpeq_epi64, _mm_cmpgt_epi64,
529 _mm_mullo_epi32, _mm_mul_epi32, _mm_packus_epi32): New.
530 * config/rs6000/nmmintrin.h: Copy from i386, tweak to suit.
532 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
534 * config/rs6000/smmintrin.h (_mm_cvtepi8_epi16, _mm_cvtepi8_epi32,
535 _mm_cvtepi8_epi64, _mm_cvtepi16_epi32, _mm_cvtepi16_epi64,
536 _mm_cvtepi32_epi64, _mm_cvtepu8_epi16, _mm_cvtepu8_epi32,
537 _mm_cvtepu8_epi64, _mm_cvtepu16_epi32, _mm_cvtepu16_epi64,
538 _mm_cvtepu32_epi64): New.
540 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
542 * config/rs6000/smmintrin.h (_mm_test_all_zeros,
543 _mm_test_all_ones, _mm_test_mix_ones_zeros): Rewrite as macro.
545 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
547 * config/rs6000/smmintrin.h (_mm_min_epi8, _mm_min_epu16,
548 _mm_min_epi32, _mm_min_epu32, _mm_max_epi8, _mm_max_epu16,
549 _mm_max_epi32, _mm_max_epu32): New.
551 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
553 * ipa-modref-tree.h (struct modref_access_node): Revert
555 (struct modref_ref_node): Likewise.
557 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
559 * ipa-modref-tree.h (modref_tree::global_access_p): New member
562 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
563 ignore_stores_eaf_flags): Move to ipa-modref.h
564 (remove_useless_eaf_flags): Remove early exit on NOCLOBBER.
565 (modref_summary::global_memory_read_p): New member function.
566 (modref_summary::global_memory_written_p): New member function.
567 * ipa-modref.h (modref_summary::global_memory_read_p,
568 modref_summary::global_memory_written_p): Declare.
569 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
570 ignore_stores_eaf_flags): move here.
571 * tree-ssa-structalias.c: Include ipa-modref-tree.h, ipa-modref.h
573 (handle_rhs_call): Rewrite.
574 (handle_call_arg): New function.
575 (determine_global_memory_access): New function.
576 (handle_const_call): Remove
577 (handle_pure_call): Remove
578 (find_func_aliases_for_call): Update use of handle_rhs_call.
579 (compute_points_to_sets): Handle global memory acccesses
582 2021-10-11 Diane Meirowitz <diane.meirowitz@oracle.com>
584 * doc/invoke.texi: Add link to UndefinedBehaviorSanitizer
585 documentation, mention UBSAN_OPTIONS, similar to what is done
586 for AddressSanitizer.
588 2021-10-11 Richard Biener <rguenther@suse.de>
591 * internal-fn.c (expand_DEFERRED_INIT): Check for mode
592 availability before building an integer type for storage
595 2021-10-11 Richard Biener <rguenther@suse.de>
598 * gimple.c (gimple_call_fnspec): Do not mark operator new/delete
601 2021-10-11 Martin Liska <mliska@suse.cz>
603 * common.opt: Remove Init(2) for some options.
604 * toplev.c (process_options): Do not use AUTODETECT_VALUE, but
605 use rather OPTION_SET_P.
607 2021-10-11 Martin Liska <mliska@suse.cz>
609 * common.opt: Remove usage of IRA_REGION_AUTODETECT.
610 * flag-types.h (enum ira_region): Likewise.
611 * toplev.c (process_options): Use OPTION_SET_P instead of
612 IRA_REGION_AUTODETECT.
614 2021-10-11 Jakub Jelinek <jakub@redhat.com>
616 * omp-low.c (omp_runtime_api_call): Handle omp_get_max_teams,
617 omp_[sg]et_teams_thread_limit and omp_set_num_teams.
619 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
621 * config/s390/s390-protos.h (s390_rawmemchr): Add prototype.
622 * config/s390/s390.c (s390_rawmemchr): New function.
623 * config/s390/s390.md (rawmemchr<SINT:mode>): New expander.
624 * config/s390/vector.md (@vec_vfees<mode>): Basically a copy of
625 the pattern vfees<mode> from vx-builtins.md.
626 * config/s390/vx-builtins.md (*vfees<mode>): Remove.
628 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
630 * builtins.c (get_memory_rtx): Change to external linkage.
631 * builtins.h (get_memory_rtx): Add function prototype.
632 * doc/md.texi (rawmemchr<mode>): Document.
633 * internal-fn.c (expand_RAWMEMCHR): Define.
634 * internal-fn.def (RAWMEMCHR): Add.
635 * optabs.def (rawmemchr_optab): Add.
636 * tree-loop-distribution.c (find_single_drs): Change return code
637 behaviour by also returning true if no single store was found
639 (loop_distribution::classify_partition): Respect the new return
640 code behaviour of function find_single_drs.
641 (loop_distribution::execute): Call new function
642 transform_reduction_loop in order to replace rawmemchr or strlen
643 like loops by calls into builtins.
644 (generate_reduction_builtin_1): New function.
645 (generate_rawmemchr_builtin): New function.
646 (generate_strlen_builtin_1): New function.
647 (generate_strlen_builtin): New function.
648 (generate_strlen_builtin_using_rawmemchr): New function.
649 (reduction_var_overflows_first): New function.
650 (determine_reduction_stmt_1): New function.
651 (determine_reduction_stmt): New function.
652 (loop_distribution::transform_reduction_loop): New function.
654 2021-10-11 Martin Liska <mliska@suse.cz>
656 * tree.c (cl_option_hasher::hash): Use cl_optimization_hash
657 and remove legacy hashing code.
659 2021-10-11 Kito Cheng <kito.cheng@sifive.com>
662 * builtins.c (maybe_emit_call_builtin___clear_cache): Allow
663 CONST_INT for BEGIN and END, and use gcc_assert rather than
666 2021-10-10 Jakub Jelinek <jakub@redhat.com>
669 * var-tracking.c (add_stores): For cselib_sp_derived_value_p values
670 use MO_VAL_SET if loc is not sp.
672 2021-10-10 Andrew Pinski <apinski@marvell.com>
674 PR tree-optimization/102622
675 * match.pd: Swap the order of a?pow2cst:0 and a?-1:0 transformations.
676 Swap the order of a?0:pow2cst and a?0:-1 transformations.
678 2021-10-09 liuhongt <hongtao.liu@intel.com>
681 * config/i386/i386-expand.c (ix86_valid_mask_cmp_mode): Handle
683 (ix86_use_mask_cmp_p): Ditto.
684 (ix86_expand_sse_movcc): Ditto.
685 * config/i386/i386.md (setcc_hf_mask): New define_insn.
687 (UNSPEC_MOVCC_MASK): New unspec.
688 * config/i386/sse.md (UNSPEC_PCMP): Move to i386.md.
690 2021-10-08 Vladimir N. Makarov <vmakarov@redhat.com>
692 PR rtl-optimization/102627
693 * lra-constraints.c (split_reg): Use at least natural mode of hard reg.
695 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
697 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Grow
700 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
702 * value-range.cc (irange::debug): New.
703 * value-range.h (irange::debug): New.
705 2021-10-08 Richard Sandiford <richard.sandiford@arm.com>
707 PR tree-optimization/102385
708 * predict.h (change_edge_frequency): Declare.
709 * predict.c (change_edge_frequency): New function.
710 * tree-ssa-loop-manip.h (tree_transform_and_unroll_loop): Remove
712 (tree_unroll_loop): Likewise.
713 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Update accordingly.
714 * tree-predcom.c (pcom_worker::tree_predictive_commoning_loop):
716 * tree-ssa-loop-prefetch.c (loop_prefetch_arrays): Likewise.
717 * tree-ssa-loop-manip.c (tree_unroll_loop): Likewise.
718 (tree_transform_and_unroll_loop): Likewise. Use single_dom_exit
719 to retrieve the exit edges. Make all the old profile update code
720 conditional on !single_loop_p -- the case it was written for --
721 and use a different approach for the single-loop case.
723 2021-10-08 Martin Liska <mliska@suse.cz>
725 * config/alpha/alpha.c (alpha_option_override): Use new macro
727 * config/arc/arc.c (arc_override_options): Likewise.
728 * config/arm/arm.c (arm_option_override): Likewise.
729 * config/bfin/bfin.c (bfin_load_pic_reg): Likewise.
730 * config/c6x/c6x.c (c6x_option_override): Likewise.
731 * config/csky/csky.c: Likewise.
732 * config/darwin.c (darwin_override_options): Likewise.
733 * config/frv/frv.c (frv_option_override): Likewise.
734 * config/i386/djgpp.h: Likewise.
735 * config/i386/i386.c (ix86_stack_protect_guard): Likewise.
736 (ix86_max_noce_ifcvt_seq_cost): Likewise.
737 * config/ia64/ia64.c (ia64_option_override): Likewise.
738 (ia64_override_options_after_change): Likewise.
739 * config/m32c/m32c.c (m32c_option_override): Likewise.
740 * config/m32r/m32r.c (m32r_init): Likewise.
741 * config/m68k/m68k.c (m68k_option_override): Likewise.
742 * config/microblaze/microblaze.c (microblaze_option_override): Likewise.
743 * config/mips/mips.c (mips_option_override): Likewise.
744 * config/nios2/nios2.c (nios2_option_override): Likewise.
745 * config/nvptx/nvptx.c (nvptx_option_override): Likewise.
746 * config/pa/pa.c (pa_option_override): Likewise.
747 * config/riscv/riscv.c (riscv_option_override): Likewise.
748 * config/rs6000/aix71.h: Likewise.
749 * config/rs6000/aix72.h: Likewise.
750 * config/rs6000/aix73.h: Likewise.
751 * config/rs6000/rs6000.c (darwin_rs6000_override_options): Likewise.
752 (rs6000_override_options_after_change): Likewise.
753 (rs6000_linux64_override_options): Likewise.
754 (glibc_supports_ieee_128bit): Likewise.
755 (rs6000_option_override_internal): Likewise.
756 (rs6000_file_start): Likewise.
757 (rs6000_darwin_file_start): Likewise.
758 * config/rs6000/rtems.h: Likewise.
759 * config/rs6000/sysv4.h: Likewise.
760 * config/rs6000/vxworks.h (SUB3TARGET_OVERRIDE_OPTIONS): Likewise.
761 * config/s390/s390.c (s390_option_override): Likewise.
762 * config/sh/linux.h: Likewise.
763 * config/sh/netbsd-elf.h (while): Likewise.
764 * config/sh/sh.c (sh_option_override): Likewise.
765 * config/sol2.c (solaris_override_options): Likewise.
766 * config/sparc/sparc.c (sparc_option_override): Likewise.
767 * config/tilegx/tilegx.c (tilegx_option_override): Likewise.
768 * config/visium/visium.c (visium_option_override): Likewise.
769 * config/vxworks.c (vxworks_override_options): Likewise.
770 * lto-opts.c (lto_write_options): Likewise.
771 * omp-expand.c (expand_omp_simd): Likewise.
772 * omp-general.c (omp_max_vf): Likewise.
773 * omp-offload.c (oacc_xform_loop): Likewise.
774 * opts.h (OPTION_SET_P): Likewise.
775 * targhooks.c (default_max_noce_ifcvt_seq_cost): Likewise.
776 * toplev.c (process_options): Likewise.
777 * tree-predcom.c: Likewise.
778 * tree-sra.c (analyze_all_variable_accesses): Likewise.
780 2021-10-08 liuhongt <hongtao.liu@intel.com>
783 * config/i386/i386.c (ix86_optab_supported_p):
784 Return true for HFmode.
785 * match.pd: Simplify (_Float16) ceil ((double) x) to
786 __builtin_ceilf16 (a) when a is _Float16 type and
787 direct_internal_fn_supported_p.
789 2021-10-08 liuhongt <hongtao.liu@intel.com>
792 * config/i386/i386-expand.c (emit_reduc_half): Hanlde V4HImode.
793 * config/i386/mmx.md (reduc_plus_scal_v4hi): New.
794 (reduc_<code>_scal_v4hi): New.
796 2021-10-08 liuhongt <hongtao.liu@intel.com>
798 * common.opt (ftree-vectorize): Add Var(flag_tree_vectorize).
799 * doc/invoke.texi (Options That Control Optimization): Update
801 * opts.c (default_options_table): Enable auto-vectorization at
802 O2 with very-cheap cost model.
803 (finish_options): Use cheap cost model for
804 explicit -ftree{,-loop}-vectorize.
806 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
808 * ctfc.c (ctfc_delete_container): Free hash table contents.
810 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
812 * toplev.c (process_options): Do not warn for GNU GIMPLE.
814 2021-10-07 Siddhesh Poyarekar <siddhesh@gotplt.org>
816 * tree-object-size.c (addr_object_size,
817 compute_builtin_object_size): Drop PDECL and POFF arguments.
818 (addr_object_size): Adjust calls.
819 * tree-object-size.h (compute_builtin_object_size): Drop PDECL
822 2021-10-07 Roger Sayle <roger@nextmovesoftware.com>
824 * rtl.def (SMUL_HIGHPART, UMUL_HIGHPART): New RTX codes for
825 representing signed and unsigned high-part multiplication resp.
826 * simplify-rtx.c (simplify_binary_operation_1) [SMUL_HIGHPART,
827 UMUL_HIGHPART]: Simplify high-part multiplications by zero.
828 [SS_PLUS, US_PLUS, SS_MINUS, US_MINUS, SS_MULT, US_MULT,
829 SS_DIV, US_DIV]: Similar simplifications for saturating
831 (simplify_const_binary_operation) [SS_PLUS, US_PLUS, SS_MINUS,
832 US_MINUS, SS_MULT, US_MULT, SMUL_HIGHPART, UMUL_HIGHPART]:
833 Implement compile-time evaluation for constant operands.
834 * dwarf2out.c (mem_loc_descriptor): Skip SMUL_HIGHPART and
836 * doc/rtl.texi (smul_highpart, umul_highpart): Document RTX codes.
837 * doc/md.texi (smul@var{m}3_highpart, umul@var{m3}_highpart):
838 Mention the new smul_highpart and umul_highpart RTX codes.
839 * doc/invoke.texi: Silence @xref "compilation" warnings.
841 2021-10-07 Martin Jambor <mjambor@suse.cz>
844 * ipa-prop.c (ipa_edge_args_sum_t::duplicate): Also handle the
845 case when the source reference description corresponds to a
846 referance taken in a function src->caller is inlined to.
848 2021-10-07 Jan Hubicka <hubicka@ucw.cz>
851 * ipa-modref-tree.h (modref_access_node::contains_p): Handle offsets
853 (modref_access_node::try_merge_with): Add sanity check that there
854 are no redundant entries in the list.
856 2021-10-07 Richard Biener <rguenther@suse.de>
858 PR tree-optimization/102608
859 * tree-ssa-sccvn.c (visit_stmt): Drop .DEFERRED_INIT to
862 2021-10-07 Martin Liska <mliska@suse.cz>
864 * toplev.c (toplev::main): Make
865 save_opt_decoded_options a pointer type
866 * toplev.h: Likewise.
868 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
870 * config/gcn/gcn-valu.md (gather<mode>_insn_2offsets<exec>): Apply
871 HAVE_GCN_ASM_GLOBAL_LOAD_FIXED.
872 (scatter<mode>_insn_2offsets<exec_scatter>): Likewise.
874 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
876 * config/gcn/gcn-hsa.h (SRAMOPT): Include the whole option string.
877 Adjust for new -msram-ecc=any behaviour.
878 (ASM_SPEC): Adjust -mxnack and -msram-ecc usage.
879 * config/gcn/gcn.c (output_file_start): Implement -msram-ecc=any.
880 * config/gcn/mkoffload.c (EF_AMDGPU_XNACK): Rename to ...
881 (EF_AMDGPU_XNACK_V3): ... this.
882 (EF_AMDGPU_SRAM_ECC): Rename to ...
883 (EF_AMDGPU_SRAM_ECC_V3): ... this.
884 (EF_AMDGPU_FEATURE_XNACK_V4): New.
885 (EF_AMDGPU_FEATURE_XNACK_UNSUPPORTED_V4): New.
886 (EF_AMDGPU_FEATURE_XNACK_ANY_V4): New.
887 (EF_AMDGPU_FEATURE_XNACK_OFF_V4): New.
888 (EF_AMDGPU_FEATURE_XNACK_ON_V4): New.
889 (EF_AMDGPU_FEATURE_SRAMECC_V4): New.
890 (EF_AMDGPU_FEATURE_SRAMECC_UNSUPPORTED_V4): New.
891 (EF_AMDGPU_FEATURE_SRAMECC_ANY_V4): New.
892 (EF_AMDGPU_FEATURE_SRAMECC_OFF_V4): New.
893 (EF_AMDGPU_FEATURE_SRAMECC_ON_V4): New.
895 (SET_XNACK_OFF): New.
897 (SET_SRAM_ECC_ON): New.
898 (SET_SRAM_ECC_ANY): New.
899 (SET_SRAM_ECC_OFF): New.
900 (TEST_SRAM_ECC_ANY): New.
901 (TEST_SRAM_ECC_ON): New.
902 (main): Implement HSACOv4 and -msram-ecc=any.
904 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
906 * config.in: Regenerate.
907 * config/gcn/gcn-hsa.h (X_FIJI): New macro.
911 (A_FIJI): Rename to ...
913 (A_900): Rename to ...
915 (A_906): Rename to ...
917 (A_908): Rename to ...
919 (SRAMOPT): New macro.
920 (ASM_SPEC): Adjust xnack option usage.
921 * config/gcn/gcn.c (output_file_start): Adjust amdgcn_target usage.
922 * configure: Regenerate.
923 * configure.ac: Detect LLVM assembler dialect.
925 2021-10-07 Richard Biener <rguenther@suse.de>
927 * tree-pretty-print.c (dump_generic_node): Do not elide
928 printing '&' when dumping with -gimple.
930 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
932 * gimple-range-cache.cc (non_null_ref::adjust_range): Call new
934 * gimple-range-fold.cc (adjust_pointer_diff_expr): Ditto.
935 (adjust_imagpart_expr): Ditto.
936 * value-range.cc (irange::irange_intersect): Call new routine if
937 RHS is a single pair.
938 (irange::intersect): New wide_int version.
939 * value-range.h (class irange): New prototype.
941 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
943 * gimple-range-edge.cc (gimple_outgoing_range::gimple_outgoing_range):
944 Add parameter to limit size when recognizing switches.
945 (gimple_outgoing_range::edge_range_p): Check size limit.
946 * gimple-range-edge.h (gimple_outgoing_range): Add size field.
947 * gimple-range-gori.cc (gori_map::calculate_gori): Ignore switches
948 that exceed the size limit.
949 (gori_compute::gori_compute): Add initializer.
950 * params.opt (evrp-switch-limit): New.
951 * doc/invoke.texi: Update docs.
953 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
955 * value-range.h (irange::set_varying): Use TYPE_MIN_VALUE and
956 TYPE_MAX_VALUE instead of creating new trees when possible.
958 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
960 * gimple-range-cache.cc (non_null_ref::adjust_range): Check for
961 zero and non-zero more efficently.
963 2021-10-06 Richard Biener <rguenther@suse.de>
966 * dumpfile.h (TDF_GIMPLE_VAL): New.
967 (dump_flag): Re-order and adjust TDF_* flags. Make
968 the enum uint32_t. Use std::underlying_type in the
970 (optgroup_flag): Likewise for the operator overloads.
971 * tree-pretty-print.c (dump_generic_node): Wrap ADDR_EXPR
972 in _Literal if TDF_GIMPLE_VAL.
973 * gimple-pretty-print.c (dump_gimple_assign): Add
974 TDF_GIMPLE_VAL to flags when dumping operands where only
975 is_gimple_val are allowed.
976 (dump_gimple_cond): Likewise.
978 2021-10-06 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
980 * gimple-isel.cc (gimple_expand_vec_cond_expr): Remove redundant if
983 2021-10-05 qing zhao <qing.zhao@oracle.com>
986 * gimplify.c (gimplify_decl_expr): Not add initialization for an
987 auto variable when it has been initialized by frontend.
989 2021-10-05 Aldy Hernandez <aldyh@redhat.com>
991 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
994 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
996 * common/config/avr/avr-common.c (avr_handle_option): Mark
997 argument as ATTRIBUTE_UNUSED.
999 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
1001 * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC):
1002 Undefine before redefinition.
1004 2021-10-05 Richard Biener <rguenther@suse.de>
1006 * toplev.c (no_backend): Remove global var.
1007 (process_options): Pass in no_backend, move post_options
1008 langhook call to toplev::main.
1009 (do_compile): Pass in no_backend, move process_options call
1011 (toplev::run_self_tests): Check no_backend at the caller.
1012 (toplev::main): Call post_options and process_options
1013 split out from do_compile, do self-tests only if
1014 no_backend is initialized.
1016 2021-10-05 Richard Biener <rguenther@suse.de>
1018 * tree-cfg.c (dump_function_to_file): Dump the UID of the
1019 function as part of the name when requested.
1020 * tree-pretty-print.c (dump_function_name): Dump the UID when
1021 requested and the langhook produced the actual name.
1023 2021-10-05 Richard Biener <rguenther@suse.de>
1025 PR middle-end/102587
1026 PR middle-end/102285
1027 * internal-fn.c (expand_DEFERRED_INIT): Fall back to
1028 zero-initialization as last resort, use the constant
1029 size as given by the DEFERRED_INIT argument to build
1032 2021-10-04 Marek Polacek <polacek@redhat.com>
1035 * doc/invoke.texi: Document -Warray-compare.
1037 2021-10-04 Richard Biener <rguenther@suse.de>
1039 * gimplify.c (is_var_need_auto_init): DECL_HARD_REGISTER
1040 variables are not to be initialized.
1042 2021-10-04 Richard Biener <rguenther@suse.de>
1044 * expr.h (non_mem_decl_p): Declare.
1045 (mem_ref_refers_to_non_mem_p): Likewise.
1046 * expr.c (non_mem_decl_p): Export.
1047 (mem_ref_refers_to_non_mem_p): Likewise.
1048 * internal-fn.c (expand_DEFERRED_INIT): Do not expand the LHS
1049 but check the base with mem_ref_refers_to_non_mem_p
1052 2021-10-04 Richard Biener <rguenther@suse.de>
1054 PR tree-optimization/102570
1055 * tree-ssa-sccvn.h (vn_reference_op_struct): Document
1056 we are using clique for the internal function code.
1057 * tree-ssa-sccvn.c (vn_reference_op_eq): Compare the
1058 internal function code.
1059 (print_vn_reference_ops): Print the internal function code.
1060 (vn_reference_op_compute_hash): Hash it.
1061 (copy_reference_ops_from_call): Record it.
1062 (visit_stmt): Remove the restriction around internal function
1064 (fully_constant_vn_reference_p): Use fold_const_call and handle
1066 (vn_reference_eq): Compare call return types.
1067 * tree-ssa-pre.c (create_expression_by_pieces): Handle
1068 generating calls to internal functions.
1069 (compute_avail): Remove the restriction around internal function
1072 2021-10-04 Aldy Hernandez <aldyh@redhat.com>
1074 PR tree-optimization/102560
1075 * gimple-ssa-warn-alloca.c (alloca_call_type): Remove static
1076 marker for invalid_range.
1078 2021-10-04 Richard Biener <rguenther@suse.de>
1080 PR middle-end/102587
1081 * internal-fn.c (expand_DEFERRED_INIT): Guard register
1082 initialization path an avoid initializing VLA registers
1085 2021-10-04 Eric Botcazou <ebotcazou@adacore.com>
1087 * config/rs6000/vxworks.h (TARGET_INIT_LIBFUNCS): Delete.
1089 2021-10-03 Martin Liska <mliska@suse.cz>
1091 * toplev.c (toplev::main): Check opt_index if it is a part
1094 2021-10-02 Aldy Hernandez <aldyh@redhat.com>
1096 PR tree-optimization/102563
1097 * range-op.cc (operator_lshift::op1_range): Do not clobber
1100 2021-10-02 Martin Liska <mliska@suse.cz>
1102 * toplev.c (toplev::main): save_decoded_options[0] is program
1103 name and so it should be skipped.
1105 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
1107 PR tree-optimization/102546
1108 * range-op.cc (operator_lshift::op1_range): Teach range-ops that
1109 X << Y is non-zero implies X is also non-zero.
1111 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1113 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
1115 * config/aarch64/aarch64-tune.md: Regenerate.
1116 * doc/invoke.texi: Update docs.
1118 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1120 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
1122 * config/aarch64/aarch64-tune.md: Regenerate.
1123 * doc/invoke.texi: Update docs.
1125 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1127 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
1129 * config/aarch64/aarch64-tune.md: Regenerate.
1130 * doc/invoke.texi: Update docs.
1132 2021-10-01 Martin Sebor <msebor@redhat.com>
1135 * doc/invoke.texi (-Waddress): Update.
1136 * gengtype.c (write_types): Avoid -Waddress.
1137 * poly-int.h (POLY_SET_COEFF): Avoid using null.
1139 2021-10-01 John David Anglin <danglin@gcc.gnu.org>
1142 * config/pa/pa.c (pa_option_override): Default to dwarf version 4
1145 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1147 * config/aarch64/aarch64.h (AARCH64_FL_V9): Update value.
1149 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
1151 * gimple-range-path.cc (path_range_query::compute_ranges): Use
1153 * gimple-range-path.h (class path_range_query): Remove shadowed
1155 (path_range_query::get_path_oracle): New.
1157 2021-10-01 Jakub Jelinek <jakub@redhat.com>
1158 Richard Biener <rguenther@suse.de>
1161 * doc/invoke.texi (-fsanitize=integer-divide-by-zero): Remove
1162 INT_MIN / -1 division detection from here ...
1163 (-fsanitize=signed-integer-overflow): ... and add it here.
1165 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1167 * config/aarch64/aarch64-arches.def (AARCH64_ARCH): Added
1169 * config/aarch64/aarch64.h (AARCH64_FL_V9): New.
1170 (AARCH64_FL_FOR_ARCH9): New flags for Armv9-A.
1171 (AARCH64_ISA_V9): New ISA flag.
1172 * doc/invoke.texi: Update docs.
1174 2021-10-01 Martin Liska <mliska@suse.cz>
1176 * toplev.c (toplev::main): Save decoded optimization options.
1177 * toplev.h (save_opt_decoded_options): New.
1178 * doc/extend.texi: Be more clear about optimize and target
1181 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
1183 * explow.c: Include langhooks.h.
1184 (set_stack_check_libfunc): Build a proper function type.
1186 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
1189 * config/i386/i386.c (legitimate_pic_address_disp_p): For PE-COFF do
1190 not return true for external weak function symbols in medium model.
1192 2021-10-01 Jakub Jelinek <jakub@redhat.com>
1194 * tree.h (OMP_CLAUSE_ORDER_REPRODUCIBLE): Define.
1195 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_ORDER>: Print
1196 reproducible: for OMP_CLAUSE_ORDER_REPRODUCIBLE.
1197 * omp-general.c (omp_extract_for_data): If OMP_CLAUSE_ORDER is seen
1198 without OMP_CLAUSE_ORDER_UNCONSTRAINED, overwrite sched_kind to
1199 OMP_CLAUSE_SCHEDULE_STATIC.
1201 2021-10-01 Richard Biener <rguenther@suse.de>
1203 PR middle-end/102518
1204 * tree-inline.c (setup_one_parameter): Avoid substituting
1205 an invariant into contexts where a GIMPLE register is not valid.
1207 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1209 * config/arm/arm-cpus.in: Add Cortex-R52+ CPU.
1210 * config/arm/arm-tables.opt: Regenerate.
1211 * config/arm/arm-tune.md: Regenerate.
1212 * doc/invoke.texi: Update docs.
1214 2021-09-30 Uroš Bizjak <ubizjak@gmail.com>
1217 * config/i386/i386.md
1218 (sign_extend:WIDE (any_logic:NARROW (memory, immediate)) splitters):
1221 2021-09-30 Tobias Burnus <tobias@codesourcery.com>
1223 * omp-low.c (omp_runtime_api_call): Add omp_aligned_{,c}alloc and
1224 omp_{c,re}alloc, fix omp_alloc/omp_free.
1226 2021-09-30 Martin Liska <mliska@suse.cz>
1228 * defaults.h (ASM_OUTPUT_ASCII): Do not hide global variable
1229 asm_out_file and stream directly to MYFILE.
1231 2021-09-30 Richard Biener <rguenther@suse.de>
1233 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
1234 Restore and fix condition under which we apply npeel to
1235 the DRs misalignment value.
1237 2021-09-30 Richard Biener <rguenther@suse.de>
1239 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
1240 Fix npeel check for variable amount of peeling.
1242 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
1244 * lto-wrapper.c (run_gcc): Plug snprintf overflow.
1246 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
1248 * gimple-range.cc (gimple_ranger::debug): New.
1249 * gimple-range.h (class gimple_ranger): Add debug.
1251 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
1253 PR middle-end/102519
1254 * tree-vrp.c (hybrid_threader::~hybrid_threader): Free m_query.
1256 2021-09-29 Indu Bhagat <indu.bhagat@oracle.com>
1259 * btfout.c (GTY): Add GTY (()) albeit for cosmetic only purpose.
1260 (btf_finalize): Empty the hash_map btf_var_ids.
1262 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
1264 * tree-vrp.c (thread_through_all_blocks): Return bool.
1265 (execute_vrp_threader): Return TODO_* flags.
1266 (pass_data_vrp_threader): Set todo_flags_finish to 0.
1268 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
1270 * timevar.def (TV_TREE_VRP_THREADER): New.
1271 * tree-vrp.c: Use TV_TREE_VRP_THREADER for VRP threader pass.
1273 2021-09-29 David Faust <david.faust@oracle.com>
1275 * config.gcc (bpf-*-*): Do not overwrite extra_headers.
1277 2021-09-29 Jonathan Wright <jonathan.wright@arm.com>
1279 * config/aarch64/aarch64-builtins.c (TYPES_BINOP_PPU): Define
1280 new type qualifier enum.
1281 (TYPES_TERNOP_SSSU): Likewise.
1282 (TYPES_TERNOP_PPPU): Likewise.
1283 * config/aarch64/aarch64-simd-builtins.def: Define PPU, SSU,
1284 PPPU and SSSU builtin generator macros for qtbl1 and qtbx1
1286 * config/aarch64/arm_neon.h (vqtbl1_p8): Use type-qualified
1287 builtin and remove casts.
1288 (vqtbl1_s8): Likewise.
1289 (vqtbl1q_p8): Likewise.
1290 (vqtbl1q_s8): Likewise.
1291 (vqtbx1_s8): Likewise.
1292 (vqtbx1_p8): Likewise.
1293 (vqtbx1q_s8): Likewise.
1294 (vqtbx1q_p8): Likewise.
1295 (vtbl1_p8): Likewise.
1296 (vtbl2_p8): Likewise.
1297 (vtbx2_p8): Likewise.
1299 2021-09-29 Richard Biener <rguenther@suse.de>
1301 * tree-vect-data-refs.c (vect_dr_misalign_for_aligned_access):
1303 (vect_update_misalignment_for_peel): Use it to update
1304 misaligned to the value necessary for an aligned access.
1305 (vect_get_peeling_costs_all_drs): Likewise.
1306 (vect_enhance_data_refs_alignment): Likewise.
1308 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1310 * config/aarch64/aarch64.c (aarch64_expand_cpymem): Count number of
1311 emitted operations and adjust heuristic for code size.
1313 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1315 * config/aarch64/aarch64.c (aarch64_expand_setmem): Count number of
1316 emitted operations and adjust heuristic for code size.
1318 2021-09-29 Jakub Jelinek <jakub@redhat.com>
1320 PR middle-end/102504
1321 * gimplify.c (gimplify_scan_omp_clauses): Use omp_check_private even
1322 in OMP_SCOPE clauses, not just on worksharing construct clauses.
1324 2021-09-28 Geng Qi <gengqi@linux.alibaba.com>
1326 * config/riscv/riscv.md (mulv<mode>4): Call gen_smul<mode>3_highpart.
1327 (<u>mulditi3): Call <su>muldi3_highpart.
1328 (<u>muldi3_highpart): Rename to <su>muldi3_highpart.
1329 (<u>mulsidi3): Call <su>mulsi3_highpart.
1330 (<u>mulsi3_highpart): Rename to <su>mulsi3_highpart.
1332 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
1334 * config/darwin.h (DSYMUTIL_SPEC): Recognize D sources.
1336 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
1338 * config/rs6000/darwin.h (FIXED_R13): Add for PPC64.
1339 (FIRST_SAVED_GP_REGNO): Save from R13 even when it is one
1342 2021-09-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1344 * config/aarch64/aarch64.h (AARCH64_FL_LS64): Define
1345 (AARCH64_FL_V8_7): Likewise.
1346 (AARCH64_FL_FOR_ARCH8_7): Likewise.
1347 * config/aarch64/aarch64-arches.def (armv8.7-a): Define.
1348 * config/aarch64/aarch64-option-extensions.def (ls64): Define.
1349 * doc/invoke.texi: Document the above.
1351 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
1353 * dbgcnt.c (dbg_cnt_counter): New.
1354 * dbgcnt.h (dbg_cnt_counter): New.
1355 * dumpfile.c (dump_options): Add entry for TDF_THREADING.
1356 * dumpfile.h (enum dump_flag): Add TDF_THREADING.
1357 * gimple-range-path.cc (DEBUG_SOLVER): Use TDF_THREADING.
1358 * tree-ssa-threadupdate.c (dump_jump_thread_path): Dump out
1361 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
1363 * cfgcleanup.c (pass_jump::execute): Check
1364 flag_expensive_optimizations.
1365 (pass_jump_after_combine::gate): Same.
1366 * doc/invoke.texi (-fthread-jumps): Enable for -O1.
1367 * opts.c (default_options_table): Enable -fthread-jumps at -O1.
1368 * tree-ssa-threadupdate.c
1369 (fwd_jt_path_registry::remove_jump_threads_including): Bail unless
1372 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
1374 * tree-ssa-reassoc.c (biased_names): New global.
1375 (propagate_bias_p): New function.
1376 (loop_carried_phi): Remove.
1377 (propagate_rank): Propagate bias along single uses.
1378 (get_rank): Update biased_names when needed.
1380 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
1382 * passes.def (pass_reassoc): Rename parameter to early_p.
1383 * tree-ssa-reassoc.c (reassoc_bias_loop_carried_phi_ranks_p):
1385 (phi_rank): Don't bias loop-carried phi ranks
1386 before vectorization pass.
1387 (execute_reassoc): Add bias_loop_carried_phi_ranks_p parameter.
1388 (pass_reassoc::pass_reassoc): Add bias_loop_carried_phi_ranks_p
1390 (pass_reassoc::set_param): Set bias_loop_carried_phi_ranks_p
1392 (pass_reassoc::execute): Pass bias_loop_carried_phi_ranks_p to
1394 (pass_reassoc::bias_loop_carried_phi_ranks_p): New member.
1396 2021-09-28 Jakub Jelinek <jakub@redhat.com>
1399 * config/i386/i386.c (standard_80387_constant_p): Don't recognize
1400 special 80387 instruction XFmode constants if flag_rounding_math.
1402 2021-09-28 Richard Biener <rguenther@suse.de>
1404 PR tree-optimization/100112
1405 * tree-ssa-sccvn.c (visit_reference_op_load): Record the
1406 referece into the hashtable twice in case last_vuse is
1407 different from the original vuse on the stmt.
1409 2021-09-28 Jakub Jelinek <jakub@redhat.com>
1411 PR middle-end/102492
1412 * gimplify.c (gimplify_adjust_omp_clauses_1): Don't call the
1413 omp_finish_clause langhook on implicitly added OMP_CLAUSE_PRIVATE
1414 clauses on SIMD constructs.
1416 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
1418 PR tree-optimization/102511
1419 * gimple-range-path.cc (path_range_query::range_on_path_entry):
1420 Return VARYING when nothing found.
1422 2021-09-28 Hongyu Wang <hongyu.wang@intel.com>
1425 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): Add
1427 (VALID_SSE2_REG_VHF_MODE): Add V4HFmode and V2HFmode.
1428 (VALID_MMX_REG_MODE): Add V4HFmode.
1429 (SSE_REG_MODE_P): Replace VALID_AVX512FP16_REG_MODE with
1430 vector mode condition.
1431 * config/i386/i386.c (classify_argument): Parse V4HF/V2HF
1433 (function_arg_32): Add V4HFmode.
1434 (function_arg_advance_32): Likewise.
1435 * config/i386/i386.md (mode): Add V4HF/V2HF.
1436 (MODE_SIZE): Likewise.
1437 * config/i386/mmx.md (MMXMODE): Add V4HF mode.
1438 (V_32): Add V2HF mode.
1439 (VHF_32_64): New mode iterator.
1440 (*mov<mode>_internal): Adjust sse alternatives to support
1442 (*mov<mode>_internal): Adjust sse alternatives to support
1444 (<insn><mode>3): New define_insn for add/sub/mul/div.
1446 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
1448 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Check
1450 (pass_early_thread_jumps::gate): Same.
1451 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
1452 Return if !flag_thread_jumps.
1453 * tree-ssa-threadupdate.c
1454 (jt_path_registry::register_jump_thread): Assert that
1455 flag_thread_jumps is true.
1457 2021-09-28 liuhongt <hongtao.liu@intel.com>
1460 (simplify_context::simplify_binary_operation_1): Relax
1461 condition of simplifying (vec_concat:M (vec_select op0
1462 index0)(vec_select op1 index1)) to allow different modes
1463 between op0 and M, but have same inner mode.
1465 2021-09-28 liuhongt <hongtao.liu@intel.com>
1467 * config/i386/i386-expand.c (emit_reduc_half): Handle
1468 V8HF/V16HF/V32HFmode.
1469 * config/i386/sse.md (REDUC_SSE_PLUS_MODE): Add V8HF.
1470 (REDUC_SSE_SMINMAX_MODE): Ditto.
1471 (REDUC_PLUS_MODE): Add V16HF and V32HF.
1472 (REDUC_SMINMAX_MODE): Ditto.
1474 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
1476 * gimple-range-path.cc
1477 (path_range_query::precompute_ranges_in_block): Rename to...
1478 (path_range_query::compute_ranges_in_block): ...this.
1479 (path_range_query::precompute_ranges): Rename to...
1480 (path_range_query::compute_ranges): ...this.
1481 (path_range_query::precompute_relations): Rename to...
1482 (path_range_query::compute_relations): ...this.
1483 (path_range_query::precompute_phi_relations): Rename to...
1484 (path_range_query::compute_phi_relations): ...this.
1485 * gimple-range-path.h: Rename precompute* to compute*.
1486 * tree-ssa-threadbackward.c
1487 (back_threader::find_taken_edge_switch): Same.
1488 (back_threader::find_taken_edge_cond): Same.
1489 * tree-ssa-threadedge.c
1490 (hybrid_jt_simplifier::compute_ranges_from_state): Same.
1491 (hybrid_jt_state::register_equivs_stmt): Inline...
1492 * tree-ssa-threadedge.h: ...here.
1494 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
1496 * tree-vrp.c (lhs_of_dominating_assert): Remove.
1497 (class vrp_jt_state): Remove.
1498 (class vrp_jt_simplifier): Remove.
1499 (vrp_jt_simplifier::simplify): Remove.
1500 (class vrp_jump_threader): Remove.
1501 (vrp_jump_threader::vrp_jump_threader): Remove.
1502 (vrp_jump_threader::~vrp_jump_threader): Remove.
1503 (vrp_jump_threader::before_dom_children): Remove.
1504 (vrp_jump_threader::after_dom_children): Remove.
1506 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
1508 * passes.def (pass_vrp_threader): New.
1509 * tree-pass.h (make_pass_vrp_threader): Add make_pass_vrp_threader.
1510 * tree-ssa-threadedge.c (hybrid_jt_state::register_equivs_stmt): New.
1511 (hybrid_jt_simplifier::hybrid_jt_simplifier): New.
1512 (hybrid_jt_simplifier::simplify): New.
1513 (hybrid_jt_simplifier::compute_ranges_from_state): New.
1514 * tree-ssa-threadedge.h (class hybrid_jt_state): New.
1515 (class hybrid_jt_simplifier): New.
1516 * tree-vrp.c (execute_vrp): Remove ASSERT_EXPR based jump
1518 (class hybrid_threader): New.
1519 (hybrid_threader::hybrid_threader): New.
1520 (hybrid_threader::~hybrid_threader): New.
1521 (hybrid_threader::before_dom_children): New.
1522 (hybrid_threader::after_dom_children): New.
1523 (execute_vrp_threader): New.
1524 (class pass_vrp_threader): New.
1525 (make_pass_vrp_threader): New.
1527 2021-09-27 Martin Liska <mliska@suse.cz>
1529 * output.h (enum section_flag): New.
1530 (SECTION_FORGET): Remove.
1531 (SECTION_ENTSIZE): Make it (1UL << 8) - 1.
1532 (SECTION_STYLE_MASK): Define it based on other enum
1534 * varasm.c (switch_to_section): Remove unused handling of
1537 2021-09-27 Martin Liska <mliska@suse.cz>
1539 * common.opt: Add new variable flag_default_complex_method.
1540 * opts.c (finish_options): Handle flags related to
1541 x_flag_complex_method.
1542 * toplev.c (process_options): Remove option handling related
1543 to flag_complex_method.
1545 2021-09-27 Richard Biener <rguenther@suse.de>
1547 PR middle-end/102450
1548 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid using
1549 type_for_size, instead use int_mode_for_size.
1551 2021-09-27 Andrew Pinski <apinski@marvell.com>
1554 * gimplify.c (gimplify_save_expr): Return early
1555 if the type of val is error_mark_node.
1557 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
1559 * tree-ssanames.c (ssa_name_has_boolean_range): Use
1562 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
1564 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
1565 vrp_visit_cond_stmt.
1566 * tree-ssa-dom.c (cprop_operand): Convert to range_query API.
1567 (cprop_into_stmt): Same.
1568 (dom_opt_dom_walker::optimize_stmt): Same.
1570 2021-09-27 Richard Biener <rguenther@suse.de>
1572 PR tree-optimization/97351
1573 PR tree-optimization/97352
1574 PR tree-optimization/82426
1575 * tree-vectorizer.h (dr_misalignment): Add vector type
1577 (aligned_access_p): Likewise.
1578 (known_alignment_for_access_p): Likewise.
1579 (vect_supportable_dr_alignment): Likewise.
1580 (vect_known_alignment_in_bytes): Likewise. Refactor.
1581 (DR_MISALIGNMENT): Remove.
1582 (vect_update_shared_vectype): Likewise.
1583 * tree-vect-data-refs.c (dr_misalignment): Refactor, handle
1584 a vector type with larger alignment requirement and apply
1585 the negative step adjustment here.
1586 (vect_calculate_target_alignment): Remove.
1587 (vect_compute_data_ref_alignment): Get explicit vector type
1588 argument, do not apply a negative step alignment adjustment
1590 (vect_slp_analyze_node_alignment): Re-analyze alignment
1591 when we re-visit the DR with a bigger desired alignment but
1592 keep more precise results from smaller alignments.
1593 * tree-vect-slp.c (vect_update_shared_vectype): Remove.
1594 (vect_slp_analyze_node_operations_1): Do not update the
1595 shared vector type on stmts.
1596 * tree-vect-stmts.c (vect_analyze_stmt): Push/pop the
1597 vector type of an SLP node to the representative stmt-info.
1598 (vect_transform_stmt): Likewise.
1600 2021-09-27 liuhongt <hongtao.liu@intel.com>
1603 2021-09-09 liuhongt <hongtao.liu@intel.com>
1606 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
1607 (reduc_plus_scal_v4sf): .. this, New define_expand.
1608 (reduc_plus_scal_v2df): .. and this, New define_expand.
1610 2021-09-26 liuhongt <hongtao.liu@intel.com>
1612 * doc/extend.texi (Half-Precision): Remove storage only
1613 description for _Float16 w/o avx512fp16.
1615 2021-09-25 Dimitar Dimitrov <dimitar@dinux.eu>
1617 * config/pru/constraints.md (Rrio): New constraint.
1618 * config/pru/predicates.md (regio_operand): New predicate.
1619 * config/pru/pru-pragma.c (pru_register_pragmas): Register
1620 the __regio_symbol address space.
1621 * config/pru/pru-protos.h (pru_symref2ioregno): Declaration.
1622 * config/pru/pru.c (pru_symref2ioregno): New helper function.
1623 (pru_legitimate_address_p): Remove.
1624 (pru_addr_space_legitimate_address_p): Use the address space
1626 (pru_nongeneric_pointer_addrspace): New helper function.
1627 (pru_insert_attributes): New function to validate __regio_symbol
1629 (TARGET_INSERT_ATTRIBUTES): New macro.
1630 (TARGET_LEGITIMATE_ADDRESS_P): Remove.
1631 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): New macro.
1632 * config/pru/pru.h (enum reg_class): Add REGIO_REGS class.
1633 * config/pru/pru.md (*regio_readsi): New pattern to read I/O
1635 (*regio_nozext_writesi): New pattern to write to I/O registers.
1636 (*regio_zext_write_r30<EQS0:mode>): Ditto.
1637 * doc/extend.texi: Document the new PRU Named Address Space.
1639 2021-09-24 Patrick Palka <ppalka@redhat.com>
1643 * real.c (encode_ieee_double): Avoid unwanted sign extension.
1644 (encode_ieee_quad): Likewise.
1646 2021-09-24 Vladimir Makarov <vmakarov@redhat.com>
1648 PR rtl-optimization/102147
1649 * ira-build.c (ira_conflict_vector_profitable_p): Make
1650 profitability calculation independent of host compiler pointer and
1653 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
1655 * gimple-range-path.cc (path_range_query::path_range_query):
1656 Move debugging header...
1657 (path_range_query::precompute_ranges): ...here.
1658 (path_range_query::internal_range_of_expr): Do not call
1659 range_on_path_entry if NAME is defined in the current block.
1661 2021-09-24 Richard Biener <rguenther@suse.de>
1663 * cfghooks.c (verify_flow_info): Verify unallocated BB and
1664 edge flags are not set.
1666 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
1668 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
1670 (jt_path_registry::register_jump_thread): Call
1671 cancel_invalid_paths.
1672 * tree-ssa-threadupdate.h (class jt_path_registry): Add
1673 cancel_invalid_paths.
1675 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
1677 PR tree-optimization/102400
1678 * tree-ssa-sccvn.c (vn_reference_insert_pieces): Initialize
1679 result_vdef to zero value.
1681 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
1683 PR tree-optimization/102451
1684 * tree-ssa-dse.c (delete_dead_or_redundant_call): Record bb of stmt
1687 2021-09-24 Hongyu Wang <hongyu.wang@intel.com>
1689 * config/i386/sse.md (cond_<insn><mode>): Extend to support
1691 (cond_mul<mode>): Likewise.
1692 (cond_div<mode>): Likewise.
1693 (cond_<code><mode>): Likewise.
1694 (cond_fma<mode>): Likewise.
1695 (cond_fms<mode>): Likewise.
1696 (cond_fnma<mode>): Likewise.
1697 (cond_fnms<mode>): Likewise.
1699 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
1701 PR tree-optimization/102463
1702 * gimple-range-fold.cc (fold_using_range::relation_fold_and_or): If
1703 there is no range-ops handler, don't look for a relation.
1705 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
1707 * gimple-range-cache.cc (ranger_cache::ranger_cache): Take
1708 non-executable_edge flag as parameter.
1709 * gimple-range-cache.h (ranger_cache): Adjust prototype.
1710 * gimple-range-gori.cc (gori_compute::gori_compute): Take
1711 non-executable_edge flag as parameter.
1712 (gori_compute::outgoing_edge_range_p): Check new flag.
1713 * gimple-range-gori.h (gori_compute): Adjust prototype.
1714 * gimple-range.cc (gimple_ranger::gimple_ranger): Create new flag.
1715 (gimple_ranger::range_on_edge): Check new flag.
1716 * gimple-range.h (gimple_ranger::non_executable_edge_flag): New.
1717 * gimple-ssa-evrp.c (rvrp_folder): Pass ranger flag to simplifer.
1718 (hybrid_folder::hybrid_folder): Set ranger non-executable flag value.
1719 (hybrid_folder::fold_stmt): Set flag value in the simplifer.
1720 * vr-values.c (simplify_using_ranges::set_and_propagate_unexecutable):
1721 Use not_executable flag if provided inmstead of EDGE_EXECUTABLE.
1722 (simplify_using_ranges::simplify_switch_using_ranges): Clear
1723 EDGE_EXECUTABLE like it originally did.
1724 (simplify_using_ranges::cleanup_edges_and_switches): Clear any
1725 NON_EXECUTABLE flags.
1726 (simplify_using_ranges::simplify_using_ranges): Adjust.
1727 * vr-values.h (class simplify_using_ranges): Adjust.
1728 (simplify_using_ranges::set_range_query): Add non-executable flag param.
1730 2021-09-23 Bill Schmidt <wschmidt@linux.ibm.com>
1733 * config/rs6000/rs6000-call.c (rs6000_aggregate_candidate): Detect
1734 zero-width bit fields and return indicator.
1735 (rs6000_discover_homogeneous_aggregate): Diagnose when the
1736 presence of a zero-width bit field changes parameter passing in
1739 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
1741 * gimple-range-fold.cc (fold_using_range::range_of_phi):
1742 Remove dominator check.
1744 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
1746 * gimple-range-path.cc (path_range_query::precompute_relations):
1747 Hoist edge calculations before using EDGE_SUCC.
1749 2021-09-23 Jonathan Wakely <jwakely@redhat.com>
1751 * configure.ac: Fix --with-multilib-list description.
1752 * configure: Regenerate.
1754 2021-09-23 Richard Biener <rguenther@suse.de>
1756 PR tree-optimization/102448
1757 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
1758 Clear alignment info copied from DR_PTR_INFO.
1760 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
1762 * config/i386/i386-expand.c (ix86_use_mask_cmp_p): Enable
1764 * config/i386/sse.md (sseintvecmodelower): Add HF vector modes.
1765 (<avx512>_store<mode>_mask): Extend to support HF vector modes.
1766 (vec_cmp<mode><avx512fmaskmodelower>): Likewise.
1767 (vcond_mask_<mode><avx512fmaskmodelower>): Likewise.
1768 (vcond<mode><mode>): New expander.
1769 (vcond<mode><sseintvecmodelower>): Likewise.
1770 (vcond<sseintvecmodelower><mode>): Likewise.
1771 (vcondu<mode><sseintvecmodelower>): Likewise.
1773 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
1775 * config/i386/sse.md (extend<ssePHmodelower><mode>2):
1777 (extendv4hf<mode>2): Likewise.
1778 (extendv2hfv2df2): Likewise.
1779 (trunc<mode><ssePHmodelower>2): Likewise.
1780 (avx512fp16_vcvt<castmode>2ph_<mode>): Rename to ...
1781 (trunc<mode>v4hf2): ... this, and drop constraints.
1782 (avx512fp16_vcvtpd2ph_v2df): Rename to ...
1783 (truncv2dfv2hf2): ... this, and likewise.
1785 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
1787 * config/i386/sse.md (float<floatunssuffix><mode><ssePHmodelower>2):
1789 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>):
1791 (float<floatunssuffix><mode>v4hf2): ... this, and drop constraints.
1792 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Rename to ...
1793 (float<floatunssuffix>v2div2hf2): ... this, and likewise.
1795 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
1797 * config/i386/i386.md (fix<fixunssuffix>_trunchf<mode>2): New expander.
1798 (fixuns_trunchfhi2): Likewise.
1799 (*fixuns_trunchfsi2zext): New define_insn.
1800 * config/i386/sse.md (ssePHmodelower): New mode_attr.
1801 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
1802 New expander for same element vector fix_truncate.
1803 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
1804 Likewise for V4HF to V4SI/V4DI fix_truncate.
1805 (fix<fixunssuffix>_truncv2hfv2di2):
1806 Likeise for V2HF to V2DI fix_truncate.
1808 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
1810 * config/i386/i386.md (<code>hf3): New expander.
1812 2021-09-23 liuhongt <hongtao.liu@intel.com>
1814 * config/i386/sse.md (FMAMODEM): extend to handle FP16.
1815 (VFH_SF_AVX512VL): Extend to handle HFmode.
1816 (VF_SF_AVX512VL): Deleted.
1818 2021-09-23 liuhongt <hongtao.liu@intel.com>
1820 * config/i386/i386.md (rinthf2): New expander.
1821 (nearbyinthf2): New expander.
1823 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
1825 * tree-ssa-dom.c (class dom_jump_threader_simplifier): Rename...
1826 (class dom_jt_state): ...this and provide virtual overrides.
1827 (dom_jt_state::register_equiv): New.
1828 (class dom_jt_simplifier): Rename from
1829 dom_jump_threader_simplifier.
1830 (dom_jump_threader_simplifier::simplify): Rename...
1831 (dom_jt_simplifier::simplify): ...to this.
1832 (pass_dominator::execute): Use dom_jt_simplifier and
1834 * tree-ssa-threadedge.c (jump_threader::jump_threader):
1836 (jt_state::register_equivs_stmt): Abstract out...
1837 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
1839 (jump_threader::thread_around_empty_blocks): Update state.
1840 (jump_threader::thread_through_normal_block): Same.
1841 (jt_state::jt_state): Remove.
1842 (jt_state::push): Remove pass specific bits. Keep block vector
1844 (jt_state::append_path): New.
1845 (jt_state::pop): Remove pass specific bits.
1846 (jt_state::register_equiv): Same.
1847 (jt_state::record_ranges_from_stmt): Same.
1848 (jt_state::register_equivs_on_edge): Same. Rename...
1849 (jt_state::register_equivs_edge): ...to this.
1850 (jt_state::dump): New.
1851 (jt_state::debug): New.
1852 (jump_threader_simplifier::simplify): Remove.
1853 (jt_state::get_path): New.
1854 * tree-ssa-threadedge.h (class jt_simplifier): Make into a base
1855 class. Expose common functionality as virtual methods.
1856 (class jump_threader_simplifier): Same. Rename...
1857 (class jt_simplifier): ...to this.
1858 * tree-vrp.c (class vrp_jump_threader_simplifier): Rename...
1859 (class vrp_jt_simplifier): ...to this. Provide pass specific
1861 (class vrp_jt_state): New.
1862 (vrp_jump_threader_simplifier::simplify): Rename...
1863 (vrp_jt_simplifier::simplify): ...to this. Inline code from
1864 what used to be the base class.
1865 (vrp_jump_threader::vrp_jump_threader): Use vrp_jt_state and
1868 2021-09-22 Tobias Burnus <tobias@codesourcery.com>
1871 * doc/invoke.texi (-Wno-missing-include-dirs.): Document Fortran
1874 2021-09-22 Roger Sayle <roger@nextmovesoftware.com>
1875 Richard Biener <rguenther@suse.de>
1877 * match.pd (negation simplifications): Implement some negation
1878 folding transformations from fold-const.c's fold_negate_expr.
1879 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Add a SIMPLIFY
1880 argument, to control whether the op should be simplified prior
1881 to looking up/assigning a value number.
1882 (vn_nary_build_or_lookup): Update call to vn_nary_build_or_lookup_1.
1883 (vn_nary_simplify): Likewise.
1884 (visit_nary_op): Likewise, but when constructing a NEGATE_EXPR
1885 now call vn_nary_build_or_lookup_1 disabling simplification.
1887 2021-09-22 Jiufu Guo <guojiufu@linux.ibm.com>
1889 PR tree-optimization/102087
1890 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
1891 Update bound/cmp/control for niter.
1893 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
1895 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
1896 Move check for non-empty BB here.
1897 (fur_source::register_outgoing_edges): ...from here.
1899 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
1901 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
1902 Remove call to improve_range_with_equivs.
1903 (path_range_query::improve_range_with_equivs): Remove
1904 * gimple-range-path.h: Remove improve_range_with_equivs.
1906 2021-09-22 dianhong xu <dianhong.xu@intel.com>
1908 * config/i386/avx512fp16intrin.h:
1909 (_mm512_mask_blend_ph): New intrinsic.
1910 (_mm512_permutex2var_ph): Ditto.
1911 (_mm512_permutexvar_ph): Ditto.
1912 * config/i386/avx512fp16vlintrin.h:
1913 (_mm256_mask_blend_ph): New intrinsic.
1914 (_mm256_permutex2var_ph): Ditto.
1915 (_mm256_permutexvar_ph): Ditto.
1916 (_mm_mask_blend_ph): Ditto.
1917 (_mm_permutex2var_ph): Ditto.
1918 (_mm_permutexvar_ph): Ditto.
1920 2021-09-22 dianhong xu <dianhong.xu@intel.com>
1922 * config/i386/avx512fp16intrin.h: Add new intrinsics.
1923 (_mm512_conj_pch): New intrinsic.
1924 (_mm512_mask_conj_pch): Ditto.
1925 (_mm512_maskz_conj_pch): Ditto.
1926 * config/i386/avx512fp16vlintrin.h: Add new intrinsics.
1927 (_mm256_conj_pch): New intrinsic.
1928 (_mm256_mask_conj_pch): Ditto.
1929 (_mm256_maskz_conj_pch): Ditto.
1930 (_mm_conj_pch): Ditto.
1931 (_mm_mask_conj_pch): Ditto.
1932 (_mm_maskz_conj_pch): Ditto.
1934 2021-09-22 dianhong xu <dianhong.xu@intel.com>
1936 * config/i386/avx512fp16intrin.h (_MM512_REDUCE_OP): New macro
1937 (_mm512_reduce_add_ph): New intrinsic.
1938 (_mm512_reduce_mul_ph): Ditto.
1939 (_mm512_reduce_min_ph): Ditto.
1940 (_mm512_reduce_max_ph): Ditto.
1941 * config/i386/avx512fp16vlintrin.h
1942 (_MM256_REDUCE_OP/_MM_REDUCE_OP): New macro.
1943 (_mm256_reduce_add_ph): New intrinsic.
1944 (_mm256_reduce_mul_ph): Ditto.
1945 (_mm256_reduce_min_ph): Ditto.
1946 (_mm256_reduce_max_ph): Ditto.
1947 (_mm_reduce_add_ph): Ditto.
1948 (_mm_reduce_mul_ph): Ditto.
1949 (_mm_reduce_min_ph): Ditto.
1950 (_mm_reduce_max_ph): Ditto.
1952 2021-09-22 dianhong xu <dianhong.xu@intel.com>
1954 * config/i386/avx512fp16intrin.h (__m512h_u, __m256h_u,
1955 __m128h_u): New typedef.
1956 (_mm512_load_ph): New intrinsic.
1957 (_mm256_load_ph): Ditto.
1958 (_mm_load_ph): Ditto.
1959 (_mm512_loadu_ph): Ditto.
1960 (_mm256_loadu_ph): Ditto.
1961 (_mm_loadu_ph): Ditto.
1962 (_mm512_store_ph): Ditto.
1963 (_mm256_store_ph): Ditto.
1964 (_mm_store_ph): Ditto.
1965 (_mm512_storeu_ph): Ditto.
1966 (_mm256_storeu_ph): Ditto.
1967 (_mm_storeu_ph): Ditto.
1968 (_mm512_abs_ph): Ditto.
1969 * config/i386/avx512fp16vlintrin.h
1970 (_mm_abs_ph): Ditto.
1971 (_mm256_abs_ph): Ditto.
1973 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
1975 * config/s390/tpf.md (prologue_tpf, epilogue_tpf): Add cc clobber.
1977 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
1980 * config/s390/s390.c (s390_expand_insv): Emit a normal move if it
1981 is actually a full copy of the source operand into the target.
1982 Don't emit a strict low part move if source and target mode match.
1984 2021-09-22 Jakub Jelinek <jakub@redhat.com>
1986 PR middle-end/102415
1987 * omp-expand.c (expand_omp_single): If region->exit is NULL,
1988 assert region->entry is GIMPLE_OMP_SCOPE region and return.
1990 2021-09-22 Jakub Jelinek <jakub@redhat.com>
1992 * tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
1993 * tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
1994 arguments from 2 to 3.
1995 * tree-pretty-print.c (dump_omp_clause): Print allocator() around
1996 allocate clause allocator and print align if present.
1997 * omp-low.c (scan_sharing_clauses): Force allocate_map entry even
1998 for omp_default_mem_alloc if align modifier is present. If align
1999 modifier is present, use TREE_LIST to encode both allocator and
2001 (lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
2002 Handle align modifier on allocator clause if present.
2004 2021-09-22 liuhongt <hongtao.liu@intel.com>
2006 * config/i386/i386.md (define_attr "isa"): Add
2008 (define_attr "enabled"): Correspond fma_or_avx512vl to
2009 TARGET_FMA || TARGET_AVX512VL.
2010 * config/i386/mmx.md (fmav2sf4): Extend to AVX512 fma.
2015 2021-09-22 liuhongt <hongtao.liu@intel.com>
2017 * config/i386/i386.md (cstorehf3): New define_expand.
2019 2021-09-22 liuhongt <hongtao.liu@intel.com>
2021 * config/i386/i386.md (<rounding_insn>hf2): New expander.
2022 (sse4_1_round<mode>2): Extend from MODEF to MODEFH.
2023 * config/i386/sse.md (*sse4_1_round<ssescalarmodesuffix>):
2024 Extend from VF_128 to VFH_128.
2026 2021-09-22 liuhongt <hongtao.liu@intel.com>
2028 * config/i386/i386-features.c (i386-features.c): Handle
2030 * config/i386/i386.md (sqrthf2): New expander.
2031 (*sqrthf2): New define_insn.
2032 * config/i386/sse.md
2033 (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
2036 2021-09-22 liuhongt <hongtao.liu@intel.com>
2038 * config/i386/avx512fp16intrin.h (_mm_mask_fcmadd_sch):
2040 (_mm_mask3_fcmadd_sch): Likewise.
2041 (_mm_maskz_fcmadd_sch): Likewise.
2042 (_mm_fcmadd_sch): Likewise.
2043 (_mm_mask_fmadd_sch): Likewise.
2044 (_mm_mask3_fmadd_sch): Likewise.
2045 (_mm_maskz_fmadd_sch): Likewise.
2046 (_mm_fmadd_sch): Likewise.
2047 (_mm_mask_fcmadd_round_sch): Likewise.
2048 (_mm_mask3_fcmadd_round_sch): Likewise.
2049 (_mm_maskz_fcmadd_round_sch): Likewise.
2050 (_mm_fcmadd_round_sch): Likewise.
2051 (_mm_mask_fmadd_round_sch): Likewise.
2052 (_mm_mask3_fmadd_round_sch): Likewise.
2053 (_mm_maskz_fmadd_round_sch): Likewise.
2054 (_mm_fmadd_round_sch): Likewise.
2055 (_mm_fcmul_sch): Likewise.
2056 (_mm_mask_fcmul_sch): Likewise.
2057 (_mm_maskz_fcmul_sch): Likewise.
2058 (_mm_fmul_sch): Likewise.
2059 (_mm_mask_fmul_sch): Likewise.
2060 (_mm_maskz_fmul_sch): Likewise.
2061 (_mm_fcmul_round_sch): Likewise.
2062 (_mm_mask_fcmul_round_sch): Likewise.
2063 (_mm_maskz_fcmul_round_sch): Likewise.
2064 (_mm_fmul_round_sch): Likewise.
2065 (_mm_mask_fmul_round_sch): Likewise.
2066 (_mm_maskz_fmul_round_sch): Likewise.
2067 * config/i386/i386-builtin.def: Add corresponding new builtins.
2068 * config/i386/sse.md
2069 (avx512fp16_fmaddcsh_v8hf_maskz<round_expand_name>): New expander.
2070 (avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>): Ditto.
2071 (avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>):
2073 (avx512fp16_<complexopname>sh_v8hf_mask<round_name>): Ditto.
2074 (avx512fp16_<complexopname>sh_v8hf<mask_scalarc_name><round_scalarcz_name>):
2076 * config/i386/subst.md (mask_scalarcz_name): New.
2077 (mask_scalarc_name): Ditto.
2078 (mask_scalarc_operand3): Ditto.
2079 (mask_scalarcz_operand4): Ditto.
2080 (round_scalarcz_name): Ditto.
2081 (round_scalarc_mask_operand3): Ditto.
2082 (round_scalarcz_mask_operand4): Ditto.
2083 (round_scalarc_mask_op3): Ditto.
2084 (round_scalarcz_mask_op4): Ditto.
2085 (round_scalarcz_constraint): Ditto.
2086 (round_scalarcz_nimm_predicate): Ditto.
2087 (mask_scalarcz): Ditto.
2088 (mask_scalarc): Ditto.
2089 (round_scalarcz): Ditto.
2091 2021-09-22 liuhongt <hongtao.liu@intel.com>
2093 * config/i386/avx512fp16intrin.h (_mm512_fcmadd_pch):
2095 (_mm512_mask_fcmadd_pch): Likewise.
2096 (_mm512_mask3_fcmadd_pch): Likewise.
2097 (_mm512_maskz_fcmadd_pch): Likewise.
2098 (_mm512_fmadd_pch): Likewise.
2099 (_mm512_mask_fmadd_pch): Likewise.
2100 (_mm512_mask3_fmadd_pch): Likewise.
2101 (_mm512_maskz_fmadd_pch): Likewise.
2102 (_mm512_fcmadd_round_pch): Likewise.
2103 (_mm512_mask_fcmadd_round_pch): Likewise.
2104 (_mm512_mask3_fcmadd_round_pch): Likewise.
2105 (_mm512_maskz_fcmadd_round_pch): Likewise.
2106 (_mm512_fmadd_round_pch): Likewise.
2107 (_mm512_mask_fmadd_round_pch): Likewise.
2108 (_mm512_mask3_fmadd_round_pch): Likewise.
2109 (_mm512_maskz_fmadd_round_pch): Likewise.
2110 (_mm512_fcmul_pch): Likewise.
2111 (_mm512_mask_fcmul_pch): Likewise.
2112 (_mm512_maskz_fcmul_pch): Likewise.
2113 (_mm512_fmul_pch): Likewise.
2114 (_mm512_mask_fmul_pch): Likewise.
2115 (_mm512_maskz_fmul_pch): Likewise.
2116 (_mm512_fcmul_round_pch): Likewise.
2117 (_mm512_mask_fcmul_round_pch): Likewise.
2118 (_mm512_maskz_fcmul_round_pch): Likewise.
2119 (_mm512_fmul_round_pch): Likewise.
2120 (_mm512_mask_fmul_round_pch): Likewise.
2121 (_mm512_maskz_fmul_round_pch): Likewise.
2122 * config/i386/avx512fp16vlintrin.h (_mm_fmadd_pch):
2124 (_mm_mask_fmadd_pch): Likewise.
2125 (_mm_mask3_fmadd_pch): Likewise.
2126 (_mm_maskz_fmadd_pch): Likewise.
2127 (_mm256_fmadd_pch): Likewise.
2128 (_mm256_mask_fmadd_pch): Likewise.
2129 (_mm256_mask3_fmadd_pch): Likewise.
2130 (_mm256_maskz_fmadd_pch): Likewise.
2131 (_mm_fcmadd_pch): Likewise.
2132 (_mm_mask_fcmadd_pch): Likewise.
2133 (_mm_mask3_fcmadd_pch): Likewise.
2134 (_mm_maskz_fcmadd_pch): Likewise.
2135 (_mm256_fcmadd_pch): Likewise.
2136 (_mm256_mask_fcmadd_pch): Likewise.
2137 (_mm256_mask3_fcmadd_pch): Likewise.
2138 (_mm256_maskz_fcmadd_pch): Likewise.
2139 (_mm_fmul_pch): Likewise.
2140 (_mm_mask_fmul_pch): Likewise.
2141 (_mm_maskz_fmul_pch): Likewise.
2142 (_mm256_fmul_pch): Likewise.
2143 (_mm256_mask_fmul_pch): Likewise.
2144 (_mm256_maskz_fmul_pch): Likewise.
2145 (_mm_fcmul_pch): Likewise.
2146 (_mm_mask_fcmul_pch): Likewise.
2147 (_mm_maskz_fcmul_pch): Likewise.
2148 (_mm256_fcmul_pch): Likewise.
2149 (_mm256_mask_fcmul_pch): Likewise.
2150 (_mm256_maskz_fcmul_pch): Likewise.
2151 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF_V8HF_V8HF,
2152 V8HF_FTYPE_V16HF_V16HF_V16HF, V16HF_FTYPE_V16HF_V16HF_V16HF_UQI,
2153 V32HF_FTYPE_V32HF_V32HF_V32HF_INT,
2154 V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT): Add new builtin types.
2155 * config/i386/i386-builtin.def: Add new builtins.
2156 * config/i386/i386-expand.c: Handle new builtin types.
2157 * config/i386/subst.md (SUBST_CV): New.
2158 (maskc_name): Ditto.
2159 (maskc_operand3): Ditto.
2161 (sdc_maskz_name): Ditto.
2162 (sdc_mask_op4): Ditto.
2163 (sdc_mask_op5): Ditto.
2164 (sdc_mask_mode512bit_condition): Ditto.
2166 (round_maskc_operand3): Ditto.
2167 (round_sdc_mask_operand4): Ditto.
2168 (round_maskc_op3): Ditto.
2169 (round_sdc_mask_op4): Ditto.
2170 (round_saeonly_sdc_mask_operand5): Ditto.
2171 * config/i386/sse.md (unspec): Add complex fma unspecs.
2172 (avx512fmaskcmode): New.
2173 (UNSPEC_COMPLEX_F_C_MA): Ditto.
2174 (UNSPEC_COMPLEX_F_C_MUL): Ditto.
2175 (complexopname): Ditto.
2176 (<avx512>_fmaddc_<mode>_maskz<round_expand_name>): New expander.
2177 (<avx512>_fcmaddc_<mode>_maskz<round_expand_name>): Ditto.
2178 (fma_<complexopname>_<mode><sdc_maskz_name><round_name>): New
2180 (<avx512>_<complexopname>_<mode>_mask<round_name>): Ditto.
2181 (<avx512>_<complexopname>_<mode><maskc_name><round_name>): Ditto.
2183 2021-09-22 Kewen Lin <linkw@linux.ibm.com>
2185 * config/rs6000/rs6000.opt (rs6000-density-pct-threshold,
2186 rs6000-density-size-threshold, rs6000-density-penalty,
2187 rs6000-density-load-pct-threshold,
2188 rs6000-density-load-num-threshold): New parameter.
2189 * config/rs6000/rs6000.c (rs6000_density_test): Adjust with
2190 corresponding parameters.
2192 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2194 * gimple-range-path.cc (path_range_query::defined_outside_path):
2196 (path_range_query::range_on_path_entry): New.
2197 (path_range_query::internal_range_of_expr): Resolve unknowns
2199 (path_range_query::improve_range_with_equivs): New.
2200 (path_range_query::ssa_range_in_phi): Resolve unknowns with
2202 * gimple-range-path.h (class path_range_query): Add
2203 defined_outside_path, range_on_path_entry, and
2204 improve_range_with_equivs.
2206 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2208 * gimple-range-path.cc (path_range_query::add_to_imports): New.
2209 (path_range_query::add_copies_to_imports): New.
2210 (path_range_query::precompute_ranges): Call
2211 add_copies_to_imports.
2212 * gimple-range-path.h (class path_range_query): Add prototypes
2213 for add_copies_to_imports and add_to_imports.
2215 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2217 * gimple-range-path.cc (path_range_query::range_defined_in_block):
2218 Remove useless code.
2220 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2222 * gimple-range-fold.h (class fur_source): Make oracle protected.
2223 * gimple-range-path.cc (path_range_query::path_range_query): Add
2224 resolve argument. Initialize oracle.
2225 (path_range_query::~path_range_query): Delete oracle.
2226 (path_range_query::range_of_stmt): Adapt to use relations.
2227 (path_range_query::precompute_ranges): Pre-compute relations.
2228 (class jt_fur_source): New
2229 (jt_fur_source::jt_fur_source): New.
2230 (jt_fur_source::register_relation): New.
2231 (jt_fur_source::query_relation): New.
2232 (path_range_query::precompute_relations): New.
2233 (path_range_query::precompute_phi_relations): New.
2234 * gimple-range-path.h (path_range_query): Add resolve argument.
2235 Add oracle, precompute_relations, precompute_phi_relations.
2236 * tree-ssa-threadbackward.c (back_threader::back_threader): Pass
2237 resolve argument to solver.
2239 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2241 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
2242 Rename postfold_gcond_edges to register_outgoing_edges and
2244 (fold_using_range::postfold_gcond_edges): Rename...
2245 (fur_source::register_outgoing_edges): ...to this.
2246 * gimple-range-fold.h (postfold_gcond_edges): Rename to
2247 register_outgoing_edges and move to fur_source.
2249 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2251 * gimple-range-fold.cc (fold_using_range::range_of_phi): Check
2252 dom_info_available_p.
2254 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
2256 * gimple-range-cache.cc (non_null_ref::non_null_ref): Use create
2257 and quick_grow_cleared instead of safe_grow_cleared.
2259 2021-09-21 Thomas Schwinge <thomas@codesourcery.com>
2262 * omp-oacc-neuter-broadcast.cc (oacc_do_neutering): Evaluate
2265 2021-09-21 Richard Earnshaw <rearnsha@arm.com>
2267 * configure.ac: Detect when the assembler supports new-style
2268 architecture extensions.
2269 * common/config/arm/arm-common.c (arm_rewrite_mcpu): Return
2270 the full CPU string if the assembler can grok it.
2271 (arm_rewrite_march): Likewise but for the architecture.
2272 * config.in: Regenerate.
2273 * configure: Regenerate.
2275 2021-09-21 Richard Biener <rguenther@suse.de>
2277 PR tree-optimization/102421
2278 * tree-vect-loop.c (vect_dissolve_slp_only_groups): Copy and
2279 adjust alignment info.
2281 2021-09-21 Kewen Lin <linkw@linux.ibm.com>
2283 * ipa-fnsummary.c (ipa_fn_summary_write): Remove inconsistent
2284 bitfield stream out.
2286 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
2288 * gimple-range-fold.cc (fold_using_range::range_of_phi): Ignore
2289 undefined edges, apply an equivalence if appropriate.
2290 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Return
2291 UNDEFINED if EDGE_EXECUTABLE is not set.
2292 * gimple-range.cc (gimple_ranger::gimple_ranger): Set all edges
2293 as EXECUTABLE upon startup.
2294 (gimple_ranger::range_on_edge): Return UNDEFINED for edges without
2295 EDGE_EXECUTABLE set.
2296 * vr-values.c (set_and_propagate_unexecutable): New.
2297 (simplify_using_ranges::fold_cond): Call set_and_propagate.
2298 (simplify_using_ranges::simplify_switch_using_ranges): Ditto.
2299 * vr-values.h: Add prototype.
2301 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
2303 * value-relation.cc (equiv_oracle::register_initial_def): New.
2304 (equiv_oracle::register_relation): Call register_initial_def.
2305 (equiv_oracle::add_equiv_to_block): New. Split register_relation.
2306 (relation_oracle::register_stmt): Check def block of PHI arguments.
2307 * value-relation.h (equiv_oracle): Add new prototypes.
2309 2021-09-20 Matthias Kretz <m.kretz@gsi.de>
2311 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
2312 Define __RECIPROCAL_MATH__, __NO_SIGNED_ZEROS__,
2313 __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__, and
2314 __ROUNDING_MATH__ according to their corresponding flags.
2315 * doc/cpp.texi: Document __RECIPROCAL_MATH__,
2316 __NO_SIGNED_ZEROS__, __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__,
2317 and __ROUNDING_MATH__.
2319 2021-09-20 Richard Biener <rguenther@suse.de>
2321 * tree-vect-stmts.c (vectorizable_load): Use the vectype
2324 2021-09-20 Richard Biener <rguenther@suse.de>
2326 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
2327 Do not compute alignment of the vectorized access here.
2329 2021-09-20 Richard Biener <rguenther@suse.de>
2331 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
2332 Store -1 for runtime alias peeling iterations.
2334 2021-09-20 Richard Biener <rguenther@suse.de>
2336 * config.gcc: Obsolete hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11*.
2338 2021-09-20 Thomas Schwinge <thomas@codesourcery.com>
2340 * input.c (string_concat_db::record_string_concatenation)
2341 (string_concat_db::get_string_concatenation): Skip for
2342 'RESERVED_LOCATION_P'.
2344 2021-09-20 Richard Biener <rguenther@suse.de>
2346 PR tree-optimization/65206
2347 * tree-data-ref.h (struct data_reference): Add alt_indices,
2349 * tree-data-ref.c (free_data_ref): Release alt_indices.
2350 (dr_analyze_indices): Work on struct indices and get DR_REF as tree.
2351 (create_data_ref): Adjust.
2352 (initialize_data_dependence_relation): Split into head
2353 and tail. When the base objects fail to match up try
2354 again with pointer-based analysis of indices.
2355 * tree-vectorizer.c (vec_info_shared::check_datarefs): Do
2356 not compare the lazily computed alternate set of indices.
2358 2021-09-20 Iain Sandoe <iain@sandoe.co.uk>
2360 * gcc.c: Test for execute OK when we find the
2361 programs for assembler linker and dsymutil and those
2362 were specified at configure-time.
2364 2021-09-19 Martin Sebor <msebor@redhat.com>
2366 PR middle-end/102403
2367 * gimple-predicate-analysis.cc (predicate::init_from_control_deps):
2368 Correct a function pre/postcondition.
2370 2021-09-19 Martin Sebor <msebor@redhat.com>
2372 PR middle-end/102243
2373 * tree-ssa-strlen.c (get_range): Handle null cfun.
2375 2021-09-19 Iain Sandoe <iain@sandoe.co.uk>
2377 * config/darwin.h (LINK_COMMAND_SPEC_A): Use Darwin10
2378 unwinder shim as a convenience library.
2380 2021-09-19 Andrew Pinski <apinski@marvell.com>
2382 * doc/install.texi: Add note about
2383 binutils 2.35 is required for LTO usage.
2385 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
2387 * tree-ssa-threadbackward.c
2388 (back_threader_registry::register_path): Use push_edge.
2389 * tree-ssa-threadedge.c
2390 (jump_threader::thread_around_empty_blocks): Same.
2391 (jump_threader::thread_through_normal_block): Same.
2392 (jump_threader::thread_across_edge): Same. Also, use auto_bitmap.
2394 * tree-ssa-threadupdate.c
2395 (jt_path_registry::allocate_thread_edge): Remove.
2396 (jt_path_registry::push_edge): New.
2397 (dump_jump_thread_path): Make static.
2398 * tree-ssa-threadupdate.h (allocate_thread_edge): Remove.
2401 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
2403 * gimple-range-path.cc (path_range_query::path_range_query): Add
2405 (path_range_query::dump): Remove extern declaration of dump_ranger.
2406 * gimple-range-trace.cc (dump_ranger): Add DEBUG_FUNCTION marker.
2407 * gimple-range-trace.h (dump_ranger): Add prototype.
2409 2021-09-19 John Ericson <git@JohnEricson.me>
2411 * gcc.c (find_a_program): New function, factored out of...
2412 (find_a_file): Here.
2413 (execute): Use find_a_program when looking for programs rather
2416 2021-09-19 Matwey V. Kornilov <matwey.kornilov@gmail.com>
2418 * config/avr/avr-mcus.def: Add atmega324pb.
2419 * doc/avr-mmcu.texi: Corresponding changes.
2421 2021-09-19 Roger Sayle <roger@nextmovesoftware.com>
2424 * match.pd (cmp @0 REAL_CST@1): When @0 is also REAL_CST, apply
2425 the same transformations as to @1. For comparisons against NaN,
2426 don't check HONOR_SNANS but confirm that neither operand is a
2429 2021-09-19 Benjamin Peterson <benjamin@locrian.net>
2431 * attribs.c (make_unique_name): Delete.
2432 * attribs.h (make_unique_name): Delete.
2434 2021-09-19 Andrew Pinski <apinski@marvell.com>
2436 * lra-constraints.c (check_and_process_move): Assert
2437 that dclass and sclass are greater than or equal to NO_REGS.
2439 2021-09-18 Jakub Jelinek <jakub@redhat.com>
2441 * tree.h (OMP_CLAUSE_ORDER_UNCONSTRAINED): Define.
2442 * tree-pretty-print.c (dump_omp_clause): Print unconstrained:
2443 for OMP_CLAUSE_ORDER_UNCONSTRAINED.
2445 2021-09-18 liuhongt <hongtao.liu@intel.com>
2447 * config/i386/i386-features.c (remove_partial_avx_dependency):
2448 Restrict TARGET_USE_VECTOR_FP_CONVERTS and
2449 TARGET_USE_VECTOR_CONVERTS to conversion instructions only.
2451 2021-09-18 Jakub Jelinek <jakub@redhat.com>
2453 * gimplify.c (omp_default_clause): For C/C++ default({,first}private),
2454 if file/namespace scope variable doesn't have predetermined sharing,
2455 treat it as if there was default(none).
2457 2021-09-18 liuhongt <hongtao.liu@intel.com>
2459 * config/i386/avx512fp16intrin.h (_mm_fmadd_sh):
2461 (_mm_mask_fmadd_sh): Likewise.
2462 (_mm_mask3_fmadd_sh): Likewise.
2463 (_mm_maskz_fmadd_sh): Likewise.
2464 (_mm_fmadd_round_sh): Likewise.
2465 (_mm_mask_fmadd_round_sh): Likewise.
2466 (_mm_mask3_fmadd_round_sh): Likewise.
2467 (_mm_maskz_fmadd_round_sh): Likewise.
2468 (_mm_fnmadd_sh): Likewise.
2469 (_mm_mask_fnmadd_sh): Likewise.
2470 (_mm_mask3_fnmadd_sh): Likewise.
2471 (_mm_maskz_fnmadd_sh): Likewise.
2472 (_mm_fnmadd_round_sh): Likewise.
2473 (_mm_mask_fnmadd_round_sh): Likewise.
2474 (_mm_mask3_fnmadd_round_sh): Likewise.
2475 (_mm_maskz_fnmadd_round_sh): Likewise.
2476 (_mm_fmsub_sh): Likewise.
2477 (_mm_mask_fmsub_sh): Likewise.
2478 (_mm_mask3_fmsub_sh): Likewise.
2479 (_mm_maskz_fmsub_sh): Likewise.
2480 (_mm_fmsub_round_sh): Likewise.
2481 (_mm_mask_fmsub_round_sh): Likewise.
2482 (_mm_mask3_fmsub_round_sh): Likewise.
2483 (_mm_maskz_fmsub_round_sh): Likewise.
2484 (_mm_fnmsub_sh): Likewise.
2485 (_mm_mask_fnmsub_sh): Likewise.
2486 (_mm_mask3_fnmsub_sh): Likewise.
2487 (_mm_maskz_fnmsub_sh): Likewise.
2488 (_mm_fnmsub_round_sh): Likewise.
2489 (_mm_mask_fnmsub_round_sh): Likewise.
2490 (_mm_mask3_fnmsub_round_sh): Likewise.
2491 (_mm_maskz_fnmsub_round_sh): Likewise.
2492 * config/i386/i386-builtin-types.def
2493 (V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT): New builtin type.
2494 * config/i386/i386-builtin.def: Add new builtins.
2495 * config/i386/i386-expand.c: Handle new builtin type.
2496 * config/i386/sse.md (fmai_vmfmadd_<mode><round_name>):
2497 Ajdust to support FP16.
2498 (fmai_vmfmsub_<mode><round_name>): Ditto.
2499 (fmai_vmfnmadd_<mode><round_name>): Ditto.
2500 (fmai_vmfnmsub_<mode><round_name>): Ditto.
2501 (*fmai_fmadd_<mode>): Ditto.
2502 (*fmai_fmsub_<mode>): Ditto.
2503 (*fmai_fnmadd_<mode><round_name>): Ditto.
2504 (*fmai_fnmsub_<mode><round_name>): Ditto.
2505 (avx512f_vmfmadd_<mode>_mask<round_name>): Ditto.
2506 (avx512f_vmfmadd_<mode>_mask3<round_name>): Ditto.
2507 (avx512f_vmfmadd_<mode>_maskz<round_expand_name>): Ditto.
2508 (avx512f_vmfmadd_<mode>_maskz_1<round_name>): Ditto.
2509 (*avx512f_vmfmsub_<mode>_mask<round_name>): Ditto.
2510 (avx512f_vmfmsub_<mode>_mask3<round_name>): Ditto.
2511 (*avx512f_vmfmsub_<mode>_maskz_1<round_name>): Ditto.
2512 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
2513 (*avx512f_vmfnmsub_<mode>_mask3<round_name>): Ditto.
2514 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
2515 (*avx512f_vmfnmadd_<mode>_mask<round_name>): Renamed to ...
2516 (avx512f_vmfnmadd_<mode>_mask<round_name>) ... this, and
2517 adjust to support FP16.
2518 (avx512f_vmfnmadd_<mode>_mask3<round_name>): Ditto.
2519 (avx512f_vmfnmadd_<mode>_maskz_1<round_name>): Ditto.
2520 (avx512f_vmfnmadd_<mode>_maskz<round_expand_name>): New
2523 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
2525 * config/i386/sse.md (avx512fmaskmodelower): Extend to support
2527 (maskload<mode><avx512fmaskmodelower>): Ditto.
2528 (maskstore<mode><avx512fmaskmodelower>): Ditto.
2530 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
2532 * config/i386/i386-expand.c (ix86_expand_fp_absneg_operator):
2534 (ix86_expand_copysign): Ditto.
2535 (ix86_expand_xorsign): Ditto.
2536 * config/i386/i386.c (ix86_build_const_vector): Handle HF vector
2538 (ix86_build_signbit_mask): Ditto.
2539 (ix86_can_change_mode_class): Ditto.
2540 * config/i386/i386.md
2541 (SSEMODEF): Add HFmode.
2542 (ssevecmodef): Ditto.
2543 (<code>hf2): New define_expand.
2544 (*<code>hf2_1): New define_insn_and_split.
2545 (copysign<mode>): Extend to support HFmode under AVX512FP16.
2546 (xorsign<mode>): Ditto.
2547 * config/i386/sse.md (VFB): New mode iterator.
2548 (VFB_128_256): Ditto.
2550 (sseintvecmode2): Support HF vector mode.
2551 (<code><mode>2): Use new mode iterator.
2552 (*<code><mode>2): Ditto.
2553 (copysign<mode>3): Ditto.
2554 (xorsign<mode>3): Ditto.
2555 (<code><mode>3<mask_name>): Ditto.
2556 (<code><mode>3<mask_name>): Ditto.
2557 (<sse>_andnot<mode>3<mask_name>): Adjust for HF vector mode.
2558 (<sse>_andnot<mode>3<mask_name>): Ditto.
2559 (*<code><mode>3<mask_name>): Ditto.
2560 (*<code><mode>3<mask_name>): Ditto.
2562 2021-09-18 liuhongt <hongtao.liu@intel.com>
2564 * config/i386/avx512fp16intrin.h (_mm512_mask_fmadd_ph):
2566 (_mm512_mask3_fmadd_ph): Likewise.
2567 (_mm512_maskz_fmadd_ph): Likewise.
2568 (_mm512_fmadd_round_ph): Likewise.
2569 (_mm512_mask_fmadd_round_ph): Likewise.
2570 (_mm512_mask3_fmadd_round_ph): Likewise.
2571 (_mm512_maskz_fmadd_round_ph): Likewise.
2572 (_mm512_fnmadd_ph): Likewise.
2573 (_mm512_mask_fnmadd_ph): Likewise.
2574 (_mm512_mask3_fnmadd_ph): Likewise.
2575 (_mm512_maskz_fnmadd_ph): Likewise.
2576 (_mm512_fnmadd_round_ph): Likewise.
2577 (_mm512_mask_fnmadd_round_ph): Likewise.
2578 (_mm512_mask3_fnmadd_round_ph): Likewise.
2579 (_mm512_maskz_fnmadd_round_ph): Likewise.
2580 (_mm512_fmsub_ph): Likewise.
2581 (_mm512_mask_fmsub_ph): Likewise.
2582 (_mm512_mask3_fmsub_ph): Likewise.
2583 (_mm512_maskz_fmsub_ph): Likewise.
2584 (_mm512_fmsub_round_ph): Likewise.
2585 (_mm512_mask_fmsub_round_ph): Likewise.
2586 (_mm512_mask3_fmsub_round_ph): Likewise.
2587 (_mm512_maskz_fmsub_round_ph): Likewise.
2588 (_mm512_fnmsub_ph): Likewise.
2589 (_mm512_mask_fnmsub_ph): Likewise.
2590 (_mm512_mask3_fnmsub_ph): Likewise.
2591 (_mm512_maskz_fnmsub_ph): Likewise.
2592 (_mm512_fnmsub_round_ph): Likewise.
2593 (_mm512_mask_fnmsub_round_ph): Likewise.
2594 (_mm512_mask3_fnmsub_round_ph): Likewise.
2595 (_mm512_maskz_fnmsub_round_ph): Likewise.
2596 * config/i386/avx512fp16vlintrin.h (_mm256_fmadd_ph):
2598 (_mm256_mask_fmadd_ph): Likewise.
2599 (_mm256_mask3_fmadd_ph): Likewise.
2600 (_mm256_maskz_fmadd_ph): Likewise.
2601 (_mm_fmadd_ph): Likewise.
2602 (_mm_mask_fmadd_ph): Likewise.
2603 (_mm_mask3_fmadd_ph): Likewise.
2604 (_mm_maskz_fmadd_ph): Likewise.
2605 (_mm256_fnmadd_ph): Likewise.
2606 (_mm256_mask_fnmadd_ph): Likewise.
2607 (_mm256_mask3_fnmadd_ph): Likewise.
2608 (_mm256_maskz_fnmadd_ph): Likewise.
2609 (_mm_fnmadd_ph): Likewise.
2610 (_mm_mask_fnmadd_ph): Likewise.
2611 (_mm_mask3_fnmadd_ph): Likewise.
2612 (_mm_maskz_fnmadd_ph): Likewise.
2613 (_mm256_fmsub_ph): Likewise.
2614 (_mm256_mask_fmsub_ph): Likewise.
2615 (_mm256_mask3_fmsub_ph): Likewise.
2616 (_mm256_maskz_fmsub_ph): Likewise.
2617 (_mm_fmsub_ph): Likewise.
2618 (_mm_mask_fmsub_ph): Likewise.
2619 (_mm_mask3_fmsub_ph): Likewise.
2620 (_mm_maskz_fmsub_ph): Likewise.
2621 (_mm256_fnmsub_ph): Likewise.
2622 (_mm256_mask_fnmsub_ph): Likewise.
2623 (_mm256_mask3_fnmsub_ph): Likewise.
2624 (_mm256_maskz_fnmsub_ph): Likewise.
2625 (_mm_fnmsub_ph): Likewise.
2626 (_mm_mask_fnmsub_ph): Likewise.
2627 (_mm_mask3_fnmsub_ph): Likewise.
2628 (_mm_maskz_fnmsub_ph): Likewise.
2629 * config/i386/i386-builtin.def: Add corresponding new builtins.
2630 * config/i386/sse.md
2631 (<avx512>_fmadd_<mode>_maskz<round_expand_name>): Adjust to
2632 support HF vector modes.
2633 (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
2635 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
2636 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
2637 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
2638 (<avx512>_fmadd_<mode>_mask<round_name>): Ditto.
2639 (<avx512>_fmadd_<mode>_mask3<round_name>): Ditto.
2640 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Ditto.
2641 (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>):
2643 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
2644 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
2645 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
2646 (<avx512>_fmsub_<mode>_mask<round_name>): Ditto.
2647 (<avx512>_fmsub_<mode>_mask3<round_name>): Ditto.
2648 (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>):
2650 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
2651 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
2652 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
2653 (<avx512>_fnmadd_<mode>_mask<round_name>): Ditto.
2654 (<avx512>_fnmadd_<mode>_mask3<round_name>): Ditto.
2655 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Ditto.
2656 (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>):
2658 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
2659 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
2660 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
2661 (<avx512>_fnmsub_<mode>_mask<round_name>): Ditto.
2662 (<avx512>_fnmsub_<mode>_mask3<round_name>): Ditto.
2664 2021-09-18 liuhongt <hongtao.liu@intel.com>
2666 * config/i386/avx512fp16intrin.h (_mm512_fmaddsub_ph):
2668 (_mm512_mask_fmaddsub_ph): Likewise.
2669 (_mm512_mask3_fmaddsub_ph): Likewise.
2670 (_mm512_maskz_fmaddsub_ph): Likewise.
2671 (_mm512_fmaddsub_round_ph): Likewise.
2672 (_mm512_mask_fmaddsub_round_ph): Likewise.
2673 (_mm512_mask3_fmaddsub_round_ph): Likewise.
2674 (_mm512_maskz_fmaddsub_round_ph): Likewise.
2675 (_mm512_mask_fmsubadd_ph): Likewise.
2676 (_mm512_mask3_fmsubadd_ph): Likewise.
2677 (_mm512_maskz_fmsubadd_ph): Likewise.
2678 (_mm512_fmsubadd_round_ph): Likewise.
2679 (_mm512_mask_fmsubadd_round_ph): Likewise.
2680 (_mm512_mask3_fmsubadd_round_ph): Likewise.
2681 (_mm512_maskz_fmsubadd_round_ph): Likewise.
2682 * config/i386/avx512fp16vlintrin.h (_mm256_fmaddsub_ph):
2684 (_mm256_mask_fmaddsub_ph): Likewise.
2685 (_mm256_mask3_fmaddsub_ph): Likewise.
2686 (_mm256_maskz_fmaddsub_ph): Likewise.
2687 (_mm_fmaddsub_ph): Likewise.
2688 (_mm_mask_fmaddsub_ph): Likewise.
2689 (_mm_mask3_fmaddsub_ph): Likewise.
2690 (_mm_maskz_fmaddsub_ph): Likewise.
2691 (_mm256_fmsubadd_ph): Likewise.
2692 (_mm256_mask_fmsubadd_ph): Likewise.
2693 (_mm256_mask3_fmsubadd_ph): Likewise.
2694 (_mm256_maskz_fmsubadd_ph): Likewise.
2695 (_mm_fmsubadd_ph): Likewise.
2696 (_mm_mask_fmsubadd_ph): Likewise.
2697 (_mm_mask3_fmsubadd_ph): Likewise.
2698 (_mm_maskz_fmsubadd_ph): Likewise.
2699 * config/i386/i386-builtin.def: Add corresponding new builtins.
2700 * config/i386/sse.md (VFH_SF_AVX512VL): New mode iterator.
2701 * (<avx512>_fmsubadd_<mode>_maskz<round_expand_name>): New expander.
2702 * (<avx512>_fmaddsub_<mode>_maskz<round_expand_name>): Use
2704 * (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
2706 * (<avx512>_fmaddsub_<mode>_mask<round_name>): Ditto.
2707 * (<avx512>_fmaddsub_<mode>_mask3<round_name>): Ditto.
2708 * (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
2710 * (<avx512>_fmsubadd_<mode>_mask<round_name>): Ditto.
2711 * (<avx512>_fmsubadd_<mode>_mask3<round_name>): Ditto.
2713 2021-09-18 liuhongt <hongtao.liu@intel.com>
2716 * config/i386/i386.c (ix86_print_operand): Handle
2717 V8HF/V16HF/V32HFmode.
2718 * config/i386/i386.h (VALID_BCST_MODE_P): Add HFmode.
2719 * config/i386/sse.md (avx512bcst): Remove.
2721 2021-09-17 Martin Sebor <msebor@redhat.com>
2723 * Makefile.in (OBJS): Add gimple-predicate-analysis.o.
2724 * tree-ssa-uninit.c (max_phi_args): Move to gimple-predicate-analysis.
2725 (MASK_SET_BIT, MASK_TEST_BIT, MASK_EMPTY): Same.
2726 (check_defs): Add comment.
2727 (can_skip_redundant_opnd): Update comment.
2728 (compute_uninit_opnds_pos): Adjust to namespace change.
2729 (find_pdom): Move to gimple-predicate-analysis.cc.
2731 (struct uninit_undef_val_t): New.
2732 (is_non_loop_exit_postdominating): Move to gimple-predicate-analysis.cc.
2733 (find_control_equiv_block): Same.
2734 (MAX_NUM_CHAINS, MAX_CHAIN_LEN, MAX_POSTDOM_CHECK): Same.
2735 (MAX_SWITCH_CASES): Same.
2736 (compute_control_dep_chain): Same.
2737 (find_uninit_use): Use predicate analyzer.
2738 (struct pred_info): Move to gimple-predicate-analysis.
2739 (convert_control_dep_chain_into_preds): Same.
2740 (find_predicates): Same.
2741 (collect_phi_def_edges): Same.
2742 (warn_uninitialized_phi): Use predicate analyzer.
2743 (find_def_preds): Move to gimple-predicate-analysis.
2744 (dump_pred_info): Same.
2745 (dump_pred_chain): Same.
2746 (dump_predicates): Same.
2747 (destroy_predicate_vecs): Remove.
2748 (execute_late_warn_uninitialized): New.
2749 (get_cmp_code): Move to gimple-predicate-analysis.
2750 (is_value_included_in): Same.
2751 (value_sat_pred_p): Same.
2752 (find_matching_predicate_in_rest_chains): Same.
2753 (is_use_properly_guarded): Same.
2754 (prune_uninit_phi_opnds): Same.
2755 (find_var_cmp_const): Same.
2756 (use_pred_not_overlap_with_undef_path_pred): Same.
2757 (pred_equal_p): Same.
2758 (is_neq_relop_p): Same.
2759 (is_neq_zero_form_p): Same.
2760 (pred_expr_equal_p): Same.
2761 (is_pred_expr_subset_of): Same.
2762 (is_pred_chain_subset_of): Same.
2763 (is_included_in): Same.
2764 (is_superset_of): Same.
2766 (simplify_pred): Same.
2767 (simplify_preds_2): Same.
2768 (simplify_preds_3): Same.
2769 (simplify_preds_4): Same.
2770 (simplify_preds): Same.
2772 (push_to_worklist): Same.
2773 (get_pred_info_from_cmp): Same.
2774 (is_degenerated_phi): Same.
2775 (normalize_one_pred_1): Same.
2776 (normalize_one_pred): Same.
2777 (normalize_one_pred_chain): Same.
2778 (normalize_preds): Same.
2779 (can_one_predicate_be_invalidated_p): Same.
2780 (can_chain_union_be_invalidated_p): Same.
2781 (uninit_uses_cannot_happen): Same.
2782 (pass_late_warn_uninitialized::execute): Define.
2783 * gimple-predicate-analysis.cc: New file.
2784 * gimple-predicate-analysis.h: New file.
2786 2021-09-17 Julian Brown <julian@codesourcery.com>
2788 * config/gcn/gcn.c (gimple.h): Include.
2789 (gcn_fork_join): Emit barrier for worker-level joins.
2790 * omp-oacc-neuter-broadcast.cc (find_local_vars_to_propagate): Add
2791 writes_gang_private bitmap parameter. Set bit for blocks
2792 containing gang-private variable writes.
2793 (worker_single_simple): Don't emit barrier after predicated block.
2794 (worker_single_copy): Don't emit barrier if we're not broadcasting
2795 anything and the block contains no gang-private writes.
2796 (neuter_worker_single): Don't predicate blocks that only contain
2797 NOPs or internal marker functions. Pass has_gang_private_write
2798 argument to worker_single_copy.
2799 (oacc_do_neutering): Add writes_gang_private bitmap handling.
2801 2021-09-17 Julian Brown <julian@codesourcery.com>
2803 * config/gcn/gcn-protos.h
2804 (gcn_goacc_create_worker_broadcast_record): Update prototype.
2805 * config/gcn/gcn-tree.c (gcn_goacc_get_worker_red_decl): Use
2806 preallocated block of LDS memory. Do not cache/share decls for
2807 reduction temporaries between invocations.
2808 (gcn_goacc_reduction_teardown): Unshare VAR on second use.
2809 (gcn_goacc_create_worker_broadcast_record): Add OFFSET parameter
2810 and return temporary LDS space at that offset. Return pointer in
2812 * config/gcn/gcn.c (acc_lds_size, gang_private_hwm, lds_allocs):
2814 (ACC_LDS_SIZE): Define as acc_lds_size.
2815 (gcn_init_machine_status): Don't initialise lds_allocated,
2816 lds_allocs, reduc_decls fields of machine function struct.
2817 (gcn_option_override): Handle default size for gang-private
2818 variables and -mgang-private-size option.
2819 (gcn_expand_prologue): Use LDS_SIZE instead of LDS_SIZE-1 when
2820 initialising M0_REG.
2821 (gcn_shared_mem_layout): New function.
2822 (gcn_print_lds_decl): Update comment. Use global lds_allocs map and
2823 gang_private_hwm variable.
2824 (TARGET_GOACC_SHARED_MEM_LAYOUT): Define target hook.
2825 * config/gcn/gcn.h (machine_function): Remove lds_allocated,
2826 lds_allocs, reduc_decls. Add reduction_base, reduction_limit.
2827 * config/gcn/gcn.opt (gang_private_size_opt): New global.
2828 (mgang-private-size=): New option.
2829 * doc/tm.texi.in (TARGET_GOACC_SHARED_MEM_LAYOUT): Place
2831 * doc/tm.texi: Regenerate.
2832 * omp-oacc-neuter-broadcast.cc (targhooks.h, diagnostic-core.h):
2834 (build_sender_ref): Handle sender_decl being pointer.
2835 (worker_single_copy): Add PLACEMENT and ISOLATE_BROADCASTS
2836 parameters. Pass placement argument to
2837 create_worker_broadcast_record hook invocations. Handle
2838 sender_decl being pointer and isolate_broadcasts inserting extra
2840 (blk_offset_map_t): Add typedef.
2841 (neuter_worker_single): Add BLK_OFFSET_MAP parameter. Pass
2842 preallocated range to worker_single_copy call.
2843 (dfs_broadcast_reachable_1): New function.
2844 (idx_decl_pair_t, used_range_vec_t): New typedefs.
2845 (sort_size_descending): New function.
2846 (addr_range): New class.
2847 (splay_tree_compare_addr_range, splay_tree_free_key)
2848 (first_fit_range, merge_ranges_1, merge_ranges): New functions.
2849 (execute_omp_oacc_neuter_broadcast): Rename to...
2850 (oacc_do_neutering): ... this. Add BOUNDS_LO, BOUNDS_HI
2851 parameters. Arrange layout of shared memory for broadcast
2853 (execute_omp_oacc_neuter_broadcast): New function.
2854 (pass_omp_oacc_neuter_broadcast::gate): Remove num_workers==1
2855 handling from here. Enable pass for all OpenACC routines in order
2856 to call shared memory-layout hook.
2857 * target.def (create_worker_broadcast_record): Add OFFSET
2859 (shared_mem_layout): New hook.
2861 2021-09-17 Julian Brown <julian@codesourcery.com>
2862 Thomas Schwinge <thomas@codesourcery.com>
2864 * omp-oacc-neuter-broadcast.cc
2865 (pass_omp_oacc_neuter_broadcast::gate): Disable if num_workers is
2867 (execute_omp_oacc_neuter_broadcast): Adjust.
2869 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
2871 * value-relation.cc (class equiv_chain): Move to header file.
2872 (path_oracle::path_oracle): New.
2873 (path_oracle::~path_oracle): New.
2874 (path_oracle::register_relation): New.
2875 (path_oracle::query_relation): New.
2876 (path_oracle::reset_path): New.
2877 (path_oracle::dump): New.
2878 * value-relation.h (class equiv_chain): Move to here.
2879 (class path_oracle): New.
2881 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
2883 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a DOM
2885 * gimple-range-fold.cc (fur_depend::register_relation): Use
2886 register_stmt/edge routines.
2887 * value-relation.cc (equiv_chain::find): Relocate from equiv_oracle.
2888 (equiv_oracle::equiv_oracle): Create self equivalence cache.
2889 (equiv_oracle::~equiv_oracle): Release same.
2890 (equiv_oracle::equiv_set): Return entry from self equiv cache if there
2891 are no equivalences.
2892 (equiv_oracle::find_equiv_block): Move list find to equiv_chain.
2893 (equiv_oracle::register_relation): Rename from register_equiv.
2894 (relation_chain_head::find_relation): Relocate from dom_oracle.
2895 (relation_oracle::register_stmt): New.
2896 (relation_oracle::register_edge): New.
2897 (dom_oracle::*): Rename from relation_oracle.
2898 (dom_oracle::register_relation): Adjust to call equiv_oracle.
2899 (dom_oracle::set_one_relation): Split from register_relation.
2900 (dom_oracle::register_transitives): Consolidate 2 methods.
2901 (dom_oracle::find_relation_block): Move core to relation_chain.
2902 (dom_oracle::query_relation): Rename from find_relation_dom and adjust.
2903 * value-relation.h (class relation_oracle): New pure virtual base.
2904 (class equiv_oracle): Inherit from relation_oracle and adjust.
2905 (class dom_oracle): Rename from old relation_oracle and adjust.
2907 2021-09-17 Martin Sebor <msebor@redhat.com>
2909 PR middle-end/102200
2910 * pointer-query.cc (access_ref::inform_access): Handle MIN/MAX_EXPR.
2911 (handle_min_max_size): Change argument. Store original SSA_NAME for
2912 operands to potentially distinct (sub)objects.
2913 (compute_objsize_r): Adjust call to the above.
2915 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
2917 * config/rs6000/rs6000.c (rs6000-builtins.h): New include.
2918 (rs6000_new_builtin_vectorized_function): New function.
2919 (rs6000_new_builtin_md_vectorized_function): Likewise.
2920 (rs6000_builtin_vectorized_function): Call
2921 rs6000_new_builtin_vectorized_function.
2922 (rs6000_builtin_md_vectorized_function): Call
2923 rs6000_new_builtin_md_vectorized_function.
2925 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
2927 * config/rs6000/rs6000-builtin-new.def (ASSEMBLE_ACC): Add mmaint flag.
2928 (ASSEMBLE_PAIR): Likewise.
2929 (BUILD_ACC): Likewise.
2930 (DISASSEMBLE_ACC): Likewise.
2931 (DISASSEMBLE_PAIR): Likewise.
2932 (PMXVBF16GER2): Likewise.
2933 (PMXVBF16GER2NN): Likewise.
2934 (PMXVBF16GER2NP): Likewise.
2935 (PMXVBF16GER2PN): Likewise.
2936 (PMXVBF16GER2PP): Likewise.
2937 (PMXVF16GER2): Likewise.
2938 (PMXVF16GER2NN): Likewise.
2939 (PMXVF16GER2NP): Likewise.
2940 (PMXVF16GER2PN): Likewise.
2941 (PMXVF16GER2PP): Likewise.
2942 (PMXVF32GER): Likewise.
2943 (PMXVF32GERNN): Likewise.
2944 (PMXVF32GERNP): Likewise.
2945 (PMXVF32GERPN): Likewise.
2946 (PMXVF32GERPP): Likewise.
2947 (PMXVF64GER): Likewise.
2948 (PMXVF64GERNN): Likewise.
2949 (PMXVF64GERNP): Likewise.
2950 (PMXVF64GERPN): Likewise.
2951 (PMXVF64GERPP): Likewise.
2952 (PMXVI16GER2): Likewise.
2953 (PMXVI16GER2PP): Likewise.
2954 (PMXVI16GER2S): Likewise.
2955 (PMXVI16GER2SPP): Likewise.
2956 (PMXVI4GER8): Likewise.
2957 (PMXVI4GER8PP): Likewise.
2958 (PMXVI8GER4): Likewise.
2959 (PMXVI8GER4PP): Likewise.
2960 (PMXVI8GER4SPP): Likewise.
2961 (XVBF16GER2): Likewise.
2962 (XVBF16GER2NN): Likewise.
2963 (XVBF16GER2NP): Likewise.
2964 (XVBF16GER2PN): Likewise.
2965 (XVBF16GER2PP): Likewise.
2966 (XVF16GER2): Likewise.
2967 (XVF16GER2NN): Likewise.
2968 (XVF16GER2NP): Likewise.
2969 (XVF16GER2PN): Likewise.
2970 (XVF16GER2PP): Likewise.
2971 (XVF32GER): Likewise.
2972 (XVF32GERNN): Likewise.
2973 (XVF32GERNP): Likewise.
2974 (XVF32GERPN): Likewise.
2975 (XVF32GERPP): Likewise.
2976 (XVF64GER): Likewise.
2977 (XVF64GERNN): Likewise.
2978 (XVF64GERNP): Likewise.
2979 (XVF64GERPN): Likewise.
2980 (XVF64GERPP): Likewise.
2981 (XVI16GER2): Likewise.
2982 (XVI16GER2PP): Likewise.
2983 (XVI16GER2S): Likewise.
2984 (XVI16GER2SPP): Likewise.
2985 (XVI4GER8): Likewise.
2986 (XVI4GER8PP): Likewise.
2987 (XVI8GER4): Likewise.
2988 (XVI8GER4PP): Likewise.
2989 (XVI8GER4SPP): Likewise.
2990 (XXMFACC): Likewise.
2991 (XXMTACC): Likewise.
2992 (XXSETACCZ): Likewise.
2993 (ASSEMBLE_PAIR_V): Likewise.
2994 (BUILD_PAIR): Likewise.
2995 (DISASSEMBLE_PAIR_V): Likewise.
2998 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_mma_builtin):
2999 Handle RS6000_BIF_LXVP and RS6000_BIF_STXVP.
3000 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add ismmaint.
3001 (parse_bif_attrs): Handle ismmaint.
3002 (write_decls): Add bif_mmaint_bit and bif_is_mmaint.
3003 (write_bif_static_init): Handle ismmaint.
3005 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
3007 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin): New
3009 (rs6000_gimple_fold_builtin): Call rs6000_gimple_fold_new_builtin.
3010 (rs6000_new_builtin_valid_without_lhs): New function.
3011 (rs6000_gimple_fold_new_mma_builtin): Likewise.
3012 (rs6000_gimple_fold_new_builtin): Likewise.
3014 2021-09-17 Thomas Schwinge <thomas@codesourcery.com>
3016 * hash-table.h (hash_table<Descriptor, Lazy, Allocator>::expand):
3017 Destruct stale Value objects.
3018 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
3021 2021-09-17 Roger Sayle <roger@nextmovesoftware.com>
3024 * match.pd (shift optimizations): Disable recent sign-changing
3025 optimization for shifts by zero, these will be folded later.
3027 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
3029 * config/rs6000/rs6000-builtin-new.def (__builtin_mffsl): Move from
3030 [power9] to [always].
3032 2021-09-17 Richard Biener <rguenther@suse.de>
3034 * tree-vect-stmts.c (vectorizable_load): Do not frob
3037 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
3039 * config/i386/i386-features.c (remove_partial_avx_dependency):
3040 Also check TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY and
3041 and TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY before generating
3043 * config/i386/i386.h (TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY):
3045 (TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
3046 * config/i386/i386.md (SSE FP to FP splitters): Replace
3047 TARGET_SSE_PARTIAL_REG_DEPENDENCY with
3048 TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY.
3049 (SSE INT to FP splitter): Replace TARGET_SSE_PARTIAL_REG_DEPENDENCY
3050 with TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY.
3051 * config/i386/x86-tune.def
3052 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): New.
3053 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
3055 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
3058 * config/i386/i386-features.c (remove_partial_avx_dependency):
3059 Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
3060 before generating vxorps.
3062 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
3064 * config/i386/i386-options.c (processor_cost_table): Use
3065 tremont_cost for Tremont.
3066 * config/i386/x86-tune-costs.h (tremont_memcpy): New.
3067 (tremont_memset): Likewise.
3068 (tremont_cost): Likewise.
3069 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
3072 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
3074 * common/config/i386/i386-common.c: Use Haswell scheduling model
3076 * config/i386/i386.c (ix86_sched_init_global): Prepare for Tremont
3078 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Tremont
3080 (ix86_adjust_cost): Handle Tremont.
3081 * config/i386/x86-tune.def (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY):
3083 (X86_TUNE_USE_LEAVE): Likewise.
3084 (X86_TUNE_PUSH_MEMORY): Likewise.
3085 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
3086 (X86_TUNE_USE_CLTD): Likewise.
3087 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
3088 (X86_TUNE_AVOID_MFENCE): Likewise.
3089 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
3090 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
3091 (X86_TUNE_ACCUMULATE_OUTGOING_ARGS): Disable for Tremont.
3092 (X86_TUNE_FOUR_JUMP_LIMIT): Likewise.
3093 (X86_TUNE_OPT_AGU): Likewise.
3094 (X86_TUNE_AVOID_LEA_FOR_ADDR): Likewise.
3095 (X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE): Likewise.
3096 (X86_TUNE_EXPAND_ABS): Likewise.
3097 (X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS): Likewise.
3098 (X86_TUNE_SLOW_PSHUFB): Likewise.
3100 2021-09-17 Eric Botcazou <ebotcazou@adacore.com>
3102 PR rtl-optimization/102306
3103 * combine.c (try_combine): Abort the combination if we are about to
3104 duplicate volatile references.
3106 2021-09-17 liuhongt <hongtao.liu@intel.com>
3108 * config/i386/avx512fp16intrin.h (_mm_undefined_ph):
3110 (_mm256_undefined_ph): Likewise.
3111 (_mm512_undefined_ph): Likewise.
3112 (_mm_cvtsh_h): Likewise.
3113 (_mm256_cvtsh_h): Likewise.
3114 (_mm512_cvtsh_h): Likewise.
3115 (_mm512_castph_ps): Likewise.
3116 (_mm512_castph_pd): Likewise.
3117 (_mm512_castph_si512): Likewise.
3118 (_mm512_castph512_ph128): Likewise.
3119 (_mm512_castph512_ph256): Likewise.
3120 (_mm512_castph128_ph512): Likewise.
3121 (_mm512_castph256_ph512): Likewise.
3122 (_mm512_zextph128_ph512): Likewise.
3123 (_mm512_zextph256_ph512): Likewise.
3124 (_mm512_castps_ph): Likewise.
3125 (_mm512_castpd_ph): Likewise.
3126 (_mm512_castsi512_ph): Likewise.
3127 * config/i386/avx512fp16vlintrin.h (_mm_castph_ps):
3129 (_mm256_castph_ps): Likewise.
3130 (_mm_castph_pd): Likewise.
3131 (_mm256_castph_pd): Likewise.
3132 (_mm_castph_si128): Likewise.
3133 (_mm256_castph_si256): Likewise.
3134 (_mm_castps_ph): Likewise.
3135 (_mm256_castps_ph): Likewise.
3136 (_mm_castpd_ph): Likewise.
3137 (_mm256_castpd_ph): Likewise.
3138 (_mm_castsi128_ph): Likewise.
3139 (_mm256_castsi256_ph): Likewise.
3140 (_mm256_castph256_ph128): Likewise.
3141 (_mm256_castph128_ph256): Likewise.
3142 (_mm256_zextph128_ph256): Likewise.
3144 2021-09-17 liuhongt <hongtao.liu@intel.com>
3146 * config/i386/avx512fp16intrin.h (_mm_cvtsh_ss):
3148 (_mm_mask_cvtsh_ss): Likewise.
3149 (_mm_maskz_cvtsh_ss): Likewise.
3150 (_mm_cvtsh_sd): Likewise.
3151 (_mm_mask_cvtsh_sd): Likewise.
3152 (_mm_maskz_cvtsh_sd): Likewise.
3153 (_mm_cvt_roundsh_ss): Likewise.
3154 (_mm_mask_cvt_roundsh_ss): Likewise.
3155 (_mm_maskz_cvt_roundsh_ss): Likewise.
3156 (_mm_cvt_roundsh_sd): Likewise.
3157 (_mm_mask_cvt_roundsh_sd): Likewise.
3158 (_mm_maskz_cvt_roundsh_sd): Likewise.
3159 (_mm_cvtss_sh): Likewise.
3160 (_mm_mask_cvtss_sh): Likewise.
3161 (_mm_maskz_cvtss_sh): Likewise.
3162 (_mm_cvtsd_sh): Likewise.
3163 (_mm_mask_cvtsd_sh): Likewise.
3164 (_mm_maskz_cvtsd_sh): Likewise.
3165 (_mm_cvt_roundss_sh): Likewise.
3166 (_mm_mask_cvt_roundss_sh): Likewise.
3167 (_mm_maskz_cvt_roundss_sh): Likewise.
3168 (_mm_cvt_roundsd_sh): Likewise.
3169 (_mm_mask_cvt_roundsd_sh): Likewise.
3170 (_mm_maskz_cvt_roundsd_sh): Likewise.
3171 * config/i386/i386-builtin-types.def
3172 (V8HF_FTYPE_V2DF_V8HF_V8HF_UQI_INT,
3173 V8HF_FTYPE_V4SF_V8HF_V8HF_UQI_INT,
3174 V2DF_FTYPE_V8HF_V2DF_V2DF_UQI_INT,
3175 V4SF_FTYPE_V8HF_V4SF_V4SF_UQI_INT): Add new builtin types.
3176 * config/i386/i386-builtin.def: Add corrresponding new builtins.
3177 * config/i386/i386-expand.c: Handle new builtin types.
3178 * config/i386/sse.md (VF48_128): New mode iterator.
3179 (avx512fp16_vcvtsh2<ssescalarmodesuffix><mask_scalar_name><round_saeonly_scalar_name>):
3181 (avx512fp16_vcvt<ssescalarmodesuffix>2sh<mask_scalar_name><round_scalar_name>):
3184 2021-09-17 liuhongt <hongtao.liu@intel.com>
3186 * config/i386/avx512fp16intrin.h (_mm512_cvtph_pd):
3188 (_mm512_mask_cvtph_pd): Likewise.
3189 (_mm512_maskz_cvtph_pd): Likewise.
3190 (_mm512_cvt_roundph_pd): Likewise.
3191 (_mm512_mask_cvt_roundph_pd): Likewise.
3192 (_mm512_maskz_cvt_roundph_pd): Likewise.
3193 (_mm512_cvtxph_ps): Likewise.
3194 (_mm512_mask_cvtxph_ps): Likewise.
3195 (_mm512_maskz_cvtxph_ps): Likewise.
3196 (_mm512_cvtx_roundph_ps): Likewise.
3197 (_mm512_mask_cvtx_roundph_ps): Likewise.
3198 (_mm512_maskz_cvtx_roundph_ps): Likewise.
3199 (_mm512_cvtxps_ph): Likewise.
3200 (_mm512_mask_cvtxps_ph): Likewise.
3201 (_mm512_maskz_cvtxps_ph): Likewise.
3202 (_mm512_cvtx_roundps_ph): Likewise.
3203 (_mm512_mask_cvtx_roundps_ph): Likewise.
3204 (_mm512_maskz_cvtx_roundps_ph): Likewise.
3205 (_mm512_cvtpd_ph): Likewise.
3206 (_mm512_mask_cvtpd_ph): Likewise.
3207 (_mm512_maskz_cvtpd_ph): Likewise.
3208 (_mm512_cvt_roundpd_ph): Likewise.
3209 (_mm512_mask_cvt_roundpd_ph): Likewise.
3210 (_mm512_maskz_cvt_roundpd_ph): Likewise.
3211 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_pd):
3213 (_mm_mask_cvtph_pd): Likewise.
3214 (_mm_maskz_cvtph_pd): Likewise.
3215 (_mm256_cvtph_pd): Likewise.
3216 (_mm256_mask_cvtph_pd): Likewise.
3217 (_mm256_maskz_cvtph_pd): Likewise.
3218 (_mm_cvtxph_ps): Likewise.
3219 (_mm_mask_cvtxph_ps): Likewise.
3220 (_mm_maskz_cvtxph_ps): Likewise.
3221 (_mm256_cvtxph_ps): Likewise.
3222 (_mm256_mask_cvtxph_ps): Likewise.
3223 (_mm256_maskz_cvtxph_ps): Likewise.
3224 (_mm_cvtxps_ph): Likewise.
3225 (_mm_mask_cvtxps_ph): Likewise.
3226 (_mm_maskz_cvtxps_ph): Likewise.
3227 (_mm256_cvtxps_ph): Likewise.
3228 (_mm256_mask_cvtxps_ph): Likewise.
3229 (_mm256_maskz_cvtxps_ph): Likewise.
3230 (_mm_cvtpd_ph): Likewise.
3231 (_mm_mask_cvtpd_ph): Likewise.
3232 (_mm_maskz_cvtpd_ph): Likewise.
3233 (_mm256_cvtpd_ph): Likewise.
3234 (_mm256_mask_cvtpd_ph): Likewise.
3235 (_mm256_maskz_cvtpd_ph): Likewise.
3236 * config/i386/i386-builtin.def: Add corresponding new builtins.
3237 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3238 * config/i386/i386-expand.c: Handle new builtin types.
3239 * config/i386/sse.md
3240 (VF4_128_8_256): New.
3241 (VF48H_AVX512VL): Ditto.
3242 (ssePHmode): Add HF vector modes.
3243 (castmode): Add new convertable modes.
3246 (avx512fp16_vcvt<castmode>2ph_<mode><mask_name><round_name>): Ditto.
3247 (avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
3248 (*avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
3249 (avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
3250 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
3251 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask_1): Ditto.
3252 (avx512fp16_float_extend_ph<mode>2<mask_name><round_saeonly_name>):
3254 (avx512fp16_float_extend_ph<mode>2<mask_name>): Ditto.
3255 (*avx512fp16_float_extend_ph<mode>2_load<mask_name>): Ditto.
3256 (avx512fp16_float_extend_phv2df2<mask_name>): Ditto.
3257 (*avx512fp16_float_extend_phv2df2_load<mask_name>): Ditto.
3259 2021-09-17 liuhongt <hongtao.liu@intel.com>
3261 * config/i386/avx512fp16intrin.h (_mm_cvttsh_i32):
3263 (_mm_cvttsh_u32): Likewise.
3264 (_mm_cvtt_roundsh_i32): Likewise.
3265 (_mm_cvtt_roundsh_u32): Likewise.
3266 (_mm_cvttsh_i64): Likewise.
3267 (_mm_cvttsh_u64): Likewise.
3268 (_mm_cvtt_roundsh_i64): Likewise.
3269 (_mm_cvtt_roundsh_u64): Likewise.
3270 * config/i386/i386-builtin.def: Add corresponding new builtins.
3271 * config/i386/sse.md
3272 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<round_saeonly_name>):
3275 2021-09-17 liuhongt <hongtao.liu@intel.com>
3277 * config/i386/avx512fp16intrin.h (_mm512_cvttph_epi32):
3279 (_mm512_mask_cvttph_epi32): Likewise.
3280 (_mm512_maskz_cvttph_epi32): Likewise.
3281 (_mm512_cvtt_roundph_epi32): Likewise.
3282 (_mm512_mask_cvtt_roundph_epi32): Likewise.
3283 (_mm512_maskz_cvtt_roundph_epi32): Likewise.
3284 (_mm512_cvttph_epu32): Likewise.
3285 (_mm512_mask_cvttph_epu32): Likewise.
3286 (_mm512_maskz_cvttph_epu32): Likewise.
3287 (_mm512_cvtt_roundph_epu32): Likewise.
3288 (_mm512_mask_cvtt_roundph_epu32): Likewise.
3289 (_mm512_maskz_cvtt_roundph_epu32): Likewise.
3290 (_mm512_cvttph_epi64): Likewise.
3291 (_mm512_mask_cvttph_epi64): Likewise.
3292 (_mm512_maskz_cvttph_epi64): Likewise.
3293 (_mm512_cvtt_roundph_epi64): Likewise.
3294 (_mm512_mask_cvtt_roundph_epi64): Likewise.
3295 (_mm512_maskz_cvtt_roundph_epi64): Likewise.
3296 (_mm512_cvttph_epu64): Likewise.
3297 (_mm512_mask_cvttph_epu64): Likewise.
3298 (_mm512_maskz_cvttph_epu64): Likewise.
3299 (_mm512_cvtt_roundph_epu64): Likewise.
3300 (_mm512_mask_cvtt_roundph_epu64): Likewise.
3301 (_mm512_maskz_cvtt_roundph_epu64): Likewise.
3302 (_mm512_cvttph_epi16): Likewise.
3303 (_mm512_mask_cvttph_epi16): Likewise.
3304 (_mm512_maskz_cvttph_epi16): Likewise.
3305 (_mm512_cvtt_roundph_epi16): Likewise.
3306 (_mm512_mask_cvtt_roundph_epi16): Likewise.
3307 (_mm512_maskz_cvtt_roundph_epi16): Likewise.
3308 (_mm512_cvttph_epu16): Likewise.
3309 (_mm512_mask_cvttph_epu16): Likewise.
3310 (_mm512_maskz_cvttph_epu16): Likewise.
3311 (_mm512_cvtt_roundph_epu16): Likewise.
3312 (_mm512_mask_cvtt_roundph_epu16): Likewise.
3313 (_mm512_maskz_cvtt_roundph_epu16): Likewise.
3314 * config/i386/avx512fp16vlintrin.h (_mm_cvttph_epi32):
3316 (_mm_mask_cvttph_epi32): Likewise.
3317 (_mm_maskz_cvttph_epi32): Likewise.
3318 (_mm256_cvttph_epi32): Likewise.
3319 (_mm256_mask_cvttph_epi32): Likewise.
3320 (_mm256_maskz_cvttph_epi32): Likewise.
3321 (_mm_cvttph_epu32): Likewise.
3322 (_mm_mask_cvttph_epu32): Likewise.
3323 (_mm_maskz_cvttph_epu32): Likewise.
3324 (_mm256_cvttph_epu32): Likewise.
3325 (_mm256_mask_cvttph_epu32): Likewise.
3326 (_mm256_maskz_cvttph_epu32): Likewise.
3327 (_mm_cvttph_epi64): Likewise.
3328 (_mm_mask_cvttph_epi64): Likewise.
3329 (_mm_maskz_cvttph_epi64): Likewise.
3330 (_mm256_cvttph_epi64): Likewise.
3331 (_mm256_mask_cvttph_epi64): Likewise.
3332 (_mm256_maskz_cvttph_epi64): Likewise.
3333 (_mm_cvttph_epu64): Likewise.
3334 (_mm_mask_cvttph_epu64): Likewise.
3335 (_mm_maskz_cvttph_epu64): Likewise.
3336 (_mm256_cvttph_epu64): Likewise.
3337 (_mm256_mask_cvttph_epu64): Likewise.
3338 (_mm256_maskz_cvttph_epu64): Likewise.
3339 (_mm_cvttph_epi16): Likewise.
3340 (_mm_mask_cvttph_epi16): Likewise.
3341 (_mm_maskz_cvttph_epi16): Likewise.
3342 (_mm256_cvttph_epi16): Likewise.
3343 (_mm256_mask_cvttph_epi16): Likewise.
3344 (_mm256_maskz_cvttph_epi16): Likewise.
3345 (_mm_cvttph_epu16): Likewise.
3346 (_mm_mask_cvttph_epu16): Likewise.
3347 (_mm_maskz_cvttph_epu16): Likewise.
3348 (_mm256_cvttph_epu16): Likewise.
3349 (_mm256_mask_cvttph_epu16): Likewise.
3350 (_mm256_maskz_cvttph_epu16): Likewise.
3351 * config/i386/i386-builtin.def: Add new builtins.
3352 * config/i386/sse.md
3353 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name><round_saeonly_name>):
3355 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name>): Ditto.
3356 (*avx512fp16_fix<fixunssuffix>_trunc<mode>2_load<mask_name>): Ditto.
3357 (avx512fp16_fix<fixunssuffix>_truncv2di2<mask_name>): Ditto.
3358 (avx512fp16_fix<fixunssuffix>_truncv2di2_load<mask_name>): Ditto.
3360 2021-09-17 liuhongt <hongtao.liu@intel.com>
3362 * config/i386/avx512fp16intrin.h (_mm_cvtsh_i32): New intrinsic.
3363 (_mm_cvtsh_u32): Likewise.
3364 (_mm_cvt_roundsh_i32): Likewise.
3365 (_mm_cvt_roundsh_u32): Likewise.
3366 (_mm_cvtsh_i64): Likewise.
3367 (_mm_cvtsh_u64): Likewise.
3368 (_mm_cvt_roundsh_i64): Likewise.
3369 (_mm_cvt_roundsh_u64): Likewise.
3370 (_mm_cvti32_sh): Likewise.
3371 (_mm_cvtu32_sh): Likewise.
3372 (_mm_cvt_roundi32_sh): Likewise.
3373 (_mm_cvt_roundu32_sh): Likewise.
3374 (_mm_cvti64_sh): Likewise.
3375 (_mm_cvtu64_sh): Likewise.
3376 (_mm_cvt_roundi64_sh): Likewise.
3377 (_mm_cvt_roundu64_sh): Likewise.
3378 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3379 * config/i386/i386-builtin.def: Add corresponding new builtins.
3380 * config/i386/i386-expand.c (ix86_expand_round_builtin):
3381 Handle new builtin types.
3382 * config/i386/sse.md
3383 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix><round_name>):
3385 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix>_2): Likewise.
3386 (avx512fp16_vcvt<floatsuffix>si2sh<rex64namesuffix><round_name>): Likewise.
3388 2021-09-16 Bill Schmidt <wschmidt@linux.ibm.com>
3390 * config/rs6000/rs6000-c.c (rs6000-builtins.h): New include.
3391 (altivec_resolve_new_overloaded_builtin): New forward decl.
3392 (rs6000_new_builtin_type_compatible): New function.
3393 (altivec_resolve_overloaded_builtin): Call
3394 altivec_resolve_new_overloaded_builtin.
3395 (altivec_build_new_resolved_builtin): New function.
3396 (altivec_resolve_new_overloaded_builtin): Likewise.
3397 * config/rs6000/rs6000-call.c (rs6000_new_builtin_is_supported):
3399 * config/rs6000/rs6000-gen-builtins.c (write_decls): Remove _p from
3400 name of rs6000_new_builtin_is_supported.
3402 2021-09-16 Uroš Bizjak <ubizjak@gmail.com>
3404 * config/i386/i386-protos.h (ix86_decompose_address):
3405 Change return type to bool.
3406 * config/i386/i386.c (ix86_decompose_address): Ditto.
3408 2021-09-16 Tobias Burnus <tobias@codesourcery.com>
3411 * config/rs6000/t-rs6000 (build/rs6000-gen-builtins.o, build/rbtree.o):
3412 Added 'build/' to target, use build/%.o rule.
3413 (build/rs6000-gen-builtins$(build_exeext)): Add 'build/' and
3414 '$(build_exeext)' to target and 'build/' for the *.o files.
3415 (rs6000-builtins.c): Update for those changes; run rs6000-gen-builtins
3418 2021-09-16 Martin Jambor <mjambor@suse.cz>
3420 * cgraph.c (cgraph_node::dump): Do not check caller count sums if
3421 the body has been removed. Remove trailing whitespace.
3423 2021-09-16 Richard Biener <rguenther@suse.de>
3425 PR middle-end/102360
3426 * internal-fn.c (expand_DEFERRED_INIT): Make pattern-init
3427 of non-memory more robust.
3429 2021-09-16 Daniel Cederman <cederman@gaisler.com>
3431 * config/sparc/sparc-opts.h (enum sparc_processor_type): Add LEON5
3432 * config/sparc/sparc.c (struct processor_costs): Add LEON5 costs
3433 (leon5_adjust_cost): Increase cost of store with data dependency
3434 on ALU instruction and FPU anti-dependencies.
3435 (sparc_option_override): Add LEON5 costs
3436 (sparc_adjust_cost): Add LEON5 cost adjustments
3437 * config/sparc/sparc.h: Add LEON5
3438 * config/sparc/sparc.md: Include LEON5 scheduling information
3439 * config/sparc/sparc.opt: Add LEON5
3440 * doc/invoke.texi: Add LEON5
3441 * config/sparc/leon5.md: New file.
3443 2021-09-16 Daniel Cederman <cederman@gaisler.com>
3445 * config/sparc/sparc.md (stack_protect_set32): Add NOP to prevent
3446 sensitive sequence for B2BST errata workaround.
3448 2021-09-16 Daniel Cederman <cederman@gaisler.com>
3450 * config/sparc/sparc.c (sparc_do_work_around_errata): Do not begin
3451 functions with atomic instruction in the UT700 errata workaround.
3453 2021-09-16 Daniel Cederman <cederman@gaisler.com>
3455 * config/sparc/sparc.c (next_active_non_empty_insn): New function
3456 that returns next active non empty assembly instruction.
3457 (sparc_do_work_around_errata): Use new function.
3459 2021-09-16 Daniel Cederman <cederman@gaisler.com>
3461 * config/sparc/sparc.c (store_insn_p): Add predicate for store
3463 (load_insn_p): Add predicate for load attributes.
3464 (sparc_do_work_around_errata): Use new predicates.
3466 2021-09-16 Andreas Larsson <andreas@gaisler.com>
3468 * config/sparc/sparc.c (dump_target_flag_bits): Print bit names for
3471 2021-09-16 Martin Liska <mliska@suse.cz>
3473 * config/mips/netbsd.h: Fix typo in name of a macro.
3475 2021-09-16 liuhongt <hongtao.liu@intel.com>
3477 PR middle-end/102080
3478 * match.pd: Check mask type when doing cond_op related gimple
3480 * tree.c (is_truth_type_for): New function.
3481 * tree.h (is_truth_type_for): New declaration.
3483 2021-09-16 liuhongt <hongtao.liu@intel.com>
3485 * config/i386/avx512fp16intrin.h (_mm512_cvtepi32_ph): New
3487 (_mm512_mask_cvtepi32_ph): Likewise.
3488 (_mm512_maskz_cvtepi32_ph): Likewise.
3489 (_mm512_cvt_roundepi32_ph): Likewise.
3490 (_mm512_mask_cvt_roundepi32_ph): Likewise.
3491 (_mm512_maskz_cvt_roundepi32_ph): Likewise.
3492 (_mm512_cvtepu32_ph): Likewise.
3493 (_mm512_mask_cvtepu32_ph): Likewise.
3494 (_mm512_maskz_cvtepu32_ph): Likewise.
3495 (_mm512_cvt_roundepu32_ph): Likewise.
3496 (_mm512_mask_cvt_roundepu32_ph): Likewise.
3497 (_mm512_maskz_cvt_roundepu32_ph): Likewise.
3498 (_mm512_cvtepi64_ph): Likewise.
3499 (_mm512_mask_cvtepi64_ph): Likewise.
3500 (_mm512_maskz_cvtepi64_ph): Likewise.
3501 (_mm512_cvt_roundepi64_ph): Likewise.
3502 (_mm512_mask_cvt_roundepi64_ph): Likewise.
3503 (_mm512_maskz_cvt_roundepi64_ph): Likewise.
3504 (_mm512_cvtepu64_ph): Likewise.
3505 (_mm512_mask_cvtepu64_ph): Likewise.
3506 (_mm512_maskz_cvtepu64_ph): Likewise.
3507 (_mm512_cvt_roundepu64_ph): Likewise.
3508 (_mm512_mask_cvt_roundepu64_ph): Likewise.
3509 (_mm512_maskz_cvt_roundepu64_ph): Likewise.
3510 (_mm512_cvtepi16_ph): Likewise.
3511 (_mm512_mask_cvtepi16_ph): Likewise.
3512 (_mm512_maskz_cvtepi16_ph): Likewise.
3513 (_mm512_cvt_roundepi16_ph): Likewise.
3514 (_mm512_mask_cvt_roundepi16_ph): Likewise.
3515 (_mm512_maskz_cvt_roundepi16_ph): Likewise.
3516 (_mm512_cvtepu16_ph): Likewise.
3517 (_mm512_mask_cvtepu16_ph): Likewise.
3518 (_mm512_maskz_cvtepu16_ph): Likewise.
3519 (_mm512_cvt_roundepu16_ph): Likewise.
3520 (_mm512_mask_cvt_roundepu16_ph): Likewise.
3521 (_mm512_maskz_cvt_roundepu16_ph): Likewise.
3522 * config/i386/avx512fp16vlintrin.h (_mm_cvtepi32_ph): New
3524 (_mm_mask_cvtepi32_ph): Likewise.
3525 (_mm_maskz_cvtepi32_ph): Likewise.
3526 (_mm256_cvtepi32_ph): Likewise.
3527 (_mm256_mask_cvtepi32_ph): Likewise.
3528 (_mm256_maskz_cvtepi32_ph): Likewise.
3529 (_mm_cvtepu32_ph): Likewise.
3530 (_mm_mask_cvtepu32_ph): Likewise.
3531 (_mm_maskz_cvtepu32_ph): Likewise.
3532 (_mm256_cvtepu32_ph): Likewise.
3533 (_mm256_mask_cvtepu32_ph): Likewise.
3534 (_mm256_maskz_cvtepu32_ph): Likewise.
3535 (_mm_cvtepi64_ph): Likewise.
3536 (_mm_mask_cvtepi64_ph): Likewise.
3537 (_mm_maskz_cvtepi64_ph): Likewise.
3538 (_mm256_cvtepi64_ph): Likewise.
3539 (_mm256_mask_cvtepi64_ph): Likewise.
3540 (_mm256_maskz_cvtepi64_ph): Likewise.
3541 (_mm_cvtepu64_ph): Likewise.
3542 (_mm_mask_cvtepu64_ph): Likewise.
3543 (_mm_maskz_cvtepu64_ph): Likewise.
3544 (_mm256_cvtepu64_ph): Likewise.
3545 (_mm256_mask_cvtepu64_ph): Likewise.
3546 (_mm256_maskz_cvtepu64_ph): Likewise.
3547 (_mm_cvtepi16_ph): Likewise.
3548 (_mm_mask_cvtepi16_ph): Likewise.
3549 (_mm_maskz_cvtepi16_ph): Likewise.
3550 (_mm256_cvtepi16_ph): Likewise.
3551 (_mm256_mask_cvtepi16_ph): Likewise.
3552 (_mm256_maskz_cvtepi16_ph): Likewise.
3553 (_mm_cvtepu16_ph): Likewise.
3554 (_mm_mask_cvtepu16_ph): Likewise.
3555 (_mm_maskz_cvtepu16_ph): Likewise.
3556 (_mm256_cvtepu16_ph): Likewise.
3557 (_mm256_mask_cvtepu16_ph): Likewise.
3558 (_mm256_maskz_cvtepu16_ph): Likewise.
3559 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3560 * config/i386/i386-builtin.def: Add corresponding new builtins.
3561 * config/i386/i386-expand.c
3562 (ix86_expand_args_builtin): Handle new builtin types.
3563 (ix86_expand_round_builtin): Ditto.
3564 * config/i386/i386-modes.def: Declare V2HF and V6HF.
3565 * config/i386/sse.md (VI2H_AVX512VL): New.
3567 (sseintvecmode): Add HF vector modes.
3568 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode><mask_name><round_name>):
3570 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
3571 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
3572 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
3573 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
3574 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask_1): Ditto.
3575 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
3576 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
3577 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
3578 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
3579 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask_1): Ditto.
3580 * config/i386/subst.md (round_qq2phsuff): New subst_attr.
3582 2021-09-16 liuhongt <hongtao.liu@intel.com>
3584 * config/i386/avx512fp16intrin.h (_mm512_cvtph_epi32):
3586 (_mm512_mask_cvtph_epi32): Likewise.
3587 (_mm512_maskz_cvtph_epi32): Likewise.
3588 (_mm512_cvt_roundph_epi32): Likewise.
3589 (_mm512_mask_cvt_roundph_epi32): Likewise.
3590 (_mm512_maskz_cvt_roundph_epi32): Likewise.
3591 (_mm512_cvtph_epu32): Likewise.
3592 (_mm512_mask_cvtph_epu32): Likewise.
3593 (_mm512_maskz_cvtph_epu32): Likewise.
3594 (_mm512_cvt_roundph_epu32): Likewise.
3595 (_mm512_mask_cvt_roundph_epu32): Likewise.
3596 (_mm512_maskz_cvt_roundph_epu32): Likewise.
3597 (_mm512_cvtph_epi64): Likewise.
3598 (_mm512_mask_cvtph_epi64): Likewise.
3599 (_mm512_maskz_cvtph_epi64): Likewise.
3600 (_mm512_cvt_roundph_epi64): Likewise.
3601 (_mm512_mask_cvt_roundph_epi64): Likewise.
3602 (_mm512_maskz_cvt_roundph_epi64): Likewise.
3603 (_mm512_cvtph_epu64): Likewise.
3604 (_mm512_mask_cvtph_epu64): Likewise.
3605 (_mm512_maskz_cvtph_epu64): Likewise.
3606 (_mm512_cvt_roundph_epu64): Likewise.
3607 (_mm512_mask_cvt_roundph_epu64): Likewise.
3608 (_mm512_maskz_cvt_roundph_epu64): Likewise.
3609 (_mm512_cvtph_epi16): Likewise.
3610 (_mm512_mask_cvtph_epi16): Likewise.
3611 (_mm512_maskz_cvtph_epi16): Likewise.
3612 (_mm512_cvt_roundph_epi16): Likewise.
3613 (_mm512_mask_cvt_roundph_epi16): Likewise.
3614 (_mm512_maskz_cvt_roundph_epi16): Likewise.
3615 (_mm512_cvtph_epu16): Likewise.
3616 (_mm512_mask_cvtph_epu16): Likewise.
3617 (_mm512_maskz_cvtph_epu16): Likewise.
3618 (_mm512_cvt_roundph_epu16): Likewise.
3619 (_mm512_mask_cvt_roundph_epu16): Likewise.
3620 (_mm512_maskz_cvt_roundph_epu16): Likewise.
3621 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_epi32):
3623 (_mm_mask_cvtph_epi32): Likewise.
3624 (_mm_maskz_cvtph_epi32): Likewise.
3625 (_mm256_cvtph_epi32): Likewise.
3626 (_mm256_mask_cvtph_epi32): Likewise.
3627 (_mm256_maskz_cvtph_epi32): Likewise.
3628 (_mm_cvtph_epu32): Likewise.
3629 (_mm_mask_cvtph_epu32): Likewise.
3630 (_mm_maskz_cvtph_epu32): Likewise.
3631 (_mm256_cvtph_epu32): Likewise.
3632 (_mm256_mask_cvtph_epu32): Likewise.
3633 (_mm256_maskz_cvtph_epu32): Likewise.
3634 (_mm_cvtph_epi64): Likewise.
3635 (_mm_mask_cvtph_epi64): Likewise.
3636 (_mm_maskz_cvtph_epi64): Likewise.
3637 (_mm256_cvtph_epi64): Likewise.
3638 (_mm256_mask_cvtph_epi64): Likewise.
3639 (_mm256_maskz_cvtph_epi64): Likewise.
3640 (_mm_cvtph_epu64): Likewise.
3641 (_mm_mask_cvtph_epu64): Likewise.
3642 (_mm_maskz_cvtph_epu64): Likewise.
3643 (_mm256_cvtph_epu64): Likewise.
3644 (_mm256_mask_cvtph_epu64): Likewise.
3645 (_mm256_maskz_cvtph_epu64): Likewise.
3646 (_mm_cvtph_epi16): Likewise.
3647 (_mm_mask_cvtph_epi16): Likewise.
3648 (_mm_maskz_cvtph_epi16): Likewise.
3649 (_mm256_cvtph_epi16): Likewise.
3650 (_mm256_mask_cvtph_epi16): Likewise.
3651 (_mm256_maskz_cvtph_epi16): Likewise.
3652 (_mm_cvtph_epu16): Likewise.
3653 (_mm_mask_cvtph_epu16): Likewise.
3654 (_mm_maskz_cvtph_epu16): Likewise.
3655 (_mm256_cvtph_epu16): Likewise.
3656 (_mm256_mask_cvtph_epu16): Likewise.
3657 (_mm256_maskz_cvtph_epu16): Likewise.
3658 * config/i386/i386-builtin-types.def: Add new builtin types.
3659 * config/i386/i386-builtin.def: Add new builtins.
3660 * config/i386/i386-expand.c
3661 (ix86_expand_args_builtin): Handle new builtin types.
3662 (ix86_expand_round_builtin): Ditto.
3663 * config/i386/sse.md (sseintconvert): New.
3665 (UNSPEC_US_FIX_NOTRUNC): Ditto.
3666 (sseintconvertsignprefix): Ditto.
3667 (avx512fp16_vcvtph2<sseintconvertsignprefix><sseintconvert>_<mode><mask_name><round_name>):
3670 2021-09-16 liuhongt <hongtao.liu@intel.com>
3672 * config/i386/avx512fp16intrin.h: (_mm_cvtsi16_si128):
3674 (_mm_cvtsi128_si16): Likewise.
3675 (_mm_mask_load_sh): Likewise.
3676 (_mm_maskz_load_sh): Likewise.
3677 (_mm_mask_store_sh): Likewise.
3678 (_mm_move_sh): Likewise.
3679 (_mm_mask_move_sh): Likewise.
3680 (_mm_maskz_move_sh): Likewise.
3681 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3682 * config/i386/i386-builtin.def: Add corresponding new builtins.
3683 * config/i386/i386-expand.c
3684 (ix86_expand_special_args_builtin): Handle new builtin types.
3685 (ix86_expand_vector_init_one_nonzero): Adjust for FP16 target.
3686 * config/i386/sse.md (VI2F): New mode iterator.
3687 (vec_set<mode>_0): Use new mode iterator.
3688 (avx512f_mov<ssescalarmodelower>_mask): Adjust for HF vector mode.
3689 (avx512f_store<mode>_mask): Ditto.
3691 2021-09-16 Kewen Lin <linkw@linux.ibm.com>
3693 * config/rs6000/rs6000.opt (-mtoc-fusion): Remove.
3695 2021-09-15 David Edelsohn <dje.gcc@gmail.com>
3697 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
3698 Proceed if no symbol summary or the symbol alias flag is false.
3700 2021-09-15 Jakub Jelinek <jakub@redhat.com>
3704 * varasm.c (output_constructor_regular_field): Instead of assertion
3705 that array_size_for_constructor result is equal to size of
3706 TREE_TYPE (local->val) in bytes, assert that the type size is greater
3707 or equal to array_size_for_constructor result and use type size as
3710 2021-09-15 Martin Liska <mliska@suse.cz>
3713 * config/i386/vxworks.h: Use new macro TARGET_CPU_P.
3715 2021-09-15 Martin Liska <mliska@suse.cz>
3718 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
3719 Check that we have a symbol summary for a symbol.
3721 2021-09-15 Richard Biener <rguenther@suse.de>
3724 * config/rs6000/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
3725 to inherit from elfos.h
3727 2021-09-15 liuhongt <hongtao.liu@intel.com>
3730 * config/i386/i386-expand.c
3731 (ix86_expand_vector_init_interleave): Use puncklwd to pack 2
3733 (ix86_expand_vector_set): Use blendw instead of pinsrw.
3734 * config/i386/i386.c (ix86_can_change_mode_class): Adjust for
3735 AVX512FP16 which supports 16bit vector load.
3736 * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>):
3738 (avx512bw_interleave_high<mode><mask_name>): .. this, and
3739 extend to V32HFmode.
3740 (avx2_interleave_highv16hi<mask_name>): Rename to ..
3741 (avx2_interleave_high<mode><mask_name>): .. this, and extend
3743 (vec_interleave_highv8hi<mask_name>): Rename to ..
3744 (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode.
3745 (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>):
3747 (<mask_codefor>avx512bw_interleave_low<mode><mask_name>):
3748 this, and extend to V32HFmode.
3749 (avx2_interleave_lowv16hi<mask_name>): Rename to ..
3750 (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode.
3751 (vec_interleave_lowv8hi<mask_name>): Rename to ..
3752 (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode.
3753 (sse4_1_pblendw): Rename to ..
3754 (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode.
3755 (avx2_pblendph): New define_expand.
3756 (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use
3757 sseintmodesuffix instead of ssemodesuffix.
3758 (blendsuf): New mode attr.
3760 2021-09-15 Richard Biener <rguenther@suse.de>
3762 * tree-vectorizer.h (dr_misalignment): Move out of line.
3763 (dr_target_alignment): New.
3764 (DR_TARGET_ALIGNMENT): Wrap dr_target_alignment.
3765 (set_dr_target_alignment): New.
3766 (SET_DR_TARGET_ALIGNMENT): Wrap set_dr_target_alignment.
3767 * tree-vect-data-refs.c (dr_misalignment): Compute and
3768 return the group members misalignment.
3769 (vect_compute_data_ref_alignment): Use SET_DR_TARGET_ALIGNMENT.
3770 (vect_analyze_data_refs_alignment): Compute alignment only
3771 for the first element of a DR group.
3772 (vect_slp_analyze_node_alignment): Likewise.
3774 2021-09-15 Hongyu Wang <hongyu.wang@intel.com>
3776 * config/i386/avx512fp16intrin.h: Adjust all builtin calls.
3777 * config/i386/avx512fp16vlintrin.h: Likewise.
3778 * config/i386/i386-builtin.def: Adjust builtin name and
3779 enumeration to match AVX512F style.
3781 2021-09-15 Richard Biener <rguenther@suse.de>
3783 PR tree-optimization/102318
3784 * tree-vect-loop.c (vect_transform_cycle_phi): Revert
3785 previous change and do the mode conversion separately from
3786 the sign conversion.
3788 2021-09-15 Hongtao Liu <hongtao.liu@intel.com>
3789 Peter Cordes <peter@cordes.ca>
3792 * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
3793 (*vec_extract<mode><ssescalarmodelower>_valign): Output
3794 vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
3797 2021-09-15 Richard Biener <rguenther@suse.de>
3799 * config.gcc: Remove vax-*-openbsd* configuration.
3801 2021-09-15 Richard Biener <rguenther@suse.de>
3803 * config.gcc: Remove m68k-openbsd.
3805 2021-09-15 Max Filippov <jcmvbkbc@gmail.com>
3808 * config/xtensa/t-xtensa (TM_H): Add include/xtensa-config.h.
3810 2021-09-14 Peter Bergner <bergner@linux.ibm.com>
3812 * config/rs6000/mma.md (unspec): Delete UNSPEC_MMA_XXSETACCZ.
3813 (unspecv): Add UNSPECV_MMA_XXSETACCZ.
3814 (*mma_xxsetaccz): Delete.
3815 (mma_xxsetaccz): Change to define_insn. Remove operand 1.
3816 Use UNSPECV_MMA_XXSETACCZ. Update comment.
3817 * config/rs6000/rs6000.c (rs6000_rtx_costs): Use UNSPECV_MMA_XXSETACCZ.
3819 2021-09-14 Iain Sandoe <iain@sandoe.co.uk>
3821 * Makefile.in: Remove variables related to applying no-PIE
3822 to the exes on $build.
3823 * configure: Regenerate.
3824 * configure.ac: Remove configuration related to applying
3825 no-PIE to the exes on $build.
3827 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
3829 * config/arc/arc.md (doloop_end): Add missing mode.
3830 (loop_end): Likewise.
3832 2021-09-14 Jakub Jelinek <jakub@redhat.com>
3834 * gimplify.c (goa_stabilize_expr): Add depth argument, propagate
3835 it to recursive calls, for depth above 7 just gimplify or return.
3836 Perform a test even for MODIFY_EXPR, ADDR_EXPR, COMPOUND_EXPR with
3837 __builtin_clear_padding and TARGET_EXPR.
3838 (gimplify_omp_atomic): Adjust goa_stabilize_expr callers.
3840 2021-09-14 liuhongt <hongtao.liu@intel.com>
3842 * config/i386/avx512fp16intrin.h (_mm_fpclass_sh_mask):
3844 (_mm_mask_fpclass_sh_mask): Likewise.
3845 (_mm512_mask_fpclass_ph_mask): Likewise.
3846 (_mm512_fpclass_ph_mask): Likewise.
3847 (_mm_getexp_sh): Likewise.
3848 (_mm_mask_getexp_sh): Likewise.
3849 (_mm_maskz_getexp_sh): Likewise.
3850 (_mm512_getexp_ph): Likewise.
3851 (_mm512_mask_getexp_ph): Likewise.
3852 (_mm512_maskz_getexp_ph): Likewise.
3853 (_mm_getexp_round_sh): Likewise.
3854 (_mm_mask_getexp_round_sh): Likewise.
3855 (_mm_maskz_getexp_round_sh): Likewise.
3856 (_mm512_getexp_round_ph): Likewise.
3857 (_mm512_mask_getexp_round_ph): Likewise.
3858 (_mm512_maskz_getexp_round_ph): Likewise.
3859 (_mm_getmant_sh): Likewise.
3860 (_mm_mask_getmant_sh): Likewise.
3861 (_mm_maskz_getmant_sh): Likewise.
3862 (_mm512_getmant_ph): Likewise.
3863 (_mm512_mask_getmant_ph): Likewise.
3864 (_mm512_maskz_getmant_ph): Likewise.
3865 (_mm_getmant_round_sh): Likewise.
3866 (_mm_mask_getmant_round_sh): Likewise.
3867 (_mm_maskz_getmant_round_sh): Likewise.
3868 (_mm512_getmant_round_ph): Likewise.
3869 (_mm512_mask_getmant_round_ph): Likewise.
3870 (_mm512_maskz_getmant_round_ph): Likewise.
3871 * config/i386/avx512fp16vlintrin.h (_mm_mask_fpclass_ph_mask):
3873 (_mm_fpclass_ph_mask): Likewise.
3874 (_mm256_mask_fpclass_ph_mask): Likewise.
3875 (_mm256_fpclass_ph_mask): Likewise.
3876 (_mm256_getexp_ph): Likewise.
3877 (_mm256_mask_getexp_ph): Likewise.
3878 (_mm256_maskz_getexp_ph): Likewise.
3879 (_mm_getexp_ph): Likewise.
3880 (_mm_mask_getexp_ph): Likewise.
3881 (_mm_maskz_getexp_ph): Likewise.
3882 (_mm256_getmant_ph): Likewise.
3883 (_mm256_mask_getmant_ph): Likewise.
3884 (_mm256_maskz_getmant_ph): Likewise.
3885 (_mm_getmant_ph): Likewise.
3886 (_mm_mask_getmant_ph): Likewise.
3887 (_mm_maskz_getmant_ph): Likewise.
3888 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3889 * config/i386/i386-builtin.def: Add corresponding new builtins.
3890 * config/i386/i386-expand.c
3891 (ix86_expand_args_builtin): Handle new builtin types.
3892 (ix86_expand_round_builtin): Ditto.
3893 * config/i386/sse.md (vecmemsuffix): Add HF vector modes.
3894 (<avx512>_getexp<mode><mask_name><round_saeonly_name>): Adjust
3895 to support HF vector modes.
3896 (avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name):
3898 (avx512dq_fpclass<mode><mask_scalar_merge_name>): Ditto.
3899 (avx512dq_vmfpclass<mode><mask_scalar_merge_name>): Ditto.
3900 (<avx512>_getmant<mode><mask_name><round_saeonly_name>): Ditto.
3901 (avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3904 2021-09-14 liuhongt <hongtao.liu@intel.com>
3906 * config/i386/avx512fp16intrin.h (_mm512_reduce_ph):
3908 (_mm512_mask_reduce_ph): Likewise.
3909 (_mm512_maskz_reduce_ph): Likewise.
3910 (_mm512_reduce_round_ph): Likewise.
3911 (_mm512_mask_reduce_round_ph): Likewise.
3912 (_mm512_maskz_reduce_round_ph): Likewise.
3913 (_mm_reduce_sh): Likewise.
3914 (_mm_mask_reduce_sh): Likewise.
3915 (_mm_maskz_reduce_sh): Likewise.
3916 (_mm_reduce_round_sh): Likewise.
3917 (_mm_mask_reduce_round_sh): Likewise.
3918 (_mm_maskz_reduce_round_sh): Likewise.
3919 (_mm512_roundscale_ph): Likewise.
3920 (_mm512_mask_roundscale_ph): Likewise.
3921 (_mm512_maskz_roundscale_ph): Likewise.
3922 (_mm512_roundscale_round_ph): Likewise.
3923 (_mm512_mask_roundscale_round_ph): Likewise.
3924 (_mm512_maskz_roundscale_round_ph): Likewise.
3925 (_mm_roundscale_sh): Likewise.
3926 (_mm_mask_roundscale_sh): Likewise.
3927 (_mm_maskz_roundscale_sh): Likewise.
3928 (_mm_roundscale_round_sh): Likewise.
3929 (_mm_mask_roundscale_round_sh): Likewise.
3930 (_mm_maskz_roundscale_round_sh): Likewise.
3931 * config/i386/avx512fp16vlintrin.h: (_mm_reduce_ph):
3933 (_mm_mask_reduce_ph): Likewise.
3934 (_mm_maskz_reduce_ph): Likewise.
3935 (_mm256_reduce_ph): Likewise.
3936 (_mm256_mask_reduce_ph): Likewise.
3937 (_mm256_maskz_reduce_ph): Likewise.
3938 (_mm_roundscale_ph): Likewise.
3939 (_mm_mask_roundscale_ph): Likewise.
3940 (_mm_maskz_roundscale_ph): Likewise.
3941 (_mm256_roundscale_ph): Likewise.
3942 (_mm256_mask_roundscale_ph): Likewise.
3943 (_mm256_maskz_roundscale_ph): Likewise.
3944 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3945 * config/i386/i386-builtin.def: Add corresponding new builtins.
3946 * config/i386/i386-expand.c
3947 (ix86_expand_args_builtin): Handle new builtin types.
3948 (ix86_expand_round_builtin): Ditto.
3949 * config/i386/sse.md (<mask_codefor>reducep<mode><mask_name>):
3951 (<mask_codefor>reducep<mode><mask_name><round_saeonly_name>):
3952 ... this, and adjust for round operands.
3953 (reduces<mode><mask_scalar_name>): Likewise, with ...
3954 (reduces<mode><mask_scalar_name><round_saeonly_scalar_name):
3956 (<avx512>_rndscale<mode><mask_name><round_saeonly_name>):
3957 Adjust for HF vector modes.
3958 (avx512f_rndscale<mode><mask_scalar_name><round_saeonly_scalar_name>):
3960 (*avx512f_rndscale<mode><round_saeonly_name>): Ditto.
3962 2021-09-14 liuhongt <hongtao.liu@intel.com>
3964 * config/i386/avx512fp16intrin.h: (_mm512_rcp_ph):
3966 (_mm512_mask_rcp_ph): Likewise.
3967 (_mm512_maskz_rcp_ph): Likewise.
3968 (_mm_rcp_sh): Likewise.
3969 (_mm_mask_rcp_sh): Likewise.
3970 (_mm_maskz_rcp_sh): Likewise.
3971 (_mm512_scalef_ph): Likewise.
3972 (_mm512_mask_scalef_ph): Likewise.
3973 (_mm512_maskz_scalef_ph): Likewise.
3974 (_mm512_scalef_round_ph): Likewise.
3975 (_mm512_mask_scalef_round_ph): Likewise.
3976 (_mm512_maskz_scalef_round_ph): Likewise.
3977 (_mm_scalef_sh): Likewise.
3978 (_mm_mask_scalef_sh): Likewise.
3979 (_mm_maskz_scalef_sh): Likewise.
3980 (_mm_scalef_round_sh): Likewise.
3981 (_mm_mask_scalef_round_sh): Likewise.
3982 (_mm_maskz_scalef_round_sh): Likewise.
3983 * config/i386/avx512fp16vlintrin.h (_mm_rcp_ph):
3985 (_mm256_rcp_ph): Likewise.
3986 (_mm_mask_rcp_ph): Likewise.
3987 (_mm256_mask_rcp_ph): Likewise.
3988 (_mm_maskz_rcp_ph): Likewise.
3989 (_mm256_maskz_rcp_ph): Likewise.
3990 (_mm_scalef_ph): Likewise.
3991 (_mm256_scalef_ph): Likewise.
3992 (_mm_mask_scalef_ph): Likewise.
3993 (_mm256_mask_scalef_ph): Likewise.
3994 (_mm_maskz_scalef_ph): Likewise.
3995 (_mm256_maskz_scalef_ph): Likewise.
3996 * config/i386/i386-builtin.def: Add new builtins.
3997 * config/i386/sse.md (VFH_AVX512VL): New.
3998 (avx512fp16_rcp<mode>2<mask_name>): Ditto.
3999 (avx512fp16_vmrcpv8hf2<mask_scalar_name>): Ditto.
4000 (avx512f_vmscalef<mode><mask_scalar_name><round_scalar_name>):
4001 Adjust to support HF vector modes.
4002 (<avx512>_scalef<mode><mask_name><round_name>): Ditto.
4004 2021-09-14 liuhongt <hongtao.liu@intel.com>
4006 * config/i386/avx512fp16intrin.h: (_mm512_sqrt_ph):
4008 (_mm512_mask_sqrt_ph): Likewise.
4009 (_mm512_maskz_sqrt_ph): Likewise.
4010 (_mm512_sqrt_round_ph): Likewise.
4011 (_mm512_mask_sqrt_round_ph): Likewise.
4012 (_mm512_maskz_sqrt_round_ph): Likewise.
4013 (_mm512_rsqrt_ph): Likewise.
4014 (_mm512_mask_rsqrt_ph): Likewise.
4015 (_mm512_maskz_rsqrt_ph): Likewise.
4016 (_mm_rsqrt_sh): Likewise.
4017 (_mm_mask_rsqrt_sh): Likewise.
4018 (_mm_maskz_rsqrt_sh): Likewise.
4019 (_mm_sqrt_sh): Likewise.
4020 (_mm_mask_sqrt_sh): Likewise.
4021 (_mm_maskz_sqrt_sh): Likewise.
4022 (_mm_sqrt_round_sh): Likewise.
4023 (_mm_mask_sqrt_round_sh): Likewise.
4024 (_mm_maskz_sqrt_round_sh): Likewise.
4025 * config/i386/avx512fp16vlintrin.h (_mm_sqrt_ph): New intrinsic.
4026 (_mm256_sqrt_ph): Likewise.
4027 (_mm_mask_sqrt_ph): Likewise.
4028 (_mm256_mask_sqrt_ph): Likewise.
4029 (_mm_maskz_sqrt_ph): Likewise.
4030 (_mm256_maskz_sqrt_ph): Likewise.
4031 (_mm_rsqrt_ph): Likewise.
4032 (_mm256_rsqrt_ph): Likewise.
4033 (_mm_mask_rsqrt_ph): Likewise.
4034 (_mm256_mask_rsqrt_ph): Likewise.
4035 (_mm_maskz_rsqrt_ph): Likewise.
4036 (_mm256_maskz_rsqrt_ph): Likewise.
4037 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
4038 * config/i386/i386-builtin.def: Add corresponding new builtins.
4039 * config/i386/i386-expand.c
4040 (ix86_expand_args_builtin): Handle new builtins.
4041 (ix86_expand_round_builtin): Ditto.
4042 * config/i386/sse.md (VF_AVX512FP16VL): New.
4043 (sqrt<mode>2): Adjust for HF vector modes.
4044 (<sse>_sqrt<mode>2<mask_name><round_name>): Likewise.
4045 (<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
4047 (<sse>_rsqrt<mode>2<mask_name>): New.
4048 (avx512fp16_vmrsqrtv8hf2<mask_scalar_name>): Likewise.
4050 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
4053 * diagnostic-spec.c (warning_suppressed_at, copy_warning): Handle
4054 'RESERVED_LOCATION_P' locations.
4055 * warning-control.cc (get_nowarn_spec, suppress_warning)
4056 (copy_warning): Likewise.
4058 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
4060 * diagnostic-spec.h (typedef xint_hash_t): Use 'location_t' instead of...
4061 (typedef key_type_t): ... this. Remove.
4062 (nowarn_map): Document.
4063 * diagnostic-spec.c (nowarn_map): Likewise.
4064 * warning-control.cc (convert_to_key): Evolve functions into...
4065 (get_location): ... these. Adjust all users.
4067 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
4069 * warning-control.cc (copy_warning): Remove 'nowarn_map' setup.
4071 2021-09-13 Jason Merrill <jason@redhat.com>
4073 * params.opt: Add destructive-interference-size and
4074 constructive-interference-size.
4075 * doc/invoke.texi: Document them.
4076 * config/aarch64/aarch64.c (aarch64_override_options_internal):
4078 * config/arm/arm.c (arm_option_override): Set them.
4079 * config/i386/i386-options.c (ix86_option_override_internal):
4082 2021-09-13 Martin Liska <mliska@suse.cz>
4083 H.J. Lu <hjl.tools@gmail.com>
4086 * common/config/i386/cpuinfo.h (cpu_indicator_init): Add support
4087 for x86-64 micro levels for __builtin_cpu_supports.
4088 * common/config/i386/i386-cpuinfo.h (enum feature_priority):
4089 Add priorities for the micro-arch levels.
4090 (enum processor_features): Add new features.
4091 * common/config/i386/i386-isas.h: Add micro-arch features.
4092 * config/i386/i386-builtins.c (get_builtin_code_for_version):
4093 Support the micro-arch levels by callsing
4094 __builtin_cpu_supports.
4095 * doc/extend.texi: Document that the levels are support by
4096 __builtin_cpu_supports.
4098 2021-09-13 Andrew Pinski <apinski@marvell.com>
4101 * config/aarch64/aarch64-builtins.c (aarch64_fold_builtin_lane_check):
4103 (aarch64_general_fold_builtin): Handle AARCH64_SIMD_BUILTIN_LANE_CHECK.
4104 (aarch64_general_gimple_fold_builtin): Likewise.
4106 2021-09-13 Andrew Pinski <apinski@marvell.com>
4108 * config.gcc: Add m32r-*-linux* and m32rle-*-linux*
4109 to the Unsupported targets list.
4110 Remove support for m32r-*-linux* and m32rle-*-linux*.
4111 * config/m32r/linux.h: Removed.
4112 * config/m32r/t-linux: Removed.
4114 2021-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4117 * config/aarch64/aarch64.c (aarch64_classify_address): Don't allow
4118 register index for SVE predicate modes.
4120 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
4122 * tree-ssa-threadbackward.c
4123 (back_threader_profitability::profitable_path_p): Remove FSM
4125 (back_threader_registry::register_path): Same.
4126 * tree-ssa-threadedge.c
4127 (jump_threader::simplify_control_stmt_condition): Same.
4128 * tree-ssa-threadupdate.c (jt_path_registry::jt_path_registry):
4129 Add backedge_threads argument.
4130 (fwd_jt_path_registry::fwd_jt_path_registry): Pass
4131 backedge_threads argument.
4132 (back_jt_path_registry::back_jt_path_registry): Same.
4133 (dump_jump_thread_path): Adjust for FSM removal.
4134 (back_jt_path_registry::rewire_first_differing_edge): Same.
4135 (back_jt_path_registry::adjust_paths_after_duplication): Same.
4136 (back_jt_path_registry::update_cfg): Same.
4137 (jt_path_registry::register_jump_thread): Same.
4138 * tree-ssa-threadupdate.h (enum jump_thread_edge_type): Remove
4140 (class back_jt_path_registry): Add backedge_threads to
4143 2021-09-13 Martin Liska <mliska@suse.cz>
4146 * asan.h (sanitize_coverage_p): Handle when fn == NULL.
4148 2021-09-13 H.J. Lu <hjl.tools@gmail.com>
4151 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): New.
4152 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
4153 (MOVE_MAX): Check TARGET_AVX256_MOVE_BY_PIECES and
4154 TARGET_AVX256_STORE_BY_PIECES instead of
4155 TARGET_AVX256_SPLIT_UNALIGNED_LOAD and
4156 TARGET_AVX256_SPLIT_UNALIGNED_STORE.
4157 (STORE_MAX_PIECES): Check TARGET_AVX256_STORE_BY_PIECES instead
4158 of TARGET_AVX256_SPLIT_UNALIGNED_STORE.
4159 * config/i386/x86-tune.def (X86_TUNE_AVX256_MOVE_BY_PIECES): New.
4160 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
4162 2021-09-13 liuhongt <hongtao.liu@intel.com>
4165 * expmed.c (extract_bit_field_using_extv): Use
4166 gen_lowpart_if_possible instead of gen_lowpart to avoid ICE.
4168 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
4170 * Makefile.in (OBJS): Add value-pointer-equiv.o.
4171 * gimple-ssa-evrp.c (class ssa_equiv_stack): Move to
4172 value-pointer-equiv.*.
4173 (ssa_equiv_stack::ssa_equiv_stack): Same.
4174 (ssa_equiv_stack::enter): Same.
4175 (ssa_equiv_stack::leave): Same.
4176 (ssa_equiv_stack::push_replacement): Same.
4177 (ssa_equiv_stack::get_replacement): Same.
4178 (is_pointer_ssa): Same.
4179 (class pointer_equiv_analyzer): Same.
4180 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
4181 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Same.
4182 (pointer_equiv_analyzer::set_global_equiv): Same.
4183 (pointer_equiv_analyzer::set_cond_equiv): Same.
4184 (pointer_equiv_analyzer::get_equiv): Same.
4185 (pointer_equiv_analyzer::enter): Same.
4186 (pointer_equiv_analyzer::leave): Same.
4187 (pointer_equiv_analyzer::get_equiv_expr): Same.
4188 (pta_valueize): Same.
4189 (pointer_equiv_analyzer::visit_stmt): Same.
4190 (pointer_equiv_analyzer::visit_edge): Same.
4191 (hybrid_folder::value_of_expr): Same.
4192 (hybrid_folder::value_on_edge): Same.
4193 * value-pointer-equiv.cc: New file.
4194 * value-pointer-equiv.h: New file.
4196 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
4199 * gimple-fold.c (gimple_fold_builtin_memory_op): Allow folding
4200 memcpy if the size is not more than MOVE_MAX * MOVE_RATIO.
4202 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
4205 * config/arm/arm.md (movmisaligndi): New define_expand.
4206 * config/arm/vec-common.md (movmisalign<mode>): Iterate over VDQ mode.
4208 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
4211 * emit-rtl.c (gen_highpart): Use adjust_address to handle
4212 MEM rather than calling simplify_gen_subreg.
4214 2021-09-13 Jan-Benedict Glaw <jbglaw@ług-owl.de>
4216 * config/alpha/vms.h (INIT_CUMULATIVE_ARGS): Wrap multi-statment
4217 define into a block.
4219 2021-09-13 Richard Biener <rguenther@suse.de>
4221 * config/darwin.h (DARWIN_PREFER_DWARF): Do not define.
4222 * config/i386/darwin.h (PREFERRED_DEBUGGING_TYPE): Do not
4223 change based on DARWIN_PREFER_DWARF not being defined.
4225 2021-09-13 Richard Biener <rguenther@suse.de>
4227 * config/i386/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
4228 to inherit from elfos.h
4230 2021-09-13 Richard Biener <rguenther@suse.de>
4232 * config.gcc: Add cr16-*-* to the list of obsoleted targets.
4234 2021-09-13 Richard Biener <rguenther@suse.de>
4236 * config/avr/elf.h (PREFERRED_DEBUGGING_TYPE): Remove
4237 override, pick up DWARF2_DEBUG define from elfos.h
4239 2021-09-13 Richard Biener <rguenther@suse.de>
4241 * config/rx/rx.h (PREFERRED_DEBUGGING_TYPE): Always define to
4244 2021-09-13 Richard Biener <rguenther@suse.de>
4246 * config/alpha/vms.h (PREFERRED_DEBUGGING_TYPE): Define to
4249 2021-09-13 Richard Biener <rguenther@suse.de>
4251 * config/i386/cygming.h: Always default to DWARF2 debugging.
4252 Do not define DBX_DEBUGGING_INFO, that's done via dbxcoff.h
4254 * doc/install.texi: Document binutils 2.16 as minimum
4255 requirement for mingw.
4257 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
4259 * config/rs6000/rs6000.c (struct rs6000_cost_data): New members
4260 nstmts, nloads and extra_ctor_cost.
4261 (rs6000_density_test): Add load density related heuristics. Do
4262 extra costing on vector construction statements if need.
4263 (rs6000_init_cost): Init new members.
4264 (rs6000_update_target_cost_per_stmt): New function.
4265 (rs6000_add_stmt_cost): Factor vect_nonmem hunk out to function
4266 rs6000_update_target_cost_per_stmt and call it.
4268 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
4270 * config/rs6000/rs6000.c (struct rs6000_cost_data): Remove typedef.
4271 (rs6000_init_cost): Adjust.
4273 2021-09-13 liuhongt <hongtao.liu@intel.com>
4275 * config/i386/i386.md: (UNSPEC_COPYSIGN): Remove.
4276 (UNSPEC_XORSIGN): Ditto.
4278 2021-09-12 Roger Sayle <roger@nextmovesoftware.com>
4280 * expr.c (convert_move): Preserve SUBREG_PROMOTED_VAR_P when
4281 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
4284 2021-09-11 Aldy Hernandez <aldyh@redhat.com>
4286 * tree-ssa-threadbackward.c (class back_threader_registry): Use
4287 back_jt_path_registry.
4288 * tree-ssa-threadedge.c (jump_threader::jump_threader): Use
4289 fwd_jt_path_registry.
4290 * tree-ssa-threadedge.h (class jump_threader): Same..
4291 * tree-ssa-threadupdate.c
4292 (jump_thread_path_registry::jump_thread_path_registry): Rename...
4293 (jt_path_registry::jt_path_registry): ...to this.
4294 (jump_thread_path_registry::~jump_thread_path_registry): Rename...
4295 (jt_path_registry::~jt_path_registry): ...this.
4296 (fwd_jt_path_registry::fwd_jt_path_registry): New.
4297 (fwd_jt_path_registry::~fwd_jt_path_registry): New.
4298 (jump_thread_path_registry::allocate_thread_edge): Rename...
4299 (jt_path_registry::allocate_thread_edge): ...to this.
4300 (jump_thread_path_registry::allocate_thread_path): Rename...
4301 (jt_path_registry::allocate_thread_path): ...to this.
4302 (jump_thread_path_registry::lookup_redirection_data): Rename...
4303 (fwd_jt_path_registry::lookup_redirection_data): ...to this.
4304 (jump_thread_path_registry::thread_block_1): Rename...
4305 (fwd_jt_path_registry::thread_block_1): ...to this.
4306 (jump_thread_path_registry::thread_block): Rename...
4307 (fwd_jt_path_registry::thread_block): ...to this.
4308 (jt_path_registry::thread_through_loop_header): Rename...
4309 (fwd_jt_path_registry::thread_through_loop_header): ...to this.
4310 (jump_thread_path_registry::mark_threaded_blocks): Rename...
4311 (fwd_jt_path_registry::mark_threaded_blocks): ...to this.
4312 (jump_thread_path_registry::debug_path): Rename...
4313 (jt_path_registry::debug_path): ...to this.
4314 (jump_thread_path_registry::dump): Rename...
4315 (jt_path_registry::debug): ...to this.
4316 (jump_thread_path_registry::rewire_first_differing_edge): Rename...
4317 (back_jt_path_registry::rewire_first_differing_edge): ...to this.
4318 (jump_thread_path_registry::adjust_paths_after_duplication): Rename...
4319 (back_jt_path_registry::adjust_paths_after_duplication): ...to this.
4320 (jump_thread_path_registry::duplicate_thread_path): Rename...
4321 (back_jt_path_registry::duplicate_thread_path): ...to this. Also,
4322 drop ill-formed candidates.
4323 (jump_thread_path_registry::remove_jump_threads_including): Rename...
4324 (fwd_jt_path_registry::remove_jump_threads_including): ...to this.
4325 (jt_path_registry::thread_through_all_blocks): New.
4326 (back_jt_path_registry::update_cfg): New.
4327 (fwd_jt_path_registry::update_cfg): New.
4328 (jump_thread_path_registry::register_jump_thread): Rename...
4329 (jt_path_registry::register_jump_thread): ...to this.
4330 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
4332 (class jt_path_registry): ...here.
4333 (class fwd_jt_path_registry): New.
4334 (class back_jt_path_registry): New.
4336 2021-09-10 liuhongt <hongtao.liu@intel.com>
4339 2021-09-01 liuhongt <hongtao.liu@intel.com>
4341 * emit-rtl.c (validate_subreg): Get rid of all float-int
4344 2021-09-10 Jakub Jelinek <jakub@redhat.com>
4346 * tree-core.h (enum omp_memory_order): Add OMP_MEMORY_ORDER_MASK,
4347 OMP_FAIL_MEMORY_ORDER_UNSPECIFIED, OMP_FAIL_MEMORY_ORDER_RELAXED,
4348 OMP_FAIL_MEMORY_ORDER_ACQUIRE, OMP_FAIL_MEMORY_ORDER_RELEASE,
4349 OMP_FAIL_MEMORY_ORDER_ACQ_REL, OMP_FAIL_MEMORY_ORDER_SEQ_CST and
4350 OMP_FAIL_MEMORY_ORDER_MASK enumerators.
4351 (OMP_FAIL_MEMORY_ORDER_SHIFT): Define.
4352 * gimple-pretty-print.c (dump_gimple_omp_atomic_load,
4353 dump_gimple_omp_atomic_store): Print [weak] for weak atomic
4355 * gimple.h (enum gf_mask): Change GF_OMP_ATOMIC_MEMORY_ORDER
4356 to 6-bit mask, adjust GF_OMP_ATOMIC_NEED_VALUE value and add
4358 (gimple_omp_atomic_weak_p, gimple_omp_atomic_set_weak): New inline
4360 * tree.h (OMP_ATOMIC_WEAK): Define.
4361 * tree-pretty-print.c (dump_omp_atomic_memory_order): Adjust for
4362 fail memory order being encoded in the same enum and also print
4363 fail clause if present.
4364 (dump_generic_node): Print weak clause if OMP_ATOMIC_WEAK.
4365 * gimplify.c (goa_stabilize_expr): Add target_expr and rhs arguments,
4366 handle pre_p == NULL case as a test mode that only returns value
4367 but doesn't change gimplify nor change anything otherwise, adjust
4368 recursive calls, add MODIFY_EXPR, ADDR_EXPR, COND_EXPR, TARGET_EXPR
4369 and CALL_EXPR handling, adjust COMPOUND_EXPR handling for
4370 __builtin_clear_padding calls, for !rhs gimplify as lvalue rather
4372 (gimplify_omp_atomic): Adjust goa_stabilize_expr caller. Handle
4373 COND_EXPR rhs. Set weak flag on gimple load/store for
4375 * omp-expand.c (omp_memory_order_to_fail_memmodel): New function.
4376 (omp_memory_order_to_memmodel): Adjust for fail clause encoded
4378 (expand_omp_atomic_cas): New function.
4379 (expand_omp_atomic_pipeline): Use omp_memory_order_to_fail_memmodel
4381 (expand_omp_atomic): Attempt to optimize atomic compare and exchange
4382 using expand_omp_atomic_cas.
4384 2021-09-10 Aldy Hernandez <aldyh@redhat.com>
4385 Michael Matz <matz@suse.de>
4387 * tree-pass.h (PROP_loop_opts_done): New.
4388 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
4389 Intersect with global range.
4390 * tree-ssa-loop.c (tree_ssa_loop_done): Set PROP_loop_opts_done.
4391 * tree-ssa-threadbackward.c
4392 (back_threader_profitability::profitable_path_p): Disable
4393 threading through latches until after loop optimizations have run.
4395 2021-09-10 David Faust <david.faust@oracle.com>
4397 * doc/invoke.texi: Document BPF -mcpu, -mjmpext, -mjmp32 and -malu32
4400 2021-09-10 David Faust <david.faust@oracle.com>
4402 * config/bpf/bpf-opts.h (bpf_isa_version): New enum.
4403 * config/bpf/bpf-protos.h (bpf_expand_cbranch): New.
4404 * config/bpf/bpf.c (bpf_option_override): Handle -mcpu option.
4405 (bpf_expand_cbranch): New function.
4406 * config/bpf/bpf.md (AM mode iterator): Conditionalize support for SI
4408 (zero_extendsidi2): Only use mov32 instruction if it is available.
4409 (SIM mode iterator): Conditionalize support for SI mode.
4410 (JM mode iterator): New.
4411 (cbranchdi4): Update name, use new JM iterator. Use bpf_expand_cbranch.
4412 (*branch_on_di): Update name, use new JM iterator.
4413 * config/bpf/bpf.opt: (mjmpext): New option.
4417 (bpf_isa): New enum.
4419 2021-09-10 David Faust <david.faust@oracle.com>
4421 * config/bpf/bpf.md (zero_extendhidi2): Add new output template
4422 for register-to-register extensions.
4423 (zero_extendqidi2): Likewise.
4425 2021-09-10 Richard Biener <rguenther@suse.de>
4427 PR middle-end/102273
4428 * internal-fn.c (expand_DEFERRED_INIT): Always expand non-SSA vars.
4430 2021-09-10 Richard Biener <rguenther@suse.de>
4432 PR middle-end/102269
4433 * gimplify.c (is_var_need_auto_init): Empty types do not need
4436 2021-09-10 Richard Biener <rguenther@suse.de>
4438 * configure.ac (--with-stabs): Remove.
4439 * configure: Regenerate.
4440 * doc/install.texi: Remove --with-stabs documentation.
4442 2021-09-10 liuhongt <hongtao.liu@intel.com>
4444 * config/i386/avx512fp16intrin.h: (_mm512_cmp_ph_mask):
4446 (_mm512_mask_cmp_ph_mask): Likewise.
4447 (_mm512_cmp_round_ph_mask): Likewise.
4448 (_mm512_mask_cmp_round_ph_mask): Likewise.
4449 (_mm_cmp_sh_mask): Likewise.
4450 (_mm_mask_cmp_sh_mask): Likewise.
4451 (_mm_cmp_round_sh_mask): Likewise.
4452 (_mm_mask_cmp_round_sh_mask): Likewise.
4453 (_mm_comieq_sh): Likewise.
4454 (_mm_comilt_sh): Likewise.
4455 (_mm_comile_sh): Likewise.
4456 (_mm_comigt_sh): Likewise.
4457 (_mm_comige_sh): Likewise.
4458 (_mm_comineq_sh): Likewise.
4459 (_mm_ucomieq_sh): Likewise.
4460 (_mm_ucomilt_sh): Likewise.
4461 (_mm_ucomile_sh): Likewise.
4462 (_mm_ucomigt_sh): Likewise.
4463 (_mm_ucomige_sh): Likewise.
4464 (_mm_ucomineq_sh): Likewise.
4465 (_mm_comi_round_sh): Likewise.
4466 (_mm_comi_sh): Likewise.
4467 * config/i386/avx512fp16vlintrin.h (_mm_cmp_ph_mask): New intrinsic.
4468 (_mm_mask_cmp_ph_mask): Likewise.
4469 (_mm256_cmp_ph_mask): Likewise.
4470 (_mm256_mask_cmp_ph_mask): Likewise.
4471 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
4472 * config/i386/i386-builtin.def: Add corresponding new builtins.
4473 * config/i386/i386-expand.c
4474 (ix86_expand_args_builtin): Handle new builtin types.
4475 (ix86_expand_round_builtin): Ditto.
4476 * config/i386/i386.md (ssevecmode): Add HF mode.
4477 (MODEFH): New mode iterator.
4478 * config/i386/sse.md
4479 (V48H_AVX512VL): New mode iterator to support HF vector modes.
4480 Ajdust corresponding description.
4481 (ssecmpintprefix): New.
4482 (VI12_AVX512VL): Adjust to support HF vector modes.
4483 (cmp_imm_predicate): Likewise.
4484 (<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
4486 (avx512f_vmcmp<mode>3<round_saeonly_name>): Likewise.
4487 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Likewise.
4488 (<sse>_<unord>comi<round_saeonly_name>): Likewise.
4490 2021-09-10 liuhongt <hongtao.liu@intel.com>
4492 * config/i386/avx512fp16intrin.h: (_mm512_max_ph): New intrinsic.
4493 (_mm512_mask_max_ph): Likewise.
4494 (_mm512_maskz_max_ph): Likewise.
4495 (_mm512_min_ph): Likewise.
4496 (_mm512_mask_min_ph): Likewise.
4497 (_mm512_maskz_min_ph): Likewise.
4498 (_mm512_max_round_ph): Likewise.
4499 (_mm512_mask_max_round_ph): Likewise.
4500 (_mm512_maskz_max_round_ph): Likewise.
4501 (_mm512_min_round_ph): Likewise.
4502 (_mm512_mask_min_round_ph): Likewise.
4503 (_mm512_maskz_min_round_ph): Likewise.
4504 (_mm_max_sh): Likewise.
4505 (_mm_mask_max_sh): Likewise.
4506 (_mm_maskz_max_sh): Likewise.
4507 (_mm_min_sh): Likewise.
4508 (_mm_mask_min_sh): Likewise.
4509 (_mm_maskz_min_sh): Likewise.
4510 (_mm_max_round_sh): Likewise.
4511 (_mm_mask_max_round_sh): Likewise.
4512 (_mm_maskz_max_round_sh): Likewise.
4513 (_mm_min_round_sh): Likewise.
4514 (_mm_mask_min_round_sh): Likewise.
4515 (_mm_maskz_min_round_sh): Likewise.
4516 * config/i386/avx512fp16vlintrin.h (_mm_max_ph): New intrinsic.
4517 (_mm256_max_ph): Likewise.
4518 (_mm_mask_max_ph): Likewise.
4519 (_mm256_mask_max_ph): Likewise.
4520 (_mm_maskz_max_ph): Likewise.
4521 (_mm256_maskz_max_ph): Likewise.
4522 (_mm_min_ph): Likewise.
4523 (_mm256_min_ph): Likewise.
4524 (_mm_mask_min_ph): Likewise.
4525 (_mm256_mask_min_ph): Likewise.
4526 (_mm_maskz_min_ph): Likewise.
4527 (_mm256_maskz_min_ph): Likewise.
4528 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
4529 * config/i386/i386-builtin.def: Add corresponding new builtins.
4530 * config/i386/i386-expand.c
4531 (ix86_expand_args_builtin): Handle new builtin types.
4532 * config/i386/sse.md
4533 (<code><mode>3<mask_name><round_saeonly_name>): Adjust to
4534 support HF vector modes.
4535 (*<code><mode>3<mask_name><round_saeonly_name>): Likewise.
4536 (ieee_<ieee_maxmin><mode>3<mask_name><round_saeonly_name>):
4538 (<sse>_vm<code><mode>3<mask_scalar_name><round_saeonly_scalar_name>):
4540 * config/i386/subst.md (round_saeonly_mode512bit_condition):
4541 Adjust for HF vector modes.
4543 2021-09-10 Liu, Hongtao <hongtao.liu@intel.com>
4545 * config/i386/avx512fp16intrin.h (_mm_add_sh): New intrinsic.
4546 (_mm_mask_add_sh): Likewise.
4547 (_mm_maskz_add_sh): Likewise.
4548 (_mm_sub_sh): Likewise.
4549 (_mm_mask_sub_sh): Likewise.
4550 (_mm_maskz_sub_sh): Likewise.
4551 (_mm_mul_sh): Likewise.
4552 (_mm_mask_mul_sh): Likewise.
4553 (_mm_maskz_mul_sh): Likewise.
4554 (_mm_div_sh): Likewise.
4555 (_mm_mask_div_sh): Likewise.
4556 (_mm_maskz_div_sh): Likewise.
4557 (_mm_add_round_sh): Likewise.
4558 (_mm_mask_add_round_sh): Likewise.
4559 (_mm_maskz_add_round_sh): Likewise.
4560 (_mm_sub_round_sh): Likewise.
4561 (_mm_mask_sub_round_sh): Likewise.
4562 (_mm_maskz_sub_round_sh): Likewise.
4563 (_mm_mul_round_sh): Likewise.
4564 (_mm_mask_mul_round_sh): Likewise.
4565 (_mm_maskz_mul_round_sh): Likewise.
4566 (_mm_div_round_sh): Likewise.
4567 (_mm_mask_div_round_sh): Likewise.
4568 (_mm_maskz_div_round_sh): Likewise.
4569 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
4570 * config/i386/i386-builtin.def: Add corresponding new builtins.
4571 * config/i386/i386-expand.c
4572 (ix86_expand_round_builtin): Handle new builtins.
4573 * config/i386/sse.md (VF_128): Change description.
4574 (<sse>_vm<plusminus_insn><mode>3<mask_scalar_name><round_scalar_name>):
4575 Adjust to support HF vector modes.
4576 (<sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name><round_scalar_name>):
4579 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
4581 * config/i386/i386-expand.c
4582 (ix86_avx256_split_vector_move_misalign): Handle V16HF mode.
4583 * config/i386/i386.c
4584 (ix86_preferred_simd_mode): Handle HF mode.
4585 * config/i386/sse.md (V_256H): New mode iterator.
4586 (avx_vextractf128<mode>): Use it.
4587 (VEC_INIT_MODE): Align vector HFmode condition to vector
4588 HImodes since there're no real HF instruction used.
4589 (VEC_INIT_HALF_MODE): Ditto.
4591 (VIHF_AVX512BW): Ditto.
4592 (*vec_extracthf): Ditto.
4593 (VEC_EXTRACT_MODE): Ditto.
4595 2021-09-10 Richard Biener <rguenther@suse.de>
4598 * config/dbx.h: Remove.
4599 * config/dbxcoff.h: Do not define PREFERRED_DEBUGGING_TYPE.
4600 * config/lynx.h: Likewise.
4602 2021-09-10 liuhongt <hongtao.liu@intel.com>
4604 * config/i386/i386-expand.c (ix86_expand_copysign): Expand
4605 right into ANDNOT + AND + IOR, using paradoxical subregs.
4606 (ix86_split_copysign_const): Remove.
4607 (ix86_split_copysign_var): Ditto.
4608 * config/i386/i386-protos.h (ix86_split_copysign_const): Dotto.
4609 (ix86_split_copysign_var): Ditto.
4610 * config/i386/i386.md (@copysign<mode>3_const): Ditto.
4611 (@copysign<mode>3_var): Ditto.
4613 2021-09-09 qing zhao <qing.zhao@oracle.com>
4615 * builtins.c (expand_builtin_memset): Make external visible.
4616 * builtins.h (expand_builtin_memset): Declare extern.
4617 * common.opt (ftrivial-auto-var-init=): New option.
4618 * doc/extend.texi: Document the uninitialized attribute.
4619 * doc/invoke.texi: Document -ftrivial-auto-var-init.
4620 * flag-types.h (enum auto_init_type): New enumerated type
4622 * gimple-fold.c (clear_padding_type): Add one new parameter.
4623 (clear_padding_union): Likewise.
4624 (clear_padding_emit_loop): Likewise.
4625 (clear_type_padding_in_mask): Likewise.
4626 (gimple_fold_builtin_clear_padding): Handle this new parameter.
4627 * gimplify.c (gimple_add_init_for_auto_var): New function.
4628 (gimple_add_padding_init_for_auto_var): New function.
4629 (is_var_need_auto_init): New function.
4630 (gimplify_decl_expr): Add initialization to automatic variables per
4632 (gimplify_call_expr): Add one new parameter for call to
4633 __builtin_clear_padding.
4634 (gimplify_init_constructor): Add padding initialization in the end.
4635 * internal-fn.c (INIT_PATTERN_VALUE): New macro.
4636 (expand_DEFERRED_INIT): New function.
4637 * internal-fn.def (DEFERRED_INIT): New internal function.
4638 * tree-cfg.c (verify_gimple_call): Verify calls to .DEFERRED_INIT.
4639 * tree-sra.c (generate_subtree_deferred_init): New function.
4640 (scan_function): Avoid setting cannot_scalarize_away_bitmap for
4641 calls to .DEFERRED_INIT.
4642 (sra_modify_deferred_init): New function.
4643 (sra_modify_function_body): Handle calls to DEFERRED_INIT specially.
4644 * tree-ssa-structalias.c (find_func_aliases_for_call): Likewise.
4645 * tree-ssa-uninit.c (warn_uninit): Handle calls to DEFERRED_INIT
4647 (check_defs): Likewise.
4648 (warn_uninitialized_vars): Likewise.
4649 * tree-ssa.c (ssa_undefined_value_p): Likewise.
4650 * tree.c (build_common_builtin_nodes): Build tree node for
4651 BUILT_IN_CLEAR_PADDING when needed.
4653 2021-09-09 Richard Biener <rguenther@suse.de>
4655 * tree-ssa-loop-im.c (fill_always_executed_in_1): Walk
4658 2021-09-09 Richard Biener <rguenther@suse.de>
4660 * tree-ssa-loop-im.c (fill_always_executed_in_1): Integrate
4661 DOM walk from get_loop_body_in_dom_order using a worklist
4664 2021-09-09 liuhongt <hongtao.liu@intel.com>
4666 * config.gcc: Add avx512fp16vlintrin.h.
4667 * config/i386/avx512fp16intrin.h: (_mm512_add_ph): New intrinsic.
4668 (_mm512_mask_add_ph): Likewise.
4669 (_mm512_maskz_add_ph): Likewise.
4670 (_mm512_sub_ph): Likewise.
4671 (_mm512_mask_sub_ph): Likewise.
4672 (_mm512_maskz_sub_ph): Likewise.
4673 (_mm512_mul_ph): Likewise.
4674 (_mm512_mask_mul_ph): Likewise.
4675 (_mm512_maskz_mul_ph): Likewise.
4676 (_mm512_div_ph): Likewise.
4677 (_mm512_mask_div_ph): Likewise.
4678 (_mm512_maskz_div_ph): Likewise.
4679 (_mm512_add_round_ph): Likewise.
4680 (_mm512_mask_add_round_ph): Likewise.
4681 (_mm512_maskz_add_round_ph): Likewise.
4682 (_mm512_sub_round_ph): Likewise.
4683 (_mm512_mask_sub_round_ph): Likewise.
4684 (_mm512_maskz_sub_round_ph): Likewise.
4685 (_mm512_mul_round_ph): Likewise.
4686 (_mm512_mask_mul_round_ph): Likewise.
4687 (_mm512_maskz_mul_round_ph): Likewise.
4688 (_mm512_div_round_ph): Likewise.
4689 (_mm512_mask_div_round_ph): Likewise.
4690 (_mm512_maskz_div_round_ph): Likewise.
4691 * config/i386/avx512fp16vlintrin.h: New header.
4692 * config/i386/i386-builtin-types.def (V16HF, V8HF, V32HF):
4693 Add new builtin types.
4694 * config/i386/i386-builtin.def: Add corresponding builtins.
4695 * config/i386/i386-expand.c
4696 (ix86_expand_args_builtin): Handle new builtin types.
4697 (ix86_expand_round_builtin): Likewise.
4698 * config/i386/immintrin.h: Include avx512fp16vlintrin.h
4699 * config/i386/sse.md (VFH): New mode_iterator.
4701 (avx512fmaskmode): Add HF vector modes.
4702 (avx512fmaskhalfmode): Likewise.
4703 (<plusminus_insn><mode>3<mask_name><round_name>): Adjust to for
4705 (*<plusminus_insn><mode>3<mask_name><round_name>): Likewise.
4706 (mul<mode>3<mask_name><round_name>): Likewise.
4707 (*mul<mode>3<mask_name><round_name>): Likewise.
4708 (div<mode>3): Likewise.
4709 (<sse>_div<mode>3<mask_name><round_name>): Likewise.
4710 * config/i386/subst.md (SUBST_V): Add HF vector modes.
4711 (SUBST_A): Likewise.
4712 (round_mode512bit_condition): Adjust for V32HFmode.
4714 2021-09-09 liuhongt <hongtao.liu@intel.com>
4717 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
4718 (reduc_plus_scal_v4sf): .. this, New define_expand.
4719 (reduc_plus_scal_v2df): .. and this, New define_expand.
4721 2021-09-09 liuhongt <hongtao.liu@intel.com>
4724 * config/i386/sse.md (*vec_extract<mode><ssescalarmodelower>_valign):
4727 2021-09-08 Jonathan Wakely <jwakely@redhat.com>
4730 * doc/trouble.texi (Copy Assignment): Fix description of
4731 behaviour and fix code in example.
4733 2021-09-08 Segher Boessenkool <segher@kernel.crashing.org>
4736 * config/rs6000/rs6000-logue.c (rs6000_emit_epilogue): For ELFv2 use
4737 r11 instead of r12 for restoring CR.
4739 2021-09-08 Jakub Jelinek <jakub@redhat.com>
4740 liuhongt <hongtao.liu@intel.com>
4743 * config/i386/i386.md (@xorsign<mode>3_1): Remove.
4744 * config/i386/i386-expand.c (ix86_expand_xorsign): Expand right away
4745 into AND with mask and XOR, using paradoxical subregs.
4746 (ix86_split_xorsign): Remove.
4747 * config/i386/i386-protos.h (ix86_split_xorsign): Remove.
4749 2021-09-08 Di Zhao <dizhao@os.amperecomputing.com>
4751 * tree-ssa-sccvn.c (vn_nary_op_insert_into): fix result compare
4753 2021-09-08 Jakub Jelinek <jakub@redhat.com>
4756 * config/i386/i386.md (xorsign<mode>3): If operands[1] is equal to
4757 operands[2], emit abs<mode>2 instead.
4758 (@xorsign<mode>3_1): Add early-clobbers for output operand, enable
4759 first alternative even for avx, add another alternative with
4760 =&Yv <- 0, Yv, Yvm constraints.
4761 * config/i386/i386-expand.c (ix86_split_xorsign): If op0 is equal
4762 to op1, emit vpandn instead.
4764 2021-09-08 liuhongt <hongtao.liu@intel.com>
4766 * config/i386/avx512fp16intrin.h (_mm_set_ph): New intrinsic.
4767 (_mm256_set_ph): Likewise.
4768 (_mm512_set_ph): Likewise.
4769 (_mm_setr_ph): Likewise.
4770 (_mm256_setr_ph): Likewise.
4771 (_mm512_setr_ph): Likewise.
4772 (_mm_set1_ph): Likewise.
4773 (_mm256_set1_ph): Likewise.
4774 (_mm512_set1_ph): Likewise.
4775 (_mm_setzero_ph): Likewise.
4776 (_mm256_setzero_ph): Likewise.
4777 (_mm512_setzero_ph): Likewise.
4778 (_mm_set_sh): Likewise.
4779 (_mm_load_sh): Likewise.
4780 (_mm_store_sh): Likewise.
4781 * config/i386/i386-builtin-types.def (V8HF): New type.
4782 (DEF_FUNCTION_TYPE (V8HF, V8HI)): New builtin function type
4783 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
4784 Support vector HFmodes.
4785 (ix86_expand_vector_init_one_nonzero): Likewise.
4786 (ix86_expand_vector_init_one_var): Likewise.
4787 (ix86_expand_vector_init_interleave): Likewise.
4788 (ix86_expand_vector_init_general): Likewise.
4789 (ix86_expand_vector_set): Likewise.
4790 (ix86_expand_vector_extract): Likewise.
4791 (ix86_expand_vector_init_concat): Likewise.
4792 (ix86_expand_sse_movcc): Handle vector HFmodes.
4793 (ix86_expand_vector_set_var): Ditto.
4794 * config/i386/i386-modes.def: Add HF vector modes in comment.
4795 * config/i386/i386.c (classify_argument): Add HF vector modes.
4796 (ix86_hard_regno_mode_ok): Allow HF vector modes for AVX512FP16.
4797 (ix86_vector_mode_supported_p): Likewise.
4798 (ix86_set_reg_reg_cost): Handle vector HFmode.
4799 (ix86_get_ssemov): Handle vector HFmode.
4800 (function_arg_advance_64): Pass unamed V16HFmode and V32HFmode
4802 (function_arg_advance_32): Pass V8HF/V16HF/V32HF by sse reg for 32bit
4804 (function_arg_advance_32): Ditto.
4805 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): New.
4806 (VALID_AVX256_REG_OR_OI_MODE): Rename to ..
4807 (VALID_AVX256_REG_OR_OI_VHF_MODE): .. this, and add V16HF.
4808 (VALID_SSE2_REG_VHF_MODE): New.
4809 (VALID_AVX512VL_128_REG_MODE): Add V8HF and TImode.
4810 (SSE_REG_MODE_P): Add vector HFmode.
4811 * config/i386/i386.md (mode): Add HF vector modes.
4812 (MODE_SIZE): Likewise.
4813 (ssemodesuffix): Add ph suffix for HF vector modes.
4814 * config/i386/sse.md (VFH_128): New mode iterator.
4815 (VMOVE): Adjust for HF vector modes.
4817 (V_256_512): Likewise.
4819 (avx512fmaskmode): Likewise.
4820 (shuffletype): Likewise.
4821 (sseinsnmode): Likewise.
4822 (ssedoublevecmode): Likewise.
4823 (ssehalfvecmode): Likewise.
4824 (ssehalfvecmodelower): Likewise.
4825 (ssePScmode): Likewise.
4826 (ssescalarmode): Likewise.
4827 (ssescalarmodelower): Likewise.
4828 (sseintprefix): Likewise.
4830 (bcstscalarsuff): Likewise.
4831 (xtg_mode): Likewise.
4832 (VI12HF_AVX512VL): New mode_iterator.
4833 (VF_AVX512FP16): Likewise.
4835 (VIHF_256): Likewise.
4836 (VIHF_AVX512BW): Likewise.
4837 (V16_256): Likewise.
4838 (V32_512): Likewise.
4839 (sseintmodesuffix): New mode_attr.
4840 (sse): Add scalar and vector HFmodes.
4841 (ssescalarmode): Add vector HFmode mapping.
4842 (ssescalarmodesuffix): Add sh suffix for HFmode.
4843 (*<sse>_vm<insn><mode>3): Use VFH_128.
4844 (*<sse>_vm<multdiv_mnemonic><mode>3): Likewise.
4845 (*ieee_<ieee_maxmin><mode>3): Likewise.
4846 (<avx512>_blendm<mode>): New define_insn.
4847 (vec_setv8hf): New define_expand.
4848 (vec_set<mode>_0): New define_insn for HF vector set.
4849 (*avx512fp16_movsh): Likewise.
4850 (avx512fp16_movsh): Likewise.
4851 (vec_extract_lo_v32hi): Rename to ...
4852 (vec_extract_lo_<mode>): ... this, and adjust to allow HF
4854 (vec_extract_hi_v32hi): Likewise.
4855 (vec_extract_hi_<mode>): Likewise.
4856 (vec_extract_lo_v16hi): Likewise.
4857 (vec_extract_lo_<mode>): Likewise.
4858 (vec_extract_hi_v16hi): Likewise.
4859 (vec_extract_hi_<mode>): Likewise.
4860 (vec_set_hi_v16hi): Likewise.
4861 (vec_set_hi_<mode>): Likewise.
4862 (vec_set_lo_v16hi): Likewise.
4863 (vec_set_lo_<mode>): Likewise.
4864 (*vec_extract<mode>_0): New define_insn_and_split for HF
4866 (*vec_extracthf): New define_insn.
4867 (VEC_EXTRACT_MODE): Add HF vector modes.
4868 (PINSR_MODE): Add V8HF.
4869 (sse2p4_1): Likewise.
4870 (pinsr_evex_isa): Likewise.
4871 (<sse2p4_1>_pinsr<ssemodesuffix>): Adjust to support
4872 insert for V8HFmode.
4873 (pbroadcast_evex_isa): Add HF vector modes.
4874 (AVX2_VEC_DUP_MODE): Likewise.
4875 (VEC_INIT_MODE): Likewise.
4876 (VEC_INIT_HALF_MODE): Likewise.
4877 (avx2_pbroadcast<mode>): Adjust to support HF vector mode
4879 (avx2_pbroadcast<mode>_1): Likewise.
4880 (<avx512>_vec_dup<mode>_1): Likewise.
4881 (<avx512>_vec_dup<mode><mask_name>): Likewise.
4882 (<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>):
4885 2021-09-08 Guo, Xuepeng <xuepeng.guo@intel.com>
4886 H.J. Lu <hongjiu.lu@intel.com>
4887 Liu Hongtao <hongtao.liu@intel.com>
4888 Wang Hongyu <hongyu.wang@intel.com>
4889 Xu Dianhong <dianhong.xu@intel.com>
4891 * common/config/i386/cpuinfo.h (get_available_features):
4892 Detect FEATURE_AVX512FP16.
4893 * common/config/i386/i386-common.c
4894 (OPTION_MASK_ISA_AVX512FP16_SET,
4895 OPTION_MASK_ISA_AVX512FP16_UNSET,
4896 OPTION_MASK_ISA2_AVX512FP16_SET,
4897 OPTION_MASK_ISA2_AVX512FP16_UNSET): New.
4898 (OPTION_MASK_ISA2_AVX512BW_UNSET,
4899 OPTION_MASK_ISA2_AVX512BF16_UNSET): Add AVX512FP16.
4900 (ix86_handle_option): Handle -mavx512fp16.
4901 * common/config/i386/i386-cpuinfo.h (enum processor_features):
4902 Add FEATURE_AVX512FP16.
4903 * common/config/i386/i386-isas.h: Add entry for AVX512FP16.
4904 * config.gcc: Add avx512fp16intrin.h.
4905 * config/i386/avx512fp16intrin.h: New intrinsic header.
4906 * config/i386/cpuid.h: Add bit_AVX512FP16.
4907 * config/i386/i386-builtin-types.def: (FLOAT16): New primitive type.
4908 * config/i386/i386-builtins.c: Support _Float16 type for i386
4910 (ix86_register_float16_builtin_type): New function.
4911 (ix86_float16_type_node): New.
4912 * config/i386/i386-c.c (ix86_target_macros_internal): Define
4914 * config/i386/i386-expand.c (ix86_expand_branch): Support
4916 (ix86_prepare_fp_compare_args): Adjust TARGET_SSE_MATH &&
4917 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
4918 (ix86_expand_fp_movcc): Ditto.
4919 * config/i386/i386-isa.def: Add PTA define for AVX512FP16.
4920 * config/i386/i386-options.c (isa2_opts): Add -mavx512fp16.
4921 (ix86_valid_target_attribute_inner_p): Add avx512fp16 attribute.
4922 * config/i386/i386.c (ix86_get_ssemov): Use
4923 vmovdqu16/vmovw/vmovsh for HFmode/HImode scalar or vector.
4924 (ix86_get_excess_precision): Use
4925 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when TARGET_AVX512FP16
4927 (sse_store_index): Use SFmode cost for HFmode cost.
4928 (inline_memory_move_cost): Add HFmode, and perfer SSE cost over
4929 GPR cost for HFmode.
4930 (ix86_hard_regno_mode_ok): Allow HImode in sse register.
4931 (ix86_mangle_type): Add manlging for _Float16 type.
4932 (inline_secondary_memory_needed): No memory is needed for
4933 16bit movement between gpr and sse reg under
4935 (ix86_multiplication_cost): Adjust TARGET_SSE_MATH &&
4936 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
4937 (ix86_division_cost): Ditto.
4938 (ix86_rtx_costs): Ditto.
4939 (ix86_add_stmt_cost): Ditto.
4940 (ix86_optab_supported_p): Ditto.
4941 * config/i386/i386.h (VALID_AVX512F_SCALAR_MODE): Add HFmode.
4942 (SSE_FLOAT_MODE_SSEMATH_OR_HF_P): Add HFmode.
4943 (PTA_SAPPHIRERAPIDS): Add PTA_AVX512FP16.
4944 * config/i386/i386.md (mode): Add HFmode.
4945 (MODE_SIZE): Add HFmode.
4946 (isa): Add avx512fp16.
4947 (enabled): Handle avx512fp16.
4948 (ssemodesuffix): Add sh suffix for HFmode.
4949 (comm): Add mult, div.
4950 (plusminusmultdiv): New code iterator.
4951 (insn): Add mult, div.
4952 (*movhf_internal): Adjust for avx512fp16 instruction.
4953 (*movhi_internal): Ditto.
4954 (*cmpi<unord>hf): New define_insn for HFmode.
4955 (*ieee_s<ieee_maxmin>hf3): Likewise.
4956 (extendhf<mode>2): Likewise.
4957 (trunc<mode>hf2): Likewise.
4958 (float<floatunssuffix><mode>hf2): Likewise.
4959 (*<insn>hf): Likewise.
4960 (cbranchhf4): New expander.
4961 (movhfcc): Likewise.
4962 (<insn>hf3): Likewise.
4965 * config/i386/i386.opt: Add mavx512fp16.
4966 * config/i386/immintrin.h: Include avx512fp16intrin.h.
4967 * doc/invoke.texi: Add mavx512fp16.
4968 * doc/extend.texi: Add avx512fp16 Usage Notes.
4970 2021-09-08 liuhongt <hongtao.liu@intel.com>
4972 * common.opt: Support -fexcess-precision=16.
4973 * config/aarch64/aarch64.c (aarch64_excess_precision): Return
4974 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when
4975 EXCESS_PRECISION_TYPE_FLOAT16.
4976 * config/arm/arm.c (arm_excess_precision): Ditto.
4977 * config/i386/i386.c (ix86_get_excess_precision): Ditto.
4978 * config/m68k/m68k.c (m68k_excess_precision): Issue an error
4979 when EXCESS_PRECISION_TYPE_FLOAT16.
4980 * config/s390/s390.c (s390_excess_precision): Ditto.
4981 * coretypes.h (enum excess_precision_type): Add
4982 EXCESS_PRECISION_TYPE_FLOAT16.
4983 * doc/tm.texi (TARGET_C_EXCESS_PRECISION): Update documents.
4984 * doc/tm.texi.in (TARGET_C_EXCESS_PRECISION): Ditto.
4985 * doc/extend.texi (Half-Precision): Document
4986 -fexcess-precision=16.
4987 * flag-types.h (enum excess_precision): Add
4988 EXCESS_PRECISION_FLOAT16.
4989 * target.def (excess_precision): Update document.
4990 * tree.c (excess_precision_type): Set excess_precision_type to
4991 EXCESS_PRECISION_FLOAT16 when -fexcess-precision=16.
4993 2021-09-08 liuhongt <hongtao.liu@intel.com>
4995 * doc/extend.texi: (@node Floating Types): Adjust the wording.
4996 (@node Half-Precision): Ditto.
4998 2021-09-07 Takayuki 'January June' Suwa <jjsuwa_sys3175@yahoo.co.jp>
5001 * config/xtensa/xtensa.c (xtensa_emit_move_sequence): Add
5002 'CONST_INT_P (src)' to the condition of the block that tries to
5003 eliminate literal when loading integer contant.
5005 2021-09-07 David Faust <david.faust@oracle.com>
5007 * doc/extend.texi (BPF Type Attributes) New node.
5008 Document new preserve_access_index attribute.
5009 Document new preserve_access_index builtin.
5010 * doc/invoke.texi: Document -mco-re and -mno-co-re options.
5012 2021-09-07 David Faust <david.faust@oracle.com>
5014 * config/bpf/bpf.c: Adjust includes.
5015 (bpf_handle_preserve_access_index_attribute): New function.
5016 (bpf_attribute_table): Use it here.
5017 (bpf_builtins): Add BPF_BUILTIN_PRESERVE_ACCESS_INDEX.
5018 (bpf_option_override): Handle "-mco-re" option.
5019 (bpf_asm_init_sections): New.
5020 (TARGET_ASM_INIT_SECTIONS): Redefine.
5021 (bpf_file_end): New.
5022 (TARGET_ASM_FILE_END): Redefine.
5023 (bpf_init_builtins): Add "__builtin_preserve_access_index".
5024 (bpf_core_compute, bpf_core_get_index): New.
5025 (is_attr_preserve_access): New.
5026 (bpf_expand_builtin): Handle new builtins.
5027 (bpf_core_newdecl, bpf_core_is_maybe_aggregate_access): New.
5028 (bpf_core_walk): New.
5029 (bpf_resolve_overloaded_builtin): New.
5030 (TARGET_RESOLVE_OVERLOADED_BUILTIN): Redefine.
5032 (pass_bpf_core_attr): New RTL pass.
5033 * config/bpf/bpf-passes.def: New file.
5034 * config/bpf/bpf-protos.h (make_pass_bpf_core_attr): New.
5035 * config/bpf/coreout.c: New file.
5036 * config/bpf/coreout.h: Likewise.
5037 * config/bpf/t-bpf (TM_H): Add $(srcdir)/config/bpf/coreout.h.
5038 (coreout.o): New rule.
5039 (PASSES_EXTRA): Add $(srcdir)/config/bpf/bpf-passes.def.
5040 * config.gcc (bpf): Add coreout.h to extra_headers.
5041 Add coreout.o to extra_objs.
5042 Add $(srcdir)/config/bpf/coreout.c to target_gtfiles.
5044 2021-09-07 David Faust <david.faust@oracle.com>
5046 * btfout.c (get_btf_id): Function is no longer static.
5047 * ctfc.h: Expose it here.
5049 2021-09-07 David Faust <david.faust@oracle.com>
5051 * ctfc.c (ctf_lookup_tree_type): New function.
5054 2021-09-07 David Faust <david.faust@oracle.com>
5056 * ctfc.c (ctf_dtd_lookup): Function is no longer static.
5057 * ctfc.h: Analogous change.
5059 2021-09-07 David Faust <david.faust@oracle.com>
5061 * dwarf2out.c (lookup_type_die): Function is no longer static.
5062 * dwarf2out.h: Expose it here.
5064 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
5066 * dwarf2ctf.c (ctf_debug_finalize): Make it static.
5067 (ctf_debug_early_finish): New definition.
5068 (ctf_debug_finish): Likewise.
5069 * dwarf2ctf.h (ctf_debug_finalize): Remove declaration.
5070 (ctf_debug_early_finish): New declaration.
5071 (ctf_debug_finish): Likewise.
5072 * dwarf2out.c (dwarf2out_finish): Invoke ctf_debug_finish.
5073 (dwarf2out_early_finish): Invoke ctf_debug_early_finish.
5075 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
5077 * config/bpf/bpf.c (bpf_option_override): For BPF backend, disable LTO
5078 support when compiling for CO-RE.
5079 * config/bpf/bpf.opt: Add new command line option -mco-re.
5081 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
5083 * flag-types.h (enum debug_info_type): Add new enum
5084 DINFO_TYPE_BTF_WITH_CORE.
5085 (BTF_WITH_CORE_DEBUG): New bitmask.
5086 * flags.h (btf_with_core_debuginfo_p): New declaration.
5087 * opts.c (btf_with_core_debuginfo_p): New definition.
5089 2021-09-07 Jason Merrill <jason@redhat.com>
5091 * tree.h (error_operand_p): Change to inline function.
5093 2021-09-07 Aldy Hernandez <aldyh@redhat.com>
5095 * tree-ssa-threadedge.c (forwarder_block_p): Rename to...
5096 (empty_block_with_phis_p): ...this.
5097 (potentially_threadable_block): Same.
5098 (jump_threader::thread_through_normal_block): Same.
5100 2021-09-07 Eric Botcazou <ebotcazou@adacore.com>
5103 * dwarf2out.c (mark_base_types): New overloaded function.
5104 (dwarf2out_early_finish): Invoke it on the COMDAT type list as well
5105 as the compilation unit, and call move_marked_base_types afterward.
5107 2021-09-07 H.J. Lu <hjl.tools@gmail.com>
5110 * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
5112 (ix86_expand_vector_convert_uns_vsivsf): Likewise.
5114 2021-09-07 Richard Biener <rguenther@suse.de>
5116 PR tree-optimization/102226
5117 * tree-vect-loop.c (vect_transform_cycle_phi): Record
5118 the converted value for the epilogue PHI use.
5120 2021-09-07 Martin Liska <mliska@suse.cz>
5122 PR gcov-profile/80223
5123 * ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
5124 options, do not inline when no_profile_instrument_function
5125 attributes are different in early inliner. It's fine to inline
5126 it after PGO instrumentation.
5128 2021-09-07 Richard Biener <rguenther@suse.de>
5130 PR tree-optimization/101555
5131 * tree-ssa-pre.c (translate_vuse_through_block): Do not
5132 perform an alias walk to determine the validity of the
5133 mem at the start of the block which is already guaranteed
5134 by means of prune_clobbered_mems.
5135 (phi_translate_1): Pass edge to translate_vuse_through_block.
5137 2021-09-07 Xionghu Luo <luoxhu@linux.ibm.com>
5140 * config/rs6000/rs6000.md (fmod<mode>3): New define_expand.
5141 (remainder<mode>3): Likewise.
5143 2021-09-07 YunQiang Su <yunqiang.su@cipunited.com>
5145 * config/mips/mips.c (mips_file_start): add .module for
5148 2021-09-06 Roger Sayle <roger@nextmovesoftware.com>
5150 * wide-int.cc (wi::clz): Reorder tests to ensure the result
5151 is zero for all negative values.
5153 2021-09-06 Tobias Burnus <tobias@codesourcery.com>
5155 * doc/invoke.texi (-foffload-options): Fix @opindex.
5157 2021-09-06 H.J. Lu <hjl.tools@gmail.com>
5160 * config/i386/i386-expand.c (ix86_split_xorsign): Use operands[2].
5161 * config/i386/i386.md (@xorsign<mode>3_1): Add non-destructive
5162 source alternative for AVX.
5164 2021-09-06 liuhongt <hongtao.liu@intel.com>
5166 PR middle-end/102182
5167 * optabs.c (expand_fix): Add from1 to avoid from being
5170 2021-09-06 Eric Botcazou <ebotcazou@adacore.com>
5172 * dwarf2out.c (modified_type_die): Deal with all array types earlier
5173 and use local variable consistently throughout the function.
5175 2021-09-06 Jakub Jelinek <jakub@redhat.com>
5177 PR tree-optimization/102207
5178 * match.pd: Don't demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
5179 were promoted from signed to wider unsigned type.
5181 2021-09-06 Andrew Pinski <apinski@marvell.com>
5183 PR tree-optimization/63184
5184 * match.pd: Add simplification of pointer_diff of two pointer_plus
5185 with addr_expr in the first operand of each pointer_plus.
5186 Add simplificatoin of ne/eq of two pointer_plus with addr_expr
5187 in the first operand of each pointer_plus.
5189 2021-09-06 Richard Biener <rguenther@suse.de>
5191 PR tree-optimization/102176
5192 * tree-vect-slp.c (vect_slp_gather_vectorized_scalar_stmts):
5194 (vect_bb_slp_scalar_cost): Use the computed set of
5195 vectorized scalar stmts instead of relying on the out-of-date
5196 and not accurate PURE_SLP_STMT.
5197 (vect_bb_vectorization_profitable_p): Compute the set
5198 of vectorized scalar stmts.
5200 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
5202 * gimple-range-path.cc (path_range_query::range_of_stmt): Remove
5203 GIMPLE_COND special casing.
5204 (path_range_query::range_defined_in_block): Use range_of_stmt
5205 instead of calling fold_range directly.
5207 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
5209 * gimple-range-path.cc (path_range_query::range_of_expr): Set
5210 m_undefined_path when appropriate.
5211 (path_range_query::internal_range_of_expr): Copy from range_of_expr.
5212 (path_range_query::unreachable_path_p): New.
5213 (path_range_query::precompute_ranges): Set m_undefined_path.
5214 * gimple-range-path.h (path_range_query::unreachable_path_p): New.
5215 (path_range_query::internal_range_of_expr): New.
5216 * tree-ssa-threadbackward.c (back_threader::find_taken_edge_cond):
5217 Use unreachable_path_p.
5219 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
5221 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
5222 Remove argument and call find_taken_edge.
5223 (back_threader::resolve_phi): Do not calculate taken edge before
5224 calling maybe_register_path.
5225 (back_threader::find_paths_to_names): Same.
5227 2021-09-05 Jeff Law <jlaw@localhost.localdomain>
5229 * config/h8300/h8300.md (QHSI2 mode iterator): New mode iterator.
5230 * config/h8300/testcompare.md (store_c): Update name, use new
5232 (store_neg_c, store_shifted_c): New patterns.
5234 2021-09-03 Segher Boessenkool <segher@kernel.crashing.org>
5237 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): On ELFv2 use r11
5238 instead of r12 for CR save, in all cases.
5240 2021-09-03 Andrew Pinski <apinski@marvell.com>
5242 * config/aarch64/aarch64-sve-builtins.cc (register_vector_type):
5243 Handle error_mark_node as the type of the type_decl.
5245 2021-09-03 Andrew Pinski <apinski@marvell.com>
5247 * config/aarch64/aarch64-builtins.c (struct aarch64_simd_type_info):
5249 (aarch64_simd_types): Likewise.
5250 (aarch64_simd_intOI_type_node): Likewise.
5251 (aarch64_simd_intCI_type_node): Likewise.
5252 (aarch64_simd_intXI_type_node): Likewise.
5253 * config/aarch64/aarch64.h (aarch64_fp16_type_node): Likewise.
5254 (aarch64_fp16_ptr_type_node): Likewise.
5255 (aarch64_bf16_type_node): Likewise.
5256 (aarch64_bf16_ptr_type_node): Likewise.
5258 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5260 * range-op.cc (operator_minus::op1_op2_relation_effect): Abstract
5262 (minus_op1_op2_relation_effect): ...here.
5263 (class operator_pointer_diff): New.
5264 (operator_pointer_diff::op1_op2_relation_effect): Call
5265 minus_op1_op2_relation_effect.
5266 (integral_table::integral_table): Add entry for POINTER_DIFF_EXPR.
5268 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5270 * tree-ssa-threadbackward.c (back_threader::thread_through_all_blocks):
5271 Add may_peel_loop_headers.
5272 (back_threader_registry::thread_through_all_blocks): Same.
5273 (try_thread_blocks): Pass may_peel_loop_headers argument.
5274 (pass_early_thread_jumps::execute): Same.
5276 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5278 * tree-ssa-threadedge.c (has_phis_p): New.
5279 (forwarder_block_p): New.
5280 (potentially_threadable_block): Call forwarder_block_p.
5281 (jump_threader::thread_around_empty_blocks): Call has_phis_p.
5282 (jump_threader::thread_through_normal_block): Call
5285 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5287 * tree-ssa-threadbackward.c (back_threader::dump): New.
5288 (back_threader::debug): New.
5289 (back_threader_profitability::profitable_path_p): Dump blocks
5290 even if we are bailing early.
5292 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5294 * tree-ssa-threadupdate.c (cancel_thread): New.
5295 (jump_thread_path_registry::thread_block_1): Use cancel_thread.
5296 (jump_thread_path_registry::mark_threaded_blocks): Same.
5297 (jump_thread_path_registry::register_jump_thread): Same.
5299 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5301 * tree-ssa-threadedge.c (jt_state::push): Only call methods for
5302 which objects are available.
5303 (jt_state::pop): Same.
5304 (jt_state::register_equiv): Same.
5305 (jt_state::register_equivs_on_edge): Same.
5307 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5309 * tree-ssa-threadedge.c (jump_threader::thread_across_edge):
5310 Move pop until after a thread is registered.
5312 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5314 * tree-ssa-threadupdate.c (debug): New.
5316 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5318 * gimple-range-trace.cc (push_dump_file::push_dump_file): New.
5319 (push_dump_file::~push_dump_file): New.
5320 (dump_ranger): Change dump_file temporarily while dumping
5322 * gimple-range-trace.h (class push_dump_file): New.
5324 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5326 * gimple-range-trace.cc (debug_seed_ranger): Remove static.
5327 (dump_ranger): Dump function name.
5329 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5331 * gimple-range-path.cc (path_range_query::range_defined_in_block):
5332 Adjust for non-null.
5333 (path_range_query::adjust_for_non_null_uses): New.
5334 (path_range_query::precompute_ranges): Call
5335 adjust_for_non_null_uses.
5336 * gimple-range-path.h: Add m_non_null and
5337 adjust_for_non_null_uses.
5339 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5341 * gimple-range-path.cc (path_range_query::dump): Dump path
5343 (path_range_query::precompute_ranges): Dump entire path.
5345 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5347 * value-relation.cc (relation_oracle::debug): New.
5348 * value-relation.h (relation_oracle::debug): New.
5350 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5352 * tree-ssa-loop-ch.c: Remove unnecessary include file.
5354 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5356 * gimple-range-fold.cc (fold_using_range::postfold_gcond_edges):
5357 Skip statements with no defining BB.
5358 * gimple-range-path.cc (path_range_query::range_defined_in_block):
5359 Do not get confused by statements with no defining BB.
5361 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
5363 * gimple-range-fold.cc (adjust_imagpart_expr): Move from
5364 gimple_range_adjustment. Add support for constants.
5365 (adjust_realpart_expr): New.
5366 (gimple_range_adjustment): Move IMAGPART_EXPR code to
5367 adjust_imagpart_expr.
5368 * range-op.cc (integral_table::integral_table): Add entry for
5371 2021-09-03 Jakub Jelinek <jakub@redhat.com>
5373 * omp-expand.c (expand_omp_atomic_pipeline): Use
5374 IFN_ATOMIC_COMPARE_EXCHANGE instead of
5375 BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_? so that memory order
5378 2021-09-03 Jakub Jelinek <jakub@redhat.com>
5381 * tree.h (DECL_FIELD_ABI_IGNORED): Changed into rvalue only macro
5382 that is false if DECL_BIT_FIELD.
5383 (SET_DECL_FIELD_ABI_IGNORED, DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD,
5384 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD): Define.
5385 * tree-streamer-out.c (pack_ts_decl_common_value_fields): For
5386 DECL_BIT_FIELD stream DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead
5387 of DECL_FIELD_ABI_IGNORED.
5388 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
5389 SET_DECL_FIELD_ABI_IGNORED instead of writing to
5390 DECL_FIELD_ABI_IGNORED and for DECL_BIT_FIELD use
5391 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead.
5392 * lto-streamer-out.c (hash_tree): For DECL_BIT_FIELD hash
5393 DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead of DECL_FIELD_ABI_IGNORED.
5395 2021-09-03 liuhongt <hongtao.liu@intel.com>
5398 * config/i386/amxbf16intrin.h : Remove macro check for __AMX_BF16__.
5399 * config/i386/amxint8intrin.h : Remove macro check for __AMX_INT8__.
5400 * config/i386/amxtileintrin.h : Remove macro check for __AMX_TILE__.
5402 2021-09-02 Martin Sebor <msebor@redhat.com>
5404 PR tree-optimization/17506
5406 * tree-ssa-uninit.c (warn_uninit): Remove conditional guarding note.
5408 2021-09-02 Richard Biener <rguenther@suse.de>
5410 * tree-ssa-loop-im.c (fill_always_executed_in_1): Refine
5411 fix for PR78185 and continue processing when leaving
5414 2021-09-02 Jakub Jelinek <jakub@redhat.com>
5416 PR tree-optimization/99591
5417 * match.pd: Demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
5420 2021-09-02 Richard Biener <rguenther@suse.de>
5423 2021-09-02 Richard Biener <rguenther@suse.de>
5425 PR tree-optimization/102155
5426 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
5427 over a part of the RPO array and do not recurse here.
5428 Dump blocks marked as always executed.
5429 (fill_always_executed_in): Walk over the RPO array and
5430 process loops whose header we run into.
5431 (loop_invariant_motion_in_fun): Compute the first RPO
5432 using rev_post_order_and_mark_dfs_back_seme in iteration
5433 order and pass that to fill_always_executed_in.
5435 2021-09-02 liuhongt <hongtao.liu@intel.com>
5437 * config/i386/i386-modes.def (FLOAT_MODE): Define ieee HFmode.
5438 * config/i386/i386.c (enum x86_64_reg_class): Add
5440 (merge_classes): Handle X86_64_SSEHF_CLASS.
5441 (examine_argument): Ditto.
5442 (construct_container): Ditto.
5443 (classify_argument): Ditto, and set HFmode/HCmode to
5445 (function_value_32): Return _FLoat16/Complex Float16 by
5447 (function_value_64): Return _Float16/Complex Float16 by SSE
5449 (ix86_print_operand): Handle CONST_DOUBLE HFmode.
5450 (ix86_secondary_reload): Require gpr as intermediate register
5451 to store _Float16 from sse register when sse4 is not
5453 (ix86_libgcc_floating_mode_supported_p): Enable _FLoat16 under
5455 (ix86_scalar_mode_supported_p): Ditto.
5456 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Defined.
5457 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add HFmode.
5458 (VALID_INT_MODE_P): Add HFmode and HCmode.
5459 * config/i386/i386.md (*pushhf_rex64): New define_insn.
5461 (*movhf_internal): Ditto.
5462 * doc/extend.texi (Half-Precision Floating Point): Documemt
5465 2021-09-02 Richard Biener <rguenther@suse.de>
5467 PR tree-optimization/102155
5468 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
5469 over a part of the RPO array and do not recurse here.
5470 Dump blocks marked as always executed.
5471 (fill_always_executed_in): Walk over the RPO array and
5472 process loops whose header we run into.
5473 (loop_invariant_motion_in_fun): Compute the first RPO
5474 using rev_post_order_and_mark_dfs_back_seme in iteration
5475 order and pass that to fill_always_executed_in.
5477 2021-09-02 YunQiang Su <syq@debian.org>
5480 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
5482 * config/mips/mips.c (mips_module_isa_name): New.
5483 mips_file_start: add .module mipsREV to all asm output
5485 2021-09-01 Jeff Law <jlaw@localhost.localdomain>
5487 PR tree-optimization/102152
5488 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Reduce a vector
5489 comparison to a scalar comparison before calling
5490 update_stmt_if_modified.
5492 2021-09-01 Andrew Pinski <apinski@marvell.com>
5495 * config/aarch64/aarch64.c (aarch64_expand_setmem):
5496 Check STRICT_ALIGNMENT before creating an overlapping
5499 2021-09-01 Martin Sebor <msebor@redhat.com>
5501 * gimple-ssa-warn-access.cc (get_size_range): Add argument.
5502 (check_access): Pass additional argument.
5503 (check_memop_access): Remove template and make a member function.
5504 (maybe_check_dealloc_call): Make a pass_waccess member function.
5505 (class pass_waccess): Add, rename, and remove members.
5506 (pass_waccess::pass_waccess): Adjust to name change.
5507 (pass_waccess::~pass_waccess): Same.
5508 (check_alloca): Make a member function.
5509 (check_alloc_size_call): Same.
5510 (check_strcat): Same.
5511 (check_strncat): Same.
5512 (check_stxcpy): Same.
5513 (check_stxncpy): Same.
5514 (check_strncmp): Same.
5515 (maybe_warn_rdwr_sizes): Rename...
5516 (pass_waccess::maybe_check_access_sizes): ...to this.
5517 (pass_waccess::check_call): Adjust to name changes.
5518 (pass_waccess::maybe_check_dealloc_call): Make a pass_waccess member
5520 (pass_waccess::execute): Adjust to name changes.
5521 * gimple-ssa-warn-access.h (check_memop_access): Remove.
5522 * pointer-query.cc (access_ref::phi): Handle null pointer.
5523 (access_ref::inform_access): Same.
5524 (pointer_query::put_ref): Modify a cached value, not a copy of it.
5525 (pointer_query::dump): New function.
5526 (compute_objsize_r): Avoid overwriting access_ref::bndrng. Cache
5528 * pointer-query.h (pointer_query::dump): Declare.
5529 * tree-ssa-strlen.c (get_range): Simplify. Use function query.
5530 (dump_strlen_info): Use function query.
5531 (printf_strlen_execute): Factor code out into pointer_query::put_ref.
5533 2021-09-01 Thomas Schwinge <thomas@codesourcery.com>
5535 * tree.c (walk_tree_1) <OMP_CLAUSE>: Simplify.
5537 2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
5539 * doc/extend.texi: Document unavailable attribute.
5540 * print-tree.c (print_node): Handle unavailable attribute.
5541 * tree-core.h (struct tree_base): Add a bit to carry unavailability.
5542 * tree.c (error_unavailable_use): New.
5543 * tree.h (TREE_UNAVAILABLE): New.
5544 (error_unavailable_use): New.
5546 2021-09-01 Jakub Jelinek <jakub@redhat.com>
5548 PR tree-optimization/102124
5549 * tree-vect-patterns.c (vect_recog_widen_op_pattern): For ORIG_CODE
5550 MINUS_EXPR, if itype is unsigned with smaller precision than type,
5551 add an extra cast to signed variant of itype to ensure sign-extension.
5553 2021-09-01 Martin Liska <mliska@suse.cz>
5555 * graph.c (draw_cfg_node_succ_edges): Do not color fallthru
5556 edges and rather use colors for TRUE and FALSE edges.
5558 2021-09-01 Richard Biener <rguenther@suse.de>
5560 PR tree-optimization/93491
5561 * tree-ssa-pre.c (compute_avail): Set BB_MAY_NOTRETURN
5562 after processing the stmt itself. Do not consider
5563 pure functions possibly not returning. Properly avoid
5564 adding possibly trapping calls to EXP_GEN when there's
5565 a preceeding possibly not returning call.
5566 * tree-ssa-sccvn.c (vn_reference_may_trap): Conservatively
5569 2021-09-01 Richard Biener <rguenther@suse.de>
5571 PR tree-optimization/102139
5572 * tree-vectorizer.h (vec_base_alignments): Adjust hash-map
5573 type to record a std::pair of the stmt-info and the innermost
5575 (dr_vec_info::group): New member.
5576 * tree-vect-data-refs.c (vect_record_base_alignment): Adjust.
5577 (vect_compute_data_ref_alignment): Verify the recorded
5578 base alignment can be used.
5579 (data_ref_pair): Remove.
5580 (dr_group_sort_cmp): Adjust.
5581 (vect_analyze_data_ref_accesses): Store the group-ID in the
5582 dr_vec_info and operate on a vector of dr_vec_infos.
5584 2021-09-01 YunQiang Su <yunqiang.su@cipunited.com>
5586 * read-md.c (md_reader::handle_enum): support value assignation.
5587 * doc/md.texi: record define_c_enum value assignation support.
5589 2021-09-01 Jakub Jelinek <jakub@redhat.com>
5591 PR tree-optimization/102141
5592 * gimple-ssa-store-merging.c (bswap_view_convert): Add BEFORE
5593 argument. If false, emit stmts after gsi instead of before, and
5595 (bswap_replace): Adjust callers. When converting output of bswap,
5596 emit VIEW_CONVERT prepratation stmts after a copy of gsi instead
5599 2021-09-01 liuhongt <hongtao.liu@intel.com>
5601 * emit-rtl.c (validate_subreg): Get rid of all float-int
5604 2021-09-01 liuhongt <hongtao.liu@intel.com>
5607 2021-08-30 liuhongt <hongtao.liu@intel.com>
5609 * expmed.c (extract_bit_field_1): Make sure we're playing with
5610 integral modes before call extract_integral_bit_field.
5611 (extract_integral_bit_field): Add a parameter of type
5612 scalar_int_mode which corresponds to of tmode.
5613 And call extract_and_convert_fixed_bit_field instead of
5614 extract_fixed_bit_field and convert_extracted_bit_field.
5615 (extract_and_convert_fixed_bit_field): New function, it's a
5616 combination of extract_fixed_bit_field and
5617 convert_extracted_bit_field.
5619 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
5621 * tree.c (walk_tree_1) <OMP_CLAUSE_TILE>: Handle three operands.
5623 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
5625 * omp-general.h (omp_is_reference): Rename to...
5626 (omp_privatize_by_reference): ... this. Adjust all users...
5627 * omp-general.c: ... here, ...
5628 * gimplify.c: ... here, ...
5629 * omp-expand.c: ... here, ...
5630 * omp-low.c: ... here.
5632 2021-08-31 Martin Sebor <msebor@redhat.com>
5634 * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Test
5635 pointer element for equality to zero, not that of the cotaining
5638 2021-08-31 Martin Sebor <msebor@redhat.com>
5640 * gcc-rich-location.h (gcc_rich_location): Make ctor explicit.
5642 2021-08-31 Martin Sebor <msebor@redhat.com>
5644 * function.h (function): Add comments.
5645 (get_range_query): Same. Add attribute returns nonnull.
5647 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
5649 * expr.c (convert_modes): Don't use subreg_promoted_mode on a
5650 SUBREG if it can't be guaranteed to a SUBREG_PROMOTED_VAR_P set.
5651 Instead use the standard (safer) is_a <scalar_int_mode> idiom.
5653 2021-08-31 Jeff Law <jlaw@localhost.localdomain>
5655 * config.gcc (cris-*-elf, cris-*-none): Remove dbxelf.h from
5657 (m32r-*-elf, m32rle-*-elf, m32r-*-linux): Likewise.
5658 (mn10300-*-*, am33_2.0-*-linux*): Likewise.
5659 (xtensa*-*-elf, xtensa*-*-linux, xtensa*-*-uclinux): Likewise.
5660 (m32c-*-elf*, m32c-*-rtems*): Likewise.
5661 * config/cris/cris.h (DBX_NO_XREFS): Remove.
5662 (DBX_CONTIN_LENGTH, DBX_CONTIN_CHAR): Likewise.
5663 * config/m32r/m32r.h (DBXOUT_SOURCE_LINE): Likewise.
5664 (DBX_DEBUGGING_INFO, DBX_CONTIN_LENGTH): Likewise.
5665 * config/mn10300/mn10300.h (DEFAULT_GDB_EXTENSIONS): Likewise.
5666 * config/mn10300/linux.h (DBX_REGISTER_NAMES): Likewise.
5668 2021-08-31 Marcel Vollweiler <marcel@codesourcery.com>
5670 * gimplify.c (gimplify_scan_omp_clauses): Error handling. 'ancestor' only
5671 allowed on target constructs and only with particular other clauses.
5672 * omp-expand.c (expand_omp_target): Output of 'sorry, not supported' if
5674 * omp-low.c (check_omp_nesting_restrictions): Error handling. No nested OpenMP
5675 structs when 'ancestor' is used.
5676 (scan_omp_1_stmt): No usage of OpenMP runtime routines in a target region when
5678 * tree-pretty-print.c (dump_omp_clause): Append 'ancestor'.
5679 * tree.h (OMP_CLAUSE_DEVICE_ANCESTOR): Define macro.
5681 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
5683 * expr.c (convert_modes): Preserve SUBREG_PROMOTED_VAR_P when
5684 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
5686 * simplify-rtx.c (simplify_unary_operation_1) [SIGN_EXTEND]:
5687 Likewise, preserve SUBREG_PROMOTED_VAR_P when creating a (wider)
5688 partial subreg from a SUBREG_PROMOTED_VAR_P subreg. Generate
5689 SIGN_EXTEND of the SUBREG_REG when a subreg would be paradoxical.
5690 [ZERO_EXTEND]: Likewise, preserve SUBREG_PROMOTED_VAR_P when
5691 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
5692 subreg. Generate ZERO_EXTEND of the SUBREG_REG when a subreg
5693 would be paradoxical.
5695 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
5697 * combine.c (combine_simplify_rtx): Avoid converting an explicit
5698 TRUNCATE into a lowpart SUBREG on !TRULY_NOOP_TRUNCATION targets.
5699 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
5701 2021-08-31 Richard Biener <rguenther@suse.de>
5703 PR tree-optimization/102142
5704 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Fix
5705 condition under which to unset the visited flag.
5707 2021-08-31 Richard Biener <rguenther@suse.de>
5709 PR middle-end/102129
5710 * tree-ssa-ter.c (find_replaceable_in_bb): Do not move
5711 possibly trapping expressions across calls.
5713 2021-08-31 Jakub Jelinek <jakub@redhat.com>
5715 PR tree-optimization/102134
5716 * tree-ssa-ccp.c (bit_value_binop) <case RSHIFT_EXPR>: If sgn is
5717 UNSIGNED and r1val | r1mask has MSB set, ensure lzcount doesn't
5720 2021-08-31 Andrew Pinski <apinski@marvell.com>
5723 * collect-utils.c (setup_signals): New declaration.
5724 * collect-utils.h (setup_signals): New function.
5725 * collect2.c (handler): Delete.
5726 (main): Instead of manually setting up the signals,
5727 just call setup_signals.
5728 * lto-wrapper.c (main): Likewise.
5730 2021-08-31 Andrew Pinski <apinski@marvell.com>
5733 * config/i386/i386-protos.h (x86_output_aligned_bss):
5734 Change align argument to unsigned type.
5735 (x86_elf_aligned_decl_common): Likewise.
5736 * config/i386/i386.c (x86_elf_aligned_decl_common): Likewise.
5737 (x86_output_aligned_bss): Likewise.
5739 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
5741 * config/mips/mips.c (mips_module_isa_name): New.
5742 mips_file_start: add .module mipsREV to all asm output
5744 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
5746 * config/mips/mips.h (struct mips_cpu_info): define enum mips_isa;
5747 use enum instead of int for 'isa' member.
5748 * config.gcc, config/mips/mips.c, config/mips/mips-cpus.def,
5749 config/mips/netbsd.h: replace hardcoded numbers with enum.
5751 2021-08-31 liuhongt <hongtao.liu@intel.com>
5753 * config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
5754 define_split to define_insn_and_split.
5755 (*avx2_eq<mode>3): Removed.
5756 (<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
5757 (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
5758 (*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
5760 (*avx2_gt<mode>3): Removed.
5761 (<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
5762 define_insn to define_expand, and adjust pattern.
5763 (UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.
5765 2021-08-30 David Malcolm <dmalcolm@redhat.com>
5768 * Makefile.in (ANALYZER_OBJS): Add analyzer/call-info.o.
5770 2021-08-30 Jason Merrill <jason@redhat.com>
5772 * doc/invoke.texi: Document -Wmissing-requires.
5774 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
5776 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Remove
5777 TARGET_EXTRA_BUILTINS guard.
5779 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
5781 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Change
5782 initialization of V2DI_type_node and unsigned_V2DI_type_node.
5784 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
5786 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Use the new
5787 decl when new_builtins_are_live.
5788 * config/rs6000/rs6000-builtin-new.def (__builtin_cfstring): New
5791 2021-08-30 Pat Haugen <pthaugen@linux.ibm.com>
5793 * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add
5794 OPTION_MASK_P10_FUSION_2STORE.
5795 (POWERPC_MASKS): Likewise.
5796 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
5797 store fusion for Power10.
5798 (is_fusable_store): New.
5799 (power10_sched_reorder): Likewise.
5800 (rs6000_sched_reorder): Do Power10 specific reordering.
5801 (rs6000_sched_reorder2): Likewise.
5802 * config/rs6000/rs6000.opt: Add new option.
5804 2021-08-30 Richard Biener <rguenther@suse.de>
5806 PR tree-optimization/102128
5807 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
5808 Move scanning for if-converted scalar code to the caller
5809 and instead delay clearing the visited flag for profitable
5811 (vect_slp_region): Cost all subgraphs before scheduling.
5812 For if-converted BB vectorization scan for scalar COND_EXPRs
5813 and do not vectorize if any found and the cost model is
5816 2021-08-30 Richard Biener <rguenther@suse.de>
5818 * common.opt (fexceptions): Mark
5819 EnabledBy(fnon-call-exceptions).
5820 * doc/invoke.texi (fnon-call-exceptions): Document this
5821 enables -fexceptions.
5823 2021-08-30 Sebastian Huber <sebastian.huber@embedded-brains.de>
5825 * tsystem.h (abort): Define abort() if inhibit_libc is defined and it
5826 is not already defined.
5828 2021-08-30 liuhongt <hongtao.liu@intel.com>
5830 * expmed.c (extract_bit_field_1): Make sure we're playing with
5831 integral modes before call extract_integral_bit_field.
5832 (extract_integral_bit_field): Add a parameter of type
5833 scalar_int_mode which corresponds to of tmode.
5834 And call extract_and_convert_fixed_bit_field instead of
5835 extract_fixed_bit_field and convert_extracted_bit_field.
5836 (extract_and_convert_fixed_bit_field): New function, it's a
5837 combination of extract_fixed_bit_field and
5838 convert_extracted_bit_field.
5840 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
5842 * config/darwin.c (darwin_libc_has_function): Do not run
5843 the checks for x86 or modern Darwin. Make sure that there
5844 is a value set for darwin_macosx_version_min before testing.
5846 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
5848 * config/i386/darwin.h (CLEAR_INSN_CACHE): New.
5850 2021-08-28 Jan Hubicka <hubicka@ucw.cz>
5852 * ipa-modref-tree.h (modref_access_node::merge): Break out
5853 logic combining offsets and logic merging ranges to ...
5854 (modref_access_node::combined_offsets): ... here
5855 (modref_access_node::update2): ... here
5856 (modref_access_node::closer_pair_p): New member function.
5857 (modref_access_node::forced_merge): New member function.
5858 (modre_ref_node::insert): Do merging when table is full.
5860 2021-08-28 YunQiang Su <yunqiang.su@cipunited.com>
5863 * config.gcc: MIPS: use N64 ABI by default if the triple end
5864 with -gnuabi64, which is used by Debian since 2013.
5866 2021-08-28 Alexandre Oliva <oliva@adacore.com>
5868 * ipa-modref.c (analyze_function): Skip debug stmts.
5869 * tree-inline.c (estimate_num_insn): Consider builtins even
5870 without a cgraph_node.
5872 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
5874 * config/h8300/bitfield.md (cstore<mode>4): Remove expander.
5875 * config/h8300/h8300.c (h8300_expand_branch): Remove function.
5876 * config/h8300/h8300-protos.h (h8300_expadn_branch): Remove prototype.
5877 * config/h8300/h8300.md (eqne): New code iterator.
5878 (geultu, geultu_to_c): Similarly.
5879 * config/h8300/testcompare.md (cstore<mode>4): Dummy expander.
5880 (store_c_<mode>, store_c_i_<mode>): New define_insn_and_splits
5881 (cmp<mode>_c): New pattern
5883 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
5885 * tree-ssa-dom.c (reduce_vector_comparison_to_scalar_comparison): New
5887 (dom_opt_dom_walker::optimize_stmt): Use it.
5889 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
5891 * config/darwin.c (finalize_ctors): Add a section-start linker-
5893 (finalize_dtors): Likewise.
5894 * config/darwin.h (MIN_LD64_INIT_TERM_START_LABELS): New.
5896 2021-08-27 Bill Schmidt <wschmidt@linux.ibm.com>
5898 * config/rs6000/rs6000-call.c (rs6000-builtins.h): New #include.
5899 (rs6000_init_builtins): Call rs6000_init_generated_builtins. Skip the
5900 old initialization logic when new builtins are enabled.
5901 * config/rs6000/rs6000-gen-builtins.c (write_decls): Rename
5902 rs6000_autoinit_builtins to rs6000_init_generated_builtins.
5903 (write_init_file): Likewise.
5905 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
5907 * configure.ac (darwin2[[0-9]]* | darwin19*): Alter use of
5908 gcc_GAS_CHECK_FEATURE to remove an extraneous parameter.
5909 (amdgcn-* | gcn-*) Likewise.
5911 2021-08-27 Anthony Sharp <anthonysharp15@gmail.com>
5913 * symbol-summary.h: Added missing template keyword.
5915 2021-08-27 Richard Biener <rguenther@suse.de>
5917 PR tree-optimization/45178
5918 * tree-ssa-dce.c (find_obviously_necessary_stmts): For
5919 infinite loops without exit do not mark control dependent
5920 edges of the latch necessary.
5922 2021-08-27 konglin1 <lingling.kong@intel.com>
5925 * config/i386/sse.md: (<avx512>scattersi<mode>): Add mask operand to
5927 (<avx512>scattersi<mode>): Likewise.
5928 (*avx512f_scattersi<VI48F:mode>): Merge mask operand to set_dest.
5929 (*avx512f_scatterdi<VI48F:mode>): Likewise
5931 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
5933 * config/rs6000/rs6000.c (rs6000_builtin_md_vectorized_function): Add
5934 support for built-in functions MISC_BUILTIN_DIVWE, MISC_BUILTIN_DIVWEU,
5935 MISC_BUILTIN_DIVDE, MISC_BUILTIN_DIVDEU, P10_BUILTIN_CFUGED,
5936 P10_BUILTIN_CNTLZDM, P10_BUILTIN_CNTTZDM, P10_BUILTIN_PDEPD and
5937 P10_BUILTIN_PEXTD on Power10.
5939 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
5941 * config/rs6000/rs6000-call.c (builtin_function_type): Add unsigned
5942 signedness for some Power10 bifs.
5944 2021-08-27 David Edelsohn <dje.gcc@gmail.com>
5947 * config/rs6000/rs6000.c (rs6000_adjust_field_align): Use
5948 computed alignment if the entire struct has attribute packed.
5950 2021-08-27 liuhongt <hongtao.liu@intel.com>
5954 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
5955 IX86_BUILTIN_SHUFPD512, IX86_BUILTIN_SHUFPS512,
5956 IX86_BUILTIN_SHUFPD256, IX86_BUILTIN_SHUFPS,
5957 IX86_BUILTIN_SHUFPS256.
5958 (ix86_masked_all_ones): New function.
5960 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
5962 * config/i386/i386.md (*btr<mode>_1): Call force_reg unconditionally.
5963 (conditional moves with memory inputs splitters): Ditto.
5964 * config/i386/sse.md (one_cmpl<mode>2): Simplify.
5966 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
5968 * ipa-modref-tree.h (modref_access_node::try_merge_with): Restart
5969 search after merging.
5971 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
5973 * config/rs6000/rs6000-overload.def: Add remaining overloads.
5975 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
5977 * config/rs6000/rs6000-builtin-new.def: Add cell stanza.
5979 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
5981 * config/rs6000/rs6000-builtin-new.def: Add ieee128-hw, dfp,
5982 crypto, and htm stanzas.
5984 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
5986 * config/rs6000/rs6000-builtin-new.def: Add mma stanza.
5988 2021-08-26 Martin Sebor <msebor@redhat.com>
5990 * tree-ssa-uninit.c (warn_uninit): Refactor and simplify.
5991 (warn_uninit_phi_uses): Remove argument from calls to warn_uninit.
5992 (warn_uninitialized_vars): Same. Reduce visibility of locals.
5993 (warn_uninitialized_phi): Same.
5995 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
5997 * tree-ssa-ccp.c (get_individual_bits): Helper function to
5998 extract the individual bits from a widest_int constant (mask).
5999 (gray_code_bit_flips): New read-only table for effiently
6000 enumerating permutations/combinations of bits.
6001 (bit_value_binop) [LROTATE_EXPR, RROTATE_EXPR]: Handle rotates
6002 by unknown counts that are guaranteed less than the target
6003 precision and four or fewer unknown bits by enumeration.
6004 [LSHIFT_EXPR, RSHIFT_EXPR]: Likewise, also handle shifts by
6005 enumeration under the same conditions. Handle remaining
6006 shifts as a mask based upon the minimum possible shift value.
6008 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
6009 Richard Biener <rguenther@suse.de>
6011 * match.pd (shift transformations): Remove a redundant
6012 !POINTER_TYPE_P check.
6014 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
6017 * config/i386/i386.md (cmove reg-reg move elimination peephole2s):
6018 Set all_regs to true in the call to replace_rtx.
6020 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
6022 * ipa-modref-tree.c (test_insert_search_collapse): Update test.
6023 * ipa-modref-tree.h (modref_base_node::insert): Be smarter when
6024 hiting --param modref-max-refs limit.
6025 (modref_tree:insert_base): Be smarter when hitting
6026 --param modref-max-bases limit. Add new parameter REF.
6027 (modref_tree:insert): Update.
6028 (modref_tree:merge): Update.
6029 * ipa-modref.c (read_modref_records): Update.
6031 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
6033 * params.opt: (modref-max-adjustments): Add full stop.
6035 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
6037 * ipa-modref-tree.h (modref_ref_node::verify): New member
6039 (modref_ref_node::insert): Use it.
6040 (modref_ref_node::try_mere_with): Fix off by one error.
6042 2021-08-26 Martin Liska <mliska@suse.cz>
6043 Stefan Kneifel <stefan.kneifel@bluewin.ch>
6045 * cgraph.h (create_version_clone_with_body): Add new parameter.
6046 * cgraphclones.c: Likewise.
6047 * multiple_target.c (create_dispatcher_calls): Do not use
6049 (create_target_clone): Likewise here.
6051 2021-08-26 Jonathan Yong <10walls@gmail.com>
6053 * doc/extend.texi: Add note about reserved priorities
6054 to the constructor attribute.
6056 2021-08-25 Martin Sebor <msebor@redhat.com>
6058 * gimple-range-cache.cc (ssa_global_cache::dump): Avoid printing
6059 range table header alone.
6060 * gimple-range.cc (gimple_ranger::export_global_ranges): Same.
6062 2021-08-25 Jan Hubicka <hubicka@ucw.cz>
6064 * doc/invoke.texi: Document --param modref-max-adjustments.
6065 * ipa-modref-tree.c (test_insert_search_collapse): Update.
6066 (test_merge): Update.
6067 * ipa-modref-tree.h (struct modref_access_node): Add adjustments;
6068 (modref_access_node::operator==): Fix handling of access ranges.
6069 (modref_access_node::contains): Constify parameter; handle also
6070 mismatched parm offsets.
6071 (modref_access_node::update): New function.
6072 (modref_access_node::merge): New function.
6073 (unspecified_modref_access_node): Update constructor.
6074 (modref_ref_node::insert_access): Add record_adjustments parameter;
6076 (modref_ref_node::try_merge_with): New private function.
6077 (modref_tree::insert): New record_adjustments parameter.
6078 (modref_tree::merge): New record_adjustments parameter.
6079 (modref_tree::copy_from): Update.
6080 * ipa-modref.c (dump_access): Dump adjustments field.
6081 (get_access): Update constructor.
6082 (record_access): Update call of insert.
6083 (record_access_lto): Update call of insert.
6084 (merge_call_side_effects): Add record_adjustments parameter.
6085 (get_access_for_fnspec): Update.
6086 (process_fnspec): Update.
6087 (analyze_call): Update.
6088 (analyze_function): Update.
6089 (read_modref_records): Update.
6090 (ipa_merge_modref_summary_after_inlining): Update.
6091 (propagate_unknown_call): Update.
6092 (modref_propagate_in_scc): Update.
6093 * params.opt (param-max-modref-adjustments=): New.
6095 2021-08-25 Michael Meissner <meissner@linux.ibm.com>
6097 * config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
6099 (xxspltiw_v4si): Use vecperm type attribute.
6100 (xxspltiw_v4si_inst): Use vecperm type attribute.
6101 (xxspltiw_v4sf_inst): Likewise.
6102 (xxspltidp_v2df): Use vecperm type attribute. Use
6103 UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
6104 (xxspltidp_v2df_inst): Likewise.
6105 (xxsplti32dx_v4si): Use vecperm type attribute.
6106 (xxsplti32dx_v4si_inst): Likewise.
6107 (xxsplti32dx_v4sf_inst): Likewise.
6108 (xxblend_<mode>): Likewise.
6109 (xxpermx): Likewise.
6110 (xxpermx_inst): Likewise.
6113 2021-08-25 Lewis Hyatt <lhyatt@gmail.com>
6116 * coretypes.h (typedef diagnostic_input_charset_callback): Declare.
6117 * diagnostic.c (diagnostic_initialize_input_context): New function.
6118 * diagnostic.h (diagnostic_initialize_input_context): Declare.
6119 * input.c (default_charset_callback): New function.
6120 (file_cache::initialize_input_context): New function.
6121 (file_cache_slot::create): Added ability to convert the input
6122 according to the input context.
6123 (file_cache::file_cache): Initialize the new input context.
6124 (class file_cache_slot): Added new m_alloc_offset member.
6125 (file_cache_slot::file_cache_slot): Initialize the new member.
6126 (file_cache_slot::~file_cache_slot): Handle potentially offset buffer.
6127 (file_cache_slot::maybe_grow): Likewise.
6128 (file_cache_slot::needs_read_p): Handle NULL fp, which is now possible.
6129 (file_cache_slot::get_next_line): Likewise.
6130 * input.h (class file_cache): Added input context member.
6132 2021-08-25 Richard Biener <rguenther@suse.de>
6134 PR tree-optimization/102046
6135 * tree-vect-slp.c (vect_build_slp_tree_2): Conservatively
6136 update ->any_pattern when swapping operands.
6138 2021-08-25 Hongyu Wang <hongyu.wang@intel.com>
6141 * config/i386/i386.c (ix86_live_on_entry): Adjust comment.
6142 (ix86_decompose_address): Remove retval check for ASHIFT,
6143 allow non-canonical zero extend if AND mask covers ASHIFT
6145 (ix86_legitimate_address_p): Adjust condition for decompose.
6146 (ix86_rtx_costs): Adjust cost for lea with non-canonical
6148 Co-Authored by: Uros Bizjak <ubizjak@gmail.com>
6150 2021-08-25 Jiufu Guo <guojiufu@linux.ibm.com>
6152 PR tree-optimization/101145
6153 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
6155 (number_of_iterations_lt): Invoke above function.
6156 (adjust_cond_for_loop_until_wrap):
6157 Merge to number_of_iterations_until_wrap.
6158 (number_of_iterations_cond): Update invokes for
6159 adjust_cond_for_loop_until_wrap and number_of_iterations_lt.
6161 2021-08-25 konglin1 <lingling.kong@intel.com>
6164 * config/i386/avx512dqintrin.h (_mm512_fpclass_ps_mask): Fix
6166 (_mm512_mask_fpclass_ps_mask): Ditto.
6168 2021-08-25 Kewen Lin <linkw@linux.ibm.com>
6170 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Remove.
6171 (vec_unpacku_hi_v8hi): Likewise.
6172 (vec_unpacku_lo_v16qi): Likewise.
6173 (vec_unpacku_lo_v8hi): Likewise.
6174 (vec_unpacku_hi_<VP_small_lc>): New define_expand.
6175 (vec_unpacku_lo_<VP_small_lc>): Likewise.
6177 2021-08-24 David Edelsohn <dje.gcc@gmail.com>
6179 * config/rs6000/aix.h (SYSTEM_IMPLICIT_EXTERN_C): Delete.
6180 * config/rs6000/aix71.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
6181 * config/rs6000/aix72.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
6182 * config/rs6000/aix73.h (TARGET_AIX_VERSION): Increase to 73.
6184 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
6186 PR middle-end/102031
6187 * simplify-rtx.c (simplify_truncation): When comparing precisions
6188 use "subreg_prec" variable, not "subreg_mode".
6190 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
6192 * config/rs6000/rs6000-builtin-new.def: Add power10 and power10-64
6195 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
6197 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Initialize
6198 various pointer type nodes.
6199 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add enum
6200 values for various pointer types.
6201 (ptr_V16QI_type_node): New macro.
6202 (ptr_V1TI_type_node): New macro.
6203 (ptr_V2DI_type_node): New macro.
6204 (ptr_V2DF_type_node): New macro.
6205 (ptr_V4SI_type_node): New macro.
6206 (ptr_V4SF_type_node): New macro.
6207 (ptr_V8HI_type_node): New macro.
6208 (ptr_unsigned_V16QI_type_node): New macro.
6209 (ptr_unsigned_V1TI_type_node): New macro.
6210 (ptr_unsigned_V8HI_type_node): New macro.
6211 (ptr_unsigned_V4SI_type_node): New macro.
6212 (ptr_unsigned_V2DI_type_node): New macro.
6213 (ptr_bool_V16QI_type_node): New macro.
6214 (ptr_bool_V8HI_type_node): New macro.
6215 (ptr_bool_V4SI_type_node): New macro.
6216 (ptr_bool_V2DI_type_node): New macro.
6217 (ptr_bool_V1TI_type_node): New macro.
6218 (ptr_pixel_type_node): New macro.
6219 (ptr_intQI_type_node): New macro.
6220 (ptr_uintQI_type_node): New macro.
6221 (ptr_intHI_type_node): New macro.
6222 (ptr_uintHI_type_node): New macro.
6223 (ptr_intSI_type_node): New macro.
6224 (ptr_uintSI_type_node): New macro.
6225 (ptr_intDI_type_node): New macro.
6226 (ptr_uintDI_type_node): New macro.
6227 (ptr_intTI_type_node): New macro.
6228 (ptr_uintTI_type_node): New macro.
6229 (ptr_long_integer_type_node): New macro.
6230 (ptr_long_unsigned_type_node): New macro.
6231 (ptr_float_type_node): New macro.
6232 (ptr_double_type_node): New macro.
6233 (ptr_long_double_type_node): New macro.
6234 (ptr_dfloat64_type_node): New macro.
6235 (ptr_dfloat128_type_node): New macro.
6236 (ptr_ieee128_type_node): New macro.
6237 (ptr_ibm128_type_node): New macro.
6238 (ptr_vector_pair_type_node): New macro.
6239 (ptr_vector_quad_type_node): New macro.
6240 (ptr_long_long_integer_type_node): New macro.
6241 (ptr_long_long_unsigned_type_node): New macro.
6243 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
6245 * config/rs6000/rs6000-builtin-new.def: Add power9-vector, power9,
6246 and power9-64 stanzas.
6248 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
6249 Tom de Vries <tdevries@suse.de>
6251 * config.gcc (nvptx-*-*): Define {c,c++}_target_objs.
6252 * config/nvptx/nvptx-protos.h (nvptx_cpu_cpp_builtins): Prototype.
6253 * config/nvptx/nvptx.h (TARGET_CPU_CPP_BUILTINS): Implement with
6254 a call to the new nvptx_cpu_cpp_builtins function in nvptx-c.c.
6255 * config/nvptx/t-nvptx (nvptx-c.o): New rule.
6256 * config/nvptx/nvptx-c.c: New source file.
6257 (nvptx_cpu_cpp_builtins): Move implementation here.
6259 2021-08-24 Martin Sebor <msebor@redhat.com>
6261 PR middle-end/101600
6262 PR middle-end/101977
6263 * gimple-ssa-warn-access.cc (maybe_warn_for_bound): Tighten up
6264 the phrasing of a warning.
6265 (check_access): Use the remaining size after subtracting any offset
6266 rather than the whole object size.
6267 * pointer-query.cc (access_ref::get_ref): Clear BASE0 flag if it's
6268 clear for any nonnull PHI argument.
6269 (compute_objsize): Clear argument.
6271 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
6273 * config/rs6000/rs6000-builtin-new.def: Add power8-vector stanza.
6275 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
6277 * config/rs6000/rs6000-builtin-new.def: Add power7 and power7-64
6280 2021-08-24 Andrew MacLeod <amacleod@redhat.com>
6282 * value-relation.cc (rr_transitive_table): New.
6283 (relation_transitive): New.
6284 (value_relation::swap): Remove.
6285 (value_relation::apply_transitive): New.
6286 (relation_oracle::relation_oracle): Allocate a new tmp bitmap.
6287 (relation_oracle::register_relation): Call register_transitives.
6288 (relation_oracle::register_transitives): New.
6289 * value-relation.h (relation_oracle): Add new temporary bitmap and
6292 2021-08-24 H.J. Lu <hjl.tools@gmail.com>
6295 * config/i386/i386-expand.c (ix86_expand_vector_move): Broadcast
6296 from integer to a pseudo vector register.
6298 2021-08-24 Richard Biener <rguenther@suse.de>
6300 PR tree-optimization/100089
6301 * tree-vectorizer.h (vect_slp_bb): Rename to ...
6302 (vect_slp_if_converted_bb): ... this and get the original
6303 loop as new argument.
6304 * tree-vectorizer.c (try_vectorize_loop_1): Revert previous fix,
6305 pass original loop to vect_slp_if_converted_bb.
6306 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
6307 If orig_loop was passed scan the not vectorized stmts
6308 for COND_EXPRs and force not profitable if found.
6309 (vect_slp_region): Pass down all SLP instances to costing
6310 if orig_loop was specified.
6311 (vect_slp_bbs): Pass through orig_loop.
6312 (vect_slp_bb): Rename to ...
6313 (vect_slp_if_converted_bb): ... this and get the original
6314 loop as new argument.
6315 (vect_slp_function): Adjust.
6317 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
6320 * config/arm/arm.md (attribute arch): Add fix_vlldm.
6321 (arch_enabled): Use it.
6322 * config/arm/vfp.md (lazy_store_multiple_insn): Add alternative to
6323 use when erratum mitigation is needed.
6325 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
6328 * config/arm/arm.opt (mfix-cmse-cve-2021-35465): New option.
6329 * doc/invoke.texi (Arm Options): Document it.
6330 * config/arm/arm-cpus.in (quirk_vlldm): New feature bit.
6331 (ALL_QUIRKS): Add quirk_vlldm.
6332 (cortex-m33): Add quirk_vlldm.
6333 (cortex-m35p, cortex-m55): Likewise.
6334 * config/arm/arm.c (arm_option_override): Enable fix_vlldm if
6335 targetting an affected CPU and not explicitly controlled on
6338 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
6340 * config/arm/vfp.md (lazy_store_multiple_insn): Rewrite as valid RTL.
6341 (lazy_load_multiple_insn): Likewise.
6343 2021-08-24 liuhongt <hongtao.liu@intel.com>
6346 * config/i386/sse.md (<avx512>_vternlog<mode><sd_maskz_name>):
6347 Enable avx512 embedded broadcast.
6348 (*<avx512>_vternlog<mode>_all): Ditto.
6349 (<avx512>_vternlog<mode>_mask): Ditto.
6351 2021-08-24 liuhongt <hongtao.liu@intel.com>
6354 * config/i386/i386.c (ix86_rtx_costs): Define cost for
6356 * config/i386/i386.h (STRIP_UNARY): New macro.
6357 * config/i386/predicates.md (reg_or_notreg_operand): New
6359 * config/i386/sse.md (*<avx512>_vternlog<mode>_all): New define_insn.
6360 (*<avx512>_vternlog<mode>_1): New pre_reload
6361 define_insn_and_split.
6362 (*<avx512>_vternlog<mode>_2): Ditto.
6363 (*<avx512>_vternlog<mode>_3): Ditto.
6364 (any_logic1,any_logic2): New code iterator.
6365 (logic_op): New code attribute.
6366 (ternlogsuffix): Extend to VNxDF and VNxSF.
6368 2021-08-24 Richard Biener <rguenther@suse.de>
6370 * doc/invoke.texi (vect-inner-loop-cost-factor): Adjust.
6371 * params.opt (--param vect-inner-loop-cost-factor): Adjust
6373 * tree-vect-loop.c (vect_analyze_loop_form): Initialize
6374 inner_loop_cost_factor to the minimum of the estimated number
6375 of iterations of the inner loop and vect-inner-loop-cost-factor.
6377 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
6378 Richard Biener <rguenther@suse.de>
6380 * config/i386/i386-features.c (compute_convert_gain): Provide
6381 more accurate values for CONST_INT, when optimizing for size.
6382 * config/i386/i386.c (COSTS_N_BYTES): Move definition from here...
6383 * config/i386/i386.h (COSTS_N_BYTES): to here.
6385 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
6386 Jakub Jelinek <jakub@redhat.com>
6388 PR middle-end/102029
6389 * match.pd (shift transformations): Add an additional check for
6390 !POINTER_TYPE_P in the recently added left shift transformation.
6392 2021-08-24 liuhongt <hongtao.liu@intel.com>
6394 PR tree-optimization/100089
6395 * tree-vectorizer.c (try_vectorize_loop_1): Disable slp in
6396 loop vectorizer when cost model is very-cheap.
6398 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
6400 * config/rs6000/rs6000-gen-builtins.c (parse_bif_entry): Don't call
6401 asprintf, which is not available on AIX.
6403 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
6405 * config.gcc (target_gtfiles): Add ./rs6000-builtins.h.
6406 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Set.
6408 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
6410 * config.gcc (powerpc*-*-*): Add rs6000-builtins.o to extra_objs.
6411 * config/rs6000/rs6000-gen-builtins.c (main): Close init_file
6413 * config/rs6000/t-rs6000 (rs6000-gen-builtins.o): New target.
6414 (rbtree.o): Likewise.
6415 (rs6000-gen-builtins): Likewise.
6416 (rs6000-builtins.c): Likewise.
6417 (rs6000-builtins.h): Likewise.
6418 (rs6000.o): Add dependency.
6419 (EXTRA_HEADERS): Add rs6000-vecdefines.h.
6420 (rs6000-vecdefines.h): New target.
6421 (rs6000-builtins.o): Likewise.
6422 (rs6000-call.o): Add rs6000-builtins.h as a dependency.
6423 (rs6000-c.o): Likewise.
6425 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
6428 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace):
6429 Diagnose buffer overrun.
6430 (safe_inc_pos): Fix overrun detection.
6431 (match_identifier): Diagnose buffer overrun.
6432 (match_integer): Likewise.
6433 (match_to_right_bracket): Likewise.
6435 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
6437 * ipa-modref-tree.h (modref_access_node::range_info_useful_p):
6438 Improve range compare.
6439 (modref_access_node::contains): New member function.
6440 (modref_access_node::search): Remove.
6441 (modref_access_node::insert): Be smarter about subaccesses.
6443 2021-08-23 Thomas Schwinge <thomas@codesourcery.com>
6445 * config/i386/i386-options.c (ix86_omp_device_kind_arch_isa)
6446 <omp_device_arch> [ACCEL_COMPILER]: Match "intel_mic".
6447 * config/i386/t-omp-device (omp-device-properties-i386) <arch>:
6450 2021-08-23 Jeff Law <jlaw@localhost.localdomain>
6452 * config/h8300/h8300-protos.h (h8300_expand_epilogue): Add new
6454 * config/h8300/jumpcall.md (call, call_value): Restrict to
6455 !SIBLING_CALL_P cases.
6456 (subcall, sibcall_value): New patterns & expanders.
6457 * config/h8300/proepi.md (epilogue): Pass new argument to
6458 h8300_expand_epilogue.
6459 (sibcall_epilogue): New expander.
6460 * config/h8300/h8300.c (h8300_expand_epilogue): Handle sibcall
6462 (h8300_ok_for_sibcall_p): New function.
6463 (TARGET_FUNCTION_OK_FOR_SIBCALL): define.
6465 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
6467 * simplify-rtx.c (simplify_unary_operation_1): [TRUNCATE]:
6468 Handle case where the operand is already the desired mode.
6470 2021-08-23 Richard Biener <rguenther@suse.de>
6473 * tree-ssa-structalias.c (ipa_pta_execute): Check in_other_partition
6474 in addition to has_gimple_body.
6476 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
6478 PR middle-end/101949
6479 * ipa-modref.c (analyze_ssa_name_flags): Fix merging of
6482 2021-08-23 Martin Liska <mliska@suse.cz>
6484 * doc/invoke.texi: Put the option out of -mxl-mode-app-model
6487 2021-08-23 Richard Biener <rguenther@suse.de>
6489 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6490 Properly scale the inner loop cost only once.
6492 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
6494 * tree-ssa-ccp.c (bit_value_binop) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]:
6495 Provide bounds for unsigned (and signed with non-negative operands)
6496 division and modulus.
6498 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
6500 * simplify-rtx.c (simplify_truncation): Generalize simplification
6501 of (truncate:A (subreg:B X)).
6502 (simplify_unary_operation_1) [FLOAT_TRUNCATE, FLOAT_EXTEND,
6503 SIGN_EXTEND, ZERO_EXTEND]: Handle cases where the operand
6504 already has the desired machine mode.
6505 (test_scalar_int_ops): Add tests that useless extensions and
6506 truncations are optimized away.
6507 (test_scalar_int_ext_ops): New self-test function to confirm
6508 that truncations of extensions are correctly simplified.
6509 (test_scalar_int_ext_ops2): New self-test function to check
6510 truncations of truncations, extensions of extensions, and
6511 truncations of extensions.
6512 (test_scalar_ops): Call the above two functions with a
6513 representative sampling of integer machine modes.
6515 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
6517 * match.pd (shift transformations): Change the sign of an
6518 LSHIFT_EXPR if it reduces the number of explicit conversions.
6520 2021-08-23 Jakub Jelinek <jakub@redhat.com>
6522 PR tree-optimization/86723
6523 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
6524 cast64_to_32 argument, set *cast64_to_32 to false, unless n is
6525 non-memory permutation of 64-bit src which only has bytes of
6526 0 or [5..8] and n->range is 4.
6527 (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
6528 find_bswap_or_nop_finalize caller, support bswap with some bytes
6529 zeroed, as long as at least two bytes are not zeroed.
6530 (bswap_replace): Add mask argument and handle masking of bswap
6532 (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
6533 caller, punt if cast64_to_32 or mask is not all ones.
6534 (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
6535 caller, for now punt if cast64_to_32.
6537 2021-08-23 Richard Biener <rguenther@suse.de>
6539 PR tree-optimization/79334
6540 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Record
6541 a type also for COMPONENT_REFs.
6542 (vn_reference_may_trap): Check ARRAY_REF with constant index
6543 against the array domain.
6545 2021-08-23 liuhongt <hongtao.liu@intel.com>
6548 * config/i386/sse.md (*avx512f_pshufb_truncv8hiv8qi_1): Add
6549 TARGET_AVX512BW to condition.
6551 2021-08-23 Jakub Jelinek <jakub@redhat.com>
6554 * dwarf2out.c (gen_variable_die): Add DW_AT_location for global
6555 register variables already during early_dwarf if possible.
6557 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
6559 * config/arm/arm_mve.h: Fix __arm_vctp16q return type.
6561 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
6564 * config/arm/arm.opt: Fix typo.
6565 * config/arm/t-rmprofile: Fix typo.
6567 2021-08-23 Jakub Jelinek <jakub@redhat.com>
6569 * tree.h (OMP_CLAUSE_GRAINSIZE_STRICT): Define.
6570 (OMP_CLAUSE_NUM_TASKS_STRICT): Define.
6571 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_GRAINSIZE,
6572 case OMP_CLAUSE_NUM_TASKS>: Print strict: modifier.
6573 * omp-expand.c (expand_task_call): Use GOMP_TASK_FLAG_STRICT in iflags
6574 if either grainsize or num_tasks clause has the strict modifier.
6576 2021-08-23 Martin Liska <mliska@suse.cz>
6578 * dbgcnt.def (DEBUG_COUNTER): New counter.
6579 * gimple.c (gimple_call_arg_flags): Use it in IPA PTA.
6581 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
6583 * ipa-modref.c (analyze_ssa_name_flags): Improve handling of return slot.
6585 2021-08-23 Xi Ruoyao <xry111@mengyan1223.wang>
6588 * config/mips/mips-protos.h (mips_msa_output_shift_immediate):
6590 * config/mips/mips.c (mips_msa_output_shift_immediate): New
6592 * config/mips/mips-msa.md (vashl<mode>3, vashr<mode>3,
6593 vlshr<mode>3): Call it.
6595 2021-08-22 Jan Hubicka <hubicka@ucw.cz>
6596 Martin Liska <mliska@suse.cz>
6598 PR middle-end/101949
6599 * ipa-modref.c (analyze_ssa_name_flags): Indirect call implies
6602 2021-08-21 Dragan Mladjenovic <OT_Dragan.Mladjenovic@mediatek.com>
6604 * config/mips/mips.c (mips_function_rodata_section,
6605 TARGET_ASM_FUNCTION_RODATA_SECTION): Removed.
6607 2021-08-21 John David Anglin <danglin@gcc.gnu.org>
6609 * config/pa/pa.c (pa_asm_output_aligned_common): Remove warning.
6611 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
6613 * configure.ac (thread-local storage support): Remove tls_first_major
6614 and tls_first_minor. Use "$conftest_s" to check support.
6615 * configure: Regenerate.
6617 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
6619 * configure.ac: Fixup formatting.
6621 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
6623 * acinclude.m4 (gcc_GAS_CHECK_FEATURE): Remove third argument and ...
6624 * configure.ac: ... update all callers.
6626 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
6629 * acinclude.m4 (_gcc_COMPUTE_GAS_VERSION, _gcc_GAS_VERSION_GTE_IFELSE)
6630 (gcc_GAS_VERSION_GTE_IFELSE): Remove.
6631 (gcc_GAS_CHECK_FEATURE): Do not handle in-tree case specially.
6632 * configure.ac: Remove gcc_cv_gas_major_version, gcc_cv_gas_minor_version.
6633 Remove remaining checks for in-tree assembler.
6634 * configure: Regenerate.
6636 2021-08-20 Jeff Law <jlaw@localhost.localdomain>
6638 * config/h8300/h8300.c (shift_alg_hi): Improve arithmetic shift right
6639 by 15 bits for H8/300H and H8/S. Improve logical shifts by 12
6641 (shift_alg_si): Improve arithmetic right shift by 28-30 bits for
6642 H8/300H. Improve arithmetic shift right by 15 bits for H8/S.
6643 Improve logical shifts by 27 bits for H8/S.
6644 (get_shift_alg): Corresponding changes.
6645 (h8300_option_override): Revert to loops for -Os when profitable.
6647 2021-08-20 Richard Biener <rguenther@suse.de>
6649 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
6651 (vect_analyze_data_ref_accesses): Likewise. Assign the BB
6652 index as group_id when dataref_groups were not computed.
6653 * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
6654 we advace to the next BB.
6656 2021-08-20 Jakub Jelinek <jakub@redhat.com>
6658 * omp-builtins.def (BUILT_IN_GOMP_WARNING, BUILT_IN_GOMP_ERROR): New
6661 2021-08-20 Martin Liska <mliska@suse.cz>
6663 PR gcov-profile/89961
6664 * gcov.c (make_gcov_file_name): Rewrite using std::string.
6665 (mangle_name): Simplify, do not used the second argument.
6666 (strip_extention): New function.
6667 (get_md5sum): Likewise.
6668 (get_gcov_intermediate_filename): Handle properly -p and -x
6670 (output_gcov_file): Use string type.
6671 (generate_results): Likewise.
6672 (md5sum_to_hex): Remove.
6674 2021-08-20 Michael Meissner <meissner@linux.ibm.com>
6676 * config/rs6000/altivec.md (UNSPEC_XXEVAL): Move to vsx.md.
6677 (UNSPEC_XXSPLTIW): Move to vsx.md.
6678 (UNSPEC_XXSPLTID): Move to vsx.md.
6679 (UNSPEC_XXSPLTI32DX): Move to vsx.md.
6680 (UNSPEC_XXBLEND): Move to vsx.md.
6681 (UNSPEC_XXPERMX): Move to vsx.md.
6682 (VM3): Move to vsx.md.
6683 (VM3_char): Move to vsx.md.
6684 (xxspltiw_v4si): Move to vsx.md.
6685 (xxspltiw_v4sf): Move to vsx.md.
6686 (xxspltiw_v4sf_inst): Move to vsx.md.
6687 (xxspltidp_v2df): Move to vsx.md.
6688 (xxspltidp_v2df_inst): Move to vsx.md.
6689 (xxsplti32dx_v4si_inst): Move to vsx.md.
6690 (xxsplti32dx_v4sf): Move to vsx.md.
6691 (xxsplti32dx_v4sf_inst): Move to vsx.md.
6692 (xxblend_<mode>): Move to vsx.md.
6693 (xxpermx): Move to vsx.md.
6694 (xxpermx_inst): Move to vsx.md.
6695 * config/rs6000/vsx.md (UNSPEC_XXEVAL): Move from altivec.md.
6696 (UNSPEC_XXSPLTIW): Move from altivec.md.
6697 (UNSPEC_XXSPLTID): Move from altivec.md.
6698 (UNSPEC_XXSPLTI32DX): Move from altivec.md.
6699 (UNSPEC_XXBLEND): Move from altivec.md.
6700 (UNSPEC_XXPERMX): Move from altivec.md.
6701 (VM3): Move from altivec.md.
6702 (VM3_char): Move from altivec.md.
6703 (xxspltiw_v4si): Move from altivec.md.
6704 (xxspltiw_v4sf): Move from altivec.md.
6705 (xxspltiw_v4sf_inst): Move from altivec.md.
6706 (xxspltidp_v2df): Move from altivec.md.
6707 (xxspltidp_v2df_inst): Move from altivec.md.
6708 (xxsplti32dx_v4si_inst): Move from altivec.md.
6709 (xxsplti32dx_v4sf): Move from altivec.md.
6710 (xxsplti32dx_v4sf_inst): Move from altivec.md.
6711 (xxblend_<mode>): Move from altivec.md.
6712 (xxpermx): Move from altivec.md.
6713 (xxpermx_inst): Move from altivec.md.
6715 2021-08-19 Roger Sayle <roger@nextmovesoftware.com>
6717 * tree-vect-generic.c (expand_vector_operations_1): Use either
6718 gimplify_build1 or gimplify_build2 instead of gimple_build_assign
6719 when constructing scalar splat expressions.
6721 2021-08-19 Peter Bergner <bergner@linux.ibm.com>
6724 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Cast
6725 pointer to __vector_pair *.
6727 2021-08-19 Martin Sebor <msebor@redhat.com>
6729 * gimple-range.cc: Add comments.
6730 * gimple-range.h: Same.
6732 2021-08-19 Martin Sebor <msebor@redhat.com>
6734 PR middle-end/101984
6735 * gimple-ssa-warn-access.cc (pass_waccess::execute): Also call
6738 2021-08-19 Jeff Law <jlaw@localhost.localdomain>
6740 * config.gcc (h8300-*-elf*): Do not include dbxelf.h.
6741 (h8300-*-linux*, v850-*-rtems*, v850*-elf*): Likewise.
6742 * config/v850/v850.h (DEFAULT_GDB_EXTENSIONS): Remove.
6744 2021-08-19 Jakub Jelinek <jakub@redhat.com>
6746 PR middle-end/101950
6747 * optabs.c (expand_clrsb_using_clz): New function.
6748 (expand_unop): Use it as another clrsb expansion fallback.
6750 2021-08-19 liuhongt <hongtao.liu@intel.com>
6753 2021-07-28 liuhongt <hongtao.liu@intel.com>
6756 * config/i386/i386.h (processor_costs): Add new member
6758 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
6759 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
6760 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
6761 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
6762 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
6763 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
6764 generic_cost, core_cost): Initialize integer_to_sse same value
6766 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
6767 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6768 Use integer_to_sse instead of sse_op to calculate the cost of
6771 2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
6773 * config.gcc: Include rpath.opt for Darwin.
6774 * config/darwin.h (DRIVER_SELF_SPECS): Handle -rpath.
6776 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
6779 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
6782 2021-08-18 Jonathan Wright <jonathan.wright@arm.com>
6784 * config/aarch64/arm_neon.h (vld3_lane_f64): Use float RTL
6785 pattern and type cast.
6786 (vld4_lane_f32): Use float RTL pattern.
6787 (vld4q_lane_f64): Use float type cast.
6789 2021-08-18 Jan Hubicka <hubicka@ucw.cz>
6791 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Check also
6794 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
6796 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor): Extend.
6797 (test_map_of_type_with_ctor_and_dtor_expand): Add function.
6798 (hash_map_tests_c_tests): Call it.
6800 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
6802 * ggc.h (enum ggc_collect): New.
6803 (ggc_collect): Use it.
6804 * ggc-page.c: Adjust.
6805 * ggc-common.c: Likewise.
6806 * ggc-tests.c: Likewise.
6807 * read-rtl-function.c: Likewise.
6808 * selftest-run-tests.c: Likewise.
6809 * doc/gty.texi (Invoking the garbage collector): Likewise.
6811 2021-08-18 liuhongt <hongtao.liu@intel.com>
6814 * config/i386/i386.h (TARGET_V2DF_REDUCTION_PREFER_HADDPD):
6816 * config/i386/sse.md (*sse3_haddv2df3_low): Add
6817 TARGET_V2DF_REDUCTION_PREFER_HADDPD.
6818 (*sse3_hsubv2df3_low): Ditto.
6819 * config/i386/x86-tune.def
6820 (X86_TUNE_V2DF_REDUCTION_PREFER_HADDPD): New tune.
6822 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
6824 * gimple-range-gori.cc (gori_compute::gori_compute): Enable tracing.
6825 (gori_compute::compute_operand_range): Add tracing.
6826 (gori_compute::logical_combine): Ditto.
6827 (gori_compute::compute_logical_operands): Ditto.
6828 (gori_compute::compute_operand1_range): Ditto.
6829 (gori_compute::compute_operand2_range): Ditto.
6830 (gori_compute::outgoing_edge_range_p): Ditto.
6831 * gimple-range-gori.h (class gori_compute): Add range_tracer.
6833 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
6835 * flag-types.h (enum evrp_mode): Adjust evrp-mode values.
6836 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Relocate from.
6837 * gimple-range-trace.h (DEBUG_RANGE_CACHE): Here.
6838 * params.opt (--param=evrp-mode): Adjust options.
6840 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
6842 * Makefile.in (OBJS): Add gimple-range-trace.o.
6843 * gimple-range-cache.h (enable_new_values): Remove unused prototype.
6844 * gimple-range-fold.cc: Adjust headers.
6845 * gimple-range-trace.cc: New.
6846 * gimple-range-trace.h: New.
6847 * gimple-range.cc (gimple_ranger::gimple_ranger): Enable tracer.
6848 (gimple_ranger::range_of_expr): Add tracing.
6849 (gimple_ranger::range_on_entry): Ditto.
6850 (gimple_ranger::range_on_exit): Ditto.
6851 (gimple_ranger::range_on_edge): Ditto.
6852 (gimple_ranger::fold_range_internal): Ditto.
6853 (gimple_ranger::dump_bb): Do not calculate edge range twice.
6854 (trace_ranger::*): Remove.
6855 (enable_ranger): Never create a trace_ranger.
6856 (debug_seed_ranger): Move to gimple-range-trace.cc.
6857 (dump_ranger): Ditto.
6858 (debug_ranger): Ditto.
6859 * gimple-range.h: Include gimple-range-trace.h.
6860 (range_on_entry, range_on_exit): No longer virtual.
6861 (class trace_ranger): Remove.
6862 (DEBUG_RANGE_CACHE): Move to gimple-range-trace.h.
6864 2021-08-17 Martin Sebor <msebor@redhat.com>
6866 PR middle-end/101854
6867 * builtins.c (expand_builtin_alloca): Move warning code to check_alloca
6868 in gimple-ssa-warn-access.cc.
6869 * calls.c (alloc_max_size): Move code to check_alloca.
6870 (get_size_range): Move to pointer-query.cc.
6871 (maybe_warn_alloc_args_overflow): Move to gimple-ssa-warn-access.cc.
6872 (get_attr_nonstring_decl): Move to tree.c.
6873 (fntype_argno_type): Move to gimple-ssa-warn-access.cc.
6874 (append_attrname): Same.
6875 (maybe_warn_rdwr_sizes): Same.
6876 (initialize_argument_information): Move code to
6877 gimple-ssa-warn-access.cc.
6878 * calls.h (maybe_warn_alloc_args_overflow): Move to
6879 gimple-ssa-warn-access.h.
6880 (get_attr_nonstring_decl): Move to tree.h.
6881 (maybe_warn_nonstring_arg): Move to gimple-ssa-warn-access.h.
6882 (enum size_range_flags): Move to pointer-query.h.
6883 (get_size_range): Same.
6884 * gimple-ssa-warn-access.cc (has_location): Remove unused overload
6885 to avoid Clang -Wunused-function.
6886 (get_size_range): Declare static.
6887 (maybe_emit_free_warning): Rename...
6888 (maybe_check_dealloc_call): ...to this for consistency.
6889 (class pass_waccess): Add members.
6890 (pass_waccess::~pass_waccess): Defined.
6891 (alloc_max_size): Move here from calls.c.
6892 (maybe_warn_alloc_args_overflow): Same.
6893 (check_alloca): New function.
6894 (check_alloc_size_call): New function.
6895 (check_strncat): Handle another warning flag.
6896 (pass_waccess::check_builtin): Handle alloca.
6897 (fntype_argno_type): Move here from calls.c.
6898 (append_attrname): Same.
6899 (maybe_warn_rdwr_sizes): Same.
6900 (pass_waccess::check_call): Define.
6901 (check_nonstring_args): New function.
6902 (pass_waccess::check): Call new member functions.
6903 (pass_waccess::execute): Enable ranger.
6904 * gimple-ssa-warn-access.h (get_size_range): Move here from calls.h.
6905 (maybe_warn_nonstring_arg): Same.
6906 * gimple-ssa-warn-restrict.c: Remove #include.
6907 * pointer-query.cc (get_size_range): Move here from calls.c.
6908 * pointer-query.h (enum size_range_flags): Same.
6909 (get_size_range): Same.
6910 * tree.c (get_attr_nonstring_decl): Move here from calls.c.
6911 * tree.h (get_attr_nonstring_decl): Move here from calls.h.
6913 2021-08-17 Thomas Schwinge <thomas@codesourcery.com>
6915 * ggc.h (ggc_collect): Add 'force_collect' parameter.
6916 * ggc-page.c (ggc_collect): Use that one instead of global
6917 'ggc_force_collect'. Adjust all users.
6918 * doc/gty.texi (Invoking the garbage collector): Update.
6919 * ggc-internal.h (ggc_force_collect): Remove.
6920 * ggc-common.c (ggc_force_collect): Likewise.
6921 * selftest.h (forcibly_ggc_collect): Remove.
6922 * ggc-tests.c (selftest::forcibly_ggc_collect): Likewise.
6923 * read-rtl-function.c (test_loading_labels): Adjust.
6924 * selftest-run-tests.c (run_tests): Likewise.
6926 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
6928 * config/darwin.c (darwin_file_end): Reset and reclaim the
6929 section names table at the end of compile.
6931 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
6934 * config.in: Regenerate.
6935 * config/i386/darwin.h (EXTRA_ASM_OPTS): New
6936 (ASM_SPEC): Pass options to disable branch shortening where
6938 * configure: Regenerate.
6939 * configure.ac: Detect versions of 'as' that support the
6940 optimisation which has the bug.
6942 2021-08-17 Richard Biener <rguenther@suse.de>
6944 * optabs-query.c (supports_vec_gather_load_p): Also check
6946 (supports_vec_scatter_store_p): Likewise.
6947 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Fall
6948 back to masked variants if non-masked are not supported.
6949 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
6950 When we need to use masked gather/scatter but do not have
6951 a mask set up a constant true one.
6952 * tree-vect-stmts.c (vect_check_scalar_mask): Also allow
6955 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
6957 * tree-ssa-ccp.c (bit_value_binop) [MINUS_EXPR]: Use same
6958 algorithm as PLUS_EXPR to improve subtraction bit bounds.
6959 [POINTER_DIFF_EXPR]: Treat as synonymous with MINUS_EXPR.
6961 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
6963 * tree-ssa-ccp.c (bit_value_mult_const): New helper function to
6964 calculate the mask-value pair result of a multiplication by an
6966 (bit_value_binop) [MULT_EXPR]: Call it from here for
6967 multiplications by (sparse) non-negative constants.
6969 2021-08-17 Christophe Lyon <christophe.lyon@foss.st.com>
6972 * config.gcc (gcc_cv_initfini_array): Leave undefined for
6973 uclinuxfdpiceabi targets.
6975 2021-08-17 Alexandre Oliva <oliva@adacore.com>
6977 * tree-inline.c (maybe_move_debug_stmts_to_successors): Don't
6978 reverse debug stmts.
6980 2021-08-17 Alexandre Oliva <oliva@adacore.com>
6982 * tree-cfg.c (dump_function_to_file): Use fun, not cfun.
6984 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
6986 * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete.
6987 (__LD4Q_LANE_FUNC): Likewise.
6988 (vld4_lane_u8): Define without macro.
6989 (vld4_lane_u16): Likewise.
6990 (vld4_lane_u32): Likewise.
6991 (vld4_lane_u64): Likewise.
6992 (vld4_lane_s8): Likewise.
6993 (vld4_lane_s16): Likewise.
6994 (vld4_lane_s32): Likewise.
6995 (vld4_lane_s64): Likewise.
6996 (vld4_lane_f16): Likewise.
6997 (vld4_lane_f32): Likewise.
6998 (vld4_lane_f64): Likewise.
6999 (vld4_lane_p8): Likewise.
7000 (vld4_lane_p16): Likewise.
7001 (vld4_lane_p64): Likewise.
7002 (vld4q_lane_u8): Likewise.
7003 (vld4q_lane_u16): Likewise.
7004 (vld4q_lane_u32): Likewise.
7005 (vld4q_lane_u64): Likewise.
7006 (vld4q_lane_s8): Likewise.
7007 (vld4q_lane_s16): Likewise.
7008 (vld4q_lane_s32): Likewise.
7009 (vld4q_lane_s64): Likewise.
7010 (vld4q_lane_f16): Likewise.
7011 (vld4q_lane_f32): Likewise.
7012 (vld4q_lane_f64): Likewise.
7013 (vld4q_lane_p8): Likewise.
7014 (vld4q_lane_p16): Likewise.
7015 (vld4q_lane_p64): Likewise.
7016 (vld4_lane_bf16): Likewise.
7017 (vld4q_lane_bf16): Likewise.
7019 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
7021 * config/aarch64/arm_neon.h (__LD3_LANE_FUNC): Delete.
7022 (__LD3Q_LANE_FUNC): Delete.
7023 (vld3_lane_u8): Define without macro.
7024 (vld3_lane_u16): Likewise.
7025 (vld3_lane_u32): Likewise.
7026 (vld3_lane_u64): Likewise.
7027 (vld3_lane_s8): Likewise.
7028 (vld3_lane_s16): Likewise.
7029 (vld3_lane_s32): Likewise.
7030 (vld3_lane_s64): Likewise.
7031 (vld3_lane_f16): Likewise.
7032 (vld3_lane_f32): Likewise.
7033 (vld3_lane_f64): Likewise.
7034 (vld3_lane_p8): Likewise.
7035 (vld3_lane_p16): Likewise.
7036 (vld3_lane_p64): Likewise.
7037 (vld3q_lane_u8): Likewise.
7038 (vld3q_lane_u16): Likewise.
7039 (vld3q_lane_u32): Likewise.
7040 (vld3q_lane_u64): Likewise.
7041 (vld3q_lane_s8): Likewise.
7042 (vld3q_lane_s16): Likewise.
7043 (vld3q_lane_s32): Likewise.
7044 (vld3q_lane_s64): Likewise.
7045 (vld3q_lane_f16): Likewise.
7046 (vld3q_lane_f32): Likewise.
7047 (vld3q_lane_f64): Likewise.
7048 (vld3q_lane_p8): Likewise.
7049 (vld3q_lane_p16): Likewise.
7050 (vld3q_lane_p64): Likewise.
7051 (vld3_lane_bf16): Likewise.
7052 (vld3q_lane_bf16): Likewise.
7054 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
7056 * config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Delete.
7057 (__LD2Q_LANE_FUNC): Likewise.
7058 (vld2_lane_u8): Define without macro.
7059 (vld2_lane_u16): Likewise.
7060 (vld2_lane_u32): Likewise.
7061 (vld2_lane_u64): Likewise.
7062 (vld2_lane_s8): Likewise.
7063 (vld2_lane_s16): Likewise.
7064 (vld2_lane_s32): Likewise.
7065 (vld2_lane_s64): Likewise.
7066 (vld2_lane_f16): Likewise.
7067 (vld2_lane_f32): Likewise.
7068 (vld2_lane_f64): Likewise.
7069 (vld2_lane_p8): Likewise.
7070 (vld2_lane_p16): Likewise.
7071 (vld2_lane_p64): Likewise.
7072 (vld2q_lane_u8): Likewise.
7073 (vld2q_lane_u16): Likewise.
7074 (vld2q_lane_u32): Likewise.
7075 (vld2q_lane_u64): Likewise.
7076 (vld2q_lane_s8): Likewise.
7077 (vld2q_lane_s16): Likewise.
7078 (vld2q_lane_s32): Likewise.
7079 (vld2q_lane_s64): Likewise.
7080 (vld2q_lane_f16): Likewise.
7081 (vld2q_lane_f32): Likewise.
7082 (vld2q_lane_f64): Likewise.
7083 (vld2q_lane_p8): Likewise.
7084 (vld2q_lane_p16): Likewise.
7085 (vld2q_lane_p64): Likewise.
7086 (vld2_lane_bf16): Likewise.
7087 (vld2q_lane_bf16): Likewise.
7089 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
7091 * haifa-sched.c (advance_one_cycle): Output more context-synchronization
7094 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
7096 * haifa-sched.c (enum rfs_decision, rfs_str): Add RFS_AUTOPREF.
7097 (rank_for_schedule): Use it.
7099 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
7101 PR rtl-optimization/91598
7102 * haifa-sched.c (autopref_rank_for_schedule): Prioritize "irrelevant"
7103 insns after memory reads and before memory writes.
7105 2021-08-17 Alistair_Lee <alistair.lee@arm.com>
7107 * rtl.h (CONST_VECTOR_P): New macro.
7108 * config/aarch64/aarch64.c (aarch64_get_sve_pred_bits): Use RTL
7109 code testing macros.
7110 (aarch64_ptrue_all_mode): Likewise.
7111 (aarch64_expand_mov_immediate): Likewise.
7112 (aarch64_const_vec_all_in_range_p): Likewise.
7113 (aarch64_rtx_costs): Likewise.
7114 (aarch64_legitimate_constant_p): Likewise.
7115 (aarch64_simd_valid_immediate): Likewise.
7116 (aarch64_simd_make_constant): Likewise.
7117 (aarch64_convert_mult_to_shift): Likewise.
7118 (aarch64_expand_sve_vec_perm): Likewise.
7119 (aarch64_vec_fpconst_pow_of_2): Likewise.
7121 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
7123 PR tree-optimization/101938
7124 * range-op.cc (operator_abs::op1_range): Special case
7125 -TYPE_MIN_VALUE for flag_wrapv.
7127 2021-08-17 Kewen Lin <linkw@linux.ibm.com>
7129 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Add the cost for
7132 2021-08-17 Jakub Jelinek <jakub@redhat.com>
7134 * tree.def (OMP_SCOPE): New tree code.
7135 * tree.h (OMP_SCOPE_BODY, OMP_SCOPE_CLAUSES): Define.
7136 * tree-nested.c (convert_nonlocal_reference_stmt,
7137 convert_local_reference_stmt, convert_gimple_call): Handle
7139 * tree-pretty-print.c (dump_generic_node): Handle OMP_SCOPE.
7140 * gimple.def (GIMPLE_OMP_SCOPE): New gimple code.
7141 * gimple.c (gimple_build_omp_scope): New function.
7142 (gimple_copy): Handle GIMPLE_OMP_SCOPE.
7143 * gimple.h (gimple_build_omp_scope): Declare.
7144 (gimple_has_substatements): Handle GIMPLE_OMP_SCOPE.
7145 (gimple_omp_scope_clauses, gimple_omp_scope_clauses_ptr,
7146 gimple_omp_scope_set_clauses): New inline functions.
7147 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_SCOPE.
7148 * gimple-pretty-print.c (dump_gimple_omp_scope): New function.
7149 (pp_gimple_stmt_1): Handle GIMPLE_OMP_SCOPE.
7150 * gimple-walk.c (walk_gimple_stmt): Likewise.
7151 * gimple-low.c (lower_stmt): Likewise.
7152 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
7153 (gimplify_scan_omp_clauses): For task reductions, handle OMP_SCOPE
7154 like ORT_WORKSHARE constructs. Adjust diagnostics for %<scope%>
7155 allowing task reductions. Reject inscan reductions on scope.
7156 (omp_find_stores_stmt): Handle GIMPLE_OMP_SCOPE.
7157 (gimplify_omp_workshare, gimplify_expr): Handle OMP_SCOPE.
7158 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_SCOPE.
7159 (estimate_num_insns): Likewise.
7160 * omp-low.c (build_outer_var_ref): Look through GIMPLE_OMP_SCOPE
7161 contexts if var isn't privatized there.
7162 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_SCOPE.
7163 (scan_omp_1_stmt): Likewise.
7164 (maybe_add_implicit_barrier_cancel): Look through outer
7166 (lower_omp_scope): New function.
7167 (lower_omp_task_reductions): Handle OMP_SCOPE.
7168 (lower_omp_1): Handle GIMPLE_OMP_SCOPE.
7169 (diagnose_sb_1, diagnose_sb_2): Likewise.
7170 * omp-expand.c (expand_omp_single): Support also GIMPLE_OMP_SCOPE.
7171 (expand_omp): Handle GIMPLE_OMP_SCOPE.
7172 (omp_make_gimple_edges): Likewise.
7173 * omp-builtins.def (BUILT_IN_GOMP_SCOPE_START): New built-in.
7175 2021-08-17 Richard Biener <rguenther@suse.de>
7177 PR tree-optimization/101925
7178 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Set
7179 reverse on COMPONENT_REF and ARRAY_REF according to
7180 what reverse_storage_order_for_component_p does.
7181 (vn_reference_eq): Compare reversed on reference ops.
7182 (reverse_storage_order_for_component_p): New overload.
7183 (vn_reference_lookup_3): Check reverse_storage_order_for_component_p
7184 on the reference looked up.
7186 2021-08-17 Jeff Law <jlaw@localhost.localdomain>
7188 * config/h8300/h8300.c (shift_alg_si): Avoid loops for most SImode
7190 (h8300_option_override): Use loops on H8/S more often when optimizing
7192 (get_shift_alg): Handle new "special" cases on H8/S. Simplify
7193 accordingly. Handle various arithmetic right shifts with special
7194 sequences that we couldn't handle before.
7196 2021-08-16 Jeff Law <jlaw@localhost.localdomain>
7198 * config.gcc (rl78-*-elf*): Do not include dbxelf.h.
7200 2021-08-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
7202 * config/sparc/rtemself.h (SPARC_GCOV_TYPE_SIZE): Define.
7203 * config/sparc/sparc.c (sparc_gcov_type_size): New.
7204 (TARGET_GCOV_TYPE_SIZE): Redefine if SPARC_GCOV_TYPE_SIZE is defined.
7205 * coverage.c (get_gcov_type): Use targetm.gcov_type_size().
7206 * doc/tm.texi (TARGET_GCOV_TYPE_SIZE): Add hook under "Misc".
7207 * doc/tm.texi.in: Regenerate.
7208 * target.def (gcov_type_size): New target hook.
7209 * targhooks.c (default_gcov_type_size): New.
7210 * targhooks.h (default_gcov_type_size): Declare.
7211 * tree-profile.c (gimple_gen_edge_profiler): Use precision of
7213 (gimple_gen_time_profiler): Likewise.
7215 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
7217 * dwarf2out.c (add_scalar_info): Deal with DW_AT_data_bit_offset.
7219 2021-08-16 Tobias Burnus <tobias@codesourcery.com>
7221 PR middle-end/101931
7222 * omp-low.c (omp_runtime_api_call): Update for routines
7223 added in the meanwhile.
7225 2021-08-16 Martin Liska <mliska@suse.cz>
7227 PR tree-optimization/100393
7228 * tree-switch-conversion.c (group_cluster::dump): Use
7229 get_comparison_count.
7230 (jump_table_cluster::find_jump_tables): Pre-compute number of
7231 comparisons and then decrement it. Cache also max_ratio.
7232 (jump_table_cluster::can_be_handled): Change signature.
7233 * tree-switch-conversion.h (get_comparison_count): New.
7235 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
7237 * dwarf2out.c (add_data_member_location_attribute): Use GNAT
7238 encodings only when -fgnat-encodings=all is specified.
7239 (add_bound_info): Likewise.
7240 (add_byte_size_attribute): Likewise.
7241 (gen_member_die): Likewise.
7243 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
7245 * omp-oacc-neuter-broadcast.cc
7246 (execute_omp_oacc_neuter_broadcast): Plug 'par' memory leak.
7248 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
7250 * omp-oacc-neuter-broadcast.cc
7251 (execute_omp_oacc_neuter_broadcast): Clarify memory management for
7254 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
7256 * omp-oacc-neuter-broadcast.cc (field_map): Move variable into...
7257 (execute_omp_oacc_neuter_broadcast): ... here.
7258 (install_var_field, build_receiver_ref, build_sender_ref): Take
7259 'field_map_t *' parameter. Adjust all users.
7260 (worker_single_copy, neuter_worker_single): Take a
7261 'record_field_map_t *' parameter. Adjust all users.
7263 2021-08-16 liuhongt <hongtao.liu@intel.com>
7266 * config/i386/i386.md (ldexp<mode>3): Force operands[1] to
7269 2021-08-16 Martin Liska <mliska@suse.cz>
7272 * multiple_target.c (create_dispatcher_calls): Make default
7273 function local only if it is a definition.
7275 2021-08-16 Martin Liska <mliska@suse.cz>
7278 * ipa-icf-gimple.c (func_checker::compare_ssa_name): Do not
7279 consider equal SSA_NAMEs when one is a param.
7281 2021-08-16 liuhongt <hongtao.liu@intel.com>
7284 * config/i386/i386-expand.c (ix86_expand_vec_perm_vpermt2):
7285 Support vpermi2b for V32QI/V16QImode.
7286 (ix86_extract_perm_from_pool_constant): New function.
7287 (ix86_expand_vec_one_operand_perm_avx512): Support
7288 vpermw/vpermb under TARGET_AVX512BW/TARGET_AVX512VBMI.
7289 (expand_vec_perm_1): Adjust comments for upper.
7290 * config/i386/i386-protos.h (ix86_extract_perm_from_pool_constant):
7292 * config/i386/predicates.md (permvar_truncate_operand): New predicate.
7293 (pshufb_truncv4siv4hi_operand): Ditto.
7294 (pshufb_truncv8hiv8qi_operand): Ditto.
7295 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1):
7296 New pre_reload define_insn_and_split.
7297 (*avx512f_permvar_truncv8siv8hi_1): Ditto.
7298 (*avx512f_vpermvar_truncv8div8si_1): Ditto.
7299 (*avx512f_permvar_truncv32hiv32qi_1): Ditto.
7300 (*avx512f_permvar_truncv16hiv16qi_1): Ditto.
7301 (*avx512f_permvar_truncv4div4si_1): Ditto.
7302 (*avx512f_pshufb_truncv8hiv8qi_1): Ditto.
7303 (*avx512f_pshufb_truncv4siv4hi_1): Ditto.
7304 (*avx512f_pshufd_truncv2div2si_1): Ditto.
7306 2021-08-16 Kito Cheng <kito.cheng@sifive.com>
7308 * config/riscv/multilib-generator: Support code model option for
7310 * doc/install.texi: Add document of new option for
7311 --with-multilib-generator.
7313 2021-08-15 Clément Chigot <clement.chigot@atos.net>
7315 * config/rs6000/rs6000.c (xcoff_tls_exec_model_detected): New.
7316 (rs6000_legitimize_tls_address_aix): Use it.
7317 (rs6000_xcoff_file_end): Add ".ref __tls_get_addr" when
7318 xcoff_tls_exec_model_detected is true.
7320 2021-08-15 Jeff Law <jlaw@localhost.localdomain>
7322 * config/h8300/h8300.c (shift_alg_si): Retune H8/300H shifts
7323 to allow a bit more code growth, saving many dozens of cycles.
7324 (h8300_option_override): Adjus shift_alg_si if optimizing for
7326 (get_shift_alg): Use special + inline shifts for residuals
7329 2021-08-14 Stafford Horne <shorne@gmail.com>
7332 * config/or1k/or1k-opts.h: New file.
7333 * config/or1k/or1k.c (or1k_legitimize_address_1, print_reloc):
7334 Support generating gotha relocations if -mcmodel=large is
7336 * config/or1k/or1k.h (TARGET_CMODEL_SMALL, TARGET_CMODEL_LARGE):
7338 * config/or1k/or1k.opt (mcmodel=): New option.
7339 * doc/invoke.texi (OpenRISC Options): Document mcmodel.
7341 2021-08-14 Martin Sebor <msebor@redhat.com>
7343 PR middle-end/101791
7344 * gimple-ssa-warn-access.cc (new_delete_mismatch_p): Use new argument
7345 to valid_new_delete_pair_p.
7346 * tree.c (valid_new_delete_pair_p): Add argument.
7347 * tree.h (valid_new_delete_pair_p): Same.
7349 2021-08-14 Jakub Jelinek <jakub@redhat.com>
7352 * config/i386/i386-expand.c (expand_vec_perm_broadcast_1)
7353 <case E_V64QImode>: For this mode assert
7354 !TARGET_AVX512BW || d->perm[0] rather than !TARGET_AVX2 || d->perm[0].
7356 2021-08-13 Michael Meissner <meissner@linux.ibm.com>
7359 * config/rs6000/altivec.md (xxeval): Use register_predicate
7360 instead of altivec_register_predicate.
7362 2021-08-13 Martin Sebor <msebor@redhat.com>
7364 PR middle-end/101734
7365 * tree-ssa-uninit.c (maybe_warn_read_write_only): New function.
7366 (maybe_warn_operand): Call it.
7368 2021-08-13 Martin Liska <mliska@suse.cz>
7371 * attribs.c (decl_attributes): Make naked functions "noipa"
7374 2021-08-13 Martin Liska <mliska@suse.cz>
7377 * symtab.c (symtab_node::noninterposable_alias): Do not create
7378 local aliases for target_clone functions as the clonning pass
7381 2021-08-13 Martin Liska <mliska@suse.cz>
7383 * opts.c (LIVE_PATCHING_OPTION): Define.
7384 (control_options_for_live_patching): Use it in error messages.
7386 2021-08-13 Jan Hubicka <hubicka@ucw.cz>
7388 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOREAD.
7389 (implicit_const_eaf_flags, implicit_pure_eaf_flags,
7390 ignore_stores_eaf_flags): New constants.
7391 (remove_useless_eaf_flags): New function.
7392 (eaf_flags_useful_p): Use it.
7393 (deref_flags): Add EAF_NOT_RETURNED if flag is unused;
7395 (modref_lattice::init): Add EAF_NOREAD.
7396 (modref_lattice::add_escape_point): Do not reacord escape point if
7398 (modref_lattice::merge): EAF_NOESCAPE implies EAF_NODIRECTESCAPE;
7399 use remove_useless_eaf_flags.
7400 (modref_lattice::merge_deref): Use ignore_stores_eaf_flags.
7401 (modref_lattice::merge_direct_load): Add EAF_NOREAD
7402 (analyze_ssa_name_flags): Fix handling EAF_NOT_RETURNED
7403 (analyze_parms): Use remove_useless_eaf_flags.
7404 (ipa_merge_modref_summary_after_inlining): Use ignore_stores_eaf_flags.
7405 (modref_merge_call_site_flags): Add caller and ecf_flags parameter;
7406 use remove_useless_eaf_flags.
7407 (modref_propagate_flags_in_scc): Update.
7408 * ipa-modref.h: Turn eaf_flags_t back to char.
7409 * tree-core.h (EAF_NOT_RETURNED): Fix.
7410 (EAF_NOREAD): New constant
7411 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1): Check for
7413 * tree-ssa-structalias.c (handle_rhs_call): Handle new flags.
7414 (handle_pure_call): Likewise.
7416 2021-08-12 Jakub Jelinek <jakub@redhat.com>
7418 * tree.def (OMP_MASKED): New tree code.
7419 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_FILTER.
7420 * tree.h (OMP_MASKED_BODY, OMP_MASKED_CLAUSES, OMP_MASKED_COMBINED,
7421 OMP_CLAUSE_FILTER_EXPR): Define.
7422 * tree.c (omp_clause_num_ops): Add OMP_CLAUSE_FILTER entry.
7423 (omp_clause_code_name): Likewise.
7424 (walk_tree_1): Handle OMP_CLAUSE_FILTER.
7425 * tree-nested.c (convert_nonlocal_omp_clauses,
7426 convert_local_omp_clauses): Handle OMP_CLAUSE_FILTER.
7427 (convert_nonlocal_reference_stmt, convert_local_reference_stmt,
7428 convert_gimple_call): Handle GIMPLE_OMP_MASTER.
7429 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_FILTER.
7430 (dump_generic_node): Handle OMP_MASTER.
7431 * gimple.def (GIMPLE_OMP_MASKED): New gimple code.
7432 * gimple.c (gimple_build_omp_masked): New function.
7433 (gimple_copy): Handle GIMPLE_OMP_MASKED.
7434 * gimple.h (gimple_build_omp_masked): Declare.
7435 (gimple_has_substatements): Handle GIMPLE_OMP_MASKED.
7436 (gimple_omp_masked_clauses, gimple_omp_masked_clauses_ptr,
7437 gimple_omp_masked_set_clauses): New inline functions.
7438 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_MASKED.
7439 * gimple-pretty-print.c (dump_gimple_omp_masked): New function.
7440 (pp_gimple_stmt_1): Handle GIMPLE_OMP_MASKED.
7441 * gimple-walk.c (walk_gimple_stmt): Likewise.
7442 * gimple-low.c (lower_stmt): Likewise.
7443 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
7444 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_FILTER. For clauses
7445 that take one expression rather than decl or constant, force
7446 gimplification of that into a SSA_NAME or temporary unless min
7448 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_FILTER.
7449 (gimplify_expr): Handle OMP_MASKED.
7450 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_MASKED.
7451 (estimate_num_insns): Likewise.
7452 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_FILTER.
7453 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_MASKED. Adjust
7454 diagnostics for existence of masked construct.
7455 (scan_omp_1_stmt, lower_omp_master, lower_omp_1, diagnose_sb_1,
7456 diagnose_sb_2): Handle GIMPLE_OMP_MASKED.
7457 * omp-expand.c (expand_omp_synch, expand_omp, omp_make_gimple_edges):
7460 2021-08-12 Uroš Bizjak <ubizjak@gmail.com>
7463 * config/i386/i386.md (avx512f_scalef<mode>2): New insn pattern.
7464 (ldexp<mode>3): Use avx512f_scalef<mode>2.
7465 (UNSPEC_SCALEF): Move from ...
7466 * config/i386/sse.md (UNSPEC_SCALEF): ... here.
7468 2021-08-12 Jan Hubicka <hubicka@ucw.cz>
7470 * ipa-split.c (consider_split): Fix condition testing void functions.
7472 2021-08-12 Aldy Hernandez <aldyh@redhat.com>
7474 * doc/invoke.texi: Remove docs for threader-mode param.
7475 * flag-types.h (enum threader_mode): Remove.
7476 * params.opt: Remove threader-mode param.
7477 * tree-ssa-threadbackward.c (class back_threader): Remove
7478 path_is_unreachable_p.
7479 Make find_paths private.
7480 Add maybe_thread and thread_through_all_blocks.
7481 Remove reference marker for m_registry.
7482 Remove reference marker for m_profit.
7483 (back_threader::back_threader): Adjust for registry and profit not
7485 (dump_path): Move down.
7487 (class thread_jumps): Remove.
7488 (class back_threader_registry): Remove m_all_paths.
7490 (thread_jumps::thread_through_all_blocks): Move to back_threader
7492 (fsm_find_thread_path): Remove
7493 (back_threader::maybe_thread): New.
7494 (back_threader::thread_through_all_blocks): Move from
7496 (back_threader_registry::back_threader_registry): Remove
7498 (back_threader_registry::~back_threader_registry): Remove.
7499 (thread_jumps::find_taken_edge): Remove.
7500 (thread_jumps::check_subpath_and_update_thread_path): Remove.
7501 (thread_jumps::maybe_register_path): Remove.
7502 (thread_jumps::handle_phi): Remove.
7503 (handle_assignment_p): Remove.
7504 (thread_jumps::handle_assignment): Remove.
7505 (thread_jumps::fsm_find_control_statement_thread_paths): Remove.
7506 (thread_jumps::find_jump_threads_backwards): Remove.
7507 (thread_jumps::find_jump_threads_backwards_with_ranger): Remove.
7508 (try_thread_blocks): Rename find_jump_threads_backwards to
7510 (pass_early_thread_jumps::execute): Same.
7512 2021-08-12 Tobias Burnus <tobias@codesourcery.com>
7514 * tree-core.h (omp_clause_proc_bind_kind): Add
7515 OMP_CLAUSE_PROC_BIND_PRIMARY.
7516 * tree-pretty-print.c (dump_omp_clause): Add TODO comment to
7517 change 'master' to 'primary' in proc_bind for OpenMP 5.1.
7519 2021-08-12 Claudiu Zissulescu <claziss@synopsys.com>
7521 * common/config/arc/arc-common.c (arc_option_init_struct): Remove
7522 fno-common reference.
7523 * config/arc/arc.c (arc_override_options): Remove overriding of
7526 2021-08-12 Jakub Jelinek <jakub@redhat.com>
7529 * config/i386/i386-expand.c (ix86_expand_vec_one_operand_perm_avx512):
7530 If d->testing_p, return true after performing checks instead of
7531 actually expanding the insn.
7532 (expand_vec_perm_broadcast_1): Handle V32HImode - assert
7533 !TARGET_AVX512BW and return false.
7535 2021-08-12 Eric Botcazou <ebotcazou@gcc.gnu.org>
7537 * configure.ac (PE linker --disable-dynamicbase support): New check.
7538 * configure: Regenerate.
7539 * config.in: Likewise.
7540 * config/i386/mingw32.h (LINK_SPEC_DISABLE_DYNAMICBASE): New define.
7541 (LINK_SPEC): Use it.
7542 * config/i386/mingw-w64.h (LINK_SPEC_DISABLE_DYNAMICBASE): Likewise.
7543 (LINK_SPEC): Likewise.
7545 2021-08-12 liuhongt <hongtao.liu@intel.com>
7548 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_2): New
7549 post_reload define_insn_and_split.
7550 (*avx512bw_zero_extendv32qiv32hi2_2): Ditto.
7551 (*sse4_1_zero_extendv8qiv8hi2_4): Ditto.
7552 (*avx512f_zero_extendv16hiv16si2_2): Ditto.
7553 (*avx2_zero_extendv8hiv8si2_2): Ditto.
7554 (*sse4_1_zero_extendv4hiv4si2_4): Ditto.
7555 (*avx512f_zero_extendv8siv8di2_2): Ditto.
7556 (*avx2_zero_extendv4siv4di2_2): Ditto.
7557 (*sse4_1_zero_extendv2siv2di2_4): Ditto.
7558 (VI248_256, VI248_512, VI148_512, VI148_256, VI148_128): New
7561 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
7563 * config/rs6000/rs6000-builtin-new.def: Add always, power5, and
7566 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
7568 * config/rs6000/rs6000-builtin-new.def: Add vsx stanza.
7570 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
7572 * config/rs6000/rs6000-builtin-new.def: Finish altivec stanza.
7573 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
7574 initialization of pcvoid_type_node here...
7575 (altivec_init_builtins): ...from here.
7576 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add
7577 RS6000_BTI_const_ptr_void.
7578 (pcvoid_type_node): New macro.
7580 2021-08-11 Richard Biener <rguenther@suse.de>
7583 * tree-ssa-forwprop.c (pass_forwprop::execute): Do not decompose
7584 hard-register accesses.
7586 2021-08-11 Richard Biener <rguenther@suse.de>
7588 * tree-ssa-operands.c (operands_scanner::get_expr_operands):
7589 Do not look at COMPONENT_REF FIELD_DECLs TREE_THIS_VOLATILE
7590 to determine has_volatile_ops.
7592 2021-08-11 Eric Botcazou <ebotcazou@gcc.gnu.org>
7594 * cfgexpand.c (expand_used_vars): Reuse attribs local variable.
7596 2021-08-11 Jan Hubicka <hubicka@ucw.cz>
7597 Alexandre Oliva <oliva@adacore.com>
7599 * ipa-modref.c (modref_lattice::dump): Fix escape_point's min_flags
7601 (modref_lattice::merge_deref): Fix handling of indirect scape points.
7602 (update_escape_summary_1): Likewise.
7603 (update_escape_summary): Likewise.
7604 (ipa_merge_modref_summary_after_inlining): Likewise.
7606 2021-08-11 Richard Biener <rguenther@suse.de>
7608 PR middle-end/101858
7609 * fold-const.c (fold_binary_loc): Guard simplification
7610 of X < (cast) (1 << Y) to integer types.
7612 2021-08-11 Richard Biener <rguenther@suse.de>
7614 PR tree-optimization/101861
7615 * tree-vect-stmts.c (vectorizable_load): Fix error in
7616 previous change with regard to gather vectorization.
7618 2021-08-11 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
7621 * config/arm/arm_neon.h (vdup_n_s8): Replace call to builtin
7623 (vdup_n_s16): Likewise.
7624 (vdup_n_s32): Likewise.
7625 (vdup_n_s64): Likewise.
7626 (vdup_n_u8): Likewise.
7627 (vdup_n_u16): Likewise.
7628 (vdup_n_u32): Likewise.
7629 (vdup_n_u64): Likewise.
7630 (vdup_n_p8): Likewise.
7631 (vdup_n_p16): Likewise.
7632 (vdup_n_p64): Likewise.
7633 (vdup_n_f16): Likewise.
7634 (vdup_n_f32): Likewise.
7635 (vdupq_n_s8): Likewise.
7636 (vdupq_n_s16): Likewise.
7637 (vdupq_n_s32): Likewise.
7638 (vdupq_n_s64): Likewise.
7639 (vdupq_n_u8): Likewise.
7640 (vdupq_n_u16): Likewise.
7641 (vdupq_n_u32): Likewise.
7642 (vdupq_n_u64): Likewise.
7643 (vdupq_n_p8): Likewise.
7644 (vdupq_n_p16): Likewise.
7645 (vdupq_n_p64): Likewise.
7646 (vdupq_n_f16): Likewise.
7647 (vdupq_n_f32): Likewise.
7648 (vmov_n_s8): Replace call to builtin with call to corresponding
7650 (vmov_n_s16): Likewise.
7651 (vmov_n_s32): Likewise.
7652 (vmov_n_s64): Likewise.
7653 (vmov_n_u8): Likewise.
7654 (vmov_n_u16): Likewise.
7655 (vmov_n_u32): Likewise.
7656 (vmov_n_u64): Likewise.
7657 (vmov_n_p8): Likewise.
7658 (vmov_n_p16): Likewise.
7659 (vmov_n_f16): Likewise.
7660 (vmov_n_f32): Likewise.
7661 (vmovq_n_s8): Likewise.
7662 (vmovq_n_s16): Likewise.
7663 (vmovq_n_s32): Likewise.
7664 (vmovq_n_s64): Likewise.
7665 (vmovq_n_u8): Likewise.
7666 (vmovq_n_u16): Likewise.
7667 (vmovq_n_u32): Likewise.
7668 (vmovq_n_u64): Likewise.
7669 (vmovq_n_p8): Likewise.
7670 (vmovq_n_p16): Likewise.
7671 (vmovq_n_f16): Likewise.
7672 (vmovq_n_f32): Likewise.
7673 * config/arm/arm_neon_builtins.def: Remove entries for vdup_n.
7675 2021-08-11 liuhongt <hongtao.liu@intel.com>
7678 * config/i386/i386.md (ldexp<mode>3): Extend to vscalefs[sd]
7679 when TARGET_AVX512F and TARGET_SSE_MATH.
7681 2021-08-10 Jakub Jelinek <jakub@redhat.com>
7684 * config/i386/i386-expand.c (expand_vec_perm_even_odd): Return false
7685 for V32HImode if !TARGET_AVX512BW.
7686 (ix86_vectorize_vec_perm_const) <case E_V32HImode, case E_V64QImode>:
7687 If !TARGET_AVX512BW and TARGET_AVX512F and d.testing_p, don't fail
7688 early, but actually check the permutation.
7690 2021-08-10 Richard Biener <rguenther@suse.de>
7692 PR tree-optimization/101809
7693 * tree-vect-stmts.c (get_load_store_type): Allow emulated
7694 gathers with offset vector nunits being a constant multiple
7695 of the data vector nunits.
7696 (vect_get_gather_scatter_ops): Use the appropriate nunits
7697 for the offset vector defs.
7698 (vectorizable_store): Adjust call to
7699 vect_get_gather_scatter_ops.
7700 (vectorizable_load): Likewise. Handle the case of less
7701 offset vectors than data vectors.
7703 2021-08-10 Jakub Jelinek <jakub@redhat.com>
7706 * config/i386/sse.md (*avx512f_shuf_<shuffletype>64x2_1<mask_name>_1,
7707 *avx512f_shuf_<shuffletype>32x4_1<mask_name>_1): New define_insn
7710 2021-08-10 Richard Biener <rguenther@suse.de>
7712 PR tree-optimization/101801
7713 PR tree-optimization/101819
7714 * tree-vectorizer.h (vect_emulated_vector_p): Declare.
7715 * tree-vect-loop.c (vect_emulated_vector_p): New function.
7716 (vectorizable_reduction): Re-instantiate a check for emulated
7718 * tree-vect-stmts.c (vectorizable_shift): Likewise.
7719 (vectorizable_operation): Likewise. Cost emulated vector
7720 operations according to the scalar sequence synthesized by
7723 2021-08-10 Richard Biener <rguenther@suse.de>
7725 PR middle-end/101824
7726 * tree-nested.c (get_frame_field): Mark the COMPONENT_REF as
7727 volatile in case the variable was.
7729 2021-08-10 H.J. Lu <hjl.tools@gmail.com>
7732 * config/i386/constraints.md (BC): Document for integer SSE
7733 constant all bits set operand.
7734 (BF): New constraint for const floating-point all bits set
7736 * config/i386/i386.c (standard_sse_constant_p): Likewise.
7737 (standard_sse_constant_opcode): Likewise.
7738 * config/i386/sse.md (sseconstm1): New mode attribute.
7739 (mov<mode>_internal): Replace BC with <sseconstm1>.
7741 2021-08-10 liuhongt <hongtao.liu@intel.com>
7743 * config/i386/sse.md (cond_<insn><mode>): New expander.
7744 (VI248_AVX512VLBW): New mode iterator.
7745 * config/i386/predicates.md
7746 (nonimmediate_or_const_vec_dup_operand): New predicate.
7748 2021-08-09 Andrew MacLeod <amacleod@redhat.com>
7750 PR tree-optimization/101741
7751 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Check
7752 type of parameter for toupper/tolower.
7754 2021-08-09 Martin Jambor <mjambor@suse.cz>
7757 * ipa-prop.c (propagate_controlled_uses): Removed a spurious space.
7759 2021-08-09 Pat Haugen <pthaugen@linux.ibm.com>
7761 * config/rs6000/rs6000.c (is_load_insn1): Verify destination is a
7763 (is_store_insn1): Verify source is a register.
7765 2021-08-09 Uroš Bizjak <ubizjak@gmail.com>
7768 * config/i386/mmx.md (<any_logic:code>v2sf3):
7769 Rename from *mmx_<any_logic:code>v2sf3
7771 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7773 * config/nvptx/nvptx.c: Cross-reference parts adapted in
7774 'gcc/omp-oacc-neuter-broadcast.cc'.
7775 * omp-low.c: Likewise.
7776 * omp-oacc-neuter-broadcast.cc: Cross-reference parts adapted from
7779 2021-08-09 Julian Brown <julian@codesourcery.com>
7780 Kwok Cheung Yeung <kcy@codesourcery.com>
7781 Thomas Schwinge <thomas@codesourcery.com>
7783 * config/gcn/gcn.c (gcn_init_builtins): Override decls for
7784 BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START,
7785 BUILT_IN_GOACC_SINGLE_COPY_END and BUILT_IN_GOACC_BARRIER.
7786 (gcn_goacc_validate_dims): Turn on worker partitioning unconditionally.
7787 (gcn_fork_join): Update comment.
7788 * config/gcn/gcn.opt (flag_worker_partitioning): Remove.
7789 (macc_experimental_workers): Remove unused option.
7791 2021-08-09 Julian Brown <julian@codesourcery.com>
7792 Nathan Sidwell <nathan@codesourcery.com> (via 'gcc/config/nvptx/nvptx.c' master)
7793 Kwok Cheung Yeung <kcy@codesourcery.com>
7794 Thomas Schwinge <thomas@codesourcery.com>
7796 * Makefile.in (OBJS): Add omp-oacc-neuter-broadcast.o.
7797 * doc/tm.texi.in (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD):
7798 Add documentation hook.
7799 * doc/tm.texi: Regenerate.
7800 * omp-oacc-neuter-broadcast.cc: New file.
7801 * omp-builtins.def (BUILT_IN_GOACC_BARRIER)
7802 (BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START)
7803 (BUILT_IN_GOACC_SINGLE_COPY_END): New builtins.
7804 * passes.def (pass_omp_oacc_neuter_broadcast): Add pass.
7805 * target.def (goacc.create_worker_broadcast_record): Add target
7807 * tree-pass.h (make_pass_omp_oacc_neuter_broadcast): Add
7809 * config/gcn/gcn-protos.h (gcn_goacc_adjust_propagation_record):
7810 Rename prototype to...
7811 (gcn_goacc_create_worker_broadcast_record): ... this.
7812 * config/gcn/gcn-tree.c (gcn_goacc_adjust_propagation_record): Rename
7814 (gcn_goacc_create_worker_broadcast_record): ... this.
7815 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_PROPAGATION_RECORD):
7817 (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD): ... this.
7819 2021-08-09 Tejas Belagod <tejas.belagod@arm.com>
7822 * config/aarch64/aarch64-simd.md (vlshr<mode>3, vashr<mode>3): Use
7825 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7827 * Makefile.in (GTFILES): Remove '$(srcdir)/omp-offload.c'.
7829 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7831 * builtins.def (DEF_GOACC_BUILTIN, DEF_GOMP_BUILTIN): Don't
7832 consider '-foffload-abi'.
7833 * common.opt (-foffload-abi): Remove 'Var', 'Init'.
7834 * opts.c (common_handle_option) <-foffload-abi> [ACCEL_COMPILER]:
7837 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7839 * optc-gen.awk: Sanity check that 'Init' doesn't appear without
7842 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7844 * omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE): Remove.
7846 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7848 * doc/gty.texi (Files): Update.
7850 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7852 * doc/gty.texi (Files): Fix GTY header file example.
7854 2021-08-09 Roger Sayle <roger@nextmovesoftware.com>
7856 * tree-ssa-ccp.c (value_mask_to_min_max): Helper function to
7857 determine the upper and lower bounds from a mask-value pair.
7858 (bit_value_unop) [ABS_EXPR, ABSU_EXPR]: Add support for
7859 absolute value and unsigned absolute value expressions.
7860 (bit_value_binop): Initialize *VAL's precision.
7861 [LT_EXPR, LE_EXPR]: Use value_mask_to_min_max to determine
7862 upper and lower bounds of operands. Add LE_EXPR/GE_EXPR
7863 support when the operands are unknown but potentially equal.
7864 [MIN_EXPR, MAX_EXPR]: Support minimum/maximum expressions.
7866 2021-08-09 Bin Cheng <bin.cheng@linux.alibaba.com>
7868 * config/aarch64/aarch64.md
7869 (*extend<SHORT:mode><GPI:mode>2_aarch64): Use %<GPI:w>0.
7871 2021-08-08 Sergei Trofimovich <siarheit@google.com>
7873 * lra-constraints.c: Fix s/otput/output/ typo.
7875 2021-08-06 Martin Sebor <msebor@redhat.com>
7877 * builtins.c (expand_builtin_memchr): Move to gimple-ssa-warn-access.cc.
7878 (expand_builtin_strcat): Same.
7879 (expand_builtin_stpncpy): Same.
7880 (expand_builtin_strncat): Same.
7881 (check_read_access): Same.
7882 (check_memop_access): Same.
7883 (expand_builtin_strlen): Move checks to gimple-ssa-warn-access.cc.
7884 (expand_builtin_strnlen): Same.
7885 (expand_builtin_memcpy): Same.
7886 (expand_builtin_memmove): Same.
7887 (expand_builtin_mempcpy): Same.
7888 (expand_builtin_strcpy): Same.
7889 (expand_builtin_strcpy_args): Same.
7890 (expand_builtin_stpcpy_1): Same.
7891 (expand_builtin_strncpy): Same.
7892 (expand_builtin_memset): Same.
7893 (expand_builtin_bzero): Same.
7894 (expand_builtin_strcmp): Same.
7895 (expand_builtin_strncmp): Same.
7896 (expand_builtin): Remove handlers.
7897 (fold_builtin_strlen): Add a comment.
7898 * builtins.h (check_access): Move to gimple-ssa-warn-access.cc.
7899 * calls.c (maybe_warn_nonstring_arg): Same.
7900 * diagnostic-spec.c (nowarn_spec_t::nowarn_spec_t): Add warning option.
7901 * gimple-fold.c (gimple_fold_builtin_strcpy): Pass argument to callee.
7902 (gimple_fold_builtin_stpcpy): Same.
7903 * gimple-ssa-warn-access.cc (has_location): New function.
7904 (get_location): Same.
7905 (get_callee_fndecl): Same.
7908 (warn_string_no_nul): Define.
7909 (unterminated_array): Same.
7910 (check_nul_terminated_array): Same.
7911 (maybe_warn_nonstring_arg): Same.
7912 (maybe_warn_for_bound): Same.
7913 (warn_for_access): Same.
7914 (check_access): Same.
7915 (check_memop_access): Same.
7916 (check_read_access): Same.
7917 (warn_dealloc_offset): Use helper functions.
7918 (maybe_emit_free_warning): Same.
7919 (class pass_waccess): Add members.
7920 (check_strcat): New function.
7921 (check_strncat): New function.
7922 (check_stxcpy): New function.
7923 (check_stxncpy): New function.
7924 (check_strncmp): New function.
7925 (pass_waccess::check_builtin): New function.
7926 (pass_waccess::check): Call it.
7927 * gimple-ssa-warn-access.h (warn_string_no_nul): Move here from
7929 (maybe_warn_for_bound): Same.
7930 (check_access): Same.
7931 (check_memop_access): Same.
7932 (check_read_access): Same.
7933 * pointer-query.h (struct access_data): Define a ctor overload.
7935 2021-08-06 Richard Biener <rguenther@suse.de>
7937 PR tree-optimization/101801
7938 * tree-vectorizer.h (vect_worthwhile_without_simd_p): Rename...
7939 (vect_can_vectorize_without_simd_p): ... to this.
7940 * tree-vect-loop.c (vect_worthwhile_without_simd_p): Rename...
7941 (vect_can_vectorize_without_simd_p): ... to this and fold
7942 in vect_min_worthwhile_factor.
7943 (vect_min_worthwhile_factor): Remove.
7944 (vectorizable_reduction): Adjust and remove the cost part.
7945 * tree-vect-stmts.c (vectorizable_shift): Likewise.
7946 (vectorizable_operation): Likewise.
7948 2021-08-06 Uroš Bizjak <ubizjak@gmail.com>
7951 * config/i386/i386.md (cmove reg-to-reg move elimination peephole2s):
7952 Add general_gr_operand predicate to operand 3.
7954 2021-08-06 Roger Sayle <roger@nextmovesoftware.com>
7956 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern): Use
7957 CFN_BUILT_IN_CLRSB* instead of BUILT_IN_CLRSB* for consistency.
7959 2021-08-06 Tamar Christina <tamar.christina@arm.com>
7961 * config/aarch64/aarch64-sve-builtins.cc (register_svpattern,
7962 register_svprfop): Pass vec<> by pointer.
7963 * langhooks-def.h (lhd_simulate_enum_decl): Likewise.
7964 * langhooks.c (lhd_simulate_enum_decl): Likewise.
7965 * langhooks.h (struct lang_hooks_for_types): Likewise.
7967 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
7969 * config/aarch64/arm_neon.h (vst1_bf16_x2): Use
7970 __builtin_memcpy instead of constructing an additional
7971 __builtin_aarch64_simd_oi one vector at a time.
7972 (vst1q_bf16_x2): Likewise.
7973 (vst1_bf16_x3): Use __builtin_memcpy instead of constructing
7974 an additional __builtin_aarch64_simd_ci one vector at a time.
7975 (vst1q_bf16_x3): Likewise.
7976 (vst1_bf16_x4): Use __builtin_memcpy instead of a union.
7977 (vst1q_bf16_x4): Likewise.
7978 (vst2_bf16): Use __builtin_memcpy instead of constructing an
7979 additional __builtin_aarch64_simd_oi one vector at a time.
7980 (vst2q_bf16): Likewise.
7981 (vst3_bf16): Use __builtin_memcpy instead of constructing an
7982 additional __builtin_aarch64_simd_ci mode one vector at a
7984 (vst3q_bf16): Likewise.
7985 (vst4_bf16): Use __builtin_memcpy instead of constructing an
7986 additional __builtin_aarch64_simd_xi one vector at a time.
7987 (vst4q_bf16): Likewise.
7989 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
7991 * config/aarch64/arm_neon.h (__ST2_LANE_FUNC): Delete.
7992 (__ST2Q_LANE_FUNC): Delete.
7993 (vst2_lane_f16): Use __builtin_memcpy to copy vector
7994 structure instead of constructing __builtin_aarch64_simd_oi
7995 one vector at a time.
7996 (vst2_lane_f32): Likewise.
7997 (vst2_lane_f64): Likewise.
7998 (vst2_lane_p8): Likewise.
7999 (vst2_lane_p16): Likewise.
8000 (vst2_lane_p64): Likewise.
8001 (vst2_lane_s8): Likewise.
8002 (vst2_lane_s16): Likewise.
8003 (vst2_lane_s32): Likewise.
8004 (vst2_lane_s64): Likewise.
8005 (vst2_lane_u8): Likewise.
8006 (vst2_lane_u16): Likewise.
8007 (vst2_lane_u32): Likewise.
8008 (vst2_lane_u64): Likewise.
8009 (vst2_lane_bf16): Likewise.
8010 (vst2q_lane_f16): Use __builtin_memcpy to copy vector
8011 structure instead of using a union.
8012 (vst2q_lane_f32): Likewise.
8013 (vst2q_lane_f64): Likewise.
8014 (vst2q_lane_p8): Likewise.
8015 (vst2q_lane_p16): Likewise.
8016 (vst2q_lane_p64): Likewise.
8017 (vst2q_lane_s8): Likewise.
8018 (vst2q_lane_s16): Likewise.
8019 (vst2q_lane_s32): Likewise.
8020 (vst2q_lane_s64): Likewise.
8021 (vst2q_lane_u8): Likewise.
8022 (vst2q_lane_u16): Likewise.
8023 (vst2q_lane_u32): Likewise.
8024 (vst2q_lane_u64): Likewise.
8025 (vst2q_lane_bf16): Likewise.
8027 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
8029 * config/aarch64/arm_neon.h (__ST3_LANE_FUNC): Delete.
8030 (__ST3Q_LANE_FUNC): Delete.
8031 (vst3_lane_f16): Use __builtin_memcpy to copy vector
8032 structure instead of constructing __builtin_aarch64_simd_ci
8033 one vector at a time.
8034 (vst3_lane_f32): Likewise.
8035 (vst3_lane_f64): Likewise.
8036 (vst3_lane_p8): Likewise.
8037 (vst3_lane_p16): Likewise.
8038 (vst3_lane_p64): Likewise.
8039 (vst3_lane_s8): Likewise.
8040 (vst3_lane_s16): Likewise.
8041 (vst3_lane_s32): Likewise.
8042 (vst3_lane_s64): Likewise.
8043 (vst3_lane_u8): Likewise.
8044 (vst3_lane_u16): Likewise.
8045 (vst3_lane_u32): Likewise.
8046 (vst3_lane_u64): Likewise.
8047 (vst3_lane_bf16): Likewise.
8048 (vst3q_lane_f16): Use __builtin_memcpy to copy vector
8049 structure instead of using a union.
8050 (vst3q_lane_f32): Likewise.
8051 (vst3q_lane_f64): Likewise.
8052 (vst3q_lane_p8): Likewise.
8053 (vst3q_lane_p16): Likewise.
8054 (vst3q_lane_p64): Likewise.
8055 (vst3q_lane_s8): Likewise.
8056 (vst3q_lane_s16): Likewise.
8057 (vst3q_lane_s32): Likewise.
8058 (vst3q_lane_s64): Likewise.
8059 (vst3q_lane_u8): Likewise.
8060 (vst3q_lane_u16): Likewise.
8061 (vst3q_lane_u32): Likewise.
8062 (vst3q_lane_u64): Likewise.
8063 (vst3q_lane_bf16): Likewise.
8065 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
8067 * config/aarch64/arm_neon.h (__ST4_LANE_FUNC): Delete.
8068 (__ST4Q_LANE_FUNC): Delete.
8069 (vst4_lane_f16): Use __builtin_memcpy to copy vector
8070 structure instead of constructing __builtin_aarch64_simd_xi
8071 one vector at a time.
8072 (vst4_lane_f32): Likewise.
8073 (vst4_lane_f64): Likewise.
8074 (vst4_lane_p8): Likewise.
8075 (vst4_lane_p16): Likewise.
8076 (vst4_lane_p64): Likewise.
8077 (vst4_lane_s8): Likewise.
8078 (vst4_lane_s16): Likewise.
8079 (vst4_lane_s32): Likewise.
8080 (vst4_lane_s64): Likewise.
8081 (vst4_lane_u8): Likewise.
8082 (vst4_lane_u16): Likewise.
8083 (vst4_lane_u32): Likewise.
8084 (vst4_lane_u64): Likewise.
8085 (vst4_lane_bf16): Likewise.
8086 (vst4q_lane_f16): Use __builtin_memcpy to copy vector
8087 structure instead of using a union.
8088 (vst4q_lane_f32): Likewise.
8089 (vst4q_lane_f64): Likewise.
8090 (vst4q_lane_p8): Likewise.
8091 (vst4q_lane_p16): Likewise.
8092 (vst4q_lane_p64): Likewise.
8093 (vst4q_lane_s8): Likewise.
8094 (vst4q_lane_s16): Likewise.
8095 (vst4q_lane_s32): Likewise.
8096 (vst4q_lane_s64): Likewise.
8097 (vst4q_lane_u8): Likewise.
8098 (vst4q_lane_u16): Likewise.
8099 (vst4q_lane_u32): Likewise.
8100 (vst4q_lane_u64): Likewise.
8101 (vst4q_lane_bf16): Likewise.
8103 2021-08-06 Martin Liska <mliska@suse.cz>
8105 * config/rs6000/rs6000.c (rs6000_option_override_internal): When
8106 a target option is restored, it can have
8107 rs6000_long_double_type_size set to FLOAT_PRECISION_TFmode
8108 and error should not be emitted.
8110 2021-08-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
8112 * gcov-io.h (gcov_write): Declare.
8113 * gcov-io.c (gcov_write): New.
8114 (gcov_write_counter): Remove.
8115 (gcov_write_tag_length): Likewise.
8116 (gcov_write_summary): Replace gcov_write_tag_length() with calls to
8117 gcov_write_unsigned().
8118 * doc/invoke.texi (fprofile-info-section): Mention
8119 __gcov_info_to_gdca().
8121 2021-08-06 Martin Sebor <msebor@redhat.com>
8123 * dominance.c (prune_bbs_to_update_dominators): Adjust by-value vec
8124 arguments to by-reference.
8125 (iterate_fix_dominators): Same.
8126 * dominance.h (iterate_fix_dominators): Same.
8127 * ipa-prop.h: Call auto_vec::to_vec_legacy.
8128 * tree-data-ref.c (dump_data_dependence_relation): Adjust by-value vec
8129 arguments to by-reference.
8130 (debug_data_dependence_relation): Same.
8131 (dump_data_dependence_relations): Same.
8132 * tree-data-ref.h (debug_data_dependence_relation): Same.
8133 (dump_data_dependence_relations): Same.
8134 * tree-predcom.c (dump_chains): Same.
8135 (initialize_root_vars_lm): Same.
8136 (determine_unroll_factor): Same.
8137 (replace_phis_by_defined_names): Same.
8138 (insert_init_seqs): Same.
8139 (pcom_worker::tree_predictive_commoning_loop): Call
8140 auto_vec::to_vec_legacy.
8141 * tree-ssa-pre.c (insert_into_preds_of_block): Adjust by-value vec
8142 arguments to by-reference.
8143 * tree-ssa-threadbackward.c (populate_worklist): Same.
8144 (back_threader::resolve_def): Same.
8145 * tree-vect-data-refs.c (vect_check_nonzero_value): Same.
8146 (vect_enhance_data_refs_alignment): Same.
8147 (vect_check_lower_bound): Same.
8148 (vect_prune_runtime_alias_test_list): Same.
8149 (vect_permute_store_chain): Same.
8150 * tree-vect-slp-patterns.c (vect_normalize_conj_loc): Same.
8151 * tree-vect-stmts.c (vect_create_vectorized_demotion_stmts): Same.
8152 * tree-vectorizer.h (vect_permute_store_chain): Same.
8153 * vec.c (test_init): New function.
8154 (vec_c_tests): Call new function.
8155 * vec.h (vec): Declare ctors, dtor, and assignment.
8156 (auto_vec::vec_to_legacy): New function.
8157 (vec::copy): Adjust initialization.
8159 2021-08-05 H.J. Lu <hjl.tools@gmail.com>
8162 * config/i386/i386.c (ix86_can_inline_p): Ignore MASK_80387 if
8163 callee only uses GPRs.
8164 * config/i386/ia32intrin.h: Revert commit 5463cee2770.
8165 * config/i386/serializeintrin.h: Revert commit 71958f740f1.
8166 * config/i386/x86gprintrin.h: Add
8167 #pragma GCC target("general-regs-only") and #pragma GCC pop_options
8168 to disable non-GPR ISAs.
8170 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
8172 PR middle-end/101787
8173 * doc/md.texi (cond_ashl, cond_ashr, cond_lshr): Document.
8175 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
8177 * tree-vectorizer.h (vect_is_store_elt_extraction, vect_is_reduction)
8178 (vect_reduc_type, vect_embedded_comparison_type, vect_comparison_type)
8179 (vect_is_extending_load, vect_is_integer_truncation): New functions,
8180 moved from aarch64.c but given different names.
8181 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction)
8182 (aarch64_is_reduction, aarch64_reduc_type)
8183 (aarch64_embedded_comparison_type, aarch64_comparison_type)
8184 (aarch64_extending_load_p, aarch64_integer_truncation_p): Delete
8185 in favor of the above. Update callers accordingly.
8187 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
8190 * config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
8191 writing .cpu directive in asm output.
8192 * config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
8193 (arm_last_printed_arch_string): Delete.
8194 (arm_last-printed_fpu_string): Delete.
8195 (arm_configure_build_target): If use of floating-point/SIMD is
8196 disabled, remove all fp/simd related features from the target ISA.
8197 (last_arm_targ_options): New variable.
8198 (arm_print_asm_arch_directives): Add new parameters. Change order
8199 of emitted directives and handle all cases here.
8200 (arm_file_start): Always call arm_print_asm_arch_directives, move
8201 all generation of .arch/.arch_extension here.
8202 (arm_file_end): Call arm_print_asm_arch.
8203 (arm_declare_function_name): Call arm_print_asm_arch_directives
8204 instead of printing .arch/.fpu directives directly.
8206 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
8208 * config/arm/arm.c (arm_configure_build_target): Don't call
8209 arm_option_reconfigure_globals.
8210 (arm_option_restore): Call arm_option_reconfigure_globals after
8211 reconfiguring the target.
8212 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
8214 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
8216 * config/arm/arm.c (arm_configure_build_target): Ensure the target's
8217 arch_name is always set.
8219 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
8221 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
8222 of vec_select high-half from being added into Neon subtract
8225 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
8227 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
8228 of vec_select high-half from being added into Neon add cost.
8230 2021-08-05 Kewen Lin <linkw@linux.ibm.com>
8232 * cfgloop.h (loops_list::loops_list): Add one optional argument
8233 root and adjust accordingly, update loop tree walking and factor
8235 * cfgloop.c (loops_list::walk_loop_tree): ... this. New function.
8237 2021-08-05 Eric Botcazou <ebotcazou@gcc.gnu.org>
8239 PR tree-optimization/101626
8240 * tree-sra.c (propagate_subaccesses_from_rhs): Do not set the
8241 reverse scalar storage order on a pointer or vector component.
8243 2021-08-05 liuhongt <hongtao.liu@intel.com>
8245 * config/i386/sse.md (cond_<code><mode>): New expander.
8247 2021-08-05 liuhongt <hongtao.liu@intel.com>
8249 * config/i386/sse.md (cond_<code><mode>): New expander.
8251 2021-08-05 liuhongt <hongtao.liu@intel.com>
8253 * config/i386/sse.md (cond_<code><mode>): New expander.
8255 2021-08-04 David Malcolm <dmalcolm@redhat.com>
8258 * Makefile.in (ANALYZER_OBJS): Add analyzer/region-model-asm.o.
8260 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
8263 * config/i386/i386.h (STORE_MAX_PIECES): Allow 16/32/64 bytes
8264 only if TARGET_INTER_UNIT_MOVES_TO_VEC is true.
8266 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
8269 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
8270 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
8271 data with SSE register from one memory location to another.
8273 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
8275 * config/s390/s390.c (expand_perm_with_vpdi): New function.
8276 (vectorize_vec_perm_const_1): Call expand_perm_with_vpdi.
8277 * config/s390/vector.md (*vpdi1<mode>, @vpdi1<mode>): Enable a
8278 parameterized expander.
8279 (*vpdi4<mode>, @vpdi4<mode>): Likewise.
8281 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
8283 * config/s390/s390.c (MAX_VECT_LEN): Define macro.
8284 (struct expand_vec_perm_d): Define struct.
8285 (expand_perm_with_merge): New function.
8286 (vectorize_vec_perm_const_1): New function.
8287 (s390_vectorize_vec_perm_const): New function.
8288 (TARGET_VECTORIZE_VEC_PERM_CONST): Define target macro.
8290 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
8292 * config/s390/vector.md (V_HW_64): Remove mode iterator.
8293 (*vec_load_pair<mode>): Use V_HW_2 instead of V_HW_64.
8294 * config/s390/vx-builtins.md
8295 (vec_scatter_element<V_HW_2:mode>_SI): Use V_HW_2 instead of
8298 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
8300 * config/s390/s390.md (UNSPEC_VEC_PERMI): Remove constant
8302 * config/s390/vector.md (*vpdi1<mode>, *vpdi4<mode>): New pattern
8304 * config/s390/vx-builtins.md (*vec_permi<mode>): Emit generic rtx
8305 instead of an unspec.
8307 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
8309 * config/s390/s390-modes.def: Add more vector modes to support
8310 concatenation of two vectors.
8311 * config/s390/s390-protos.h (s390_expand_merge_perm_const): Add
8313 (s390_expand_merge): Likewise.
8314 * config/s390/s390.c (s390_expand_merge_perm_const): New function.
8315 (s390_expand_merge): New function.
8316 * config/s390/s390.md (UNSPEC_VEC_MERGEH, UNSPEC_VEC_MERGEL):
8317 Remove constant definitions.
8318 * config/s390/vector.md (V_HW_2): Add mode iterators.
8319 (VI_HW_4, V_HW_4): Rename VI_HW_4 to V_HW_4.
8320 (vec_2x_nelts, vec_2x_wide): New mode attributes.
8321 (*vmrhb, *vmrlb, *vmrhh, *vmrlh, *vmrhf, *vmrlf, *vmrhg, *vmrlg):
8322 New pattern definitions.
8323 (vec_widen_umult_lo_<mode>, vec_widen_umult_hi_<mode>)
8324 (vec_widen_smult_lo_<mode>, vec_widen_smult_hi_<mode>)
8325 (vec_unpacks_lo_v4sf, vec_unpacks_hi_v4sf, vec_unpacks_lo_v2df)
8326 (vec_unpacks_hi_v2df): Adjust expanders to emit non-unspec RTX for
8328 * config/s390/vx-builtins.md (V_HW_4): Remove mode iterator. Now
8330 (vec_mergeh<mode>, vec_mergel<mode>): Use s390_expand_merge to
8331 emit vec merge pattern.
8333 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
8335 * config/aarch64/aarch64.c (aarch64_strip_extend_vec_half):
8337 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent cost of
8338 vec_select high-half from being added into Neon multiply
8340 * rtlanal.c (vec_series_highpart_p): Define.
8341 * rtlanal.h (vec_series_highpart_p): Declare.
8343 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
8345 * config/aarch64/aarch64.c (aarch64_strip_duplicate_vec_elt):
8347 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent
8348 vec_select cost from being added into Neon multiply cost.
8350 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
8352 * tree-vect-loop.c (vect_better_loop_vinfo_p): Detect cases in
8353 which old_loop_vinfo is an epilogue loop that handles a constant
8354 number of iterations.
8356 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
8358 * tree-vect-loop.c (vect_analyze_loop): Print a dump message
8359 when a reanalyzed loop fails to be cheaper than the current
8362 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
8364 * config/aarch64/aarch64.c: Fix a typo.
8366 2021-08-04 Vincent Lefèvre <vincent-gcc@vinc17.net>
8368 PR gcov-profile/101773
8369 * gcov-io.c (gcov_close): Check return code of a fclose.
8371 2021-08-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
8374 * dwarf2out.c (dwarf2out_assembly_start): Emit a dummy
8375 .file statement when needed.
8377 2021-08-04 Richard Biener <rguenther@suse.de>
8379 * tree-vect-data-refs.c (vect_check_gather_scatter):
8380 Include widening conversions only when the result is
8381 still handed by native gather or the current offset
8382 size not already matches the data size.
8383 Also succeed analysis in case there's no native support,
8384 noted by a IFN_LAST ifn and a NULL decl.
8385 (vect_analyze_data_refs): Always consider gathers.
8386 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
8387 Test for no IFN gather rather than decl gather.
8388 * tree-vect-stmts.c (vect_model_load_cost): Pass in the
8389 gather-scatter info and cost emulated gathers accordingly.
8390 (vect_truncate_gather_scatter_offset): Properly test for
8392 (vect_use_strided_gather_scatters_p): Likewise.
8393 (get_load_store_type): Handle emulated gathers and its
8395 (vectorizable_load): Likewise. Emulate them by extracting
8396 scalar offsets, doing scalar loads and a vector construct.
8398 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
8401 * expr.c (op_by_pieces_d::op_by_pieces_d): Add a max_pieces
8402 argument to set m_max_size.
8403 (move_by_pieces_d): Pass MOVE_MAX_PIECES to op_by_pieces_d.
8404 (store_by_pieces_d): Pass STORE_MAX_PIECES to op_by_pieces_d.
8405 (compare_by_pieces_d): Pass COMPARE_MAX_PIECES to op_by_pieces_d.
8407 2021-08-04 Roger Sayle <roger@nextmovesoftware.com>
8408 Marc Glisse <marc.glisse@inria.fr>
8410 * match.pd (bit_ior, bit_xor): Canonicalize (X*C1)|(X*C2) and
8411 (X*C1)^(X*C2) as X*(C1+C2), and related variants, using
8412 tree_nonzero_bits to ensure that operands are bit-wise disjoint.
8414 2021-08-04 Richard Biener <rguenther@suse.de>
8416 * tree-ssa-forwprop.c (pass_forwprop::execute): Split
8417 out code to decompose vector loads ...
8418 (optimize_vector_load): ... here. Generalize it to
8419 handle intermediate widening and TARGET_MEM_REF loads
8420 and apply it to loads with a supported vector mode as well.
8422 2021-08-04 Richard Biener <rguenther@suse.de>
8424 PR tree-optimization/101756
8425 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Make sure
8426 the result of the reduction epilogue is compatible to the original
8429 2021-08-04 liuhongt <hongtao.liu@intel.com>
8432 * config/i386/i386.md (peephole2): Refine predicate from
8433 register_operand to general_reg_operand.
8435 2021-08-04 Aldy Hernandez <aldyh@redhat.com>
8437 * gimple-range-path.h (path_range_query::dump): Mark override.
8439 2021-08-04 Richard Biener <rguenther@suse.de>
8441 PR tree-optimization/101769
8442 * tree-tailcall.c (eliminate_tail_call): Add the created loop
8443 for the first recursion and return it via the new output parameter.
8444 (optimize_tail_call): Pass through new output param.
8445 (tree_optimize_tail_calls_1): After creating all latches,
8446 add the created loop to the loop tree. Do not mark loops for fixup.
8448 2021-08-04 Martin Liska <mliska@suse.cz>
8450 * doc/invoke.texi: Document threader-mode param.
8452 2021-08-04 liuhongt <hongtao.liu@intel.com>
8454 * config/i386/sse.md (cond_fma<mode>): New expander.
8455 (cond_fms<mode>): Ditto.
8456 (cond_fnma<mode>): Ditto.
8457 (cond_fnms<mode>): Ditto.
8459 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
8461 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode>): Use && instead of &.
8463 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
8465 * config/rs6000/constraints.md: Remove "e" from the list of available
8466 constraint characters.
8468 2021-08-03 Eugene Rozenfeld <erozen@microsoft.com>
8470 PR gcov-profile/71672
8471 * auto-profile.c (afdo_indirect_call): Fix setup of the historgram value for indirect calls.
8473 2021-08-03 Paul A. Clarke <pc@us.ibm.com>
8475 * config/rs6000/smmintrin.h (_mm_minpos_epu16): New.
8477 2021-08-03 H.J. Lu <hjl.tools@gmail.com>
8479 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): In 64-bit mode,
8480 try XMM31 to avoid vzeroupper.
8482 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8484 * doc/invoke.texi: Document -mtune=neoverse-512tvb and
8485 -mcpu=neoverse-512tvb.
8486 * config/aarch64/aarch64-cores.def (neoverse-512tvb): New entry.
8487 * config/aarch64/aarch64-tune.md: Regenerate.
8488 * config/aarch64/aarch64.c (neoverse512tvb_sve_vector_cost)
8489 (neoverse512tvb_sve_issue_info, neoverse512tvb_vec_issue_info)
8490 (neoverse512tvb_vector_cost, neoverse512tvb_tunings): New structures.
8491 (aarch64_adjust_body_cost_sve): Handle -mtune=neoverse-512tvb.
8492 (aarch64_adjust_body_cost): Likewise.
8494 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8496 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Only
8497 record issue information for operations that occur in the
8500 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8502 * config/aarch64/aarch64.c (aarch64_multiply_add_p): Add a vec_flags
8503 parameter. Detect cases in which an Advanced SIMD MLA would almost
8504 certainly require a MOV.
8505 (aarch64_count_ops): Update accordingly.
8507 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8509 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction): New
8510 function, split out from...
8511 (aarch64_detect_vector_stmt_subtype): ...here.
8512 (aarch64_add_stmt_cost): Treat extracting element 0 as free.
8514 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8516 * config/aarch64/aarch64-protos.h (sve_vec_cost):
8517 Add gather_load_x32_cost and gather_load_x64_cost.
8518 * config/aarch64/aarch64.c (generic_sve_vector_cost)
8519 (a64fx_sve_vector_cost, neoversev1_sve_vector_cost): Update
8520 accordingly, using the values given by the scalar_load * number
8521 of elements calculation that we used previously.
8522 (aarch64_detect_vector_stmt_subtype): Use the new fields.
8524 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8526 * config/aarch64/aarch64.c (aarch64_adjust_body_cost_sve): New
8527 function, split out from...
8528 (aarch64_adjust_body_cost): ...here.
8530 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8532 * config/aarch64/fractional-cost.h: New file.
8533 * config/aarch64/aarch64.c: Include <algorithm> (indirectly)
8534 and cost_fraction.h.
8535 (vec_cost_fraction): New typedef.
8536 (aarch64_detect_scalar_stmt_subtype): Use it for statement costs.
8537 (aarch64_detect_vector_stmt_subtype): Likewise.
8538 (aarch64_sve_adjust_stmt_cost, aarch64_adjust_stmt_cost): Likewise.
8539 (aarch64_estimate_min_cycles_per_iter): Use vec_cost_fraction
8541 (aarch64_adjust_body_cost): Likewise.
8542 (aarch64_test_cost_fraction): New function.
8543 (aarch64_run_selftests): Call it.
8545 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8547 * config/aarch64/aarch64-protos.h (tune_params::sve_width): Turn
8549 * config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Update
8551 (aarch64_estimated_poly_value): Likewise. Use the least significant
8552 set bit for the minimum and likely values. Use the most significant
8553 set bit for the maximum value.
8555 2021-08-03 liuhongt <hongtao.liu@intel.com>
8557 * config/i386/sse.md (cond_<insn><mode>): New expander.
8558 (cond_mul<mode>): Ditto.
8560 2021-08-03 Kewen Lin <linkw@linux.ibm.com>
8562 * tree-cfg.c (move_sese_region_to_fn): Fix typos on dloop.
8564 2021-08-03 liuhongt <hongtao.liu@intel.com>
8566 * config/i386/sse.md (cond_<insn><mode>):New expander.
8567 (cond_mul<mode>): Ditto.
8568 (cond_div<mode>): Ditto.
8570 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
8572 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Also
8573 check stack_realign_needed for stack realignment.
8574 (ix86_legitimate_constant_p): Always allow CONST_WIDE_INT smaller
8575 than the largest integer supported by vector register.
8576 * config/i386/i386.h (MAX_MOVE_MAX): New. Set to 64.
8577 (MOVE_MAX): Set to bytes of the largest integer supported by
8579 (STORE_MAX_PIECES): New.
8581 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
8583 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
8584 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
8585 data from one memory location to another.
8587 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
8590 * config/i386/i386.c (TARGET_GEN_MEMSET_SCRATCH_RTX): New.
8592 2021-08-02 Aldy Hernandez <aldyh@redhat.com>
8594 PR tree-optimization/101724
8595 * params.opt: Remove --param=threader-iterative.
8596 * tree-ssa-threadbackward.c (pass_thread_jumps::execute): Remove
8599 2021-08-02 Tom de Vries <tdevries@suse.de>
8601 PR middle-end/101665
8602 * doc/extend.texi (nonnull attribute): Improve documentation.
8604 2021-08-02 Andrew Pinski <apinski@marvell.com>
8606 PR rtl-optimization/101683
8607 * rtlanal.c (may_trap_p_1): Handle UNSIGNED_FIX.
8609 2021-08-02 Roger Sayle <roger@nextmovesoftware.com>
8611 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern):
8612 Renamed from cond_removal_in_popcount_clz_ctz_pattern.
8613 Add support for BSWAP, FFS, PARITY and CLRSB builtins.
8614 (tree_ssa_phiop_worker): Update call to function above.
8616 2021-08-01 H.J. Lu <hjl.tools@gmail.com>
8619 * config/i386/i386.md (bsr_rex64_1_zext): New.
8620 (combine splitter for constant - clzll): Replace gen_bsr_rex64_1
8621 with gen_bsr_rex64_1_zext.
8623 2021-07-31 Jakub Jelinek <jakub@redhat.com>
8626 * config/i386/i386.md (bsr_rex64_1, bsr_1, bsr_zext_1): New
8627 define_insn patterns.
8628 (*bsr_rex64_2, *bsr_2): New define_insn_and_split patterns.
8629 Add combine splitters for constant - clz.
8630 (clz<mode>2): Use a temporary pseudo for bsr result.
8632 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
8634 * config/rs6000/smmintrin.h (_mm_floor_pd, _mm_floor_ps,
8635 _mm_floor_sd, _mm_floor_ss): New.
8637 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
8639 * config/rs6000/smmintrin.h (_mm_ceil_pd, _mm_ceil_ps,
8640 _mm_ceil_sd, _mm_ceil_ss): New.
8642 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
8644 * config/rs6000/smmintrin.h (_mm_blend_pd, _mm_blendv_pd,
8645 _mm_blend_ps, _mm_blendv_ps): New.
8647 2021-07-30 Roger Sayle <roger@nextmovesoftware.com>
8648 Uroš Bizjak <ubizjak@gmail.com>
8650 * config/i386/i386.md (*dec_cmov<mode>): New define_insn_and_split
8651 to generate a conditional move using the carry flag after sub $1.
8652 (peephole2): Eliminate a register-to-register move by inverting
8653 the condition of a conditional move.
8655 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
8657 * config/mmix/mmix.md ("call", "call_value", "*call_real")
8658 ("*call_value_real"): Don't generate rtx mentioning the generic
8659 operands 1 and 2 to "call", and similarly for "call_value".
8660 * config/mmix/mmix.c (mmix_print_operand_punct_valid_p)
8661 (mmix_print_operand): Use '!' instead of 'p'.
8663 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
8665 * doc/md.texi (call): Correct information about operand 2.
8666 * config/mmix/mmix.md ("call", "call_value"): Remove fixed FIXMEs.
8668 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
8670 * range-op.cc (operator_trunc_mod::wi_fold): Fold constants.
8672 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
8674 * range-op.cc (operator_div::wi_fold): Return UNDEFINED for [0, 0] divisor.
8676 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
8678 * gimple-range-cache.cc (*::set_bb_range): Change const basic_block to
8680 (*::get_bb_range): Ditto.
8681 (*::bb_range_p): Ditto.
8682 * gimple-range-cache.h: Change prototypes.
8684 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
8687 * builtins.c (builtin_memcpy_read_str): Change the mode argument
8688 from scalar_int_mode to fixed_size_mode.
8689 (builtin_strncpy_read_str): Likewise.
8690 (gen_memset_value_from_prev): New function.
8691 (builtin_memset_read_str): Change the mode argument from
8692 scalar_int_mode to fixed_size_mode. Use gen_memset_value_from_prev
8693 and support CONST_VECTOR.
8694 (builtin_memset_gen_str): Likewise.
8695 (try_store_by_multiple_pieces): Use by_pieces_constfn to declare
8697 * builtins.h (builtin_strncpy_read_str): Replace scalar_int_mode
8698 with fixed_size_mode.
8699 (builtin_memset_read_str): Likewise.
8700 * expr.c (widest_int_mode_for_size): Renamed to ...
8701 (widest_fixed_size_mode_for_size): Add a bool argument to
8702 indicate if QI vector mode can be used.
8703 (by_pieces_ninsns): Call widest_fixed_size_mode_for_size
8704 instead of widest_int_mode_for_size.
8705 (pieces_addr::adjust): Change the mode argument from
8706 scalar_int_mode to fixed_size_mode.
8707 (op_by_pieces_d): Make m_len read-only. Add a bool member,
8708 m_qi_vector_mode, to indicate that QI vector mode can be used.
8709 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
8710 initialize m_qi_vector_mode. Call widest_fixed_size_mode_for_size
8711 instead of widest_int_mode_for_size.
8712 (op_by_pieces_d::get_usable_mode): Change the mode argument from
8713 scalar_int_mode to fixed_size_mode. Call
8714 widest_fixed_size_mode_for_size instead of
8715 widest_int_mode_for_size.
8716 (op_by_pieces_d::smallest_fixed_size_mode_for_size): New member
8717 function to return the smallest integer or QI vector mode.
8718 (op_by_pieces_d::run): Call widest_fixed_size_mode_for_size
8719 instead of widest_int_mode_for_size. Call
8720 smallest_fixed_size_mode_for_size instead of
8721 smallest_int_mode_for_size.
8722 (store_by_pieces_d::store_by_pieces_d): Add a bool argument to
8723 indicate that QI vector mode can be used and pass it to
8724 op_by_pieces_d::op_by_pieces_d.
8725 (can_store_by_pieces): Call widest_fixed_size_mode_for_size
8726 instead of widest_int_mode_for_size. Pass memsetp to
8727 widest_fixed_size_mode_for_size to support QI vector mode.
8728 Allow all CONST_VECTORs for memset if vec_duplicate is supported.
8729 (store_by_pieces): Pass memsetp to
8730 store_by_pieces_d::store_by_pieces_d.
8731 (clear_by_pieces_1): Removed.
8732 (clear_by_pieces): Replace clear_by_pieces_1 with
8733 builtin_memset_read_str and pass true to store_by_pieces_d to
8734 support vector mode broadcast.
8735 (string_cst_read_str): Change the mode argument from
8736 scalar_int_mode to fixed_size_mode.
8737 * expr.h (by_pieces_constfn): Change scalar_int_mode to
8739 (by_pieces_prev): Likewise.
8740 * rtl.h (lowpart_subreg_regno): New.
8741 * rtlanal.c (lowpart_subreg_regno): New. A wrapper around
8742 simplify_subreg_regno.
8743 * target.def (gen_memset_scratch_rtx): New hook.
8744 * doc/tm.texi.in: Add TARGET_GEN_MEMSET_SCRATCH_RTX.
8745 * doc/tm.texi: Regenerated.
8747 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
8750 * config/mips/mips.c (mips_atomic_assign_expand_fenv): Use
8751 TARGET_EXPR instead of MODIFY_EXPR.
8753 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
8756 * config/mips/mips-protos.h (mips_expand_vec_cmp_expr): Declare.
8757 * config/mips/mips.c (mips_expand_vec_cmp_expr): New function.
8758 * config/mips/mips-msa.md (vec_cmp<MSA:mode><mode_i>): New
8760 (vec_cmpu<IMSA:mode><mode_i>): New expander.
8762 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
8765 * config/i386/i386-options.c (ix86_option_override_internal):
8766 Don't enable LZCNT/POPCNT if they have been disabled explicitly.
8768 2021-07-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
8771 * config/arm/arm_neon.h (vld1_p64): Replace call to builtin by
8772 explicitly dereferencing __a.
8773 (vld1_s64): Likewise.
8774 (vld1_u64): Likewise.
8775 * config/arm/arm_neon_builtins.def (vld1): Remove entry for di
8776 and change to VAR13.
8778 2021-07-30 Aldy Hernandez <aldyh@redhat.com>
8780 * gimple-loop-versioning.cc (lv_dom_walker::lv_dom_walker): Remove
8781 use of m_range_analyzer.
8782 (loop_versioning::lv_dom_walker::before_dom_children): Same.
8783 (loop_versioning::lv_dom_walker::after_dom_children): Remove.
8784 (loop_versioning::prune_loop_conditions): Replace vr_values use
8785 with range_query interface.
8786 (pass_loop_versioning::execute): Use ranger.
8788 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
8791 * ipa-devirt.c (ipa_odr_read_section): Compare the precision of
8792 enum values, and emit a warning if they mismatch.
8794 2021-07-30 Kewen Lin <linkw@linux.ibm.com>
8796 * cfgloop.h (as_const): New function.
8797 (class loop_iterator): Rename to ...
8798 (class loops_list): ... this.
8799 (loop_iterator::next): Rename to ...
8800 (loops_list::Iter::fill_curr_loop): ... this and adjust.
8801 (loop_iterator::loop_iterator): Rename to ...
8802 (loops_list::loops_list): ... this and adjust.
8803 (loops_list::Iter): New class.
8804 (loops_list::iterator): New type.
8805 (loops_list::const_iterator): New type.
8806 (loops_list::begin): New function.
8807 (loops_list::end): Likewise.
8808 (loops_list::begin const): Likewise.
8809 (loops_list::end const): Likewise.
8810 (FOR_EACH_LOOP): Remove.
8811 (FOR_EACH_LOOP_FN): Remove.
8812 * cfgloop.c (flow_loops_dump): Adjust FOR_EACH_LOOP* with range-based
8813 for loop with loops_list instance.
8814 (sort_sibling_loops): Likewise.
8815 (disambiguate_loops_with_multiple_latches): Likewise.
8816 (verify_loop_structure): Likewise.
8817 * cfgloopmanip.c (create_preheaders): Likewise.
8818 (force_single_succ_latches): Likewise.
8819 * config/aarch64/falkor-tag-collision-avoidance.c
8820 (execute_tag_collision_avoidance): Likewise.
8821 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Likewise.
8822 * config/s390/s390.c (s390_adjust_loops): Likewise.
8823 * doc/loop.texi: Likewise.
8824 * gimple-loop-interchange.cc (pass_linterchange::execute): Likewise.
8825 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
8826 * gimple-loop-versioning.cc (loop_versioning::analyze_blocks): Likewise.
8827 (loop_versioning::make_versioning_decisions): Likewise.
8828 * gimple-ssa-split-paths.c (split_paths): Likewise.
8829 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
8830 * graphite.c (canonicalize_loop_form): Likewise.
8831 (graphite_transform_loops): Likewise.
8832 * ipa-fnsummary.c (analyze_function_body): Likewise.
8833 * ipa-pure-const.c (analyze_function): Likewise.
8834 * loop-doloop.c (doloop_optimize_loops): Likewise.
8835 * loop-init.c (loop_optimizer_finalize): Likewise.
8836 (fix_loop_structure): Likewise.
8837 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
8838 (move_loop_invariants): Likewise.
8839 * loop-unroll.c (decide_unrolling): Likewise.
8840 (unroll_loops): Likewise.
8841 * modulo-sched.c (sms_schedule): Likewise.
8842 * predict.c (predict_loops): Likewise.
8843 (pass_profile::execute): Likewise.
8844 * profile.c (branch_prob): Likewise.
8845 * sel-sched-ir.c (sel_finish_pipelining): Likewise.
8846 (sel_find_rgns): Likewise.
8847 * tree-cfg.c (replace_loop_annotate): Likewise.
8848 (replace_uses_by): Likewise.
8849 (move_sese_region_to_fn): Likewise.
8850 * tree-if-conv.c (pass_if_conversion::execute): Likewise.
8851 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
8852 * tree-parloops.c (parallelize_loops): Likewise.
8853 * tree-predcom.c (tree_predictive_commoning): Likewise.
8854 * tree-scalar-evolution.c (scev_initialize): Likewise.
8855 (scev_reset): Likewise.
8856 * tree-ssa-dce.c (find_obviously_necessary_stmts): Likewise.
8857 * tree-ssa-live.c (remove_unused_locals): Likewise.
8858 * tree-ssa-loop-ch.c (ch_base::copy_headers): Likewise.
8859 * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
8860 (tree_ssa_lim_initialize): Likewise.
8861 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Likewise.
8862 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize): Likewise.
8863 * tree-ssa-loop-manip.c (get_loops_exits): Likewise.
8864 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations): Likewise.
8865 (free_numbers_of_iterations_estimates): Likewise.
8866 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Likewise.
8867 * tree-ssa-loop-split.c (tree_ssa_split_loops): Likewise.
8868 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Likewise.
8869 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
8870 (pass_scev_cprop::execute): Likewise.
8871 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
8872 * tree-ssa-sccvn.c (do_rpo_vn): Likewise.
8873 * tree-ssa-threadupdate.c
8874 (jump_thread_path_registry::thread_through_all_blocks): Likewise.
8875 * tree-vectorizer.c (vectorize_loops): Likewise.
8876 * tree-vrp.c (vrp_asserts::find_assert_locations): Likewise.
8878 2021-07-29 Hans-Peter Nilsson <hp@bitrange.com>
8880 * config/mmix/mmix.c (mmix_function_arg_1): Avoid
8881 generating a VOIDmode register for e.g the
8882 function_arg_info::end_marker.
8884 2021-07-29 Jeff Law <jeffreyalaw@gmail.com>
8886 * config/h8300/h8300-modes.def: Add CCZ, CCV and CCC, drop CCZNV.
8887 * config/h8300/h8300.md (H8cc mode iterator): Add CCZ.
8888 (cc mode_attr): Similarly.
8889 (ccz subst_attr): Similarly.
8890 * config/h8300/jumpcall.md: Add new patterns for branch-on-bit.
8891 * config/h8300/testcompare.md: Remove various cc0 based patterns
8892 that had been commented out. Add pattern to set CCZ from a bit
8895 2021-07-29 Thomas Schwinge <thomas@codesourcery.com>
8896 Julian Brown <julian@codesourcery.com>
8897 Kwok Cheung Yeung <kcy@codesourcery.com>
8899 * omp-offload.c (oacc_loop_xform_head_tail, oacc_loop_process):
8900 'update_stmt' after modification.
8901 (pass_oacc_loop_designation): New function, extracted out of...
8902 (pass_oacc_device_lower): ... this.
8903 (pass_data_oacc_loop_designation, pass_oacc_loop_designation)
8904 (make_pass_oacc_loop_designation): New
8905 * passes.def: Add it.
8906 * tree-parloops.c (create_parallel_loop): Adjust.
8907 * tree-pass.h (make_pass_oacc_loop_designation): New.
8909 2021-07-29 Aldy Hernandez <aldyh@redhat.com>
8911 * flag-types.h (enum threader_mode): New.
8912 * params.opt: Add entry for --param=threader-mode.
8913 * tree-ssa-threadbackward.c (THREADER_ITERATIVE_MODE): New.
8914 (class back_threader): New.
8915 (back_threader::back_threader): New.
8916 (back_threader::~back_threader): New.
8917 (back_threader::maybe_register_path): New.
8918 (back_threader::find_taken_edge): New.
8919 (back_threader::find_taken_edge_switch): New.
8920 (back_threader::find_taken_edge_cond): New.
8921 (back_threader::resolve_def): New.
8922 (back_threader::resolve_phi): New.
8923 (back_threader::find_paths_to_names): New.
8924 (back_threader::find_paths): New.
8927 (thread_jumps::find_jump_threads_backwards): Call ranger threader.
8928 (thread_jumps::find_jump_threads_backwards_with_ranger): New.
8929 (pass_thread_jumps::execute): Abstract out code...
8930 (try_thread_blocks): ...here.
8931 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
8932 Abstract out threading candidate code to...
8933 (single_succ_to_potentially_threadable_block): ...here.
8934 * tree-ssa-threadedge.h (single_succ_to_potentially_threadable_block):
8936 * tree-ssa-threadupdate.c (register_jump_thread): Return boolean.
8937 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
8938 Return bool from register_jump_thread.
8940 2021-07-29 Andreas Krebbel <krebbel@linux.ibm.com>
8942 * target.def: in0 and in1 do not need to be registers.
8943 * doc/tm.texi: Regenerate.
8945 2021-07-29 liuhongt <hongtao.liu@intel.com>
8948 * config/i386/i386.c (ix86_widen_mult_cost): New function.
8949 (ix86_add_stmt_cost): Use ix86_widen_mult_cost for
8952 2021-07-29 Jiufu Guo <guojiufu@linux.ibm.com>
8955 * config/rs6000/rs6000.c (TARGET_PREFERRED_DOLOOP_MODE): New hook.
8956 (rs6000_preferred_doloop_mode): New hook.
8957 * doc/tm.texi: Regenerate.
8958 * doc/tm.texi.in: Add hook preferred_doloop_mode.
8959 * target.def (preferred_doloop_mode): New hook.
8960 * targhooks.c (default_preferred_doloop_mode): New hook.
8961 * targhooks.h (default_preferred_doloop_mode): New hook.
8962 * tree-ssa-loop-ivopts.c (compute_doloop_base_on_mode): New function.
8963 (add_iv_candidate_for_doloop): Call targetm.preferred_doloop_mode
8964 and compute_doloop_base_on_mode.
8966 2021-07-28 Martin Sebor <msebor@redhat.com>
8968 PR middle-end/101494
8969 * tree-ssa-uninit.c (maybe_warn_operand): Correct object offset
8970 and size computation.
8972 2021-07-28 Martin Sebor <msebor@redhat.com>
8974 PR middle-end/101601
8975 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Remove
8977 Handle pointers to functions.
8979 2021-07-28 Martin Sebor <msebor@redhat.com>
8981 * Makefile.in (OBJS): Add gimple-ssa-warn-access.o and pointer-query.o.
8982 * attribs.h (fndecl_dealloc_argno): Move fndecl_dealloc_argno to tree.h.
8983 * builtins.c (compute_objsize_r): Move to pointer-query.cc.
8984 (access_ref::access_ref): Same.
8985 (access_ref::phi): Same.
8986 (access_ref::get_ref): Same.
8987 (access_ref::size_remaining): Same.
8988 (access_ref::offset_in_range): Same.
8989 (access_ref::add_offset): Same.
8990 (access_ref::inform_access): Same.
8991 (ssa_name_limit_t::visit_phi): Same.
8992 (ssa_name_limit_t::leave_phi): Same.
8993 (ssa_name_limit_t::next): Same.
8994 (ssa_name_limit_t::next_phi): Same.
8995 (ssa_name_limit_t::~ssa_name_limit_t): Same.
8996 (pointer_query::pointer_query): Same.
8997 (pointer_query::get_ref): Same.
8998 (pointer_query::put_ref): Same.
8999 (pointer_query::flush_cache): Same.
9000 (warn_string_no_nul): Move to gimple-ssa-warn-access.cc.
9001 (check_nul_terminated_array): Same.
9002 (unterminated_array): Same.
9003 (maybe_warn_for_bound): Same.
9004 (check_read_access): Same.
9005 (warn_for_access): Same.
9006 (get_size_range): Same.
9007 (check_access): Same.
9008 (gimple_call_alloc_size): Move to tree.c.
9009 (gimple_parm_array_size): Move to pointer-query.cc.
9010 (get_offset_range): Same.
9011 (gimple_call_return_array): Same.
9012 (handle_min_max_size): Same.
9013 (handle_array_ref): Same.
9014 (handle_mem_ref): Same.
9015 (compute_objsize): Same.
9016 (gimple_call_alloc_p): Move to gimple-ssa-warn-access.cc.
9017 (call_dealloc_argno): Same.
9018 (fndecl_dealloc_argno): Same.
9019 (new_delete_mismatch_p): Same.
9020 (matching_alloc_calls_p): Same.
9021 (warn_dealloc_offset): Same.
9022 (maybe_emit_free_warning): Same.
9023 * builtins.h (check_nul_terminated_array): Move to
9024 gimple-ssa-warn-access.h.
9025 (check_nul_terminated_array): Same.
9026 (warn_string_no_nul): Same.
9027 (unterminated_array): Same.
9028 (class ssa_name_limit_t): Same.
9029 (class pointer_query): Same.
9030 (struct access_ref): Same.
9031 (class range_query): Same.
9032 (struct access_data): Same.
9033 (gimple_call_alloc_size): Same.
9034 (gimple_parm_array_size): Same.
9035 (compute_objsize): Same.
9036 (class access_data): Same.
9037 (maybe_emit_free_warning): Same.
9038 * calls.c (initialize_argument_information): Remove call to
9039 maybe_emit_free_warning.
9040 * gimple-array-bounds.cc: Include new header..
9041 * gimple-fold.c: Same.
9042 * gimple-ssa-sprintf.c: Same.
9043 * gimple-ssa-warn-restrict.c: Same.
9044 * passes.def: Add pass_warn_access.
9045 * tree-pass.h (make_pass_warn_access): Declare.
9046 * tree-ssa-strlen.c: Include new headers.
9047 * tree.c (fndecl_dealloc_argno): Move here from builtins.c.
9048 * tree.h (fndecl_dealloc_argno): Move here from attribs.h.
9049 * gimple-ssa-warn-access.cc: New file.
9050 * gimple-ssa-warn-access.h: New file.
9051 * pointer-query.cc: New file.
9052 * pointer-query.h: New file.
9054 2021-07-28 Jakub Jelinek <jakub@redhat.com>
9056 PR middle-end/101624
9057 * ubsan.c (maybe_instrument_pointer_overflow,
9058 instrument_object_size): Only test DECL_REGISTER on VAR_DECLs,
9059 PARM_DECLs or RESULT_DECLs.
9060 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
9062 2021-07-28 Jakub Jelinek <jakub@redhat.com>
9064 PR middle-end/101642
9065 * match.pd (bswap16 (x) == bswap16 (y)): Cast both operands
9066 to type of bswap16 for comparison.
9067 (bswap16 (x) == cst): Cast bswap16 operand to type of cst.
9069 2021-07-28 Richard Biener <rguenther@suse.de>
9071 PR tree-optimization/101615
9072 * tree-vect-slp.c (vect_optimize_slp): Materialize permutes
9073 at CTOR SLP graph entries.
9075 2021-07-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9077 * config/aarch64/aarch64.md (*extend<SHORT:mode><GPI:mode>2_aarch64):
9078 Add "r,w" alternative.
9080 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
9083 * config/i386/i386.c (ix86_avx_u128_mode_needed): Don't set
9084 AVX_U128_DIRTY when all bits are zero.
9086 2021-07-28 Richard Biener <rguenther@suse.de>
9088 PR tree-optimization/101615
9089 * tree-vect-slp.c (vect_optimize_slp): Pre-existing vector
9090 external nodes cannot be permuted so make them perm_out 0.
9092 2021-07-28 Andrew Stubbs <ams@codesourcery.com>
9095 * config.in: Regenerate.
9096 * config/gcn/gcn-hsa.h (A_FIJI): New define.
9097 (A_900): New define.
9098 (A_906): New define.
9099 (A_908): New define.
9100 (ASM_SPEC): Use A_FIJI, A_900, A_906 and A_908.
9101 * config/gcn/gcn.c (output_file_start): Adjust attributes according
9102 to the assembler capabilities.
9103 * config/gcn/mkoffload.c (main): Likewise.
9104 * configure: Regenerate.
9105 * configure.ac: Add tests for LLVM assembler attribute features.
9107 2021-07-28 Andrew MacLeod <amacleod@redhat.com>
9109 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Check for
9110 cond_false and cond_true on branches.
9112 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
9114 * config/aarch64/aarch64.c (aarch64_gen_adjusted_ldpstp): use
9117 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
9119 * alias.c (init_alias_analysis): Don't skip prologue/epilogue.
9121 2021-07-28 Jakub Jelinek <jakub@redhat.com>
9124 * config/i386/sse.md (vashr<mode>3): Split into vashrv8di3 expander
9125 and vashrv4di3 expander, where the latter requires just TARGET_AVX2
9126 and has special !TARGET_AVX512VL expansion.
9127 (vashrv2di3<mask_name>): Rename to ...
9128 (vashrv2di3): ... this. Change condition to TARGET_XOP || TARGET_AVX2
9129 and add special !TARGET_XOP && !TARGET_AVX512VL expansion.
9131 2021-07-28 Martin Uecker <muecker@gwdg.de>
9133 * calls.c (maybe_warn_rdwr_sizes): Correct argument
9134 numbers in warning that were switched.
9136 2021-07-28 Kewen Lin <linkw@linux.ibm.com>
9138 PR tree-optimization/101596
9139 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Fix wrong check
9140 by using new_type's precision instead.
9142 2021-07-28 liuhongt <hongtao.liu@intel.com>
9145 * config/i386/i386.h (processor_costs): Add new member
9147 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
9148 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
9149 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
9150 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
9151 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
9152 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
9153 generic_cost, core_cost): Initialize integer_to_sse same value
9155 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
9156 * config/i386/i386.c (ix86_builtin_vectorization_cost):
9157 Use integer_to_sse instead of sse_op to calculate the cost of
9160 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
9162 * config/rs6000/rs6000-gen-builtins.c (write_ovld_static_init): New
9164 (write_init_file): Call write_ovld_static_init.
9166 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
9168 * config/rs6000/rs6000-gen-builtins.c (write_bif_static_init): New
9170 (write_init_file): Call write_bif_static_init.
9172 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
9174 * config/rs6000/rs6000-gen-builtins.c (typemap): New struct.
9175 (TYPE_MAP_SIZE): New macro.
9176 (type_map): New initialized variable.
9177 (typemap_cmp): New function.
9178 (write_type_node): Likewise.
9179 (write_fntype_init): Implement.
9181 2021-07-27 Martin Sebor <msebor@redhat.com>
9183 PR tree-optimization/101584
9184 * tree-ssa-uninit.c (builtin_call_nomodifying_p): New function.
9185 (check_defs): Call it.
9187 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
9189 * tree-ssa-dom.c (dom_jump_threader_simplifier):
9190 Put avail_exprs_stack in the class, instead of passing it to
9191 jump_threader_simplifier.
9192 (dom_jump_threader_simplifier::simplify): Add state argument.
9193 (dom_opt_dom_walker): Add state.
9194 (pass_dominator::execute): Pass state to threader.
9195 (dom_opt_dom_walker::before_dom_children): Use state.
9196 * tree-ssa-threadedge.c (jump_threader::jump_threader): Replace
9198 (jump_threader::record_temporary_equivalences_from_phis):
9199 Register equivalences through the state variable.
9200 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
9201 Record ranges in a statement through the state variable.
9202 (jump_threader::simplify_control_stmt_condition): Pass state to
9204 (jump_threader::simplify_control_stmt_condition_1): Same.
9205 (jump_threader::thread_around_empty_blocks): Remove obsolete
9207 (jump_threader::thread_through_normal_block): Record equivalences
9208 on edge through the state variable.
9209 (jump_threader::thread_across_edge): Abstract state pushing.
9210 (jt_state::jt_state): New.
9211 (jt_state::push): New.
9212 (jt_state::pop): New.
9213 (jt_state::register_equiv): New.
9214 (jt_state::record_ranges_from_stmt): New.
9215 (jt_state::register_equivs_on_edge): New.
9216 (jump_threader_simplifier::jump_threader_simplifier): Move from
9218 (jump_threader_simplifier::simplify): Add state argument.
9219 * tree-ssa-threadedge.h (class jt_state): New.
9220 (class jump_threader): Add state to constructor.
9221 (class jump_threader_simplifier): Add state to simplify. Remove
9222 avail_exprs_stack from class.
9223 * tree-vrp.c (vrp_jump_threader_simplifier::simplify): Add state
9225 (vrp_jump_threader::vrp_jump_threader): Add state.
9226 (vrp_jump_threader::~vrp_jump_threader): Cleanup state.
9228 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
9230 * Makefile.in (OBJS): Add gimple-range-path.o.
9231 * gimple-range-path.cc: New file.
9232 * gimple-range-path.h: New file.
9234 2021-07-27 Jonathan Wright <jonathan.wright@arm.com>
9236 * config/aarch64/aarch64-simd.md: Push sign/zero-extension
9237 inside vec_duplicate for all patterns.
9238 * simplify-rtx.c (simplify_context::simplify_unary_operation_1):
9239 Push sign/zero-extension inside vec_duplicate.
9241 2021-07-27 Richard Biener <rguenther@suse.de>
9243 PR tree-optimization/101573
9244 * tree-ssa-uninit.c (warn_uninit_phi_uses): New function
9245 looking at uninitialized PHI arg defs in some constrained cases.
9246 (warn_uninitialized_vars): Call it.
9247 (execute_early_warn_uninitialized): Calculate dominators.
9249 2021-07-27 Richard Biener <rguenther@suse.de>
9251 PR tree-optimization/39821
9252 * tree-vect-stmts.c (vect_model_promotion_demotion_cost): Use
9253 vector_stmt for widening arithmetic.
9254 (vectorizable_conversion): Adjust.
9256 2021-07-27 Martin Jambor <mjambor@suse.cz>
9258 * cgraph.h (ipa_replace_map): New field force_load_ref.
9259 * ipa-prop.h (ipa_param_descriptor): Reduce precision of move_cost,
9260 aded new flag load_dereferenced, adjusted comments.
9261 (ipa_get_param_dereferenced): New function.
9262 (ipa_set_param_dereferenced): Likewise.
9263 * cgraphclones.c (cgraph_node::create_virtual_clone): Follow it.
9264 * ipa-cp.c: Include gimple.h.
9265 (ipcp_discover_new_direct_edges): Take into account dereferenced flag.
9266 (get_replacement_map): New parameter force_load_ref, set the
9267 appropriate flag in ipa_replace_map if set.
9268 (struct symbol_and_index_together): New type.
9269 (adjust_refs_in_act_callers): New function.
9270 (adjust_references_in_caller): Likewise.
9271 (create_specialized_node): When appropriate, call
9272 adjust_references_in_caller and force only load references.
9273 * ipa-prop.c (load_from_dereferenced_name): New function.
9274 (ipa_analyze_controlled_uses): Also detect loads from a
9275 dereference, harden testing of call statements.
9276 (ipa_write_node_info): Stream the dereferenced flag.
9277 (ipa_read_node_info): Likewise.
9278 (ipa_set_jf_constant): Also create refdesc when jump function
9279 references a variable.
9280 (cgraph_node_for_jfunc): Rename to symtab_node_for_jfunc, work
9281 also on references of variables and return a symtab_node. Adjust
9283 (propagate_controlled_uses): Also remove references to VAR_DECLs.
9285 2021-07-27 Jakub Jelinek <jakub@redhat.com>
9287 PR middle-end/101586
9288 * gimple-fold.c (clear_padding_type): Ignore FIELD_DECLs with byte
9289 positions above or equal to sz except for diagnostics of flexible
9292 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
9294 PR tree-optimization/78888
9295 * gimple-range-fold.cc (get_letter_range): New.
9296 (fold_using_range::range_of_builtin_call): Call get_letter_range.
9298 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
9300 PR tree-optimization/78888
9301 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Add cases
9302 for CFN_BUILT_IN_TOUPPER and CFN_BUILT_IN_TOLOWER.
9304 2021-07-26 Roger Sayle <roger@nextmovesoftware.com>
9305 Marc Glisse <marc.glisse@inria.fr>
9307 * match.pd (rotate): Simplify equality/inequality of rotations.
9308 (bswap): Simplify equality/inequality tests of byte swapping.
9310 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
9312 * range-op.cc (operator_bitwise_xor::op1_op2_relation_effect):
9315 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
9317 * range-op.cc (operator_lshift::fold_range): Pass rel to
9318 base class fold_range.
9319 (operator_rshift::fold_range): Same.
9321 2021-07-26 Ashimida <ashimida@linux.alibaba.com>
9324 * toplev.h (min_align_loops_log): Remove declaration.
9325 (min_align_jumps_log, min_align_labels_log): Likewise.
9326 (min_align_functions_log): Likewise.
9328 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
9330 * tree-vrp.c (vrp_simplify_cond_using_ranges): Rename vr_values
9332 (execute_vrp): Abstract out simplification of conditionals...
9333 (simplify_casted_conds): ...here.
9335 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
9337 * gimple-array-bounds.cc (array_bounds_checker::get_value_range):
9338 Add gimple argument.
9339 (array_bounds_checker::check_array_ref): Same.
9340 (array_bounds_checker::check_addr_expr): Same.
9341 (array_bounds_checker::check_array_bounds): Pass statement to
9342 check_array_bounds and check_addr_expr.
9343 * gimple-array-bounds.h (check_array_bounds): Add gimple argument.
9344 (check_addr_expr): Same.
9345 (get_value_range): Same.
9347 2021-07-26 Tamar Christina <tamar.christina@arm.com>
9349 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): Rename to..
9350 (sdot_prod, udot_prod): ... This.
9351 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): Merged
9353 (<sur>dot_prod<vsi2qi>): ... this.
9354 (aarch64_<sur>dot_lane<vsi2qi>, aarch64_<sur>dot_laneq<vsi2qi>):
9355 Change operands order.
9356 (<sur>sadv16qi): Use new operands order.
9357 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32,
9358 vdotq_s32): Use new RTL ordering.
9360 2021-07-26 Tamar Christina <tamar.christina@arm.com>
9362 * config/aarch64/aarch64-builtins.c (TYPES_TERNOP_SUSS,
9363 aarch64_types_ternop_suss_qualifiers): New.
9364 * config/aarch64/aarch64-simd-builtins.def (usdot_prod): Use it.
9365 * config/aarch64/aarch64-simd.md (usdot_prod<vsi2qi>): Re-organize RTL.
9366 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Use it.
9368 2021-07-23 Jakub Jelinek <jakub@redhat.com>
9370 PR rtl-optimization/101562
9371 * expmed.c (store_integral_bit_field): Only use movstrict_optab
9372 if the operand isn't paradoxical.
9374 2021-07-23 Aldy Hernandez <aldyh@redhat.com>
9376 * gimple-array-bounds.h (class array_bounds_checker): Change
9377 ranges type to range_query.
9379 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9381 * config/aarch64/arm_neon.h (vst1_s64_x2): Use
9382 __builtin_memcpy instead of constructing
9383 __builtin_aarch64_simd_oi one vector at a time.
9384 (vst1_u64_x2): Likewise.
9385 (vst1_f64_x2): Likewise.
9386 (vst1_s8_x2): Likewise.
9387 (vst1_p8_x2): Likewise.
9388 (vst1_s16_x2): Likewise.
9389 (vst1_p16_x2): Likewise.
9390 (vst1_s32_x2): Likewise.
9391 (vst1_u8_x2): Likewise.
9392 (vst1_u16_x2): Likewise.
9393 (vst1_u32_x2): Likewise.
9394 (vst1_f16_x2): Likewise.
9395 (vst1_f32_x2): Likewise.
9396 (vst1_p64_x2): Likewise.
9397 (vst1q_s8_x2): Likewise.
9398 (vst1q_p8_x2): Likewise.
9399 (vst1q_s16_x2): Likewise.
9400 (vst1q_p16_x2): Likewise.
9401 (vst1q_s32_x2): Likewise.
9402 (vst1q_s64_x2): Likewise.
9403 (vst1q_u8_x2): Likewise.
9404 (vst1q_u16_x2): Likewise.
9405 (vst1q_u32_x2): Likewise.
9406 (vst1q_u64_x2): Likewise.
9407 (vst1q_f16_x2): Likewise.
9408 (vst1q_f32_x2): Likewise.
9409 (vst1q_f64_x2): Likewise.
9410 (vst1q_p64_x2): Likewise.
9412 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9414 * config/aarch64/arm_neon.h (vst1_s64_x3): Use
9415 __builtin_memcpy instead of constructing
9416 __builtin_aarch64_simd_ci one vector at a time.
9417 (vst1_u64_x3): Likewise.
9418 (vst1_f64_x3): Likewise.
9419 (vst1_s8_x3): Likewise.
9420 (vst1_p8_x3): Likewise.
9421 (vst1_s16_x3): Likewise.
9422 (vst1_p16_x3): Likewise.
9423 (vst1_s32_x3): Likewise.
9424 (vst1_u8_x3): Likewise.
9425 (vst1_u16_x3): Likewise.
9426 (vst1_u32_x3): Likewise.
9427 (vst1_f16_x3): Likewise.
9428 (vst1_f32_x3): Likewise.
9429 (vst1_p64_x3): Likewise.
9430 (vst1q_s8_x3): Likewise.
9431 (vst1q_p8_x3): Likewise.
9432 (vst1q_s16_x3): Likewise.
9433 (vst1q_p16_x3): Likewise.
9434 (vst1q_s32_x3): Likewise.
9435 (vst1q_s64_x3): Likewise.
9436 (vst1q_u8_x3): Likewise.
9437 (vst1q_u16_x3): Likewise.
9438 (vst1q_u32_x3): Likewise.
9439 (vst1q_u64_x3): Likewise.
9440 (vst1q_f16_x3): Likewise.
9441 (vst1q_f32_x3): Likewise.
9442 (vst1q_f64_x3): Likewise.
9443 (vst1q_p64_x3): Likewise.
9445 2021-07-23 H.J. Lu <hjl.tools@gmail.com>
9448 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): Don't return
9449 hard register when LRA is in progress.
9451 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9453 * config/aarch64/arm_neon.h (vst1_s8_x4): Use
9454 __builtin_memcpy instead of using a union.
9455 (vst1q_s8_x4): Likewise.
9456 (vst1_s16_x4): Likewise.
9457 (vst1q_s16_x4): Likewise.
9458 (vst1_s32_x4): Likewise.
9459 (vst1q_s32_x4): Likewise.
9460 (vst1_u8_x4): Likewise.
9461 (vst1q_u8_x4): Likewise.
9462 (vst1_u16_x4): Likewise.
9463 (vst1q_u16_x4): Likewise.
9464 (vst1_u32_x4): Likewise.
9465 (vst1q_u32_x4): Likewise.
9466 (vst1_f16_x4): Likewise.
9467 (vst1q_f16_x4): Likewise.
9468 (vst1_f32_x4): Likewise.
9469 (vst1q_f32_x4): Likewise.
9470 (vst1_p8_x4): Likewise.
9471 (vst1q_p8_x4): Likewise.
9472 (vst1_p16_x4): Likewise.
9473 (vst1q_p16_x4): Likewise.
9474 (vst1_s64_x4): Likewise.
9475 (vst1_u64_x4): Likewise.
9476 (vst1_p64_x4): Likewise.
9477 (vst1q_s64_x4): Likewise.
9478 (vst1q_u64_x4): Likewise.
9479 (vst1q_p64_x4): Likewise.
9480 (vst1_f64_x4): Likewise.
9481 (vst1q_f64_x4): Likewise.
9483 2021-07-23 Jonathan Wrightt <jonathan.wright@arm.com>
9485 * config/aarch64/arm_neon.h (vst2_s64): Use __builtin_memcpy
9486 instead of constructing __builtin_aarch64_simd_oi one vector
9488 (vst2_u64): Likewise.
9489 (vst2_f64): Likewise.
9490 (vst2_s8): Likewise.
9491 (vst2_p8): Likewise.
9492 (vst2_s16): Likewise.
9493 (vst2_p16): Likewise.
9494 (vst2_s32): Likewise.
9495 (vst2_u8): Likewise.
9496 (vst2_u16): Likewise.
9497 (vst2_u32): Likewise.
9498 (vst2_f16): Likewise.
9499 (vst2_f32): Likewise.
9500 (vst2_p64): Likewise.
9501 (vst2q_s8): Likewise.
9502 (vst2q_p8): Likewise.
9503 (vst2q_s16): Likewise.
9504 (vst2q_p16): Likewise.
9505 (vst2q_s32): Likewise.
9506 (vst2q_s64): Likewise.
9507 (vst2q_u8): Likewise.
9508 (vst2q_u16): Likewise.
9509 (vst2q_u32): Likewise.
9510 (vst2q_u64): Likewise.
9511 (vst2q_f16): Likewise.
9512 (vst2q_f32): Likewise.
9513 (vst2q_f64): Likewise.
9514 (vst2q_p64): Likewise.
9516 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9518 * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy
9519 instead of constructing __builtin_aarch64_simd_ci one vector
9521 (vst3_u64): Likewise.
9522 (vst3_f64): Likewise.
9523 (vst3_s8): Likewise.
9524 (vst3_p8): Likewise.
9525 (vst3_s16): Likewise.
9526 (vst3_p16): Likewise.
9527 (vst3_s32): Likewise.
9528 (vst3_u8): Likewise.
9529 (vst3_u16): Likewise.
9530 (vst3_u32): Likewise.
9531 (vst3_f16): Likewise.
9532 (vst3_f32): Likewise.
9533 (vst3_p64): Likewise.
9534 (vst3q_s8): Likewise.
9535 (vst3q_p8): Likewise.
9536 (vst3q_s16): Likewise.
9537 (vst3q_p16): Likewise.
9538 (vst3q_s32): Likewise.
9539 (vst3q_s64): Likewise.
9540 (vst3q_u8): Likewise.
9541 (vst3q_u16): Likewise.
9542 (vst3q_u32): Likewise.
9543 (vst3q_u64): Likewise.
9544 (vst3q_f16): Likewise.
9545 (vst3q_f32): Likewise.
9546 (vst3q_f64): Likewise.
9547 (vst3q_p64): Likewise.
9549 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9551 * config/aarch64/arm_neon.h (vst4_s64): Use __builtin_memcpy
9552 instead of constructing __builtin_aarch64_simd_xi one vector
9554 (vst4_u64): Likewise.
9555 (vst4_f64): Likewise.
9556 (vst4_s8): Likewise.
9557 (vst4_p8): Likewise.
9558 (vst4_s16): Likewise.
9559 (vst4_p16): Likewise.
9560 (vst4_s32): Likewise.
9561 (vst4_u8): Likewise.
9562 (vst4_u16): Likewise.
9563 (vst4_u32): Likewise.
9564 (vst4_f16): Likewise.
9565 (vst4_f32): Likewise.
9566 (vst4_p64): Likewise.
9567 (vst4q_s8): Likewise.
9568 (vst4q_p8): Likewise.
9569 (vst4q_s16): Likewise.
9570 (vst4q_p16): Likewise.
9571 (vst4q_s32): Likewise.
9572 (vst4q_s64): Likewise.
9573 (vst4q_u8): Likewise.
9574 (vst4q_u16): Likewise.
9575 (vst4q_u32): Likewise.
9576 (vst4q_u64): Likewise.
9577 (vst4q_f16): Likewise.
9578 (vst4q_f32): Likewise.
9579 (vst4q_f64): Likewise.
9580 (vst4q_p64): Likewise.
9582 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9584 * config/aarch64/arm_neon.h (vtbx4_s8): Use __builtin_memcpy
9585 instead of constructing __builtin_aarch64_simd_oi one vector
9587 (vtbx4_u8): Likewise.
9588 (vtbx4_p8): Likewise.
9590 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9592 * config/aarch64/arm_neon.h (vtbl3_s8): Use __builtin_memcpy
9593 instead of constructing __builtin_aarch64_simd_oi one vector
9595 (vtbl3_u8): Likewise.
9596 (vtbl3_p8): Likewise.
9597 (vtbl4_s8): Likewise.
9598 (vtbl4_u8): Likewise.
9599 (vtbl4_p8): Likewise.
9601 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9603 * config/aarch64/arm_neon.h (vqtbx2_s8): Use __builtin_memcpy
9604 instead of constructing __builtin_aarch64_simd_oi one vector
9606 (vqtbx2_u8): Likewise.
9607 (vqtbx2_p8): Likewise.
9608 (vqtbx2q_s8): Likewise.
9609 (vqtbx2q_u8): Likewise.
9610 (vqtbx2q_p8): Likewise.
9611 (vqtbx3_s8): Use __builtin_memcpy instead of constructing
9612 __builtin_aarch64_simd_ci one vector at a time.
9613 (vqtbx3_u8): Likewise.
9614 (vqtbx3_p8): Likewise.
9615 (vqtbx3q_s8): Likewise.
9616 (vqtbx3q_u8): Likewise.
9617 (vqtbx3q_p8): Likewise.
9618 (vqtbx4_s8): Use __builtin_memcpy instead of constructing
9619 __builtin_aarch64_simd_xi one vector at a time.
9620 (vqtbx4_u8): Likewise.
9621 (vqtbx4_p8): Likewise.
9622 (vqtbx4q_s8): Likewise.
9623 (vqtbx4q_u8): Likewise.
9624 (vqtbx4q_p8): Likewise.
9626 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9628 * config/aarch64/arm_neon.h (vqtbl2_s8): Use __builtin_memcpy
9629 instead of constructing __builtin_aarch64_simd_oi one vector
9631 (vqtbl2_u8): Likewise.
9632 (vqtbl2_p8): Likewise.
9633 (vqtbl2q_s8): Likewise.
9634 (vqtbl2q_u8): Likewise.
9635 (vqtbl2q_p8): Likewise.
9636 (vqtbl3_s8): Use __builtin_memcpy instead of constructing
9637 __builtin_aarch64_simd_ci one vector at a time.
9638 (vqtbl3_u8): Likewise.
9639 (vqtbl3_p8): Likewise.
9640 (vqtbl3q_s8): Likewise.
9641 (vqtbl3q_u8): Likewise.
9642 (vqtbl3q_p8): Likewise.
9643 (vqtbl4_s8): Use __builtin_memcpy instead of constructing
9644 __builtin_aarch64_simd_xi one vector at a time.
9645 (vqtbl4_u8): Likewise.
9646 (vqtbl4_p8): Likewise.
9647 (vqtbl4q_s8): Likewise.
9648 (vqtbl4q_u8): Likewise.
9649 (vqtbl4q_p8): Likewise.
9651 2021-07-23 Haochen Gui <guihaoc@gcc.gnu.org>
9654 * config/rs6000/rs6000.md (cstore<mode>4): Fix wrong fall through.
9656 2021-07-22 Andrew Pinski <apinski@marvell.com>
9658 PR tree-optimization/10153
9659 * tree-tailcall.c (create_tailcall_accumulator):
9660 Don't call fold_convert as the type should be correct already.
9661 (tree_optimize_tail_calls_1): Use build_{one,zero}_cst instead
9662 of integer_{one,zero}_node for the call of create_tailcall_accumulator.
9664 2021-07-22 Aldy Hernandez <aldyh@redhat.com>
9666 * gimple-range-cache.cc (non_null_ref::adjust_range): Replace
9667 varying_p check for null/non-null check.
9669 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
9671 PR tree-optimization/101511
9672 * value-relation.cc (relation_oracle::query_relation): Check if ssa1
9673 is in ssa2's equiv set, and don't trap if so.
9675 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
9677 PR tree-optimization/101497
9678 * gimple-range-fold.cc (fold_using_range::range_of_cond_expr): Check
9681 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
9683 PR tree-optimization/101496
9684 * vr-values.c (simplify_using_ranges::fold_cond): Call range_of_stmt
9685 first, then vrp_visit_cond_Stmt.
9687 2021-07-22 liuhongt <hongtao.liu@intel.com>
9689 * config/i386/i386-expand.c
9690 (ix86_broadcast_from_integer_constant): Rename to ..
9691 (ix86_broadcast_from_constant): .. this, and extend it to
9693 (ix86_expand_vector_move): Extend to float mode.
9694 * config/i386/i386-features.c
9695 (replace_constant_pool_with_broadcast): Remove.
9696 (remove_partial_avx_dependency_gate): Ditto.
9697 (constant_pool_broadcast): Ditto.
9698 (class pass_constant_pool_broadcast): Ditto.
9699 (make_pass_constant_pool_broadcast): Ditto.
9700 (remove_partial_avx_dependency): Adjust gate.
9701 * config/i386/i386-passes.def: Remove pass_constant_pool_broadcast.
9702 * config/i386/i386-protos.h
9703 (make_pass_constant_pool_broadcast): Remove.
9705 2021-07-22 liuhongt <hongtao.liu@intel.com>
9707 * config/i386/constraints.md (Wb): New constraint.
9709 * config/i386/i386.md (*ashlhi3_1): Extend to avx512 mask
9711 (*ashlqi3_1): Ditto.
9712 (*<insn><mode>3_1): Split to ..
9713 (*ashr<mode>3_1): this, ...
9714 (*lshr<mode>3_1): and this, also extend this pattern to avx512
9716 (*<insn><mode>3_1): Split to ..
9717 (*ashr<mode>3_1): this, ...
9718 (*lshrqi3_1): and this, also extend this pattern to avx512
9720 (*lshrhi3_1): And this, also extend this pattern to avx512
9722 * config/i386/sse.md (k<code><mode>): New define_split after
9723 it to convert generic shift pattern to mask shift ones.
9725 2021-07-21 Thomas Schwinge <thomas@codesourcery.com>
9726 Joseph Myers <joseph@codesourcery.com>
9727 Cesar Philippidis <cesar@codesourcery.com>
9729 * tree-core.h (omp_clause_code): Add 'OMP_CLAUSE_NOHOST'.
9730 * tree.c (omp_clause_num_ops, omp_clause_code_name, walk_tree_1):
9732 * tree-pretty-print.c (dump_omp_clause): Likewise.
9733 * omp-general.c (oacc_verify_routine_clauses): Likewise.
9734 * gimplify.c (gimplify_scan_omp_clauses)
9735 (gimplify_adjust_omp_clauses): Likewise.
9736 * tree-nested.c (convert_nonlocal_omp_clauses)
9737 (convert_local_omp_clauses): Likewise.
9738 * omp-low.c (scan_sharing_clauses): Likewise.
9739 * omp-offload.c (execute_oacc_device_lower): Update.
9741 2021-07-21 Martin Sebor <msebor@redhat.com>
9743 * tree-ssa-alias.c (walk_aliased_vdefs_1): Fix typos in a comment.
9745 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9747 * config/rs6000/rs6000-gen-builtins.c (write_init_bif_table):
9750 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9752 * config/rs6000/rs6000-gen-builtins.c (write_fntype): New
9754 (write_fntype_init): New stub function.
9755 (write_init_bif_table): Likewise.
9756 (write_init_ovld_table): New function.
9757 (write_init_file): Implement.
9759 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9761 * config/rs6000/rs6000-gen-builtins.c
9762 (write_autogenerated_header): New function.
9763 (write_decls): Likewise.
9764 (write_extern_fntype): New callback function.
9765 (write_header_file): Implement.
9767 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9769 * config/rs6000/rs6000-gen-builtins.c (write_defines_file):
9772 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9774 * config/rs6000/rs6000-gen-builtins.c (complete_vector_type): New
9776 (complete_base_type): Likewise.
9777 (construct_fntype_id): Likewise.
9778 (parse_bif_entry): Call contruct_fntype_id.
9779 (parse_ovld_entry): Likewise.
9781 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9783 * config/rs6000/rs6000-gen-builtins.c (ovld_stanza): New struct.
9784 (MAXOVLDSTANZAS): New macro.
9785 (ovld_stanzas): New variable.
9786 (curr_ovld_stanza): Likewise.
9787 (MAXOVLDS): New macro.
9788 (ovlddata): New struct.
9789 (ovlds): New variable.
9790 (curr_ovld): Likewise.
9791 (max_ovld_args): Likewise.
9792 (parse_ovld_entry): New function.
9793 (parse_ovld_stanza): Likewise.
9794 (parse_ovld): Implement.
9796 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9798 * config/rs6000/rs6000-gen-builtins.c (parse_bif_attrs):
9801 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9803 * config/rs6000/rs6000-gen-builtins.c (parse_args): New function.
9804 (parse_prototype): Implement.
9806 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9808 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): New enum.
9809 (curr_bif_stanza): New variable.
9810 (stanza_entry): New struct.
9811 (stanza_map): New initialized variable.
9812 (enable_string): Likewise.
9813 (fnkinds): New enum.
9814 (typelist): New struct.
9815 (attrinfo): Likewise.
9816 (MAXRESTROPNDS): New macro.
9817 (prototype): New struct.
9818 (MAXBIFS): New macro.
9819 (bifdata): New struct.
9820 (bifs): New variable.
9821 (curr_bif): Likewise.
9822 (bif_order): Likewise.
9823 (bif_index): Likewise.
9824 (fatal): New function.
9825 (stanza_name_to_stanza): Likewise.
9826 (parse_bif_attrs): New stub function.
9827 (parse_prototype): Likewise.
9828 (parse_bif_entry): New function.
9829 (parse_bif_stanza): Likewise.
9830 (parse_bif): Implement.
9831 (set_bif_order): New function.
9832 (create_bif_order): Implement.
9834 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9836 * config/rs6000/rs6000-gen-builtins.c (rbtree.h): New #include.
9837 (num_bifs): New variable.
9838 (num_ovld_stanzas): Likewise.
9839 (num_ovlds): Likewise.
9840 (parse_codes): New enum.
9841 (bif_rbt): New variable.
9842 (ovld_rbt): Likewise.
9843 (fntype_rbt): Likewise.
9844 (bifo_rbt): Likewise.
9845 (parse_bif): New stub function.
9846 (create_bif_order): Likewise.
9847 (parse_ovld): Likewise.
9848 (write_header_file): Likewise.
9849 (write_init_file): Likewise.
9850 (write_defines_file): Likewise.
9851 (delete_output_files): New function.
9854 2021-07-21 H.J. Lu <hjl.tools@gmail.com>
9857 * config/i386/i386-builtin.def: Remove OPTION_MASK_ISA_SSE4_2
9858 from CRC32 _builtin functions.
9860 2021-07-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
9862 * coverage.c (build_gcov_info_var_registration): Mark the object placed
9863 in the linker set as referenced so that it does not get optimized away.
9865 2021-07-21 Kito Cheng <kito.cheng@sifive.com>
9868 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
9870 * config.gcc (riscv*-*-*): Detect which python is available.
9872 2021-07-21 Jakub Jelinek <jakub@redhat.com>
9874 PR middle-end/101535
9875 * gimplify.c (omp_check_private): Properly skip ORT_TARGET_DATA
9876 contexts in which decl isn't privatized and for ORT_TARGET return
9877 false if decl is mapped.
9879 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
9881 * gimple-loop-jam.c: Include tree-ssa-sccvn.h.
9882 (tree_loop_unroll_and_jam): Run value-numbering on a loop that
9883 has been successfully unrolled.
9885 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
9887 * tree-ssa-loop-manip.c (determine_exit_conditions): Return a null
9888 exit condition if no tail loop is needed, and if the original exit
9889 condition should therefore be kept as-is.
9890 (tree_transform_and_unroll_loop): Handle that case here too.
9892 2021-07-21 Kewen Lin <linkw@linux.ibm.com>
9894 * tree-data-ref.c (free_dependence_relations): Adjust to pass vec
9896 (free_data_refs): Likewise.
9897 * tree-data-ref.h (free_dependence_relations): Likewise.
9898 (free_data_refs): Likewise.
9899 * tree-predcom.c (struct chain): Use auto_vec instead of vec for
9901 (struct component): Likewise.
9902 (pcom_worker::pcom_worker): Adjust for auto_vec and renaming changes.
9903 (pcom_worker::~pcom_worker): Likewise.
9904 (pcom_worker::release_chain): Adjust as auto_vec changes.
9905 (pcom_worker::loop): Rename to ...
9906 (pcom_worker::m_loop): ... this.
9907 (pcom_worker::datarefs): Rename to ...
9908 (pcom_worker::m_datarefs): ... this. Use auto_vec instead of vec.
9909 (pcom_worker::dependences): Rename to ...
9910 (pcom_worker::m_dependences): ... this. Use auto_vec instead of vec.
9911 (pcom_worker::chains): Rename to ...
9912 (pcom_worker::m_chains): ... this. Use auto_vec instead of vec.
9913 (pcom_worker::looparound_phis): Rename to ...
9914 (pcom_worker::m_looparound_phis): ... this. Use auto_vec instead of
9916 (pcom_worker::cache): Rename to ...
9917 (pcom_worker::m_cache): ... this. Use auto_vec instead of vec.
9918 (pcom_worker::release_chain): Adjust for auto_vec changes.
9919 (pcom_worker::release_chains): Adjust for auto_vec and renaming
9921 (release_component): Remove.
9922 (release_components): Adjust for release_component removal.
9923 (component_of): Adjust to use vec.
9924 (merge_comps): Likewise.
9925 (pcom_worker::aff_combination_dr_offset): Adjust for renaming changes.
9926 (pcom_worker::determine_offset): Likewise.
9927 (class comp_ptrs): Remove.
9928 (pcom_worker::split_data_refs_to_components): Adjust for renaming
9929 changes, for comp_ptrs removal with auto_vec.
9930 (pcom_worker::suitable_component_p): Adjust for renaming changes.
9931 (pcom_worker::filter_suitable_components): Adjust for release_component
9933 (pcom_worker::valid_initializer_p): Adjust for renaming changes.
9934 (pcom_worker::find_looparound_phi): Likewise.
9935 (pcom_worker::add_looparound_copies): Likewise.
9936 (pcom_worker::determine_roots_comp): Likewise.
9937 (pcom_worker::single_nonlooparound_use): Likewise.
9938 (pcom_worker::execute_pred_commoning_chain): Likewise.
9939 (pcom_worker::execute_pred_commoning): Likewise.
9940 (pcom_worker::try_combine_chains): Likewise.
9941 (pcom_worker::prepare_initializers_chain): Likewise.
9942 (pcom_worker::prepare_initializers): Likewise.
9943 (pcom_worker::prepare_finalizers_chain): Likewise.
9944 (pcom_worker::prepare_finalizers): Likewise.
9945 (pcom_worker::tree_predictive_commoning_loop): Likewise.
9947 2021-07-20 Martin Sebor <msebor@redhat.com>
9949 PR middle-end/101397
9950 * builtins.c (gimple_call_return_array): Add argument. Correct
9951 offsets for memchr, mempcpy, stpcpy, and stpncpy.
9952 (compute_objsize_r): Adjust offset computation for argument returning
9955 2021-07-20 Martin Sebor <msebor@redhat.com>
9957 PR middle-end/101300
9958 * tree-ssa-uninit.c (check_defs): Handle UBSAN built-ins.
9960 2021-07-20 Jeff Law <jlaw@localhost.localdomain>
9962 * function.c (assign_parm_setup_block): Use adjust_address instead
9963 of change_address to preserve MEM_EXPR and friends.
9965 2021-07-20 Martin Sebor <msebor@redhat.com>
9967 * cfgloop.h (single_likely_exit): Adjust by-value argument to
9969 * cfgloopanal.c (single_likely_exit): Same.
9970 * cgraph.h (struct cgraph_node): Same.
9971 * cgraphclones.c (cgraph_node::create_virtual_clone): Same.
9972 * genautomata.c (merge_states): Same.
9973 * genextract.c (VEC_char_to_string): Same.
9974 * genmatch.c (dt_node::gen_kids_1): Same.
9975 (walk_captures): Adjust by-value argument to by-reference.
9976 * gimple-ssa-store-merging.c (check_no_overlap): Adjust by-value argument
9977 to by-const-reference.
9978 * gimple.c (gimple_build_call_vec): Same.
9979 (gimple_build_call_internal_vec): Same.
9980 (gimple_build_switch): Same.
9981 (sort_case_labels): Same.
9982 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
9984 * gimple.h (gimple_build_call_vec): Adjust by-value argument to
9986 (gimple_build_call_internal_vec): Same.
9987 (gimple_build_switch): Same.
9988 (sort_case_labels): Same.
9989 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
9991 * haifa-sched.c (calc_priorities): Adjust by-value argument to
9993 (sched_init_luids): Same.
9994 (haifa_init_h_i_d): Same.
9995 * ipa-cp.c (ipa_get_indirect_edge_target_1): Same.
9996 (adjust_callers_for_value_intersection): Adjust by-value argument to
9998 (find_more_scalar_values_for_callers_subset): Adjust by-value argument to
10000 (find_more_contexts_for_caller_subset): Same.
10001 (find_aggregate_values_for_callers_subset): Same.
10002 (copy_useful_known_contexts): Same.
10003 * ipa-fnsummary.c (remap_edge_summaries): Same.
10004 (remap_freqcounting_predicate): Same.
10005 * ipa-inline.c (add_new_edges_to_heap): Adjust by-value argument to
10007 * ipa-predicate.c (predicate::remap_after_inlining): Adjust by-value argument
10008 to by-const-reference.
10009 * ipa-predicate.h (predicate::remap_after_inlining): Same.
10010 * ipa-prop.c (ipa_find_agg_cst_for_param): Same.
10011 * ipa-prop.h (ipa_find_agg_cst_for_param): Same.
10012 * ira-build.c (ira_loop_tree_body_rev_postorder): Same.
10013 * read-rtl.c (add_overload_instance): Same.
10014 * rtl.h (native_decode_rtx): Same.
10015 (native_decode_vector_rtx): Same.
10016 * sched-int.h (sched_init_luids): Same.
10017 (haifa_init_h_i_d): Same.
10018 * simplify-rtx.c (native_decode_vector_rtx): Same.
10019 (native_decode_rtx): Same.
10020 * tree-call-cdce.c (gen_shrink_wrap_conditions): Same.
10021 (shrink_wrap_one_built_in_call_with_conds): Same.
10022 (shrink_wrap_conditional_dead_built_in_calls): Same.
10023 * tree-data-ref.c (create_runtime_alias_checks): Same.
10024 (compute_all_dependences): Same.
10025 * tree-data-ref.h (compute_all_dependences): Same.
10026 (create_runtime_alias_checks): Same.
10027 (index_in_loop_nest): Same.
10028 * tree-if-conv.c (mask_exists): Same.
10029 * tree-loop-distribution.c (class loop_distribution): Same.
10030 (loop_distribution::create_rdg_vertices): Same.
10031 (dump_rdg_partitions): Same.
10032 (debug_rdg_partitions): Same.
10033 (partition_contains_all_rw): Same.
10034 (loop_distribution::distribute_loop): Same.
10035 * tree-parloops.c (oacc_entry_exit_ok_1): Same.
10036 (oacc_entry_exit_single_gang): Same.
10037 * tree-ssa-loop-im.c (hoist_memory_references): Same.
10038 (loop_suitable_for_sm): Same.
10039 * tree-ssa-loop-niter.c (bound_index): Same.
10040 * tree-ssa-reassoc.c (update_ops): Same.
10041 (swap_ops_for_binary_stmt): Same.
10042 (rewrite_expr_tree): Same.
10043 (rewrite_expr_tree_parallel): Same.
10044 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Same.
10045 * tree-ssa-sccvn.h (ao_ref_init_from_vn_reference): Same.
10046 * tree-ssa-structalias.c (process_all_all_constraints): Same.
10047 (make_constraints_to): Same.
10048 (handle_lhs_call): Same.
10049 (find_func_aliases_for_builtin_call): Same.
10050 (sort_fieldstack): Same.
10051 (check_for_overlaps): Same.
10052 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Same.
10053 (vect_create_cond_for_unequal_addrs): Same.
10054 (vect_create_cond_for_lower_bounds): Same.
10055 (vect_create_cond_for_alias_checks): Same.
10056 * tree-vect-slp-patterns.c (vect_validate_multiplication): Same.
10057 * tree-vect-slp.c (vect_analyze_slp_instance): Same.
10058 (vect_make_slp_decision): Same.
10059 (vect_slp_bbs): Same.
10060 (duplicate_and_interleave): Same.
10061 (vect_transform_slp_perm_load): Same.
10062 (vect_schedule_slp): Same.
10063 * tree-vectorizer.h (vect_transform_slp_perm_load): Same.
10064 (vect_schedule_slp): Same.
10065 (duplicate_and_interleave): Same.
10066 * tree.c (build_vector_from_ctor): Same.
10067 (build_vector): Same.
10068 (check_vector_cst): Same.
10069 (check_vector_cst_duplicate): Same.
10070 (check_vector_cst_fill): Same.
10071 (check_vector_cst_stepped): Same.
10072 * tree.h (build_vector_from_ctor): Same.
10074 2021-07-20 Jakub Jelinek <jakub@redhat.com>
10077 * config/rs6000/rs6000-protos.h (easy_altivec_constant): Change return
10078 type from bool to int.
10079 * config/rs6000/rs6000.c (vspltis_constant): Fix up handling the
10080 EASY_VECTOR_MSB case if either step or copies is not 1.
10081 (vspltis_shifted): Fix comment typo.
10082 (easy_altivec_constant): Change return type from bool to int, instead
10083 of returning true return byte size of the element mode that should be
10084 used to synthetize the constant.
10085 * config/rs6000/predicates.md (easy_vector_constant_msb): Require
10086 that vspltis_shifted is 0, handle the case where easy_altivec_constant
10087 assumes using different vector mode from CONST_VECTOR's mode.
10088 * config/rs6000/altivec.md (easy_vector_constant_msb splitter): Use
10089 easy_altivec_constant to determine mode in which -1 >> -1 should be
10090 performed, use rs6000_expand_vector_init instead of gen_vec_initv4sisi.
10092 2021-07-20 Richard Biener <rguenther@suse.de>
10095 * dwarf2out.h (dwarf_file_data): Add key member.
10096 * dwarf2out.c (dwarf_file_hasher::equal): Compare key.
10097 (dwarf_file_hasher::hash): Hash key.
10098 (lookup_filename): Remap the filename and store it in the
10099 filename member of dwarf_file_data when creating a new
10101 (file_name_acquire): Do not remap the filename again.
10102 (maybe_emit_file): Likewise.
10104 2021-07-20 Jonathan Wright <jonathan.wright@arm.com>
10106 * config/aarch64/aarch64-simd-builtins.def: Use two variant
10107 generators for all TBL/TBX intrinsics and rename to
10108 consistent forms: qtbl[1234] or qtbx[1234].
10109 * config/aarch64/aarch64-simd.md (aarch64_tbl1<mode>):
10111 (aarch64_qtbl1<mode>): This.
10112 (aarch64_tbx1<mode>): Rename to...
10113 (aarch64_qtbx1<mode>): This.
10114 (aarch64_tbl2v16qi): Delete.
10115 (aarch64_tbl3<mode>): Rename to...
10116 (aarch64_qtbl2<mode>): This.
10117 (aarch64_tbx4<mode>): Rename to...
10118 (aarch64_qtbx2<mode>): This.
10119 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_1): Use
10120 renamed qtbl1 and qtbl2 RTL patterns.
10121 * config/aarch64/arm_neon.h (vqtbl1_p8): Use renamed qtbl1
10123 (vqtbl1_s8): Likewise.
10124 (vqtbl1_u8): Likewise.
10125 (vqtbl1q_p8): Likewise.
10126 (vqtbl1q_s8): Likewise.
10127 (vqtbl1q_u8): Likewise.
10128 (vqtbx1_s8): Use renamed qtbx1 RTL pattern.
10129 (vqtbx1_u8): Likewise.
10130 (vqtbx1_p8): Likewise.
10131 (vqtbx1q_s8): Likewise.
10132 (vqtbx1q_u8): Likewise.
10133 (vqtbx1q_p8): Likewise.
10134 (vtbl1_s8): Use renamed qtbl1 RTL pattern.
10135 (vtbl1_u8): Likewise.
10136 (vtbl1_p8): Likewise.
10137 (vtbl2_s8): Likewise
10138 (vtbl2_u8): Likewise.
10139 (vtbl2_p8): Likewise.
10140 (vtbl3_s8): Use renamed qtbl2 RTL pattern.
10141 (vtbl3_u8): Likewise.
10142 (vtbl3_p8): Likewise.
10143 (vtbl4_s8): Likewise.
10144 (vtbl4_u8): Likewise.
10145 (vtbl4_p8): Likewise.
10146 (vtbx2_s8): Use renamed qtbx2 RTL pattern.
10147 (vtbx2_u8): Likewise.
10148 (vtbx2_p8): Likewise.
10149 (vqtbl2_s8): Use renamed qtbl2 RTL pattern.
10150 (vqtbl2_u8): Likewise.
10151 (vqtbl2_p8): Likewise.
10152 (vqtbl2q_s8): Likewise.
10153 (vqtbl2q_u8): Likewise.
10154 (vqtbl2q_p8): Likewise.
10155 (vqtbx2_s8): Use renamed qtbx2 RTL pattern.
10156 (vqtbx2_u8): Likewise.
10157 (vqtbx2_p8): Likewise.
10158 (vqtbx2q_s8): Likewise.
10159 (vqtbx2q_u8): Likewise.
10160 (vqtbx2q_p8): Likewise.
10161 (vtbx4_s8): Likewise.
10162 (vtbx4_u8): Likewise.
10163 (vtbx4_p8): Likewise.
10165 2021-07-20 Uroš Bizjak <ubizjak@gmail.com>
10168 * config/i386/sync.md (define_peephole2 atomic_storedi_fpu):
10170 (define_peephole2 atomic_loaddi_fpu): Ditto.
10172 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
10174 * config.gcc (riscv*-*-*): Detect which python is available.
10176 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
10178 * config/rs6000/vsx.md (mulhs_<mode>): Rename to...
10179 (smul<mode>3_highpart): ... this.
10180 (mulhu_<mode>): Rename to...
10181 (umul<mode>3_highpart): ... this.
10182 * config/rs6000/rs6000-builtin.def (MULHS_V2DI, MULHS_V4SI,
10183 MULHU_V2DI, MULHU_V4SI): Adjust.
10185 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
10187 PR tree-optimization/100696
10188 * internal-fn.c (first_commutative_argument): Add info for IFN_MULH.
10189 * internal-fn.def (IFN_MULH): New internal function.
10190 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Add support to
10191 recog normal multiply highpart as IFN_MULH.
10192 * config/i386/i386.c (ix86_add_stmt_cost): Adjust for combined
10195 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
10197 * config/elfos.h (CTF_DEBUGGING_INFO): New definition.
10198 (BTF_DEBUGGING_INFO): Likewise.
10199 * doc/tm.texi.in: Document the new macros.
10200 * doc/tm.texi: Regenerated.
10201 * toplev.c: Guard initialization of debug hooks.
10203 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
10205 * flags.h (ctf_debuginfo_p): New function declaration.
10206 * opts.c (ctf_debuginfo_p): New function definition.
10208 2021-07-19 Andrew Stubbs <ams@codesourcery.com>
10211 * config/gcn/gcn-hsa.h (DRIVER_SELF_SPECS): New.
10212 (ASM_SPEC): Set -mattr for xnack and sram-ecc.
10213 * config/gcn/gcn-opts.h (enum sram_ecc_type): New.
10214 * config/gcn/gcn-valu.md: Add a warning comment.
10215 * config/gcn/gcn.c (gcn_option_override): Add "sorry" for -mxnack.
10216 (output_file_start): Add xnack and sram-ecc state to ".amdgcn_target".
10217 * config/gcn/gcn.md: Add a warning comment.
10218 * config/gcn/gcn.opt: Add -mxnack and -msram-ecc.
10219 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): Remove
10221 (EF_AMDGPU_XNACK): New.
10222 (EF_AMDGPU_SRAM_ECC): New.
10224 (copy_early_debug_info): Use elf_flags.
10225 (main): Handle -mxnack and -msram-ecc options.
10226 * doc/invoke.texi: Document -mxnack and -msram-ecc.
10228 2021-07-19 Andrew Pinski <apinski@marvell.com>
10231 * config/aarch64/aarch64.md (csneg3_uxtw_insn): Rename to ...
10232 (*cs<neg_not_cs>3_uxtw_insn4): and extend to NEG_NOT.
10234 2021-07-19 Richard Biener <rguenther@suse.de>
10236 PR tree-optimization/101505
10237 * tree-vect-patterns.c (vect_determine_precisions): Walk
10238 PHIs also for loop vectorization.
10240 2021-07-19 Richard Biener <rguenther@suse.de>
10242 * gimple.h (gimple_expr_type): Remove.
10243 * doc/gimple.texi: Remove gimple_expr_type documentation.
10245 2021-07-19 Richard Biener <rguenther@suse.de>
10247 * tree-ssa-sccvn.c (vn_reference_eq): Handle NULL vr->type.
10248 (ao_ref_init_from_vn_reference): Likewise.
10249 (fully_constant_reference): Likewise.
10250 (vn_reference_lookup_call): Do not set vr->type to random
10252 * tree-ssa-pre.c (compute_avail): Do not try to PRE calls
10254 * tree-vect-generic.c (expand_vector_piecewise): Pass in
10255 whether we expanded parallel.
10256 (expand_vector_parallel): Adjust.
10257 (expand_vector_addition): Likewise.
10258 (expand_vector_comparison): Likewise.
10259 (expand_vector_operation): Likewise.
10260 (expand_vector_scalar_condition): Likewise.
10261 (expand_vector_conversion): Likewise.
10263 2021-07-19 Richard Biener <rguenther@suse.de>
10265 * tree-vrp.c (register_edge_assert_for_2): Use the
10267 (vrp_folder::fold_predicate_in): Likewise.
10268 * vr-values.c (gimple_assign_nonzero_p): Likewise.
10269 (vr_values::extract_range_from_comparison): Likewise.
10270 (vr_values::extract_range_from_ubsan_builtin): Use the
10271 type of the first operand.
10272 (vr_values::extract_range_basic): Push down type
10273 computation, use the appropriate LHS.
10274 (vr_values::extract_range_from_assignment): Use the
10277 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
10280 * common/config/i386/i386-common.c (ix86_handle_option): For
10281 -mgeneral-regs-only, enable the GPR only instructions which are
10282 enabled implicitly by SSE ISAs unless they have been disabled
10285 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
10288 * config/i386/i386.c (ix86_check_avx_upper_stores): Moved before
10289 ix86_avx_u128_mode_needed.
10290 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY if callee
10291 returns AVX register.
10293 2021-07-17 Jan Hubicka <hubicka@ucw.cz>
10295 * tree-ssa-structalias.c (handle_rhs_call): Support EAF_NOT_RETURNED.
10296 (handle_const_call): Liekise
10297 (handle_pure_call): Liekise
10299 2021-07-17 Andrew MacLeod <amacleod@redhat.com>
10301 PR tree-optimization/96542
10302 * range-op.cc (range_operator::wi_fold_in_parts): New.
10303 (range_operator::fold_range): Call wi_fold_in_parts.
10304 (operator_lshift::wi_fold): Fix broken lshift by [0,0].
10305 * range-op.h (wi_fold_in_parts): Add prototype.
10307 2021-07-16 David Malcolm <dmalcolm@redhat.com>
10309 * doc/analyzer.texi: Add __analyzer_dump_state.
10311 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10313 * config/rs6000/rbtree.c: New file.
10314 * config/rs6000/rbtree.h: New file.
10316 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10318 * config/rs6000/rs6000-gen-builtins.c (restriction): New enum.
10319 (typeinfo): Add restr field.
10320 (match_bracketed_pair): New function.
10321 (match_const_restriction): Implement.
10323 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10325 * config/rs6000/rs6000-gen-builtins.c (match_basetype): Implement.
10327 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10329 * config/rs6000/rs6000-gen-builtins.c (void_status): New enum.
10330 (basetype): Likewise.
10331 (typeinfo): Likewise.
10332 (handle_pointer): New function.
10333 (match_basetype): New stub function.
10334 (match_const_restriction): Likewise.
10335 (match_type): New function.
10337 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10339 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace): New
10341 (advance_line): Likewise.
10342 (safe_inc_pos): Likewise.
10343 (match_identifier): Likewise.
10344 (match_integer): Likewise.
10345 (match_to_right_bracket): Likewise.
10347 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10349 * config/rs6000/rs6000-gen-builtins.c (bif_file): New variable.
10350 (ovld_file): Likewise.
10351 (header_file): Likewise.
10352 (init_file): Likewise.
10353 (defines_file): Likewise.
10354 (pgm_path): Likewise.
10355 (bif_path): Likewise.
10356 (ovld_path): Likewise.
10357 (header_path): Likewise.
10358 (init_path): Likewise.
10359 (defines_path): Likewise.
10360 (LINELEN): New macro.
10361 (linebuf): New variable.
10365 (bif_diag): New function.
10366 (ovld_diag): Likewise.
10368 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10370 * config/rs6000/rs6000-builtin-new.def: New.
10371 * config/rs6000/rs6000-overload.def: New.
10373 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10375 * config/rs6000/rs6000-gen-builtins.c: New.
10377 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
10379 * Makefile.in (EXTRA_GTYPE_DEPS): New variable.
10380 (s-gtype): Depend on EXTRA_GTYPE_DEPS.
10381 * gengtype-state.c (state_writer::write_state_file_list): Add a
10382 parameter to the fileslist expression for the number of build
10384 (read_state_files_list): Detect build headers and strip the
10385 initial "./" or ".\" from their names.
10386 * gengtype.c (build_headers): New global variable.
10387 (num_build_headers): Likewise.
10388 (open_base_files): Emit #include for each build header.
10389 (main): Detect and count build headers.
10390 * gengtype.h (build_headers): New extern variable.
10391 (num_build_headers): Likewise.
10393 2021-07-16 Richard Biener <rguenther@suse.de>
10395 * gimple-ssa-store-merging.c (verify_symbolic_number_p): Use
10396 the type of the LHS.
10397 (find_bswap_or_nop_1): Likewise.
10398 (find_bswap_or_nop): Likewise.
10399 * tree-vectorizer.h (vect_get_smallest_scalar_type): Adjust
10401 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
10402 Remove unused parameters, pass in the scalar type. Fix
10403 internal store function handling.
10404 * tree-vect-stmts.c (vect_analyze_stmt): Remove assert.
10405 (vect_get_vector_types_for_stmt): Move down check for
10406 existing vector stmt after we've determined a scalar type.
10407 Pass down the used scalar type to vect_get_smallest_scalar_type.
10408 * tree-vect-generic.c (expand_vector_condition): Use
10409 the type of the LHS.
10410 (expand_vector_scalar_condition): Likewise.
10411 (expand_vector_operations_1): Likewise.
10412 * tree-vect-patterns.c (vect_widened_op_tree): Likewise.
10413 (vect_recog_dot_prod_pattern): Likewise.
10414 (vect_recog_sad_pattern): Likewise.
10415 (vect_recog_widen_op_pattern): Likewise.
10416 (vect_recog_widen_sum_pattern): Likewise.
10417 (vect_recog_mixed_size_cond_pattern): Likewise.
10419 2021-07-16 Jan Hubicka <hubicka@ucw.cz>
10421 * ipa-modref.c (struct escape_entry): Use eaf_fleags_t.
10422 (dump_eaf_flags): Dump EAF_NOT_RETURNED
10423 (eaf_flags_useful_p): Use eaf_fleags_t; handle const functions
10424 and EAF_NOT_RETURNED.
10425 (modref_summary::useful_p): Likewise.
10426 (modref_summary_lto::useful_p): Likewise.
10427 (struct) modref_summary_lto: Use eaf_fleags_t.
10428 (deref_flags): Handle EAF_NOT_RETURNED.
10429 (struct escape_point): Use min_flags.
10430 (modref_lattice::init): Add EAF_NOT_RETURNED.
10431 (merge_call_lhs_flags): Ignore EAF_NOT_RETURNED functions
10432 (analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
10434 (analyze_parms): Also analyze const functions; update conition on
10436 (modref_write): Update streaming.
10437 (read_section): Update streaming.
10438 (remap_arg_flags): Use eaf_flags_t.
10439 (modref_merge_call_site_flags): Hanlde EAF_NOT_RETURNED.
10440 * ipa-modref.h: (eaf_flags_t): New typedef.
10441 (struct modref_summary): Use eaf_flags_t.
10442 * tree-core.h (EAF_NOT_RETURNED): New constant.
10444 2021-07-16 Richard Biener <rguenther@suse.de>
10446 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Use
10447 the type of the LHS.
10448 (gimple_assign_nonnegative_warnv_p): Likewise.
10449 (gimple_call_nonnegative_warnv_p): Likewise. Return false
10450 if the call has no LHS.
10451 * gimple.c (gimple_could_trap_p_1): Use the type of the LHS.
10452 * tree-eh.c (stmt_could_throw_1_p): Likewise.
10453 * tree-inline.c (insert_init_stmt): Likewise.
10454 * tree-ssa-loop-niter.c (get_val_for): Likewise.
10455 * tree-outof-ssa.c (ssa_is_replaceable_p): Use the type of
10457 * tree-ssa-sccvn.c (init_vn_nary_op_from_stmt): Take a
10458 gassign *. Use the type of the lhs.
10459 (vn_nary_op_lookup_stmt): Adjust.
10460 (vn_nary_op_insert_stmt): Likewise.
10462 2021-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
10464 * config/s390/predicates.md (bras_sym_operand): Accept all
10465 functions in 64-bit mode, use UNSPEC_PLT31.
10466 (larl_operand): Use UNSPEC_PLT31.
10467 * config/s390/s390.c (s390_loadrelative_operand_p): Likewise.
10468 (legitimize_pic_address): Likewise.
10469 (s390_emit_tls_call_insn): Mark __tls_get_offset as function,
10471 (s390_delegitimize_address): Use UNSPEC_PLT31.
10472 (s390_output_addr_const_extra): Likewise.
10473 (print_operand): Add @PLT to TLS calls, handle %K.
10474 (s390_function_profiler): Mark __fentry__/_mcount as function,
10475 use %K, use UNSPEC_PLT31.
10476 (s390_output_mi_thunk): Use only UNSPEC_GOT, use %K.
10477 (s390_emit_call): Use UNSPEC_PLT31.
10478 (s390_emit_tpf_eh_return): Mark __tpf_eh_return as function.
10479 * config/s390/s390.md (UNSPEC_PLT31): Rename from UNSPEC_PLT.
10480 (*movdi_64): Use %K.
10481 (reload_base_64): Likewise.
10482 (*sibcall_brc): Likewise.
10483 (*sibcall_brcl): Likewise.
10484 (*sibcall_value_brc): Likewise.
10485 (*sibcall_value_brcl): Likewise.
10487 (*brasl): Likewise.
10488 (*bras_r): Likewise.
10489 (*brasl_r): Likewise.
10490 (*bras_tls): Likewise.
10491 (*brasl_tls): Likewise.
10492 (main_base_64): Likewise.
10493 (reload_base_64): Likewise.
10494 (@split_stack_call<mode>): Likewise.
10496 2021-07-16 Richard Biener <rguenther@suse.de>
10498 PR tree-optimization/101467
10499 * tree-vect-stmts.c (vect_gen_while): Properly guard
10500 make_temp_ssa_name usage.
10502 2021-07-16 Cooper Qu <cooper.qu@linux.alibaba.com>
10504 * config.gcc: Don't use forked print-sysroot-suffix.sh and
10505 t-sysroot-suffix for C-SKY.
10506 * config/csky/print-sysroot-suffix.sh: Delete.
10507 * config/csky/t-csky-linux: Delete.
10508 * config/csky/t-sysroot-suffix: Define MULTILIB_DIRNAMES
10509 instead of CSKY_MULTILIB_DIRNAMES.
10511 2021-07-16 Richard Biener <rguenther@suse.de>
10513 * tree-vect-loop.c (vect_transform_cycle_phi): Correct sign
10514 conversion issues with the partial reduction of the reused
10515 vector accumulator.
10517 2021-07-16 Richard Biener <rguenther@suse.de>
10519 * config/i386/i386-options.c (ix86_option_override_internal): Set
10520 param_vect_partial_vector_usage to zero if not set.
10522 2021-07-15 Uroš Bizjak <ubizjak@gmail.com>
10525 * config/i386/i386.h (VALID_SSE_REG_MODE): Add TDmode.
10526 (VALID_INT_MODE_P): Add SDmode and DDmode.
10527 Add TDmode for TARGET_64BIT.
10528 (VALID_DFP_MODE_P): Remove.
10529 * config/i386/i386.c (ix86_hard_regno_mode_ok):
10530 Do not use VALID_DFP_MODE_P.
10532 2021-07-15 Andrew MacLeod <amacleod@redhat.com>
10534 * gimple-range-fold.cc (adjust_pointer_diff_expr): Use
10536 (fold_using_range::fold_stmt): Ditto.
10537 (fold_using_range::range_of_range_op): Ditto.
10538 (fold_using_range::range_of_phi): Ditto.
10539 (fold_using_range::range_of_call): Ditto.
10540 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
10541 (fold_using_range::range_of_builtin_call): Ditto.
10542 (fold_using_range::range_of_cond_expr): Ditto.
10543 * gimple-range-fold.h (gimple_range_type): New.
10545 2021-07-15 Martin Sebor <msebor@redhat.com>
10547 PR middle-end/97027
10548 * tree-ssa-strlen.c (handle_assign): New function.
10549 (maybe_warn_overflow): Add argument.
10550 (nonzero_bytes_for_type): New function.
10551 (count_nonzero_bytes): Handle more tree types. Call
10552 nonzero_bytes_for_tye.
10553 (count_nonzero_bytes): Handle types.
10554 (handle_store): Handle stores from function calls.
10555 (strlen_check_and_optimize_call): Move code to handle_assign. Call
10556 it for assignments from function calls.
10558 2021-07-15 David Malcolm <dmalcolm@redhat.com>
10563 * doc/invoke.texi: Add -Wanalyzer-use-of-uninitialized-value.
10565 2021-07-15 David Malcolm <dmalcolm@redhat.com>
10567 * doc/invoke.texi (-fdump-analyzer-exploded-paths): New.
10569 2021-07-15 Martin Sebor <msebor@redhat.com>
10573 * fold-const.c (operand_compare::operand_equal_p): Handle OEP_DECL_NAME.
10574 (operand_compare::verify_hash_value): Same.
10575 * tree-core.h (OEP_DECL_NAME): New.
10577 2021-07-15 Martin Jambor <mjambor@suse.cz>
10579 * profile-count.h (profile_count::value): Change the return type to
10581 * gimple-pretty-print.c (dump_gimple_bb_header): Adjust print
10583 * tree-cfg.c (dump_function_to_file): Likewise.
10585 2021-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
10588 * config/rs6000/rs6000-p8swap.c (has_part_mult): New.
10589 (rs6000_analyze_swaps): Insns containing a subreg of a mult are
10592 2021-07-15 Richard Biener <rguenther@suse.de>
10594 * tree-vectorizer.h (vect_gen_while): Match up with
10595 vect_gen_while_not.
10596 * tree-vect-stmts.c (vect_gen_while): Adjust API to that
10597 of vect_gen_while_not.
10598 (vect_gen_while_not): Adjust.
10599 * tree-vect-loop-manip.c (vect_set_loop_controls_directly): Likewise.
10601 2021-07-15 Aldy Hernandez <aldyh@redhat.com>
10603 * gimple-range-cache.cc (non_null_ref::adjust_range): New.
10604 (ranger_cache::range_of_def): Call adjust_range.
10605 (ranger_cache::entry_range): Same.
10606 * gimple-range-cache.h (non_null_ref::adjust_range): New.
10607 * gimple-range.cc (gimple_ranger::range_of_expr): Call
10609 (gimple_ranger::range_on_entry): Same.
10611 2021-07-15 Tamar Christina <tamar.christina@arm.com>
10614 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10616 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
10618 2021-07-15 Tamar Christina <tamar.christina@arm.com>
10621 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10623 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
10624 (sdot_prod, udot_prod): ...These.
10625 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
10626 (aarch64_<sur>dot<vsi2qi>): Rename to...
10627 (<sur>dot_prod<vsi2qi>): ...This.
10628 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
10631 2021-07-15 Jakub Jelinek <jakub@redhat.com>
10633 PR middle-end/101437
10634 * gimplify.c (gimplify_expr): Throw away volatile reads from empty
10635 types even if they have non-BLKmode TYPE_MODE.
10637 2021-07-15 Richard Biener <rguenther@suse.de>
10640 * gcc.c (process_command): Process -gtoggle like process_options
10641 would after parsing options.
10643 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
10645 * cfgexpand.c (expand_asm_loc): Adjust.
10646 (expand_asm_stmt): Likewise.
10647 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Likewise.
10648 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
10649 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
10650 * config/avr/avr.c (avr_md_asm_adjust): Likewise.
10651 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
10652 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
10653 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
10654 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
10655 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
10656 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
10657 * config/s390/s390.c (s390_md_asm_adjust): Likewise.
10658 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
10659 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
10660 * doc/tm.texi: Regenerate.
10661 * target.def: Add location argument to md_asm_adjust.
10663 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
10665 * tree-diagnostic.c (diagnostic_report_current_function): Use the
10666 diagnostic's location, not input_location.
10668 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
10670 * cfgexpand.c (tree_conflicts_with_clobbers_p): Pass location to
10672 (expand_asm_stmt): Likewise.
10674 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
10676 * config/rs6000/rs6000.c (adjacent_mem_locations): Return the lower
10677 addressed memory rtx, if any.
10678 (rs6000_split_multireg_move): Fix code formatting.
10679 Handle MMA build built-ins with operands in adjacent memory locations.
10681 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
10683 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Move to later
10686 2021-07-14 Jason Merrill <jason@redhat.com>
10688 * sel-sched-ir.h (get_all_loop_exits): Use auto_vec.
10690 2021-07-14 Jason Merrill <jason@redhat.com>
10692 * doc/invoke.texi: -fdelete-dead-exceptions is on by default for
10695 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10697 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
10698 Remove erroneous line.
10700 2021-07-14 Andrew MacLeod <amacleod@redhat.com>
10702 * params.opt (param_evrp_mode): Change default.
10704 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10706 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
10707 (sdot_prod, udot_prod): ...These.
10708 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
10709 (aarch64_<sur>dot<vsi2qi>): Rename to...
10710 (<sur>dot_prod<vsi2qi>): ...This.
10711 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
10714 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10716 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
10718 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10720 * doc/sourcebuild.texi (arm_v8_2a_i8mm_neon_hw): Document.
10722 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10724 * config/arm/neon.md (usdot_prod<vsi2qi>): New.
10726 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10728 * config/aarch64/aarch64-simd.md (aarch64_usdot<vsi2qi>): Rename to...
10729 (usdot_prod<vsi2qi>): ... This.
10730 * config/aarch64/aarch64-simd-builtins.def (usdot): Rename to...
10731 (usdot_prod): ...This.
10732 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Likewise.
10733 * config/aarch64/aarch64-sve.md (@aarch64_<sur>dot_prod<vsi2qi>):
10735 (@<sur>dot_prod<vsi2qi>): ...This.
10736 * config/aarch64/aarch64-sve-builtins-base.cc
10737 (svusdot_impl::expand): Use it.
10739 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10741 * optabs.def (usdot_prod_optab): New.
10742 * doc/md.texi: Document it and clarify other dot prod optabs.
10743 * optabs-tree.h (enum optab_subtype): Add optab_vector_mixed_sign.
10744 * optabs-tree.c (optab_for_tree_code): Support usdot_prod_optab.
10745 * optabs.c (expand_widen_pattern_expr): Likewise.
10746 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
10747 * tree-vect-loop.c (vectorizable_reduction): Query dot-product kind.
10748 * tree-vect-patterns.c (vect_supportable_direct_optab_p): Take optional
10750 (vect_widened_op_tree): Optionally ignore
10752 (vect_recog_dot_prod_pattern): Support usdot_prod_optab.
10754 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
10757 * config/i386/driver-i386.c (host_detect_local_cpu): Check
10758 "arch [32|64]" and "tune [32|64]" for 32-bit and 64-bit codegen.
10759 Enable UINTR only for 64-bit codegen.
10760 * config/i386/i386-options.c
10761 (ix86_option_override_internal::DEF_PTA): Skip PTA_UINTR if not
10763 * config/i386/i386.h (ARCH_ARG): New.
10764 (CC1_CPU_SPEC): Pass "[arch|tune] 32" for 32-bit codegen and
10765 "[arch|tune] 64" for 64-bit codegen.
10767 2021-07-14 Richard Biener <rguenther@suse.de>
10769 PR tree-optimization/101445
10770 * tree-vect-stmts.c (vectorizable_load): Do the gap adjustment
10771 of the IV in the correct direction for negative stride
10774 2021-07-14 Jakub Jelinek <jakub@redhat.com>
10777 * godump.c (godump_str_hash): New type.
10778 (godump_container::pot_dummy_types): Use string_hash instead of
10779 ptr_hash in the hash_set.
10781 2021-07-14 Richard Biener <rguenther@suse.de>
10783 * tree-vect-loop.c (vect_find_reusable_accumulator): Handle
10784 vector types where the old vector type has a multiple of
10785 the new vector type elements.
10786 (vect_create_partial_epilog): New function, split out from...
10787 (vect_create_epilog_for_reduction): ... here.
10788 (vect_transform_cycle_phi): Reduce the re-used accumulator
10789 to the new vector type.
10791 2021-07-14 Alexandre Oliva <oliva@adacore.com>
10793 * tree-ssa-alias.c (attr_fnspec::verify): Fix index in
10794 non-'t'-sized arg check.
10796 2021-07-14 Alexandre Oliva <oliva@adacore.com>
10798 * tree-cfg.c (cleanup_dead_labels_eh): Update
10799 post_landing_pad label upon change of landing pad block's
10801 (cleanup_dead_labels): Check that a removed label is not that
10804 2021-07-13 Jonathan Wright <jonathan.wright@arm.com>
10806 * combine.c (combine_simplify_rtx): Add vec_select -> subreg
10808 * config/aarch64/aarch64.md (*zero_extend<SHORT:mode><GPI:mode>2_aarch64):
10809 Add Neon to general purpose register case for zero-extend
10811 * config/arm/vfp.md (*arm_movsi_vfp): Remove "*" from *t -> r
10812 case to prevent some cases opting to go through memory.
10813 * cse.c (fold_rtx): Add vec_select -> subreg simplification.
10814 * rtl.c (rtvec_series_p): Define predicate to determine
10815 whether a vector contains a linear series of integers.
10816 * rtl.h (rtvec_series_p): Define.
10817 * rtlanal.c (vec_series_lowpart_p): Define predicate to
10818 determine if a vector selection is equivalent to the low part
10820 * rtlanal.h (vec_series_lowpart_p): Define.
10821 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
10822 Add vec_select -> subreg simplification.
10824 2021-07-13 Paul A. Clarke <pc@us.ibm.com>
10826 * config/rs6000/smmintrin.h (_mm_testz_si128, _mm_testc_si128,
10827 _mm_testnzc_si128, _mm_test_all_ones, _mm_test_all_zeros,
10828 _mm_test_mix_ones_zeros): New.
10830 2021-07-13 Roger Sayle <roger@nextmovesoftware.com>
10831 Richard Biener <rguenther@suse.de>
10833 * gimple.c (gimple_could_trap_p_1): Make S argument a
10834 "const gimple*". Preserve constness in call to
10835 gimple_asm_volatile_p.
10836 (gimple_could_trap_p): Make S argument a "const gimple*".
10837 * gimple.h (gimple_could_trap_p_1, gimple_could_trap_p):
10838 Update function prototypes.
10840 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10842 * tree-vectorizer.h (vect_reusable_accumulator): New structure.
10843 (_loop_vec_info::main_loop_edge): New field.
10844 (_loop_vec_info::skip_main_loop_edge): Likewise.
10845 (_loop_vec_info::skip_this_loop_edge): Likewise.
10846 (_loop_vec_info::reusable_accumulators): Likewise.
10847 (_stmt_vec_info::reduc_scalar_results): Likewise.
10848 (_stmt_vec_info::reused_accumulator): Likewise.
10849 (vect_get_main_loop_result): Declare.
10850 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize
10851 reduc_scalar_inputs.
10852 (vec_info::free_stmt_vec_info): Free reduc_scalar_inputs.
10853 * tree-vect-loop-manip.c (vect_get_main_loop_result): New function.
10854 (vect_do_peeling): Fill an epilogue loop's main_loop_edge,
10855 skip_main_loop_edge and skip_this_loop_edge fields.
10856 * tree-vect-loop.c (INCLUDE_ALGORITHM): Define.
10857 (vect_emit_reduction_init_stmts): New function.
10858 (get_initial_def_for_reduction): Use it.
10859 (get_initial_defs_for_reduction): Likewise. Change the vinfo
10860 parameter to a loop_vec_info.
10861 (vect_create_epilog_for_reduction): Store the scalar results
10862 in the reduc_info. If an epilogue loop is reusing an accumulator
10863 from the main loop, and if the epilogue loop can also be skipped,
10864 try to place the reduction code in the join block. Record
10865 accumulators that could potentially be reused by epilogue loops.
10866 (vect_transform_cycle_phi): When vectorizing epilogue loops,
10867 try to reuse accumulators from the main loop. Record the initial
10868 value in reduc_info for non-SLP reductions too.
10870 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10872 * tree-vect-loop.c (get_initial_def_for_reduction): Remove
10873 adjustment handling. Take the neutral value as an argument,
10874 in place of the code argument.
10875 (vect_transform_cycle_phi): Update accordingly. Handle the
10876 initial values of cond reductions separately from code reductions.
10877 Choose the adjustment here rather than in
10878 get_initial_def_for_reduction. Sink the splat of vec_initial_def.
10880 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10882 * tree-vect-loop.c (neutral_op_for_slp_reduction): Replace with...
10883 (neutral_op_for_reduction): ...this, providing a more general
10885 (vect_create_epilog_for_reduction): Update accordingly.
10886 (vectorizable_reduction): Likewise.
10887 (vect_transform_cycle_phi): Likewise.
10889 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10891 * tree-vect-loop.c (get_initial_def_for_reduction): Take the
10892 reduc_info instead of the original stmt_vec_info.
10893 (vect_transform_cycle_phi): Update accordingly.
10895 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10897 * tree-vect-loop.c (get_initial_defs_for_reduction): Take the
10898 reduc_info as an additional parameter.
10899 (vect_transform_cycle_phi): Update accordingly.
10901 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10903 * tree-vectorizer.h: Include tree-ssa-operands.h.
10904 (vect_phi_initial_value): New function.
10905 * tree-vect-loop.c (neutral_op_for_slp_reduction): Use it.
10906 (get_initial_defs_for_reduction, info_for_reduction): Likewise.
10907 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
10908 (vect_transform_cycle_phi, vectorizable_induction): Likewise.
10910 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10912 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
10913 the phi results to vectype after creating them. Remove later
10914 conversion code that thus becomes redundant.
10916 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10918 * tree-vect-loop.c (vect_create_epilog_for_reduction): Replace
10919 the new_phis vector with a reduc_inputs vector. Combine handling
10920 of reduction chains and ncopies > 1.
10922 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10924 * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
10925 scalar_results to group_size elements after reducing down from
10926 N*group_size elements. Construct an array_slice of the live-out
10927 stmts and assert that there is one stmt per scalar result.
10929 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10931 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
10932 nested_in_vect_loop and use double_reduc everywhere. Remove dead
10933 assignment to "loop".
10935 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10937 * internal-fn.c (vectorized_internal_fn_supported_p): Handle
10938 vector types first. For scalar types, consider both the preferred
10939 vector mode and the alternative vector modes.
10940 * optabs-query.c (can_vec_mask_load_store_p): Use the same
10941 structure as above, in particular using related_vector_mode
10942 for modes provided by autovectorize_vector_modes.
10944 2021-07-13 Jakub Jelinek <jakub@redhat.com>
10945 Richard Biener <rguenther@suse.de>
10947 PR tree-optimization/101419
10948 * tree-pass.h (PROP_objsz): Define.
10949 (make_pass_early_object_sizes): Declare.
10950 * passes.def (pass_all_early_optimizations): Rename pass_object_sizes
10951 there to pass_early_object_sizes, drop parameter.
10952 (pass_all_optimizations): Move pass_object_sizes right after pass_ccp,
10953 drop parameter, move pass_post_ipa_warn right after that.
10954 * tree-object-size.c (pass_object_sizes::execute): Rename to...
10955 (object_sizes_execute): ... this. Add insert_min_max_p argument.
10956 (pass_data_object_sizes): Move after object_sizes_execute.
10957 (pass_object_sizes): Likewise. In execute method call
10958 object_sizes_execute, drop set_pass_param method and insert_min_max_p
10959 non-static data member and its initializer in the ctor.
10960 (pass_data_early_object_sizes, pass_early_object_sizes,
10961 make_pass_early_object_sizes): New.
10962 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Use
10963 (cfun->curr_properties & PROP_objsz) instead of cfun->after_inlining.
10965 2021-07-13 Kito Cheng <kito.cheng@sifive.com>
10968 * config/riscv/constraints.md ("S"): Update description and remove
10970 * doc/md.texi (Machine Constraints): Document the 'S' constraints
10973 2021-07-13 Richard Biener <rguenther@suse.de>
10976 2021-07-12 Richard Biener <rguenther@suse.de>
10978 * tree-vect-slp.c (vect_slp_region): Show the number of
10979 SLP graph entries in the optimization message.
10981 2021-07-13 Michael Meissner <meissner@linux.ibm.com>
10983 * config/rs6000/altivec.md (xxspltiw_v4sf): Change local variable
10985 * config/rs6000/rs6000-protos.h (rs6000_const_f32_to_i32): Change
10986 return type to long.
10987 * config/rs6000/rs6000.c (rs6000_const_f32_to_i32): Change return
10990 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
10992 * gimple-range-fold.cc (fold_using_range::range_of_builtin_ubsan_call):
10993 Query relation between the 2 operands and use it.
10995 2021-07-12 Sergei Trofimovich <siarheit@google.com>
10997 * doc/cfg.texi: Fix s/ei_safe_safe/ei_safe_edge/ typo.
10999 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
11002 * config/i386/predicates.md (vec_setm_sse41_operand):
11003 Rename from vec_setm_operand.
11004 (vec_setm_avx2_operand): New predicate.
11005 * config/i386/sse.md (vec_set<V_128:mode>): Use V_128 mode iterator.
11006 Use vec_setm_sse41_operand as operand 2 predicate.
11007 (vec_set<V_256_512:mode): New expander.
11008 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_sse41_operand
11009 as operand 2 predicate.
11011 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
11013 PR tree-optimization/101335
11014 * range-op.cc (operator_cast::lhs_op1_relation): Delete.
11016 2021-07-12 Andrew Pinski <apinski@marvell.com>
11018 * tree-ssa-phiopt.c (match_simplify_replacement): Move
11019 insert of the sequence before the movement of the
11020 statement. Check if to see if the statement is used
11021 outside of the original phi to see if we should move it.
11023 2021-07-12 Richard Biener <rguenther@suse.de>
11025 * dump-context.h (debug_dump_context::debug_dump_context):
11026 Add FILE * parameter defaulted to stderr.
11027 * dumpfile.c (debug_dump_context::debug_dump_context): Adjust.
11028 * tree-vect-slp.c (dot_slp_tree): New functions.
11030 2021-07-12 Richard Biener <rguenther@suse.de>
11032 PR tree-optimization/101373
11033 * tree-ssa-pre.c (prune_clobbered_mems): Also prune trapping
11034 references when the BB may not return.
11035 (compute_avail): Pass in the function we're working on and
11036 replace cfun references with it. Externally throwing
11037 const calls also possibly terminate the function.
11038 (pass_pre::execute): Pass down the function we're working on.
11039 * gcse.c (compute_hash_table_work): Externally throwing
11040 const/pure calls also need record_last_mem_set_info.
11041 * postreload-gcse.c (record_opr_changes): Looping or externally
11042 throwing const/pure calls also need record_last_mem_set_info.
11044 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
11046 * recog.c (memory_address_addr_space_p): Change the type to bool.
11047 Return true/false instead of 1/0.
11048 (offsettable_memref_p): Ditto.
11049 (offsettable_nonstrict_memref_p): Ditto.
11050 (offsettable_address_addr_space_p): Ditto.
11051 Change the type of addressp indirect function to bool.
11052 * recog.h (memory_address_addr_space_p): Change the type to bool.
11053 (strict_memory_address_addr_space_p): Ditto.
11054 (offsettable_memref_p): Ditto.
11055 (offsettable_nonstrict_memref_p): Ditto.
11056 (offsettable_address_addr_space_p): Ditto.
11057 * reload.c (maybe_memory_address_addr_space_p): Ditto.
11058 (strict_memory_address_addr_space_p): Change the type to bool.
11059 Return true/false instead of 1/0.
11060 (maybe_memory_address_addr_space_p): Change the type to bool.
11062 2021-07-12 Richard Biener <rguenther@suse.de>
11064 * tree-vect-slp.c (vect_slp_region): Show the number of
11065 SLP graph entries in the optimization message.
11067 2021-07-12 Richard Biener <rguenther@suse.de>
11069 PR tree-optimization/101394
11070 * tree-ssa-pre.c (do_pre_regular_insertion): Avoid inserting
11071 copies from abnormals for a full redundancy.
11073 2021-07-12 Richard Biener <rguenther@suse.de>
11075 PR middle-end/101423
11076 * gimple.c (gimple_could_trap_p_1): Internal function calls
11078 * tree-eh.c (tree_could_trap_p): Likewise.
11080 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
11083 * config/arm/arm_neon.h (vmul_n_u32): Replace call to builtin with
11085 (vmulq_n_u32): Likewise.
11086 (vmul_n_f32): Gate __a * __b on __FAST_MATH__.
11087 (vmulq_n_f32): Likewise.
11088 (vmul_n_f16): Likewise.
11089 (vmulq_n_f16): Likewise.
11091 2021-07-12 Martin Liska <mliska@suse.cz>
11093 PR sanitizer/101425
11094 * gcc.c (check_offload_target_name): Call
11095 candidates_list_and_hint only if we have a candidate.
11097 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
11100 * config/arm/neon.md (vec_init): Move to ...
11101 * config/arm/vec-common.md (vec_init): ... here.
11102 Change the pattern's mode to VDQX and gate it on VALID_MVE_MODE.
11104 2021-07-12 Roger Sayle <roger@nextmovesoftware.com>
11106 PR tree-optimization/101403
11107 * match.pd ((T)bswap(X)>>C): Correctly handle cases where
11108 signedness of the shift is not the same as the signedness of
11109 the type extension.
11111 2021-07-09 Roger Sayle <roger@nextmovesoftware.com>
11112 Uroš Bizjak <ubizjak@gmail.com>
11114 * config/i386/i386.md (*divmodsi4_const): Optimize SImode
11115 divmod of a constant numerator with new define_insn_and_split.
11117 2021-07-09 Iain Sandoe <iain@sandoe.co.uk>
11120 * config/i386/i386-expand.c (ix86_expand_call): If a call is
11121 to a non-local-binding, or local but to a public symbol, then
11122 assume that it might be indirected via the lazy symbol binder.
11123 Mark R10 and R10 as clobbered in that case.
11125 2021-07-09 Eric Botcazou <ebotcazou@adacore.com>
11128 * gcc.c (ASM_DEBUG_DWARF_OPTION): Set again to --gdwarf2 in
11129 the case where HAVE_AS_WORKING_DWARF_N_FLAG is not defined
11130 and HAVE_LD_BROKEN_PE_DWARF5 is defined.
11132 2021-07-09 Uroš Bizjak <ubizjak@gmail.com>
11134 * config/i386/i386.md (*udivmodsi4_pow2_zext_1): Limit the
11135 log2 range of operands[3] to [1,31].
11136 (*udivmodsi4_pow2_zext_2): Ditto. Correct insn RTX pattern.
11138 2021-07-09 Sergei Trofimovich <siarheit@google.com>
11140 * doc/md.texi: Don't split @smallexample in multiple @groups.
11142 2021-07-09 Sergei Trofimovich <siarheit@google.com>
11144 * doc/md.texi: Add missing 'see' word.
11146 2021-07-09 Andrew Pinski <apinski@marvell.com>
11148 * tree-ssa-phiopt.c (phiopt_early_allow): Change arguments
11149 to take sequence and gimple_match_op. Accept the case where
11150 op is a SSA_NAME and one statement in the sequence.
11151 Also allow constants.
11152 (gimple_simplify_phiopt): Always pass a sequence to resimplify.
11153 Update call to phiopt_early_allow. Discard the sequence if not
11156 2021-07-09 Xi Ruoyao <xry111@mengyan1223.wang>
11161 * config/mips/mips.c (mips_const_insns): Use MSA_SUPPORTED_MODE_P
11162 instead of ISA_HAS_MSA.
11163 (mips_expand_vec_unpack): Likewise.
11164 (mips_expand_vector_init): Likewise.
11166 2021-07-09 Kewen Lin <linkw@linux.ibm.com>
11168 * config/rs6000/vsx.md (mods_<mode>): Rename to...
11169 (mod<mode>3): ... this.
11170 (modu_<mode>): Rename to...
11171 (umod<mode>3): ... this.
11172 * config/rs6000/rs6000-builtin.def (MODS_V2DI, MODS_V4SI, MODU_V2DI,
11173 MODU_V4SI): Adjust.
11175 2021-07-08 Jeff Law <jeffreyalaw@gmail.com>
11177 * config/h8300/shiftrotate.md (variable shifts): Expose condition
11178 code handling for the test before the loop.
11180 2021-07-08 Martin Jambor <mjambor@suse.cz>
11183 * ipa-sra.c (class isra_call_summary): New member
11184 m_before_any_store, initialize it in the constructor.
11185 (isra_call_summary::dump): Dump the new field.
11186 (ipa_sra_call_summaries::duplicate): Copy it.
11187 (process_scan_results): Set it.
11188 (isra_write_edge_summary): Stream it.
11189 (isra_read_edge_summary): Likewise.
11190 (param_splitting_across_edge): Only override
11191 safe_to_import_accesses if m_before_any_store is set.
11193 2021-07-08 Martin Sebor <msebor@redhat.com>
11195 PR bootstrap/101374
11196 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref):
11197 Use Object Size Type 0 instead of 1.
11199 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
11201 * tree-vect-loop.c (vectorizable_reduction): Remove always-true
11204 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
11206 * match.pd: Simplify an extend-operate-truncate sequence involving
11209 2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
11210 Richard Biener <rguenther@suse.de>
11212 PR tree-optimization/40210
11213 * match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
11214 (x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
11215 when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
11217 2021-07-08 Uroš Bizjak <ubizjak@gmail.com>
11220 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
11222 * config/i386/mmx.md (V_32): New mode iterator.
11223 (mov<V_32:mode>): Use V_32 mode iterator.
11224 (*mov<V_32:mode>_internal): Ditto.
11225 (*push<V_32:mode>2_rex64): Ditto.
11226 (*push<V_32:mode>2): Ditto.
11227 (movmisalign<V_32:mode>): Ditto.
11228 (mmx_<any_shiftrt:insn>v1si3): New insn pattern.
11229 (sse4_1_<any_extend:code>v2qiv2hi2): Ditto.
11230 (vec_unpacks_lo_v4qi): New expander.
11231 (vec_unpacks_hi_v4qi): Ditto.
11232 (vec_unpacku_lo_v4qi): Ditto.
11233 (vec_unpacku_hi_v4qi): Ditto.
11234 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add V1SImode.
11235 (VALID_INT_MODE_P): Ditto.
11237 2021-07-08 Michael Meissner <meissner@linux.ibm.com>
11240 * config/rs6000/rs6000.md (udivti3): New insn.
11241 (divti3): New insn.
11242 (umodti3): New insn.
11243 (modti3): New insn.
11245 2021-07-07 Martin Sebor <msebor@redhat.com>
11247 PR tree-optimization/100137
11248 PR tree-optimization/99121
11249 PR tree-optimization/97027
11250 * builtins.c (access_ref::access_ref): Also set offmax.
11251 (access_ref::offset_in_range): Define new function.
11252 (access_ref::add_offset): Set offmax.
11253 (access_ref::inform_access): Handle access_none.
11254 (handle_mem_ref): Clear ostype.
11255 (compute_objsize_r): Handle ASSERT_EXPR.
11256 * builtins.h (struct access_ref): Add offmax member.
11257 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Use
11258 compute_objsize() and simplify.
11260 2021-07-07 Peter Bergner <bergner@linux.ibm.com>
11262 * config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
11263 and VSX_BUILTIN_STXVP.
11265 2021-07-07 Martin Sebor <msebor@redhat.com>
11268 * config/aarch64/aarch64.c (aarch64_simd_lane_bounds): Remove
11269 a stray %K from error_at() missed in r12-2088.
11271 2021-07-07 Richard Biener <rguenther@suse.de>
11273 PR tree-optimization/99728
11274 * tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
11276 (mem_refs_may_alias_p): Add assert we handled aggregate
11278 (sm_seq_valid_bb): Give up when running into aggregate copies.
11279 (ref_indep_loop_p): Handle aggregate copies as never
11280 being invariant themselves but allow other refs to be
11281 disambiguated against them.
11282 (can_sm_ref_p): Do not try to apply store-motion to aggregate
11285 2021-07-06 Indu Bhagat <indu.bhagat@oracle.com>
11288 * dwarf2ctf.c (ctf_get_AT_data_member_location): Multiply by 8 to get
11291 2021-07-06 Martin Sebor <msebor@redhat.com>
11293 * gimple-pretty-print.c (percent_G_format): Remove.
11294 * tree-diagnostic.c (default_tree_printer): Remove calls.
11295 * tree-pretty-print.c (percent_K_format): Remove.
11296 * tree-pretty-print.h (percent_K_format): Remove.
11298 2021-07-06 Martin Sebor <msebor@redhat.com>
11300 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_builtin):
11301 Remove %K and use error_at.
11302 (aarch64_expand_fcmla_builtin): Same.
11303 (aarch64_expand_builtin_tme): Same.
11304 (aarch64_expand_builtin_memtag): Same.
11305 * config/arm/arm-builtins.c (arm_expand_acle_builtin): Same.
11306 (arm_expand_builtin): Same.
11307 * config/arm/arm.c (bounds_check): Same.
11309 2021-07-06 Martin Sebor <msebor@redhat.com>
11311 * builtins.c (warn_string_no_nul): Remove %G.
11312 (maybe_warn_for_bound): Same.
11313 (warn_for_access): Same.
11314 (check_access): Same.
11315 (check_strncat_sizes): Same.
11316 (expand_builtin_strncat): Same.
11317 (expand_builtin_strncmp): Same.
11318 (expand_builtin): Same.
11319 (expand_builtin_object_size): Same.
11320 (warn_dealloc_offset): Same.
11321 (maybe_emit_free_warning): Same.
11322 * calls.c (maybe_warn_alloc_args_overflow): Same.
11323 (maybe_warn_nonstring_arg): Same.
11324 (maybe_warn_rdwr_sizes): Same.
11325 * expr.c (expand_expr_real_1): Remove %K.
11326 * gimple-fold.c (gimple_fold_builtin_strncpy): Remove %G.
11327 (gimple_fold_builtin_strncat): Same.
11328 * gimple-ssa-sprintf.c (format_directive): Same.
11329 (handle_printf_call): Same.
11330 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Same.
11331 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
11332 (maybe_diag_access_bounds): Same. Call gimple_location.
11333 (check_bounds_or_overlap): Same.
11334 * trans-mem.c (ipa_tm_scan_irr_block): Remove %K. Simplify.
11335 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Remove %G.
11336 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
11337 (maybe_diag_stxncpy_trunc): Same.
11338 (handle_builtin_stxncpy_strncat): Same.
11339 (maybe_warn_pointless_strcmp): Same.
11340 * tree-ssa-uninit.c (maybe_warn_operand): Same.
11342 2021-07-06 Uroš Bizjak <ubizjak@gmail.com>
11345 * config/i386/predicates.md (vec_setm_operand): Enable
11346 register_operand for TARGET_SSE4_1.
11347 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_operand
11348 as operand 2 predicate. Call ix86_expand_vector_set_var
11349 for non-constant index operand.
11350 (vec_setv4qi): Use vec_setm_mmx_operand as operand 2 predicate.
11351 Call ix86_expand_vector_set_var for non-constant index operand.
11353 2021-07-06 Jeff Law <jeffreyalaw@gmail.com>
11355 * config/h8300/jumpcall.md (*branch): When possible, generate
11356 the comparison in CCZN mode.
11357 * config/h8300/predicates.md (simple_memory_operand): Reject all
11358 auto-increment addressing modes.
11360 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
11362 PR bootstrap/100246
11363 * config/i386/i386.h (struct stringop_algs): Define a CTOR for
11366 2021-07-06 Richard Biener <rguenther@suse.de>
11368 * doc/md.texi (vec_fmaddsub<mode>4): Document.
11369 (vec_fmsubadd<mode>4): Likewise.
11370 * optabs.def (vec_fmaddsub$a4): Add.
11371 (vec_fmsubadd$a4): Likewise.
11372 * internal-fn.def (IFN_VEC_FMADDSUB): Add.
11373 (IFN_VEC_FMSUBADD): Likewise.
11374 * tree-vect-slp-patterns.c (addsub_pattern::recognize):
11375 Refactor to handle IFN_VEC_FMADDSUB and IFN_VEC_FMSUBADD.
11376 (addsub_pattern::build): Likewise.
11377 * tree-vect-slp.c (vect_optimize_slp): CFN_VEC_FMADDSUB
11378 and CFN_VEC_FMSUBADD are not transparent for permutes.
11379 * config/i386/sse.md (vec_fmaddsub<mode>4): New expander.
11380 (vec_fmsubadd<mode>4): Likewise.
11382 2021-07-06 Richard Biener <rguenther@suse.de>
11384 * doc/invoke.texi (fmove-loop-stores): Document.
11385 * common.opt (fmove-loop-stores): New option.
11386 * opts.c (default_options_table): Enable -fmove-loop-stores
11387 at -O1 but not -Og.
11388 * tree-ssa-loop-im.c (pass_lim::execute): Pass
11389 flag_move_loop_stores instead of true to
11390 loop_invariant_motion_in_fun.
11392 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
11394 * doc/install.texi: Document --with-dsymutil.
11396 2021-07-06 Andrew Pinski <apinski@marvell.com>
11398 PR tree-optimization/101256
11399 * dbgcnt.def (phiopt_edge_range): New counter.
11400 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
11401 Check to make sure the new name is defined in the same
11402 bb as the conditional before duplicating range info.
11403 Also add debug counter.
11405 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
11407 PR rtl-optimization/100328
11408 * config/i386/i386-options.c (ix86_option_override_internal):
11409 Set param_ira_consider_dup_in_all_alts to 0.
11411 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
11413 PR rtl-optimization/100328
11414 * doc/invoke.texi (ira-consider-dup-in-all-alts): Document new
11416 * ira.c (ira_get_dup_out_num): Adjust as parameter
11417 param_ira_consider_dup_in_all_alts.
11418 * params.opt (ira-consider-dup-in-all-alts): New.
11419 * ira-conflicts.c (process_regs_for_copy): Add one parameter
11420 single_input_op_has_cstr_p.
11421 (get_freq_for_shuffle_copy): New function.
11422 (add_insn_allocno_copies): Adjust as single_input_op_has_cstr_p.
11423 * ira-int.h (ira_get_dup_out_num): Add one bool parameter.
11425 2021-07-05 Jeff Law <jeffreyalaw@gmail.com>
11427 * config/h8300/shiftrotate.md (shift-by-variable patterns): Update to
11428 generate condition code aware RTL directly.
11430 2021-07-05 Andrew Pinski <apinski@marvell.com>
11432 PR tree-optimization/101039
11433 * match.pd (A CMP 0 ? A : -A): New patterns.
11434 * tree-ssa-phiopt.c (abs_replacement): Delete function.
11435 (tree_ssa_phiopt_worker): Don't call abs_replacement.
11436 Update comment about abs_replacement.
11438 2021-07-05 Andrew Pinski <apinski@marvell.com>
11440 * tree-ssa-phiopt.c (gimple_simplify_phiopt):
11441 If "A ? B : C" fails to simplify, try "(!A) ? C : B".
11443 2021-07-05 Andrew Pinski <apinski@marvell.com>
11445 * tree-ssa-phiopt.c (match_simplify_replacement):
11446 Add early_p argument. Call gimple_simplify_phiopt
11447 instead of gimple_simplify.
11448 (tree_ssa_phiopt_worker): Update call to
11449 match_simplify_replacement and allow unconditionally.
11450 (phiopt_early_allow): New function.
11451 (gimple_simplify_phiopt): New function.
11453 2021-07-05 Andrew Pinski <apinski@marvell.com>
11455 PR middle-end/101237
11456 * fold-const.c (negate_expr_p): Remove call to element_mode
11457 and TREE_MODE/TREE_TYPE when calling HONOR_SIGNED_ZEROS,
11458 HONOR_SIGN_DEPENDENT_ROUNDING, and HONOR_SNANS.
11459 (fold_negate_expr_1): Likewise.
11460 (const_unop): Likewise.
11461 (fold_cond_expr_with_comparison): Likewise.
11462 (fold_binary_loc): Likewise.
11463 (fold_ternary_loc): Likewise.
11464 (tree_call_nonnegative_warnv_p): Likewise.
11465 * match.pd (-(A + B) -> (-B) - A): Likewise.
11467 2021-07-05 Iain Sandoe <iain@sandoe.co.uk>
11469 * configure.ac: Handle --with-dsymutil in the same way as we
11470 do for the assembler and linker. (DEFAULT_DSYMUTIL): New.
11471 Extract the type and version for the dsymutil configured or
11472 found by the default searches.
11473 * config.in: Regenerated.
11474 * configure: Regenerated.
11475 * collect2.c (do_dsymutil): Handle locating dsymutil in the
11476 same way as for the assembler and linker.
11477 * config/darwin.h (DSYMUTIL): Delete.
11478 * gcc.c: Report a configured dsymutil correctly.
11479 * exec-tool.in: Allow for dsymutil.
11481 2021-07-05 Uroš Bizjak <ubizjak@gmail.com>
11483 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
11484 Handle V4QI and V2HI modes.
11485 (expand_vec_perm_blend): Allow 4-byte vector modes with TARGET_SSE4_1.
11486 Handle V4QI mode. Emit mmx_pblendvb32 for 4-byte modes.
11487 (expand_vec_perm_pshufb): Rewrite to use switch statemets.
11488 Handle 4-byte dual operands with TARGET_XOP and single operands
11489 with TARGET_SSSE3. Emit mmx_ppermv32 for TARGET_XOP and
11490 mmx_pshufbv4qi3 for TARGET_SSSE3.
11491 (expand_vec_perm_pblendv): Allow 4-byte vector modes with TARGET_SSE4_1.
11492 (expand_vec_perm_interleave2): Allow 4-byte vector modes.
11493 (expand_vec_perm_pshufb2): Allow 4-byte vector modes with TARGET_SSSE3.
11494 (expand_vec_perm_even_odd_1): Handle V4QI mode.
11495 (expand_vec_perm_broadcast_1): Handle V4QI mode.
11496 (ix86_vectorize_vec_perm_const): Handle V4QI mode.
11497 * config/i386/mmx.md (mmx_ppermv32): New insn pattern.
11498 (mmx_pshufbv4qi3): Ditto.
11499 (*mmx_pblendw32): Ditto.
11500 (*mmx_pblendw64): Rename from *mmx_pblendw.
11501 (mmx_punpckhbw_low): New insn_and_split pattern.
11502 (mmx_punpcklbw_low): Ditto.
11504 2021-07-05 Richard Biener <rguenther@suse.de>
11506 * tree-vect-loop-manip.c (vect_loop_versioning): Do not
11507 set LOOP_C_INFINITE on the vectorized loop.
11509 2021-07-05 Richard Biener <rguenther@suse.de>
11511 PR middle-end/101291
11512 * cfgloopmanip.c (loop_version): Set the loop copy of the
11513 versioned loop to the new loop.
11515 2021-07-04 Iain Sandoe <iain@sandoe.co.uk>
11518 * config.gcc: Ensure that Darwin biarch definitions are
11519 added before i386.h.
11520 * config/i386/darwin.h (TARGET_64BIT): Remove.
11521 (PR80556_WORKAROUND): New.
11522 (REAL_LIBGCC_SPEC): Amend to use PR80556_WORKAROUND.
11523 (DARWIN_SUBARCH_SPEC): New.
11524 * config/i386/darwin32-biarch.h (TARGET_64BIT_DEFAULT,
11525 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
11526 (REAL_LIBGCC_SPEC): Remove.
11527 * config/i386/darwin64-biarch.h (TARGET_64BIT_DEFAULT,
11528 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
11529 (REAL_LIBGCC_SPEC): Remove.
11531 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
11533 PR middle-end/101294
11534 * expr.c (store_constructor): Don't use vec_duplicate on vector.
11536 2021-07-02 Martin Sebor <msebor@redhat.com>
11538 PR middle-end/98871
11539 PR middle-end/98512
11540 * diagnostic.c (get_any_inlining_info): New.
11541 (update_effective_level_from_pragmas): Handle inlining context.
11542 (diagnostic_enabled): Same.
11543 (diagnostic_report_diagnostic): Same.
11544 * diagnostic.h (struct diagnostic_info): Add ctor.
11545 (struct diagnostic_context): Add new member.
11546 * tree-diagnostic.c (set_inlining_locations): New.
11547 (tree_diagnostics_defaults): Set new callback pointer.
11549 2021-07-02 Peter Bergner <bergner@linux.ibm.com>
11551 * config/rs6000/rs6000-builtin.def (BU_MMA_PAIR_LD, BU_MMA_PAIR_ST):
11553 (__builtin_vsx_lxvp, __builtin_vsx_stxvp): New built-ins.
11554 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Expand
11555 lxvp and stxvp built-ins.
11556 (mma_init_builtins): Handle lxvp and stxvp built-ins.
11557 (builtin_function_type): Likewise.
11558 * doc/extend.texi (__builtin_vsx_lxvp, __builtin_mma_stxvp): Document.
11560 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
11562 * config/h8300/h8300-protos.h (compute_a_shift_cc): Accept
11563 additional argument for the code.
11564 * config/h8300/h8300.c (compute_a_shift_cc): Accept additional
11565 argument for the code. Just return if the ZN bits are useful or
11566 not rather than the old style CC_* enums.
11567 * config/h8300/shiftrotate.md (shiftqi_noscratch): Move before
11568 more generic shiftqi patterns.
11569 (shifthi_noscratch, shiftsi_noscratch): Similarly.
11570 (shiftqi_noscratch_set_flags): New pattern.
11571 (shifthi_noscratch_set_flags, shiftsi_noscratch_set_flags): Likewise.
11573 2021-07-02 Andrew MacLeod <amacleod@redhat.com>
11575 PR tree-optimization/101223
11576 * range-op.cc (build_lt): Add -1 for signed values.
11577 (built_gt): Subtract -1 for signed values.
11579 2021-07-02 David Faust <david.faust@oracle.com>
11581 * btfout.c (get_btf_kind): Support BTF_KIND_FLOAT.
11582 (btf_asm_type): Likewise.
11584 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
11586 * config/h8300/h8300-protos.h (output_a_shift): Make first argument
11587 an array of rtx rather than a pointer to rtx. Add code argument.
11588 (compute_a_shift_length): Similarly.
11589 * config/h8300/h8300.c (h8300_shift_costs): Adjust now that the
11590 shift itself isn't an operand. Create dummy operand[0] to carry
11591 a mode and pass a suitable rtx code to compute_a_shift_length.
11592 (get_shift_alg): Adjust operand number of clobber in output templates.
11593 (output_a_shift): Make first argument an array of rtx rather than
11594 a pointer to rtx. Add code argument for the type of shift.
11595 Adjust now that the shift itself is no longer an operand.
11596 (compute_a_shift_length): Similarly.
11597 * config/h8300/shiftrotate.md (shiftqi, shifthi, shiftsi): Use an
11598 iterator rather than nshift_operator.
11599 (shiftqi_noscratch, shifthi_noscratch, shiftsi_noscratch): Likewise.
11600 (shiftqi_clobber_flags): Adjust to API changes in output_a_shift
11601 and compute_a_shift_length.
11602 (shiftqi_noscratch_clobber_flags): Likewise.
11603 (shifthi_noscratch_clobber_flags): Likewise.
11604 (shiftsi_noscratch_clobber_flags): Likewise.
11606 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
11609 * config/darwin.h (DSYMUTIL_SPEC): Do not try to run
11610 dsymutil for BTF/CTF.
11612 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
11615 * config/darwin.h (CTF_INFO_SECTION_NAME): Update the
11616 segment to include BTF.
11617 (BTF_INFO_SECTION_NAME): New.
11619 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
11621 * config/m32r/m32r-protos.h (call_operand): Adjust return type.
11622 (small_data_operand, memreg_operand, small_insn_p): Likewise.
11623 * config/m32r/m32r.c (call_operand): Adjust return type.
11624 (small_data_operand, memreg_operand): Likewise.
11626 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
11628 * config/frv/frv-protos.h (integer_register_operand): Adjust return
11630 (frv_load_operand, gpr_or_fpr_operand, gpr_no_subreg_operand): Likewise.
11631 (fpr_or_int6_operand, gpr_or_int_operand); Likewise.
11632 (gpr_or_int12_operand, gpr_or_int10_operand); Likewise.
11633 (move_source_operand, move_destination_operand): Likewise.
11634 (condexec_source_operand, condexec_dest_operand): Likewise.
11635 (lr_operand, gpr_or_memory_operand, fpr_or_memory_operand): Likewise.
11636 (reg_or_0_operand, fcc_operand, icc_operand, cc_operand): Likewise.
11637 (fcr_operand, icr_operand, cr_operand, call_operand): Likewise.
11638 (fpr_operand, even_reg_operand, odd_reg_operand): Likewise.
11639 (even_gpr_operand, odd_gpr_operand, quad_fpr_operand): Likewise.
11640 (even_fpr_operand, odd_fpr_operand): Likewise.
11641 (dbl_memory_one_insn_operand, dbl_memory_two_insn_operand): Likewise.
11642 (int12_operand, int6_operand, int5_operand, uint5_operand): Likewise.
11643 (uint4_operand, uint1_operand, int_2word_operand): Likewise
11644 (upper_int16_operand, uint16_operand, symbolic_operand): Likewise.
11645 (relational_operator, float_relational_operator): Likewise.
11646 (ccr_eqne_operator, minmax_operator): Likewise.
11647 (condexec_si_binary_operator, condexec_si_media_operator): Likewise.
11648 (condexec_si_divide_operator, condexec_si_unary_operator): Likewise.
11649 (condexec_sf_conv_operator, condexec_sf_add_operator): Likewise.
11650 (intop_compare_operator, acc_operand, even_acc_operand): Likewise.
11651 (quad_acc_operand, accg_operand): Likewise.
11653 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
11655 * config/stormy16/stormy16-protos.h (xstormy16_below_100_symbol): Change
11656 return type to a bool.
11657 (nonimmediate_nonstack_operand): Likewise.
11658 (xstormy16_splittable_below100_operand): Likewise.
11659 * config/stormy16/stormy16.c (xstormy16_below_100_symbol): Fix
11661 (xstormy16_splittable_below100_operand): Likewise.
11663 2021-07-02 Richard Biener <rguenther@suse.de>
11665 PR tree-optimization/101293
11666 * tree-ssa-loop-im.c (mem_ref_hasher::equal): Compare MEM_REF bases
11667 with combined offsets.
11668 (gather_mem_refs_stmt): Hash MEM_REFs as if their offset were
11669 combined with the rest of the offset.
11671 2021-07-02 Eric Botcazou <ebotcazou@adacore.com>
11673 * config/i386/i386.c (asm_preferred_eh_data_format): Always use the
11674 PIC encodings for PE-COFF targets.
11676 2021-07-02 Jakub Jelinek <jakub@redhat.com>
11679 * config/i386/i386-expand.c (ix86_broadcast_from_integer_constant):
11680 Return nullptr for TImode inner mode.
11682 2021-07-02 Richard Biener <rguenther@suse.de>
11684 PR tree-optimization/101280
11685 PR tree-optimization/101173
11686 * gimple-loop-interchange.cc
11687 (tree_loop_interchange::valid_data_dependences): Properly
11688 guard all dependence checks with DDR_REVERSED_P or its
11691 2021-07-02 Hongyu Wang <hongyu.wang@intel.com>
11693 * config/i386/i386-expand.c (ix86_expand_builtin):
11694 Add branch to clear odata when ZF is set for asedecenc_expand
11695 and wideaesdecenc_expand.
11697 2021-07-02 Eugene Rozenfeld <erozen@microsoft.com>
11699 * config/i386/gcc-auto-profile: regenerate
11701 2021-07-02 liuhongt <hongtao.liu@intel.com>
11703 * config/i386/sse.md (trunc<mode><pmov_dst_4>2): Refined to ..
11704 (trunc<mode><pmov_dst_4_lower>2): this.
11706 2021-07-01 David Malcolm <dmalcolm@redhat.com>
11708 * diagnostic.h (diagnostic_context::m_file_cache): New field.
11709 * input.c (class fcache): Rename to...
11710 (class file_cache_slot): ...this, making most members private and
11711 prefixing fields with "m_".
11712 (file_cache_slot::get_file_path): New accessor.
11713 (file_cache_slot::get_use_count): New accessor.
11714 (file_cache_slot::missing_trailing_newline_p): New accessor.
11715 (file_cache_slot::inc_use_count): New.
11716 (fcache_buffer_size): Move to...
11717 (file_cache_slot::buffer_size): ...here.
11718 (fcache_line_record_size): Move to...
11719 (file_cache_slot::line_record_size): ...here.
11720 (fcache_tab): Delete, in favor of global_dc->m_file_cache.
11721 (fcache_tab_size): Move to file_cache::num_file_slots.
11722 (diagnostic_file_cache_init): Update for move of fcache_tab
11723 to global_dc->m_file_cache.
11724 (diagnostic_file_cache_fini): Likewise.
11725 (lookup_file_in_cache_tab): Convert to...
11726 (file_cache::lookup_file): ...this.
11727 (diagnostics_file_cache_forcibly_evict_file): Update for move of
11728 fcache_tab to global_dc->m_file_cache, moving most of
11729 implementation to...
11730 (file_cache::forcibly_evict_file): ...this new function and...
11731 (file_cache_slot::evict): ...this new function.
11732 (evicted_cache_tab_entry): Convert to...
11733 (file_cache::evicted_cache_tab_entry): ...this.
11734 (add_file_to_cache_tab): Convert to...
11735 (file_cache::add_file): ...this, moving bulk of implementation
11737 (file_cache_slot::create): ..this new function.
11738 (file_cache::file_cache): New.
11739 (file_cache::~file_cache): New.
11740 (lookup_or_add_file_to_cache_tab): Convert to...
11741 (file_cache::lookup_or_add_file): ..this new function.
11742 (fcache::fcache): Rename to...
11743 (file_cache_slot::file_cache_slot): ...this, adding "m_" prefixes
11745 (fcache::~fcache): Rename to...
11746 (file_cache_slot::~file_cache_slot): ...this, adding "m_" prefixes
11748 (needs_read): Convert to...
11749 (file_cache_slot::needs_read_p): ...this.
11750 (needs_grow): Convert to...
11751 (file_cache_slot::needs_grow_p): ...this.
11752 (maybe_grow): Convert to...
11753 (file_cache_slot::maybe_grow): ...this.
11754 (read_data): Convert to...
11755 (file_cache_slot::read_data): ...this.
11756 (maybe_read_data): Convert to...
11757 (file_cache_slot::maybe_read_data): ...this.
11758 (get_next_line): Convert to...
11759 (file_cache_slot::get_next_line): ...this.
11760 (goto_next_line): Convert to...
11761 (file_cache_slot::goto_next_line): ...this.
11762 (read_line_num): Convert to...
11763 (file_cache_slot::read_line_num): ...this.
11764 (location_get_source_line): Update for moving of globals to
11765 global_dc->m_file_cache.
11766 (location_missing_trailing_newline): Likewise.
11767 * input.h (class file_cache_slot): New forward decl.
11768 (class file_cache): New.
11770 2021-07-01 Michael Meissner <meissner@linux.ibm.com>
11772 * config/rs6000/rs6000.c (rs6000_maybe_emit_fp_cmove): Add IEEE
11773 128-bit floating point conditional move support.
11774 (have_compare_and_set_mask): Add IEEE 128-bit floating point
11776 * config/rs6000/rs6000.md (mov<mode>cc, IEEE128 iterator): New insn.
11777 (mov<mode>cc_p10, IEEE128 iterator): New insn.
11778 (mov<mode>cc_invert_p10, IEEE128 iterator): New insn.
11779 (fpmask<mode>, IEEE128 iterator): New insn.
11780 (xxsel<mode>, IEEE128 iterator): New insn.
11782 2021-07-01 Iain Sandoe <iain@sandoe.co.uk>
11785 * config/darwin.h (CTF_INFO_SECTION_NAME): New.
11787 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
11789 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
11791 * config/i386/i386-protos.h (ix86_expand_vector_init_duplicate):
11793 * config/i386/sse.md (INT_BROADCAST_MODE): New mode iterator.
11794 (vec_duplicate<mode>): New expander.
11796 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
11799 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
11801 (ix86_byte_broadcast): New function.
11802 (ix86_convert_const_wide_int_to_broadcast): Likewise.
11803 (ix86_expand_move): Convert CONST_WIDE_INT to broadcast if mode
11804 size is 16 bytes or bigger.
11805 (ix86_broadcast_from_integer_constant): New function.
11806 (ix86_expand_vector_move): Convert CONST_WIDE_INT and CONST_VECTOR
11807 to broadcast if mode size is 16 bytes or bigger.
11808 * config/i386/i386-protos.h (ix86_gen_scratch_sse_rtx): New
11810 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): New function.
11812 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
11814 * config/i386/predicates.md (ix86_endbr_immediate_operand):
11815 Return true/false instead of 1/0.
11816 (movq_parallel): Ditto.
11818 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
11820 * recog.c (general_operand): Return true/false instead of 1/0.
11821 (register_operand): Ditto.
11822 (immediate_operand): Ditto.
11823 (const_int_operand): Ditto.
11824 (const_scalar_int_operand): Ditto.
11825 (const_double_operand): Ditto.
11826 (push_operand): Ditto.
11827 (pop_operand): Ditto.
11828 (memory_operand): Ditto.
11829 (indirect_operand): Ditto.
11831 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
11833 * genpreds.c (write_predicate_subfunction):
11834 Change the type of written subfunction to bool.
11835 (write_one_predicate_function):
11836 Change the type of written function to bool.
11837 (write_tm_preds_h): Ditto.
11838 * recog.h (*insn_operand_predicate_fn): Change the type to bool.
11839 * recog.c (general_operand): Change the type to bool.
11840 (address_operand): Ditto.
11841 (register_operand): Ditto.
11842 (pmode_register_operand): Ditto.
11843 (scratch_operand): Ditto.
11844 (immediate_operand): Ditto.
11845 (const_int_operand): Ditto.
11846 (const_scalar_int_operand): Ditto.
11847 (const_double_operand): Ditto.
11848 (nonimmediate_operand): Ditto.
11849 (nonmemory_operand): Ditto.
11850 (push_operand): Ditto.
11851 (pop_operand): Ditto.
11852 (memory_operand): Ditto.
11853 (indirect_operand): Ditto.
11854 (ordered_comparison_operator): Ditto.
11855 (comparison_operator): Ditto.
11856 * config/i386/i386-expand.c (ix86_expand_sse_cmp):
11857 Change the type of indirect predicate function to bool.
11858 * config/rs6000/rs6000.c (easy_vector_constant):
11859 Change the type to bool.
11860 * config/mips/mips-protos.h (m16_based_address_p):
11861 Change the type of operand 3 to bool.
11863 2021-07-01 Richard Biener <rguenther@suse.de>
11865 PR tree-optimization/101280
11866 PR tree-optimization/101173
11867 * gimple-loop-interchange.cc
11868 (tree_loop_interchange::valid_data_dependences): Revert
11869 previous change and instead correctly handle DDR_REVERSED_P
11872 2021-07-01 Richard Biener <rguenther@suse.de>
11874 PR tree-optimization/101278
11875 * tree-ssa-dse.c (dse_classify_store): First check for
11876 uses, then ignore stmt for chaining purposes.
11878 2021-07-01 Richard Biener <rguenther@suse.de>
11880 PR tree-optimization/100778
11881 * tree-vect-slp.c (vect_schedule_slp_node): Do not place trapping
11882 vectorized ops ahead of their scalar BB.
11884 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
11887 * config/i386/i386.md (*nabs<dwi>2_doubleword):
11888 New insn_and_split pattern.
11889 (*nabs<dwi>2_1): Ditto.
11890 * config/i386/i386-features.c
11891 (general_scalar_chain::compute_convert_gain):
11892 Handle (NEG (ABS (...))) RTX. Rewrite src code
11893 scanner as switch statement.
11894 (general_scalar_chain::convert_insn):
11895 Handle (NEG (ABS (...))) RTX.
11896 (general_scalar_to_vector_candidate_p):
11897 Detect (NEG (ABS (...))) RTX. Reorder case statements
11898 for (AND (NOT (...) ...)) fallthrough.
11900 2021-07-01 Richard Biener <rguenther@suse.de>
11902 PR tree-optimization/101178
11903 * tree-vect-slp.c (slpg_vertex::materialize): Remove.
11904 (slpg::perm_in): Add.
11905 (slpg::get_perm_in): Remove.
11906 (slpg::get_perm_materialized): Add.
11907 (vect_optimize_slp): Handle VEC_PERM nodes more optimally
11908 during permute propagation and materialization.
11910 2021-07-01 Jakub Jelinek <jakub@redhat.com>
11913 * dwarf2out.c (loc_list_from_tree_1): Handle COMPOUND_LITERAL_EXPR.
11915 2021-07-01 Jakub Jelinek <jakub@redhat.com>
11917 PR middle-end/94366
11918 * omp-low.c (lower_rec_input_clauses): Rename is_fp_and_or to
11919 is_truth_op, set it for TRUTH_*IF_EXPR regardless of new_var's type,
11920 use boolean_type_node instead of integer_type_node as NE_EXPR type.
11921 (lower_reduction_clauses): Likewise.
11923 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
11925 * config/gcn/gcn.c: Include dwarf2.h.
11926 (gcn_addr_space_debug): New function.
11927 (TARGET_ADDR_SPACE_DEBUG): New hook.
11929 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
11931 * common/config/gcn/gcn-common.c
11932 (gcn_option_optimization_table): Change OPT_fomit_frame_pointer to -O3.
11933 * config/gcn/gcn.c (gcn_expand_prologue): Prefer the frame pointer
11935 (gcn_expand_prologue): Prefer the frame pointer when emitting CFI.
11936 (gcn_frame_pointer_rqd): New function.
11937 (TARGET_FRAME_POINTER_REQUIRED): New hook.
11939 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
11941 * config/gcn/gcn.c (move_callee_saved_registers): Emit CFI notes for
11942 prologue register saves.
11943 (gcn_debug_unwind_info): Use UI_DWARF2.
11944 (gcn_dwarf_register_number): Map DWARF_LINK_REGISTER to DWARF PC.
11945 (gcn_dwarf_register_span): DWARF_LINK_REGISTER doesn't span.
11946 * config/gcn/gcn.h: (DWARF_FRAME_RETURN_COLUMN): New define.
11947 (DWARF_LINK_REGISTER): New define.
11948 (FIRST_PSEUDO_REGISTER): Increment.
11949 (FIXED_REGISTERS): Add entry for DWARF_LINK_REGISTER.
11950 (CALL_USED_REGISTERS): Likewise.
11951 (REGISTER_NAMES): Likewise.
11953 2021-06-30 Richard Biener <rguenther@suse.de>
11955 PR tree-optimization/101267
11956 * tree-vect-stmts.c (vect_check_scalar_mask): Adjust
11957 API and use SLP compatible interface of vect_is_simple_use.
11958 Reject not vectorized SLP defs for callers that do not support
11960 (vect_check_store_rhs): Handle masked stores and pass down
11961 the appropriate operator index.
11962 (vectorizable_call): Adjust.
11963 (vectorizable_store): Likewise.
11964 (vectorizable_load): Likewise. Handle SLP pecularity of
11966 (vect_is_simple_use): Remove special-casing of masked stores.
11968 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
11970 * common.opt (foffload): Remove help as Driver only.
11971 * gcc.c (display_help): Add -foffload.
11973 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
11975 * gcc.c (close_at_file, execute): Replace alloca by XALLOCAVEC.
11976 (check_offload_target_name): Fix splitting OFFLOAD_TARGETS into
11977 a candidate list; better inform no offload target is configured
11978 and fix hint extraction when passed target is not '\0' at [len].
11979 * common.opt (foffload): Add tailing '.'.
11980 (foffload-options): Likewise; fix flag name in the help string.
11982 2021-06-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
11985 * config/arm/arm_neon.h: Move vabs intrinsics before vcage_f32.
11986 (vcage_f32): Gate comparison on __FAST_MATH__.
11987 (vcageq_f32): Likewise.
11988 (vcale_f32): Likewise.
11989 (vcaleq_f32): Likewise.
11990 (vcagt_f32): Likewise.
11991 (vcagtq_f32): Likewise.
11992 (vcalt_f32): Likewise.
11993 (vcaltq_f32): Likewise.
11994 (vcage_f16): Likewise.
11995 (vcageq_f16): Likewise.
11996 (vcale_f16): Likewise.
11997 (vcaleq_f16): Likewise.
11998 (vcagt_f16): Likewise.
11999 (vcagtq_f16): Likewise.
12000 (vcalt_f16): Likewise.
12001 (vcaltq_f16): Likewise.
12003 2021-06-30 Richard Biener <rguenther@suse.de>
12005 PR tree-optimization/101264
12006 * tree-vect-slp.c (vect_optimize_slp): Propagate the
12007 computed perm_in to all "any" permute successors
12008 we cannot de-duplicate immediately.
12010 2021-06-30 liuhongt <hongtao.liu@intel.com>
12013 * config/i386/sse.md
12014 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>):
12016 (avx512f_sfixupimm<mode><maskz_scalar_name><round_saeonly_name>):
12018 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>"): Refined.
12019 * config/i386/subst.md (maskz_scalar): New define_subst.
12020 (maskz_scalar_name): New subst_attr.
12021 (maskz_scalar_op5): Ditto.
12022 (round_saeonly_maskz_scalar_op5): Ditto.
12023 (round_saeonly_maskz_scalar_operand5): Ditto.
12025 2021-06-30 David Edelsohn <dje.gcc@gmail.com>
12027 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
12028 Increase code CSECT alignment to at least 32 bytes.
12029 * config/rs6000/xcoff.h (TEXT_SECTION_ASM_OP): Add 32 byte
12030 alignment designation.
12032 2021-06-29 Sergei Trofimovich <siarheit@google.com>
12034 * doc/generic.texi: Fix s/net yet/not yet/ typo.
12036 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
12038 PR tree-optimization/101254
12039 * range-op.cc (operator_minus::op1_op2_relation_effect): Check for
12040 wrapping/non-wrapping when setting the result range.
12042 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
12044 * value-query.cc (gimple_range_global): Allow phis.
12046 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
12048 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): Use stmt.
12049 (simplify_using_ranges::op_with_boolean_value_range_p): Add a
12050 statement for location context.
12051 (check_for_binary_op_overflow): Ditto.
12052 (simplify_using_ranges::get_vr_for_comparison): Ditto.
12053 (simplify_using_ranges::compare_name_with_value): Ditto.
12054 (simplify_using_ranges::compare_names): Ditto.
12055 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Ditto.
12056 (simplify_using_ranges::simplify_truth_ops_using_ranges): Ditto.
12057 (simplify_using_ranges::simplify_min_or_max_using_ranges): Ditto.
12058 (simplify_using_ranges::simplify_internal_call_using_ranges): Ditto.
12059 (simplify_using_ranges::two_valued_val_range_p): Ditto.
12060 (simplify_using_ranges::simplify): Ditto.
12061 * vr-values.h: Adjust prototypes.
12063 2021-06-29 Uroš Bizjak <ubizjak@gmail.com>
12066 * config/i386/mmx.md (vec_addsubv2sf3): New insn pattern.
12068 2021-06-29 Julian Brown <julian@codesourcery.com>
12070 * config/gcn/gcn.c (gcn_init_libfuncs): New function.
12071 (TARGET_INIT_LIBFUNCS): Define target hook using above function.
12072 * config/gcn/gcn.h (UNITS_PER_WORD): Define to 8 for IN_LIBGCC2, 4
12074 (LIBGCC2_UNITS_PER_WORD, BITS_PER_WORD): Remove definitions.
12075 (MAX_FIXED_MODE_SIZE): Change to 128.
12077 2021-06-29 Julian Brown <julian@codesourcery.com>
12079 * config/gcn/gcn.md (UNSPEC_FLBIT_INT): New unspec constant.
12080 (s_mnemonic): Add clrsb.
12081 (gcn_flbit<mode>_int): Add insn pattern for SImode/DImode.
12082 (clrsb<mode>2): Add expander for SImode/DImode.
12084 2021-06-29 Julian Brown <julian@codesourcery.com>
12086 * config/gcn/gcn.md (<su>mulsidi3, <su>mulsidi3_reg, <su>mulsidi3_imm,
12087 muldi3): Add patterns.
12089 2021-06-29 Julian Brown <julian@codesourcery.com>
12091 * config/gcn/gcn.md (<su>mulsi3_highpart): Change to expander.
12092 (<su>mulsi3_highpart_reg, <su>mulsi3_highpart_imm): New patterns.
12094 2021-06-29 Julian Brown <julian@codesourcery.com>
12096 * config/gcn/gcn.md (mulsi3): Make s_mulk_i32 variant clobber SCC.
12098 2021-06-29 Joseph Myers <joseph@codesourcery.com>
12100 * btfout.c, ctfout.c: Include "memmodel.h".
12102 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
12104 * gcc.c (check_offload_target_name): Cast len argument to
12105 %q.*s to 'int'; avoid -Wstringop-truncation warning.
12107 2021-06-29 Richard Biener <rguenther@suse.de>
12109 * tree-vect-slp.c (vect_optimize_slp): Forward propagate
12110 to "any" permute nodes and relax "any" permute proapgation
12111 during iterative backward propagation.
12113 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
12116 * common.opt (-foffload=): Update description.
12117 (-foffload-options=): New.
12118 * doc/invoke.texi (C Language Options): Document
12119 -foffload and -foffload-options.
12120 * gcc.c (check_offload_target_name): New, split off from
12121 handle_foffload_option.
12122 (check_foffload_target_names): New.
12123 (handle_foffload_option): Handle -foffload=default.
12124 (driver_handle_option): Update for -foffload-options.
12125 * lto-opts.c (lto_write_options): Use -foffload-options
12126 instead of -foffload.
12127 * lto-wrapper.c (merge_and_complain, append_offload_options):
12129 * opts.c (common_handle_option): Likewise.
12131 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
12133 * doc/invoke.texi (C Language Options): Sort options
12134 alphabetically in optlist and also the description itself.
12135 Remove leftover -fallow-single-precision from and add missing
12136 -fgnu-tm to the optlist.
12138 2021-06-29 Richard Biener <rguenther@suse.de>
12140 * tree-vect-slp.c (slpg_vertex::visited): Remove.
12141 (vect_slp_perms_eq): Handle -1 permutes.
12142 (vect_optimize_slp): Rewrite permute propagation.
12144 2021-06-29 Jakub Jelinek <jakub@redhat.com>
12147 * match.pd ((intptr_t)x eq/ne CST to x eq/ne (typeof x) CST): Don't
12148 perform the optimization in GENERIC when sanitizing and x has a
12151 2021-06-29 Richard Biener <rguenther@suse.de>
12153 PR tree-optimization/101242
12154 * tree-vect-slp.c (vect_slp_build_vertices): Force-add
12155 PHIs with not represented initial values as leafs.
12157 2021-06-29 Jan-Benedict Glaw <jbglaw@getslash.de>
12159 * config/pdp11/pdp11.h (ASM_OUTPUT_SKIP): Fix signedness warning.
12160 * config/pdp11/pdp11.c (pdp11_asm_print_operand_punct_valid_p): Remove
12161 "register" keyword.
12162 (pdp11_initial_elimination_offset) Remove unused variable.
12163 (pdp11_cmp_length) Ditto.
12164 (pdp11_insn_cost): Ditto, and fix signedness warning.
12166 2021-06-29 David Edelsohn <dje.gcc@gmail.com>
12168 * btfout.c: Include tm_p.h.
12171 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
12173 * config/bpf/bpf.c (bpf_expand_prologue): Do not mark insns as
12175 (bpf_expand_epilogue): Likewise.
12176 * config/bpf/bpf.h (DWARF2_FRAME_INFO): Define to 0.
12177 Do not define DBX_DEBUGGING_INFO.
12179 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
12181 * doc/invoke.texi: Document the CTF and BTF debug info options.
12183 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
12184 David Faust <david.faust@oracle.com>
12185 Jose E. Marchesi <jose.marchesi@oracle.com>
12186 Weimin Pan <weimin.pan@oracle.com>
12188 * Makefile.in: Add ctfc.*, ctfout.c and btfout.c files to
12189 GTFILES. Add new object files.
12190 * common.opt: Add CTF and BTF debug info options.
12191 * btfout.c: New file.
12192 * ctfc.c: Likewise.
12193 * ctfc.h: Likewise.
12194 * ctfout.c: Likewise.
12195 * dwarf2ctf.c: Likewise.
12196 * dwarf2ctf.h: Likewise.
12197 * dwarf2cfi.c (dwarf2out_do_frame): Acknowledge CTF_DEBUG and
12199 * dwarf2out.c (dwarf2out_source_line): Likewise.
12200 (dwarf2out_finish): Skip emitting DWARF if CTF or BTF are to
12202 (debug_format_do_cu): New function.
12203 (dwarf2out_early_finish): Traverse DIEs and emit CTF/BTF for
12205 Include dwarf2ctf.c.
12206 * final.c (dwarf2_debug_info_emitted_p): Acknowledge DWARF-based debug
12208 * flag-types.h (enum debug_info_type): Add CTF_DEBUG and BTF_DEBUG.
12209 (CTF_DEBUG): New bitmask.
12210 (BTF_DEBUG): Likewise.
12211 (enum ctf_debug_info_levels): New enum.
12212 * gengtype.c (open_base_files): Handle ctfc.h.
12213 (main): Handle uint32_t type.
12214 * flags.h (btf_debuginfo_p): New definition.
12215 (dwarf_based_debuginfo_p): Likewise.
12216 * opts.c (debug_type_names): Add entries for CTF and BTF.
12217 (btf_debuginfo_p): New function.
12218 (dwarf_based_debuginfo_p): Likewise.
12219 (common_handle_option): Handle -gctfN and -gbtf options.
12220 (set_debug_level): Set CTF_DEBUG, BTF_DEBUG whenever appropriate.
12221 * toplev.c (process_options): Inform the user and ignore -gctfLEVEL if
12224 2021-06-28 Jose E. Marchesi <jose.marchesi@oracle.com>
12226 * dwarf2out.c (AT_class): Function is no longer static.
12227 (AT_int): Likewise.
12228 (AT_unsigned): Likewise.
12229 (AT_loc): Likewise.
12230 (get_AT): Likewise.
12231 (get_AT_string): Likewise.
12232 (get_AT_flag): Likewise.
12233 (get_AT_unsigned): Likewise.
12234 (get_AT_ref): Likewise.
12235 (new_die_raw): Likewise.
12236 (lookup_decl_die): Likewise.
12237 (base_type_die): Likewise.
12238 (add_name_attribute): Likewise.
12239 (add_AT_int): Likewise.
12240 (add_AT_unsigned): Likewise.
12241 (add_AT_loc): Likewise.
12242 (dw_get_die_tag): New function.
12243 (dw_get_die_child): Likewise.
12244 (dw_get_die_sib): Likewise.
12245 (struct dwarf_file_data): Move from here to dwarf2out.h
12246 (struct dw_attr_struct): Likewise.
12247 * dwarf2out.h: Analogous changes.
12249 2021-06-28 Martin Jambor <mjambor@suse.cz>
12252 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
12253 members m_dead_stmts and m_dead_ssas.
12254 * ipa-param-manipulation.c
12255 (ipa_param_body_adjustments::mark_dead_statements): New function.
12256 (ipa_param_body_adjustments::common_initialization): Call it on
12257 all removed but not split parameters.
12258 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
12260 (ipa_param_body_adjustments::modify_call_stmt): Remove arguments that
12262 * tree-inline.c (remap_gimple_stmt): Do not copy dead statements, reset
12263 dead debug statements.
12264 (copy_phis_for_bb): Do not copy dead PHI nodes.
12266 2021-06-28 Martin Jambor <mjambor@suse.cz>
12269 * symtab-clones.h (clone_info): Removed member param_adjustments.
12270 * ipa-param-manipulation.h: Adjust initial comment to reflect how we
12271 deal with pass-through splits now.
12272 (ipa_param_performed_split): Removed.
12273 (ipa_param_adjustments::modify_call): Adjusted parameters.
12274 (class ipa_param_body_adjustments): Adjusted parameters of
12275 register_replacement, modify_gimple_stmt and modify_call_stmt.
12276 (ipa_verify_edge_has_no_modifications): Declare.
12277 (ipa_edge_modifications_finalize): Declare.
12278 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Remove
12279 performed_splits processing, pas only edge to padjs->modify_call,
12280 check that call arguments were not modified if they should not have
12282 * cgraphclones.c (cgraph_node::create_clone): Do not copy performed
12284 * ipa-param-manipulation.c (struct pass_through_split_map): New type.
12285 (ipa_edge_modification_info): Likewise.
12286 (ipa_edge_modification_sum): Likewise.
12287 (ipa_edge_modifications): New edge summary.
12288 (ipa_verify_edge_has_no_modifications): New function.
12289 (transitive_split_p): Removed.
12290 (transitive_split_map): Likewise.
12291 (init_transitive_splits): Likewise.
12292 (ipa_param_adjustments::modify_call): Adjusted to use the new edge
12293 summary instead of performed_splits.
12294 (ipa_param_body_adjustments::register_replacement): Drop dummy
12295 parameter, set base_index of the created ipa_param_body_replacement.
12296 (phi_arg_will_live_p): New function.
12297 (ipa_param_body_adjustments::common_initialization): Do not create
12298 IPA_SRA dummy decls.
12299 (simple_tree_swap_info): Removed.
12300 (remap_split_decl_to_dummy): Likewise.
12301 (record_argument_state_1): New function.
12302 (record_argument_state): Likewise.
12303 (ipa_param_body_adjustments::modify_call_stmt): New parameter
12304 orig_stmt. Do not work with dummy decls, save necessary info about
12305 changes to ipa_edge_modifications.
12306 (ipa_param_body_adjustments::modify_gimple_stmt): New parameter
12307 orig_stmt, pass it to modify_call_stmt.
12308 (ipa_param_body_adjustments::modify_cfun_body): Adjust call to
12309 modify_gimple_stmt.
12310 (ipa_edge_modifications_finalize): New function.
12311 * tree-inline.c (remap_gimple_stmt): Pass original statement to
12312 modify_gimple_stmt.
12313 (copy_phis_for_bb): Do not copy dead PHI nodes.
12314 (expand_call_inline): Do not remap performed_splits.
12315 (update_clone_info): Likewise.
12316 * toplev.c: Include ipa-param-manipulation.h.
12317 (toplev::finalize): Call ipa_edge_modifications_finalize.
12319 2021-06-28 Andrew Pinski <apinski@marvell.com>
12321 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Duplicate range
12322 info if we're the only things setting the target PHI.
12323 (value_replacement): Don't duplicate range here.
12324 (minmax_replacement): Likewise.
12326 2021-06-28 Richard Biener <rguenther@suse.de>
12328 PR tree-optimization/101229
12329 * gimple-walk.c (gimple_walk_op): Handle PHIs.
12331 2021-06-28 Martin Liska <mliska@suse.cz>
12333 * config/v850/v850.c (construct_dispose_instruction): Allocate
12335 (construct_prepare_instruction): Likewise.
12337 2021-06-28 Martin Liska <mliska@suse.cz>
12339 * config/v850/v850.c (v850_option_override): Build default
12341 (v850_can_inline_p): New. Allow MASK_PROLOG_FUNCTION to be
12342 ignored for inlining.
12343 (TARGET_CAN_INLINE_P): New.
12345 2021-06-28 Richard Biener <rguenther@suse.de>
12347 PR tree-optimization/101207
12348 * tree-vect-slp.c (vect_optimize_slp): Do BB reduction
12349 permute eliding for load permutations properly.
12351 2021-06-28 Richard Biener <rguenther@suse.de>
12353 PR tree-optimization/101173
12354 * gimple-loop-interchange.cc
12355 (tree_loop_interchange::valid_data_dependences): Disallow outer
12356 loop dependence distance of zero.
12358 2021-06-28 liuhongt <hongtao.liu@intel.com>
12361 * config/i386/sse.md (*avx_cmp<mode>3_lt): New
12362 define_insn_and_split.
12363 (*avx_cmp<mode>3_ltint): Ditto.
12364 (*avx2_pcmp<mode>3_3): Ditto.
12365 (*avx2_pcmp<mode>3_4): Ditto.
12366 (*avx2_pcmp<mode>3_5): Ditto.
12368 2021-06-28 liuhongt <hongtao.liu@intel.com>
12370 * config/i386/i386-builtin.def (IX86_BUILTIN_BLENDVPD256,
12371 IX86_BUILTIN_BLENDVPS256, IX86_BUILTIN_PBLENDVB256,
12372 IX86_BUILTIN_BLENDVPD, IX86_BUILTIN_BLENDVPS,
12373 IX86_BUILTIN_PBLENDVB128): Replace icode with
12375 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold blendv
12377 * config/i386/sse.md (*<sse4_1_avx2>_pblendvb_lt_subreg_not):
12378 New pre_reload splitter.
12380 2021-06-27 Andrew Pinski <apinski@marvell.com>
12382 PR middle-end/101230
12383 * fold-const.c (fold_ternary_loc): Check
12384 the return value of invert_tree_comparison.
12386 2021-06-27 David Edelsohn <dje.gcc@gmail.com>
12388 * config.gcc: Add SPDX License Identifier.
12389 (powerpc-ibm-aix789): Default to aix73.h.
12390 (powerpc-ibm-aix7.2.*.*): New stanza.
12391 * config/rs6000/aix72.h: Add SPDX License Identifier.
12392 * config/rs6000/aix73.h: New file.
12394 2021-06-26 Jason Merrill <jason@redhat.com>
12396 * except.c: #include "dwarf2.h" instead of "dwarf2out.h".
12398 2021-06-26 Andrew Pinski <apinski@marvell.com>
12400 * genmatch.c (lower_cond): Copy for_subst_vec
12401 for the simplify also.
12402 (lower): Swap the order for lower_for and lower_cond.
12404 2021-06-26 Andrew Pinski <apinski@marvell.com>
12406 * tree-ssa-phiopt.c (match_simplify_replacement): Reset
12407 flow senatitive info on the moved ssa set.
12409 2021-06-26 Andrew Pinski <apinski@marvell.com>
12411 * fold-const.c (fold_cond_expr_with_comparison):
12412 Exand arg0 into comp_code, arg00, and arg01.
12413 (fold_ternary_loc): Use invert_tree_comparison
12414 instead of fold_invert_truthvalue for the case
12415 where we have A CMP B ? C : A.
12417 2021-06-25 Martin Sebor <msebor@redhat.com>
12419 PR middle-end/101216
12420 * calls.c (maybe_warn_rdwr_sizes): Use the no_warning constant.
12422 2021-06-25 Jeff Law <jeffreyalaw@gmail.com>
12424 * config/h8300/h8300.c (select_cc_mode): Handle ASHIFTRT and LSHIFTRT.
12426 2021-06-25 Richard Biener <rguenther@suse.de>
12428 PR tree-optimization/101202
12429 * tree-vect-slp.c (vect_optimize_slp): Explicitely handle
12432 2021-06-25 Richard Biener <rguenther@suse.de>
12434 * tree-vect-slp-patterns.c (addsub_pattern::build): Copy
12435 STMT_VINFO_REDUC_DEF from the original representative.
12437 2021-06-25 Martin Sebor <msebor@redhat.com>
12439 * builtins.c (warn_string_no_nul): Replace uses of TREE_NO_WARNING,
12440 gimple_no_warning_p and gimple_set_no_warning with
12441 warning_suppressed_p, and suppress_warning.
12443 (maybe_warn_for_bound): Same.
12444 (warn_for_access): Same.
12445 (check_access): Same.
12446 (expand_builtin_strncmp): Same.
12447 (fold_builtin_varargs): Same.
12448 * calls.c (maybe_warn_nonstring_arg): Same.
12449 (maybe_warn_rdwr_sizes): Same.
12450 * cfgexpand.c (expand_call_stmt): Same.
12451 * cgraphunit.c (check_global_declaration): Same.
12452 * fold-const.c (fold_undefer_overflow_warnings): Same.
12453 (fold_truth_not_expr): Same.
12454 (fold_unary_loc): Same.
12455 (fold_checksum_tree): Same.
12456 * gimple-array-bounds.cc (array_bounds_checker::check_array_ref): Same.
12457 (array_bounds_checker::check_mem_ref): Same.
12458 (array_bounds_checker::check_addr_expr): Same.
12459 (array_bounds_checker::check_array_bounds): Same.
12460 * gimple-expr.c (copy_var_decl): Same.
12461 * gimple-fold.c (gimple_fold_builtin_strcpy): Same.
12462 (gimple_fold_builtin_strncat): Same.
12463 (gimple_fold_builtin_stxcpy_chk): Same.
12464 (gimple_fold_builtin_stpcpy): Same.
12465 (gimple_fold_builtin_sprintf): Same.
12466 (fold_stmt_1): Same.
12467 * gimple-ssa-isolate-paths.c (diag_returned_locals): Same.
12468 * gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Same.
12469 * gimple-ssa-sprintf.c (handle_printf_call): Same.
12470 * gimple-ssa-store-merging.c (imm_store_chain_info::output_merged_store): Same.
12471 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
12472 * gimple-ssa-warn-restrict.h: Adjust declarations.
12473 (maybe_diag_access_bounds): Replace uses of TREE_NO_WARNING,
12474 gimple_no_warning_p and gimple_set_no_warning with
12475 warning_suppressed_p, and suppress_warning.
12476 (check_call): Same.
12477 (check_bounds_or_overlap): Same.
12478 * gimple.c (gimple_build_call_from_tree): Same.
12479 * gimplify.c (gimplify_return_expr): Same.
12480 (gimplify_cond_expr): Same.
12481 (gimplify_modify_expr_complex_part): Same.
12482 (gimplify_modify_expr): Same.
12483 (gimple_push_cleanup): Same.
12484 (gimplify_expr): Same.
12485 * omp-expand.c (expand_omp_for_generic): Same.
12486 (expand_omp_taskloop_for_outer): Same.
12487 * omp-low.c (lower_rec_input_clauses): Same.
12488 (lower_lastprivate_clauses): Same.
12489 (lower_send_clauses): Same.
12490 (lower_omp_target): Same.
12491 * tree-cfg.c (pass_warn_function_return::execute): Same.
12492 * tree-complex.c (create_one_component_var): Same.
12493 * tree-inline.c (remap_gimple_op_r): Same.
12494 (copy_tree_body_r): Same.
12495 (declare_return_variable): Same.
12496 (expand_call_inline): Same.
12497 * tree-nested.c (lookup_field_for_decl): Same.
12498 * tree-sra.c (create_access_replacement): Same.
12499 (generate_subtree_copies): Same.
12500 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Same.
12501 * tree-ssa-forwprop.c (combine_cond_expr_cond): Same.
12502 * tree-ssa-loop-ch.c (ch_base::copy_headers): Same.
12503 * tree-ssa-loop-im.c (execute_sm): Same.
12504 * tree-ssa-phiopt.c (cond_store_replacement): Same.
12505 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
12506 (handle_builtin_strcpy): Same.
12507 (maybe_diag_stxncpy_trunc): Same.
12508 (handle_builtin_stxncpy_strncat): Same.
12509 (handle_builtin_strcat): Same.
12510 * tree-ssa-uninit.c (get_no_uninit_warning): Same.
12511 (set_no_uninit_warning): Same.
12512 (uninit_undefined_value_p): Same.
12513 (warn_uninit): Same.
12514 (maybe_warn_operand): Same.
12515 * tree-vrp.c (compare_values_warnv): Same.
12516 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr): Same.
12517 (test_for_singularity): Same.
12518 * gimple.h (warning_suppressed_p): New function.
12519 (suppress_warning): Same.
12520 (copy_no_warning): Same.
12521 (gimple_set_block): Call gimple_set_location.
12522 (gimple_set_location): Call copy_warning.
12524 2021-06-25 Martin Sebor <msebor@redhat.com>
12526 * tree.h (warning_suppressed_at, copy_warning,
12527 warning_suppressed_p, suppress_warning): New functions.
12529 2021-06-25 Martin Sebor <msebor@redhat.com>
12531 * Makefile.in (OBJS-libcommon): Add diagnostic-spec.o.
12532 * gengtype.c (open_base_files): Add diagnostic-spec.h.
12533 * diagnostic-spec.c: New file.
12534 * diagnostic-spec.h: New file.
12535 * tree.h (no_warning, all_warnings, suppress_warning_at): New
12537 * warning-control.cc: New file.
12539 2021-06-25 liuhongt <hongtao.liu@intel.com>
12542 * config/i386/i386.c (x86_order_regs_for_local_alloc):
12545 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
12547 PR tree-optimization/101189
12548 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Pass
12549 LHS range of condition to postfold routine.
12550 (fold_using_range::postfold_gcond_edges): Only process the TRUE or
12551 FALSE edge if the LHS range supports it being taken.
12552 * gimple-range-fold.h (postfold_gcond_edges): Add range parameter.
12554 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
12556 * value-relation.cc (equiv_oracle::dump): Do not dump NULL blocks.
12557 (relation_oracle::find_relation_block): Check correct bitmap.
12558 (relation_oracle::dump): Do not dump NULL blocks.
12560 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
12562 * gimple-range-cache.cc (ranger_cache::propagate_cache): Call
12563 range_on_edge instead of manually calculating.
12565 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
12567 * range-op.cc: Fix comment.
12569 2021-06-24 Uroš Bizjak <ubizjak@gmail.com>
12572 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
12573 Handle V8QI and V4HI modes.
12574 * config/i386/mmx.md (sse4_1_<any_extend:code>v4qiv4hi2):
12576 (sse4_1_<any_extend:code>v4qiv4hi2): Ditto.
12577 (mmxpackmode): New mode attribute.
12578 (vec_pack_trunc_<mmxpackmode:mode>): New expander.
12579 (mmxunpackmode): New mode attribute.
12580 (vec_unpacks_lo_<mmxunpackmode:mode>): New expander.
12581 (vec_unpacks_hi_<mmxunpackmode:mode>): Ditto.
12582 (vec_unpacku_lo_<mmxunpackmode:mode>): Ditto.
12583 (vec_unpacku_hi_<mmxunpackmode:mode>): Ditto.
12584 * config/i386/i386.md (extsuffix): Move from ...
12585 * config/i386/sse.md: ... here.
12587 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
12589 * dwarf2out.c (dwarf2out_assembly_start): Emit .file 0 marker here..
12590 (dwarf2out_finish): ...instead of here.
12592 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
12594 * configure.ac (--gdwarf-5 option): Use objdump instead of readelf.
12595 (working --gdwarf-4/--gdwarf-5 for all sources): Likewise.
12596 (--gdwarf-4 not refusing generated .debug_line): Adjust for Windows.
12597 * configure: Regenerate.
12599 2021-06-24 Richard Biener <rguenther@suse.de>
12601 * config/i386/sse.md (vec_addsubv4df3, vec_addsubv2df3,
12602 vec_addsubv8sf3, vec_addsubv4sf3): Merge into ...
12603 (vec_addsub<mode>3): ... using a new addsub_cst mode attribute.
12605 2021-06-24 Richard Biener <rguenther@suse.de>
12607 * config/i386/sse.md (avx_addsubv4df3): Rename to
12609 (avx_addsubv8sf3): Rename to vec_addsubv8sf3.
12610 (sse3_addsubv2df3): Rename to vec_addsubv2df3.
12611 (sse3_addsubv4sf3): Rename to vec_addsubv4sf3.
12612 * config/i386/i386-builtin.def: Adjust.
12613 * internal-fn.def (VEC_ADDSUB): New internal optab fn.
12614 * optabs.def (vec_addsub_optab): New optab.
12615 * tree-vect-slp-patterns.c (class addsub_pattern): New.
12616 (slp_patterns): Add addsub_pattern.
12617 * tree-vect-slp.c (vect_optimize_slp): Disable propagation
12618 across CFN_VEC_ADDSUB.
12619 * tree-vectorizer.h (vect_pattern::vect_pattern): Make
12621 * doc/md.texi (vec_addsub<mode>3): Document.
12623 2021-06-24 Jakub Jelinek <jakub@redhat.com>
12625 PR middle-end/101170
12626 * df-scan.c (df_ref_record): For paradoxical big-endian SUBREGs
12627 where regno + subreg_regno_offset wraps around use 0 as starting
12630 2021-06-24 Jakub Jelinek <jakub@redhat.com>
12632 PR middle-end/101172
12633 * stor-layout.c (finish_bitfield_representative): If nextf has
12634 error_mark_node type, set repr type to error_mark_node too.
12636 2021-06-24 Ilya Leoshkevich <iii@linux.ibm.com>
12638 * config/s390/s390.c (s390_function_profiler): Ignore labelno
12640 * config/s390/s390.h (NO_PROFILE_COUNTERS): Define.
12642 2021-06-24 Richard Biener <rguenther@suse.de>
12644 * tree-vect-slp.c (vect_optimize_slp): Do not propagate
12645 across operations that have different semantics on different
12648 2021-06-24 Jakub Jelinek <jakub@redhat.com>
12650 * tree.h (OMP_CLAUSE_MAP_IN_REDUCTION): Document meaning for OpenMP.
12651 * gimplify.c (gimplify_scan_omp_clauses): For OpenMP map clauses
12652 with OMP_CLAUSE_MAP_IN_REDUCTION flag partially defer gimplification
12653 of non-decl OMP_CLAUSE_DECL. For OMP_CLAUSE_IN_REDUCTION on
12654 OMP_TARGET user outer_ctx instead of ctx for placeholders and
12655 initializer/combiner gimplification.
12656 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_MAP_IN_REDUCTION
12657 on target constructs.
12658 (lower_rec_input_clauses): Likewise.
12659 (lower_omp_target): Likewise.
12660 * omp-expand.c (expand_omp_target): Temporarily ignore nowait clause
12661 on target if in_reduction is present.
12663 2021-06-24 Kewen Lin <linkw@linux.ibm.com>
12665 * tree-predcom.c (class pcom_worker): New class.
12666 (release_chain): Renamed to...
12667 (pcom_worker::release_chain): ...this.
12668 (release_chains): Renamed to...
12669 (pcom_worker::release_chains): ...this.
12670 (aff_combination_dr_offset): Renamed to...
12671 (pcom_worker::aff_combination_dr_offset): ...this.
12672 (determine_offset): Renamed to...
12673 (pcom_worker::determine_offset): ...this.
12674 (class comp_ptrs): New class.
12675 (split_data_refs_to_components): Renamed to...
12676 (pcom_worker::split_data_refs_to_components): ...this,
12677 and update with class comp_ptrs.
12678 (suitable_component_p): Renamed to...
12679 (pcom_worker::suitable_component_p): ...this.
12680 (filter_suitable_components): Renamed to...
12681 (pcom_worker::filter_suitable_components): ...this.
12682 (valid_initializer_p): Renamed to...
12683 (pcom_worker::valid_initializer_p): ...this.
12684 (find_looparound_phi): Renamed to...
12685 (pcom_worker::find_looparound_phi): ...this.
12686 (add_looparound_copies): Renamed to...
12687 (pcom_worker::add_looparound_copies): ...this.
12688 (determine_roots_comp): Renamed to...
12689 (pcom_worker::determine_roots_comp): ...this.
12690 (determine_roots): Renamed to...
12691 (pcom_worker::determine_roots): ...this.
12692 (single_nonlooparound_use): Renamed to...
12693 (pcom_worker::single_nonlooparound_use): ...this.
12694 (remove_stmt): Renamed to...
12695 (pcom_worker::remove_stmt): ...this.
12696 (execute_pred_commoning_chain): Renamed to...
12697 (pcom_worker::execute_pred_commoning_chain): ...this.
12698 (execute_pred_commoning): Renamed to...
12699 (pcom_worker::execute_pred_commoning): ...this.
12700 (struct epcc_data): New member worker.
12701 (execute_pred_commoning_cbck): Call execute_pred_commoning
12702 with pcom_worker pointer.
12703 (find_use_stmt): Renamed to...
12704 (pcom_worker::find_use_stmt): ...this.
12705 (find_associative_operation_root): Renamed to...
12706 (pcom_worker::find_associative_operation_root): ...this.
12707 (find_common_use_stmt): Renamed to...
12708 (pcom_worker::find_common_use_stmt): ...this.
12709 (combinable_refs_p): Renamed to...
12710 (pcom_worker::combinable_refs_p): ...this.
12711 (reassociate_to_the_same_stmt): Renamed to...
12712 (pcom_worker::reassociate_to_the_same_stmt): ...this.
12713 (stmt_combining_refs): Renamed to...
12714 (pcom_worker::stmt_combining_refs): ...this.
12715 (combine_chains): Renamed to...
12716 (pcom_worker::combine_chains): ...this.
12717 (try_combine_chains): Renamed to...
12718 (pcom_worker::try_combine_chains): ...this.
12719 (prepare_initializers_chain): Renamed to...
12720 (pcom_worker::prepare_initializers_chain): ...this.
12721 (prepare_initializers): Renamed to...
12722 (pcom_worker::prepare_initializers): ...this.
12723 (prepare_finalizers_chain): Renamed to...
12724 (pcom_worker::prepare_finalizers_chain): ...this.
12725 (prepare_finalizers): Renamed to...
12726 (pcom_worker::prepare_finalizers): ...this.
12727 (tree_predictive_commoning_loop): Renamed to...
12728 (pcom_worker::tree_predictive_commoning_loop): ...this, adjust
12729 some calls and remove some cleanup code.
12730 (tree_predictive_commoning): Adjusted to use pcom_worker instance.
12731 (static variable looparound_phis): Remove.
12732 (static variable name_expansions): Remove.
12734 2021-06-24 Richard Biener <rguenther@suse.de>
12736 * tree-vect-slp.c (slpg_vertex): New struct.
12737 (vect_slp_build_vertices): Adjust.
12738 (vect_optimize_slp): Likewise. Maintain an outgoing permute
12739 and a materialized one.
12741 2021-06-24 Richard Biener <rguenther@suse.de>
12743 PR tree-optimization/101105
12744 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
12745 Only ignore steps when they are equal or scalar order is preserved.
12747 2021-06-24 liuhongt <hongtao.liu@intel.com>
12750 * config/i386/i386-expand.c (ix86_expand_vec_interleave):
12751 Adjust comments for ix86_expand_vecop_qihi2.
12752 (ix86_expand_vecmul_qihi): Renamed to ..
12753 (ix86_expand_vecop_qihi2): Adjust function prototype to
12754 support shift operation, add static to definition.
12755 (ix86_expand_vec_shift_qihi_constant): Add static to definition.
12756 (ix86_expand_vecop_qihi): Call ix86_expand_vecop_qihi2 and
12757 ix86_expand_vec_shift_qihi_constant.
12758 * config/i386/i386-protos.h (ix86_expand_vecmul_qihi): Deleted.
12759 (ix86_expand_vec_shift_qihi_constant): Deleted.
12760 * config/i386/sse.md (VI12_256_512_AVX512VL): New mode
12762 (mulv8qi3): Call ix86_expand_vecop_qihi directly, add
12763 condition TARGET_64BIT.
12764 (mul<mode>3): Ditto.
12765 (<insn><mode>3): Ditto.
12766 (vlshr<mode>3): Extend to support avx512 vlshr.
12767 (v<insn><mode>3): New expander for
12769 (v<insn>v8qi3): Ditto.
12770 (vashrv8hi3<mask_name>): Renamed to ..
12771 (vashr<mode>3): And extend to support V16QImode for avx512.
12772 (vashrv16qi3): Deleted.
12773 (vashrv2di3<mask_name>): Extend expander to support avx512
12776 2021-06-23 Dimitar Dimitrov <dimitar@dinux.eu>
12778 * doc/lto.texi (Design Overview): Update that slim objects are
12781 2021-06-23 Aaron Sawdey <acsawdey@linux.ibm.com>
12783 * config/rs6000/rs6000-cpus.def: Take OPTION_MASK_PCREL_OPT out
12784 of OTHER_POWER10_MASKS so it will not be enabled by default.
12786 2021-06-23 Richard Biener <rguenther@suse.de>
12787 Martin Jambor <mjambor@suse.cz>
12789 * tree-inline.c (setup_one_parameter): Set TREE_READONLY of the
12790 param replacement unconditionally. Adjust comment.
12792 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
12794 * Makefile.in (OBJS): Add gimple-range-fold.o
12795 * gimple-range-fold.cc: New.
12796 * gimple-range-fold.h: New.
12797 * gimple-range-gori.cc (gimple_range_calc_op1): Move to here.
12798 (gimple_range_calc_op2): Ditto.
12799 * gimple-range-gori.h: Move prototypes to here.
12800 * gimple-range.cc: Adjust include files.
12801 (fur_source:fur_source): Relocate to gimple-range-fold.cc.
12802 (fur_source::get_operand): Ditto.
12803 (fur_source::get_phi_operand): Ditto.
12804 (fur_source::query_relation): Ditto.
12805 (fur_source::register_relation): Ditto.
12806 (class fur_edge): Ditto.
12807 (fur_edge::fur_edge): Ditto.
12808 (fur_edge::get_operand): Ditto.
12809 (fur_edge::get_phi_operand): Ditto.
12810 (fur_stmt::fur_stmt): Ditto.
12811 (fur_stmt::get_operand): Ditto.
12812 (fur_stmt::get_phi_operand): Ditto.
12813 (fur_stmt::query_relation): Ditto.
12814 (class fur_depend): Relocate to gimple-range-fold.h.
12815 (fur_depend::fur_depend): Relocate to gimple-range-fold.cc.
12816 (fur_depend::register_relation): Ditto.
12817 (fur_depend::register_relation): Ditto.
12818 (class fur_list): Ditto.
12819 (fur_list::fur_list): Ditto.
12820 (fur_list::get_operand): Ditto.
12821 (fur_list::get_phi_operand): Ditto.
12822 (fold_range): Ditto.
12823 (adjust_pointer_diff_expr): Ditto.
12824 (gimple_range_adjustment): Ditto.
12825 (gimple_range_base_of_assignment): Ditto.
12826 (gimple_range_operand1): Ditto.
12827 (gimple_range_operand2): Ditto.
12828 (gimple_range_calc_op1): Relocate to gimple-range-gori.cc.
12829 (gimple_range_calc_op2): Ditto.
12830 (fold_using_range::fold_stmt): Relocate to gimple-range-fold.cc.
12831 (fold_using_range::range_of_range_op): Ditto.
12832 (fold_using_range::range_of_address): Ditto.
12833 (fold_using_range::range_of_phi): Ditto.
12834 (fold_using_range::range_of_call): Ditto.
12835 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
12836 (fold_using_range::range_of_builtin_call): Ditto.
12837 (fold_using_range::range_of_cond_expr): Ditto.
12838 (fold_using_range::range_of_ssa_name_with_loop_info): Ditto.
12839 (fold_using_range::relation_fold_and_or): Ditto.
12840 (fold_using_range::postfold_gcond_edges): Ditto.
12841 * gimple-range.h: Add gimple-range-fold.h to include files. Change
12842 GIMPLE_RANGE_STMT_H to GIMPLE_RANGE_H.
12843 (gimple_range_handler): Relocate to gimple-range-fold.h.
12844 (gimple_range_ssa_p): Ditto.
12845 (range_compatible_p): Ditto.
12846 (class fur_source): Ditto.
12847 (class fur_stmt): Ditto.
12848 (class fold_using_range): Ditto.
12849 (gimple_range_calc_op1): Relocate to gimple-range-gori.h
12850 (gimple_range_calc_op2): Ditto.
12852 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
12854 PR tree-optimization/101148
12855 PR tree-optimization/101014
12856 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
12857 (ranger_cache::~ranger_cache): Adjust.
12858 (ranger_cache::block_range): Check if propagation disallowed.
12859 (ranger_cache::propagate_cache): Disallow propagation if new value
12860 can't be stored properly.
12861 * gimple-range-cache.h (ranger_cache::m_propfail): New member.
12863 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
12865 * gimple-range-cache.cc (class ssa_block_ranges): Adjust prototype.
12866 (sbr_vector::set_bb_range): Return true.
12867 (class sbr_sparse_bitmap): Adjust.
12868 (sbr_sparse_bitmap::set_bb_range): Return value.
12869 (block_range_cache::set_bb_range): Return value.
12870 (ranger_cache::propagate_cache): Use return value to print msg.
12871 * gimple-range-cache.h (class block_range_cache): Adjust.
12873 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
12875 * gimple-range.cc (dump_bb): Use range_on_edge from the cache.
12877 2021-06-23 Jeff Law <jeffreyalaw@gmail.com>
12879 * config/h8300/logical.md (<code><mode>3<ccnz>): Use <cczn>
12880 so this pattern can be used for test/compare removal. Pass
12881 current insn to compute_logical_op_length and output_logical_op.
12882 * config/h8300/h8300.c (compute_logical_op_cc): Remove.
12883 (h8300_and_costs): Add argument to compute_logical_op_length.
12884 (output_logical_op): Add new argument. Use it to determine if the
12885 condition codes are used and adjust the output accordingly.
12886 (compute_logical_op_length): Add new argument and update length
12887 computations when condition codes are used.
12888 * config/h8300/h8300-protos.h (compute_logical_op_length): Update
12890 (output_logical_op): Likewise.
12892 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
12895 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
12896 Handle 64bit modes for TARGET_XOP. Use indirect gen_* functions.
12897 * config/i386/mmx.md (mmx_ppermv64): New insn pattern.
12898 * config/i386/i386.md (unspec): Move UNSPEC_XOP_PERMUTE from ...
12899 * config/i386/sse.md (unspec): ... here.
12901 2021-06-23 Martin Liska <mliska@suse.cz>
12904 * optc-save-gen.awk: Put back arm_fp16_format to
12907 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
12910 * config/i386/i386.md (bsr_rex64): Add zero-flag setting RTX.
12913 (clz<mode>2): Update RTX pattern for additions.
12915 2021-06-23 Jakub Jelinek <jakub@redhat.com>
12917 PR middle-end/101167
12918 * omp-low.c (lower_omp_regimplify_p): Regimplify also PARM_DECLs
12919 and RESULT_DECLs that have DECL_HAS_VALUE_EXPR_P set.
12921 2021-06-22 Sergei Trofimovich <siarheit@google.com>
12923 * doc/rtl.texi: drop unbalanced parenthesis.
12925 2021-06-22 Richard Biener <rguenther@suse.de>
12927 PR middle-end/101156
12928 * gimplify.c (gimplify_expr): Remove premature incorrect
12931 2021-06-22 Jakub Jelinek <jakub@redhat.com>
12933 PR tree-optimization/101159
12934 * tree-vect-patterns.c (vect_recog_popcount_pattern): Fix some
12937 2021-06-22 Jakub Jelinek <jakub@redhat.com>
12939 PR middle-end/101160
12940 * function.c (assign_parms): For decl_result with TYPE_EMPTY_P type
12941 clear crtl->return_rtx instead of keeping it referencing a pseudo.
12943 2021-06-22 Jakub Jelinek <jakub@redhat.com>
12944 Andrew Pinski <apinski@marvell.com>
12946 PR tree-optimization/101162
12947 * fold-const.c (range_check_type): Handle OFFSET_TYPE like pointer
12950 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12952 * range-op.cc (range_relational_tests): New.
12953 (range_op_tests): Call range_relational_tests.
12955 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12957 * range-op.cc (operator_cast::lhs_op1_relation): New.
12958 (operator_identity::lhs_op1_relation): Mew.
12960 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12962 * range-op.cc (operator_minus::op1_op2_relation_effect): New.
12964 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12966 * range-op.cc (operator_plus::lhs_op1_relation): New.
12967 (operator_plus::lhs_op2_relation): New.
12969 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12971 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a
12972 relation_oracle if dominators exist.
12973 (ranger_cache::~ranger_cache): Dispose of oracle.
12974 (ranger_cache::dump_bb): Dump oracle.
12975 * gimple-range.cc (fur_source::fur_source): New.
12976 (fur_source::get_operand): Use mmeber query.
12977 (fur_source::get_phi_operand): Use member_query.
12978 (fur_source::query_relation): New.
12979 (fur_source::register_dependency): Delete.
12980 (fur_source::register_relation): New.
12981 (fur_edge::fur_edge): Adjust.
12982 (fur_edge::get_phi_operand): Fix comment.
12983 (fur_edge::query): Delete.
12984 (fur_stmt::fur_stmt): Adjust.
12985 (fur_stmt::query): Delete.
12986 (fur_depend::fur_depend): Adjust.
12987 (fur_depend::register_relation): New.
12988 (fur_depend::register_relation): New.
12989 (fur_list::fur_list): Adjust.
12990 (fur_list::get_operand): Use member query.
12991 (fold_using_range::range_of_range_op): Process and query relations.
12992 (fold_using_range::range_of_address): Adjust dependency call.
12993 (fold_using_range::range_of_phi): Ditto.
12994 (gimple_ranger::gimple_ranger): New. Use ranger_ache oracle.
12995 (fold_using_range::relation_fold_and_or): New.
12996 (fold_using_range::postfold_gcond_edges): New.
12997 * gimple-range.h (class gimple_ranger): Adjust.
12998 (class fur_source): Adjust members.
12999 (class fur_stmt): Ditto.
13000 (class fold_using_range): Ditto.
13002 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
13004 * range-op.cc (range_operator::wi_fold): Apply relation effect.
13005 (range_operator::fold_range): Adjust and apply relation effect.
13006 (*::fold_range): Add relation parameters.
13007 (*::op1_range): Ditto.
13008 (*::op2_range): Ditto.
13009 (range_operator::lhs_op1_relation): New.
13010 (range_operator::lhs_op2_relation): New.
13011 (range_operator::op1_op2_relation): New.
13012 (range_operator::op1_op2_relation_effect): New.
13013 (relop_early_resolve): New.
13014 (operator_equal::op1_op2_relation): New.
13015 (operator_equal::fold_range): Call relop_early_resolve.
13016 (operator_not_equal::op1_op2_relation): New.
13017 (operator_not_equal::fold_range): Call relop_early_resolve.
13018 (operator_lt::op1_op2_relation): New.
13019 (operator_lt::fold_range): Call relop_early_resolve.
13020 (operator_le::op1_op2_relation): New.
13021 (operator_le::fold_range): Call relop_early_resolve.
13022 (operator_gt::op1_op2_relation): New.
13023 (operator_gt::fold_range): Call relop_early_resolve.
13024 (operator_ge::op1_op2_relation): New.
13025 (operator_ge::fold_range): Call relop_early_resolve.
13026 * range-op.h (class range_operator): Adjust parameters and methods.
13028 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
13030 * Makefile.in (OBJS): Add value-relation.o.
13031 * gimple-range.h: Adjust include files.
13032 * tree-data-ref.c: Adjust include file order.
13033 * value-query.cc (range_query::get_value_range): Default to no oracle.
13034 (range_query::query_relation): New.
13035 (range_query::query_relation): New.
13036 * value-query.h (class range_query): Adjust.
13037 * value-relation.cc: New.
13038 * value-relation.h: New.
13040 2021-06-22 Richard Biener <rguenther@suse.de>
13042 PR tree-optimization/101151
13043 * tree-ssa-sink.c (statement_sink_location): Expand irreducible
13046 2021-06-22 Jojo R <rjiejie@linux.alibaba.com>
13048 * config/riscv/riscv.c (thead_c906_tune_info): New.
13049 (riscv_tune_info_table): Use new tune.
13051 2021-06-22 Richard Biener <rguenther@suse.de>
13053 PR tree-optimization/101158
13054 * tree-vect-slp.c (vect_build_slp_tree_1): Move same operand
13055 checking after checking for matching operation.
13057 2021-06-22 Richard Biener <rguenther@suse.de>
13059 PR tree-optimization/101159
13060 * tree-vect-patterns.c (vect_recog_popcount_pattern): Add
13061 missing NULL vectype check.
13063 2021-06-22 Richard Biener <rguenther@suse.de>
13065 PR tree-optimization/101154
13066 * tree-vect-slp.c (vect_build_slp_tree_2): Fix out-of-bound access.
13068 2021-06-22 Jakub Jelinek <jakub@redhat.com>
13071 * config/i386/i386-protos.h (ix86_last_zero_store_uid): Declare.
13072 * config/i386/i386-expand.c (ix86_last_zero_store_uid): New variable.
13073 * config/i386/i386.c (ix86_expand_prologue): Clear it.
13074 * config/i386/i386.md (peephole2s for 1/2/4 stores of const0_rtx):
13075 Remove "" from match_operand. Emit new insns using emit_move_insn and
13076 set ix86_last_zero_store_uid to INSN_UID of the last store.
13077 Add peephole2s for 1/2/4 stores of const0_rtx following previous
13080 2021-06-22 Martin Liska <mliska@suse.cz>
13082 * auto-profile.c (AUTO_PROFILE_VERSION): Bump as string format
13085 2021-06-22 Martin Liska <mliska@suse.cz>
13087 * gcov-io.h: Remove padding entries.
13089 2021-06-22 liuhongt <hongtao.liu@intel.com>
13091 PR tree-optimization/97770
13092 * tree-vect-patterns.c (vect_recog_popcount_pattern):
13094 (vect_recog_func vect_vect_recog_func_ptrs): Add new pattern.
13096 2021-06-22 liuhongt <hongtao.liu@intel.com>
13099 * config/i386/i386-builtin.def (BDESC): Adjust builtin name.
13100 * config/i386/sse.md (<avx512>_expand<mode>_mask): Rename to ..
13101 (expand<mode>_mask): this ..
13102 (*expand<mode>_mask): New pre_reload splitter to transform
13103 v{,p}expand* to vmov* when mask is zero, all ones, or has all
13104 ones in it's lower part, otherwise still generate
13107 2021-06-22 liuhongt <hongtao.liu@intel.com>
13110 * config/i386/i386-expand.c
13111 (ix86_expand_special_args_builtin): Keep constm1_operand only
13112 if it satisfies insn's operand predicate.
13114 2021-06-21 Jason Merrill <jason@redhat.com>
13117 * df-scan.c (df_ref_record): Check that regno < endregno.
13118 * function.c (assign_parms, expand_function_end): Do nothing with a
13119 TYPE_EMPTY_P result.
13121 2021-06-21 Richard Biener <rguenther@suse.de>
13123 PR tree-optimization/101120
13124 * tree-vect-data-refs.c (bump_vector_ptr): Fold the
13126 * tree-vect-slp.c (vect_transform_slp_perm_load): Add
13127 DR chain DCE capability.
13128 * tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
13129 * tree-vect-stmts.c (vectorizable_load): Remove unused
13130 loads in the DR chain for SLP.
13132 2021-06-21 Jakub Jelinek <jakub@redhat.com>
13134 PR inline-asm/100785
13135 * gimplify.c (gimplify_asm_expr): Don't diagnose errors if
13136 output or input operands were already error_mark_node.
13137 * cfgexpand.c (expand_asm_stmt): If errors are emitted,
13138 remove all inputs, outputs and clobbers from the asm and
13139 set template to "".
13141 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
13143 * config/arm/arm_neon.h (vceq_s8): Replace builtin with __a == __b.
13144 (vceq_s16): Likewise.
13145 (vceq_s32): Likewise.
13146 (vceq_u8): Likewise.
13147 (vceq_u16): Likewise.
13148 (vceq_u32): Likewise.
13149 (vceq_p8): Likewise.
13150 (vceqq_s8): Likewise.
13151 (vceqq_s16): Likewise.
13152 (vceqq_s32): Likewise.
13153 (vceqq_u8): Likewise.
13154 (vceqq_u16): Likewise.
13155 (vceqq_u32): Likewise.
13156 (vceqq_p8): Likewise.
13157 (vceq_f32): Gate __a == __b on __FAST_MATH__.
13158 (vceqq_f32): Likewise.
13159 (vceq_f16): Likewise.
13160 (vceqq_f16): Likewise.
13162 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
13165 * config/arm/iterators.md (NEON_VACMP): Remove.
13166 * config/arm/neon.md (neon_vca<cmp_op><mode>): Use GLTE instead of GTGE
13168 (neon_vca<cmp_op><mode>_insn): Likewise.
13169 (neon_vca<cmp_op_unsp><mode>_insn_unspec): Use NEON_VAGLTE instead of
13172 2021-06-21 Richard Biener <rguenther@suse.de>
13174 PR tree-optimization/101121
13175 * tree-vect-slp.c (vect_build_slp_tree_2): To not fail fatally
13176 when we just lack a stmt with the desired op when doing permutation.
13177 (vect_build_slp_tree): When caching a failed SLP build attempt
13178 assert that at least one lane is marked as not matching.
13180 2021-06-21 liuhongt <hongtao.liu@intel.com>
13183 * config/i386/i386.md: (*anddi_1): Disparage slightly the mask
13184 register alternative.
13185 (*and<mode>_1): Ditto.
13187 (*andn<mode>_1): Ditto.
13188 (*<code><mode>_1): Ditto.
13189 (*<code>qi_1): Ditto.
13190 (*one_cmpl<mode>2_1): Ditto.
13191 (*one_cmplsi2_1_zext): Ditto.
13192 (*one_cmplqi2_1): Ditto.
13193 * config/i386/i386.c (x86_order_regs_for_local_alloc): Change
13194 the order of mask registers to be before general registers.
13196 2021-06-21 Roger Sayle <roger@nextmovesoftware.com>
13199 * config/i386/i386.md: New define_peephole2s to shrink writing
13200 1, 2 or 4 consecutive zeros to memory when optimizing for size.
13202 2021-06-18 Jeff Law <jeffreyalaw@gmail.com>
13204 * config/h8300/h8300.c (h8300_select_cc_mode): Handle SYMBOL_REF.
13205 * config/h8300/logical.md (<code><mode>3 logcial expander): Generate
13206 more efficient code when the source can be trivially simplified.
13208 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
13210 * gimple-range-cache.cc (ranger_cache::range_of_def): Calculate
13211 a range if global is not available.
13212 (ranger_cache::entry_range): Fallback to range_of_def.
13213 * gimple-range-cache.h (range_of_def): Adjust prototype.
13215 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
13217 PR tree-optimization/101014
13218 * gimple-range-cache.cc (ranger_cache::ranger_cache): Remove poor
13220 (ranger_cache::~ranger_cache): Ditto.
13221 (ranger_cache::enable_new_values): Delete.
13222 (ranger_cache::push_poor_value): Delete.
13223 (ranger_cache::range_of_def): Remove poor value processing.
13224 (ranger_cache::entry_range): Ditto.
13225 (ranger_cache::fill_block_cache): Ditto.
13226 * gimple-range-cache.h (class ranger_cache): Remove poor value members.
13227 * gimple-range.cc (gimple_ranger::range_of_expr): Remove call.
13228 * gimple-range.h (class gimple_ranger): Adjust.
13230 2021-06-18 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
13233 * common/config/arm/arm-common.c (arm_canon_arch_option_1): New function
13234 derived from arm_canon_arch.
13235 (arm_canon_arch_option): Call it.
13236 (arm_canon_arch_multilib_option): New function.
13237 * config/arm/arm-cpus.in (IGNORE_FOR_MULTILIB): New fgroup.
13238 * config/arm/arm.h (arm_canon_arch_multilib_option): New prototype.
13239 (CANON_ARCH_MULTILIB_SPEC_FUNCTION): New macro.
13240 (MULTILIB_ARCH_CANONICAL_SPECS): New macro.
13241 (DRIVER_SELF_SPECS): Add MULTILIB_ARCH_CANONICAL_SPECS.
13242 * config/arm/arm.opt (mlibarch): New option.
13243 * config/arm/t-rmprofile (MULTILIB_MATCHES): For armv8*-m, replace use
13244 of march on RHS with mlibarch.
13246 2021-06-18 Marcel Vollweiler <marcel@codesourcery.com>
13248 * config.in: Regenerate.
13249 * config/gcn/gcn.c (print_operand_address): Fix for global_load assembler
13251 * configure: Regenerate.
13252 * configure.ac: Fix for global_load assembler functions.
13254 2021-06-18 Richard Biener <rguenther@suse.de>
13256 PR tree-optimization/101112
13257 * tree-vect-slp.c (vect_slp_linearize_chain): Fix condition
13258 to lookup a pattern stmt def.
13260 2021-06-18 Jakub Jelinek <jakub@redhat.com>
13262 PR middle-end/101062
13263 * stor-layout.c (finish_bitfield_layout): Don't add bitfield
13264 representatives in QUAL_UNION_TYPE.
13266 2021-06-18 Andrew Pinski <apinski@marvell.com>
13268 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
13269 Add counting of how many times it is done.
13270 (factor_out_conditional_conversion): Likewise.
13271 (match_simplify_replacement): Likewise.
13272 (value_replacement): Likewise.
13273 (spaceship_replacement): Likewise.
13274 (cond_store_replacement): Likewise.
13275 (cond_if_else_store_replacement_1): Likewise.
13276 (hoist_adjacent_loads): Likewise.
13278 2021-06-18 Andrew Pinski <apinski@marvell.com>
13280 * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
13281 types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and CNONJ_EXPR.
13282 (verify_gimple_assign_binary): Reject point and offset types on
13283 MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
13284 FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
13285 FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.
13287 2021-06-18 Michael Meissner <meissner@linux.ibm.com>
13289 * config/rs6000/rs6000.c (rs6000_emit_minmax): Add support for ISA
13290 3.1 IEEE 128-bit floating point xsmaxcqp/xsmincqp instructions.
13291 * config/rs6000/rs6000.md (s<minmax><mode>3, IEEE128 iterator):
13294 2021-06-17 Aaron Sawdey <acsawdey@linux.ibm.com>
13296 * config/rs6000/genfusion.pl (gen_logical_addsubf): Add
13297 earlyclobber to alts 0/1.
13298 (gen_addadd): Add earlyclobber to alts 0/1.
13299 * config/rs6000/fusion.md: Regenerate file.
13301 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13303 * cfgloopanal.c (get_loop_hot_path): Make path an auto_vec.
13305 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
13307 * gimple-range-cache.cc: Comment cleanups.
13308 * gimple-range-gori.cc: Comment cleanups.
13309 * gimple-range.cc: Comment/spacing cleanups
13310 * value-range.h: Comment cleanups.
13312 2021-06-17 H.J. Lu <hjl.tools@gmail.com>
13315 * calls.c (expand_call): Replace PUSH_ARGS with
13316 targetm.calls.push_argument (0).
13317 (emit_library_call_value_1): Likewise.
13318 * defaults.h (PUSH_ARGS): Removed.
13319 (PUSH_ARGS_REVERSED): Replace PUSH_ARGS with
13320 targetm.calls.push_argument (0).
13321 * expr.c (block_move_libcall_safe_for_call_parm): Likewise.
13322 (emit_push_insn): Pass the number bytes to push to
13323 targetm.calls.push_argument and pass 0 if ARGS_ADDR is 0.
13324 * hooks.c (hook_bool_uint_true): New.
13325 * hooks.h (hook_bool_uint_true): Likewise.
13326 * rtlanal.c (nonzero_bits1): Replace PUSH_ARGS with
13327 targetm.calls.push_argument (0).
13328 * target.def (push_argument): Add a targetm.calls hook.
13329 * targhooks.c (default_push_argument): New.
13330 * targhooks.h (default_push_argument): Likewise.
13331 * config/bpf/bpf.h (PUSH_ARGS): Removed.
13332 * config/cr16/cr16.c (TARGET_PUSH_ARGUMENT): New.
13333 * config/cr16/cr16.h (PUSH_ARGS): Removed.
13334 * config/i386/i386.c (ix86_push_argument): New.
13335 (TARGET_PUSH_ARGUMENT): Likewise.
13336 * config/i386/i386.h (PUSH_ARGS): Removed.
13337 * config/m32c/m32c.c (TARGET_PUSH_ARGUMENT): New.
13338 * config/m32c/m32c.h (PUSH_ARGS): Removed.
13339 * config/nios2/nios2.h (PUSH_ARGS): Likewise.
13340 * config/pru/pru.h (PUSH_ARGS): Likewise.
13341 * doc/tm.texi.in: Remove PUSH_ARGS documentation. Add
13342 TARGET_PUSH_ARGUMENT hook.
13343 * doc/tm.texi: Regenerated.
13345 2021-06-17 Uroš Bizjak <ubizjak@gmail.com>
13348 * config/i386/i386-expand.c (expand_vector_set_var):
13349 Handle V2FS mode remapping. Pass TARGET_MMX_WITH_SSE to
13350 ix86_expand_vector_init_duplicate.
13351 (ix86_expand_vector_init_duplicate): Emit insv_1 for
13352 QImode for !TARGET_PARTIAL_REG_STALL.
13353 * config/i386/predicates.md (vec_setm_mmx_operand): New predicate.
13354 * config/i386/mmx.md (vec_setv2sf): Use vec_setm_mmx_operand
13355 as operand 2 predicate. Call ix86_expand_vector_set_var
13356 for non-constant index operand.
13357 (vec_setv2si): Ditto.
13358 (vec_setv4hi): Ditto.
13359 (vec_setv8qi): ditto.
13361 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
13363 PR tree-optimization/100790
13364 * gimple-range.cc (range_of_builtin_call): Cleanup clz and ctz
13367 2021-06-17 Martin Liska <mliska@suse.cz>
13369 * doc/invoke.texi: Use consistently -O1 instead of -O.
13371 2021-06-17 Martin Liska <mliska@suse.cz>
13373 * gcov-io.h: Update documentation entry about string format.
13375 2021-06-17 Marius Hillenbrand <mhillen@linux.ibm.com>
13378 * config/s390/vecintrin.h (vec_doublee): Fix to use
13379 __builtin_s390_vflls.
13380 (vec_floate): Fix to use __builtin_s390_vflrd.
13382 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13384 * dominance.c (get_dominated_to_depth): Return auto_vec<basic_block>.
13385 * dominance.h (get_dominated_to_depth): Likewise.
13386 (get_all_dominated_blocks): Likewise.
13387 * cfgcleanup.c (delete_unreachable_blocks): Adjust.
13388 * gcse.c (hoist_code): Likewise.
13389 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
13390 * tree-parloops.c (oacc_entry_exit_ok): Likewise.
13391 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Likewise.
13392 * tree-ssa-phiprop.c (pass_phiprop::execute): Likewise.
13394 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13396 * dominance.c (get_dominated_by_region): Return auto_vec<basic_block>.
13397 * dominance.h (get_dominated_by_region): Likewise.
13398 * tree-cfg.c (gimple_duplicate_sese_region): Adjust.
13399 (gimple_duplicate_sese_tail): Likewise.
13400 (move_sese_region_to_fn): Likewise.
13402 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13404 * dominance.c (get_dominated_by): Return auto_vec<basic_block>.
13405 * dominance.h (get_dominated_by): Likewise.
13406 * auto-profile.c (afdo_find_equiv_class): Adjust.
13407 * cfgloopmanip.c (duplicate_loop_to_header_edge): Likewise.
13408 * loop-unroll.c (unroll_loop_runtime_iterations): Likewise.
13409 * tree-cfg.c (test_linear_chain): Likewise.
13410 (test_diamond): Likewise.
13412 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13414 * cfgloop.h (get_loop_hot_path): Return auto_vec<basic_block>.
13415 * cfgloopanal.c (get_loop_hot_path): Likewise.
13416 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
13418 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13420 * cgraph.c (cgraph_node::collect_callers): Return
13421 auto_vec<cgraph_edge *>.
13422 * cgraph.h (cgraph_node::collect_callers): Likewise.
13423 * ipa-cp.c (create_specialized_node): Adjust.
13424 (decide_about_value): Likewise.
13425 (decide_whether_version_node): Likewise.
13426 * ipa-sra.c (process_isra_node_results): Likewise.
13428 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
13430 * vec.h (vl_ptr>::using_auto_storage): Handle null m_vec.
13431 (auto_vec<T, 0>::auto_vec): Define move constructor, and delete copy
13433 (auto_vec<T, 0>::operator=): Define move assignment and delete copy
13436 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
13438 * gimple-range.cc (debug_seed_ranger): New.
13439 (dump_ranger): New.
13440 (debug_ranger): New.
13442 2021-06-17 Richard Biener <rguenther@suse.de>
13444 PR tree-optimization/54400
13445 * tree-vectorizer.h (enum slp_instance_kind): Add
13446 slp_inst_kind_bb_reduc.
13447 (reduction_fn_for_scalar_code): Declare.
13448 * tree-vect-data-refs.c (vect_slp_analyze_instance_dependence):
13449 Check SLP_INSTANCE_KIND instead of looking at the
13451 (vect_slp_analyze_instance_alignment): Likewise.
13452 * tree-vect-loop.c (reduction_fn_for_scalar_code): Export.
13453 * tree-vect-slp.c (vect_slp_linearize_chain): Split out
13454 chain linearization from vect_build_slp_tree_2 and generalize
13455 for the use of BB reduction vectorization.
13456 (vect_build_slp_tree_2): Adjust accordingly.
13457 (vect_optimize_slp): Elide permutes at the root of BB reduction
13459 (vectorizable_bb_reduc_epilogue): New function.
13460 (vect_slp_prune_covered_roots): Likewise.
13461 (vect_slp_analyze_operations): Use them.
13462 (vect_slp_check_for_constructors): Recognize associatable
13463 chains for BB reduction vectorization.
13464 (vectorize_slp_instance_root_stmt): Generate code for the
13465 BB reduction epilogue.
13467 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
13469 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Check with
13471 (gori_compute::may_recompute_p): New.
13472 (gori_compute::outgoing_edge_range_p): Perform recomputations.
13473 * gimple-range-gori.h (class gori_compute): Add prototype.
13475 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
13477 * gimple-range-cache.cc (ranger_cache::range_on_edge): Always return
13478 true when a range can be calculated.
13479 * gimple-range.cc (gimple_ranger::dump_bb): Check has_edge_range_p.
13481 2021-06-16 Martin Sebor <msebor@redhat.com>
13483 * doc/invoke.texi (-Wmismatched-dealloc, -Wmismatched-new-delete):
13484 Correct documented defaults.
13486 2021-06-16 Andrew MacLeod <amacleod@redhat.com>
13488 * gimple-range-cache.cc (ranger_cache::ranger_cache): Initialize
13489 m_new_value_p directly.
13491 2021-06-16 Uroš Bizjak <ubizjak@gmail.com>
13494 * config/i386/i386-expand.c (expand_vec_perm_2perm_pblendv):
13495 Handle 64bit modes for TARGET_SSE4_1.
13496 (expand_vec_perm_pshufb2): Handle 64bit modes for TARGET_SSSE3.
13497 (expand_vec_perm_even_odd_pack): Handle V4HI mode.
13498 (expand_vec_perm_even_odd_1) <case E_V4HImode>: Expand via
13499 expand_vec_perm_pshufb2 for TARGET_SSSE3 and via
13500 expand_vec_perm_even_odd_pack for TARGET_SSE4_1.
13501 * config/i386/mmx.md (mmx_packusdw): New insn pattern.
13503 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
13505 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn<mode>):
13506 Change to an expander that emits the correct instruction
13507 depending on endianness.
13508 (aarch64_<sur><addsub>hn<mode>_insn_le): Define.
13509 (aarch64_<sur><addsub>hn<mode>_insn_be): Define.
13511 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
13513 * config/aarch64/aarch64-simd-builtins.def: Split generator
13514 for aarch64_<su>qmovn builtins into scalar and vector
13516 * config/aarch64/aarch64-simd.md (aarch64_<su>qmovn<mode>_insn_le):
13518 (aarch64_<su>qmovn<mode>_insn_be): Define.
13519 (aarch64_<su>qmovn<mode>): Split into scalar and vector
13520 variants. Change vector variant to an expander that emits the
13521 correct instruction depending on endianness.
13523 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
13525 * config/aarch64/aarch64-simd-builtins.def: Split generator
13526 for aarch64_sqmovun builtins into scalar and vector variants.
13527 * config/aarch64/aarch64-simd.md (aarch64_sqmovun<mode>):
13528 Split into scalar and vector variants. Change vector variant
13529 to an expander that emits the correct instruction depending
13531 (aarch64_sqmovun<mode>_insn_le): Define.
13532 (aarch64_sqmovun<mode>_insn_be): Define.
13534 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
13536 * config/aarch64/aarch64-simd.md (aarch64_xtn<mode>_insn_le):
13537 Define - modeling zero-high-half semantics.
13538 (aarch64_xtn<mode>): Change to an expander that emits the
13539 appropriate instruction depending on endianness.
13540 (aarch64_xtn<mode>_insn_be): Define - modeling zero-high-half
13542 (aarch64_xtn2<mode>_le): Rename to...
13543 (aarch64_xtn2<mode>_insn_le): This.
13544 (aarch64_xtn2<mode>_be): Rename to...
13545 (aarch64_xtn2<mode>_insn_be): This.
13546 (vec_pack_trunc_<mode>): Emit truncation instruction instead
13548 * config/aarch64/iterators.md (Vnarrowd): Add Vnarrowd mode
13549 attribute iterator.
13551 2021-06-16 Martin Jambor <mjambor@suse.cz>
13553 PR tree-optimization/100453
13554 * tree-sra.c (create_access): Disqualify any const candidates
13555 which are written to.
13556 (sra_modify_expr): Do not store sub-replacements back to a const base.
13557 (handle_unscalarized_data_in_subtree): Likewise.
13558 (sra_modify_assign): Likewise. Earlier, use TREE_READONLy test
13559 instead of constant_decl_p.
13561 2021-06-16 Jakub Jelinek <jakub@redhat.com>
13563 PR middle-end/101062
13564 * stor-layout.c (finish_bitfield_representative): For fields in unions
13565 assume nextf is always NULL.
13566 (finish_bitfield_layout): Compute bit field representatives also in
13567 unions, but handle it as if each bitfield was the only field in the
13570 2021-06-16 Richard Biener <rguenther@suse.de>
13572 PR tree-optimization/101088
13573 * tree-ssa-loop-im.c (sm_seq_valid_bb): Only look for
13574 supported refs on edges. Do not assert same ref but
13575 different kind stores are unsuported but mark them so.
13576 (hoist_memory_references): Only look for supported refs
13579 2021-06-16 Roger Sayle <roger@nextmovesoftware.com>
13581 PR rtl-optimization/46235
13582 * config/i386/i386.md: New define_split for bt followed by cmov.
13583 (*bt<mode>_setcqi): New define_insn_and_split for bt followed by setc.
13584 (*bt<mode>_setncqi): New define_insn_and_split for bt then setnc.
13585 (*bt<mode>_setnc<mode>): New define_insn_and_split for bt followed
13586 by setnc with zero extension.
13588 2021-06-16 Richard Biener <rguenther@suse.de>
13590 PR tree-optimization/101083
13591 * tree-vect-slp.c (vect_slp_build_two_operator_nodes): Get
13592 vectype as argument.
13593 (vect_build_slp_tree_2): Adjust.
13595 2021-06-15 Martin Sebor <msebor@redhat.com>
13597 PR middle-end/100876
13598 * builtins.c: (gimple_call_return_array): Account for size_t
13599 mangling as either unsigned int or unsigned long
13601 2021-06-15 Jeff Law <jeffreyalaw@gmail.com>
13603 * compare-elim.c (try_eliminate_compare): Run DCE to clean things
13604 up before eliminating comparisons.
13606 2021-06-15 Aldy Hernandez <aldyh@redhat.com>
13608 * range-op.cc (operator_bitwise_or::wi_fold): Make sure
13609 nonzero|X is nonzero.
13610 (range_op_bitwise_and_tests): Add tests for above.
13612 2021-06-15 Carl Love <cel@us.ibm.com>
13615 * config/rs6000/rs6000-builtin.def (VCMPEQUT): Fix the ICODE for the
13617 (VRLQ, VSLQ, VSRQ, VSRAQ): Remove unused BU_P10_OVERLOAD_2
13620 2021-06-15 Tobias Burnus <tobias@codesourcery.com>
13623 * gimplify.c (enum gimplify_defaultmap_kind): Add GDMK_SCALAR_TARGET.
13624 (struct gimplify_omp_ctx): Extend defaultmap array by one.
13625 (new_omp_context): Init defaultmap[GDMK_SCALAR_TARGET].
13626 (omp_notice_variable): Update type classification for Fortran.
13627 (gimplify_scan_omp_clauses): Update calls for new argument; handle
13628 GDMK_SCALAR_TARGET; for Fortran, GDMK_POINTER avoid GOVD_MAP_0LEN_ARRAY.
13629 * langhooks-def.h (lhd_omp_scalar_p): Add 'ptr_ok' argument.
13630 * langhooks.c (lhd_omp_scalar_p): Likewise.
13631 (LANG_HOOKS_OMP_ALLOCATABLE_P, LANG_HOOKS_OMP_SCALAR_TARGET_P): New.
13632 (LANG_HOOKS_DECLS): Add them.
13633 * langhooks.h (struct lang_hooks_for_decls): Add new hooks, update
13634 omp_scalar_p pointer type to include the new bool argument.
13636 2021-06-15 David Malcolm <dmalcolm@redhat.com>
13638 * doc/analyzer.texi
13639 (Special Functions for Debugging the Analyzer): Add
13640 __analyzer_dump_capacity.
13642 2021-06-15 Jakub Jelinek <jakub@redhat.com>
13645 * expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
13646 case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.
13648 2021-06-15 Richard Biener <rguenther@suse.de>
13650 * cfgloopanal.c (mark_irreducible_loops): Use a dominance
13651 check to identify loop latches.
13652 * cfgloop.c (verify_loop_structure): Likewise.
13653 * loop-init.c (apply_loop_flags): Allow marked irreducible
13654 regions even with multiple latches.
13655 * predict.c (rebuild_frequencies): Simplify.
13657 2021-06-15 Richard Biener <rguenther@suse.de>
13659 * tree-ssa-threadupdate.c
13660 (jump_thread_path_registry::mark_threaded_blocks): Assert we
13661 have marked irreducible regions.
13663 2021-06-14 Martin Sebor <msebor@redhat.com>
13666 * builtins.c (gimple_call_return_array): Check for attribute fn spec.
13667 Handle calls to placement new.
13668 (ndecl_dealloc_argno): Avoid placement delete.
13670 2021-06-14 Peter Bergner <bergner@linux.ibm.com>
13673 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
13674 create_tmp_reg_or_ssa_name().
13676 2021-06-14 Andrew MacLeod <amacleod@redhat.com>
13678 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
13679 (ranger_cache::enable_new_values): Set to specified value and
13680 return the old value.
13681 (ranger_cache::disable_new_values): Delete.
13682 (ranger_cache::fill_block_cache): Disable non 1st order derived
13684 * gimple-range-cache.h (ranger_cache): Adjust prototypes.
13685 * gimple-range.cc (gimple_ranger::range_of_expr): Adjust.
13687 2021-06-14 Uroš Bizjak <ubizjak@gmail.com>
13690 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
13691 Return true early when testing with V2HImode.
13692 * config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.
13694 2021-06-14 Christophe Lyon <christophe.lyon@linaro.org>
13696 * config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
13697 (mve_vec_unpack<US>_hi_<mode>): New pattern.
13698 (@mve_vec_pack_trunc_lo_<mode>): New pattern.
13699 (mve_vmovntq_<supf><mode>): Prefix with '@'.
13700 * config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
13702 (vec_unpack<US>_lo_<mode>): Likewise.
13703 (vec_pack_trunc_<mode>): Rename to
13704 neon_quad_vec_pack_trunc_<mode>.
13705 * config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
13707 (vec_unpack<US>_lo_<mode>): New.
13708 (vec_pack_trunc_<mode>): New.
13710 2021-06-14 Richard Biener <rguenther@suse.de>
13712 PR tree-optimization/100934
13713 * tree-ssa-dom.c (pass_dominator::execute): Properly
13714 mark irreducible regions.
13716 2021-06-14 Martin Liska <mliska@suse.cz>
13718 * doc/invoke.texi: Put r{...} on the same line as @item.
13720 2021-06-14 Martin Liska <mliska@suse.cz>
13722 * doc/invoke.texi: Add missing newline.
13724 2021-06-14 Martin Liska <mliska@suse.cz>
13726 * doc/invoke.texi: Remove '+' charasters.
13728 2021-06-14 Claudiu Zissulescu <claziss@synopsys.com>
13730 * config.gcc (arc): Add support for with_cpu option.
13731 * config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.
13733 2021-06-14 Richard Biener <rguenther@suse.de>
13735 PR tree-optimization/101031
13736 * tree-ssa-strlen.c (maybe_invalidate): Increment max_size
13737 instead of size when accounting for a possibly string
13740 2021-06-14 Martin Liska <mliska@suse.cz>
13742 * gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].
13744 2021-06-14 Aldy Hernandez <aldyh@redhat.com>
13746 * value-query.cc (gimple_range_global): Call get_range_global
13747 if called after inlining.
13749 2021-06-13 Uroš Bizjak <ubizjak@gmail.com>
13752 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
13753 Emit constant permutation insn directly from here.
13755 2021-06-13 Trevor Saunders <tbsaunde@tbsaunde.org>
13757 * attribs.c (find_attribute_namespace): Iterate over vec<> with
13759 * auto-profile.c (afdo_find_equiv_class): Likewise.
13760 * gcc.c (do_specs_vec): Likewise.
13761 (do_spec_1): Likewise.
13762 (driver::set_up_specs): Likewise.
13763 * gimple-loop-jam.c (any_access_function_variant_p): Likewise.
13764 * gimple-ssa-store-merging.c (compatible_load_p): Likewise.
13765 (imm_store_chain_info::try_coalesce_bswap): Likewise.
13766 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
13767 (get_location_for_stmts): Likewise.
13768 * graphite-poly.c (print_iteration_domains): Likewise.
13769 (free_poly_bb): Likewise.
13770 (remove_gbbs_in_scop): Likewise.
13771 (free_scop): Likewise.
13772 (dump_gbb_cases): Likewise.
13773 (dump_gbb_conditions): Likewise.
13774 (print_pdrs): Likewise.
13775 (print_scop): Likewise.
13776 * ifcvt.c (cond_move_process_if_block): Likewise.
13777 * lower-subreg.c (decompose_multiword_subregs): Likewise.
13778 * regcprop.c (pass_cprop_hardreg::execute): Likewise.
13779 * sanopt.c (sanitize_rewrite_addressable_params): Likewise.
13780 * sel-sched-dump.c (dump_insn_vector): Likewise.
13781 * store-motion.c (store_ops_ok): Likewise.
13782 (store_killed_in_insn): Likewise.
13783 * timevar.c (timer::named_items::print): Likewise.
13784 * tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
13785 (cleanup_tree_cfg_noloop): Likewise.
13786 * tree-data-ref.c (dump_data_references): Likewise.
13787 (print_dir_vectors): Likewise.
13788 (print_dist_vectors): Likewise.
13789 (dump_data_dependence_relations): Likewise.
13790 (dump_dist_dir_vectors): Likewise.
13791 (dump_ddrs): Likewise.
13792 (create_runtime_alias_checks): Likewise.
13793 (free_subscripts): Likewise.
13794 (save_dist_v): Likewise.
13795 (save_dir_v): Likewise.
13796 (invariant_access_functions): Likewise.
13797 (same_access_functions): Likewise.
13798 (access_functions_are_affine_or_constant_p): Likewise.
13799 (find_data_references_in_stmt): Likewise.
13800 (graphite_find_data_references_in_stmt): Likewise.
13801 (free_dependence_relations): Likewise.
13802 (free_data_refs): Likewise.
13803 * tree-inline.c (copy_debug_stmts): Likewise.
13804 * tree-into-ssa.c (dump_currdefs): Likewise.
13805 (rewrite_update_phi_arguments): Likewise.
13806 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
13807 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
13809 (vect_slp_analyze_node_dependences): Likewise.
13810 (vect_slp_analyze_instance_dependence): Likewise.
13811 (vect_record_base_alignments): Likewise.
13812 (vect_get_peeling_costs_all_drs): Likewise.
13813 (vect_peeling_supportable): Likewise.
13814 * tree-vectorizer.c (vec_info::~vec_info): Likewise.
13815 (vec_info::free_stmt_vec_infos): Likewise.
13817 2021-06-13 Jeff Law <jeffreyalaw@gmail.com>
13819 * config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
13820 (andqi3_1<cczn>): Removed.
13821 (<ors>qi3_1): Do not split for IOR/XOR a single bit.
13822 (H8/SX bit logicals): Split out from other patterns.
13823 * config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
13824 mulqihi3_const_clobber_flags.
13825 (mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly
13827 2021-06-13 H.J. Lu <hjl.tools@gmail.com>
13830 * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
13831 to true if red zone is used.
13832 (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
13833 ix86_red_zone_used.
13834 * config/i386/i386.h (machine_function): Add red_zone_used.
13835 (ix86_red_zone_size): Removed.
13836 (ix86_red_zone_used): New.
13837 * config/i386/i386.md (peephole2 patterns): Replace
13838 ix86_red_zone_size with ix86_red_zone_used.
13840 2021-06-12 Jason Merrill <jason@redhat.com>
13842 * doc/extend.texi (unused variable attribute): Applies to
13843 structure fields as well.
13845 2021-06-12 Eugene Rozenfeld <erozen@microsoft.com>
13847 * auto-profile.c (read_profile): fix a typo in an error string
13849 2021-06-11 Thomas Schwinge <thomas@codesourcery.com>
13851 * tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
13853 * tree-pretty-print.c (dump_omp_clauses): Update.
13854 (dump_generic_node) <OMP_CLAUSE>: Use it.
13856 2021-06-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
13859 * config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
13860 int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
13861 the polymorphic variants matching code.
13862 (__arm_vld1q_z): Likewise.
13863 (__arm_vld2q): Likewise.
13864 (__arm_vld4q): Likewise.
13865 (__arm_vldrbq_gather_offset): Likewise.
13866 (__arm_vldrbq_gather_offset_z): Likewise.
13868 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
13870 PR tree-optimization/96392
13871 * fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.
13873 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
13875 PR tree-optimization/96392
13876 * fold-const.c (fold_real_zero_addition_p): Take both arguments
13877 of the addition or subtraction, not just the zero. Use this
13878 other argument in tests for signaling NaNs and signed zeros.
13879 (tree_expr_maybe_real_minus_zero_p): New predicate.
13880 * fold-const.h (fold_real_zero_addition_p): Update prototype.
13881 (tree_expr_maybe_real_minus_zero_p): New function prototype.
13882 * match.pd: Update calls to fold_real_zero_addition_p.
13883 Replace HONOR_NANS with tree_expr_maybe_nan_p.
13884 Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
13885 Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
13886 * tree-ssa-reassoc.c (eliminate_using_constants): Update
13887 call to fold_real_zero_addition_p.
13889 2021-06-11 Richard Biener <rguenther@suse.de>
13891 PR tree-optimization/101025
13892 * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
13893 all refs that require dependence checking.
13895 2021-06-11 Richard Biener <rguenther@suse.de>
13897 PR tree-optimization/101028
13898 * tree-vect-slp.c (vect_build_slp_tree_2): When SLP
13899 reassoc discovery fails fatally, mark appropriate lanes
13902 2021-06-11 Richard Biener <rguenther@suse.de>
13904 PR tree-optimization/101026
13905 * tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
13906 have a representative for the associated chain nodes.
13908 2021-06-11 Jakub Jelinek <jakub@redhat.com>
13910 PR rtl-optimization/101008
13911 * simplify-rtx.c (relational_result): New function.
13912 (simplify_logical_relational_operation,
13913 simplify_relational_operation): Use it.
13915 2021-06-11 Jakub Jelinek <jakub@redhat.com>
13918 * config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.
13920 2021-06-11 Uroš Bizjak <ubizjak@gmail.com>
13923 * config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
13924 false if the permutation can be implemented with constant
13925 permutation instruction in wider mode.
13926 (canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
13927 Handle V8QImode and V4HImode.
13929 2021-06-11 Martin Liska <mliska@suse.cz>
13931 PR gcov-profile/100788
13932 * common.opt: Add new option.
13933 * coverage.c (coverage_begin_function): Emit warning instead on
13934 the internal compiler error.
13935 * doc/invoke.texi: Document the option.
13936 * toplev.c (process_options): Enable it by default.
13938 2021-06-11 Richard Biener <rguenther@suse.de>
13940 PR middle-end/101009
13941 * tree-data-ref.c (build_classic_dist_vector_1): Make sure
13942 to set *init_b to true when we encounter a constant equal
13944 (compute_affine_dependence): Also dump the actual DR_REF.
13946 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
13948 PR tree-optimization/100984
13949 * gimple-ssa-evrp.c (ssa_equiv_stack): Use auto_vec for
13950 replacements table.
13951 (ssa_equiv_stack::~ssa_equiv_stack): Remove.
13953 2021-06-11 Kewen Lin <linkw@linux.ibm.com>
13955 * config/rs6000/rs6000.md
13956 (floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
13957 define_insn_and_split.
13959 2021-06-11 Richard Biener <rguenther@suse.de>
13961 * tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
13962 to sort operands of the associative chain.
13964 2021-06-11 Richard Biener <rguenther@suse.de>
13966 * system.h (gcc_stablesort_r): Declare.
13967 * sort.cc (gcc_sort_r): Support stable sort.
13968 (gcc_stablesort_r): Define.
13969 * vec.h (vec<>::stablesort): Add.
13971 2021-06-10 Uroš Bizjak <ubizjak@gmail.com>
13974 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
13975 Handle V2SF mode. Emit SHUFPS to fixup unpack-high for V2SF mode.
13976 (expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
13977 (expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
13978 (expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
13979 (expand_vec_perm_interleave2): Handle 64bit modes.
13980 (expand_vec_perm_even_odd_pack): Handle V8QI mode.
13981 (expand_vec_perm_even_odd_1): Ditto.
13982 (ix86_vectorize_vec_perm_const): Ditto.
13983 * config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
13984 * config/i386/sse.md: ... here.
13985 * config/i386/mmx.md (*vec_interleave_lowv2sf):
13986 New insn_and_split pattern.
13987 (*vec_interleave_highv2sf): Ditto.
13988 (mmx_pshufbv8qi3): New insn pattern.
13989 (*mmx_pblendw): Ditto.
13991 2021-06-10 Peter Bergner <bergner@linux.ibm.com>
13993 * config/rs6000/rs6000-builtin.def (build_pair): New built-in.
13994 (build_acc): Likewise.
13995 * config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
13996 source operands in little-endian mode.
13997 (rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
13998 (mma_init_builtins): Likewise.
13999 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
14000 ordering for the MMA assemble and build source operands.
14001 * doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
14003 (__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
14006 2021-06-10 Jeff Law <jeffreyalaw@gmail.com>
14008 * config/h8300/h8300.c (select_cc_mode): Handle MEM. Use
14010 * config/h8300/extensions.md: Replace _clobber_flags patterns
14013 2021-06-10 Robin Dapp <rdapp@linux.ibm.com>
14015 * config/s390/vector.md (vcond_mask_<mode><mode>): Change to
14016 (vcond_mask_<mode><tointvec>): this.
14018 2021-06-10 Andrew Stubbs <ams@codesourcery.com>
14019 Thomas Schwinge <thomas@codesourcery.com>
14021 * omp-builtins.def (BUILT_IN_GOACC_ENTER_EXIT_DATA): Split into...
14022 (BUILT_IN_GOACC_ENTER_DATA, BUILT_IN_GOACC_EXIT_DATA): ... these.
14023 * gimple.h (enum gf_mask): Split
14024 'GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA' into
14025 'GF_OMP_TARGET_KIND_OACC_ENTER_DATA' and
14026 'GF_OMP_TARGET_KIND_OACC_EXIT_DATA'.
14027 (is_gimple_omp_oacc): Update.
14028 * gimple-pretty-print.c (dump_gimple_omp_target): Likewise.
14029 * gimplify.c (gimplify_omp_target_update): Likewise.
14030 * omp-expand.c (expand_omp_target, build_omp_regions_1)
14031 (omp_make_gimple_edges): Likewise.
14032 * omp-low.c (check_omp_nesting_restrictions, lower_omp_target):
14035 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
14037 * value-query.cc (value_query::value_on_edge): Rename name to
14039 (range_query::range_on_edge): Same.
14040 (range_query::value_of_expr): Same.
14041 (range_query::value_on_edge): Same.
14042 * value-query.h (class value_query): Same.
14043 (class range_query): Same.
14045 2021-06-10 Richard Biener <rguenther@suse.de>
14047 PR tree-optimization/101003
14048 * tree-vect-slp.c (vect_build_slp_tree_2): Appropriately
14049 use the pattern stmt defs when linearizing a chain.
14051 2021-06-10 Jakub Jelinek <jakub@redhat.com>
14054 * ifcvt.c (noce_get_alt_condition, noce_try_abs): Use
14055 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
14057 2021-06-10 Clement Chigot <clement.chigot@atos.net>
14059 * config/rs6000/aix71.h (ASM_CPU_SPEC): Add Power10 directive.
14060 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
14062 2021-06-09 Andrew Pinski <apinski@marvell.com>
14064 PR tree-optimization/100925
14065 * match.pd (a ? CST1 : CST2): Limit transformations
14066 that would produce a negative to integeral types only.
14067 Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.
14069 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
14072 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
14074 * doc/tm.texi: Correctly update.
14076 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
14078 * doc/tm.texi: Correctly update.
14080 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
14083 * doc/tm.texi.in (Trampolines): Add a missing blank line.
14085 2021-06-09 Paul Eggert <eggert@cs.ucla.edu>
14088 * doc/invoke.texi (Code Gen Options); Document that -fno-trampolines
14089 and -ftrampolines work only with Ada.
14090 * doc/tm.texi.in (Trampolines): Likewise.
14091 * doc/tm.texi: Regenerated.
14093 2021-06-09 Carl Love <cel@us.ibm.com>
14095 * config/rs6000/altivec.h (vec_signextll, vec_signexti, vec_signextq):
14096 Add define for new builtins.
14097 * config/rs6000/altivec.md(altivec_vreveti2): Add define_expand.
14098 * config/rs6000/rs6000-builtin.def (VSIGNEXTI, VSIGNEXTLL): Add
14099 overloaded builtin definitions.
14100 (VSIGNEXTSB2W, VSIGNEXTSH2W, VSIGNEXTSB2D, VSIGNEXTSH2D,VSIGNEXTSW2D,
14101 VSIGNEXTSD2Q): Add builtin expansions.
14102 (SIGNEXT): Add P10 overload definition.
14103 * config/rs6000/rs6000-call.c (P9V_BUILTIN_VEC_VSIGNEXTI, P9V_BUILTIN_VEC_VSIGNEXTLL,
14104 P10_BUILTIN_VEC_SIGNEXT): Add overloaded argument definitions.
14105 * config/rs6000/vsx.md (vsx_sign_extend_v2di_v1ti): Add define_insn.
14106 (vsignextend_v2di_v1ti, vsignextend_qi_<mode>, vsignextend_hi_<mode>,
14107 vsignextend_si_v2di)[VIlong]: Add define_expand.
14108 Make define_insn vsx_sign_extend_si_v2di visible.
14109 * doc/extend.texi: Add documentation for the vec_signexti,
14110 vec_signextll builtins and vec_signextq.
14112 2021-06-09 Carl Love <cel@us.ibm.com>
14114 * config/rs6000/rs6000.c (__fixkfti, __fixunskfti, __floattikf,
14115 __floatuntikf): Names changed to __fixkfti_sw, __fixunskfti_sw,
14116 __floattikf_sw, __floatuntikf_sw respectively.
14117 * config/rs6000/rs6000.md (floatti<mode>2, floatunsti<mode>2,
14118 fix_trunc<mode>ti2, fixuns_trunc<mode>ti2): Add
14119 define_insn for mode IEEE 128.
14121 2021-06-09 Carl Love <cel@us.ibm.com>
14123 * config/rs6000/altivec.md (altivec_vslq, altivec_vsrq):
14124 Rename to altivec_vslq_<mode>, altivec_vsrq_<mode>, mode VEC_TI.
14125 * config/rs6000/vector.md (VEC_TI): Was named VSX_TI in vsx.md.
14126 (vashlv1ti3): Change to vashl<mode>3, mode VEC_TI.
14127 (vlshrv1ti3): Change to vlshr<mode>3, mode VEC_TI.
14128 * config/rs6000/vsx.md (VSX_TI): Remove define_mode_iterator. Update
14129 uses of VSX_TI to VEC_TI.
14131 2021-06-09 Carl Love <cel@us.ibm.com>
14133 * config/rs6000/dfp.md (floattitd2, fixtdti2): New define_insns.
14135 2021-06-09 Carl Love <cel@us.ibm.com>
14137 * config/rs6000/altivec.h (vec_dive, vec_mod): Add define for new
14139 * config/rs6000/altivec.md (UNSPEC_VMULEUD, UNSPEC_VMULESD,
14140 UNSPEC_VMULOUD, UNSPEC_VMULOSD): New unspecs.
14141 (altivec_eqv1ti, altivec_gtv1ti, altivec_gtuv1ti, altivec_vmuleud,
14142 altivec_vmuloud, altivec_vmulesd, altivec_vmulosd, altivec_vrlq,
14143 altivec_vrlqmi, altivec_vrlqmi_inst, altivec_vrlqnm,
14144 altivec_vrlqnm_inst, altivec_vslq, altivec_vsrq, altivec_vsraq,
14145 altivec_vcmpequt_p, altivec_vcmpgtst_p, altivec_vcmpgtut_p): New
14147 (vec_widen_umult_even_v2di, vec_widen_smult_even_v2di,
14148 vec_widen_umult_odd_v2di, vec_widen_smult_odd_v2di, altivec_vrlqmi,
14149 altivec_vrlqnm): New define_expands.
14150 * config/rs6000/rs6000-builtin.def (VCMPEQUT_P, VCMPGTST_P,
14151 VCMPGTUT_P): Add macro expansions.
14152 (BU_P10V_AV_P): Add builtin predicate definition.
14153 (VCMPGTUT, VCMPGTST, VCMPEQUT, CMPNET, CMPGE_1TI,
14154 CMPGE_U1TI, CMPLE_1TI, CMPLE_U1TI, VNOR_V1TI_UNS, VNOR_V1TI, VCMPNET_P,
14155 VCMPAET_P, VMULEUD, VMULESD, VMULOUD, VMULOSD, VRLQ,
14156 VSLQ, VSRQ, VSRAQ, VRLQNM, DIV_V1TI, UDIV_V1TI, DIVES_V1TI, DIVEU_V1TI,
14157 MODS_V1TI, MODU_V1TI, VRLQMI): New macro expansions.
14158 (VRLQ, VSLQ, VSRQ, VSRAQ, DIVE, MOD): New overload expansions.
14159 * config/rs6000/rs6000-call.c (P10_BUILTIN_VCMPEQUT,
14160 P10V_BUILTIN_CMPGE_1TI, P10V_BUILTIN_CMPGE_U1TI,
14161 P10V_BUILTIN_VCMPGTUT, P10V_BUILTIN_VCMPGTST,
14162 P10V_BUILTIN_CMPLE_1TI, P10V_BUILTIN_VCMPLE_U1TI,
14163 P10V_BUILTIN_DIV_V1TI, P10V_BUILTIN_UDIV_V1TI,
14164 P10V_BUILTIN_VMULESD, P10V_BUILTIN_VMULEUD,
14165 P10V_BUILTIN_VMULOSD, P10V_BUILTIN_VMULOUD,
14166 P10V_BUILTIN_VNOR_V1TI, P10V_BUILTIN_VNOR_V1TI_UNS,
14167 P10V_BUILTIN_VRLQ, P10V_BUILTIN_VRLQMI,
14168 P10V_BUILTIN_VRLQNM, P10V_BUILTIN_VSLQ,
14169 P10V_BUILTIN_VSRQ, P10V_BUILTIN_VSRAQ,
14170 P10V_BUILTIN_VCMPGTUT_P, P10V_BUILTIN_VCMPGTST_P,
14171 P10V_BUILTIN_VCMPEQUT_P, P10V_BUILTIN_VCMPGTUT_P,
14172 P10V_BUILTIN_VCMPGTST_P, P10V_BUILTIN_CMPNET,
14173 P10V_BUILTIN_VCMPNET_P, P10V_BUILTIN_VCMPAET_P,
14174 P10V_BUILTIN_DIVES_V1TI, P10V_BUILTIN_MODS_V1TI,
14175 P10V_BUILTIN_MODU_V1TI):
14176 New overloaded definitions.
14177 (rs6000_gimple_fold_builtin) [P10V_BUILTIN_VCMPEQUT,
14178 P10V_BUILTIN_CMPNET, P10V_BUILTIN_CMPGE_1TI,
14179 P10V_BUILTIN_CMPGE_U1TI, P10V_BUILTIN_VCMPGTUT,
14180 P10V_BUILTIN_VCMPGTST, P10V_BUILTIN_CMPLE_1TI,
14181 P10V_BUILTIN_CMPLE_U1TI]: New case statements.
14182 (rs6000_init_builtins) [bool_V1TI_type_node, int_ftype_int_v1ti_v1ti]:
14184 (altivec_init_builtins): New E_V1TImode case statement.
14185 (builtin_function_type)[P10_BUILTIN_128BIT_VMULEUD,
14186 P10_BUILTIN_128BIT_VMULOUD, P10_BUILTIN_128BIT_DIVEU_V1TI,
14187 P10_BUILTIN_128BIT_MODU_V1TI, P10_BUILTIN_CMPGE_U1TI,
14188 P10_BUILTIN_VCMPGTUT, P10_BUILTIN_VCMPEQUT]: New case statements.
14189 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute) [E_TImode,
14190 E_V1TImode]: New case statements.
14191 * config/rs6000/rs6000.h (rs6000_builtin_type_index): New enum
14192 value RS6000_BTI_bool_V1TI.
14193 * config/rs6000/vector.md (vector_gtv1ti,vector_nltv1ti,
14194 vector_gtuv1ti, vector_nltuv1ti, vector_ngtv1ti, vector_ngtuv1ti,
14195 vector_eq_v1ti_p, vector_ne_v1ti_p, vector_ae_v1ti_p,
14196 vector_gt_v1ti_p, vector_gtu_v1ti_p, vrotlv1ti3, vashlv1ti3,
14197 vlshrv1ti3, vashrv1ti3): New define_expands.
14198 * config/rs6000/vsx.md (UNSPEC_VSX_DIVSQ, UNSPEC_VSX_DIVUQ,
14199 UNSPEC_VSX_DIVESQ, UNSPEC_VSX_DIVEUQ, UNSPEC_VSX_MODSQ,
14200 UNSPEC_VSX_MODUQ): New unspecs.
14201 (mulv2di3, vsx_div_v1ti, vsx_udiv_v1ti, vsx_dives_v1ti,
14202 vsx_diveu_v1ti, vsx_mods_v1ti, vsx_modu_v1ti, xxswapd_v1ti): New
14204 (vcmpnet): New define_expand.
14205 * doc/extend.texi: Add documentation for the new builtins vec_rl,
14206 vec_rlmi, vec_rlnm, vec_sl, vec_sr, vec_sra, vec_mule, vec_mulo,
14207 vec_div, vec_dive, vec_mod, vec_cmpeq, vec_cmpne, vec_cmpgt, vec_cmplt,
14208 vec_cmpge, vec_cmple, vec_all_eq, vec_all_ne, vec_all_gt, vec_all_lt,
14209 vec_all_ge, vec_all_le, vec_any_eq, vec_any_ne, vec_any_gt, vec_any_lt,
14210 vec_any_ge, vec_any_le.
14212 2021-06-09 Carl Love <cel@us.ibm.com>
14214 * config/rs6000/altivec.md (altivec_vrl<VI_char>mi): Fix
14215 bug in argument generation.
14217 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
14219 * config/arm/iterators.md (<supf>): Remove VCLZQ_U, VCLZQ_S.
14221 * config/arm/mve.md (mve_vclzq_<supf><mode>): Add '@' prefix,
14222 remove <supf> iterator.
14223 (mve_vclzq_u<mode>): New.
14224 * config/arm/neon.md (clz<mode>2): Rename to neon_vclz<mode>.
14225 (neon_vclz<mode): Move to ...
14226 * config/arm/unspecs.md (VCLZQ_U, VCLZQ_S): Remove.
14227 * config/arm/vec-common.md: ... here. Add support for MVE.
14229 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
14231 * config/arm/mve.md (mve_vhaddq_<supf><mode>): Prefix with '@'.
14232 (@mve_vrhaddq_<supf><mode): Likewise.
14233 * config/arm/neon.md (neon_v<r>hadd<sup><mode>): Likewise.
14234 * config/arm/vec-common.md (avg<mode>3_floor, uavg<mode>3_floor)
14235 (avg<mode>3_ceil", uavg<mode>3_ceil): New patterns.
14237 2021-06-09 imba-tjd <109224573@qq.com>
14239 * doc/invoke.texi: Fix typo.
14241 2021-06-09 Roger Sayle <roger@nextmovesoftware.com>
14243 PR middle-end/53267
14244 * fold-const-call.c (fold_const_call_sss) [CASE_CFN_FMOD]:
14245 Support evaluation of fmod/fmodf/fmodl at compile-time.
14247 2021-06-09 Richard Biener <rguenther@suse.de>
14249 PR tree-optimization/100981
14250 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
14251 gimple_get_lhs to also handle calls.
14252 * tree-vect-slp-patterns.c (complex_pattern::build): Transfer
14255 2021-06-09 Richard Biener <rguenther@suse.de>
14257 PR tree-optimization/97832
14258 * tree-vectorizer.h (_slp_tree::failed): New.
14259 * tree-vect-slp.c (_slp_tree::_slp_tree): Initialize
14261 (_slp_tree::~_slp_tree): Free failed.
14262 (vect_build_slp_tree): Retain failed nodes and record
14263 matches in them, copying that back out when running
14264 into a cached fail. Dump start and end of discovery.
14265 (dt_sort_cmp): New.
14266 (vect_build_slp_tree_2): Handle associatable chains
14267 together doing more aggressive operand swapping.
14269 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
14272 * config.gcc (gcc_cv_initfini_array): Set to yes for Linux and
14274 * doc/install.texi: Require glibc 2.1 and binutils 2.12 for
14275 Linux and GNU targets.
14277 2021-06-09 Richard Biener <rguenther@suse.de>
14279 * tree-vect-stmts.c (vect_is_simple_use): Always get dt
14282 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
14284 * config/arc/arc.md (loop_end): Change it to
14285 define_insn_and_split.
14287 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
14289 * config/arc/arc.md (maddhisi4): Use VMAC2H instruction.
14290 (machi): New pattern.
14291 (umaddhisi4): Use VMAC2HU instruction.
14292 (umachi): New pattern.
14294 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
14296 * config/arc/arc-protos.h (arc_split_move_p): New prototype.
14297 * config/arc/arc.c (arc_split_move_p): New function.
14298 (arc_split_move): Clean up.
14299 * config/arc/arc.md (movdi_insn): Clean up, use arc_split_move_p.
14300 (movdf_insn): Likewise.
14301 * config/arc/simdext.md (mov<VWH>_insn): Likewise.
14303 2021-06-09 Uroš Bizjak <ubizjak@gmail.com>
14306 * config/i386/i386.c (print_operand_address_as): Rename "no_rip"
14307 argument to "raw". Do not emit segment overrides when "raw" is true.
14309 2021-06-09 Martin Liska <mliska@suse.cz>
14311 * doc/gcov.texi: Create a proper JSON files.
14312 * doc/invoke.texi: Remove dots in order to make it a valid
14315 2021-06-09 Xionghu Luo <luoxhu@linux.ibm.com>
14317 * config/rs6000/rs6000-p8swap.c (pattern_is_rotate64): New.
14318 (insn_is_load_p): Use pattern_is_rotate64.
14319 (insn_is_swap_p): Likewise.
14320 (quad_aligned_load_p): Likewise.
14321 (const_load_sequence_p): Likewise.
14322 (replace_swapped_aligned_load): Likewise.
14323 (recombine_lvx_pattern): Likewise.
14324 (recombine_stvx_pattern): Likewise.
14326 2021-06-09 Andrew MacLeod <amacleod@redhat.com>
14328 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Use a
14329 fur_stmt source record.
14330 * gimple-range.cc (fur_source::get_operand): Generic range query.
14331 (fur_source::get_phi_operand): New.
14332 (fur_source::register_dependency): New.
14333 (fur_source::query): New.
14334 (class fur_edge): New. Edge source for operands.
14335 (fur_edge::fur_edge): New.
14336 (fur_edge::get_operand): New.
14337 (fur_edge::get_phi_operand): New.
14338 (fur_edge::query): New.
14339 (fur_stmt::fur_stmt): New.
14340 (fur_stmt::get_operand): New.
14341 (fur_stmt::get_phi_operand): New.
14342 (fur_stmt::query): New.
14343 (class fur_depend): New. Statement source and process dependencies.
14344 (fur_depend::fur_depend): New.
14345 (fur_depend::register_dependency): New.
14346 (class fur_list): New. List source for operands.
14347 (fur_list::fur_list): New.
14348 (fur_list::get_operand): New.
14349 (fur_list::get_phi_operand): New.
14350 (fold_range): New. Instantiate appropriate fur_source class and fold.
14351 (fold_using_range::range_of_range_op): Use new API.
14352 (fold_using_range::range_of_address): Ditto.
14353 (fold_using_range::range_of_phi): Ditto.
14354 (imple_ranger::fold_range_internal): Use fur_depend class.
14355 (fold_using_range::range_of_ssa_name_with_loop_info): Use new API.
14356 * gimple-range.h (class fur_source): Now a base class.
14357 (class fur_stmt): New.
14358 (fold_range): New prototypes.
14359 (fur_source::fur_source): Delete.
14361 2021-06-08 Andrew Pinski <apinski@marvell.com>
14363 PR tree-optimization/25290
14364 * tree-ssa-phiopt.c (xor_replacement): Delete.
14365 (tree_ssa_phiopt_worker): Delete use of xor_replacement.
14366 (match_simplify_replacement): Allow one cheap preparation
14367 statement that can be moved to before the if.
14369 2021-06-08 Pat Haugen <pthaugen@linux.ibm.com>
14371 * config/rs6000/power10.md (power10-fused-load, power10-fused-store,
14372 power10-fused_alu, power10-fused-vec, power10-fused-branch): New.
14374 2021-06-08 Jeff Law <jeffreyalaw@gmail.com>
14376 * config/h8300/logical.md (andqi3_1): Move BCLR case into define_insn_and_split.
14377 Create length attribute on define_insn_and_split. Only split for cases which we
14379 (andqi3_1<cczn>): Renamed from andqi3_1_clobber_flags. Only handle AND here and
14380 fix length computation.
14381 (b<code><mode>msx): Combine QImode and HImode H8/SX patterns using iterator.
14383 2021-06-08 Richard Biener <rguenther@suse.de>
14385 PR tree-optimization/100923
14386 * tree-ssa-sccvn.c (valueize_refs_1): Take a pointer to
14387 the operand vector to be valueized.
14388 (valueize_refs): Likewise.
14389 (valueize_shared_reference_ops_from_ref): Adjust.
14390 (valueize_shared_reference_ops_from_call): Likewise.
14391 (vn_reference_lookup_3): Likewise.
14392 (vn_reference_lookup_pieces): Likewise. Re-valueize
14393 with honoring availability when we are about to create
14394 the ao_ref and valueized before.
14395 (vn_reference_lookup): Likewise.
14396 (vn_reference_insert_pieces): Adjust.
14398 2021-06-08 Richard Biener <rguenther@suse.de>
14400 * tree-vectorizer.h (_slp_instance::root_stmt): Change to...
14401 (_slp_instance::root_stmts): ... a vector.
14402 (SLP_INSTANCE_ROOT_STMT): Rename to ...
14403 (SLP_INSTANCE_ROOT_STMTS): ... this.
14404 (slp_root::root): Change to...
14405 (slp_root::roots): ... a vector.
14406 (slp_root::slp_root): Adjust.
14407 * tree-vect-slp.c (_slp_instance::location): Adjust.
14408 (vect_free_slp_instance): Release the root stmt vector.
14409 (vect_build_slp_instance): Adjust.
14410 (vect_analyze_slp): Likewise.
14411 (_bb_vec_info::~_bb_vec_info): Likewise.
14412 (vect_slp_analyze_operations): Likewise.
14413 (vect_bb_vectorization_profitable_p): Likewise. Adjust
14414 costs for the root stmt.
14415 (vect_slp_check_for_constructors): Gather all BIT_INSERT_EXPRs
14417 (vect_slp_analyze_bb_1): Simplify by marking all root stmts
14419 (vectorize_slp_instance_root_stmt): Adjust.
14420 (vect_schedule_slp): Likewise.
14422 2021-06-08 Aldy Hernandez <aldyh@redhat.com>
14424 * gimple-ssa-evrp.c (class ssa_equiv_stack): New.
14425 (ssa_equiv_stack::ssa_equiv_stack): New.
14426 (ssa_equiv_stack::~ssa_equiv_stack): New.
14427 (ssa_equiv_stack::enter): New.
14428 (ssa_equiv_stack::leave): New.
14429 (ssa_equiv_stack::push_replacement): New.
14430 (ssa_equiv_stack::get_replacement): New.
14431 (is_pointer_ssa): New.
14432 (class pointer_equiv_analyzer): New.
14433 (pointer_equiv_analyzer::pointer_equiv_analyzer): New.
14434 (pointer_equiv_analyzer::~pointer_equiv_analyzer): New.
14435 (pointer_equiv_analyzer::set_global_equiv): New.
14436 (pointer_equiv_analyzer::set_cond_equiv): New.
14437 (pointer_equiv_analyzer::get_equiv): New.
14438 (pointer_equiv_analyzer::enter): New.
14439 (pointer_equiv_analyzer::leave): New.
14440 (pointer_equiv_analyzer::get_equiv_expr): New.
14441 (pta_valueize): New.
14442 (pointer_equiv_analyzer::visit_stmt): New.
14443 (pointer_equiv_analyzer::visit_edge): New.
14444 (hybrid_folder::value_of_expr): Call PTA.
14445 (hybrid_folder::value_on_edge): Same.
14446 (hybrid_folder::pre_fold_bb): New.
14447 (hybrid_folder::post_fold_bb): New.
14448 (hybrid_folder::pre_fold_stmt): New.
14449 (rvrp_folder::pre_fold_bb): New.
14450 (rvrp_folder::post_fold_bb): New.
14451 (rvrp_folder::pre_fold_stmt): New.
14452 (rvrp_folder::value_of_expr): Call PTA.
14453 (rvrp_folder::value_on_edge): Same.
14455 2021-06-08 Jakub Jelinek <jakub@redhat.com>
14458 * tree-inline.c (copy_tree_body_r): For OMP_CLAUSE_DEPEND don't
14459 check TREE_CODE if OMP_CLAUSE_DECL is NULL.
14461 2021-06-08 Richard Biener <rguenther@suse.de>
14463 PR middle-end/100951
14464 * tree-vect-generic.c (expand_vector_piecewise): Build a
14465 VECTOR_CST if all elements are constant.
14466 (expand_vector_condition): Likewise.
14467 (lower_vec_perm): Likewise.
14468 (expand_vector_conversion): Likewise.
14470 2021-06-08 Martin Liska <mliska@suse.cz>
14472 * doc/invoke.texi: Document new param evrp-sparse-threshold.
14474 2021-06-08 Martin Liska <mliska@suse.cz>
14476 * genautomata.c (create_automata): Fix typo.
14478 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
14480 PR tree-optimization/100794
14481 * tree-predcom.c (tree_predictive_commoning_loop): Add parameter
14482 allow_unroll_p and only allow unrolling when it's true.
14483 (tree_predictive_commoning): Add parameter allow_unroll_p and
14485 (run_tree_predictive_commoning): Likewise.
14486 (pass_predcom::gate): Check flag_tree_loop_vectorize and
14487 global_options_set.x_flag_predictive_commoning.
14488 (pass_predcom::execute): Adjust for allow_unroll_p.
14490 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
14492 * tree-predcom.c (execute_pred_commoning): Remove update_ssa call.
14493 (tree_predictive_commoning_loop): Factor some cleanup stuffs into
14494 lambda function cleanup, remove scev_reset call, and adjust return
14496 (tree_predictive_commoning): Adjust for different changed values,
14497 only set flag TODO_update_ssa_only_virtuals if changed.
14498 (pass_data pass_data_predcom): Remove TODO_update_ssa_only_virtuals
14499 from todo_flags_finish.
14501 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
14503 * gimple-range-cache.cc (class sbr_sparse_bitmap): New.
14504 (sbr_sparse_bitmap::sbr_sparse_bitmap): New.
14505 (sbr_sparse_bitmap::bitmap_set_quad): New.
14506 (sbr_sparse_bitmap::bitmap_get_quad): New.
14507 (sbr_sparse_bitmap::set_bb_range): New.
14508 (sbr_sparse_bitmap::get_bb_range): New.
14509 (sbr_sparse_bitmap::bb_range_p): New.
14510 (block_range_cache::block_range_cache): initialize bitmap obstack.
14511 (block_range_cache::~block_range_cache): Destruct obstack.
14512 (block_range_cache::set_bb_range): Decide when to utilze the
14513 sparse on entry cache.
14514 * gimple-range-cache.h (block_range_cache): Add bitmap obstack.
14515 * params.opt (-param=evrp-sparse-threshold): New.
14517 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
14519 * bitmap.c (bitmap_set_aligned_chunk): New.
14520 (bitmap_get_aligned_chunk): New.
14521 (test_aligned_chunk): New.
14522 (bitmap_c_tests): Call test_aligned_chunk.
14523 * bitmap.h (bitmap_set_aligned_chunk, bitmap_get_aligned_chunk): New.
14525 2021-06-07 Uroš Bizjak <ubizjak@gmail.com>
14528 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
14530 (ix86_expand_vector_init_one_nonzero): Ditto.
14531 (ix86_expand_vector_init_one_var): Ditto.
14532 (ix86_expand_vector_init_general): Ditto.
14533 * config/i386/mmx.md (vec_initv4qiqi): New expander.
14535 2021-06-07 Jeff Law <jeffreyalaw@gmail.com>
14537 * config/h8300/movepush.md: Change most _clobber_flags
14538 patterns to instead use <cczn> subst.
14539 (movsi_cczn): New pattern with usable CC cases split out.
14540 (movsi_h8sx_cczn): Likewise.
14542 2021-06-07 Martin Liska <mliska@suse.cz>
14544 * common/common-target.def: Split long lines and replace them
14546 * target.def: Likewise.
14547 * doc/tm.texi: Re-generated.
14549 2021-06-07 Jakub Jelinek <jakub@redhat.com>
14552 * fold-const.c (fold_read_from_vector): Return NULL if trying to
14553 read from a CONSTRUCTOR with vector type elements.
14555 2021-06-07 Jakub Jelinek <jakub@redhat.com>
14557 PR middle-end/100898
14558 * tree-inline.c (copy_bb): Only use gimple_call_arg_ptr if memcpy
14559 should copy any arguments. Don't call gimple_call_num_args
14560 on id->call_stmt or call_stmt more than once.
14562 2021-06-07 liuhongt <hongtao.liu@intel.com>
14565 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): Refine
14567 (<insn>v4siv4di2): Delete constraints for define_expand.
14569 2021-06-07 liuhongt <hongtao.liu@intel.com>
14572 * config/i386/i386-expand.c (ix86_expand_builtin): Remove
14573 assignment of cfun->machine->has_explicit_vzeroupper.
14574 * config/i386/i386-features.c
14575 (ix86_add_reg_usage_to_vzerouppers): Delete.
14576 (ix86_add_reg_usage_to_vzeroupper): Ditto.
14577 (rest_of_handle_insert_vzeroupper): Remove
14578 ix86_add_reg_usage_to_vzerouppers, add df_analyze at the end
14580 (gate): Remove cfun->machine->has_explicit_vzeroupper.
14581 * config/i386/i386-protos.h (ix86_expand_avx_vzeroupper):
14583 * config/i386/i386.c (ix86_insn_callee_abi): New function.
14584 (ix86_initialize_callee_abi): Ditto.
14585 (ix86_expand_avx_vzeroupper): Ditto.
14586 (ix86_hard_regno_call_part_clobbered): Adjust for vzeroupper
14588 (TARGET_INSN_CALLEE_ABI): Define as ix86_insn_callee_abi.
14589 (ix86_emit_mode_set): Call ix86_expand_avx_vzeroupper
14591 * config/i386/i386.h (struct GTY(()) machine_function): Delete
14592 has_explicit_vzeroupper.
14593 * config/i386/i386.md (enum unspec): New member
14595 (ABI_DEFAULT,ABI_VZEROUPPER,ABI_UNKNOWN): New
14596 define_constants for insn callee abi index.
14597 * config/i386/predicates.md (vzeroupper_pattern): Adjust.
14598 * config/i386/sse.md (UNSPECV_VZEROUPPER): Deleted.
14599 (avx_vzeroupper): Call ix86_expand_avx_vzeroupper.
14600 (*avx_vzeroupper): Rename to ..
14601 (avx_vzeroupper_callee_abi): .. this, and adjust pattern as
14602 call_insn which has a special vzeroupper ABI.
14603 (*avx_vzeroupper_1): Deleted.
14605 2021-06-07 liuhongt <hongtao.liu@intel.com>
14608 * df-scan.c (df_get_call_refs): When call_insn is a fake call,
14609 it won't use stack pointer reg.
14610 * final.c (leaf_function_p): When call_insn is a fake call, it
14611 won't affect caller as a leaf function.
14612 * reg-stack.c (callee_clobbers_any_stack_reg): New.
14613 (subst_stack_regs): When call_insn doesn't clobber any stack
14614 reg, don't clear the arguments.
14615 * rtl.c (shallow_copy_rtx): Don't clear flag used when orig is
14617 * shrink-wrap.c (requires_stack_frame_p): No need for stack
14618 frame for a fake call.
14619 * rtl.h (FAKE_CALL_P): New macro.
14621 2021-06-06 Eric Botcazou <ebotcazou@adacore.com>
14623 * config/sparc/sparc-protos.h (order_regs_for_local_alloc): Rename
14625 (sparc_order_regs_for_local_alloc): ...this.
14626 (sparc_leaf_reg_remap): Declare.
14627 * config/sparc/sparc.h (ADJUST_REG_ALLOC_ORDER): Adjust.
14628 (LEAF_REG_REMAP): Reimplement as call to sparc_leaf_reg_remap.
14629 * config/sparc/sparc.c (leaf_reg_remap): Delete.
14630 (order_regs_for_local_alloc): Rename to...
14631 (sparc_order_regs_for_local_alloc): ...this.
14632 (sparc_leaf_reg_remap): New function.
14633 (sparc_conditional_register_usage): Do not modify leaf_reg_remap.
14635 2021-06-06 David Edelsohn <dje.gcc@gmail.com>
14637 * config/rs6000/rs6000.c (rs6000_xcoff_asm_output_aligned_decl_common):
14638 Use assemble_name to output BSS section name.
14640 2021-06-06 Uroš Bizjak <ubizjak@gmail.com>
14642 * config/i386/constraints.md (Bs):
14643 Remove boolean operators from match_test RTX.
14646 (M): Use "mode" variable instead of GET_MODE (op) in match_test RTX.
14649 2021-06-06 Martin Liska <mliska@suse.cz>
14651 * doc/extend.texi: Add missing @headitem.
14652 * doc/invoke.texi: Likewise.
14653 * doc/objc.texi: Likewise.
14655 2021-06-06 Martin Liska <mliska@suse.cz>
14657 * genhooks.c (emit_findices): Remove unused function.
14658 (emit_documentation): Do not call emit_findices
14659 and do not search for @Fcode directives.
14661 2021-06-06 Martin Liska <mliska@suse.cz>
14663 * doc/invoke.texi: Remove extra character.
14665 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
14667 * config/sh/sh.md (doloop_end_split): Fix empty split condition.
14669 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
14671 * config/sparc/sparc.md (*snedi<W:mode>_zero_vis3,
14672 *neg_snedi<W:mode>_zero_subxc, *plus_snedi<W:mode>_zero,
14673 *plus_plus_snedi<W:mode>_zero, *minus_snedi<W:mode>_zero,
14674 *minus_minus_snedi<W:mode>_zero): Fix empty split condition.
14676 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
14678 * config/or1k/or1k.md (*movdi): Fix empty split condition.
14680 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
14682 * config/mips/mips.md (<anonymous>, bswapsi2, bswapdi2): Fix empty
14685 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
14687 * config/m68k/m68k.md (*zero_extend_inc, *zero_extend_dec,
14688 *zero_extendsidi2): Fix empty split condition.
14690 2021-06-05 Jeff Law <jeffreyalaw@gmail.com>
14692 * config/h8300/addsub.md: Fix split condition in define_insn_and_split
14694 * config/h8300/bitfield.md: Likewise.
14695 * config/h8300/combiner.md: Likewise.
14696 * config/h8300/divmod.md: Likewise.
14697 * config/h8300/extensions.md: Likewise.
14698 * config/h8300/jumpcall.md: Likewise.
14699 * config/h8300/movepush.md: Likewise.
14700 * config/h8300/multiply.md: Likewise.
14701 * config/h8300/other.md: Likewise.
14702 * config/h8300/shiftrotate.md: Likewise.
14703 * config/h8300/logical.md: Likewise. Fix split pattern to use
14704 code iterator that somehow slipped through.
14706 2021-06-04 Tobias Burnus <tobias@codesourcery.com>
14708 PR middle-end/100905
14709 * tree-nested.c (convert_nonlocal_omp_clauses,
14710 convert_local_omp_clauses): Handle OMP_CLAUSE_BIND.
14712 2021-06-04 Martin Sebor <msebor@redhat.com>
14714 PR middle-end/100732
14715 * gimple-fold.c (gimple_fold_builtin_sprintf): Avoid folding calls
14716 with either source or destination argument of invalid type.
14717 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Avoid checking
14718 calls with arguments of invalid type.
14720 2021-06-04 Martin Sebor <msebor@redhat.com>
14722 * attribs.c (init_attr_rdwr_indices): Use VLA bounds in the expected
14724 (attr_access::vla_bounds): Also handle VLA bounds.
14726 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
14728 * config/i386/predicates.md (GOT_memory_operand):
14729 Implement using match_code RTXes.
14730 (GOT32_symbol_operand): Ditto.
14732 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
14735 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
14737 (ix86_expand_vector_init_general): Ditto.
14738 Use SImode instead of word_mode for logic operations
14739 when GET_MODE_SIZE (mode) < UNITS_PER_WORD.
14740 (expand_vec_perm_even_odd_1): Assert that V2HI mode should be
14741 implemented by expand_vec_perm_1.
14742 (expand_vec_perm_broadcast_1): Assert that V2HI and V4HI modes
14743 should be implemented using standard shuffle patterns.
14744 (ix86_vectorize_vec_perm_const): Handle V2HImode. Add V4HI and
14745 V2HI modes to modes, implementable with shuffle for one operand.
14746 * config/i386/mmx.md (*punpckwd): New insn_and_split pattern.
14747 (*pshufw_1): New insn pattern.
14748 (*vec_dupv2hi): Ditto.
14749 (vec_initv2hihi): New expander.
14751 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
14753 * config/arm/vfp.md (no_literal_pool_df_immediate,
14754 no_literal_pool_sf_immediate): Fix empty split condition.
14756 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
14758 * config/i386/i386.md (*load_tp_x32_zext, *add_tp_x32_zext,
14759 *tls_dynamic_gnu2_combine_32): Fix empty split condition.
14760 * config/i386/sse.md (*<sse2_avx2>_pmovmskb_lt,
14761 *<sse2_avx2>_pmovmskb_zext_lt, *sse2_pmovmskb_ext_lt,
14762 *<sse4_1_avx2>_pblendvb_lt): Likewise.
14764 2021-06-04 Jakub Jelinek <jakub@redhat.com>
14767 * config/i386/i386-expand.c (ix86_expand_vector_init): Handle
14768 concatenation from half-sized modes with TImode elements.
14770 2021-06-04 Claudiu Zissulescu <claziss@synopsys.com>
14772 * config/arc/arc.c (arc_override_options): Disable millicode
14773 thunks when RF16 is on.
14775 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
14777 * config/rs6000/rs6000.h (PROMOTE_MODE): Remove.
14779 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
14781 * config/rs6000/rs6000-call.c (rs6000_promote_function_mode):
14782 Replace PROMOTE_MODE marco with its content.
14784 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
14786 * config/cris/cris.md (*addi_reload): Fix empty split condition.
14788 2021-06-03 Jim Wilson <jimw@sifive.com>
14790 * config.gcc (riscv*-*-*): If --with-riscv-attribute not used,
14791 turn it on for all riscv targets.
14793 2021-06-03 Uroš Bizjak <ubizjak@gmail.com>
14796 * config/i386/i386-expand.c (ix86_expand_vector_set):
14797 Handle V2HI and V4QI modes.
14798 (ix86_expand_vector_extract): Ditto.
14799 * config/i386/mmx.md (*pinsrw): New insn pattern.
14802 (*pextrw_zext): Ditto.
14804 (*pextrb_zext): Ditto.
14805 (vec_setv2hi): New expander.
14806 (vec_extractv2hihi): Ditto.
14807 (vec_setv4qi): Ditto.
14808 (vec_extractv4qiqi): Ditto.
14809 (vec_setv8qi): Enable only for TARGET_SSE4_1.
14810 (vec_extractv8qiqi): Ditto.
14812 2021-06-03 Aaron Sawdey <acsawdey@linux.ibm.com>
14814 * config/rs6000/genfusion.pl (gen_logical_addsubf): Fix input
14815 order to subf instruction.
14816 * config/rs6000/fusion.md: Regenerate.
14818 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
14820 * calls.c (get_size_range): Use range_of_expr instead of
14821 determine_value_range.
14822 * tree-affine.c (expr_to_aff_combination): Same.
14823 * tree-data-ref.c (split_constant_offset): Same.
14824 * tree-vrp.c (determine_value_range_1): Remove.
14825 (determine_value_range): Remove.
14826 * tree-vrp.h (determine_value_range): Remove.
14828 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
14830 * function-tests.c (test_ranges): Call gimple_range_tests.
14831 * gimple-range-cache.cc (ranger_cache::range_of_expr): Pass stmt
14833 * gimple-range.cc (fur_source::get_operand): Do not call
14834 get_tree_range or gimple_range_global.
14836 (get_tree_range): Move to value-query.cc.
14837 Call get_arith_expr_range.
14838 (gimple_ranger::range_of_expr): Add argument to get_tree_range.
14839 Include gimple-range-tests.cc.
14840 * gimple-range.h (fold_range): Add argument.
14841 (get_tree_range): Remove.
14842 * selftest.h (gimple_range_tests): New.
14843 * value-query.cc (global_range_query::range_of_expr): Add
14845 (range_query::get_tree_range): Move from gimple-range.cc.
14846 * value-query.h (class range_query): Add get_tree_range and
14847 get_arith_expr_range. Make fur_source a friend.
14848 * vr-values.c (vr_values::range_of_expr): Pass stmt to
14850 * gimple-range-tests.cc: New file.
14852 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
14854 * gimple-range.cc (gimple_ranger::export_global_ranges): Call
14855 update_global_range.
14856 * value-query.cc (update_global_range): New.
14857 * value-query.h (update_global_range): New.
14859 2021-06-03 David Malcolm <dmalcolm@redhat.com>
14861 * diagnostic-show-locus.c (diagnostic_show_locus): Don't reject
14862 printing the same location twice if there are fix-it hints,
14863 multiple locations, or a label.
14865 2021-06-03 Andre Vieira <andre.simoesdiasvieira@arm.com>
14867 * tree-vect-loop.c (vect_transform_loop): Use main loop's various'
14868 thresholds to narrow the upper bound on epilogue iterations.
14870 2021-06-03 Christophe Lyon <christophe.lyon@linaro.org>
14872 * config/arm/mve.md (mve_vabsq_f<mode>): Use 'abs' instead of unspec.
14873 (mve_vabsq_s<mode>): Likewise.
14874 * config/arm/neon.md (abs<mode>2): Rename to neon_abs<mode>2.
14875 * config/arm/unspecs.md (VABSQ_F, VABSQ_S): Delete.
14876 * config/arm/vec-common.md (neg<mode>2): Rename to
14877 <absneg_str><mode>2.
14879 2021-06-03 Claudiu Zissulescu <claziss@synopsys.com>
14881 * common/config/arc/arc-common.c (arc_option_optimization_table):
14882 Remove malign-call.
14883 * config/arc/arc.c (arc_unalign_branch_p): Remove unused function.
14884 * config/arc/arc.h (TARGET_MIXED_CODE): Remove macro.
14885 (INDEX_REG_CLASS): Only refer to GENERAL_REGS.
14886 * config/arc/arc.md (abssi2_mixed): Remove pattern.
14887 * config/arc/arc.opt (munalign-prob-threshold): Mark it obsolete.
14888 (malign-call): Likewise.
14889 (mmixed-code): Likewise.
14890 * doc/invoke.texi (ARC): Update doc.
14892 2021-06-03 Martin Liska <mliska@suse.cz>
14894 * common.opt: Use proper Enum values.
14895 * opts.c (COVERAGE_SANITIZER_OPT): Remove.
14896 (parse_sanitizer_options): Handle only sanitizer_opts.
14897 (common_handle_option): Just assign value.
14899 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
14902 * tree-inline.c (inline_forbidden_p): Remove test on return type.
14904 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
14906 * dwarf2out.c (loc_list_from_tree_1) <FUNCTION_DECL>: Also generate
14907 DW_OP_GNU_variable_value referencing an existing DIE at file scope.
14908 (type_byte_size): Inline into...
14909 (add_byte_size_attribute): ...this and call add_scalar_info.
14911 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
14913 * dwarf2out.c (mem_loc_descriptor) <UDIV>: Fix typo.
14914 (typed_binop_from_tree): New function.
14915 (loc_list_from_tree_1) <EXACT_DIV_EXPR>: For an unsigned type,
14916 turn a divide by a power of 2 into a shift.
14917 <CEIL_DIV_EXPR>: For an unsigned type, use a signed divide if the
14918 size of the mode is lower than DWARF2_ADDR_SIZE; otherwise, do a
14919 typed divide by calling typed_binop_from_tree.
14921 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
14923 * dwarf2out.c (scompare_loc_descriptor): Fix head comment.
14924 (is_handled_procedure_type): Likewise.
14925 (struct loc_descr_context): Add strict_signedness field.
14926 (resolve_args_picking_1): Deal with DW_OP_[GNU_]deref_type,
14927 DW_OP_[GNU_]convert and DW_OP_[GNU_]reinterpret.
14928 (resolve_args_picking): Minor tweak.
14929 (function_to_dwarf_procedure): Initialize strict_signedness field.
14930 (type_byte_size): Likewise.
14931 (field_byte_offset): Likewise.
14932 (gen_descr_array_type_die): Likewise.
14933 (gen_variant_part): Likewise.
14934 (loc_list_from_tree_1) <CALL_EXPR>: Tidy up and set strict_signedness
14935 to true when a context is present before evaluating the arguments.
14936 <COND_EXPR>: Do not generate a useless comparison with zero.
14937 When dereferencing an address, if strict_signedness is true and the
14938 type is small and signed, use DW_OP_deref_type to do the dereference
14939 and then DW_OP_convert to convert back to the generic type.
14941 2021-06-03 Jakub Jelinek <jakub@redhat.com>
14944 * tree-inline.c (copy_tree_body_r): Handle iterators on
14945 OMP_CLAUSE_AFFINITY or OMP_CLAUSE_DEPEND.
14947 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
14949 * config/arc/arc.md (*bbit_di): Remove.
14951 2021-06-02 Christoph Muellner <cmuellner@gcc.gnu.org>
14953 PR rtl-optimization/100264
14954 * ree.c (get_sub_rtx): Ignore SET expressions without register
14955 destinations and remove assertion, as it is not valid anymore
14956 with this new behaviour.
14957 (merge_def_and_ext): Eliminate destination check for register
14958 as such SET expressions can't occur anymore.
14959 (combine_reaching_defs): Likewise.
14961 2021-06-02 Jakub Jelinek <jakub@redhat.com>
14964 * config/xtensa/xtensa.h (LEAF_REG_REMAP): Cast REGNO to int to avoid
14965 -Wtype-limits warnings.
14966 (DWARF_FRAME_REGISTER): Rewrite into ternary operator with addition
14967 in operands to avoid -Wsign-compare warnings.
14969 2021-06-02 Pat Haugen <pthaugen@linux.ibm.com>
14971 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): Use
14974 2021-06-02 Vineet Gupta <vgupta@synopsys.com>
14976 * config/arc/arc.h (TARGET_CPU_DEFAULT): Change to hs38_linux.
14978 2021-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
14980 * config/s390/s390.md(*ashrdi3_31<setcc><cconly>): Use a single
14982 * config/s390/subst.md(cconly_subst): Use a single constraint
14983 in (match_scratch).
14985 2021-06-02 Martin Liska <mliska@suse.cz>
14987 * ipa-icf.h: Use auto_vec for memory_access_types.
14989 2021-06-02 Jeff Law <jeffreyalaw@gmail.com>
14991 * config/h8300/h8300-protos.h (compute_a_shift_length): Drop unused
14992 argument from prototype.
14993 (output_logical_op): Add rtx_code argument.
14994 (compute_logical_op_length): Likewise.
14995 * config/h8300/h8300.c (h8300_and_costs): Pass additional argument
14996 to compute_a_shift_length.
14997 (output_logical_op); New argument with the rtx code rather than
14998 extracting it from an operand. Handle QImode too.
14999 (compute_logical_op_length): Similary.
15000 (compute_a_shift_length): Drop unused argument.
15001 * config/h8300/h8300.md (logicals): New code iterator.
15002 * config/h8300/logical.md (<code><mode>3 expander): Combine
15003 the "and" expander with the "ior"/"xor" expander.
15004 (bclr<mode>msx): Combine the QI/HI mode patterns.
15005 (<logical><mode>3 insns): Use code iterator rather than match_operator.
15006 Handle QImode as well. Update call to output_logical_op and
15007 compute_logical_op_length to pass in rtx_code
15008 Fix split condition on all define_insn_and_split patterns.
15009 (one_cmpl<mode>2<cczn>): Use <cczn> to support both clobbering
15010 the flags and setting ZN via existing define_subst.
15011 * config/h8300/shiftrotate.md: Drop unused argument from
15012 calls to compute_a_shift_length.
15013 Signed-off-by: Jeff Law <jeffreyalaw@gmail.com>
15015 2021-06-01 Andrew Pinski <apinski@marvell.com>
15017 PR tree-optimization/25290
15018 * tree-ssa-phiopt.c (match_simplify_replacement):
15020 (tree_ssa_phiopt_worker): Use match_simplify_replacement.
15021 (two_value_replacement): Change the comment about
15022 conditional_replacement.
15023 (conditional_replacement): Delete.
15025 2021-06-01 Andrew Pinski <apinski@marvell.com>
15027 PR tree-optimization/95481
15028 * tree-tailcall.c (find_tail_calls): Handle empty typed
15031 2021-06-01 Andrew Pinski <apinski@marvell.com>
15033 * gimplify.c (zero_sized_field_decl): Delete
15034 (zero_sized_type): Delete
15035 (gimplify_init_ctor_eval): Use is_empty_type instead
15036 of zero_sized_field_decl.
15037 (gimplify_modify_expr): Use is_empty_type instead of
15040 2021-06-01 Jason Merrill <jason@redhat.com>
15043 * tree.h (CALL_FROM_NEW_OR_DELETE_P): Adjust comment.
15045 2021-06-01 Jason Merrill <jason@redhat.com>
15048 * diagnostic.h (warning_enabled_at): Declare.
15049 * diagnostic.c (diagnostic_enabled): Factor out from...
15050 (diagnostic_report_diagnostic): ...here.
15051 (warning_enabled_at): New.
15053 2021-06-01 Aldy Hernandez <aldyh@redhat.com>
15055 * gimple-ssa-evrp.c: Enable exporting of global ranges.
15057 2021-06-01 Martin Liska <mliska@suse.cz>
15060 * doc/invoke.texi: Mention that -fgcse-after-reload
15061 is enabled with -O3.
15063 2021-06-01 liuhongt <hongtao.liu@intel.com>
15065 PR tree-optimization/98365
15066 * tree-if-conv.c (strip_nop_cond_scalar_reduction): New function.
15067 (is_cond_scalar_reduction): Handle nop_expr in cond scalar reduction.
15068 (convert_scalar_cond_reduction): Ditto.
15069 (predicate_scalar_phi): Ditto.
15071 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
15073 PR tree-optimization/100781
15074 * gimple-range-cache.cc (ranger_cache::ranger_cache): Enable new
15075 value calculation by default.
15076 (ranger_cache::enable_new_values): New.
15077 (ranger_cache::disable_new_values): New.
15078 (ranger_cache::push_poor_value): Check if new values are allowed.
15079 * gimple-range-cache.h (class ranger_cache): New member/methods.
15080 * gimple-range.cc (gimple_ranger::range_of_expr): Check for debug
15081 statement, and disable/renable new value calculation.
15083 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
15085 * gimple-range-cache.cc (ranger_cache::ssa_range_in_bb): Delete.
15086 (ranger_cache::range_of_def): New.
15087 (ranger_cache::entry_range): New.
15088 (ranger_cache::exit_range): New.
15089 (ranger_cache::range_of_expr): Adjust.
15090 (ranger_cache::range_on_edge): Adjust.
15091 (ranger_cache::propagate_cache): Call exit_range directly.
15092 * gimple-range-cache.h (class ranger_cache): Adjust.
15094 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
15096 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust for
15097 gori_compute being a member rather than base class.
15098 dervied call to member call.
15099 (ranger_cache::dump): No longer dump gori_map.
15100 (ranger_cache::dump_bb): New.
15101 (ranger_cache::get_non_stale_global_range): Adjust for gori_compute
15102 being a member rather than base class.
15103 (ranger_cache::set_global_range): Ditto.
15104 (ranger_cache::ssa_range_in_bb): Ditto.
15105 (ranger_cache::range_of_expr): New.
15106 (ranger_cache::range_on_edge): New.
15107 (ranger_cache::block_range): Adjust for gori_computes. Debug changes.
15108 (ranger_cache::propagate_cache): Adjust debugging output.
15109 (ranger_cache::fill_block_cache): Adjust for gori_computes. Debug
15111 * gimple-range-cache.h (class ranger_cache): Make gori_compute a
15112 member, and inherit from range_query instead.
15113 (ranger_cache::dump_bb): New. split from dump.
15114 * gimple-range-gori.cc (gori_compute::ssa_range_in_bb): Delete.
15115 (gori_compute::expr_range_at_stmt): Delete.
15116 (gori_compute::compute_name_range_op): Delete.
15117 (gori_compute::compute_operand_range_switch): Add fur_source.
15118 (gori_compute::compute_operand_range): Add fur_source param, inline
15119 old compute_name_range_op and optimize_logical_operands.
15120 (struct tf_range): Delete.
15121 (gori_compute::logical_combine): Adjust
15122 (gori_compute::optimize_logical_operands): Delete.
15123 (gori_compute::compute_logical_operands_in_chain): Delete.
15124 (gori_compute::compute_logical_operands): Adjust.
15125 (gori_compute::compute_operand1_range): Adjust to fur_source.
15126 (gori_compute::compute_operand2_range): Ditto.
15127 (gori_compute::compute_operand1_and_operand2_range): Ditto.
15128 (gori_compute::outgoing_edge_range_p): Add range_query parameter,
15129 and adjust to fur_source.
15130 * gimple-range-gori.h (class gori_compute): Simplify and adjust to
15131 range_query and fur_source.
15132 * gimple-range.cc (gimple_ranger::range_on_edge): Query range_on_edge
15133 from the ranger_cache..
15134 (gimple_ranger::fold_range_internal): Adjust to base class change of
15136 (gimple_ranger::dump_bb): Adjust dump.
15137 * gimple-range.h (gimple_ranger):export gori computes object.
15139 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
15141 PR tree-optimization/100774
15142 * gimple-range-cache.cc (ranger_cache::get_non_stale_global_range):
15143 Constant values are also not stale.
15144 (ranger_cache::set_global_range): Range invariant values should also
15145 have the correct timestamp.
15147 2021-05-31 Martin Liska <mliska@suse.cz>
15149 * tree-streamer-in.c (unpack_ts_function_decl_value_fields):
15150 Unpack FUNCTION_DECL_DECL_TYPE.
15151 * tree-streamer-out.c (pack_ts_function_decl_value_fields):
15152 Stream FUNCTION_DECL_DECL_TYPE instead of
15153 DECL_IS_OPERATOR_NEW_P.
15154 * tree.h (set_function_decl_type): Use FUNCTION_DECL_DECL_TYPE
15156 (DECL_IS_OPERATOR_NEW_P): Likewise.
15157 (DECL_IS_OPERATOR_DELETE_P): Likewise.
15158 (DECL_LAMBDA_FUNCTION_P): Likewise.
15160 2021-05-31 Richard Biener <rguenther@suse.de>
15163 * internal-fn.c (expand_SHUFFLEVECTOR): Define.
15164 * internal-fn.def (SHUFFLEVECTOR): New.
15165 * internal-fn.h (expand_SHUFFLEVECTOR): Declare.
15166 * doc/extend.texi: Document __builtin_shufflevector.
15168 2021-05-31 Peter Bergner <bergner@linux.ibm.com>
15171 * config/rs6000/predicates.md(mma_assemble_input_operand): Allow
15172 indexed form addresses.
15174 2021-05-29 Jeff Law <jlaw@tachyum.com>
15176 * config/h8300/h8300.c (h8300_emit_stack_adjustment): Drop unused
15177 parameter. Call callers fixed.
15179 (output_plussi): Add FALLTHRU markers.
15180 (h8300_shift_needs_scratch_p): Add gcc_unreachable marker.
15182 2021-05-29 Jakub Jelinek <jakub@redhat.com>
15184 PR middle-end/99928
15185 * gimplify.c (gimplify_scan_omp_clauses): For taskloop simd
15186 combined with parallel, make sure to add shared clause to
15187 parallel for explicit linear clause.
15189 2021-05-29 Aldy Hernandez <aldyh@redhat.com>
15191 PR tree-optimization/100787
15192 * gimple-ssa-evrp.c: Disable exporting of global ranges.
15194 2021-05-28 Jason Merrill <jason@redhat.com>
15196 * tree-iterator.h (struct tree_stmt_iterator): Add operator++,
15197 operator--, operator*, operator==, and operator!=.
15198 (class tsi_range): New.
15200 2021-05-28 Richard Biener <rguenther@suse.de>
15202 PR tree-optimization/100778
15203 * tree-vect-slp.c (vect_build_slp_tree_1): Prevent possibly
15204 trapping ops in different BBs.
15206 2021-05-28 Richard Biener <rguenther@suse.de>
15209 * tree-inline.c (copy_bb): When processing __builtin_va_arg_pack
15210 copy fntype from original call.
15212 2021-05-28 Martin Liska <mliska@suse.cz>
15214 PR gcov-profile/100751
15215 * doc/gcov.texi: Revert partially a hunk that was wrong.
15217 2021-05-28 Cooper Qu <cooper.qu@linux.alibaba.com>
15219 * config/csky/csky-linux-elf.h (HAVE_sync_compare_and_swapqi):
15221 (HAVE_sync_compare_and_swaphi): Likewise.
15222 (HAVE_sync_compare_and_swapsi): Likewise.
15224 2021-05-28 Jakub Jelinek <jakub@redhat.com>
15226 PR middle-end/99928
15227 * tree.h (OMP_CLAUSE_MAP_IMPLICIT): Define.
15229 2021-05-28 Tobias Burnus <tobias@codesourcery.com>
15231 * gimplify.c (gimplify_omp_affinity): New.
15232 (gimplify_scan_omp_clauses): Call it; remove affinity clause afterwards.
15233 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_AFFINITY.
15234 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_AFFINITY.
15235 * tree.c (omp_clause_num_ops, omp_clause_code_name): Add clause.
15236 (walk_tree_1): Handle OMP_CLAUSE_AFFINITY.
15238 2021-05-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
15239 Richard Biener <rguenther@suse.de>
15241 * match.pd <popcount & / + pattern matching>:
15242 When generating popcount directly fails, try doing it in two halves.
15244 2021-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
15246 * Makefile.in (generated_files): Add gimple-match.c and
15249 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
15251 * gensupport.c (alter_predicate_for_insn): Handle MATCH_DUP.
15253 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
15255 * gensupport.c (alter_constraints): Add MATCH_SCRATCH case.
15257 2021-05-28 Kewen Lin <linkw@linux.ibm.com>
15259 PR tree-optimization/99398
15260 * tree-ssa-forwprop.c (simplify_permutation): Optimize some cases
15261 where the fed operands are CTOR/CST and propagated through
15262 VIEW_CONVERT_EXPR. Call vec_perm_indices::new_shrunk_vector.
15263 * vec-perm-indices.c (vec_perm_indices::new_shrunk_vector): New
15265 * vec-perm-indices.h (vec_perm_indices::new_shrunk_vector): New
15268 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
15270 * config/i386/mmx.md (addv2sf3): Do not call
15271 ix86_fixup_binary_operands_no_copy.
15274 (<smaxmin:code>v2sf3): Ditto.
15275 (<plusminus:insn><MMXMODEI:mode>3): Ditto.
15276 (<plusminus:insn><VI_32:mode>3): Remove expander.
15277 (<plusminus:insn><VI_32:mode>3): Rename from
15278 "*<plusminus:insn><VI_32:mode>3".
15279 (mulv4hi): Do not call ix86_fixup_binary_operands_no_copy.
15280 (mulv2hi3): Remove expander.
15281 (mulv2hi3): Rename from *mulv2hi3.
15282 (<s>mulv2hi3_highpart): Remove expander.
15283 (<s>mulv2hi3_highpart): Rename from *<s>mulv2hi3_highpart.
15284 (<smaxmin:code><MMXMODE14:mode>3): Rename from
15285 "*mmx_<smaxmin:code><MMXMODE14:mode>3".
15286 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Remove expander.
15287 (SMAXMIN_MMXMODEI): Remove mode iterator.
15288 (<smaxmin:code>v4hi3): New expander.
15289 (<smaxmin:code>v4qi3): Rename from *<smaxmin:code>v4qi3.
15290 (<smaxmin:code>v2hi3): Rename from *<smaxmin:code>v2hi3.
15291 (<smaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
15292 (SMAXMIN_VI_32): Remove mode iterator.
15293 (<umaxmin:code><MMXMODE24:mode>3): Rename from
15294 "*mmx_<umaxmin:code><MMXMODE24:mode>3".
15295 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Remove expander.
15296 (UMAXMIN_MMXMODEI): Remove mode iterator.
15297 (<umaxmin:code>v8qi3): New expander.
15298 (<umaxmin:code>v4qi3): Rename from *<umaxmin:code>v4qi3.
15299 (<umaxmin:code>v2hi3): Rename from *<umaxmin:code>v2hi3.
15300 (<umaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
15301 (UMAXMIN_VI_32): Remove mode iterator.
15302 (<any_shift:insn>v2hi3): Remove expander.
15303 (<any_shift:insn>v2hi3): Rename from *<any_shift:insn>v2hi3.
15304 (<any_logic:code><MMXMODEI:mode>3): Do not call
15305 ix86_fixup_binary_operands_no_copy.
15306 (<any_logic:code><VI_32:mode>3): Remove expander.
15307 (<any_logic:code><VI_32:mode>3): Rename from
15308 "*<any_logic:code><VI_32:mode>3".
15309 (uavg<mode>3_ceil): Do not call ix86_fixup_binary_operands_no_copy.
15310 * config/i386/sse.md (div<VF2:mode>3): Do not call
15311 ix86_fixup_binary_operands_no_copy.
15312 (div<VF1:mode>3): Ditto.
15313 (<maxmin:code><VI8_AVX2_AVX512F:mode>3): Ditto.
15314 (smulhrsv4hi3): Ditto.
15315 (smulhrsv2hi3): Ditto.
15317 2021-05-27 Martin Sebor <msebor@redhat.com>
15319 * ggc.h (gt_ggc_mx): Add overloads for all integers.
15321 * hash-map.h (class hash_map): Add pch_nx_helper overloads for all
15323 (hash_map::operator==): New function.
15325 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
15328 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
15329 For TARGET_XOP bypass SSE comparisons for all supported vector modes.
15330 * config/i386/mmx.md (*xop_maskcmp<MMXMODEI:mode>3): New insn pattern.
15331 (*xop_maskcmp<VI_32:mode>3): Ditto.
15332 (*xop_maskcmp_uns<MMXMODEI:mode>3): Ditto.
15333 (*xop_maskcmp_uns<VI_32:mode>3): Ditto.
15335 2021-05-27 Richard Earnshaw <rearnsha@arm.com>
15338 * config/arm/arm.c (arm_configure_build_target): Remove parameter
15339 opts_set, directly check opts parameters for being non-null.
15340 (arm_option_restore): Update call to arm_configure_build_target.
15341 (arm_option_override): Likewise.
15342 (arm_can_inline_p): Likewise.
15343 (arm_valid_target_attribute_tree): Likewise.
15344 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
15345 * config/arm/arm-protos.h (arm_configure_build_target): Adjust
15348 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
15350 * vr-values.c (simplify_conversion_using_ranges): Use
15351 get_range_query instead of get_global_range_query.
15353 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
15355 * gimple-range.cc (get_range_global): Move to value-query.cc.
15356 (gimple_range_global): Same.
15357 (get_global_range_query): Same.
15358 (global_range_query::range_of_expr): Same.
15359 * gimple-range.h (class global_range_query): Move to
15361 (gimple_range_global): Same.
15362 * tree-ssanames.c (get_range_info): Move to value-query.cc.
15363 (get_ptr_nonnull): Same.
15364 * tree-ssanames.h (get_range_info): Remove.
15365 (get_ptr_nonnull): Remove.
15366 * value-query.cc (get_ssa_name_range_info): Move from
15368 (get_ssa_name_ptr_info_nonnull): Same.
15369 (get_range_global): Move from gimple-range.cc.
15370 (gimple_range_global): Same.
15371 (get_global_range_query): Same.
15372 (global_range_query::range_of_expr): Same.
15373 * value-query.h (class global_range_query): Move from
15375 (gimple_range_global): Same.
15377 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
15380 * config/i386/mmx.md (uavgv4qi3_ceil): New insn pattern.
15381 (uavgv2hi3_ceil): Ditto.
15383 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
15386 * doc/extend.texi (scalar_storage_order): Rephrase slightly.
15388 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
15390 * tree-ssanames.c (get_range_info): Merge both copies of
15391 get_range_info into one that works with irange.
15392 * tree-ssanames.h (get_range_info): Remove version that works on
15395 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
15397 * builtins.c (check_nul_terminated_array): Convert to get_range_query.
15398 (expand_builtin_strnlen): Same.
15399 (determine_block_size): Same.
15400 * fold-const.c (expr_not_equal_to): Same.
15401 * gimple-fold.c (size_must_be_zero_p): Same.
15402 * gimple-match-head.c: Include gimple-range.h.
15403 * gimple-pretty-print.c (dump_ssaname_info): Convert to get_range_query.
15404 * gimple-ssa-warn-restrict.c
15405 (builtin_memref::extend_offset_range): Same.
15406 * graphite-sese-to-poly.c (add_param_constraints): Same.
15407 * internal-fn.c (get_min_precision): Same.
15408 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): Same.
15409 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Same.
15411 * tree-data-ref.c (split_constant_offset): Same.
15412 (dr_step_indicator): Same.
15413 * tree-dfa.c (get_ref_base_and_extent): Same.
15414 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
15415 * tree-ssa-loop-niter.c (refine_value_range_using_guard): Same.
15416 (determine_value_range): Same.
15417 (record_nonwrapping_iv): Same.
15418 (infer_loop_bounds_from_signedness): Same.
15419 (scev_var_range_cant_overflow): Same.
15420 * tree-ssa-phiopt.c (two_value_replacement): Same.
15421 * tree-ssa-pre.c (insert_into_preds_of_block): Same.
15422 * tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Same.
15423 * tree-ssa-strlen.c (handle_builtin_stxncpy_strncat): Same.
15425 (dump_strlen_info): Same.
15426 (set_strlen_range): Same.
15427 (maybe_diag_stxncpy_trunc): Same.
15428 (get_len_or_size): Same.
15429 (handle_integral_assign): Same.
15430 * tree-ssa-structalias.c (find_what_p_points_to): Same.
15431 * tree-ssa-uninit.c (find_var_cmp_const): Same.
15432 * tree-switch-conversion.c (bit_test_cluster::emit): Same.
15433 * tree-vect-patterns.c (vect_get_range_info): Same.
15434 (vect_recog_divmod_pattern): Same.
15435 * tree-vrp.c (intersect_range_with_nonzero_bits): Same.
15436 (register_edge_assert_for_2): Same.
15437 (determine_value_range_1): Same.
15438 * tree.c (get_range_pos_neg): Same.
15439 * vr-values.c (vr_values::get_lattice_entry): Same.
15440 (vr_values::update_value_range): Same.
15441 (simplify_conversion_using_ranges): Same.
15443 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
15445 * gimple-ssa-warn-alloca.c (alloca_call_type): Use
15446 get_range_query instead of query argument.
15447 (pass_walloca::execute): Enable and disable global ranger.
15449 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
15451 * gimple-ssa-evrp.c (rvrp_folder::rvrp_folder): Call
15453 (rvrp_folder::~rvrp_folder): Call disable_ranger.
15454 (hybrid_folder::hybrid_folder): Call enable_ranger.
15455 (hybrid_folder::~hybrid_folder): Call disable_ranger.
15457 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
15459 * function.c (allocate_struct_function): Set cfun->x_range_query.
15460 * function.h (struct function): Declare x_range_query.
15461 (get_range_query): New.
15462 (get_global_range_query): New.
15463 * gimple-range-cache.cc (ssa_global_cache::ssa_global_cache):
15464 Remove call to safe_grow_cleared.
15465 * gimple-range.cc (get_range_global): New.
15466 (gimple_range_global): Move from gimple-range.h.
15467 (get_global_range_query): New.
15468 (global_range_query::range_of_expr): New.
15469 (enable_ranger): New.
15470 (disable_ranger): New.
15471 * gimple-range.h (gimple_range_global): Move to gimple-range.cc.
15472 (class global_range_query): New.
15473 (enable_ranger): New.
15474 (disable_ranger): New.
15475 * gimple-ssa-evrp.c (evrp_folder::~evrp_folder): Rename
15476 dump_all_value_ranges to dump.
15477 * tree-vrp.c (vrp_prop::finalize): Same.
15478 * value-query.cc (range_query::dump): New.
15479 * value-query.h (range_query::dump): New.
15480 * vr-values.c (vr_values::dump_all_value_ranges): Rename to...
15481 (vr_values::dump): ...this.
15482 * vr-values.h (class vr_values): Rename dump_all_value_ranges to
15483 dump and make virtual.
15485 2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
15487 * config/i386/i386.c (ix86_autovectorize_vector_modes):
15488 Add V4QImode and V16QImode for TARGET_SSE2.
15489 * doc/sourcebuild.texi (Vector-specific attributes):
15490 Add vect64 and vect32 description.
15492 2021-05-26 Bernd Edlinger <bernd.edlinger@hotmail.de>
15494 * gimple-range-gori.cc (range_def_chain::register_dependency):
15495 Resize m_def_chain when needed.
15497 2021-05-26 Christophe Lyon <christophe.lyon@linaro.org>
15499 * config/arm/mve.md (mve_vaddvq_<supf><mode>): Prefix with '@'.
15500 * config/arm/neon.md (reduc_plus_scal_<mode>): Move to ..
15501 * config/arm/vec-common.md: .. here. Add support for MVE.
15503 2021-05-26 Jakub Jelinek <jakub@redhat.com>
15505 * config/epiphany/epiphany.c (epiphany_print_operand_address): Remove
15507 * config/microblaze/microblaze.c (microblaze_legitimize_address,
15509 microblaze_option_override, print_operand): Likewise.
15510 * config/microblaze/microblaze.md (call_internal_plt,
15511 call_value_intern_plt, call_value_intern): Likewise.
15512 * config/arm/aout.h (ASM_OUTPUT_ALIGN): Likewise.
15513 * config/iq2000/iq2000.md (call_internal1, call_value_internal1,
15514 call_value_multiple_internal1): Likewise.
15515 * config/bfin/bfin.c (symbolic_reference_mentioned_p): Likewise.
15517 2021-05-26 Jan-Benedict Glaw <jbglaw@lug-owl.de>
15519 * config/arc/arc.c (arc_address_cost, arc_print_operand_address,
15520 arc_ccfsm_advance, symbolic_reference_mentioned_p,
15521 arc_raw_symbolic_reference_mentioned_p): Remove register
15524 2021-05-26 Jakub Jelinek <jakub@redhat.com>
15527 * omp-low.c: Include omp-offload.h.
15528 (create_omp_child_function): If current_function_decl has
15529 "omp declare target" attribute and is_gimple_omp_offloaded,
15530 remove that attribute from the copy of attribute list and
15531 add "omp target entrypoint" attribute instead.
15532 (lower_omp_target): Mark .omp_data_sizes.* and .omp_data_kinds.*
15533 variables for offloading if in omp_maybe_offloaded_ctx.
15534 * omp-offload.c (pass_omp_target_link::execute): Nullify second
15535 argument to GOMP_target_data_ext in offloaded code.
15537 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
15539 * config/csky/csky.c (csky_can_change_mode_class): Delete.
15540 For csky, HF/SF mode use the low bits of VREGS.
15542 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
15544 * gimplify.c (gimplify_decl_expr): Do not clear TREE_READONLY on a
15545 DECL which is a reference for OMP.
15547 2021-05-26 Martin Liska <mliska@suse.cz>
15549 PR gcov-profile/100751
15550 * doc/gcov.texi: Document that __gcov_dump can be called just
15551 once and that __gcov_reset resets run-time counters.
15553 2021-05-26 Martin Liska <mliska@suse.cz>
15555 * doc/install.texi: Port relevant part from install-old.texi
15556 and re-generate list of CPUs and systems.
15558 2021-05-26 Martin Liska <mliska@suse.cz>
15560 * Makefile.in: Remove it.
15561 * doc/include/fdl.texi: Update next/previous chapters.
15562 * doc/install.texi: Likewise.
15563 * doc/install-old.texi: Removed.
15565 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
15567 * config/csky/csky.c (ck810_legitimate_index_p): Support
15568 "base + index" with DF mode.
15569 * config/csky/constraints.md ("Y"): New constraint for memory operands
15570 without index register.
15571 * config/csky/csky_insn_fpuv2.md (fpuv3_movdf): Use "Y" instead of "m"
15572 when mov between memory and general registers, and lower their priority.
15573 * config/csky/csky_insn_fpuv3.md (fpuv2_movdf): Likewise.
15575 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
15577 * config/csky/csky.c (TARGET_PROMOTE_PROTOTYPES): Delete.
15579 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
15581 * config/csky/csky.md (untyped_call): Emit clobber for return
15582 registers to mark them used.
15584 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
15586 * config/csky/csky.md (cskyv2_sextend_ldbs): New.
15588 2021-05-26 Andrew Pinski <apinski@marvell.com>
15590 * match.pd (x < 0 ? ~y : y): New patterns.
15592 2021-05-26 Andrew Pinski <apinski@marvell.com>
15594 * match.pd (A?CST1:CST2): Add simplifcations for A?0:+-1, A?+-1:0,
15595 A?POW2:0 and A?0:POW2.
15597 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15599 * gimple-range-gori.cc (class logical_stmt_cache): Delete
15600 (logical_stmt_cache::logical_stmt_cache ): Delete.
15601 (logical_stmt_cache::~logical_stmt_cache): Delete.
15602 (logical_stmt_cache::cache_entry::dump): Delete.
15603 (logical_stmt_cache::get_range): Delete.
15604 (logical_stmt_cache::cached_name ): Delete.
15605 (logical_stmt_cache::same_cached_name): Delete.
15606 (logical_stmt_cache::cacheable_p): Delete.
15607 (logical_stmt_cache::slot_diagnostics ): Delete.
15608 (logical_stmt_cache::dump): Delete.
15609 (gori_compute_cache::gori_compute_cache): Delete.
15610 (gori_compute_cache::~gori_compute_cache): Delete.
15611 (gori_compute_cache::compute_operand_range): Delete.
15612 (gori_compute_cache::cache_stmt): Delete.
15613 * gimple-range-gori.h (gori_compute::compute_operand_range): Remove
15615 (class gori_compute_cache): Delete.
15617 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15619 * gimple-range.cc (fold_using_range::range_of_range_op): Use m_gori
15621 (fold_using_range::range_of_address): Adjust.
15622 (fold_using_range::range_of_phi): Adjust.
15623 * gimple-range.h (class fur_source): Adjust.
15624 (fur_source::fur_source): Adjust.
15626 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15628 * gimple-range-gori.cc (gori_compute::expr_range_at_stmt): Rename
15629 from expr_range_in_bb and adjust.
15630 (gori_compute::compute_name_range_op): Adjust.
15631 (gori_compute::optimize_logical_operands): Adjust.
15632 (gori_compute::compute_logical_operands_in_chain): Adjust.
15633 (gori_compute::compute_operand1_range): Adjust.
15634 (gori_compute::compute_operand2_range): Adjust.
15635 (ori_compute_cache::cache_stmt): Adjust.
15636 * gimple-range-gori.h (gori_compute): Rename prototype.
15638 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15640 * gimple-range.cc (gimple_ranger::range_of_expr): Non-null should be
15641 checked only after range_of_stmt, not range_on_entry.
15642 (gimple_ranger::range_on_entry): Check for non-null in any
15643 predecessor block, if it is not already non-null.
15644 (gimple_ranger::range_on_exit): DOnt check for non-null after
15645 range on entry call.
15646 (gimple_ranger::dump_bb): New. Split from dump.
15647 (gimple_ranger::dump): Adjust.
15648 * gimple-range.h (class gimple_ranger): Adjust.
15650 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15652 * gimple-range-cache.cc (struct range_timestamp): Delete.
15653 (class temporal_cache): Adjust.
15654 (temporal_cache::get_timestamp): Delete.
15655 (temporal_cache::set_dependency): Delete.
15656 (temporal_cache::temporal_value): Adjust.
15657 (temporal_cache::current_p): Take dependencies as params.
15658 (temporal_cache::set_timestamp): Adjust.
15659 (temporal_cache::set_always_current): Adjust.
15660 (ranger_cache::get_non_stale_global_range): Adjust.
15661 (ranger_cache::register_dependency): Delete.
15662 * gimple-range-cache.h (class range_cache): Adjust.
15664 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15666 * gimple-range-gori.cc (range_def_chain::range_def_chain): init
15668 (range_def_chain::~range_def_chain): Dispose of obstack rather than
15669 each individual bitmap.
15670 (range_def_chain::set_import): New.
15671 (range_def_chain::get_imports): New.
15672 (range_def_chain::chain_import_p): New.
15673 (range_def_chain::register_dependency): Rename from build_def_chain
15675 (range_def_chain::def_chain_in_bitmap_p): New.
15676 (range_def_chain::add_def_chain_to_bitmap): New.
15677 (range_def_chain::has_def_chain): Just check first depenedence.
15678 (range_def_chain::get_def_chain): Process imports, use generic
15679 register_dependency routine.
15680 (range_def_chain::dump): New.
15681 (gori_map::gori_map): Allocate import list.
15682 (gori_map::~gori_map): Release imports.
15683 (gori_map::exports): Check for past allocated block size.
15684 (gori_map::imports): New.
15685 (gori_map::def_chain_in_export_p): Delete.
15686 (gori_map::is_import_p): New.
15687 (gori_map::maybe_add_gori): Handle imports.
15688 (gori_map::dump): Adjust output, add imports.
15689 (gori_compute::has_edge_range_p): Remove def_chain_in_export call.
15690 (gori_export_iterator::gori_export_iterator): New.
15691 (gori_export_iterator::next): New.
15692 (gori_export_iterator::get_name): New.
15693 * gimple-range-gori.h (range_def_chain): Add imports and direct
15694 dependecies via struct rdc.
15695 (range_def_chain::depend1): New.
15696 (range_def_chain::depend2): New.
15697 (class gori_map): Adjust.
15698 (FOR_EACH_GORI_IMPORT_NAME): New.
15699 (FOR_EACH_GORI_EXPORT_NAME): New.
15700 (class gori_export_iterator): New.
15702 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15704 * gimple-range-cache.cc (ranger_cache::ranger_cache): Move initial
15705 export cache filling to here.
15706 * gimple-range-gori.cc (gori_compute::gori_compute) : From Here.
15708 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15710 * gimple-range-gori.cc (range_def_chain): Move to gimple-range-gori.h.
15711 (gori_map): Move to gimple-range-gori.h.
15712 (gori_compute::gori_compute): Adjust.
15713 (gori_compute::~gori_compute): Delete.
15714 (gori_compute::compute_operand_range_switch): Adjust.
15715 (gori_compute::compute_operand_range): Adjust.
15716 (gori_compute::compute_logical_operands): Adjust.
15717 (gori_compute::has_edge_range_p ): Adjust.
15718 (gori_compute::set_range_invariant): Delete.
15719 (gori_compute::dump): Adjust.
15720 (gori_compute::outgoing_edge_range_p): Adjust.
15721 * gimple-range-gori.h (class range_def_chain): Relocate here.
15722 (class gori_map): Relocate here.
15723 (class gori_compute): Inherit from gori_map, and adjust.
15725 2021-05-25 Aldy Hernandez <aldyh@redhat.com>
15727 * value-range.cc (range_tests_legacy): Use
15728 build_nonstandard_integer_type instead of int and short.
15730 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
15732 * gimplify.c (gimplify_decl_expr): Clear TREE_READONLY on the DECL
15733 when really creating an initialization statement for it.
15735 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
15737 * tree-inline.c (setup_one_parameter): Fix thinko in new condition.
15739 2021-05-25 Kito Cheng <kito.cheng@sifive.com>
15741 * config/riscv/riscv.h (ASM_SPEC): Pass -mno-relax.
15743 2021-05-25 Martin Liska <mliska@suse.cz>
15745 PR tree-optimization/92860
15747 * optc-save-gen.awk: Remove exceptions.
15749 2021-05-25 Martin Liska <mliska@suse.cz>
15751 * asan.h (sanitize_coverage_p): New function.
15752 * doc/extend.texi: Document it.
15753 * fold-const.c (fold_range_test): Use sanitize_flags_p
15754 instead of flag_sanitize_coverage.
15755 (fold_truth_andor): Likewise.
15756 * sancov.c: Likewise.
15757 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
15758 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Handle
15759 -fsanitize-coverage when inlining.
15761 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
15763 * config/csky/csky-modes.def : Fix copyright.
15765 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
15767 * config/csky/csky-modes.def : Amend copyright.
15768 * config/csky/csky_insn_fpuv2.md : Likewise.
15769 * config/csky/csky_insn_fpuv3.md : Likewise.
15771 2021-05-25 Richard Biener <rguenther@suse.de>
15773 PR middle-end/100727
15774 * calls.c (initialize_argument_information): Explicitely test
15775 for WITH_SIZE_EXPR.
15776 * gimple-expr.c (mark_addressable): Skip outer WITH_SIZE_EXPR.
15778 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
15780 * config/csky/csky.h (FRAME_POINTER_REGNUM): Use
15781 HARD_FRAME_POINTER_REGNUM and FRAME_POINTER_REGNUM instead of
15782 the signle definition. The signle definition may not work well
15783 at simplify_subreg_regno().
15784 (HARD_FRAME_POINTER_REGNUM): New.
15785 (ELIMINABLE_REGS): Add for HARD_FRAME_POINTER_REGNUM.
15786 * config/csky/csky.c (get_csky_live_regs, csky_can_eliminate,
15787 csky_initial_elimination_offset, csky_expand_prologue,
15788 csky_expand_epilogue): Add for HARD_FRAME_POINTER_REGNUM.
15790 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
15792 * config/csky/csky.c (csky_option_override):
15793 Init csky_arch_isa_features[] in advance, so TARGET_DSP
15794 and TARGET_DIV can be set well.
15796 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
15798 * config/csky/constraints.md ("l", "h"): Delete.
15799 * config/csky/csky.h (reg_class, REG_CLASS_NAMES,
15800 REG_CLASS_CONTENTS): Delete LO_REGS and HI_REGS.
15801 * config/csky/csky.c (regno_reg_classm,
15802 csky_secondary_reload, csky_register_move_cost):
15803 Use HILO_REGS instead of LO_REGS and HI_REGS.
15805 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
15807 * config/csky/constraints.md ("W"): New constriant for mem operand
15808 with base reg, index register.
15809 ("Q"): Renamed and modified "csky_valid_fpuv2_mem_operand" to
15810 "csky_valid_mem_constraint_operand" to deal with both "Q" and "W"
15812 ("Dv"): New constraint for const double value that can be used at
15814 * config/csky/csky-modes.def (HFmode): New mode.
15815 * config/csky/csky-protos.h (csky_valid_fpuv2_mem_operand): Rename
15816 to "csky_valid_mem_constraint_operand" and support new constraint
15818 (csky_get_movedouble_length): New.
15819 (fpuv3_output_move): New.
15820 (fpuv3_const_double): New.
15821 * config/csky/csky.c (csky_option_override): New arch CK860 with fpv3.
15822 (decompose_csky_address): Refine.
15823 (csky_print_operand): New "CONST_DOUBLE" operand.
15824 (csky_output_move): Support fpv3 instructions.
15825 (csky_get_movedouble_length): New.
15826 (fpuv3_output_move): New.
15827 (fpuv3_const_double): New.
15828 (csky_emit_compare): Cover float comparsion.
15829 (csky_emit_compare_float): Refine.
15830 (csky_vaild_fpuv2_mem_operand): Rename to
15831 "csky_valid_mem_constraint_operand" and support new constraint "W".
15832 (ck860_rtx_costs): New.
15833 (csky_rtx_costs): Add the cost calculation of CK860.
15834 (regno_reg_class): New vregs for fpuv3.
15835 (csky_dbx_regno): Likewise.
15836 (csky_cpu_cpp_builtins): New builtin macro for fpuv3.
15837 (csky_conditional_register_usage): Suporrot fpuv3.
15838 (csky_dwarf_register_span): Suporrot fpuv3.
15839 (csky_init_builtins, csky_mangle_type): Support "__fp16" type.
15840 (ck810_legitimate_index_p): Support fp16.
15841 * config/csky/csky.h (TARGET_TLS): ADD CK860.
15842 (CSKY_VREG_P, CSKY_VREG_LO_P, CSKY_VREG_HI_P): Support fpuv3.
15843 (TARGET_SINGLE_FPU): Support fpuv3.
15844 (TARGET_SUPPORT_FPV3): New.
15845 (FIRST_PSEUDO_REGISTER): Change to 202 to hold the new fpuv3 registers.
15846 (FIXED_REGISTERS, CALL_REALLY_USED_REGISTERS, REGISTER_NAMES,
15847 REG_CLASS_CONTENTS): Support fpuv3.
15848 * config/csky/csky.md (movsf): Move to cksy_insn_fpu.md and refine.
15849 (csky_movsf_fpv2): Likewise.
15850 (ck801_movsf): Likewise.
15851 (csky_movsf): Likewise.
15853 (csky_movdf_fpv2): Likewise.
15854 (ck801_movdf): Likewise.
15855 (csky_movdf): Likewise.
15856 (movsicc): Refine. Use "comparison_operatior" instead of
15857 "ordered_comparison_operatior".
15858 (addsicc): Likewise.
15859 (CSKY_FIRST_VFP3_REGNUM, CSKY_LAST_VFP3_REGNUM): New constant.
15860 (call_value_internal_vh): New.
15861 * config/csky/csky_cores.def (CK860): New arch and cpu.
15866 * config/csky/csky_insn_fpu.md: Refactor. Separate all float patterns
15867 into emit-patterns and match-patterns, remain the emit-patterns here,
15868 and move the match-patterns to csky_insn_fpuv2.md or
15869 csky_insn_fpuv3.md.
15870 * config/csky/csky_insn_fpuv2.md: New file for fpuv2 instructions.
15871 * config/csky/csky_insn_fpuv3.md: New file and new patterns for fpuv3
15873 * config/csky/csky_isa.def (fcr): New.
15878 (CK860): New definition for ck860.
15879 * config/csky/csky_tables.opt (ck860): New processors ck860,
15880 ck860f. And new arch ck860.
15885 * config/csky/predicates.md (csky_float_comparsion_operator): Delete
15886 "geu", "gtu", "leu", "ltu", which will never appear at float comparison.
15887 * config/csky/t-csky-elf: Support 860.
15888 * config/csky/t-csky-linux: Likewise.
15889 * doc/md.texi: Add "Q" and "W" constraints for C-SKY.
15891 2021-05-24 Aaron Sawdey <acsawdey@linux.ibm.com>
15893 * config/rs6000/genfusion.pl (gen_logical_addsubf): Refactor to
15894 add generation of logical-add and add-logical fusion pairs.
15895 * config/rs6000/rs6000-cpus.def: Add new fusion to ISA 3.1 mask
15897 * config/rs6000/rs6000.c (rs6000_option_override_internal): Turn on
15898 logical-add and add-logical fusion by default.
15899 * config/rs6000/rs6000.opt: Add -mpower10-fusion-logical-add and
15900 -mpower10-fusion-add-logical options.
15901 * config/rs6000/fusion.md: Regenerate file.
15903 2021-05-24 Aldy Hernandez <aldyh@redhat.com>
15905 * value-range.cc (irange::legacy_equal_p): Check type when
15906 comparing VR_VARYING types.
15907 (range_tests_legacy): Test comparing VARYING ranges of different
15910 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
15912 * config/aarch64/aarch64.c (neoversen1_tunings):
15913 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
15915 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
15917 * config/aarch64/aarch64.c (aarch64_classify_symbol): Use GOT for
15918 extern weak symbols. Limit symbol offsets for non-GOT symbols with
15921 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
15923 * config/arm/neon.md (vec_load_lanesxi<mode>)
15924 (vec_store_lanexoi<mode>): Move ...
15925 * config/arm/vec-common.md: here.
15927 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
15929 * config/arm/neon.md (vec_load_lanesoi<mode>)
15930 (vec_store_lanesoi<mode>): Move ...
15931 * config/arm/vec-common.md: here.
15933 2021-05-24 liuhongt <hongtao.liu@intel.com>
15936 * config/i386/i386.c (ix86_gimple_fold_builtin): Replacing
15937 stmt with GIMPLE_NOP when lhs doesn't exist.
15939 2021-05-23 Uroš Bizjak <ubizjak@gmail.com>
15942 * config/i386/mmx.md (*push<VI_32:mode>2_rex64):
15943 New instruction pattern.
15944 (*push<VI_32:mode>2): Ditto.
15945 (push splitter for SSE registers): New splitter.
15947 2021-05-23 Andrew Pinski <apinski@marvell.com>
15949 * match.pd ((A & C) != 0 ? D : 0): Limit to non pointer types.
15951 2021-05-22 Aaron Sawdey <acsawdey@linux.ibm.com>
15953 * config/rs6000/genfusion.pl (gen_addadd): Fix incorrect attr types.
15954 * config/rs6000/fusion.md: Regenerate file.
15956 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
15958 * config/rs6000/genfusion.pl (gen_addadd): New function.
15959 * config/rs6000/fusion.md: Regenerate file.
15960 * config/rs6000/rs6000-cpus.def: Add
15961 OPTION_MASK_P10_FUSION_2ADD to masks.
15962 * config/rs6000/rs6000.c (rs6000_option_override_internal):
15963 Handle default value of OPTION_MASK_P10_FUSION_2ADD.
15964 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2add.
15966 2021-05-21 Jakub Jelinek <jakub@redhat.com>
15968 PR middle-end/99928
15969 * tree.h (OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET): Define.
15970 * gimplify.c (enum gimplify_omp_var_data): Fix up
15971 GOVD_MAP_HAS_ATTACHMENTS value, add GOVD_FIRSTPRIVATE_IMPLICIT.
15972 (omp_lastprivate_for_combined_outer_constructs): If combined target
15973 has GOVD_FIRSTPRIVATE_IMPLICIT set for the decl, change it to
15974 GOVD_MAP | GOVD_SEEN.
15975 (gimplify_scan_omp_clauses): Set GOVD_FIRSTPRIVATE_IMPLICIT for
15976 firstprivate clauses with OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT.
15977 (gimplify_adjust_omp_clauses): For firstprivate clauses with
15978 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT either clear that bit and
15979 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET too, or remove it and
15980 let it be replaced by implicit map clause.
15982 2021-05-21 Jakub Jelinek <jakub@redhat.com>
15984 PR middle-end/99928
15985 * gimplify.c (omp_lastprivate_for_combined_outer_constructs): New
15987 (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LASTPRIVATE>: Use it.
15988 (gimplify_omp_for): Likewise.
15990 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15992 PR middle-end/90115
15993 * omp-low.c (oacc_privatization_candidate_p): Reject 'static',
15994 'external' in blocks.
15996 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15998 PR middle-end/90115
15999 * flag-types.h (enum openacc_privatization): New.
16000 * params.opt (-param=openacc-privatization): New.
16001 * doc/invoke.texi (openacc-privatization): Document it.
16002 * omp-general.h (get_openacc_privatization_dump_flags): New
16004 * omp-low.c (oacc_privatization_candidate_p): Add diagnostics.
16005 * omp-offload.c (execute_oacc_device_lower)
16006 <IFN_UNIQUE_OACC_PRIVATE>: Re-work diagnostics.
16007 * target.def (goacc.adjust_private_decl): Add 'location_t'
16009 * doc/tm.texi: Regenerate.
16010 * config/gcn/gcn-protos.h (gcn_goacc_adjust_private_decl): Adjust.
16011 * config/gcn/gcn-tree.c (gcn_goacc_adjust_private_decl): Likewise.
16012 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl):
16013 Likewise. Preserve it for...
16014 (nvptx_goacc_expand_var_decl): ... use here.
16016 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16018 * doc/sourcebuild.texi (Other attributes): Document '__OPTIMIZE__'
16021 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16023 PR middle-end/90115
16024 * omp-low.c (oacc_privatization_candidate_p): New function.
16025 (oacc_privatization_scan_clause_chain)
16026 (oacc_privatization_scan_decl_chain): Use it. Also
16027 'gcc_checking_assert' that we're not seeing duplicates.
16029 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16031 PR middle-end/90115
16032 * omp-offload.c (execute_oacc_device_lower): Skip processing if no
16035 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16037 PR middle-end/90115
16038 * omp-offload.c (execute_oacc_device_lower): Explain.
16040 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16042 PR middle-end/90115
16043 * omp-offload.c (execute_oacc_device_lower)
16044 <IFN_UNIQUE_OACC_PRIVATE>: Diagnose and handle for 'level == -1'
16046 * internal-fn.c (expand_UNIQUE): Don't expect
16047 'IFN_UNIQUE_OACC_PRIVATE'.
16049 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16051 PR middle-end/90115
16052 * omp-low.c (lower_omp_for): Don't evaluate OpenMP 'for' clauses.
16054 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
16056 PR middle-end/90115
16057 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl)
16058 (nvptx_goacc_expand_var_decl): Tighten.
16060 2021-05-21 Julian Brown <julian@codesourcery.com>
16061 Chung-Lin Tang <cltang@codesourcery.com>
16062 Thomas Schwinge <thomas@codesourcery.com>
16064 PR middle-end/90115
16065 * doc/tm.texi.in (TARGET_GOACC_EXPAND_VAR_DECL)
16066 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Add documentation hooks.
16067 * doc/tm.texi: Regenerate.
16068 * expr.c (expand_expr_real_1): Expand decls using the
16069 expand_var_decl OpenACC hook if defined.
16070 * internal-fn.c (expand_UNIQUE): Handle IFN_UNIQUE_OACC_PRIVATE.
16071 * internal-fn.h (IFN_UNIQUE_CODES): Add OACC_PRIVATE.
16072 * omp-low.c (omp_context): Add oacc_privatization_candidates
16074 (lower_oacc_reductions): Add PRIVATE_MARKER parameter. Insert
16076 (lower_oacc_head_tail): Add PRIVATE_MARKER parameter. Modify
16077 private marker's gimple call arguments, and pass it to
16078 lower_oacc_reductions.
16079 (oacc_privatization_scan_clause_chain)
16080 (oacc_privatization_scan_decl_chain, lower_oacc_private_marker):
16082 (lower_omp_for, lower_omp_target, lower_omp_1): Use these.
16083 * omp-offload.c (convert.h): Include.
16084 (oacc_loop_xform_head_tail): Treat private-variable markers like
16085 fork/join when transforming head/tail sequences.
16086 (struct var_decl_rewrite_info): Add struct.
16087 (oacc_rewrite_var_decl, is_sync_builtin_call): New functions.
16088 (execute_oacc_device_lower): Support rewriting gang-private
16089 variables using target hook, and fix up addr_expr and var_decl
16091 * target.def (adjust_private_decl, expand_var_decl): New hooks.
16092 * config/gcn/gcn-protos.h (gcn_goacc_adjust_gangprivate_decl):
16094 (gcn_goacc_adjust_private_decl): ...this.
16095 * config/gcn/gcn-tree.c (gcn_goacc_adjust_gangprivate_decl):
16097 (gcn_goacc_adjust_private_decl): ...this. Add LEVEL parameter.
16098 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_GANGPRIVATE_DECL): Rename
16099 definition using gcn_goacc_adjust_gangprivate_decl...
16100 (TARGET_GOACC_ADJUST_PRIVATE_DECL): ...to this, using
16101 gcn_goacc_adjust_private_decl.
16102 * config/nvptx/nvptx.c (tree-pretty-print.h): Include.
16103 (gang_private_shared_size): New global variable.
16104 (gang_private_shared_align): Likewise.
16105 (gang_private_shared_sym): Likewise.
16106 (gang_private_shared_hmap): Likewise.
16107 (nvptx_option_override): Initialize these.
16108 (nvptx_file_end): Output gang_private_shared_sym.
16109 (nvptx_goacc_adjust_private_decl, nvptx_goacc_expand_var_decl):
16111 (nvptx_set_current_function): Clear gang_private_shared_hmap.
16112 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Define hook.
16113 (TARGET_GOACC_EXPAND_VAR_DECL): Likewise.
16115 2021-05-21 H.J. Lu <hjl.tools@gmail.com>
16117 * config/i386/i386-modes.def (MAX_BITSIZE_MODE_ANY_INT): Removed.
16119 2021-05-21 Richard Biener <rguenther@suse.de>
16120 H.J. Lu <hjl.tools@gmail.com>
16122 PR middle-end/90773
16123 * expr.c (expand_constructor): Elide expand_constructor if
16124 move by pieces is preferred.
16126 2021-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16128 * config/aarch64/aarch64-builtins.c (aarch64_call_properties):
16129 Take a flag and mode value as arguments.
16130 (aarch64_modifies_global_state_p): Likewise.
16131 (aarch64_reads_global_state_p): Likewise.
16132 (aarch64_could_trap_p): Likewise.
16133 (aarch64_get_attributes): Likewise.
16134 (aarch64_init_simd_builtins): Adjust callsite of above.
16135 (aarch64_init_fcmla_laneq_builtins): Use aarch64_get_attributes to get
16136 function attributes to apply to builtins.
16137 (aarch64_init_crc32_builtins): Likewise.
16138 (aarch64_init_builtin_rsqrt): Likewise.
16140 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
16142 * config/rs6000/rs6000.md (define_attr "type"): Add types for fusion.
16143 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10): Use new fusion types.
16144 (gen_2logical): Use new fusion types.
16145 * config/rs6000/fusion.md: Regenerate.
16147 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
16150 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
16151 Handle V4QI and V2HI modes.
16152 (ix86_expand_sse_movcc): Ditto.
16153 * config/i386/mmx.md (*<sat_plusminus:insn><VI_32:mode>3):
16154 New instruction pattern.
16155 (*eq<VI_32:mode>3): Ditto.
16156 (*gt<VI_32:mode>3): Ditto.
16157 (*xop_pcmov_<VI_32:mode>): Ditto.
16158 (mmx_pblendvb32): Ditto.
16159 (mmx_pblendvb64): Rename from mmx_pblendvb.
16160 (vec_cmp<VI_32:mode><VI_32:mode>): New expander.
16161 (vec_cmpu<VI_32:mode><VI_32:mode>): Ditto.
16162 (vcond<VI_32:mode><VI_32:mode>): Ditto.
16163 (vcondu<VI_32:mode><VI_32:mode>): Ditto.
16164 (vcond_mask_<VI_32:mode><VI_32:mode>): Ditto.
16166 2021-05-21 Jakub Jelinek <jakub@redhat.com>
16168 PR tree-optimization/94589
16169 * tree-ssa-phiopt.c (spaceship_replacement): For integral rhs1 and
16170 rhs2, treat x <= 4 equivalently to x < 5 etc. In cmp1 and cmp2 (if
16171 not the same as cmp3) treat <= the same as < and >= the same as >.
16172 Don't require that cond2_phi_edge is true edge, instead take
16173 false/true edges into account based on cmp1/cmp2 comparison kinds.
16175 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
16178 * config/i386/mmx.md (SMAXMIN_MMXMODEI): New mode iterator.
16179 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Macroize expander
16180 from <smaxmin:code>v4hi3> and <smaxmin:code><MMXMODE14:mode>3
16181 using SMAXMIN_MMXMODEI mode iterator.
16182 (*<smaxmin:code>v4qi3): New insn pattern.
16183 (*<smaxmin:code>v2hi3): Ditto.
16184 (SMAXMIN_VI_32): New mode iterator.
16185 (<smaxmin:code><SMAXMIN_VI_32>mode3): New expander.
16186 (UMAXMIN_MMXMODEI): New mode iterator.
16187 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Macroize expander
16188 from <umaxmin:code>v8qi3> and <umaxmin:code><MMXMODE24:mode>3
16189 using UMAXMIN_MMXMODEI mode iterator.
16190 (*<umaxmin:code>v4qi3): New insn pattern.
16191 (*<umaxmin:code>v2hi3): Ditto.
16192 (UMAXMIN_VI_32): New mode iterator.
16193 (<umaxmin:code><UMAXMIN_VI_32>mode3): New expander.
16194 (abs<VI_32:mode>2): New insn pattern.
16195 (ssse3_abs<MMXMODEI:mode>2, abs<MMXMODEI:mode>2): Move from ...
16196 * config/i386/sse.md: ... here.
16198 2021-05-20 Clement Chigot <clement.chigot@atos.net>
16199 David Edelsohn <dje.gcc@gmail.com>
16201 * collect2.c (scan_prog_file): Issue non-fatal warning for
16204 2021-05-20 Jonathan Wakely <jwakely@redhat.com>
16206 * doc/invoke.texi (-Wno-c++11-extensions)
16207 (-Wno-c++14-extensions, -Wno-c++17-extensions)
16208 (-Wno-c++20-extensions, -Wno-c++23-extensions): Document
16211 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
16213 * config/c6x/c6x.c (c6x_output_file_unwind): Use dwarf_debuginfo_p.
16214 * config/darwin.c (darwin_override_options): Likewise.
16215 * config/i386/cygming.h (DBX_REGISTER_NUMBER): Likewise.
16216 * config/i386/darwin.h (DBX_REGISTER_NUMBER): Likewise.
16217 (DWARF2_FRAME_REG_OUT): Likewise.
16218 * config/mips/mips.c (mips_output_filename): Likewise.
16219 * config/rs6000/rs6000.c (rs6000_xcoff_declare_function_name):
16221 (rs6000_dbx_register_number): Likewise.
16222 * dbxout.c: Include flags.h.
16223 * dwarf2cfi.c (cfi_label_required_p): Likewise.
16224 (dwarf2out_do_frame): Likewise.
16225 * except.c: Include flags.h.
16226 * final.c (dwarf2_debug_info_emitted_p): Likewise.
16227 (final_scan_insn_1): Likewise.
16228 * flags.h (dwarf_debuginfo_p): New function declaration.
16229 * opts.c (dwarf_debuginfo_p): New function definition.
16230 * targhooks.c (default_debug_unwind_info): Use dwarf_debuginfo_p.
16231 * toplev.c (process_options): Likewise.
16233 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
16235 * common.opt: Change type to support bitmasks.
16236 * flag-types.h (enum debug_info_type): Rename enumerator constants.
16237 (NO_DEBUG): New bitmask.
16238 (DBX_DEBUG): Likewise.
16239 (DWARF2_DEBUG): Likewise.
16240 (XCOFF_DEBUG): Likewise.
16241 (VMS_DEBUG): Likewise.
16242 (VMS_AND_DWARF2_DEBUG): Likewise.
16243 * flags.h (debug_set_to_format): New function declaration.
16244 (debug_set_count): Likewise.
16245 (debug_set_names): Likewise.
16246 * opts.c (debug_type_masks): Array of bitmasks for debug formats.
16247 (debug_set_to_format): New function definition.
16248 (debug_set_count): Likewise.
16249 (debug_set_names): Likewise.
16250 (set_debug_level): Update access to debug_type_names.
16251 * toplev.c: Likewise.
16253 2021-05-20 Martin Sebor <msebor@redhat.com>
16255 PR middle-end/100684
16256 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Handle C++ lambda.
16258 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
16261 * config/i386/i386.md (isa): Remove x64_bmi.
16262 (enabled): Remove x64_bmi.
16263 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
16264 Remove general register alternative.
16265 (*andnot<VI_32:mode>3): Ditto.
16266 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
16267 (*<any_logic:code><VI_32:mode>3): Ditto.
16269 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
16271 * config/arm/arm.c: Include head files tree-vectorizer.h and
16274 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
16277 * config/i386/mmx.md (Yv_Yw): Revert adding V4QI and V2HI modes.
16278 (*<plusminus:insn><VI_32:mode>3): Use Yw instad of <Yv_Yw> constrint.
16279 (<s>mulv4hi3_highpart): New expander.
16280 (*<s>mulv2hi3_highpart): New insn pattern.
16281 (<s>mulv2hi3_higpart): New expander.
16282 (*<any_shift:insn>v2hi3): New insn pattern.
16283 (<any_shift:insn>v2hi3): New expander.
16284 * config/i386/sse.md (smulhrsv2hi3): New expander.
16285 (*smulhrsv2hi3): New insn pattern.
16287 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
16289 * doc/invoke.texi (vect-inner-loop-cost-factor): Document new
16291 * params.opt (vect-inner-loop-cost-factor): New.
16292 * targhooks.c (default_add_stmt_cost): Replace hardcoded factor
16293 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR, include head file
16294 tree-vectorizer.h and its required ones.
16295 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Replace
16296 hardcoded factor 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR.
16297 * config/arm/arm.c (arm_add_stmt_cost): Likewise.
16298 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
16299 * config/rs6000/rs6000.c (rs6000_add_stmt_cost): Likewise.
16300 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
16302 (_loop_vec_info::_loop_vec_info): Init inner_loop_cost_factor.
16303 * tree-vectorizer.h (_loop_vec_info): Add inner_loop_cost_factor.
16304 (LOOP_VINFO_INNER_LOOP_COST_FACTOR): New macro.
16306 2021-05-20 Christophe Lyon <christophe.lyon@linaro.org>
16307 Torbjörn Svensson <torbjorn.svensson@st.com>
16310 * doc/cpp.texi (Common Predefined Macros): Document __FILE_NAME__.
16312 2021-05-20 Jakub Jelinek <jakub@redhat.com>
16314 PR middle-end/99928
16315 * gimplify.c (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LINEAR>: For
16316 explicit linear clause when combined with target, make it map(tofrom:)
16317 instead of no clause or firstprivate.
16319 2021-05-20 Jakub Jelinek <jakub@redhat.com>
16321 PR tree-optimization/94589
16322 * match.pd ((X & Y) == X -> (X & ~Y) == 0): Simplify even in presence
16323 of integral conversions.
16325 2021-05-19 Andrew MacLeod <amacleod@redhat.com>
16327 * gimple-range.cc (fur_source::get_operand): New.
16328 (gimple_range_fold): Delete.
16329 (fold_using_range::fold_stmt): Move from gimple_ranger::calc_stmt.
16330 (fold_using_range::range_of_range_op): Move from gimple_ranger.
16331 (fold_using_range::range_of_address): Ditto.
16332 (fold_using_range::range_of_phi): Ditto.
16333 (fold_using_range::range_of_call): Ditto.
16334 (fold_using_range::range_of_builtin_ubsan_call): Move from
16335 range_of_builtin_ubsan_call.
16336 (fold_using_range::range_of_builtin_call): Move from
16337 range_of_builtin_call.
16338 (gimple_ranger::range_of_builtin_call): Delete.
16339 (fold_using_range::range_of_cond_expr): Move from gimple_ranger.
16340 (gimple_ranger::fold_range_internal): New.
16341 (gimple_ranger::range_of_stmt): Use new fold_using_range API.
16342 (fold_using_range::range_of_ssa_name_with_loop_info): Move from
16343 gimple_ranger. Improve ranges of SSA_NAMES when possible.
16344 * gimple-range.h (gimple_ranger): Remove various range_of routines.
16345 (class fur_source): New.
16346 (class fold_using_range): New.
16347 (fur_source::fur_source): New.
16349 * vr-values.c (vr_values::extract_range_basic): Use fold_using_range
16350 instead of range_of_builtin_call.
16352 2021-05-19 Jonathan Wakely <jwakely@redhat.com>
16354 * doc/cpp.texi (Common Predefined Macros): Update documentation
16355 for the __GXX_EXPERIMENTAL_CXX0X__ macro.
16357 2021-05-19 Alex Coplan <alex.coplan@arm.com>
16360 * config/arm/arm.md (nonsecure_call_internal): Always ensure
16361 callee's address is in a register.
16363 2021-05-19 Geng Qi <gengqi@linux.alibaba.com>
16365 * common/config/riscv/riscv-common.c
16366 (riscv_subset_list::parsing_subset_version): Properly parse the letter
16368 (riscv_subset_list::parse_std_ext,
16369 riscv_subset_list::parse_multiletter_ext): To handle errors generated
16370 in riscv_subset_list::parsing_subset_version.
16372 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
16374 * config/aarch64/aarch64-simd.md: Use "neon_move_narrow_q"
16375 type attribute in patterns generating XTN(2).
16377 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
16379 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_pack_trunc_<mode>):
16380 Remove as duplicate of...
16381 (aarch64_xtn<mode>): This.
16382 (aarch64_xtn2<mode>_le): Move position in file.
16383 (aarch64_xtn2<mode>_be): Move position in file.
16384 (aarch64_xtn2<mode>): Move position in file.
16385 (vec_pack_trunc_<mode>): Define as an expander.
16387 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
16389 * config/aarch64/aarch64-simd-builtins.def: Split builtin
16390 generation for aarch64_<sur>q<r>shr<u>n_n<mode> pattern into
16391 separate scalar and vector generators.
16392 * config/aarch64/aarch64-simd.md
16393 (aarch64_<sur>q<r>shr<u>n_n<mode>): Define as an expander and
16395 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_le): This and...
16396 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_be): This.
16397 * config/aarch64/iterators.md: Define SD_HSDI iterator.
16399 2021-05-19 Jonathn Wright <jonathan.wright@arm.com>
16401 * config/aarch64/aarch64-simd.md: Use UNSPEC_SQXTUN instead
16403 * config/aarch64/iterators.md: Remove UNSPEC_SQXTUN2.
16405 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
16407 * config/aarch64/aarch64-simd.md (aarch64_<sur>q<r>shr<u>n2_n<mode>):
16408 Implement as an expand emitting a big/little endian
16409 instruction pattern.
16410 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_le): Define.
16411 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_be): Define.
16413 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
16415 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn2<mode>):
16416 Implement as an expand emitting a big/little endian
16417 instruction pattern.
16418 (aarch64_<sur><addsub>hn2<mode>_insn_le): Define.
16419 (aarch64_<sur><addsub>hn2<mode>_insn_be): Define.
16420 * config/aarch64/iterators.md: Remove UNSPEC_[R]ADDHN2 and
16421 UNSPEC_[R]SUBHN2 unspecs and ADDSUBHN2 iterator.
16423 2021-05-19 Richard Biener <rguenther@suse.de>
16425 PR middle-end/100672
16426 * fold-const.c (fold_negate_expr_1): Use element_precision.
16427 (negate_expr_p): Likewise.
16429 2021-05-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
16431 * config/aarch64/iterators.md (SVE_PRED_LOAD): New iterator.
16432 (pred_load): New int attribute.
16433 * config/aarch64/aarch64-sve.md
16434 (aarch64_load_<ANY_EXTEND:optab><SVE_HSDI:mode><SVE_PARTIAL_I:mode>): Use
16435 SVE_PRED_LOAD enum iterator and corresponding pred_load attribute.
16436 * config/aarch64/aarch64-sve-builtins-base.cc (expand): Update call to
16437 code_for_aarch64_load.
16439 2021-05-19 Richard Biener <rguenther@suse.de>
16441 * cfgexpand.c (discover_nonconstant_array_refs_r): Make
16442 sure TARGET_MEM_REF bases are expanded as memory.
16443 * tree-ssa-operands.c (operands_scanner::get_tmr_operands):
16444 Do not mark TARGET_MEM_REF bases addressable.
16445 * tree-ssa.c (non_rewritable_mem_ref_base): Handle
16446 TARGET_MEM_REF bases as never rewritable.
16447 * gimple-walk.c (walk_stmt_load_store_addr_ops): Do not
16448 walk TARGET_MEM_REF bases as address-takens.
16449 * tree-ssa-dce.c (ref_may_be_aliased): Handle TARGET_MEM_REF.
16451 2021-05-19 Richard Biener <rguenther@suse.de>
16453 * builtins.c (get_object_alignment_1): Strip outer
16455 * tree-dfa.c (get_ref_base_and_extent): Handle outer
16456 WITH_SIZE_EXPR for size processing and process the
16458 * tree-ssa-alias.c (ao_ref_base_alias_set): Strip
16459 outer WITH_SIZE_EXPR.
16460 (ao_ref_base_alias_ptr_type): Likewise.
16461 (refs_may_alias_p_2): Allow WITH_SIZE_EXPR in ref->ref
16462 and handle that accordingly, stripping it for the
16463 core alias workers.
16464 * tree.c (get_base_address): Handle WITH_SIZE_EXPR by
16465 looking through it instead of returning NULL.
16467 2021-05-19 Jakub Jelinek <jakub@redhat.com>
16469 PR middle-end/100576
16470 * builtins.c (check_read_access): Convert bound to size_type_node if
16473 2021-05-19 Richard Biener <rguenther@suse.de>
16475 * tree-cfg.c (verify_types_in_gimple_min_lval): Inline...
16476 (verify_types_in_gimple_reference): ... here. Sanitize.
16477 (verify_gimple_call): Verify references in LHS and arguments.
16478 (verify_gimple_assign_single): Reject WITH_SIZE_EXPR.
16480 2021-05-19 Uroš Bizjak <ubizjak@gmail.com>
16482 * config/i386/i386.h (VALID_INT_MODE_P):
16483 Add V8QI, V4HI and V2SI modes for TARGET_64BIT.
16484 * config/i386/i386.md (isa): Add x64_bmi.
16485 (enabled): Handle x64_bmi.
16486 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
16487 Add alternative using 64bit general registers.
16488 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
16490 2021-05-19 Jakub Jelinek <jakub@redhat.com>
16492 PR middle-end/99928
16493 * tree.h (OMP_MASTER_COMBINED): Define.
16494 * gimplify.c (gimplify_scan_omp_clauses): Rewrite lastprivate
16495 handling for outer combined/composite constructs to a loop.
16496 Handle lastprivate on combined target.
16497 (gimplify_expr): Formatting fix.
16499 2021-05-19 Xionghu Luo <luoxhu@linux.ibm.com>
16501 * passes.def: Add sink_code pass before store_merging.
16502 * tree-ssa-sink.c (pass_sink_code:clone): New.
16504 2021-05-18 Bill Schmidt <wschmidt@linux.ibm.com>
16506 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Remove call to
16507 rs6000_special_adjust_field_align_p.
16508 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
16509 * config/rs6000/rs6000-call.c (rs6000_function_arg_boundary):
16510 Remove ABI warning.
16511 (rs6000_function_arg): Likewise.
16512 * config/rs6000/rs6000-protos.h
16513 (rs6000_special_adjust_field_align_p): Remove prototype.
16514 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
16516 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Remove call to
16517 rs6000_special_adjust_field_align_p.
16519 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
16522 * config/i386/i386.h (VALID_SSE2_REG_MODE):
16523 Add V4QI and V2HI modes.
16524 (VALID_INT_MODE_P): Ditto.
16525 * config/i386/mmx.md (VI_32): New mode iterator.
16526 (mmxvecsize): Handle V4QI and V2HI.
16528 (mov<VI_32:mode>): New expander.
16529 (*mov<mode>_internal): New insn pattern.
16530 (movmisalign<VI_32:mode>): New expander.
16531 (neg<VI_32:mode>): New expander.
16532 (<plusminus:insn><VI_32:mode>3): New expander.
16533 (*<plusminus:insn><VI_32:mode>3): New insn pattern.
16534 (mulv2hi3): New expander.
16535 (*mulv2hi3): New insn pattern.
16536 (one_cmpl<VI_32:mode>2): New expander.
16537 (*andnot<VI_32:mode>3): New insn pattern.
16538 (<any_logic:code><VI_32:mode>3): New expander.
16539 (*<any_logic:code><VI_32:mode>3): New insn pattern.
16541 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
16543 * config/i386/sse.md (<any_extend:insn>v4qiv4di2):
16544 Fix a mode mismatch with operand 1.
16546 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
16549 * config/i386/i386-expand.c (split_double_mode): Return
16550 temporary register when simplify_gen_subreg fails with
16551 the high half od the paradoxical subreg.
16553 2021-05-18 Richard Biener <rguenther@suse.de>
16555 * cfgexpand.c (expand_one_var): Pass in forced_stack_var
16556 and honor it when expanding.
16557 (expand_used_vars_for_block): Pass through forced_stack_var.
16558 (expand_used_vars): Likewise.
16559 (discover_nonconstant_array_refs_r): Set bits in
16560 forced_stack_vars instead of marking vars TREE_ADDRESSABLE.
16561 (avoid_type_punning_on_regs): Likewise.
16562 (discover_nonconstant_array_refs): Likewise.
16563 (pass_expand::execute): Create and pass down forced_stack_var
16564 bitmap. For parameters and returns temporarily set
16565 TREE_ADDRESSABLE when expand_function_start.
16567 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
16569 * doc/sourcebuild.texi: Document 'dg-note'.
16571 2021-05-18 Tobias Burnus <tobias@codesourcery.com>
16574 * configure: Regenerate.
16575 * configure.ac (BUILD_CFLAG, BUILD_CXXFLAGS): Add $(CFLAGS-$@).
16577 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
16579 * gimple.h (is_gimple_omp_oacc): Tighten.
16580 * omp-low.c (check_omp_nesting_restrictions): Adjust.
16582 2021-05-18 Richard Biener <rguenther@suse.de>
16584 * tree-ssa-operands.c (mark_address_taken): Simplify.
16586 2021-05-18 Martin Liska <mliska@suse.cz>
16588 * config/gcn/mkoffload.c (STR): Redefine.
16589 * config/i386/intelmic-mkoffload.c (STR): Likewise.
16590 * config/nvptx/mkoffload.c (STR): Likewise.
16592 2021-05-18 Martin Liska <mliska@suse.cz>
16594 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
16595 Use startswith function instead of strncmp.
16596 * common/config/bfin/bfin-common.c (bfin_handle_option): Likewise.
16597 * common/config/riscv/riscv-common.c (riscv_subset_list::parse): Likewise.
16598 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): Likewise.
16599 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Likewise.
16600 * config/alpha/alpha.c (alpha_elf_section_type_flags): Likewise.
16601 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
16602 * config/arm/arm.c (arm_file_start): Likewise.
16603 (arm_valid_target_attribute_rec): Likewise.
16604 (thumb1_md_asm_adjust): Likewise.
16605 * config/arm/driver-arm.c (host_detect_local_cpu): Likewise.
16606 * config/avr/avr.c (STR_PREFIX_P): Likewise.
16607 (avr_set_current_function): Likewise.
16608 (avr_handle_addr_attribute): Likewise.
16609 (avr_asm_output_aligned_decl_common): Likewise.
16610 (avr_asm_named_section): Likewise.
16611 (avr_section_type_flags): Likewise.
16612 (avr_asm_select_section): Likewise.
16613 * config/c6x/c6x.c (c6x_in_small_data_p): Likewise.
16614 (c6x_section_type_flags): Likewise.
16615 * config/darwin-c.c (darwin_cfstring_ref_p): Likewise.
16616 (darwin_objc_declare_unresolved_class_reference): Likewise.
16617 (darwin_objc_declare_class_definition): Likewise.
16618 * config/darwin.c (indirect_data): Likewise.
16619 (darwin_encode_section_info): Likewise.
16620 (darwin_objc2_section): Likewise.
16621 (darwin_objc1_section): Likewise.
16622 (machopic_select_section): Likewise.
16623 (darwin_globalize_label): Likewise.
16624 (darwin_label_is_anonymous_local_objc_name): Likewise.
16625 (darwin_asm_named_section): Likewise.
16626 (darwin_asm_output_dwarf_offset): Likewise.
16627 * config/frv/frv.c (frv_string_begins_with): Likewise.
16628 (frv_in_small_data_p): Likewise.
16629 * config/gcn/mkoffload.c (STR): Likewise.
16631 * config/i386/i386-builtins.c (get_builtin_code_for_version): Likewise.
16632 * config/i386/i386-options.c (ix86_option_override_internal): Likewise.
16633 * config/i386/i386.c (x86_64_elf_section_type_flags): Likewise.
16634 (ix86_md_asm_adjust): Likewise.
16635 * config/i386/intelmic-mkoffload.c (STR): Likewise.
16636 * config/i386/winnt.c (i386_pe_asm_named_section): Likewise.
16637 (i386_pe_file_end): Likewise.
16638 * config/ia64/ia64.c (ia64_in_small_data_p): Likewise.
16639 (ia64_section_type_flags): Likewise.
16640 * config/mips/driver-native.c (host_detect_local_cpu): Likewise.
16641 * config/mips/mips.c (mips_handle_interrupt_attr): Likewise.
16642 (mips16_stub_function_p): Likewise.
16643 (mips_function_rodata_section): Likewise.
16644 * config/msp430/msp430.c (msp430_mcu_name): Likewise.
16645 (msp430_function_section): Likewise.
16646 (msp430_section_type_flags): Likewise.
16647 (msp430_expand_helper): Likewise.
16648 * config/nios2/nios2.c (nios2_small_section_name_p): Likewise.
16649 (nios2_valid_target_attribute_rec): Likewise.
16650 * config/nvptx/mkoffload.c (process): Likewise.
16652 * config/pa/som.h: Likewise.
16653 * config/pdp11/pdp11.c (pdp11_output_ident): Likewise.
16654 * config/riscv/riscv.c (riscv_elf_select_rtx_section): Likewise.
16655 * config/rs6000/rs6000.c (VTABLE_NAME_P): Likewise.
16656 (rs6000_inner_target_options): Likewise.
16657 * config/s390/driver-native.c (s390_host_detect_local_cpu): Likewise.
16658 * config/sparc/driver-sparc.c (host_detect_local_cpu): Likewise.
16659 * config/vax/vax.c (vax_output_int_move): Likewise.
16660 * config/vms/vms-ld.c (startswith): Likewise.
16661 (process_args): Likewise.
16663 * config/vms/vms.c: Likewise.
16665 2021-05-18 Jakub Jelinek <jakub@redhat.com>
16667 PR rtl-optimization/100590
16668 * regcprop.c (copyprop_hardreg_forward_1): Only DCE dead sets if
16669 they are NONJUMP_INSN_P.
16671 2021-05-18 Jakub Jelinek <jakub@redhat.com>
16674 * function.c (push_dummy_function): Set DECL_ARTIFICIAL and
16675 DECL_ASSEMBLER_NAME on the fn_decl.
16677 2021-05-18 Jakub Jelinek <jakub@redhat.com>
16679 PR tree-optimization/94589
16680 * tree-ssa-phiopt.c (spaceship_replacement): Pattern match
16681 phi result used in (res & ~1) == 0 comparison as res >= 0 as
16682 res == 2 would be UB with -ffinite-math-only.
16684 2021-05-18 Martin Liska <mliska@suse.cz>
16686 * Makefile.in: genversion.o should depend on DATESTAMP.
16688 2021-05-18 Claudiu Zissulescu <claziss@synopsys.com>
16690 * config/arc/simdext.md (negv2si2): Remove round bracket.
16692 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
16694 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Define
16695 _Bool as macro expanding to _Bool.
16697 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
16700 * tree.c (build_reference_type_for_mode)
16701 (build_pointer_type_for_mode): Pick pointer mode if MODE argument
16703 (build_reference_type, build_pointer_type): Invoke
16704 build_*_type_for_mode with VOIDmode.
16706 2021-05-17 Andrew MacLeod <amacleod@redhat.com>
16708 PR tree-optimization/100512
16709 * gimple-range-cache.cc (ranger_cache::set_global_range): Mark const
16710 and non-zero pointer ranges as invariant.
16711 * gimple-range.cc (gimple_ranger::range_of_stmt): Remove pointer
16712 processing from here.
16714 2021-05-17 Tom de Vries <tdevries@suse.de>
16717 * config/nvptx/nvptx-protos.h (nvptx_output_atomic_insn): Declare
16718 * config/nvptx/nvptx.c (nvptx_output_barrier)
16719 (nvptx_output_atomic_insn): New function.
16720 (nvptx_print_operand): Add support for 'B'.
16721 * config/nvptx/nvptx.md: Use nvptx_output_atomic_insn for atomic
16724 2021-05-17 Aldy Hernandez <aldyh@redhat.com>
16726 PR tree-optimization/100349
16727 * vr-values.c (bounds_of_var_in_loop): Bail if scev returns
16730 2021-05-17 Tamar Christina <tamar.christina@arm.com>
16732 * config/aarch64/driver-aarch64.c (DEFAULT_ARCH): New.
16733 (host_detect_local_cpu): Use it.
16735 2021-05-17 Martin Liska <mliska@suse.cz>
16737 * doc/invoke.texi: Add 2 missing dots.
16739 2021-05-17 Marius Hillenbrand <mhillen@linux.ibm.com>
16741 PR bootstrap/100552
16742 * configure.ac: Replace pattern substitution with call to sed.
16743 * configure: Regenerate.
16745 2021-05-17 Richard Biener <rguenther@suse.de>
16747 PR middle-end/100582
16748 * tree.c (array_at_struct_end_p): Get to the base of the
16749 reference before looking for the underlying decl.
16751 2021-05-17 Joern Rennecke <joern.rennecke@embecosm.com>
16753 * genoutput.c (validate_insn_alternatives) Make "wrong number of
16754 alternatives" message more specific, and remove assumption on where
16757 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
16759 * config/arm/iterators.md (V16): New iterator.
16760 (VH_cvtto): New iterator.
16761 (v_cmp_result): Added V4HF and V8HF support.
16762 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result>): Use VDQWH.
16763 (vcond<mode><mode>): Likewise.
16764 (vcond_mask_<mode><v_cmp_result>): Likewise.
16765 (vcond<VH_cvtto><mode>): New expander.
16767 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
16769 * config/arm/arm-protos.h (arm_expand_vector_compare): Update
16771 * config/arm/arm.c (arm_expand_vector_compare): Add support for
16773 (arm_expand_vcond): Likewise.
16774 * config/arm/iterators.md (supf): Remove VCMPNEQ_S, VCMPEQQ_S,
16775 VCMPEQQ_N_S, VCMPNEQ_N_S.
16776 (VCMPNEQ, VCMPEQQ, VCMPEQQ_N, VCMPNEQ_N): Remove.
16777 * config/arm/mve.md (@mve_vcmp<mve_cmp_op>q_<mode>): Add '@' prefix.
16778 (@mve_vcmp<mve_cmp_op>q_f<mode>): Likewise.
16779 (@mve_vcmp<mve_cmp_op>q_n_f<mode>): Likewise.
16780 (@mve_vpselq_<supf><mode>): Likewise.
16781 (@mve_vpselq_f<mode>"): Likewise.
16782 * config/arm/neon.md (vec_cmp<mode><v_cmp_result): Enable for MVE
16783 and move to vec-common.md.
16784 (vec_cmpu<mode><mode>): Likewise.
16785 (vcond<mode><mode>): Likewise.
16786 (vcond<V_cvtto><mode>): Likewise.
16787 (vcondu<mode><v_cmp_result>): Likewise.
16788 (vcond_mask_<mode><v_cmp_result>): Likewise.
16789 * config/arm/unspecs.md (VCMPNEQ_U, VCMPNEQ_S, VCMPEQQ_S)
16790 (VCMPEQQ_N_S, VCMPNEQ_N_S, VCMPEQQ_U, CMPEQQ_N_U, VCMPNEQ_N_U)
16791 (VCMPGEQ_N_S, VCMPGEQ_S, VCMPGTQ_N_S, VCMPGTQ_S, VCMPLEQ_N_S)
16792 (VCMPLEQ_S, VCMPLTQ_N_S, VCMPLTQ_S, VCMPCSQ_N_U, VCMPCSQ_U)
16793 (VCMPHIQ_N_U, VCMPHIQ_U): Remove.
16794 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result): Moved
16796 (vec_cmpu<mode><mode>): Likewise.
16797 (vcond<mode><mode>): Likewise.
16798 (vcond<V_cvtto><mode>): Likewise.
16799 (vcondu<mode><v_cmp_result>): Likewise.
16800 (vcond_mask_<mode><v_cmp_result>): Likewise. Added unsafe math
16803 2021-05-17 liuhongt <hongtao.liu@intel.com>
16806 * config/i386/i386.c (ix86_gimple_fold_builtin): Use
16807 gsi_insert_seq_before instead.
16809 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
16811 * doc/sourcebuild.texi (arm_qbit_ok): Rename into...
16812 (arm_sat_ok): ...this.
16814 2021-05-17 Martin Liska <mliska@suse.cz>
16816 * lto-wrapper.c (merge_flto_options): Factor out a new function.
16817 (merge_and_complain): Use it.
16818 (run_gcc): Merge also linker command line -flto=foo argument
16821 2021-05-16 Christophe Lyon <christophe.lyon@linaro.org>
16823 * config/arm/arm.h (CPP_SPEC): Remove error message about
16824 -mlittle-endian/-mbig-endian conflict.
16826 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
16828 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
16829 __ROP_PROTECT__ if -mrop-protect is selected.
16831 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
16833 * config/rs6000/rs6000-internal.h (rs6000_stack): Add
16834 rop_hash_save_offset and rop_hash_size.
16835 * config/rs6000/rs6000-logue.c (rs6000_stack_info): Compute
16836 rop_hash_size and rop_hash_save_offset.
16837 (debug_stack_info): Dump rop_hash_save_offset and rop_hash_size.
16838 (rs6000_emit_prologue): Emit hashst[p] in prologue.
16839 (rs6000_emit_epilogue): Emit hashchk[p] in epilogue.
16840 * config/rs6000/rs6000.md (unspec): Add UNSPEC_HASHST and
16842 (hashst): New define_insn.
16843 (hashchk): Likewise.
16845 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
16847 * config/rs6000/rs6000.c (rs6000_option_override_internal):
16848 Disable shrink wrap when inserting ROP-protect instructions.
16849 * config/rs6000/rs6000.opt (mrop-protect): New option.
16850 (mprivileged): Likewise.
16851 * doc/invoke.texi: Document mrop-protect and mprivileged.
16853 2021-05-15 Hans-Peter Nilsson <hp@axis.com>
16855 * reorg.c (fill_slots_from_thread): Reinstate code typoed out in
16858 2021-05-15 Martin Jambor <mjambor@suse.cz>
16861 2021-05-13 Martin Jambor <mjambor@suse.cz>
16863 PR tree-optimization/100453
16864 * tree-sra.c (sra_modify_assign): All const base accesses do not
16865 need refreshing, not just those from decl_pool.
16866 (sra_modify_assign): Do not refresh into a const base decl.
16868 2021-05-15 Jakub Jelinek <jakub@redhat.com>
16870 PR rtl-optimization/100342
16871 * regcprop.c (copy_value): When copying a source reg in a wider
16872 mode than it has recorded for the value, adjust recorded destination
16873 mode too or punt if !REG_CAN_CHANGE_MODE_P.
16875 2021-05-14 Jason Merrill <jason@redhat.com>
16877 * intl.h: Add comments.
16879 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16881 * config/aarch64/aarch64-simd.md
16882 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Split into...
16883 (aarch64_sqdmlsl2_lane<mode>_internal): ... This...
16884 (aarch64_sqdmlal2_lane<mode>_internal): ... And this.
16885 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Split into ...
16886 (aarch64_sqdmlsl2_laneq<mode>_internal): ... This...
16887 (aarch64_sqdmlal2_laneq<mode>_internal): ... And this.
16888 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>_internal): Split into...
16889 (aarch64_sqdmlsl2_n<mode>_internal): ... This...
16890 (aarch64_sqdmlal2_n<mode>_internal): ... And this.
16892 2021-05-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16895 * config/arm/arm_neon.h (vtst_s8): Replace call to vtst builtin with it's
16896 boolean logic equivalent.
16897 (vtst_s16): Likewise.
16898 (vtst_s32): Likewise.
16899 (vtst_u8): Likewise.
16900 (vtst_u16): Likewise.
16901 (vtst_u32): Likewise.
16902 (vtst_p8): Likewise.
16903 (vtst_p16): Likewise.
16904 (vtstq_s8): Likewise.
16905 (vtstq_s16): Likewise.
16906 (vtstq_s32): Likewise.
16907 (vtstq_u8): Likewise.
16908 (vtstq_u16): Likewise.
16909 (vtstq_u32): Likewise.
16910 (vtstq_p8): Likewise.
16911 (vtstq_p16): Likewise.
16912 * config/arm/arm_neon_builtins.def: Remove entry for vtst.
16913 * config/arm/neon.md (neon_vtst<mode>): Remove pattern.
16915 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16917 * config/aarch64/aarch64-simd.md (aarch64_sqdmlal2<mode>): Merge into...
16918 (aarch64_sqdml<SBINQOPS:as>l2<mode>): ... This.
16919 (aarch64_sqdmlsl2<mode>): Delete.
16920 (aarch64_sqdmlal2_lane<mode>): Merge this...
16921 (aarch64_sqdmlsl2_lane<mode>): ... And this...
16922 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>): ... Into this.
16923 (aarch64_sqdmlal2_laneq<mode>): Merge this...
16924 (aarch64_sqdmlsl2_laneq<mode>): ... And this...
16925 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>): ... Into this.
16926 (aarch64_sqdmlal2_n<mode>): Merge this...
16927 (aarch64_sqdmlsl2_n<mode>): ... And this...
16928 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>): ... Into this.
16930 2021-05-13 Martin Sebor <msebor@redhat.com>
16932 PR middle-end/100574
16933 * builtins.c (access_ref::get_ref): Improve detection of PHIs with
16934 all null arguments.
16936 2021-05-13 Martin Sebor <msebor@redhat.com>
16938 PR tree-optimization/93100
16939 PR middle-end/98583
16940 * tree-ssa-uninit.c (check_defs): Exclude intrinsic functions that
16941 don't modify referenced objects.
16943 2021-05-13 Martin Jambor <mjambor@suse.cz>
16945 PR tree-optimization/100453
16946 * tree-sra.c (sra_modify_assign): All const base accesses do not
16947 need refreshing, not just those from decl_pool.
16948 (sra_modify_assign): Do not refresh into a const base decl.
16950 2021-05-13 Martin Liska <mliska@suse.cz>
16952 * tree-ssa-dom.c: Remove m_simplifier.
16954 2021-05-13 Richard Earnshaw <rearnsha@arm.com>
16957 * config/arm/arm.c (arm_canonicalize_comparison): Correctly
16958 canonicalize DImode inequality comparisons against the
16959 maximum integral value.
16961 2021-05-13 Jakub Jelinek <jakub@redhat.com>
16963 PR tree-optimization/98856
16964 * config/i386/i386.c (ix86_shift_rotate_cost): Add CODE argument.
16965 Expect V2DI and V4DI arithmetic right shifts to be emulated.
16966 (ix86_rtx_costs, ix86_add_stmt_cost): Adjust ix86_shift_rotate_cost
16968 * config/i386/i386-expand.c (expand_vec_perm_2perm_interleave,
16969 expand_vec_perm_2perm_pblendv): New functions.
16970 (ix86_expand_vec_perm_const_1): Use them.
16971 * config/i386/sse.md (ashr<mode>3<mask_name>): Rename to ...
16972 (<mask_codefor>ashr<mode>3<mask_name>): ... this.
16973 (ashr<mode>3): New define_expand with VI248_AVX512BW iterator.
16974 (ashrv4di3): New define_expand.
16975 (ashrv2di3): Change condition to TARGET_SSE2, handle !TARGET_XOP
16976 and !TARGET_AVX512VL expansion.
16978 2021-05-13 Uroš Bizjak <ubizjak@gmail.com>
16981 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Force mode
16982 sizes < 16 to a register when constructing vpcmov pattern.
16983 * config/i386/mmx.md (*xop_pcmov_<mode>): Use MMXMODE124 mode.
16985 2021-05-13 Martin Liska <mliska@suse.cz>
16987 * gcov-io.c (gcov_write_block): Remove.
16988 (gcov_write_words): Likewise.
16989 (gcov_read_words): Re-implement using gcov_read_bytes.
16990 (gcov_allocate): Remove.
16991 (GCOV_BLOCK_SIZE): Likewise.
16992 (struct gcov_var): Remove most of the fields.
16993 (gcov_position): Implement with ftell.
16994 (gcov_rewrite): Remove setting of start and offset fields.
16995 (from_file): Re-format.
16996 (gcov_open): Remove setbuf call. It should not be needed.
16997 (gcov_close): Remove internal buffer handling.
16998 (gcov_magic): Use __builtin_bswap32.
16999 (gcov_write_counter): Use directly gcov_write_unsigned.
17000 (gcov_write_string): Use direct fwrite and do not round
17002 (gcov_seek): Use directly fseek.
17003 (gcov_write_tag): Use gcov_write_unsigned directly.
17004 (gcov_write_length): Likewise.
17005 (gcov_write_tag_length): Likewise.
17006 (gcov_read_bytes): Use directly fread.
17007 (gcov_read_unsigned): Use gcov_read_words.
17008 (gcov_read_counter): Likewise.
17009 (gcov_read_string): Use gcov_read_bytes.
17010 * gcov-io.h (GCOV_WORD_SIZE): Adjust to reflect
17011 that size is not in bytes, but words (4B).
17012 (GCOV_TAG_FUNCTION_LENGTH): Likewise.
17013 (GCOV_TAG_ARCS_LENGTH): Likewise.
17014 (GCOV_TAG_ARCS_NUM): Likewise.
17015 (GCOV_TAG_COUNTER_LENGTH): Likewise.
17016 (GCOV_TAG_COUNTER_NUM): Likewise.
17017 (GCOV_TAG_SUMMARY_LENGTH): Likewise.
17019 2021-05-13 liuhongt <hongtao.liu@intel.com>
17022 * config/i386/sse.md (ssedoublevecmode): Add attribute for
17023 V64QI/V32HI/V16SI/V4DI.
17024 (ssehalfvecmode): Add attribute for V2DI/V2DF.
17025 (*vec_concatv4si_0): Extend to VI124_128.
17026 (*vec_concat<mode>_0): New pre-reload splitter.
17027 * config/i386/predicates.md (movq_parallel): New predicate.
17029 2021-05-13 Alexandre Oliva <oliva@adacore.com>
17031 * targhooks.c (default_zero_call_used_regs): Retry using
17032 successfully-zeroed registers as sources.
17034 2021-05-12 Tobias Burnus <tobias@codesourcery.com>
17036 * omp-low.c (finish_taskreg_scan): Use the proper detach decl.
17038 2021-05-12 Aldy Hernandez <aldyh@redhat.com>
17041 * gimple-range.cc (range_of_builtin_call): Skip out on
17042 processing __builtin_clz when varying.
17044 2021-05-12 Tom de Vries <tdevries@suse.de>
17047 * config/nvptx/nvptx-opts.h (enum ptx_version): New enum.
17048 * config/nvptx/nvptx.c (nvptx_file_start): Print .version according
17049 to ptx_version_option.
17050 * config/nvptx/nvptx.h (TARGET_PTX_6_3): Define.
17051 * config/nvptx/nvptx.md (define_insn "nvptx_shuffle<mode>")
17052 (define_insn "nvptx_vote_ballot"): Use sync variant for
17054 * config/nvptx/nvptx.opt (ptx_version): Add enum.
17055 (mptx): Add option.
17056 * doc/invoke.texi (Nvidia PTX Options): Add mptx item.
17058 2021-05-12 Richard Biener <rguenther@suse.de>
17060 PR tree-optimization/100566
17061 * tree-ssa-sccvn.c (dominated_by_p_w_unex): Properly handle
17062 allow_back for all edge queries.
17064 2021-05-12 liuhongt <hongtao.liu@intel.com>
17067 * config/i386/sse.md (<sse4_1_avx2>_pblendvb): Add
17068 splitters for pblendvb of NOT mask register.
17070 2021-05-12 Richard Biener <rguenther@suse.de>
17072 PR tree-optimization/100519
17073 * tree-ssa-reassoc.c (can_associate_p): Split into...
17074 (can_associate_op_p): ... this
17075 (can_associate_type_p): ... and this.
17076 (is_reassociable_op): Call can_associate_op_p.
17077 (break_up_subtract_bb): Call the appropriate predicates.
17078 (reassociate_bb): Likewise.
17080 2021-05-12 Martin Liska <mliska@suse.cz>
17082 * lto-wrapper.c (merge_and_complain): Merge -flto=arg options.
17083 (run_gcc): Use -flto argument detection for merged
17086 2021-05-12 Martin Liska <mliska@suse.cz>
17088 * lto-wrapper.c (print_lto_docs_link): New function.
17089 (run_gcc): Print warning about missing job server detection
17090 after we know NR of partitions. Do the same for -flto{,=1}.
17091 * opts.c (get_option_html_page): Support -flto option.
17093 2021-05-12 Martin Liska <mliska@suse.cz>
17095 * lto-wrapper.c (get_options_from_collect_gcc_options): Change
17097 (append_option): Remove.
17098 (find_option): Rework to use the vector type.
17099 (remove_option): Remove.
17100 (merge_and_complain): Use vectors for cl_decoded_option data
17102 (append_compiler_options): Likewise.
17103 (append_diag_options): Likewise.
17104 (append_linker_options): Likewise.
17105 (append_offload_options): Likewise.
17106 (compile_offload_image): Likewise.
17107 (compile_images_for_offload_targets): Likewise.
17108 (find_and_merge_options): Likewise.
17109 (run_gcc): Likewise.
17111 2021-05-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
17114 * dwarf2out.c (dwarf2out_finish): Set
17115 have_multiple_function_sections with multi-range text_section.
17117 2021-05-12 Martin Liska <mliska@suse.cz>
17119 PR bootstrap/100560
17120 * Makefile.in: Remove version.h from linker command line.
17122 2021-05-12 Richard Biener <rguenther@suse.de>
17124 PR middle-end/100547
17125 * rtl.h (rtvec_alloc): Make argument size_t.
17126 * rtl.c (rtvec_alloc): Verify the count is less than INT_MAX.
17128 2021-05-12 Jakub Jelinek <jakub@redhat.com>
17130 PR middle-end/100508
17131 * cfgexpand.c (expand_debug_expr): For DEBUG_EXPR_DECL with vector
17132 type, don't reuse DECL_RTL if it has different mode, instead force
17133 creation of a new DEBUG_EXPR.
17135 2021-05-12 Jakub Jelinek <jakub@redhat.com>
17136 Marc Glisse <marc.glisse@inria.fr>
17138 PR tree-optimization/94589
17139 * match.pd ((X & Y) == X -> (X & ~Y) == 0,
17140 (X | Y) == Y -> (X & ~Y) == 0): New GIMPLE simplifications.
17142 2021-05-12 Uroš Bizjak <ubizjak@gmail.com>
17145 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Handle V2SF mode.
17146 * config/i386/mmx.md (MMXMODE124): New mode iterator.
17148 (mmxintvecmode): New mode attribute.
17149 (mmxintvecmodelower): Ditto.
17150 (*mmx_maskcmpv2sf3_comm): New insn pattern.
17151 (*mmx_maskcmpv2sf3): Ditto.
17152 (vec_cmpv2sfv2si): New expander.
17153 (vcond<V2FI:mode>v2si): Ditto.
17154 (mmx_vlendvps): New insn pattern.
17155 (vcond<MMXMODE124:mode><MMXMODEI:mode>): Also handle V2SFmode.
17156 (vcondu<MMXMODE124:mode><MMXMODEI:mode>): Ditto.
17157 (vcond_mask_<mode><mmxintvecmodelower>): Ditto.
17159 2021-05-11 Martin Sebor <msebor@redhat.com>
17161 PR middle-end/21433
17162 * expr.c (expand_expr_real_1): Replace unreachable code with an assert.
17164 2021-05-11 Richard Biener <rguenther@suse.de>
17166 * gimple-fold.c (gimple_fold_call): Do not call
17167 maybe_fold_reference on call arguments or the static chain.
17168 (fold_stmt_1): Do not call maybe_fold_reference on GIMPLE_ASM
17171 2021-05-11 Martin Liska <mliska@suse.cz>
17173 * builtins.def (DEF_HSAIL_BUILTIN): Remove.
17174 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
17175 (DEF_HSAIL_SAT_BUILTIN): Likewise.
17176 (DEF_HSAIL_INTR_BUILTIN): Likewise.
17177 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
17178 * doc/frontends.texi: Remove BRIG.
17179 * doc/install.texi: Likewise.
17180 * doc/invoke.texi: Likewise.
17181 * doc/standards.texi: Likewise.
17182 * brig-builtins.def: Removed.
17183 * brig/ChangeLog: Removed.
17184 * brig/Make-lang.in: Removed.
17185 * brig/brig-builtins.h: Removed.
17186 * brig/brig-c.h: Removed.
17187 * brig/brig-lang.c: Removed.
17188 * brig/brigfrontend/brig-arg-block-handler.cc: Removed.
17189 * brig/brigfrontend/brig-atomic-inst-handler.cc: Removed.
17190 * brig/brigfrontend/brig-basic-inst-handler.cc: Removed.
17191 * brig/brigfrontend/brig-branch-inst-handler.cc: Removed.
17192 * brig/brigfrontend/brig-cmp-inst-handler.cc: Removed.
17193 * brig/brigfrontend/brig-code-entry-handler.cc: Removed.
17194 * brig/brigfrontend/brig-code-entry-handler.h: Removed.
17195 * brig/brigfrontend/brig-comment-handler.cc: Removed.
17196 * brig/brigfrontend/brig-control-handler.cc: Removed.
17197 * brig/brigfrontend/brig-copy-move-inst-handler.cc: Removed.
17198 * brig/brigfrontend/brig-cvt-inst-handler.cc: Removed.
17199 * brig/brigfrontend/brig-fbarrier-handler.cc: Removed.
17200 * brig/brigfrontend/brig-function-handler.cc: Removed.
17201 * brig/brigfrontend/brig-function.cc: Removed.
17202 * brig/brigfrontend/brig-function.h: Removed.
17203 * brig/brigfrontend/brig-inst-mod-handler.cc: Removed.
17204 * brig/brigfrontend/brig-label-handler.cc: Removed.
17205 * brig/brigfrontend/brig-lane-inst-handler.cc: Removed.
17206 * brig/brigfrontend/brig-machine.c: Removed.
17207 * brig/brigfrontend/brig-machine.h: Removed.
17208 * brig/brigfrontend/brig-mem-inst-handler.cc: Removed.
17209 * brig/brigfrontend/brig-module-handler.cc: Removed.
17210 * brig/brigfrontend/brig-queue-inst-handler.cc: Removed.
17211 * brig/brigfrontend/brig-seg-inst-handler.cc: Removed.
17212 * brig/brigfrontend/brig-signal-inst-handler.cc: Removed.
17213 * brig/brigfrontend/brig-to-generic.cc: Removed.
17214 * brig/brigfrontend/brig-to-generic.h: Removed.
17215 * brig/brigfrontend/brig-util.cc: Removed.
17216 * brig/brigfrontend/brig-util.h: Removed.
17217 * brig/brigfrontend/brig-variable-handler.cc: Removed.
17218 * brig/brigfrontend/hsa-brig-format.h: Removed.
17219 * brig/brigfrontend/phsa.h: Removed.
17220 * brig/brigspec.c: Removed.
17221 * brig/config-lang.in: Removed.
17222 * brig/gccbrig.texi: Removed.
17223 * brig/lang-specs.h: Removed.
17224 * brig/lang.opt: Removed.
17226 2021-05-11 Richard Biener <rguenther@suse.de>
17229 * ipa-param-manipulation.c
17230 (ipa_param_body_adjustments::modify_call_stmt): Avoid
17231 altering SSA_NAME_DEF_STMT by adjusting the calls LHS
17232 via gimple_call_lhs_ptr.
17234 2021-05-11 Alex Coplan <alex.coplan@arm.com>
17237 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear):
17238 Avoid emitting CFA adjusts on the sp if we have the fp.
17240 2021-05-11 Richard Sandiford <richard.sandiford@arm.com>
17242 * config/aarch64/iterators.md (VMUL_CHANGE_NLANES): Delete.
17243 (VMULD): New iterator.
17244 (VCOND): Handle V4HF and V8HF.
17245 (VCONQ): Fix entry for V2SF.
17246 * config/aarch64/aarch64-simd.md (mul_lane<mode>3): Use VMULD
17247 instead of VMUL. Use a 64-bit vector mode for the indexed operand.
17248 (*aarch64_mul3_elt_<vswap_width_name><mode>): Merge with...
17249 (mul_laneq<mode>3): ...this define_insn. Use VMUL instead of VDQSF.
17250 Use a 128-bit vector mode for the indexed operand. Use stype for
17251 the scheduling type.
17253 2021-05-11 Richard Biener <rguenther@suse.de>
17255 * gimple-fold.c (maybe_fold_reference): Only return
17256 is_gimple_min_invariant values.
17258 2021-05-11 Richard Biener <rguenther@suse.de>
17260 PR middle-end/100509
17261 * gimple-fold.c (fold_gimple_assign): Only call
17262 get_symbol_constant_value on register type symbols.
17264 2021-05-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
17265 Joe Ramsay <joe.ramsay@arm.com>
17268 * config/arm/arm_mve.h (__arm_vstrwq_scatter_offset): Fix wrong arguments.
17269 (__arm_vcmpneq): Remove duplicate definition.
17270 (__arm_vstrwq_scatter_offset_p): Likewise.
17271 (__arm_vmaxq_x): Likewise.
17272 (__arm_vmlsdavaq): Likewise.
17273 (__arm_vmlsdavaxq): Likewise.
17274 (__arm_vmlsdavq_p): Likewise.
17275 (__arm_vmlsdavxq_p): Likewise.
17276 (__arm_vrmlaldavhaq): Likewise.
17277 (__arm_vstrbq_p): Likewise.
17278 (__arm_vstrbq_scatter_offset): Likewise.
17279 (__arm_vstrbq_scatter_offset_p): Likewise.
17280 (__arm_vstrdq_scatter_offset): Likewise.
17281 (__arm_vstrdq_scatter_offset_p): Likewise.
17282 (__arm_vstrdq_scatter_shifted_offset): Likewise.
17283 (__arm_vstrdq_scatter_shifted_offset_p): Likewise.
17285 2021-05-11 Jakub Jelinek <jakub@redhat.com>
17287 PR middle-end/100471
17288 * omp-low.c (lower_omp_task_reductions): For OMP_TASKLOOP, if data
17289 is 0, bypass the reduction loop including
17290 GOMP_taskgroup_reduction_unregister call.
17292 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
17294 * config/rs6000/rs6000.c (struct rs6000_cost_data): New member
17295 costing_for_scalar.
17296 (rs6000_density_test): Early return if costing_for_scalar is true.
17297 (rs6000_init_cost): Init costing_for_scalar of rs6000_cost_data.
17299 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
17301 * doc/tm.texi: Regenerated.
17302 * target.def (init_cost): Add new parameter costing_for_scalar.
17303 * targhooks.c (default_init_cost): Adjust for new parameter.
17304 * targhooks.h (default_init_cost): Likewise.
17305 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
17306 (vect_compute_single_scalar_iteration_cost): Likewise.
17307 (vect_analyze_loop_2): Likewise.
17308 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
17309 (vect_bb_vectorization_profitable_p): Likewise.
17310 * tree-vectorizer.h (init_cost): Likewise.
17311 * config/aarch64/aarch64.c (aarch64_init_cost): Likewise.
17312 * config/i386/i386.c (ix86_init_cost): Likewise.
17313 * config/rs6000/rs6000.c (rs6000_init_cost): Likewise.
17315 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
17317 * config/rs6000/rs6000.c (rs6000_vect_nonmem): Renamed to
17318 vect_nonmem and moved into...
17319 (struct rs6000_cost_data): ...here.
17320 (rs6000_init_cost): Use vect_nonmem of cost_data instead.
17321 (rs6000_add_stmt_cost): Likewise.
17322 (rs6000_finish_cost): Likewise.
17324 2021-05-10 Eric Botcazou <ebotcazou@adacore.com>
17326 * range-op.cc (get_bool_state): Adjust head comment.
17327 (operator_not_equal::op1_range): Fix comment.
17328 (operator_bitwise_xor::op1_range): Remove call to gcc_unreachable.
17330 2021-05-10 Martin Sebor <msebor@redhat.com>
17332 PR middle-end/100425
17333 PR middle-end/100510
17334 * gimple-ssa-warn-alloca.c (pass_walloca::firast_time_p): Rename...
17335 (pass_walloca::xlimit_certain_p): ...to this.
17336 (pass_walloca::gate): Execute for any kind of handled warning.
17337 (pass_walloca::execute): Avoid issuing "maybe" and "unbounded"
17338 warnings when xlimit_certain_p is set.
17340 2021-05-10 Pat Haugen <pthaugen@linux.ibm.com>
17342 * config/rs6000/rs6000.c (rs6000_ira_change_pseudo_allocno_class):
17343 Return ALTIVEC_REGS if that is best_class.
17344 (rs6000_compute_pressure_classes): Add ALTIVEC_REGS.
17346 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17348 * config/arm/arm.h (CPP_SPEC): Remove error message about
17351 2021-05-10 Martin Jambor <mjambor@suse.cz>
17353 * ipa-prop.h (IPA_NODE_REF): Removed.
17354 (IPA_NODE_REF_GET_CREATE): Likewise.
17355 (IPA_EDGE_REF): Likewise.
17356 (IPA_EDGE_REF_GET_CREATE): Likewise.
17357 (IS_VALID_JUMP_FUNC_INDEX): Likewise.
17358 * ipa-cp.c (print_all_lattices): Replaced IPA_NODE_REF with a direct
17359 use of ipa_node_params_sum.
17360 (ipcp_versionable_function_p): Likewise.
17361 (push_node_to_stack): Likewise.
17362 (pop_node_from_stack): Likewise.
17363 (set_single_call_flag): Replaced two IPA_NODE_REF with one single
17364 direct use of ipa_node_params_sum.
17365 (initialize_node_lattices): Replaced IPA_NODE_REF with a direct use of
17366 ipa_node_params_sum.
17367 (ipa_context_from_jfunc): Replaced IPA_EDGE_REF with a direct use of
17369 (ipcp_verify_propagated_values): Replaced IPA_NODE_REF with a direct
17370 use of ipa_node_params_sum.
17371 (self_recursively_generated_p): Likewise.
17372 (propagate_scalar_across_jump_function): Likewise.
17373 (propagate_context_across_jump_function): Replaced IPA_EDGE_REF with a
17374 direct use of ipa_edge_args_sum, moved the lookup after the early
17375 exit. Replaced IPA_NODE_REF with a direct use of ipa_node_params_sum.
17376 (propagate_bits_across_jump_function): Replaced IPA_NODE_REF with
17377 direct uses of ipa_node_params_sum.
17378 (propagate_vr_across_jump_function): Likewise.
17379 (propagate_aggregate_lattice): Likewise.
17380 (propagate_aggs_across_jump_function): Likewise.
17381 (propagate_constants_across_call): Likewise, also replaced
17382 IPA_EDGE_REF with a direct use of ipa_edge_args_sum.
17383 (good_cloning_opportunity_p): Replaced IPA_NODE_REF with a direct use
17384 of ipa_node_params_sum.
17385 (estimate_local_effects): Likewise.
17386 (add_all_node_vals_to_toposort): Likewise.
17387 (propagate_constants_topo): Likewise.
17388 (ipcp_propagate_stage): Likewise.
17389 (ipcp_discover_new_direct_edges): Likewise.
17390 (calls_same_node_or_its_all_contexts_clone_p): Likewise.
17391 (cgraph_edge_brings_value_p): Likewise (in both overloaded functions).
17392 (get_info_about_necessary_edges): Likewise.
17393 (want_remove_some_param_p): Likewise.
17394 (create_specialized_node): Likewise.
17395 (self_recursive_pass_through_p): Likewise.
17396 (self_recursive_agg_pass_through_p): Likewise.
17397 (find_more_scalar_values_for_callers_subset): Likewise and also
17398 replaced IPA_EDGE_REF with direct uses of ipa_edge_args_sum, in one
17399 case replacing two of those with a single query.
17400 (find_more_contexts_for_caller_subset): Likewise for the
17401 ipa_polymorphic_call_context overload.
17402 (intersect_aggregates_with_edge): Replaced IPA_EDGE_REF with a direct
17403 use of ipa_edge_args_sum. Replaced IPA_NODE_REF with direct uses of
17404 ipa_node_params_sum.
17405 (find_aggregate_values_for_callers_subset): Likewise, also reusing
17406 results of ipa_edge_args_sum->get.
17407 (cgraph_edge_brings_all_scalars_for_node): Replaced IPA_NODE_REF with
17408 direct uses of ipa_node_params_sum, replaced IPA_EDGE_REF with a
17409 direct use of ipa_edge_args_sum.
17410 (cgraph_edge_brings_all_agg_vals_for_node): Likewise, moved node
17411 summary query after the early exit and reused the result later.
17412 (decide_about_value): Replaced IPA_NODE_REF with a direct use of
17413 ipa_node_params_sum.
17414 (decide_whether_version_node): Likewise. Removed re-querying for
17415 summaries after cloning.
17416 (spread_undeadness): Replaced IPA_NODE_REF with a direct use of
17417 ipa_node_params_sum.
17418 (has_undead_caller_from_outside_scc_p): Likewise, reusing results of
17420 (identify_dead_nodes): Likewise.
17421 (ipcp_store_bits_results): Replaced IPA_NODE_REF with direct uses of
17422 ipa_node_params_sum.
17423 (ipcp_store_vr_results): Likewise.
17424 * ipa-fnsummary.c (evaluate_properties_for_edge): Likewise.
17425 (ipa_fn_summary_t::duplicate): Likewise.
17426 (analyze_function_body): Likewise.
17427 (estimate_calls_size_and_time): Likewise.
17428 (ipa_cached_call_context::duplicate_from): Likewise.
17429 (ipa_call_context::equal_to): Likewise.
17430 (remap_edge_params): Likewise.
17431 (ipa_merge_fn_summary_after_inlining): Likewise.
17432 (inline_read_section): Likewise.
17433 * ipa-icf.c (sem_function::param_used_p): Likewise.
17434 * ipa-modref.c (compute_parm_map): Likewise.
17435 (compute_parm_map): Replaced IPA_EDGE_REF with a direct use of
17437 (get_access_for_fnspec): Replaced IPA_NODE_REF with a direct use of
17438 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
17440 * ipa-profile.c (check_argument_count): Likewise.
17441 * ipa-prop.c (ipa_alloc_node_params): Replaced IPA_NODE_REF_GET_CREATE
17442 with a direct use of ipa_node_params_sum.
17443 (ipa_initialize_node_params): Likewise.
17444 (ipa_print_node_jump_functions_for_edge): Replaced IPA_EDGE_REF with a
17445 direct use of ipa_edge_args_sum and reused the query result.
17446 (ipa_compute_jump_functions_for_edge): Replaced IPA_NODE_REF with a
17447 direct use of ipa_node_params_sum and replaced IPA_EDGE_REF with a
17448 direct use of ipa_edge_args_sum.
17449 (ipa_note_param_call): Replaced IPA_NODE_REF with a direct use of
17450 ipa_node_params_sum and reused the result of the query.
17451 (ipa_analyze_node): Likewise.
17452 (ipa_analyze_controlled_uses): Replaced IPA_NODE_REF with a direct use
17453 of ipa_node_params_sum.
17454 (update_jump_functions_after_inlining): Replaced IPA_EDGE_REF with
17455 direct uses of ipa_edge_args_sum.
17456 (update_indirect_edges_after_inlining): Replaced IPA_NODE_REF with
17457 direct uses of ipa_node_params_sum and replaced IPA_EDGE_REF with a
17458 direct use of ipa_edge_args_sum. Removed superficial re-querying the
17460 (propagate_controlled_uses): Replaced IPA_NODE_REF with direct uses of
17461 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
17463 (ipa_propagate_indirect_call_infos): Replaced IPA_EDGE_REF with a
17464 direct use of ipa_edge_args_sum.
17465 (ipa_edge_args_sum_t::duplicate): Replaced IPA_NODE_REF with a direct
17466 use of ipa_node_params_sum.
17467 (ipa_print_node_params): Likewise.
17468 (ipa_write_node_info): Likewise and also replaced IPA_EDGE_REF with
17469 direct uses of ipa_edge_args_sum.
17470 (ipa_read_edge_info): Replaced IPA_EDGE_REF with a direct use of
17472 (ipa_read_node_info): Replaced IPA_NODE_REF with a direct use of
17473 ipa_node_params_sum.
17474 (ipa_prop_write_jump_functions): Likewise. Move variable node to the
17475 scopes where it is used.
17477 2021-05-10 Uroš Bizjak <ubizjak@gmail.com>
17479 * config/i386/i386-expand.c (ix86_expand_sse_movcc)
17480 <case E_V2SImode>: Force op_true to register.
17482 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17484 * config/arm/iterators.md (MVE_FP_COMPARISONS): New.
17485 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_f<mode>)
17486 (mve_vcmp<mve_cmp_op>q_n_f<mode>): New, merge all vcmp_*f*
17488 (mve_vcmpeqq_f<mode>, mve_vcmpeqq_n_f<mode>, mve_vcmpgeq_f<mode>)
17489 (mve_vcmpgeq_n_f<mode>, mve_vcmpgtq_f<mode>)
17490 (mve_vcmpgtq_n_f<mode>, mve_vcmpleq_f<mode>)
17491 (mve_vcmpleq_n_f<mode>, mve_vcmpltq_f<mode>)
17492 (mve_vcmpltq_n_f<mode>, mve_vcmpneq_f<mode>)
17493 (mve_vcmpneq_n_f<mode>): Remove.
17494 * config/arm/unspecs.md (VCMPEQQ_F, VCMPEQQ_N_F, VCMPGEQ_F)
17495 (VCMPGEQ_N_F, VCMPGTQ_F, VCMPGTQ_N_F, VCMPLEQ_F, VCMPLEQ_N_F)
17496 (VCMPLTQ_F, VCMPLTQ_N_F, VCMPNEQ_F, VCMPNEQ_N_F): Remove.
17498 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17500 * config/arm/iterators.md (MVE_COMPARISONS): New.
17502 (mve_cmp_type): New.
17503 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_<mode>): New, merge all
17505 (mve_vcmpneq_<mode>, mve_vcmpcsq_n_<mode>, mve_vcmpcsq_<mode>)
17506 (mve_vcmpeqq_n_<mode>, mve_vcmpeqq_<mode>, mve_vcmpgeq_n_<mode>)
17507 (mve_vcmpgeq_<mode>, mve_vcmpgtq_n_<mode>, mve_vcmpgtq_<mode>)
17508 (mve_vcmphiq_n_<mode>, mve_vcmphiq_<mode>, mve_vcmpleq_n_<mode>)
17509 (mve_vcmpleq_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
17510 (mve_vcmpneq_n_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
17511 (mve_vcmpneq_n_<mode>): Remove.
17513 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17515 * config/arm/arm_mve.h (__arm_vcmp*): Remove 's' suffix.
17516 * config/arm/arm_mve_builtins.def (vcmp*): Remove 's' suffix.
17517 * config/arm/mve.md (mve_vcmp*): Remove 's' suffix in pattern
17520 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17522 * config/arm/arm_mve_builtins.def (vcmpneq_u): Remove.
17523 (vcmpneq_n_u): Likewise.
17524 (vcmpeqq_u,): Likewise.
17525 (vcmpeqq_n_u): Likewise.
17526 * config/arm/iterators.md (supf): Remove VCMPNEQ_U, VCMPEQQ_U,
17527 VCMPEQQ_N_U and VCMPNEQ_N_U.
17528 * config/arm/mve.md (mve_vcmpneq): Remove <supf> iteration.
17529 (mve_vcmpeqq_n): Likewise.
17530 (mve_vcmpeqq): Likewise.
17531 (mve_vcmpneq_n): Likewise.
17533 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17535 * config/arm/arm_mve.h (__arm_vcmpeq*u*, __arm_vcmpne*u*): Call
17536 the 's' version of the builtin.
17538 2021-05-10 Richard Biener <rguenther@suse.de>
17540 PR tree-optimization/100492
17541 * tree-loop-distribution.c (find_seed_stmts_for_distribution):
17542 Find nothing when the loop contains an irreducible region.
17544 2021-05-10 Richard Biener <rguenther@suse.de>
17546 PR middle-end/100464
17548 * gimple-fold.c (canonicalize_constructor_val): Do not set
17551 2021-05-10 Richard Biener <rguenther@suse.de>
17553 PR tree-optimization/100434
17554 * tree-ssa-dse.c (initialize_ao_ref_for_dse): Handle
17556 (dse_optimize_stmt): Handle call LHS by dropping the
17557 LHS or the whole call if it doesn't have other
17559 (pass_dse::execute): Adjust.
17561 2021-05-10 Martin Liska <mliska@suse.cz>
17563 * Makefile.in: Add missing genversion rule.
17565 2021-05-10 Alex Coplan <alex.coplan@arm.com>
17568 * config/arm/mve.md (*mve_mov<mode>): Simplify output code. Use
17569 vldrw.u32 and vstrw.32 for V2D[IF]mode loads and stores.
17571 2021-05-10 Martin Liska <mliska@suse.cz>
17573 * builtins.c (is_builtin_name): Use startswith
17574 function instead of strncmp.
17575 * collect2.c (main): Likewise.
17576 (has_lto_section): Likewise.
17577 (scan_libraries): Likewise.
17578 * coverage.c (coverage_checksum_string): Likewise.
17579 (coverage_init): Likewise.
17580 * dwarf2out.c (is_cxx): Likewise.
17581 (gen_compile_unit_die): Likewise.
17582 * gcc-ar.c (main): Likewise.
17583 * gcc.c (init_spec): Likewise.
17584 (read_specs): Likewise.
17585 (execute): Likewise.
17586 (check_live_switch): Likewise.
17587 * genattrtab.c (write_attr_case): Likewise.
17588 (IS_ATTR_GROUP): Likewise.
17589 * gencfn-macros.c (main): Likewise.
17590 * gengtype.c (type_for_name): Likewise.
17591 (gen_rtx_next): Likewise.
17592 (get_file_langdir): Likewise.
17593 (write_local): Likewise.
17594 * genmatch.c (get_operator): Likewise.
17595 (get_operand_type): Likewise.
17596 (expr::gen_transform): Likewise.
17597 * genoutput.c (validate_optab_operands): Likewise.
17598 * incpath.c (add_sysroot_to_chain): Likewise.
17599 * langhooks.c (lang_GNU_C): Likewise.
17600 (lang_GNU_CXX): Likewise.
17601 (lang_GNU_Fortran): Likewise.
17602 (lang_GNU_OBJC): Likewise.
17603 * lto-wrapper.c (run_gcc): Likewise.
17604 * omp-general.c (omp_max_simt_vf): Likewise.
17605 * omp-low.c (omp_runtime_api_call): Likewise.
17606 * opts-common.c (parse_options_from_collect_gcc_options): Likewise.
17607 * read-rtl-function.c (function_reader::read_rtx_operand_r): Likewise.
17608 * real.c (real_from_string): Likewise.
17609 * selftest.c (assert_str_startswith): Likewise.
17610 * timevar.c (timer::validate_phases): Likewise.
17611 * tree.c (get_file_function_name): Likewise.
17612 * ubsan.c (ubsan_use_new_style_p): Likewise.
17613 * varasm.c (default_function_rodata_section): Likewise.
17614 (incorporeal_function_p): Likewise.
17615 (default_section_type_flags): Likewise.
17616 * system.h (startswith): Define startswith.
17618 2021-05-10 Martin Liska <mliska@suse.cz>
17620 * bitmap.h (class auto_bitmap): Remove
17621 __cplusplus >= 201103.
17622 * config/aarch64/aarch64.c: Likewise.
17623 * gimple-ssa-store-merging.c (store_immediate_info::store_immediate_info):
17625 * sbitmap.h: Likewise.
17627 2021-05-10 Martin Liska <mliska@suse.cz>
17629 * Makefile.in: Rename gcov-iov to genversion and depend
17630 on version.h (instead of gcov-iov.h).
17631 * gcov-io.h: Include version.h instread of gcov-iov.h.
17632 * gengtype-state.c (read_state_version): Likewise.
17633 * gcov-iov.c: Moved to...
17634 * genversion.c: ...here.
17635 * lto-streamer.h (LTO_major_version): Define it with
17637 * version.c: Removed.
17638 * version.h: Removed.
17640 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17642 * config/arc/arc.md (UNSPEC_ARC_DMPYWH): Define.
17643 * config/arc/simdext.md (VCT): Add predicates for iterator
17646 (voptab): Likewise.
17647 (vec_widen_<V_US>mult_hi_v4hi): Change pattern predicate.
17648 (<voptab>v2si3): New patterns.
17650 (reduc_plus_scal_v4hi): Likewise.
17651 (reduc_plus_scal_v2si): Likewise.
17652 (vec_duplicatev2si): Likewise.
17653 (vec_duplicatev4hi): Likewise.
17655 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17657 * config/arc/simdext.md: Format and cleanup file.
17659 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17661 * config/arc/simdext.md (movmisalignv2hi): Allow misaligned access
17662 only when munaligned-access option is on.
17663 (movmisalign<mode>): Likewise.
17665 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17667 * common/config/arc/arc-common.c (arc_handle_option): Remove dot
17669 * config/arc/arc.c (arc_reorg): Remove underscore from string.
17671 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17673 * config/arc/arc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
17674 (CTZ_DEFINED_VALUE_AT_ZERO): Likewise.
17675 * config/arc/arc.md (clrsbsi2): Cleanup pattern.
17676 (norm_f): Likewise.
17679 (clzsi2): Use fls instruction when available.
17680 (arc_clzsi2): Likewise.
17682 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17684 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add r26 and r27.
17686 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17688 * doc/extend.texi (__builtin_arc_sr): Swap arguments.
17690 2021-05-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
17692 PR middle-end/100467
17693 * toplev.c (compile_file): Call insn_locations_init before
17694 targetm.asm_out.code_end.
17696 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
17699 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
17701 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
17703 2021-05-07 Jakub Jelinek <jakub@redhat.com>
17704 Andrew Stubbs <amd@codesourcery.com>
17707 * builtins.c (try_store_by_multiple_pieces): Use force_operand for
17708 emit_move_insn operands.
17710 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
17712 * cfgexpand.c (expand_gimple_basic_block): Do not inherit a current
17713 location for the outgoing edges of an empty block.
17714 * dwarf2out.c (add_subscript_info): Retrieve the bounds and index
17715 type by means of the get_array_descr_info langhook, if it is set and
17716 returns true. Remove obsolete code dealing with unnamed subtypes.
17718 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
17720 * gimple-range-cache.cc (ssa_block_ranges): Virtualize.
17721 (sbr_vector): Renamed from ssa_block_cache.
17722 (sbr_vector::sbr_vector): Allocate from obstack abd initialize.
17723 (ssa_block_ranges::~ssa_block_ranges): Remove.
17724 (sbr_vector::set_bb_range): Use varying and undefined cached values.
17725 (ssa_block_ranges::set_bb_varying): Remove.
17726 (sbr_vector::get_bb_range): Adjust assert.
17727 (sbr_vector::bb_range_p): Adjust assert.
17728 (~block_range_cache): No freeing loop required.
17729 (block_range_cache::get_block_ranges): Remove.
17730 (block_range_cache::set_bb_range): Inline get_block_ranges.
17731 (block_range_cache::set_bb_varying): Remove.
17732 * gimple-range-cache.h (set_bb_varying): Remove prototype.
17733 * value-range.h (irange_allocator::get_memory): New.
17735 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
17737 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Search
17738 dominator tree is available and requested.
17739 (ranger_cache::ssa_range_in_bb): Don't search dom tree here.
17740 (ranger_cache::fill_block_cache): Don't search dom tree here either.
17741 * gimple-range-cache.h (non_null_deref_p): Add dom_search param.
17743 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
17745 * gimple-range.cc (gimple_ranger::range_on_exit): Handle block with
17746 only PHI nodes better.
17748 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
17750 * gimple-range-edge.h (gimple_outgoing_range): Rename from
17752 (gcond_edge_range): Export prototype.
17753 * gimple-range-edge.cc (gcond_edge_range): New.
17754 (gimple_outgoing_range::edge_range_p): Use gcond_edge_range.
17755 * gimple-range-gori.h (gori_compute): Use gimple_outgoing_range.
17757 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
17759 * gimple-range-edge.cc (outgoing_range::calc_switch_ranges): Compute
17760 default range into a temp and allocate only what is needed.
17762 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
17764 * range-op.cc (operator_trunc_mod::wi_fold): x % 0 is UNDEFINED.
17766 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
17768 * gimple-range.h (gimple_range_global): Pick up parameter initial
17769 values, and use-before defined locals are UNDEFINED.
17771 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
17773 * doc/extend.texi (scalar_storage_order): Mention effect on pointer
17775 * tree.h (reverse_storage_order_for_component_p): Return false if
17776 the type is a pointer.
17778 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
17780 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
17782 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
17785 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
17786 Handle V8QI, V4HI and V2SI modes.
17787 * config/i386/mmx.md (mmx_pblendvb): New insn pattern.
17788 * config/i386/sse.md (unspec): Move UNSPEC_BLENDV ...
17789 * config/i386/i386.md (unspec): ... here.
17791 2021-05-07 Tobias Burnus <tobias@codesourcery.com>
17792 Tom de Vries <tdevries@suse.de>
17794 * omp-low.c (lower_rec_simd_input_clauses): Set max_vf = 1 if
17795 a truth_value_p reduction variable is nonintegral.
17797 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
17800 * config/i386/i386-expand.c (ix86_use_mask_cmp_p):
17801 Return false for mode sizes < 16.
17803 2021-05-07 Jakub Jelinek <jakub@redhat.com>
17806 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn.
17808 2021-05-06 Martin Jambor <mjambor@suse.cz>
17810 * ipa-sra.c (ipa_sra_dump_all_summaries): Dump edge summaries even
17811 when there is no function summary.
17812 (ipa_sra_summarize_function): produce edge summaries even when
17815 2021-05-06 Tom Tromey <tom@tromey.com>
17817 * godump.c (string_hash_eq): Remove.
17818 (go_finish): Use htab_eq_string.
17820 2021-05-06 Tom Tromey <tom@tromey.com>
17822 * gengtype-state.c (read_state): Use htab_eq_string.
17823 (string_eq): Remove.
17825 2021-05-06 Tom Tromey <tom@tromey.com>
17827 * gensupport.c (htab_eq_string): Remove.
17829 2021-05-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
17832 * debug.h (gcc_debug_hooks): Add set_ignored_loc function pointer.
17833 * dwarf2out.h (dw_fde_node::ignored_debug): New data item.
17834 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Add dummy
17835 set_ignored_loc callbacks.
17836 * debug.c (do_nothing_debug_hooks): Likewise.
17837 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
17838 * dwarf2out.c (text_section_used, cold_text_section_used): Remove.
17839 (in_text_section_p, last_text_label, last_cold_label,
17840 switch_text_ranges, switch_cold_ranges): New data items.
17841 (dwarf2out_note_section_used): Remove.
17842 (dwarf2out_begin_prologue): Set fde->ignored_debug and
17844 (mark_ignored_debug_section): New helper function.
17845 (dwarf2out_end_epilogue, dwarf2out_switch_text_section): Call
17846 mark_ignored_debug_section.
17847 (dwarf2_debug_hooks): Use dwarf2out_set_ignored_loc.
17848 (dwarf2_lineno_debug_hooks): Use dummy for set_ignored_loc.
17849 (size_of_aranges): Adjust formula for multi-part text ranges size.
17850 (output_aranges): Output multi-part text ranges.
17851 (dwarf2out_set_ignored_loc): New callback function.
17852 (dwarf2out_finish): Output multi-part text ranges.
17853 (dwarf2out_c_finalize): Clear new data items.
17854 * final.c (final_start_function_1): Call set_ignored_loc callback.
17855 (final_scan_insn_1): Likewise.
17856 * ggc-page.c (gt_ggc_mx): New helper function.
17857 * stringpool.c (gt_pch_nx): Likewise.
17859 2021-05-06 Richard Biener <rguenther@suse.de>
17861 * timevar.def (TV_TREE_INSERT_PHI_NODES): Remove.
17862 (TV_TREE_SSA_REWRITE_BLOCKS): Likewise.
17863 (TV_TREE_INTO_SSA): New.
17864 * tree-into-ssa.c (insert_phi_nodes): Do not account separately.
17865 (rewrite_blocks): Likewise.
17866 (pass_data_build_ssa): Account to TV_TREE_INTO_SSA.
17868 2021-05-06 Jakub Jelinek <jakub@redhat.com>
17870 * tree-ssa-phiopt.c (value_replacement, minmax_replacement,
17871 abs_replacement, xor_replacement,
17872 cond_removal_in_popcount_clz_ctz_pattern,
17873 replace_phi_edge_with_variable): Change type of phi argument from
17874 gimple * to gphi *.
17876 2021-05-06 Richard Biener <rguenther@suse.de>
17878 * tree-ssa-loop-split.c (split_loop): Delay updating SSA form.
17879 Output an opt-info message.
17880 (do_split_loop_on_cond): Likewise.
17881 (tree_ssa_split_loops): Update SSA form here.
17883 2021-05-06 Richard Biener <rguenther@suse.de>
17885 * tree-inline.c (tree_function_versioning): Fix DECL_BY_REFERENCE
17886 return variable removal.
17888 2021-05-06 Marius Hillenbrand <mhillen@linux.ibm.com>
17890 * config/s390/s390-builtins.def (O_M5, O1_M5, ...): Remove unused macros.
17891 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
17892 (s390_vec_permi_dbl, s390_vpdi): Use the O3_U2 type for the immediate
17894 * config/s390/s390.c (s390_const_operand_ok): Remove unused
17897 2021-05-06 Jakub Jelinek <jakub@redhat.com>
17899 PR tree-optimization/94589
17900 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Call
17901 spaceship_replacement.
17902 (cond_only_block_p, spaceship_replacement): New functions.
17904 2021-05-06 Richard Biener <rguenther@suse.de>
17907 * tree-emutls.c (gen_emutls_addr): Pass in whether we're
17908 dealing with a debug use and only query existing addresses
17910 (lower_emutls_1): Avoid splitting out addresses for debug
17911 stmts, reset the debug stmt when we fail to find existing
17913 (lower_emutls_phi_arg): Set wi.stmt.
17915 2021-05-06 Christoph Muellner <cmuellner@gcc.gnu.org>
17918 * config/riscv/riscv.c (riscv_block_move_loop): Use cbranch helper.
17919 * config/riscv/riscv.md (cbranch<mode>4): Generate helpers.
17920 (stack_protect_test): Use cbranch helper.
17922 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
17925 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
17926 always return the establisher frame for __builtin_frame_address (0).
17928 2021-05-05 Ivan Sorokin <vanyacpp@gmail.com>
17931 * config/i386/i386-builtins.c (ix86_cpu_model_type_node): New.
17932 (ix86_cpu_model_var): Likewise.
17933 (ix86_cpu_features2_type_node): Likewise.
17934 (ix86_cpu_features2_var): Likewise.
17935 (fold_builtin_cpu): Cache __cpu_model and __cpu_features2 with
17938 2021-05-05 Martin Sebor <msebor@redhat.com>
17940 * passes.def (pass_warn_printf): Run after SSA.
17942 2021-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17944 * config/arm/neon.md (neon_vtst_combine<mode>): New pattern.
17945 * config/arm/predicates.md (minus_one_operand): New predicate.
17947 2021-05-05 Jeff Law <jlaw@tachyum.com>
17949 * config/avr/avr.md: Remove references to CC_STATUS_INIT.
17951 2021-05-05 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
17953 PR rtl-optimization/100263
17954 * postreload.c (move2add_valid_value_p): Ensure register can
17957 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
17959 PR rtl-optimization/100411
17960 * cfgcleanup.c (try_crossjump_to_edge): Also skip end of prologue
17961 and beginning of function markers.
17963 2021-05-05 Jeff Law <jlaw@tachyum.com>
17965 * config/cr16/cr16.h (NOTICE_UPDATE_CC): Remove.
17966 * config/cr16/cr16.c (notice_update_cc): Remove.
17967 * config/cr16/cr16-protos.h (notice_update_cc): Remove.
17969 2021-05-05 Uroš Bizjak <ubizjak@gmail.com>
17972 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
17973 Handle V8QI, V4HI and V2SI modes.
17974 * config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
17975 (ix86_build_signbit_mask): Ditto.
17976 * config/i386/mmx.md (MMXMODE14): New mode iterator.
17977 (<smaxmin:code><MMXMODE14:mode>3): New expander.
17978 (*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
17979 (<umaxmin:code><MMXMODE24:mode>3): New expander.
17980 (*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
17981 (vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
17982 (vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
17983 (vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
17984 (vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
17985 (vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
17987 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
17989 * dwarf2out.c (loc_list_from_tree_1) <DECL>: During early DWARF, do
17990 not expand the VALUE_EXPR of variables put in the non-local frame.
17991 * gimplify.c (gimplify_type_sizes) <RECORD_TYPE>: If the type is not
17992 to be ignored for debug info, ensure its variable offsets are not.
17994 2021-05-05 Richard Biener <rguenther@suse.de>
17996 PR tree-optimization/79333
17997 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
17998 Fold stmt following SSA edges.
18000 2021-05-05 Richard Biener <rguenther@suse.de>
18002 PR middle-end/100394
18003 * calls.c (expand_call): Preserve possibly throwing calls.
18004 * cfgexpand.c (expand_call_stmt): When a call can throw signal
18005 RTL expansion there are side-effects.
18006 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Simplify,
18007 mark all possibly throwing stmts necessary unless we can elide
18009 * tree-ssa-dse.c (pass_dse::execute): Preserve exceptions unless
18010 -fdelete-dead-exceptions.
18011 * tree.h (DECL_PURE_P): Add note about exceptions.
18013 2021-05-05 Alexandre Oliva <oliva@adacore.com>
18015 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Make it
18018 2021-05-04 David Edelsohn <dje.gcc@gmail.com>
18020 * config/rs6000/rs6000-call.c (rs6000_output_mi_thunk): Use
18021 get_fnname_from_decl for name of thunk.
18022 * config/rs6000/rs6000.c (rs6000_declare_alias): Use assemble_name
18023 and ASM_OUTPUT_LABEL.
18024 (rs6000_xcoff_declare_function_name): Use assemble_name and
18026 (rs6000_xcoff_declare_object_name): Use ASM_OUTPUT_LABEL.
18027 (rs6000_xcoff_encode_section_info): Don't add mapping class
18028 for aliases. Always add [DS] mapping class to primary
18030 (rs6000_asm_weaken_decl): Don't explicitly add [DS].
18032 2021-05-04 Martin Sebor <msebor@redhat.com>
18034 PR middle-end/100307
18035 * builtins.c (compute_objsize_r): Clear base0 for pointers.
18037 2021-05-04 Jeff Law <jlaw@tachyum.com>
18039 * config/bfin/bfin.h (NOTICE_UPDATE_CC): Remove.
18041 2021-05-04 Segher Boessenkool <segher@kernel.crashing.org>
18043 * caller-save.c: Remove CC0.
18044 * cfgcleanup.c: Remove CC0.
18045 * cfgrtl.c: Remove CC0.
18046 * combine.c: Remove CC0.
18047 * compare-elim.c: Remove CC0.
18048 * conditions.h: Remove CC0.
18049 * config/h8300/h8300.h: Remove CC0.
18050 * config/h8300/h8300-protos.h: Remove CC0.
18051 * config/h8300/peepholes.md: Remove CC0.
18052 * config/i386/x86-tune-sched.c: Remove CC0.
18053 * config/m68k/m68k.c: Remove CC0.
18054 * config/rl78/rl78.c: Remove CC0.
18055 * config/sparc/sparc.c: Remove CC0.
18056 * config/xtensa/xtensa.c: Remove CC0.
18057 (gen_conditional_move): Use pc_rtx instead of cc0_rtx in a piece of
18058 RTL where that is used as a placeholder only.
18059 * cprop.c: Remove CC0.
18060 * cse.c: Remove CC0.
18061 * cselib.c: Remove CC0.
18062 * df-problems.c: Remove CC0.
18063 * df-scan.c: Remove CC0.
18064 * doc/md.texi: Remove CC0. Adjust an example.
18065 * doc/rtl.texi: Remove CC0. Adjust an example.
18066 * doc/tm.texi: Regenerate.
18067 * doc/tm.texi.in: Remove CC0.
18068 * emit-rtl.c: Remove CC0.
18069 * final.c: Remove CC0.
18070 * fwprop.c: Remove CC0.
18071 * gcse-common.c: Remove CC0.
18072 * gcse.c: Remove CC0.
18073 * genattrtab.c: Remove CC0.
18074 * genconfig.c: Remove CC0.
18075 * genemit.c: Remove CC0.
18076 * genextract.c: Remove CC0.
18077 * gengenrtl.c: Remove CC0.
18078 * genrecog.c: Remove CC0.
18079 * haifa-sched.c: Remove CC0.
18080 * ifcvt.c: Remove CC0.
18081 * ira-costs.c: Remove CC0.
18082 * ira.c: Remove CC0.
18083 * jump.c: Remove CC0.
18084 * loop-invariant.c: Remove CC0.
18085 * lra-constraints.c: Remove CC0.
18086 * lra-eliminations.c: Remove CC0.
18087 * optabs.c: Remove CC0.
18088 * postreload-gcse.c: Remove CC0.
18089 * postreload.c: Remove CC0.
18090 * print-rtl.c: Remove CC0.
18091 * read-rtl-function.c: Remove CC0.
18092 * reg-notes.def: Remove CC0.
18093 * reg-stack.c: Remove CC0.
18094 * reginfo.c: Remove CC0.
18095 * regrename.c: Remove CC0.
18096 * reload.c: Remove CC0.
18097 * reload1.c: Remove CC0.
18098 * reorg.c: Remove CC0.
18099 * resource.c: Remove CC0.
18100 * rtl.c: Remove CC0.
18101 * rtl.def: Remove CC0.
18102 * rtl.h: Remove CC0.
18103 * rtlanal.c: Remove CC0.
18104 * sched-deps.c: Remove CC0.
18105 * sched-rgn.c: Remove CC0.
18106 * shrink-wrap.c: Remove CC0.
18107 * simplify-rtx.c: Remove CC0.
18108 * system.h: Remove CC0. Poison NOTICE_UPDATE_CC, CC_STATUS_MDEP_INIT,
18109 CC_STATUS_MDEP, and CC_STATUS.
18110 * target.def: Remove CC0.
18111 * valtrack.c: Remove CC0.
18112 * var-tracking.c: Remove CC0.
18114 2021-05-04 Richard Biener <rguenther@suse.de>
18116 PR tree-optimization/100414
18117 * tree-ssa-phiopt.c (get_non_trapping): Do not compute dominance
18119 (tree_ssa_phiopt_worker): But unconditionally here.
18121 2021-05-04 Tobias Burnus <tobias@codesourcery.com>
18123 * omp-low.c (lower_rec_input_clauses, lower_reduction_clauses): Handle
18124 && and || with floating-point and complex arguments.
18126 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
18128 * tree-inline.c (insert_debug_decl_map): Delete.
18129 (copy_debug_stmt): Minor tweak.
18130 (setup_one_parameter): Do not use a variable if the value is either
18131 a read-only DECL or a non-addressable local variable in the caller.
18132 In this case, insert the debug-only variable in the map manually.
18133 (expand_call_inline): Do not generate a CLOBBER for these values.
18134 * tree-inline.h (debug_map): Minor tweak.
18136 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
18138 * builtins.c (builtin_with_linkage_p): Return true for stp[n]cpy.
18139 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Tidy up.
18141 2021-05-04 Richard Biener <rguenther@suse.de>
18143 PR tree-optimization/100329
18144 * tree-ssa-reassoc.c (can_reassociate_p): Do not reassociate
18146 (insert_stmt_after): Assert we're not running into asm goto.
18148 2021-05-04 Richard Biener <rguenther@suse.de>
18150 PR tree-optimization/100398
18151 * tree-ssa-dse.c (pass_dse::execute): Preserve control
18154 2021-05-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18156 * builtins.c (try_store_by_multiple_pieces): Fix constfun's prototype.
18158 2021-05-04 Alexandre Oliva <oliva@adacore.com>
18160 * builtins.c (try_store_by_multiple_pieces): New.
18161 (expand_builtin_memset_args): Use it. If target_char_cast
18162 fails, proceed as for non-constant val. Pass len's ctz to...
18163 * expr.c (clear_storage_hints): ... this. Try store by
18164 multiple pieces after setmem.
18165 (clear_storage): Adjust.
18166 * expr.h (clear_storage_hints): Likewise.
18167 (try_store_by_multiple_pieces): Declare.
18168 * passes.def: Replace the last copy_prop with ccp.
18170 2021-05-03 Tom de Vries <tdevries@suse.de>
18173 * omp-low.c (lower_rec_input_clauses): Disable SIMT for user-defined
18176 2021-05-03 Richard Biener <rguenther@suse.de>
18178 * tree-ssa-dse.c (dse_classify_store): Track two PHI defs.
18180 2021-05-03 Richard Biener <rguenther@suse.de>
18182 * tree-ssa-dse.c: Do not include domwalk.h but cfganal.h.
18183 (dse_dom_walker): Remove.
18184 (dse_dom_walker::dse_optimize_stmt): Rename...
18185 (dse_optimize_stmt): ... to this, pass in live_bytes sbitmap.
18186 (dse_dom_walker::before_dom_children): Inline ...
18187 (pass_dse::execute): ... here. Perform a reverse program
18190 2021-05-03 H.J. Lu <hjl.tools@gmail.com>
18193 * configure: Regenerated.
18195 2021-05-03 Ilya Leoshkevich <iii@linux.ibm.com>
18198 * config/s390/s390.c (s390_hard_fp_reg_p): New function.
18199 (s390_md_asm_adjust): Handle hard registers.
18201 2021-05-03 Jakub Jelinek <jakub@redhat.com>
18203 PR tree-optimization/100382
18204 * tree-ssa-dse.c: Include tree-eh.h.
18205 (dse_dom_walker::before_dom_children): Don't remove stmts if
18206 stmt_unremovable_because_of_non_call_eh_p is true.
18208 2021-05-02 David Edelsohn <dje.gcc@gmail.com>
18210 * varasm.c (compute_reloc_for_var): Split out from...
18211 (get_variable_section): Use it.
18212 * output.h (compute_reloc_for_var): Declare.
18213 * config/rs6000/rs6000-protos.h
18214 (rs6000_xcoff_asm_output_aligned_decl_common): Change alignment to
18216 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address_aix):
18217 Don't append storage mapping class to symbol.
18218 (rs6000_xcoff_asm_named_section): Add BS and UL mapping classes.
18219 Don't convert TLS BSS to common.
18220 (rs6000_xcoff_unique_section): Don't fall back to select_secton.
18221 (rs6000_xcoff_section_type_flags): Add SECTION_BSS if DECL is
18223 (rs6000_xcoff_asm_globalize_decl_name): Don't strip storage
18225 (rs6000_xcoff_asm_output_aligned_decl_common): Align is unsigned int.
18226 If align is 0 from TLS class, use the same rules as varasm.c
18227 If not common, switch to BSS section manually.
18228 If common, emit appropriate comm or lcomm directive.
18229 (rs6000_xcoff_encode_section_info): Add logic to append all
18230 storage mapping classes.
18231 (rs6000_asm_weaken_decl): Adjust for qualname symbols.
18232 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
18233 rs6000_xcoff_asm_output_aligned_decl_common.
18234 (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
18235 rs6000_xcoff_asm_output_aligned_decl_common.
18236 (ASM_OUTPUT_TLS_COMMON): Use
18237 rs6000_xcoff_asm_output_aligned_decl_common.
18239 2021-05-02 Jakub Jelinek <jakub@redhat.com>
18242 * config/nvptx/nvptx.c (nvptx_sese_pseudo): Use nullptr instead of 0
18243 as first argument of pseudo_node_t constructors.
18245 2021-05-02 Jakub Jelinek <jakub@redhat.com>
18248 * config/i386/t-i386 (TM_H): Add $(srcdir)/config/i386/i386-isa.def.
18250 2021-05-01 Aldy Hernandez <aldyh@redhat.com>
18252 * value-range.cc (DEFINE_INT_RANGE_GC_STUBS): Remove.
18253 (gt_pch_nx (int_range<1> *&)): New.
18254 (gt_ggc_mx (int_range<1> *&)): New.
18255 * value-range.h (class irange): Add GTY support for
18258 2021-05-01 Geng Qi <gengqi@linux.alibaba.com>
18260 * doc/options.texi (Negative): Change either or to both and.
18262 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
18264 * config/aarch64/aarch64-simd-builtins.def: Add
18265 float_ml[as][q]_laneq builtin generator macros.
18266 * config/aarch64/aarch64-simd.md (mul_laneq<mode>3): Define.
18267 (aarch64_float_mla_laneq<mode>): Define.
18268 (aarch64_float_mls_laneq<mode>): Define.
18269 * config/aarch64/arm_neon.h (vmla_laneq_f32): Use RTL builtin
18270 instead of GCC vector extensions.
18271 (vmlaq_laneq_f32): Likewise.
18272 (vmls_laneq_f32): Likewise.
18273 (vmlsq_laneq_f32): Likewise.
18275 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
18277 * config/aarch64/aarch64-simd-builtins.def: Add
18278 float_ml[as]_lane builtin generator macros.
18279 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt<mode>):
18281 (mul_lane<mode>3): This, and re-order arguments.
18282 (aarch64_float_mla_lane<mode>): Define.
18283 (aarch64_float_mls_lane<mode>): Define.
18284 * config/aarch64/arm_neon.h (vmla_lane_f32): Use RTL builtin
18285 instead of GCC vector extensions.
18286 (vmlaq_lane_f32): Likewise.
18287 (vmls_lane_f32): Likewise.
18288 (vmlsq_lane_f32): Likewise.
18290 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
18292 * config/aarch64/aarch64-simd-builtins.def: Add float_ml[as]
18293 builtin generator macros.
18294 * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>):
18296 (aarch64_float_mls<mode>): Define.
18297 * config/aarch64/arm_neon.h (vmla_f32): Use RTL builtin
18298 instead of relying on GCC vector extensions.
18299 (vmla_f64): Likewise.
18300 (vmlaq_f32): Likewise.
18301 (vmlaq_f64): Likewise.
18302 (vmls_f32): Likewise.
18303 (vmls_f64): Likewise.
18304 (vmlsq_f32): Likewise.
18305 (vmlsq_f64): Likewise.
18306 * config/aarch64/iterators.md: Define VDQF_DF mode iterator.
18308 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
18310 * config/aarch64/aarch64-simd-builtins.def: Add
18311 float_ml[as]_n_builtin generator macros.
18312 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt_from_dup<mode>):
18314 (mul_n<mode>3): This, and re-order arguments.
18315 (aarch64_float_mla_n<mode>): Define.
18316 (aarch64_float_mls_n<mode>): Define.
18317 * config/aarch64/arm_neon.h (vmla_n_f32): Use RTL builtin
18318 instead of inline asm.
18319 (vmlaq_n_f32): Likewise.
18320 (vmls_n_f32): Likewise.
18321 (vmlsq_n_f32): Likewise.
18323 2021-04-30 Jonathan Wright <joanthan.wright@arm.com>
18325 * config/aarch64/aarch64-simd-builtins.def: Add pmull[2]
18326 builtin generator macros.
18327 * config/aarch64/aarch64-simd.md (aarch64_pmullv8qi): Define.
18328 (aarch64_pmull_hiv16qi_insn): Define.
18329 (aarch64_pmull_hiv16qi): Define.
18330 * config/aarch64/arm_neon.h (vmull_high_p8): Use RTL builtin
18331 instead of inline asm.
18332 (vmull_p8): Likewise.
18334 2021-04-30 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
18336 * config/avr/avr.md: Adjust peepholes to match and
18337 generate parallels with clobber of REG_CC.
18338 (mov<mode>_insn): Rename to mov<mode>_insn_split.
18339 (*mov<mode>_insn): Rename to mov<mode>_insn.
18341 2021-04-30 David Edelsohn <dje.gcc@gmail.com>
18343 * varasm.c (use_blocks_for_decl_p): Don't use section anchors
18344 for VAR_DECLs if -fdata-sections enabled.
18346 2021-04-30 Michael Meissner <meissner@linux.ibm.com>
18348 PR bootstrap/100327
18349 * config/rs6000/rs6000.c
18350 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Define.
18351 (rs6000_libgcc_floating_mode_supported_p): New target hook.
18353 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
18355 * tree-ssa-threadbackward.c (class thread_jumps): Split out code
18357 (class back_threader_registry): ...to here...
18358 (class back_threader_profitability): ...and here...
18359 (thread_jumps::thread_through_all_blocks): Remove argument.
18360 (back_threader_registry::back_threader_registry): New.
18361 (back_threader_registry::~back_threader_registry): New.
18362 (back_threader_registry::thread_through_all_blocks): New.
18363 (thread_jumps::profitable_jump_thread_path): Move from here...
18364 (back_threader_profitability::profitable_path_p): ...to here.
18365 (thread_jumps::find_taken_edge): New.
18366 (thread_jumps::convert_and_register_current_path): Move...
18367 (back_threader_registry::register_path): ...to here.
18368 (thread_jumps::register_jump_thread_path_if_profitable): Move...
18369 (thread_jumps::maybe_register_path): ...to here.
18370 (thread_jumps::handle_phi): Call find_taken_edge and
18371 maybe_register_path.
18372 (thread_jumps::handle_assignment): Same.
18373 (thread_jumps::fsm_find_control_statement_thread_paths): Remove
18374 tree argument to handle_phi and handle_assignment.
18375 (thread_jumps::find_jump_threads_backwards): Set m_name. Remove
18376 set of m_speed_p and m_max_threaded_paths.
18377 (pass_thread_jumps::execute): Remove second argument from
18378 find_jump_threads_backwards.
18379 (pass_early_thread_jumps::execute): Same.
18381 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
18383 * tree-ssa-dom.c (class dom_jump_threader_simplifier): New.
18384 (class dom_opt_dom_walker): Initialize some class variables.
18385 (pass_dominator::execute): Pass evrp_range_analyzer and
18386 dom_jump_threader_simplifier to dom_opt_dom_walker.
18387 Adjust for some functions moving into classes.
18388 (simplify_stmt_for_jump_threading): Adjust and move to...
18389 (jump_threader_simplifier::simplify): ...here.
18390 (dom_opt_dom_walker::before_dom_children): Adjust for
18391 m_evrp_range_analyzer.
18392 (dom_opt_dom_walker::after_dom_children): Remove x_vr_values hack.
18393 (test_for_singularity): Place in dom_opt_dom_walker class.
18394 (dom_opt_dom_walker::optimize_stmt): The argument
18395 evrp_range_analyzer is now a class field.
18396 * tree-ssa-threadbackward.c (class thread_jumps): Add m_registry.
18397 (thread_jumps::thread_through_all_blocks): New.
18398 (thread_jumps::convert_and_register_current_path): Use m_registry.
18399 (pass_thread_jumps::execute): Adjust for thread_through_all_blocks
18400 being in the threader class.
18401 (pass_early_thread_jumps::execute): Same.
18402 * tree-ssa-threadedge.c (threadedge_initialize_values): Move...
18403 (jump_threader::jump_threader): ...here.
18404 (threadedge_finalize_values): Move...
18405 (jump_threader::~jump_threader): ...here.
18406 (jump_threader::remove_jump_threads_including): New.
18407 (jump_threader::thread_through_all_blocks): New.
18408 (record_temporary_equivalences_from_phis): Move...
18409 (jump_threader::record_temporary_equivalences_from_phis): ...here.
18410 (record_temporary_equivalences_from_stmts_at_dest): Move...
18411 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
18413 (simplify_control_stmt_condition_1): Move to jump_threader class.
18414 (simplify_control_stmt_condition): Move...
18415 (jump_threader::simplify_control_stmt_condition): ...here.
18416 (thread_around_empty_blocks): Move...
18417 (jump_threader::thread_around_empty_blocks): ...here.
18418 (thread_through_normal_block): Move...
18419 (jump_threader::thread_through_normal_block): ...here.
18420 (thread_across_edge): Move...
18421 (jump_threader::thread_across_edge): ...here.
18422 (thread_outgoing_edges): Move...
18423 (jump_threader::thread_outgoing_edges): ...here.
18424 * tree-ssa-threadedge.h: Move externally facing functings...
18425 (class jump_threader): ...here...
18426 (class jump_threader_simplifier): ...and here.
18427 * tree-ssa-threadupdate.c (struct redirection_data): Remove comment.
18428 (jump_thread_path_allocator::jump_thread_path_allocator): New.
18429 (jump_thread_path_allocator::~jump_thread_path_allocator): New.
18430 (jump_thread_path_allocator::allocate_thread_edge): New.
18431 (jump_thread_path_allocator::allocate_thread_path): New.
18432 (jump_thread_path_registry::jump_thread_path_registry): New.
18433 (jump_thread_path_registry::~jump_thread_path_registry): New.
18434 (jump_thread_path_registry::allocate_thread_edge): New.
18435 (jump_thread_path_registry::allocate_thread_path): New.
18436 (dump_jump_thread_path): Make extern.
18437 (debug (const vec<jump_thread_edge *> &path)): New.
18438 (struct removed_edges): Move to tree-ssa-threadupdate.h.
18439 (struct thread_stats_d): Remove.
18440 (remove_ctrl_stmt_and_useless_edges): Make static.
18441 (lookup_redirection_data): Move...
18442 (jump_thread_path_registry::lookup_redirection_data): ...here.
18443 (ssa_redirect_edges): Make static.
18444 (thread_block_1): Move...
18445 (jump_thread_path_registry::thread_block_1): ...here.
18446 (thread_block): Move...
18447 (jump_thread_path_registry::thread_block): ...here.
18448 (thread_through_loop_header): Move...
18449 (jump_thread_path_registry::thread_through_loop_header): ...here.
18450 (mark_threaded_blocks): Move...
18451 (jump_thread_path_registry::mark_threaded_blocks): ...here.
18452 (debug_path): Move...
18453 (jump_thread_path_registry::debug_path): ...here.
18454 (debug_all_paths): Move...
18455 (jump_thread_path_registry::dump): ..here.
18456 (rewire_first_differing_edge): Move...
18457 (jump_thread_path_registry::rewire_first_differing_edge): ...here.
18458 (adjust_paths_after_duplication): Move...
18459 (jump_thread_path_registry::adjust_paths_after_duplication): ...here.
18460 (duplicate_thread_path): Move...
18461 (jump_thread_path_registry::duplicate_thread_path): ..here.
18462 (remove_jump_threads_including): Move...
18463 (jump_thread_path_registry::remove_jump_threads_including): ...here.
18464 (thread_through_all_blocks): Move to...
18465 (jump_thread_path_registry::thread_through_all_blocks): ...here.
18466 (delete_jump_thread_path): Remove.
18467 (register_jump_thread): Move...
18468 (jump_thread_path_registry::register_jump_thread): ...here.
18469 * tree-ssa-threadupdate.h: Move externally facing functions...
18470 (class jump_thread_path_allocator): ...here...
18471 (class jump_thread_path_registry): ...and here.
18472 (thread_through_all_blocks): Remove.
18473 (struct removed_edges): New.
18474 (register_jump_thread): Remove.
18475 (remove_jump_threads_including): Remove.
18476 (delete_jump_thread_path): Remove.
18477 (remove_ctrl_stmt_and_useless_edges): Remove.
18478 (free_dom_edge_info): New prototype.
18479 * tree-vrp.c: Remove x_vr_values hack.
18480 (class vrp_jump_threader_simplifier): New.
18481 (vrp_jump_threader_simplifier::simplify): New.
18482 (vrp_jump_threader::vrp_jump_threader): Adjust method signature.
18483 Remove m_dummy_cond.
18484 Instantiate m_simplifier and m_threader.
18485 (vrp_jump_threader::thread_through_all_blocks): New.
18486 (vrp_jump_threader::simplify_stmt): Remove.
18487 (vrp_jump_threader::after_dom_children): Do not set m_dummy_cond.
18488 Remove x_vr_values hack.
18489 (execute_vrp): Adjust for thread_through_all_blocks being in a
18492 2021-04-30 Christophe Lyon <christophe.lyon@linaro.org>
18494 * genflags.c (gen_insn): Print failed expansion string.
18496 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
18498 * expr.c (alignment_for_piecewise_move): Call mode_for_size
18499 without limit to MAX_FIXED_MODE_SIZE.
18501 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
18503 PR middle-end/90773
18504 * builtins.c (builtin_memset_gen_str): Don't use return from
18505 simplify_gen_subreg.
18507 2021-04-30 Uroš Bizjak <ubizjak@gmail.com>
18510 * config/i386/i386.md (*add<mode>3_carry_0r): New insn pattern.
18511 (*addsi3_carry_zext_0r): Ditto.
18512 (*sub<mode>3_carry_0): Ditto.
18513 (*subsi3_carry_zext_0r): Ditto.
18514 * config/i386/predicates.md (ix86_carry_flag_unset_operator):
18516 * config/i386/i386.c (ix86_rtx_costs) <case PLUS, case MINUS>:
18517 Also consider ix86_carry_flag_unset_operator to calculate
18518 the cost of adc/sbb insn.
18520 2021-04-30 Roman Zhuykov <zhroma@ispras.ru>
18522 PR rtl-optimization/100225
18523 PR rtl-optimization/84878
18524 * modulo-sched.c (sms_schedule): Use note_stores to skip loops
18525 where we have an instruction which touches (writes) any hard
18526 register from df->regular_block_artificial_uses set.
18527 Allow not-single-set instruction only right before basic block
18530 2021-04-30 Geng Qi <gengqi@linux.alibaba.com>
18532 * config/riscv/riscv.opt (march=,mabi=): Negative itself.
18534 2021-04-30 LevyHsu <admin@levyhsu.com>
18536 * config/riscv/riscv.c (riscv_min_arithmetic_precision): New.
18537 * config/riscv/riscv.h (TARGET_MIN_ARITHMETIC_PRECISION): New.
18538 * config/riscv/riscv.md (addv<mode>4, uaddv<mode>4): New.
18539 (subv<mode>4, usubv<mode>4, mulv<mode>4, umulv<mode>4): New.
18541 2021-04-29 Alexandre Oliva <oliva@adacore.com>
18543 * config.gcc: Merged x86 and x86_64 cpu_type-setting cases.
18545 2021-04-29 Alexandre Oliva <oliva@adacore.com>
18547 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Rename to...
18548 (ASM_OUTPUT_MAX_SKIP_ALIGN): ... this. Enclose in do/while(0).
18549 * config/i386/i386.c: Adjust.
18550 * config/i386/i386.md: Adjust.
18551 * config/i386/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Drop.
18552 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18553 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18554 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18555 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18556 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18557 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18558 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18559 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18560 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18561 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
18563 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
18565 * config/i386/i386-expand.c (ix86_expand_int_compare):
18566 Swap operands of GTU and LEU comparison to emit carry flag comparison.
18567 * config/i386/i386.md (*add<mode>3_carry_0): Change insn
18568 predicate to allow more combine opportunities with memory operands.
18569 (*sub<mode>3_carry_0): Ditto.
18571 2021-04-29 Richard Sandiford <richard.sandiford@arm.com>
18573 PR rtl-optimization/100303
18574 * rtl-ssa/accesses.cc (function_info::make_use_available): Take a
18575 boolean that indicates whether the use will only be used in
18576 debug instructions. Treat it in the same way that existing
18577 cross-EBB debug references would be handled if so.
18578 (function_info::make_uses_available): Likewise.
18579 * rtl-ssa/functions.h (function_info::make_uses_available): Update
18580 prototype accordingly.
18581 (function_info::make_uses_available): Likewise.
18582 * fwprop.c (try_fwprop_subst): Update call accordingly.
18584 2021-04-29 Jeff Law <jlaw@tachyum.com>
18586 * config/nios2/nios2-protos.h (nios2_fpu_insn_enabled): Move outside
18589 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
18590 Richard Biener <rguenther@suse.de>
18593 * config/i386/i386-builtin.def (IX86_BUILTIN_MASKLOADPD)
18594 (IX86_BUILTIN_MASKLOADPS, IX86_BUILTIN_MASKLOADPD256)
18595 (IX86_BUILTIN_MASKLOADPS256, IX86_BUILTIN_MASKLOADD)
18596 (IX86_BUILTIN_MASKLOADQ, IX86_BUILTIN_MASKLOADD256)
18597 (IX86_BUILTIN_MASKLOADQ256): Move from SPECIAL_ARGS
18598 to PURE_ARGS category.
18599 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
18600 Handle PURE_ARGS category.
18601 * config/i386/i386-expand.c (ix86_expand_builtin): Ditto.
18603 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
18605 * configure.ac: Check for the presence of sys/locking.h header and
18606 for whether _LK_LOCK is supported by _locking.
18607 * configure: Regenerate.
18608 * config.in: Likewise.
18609 * gcov-io.h: Define GCOV_LOCKED_WITH_LOCKING if HOST_HAS_LK_LOCK.
18610 * gcov-io.c (gcov_open): Add support for GCOV_LOCKED_WITH_LOCKING.
18611 * system.h: Include <sys/locking.h> if HAVE_SYS_LOCKING_H.
18613 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
18615 * config/i386/predicates.md (fcmov_comparison_operator):
18616 Do not check for trivial FP comparison operator.
18617 <case GEU, case LTU>: Allow CCGZmode.
18618 <case GTU, case LEU>: Do not allow CCCmode.
18619 (ix86_comparison_operator) <case GTU, case LEU>: Allow only CCmode.
18620 (ix86_carry_flag_operator): Match only LTU and UNLT code.
18621 Do not check for trivial FP comparison operator. Allow CCGZmode.
18623 2021-04-29 Tom de Vries <tdevries@suse.de>
18625 * omp-expand.c (expand_omp_simd): Add step_orig, and replace uses of
18626 fd->loop.step by either step or orig_step.
18628 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
18630 * config/sparc/sparc.c (gen_load_pcrel_sym): Delete.
18631 (load_got_register): Do the PIC dance here.
18632 (sparc_legitimize_tls_address): Simplify.
18633 (sparc_emit_probe_stack_range): Likewise.
18634 (sparc32_initialize_trampoline): Likewise.
18635 (sparc64_initialize_trampoline): Likewise.
18636 * config/sparc/sparc.md (load_pcrel_sym<P:mode>): Add @ marker.
18637 (probe_stack_range<P:mode>): Likewise.
18638 (flush<P:mode>): Likewise.
18639 (tgd_hi22<P:mode>): Likewise.
18640 (tgd_lo10<P:mode>): Likewise.
18641 (tgd_add<P:mode>): Likewise.
18642 (tgd_call<P:mode>): Likewise.
18643 (tldm_hi22<P:mode>): Likewise.
18644 (tldm_lo10<P:mode>): Likewise.
18645 (tldm_add<P:mode>): Likewise.
18646 (tldm_call<P:mode>): Likewise.
18647 (tldo_hix22<P:mode>): Likewise.
18648 (tldo_lox10<P:mode>): Likewise.
18649 (tldo_add<P:mode>): Likewise.
18650 (tie_hi22<P:mode>): Likewise.
18651 (tie_lo10<P:mode>): Likewise.
18652 (tie_add<P:mode>): Likewise.
18653 (tle_hix22<P:mode>): Likewise.
18654 (tle_lox10<P:mode>): Likewise.
18655 (stack_protect_setsi): Rename to...
18656 (stack_protect_set32): ...this.
18657 (stack_protect_setdi): Rename to...
18658 (stack_protect_set64): ...this.
18659 (stack_protect_set): Adjust calls to above.
18660 (stack_protect_testsi): Rename to...
18661 (stack_protect_test32): ...this.
18662 (stack_protect_testdi): Rename to...
18663 (stack_protect_test64): ...this.
18664 (stack_protect_test): Adjust calls to above.
18666 2021-04-29 H.J. Lu <hjl.tools@gmail.com>
18668 PR middle-end/90773
18669 * builtins.c (builtin_memcpy_read_str): Add a dummy argument.
18670 (builtin_strncpy_read_str): Likewise.
18671 (builtin_memset_read_str): Add an argument for the previous RTL
18672 information and generate the new RTL from the previous RTL info.
18673 (builtin_memset_gen_str): Likewise.
18674 * builtins.h (builtin_strncpy_read_str): Update the prototype.
18675 (builtin_memset_read_str): Likewise.
18676 * expr.c (by_pieces_ninsns): If targetm.overlap_op_by_pieces_p()
18677 returns true, round up size and alignment to the widest integer
18678 mode for maximum size.
18679 (pieces_addr::adjust): Add a pointer to by_pieces_prev argument
18680 and pass it to m_constfn.
18681 (op_by_pieces_d): Add m_push and m_overlap_op_by_pieces.
18682 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
18683 initialize m_push. Initialize m_overlap_op_by_pieces with
18684 targetm.overlap_op_by_pieces_p ().
18685 (op_by_pieces_d::run): Pass the previous RTL information to
18686 pieces_addr::adjust and generate overlapping operations if
18687 m_overlap_op_by_pieces is true.
18689 (move_by_pieces_d::move_by_pieces_d): Updated for op_by_pieces_d
18691 (store_by_pieces_d::store_by_pieces_d): Updated for op_by_pieces_d
18693 (can_store_by_pieces): Use by_pieces_constfn on constfun.
18694 (store_by_pieces): Use by_pieces_constfn on constfun. Updated
18695 for op_by_pieces_d change.
18696 (clear_by_pieces_1): Add a dummy argument.
18697 (clear_by_pieces): Updated for op_by_pieces_d change.
18698 (compare_by_pieces_d::compare_by_pieces_d): Likewise.
18699 (string_cst_read_str): Add a dummy argument.
18700 * expr.h (by_pieces_constfn): Add a dummy argument.
18701 (by_pieces_prev): New.
18702 * target.def (overlap_op_by_pieces_p): New target hook.
18703 * config/i386/i386.c (TARGET_OVERLAP_OP_BY_PIECES_P): New.
18704 * doc/tm.texi.in: Add TARGET_OVERLAP_OP_BY_PIECES_P.
18705 * doc/tm.texi: Regenerated.
18707 2021-04-29 Richard Biener <rguenther@suse.de>
18709 PR tree-optimization/100253
18710 * tree-vect-stmts.c (vectorizable_load): Do not assume
18711 element alignment when DR_MISALIGNMENT is -1.
18712 (vectorizable_store): Likewise.
18714 2021-04-29 Jakub Jelinek <jakub@redhat.com>
18717 * config/aarch64/aarch64.c (aarch64_add_offset_1_temporaries): Use
18718 absu_hwi instead of abs_hwi.
18720 2021-04-29 Richard Biener <rguenther@suse.de>
18722 PR middle-end/38474
18723 * tree-ssa-structalias.c (add_graph_edge): Avoid direct
18724 forwarding when indirect forwarding through ESCAPED
18727 2021-04-29 Tom de Vries <tdevries@suse.de>
18730 * internal-fn.c (expand_GOMP_SIMT_ENTER_ALLOC)
18731 (expand_GOMP_SIMT_LAST_LANE, expand_GOMP_SIMT_ORDERED_PRED)
18732 (expand_GOMP_SIMT_VOTE_ANY, expand_GOMP_SIMT_XCHG_BFLY)
18733 (expand_GOMP_SIMT_XCHG_IDX): Ensure target is assigned to.
18735 2021-04-29 Richard Biener <rguenther@suse.de>
18737 PR tree-optimization/99912
18738 * tree-ssa-dse.c (dse_dom_walker::m_need_cfg_cleanup): New.
18739 (dse_dom_walker::todo): Likewise.
18740 (dse_dom_walker::dse_optimize_stmt): Move VDEF check to the
18742 (dse_dom_walker::before_dom_children): Remove trivially
18743 dead SSA defs and schedule CFG cleanup if we removed all
18745 (pass_dse::execute): Get TODO as computed by the DOM walker
18746 and return it. Wipe dominator info earlier.
18748 2021-04-29 Richard Biener <rguenther@suse.de>
18751 * ipa-prop.c (ipcp_modif_dom_walker::before_dom_children):
18752 Track blocks to cleanup EH in new m_need_eh_cleanup.
18753 (ipcp_modif_dom_walker::cleanup_eh): New.
18754 (ipcp_transform_function): Release dominator info before
18757 2021-04-29 Martin Sebor <msebor@redhat.com>
18759 PR middle-end/100250
18760 * attribs.c (attr_access::array_as_string): Avoid dereferencing
18761 a pointer when it's null.
18763 2021-04-29 Martin Sebor <msebor@redhat.com>
18765 * Makefile.in (OBJS): Add ipa-free-lang-data.o.
18766 * ipa-free-lang-data.cc: New file.
18767 * tree.c: Move pass free_lang_data to file above.
18768 (build_array_type_1): Declare extern.
18769 * tree.h (build_array_type_1): Declare.
18771 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18773 * config/aarch64/aarch64-simd-builtins.def: Modify comment to
18774 make consistent with updated RTL pattern.
18775 * config/aarch64/aarch64-simd.md (aarch64_<sur>qmovn<mode>):
18776 Implement using ss_truncate and us_truncate rather than
18778 * config/aarch64/iterators.md: Remove redundant unspecs and
18779 iterator: UNSPEC_[SU]QXTN and SUQMOVN respectively.
18781 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18783 * config/aarch64/arm_acle.h (__attribute__): Make intrinsic
18784 attributes consistent with those defined in arm_neon.h.
18786 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18788 * config/aarch64/arm_fp16.h (__attribute__): Make intrinsic
18789 attributes consistent with those defined in arm_neon.h.
18791 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18793 * config/aarch64/aarch64-simd-builtins.def: Add
18794 float_trunc_rodd builtin generator macros.
18795 * config/aarch64/aarch64-simd.md (aarch64_float_trunc_rodd_df):
18797 (aarch64_float_trunc_rodd_lo_v2sf): Define.
18798 (aarch64_float_trunc_rodd_hi_v4sf_le): Define.
18799 (aarch64_float_trunc_rodd_hi_v4sf_be): Define.
18800 (aarch64_float_trunc_rodd_hi_v4sf): Define.
18801 * config/aarch64/arm_neon.h (vcvtx_f32_f64): Use RTL builtin
18802 instead of inline asm.
18803 (vcvtx_high_f32_f64): Likewise.
18804 (vcvtxd_f32_f64): Likewise.
18805 * config/aarch64/iterators.md: Add FCVTXN unspec.
18807 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18809 * config/aarch64/aarch64-simd-builtins.def: Add tbx1 builtin
18811 * config/aarch64/aarch64-simd.md (aarch64_tbx1<mode>):
18813 * config/aarch64/arm_neon.h (vqtbx1_s8): USE RTL builtin
18814 instead of inline asm.
18815 (vqtbx1_u8): Likewise.
18816 (vqtbx1_p8): Likewise.
18817 (vqtbx1q_s8): Likewise.
18818 (vqtbx1q_u8): Likewise.
18819 (vqtbx1q_p8): Likewise.
18820 (vtbx2_s8): Likewise.
18821 (vtbx2_u8): Likewise.
18822 (vtbx2_p8): Likewise.
18824 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18826 * config/aarch64/aarch64-simd-builtins.def: Add tbl1 builtin
18828 * config/aarch64/arm_neon.h (vqtbl1_p8): Use RTL builtin
18829 instead of inline asm.
18830 (vqtbl1_s8): Likewise.
18831 (vqtbl1_u8): Likewise.
18832 (vqtbl1q_p8): Likewise.
18833 (vqtbl1q_s8): Likewise.
18834 (vqtbl1q_u8): Likewise.
18835 (vtbl1_s8): Likewise.
18836 (vtbl1_u8): Likewise.
18837 (vtbl1_p8): Likewise.
18838 (vtbl2_s8): Likewise.
18839 (vtbl2_u8): Likewise.
18840 (vtbl2_p8): Likewise.
18842 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18844 * config/aarch64/aarch64-simd-builtins.def: Add polynomial
18845 ssri_n buitin generator macro.
18846 * config/aarch64/arm_neon.h (vsri_n_p8): Use RTL builtin
18847 instead of inline asm.
18848 (vsri_n_p16): Likewise.
18849 (vsri_n_p64): Likewise.
18850 (vsriq_n_p8): Likewise.
18851 (vsriq_n_p16): Likewise.
18852 (vsriq_n_p64): Likewise.
18854 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18856 * config/aarch64/aarch64-simd-builtins.def: Use VALLP mode
18857 iterator for polynomial ssli_n builtin generator macro.
18858 * config/aarch64/arm_neon.h (vsli_n_p8): Use RTL builtin
18859 instead of inline asm.
18860 (vsli_n_p16): Likewise.
18861 (vsliq_n_p8): Likewise.
18862 (vsliq_n_p16): Likewise.
18863 * config/aarch64/iterators.md: Define VALLP mode iterator.
18865 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18867 * config/aarch64/aarch64-simd-builtins.def: Use VDQV_L
18868 iterator to generate [su]adalp RTL builtins.
18869 * config/aarch64/aarch64-simd.md: Use VDQV_L iterator in
18870 [su]adalp RTL pattern.
18871 * config/aarch64/arm_neon.h (vpadal_s32): Use RTL builtin
18872 instead of inline asm.
18873 (vpadal_u32): Likewise.
18875 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18877 * config/aarch64/aarch64-simd-builtins.def: Add [su]addlp
18878 builtin generator macros.
18879 * config/aarch64/aarch64-simd.md (aarch64_<su>addlp<mode>):
18881 * config/aarch64/arm_neon.h (vpaddl_s8): Use RTL builtin
18882 instead of inline asm.
18883 (vpaddl_s16): Likewise.
18884 (vpaddl_s32): Likewise.
18885 (vpaddl_u8): Likewise.
18886 (vpaddl_u16): Likewise.
18887 (vpaddl_u32): Likewise.
18888 (vpaddlq_s8): Likewise.
18889 (vpaddlq_s16): Likewise.
18890 (vpaddlq_s32): Likewise.
18891 (vpaddlq_u8): Likewise.
18892 (vpaddlq_u16): Likewise.
18893 (vpaddlq_u32): Liwewise.
18894 * config/aarch64/iterators.md: Define [SU]ADDLP unspecs with
18895 appropriate attributes.
18897 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18899 * config/aarch64/aarch64-simd-builtins.def: Use VDQ_I iterator
18900 for aarch64_addp<mode> builtin macro generator.
18901 * config/aarch64/aarch64-simd.md: Use VDQ_I iterator in
18902 aarch64_addp<mode> RTL pattern.
18903 * config/aarch64/arm_neon.h (vpaddq_s8): Use RTL builtin
18904 instead of inline asm.
18905 (vpaddq_s16): Likewise.
18906 (vpaddq_s32): Likewise.
18907 (vpaddq_s64): Likewise.
18908 (vpaddq_u8): Likewise.
18909 (vpaddq_u16): Likewise.
18910 (vpaddq_u32): Likewise.
18911 (vpaddq_u64): Likewise.
18913 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18915 * config/aarch64/aarch64-simd-builtins.def: Add sq[r]dmulh_n
18916 builtin generator macros.
18917 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_n<mode>):
18919 * config/aarch64/arm_neon.h (vqdmulh_n_s16): Use RTL builtin
18920 instead of inline asm.
18921 (vqdmulh_n_s32): Likewise.
18922 (vqdmulhq_n_s16): Likewise.
18923 (vqdmulhq_n_s32): Likewise.
18924 (vqrdmulh_n_s16): Likewise.
18925 (vqrdmulh_n_s32): Likewise.
18926 (vqrdmulhq_n_s16): Likewise.
18927 (vqrdmulhq_n_s32): Likewise.
18929 2021-04-28 Tobias Burnus <tobias@codesourcery.com>
18931 * doc/install.texi (--enable-offload-defaulted): Document.
18933 2021-04-28 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
18935 * config/avr/avr-dimode.md: Turn existing patterns into
18936 define_insn_and_split style patterns where the splitter
18937 adds a clobber of the condition code register. Drop "cc"
18938 attribute. Add new patterns to match output of
18940 * config/avr/avr-fixed.md: Likewise.
18941 * config/avr/avr.c (cc_reg_rtx): New.
18942 (avr_parallel_insn_from_insns): Adjust insn count
18943 for removal of set of cc0.
18944 (avr_is_casesi_sequence): Likewise.
18945 (avr_casei_sequence_check_operands): Likewise.
18946 (avr_optimize_casesi): Likewise. Also insert
18947 new insns after jump_insn.
18948 (avr_pass_casesi::avr_rest_of_handle_casesi): Adjust
18949 for removal of set of cc0.
18950 (avr_init_expanders): Initialize cc_reg_rtx.
18951 (avr_regno_reg_class): Handle REG_CC.
18952 (cond_string): Remove usage of CC_OVERFLOW_UNUSABLE.
18953 (avr_notice_update_cc): Remove function.
18954 (ret_cond_branch): Remove usage of CC_OVERFLOW_UNUSABLE.
18955 (compare_condition): Adjust for PARALLEL with
18957 (out_shift_with_cnt): Likewise.
18958 (ashlhi3_out): Likewise.
18959 (ashrhi3_out): Likewise.
18960 (lshrhi3_out): Likewise.
18961 (avr_class_max_nregs): Return single reg for REG_CC.
18962 (avr_compare_pattern): Check for REG_CC instead
18964 (avr_reorg_remove_redundant_compare): Likewise.
18965 (avr_reorg):Adjust for PARALLEL with REG_CC clobber.
18966 (avr_hard_regno_nregs): Return single reg for REG_CC.
18967 (avr_hard_regno_mode_ok): Allow only CCmode for REG_CC.
18968 (avr_md_asm_adjust): Clobber REG_CC.
18969 (TARGET_HARD_REGNO_NREGS): Define.
18970 (TARGET_CLASS_MAX_NREGS): Define.
18971 (TARGET_MD_ASM_ADJUST): Define.
18972 * config/avr/avr.h (FIRST_PSEUDO_REGISTER): Adjust
18974 (enum reg_class): Add CC_REG class.
18975 (NOTICE_UPDATE_CC): Remove.
18976 (CC_OVERFLOW_UNUSABLE): Remove.
18977 (CC_NO_CARRY): Remove.
18978 * config/avr/avr.md: Turn existing patterns into
18979 define_insn_and_split style patterns where the splitter
18980 adds a clobber of the condition code register. Drop "cc"
18981 attribute. Add new patterns to match output of
18983 (sez): Remove unused pattern.
18985 2021-04-28 Richard Earnshaw <rearnsha@arm.com>
18988 * config/arm/arm.c (arm_hard_regno_mode_ok): Only allow VPR to be
18991 2021-04-28 Richard Sandiford <richard.sandiford@arm.com>
18994 * config/aarch64/constraints.md (Utq): Require the address to
18995 be valid for both the element mode and for V2DImode.
18997 2021-04-28 Jakub Jelinek <jakub@redhat.com>
18998 Tobias Burnus <tobias@codesourcery.com>
19000 * configure.ac (OFFLOAD_DEFAULTED): AC_DEFINE if offload-defaulted.
19001 * gcc.c (process_command): New variable.
19002 (driver::maybe_putenv_OFFLOAD_TARGETS): If OFFLOAD_DEFAULTED,
19003 set it if -foffload is defaulted.
19004 * lto-wrapper.c (OFFLOAD_TARGET_DEFAULT_ENV): Define.
19005 (compile_offload_image): If OFFLOAD_DEFAULTED and
19006 OFFLOAD_TARGET_DEFAULT is in the environment, don't fail
19007 if corresponding mkoffload can't be found.
19008 (compile_images_for_offload_targets): Likewise. Free and clear
19009 offload_names if no valid offload is found.
19010 * config.in: Regenerate.
19011 * configure: Regenerate.
19013 2021-04-28 Richard Biener <rguenther@suse.de>
19015 PR tree-optimization/100292
19016 * tree-vect-generic.c (expand_vector_condition): Do not fold
19019 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
19021 * config/rs6000/aix.h (SUBTARGET_DRIVER_SELF_SPECS): New.
19022 * config/rs6000/aix64.opt (m64): New.
19025 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
19027 * config/vax/vax.c (print_operand_address, vax_address_cost_1)
19028 (index_term_p): Handle ASHIFT too.
19030 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
19032 * config/vax/builtins.md (jbb<ccss>i<mode>): Remove operand #3.
19033 (sync_lock_test_and_set<mode>): Adjust accordingly.
19034 (sync_lock_release<mode>): Likewise.
19036 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
19038 * config/vax/vax-protos.h (adjacent_operands_p): Remove
19040 * config/vax/vax.c (adjacent_operands_p): Remove.
19042 2021-04-27 Maciej W. Rozycki <macro@linux-mips.org>
19044 * ifcvt.c (dead_or_predicable) [!IFCVT_MODIFY_TESTS]: Fall
19045 through to the non-conditional execution case if getting the
19046 condition for conditional execution has failed.
19048 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
19050 PR middle-end/100284
19051 * gimple.c (gimple_could_trap_p_1): Remove VEC_COND_EXPR test.
19052 * tree-eh.c (operation_could_trap_p): Handle VEC_COND_EXPR rather
19053 than asserting on it.
19055 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
19057 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): Protect
19058 with TARGET_AIX_OS.
19060 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
19063 * calls.c (precompute_register_parameters): Additionally test
19064 targetm.precompute_tls_p to pre-compute argument.
19065 * config/rs6000/aix.h (TARGET_PRECOMPUTE_TLS_P): Define.
19066 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): New.
19067 * target.def (precompute_tls_p): New.
19068 * doc/tm.texi.in (TARGET_PRECOMPUTE_TLS_P): Add hook documentation.
19069 * doc/tm.texi: Regenerated.
19071 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19074 * config/aarch64/aarch64.c (aarch64_print_operand): Cast -UINTVAL
19075 back to HOST_WIDE_INT.
19077 2021-04-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
19080 * simplify-rtx.c (simplify_context::simplify_subreg): Check the
19081 memory alignment for the outer mode.
19083 2021-04-27 H.J. Lu <hjl.tools@gmail.com>
19085 PR middle-end/90773
19086 * expr.c (op_by_pieces_d::get_usable_mode): New member function.
19087 (op_by_pieces_d::run): Cange a while loop to a do-while loop.
19089 2021-04-27 Alex Coplan <alex.coplan@arm.com>
19092 * config/arm/arm.c (arm_split_compare_and_swap): Fix up codegen
19093 with negative immediates: ensure we expand cbranchsi4_scratch
19094 correctly and ensure we satisfy its constraints.
19095 * config/arm/sync.md
19096 (@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): Don't
19097 attempt to tie two output operands together with constraints;
19098 collapse two alternatives.
19099 (@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
19100 * config/arm/thumb1.md (cbranchsi4_neg_late): New.
19102 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19105 * config/aarch64/predicates.md (aarch64_sub_immediate,
19106 aarch64_plus_immediate): Use -UINTVAL instead of -INTVAL.
19107 * config/aarch64/aarch64.md (casesi, rotl<mode>3): Likewise.
19108 * config/aarch64/aarch64.c (aarch64_print_operand,
19109 aarch64_split_atomic_op, aarch64_expand_subvti): Likewise.
19111 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19113 PR tree-optimization/100239
19114 * tree-vect-generic.c (lower_vec_perm): Don't accept constant
19115 permutations with all indices from the first zero element as vec_shl.
19117 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19119 PR rtl-optimization/100254
19120 * cfgcleanup.c (outgoing_edges_match): Check REG_EH_REGION on
19121 last1 and last2 insns rather than BB_END (bb1) and BB_END (bb2) insns.
19123 2021-04-27 Richard Biener <rguenther@suse.de>
19125 PR tree-optimization/99912
19126 * passes.def: Add comment about new TODO_remove_unused_locals.
19127 * tree-stdarg.c (pass_data_stdarg): Run TODO_remove_unused_locals
19130 2021-04-27 Richard Biener <rguenther@suse.de>
19132 PR tree-optimization/99912
19133 * passes.def (pass_all_optimizations): Add pass_dse before
19134 the first pass_dce, move the first pass_dse before the
19135 pass_dce following pass_pre.
19137 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19139 PR tree-optimization/95527
19140 * generic-match-head.c: Include tm.h.
19141 * gimple-match-head.c: Include tm.h.
19142 * match.pd (CLZ == INTEGER_CST): Don't use
19143 #ifdef CLZ_DEFINED_VALUE_AT_ZERO, only test CLZ_DEFINED_VALUE_AT_ZERO
19144 if clz == CFN_CLZ. Add missing val declaration.
19145 (CTZ cmp CST): New simplifications.
19147 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19149 PR tree-optimization/96696
19150 * expr.c (expand_expr_divmod): New function.
19151 (expand_expr_real_2) <case TRUNC_DIV_EXPR>: Use it for truncations and
19152 divisions. Formatting fixes.
19153 <case MULT_EXPR>: Optimize x / y * y as x - x % y if the latter is
19156 2021-04-27 Martin Jambor <mjambor@suse.cz>
19159 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
19160 If removing a call statement LHS SSA name, release it.
19162 2021-04-27 Richard Earnshaw <rearnsha@arm.com>
19165 * config/arm/arm.c (THUMB2_WORK_REGS): Check PIC_OFFSET_TABLE_REGNUM
19166 is valid before including it in the mask.
19168 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
19171 * config/aarch64/aarch64.c (aarch64_comp_type_attributes): Handle
19174 2021-04-27 Richard Biener <rguenther@suse.de>
19176 PR tree-optimization/100051
19177 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Add
19178 disambiguator based on access size vs. decl size.
19180 2021-04-27 Richard Biener <rguenther@suse.de>
19182 PR tree-optimization/100278
19183 * tree-ssa-pre.c (compute_avail): Give up when we cannot
19184 adjust TBAA beacuse of mismatching bases.
19186 2021-04-27 Jakub Jelinek <jakub@redhat.com>
19189 * config/i386/i386.md (*<insn><mode>3_mask, *<insn><mode>3_mask_1):
19190 For any_rotate define_insn_split and following splitters, use
19191 SWI iterator instead of SWI48.
19193 2021-04-27 Richard Biener <rguenther@suse.de>
19195 PR tree-optimization/99776
19196 * match.pd (bit_field_ref (ctor)): Relax element extract
19197 type compatibility checks.
19199 2021-04-27 Cui,Lili <lili.cui@intel.com>
19201 * common/config/i386/i386-common.c (processor_names):
19202 Sync processor_names with processor_type.
19203 * config/i386/i386-options.c (processor_cost_table):
19204 Sync processor_cost_table with processor_type.
19206 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
19208 * value-range.cc (irange::irange_set_1bit_anti_range): Add assert.
19209 (irange::set): Call irange_set_1bit_anti_range for handling all
19210 1-bit ranges. Fall through on ~[MIN,MAX].
19212 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
19214 * value-range.cc (irange::legacy_num_pairs): Remove.
19215 (irange::invert): Change gcc_assert to gcc_checking_assert.
19216 * value-range.h (irange::num_pairs): Adjust for a cached
19217 num_pairs(). Also, rename all gcc_assert's to
19218 gcc_checking_assert's.
19220 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
19222 * value-range.cc (irange::operator=): Set m_kind.
19223 (irange::copy_to_legacy): Handle varying and undefined sources
19224 as a legacy copy since they can be easily copied.
19225 (irange::irange_set): Set m_kind.
19226 (irange::irange_set_anti_range): Same.
19227 (irange::set): Rename normalize_min_max to normalize_kind.
19228 (irange::verify_range): Adjust for multi-ranges having the
19230 (irange::irange_union): Set m_kind.
19231 (irange::irange_intersect): Same.
19232 (irange::invert): Same.
19233 * value-range.h (irange::kind): Always return m_kind.
19234 (irange::varying_p): Rename to...
19235 (irange::varying_comptaible_p): ...this.
19236 (irange::undefined_p): Only look at m_kind.
19237 (irange::irange): Always set VR_UNDEFINED if applicable.
19238 (irange::set_undefined): Always set VR_UNDEFINED.
19239 (irange::set_varying): Always set m_kind to VR_VARYING.
19240 (irange::normalize_min_max): Rename to...
19241 (irange::normalize_kind): ...this.
19243 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
19245 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
19246 Adjust for constant_p including varying_p.
19247 * tree-vrp.c (vrp_prop::finalize): Same.
19248 (determine_value_range): Same.
19249 * vr-values.c (vr_values::range_of_expr): Same.
19250 * value-range.cc (irange::symbolic_p): Do not check varying_p.
19251 (irange::constant_p): Same.
19253 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
19255 * value-range.cc (irange::legacy_lower_bound): Replace
19256 !undefined_p check with num_ranges > 0.
19257 (irange::legacy_upper_bound): Same.
19258 * value-range.h (irange::type): Same.
19259 (irange::lower_bound): Same.
19260 (irange::upper_bound): Same.
19262 2021-04-26 Richard Biener <rguenther@suse.de>
19264 PR tree-optimization/99956
19265 * gimple-loop-interchange.cc (compute_access_stride):
19266 Try instantiating the access in a shallower loop nest
19267 if instantiating failed.
19268 (compute_access_strides): Pass adjustable loop_nest
19269 to compute_access_stride.
19271 2021-04-26 Christophe Lyon <christophe.lyon@linaro.org>
19273 * doc/sourcebuild.texi (arm_cmse_hw): Document.
19275 2021-04-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19277 * config/aarch64/iterators.md (vwcore): Handle V4BF, V8BF.
19279 2021-04-26 Thomas Schwinge <thomas@codesourcery.com>
19280 Nathan Sidwell <nathan@codesourcery.com>
19281 Tom de Vries <vries@codesourcery.com>
19282 Julian Brown <julian@codesourcery.com>
19283 Kwok Cheung Yeung <kcy@codesourcery.com>
19285 * omp-offload.c (oacc_validate_dims): Implement
19286 '-Wopenacc-parallelism'.
19287 * doc/invoke.texi (-Wopenacc-parallelism): Document.
19289 2021-04-26 Richard Biener <rguenther@suse.de>
19291 * tree-cfg.h (gimplify_build1): Remove.
19292 (gimplify_build2): Likewise.
19293 (gimplify_build3): Likewise.
19294 * tree-cfg.c (gimplify_build1): Move to tree-vect-generic.c.
19295 (gimplify_build2): Likewise.
19296 (gimplify_build3): Likewise.
19297 * tree-vect-generic.c (gimplify_build1): Move from tree-cfg.c.
19299 (gimplify_build2): Likewise.
19300 (gimplify_build3): Likewise.
19301 (tree_vec_extract): Use resimplify with following SSA edges.
19302 (expand_vector_parallel): Avoid passing NULL size/bitpos
19303 to tree_vec_extract.
19304 * expr.c (store_constructor): Deal with zero-element CTORs.
19305 * match.pd (bit_field_ref <vector CTOR>): Make sure to
19306 produce vector constants when possible.
19308 2021-04-26 Richard Biener <rguenther@suse.de>
19310 * tree-complex.c: Include gimple-fold.h.
19311 (expand_complex_addition): Use gimple_build.
19312 (expand_complex_multiplication_components): Likewise.
19313 (expand_complex_multiplication): Likewise.
19314 (expand_complex_div_straight): Likewise.
19315 (expand_complex_div_wide): Likewise.
19316 (expand_complex_division): Likewise.
19317 (expand_complex_conjugate): Likewise.
19318 (expand_complex_comparison): Likewise.
19320 2021-04-26 Richard Biener <rguenther@suse.de>
19322 * tree-ssa-phiopt.c (two_value_replacement): Remove use
19323 of legacy gimplify_buildN API.
19325 2021-04-26 Richard Biener <rguenther@suse.de>
19327 PR tree-optimization/99473
19328 * tree-ssa-phiopt.c (cond_store_replacement): Handle all
19331 2021-04-26 Richard Biener <rguenther@suse.de>
19333 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin):
19334 Use replace_call_with_value.
19336 2021-04-26 Richard Biener <rguenther@suse.de>
19338 * tree-ssa-propagate.h (valid_gimple_rhs_p): Remove.
19339 (update_gimple_call): Likewise.
19340 (update_call_from_tree): Likewise.
19341 * tree-ssa-propagate.c (valid_gimple_rhs_p): Remove.
19342 (valid_gimple_call_p): Likewise.
19343 (move_ssa_defining_stmt_for_defs): Likewise.
19344 (finish_update_gimple_call): Likewise.
19345 (update_gimple_call): Likewise.
19346 (update_call_from_tree): Likewise.
19347 (propagate_tree_value_into_stmt): Use replace_call_with_value.
19348 * gimple-fold.h (update_gimple_call): Declare.
19349 * gimple-fold.c (valid_gimple_rhs_p): Move here from
19350 tree-ssa-propagate.c.
19351 (update_gimple_call): Likewise.
19352 (valid_gimple_call_p): Likewise.
19353 (finish_update_gimple_call): Likewise, and simplify.
19354 (gimplify_and_update_call_from_tree): Implement
19355 update_call_from_tree functionality, avoid excessive
19356 push/pop_gimplify_context.
19357 (gimple_fold_builtin): Use only gimplify_and_update_call_from_tree.
19358 (gimple_fold_call): Likewise.
19359 * gimple-ssa-sprintf.c (try_substitute_return_value): Likewise.
19360 * tree-ssa-ccp.c (ccp_folder::fold_stmt): Likewise.
19361 (pass_fold_builtins::execute): Likewise.
19362 (optimize_stack_restore): Use replace_call_with_value.
19363 * tree-cfg.c (fold_loop_internal_call): Likewise.
19364 * tree-ssa-dce.c (maybe_optimize_arith_overflow): Use
19365 only gimplify_and_update_call_from_tree.
19366 * tree-ssa-strlen.c (handle_builtin_strlen): Likewise.
19367 (handle_builtin_strchr): Likewise.
19368 * tsan.c: Include gimple-fold.h instead of tree-ssa-propagate.h.
19370 2021-04-26 Jakub Jelinek <jakub@redhat.com>
19373 * vmsdbgout.c (ASM_OUTPUT_DEBUG_STRING, vmsdbgout_begin_block,
19374 vmsdbgout_end_block, lookup_filename, vmsdbgout_source_line): Remove
19377 2021-04-25 liuhongt <hongtao.liu@intel.com>
19380 * config/i386/i386-builtin.def (BDESC): Change the icode of
19381 the following builtins to CODE_FOR_nothing.
19382 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
19383 IX86_BUILTIN_PCMPEQB128, IX86_BUILTIN_PCMPEQW128,
19384 IX86_BUILTIN_PCMPEQD128, IX86_BUILTIN_PCMPEQQ,
19385 IX86_BUILTIN_PCMPEQB256, IX86_BUILTIN_PCMPEQW256,
19386 IX86_BUILTIN_PCMPEQD256, IX86_BUILTIN_PCMPEQQ256,
19387 IX86_BUILTIN_PCMPGTB128, IX86_BUILTIN_PCMPGTW128,
19388 IX86_BUILTIN_PCMPGTD128, IX86_BUILTIN_PCMPGTQ,
19389 IX86_BUILTIN_PCMPGTB256, IX86_BUILTIN_PCMPGTW256,
19390 IX86_BUILTIN_PCMPGTD256, IX86_BUILTIN_PCMPGTQ256.
19391 * config/i386/sse.md (avx2_eq<mode>3): Deleted.
19392 (sse2_eq<mode>3): Ditto.
19393 (sse4_1_eqv2di3): Ditto.
19394 (sse2_gt<mode>3): Rename to ..
19395 (*sse2_gt<mode>3): .. this.
19397 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
19400 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
19403 * config/darwin.c (darwin_binds_local_p): Assume that any
19404 public symbol might be interposed for PIC code. Update function
19405 header comment to reflect current Darwin capability.
19407 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
19410 * config/darwin.c (darwin_binds_local_p): Assume that any
19411 public symbol might be interposed for PIC code. Update function
19412 header comment to reflect current Darwin capability.
19414 2021-04-24 Richard Sandiford <richard.sandiford@arm.com>
19416 * doc/sourcebuild.texi: Document no-opts and any-opts target
19419 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
19421 * config/bpf/bpf.h (ASM_OUTPUT_ALIGNED_BSS): Use .type and .lcomm.
19423 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
19425 * config/bpf/bpf.h (FUNCTION_BOUNDARY): Set to 64.
19427 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
19430 * config/i386/i386-options.c (ix86_option_override_internal):
19431 Error out when -m96bit-long-double is used with 64bit targets.
19432 * config/i386/i386.md (*pushxf_rounded): Remove pattern.
19434 2021-04-23 Martin Liska <mliska@suse.cz>
19436 * lto-wrapper.c: Remove FIXME about usage of
19437 hardware_concurrency. The function is not on par with
19440 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
19443 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
19444 Copy operand 3 to operand 4. Use sse_reg_operand
19445 as operand 3 predicate.
19446 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): Ditto.
19447 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
19448 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
19449 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2):
19450 Copy operand 1 to operand 0.
19451 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): Ditto.
19452 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
19453 (LDX_ATOMIC/LDX_ATOMIC FP store peephole2 with mem blockage): Ditto.
19455 2021-04-23 Alex Coplan <alex.coplan@arm.com>
19457 PR rtl-optimization/100230
19458 * early-remat.c (early_remat::sort_candidates): Use delete[]
19459 instead of delete for array allocated with new[].
19461 2021-04-23 Richard Biener <rguenther@suse.de>
19463 * genmatch.c (lower_cond): Remove VEC_COND_EXPR special-casing.
19464 (capture_info::capture_info): Likewise.
19465 (capture_info::walk_match): Likewise.
19466 (expr::gen_transform): Likewise.
19467 (dt_simplify::gen_1): Likewise.
19468 * gimple-match-head.c (maybe_resimplify_conditional_op):
19469 Remove VEC_COND_EXPR special-casing.
19470 (gimple_simplify): Likewise.
19471 * gimple.c (gimple_could_trap_p_1): Adjust.
19472 * tree-ssa-pre.c (compute_avail): Allow VEC_COND_EXPR
19473 to participate in PRE.
19475 2021-04-23 Richard Biener <rguenther@suse.de>
19477 * cfganal.c (connect_infinite_loops_to_exit): First call
19478 add_noreturn_fake_exit_edges.
19479 * ipa-sra.c (process_scan_results): Do not call the now redundant
19480 add_noreturn_fake_exit_edges.
19481 * predict.c (tree_estimate_probability): Likewise.
19482 (rebuild_frequencies): Likewise.
19483 * store-motion.c (one_store_motion_pass): Likewise.
19485 2021-04-23 Richard Biener <rguenther@suse.de>
19487 PR tree-optimization/100222
19488 * predict.c (pass_profile::execute): Remove redundant call to
19489 mark_irreducible_loops.
19490 (report_predictor_hitrates): Likewise.
19492 2021-04-23 Richard Biener <rguenther@suse.de>
19494 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Avoid
19495 valid_gimple_rhs_p by instead gimplifying to one.
19497 2021-04-23 Richard Biener <rguenther@suse.de>
19499 PR tree-optimization/99971
19500 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
19501 Always use TBAA for loads.
19503 2021-04-23 liuhongt <hongtao.liu@intel.com>
19506 * config/i386/i386-options.c (ix86_option_override_internal):
19507 Clear MASK_AVX256_SPLIT_UNALIGNED_LOAD/STORE in x_target_flags
19508 when X86_TUNE_AVX256_UNALIGNED_LOAD/STORE_OPTIMAL is enabled
19509 by target attribute.
19511 2021-04-23 David Edelsohn <dje.gcc@gmail.com>
19513 * config/rs6000/aix71.h (PREFERRED_DEBUGGING_TYPE): Change to
19515 * config/rs6000/aix72.h (PREFERRED_DEBUGGING_TYPE): Same.
19517 2021-04-22 David Edelsohn <dje.gcc@gmail.com>
19519 * config.gcc (powerpc-ibm-aix6.*): Remove.
19520 * config/rs6000/aix61.h: Delete.
19522 2021-04-22 Martin Liska <mliska@suse.cz>
19524 PR testsuite/100159
19525 PR testsuite/100192
19526 * builtins.c (expand_builtin): Fix typos and missing comments.
19527 * dwarf2out.c (gen_subprogram_die): Likewise.
19528 (gen_struct_or_union_type_die): Likewise.
19530 2021-04-22 Uroš Bizjak <ubizjak@gmail.com>
19533 * config/i386/i386-expand.c (ix86_expand_convert_uns_sidf_sse):
19534 Remove the sign with FE_DOWNWARD, where x - x = -0.0.
19536 2021-04-21 Iain Sandoe <iain@sandoe.co.uk>
19538 * config/i386/darwin.h (TARGET_64BIT): Remove definition
19539 based on TARGET_ISA_64BIT.
19540 (TARGET_64BIT_P): Remove definition based on
19541 TARGET_ISA_64BIT_P().
19543 2021-04-21 Martin Liska <mliska@suse.cz>
19546 2021-04-21 Martin Liska <mliska@suse.cz>
19548 * lto-wrapper.c (cpuset_popcount): Remove.
19549 (init_num_threads): Remove and use hardware_concurrency.
19551 2021-04-21 Martin Liska <mliska@suse.cz>
19554 * main.c (main): Call toplev::finalize in CHECKING_P mode.
19555 * ipa-modref.c (ipa_modref_c_finalize): summaries are NULL
19556 when incremental LTO linking happens.
19558 2021-04-21 Martin Liska <mliska@suse.cz>
19560 * lto-wrapper.c (run_gcc): When -flto=jobserver is used, but the
19561 makeserver cannot be detected, then use -flto=N fallback.
19563 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
19565 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): When cross-compiling,
19566 default to yes for aarch64-linux-gnu.
19567 * configure: Regenerate.
19569 2021-04-21 Martin Liska <mliska@suse.cz>
19571 * lto-wrapper.c (cpuset_popcount): Remove.
19572 (init_num_threads): Remove and use hardware_concurrency.
19574 2021-04-21 Martin Liska <mliska@suse.cz>
19576 * config/i386/i386.c: Remove superfluous || TARGET_MACHO
19577 which remains to be '(... || 0)' and clang complains about it.
19578 * dwarf2out.c (AT_vms_delta): Declare conditionally.
19579 (add_AT_vms_delta): Likewise.
19580 * tree.c (fld_simplified_type): Use rather more common pattern
19581 for disabling of something (#if 0).
19582 (get_tree_code_name): Likewise.
19583 (verify_type_variant): Likewise.
19585 2021-04-21 Martin Liska <mliska@suse.cz>
19587 * config/i386/i386-expand.c (decide_alignment): Use newly named
19588 macro TARGET_CPU_P.
19589 * config/i386/i386.c (ix86_decompose_address): Likewise.
19590 (ix86_address_cost): Likewise.
19591 (ix86_lea_outperforms): Likewise.
19592 (ix86_avoid_lea_for_addr): Likewise.
19593 (ix86_add_stmt_cost): Likewise.
19594 * config/i386/i386.h (TARGET_*): Remove.
19595 (TARGET_CPU_P): New macro.
19596 * config/i386/i386.md: Use newly named macro TARGET_CPU_P.
19597 * config/i386/x86-tune-sched-atom.c (do_reorder_for_imul): Likewise.
19598 (swap_top_of_ready_list): Likewise.
19599 (ix86_atom_sched_reorder): Likewise.
19600 * config/i386/x86-tune-sched-bd.c (ix86_bd_has_dispatch): Likewise.
19601 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Likewise.
19603 2021-04-21 Martin Liska <mliska@suse.cz>
19605 * config/i386/i386-options.c (TARGET_EXPLICIT_NO_SAHF_P):
19607 (SET_TARGET_NO_SAHF): Likewise.
19608 (TARGET_EXPLICIT_PREFETCH_SSE_P): Likewise.
19609 (SET_TARGET_PREFETCH_SSE): Likewise.
19610 (TARGET_EXPLICIT_NO_TUNE_P): Likewise.
19611 (SET_TARGET_NO_TUNE): Likewise.
19612 (TARGET_EXPLICIT_NO_80387_P): Likewise.
19613 (SET_TARGET_NO_80387): Likewise.
19615 * config/i386/i386.h (TARGET_*): Remove.
19616 * opth-gen.awk: Generate new used macros.
19618 2021-04-21 Martin Liska <mliska@suse.cz>
19620 * config/i386/i386.h (PTA_*): Remove.
19621 (enum pta_flag): New.
19622 (DEF_PTA): Generate PTA_* values from i386-isa.def.
19623 * config/i386/i386-isa.def: New file.
19625 2021-04-21 Alex Coplan <alex.coplan@arm.com>
19628 * config/aarch64/aarch64-bti-insert.c (aarch64_bti_j_insn_p): New.
19629 (rest_of_insert_bti): Avoid inserting duplicate bti j insns for
19630 jump table targets.
19632 2021-04-21 H.J. Lu <hjl.tools@gmail.com>
19634 * config.gcc: Install mwaitintrin.h for i[34567]86-*-* and
19635 x86_64-*-* targets.
19636 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_MWAIT_SET):
19638 (OPTION_MASK_ISA2_MWAIT_UNSET): Likewise.
19639 (ix86_handle_option): Handle -mmwait.
19640 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
19641 Replace OPTION_MASK_ISA_SSE3 with OPTION_MASK_ISA2_MWAIT on
19642 __builtin_ia32_monitor and __builtin_ia32_mwait.
19643 * config/i386/i386-options.c (isa2_opts): Add -mmwait.
19644 (ix86_valid_target_attribute_inner_p): Likewise.
19645 (ix86_option_override_internal): Enable mwait/monitor
19646 instructions for -msse3.
19647 * config/i386/i386.h (TARGET_MWAIT): New.
19648 (TARGET_MWAIT_P): Likewise.
19649 * config/i386/i386.opt: Add -mmwait.
19650 * config/i386/mwaitintrin.h: New file.
19651 * config/i386/pmmintrin.h: Include <mwaitintrin.h>.
19652 * config/i386/sse.md (sse3_mwait): Replace TARGET_SSE3 with
19654 (@sse3_monitor_<mode>): Likewise.
19655 * config/i386/x86gprintrin.h: Include <mwaitintrin.h>.
19656 * doc/extend.texi: Document mwait target attribute.
19657 * doc/invoke.texi: Document -mmwait.
19659 2021-04-21 Martin Liska <mliska@suse.cz>
19661 * config/i386/i386-options.c (DEF_ENUM): Remove it.
19662 * config/i386/i386-opts.h (DEF_ENUM): Likewise.
19663 * config/i386/stringop.def (DEF_ENUM): Likewise.
19665 2021-04-21 Martin Liska <mliska@suse.cz>
19667 * tree-cfg.c (gimple_verify_flow_info): Use qD instead
19668 of print_generic_expr.
19670 2021-04-21 Jakub Jelinek <jakub@redhat.com>
19672 PR rtl-optimization/100148
19673 * cprop.c (constprop_register): Use next_nondebug_insn instead of
19676 2021-04-21 Martin Liska <mliska@suse.cz>
19679 * cgraphunit.c (cgraph_node::analyze): Remove duplicate
19680 free_dominance_info calls.
19682 2021-04-21 Richard Biener <rguenther@suse.de>
19684 * gimple-fold.c (maybe_fold_reference): Remove is_lhs
19685 parameter (and assume it to be false).
19686 (fold_gimple_assign): Adjust, remove all callers of
19687 maybe_fold_reference calling it with is_lhs true.
19688 (gimple_fold_call): Likewise.
19689 (fold_stmt_1): Likewise.
19691 2021-04-21 Richard Biener <rguenther@suse.de>
19693 * fold-const.c (pedantic_non_lvalue_loc): Remove.
19694 (fold_binary_loc): Adjust.
19695 (fold_ternary_loc): Likewise.
19697 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
19699 PR middle-end/100130
19700 * varasm.c (get_block_for_decl): Make sure that any use of the
19701 retain attribute matches the section's retain flag.
19702 (switch_to_section): Check for retain mismatches even when
19703 changing sections, but do not warn if the given decl is the
19704 section's named.decl.
19705 (output_object_block): Pass the first decl in the block (if any)
19706 to switch_to_section.
19708 2021-04-20 H.J. Lu <hjl.tools@gmail.com>
19710 * config/i386/i386-c.c (ix86_target_macros_internal): Define
19711 __CRC32__ for -mcrc32.
19712 * config/i386/i386-options.c (ix86_option_override_internal):
19713 Enable crc32 instruction for -msse4.2.
19714 * config/i386/i386.md (sse4_2_crc32<mode>): Remove TARGET_SSE4_2
19716 (sse4_2_crc32di): Likewise.
19717 * config/i386/ia32intrin.h: Use crc32 target option for CRC32
19720 2021-04-20 Segher Boessenkool <segher@kernel.crashing.org>
19723 * config/rs6000/rs6000.c (rs6000_machine_from_flags): Do not consider
19726 2021-04-20 Martin Liska <mliska@suse.cz>
19728 * doc/invoke.texi: Fix typo.
19729 * params.opt: Likewise.
19731 2021-04-20 Martin Liska <mliska@suse.cz>
19733 * doc/invoke.texi: Document new param.
19735 2021-04-19 Andrew MacLeod <amacleod@redhat.com>
19737 PR tree-optimization/100081
19738 * gimple-range-cache.h (ranger_cache): Inherit from gori_compute
19739 rather than gori_compute_cache.
19740 * gimple-range-gori.cc (is_gimple_logical_p): Move to top of file.
19741 (range_def_chain::m_logical_depth): New member.
19742 (range_def_chain::range_def_chain): Initialize m_logical_depth.
19743 (range_def_chain::get_def_chain): Don't build defchains through more
19744 than LOGICAL_LIMIT logical expressions.
19745 * params.opt (param_ranger_logical_depth): New.
19747 2021-04-19 Richard Earnshaw <rearnsha@arm.com>
19750 * config/arm/arm.c (arm_configure_build_target): Do not strip
19751 extended FPU/SIMD feature bits from the target ISA when -mfpu
19752 is specified (partial revert of r11-8168).
19754 2021-04-19 Thomas Schwinge <thomas@codesourcery.com>
19756 * params.opt (-param=openacc-kernels=): Add.
19757 * omp-oacc-kernels-decompose.cc
19758 (pass_omp_oacc_kernels_decompose::gate): Use it.
19759 * doc/invoke.texi (-fopenacc-kernels=@var{mode}): Move...
19760 (--param): ... here, 'openacc-kernels'.
19762 2021-04-19 Martin Liska <mliska@suse.cz>
19765 * gengtype.c (finish_root_table): Align function arguments
19766 in between declaration and definition.
19768 2021-04-19 Eric Botcazou <ebotcazou@adacore.com>
19770 * config/i386/winnt.c (i386_pe_seh_cold_init): Properly deal with
19771 frames larger than the SEH maximum frame size.
19773 2021-04-18 Segher Boessenkool <segher@kernel.crashing.org>
19775 PR rtl-optimization/99927
19776 * combine.c (distribute_notes) [REG_UNUSED]: If the register already
19777 is dead, just drop it.
19779 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
19782 * config/i386/winnt-d.c (TARGET_D_TEMPLATES_ALWAYS_COMDAT): Define.
19783 * doc/tm.texi: Regenerate.
19784 * doc/tm.texi.in (D language and ABI): Add @hook for
19785 TARGET_D_TEMPLATES_ALWAYS_COMDAT.
19787 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
19789 * config/darwin-d.c (darwin_d_handle_target_object_format): New
19791 (darwin_d_register_target_info): New function.
19792 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19793 * config/dragonfly-d.c (dragonfly_d_handle_target_object_format): New
19795 (dragonfly_d_register_target_info): New function.
19796 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19797 * config/freebsd-d.c (freebsd_d_handle_target_object_format): New
19799 (freebsd_d_register_target_info): New function.
19800 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19801 * config/glibc-d.c (glibc_d_handle_target_object_format): New
19803 (glibc_d_register_target_info): New function.
19804 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19805 * config/i386/i386-d.c (ix86_d_handle_target_object_format): New
19807 (ix86_d_register_target_info): Add ix86_d_handle_target_object_format
19808 as handler for objectFormat key.
19809 * config/i386/winnt-d.c (winnt_d_handle_target_object_format): New
19811 (winnt_d_register_target_info): New function.
19812 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19813 * config/netbsd-d.c (netbsd_d_handle_target_object_format): New
19815 (netbsd_d_register_target_info): New function.
19816 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19817 * config/openbsd-d.c (openbsd_d_handle_target_object_format): New
19819 (openbsd_d_register_target_info): New function.
19820 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19821 * config/pa/pa-d.c (pa_d_handle_target_object_format): New function.
19822 (pa_d_register_target_info): Add pa_d_handle_target_object_format as
19823 handler for objectFormat key.
19824 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_object_format): New
19826 (rs6000_d_register_target_info): Add
19827 rs6000_d_handle_target_object_format as handler for objectFormat key.
19828 * config/sol2-d.c (solaris_d_handle_target_object_format): New
19830 (solaris_d_register_target_info): New function.
19831 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19833 2021-04-16 Jakub Jelinek <jakub@redhat.com>
19836 * config/aarch64/aarch64.c (aarch64_function_arg_alignment): Change
19837 abi_break argument from bool * to unsigned *, store there the pre-GCC 9
19839 (aarch64_layout_arg, aarch64_gimplify_va_arg_expr): Adjust callers.
19840 (aarch64_function_arg_regno_p): Likewise. Only emit -Wpsabi note if
19841 the old and new alignment after applying MIN/MAX to it is different.
19843 2021-04-16 Tamar Christina <tamar.christina@arm.com>
19846 * config/aarch64/aarch64-sve.md (@aarch64_sve_trn1_conv<mode>): New.
19847 * config/aarch64/aarch64.c (aarch64_expand_sve_const_pred_trn): Use new
19849 * config/aarch64/iterators.md (UNSPEC_TRN1_CONV): New.
19851 2021-04-16 Bill Schmidt <wschmidt@linux.ibm.com>
19853 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Revise
19854 this section and its subsections.
19856 2021-04-16 Jakub Jelinek <jakub@redhat.com>
19859 * config/aarch64/aarch64.md (*neg_asr_si2_extr, *extrsi5_insn_di): New
19860 define_insn patterns.
19862 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
19864 PR rtl-optimization/98689
19865 * reg-notes.def (UNTYPED_CALL): New note.
19866 * combine.c (distribute_notes): Handle it.
19867 * emit-rtl.c (try_split): Likewise.
19868 * rtlanal.c (rtx_properties::try_to_add_insn): Likewise. Assume
19869 that calls with the note implicitly set all return value registers.
19870 * builtins.c (expand_builtin_apply): Add a REG_UNTYPED_CALL
19873 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
19875 PR rtl-optimization/99596
19876 * rtlanal.c (rtx_properties::try_to_add_insn): Don't add global
19877 register accesses for const calls. Assume that pure functions
19878 can only read from global registers. Ignore cases in which
19879 the stack pointer has been marked global.
19881 2021-04-16 Jakub Jelinek <jakub@redhat.com>
19884 * tree-vect-loop.c (vect_transform_loop): Don't remove just
19885 dead scalar .MASK_LOAD calls, but also dead .COND_* calls - replace
19886 them by their last argument.
19888 2021-04-15 Martin Liska <mliska@suse.cz>
19890 * doc/invoke.texi: Other params don't use it, remove it.
19892 2021-04-15 Richard Biener <rguenther@suse.de>
19894 * gimple-builder.h: Add deprecation note.
19896 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
19899 * attribs.h (restrict_type_identity_attributes_to): Declare.
19900 * attribs.c (restrict_type_identity_attributes_to): New function.
19902 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
19905 * attribs.h (affects_type_identity_attributes): Declare.
19906 * attribs.c (remove_attributes_matching): New function.
19907 (affects_type_identity_attributes): Likewise.
19909 2021-04-15 Jakub Jelinek <jakub@redhat.com>
19912 * config/aarch64/aarch64.md (*<LOGICAL:optab>_<SHIFT:optab><mode>3):
19913 Add combine splitters for *<LOGICAL:optab>_ashl<mode>3 with
19914 ZERO_EXTEND, SIGN_EXTEND or AND.
19916 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
19918 PR rtl-optimization/99929
19919 * rtl.h (same_vector_encodings_p): New function.
19920 * cse.c (exp_equiv_p): Check that CONST_VECTORs have the same encoding.
19921 * cselib.c (rtx_equal_for_cselib_1): Likewise.
19922 * jump.c (rtx_renumbered_equal_p): Likewise.
19923 * lra-constraints.c (operands_match_p): Likewise.
19924 * reload.c (operands_match_p): Likewise.
19925 * rtl.c (rtx_equal_p_cb, rtx_equal_p): Likewise.
19927 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
19929 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Print
19930 more information about variable-length CONST_VECTORs.
19932 2021-04-14 Vladimir N. Makarov <vmakarov@redhat.com>
19934 PR rtl-optimization/100066
19935 * lra-constraints.c (split_reg): Check paradoxical_subreg_p for
19936 ordered modes when choosing splitting mode for hard reg.
19938 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
19941 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector_sel):
19943 (aarch64_expand_sve_const_vector): Use it for nelts_per_pattern==2.
19945 2021-04-14 Andreas Krebbel <krebbel@linux.ibm.com>
19947 * config/s390/s390-builtins.def (O_M5, O_M12, ...): Add new macros
19948 for mask operand types.
19949 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
19950 (s390_vec_permi_dbl, s390_vpdi): Use the M5 type for the immediate
19952 (s390_vec_msum_u128, s390_vmslg): Use the M12 type for the
19954 * config/s390/s390.c (s390_const_operand_ok): Check the new
19955 operand types and generate a list of valid values.
19957 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
19959 * doc/tm.texi: Regenerate.
19960 * doc/tm.texi.in (D language and ABI): Add @hook for
19961 TARGET_D_REGISTER_OS_TARGET_INFO.
19963 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
19965 * config/aarch64/aarch64-d.c (aarch64_d_handle_target_float_abi): New
19967 (aarch64_d_register_target_info): New function.
19968 * config/aarch64/aarch64-protos.h (aarch64_d_register_target_info):
19970 * config/aarch64/aarch64.h (TARGET_D_REGISTER_CPU_TARGET_INFO):
19972 * config/arm/arm-d.c (arm_d_handle_target_float_abi): New function.
19973 (arm_d_register_target_info): New function.
19974 * config/arm/arm-protos.h (arm_d_register_target_info): Declare.
19975 * config/arm/arm.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19976 * config/i386/i386-d.c (ix86_d_handle_target_float_abi): New function.
19977 (ix86_d_register_target_info): New function.
19978 * config/i386/i386-protos.h (ix86_d_register_target_info): Declare.
19979 * config/i386/i386.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19980 * config/mips/mips-d.c (mips_d_handle_target_float_abi): New function.
19981 (mips_d_register_target_info): New function.
19982 * config/mips/mips-protos.h (mips_d_register_target_info): Declare.
19983 * config/mips/mips.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19984 * config/pa/pa-d.c (pa_d_handle_target_float_abi): New function.
19985 (pa_d_register_target_info): New function.
19986 * config/pa/pa-protos.h (pa_d_register_target_info): Declare.
19987 * config/pa/pa.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19988 * config/riscv/riscv-d.c (riscv_d_handle_target_float_abi): New
19990 (riscv_d_register_target_info): New function.
19991 * config/riscv/riscv-protos.h (riscv_d_register_target_info): Declare.
19992 * config/riscv/riscv.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19993 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_float_abi): New
19995 (rs6000_d_register_target_info): New function.
19996 * config/rs6000/rs6000-protos.h (rs6000_d_register_target_info):
19998 * config/rs6000/rs6000.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19999 * config/s390/s390-d.c (s390_d_handle_target_float_abi): New function.
20000 (s390_d_register_target_info): New function.
20001 * config/s390/s390-protos.h (s390_d_register_target_info): Declare.
20002 * config/s390/s390.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
20003 * config/sparc/sparc-d.c (sparc_d_handle_target_float_abi): New
20005 (sparc_d_register_target_info): New function.
20006 * config/sparc/sparc-protos.h (sparc_d_register_target_info): Declare.
20007 * config/sparc/sparc.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
20008 * doc/tm.texi: Regenerate.
20009 * doc/tm.texi.in (D language and ABI): Add @hook for
20010 TARGET_D_REGISTER_CPU_TARGET_INFO.
20012 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
20014 * config/i386/i386-d.c (ix86_d_has_stdcall_convention): New function.
20015 * config/i386/i386-protos.h (ix86_d_has_stdcall_convention): Declare.
20016 * config/i386/i386.h (TARGET_D_HAS_STDCALL_CONVENTION): Define.
20017 * doc/tm.texi: Regenerate.
20018 * doc/tm.texi.in (D language and ABI): Add @hook for
20019 TARGET_D_HAS_STDCALL_CONVENTION.
20021 2021-04-14 Richard Biener <rguenther@suse.de>
20023 * tree-cfg.c (verify_gimple_assign_ternary): Verify that
20024 VEC_COND_EXPRs have a gimple_val condition.
20025 * tree-ssa-propagate.c (valid_gimple_rhs_p): VEC_COND_EXPR
20026 can no longer have a GENERIC condition.
20028 2021-04-14 Richard Earnshaw <rearnsha@arm.com>
20031 * config/arm/arm.c (arm_configure_build_target): Strip isa_all_fpbits
20032 from the isa_delta when -mfpu has been used.
20033 (arm_options_perform_arch_sanity_checks): It's the architecture that
20034 lacks an FPU not the processor.
20036 2021-04-13 Richard Biener <rguenther@suse.de>
20038 PR tree-optimization/100053
20039 * tree-ssa-sccvn.c (vn_nary_op_get_predicated_value): Do
20040 not use optimistic dominance queries for backedges to validate
20042 (dominated_by_p_w_unex): Add parameter to ignore executable
20043 state on backedges.
20044 (rpo_elim::eliminate_avail): Adjust.
20046 2021-04-13 Jakub Jelinek <jakub@redhat.com>
20049 * config/aarch64/aarch64.md (*aarch64_bfxil<mode>_extr,
20050 *aarch64_bfxilsi_extrdi): New define_insn patterns.
20052 2021-04-13 Jakub Jelinek <jakub@redhat.com>
20055 * simplify-rtx.c (simplify_immed_subreg): For MODE_COMPOSITE_P
20056 outermode, return NULL if the result doesn't encode back to the
20057 original byte sequence.
20058 (simplify_gen_subreg): Don't create SUBREGs from constants to
20059 MODE_COMPOSITE_P outermode.
20061 2021-04-12 Jakub Jelinek <jakub@redhat.com>
20063 PR rtl-optimization/99905
20064 * combine.c (expand_compound_operation): If pos + len > modewidth,
20065 perform the right shift by pos in inner_mode and then convert to mode,
20066 instead of trying to simplify a shift of rtx with inner_mode by pos
20067 as if it was a shift in mode.
20069 2021-04-12 Jakub Jelinek <jakub@redhat.com>
20072 * combine.c (simplify_and_const_int_1): Don't optimize varop
20073 away if it has side-effects.
20075 2021-04-12 Martin Liska <mliska@suse.cz>
20077 * doc/extend.texi: Escape @smallexample content.
20079 2021-04-12 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
20081 * config/s390/s390.md ("*movdi_31", "*movdi_64"): Add
20082 alternative in order to load a DFP zero.
20084 2021-04-12 Martin Liska <mliska@suse.cz>
20086 * doc/extend.texi: Be more precise in documentation
20087 of symver attribute.
20089 2021-04-12 Martin Liska <mliska@suse.cz>
20092 * gimplify.c (gimplify_expr): Right now, we unpoison all
20093 variables before a goto <dest>. We should not do it if we are
20096 2021-04-12 Cui,Lili <lili.cui@intel.com>
20098 * common/config/i386/cpuinfo.h (get_intel_cpu): Handle
20100 * common/config/i386/i386-common.c (processor_names): Add
20102 (processor_alias_table): Add rocketlake.
20103 * common/config/i386/i386-cpuinfo.h (processor_subtypes): Add
20104 INTEL_COREI7_ROCKETLAKE.
20105 * config.gcc: Add -march=rocketlake.
20106 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20108 * config/i386/i386-options.c (m_ROCKETLAKE) : Define.
20109 (processor_cost_table): Add rocketlake cost.
20110 * config/i386/i386.h (ix86_size_cost) : Define
20112 (processor_type) : Add PROCESSOR_ROCKETLAKE.
20113 (PTA_ROCKETLAKE): Ditto.
20114 * doc/extend.texi: Add rocketlake.
20115 * doc/invoke.texi: Add rocketlake.
20117 2021-04-12 Cui,Lili <lili.cui@intel.com>
20119 * config/i386/i386.h (PTA_ALDERLAKE): Change alderlake ISA list.
20120 * config/i386/i386-options.c (m_CORE_AVX2): Add m_ALDERLAKE.
20121 * common/config/i386/cpuinfo.h (get_intel_cpu): Add AlderLake model.
20122 * doc/invoke.texi: Change alderlake ISA list.
20124 2021-04-11 Hafiz Abid Qadeer <abidh@codesourcery.com>
20126 PR middle-end/98088
20127 * omp-expand.c (expand_oacc_collapse_init): Update condition in
20130 2021-04-10 H.J. Lu <hjl.tools@gmail.com>
20133 * config/i386/serializeintrin.h (_serialize): Defined as macro.
20135 2021-04-10 Jakub Jelinek <jakub@redhat.com>
20138 * expr.c (expand_expr_addr_expr_1): Test is_global_var rather than
20139 just TREE_STATIC on COMPOUND_LITERAL_EXPR_DECLs.
20141 2021-04-10 Jakub Jelinek <jakub@redhat.com>
20143 PR middle-end/99989
20144 * gimple-ssa-warn-alloca.c
20145 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit to
20146 0 with integer precision unconditionally.
20148 2021-04-10 Jakub Jelinek <jakub@redhat.com>
20150 PR rtl-optimization/98601
20151 * rtlanal.c (rtx_addr_can_trap_p_1): Allow in assert unknown size
20152 not just for BLKmode, but also for VOIDmode. For STRICT_ALIGNMENT
20153 unaligned_mems handle VOIDmode like BLKmode.
20155 2021-04-10 Jan Hubicka <hubicka@ucw.cz>
20158 * tree.c (free_lang_data_in_decl): Do not release body of
20159 declare_variant_alt.
20161 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
20163 * config/aarch64/aarch64.c (aarch64_option_restore): If the
20164 architecture was specified explicitly and the tuning wasn't,
20165 tune for the architecture rather than the configured default CPU.
20167 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
20169 * config/aarch64/aarch64.md (tlsdesc_small_sve_<mode>): Use X30
20170 as the temporary register.
20172 2021-04-09 Martin Liska <mliska@suse.cz>
20174 * doc/extend.texi: Move non-target attributes on the top level.
20176 2021-04-09 Martin Liska <mliska@suse.cz>
20178 * doc/invoke.texi: Document minimum and maximum value of the
20179 argument for both supported compression algorithms.
20181 2021-04-08 David Edelsohn <dje.gcc@gmail.com>
20183 * config/rs6000/rs6000.c (rs6000_xcoff_select_section): Select
20184 TLS BSS before TLS data.
20185 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .comm.
20187 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
20189 * doc/sourcebuild.texi (stdint_types_mbig_endian): Document.
20191 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
20193 * match.pd: Extend vec_cond folds to handle shifts.
20195 2021-04-08 Maciej W. Rozycki <macro@orcam.me.uk>
20197 * config/vax/vax.md: Fix comment for `*bit<mode>' pattern's
20200 2021-04-08 Alex Coplan <alex.coplan@arm.com>
20203 * config/arm/iterators.md (MVE_vecs): New.
20204 (V_elem): Also handle V2DF.
20205 * config/arm/mve.md (*mve_mov<mode>): Rename to ...
20206 (*mve_vdup<mode>): ... this. Remove second alternative since
20207 vec_duplicate of const_int is not canonical RTL, and we don't
20208 want to match symbol_refs.
20209 (*mve_vec_duplicate<mode>): Delete (pattern is redundant).
20211 2021-04-08 Xionghu Luo <luoxhu@linux.ibm.com>
20213 * fold-const.c (fold_single_bit_test): Fix typo.
20214 * print-rtl.c (print_rtx_insn_vec): Call print_rtl_single
20217 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
20219 PR tree-optimization/97513
20220 * tree-vect-slp.c (vect_add_slp_permutation): New function,
20222 (vectorizable_slp_permutation): ...here. Detect cases in which
20223 all VEC_PERM_EXPRs are guaranteed to have the same stepped
20224 permute vector and only generate one permute vector for that case.
20225 Extend that case to handle variable-length vectors.
20227 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
20229 PR tree-optimization/99873
20230 * tree-vect-slp.c (vect_slp_prefer_store_lanes_p): New function.
20231 (vect_build_slp_instance): Don't split store groups that could
20232 use IFN_STORE_LANES.
20234 2021-04-07 Jakub Jelinek <jakub@redhat.com>
20237 * varasm.c (output_constant_pool_contents): Don't strip name encoding
20238 from XSTR (desc->sym, 0) or from label before passing those to
20241 2021-04-07 Richard Biener <rguenther@suse.de>
20243 PR tree-optimization/99954
20244 * tree-loop-distribution.c: Include tree-affine.h.
20245 (generate_memcpy_builtin): Try using tree-affine to prove
20247 (loop_distribution::classify_builtin_ldst): Always classify
20250 2021-04-07 Richard Biener <rguenther@suse.de>
20252 PR tree-optimization/99947
20253 * tree-vect-loop.c (vectorizable_induction): Pre-allocate
20254 steps vector to avoid pushing elements from the reallocated
20257 2021-04-07 Richard Biener <rguenther@suse.de>
20259 * tree-ssa-sccvn.h (print_vn_reference_ops): Declare.
20260 * tree-ssa-pre.c (print_pre_expr): Factor out VN reference operand
20262 * tree-ssa-sccvn.c (print_vn_reference_ops): ... into this new
20264 (debug_vn_reference_ops): New.
20266 2021-04-07 Bin Cheng <bin.cheng@linux.alibaba.com>
20268 PR tree-optimization/98736
20269 * tree-loop-distribution.c
20270 * (loop_distribution::bb_top_order_init):
20271 Compute RPO with programing order preserved by calling function
20272 rev_post_order_and_mark_dfs_back_seme.
20274 2021-04-06 Vladimir N. Makarov <vmakarov@redhat.com>
20277 * lra-constraints.c (split_reg): Don't check paradoxical_subreg_p.
20278 * lra-lives.c (clear_sparseset_regnos, regnos_in_sparseset_p): New
20280 (process_bb_lives): Don't update biggest mode of hard reg for
20281 implicit in multi-register group. Use the new functions for
20282 updating dead_set and unused_set by register notes.
20284 2021-04-06 Xianmiao Qu <xianmiao_qu@c-sky.com>
20286 * config/csky/csky_pipeline_ck802.md : Use insn reservation name
20289 2021-04-06 H.J. Lu <hjl.tools@gmail.com>
20291 * config/i386/x86-tune-costs.h (skylake_memcpy): Updated.
20292 (skylake_memset): Likewise.
20293 (skylake_cost): Change CLEAR_RATIO to 17.
20294 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
20295 Replace m_CANNONLAKE, m_ICELAKE_CLIENT, m_ICELAKE_SERVER,
20296 m_TIGERLAKE and m_SAPPHIRERAPIDS with m_SKYLAKE and m_CORE_AVX512.
20298 2021-04-06 Richard Biener <rguenther@suse.de>
20300 PR tree-optimization/99880
20301 * tree-vect-loop.c (maybe_set_vectorized_backedge_value): Only
20302 set vectorized defs of relevant PHIs.
20304 2021-04-06 Richard Biener <rguenther@suse.de>
20306 PR tree-optimization/99924
20307 * tree-vect-slp.c (vect_bb_partition_graph_r): Do not mark
20308 nodes w/o scalar stmts as visited.
20310 2021-04-06 Alex Coplan <alex.coplan@arm.com>
20313 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Also use base
20314 PCS for [su]fix_optab.
20316 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
20318 * config/darwin.c (machopic_legitimize_pic_address): Check
20319 that the current pic register is one of the hard reg set
20320 before setting liveness.
20322 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
20324 * config/darwin.c (machopic_legitimize_pic_address): Fix
20325 whitespace, remove unused code.
20327 2021-04-03 Jakub Jelinek <jakub@redhat.com>
20329 PR tree-optimization/99882
20330 * gimple-ssa-store-merging.c (bswap_view_convert): Handle val with
20333 2021-04-03 Jakub Jelinek <jakub@redhat.com>
20335 PR rtl-optimization/99863
20336 * dse.c (replace_read): Drop regs_live argument. Instead of
20337 regs_live, use store_insn->fixed_regs_live if non-NULL,
20338 otherwise punt if insns sequence clobbers or sets any hard
20341 2021-04-03 Jakub Jelinek <jakub@redhat.com>
20344 * targhooks.h (default_print_patchable_function_entry_1): Declare.
20345 * targhooks.c (default_print_patchable_function_entry_1): New function,
20346 copied from default_print_patchable_function_entry with an added flags
20348 (default_print_patchable_function_entry): Rewritten into a small
20349 wrapper around default_print_patchable_function_entry_1.
20350 * config/rs6000/rs6000.c (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
20352 (rs6000_print_patchable_function_entry): New function.
20354 2021-04-02 Eric Botcazou <ebotcazou@adacore.com>
20356 * doc/invoke.texi (fdelete-dead-exceptions): Minor tweak.
20358 2021-04-01 Jason Merrill <jason@redhat.com>
20361 * common.opt: Document v15 and v16.
20363 2021-04-01 Richard Biener <rguenther@suse.de>
20365 PR tree-optimization/99863
20366 * gimplify.c (gimplify_init_constructor): Recompute vector
20369 2021-04-01 Jakub Jelinek <jakub@redhat.com>
20371 * doc/extend.texi (symver attribute): Fix up syntax errors
20374 2021-04-01 Jakub Jelinek <jakub@redhat.com>
20376 PR tree-optimization/96573
20377 * gimple-ssa-store-merging.c (init_symbolic_number): Handle
20378 also pointer types.
20380 2021-04-01 Richard Biener <rguenther@suse.de>
20382 PR tree-optimization/99856
20383 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Promote
20384 precision to vector element precision.
20386 2021-04-01 Martin Jambor <mjambor@suse.cz>
20388 PR tree-optimization/97009
20389 * tree-sra.c (access_or_its_child_written): New function.
20390 (propagate_subaccesses_from_rhs): Use it instead of a simple grp_write
20393 2021-03-31 Jan Hubicka <hubicka@ucw.cz>
20396 * cif-code.def (USES_COMDAT_LOCAL): Make CIF_FINAL_NORMAL.
20398 2021-03-31 Pat Haugen <pthaugen@linux.ibm.com>
20401 * config/rs6000/altivec.md (xxspltiw_v4si, xxspltiw_v4sf_inst,
20402 xxspltidp_v2df_inst, xxsplti32dx_v4si_inst, xxsplti32dx_v4sf_inst,
20403 xxblend_<mode>, xxpermx_inst, xxeval): Mark prefixed.
20404 * config/rs6000/mma.md (mma_<vvi4i4i8>, mma_<avvi4i4i8>,
20405 mma_<vvi4i4i2>, mma_<avvi4i4i2>, mma_<vvi4i4>, mma_<avvi4i4>,
20406 mma_<pvi4i2>, mma_<apvi4i2>, mma_<vvi4i4i4>, mma_<avvi4i4i4>):
20408 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Adjust test.
20409 * config/rs6000/rs6000.md (define_attr "maybe_prefixed"): New.
20410 (define_attr "prefixed"): Update initializer.
20412 2021-03-31 Jakub Jelinek <jakub@redhat.com>
20415 * dwarf2out.c (debug_ranges_dwo_section): New variable.
20416 (DW_RANGES_IDX_SKELETON): Define.
20417 (struct dw_ranges): Add begin_entry and end_entry members.
20418 (DEBUG_DWO_RNGLISTS_SECTION): Define.
20419 (add_ranges_num): Adjust r initializer for addition of *_entry
20421 (add_ranges_by_labels): For -gsplit-dwarf and force_direct,
20422 set idx to DW_RANGES_IDX_SKELETON.
20423 (use_distinct_base_address_for_range): New function.
20424 (index_rnglists): Don't set r->idx if it is equal to
20425 DW_RANGES_IDX_SKELETON. Initialize r->begin_entry and
20426 r->end_entry for -gsplit-dwarf if those will be needed by
20428 (output_rnglists): Add DWO argument. If true, switch to
20429 debug_ranges_dwo_section rather than debug_ranges_section.
20430 Adjust l1/l2 label indexes. Only output the offset table when
20431 dwo is true and don't include in there the skeleton range
20432 entry if present. For -gsplit-dwarf, skip ranges that belong
20433 to the other rnglists section. Change return type from void
20434 to bool and return true if there are any range entries for
20435 the other section. For dwarf_split_debug_info use
20436 DW_RLE_startx_endx, DW_RLE_startx_length and DW_RLE_base_addressx
20437 entries instead of DW_RLE_start_end, DW_RLE_start_length and
20438 DW_RLE_base_address. Use use_distinct_base_address_for_range.
20439 (init_sections_and_labels): Initialize debug_ranges_dwo_section
20440 if -gsplit-dwarf and DWARF >= 5. Adjust ranges_section_label
20441 and range_base_label indexes.
20442 (dwarf2out_finish): Call index_rnglists earlier before finalizing
20443 .debug_addr. Never emit DW_AT_rnglists_base attribute. For
20444 -gsplit-dwarf and DWARF >= 5 call output_rnglists up to twice
20445 with different dwo arguments.
20446 (dwarf2out_c_finalize): Clear debug_ranges_dwo_section.
20448 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
20450 PR tree-optimization/98268
20451 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Call
20452 recompute_tree_invariant_for_addr_expr after successfully
20453 folding a TARGET_MEM_REF that occurs inside an ADDR_EXPR.
20455 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
20457 PR tree-optimization/99726
20458 * tree-data-ref.c (create_intersect_range_checks_index): Bail
20459 out if there is more than one access function SCEV for the loop
20462 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
20464 PR rtl-optimization/97141
20465 PR rtl-optimization/98726
20466 * emit-rtl.c (valid_for_const_vector_p): Return true for
20468 * rtx-vector-builder.h (rtx_vector_builder::step): Return a
20469 poly_wide_int instead of a wide_int.
20470 (rtx_vector_builder::apply_set): Take a poly_wide_int instead
20472 * rtx-vector-builder.c (rtx_vector_builder::apply_set): Likewise.
20473 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Return
20474 false for CONST_VECTORs that cannot be forced to memory.
20475 * config/aarch64/aarch64-simd.md (mov<mode>): If a CONST_VECTOR
20476 is too complex to force to memory, build it up from individual
20479 2021-03-31 Jan Hubicka <jh@suse.cz>
20482 * cgraph.c (cgraph_node::release_body): Fix overactive check.
20484 2021-03-31 Christophe Lyon <christophe.lyon@linaro.org>
20487 * config/arm/vec-common.md (mul<mode>3): Disable on iwMMXT, expect
20490 2021-03-31 H.J. Lu <hjl.tools@gmail.com>
20492 * config/i386/i386-expand.c (expand_set_or_cpymem_via_rep):
20493 For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, don't convert QImode
20495 (decide_alg): For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, use
20496 "rep movsb/stosb" only for known sizes.
20497 * config/i386/i386-options.c (processor_cost_table): Use Ice
20498 Lake cost for Cannon Lake, Ice Lake, Tiger Lake, Sapphire
20499 Rapids and Alder Lake.
20500 * config/i386/i386.h (TARGET_PREFER_KNOWN_REP_MOVSB_STOSB): New.
20501 * config/i386/x86-tune-costs.h (icelake_memcpy): New.
20502 (icelake_memset): Likewise.
20503 (icelake_cost): Likewise.
20504 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
20507 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
20510 * config/aarch64/aarch64.c
20511 (aarch64_vectorize_preferred_vector_alignment): Query the size
20512 of the provided SVE vector; do not assume that all SVE vectors
20513 have the same size.
20515 2021-03-31 Jan Hubicka <jh@suse.cz>
20518 * cgraph.c (cgraph_node::release_body): Remove all callers and
20520 * cgraphclones.c (cgraph_node::materialize_clone): Do not do it here.
20521 * cgraphunit.c (cgraph_node::expand): And here.
20523 2021-03-31 Martin Liska <mliska@suse.cz>
20525 * ipa-modref.c (analyze_ssa_name_flags): Fix coding style
20526 and one negated condition.
20528 2021-03-31 Jakub Jelinek <jakub@redhat.com>
20529 Richard Sandiford <richard.sandiford@arm.com>
20532 * config/aarch64/aarch64.md (*add<mode>3_poly_1): Swap Uai and Uav
20533 constraints on operands[2] and similarly 0 and rk constraints
20534 on operands[1] corresponding to that.
20536 2021-03-31 Jakub Jelinek <jakub@redhat.com>
20539 * configure.ac (HAVE_LD_BROKEN_PE_DWARF5): New AC_DEFINE if PECOFF
20540 linker doesn't support DWARF sections new in DWARF5.
20541 * config/i386/i386-options.c (ix86_option_override_internal): Default
20542 to dwarf_version 4 if HAVE_LD_BROKEN_PE_DWARF5 for TARGET_PECOFF
20544 * config.in: Regenerated.
20545 * configure: Regenerated.
20547 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20550 * config/aarch64/aarch64.c (aarch64_analyze_loop_vinfo): Check for
20551 available issue_info before using it.
20553 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20556 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): Do not allow zero
20559 2021-03-30 Xionghu Luo <luoxhu@linux.ibm.com>
20562 * config/rs6000/altivec.md (altivec_lvsl_reg): Change to ...
20563 (altivec_lvsl_reg_<mode>): ... this.
20564 (altivec_lvsr_reg): Change to ...
20565 (altivec_lvsr_reg_<mode>): ... this.
20566 * config/rs6000/predicates.md (vec_set_index_operand): New.
20567 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20568 Enable 32bit variable vec_insert for all TARGET_VSX.
20569 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
20570 Enable 32bit variable vec_insert for p9 and above.
20571 (rs6000_expand_vector_set_var_p8): Rename to ...
20572 (rs6000_expand_vector_set_var_p7): ... this.
20573 (rs6000_expand_vector_set): Use TARGET_VSX and adjust assert
20575 * config/rs6000/vector.md (vec_set<mode>): Use vec_set_index_operand.
20576 * config/rs6000/vsx.md (xl_len_r): Use gen_altivec_lvsl_reg_di and
20577 gen_altivec_lvsr_reg_di.
20579 2021-03-30 H.J. Lu <hjl.tools@gmail.com>
20582 * config/i386/ia32intrin.h (__rdtsc): Defined as macro.
20583 (__rdtscp): Likewise.
20585 2021-03-30 Tamar Christina <tamar.christina@arm.com>
20587 PR tree-optimization/99825
20588 * tree-vect-slp-patterns.c (vect_check_evenodd_blend):
20589 Reject non-mult 2 lanes.
20591 2021-03-30 Richard Earnshaw <rearnsha@arm.com>
20594 * config/arm/arm.c (arm_file_start): Fix emission of
20595 Tag_ABI_VFP_args attribute.
20597 2021-03-30 Richard Biener <rguenther@suse.de>
20599 PR tree-optimization/99824
20600 * stor-layout.c (set_min_and_max_values_for_integral_type):
20601 Assert the precision is within the bounds of
20602 WIDE_INT_MAX_PRECISION.
20603 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
20604 the outermost component ref only to lower the access size
20605 and initialize that from the access type.
20607 2021-03-30 Richard Sandiford <richard.sandiford@arm.com>
20610 * config/aarch64/aarch64.md (mov<mode>): Pass multi-instruction
20611 CONST_INTs to aarch64_expand_mov_immediate when called after RA.
20613 2021-03-30 Mihailo Stojanovic <mihailo.stojanovic@typhoon-hil.com>
20615 * config/aarch64/aarch64.md
20616 (<optab>_trunc<fcvt_target><GPI:mode>2): Set the "arch"
20617 attribute to disambiguate between SIMD and FP variants of the
20620 2021-03-29 Jan Hubicka <hubicka@ucw.cz>
20622 * ipa-modref.c (merge_call_lhs_flags): Correct handling of deref.
20623 (analyze_ssa_name_flags): Fix typo in comment.
20625 2021-03-29 Alex Coplan <alex.coplan@arm.com>
20628 * config/aarch64/aarch64-sve-builtins.cc
20629 (function_builder::add_function): Add placeholder_p argument, use
20630 placeholder decls if this is set.
20631 (function_builder::add_unique_function): Instead of conditionally adding
20632 direct overloads, unconditionally add either a direct overload or a
20634 (function_builder::add_overloaded_function): Set placeholder_p if we're
20635 using C++ overloads. Use the obstack for string storage instead
20636 of relying on the tree nodes.
20637 (function_builder::add_overloaded_functions): Don't return early for
20638 m_direct_overloads: we need to add placeholders.
20639 * config/aarch64/aarch64-sve-builtins.h
20640 (function_builder::add_function): Add placeholder_p argument.
20642 2021-03-29 Richard Biener <rguenther@suse.de>
20644 PR tree-optimization/99807
20645 * tree-vect-slp.c (vect_slp_analyze_node_operations_1): Move
20646 assert below VEC_PERM handling.
20648 2021-03-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20651 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>): Use
20652 aarch64_simd_or_scalar_imm_zero to match zeroes. Remove pattern
20653 matching const_int 0.
20654 (move_lo_quad_internal_be_<mode>): Likewise.
20655 (move_lo_quad_<mode>): Update for the above.
20656 * config/aarch64/iterators.md (VQ_2E): Delete.
20658 2021-03-29 Jakub Jelinek <jakub@redhat.com>
20660 PR tree-optimization/99777
20661 * fold-const.c (extract_muldiv_1): For conversions, punt on casts from
20662 types other than scalar integral types.
20664 2021-03-28 David Edelsohn <dje.gcc@gmail.com>
20666 * config/rs6000/rs6000.c (rs6000_output_dwarf_dtprel): Do not add
20667 XCOFF TLS reloc decorations.
20669 2021-03-28 Gerald Pfeifer <gerald@pfeifer.com>
20671 * doc/analyzer.texi (Analyzer Internals): Update link to
20672 "A Memory Model for Static Analysis of C Programs".
20674 2021-03-26 David Edelsohn <dje.gcc@gmail.com>
20676 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Call function.
20677 * config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align):
20679 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align): New.
20680 (rs6000_special_round_type_align): Recursively check innermost first
20683 2021-03-26 Jakub Jelinek <jakub@redhat.com>
20686 * dwarf2out.h (struct dw_fde_node): Add rule18 member.
20687 * dwarf2cfi.c (dwarf2out_frame_debug_expr): When handling (set hfp sp)
20688 assignment with drap_reg active, queue reg save for hfp with offset 0
20689 and flush queued reg saves. When handling a push with rule18,
20690 defer queueing reg save for hfp and just assert the offset is 0.
20691 (scan_trace): Assert that fde->rule18 is false.
20693 2021-03-26 Vladimir Makarov <vmakarov@redhat.com>
20696 * ira-costs.c (record_reg_classes): Put case with
20697 CT_RELAXED_MEMORY adjacent to one with CT_MEMORY.
20698 * ira.c (ira_setup_alts): Ditto.
20699 * lra-constraints.c (process_alt_operands): Ditto.
20700 * recog.c (asm_operand_ok): Ditto.
20701 * reload.c (find_reloads): Ditto.
20703 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20705 * config/aarch64/aarch64-protos.h
20706 (cpu_addrcost_table::post_modify_ld3_st3): New member variable.
20707 (cpu_addrcost_table::post_modify_ld4_st4): Likewise.
20708 * config/aarch64/aarch64.c (generic_addrcost_table): Update
20709 accordingly, using the same costs as for post_modify.
20710 (exynosm1_addrcost_table, xgene1_addrcost_table): Likewise.
20711 (thunderx2t99_addrcost_table, thunderx3t110_addrcost_table):
20712 (tsv110_addrcost_table, qdf24xx_addrcost_table): Likewise.
20713 (a64fx_addrcost_table): Likewise.
20714 (neoversev1_addrcost_table): New.
20715 (neoversev1_tunings): Use neoversev1_addrcost_table.
20716 (aarch64_address_cost): Use the new post_modify costs for CImode
20719 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20721 * config/aarch64/aarch64.opt
20722 (-param=aarch64-loop-vect-issue-rate-niters=): New parameter.
20723 * doc/invoke.texi: Document it.
20724 * config/aarch64/aarch64-protos.h (aarch64_base_vec_issue_info)
20725 (aarch64_scalar_vec_issue_info, aarch64_simd_vec_issue_info)
20726 (aarch64_advsimd_vec_issue_info, aarch64_sve_vec_issue_info)
20727 (aarch64_vec_issue_info): New structures.
20728 (cpu_vector_cost): Write comments above the variables rather
20730 (cpu_vector_cost::issue_info): New member variable.
20731 * config/aarch64/aarch64.c: Include gimple-pretty-print.h
20732 and tree-ssa-loop-niter.h.
20733 (generic_vector_cost, a64fx_vector_cost, qdf24xx_vector_cost)
20734 (thunderx_vector_cost, tsv110_vector_cost, cortexa57_vector_cost)
20735 (exynosm1_vector_cost, xgene1_vector_cost, thunderx2t99_vector_cost)
20736 (thunderx3t110_vector_cost): Initialize issue_info to null.
20737 (neoversev1_scalar_issue_info, neoversev1_advsimd_issue_info)
20738 (neoversev1_sve_issue_info, neoversev1_vec_issue_info): New structures.
20739 (neoversev1_vector_cost): Use them.
20740 (aarch64_vec_op_count, aarch64_sve_op_count): New structures.
20741 (aarch64_vector_costs::saw_sve_only_op): New member variable.
20742 (aarch64_vector_costs::num_vector_iterations): Likewise.
20743 (aarch64_vector_costs::scalar_ops): Likewise.
20744 (aarch64_vector_costs::advsimd_ops): Likewise.
20745 (aarch64_vector_costs::sve_ops): Likewise.
20746 (aarch64_vector_costs::seen_loads): Likewise.
20747 (aarch64_simd_vec_costs_for_flags): New function.
20748 (aarch64_analyze_loop_vinfo): Initialize num_vector_iterations.
20749 Count the number of predicate operations required by SVE WHILE
20751 (aarch64_comparison_type, aarch64_multiply_add_p): New functions.
20752 (aarch64_sve_only_stmt_p, aarch64_in_loop_reduction_latency): Likewise.
20753 (aarch64_count_ops): Likewise.
20754 (aarch64_add_stmt_cost): Record whether see an SVE operation
20755 that cannot currently be implementing using Advanced SIMD.
20756 Record issue information about the scalar, Advanced SIMD
20757 and (where relevant) SVE versions of a loop.
20758 (aarch64_vec_op_count::dump): New function.
20759 (aarch64_sve_op_count::dump): Likewise.
20760 (aarch64_estimate_min_cycles_per_iter): Likewise.
20761 (aarch64_adjust_body_cost): If issue information is available,
20762 try to compare the issue rates of the various loop implementations
20763 and increase or decrease the vector body cost accordingly.
20765 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20767 * config/aarch64/aarch64.c (aarch64_detect_vector_stmt_subtype):
20768 Assume a zero cost for induction phis.
20770 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20772 * config/aarch64/aarch64.c (aarch64_embedded_comparison_type): New
20774 (aarch64_adjust_stmt_cost): Add the costs of embedded scalar and
20775 vector comparisons.
20777 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20779 * config/aarch64/aarch64.c (aarch64_detect_scalar_stmt_subtype):
20781 (aarch64_add_stmt_cost): Call it.
20783 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20785 * config/aarch64/aarch64-tuning-flags.def (matched_vector_throughput):
20786 New tuning parameter.
20787 * config/aarch64/aarch64.c (neoversev1_tunings): Use it.
20788 (aarch64_estimated_sve_vq): New function.
20789 (aarch64_vector_costs::analyzed_vinfo): New member variable.
20790 (aarch64_vector_costs::is_loop): Likewise.
20791 (aarch64_vector_costs::unrolled_advsimd_niters): Likewise.
20792 (aarch64_vector_costs::unrolled_advsimd_stmts): Likewise.
20793 (aarch64_record_potential_advsimd_unrolling): New function.
20794 (aarch64_analyze_loop_vinfo, aarch64_analyze_bb_vinfo): Likewise.
20795 (aarch64_add_stmt_cost): Call aarch64_analyze_loop_vinfo or
20796 aarch64_analyze_bb_vinfo on the first use of a costs structure.
20797 Detect whether we're vectorizing a loop for SVE that might be
20798 completely unrolled if it used Advanced SIMD instead.
20799 (aarch64_adjust_body_cost_for_latency): New function.
20800 (aarch64_finish_cost): Call it.
20802 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20804 * config/aarch64/aarch64.c (aarch64_vector_costs): New structure.
20805 (aarch64_init_cost): New function.
20806 (aarch64_add_stmt_cost): Use aarch64_vector_costs instead of
20807 the default unsigned[3].
20808 (aarch64_finish_cost, aarch64_destroy_cost_data): New functions.
20809 (TARGET_VECTORIZE_INIT_COST): Override.
20810 (TARGET_VECTORIZE_FINISH_COST): Likewise.
20811 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
20813 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20815 * config/aarch64/aarch64.c (neoversev1_advsimd_vector_cost)
20816 (neoversev1_sve_vector_cost): New cost structures.
20817 (neoversev1_vector_cost): Likewise.
20818 (neoversev1_tunings): Use them. Enable use_new_vector_costs.
20820 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20822 * config/aarch64/aarch64-protos.h
20823 (sve_vec_cost::scatter_store_elt_cost): New member variable.
20824 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
20825 accordingly, taking the cost from the cost of a scalar_store.
20826 (a64fx_sve_vector_cost): Likewise.
20827 (aarch64_detect_vector_stmt_subtype): Detect scatter stores.
20829 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20831 * config/aarch64/aarch64-protos.h
20832 (simd_vec_cost::store_elt_extra_cost): New member variable.
20833 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
20834 accordingly, using the vec_to_scalar cost for the new field.
20835 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
20836 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
20837 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
20838 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
20839 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
20840 (thunderx3t110_advsimd_vector_cost): Likewise.
20841 (aarch64_detect_vector_stmt_subtype): Detect single-element stores.
20843 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20845 * config/aarch64/aarch64-protos.h (simd_vec_cost::ld2_st2_permute_cost)
20846 (simd_vec_cost::ld3_st3_permute_cost): New member variables.
20847 (simd_vec_cost::ld4_st4_permute_cost): Likewise.
20848 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
20849 accordingly, using zero for the new costs.
20850 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
20851 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
20852 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
20853 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
20854 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
20855 (thunderx3t110_advsimd_vector_cost): Likewise.
20856 (aarch64_ld234_st234_vectors): New function.
20857 (aarch64_adjust_stmt_cost): Likewise.
20858 (aarch64_add_stmt_cost): Call aarch64_adjust_stmt_cost if using
20859 the new vector costs.
20861 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20863 * config/aarch64/aarch64-protos.h (sve_vec_cost): Turn into a
20864 derived class of simd_vec_cost. Add information about CLAST[AB]
20865 and FADDA instructions.
20866 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
20867 accordingly, using the vec_to_scalar costs for the new fields.
20868 (a64fx_sve_vector_cost): Likewise.
20869 (aarch64_reduc_type): New function.
20870 (aarch64_sve_in_loop_reduction_latency): Likewise.
20871 (aarch64_detect_vector_stmt_subtype): Take a vinfo parameter.
20872 Use aarch64_sve_in_loop_reduction_latency to handle SVE reductions
20873 that occur in the loop body.
20874 (aarch64_add_stmt_cost): Update call accordingly.
20876 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20878 * config/aarch64/aarch64-tuning-flags.def (use_new_vector_costs):
20880 * config/aarch64/aarch64-protos.h (simd_vec_cost): Put comments
20881 above the fields rather than to the right.
20882 (simd_vec_cost::reduc_i8_cost): New member variable.
20883 (simd_vec_cost::reduc_i16_cost): Likewise.
20884 (simd_vec_cost::reduc_i32_cost): Likewise.
20885 (simd_vec_cost::reduc_i64_cost): Likewise.
20886 (simd_vec_cost::reduc_f16_cost): Likewise.
20887 (simd_vec_cost::reduc_f32_cost): Likewise.
20888 (simd_vec_cost::reduc_f64_cost): Likewise.
20889 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
20890 accordingly, using the vec_to_scalar_cost for the new fields.
20891 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
20892 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
20893 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
20894 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
20895 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
20896 (thunderx3t110_advsimd_vector_cost): Likewise.
20897 (aarch64_use_new_vector_costs_p): New function.
20898 (aarch64_simd_vec_costs): New function, split out from...
20899 (aarch64_builtin_vectorization_cost): ...here.
20900 (aarch64_is_reduction): New function.
20901 (aarch64_detect_vector_stmt_subtype): Likewise.
20902 (aarch64_add_stmt_cost): Call aarch64_detect_vector_stmt_subtype if
20903 using the new vector costs.
20905 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
20908 * tree-emutls.c (get_emutls_init_templ_addr): Mark initializer of weak
20909 TLS declarations as public.
20911 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
20913 * config/aarch64/aarch64-d.c (IN_TARGET_CODE): Define.
20914 * config/arm/arm-d.c (IN_TARGET_CODE): Likewise.
20915 * config/i386/i386-d.c (IN_TARGET_CODE): Likewise.
20916 * config/mips/mips-d.c (IN_TARGET_CODE): Likewise.
20917 * config/pa/pa-d.c (IN_TARGET_CODE): Likewise.
20918 * config/riscv/riscv-d.c (IN_TARGET_CODE): Likewise.
20919 * config/rs6000/rs6000-d.c (IN_TARGET_CODE): Likewise.
20920 * config/s390/s390-d.c (IN_TARGET_CODE): Likewise.
20921 * config/sparc/sparc-d.c (IN_TARGET_CODE): Likewise.
20923 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
20926 * config.gcc (*-*-cygwin*): Add winnt-d.o
20927 (*-*-mingw*): Likewise.
20928 * config/i386/cygwin.h (EXTRA_TARGET_D_OS_VERSIONS): New macro.
20929 * config/i386/mingw32.h (EXTRA_TARGET_D_OS_VERSIONS): Likewise.
20930 * config/i386/t-cygming: Add winnt-d.o.
20931 * config/i386/winnt-d.c: New file.
20933 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
20935 * config/freebsd-d.c: Include memmodel.h.
20937 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
20940 * config.gcc (*-*-openbsd*): Add openbsd-d.o.
20941 * config/t-openbsd: Add openbsd-d.o.
20942 * config/openbsd-d.c: New file.
20944 2021-03-25 Stam Markianos-Wright <stam.markianos-wright@arm.com>
20946 PR tree-optimization/96974
20947 * tree-vect-stmts.c (vect_get_vector_types_for_stmt): Replace assert
20948 with graceful exit.
20950 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
20953 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
20957 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
20958 always_inline in system headers.
20960 2021-03-25 Kewen Lin <linkw@linux.ibm.com>
20962 * tree-vect-loop.c (vect_model_reduction_cost): Init inside_cost.
20964 2021-03-25 Jakub Jelinek <jakub@redhat.com>
20967 * tree-core.h (enum operand_equal_flag): Add OEP_ADDRESS_OF_SAME_FIELD.
20968 * fold-const.c (operand_compare::operand_equal_p): Don't compare
20969 field offsets if OEP_ADDRESS_OF_SAME_FIELD.
20971 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
20975 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
20976 always_inline in system headers.
20978 2021-03-25 Richard Biener <rguenther@suse.de>
20980 PR tree-optimization/99746
20981 * tree-vect-slp-patterns.c (complex_pattern::build): Do not mark
20982 the scalar stmt as patterned. Instead set up required things
20985 2021-03-25 Xionghu Luo <luoxhu@linux.ibm.com>
20987 * config/rs6000/rs6000.c (power8_costs): Change l2 cache
20990 2021-03-24 Martin Liska <mliska@suse.cz>
20993 * common/config/i386/i386-common.c (ARRAY_SIZE): Fix off-by-one
20995 * config/i386/i386-options.c (ix86_option_override_internal):
20996 Add run-time assert.
20998 2021-03-24 Martin Jambor <mjambor@suse.cz>
21001 * ipa-cp.c (initialize_node_lattices): Mark as bottom all
21002 parameters with unknown type.
21003 (ipacp_value_safe_for_type): New function.
21004 (propagate_vals_across_arith_jfunc): Verify that the constant type
21005 can be used for a type of the formal parameter.
21006 (propagate_vals_across_ancestor): Likewise.
21007 (propagate_scalar_across_jump_function): Likewise. Pass the type
21008 also to propagate_vals_across_ancestor.
21010 2021-03-24 Christophe Lyon <christophe.lyon@linaro.org>
21013 * config/arm/mve.md (movmisalign<mode>_mve_store): Use Ux
21015 (movmisalign<mode>_mve_load): Likewise.
21017 2021-03-24 Jakub Jelinek <jakub@redhat.com>
21020 * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
21021 movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.
21023 2021-03-24 Alexandre Oliva <oliva@adacore.com>
21025 * doc/sourcebuild.texi (sysconf): New effective target.
21027 2021-03-24 Alexandre Oliva <oliva@adacore.com>
21029 * config/i386/predicates.md (reg_or_const_vec_operand): New.
21030 * config/i386/sse.md (ssse3_pshufbv8qi3): Add an expander for
21031 the now *-prefixed insn_and_split, turn the splitter const vec
21032 into an input for the insn, making it an ignored immediate for
21033 non-split cases, and loaded into the scratch register
21036 2021-03-23 Vladimir N. Makarov <vmakarov@redhat.com>
21039 * config/aarch64/constraints.md (Utq, UOb, UOh, UOw, UOd, UOty):
21040 Use define_relaxed_memory_constraint for them.
21042 2021-03-23 Iain Sandoe <iain@sandoe.co.uk>
21045 * config/host-darwin.c (darwin_gt_pch_use_address): Add a
21046 colon to the diagnostic message.
21048 2021-03-23 Ilya Leoshkevich <iii@linux.ibm.com>
21050 * fwprop.c (fwprop_propagation::fwprop_propagation): Look at
21052 (try_fwprop_subst_note): Use set_info instead of insn_info.
21053 (try_fwprop_subst_pattern): Likewise.
21054 (try_fwprop_subst_notes): Likewise.
21055 (try_fwprop_subst): Likewise.
21056 (forward_propagate_subreg): Likewise.
21057 (forward_propagate_and_simplify): Likewise.
21058 (forward_propagate_into): Likewise.
21059 * rtl-ssa/accesses.h (set_info::single_nondebug_use) New
21061 (set_info::single_nondebug_insn_use): Likewise.
21062 (set_info::single_phi_use): Likewise.
21063 * rtl-ssa/member-fns.inl (set_info::single_nondebug_use) New
21065 (set_info::single_nondebug_insn_use): Likewise.
21066 (set_info::single_phi_use): Likewise.
21068 2021-03-23 Christophe Lyon <christophe.lyon@linaro.org>
21070 * doc/sourcebuild.texi (arm_dsp_ok, arm_dsp): Document.
21072 2021-03-23 Jakub Jelinek <jakub@redhat.com>
21075 * config/aarch64/aarch64.c (aarch64_add_offset): Tell
21076 expand_mult to perform an unsigned rather than a signed
21079 2021-03-23 H.J. Lu <hjl.tools@gmail.com>
21082 * config/i386/cpuid.h (__cpuid): Add __volatile__.
21083 (__cpuid_count): Likewise.
21085 2021-03-23 Richard Biener <rguenther@suse.de>
21087 PR tree-optimization/99721
21088 * tree-vect-slp.c (vect_slp_analyze_node_operations):
21089 Make sure we can schedule the node.
21091 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
21093 * config/riscv/riscv.c (riscv_subword): Take endianness into
21094 account when calculating the byte offset.
21096 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
21098 * config/riscv/predicates.md (subreg_lowpart_operator): New predicate
21099 * config/riscv/riscv.md (*addsi3_extended2, *subsi3_extended2)
21100 (*negsi2_extended2, *mulsi3_extended2, *<optab>si3_mask)
21101 (*<optab>si3_mask_1, *<optab>di3_mask, *<optab>di3_mask_1)
21102 (*<optab>si3_extend_mask, *<optab>si3_extend_mask_1): Use
21103 new predicate "subreg_lowpart_operator"
21105 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
21107 * config/riscv/riscv.c (riscv_swap_instruction): New function
21108 to byteswap an SImode rtx containing an instruction.
21109 (riscv_trampoline_init): Byteswap the generated instructions
21112 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
21114 * common/config/riscv/riscv-common.c
21115 (TARGET_DEFAULT_TARGET_FLAGS): Set default endianness.
21116 * config.gcc (riscv32be-*, riscv64be-*): Set
21117 TARGET_BIG_ENDIAN_DEFAULT to 1.
21118 * config/riscv/elf.h (LINK_SPEC): Change -melf* value
21119 depending on default endianness.
21120 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
21121 * config/riscv/linux.h (LINK_SPEC): Likewise.
21122 * config/riscv/riscv.c (TARGET_DEFAULT_TARGET_FLAGS): Set
21123 default endianness.
21124 * config/riscv/riscv.h (DEFAULT_ENDIAN_SPEC): New macro.
21126 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
21128 * config/riscv/elf.h (LINK_SPEC): Pass linker endianness flag.
21129 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
21130 * config/riscv/linux.h (LINK_SPEC): Likewise.
21131 * config/riscv/riscv.h (ASM_SPEC): Pass -mbig-endian and
21133 (BYTES_BIG_ENDIAN): Handle big endian.
21134 (WORDS_BIG_ENDIAN): Define to BYTES_BIG_ENDIAN.
21135 * config/riscv/riscv.opt (-mbig-endian, -mlittle-endian): New
21137 * doc/invoke.texi (-mbig-endian, -mlittle-endian): Document.
21139 2021-03-23 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
21141 * regcprop.c (find_oldest_value_reg): Ask target whether
21142 different mode is fine for replacement register.
21144 2021-03-23 Aldy Hernandez <aldyh@redhat.com>
21146 PR tree-optimization/99296
21147 * value-range.cc (irange::irange_set_1bit_anti_range): New.
21148 (irange::irange_set_anti_range): Call irange_set_1bit_anti_range
21149 * value-range.h (irange::irange_set_1bit_anti_range): New.
21151 2021-03-22 Vladimir N. Makarov <vmakarov@redhat.com>
21154 * config/aarch64/constraints.md (UtQ): Use
21155 define_relaxed_memory_constraint for it.
21156 * doc/md.texi (define_relaxed_memory_constraint): Describe it.
21157 * genoutput.c (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
21158 * genpreds.c (constraint_data): Add bitfield is_relaxed_memory.
21159 (have_relaxed_memory_constraints): New static var.
21160 (relaxed_memory_start, relaxed_memory_end): Ditto.
21161 (add_constraint): Add arg is_relaxed_memory. Check name for
21162 relaxed memory. Set up is_relaxed_memory in constraint_data and
21163 have_relaxed_memory_constraints. Adjust calls.
21164 (choose_enum_order): Process relaxed memory.
21165 (write_tm_preds_h): Ditto.
21166 (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
21167 * gensupport.c (process_rtx): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
21168 * ira-costs.c (record_reg_classes): Process CT_RELAXED_MEMORY.
21169 * ira-lives.c (single_reg_class): Use
21170 insn_extra_relaxed_memory_constraint.
21171 * ira.c (ira_setup_alts): CT_RELAXED_MEMORY.
21172 * lra-constraints.c (valid_address_p): Use
21173 insn_extra_relaxed_memory_constraint instead of other memory
21175 (process_alt_operands): Process CT_RELAXED_MEMORY.
21176 (curr_insn_transform): Use insn_extra_relaxed_memory_constraint.
21177 * recog.c (asm_operand_ok, preprocess_constraints): Process
21179 * reload.c (find_reloads): Ditto.
21180 * rtl.def (DEFINE_RELAXED_MEMORY_CONSTRAINT): New.
21181 * stmt.c (parse_input_constraint): Use
21182 insn_extra_relaxed_memory_constraint.
21184 2021-03-22 Segher Boessenkool <segher@kernel.crashing.org>
21187 * ubsan.c (ubsan_instrument_float_cast): Don't test for unordered if
21190 2021-03-22 Alex Coplan <alex.coplan@arm.com>
21193 * config/arm/arm-protos.h (neon_make_constant): Add generate
21194 argument to guard emitting insns, default to true.
21195 * config/arm/arm.c (arm_legitimate_constant_p_1): Reject
21196 CONST_VECTORs which neon_make_constant can't handle.
21197 (neon_vdup_constant): Add generate argument, avoid emitting
21198 insns if it's not set.
21199 (neon_make_constant): Plumb new generate argument through.
21200 * config/arm/constraints.md (Ui): New. Use it...
21201 * config/arm/mve.md (*mve_mov<mode>): ... here.
21202 * config/arm/vec-common.md (movv8hf): Use neon_make_constant to
21203 synthesize constants.
21205 2021-03-22 Richard Biener <rguenther@suse.de>
21207 * debug.h: Add deprecation warning.
21209 2021-03-22 Richard Biener <rguenther@suse.de>
21211 PR tree-optimization/99694
21212 * tree-ssa-sccvn.c (visit_phi): Ignore edges with the
21215 2021-03-22 Kito Cheng <kito.cheng@sifive.com>
21218 * config/riscv/riscv.c (riscv_expand_block_move): Get RTL value
21219 after type checking.
21221 2021-03-22 Jakub Jelinek <jakub@redhat.com>
21225 * dwarf2out.c (get_full_len): Use get_precision rather than
21227 (add_const_value_attribute): Make sure add_AT_wide argument has
21228 precision prec rather than some very wide one.
21230 2021-03-22 Kewen Lin <linkw@linux.ibm.com>
21232 * config/rs6000/rs6000.md (*rotldi3_insert_sf,
21233 *mov<SFDF:mode><SFDF2:mode>cc_p9, floatsi<mode>2_lfiwax,
21234 floatsi<mode>2_lfiwax_mem, floatunssi<mode>2_lfiwzx,
21235 floatunssi<mode>2_lfiwzx_mem, *floatsidf2_internal,
21236 *floatunssidf2_internal, fix_trunc<mode>si2_stfiwx,
21237 fix_trunc<mode>si2_internal, fixuns_trunc<mode>si2_stfiwx,
21238 *round32<mode>2_fprs, *roundu32<mode>2_fprs,
21239 *fix_trunc<mode>si2_internal): Fix empty split condition.
21240 * config/rs6000/vsx.md (*vsx_le_undo_permute_<mode>,
21241 vsx_reduc_<VEC_reduc_name>_v2df, vsx_reduc_<VEC_reduc_name>_v4sf,
21242 *vsx_reduc_<VEC_reduc_name>_v2df_scalar,
21243 *vsx_reduc_<VEC_reduc_name>_v4sf_scalar): Likewise.
21245 2021-03-22 Xionghu Luo <luoxhu@linux.ibm.com>
21248 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
21249 Convert idx to DImode.
21250 (rs6000_expand_vector_set_var_p8): Likewise.
21252 2021-03-21 Jakub Jelinek <jakub@redhat.com>
21255 * dwarf2out.c (insert_float): Change return type from void to
21256 unsigned, handle GET_MODE_SIZE (mode) == 2 and return element size.
21257 (mem_loc_descriptor, loc_descriptor, add_const_value_attribute):
21260 2021-03-20 H.J. Lu <hjl.tools@gmail.com>
21263 * config/i386/i386.c (construct_container): Check cfun != NULL
21264 before accessing silent_p.
21266 2021-03-20 Ahamed Husni <ahamedhusni73@gmail.com>
21268 * asan.c: Fix typos in comments.
21270 2021-03-20 Vladimir N. Makarov <vmakarov@redhat.com>
21272 PR rtl-optimization/99680
21273 * lra-constraints.c (skip_contraint_modifiers): Rename to skip_constraint_modifiers.
21274 (process_address_1): Check empty constraint before using
21277 2021-03-19 Pat Haugen <pthaugen@linux.ibm.com>
21279 * config/rs6000/rs6000.c (power10_cost): New.
21280 (rs6000_option_override_internal): Set Power10 costs.
21281 (rs6000_issue_rate): Set Power10 issue rate.
21282 * config/rs6000/power10.md: Rewrite for Power10.
21284 2021-03-19 Vladimir N. Makarov <vmakarov@redhat.com>
21287 * lra-constraints.c (process_address_1): Don't use unknown
21288 constraint for address constraint.
21290 2021-03-19 Iain Sandoe <iain@sandoe.co.uk>
21293 * config.gcc (powerpc-*-darwin8): Delete the reference to
21294 the now removed darwin8.h.
21296 2021-03-19 Olivier Hainque <hainque@adacore.com>
21299 * config/vxworksae.h (VX_CPU_PREFIX): Define.
21301 2021-03-19 John David Anglin <danglin@gcc.gnu.org>
21303 * config/pa/pa.c (import_milli): Use memcpy instead of strncpy.
21305 2021-03-19 Tamar Christina <tamar.christina@arm.com>
21307 PR tree-optimization/99656
21308 * tree-vect-slp-patterns.c (linear_loads_p,
21309 complex_add_pattern::matches, is_eq_or_top,
21310 vect_validate_multiplication, complex_mul_pattern::matches,
21311 complex_fms_pattern::matches): Remove complex_perm_kinds_t.
21312 * tree-vectorizer.h: (complex_load_perm_t): Removed.
21313 (slp_tree_to_load_perm_map_t): Use complex_perm_kinds_t instead of
21314 complex_load_perm_t.
21316 2021-03-19 H.J. Lu <hjl.tools@gmail.com>
21319 * config/i386/i386-options.c (ix86_init_machine_status): Set
21321 * config/i386/i386.c (init_cumulative_args): Set silent_p to
21323 (construct_container): Return early for return and argument
21324 errors if silent_p is true.
21325 * config/i386/i386.h (machine_function): Add silent_p.
21327 2021-03-19 Jakub Jelinek <jakub@redhat.com>
21330 * config/arm/constraints.md (Ds): New constraint.
21331 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): Use w,Ds
21332 constraint instead of w,Dm.
21334 2021-03-19 Andrew Stubbs <ams@codesourcery.com>
21336 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Fix quotes
21339 2021-03-19 Eric Botcazou <ebotcazou@adacore.com>
21341 PR middle-end/99641
21342 * fold-const.c (native_encode_initializer) <CONSTRUCTOR>: For an
21343 array type, do the computation of the current position in sizetype.
21345 2021-03-18 Vladimir N. Makarov <vmakarov@redhat.com>
21348 * lra-constraints.c (process_address_1): Use lookup_constraint
21349 only for a single constraint.
21351 2021-03-18 Martin Sebor <msebor@redhat.com>
21353 PR middle-end/99502
21354 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): Rename...
21355 (inbounds_memaccess_p): ...to this. Check the ending offset of
21356 the accessed member.
21358 2021-03-18 Andrew Stubbs <ams@codesourcery.com>
21360 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Add %< and
21361 %> quote markers to error messages.
21362 (gcn_goacc_validate_dims): Likewise.
21363 (gcn_conditional_register_usage): Remove exclaimation mark from error
21365 (gcn_vectorize_vec_perm_const): Ensure perm is fully uninitialized.
21367 2021-03-18 Jan Hubicka <hubicka@ucw.cz>
21369 * config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
21372 2021-03-18 Sinan Lin <sinan@isrc.iscas.ac.cn>
21373 Kito Cheng <kito.cheng@sifive.com>
21375 * config/riscv/riscv.c (riscv_block_move_straight): Change type
21376 to unsigned HOST_WIDE_INT for parameter and local variable with
21377 HOST_WIDE_INT type.
21378 (riscv_adjust_block_mem): Ditto.
21379 (riscv_block_move_loop): Ditto.
21380 (riscv_expand_block_move): Ditto.
21382 2021-03-18 Nick Clifton <nickc@redhat.com>
21384 * config/v850/v850.c (construct_restore_jr): Increase static
21386 (construct_save_jarl): Likewise.
21387 * config/v850/v850.h (DWARF2_DEBUGGING_INFO): Define.
21389 2021-03-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21391 * config/aarch64/aarch64.c (aarch64_adjust_generic_arch_tuning): Define.
21392 (aarch64_override_options_internal): Use it.
21393 (generic_tunings): Add AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS to
21396 2021-03-17 Sandra Loosemore <sandra@codesourcery.com>
21398 * config/nios2/nios2.c (nios2_custom_check_insns): Clean up
21399 error message format issues.
21400 (nios2_option_override): Likewise.
21401 (nios2_expand_fpu_builtin): Likewise.
21402 (nios2_init_custom_builtins): Adjust to avoid bogus strncpy
21403 truncation warning.
21404 (nios2_expand_custom_builtin): More error message format fixes.
21405 (nios2_expand_rdwrctl_builtin): Likewise.
21406 (nios2_expand_rdprs_builtin): Likewise.
21407 (nios2_expand_eni_builtin): Likewise.
21408 (nios2_expand_builtin): Likewise.
21409 (nios2_register_custom_code): Likewise.
21410 (nios2_valid_target_attribute_rec): Likewise.
21411 (nios2_add_insn_asm): Fix uninitialized variable warning.
21413 2021-03-17 Jan Hubicka <jh@suse.cz>
21415 * config/i386/x86-tune-costs.h (struct processor_costs): Update costs
21416 of gather to match reality.
21417 * config/i386/x86-tune.def (X86_TUNE_USE_GATHER): Enable for znver3.
21419 2021-03-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21421 * config/aarch64/aarch64-builtins.c (aarch64_expand_rng_builtin): Use EQ
21422 to compare against CC_REG rather than NE.
21424 2021-03-17 H.J. Lu <hjl.tools@gmail.com>
21427 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support
21428 inline assembly statements.
21429 (ix86_print_operand): Update 'P' handling for -fno-plt.
21431 2021-03-17 Tamar Christina <tamar.christina@arm.com>
21434 * config/aarch64/aarch64.c
21435 (aarch64_simd_clone_compute_vecsize_and_simdlen): Remove unused var.
21437 2021-03-16 Segher Boessenkool <segher@kernel.crashing.org>
21440 * config/rs6000/predicates.md (branch_comparison_operator): Allow
21441 ordered and unordered for CCFPmode, if flag_finite_math_only.
21443 2021-03-16 Jakub Jelinek <jakub@redhat.com>
21446 * config/i386/i386-expand.c (ix86_split_lea_for_addr): Emit a MULT
21447 rather than ASHIFT.
21448 * config/i386/i386.md (mult by 1248 into ashift): New splitter.
21450 2021-03-16 Martin Liska <mliska@suse.cz>
21453 * optc-save-gen.awk: Add flag_ipa_ra to exceptions for
21454 cl_optimization_compare function.
21456 2021-03-16 Ilya Leoshkevich <iii@linux.ibm.com>
21458 * config/s390/s390.c (f_constraint_p): Treat "fv" constraints
21461 2021-03-16 Jakub Jelinek <jakub@redhat.com>
21464 * config/i386/i386.h (struct machine_function): Add
21465 has_explicit_vzeroupper bitfield.
21466 * config/i386/i386-expand.c (ix86_expand_builtin): Set
21467 cfun->machine->has_explicit_vzeroupper when expanding
21468 IX86_BUILTIN_VZEROUPPER.
21469 * config/i386/i386-features.c (rest_of_handle_insert_vzeroupper):
21470 Do the mode switching only when TARGET_VZEROUPPER, expensive
21471 optimizations turned on and not optimizing for size.
21472 (pass_insert_vzeroupper::gate): Enable even when
21473 cfun->machine->has_explicit_vzeroupper is set.
21475 2021-03-16 Jakub Jelinek <jakub@redhat.com>
21478 * config/aarch64/aarch64.c
21479 (aarch64_simd_clone_compute_vecsize_and_simdlen): If not a function
21480 definition, walk TYPE_ARG_TYPES list if non-NULL for argument types
21481 instead of DECL_ARGUMENTS. Ignore types for uniform arguments.
21483 2021-03-15 Richard Biener <rguenther@suse.de>
21485 PR tree-optimization/98834
21486 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle missing
21487 subsetting by truncating the access size.
21489 2021-03-15 Jan Hubicka <hubicka@ucw.cz>
21491 * config/i386/i386-options.c (processor_cost_table): Add znver3_cost.
21492 * config/i386/x86-tune-costs.h (znver3_cost): New gobal variable; copy
21495 2021-03-15 Martin Liska <mliska@suse.cz>
21497 * spellcheck.c: Add missing comma in initialization.
21499 2021-03-14 Uroš Bizjak <ubizjak@gmail.com>
21501 * config/i386/sse.md (*vec_extract<mode>): Merge alternative 0 with
21502 alternative 2 and alternative 1 with alternative 3 using
21503 YW register constraint.
21504 (*vec_extract<PEXTR_MODE12:mode>_zext): Merge alternatives
21505 using YW register constraint.
21506 (*vec_extractv16qi_zext): Ditto.
21507 (*vec_extractv4si): Merge alternatives 4 and 5
21508 using Yw register constraint.
21509 (*ssse3_palignr<mode>_perm): Use Yw instead of v for alternative 3.
21511 2021-03-13 Martin Sebor <msebor@redhat.com>
21513 PR tree-optimization/99489
21514 * builtins.c (gimple_call_alloc_size): Fail gracefully when argument
21515 is not a call statement.
21517 2021-03-13 Jakub Jelinek <jakub@redhat.com>
21519 PR tree-optimization/99544
21520 * match.pd (X + (X << C) -> X * (1 + (1 << C))): Don't simplify
21521 if for vector types multiplication can't be done in type's mode.
21523 2021-03-12 Eric Botcazou <ebotcazou@adacore.com>
21526 * config/sparc/constraints.md (w): Rename to...
21527 (W): ... this and ditch previous implementation.
21528 * config/sparc/sparc.md (*movdi_insn_sp64): Replace W with m.
21529 (*movdf_insn_sp64): Likewise.
21530 (*mov<VM64:mode>_insn_sp64): Likewise.
21531 * config/sparc/sync.md (*atomic_compare_and_swap<mode>_1): Replace
21533 (atomic_compare_and_swap_leon3_1): Likewise.
21534 (*atomic_compare_and_swapdi_v8plus): Likewise.
21535 * config/sparc/sparc.c (memory_ok_for_ldd): Remove useless test on
21536 architecture and add missing address validity check during LRA.
21538 2021-03-12 Tobias Burnus <tobias@codesourcery.com>
21541 * gimplify.c (omp_add_variable): Handle NULL_TREE as size
21542 occuring for assumed-size arrays in use_device_{ptr,addr}.
21544 2021-03-12 Jakub Jelinek <jakub@redhat.com>
21547 * config/i386/constraints.md (YW): New internal constraint.
21548 * config/i386/sse.md (v_Yw): Add V4TI, V2TI, V1TI and TI cases.
21549 (*<sse2_avx2>_<insn><mode>3<mask_name>,
21550 *<sse2_avx2>_uavg<mode>3<mask_name>, *abs<mode>2,
21551 *<s>mul<mode>3_highpart<mask_name>): Use <v_Yw> instead of v in
21553 (<sse2_avx2>_psadbw): Use YW instead of v in constraints.
21554 (*avx2_pmaddwd, *sse2_pmaddwd, *<code>v8hi3, *<code>v16qi3,
21555 avx2_pmaddubsw256, ssse3_pmaddubsw128): Merge last two alternatives
21556 into one, use Yw instead of former x,v.
21557 (ashr<mode>3, <insn><mode>3): Use <v_Yw> instead of x in constraints of
21558 the last alternative.
21559 (<sse2_avx2>_packsswb<mask_name>, <sse2_avx2>_packssdw<mask_name>,
21560 <sse2_avx2>_packuswb<mask_name>, <sse4_1_avx2>_packusdw<mask_name>,
21561 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, <ssse3_avx2>_palignr<mode>,
21562 <ssse3_avx2>_pshufb<mode>3<mask_name>): Merge last two alternatives
21563 into one, use <v_Yw> instead of former x,v.
21564 (avx2_interleave_highv32qi<mask_name>,
21565 vec_interleave_highv16qi<mask_name>): Use Yw instead of v in
21566 constraints. Add && <mask_avx512bw_condition> to condition.
21567 (avx2_interleave_lowv32qi<mask_name>,
21568 vec_interleave_lowv16qi<mask_name>,
21569 avx2_interleave_highv16hi<mask_name>,
21570 vec_interleave_highv8hi<mask_name>,
21571 avx2_interleave_lowv16hi<mask_name>, vec_interleave_lowv8hi<mask_name>,
21572 avx2_pshuflw_1<mask_name>, sse2_pshuflw_1<mask_name>,
21573 avx2_pshufhw_1<mask_name>, sse2_pshufhw_1<mask_name>,
21574 avx2_<code>v16qiv16hi2<mask_name>, sse4_1_<code>v8qiv8hi2<mask_name>,
21575 *sse4_1_<code>v8qiv8hi2<mask_name>_1, <sse2_avx2>_<insn><mode>3): Use
21576 Yw instead of v in constraints.
21577 * config/i386/mmx.md (Yv_Yw): New define_mode_attr.
21578 (*mmx_<insn><mode>3, mmx_ashr<mode>3, mmx_<insn><mode>3): Use <Yv_Yw>
21579 instead of Yv in constraints.
21580 (*mmx_<insn><mode>3, *mmx_mulv4hi3, *mmx_smulv4hi3_highpart,
21581 *mmx_umulv4hi3_highpart, *mmx_pmaddwd, *mmx_<code>v4hi3,
21582 *mmx_<code>v8qi3, mmx_pack<s_trunsuffix>swb, mmx_packssdw,
21583 mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
21584 *mmx_uavgv8qi3, *mmx_uavgv4hi3, mmx_psadbw): Use Yw instead of Yv in
21586 (*mmx_pinsrw, *mmx_pinsrb, *mmx_pextrw, *mmx_pextrw_zext, *mmx_pextrb,
21587 *mmx_pextrb_zext): Use YW instead of Yv in constraints.
21588 (*mmx_eq<mode>3, mmx_gt<mode>3): Use x instead of Yv in constraints.
21589 (mmx_andnot<mode>3, *mmx_<code><mode>3): Split last alternative into
21590 two, one with just x, another isa avx512vl with v.
21592 2021-03-12 Martin Liska <mliska@suse.cz>
21594 * doc/invoke.texi: Add missing param documentation.
21596 2021-03-11 David Malcolm <dmalcolm@redhat.com>
21599 * Makefile.in (ANALYZER_OBJS): Add analyzer/feasible-graph.o and
21600 analyzer/trimmed-graph.o.
21601 * doc/analyzer.texi (Analyzer Paths): Rewrite description of
21602 feasibility checking to reflect new implementation.
21603 * doc/invoke.texi (-fdump-analyzer-feasibility): Document new
21605 * shortest-paths.h (shortest_paths::get_shortest_distance): New.
21607 2021-03-11 David Malcolm <dmalcolm@redhat.com>
21609 * digraph.cc (selftest::test_shortest_paths): Update
21610 shortest_paths init for new param. Add test of
21611 SPS_TO_GIVEN_TARGET.
21612 * shortest-paths.h (enum shortest_path_sense): New.
21613 (shortest_paths::shortest_paths): Add "sense" param.
21614 Update for renamings. Generalize to use "sense" param.
21615 (shortest_paths::get_shortest_path): Rename param.
21616 (shortest_paths::m_sense): New field.
21617 (shortest_paths::m_prev): Rename...
21618 (shortest_paths::m_best_edge): ...to this.
21619 (shortest_paths::get_shortest_path): Update for renamings.
21620 Conditionalize flipping of path on sense of traversal.
21622 2021-03-11 David Malcolm <dmalcolm@redhat.com>
21624 * digraph.cc (selftest::test_shortest_paths): Add test coverage
21625 for paths from B and C.
21626 * shortest-paths.h (shortest_paths::shortest_paths): Handle
21627 unreachable nodes, rather than asserting.
21629 2021-03-11 David Edelsohn <dje.gcc@gmail.com>
21632 * config/rs6000/rs6000.c (rs6000_xcoff_file_start): Don't create
21633 xcoff_tbss_section_name.
21634 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .lcomm.
21635 * xcoffout.c (xcoff_tbss_section_name): Delete.
21636 * xcoffout.h (xcoff_tbss_section_name): Delete.
21638 2021-03-11 Richard Biener <rguenther@suse.de>
21640 PR tree-optimization/99523
21641 * tree-cfg.c (dump_function_to_file): Dump SSA names
21642 w/o identifier to the decls section as well, not only those
21643 without a VAR_DECL.
21645 2021-03-11 Jakub Jelinek <jakub@redhat.com>
21648 * ipa-icf-gimple.c (func_checker::compare_gimple_call): For internal
21649 function calls with lhs fail if the lhs don't have compatible types.
21651 2021-03-11 Hans-Peter Nilsson <hp@axis.com>
21653 * config/cris/cris.h (HARD_FRAME_POINTER_REGNUM): Define.
21654 Change FRAME_POINTER_REGNUM to correspond to a new faked
21655 register faked_fp, part of GENNONACR_REGS like faked_ap.
21656 (CRIS_FAKED_REGS_CONTENTS): New helper macro.
21657 (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS, CALL_USED_REGISTERS):
21658 (REG_ALLOC_ORDER, REG_CLASS_CONTENTS, REGNO_OK_FOR_BASE_P)
21659 (ELIMINABLE_REGS, REGISTER_NAMES): Adjust accordingly.
21660 * config/cris/cris.md (CRIS_FP_REGNUM): Renumber to new faked
21662 (CRIS_REAL_FP_REGNUM): New constant.
21663 * config/cris/cris.c (cris_reg_saved_in_regsave_area): Check
21664 for HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM.
21665 (cris_initial_elimination_offset): Handle elimination changes
21666 to HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM
21667 and add one from FRAME_POINTER_REGNUM to
21668 HARD_FRAME_POINTER_REGNUM.
21669 (cris_expand_prologue, cris_expand_epilogue): Emit code for
21670 hard_frame_pointer_rtx instead of frame_pointer_rtx.
21672 2021-03-10 David Edelsohn <dje.gcc@gmail.com>
21675 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Add check for DCmode.
21676 * config/rs6000/rs6000.c (rs6000_special_round_type_align): Same.
21678 2021-03-10 Vladimir N. Makarov <vmakarov@redhat.com>
21681 * lra-constraints.c (process_address_1): Don't check unknown
21682 constraint, use X for empty constraint.
21684 2021-03-10 Alex Coplan <alex.coplan@arm.com>
21686 * config/aarch64/aarch64.c (aarch64_vfp_is_call_or_return_candidate):
21687 Fix typo in comment describing "is_ha" argument.
21689 2021-03-10 John David Anglin <danglin@gcc.gnu.org>
21691 * doc/sourcebuild.texi: Document LRA target selector.
21693 2021-03-10 David Malcolm <dmalcolm@redhat.com>
21695 * doc/ux.texi: Add subsection contrasting interactive versus
21696 batch usage of GCC.
21698 2021-03-10 Joel Hutton <joel.hutton@arm.com>
21701 * tree-vect-stmts.c (vectorizable_store): Fix scatter store mask
21703 (vectorizable_load): Fix gather load mask check condition.
21705 2021-03-10 Richard Biener <rguenther@suse.de>
21707 PR tree-optimization/99510
21708 * tree.c (check_aligned_type): Check that the candidate
21709 has TYPE_USER_ALIGN set instead of matching with the
21712 2021-03-10 Eric Botcazou <ebotcazou@adacore.com>
21714 * config/sparc/sparc.c (sparc_regmode_natural_size): Return 4 for
21715 float and vector integer modes only if the mode is not larger.
21717 2021-03-10 Hans-Peter Nilsson <hp@axis.com>
21719 * config/cris/cris.h (DWARF_FRAME_REGISTERS): Define.
21721 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
21723 * ira.c (ira_setup_alts, ira_get_dup_out_num): Process digital
21725 * ira-lives.c (single_reg_class): Ditto.
21727 2021-03-09 Sebastian Huber <sebastian.huber@embedded-brains.de>
21729 * config.gcc (aarch64-*-rtems*): Include general rtems.h after
21730 the architecture-specific rtems.h.
21731 (aarch64-*-rtems*): Likewise.
21732 (arm*-*-rtems*): Likewise.
21733 (epiphany-*-rtems*): Likewise.
21734 (riscv*-*-rtems*): Likewise.
21736 2021-03-09 Jakub Jelinek <jakub@redhat.com>
21738 PR tree-optimization/99305
21739 * tree-ssa-phiopt.c (conditional_replacement): Test integer_pow2p
21740 before integer_all_onesp instead of vice versa.
21742 2021-03-09 Richard Earnshaw <rearnsha@arm.com>
21744 * common/config/arm/arm-common.c (arm_config_default): Change type
21745 of 'i' to unsigned.
21747 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
21750 * lra-constraints.c (process_address_1): Process constraint 'g'
21751 separately and digital constraints containing more one digit.
21753 2021-03-09 Nick Clifton <nickc@redhat.com>
21755 * config/rx/rx.h (DBX_DEBUGGING_INFO): Define.
21756 (DWARF"_DEBUGGING_INFO): Define.
21758 2021-03-09 Eric Botcazou <ebotcazou@adacore.com>
21761 * calls.c (initialize_argument_information): When the argument
21762 is passed by reference, do not make a copy in a thunk only if
21763 the argument is already in memory. Remove redundant test for
21764 the case of callee copy.
21766 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
21769 * lra-constraints.c (process_address_1): Process 0..9 constraints
21770 in process_address_1.
21772 2021-03-09 Andreas Krebbel <krebbel@linux.ibm.com>
21774 * config/s390/s390.c (struct s390_processor processor_table):
21775 Binutils name string must not be empty.
21777 2021-03-09 Claudiu Zissulescu <claziss@synopsys.com>
21779 * config/arc/arc.c (arc_attr_type): Remove function.
21781 2021-03-09 Martin Liska <mliska@suse.cz>
21784 * config/i386/i386-options.c (ix86_option_override_internal):
21785 Set isa_flags for OPTS argument and not for the global
21788 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
21790 * config/rs6000/predicates.md (ds_form_mem_operand): Check
21793 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
21796 * config/rs6000/predicates.md (ds_form_mem_operand) New
21798 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10) Use
21799 ds_form_mem_operand in ld/lwa patterns.
21800 * config/rs6000/fusion.md: Regenerate file.
21802 2021-03-08 Martin Sebor <msebor@redhat.com>
21804 PR middle-end/98266
21805 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): New function.
21806 (array_bounds_checker::check_array_bounds): Call it.
21808 2021-03-08 Martin Sebor <msebor@redhat.com>
21810 PR middle-end/97631
21811 * tree-ssa-strlen.c (maybe_warn_overflow): Test rawmem.
21812 (handle_builtin_stxncpy_strncat): Rename locals. Determine
21813 destination size from allocation calls. Issue a more appropriate
21815 (handle_builtin_memcpy): Pass true as rawmem to maybe_warn_overflow.
21816 (handle_builtin_memset): Same.
21818 2021-03-08 Peter Bergner <bergner@linux.ibm.com>
21821 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_permute): Add an assert
21822 to ensure we do not have an Altivec style address.
21823 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): Disable if passed
21824 an Altivec style address.
21825 (*vsx_le_perm_store_<mode>): Likewise.
21826 (splitters after *vsx_le_perm_store_<mode>): Likewise.
21827 (vsx_load_<mode>): Disable special expander if passed an Altivec
21829 (vsx_store_<mode>): Likewise.
21831 2021-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21834 * config/aarch64/predicates.md (aarch64_simd_shift_imm_vec_qi): Define.
21835 (aarch64_simd_shift_imm_vec_hi): Likewise.
21836 (aarch64_simd_shift_imm_vec_si): Likewise.
21837 (aarch64_simd_shift_imm_vec_di): Likewise.
21838 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le): Use
21839 predicate from above.
21840 (aarch64_shrn<mode>_insn_be): Likewise.
21841 (aarch64_rshrn<mode>_insn_le): Likewise.
21842 (aarch64_rshrn<mode>_insn_be): Likewise.
21843 (aarch64_shrn2<mode>_insn_le): Likewise.
21844 (aarch64_shrn2<mode>_insn_be): Likewise.
21845 (aarch64_rshrn2<mode>_insn_le): Likewise.
21846 (aarch64_rshrn2<mode>_insn_be): Likewise.
21848 2021-03-08 Vladimir N. Makarov <vmakarov@redhat.com>
21851 * lra-constraints.c (skip_contraint_modifiers): New function.
21852 (process_address_1): Use it before lookup_constraint call.
21854 2021-03-08 Martin Liska <mliska@suse.cz>
21857 * config/i386/i386-options.c (ix86_option_override_internal):
21858 Enable UINTR and HRESET for -march that supports it.
21860 2021-03-08 Ilya Leoshkevich <iii@linux.ibm.com>
21862 * config/s390/s390.c (f_constraint_p): New function.
21863 (s390_md_asm_adjust): Implement TARGET_MD_ASM_ADJUST.
21864 (TARGET_MD_ASM_ADJUST): Likewise.
21866 2021-03-08 Tobias Burnus <tobias@codesourcery.com>
21869 * tree-nested.c (convert_local_reference_stmt): Avoid calling
21870 lookup_field_for_decl for Fortran module (= namespace context).
21872 2021-03-08 Andreas Krebbel <krebbel@linux.ibm.com>
21874 * config/s390/s390.c (s390_expand_vec_compare): Implement <0
21875 comparison with arithmetic right shift.
21876 (s390_expand_vcond): No need for a force_reg anymore.
21877 s390_vec_compare will do it.
21878 * config/s390/vector.md ("vec_cmp<mode><tointvec>"): Accept also
21879 immediate operands.
21881 2021-03-07 Jakub Jelinek <jakub@redhat.com>
21884 * config/i386/constraints.md (Yw): Use SSE_REGS if TARGET_SSE
21885 but TARGET_AVX512BW or TARGET_AVX512VL is not set. Adjust description
21887 * config/i386/sse.md (v_Yw): New define_mode_attr.
21888 (*<insn><mode>3, *mul<mode>3<mask_name>, *avx2_<code><mode>3,
21889 *sse4_1_<code><mode>3<mask_name>): Use <v_Yw> instead of v
21891 * config/i386/mmx.md (mmx_pshufw_1, *vec_dupv4hi): Use Yw instead of
21892 xYw in constraints.
21894 2021-03-06 Julian Brown <julian@codesourcery.com>
21896 * tree-pretty-print.c (dump_generic_node): Emit non-generic
21897 address space info for aggregates.
21899 2021-03-06 Hans-Peter Nilsson <hp@axis.com>
21901 * config/cris/cris.h (MAX_FIXED_MODE_SIZE): Don't define.
21903 2021-03-05 Jakub Jelinek <jakub@redhat.com>
21905 PR middle-end/99322
21906 * tree-cfg.c (bb_to_omp_idx): New variable.
21907 (execute_build_cfg): Release the bb_to_omp_idx vector after
21908 cleanup_tree_cfg returns.
21909 (handle_abnormal_edges): Remove bb_to_omp_idx argument, adjust
21910 for bb_to_omp_idx being a vec<int> instead of pointer to array
21912 (make_edges): Remove bb_to_omp_idx local variable, don't pass
21913 it to handle_abnormal_edges, adjust for bb_to_omp_idx being a
21914 vec<int> instead of pointer to array of ints and don't free/release
21916 (remove_bb): When removing a bb and placing forced label somewhere
21917 else, ensure it is put into the same OpenMP region during cfg
21918 pass if possible or to entry successor as fallback. Unregister
21919 bb from bb_to_omp_idx.
21921 2021-03-05 Vladimir N. Makarov <vmakarov@redhat.com>
21924 * lra-constraints.c (process_address_1): Skip decomposing address
21925 for asm insn operand with unknown constraint.
21927 2021-03-05 Martin Jambor <mjambor@suse.cz>
21930 * cgraph.c (cgraph_edge::set_call_stmt): Do not update all
21931 corresponding speculative edges if we are about to resolve
21932 sepculation. Make edge direct (and so resolve speculations) before
21933 removing it from call_site_hash.
21934 (cgraph_edge::make_direct): Relax the initial assert to allow calling
21935 the function on speculative direct edges.
21937 2021-03-05 Eric Botcazou <ebotcazou@adacore.com>
21939 PR rtl-optimization/99376
21940 * rtlanal.c (nonzero_bits1) <arithmetic operators>: If the number
21941 of low-order zero bits is too large, set the result to 0 directly.
21943 2021-03-04 Jakub Jelinek <jakub@redhat.com>
21945 PR middle-end/93235
21946 * expmed.c (store_bit_field_using_insv): Return false of xop0 is a
21947 SUBREG and a SUBREG to op_mode can't be created.
21949 2021-03-04 Alex Coplan <alex.coplan@arm.com>
21952 * config/aarch64/aarch64-sve-builtins.cc
21953 (function_resolver::require_vector_type): Handle error_mark_node.
21955 2021-03-04 Ilya Leoshkevich <iii@linux.ibm.com>
21957 * cfgexpand.c (expand_asm_loc): Pass new parameter.
21958 (expand_asm_stmt): Likewise.
21959 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Add new
21961 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
21962 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
21963 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
21964 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
21965 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
21966 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
21967 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
21968 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
21969 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
21970 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
21971 * doc/tm.texi (md_asm_adjust): Likewise.
21972 * target.def (md_asm_adjust): Likewise.
21974 2021-03-04 Richard Biener <rguenther@suse.de>
21976 PR middle-end/97855
21977 * tree-pretty-print.c: Poison pp_printf.
21978 (dump_decl_name): Avoid use of pp_printf.
21979 (dump_block_node): Likewise.
21980 (dump_generic_node): Likewise.
21982 2021-03-04 Martin Sebor <msebor@redhat.com>
21984 PR middle-end/96963
21985 PR middle-end/94655
21986 * builtins.c (handle_array_ref): New helper.
21987 (handle_mem_ref): New helper.
21988 (compute_objsize_r): Factor out ARRAY_REF and MEM_REF handling
21989 into new helper functions. Correct a workaround for vectorized
21992 2021-03-03 Pat Haugen <pthaugen@linux.ibm.com>
21994 * config/rs6000/dfp.md (extendddtd2, trunctddd2, *cmp<mode>_internal1,
21995 floatditd2, ftrunc<mode>2, fix<mode>di2, dfp_ddedpd_<mode>,
21996 dfp_denbcd_<mode>, dfp_dxex_<mode>, dfp_diex_<mode>,
21997 *dfp_sgnfcnc_<mode>, dfp_dscli_<mode>, dfp_dscri_<mode>): Update size
21998 attribute for Power10.
21999 * config/rs6000/mma.md (*movoo): Likewise.
22000 * config/rs6000/rs6000.md (define_attr "size"): Add 256.
22001 (define_mode_attr bits): Add DD/TD modes.
22002 * config/rs6000/sync.md (load_quadpti, store_quadpti, load_lockedpti,
22003 store_conditionalpti): Update size attribute for Power10.
22005 2021-03-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22008 * config/sparc/t-sparc (tree-ssanames.o-warn): Don't error for
22009 -Wuninitialized, -Wmaybe-uninitialized.
22010 (wide-int.o-warn): Likewise.
22012 2021-03-03 Richard Earnshaw <rearnsha@arm.com>
22014 * common/config/arm/arm-common.c: Include configargs.h.
22015 (arm_config_default): New function.
22016 (arm_target_mode): Renamed from arm_target_thumb_only. Handle
22017 processors that do not support Thumb. Take into account the
22018 --with-mode configuration setting for selecting the default.
22019 * config/arm/arm.h (OPTION_DEFAULT_SPECS): Remove entry for 'mode'.
22020 (TARGET_MODE_SPEC_FUNCTIONS): Update for function name change.
22022 2021-03-03 Martin Liska <mliska@suse.cz>
22024 PR gcov-profile/97461
22025 * gcov-io.h (GCOV_PREALLOCATED_KVP): Remove.
22027 2021-03-03 Eric Botcazou <ebotcazou@adacore.com>
22030 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
22031 point back the hard frame pointer to its default location when the
22032 frame is larger than SEH_MAX_FRAME_SIZE.
22034 2021-03-03 Jakub Jelinek <jakub@redhat.com>
22037 * config/i386/predicates.md (logic_operator): New define_predicate.
22038 * config/i386/i386.md (mov + mem using comm arith peephole2):
22039 Punt if operands[1] is EXT_REX_SSE_REGNO_P, AVX512BW is not enabled
22040 and the inner mode is [QH]Imode.
22042 2021-03-03 Jakub Jelinek <jakub@redhat.com>
22045 * dwarf2out.c (dw_loc_list_struct): Add end_entry member.
22046 (new_loc_list): Clear end_entry.
22047 (output_loc_list): Only use DW_LLE_startx_length for -gsplit-dwarf
22048 if HAVE_AS_LEB128, otherwise use DW_LLE_startx_endx. Fix comment
22050 (index_location_lists): For dwarf_version >= 5 without HAVE_AS_LEB128,
22051 initialize also end_entry.
22053 2021-03-03 Jakub Jelinek <jakub@redhat.com>
22056 * cfgrtl.c (fixup_partitions): When changing some bbs from hot to cold
22057 partitions, if in non-layout mode after reorder_blocks also move
22058 affected blocks to ensure a single partition transition.
22060 2021-03-03 Jason Merrill <jason@redhat.com>
22063 * cgraphunit.c (process_function_and_variable_attributes): Don't
22064 warn about flatten on an alias if the target also has it.
22065 * cgraph.h (symtab_node::get_alias_target_tree): New.
22067 2021-03-02 David Edelsohn <dje.gcc@gmail.com>
22069 * config/rs6000/rs6000.md (tls_get_tpointer_internal): Prepend
22070 period to symbol name.
22071 (tls_get_addr_internal<mode>): Same.
22073 2021-03-02 David Malcolm <dmalcolm@redhat.com>
22076 * diagnostic-show-locus.c
22077 (selftest::test_one_liner_many_fixits_2): Fix accidental usage of
22080 2021-03-02 Martin Sebor <msebor@redhat.com>
22082 PR middle-end/99276
22083 * builtins.c (warn_for_access): Remove stray warning text.
22085 2021-03-02 Martin Sebor <msebor@redhat.com>
22087 PR middle-end/99295
22088 * doc/extend.texi (attribute malloc): Reword and clarify nonaliasing
22091 2021-03-02 Jakub Jelinek <jakub@redhat.com>
22094 * dwarf2out.c (output_macinfo_op): Use DW_MACRO_*_str* even with
22095 -gdwarf-5 -gstrict-dwarf. For -gsplit-dwarf -gdwarf-5 use
22096 DW_MACRO_*_strx instead of DW_MACRO_*_strp. Handle
22097 DW_MACRO_define_strx and DW_MACRO_undef_strx.
22098 (save_macinfo_strings): Use DW_MACRO_*_str* even with
22099 -gdwarf-5 -gstrict-dwarf. Handle DW_MACRO_define_strx and
22100 DW_MACRO_undef_strx.
22102 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
22104 * config/s390/s390-builtin-types.def (BT_FN_V4SF_V8HI_UINT): New
22106 (BT_FN_V8HI_V8HI_UINT): Likewise.
22107 (BT_FN_V8HI_V4SF_V4SF_UINT): Likewise.
22108 * config/s390/s390-builtins.def (B_NNPA): New macro definition.
22109 (s390_vclfnhs, s390_vclfnls, s390_vcrnfs, s390_vcfn, s390_vcnf):
22110 New builtin definitions.
22111 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Bump
22112 vector extension version.
22113 * config/s390/s390.c (s390_expand_builtin): Check if builtins are
22114 available with current -march level.
22115 * config/s390/s390.md (UNSPEC_NNPA_VCLFNHS_V8HI)
22116 (UNSPEC_NNPA_VCLFNLS_V8HI, UNSPEC_NNPA_VCRNFS_V8HI)
22117 (UNSPEC_NNPA_VCFN_V8HI, UNSPEC_NNPA_VCNF_V8HI): New constants.
22118 * config/s390/vecintrin.h (vec_extend_to_fp32_hi): New macro.
22119 (vec_extend_to_fp32_lo): Likewise.
22120 (vec_round_from_fp32): Likewise.
22121 (vec_convert_to_fp16): Likewise.
22122 (vec_convert_from_fp16): Likewise.
22123 * config/s390/vx-builtins.md (vclfnhs_v8hi): New insn pattern.
22124 (vclfnls_v8hi): Likewise.
22125 (vcrnfs_v8hi): Likewise.
22126 (vcfn_v8hi): Likewise.
22127 (vcnf_v8hi): Likewise.
22129 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
22131 * common/config/s390/s390-common.c (processor_flags_table): New entry.
22132 * config.gcc: Enable arch14 for --with-arch and --with-tune.
22133 * config/s390/driver-native.c (s390_host_detect_local_cpu): Pick
22134 arch14 for unknown CPU models.
22135 * config/s390/s390-opts.h (enum processor_type): Add PROCESSOR_ARCH14.
22136 * config/s390/s390.c (s390_issue_rate): Add case for PROCESSOR_ARCH14.
22137 (s390_get_sched_attrmask): Likewise.
22138 (s390_get_unit_mask): Likewise.
22139 * config/s390/s390.h (enum processor_flags): Add PF_NNPA and PF_ARCH14.
22140 (TARGET_CPU_ARCH14, TARGET_CPU_ARCH14_P, TARGET_CPU_NNPA)
22141 (TARGET_CPU_NNPA_P, TARGET_ARCH14, TARGET_ARCH14_P, TARGET_NNPA)
22142 (TARGET_NNPA_P): New macro definitions.
22143 * config/s390/s390.md ("cpu_facility", "enabled"): Add arch14 and nnpa.
22144 * config/s390/s390.opt: Add PROCESSOR_ARCH14.
22146 2021-03-02 Jakub Jelinek <jakub@redhat.com>
22148 PR middle-end/95757
22149 * tree-vrp.c (register_edge_assert_for): Remove superfluous ()s around
22150 condition. Call register_edge_assert_for_1 for == 0, != 0, == 1 and
22151 != 1 comparisons if name is lhs of a comparison.
22153 2021-03-01 Iain Sandoe <iain@sandoe.co.uk>
22157 * config/darwin-protos.h (darwin_should_restore_cfa_state): New.
22158 * config/darwin.c (darwin_should_restore_cfa_state): New.
22159 * config/darwin.h (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New.
22160 * doc/tm.texi: Regenerated.
22161 * doc/tm.texi.in: Document TARGET_ASM_SHOULD_RESTORE_CFA_STATE.
22162 * dwarf2cfi.c (connect_traces): If the target requests, restore
22163 the CFA expression after a DW_CFA_restore.
22164 * target.def (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New hook.
22166 2021-03-01 Martin Liska <mliska@suse.cz>
22169 * optc-save-gen.awk: Add 4 more exceptions.
22171 2021-03-01 Nathan Sidwell <nathan@acm.org>
22174 * tree.h (TYPE_ALIGN_RAW): New accessor.
22175 (TYPE_ALIGN): Use it.
22177 2021-03-01 Jan Hubicka <jh@suse.cz>
22180 * ipa-fnsummary.c (compute_fn_summary): Fix sanity check.
22182 2021-03-01 Eric Botcazou <ebotcazou@adacore.com>
22185 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
22186 point the hard frame pointer to the SSE register save area instead
22187 of the general register save area. Perform only minimal adjustment
22188 for small frames if it is initially not correctly aligned.
22189 (ix86_expand_prologue): Remove early saves for a SEH target.
22190 * config/i386/winnt.c (struct seh_frame_state): Document constraint.
22192 2021-02-28 Jakub Jelinek <jakub@redhat.com>
22195 * ipa.c (symbol_table::remove_unreachable_nodes): Fix a comment
22196 typo - referneced -> referenced.
22197 * tree.c (component_ref_size): Fix comment typo -
22198 refernce -> reference.
22199 * tree-ssa-alias.c (access_path_may_continue_p): Fix comment typo -
22200 traling -> trailing.
22201 (aliasing_component_refs_p): Fix comment typos -
22202 refernce -> reference and refernece -> reference and
22203 traling -> trailing.
22204 (nonoverlapping_refs_since_match_p): Fix comment typo -
22205 referneces -> references.
22206 * doc/invoke.texi (--param modref-max-bases): Fix a typo -
22207 referneces -> references.
22209 2021-02-27 Iain Sandoe <iain@sandoe.co.uk>
22211 * config/host-darwin.c (darwin_gt_pch_use_address): Modify
22212 diagnostic message to avoid use of a contraction and format
22215 2021-02-27 Jakub Jelinek <jakub@redhat.com>
22218 * gcse.c (gcse_or_cprop_is_too_expensive): Use %wu instead of
22219 HOST_WIDE_INT_PRINT_UNSIGNED in warning format string.
22220 * ipa-devirt.c (ipa_odr_read_section): Use %wd instead of
22221 HOST_WIDE_INT_PRINT_DEC in inform format string. Fix comment
22224 2021-02-26 Richard Biener <rguenther@suse.de>
22226 PR middle-end/99281
22227 * expr.c (store_field): For calls with return-slot optimization
22228 and addressable return type expand the store directly.
22230 2021-02-26 Richard Biener <rguenther@suse.de>
22233 * builtins.c (warn_string_no_nul): Fix diagnostic formatting.
22235 2021-02-26 Peter Bergner <bergner@linux.ibm.com>
22238 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Replace assert
22241 2021-02-26 Aaron Sawdey <acsawdey@linux.ibm.com>
22243 * config.gcc: Add rs6000-pcrel-opt.o.
22244 * config/rs6000/rs6000-pcrel-opt.c: New file.
22245 * config/rs6000/pcrel-opt.md: New file.
22246 * config/rs6000/predicates.md: Add d_form_memory predicate.
22247 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_PCREL_OPT.
22248 * config/rs6000/rs6000-passes.def: Add pass_pcrel_opt.
22249 * config/rs6000/rs6000-protos.h: Add reg_to_non_prefixed(),
22250 pcrel_opt_valid_mem_p(), output_pcrel_opt_reloc(),
22251 and make_pass_pcrel_opt().
22252 * config/rs6000/rs6000.c (reg_to_non_prefixed): Make global.
22253 (rs6000_option_override_internal): Add pcrel-opt.
22254 (rs6000_delegitimize_address): Support pcrel-opt.
22255 (rs6000_opt_masks): Add pcrel-opt.
22256 (pcrel_opt_valid_mem_p): New function.
22257 (reg_to_non_prefixed): Make global.
22258 (rs6000_asm_output_opcode): Reset prepend_p_to_next_insn.
22259 (output_pcrel_opt_reloc): New function.
22260 * config/rs6000/rs6000.md (loads_extern_addr): New attr.
22261 (pcrel_extern_addr): Set loads_extern_addr.
22262 Add include for pcrel-opt.md.
22263 * config/rs6000/rs6000.opt: Add -mpcrel-opt.
22264 * config/rs6000/t-rs6000: Add rules for pcrel-opt.c and
22267 2021-02-26 YunQiang Su <yunqiang.su@cipunited.com>
22270 * config/mips/mips.c (mips_expand_ext_as_unaligned_load):
22271 If TARGET_64BIT and dest is SUBREG, we check the width, if it
22272 equal to SImode, we use SImode operation, just like what we are
22275 2021-02-26 Marek Polacek <polacek@redhat.com>
22277 * builtins.c (warn_for_access): Fix typos.
22279 2021-02-25 Iain Sandoe <iain@sandoe.co.uk>
22281 * config/aarch64/aarch64.md (<optab>_rol<mode>3): Add a '#'
22282 mark in front of the immediate quantity.
22283 (<optab>_rolsi3_uxtw): Likewise.
22285 2021-02-25 Richard Earnshaw <rearnsha@arm.com>
22288 * config/arm/thumb2.md (nonsecure_call_reg_thumb2_fpcxt): New pattern.
22289 (nonsecure_call_value_reg_thumb2_fpcxt): Likewise.
22290 (nonsecure_call_reg_thumb2): Restrict to using r4 for the callee
22291 address and disable when the FPCXT is not available.
22292 (nonsecure_call_value_reg_thumb2): Likewise.
22294 2021-02-25 Nathan Sidwell <nathan@acm.org>
22297 * doc/invoke.texi (flang-info-module-cmi): Renamed option.
22299 2021-02-25 Tamar Christina <tamar.christina@arm.com>
22301 * tree-vect-slp.c (optimize_load_redistribution_1): Abort on NULL nodes.
22303 2021-02-25 Richard Biener <rguenther@suse.de>
22305 PR tree-optimization/99253
22306 * tree-vect-loop.c (check_reduction_path): First compute
22307 code, then verify out-of-loop uses.
22309 2021-02-25 Jakub Jelinek <jakub@redhat.com>
22312 * match.pd ((T)(A) + CST -> (T)(A + CST)): Add :s to convert.
22314 2021-02-25 Jakub Jelinek <jakub@redhat.com>
22316 PR tree-optimization/80635
22317 * tree-vrp.c (vrp_simplify_cond_using_ranges): Also handle
22318 VIEW_CONVERT_EXPR if modes are the same, innerop is integral and
22319 has mode precision.
22321 2021-02-25 Richard Biener <rguenther@suse.de>
22323 * tree-vect-slp.c (optimize_load_redistribution_1): Delay
22324 load_map population.
22325 (vect_match_slp_patterns_2): Revert part of last change.
22326 (vect_analyze_slp): Do not interleave optimize_load_redistribution
22327 with pattern detection but do it afterwards. Dump the
22328 whole SLP graph after pattern recognition and load
22329 redistribution optimization finished.
22331 2021-02-24 Jakub Jelinek <jakub@redhat.com>
22334 * omp-low.c (struct omp_context): Add teams_nested_p and
22335 nonteams_nested_p members.
22336 (scan_omp_target): Diagnose teams nested inside of target with other
22337 directives strictly nested inside of the same target.
22338 (check_omp_nesting_restrictions): Set ctx->teams_nested_p or
22339 ctx->nonteams_nested_p as needed.
22341 2021-02-24 Vladimir N. Makarov <vmakarov@redhat.com>
22343 PR inline-asm/99123
22344 * lra-constraints.c (uses_hard_regs_p): Don't use decompose_mem_address.
22346 2021-02-24 Hans-Peter Nilsson <hp@axis.com>
22348 * config/cris/cris.c (cris_expand_prologue): Set
22349 current_function_static_stack_size, if flag_stack_usage_info.
22351 2021-02-24 Pat Haugen <pthaugen@linux.ibm.com>
22353 * config/rs6000/rs6000.c (next_insn_prefixed_p): Rename.
22354 (rs6000_final_prescan_insn): Adjust.
22355 (rs6000_asm_output_opcode): Likewise.
22357 2021-02-24 Martin Sebor <msebor@redhat.com>
22359 PR middle-end/97172
22360 * attribs.c (attr_access::free_lang_data): Clear attribute arg spec
22361 from function arguments.
22363 2021-02-24 Tamar Christina <tamar.christina@arm.com>
22365 PR tree-optimization/99220
22366 * tree-vect-slp.c (optimize_load_redistribution_1): Remove
22367 node from cache when it's about to be deleted.
22369 2021-02-24 Jakub Jelinek <jakub@redhat.com>
22371 PR tree-optimization/99225
22372 * fold-const.c (fold_binary_loc) <case NE_EXPR>: In (x & (1 << y)) != 0
22373 to ((x >> y) & 1) != 0 simplifications use build_one_cst instead of
22374 build_int_cst (..., 1). Formatting fixes.
22376 2021-02-24 Tamar Christina <tamar.christina@arm.com>
22378 PR tree-optimization/99149
22379 * tree-vect-slp-patterns.c (vect_detect_pair_op): Don't recreate the
22381 (vect_slp_reset_pattern): Remove.
22382 (complex_fma_pattern::matches): Remove call to vect_slp_reset_pattern.
22383 (complex_mul_pattern::build, complex_fma_pattern::build,
22384 complex_fms_pattern::build): Fix ref counts.
22385 * tree-vect-slp.c (vect_free_slp_tree): Undo SLP only pattern relevancy
22386 when node is being deleted.
22387 (vect_match_slp_patterns_2): Correct result of cache hit on patterns.
22388 (vect_schedule_slp): Invalidate SLP_TREE_REPRESENTATIVE of removed
22390 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize value.
22392 2021-02-24 Matthias Klose <doko@ubuntu.com>
22395 2020-12-07 Matthias Klose <doko@ubuntu.com>
22397 * genextract.c (print_header): Undefine ENABLE_RTL_CHECKING
22398 and ENABLE_RTL_FLAG_CHECKING.
22400 2021-02-24 Richard Biener <rguenther@suse.de>
22403 * builtins.c (fold_builtin_next_arg): Avoid NULL arg.
22405 2021-02-23 Peter Bergner <bergner@linux.ibm.com>
22407 * config/rs6000/mma.md (mma_assemble_pair): Rename from this...
22408 (vsx_assemble_pair): ...to this.
22409 (*mma_assemble_pair): Rename from this...
22410 (*vsx_assemble_pair): ...to this.
22411 (mma_disassemble_pair): Rename from this...
22412 (vsx_disassemble_pair): ...to this.
22413 (*mma_disassemble_pair): Rename from this...
22414 (*vsx_disassemble_pair): ...to this.
22415 * config/rs6000/rs6000-builtin.def (BU_MMA_V2, BU_MMA_V3,
22416 BU_COMPAT): New macros.
22417 (mma_assemble_pair): Rename from this...
22418 (vsx_assemble_pair): ...to this.
22419 (mma_disassemble_pair): Rename from this...
22420 (vsx_disassemble_pair): ...to this.
22421 (mma_assemble_pair): New compatibility built-in.
22422 (mma_disassemble_pair): Likewise.
22423 * config/rs6000/rs6000-call.c (struct builtin_compatibility): New.
22424 (RS6000_BUILTIN_COMPAT): Define.
22425 (bdesc_compat): New.
22426 (mma_expand_builtin): Use VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
22427 (rs6000_gimple_fold_mma_builtin): Use MMA_BUILTIN_DISASSEMBLE_PAIR
22428 and VSX_BUILTIN_ASSEMBLE_PAIR.
22429 (rs6000_init_builtins): Register compatibility built-ins.
22430 (mma_init_builtins): Use VSX_BUILTIN_ASSEMBLE_PAIR,
22431 VSX_BUILTIN_ASSEMBLE_PAIR_INTERNAL, VSX_BUILTIN_DISASSEMBLE_PAIR and
22432 VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
22433 * doc/extend.texi (__builtin_mma_assemble_pair): Rename from this...
22434 (__builtin_vsx_assemble_pair): ...to this.
22435 (__builtin_mma_disassemble_pair): Rename from this...
22436 (__builtin_vsx_disassemble_pair): ...to this.
22438 2021-02-23 Martin Liska <mliska@suse.cz>
22441 * ipa-icf.c (sem_variable::merge): Do not merge 2 variables
22442 with different alignment. That leads to an invalid red zone
22443 size allocated in runtime.
22445 2021-02-23 Jakub Jelinek <jakub@redhat.com>
22447 PR tree-optimization/99204
22448 * fold-const.c (fold_read_from_constant_string): Check that
22449 tree_fits_uhwi_p (index) rather than just that index is INTEGER_CST.
22451 2021-02-23 Segher Boessenkool <segher@kernel.crashing.org>
22452 Kewen Lin <linkw@gcc.gnu.org>
22454 * config/rs6000/rs6000.md (*rotl<mode>3_insert_3): Renamed to...
22455 (rotl<mode>3_insert_3): ...this.
22456 (plus_ior_xor): New code_iterator.
22457 (define_split for GPR rl*imi): New splitter.
22458 * config/rs6000/vsx.md (vsx_init_v4si): Use gen_rotldi3_insert_3
22459 for integer merging.
22461 2021-02-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22463 * config/aarch64/aarch64-tuning-flags.def (cse_sve_vl_constants):
22465 * config/aarch64/aarch64.md (add<mode>3): Force CONST_POLY_INT immediates
22466 into a register when the above is enabled.
22467 * config/aarch64/aarch64.c (neoversev1_tunings):
22468 AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
22469 (aarch64_rtx_costs): Use AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
22471 2021-02-22 Hans-Peter Nilsson <hp@axis.com>
22473 * config/cris/cris.c (cris_print_operand) <'T'>: Change
22474 valid operand from is now an addi mult-value to shift-value.
22475 * config/cris/cris.md (*addi): Change expression of scaled
22476 operand from mult to ashift.
22477 * config/cris/cris.md (*addi_reload): New insn_and_split.
22479 2021-02-22 John David Anglin <danglin@gcc.gnu.org>
22482 * config/pa/pa.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Define as
22483 hook_bool_const_tree_hwi_hwi_const_tree_true.
22484 (pa_asm_output_mi_thunk): Add support for nonzero vcall_offset.
22486 2021-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
22488 PR rtl-optimization/98791
22489 * ira-conflicts.c (process_regs_for_copy): Don't create allocno copies
22490 for unordered modes.
22492 2021-02-22 Martin Liska <mliska@suse.cz>
22494 * tree-inline.c (inline_forbidden_p): Set
22495 inline_forbidden_reason.
22497 2021-02-22 Richard Biener <rguenther@suse.de>
22499 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Dump
22502 2021-02-22 Richard Biener <rguenther@suse.de>
22504 PR tree-optimization/99165
22505 * gimple-ssa-store-merging.c (pass_store_merging::process_store):
22506 Accumulate changed to ret.
22508 2021-02-21 Uros Bizjak <ubizjak@gmail.com>
22511 2020-12-09 Uroš Bizjak <ubizjak@gmail.com>
22513 * config/i386/i386.h (REG_ALLOC_ORDER): Remove
22515 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
22518 * config/s390/vector.md (trunctf<DFP_ALL:mode>2_vr): New
22520 (trunctf<DFP_ALL:mode>2): Likewise.
22521 (trunctdtf2_vr): Likewise.
22522 (trunctdtf2): Likewise.
22523 (extend<DFP_ALL:mode>tf2_vr): Likewise.
22524 (extend<DFP_ALL:mode>tf2): Likewise.
22525 (extendtftd2_vr): Likewise.
22526 (extendtftd2): Likewise.
22528 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
22530 * config/s390/vector.md (*fprx2_to_tf): Rename to fprx2_to_tf,
22531 add memory alternative.
22532 (tf_to_fprx2): New pattern.
22534 2021-02-19 Martin Sebor <msebor@redhat.com>
22537 * attribs.c (init_attr_rdwr_indices): Guard vblist use.
22538 (attr_access::free_lang_data): Remove a spurious test.
22540 2021-02-19 Nathan Sidwell <nathan@acm.org>
22542 * doc/invoke.texi (flang-info-module-read): Document.
22544 2021-02-19 Martin Liska <mliska@suse.cz>
22546 PR translation/99167
22547 * params.opt: Fix typo.
22549 2021-02-19 Richard Biener <rguenther@suse.de>
22551 PR middle-end/99122
22552 * tree-inline.c (inline_forbidden_p): Do not inline functions
22553 with VLA arguments or return value.
22555 2021-02-19 Jakub Jelinek <jakub@redhat.com>
22558 * config/arm/arm.md (*stack_protect_combined_set_insn,
22559 *stack_protect_combined_test_insn): If force_const_mem result
22560 is not valid general operand, force its address into the destination
22563 2021-02-19 Jakub Jelinek <jakub@redhat.com>
22566 * tree-cfg.c (gimple_merge_blocks): If bb a starts with eh landing
22567 pad or non-local label, put FORCED_LABELs from bb b after that label
22568 rather than before it.
22570 2021-02-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
22573 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3): Use
22574 expand_vector_broadcast' to emit the vec_duplicate operand.
22576 2021-02-18 Vladimir N. Makarov <vmakarov@redhat.com>
22578 PR rtl-optimization/96264
22579 * lra-remat.c (reg_overlap_for_remat_p): Check also output insn
22582 2021-02-18 H.J. Lu <hjl.tools@gmail.com>
22585 * varasm.c (get_section): Replace SUPPORTS_SHF_GNU_RETAIN with
22586 looking up the retain attribute.
22587 (resolve_unique_section): Likewise.
22588 (get_variable_section): Likewise.
22589 (switch_to_section): Likewise. Warn when a symbol without the
22590 retain attribute and a symbol with the retain attribute are
22591 placed in the section with the same name, instead of the used
22593 * doc/extend.texi: Document the "retain" attribute.
22595 2021-02-18 Nathan Sidwell <nathan@acm.org>
22598 * doc/invoke.texi (flang-info-include-translate): Document header
22601 2021-02-18 Richard Biener <rguenther@suse.de>
22603 PR middle-end/99122
22604 * ipa-fnsummary.c (analyze_function_body): Set
22605 CIF_FUNCTION_NOT_INLINABLE for VLA parameter calls.
22606 * tree-inline.c (insert_init_debug_bind): Pass NULL for
22607 error_mark_node values.
22608 (force_value_to_type): Do not build V_C_Es for WITH_SIZE_EXPR
22610 (setup_one_parameter): Delay force_value_to_type until when
22613 2021-02-18 Hans-Peter Nilsson <hp@axis.com>
22615 PR tree-optimization/99142
22616 * match.pd (clz cmp 0): Gate replacement on single_use of clz result.
22618 2021-02-18 Jakub Jelinek <jakub@redhat.com>
22620 * wide-int-bitmask.h (wide_int_bitmask::wide_int_bitmask (),
22621 wide_int_bitmask::wide_int_bitmask (uint64_t),
22622 wide_int_bitmask::wide_int_bitmask (uint64_t, uint64_t),
22623 wide_int_bitmask::operator ~ () const,
22624 wide_int_bitmask::operator | (wide_int_bitmask) const,
22625 wide_int_bitmask::operator & (wide_int_bitmask) const): Use constexpr
22627 * config/i386/i386.h (PTA_3DNOW, PTA_3DNOW_A, PTA_64BIT, PTA_ABM,
22628 PTA_AES, PTA_AVX, PTA_BMI, PTA_CX16, PTA_F16C, PTA_FMA, PTA_FMA4,
22629 PTA_FSGSBASE, PTA_LWP, PTA_LZCNT, PTA_MMX, PTA_MOVBE, PTA_NO_SAHF,
22630 PTA_PCLMUL, PTA_POPCNT, PTA_PREFETCH_SSE, PTA_RDRND, PTA_SSE, PTA_SSE2,
22631 PTA_SSE3, PTA_SSE4_1, PTA_SSE4_2, PTA_SSE4A, PTA_SSSE3, PTA_TBM,
22632 PTA_XOP, PTA_AVX2, PTA_BMI2, PTA_RTM, PTA_HLE, PTA_PRFCHW, PTA_RDSEED,
22633 PTA_ADX, PTA_FXSR, PTA_XSAVE, PTA_XSAVEOPT, PTA_AVX512F, PTA_AVX512ER,
22634 PTA_AVX512PF, PTA_AVX512CD, PTA_NO_TUNE, PTA_SHA, PTA_PREFETCHWT1,
22635 PTA_CLFLUSHOPT, PTA_XSAVEC, PTA_XSAVES, PTA_AVX512DQ, PTA_AVX512BW,
22636 PTA_AVX512VL, PTA_AVX512IFMA, PTA_AVX512VBMI, PTA_CLWB, PTA_MWAITX,
22637 PTA_CLZERO, PTA_NO_80387, PTA_PKU, PTA_AVX5124VNNIW, PTA_AVX5124FMAPS,
22638 PTA_AVX512VPOPCNTDQ, PTA_SGX, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES,
22639 PTA_AVX512VBMI2, PTA_VPCLMULQDQ, PTA_AVX512BITALG, PTA_RDPID,
22640 PTA_PCONFIG, PTA_WBNOINVD, PTA_AVX512VP2INTERSECT, PTA_PTWRITE,
22641 PTA_AVX512BF16, PTA_WAITPKG, PTA_MOVDIRI, PTA_MOVDIR64B, PTA_ENQCMD,
22642 PTA_CLDEMOTE, PTA_SERIALIZE, PTA_TSXLDTRK, PTA_AMX_TILE, PTA_AMX_INT8,
22643 PTA_AMX_BF16, PTA_UINTR, PTA_HRESET, PTA_KL, PTA_WIDEKL, PTA_AVXVNNI,
22644 PTA_X86_64_BASELINE, PTA_X86_64_V2, PTA_X86_64_V3, PTA_X86_64_V4,
22645 PTA_CORE2, PTA_NEHALEM, PTA_WESTMERE, PTA_SANDYBRIDGE, PTA_IVYBRIDGE,
22646 PTA_HASWELL, PTA_BROADWELL, PTA_SKYLAKE, PTA_SKYLAKE_AVX512,
22647 PTA_CASCADELAKE, PTA_COOPERLAKE, PTA_CANNONLAKE, PTA_ICELAKE_CLIENT,
22648 PTA_ICELAKE_SERVER, PTA_TIGERLAKE, PTA_SAPPHIRERAPIDS, PTA_ALDERLAKE,
22649 PTA_KNL, PTA_BONNELL, PTA_SILVERMONT, PTA_GOLDMONT, PTA_GOLDMONT_PLUS,
22650 PTA_TREMONT, PTA_KNM): Use constexpr instead of const.
22652 2021-02-18 Jakub Jelinek <jakub@redhat.com>
22654 PR middle-end/99109
22655 * gimple-array-bounds.cc (build_zero_elt_array_type): Rename to ...
22656 (build_printable_array_type): ... this. Add nelts argument. For
22657 overaligned eltype, use TYPE_MAIN_VARIANT (eltype) instead. If
22658 nelts, call build_array_type_nelts.
22659 (array_bounds_checker::check_mem_ref): Use build_printable_array_type
22660 instead of build_zero_elt_array_type and build_array_type_nelts.
22662 2021-02-18 Jakub Jelinek <jakub@redhat.com>
22665 * config/i386/i386.c (distance_non_agu_define): Don't call
22666 extract_insn_cached here.
22667 (ix86_lea_outperforms): Save and restore recog_data around call
22668 to distance_non_agu_define and distance_agu_use.
22669 (ix86_ok_to_clobber_flags): Remove.
22670 (ix86_avoid_lea_for_add): Don't call ix86_ok_to_clobber_flags.
22671 (ix86_avoid_lea_for_addr): Likewise. Adjust function comment.
22672 * config/i386/i386.md (*lea<mode>): Change from define_insn_and_split
22673 into define_insn. Move the splitting to define_peephole2 and
22674 check there using peep2_regno_dead_p if FLAGS_REG is dead.
22676 2021-02-17 Julian Brown <julian@codesourcery.com>
22678 * gimplify.c (gimplify_scan_omp_clauses): Handle ATTACH_DETACH
22681 2021-02-17 Xi Ruoyao <xry111@mengyan1223.wang>
22684 * config/mips/mips.c (mips_symbol_insns): Do not use
22685 MSA_SUPPORTED_MODE_P if mode is MAX_MACHINE_MODE.
22687 2021-02-16 Vladimir N. Makarov <vmakarov@redhat.com>
22689 PR inline-asm/98096
22690 * stmt.c (resolve_operand_name_1): Take inout operands into account
22691 for access to labels by names.
22692 * doc/extend.texi: Describe counting operands for accessing labels.
22694 2021-02-16 Richard Biener <rguenther@suse.de>
22696 PR tree-optimization/38474
22697 * tree-ssa-structalias.c (variable_info::address_taken): New.
22698 (new_var_info): Initialize address_taken.
22699 (process_constraint): Set address_taken.
22700 (solve_constraints): Use the new address_taken flag rather
22701 than is_reg_var for sorting variables.
22702 (dump_constraint): Dump the variable number if the name
22705 2021-02-16 Jakub Jelinek <jakub@redhat.com>
22708 * tree-vect-stmts.c (vectorizable_simd_clone_call): For num_calls != 1
22709 multiply by 4096 and for inbranch by 8192.
22710 * config/i386/i386.c (ix86_simd_clone_usable): For TARGET_AVX512F,
22711 return 3, 2 or 1 for mangle letters 'b', 'c' or 'd'.
22713 2021-02-15 Maya Rashish <coypu@sdf.org>
22715 * config/aarch64/aarch64.c (aarch64_init_builtins):
22716 Call SUBTARGET_INIT_BUILTINS.
22718 2021-02-15 Peter Bergner <bergner@linux.ibm.com>
22720 PR rtl-optimization/98872
22721 * init-regs.c (initialize_uninitialized_regs): Skip initialization
22722 if CONST0_RTX is NULL.
22724 2021-02-15 Richard Sandiford <richard.sandiford@arm.com>
22726 PR rtl-optimization/98863
22727 * rtl-ssa/functions.h (function_info::bb_live_out_info): Delete.
22728 (function_info::build_info): Turn into a declaration, moving the
22729 definition to internals.h.
22730 (function_info::bb_walker): Declare.
22731 (function_info::create_reg_use): Likewise.
22732 (function_info::calculate_potential_phi_regs): Take a build_info
22734 (function_info::place_phis, function_info::create_ebbs): Declare.
22735 (function_info::calculate_ebb_live_in_for_debug): Likewise.
22736 (function_info::populate_backedge_phis): Delete.
22737 (function_info::start_block, function_info::end_block): Declare.
22738 (function_info::populate_phi_inputs): Delete.
22739 (function_info::m_potential_phi_regs): Move information to build_info.
22740 * rtl-ssa/internals.h: New file.
22741 (function_info::bb_phi_info): New class.
22742 (function_info::build_info): Moved from functions.h.
22743 Add a constructor and destructor.
22744 (function_info::build_info::ebb_use): Delete.
22745 (function_info::build_info::ebb_def): Likewise.
22746 (function_info::build_info::bb_live_out): Likewise.
22747 (function_info::build_info::tmp_ebb_live_in_for_debug): New variable.
22748 (function_info::build_info::potential_phi_regs): Likewise.
22749 (function_info::build_info::potential_phi_regs_for_debug): Likewise.
22750 (function_info::build_info::ebb_def_regs): Likewise.
22751 (function_info::build_info::bb_phis): Likewise.
22752 (function_info::build_info::bb_mem_live_out): Likewise.
22753 (function_info::build_info::bb_to_rpo): Likewise.
22754 (function_info::build_info::def_stack): Likewise.
22755 (function_info::build_info::old_def_stack_limit): Likewise.
22756 * rtl-ssa/internals.inl (function_info::build_info::record_reg_def):
22757 Remove the regno argument. Push the previous definition onto the
22758 definition stack where necessary.
22759 * rtl-ssa/accesses.cc: Include internals.h.
22760 * rtl-ssa/changes.cc: Likewise.
22761 * rtl-ssa/blocks.cc: Likewise.
22762 (function_info::build_info::build_info): Define.
22763 (function_info::build_info::~build_info): Likewise.
22764 (function_info::bb_walker): New class.
22765 (function_info::bb_walker::bb_walker): Define.
22766 (function_info::add_live_out_use): Convert a logarithmic-complexity
22767 test into a linear one. Allow the same definition to be passed
22769 (function_info::calculate_potential_phi_regs): Moved from
22770 functions.cc. Take a build_info parameter and store the
22771 information there instead.
22772 (function_info::place_phis): New function.
22773 (function_info::add_entry_block_defs): Update call to record_reg_def.
22774 (function_info::calculate_ebb_live_in_for_debug): New function.
22775 (function_info::add_phi_nodes): Use bb_phis to decide which
22776 registers need phi nodes and initialize ebb_def_regs accordingly.
22777 Do not add degenerate phis here.
22778 (function_info::add_artificial_accesses): Use create_reg_use.
22779 Assert that all definitions are listed in the DF LR sets.
22780 Update call to record_reg_def.
22781 (function_info::record_block_live_out): Record live-out register
22782 values in the phis of successor blocks. Use the live-out set
22783 when processing the last block in an EBB, instead of always
22784 using the live-in sets of successor blocks. AND the live sets
22785 with the set of registers that have been defined in the EBB,
22786 rather than with all potential phi registers. Cope correctly
22787 with branches back to the start of the current EBB.
22788 (function_info::start_block): New function.
22789 (function_info::end_block): Likewise.
22790 (function_info::populate_phi_inputs): Likewise.
22791 (function_info::create_ebbs): Likewise.
22792 (function_info::process_all_blocks): Rewrite into a multi-phase
22794 * rtl-ssa/functions.cc: Include internals.h.
22795 (function_info::calculate_potential_phi_regs): Move to blocks.cc.
22796 (function_info::init_function_data): Remove caller.
22797 * rtl-ssa/insns.cc: Include internals.h
22798 (function_info::create_reg_use): New function. Lazily any
22799 degenerate phis needed by the linear RPO view.
22800 (function_info::record_use): Use create_reg_use. When processing
22801 debug uses, use potential_phi_regs and test it before checking
22802 whether the register is live on entry to the current EBB. Lazily
22803 calculate ebb_live_in_for_debug.
22804 (function_info::record_call_clobbers): Update call to record_reg_def.
22805 (function_info::record_def): Likewise.
22807 2021-02-15 Martin Liska <mliska@suse.cz>
22809 * toplev.c (init_asm_output): Free output of
22810 gen_command_line_string function.
22811 (process_options): Likewise.
22813 2021-02-15 Martin Liska <mliska@suse.cz>
22815 * params.opt: Add 2 missing Param keywords.
22817 2021-02-15 Eric Botcazou <ebotcazou@adacore.com>
22819 * df-core.c (df_worklist_dataflow_doublequeue): Use proper cast.
22821 2021-02-15 Jakub Jelinek <jakub@redhat.com>
22823 PR tree-optimization/99079
22824 * match.pd (A % (pow2pcst << N) -> A & ((pow2pcst << N) - 1)): Remove
22825 useless tree_nop_conversion_p (type, TREE_TYPE (@3)) check. Instead
22826 require both type and TREE_TYPE (@1) to be integral types and either
22827 type having smaller or equal precision, or TREE_TYPE (@1) being
22828 unsigned type, or type being signed type. If TREE_TYPE (@1)
22829 doesn't have wrapping overflow, perform the subtraction of one in
22832 2021-02-14 Jan Hubicka <hubicka@ucw.cz>
22833 Richard Biener <rguether@suse.de>
22836 * ipa-reference.c (ipa_init): Only conditinally initialize
22837 reference_vars_to_consider.
22838 (propagate): Conditionally deninitialize reference_vars_to_consider.
22839 (ipa_reference_write_optimization_summary): Sanity check that
22840 reference_vars_to_consider is not allocated.
22842 2021-02-13 Levy Hsu <admin@levyhsu.com>
22845 * config/riscv/riscv-shorten-memrefs.c (pass_shorten_memrefs): Add
22846 extend parameter to get_si_mem_base_reg declaration.
22847 (get_si_mem_base_reg): Add extend parameter. Set it.
22848 (analyze): Pass extend arg to get_si_mem_base_reg.
22849 (transform): Likewise. Use it when rewriting mems.
22850 * config/riscv/riscv.c (riscv_legitimize_move): Check for subword
22851 loads and emit sign/zero extending load followed by subreg move.
22853 2021-02-13 Jim Wilson <jimw@sifive.com>
22856 * config/riscv/riscv.c (riscv_compressed_lw_address_p): Drop early
22857 exit when !reload_completed. Only perform check for compressed reg
22858 if reload_completed.
22859 (riscv_rtx_costs): In MEM case, when optimizing for size and
22860 shorten memrefs, if not compressible, then increase cost.
22862 2021-02-13 Jakub Jelinek <jakub@redhat.com>
22864 PR rtl-optimization/98439
22865 * recog.c (pass_split_before_regstack::gate): Enable even when
22866 pass_split_before_sched2 is enabled if -fselective-scheduling2 is
22869 2021-02-13 Jakub Jelinek <jakub@redhat.com>
22872 * config/i386/mmx.md (*mmx_pshufd_1): Add a combine splitter for
22873 swap of V2SImode elements in memory into DImode memory rotate by 32.
22875 2021-02-12 Martin Sebor <msebor@redhat.com>
22877 * tree-pretty-print.c (print_generic_expr_to_str): Update comment.
22879 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
22881 * rtl-ssa/accesses.cc (function_info::make_use_available): Use
22882 m_temp_obstack rather than m_obstack to allocate the temporary use.
22884 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
22886 * df-problems.c (df_lr_bb_local_compute): Treat partial definitions
22887 as read-modify operations.
22889 2021-02-12 Richard Biener <rguenther@suse.de>
22891 PR middle-end/38474
22892 * ipa-fnsummary.c (unmodified_parm_1): Only walk when
22893 fbi->aa_walk_budget is bigger than zero. Update
22894 fbi->aa_walk_budget.
22895 (param_change_prob): Likewise.
22896 * ipa-prop.c (detect_type_change_from_memory_writes):
22897 Properly account walk_aliased_vdefs.
22898 (parm_preserved_before_stmt_p): Canonicalize updates.
22899 (parm_ref_data_preserved_p): Likewise.
22900 (parm_ref_data_pass_through_p): Likewise.
22901 (determine_known_aggregate_parts): Account own alias queries.
22903 2021-02-12 Martin Liska <mliska@suse.cz>
22905 * opts-common.c (decode_cmdline_option): Release werror_arg.
22906 * opts.c (gen_producer_string): Release output of
22907 gen_command_line_string.
22909 2021-02-12 Richard Biener <rguenther@suse.de>
22911 PR tree-optimization/38474
22912 * params.opt (-param=max-store-chains-to-track=): New param.
22913 (-param=max-stores-to-track=): Likewise.
22914 * doc/invoke.texi (max-store-chains-to-track): Document.
22915 (max-stores-to-track): Likewise.
22916 * gimple-ssa-store-merging.c (pass_store_merging::m_n_chains):
22918 (pass_store_merging::m_n_stores): Likewise.
22919 (pass_store_merging::terminate_and_process_chain): Update
22920 m_n_stores and m_n_chains.
22921 (pass_store_merging::process_store): Likewise. Terminate
22922 oldest chains if the number of stores or chains get too large.
22923 (imm_store_chain_info::terminate_and_process_chain): Dump
22926 2021-02-11 Eric Botcazou <ebotcazou@adacore.com>
22928 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
22929 the cold section, emit a nop before the directive if the previous
22930 active instruction can throw.
22932 2021-02-11 Peter Bergner <bergner@linux.ibm.com>
22935 * config/rs6000/predicates.md (mma_assemble_input_operand): Restrict
22936 memory addresses that are legal for quad word accesses.
22938 2021-02-11 Andrea Corallo <andrea.corallo@arm.com>
22941 * config/arm/thumb2.md (*doloop_end_internal): Generate
22942 alternative sequence to handle long range branches.
22944 2021-02-11 Joel Hutton <joel.hutton@arm.com>
22946 PR tree-optimization/98772
22947 * optabs-tree.c (supportable_half_widening_operation): New function
22948 to check for supportable V8QI->V8HI widening patterns.
22949 * optabs-tree.h (supportable_half_widening_operation): New function.
22950 * tree-vect-stmts.c (vect_create_half_widening_stmts): New function
22951 to create promotion stmts for V8QI->V8HI widening patterns.
22952 (vectorizable_conversion): Add case for V8QI->V8HI.
22954 2021-02-11 Richard Biener <rguenther@suse.de>
22956 * sparseset.h (SPARSESET_ELT_BITS): Remove.
22957 (SPARSESET_ELT_TYPE): Use unsigned int.
22958 * fwprop.c: Do not include sparseset.h.
22960 2021-02-10 Jakub Jelinek <jakub@redhat.com>
22963 * varasm.c (declare_weak): For -fsyntax-only, allow even
22964 TREE_ASM_WRITTEN function decls.
22966 2021-02-10 Jakub Jelinek <jakub@redhat.com>
22969 * config/i386/sse.md (fix<fixunssuffix>_truncv2sfv2di2,
22970 <insn>v8qiv8hi2, <insn>v8qiv8si2, <insn>v4qiv4si2, <insn>v4hiv4si2,
22971 <insn>v8qiv8di2, <insn>v4qiv4di2, <insn>v2qiv2di2, <insn>v4hiv4di2,
22972 <insn>v2hiv2di2, <insn>v2siv2di2): Force operands[1] into REG before
22973 calling simplify_gen_subreg on it.
22975 2021-02-10 Martin Liska <mliska@suse.cz>
22977 * config/nvptx/nvptx.c (nvptx_option_override): Use
22978 flag_patchable_function_entry instead of the removed
22979 function_entry_patch_area_size.
22981 2021-02-10 Martin Liska <mliska@suse.cz>
22983 PR tree-optimization/99002
22984 PR tree-optimization/99026
22985 * gimple-if-to-switch.cc (if_chain::is_beneficial): Fix memory
22986 leak when adjacent cases are merged.
22987 * tree-switch-conversion.c (switch_decision_tree::analyze_switch_statement): Use
22989 (make_pass_lower_switch): Remove trailing whitespace.
22990 * tree-switch-conversion.h (release_clusters): New.
22992 2021-02-10 Richard Biener <rguenther@suse.de>
22994 PR rtl-optimization/99054
22995 * cfgrtl.c (rtl-optimization/99054): Return an auto_vec.
22996 (fixup_partitions): Adjust.
22997 (rtl_verify_edges): Likewise.
22999 2021-02-10 Jakub Jelinek <jakub@redhat.com>
23001 PR middle-end/99007
23002 * gimplify.c (gimplify_scan_omp_clauses): For MEM_REF on reductions,
23003 temporarily disable gimplify_ctxp->into_ssa around gimplify_expr
23006 2021-02-10 Richard Biener <rguenther@suse.de>
23009 * ipa-pure-const.c (propagate_malloc): Use an auto_vec<>
23012 2021-02-10 Richard Biener <rguenther@suse.de>
23014 PR tree-optimization/99024
23015 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Only
23016 clear loop->aux if it is associated with the destroyed loop_vinfo.
23018 2021-02-10 Martin Liska <mliska@suse.cz>
23020 PR tree-optimization/99002
23021 * gimple-if-to-switch.cc (find_conditions): Fix memory leak
23024 2021-02-10 Martin Liska <mliska@suse.cz>
23027 * ipa-icf.c (sem_item::add_reference): Fix memory leak when
23028 a reference exists.
23030 2021-02-10 Jakub Jelinek <jakub@redhat.com>
23033 * dwarf2out.c (prune_unused_types_walk): Mark DW_TAG_variable DIEs
23034 at class scope for DWARF5+.
23036 2021-02-09 Eric Botcazou <ebotcazou@adacore.com>
23038 PR rtl-optimization/96015
23039 * reorg.c (skip_consecutive_labels): Minor comment tweaks.
23040 (relax_delay_slots): When deleting a jump to the next active
23041 instruction over a barrier, first delete the barrier if the
23042 jump is the only way to reach the target label.
23044 2021-02-09 Andre Vieira <andre.simoesdiasvieira@arm.com>
23046 * config/aarch64/aarch64-cost-tables.h: Add entries for vect.mul.
23047 * config/aarch64/aarch64.c (aarch64_rtx_mult_cost): Use vect.mul for
23048 vector multiplies and vect.alu for SSRA.
23049 * config/arm/aarch-common-protos.h (struct vector_cost_table): Define
23050 vect.mul cost field.
23051 * config/arm/aarch-cost-tables.h: Add entries for vect.mul.
23052 * config/arm/arm.c: Likewise.
23054 2021-02-09 Richard Biener <rguenther@suse.de>
23056 PR tree-optimization/98863
23057 * tree-ssa-sccvn.h (vn_avail::next_undo): Add.
23058 * tree-ssa-sccvn.c (last_pushed_avail): New global.
23059 (rpo_elim::eliminate_push_avail): Chain pushed avails.
23060 (unwind_state::avail_top): Add.
23061 (do_unwind): Rewrite unwinding of avail entries.
23062 (do_rpo_vn): Initialize last_pushed_avail and
23063 avail_top of the undo state.
23065 2021-02-09 Jakub Jelinek <jakub@redhat.com>
23067 PR middle-end/99004
23068 * calls.c (maybe_warn_rdwr_sizes): Change s0 and s1 type from
23069 const char * to char * and free those pointers after use.
23071 2021-02-09 Richard Biener <rguenther@suse.de>
23073 PR tree-optimization/99017
23074 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Allow
23075 zero vector cost entries.
23077 2021-02-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
23079 PR middle-end/98974
23080 * tree-vect-stmts.c (vectorizable_condition): Remove shadow vec_num
23081 parameter in vectorizable_condition.
23083 2021-02-08 Richard Biener <rguenther@suse.de>
23086 * tree.c (walk_tree_1): Walk VECTOR_CST elements.
23088 2021-02-08 Martin Liska <mliska@suse.cz>
23091 * cfgexpand.c (pass_expand::execute): Parse per-function option
23092 flag_patchable_function_entry and use it.
23093 * common.opt: Remove function_entry_patch_area_size and
23094 function_entry_patch_area_start global variables.
23095 * opts.c (parse_and_check_patch_area): New function.
23096 (common_handle_option): Use it.
23097 * opts.h (parse_and_check_patch_area): New function.
23098 * toplev.c (process_options): Parse and use
23099 function_entry_patch_area_size.
23101 2021-02-08 Martin Sebor <msebor@redhat.com>
23103 * doc/extend.texi (attribute malloc): Correct typos.
23105 2021-02-05 Nathan Sidwell <nathan@acm.org>
23108 * gcc.c (driver::maybe_run_linker): Check for input file
23109 accessibility if not linking.
23111 2021-02-05 Richard Biener <rguenther@suse.de>
23113 PR tree-optimization/98855
23114 * tree-vectorizer.h (add_stmt_cost): New overload.
23115 * tree-vect-slp.c (li_cost_vec_cmp): New.
23116 (vect_bb_slp_scalar_cost): Cost individual loop regions
23117 separately. Account for the scalar instance root stmt.
23119 2021-02-05 Tom de Vries <tdevries@suse.de>
23122 * tree-switch-conversion.c (jump_table_cluster::emit): Add loc
23124 (bit_test_cluster::emit): Reuse location_t for newly created
23126 (switch_decision_tree::try_switch_expansion): Preserve
23128 * tree-switch-conversion.h: Change function signatures.
23130 2021-02-05 Jakub Jelinek <jakub@redhat.com>
23133 * config/i386/i386-options.c (m_NONE, m_ALL): Define.
23134 * config/i386/x86-tune.def (X86_TUNE_BRANCH_PREDICTION_HINTS,
23135 X86_TUNE_PROMOTE_QI_REGS): Use m_NONE instead of 0U.
23136 (X86_TUNE_QIMODE_MATH): Use m_ALL instead of ~0U.
23138 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23140 * config/aarch64/aarch64-simd-builtins.def (get_high): Define builtin.
23141 * config/aarch64/aarch64-simd.md (aarch64_get_high<mode>): Define.
23142 * config/aarch64/arm_neon.h (__GET_HIGH): Delete.
23143 (vget_high_f16): Reimplement using new builtin.
23144 (vget_high_f32): Likewise.
23145 (vget_high_f64): Likewise.
23146 (vget_high_p8): Likewise.
23147 (vget_high_p16): Likewise.
23148 (vget_high_p64): Likewise.
23149 (vget_high_s8): Likewise.
23150 (vget_high_s16): Likewise.
23151 (vget_high_s32): Likewise.
23152 (vget_high_s64): Likewise.
23153 (vget_high_u8): Likewise.
23154 (vget_high_u16): Likewise.
23155 (vget_high_u32): Likewise.
23156 (vget_high_u64): Likewise.
23158 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23160 * config/aarch64/aarch64-simd-builtins.def (get_low): Define builtin.
23161 * config/aarch64/aarch64-simd.md (aarch64_get_low<mode>): Define.
23162 * config/aarch64/arm_neon.h (__GET_LOW): Delete.
23163 (vget_low_f16): Reimplement using new builtin.
23164 (vget_low_f32): Likewise.
23165 (vget_low_f64): Likewise.
23166 (vget_low_p8): Likewise.
23167 (vget_low_p16): Likewise.
23168 (vget_low_p64): Likewise.
23169 (vget_low_s8): Likewise.
23170 (vget_low_s16): Likewise.
23171 (vget_low_s32): Likewise.
23172 (vget_low_s64): Likewise.
23173 (vget_low_u8): Likewise.
23174 (vget_low_u16): Likewise.
23175 (vget_low_u32): Likewise.
23176 (vget_low_u64): Likewise.
23178 2021-02-05 Kito Cheng <kito.cheng@sifive.com>
23180 * gcc.c (print_multilib_info): Check all required argument is provided
23183 2021-02-05 liuhongt <hongtao.liu@intel.com>
23186 * config/i386/i386-expand.c (ix86_expand_sse_cmp): Don't
23187 generate integer mask comparison for 128/256-bits vector when
23188 op_true/op_false is NULL_RTX or CONSTM1_RTX/CONST0_RTX. Also
23189 delete redundant !maskcmp condition.
23190 (ix86_expand_int_vec_cmp): Ditto but no redundant deletion
23192 (ix86_expand_sse_movcc): Delete definition of maskcmp, add the
23193 condition directly to if (maskcmp), add extra check for
23194 cmpmode, it should be MODE_INT.
23195 (ix86_expand_fp_vec_cmp): Pass NULL to ix86_expand_sse_cmp's
23196 parameters op_true/op_false.
23197 (ix86_use_mask_cmp_p): New.
23199 2021-02-05 liuhongt <hongtao.liu@intel.com>
23202 * config/i386/x86-tune.def (X86_TUNE_AVX256_UNALIGNED_LOAD_OPTIMAL):
23203 Remove m_GENERIC from ~list.
23204 (X86_TUNE_AVX256_UNALIGNED_STORE_OPTIMAL): Ditto.
23206 2021-02-04 David Malcolm <dmalcolm@redhat.com>
23209 * diagnostic-show-locus.c (compatible_locations_p): Require
23210 locations in the same macro map to be either both from the
23211 macro definition, or both from the macro arguments.
23213 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
23215 * config/aarch64/aarch64-simd-builtins.def: Add
23216 [su]mull_hi_lane[q] builtin generator macros.
23217 * config/aarch64/aarch64-simd.md
23218 (aarch64_<su>mull_hi_lane<mode>_insn): Define.
23219 (aarch64_<su>mull_hi_lane<mode>): Define.
23220 (aarch64_<su>mull_hi_laneq<mode>_insn): Define.
23221 (aarch64_<su>mull_hi_laneq<mode>): Define.
23222 * config/aarch64/arm_neon.h (vmull_high_lane_s16): Use RTL
23223 builtin instead of inline asm.
23224 (vmull_high_lane_s32): Likewise.
23225 (vmull_high_lane_u16): Likewise.
23226 (vmull_high_lane_u32): Likewise.
23227 (vmull_high_laneq_s16): Likewise.
23228 (vmull_high_laneq_s32): Likewise.
23229 (vmull_high_laneq_u16): Likewise.
23230 (vmull_high_laneq_u32): Liekwise.
23232 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
23234 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_hi_n
23235 builtin generator macros.
23236 * config/aarch64/aarch64-simd.md
23237 (aarch64_<su>mull_hi_n<mode>_insn): Define.
23238 (aarch64_<su>mull_hi_n<mode>): Define.
23239 * config/aarch64/arm_neon.h (vmull_high_n_s16): Use RTL builtin
23240 instead of inline asm.
23241 (vmull_high_n_s32): Likewise.
23242 (vmull_high_n_u16): Likewise.
23243 (vmull_high_n_u32): Likewise.
23245 2021-02-04 Richard Biener <rguenther@suse.de>
23247 PR tree-optimization/98855
23248 * tree-vect-loop.c (vectorizable_phi): Do not cost
23249 single-argument PHIs.
23250 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Likewise.
23251 * tree-vect-stmts.c (vectorizable_bswap): Also perform
23252 costing for SLP operation.
23254 2021-02-04 Martin Liska <mliska@suse.cz>
23256 * doc/extend.texi: Mention -mprefer-vector-width in target
23259 2021-02-03 Martin Sebor <msebor@redhat.com>
23261 PR tree-optimization/98937
23262 * tree-ssa-strlen.c (strlen_dom_walker::~strlen_dom_walker): Define.
23263 Flush pointer_query cache.
23265 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
23267 * config/rs6000/genfusion.pl (gen_2logical): Add missing
23268 fixes based on patch review.
23269 * config/rs6000/fusion.md: Regenerate file.
23271 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
23273 * config/rs6000/t-rs6000: Comment out auto generation of
23276 2021-02-03 Andrew Stubbs <ams@codesourcery.com>
23278 * config/gcn/gcn-opts.h (enum processor_type): Add PROCESSOR_GFX908.
23279 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Add gfx908.
23280 (output_file_start): Add gfx908.
23281 * config/gcn/gcn.opt (gpu_type): Add gfx908.
23282 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Add march=gfx908.
23283 (MULTILIB_DIRNAMES): Add gfx908.
23284 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): New define.
23285 (main): Recognize gfx908.
23286 * config/gcn/t-omp-device: Add gfx908.
23288 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
23290 * config/aarch64/aarch64-simd-builtins.def: Add
23291 [su]mlsl_hi_lane[q] builtin macro generators.
23292 * config/aarch64/aarch64-simd.md
23293 (aarch64_<su>mlsl_hi_lane<mode>_insn): Define.
23294 (aarch64_<su>mlsl_hi_lane<mode>): Define.
23295 (aarch64_<su>mlsl_hi_laneq<mode>_insn): Define.
23296 (aarch64_<su>mlsl_hi_laneq<mode>): Define.
23297 * config/aarch64/arm_neon.h (vmlsl_high_lane_s16): Use RTL
23298 builtin instead of inline asm.
23299 (vmlsl_high_lane_s32): Likewise.
23300 (vmlsl_high_lane_u16): Likewise.
23301 (vmlsl_high_lane_u32): Likewise.
23302 (vmlsl_high_laneq_s16): Likewise.
23303 (vmlsl_high_laneq_s32): Likewise.
23304 (vmlsl_high_laneq_u16): Likewise.
23305 (vmlsl_high_laneq_u32): Likewise.
23306 (vmlal_high_laneq_u32): Likewise.
23308 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
23310 * config/aarch64/aarch64-simd-builtins.def: Add
23311 [su]mlal_hi_lane[q] builtin generator macros.
23312 * config/aarch64/aarch64-simd.md
23313 (aarch64_<su>mlal_hi_lane<mode>_insn): Define.
23314 (aarch64_<su>mlal_hi_lane<mode>): Define.
23315 (aarch64_<su>mlal_hi_laneq<mode>_insn): Define.
23316 (aarch64_<su>mlal_hi_laneq<mode>): Define.
23317 * config/aarch64/arm_neon.h (vmlal_high_lane_s16): Use RTL
23318 builtin instead of inline asm.
23319 (vmlal_high_lane_s32): Likewise.
23320 (vmlal_high_lane_u16): Likewise.
23321 (vmlal_high_lane_u32): Likewise.
23322 (vmlal_high_laneq_s16): Likewise.
23323 (vmlal_high_laneq_s32): Likewise.
23324 (vmlal_high_laneq_u16): Likewise.
23325 (vmlal_high_laneq_u32): Likewise.
23327 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
23329 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_hi_n
23330 builtin generator macros.
23331 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_hi_n<mode>_insn):
23333 (aarch64_<su>mlsl_hi_n<mode>): Define.
23334 * config/aarch64/arm_neon.h (vmlsl_high_n_s16): Use RTL builtin
23335 instead of inline asm.
23336 (vmlsl_high_n_s32): Likewise.
23337 (vmlsl_high_n_u16): Likewise.
23338 (vmlsl_high_n_u32): Likewise.
23340 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
23342 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_hi_n
23343 builtin generator macros.
23344 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_hi_n<mode>_insn):
23346 (aarch64_<su>mlal_hi_n<mode>): Define.
23347 * config/aarch64/arm_neon.h (vmlal_high_n_s16): Use RTL builtin
23348 instead of inline asm.
23349 (vmlal_high_n_s32): Likewise.
23350 (vmlal_high_n_u16): Likewise.
23351 (vmlal_high_n_u32): Likewise.
23353 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
23355 * config/aarch64/aarch64-simd-builtins.def: Add RTL builtin
23357 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal_hi<mode>):
23359 (aarch64_<su>mlal_hi<mode>_insn): This.
23360 (aarch64_<su>mlal_hi<mode>): Define.
23361 * config/aarch64/arm_neon.h (vmlal_high_s8): Use RTL builtin
23362 instead of inline asm.
23363 (vmlal_high_s16): Likewise.
23364 (vmlal_high_s32): Likewise.
23365 (vmlal_high_u8): Likewise.
23366 (vmlal_high_u16): Likewise.
23367 (vmlal_high_u32): Likewise.
23369 2021-02-03 Ilya Leoshkevich <iii@linux.ibm.com>
23371 * lra-spills.c (remove_pseudos): Call lra_update_insn_recog_data()
23372 after calling alter_subreg() on a (mem).
23374 2021-02-03 Martin Liska <mliska@suse.cz>
23377 * lto-streamer-out.c (produce_lto_section): Fill up missing
23379 * lto-streamer.h (struct lto_section): Add _padding field.
23381 2021-02-03 Richard Biener <rguenther@suse.de>
23383 * lto-streamer.c (lto_get_section_name): Free temporary
23385 * tree-loop-distribution.c
23386 (loop_distribution::merge_dep_scc_partitions): Free edge data.
23388 2021-02-03 Jakub Jelinek <jakub@redhat.com>
23390 PR middle-end/97487
23391 * ifcvt.c (noce_can_force_operand): New function.
23392 (noce_emit_move_insn): Use it.
23393 (noce_try_sign_mask): Likewise. Formatting fix.
23395 2021-02-03 Jakub Jelinek <jakub@redhat.com>
23397 PR middle-end/97971
23398 * lra-constraints.c (process_alt_operands): For inline asm, don't call
23399 fatal_insn, but instead return false.
23401 2021-02-03 Jakub Jelinek <jakub@redhat.com>
23403 PR tree-optimization/98287
23404 * config/i386/mmx.md (<insn><mode>3): For shifts don't enable expander
23407 2021-02-03 Tamar Christina <tamar.christina@arm.com>
23409 PR tree-optimization/98928
23410 * tree-vect-loop.c (vect_analyze_loop_2): Change
23411 STMT_VINFO_SLP_VECT_ONLY to STMT_VINFO_SLP_VECT_ONLY_PATTERN.
23412 * tree-vect-slp-patterns.c (complex_pattern::build): Likewise.
23413 * tree-vectorizer.h (STMT_VINFO_SLP_VECT_ONLY_PATTERN): New.
23414 (class _stmt_vec_info): Add slp_vect_pattern_only_p.
23416 2021-02-02 Richard Biener <rguenther@suse.de>
23418 * gimple-loop-interchange.cc (prepare_data_references):
23420 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
23421 * tree-ssa-loop-im.c (hoist_memory_references): Likewise.
23422 * tree-vect-stmts.c (vectorizable_condition): Do not
23424 (vectorizable_comparison): Likewise.
23426 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23428 * config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
23429 * config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
23430 * config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
23431 (vrsqrteq_u32): Likewise.
23433 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23435 * config/aarch64/aarch64-simd-builtins.def (sqxtun2): Define builtin.
23436 * config/aarch64/aarch64-simd.md (aarch64_sqxtun2<mode>_le): Define.
23437 (aarch64_sqxtun2<mode>_be): Likewise.
23438 (aarch64_sqxtun2<mode>): Likewise.
23439 * config/aarch64/arm_neon.h (vqmovun_high_s16): Reimplement using builtin.
23440 (vqmovun_high_s32): Likewise.
23441 (vqmovun_high_s64): Likewise.
23442 * config/aarch64/iterators.md (UNSPEC_SQXTUN2): Define.
23444 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23446 * config/aarch64/aarch64-simd-builtins.def (bfdot_lane, bfdot_laneq): Use
23448 (bfmlalb_lane, bfmlalt_lane, bfmlalb_lane_q, bfmlalt_lane_q): Use FP flags.
23450 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23452 * config/aarch64/aarch64-simd-builtins.def (fcmla_lane0, fcmla_lane90,
23453 fcmla_lane180, fcmla_lane270, fcmlaq_lane0, fcmlaq_lane90, fcmlaq_lane180,
23454 fcmlaq_lane270, scvtf, ucvtf, fcvtzs, fcvtzu, scvtfsi, scvtfdi, ucvtfsi,
23455 ucvtfdi, fcvtzshf, fcvtzuhf, fmlal_lane_low, fmlsl_lane_low,
23456 fmlal_laneq_low, fmlsl_laneq_low, fmlalq_lane_low, fmlslq_lane_low,
23457 fmlalq_laneq_low, fmlslq_laneq_low, fmlal_lane_high, fmlsl_lane_high,
23458 fmlal_laneq_high, fmlsl_laneq_high, fmlalq_lane_high, fmlslq_lane_high,
23459 fmlalq_laneq_high, fmlslq_laneq_high): Use FP flags.
23461 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23463 * config/aarch64/aarch64-builtins.c (FLAG_LOAD): Define.
23464 * config/aarch64/aarch64-simd-builtins.def (ld1x2, ld2, ld3, ld4, ld2r,
23465 ld3r, ld4r, ld1, ld1x3, ld1x4): Use LOAD flags.
23467 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23469 * config/aarch64/aarch64-simd-builtins.def (combine, zip1, zip2,
23470 uzp1, uzp2, trn1, trn2, simd_bsl): Use AUTO_FP flags.
23472 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23474 * config/aarch64/aarch64-simd-builtins.def (clrsb, clz, ctz, popcount,
23475 vec_smult_lane_, vec_smlal_lane_, vec_smult_laneq_, vec_smlal_laneq_,
23476 vec_umult_lane_, vec_umlal_lane_, vec_umult_laneq_, vec_umlal_laneq_,
23477 ashl, sshl, ushl, srshl, urshl, sdot_lane, udot_lane, sdot_laneq,
23478 udot_laneq, usdot_lane, usdot_laneq, sudot_lane, sudot_laneq, ashr,
23479 ashr_simd, lshr, lshr_simd, srshr_n, urshr_n, ssra_n, usra_n, srsra_n,
23480 ursra_n, sshll_n, ushll_n, sshll2_n, ushll2_n, ssri_n, usri_n, ssli_n,
23481 ssli_n, usli_n, bswap, rbit, simd_bsl, eor3q, rax1q, xarq, bcaxq): Use
23482 NONE builtin flags.
23484 2021-02-02 Jakub Jelinek <jakub@redhat.com>
23486 PR tree-optimization/98848
23487 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
23488 STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
23490 2021-02-02 Kito Cheng <kito.cheng@sifive.com>
23493 * expr.c: Check mode before calling store_expr.
23495 2021-02-02 Christophe Lyon <christophe.lyon@linaro.org>
23497 * config/arm/iterators.md (supf): Remove VORNQ_S and VORNQ_U.
23499 * config/arm/mve.md (mve_vornq_s<mode>): New entry for vorn
23500 instruction using expression ior.
23501 (mve_vornq_u<mode>): New expander.
23502 (mve_vornq_f<mode>): Use ior code instead of unspec.
23503 * config/arm/unspecs.md (VORNQ_S, VORNQ_U, VORNQ_F): Remove.
23505 2021-02-02 Alexandre Oliva <oliva@adacore.com>
23507 * tree-nested.c (convert_nonlocal_reference_op): Move
23508 current_function_decl restore after re-gimplification.
23509 (convert_local_reference_op): Likewise.
23511 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23513 * config/aarch64/aarch64-simd-builtins.def (rshrn, rshrn2):
23515 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>_insn_le):
23517 (aarch64_rshrn<mode>_insn_be): Likewise.
23518 (aarch64_rshrn<mode>): Likewise.
23519 (aarch64_rshrn2<mode>_insn_le): Likewise.
23520 (aarch64_rshrn2<mode>_insn_be): Likewise.
23521 (aarch64_rshrn2<mode>): Likewise.
23522 * config/aarch64/aarch64.md (unspec): Add UNSPEC_RSHRN.
23523 * config/aarch64/arm_neon.h (vrshrn_high_n_s16): Reimplement
23525 (vrshrn_high_n_s32): Likewise.
23526 (vrshrn_high_n_s64): Likewise.
23527 (vrshrn_high_n_u16): Likewise.
23528 (vrshrn_high_n_u32): Likewise.
23529 (vrshrn_high_n_u64): Likewise.
23530 (vrshrn_n_s16): Likewise.
23531 (vrshrn_n_s32): Likewise.
23532 (vrshrn_n_s64): Likewise.
23533 (vrshrn_n_u16): Likewise.
23534 (vrshrn_n_u32): Likewise.
23535 (vrshrn_n_u64): Likewise.
23537 2021-02-01 Sergei Trofimovich <siarheit@google.com>
23539 PR tree-optimization/98499
23540 * ipa-modref.c (analyze_ssa_name_flags): treat RVO
23541 conservatively and assume all possible side-effects.
23543 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23545 * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi,
23546 vec_unpacku_hi_): Define builtins.
23547 * config/aarch64/arm_neon.h (vmovl_high_s8): Reimplement using
23549 (vmovl_high_s16): Likewise.
23550 (vmovl_high_s32): Likewise.
23551 (vmovl_high_u8): Likewise.
23552 (vmovl_high_u16): Likewise.
23553 (vmovl_high_u32): Likewise.
23555 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23557 * config/aarch64/aarch64-simd-builtins.def (sabdl, uabdl):
23559 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl<mode>): New
23561 * config/aarch64/aarch64.md (unspec): Define UNSPEC_SABDL,
23563 * config/aarch64/arm_neon.h (vabdl_s8): Reimplemet using
23565 (vabdl_s16): Likewise.
23566 (vabdl_s32): Likewise.
23567 (vabdl_u8): Likewise.
23568 (vabdl_u16): Likewise.
23569 (vabdl_u32): Likewise.
23570 * config/aarch64/iterators.md (ABDL): New int iterator.
23571 (sur): Handle UNSPEC_SABDL, UNSPEC_UABDL.
23573 2021-02-01 Martin Sebor <msebor@redhat.com>
23575 * tree.h (BLOCK_VARS): Add comment.
23576 (BLOCK_SUBBLOCKS): Same.
23577 (BLOCK_SUPERCONTEXT): Same.
23578 (BLOCK_ABSTRACT_ORIGIN): Same.
23579 (inlined_function_outer_scope_p): Same.
23581 2021-02-01 Martin Sebor <msebor@redhat.com>
23583 PR middle-end/97172
23584 * attribs.c (attr_access::free_lang_data): Define new function.
23585 * attribs.h (attr_access::free_lang_data): Declare new function.
23587 2021-02-01 Richard Biener <rguenther@suse.de>
23589 * vec.h (auto_vec::auto_vec): Add memory stat parameters
23591 * bitmap.h (auto_bitmap::auto_bitmap): Likewise.
23593 2021-02-01 Tamar Christina <tamar.christina@arm.com>
23595 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>,
23596 aarch64_<su>mlsl<mode>, aarch64_<su>mlsl_n<mode>): Flip mult operands.
23598 2021-02-01 Richard Biener <rguenther@suse.de>
23600 PR rtl-optimization/98863
23601 * config/i386/i386-features.c (convert_scalars_to_vector):
23602 Set DF_RD_PRUNE_DEAD_DEFS.
23604 2021-01-31 Eric Botcazou <ebotcazou@adacore.com>
23606 * system.h (SIZE_MAX): Define if not already defined.
23608 2021-01-30 Aaron Sawdey <acsawdey@linux.ibm.com>
23610 * config/rs6000/genfusion.pl (gen_2logical): New function to
23611 generate patterns for logical-logical fusion.
23612 * config/rs6000/fusion.md: Regenerated patterns.
23613 * config/rs6000/rs6000-cpus.def: Add
23614 OPTION_MASK_P10_FUSION_2LOGICAL.
23615 * config/rs6000/rs6000.c (rs6000_option_override_internal):
23616 Enable logical-logical fusion for p10.
23617 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2logical.
23619 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
23621 * config/rs6000/rs6000.opt: Add periods to new AIX options.
23623 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
23625 * config/rs6000/rs6000.opt (mabi=vec-extabi): New.
23626 (mabi=vec-default): New.
23627 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
23628 __EXTABI__ for AIX Vector extended ABI.
23629 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print AIX Vector
23631 (conditional_register_usage): If AIX vec_extabi enabled, vs20-vs31
23633 * doc/invoke.texi (PowerPC mabi): Add AIX vec-extabi and vec-default.
23635 2021-01-30 Jakub Jelinek <jakub@redhat.com>
23637 * config/i386/i386-features.c (remove_partial_avx_dependency): Clear
23638 DF_DEFER_INSN_RESCAN after calling df_process_deferred_rescans.
23640 2021-01-29 Vladimir N. Makarov <vmakarov@redhat.com>
23643 * lra-constraints.c (in_class_p): Don't narrow class only for REG
23646 2021-01-29 Will Schmidt <will_schmidt@vnet.ibm.com>
23648 * config/rs6000/rs6000-call.c (rs6000_expand_binup_builtin): Add
23649 clauses for CODE_FOR_vsx_xvcvuxddp_scale and
23650 CODE_FOR_vsx_xvcvsxddp_scale to the parameter checking code.
23652 2021-01-29 Andrew MacLeod <amacleod@redhat.com>
23654 PR tree-optimization/98866
23655 * gimple-range-gori.h (gori_compute:set_range_invariant): New.
23656 * gimple-range-gori.cc (gori_map::set_range_invariant): New.
23657 (gori_map::m_maybe_invariant): Rename from all_outgoing.
23658 (gori_map::gori_map): Rename all_outgoing to m_maybe_invariant.
23659 (gori_map::is_export_p): Ditto.
23660 (gori_map::calculate_gori): Ditto.
23661 (gori_compute::set_range_invariant): New.
23662 * gimple-range.cc (gimple_ranger::range_of_stmt): Set range
23663 invariant for pointers evaluating to [1, +INF].
23665 2021-01-29 Richard Biener <rguenther@suse.de>
23667 PR rtl-optimization/98863
23668 * config/i386/i386-features.c (remove_partial_avx_dependency):
23669 Do not perform DF analysis.
23670 (pass_data_remove_partial_avx_dependency): Remove
23673 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
23675 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_n
23676 builtin generator macros.
23677 * config/aarch64/aarch64-simd.md (aarch64_<su>mull_n<mode>):
23679 * config/aarch64/arm_neon.h (vmull_n_s16): Use RTL builtin
23680 instead of inline asm.
23681 (vmull_n_s32): Likewise.
23682 (vmull_n_u16): Likewise.
23683 (vmull_n_u32): Likewise.
23685 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23687 * config/aarch64/aarch64-simd-builtins.def (sabdl2, uabdl2):
23689 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
23691 (aarch64_<sur>abdl2<mode>): ... This.
23692 (<sur>sadv16qi): Adjust use of above.
23693 * config/aarch64/arm_neon.h (vabdl_high_s8): Reimplement using
23695 (vabdl_high_s16): Likewise.
23696 (vabdl_high_s32): Likewise.
23697 (vabdl_high_u8): Likewise.
23698 (vabdl_high_u16): Likewise.
23699 (vabdl_high_u32): Likewise.
23701 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23703 * config/aarch64/aarch64-simd-builtins.def (sabal2): Define
23705 (uabal2): Likewise.
23706 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal2<mode>): New
23708 * config/aarch64/aarch64.md (unspec): Add UNSPEC_SABAL2 and
23710 * config/aarch64/arm_neon.h (vabal_high_s8): Reimplement using
23712 (vabal_high_s16): Likewise.
23713 (vabal_high_s32): Likewise.
23714 (vabal_high_u8): Likewise.
23715 (vabal_high_u16): Likewise.
23716 (vabal_high_u32): Likewise.
23717 * config/aarch64/iterators.md (ABAL2): New mode iterator.
23718 (sur): Handle UNSPEC_SABAL2, UNSPEC_UABAL2.
23720 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23722 * config/aarch64/aarch64-simd-builtins.def (sabal): Define
23725 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal<mode>_4):
23727 (aarch64_<sur>abal<mode>): ... This
23728 (<sur>sadv16qi): Adust use of the above.
23729 * config/aarch64/arm_neon.h (vabal_s8): Reimplement using
23731 (vabal_s16): Likewise.
23732 (vabal_s32): Likewise.
23733 (vabal_u8): Likewise.
23734 (vabal_u16): Likewise.
23735 (vabal_u32): Likewise.
23737 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23739 * config/aarch64/aarch64-simd-builtins.def (saddlv, uaddlv):
23741 * config/aarch64/aarch64-simd.md (aarch64_<su>addlv<mode>):
23743 * config/aarch64/arm_neon.h (vaddlv_s8): Reimplement using
23745 (vaddlv_s16): Likewise.
23746 (vaddlv_u8): Likewise.
23747 (vaddlv_u16): Likewise.
23748 (vaddlvq_s8): Likewise.
23749 (vaddlvq_s16): Likewise.
23750 (vaddlvq_s32): Likewise.
23751 (vaddlvq_u8): Likewise.
23752 (vaddlvq_u16): Likewise.
23753 (vaddlvq_u32): Likewise.
23754 (vaddlv_s32): Likewise.
23755 (vaddlv_u32): Likewise.
23756 * config/aarch64/iterators.md (VDQV_L): New mode iterator.
23757 (unspec): Add UNSPEC_SADDLV, UNSPEC_UADDLV.
23758 (Vwstype): New mode attribute.
23760 (VWIDE_S): Likewise.
23761 (USADDLV): New int iterator.
23762 (su): Handle UNSPEC_SADDLV, UNSPEC_UADDLV.
23764 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
23766 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_lane[q]
23767 builtin generator macros.
23768 * config/aarch64/aarch64-simd.md (aarch64_vec_<su>mlsl_lane<Qlane>):
23770 * config/aarch64/arm_neon.h (vmlsl_lane_s16): Use RTL builtin
23771 instead of inline asm.
23772 (vmlsl_lane_s32): Likewise.
23773 (vmlsl_lane_u16): Likewise.
23774 (vmlsl_lane_u32): Likewise.
23775 (vmlsl_laneq_s16): Likewise.
23776 (vmlsl_laneq_s32): Likewise.
23777 (vmlsl_laneq_u16): Likewise.
23778 (vmlsl_laneq_u32): Likewise.
23780 2021-01-29 Richard Biener <rguenther@suse.de>
23782 * doc/invoke.texi (--param max-gcse-memory): Document unit
23784 * gcse.c (gcse_or_cprop_is_too_expensive): Adjust.
23785 * params.opt (--param max-gcse-memory): Adjust default and
23786 document unit of size.
23788 2021-01-29 Richard Biener <rguenther@suse.de>
23790 PR rtl-optimization/98863
23791 * gcse.c (gcse_or_cprop_is_too_expensive): Use unsigned
23792 HOST_WIDE_INT for the memory estimate.
23794 2021-01-29 Bin Cheng <bin.cheng@linux.alibaba.com>
23795 Richard Biener <rguenther@suse.de>
23797 PR tree-optimization/97627
23798 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
23799 Do not analyze fake edges.
23801 2021-01-29 Richard Biener <rguenther@suse.de>
23803 PR rtl-optimization/98144
23804 * df.h (df_mir_bb_info): Add con_visited member.
23805 * df-problems.c (df_mir_alloc): Initialize con_visited,
23806 do not fully populate IN and OUT.
23807 (df_mir_reset): Likewise.
23808 (df_mir_confluence_0): Set con_visited.
23809 (df_mir_confluence_n): Properly handle implicitely
23810 fully populated IN and OUT as designated by con_visited
23811 and update con_visited accordingly.
23813 2021-01-29 Jakub Jelinek <jakub@redhat.com>
23816 * config/arm/vec-common.md (mve_vshlq_<supf><mode>,
23817 vashl<mode>3, vashr<mode>3, vlshr<mode>3): Add
23818 && !TARGET_REALLY_IWMMXT to conditions.
23820 2021-01-29 Jakub Jelinek <jakub@redhat.com>
23823 * cfgbuild.c (find_bb_boundaries): Reset debug_insn when seeing
23826 2021-01-28 Marek Polacek <polacek@redhat.com>
23829 * stor-layout.c (finalize_type_size): If we reset TYPE_USER_ALIGN in
23830 the main variant, maybe reset it in its variants too.
23831 * tree.c (check_base_type): Return true only if TYPE_USER_ALIGN match.
23832 (check_aligned_type): Check if TYPE_USER_ALIGN match.
23834 2021-01-28 Christophe Lyon <christophe.lyon@linaro.org>
23837 * config/arm/arm.c (arm_rtx_costs_internal): Adjust cost of vector
23838 of constant zero for comparisons.
23840 2021-01-28 Michael Meissner <meissner@linux.ibm.com>
23842 * config/rs6000/rs6000.c (rs6000_mangle_decl_assembler_name): Add
23843 support for mapping built-in function names for long double
23844 built-in functions if long double is IEEE 128-bit.
23846 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
23848 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_n
23849 builtin generator macros.
23850 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_n<mode>):
23852 * config/aarch64/arm_neon.h (vmlsl_n_s16): Use RTL builtin
23853 instead of inline asm.
23854 (vmlsl_n_s32): Likewise.
23855 (vmlsl_n_u16): Likewise.
23856 (vmlsl_n_u32): Likewise.
23858 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
23860 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_n
23861 builtin generator macros.
23862 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>):
23864 * config/aarch64/arm_neon.h (vmlal_n_s16): Use RTL builtin
23865 instead of inline asm.
23866 (vmlal_n_s32): Likewise.
23867 (vmlal_n_u16): Likewise.
23868 (vmlal_n_u32): Likewise.
23870 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23872 * config/aarch64/aarch64-simd-builtins.def (shrn2): Define
23874 * config/aarch64/aarch64-simd.md (aarch64_shrn2<mode>_insn_le):
23876 (aarch64_shrn2<mode>_insn_be): Likewise.
23877 (aarch64_shrn2<mode>): Likewise.
23878 * config/aarch64/arm_neon.h (vshrn_high_n_s16): Reimlplement
23880 (vshrn_high_n_s32): Likewise.
23881 (vshrn_high_n_s64): Likewise.
23882 (vshrn_high_n_u16): Likewise.
23883 (vshrn_high_n_u32): Likewise.
23884 (vshrn_high_n_u64): Likewise.
23886 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23888 * config/aarch64/aarch64-simd-builtins.def (shrn): Define
23890 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le):
23892 (aarch64_shrn<mode>_insn_be): Likewise.
23893 (aarch64_shrn<mode>): Likewise.
23894 * config/aarch64/arm_neon.h (vshrn_n_s16): Reimplement using
23896 (vshrn_n_s32): Likewise.
23897 (vshrn_n_s64): Likewise.
23898 (vshrn_n_u16): Likewise.
23899 (vshrn_n_u32): Likewise.
23900 (vshrn_n_u64): Likewise.
23901 * config/aarch64/iterators.md (vn_mode): New mode attribute.
23903 2021-01-28 Richard Biener <rguenther@suse.de>
23905 PR rtl-optimization/80960
23906 * dse.c (check_mem_read_rtx): Call get_addr on the
23909 2021-01-28 Xionghu Luo <luoxhu@linux.ibm.com>
23910 David Edelsohn <dje.gcc@gmail.com>
23913 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
23914 Don't generate VIEW_CONVERT_EXPR for fcode ALTIVEC_BUILTIN_VEC_INSERT
23916 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
23918 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Remove the
23919 wrapper call rs6000_expand_vector_set_var for cleanup. Call
23920 rs6000_expand_vector_set_var_p9 and rs6000_expand_vector_set_var_p8
23922 (rs6000_expand_vector_set_var): Delete.
23923 (rs6000_expand_vector_set_var_p9): Make static.
23924 (rs6000_expand_vector_set_var_p8): Make static.
23926 2021-01-28 Xing GUO <higuoxing@gmail.com>
23928 * common/config/riscv/riscv-common.c
23929 (riscv_subset_list::parsing_subset_version): Fix -march option parsing
23930 when `p` extension exists.
23932 2021-01-27 Vladimir N. Makarov <vmakarov@redhat.com>
23934 PR rtl-optimization/97684
23935 * ira.c (ira): Call ira_set_pseudo_classes before
23936 update_equiv_regs when it is necessary.
23938 2021-01-27 Jakub Jelinek <jakub@redhat.com>
23941 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Use
23942 %w0, %w1 and %2 instead of %0, %1 and %2.
23944 2021-01-27 Aaron Sawdey <acsawdey@linux.ibm.com>
23946 * config/rs6000/genfusion.pl: New script to generate
23947 define_insn_and_split patterns so combine can arrange fused
23948 instructions next to each other.
23949 * config/rs6000/fusion.md: New file, generated fused instruction
23950 patterns for combine.
23951 * config/rs6000/predicates.md (const_m1_to_1_operand): New predicate.
23952 (non_update_memory_operand): New predicate.
23953 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_P10_FUSION and
23954 OPTION_MASK_P10_FUSION_LD_CMPI to ISA_3_1_MASKS_SERVER and
23956 * config/rs6000/rs6000-protos.h (address_is_non_pfx_d_or_x): Add
23958 * config/rs6000/rs6000.c (rs6000_option_override_internal):
23959 Automatically set OPTION_MASK_P10_FUSION and
23960 OPTION_MASK_P10_FUSION_LD_CMPI if target is power10.
23961 (rs600_opt_masks): Allow -mpower10-fusion
23962 in function attributes.
23963 (address_is_non_pfx_d_or_x): New function.
23964 * config/rs6000/rs6000.h: Add MASK_P10_FUSION.
23965 * config/rs6000/rs6000.md: Include fusion.md.
23966 * config/rs6000/rs6000.opt: Add -mpower10-fusion
23967 and -mpower10-fusion-ld-cmpi.
23968 * config/rs6000/t-rs6000: Add dependencies involving fusion.md.
23970 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
23972 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal
23973 builtin generator macros.
23974 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal<mode>):
23976 (aarch64_<su>mlal<mode>): This.
23977 * config/aarch64/arm_neon.h (vmlal_s8): Use RTL builtin
23978 instead of inline asm.
23979 (vmlal_s16): Likewise.
23980 (vmlal_s32): Likewise.
23981 (vmlal_u8): Likewise.
23982 (vmlal_u16): Likewise.
23983 (vmlal_u32): Likewise.
23985 2021-01-27 Richard Biener <rguenther@suse.de>
23987 PR tree-optimization/98854
23988 * tree-vect-slp.c (vect_build_slp_tree_2): Also build
23989 PHIs from scalars when the number of CTORs matches the
23990 number of children.
23992 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
23994 * config/aarch64/aarch64-simd-builtins.def: Add mls_n builtin
23996 * config/aarch64/aarch64-simd.md (*aarch64_mls_elt_merge<mode>):
23998 (aarch64_mls_n<mode>): This.
23999 * config/aarch64/arm_neon.h (vmls_n_s16): Use RTL builtin
24001 (vmls_n_s32): Likewise.
24002 (vmls_n_u16): Likewise.
24003 (vmls_n_u32): Likewise.
24004 (vmlsq_n_s16): Likewise.
24005 (vmlsq_n_s32): Likewise.
24006 (vmlsq_n_u16): Likewise.
24007 (vmlsq_n_u32): Likewise.
24009 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
24011 * config/aarch64/aarch64-simd-builtins.def: Add mls builtin
24013 * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather
24015 (vmls_s16): Likewise.
24016 (vmls_s32): Likewise.
24017 (vmls_u8): Likewise.
24018 (vmls_u16): Likewise.
24019 (vmls_u32): Likewise.
24020 (vmlsq_s8): Likewise.
24021 (vmlsq_s16): Likewise.
24022 (vmlsq_s32): Likewise.
24023 (vmlsq_u8): Likewise.
24024 (vmlsq_u16): Likewise.
24025 (vmlsq_u32): Likewise.
24027 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
24029 * config/aarch64/aarch64-simd-builtins.def: Add mla_n builtin
24031 * config/aarch64/aarch64-simd.md (*aarch64_mla_elt_merge<mode>):
24033 (aarch64_mla_n<mode>): This.
24034 * config/aarch64/arm_neon.h (vmla_n_s16): Use RTL builtin
24036 (vmla_n_s32): Likewise.
24037 (vmla_n_u16): Likewise.
24038 (vmla_n_u32): Likewise.
24039 (vmlaq_n_s16): Likewise.
24040 (vmlaq_n_s32): Likewise.
24041 (vmlaq_n_u16): Likewise.
24042 (vmlaq_n_u32): Likewise.
24044 2021-01-27 liuhongt <hongtao.liu@intel.com>
24047 * config/i386/sse.md (sse2_gt<mode>3): Drop !TARGET_XOP in condition.
24048 (*sse2_eq<mode>3): Ditto.
24050 2021-01-27 Jakub Jelinek <jakub@redhat.com>
24052 * tree-pass.h (PROP_trees): Rename to ...
24053 (PROP_gimple): ... this.
24054 * cfgexpand.c (pass_data_expand): Replace PROP_trees with PROP_gimple.
24055 * passes.c (execute_function_dump, execute_function_todo,
24056 execute_one_ipa_transform_pass, execute_one_pass): Likewise.
24057 * varpool.c (ctor_for_folding): Likewise.
24059 2021-01-27 Jakub Jelinek <jakub@redhat.com>
24061 PR tree-optimization/97260
24062 * varpool.c: Include tree-pass.h.
24063 (ctor_for_folding): In GENERIC return DECL_INITIAL for TREE_READONLY
24064 non-TREE_SIDE_EFFECTS automatic variables.
24066 2021-01-26 Paul Fee <paul.f.fee@gmail.com>
24068 * doc/cpp.texi (__cplusplus): Document value for -std=c++23
24070 * doc/invoke.texi: Document -std=c++23 and -std=gnu++23.
24071 * dwarf2out.c (highest_c_language): Recognise C++20 and C++23.
24072 (gen_compile_unit_die): Recognise C++23.
24074 2021-01-26 Jakub Jelinek <jakub@redhat.com>
24077 * dwarf2asm.c (dw2_assemble_integer): Cast DWARF2_ADDR_SIZE to int
24080 2021-01-26 Jakub Jelinek <jakub@redhat.com>
24083 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
24084 Use UINTVAL (shft_amnt) and UINTVAL (mask) instead of INTVAL (shft_amnt)
24085 and INTVAL (mask). Add && INTVAL (mask) > 0 condition.
24087 2021-01-26 Richard Biener <rguenther@suse.de>
24089 * gimple-pretty-print.c (dump_binary_rhs): Handle
24090 VEC_WIDEN_{PLUS,MINUS}_{LO,HI}_EXPR.
24092 2021-01-26 Richard Biener <rguenther@suse.de>
24094 PR middle-end/98726
24095 * tree.h (vector_cst_int_elt): Remove.
24096 * tree.c (vector_cst_int_elt): Use poly_wide_int for computations,
24099 2021-01-26 Andrew Stubbs <ams@codesourcery.com>
24101 * config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
24102 for V64DFmode min/max reductions.
24104 2021-01-26 Jakub Jelinek <jakub@redhat.com>
24106 * dwarf2asm.c (dw2_assemble_integer): Handle size twice as large
24107 as DWARF2_ADDR_SIZE if x is not a scalar int by emitting it as
24108 two halves, one with x and the other with const0_rtx, ordered
24109 depending on endianity.
24111 2021-01-26 Alexandre Oliva <oliva@adacore.com>
24113 * gimplify.c (gimplify_decl_expr): Skip asan marking calls for
24114 temporaries not seen in binding block, and not about to be
24115 added as gimple variables.
24117 2021-01-25 Martin Sebor <msebor@redhat.com>
24120 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust warning text.
24122 2021-01-25 Martin Liska <mliska@suse.cz>
24124 * value-prof.c (get_nth_most_common_value): Use %s instead
24127 2021-01-25 Jakub Jelinek <jakub@redhat.com>
24130 * configure.ac (HAVE_AS_GDWARF_5_DEBUG_FLAG): Only define if
24131 readelf -wi is able to read the emitted .debug_info back.
24132 * configure: Regenerated.
24134 2021-01-25 Martin Liska <mliska@suse.cz>
24136 PR gcov-profile/98739
24137 * common.opt: Add missing sign symbol.
24138 * value-prof.c (get_nth_most_common_value): Restore handling
24139 of PROFILE_REPRODUCIBILITY_PARALLEL_RUNS and
24140 PROFILE_REPRODUCIBILITY_MULTITHREADED.
24142 2021-01-25 Richard Biener <rguenther@suse.de>
24144 PR middle-end/98807
24145 * tree.c (vector_element_bits): Always use precision of
24146 the element type for boolean vectors.
24148 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
24150 * config/rtems.h (STARTFILE_SPEC): Remove qnolinkcmds.
24151 (ENDFILE_SPEC): Evaluate qnolinkcmds.
24153 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
24155 * config/rtems.h (STARTFILE_SPEC): Remove nostdlib and
24156 nostartfiles handling since this is already done by
24157 LINK_COMMAND_SPEC. Evaluate qnolinkcmds.
24158 (ENDFILE_SPEC): Remove nostdlib and nostartfiles handling since this
24159 is already done by LINK_COMMAND_SPEC.
24160 (LIB_SPECS): Remove nostdlib and nodefaultlibs handling since
24161 this is already done by LINK_COMMAND_SPEC. Remove qnolinkcmds
24164 2021-01-25 Jakub Jelinek <jakub@redhat.com>
24167 * fold-const-call.c (host_size_t_cst_p): Renamed to ...
24168 (size_t_cst_p): ... this. Check and store unsigned HOST_WIDE_INT
24169 value rather than host size_t.
24170 (fold_const_call): Change type of s2 from size_t to
24171 unsigned HOST_WIDE_INT. Use size_t_cst_p instead of
24172 host_size_t_cst_p. For strncmp calls, pass MIN (s2, SIZE_MAX)
24173 instead of s2 as last argument.
24175 2021-01-25 Tamar Christina <tamar.christina@arm.com>
24177 * config/arm/iterators.md (rotsplit1, rotsplit2, conj_op, fcmac1,
24178 VCMLA_OP, VCMUL_OP): New.
24179 * config/arm/mve.md (mve_vcmlaq<mve_rot><mode>): Support vec_dup 0.
24180 * config/arm/neon.md (cmul<conj_op><mode>3): New.
24181 * config/arm/unspecs.md (UNSPEC_VCMLA_CONJ, UNSPEC_VCMLA180_CONJ,
24182 UNSPEC_VCMUL_CONJ): New.
24183 * config/arm/vec-common.md (cmul<conj_op><mode>3, arm_vcmla<rot><mode>,
24184 cml<fcmac1><conj_op><mode>4): New.
24186 2021-01-23 Jakub Jelinek <jakub@redhat.com>
24189 * config/rs6000/mmintrin.h (__m64): Add __may_alias__ attribute.
24191 2021-01-22 Jonathan Wright <jonathan.wright@arm.com>
24193 * config/aarch64/aarch64-simd-builtins.def: Add mla builtin
24195 * config/aarch64/arm_neon.h (vmla_s8): Use RTL builtin rather
24197 (vmla_s16): Likewise.
24198 (vmla_s32): Likewise.
24199 (vmla_u8): Likewise.
24200 (vmla_u16): Likewise.
24201 (vmla_u32): Likewise.
24202 (vmlaq_s8): Likewise.
24203 (vmlaq_s16): Likewise.
24204 (vmlaq_s32): Likewise.
24205 (vmlaq_u8): Likewise.
24206 (vmlaq_u16): Likewise.
24207 (vmlaq_u32): Likewise.
24209 2021-01-22 David Malcolm <dmalcolm@redhat.com>
24211 * doc/invoke.texi (GCC_EXTRA_DIAGNOSTIC_OUTPUT): Add @findex
24214 2021-01-22 Jakub Jelinek <jakub@redhat.com>
24217 * dwarf2out.c (output_file_names): For -gdwarf-5, if there are no
24218 filenames to emit, still emit the required 0 index directory and
24219 filename entries that match DW_AT_comp_dir and DW_AT_name of the
24222 2021-01-22 Marek Polacek <polacek@redhat.com>
24225 * doc/invoke.texi: Update C++ ABI Version 15 description.
24227 2021-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24229 PR tree-optimization/98766
24230 * tree-ssa-math-opts.c (convert_mult_to_fma): Use maybe_le when
24231 comparing against type size with param_avoid_fma_max_bits.
24233 2021-01-22 Richard Biener <rguenther@suse.de>
24235 PR middle-end/98793
24236 * tree.c (vector_element_bits): Key single-bit bool vector on
24237 integer mode rather than not vector mode.
24239 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
24242 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
24243 Generate ARRAY_REF(VIEW_CONVERT_EXPR) for P8 and later
24245 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): Update
24246 to call different path for P8 and P9.
24247 (rs6000_expand_vector_set_var_p9): New function.
24248 (rs6000_expand_vector_set_var_p8): New function.
24250 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
24254 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
24255 Ajdust variable index vec_insert from address dereference to
24256 ARRAY_REF(VIEW_CONVERT_EXPR) tree expression.
24257 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
24259 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): New function.
24261 2021-01-22 Martin Liska <mliska@suse.cz>
24263 PR gcov-profile/98739
24264 * profile.c (compute_value_histograms): Drop time profile for
24265 -fprofile-reproducible=multithreaded.
24267 2021-01-22 Nathan Sidwell <nathan@acm.org>
24269 * gcc.c (process_command): Don't check OPT_SPECIAL_input_file
24272 2021-01-22 Richard Biener <rguenther@suse.de>
24274 PR middle-end/98773
24275 * tree-data-ref.c (initalize_matrix_A): Revert previous
24276 change, retaining failing on HOST_WIDE_INT_MIN CHREC_RIGHT.
24278 2021-01-22 Jakub Jelinek <jakub@redhat.com>
24280 PR tree-optimization/90248
24281 * match.pd (X cmp 0.0 ? 1.0 : -1.0 -> copysign(1, +-X),
24282 X cmp 0.0 ? -1.0 : +1.0 -> copysign(1, -+X)): Remove
24284 (X * (X cmp 0.0 ? 1.0 : -1.0) -> +-abs(X),
24285 X * (X cmp 0.0 ? -1.0 : 1.0) -> +-abs(X)): New simplifications.
24287 2021-01-22 Jakub Jelinek <jakub@redhat.com>
24289 PR tree-optimization/98255
24290 * tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
24291 extend index - low_bound from sizetype's precision rather than index
24293 (get_addr_base_and_unit_offset_1): Likewise.
24294 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
24295 * gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
24297 2021-01-22 Richard Biener <rguenther@suse.de>
24299 PR tree-optimization/98786
24300 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Avoid
24301 adding new uses of abnormals. Verify we deal with a conditional
24304 2021-01-22 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24307 * optc-save-gen.awk: Add arm_fp16_format to checked_options.
24309 2021-01-22 liuhongt <hongtao.liu@intel.com>
24313 * config/i386/sse.md (VI_128_256): New mode iterator.
24314 (*avx_cmp<mode>3_1, *avx_cmp<mode>3_2, *avx_cmp<mode>3_3,
24315 *avx_cmp<mode>3_4, *avx2_eq<mode>3, *avx2_pcmp<mode>3_1,
24316 *avx2_pcmp<mode>3_2, *avx2_gt<mode>3): New
24317 define_insn_and_split to lower avx512 vector comparison to avx
24318 version when dest is vector.
24319 (*<avx512>_cmp<mode>3,*<avx512>_cmp<mode>3,*<avx512>_ucmp<mode>3):
24320 define_insn_and_split for negating the comparison result.
24321 * config/i386/predicates.md (float_vector_all_ones_operand):
24323 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Use
24324 general NOT operator without UNSPEC_MASKOP.
24326 2021-01-21 Vladimir N. Makarov <vmakarov@redhat.com>
24328 PR rtl-optimization/98777
24329 * lra-int.h (lra_pmode_pseudo): New extern.
24330 * lra.c (lra_pmode_pseudo): New global.
24332 * lra-eliminations.c (eliminate_regs_in_insn): Use it.
24334 2021-01-21 Ilya Leoshkevich <iii@linux.ibm.com>
24336 * fwprop.c (fwprop_propagation::classify_result): Allow
24337 (subreg (mem)) simplifications.
24339 2021-01-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24341 * config/aarch64/aarch64-simd.md (aarch64_sqdml<SBINQOPS:as>l<mode>):
24343 (aarch64_sqdmlal<mode>): ... This...
24344 (aarch64_sqdmlsl<mode>): ... And this.
24345 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Split into...
24346 (aarch64_sqdmlal_lane<mode>): ... This...
24347 (aarch64_sqdmlsl_lane<mode>): ... And this.
24348 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Split into...
24349 (aarch64_sqdmlsl_laneq<mode>): ... This...
24350 (aarch64_sqdmlal_laneq<mode>): ... And this.
24351 (aarch64_sqdml<SBINQOPS:as>l_n<mode>): Split into...
24352 (aarch64_sqdmlsl_n<mode>): ... This...
24353 (aarch64_sqdmlal_n<mode>): ... And this.
24354 (aarch64_sqdml<SBINQOPS:as>l2<mode>_internal): Split into...
24355 (aarch64_sqdmlal2<mode>_internal): ... This...
24356 (aarch64_sqdmlsl2<mode>_internal): ... And this.
24358 2021-01-21 Christophe Lyon <christophe.lyon@linaro.org>
24360 * config/arm/arm_mve.h (__arm_vcmpneq_s8): Fix return type.
24362 2021-01-21 Andrea Corallo <andrea.corallo@arm.com>
24365 * doc/sourcebuild.texi (arm_thumb2_no_arm_v8_1_lob): Document.
24367 2021-01-21 liuhongt <hongtao.liu@intel.com>
24369 PR rtl-optimization/98694
24370 * regcprop.c (copy_value): If SRC had been assigned a mode
24371 narrower than the copy, we can't link DEST into the chain even
24372 they have same hard_regno_nregs(i.e. HImode/SImode in i386
24375 2021-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24377 * config/aarch64/aarch64-simd.md (aarch64_get_lane<mode>):
24378 Convert to define_insn_and_split. Split into simple move when moving
24381 2021-01-20 Segher Boessenkool <segher@kernel.crashing.org>
24383 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Change assert.
24384 Adjust comment. Simplify code.
24386 2021-01-20 Jakub Jelinek <jakub@redhat.com>
24389 * dwarf2out.c (reset_indirect_string): Also reset indirect strings
24390 with DW_FORM_line_strp form.
24391 (prune_unused_types_update_strings): Don't add into debug_str_hash
24392 indirect strings with DW_FORM_line_strp form.
24393 (adjust_name_comp_dir): New function.
24394 (dwarf2out_finish): Call it on CU DIEs after resetting
24395 debug_line_str_hash.
24397 2021-01-20 Vladimir N. Makarov <vmakarov@redhat.com>
24399 PR rtl-optimization/98722
24400 * lra-eliminations.c (eliminate_regs_in_insn): Check that target
24401 has no 3-op add insn to transform insns containing two pluses.
24403 2021-01-20 Richard Biener <rguenther@suse.de>
24405 * hwint.h (add_hwi): New function.
24406 (mul_hwi): Likewise.
24407 * tree-data-ref.c (initialize_matrix_A): Properly translate
24408 tree constants and avoid HOST_WIDE_INT_MIN.
24409 (lambda_matrix_row_add): Avoid undefined integer overflow
24410 and return true on such overflow.
24411 (lambda_matrix_right_hermite): Handle overflow from
24412 lambda_matrix_row_add gracefully. Simplify previous fix.
24413 (analyze_subscript_affine_affine): Likewise.
24415 2021-01-20 Eugene Rozenfeld <erozen@microsoft.com>
24417 PR tree-optimization/96674
24418 * match.pd: New patterns: x < y || y == XXX_MIN --> x <= y - 1
24419 x >= y && y != XXX_MIN --> x > y - 1
24421 2021-01-20 Richard Sandiford <richard.sandiford@arm.com>
24423 PR tree-optimization/98535
24424 * tree-vect-slp.c (duplicate_and_interleave): Use quick_grow_cleared.
24425 If the high and low permutes are the same, remove the high permutes
24426 from the working set and only continue with the low ones.
24428 2021-01-20 Jakub Jelinek <jakub@redhat.com>
24430 PR tree-optimization/98721
24431 * builtins.c (access_ref::inform_access): Don't assume
24432 SSA_NAME_IDENTIFIER must be non-NULL. Print messages about
24433 object whenever allocfn is NULL, rather than only when DECL_P
24434 is true. Use %qE instead of %qD for that. Formatting fixes.
24436 2021-01-20 Richard Biener <rguenther@suse.de>
24438 PR tree-optimization/98758
24439 * tree-data-ref.c (int_divides_p): Use lambda_int arguments.
24440 (lambda_matrix_right_hermite): Avoid undefinedness with
24441 signed integer abs and multiplication.
24442 (analyze_subscript_affine_affine): Use lambda_int.
24444 2021-01-20 David Malcolm <dmalcolm@redhat.com>
24447 * dwarf2out.c (output_line_info): Rename static variable
24448 "generation", moving it out of the function to...
24449 (output_line_info_generation): New.
24450 (init_sections_and_labels): Likewise, renaming the variable to...
24451 (init_sections_and_labels_generation): New.
24452 (dwarf2out_c_finalize): Reset the new variables.
24454 2021-01-19 Martin Sebor <msebor@redhat.com>
24456 PR middle-end/98664
24457 * tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
24458 all functions, even if they're not declared artificial or inline.
24459 * tree.c (tree_inlined_location): Use macro expansion location
24460 only if scope traversal fails to expose one.
24462 2021-01-19 Richard Sandiford <richard.sandiford@arm.com>
24464 PR rtl-optimization/92294
24465 * alias.c (compare_base_symbol_refs): Take an extra parameter
24466 and add the distance between two symbols to it. Enshrine in
24467 comments that -1 means "either 0 or 1, but we can't tell
24468 which at compile time".
24469 (memrefs_conflict_p): Update call accordingly.
24470 (rtx_equal_for_memref_p): Likewise. Take the distance between symbols
24473 2021-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24475 * config/aarch64/aarch64-simd-builtins.def (sqshl, uqshl,
24476 sqrshl, uqrshl, sqadd, uqadd, sqsub, uqsub, suqadd, usqadd, sqmovn,
24477 uqmovn, sqxtn2, uqxtn2, sqabs, sqneg, sqdmlal, sqdmlsl, sqdmlal_lane,
24478 sqdmlsl_lane, sqdmlal_laneq, sqdmlsl_laneq, sqdmlal_n, sqdmlsl_n,
24479 sqdmlal2, sqdmlsl2, sqdmlal2_lane, sqdmlsl2_lane, sqdmlal2_laneq,
24480 sqdmlsl2_laneq, sqdmlal2_n, sqdmlsl2_n, sqdmull, sqdmull_lane,
24481 sqdmull_laneq, sqdmull_n, sqdmull2, sqdmull2_lane, sqdmull2_laneq,
24482 sqdmull2_n, sqdmulh, sqrdmulh, sqdmulh_lane, sqdmulh_laneq,
24483 sqrdmulh_lane, sqrdmulh_laneq, sqshrun_n, sqrshrun_n, sqshrn_n,
24484 uqshrn_n, sqrshrn_n, uqrshrn_n, sqshlu_n, sqshl_n, uqshl_n, sqrdmlah,
24485 sqrdmlsh, sqrdmlah_lane, sqrdmlsh_lane, sqrdmlah_laneq, sqrdmlsh_laneq,
24486 sqmovun): Use NONE flags.
24488 2021-01-19 Richard Biener <rguenther@suse.de>
24491 * ipa-modref.c (analyze_stmt): Only record a summary for a
24494 2021-01-19 Richard Biener <rguenther@suse.de>
24496 PR middle-end/98638
24497 * tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
24499 2021-01-19 Daniel Hellstrom <daniel@gaisler.com>
24501 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
24502 built-in define __FIX_LEON3FT_TN0018.
24504 2021-01-19 Richard Biener <rguenther@suse.de>
24507 * tree-inline.c (tree_function_versioning): Set input_location
24508 to UNKNOWN_LOCATION throughout the function.
24510 2021-01-19 Tobias Burnus <tobias@codesourcery.com>
24513 * omp-low.c (lower_omp_target): Handle nonpointer is_device_ptr.
24515 2021-01-19 Martin Jambor <mjambor@suse.cz>
24518 * ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
24519 whether non-call exceptions allow removal of a statement.
24520 (isra_analyze_call): Pass the appropriate function to
24521 ssa_name_only_returned_p.
24523 2021-01-19 Geng Qi <gengqi@linux.alibaba.com>
24525 * config/riscv/arch-canonicalize (longext_sort): New function for
24526 sorting 'multi-letter'.
24527 * config/riscv/multilib-generator: Adjusting the loop of 'alt' in
24528 'alts'. The 'arch' may not be the first of 'alts'.
24529 (_expand_combination): Add underline for the 'ext' without '*'.
24530 This is because, a single-letter extension can always be treated well
24531 with a '_' prefix, but it cannot be separated out if it is appended
24534 2021-01-18 Vladimir N. Makarov <vmakarov@redhat.com>
24537 * ira.c (ira): Skip abnormal critical edge splitting.
24539 2021-01-18 Jakub Jelinek <jakub@redhat.com>
24541 PR tree-optimization/98727
24542 * tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
24543 second .MUL_OVERFLOW operand for signed multiplication with overflow
24544 checking if the second operand of multiplication is not constant.
24546 2021-01-18 David Edelsohn <dje.gcc@gmail.com>
24548 * doc/invoke.texi (-gdwarf): TPF defaults to version 2 and AIX
24549 defaults to version 4.
24551 2021-01-18 David Malcolm <dmalcolm@redhat.com>
24553 * attribs.h (fndecl_dealloc_argno): New decl.
24554 * builtins.c (call_dealloc_argno): Split out second half of
24556 (fndecl_dealloc_argno): New.
24557 * doc/extend.texi (Common Function Attributes): Document the
24558 interaction between the analyzer and the malloc attribute.
24559 * doc/invoke.texi (Static Analyzer Options): Likewise.
24561 2021-01-17 David Edelsohn <dje.gcc@gmail.com>
24563 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Override
24564 dwarf_version to 4.
24565 * config/rs6000/aix72.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
24567 2021-01-17 Martin Jambor <mjambor@suse.cz>
24570 * cgraph.c (clone_of_p): Check also former_clone_of as we climb
24573 2021-01-17 Mark Wielaard <mark@klomp.org>
24575 * common.opt (gdwarf-): Init(5).
24576 * doc/invoke.texi (-gdwarf): Document default to 5.
24578 2021-01-16 Kwok Cheung Yeung <kcy@codesourcery.com>
24580 * builtin-types.def
24581 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
24583 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
24584 ...this. Add extra argument.
24585 * gimplify.c (omp_default_clause): Ensure that event handle is
24586 firstprivate in a task region.
24587 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_DETACH.
24588 (gimplify_adjust_omp_clauses): Likewise.
24589 * omp-builtins.def (BUILT_IN_GOMP_TASK): Change function type to
24590 BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR.
24591 * omp-expand.c (expand_task_call): Add GOMP_TASK_FLAG_DETACH to flags
24592 if detach clause specified. Add detach argument when generating
24594 * omp-low.c (scan_sharing_clauses): Setup data environment for detach
24596 (finish_taskreg_scan): Move field for variable containing the event
24597 handle to the front of the struct.
24598 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_DETACH. Fix
24600 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
24601 OMP_CLAUSE_DETACH clause.
24602 (convert_local_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
24603 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_DETACH.
24604 * tree.c (omp_clause_num_ops): Add entry for OMP_CLAUSE_DETACH.
24606 (omp_clause_code_name): Add entry for OMP_CLAUSE_DETACH. Fix
24608 (walk_tree_1): Handle OMP_CLAUSE_DETACH.
24610 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
24612 * config/nios2/t-rtems: Reset all MULTILIB_* variables. Shorten
24613 multilib directory names. Use MULTILIB_REQUIRED instead of
24614 MULTILIB_EXCEPTIONS. Add -mhw-mul -mhw-mulx -mhw-div
24615 -mcustom-fpu-cfg=fph2 multilib.
24617 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
24619 * config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
24620 (nios2_init_fpu_configs): Provide register values for new
24621 -mcustom-fpu-cfg=fph2 option variant.
24622 * doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
24625 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
24627 * config/nios2/nios2.c (nios2_custom_check_insns): Remove
24628 custom instruction warnings.
24630 2021-01-16 Jakub Jelinek <jakub@redhat.com>
24632 PR tree-optimization/96669
24633 * match.pd ((CST << x) & 1 -> x == 0): New simplification.
24635 2021-01-16 Jakub Jelinek <jakub@redhat.com>
24637 PR tree-optimization/96271
24638 * passes.def: Pass false argument to first two pass_cd_dce
24639 instances and true to last instance. Add comment that
24640 last instance rewrites no longer addressed locals.
24641 * tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
24643 (pass_cd_dce::set_pass_param): New method.
24644 (pass_cd_dce::execute): Return TODO_update_address_taken from
24645 last cd_dce instance.
24647 2021-01-15 Carl Love <cel@us.ibm.com>
24649 * config/rs6000/altivec.h (vec_mulh, vec_div, vec_dive, vec_mod):
24651 * config/rs6000/altivec.md (VIlong): Move define to file vsx.md.
24652 * config/rs6000/rs6000-builtin.def (DIVES_V4SI, DIVES_V2DI,
24653 DIVEU_V4SI, DIVEU_V2DI, DIVS_V4SI, DIVS_V2DI, DIVU_V4SI,
24654 DIVU_V2DI, MODS_V2DI, MODS_V4SI, MODU_V2DI, MODU_V4SI,
24655 MULHS_V2DI, MULHS_V4SI, MULHU_V2DI, MULHU_V4SI, MULLD_V2DI):
24656 Add builtin define.
24657 (MULH, DIVE, MOD): Add new BU_P10_OVERLOAD_2 definitions.
24658 * config/rs6000/rs6000-call.c (VSX_BUILTIN_VEC_DIV,
24659 VSX_BUILTIN_VEC_DIVE, P10_BUILTIN_VEC_MOD, P10_BUILTIN_VEC_MULH):
24660 New overloaded definitions.
24661 (builtin_function_type) [P10V_BUILTIN_DIVEU_V4SI,
24662 P10V_BUILTIN_DIVEU_V2DI, P10V_BUILTIN_DIVU_V4SI,
24663 P10V_BUILTIN_DIVU_V2DI, P10V_BUILTIN_MODU_V2DI,
24664 P10V_BUILTIN_MODU_V4SI, P10V_BUILTIN_MULHU_V2DI,
24665 P10V_BUILTIN_MULHU_V4SI]: Add case
24666 statement for builtins.
24667 * config/rs6000/rs6000.md (bits): Add new attribute sizes V4SI, V2DI.
24668 * config/rs6000/vsx.md (VIlong): Moved from config/rs6000/altivec.md.
24669 (UNSPEC_VDIVES, UNSPEC_VDIVEU): New unspec definitions.
24670 (vsx_mul_v2di): Add if TARGET_POWER10 statement.
24671 (vsx_udiv_v2di): Add if TARGET_POWER10 statement.
24672 (dives_<mode>, diveu_<mode>, div<mode>3, uvdiv<mode>3,
24673 mods_<mode>, modu_<mode>, mulhs_<mode>, mulhu_<mode>, mulv2di3):
24674 Add define_insn, mode is VIlong.
24675 * doc/extend.texi (vec_mulh, vec_mul, vec_div, vec_dive, vec_mod):
24676 Add builtin descriptions.
24678 2021-01-15 Eric Botcazou <ebotcazou@adacore.com>
24680 * final.c (final_start_function_1): Reset force_source_line.
24682 2021-01-15 Jakub Jelinek <jakub@redhat.com>
24684 PR tree-optimization/96669
24685 * match.pd (((1 << A) & 1) != 0 -> A == 0,
24686 ((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
24687 possibly different power of two constants and to right shift too.
24689 2021-01-15 Jakub Jelinek <jakub@redhat.com>
24691 PR tree-optimization/96681
24692 * match.pd ((x < 0) ^ (y < 0) to (x ^ y) < 0): New simplification.
24693 ((x >= 0) ^ (y >= 0) to (x ^ y) < 0): Likewise.
24694 ((x < 0) ^ (y >= 0) to (x ^ y) >= 0): Likewise.
24695 ((x >= 0) ^ (y < 0) to (x ^ y) >= 0): Likewise.
24697 2021-01-15 Alexandre Oliva <oliva@adacore.com>
24699 * opts.c (gen_command_line_string): Exclude -dumpbase-ext.
24701 2021-01-15 Tamar Christina <tamar.christina@arm.com>
24703 * config/aarch64/aarch64-simd.md (cml<fcmac1><conj_op><mode>4,
24704 cmul<conj_op><mode>3): New.
24705 * config/aarch64/iterators.md (UNSPEC_FCMUL,
24706 UNSPEC_FCMUL180, UNSPEC_FCMLA_CONJ, UNSPEC_FCMLA180_CONJ,
24707 UNSPEC_CMLA_CONJ, UNSPEC_CMLA180_CONJ, UNSPEC_CMUL, UNSPEC_CMUL180,
24708 FCMLA_OP, FCMUL_OP, conj_op, rotsplit1, rotsplit2, fcmac1, sve_rot1,
24709 sve_rot2, SVE2_INT_CMLA_OP, SVE2_INT_CMUL_OP, SVE2_INT_CADD_OP): New.
24710 (rot): Add UNSPEC_FCMUL, UNSPEC_FCMUL180.
24711 (rot_op): Renamed to conj_op.
24712 * config/aarch64/aarch64-sve.md (cml<fcmac1><conj_op><mode>4,
24713 cmul<conj_op><mode>3): New.
24714 * config/aarch64/aarch64-sve2.md (cml<fcmac1><conj_op><mode>4,
24715 cmul<conj_op><mode>3): New.
24717 2021-01-15 David Malcolm <dmalcolm@redhat.com>
24721 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
24722 Escape the tempfile name when constructing the expected output.
24724 2021-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24726 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlsl_hi<mode>):
24728 (aarch64_<su>mlsl_hi<mode>): ... This.
24729 (aarch64_<su>mlsl_hi<mode>): Define.
24730 (*aarch64_<su>mlsl<mode): Rename to...
24731 (aarch64_<su>mlsl<mode): ... This.
24732 * config/aarch64/aarch64-simd-builtins.def (smlsl, umlsl,
24733 smlsl_hi, umlsl_hi): Define builtins.
24734 * config/aarch64/arm_neon.h (vmlsl_high_s8, vmlsl_high_s16,
24735 vmlsl_high_s32, vmlsl_high_u8, vmlsl_high_u16, vmlsl_high_u32,
24736 vmlsl_s8, vmlsl_s16, vmlsl_s32, vmlsl_u8,
24737 vmlsl_u16, vmlsl_u32): Reimplement with builtins.
24739 2021-01-15 Uroš Bizjak <ubizjak@gmail.com>
24741 * config/i386/i386-c.c (ix86_target_macros):
24742 Use cpp_define_formatted for __SIZEOF_FLOAT80__ definition.
24744 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
24747 * config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
24748 * Makefile.in (RTL_SSA_H): New variable.
24749 * config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
24750 * config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
24751 * config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
24753 * config/aarch64/aarch64-cc-fusion.cc: New file.
24755 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
24757 * recog.h (insn_change_watermark::~insn_change_watermark): Avoid
24758 calling cancel_changes for changes that no longer exist.
24760 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
24762 * rtl-ssa/functions.h (function_info::ref_defs): Rename to...
24763 (function_info::reg_defs): ...this.
24764 * rtl-ssa/member-fns.inl (function_info::ref_defs): Rename to...
24765 (function_info::reg_defs): ...this.
24767 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
24770 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
24772 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
24775 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
24778 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
24780 2021-01-15 Richard Biener <rguenther@suse.de>
24782 PR tree-optimization/96376
24783 * tree-vect-stmts.c (get_load_store_type): Disregard alignment
24784 for VMAT_INVARIANT.
24786 2021-01-15 Martin Liska <mliska@suse.cz>
24788 * doc/install.texi: Document that some tests need pytest module.
24789 * doc/sourcebuild.texi: Likewise.
24791 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
24794 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
24796 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
24798 * config/arm/mve.md (mve_vshrq_n_s<mode>_imm): New entry.
24799 (mve_vshrq_n_u<mode>_imm): Likewise.
24800 * config/arm/neon.md (vashr<mode>3, vlshr<mode>3): Move to ...
24801 * config/arm/vec-common.md: ... here.
24803 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
24805 * config/arm/mve.md (mve_vshlq_<supf><mode>): Move to
24807 * config/arm/neon.md (vashl<mode>3): Delete.
24808 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): New.
24809 (vasl<mode>3): New expander.
24811 2021-01-15 Richard Biener <rguenther@suse.de>
24813 PR tree-optimization/98685
24814 * tree-vect-slp.c (vect_schedule_slp_node): Refactor handling
24815 of vector extern defs.
24817 2021-01-14 David Malcolm <dmalcolm@redhat.com>
24820 * diagnostic.c (diagnostic_kind_text): Break out this array
24822 (diagnostic_build_prefix): ...here.
24823 (fancy_abort): Detect when diagnostic_initialize has not yet been
24824 called and fall back to a minimal implementation of printing the
24825 ICE, rather than segfaulting in internal_error.
24827 2021-01-14 David Malcolm <dmalcolm@redhat.com>
24829 * diagnostic.c (diagnostic_initialize): Eliminate
24830 parseable_fixits_p in favor of initializing extra_output_kind from
24831 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
24832 (convert_column_unit): New function, split out from...
24833 (diagnostic_converted_column): ...this.
24834 (print_parseable_fixits): Add "column_unit" and "tabstop" params.
24835 Use them to call convert_column_unit on the column values.
24836 (diagnostic_report_diagnostic): Eliminate conditional on
24837 parseable_fixits_p in favor of a switch statement on
24838 extra_output_kind, passing the appropriate values to the new
24839 params of print_parseable_fixits.
24840 (selftest::test_print_parseable_fixits_none): Update for new
24841 params of print_parseable_fixits.
24842 (selftest::test_print_parseable_fixits_insert): Likewise.
24843 (selftest::test_print_parseable_fixits_remove): Likewise.
24844 (selftest::test_print_parseable_fixits_replace): Likewise.
24845 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
24847 (selftest::diagnostic_c_tests): Call it.
24848 * diagnostic.h (enum diagnostics_extra_output_kind): New.
24849 (diagnostic_context::parseable_fixits_p): Delete field in favor
24851 (diagnostic_context::extra_output_kind): ...this new field.
24852 * doc/invoke.texi (Environment Variables): Add
24853 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
24854 * opts.c (common_handle_option): Update handling of
24855 OPT_fdiagnostics_parseable_fixits for change to diagnostic_context
24858 2021-01-14 Tamar Christina <tamar.christina@arm.com>
24860 * tree-vect-slp-patterns.c (class complex_operations_pattern,
24861 complex_operations_pattern::matches,
24862 complex_operations_pattern::recognize,
24863 complex_operations_pattern::build): New.
24864 (slp_patterns): Use it.
24866 2021-01-14 Tamar Christina <tamar.christina@arm.com>
24868 * internal-fn.def (COMPLEX_FMS, COMPLEX_FMS_CONJ): New.
24869 * optabs.def (cmls_optab, cmls_conj_optab): New.
24870 * doc/md.texi: Document them.
24871 * tree-vect-slp-patterns.c (class complex_fms_pattern,
24872 complex_fms_pattern::matches, complex_fms_pattern::recognize,
24873 complex_fms_pattern::build): New.
24875 2021-01-14 Tamar Christina <tamar.christina@arm.com>
24877 * internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
24878 * optabs.def (cmla_optab, cmla_conj_optab): New.
24879 * doc/md.texi: Document them.
24880 * tree-vect-slp-patterns.c (vect_match_call_p,
24881 class complex_fma_pattern, vect_slp_reset_pattern,
24882 complex_fma_pattern::matches, complex_fma_pattern::recognize,
24883 complex_fma_pattern::build): New.
24885 2021-01-14 Tamar Christina <tamar.christina@arm.com>
24887 * internal-fn.def (COMPLEX_MUL, COMPLEX_MUL_CONJ): New.
24888 * optabs.def (cmul_optab, cmul_conj_optab): New.
24889 * doc/md.texi: Document them.
24890 * tree-vect-slp-patterns.c (vect_match_call_complex_mla,
24891 vect_normalize_conj_loc, is_eq_or_top, vect_validate_multiplication,
24892 vect_build_combine_node, class complex_mul_pattern,
24893 complex_mul_pattern::matches, complex_mul_pattern::recognize,
24894 complex_mul_pattern::build): New.
24896 2021-01-14 Tamar Christina <tamar.christina@arm.com>
24898 * tree-vect-slp.c (optimize_load_redistribution_1): New.
24899 (optimize_load_redistribution, vect_is_slp_load_node): New.
24900 (vect_match_slp_patterns): Use it.
24902 2021-01-14 Tamar Christina <tamar.christina@arm.com>
24904 * tree-vect-slp-patterns.c (complex_add_pattern::build):
24907 2021-01-14 Thomas Schwinge <thomas@codesourcery.com>
24909 * config/gcn/mkoffload.c (main): Create an offload image only in
24910 64-bit configurations.
24912 2021-01-14 H.J. Lu <hjl.tools@gmail.com>
24915 * config/i386/i386-options.c (ix86_option_override_internal):
24916 Issue an error for -fcf-protection with CF_BRANCH when compiling
24917 for 32-bit non-TARGET_CMOV targets.
24919 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
24922 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p):
24923 Remove declaration and initialization of shadow variable "ret".
24924 (ix86_option_override_internal): Remove delcaration of
24925 shadow variable "i". Redeclare shadowed variable to unsigned.
24926 * common/config/i386/i386-common.c (pta_size): Redeclare to unsigned.
24927 * config/i386/i386-builtins.c (get_builtin_code_for_version):
24928 Update for redeclaration.
24929 * config/i386/i386.h (pta_size): Ditto.
24931 2021-01-14 Richard Biener <rguenther@suse.de>
24933 PR tree-optimization/98674
24934 * tree-data-ref.c (base_supports_access_fn_components_p): New.
24935 (initialize_data_dependence_relation): For two bases without
24936 possible access fns resort to type size equality when determining
24937 shape compatibility.
24939 2021-01-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24942 * config/arm/arm_neon.h: Replace calls to __builtin_vcge* by
24943 <=, >= operators in vcle and vcge intrinsics respectively.
24944 * config/arm/arm_neon_builtins.def: Remove entry for
24947 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
24950 * config/i386/i386-options.c (ix86_function_specific_save):
24951 Remove redundant assignment to opts->x_ix86_branch_cost.
24952 * config/i386/i386.c (ix86_prefetch_sse):
24953 Rename from x86_prefetch_sse. Update all uses.
24954 * config/i386/i386.h: Update for rename.
24955 * config/i386/i386-options.h: Ditto.
24957 2021-01-14 Jakub Jelinek <jakub@redhat.com>
24960 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3,
24961 *sse4_1_zero_extendv4hiv4si2_3, *sse4_1_zero_extendv2siv2di2_3):
24962 Use Bm instead of m for non-avx. Add isa attribute.
24964 2021-01-14 Jakub Jelinek <jakub@redhat.com>
24966 PR tree-optimization/96688
24967 * match.pd (~(X >> Y) -> ~X >> Y): New simplification if
24968 ~X can be simplified.
24970 2021-01-14 Richard Sandiford <richard.sandiford@arm.com>
24972 * tree-vect-stmts.c (vect_model_load_cost): Account for unused
24973 IFN_LOAD_LANES results.
24975 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24977 * config/aarch64/aarch64-simd.md (aarch64_<su>xtl<mode>):
24979 (aarch64_xtn<mode>): Likewise.
24980 * config/aarch64/aarch64-simd-builtins.def (sxtl, uxtl, xtn):
24983 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
24985 (vmovl_s16): Likewise.
24986 (vmovl_s32): Likewise.
24987 (vmovl_u8): Likewise.
24988 (vmovl_u16): Likewise.
24989 (vmovl_u32): Likewise.
24990 (vmovn_s16): Likewise.
24991 (vmovn_s32): Likewise.
24992 (vmovn_s64): Likewise.
24993 (vmovn_u16): Likewise.
24994 (vmovn_u32): Likewise.
24995 (vmovn_u64): Likewise.
24997 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24999 * config/aarch64/aarch64-simd.md (aarch64_<su>qxtn2<mode>_le):
25001 (aarch64_<su>qxtn2<mode>_be): Likewise.
25002 (aarch64_<su>qxtn2<mode>): Likewise.
25003 * config/aarch64/aarch64-simd-builtins.def (sqxtn2, uqxtn2):
25005 * config/aarch64/iterators.md (SAT_TRUNC): Define code_iterator.
25006 (su): Handle ss_truncate and us_truncate.
25007 * config/aarch64/arm_neon.h (vqmovn_high_s16): Reimplement using
25009 (vqmovn_high_s32): Likewise.
25010 (vqmovn_high_s64): Likewise.
25011 (vqmovn_high_u16): Likewise.
25012 (vqmovn_high_u32): Likewise.
25013 (vqmovn_high_u64): Likewise.
25015 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25017 * config/aarch64/aarch64-simd.md (aarch64_xtn2<mode>_le):
25019 (aarch64_xtn2<mode>_be): Likewise.
25020 (aarch64_xtn2<mode>): Likewise.
25021 * config/aarch64/aarch64-simd-builtins.def (xtn2): Define
25023 * config/aarch64/arm_neon.h (vmovn_high_s16): Reimplement using
25025 (vmovn_high_s32): Likewise.
25026 (vmovn_high_s64): Likewise.
25027 (vmovn_high_u16): Likewise.
25028 (vmovn_high_u32): Likewise.
25029 (vmovn_high_u64): Likewise.
25031 2021-01-13 Stafford Horne <shorne@gmail.com>
25033 * config/or1k/or1k.h (ASM_PREFERRED_EH_DATA_FORMAT): New macro.
25035 2021-01-13 Stafford Horne <shorne@gmail.com>
25037 * config/or1k/linux.h (TARGET_ASM_FILE_END): Define macro.
25039 2021-01-13 Stafford Horne <shorne@gmail.com>
25041 * config/or1k/or1k.h (TARGET_CPU_CPP_BUILTINS): Add builtin
25042 define for __or1k_hard_float__.
25044 2021-01-13 Stafford Horne <shorne@gmail.com>
25046 * config/or1k/or1k.h (NO_PROFILE_COUNTERS): Define as 1.
25047 (PROFILE_HOOK): Define to call _mcount.
25048 (FUNCTION_PROFILER): Change from abort to no-op.
25050 2021-01-13 Jakub Jelinek <jakub@redhat.com>
25052 PR tree-optimization/96691
25053 * match.pd ((~X | C) ^ D -> (X | C) ^ (~D ^ C),
25054 (~X & C) ^ D -> (X & C) ^ (D ^ C)): New simplifications if
25055 (~D ^ C) or (D ^ C) can be simplified.
25057 2021-01-13 Richard Biener <rguenther@suse.de>
25059 PR tree-optimization/92645
25060 * match.pd (BIT_FIELD_REF to conversion): Delay canonicalization
25061 until after vector lowering.
25063 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
25065 * config/aarch64/aarch64-sve.md (fnma<mode>4): Extend from SVE_FULL_I
25067 (@aarch64_pred_fnma<mode>, cond_fnma<mode>, *cond_fnma<mode>_2)
25068 (*cond_fnma<mode>_4, *cond_fnma<mode>_any): Likewise.
25070 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
25072 * config/aarch64/aarch64-sve.md (fma<mode>4): Extend from SVE_FULL_I
25074 (@aarch64_pred_fma<mode>, cond_fma<mode>, *cond_fma<mode>_2)
25075 (*cond_fma<mode>_4, *cond_fma<mode>_any): Likewise.
25077 2021-01-13 Richard Biener <rguenther@suse.de>
25079 PR tree-optimization/92645
25080 * tree-vect-slp.c (vect_build_slp_tree_1): Relax supported
25081 BIT_FIELD_REF argument.
25082 (vect_build_slp_tree_2): Record the desired vector type
25083 on the external vector def.
25084 (vectorizable_slp_permutation): Handle required punning
25085 of existing vector defs.
25087 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
25089 * rtl-ssa/accesses.h (def_lookup): Fix order of comparison results.
25091 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
25093 * config/sh/sh.md (movsf_ie): Remove operands[2] test.
25095 2021-01-13 Samuel Thibault <samuel.thibault@ens-lyon.org>
25097 * config.gcc [$target == *-*-gnu*]: Enable
25098 'default_gnu_indirect_function'.
25100 2021-01-13 Jakub Jelinek <jakub@redhat.com>
25103 * optabs.c (expand_vec_perm_const): Don't force v0 and v1 into
25104 registers before calling targetm.vectorize.vec_perm_const, only after
25106 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const): Handle
25107 two argument permutation when one operand is zero vector and only
25108 after that force operands into registers.
25109 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_1): New
25110 define_insn_and_split pattern.
25111 (*avx512bw_zero_extendv32qiv32hi2_1): Likewise.
25112 (*avx512f_zero_extendv16hiv16si2_1): Likewise.
25113 (*avx2_zero_extendv8hiv8si2_1): Likewise.
25114 (*avx512f_zero_extendv8siv8di2_1): Likewise.
25115 (*avx2_zero_extendv4siv4di2_1): Likewise.
25116 * config/mips/mips.c (mips_vectorize_vec_perm_const): Force operands
25118 * config/arm/arm.c (arm_vectorize_vec_perm_const): Likewise.
25119 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Likewise.
25120 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const): Likewise.
25121 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const): Likewise.
25122 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const): Likewise.
25123 * config/gcn/gcn.c (gcn_vectorize_vec_perm_const): Likewise. Use std::swap.
25125 2021-01-13 Martin Liska <mliska@suse.cz>
25127 PR tree-optimization/98455
25128 * gimple-if-to-switch.cc (condition_info::record_phi_mapping):
25129 Record also virtual PHIs.
25130 (pass_if_to_switch::execute): Return TODO_cleanup_cfg only
25133 2021-01-13 Jonathan Wakely <jwakely@redhat.com>
25135 * doc/invoke.texi (C++ Modules): Fix typos.
25137 2021-01-13 Richard Biener <rguenther@suse.de>
25139 PR tree-optimization/98640
25140 * tree-ssa-sccvn.c (visit_nary_op): Do not try to
25141 handle plus or minus from a truncated operand to be
25144 2021-01-13 Jakub Jelinek <jakub@redhat.com>
25147 * config/i386/i386.md (*btr<mode>_1, *btr<mode>_2): New
25148 define_insn_and_split patterns.
25149 (splitter after *btr<mode>_2): New splitter.
25151 2021-01-13 Martin Liska <mliska@suse.cz>
25154 * cgraphunit.c (analyze_functions): Remove dead code.
25156 2021-01-13 Qian Jianhua <qianjh@cn.fujitsu.com>
25158 * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
25159 * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
25160 (a64fx_regmove_cost, a64fx_vector_cost): New.
25161 (a64fx_tunings): Use the new added cost tables.
25163 2021-01-13 Jakub Jelinek <jakub@redhat.com>
25166 * config/i386/predicates.md (pmovzx_parallel): New predicate.
25167 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): New
25168 define_insn_and_split pattern.
25169 (*sse4_1_zero_extendv4hiv4si2_3): Likewise.
25170 (*sse4_1_zero_extendv2siv2di2_3): Likewise.
25172 2021-01-13 Julian Brown <julian@codesourcery.com>
25174 * config/gcn/gcn.c (gcn_conditional_register_usage): Remove dead code
25175 to fix v0 register.
25177 2021-01-13 Julian Brown <julian@codesourcery.com>
25179 * config/gcn/gcn.c (gcn_md_reorg): Fix case where EXEC reg is live
25182 2021-01-13 Julian Brown <julian@codesourcery.com>
25184 * config/gcn/gcn-valu.md (recip<mode>2<exec>, recip<mode>2): Use unspec
25185 for reciprocal-approximation instructions.
25186 (div<mode>3): Use fused multiply-accumulate operations for reciprocal
25187 refinement and division result.
25188 * config/gcn/gcn.md (UNSPEC_RCP): New unspec constant.
25190 2021-01-13 Julian Brown <julian@codesourcery.com>
25192 * config/gcn/gcn-valu.md (subdf): Rename to...
25195 2021-01-12 Martin Liska <mliska@suse.cz>
25197 * gcov.c (source_info::debug): Fix printf format for 32-bit hosts.
25199 2021-01-12 Andrea Corallo <andrea.corallo@arm.com>
25201 * function-abi.h: Fix typo.
25203 2021-01-12 Christophe Lyon <christophe.lyon@linaro.org>
25207 * config/arm/arm.h (ARM_HAVE_NEON_V8QI_LDST): New macro.
25208 (ARM_HAVE_NEON_V16QI_LDST, ARM_HAVE_NEON_V4HI_LDST): Likewise.
25209 (ARM_HAVE_NEON_V8HI_LDST, ARM_HAVE_NEON_V2SI_LDST): Likewise.
25210 (ARM_HAVE_NEON_V4SI_LDST, ARM_HAVE_NEON_V4HF_LDST): Likewise.
25211 (ARM_HAVE_NEON_V8HF_LDST, ARM_HAVE_NEON_V4BF_LDST): Likewise.
25212 (ARM_HAVE_NEON_V8BF_LDST, ARM_HAVE_NEON_V2SF_LDST): Likewise.
25213 (ARM_HAVE_NEON_V4SF_LDST, ARM_HAVE_NEON_DI_LDST): Likewise.
25214 (ARM_HAVE_NEON_V2DI_LDST): Likewise.
25215 (ARM_HAVE_V8QI_LDST, ARM_HAVE_V16QI_LDST): Likewise.
25216 (ARM_HAVE_V4HI_LDST, ARM_HAVE_V8HI_LDST): Likewise.
25217 (ARM_HAVE_V2SI_LDST, ARM_HAVE_V4SI_LDST, ARM_HAVE_V4HF_LDST): Likewise.
25218 (ARM_HAVE_V8HF_LDST, ARM_HAVE_V4BF_LDST, ARM_HAVE_V8BF_LDST): Likewise.
25219 (ARM_HAVE_V2SF_LDST, ARM_HAVE_V4SF_LDST, ARM_HAVE_DI_LDST): Likewise.
25220 (ARM_HAVE_V2DI_LDST): Likewise.
25221 * config/arm/mve.md (*movmisalign<mode>_mve_store): New pattern.
25222 (*movmisalign<mode>_mve_load): New pattern.
25223 * config/arm/neon.md (movmisalign<mode>): Move to ...
25224 * config/arm/vec-common.md: ... here.
25226 2021-01-12 Vladimir N. Makarov <vmakarov@redhat.com>
25229 * lra-eliminations.c (eliminate_regs_in_insn): Add transformation
25230 of pattern 'plus (plus (hard reg, const), pseudo)'.
25232 2021-01-12 Richard Biener <rguenther@suse.de>
25234 PR tree-optimization/98550
25235 * tree-vect-slp.c (vect_record_max_nunits): Check whether
25236 the group size is a multiple of the vector element count.
25237 (vect_build_slp_tree_1): When we need to fail because
25238 the vector type choosen causes unrolling do so lazily
25239 without affecting matches only at the end to guide group splitting.
25241 2021-01-12 Martin Liska <mliska@suse.cz>
25244 * optc-save-gen.awk: Compare also n_target_save vars with
25247 2021-01-12 Martin Liska <mliska@suse.cz>
25249 * gcov.c (source_info::debug): New.
25250 (print_usage): Add --debug (-D) option.
25251 (process_args): Likewise.
25252 (generate_results): Call src->debug after
25253 accumulate_line_counts.
25254 (read_graph_file): Properly assign id for EXIT_BLOCK.
25255 * profile.c (branch_prob): Dump function body before it is
25258 2021-01-12 Jakub Jelinek <jakub@redhat.com>
25260 PR tree-optimization/98629
25261 * tree-ssa-math-opts.c (arith_overflow_check_p): Don't update use_stmt
25262 unless returning non-zero.
25264 2021-01-12 Jakub Jelinek <jakub@redhat.com>
25266 PR tree-optimization/95731
25267 * tree-ssa-reassoc.c (optimize_range_tests_cmp_bitwise): Also optimize
25268 x < 0 && y < 0 && z < 0 into (x | y | z) < 0 for signed x, y, z.
25269 (optimize_range_tests): Call optimize_range_tests_cmp_bitwise
25270 only after optimize_range_tests_var_bound.
25272 2021-01-12 Jakub Jelinek <jakub@redhat.com>
25274 * configure.ac: Ensure c/Make-lang.in comes first in @all_lang_makefrags@.
25275 * configure: Regenerated.
25277 2021-01-12 liuhongt <hongtao.liu@intel.com>
25280 * config/i386/i386-builtins.h (BUILTIN_DESC_SWAP_OPERANDS):
25282 * config/i386/i386-expand.c (ix86_expand_sse_comi): Delete
25285 2021-01-12 Alexandre Oliva <oliva@adacore.com>
25287 * ssa-iterators.h (end_imm_use_stmt_traverse): Forward
25289 (auto_end_imm_use_stmt_traverse): New struct.
25290 (FOR_EACH_IMM_USE_STMT): Use it.
25291 (BREAK_FROM_IMM_USE_STMT, RETURN_FROM_IMM_USE_STMT): Remove,
25293 * gimple-ssa-strength-reduction.c: ... here, ...
25294 * graphite-scop-detection.c: ... here, ...
25295 * ipa-modref.c, ipa-pure-const.c, ipa-sra.c: ... here, ...
25296 * tree-predcom.c, tree-ssa-ccp.c: ... here, ...
25297 * tree-ssa-dce.c, tree-ssa-dse.c: ... here, ...
25298 * tree-ssa-loop-ivopts.c, tree-ssa-math-opts.c: ... here, ...
25299 * tree-ssa-phiprop.c, tree-ssa.c: ... here, ...
25300 * tree-vect-slp.c: ... and here, ...
25301 * doc/tree-ssa.texi: ... and the example here.
25303 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25305 * config/aarch64/aarch64-sve.md (sdiv_pow2<mode>3): Extend from
25306 SVE_FULL_I to SVE_I. Generate an UNSPEC_PRED_X.
25307 (*sdiv_pow2<mode>3): New pattern.
25308 (@cond_<sve_int_op><mode>): Extend from SVE_FULL_I to SVE_I.
25309 Wrap the ASRD in an UNSPEC_PRED_X.
25310 (*cond_<sve_int_op><mode>_2): Likewise. Replace the UNSPEC_PRED_X
25311 predicate with a constant PTRUE, if it isn't already.
25312 (*cond_<sve_int_op><mode>_z): Replace with...
25313 (*cond_<sve_int_op><mode>_any): ...this new pattern.
25315 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25317 * config/aarch64/aarch64-sve.md (*cond_bic<mode>_2): Extend from
25318 SVE_FULL_I to SVE_I.
25319 (*cond_bic<mode>_any): Likewise.
25321 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25323 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart)
25324 (@aarch64_pred_<MUL_HIGHPART:optab><mode>): Extend from SVE_FULL_I
25327 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25329 * config/aarch64/aarch64-sve.md (<su>abd<mode>_3): Extend from
25330 SVE_FULL_I to SVE_I.
25331 (*aarch64_cond_<su>abd<mode>_2): Likewise.
25332 (*aarch64_cond_<su>abd<mode>_any): Likewise.
25333 (@aarch64_pred_<su>abd<mode>): Likewise. Use UNSPEC_PRED_X
25334 for the max and min but not for the minus.
25335 (*aarch64_cond_<su>abd<mode>_3): New pattern.
25337 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25339 * config/aarch64/iterators.md (SVE_24I): New iterator.
25340 * config/aarch64/aarch64-sve.md (*aarch64_adr<mode>_shift): Extend from
25341 SVE_FULL_SDI to SVE_24I. Use containers rather than elements.
25343 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25345 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_BINARY:optab><mode>)
25346 (*cond_<SVE_INT_BINARY:optab><mode>_2): Extend from SVE_FULL_I
25348 (*cond_<SVE_INT_BINARY:optab><mode>_3): Likewise.
25349 (*cond_<SVE_INT_BINARY:optab><mode>_any): Likewise.
25350 (*cond_<SVE_INT_BINARY:optab><mode>_2_const): Likewise.
25351 (*cond_<SVE_INT_BINARY:optab><mode>_any_const): Likewise.
25353 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25355 * config/aarch64/aarch64-sve.md (<SVE_INT_BINARY_IMM:optab><mode>3)
25356 (@aarch64_pred_<SVE_INT_BINARY_IMM:optab><mode>)
25357 (*post_ra_<SVE_INT_BINARY_IMM:optab><mode>3): Extend from SVE_FULL_I
25360 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
25362 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3)
25363 (v<ASHIFT:optab><mode>3, @aarch64_pred_<optab><mode>)
25364 (*post_ra_v<ASHIFT:optab><mode>3): Extend from SVE_FULL_I to SVE_I.
25366 2021-01-11 Martin Liska <mliska@suse.cz>
25369 * symtab-clones.h (clone_info::release): Release
25370 symtab::m_clones with ggc_delete as it's a GGC memory.
25372 2021-01-11 Matthias Klose <doko@ubuntu.com>
25374 * Makefile.in (LINK_PROGRESS): Show the link target.
25376 2021-01-11 Richard Biener <rguenther@suse.de>
25378 PR tree-optimization/91403
25379 * tree-vect-data-refs.c (vect_analyze_group_access_1): Cap
25380 single-element interleaving group size at 4096 elements.
25382 2021-01-11 Richard Biener <rguenther@suse.de>
25384 PR tree-optimization/98526
25385 * tree-vect-loop.c (vect_model_reduction_cost): Remove costing
25386 of the actual reduction op for the regular case.
25387 (vectorizable_reduction): Cost the stmts
25388 vect_transform_reduction produces here.
25390 2021-01-11 Andreas Krebbel <krebbel@linux.ibm.com>
25392 * tree-ssa-forwprop.c (simplify_vector_constructor): For
25393 big-endian, use UNPACK[_FLOAT]_HI.
25395 2021-01-11 Tamar Christina <tamar.christina@arm.com>
25397 * tree-vect-slp-patterns.c (class complex_pattern,
25398 class complex_add_pattern): Add parameters to matches.
25399 (complex_add_pattern::build): Free memory.
25400 (complex_add_pattern::matches): Move validation end of match.
25401 (complex_add_pattern::recognize): Likewise.
25403 2021-01-11 Tamar Christina <tamar.christina@arm.com>
25405 * tree-vect-slp-patterns.c (linear_loads_p): Fix externals.
25407 2021-01-11 Tamar Christina <tamar.christina@arm.com>
25409 * tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
25411 2021-01-11 Jakub Jelinek <jakub@redhat.com>
25413 PR tree-optimization/95867
25414 * tree-ssa-math-opts.h: New header.
25415 * tree-ssa-math-opts.c: Include tree-ssa-math-opts.h.
25416 (powi_as_mults): No longer static. Use build_one_cst instead of
25417 build_real. Formatting fix.
25418 * tree-ssa-reassoc.c: Include tree-ssa-math-opts.h.
25419 (attempt_builtin_powi): Handle multiplication reassociation without
25420 powi_fndecl using powi_as_mults.
25421 (reassociate_bb): For integral types don't require
25422 -funsafe-math-optimizations to call attempt_builtin_powi.
25424 2021-01-11 Jakub Jelinek <jakub@redhat.com>
25426 PR tree-optimization/95852
25427 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): Change
25428 mul_stmts parameter type to vec<gimple *> &. Before cond_stmt
25429 allow in the bb any of the stmts in that vector, div_stmt and
25430 up to 3 cast stmts.
25431 (arith_cast_equal_p): New function.
25432 (arith_overflow_check_p): Add cast_stmt argument, handle signed
25433 multiply overflow checks.
25434 (match_arith_overflow): Adjust caller. Handle signed multiply
25437 2021-01-11 Jakub Jelinek <jakub@redhat.com>
25439 PR tree-optimization/95852
25440 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): New function.
25441 (uaddsub_overflow_check_p): Renamed to ...
25442 (arith_overflow_check_p): ... this. Handle also multiplication
25443 with overflow check.
25444 (match_uaddsub_overflow): Renamed to ...
25445 (match_arith_overflow): ... this. Add cfg_changed argument. Handle
25446 also multiplication with overflow check. Adjust function comment.
25447 (math_opts_dom_walker::after_dom_children): Adjust callers. Call
25448 match_arith_overflow also for MULT_EXPR.
25450 2021-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25452 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
25453 __builtin_convertvector.
25454 (vmovl_s16): Likewise.
25455 (vmovl_s32): Likewise.
25456 (vmovl_u8): Likewise.
25457 (vmovl_u16): Likewise.
25458 (vmovl_u32): Likewise.
25459 (vmovn_s16): Likewise.
25460 (vmovn_s32): Likewise.
25461 (vmovn_s64): Likewise.
25462 (vmovn_u16): Likewise.
25463 (vmovn_u32): Likewise.
25464 (vmovn_u64): Likewise.
25466 2021-01-11 Martin Liska <mliska@suse.cz>
25468 * gimple-if-to-switch.cc (struct condition_info): Use auto_var.
25469 (if_chain::is_beneficial): Delete clusters
25470 (find_conditions): Make second argument of conditions_in_bbs a
25471 pointer so that we control over it's lifetime.
25472 (pass_if_to_switch::execute): Delete them.
25474 2021-01-11 Kewen Lin <linkw@linux.ibm.com>
25476 * ira.c (move_unallocated_pseudos): Check other_reg and skip if
25479 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
25481 * config/vax/vax.md (cc): Remove mode attribute.
25482 (subst_<cc>, subst_f<cc>): Rename to...
25483 (subst_<mode>, subst_f<VAXccnz:mode>): ... these respectively.
25484 (*cbranch<VAXint:mode>4_<VAXcc:mode>): Update for `cc' removal.
25485 (*cbranch<VAXfp:mode>4_<VAXccnz:mode>): Likewise.
25486 (*branch_<mode>, *branch_<mode>_reversed): Likewise.
25488 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
25490 * config/vax/vax.md (subst_f<cc>): Add mode to operands and
25491 `const_double_zero'.
25493 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
25495 * config/pdp11/pdp11.md (PDPfp): New mode iterator.
25496 (fcc_cc, fcc_ccnz): Use it. Add mode to `const_double_zero' and
25499 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
25501 * genemit.c (gen_exp) <CONST_DOUBLE>: Handle `const_double_zero'
25503 * read-rtl.c (rtx_reader::read_rtx_code): Handle machine mode
25504 with `const_double_zero'.
25505 * doc/rtl.texi (Constant Expression Types): Document it.
25507 2021-01-09 Jakub Jelinek <jakub@redhat.com>
25510 * tree-cfg.c (verify_gimple_assign_binary): Allow lhs of
25511 POINTER_DIFF_EXPR to be any integral type.
25513 2021-01-09 Jakub Jelinek <jakub@redhat.com>
25515 PR rtl-optimization/98603
25516 * function.c (instantiate_virtual_regs_in_insn): For asm goto
25517 with impossible constraints, drop all SETs, CLOBBERs, drop PARALLEL
25518 if any, set ASM_OPERANDS mode to VOIDmode and change
25519 ASM_OPERANDS_OUTPUT_CONSTRAINT and ASM_OPERANDS_OUTPUT_IDX.
25521 2021-01-09 Alexandre Oliva <oliva@gnu.org>
25524 * final.c (notice_source_line): Narrow down the condition to
25525 skip a line-0 marker.
25527 2021-01-08 Sergei Trofimovich <siarheit@google.com>
25529 * ipa-modref.c (merge_call_side_effects): Fix
25530 linebreak split by reordering two print calls.
25532 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
25534 * config/s390/vector.md (*tf_to_fprx2_0): Rename from
25535 "*mov_tf_to_fprx2_0" for consistency, fix constraint.
25536 (*tf_to_fprx2_1): Rename from "*mov_tf_to_fprx2_1" for
25537 consistency, fix constraint.
25539 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
25541 * config/s390/s390-c.c (s390_def_or_undef_macro): Accept
25542 callables instead of mask values.
25543 (struct target_flag_set_p): New predicate.
25544 (s390_cpu_cpp_builtins_internal): Define or undefine
25545 __LONG_DOUBLE_VX__ macro.
25547 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
25550 * config/i386/i386.c (x86_function_profiler): Use R10 and R11
25551 to call mcount in large model with PIC for NO_PROFILE_COUNTERS
25554 2021-01-08 Richard Biener <rguenther@suse.de>
25556 * tree-ssa-sccvn.c (pass_fre::execute): Reset the SCEV hash table.
25558 2021-01-08 Richard Biener <rguenther@suse.de>
25560 * tree-vect-slp.c (scalar_stmts_to_slp_tree_map_t): Fix.
25561 (vect_build_slp_tree): On cache hit release the matched
25562 scalar stmts vector.
25563 * tree-vect-stmts.c (vectorizable_store): Properly free
25564 vec_oprnds before possibly gathering them again.
25566 2021-01-08 Richard Biener <rguenther@suse.de>
25568 PR tree-optimization/98544
25569 * tree-vect-slp.c (vect_optimize_slp): Always materialize
25570 permutes at a permute node.
25572 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
25575 * config/i386/i386.c (x86_function_profiler): Use R10 to call
25576 mcount in large model. Sorry for large model with PIC.
25578 2021-01-08 Jakub Jelinek <jakub@redhat.com>
25581 * config/i386/i386.opt (ix86_cmodel, ix86_incoming_stack_boundary_arg,
25582 ix86_pmode, ix86_preferred_stack_boundary_arg, ix86_regparm,
25583 ix86_veclibabi_type): Remove x_ prefix, use TargetVariable instead of
25584 TargetSave and initialize for variables with enum types.
25585 (mfentry, mstack-protector-guard-reg=, mstack-protector-guard-offset=,
25586 mstack-protector-guard-symbol=): Add Save.
25587 * config/i386/i386-options.c (ix86_function_specific_save,
25588 ix86_function_specific_restore): Don't save or restore x_ix86_cmodel,
25589 x_ix86_incoming_stack_boundary_arg, x_ix86_pmode,
25590 x_ix86_preferred_stack_boundary_arg, x_ix86_regparm,
25591 x_ix86_veclibabi_type.
25593 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
25595 * config/aarch64/aarch64-sve.md (*cnot<mode>): Extend from
25596 SVE_FULL_I to SVE_I.
25597 (*cond_cnot<mode>_2, *cond_cnot<mode>_any): Likewise.
25599 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
25601 * config/aarch64/aarch64-sve.md (*cond_uxt<mode>_2): Extend from
25602 SVE_FULL_I to SVE_I.
25603 (*cond_uxt<mode>_any): Likewise.
25605 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25607 * config/aarch64/iterators.md (Vwhalf): New iterator.
25608 * config/aarch64/aarch64-simd.md (aarch64_<sur>adalp<mode>_3):
25610 (aarch64_<sur>adalp<mode>): ... This. Make more
25612 (<sur>sadv16qi): Adjust callsite of the above.
25613 * config/aarch64/aarch64-simd-builtins.def (sadalp, uadalp): New
25615 * config/aarch64/arm_neon.h (vpadal_s8): Reimplement using
25617 (vpadal_s16): Likewise.
25618 (vpadal_u8): Likewise.
25619 (vpadal_u16): Likewise.
25620 (vpadalq_s8): Likewise.
25621 (vpadalq_s16): Likewise.
25622 (vpadalq_s32): Likewise.
25623 (vpadalq_u8): Likewise.
25624 (vpadalq_u16): Likewise.
25625 (vpadalq_u32): Likewise.
25627 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25629 * config/aarch64/aarch64-simd.md (aarch64_<su>abd<mode>_3):
25631 (aarch64_<su>abd<mode>): ... This.
25632 (<sur>sadv16qi): Adjust callsite of the above.
25633 * config/aarch64/aarch64-simd-builtins.def (sabd, uabd): Define
25635 * config/aarch64/arm_neon.h (vabd_s8): Reimplement using
25637 (vabd_s16): Likewise.
25638 (vabd_s32): Likewise.
25639 (vabd_u8): Likewise.
25640 (vabd_u16): Likewise.
25641 (vabd_u32): Likewise.
25642 (vabdq_s8): Likewise.
25643 (vabdq_s16): Likewise.
25644 (vabdq_s32): Likewise.
25645 (vabdq_u8): Likewise.
25646 (vabdq_u16): Likewise.
25647 (vabdq_u32): Likewise.
25649 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25651 * config/aarch64/aarch64-simd-builtins.def (saba, uaba): Define
25653 * config/aarch64/arm_neon.h (vaba_s8): Implement using builtin.
25654 (vaba_s16): Likewise.
25655 (vaba_s32): Likewise.
25656 (vaba_u8): Likewise.
25657 (vaba_u16): Likewise.
25658 (vaba_u32): Likewise.
25659 (vabaq_s8): Likewise.
25660 (vabaq_s16): Likewise.
25661 (vabaq_s32): Likewise.
25662 (vabaq_u8): Likewise.
25663 (vabaq_u16): Likewise.
25664 (vabaq_u32): Likewise.
25666 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25668 * config/aarch64/aarch64-simd.md (aba<mode>_3): Rename to...
25669 (aarch64_<su>aba<mode>): ... This. Handle uaba as well.
25670 Change RTL pattern to match.
25672 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
25674 * common/config/riscv/riscv-common.c (riscv_current_subset_list): New.
25675 * config/riscv/riscv-c.c (riscv-subset.h): New.
25676 (INCLUDE_STRING): Define.
25677 (riscv_cpu_cpp_builtins): Add new style architecture extension
25679 * config/riscv/riscv-subset.h (riscv_subset_list::begin): New.
25680 (riscv_subset_list::end): New.
25681 (riscv_current_subset_list): New.
25683 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
25685 * common/config/riscv/riscv-common.c (RISCV_DONT_CARE_VERSION):
25686 Move to riscv-subset.h.
25687 (struct riscv_subset_t): Ditto.
25688 (class riscv_subset_list): Ditto.
25689 * config/riscv/riscv-subset.h (RISCV_DONT_CARE_VERSION): Move
25690 from riscv-common.c.
25691 (struct riscv_subset_t): Ditto.
25692 (class riscv_subset_list): Ditto.
25693 * config/riscv/t-riscv ($(common_out_file)): Add file
25696 2021-01-07 Jakub Jelinek <jakub@redhat.com>
25699 * config/i386/i386.md (*bmi_blsi_<mode>_cmp, *bmi_blsi_<mode>_ccno):
25700 New define_insn patterns.
25702 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
25704 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_UNARY:optab><mode>)
25705 (*cond_<SVE_INT_UNARY:optab><mode>_2): Extend from SVE_FULL_I to SVE_I.
25706 (*cond_<SVE_INT_UNARY:optab><mode>_any): Likewise.
25708 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
25710 PR tree-optimization/98560
25711 * internal-fn.def (IFN_VCONDU, IFN_VCONDEQ): Use type vec_cond.
25712 * internal-fn.c (vec_cond_mask_direct): Get the data mode from
25714 (vec_cond_direct): Likewise argument 2.
25715 (vec_condu_direct, vec_condeq_direct): Delete.
25716 (expand_vect_cond_optab_fn): Rename to...
25717 (expand_vec_cond_optab_fn): ...this, replacing old macro.
25718 (expand_vec_condu_optab_fn, expand_vec_condeq_optab_fn): Delete.
25719 (expand_vect_cond_mask_optab_fn): Rename to...
25720 (expand_vec_cond_mask_optab_fn): ...this, replacing old macro.
25721 (direct_vec_cond_mask_optab_supported_p): Treat the optab as a
25723 (direct_vec_cond_optab_supported_p): Likewise.
25724 (direct_vec_condu_optab_supported_p): Delete.
25725 (direct_vec_condeq_optab_supported_p): Delete.
25726 * gimple-isel.cc: Include internal-fn.h.
25727 (gimple_expand_vec_cond_expr): Check that IFN_VCONDEQ is supported
25730 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
25732 PR tree-optimization/98560
25733 * gimple-isel.cc (gimple_expand_vec_cond_expr): If we fail to use
25734 IFN_VCOND{,U,EQ}, fall back on IFN_VCOND_MASK.
25736 2021-01-07 Uroš Bizjak <ubizjak@gmail.com>
25738 * config/i386/i386.md (insn): Merge from plusminus_insn, shift_insn,
25739 rotate_insn and optab code attributes.
25740 Update all uses to merged code attribute.
25741 * config/i386/sse.md: Update all uses to merged code attribute.
25742 * config/i386/mmx.md: Update all uses to merged code attribute.
25744 2021-01-07 Jakub Jelinek <jakub@redhat.com>
25746 PR tree-optimization/98568
25747 * gimple-ssa-store-merging.c (bswap_view_convert): New function.
25748 (bswap_replace): Use it.
25750 2021-01-06 Vladimir N. Makarov <vmakarov@redhat.com>
25752 PR rtl-optimization/97978
25753 * lra-int.h (lra_hard_reg_split_p): New external.
25754 * lra.c (lra_hard_reg_split_p): New global.
25755 (lra): Set up lra_hard_reg_split_p after splitting a hard reg.
25756 * lra-assigns.c (lra_assign): Don't check allocation correctness
25757 after hard reg splitting.
25759 2021-01-06 Martin Sebor <msebor@redhat.com>
25762 * builtins.c (new_delete_mismatch_p): New overload.
25763 (new_delete_mismatch_p (tree, tree)): Call it.
25765 2021-01-06 Alexandre Oliva <oliva@adacore.com>
25767 * Makefile.in (T_GLIMITS_H): New.
25768 (stmp-int-hdrs): Depend on it, use it.
25769 * config/t-vxworks (T_GLIMITS_H): Override it.
25770 (vxw-glimits.h): New.
25772 2021-01-06 Richard Biener <rguenther@suse.de>
25774 PR tree-optimization/98513
25775 * value-range.cc (intersect_ranges): Compare the upper bounds
25776 for the expected relation.
25778 2021-01-06 Gerald Pfeifer <gerald@pfeifer.com>
25781 2020-12-28 Gerald Pfeifer <gerald@pfeifer.com>
25783 * doc/standards.texi (HSAIL): Remove section.
25785 2021-01-05 Samuel Thibault <samuel.thibault@ens-lyon.org>
25787 * configure: Re-generate.
25789 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25791 * doc/invoke.texi (-std=c++20): Adjust for the publication of
25792 ISO 14882:2020 standard.
25793 * doc/standards.texi: Likewise.
25795 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25797 PR tree-optimization/94802
25798 * expr.h (maybe_optimize_sub_cmp_0): Declare.
25799 * expr.c: Include tree-pretty-print.h and flags.h.
25800 (maybe_optimize_sub_cmp_0): New function.
25801 (do_store_flag): Use it.
25802 * cfgexpand.c (expand_gimple_cond): Likewise.
25804 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
25806 * mux-utils.h (pointer_mux::m_ptr): Tweak description of contents.
25807 * rtlanal.c (simple_regno_set): Tweak description to clarify the
25810 2021-01-05 Richard Biener <rguenther@suse.de>
25812 PR tree-optimization/98516
25813 * tree-vect-slp.c (vect_optimize_slp): Permute the incoming
25814 lanes when materializing on a VEC_PERM node.
25815 (vectorizable_slp_permutation): Dump the permute properly.
25817 2021-01-05 Richard Biener <rguenther@suse.de>
25819 * tree-vect-slp.c (vect_slp_region): Move debug counter
25820 to cover individual subgraphs.
25822 2021-01-05 Richard Biener <rguenther@suse.de>
25824 PR tree-optimization/98428
25825 * tree-vect-slp.c (vect_build_slp_tree_1): Properly reject
25826 vector lane extracts for loop vectorization.
25828 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25830 PR tree-optimization/98514
25831 * tree-ssa-reassoc.c (bb_rank): Change type from long * to
25833 (operand_rank): Change type from hash_map<tree, long> to
25834 hash_map<tree, int64_t>.
25835 (phi_rank): Change return type from long to int64_t.
25836 (loop_carried_phi): Change block_rank variable type from long to
25838 (propagate_rank): Change return type, rank parameter type and
25839 op_rank variable type from long to int64_t.
25840 (find_operand_rank): Change return type from long to int64_t
25841 and change slot variable type from long * to int64_t *.
25842 (insert_operand_rank): Change rank parameter type from long to
25844 (get_rank): Change return type and rank variable type from long to
25845 int64_t. Use PRId64 instead of ld to print the rank.
25846 (init_reassoc): Change rank variable type from long to int64_t
25847 and adjust correspondingly bb_rank and operand_rank initialization.
25849 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25851 PR tree-optimization/96928
25852 * tree-ssa-phiopt.c (xor_replacement): New function.
25853 (tree_ssa_phiopt_worker): Call it.
25855 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25857 PR tree-optimization/96930
25858 * match.pd ((A / (1 << B)) -> (A >> B)): If A is extended
25859 from narrower value which has the same type as 1 << B, perform
25860 the right shift on the narrower value followed by extension.
25862 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25864 PR tree-optimization/96239
25865 * gimple-ssa-store-merging.c (maybe_optimize_vector_constructor): New
25867 (get_status_for_store_merging): Don't return BB_INVALID for blocks
25868 with potential bswap optimizable CONSTRUCTORs.
25869 (pass_store_merging::execute): Optimize vector CONSTRUCTORs with bswap
25872 2021-01-05 Richard Biener <rguenther@suse.de>
25874 PR tree-optimization/98381
25875 * tree.c (vector_element_bits): Properly compute bool vector
25877 * tree-vect-loop.c (vectorizable_live_operation): Properly
25878 compute the last lane bit offset.
25880 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
25883 * config/i386/sse.md (sse_cvtps2pi): Redefine as define_insn_and_split.
25884 Clear the top 64 bytes of the input XMM register.
25885 (sse_cvttps2pi): Ditto.
25887 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
25890 * config/i386/xopintrin.h (_mm256_cmov_si256): New.
25892 2021-01-05 H.J. Lu <hjl.tools@gmail.com>
25895 * config/i386/xmmintrin.h (_mm_extract_pi16): Cast to unsigned
25898 2021-01-05 Claudiu Zissulescu <claziss@synopsys.com>
25900 * config/arc/arc.md (maddsidi4_split): Use ACC_REG_FIRST.
25901 (umaddsidi4_split): Likewise.
25903 2021-01-05 liuhongt <hongtao.liu@intel.com>
25906 * config/i386/sse.md (*sse2_pmovskb_zexthisi): New
25907 define_insn_and_split for zero_extend of subreg HI of pmovskb
25909 (*sse2_pmovskb_zexthisi): Add new combine splitters for
25910 zero_extend of not of subreg HI of pmovskb result.
25912 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
25915 * explow.c (convert_memory_address_addr_space_1): Handle UNSPECs
25917 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate): Use
25918 convert_memory_address to convert symbolic immediates to ptr_mode
25919 before forcing them to memory.
25921 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
25923 PR rtl-optimization/97144
25924 * recog.c (constrain_operands): Initialize matching_operand
25925 for each alternative, rather than only doing it once.
25927 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
25929 PR rtl-optimization/98403
25930 * rtl-ssa/changes.cc (function_info::finalize_new_accesses): Explain
25931 why we don't remove call clobbers.
25932 (function_info::apply_changes_to_insn): Don't attempt to add
25933 call clobbers here.
25935 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
25937 PR tree-optimization/98371
25938 * tree-vect-loop.c (vect_reanalyze_as_main_loop): New function.
25939 (vect_analyze_loop): If an epilogue loop appears to be cheaper
25940 than the main loop, re-analyze it as a main loop before adopting
25943 2021-01-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
25946 * configure.ac (NETLIBS): Determine using AX_LIB_SOCKET_NSL.
25947 * aclocal.m4, configure: Regenerate.
25948 * Makefile.in (NETLIBS): Define.
25949 (BACKEND): Remove $(CODYLIB).
25951 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25953 PR rtl-optimization/98334
25954 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
25955 Optimize (X - 1) * Y + Y to X * Y or (X + 1) * Y - Y to X * Y.
25957 2021-01-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
25959 * tree-inline.c (expand_call_inline): Restore input_location.
25960 Return result from recursive call.
25962 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
25964 PR tree-optimization/95401
25965 * config/aarch64/aarch64-sve-builtins.cc
25966 (gimple_folder::load_store_cookie): Use bits rather than bytes
25967 for the alignment argument to IFN_MASK_LOAD and IFN_MASK_STORE.
25968 * gimple-fold.c (gimple_fold_mask_load_store_mem_ref): Likewise.
25969 * tree-vect-stmts.c (vectorizable_store): Likewise.
25970 (vectorizable_load): Likewise.
25972 2021-01-04 Richard Biener <rguenther@suse.de>
25974 PR tree-optimization/98308
25975 * tree-vect-stmts.c (vectorizable_load): Set invariant mask
25978 2021-01-04 Jakub Jelinek <jakub@redhat.com>
25980 PR tree-optimization/95771
25981 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Handle types
25982 with precision smaller than int's precision and types with precision
25983 twice as large as long long. Formatting fixes.
25985 2021-01-04 Richard Biener <rguenther@suse.de>
25987 PR tree-optimization/98464
25988 * tree-ssa-sccvn.c (vn_valueize_for_srt): Rename from ...
25989 (vn_valueize_wrapper): ... this. Temporarily adjust vn_context_bb.
25990 (process_bb): Adjust.
25992 2021-01-04 Matthew Malcomson <matthew.malcomson@arm.com>
25995 * doc/invoke.texi (-fsanitize=address): Fix wording describing
25996 clash with -fsanitize=hwaddress.
25998 2021-01-04 Richard Biener <rguenther@suse.de>
26000 PR tree-optimization/98282
26001 * tree-ssa-sccvn.c (vn_get_stmt_kind): Classify tcc_reference on
26002 invariants as VN_NARY.
26004 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
26007 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Accept
26008 aarch64_simd_reg_or_zero for operand 2. Use the combinez patterns
26009 to handle zero operands.
26011 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
26013 * config/aarch64/aarch64.c (offset_6bit_signed_scaled_p): New function.
26014 (offset_6bit_unsigned_scaled_p): Fix typo in comment.
26015 (aarch64_sve_prefetch_operand_p): Accept MUL VLs in the range
26018 2021-01-04 Richard Biener <rguenther@suse.de>
26020 PR tree-optimization/98393
26021 * tree-vect-slp.c (vect_build_slp_tree): Properly zero matches
26022 when hitting the limit.
26024 2021-01-04 Richard Biener <rguenther@suse.de>
26026 PR tree-optimization/98291
26027 * tree-vect-loop.c (vectorizable_reduction): Bypass
26028 associativity check for SLP reductions with VF 1.
26030 2021-01-04 Jakub Jelinek <jakub@redhat.com>
26032 PR tree-optimization/96782
26033 * match.pd (x == ~x -> false, x != ~x -> true): New simplifications.
26035 2021-01-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
26037 * collect-utils.c (collect_execute): Check dumppfx.
26038 * collect2.c (maybe_run_lto_and_relink, do_link): Pass atsuffix
26039 to collect_execute.
26040 (do_link): Add new parameter atsuffix.
26041 (main): Handle -dumpdir option. Skip one argument for
26042 -o, -isystem and -B options.
26043 * gcc.c (make_at_file): New helper function.
26044 (close_at_file): Use it.
26046 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26048 * config/darwin.h (MIN_LD64_NO_COAL_SECTS): Adjust.
26049 Amend handling for LD64_VERSION fallback defaults.
26051 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26053 * config.gcc: Compute default version information
26054 from the configured target. Likewise defaults for
26056 * config/darwin10.h: Removed.
26057 * config/darwin12.h: Removed.
26058 * config/darwin9.h: Removed.
26059 * config/rs6000/darwin8.h: Removed.
26061 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26063 * config/darwin9.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
26065 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26067 * config/darwin9.h (STACK_CHECK_STATIC_BUILTIN): Move from here..
26068 * config/darwin.h (STACK_CHECK_STATIC_BUILTIN): .. to here.
26070 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26072 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move from
26074 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): ... to here.
26076 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26078 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move the spec
26079 for the Darwin10 unwinder stub from here ...
26080 * config/darwin.h (LINK_COMMAND_SPEC_A): ... to here.
26082 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
26084 * config/darwin.h (DSYMUTIL_SPEC): Default to DWARF
26085 (ASM_DEBUG_SPEC):Only define if the assembler supports
26087 (PREFERRED_DEBUGGING_TYPE): Default to DWARF.
26088 (DARWIN_PREFER_DWARF): Define.
26089 * config/darwin9.h (PREFERRED_DEBUGGING_TYPE): Remove.
26090 (DARWIN_PREFER_DWARF): Likewise
26091 (DSYMUTIL_SPEC): Likewise.
26092 (COLLECT_RUN_DSYMUTIL): Likewise.
26093 (ASM_DEBUG_SPEC): Likewise.
26094 (ASM_DEBUG_OPTION_SPEC): Likewise.
26096 2021-01-02 Jan Hubicka <jh@suse.cz>
26098 * cfg.c (free_block): ggc_free bb.
26100 2021-01-01 Jakub Jelinek <jakub@redhat.com>
26102 * gcc.c (process_command): Update copyright notice dates.
26103 * gcov-dump.c (print_version): Ditto.
26104 * gcov.c (print_version): Ditto.
26105 * gcov-tool.c (print_version): Ditto.
26106 * gengtype.c (create_file): Ditto.
26107 * doc/cpp.texi: Bump @copying's copyright year.
26108 * doc/cppinternals.texi: Ditto.
26109 * doc/gcc.texi: Ditto.
26110 * doc/gccint.texi: Ditto.
26111 * doc/gcov.texi: Ditto.
26112 * doc/install.texi: Ditto.
26113 * doc/invoke.texi: Ditto.
26115 2021-01-01 Jakub Jelinek <jakub@redhat.com>
26117 * ChangeLog-2020: Rotate ChangeLog. New file.
26120 Copyright (C) 2021 Free Software Foundation, Inc.
26122 Copying and distribution of this file, with or without modification,
26123 are permitted in any medium without royalty provided the copyright
26124 notice and this notice are preserved.