1 2021-11-12 Stafford Horne <shorne@gmail.com>
3 * config/or1k/or1k-protos.h (or1k_profile_hook): New function.
4 * config/or1k/or1k.h (PROFILE_HOOK): Change macro to reference
5 new function or1k_profile_hook.
6 * config/or1k/or1k.c (struct machine_function): Add new field
8 (or1k_profile_hook): New function.
9 (or1k_init_pic_reg): Update to inject pic rtx after _mcount arg
11 (or1k_frame_pointer_required): Frame pointer no longer needed
14 2021-11-12 Jan Hubicka <jh@suse.cz>
16 PR tree-optimization/103209
17 * tree-ssa-structalias.c (find_func_aliases_for_call): Fix
18 use of handle_rhs_call
20 2021-11-12 Aldy Hernandez <aldyh@redhat.com>
22 PR tree-optimization/103202
23 * gimple-range-path.cc
24 (path_range_query::compute_ranges_in_block): Solve PHI imports first.
26 2021-11-12 Jan Hubicka <jh@suse.cz>
28 * ipa-pure-const.c (propagate_pure_const): Remove redundant check;
29 fix call of ipa_make_function_const and ipa_make_function_pure.
31 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
33 * config/aarch64/aarch64.c (aarch64_vector_op::n_advsimd_ops): Delete.
34 (aarch64_vector_op::m_seen_loads): Likewise.
35 (aarch64_vector_costs::aarch64_vector_costs): Don't push to
37 (aarch64_vector_op::count_ops): Remove vectype and factor parameters.
38 Remove code that tries to predict different vec_flags from the
40 (aarch64_vector_costs::add_stmt_cost): Update accordingly.
41 Remove m_advsimd_ops handling.
43 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
45 * config/aarch64/aarch64.c (aarch64_vector_costs::m_saw_sve_only_op)
46 (aarch64_sve_only_stmt_p): Delete.
47 (aarch64_vector_costs::prefer_unrolled_loop): New function,
48 extracted from adjust_body_cost.
49 (aarch64_vector_costs::better_main_loop_than_p): New function,
50 using heuristics extracted from adjust_body_cost and
52 (aarch64_vector_costs::adjust_body_cost_sve): Remove
53 advsimd_cycles_per_iter and could_use_advsimd parameters.
54 Update after changes above.
55 (aarch64_vector_costs::adjust_body_cost): Update after changes above.
57 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
59 * config/aarch64/aarch64.c (aarch64_vec_op_count::m_vf_factor):
61 (aarch64_vec_op_count::aarch64_vec_op_count): Add a parameter for it.
62 (aarch64_vec_op_count::vf_factor): New function.
63 (aarch64_vector_costs::aarch64_vector_costs): When costing for
64 neoverse-512tvb, pass a vf_factor of 2 for the Neoverse V1 version
66 (aarch64_vector_costs::adjust_body_cost): Read the vf factor
67 instead of hard-coding 2.
69 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
71 * config/aarch64/aarch64.c
72 (aarch64_vec_op_count::rename_cycles_per_iter): New function.
73 (aarch64_vec_op_count::min_nonpred_cycles_per_iter): Likewise.
74 (aarch64_vec_op_count::min_pred_cycles_per_iter): Likewise.
75 (aarch64_vec_op_count::min_cycles_per_iter): Likewise.
76 (aarch64_vec_op_count::dump): Move earlier in file. Dump the
78 (aarch64_estimate_min_cycles_per_iter): Delete.
79 (adjust_body_cost): Use aarch64_vec_op_count::min_cycles_per_iter
80 instead of aarch64_estimate_min_cycles_per_iter. Rely on the dump
81 routine to print CPI estimates.
82 (adjust_body_cost_sve): Likewise. Use the other functions above
83 instead of doing the work inline.
85 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
87 * config/aarch64/aarch64.c (aarch64_vec_op_count): Allow default
89 (aarch64_vec_op_count::base_issue_info): Remove handling of null
91 (aarch64_vec_op_count::simd_issue_info): Likewise.
92 (aarch64_vec_op_count::sve_issue_info): Likewise.
93 (aarch64_vector_costs::m_ops): Turn into a vector.
94 (aarch64_vector_costs::m_advsimd_ops): Likewise.
95 (aarch64_vector_costs::aarch64_vector_costs): Add entries to
96 the vectors based on aarch64_tune_params.
97 (aarch64_vector_costs::analyze_loop_vinfo): Update the pred_ops
98 of all entries in m_ops.
99 (aarch64_vector_costs::add_stmt_cost): Call count_ops for all
101 (aarch64_estimate_min_cycles_per_iter): Remove issue_info
102 parameter and get the information from the ops instead.
103 (aarch64_vector_costs::adjust_body_cost_sve): Take a
104 aarch64_vec_issue_info instead of a aarch64_vec_op_count.
105 (aarch64_vector_costs::adjust_body_cost): Update call accordingly.
106 Exit earlier if m_ops is empty for either cost structure.
108 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
110 * config/aarch64/aarch64.c (aarch64_vector_costs::m_scalar_ops)
111 (aarch64_vector_costs::m_sve_ops): Replace with...
112 (aarch64_vector_costs::m_ops): ...this.
113 (aarch64_vector_costs::analyze_loop_vinfo): Update accordingly.
114 (aarch64_vector_costs::adjust_body_cost_sve): Likewise.
115 (aarch64_vector_costs::aarch64_vector_costs): Likewise.
116 Initialize m_vec_flags here rather than in add_stmt_cost.
117 (aarch64_vector_costs::count_ops): Test for scalar reductions too.
118 Allow vectype to be null.
119 (aarch64_vector_costs::add_stmt_cost): Call count_ops for scalar
120 code too. Don't require vectype to be nonnull.
121 (aarch64_vector_costs::adjust_body_cost): Take the loop_vec_info
122 and scalar costs as parameters. Use the scalar costs to determine
123 the cycles per iteration of the scalar loop, then multiply it
125 (aarch64_vector_costs::finish_cost): Update call accordingly.
127 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
129 * config/aarch64/aarch64.c (aarch64_dr_type): New function.
130 (aarch64_vector_costs::count_ops): Use it rather than the
131 vectype to determine floatness.
133 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
135 * config/aarch64/aarch64.c (aarch64_sve_in_loop_reduction_latency):
136 Remove vectype parameter and get floatness from the type of the
138 (arch64_in_loop_reduction_latency): Likewise.
139 (aarch64_detect_vector_stmt_subtype): Update caller.
140 (aarch64_vector_costs::count_ops): Likewise.
142 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
144 * config/aarch64/aarch64.c (aarch64_sve_op_count): Fold into...
145 (aarch64_vec_op_count): ...this. Add a constructor.
146 (aarch64_vec_op_count::vec_flags): New function.
147 (aarch64_vec_op_count::base_issue_info): Likewise.
148 (aarch64_vec_op_count::simd_issue_info): Likewise.
149 (aarch64_vec_op_count::sve_issue_info): Likewise.
150 (aarch64_vec_op_count::m_issue_info): New member variable.
151 (aarch64_vec_op_count::m_vec_flags): Likewise.
152 (aarch64_vector_costs): Add a constructor.
153 (aarch64_vector_costs::m_sve_ops): Change type to aarch64_vec_op_count.
154 (aarch64_vector_costs::aarch64_vector_costs): New function.
155 Initialize m_scalar_ops, m_advsimd_ops and m_sve_ops.
156 (aarch64_vector_costs::count_ops): Remove vec_flags and
157 issue_info parameters, using the new aarch64_vec_op_count
159 (aarch64_vector_costs::add_stmt_cost): Update call accordingly.
160 (aarch64_sve_op_count::dump): Fold into...
161 (aarch64_vec_op_count::dump): ..here.
163 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
165 * config/aarch64/aarch64.c: Include tree-dfa.h.
166 (aarch64_check_consecutive_mems): New function that takes MEM_EXPR
167 and MEM_OFFSET into account.
168 (aarch64_swap_ldrstr_operands): Use it.
169 (aarch64_operands_ok_for_ldpstp): Likewise. Check that the
170 address of the second memory doesn't depend on the result of
173 2021-11-12 Jan Hubicka <hubicka@ucw.cz>
175 * ipa-pure-const.c (ipa_make_function_pure): Fix exit condition.
177 2021-11-12 Jan Hubicka <jh@suse.cz>
179 PR tree-optimization/103175
180 * ipa-modref.c (modref_lattice::merge): Add sanity check.
181 (callee_to_caller_flags): Make flags adjustment sane.
182 (modref_eaf_analysis::analyze_ssa_name): Likewise.
184 2021-11-12 Jan Hubicka <jh@suse.cz>
187 * ipa-modref.c (analyze_function, modref_propagate_in_scc): Do
188 not mark pure/const function if there are side-effects.
190 2021-11-12 Chung-Lin Tang <cltang@codesourcery.com>
192 * tree.h (OMP_CLAUSE_MAP_RUNTIME_IMPLICIT_P): New access macro for
193 'implicit' bit, using 'base.deprecated_flag' field of tree_node.
194 * tree-pretty-print.c (dump_omp_clause): Add support for printing
195 implicit attribute in tree dumping.
196 * gimplify.c (gimplify_adjust_omp_clauses_1):
197 Set OMP_CLAUSE_MAP_RUNTIME_IMPLICIT_P to 1 if map clause is implicitly
199 (gimplify_adjust_omp_clauses): Adjust place of adding implicitly created
200 clauses, from simple append, to starting of list, after non-map clauses.
201 * omp-low.c (lower_omp_target): Add GOMP_MAP_IMPLICIT bits into kind
202 values passed to libgomp for implicit maps.
204 2021-11-12 Jakub Jelinek <jakub@redhat.com>
206 * omp-builtins.def (BUILT_IN_GOMP_TEAMS): Remove.
207 (BUILT_IN_GOMP_TEAMS4): New.
208 * builtin-types.def (BT_FN_VOID_UINT_UINT): Remove.
209 (BT_FN_BOOL_UINT_UINT_UINT_BOOL): New.
210 * omp-low.c (lower_omp_teams): Use GOMP_teams4 instead of
211 GOMP_teams, pass to it also num_teams lower-bound expression
212 or a dup of upper-bound if it is missing and a flag whether
213 it is the first call or not.
215 2021-11-12 Martin Liska <mliska@suse.cz>
217 PR tree-optimization/102497
218 * gimple-predicate-analysis.cc (add_pred): Remove unused
221 2021-11-12 Richard Biener <rguenther@suse.de>
223 PR tree-optimization/103204
224 * tree-ssa-sccvn.c (valueize_refs_1): Re-valueize the
225 top operand after folding in an address.
227 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
229 * gimple-range-path.cc (path_range_query::path_range_query): New
230 ctor without a ranger.
231 (path_range_query::~path_range_query): Free ranger if necessary.
232 (path_range_query::range_on_path_entry): Adjust m_ranger for pointer.
233 (path_range_query::ssa_range_in_phi): Same.
234 (path_range_query::compute_ranges_in_block): Same.
235 (path_range_query::compute_imports): Same.
236 (path_range_query::compute_ranges): Same.
237 (path_range_query::range_of_stmt): Same.
238 (path_range_query::compute_outgoing_relations): Same.
239 * gimple-range-path.h (class path_range_query): New ctor.
240 * tree-ssa-loop-ch.c (ch_base::copy_headers): Remove gimple_ranger
241 as path_range_query allocates one.
242 * tree-ssa-threadbackward.c (class back_threader): Remove m_ranger.
243 (back_threader::~back_threader): Same.
245 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
247 * tree-ssa-threadbackward.c
248 (back_threader_profitability::profitable_path_p): Remove loop
249 crossing restriction.
251 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
253 * ipa-modref.c (analyze_function): Do pure/const discovery, return
255 (pass_modref::execute): If pure/const is discovered fixup cfg.
256 (ignore_edge): Do not ignore pure/const edges.
257 (modref_propagate_in_scc): Do pure/const discovery, return true if
258 cdtor was promoted pure/const.
259 (pass_ipa_modref::execute): If needed remove unreachable functions.
260 * ipa-pure-const.c (warn_function_noreturn): Fix whitespace.
261 (warn_function_cold): Likewise.
262 (skip_function_for_local_pure_const): Move earlier.
263 (ipa_make_function_const): Break out from ...
264 (ipa_make_function_pure): Break out from ...
265 (propagate_pure_const): ... here.
266 (pass_local_pure_const::execute): Use it.
267 * ipa-utils.h (ipa_make_function_const): Declare.
268 (ipa_make_function_pure): Declare.
269 * passes.def: Move early modref after pure-const.
271 2021-11-11 David Malcolm <dmalcolm@redhat.com>
274 * diagnostic-show-locus.c (def_policy): Use def_tabstop.
276 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
278 * config/aarch64/aarch64-builtins.c (TYPES_COMBINE): Delete.
279 (TYPES_COMBINEP): Delete.
280 * config/aarch64/aarch64-simd-builtins.def: Declare type-
281 qualified builtins for vcombine_* intrinsics.
282 * config/aarch64/arm_neon.h (vcombine_s8): Remove unnecessary
284 (vcombine_s16): Likewise.
285 (vcombine_s32): Likewise.
286 (vcombine_f32): Likewise.
287 (vcombine_u8): Use type-qualified builtin and remove casts.
288 (vcombine_u16): Likewise.
289 (vcombine_u32): Likewise.
290 (vcombine_u64): Likewise.
291 (vcombine_p8): Likewise.
292 (vcombine_p16): Likewise.
293 (vcombine_p64): Likewise.
294 (vcombine_bf16): Remove unnecessary cast.
295 * config/aarch64/iterators.md (VD_I): New mode iterator.
296 (VDC_P): New mode iterator.
298 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
300 * config/aarch64/aarch64-builtins.c (TYPES_LOAD1_U): Define.
301 (TYPES_LOAD1_P): Define.
302 (TYPES_STORE1_U): Define.
303 (TYPES_STORE1P): Rename to...
304 (TYPES_STORE1_P): This.
305 (get_mem_type_for_load_store): Add unsigned and poly types.
306 (aarch64_general_gimple_fold_builtin): Add unsigned and poly
307 type-qualified builtin declarations.
308 * config/aarch64/aarch64-simd-builtins.def: Declare type-
309 qualified builtins for LD1/ST1.
310 * config/aarch64/arm_neon.h (vld1_p8): Use type-qualified
311 builtin and remove cast.
312 (vld1_p16): Likewise.
314 (vld1_u16): Likewise.
315 (vld1_u32): Likewise.
316 (vld1q_p8): Likewise.
317 (vld1q_p16): Likewise.
318 (vld1q_p64): Likewise.
319 (vld1q_u8): Likewise.
320 (vld1q_u16): Likewise.
321 (vld1q_u32): Likewise.
322 (vld1q_u64): Likewise.
324 (vst1_p16): Likewise.
326 (vst1_u16): Likewise.
327 (vst1_u32): Likewise.
328 (vst1q_p8): Likewise.
329 (vst1q_p16): Likewise.
330 (vst1q_p64): Likewise.
331 (vst1q_u8): Likewise.
332 (vst1q_u16): Likewise.
333 (vst1q_u32): Likewise.
334 (vst1q_u64): Likewise.
335 * config/aarch64/iterators.md (VALLP_NO_DI): New iterator.
337 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
339 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
340 builtins for vector reduction.
341 * config/aarch64/arm_neon.h (vaddv_u8): Use type-qualified
342 builtin and remove casts.
343 (vaddv_u16): Likewise.
344 (vaddv_u32): Likewise.
345 (vaddvq_u8): Likewise.
346 (vaddvq_u16): Likewise.
347 (vaddvq_u32): Likewise.
348 (vaddvq_u64): Likewise.
350 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
352 * config/aarch64/aarch64-simd-builtins.def:
353 * config/aarch64/arm_neon.h (vpaddq_u8): Use type-qualified
354 builtin and remove casts.
355 (vpaddq_u16): Likewise.
356 (vpaddq_u32): Likewise.
357 (vpaddq_u64): Likewise.
358 (vpadd_u8): Likewise.
359 (vpadd_u16): Likewise.
360 (vpadd_u32): Likewise.
361 (vpaddd_u64): Likewise.
363 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
365 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
366 builtins for [r]subhn[2].
367 * config/aarch64/arm_neon.h (vsubhn_s16): Remove unnecessary
369 (vsubhn_s32): Likewise.
370 (vsubhn_s64): Likewise.
371 (vsubhn_u16): Use type-qualified builtin and remove casts.
372 (vsubhn_u32): Likewise.
373 (vsubhn_u64): Likewise.
374 (vrsubhn_s16): Remove unnecessary cast.
375 (vrsubhn_s32): Likewise.
376 (vrsubhn_s64): Likewise.
377 (vrsubhn_u16): Use type-qualified builtin and remove casts.
378 (vrsubhn_u32): Likewise.
379 (vrsubhn_u64): Likewise.
380 (vrsubhn_high_s16): Remove unnecessary cast.
381 (vrsubhn_high_s32): Likewise.
382 (vrsubhn_high_s64): Likewise.
383 (vrsubhn_high_u16): Use type-qualified builtin and remove
385 (vrsubhn_high_u32): Likewise.
386 (vrsubhn_high_u64): Likewise.
387 (vsubhn_high_s16): Remove unnecessary cast.
388 (vsubhn_high_s32): Likewise.
389 (vsubhn_high_s64): Likewise.
390 (vsubhn_high_u16): Use type-qualified builtin and remove
392 (vsubhn_high_u32): Likewise.
393 (vsubhn_high_u64): Likewise.
395 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
397 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
398 builtins for [r]addhn[2].
399 * config/aarch64/arm_neon.h (vaddhn_s16): Remove unnecessary
401 (vaddhn_s32): Likewise.
402 (vaddhn_s64): Likewise.
403 (vaddhn_u16): Use type-qualified builtin and remove casts.
404 (vaddhn_u32): Likewise.
405 (vaddhn_u64): Likewise.
406 (vraddhn_s16): Remove unnecessary cast.
407 (vraddhn_s32): Likewise.
408 (vraddhn_s64): Likewise.
409 (vraddhn_u16): Use type-qualified builtin and remove casts.
410 (vraddhn_u32): Likewise.
411 (vraddhn_u64): Likewise.
412 (vaddhn_high_s16): Remove unnecessary cast.
413 (vaddhn_high_s32): Likewise.
414 (vaddhn_high_s64): Likewise.
415 (vaddhn_high_u16): Use type-qualified builtin and remove
417 (vaddhn_high_u32): Likewise.
418 (vaddhn_high_u64): Likewise.
419 (vraddhn_high_s16): Remove unnecessary cast.
420 (vraddhn_high_s32): Likewise.
421 (vraddhn_high_s64): Likewise.
422 (vraddhn_high_u16): Use type-qualified builtin and remove
424 (vraddhn_high_u32): Likewise.
425 (vraddhn_high_u64): Likewise.
427 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
429 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
430 qualifiers in generator macros for uhsub builtins.
431 * config/aarch64/arm_neon.h (vhsub_s8): Remove unnecessary
433 (vhsub_s16): Likewise.
434 (vhsub_s32): Likewise.
435 (vhsub_u8): Use type-qualified builtin and remove casts.
436 (vhsub_u16): Likewise.
437 (vhsub_u32): Likewise.
438 (vhsubq_s8): Remove unnecessary cast.
439 (vhsubq_s16): Likewise.
440 (vhsubq_s32): Likewise.
441 (vhsubq_u8): Use type-qualified builtin and remove casts.
442 (vhsubq_u16): Likewise.
443 (vhsubq_u32): Likewise.
445 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
447 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
448 qualifiers in generator macros for u[r]hadd builtins.
449 * config/aarch64/arm_neon.h (vhadd_s8): Remove unnecessary
451 (vhadd_s16): Likewise.
452 (vhadd_s32): Likewise.
453 (vhadd_u8): Use type-qualified builtin and remove casts.
454 (vhadd_u16): Likewise.
455 (vhadd_u32): Likewise.
456 (vhaddq_s8): Remove unnecessary cast.
457 (vhaddq_s16): Likewise.
458 (vhaddq_s32): Likewise.
459 (vhaddq_u8): Use type-qualified builtin and remove casts.
460 (vhaddq_u16): Likewise.
461 (vhaddq_u32): Likewise.
462 (vrhadd_s8): Remove unnecessary cast.
463 (vrhadd_s16): Likewise.
464 (vrhadd_s32): Likewise.
465 (vrhadd_u8): Use type-qualified builtin and remove casts.
466 (vrhadd_u16): Likewise.
467 (vrhadd_u32): Likewise.
468 (vrhaddq_s8): Remove unnecessary cast.
469 (vrhaddq_s16): Likewise.
470 (vrhaddq_s32): Likewise.
471 (vrhaddq_u8): Use type-wualified builtin and remove casts.
472 (vrhaddq_u16): Likewise.
473 (vrhaddq_u32): Likewise.
475 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
477 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
478 qualifiers in generator macros for usub[lw][2] builtins.
479 * config/aarch64/arm_neon.h (vsubl_s8): Remove unnecessary
481 (vsubl_s16): Likewise.
482 (vsubl_s32): Likewise.
483 (vsubl_u8): Use type-qualified builtin and remove casts.
484 (vsubl_u16): Likewise.
485 (vsubl_u32): Likewise.
486 (vsubl_high_s8): Remove unnecessary cast.
487 (vsubl_high_s16): Likewise.
488 (vsubl_high_s32): Likewise.
489 (vsubl_high_u8): Use type-qualified builtin and remove casts.
490 (vsubl_high_u16): Likewise.
491 (vsubl_high_u32): Likewise.
492 (vsubw_s8): Remove unnecessary casts.
493 (vsubw_s16): Likewise.
494 (vsubw_s32): Likewise.
495 (vsubw_u8): Use type-qualified builtin and remove casts.
496 (vsubw_u16): Likewise.
497 (vsubw_u32): Likewise.
498 (vsubw_high_s8): Remove unnecessary cast.
499 (vsubw_high_s16): Likewise.
500 (vsubw_high_s32): Likewise.
501 (vsubw_high_u8): Use type-qualified builtin and remove casts.
502 (vsubw_high_u16): Likewise.
503 (vsubw_high_u32): Likewise.
505 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
507 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
508 qualifiers in generator macros for uadd[lw][2] builtins.
509 * config/aarch64/arm_neon.h (vaddl_s8): Remove unnecessary
511 (vaddl_s16): Likewise.
512 (vaddl_s32): Likewise.
513 (vaddl_u8): Use type-qualified builtin and remove casts.
514 (vaddl_u16): Likewise.
515 (vaddl_u32): Likewise.
516 (vaddl_high_s8): Remove unnecessary cast.
517 (vaddl_high_s16): Likewise.
518 (vaddl_high_s32): Likewise.
519 (vaddl_high_u8): Use type-qualified builtin and remove casts.
520 (vaddl_high_u16): Likewise.
521 (vaddl_high_u32): Likewise.
522 (vaddw_s8): Remove unnecessary cast.
523 (vaddw_s16): Likewise.
524 (vaddw_s32): Likewise.
525 (vaddw_u8): Use type-qualified builtin and remove casts.
526 (vaddw_u16): Likewise.
527 (vaddw_u32): Likewise.
528 (vaddw_high_s8): Remove unnecessary cast.
529 (vaddw_high_s16): Likewise.
530 (vaddw_high_s32): Likewise.
531 (vaddw_high_u8): Use type-qualified builtin and remove casts.
532 (vaddw_high_u16): Likewise.
533 (vaddw_high_u32): Likewise.
535 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
537 * config/aarch64/aarch64-simd-builtins.def: Declare type-
538 qualified builtins for [R]SHRN[2].
539 * config/aarch64/arm_neon.h (vshrn_n_u16): Use type-qualified
540 builtin and remove casts.
541 (vshrn_n_u32): Likewise.
542 (vshrn_n_u64): Likewise.
543 (vrshrn_high_n_u16): Likewise.
544 (vrshrn_high_n_u32): Likewise.
545 (vrshrn_high_n_u64): Likewise.
546 (vrshrn_n_u16): Likewise.
547 (vrshrn_n_u32): Likewise.
548 (vrshrn_n_u64): Likewise.
549 (vshrn_high_n_u16): Likewise.
550 (vshrn_high_n_u32): Likewise.
551 (vshrn_high_n_u64): Likewise.
553 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
555 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
556 type-qualified builtins for XTN[2].
557 * config/aarch64/arm_neon.h (vmovn_high_u16): Use type-
558 qualified builtin and remove casts.
559 (vmovn_high_u32): Likewise.
560 (vmovn_high_u64): Likewise.
561 (vmovn_u16): Likewise.
562 (vmovn_u32): Likewise.
563 (vmovn_u64): Likewise.
565 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
567 * config/aarch64/aarch64-simd-builtins.def: Use poly type
568 qualifier in builtin generator macros.
569 * config/aarch64/arm_neon.h (vmul_p8): Use type-qualified
570 builtin and remove casts.
571 (vmulq_p8): Likewise.
572 (vmull_high_p8): Likewise.
573 (vmull_p8): Likewise.
575 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
577 * config/aarch64/aarch64-simd-builtins.def: Declare type-
578 qualified builtin generators for unsigned MLA/MLS intrinsics.
579 * config/aarch64/arm_neon.h (vmla_n_u16): Use type-qualified
581 (vmla_n_u32): Likewise.
583 (vmla_u16): Likewise.
584 (vmla_u32): Likewise.
585 (vmlaq_n_u16): Likewise.
586 (vmlaq_n_u32): Likewise.
587 (vmlaq_u8): Likewise.
588 (vmlaq_u16): Likewise.
589 (vmlaq_u32): Likewise.
590 (vmls_n_u16): Likewise.
591 (vmls_n_u32): Likewise.
593 (vmls_u16): Likewise.
594 (vmls_u32): Likewise.
595 (vmlsq_n_u16): Likewise.
596 (vmlsq_n_u32): Likewise.
597 (vmlsq_u8): Likewise.
598 (vmlsq_u16): Likewise.
599 (vmlsq_u32): Likewise.
601 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
603 * ipa-modref.c (modref_summary::useful_p): Check also for side-effects
604 with looping const/pure.
605 (modref_summary_lto::useful_p): Likewise.
606 (merge_call_side_effects): Merge side effects before early exit
608 (process_fnspec): Also handle pure functions.
609 (analyze_call): Do not early exit on looping pure const.
610 (propagate_unknown_call): Also handle nontrivial SCC as side-effect.
611 (modref_propagate_in_scc): Update.
613 2021-11-11 Richard Biener <rguenther@suse.de>
615 PR tree-optimization/103190
616 * tree-ssa-reassoc.c (insert_stmt_after): Only assert on asm goto.
618 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
620 * gimple-range-path.cc (path_range_query::add_copies_to_imports):
622 (path_range_query::compute_imports): ...this. Adapt it so it can
623 be passed the imports bitmap instead of working on m_imports.
624 (path_range_query::compute_ranges): Call compute_imports in all
625 cases unless an imports bitmap is passed.
626 * gimple-range-path.h (path_range_query::compute_imports): New.
627 (path_range_query::add_copies_to_imports): Remove.
628 * tree-ssa-threadbackward.c (back_threader::resolve_def): Remove.
629 (back_threader::find_paths_to_names): Inline resolve_def.
630 (back_threader::find_paths): Call compute_imports.
631 (back_threader::resolve_phi): Adjust comment.
633 2021-11-11 Richard Biener <rguenther@suse.de>
635 PR tree-optimization/103188
636 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
637 Remove query parameter, split out check for size
639 (ch_base::m_ranger, cb_base::m_query): Remove.
640 (ch_base::copy_headers): Split processing loop into
641 analysis around which we allocate and use ranger and
642 transform where we do not.
643 (pass_ch::execute): Do not allocate/free ranger here.
644 (pass_ch_vect::execute): Likewise.
646 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
648 * ipa-pure-const.c (propagate_pure_const): Self recursion is
651 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
653 * cgraph.c (set_noreturn_flag_1): New function.
654 (cgraph_node::set_noreturn_flag): New member function
655 * cgraph.h (cgraph_node::set_noreturn_flags): Declare.
656 * ipa-pure-const.c (pass_local_pure_const::execute): Use it.
658 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
659 Richard Biener <rguenther@suse.de>
661 * tree-ssa-loop-ch.c (entry_loop_condition_is_static): Resolve
662 statically to the edge remaining in the loop.
664 2021-11-11 Richard Biener <rguenther@suse.de>
667 * tree-eh.c (operation_could_trap_helper_p): Properly
668 check vector constants for a zero element for integer
669 division. Separate floating point and integer division code.
671 2021-11-11 Jakub Jelinek <jakub@redhat.com>
674 * dwarf2out.c (field_byte_offset): Do the PCC_BITFIELD_TYPE_MATTERS
675 handling only for DECL_BIT_FIELD_TYPE decls.
677 2021-11-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
680 * config/aarch64/aarch64.c (aarch64_process_target_attr): Check if
681 token is arch extension without leading '+' and emit appropriate
682 diagnostic for the same.
684 2021-11-11 Jakub Jelinek <jakub@redhat.com>
686 * tree.h (OMP_CLAUSE_NUM_TEAMS_EXPR): Rename to ...
687 (OMP_CLAUSE_NUM_TEAMS_UPPER_EXPR): ... this.
688 (OMP_CLAUSE_NUM_TEAMS_LOWER_EXPR): Define.
689 * tree.c (omp_clause_num_ops): Increase num ops for
690 OMP_CLAUSE_NUM_TEAMS to 2.
691 * tree-pretty-print.c (dump_omp_clause): Print optional lower bound
692 for OMP_CLAUSE_NUM_TEAMS.
693 * gimplify.c (gimplify_scan_omp_clauses): Gimplify
694 OMP_CLAUSE_NUM_TEAMS_LOWER_EXPR if non-NULL.
695 (optimize_target_teams): Use OMP_CLAUSE_NUM_TEAMS_UPPER_EXPR instead
696 of OMP_CLAUSE_NUM_TEAMS_EXPR. Handle OMP_CLAUSE_NUM_TEAMS_LOWER_EXPR.
697 * omp-low.c (lower_omp_teams): Use OMP_CLAUSE_NUM_TEAMS_UPPER_EXPR
698 instead of OMP_CLAUSE_NUM_TEAMS_EXPR.
699 * omp-expand.c (expand_teams_call, get_target_arguments): Likewise.
701 2021-11-11 Richard Biener <rguenther@suse.de>
703 * cfganal.c (find_pdom): Remove.
704 (control_dependences::find_control_dependence): Remove
705 special-casing of entry block, call get_immediate_dominator
707 * gimple-predicate-analysis.cc (find_pdom): Remove.
708 (find_dom): Likewise.
709 (find_control_equiv_block): Call get_immediate_dominator
711 (compute_control_dep_chain): Likewise.
712 (predicate::init_from_phi_def): Likewise.
714 2021-11-11 Richard Biener <rguenther@suse.de>
716 * cfganal.h (control_dependences::control_dependence_map):
718 (control_dependences::m_bitmaps): New.
719 * cfganal.c (control_dependences::set_control_dependence_map_bit):
721 (control_dependences::clear_control_dependence_bitmap):
723 (control_dependences::find_control_dependence): Do not
724 find_edge for the abnormal edge test.
725 (control_dependences::control_dependences): Instead do not
726 add abnormal edges to the edge list. Adjust.
727 (control_dependences::~control_dependences): Likewise.
728 (control_dependences::get_edges_dependent_on): Likewise.
729 * function-tests.c: Include bitmap.h.
731 2021-11-11 Kewen Lin <linkw@linux.ibm.com>
733 * doc/invoke.texi: Change references to "future cpu" to "power10",
734 "-mcpu=future" to "-mcpu=power10". Adjust words for float128.
736 2021-11-11 Cui,Lili <lili.cui@intel.com>
738 * config/i386/i386-options.c (m_CORE_AVX2): Remove Alderlake
740 (processor_cost_table): Use alderlake_cost for Alderlake.
741 * config/i386/i386.c (ix86_sched_init_global): Handle Alderlake.
742 * config/i386/x86-tune-costs.h (struct processor_costs): Add alderlake
744 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Alderlake
746 (ix86_adjust_cost): Handle Alderlake.
747 * config/i386/x86-tune.def (X86_TUNE_SCHEDULE): Enable for Alderlake.
748 (X86_TUNE_PARTIAL_REG_DEPENDENCY): Likewise.
749 (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY): Likewise.
750 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): Likewise.
751 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
752 (X86_TUNE_MEMORY_MISMATCH_STALL): Likewise.
753 (X86_TUNE_USE_LEAVE): Likewise.
754 (X86_TUNE_PUSH_MEMORY): Likewise.
755 (X86_TUNE_USE_INCDEC): Likewise.
756 (X86_TUNE_INTEGER_DFMODE_MOVES): Likewise.
757 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
758 (X86_TUNE_USE_SAHF): Likewise.
759 (X86_TUNE_USE_BT): Likewise.
760 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
761 (X86_TUNE_ONE_IF_CONV_INSN): Likewise.
762 (X86_TUNE_AVOID_MFENCE): Likewise.
763 (X86_TUNE_USE_SIMODE_FIOP): Likewise.
764 (X86_TUNE_EXT_80387_CONSTANTS): Likewise.
765 (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL): Likewise.
766 (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Likewise.
767 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
768 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
769 (X86_TUNE_AVOID_4BYTE_PREFIXES): Likewise.
770 (X86_TUNE_USE_GATHER): Disable for Alderlake.
771 (X86_TUNE_AVX256_MOVE_BY_PIECES): Likewise.
772 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
774 2021-11-11 liuhongt <hongtao.liu@intel.com>
777 * config/i386/sse.md (V_128_256): Extend to V8HF/V16HF.
778 (avxsizesuffix): Ditto.
780 2021-11-11 Kito Cheng <kito.cheng@sifive.com>
782 * common/config/riscv/riscv-common.c (riscv_subset_list::to_string): Fix
783 wrong marco checking.
785 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
787 PR tree-optimization/102906
788 * tree-ssa-loop-ch.c (entry_loop_condition_is_static): New.
789 (should_duplicate_loop_header_p): Call entry_loop_condition_is_static.
790 (class ch_base): Add m_ranger and m_query.
791 (ch_base::copy_headers): Pass m_query to
792 entry_loop_condition_is_static.
793 (pass_ch::execute): Allocate and deallocate m_ranger and
795 (pass_ch_vect::execute): Same.
797 2021-11-10 Andrew Pinski <apinski@marvell.com>
800 * config/aarch64/aarch64-simd.md (aarch64_simd_dup<mode>):
801 Use vwcore iterator for the r constraint output string.
803 2021-11-10 qing zhao <qing.zhao@oracle.com>
805 * internal-fn.c (expand_DEFERRED_INIT): Apply pattern initialization
806 only when have_insn_for return true for the mode. Fix a memory leak.
808 2021-11-10 Christophe Lyon <christophe.lyon@foss.st.com>
810 * config/arm/arm.c (cortexa9_extra_costs, cortexa8_extra_costs,
811 cortexa5_extra_costs, cortexa7_extra_costs,
812 cortexa12_extra_costs, cortexa15_extra_costs, v7m_extra_costs):
813 Initialize movi, dup and extract costing fields.
815 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
817 * gimple-range-path.cc (path_range_query::path_range_query): Do
819 (path_range_query::dump): Change m_path uses to non-pointer.
820 (path_range_query::defined_outside_path): Same.
821 (path_range_query::set_path): Same.
822 (path_range_query::add_copies_to_imports): Same.
823 (path_range_query::range_of_stmt): Same.
824 (path_range_query::compute_outgoing_relations): Same.
825 (path_range_query::compute_ranges): Imports are now optional.
826 Implement overload that takes an edge.
827 * gimple-range-path.h (class path_range_query): Make imports
828 optional for compute_ranges. Add compute_ranges(edge) overload.
829 Make m_path an auto_vec instead of a pointer and adjust
832 2021-11-10 Tamar Christina <tamar.christina@arm.com>
834 * tree-vectorizer.h (struct scalar_cond_masked_key): Add inverted_p.
835 (default_hash_traits<scalar_conf_masked_key>): Likewise.
836 * tree-vect-stmts.c (vectorizable_condition): Check if inverse of mask
838 * tree-vectorizer.c (scalar_cond_masked_key::get_cond_ops_from_tree):
839 Register mask inverses.
841 2021-11-10 Tamar Christina <tamar.christina@arm.com>
843 * tree-vectorizer.c (vectorize_loops): Do local CSE through RPVN upon
844 successful vectorization.
846 2021-11-10 Andrew MacLeod <amacleod@redhat.com>
848 * gimple-range-cache.cc (sbr_vector::grow): New.
849 (sbr_vector::set_bb_range): Call grow.
850 (sbr_vector::get_bb_range): Same.
851 (sbr_vector::bb_range_p): Remove assert.
853 2021-11-10 Tamar Christina <tamar.christina@arm.com>
855 * config/aarch64/aarch64-simd.md (*aarch64_topbits_shuffle<mode>_le
856 ,*aarch64_topbits_shuffle<mode>_be): Remove.
858 2021-11-10 Jan Hubicka <jh@suse.cz>
860 * ipa-modref.c: Include tree-eh.h
861 (modref_summary::modref_summary): Initialize side_effects.
862 (struct modref_summary_lto): New bool field side_effects.
863 (modref_summary_lto::modref_summary_lto): Initialize side_effects.
864 (modref_summary::dump): Dump side_effects.
865 (modref_summary_lto::dump): Dump side_effects.
866 (merge_call_side_effects): Merge side effects.
867 (process_fnspec): Calls to non-const/pure or looping
868 function is a side effect.
869 (analyze_call): Self-recursion is a side-effect; handle
871 (analyze_load): Watch for volatile and throwing memory.
872 (analyze_store): Likewise.
873 (analyze_stmt): Watch for volatitle asm.
874 (analyze_function): Handle side_effects.
875 (modref_summaries::duplicate): Duplicate side_effects.
876 (modref_summaries_lto::duplicate): Likewise.
877 (modref_write): Stream side_effects.
878 (read_section): Likewise.
879 (update_signature): Update.
880 (propagate_unknown_call): Handle side_effects.
881 (modref_propagate_in_scc): Likewise.
882 * ipa-modref.h (struct modref_summary): Add side_effects.
883 * ipa-pure-const.c (special_builtin_state): Rename to ...
884 (builtin_safe_for_const_function_p): ... this one.
885 (check_call): Update.
886 (finite_function_p): Break out from ...
887 (propagate_pure_const): ... here
888 * ipa-utils.h (finite_function): Declare.
890 2021-11-10 Lucas A. M. Magalhães <lamm@linux.ibm.com>
892 * config.gcc (powerpc*-*-*): Remove -rpath from
893 --with-advance-toolchain.
895 2021-11-10 Marek Polacek <polacek@redhat.com>
898 * attribs.c (struct scoped_attributes): Add a bool member.
899 (lookup_scoped_attribute_spec): Forward declare.
900 (register_scoped_attributes): New bool parameter, defaulted to
902 (handle_ignored_attributes_option): New function.
903 (free_attr_data): New function.
904 (init_attributes): Call handle_ignored_attributes_option.
905 (attr_namespace_ignored_p): New function.
906 (decl_attributes): Check attr_namespace_ignored_p before
908 * attribs.h (free_attr_data): Declare.
909 (register_scoped_attributes): Adjust declaration.
910 (handle_ignored_attributes_option): Declare.
911 (canonicalize_attr_name): New function template.
912 (canonicalize_attr_name): Use it.
913 * common.opt (Wattributes=): New option with a variable.
914 * doc/extend.texi: Document #pragma GCC diagnostic ignored_attributes.
915 * doc/invoke.texi: Document -Wno-attributes=.
916 * opts.c (common_handle_option) <case OPT_Wattributes_>: Handle.
917 * plugin.h (register_scoped_attributes): Adjust declaration.
918 * toplev.c (compile_file): Call free_attr_data.
920 2021-11-10 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
922 * config/arm/arm-cpus.in (cortex-a710): New CPU.
923 * config/arm/arm-tables.opt: Regenerate.
924 * config/arm/arm-tune.md: Regenerate.
925 * doc/invoke.texi: Update docs.
927 2021-11-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
929 * config/aarch64/aarch64-builtins.c
930 (aarch64_general_gimple_fold_builtin): Mark argument as unused.
932 2021-11-10 Martin Liska <mliska@suse.cz>
934 * lto-wrapper.c (merge_and_complain): Make the first argument
937 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
939 * config/aarch64/iterators.md (optab): Use fmax_nan instead of
940 smax_nan and fmin_nan instead of smin_nan.
941 (maxmin_uns): Rename to...
942 (fmaxmin): ...this and make the same changes. Remove entries
943 unrelated to fmax* and fmin*.
944 * config/aarch64/aarch64.md (<maxmin_uns><mode>3): Rename to...
945 (<fmaxmin><mode>3): ...this.
946 * config/aarch64/aarch64-simd.md (aarch64_<maxmin_uns>p<mode>):
948 (aarch64_<optab>p<mode>): ...this.
949 (<maxmin_uns><mode>3): Rename to...
950 (<fmaxmin><mode>3): ...this.
951 (reduc_<maxmin_uns>_scal_<mode>): Rename to...
952 (reduc_<optab>_scal_<mode>): ...this and update gen* call.
953 (aarch64_reduc_<maxmin_uns>_internal<mode>): Rename to...
954 (aarch64_reduc_<optab>_internal<mode>): ...this.
955 (aarch64_reduc_<maxmin_uns>_internalv2si): Rename to...
956 (aarch64_reduc_<optab>_internalv2si): ...this.
957 * config/aarch64/aarch64-sve.md (<maxmin_uns><mode>3): Rename to...
958 (<fmaxmin><mode>3): ...this.
959 * config/aarch64/aarch64-simd-builtins.def (smax_nan, smin_nan)
961 (fmax_nan, fmin_nan): ...this.
962 * config/aarch64/arm_neon.h (vmax_f32, vmax_f64, vmaxq_f32, vmaxq_f64)
963 (vmin_f32, vmin_f64, vminq_f32, vminq_f64, vmax_f16, vmaxq_f16)
964 (vmin_f16, vminq_f16): Update accordingly.
966 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
968 * tree-vectorizer.h (vector_costs::finish_cost): Take the
969 corresponding scalar costs as a parameter.
970 (finish_cost): Likewise.
971 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
972 (vect_estimate_min_profitable_iters): Update accordingly.
973 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Likewise.
974 * tree-vectorizer.c (vector_costs::finish_cost): Likewise.
975 * config/aarch64/aarch64.c (aarch64_vector_costs::finish_cost):
977 * config/rs6000/rs6000.c (rs6000_cost_data::finish_cost): Likewise.
979 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
981 * tree-vectorizer.h (_loop_vec_info::scalar_costs): New member
983 (_loop_vec_info::single_scalar_iteration_cost): Delete.
984 (LOOP_VINFO_SINGLE_SCALAR_ITERATION_COST): Delete.
985 (vector_costs::total_cost): New function.
986 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
988 (_loop_vec_info::~_loop_vec_info): Delete scalar_costs.
989 (vect_compute_single_scalar_iteration_cost): Store the costs
990 in loop_vinfo->scalar_costs.
991 (vect_estimate_min_profitable_iters): Get the scalar cost from
992 loop_vinfo->scalar_costs.
994 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
996 * tree-vectorizer.h (vector_costs::better_main_loop_than_p)
997 (vector_costs::better_epilogue_loop_than_p)
998 (vector_costs::compare_inside_loop_cost)
999 (vector_costs::compare_outside_loop_cost): Likewise.
1000 * tree-vectorizer.c (vector_costs::better_main_loop_than_p)
1001 (vector_costs::better_epilogue_loop_than_p)
1002 (vector_costs::compare_inside_loop_cost)
1003 (vector_costs::compare_outside_loop_cost): New functions,
1004 containing code moved from...
1005 * tree-vect-loop.c (vect_better_loop_vinfo_p): ...here.
1007 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
1009 * tree-vectorizer.h (_loop_vec_info): Remove vec_outside_cost
1010 and vec_inside_cost.
1011 (vector_costs::outside_cost): New function.
1012 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
1014 (vect_estimate_min_profitable_iters): Likewise.
1015 (vect_better_loop_vinfo_p): Get the inside and outside costs
1016 from the loop_vec_infos' vector_costs.
1018 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
1020 * tree-vectorizer.h (vec_info::target_cost_data): Replace with...
1021 (_loop_vec_info::vector_costs): ...this.
1022 (LOOP_VINFO_TARGET_COST_DATA): Delete.
1023 * tree-vectorizer.c (vec_info::vec_info): Remove target_cost_data
1025 (vec_info::~vec_info): Remove corresponding delete.
1026 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1027 vector_costs to null.
1028 (_loop_vec_info::~_loop_vec_info): Delete vector_costs.
1029 (vect_analyze_loop_operations): Update after above changes.
1030 (vect_analyze_loop_2): Likewise.
1031 (vect_estimate_min_profitable_iters): Likewise.
1032 * tree-vect-slp.c (vect_slp_analyze_operations): Likewise.
1034 2021-11-10 Jan Hubicka <hubicka@ucw.cz>
1036 * tree-core.h (EAF_DIRECT): Remove.
1037 (EAF_NOCLOBBER): Remove.
1038 (EAF_UNUSED): Remove.
1039 (EAF_NOESCAPE): Remove.
1040 (EAF_NO_DIRECT_CLOBBER): New.
1041 (EAF_NO_INDIRECT_CLOBBER): New.
1042 (EAF_NODIRECTESCAPE): Remove.
1043 (EAF_NO_DIRECT_ESCAPE): New.
1044 (EAF_NO_INDIRECT_ESCAPE): New.
1045 (EAF_NOT_RETURNED): Remove.
1046 (EAF_NOT_RETURNED_INDIRECTLY): New.
1047 (EAF_NOREAD): Remove.
1048 (EAF_NO_DIRECT_READ): New.
1049 (EAF_NO_INDIRECT_READ): New.
1050 * gimple.c (gimple_call_arg_flags): Update for new flags.
1051 (gimple_call_retslot_flags): Update for new flags.
1052 * ipa-modref.c (dump_eaf_flags): Likewise.
1053 (remove_useless_eaf_flags): Likewise.
1054 (deref_flags): Likewise.
1055 (modref_lattice::init): Likewise.
1056 (modref_lattice::merge): Likewise.
1057 (modref_lattice::merge_direct_load): Likewise.
1058 (modref_lattice::merge_direct_store): Likewise.
1059 (modref_eaf_analysis::merge_call_lhs_flags): Likewise.
1060 (callee_to_caller_flags): Likewise.
1061 (modref_eaf_analysis::analyze_ssa_name): Likewise.
1062 (modref_eaf_analysis::propagate): Likewise.
1063 (modref_merge_call_site_flags): Likewise.
1064 * ipa-modref.h (interposable_eaf_flags): Likewise.
1065 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1) Likewise.
1066 * tree-ssa-structalias.c (handle_call_arg): Likewise.
1067 (handle_rhs_call): Likewise.
1068 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Likewise.
1070 2021-11-10 Tamar Christina <tamar.christina@arm.com>
1072 * tree-vect-slp-patterns.c (complex_mul_pattern::matches): Remove l1node.
1074 2021-11-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
1076 * config/aarch64/aarch64-builtins.c
1077 (aarch64_general_gimple_fold_builtin): Change pointer alignment and
1080 2021-11-10 Jan Hubicka <jh@suse.cz>
1082 * ipa-modref-tree.h (modref_tree::remap_params): Fix off-by-one error.
1084 2021-11-10 H.J. Lu <hongjiu.lu@intel.com>
1085 Hongtao Liu <hongtao.liu@intel.com>
1087 PR middle-end/102566
1088 * match.pd (nop_atomic_bit_test_and_p): New match.
1089 * tree-ssa-ccp.c (convert_atomic_bit_not): New function.
1090 (gimple_nop_atomic_bit_test_and_p): New prototype.
1091 (optimize_atomic_bit_test_and): Transform equivalent, but slighly
1092 different cases to their canonical forms.
1094 2021-11-10 liuhongt <hongtao.liu@intel.com>
1096 PR tree-optimization/103126
1097 * tree-vect-loop.c (neutral_op_for_reduction): Remove static.
1098 * tree-vectorizer.h (neutral_op_for_reduction): Declare.
1099 * tree-if-conv.c : Include tree-vectorizer.h.
1100 (is_cond_scalar_reduction): Handle
1101 BIT_XOR_EXPR/BIT_IOR_EXPR/BIT_AND_EXPR.
1102 (convert_scalar_cond_reduction): Ditto.
1104 2021-11-10 konglin1 <lingling.kong@intel.com>
1106 * config/i386/sse.md (cmul<conj_op><mode>3): add new define_expand.
1107 (cmla<conj_op><mode>4): Likewise
1109 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
1111 * tree-ssa-threadedge.c: Do not include
1112 gimple-ssa-evrp-analyze.h.
1113 * value-pointer-equiv.cc: Same.
1115 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
1117 * tree-ssa-threadbackward.c
1118 (back_threader::maybe_register_path_dump): Abstract path dumping...
1119 (dump_path): ...here.
1120 (back_threader::resolve_phi): Call dump_path.
1123 2021-11-10 konglin1 <lingling.kong@intel.com>
1125 * config/i386/sse.md (fma_<complexpairopname>_<mode>_pair):
1126 Add new define_insn.
1127 (fma_<mode>_fmaddc_bcst): Add new define_insn_and_split.
1128 (fma_<mode>_fcmaddc_bcst): Likewise
1130 2021-11-10 liuhongt <hongtao.liu@intel.com>
1133 * match.pd: Simplify (trunc)fmax/fmin((extend)a, (extend)b) to
1136 2021-11-10 Andrew Pinski <apinski@marvell.com>
1139 * config/aarch64/aarch64.c (aarch64_evpc_ins): Don't use target
1140 as an input, use original one.
1142 2021-11-10 Sandra Loosemore <sandra@codesourcery.com>
1144 * config/nios2/nios2.c (nios2_can_inline_p): New.
1145 (TARGET_CAN_INLINE_P): Define.
1147 2021-11-09 Bill Schmidt <wschmidt@linux.ibm.com>
1149 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin):
1150 Disable gimple fold for RS6000_BIF_{XVMINDP,XVMINSP,VMINFP} and
1151 RS6000_BIF_{XVMAXDP,XVMAXSP,VMAXFP} when fast-math is not set.
1152 (lxvrse_expand_builtin): Modify the expansion for sign extension.
1153 All extensions are done within VSX registers.
1155 2021-11-09 Andrew Pinski <apinski@marvell.com>
1157 PR tree-optimization/10352
1158 * match.pd: Remove check of TYPE_PRECISION for
1161 2021-11-09 Andrew MacLeod <amacleod@redhat.com>
1163 * function.c (allocate_struct_function): Don't set x_range_query.
1164 * function.h (get_range_query): Move to value-query.h.
1165 * gimple-range.cc (enable_ranger): Check that query is currently NULL.
1166 (disable_ranger): Clear function current query field.
1167 * value-query.cc (get_global_range_query): Relocate to:
1168 * value-query.h (get_global_range_query): Here and inline.
1169 (get_range_query): Relocate here from function.h.
1171 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
1173 * tree-ssa-threadbackward.c
1174 (back_threader::maybe_register_path_dump): New.
1175 (back_threader::maybe_register_path): Call maybe_register_path_dump.
1177 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
1179 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
1180 Return NULL when unprofitable.
1182 2021-11-09 Martin Jambor <mjambor@suse.cz>
1184 * tree.h (build_debug_expr_decl): Declare.
1185 * tree.c (build_debug_expr_decl): New function.
1186 * cfgexpand.c (avoid_deep_ter_for_debug): Use build_debug_expr_decl
1187 instead of building a DEBUG_EXPR_DECL.
1188 * ipa-param-manipulation.c
1189 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
1190 * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Likewise.
1191 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
1192 * tree-ssa-phiopt.c (spaceship_replacement): Likewise.
1193 * tree-ssa-reassoc.c (make_new_ssa_for_def): Likewise.
1195 2021-11-09 Bill Schmidt <wschmidt@linux.ibm.com>
1197 * config/rs6000/rs6000-call.c (rs6000_type_string): New function.
1198 (def_builtin): Change debug formatting for easier parsing and
1199 include more information.
1200 (rs6000_init_builtins): Add dump of autogenerated builtins.
1201 (altivec_init_builtins): Dump __builtin_altivec_mask_for_load for
1204 2021-11-09 Marek Polacek <polacek@redhat.com>
1206 * ipa-modref.c (analyze_parms): Fix a typo.
1208 2021-11-09 Kito Cheng <kito.cheng@sifive.com>
1211 * common/config/riscv/riscv-common.c (multi_letter_subset_rank): Remove
1212 assertion for Z*-ext.
1214 2021-11-09 Jan Hubicka <hubicka@ucw.cz>
1216 * ipa-modref.c (analyze_parms): Add past_flags, past_retslot_flags
1217 and past_static_chain; merge past summary with current one.
1218 (analyze_function): Update.
1220 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
1222 * doc/invoke.texi (Invoking GCC): Document --param=threader-debug.
1224 2021-11-09 Martin Liska <mliska@suse.cz>
1226 * print-rtl.c (rtx_writer::rtx_writer): Make the compilation
1227 conditional based on
1228 * print-rtl.h (class rtx_writer): Likewise.
1230 2021-11-09 Thomas Schwinge <thomas@codesourcery.com>
1232 * input.h (location_hash): Use 'BUILTINS_LOCATION' as spare value
1233 for 'Deleted'. Turn into a '#define'.
1235 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
1237 * gimple-range-path.cc (path_range_query::dump): Clean up.
1238 (path_range_query::compute_ranges): Same.
1239 * value-relation.cc (path_oracle::dump): Same.
1241 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
1243 * dumpfile.c (dump_options): Remove TDF_THREADING entry.
1244 * dumpfile.h (enum dump_flag): Remove TDF_THREADING and adjust
1246 * flag-types.h (enum threader_debug): New.
1247 * gimple-range-path.cc (DEBUG_SOLVER): Use param_threader_debug.
1248 * params.opt: Add entry for --param=threader-debug=.
1250 2021-11-09 Thomas Schwinge <thomas@codesourcery.com>
1252 PR middle-end/101204
1254 * diagnostic-spec.h (typedef xint_hash_t): Turn into...
1255 (struct xint_hash_t): ... this.
1256 * doc/gty.texi: Update.
1258 2021-11-09 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
1260 * config/arm/arm-cpus.in (armv9): New define.
1261 (ARMv9a): New group.
1262 (armv9-a): New arch definition.
1263 * config/arm/arm-tables.opt: Regenerate.
1264 * config/arm/arm.h (BASE_ARCH_9A): New arch enum value.
1265 * config/arm/t-aprofile: Added armv9-a and armv9+simd.
1266 * config/arm/t-arm-elf: Added arm9-a, v9_fps and all_v9_archs
1267 to MULTILIB_MATCHES.
1268 * config/arm/t-multilib: Added v9_a_nosimd_variants and
1269 v9_a_simd_variants to MULTILIB_MATCHES.
1270 * doc/invoke.texi: Update docs.
1272 2021-11-09 Martin Jambor <mjambor@suse.cz>
1275 * ipa-param-manipulation.c (replace_with_mapped_expr): Early
1276 return with error_mark_mode when part of expression is mapped to
1278 (ipa_param_body_adjustments::remap_with_debug_expressions): Set
1279 mapped value to NULL if walk_tree returns error_mark_mode.
1281 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
1283 * tree-ssa-strlen.c (compare_nonzero_chars): Pass statement
1285 (get_addr_stridx): Same.
1287 (get_range_strlen_dynamic): Same.
1288 (handle_builtin_strlen): Same.
1289 (handle_builtin_strchr): Same.
1290 (handle_builtin_strcpy): Same.
1291 (maybe_diag_stxncpy_trunc): Same.
1292 (handle_builtin_stxncpy_strncat): Same.
1293 (handle_builtin_memcpy): Same.
1294 (handle_builtin_strcat): Same.
1295 (handle_alloc_call): Same.
1296 (handle_builtin_memset): Same.
1297 (handle_builtin_string_cmp): Same.
1298 (handle_pointer_plus): Same.
1299 (count_nonzero_bytes_addr): Same.
1300 (count_nonzero_bytes): Same.
1301 (handle_store): Same.
1302 (fold_strstr_to_strncmp): Same.
1303 (handle_integral_assign): Same.
1304 (check_and_optimize_stmt): Same.
1305 (class strlen_dom_walker): Replace evrp with ranger.
1306 (strlen_dom_walker::before_dom_children): Remove evrp.
1307 (strlen_dom_walker::after_dom_children): Remove evrp.
1308 * gimple-ssa-warn-access.cc (maybe_check_access_sizes):
1309 Restrict sprintf output.
1311 2021-11-09 Martin Liska <mliska@suse.cz>
1313 * genconditions.c (write_one_condition): Add const qualifier
1316 2021-11-09 Jeff Law <jeffreyalaw@gmail.com>
1318 * match.pd: New pattern to simplify (1 << n) & M ==/!= 0 for M
1321 2021-11-08 Bill Schmidt <wschmidt@linux.ibm.com>
1323 * config/rs6000/rs6000-builtin-new.def (VEC_INIT_V16QI): Use
1324 escape-newline support.
1325 (VEC_INIT_V4SI): Likewise.
1326 (VEC_INIT_V8HI): Likewise.
1327 (PACK_V1TI): Likewise.
1329 (VFIRSTMISMATCHOREOSINDEX_V16QI): Likewise.
1330 (VFIRSTMISMATCHOREOSINDEX_V8HI): Likewise.
1331 (VFIRSTMISMATCHOREOSINDEX_V4SI): Likewise.
1335 (FMAF128_ODD): Likewise.
1336 (VSCEQPUO): Likewise.
1338 (VSIEQPF): Likewise.
1339 (VSTDCQP): Likewise.
1340 (PACK_TD): Likewise.
1341 (TABORTDC): Likewise.
1342 (TABORTDCI): Likewise.
1343 (SE_LXVRBX): Likewise.
1344 (SE_LXVRHX): Likewise.
1345 (SE_LXVRWX): Likewise.
1346 (SE_LXVRDX): Likewise.
1347 (VREPLACE_UN_UV2DI): Likewise.
1348 (VREPLACE_UN_UV4SI): Likewise.
1349 (VREPLACE_UN_V2DI): Likewise.
1350 (VREPLACE_ELT_UV2DI): Likewise.
1351 (VREPLACE_ELT_V2DI): Likewise.
1352 (ZE_LXVRBX): Likewise.
1353 (ZE_LXVRHX): Likewise.
1354 (ZE_LXVRWX): Likewise.
1355 (ZE_LXVRDX): Likewise.
1357 (CNTLZDM): Likewise.
1358 (CNTTZDM): Likewise.
1361 (PMXVBF16GER2): Likewise.
1362 (PMXVBF16GER2_INTERNAL): Likewise.
1363 (PMXVBF16GER2NN): Likewise.
1364 (PMXVBF16GER2NN_INTERNAL): Likewise.
1365 (PMXVBF16GER2NP): Likewise.
1366 (PMXVBF16GER2NP_INTERNAL): Likewise.
1367 (PMXVBF16GER2PN): Likewise.
1368 (PMXVBF16GER2PN_INTERNAL): Likewise.
1369 (PMXVBF16GER2PP): Likewise.
1370 (PMXVBF16GER2PP_INTERNAL): Likewise.
1371 (PMXVF16GER2): Likewise.
1372 (PMXVF16GER2_INTERNAL): Likewise.
1373 (PMXVF16GER2NN): Likewise.
1374 (PMXVF16GER2NN_INTERNAL): Likewise.
1375 (PMXVF16GER2NP): Likewise.
1376 (PMXVF16GER2NP_INTERNAL): Likewise.
1377 (PMXVF16GER2PN): Likewise.
1378 (PMXVF16GER2PN_INTERNAL): Likewise.
1379 (PMXVF16GER2PP): Likewise.
1380 (PMXVF16GER2PP_INTERNAL): Likewise.
1381 (PMXVF32GER_INTERNAL): Likewise.
1382 (PMXVF32GERNN): Likewise.
1383 (PMXVF32GERNN_INTERNAL): Likewise.
1384 (PMXVF32GERNP): Likewise.
1385 (PMXVF32GERNP_INTERNAL): Likewise.
1386 (PMXVF32GERPN): Likewise.
1387 (PMXVF32GERPN_INTERNAL): Likewise.
1388 (PMXVF32GERPP): Likewise.
1389 (PMXVF32GERPP_INTERNAL): Likewise.
1390 (PMXVF64GER): Likewise.
1391 (PMXVF64GER_INTERNAL): Likewise.
1392 (PMXVF64GERNN): Likewise.
1393 (PMXVF64GERNN_INTERNAL): Likewise.
1394 (PMXVF64GERNP): Likewise.
1395 (PMXVF64GERNP_INTERNAL): Likewise.
1396 (PMXVF64GERPN): Likewise.
1397 (PMXVF64GERPN_INTERNAL): Likewise.
1398 (PMXVF64GERPP): Likewise.
1399 (PMXVF64GERPP_INTERNAL): Likewise.
1400 (PMXVI16GER2): Likewise.
1401 (PMXVI16GER2_INTERNAL): Likewise.
1402 (PMXVI16GER2PP): Likewise.
1403 (PMXVI16GER2PP_INTERNAL): Likewise.
1404 (PMXVI16GER2S): Likewise.
1405 (PMXVI16GER2S_INTERNAL): Likewise.
1406 (PMXVI16GER2SPP): Likewise.
1407 (PMXVI16GER2SPP_INTERNAL): Likewise.
1408 (PMXVI4GER8): Likewise.
1409 (PMXVI4GER8_INTERNAL): Likewise.
1410 (PMXVI4GER8PP): Likewise.
1411 (PMXVI4GER8PP_INTERNAL): Likewise.
1412 (PMXVI8GER4): Likewise.
1413 (PMXVI8GER4_INTERNAL): Likewise.
1414 (PMXVI8GER4PP): Likewise.
1415 (PMXVI8GER4PP_INTERNAL): Likewise.
1416 (PMXVI8GER4SPP): Likewise.
1417 (PMXVI8GER4SPP_INTERNAL): Likewise.
1418 * config/rs6000/rs6000-gen-builtins.c (MAXLINES): New macro.
1419 (linebuf): Increase size.
1420 (lines): New variable.
1421 (lastline): Likewise.
1422 (real_line_pos): New function.
1423 (diag): Change signature.
1424 (bif_diag): Change signature; support escape-newline handling.
1425 (ovld_diag): Likewise.
1426 (fatal): Move earlier.
1427 (consume_whitespace): Adjust diag call.
1428 (advance_line): Add escape-newline handling; call fatal.
1429 (safe_inc_pos): Adjust diag call.
1430 (match_identifier): Likewise.
1431 (match_integer): Likewise.
1432 (match_to_right_bracket): Call fatal instead of diag; adjust diag
1434 (match_basetype): Adjust diag calls.
1435 (match_bracketed_pair): Likewise.
1436 (match_const_restriction): Likewise.
1437 (match_type): Likewise.
1438 (parse_args): Likewise.
1439 (parse_bif_attrs): Likewise.
1440 (complete_vector_type): Likewise.
1441 (complete_base_type): Likewise.
1442 (parse_prototype): Likewise.
1443 (parse_bif_entry): Likewise.
1444 (parse_bif_stanza): Likewise.
1445 (parse_ovld_entry): Likewise.
1446 (parse_ovld_stanza): Likewise.
1447 (main): Allocate buffers for lines[].
1449 2021-11-08 Eric Botcazou <ebotcazou@adacore.com>
1451 * config/i386/i386.c (legitimize_pic_address): Adjust comment and
1452 use the REG argument on the CM_LARGE_PIC code path as well.
1453 * config/i386/predicates.md (gotoff_operand): Do not treat VxWorks
1454 specially with the large code models.
1456 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
1458 * gimple.c (gimple_call_static_chain_flags): Revert the workaround
1459 allowing interposition since issues with binds_to_local_def were
1462 2021-11-08 Andrew MacLeod <amacleod@redhat.com>
1464 PR tree-optimization/103122
1465 * gimple-range.cc (gimple_ranger::range_of_expr): Request the cache
1466 entry with "calulate new values" set to false.
1468 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
1470 * builtins.c (is_simple_builtin): Add builitin_dwarf_cfa
1471 and builtin_return_address.
1472 (builtin_fnspec): Annotate builtin_return,
1473 bulitin_eh_pointer, builtin_eh_filter, builtin_unwind_resume,
1474 builtin_cxa_end_cleanup, builtin_eh_copy_values,
1475 builtin_frame_address, builtin_apply_args,
1476 builtin_asan_before_dynamic_init, builtin_asan_after_dynamic_init,
1477 builtin_prefetch, builtin_dwarf_cfa, builtin_return_addrss
1479 * ipa-pure-const.c (special_builtin_state): Add builtin_dwarf_cfa
1480 and builtin_return_address.
1482 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
1484 PR tree-optimization/103177
1485 * passes.def: Move uncprop after pure/const and modref.
1487 2021-11-08 Martin Jambor <mjambor@suse.cz>
1491 * tree-inline.c (remap_gimple_stmt): Unshare the expression without
1492 location before invoking remap_with_debug_expressions on it.
1493 * ipa-param-manipulation.c
1494 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
1496 2021-11-08 David Edelsohn <dje.gcc@gmail.com>
1498 * config/rs6000/vsx.md (vsx_splat_v4si_di): Revert "wa"
1501 2021-11-08 Richard Biener <rguenther@suse.de>
1503 * cfgloop.c (verify_loop_structure): Use a temporary BB flag
1504 instead of an sbitmap to cache irreducible state.
1506 2021-11-08 Aldy Hernandez <aldyh@redhat.com>
1508 PR tree-optimization/103120
1509 * gimple-range-path.cc (path_range_query::range_defined_in_block):
1510 Bail if there's a cache entry.
1512 2021-11-08 Bill Schmidt <wschmidt@linux.ibm.com>
1514 * config/rs6000/rs6000.c (rs6000_builtin_reciprocal): Use
1515 rs6000_builtin_decls_x when appropriate.
1516 (add_condition_to_bb): Likewise.
1517 (rs6000_atomic_assign_expand_fenv): Likewise.
1519 2021-11-08 Bill Schmidt <wschmidt@linux.ibm.com>
1521 * config/rs6000/rs6000-call.c (rs6000_new_builtin_decl): New function.
1522 (rs6000_builtin_decl): Call it.
1524 2021-11-08 Martin Liska <mliska@suse.cz>
1526 * cgraph.c (cgraph_node::dump): Dump it from decl.
1528 2021-11-08 Martin Liska <mliska@suse.cz>
1530 PR gcov-profile/100520
1531 * coverage.c (coverage_compute_profile_id): Strip .gk when
1532 compare debug is used.
1533 * system.h (endswith): New function.
1535 2021-11-08 Martin Liska <mliska@suse.cz>
1537 * cgraph.c (cgraph_node::dump): Dump static_chain_decl.
1539 2021-11-08 Thomas Schwinge <thomas@codesourcery.com>
1541 * config/rs6000/rbtree.c: Fix 'Copyright (C) 2020-21' into '2020-2021'
1542 * config/rs6000/rbtree.h: Likewise.
1543 * config/rs6000/rs6000-builtin-new.def: Likewise.
1544 * config/rs6000/rs6000-gen-builtins.c: Likewise.
1545 * config/rs6000/rs6000-overload.def: Likewise.
1547 2021-11-08 Richard Sandiford <richard.sandiford@arm.com>
1549 * config/aarch64/aarch64.c (aarch64_ldn_stn_vectors): New function.
1550 (aarch64_address_cost): Use it instead of testing for CImode and
1553 2021-11-08 Richard Sandiford <richard.sandiford@arm.com>
1555 * genmodes.c (emit_insn_modes_h): Define NUM_MODE_* macros.
1556 * expmed.h (NUM_MODE_INT): Delete in favor of genmodes definitions.
1557 (NUM_MODE_PARTIAL_INT, NUM_MODE_VECTOR_INT): Likewise.
1558 * real.h (real_format_for_mode): Use NUM_MODE_FLOAT and
1559 NUM_MODE_DECIMAL_FLOAT.
1560 (REAL_MODE_FORMAT): Likewise.
1562 2021-11-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
1564 * tree-vect-loop.c (vect_better_loop_vinfo_p): Change how epilogue loop
1567 2021-11-08 Richard Biener <rguenther@suse.de>
1569 * tree-vectorizer.h (vect_create_loop_vinfo): Add main_loop_info
1571 * tree-vect-loop.c (vect_create_loop_vinfo): Likewise. Set
1572 LOOP_VINFO_ORIG_LOOP_INFO and conditionalize set of
1573 LOOP_VINFO_NITERS_ASSUMPTIONS.
1574 (vect_analyze_loop_1): Adjust.
1575 (vect_analyze_loop): Move loop constraint setting and
1576 SCEV/niter reset here from vect_create_loop_vinfo to perform
1578 (vect_analyze_loop_form): Move dumping of symbolic niters
1579 here from vect_create_loop_vinfo.
1581 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
1583 * ipa-modref-tree.h (enum modref_special_parms): New enum.
1584 (struct modref_access_node): update for special parms.
1585 (struct modref_ref_node): Likewise.
1586 (struct modref_parm_map): Likewise.
1587 (struct modref_tree): Likewise.
1588 * ipa-modref.c (dump_access): Likewise.
1589 (get_access): Detect static chain.
1590 (parm_map_for_arg): Take tree as arg instead of
1592 (merge_call_side_effects): Compute map for static chain.
1593 (process_fnspec): Update.
1594 (struct escape_point): Remove retslot_arg and static_chain_arg.
1595 (analyze_parms): Update.
1596 (compute_parm_map): Update.
1597 (propagate_unknown_call): Update.
1598 (modref_propagate_in_scc): Update.
1599 (modref_merge_call_site_flags): Update.
1600 (ipa_merge_modref_summary_after_inlining): Update.
1601 * tree-ssa-alias.c (modref_may_conflict): Handle static chain.
1602 * ipa-modref-tree.c (test_merge): Update.
1604 2021-11-08 Haochen Gui <guihaoc@gcc.gnu.org>
1606 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin): Disable
1607 gimple fold for VSX_BUILTIN_XVMINDP, ALTIVEC_BUILTIN_VMINFP,
1608 VSX_BUILTIN_XVMAXDP, ALTIVEC_BUILTIN_VMAXFP when fast-math is not
1611 2021-11-08 liuhongt <hongtao.liu@intel.com>
1613 PR tree-optimization/103077
1614 * doc/invoke.texi (Options That Control Optimization):
1615 Update documentation for -ftree-loop-vectorize and
1616 -ftree-slp-vectorize which are enabled by default at -02.
1618 2021-11-08 liuhongt <hongtao.liu@intel.com>
1621 * match.pd (Simplifcation (trunc)copysign((extend)a, (extend)b)
1622 to .COPYSIGN (a, b)): Add !HONOR_SNANS.
1624 2021-11-08 liuhongt <hongtao.liu@intel.com>
1627 * match.pd: Simplify
1628 (trunc)fma ((extend)a, (extend)b, (extend)c) to IFN_FMA (a, b,
1629 c) under flag_unsafe_math_optimizations.
1631 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
1634 * params.opt (modref-max-depth): Add range.
1635 (modref-max-adjustments): Fix range.
1637 2021-11-07 Aldy Hernandez <aldyh@redhat.com>
1639 * tree-pass.h (make_pass_vrp_threader): Remove.
1640 * tree-ssa-threadbackward.c
1641 (back_threader_profitability::profitable_path_p): Remove
1642 ASSERT_EXPR references.
1643 * tree-ssa-threadedge.c (jt_state::register_equivs_stmt): Same.
1644 * tree-vrp.c (vrp_folder::simplify_casted_conds): Same.
1645 (execute_vrp): Same.
1646 (class hybrid_threader): Remove.
1647 (hybrid_threader::hybrid_threader): Remove.
1648 (hybrid_threader::~hybrid_threader): Remove.
1649 (hybrid_threader::before_dom_children): Remove.
1650 (hybrid_threader::after_dom_children): Remove.
1651 (execute_vrp_threader): Remove.
1652 (class pass_vrp_threader): Remove.
1653 (make_pass_vrp_threader): Remove.
1655 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
1657 * gimple.c (gimple_call_arg_flags): Use interposable_eaf_flags.
1658 (gimple_call_retslot_flags): Likewise.
1659 (gimple_call_static_chain_flags): Likewise.
1660 * ipa-modref.c (remove_useless_eaf_flags): Do not remove everything for
1662 (modref_summary::useful_p): Likewise.
1663 (modref_summary_lto::useful_p): Likewise.
1664 (analyze_parms): Do not give up on NOVOPS.
1665 (analyze_function): When dumping report chnages in EAF flags
1666 between IPA and local pass.
1667 (modref_merge_call_site_flags): Compute implicit eaf flags
1668 based on callee ecf_flags and fnspec; if the function does not
1669 bind to current defs use interposable_eaf_flags.
1670 (modref_propagate_flags_in_scc): Update.
1671 * ipa-modref.h (interposable_eaf_flags): New function.
1673 2021-11-07 Bill Schmidt <wschmidt@linux.ibm.com>
1675 * config/rs6000/rs6000-call.c (rs6000_expand_new_builtin): New
1677 (rs6000_invalid_new_builtin): New function.
1678 (rs6000_expand_builtin): Call rs6000_expand_new_builtin.
1679 (rs6000_expand_ldst_mask): New function.
1680 (new_cpu_expand_builtin): Likewise.
1681 (elemrev_icode): Likewise.
1682 (ldv_expand_builtin): Likewise.
1683 (lxvrse_expand_builtin): Likewise.
1684 (lxvrze_expand_builtin): Likewise.
1685 (stv_expand_builtin): Likewise.
1686 (new_mma_expand_builtin): Likewise.
1687 (new_htm_spr_num): Likewise.
1688 (new_htm_expand_builtin): Likewise.
1689 (rs6000_expand_new_builtin): Likewise.
1690 (rs6000_init_builtins): Initialize altivec_builtin_mask_for_load.
1692 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
1694 * ipa-modref.c (modref_lattice): Add do_dataflow,
1695 changed and propagate_to fields.
1696 (modref_lattice::release): Free propagate_to
1697 (modref_lattice::merge): Do not give up early on unknown
1699 (modref_lattice::merge_deref): Likewise.
1700 (modref_eaf_analysis): Update toplevel comment.
1701 (modref_eaf_analysis::analyze_ssa_name): Record postponned ssa names;
1702 do optimistic dataflow initialization.
1703 (modref_eaf_analysis::merge_with_ssa_name): Build dataflow graph.
1704 (modref_eaf_analysis::propagate): New member function.
1705 (analyze_parms): Update to new API of modref_eaf_analysis.
1707 2021-11-06 Jan Hubicka <hubicka@ucw.cz>
1709 * cgraph.h (cgraph_node::can_be_discarded_p): Do not
1710 return true on functions from other partition.
1712 2021-11-06 David Edelsohn <dje.gcc@gmail.com>
1714 * config/rs6000/vsx.md (vsx_splat_v4si): Change constraints to "wa".
1715 (vsx_splat_v4si_di): Change constraint to "wa".
1717 2021-11-06 Aldy Hernandez <aldyh@redhat.com>
1719 PR tree-optimization/103061
1720 * value-relation.cc (path_oracle::path_oracle): Initialize
1722 (path_oracle::killing_def): Set m_killed_defs.
1723 (path_oracle::query_relation): Do not look at the root oracle for
1725 * value-relation.h (class path_oracle): Add m_killed_defs.
1727 2021-11-06 Aldy Hernandez <aldyh@redhat.com>
1729 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
1730 Remove gotos and other cleanups.
1732 2021-11-05 Jan Hubicka <hubicka@ucw.cz>
1735 * ipa-modref-tree.h (modref_tree::insert): Do nothing for
1736 paradoxical and zero sized accesses.
1738 2021-11-05 Jan Hubicka <hubicka@ucw.cz>
1741 * ipa-modref-tree.h (struct modref_access_node): Avoid left shift
1744 2021-11-05 Iain Sandoe <iain@sandoe.co.uk>
1746 * config/darwin.h (ASM_GENERATE_INTERNAL_LABEL): Add LTRAMP
1747 to the list of symbol prefixes that must be made linker-
1750 2021-11-05 Iain Sandoe <iain@sandoe.co.uk>
1752 * config.host: Add support for aarch64-*-darwin.
1753 * config/aarch64/host-aarch64-darwin.c: New file.
1754 * config/aarch64/x-darwin: New file.
1756 2021-11-05 Andrew MacLeod <amacleod@redhat.com>
1758 PR tree-optimization/103093
1759 * gimple-range-gori.cc (range_def_chain::get_imports): Remove assert.
1761 2021-11-05 Andrew MacLeod <amacleod@redhat.com>
1763 PR tree-optimization/102943
1764 * gimple-range-cache.cc (class update_list): New.
1765 (update_list::add): Replace add_to_update.
1766 (update_list::pop): New.
1767 (ranger_cache::ranger_cache): Adjust.
1768 (ranger_cache::~ranger_cache): Adjust.
1769 (ranger_cache::add_to_update): Delete.
1770 (ranger_cache::propagate_cache): Adjust to new class.
1771 (ranger_cache::propagate_updated_value): Ditto.
1772 (ranger_cache::fill_block_cache): Ditto.
1773 * gimple-range-cache.h (class ranger_cache): Adjust to update class.
1775 2021-11-05 Richard Biener <rguenther@suse.de>
1777 * tree-vect-loop.c (vect_analyze_loop): Remove obsolete
1778 comment and expand on another one. Combine nested if.
1780 2021-11-05 John David Anglin <danglin@gcc.gnu.org>
1783 * config/pa/pa-modes.def: Add OImode integer type.
1784 * config/pa/pa.c (pa_scalar_mode_supported_p): Allow TImode
1786 * config/pa/pa.h (MIN_UNITS_PER_WORD) Define to MIN_UNITS_PER_WORD
1787 to UNITS_PER_WORD if IN_LIBGCC2.
1788 * config/pa/pa.md (addti3, addvti3, subti3, subvti3, negti2,
1789 negvti2, ashlti3, shrpd_internal): New patterns.
1790 Change some multi instruction types to multi.
1792 2021-11-05 Jakub Jelinek <jakub@redhat.com>
1795 * config/i386/i386.h
1796 (stringop_algs::stringop_strategy::stringop_strategy): Make the ctor
1799 2021-11-05 Wilco Dijkstra <wdijkstr@arm.com>
1802 * config/aarch64/aarch64.c (aarch64_mov_operand_p): Strip the salt
1804 * config/aarch64/constraints.md: Support const in Usw.
1806 2021-11-05 John David Anglin <danglin@gcc.gnu.org>
1808 * config/pa/pa.h (PREFERRED_DEBUGGING_TYPE): Define to DWARF2_DEBUG.
1809 * config/pa/pa64-hpux.h (PREFERRED_DEBUGGING_TYPE): Remove define.
1811 2021-11-05 Richard Biener <rguenther@suse.de>
1813 * tree-vectorizer.h (vec_info_shared::n_stmts): Add.
1814 (LOOP_VINFO_N_STMTS): Likewise.
1815 (vec_info_for_bb): Remove unused function.
1816 * tree-vectorizer.c (vec_info_shared::vec_info_shared):
1817 Initialize n_stmts member.
1818 * tree-vect-loop.c: Remove INCLUDE_FUNCTIONAL.
1819 (vect_create_loop_vinfo): Do not set loop->aux.
1820 (vect_analyze_loop_2): Do not get n_stmts as argument,
1821 instead use LOOP_VINFO_N_STMTS. Set LOOP_VINFO_VECTORIZABLE_P
1823 (vect_analyze_loop_1): Remove callback, get the mode iterator
1824 and autodetected_vector_mode as argument, advancing the
1825 iterator and initializing autodetected_vector_mode here.
1826 (vect_analyze_loop): Split analysis loop into two, first
1827 processing main loops only and then epilogues.
1829 2021-11-05 Martin Jambor <mjambor@suse.cz>
1831 * ipa-prop.c (compute_complex_assign_jump_func): Remove
1832 unnecessary check for RECORD_TYPE.
1834 2021-11-05 Gerald Pfeifer <gerald@pfeifer.com>
1836 * doc/install.texi2html: Do not generate old.html any longer.
1838 2021-11-05 Martin Liska <mliska@suse.cz>
1841 * opts.c (finish_options): Reset flag_gtoggle when it is used.
1843 2021-11-05 Jakub Jelinek <jakub@redhat.com>
1846 * dwarf2out.c (add_const_value_attribute): Add MODE argument, use it
1847 in CONST_WIDE_INT handling. Adjust recursive calls.
1848 (add_location_or_const_value_attribute): Pass DECL_MODE (decl) to
1849 new add_const_value_attribute argument.
1850 (tree_add_const_value_attribute): Pass TYPE_MODE (type) to new
1851 add_const_value_attribute argument.
1853 2021-11-05 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
1855 * config/vx-common.h: Test value of TARGET_VXWORKS7 rather
1858 2021-11-05 Richard Biener <rguenther@suse.de>
1860 * tree-vectorizer.h (struct vect_loop_form_info): New.
1861 (vect_analyze_loop_form): Adjust.
1862 (vect_create_loop_vinfo): New.
1863 * tree-parloops.c (gather_scalar_reductions): Adjust for
1864 vect_analyze_loop_form API change.
1865 * tree-vect-loop.c: Include <functional>.
1866 (vect_analyze_loop_form_1): Rename to vect_analyze_loop_form,
1867 take struct vect_loop_form_info as output parameter and adjust.
1868 (vect_analyze_loop_form): Rename to vect_create_loop_vinfo and
1869 split out call to the original vect_analyze_loop_form_1.
1870 (vect_reanalyze_as_main_loop): Rename to...
1871 (vect_analyze_loop_1): ... this, factor out the call to
1872 vect_analyze_loop_form and generalize to be able to use it twice ...
1873 (vect_analyze_loop): ... here. Perform vect_analyze_loop_form
1876 2021-11-05 Xionghu Luo <luoxhu@linux.ibm.com>
1879 * config/rs6000/fusion.md: Regenerate.
1880 * config/rs6000/genfusion.pl: Fix incorrect clobber constraint.
1882 2021-11-04 Andreas Krebbel <krebbel@linux.ibm.com>
1884 * config/s390/s390.h (STACK_CHECK_MOVING_SP): New macro
1887 2021-11-04 Tamar Christina <tamar.christina@arm.com>
1889 * config/aarch64/aarch64-builtins.c
1890 (aarch64_general_gimple_fold_builtin): Add ashl, sshl, ushl, ashr,
1891 ashr_simd, lshr, lshr_simd.
1892 * config/aarch64/aarch64-simd-builtins.def (lshr): Use USHIFTIMM.
1893 * config/aarch64/arm_neon.h (vshr_n_u8, vshr_n_u16, vshr_n_u32,
1894 vshrq_n_u8, vshrq_n_u16, vshrq_n_u32, vshrq_n_u64): Fix type hack.
1896 2021-11-04 Tamar Christina <tamar.christina@arm.com>
1898 * match.pd: New negate+shift pattern.
1900 2021-11-04 Andrew MacLeod <amacleod@redhat.com>
1902 PR tree-optimization/103079
1903 * gimple-range-gori.cc (gimple_range_calc_op1): Treat undefined as
1905 (gimple_range_calc_op2): Ditto.
1907 2021-11-04 Martin Jambor <mjambor@suse.cz>
1910 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
1911 members remap_with_debug_expressions, m_dead_ssa_debug_equiv,
1912 m_dead_stmt_debug_equiv and prepare_debug_expressions. Added
1913 parameter to mark_dead_statements.
1914 * ipa-param-manipulation.c: Include tree-phinodes.h and cfgexpand.h.
1915 (ipa_param_body_adjustments::mark_dead_statements): New parameter
1916 debugstack, push into it all SSA names used in debug statements,
1917 produce m_dead_ssa_debug_equiv mapping for the removed param.
1918 (replace_with_mapped_expr): New function.
1919 (ipa_param_body_adjustments::remap_with_debug_expressions): Likewise.
1920 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
1921 (ipa_param_body_adjustments::common_initialization): Gather and
1922 procecc SSA which will be removed but are in debug statements. Simplify.
1923 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
1925 * tree-inline.c (remap_gimple_stmt): Create a debug bind when possible
1926 when avoiding a copy of an unnecessary statement. Remap removed SSA
1927 names in existing debug statements.
1928 (tree_function_versioning): Do not create DEBUG_EXPR_DECL for removed
1929 parameters if we have already done so.
1931 2021-11-04 Jan Hubicka <hubicka@ucw.cz>
1934 * gimple.c (gimple_call_static_chain_flags): Handle case when
1935 nested function does not bind locally.
1937 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
1939 * config/aarch64/aarch64.c (aarch64_function_value): Generate
1940 a register rtx for Neon vector-tuple modes.
1941 (aarch64_layout_arg): Likewise.
1943 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
1945 * lower-subreg.c (simple_move): Prevent decomposition if
1946 modes are not tieable.
1948 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
1949 Richard Sandiford <richard.sandiford@arm.com>
1951 * config/aarch64/aarch64-builtins.c (v2x8qi_UP): Define.
1952 (v2x4hi_UP): Likewise.
1953 (v2x4hf_UP): Likewise.
1954 (v2x4bf_UP): Likewise.
1955 (v2x2si_UP): Likewise.
1956 (v2x2sf_UP): Likewise.
1957 (v2x1di_UP): Likewise.
1958 (v2x1df_UP): Likewise.
1959 (v2x16qi_UP): Likewise.
1960 (v2x8hi_UP): Likewise.
1961 (v2x8hf_UP): Likewise.
1962 (v2x8bf_UP): Likewise.
1963 (v2x4si_UP): Likewise.
1964 (v2x4sf_UP): Likewise.
1965 (v2x2di_UP): Likewise.
1966 (v2x2df_UP): Likewise.
1967 (v3x8qi_UP): Likewise.
1968 (v3x4hi_UP): Likewise.
1969 (v3x4hf_UP): Likewise.
1970 (v3x4bf_UP): Likewise.
1971 (v3x2si_UP): Likewise.
1972 (v3x2sf_UP): Likewise.
1973 (v3x1di_UP): Likewise.
1974 (v3x1df_UP): Likewise.
1975 (v3x16qi_UP): Likewise.
1976 (v3x8hi_UP): Likewise.
1977 (v3x8hf_UP): Likewise.
1978 (v3x8bf_UP): Likewise.
1979 (v3x4si_UP): Likewise.
1980 (v3x4sf_UP): Likewise.
1981 (v3x2di_UP): Likewise.
1982 (v3x2df_UP): Likewise.
1983 (v4x8qi_UP): Likewise.
1984 (v4x4hi_UP): Likewise.
1985 (v4x4hf_UP): Likewise.
1986 (v4x4bf_UP): Likewise.
1987 (v4x2si_UP): Likewise.
1988 (v4x2sf_UP): Likewise.
1989 (v4x1di_UP): Likewise.
1990 (v4x1df_UP): Likewise.
1991 (v4x16qi_UP): Likewise.
1992 (v4x8hi_UP): Likewise.
1993 (v4x8hf_UP): Likewise.
1994 (v4x8bf_UP): Likewise.
1995 (v4x4si_UP): Likewise.
1996 (v4x4sf_UP): Likewise.
1997 (v4x2di_UP): Likewise.
1998 (v4x2df_UP): Likewise.
1999 (TYPES_GETREGP): Delete.
2000 (TYPES_SETREGP): Likewise.
2001 (TYPES_LOADSTRUCT_U): Define.
2002 (TYPES_LOADSTRUCT_P): Likewise.
2003 (TYPES_LOADSTRUCT_LANE_U): Likewise.
2004 (TYPES_LOADSTRUCT_LANE_P): Likewise.
2005 (TYPES_STORE1P): Move for consistency.
2006 (TYPES_STORESTRUCT_U): Define.
2007 (TYPES_STORESTRUCT_P): Likewise.
2008 (TYPES_STORESTRUCT_LANE_U): Likewise.
2009 (TYPES_STORESTRUCT_LANE_P): Likewise.
2010 (aarch64_simd_tuple_types): Define.
2011 (aarch64_lookup_simd_builtin_type): Handle tuple type lookup.
2012 (aarch64_init_simd_builtin_functions): Update frontend lookup
2013 for builtin functions after handling arm_neon.h pragma.
2014 (register_tuple_type): Manually set modes of single-integer
2015 tuple types. Record tuple types.
2016 * config/aarch64/aarch64-modes.def
2017 (ADV_SIMD_D_REG_STRUCT_MODES): Define D-register tuple modes.
2018 (ADV_SIMD_Q_REG_STRUCT_MODES): Define Q-register tuple modes.
2019 (SVE_MODES): Give single-vector modes priority over vector-
2021 (VECTOR_MODES_WITH_PREFIX): Set partial-vector mode order to
2022 be after all single-vector modes.
2023 * config/aarch64/aarch64-simd-builtins.def: Update builtin
2024 generator macros to reflect modifications to the backend
2026 * config/aarch64/aarch64-simd.md (aarch64_simd_ld2<mode>):
2027 Use vector-tuple mode iterator and rename to...
2028 (aarch64_simd_ld2<vstruct_elt>): This.
2029 (aarch64_simd_ld2r<mode>): Use vector-tuple mode iterator and
2031 (aarch64_simd_ld2r<vstruct_elt>): This.
2032 (aarch64_vec_load_lanesoi_lane<mode>): Use vector-tuple mode
2033 iterator and rename to...
2034 (aarch64_vec_load_lanes<mode>_lane<vstruct_elt>): This.
2035 (vec_load_lanesoi<mode>): Use vector-tuple mode iterator and
2037 (vec_load_lanes<mode><vstruct_elt>): This.
2038 (aarch64_simd_st2<mode>): Use vector-tuple mode iterator and
2040 (aarch64_simd_st2<vstruct_elt>): This.
2041 (aarch64_vec_store_lanesoi_lane<mode>): Use vector-tuple mode
2042 iterator and rename to...
2043 (aarch64_vec_store_lanes<mode>_lane<vstruct_elt>): This.
2044 (vec_store_lanesoi<mode>): Use vector-tuple mode iterator and
2046 (vec_store_lanes<mode><vstruct_elt>): This.
2047 (aarch64_simd_ld3<mode>): Use vector-tuple mode iterator and
2049 (aarch64_simd_ld3<vstruct_elt>): This.
2050 (aarch64_simd_ld3r<mode>): Use vector-tuple mode iterator and
2052 (aarch64_simd_ld3r<vstruct_elt>): This.
2053 (aarch64_vec_load_lanesci_lane<mode>): Use vector-tuple mode
2054 iterator and rename to...
2055 (vec_load_lanesci<mode>): This.
2056 (aarch64_simd_st3<mode>): Use vector-tuple mode iterator and
2058 (aarch64_simd_st3<vstruct_elt>): This.
2059 (aarch64_vec_store_lanesci_lane<mode>): Use vector-tuple mode
2060 iterator and rename to...
2061 (vec_store_lanesci<mode>): This.
2062 (aarch64_simd_ld4<mode>): Use vector-tuple mode iterator and
2064 (aarch64_simd_ld4<vstruct_elt>): This.
2065 (aarch64_simd_ld4r<mode>): Use vector-tuple mode iterator and
2067 (aarch64_simd_ld4r<vstruct_elt>): This.
2068 (aarch64_vec_load_lanesxi_lane<mode>): Use vector-tuple mode
2069 iterator and rename to...
2070 (vec_load_lanesxi<mode>): This.
2071 (aarch64_simd_st4<mode>): Use vector-tuple mode iterator and
2073 (aarch64_simd_st4<vstruct_elt>): This.
2074 (aarch64_vec_store_lanesxi_lane<mode>): Use vector-tuple mode
2075 iterator and rename to...
2076 (vec_store_lanesxi<mode>): This.
2077 (mov<mode>): Define for Neon vector-tuple modes.
2078 (aarch64_ld1x3<VALLDIF:mode>): Use vector-tuple mode iterator
2080 (aarch64_ld1x3<vstruct_elt>): This.
2081 (aarch64_ld1_x3_<mode>): Use vector-tuple mode iterator and
2083 (aarch64_ld1_x3_<vstruct_elt>): This.
2084 (aarch64_ld1x4<VALLDIF:mode>): Use vector-tuple mode iterator
2086 (aarch64_ld1x4<vstruct_elt>): This.
2087 (aarch64_ld1_x4_<mode>): Use vector-tuple mode iterator and
2089 (aarch64_ld1_x4_<vstruct_elt>): This.
2090 (aarch64_st1x2<VALLDIF:mode>): Use vector-tuple mode iterator
2092 (aarch64_st1x2<vstruct_elt>): This.
2093 (aarch64_st1_x2_<mode>): Use vector-tuple mode iterator and
2095 (aarch64_st1_x2_<vstruct_elt>): This.
2096 (aarch64_st1x3<VALLDIF:mode>): Use vector-tuple mode iterator
2098 (aarch64_st1x3<vstruct_elt>): This.
2099 (aarch64_st1_x3_<mode>): Use vector-tuple mode iterator and
2101 (aarch64_st1_x3_<vstruct_elt>): This.
2102 (aarch64_st1x4<VALLDIF:mode>): Use vector-tuple mode iterator
2104 (aarch64_st1x4<vstruct_elt>): This.
2105 (aarch64_st1_x4_<mode>): Use vector-tuple mode iterator and
2107 (aarch64_st1_x4_<vstruct_elt>): This.
2108 (*aarch64_mov<mode>): Define for vector-tuple modes.
2109 (*aarch64_be_mov<mode>): Likewise.
2110 (aarch64_ld<VSTRUCT:nregs>r<VALLDIF:mode>): Use vector-tuple
2111 mode iterator and rename to...
2112 (aarch64_ld<nregs>r<vstruct_elt>): This.
2113 (aarch64_ld2<mode>_dreg): Use vector-tuple mode iterator and
2115 (aarch64_ld2<vstruct_elt>_dreg): This.
2116 (aarch64_ld3<mode>_dreg): Use vector-tuple mode iterator and
2118 (aarch64_ld3<vstruct_elt>_dreg): This.
2119 (aarch64_ld4<mode>_dreg): Use vector-tuple mode iterator and
2121 (aarch64_ld4<vstruct_elt>_dreg): This.
2122 (aarch64_ld<VSTRUCT:nregs><VDC:mode>): Use vector-tuple mode
2123 iterator and rename to...
2124 (aarch64_ld<nregs><vstruct_elt>): Use vector-tuple mode
2125 iterator and rename to...
2126 (aarch64_ld<VSTRUCT:nregs><VQ:mode>): Use vector-tuple mode
2127 (aarch64_ld1x2<VQ:mode>): Delete.
2128 (aarch64_ld1x2<VDC:mode>): Use vector-tuple mode iterator and
2130 (aarch64_ld1x2<vstruct_elt>): This.
2131 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use vector-
2132 tuple mode iterator and rename to...
2133 (aarch64_ld<nregs>_lane<vstruct_elt>): This.
2134 (aarch64_get_dreg<VSTRUCT:mode><VDC:mode>): Delete.
2135 (aarch64_get_qreg<VSTRUCT:mode><VQ:mode>): Likewise.
2136 (aarch64_st2<mode>_dreg): Use vector-tuple mode iterator and
2138 (aarch64_st2<vstruct_elt>_dreg): This.
2139 (aarch64_st3<mode>_dreg): Use vector-tuple mode iterator and
2141 (aarch64_st3<vstruct_elt>_dreg): This.
2142 (aarch64_st4<mode>_dreg): Use vector-tuple mode iterator and
2144 (aarch64_st4<vstruct_elt>_dreg): This.
2145 (aarch64_st<VSTRUCT:nregs><VDC:mode>): Use vector-tuple mode
2146 iterator and rename to...
2147 (aarch64_st<nregs><vstruct_elt>): This.
2148 (aarch64_st<VSTRUCT:nregs><VQ:mode>): Use vector-tuple mode
2149 iterator and rename to aarch64_st<nregs><vstruct_elt>.
2150 (aarch64_st<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use vector-
2151 tuple mode iterator and rename to...
2152 (aarch64_st<nregs>_lane<vstruct_elt>): This.
2153 (aarch64_set_qreg<VSTRUCT:mode><VQ:mode>): Delete.
2154 (aarch64_simd_ld1<mode>_x2): Use vector-tuple mode iterator
2156 (aarch64_simd_ld1<vstruct_elt>_x2): This.
2157 * config/aarch64/aarch64.c (aarch64_advsimd_struct_mode_p):
2158 Refactor to include new vector-tuple modes.
2159 (aarch64_classify_vector_mode): Add cases for new vector-
2161 (aarch64_advsimd_partial_struct_mode_p): Define.
2162 (aarch64_advsimd_full_struct_mode_p): Likewise.
2163 (aarch64_advsimd_vector_array_mode): Likewise.
2164 (aarch64_sve_data_mode): Change location in file.
2165 (aarch64_array_mode): Handle case of Neon vector-tuple modes.
2166 (aarch64_hard_regno_nregs): Handle case of partial Neon
2168 (aarch64_classify_address): Refactor to include handling of
2169 Neon vector-tuple modes.
2170 (aarch64_print_operand): Print "d" for "%R" for a partial
2171 Neon vector structure.
2172 (aarch64_expand_vec_perm_1): Use new vector-tuple mode.
2173 (aarch64_modes_tieable_p): Prevent tieing Neon partial struct
2174 modes with scalar machines modes larger than 8 bytes.
2175 (aarch64_can_change_mode_class): Don't allow changes between
2176 partial and full Neon vector-structure modes.
2177 * config/aarch64/arm_neon.h (vst2_lane_f16): Use updated
2178 builtin and remove boiler-plate code for opaque mode.
2179 (vst2_lane_f32): Likewise.
2180 (vst2_lane_f64): Likewise.
2181 (vst2_lane_p8): Likewise.
2182 (vst2_lane_p16): Likewise.
2183 (vst2_lane_p64): Likewise.
2184 (vst2_lane_s8): Likewise.
2185 (vst2_lane_s16): Likewise.
2186 (vst2_lane_s32): Likewise.
2187 (vst2_lane_s64): Likewise.
2188 (vst2_lane_u8): Likewise.
2189 (vst2_lane_u16): Likewise.
2190 (vst2_lane_u32): Likewise.
2191 (vst2_lane_u64): Likewise.
2192 (vst2q_lane_f16): Likewise.
2193 (vst2q_lane_f32): Likewise.
2194 (vst2q_lane_f64): Likewise.
2195 (vst2q_lane_p8): Likewise.
2196 (vst2q_lane_p16): Likewise.
2197 (vst2q_lane_p64): Likewise.
2198 (vst2q_lane_s8): Likewise.
2199 (vst2q_lane_s16): Likewise.
2200 (vst2q_lane_s32): Likewise.
2201 (vst2q_lane_s64): Likewise.
2202 (vst2q_lane_u8): Likewise.
2203 (vst2q_lane_u16): Likewise.
2204 (vst2q_lane_u32): Likewise.
2205 (vst2q_lane_u64): Likewise.
2206 (vst3_lane_f16): Likewise.
2207 (vst3_lane_f32): Likewise.
2208 (vst3_lane_f64): Likewise.
2209 (vst3_lane_p8): Likewise.
2210 (vst3_lane_p16): Likewise.
2211 (vst3_lane_p64): Likewise.
2212 (vst3_lane_s8): Likewise.
2213 (vst3_lane_s16): Likewise.
2214 (vst3_lane_s32): Likewise.
2215 (vst3_lane_s64): Likewise.
2216 (vst3_lane_u8): Likewise.
2217 (vst3_lane_u16): Likewise.
2218 (vst3_lane_u32): Likewise.
2219 (vst3_lane_u64): Likewise.
2220 (vst3q_lane_f16): Likewise.
2221 (vst3q_lane_f32): Likewise.
2222 (vst3q_lane_f64): Likewise.
2223 (vst3q_lane_p8): Likewise.
2224 (vst3q_lane_p16): Likewise.
2225 (vst3q_lane_p64): Likewise.
2226 (vst3q_lane_s8): Likewise.
2227 (vst3q_lane_s16): Likewise.
2228 (vst3q_lane_s32): Likewise.
2229 (vst3q_lane_s64): Likewise.
2230 (vst3q_lane_u8): Likewise.
2231 (vst3q_lane_u16): Likewise.
2232 (vst3q_lane_u32): Likewise.
2233 (vst3q_lane_u64): Likewise.
2234 (vst4_lane_f16): Likewise.
2235 (vst4_lane_f32): Likewise.
2236 (vst4_lane_f64): Likewise.
2237 (vst4_lane_p8): Likewise.
2238 (vst4_lane_p16): Likewise.
2239 (vst4_lane_p64): Likewise.
2240 (vst4_lane_s8): Likewise.
2241 (vst4_lane_s16): Likewise.
2242 (vst4_lane_s32): Likewise.
2243 (vst4_lane_s64): Likewise.
2244 (vst4_lane_u8): Likewise.
2245 (vst4_lane_u16): Likewise.
2246 (vst4_lane_u32): Likewise.
2247 (vst4_lane_u64): Likewise.
2248 (vst4q_lane_f16): Likewise.
2249 (vst4q_lane_f32): Likewise.
2250 (vst4q_lane_f64): Likewise.
2251 (vst4q_lane_p8): Likewise.
2252 (vst4q_lane_p16): Likewise.
2253 (vst4q_lane_p64): Likewise.
2254 (vst4q_lane_s8): Likewise.
2255 (vst4q_lane_s16): Likewise.
2256 (vst4q_lane_s32): Likewise.
2257 (vst4q_lane_s64): Likewise.
2258 (vst4q_lane_u8): Likewise.
2259 (vst4q_lane_u16): Likewise.
2260 (vst4q_lane_u32): Likewise.
2261 (vst4q_lane_u64): Likewise.
2262 (vtbl3_s8): Likewise.
2263 (vtbl3_u8): Likewise.
2264 (vtbl3_p8): Likewise.
2265 (vtbl4_s8): Likewise.
2266 (vtbl4_u8): Likewise.
2267 (vtbl4_p8): Likewise.
2268 (vld1_u8_x3): Likewise.
2269 (vld1_s8_x3): Likewise.
2270 (vld1_u16_x3): Likewise.
2271 (vld1_s16_x3): Likewise.
2272 (vld1_u32_x3): Likewise.
2273 (vld1_s32_x3): Likewise.
2274 (vld1_u64_x3): Likewise.
2275 (vld1_s64_x3): Likewise.
2276 (vld1_f16_x3): Likewise.
2277 (vld1_f32_x3): Likewise.
2278 (vld1_f64_x3): Likewise.
2279 (vld1_p8_x3): Likewise.
2280 (vld1_p16_x3): Likewise.
2281 (vld1_p64_x3): Likewise.
2282 (vld1q_u8_x3): Likewise.
2283 (vld1q_s8_x3): Likewise.
2284 (vld1q_u16_x3): Likewise.
2285 (vld1q_s16_x3): Likewise.
2286 (vld1q_u32_x3): Likewise.
2287 (vld1q_s32_x3): Likewise.
2288 (vld1q_u64_x3): Likewise.
2289 (vld1q_s64_x3): Likewise.
2290 (vld1q_f16_x3): Likewise.
2291 (vld1q_f32_x3): Likewise.
2292 (vld1q_f64_x3): Likewise.
2293 (vld1q_p8_x3): Likewise.
2294 (vld1q_p16_x3): Likewise.
2295 (vld1q_p64_x3): Likewise.
2296 (vld1_u8_x2): Likewise.
2297 (vld1_s8_x2): Likewise.
2298 (vld1_u16_x2): Likewise.
2299 (vld1_s16_x2): Likewise.
2300 (vld1_u32_x2): Likewise.
2301 (vld1_s32_x2): Likewise.
2302 (vld1_u64_x2): Likewise.
2303 (vld1_s64_x2): Likewise.
2304 (vld1_f16_x2): Likewise.
2305 (vld1_f32_x2): Likewise.
2306 (vld1_f64_x2): Likewise.
2307 (vld1_p8_x2): Likewise.
2308 (vld1_p16_x2): Likewise.
2309 (vld1_p64_x2): Likewise.
2310 (vld1q_u8_x2): Likewise.
2311 (vld1q_s8_x2): Likewise.
2312 (vld1q_u16_x2): Likewise.
2313 (vld1q_s16_x2): Likewise.
2314 (vld1q_u32_x2): Likewise.
2315 (vld1q_s32_x2): Likewise.
2316 (vld1q_u64_x2): Likewise.
2317 (vld1q_s64_x2): Likewise.
2318 (vld1q_f16_x2): Likewise.
2319 (vld1q_f32_x2): Likewise.
2320 (vld1q_f64_x2): Likewise.
2321 (vld1q_p8_x2): Likewise.
2322 (vld1q_p16_x2): Likewise.
2323 (vld1q_p64_x2): Likewise.
2324 (vld1_s8_x4): Likewise.
2325 (vld1q_s8_x4): Likewise.
2326 (vld1_s16_x4): Likewise.
2327 (vld1q_s16_x4): Likewise.
2328 (vld1_s32_x4): Likewise.
2329 (vld1q_s32_x4): Likewise.
2330 (vld1_u8_x4): Likewise.
2331 (vld1q_u8_x4): Likewise.
2332 (vld1_u16_x4): Likewise.
2333 (vld1q_u16_x4): Likewise.
2334 (vld1_u32_x4): Likewise.
2335 (vld1q_u32_x4): Likewise.
2336 (vld1_f16_x4): Likewise.
2337 (vld1q_f16_x4): Likewise.
2338 (vld1_f32_x4): Likewise.
2339 (vld1q_f32_x4): Likewise.
2340 (vld1_p8_x4): Likewise.
2341 (vld1q_p8_x4): Likewise.
2342 (vld1_p16_x4): Likewise.
2343 (vld1q_p16_x4): Likewise.
2344 (vld1_s64_x4): Likewise.
2345 (vld1_u64_x4): Likewise.
2346 (vld1_p64_x4): Likewise.
2347 (vld1q_s64_x4): Likewise.
2348 (vld1q_u64_x4): Likewise.
2349 (vld1q_p64_x4): Likewise.
2350 (vld1_f64_x4): Likewise.
2351 (vld1q_f64_x4): Likewise.
2352 (vld2_s64): Likewise.
2353 (vld2_u64): Likewise.
2354 (vld2_f64): Likewise.
2355 (vld2_s8): Likewise.
2356 (vld2_p8): Likewise.
2357 (vld2_p64): Likewise.
2358 (vld2_s16): Likewise.
2359 (vld2_p16): Likewise.
2360 (vld2_s32): Likewise.
2361 (vld2_u8): Likewise.
2362 (vld2_u16): Likewise.
2363 (vld2_u32): Likewise.
2364 (vld2_f16): Likewise.
2365 (vld2_f32): Likewise.
2366 (vld2q_s8): Likewise.
2367 (vld2q_p8): Likewise.
2368 (vld2q_s16): Likewise.
2369 (vld2q_p16): Likewise.
2370 (vld2q_p64): Likewise.
2371 (vld2q_s32): Likewise.
2372 (vld2q_s64): Likewise.
2373 (vld2q_u8): Likewise.
2374 (vld2q_u16): Likewise.
2375 (vld2q_u32): Likewise.
2376 (vld2q_u64): Likewise.
2377 (vld2q_f16): Likewise.
2378 (vld2q_f32): Likewise.
2379 (vld2q_f64): Likewise.
2380 (vld3_s64): Likewise.
2381 (vld3_u64): Likewise.
2382 (vld3_f64): Likewise.
2383 (vld3_s8): Likewise.
2384 (vld3_p8): Likewise.
2385 (vld3_s16): Likewise.
2386 (vld3_p16): Likewise.
2387 (vld3_s32): Likewise.
2388 (vld3_u8): Likewise.
2389 (vld3_u16): Likewise.
2390 (vld3_u32): Likewise.
2391 (vld3_f16): Likewise.
2392 (vld3_f32): Likewise.
2393 (vld3_p64): Likewise.
2394 (vld3q_s8): Likewise.
2395 (vld3q_p8): Likewise.
2396 (vld3q_s16): Likewise.
2397 (vld3q_p16): Likewise.
2398 (vld3q_s32): Likewise.
2399 (vld3q_s64): Likewise.
2400 (vld3q_u8): Likewise.
2401 (vld3q_u16): Likewise.
2402 (vld3q_u32): Likewise.
2403 (vld3q_u64): Likewise.
2404 (vld3q_f16): Likewise.
2405 (vld3q_f32): Likewise.
2406 (vld3q_f64): Likewise.
2407 (vld3q_p64): Likewise.
2408 (vld4_s64): Likewise.
2409 (vld4_u64): Likewise.
2410 (vld4_f64): Likewise.
2411 (vld4_s8): Likewise.
2412 (vld4_p8): Likewise.
2413 (vld4_s16): Likewise.
2414 (vld4_p16): Likewise.
2415 (vld4_s32): Likewise.
2416 (vld4_u8): Likewise.
2417 (vld4_u16): Likewise.
2418 (vld4_u32): Likewise.
2419 (vld4_f16): Likewise.
2420 (vld4_f32): Likewise.
2421 (vld4_p64): Likewise.
2422 (vld4q_s8): Likewise.
2423 (vld4q_p8): Likewise.
2424 (vld4q_s16): Likewise.
2425 (vld4q_p16): Likewise.
2426 (vld4q_s32): Likewise.
2427 (vld4q_s64): Likewise.
2428 (vld4q_u8): Likewise.
2429 (vld4q_u16): Likewise.
2430 (vld4q_u32): Likewise.
2431 (vld4q_u64): Likewise.
2432 (vld4q_f16): Likewise.
2433 (vld4q_f32): Likewise.
2434 (vld4q_f64): Likewise.
2435 (vld4q_p64): Likewise.
2436 (vld2_dup_s8): Likewise.
2437 (vld2_dup_s16): Likewise.
2438 (vld2_dup_s32): Likewise.
2439 (vld2_dup_f16): Likewise.
2440 (vld2_dup_f32): Likewise.
2441 (vld2_dup_f64): Likewise.
2442 (vld2_dup_u8): Likewise.
2443 (vld2_dup_u16): Likewise.
2444 (vld2_dup_u32): Likewise.
2445 (vld2_dup_p8): Likewise.
2446 (vld2_dup_p16): Likewise.
2447 (vld2_dup_p64): Likewise.
2448 (vld2_dup_s64): Likewise.
2449 (vld2_dup_u64): Likewise.
2450 (vld2q_dup_s8): Likewise.
2451 (vld2q_dup_p8): Likewise.
2452 (vld2q_dup_s16): Likewise.
2453 (vld2q_dup_p16): Likewise.
2454 (vld2q_dup_s32): Likewise.
2455 (vld2q_dup_s64): Likewise.
2456 (vld2q_dup_u8): Likewise.
2457 (vld2q_dup_u16): Likewise.
2458 (vld2q_dup_u32): Likewise.
2459 (vld2q_dup_u64): Likewise.
2460 (vld2q_dup_f16): Likewise.
2461 (vld2q_dup_f32): Likewise.
2462 (vld2q_dup_f64): Likewise.
2463 (vld2q_dup_p64): Likewise.
2464 (vld3_dup_s64): Likewise.
2465 (vld3_dup_u64): Likewise.
2466 (vld3_dup_f64): Likewise.
2467 (vld3_dup_s8): Likewise.
2468 (vld3_dup_p8): Likewise.
2469 (vld3_dup_s16): Likewise.
2470 (vld3_dup_p16): Likewise.
2471 (vld3_dup_s32): Likewise.
2472 (vld3_dup_u8): Likewise.
2473 (vld3_dup_u16): Likewise.
2474 (vld3_dup_u32): Likewise.
2475 (vld3_dup_f16): Likewise.
2476 (vld3_dup_f32): Likewise.
2477 (vld3_dup_p64): Likewise.
2478 (vld3q_dup_s8): Likewise.
2479 (vld3q_dup_p8): Likewise.
2480 (vld3q_dup_s16): Likewise.
2481 (vld3q_dup_p16): Likewise.
2482 (vld3q_dup_s32): Likewise.
2483 (vld3q_dup_s64): Likewise.
2484 (vld3q_dup_u8): Likewise.
2485 (vld3q_dup_u16): Likewise.
2486 (vld3q_dup_u32): Likewise.
2487 (vld3q_dup_u64): Likewise.
2488 (vld3q_dup_f16): Likewise.
2489 (vld3q_dup_f32): Likewise.
2490 (vld3q_dup_f64): Likewise.
2491 (vld3q_dup_p64): Likewise.
2492 (vld4_dup_s64): Likewise.
2493 (vld4_dup_u64): Likewise.
2494 (vld4_dup_f64): Likewise.
2495 (vld4_dup_s8): Likewise.
2496 (vld4_dup_p8): Likewise.
2497 (vld4_dup_s16): Likewise.
2498 (vld4_dup_p16): Likewise.
2499 (vld4_dup_s32): Likewise.
2500 (vld4_dup_u8): Likewise.
2501 (vld4_dup_u16): Likewise.
2502 (vld4_dup_u32): Likewise.
2503 (vld4_dup_f16): Likewise.
2504 (vld4_dup_f32): Likewise.
2505 (vld4_dup_p64): Likewise.
2506 (vld4q_dup_s8): Likewise.
2507 (vld4q_dup_p8): Likewise.
2508 (vld4q_dup_s16): Likewise.
2509 (vld4q_dup_p16): Likewise.
2510 (vld4q_dup_s32): Likewise.
2511 (vld4q_dup_s64): Likewise.
2512 (vld4q_dup_u8): Likewise.
2513 (vld4q_dup_u16): Likewise.
2514 (vld4q_dup_u32): Likewise.
2515 (vld4q_dup_u64): Likewise.
2516 (vld4q_dup_f16): Likewise.
2517 (vld4q_dup_f32): Likewise.
2518 (vld4q_dup_f64): Likewise.
2519 (vld4q_dup_p64): Likewise.
2520 (vld2_lane_u8): Likewise.
2521 (vld2_lane_u16): Likewise.
2522 (vld2_lane_u32): Likewise.
2523 (vld2_lane_u64): Likewise.
2524 (vld2_lane_s8): Likewise.
2525 (vld2_lane_s16): Likewise.
2526 (vld2_lane_s32): Likewise.
2527 (vld2_lane_s64): Likewise.
2528 (vld2_lane_f16): Likewise.
2529 (vld2_lane_f32): Likewise.
2530 (vld2_lane_f64): Likewise.
2531 (vld2_lane_p8): Likewise.
2532 (vld2_lane_p16): Likewise.
2533 (vld2_lane_p64): Likewise.
2534 (vld2q_lane_u8): Likewise.
2535 (vld2q_lane_u16): Likewise.
2536 (vld2q_lane_u32): Likewise.
2537 (vld2q_lane_u64): Likewise.
2538 (vld2q_lane_s8): Likewise.
2539 (vld2q_lane_s16): Likewise.
2540 (vld2q_lane_s32): Likewise.
2541 (vld2q_lane_s64): Likewise.
2542 (vld2q_lane_f16): Likewise.
2543 (vld2q_lane_f32): Likewise.
2544 (vld2q_lane_f64): Likewise.
2545 (vld2q_lane_p8): Likewise.
2546 (vld2q_lane_p16): Likewise.
2547 (vld2q_lane_p64): Likewise.
2548 (vld3_lane_u8): Likewise.
2549 (vld3_lane_u16): Likewise.
2550 (vld3_lane_u32): Likewise.
2551 (vld3_lane_u64): Likewise.
2552 (vld3_lane_s8): Likewise.
2553 (vld3_lane_s16): Likewise.
2554 (vld3_lane_s32): Likewise.
2555 (vld3_lane_s64): Likewise.
2556 (vld3_lane_f16): Likewise.
2557 (vld3_lane_f32): Likewise.
2558 (vld3_lane_f64): Likewise.
2559 (vld3_lane_p8): Likewise.
2560 (vld3_lane_p16): Likewise.
2561 (vld3_lane_p64): Likewise.
2562 (vld3q_lane_u8): Likewise.
2563 (vld3q_lane_u16): Likewise.
2564 (vld3q_lane_u32): Likewise.
2565 (vld3q_lane_u64): Likewise.
2566 (vld3q_lane_s8): Likewise.
2567 (vld3q_lane_s16): Likewise.
2568 (vld3q_lane_s32): Likewise.
2569 (vld3q_lane_s64): Likewise.
2570 (vld3q_lane_f16): Likewise.
2571 (vld3q_lane_f32): Likewise.
2572 (vld3q_lane_f64): Likewise.
2573 (vld3q_lane_p8): Likewise.
2574 (vld3q_lane_p16): Likewise.
2575 (vld3q_lane_p64): Likewise.
2576 (vld4_lane_u8): Likewise.
2577 (vld4_lane_u16): Likewise.
2578 (vld4_lane_u32): Likewise.
2579 (vld4_lane_u64): Likewise.
2580 (vld4_lane_s8): Likewise.
2581 (vld4_lane_s16): Likewise.
2582 (vld4_lane_s32): Likewise.
2583 (vld4_lane_s64): Likewise.
2584 (vld4_lane_f16): Likewise.
2585 (vld4_lane_f32): Likewise.
2586 (vld4_lane_f64): Likewise.
2587 (vld4_lane_p8): Likewise.
2588 (vld4_lane_p16): Likewise.
2589 (vld4_lane_p64): Likewise.
2590 (vld4q_lane_u8): Likewise.
2591 (vld4q_lane_u16): Likewise.
2592 (vld4q_lane_u32): Likewise.
2593 (vld4q_lane_u64): Likewise.
2594 (vld4q_lane_s8): Likewise.
2595 (vld4q_lane_s16): Likewise.
2596 (vld4q_lane_s32): Likewise.
2597 (vld4q_lane_s64): Likewise.
2598 (vld4q_lane_f16): Likewise.
2599 (vld4q_lane_f32): Likewise.
2600 (vld4q_lane_f64): Likewise.
2601 (vld4q_lane_p8): Likewise.
2602 (vld4q_lane_p16): Likewise.
2603 (vld4q_lane_p64): Likewise.
2604 (vqtbl2_s8): Likewise.
2605 (vqtbl2_u8): Likewise.
2606 (vqtbl2_p8): Likewise.
2607 (vqtbl2q_s8): Likewise.
2608 (vqtbl2q_u8): Likewise.
2609 (vqtbl2q_p8): Likewise.
2610 (vqtbl3_s8): Likewise.
2611 (vqtbl3_u8): Likewise.
2612 (vqtbl3_p8): Likewise.
2613 (vqtbl3q_s8): Likewise.
2614 (vqtbl3q_u8): Likewise.
2615 (vqtbl3q_p8): Likewise.
2616 (vqtbl4_s8): Likewise.
2617 (vqtbl4_u8): Likewise.
2618 (vqtbl4_p8): Likewise.
2619 (vqtbl4q_s8): Likewise.
2620 (vqtbl4q_u8): Likewise.
2621 (vqtbl4q_p8): Likewise.
2622 (vqtbx2_s8): Likewise.
2623 (vqtbx2_u8): Likewise.
2624 (vqtbx2_p8): Likewise.
2625 (vqtbx2q_s8): Likewise.
2626 (vqtbx2q_u8): Likewise.
2627 (vqtbx2q_p8): Likewise.
2628 (vqtbx3_s8): Likewise.
2629 (vqtbx3_u8): Likewise.
2630 (vqtbx3_p8): Likewise.
2631 (vqtbx3q_s8): Likewise.
2632 (vqtbx3q_u8): Likewise.
2633 (vqtbx3q_p8): Likewise.
2634 (vqtbx4_s8): Likewise.
2635 (vqtbx4_u8): Likewise.
2636 (vqtbx4_p8): Likewise.
2637 (vqtbx4q_s8): Likewise.
2638 (vqtbx4q_u8): Likewise.
2639 (vqtbx4q_p8): Likewise.
2640 (vst1_s64_x2): Likewise.
2641 (vst1_u64_x2): Likewise.
2642 (vst1_f64_x2): Likewise.
2643 (vst1_s8_x2): Likewise.
2644 (vst1_p8_x2): Likewise.
2645 (vst1_s16_x2): Likewise.
2646 (vst1_p16_x2): Likewise.
2647 (vst1_s32_x2): Likewise.
2648 (vst1_u8_x2): Likewise.
2649 (vst1_u16_x2): Likewise.
2650 (vst1_u32_x2): Likewise.
2651 (vst1_f16_x2): Likewise.
2652 (vst1_f32_x2): Likewise.
2653 (vst1_p64_x2): Likewise.
2654 (vst1q_s8_x2): Likewise.
2655 (vst1q_p8_x2): Likewise.
2656 (vst1q_s16_x2): Likewise.
2657 (vst1q_p16_x2): Likewise.
2658 (vst1q_s32_x2): Likewise.
2659 (vst1q_s64_x2): Likewise.
2660 (vst1q_u8_x2): Likewise.
2661 (vst1q_u16_x2): Likewise.
2662 (vst1q_u32_x2): Likewise.
2663 (vst1q_u64_x2): Likewise.
2664 (vst1q_f16_x2): Likewise.
2665 (vst1q_f32_x2): Likewise.
2666 (vst1q_f64_x2): Likewise.
2667 (vst1q_p64_x2): Likewise.
2668 (vst1_s64_x3): Likewise.
2669 (vst1_u64_x3): Likewise.
2670 (vst1_f64_x3): Likewise.
2671 (vst1_s8_x3): Likewise.
2672 (vst1_p8_x3): Likewise.
2673 (vst1_s16_x3): Likewise.
2674 (vst1_p16_x3): Likewise.
2675 (vst1_s32_x3): Likewise.
2676 (vst1_u8_x3): Likewise.
2677 (vst1_u16_x3): Likewise.
2678 (vst1_u32_x3): Likewise.
2679 (vst1_f16_x3): Likewise.
2680 (vst1_f32_x3): Likewise.
2681 (vst1_p64_x3): Likewise.
2682 (vst1q_s8_x3): Likewise.
2683 (vst1q_p8_x3): Likewise.
2684 (vst1q_s16_x3): Likewise.
2685 (vst1q_p16_x3): Likewise.
2686 (vst1q_s32_x3): Likewise.
2687 (vst1q_s64_x3): Likewise.
2688 (vst1q_u8_x3): Likewise.
2689 (vst1q_u16_x3): Likewise.
2690 (vst1q_u32_x3): Likewise.
2691 (vst1q_u64_x3): Likewise.
2692 (vst1q_f16_x3): Likewise.
2693 (vst1q_f32_x3): Likewise.
2694 (vst1q_f64_x3): Likewise.
2695 (vst1q_p64_x3): Likewise.
2696 (vst1_s8_x4): Likewise.
2697 (vst1q_s8_x4): Likewise.
2698 (vst1_s16_x4): Likewise.
2699 (vst1q_s16_x4): Likewise.
2700 (vst1_s32_x4): Likewise.
2701 (vst1q_s32_x4): Likewise.
2702 (vst1_u8_x4): Likewise.
2703 (vst1q_u8_x4): Likewise.
2704 (vst1_u16_x4): Likewise.
2705 (vst1q_u16_x4): Likewise.
2706 (vst1_u32_x4): Likewise.
2707 (vst1q_u32_x4): Likewise.
2708 (vst1_f16_x4): Likewise.
2709 (vst1q_f16_x4): Likewise.
2710 (vst1_f32_x4): Likewise.
2711 (vst1q_f32_x4): Likewise.
2712 (vst1_p8_x4): Likewise.
2713 (vst1q_p8_x4): Likewise.
2714 (vst1_p16_x4): Likewise.
2715 (vst1q_p16_x4): Likewise.
2716 (vst1_s64_x4): Likewise.
2717 (vst1_u64_x4): Likewise.
2718 (vst1_p64_x4): Likewise.
2719 (vst1q_s64_x4): Likewise.
2720 (vst1q_u64_x4): Likewise.
2721 (vst1q_p64_x4): Likewise.
2722 (vst1_f64_x4): Likewise.
2723 (vst1q_f64_x4): Likewise.
2724 (vst2_s64): Likewise.
2725 (vst2_u64): Likewise.
2726 (vst2_f64): Likewise.
2727 (vst2_s8): Likewise.
2728 (vst2_p8): Likewise.
2729 (vst2_s16): Likewise.
2730 (vst2_p16): Likewise.
2731 (vst2_s32): Likewise.
2732 (vst2_u8): Likewise.
2733 (vst2_u16): Likewise.
2734 (vst2_u32): Likewise.
2735 (vst2_f16): Likewise.
2736 (vst2_f32): Likewise.
2737 (vst2_p64): Likewise.
2738 (vst2q_s8): Likewise.
2739 (vst2q_p8): Likewise.
2740 (vst2q_s16): Likewise.
2741 (vst2q_p16): Likewise.
2742 (vst2q_s32): Likewise.
2743 (vst2q_s64): Likewise.
2744 (vst2q_u8): Likewise.
2745 (vst2q_u16): Likewise.
2746 (vst2q_u32): Likewise.
2747 (vst2q_u64): Likewise.
2748 (vst2q_f16): Likewise.
2749 (vst2q_f32): Likewise.
2750 (vst2q_f64): Likewise.
2751 (vst2q_p64): Likewise.
2752 (vst3_s64): Likewise.
2753 (vst3_u64): Likewise.
2754 (vst3_f64): Likewise.
2755 (vst3_s8): Likewise.
2756 (vst3_p8): Likewise.
2757 (vst3_s16): Likewise.
2758 (vst3_p16): Likewise.
2759 (vst3_s32): Likewise.
2760 (vst3_u8): Likewise.
2761 (vst3_u16): Likewise.
2762 (vst3_u32): Likewise.
2763 (vst3_f16): Likewise.
2764 (vst3_f32): Likewise.
2765 (vst3_p64): Likewise.
2766 (vst3q_s8): Likewise.
2767 (vst3q_p8): Likewise.
2768 (vst3q_s16): Likewise.
2769 (vst3q_p16): Likewise.
2770 (vst3q_s32): Likewise.
2771 (vst3q_s64): Likewise.
2772 (vst3q_u8): Likewise.
2773 (vst3q_u16): Likewise.
2774 (vst3q_u32): Likewise.
2775 (vst3q_u64): Likewise.
2776 (vst3q_f16): Likewise.
2777 (vst3q_f32): Likewise.
2778 (vst3q_f64): Likewise.
2779 (vst3q_p64): Likewise.
2780 (vst4_s64): Likewise.
2781 (vst4_u64): Likewise.
2782 (vst4_f64): Likewise.
2783 (vst4_s8): Likewise.
2784 (vst4_p8): Likewise.
2785 (vst4_s16): Likewise.
2786 (vst4_p16): Likewise.
2787 (vst4_s32): Likewise.
2788 (vst4_u8): Likewise.
2789 (vst4_u16): Likewise.
2790 (vst4_u32): Likewise.
2791 (vst4_f16): Likewise.
2792 (vst4_f32): Likewise.
2793 (vst4_p64): Likewise.
2794 (vst4q_s8): Likewise.
2795 (vst4q_p8): Likewise.
2796 (vst4q_s16): Likewise.
2797 (vst4q_p16): Likewise.
2798 (vst4q_s32): Likewise.
2799 (vst4q_s64): Likewise.
2800 (vst4q_u8): Likewise.
2801 (vst4q_u16): Likewise.
2802 (vst4q_u32): Likewise.
2803 (vst4q_u64): Likewise.
2804 (vst4q_f16): Likewise.
2805 (vst4q_f32): Likewise.
2806 (vst4q_f64): Likewise.
2807 (vst4q_p64): Likewise.
2808 (vtbx4_s8): Likewise.
2809 (vtbx4_u8): Likewise.
2810 (vtbx4_p8): Likewise.
2811 (vld1_bf16_x2): Likewise.
2812 (vld1q_bf16_x2): Likewise.
2813 (vld1_bf16_x3): Likewise.
2814 (vld1q_bf16_x3): Likewise.
2815 (vld1_bf16_x4): Likewise.
2816 (vld1q_bf16_x4): Likewise.
2817 (vld2_bf16): Likewise.
2818 (vld2q_bf16): Likewise.
2819 (vld2_dup_bf16): Likewise.
2820 (vld2q_dup_bf16): Likewise.
2821 (vld3_bf16): Likewise.
2822 (vld3q_bf16): Likewise.
2823 (vld3_dup_bf16): Likewise.
2824 (vld3q_dup_bf16): Likewise.
2825 (vld4_bf16): Likewise.
2826 (vld4q_bf16): Likewise.
2827 (vld4_dup_bf16): Likewise.
2828 (vld4q_dup_bf16): Likewise.
2829 (vst1_bf16_x2): Likewise.
2830 (vst1q_bf16_x2): Likewise.
2831 (vst1_bf16_x3): Likewise.
2832 (vst1q_bf16_x3): Likewise.
2833 (vst1_bf16_x4): Likewise.
2834 (vst1q_bf16_x4): Likewise.
2835 (vst2_bf16): Likewise.
2836 (vst2q_bf16): Likewise.
2837 (vst3_bf16): Likewise.
2838 (vst3q_bf16): Likewise.
2839 (vst4_bf16): Likewise.
2840 (vst4q_bf16): Likewise.
2841 (vld2_lane_bf16): Likewise.
2842 (vld2q_lane_bf16): Likewise.
2843 (vld3_lane_bf16): Likewise.
2844 (vld3q_lane_bf16): Likewise.
2845 (vld4_lane_bf16): Likewise.
2846 (vld4q_lane_bf16): Likewise.
2847 (vst2_lane_bf16): Likewise.
2848 (vst2q_lane_bf16): Likewise.
2849 (vst3_lane_bf16): Likewise.
2850 (vst3q_lane_bf16): Likewise.
2851 (vst4_lane_bf16): Likewise.
2852 (vst4q_lane_bf16): Likewise.
2853 * config/aarch64/geniterators.sh: Modify iterator regex to
2854 match new vector-tuple modes.
2855 * config/aarch64/iterators.md (insn_count): Extend mode
2856 attribute with vector-tuple type information.
2858 (Vendreg): Likewise.
2861 (VSTRUCT_2D): New mode iterator.
2862 (VSTRUCT_2DNX): Likewise.
2863 (VSTRUCT_2DX): Likewise.
2864 (VSTRUCT_2Q): Likewise.
2865 (VSTRUCT_2QD): Likewise.
2866 (VSTRUCT_3D): Likewise.
2867 (VSTRUCT_3DNX): Likewise.
2868 (VSTRUCT_3DX): Likewise.
2869 (VSTRUCT_3Q): Likewise.
2870 (VSTRUCT_3QD): Likewise.
2871 (VSTRUCT_4D): Likewise.
2872 (VSTRUCT_4DNX): Likewise.
2873 (VSTRUCT_4DX): Likewise.
2874 (VSTRUCT_4Q): Likewise.
2875 (VSTRUCT_4QD): Likewise.
2876 (VSTRUCT_D): Likewise.
2877 (VSTRUCT_Q): Likewise.
2878 (VSTRUCT_QD): Likewise.
2879 (VSTRUCT_ELT): New mode attribute.
2880 (vstruct_elt): Likewise.
2881 * genmodes.c (VECTOR_MODE): Add default prefix and order
2883 (VECTOR_MODE_WITH_PREFIX): Define.
2884 (make_vector_mode): Add mode prefix and order parameters.
2886 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
2888 * expmed.c (extract_bit_field_1): Ensure modes are tieable.
2890 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
2892 * expr.c (emit_group_load_1): Remove historic workaround.
2894 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
2896 * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
2897 Factor out main loop to...
2898 (aarch64_init_simd_builtin_functions): This new function.
2899 (register_tuple_type): Define.
2900 (aarch64_scalar_builtin_type_p): Define.
2901 (handle_arm_neon_h): Define.
2902 * config/aarch64/aarch64-c.c (aarch64_pragma_aarch64): Handle
2903 pragma for arm_neon.h.
2904 * config/aarch64/aarch64-protos.h (aarch64_advsimd_struct_mode_p):
2906 (handle_arm_neon_h): Likewise.
2907 * config/aarch64/aarch64.c (aarch64_advsimd_struct_mode_p):
2908 Remove static modifier.
2909 * config/aarch64/arm_neon.h (target): Remove Neon vector
2910 structure type definitions.
2912 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
2914 PR tree-optimization/102943
2915 * gimple-range-path.cc (path_range_query::range_on_path_entry):
2916 Prefer range_of_expr unless there are no statements in the BB.
2918 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
2920 PR tree-optimization/102943
2921 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
2922 Avoid duplicate calculation of paths.
2924 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
2926 PR tree-optimization/102943
2927 * gimple-range-path.cc (path_range_query::compute_phi_relations):
2928 Only compute relations for SSA names in the import list.
2929 (path_range_query::compute_outgoing_relations): Same.
2930 * gimple-range-path.h (path_range_query::import_p): New.
2932 2021-11-04 Richard Biener <rguenther@suse.de>
2934 PR rtl-optimization/103075
2935 * simplify-rtx.c (exact_int_to_float_conversion_p): Return
2936 false for a VOIDmode operand.
2938 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
2940 * config/aarch64/aarch64.c (aarch64_vector_costs): Make member
2941 variables private and add "m_" to their names. Remove is_loop.
2942 (aarch64_record_potential_advsimd_unrolling): Replace with...
2943 (aarch64_vector_costs::record_potential_advsimd_unrolling): ...this.
2944 (aarch64_analyze_loop_vinfo): Replace with...
2945 (aarch64_vector_costs::analyze_loop_vinfo): ...this.
2946 Move initialization of (m_)vec_flags to add_stmt_cost.
2947 (aarch64_analyze_bb_vinfo): Delete.
2948 (aarch64_count_ops): Replace with...
2949 (aarch64_vector_costs::count_ops): ...this.
2950 (aarch64_vector_costs::add_stmt_cost): Set m_vec_flags,
2951 using m_costing_for_scalar to test whether we're costing
2952 scalar or vector code.
2953 (aarch64_adjust_body_cost_sve): Replace with...
2954 (aarch64_vector_costs::adjust_body_cost_sve): ...this.
2955 (aarch64_adjust_body_cost): Replace with...
2956 (aarch64_vector_costs::adjust_body_cost): ...this.
2957 (aarch64_vector_costs::finish_cost): Use m_vinfo instead of is_loop.
2959 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
2961 * target.def (targetm.vectorize.init_cost): Replace with...
2962 (targetm.vectorize.create_costs): ...this.
2963 (targetm.vectorize.add_stmt_cost): Delete.
2964 (targetm.vectorize.finish_cost): Likewise.
2965 (targetm.vectorize.destroy_cost_data): Likewise.
2966 * doc/tm.texi.in (TARGET_VECTORIZE_INIT_COST): Replace with...
2967 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
2968 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
2969 (TARGET_VECTORIZE_FINISH_COST): Likewise.
2970 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
2971 * doc/tm.texi: Regenerate.
2972 * tree-vectorizer.h (vec_info::vec_info): Remove target_cost_data
2974 (vec_info::target_cost_data): Change from a void * to a vector_costs *.
2975 (vector_costs): New class.
2976 (init_cost): Take a vec_info and return a vector_costs.
2977 (dump_stmt_cost): Remove data parameter.
2978 (add_stmt_cost): Replace vinfo and data parameters with a vector_costs.
2979 (add_stmt_costs): Likewise.
2980 (finish_cost): Replace data parameter with a vector_costs.
2981 (destroy_cost_data): Delete.
2982 * tree-vectorizer.c (dump_stmt_cost): Remove data argument and
2984 (vec_info::vec_info): Remove the target_cost_data parameter and
2985 initialize the member variable to null instead.
2986 (vec_info::~vec_info): Delete target_cost_data instead of calling
2988 (vector_costs::add_stmt_cost): New function.
2989 (vector_costs::finish_cost): Likewise.
2990 (vector_costs::record_stmt_cost): Likewise.
2991 (vector_costs::adjust_cost_for_freq): Likewise.
2992 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
2993 call to vec_info::vec_info.
2994 (vect_compute_single_scalar_iteration_cost): Update after above
2995 changes to costing interface.
2996 (vect_analyze_loop_operations): Likewise.
2997 (vect_estimate_min_profitable_iters): Likewise.
2998 (vect_analyze_loop_2): Initialize LOOP_VINFO_TARGET_COST_DATA
2999 at the start_over point, where it needs to be recreated after
3000 trying without slp. Update retry code accordingly.
3001 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Update call
3002 to vec_info::vec_info.
3003 (vect_slp_analyze_operation): Update after above changes to costing
3005 (vect_bb_vectorization_profitable_p): Likewise.
3006 * targhooks.h (default_init_cost): Replace with...
3007 (default_vectorize_create_costs): ...this.
3008 (default_add_stmt_cost): Delete.
3009 (default_finish_cost, default_destroy_cost_data): Likewise.
3010 * targhooks.c (default_init_cost): Replace with...
3011 (default_vectorize_create_costs): ...this.
3012 (default_add_stmt_cost): Delete, moving logic to vector_costs instead.
3013 (default_finish_cost, default_destroy_cost_data): Delete.
3014 * config/aarch64/aarch64.c (aarch64_vector_costs): Inherit from
3015 vector_costs. Add a constructor.
3016 (aarch64_init_cost): Replace with...
3017 (aarch64_vectorize_create_costs): ...this.
3018 (aarch64_add_stmt_cost): Replace with...
3019 (aarch64_vector_costs::add_stmt_cost): ...this. Use record_stmt_cost
3020 to adjust the cost for inner loops.
3021 (aarch64_finish_cost): Replace with...
3022 (aarch64_vector_costs::finish_cost): ...this.
3023 (aarch64_destroy_cost_data): Delete.
3024 (TARGET_VECTORIZE_INIT_COST): Replace with...
3025 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
3026 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
3027 (TARGET_VECTORIZE_FINISH_COST): Likewise.
3028 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
3029 * config/i386/i386.c (ix86_vector_costs): New structure.
3030 (ix86_init_cost): Replace with...
3031 (ix86_vectorize_create_costs): ...this.
3032 (ix86_add_stmt_cost): Replace with...
3033 (ix86_vector_costs::add_stmt_cost): ...this. Use adjust_cost_for_freq
3034 to adjust the cost for inner loops.
3035 (ix86_finish_cost, ix86_destroy_cost_data): Delete.
3036 (TARGET_VECTORIZE_INIT_COST): Replace with...
3037 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
3038 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
3039 (TARGET_VECTORIZE_FINISH_COST): Likewise.
3040 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
3041 * config/rs6000/rs6000.c (TARGET_VECTORIZE_INIT_COST): Replace with...
3042 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
3043 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
3044 (TARGET_VECTORIZE_FINISH_COST): Likewise.
3045 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
3046 (rs6000_cost_data): Inherit from vector_costs.
3047 Add a constructor. Drop loop_info, cost and costing_for_scalar
3048 in favor of the corresponding vector_costs member variables.
3049 Add "m_" to the names of the remaining member variables and
3051 (rs6000_density_test): Replace with...
3052 (rs6000_cost_data::density_test): ...this.
3053 (rs6000_init_cost): Replace with...
3054 (rs6000_vectorize_create_costs): ...this.
3055 (rs6000_update_target_cost_per_stmt): Replace with...
3056 (rs6000_cost_data::update_target_cost_per_stmt): ...this.
3057 (rs6000_add_stmt_cost): Replace with...
3058 (rs6000_cost_data::add_stmt_cost): ...this. Use adjust_cost_for_freq
3059 to adjust the cost for inner loops.
3060 (rs6000_adjust_vect_cost_per_loop): Replace with...
3061 (rs6000_cost_data::adjust_vect_cost_per_loop): ...this.
3062 (rs6000_finish_cost): Replace with...
3063 (rs6000_cost_data::finish_cost): ...this. Group loop code
3064 into a single if statement and pass the loop_vinfo down to
3066 (rs6000_destroy_cost_data): Delete.
3068 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
3070 PR tree-optimization/103062
3071 PR tree-optimization/103062
3072 * value-pointer-equiv.cc (ssa_equiv_stack::ssa_equiv_stack):
3073 Increase size of allocation by 1.
3074 (ssa_equiv_stack::push_replacement): Grow as needed.
3075 (ssa_equiv_stack::get_replacement): Same.
3076 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
3077 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Remove delete.
3078 (pointer_equiv_analyzer::set_global_equiv): Grow as needed.
3079 (pointer_equiv_analyzer::get_equiv): Same.
3080 (pointer_equiv_analyzer::get_equiv_expr): Remove const.
3081 * value-pointer-equiv.h (class pointer_equiv_analyzer): Remove
3082 const markers. Use auto_vec instead of tree *.
3084 2021-11-04 Richard Biener <rguenther@suse.de>
3086 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Remove always
3087 true parameter and inline valueization.
3088 (vn_nary_op_lookup_1): Inline valueization from ...
3089 (vn_nary_op_compute_hash): ... here and remove it here.
3090 * tree-ssa-pre.c (phi_translate_1): Do not valueize
3091 before vn_nary_lookup_pieces.
3092 (get_representative_for): Mark created SSA representatives
3095 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
3097 * simplify-rtx.c (simplify_context::simplify_gen_vec_select): Assert
3098 that the operand has a vector mode. Use subreg_lowpart_offset
3099 to test whether an index corresponds to the low part.
3101 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
3103 * read-rtl.c: Remove dead !GENERATOR_FILE block.
3104 * read-rtl-function.c (function_reader::consolidate_singletons):
3105 Generate canonical CONST_VECTORs.
3107 2021-11-04 liuhongt <hongtao.liu@intel.com>
3110 * config/i386/predicates.md (reg_or_notreg_operand): Rename to ..
3111 (regmem_or_bitnot_regmem_operand): .. and extend to handle
3113 * config/i386/sse.md (*<avx512>_vpternlog<mode>_1): Force_reg
3114 the operands which are required to be register_operand.
3115 (*<avx512>_vpternlog<mode>_2): Ditto.
3116 (*<avx512>_vpternlog<mode>_3): Ditto.
3117 (*<avx512>_vternlog<mode>_all): Disallow embeded broadcast for
3118 vector HFmodes since it's not a real AVX512FP16 instruction.
3120 2021-11-04 liuhongt <hongtao.liu@intel.com>
3123 * match.pd: simplify (trunc)copysign((extend)a, (extend)b) to
3124 .COPYSIGN (a,b) when a and b are same type as the truncation
3125 type and has less precision than extend type.
3127 2021-11-04 Richard Biener <rguenther@suse.de>
3129 * doc/generic.texi: Update TARGET_MEM_REF and MEM_REF
3132 2021-11-04 Hongyu Wang <hongyu.wang@intel.com>
3134 * config/i386/sse.md (VI2_AVX512VNNIBW): New mode iterator.
3135 (VI1_AVX512VNNI): Likewise.
3136 (SDOT_VPDP_SUF): New mode_attr.
3139 (sdot_prod<mode>): Use VI2_AVX512F iterator, expand to
3140 vpdpwssd when VNNI targets available.
3141 (usdot_prod<mode>): New expander for vector QImode.
3143 2021-11-04 Hongyu Wang <hongyu.wang@intel.com>
3145 * config/i386/amxtileintrin.h (_tile_loadd_internal): Add
3146 parentheses to base and stride.
3147 (_tile_stream_loadd_internal): Likewise.
3148 (_tile_stored_internal): Likewise.
3150 2021-11-03 Maciej W. Rozycki <macro@embecosm.com>
3152 * config/riscv/riscv.c (riscv_class_max_nregs): Swap the
3153 arguments to `reg_class_subset_p'.
3155 2021-11-03 Joseph Myers <joseph@codesourcery.com>
3158 * fold-const.c (fold_init): New function.
3159 * fold-const.h (fold_init): New prototype.
3161 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
3163 * params.opt (param_vrp2_mode): Make ranger the default for VRP2.
3165 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
3167 * gimple-range-fold.cc (fold_using_range::range_of_phi): Don't import
3168 a range from edge if arg == phidef.
3170 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
3172 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Test
3173 for constant before any other processing.
3175 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
3177 * flag-types.h (RANGER_DEBUG_ALL): Fix values.
3179 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
3181 * gimple-range.cc (gimple_ranger::gimple_ranger): Initialize current_bb.
3182 (gimple_ranger::range_of_expr): Pick up range_on_entry when there is
3183 no explcit context and current_bb is set.
3184 (gimple_ranger::fold_stmt): New.
3185 * gimple-range.h (current_bb, fold_stmt): New.
3186 * tree-vrp.c (rvrp_folder::fold_stmt): Call ranger's fold_stmt.
3188 2021-11-03 Richard Biener <rguenther@suse.de>
3190 PR tree-optimization/102970
3191 * tree-ssa-pre.c (phi_translate_1): Drop clique and base
3192 when translating a MEM_REF over a backedge.
3194 2021-11-03 Philipp Tomsich <philipp.tomsich@vrull.eu>
3196 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New Ampere-1 core.
3197 * config/aarch64/aarch64-tune.md: Regenerate.
3198 * config/aarch64/aarch64-cost-tables.h: Add extra costs for Ampere-1.
3199 * config/aarch64/aarch64.c: Add tuning structures for Ampere-1.
3200 * doc/invoke.texi: Add documentation for Ampere-1 core.
3202 2021-11-03 Wilco Dijkstra <wdijkstr@arm.com>
3204 * config/aarch64/aarch64.md (movsi): Add alternative for GOT accesses.
3206 (ldr_got_small_<mode>): Remove pattern.
3207 (ldr_got_small_sidi): Likewise.
3208 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately): Keep
3209 GOT accesses as moves.
3210 (aarch64_print_operand): Correctly print got_lo12 in L specifier.
3211 (aarch64_mov_operand_p): Make GOT accesses valid move operands.
3212 * config/aarch64/constraints.md: Add new constraint Usw for GOT access.
3214 2021-11-03 Martin Liska <mliska@suse.cz>
3216 * gcov.c (read_line): Remove dead variable.
3218 2021-11-03 Martin Liska <mliska@suse.cz>
3221 * ipa-fnsummary.c (edge_predicate_pool): Rename predicate class to ipa_predicate.
3222 (ipa_fn_summary::account_size_time): Likewise.
3223 (edge_set_predicate): Likewise.
3224 (set_hint_predicate): Likewise.
3225 (add_freqcounting_predicate): Likewise.
3226 (evaluate_conditions_for_known_args): Likewise.
3227 (evaluate_properties_for_edge): Likewise.
3228 (remap_freqcounting_preds_after_dup): Likewise.
3229 (ipa_fn_summary_t::duplicate): Likewise.
3230 (set_cond_stmt_execution_predicate): Likewise.
3231 (set_switch_stmt_execution_predicate): Likewise.
3232 (compute_bb_predicates): Likewise.
3233 (will_be_nonconstant_expr_predicate): Likewise.
3234 (will_be_nonconstant_predicate): Likewise.
3235 (phi_result_unknown_predicate): Likewise.
3236 (predicate_for_phi_result): Likewise.
3237 (analyze_function_body): Likewise.
3238 (compute_fn_summary): Likewise.
3239 (summarize_calls_size_and_time): Likewise.
3240 (estimate_calls_size_and_time): Likewise.
3241 (ipa_call_context::estimate_size_and_time): Likewise.
3242 (remap_edge_summaries): Likewise.
3243 (remap_freqcounting_predicate): Likewise.
3244 (ipa_merge_fn_summary_after_inlining): Likewise.
3245 (ipa_update_overall_fn_summary): Likewise.
3246 (read_ipa_call_summary): Likewise.
3247 (inline_read_section): Likewise.
3248 * ipa-fnsummary.h (struct ipa_freqcounting_predicate): Likewise.
3249 * ipa-predicate.c (predicate::add_clause): Likewise.
3250 (ipa_predicate::add_clause): Likewise.
3251 (predicate::or_with): Likewise.
3252 (ipa_predicate::or_with): Likewise.
3253 (predicate::evaluate): Likewise.
3254 (ipa_predicate::evaluate): Likewise.
3255 (predicate::probability): Likewise.
3256 (ipa_predicate::probability): Likewise.
3257 (dump_condition): Likewise.
3258 (dump_clause): Likewise.
3259 (predicate::dump): Likewise.
3260 (ipa_predicate::dump): Likewise.
3261 (predicate::debug): Likewise.
3262 (ipa_predicate::debug): Likewise.
3263 (predicate::remap_after_duplication): Likewise.
3264 (ipa_predicate::remap_after_duplication): Likewise.
3265 (predicate::remap_after_inlining): Likewise.
3266 (ipa_predicate::remap_after_inlining): Likewise.
3267 (predicate::stream_in): Likewise.
3268 (ipa_predicate::stream_in): Likewise.
3269 (predicate::stream_out): Likewise.
3270 (ipa_predicate::stream_out): Likewise.
3271 (add_condition): Likewise.
3272 * ipa-predicate.h (class predicate): Likewise.
3273 (class ipa_predicate): Likewise.
3274 (add_condition): Likewise.
3276 2021-11-03 Richard Biener <rguenther@suse.de>
3278 * bitmap.h (bitmap_bit_p): Change the return type to bool.
3279 * bitmap.c (bitmap_bit_p): Likewise.
3280 * sbitmap.h (bitmap_bit_p): Likewise.
3281 (bitmap_set_bit): Return whether the bit changed.
3282 (bitmap_clear_bit): Likewise.
3283 * tree-ssa.c (verify_vssa): Make use of the changed state
3284 from bitmap_set_bit.
3286 2021-11-03 Richard Biener <rguenther@suse.de>
3288 PR middle-end/103033
3289 * internal-fn.c (expand_DEFERRED_INIT): Elide the
3290 native_interpret_expr path in favor of folding the
3291 VIEW_CONVERT_EXPR generated when punning the RHS.
3293 2021-11-03 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
3295 * config/s390/s390.c (s390_loop_unroll_adjust): In case of early
3298 2021-11-03 Jan Hubicka <hubicka@ucw.cz>
3301 * ipa-modref.c (callee_to_caller_flags): New function.
3302 (modref_eaf_analysis::analyze_ssa_name): Use it.
3303 (ipa_merge_modref_summary_after_inlining): Fix whitespace.
3305 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
3307 * ipa-modref.c (modref_eaf_analysis::analyze_ssa_name): Revert
3310 2021-11-02 Roger Sayle <roger@nextmovesoftware.com>
3311 Uroš Bizjak <ubizjak@gmail.com>
3313 * config/i386/i386.md (<any_rotate>ti3): Provide expansion for
3314 rotations by non-constant amounts.
3316 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
3318 * ipa-modref.c: Fix anonymous namespace placement.
3319 (class modref_eaf_analysis): New class.
3320 (analyze_ssa_name_flags): Turn to ...
3321 (modref_eaf_analysis::analyze_ssa_name): ... this one.
3322 (merge_call_lhs_flags): Turn to ...
3323 (modref_eaf_analysis::merge_call_lhs_flags): .. this one
3324 (modref_eaf_analysis::merge_with_ssa_name): New member function.
3325 (record_escape_points): Turn to ...
3326 (modref_eaf_analysis::record_escape_points): ... this one.
3327 (analyze_parms): Updat
3328 (ipa_merge_modref_summary_after_inlining): Move to the end of file.
3330 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
3332 * gimple.c (gimple_call_static_chain_flags): New function.
3333 * gimple.h (gimple_call_static_chain_flags): Declare
3334 * ipa-modref.c (modref_summary::modref_summary): Initialize
3336 (modref_summary_lto::modref_summary_lto): Likewise.
3337 (modref_summary::useful_p): Test static_chain_flags.
3338 (modref_summary_lto::useful_p): Likewise.
3339 (struct modref_summary_lto): Add static_chain_flags.
3340 (modref_summary::dump): Dump static_chain_flags.
3341 (modref_summary_lto::dump): Likewise.
3342 (struct escape_point): Add static_cahin_arg.
3343 (analyze_ssa_name_flags): Use gimple_call_static_chain_flags.
3344 (analyze_parms): Handle static chains.
3345 (modref_summaries::duplicate): Duplicate static_chain_flags.
3346 (modref_summaries_lto::duplicate): Likewise.
3347 (modref_write): Stream static_chain_flags.
3348 (read_section): Likewise.
3349 (modref_merge_call_site_flags): Handle static_chain_flags.
3350 * ipa-modref.h (struct modref_summary): Add static_chain_flags.
3351 * tree-ssa-structalias.c (handle_rhs_call): Use
3352 gimple_static_chain_flags.
3354 2021-11-02 Richard Biener <rguenther@suse.de>
3356 PR tree-optimization/103029
3357 * tree-vect-loop-manip.c (vect_loop_versioning): Ensure
3358 the PHI nodes in the loop maintain their original operand
3361 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
3363 * tree-core.h (EAF_NOT_RETURNED_DIRECTLY): New flag.
3364 (EAF_NOREAD): Renumber.
3365 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOT_RETURNED_DIRECTLY.
3366 (remove_useless_eaf_flags): Handle EAF_NOT_RETURNED_DIRECTLY
3367 (deref_flags): Likewise.
3368 (modref_lattice::init): Likewise.
3369 (modref_lattice::merge): Likewise.
3370 (merge_call_lhs_flags): Likewise.
3371 (analyze_ssa_name_flags): Likewise.
3372 (modref_merge_call_site_flags): Likewise.
3373 * tree-ssa-structalias.c (handle_call_arg): Likewise.
3375 2021-11-02 Maciej W. Rozycki <macro@embecosm.com>
3377 * config/riscv/riscv.c (riscv_rtx_costs): Correct a CONST_INT_P
3378 check and remove an unused local variable with shNadd/shNadd.uw
3381 2021-11-02 Tamar Christina <tamar.christina@arm.com>
3383 PR tree-optimization/103007
3384 * tree-vect-slp-patterns.c (complex_fms_pattern::matches): Add elem
3387 2021-11-02 Richard Biener <rguenther@suse.de>
3389 PR middle-end/103038
3390 * fold-const.c (native_interpret_expr): Handle OFFSET_TYPE.
3391 (can_native_interpret_type_p): Likewise.
3392 * internal-fn.c (expand_DEFERRED_INIT): View-convert the
3393 RHS if the LHS is an SSA name.
3395 2021-11-02 Richard Sandiford <richard.sandiford@arm.com>
3397 * langhooks.h (lang_hooks_for_types::simulate_record_decl): New hook.
3398 * langhooks-def.h (lhd_simulate_record_decl): Declare.
3399 (LANG_HOOKS_SIMULATE_RECORD_DECL): Define.
3400 (LANG_HOOKS_FOR_TYPES_INITIALIZER): Include it.
3401 * langhooks.c (lhd_simulate_record_decl): New function.
3403 2021-11-02 Jakub Jelinek <jakub@redhat.com>
3406 * config/i386/i386.c (ix86_vector_mode_supported_p): Reject vector
3407 modes with TImode inner mode if 32-bit.
3409 2021-11-02 liuhongt <hongtao.liu@intel.com>
3411 * doc/sourcebuild.texi (vect_slp_v4qi_store_unalign,
3412 vect_slp_v2hi_store_unalign, vect_slp_v4hi_store_unalign,
3413 vect_slp_v4si_store_unalign): Document efficient target.
3414 (vect_slp_v4qi_store_unalign_1, vect_slp_v8qi_store_unalign_1,
3415 vect_slp_v16qi_store_unalign_1): Ditto.
3416 (vect_slp_v2hi_store_align,vect_slp_v2qi_store_align,
3417 vect_slp_v2si_store_align, vect_slp_v4qi_store_align): Ditto.
3418 (struct_4char_block_move, struct_8char_block_move,
3419 struct_16char_block_move): Ditto.
3421 2021-11-02 Roger Sayle <roger@nextmovesoftware.com>
3422 Jakub Jelinek <jakub@redhat.com>
3425 * config/i386/i386-expand.c (ix86_expand_v1ti_to_ti,
3426 ix86_expand_ti_to_v1ti): New helper functions.
3427 (ix86_expand_v1ti_shift): Check if the amount operand is an
3428 integer constant, and expand as a TImode shift if it isn't.
3429 (ix86_expand_v1ti_rotate): Check if the amount operand is an
3430 integer constant, and expand as a TImode rotate if it isn't.
3431 (ix86_expand_v1ti_ashiftrt): New function to expand arithmetic
3432 right shifts of V1TImode quantities.
3433 * config/i386/i386-protos.h (ix86_expand_v1ti_ashift): Prototype.
3434 * config/i386/sse.md (ashlv1ti3, lshrv1ti3): Change constraints
3435 to QImode general_operand, and let the helper functions lower
3436 shifts by non-constant operands, as TImode shifts. Make
3437 conditional on TARGET_64BIT.
3438 (ashrv1ti3): New expander calling ix86_expand_v1ti_ashiftrt.
3439 (rotlv1ti3, rotrv1ti3): Change shift operand to QImode.
3440 Make conditional on TARGET_64BIT.
3442 2021-11-02 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
3444 * config/s390/s390.md ("*cc_to_int", "tabort", "*tabort_1",
3445 "*tabort_1_plus"): Remove operands non-null check.
3447 2021-11-01 Martin Liska <mliska@suse.cz>
3449 * opt-functions.awk: Add new sanity checking.
3450 * optc-gen.awk: Add new argument to integer_range_info.
3451 * params.opt: Update 2 params which have negative IntegerRange.
3453 2021-11-01 qing zhao <qing.zhao@oracle.com>
3455 * gimplify.c (gimplify_decl_expr): Do not add call to
3456 __builtin_clear_padding when a variable is a gimple register
3457 or it might not have padding.
3458 (gimplify_init_constructor): Likewise.
3460 2021-11-01 Tamar Christina <tamar.christina@arm.com>
3462 * config/arm/aarch-common-protos.h (struct vector_cost_table): Add
3463 movi, dup and extract costing fields.
3464 * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs,
3465 thunderx_extra_costs, thunderx2t99_extra_costs,
3466 thunderx3t110_extra_costs, tsv110_extra_costs, a64fx_extra_costs): Use
3468 * config/arm/aarch-cost-tables.h (generic_extra_costs,
3469 cortexa53_extra_costs, cortexa57_extra_costs, cortexa76_extra_costs,
3470 exynosm1_extra_costs, xgene1_extra_costs): Likewise
3471 * config/aarch64/aarch64-simd.md (aarch64_simd_dup<mode>): Add r->w dup.
3472 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add extra costs.
3474 2021-11-01 Tamar Christina <tamar.christina@arm.com>
3476 * cse.c (add_to_set): New.
3477 (find_sets_in_insn): Register constants in sets.
3478 (canonicalize_insn): Use auto_vec instead.
3479 (cse_insn): Try materializing using vec_dup.
3480 * rtl.h (simplify_context::simplify_gen_vec_select,
3481 simplify_gen_vec_select): New.
3482 * simplify-rtx.c (simplify_context::simplify_gen_vec_select): New.
3484 2021-11-01 David Malcolm <dmalcolm@redhat.com>
3486 * common.opt (fdiagnostics-escape-format=): New.
3487 (diagnostics_escape_format): New enum.
3488 (DIAGNOSTICS_ESCAPE_FORMAT_UNICODE): New enum value.
3489 (DIAGNOSTICS_ESCAPE_FORMAT_BYTES): Likewise.
3490 * diagnostic-format-json.cc (json_end_diagnostic): Add
3491 "escape-source" attribute.
3492 * diagnostic-show-locus.c
3493 (exploc_with_display_col::exploc_with_display_col): Replace
3494 "tabstop" param with a cpp_char_column_policy and add an "aspect"
3495 param. Use these to compute m_display_col accordingly.
3496 (struct char_display_policy): New struct.
3497 (layout::m_policy): New field.
3498 (layout::m_escape_on_output): New field.
3499 (def_policy): New function.
3500 (make_range): Update for changes to exploc_with_display_col ctor.
3501 (default_print_decoded_ch): New.
3502 (width_per_escaped_byte): New.
3503 (escape_as_bytes_width): New.
3504 (escape_as_bytes_print): New.
3505 (escape_as_unicode_width): New.
3506 (escape_as_unicode_print): New.
3508 (layout::layout): Initialize new fields. Update m_exploc ctor
3509 call for above change to ctor.
3510 (layout::maybe_add_location_range): Update for changes to
3511 exploc_with_display_col ctor.
3512 (layout::calculate_x_offset_display): Update for change to
3514 (layout::print_source_line): Pass policy
3515 to cpp_display_width_computation. Capture cpp_decoded_char when
3516 calling process_next_codepoint. Move printing of source code to
3517 m_policy.m_print_cb.
3518 (line_label::line_label): Pass in policy rather than context.
3519 (layout::print_any_labels): Update for change to line_label ctor.
3520 (get_affected_range): Pass in policy rather than context, updating
3521 calls to location_compute_display_column accordingly.
3522 (get_printed_columns): Likewise, also for cpp_display_width.
3523 (correction::correction): Pass in policy rather than tabstop.
3524 (correction::compute_display_cols): Pass m_policy rather than
3525 m_tabstop to cpp_display_width.
3526 (correction::m_tabstop): Replace with...
3527 (correction::m_policy): ...this.
3528 (line_corrections::line_corrections): Pass in policy rather than
3530 (line_corrections::m_context): Replace with...
3531 (line_corrections::m_policy): ...this.
3532 (line_corrections::add_hint): Update to use m_policy rather than
3534 (line_corrections::add_hint): Likewise.
3535 (layout::print_trailing_fixits): Likewise.
3536 (selftest::test_display_widths): New.
3537 (selftest::test_layout_x_offset_display_utf8): Update to use
3538 policy rather than tabstop.
3539 (selftest::test_one_liner_labels_utf8): Add test of escaping
3541 (selftest::test_diagnostic_show_locus_one_liner_utf8): Update to
3542 use policy rather than tabstop.
3543 (selftest::test_overlapped_fixit_printing): Likewise.
3544 (selftest::test_overlapped_fixit_printing_utf8): Likewise.
3545 (selftest::test_overlapped_fixit_printing_2): Likewise.
3546 (selftest::test_tab_expansion): Likewise.
3547 (selftest::test_escaping_bytes_1): New.
3548 (selftest::test_escaping_bytes_2): New.
3549 (selftest::diagnostic_show_locus_c_tests): Call the new tests.
3550 * diagnostic.c (diagnostic_initialize): Initialize
3551 context->escape_format.
3552 (convert_column_unit): Update to use default character width policy.
3553 (selftest::test_diagnostic_get_location_text): Likewise.
3554 * diagnostic.h (enum diagnostics_escape_format): New enum.
3555 (diagnostic_context::escape_format): New field.
3556 * doc/invoke.texi (-fdiagnostics-escape-format=): New option.
3557 (-fdiagnostics-format=): Add "escape-source" attribute to examples
3558 of JSON output, and document it.
3559 * input.c (location_compute_display_column): Pass in "policy"
3560 rather than "tabstop", passing to
3561 cpp_byte_column_to_display_column.
3562 (selftest::test_cpp_utf8): Update to use cpp_char_column_policy.
3563 * input.h (class cpp_char_column_policy): New forward decl.
3564 (location_compute_display_column): Pass in "policy" rather than
3566 * opts.c (common_handle_option): Handle
3567 OPT_fdiagnostics_escape_format_.
3568 * selftest.c (temp_source_file::temp_source_file): New ctor
3569 overload taking a size_t.
3570 * selftest.h (temp_source_file::temp_source_file): Likewise.
3572 2021-11-01 Aldy Hernandez <aldyh@redhat.com>
3574 * dbgcnt.def: Add debug counter for back_thread[12] and
3575 back_threadfull[12].
3576 * passes.def: Pass "first" argument to each back threading pass.
3577 * tree-ssa-threadbackward.c (back_threader::back_threader): Add
3579 (back_threader::debug_counter): New.
3580 (back_threader::maybe_register_path): Call debug_counter.
3582 2021-11-01 Aldy Hernandez <aldyh@redhat.com>
3584 * tree-ssa-threadbackward.c (BT_NONE): New.
3587 (back_threader::back_threader): Add flags.
3588 Move loop initialization here.
3589 (back_threader::~back_threader): New.
3590 (back_threader::find_taken_edge_switch): Change solver and ranger
3592 (back_threader::find_taken_edge_cond): Same.
3593 (back_threader::find_paths_to_names): Same.
3594 (back_threader::find_paths): Same.
3595 (back_threader::dump): Same.
3596 (try_thread_blocks): Merge into thread_blocks.
3597 (back_threader::thread_blocks): New.
3598 (do_early_thread_jumps): Merge into thread_blocks.
3599 (do_thread_jumps): Merge into thread_blocks.
3600 (back_threader::thread_through_all_blocks): Remove.
3602 2021-11-01 Andrew MacLeod <amacleod@redhat.com>
3604 PR tree-optimization/103003
3605 * value-relation.cc (dom_oracle::register_relation): If the 2
3606 ssa names are the same, don't register any relation.
3608 2021-11-01 Dan Li <ashimida@linux.alibaba.com>
3610 * config/aarch64/aarch64.c (aarch64_expand_epilogue): Remove
3611 redundant check for calls_eh_return.
3612 * config/aarch64/aarch64.md (*do_return): Likewise.
3614 2021-11-01 Xionghu Luo <luoxhu@linux.ibm.com>
3616 * cfghooks.c (cfg_hook_duplicate_loop_to_header_edge): Rename
3617 duplicate_loop_to_header_edge to
3618 duplicate_loop_body_to_header_edge.
3619 (cfg_hook_duplicate_loop_body_to_header_edge): Likewise.
3620 * cfghooks.h (struct cfg_hooks): Likewise.
3621 (cfg_hook_duplicate_loop_body_to_header_edge): Likewise.
3622 * cfgloopmanip.c (duplicate_loop_body_to_header_edge): Likewise.
3623 (clone_loop_to_header_edge): Likewise.
3624 * cfgloopmanip.h (duplicate_loop_body_to_header_edge): Likewise.
3625 * cfgrtl.c (struct cfg_hooks): Likewise.
3626 * doc/loop.texi: Likewise.
3627 * loop-unroll.c (unroll_loop_constant_iterations): Likewise.
3628 (unroll_loop_runtime_iterations): Likewise.
3629 (unroll_loop_stupid): Likewise.
3630 (apply_opt_in_copies): Likewise.
3631 * tree-cfg.c (struct cfg_hooks): Likewise.
3632 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
3633 (try_peel_loop): Likewise.
3634 * tree-ssa-loop-manip.c (copy_phi_node_args): Likewise.
3635 (gimple_duplicate_loop_body_to_header_edge): Likewise.
3636 (tree_transform_and_unroll_loop): Likewise.
3637 * tree-ssa-loop-manip.h (gimple_duplicate_loop_body_to_header_edge):
3640 2021-11-01 Xionghu Luo <luoxhu@linux.ibm.com>
3642 * cfgloopmanip.c (loop_version): Refactor loopify to
3643 loop_version. Move condition generation after loopify.
3645 * cfgloopmanip.h (loopify): Delete.
3647 2021-10-31 Jan Hubicka <hubicka@ucw.cz>
3649 * ipa-fnsummary.c: Include tree-dfa.h.
3650 (points_to_local_or_readonly_memory_p): Return true on return
3652 * ipa-modref.c (analyze_ssa_name_flags): Fix handling of copy
3655 2021-10-30 Tobias Burnus <tobias@codesourcery.com>
3657 PR middle-end/102972
3658 * omp-low.c (omp_runtime_api_call): Use DECL_ASSEMBLER_NAME to get
3659 internal Fortran name; new permit_num_teams arg to permit
3660 omp_get_num_teams and omp_get_team_num.
3661 (scan_omp_1_stmt): Update call to it, add missing call for
3662 reverse offload, and check for strictly nested API calls in teams.
3664 2021-10-30 Jakub Jelinek <jakub@redhat.com>
3666 * gimplify.c (gimplify_omp_for): Diagnose threadprivate iterators.
3668 2021-10-29 David Malcolm <dmalcolm@redhat.com>
3670 * selftest.c (assert_streq): Add newlines when emitting non-equal
3673 2021-10-29 David Malcolm <dmalcolm@redhat.com>
3675 * Makefile.in: Fix syntax for reference to LIBDEPS in
3678 2021-10-29 Bernhard Reutner-Fischer <aldot@gcc.gnu.org>
3680 * doc/install.texi: Bump required minimum DejaGnu version.
3682 2021-10-29 Aldy Hernandez <aldyh@redhat.com>
3683 Andrew MacLeod <amacleod@redhat.com>
3685 * value-relation.cc (path_oracle::killing_def): Add a
3686 self-equivalence so we don't look to the root oracle.
3688 2021-10-29 Aldy Hernandez <aldyh@redhat.com>
3690 * passes.def: Replace the pass_thread_jumps before VRP* with
3691 pass_thread_jumps_full. Remove all pass_vrp_threader instances.
3692 * tree-ssa-threadbackward.c (pass_data_thread_jumps_full):
3693 Remove hyphen from "thread-full" name.
3695 2021-10-29 Andrew MacLeod <amacleod@redhat.com>
3697 PR tree-optimization/102983
3698 * gimple-range-cache.h (propagate_updated_value): Make public.
3699 * gimple-range.cc (gimple_ranger::range_of_stmt): Propagate exports
3700 when processing gcond stmts.
3702 2021-10-29 Jan Hubicka <hubicka@ucw.cz>
3704 * gimple.c (gimple_call_retslot_flags): New function.
3705 * gimple.h (gimple_call_retslot_flags): Declare.
3706 * ipa-modref.c: Include tree-cfg.h.
3707 (struct escape_entry): Turn parm_index to signed.
3708 (modref_summary_lto::modref_summary_lto): Add retslot_flags.
3709 (modref_summary::modref_summary): Initialize retslot_flags.
3710 (struct modref_summary_lto): Likewise.
3711 (modref_summary::useful_p): Check retslot_flags.
3712 (modref_summary_lto::useful_p): Likewise.
3713 (modref_summary::dump): Dump retslot_flags.
3714 (modref_summary_lto::dump): Likewise.
3715 (struct escape_point): Add hidden_args enum.
3716 (analyze_ssa_name_flags): Ignore return slot return;
3717 use gimple_call_retslot_flags.
3718 (record_escape_points): Break out from ...
3719 (analyze_parms): ... here; handle retslot_flags.
3720 (modref_summaries::duplicate): Duplicate retslot_flags.
3721 (modref_summaries_lto::duplicate): Likewise.
3722 (modref_write_escape_summary): Stream parm_index as signed.
3723 (modref_read_escape_summary): Likewise.
3724 (modref_write): Stream retslot_flags.
3725 (read_section): Likewise.
3726 (struct escape_map): Fix typo in comment.
3727 (update_escape_summary_1): Fix whitespace.
3728 (ipa_merge_modref_summary_after_inlining): Drop retslot_flags.
3729 (modref_merge_call_site_flags): Merge retslot_flags.
3730 * ipa-modref.h (struct modref_summary): Add retslot_flags.
3731 * tree-ssa-structalias.c (handle_rhs_call): Handle retslot_flags.
3733 2021-10-29 Tamar Christina <tamar.christina@arm.com>
3735 PR tree-optimization/102977
3736 * tree-vect-slp-patterns.c (vect_match_call_p): Remove.
3737 (vect_detect_pair_op): Add crosslane check.
3738 (vect_match_call_complex_mla): Remove.
3739 (class complex_mul_pattern): Update comment.
3740 (complex_mul_pattern::matches): Update detection.
3741 (class complex_fma_pattern): Remove.
3742 (complex_fma_pattern::matches): Remove.
3743 (complex_fma_pattern::recognize): Remove.
3744 (complex_fma_pattern::build): Remove.
3745 (class complex_fms_pattern): Update comment.
3746 (complex_fms_pattern::matches): Remove.
3747 (complex_operations_pattern::recognize): Remove complex_fma_pattern
3749 2021-10-29 Jakub Jelinek <jakub@redhat.com>
3751 * gimple-fold.c (gimple_fold_builtin_memset): Copy over location from
3754 2021-10-29 Xionghu Luo <luoxhu@linux.ibm.com>
3757 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Add
3758 patterns match and emit for VSX xxpermdi.
3760 2021-10-29 liuhongt <hongtao.liu@intel.com>
3763 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF): New
3765 (V16HF_FTYPE_V16HF): Ditto.
3766 (V32HF_FTYPE_V32HF): Ditto.
3767 (V8HF_FTYPE_V8HF_ROUND): Ditto.
3768 (V16HF_FTYPE_V16HF_ROUND): Ditto.
3769 (V32HF_FTYPE_V32HF_ROUND): Ditto.
3770 * config/i386/i386-builtin.def ( IX86_BUILTIN_FLOORPH,
3771 IX86_BUILTIN_CEILPH, IX86_BUILTIN_TRUNCPH,
3772 IX86_BUILTIN_FLOORPH256, IX86_BUILTIN_CEILPH256,
3773 IX86_BUILTIN_TRUNCPH256, IX86_BUILTIN_FLOORPH512,
3774 IX86_BUILTIN_CEILPH512, IX86_BUILTIN_TRUNCPH512): New builtin.
3775 * config/i386/i386-builtins.c
3776 (ix86_builtin_vectorized_function): Enable vectorization for
3777 HFmode FLOOR/CEIL/TRUNC operation.
3778 * config/i386/i386-expand.c (ix86_expand_args_builtin): Handle
3780 * config/i386/sse.md (rint<mode>2, nearbyint<mode>2): Extend
3783 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
3784 Andrew MacLeod <amacleod@redhat.com>
3786 * value-relation.cc (path_oracle::killing_def): Walk the
3787 equivalency list and remove SSA from any equivalencies.
3789 2021-10-28 Stafford Horne <shorne@gmail.com>
3791 * config/or1k/or1k.h (PROFILE_HOOK): Add return address argument
3794 2021-10-28 Jakub Jelinek <jakub@redhat.com>
3796 PR tree-optimization/102951
3797 * fold-const.h (address_compare): Declare.
3798 * fold-const.c (address_compare): New function.
3799 * match.pd (cmp (convert1?@2 addr@0) (convert2? addr@1)): Use
3800 address_compare helper.
3801 (minmax cmp (convert1?@2 addr@0) (convert2?@3 addr@1)): New
3804 2021-10-28 Andrew MacLeod <amacleod@redhat.com>
3806 * vr-values.c (simplify_using_ranges::fold_cond): Change fold message.
3808 2021-10-28 Andrew MacLeod <amacleod@redhat.com>
3810 PR tree-optimization/102940
3811 * tree-vrp.c (execute_ranger_vrp): Reset scev.
3813 2021-10-28 Richard Purdie <richard.purdie@linuxfoundation.org>
3815 * config/nios2/linux.h (MUSL_DYNAMIC_LINKER): Add musl linker
3817 2021-10-28 Richard Purdie <richard.purdie@linuxfoundation.org>
3819 * configure: Regenerate.
3820 * configure.ac: Use CPPFLAGS_FOR_BUILD for GMPINC
3822 2021-10-28 Eric Botcazou <ebotcazou@adacore.com>
3824 * doc/invoke.texi (%X): Remove obsolete reference to -Wl.
3826 2021-10-28 Richard Biener <rguenther@suse.de>
3829 * fold-const.c (fold_convert_const): Avoid int to float
3830 constant folding with -frounding-math and inexact result.
3831 * simplify-rtx.c (simplify_const_unary_operation): Likewise
3832 for both float and unsigned_float.
3834 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
3836 * tree-ssa-threadbackward.c
3837 (back_threader::find_taken_edge_switch): Use find_case_label_range
3838 instead of find_taken_edge.
3840 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
3842 * tree-ssa-threadbackward.c (class back_threader_registry):
3843 Inherit from back_jt_path_registry.
3844 (back_threader_registry::thread_through_all_blocks): Remove.
3845 (back_threader_registry::register_path): Remove
3846 m_lowlevel_registry prefix.
3848 2021-10-28 Richard Biener <rguenther@suse.de>
3851 * fold-const.c (fold_convert_const_real_from_real): Honor
3852 -frounding-math if the conversion is not exact.
3853 * simplify-rtx.c (simplify_const_unary_operation): Do not
3854 simplify FLOAT_TRUNCATE with sign dependent rounding.
3856 2021-10-28 Richard Biener <rguenther@suse.de>
3858 PR tree-optimization/102949
3859 * tree-vect-stmts.c (ensure_base_align): Look at the
3860 dr_info of a group leader and assert we are looking at
3861 one with analyzed alignment.
3863 2021-10-28 Kewen Lin <linkw@linux.ibm.com>
3866 * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Consider
3867 V1T1 mode for unaligned load and store.
3869 2021-10-28 Kito Cheng <kito.cheng@sifive.com>
3871 * config/riscv/riscv.md (zero_extendsidi2_internal): Allow ZBB
3874 2021-10-28 Kito Cheng <kito.cheng@sifive.com>
3876 * config/riscv/arch-canonicalize (CANONICAL_ORDER): Add `i` to
3879 2021-10-28 Alexandre Oliva <oliva@adacore.com>
3881 * common.opt (fharden-compares): New.
3882 (fharden-conditional-branches): New.
3883 * doc/invoke.texi: Document new options.
3884 * gimple-harden-conditionals.cc: New.
3885 * Makefile.in (OBJS): Build it.
3886 * passes.def: Add new passes.
3887 * tree-pass.h (make_pass_harden_compares): Declare.
3888 (make_pass_harden_conditional_branches): Declare.
3890 2021-10-28 Xionghu Luo <luoxhu@linux.ibm.com>
3893 * config/rs6000/altivec.md: Add vsx register constraints.
3894 * config/rs6000/vsx.md (vsx_xxsel<mode>): Delete.
3895 (vsx_xxsel<mode>2): Likewise.
3896 (vsx_xxsel<mode>3): Likewise.
3897 (vsx_xxsel<mode>4): Likewise.
3899 2021-10-28 Xionghu Luo <luoxhu@linux.ibm.com>
3902 * config/rs6000/altivec.md (*altivec_vsel<mode>): Change to ...
3903 (altivec_vsel<mode>): ... this and update define.
3904 (*altivec_vsel<mode>_uns): Delete.
3905 (altivec_vsel<mode>2): New define_insn.
3906 (altivec_vsel<mode>3): Likewise.
3907 (altivec_vsel<mode>4): Likewise.
3908 * config/rs6000/rs6000-call.c (altivec_expand_vec_sel_builtin): New.
3909 (altivec_expand_builtin): Call altivec_expand_vec_sel_builtin to expand
3911 * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Use bit-wise
3912 selection instead of per element.
3913 * config/rs6000/vector.md:
3914 * config/rs6000/vsx.md (*vsx_xxsel<mode>): Change to ...
3915 (vsx_xxsel<mode>): ... this and update define.
3916 (*vsx_xxsel<mode>_uns): Delete.
3917 (vsx_xxsel<mode>2): New define_insn.
3918 (vsx_xxsel<mode>3): Likewise.
3919 (vsx_xxsel<mode>4): Likewise.
3921 2021-10-28 Hongyu Wang <hongyu.wang@intel.com>
3923 * config/i386/i386.c (use_rsqrt_p): Add mode parameter, enable
3924 HFmode rsqrt without TARGET_SSE_MATH.
3925 (ix86_optab_supported_p): Refactor rint, adjust floor, ceil,
3926 btrunc condition to be restricted by -ftrapping-math, adjust
3927 use_rsqrt_p function call.
3928 * config/i386/i386.md (rcphf2): New define_insn.
3929 (rsqrthf2): Likewise.
3930 * config/i386/sse.md (div<mode>3): Change VF2H to VF2.
3931 (div<mode>3): New expander for HF mode.
3932 (rsqrt<mode>2): Likewise.
3933 (*avx512fp16_vmrcpv8hf2): New define_insn for rpad pass.
3934 (*avx512fp16_vmrsqrtv8hf2): Likewise.
3936 2021-10-27 Saagar Jha <saagar@saagarjha.com>
3938 * config.gcc: Adjust for Darwin21.
3939 * config/darwin-c.c (macosx_version_as_macro): Likewise.
3940 * config/darwin-driver.c (validate_macosx_version_min):
3942 (darwin_find_version_from_kernel): Likewise.
3944 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
3946 * gimple-range-path.cc
3947 (path_range_query::range_defined_in_block): Call killing_def.
3949 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
3951 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Dump
3952 operands as well as relation.
3953 * gimple-range-path.cc
3954 (path_range_query::compute_ranges_in_block): Compute PHI relations
3955 first. Compute outgoing relations at the end.
3956 (path_range_query::compute_ranges): Remove call to compute_relations.
3957 (path_range_query::compute_relations): Remove.
3958 (path_range_query::maybe_register_phi_relation): New.
3959 (path_range_query::compute_phi_relations): Abstract out
3960 registering one PHI relation to...
3961 (path_range_query::compute_outgoing_relations): ...here.
3962 * gimple-range-path.h (class path_range_query): Remove
3964 Add maybe_register_phi_relation.
3966 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
3967 Andrew MacLeod <amacleod@redhat.com>
3969 * value-relation.cc (path_oracle::killing_def): Kill second
3972 2021-10-27 Martin Jambor <mjambor@suse.cz>
3974 * ipa-cp.c (good_cloning_opportunity_p): Decide whether to use
3975 profile feedback depending on their local availability.
3977 2021-10-27 Martin Jambor <mjambor@suse.cz>
3979 * params.opt (param_ipa_cp_profile_count_base): New parameter.
3980 * doc/invoke.texi (Optimize Options): Add entry for
3981 ipa-cp-profile-count-base.
3982 * ipa-cp.c (max_count): Replace with base_count, replace all
3983 occurrences too, unless otherwise stated.
3984 (ipcp_cloning_candidate_p): identify mostly-directly called
3985 functions based on their counts, not max_count.
3986 (compare_edge_profile_counts): New function.
3987 (ipcp_propagate_stage): Instead of setting max_count, find the
3988 appropriate edge count in a sorted vector of counts of eligible
3989 edges and make it the base_count.
3991 2021-10-27 Martin Jambor <mjambor@suse.cz>
3993 * ipa-cp.c (struct caller_statistics): New fields rec_count_sum,
3994 n_nonrec_calls and itself, document all fields.
3995 (init_caller_stats): Initialize the above new fields.
3996 (gather_caller_stats): Gather self-recursive counts and calls number.
3997 (get_info_about_necessary_edges): Gather counts of self-recursive and
3998 other edges bringing in the requested value separately.
3999 (dump_profile_updates): Rework to dump info about a single node only.
4000 (lenient_count_portion_handling): New function.
4001 (struct gather_other_count_struct): New type.
4002 (gather_count_of_non_rec_edges): New function.
4003 (struct desc_incoming_count_struct): New type.
4004 (analyze_clone_icoming_counts): New function.
4005 (adjust_clone_incoming_counts): Likewise.
4006 (update_counts_for_self_gen_clones): Likewise.
4007 (update_profiling_info): Rewritten.
4008 (update_specialized_profile): Adjust call to dump_profile_updates.
4009 (create_specialized_node): Do not update profiling info.
4010 (decide_about_value): New parameter self_gen_clones, either push new
4011 clones into it or updat their profile counts. For self-recursively
4012 generated values, use a portion of the node count instead of count
4013 from self-recursive edges to estimate goodness.
4014 (decide_whether_version_node): Gather clones for self-generated values
4015 in a new vector, update their profiles at once at the end.
4017 2021-10-27 Richard Biener <rguenther@suse.de>
4019 * tree-vectorizer.c (vect_transform_loops): New function,
4021 (try_vectorize_loop_1): ... here. Simplify as epilogues
4022 are now fully handled in the split part.
4024 2021-10-27 Jakub Jelinek <jakub@redhat.com>
4026 * omp-expand.c (expand_omp_for_init_counts): Handle non-rectangular
4027 iterators with pointer types.
4028 (expand_omp_for_init_vars, extract_omp_for_update_vars): Likewise.
4030 2021-10-26 Martin Sebor <msebor@redhat.com>
4032 PR tree-optimization/102238
4033 PR tree-optimization/102919
4034 * gimple-ssa-sprintf.c (get_string_length): Add an argument.
4035 (array_elt_at_offset): Move to pointer-query.
4036 (set_aggregate_size_and_offset): New function.
4037 (field_at_offset): Move to pointer-query.
4038 (get_origin_and_offset): Rename...
4039 (get_origin_and_offset_r): this. Add an argument. Make aggregate
4040 handling more robust.
4041 (get_origin_and_offset): New.
4042 (alias_offset): Add an argument.
4043 (format_string): Use subobject size determined by get_origin_and_offset.
4044 * pointer-query.cc (field_at_offset): Move from gimple-ssa-sprintf.c.
4045 Improve/correct handling of aggregates.
4046 (array_elt_at_offset): Same.
4047 * pointer-query.h (field_at_offset): Declare.
4048 (array_elt_at_offset): Declare.
4050 2021-10-26 Martin Sebor <msebor@redhat.com>
4052 * builtins.c (check_strncat_sizes): Pass access_data ctor additional
4054 (expand_builtin_memcmp): Move code to gimple-ssa-warn-access.cc.
4055 (expand_builtin_fork_or_exec): Same.
4056 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Pass
4057 compute_objsize additional arguments.
4058 (inbounds_memaccess_p): Same.
4059 (array_bounds_checker::check_array_bounds): Add an assert. Stash
4060 statement in a member.
4061 (check_array_bounds_dom_walker::before_dom_children): Same.
4062 * gimple-array-bounds.h (array_bounds_checker::m_stmt): New member.
4063 * gimple-ssa-sprintf.c (get_destination_size): Add an argument.
4064 (handle_printf_call): Pass a new argument.
4065 * gimple-ssa-warn-access.cc (get_size_range): Add an argument.
4066 (check_access): Add an argument and pass it along to callees.
4067 (check_read_access): Make a member function.
4068 (pass_waccess::check_strcat): Pass access_data ctor additional
4070 (pass_waccess::check_strncat): Same.
4071 (pass_waccess::check_stxcpy): Same.
4072 (pass_waccess::check_stxncpy): Same.
4073 (pass_waccess::check_strncmp): Same.
4074 (pass_waccess::check_read_access): Same.
4075 (pass_waccess::check_builtin): Same.
4076 (pass_waccess::maybe_check_access_sizes): Same.
4077 (pass_waccess::maybe_check_dealloc_call): Same.
4078 * gimple-ssa-warn-access.h (check_read_access): Declare a new
4080 * pointer-query.cc (compute_objsize_r): Add an argument.
4081 (gimple_call_return_array): Same.
4082 (gimple_call_alloc_size): Same.
4083 (access_ref::access_ref): Same.
4084 (access_ref::get_ref): Same.
4085 (pointer_query::get_ref): Same.
4086 (handle_min_max_size): Pass an arguments to callees.
4087 (handle_array_ref): Add an argument.
4088 (handle_mem_ref): Same.
4089 (compute_objsize): Same.
4090 * pointer-query.h (struct access_ref): Adjust signatures.
4091 (struct access_data): Same.
4092 (gimple_call_alloc_size): Add an argument.
4093 (gimple_parm_array_size): Same.
4094 (compute_objsize): Same.
4095 * tree-ssa-strlen.c (strlen_pass::adjust_last_stmt): Pass an additional
4096 argument to compute_objsize.
4097 (strlen_pass::maybe_warn_overflow): Same.
4098 (maybe_diag_stxncpy_trunc): Same.
4100 2021-10-26 Martin Sebor <msebor@redhat.com>
4102 PR middle-end/102453
4103 * gimple-ssa-warn-access.cc (pass_waccess::check_atomic_builtin): New.
4104 (pass_waccess::check_atomic_builtin): Call it.
4106 2021-10-26 Vladimir N. Makarov <vmakarov@redhat.com>
4108 PR rtl-optimization/102842
4109 * lra-constraints.c (match_reload): Ignore out in checking values
4111 (curr_insn_transform): Collect outputs before doing reloads of operands.
4113 2021-10-26 Paul A. Clarke <pc@us.ibm.com>
4116 * config/rs6000/x86intrin.h: Move some included headers to new
4117 headers. Include new immintrin.h instead of those headers.
4118 * config/rs6000/immintrin.h: New.
4119 * config/rs6000/x86gprintrin.h: New.
4120 * config.gcc (powerpc*-*-*): Add new headers to extra_headers.
4122 2021-10-26 Richard Biener <rguenther@suse.de>
4124 * tree-vectorizer.h (vect_create_addr_base_for_vector_ref):
4125 Remove byte_offset parameter.
4126 (vect_create_data_ref_ptr): Likewise.
4127 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
4129 (vect_create_data_ref_ptr): Likewise.
4130 * tree-vect-stmts.c (vectorizable_store): Adjust.
4131 (vectorizable_load): Likewise.
4133 2021-10-26 Richard Biener <rguenther@suse.de>
4135 PR tree-optimization/96109
4136 * tree-vectorizer.h (dr_misalignment): Add optional offset
4138 * tree-vect-data-refs.c (dr_misalignment): Likewise. Remove
4139 offset applied for negative stride accesses.
4140 (vect_enhance_data_refs_alignment): Compute negative stride
4141 access offset and pass it to dr_misalignment.
4142 * tree-vect-stmts.c (get_negative_load_store_type): Pass
4143 negative offset to dr_misalignment.
4144 (get_group_load_store_type): Likewise.
4145 (get_load_store_type): Likewise.
4146 (vectorizable_store): Remove asserts about alignment.
4147 (vectorizable_load): Likewise.
4149 2021-10-26 Kewen Lin <linkw@linux.ibm.com>
4151 PR tree-optimization/102897
4152 * tree-ssa-forwprop.c (simplify_permutation): Remove a wrong assertion.
4154 2021-10-26 Richard Biener <rguenther@suse.de>
4156 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
4157 Take offset in bytes.
4158 (vect_create_data_ref_ptr): Likewise.
4159 * tree-vect-loop-manip.c (get_misalign_in_elems): Multiply
4160 offset by element size.
4161 (vect_create_cond_for_align_checks): Likewise.
4162 * tree-vect-stmts.c (get_negative_load_store_type): Likewise.
4163 (vectorizable_load): Remove duplicate leftover from merge
4166 2021-10-26 Roger Sayle <roger@nextmovesoftware.com>
4168 * config/i386/i386-expand.c (ix86_expand_v1ti_shift): New helper
4169 function to expand V1TI mode logical shifts by integer constants.
4170 (ix86_expand_v1ti_rotate): New helper function to expand V1TI
4171 mode rotations by integer constants.
4172 * config/i386/i386-protos.h (ix86_expand_v1ti_shift,
4173 ix86_expand_v1ti_rotate): Prototype new functions here.
4174 * config/i386/sse.md (ashlv1ti3, lshrv1ti3, rotlv1ti3, rotrv1ti3):
4175 New TARGET_SSE2 expanders to implement V1TI shifts and rotations.
4177 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
4179 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
4180 Avoid threading circular paths.
4182 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
4184 * tree-ssa-threadbackward.c (back_threader::resolve_phi):
4185 Attempt to resolve all incoming paths to a PHI.
4186 (back_threader::resolve_def): Always return true for PHIs.
4188 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
4190 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
4191 Always try to resolve path without looking back.
4192 * tree-ssa-threadupdate.c (dump_jump_thread): Indidicate whether
4193 edge is a back edge.
4195 2021-10-26 Kewen Lin <linkw@linux.ibm.com>
4197 PR tree-optimization/102789
4198 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
4199 update inits of simd_lane_access.
4201 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
4203 * tree-vrp.c (vrp_simplify_cond_using_ranges): Add return type and
4204 move to vr-values.c.
4205 (simplify_casted_conds): Move to vrp_folder class.
4206 (execute_vrp): Call via vrp_folder now.
4207 * vr-values.c (simplify_cond_using_ranges_1): Call simplify_casted_cond.
4208 (simplify_using_ranges::simplify_casted_cond): Relocate from tree-vrp.c.
4209 * vr-values.h (simplify_casted_cond): Add prototype.
4211 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
4213 * tree-vrp.c (rvrp_folder::fold_stmt): If simplification fails, try
4216 2021-10-25 Paul A. Clarke <pc@us.ibm.com>
4218 * config/rs6000/smmintrin.h (_mm_testz_si128): Add "extern" to
4220 (_mm_testc_si128): Likewise.
4221 (_mm_testnzc_si128): Likewise.
4222 (_mm_blend_ps): Likewise.
4223 (_mm_blendv_ps): Likewise.
4224 (_mm_blend_pd): Likewise.
4225 (_mm_blendv_pd): Likewise.
4226 (_mm_ceil_pd): Likewise.
4227 (_mm_ceil_sd): Likewise.
4228 (_mm_ceil_ps): Likewise.
4229 (_mm_ceil_ss): Likewise.
4230 (_mm_floor_pd): Likewise.
4231 (_mm_floor_sd): Likewise.
4232 (_mm_floor_ps): Likewise.
4233 (_mm_floor_ss): Likewise.
4234 (_mm_minpos_epu16): Likewise.
4235 (_mm_mul_epi32): Likewise.
4236 (_mm_cvtepi8_epi16): Likewise.
4237 (_mm_packus_epi32): Likewise.
4238 (_mm_cmpgt_epi64): Likewise.
4240 2021-10-25 Roger Sayle <roger@nextmovesoftware.com>
4242 * simplify-rtx.c (simplify_binary_operation_1) [SS_ASHIFT]: Simplify
4243 shifts of the mode's smin_value and smax_value when the bit count
4244 operand doesn't have side-effects.
4245 [US_ASHIFT]: Likewise, simplify shifts of the mode's umax_value
4246 when the bit count operand doesn't have side-effects.
4247 (simplify_const_binary_operation) [SS_ASHIFT, US_ASHIFT]: Perform
4248 compile-time evaluation of saturating left shifts with constant
4251 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
4253 * gimple-range.cc (gimple_ranger::export_global_ranges): Remove check
4256 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
4258 * flag-types.h (enum ranger_debug): Adjust values.
4259 * params.opt (ranger_debug): Ditto.
4261 2021-10-25 Martin Jambor <mjambor@suse.cz>
4263 PR tree-optimization/102886
4264 * tree-sra.c (totally_scalarize_subtree): Fix the out of
4267 2021-10-25 Andrew Pinski <apinski@marvell.com>
4269 * tree-ssa-dce.c (simple_dce_from_worklist):
4270 Check stmt_unremovable_because_of_non_call_eh_p also
4271 before removing the statement.
4273 2021-10-25 Richard Biener <rguenther@suse.de>
4275 PR tree-optimization/102905
4276 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
4277 Use vect_supportable_dr_alignment again to determine whether
4278 an access is supported when not aligned.
4280 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
4282 * config/riscv/riscv.c (riscv_rtx_costs): Handle cost model
4285 2021-10-25 Jim Wilson <jimw@sifive.com>
4286 Kito Cheng <kito.cheng@sifive.com>
4287 Jia-Wei Chen <jiawei@iscas.ac.cn>
4288 Shi-Hua Liao <shihua@iscas.ac.cn>
4290 * config/riscv/bitmanip.md (shiftm1): New.
4291 (*bset<mode>): Ditto.
4292 (*bset<mode>_mask): Ditto.
4293 (*bset<mode>_1): Ditto.
4294 (*bset<mode>_1_mask): Ditto.
4295 (*bseti<mode>): Ditto.
4296 (*bclr<mode>): Ditto.
4297 (*bclri<mode>): Ditto.
4298 (*binv<mode>): Ditto.
4299 (*binvi<mode>): Ditto.
4300 (*bext<mode>): Ditto.
4302 * config/riscv/predicates.md (splittable_const_int_operand):
4304 (single_bit_mask_operand): New.
4305 (not_single_bit_mask_operand): Ditto.
4306 (const31_operand): Ditto.
4307 (const63_operand): Ditto.
4308 * config/riscv/riscv.c (riscv_build_integer_1): Handle bseti.
4309 (riscv_output_move): Ditto.
4310 (riscv_print_operand): Handle new operand type: T and S.
4311 * config/riscv/riscv.h (SINGLE_BIT_MASK_OPERAND): New.
4313 2021-10-25 Jim Wilson <jimw@sifive.com>
4315 * config/riscv/riscv.c (riscv_build_integer_1): Build integer
4318 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
4320 * config/riscv/riscv.c (riscv_extend_cost): Handle cost model
4322 (riscv_rtx_costs): Ditto.
4324 2021-10-25 Jim Wilson <jimw@sifive.com>
4325 Kito Cheng <kito.cheng@sifive.com>
4326 Jia-Wei Chen <jiawei@iscas.ac.cn>
4328 * config/riscv/bitmanip.md (bitmanip_bitwise): New.
4329 (bitmanip_minmax): New.
4330 (clz_ctz_pcnt): New.
4331 (bitmanip_optab): New.
4332 (bitmanip_insn): New.
4333 (*<optab>_not<mode>): New.
4334 (*xor_not<mode>): New.
4335 (<bitmanip_optab>si2): New.
4336 (*<bitmanip_optab>disi2): New.
4337 (<bitmanip_optab>di2): New.
4338 (*zero_extendhi<GPR:mode>2_bitmanip): New.
4339 (*extend<SHORT:mode><SUPERQI:mode>2_zbb): New.
4340 (*zero_extendhi<GPR:mode>2_zbb): New.
4343 (rotrsi3_sext): New.
4346 (rotlsi3_sext): New.
4347 (bswap<mode>2): New.
4348 (<bitmanip_optab><mode>3): New.
4349 * config/riscv/riscv.md (type): Add rotate.
4350 (zero_extendhi<GPR:mode>2): Change to define_expand pattern.
4351 (*zero_extendhi<GPR:mode>2): New.
4352 (extend<SHORT:mode><SUPERQI:mode>2): Change to define_expand pattern.
4353 (*extend<SHORT:mode><SUPERQI:mode>2): New.
4355 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
4357 * config/riscv/riscv.c (riscv_extend_cost): Handle cost model
4359 (riscv_rtx_costs): Ditto.
4361 2021-10-25 Jim Wilson <jimw@sifive.com>
4362 Kito Cheng <kito.cheng@sifive.com>
4363 Jia-Wei Chen <jiawei@iscas.ac.cn>
4365 * config/riscv/bitmanip.md (*zero_extendsidi2_bitmanip): New.
4370 (riscv_rtx_costs): Ditto.
4371 * config/riscv/riscv.md: Include bitmanip.md
4372 (type): Add bitmanip bype.
4373 (zero_extendsidi2): Change to define_expand pattern.
4374 (*zero_extendsidi2_internal): New.
4375 (zero_extendsidi2_shifted): Disable for ZBA.
4377 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
4379 * common/config/riscv/riscv-common.c (riscv_ext_version_table):
4380 Add zba, zbb, zbc and zbs.
4381 (riscv_ext_flag_table): Ditto.
4382 * config/riscv/riscv-opts.h (MASK_ZBA): New.
4386 (TARGET_ZBA): Ditto.
4387 (TARGET_ZBB): Ditto.
4388 (TARGET_ZBC): Ditto.
4389 (TARGET_ZBS): Ditto.
4390 * config/riscv/riscv.opt (riscv_zb_subext): New.
4392 2021-10-25 liuhongt <hongtao.liu@intel.com>
4395 * match.pd: Simplify (_Float16) sqrtf((float) a) to .SQRT(a)
4396 when direct_internal_fn_supported_p, similar for sqrt/sqrtl.
4398 2021-10-25 Richard Biener <rguenther@suse.de>
4400 PR tree-optimization/102920
4401 * tree-ssa-sccvn.h (expressions_equal_p): Add argument
4402 controlling VN_TOP matching behavior.
4403 * tree-ssa-sccvn.c (expressions_equal_p): Likewise.
4404 (vn_phi_eq): Do not optimistically match VN_TOP.
4406 2021-10-25 konglin1 <lingling.kong@intel.com>
4408 * config/i386/sse.md (fma_<mode>_fadd_fmul): Add new
4409 define_insn_and_split.
4410 (fma_<mode>_fadd_fcmul):Likewise
4411 (fma_<complexopname>_<mode>_fma_zero):Likewise
4413 2021-10-24 John David Anglin <danglin@gcc.gnu.org>
4415 * config/pa/pa-d.c (pa_d_handle_target_float_abi): Don't check
4416 TARGET_DISABLE_FPREGS.
4417 * config/pa/pa.c (fix_range): Use MASK_SOFT_FLOAT instead of
4418 MASK_DISABLE_FPREGS.
4419 (hppa_rtx_costs): Don't check TARGET_DISABLE_FPREGS. Adjust
4420 cost of hardware integer multiplication.
4421 (pa_conditional_register_usage): Don't check TARGET_DISABLE_FPREGS.
4422 * config/pa/pa.h (INT14_OK_STRICT): Likewise.
4423 * config/pa/pa.md: Don't check TARGET_DISABLE_FPREGS. Check
4424 TARGET_SOFT_FLOAT in patterns that use xmpyu instruction.
4425 * config/pa/pa.opt (mdisable-fpregs): Change target mask to
4426 SOFT_FLOAT. Revise comment.
4427 (msoft-float): New option.
4429 2021-10-24 John David Anglin <danglin@gcc.gnu.org>
4431 * config/pa/pa.md: Don't use 'G' constraint in integer move patterns.
4433 2021-10-24 Gerald Pfeifer <gerald@pfeifer.com>
4435 * doc/install.texi (Specific): Remove obsolete details
4436 around GNU/Linux on Itanium.
4437 (Specific): Remove reference to Windows for Itanium.
4439 2021-10-23 Bernhard Reutner-Fischer <aldot@gcc.gnu.org>
4441 * config/i386/x86-tune-sched-bd.c (dispatch_group): Commentary
4444 2021-10-23 Jan Hubicka <hubicka@ucw.cz>
4446 * tree-ssa-structalias.c (compute_points_to_sets): Cleanup.
4448 2021-10-23 Roger Sayle <roger@nextmovesoftware.com>
4450 * config/i386/sse.md (<any_logic>v1ti3): New define_insn to
4451 implement V1TImode AND, IOR and XOR on TARGET_SSE2 (and above).
4452 (one_cmplv1ti2): New define expand.
4454 2021-10-22 Eric Gallager <egallager@gcc.gnu.org>
4457 * Makefile.in: Handle dvidir and install-dvi target.
4458 * configure: Regenerate.
4459 * configure.ac: Add install-dvi to target_list.
4461 2021-10-22 Gerald Pfeifer <gerald@pfeifer.com>
4463 * doc/install.texi (Binaries): Convert mingw-w64.org to https.
4466 2021-10-22 Richard Biener <rguenther@suse.de>
4468 PR tree-optimization/102893
4469 * tree-ssa-dce.c (find_obviously_necessary_stmts): Fix the
4470 test for an exit edge.
4472 2021-10-22 Aldy Hernandez <aldyh@redhat.com>
4473 Andrew MacLeod <amacleod@redhat.com>
4475 * gimple-range-path.cc (path_range_query::compute_phi_relations):
4476 Kill any global relations we may know before registering a new
4478 * value-relation.cc (path_oracle::killing_def): New.
4479 * value-relation.h (path_oracle::killing_def): New.
4481 2021-10-22 Richard Biener <rguenther@suse.de>
4484 * tree-ssa-sccvn.c (vn_phi_insert): For undefined SSA args
4486 (vn_phi_lookup): Likewise.
4488 2021-10-21 H.J. Lu <hjl.tools@gmail.com>
4491 * doc/invoke.texi: Document -fcf-protection requires i686 or
4494 2021-10-21 Eric Botcazou <ebotcazou@adacore.com>
4496 PR middle-end/102764
4497 * cfgexpand.c (expand_gimple_basic_block): Robustify latest change.
4499 2021-10-21 Jonathan Wright <jonathan.wright@arm.com>
4501 * config/aarch64/arm_neon.h (__STRUCTN): Delete function
4502 macro and all invocations.
4504 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
4506 * doc/invoke.texi (ranger-debug): Document.
4507 * flag-types.h (enum ranger_debug): New.
4508 (enum evrp_mode): Remove debug values.
4509 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Use new debug flag.
4510 * gimple-range-gori.cc (gori_compute::gori_compute): Ditto.
4511 * gimple-range.cc (gimple_ranger::gimple_ranger): Ditto.
4512 * gimple-ssa-evrp.c (hybrid_folder::choose_value): Ditto.
4513 (execute_early_vrp): Use evrp-mode directly.
4514 * params.opt (enum evrp_mode): Remove debug values.
4515 (ranger-debug): New.
4516 (ranger-logical-depth): Relocate to be in alphabetical order.
4518 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
4520 * doc/invoke.texi: (vrp1-mode, vrp2-mode): Document.
4521 * flag-types.h: (enum vrp_mode): New.
4522 * params.opt: (vrp1-mode, vrp2-mode): New.
4523 * tree-vrp.c (vrp_pass_num): New.
4524 (pass_vrp::pass_vrp): Set pass number.
4525 (pass_vrp::execute): Choose which VRP mode to execute.
4527 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
4529 * gimple-ssa-evrp.c (class rvrp_folder): Move to tree-vrp.c.
4530 (execute_early_vrp): For ranger only mode, invoke ranger_vrp.
4531 * tree-vrp.c (class rvrp_folder): Relocate here.
4532 (execute_ranger_vrp): New.
4533 * tree-vrp.h (execute_ranger_vrp): Export.
4535 2021-10-21 Martin Liska <mliska@suse.cz>
4539 * opts.c (finish_options): Process flag_var_tracking* options
4540 here as they can be adjusted by optimize attribute.
4541 Process also flag_syntax_only and flag_gtoggle.
4542 * toplev.c (process_options): Remove it here.
4543 * common.opt: Make debug_nonbind_markers_p as PerFunction
4544 attribute as it depends on optimization level.
4546 2021-10-21 Martin Jambor <mjambor@suse.cz>
4548 PR tree-optimization/102505
4549 * tree-sra.c (totally_scalarize_subtree): Check that the
4550 encountered field fits within the acces we would like to put it
4553 2021-10-21 Aldy Hernandez <aldyh@redhat.com>
4555 * tree-ssa-threadbackward.c
4556 (back_threader::maybe_register_path): Remove circular paths check.
4558 2021-10-21 Richard Biener <rguenther@suse.de>
4560 * toplev.c (process_options): Move the initial debug_hooks
4562 (toplev::main): ... before the call of the post_options
4565 2021-10-21 Richard Biener <rguenther@suse.de>
4567 PR tree-optimization/102847
4568 * tree-vect-stmts.c (vect_model_load_cost): Add the scalar
4569 load cost in the prologue for VMAT_INVARIANT.
4571 2021-10-21 Richard Biener <rguenther@suse.de>
4573 PR tree-optimization/102847
4574 * tree-vect-stmts.c (vect_model_load_cost): Explicitely
4575 handle VMAT_INVARIANT as a splat in the prologue.
4577 2021-10-21 Hongyu Wang <hongyu.wang@intel.com>
4580 * config/i386/i386.c (ix86_get_ssemov): Adjust HFmode vector
4581 move to use the same logic as HImode.
4583 2021-10-21 Richard Biener <rguenther@suse.de>
4585 * tree-vect-slp.c (vect_build_slp_tree_1): Remove
4586 superfluous gimple_call_nothrow_p check.
4588 2021-10-21 Andrew Pinski <apinski@marvell.com>
4590 * tree-cfg.c (maybe_remove_writeonly_store): Add dce_ssa_names argument.
4591 Mark the ssa-name of the rhs as one to be removed.
4592 (execute_fixup_cfg): Update call to maybe_remove_writeonly_store.
4593 Call simple_dce_from_worklist at the end to a simple dce.
4595 2021-10-21 Andrew Pinski <apinski@marvell.com>
4597 * tree-cfg.c (maybe_remove_writeonly_store): New function
4598 factored out from ...
4599 (execute_fixup_cfg): Here. Call maybe_remove_writeonly_store.
4601 2021-10-21 Andrew Pinski <apinski@marvell.com>
4603 * tree-cfg.c (execute_fixup_cfg): Remove comment
4604 about standalone pass.
4606 2021-10-21 Andrew Pinski <apinski@marvell.com>
4608 * tree-cfg.c (execute_fixup_cfg): Output when the statement
4609 is removed when it is a write only var.
4611 2021-10-21 Aldy Hernandez <aldyh@redhat.com>
4613 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
4614 Avoid threading circular paths.
4616 2021-10-20 Alex Coplan <alex.coplan@arm.com>
4618 * calls.c (initialize_argument_information): Remove some dead
4619 code, remove handling for function_arg returning const_int.
4620 * doc/tm.texi: Delete documentation for unused target hooks.
4621 * doc/tm.texi.in: Likewise.
4622 * target.def (load_bounds_for_arg): Delete.
4623 (store_bounds_for_arg): Delete.
4624 (load_returned_bounds): Delete.
4625 (store_returned_bounds): Delete.
4626 * targhooks.c (default_load_bounds_for_arg): Delete.
4627 (default_store_bounds_for_arg): Delete.
4628 (default_load_returned_bounds): Delete.
4629 (default_store_returned_bounds): Delete.
4630 * targhooks.h (default_load_bounds_for_arg): Delete.
4631 (default_store_bounds_for_arg): Delete.
4632 (default_load_returned_bounds): Delete.
4633 (default_store_returned_bounds): Delete.
4635 2021-10-20 Jonathan Wakely <jwakely@redhat.com>
4637 * doc/extend.texi (Basic Asm): Clarify that asm is not an
4639 * doc/invoke.texi (-fno-asm): Fix description for C++.
4641 2021-10-20 Jonathan Wakely <jwakely@redhat.com>
4643 * doc/install.texi: Remove link to old.html
4645 2021-10-20 Tamar Christina <tamar.christina@arm.com>
4647 * config/aarch64/aarch64-simd.md (*aarch64_cmtst_same_<mode>): New.
4649 2021-10-20 Tamar Christina <tamar.christina@arm.com>
4651 * config/aarch64/aarch64-simd.md (*aarch64_narrow_trunc<mode>): New.
4653 2021-10-20 Tamar Christina <tamar.christina@arm.com>
4655 * config/aarch64/aarch64-simd.md (aarch64_simd_ashr<mode>): Add case cmp
4657 * config/aarch64/constraints.md (D1): New.
4659 2021-10-20 Tamar Christina <tamar.christina@arm.com>
4661 * config/aarch64/aarch64-simd.md
4662 (*aarch64_<srn_op>topbits_shuffle<mode>_le): New.
4663 (*aarch64_topbits_shuffle<mode>_le): New.
4664 (*aarch64_<srn_op>topbits_shuffle<mode>_be): New.
4665 (*aarch64_topbits_shuffle<mode>_be): New.
4666 * config/aarch64/predicates.md
4667 (aarch64_simd_shift_imm_vec_exact_top): New.
4669 2021-10-20 Tamar Christina <tamar.christina@arm.com>
4671 * config/aarch64/aarch64-simd.md (*aarch64_<srn_op>shrn<mode>_vect,
4672 *aarch64_<srn_op>shrn<mode>2_vect_le,
4673 *aarch64_<srn_op>shrn<mode>2_vect_be): New.
4674 * config/aarch64/iterators.md (srn_op): New.
4676 2021-10-20 Chung-Lin Tang <cltang@codesourcery.com>
4678 * omp-low.c (omp_copy_decl_2): For !ctx, use record_vars to add new copy
4680 (scan_sharing_clauses): Place copy of OMP_CLAUSE_IN_REDUCTION decl in
4681 ctx->outer instead of ctx.
4683 2021-10-20 Martin Liska <mliska@suse.cz>
4686 2021-10-19 Martin Liska <mliska@suse.cz>
4689 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
4690 * system.h (strip_whilespaces): New function.
4692 2021-10-20 Martin Liska <mliska@suse.cz>
4695 2021-10-19 Martin Liska <mliska@suse.cz>
4698 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
4701 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
4703 * config/aarch64/aarch64.c (aarch64_case_values_threshold):
4704 Change to 8 with -Os, 11 otherwise.
4706 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
4708 * config/aarch64/aarch64.c (neoversev1_tunings):
4709 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
4710 (neoversen2_tunings): Likewise.
4712 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
4715 * config/aarch64/aarch64.md (frint_pattern): Update comment.
4716 * config/aarch64/aarch64-simd-builtins.def: Change frintn to roundeven.
4717 * config/aarch64/arm_fp16.h: Change frintn to roundeven.
4718 * config/aarch64/arm_neon.h: Likewise.
4719 * config/aarch64/iterators.md (frint_pattern): Use roundeven for FRINTN.
4721 2021-10-20 Martin Liska <mliska@suse.cz>
4723 * config/arm/arm.c (arm_unwind_emit_sequence): Do not declare
4724 already declared global variable.
4725 (arm_unwind_emit_set): Use out_file as function argument.
4726 (arm_unwind_emit): Likewise.
4727 * config/darwin.c (machopic_output_data_section_indirection): Likewise.
4728 (machopic_output_stub_indirection): Likewise.
4729 (machopic_output_indirection): Likewise.
4730 (machopic_finish): Likewise.
4731 * config/i386/i386.c (ix86_asm_output_function_label): Likewise.
4732 * config/i386/winnt.c (i386_pe_seh_unwind_emit): Likewise.
4733 * config/ia64/ia64.c (process_epilogue): Likewise.
4734 (process_cfa_adjust_cfa): Likewise.
4735 (process_cfa_register): Likewise.
4736 (process_cfa_offset): Likewise.
4737 (ia64_asm_unwind_emit): Likewise.
4738 * config/s390/s390.c (s390_asm_output_function_label): Likewise.
4740 2021-10-20 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
4742 * config/aarch64/aarch64-builtins.c
4743 (aarch64_general_gimple_fold_builtin):
4744 lower vld1 and vst1 variants of the neon builtins
4745 * config/aarch64/aarch64-protos.h:
4746 (aarch64_general_gimple_fold_builtin): Add gsi parameter.
4747 * config/aarch64/aarch64.c (aarch64_general_gimple_fold_builtin):
4750 2021-10-20 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
4752 * match.pd: Generate IFN_TRUNC.
4754 2021-10-20 Richard Biener <rguenther@suse.de>
4756 PR tree-optimization/102853
4757 * tree-data-ref.c (split_constant_offset_1): Bail out
4758 immediately if the expression traps on overflow.
4760 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
4762 * tree-ssa-threadbackward.c (back_threader::~back_threader): Remove.
4764 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
4766 * tree-ssa-threadupdate.c (back_jt_path_registry::adjust_paths_after_duplication):
4767 Remove superflous debugging message.
4768 (back_jt_path_registry::duplicate_thread_path): Same.
4770 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
4772 * tree-ssa-threadbackward.c (back_threader_registry::back_threader_registry):
4774 (back_threader_registry::register_path): Remove m_threaded_paths.
4776 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
4778 PR tree-optimization/102814
4779 * doc/invoke.texi: Document --param=max-fsm-thread-length.
4780 * params.opt: Add --param=max-fsm-thread-length.
4781 * tree-ssa-threadbackward.c
4782 (back_threader_profitability::profitable_path_p): Fail on paths
4783 longer than max-fsm-thread-length.
4785 2021-10-20 Eric Botcazou <ebotcazou@adacore.com>
4787 PR middle-end/102764
4788 * cfgexpand.c (expand_gimple_basic_block): Disregard a final debug
4789 statement to reset the current location for the outgoing edges.
4791 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
4792 Richard Biener <rguenther@suse.de>
4794 * tree-ssa-threadupdate.c (cancel_thread): Dump threading reason
4795 on the same line as the threading cancellation.
4796 (jt_path_registry::cancel_invalid_paths): Avoid rotating loops.
4797 Avoid threading through loop headers where the path remains in the
4800 2021-10-20 Siddhesh Poyarekar <siddhesh@gotplt.org>
4802 * tree-object-size.c (unknown): Make into a function. Adjust
4804 (unknown_object_size): Simplify implementation.
4806 2021-10-20 Hongtao Liu <hongtao.liu@intel.com>
4807 Kewen Lin <linkw@linux.ibm.com>
4809 * doc/sourcebuild.texi (Effective-Target Keywords): Document
4810 vect_slp_v2qi_store, vect_slp_v4qi_store, vect_slp_v8qi_store,
4811 vect_slp_v16qi_store, vect_slp_v2hi_store,
4812 vect_slp_v4hi_store, vect_slp_v2si_store, vect_slp_v4si_store.
4814 2021-10-19 Jonathan Wakely <jwakely@redhat.com>
4816 * doc/extend.texi (Basic PowerPC Built-in Functions): Fix typo.
4818 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
4822 * config/rs6000/emmintrin.h: Guard POWER8 intrinsics.
4823 * config/rs6000/pmmintrin.h: Same.
4824 * config/rs6000/smmintrin.h: Same.
4825 * config/rs6000/tmmintrin.h: Same.
4827 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
4829 * config.gcc (extra_headers): Add nmmintrin.h.
4831 2021-10-19 Richard Biener <rguenther@suse.de>
4833 * tree-vectorizer.h (vect_supportable_dr_alignment): Add
4834 misalignment parameter.
4835 * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
4836 Do not change DR_MISALIGNMENT in place, instead pass the
4837 adjusted misalignment to vect_supportable_dr_alignment.
4838 (vect_peeling_supportable): Likewise.
4839 (vect_peeling_hash_get_lowest_cost): Adjust.
4840 (vect_enhance_data_refs_alignment): Likewise.
4841 (vect_vfa_access_size): Likewise.
4842 (vect_supportable_dr_alignment): Add misalignment
4843 parameter and simplify.
4844 * tree-vect-stmts.c (get_negative_load_store_type): Adjust.
4845 (get_group_load_store_type): Likewise.
4846 (get_load_store_type): Likewise.
4848 2021-10-19 Clément Chigot <clement.chigot@atos.net>
4850 * config/rs6000/rs6000.c (rs6000_xcoff_file_end): Move
4851 __tls_get_addr reference to .text csect.
4853 2021-10-19 Martin Liska <mliska@suse.cz>
4856 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
4859 2021-10-19 Richard Biener <rguenther@suse.de>
4861 * tree-vectorizer.h (vect_get_store_cost): Adjust signature.
4862 (vect_get_load_cost): Likewise.
4863 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
4864 alignment support scheme and misalignment as arguments
4866 (vect_get_peeling_costs_all_drs): Compute that info here
4867 and note that we shouldn't need to.
4868 * tree-vect-stmts.c (vect_model_store_cost): Get
4869 alignment support scheme and misalignment as arguments.
4870 (vect_get_store_cost): Likewise.
4871 (vect_model_load_cost): Likewise.
4872 (vect_get_load_cost): Likewise.
4873 (vectorizable_store): Pass down alignment support scheme
4874 and misalignment to costing.
4875 (vectorizable_load): Likewise.
4877 2021-10-19 Richard Biener <rguenther@suse.de>
4879 * tree-vect-stmts.c (get_negative_load_store_type): Add
4880 offset output parameter and initialize it.
4881 (get_group_load_store_type): Likewise.
4882 (get_load_store_type): Likewise.
4883 (vectorizable_store): Use offset as computed by
4884 get_load_store_type.
4885 (vectorizable_load): Likewise.
4887 2021-10-19 Richard Biener <rguenther@suse.de>
4889 PR tree-optimization/102827
4890 * tree-if-conv.c (predicate_statements): Add pe parameter
4891 and use that edge to insert invariant stmts on.
4892 (combine_blocks): Pass through pe.
4893 (tree_if_conversion): Compute the edge to insert invariant
4894 stmts on and pass it along.
4896 2021-10-19 Roger Sayle <roger@nextmovesoftware.com>
4899 * config/bfin/bfin.md (addsubv2hi3, subaddv2hi3, ssaddsubv2hi3,
4900 sssubaddv2hi3): Swap the order of operators in vec_concat.
4902 2021-10-19 Xionghu Luo <luoxhu@linux.ibm.com>
4904 * config/rs6000/altivec.md (*altivec_vmrghb_internal): Delete.
4905 (altivec_vmrghb_direct): New.
4906 (*altivec_vmrghh_internal): Delete.
4907 (altivec_vmrghh_direct): New.
4908 (*altivec_vmrghw_internal): Delete.
4909 (altivec_vmrghw_direct_<mode>): New.
4910 (altivec_vmrghw_direct): Delete.
4911 (*altivec_vmrglb_internal): Delete.
4912 (altivec_vmrglb_direct): New.
4913 (*altivec_vmrglh_internal): Delete.
4914 (altivec_vmrglh_direct): New.
4915 (*altivec_vmrglw_internal): Delete.
4916 (altivec_vmrglw_direct_<mode>): New.
4917 (altivec_vmrglw_direct): Delete.
4918 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
4919 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):
4921 * config/rs6000/vsx.md (vsx_xxmrghw_<mode>): Adjust.
4922 (vsx_xxmrglw_<mode>): Adjust.
4924 2021-10-19 Aldy Hernandez <aldyh@redhat.com>
4926 * passes.def: Change threading comment before pass_ccp pass.
4928 2021-10-19 Haochen Gui <guihaoc@gcc.gnu.org>
4930 * config/rs6000/rs6000-call.c (altivec_expand_lxvr_builtin):
4931 Modify the expansion for sign extension. All extensions are done
4932 within VSX registers.
4934 2021-10-19 Richard Biener <rguenther@suse.de>
4936 * tree-vect-stmts.c (get_group_load_store_type): Add
4937 misalignment output parameter and initialize it.
4938 (get_group_load_store_type): Likewise.
4939 (vectorizable_store): Remove now redundant queries.
4940 (vectorizable_load): Likewise.
4942 2021-10-19 Richard Biener <rguenther@suse.de>
4944 * tree-vectorizer.h (vect_supportable_dr_alignment): Remove
4945 check_aligned argument.
4946 * tree-vect-data-refs.c (vect_supportable_dr_alignment):
4948 (vect_peeling_hash_insert): Add supportable_if_not_aligned
4949 argument and do not call vect_supportable_dr_alignment here.
4950 (vect_peeling_supportable): Adjust.
4951 (vect_enhance_data_refs_alignment): Compute whether the
4952 access is supported with different alignment here and
4953 pass that down to vect_peeling_hash_insert.
4954 (vect_vfa_access_size): Adjust.
4955 * tree-vect-stmts.c (vect_get_store_cost): Likewise.
4956 (vect_get_load_cost): Likewise.
4957 (get_negative_load_store_type): Likewise.
4958 (get_group_load_store_type): Likewise.
4959 (get_load_store_type): Likewise.
4961 2021-10-19 Martin Liska <mliska@suse.cz>
4964 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
4965 * system.h (strip_whilespaces): New function.
4967 2021-10-19 dianhong xu <dianhong.xu@intel.com>
4969 * config/i386/avx512fp16intrin.h:
4970 (_mm512_set1_pch): New intrinsic.
4971 * config/i386/avx512fp16vlintrin.h:
4972 (_mm256_set1_pch): New intrinsic.
4973 (_mm_set1_pch): Ditto.
4975 2021-10-18 Andrew MacLeod <amacleod@redhat.com>
4977 PR tree-optimization/102796
4978 * gimple-range.cc (gimple_ranger::range_on_edge): Process EH edges
4979 normally. Return get_tree_range for non gimple_range_ssa_p names.
4980 (gimple_ranger::range_of_stmt): Use get_tree_range for non
4981 gimple_range_ssa_p names.
4983 2021-10-18 Uroš Bizjak <ubizjak@gmail.com>
4986 * config/i386/i386.c (ix86_print_operand_address):
4987 Error out for non-address_operand asm operands.
4989 2021-10-18 Richard Biener <rguenther@suse.de>
4991 * tree-vect-data-refs.c (vect_peeling_hash_insert): Do
4992 not auto-convert dr_alignment_support to bool.
4993 (vect_peeling_supportable): Likewise.
4994 (vect_enhance_data_refs_alignment): Likewise.
4995 (vect_supportable_dr_alignment): Commonize read/write case.
4996 * tree-vect-stmts.c (vect_get_store_cost): Use
4997 dr_alignment_support, not int, for the vect_supportable_dr_alignment
4999 (vect_get_load_cost): Likewise.
5001 2021-10-18 Siddhesh Poyarekar <siddhesh@gotplt.org>
5003 * tree-object-size.c (object_sizes_execute): Consolidate LHS
5004 null check and do it early.
5006 2021-10-18 Richard Biener <rguenther@suse.de>
5008 * tree-vect-stmts.c (vectorizable_store): Use the
5009 computed alignment scheme instead of querying
5012 2021-10-18 Richard Biener <rguenther@suse.de>
5014 * tree-vect-stmts.c (vectorizable_store): Do not recompute
5015 alignment scheme already determined by get_load_store_type.
5017 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
5019 * tree-ssa-threadbackward.c (class pass_thread_jumps_full):
5020 Clone corresponding pass.
5022 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
5024 * combine.c (recog_for_combine): For an unrecognized move/set of
5025 a constant, try force_const_mem to place it in the constant pool.
5026 * cse.c (constant_pool_entries_cost, constant_pool_entries_regcost):
5027 Delete global variables (that are no longer assigned a cost value).
5028 (cse_insn): Simplify logic for deciding whether to place a folded
5029 constant in the constant pool using force_const_mem.
5030 (cse_main): Remove zero initialization of constant_pool_entries_cost
5031 and constant_pool_entries_regcost.
5032 * config/i386/i386.c (ix86_rtx_costs): Make memory accesses
5033 fractionally more expensive, when optimizing for speed.
5035 2021-10-18 Martin Liska <mliska@suse.cz>
5037 PR gcov-profile/102746
5038 PR gcov-profile/102747
5039 * gcov.c (main): Return return_code.
5040 (output_gcov_file): Mark return_code when error happens.
5041 (generate_results): Likewise.
5042 (read_graph_file): Likewise.
5043 (read_count_file): Likewise.
5045 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
5047 * config/bfin/bfin.md (define_constants): Remove UNSPEC_ONES.
5048 (define_insn "ones"): Replace UNSPEC_ONES with a truncate of
5049 a popcount, allowing compile-time evaluation/simplification.
5050 (popcountsi2, popcounthi2): New expanders using a "ones" insn.
5052 2021-10-18 Richard Biener <rguenther@suse.de>
5054 PR tree-optimization/102788
5055 * tree-vect-patterns.c (vect_init_pattern_stmt): Allow
5057 (vect_pattern_recog_1): Likewise.
5058 (vect_recog_bool_pattern): Continue matching the pattern
5059 even if we do not have a vector type for a conversion
5062 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
5064 * simplify-rtx.c (simplify_const_unary_operation) [SS_NEG, SS_ABS]:
5065 Evalute SS_NEG and SS_ABS of a constant argument.
5067 2021-10-18 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
5070 * gimple-match-head.c (try_conditional_simplification): Add case for single operand.
5071 * internal-fn.def: Add entry for COND_NEG internal function.
5072 * internal-fn.c (FOR_EACH_CODE_MAPPING): Add entry for
5073 NEGATE_EXPR, COND_NEG mapping.
5074 * optabs.def: Add entry for cond_neg_optab.
5075 * match.pd (UNCOND_UNARY, COND_UNARY): New operator lists.
5076 (vec_cond COND (foo A) B) -> (IFN_COND_FOO COND A B): New pattern.
5077 (vec_cond COND B (foo A)) -> (IFN_COND_FOO ~COND A B): Likewise.
5079 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
5081 * tree-ssa-strlen.c (class strlen_pass): Rename from
5083 (handle_builtin_stxncpy_strncat): Move to strlen_pass.
5084 (handle_assign): Same.
5085 (adjust_last_stmt): Same.
5086 (maybe_warn_overflow): Same.
5087 (handle_builtin_strlen): Same.
5088 (handle_builtin_strchr): Same.
5089 (handle_builtin_strcpy): Same.
5090 (handle_builtin_strncat): Same.
5091 (handle_builtin_stxncpy_strncat): Same.
5092 (handle_builtin_memcpy): Same.
5093 (handle_builtin_strcat): Same.
5094 (handle_alloc_call): Same.
5095 (handle_builtin_memset): Same.
5096 (handle_builtin_memcmp): Same.
5097 (get_len_or_size): Same.
5098 (strxcmp_eqz_result): Same.
5099 (handle_builtin_string_cmp): Same.
5100 (handle_pointer_plus): Same.
5101 (count_nonzero_bytes_addr): Same.
5102 (count_nonzero_bytes): Same.
5103 (handle_store): Same.
5104 (strlen_check_and_optimize_call): Same.
5105 (handle_integral_assign): Same.
5106 (check_and_optimize_stmt): Same.
5107 (printf_strlen_execute): Rename strlen_dom_walker to strlen_pass.
5109 2021-10-18 Richard Biener <rguenther@suse.de>
5111 PR tree-optimization/102798
5112 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
5113 Only copy points-to info to newly generated SSA names.
5115 2021-10-18 Martin Liska <mliska@suse.cz>
5117 * dbgcnt.c (dbg_cnt_process_opt): Remove unused but set variable.
5118 * gcov.c (get_cycles_count): Likewise.
5119 * lto-compress.c (lto_compression_zlib): Likewise.
5120 (lto_uncompression_zlib): Likewise.
5121 * targhooks.c (default_pch_valid_p): Likewise.
5123 2021-10-17 Aldy Hernandez <aldyh@redhat.com>
5125 * tree-pass.h (make_pass_thread_jumps_full): New.
5126 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Inline.
5127 (try_thread_blocks): Add resolve and speed arguments.
5128 (pass_thread_jumps::execute): Inline.
5129 (do_early_thread_jumps): New.
5130 (do_thread_jumps): New.
5131 (make_pass_thread_jumps): Move.
5132 (pass_early_thread_jumps::gate): Inline.
5133 (pass_early_thread_jumps::execute): Inline.
5134 (class pass_thread_jumps_full): New.
5136 2021-10-16 Piotr Kubaj <pkubaj@FreeBSD.org>
5138 * configure.ac: Treat powerpc64*-*-freebsd* the same as
5139 powerpc64-*-freebsd*.
5140 * configure: Regenerate.
5142 2021-10-16 H.J. Lu <hjl.tools@gmail.com>
5144 * value-query.cc (get_ssa_name_ptr_info_nonnull): Change
5145 set_ptr_nonull to set_ptr_nonnull in comments.
5147 2021-10-16 Jan Hubicka <hubicka@ucw.cz>
5149 PR tree-optimization/102720
5150 * tree-ssa-structalias.c (compute_points_to_sets): Fix producing
5151 of call used and clobbered sets.
5153 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
5155 * config/darwin.h (LINK_COMMAND_SPEC_A): Update 'r' handling to
5156 skip gomp and itm when r or nodefaultlibs is given.
5157 (DSYMUTIL_SPEC): Do not call dsymutil for '-r' link lines.
5158 Update ordering of exclusions, remove duplicate 'v' addition
5159 (collect2 will add this from the main command line).
5161 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
5163 * config/darwin-driver.c (darwin_driver_init): Revise comments, handle
5164 filelist and framework options in specs instead of code.
5165 * config/darwin.h (SUBTARGET_DRIVER_SELF_SPECS): Update to handle link
5166 specs that are really driver ones.
5167 (DARWIN_CC1_SPEC): Likewise.
5168 (CPP_SPEC): Likewise.
5169 (SYSROOT_SPEC): Append space.
5170 (LINK_SYSROOT_SPEC): Remove most driver link specs.
5171 (STANDARD_STARTFILE_PREFIX_2): Update link-related specs.
5172 (STARTFILE_SPEC): Likewise.
5173 (ASM_MMACOSX_VERSION_MIN_SPEC): Fix line wrap.
5174 (ASM_SPEC): Update driver-related specs.
5175 (ASM_FINAL_SPEC): Likewise.
5176 * config/darwin.opt: Remove now unused option aliases.
5177 * config/i386/darwin.h (EXTRA_ASM_OPTS): Ensure space after opt.
5178 (ASM_SPEC): Update driver-related specs.
5180 2021-10-15 Roger Sayle <roger@nextmovesoftware.com>
5182 * config/i386/i386.c (ix86_hardreg_mov_ok): For vector modes,
5183 allow standard_sse_constant_p immediate constants.
5185 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
5187 * config.gcc: Add tm-dwarf2.h to tm_d-file.
5189 2021-10-15 Andrew MacLeod <amacleod@redhat.com>
5191 * gimple-range-fold.h (gimple_range_ssa_p): Don't process names
5192 that occur in abnormal phis.
5193 * gimple-range.cc (gimple_ranger::range_on_edge): Return false for
5194 abnormal and EH edges.
5195 * gimple-ssa-evrp.c (rvrp_folder::value_of_expr): Ditto.
5196 (rvrp_folder::value_on_edge): Ditto.
5197 (rvrp_folder::value_of_stmt): Ditto.
5198 (hybrid_folder::value_of_expr): Ditto for ranger queries.
5199 (hybrid_folder::value_on_edge): Ditto.
5200 (hybrid_folder::value_of_stmt): Ditto.
5201 * value-query.cc (gimple_range_global): Always return a range if
5202 the type is supported.
5204 2021-10-15 John David Anglin <danglin@gcc.gnu.org>
5206 * config/pa/pa.md: Consistently use "rG" constraint for copy
5207 instruction in move patterns.
5209 2021-10-15 Andrew Stubbs <ams@codesourcery.com>
5211 * config/gcn/gcn-hsa.h (S_FIJI): Set unconditionally.
5214 * config/gcn/gcn.c: Hard code SRAM ECC settings for old architectures.
5215 * config/gcn/mkoffload.c (ELFABIVERSION_AMDGPU_HSA): Rename to ...
5216 (ELFABIVERSION_AMDGPU_HSA_V3): ... this.
5217 (ELFABIVERSION_AMDGPU_HSA_V4): New.
5218 (SET_SRAM_ECC_UNSUPPORTED): New.
5219 (copy_early_debug_info): Create elf flags to match the other objects.
5220 (main): Just let the attribute flags pass through.
5222 2021-10-15 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
5224 * tree-loop-distribution.c (reduction_var_overflows_first):
5225 Pass the type of reduction_var as first argument as it is also
5226 done for the load type.
5227 (loop_distribution::transform_reduction_loop): Add missing
5228 TREE_TYPE while determining precission of reduction_var.
5230 2021-10-15 Richard Biener <rguenther@suse.de>
5232 * defaults.h (PREFERRED_DEBUGGING_TYPE): Choose DWARF2_DEBUG
5234 * toplev.c (process_options): Warn when STABS debugging is
5235 enabled but not the preferred format.
5236 * config/pa/som.h (PREFERRED_DEBUGGING_TYPE): Define to
5238 * config/pdp11/pdp11.h (PREFERRED_DEBUGGING_TYPE): Likewise.
5240 2021-10-15 Richard Biener <rguenther@suse.de>
5243 * tree-inline.c (copy_bb): Avoid underflowing nargs.
5245 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
5247 * config/i386/i386-expand.c (ix86_expand_vec_perm): Convert
5248 HFmode input operand to HImode.
5249 (ix86_vectorize_vec_perm_const): Likewise.
5250 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1_hf):
5252 (*avx512f_permvar_truncv8siv8hi_1_hf):
5255 2021-10-15 Richard Biener <rguenther@suse.de>
5257 PR middle-end/102682
5258 * expmed.c (store_bit_field_1): Ensure a LHS subreg would
5259 not create a paradoxical subreg.
5261 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
5263 * config/i386/i386-expand.c (ix86_expand_vector_init):
5264 For half_vector concat for HFmode, handle them like HImode.
5266 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
5268 * tree-ssa-threadbackward.c (class back_threader): Add m_resolve.
5269 (back_threader::back_threader): Same.
5270 (back_threader::resolve_phi): Try to solve without looking back if
5272 (back_threader::find_paths_to_names): Same.
5273 (try_thread_blocks): Pass resolve argument to back threader.
5274 (pass_early_thread_jumps::execute): Same.
5276 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
5278 * doc/invoke.texi: Remove max-fsm-thread-length,
5279 max-fsm-thread-paths, and fsm-maximum-phi-arguments.
5281 * tree-ssa-threadbackward.c (back_threader::back_threader): Remove
5283 (back_threader_registry::back_threader_registry): Same.
5284 (back_threader_profitability::profitable_path_p): Remove
5285 param_max_fsm_thread-length.
5286 (back_threader_registry::register_path): Remove
5287 m_max_allowable_paths.
5289 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
5291 * tree-ssa-threadbackward.c (class back_threader): Make m_imports
5293 (back_threader::~back_threader): Do not release m_path.
5295 2021-10-14 Andrew MacLeod <amacleod@redhat.com>
5297 PR tree-optimization/102738
5298 * vr-values.c (simplify_using_ranges::simplify): Handle RSHIFT_EXPR.
5300 2021-10-14 Kwok Cheung Yeung <kcy@codesourcery.com>
5302 * omp-general.c (omp_check_context_selector): Move from c-omp.c.
5303 (omp_mark_declare_variant): Move from c-omp.c.
5304 (omp_context_name_list_prop): Update for Fortran strings.
5305 * omp-general.h (omp_check_context_selector): New prototype.
5306 (omp_mark_declare_variant): New prototype.
5308 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
5310 * config/rs6000/rs6000.c (rs6000_density_test): Move early
5311 exit test further up the function.
5313 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
5315 * config/arm/arm.c (arm_add_stmt_cost): Delete.
5316 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
5318 2021-10-14 Martin Jambor <mjambor@suse.cz>
5320 * doc/invoke.texi (Optimize Options): Add entry for
5321 ipa-cp-recursive-freq-factor.
5323 2021-10-14 Tamar Christina <tamar.christina@arm.com>
5325 * match.pd: New rule.
5327 2021-10-14 Jan Hubicka <hubicka@ucw.cz>
5330 * ipa-modref-tree.h (modref_access_node::update2):
5331 Also check that parm_offset is unchanged.
5332 (modref_ref_node::insert_access): Fix updating of
5335 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
5337 * tree-ssa-threadbackward.c (back_threader::resolve_phi): Add
5340 2021-10-14 Richard Biener <rguenther@suse.de>
5342 PR tree-optimization/102659
5343 * tree-if-conv.c (if_convertible_gimple_assign_stmt_p): Also
5344 rewrite pointer typed undefined overflow operations.
5345 (predicate_statements): Likewise. Make sure to emit invariant
5346 conversions in the preheader.
5347 * tree-vectorizer.c (vect_loop_vectorized_call): Look through
5348 non-empty preheaders.
5349 * tree-data-ref.c (dr_analyze_indices): Strip useless
5350 conversions to the MEM_REF base type.
5352 2021-10-14 Martin Liska <mliska@suse.cz>
5354 * common.opt: Stop using AUTODETECT_VALUE
5355 and use EnabledBy where possible.
5356 * opts.c: Enable OPT_fvar_tracking with optimize >= 1.
5357 * toplev.c (AUTODETECT_VALUE): Remove macro.
5358 (process_options): Simplify by using EnabledBy and
5359 OPT_fvar_tracking. Use OPTION_SET_P macro instead of
5362 2021-10-14 Jonathan Wright <jonathan.wright@arm.com>
5364 * config/aarch64/arm_neon.h (vld1_s8_x3): Use signed type for
5366 (vld1_s32_x3): Likewise.
5368 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
5370 PR tree-optimization/102736
5371 PR tree-optimization/102736
5372 * gimple-range-path.cc (path_range_query::range_on_path_entry):
5373 Assert that the requested range is defined outside the path.
5374 (path_range_query::ssa_range_in_phi): Do not call
5375 range_on_path_entry for SSA names that are defined within the
5378 2021-10-14 Iain Sandoe <iain@sandoe.co.uk>
5380 * config/darwin-driver.c (darwin_find_version_from_kernel):
5381 Quote internal identifiers and avoid contractions in
5383 (darwin_default_min_version): Likewise.
5384 (darwin_driver_init): Likewise.
5386 2021-10-14 Martin Jambor <mjambor@suse.cz>
5388 * params.opt (ipa-cp-recursive-freq-factor): New.
5389 * ipa-cp.c (ipcp_value): Switch to inline initialization. New members
5390 scc_no, self_recursion_generated_level, same_scc and
5391 self_recursion_generated_p.
5392 (ipcp_lattice::add_value): Replaced parameter unlimited with
5393 same_lat_gen_level, usit it determine limit of values and store it to
5395 (ipcp_lattice<valtype>::print): Dump the new fileds.
5396 (allocate_and_init_ipcp_value): Take same_lat_gen_level as a new
5397 parameter and store it to the new value.
5398 (self_recursively_generated_p): Removed.
5399 (propagate_vals_across_arith_jfunc): Use self_recursion_generated_p
5400 instead of self_recursively_generated_p, store self generation level
5402 (value_topo_info<valtype>::add_val): Set scc_no.
5403 (value_topo_info<valtype>::propagate_effects): Multiply frequencies of
5404 recursively feeding values and self generated values by appropriate
5407 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
5409 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Remove
5410 redundant test for flag_vect_cost_model.
5412 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
5414 * bitmap.c (debug): New overloaded function for auto_bitmaps.
5415 * bitmap.h (debug): Same.
5417 2021-10-14 Hongyu Wang <hongyu.wang@intel.com>
5419 * config/i386/avx512fp16intrin.h (_mm512_mask_fcmadd_pch):
5420 Adjust builtin call.
5421 (_mm512_mask3_fcmadd_pch): Likewise.
5422 (_mm512_mask_fmadd_pch): Likewise
5423 (_mm512_mask3_fmadd_pch): Likewise
5424 (_mm512_mask_fcmadd_round_pch): Likewise
5425 (_mm512_mask3_fcmadd_round_pch): Likewise
5426 (_mm512_mask_fmadd_round_pch): Likewise
5427 (_mm512_mask3_fmadd_round_pch): Likewise
5428 (_mm_mask_fcmadd_sch): Likewise
5429 (_mm_mask3_fcmadd_sch): Likewise
5430 (_mm_mask_fmadd_sch): Likewise
5431 (_mm_mask3_fmadd_sch): Likewise
5432 (_mm_mask_fcmadd_round_sch): Likewise
5433 (_mm_mask3_fcmadd_round_sch): Likewise
5434 (_mm_mask_fmadd_round_sch): Likewise
5435 (_mm_mask3_fmadd_round_sch): Likewise
5436 (_mm_fcmadd_round_sch): Likewise
5437 * config/i386/avx512fp16vlintrin.h (_mm_mask_fmadd_pch):
5438 Adjust builtin call.
5439 (_mm_mask3_fmadd_pch): Likewise
5440 (_mm256_mask_fmadd_pch): Likewise
5441 (_mm256_mask3_fmadd_pch): Likewise
5442 (_mm_mask_fcmadd_pch): Likewise
5443 (_mm_mask3_fcmadd_pch): Likewise
5444 (_mm256_mask_fcmadd_pch): Likewise
5445 (_mm256_mask3_fcmadd_pch): Likewise
5446 * config/i386/i386-builtin.def: Add mask3 builtin for complex
5447 fma, and adjust mask_builtin to corresponding expander.
5448 * config/i386/i386-expand.c (ix86_expand_round_builtin):
5449 Skip eraseing embedded rounding for expanders that emits
5451 * config/i386/sse.md (complexmove): New mode_attr.
5452 (<avx512>_fmaddc_<mode>_mask1<round_expand_name>): New expander.
5453 (<avx512>_fcmaddc_<mode>_mask1<round_expand_name>): Likewise.
5454 (avx512fp16_fmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
5455 (avx512fp16_fcmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
5456 (avx512fp16_fcmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
5457 (avx512fp16_fmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
5458 * config/i386/subst.md (round_embedded_complex): New subst.
5460 2021-10-14 John David Anglin <danglin@gcc.gnu.org>
5462 * config/pa/pa.md (cbranchsf4): Disable if TARGET_SOFT_FLOAT.
5463 (cbranchdf4): Likewise.
5464 Add missing move patterns for TARGET_SOFT_FLOAT.
5466 2021-10-13 Roger Sayle <roger@nextmovesoftware.com>
5468 * config/i386/i386-expand.c (ix86_expand_vector_move): Use a
5469 pseudo intermediate when moving a SUBREG into a hard register,
5470 by checking ix86_hardreg_mov_ok.
5471 (ix86_expand_vector_extract): Store zero-extended SImode
5472 intermediate in a pseudo, then set target using a SUBREG_PROMOTED
5474 * config/i386/sse.md (mov<VMOVE>_internal): Prevent CSE creating
5475 complex (SUBREG) sets of (vector) hard registers before reload, by
5476 checking ix86_hardreg_mov_ok.
5478 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
5480 * ctfc.h (enum ctf_dtu_d_union_enum): Remove redundant comma.
5482 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
5484 * dwarf2ctf.c (gen_ctf_array_type): Fix typo in comment.
5486 2021-10-13 Martin Sebor <msebor@redhat.com>
5488 PR middle-end/102630
5489 * pointer-query.cc (compute_objsize_r): Handle named address spaces.
5491 2021-10-13 Iain Sandoe <iain@sandoe.co.uk>
5493 * collect2.c (is_lto_object_file): Release simple-object
5494 resources, close files.
5496 2021-10-13 Andre Vieira <andre.simoesdiasvieira@arm.com>
5498 * config/arm/arm.c (thumb2_legitimate_address_p): Use VALID_MVE_MODE
5499 when checking mve addressing modes.
5500 (mve_vector_mem_operand): Fix the way we handle pre, post and offset
5502 (arm_print_operand): Fix printing of POST_ and PRE_MODIFY.
5503 * config/arm/mve.md: Use mve_memory_operand predicate everywhere where
5504 there is a single Ux constraint.
5506 2021-10-13 John David Anglin <danglin@gcc.gnu.org>
5508 * config/pa/pa.md (muldi3): Add support for inlining 64-bit
5509 multiplication on 32-bit PA 1.1 and 2.0 targets.
5511 2021-10-13 Martin Liska <mliska@suse.cz>
5513 PR gcov-profile/90364
5514 * coverage.c (build_info): Emit checksum to the global variable.
5515 (build_info_type): Add new field for checksum.
5516 (coverage_obj_finish): Pass object_checksum.
5517 (coverage_init): Use 0 as checksum for .gcno files.
5518 * gcov-dump.c (dump_gcov_file): Dump also new checksum field.
5519 * gcov.c (read_graph_file): Read also checksum.
5520 * doc/invoke.texi: Document the behaviour change.
5522 2021-10-13 Richard Biener <rguenther@suse.de>
5524 * gimple-iterator.h (gsi_iterator_update): Add GSI_LAST_NEW_STMT,
5525 start at integer value 2.
5526 * gimple-iterator.c (gsi_insert_seq_nodes_before): Update
5527 the iterator for GSI_LAST_NEW_STMT.
5528 (gsi_insert_seq_nodes_after): Likewise.
5529 * tree-if-conv.c (predicate_statements): Use GSI_LAST_NEW_STMT.
5530 * tree-ssa.c (execute_update_addresses_taken): Correct bogus
5531 arguments to gsi_replace.
5533 2021-10-13 Martin Liska <mliska@suse.cz>
5536 * common.opt: Use EnabledBy instead of detection in
5537 finish_options and process_options.
5538 * opts.c (finish_options): Remove handling of
5539 x_flag_unroll_all_loops.
5540 * toplev.c (process_options): Likewise for flag_web and
5541 flag_rename_registers.
5543 2021-10-13 Richard Biener <rguenther@suse.de>
5545 PR tree-optimization/102659
5546 * tree-if-conv.c (need_to_rewrite_undefined): New flag.
5547 (if_convertible_gimple_assign_stmt_p): Mark the loop for
5548 rewrite when stmts with undefined behavior on integer
5550 (combine_blocks): Predicate also when we need to rewrite stmts.
5551 (predicate_statements): Rewrite affected stmts to something
5552 with well-defined behavior on overflow.
5553 (tree_if_conversion): Initialize need_to_rewrite_undefined.
5555 2021-10-13 Richard Biener <rguenther@suse.de>
5558 * ipa-sra.c (ptr_parm_has_nonarg_uses): Fix volatileness
5561 2021-10-13 Jose E. Marchesi <jose.marchesi@oracle.com>
5563 * dwarf2ctf.c: Fix typo in comment.
5565 2021-10-12 Bill Schmidt <wschmidt@linux.ibm.com>
5568 * config/rs6000/altivec.h (vec_cpsgn): Swap operand order.
5569 * config/rs6000/rs6000-overload.def (VEC_COPYSIGN): Use SKIP to
5570 avoid generating an automatic #define of vec_cpsgn. Use the
5571 correct built-in for V4SFmode that doesn't depend on VSX.
5573 2021-10-12 Uroš Bizjak <ubizjak@gmail.com>
5577 * config/i386/i386.md (*add<mode>_1_slp): Rewrite as
5578 define_insn_and_split pattern. Add alternative 1 and split it
5579 post reload to insert operand 1 into the low part of operand 0.
5580 (*sub<mode>_1_slp): Ditto.
5581 (*and<mode>_1_slp): Ditto.
5582 (*<any_or:code><mode>_1_slp): Ditto.
5583 (*ashl<mode>3_1_slp): Ditto.
5584 (*<any_shiftrt:insn><mode>3_1_slp): Ditto.
5585 (*<any_rotate:insn><mode>3_1_slp): Ditto.
5586 (*neg<mode>_1_slp): New insn_and_split pattern.
5587 (*one_cmpl<mode>_1_slp): Ditto.
5589 2021-10-12 David Edelsohn <dje.gcc@gmail.com>
5591 * doc/install.texi: Update MinGW and mingw-64 Binaries
5594 2021-10-12 Daniel Le Duc Khoi Nguyen <greenrecyclebin@gmail.com>
5596 * doc/extend.texi (Common Variable Attributes): Fix typos in
5597 alloc_size documentation.
5599 2021-10-12 Richard Biener <rguenther@suse.de>
5601 PR tree-optimization/102696
5602 * tree-vect-slp.c (vect_build_slp_tree_2): Properly mark
5603 the tree fatally failed when we reject a BIT_FIELD_REF.
5605 2021-10-12 Richard Biener <rguenther@suse.de>
5607 PR tree-optimization/102572
5608 * tree-vect-stmts.c (vect_build_gather_load_calls): When
5609 gathering the vectorized defs for the mask pass in the
5610 desired mask vector type so invariants will be handled
5613 2021-10-12 Tamar Christina <tamar.christina@arm.com>
5615 * config/aarch64/aarch64-sve.md (*fcm<cmp_op><mode>_bic_combine,
5616 *fcm<cmp_op><mode>_nor_combine, *fcmuo<mode>_bic_combine,
5617 *fcmuo<mode>_nor_combine): New.
5619 2021-10-12 Eric Botcazou <ebotcazou@adacore.com>
5622 * config/sparc/sparc-modes.def (OI): New integer mode.
5624 2021-10-12 Jakub Jelinek <jakub@redhat.com>
5626 * gimple-fold.h (clear_padding_type_may_have_padding_p): Declare.
5627 * gimple-fold.c (clear_padding_type_may_have_padding_p): No longer
5630 2021-10-12 Jakub Jelinek <jakub@redhat.com>
5632 * tree-vectorizer.h (loop_cost_model): New function.
5633 (unlimited_cost_model): Use it.
5634 * tree-vect-loop.c (vect_analyze_loop_costing): Use loop_cost_model
5635 call instead of flag_vect_cost_model.
5636 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
5637 (vect_prune_runtime_alias_test_list): Likewise. Also use it instead
5638 of flag_simd_cost_model.
5640 2021-10-12 liuhongt <hongtao.liu@intel.com>
5643 * config/i386/i386-expand.c (emit_reduc_half): Handle
5645 * config/i386/mmx.md (reduc_<code>_scal_v4qi): New expander.
5646 (reduc_plus_scal_v4qi): Ditto.
5648 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
5650 * config/rs6000/smmintrin.h (_mm_cmpeq_epi64, _mm_cmpgt_epi64,
5651 _mm_mullo_epi32, _mm_mul_epi32, _mm_packus_epi32): New.
5652 * config/rs6000/nmmintrin.h: Copy from i386, tweak to suit.
5654 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
5656 * config/rs6000/smmintrin.h (_mm_cvtepi8_epi16, _mm_cvtepi8_epi32,
5657 _mm_cvtepi8_epi64, _mm_cvtepi16_epi32, _mm_cvtepi16_epi64,
5658 _mm_cvtepi32_epi64, _mm_cvtepu8_epi16, _mm_cvtepu8_epi32,
5659 _mm_cvtepu8_epi64, _mm_cvtepu16_epi32, _mm_cvtepu16_epi64,
5660 _mm_cvtepu32_epi64): New.
5662 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
5664 * config/rs6000/smmintrin.h (_mm_test_all_zeros,
5665 _mm_test_all_ones, _mm_test_mix_ones_zeros): Rewrite as macro.
5667 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
5669 * config/rs6000/smmintrin.h (_mm_min_epi8, _mm_min_epu16,
5670 _mm_min_epi32, _mm_min_epu32, _mm_max_epi8, _mm_max_epu16,
5671 _mm_max_epi32, _mm_max_epu32): New.
5673 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
5675 * ipa-modref-tree.h (struct modref_access_node): Revert
5677 (struct modref_ref_node): Likewise.
5679 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
5681 * ipa-modref-tree.h (modref_tree::global_access_p): New member
5684 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
5685 ignore_stores_eaf_flags): Move to ipa-modref.h
5686 (remove_useless_eaf_flags): Remove early exit on NOCLOBBER.
5687 (modref_summary::global_memory_read_p): New member function.
5688 (modref_summary::global_memory_written_p): New member function.
5689 * ipa-modref.h (modref_summary::global_memory_read_p,
5690 modref_summary::global_memory_written_p): Declare.
5691 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
5692 ignore_stores_eaf_flags): move here.
5693 * tree-ssa-structalias.c: Include ipa-modref-tree.h, ipa-modref.h
5695 (handle_rhs_call): Rewrite.
5696 (handle_call_arg): New function.
5697 (determine_global_memory_access): New function.
5698 (handle_const_call): Remove
5699 (handle_pure_call): Remove
5700 (find_func_aliases_for_call): Update use of handle_rhs_call.
5701 (compute_points_to_sets): Handle global memory acccesses
5704 2021-10-11 Diane Meirowitz <diane.meirowitz@oracle.com>
5706 * doc/invoke.texi: Add link to UndefinedBehaviorSanitizer
5707 documentation, mention UBSAN_OPTIONS, similar to what is done
5708 for AddressSanitizer.
5710 2021-10-11 Richard Biener <rguenther@suse.de>
5712 PR middle-end/102683
5713 * internal-fn.c (expand_DEFERRED_INIT): Check for mode
5714 availability before building an integer type for storage
5717 2021-10-11 Richard Biener <rguenther@suse.de>
5719 PR middle-end/101480
5720 * gimple.c (gimple_call_fnspec): Do not mark operator new/delete
5723 2021-10-11 Martin Liska <mliska@suse.cz>
5725 * common.opt: Remove Init(2) for some options.
5726 * toplev.c (process_options): Do not use AUTODETECT_VALUE, but
5727 use rather OPTION_SET_P.
5729 2021-10-11 Martin Liska <mliska@suse.cz>
5731 * common.opt: Remove usage of IRA_REGION_AUTODETECT.
5732 * flag-types.h (enum ira_region): Likewise.
5733 * toplev.c (process_options): Use OPTION_SET_P instead of
5734 IRA_REGION_AUTODETECT.
5736 2021-10-11 Jakub Jelinek <jakub@redhat.com>
5738 * omp-low.c (omp_runtime_api_call): Handle omp_get_max_teams,
5739 omp_[sg]et_teams_thread_limit and omp_set_num_teams.
5741 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
5743 * config/s390/s390-protos.h (s390_rawmemchr): Add prototype.
5744 * config/s390/s390.c (s390_rawmemchr): New function.
5745 * config/s390/s390.md (rawmemchr<SINT:mode>): New expander.
5746 * config/s390/vector.md (@vec_vfees<mode>): Basically a copy of
5747 the pattern vfees<mode> from vx-builtins.md.
5748 * config/s390/vx-builtins.md (*vfees<mode>): Remove.
5750 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
5752 * builtins.c (get_memory_rtx): Change to external linkage.
5753 * builtins.h (get_memory_rtx): Add function prototype.
5754 * doc/md.texi (rawmemchr<mode>): Document.
5755 * internal-fn.c (expand_RAWMEMCHR): Define.
5756 * internal-fn.def (RAWMEMCHR): Add.
5757 * optabs.def (rawmemchr_optab): Add.
5758 * tree-loop-distribution.c (find_single_drs): Change return code
5759 behaviour by also returning true if no single store was found
5761 (loop_distribution::classify_partition): Respect the new return
5762 code behaviour of function find_single_drs.
5763 (loop_distribution::execute): Call new function
5764 transform_reduction_loop in order to replace rawmemchr or strlen
5765 like loops by calls into builtins.
5766 (generate_reduction_builtin_1): New function.
5767 (generate_rawmemchr_builtin): New function.
5768 (generate_strlen_builtin_1): New function.
5769 (generate_strlen_builtin): New function.
5770 (generate_strlen_builtin_using_rawmemchr): New function.
5771 (reduction_var_overflows_first): New function.
5772 (determine_reduction_stmt_1): New function.
5773 (determine_reduction_stmt): New function.
5774 (loop_distribution::transform_reduction_loop): New function.
5776 2021-10-11 Martin Liska <mliska@suse.cz>
5778 * tree.c (cl_option_hasher::hash): Use cl_optimization_hash
5779 and remove legacy hashing code.
5781 2021-10-11 Kito Cheng <kito.cheng@sifive.com>
5784 * builtins.c (maybe_emit_call_builtin___clear_cache): Allow
5785 CONST_INT for BEGIN and END, and use gcc_assert rather than
5788 2021-10-10 Jakub Jelinek <jakub@redhat.com>
5791 * var-tracking.c (add_stores): For cselib_sp_derived_value_p values
5792 use MO_VAL_SET if loc is not sp.
5794 2021-10-10 Andrew Pinski <apinski@marvell.com>
5796 PR tree-optimization/102622
5797 * match.pd: Swap the order of a?pow2cst:0 and a?-1:0 transformations.
5798 Swap the order of a?0:pow2cst and a?0:-1 transformations.
5800 2021-10-09 liuhongt <hongtao.liu@intel.com>
5803 * config/i386/i386-expand.c (ix86_valid_mask_cmp_mode): Handle
5805 (ix86_use_mask_cmp_p): Ditto.
5806 (ix86_expand_sse_movcc): Ditto.
5807 * config/i386/i386.md (setcc_hf_mask): New define_insn.
5808 (movhf_mask): Ditto.
5809 (UNSPEC_MOVCC_MASK): New unspec.
5810 * config/i386/sse.md (UNSPEC_PCMP): Move to i386.md.
5812 2021-10-08 Vladimir N. Makarov <vmakarov@redhat.com>
5814 PR rtl-optimization/102627
5815 * lra-constraints.c (split_reg): Use at least natural mode of hard reg.
5817 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
5819 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Grow
5822 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
5824 * value-range.cc (irange::debug): New.
5825 * value-range.h (irange::debug): New.
5827 2021-10-08 Richard Sandiford <richard.sandiford@arm.com>
5829 PR tree-optimization/102385
5830 * predict.h (change_edge_frequency): Declare.
5831 * predict.c (change_edge_frequency): New function.
5832 * tree-ssa-loop-manip.h (tree_transform_and_unroll_loop): Remove
5834 (tree_unroll_loop): Likewise.
5835 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Update accordingly.
5836 * tree-predcom.c (pcom_worker::tree_predictive_commoning_loop):
5838 * tree-ssa-loop-prefetch.c (loop_prefetch_arrays): Likewise.
5839 * tree-ssa-loop-manip.c (tree_unroll_loop): Likewise.
5840 (tree_transform_and_unroll_loop): Likewise. Use single_dom_exit
5841 to retrieve the exit edges. Make all the old profile update code
5842 conditional on !single_loop_p -- the case it was written for --
5843 and use a different approach for the single-loop case.
5845 2021-10-08 Martin Liska <mliska@suse.cz>
5847 * config/alpha/alpha.c (alpha_option_override): Use new macro
5849 * config/arc/arc.c (arc_override_options): Likewise.
5850 * config/arm/arm.c (arm_option_override): Likewise.
5851 * config/bfin/bfin.c (bfin_load_pic_reg): Likewise.
5852 * config/c6x/c6x.c (c6x_option_override): Likewise.
5853 * config/csky/csky.c: Likewise.
5854 * config/darwin.c (darwin_override_options): Likewise.
5855 * config/frv/frv.c (frv_option_override): Likewise.
5856 * config/i386/djgpp.h: Likewise.
5857 * config/i386/i386.c (ix86_stack_protect_guard): Likewise.
5858 (ix86_max_noce_ifcvt_seq_cost): Likewise.
5859 * config/ia64/ia64.c (ia64_option_override): Likewise.
5860 (ia64_override_options_after_change): Likewise.
5861 * config/m32c/m32c.c (m32c_option_override): Likewise.
5862 * config/m32r/m32r.c (m32r_init): Likewise.
5863 * config/m68k/m68k.c (m68k_option_override): Likewise.
5864 * config/microblaze/microblaze.c (microblaze_option_override): Likewise.
5865 * config/mips/mips.c (mips_option_override): Likewise.
5866 * config/nios2/nios2.c (nios2_option_override): Likewise.
5867 * config/nvptx/nvptx.c (nvptx_option_override): Likewise.
5868 * config/pa/pa.c (pa_option_override): Likewise.
5869 * config/riscv/riscv.c (riscv_option_override): Likewise.
5870 * config/rs6000/aix71.h: Likewise.
5871 * config/rs6000/aix72.h: Likewise.
5872 * config/rs6000/aix73.h: Likewise.
5873 * config/rs6000/rs6000.c (darwin_rs6000_override_options): Likewise.
5874 (rs6000_override_options_after_change): Likewise.
5875 (rs6000_linux64_override_options): Likewise.
5876 (glibc_supports_ieee_128bit): Likewise.
5877 (rs6000_option_override_internal): Likewise.
5878 (rs6000_file_start): Likewise.
5879 (rs6000_darwin_file_start): Likewise.
5880 * config/rs6000/rtems.h: Likewise.
5881 * config/rs6000/sysv4.h: Likewise.
5882 * config/rs6000/vxworks.h (SUB3TARGET_OVERRIDE_OPTIONS): Likewise.
5883 * config/s390/s390.c (s390_option_override): Likewise.
5884 * config/sh/linux.h: Likewise.
5885 * config/sh/netbsd-elf.h (while): Likewise.
5886 * config/sh/sh.c (sh_option_override): Likewise.
5887 * config/sol2.c (solaris_override_options): Likewise.
5888 * config/sparc/sparc.c (sparc_option_override): Likewise.
5889 * config/tilegx/tilegx.c (tilegx_option_override): Likewise.
5890 * config/visium/visium.c (visium_option_override): Likewise.
5891 * config/vxworks.c (vxworks_override_options): Likewise.
5892 * lto-opts.c (lto_write_options): Likewise.
5893 * omp-expand.c (expand_omp_simd): Likewise.
5894 * omp-general.c (omp_max_vf): Likewise.
5895 * omp-offload.c (oacc_xform_loop): Likewise.
5896 * opts.h (OPTION_SET_P): Likewise.
5897 * targhooks.c (default_max_noce_ifcvt_seq_cost): Likewise.
5898 * toplev.c (process_options): Likewise.
5899 * tree-predcom.c: Likewise.
5900 * tree-sra.c (analyze_all_variable_accesses): Likewise.
5902 2021-10-08 liuhongt <hongtao.liu@intel.com>
5905 * config/i386/i386.c (ix86_optab_supported_p):
5906 Return true for HFmode.
5907 * match.pd: Simplify (_Float16) ceil ((double) x) to
5908 __builtin_ceilf16 (a) when a is _Float16 type and
5909 direct_internal_fn_supported_p.
5911 2021-10-08 liuhongt <hongtao.liu@intel.com>
5914 * config/i386/i386-expand.c (emit_reduc_half): Hanlde V4HImode.
5915 * config/i386/mmx.md (reduc_plus_scal_v4hi): New.
5916 (reduc_<code>_scal_v4hi): New.
5918 2021-10-08 liuhongt <hongtao.liu@intel.com>
5920 * common.opt (ftree-vectorize): Add Var(flag_tree_vectorize).
5921 * doc/invoke.texi (Options That Control Optimization): Update
5923 * opts.c (default_options_table): Enable auto-vectorization at
5924 O2 with very-cheap cost model.
5925 (finish_options): Use cheap cost model for
5926 explicit -ftree{,-loop}-vectorize.
5928 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
5930 * ctfc.c (ctfc_delete_container): Free hash table contents.
5932 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
5934 * toplev.c (process_options): Do not warn for GNU GIMPLE.
5936 2021-10-07 Siddhesh Poyarekar <siddhesh@gotplt.org>
5938 * tree-object-size.c (addr_object_size,
5939 compute_builtin_object_size): Drop PDECL and POFF arguments.
5940 (addr_object_size): Adjust calls.
5941 * tree-object-size.h (compute_builtin_object_size): Drop PDECL
5944 2021-10-07 Roger Sayle <roger@nextmovesoftware.com>
5946 * rtl.def (SMUL_HIGHPART, UMUL_HIGHPART): New RTX codes for
5947 representing signed and unsigned high-part multiplication resp.
5948 * simplify-rtx.c (simplify_binary_operation_1) [SMUL_HIGHPART,
5949 UMUL_HIGHPART]: Simplify high-part multiplications by zero.
5950 [SS_PLUS, US_PLUS, SS_MINUS, US_MINUS, SS_MULT, US_MULT,
5951 SS_DIV, US_DIV]: Similar simplifications for saturating
5953 (simplify_const_binary_operation) [SS_PLUS, US_PLUS, SS_MINUS,
5954 US_MINUS, SS_MULT, US_MULT, SMUL_HIGHPART, UMUL_HIGHPART]:
5955 Implement compile-time evaluation for constant operands.
5956 * dwarf2out.c (mem_loc_descriptor): Skip SMUL_HIGHPART and
5958 * doc/rtl.texi (smul_highpart, umul_highpart): Document RTX codes.
5959 * doc/md.texi (smul@var{m}3_highpart, umul@var{m3}_highpart):
5960 Mention the new smul_highpart and umul_highpart RTX codes.
5961 * doc/invoke.texi: Silence @xref "compilation" warnings.
5963 2021-10-07 Martin Jambor <mjambor@suse.cz>
5966 * ipa-prop.c (ipa_edge_args_sum_t::duplicate): Also handle the
5967 case when the source reference description corresponds to a
5968 referance taken in a function src->caller is inlined to.
5970 2021-10-07 Jan Hubicka <hubicka@ucw.cz>
5973 * ipa-modref-tree.h (modref_access_node::contains_p): Handle offsets
5975 (modref_access_node::try_merge_with): Add sanity check that there
5976 are no redundant entries in the list.
5978 2021-10-07 Richard Biener <rguenther@suse.de>
5980 PR tree-optimization/102608
5981 * tree-ssa-sccvn.c (visit_stmt): Drop .DEFERRED_INIT to
5984 2021-10-07 Martin Liska <mliska@suse.cz>
5986 * toplev.c (toplev::main): Make
5987 save_opt_decoded_options a pointer type
5988 * toplev.h: Likewise.
5990 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
5992 * config/gcn/gcn-valu.md (gather<mode>_insn_2offsets<exec>): Apply
5993 HAVE_GCN_ASM_GLOBAL_LOAD_FIXED.
5994 (scatter<mode>_insn_2offsets<exec_scatter>): Likewise.
5996 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
5998 * config/gcn/gcn-hsa.h (SRAMOPT): Include the whole option string.
5999 Adjust for new -msram-ecc=any behaviour.
6000 (ASM_SPEC): Adjust -mxnack and -msram-ecc usage.
6001 * config/gcn/gcn.c (output_file_start): Implement -msram-ecc=any.
6002 * config/gcn/mkoffload.c (EF_AMDGPU_XNACK): Rename to ...
6003 (EF_AMDGPU_XNACK_V3): ... this.
6004 (EF_AMDGPU_SRAM_ECC): Rename to ...
6005 (EF_AMDGPU_SRAM_ECC_V3): ... this.
6006 (EF_AMDGPU_FEATURE_XNACK_V4): New.
6007 (EF_AMDGPU_FEATURE_XNACK_UNSUPPORTED_V4): New.
6008 (EF_AMDGPU_FEATURE_XNACK_ANY_V4): New.
6009 (EF_AMDGPU_FEATURE_XNACK_OFF_V4): New.
6010 (EF_AMDGPU_FEATURE_XNACK_ON_V4): New.
6011 (EF_AMDGPU_FEATURE_SRAMECC_V4): New.
6012 (EF_AMDGPU_FEATURE_SRAMECC_UNSUPPORTED_V4): New.
6013 (EF_AMDGPU_FEATURE_SRAMECC_ANY_V4): New.
6014 (EF_AMDGPU_FEATURE_SRAMECC_OFF_V4): New.
6015 (EF_AMDGPU_FEATURE_SRAMECC_ON_V4): New.
6016 (SET_XNACK_ON): New.
6017 (SET_XNACK_OFF): New.
6019 (SET_SRAM_ECC_ON): New.
6020 (SET_SRAM_ECC_ANY): New.
6021 (SET_SRAM_ECC_OFF): New.
6022 (TEST_SRAM_ECC_ANY): New.
6023 (TEST_SRAM_ECC_ON): New.
6024 (main): Implement HSACOv4 and -msram-ecc=any.
6026 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
6028 * config.in: Regenerate.
6029 * config/gcn/gcn-hsa.h (X_FIJI): New macro.
6033 (A_FIJI): Rename to ...
6035 (A_900): Rename to ...
6037 (A_906): Rename to ...
6039 (A_908): Rename to ...
6041 (SRAMOPT): New macro.
6042 (ASM_SPEC): Adjust xnack option usage.
6043 * config/gcn/gcn.c (output_file_start): Adjust amdgcn_target usage.
6044 * configure: Regenerate.
6045 * configure.ac: Detect LLVM assembler dialect.
6047 2021-10-07 Richard Biener <rguenther@suse.de>
6049 * tree-pretty-print.c (dump_generic_node): Do not elide
6050 printing '&' when dumping with -gimple.
6052 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
6054 * gimple-range-cache.cc (non_null_ref::adjust_range): Call new
6056 * gimple-range-fold.cc (adjust_pointer_diff_expr): Ditto.
6057 (adjust_imagpart_expr): Ditto.
6058 * value-range.cc (irange::irange_intersect): Call new routine if
6059 RHS is a single pair.
6060 (irange::intersect): New wide_int version.
6061 * value-range.h (class irange): New prototype.
6063 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
6065 * gimple-range-edge.cc (gimple_outgoing_range::gimple_outgoing_range):
6066 Add parameter to limit size when recognizing switches.
6067 (gimple_outgoing_range::edge_range_p): Check size limit.
6068 * gimple-range-edge.h (gimple_outgoing_range): Add size field.
6069 * gimple-range-gori.cc (gori_map::calculate_gori): Ignore switches
6070 that exceed the size limit.
6071 (gori_compute::gori_compute): Add initializer.
6072 * params.opt (evrp-switch-limit): New.
6073 * doc/invoke.texi: Update docs.
6075 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
6077 * value-range.h (irange::set_varying): Use TYPE_MIN_VALUE and
6078 TYPE_MAX_VALUE instead of creating new trees when possible.
6080 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
6082 * gimple-range-cache.cc (non_null_ref::adjust_range): Check for
6083 zero and non-zero more efficently.
6085 2021-10-06 Richard Biener <rguenther@suse.de>
6088 * dumpfile.h (TDF_GIMPLE_VAL): New.
6089 (dump_flag): Re-order and adjust TDF_* flags. Make
6090 the enum uint32_t. Use std::underlying_type in the
6092 (optgroup_flag): Likewise for the operator overloads.
6093 * tree-pretty-print.c (dump_generic_node): Wrap ADDR_EXPR
6094 in _Literal if TDF_GIMPLE_VAL.
6095 * gimple-pretty-print.c (dump_gimple_assign): Add
6096 TDF_GIMPLE_VAL to flags when dumping operands where only
6097 is_gimple_val are allowed.
6098 (dump_gimple_cond): Likewise.
6100 2021-10-06 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
6102 * gimple-isel.cc (gimple_expand_vec_cond_expr): Remove redundant if
6105 2021-10-05 qing zhao <qing.zhao@oracle.com>
6107 PR middle-end/102359
6108 * gimplify.c (gimplify_decl_expr): Not add initialization for an
6109 auto variable when it has been initialized by frontend.
6111 2021-10-05 Aldy Hernandez <aldyh@redhat.com>
6113 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
6116 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
6118 * common/config/avr/avr-common.c (avr_handle_option): Mark
6119 argument as ATTRIBUTE_UNUSED.
6121 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
6123 * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC):
6124 Undefine before redefinition.
6126 2021-10-05 Richard Biener <rguenther@suse.de>
6128 * toplev.c (no_backend): Remove global var.
6129 (process_options): Pass in no_backend, move post_options
6130 langhook call to toplev::main.
6131 (do_compile): Pass in no_backend, move process_options call
6133 (toplev::run_self_tests): Check no_backend at the caller.
6134 (toplev::main): Call post_options and process_options
6135 split out from do_compile, do self-tests only if
6136 no_backend is initialized.
6138 2021-10-05 Richard Biener <rguenther@suse.de>
6140 * tree-cfg.c (dump_function_to_file): Dump the UID of the
6141 function as part of the name when requested.
6142 * tree-pretty-print.c (dump_function_name): Dump the UID when
6143 requested and the langhook produced the actual name.
6145 2021-10-05 Richard Biener <rguenther@suse.de>
6147 PR middle-end/102587
6148 PR middle-end/102285
6149 * internal-fn.c (expand_DEFERRED_INIT): Fall back to
6150 zero-initialization as last resort, use the constant
6151 size as given by the DEFERRED_INIT argument to build
6154 2021-10-04 Marek Polacek <polacek@redhat.com>
6157 * doc/invoke.texi: Document -Warray-compare.
6159 2021-10-04 Richard Biener <rguenther@suse.de>
6161 * gimplify.c (is_var_need_auto_init): DECL_HARD_REGISTER
6162 variables are not to be initialized.
6164 2021-10-04 Richard Biener <rguenther@suse.de>
6166 * expr.h (non_mem_decl_p): Declare.
6167 (mem_ref_refers_to_non_mem_p): Likewise.
6168 * expr.c (non_mem_decl_p): Export.
6169 (mem_ref_refers_to_non_mem_p): Likewise.
6170 * internal-fn.c (expand_DEFERRED_INIT): Do not expand the LHS
6171 but check the base with mem_ref_refers_to_non_mem_p
6174 2021-10-04 Richard Biener <rguenther@suse.de>
6176 PR tree-optimization/102570
6177 * tree-ssa-sccvn.h (vn_reference_op_struct): Document
6178 we are using clique for the internal function code.
6179 * tree-ssa-sccvn.c (vn_reference_op_eq): Compare the
6180 internal function code.
6181 (print_vn_reference_ops): Print the internal function code.
6182 (vn_reference_op_compute_hash): Hash it.
6183 (copy_reference_ops_from_call): Record it.
6184 (visit_stmt): Remove the restriction around internal function
6186 (fully_constant_vn_reference_p): Use fold_const_call and handle
6188 (vn_reference_eq): Compare call return types.
6189 * tree-ssa-pre.c (create_expression_by_pieces): Handle
6190 generating calls to internal functions.
6191 (compute_avail): Remove the restriction around internal function
6194 2021-10-04 Aldy Hernandez <aldyh@redhat.com>
6196 PR tree-optimization/102560
6197 * gimple-ssa-warn-alloca.c (alloca_call_type): Remove static
6198 marker for invalid_range.
6200 2021-10-04 Richard Biener <rguenther@suse.de>
6202 PR middle-end/102587
6203 * internal-fn.c (expand_DEFERRED_INIT): Guard register
6204 initialization path an avoid initializing VLA registers
6207 2021-10-04 Eric Botcazou <ebotcazou@adacore.com>
6209 * config/rs6000/vxworks.h (TARGET_INIT_LIBFUNCS): Delete.
6211 2021-10-03 Martin Liska <mliska@suse.cz>
6213 * toplev.c (toplev::main): Check opt_index if it is a part
6216 2021-10-02 Aldy Hernandez <aldyh@redhat.com>
6218 PR tree-optimization/102563
6219 * range-op.cc (operator_lshift::op1_range): Do not clobber
6222 2021-10-02 Martin Liska <mliska@suse.cz>
6224 * toplev.c (toplev::main): save_decoded_options[0] is program
6225 name and so it should be skipped.
6227 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
6229 PR tree-optimization/102546
6230 * range-op.cc (operator_lshift::op1_range): Teach range-ops that
6231 X << Y is non-zero implies X is also non-zero.
6233 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
6235 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
6237 * config/aarch64/aarch64-tune.md: Regenerate.
6238 * doc/invoke.texi: Update docs.
6240 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
6242 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
6244 * config/aarch64/aarch64-tune.md: Regenerate.
6245 * doc/invoke.texi: Update docs.
6247 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
6249 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
6251 * config/aarch64/aarch64-tune.md: Regenerate.
6252 * doc/invoke.texi: Update docs.
6254 2021-10-01 Martin Sebor <msebor@redhat.com>
6257 * doc/invoke.texi (-Waddress): Update.
6258 * gengtype.c (write_types): Avoid -Waddress.
6259 * poly-int.h (POLY_SET_COEFF): Avoid using null.
6261 2021-10-01 John David Anglin <danglin@gcc.gnu.org>
6264 * config/pa/pa.c (pa_option_override): Default to dwarf version 4
6267 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
6269 * config/aarch64/aarch64.h (AARCH64_FL_V9): Update value.
6271 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
6273 * gimple-range-path.cc (path_range_query::compute_ranges): Use
6275 * gimple-range-path.h (class path_range_query): Remove shadowed
6277 (path_range_query::get_path_oracle): New.
6279 2021-10-01 Jakub Jelinek <jakub@redhat.com>
6280 Richard Biener <rguenther@suse.de>
6283 * doc/invoke.texi (-fsanitize=integer-divide-by-zero): Remove
6284 INT_MIN / -1 division detection from here ...
6285 (-fsanitize=signed-integer-overflow): ... and add it here.
6287 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
6289 * config/aarch64/aarch64-arches.def (AARCH64_ARCH): Added
6291 * config/aarch64/aarch64.h (AARCH64_FL_V9): New.
6292 (AARCH64_FL_FOR_ARCH9): New flags for Armv9-A.
6293 (AARCH64_ISA_V9): New ISA flag.
6294 * doc/invoke.texi: Update docs.
6296 2021-10-01 Martin Liska <mliska@suse.cz>
6298 * toplev.c (toplev::main): Save decoded optimization options.
6299 * toplev.h (save_opt_decoded_options): New.
6300 * doc/extend.texi: Be more clear about optimize and target
6303 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
6305 * explow.c: Include langhooks.h.
6306 (set_stack_check_libfunc): Build a proper function type.
6308 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
6311 * config/i386/i386.c (legitimate_pic_address_disp_p): For PE-COFF do
6312 not return true for external weak function symbols in medium model.
6314 2021-10-01 Jakub Jelinek <jakub@redhat.com>
6316 * tree.h (OMP_CLAUSE_ORDER_REPRODUCIBLE): Define.
6317 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_ORDER>: Print
6318 reproducible: for OMP_CLAUSE_ORDER_REPRODUCIBLE.
6319 * omp-general.c (omp_extract_for_data): If OMP_CLAUSE_ORDER is seen
6320 without OMP_CLAUSE_ORDER_UNCONSTRAINED, overwrite sched_kind to
6321 OMP_CLAUSE_SCHEDULE_STATIC.
6323 2021-10-01 Richard Biener <rguenther@suse.de>
6325 PR middle-end/102518
6326 * tree-inline.c (setup_one_parameter): Avoid substituting
6327 an invariant into contexts where a GIMPLE register is not valid.
6329 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
6331 * config/arm/arm-cpus.in: Add Cortex-R52+ CPU.
6332 * config/arm/arm-tables.opt: Regenerate.
6333 * config/arm/arm-tune.md: Regenerate.
6334 * doc/invoke.texi: Update docs.
6336 2021-09-30 Uroš Bizjak <ubizjak@gmail.com>
6339 * config/i386/i386.md
6340 (sign_extend:WIDE (any_logic:NARROW (memory, immediate)) splitters):
6343 2021-09-30 Tobias Burnus <tobias@codesourcery.com>
6345 * omp-low.c (omp_runtime_api_call): Add omp_aligned_{,c}alloc and
6346 omp_{c,re}alloc, fix omp_alloc/omp_free.
6348 2021-09-30 Martin Liska <mliska@suse.cz>
6350 * defaults.h (ASM_OUTPUT_ASCII): Do not hide global variable
6351 asm_out_file and stream directly to MYFILE.
6353 2021-09-30 Richard Biener <rguenther@suse.de>
6355 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
6356 Restore and fix condition under which we apply npeel to
6357 the DRs misalignment value.
6359 2021-09-30 Richard Biener <rguenther@suse.de>
6361 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
6362 Fix npeel check for variable amount of peeling.
6364 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
6366 * lto-wrapper.c (run_gcc): Plug snprintf overflow.
6368 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
6370 * gimple-range.cc (gimple_ranger::debug): New.
6371 * gimple-range.h (class gimple_ranger): Add debug.
6373 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
6375 PR middle-end/102519
6376 * tree-vrp.c (hybrid_threader::~hybrid_threader): Free m_query.
6378 2021-09-29 Indu Bhagat <indu.bhagat@oracle.com>
6381 * btfout.c (GTY): Add GTY (()) albeit for cosmetic only purpose.
6382 (btf_finalize): Empty the hash_map btf_var_ids.
6384 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
6386 * tree-vrp.c (thread_through_all_blocks): Return bool.
6387 (execute_vrp_threader): Return TODO_* flags.
6388 (pass_data_vrp_threader): Set todo_flags_finish to 0.
6390 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
6392 * timevar.def (TV_TREE_VRP_THREADER): New.
6393 * tree-vrp.c: Use TV_TREE_VRP_THREADER for VRP threader pass.
6395 2021-09-29 David Faust <david.faust@oracle.com>
6397 * config.gcc (bpf-*-*): Do not overwrite extra_headers.
6399 2021-09-29 Jonathan Wright <jonathan.wright@arm.com>
6401 * config/aarch64/aarch64-builtins.c (TYPES_BINOP_PPU): Define
6402 new type qualifier enum.
6403 (TYPES_TERNOP_SSSU): Likewise.
6404 (TYPES_TERNOP_PPPU): Likewise.
6405 * config/aarch64/aarch64-simd-builtins.def: Define PPU, SSU,
6406 PPPU and SSSU builtin generator macros for qtbl1 and qtbx1
6408 * config/aarch64/arm_neon.h (vqtbl1_p8): Use type-qualified
6409 builtin and remove casts.
6410 (vqtbl1_s8): Likewise.
6411 (vqtbl1q_p8): Likewise.
6412 (vqtbl1q_s8): Likewise.
6413 (vqtbx1_s8): Likewise.
6414 (vqtbx1_p8): Likewise.
6415 (vqtbx1q_s8): Likewise.
6416 (vqtbx1q_p8): Likewise.
6417 (vtbl1_p8): Likewise.
6418 (vtbl2_p8): Likewise.
6419 (vtbx2_p8): Likewise.
6421 2021-09-29 Richard Biener <rguenther@suse.de>
6423 * tree-vect-data-refs.c (vect_dr_misalign_for_aligned_access):
6425 (vect_update_misalignment_for_peel): Use it to update
6426 misaligned to the value necessary for an aligned access.
6427 (vect_get_peeling_costs_all_drs): Likewise.
6428 (vect_enhance_data_refs_alignment): Likewise.
6430 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6432 * config/aarch64/aarch64.c (aarch64_expand_cpymem): Count number of
6433 emitted operations and adjust heuristic for code size.
6435 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6437 * config/aarch64/aarch64.c (aarch64_expand_setmem): Count number of
6438 emitted operations and adjust heuristic for code size.
6440 2021-09-29 Jakub Jelinek <jakub@redhat.com>
6442 PR middle-end/102504
6443 * gimplify.c (gimplify_scan_omp_clauses): Use omp_check_private even
6444 in OMP_SCOPE clauses, not just on worksharing construct clauses.
6446 2021-09-28 Geng Qi <gengqi@linux.alibaba.com>
6448 * config/riscv/riscv.md (mulv<mode>4): Call gen_smul<mode>3_highpart.
6449 (<u>mulditi3): Call <su>muldi3_highpart.
6450 (<u>muldi3_highpart): Rename to <su>muldi3_highpart.
6451 (<u>mulsidi3): Call <su>mulsi3_highpart.
6452 (<u>mulsi3_highpart): Rename to <su>mulsi3_highpart.
6454 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
6456 * config/darwin.h (DSYMUTIL_SPEC): Recognize D sources.
6458 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
6460 * config/rs6000/darwin.h (FIXED_R13): Add for PPC64.
6461 (FIRST_SAVED_GP_REGNO): Save from R13 even when it is one
6464 2021-09-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6466 * config/aarch64/aarch64.h (AARCH64_FL_LS64): Define
6467 (AARCH64_FL_V8_7): Likewise.
6468 (AARCH64_FL_FOR_ARCH8_7): Likewise.
6469 * config/aarch64/aarch64-arches.def (armv8.7-a): Define.
6470 * config/aarch64/aarch64-option-extensions.def (ls64): Define.
6471 * doc/invoke.texi: Document the above.
6473 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
6475 * dbgcnt.c (dbg_cnt_counter): New.
6476 * dbgcnt.h (dbg_cnt_counter): New.
6477 * dumpfile.c (dump_options): Add entry for TDF_THREADING.
6478 * dumpfile.h (enum dump_flag): Add TDF_THREADING.
6479 * gimple-range-path.cc (DEBUG_SOLVER): Use TDF_THREADING.
6480 * tree-ssa-threadupdate.c (dump_jump_thread_path): Dump out
6483 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
6485 * cfgcleanup.c (pass_jump::execute): Check
6486 flag_expensive_optimizations.
6487 (pass_jump_after_combine::gate): Same.
6488 * doc/invoke.texi (-fthread-jumps): Enable for -O1.
6489 * opts.c (default_options_table): Enable -fthread-jumps at -O1.
6490 * tree-ssa-threadupdate.c
6491 (fwd_jt_path_registry::remove_jump_threads_including): Bail unless
6494 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
6496 * tree-ssa-reassoc.c (biased_names): New global.
6497 (propagate_bias_p): New function.
6498 (loop_carried_phi): Remove.
6499 (propagate_rank): Propagate bias along single uses.
6500 (get_rank): Update biased_names when needed.
6502 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
6504 * passes.def (pass_reassoc): Rename parameter to early_p.
6505 * tree-ssa-reassoc.c (reassoc_bias_loop_carried_phi_ranks_p):
6507 (phi_rank): Don't bias loop-carried phi ranks
6508 before vectorization pass.
6509 (execute_reassoc): Add bias_loop_carried_phi_ranks_p parameter.
6510 (pass_reassoc::pass_reassoc): Add bias_loop_carried_phi_ranks_p
6512 (pass_reassoc::set_param): Set bias_loop_carried_phi_ranks_p
6514 (pass_reassoc::execute): Pass bias_loop_carried_phi_ranks_p to
6516 (pass_reassoc::bias_loop_carried_phi_ranks_p): New member.
6518 2021-09-28 Jakub Jelinek <jakub@redhat.com>
6521 * config/i386/i386.c (standard_80387_constant_p): Don't recognize
6522 special 80387 instruction XFmode constants if flag_rounding_math.
6524 2021-09-28 Richard Biener <rguenther@suse.de>
6526 PR tree-optimization/100112
6527 * tree-ssa-sccvn.c (visit_reference_op_load): Record the
6528 referece into the hashtable twice in case last_vuse is
6529 different from the original vuse on the stmt.
6531 2021-09-28 Jakub Jelinek <jakub@redhat.com>
6533 PR middle-end/102492
6534 * gimplify.c (gimplify_adjust_omp_clauses_1): Don't call the
6535 omp_finish_clause langhook on implicitly added OMP_CLAUSE_PRIVATE
6536 clauses on SIMD constructs.
6538 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
6540 PR tree-optimization/102511
6541 * gimple-range-path.cc (path_range_query::range_on_path_entry):
6542 Return VARYING when nothing found.
6544 2021-09-28 Hongyu Wang <hongyu.wang@intel.com>
6547 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): Add
6549 (VALID_SSE2_REG_VHF_MODE): Add V4HFmode and V2HFmode.
6550 (VALID_MMX_REG_MODE): Add V4HFmode.
6551 (SSE_REG_MODE_P): Replace VALID_AVX512FP16_REG_MODE with
6552 vector mode condition.
6553 * config/i386/i386.c (classify_argument): Parse V4HF/V2HF
6555 (function_arg_32): Add V4HFmode.
6556 (function_arg_advance_32): Likewise.
6557 * config/i386/i386.md (mode): Add V4HF/V2HF.
6558 (MODE_SIZE): Likewise.
6559 * config/i386/mmx.md (MMXMODE): Add V4HF mode.
6560 (V_32): Add V2HF mode.
6561 (VHF_32_64): New mode iterator.
6562 (*mov<mode>_internal): Adjust sse alternatives to support
6564 (*mov<mode>_internal): Adjust sse alternatives to support
6566 (<insn><mode>3): New define_insn for add/sub/mul/div.
6568 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
6570 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Check
6572 (pass_early_thread_jumps::gate): Same.
6573 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
6574 Return if !flag_thread_jumps.
6575 * tree-ssa-threadupdate.c
6576 (jt_path_registry::register_jump_thread): Assert that
6577 flag_thread_jumps is true.
6579 2021-09-28 liuhongt <hongtao.liu@intel.com>
6582 (simplify_context::simplify_binary_operation_1): Relax
6583 condition of simplifying (vec_concat:M (vec_select op0
6584 index0)(vec_select op1 index1)) to allow different modes
6585 between op0 and M, but have same inner mode.
6587 2021-09-28 liuhongt <hongtao.liu@intel.com>
6589 * config/i386/i386-expand.c (emit_reduc_half): Handle
6590 V8HF/V16HF/V32HFmode.
6591 * config/i386/sse.md (REDUC_SSE_PLUS_MODE): Add V8HF.
6592 (REDUC_SSE_SMINMAX_MODE): Ditto.
6593 (REDUC_PLUS_MODE): Add V16HF and V32HF.
6594 (REDUC_SMINMAX_MODE): Ditto.
6596 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
6598 * gimple-range-path.cc
6599 (path_range_query::precompute_ranges_in_block): Rename to...
6600 (path_range_query::compute_ranges_in_block): ...this.
6601 (path_range_query::precompute_ranges): Rename to...
6602 (path_range_query::compute_ranges): ...this.
6603 (path_range_query::precompute_relations): Rename to...
6604 (path_range_query::compute_relations): ...this.
6605 (path_range_query::precompute_phi_relations): Rename to...
6606 (path_range_query::compute_phi_relations): ...this.
6607 * gimple-range-path.h: Rename precompute* to compute*.
6608 * tree-ssa-threadbackward.c
6609 (back_threader::find_taken_edge_switch): Same.
6610 (back_threader::find_taken_edge_cond): Same.
6611 * tree-ssa-threadedge.c
6612 (hybrid_jt_simplifier::compute_ranges_from_state): Same.
6613 (hybrid_jt_state::register_equivs_stmt): Inline...
6614 * tree-ssa-threadedge.h: ...here.
6616 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
6618 * tree-vrp.c (lhs_of_dominating_assert): Remove.
6619 (class vrp_jt_state): Remove.
6620 (class vrp_jt_simplifier): Remove.
6621 (vrp_jt_simplifier::simplify): Remove.
6622 (class vrp_jump_threader): Remove.
6623 (vrp_jump_threader::vrp_jump_threader): Remove.
6624 (vrp_jump_threader::~vrp_jump_threader): Remove.
6625 (vrp_jump_threader::before_dom_children): Remove.
6626 (vrp_jump_threader::after_dom_children): Remove.
6628 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
6630 * passes.def (pass_vrp_threader): New.
6631 * tree-pass.h (make_pass_vrp_threader): Add make_pass_vrp_threader.
6632 * tree-ssa-threadedge.c (hybrid_jt_state::register_equivs_stmt): New.
6633 (hybrid_jt_simplifier::hybrid_jt_simplifier): New.
6634 (hybrid_jt_simplifier::simplify): New.
6635 (hybrid_jt_simplifier::compute_ranges_from_state): New.
6636 * tree-ssa-threadedge.h (class hybrid_jt_state): New.
6637 (class hybrid_jt_simplifier): New.
6638 * tree-vrp.c (execute_vrp): Remove ASSERT_EXPR based jump
6640 (class hybrid_threader): New.
6641 (hybrid_threader::hybrid_threader): New.
6642 (hybrid_threader::~hybrid_threader): New.
6643 (hybrid_threader::before_dom_children): New.
6644 (hybrid_threader::after_dom_children): New.
6645 (execute_vrp_threader): New.
6646 (class pass_vrp_threader): New.
6647 (make_pass_vrp_threader): New.
6649 2021-09-27 Martin Liska <mliska@suse.cz>
6651 * output.h (enum section_flag): New.
6652 (SECTION_FORGET): Remove.
6653 (SECTION_ENTSIZE): Make it (1UL << 8) - 1.
6654 (SECTION_STYLE_MASK): Define it based on other enum
6656 * varasm.c (switch_to_section): Remove unused handling of
6659 2021-09-27 Martin Liska <mliska@suse.cz>
6661 * common.opt: Add new variable flag_default_complex_method.
6662 * opts.c (finish_options): Handle flags related to
6663 x_flag_complex_method.
6664 * toplev.c (process_options): Remove option handling related
6665 to flag_complex_method.
6667 2021-09-27 Richard Biener <rguenther@suse.de>
6669 PR middle-end/102450
6670 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid using
6671 type_for_size, instead use int_mode_for_size.
6673 2021-09-27 Andrew Pinski <apinski@marvell.com>
6676 * gimplify.c (gimplify_save_expr): Return early
6677 if the type of val is error_mark_node.
6679 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
6681 * tree-ssanames.c (ssa_name_has_boolean_range): Use
6684 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
6686 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
6687 vrp_visit_cond_stmt.
6688 * tree-ssa-dom.c (cprop_operand): Convert to range_query API.
6689 (cprop_into_stmt): Same.
6690 (dom_opt_dom_walker::optimize_stmt): Same.
6692 2021-09-27 Richard Biener <rguenther@suse.de>
6694 PR tree-optimization/97351
6695 PR tree-optimization/97352
6696 PR tree-optimization/82426
6697 * tree-vectorizer.h (dr_misalignment): Add vector type
6699 (aligned_access_p): Likewise.
6700 (known_alignment_for_access_p): Likewise.
6701 (vect_supportable_dr_alignment): Likewise.
6702 (vect_known_alignment_in_bytes): Likewise. Refactor.
6703 (DR_MISALIGNMENT): Remove.
6704 (vect_update_shared_vectype): Likewise.
6705 * tree-vect-data-refs.c (dr_misalignment): Refactor, handle
6706 a vector type with larger alignment requirement and apply
6707 the negative step adjustment here.
6708 (vect_calculate_target_alignment): Remove.
6709 (vect_compute_data_ref_alignment): Get explicit vector type
6710 argument, do not apply a negative step alignment adjustment
6712 (vect_slp_analyze_node_alignment): Re-analyze alignment
6713 when we re-visit the DR with a bigger desired alignment but
6714 keep more precise results from smaller alignments.
6715 * tree-vect-slp.c (vect_update_shared_vectype): Remove.
6716 (vect_slp_analyze_node_operations_1): Do not update the
6717 shared vector type on stmts.
6718 * tree-vect-stmts.c (vect_analyze_stmt): Push/pop the
6719 vector type of an SLP node to the representative stmt-info.
6720 (vect_transform_stmt): Likewise.
6722 2021-09-27 liuhongt <hongtao.liu@intel.com>
6725 2021-09-09 liuhongt <hongtao.liu@intel.com>
6728 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
6729 (reduc_plus_scal_v4sf): .. this, New define_expand.
6730 (reduc_plus_scal_v2df): .. and this, New define_expand.
6732 2021-09-26 liuhongt <hongtao.liu@intel.com>
6734 * doc/extend.texi (Half-Precision): Remove storage only
6735 description for _Float16 w/o avx512fp16.
6737 2021-09-25 Dimitar Dimitrov <dimitar@dinux.eu>
6739 * config/pru/constraints.md (Rrio): New constraint.
6740 * config/pru/predicates.md (regio_operand): New predicate.
6741 * config/pru/pru-pragma.c (pru_register_pragmas): Register
6742 the __regio_symbol address space.
6743 * config/pru/pru-protos.h (pru_symref2ioregno): Declaration.
6744 * config/pru/pru.c (pru_symref2ioregno): New helper function.
6745 (pru_legitimate_address_p): Remove.
6746 (pru_addr_space_legitimate_address_p): Use the address space
6748 (pru_nongeneric_pointer_addrspace): New helper function.
6749 (pru_insert_attributes): New function to validate __regio_symbol
6751 (TARGET_INSERT_ATTRIBUTES): New macro.
6752 (TARGET_LEGITIMATE_ADDRESS_P): Remove.
6753 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): New macro.
6754 * config/pru/pru.h (enum reg_class): Add REGIO_REGS class.
6755 * config/pru/pru.md (*regio_readsi): New pattern to read I/O
6757 (*regio_nozext_writesi): New pattern to write to I/O registers.
6758 (*regio_zext_write_r30<EQS0:mode>): Ditto.
6759 * doc/extend.texi: Document the new PRU Named Address Space.
6761 2021-09-24 Patrick Palka <ppalka@redhat.com>
6765 * real.c (encode_ieee_double): Avoid unwanted sign extension.
6766 (encode_ieee_quad): Likewise.
6768 2021-09-24 Vladimir Makarov <vmakarov@redhat.com>
6770 PR rtl-optimization/102147
6771 * ira-build.c (ira_conflict_vector_profitable_p): Make
6772 profitability calculation independent of host compiler pointer and
6775 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
6777 * gimple-range-path.cc (path_range_query::path_range_query):
6778 Move debugging header...
6779 (path_range_query::precompute_ranges): ...here.
6780 (path_range_query::internal_range_of_expr): Do not call
6781 range_on_path_entry if NAME is defined in the current block.
6783 2021-09-24 Richard Biener <rguenther@suse.de>
6785 * cfghooks.c (verify_flow_info): Verify unallocated BB and
6786 edge flags are not set.
6788 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
6790 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
6792 (jt_path_registry::register_jump_thread): Call
6793 cancel_invalid_paths.
6794 * tree-ssa-threadupdate.h (class jt_path_registry): Add
6795 cancel_invalid_paths.
6797 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
6799 PR tree-optimization/102400
6800 * tree-ssa-sccvn.c (vn_reference_insert_pieces): Initialize
6801 result_vdef to zero value.
6803 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
6805 PR tree-optimization/102451
6806 * tree-ssa-dse.c (delete_dead_or_redundant_call): Record bb of stmt
6809 2021-09-24 Hongyu Wang <hongyu.wang@intel.com>
6811 * config/i386/sse.md (cond_<insn><mode>): Extend to support
6813 (cond_mul<mode>): Likewise.
6814 (cond_div<mode>): Likewise.
6815 (cond_<code><mode>): Likewise.
6816 (cond_fma<mode>): Likewise.
6817 (cond_fms<mode>): Likewise.
6818 (cond_fnma<mode>): Likewise.
6819 (cond_fnms<mode>): Likewise.
6821 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
6823 PR tree-optimization/102463
6824 * gimple-range-fold.cc (fold_using_range::relation_fold_and_or): If
6825 there is no range-ops handler, don't look for a relation.
6827 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
6829 * gimple-range-cache.cc (ranger_cache::ranger_cache): Take
6830 non-executable_edge flag as parameter.
6831 * gimple-range-cache.h (ranger_cache): Adjust prototype.
6832 * gimple-range-gori.cc (gori_compute::gori_compute): Take
6833 non-executable_edge flag as parameter.
6834 (gori_compute::outgoing_edge_range_p): Check new flag.
6835 * gimple-range-gori.h (gori_compute): Adjust prototype.
6836 * gimple-range.cc (gimple_ranger::gimple_ranger): Create new flag.
6837 (gimple_ranger::range_on_edge): Check new flag.
6838 * gimple-range.h (gimple_ranger::non_executable_edge_flag): New.
6839 * gimple-ssa-evrp.c (rvrp_folder): Pass ranger flag to simplifer.
6840 (hybrid_folder::hybrid_folder): Set ranger non-executable flag value.
6841 (hybrid_folder::fold_stmt): Set flag value in the simplifer.
6842 * vr-values.c (simplify_using_ranges::set_and_propagate_unexecutable):
6843 Use not_executable flag if provided inmstead of EDGE_EXECUTABLE.
6844 (simplify_using_ranges::simplify_switch_using_ranges): Clear
6845 EDGE_EXECUTABLE like it originally did.
6846 (simplify_using_ranges::cleanup_edges_and_switches): Clear any
6847 NON_EXECUTABLE flags.
6848 (simplify_using_ranges::simplify_using_ranges): Adjust.
6849 * vr-values.h (class simplify_using_ranges): Adjust.
6850 (simplify_using_ranges::set_range_query): Add non-executable flag param.
6852 2021-09-23 Bill Schmidt <wschmidt@linux.ibm.com>
6855 * config/rs6000/rs6000-call.c (rs6000_aggregate_candidate): Detect
6856 zero-width bit fields and return indicator.
6857 (rs6000_discover_homogeneous_aggregate): Diagnose when the
6858 presence of a zero-width bit field changes parameter passing in
6861 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
6863 * gimple-range-fold.cc (fold_using_range::range_of_phi):
6864 Remove dominator check.
6866 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
6868 * gimple-range-path.cc (path_range_query::precompute_relations):
6869 Hoist edge calculations before using EDGE_SUCC.
6871 2021-09-23 Jonathan Wakely <jwakely@redhat.com>
6873 * configure.ac: Fix --with-multilib-list description.
6874 * configure: Regenerate.
6876 2021-09-23 Richard Biener <rguenther@suse.de>
6878 PR tree-optimization/102448
6879 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
6880 Clear alignment info copied from DR_PTR_INFO.
6882 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
6884 * config/i386/i386-expand.c (ix86_use_mask_cmp_p): Enable
6886 * config/i386/sse.md (sseintvecmodelower): Add HF vector modes.
6887 (<avx512>_store<mode>_mask): Extend to support HF vector modes.
6888 (vec_cmp<mode><avx512fmaskmodelower>): Likewise.
6889 (vcond_mask_<mode><avx512fmaskmodelower>): Likewise.
6890 (vcond<mode><mode>): New expander.
6891 (vcond<mode><sseintvecmodelower>): Likewise.
6892 (vcond<sseintvecmodelower><mode>): Likewise.
6893 (vcondu<mode><sseintvecmodelower>): Likewise.
6895 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
6897 * config/i386/sse.md (extend<ssePHmodelower><mode>2):
6899 (extendv4hf<mode>2): Likewise.
6900 (extendv2hfv2df2): Likewise.
6901 (trunc<mode><ssePHmodelower>2): Likewise.
6902 (avx512fp16_vcvt<castmode>2ph_<mode>): Rename to ...
6903 (trunc<mode>v4hf2): ... this, and drop constraints.
6904 (avx512fp16_vcvtpd2ph_v2df): Rename to ...
6905 (truncv2dfv2hf2): ... this, and likewise.
6907 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
6909 * config/i386/sse.md (float<floatunssuffix><mode><ssePHmodelower>2):
6911 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>):
6913 (float<floatunssuffix><mode>v4hf2): ... this, and drop constraints.
6914 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Rename to ...
6915 (float<floatunssuffix>v2div2hf2): ... this, and likewise.
6917 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
6919 * config/i386/i386.md (fix<fixunssuffix>_trunchf<mode>2): New expander.
6920 (fixuns_trunchfhi2): Likewise.
6921 (*fixuns_trunchfsi2zext): New define_insn.
6922 * config/i386/sse.md (ssePHmodelower): New mode_attr.
6923 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
6924 New expander for same element vector fix_truncate.
6925 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
6926 Likewise for V4HF to V4SI/V4DI fix_truncate.
6927 (fix<fixunssuffix>_truncv2hfv2di2):
6928 Likeise for V2HF to V2DI fix_truncate.
6930 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
6932 * config/i386/i386.md (<code>hf3): New expander.
6934 2021-09-23 liuhongt <hongtao.liu@intel.com>
6936 * config/i386/sse.md (FMAMODEM): extend to handle FP16.
6937 (VFH_SF_AVX512VL): Extend to handle HFmode.
6938 (VF_SF_AVX512VL): Deleted.
6940 2021-09-23 liuhongt <hongtao.liu@intel.com>
6942 * config/i386/i386.md (rinthf2): New expander.
6943 (nearbyinthf2): New expander.
6945 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
6947 * tree-ssa-dom.c (class dom_jump_threader_simplifier): Rename...
6948 (class dom_jt_state): ...this and provide virtual overrides.
6949 (dom_jt_state::register_equiv): New.
6950 (class dom_jt_simplifier): Rename from
6951 dom_jump_threader_simplifier.
6952 (dom_jump_threader_simplifier::simplify): Rename...
6953 (dom_jt_simplifier::simplify): ...to this.
6954 (pass_dominator::execute): Use dom_jt_simplifier and
6956 * tree-ssa-threadedge.c (jump_threader::jump_threader):
6958 (jt_state::register_equivs_stmt): Abstract out...
6959 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
6961 (jump_threader::thread_around_empty_blocks): Update state.
6962 (jump_threader::thread_through_normal_block): Same.
6963 (jt_state::jt_state): Remove.
6964 (jt_state::push): Remove pass specific bits. Keep block vector
6966 (jt_state::append_path): New.
6967 (jt_state::pop): Remove pass specific bits.
6968 (jt_state::register_equiv): Same.
6969 (jt_state::record_ranges_from_stmt): Same.
6970 (jt_state::register_equivs_on_edge): Same. Rename...
6971 (jt_state::register_equivs_edge): ...to this.
6972 (jt_state::dump): New.
6973 (jt_state::debug): New.
6974 (jump_threader_simplifier::simplify): Remove.
6975 (jt_state::get_path): New.
6976 * tree-ssa-threadedge.h (class jt_simplifier): Make into a base
6977 class. Expose common functionality as virtual methods.
6978 (class jump_threader_simplifier): Same. Rename...
6979 (class jt_simplifier): ...to this.
6980 * tree-vrp.c (class vrp_jump_threader_simplifier): Rename...
6981 (class vrp_jt_simplifier): ...to this. Provide pass specific
6983 (class vrp_jt_state): New.
6984 (vrp_jump_threader_simplifier::simplify): Rename...
6985 (vrp_jt_simplifier::simplify): ...to this. Inline code from
6986 what used to be the base class.
6987 (vrp_jump_threader::vrp_jump_threader): Use vrp_jt_state and
6990 2021-09-22 Tobias Burnus <tobias@codesourcery.com>
6993 * doc/invoke.texi (-Wno-missing-include-dirs.): Document Fortran
6996 2021-09-22 Roger Sayle <roger@nextmovesoftware.com>
6997 Richard Biener <rguenther@suse.de>
6999 * match.pd (negation simplifications): Implement some negation
7000 folding transformations from fold-const.c's fold_negate_expr.
7001 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Add a SIMPLIFY
7002 argument, to control whether the op should be simplified prior
7003 to looking up/assigning a value number.
7004 (vn_nary_build_or_lookup): Update call to vn_nary_build_or_lookup_1.
7005 (vn_nary_simplify): Likewise.
7006 (visit_nary_op): Likewise, but when constructing a NEGATE_EXPR
7007 now call vn_nary_build_or_lookup_1 disabling simplification.
7009 2021-09-22 Jiufu Guo <guojiufu@linux.ibm.com>
7011 PR tree-optimization/102087
7012 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
7013 Update bound/cmp/control for niter.
7015 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
7017 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
7018 Move check for non-empty BB here.
7019 (fur_source::register_outgoing_edges): ...from here.
7021 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
7023 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
7024 Remove call to improve_range_with_equivs.
7025 (path_range_query::improve_range_with_equivs): Remove
7026 * gimple-range-path.h: Remove improve_range_with_equivs.
7028 2021-09-22 dianhong xu <dianhong.xu@intel.com>
7030 * config/i386/avx512fp16intrin.h:
7031 (_mm512_mask_blend_ph): New intrinsic.
7032 (_mm512_permutex2var_ph): Ditto.
7033 (_mm512_permutexvar_ph): Ditto.
7034 * config/i386/avx512fp16vlintrin.h:
7035 (_mm256_mask_blend_ph): New intrinsic.
7036 (_mm256_permutex2var_ph): Ditto.
7037 (_mm256_permutexvar_ph): Ditto.
7038 (_mm_mask_blend_ph): Ditto.
7039 (_mm_permutex2var_ph): Ditto.
7040 (_mm_permutexvar_ph): Ditto.
7042 2021-09-22 dianhong xu <dianhong.xu@intel.com>
7044 * config/i386/avx512fp16intrin.h: Add new intrinsics.
7045 (_mm512_conj_pch): New intrinsic.
7046 (_mm512_mask_conj_pch): Ditto.
7047 (_mm512_maskz_conj_pch): Ditto.
7048 * config/i386/avx512fp16vlintrin.h: Add new intrinsics.
7049 (_mm256_conj_pch): New intrinsic.
7050 (_mm256_mask_conj_pch): Ditto.
7051 (_mm256_maskz_conj_pch): Ditto.
7052 (_mm_conj_pch): Ditto.
7053 (_mm_mask_conj_pch): Ditto.
7054 (_mm_maskz_conj_pch): Ditto.
7056 2021-09-22 dianhong xu <dianhong.xu@intel.com>
7058 * config/i386/avx512fp16intrin.h (_MM512_REDUCE_OP): New macro
7059 (_mm512_reduce_add_ph): New intrinsic.
7060 (_mm512_reduce_mul_ph): Ditto.
7061 (_mm512_reduce_min_ph): Ditto.
7062 (_mm512_reduce_max_ph): Ditto.
7063 * config/i386/avx512fp16vlintrin.h
7064 (_MM256_REDUCE_OP/_MM_REDUCE_OP): New macro.
7065 (_mm256_reduce_add_ph): New intrinsic.
7066 (_mm256_reduce_mul_ph): Ditto.
7067 (_mm256_reduce_min_ph): Ditto.
7068 (_mm256_reduce_max_ph): Ditto.
7069 (_mm_reduce_add_ph): Ditto.
7070 (_mm_reduce_mul_ph): Ditto.
7071 (_mm_reduce_min_ph): Ditto.
7072 (_mm_reduce_max_ph): Ditto.
7074 2021-09-22 dianhong xu <dianhong.xu@intel.com>
7076 * config/i386/avx512fp16intrin.h (__m512h_u, __m256h_u,
7077 __m128h_u): New typedef.
7078 (_mm512_load_ph): New intrinsic.
7079 (_mm256_load_ph): Ditto.
7080 (_mm_load_ph): Ditto.
7081 (_mm512_loadu_ph): Ditto.
7082 (_mm256_loadu_ph): Ditto.
7083 (_mm_loadu_ph): Ditto.
7084 (_mm512_store_ph): Ditto.
7085 (_mm256_store_ph): Ditto.
7086 (_mm_store_ph): Ditto.
7087 (_mm512_storeu_ph): Ditto.
7088 (_mm256_storeu_ph): Ditto.
7089 (_mm_storeu_ph): Ditto.
7090 (_mm512_abs_ph): Ditto.
7091 * config/i386/avx512fp16vlintrin.h
7092 (_mm_abs_ph): Ditto.
7093 (_mm256_abs_ph): Ditto.
7095 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
7097 * config/s390/tpf.md (prologue_tpf, epilogue_tpf): Add cc clobber.
7099 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
7102 * config/s390/s390.c (s390_expand_insv): Emit a normal move if it
7103 is actually a full copy of the source operand into the target.
7104 Don't emit a strict low part move if source and target mode match.
7106 2021-09-22 Jakub Jelinek <jakub@redhat.com>
7108 PR middle-end/102415
7109 * omp-expand.c (expand_omp_single): If region->exit is NULL,
7110 assert region->entry is GIMPLE_OMP_SCOPE region and return.
7112 2021-09-22 Jakub Jelinek <jakub@redhat.com>
7114 * tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
7115 * tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
7116 arguments from 2 to 3.
7117 * tree-pretty-print.c (dump_omp_clause): Print allocator() around
7118 allocate clause allocator and print align if present.
7119 * omp-low.c (scan_sharing_clauses): Force allocate_map entry even
7120 for omp_default_mem_alloc if align modifier is present. If align
7121 modifier is present, use TREE_LIST to encode both allocator and
7123 (lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
7124 Handle align modifier on allocator clause if present.
7126 2021-09-22 liuhongt <hongtao.liu@intel.com>
7128 * config/i386/i386.md (define_attr "isa"): Add
7130 (define_attr "enabled"): Correspond fma_or_avx512vl to
7131 TARGET_FMA || TARGET_AVX512VL.
7132 * config/i386/mmx.md (fmav2sf4): Extend to AVX512 fma.
7137 2021-09-22 liuhongt <hongtao.liu@intel.com>
7139 * config/i386/i386.md (cstorehf3): New define_expand.
7141 2021-09-22 liuhongt <hongtao.liu@intel.com>
7143 * config/i386/i386.md (<rounding_insn>hf2): New expander.
7144 (sse4_1_round<mode>2): Extend from MODEF to MODEFH.
7145 * config/i386/sse.md (*sse4_1_round<ssescalarmodesuffix>):
7146 Extend from VF_128 to VFH_128.
7148 2021-09-22 liuhongt <hongtao.liu@intel.com>
7150 * config/i386/i386-features.c (i386-features.c): Handle
7152 * config/i386/i386.md (sqrthf2): New expander.
7153 (*sqrthf2): New define_insn.
7154 * config/i386/sse.md
7155 (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
7158 2021-09-22 liuhongt <hongtao.liu@intel.com>
7160 * config/i386/avx512fp16intrin.h (_mm_mask_fcmadd_sch):
7162 (_mm_mask3_fcmadd_sch): Likewise.
7163 (_mm_maskz_fcmadd_sch): Likewise.
7164 (_mm_fcmadd_sch): Likewise.
7165 (_mm_mask_fmadd_sch): Likewise.
7166 (_mm_mask3_fmadd_sch): Likewise.
7167 (_mm_maskz_fmadd_sch): Likewise.
7168 (_mm_fmadd_sch): Likewise.
7169 (_mm_mask_fcmadd_round_sch): Likewise.
7170 (_mm_mask3_fcmadd_round_sch): Likewise.
7171 (_mm_maskz_fcmadd_round_sch): Likewise.
7172 (_mm_fcmadd_round_sch): Likewise.
7173 (_mm_mask_fmadd_round_sch): Likewise.
7174 (_mm_mask3_fmadd_round_sch): Likewise.
7175 (_mm_maskz_fmadd_round_sch): Likewise.
7176 (_mm_fmadd_round_sch): Likewise.
7177 (_mm_fcmul_sch): Likewise.
7178 (_mm_mask_fcmul_sch): Likewise.
7179 (_mm_maskz_fcmul_sch): Likewise.
7180 (_mm_fmul_sch): Likewise.
7181 (_mm_mask_fmul_sch): Likewise.
7182 (_mm_maskz_fmul_sch): Likewise.
7183 (_mm_fcmul_round_sch): Likewise.
7184 (_mm_mask_fcmul_round_sch): Likewise.
7185 (_mm_maskz_fcmul_round_sch): Likewise.
7186 (_mm_fmul_round_sch): Likewise.
7187 (_mm_mask_fmul_round_sch): Likewise.
7188 (_mm_maskz_fmul_round_sch): Likewise.
7189 * config/i386/i386-builtin.def: Add corresponding new builtins.
7190 * config/i386/sse.md
7191 (avx512fp16_fmaddcsh_v8hf_maskz<round_expand_name>): New expander.
7192 (avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>): Ditto.
7193 (avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>):
7195 (avx512fp16_<complexopname>sh_v8hf_mask<round_name>): Ditto.
7196 (avx512fp16_<complexopname>sh_v8hf<mask_scalarc_name><round_scalarcz_name>):
7198 * config/i386/subst.md (mask_scalarcz_name): New.
7199 (mask_scalarc_name): Ditto.
7200 (mask_scalarc_operand3): Ditto.
7201 (mask_scalarcz_operand4): Ditto.
7202 (round_scalarcz_name): Ditto.
7203 (round_scalarc_mask_operand3): Ditto.
7204 (round_scalarcz_mask_operand4): Ditto.
7205 (round_scalarc_mask_op3): Ditto.
7206 (round_scalarcz_mask_op4): Ditto.
7207 (round_scalarcz_constraint): Ditto.
7208 (round_scalarcz_nimm_predicate): Ditto.
7209 (mask_scalarcz): Ditto.
7210 (mask_scalarc): Ditto.
7211 (round_scalarcz): Ditto.
7213 2021-09-22 liuhongt <hongtao.liu@intel.com>
7215 * config/i386/avx512fp16intrin.h (_mm512_fcmadd_pch):
7217 (_mm512_mask_fcmadd_pch): Likewise.
7218 (_mm512_mask3_fcmadd_pch): Likewise.
7219 (_mm512_maskz_fcmadd_pch): Likewise.
7220 (_mm512_fmadd_pch): Likewise.
7221 (_mm512_mask_fmadd_pch): Likewise.
7222 (_mm512_mask3_fmadd_pch): Likewise.
7223 (_mm512_maskz_fmadd_pch): Likewise.
7224 (_mm512_fcmadd_round_pch): Likewise.
7225 (_mm512_mask_fcmadd_round_pch): Likewise.
7226 (_mm512_mask3_fcmadd_round_pch): Likewise.
7227 (_mm512_maskz_fcmadd_round_pch): Likewise.
7228 (_mm512_fmadd_round_pch): Likewise.
7229 (_mm512_mask_fmadd_round_pch): Likewise.
7230 (_mm512_mask3_fmadd_round_pch): Likewise.
7231 (_mm512_maskz_fmadd_round_pch): Likewise.
7232 (_mm512_fcmul_pch): Likewise.
7233 (_mm512_mask_fcmul_pch): Likewise.
7234 (_mm512_maskz_fcmul_pch): Likewise.
7235 (_mm512_fmul_pch): Likewise.
7236 (_mm512_mask_fmul_pch): Likewise.
7237 (_mm512_maskz_fmul_pch): Likewise.
7238 (_mm512_fcmul_round_pch): Likewise.
7239 (_mm512_mask_fcmul_round_pch): Likewise.
7240 (_mm512_maskz_fcmul_round_pch): Likewise.
7241 (_mm512_fmul_round_pch): Likewise.
7242 (_mm512_mask_fmul_round_pch): Likewise.
7243 (_mm512_maskz_fmul_round_pch): Likewise.
7244 * config/i386/avx512fp16vlintrin.h (_mm_fmadd_pch):
7246 (_mm_mask_fmadd_pch): Likewise.
7247 (_mm_mask3_fmadd_pch): Likewise.
7248 (_mm_maskz_fmadd_pch): Likewise.
7249 (_mm256_fmadd_pch): Likewise.
7250 (_mm256_mask_fmadd_pch): Likewise.
7251 (_mm256_mask3_fmadd_pch): Likewise.
7252 (_mm256_maskz_fmadd_pch): Likewise.
7253 (_mm_fcmadd_pch): Likewise.
7254 (_mm_mask_fcmadd_pch): Likewise.
7255 (_mm_mask3_fcmadd_pch): Likewise.
7256 (_mm_maskz_fcmadd_pch): Likewise.
7257 (_mm256_fcmadd_pch): Likewise.
7258 (_mm256_mask_fcmadd_pch): Likewise.
7259 (_mm256_mask3_fcmadd_pch): Likewise.
7260 (_mm256_maskz_fcmadd_pch): Likewise.
7261 (_mm_fmul_pch): Likewise.
7262 (_mm_mask_fmul_pch): Likewise.
7263 (_mm_maskz_fmul_pch): Likewise.
7264 (_mm256_fmul_pch): Likewise.
7265 (_mm256_mask_fmul_pch): Likewise.
7266 (_mm256_maskz_fmul_pch): Likewise.
7267 (_mm_fcmul_pch): Likewise.
7268 (_mm_mask_fcmul_pch): Likewise.
7269 (_mm_maskz_fcmul_pch): Likewise.
7270 (_mm256_fcmul_pch): Likewise.
7271 (_mm256_mask_fcmul_pch): Likewise.
7272 (_mm256_maskz_fcmul_pch): Likewise.
7273 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF_V8HF_V8HF,
7274 V8HF_FTYPE_V16HF_V16HF_V16HF, V16HF_FTYPE_V16HF_V16HF_V16HF_UQI,
7275 V32HF_FTYPE_V32HF_V32HF_V32HF_INT,
7276 V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT): Add new builtin types.
7277 * config/i386/i386-builtin.def: Add new builtins.
7278 * config/i386/i386-expand.c: Handle new builtin types.
7279 * config/i386/subst.md (SUBST_CV): New.
7280 (maskc_name): Ditto.
7281 (maskc_operand3): Ditto.
7283 (sdc_maskz_name): Ditto.
7284 (sdc_mask_op4): Ditto.
7285 (sdc_mask_op5): Ditto.
7286 (sdc_mask_mode512bit_condition): Ditto.
7288 (round_maskc_operand3): Ditto.
7289 (round_sdc_mask_operand4): Ditto.
7290 (round_maskc_op3): Ditto.
7291 (round_sdc_mask_op4): Ditto.
7292 (round_saeonly_sdc_mask_operand5): Ditto.
7293 * config/i386/sse.md (unspec): Add complex fma unspecs.
7294 (avx512fmaskcmode): New.
7295 (UNSPEC_COMPLEX_F_C_MA): Ditto.
7296 (UNSPEC_COMPLEX_F_C_MUL): Ditto.
7297 (complexopname): Ditto.
7298 (<avx512>_fmaddc_<mode>_maskz<round_expand_name>): New expander.
7299 (<avx512>_fcmaddc_<mode>_maskz<round_expand_name>): Ditto.
7300 (fma_<complexopname>_<mode><sdc_maskz_name><round_name>): New
7302 (<avx512>_<complexopname>_<mode>_mask<round_name>): Ditto.
7303 (<avx512>_<complexopname>_<mode><maskc_name><round_name>): Ditto.
7305 2021-09-22 Kewen Lin <linkw@linux.ibm.com>
7307 * config/rs6000/rs6000.opt (rs6000-density-pct-threshold,
7308 rs6000-density-size-threshold, rs6000-density-penalty,
7309 rs6000-density-load-pct-threshold,
7310 rs6000-density-load-num-threshold): New parameter.
7311 * config/rs6000/rs6000.c (rs6000_density_test): Adjust with
7312 corresponding parameters.
7314 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
7316 * gimple-range-path.cc (path_range_query::defined_outside_path):
7318 (path_range_query::range_on_path_entry): New.
7319 (path_range_query::internal_range_of_expr): Resolve unknowns
7321 (path_range_query::improve_range_with_equivs): New.
7322 (path_range_query::ssa_range_in_phi): Resolve unknowns with
7324 * gimple-range-path.h (class path_range_query): Add
7325 defined_outside_path, range_on_path_entry, and
7326 improve_range_with_equivs.
7328 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
7330 * gimple-range-path.cc (path_range_query::add_to_imports): New.
7331 (path_range_query::add_copies_to_imports): New.
7332 (path_range_query::precompute_ranges): Call
7333 add_copies_to_imports.
7334 * gimple-range-path.h (class path_range_query): Add prototypes
7335 for add_copies_to_imports and add_to_imports.
7337 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
7339 * gimple-range-path.cc (path_range_query::range_defined_in_block):
7340 Remove useless code.
7342 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
7344 * gimple-range-fold.h (class fur_source): Make oracle protected.
7345 * gimple-range-path.cc (path_range_query::path_range_query): Add
7346 resolve argument. Initialize oracle.
7347 (path_range_query::~path_range_query): Delete oracle.
7348 (path_range_query::range_of_stmt): Adapt to use relations.
7349 (path_range_query::precompute_ranges): Pre-compute relations.
7350 (class jt_fur_source): New
7351 (jt_fur_source::jt_fur_source): New.
7352 (jt_fur_source::register_relation): New.
7353 (jt_fur_source::query_relation): New.
7354 (path_range_query::precompute_relations): New.
7355 (path_range_query::precompute_phi_relations): New.
7356 * gimple-range-path.h (path_range_query): Add resolve argument.
7357 Add oracle, precompute_relations, precompute_phi_relations.
7358 * tree-ssa-threadbackward.c (back_threader::back_threader): Pass
7359 resolve argument to solver.
7361 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
7363 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
7364 Rename postfold_gcond_edges to register_outgoing_edges and
7366 (fold_using_range::postfold_gcond_edges): Rename...
7367 (fur_source::register_outgoing_edges): ...to this.
7368 * gimple-range-fold.h (postfold_gcond_edges): Rename to
7369 register_outgoing_edges and move to fur_source.
7371 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
7373 * gimple-range-fold.cc (fold_using_range::range_of_phi): Check
7374 dom_info_available_p.
7376 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
7378 * gimple-range-cache.cc (non_null_ref::non_null_ref): Use create
7379 and quick_grow_cleared instead of safe_grow_cleared.
7381 2021-09-21 Thomas Schwinge <thomas@codesourcery.com>
7384 * omp-oacc-neuter-broadcast.cc (oacc_do_neutering): Evaluate
7387 2021-09-21 Richard Earnshaw <rearnsha@arm.com>
7389 * configure.ac: Detect when the assembler supports new-style
7390 architecture extensions.
7391 * common/config/arm/arm-common.c (arm_rewrite_mcpu): Return
7392 the full CPU string if the assembler can grok it.
7393 (arm_rewrite_march): Likewise but for the architecture.
7394 * config.in: Regenerate.
7395 * configure: Regenerate.
7397 2021-09-21 Richard Biener <rguenther@suse.de>
7399 PR tree-optimization/102421
7400 * tree-vect-loop.c (vect_dissolve_slp_only_groups): Copy and
7401 adjust alignment info.
7403 2021-09-21 Kewen Lin <linkw@linux.ibm.com>
7405 * ipa-fnsummary.c (ipa_fn_summary_write): Remove inconsistent
7406 bitfield stream out.
7408 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
7410 * gimple-range-fold.cc (fold_using_range::range_of_phi): Ignore
7411 undefined edges, apply an equivalence if appropriate.
7412 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Return
7413 UNDEFINED if EDGE_EXECUTABLE is not set.
7414 * gimple-range.cc (gimple_ranger::gimple_ranger): Set all edges
7415 as EXECUTABLE upon startup.
7416 (gimple_ranger::range_on_edge): Return UNDEFINED for edges without
7417 EDGE_EXECUTABLE set.
7418 * vr-values.c (set_and_propagate_unexecutable): New.
7419 (simplify_using_ranges::fold_cond): Call set_and_propagate.
7420 (simplify_using_ranges::simplify_switch_using_ranges): Ditto.
7421 * vr-values.h: Add prototype.
7423 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
7425 * value-relation.cc (equiv_oracle::register_initial_def): New.
7426 (equiv_oracle::register_relation): Call register_initial_def.
7427 (equiv_oracle::add_equiv_to_block): New. Split register_relation.
7428 (relation_oracle::register_stmt): Check def block of PHI arguments.
7429 * value-relation.h (equiv_oracle): Add new prototypes.
7431 2021-09-20 Matthias Kretz <m.kretz@gsi.de>
7433 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
7434 Define __RECIPROCAL_MATH__, __NO_SIGNED_ZEROS__,
7435 __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__, and
7436 __ROUNDING_MATH__ according to their corresponding flags.
7437 * doc/cpp.texi: Document __RECIPROCAL_MATH__,
7438 __NO_SIGNED_ZEROS__, __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__,
7439 and __ROUNDING_MATH__.
7441 2021-09-20 Richard Biener <rguenther@suse.de>
7443 * tree-vect-stmts.c (vectorizable_load): Use the vectype
7446 2021-09-20 Richard Biener <rguenther@suse.de>
7448 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
7449 Do not compute alignment of the vectorized access here.
7451 2021-09-20 Richard Biener <rguenther@suse.de>
7453 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
7454 Store -1 for runtime alias peeling iterations.
7456 2021-09-20 Richard Biener <rguenther@suse.de>
7458 * config.gcc: Obsolete hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11*.
7460 2021-09-20 Thomas Schwinge <thomas@codesourcery.com>
7462 * input.c (string_concat_db::record_string_concatenation)
7463 (string_concat_db::get_string_concatenation): Skip for
7464 'RESERVED_LOCATION_P'.
7466 2021-09-20 Richard Biener <rguenther@suse.de>
7468 PR tree-optimization/65206
7469 * tree-data-ref.h (struct data_reference): Add alt_indices,
7471 * tree-data-ref.c (free_data_ref): Release alt_indices.
7472 (dr_analyze_indices): Work on struct indices and get DR_REF as tree.
7473 (create_data_ref): Adjust.
7474 (initialize_data_dependence_relation): Split into head
7475 and tail. When the base objects fail to match up try
7476 again with pointer-based analysis of indices.
7477 * tree-vectorizer.c (vec_info_shared::check_datarefs): Do
7478 not compare the lazily computed alternate set of indices.
7480 2021-09-20 Iain Sandoe <iain@sandoe.co.uk>
7482 * gcc.c: Test for execute OK when we find the
7483 programs for assembler linker and dsymutil and those
7484 were specified at configure-time.
7486 2021-09-19 Martin Sebor <msebor@redhat.com>
7488 PR middle-end/102403
7489 * gimple-predicate-analysis.cc (predicate::init_from_control_deps):
7490 Correct a function pre/postcondition.
7492 2021-09-19 Martin Sebor <msebor@redhat.com>
7494 PR middle-end/102243
7495 * tree-ssa-strlen.c (get_range): Handle null cfun.
7497 2021-09-19 Iain Sandoe <iain@sandoe.co.uk>
7499 * config/darwin.h (LINK_COMMAND_SPEC_A): Use Darwin10
7500 unwinder shim as a convenience library.
7502 2021-09-19 Andrew Pinski <apinski@marvell.com>
7504 * doc/install.texi: Add note about
7505 binutils 2.35 is required for LTO usage.
7507 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
7509 * tree-ssa-threadbackward.c
7510 (back_threader_registry::register_path): Use push_edge.
7511 * tree-ssa-threadedge.c
7512 (jump_threader::thread_around_empty_blocks): Same.
7513 (jump_threader::thread_through_normal_block): Same.
7514 (jump_threader::thread_across_edge): Same. Also, use auto_bitmap.
7516 * tree-ssa-threadupdate.c
7517 (jt_path_registry::allocate_thread_edge): Remove.
7518 (jt_path_registry::push_edge): New.
7519 (dump_jump_thread_path): Make static.
7520 * tree-ssa-threadupdate.h (allocate_thread_edge): Remove.
7523 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
7525 * gimple-range-path.cc (path_range_query::path_range_query): Add
7527 (path_range_query::dump): Remove extern declaration of dump_ranger.
7528 * gimple-range-trace.cc (dump_ranger): Add DEBUG_FUNCTION marker.
7529 * gimple-range-trace.h (dump_ranger): Add prototype.
7531 2021-09-19 John Ericson <git@JohnEricson.me>
7533 * gcc.c (find_a_program): New function, factored out of...
7534 (find_a_file): Here.
7535 (execute): Use find_a_program when looking for programs rather
7538 2021-09-19 Matwey V. Kornilov <matwey.kornilov@gmail.com>
7540 * config/avr/avr-mcus.def: Add atmega324pb.
7541 * doc/avr-mmcu.texi: Corresponding changes.
7543 2021-09-19 Roger Sayle <roger@nextmovesoftware.com>
7546 * match.pd (cmp @0 REAL_CST@1): When @0 is also REAL_CST, apply
7547 the same transformations as to @1. For comparisons against NaN,
7548 don't check HONOR_SNANS but confirm that neither operand is a
7551 2021-09-19 Benjamin Peterson <benjamin@locrian.net>
7553 * attribs.c (make_unique_name): Delete.
7554 * attribs.h (make_unique_name): Delete.
7556 2021-09-19 Andrew Pinski <apinski@marvell.com>
7558 * lra-constraints.c (check_and_process_move): Assert
7559 that dclass and sclass are greater than or equal to NO_REGS.
7561 2021-09-18 Jakub Jelinek <jakub@redhat.com>
7563 * tree.h (OMP_CLAUSE_ORDER_UNCONSTRAINED): Define.
7564 * tree-pretty-print.c (dump_omp_clause): Print unconstrained:
7565 for OMP_CLAUSE_ORDER_UNCONSTRAINED.
7567 2021-09-18 liuhongt <hongtao.liu@intel.com>
7569 * config/i386/i386-features.c (remove_partial_avx_dependency):
7570 Restrict TARGET_USE_VECTOR_FP_CONVERTS and
7571 TARGET_USE_VECTOR_CONVERTS to conversion instructions only.
7573 2021-09-18 Jakub Jelinek <jakub@redhat.com>
7575 * gimplify.c (omp_default_clause): For C/C++ default({,first}private),
7576 if file/namespace scope variable doesn't have predetermined sharing,
7577 treat it as if there was default(none).
7579 2021-09-18 liuhongt <hongtao.liu@intel.com>
7581 * config/i386/avx512fp16intrin.h (_mm_fmadd_sh):
7583 (_mm_mask_fmadd_sh): Likewise.
7584 (_mm_mask3_fmadd_sh): Likewise.
7585 (_mm_maskz_fmadd_sh): Likewise.
7586 (_mm_fmadd_round_sh): Likewise.
7587 (_mm_mask_fmadd_round_sh): Likewise.
7588 (_mm_mask3_fmadd_round_sh): Likewise.
7589 (_mm_maskz_fmadd_round_sh): Likewise.
7590 (_mm_fnmadd_sh): Likewise.
7591 (_mm_mask_fnmadd_sh): Likewise.
7592 (_mm_mask3_fnmadd_sh): Likewise.
7593 (_mm_maskz_fnmadd_sh): Likewise.
7594 (_mm_fnmadd_round_sh): Likewise.
7595 (_mm_mask_fnmadd_round_sh): Likewise.
7596 (_mm_mask3_fnmadd_round_sh): Likewise.
7597 (_mm_maskz_fnmadd_round_sh): Likewise.
7598 (_mm_fmsub_sh): Likewise.
7599 (_mm_mask_fmsub_sh): Likewise.
7600 (_mm_mask3_fmsub_sh): Likewise.
7601 (_mm_maskz_fmsub_sh): Likewise.
7602 (_mm_fmsub_round_sh): Likewise.
7603 (_mm_mask_fmsub_round_sh): Likewise.
7604 (_mm_mask3_fmsub_round_sh): Likewise.
7605 (_mm_maskz_fmsub_round_sh): Likewise.
7606 (_mm_fnmsub_sh): Likewise.
7607 (_mm_mask_fnmsub_sh): Likewise.
7608 (_mm_mask3_fnmsub_sh): Likewise.
7609 (_mm_maskz_fnmsub_sh): Likewise.
7610 (_mm_fnmsub_round_sh): Likewise.
7611 (_mm_mask_fnmsub_round_sh): Likewise.
7612 (_mm_mask3_fnmsub_round_sh): Likewise.
7613 (_mm_maskz_fnmsub_round_sh): Likewise.
7614 * config/i386/i386-builtin-types.def
7615 (V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT): New builtin type.
7616 * config/i386/i386-builtin.def: Add new builtins.
7617 * config/i386/i386-expand.c: Handle new builtin type.
7618 * config/i386/sse.md (fmai_vmfmadd_<mode><round_name>):
7619 Ajdust to support FP16.
7620 (fmai_vmfmsub_<mode><round_name>): Ditto.
7621 (fmai_vmfnmadd_<mode><round_name>): Ditto.
7622 (fmai_vmfnmsub_<mode><round_name>): Ditto.
7623 (*fmai_fmadd_<mode>): Ditto.
7624 (*fmai_fmsub_<mode>): Ditto.
7625 (*fmai_fnmadd_<mode><round_name>): Ditto.
7626 (*fmai_fnmsub_<mode><round_name>): Ditto.
7627 (avx512f_vmfmadd_<mode>_mask<round_name>): Ditto.
7628 (avx512f_vmfmadd_<mode>_mask3<round_name>): Ditto.
7629 (avx512f_vmfmadd_<mode>_maskz<round_expand_name>): Ditto.
7630 (avx512f_vmfmadd_<mode>_maskz_1<round_name>): Ditto.
7631 (*avx512f_vmfmsub_<mode>_mask<round_name>): Ditto.
7632 (avx512f_vmfmsub_<mode>_mask3<round_name>): Ditto.
7633 (*avx512f_vmfmsub_<mode>_maskz_1<round_name>): Ditto.
7634 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
7635 (*avx512f_vmfnmsub_<mode>_mask3<round_name>): Ditto.
7636 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
7637 (*avx512f_vmfnmadd_<mode>_mask<round_name>): Renamed to ...
7638 (avx512f_vmfnmadd_<mode>_mask<round_name>) ... this, and
7639 adjust to support FP16.
7640 (avx512f_vmfnmadd_<mode>_mask3<round_name>): Ditto.
7641 (avx512f_vmfnmadd_<mode>_maskz_1<round_name>): Ditto.
7642 (avx512f_vmfnmadd_<mode>_maskz<round_expand_name>): New
7645 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
7647 * config/i386/sse.md (avx512fmaskmodelower): Extend to support
7649 (maskload<mode><avx512fmaskmodelower>): Ditto.
7650 (maskstore<mode><avx512fmaskmodelower>): Ditto.
7652 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
7654 * config/i386/i386-expand.c (ix86_expand_fp_absneg_operator):
7656 (ix86_expand_copysign): Ditto.
7657 (ix86_expand_xorsign): Ditto.
7658 * config/i386/i386.c (ix86_build_const_vector): Handle HF vector
7660 (ix86_build_signbit_mask): Ditto.
7661 (ix86_can_change_mode_class): Ditto.
7662 * config/i386/i386.md
7663 (SSEMODEF): Add HFmode.
7664 (ssevecmodef): Ditto.
7665 (<code>hf2): New define_expand.
7666 (*<code>hf2_1): New define_insn_and_split.
7667 (copysign<mode>): Extend to support HFmode under AVX512FP16.
7668 (xorsign<mode>): Ditto.
7669 * config/i386/sse.md (VFB): New mode iterator.
7670 (VFB_128_256): Ditto.
7672 (sseintvecmode2): Support HF vector mode.
7673 (<code><mode>2): Use new mode iterator.
7674 (*<code><mode>2): Ditto.
7675 (copysign<mode>3): Ditto.
7676 (xorsign<mode>3): Ditto.
7677 (<code><mode>3<mask_name>): Ditto.
7678 (<code><mode>3<mask_name>): Ditto.
7679 (<sse>_andnot<mode>3<mask_name>): Adjust for HF vector mode.
7680 (<sse>_andnot<mode>3<mask_name>): Ditto.
7681 (*<code><mode>3<mask_name>): Ditto.
7682 (*<code><mode>3<mask_name>): Ditto.
7684 2021-09-18 liuhongt <hongtao.liu@intel.com>
7686 * config/i386/avx512fp16intrin.h (_mm512_mask_fmadd_ph):
7688 (_mm512_mask3_fmadd_ph): Likewise.
7689 (_mm512_maskz_fmadd_ph): Likewise.
7690 (_mm512_fmadd_round_ph): Likewise.
7691 (_mm512_mask_fmadd_round_ph): Likewise.
7692 (_mm512_mask3_fmadd_round_ph): Likewise.
7693 (_mm512_maskz_fmadd_round_ph): Likewise.
7694 (_mm512_fnmadd_ph): Likewise.
7695 (_mm512_mask_fnmadd_ph): Likewise.
7696 (_mm512_mask3_fnmadd_ph): Likewise.
7697 (_mm512_maskz_fnmadd_ph): Likewise.
7698 (_mm512_fnmadd_round_ph): Likewise.
7699 (_mm512_mask_fnmadd_round_ph): Likewise.
7700 (_mm512_mask3_fnmadd_round_ph): Likewise.
7701 (_mm512_maskz_fnmadd_round_ph): Likewise.
7702 (_mm512_fmsub_ph): Likewise.
7703 (_mm512_mask_fmsub_ph): Likewise.
7704 (_mm512_mask3_fmsub_ph): Likewise.
7705 (_mm512_maskz_fmsub_ph): Likewise.
7706 (_mm512_fmsub_round_ph): Likewise.
7707 (_mm512_mask_fmsub_round_ph): Likewise.
7708 (_mm512_mask3_fmsub_round_ph): Likewise.
7709 (_mm512_maskz_fmsub_round_ph): Likewise.
7710 (_mm512_fnmsub_ph): Likewise.
7711 (_mm512_mask_fnmsub_ph): Likewise.
7712 (_mm512_mask3_fnmsub_ph): Likewise.
7713 (_mm512_maskz_fnmsub_ph): Likewise.
7714 (_mm512_fnmsub_round_ph): Likewise.
7715 (_mm512_mask_fnmsub_round_ph): Likewise.
7716 (_mm512_mask3_fnmsub_round_ph): Likewise.
7717 (_mm512_maskz_fnmsub_round_ph): Likewise.
7718 * config/i386/avx512fp16vlintrin.h (_mm256_fmadd_ph):
7720 (_mm256_mask_fmadd_ph): Likewise.
7721 (_mm256_mask3_fmadd_ph): Likewise.
7722 (_mm256_maskz_fmadd_ph): Likewise.
7723 (_mm_fmadd_ph): Likewise.
7724 (_mm_mask_fmadd_ph): Likewise.
7725 (_mm_mask3_fmadd_ph): Likewise.
7726 (_mm_maskz_fmadd_ph): Likewise.
7727 (_mm256_fnmadd_ph): Likewise.
7728 (_mm256_mask_fnmadd_ph): Likewise.
7729 (_mm256_mask3_fnmadd_ph): Likewise.
7730 (_mm256_maskz_fnmadd_ph): Likewise.
7731 (_mm_fnmadd_ph): Likewise.
7732 (_mm_mask_fnmadd_ph): Likewise.
7733 (_mm_mask3_fnmadd_ph): Likewise.
7734 (_mm_maskz_fnmadd_ph): Likewise.
7735 (_mm256_fmsub_ph): Likewise.
7736 (_mm256_mask_fmsub_ph): Likewise.
7737 (_mm256_mask3_fmsub_ph): Likewise.
7738 (_mm256_maskz_fmsub_ph): Likewise.
7739 (_mm_fmsub_ph): Likewise.
7740 (_mm_mask_fmsub_ph): Likewise.
7741 (_mm_mask3_fmsub_ph): Likewise.
7742 (_mm_maskz_fmsub_ph): Likewise.
7743 (_mm256_fnmsub_ph): Likewise.
7744 (_mm256_mask_fnmsub_ph): Likewise.
7745 (_mm256_mask3_fnmsub_ph): Likewise.
7746 (_mm256_maskz_fnmsub_ph): Likewise.
7747 (_mm_fnmsub_ph): Likewise.
7748 (_mm_mask_fnmsub_ph): Likewise.
7749 (_mm_mask3_fnmsub_ph): Likewise.
7750 (_mm_maskz_fnmsub_ph): Likewise.
7751 * config/i386/i386-builtin.def: Add corresponding new builtins.
7752 * config/i386/sse.md
7753 (<avx512>_fmadd_<mode>_maskz<round_expand_name>): Adjust to
7754 support HF vector modes.
7755 (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
7757 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
7758 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
7759 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
7760 (<avx512>_fmadd_<mode>_mask<round_name>): Ditto.
7761 (<avx512>_fmadd_<mode>_mask3<round_name>): Ditto.
7762 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Ditto.
7763 (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>):
7765 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
7766 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
7767 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
7768 (<avx512>_fmsub_<mode>_mask<round_name>): Ditto.
7769 (<avx512>_fmsub_<mode>_mask3<round_name>): Ditto.
7770 (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>):
7772 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
7773 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
7774 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
7775 (<avx512>_fnmadd_<mode>_mask<round_name>): Ditto.
7776 (<avx512>_fnmadd_<mode>_mask3<round_name>): Ditto.
7777 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Ditto.
7778 (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>):
7780 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
7781 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
7782 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
7783 (<avx512>_fnmsub_<mode>_mask<round_name>): Ditto.
7784 (<avx512>_fnmsub_<mode>_mask3<round_name>): Ditto.
7786 2021-09-18 liuhongt <hongtao.liu@intel.com>
7788 * config/i386/avx512fp16intrin.h (_mm512_fmaddsub_ph):
7790 (_mm512_mask_fmaddsub_ph): Likewise.
7791 (_mm512_mask3_fmaddsub_ph): Likewise.
7792 (_mm512_maskz_fmaddsub_ph): Likewise.
7793 (_mm512_fmaddsub_round_ph): Likewise.
7794 (_mm512_mask_fmaddsub_round_ph): Likewise.
7795 (_mm512_mask3_fmaddsub_round_ph): Likewise.
7796 (_mm512_maskz_fmaddsub_round_ph): Likewise.
7797 (_mm512_mask_fmsubadd_ph): Likewise.
7798 (_mm512_mask3_fmsubadd_ph): Likewise.
7799 (_mm512_maskz_fmsubadd_ph): Likewise.
7800 (_mm512_fmsubadd_round_ph): Likewise.
7801 (_mm512_mask_fmsubadd_round_ph): Likewise.
7802 (_mm512_mask3_fmsubadd_round_ph): Likewise.
7803 (_mm512_maskz_fmsubadd_round_ph): Likewise.
7804 * config/i386/avx512fp16vlintrin.h (_mm256_fmaddsub_ph):
7806 (_mm256_mask_fmaddsub_ph): Likewise.
7807 (_mm256_mask3_fmaddsub_ph): Likewise.
7808 (_mm256_maskz_fmaddsub_ph): Likewise.
7809 (_mm_fmaddsub_ph): Likewise.
7810 (_mm_mask_fmaddsub_ph): Likewise.
7811 (_mm_mask3_fmaddsub_ph): Likewise.
7812 (_mm_maskz_fmaddsub_ph): Likewise.
7813 (_mm256_fmsubadd_ph): Likewise.
7814 (_mm256_mask_fmsubadd_ph): Likewise.
7815 (_mm256_mask3_fmsubadd_ph): Likewise.
7816 (_mm256_maskz_fmsubadd_ph): Likewise.
7817 (_mm_fmsubadd_ph): Likewise.
7818 (_mm_mask_fmsubadd_ph): Likewise.
7819 (_mm_mask3_fmsubadd_ph): Likewise.
7820 (_mm_maskz_fmsubadd_ph): Likewise.
7821 * config/i386/i386-builtin.def: Add corresponding new builtins.
7822 * config/i386/sse.md (VFH_SF_AVX512VL): New mode iterator.
7823 * (<avx512>_fmsubadd_<mode>_maskz<round_expand_name>): New expander.
7824 * (<avx512>_fmaddsub_<mode>_maskz<round_expand_name>): Use
7826 * (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
7828 * (<avx512>_fmaddsub_<mode>_mask<round_name>): Ditto.
7829 * (<avx512>_fmaddsub_<mode>_mask3<round_name>): Ditto.
7830 * (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
7832 * (<avx512>_fmsubadd_<mode>_mask<round_name>): Ditto.
7833 * (<avx512>_fmsubadd_<mode>_mask3<round_name>): Ditto.
7835 2021-09-18 liuhongt <hongtao.liu@intel.com>
7838 * config/i386/i386.c (ix86_print_operand): Handle
7839 V8HF/V16HF/V32HFmode.
7840 * config/i386/i386.h (VALID_BCST_MODE_P): Add HFmode.
7841 * config/i386/sse.md (avx512bcst): Remove.
7843 2021-09-17 Martin Sebor <msebor@redhat.com>
7845 * Makefile.in (OBJS): Add gimple-predicate-analysis.o.
7846 * tree-ssa-uninit.c (max_phi_args): Move to gimple-predicate-analysis.
7847 (MASK_SET_BIT, MASK_TEST_BIT, MASK_EMPTY): Same.
7848 (check_defs): Add comment.
7849 (can_skip_redundant_opnd): Update comment.
7850 (compute_uninit_opnds_pos): Adjust to namespace change.
7851 (find_pdom): Move to gimple-predicate-analysis.cc.
7853 (struct uninit_undef_val_t): New.
7854 (is_non_loop_exit_postdominating): Move to gimple-predicate-analysis.cc.
7855 (find_control_equiv_block): Same.
7856 (MAX_NUM_CHAINS, MAX_CHAIN_LEN, MAX_POSTDOM_CHECK): Same.
7857 (MAX_SWITCH_CASES): Same.
7858 (compute_control_dep_chain): Same.
7859 (find_uninit_use): Use predicate analyzer.
7860 (struct pred_info): Move to gimple-predicate-analysis.
7861 (convert_control_dep_chain_into_preds): Same.
7862 (find_predicates): Same.
7863 (collect_phi_def_edges): Same.
7864 (warn_uninitialized_phi): Use predicate analyzer.
7865 (find_def_preds): Move to gimple-predicate-analysis.
7866 (dump_pred_info): Same.
7867 (dump_pred_chain): Same.
7868 (dump_predicates): Same.
7869 (destroy_predicate_vecs): Remove.
7870 (execute_late_warn_uninitialized): New.
7871 (get_cmp_code): Move to gimple-predicate-analysis.
7872 (is_value_included_in): Same.
7873 (value_sat_pred_p): Same.
7874 (find_matching_predicate_in_rest_chains): Same.
7875 (is_use_properly_guarded): Same.
7876 (prune_uninit_phi_opnds): Same.
7877 (find_var_cmp_const): Same.
7878 (use_pred_not_overlap_with_undef_path_pred): Same.
7879 (pred_equal_p): Same.
7880 (is_neq_relop_p): Same.
7881 (is_neq_zero_form_p): Same.
7882 (pred_expr_equal_p): Same.
7883 (is_pred_expr_subset_of): Same.
7884 (is_pred_chain_subset_of): Same.
7885 (is_included_in): Same.
7886 (is_superset_of): Same.
7888 (simplify_pred): Same.
7889 (simplify_preds_2): Same.
7890 (simplify_preds_3): Same.
7891 (simplify_preds_4): Same.
7892 (simplify_preds): Same.
7894 (push_to_worklist): Same.
7895 (get_pred_info_from_cmp): Same.
7896 (is_degenerated_phi): Same.
7897 (normalize_one_pred_1): Same.
7898 (normalize_one_pred): Same.
7899 (normalize_one_pred_chain): Same.
7900 (normalize_preds): Same.
7901 (can_one_predicate_be_invalidated_p): Same.
7902 (can_chain_union_be_invalidated_p): Same.
7903 (uninit_uses_cannot_happen): Same.
7904 (pass_late_warn_uninitialized::execute): Define.
7905 * gimple-predicate-analysis.cc: New file.
7906 * gimple-predicate-analysis.h: New file.
7908 2021-09-17 Julian Brown <julian@codesourcery.com>
7910 * config/gcn/gcn.c (gimple.h): Include.
7911 (gcn_fork_join): Emit barrier for worker-level joins.
7912 * omp-oacc-neuter-broadcast.cc (find_local_vars_to_propagate): Add
7913 writes_gang_private bitmap parameter. Set bit for blocks
7914 containing gang-private variable writes.
7915 (worker_single_simple): Don't emit barrier after predicated block.
7916 (worker_single_copy): Don't emit barrier if we're not broadcasting
7917 anything and the block contains no gang-private writes.
7918 (neuter_worker_single): Don't predicate blocks that only contain
7919 NOPs or internal marker functions. Pass has_gang_private_write
7920 argument to worker_single_copy.
7921 (oacc_do_neutering): Add writes_gang_private bitmap handling.
7923 2021-09-17 Julian Brown <julian@codesourcery.com>
7925 * config/gcn/gcn-protos.h
7926 (gcn_goacc_create_worker_broadcast_record): Update prototype.
7927 * config/gcn/gcn-tree.c (gcn_goacc_get_worker_red_decl): Use
7928 preallocated block of LDS memory. Do not cache/share decls for
7929 reduction temporaries between invocations.
7930 (gcn_goacc_reduction_teardown): Unshare VAR on second use.
7931 (gcn_goacc_create_worker_broadcast_record): Add OFFSET parameter
7932 and return temporary LDS space at that offset. Return pointer in
7934 * config/gcn/gcn.c (acc_lds_size, gang_private_hwm, lds_allocs):
7936 (ACC_LDS_SIZE): Define as acc_lds_size.
7937 (gcn_init_machine_status): Don't initialise lds_allocated,
7938 lds_allocs, reduc_decls fields of machine function struct.
7939 (gcn_option_override): Handle default size for gang-private
7940 variables and -mgang-private-size option.
7941 (gcn_expand_prologue): Use LDS_SIZE instead of LDS_SIZE-1 when
7942 initialising M0_REG.
7943 (gcn_shared_mem_layout): New function.
7944 (gcn_print_lds_decl): Update comment. Use global lds_allocs map and
7945 gang_private_hwm variable.
7946 (TARGET_GOACC_SHARED_MEM_LAYOUT): Define target hook.
7947 * config/gcn/gcn.h (machine_function): Remove lds_allocated,
7948 lds_allocs, reduc_decls. Add reduction_base, reduction_limit.
7949 * config/gcn/gcn.opt (gang_private_size_opt): New global.
7950 (mgang-private-size=): New option.
7951 * doc/tm.texi.in (TARGET_GOACC_SHARED_MEM_LAYOUT): Place
7953 * doc/tm.texi: Regenerate.
7954 * omp-oacc-neuter-broadcast.cc (targhooks.h, diagnostic-core.h):
7956 (build_sender_ref): Handle sender_decl being pointer.
7957 (worker_single_copy): Add PLACEMENT and ISOLATE_BROADCASTS
7958 parameters. Pass placement argument to
7959 create_worker_broadcast_record hook invocations. Handle
7960 sender_decl being pointer and isolate_broadcasts inserting extra
7962 (blk_offset_map_t): Add typedef.
7963 (neuter_worker_single): Add BLK_OFFSET_MAP parameter. Pass
7964 preallocated range to worker_single_copy call.
7965 (dfs_broadcast_reachable_1): New function.
7966 (idx_decl_pair_t, used_range_vec_t): New typedefs.
7967 (sort_size_descending): New function.
7968 (addr_range): New class.
7969 (splay_tree_compare_addr_range, splay_tree_free_key)
7970 (first_fit_range, merge_ranges_1, merge_ranges): New functions.
7971 (execute_omp_oacc_neuter_broadcast): Rename to...
7972 (oacc_do_neutering): ... this. Add BOUNDS_LO, BOUNDS_HI
7973 parameters. Arrange layout of shared memory for broadcast
7975 (execute_omp_oacc_neuter_broadcast): New function.
7976 (pass_omp_oacc_neuter_broadcast::gate): Remove num_workers==1
7977 handling from here. Enable pass for all OpenACC routines in order
7978 to call shared memory-layout hook.
7979 * target.def (create_worker_broadcast_record): Add OFFSET
7981 (shared_mem_layout): New hook.
7983 2021-09-17 Julian Brown <julian@codesourcery.com>
7984 Thomas Schwinge <thomas@codesourcery.com>
7986 * omp-oacc-neuter-broadcast.cc
7987 (pass_omp_oacc_neuter_broadcast::gate): Disable if num_workers is
7989 (execute_omp_oacc_neuter_broadcast): Adjust.
7991 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
7993 * value-relation.cc (class equiv_chain): Move to header file.
7994 (path_oracle::path_oracle): New.
7995 (path_oracle::~path_oracle): New.
7996 (path_oracle::register_relation): New.
7997 (path_oracle::query_relation): New.
7998 (path_oracle::reset_path): New.
7999 (path_oracle::dump): New.
8000 * value-relation.h (class equiv_chain): Move to here.
8001 (class path_oracle): New.
8003 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
8005 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a DOM
8007 * gimple-range-fold.cc (fur_depend::register_relation): Use
8008 register_stmt/edge routines.
8009 * value-relation.cc (equiv_chain::find): Relocate from equiv_oracle.
8010 (equiv_oracle::equiv_oracle): Create self equivalence cache.
8011 (equiv_oracle::~equiv_oracle): Release same.
8012 (equiv_oracle::equiv_set): Return entry from self equiv cache if there
8013 are no equivalences.
8014 (equiv_oracle::find_equiv_block): Move list find to equiv_chain.
8015 (equiv_oracle::register_relation): Rename from register_equiv.
8016 (relation_chain_head::find_relation): Relocate from dom_oracle.
8017 (relation_oracle::register_stmt): New.
8018 (relation_oracle::register_edge): New.
8019 (dom_oracle::*): Rename from relation_oracle.
8020 (dom_oracle::register_relation): Adjust to call equiv_oracle.
8021 (dom_oracle::set_one_relation): Split from register_relation.
8022 (dom_oracle::register_transitives): Consolidate 2 methods.
8023 (dom_oracle::find_relation_block): Move core to relation_chain.
8024 (dom_oracle::query_relation): Rename from find_relation_dom and adjust.
8025 * value-relation.h (class relation_oracle): New pure virtual base.
8026 (class equiv_oracle): Inherit from relation_oracle and adjust.
8027 (class dom_oracle): Rename from old relation_oracle and adjust.
8029 2021-09-17 Martin Sebor <msebor@redhat.com>
8031 PR middle-end/102200
8032 * pointer-query.cc (access_ref::inform_access): Handle MIN/MAX_EXPR.
8033 (handle_min_max_size): Change argument. Store original SSA_NAME for
8034 operands to potentially distinct (sub)objects.
8035 (compute_objsize_r): Adjust call to the above.
8037 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
8039 * config/rs6000/rs6000.c (rs6000-builtins.h): New include.
8040 (rs6000_new_builtin_vectorized_function): New function.
8041 (rs6000_new_builtin_md_vectorized_function): Likewise.
8042 (rs6000_builtin_vectorized_function): Call
8043 rs6000_new_builtin_vectorized_function.
8044 (rs6000_builtin_md_vectorized_function): Call
8045 rs6000_new_builtin_md_vectorized_function.
8047 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
8049 * config/rs6000/rs6000-builtin-new.def (ASSEMBLE_ACC): Add mmaint flag.
8050 (ASSEMBLE_PAIR): Likewise.
8051 (BUILD_ACC): Likewise.
8052 (DISASSEMBLE_ACC): Likewise.
8053 (DISASSEMBLE_PAIR): Likewise.
8054 (PMXVBF16GER2): Likewise.
8055 (PMXVBF16GER2NN): Likewise.
8056 (PMXVBF16GER2NP): Likewise.
8057 (PMXVBF16GER2PN): Likewise.
8058 (PMXVBF16GER2PP): Likewise.
8059 (PMXVF16GER2): Likewise.
8060 (PMXVF16GER2NN): Likewise.
8061 (PMXVF16GER2NP): Likewise.
8062 (PMXVF16GER2PN): Likewise.
8063 (PMXVF16GER2PP): Likewise.
8064 (PMXVF32GER): Likewise.
8065 (PMXVF32GERNN): Likewise.
8066 (PMXVF32GERNP): Likewise.
8067 (PMXVF32GERPN): Likewise.
8068 (PMXVF32GERPP): Likewise.
8069 (PMXVF64GER): Likewise.
8070 (PMXVF64GERNN): Likewise.
8071 (PMXVF64GERNP): Likewise.
8072 (PMXVF64GERPN): Likewise.
8073 (PMXVF64GERPP): Likewise.
8074 (PMXVI16GER2): Likewise.
8075 (PMXVI16GER2PP): Likewise.
8076 (PMXVI16GER2S): Likewise.
8077 (PMXVI16GER2SPP): Likewise.
8078 (PMXVI4GER8): Likewise.
8079 (PMXVI4GER8PP): Likewise.
8080 (PMXVI8GER4): Likewise.
8081 (PMXVI8GER4PP): Likewise.
8082 (PMXVI8GER4SPP): Likewise.
8083 (XVBF16GER2): Likewise.
8084 (XVBF16GER2NN): Likewise.
8085 (XVBF16GER2NP): Likewise.
8086 (XVBF16GER2PN): Likewise.
8087 (XVBF16GER2PP): Likewise.
8088 (XVF16GER2): Likewise.
8089 (XVF16GER2NN): Likewise.
8090 (XVF16GER2NP): Likewise.
8091 (XVF16GER2PN): Likewise.
8092 (XVF16GER2PP): Likewise.
8093 (XVF32GER): Likewise.
8094 (XVF32GERNN): Likewise.
8095 (XVF32GERNP): Likewise.
8096 (XVF32GERPN): Likewise.
8097 (XVF32GERPP): Likewise.
8098 (XVF64GER): Likewise.
8099 (XVF64GERNN): Likewise.
8100 (XVF64GERNP): Likewise.
8101 (XVF64GERPN): Likewise.
8102 (XVF64GERPP): Likewise.
8103 (XVI16GER2): Likewise.
8104 (XVI16GER2PP): Likewise.
8105 (XVI16GER2S): Likewise.
8106 (XVI16GER2SPP): Likewise.
8107 (XVI4GER8): Likewise.
8108 (XVI4GER8PP): Likewise.
8109 (XVI8GER4): Likewise.
8110 (XVI8GER4PP): Likewise.
8111 (XVI8GER4SPP): Likewise.
8112 (XXMFACC): Likewise.
8113 (XXMTACC): Likewise.
8114 (XXSETACCZ): Likewise.
8115 (ASSEMBLE_PAIR_V): Likewise.
8116 (BUILD_PAIR): Likewise.
8117 (DISASSEMBLE_PAIR_V): Likewise.
8120 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_mma_builtin):
8121 Handle RS6000_BIF_LXVP and RS6000_BIF_STXVP.
8122 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add ismmaint.
8123 (parse_bif_attrs): Handle ismmaint.
8124 (write_decls): Add bif_mmaint_bit and bif_is_mmaint.
8125 (write_bif_static_init): Handle ismmaint.
8127 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
8129 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin): New
8131 (rs6000_gimple_fold_builtin): Call rs6000_gimple_fold_new_builtin.
8132 (rs6000_new_builtin_valid_without_lhs): New function.
8133 (rs6000_gimple_fold_new_mma_builtin): Likewise.
8134 (rs6000_gimple_fold_new_builtin): Likewise.
8136 2021-09-17 Thomas Schwinge <thomas@codesourcery.com>
8138 * hash-table.h (hash_table<Descriptor, Lazy, Allocator>::expand):
8139 Destruct stale Value objects.
8140 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
8143 2021-09-17 Roger Sayle <roger@nextmovesoftware.com>
8146 * match.pd (shift optimizations): Disable recent sign-changing
8147 optimization for shifts by zero, these will be folded later.
8149 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
8151 * config/rs6000/rs6000-builtin-new.def (__builtin_mffsl): Move from
8152 [power9] to [always].
8154 2021-09-17 Richard Biener <rguenther@suse.de>
8156 * tree-vect-stmts.c (vectorizable_load): Do not frob
8159 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
8161 * config/i386/i386-features.c (remove_partial_avx_dependency):
8162 Also check TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY and
8163 and TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY before generating
8165 * config/i386/i386.h (TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY):
8167 (TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
8168 * config/i386/i386.md (SSE FP to FP splitters): Replace
8169 TARGET_SSE_PARTIAL_REG_DEPENDENCY with
8170 TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY.
8171 (SSE INT to FP splitter): Replace TARGET_SSE_PARTIAL_REG_DEPENDENCY
8172 with TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY.
8173 * config/i386/x86-tune.def
8174 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): New.
8175 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
8177 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
8180 * config/i386/i386-features.c (remove_partial_avx_dependency):
8181 Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
8182 before generating vxorps.
8184 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
8186 * config/i386/i386-options.c (processor_cost_table): Use
8187 tremont_cost for Tremont.
8188 * config/i386/x86-tune-costs.h (tremont_memcpy): New.
8189 (tremont_memset): Likewise.
8190 (tremont_cost): Likewise.
8191 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
8194 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
8196 * common/config/i386/i386-common.c: Use Haswell scheduling model
8198 * config/i386/i386.c (ix86_sched_init_global): Prepare for Tremont
8200 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Tremont
8202 (ix86_adjust_cost): Handle Tremont.
8203 * config/i386/x86-tune.def (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY):
8205 (X86_TUNE_USE_LEAVE): Likewise.
8206 (X86_TUNE_PUSH_MEMORY): Likewise.
8207 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
8208 (X86_TUNE_USE_CLTD): Likewise.
8209 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
8210 (X86_TUNE_AVOID_MFENCE): Likewise.
8211 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
8212 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
8213 (X86_TUNE_ACCUMULATE_OUTGOING_ARGS): Disable for Tremont.
8214 (X86_TUNE_FOUR_JUMP_LIMIT): Likewise.
8215 (X86_TUNE_OPT_AGU): Likewise.
8216 (X86_TUNE_AVOID_LEA_FOR_ADDR): Likewise.
8217 (X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE): Likewise.
8218 (X86_TUNE_EXPAND_ABS): Likewise.
8219 (X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS): Likewise.
8220 (X86_TUNE_SLOW_PSHUFB): Likewise.
8222 2021-09-17 Eric Botcazou <ebotcazou@adacore.com>
8224 PR rtl-optimization/102306
8225 * combine.c (try_combine): Abort the combination if we are about to
8226 duplicate volatile references.
8228 2021-09-17 liuhongt <hongtao.liu@intel.com>
8230 * config/i386/avx512fp16intrin.h (_mm_undefined_ph):
8232 (_mm256_undefined_ph): Likewise.
8233 (_mm512_undefined_ph): Likewise.
8234 (_mm_cvtsh_h): Likewise.
8235 (_mm256_cvtsh_h): Likewise.
8236 (_mm512_cvtsh_h): Likewise.
8237 (_mm512_castph_ps): Likewise.
8238 (_mm512_castph_pd): Likewise.
8239 (_mm512_castph_si512): Likewise.
8240 (_mm512_castph512_ph128): Likewise.
8241 (_mm512_castph512_ph256): Likewise.
8242 (_mm512_castph128_ph512): Likewise.
8243 (_mm512_castph256_ph512): Likewise.
8244 (_mm512_zextph128_ph512): Likewise.
8245 (_mm512_zextph256_ph512): Likewise.
8246 (_mm512_castps_ph): Likewise.
8247 (_mm512_castpd_ph): Likewise.
8248 (_mm512_castsi512_ph): Likewise.
8249 * config/i386/avx512fp16vlintrin.h (_mm_castph_ps):
8251 (_mm256_castph_ps): Likewise.
8252 (_mm_castph_pd): Likewise.
8253 (_mm256_castph_pd): Likewise.
8254 (_mm_castph_si128): Likewise.
8255 (_mm256_castph_si256): Likewise.
8256 (_mm_castps_ph): Likewise.
8257 (_mm256_castps_ph): Likewise.
8258 (_mm_castpd_ph): Likewise.
8259 (_mm256_castpd_ph): Likewise.
8260 (_mm_castsi128_ph): Likewise.
8261 (_mm256_castsi256_ph): Likewise.
8262 (_mm256_castph256_ph128): Likewise.
8263 (_mm256_castph128_ph256): Likewise.
8264 (_mm256_zextph128_ph256): Likewise.
8266 2021-09-17 liuhongt <hongtao.liu@intel.com>
8268 * config/i386/avx512fp16intrin.h (_mm_cvtsh_ss):
8270 (_mm_mask_cvtsh_ss): Likewise.
8271 (_mm_maskz_cvtsh_ss): Likewise.
8272 (_mm_cvtsh_sd): Likewise.
8273 (_mm_mask_cvtsh_sd): Likewise.
8274 (_mm_maskz_cvtsh_sd): Likewise.
8275 (_mm_cvt_roundsh_ss): Likewise.
8276 (_mm_mask_cvt_roundsh_ss): Likewise.
8277 (_mm_maskz_cvt_roundsh_ss): Likewise.
8278 (_mm_cvt_roundsh_sd): Likewise.
8279 (_mm_mask_cvt_roundsh_sd): Likewise.
8280 (_mm_maskz_cvt_roundsh_sd): Likewise.
8281 (_mm_cvtss_sh): Likewise.
8282 (_mm_mask_cvtss_sh): Likewise.
8283 (_mm_maskz_cvtss_sh): Likewise.
8284 (_mm_cvtsd_sh): Likewise.
8285 (_mm_mask_cvtsd_sh): Likewise.
8286 (_mm_maskz_cvtsd_sh): Likewise.
8287 (_mm_cvt_roundss_sh): Likewise.
8288 (_mm_mask_cvt_roundss_sh): Likewise.
8289 (_mm_maskz_cvt_roundss_sh): Likewise.
8290 (_mm_cvt_roundsd_sh): Likewise.
8291 (_mm_mask_cvt_roundsd_sh): Likewise.
8292 (_mm_maskz_cvt_roundsd_sh): Likewise.
8293 * config/i386/i386-builtin-types.def
8294 (V8HF_FTYPE_V2DF_V8HF_V8HF_UQI_INT,
8295 V8HF_FTYPE_V4SF_V8HF_V8HF_UQI_INT,
8296 V2DF_FTYPE_V8HF_V2DF_V2DF_UQI_INT,
8297 V4SF_FTYPE_V8HF_V4SF_V4SF_UQI_INT): Add new builtin types.
8298 * config/i386/i386-builtin.def: Add corrresponding new builtins.
8299 * config/i386/i386-expand.c: Handle new builtin types.
8300 * config/i386/sse.md (VF48_128): New mode iterator.
8301 (avx512fp16_vcvtsh2<ssescalarmodesuffix><mask_scalar_name><round_saeonly_scalar_name>):
8303 (avx512fp16_vcvt<ssescalarmodesuffix>2sh<mask_scalar_name><round_scalar_name>):
8306 2021-09-17 liuhongt <hongtao.liu@intel.com>
8308 * config/i386/avx512fp16intrin.h (_mm512_cvtph_pd):
8310 (_mm512_mask_cvtph_pd): Likewise.
8311 (_mm512_maskz_cvtph_pd): Likewise.
8312 (_mm512_cvt_roundph_pd): Likewise.
8313 (_mm512_mask_cvt_roundph_pd): Likewise.
8314 (_mm512_maskz_cvt_roundph_pd): Likewise.
8315 (_mm512_cvtxph_ps): Likewise.
8316 (_mm512_mask_cvtxph_ps): Likewise.
8317 (_mm512_maskz_cvtxph_ps): Likewise.
8318 (_mm512_cvtx_roundph_ps): Likewise.
8319 (_mm512_mask_cvtx_roundph_ps): Likewise.
8320 (_mm512_maskz_cvtx_roundph_ps): Likewise.
8321 (_mm512_cvtxps_ph): Likewise.
8322 (_mm512_mask_cvtxps_ph): Likewise.
8323 (_mm512_maskz_cvtxps_ph): Likewise.
8324 (_mm512_cvtx_roundps_ph): Likewise.
8325 (_mm512_mask_cvtx_roundps_ph): Likewise.
8326 (_mm512_maskz_cvtx_roundps_ph): Likewise.
8327 (_mm512_cvtpd_ph): Likewise.
8328 (_mm512_mask_cvtpd_ph): Likewise.
8329 (_mm512_maskz_cvtpd_ph): Likewise.
8330 (_mm512_cvt_roundpd_ph): Likewise.
8331 (_mm512_mask_cvt_roundpd_ph): Likewise.
8332 (_mm512_maskz_cvt_roundpd_ph): Likewise.
8333 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_pd):
8335 (_mm_mask_cvtph_pd): Likewise.
8336 (_mm_maskz_cvtph_pd): Likewise.
8337 (_mm256_cvtph_pd): Likewise.
8338 (_mm256_mask_cvtph_pd): Likewise.
8339 (_mm256_maskz_cvtph_pd): Likewise.
8340 (_mm_cvtxph_ps): Likewise.
8341 (_mm_mask_cvtxph_ps): Likewise.
8342 (_mm_maskz_cvtxph_ps): Likewise.
8343 (_mm256_cvtxph_ps): Likewise.
8344 (_mm256_mask_cvtxph_ps): Likewise.
8345 (_mm256_maskz_cvtxph_ps): Likewise.
8346 (_mm_cvtxps_ph): Likewise.
8347 (_mm_mask_cvtxps_ph): Likewise.
8348 (_mm_maskz_cvtxps_ph): Likewise.
8349 (_mm256_cvtxps_ph): Likewise.
8350 (_mm256_mask_cvtxps_ph): Likewise.
8351 (_mm256_maskz_cvtxps_ph): Likewise.
8352 (_mm_cvtpd_ph): Likewise.
8353 (_mm_mask_cvtpd_ph): Likewise.
8354 (_mm_maskz_cvtpd_ph): Likewise.
8355 (_mm256_cvtpd_ph): Likewise.
8356 (_mm256_mask_cvtpd_ph): Likewise.
8357 (_mm256_maskz_cvtpd_ph): Likewise.
8358 * config/i386/i386-builtin.def: Add corresponding new builtins.
8359 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
8360 * config/i386/i386-expand.c: Handle new builtin types.
8361 * config/i386/sse.md
8362 (VF4_128_8_256): New.
8363 (VF48H_AVX512VL): Ditto.
8364 (ssePHmode): Add HF vector modes.
8365 (castmode): Add new convertable modes.
8368 (avx512fp16_vcvt<castmode>2ph_<mode><mask_name><round_name>): Ditto.
8369 (avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
8370 (*avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
8371 (avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
8372 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
8373 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask_1): Ditto.
8374 (avx512fp16_float_extend_ph<mode>2<mask_name><round_saeonly_name>):
8376 (avx512fp16_float_extend_ph<mode>2<mask_name>): Ditto.
8377 (*avx512fp16_float_extend_ph<mode>2_load<mask_name>): Ditto.
8378 (avx512fp16_float_extend_phv2df2<mask_name>): Ditto.
8379 (*avx512fp16_float_extend_phv2df2_load<mask_name>): Ditto.
8381 2021-09-17 liuhongt <hongtao.liu@intel.com>
8383 * config/i386/avx512fp16intrin.h (_mm_cvttsh_i32):
8385 (_mm_cvttsh_u32): Likewise.
8386 (_mm_cvtt_roundsh_i32): Likewise.
8387 (_mm_cvtt_roundsh_u32): Likewise.
8388 (_mm_cvttsh_i64): Likewise.
8389 (_mm_cvttsh_u64): Likewise.
8390 (_mm_cvtt_roundsh_i64): Likewise.
8391 (_mm_cvtt_roundsh_u64): Likewise.
8392 * config/i386/i386-builtin.def: Add corresponding new builtins.
8393 * config/i386/sse.md
8394 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<round_saeonly_name>):
8397 2021-09-17 liuhongt <hongtao.liu@intel.com>
8399 * config/i386/avx512fp16intrin.h (_mm512_cvttph_epi32):
8401 (_mm512_mask_cvttph_epi32): Likewise.
8402 (_mm512_maskz_cvttph_epi32): Likewise.
8403 (_mm512_cvtt_roundph_epi32): Likewise.
8404 (_mm512_mask_cvtt_roundph_epi32): Likewise.
8405 (_mm512_maskz_cvtt_roundph_epi32): Likewise.
8406 (_mm512_cvttph_epu32): Likewise.
8407 (_mm512_mask_cvttph_epu32): Likewise.
8408 (_mm512_maskz_cvttph_epu32): Likewise.
8409 (_mm512_cvtt_roundph_epu32): Likewise.
8410 (_mm512_mask_cvtt_roundph_epu32): Likewise.
8411 (_mm512_maskz_cvtt_roundph_epu32): Likewise.
8412 (_mm512_cvttph_epi64): Likewise.
8413 (_mm512_mask_cvttph_epi64): Likewise.
8414 (_mm512_maskz_cvttph_epi64): Likewise.
8415 (_mm512_cvtt_roundph_epi64): Likewise.
8416 (_mm512_mask_cvtt_roundph_epi64): Likewise.
8417 (_mm512_maskz_cvtt_roundph_epi64): Likewise.
8418 (_mm512_cvttph_epu64): Likewise.
8419 (_mm512_mask_cvttph_epu64): Likewise.
8420 (_mm512_maskz_cvttph_epu64): Likewise.
8421 (_mm512_cvtt_roundph_epu64): Likewise.
8422 (_mm512_mask_cvtt_roundph_epu64): Likewise.
8423 (_mm512_maskz_cvtt_roundph_epu64): Likewise.
8424 (_mm512_cvttph_epi16): Likewise.
8425 (_mm512_mask_cvttph_epi16): Likewise.
8426 (_mm512_maskz_cvttph_epi16): Likewise.
8427 (_mm512_cvtt_roundph_epi16): Likewise.
8428 (_mm512_mask_cvtt_roundph_epi16): Likewise.
8429 (_mm512_maskz_cvtt_roundph_epi16): Likewise.
8430 (_mm512_cvttph_epu16): Likewise.
8431 (_mm512_mask_cvttph_epu16): Likewise.
8432 (_mm512_maskz_cvttph_epu16): Likewise.
8433 (_mm512_cvtt_roundph_epu16): Likewise.
8434 (_mm512_mask_cvtt_roundph_epu16): Likewise.
8435 (_mm512_maskz_cvtt_roundph_epu16): Likewise.
8436 * config/i386/avx512fp16vlintrin.h (_mm_cvttph_epi32):
8438 (_mm_mask_cvttph_epi32): Likewise.
8439 (_mm_maskz_cvttph_epi32): Likewise.
8440 (_mm256_cvttph_epi32): Likewise.
8441 (_mm256_mask_cvttph_epi32): Likewise.
8442 (_mm256_maskz_cvttph_epi32): Likewise.
8443 (_mm_cvttph_epu32): Likewise.
8444 (_mm_mask_cvttph_epu32): Likewise.
8445 (_mm_maskz_cvttph_epu32): Likewise.
8446 (_mm256_cvttph_epu32): Likewise.
8447 (_mm256_mask_cvttph_epu32): Likewise.
8448 (_mm256_maskz_cvttph_epu32): Likewise.
8449 (_mm_cvttph_epi64): Likewise.
8450 (_mm_mask_cvttph_epi64): Likewise.
8451 (_mm_maskz_cvttph_epi64): Likewise.
8452 (_mm256_cvttph_epi64): Likewise.
8453 (_mm256_mask_cvttph_epi64): Likewise.
8454 (_mm256_maskz_cvttph_epi64): Likewise.
8455 (_mm_cvttph_epu64): Likewise.
8456 (_mm_mask_cvttph_epu64): Likewise.
8457 (_mm_maskz_cvttph_epu64): Likewise.
8458 (_mm256_cvttph_epu64): Likewise.
8459 (_mm256_mask_cvttph_epu64): Likewise.
8460 (_mm256_maskz_cvttph_epu64): Likewise.
8461 (_mm_cvttph_epi16): Likewise.
8462 (_mm_mask_cvttph_epi16): Likewise.
8463 (_mm_maskz_cvttph_epi16): Likewise.
8464 (_mm256_cvttph_epi16): Likewise.
8465 (_mm256_mask_cvttph_epi16): Likewise.
8466 (_mm256_maskz_cvttph_epi16): Likewise.
8467 (_mm_cvttph_epu16): Likewise.
8468 (_mm_mask_cvttph_epu16): Likewise.
8469 (_mm_maskz_cvttph_epu16): Likewise.
8470 (_mm256_cvttph_epu16): Likewise.
8471 (_mm256_mask_cvttph_epu16): Likewise.
8472 (_mm256_maskz_cvttph_epu16): Likewise.
8473 * config/i386/i386-builtin.def: Add new builtins.
8474 * config/i386/sse.md
8475 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name><round_saeonly_name>):
8477 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name>): Ditto.
8478 (*avx512fp16_fix<fixunssuffix>_trunc<mode>2_load<mask_name>): Ditto.
8479 (avx512fp16_fix<fixunssuffix>_truncv2di2<mask_name>): Ditto.
8480 (avx512fp16_fix<fixunssuffix>_truncv2di2_load<mask_name>): Ditto.
8482 2021-09-17 liuhongt <hongtao.liu@intel.com>
8484 * config/i386/avx512fp16intrin.h (_mm_cvtsh_i32): New intrinsic.
8485 (_mm_cvtsh_u32): Likewise.
8486 (_mm_cvt_roundsh_i32): Likewise.
8487 (_mm_cvt_roundsh_u32): Likewise.
8488 (_mm_cvtsh_i64): Likewise.
8489 (_mm_cvtsh_u64): Likewise.
8490 (_mm_cvt_roundsh_i64): Likewise.
8491 (_mm_cvt_roundsh_u64): Likewise.
8492 (_mm_cvti32_sh): Likewise.
8493 (_mm_cvtu32_sh): Likewise.
8494 (_mm_cvt_roundi32_sh): Likewise.
8495 (_mm_cvt_roundu32_sh): Likewise.
8496 (_mm_cvti64_sh): Likewise.
8497 (_mm_cvtu64_sh): Likewise.
8498 (_mm_cvt_roundi64_sh): Likewise.
8499 (_mm_cvt_roundu64_sh): Likewise.
8500 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
8501 * config/i386/i386-builtin.def: Add corresponding new builtins.
8502 * config/i386/i386-expand.c (ix86_expand_round_builtin):
8503 Handle new builtin types.
8504 * config/i386/sse.md
8505 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix><round_name>):
8507 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix>_2): Likewise.
8508 (avx512fp16_vcvt<floatsuffix>si2sh<rex64namesuffix><round_name>): Likewise.
8510 2021-09-16 Bill Schmidt <wschmidt@linux.ibm.com>
8512 * config/rs6000/rs6000-c.c (rs6000-builtins.h): New include.
8513 (altivec_resolve_new_overloaded_builtin): New forward decl.
8514 (rs6000_new_builtin_type_compatible): New function.
8515 (altivec_resolve_overloaded_builtin): Call
8516 altivec_resolve_new_overloaded_builtin.
8517 (altivec_build_new_resolved_builtin): New function.
8518 (altivec_resolve_new_overloaded_builtin): Likewise.
8519 * config/rs6000/rs6000-call.c (rs6000_new_builtin_is_supported):
8521 * config/rs6000/rs6000-gen-builtins.c (write_decls): Remove _p from
8522 name of rs6000_new_builtin_is_supported.
8524 2021-09-16 Uroš Bizjak <ubizjak@gmail.com>
8526 * config/i386/i386-protos.h (ix86_decompose_address):
8527 Change return type to bool.
8528 * config/i386/i386.c (ix86_decompose_address): Ditto.
8530 2021-09-16 Tobias Burnus <tobias@codesourcery.com>
8533 * config/rs6000/t-rs6000 (build/rs6000-gen-builtins.o, build/rbtree.o):
8534 Added 'build/' to target, use build/%.o rule.
8535 (build/rs6000-gen-builtins$(build_exeext)): Add 'build/' and
8536 '$(build_exeext)' to target and 'build/' for the *.o files.
8537 (rs6000-builtins.c): Update for those changes; run rs6000-gen-builtins
8540 2021-09-16 Martin Jambor <mjambor@suse.cz>
8542 * cgraph.c (cgraph_node::dump): Do not check caller count sums if
8543 the body has been removed. Remove trailing whitespace.
8545 2021-09-16 Richard Biener <rguenther@suse.de>
8547 PR middle-end/102360
8548 * internal-fn.c (expand_DEFERRED_INIT): Make pattern-init
8549 of non-memory more robust.
8551 2021-09-16 Daniel Cederman <cederman@gaisler.com>
8553 * config/sparc/sparc-opts.h (enum sparc_processor_type): Add LEON5
8554 * config/sparc/sparc.c (struct processor_costs): Add LEON5 costs
8555 (leon5_adjust_cost): Increase cost of store with data dependency
8556 on ALU instruction and FPU anti-dependencies.
8557 (sparc_option_override): Add LEON5 costs
8558 (sparc_adjust_cost): Add LEON5 cost adjustments
8559 * config/sparc/sparc.h: Add LEON5
8560 * config/sparc/sparc.md: Include LEON5 scheduling information
8561 * config/sparc/sparc.opt: Add LEON5
8562 * doc/invoke.texi: Add LEON5
8563 * config/sparc/leon5.md: New file.
8565 2021-09-16 Daniel Cederman <cederman@gaisler.com>
8567 * config/sparc/sparc.md (stack_protect_set32): Add NOP to prevent
8568 sensitive sequence for B2BST errata workaround.
8570 2021-09-16 Daniel Cederman <cederman@gaisler.com>
8572 * config/sparc/sparc.c (sparc_do_work_around_errata): Do not begin
8573 functions with atomic instruction in the UT700 errata workaround.
8575 2021-09-16 Daniel Cederman <cederman@gaisler.com>
8577 * config/sparc/sparc.c (next_active_non_empty_insn): New function
8578 that returns next active non empty assembly instruction.
8579 (sparc_do_work_around_errata): Use new function.
8581 2021-09-16 Daniel Cederman <cederman@gaisler.com>
8583 * config/sparc/sparc.c (store_insn_p): Add predicate for store
8585 (load_insn_p): Add predicate for load attributes.
8586 (sparc_do_work_around_errata): Use new predicates.
8588 2021-09-16 Andreas Larsson <andreas@gaisler.com>
8590 * config/sparc/sparc.c (dump_target_flag_bits): Print bit names for
8593 2021-09-16 Martin Liska <mliska@suse.cz>
8595 * config/mips/netbsd.h: Fix typo in name of a macro.
8597 2021-09-16 liuhongt <hongtao.liu@intel.com>
8599 PR middle-end/102080
8600 * match.pd: Check mask type when doing cond_op related gimple
8602 * tree.c (is_truth_type_for): New function.
8603 * tree.h (is_truth_type_for): New declaration.
8605 2021-09-16 liuhongt <hongtao.liu@intel.com>
8607 * config/i386/avx512fp16intrin.h (_mm512_cvtepi32_ph): New
8609 (_mm512_mask_cvtepi32_ph): Likewise.
8610 (_mm512_maskz_cvtepi32_ph): Likewise.
8611 (_mm512_cvt_roundepi32_ph): Likewise.
8612 (_mm512_mask_cvt_roundepi32_ph): Likewise.
8613 (_mm512_maskz_cvt_roundepi32_ph): Likewise.
8614 (_mm512_cvtepu32_ph): Likewise.
8615 (_mm512_mask_cvtepu32_ph): Likewise.
8616 (_mm512_maskz_cvtepu32_ph): Likewise.
8617 (_mm512_cvt_roundepu32_ph): Likewise.
8618 (_mm512_mask_cvt_roundepu32_ph): Likewise.
8619 (_mm512_maskz_cvt_roundepu32_ph): Likewise.
8620 (_mm512_cvtepi64_ph): Likewise.
8621 (_mm512_mask_cvtepi64_ph): Likewise.
8622 (_mm512_maskz_cvtepi64_ph): Likewise.
8623 (_mm512_cvt_roundepi64_ph): Likewise.
8624 (_mm512_mask_cvt_roundepi64_ph): Likewise.
8625 (_mm512_maskz_cvt_roundepi64_ph): Likewise.
8626 (_mm512_cvtepu64_ph): Likewise.
8627 (_mm512_mask_cvtepu64_ph): Likewise.
8628 (_mm512_maskz_cvtepu64_ph): Likewise.
8629 (_mm512_cvt_roundepu64_ph): Likewise.
8630 (_mm512_mask_cvt_roundepu64_ph): Likewise.
8631 (_mm512_maskz_cvt_roundepu64_ph): Likewise.
8632 (_mm512_cvtepi16_ph): Likewise.
8633 (_mm512_mask_cvtepi16_ph): Likewise.
8634 (_mm512_maskz_cvtepi16_ph): Likewise.
8635 (_mm512_cvt_roundepi16_ph): Likewise.
8636 (_mm512_mask_cvt_roundepi16_ph): Likewise.
8637 (_mm512_maskz_cvt_roundepi16_ph): Likewise.
8638 (_mm512_cvtepu16_ph): Likewise.
8639 (_mm512_mask_cvtepu16_ph): Likewise.
8640 (_mm512_maskz_cvtepu16_ph): Likewise.
8641 (_mm512_cvt_roundepu16_ph): Likewise.
8642 (_mm512_mask_cvt_roundepu16_ph): Likewise.
8643 (_mm512_maskz_cvt_roundepu16_ph): Likewise.
8644 * config/i386/avx512fp16vlintrin.h (_mm_cvtepi32_ph): New
8646 (_mm_mask_cvtepi32_ph): Likewise.
8647 (_mm_maskz_cvtepi32_ph): Likewise.
8648 (_mm256_cvtepi32_ph): Likewise.
8649 (_mm256_mask_cvtepi32_ph): Likewise.
8650 (_mm256_maskz_cvtepi32_ph): Likewise.
8651 (_mm_cvtepu32_ph): Likewise.
8652 (_mm_mask_cvtepu32_ph): Likewise.
8653 (_mm_maskz_cvtepu32_ph): Likewise.
8654 (_mm256_cvtepu32_ph): Likewise.
8655 (_mm256_mask_cvtepu32_ph): Likewise.
8656 (_mm256_maskz_cvtepu32_ph): Likewise.
8657 (_mm_cvtepi64_ph): Likewise.
8658 (_mm_mask_cvtepi64_ph): Likewise.
8659 (_mm_maskz_cvtepi64_ph): Likewise.
8660 (_mm256_cvtepi64_ph): Likewise.
8661 (_mm256_mask_cvtepi64_ph): Likewise.
8662 (_mm256_maskz_cvtepi64_ph): Likewise.
8663 (_mm_cvtepu64_ph): Likewise.
8664 (_mm_mask_cvtepu64_ph): Likewise.
8665 (_mm_maskz_cvtepu64_ph): Likewise.
8666 (_mm256_cvtepu64_ph): Likewise.
8667 (_mm256_mask_cvtepu64_ph): Likewise.
8668 (_mm256_maskz_cvtepu64_ph): Likewise.
8669 (_mm_cvtepi16_ph): Likewise.
8670 (_mm_mask_cvtepi16_ph): Likewise.
8671 (_mm_maskz_cvtepi16_ph): Likewise.
8672 (_mm256_cvtepi16_ph): Likewise.
8673 (_mm256_mask_cvtepi16_ph): Likewise.
8674 (_mm256_maskz_cvtepi16_ph): Likewise.
8675 (_mm_cvtepu16_ph): Likewise.
8676 (_mm_mask_cvtepu16_ph): Likewise.
8677 (_mm_maskz_cvtepu16_ph): Likewise.
8678 (_mm256_cvtepu16_ph): Likewise.
8679 (_mm256_mask_cvtepu16_ph): Likewise.
8680 (_mm256_maskz_cvtepu16_ph): Likewise.
8681 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
8682 * config/i386/i386-builtin.def: Add corresponding new builtins.
8683 * config/i386/i386-expand.c
8684 (ix86_expand_args_builtin): Handle new builtin types.
8685 (ix86_expand_round_builtin): Ditto.
8686 * config/i386/i386-modes.def: Declare V2HF and V6HF.
8687 * config/i386/sse.md (VI2H_AVX512VL): New.
8689 (sseintvecmode): Add HF vector modes.
8690 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode><mask_name><round_name>):
8692 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
8693 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
8694 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
8695 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
8696 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask_1): Ditto.
8697 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
8698 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
8699 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
8700 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
8701 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask_1): Ditto.
8702 * config/i386/subst.md (round_qq2phsuff): New subst_attr.
8704 2021-09-16 liuhongt <hongtao.liu@intel.com>
8706 * config/i386/avx512fp16intrin.h (_mm512_cvtph_epi32):
8708 (_mm512_mask_cvtph_epi32): Likewise.
8709 (_mm512_maskz_cvtph_epi32): Likewise.
8710 (_mm512_cvt_roundph_epi32): Likewise.
8711 (_mm512_mask_cvt_roundph_epi32): Likewise.
8712 (_mm512_maskz_cvt_roundph_epi32): Likewise.
8713 (_mm512_cvtph_epu32): Likewise.
8714 (_mm512_mask_cvtph_epu32): Likewise.
8715 (_mm512_maskz_cvtph_epu32): Likewise.
8716 (_mm512_cvt_roundph_epu32): Likewise.
8717 (_mm512_mask_cvt_roundph_epu32): Likewise.
8718 (_mm512_maskz_cvt_roundph_epu32): Likewise.
8719 (_mm512_cvtph_epi64): Likewise.
8720 (_mm512_mask_cvtph_epi64): Likewise.
8721 (_mm512_maskz_cvtph_epi64): Likewise.
8722 (_mm512_cvt_roundph_epi64): Likewise.
8723 (_mm512_mask_cvt_roundph_epi64): Likewise.
8724 (_mm512_maskz_cvt_roundph_epi64): Likewise.
8725 (_mm512_cvtph_epu64): Likewise.
8726 (_mm512_mask_cvtph_epu64): Likewise.
8727 (_mm512_maskz_cvtph_epu64): Likewise.
8728 (_mm512_cvt_roundph_epu64): Likewise.
8729 (_mm512_mask_cvt_roundph_epu64): Likewise.
8730 (_mm512_maskz_cvt_roundph_epu64): Likewise.
8731 (_mm512_cvtph_epi16): Likewise.
8732 (_mm512_mask_cvtph_epi16): Likewise.
8733 (_mm512_maskz_cvtph_epi16): Likewise.
8734 (_mm512_cvt_roundph_epi16): Likewise.
8735 (_mm512_mask_cvt_roundph_epi16): Likewise.
8736 (_mm512_maskz_cvt_roundph_epi16): Likewise.
8737 (_mm512_cvtph_epu16): Likewise.
8738 (_mm512_mask_cvtph_epu16): Likewise.
8739 (_mm512_maskz_cvtph_epu16): Likewise.
8740 (_mm512_cvt_roundph_epu16): Likewise.
8741 (_mm512_mask_cvt_roundph_epu16): Likewise.
8742 (_mm512_maskz_cvt_roundph_epu16): Likewise.
8743 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_epi32):
8745 (_mm_mask_cvtph_epi32): Likewise.
8746 (_mm_maskz_cvtph_epi32): Likewise.
8747 (_mm256_cvtph_epi32): Likewise.
8748 (_mm256_mask_cvtph_epi32): Likewise.
8749 (_mm256_maskz_cvtph_epi32): Likewise.
8750 (_mm_cvtph_epu32): Likewise.
8751 (_mm_mask_cvtph_epu32): Likewise.
8752 (_mm_maskz_cvtph_epu32): Likewise.
8753 (_mm256_cvtph_epu32): Likewise.
8754 (_mm256_mask_cvtph_epu32): Likewise.
8755 (_mm256_maskz_cvtph_epu32): Likewise.
8756 (_mm_cvtph_epi64): Likewise.
8757 (_mm_mask_cvtph_epi64): Likewise.
8758 (_mm_maskz_cvtph_epi64): Likewise.
8759 (_mm256_cvtph_epi64): Likewise.
8760 (_mm256_mask_cvtph_epi64): Likewise.
8761 (_mm256_maskz_cvtph_epi64): Likewise.
8762 (_mm_cvtph_epu64): Likewise.
8763 (_mm_mask_cvtph_epu64): Likewise.
8764 (_mm_maskz_cvtph_epu64): Likewise.
8765 (_mm256_cvtph_epu64): Likewise.
8766 (_mm256_mask_cvtph_epu64): Likewise.
8767 (_mm256_maskz_cvtph_epu64): Likewise.
8768 (_mm_cvtph_epi16): Likewise.
8769 (_mm_mask_cvtph_epi16): Likewise.
8770 (_mm_maskz_cvtph_epi16): Likewise.
8771 (_mm256_cvtph_epi16): Likewise.
8772 (_mm256_mask_cvtph_epi16): Likewise.
8773 (_mm256_maskz_cvtph_epi16): Likewise.
8774 (_mm_cvtph_epu16): Likewise.
8775 (_mm_mask_cvtph_epu16): Likewise.
8776 (_mm_maskz_cvtph_epu16): Likewise.
8777 (_mm256_cvtph_epu16): Likewise.
8778 (_mm256_mask_cvtph_epu16): Likewise.
8779 (_mm256_maskz_cvtph_epu16): Likewise.
8780 * config/i386/i386-builtin-types.def: Add new builtin types.
8781 * config/i386/i386-builtin.def: Add new builtins.
8782 * config/i386/i386-expand.c
8783 (ix86_expand_args_builtin): Handle new builtin types.
8784 (ix86_expand_round_builtin): Ditto.
8785 * config/i386/sse.md (sseintconvert): New.
8787 (UNSPEC_US_FIX_NOTRUNC): Ditto.
8788 (sseintconvertsignprefix): Ditto.
8789 (avx512fp16_vcvtph2<sseintconvertsignprefix><sseintconvert>_<mode><mask_name><round_name>):
8792 2021-09-16 liuhongt <hongtao.liu@intel.com>
8794 * config/i386/avx512fp16intrin.h: (_mm_cvtsi16_si128):
8796 (_mm_cvtsi128_si16): Likewise.
8797 (_mm_mask_load_sh): Likewise.
8798 (_mm_maskz_load_sh): Likewise.
8799 (_mm_mask_store_sh): Likewise.
8800 (_mm_move_sh): Likewise.
8801 (_mm_mask_move_sh): Likewise.
8802 (_mm_maskz_move_sh): Likewise.
8803 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
8804 * config/i386/i386-builtin.def: Add corresponding new builtins.
8805 * config/i386/i386-expand.c
8806 (ix86_expand_special_args_builtin): Handle new builtin types.
8807 (ix86_expand_vector_init_one_nonzero): Adjust for FP16 target.
8808 * config/i386/sse.md (VI2F): New mode iterator.
8809 (vec_set<mode>_0): Use new mode iterator.
8810 (avx512f_mov<ssescalarmodelower>_mask): Adjust for HF vector mode.
8811 (avx512f_store<mode>_mask): Ditto.
8813 2021-09-16 Kewen Lin <linkw@linux.ibm.com>
8815 * config/rs6000/rs6000.opt (-mtoc-fusion): Remove.
8817 2021-09-15 David Edelsohn <dje.gcc@gmail.com>
8819 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
8820 Proceed if no symbol summary or the symbol alias flag is false.
8822 2021-09-15 Jakub Jelinek <jakub@redhat.com>
8826 * varasm.c (output_constructor_regular_field): Instead of assertion
8827 that array_size_for_constructor result is equal to size of
8828 TREE_TYPE (local->val) in bytes, assert that the type size is greater
8829 or equal to array_size_for_constructor result and use type size as
8832 2021-09-15 Martin Liska <mliska@suse.cz>
8835 * config/i386/vxworks.h: Use new macro TARGET_CPU_P.
8837 2021-09-15 Martin Liska <mliska@suse.cz>
8840 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
8841 Check that we have a symbol summary for a symbol.
8843 2021-09-15 Richard Biener <rguenther@suse.de>
8846 * config/rs6000/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
8847 to inherit from elfos.h
8849 2021-09-15 liuhongt <hongtao.liu@intel.com>
8852 * config/i386/i386-expand.c
8853 (ix86_expand_vector_init_interleave): Use puncklwd to pack 2
8855 (ix86_expand_vector_set): Use blendw instead of pinsrw.
8856 * config/i386/i386.c (ix86_can_change_mode_class): Adjust for
8857 AVX512FP16 which supports 16bit vector load.
8858 * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>):
8860 (avx512bw_interleave_high<mode><mask_name>): .. this, and
8861 extend to V32HFmode.
8862 (avx2_interleave_highv16hi<mask_name>): Rename to ..
8863 (avx2_interleave_high<mode><mask_name>): .. this, and extend
8865 (vec_interleave_highv8hi<mask_name>): Rename to ..
8866 (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode.
8867 (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>):
8869 (<mask_codefor>avx512bw_interleave_low<mode><mask_name>):
8870 this, and extend to V32HFmode.
8871 (avx2_interleave_lowv16hi<mask_name>): Rename to ..
8872 (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode.
8873 (vec_interleave_lowv8hi<mask_name>): Rename to ..
8874 (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode.
8875 (sse4_1_pblendw): Rename to ..
8876 (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode.
8877 (avx2_pblendph): New define_expand.
8878 (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use
8879 sseintmodesuffix instead of ssemodesuffix.
8880 (blendsuf): New mode attr.
8882 2021-09-15 Richard Biener <rguenther@suse.de>
8884 * tree-vectorizer.h (dr_misalignment): Move out of line.
8885 (dr_target_alignment): New.
8886 (DR_TARGET_ALIGNMENT): Wrap dr_target_alignment.
8887 (set_dr_target_alignment): New.
8888 (SET_DR_TARGET_ALIGNMENT): Wrap set_dr_target_alignment.
8889 * tree-vect-data-refs.c (dr_misalignment): Compute and
8890 return the group members misalignment.
8891 (vect_compute_data_ref_alignment): Use SET_DR_TARGET_ALIGNMENT.
8892 (vect_analyze_data_refs_alignment): Compute alignment only
8893 for the first element of a DR group.
8894 (vect_slp_analyze_node_alignment): Likewise.
8896 2021-09-15 Hongyu Wang <hongyu.wang@intel.com>
8898 * config/i386/avx512fp16intrin.h: Adjust all builtin calls.
8899 * config/i386/avx512fp16vlintrin.h: Likewise.
8900 * config/i386/i386-builtin.def: Adjust builtin name and
8901 enumeration to match AVX512F style.
8903 2021-09-15 Richard Biener <rguenther@suse.de>
8905 PR tree-optimization/102318
8906 * tree-vect-loop.c (vect_transform_cycle_phi): Revert
8907 previous change and do the mode conversion separately from
8908 the sign conversion.
8910 2021-09-15 Hongtao Liu <hongtao.liu@intel.com>
8911 Peter Cordes <peter@cordes.ca>
8914 * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
8915 (*vec_extract<mode><ssescalarmodelower>_valign): Output
8916 vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
8919 2021-09-15 Richard Biener <rguenther@suse.de>
8921 * config.gcc: Remove vax-*-openbsd* configuration.
8923 2021-09-15 Richard Biener <rguenther@suse.de>
8925 * config.gcc: Remove m68k-openbsd.
8927 2021-09-15 Max Filippov <jcmvbkbc@gmail.com>
8930 * config/xtensa/t-xtensa (TM_H): Add include/xtensa-config.h.
8932 2021-09-14 Peter Bergner <bergner@linux.ibm.com>
8934 * config/rs6000/mma.md (unspec): Delete UNSPEC_MMA_XXSETACCZ.
8935 (unspecv): Add UNSPECV_MMA_XXSETACCZ.
8936 (*mma_xxsetaccz): Delete.
8937 (mma_xxsetaccz): Change to define_insn. Remove operand 1.
8938 Use UNSPECV_MMA_XXSETACCZ. Update comment.
8939 * config/rs6000/rs6000.c (rs6000_rtx_costs): Use UNSPECV_MMA_XXSETACCZ.
8941 2021-09-14 Iain Sandoe <iain@sandoe.co.uk>
8943 * Makefile.in: Remove variables related to applying no-PIE
8944 to the exes on $build.
8945 * configure: Regenerate.
8946 * configure.ac: Remove configuration related to applying
8947 no-PIE to the exes on $build.
8949 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
8951 * config/arc/arc.md (doloop_end): Add missing mode.
8952 (loop_end): Likewise.
8954 2021-09-14 Jakub Jelinek <jakub@redhat.com>
8956 * gimplify.c (goa_stabilize_expr): Add depth argument, propagate
8957 it to recursive calls, for depth above 7 just gimplify or return.
8958 Perform a test even for MODIFY_EXPR, ADDR_EXPR, COMPOUND_EXPR with
8959 __builtin_clear_padding and TARGET_EXPR.
8960 (gimplify_omp_atomic): Adjust goa_stabilize_expr callers.
8962 2021-09-14 liuhongt <hongtao.liu@intel.com>
8964 * config/i386/avx512fp16intrin.h (_mm_fpclass_sh_mask):
8966 (_mm_mask_fpclass_sh_mask): Likewise.
8967 (_mm512_mask_fpclass_ph_mask): Likewise.
8968 (_mm512_fpclass_ph_mask): Likewise.
8969 (_mm_getexp_sh): Likewise.
8970 (_mm_mask_getexp_sh): Likewise.
8971 (_mm_maskz_getexp_sh): Likewise.
8972 (_mm512_getexp_ph): Likewise.
8973 (_mm512_mask_getexp_ph): Likewise.
8974 (_mm512_maskz_getexp_ph): Likewise.
8975 (_mm_getexp_round_sh): Likewise.
8976 (_mm_mask_getexp_round_sh): Likewise.
8977 (_mm_maskz_getexp_round_sh): Likewise.
8978 (_mm512_getexp_round_ph): Likewise.
8979 (_mm512_mask_getexp_round_ph): Likewise.
8980 (_mm512_maskz_getexp_round_ph): Likewise.
8981 (_mm_getmant_sh): Likewise.
8982 (_mm_mask_getmant_sh): Likewise.
8983 (_mm_maskz_getmant_sh): Likewise.
8984 (_mm512_getmant_ph): Likewise.
8985 (_mm512_mask_getmant_ph): Likewise.
8986 (_mm512_maskz_getmant_ph): Likewise.
8987 (_mm_getmant_round_sh): Likewise.
8988 (_mm_mask_getmant_round_sh): Likewise.
8989 (_mm_maskz_getmant_round_sh): Likewise.
8990 (_mm512_getmant_round_ph): Likewise.
8991 (_mm512_mask_getmant_round_ph): Likewise.
8992 (_mm512_maskz_getmant_round_ph): Likewise.
8993 * config/i386/avx512fp16vlintrin.h (_mm_mask_fpclass_ph_mask):
8995 (_mm_fpclass_ph_mask): Likewise.
8996 (_mm256_mask_fpclass_ph_mask): Likewise.
8997 (_mm256_fpclass_ph_mask): Likewise.
8998 (_mm256_getexp_ph): Likewise.
8999 (_mm256_mask_getexp_ph): Likewise.
9000 (_mm256_maskz_getexp_ph): Likewise.
9001 (_mm_getexp_ph): Likewise.
9002 (_mm_mask_getexp_ph): Likewise.
9003 (_mm_maskz_getexp_ph): Likewise.
9004 (_mm256_getmant_ph): Likewise.
9005 (_mm256_mask_getmant_ph): Likewise.
9006 (_mm256_maskz_getmant_ph): Likewise.
9007 (_mm_getmant_ph): Likewise.
9008 (_mm_mask_getmant_ph): Likewise.
9009 (_mm_maskz_getmant_ph): Likewise.
9010 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
9011 * config/i386/i386-builtin.def: Add corresponding new builtins.
9012 * config/i386/i386-expand.c
9013 (ix86_expand_args_builtin): Handle new builtin types.
9014 (ix86_expand_round_builtin): Ditto.
9015 * config/i386/sse.md (vecmemsuffix): Add HF vector modes.
9016 (<avx512>_getexp<mode><mask_name><round_saeonly_name>): Adjust
9017 to support HF vector modes.
9018 (avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name):
9020 (avx512dq_fpclass<mode><mask_scalar_merge_name>): Ditto.
9021 (avx512dq_vmfpclass<mode><mask_scalar_merge_name>): Ditto.
9022 (<avx512>_getmant<mode><mask_name><round_saeonly_name>): Ditto.
9023 (avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
9026 2021-09-14 liuhongt <hongtao.liu@intel.com>
9028 * config/i386/avx512fp16intrin.h (_mm512_reduce_ph):
9030 (_mm512_mask_reduce_ph): Likewise.
9031 (_mm512_maskz_reduce_ph): Likewise.
9032 (_mm512_reduce_round_ph): Likewise.
9033 (_mm512_mask_reduce_round_ph): Likewise.
9034 (_mm512_maskz_reduce_round_ph): Likewise.
9035 (_mm_reduce_sh): Likewise.
9036 (_mm_mask_reduce_sh): Likewise.
9037 (_mm_maskz_reduce_sh): Likewise.
9038 (_mm_reduce_round_sh): Likewise.
9039 (_mm_mask_reduce_round_sh): Likewise.
9040 (_mm_maskz_reduce_round_sh): Likewise.
9041 (_mm512_roundscale_ph): Likewise.
9042 (_mm512_mask_roundscale_ph): Likewise.
9043 (_mm512_maskz_roundscale_ph): Likewise.
9044 (_mm512_roundscale_round_ph): Likewise.
9045 (_mm512_mask_roundscale_round_ph): Likewise.
9046 (_mm512_maskz_roundscale_round_ph): Likewise.
9047 (_mm_roundscale_sh): Likewise.
9048 (_mm_mask_roundscale_sh): Likewise.
9049 (_mm_maskz_roundscale_sh): Likewise.
9050 (_mm_roundscale_round_sh): Likewise.
9051 (_mm_mask_roundscale_round_sh): Likewise.
9052 (_mm_maskz_roundscale_round_sh): Likewise.
9053 * config/i386/avx512fp16vlintrin.h: (_mm_reduce_ph):
9055 (_mm_mask_reduce_ph): Likewise.
9056 (_mm_maskz_reduce_ph): Likewise.
9057 (_mm256_reduce_ph): Likewise.
9058 (_mm256_mask_reduce_ph): Likewise.
9059 (_mm256_maskz_reduce_ph): Likewise.
9060 (_mm_roundscale_ph): Likewise.
9061 (_mm_mask_roundscale_ph): Likewise.
9062 (_mm_maskz_roundscale_ph): Likewise.
9063 (_mm256_roundscale_ph): Likewise.
9064 (_mm256_mask_roundscale_ph): Likewise.
9065 (_mm256_maskz_roundscale_ph): Likewise.
9066 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
9067 * config/i386/i386-builtin.def: Add corresponding new builtins.
9068 * config/i386/i386-expand.c
9069 (ix86_expand_args_builtin): Handle new builtin types.
9070 (ix86_expand_round_builtin): Ditto.
9071 * config/i386/sse.md (<mask_codefor>reducep<mode><mask_name>):
9073 (<mask_codefor>reducep<mode><mask_name><round_saeonly_name>):
9074 ... this, and adjust for round operands.
9075 (reduces<mode><mask_scalar_name>): Likewise, with ...
9076 (reduces<mode><mask_scalar_name><round_saeonly_scalar_name):
9078 (<avx512>_rndscale<mode><mask_name><round_saeonly_name>):
9079 Adjust for HF vector modes.
9080 (avx512f_rndscale<mode><mask_scalar_name><round_saeonly_scalar_name>):
9082 (*avx512f_rndscale<mode><round_saeonly_name>): Ditto.
9084 2021-09-14 liuhongt <hongtao.liu@intel.com>
9086 * config/i386/avx512fp16intrin.h: (_mm512_rcp_ph):
9088 (_mm512_mask_rcp_ph): Likewise.
9089 (_mm512_maskz_rcp_ph): Likewise.
9090 (_mm_rcp_sh): Likewise.
9091 (_mm_mask_rcp_sh): Likewise.
9092 (_mm_maskz_rcp_sh): Likewise.
9093 (_mm512_scalef_ph): Likewise.
9094 (_mm512_mask_scalef_ph): Likewise.
9095 (_mm512_maskz_scalef_ph): Likewise.
9096 (_mm512_scalef_round_ph): Likewise.
9097 (_mm512_mask_scalef_round_ph): Likewise.
9098 (_mm512_maskz_scalef_round_ph): Likewise.
9099 (_mm_scalef_sh): Likewise.
9100 (_mm_mask_scalef_sh): Likewise.
9101 (_mm_maskz_scalef_sh): Likewise.
9102 (_mm_scalef_round_sh): Likewise.
9103 (_mm_mask_scalef_round_sh): Likewise.
9104 (_mm_maskz_scalef_round_sh): Likewise.
9105 * config/i386/avx512fp16vlintrin.h (_mm_rcp_ph):
9107 (_mm256_rcp_ph): Likewise.
9108 (_mm_mask_rcp_ph): Likewise.
9109 (_mm256_mask_rcp_ph): Likewise.
9110 (_mm_maskz_rcp_ph): Likewise.
9111 (_mm256_maskz_rcp_ph): Likewise.
9112 (_mm_scalef_ph): Likewise.
9113 (_mm256_scalef_ph): Likewise.
9114 (_mm_mask_scalef_ph): Likewise.
9115 (_mm256_mask_scalef_ph): Likewise.
9116 (_mm_maskz_scalef_ph): Likewise.
9117 (_mm256_maskz_scalef_ph): Likewise.
9118 * config/i386/i386-builtin.def: Add new builtins.
9119 * config/i386/sse.md (VFH_AVX512VL): New.
9120 (avx512fp16_rcp<mode>2<mask_name>): Ditto.
9121 (avx512fp16_vmrcpv8hf2<mask_scalar_name>): Ditto.
9122 (avx512f_vmscalef<mode><mask_scalar_name><round_scalar_name>):
9123 Adjust to support HF vector modes.
9124 (<avx512>_scalef<mode><mask_name><round_name>): Ditto.
9126 2021-09-14 liuhongt <hongtao.liu@intel.com>
9128 * config/i386/avx512fp16intrin.h: (_mm512_sqrt_ph):
9130 (_mm512_mask_sqrt_ph): Likewise.
9131 (_mm512_maskz_sqrt_ph): Likewise.
9132 (_mm512_sqrt_round_ph): Likewise.
9133 (_mm512_mask_sqrt_round_ph): Likewise.
9134 (_mm512_maskz_sqrt_round_ph): Likewise.
9135 (_mm512_rsqrt_ph): Likewise.
9136 (_mm512_mask_rsqrt_ph): Likewise.
9137 (_mm512_maskz_rsqrt_ph): Likewise.
9138 (_mm_rsqrt_sh): Likewise.
9139 (_mm_mask_rsqrt_sh): Likewise.
9140 (_mm_maskz_rsqrt_sh): Likewise.
9141 (_mm_sqrt_sh): Likewise.
9142 (_mm_mask_sqrt_sh): Likewise.
9143 (_mm_maskz_sqrt_sh): Likewise.
9144 (_mm_sqrt_round_sh): Likewise.
9145 (_mm_mask_sqrt_round_sh): Likewise.
9146 (_mm_maskz_sqrt_round_sh): Likewise.
9147 * config/i386/avx512fp16vlintrin.h (_mm_sqrt_ph): New intrinsic.
9148 (_mm256_sqrt_ph): Likewise.
9149 (_mm_mask_sqrt_ph): Likewise.
9150 (_mm256_mask_sqrt_ph): Likewise.
9151 (_mm_maskz_sqrt_ph): Likewise.
9152 (_mm256_maskz_sqrt_ph): Likewise.
9153 (_mm_rsqrt_ph): Likewise.
9154 (_mm256_rsqrt_ph): Likewise.
9155 (_mm_mask_rsqrt_ph): Likewise.
9156 (_mm256_mask_rsqrt_ph): Likewise.
9157 (_mm_maskz_rsqrt_ph): Likewise.
9158 (_mm256_maskz_rsqrt_ph): Likewise.
9159 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
9160 * config/i386/i386-builtin.def: Add corresponding new builtins.
9161 * config/i386/i386-expand.c
9162 (ix86_expand_args_builtin): Handle new builtins.
9163 (ix86_expand_round_builtin): Ditto.
9164 * config/i386/sse.md (VF_AVX512FP16VL): New.
9165 (sqrt<mode>2): Adjust for HF vector modes.
9166 (<sse>_sqrt<mode>2<mask_name><round_name>): Likewise.
9167 (<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
9169 (<sse>_rsqrt<mode>2<mask_name>): New.
9170 (avx512fp16_vmrsqrtv8hf2<mask_scalar_name>): Likewise.
9172 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
9175 * diagnostic-spec.c (warning_suppressed_at, copy_warning): Handle
9176 'RESERVED_LOCATION_P' locations.
9177 * warning-control.cc (get_nowarn_spec, suppress_warning)
9178 (copy_warning): Likewise.
9180 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
9182 * diagnostic-spec.h (typedef xint_hash_t): Use 'location_t' instead of...
9183 (typedef key_type_t): ... this. Remove.
9184 (nowarn_map): Document.
9185 * diagnostic-spec.c (nowarn_map): Likewise.
9186 * warning-control.cc (convert_to_key): Evolve functions into...
9187 (get_location): ... these. Adjust all users.
9189 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
9191 * warning-control.cc (copy_warning): Remove 'nowarn_map' setup.
9193 2021-09-13 Jason Merrill <jason@redhat.com>
9195 * params.opt: Add destructive-interference-size and
9196 constructive-interference-size.
9197 * doc/invoke.texi: Document them.
9198 * config/aarch64/aarch64.c (aarch64_override_options_internal):
9200 * config/arm/arm.c (arm_option_override): Set them.
9201 * config/i386/i386-options.c (ix86_option_override_internal):
9204 2021-09-13 Martin Liska <mliska@suse.cz>
9205 H.J. Lu <hjl.tools@gmail.com>
9208 * common/config/i386/cpuinfo.h (cpu_indicator_init): Add support
9209 for x86-64 micro levels for __builtin_cpu_supports.
9210 * common/config/i386/i386-cpuinfo.h (enum feature_priority):
9211 Add priorities for the micro-arch levels.
9212 (enum processor_features): Add new features.
9213 * common/config/i386/i386-isas.h: Add micro-arch features.
9214 * config/i386/i386-builtins.c (get_builtin_code_for_version):
9215 Support the micro-arch levels by callsing
9216 __builtin_cpu_supports.
9217 * doc/extend.texi: Document that the levels are support by
9218 __builtin_cpu_supports.
9220 2021-09-13 Andrew Pinski <apinski@marvell.com>
9223 * config/aarch64/aarch64-builtins.c (aarch64_fold_builtin_lane_check):
9225 (aarch64_general_fold_builtin): Handle AARCH64_SIMD_BUILTIN_LANE_CHECK.
9226 (aarch64_general_gimple_fold_builtin): Likewise.
9228 2021-09-13 Andrew Pinski <apinski@marvell.com>
9230 * config.gcc: Add m32r-*-linux* and m32rle-*-linux*
9231 to the Unsupported targets list.
9232 Remove support for m32r-*-linux* and m32rle-*-linux*.
9233 * config/m32r/linux.h: Removed.
9234 * config/m32r/t-linux: Removed.
9236 2021-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9239 * config/aarch64/aarch64.c (aarch64_classify_address): Don't allow
9240 register index for SVE predicate modes.
9242 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
9244 * tree-ssa-threadbackward.c
9245 (back_threader_profitability::profitable_path_p): Remove FSM
9247 (back_threader_registry::register_path): Same.
9248 * tree-ssa-threadedge.c
9249 (jump_threader::simplify_control_stmt_condition): Same.
9250 * tree-ssa-threadupdate.c (jt_path_registry::jt_path_registry):
9251 Add backedge_threads argument.
9252 (fwd_jt_path_registry::fwd_jt_path_registry): Pass
9253 backedge_threads argument.
9254 (back_jt_path_registry::back_jt_path_registry): Same.
9255 (dump_jump_thread_path): Adjust for FSM removal.
9256 (back_jt_path_registry::rewire_first_differing_edge): Same.
9257 (back_jt_path_registry::adjust_paths_after_duplication): Same.
9258 (back_jt_path_registry::update_cfg): Same.
9259 (jt_path_registry::register_jump_thread): Same.
9260 * tree-ssa-threadupdate.h (enum jump_thread_edge_type): Remove
9262 (class back_jt_path_registry): Add backedge_threads to
9265 2021-09-13 Martin Liska <mliska@suse.cz>
9268 * asan.h (sanitize_coverage_p): Handle when fn == NULL.
9270 2021-09-13 H.J. Lu <hjl.tools@gmail.com>
9273 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): New.
9274 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
9275 (MOVE_MAX): Check TARGET_AVX256_MOVE_BY_PIECES and
9276 TARGET_AVX256_STORE_BY_PIECES instead of
9277 TARGET_AVX256_SPLIT_UNALIGNED_LOAD and
9278 TARGET_AVX256_SPLIT_UNALIGNED_STORE.
9279 (STORE_MAX_PIECES): Check TARGET_AVX256_STORE_BY_PIECES instead
9280 of TARGET_AVX256_SPLIT_UNALIGNED_STORE.
9281 * config/i386/x86-tune.def (X86_TUNE_AVX256_MOVE_BY_PIECES): New.
9282 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
9284 2021-09-13 liuhongt <hongtao.liu@intel.com>
9287 * expmed.c (extract_bit_field_using_extv): Use
9288 gen_lowpart_if_possible instead of gen_lowpart to avoid ICE.
9290 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
9292 * Makefile.in (OBJS): Add value-pointer-equiv.o.
9293 * gimple-ssa-evrp.c (class ssa_equiv_stack): Move to
9294 value-pointer-equiv.*.
9295 (ssa_equiv_stack::ssa_equiv_stack): Same.
9296 (ssa_equiv_stack::enter): Same.
9297 (ssa_equiv_stack::leave): Same.
9298 (ssa_equiv_stack::push_replacement): Same.
9299 (ssa_equiv_stack::get_replacement): Same.
9300 (is_pointer_ssa): Same.
9301 (class pointer_equiv_analyzer): Same.
9302 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
9303 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Same.
9304 (pointer_equiv_analyzer::set_global_equiv): Same.
9305 (pointer_equiv_analyzer::set_cond_equiv): Same.
9306 (pointer_equiv_analyzer::get_equiv): Same.
9307 (pointer_equiv_analyzer::enter): Same.
9308 (pointer_equiv_analyzer::leave): Same.
9309 (pointer_equiv_analyzer::get_equiv_expr): Same.
9310 (pta_valueize): Same.
9311 (pointer_equiv_analyzer::visit_stmt): Same.
9312 (pointer_equiv_analyzer::visit_edge): Same.
9313 (hybrid_folder::value_of_expr): Same.
9314 (hybrid_folder::value_on_edge): Same.
9315 * value-pointer-equiv.cc: New file.
9316 * value-pointer-equiv.h: New file.
9318 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
9321 * gimple-fold.c (gimple_fold_builtin_memory_op): Allow folding
9322 memcpy if the size is not more than MOVE_MAX * MOVE_RATIO.
9324 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
9327 * config/arm/arm.md (movmisaligndi): New define_expand.
9328 * config/arm/vec-common.md (movmisalign<mode>): Iterate over VDQ mode.
9330 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
9333 * emit-rtl.c (gen_highpart): Use adjust_address to handle
9334 MEM rather than calling simplify_gen_subreg.
9336 2021-09-13 Jan-Benedict Glaw <jbglaw@ług-owl.de>
9338 * config/alpha/vms.h (INIT_CUMULATIVE_ARGS): Wrap multi-statment
9339 define into a block.
9341 2021-09-13 Richard Biener <rguenther@suse.de>
9343 * config/darwin.h (DARWIN_PREFER_DWARF): Do not define.
9344 * config/i386/darwin.h (PREFERRED_DEBUGGING_TYPE): Do not
9345 change based on DARWIN_PREFER_DWARF not being defined.
9347 2021-09-13 Richard Biener <rguenther@suse.de>
9349 * config/i386/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
9350 to inherit from elfos.h
9352 2021-09-13 Richard Biener <rguenther@suse.de>
9354 * config.gcc: Add cr16-*-* to the list of obsoleted targets.
9356 2021-09-13 Richard Biener <rguenther@suse.de>
9358 * config/avr/elf.h (PREFERRED_DEBUGGING_TYPE): Remove
9359 override, pick up DWARF2_DEBUG define from elfos.h
9361 2021-09-13 Richard Biener <rguenther@suse.de>
9363 * config/rx/rx.h (PREFERRED_DEBUGGING_TYPE): Always define to
9366 2021-09-13 Richard Biener <rguenther@suse.de>
9368 * config/alpha/vms.h (PREFERRED_DEBUGGING_TYPE): Define to
9371 2021-09-13 Richard Biener <rguenther@suse.de>
9373 * config/i386/cygming.h: Always default to DWARF2 debugging.
9374 Do not define DBX_DEBUGGING_INFO, that's done via dbxcoff.h
9376 * doc/install.texi: Document binutils 2.16 as minimum
9377 requirement for mingw.
9379 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
9381 * config/rs6000/rs6000.c (struct rs6000_cost_data): New members
9382 nstmts, nloads and extra_ctor_cost.
9383 (rs6000_density_test): Add load density related heuristics. Do
9384 extra costing on vector construction statements if need.
9385 (rs6000_init_cost): Init new members.
9386 (rs6000_update_target_cost_per_stmt): New function.
9387 (rs6000_add_stmt_cost): Factor vect_nonmem hunk out to function
9388 rs6000_update_target_cost_per_stmt and call it.
9390 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
9392 * config/rs6000/rs6000.c (struct rs6000_cost_data): Remove typedef.
9393 (rs6000_init_cost): Adjust.
9395 2021-09-13 liuhongt <hongtao.liu@intel.com>
9397 * config/i386/i386.md: (UNSPEC_COPYSIGN): Remove.
9398 (UNSPEC_XORSIGN): Ditto.
9400 2021-09-12 Roger Sayle <roger@nextmovesoftware.com>
9402 * expr.c (convert_move): Preserve SUBREG_PROMOTED_VAR_P when
9403 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
9406 2021-09-11 Aldy Hernandez <aldyh@redhat.com>
9408 * tree-ssa-threadbackward.c (class back_threader_registry): Use
9409 back_jt_path_registry.
9410 * tree-ssa-threadedge.c (jump_threader::jump_threader): Use
9411 fwd_jt_path_registry.
9412 * tree-ssa-threadedge.h (class jump_threader): Same..
9413 * tree-ssa-threadupdate.c
9414 (jump_thread_path_registry::jump_thread_path_registry): Rename...
9415 (jt_path_registry::jt_path_registry): ...to this.
9416 (jump_thread_path_registry::~jump_thread_path_registry): Rename...
9417 (jt_path_registry::~jt_path_registry): ...this.
9418 (fwd_jt_path_registry::fwd_jt_path_registry): New.
9419 (fwd_jt_path_registry::~fwd_jt_path_registry): New.
9420 (jump_thread_path_registry::allocate_thread_edge): Rename...
9421 (jt_path_registry::allocate_thread_edge): ...to this.
9422 (jump_thread_path_registry::allocate_thread_path): Rename...
9423 (jt_path_registry::allocate_thread_path): ...to this.
9424 (jump_thread_path_registry::lookup_redirection_data): Rename...
9425 (fwd_jt_path_registry::lookup_redirection_data): ...to this.
9426 (jump_thread_path_registry::thread_block_1): Rename...
9427 (fwd_jt_path_registry::thread_block_1): ...to this.
9428 (jump_thread_path_registry::thread_block): Rename...
9429 (fwd_jt_path_registry::thread_block): ...to this.
9430 (jt_path_registry::thread_through_loop_header): Rename...
9431 (fwd_jt_path_registry::thread_through_loop_header): ...to this.
9432 (jump_thread_path_registry::mark_threaded_blocks): Rename...
9433 (fwd_jt_path_registry::mark_threaded_blocks): ...to this.
9434 (jump_thread_path_registry::debug_path): Rename...
9435 (jt_path_registry::debug_path): ...to this.
9436 (jump_thread_path_registry::dump): Rename...
9437 (jt_path_registry::debug): ...to this.
9438 (jump_thread_path_registry::rewire_first_differing_edge): Rename...
9439 (back_jt_path_registry::rewire_first_differing_edge): ...to this.
9440 (jump_thread_path_registry::adjust_paths_after_duplication): Rename...
9441 (back_jt_path_registry::adjust_paths_after_duplication): ...to this.
9442 (jump_thread_path_registry::duplicate_thread_path): Rename...
9443 (back_jt_path_registry::duplicate_thread_path): ...to this. Also,
9444 drop ill-formed candidates.
9445 (jump_thread_path_registry::remove_jump_threads_including): Rename...
9446 (fwd_jt_path_registry::remove_jump_threads_including): ...to this.
9447 (jt_path_registry::thread_through_all_blocks): New.
9448 (back_jt_path_registry::update_cfg): New.
9449 (fwd_jt_path_registry::update_cfg): New.
9450 (jump_thread_path_registry::register_jump_thread): Rename...
9451 (jt_path_registry::register_jump_thread): ...to this.
9452 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
9454 (class jt_path_registry): ...here.
9455 (class fwd_jt_path_registry): New.
9456 (class back_jt_path_registry): New.
9458 2021-09-10 liuhongt <hongtao.liu@intel.com>
9461 2021-09-01 liuhongt <hongtao.liu@intel.com>
9463 * emit-rtl.c (validate_subreg): Get rid of all float-int
9466 2021-09-10 Jakub Jelinek <jakub@redhat.com>
9468 * tree-core.h (enum omp_memory_order): Add OMP_MEMORY_ORDER_MASK,
9469 OMP_FAIL_MEMORY_ORDER_UNSPECIFIED, OMP_FAIL_MEMORY_ORDER_RELAXED,
9470 OMP_FAIL_MEMORY_ORDER_ACQUIRE, OMP_FAIL_MEMORY_ORDER_RELEASE,
9471 OMP_FAIL_MEMORY_ORDER_ACQ_REL, OMP_FAIL_MEMORY_ORDER_SEQ_CST and
9472 OMP_FAIL_MEMORY_ORDER_MASK enumerators.
9473 (OMP_FAIL_MEMORY_ORDER_SHIFT): Define.
9474 * gimple-pretty-print.c (dump_gimple_omp_atomic_load,
9475 dump_gimple_omp_atomic_store): Print [weak] for weak atomic
9477 * gimple.h (enum gf_mask): Change GF_OMP_ATOMIC_MEMORY_ORDER
9478 to 6-bit mask, adjust GF_OMP_ATOMIC_NEED_VALUE value and add
9480 (gimple_omp_atomic_weak_p, gimple_omp_atomic_set_weak): New inline
9482 * tree.h (OMP_ATOMIC_WEAK): Define.
9483 * tree-pretty-print.c (dump_omp_atomic_memory_order): Adjust for
9484 fail memory order being encoded in the same enum and also print
9485 fail clause if present.
9486 (dump_generic_node): Print weak clause if OMP_ATOMIC_WEAK.
9487 * gimplify.c (goa_stabilize_expr): Add target_expr and rhs arguments,
9488 handle pre_p == NULL case as a test mode that only returns value
9489 but doesn't change gimplify nor change anything otherwise, adjust
9490 recursive calls, add MODIFY_EXPR, ADDR_EXPR, COND_EXPR, TARGET_EXPR
9491 and CALL_EXPR handling, adjust COMPOUND_EXPR handling for
9492 __builtin_clear_padding calls, for !rhs gimplify as lvalue rather
9494 (gimplify_omp_atomic): Adjust goa_stabilize_expr caller. Handle
9495 COND_EXPR rhs. Set weak flag on gimple load/store for
9497 * omp-expand.c (omp_memory_order_to_fail_memmodel): New function.
9498 (omp_memory_order_to_memmodel): Adjust for fail clause encoded
9500 (expand_omp_atomic_cas): New function.
9501 (expand_omp_atomic_pipeline): Use omp_memory_order_to_fail_memmodel
9503 (expand_omp_atomic): Attempt to optimize atomic compare and exchange
9504 using expand_omp_atomic_cas.
9506 2021-09-10 Aldy Hernandez <aldyh@redhat.com>
9507 Michael Matz <matz@suse.de>
9509 * tree-pass.h (PROP_loop_opts_done): New.
9510 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
9511 Intersect with global range.
9512 * tree-ssa-loop.c (tree_ssa_loop_done): Set PROP_loop_opts_done.
9513 * tree-ssa-threadbackward.c
9514 (back_threader_profitability::profitable_path_p): Disable
9515 threading through latches until after loop optimizations have run.
9517 2021-09-10 David Faust <david.faust@oracle.com>
9519 * doc/invoke.texi: Document BPF -mcpu, -mjmpext, -mjmp32 and -malu32
9522 2021-09-10 David Faust <david.faust@oracle.com>
9524 * config/bpf/bpf-opts.h (bpf_isa_version): New enum.
9525 * config/bpf/bpf-protos.h (bpf_expand_cbranch): New.
9526 * config/bpf/bpf.c (bpf_option_override): Handle -mcpu option.
9527 (bpf_expand_cbranch): New function.
9528 * config/bpf/bpf.md (AM mode iterator): Conditionalize support for SI
9530 (zero_extendsidi2): Only use mov32 instruction if it is available.
9531 (SIM mode iterator): Conditionalize support for SI mode.
9532 (JM mode iterator): New.
9533 (cbranchdi4): Update name, use new JM iterator. Use bpf_expand_cbranch.
9534 (*branch_on_di): Update name, use new JM iterator.
9535 * config/bpf/bpf.opt: (mjmpext): New option.
9539 (bpf_isa): New enum.
9541 2021-09-10 David Faust <david.faust@oracle.com>
9543 * config/bpf/bpf.md (zero_extendhidi2): Add new output template
9544 for register-to-register extensions.
9545 (zero_extendqidi2): Likewise.
9547 2021-09-10 Richard Biener <rguenther@suse.de>
9549 PR middle-end/102273
9550 * internal-fn.c (expand_DEFERRED_INIT): Always expand non-SSA vars.
9552 2021-09-10 Richard Biener <rguenther@suse.de>
9554 PR middle-end/102269
9555 * gimplify.c (is_var_need_auto_init): Empty types do not need
9558 2021-09-10 Richard Biener <rguenther@suse.de>
9560 * configure.ac (--with-stabs): Remove.
9561 * configure: Regenerate.
9562 * doc/install.texi: Remove --with-stabs documentation.
9564 2021-09-10 liuhongt <hongtao.liu@intel.com>
9566 * config/i386/avx512fp16intrin.h: (_mm512_cmp_ph_mask):
9568 (_mm512_mask_cmp_ph_mask): Likewise.
9569 (_mm512_cmp_round_ph_mask): Likewise.
9570 (_mm512_mask_cmp_round_ph_mask): Likewise.
9571 (_mm_cmp_sh_mask): Likewise.
9572 (_mm_mask_cmp_sh_mask): Likewise.
9573 (_mm_cmp_round_sh_mask): Likewise.
9574 (_mm_mask_cmp_round_sh_mask): Likewise.
9575 (_mm_comieq_sh): Likewise.
9576 (_mm_comilt_sh): Likewise.
9577 (_mm_comile_sh): Likewise.
9578 (_mm_comigt_sh): Likewise.
9579 (_mm_comige_sh): Likewise.
9580 (_mm_comineq_sh): Likewise.
9581 (_mm_ucomieq_sh): Likewise.
9582 (_mm_ucomilt_sh): Likewise.
9583 (_mm_ucomile_sh): Likewise.
9584 (_mm_ucomigt_sh): Likewise.
9585 (_mm_ucomige_sh): Likewise.
9586 (_mm_ucomineq_sh): Likewise.
9587 (_mm_comi_round_sh): Likewise.
9588 (_mm_comi_sh): Likewise.
9589 * config/i386/avx512fp16vlintrin.h (_mm_cmp_ph_mask): New intrinsic.
9590 (_mm_mask_cmp_ph_mask): Likewise.
9591 (_mm256_cmp_ph_mask): Likewise.
9592 (_mm256_mask_cmp_ph_mask): Likewise.
9593 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
9594 * config/i386/i386-builtin.def: Add corresponding new builtins.
9595 * config/i386/i386-expand.c
9596 (ix86_expand_args_builtin): Handle new builtin types.
9597 (ix86_expand_round_builtin): Ditto.
9598 * config/i386/i386.md (ssevecmode): Add HF mode.
9599 (MODEFH): New mode iterator.
9600 * config/i386/sse.md
9601 (V48H_AVX512VL): New mode iterator to support HF vector modes.
9602 Ajdust corresponding description.
9603 (ssecmpintprefix): New.
9604 (VI12_AVX512VL): Adjust to support HF vector modes.
9605 (cmp_imm_predicate): Likewise.
9606 (<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
9608 (avx512f_vmcmp<mode>3<round_saeonly_name>): Likewise.
9609 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Likewise.
9610 (<sse>_<unord>comi<round_saeonly_name>): Likewise.
9612 2021-09-10 liuhongt <hongtao.liu@intel.com>
9614 * config/i386/avx512fp16intrin.h: (_mm512_max_ph): New intrinsic.
9615 (_mm512_mask_max_ph): Likewise.
9616 (_mm512_maskz_max_ph): Likewise.
9617 (_mm512_min_ph): Likewise.
9618 (_mm512_mask_min_ph): Likewise.
9619 (_mm512_maskz_min_ph): Likewise.
9620 (_mm512_max_round_ph): Likewise.
9621 (_mm512_mask_max_round_ph): Likewise.
9622 (_mm512_maskz_max_round_ph): Likewise.
9623 (_mm512_min_round_ph): Likewise.
9624 (_mm512_mask_min_round_ph): Likewise.
9625 (_mm512_maskz_min_round_ph): Likewise.
9626 (_mm_max_sh): Likewise.
9627 (_mm_mask_max_sh): Likewise.
9628 (_mm_maskz_max_sh): Likewise.
9629 (_mm_min_sh): Likewise.
9630 (_mm_mask_min_sh): Likewise.
9631 (_mm_maskz_min_sh): Likewise.
9632 (_mm_max_round_sh): Likewise.
9633 (_mm_mask_max_round_sh): Likewise.
9634 (_mm_maskz_max_round_sh): Likewise.
9635 (_mm_min_round_sh): Likewise.
9636 (_mm_mask_min_round_sh): Likewise.
9637 (_mm_maskz_min_round_sh): Likewise.
9638 * config/i386/avx512fp16vlintrin.h (_mm_max_ph): New intrinsic.
9639 (_mm256_max_ph): Likewise.
9640 (_mm_mask_max_ph): Likewise.
9641 (_mm256_mask_max_ph): Likewise.
9642 (_mm_maskz_max_ph): Likewise.
9643 (_mm256_maskz_max_ph): Likewise.
9644 (_mm_min_ph): Likewise.
9645 (_mm256_min_ph): Likewise.
9646 (_mm_mask_min_ph): Likewise.
9647 (_mm256_mask_min_ph): Likewise.
9648 (_mm_maskz_min_ph): Likewise.
9649 (_mm256_maskz_min_ph): Likewise.
9650 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
9651 * config/i386/i386-builtin.def: Add corresponding new builtins.
9652 * config/i386/i386-expand.c
9653 (ix86_expand_args_builtin): Handle new builtin types.
9654 * config/i386/sse.md
9655 (<code><mode>3<mask_name><round_saeonly_name>): Adjust to
9656 support HF vector modes.
9657 (*<code><mode>3<mask_name><round_saeonly_name>): Likewise.
9658 (ieee_<ieee_maxmin><mode>3<mask_name><round_saeonly_name>):
9660 (<sse>_vm<code><mode>3<mask_scalar_name><round_saeonly_scalar_name>):
9662 * config/i386/subst.md (round_saeonly_mode512bit_condition):
9663 Adjust for HF vector modes.
9665 2021-09-10 Liu, Hongtao <hongtao.liu@intel.com>
9667 * config/i386/avx512fp16intrin.h (_mm_add_sh): New intrinsic.
9668 (_mm_mask_add_sh): Likewise.
9669 (_mm_maskz_add_sh): Likewise.
9670 (_mm_sub_sh): Likewise.
9671 (_mm_mask_sub_sh): Likewise.
9672 (_mm_maskz_sub_sh): Likewise.
9673 (_mm_mul_sh): Likewise.
9674 (_mm_mask_mul_sh): Likewise.
9675 (_mm_maskz_mul_sh): Likewise.
9676 (_mm_div_sh): Likewise.
9677 (_mm_mask_div_sh): Likewise.
9678 (_mm_maskz_div_sh): Likewise.
9679 (_mm_add_round_sh): Likewise.
9680 (_mm_mask_add_round_sh): Likewise.
9681 (_mm_maskz_add_round_sh): Likewise.
9682 (_mm_sub_round_sh): Likewise.
9683 (_mm_mask_sub_round_sh): Likewise.
9684 (_mm_maskz_sub_round_sh): Likewise.
9685 (_mm_mul_round_sh): Likewise.
9686 (_mm_mask_mul_round_sh): Likewise.
9687 (_mm_maskz_mul_round_sh): Likewise.
9688 (_mm_div_round_sh): Likewise.
9689 (_mm_mask_div_round_sh): Likewise.
9690 (_mm_maskz_div_round_sh): Likewise.
9691 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
9692 * config/i386/i386-builtin.def: Add corresponding new builtins.
9693 * config/i386/i386-expand.c
9694 (ix86_expand_round_builtin): Handle new builtins.
9695 * config/i386/sse.md (VF_128): Change description.
9696 (<sse>_vm<plusminus_insn><mode>3<mask_scalar_name><round_scalar_name>):
9697 Adjust to support HF vector modes.
9698 (<sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name><round_scalar_name>):
9701 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
9703 * config/i386/i386-expand.c
9704 (ix86_avx256_split_vector_move_misalign): Handle V16HF mode.
9705 * config/i386/i386.c
9706 (ix86_preferred_simd_mode): Handle HF mode.
9707 * config/i386/sse.md (V_256H): New mode iterator.
9708 (avx_vextractf128<mode>): Use it.
9709 (VEC_INIT_MODE): Align vector HFmode condition to vector
9710 HImodes since there're no real HF instruction used.
9711 (VEC_INIT_HALF_MODE): Ditto.
9713 (VIHF_AVX512BW): Ditto.
9714 (*vec_extracthf): Ditto.
9715 (VEC_EXTRACT_MODE): Ditto.
9717 2021-09-10 Richard Biener <rguenther@suse.de>
9720 * config/dbx.h: Remove.
9721 * config/dbxcoff.h: Do not define PREFERRED_DEBUGGING_TYPE.
9722 * config/lynx.h: Likewise.
9724 2021-09-10 liuhongt <hongtao.liu@intel.com>
9726 * config/i386/i386-expand.c (ix86_expand_copysign): Expand
9727 right into ANDNOT + AND + IOR, using paradoxical subregs.
9728 (ix86_split_copysign_const): Remove.
9729 (ix86_split_copysign_var): Ditto.
9730 * config/i386/i386-protos.h (ix86_split_copysign_const): Dotto.
9731 (ix86_split_copysign_var): Ditto.
9732 * config/i386/i386.md (@copysign<mode>3_const): Ditto.
9733 (@copysign<mode>3_var): Ditto.
9735 2021-09-09 qing zhao <qing.zhao@oracle.com>
9737 * builtins.c (expand_builtin_memset): Make external visible.
9738 * builtins.h (expand_builtin_memset): Declare extern.
9739 * common.opt (ftrivial-auto-var-init=): New option.
9740 * doc/extend.texi: Document the uninitialized attribute.
9741 * doc/invoke.texi: Document -ftrivial-auto-var-init.
9742 * flag-types.h (enum auto_init_type): New enumerated type
9744 * gimple-fold.c (clear_padding_type): Add one new parameter.
9745 (clear_padding_union): Likewise.
9746 (clear_padding_emit_loop): Likewise.
9747 (clear_type_padding_in_mask): Likewise.
9748 (gimple_fold_builtin_clear_padding): Handle this new parameter.
9749 * gimplify.c (gimple_add_init_for_auto_var): New function.
9750 (gimple_add_padding_init_for_auto_var): New function.
9751 (is_var_need_auto_init): New function.
9752 (gimplify_decl_expr): Add initialization to automatic variables per
9754 (gimplify_call_expr): Add one new parameter for call to
9755 __builtin_clear_padding.
9756 (gimplify_init_constructor): Add padding initialization in the end.
9757 * internal-fn.c (INIT_PATTERN_VALUE): New macro.
9758 (expand_DEFERRED_INIT): New function.
9759 * internal-fn.def (DEFERRED_INIT): New internal function.
9760 * tree-cfg.c (verify_gimple_call): Verify calls to .DEFERRED_INIT.
9761 * tree-sra.c (generate_subtree_deferred_init): New function.
9762 (scan_function): Avoid setting cannot_scalarize_away_bitmap for
9763 calls to .DEFERRED_INIT.
9764 (sra_modify_deferred_init): New function.
9765 (sra_modify_function_body): Handle calls to DEFERRED_INIT specially.
9766 * tree-ssa-structalias.c (find_func_aliases_for_call): Likewise.
9767 * tree-ssa-uninit.c (warn_uninit): Handle calls to DEFERRED_INIT
9769 (check_defs): Likewise.
9770 (warn_uninitialized_vars): Likewise.
9771 * tree-ssa.c (ssa_undefined_value_p): Likewise.
9772 * tree.c (build_common_builtin_nodes): Build tree node for
9773 BUILT_IN_CLEAR_PADDING when needed.
9775 2021-09-09 Richard Biener <rguenther@suse.de>
9777 * tree-ssa-loop-im.c (fill_always_executed_in_1): Walk
9780 2021-09-09 Richard Biener <rguenther@suse.de>
9782 * tree-ssa-loop-im.c (fill_always_executed_in_1): Integrate
9783 DOM walk from get_loop_body_in_dom_order using a worklist
9786 2021-09-09 liuhongt <hongtao.liu@intel.com>
9788 * config.gcc: Add avx512fp16vlintrin.h.
9789 * config/i386/avx512fp16intrin.h: (_mm512_add_ph): New intrinsic.
9790 (_mm512_mask_add_ph): Likewise.
9791 (_mm512_maskz_add_ph): Likewise.
9792 (_mm512_sub_ph): Likewise.
9793 (_mm512_mask_sub_ph): Likewise.
9794 (_mm512_maskz_sub_ph): Likewise.
9795 (_mm512_mul_ph): Likewise.
9796 (_mm512_mask_mul_ph): Likewise.
9797 (_mm512_maskz_mul_ph): Likewise.
9798 (_mm512_div_ph): Likewise.
9799 (_mm512_mask_div_ph): Likewise.
9800 (_mm512_maskz_div_ph): Likewise.
9801 (_mm512_add_round_ph): Likewise.
9802 (_mm512_mask_add_round_ph): Likewise.
9803 (_mm512_maskz_add_round_ph): Likewise.
9804 (_mm512_sub_round_ph): Likewise.
9805 (_mm512_mask_sub_round_ph): Likewise.
9806 (_mm512_maskz_sub_round_ph): Likewise.
9807 (_mm512_mul_round_ph): Likewise.
9808 (_mm512_mask_mul_round_ph): Likewise.
9809 (_mm512_maskz_mul_round_ph): Likewise.
9810 (_mm512_div_round_ph): Likewise.
9811 (_mm512_mask_div_round_ph): Likewise.
9812 (_mm512_maskz_div_round_ph): Likewise.
9813 * config/i386/avx512fp16vlintrin.h: New header.
9814 * config/i386/i386-builtin-types.def (V16HF, V8HF, V32HF):
9815 Add new builtin types.
9816 * config/i386/i386-builtin.def: Add corresponding builtins.
9817 * config/i386/i386-expand.c
9818 (ix86_expand_args_builtin): Handle new builtin types.
9819 (ix86_expand_round_builtin): Likewise.
9820 * config/i386/immintrin.h: Include avx512fp16vlintrin.h
9821 * config/i386/sse.md (VFH): New mode_iterator.
9823 (avx512fmaskmode): Add HF vector modes.
9824 (avx512fmaskhalfmode): Likewise.
9825 (<plusminus_insn><mode>3<mask_name><round_name>): Adjust to for
9827 (*<plusminus_insn><mode>3<mask_name><round_name>): Likewise.
9828 (mul<mode>3<mask_name><round_name>): Likewise.
9829 (*mul<mode>3<mask_name><round_name>): Likewise.
9830 (div<mode>3): Likewise.
9831 (<sse>_div<mode>3<mask_name><round_name>): Likewise.
9832 * config/i386/subst.md (SUBST_V): Add HF vector modes.
9833 (SUBST_A): Likewise.
9834 (round_mode512bit_condition): Adjust for V32HFmode.
9836 2021-09-09 liuhongt <hongtao.liu@intel.com>
9839 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
9840 (reduc_plus_scal_v4sf): .. this, New define_expand.
9841 (reduc_plus_scal_v2df): .. and this, New define_expand.
9843 2021-09-09 liuhongt <hongtao.liu@intel.com>
9846 * config/i386/sse.md (*vec_extract<mode><ssescalarmodelower>_valign):
9849 2021-09-08 Jonathan Wakely <jwakely@redhat.com>
9852 * doc/trouble.texi (Copy Assignment): Fix description of
9853 behaviour and fix code in example.
9855 2021-09-08 Segher Boessenkool <segher@kernel.crashing.org>
9858 * config/rs6000/rs6000-logue.c (rs6000_emit_epilogue): For ELFv2 use
9859 r11 instead of r12 for restoring CR.
9861 2021-09-08 Jakub Jelinek <jakub@redhat.com>
9862 liuhongt <hongtao.liu@intel.com>
9865 * config/i386/i386.md (@xorsign<mode>3_1): Remove.
9866 * config/i386/i386-expand.c (ix86_expand_xorsign): Expand right away
9867 into AND with mask and XOR, using paradoxical subregs.
9868 (ix86_split_xorsign): Remove.
9869 * config/i386/i386-protos.h (ix86_split_xorsign): Remove.
9871 2021-09-08 Di Zhao <dizhao@os.amperecomputing.com>
9873 * tree-ssa-sccvn.c (vn_nary_op_insert_into): fix result compare
9875 2021-09-08 Jakub Jelinek <jakub@redhat.com>
9878 * config/i386/i386.md (xorsign<mode>3): If operands[1] is equal to
9879 operands[2], emit abs<mode>2 instead.
9880 (@xorsign<mode>3_1): Add early-clobbers for output operand, enable
9881 first alternative even for avx, add another alternative with
9882 =&Yv <- 0, Yv, Yvm constraints.
9883 * config/i386/i386-expand.c (ix86_split_xorsign): If op0 is equal
9884 to op1, emit vpandn instead.
9886 2021-09-08 liuhongt <hongtao.liu@intel.com>
9888 * config/i386/avx512fp16intrin.h (_mm_set_ph): New intrinsic.
9889 (_mm256_set_ph): Likewise.
9890 (_mm512_set_ph): Likewise.
9891 (_mm_setr_ph): Likewise.
9892 (_mm256_setr_ph): Likewise.
9893 (_mm512_setr_ph): Likewise.
9894 (_mm_set1_ph): Likewise.
9895 (_mm256_set1_ph): Likewise.
9896 (_mm512_set1_ph): Likewise.
9897 (_mm_setzero_ph): Likewise.
9898 (_mm256_setzero_ph): Likewise.
9899 (_mm512_setzero_ph): Likewise.
9900 (_mm_set_sh): Likewise.
9901 (_mm_load_sh): Likewise.
9902 (_mm_store_sh): Likewise.
9903 * config/i386/i386-builtin-types.def (V8HF): New type.
9904 (DEF_FUNCTION_TYPE (V8HF, V8HI)): New builtin function type
9905 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
9906 Support vector HFmodes.
9907 (ix86_expand_vector_init_one_nonzero): Likewise.
9908 (ix86_expand_vector_init_one_var): Likewise.
9909 (ix86_expand_vector_init_interleave): Likewise.
9910 (ix86_expand_vector_init_general): Likewise.
9911 (ix86_expand_vector_set): Likewise.
9912 (ix86_expand_vector_extract): Likewise.
9913 (ix86_expand_vector_init_concat): Likewise.
9914 (ix86_expand_sse_movcc): Handle vector HFmodes.
9915 (ix86_expand_vector_set_var): Ditto.
9916 * config/i386/i386-modes.def: Add HF vector modes in comment.
9917 * config/i386/i386.c (classify_argument): Add HF vector modes.
9918 (ix86_hard_regno_mode_ok): Allow HF vector modes for AVX512FP16.
9919 (ix86_vector_mode_supported_p): Likewise.
9920 (ix86_set_reg_reg_cost): Handle vector HFmode.
9921 (ix86_get_ssemov): Handle vector HFmode.
9922 (function_arg_advance_64): Pass unamed V16HFmode and V32HFmode
9924 (function_arg_advance_32): Pass V8HF/V16HF/V32HF by sse reg for 32bit
9926 (function_arg_advance_32): Ditto.
9927 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): New.
9928 (VALID_AVX256_REG_OR_OI_MODE): Rename to ..
9929 (VALID_AVX256_REG_OR_OI_VHF_MODE): .. this, and add V16HF.
9930 (VALID_SSE2_REG_VHF_MODE): New.
9931 (VALID_AVX512VL_128_REG_MODE): Add V8HF and TImode.
9932 (SSE_REG_MODE_P): Add vector HFmode.
9933 * config/i386/i386.md (mode): Add HF vector modes.
9934 (MODE_SIZE): Likewise.
9935 (ssemodesuffix): Add ph suffix for HF vector modes.
9936 * config/i386/sse.md (VFH_128): New mode iterator.
9937 (VMOVE): Adjust for HF vector modes.
9939 (V_256_512): Likewise.
9941 (avx512fmaskmode): Likewise.
9942 (shuffletype): Likewise.
9943 (sseinsnmode): Likewise.
9944 (ssedoublevecmode): Likewise.
9945 (ssehalfvecmode): Likewise.
9946 (ssehalfvecmodelower): Likewise.
9947 (ssePScmode): Likewise.
9948 (ssescalarmode): Likewise.
9949 (ssescalarmodelower): Likewise.
9950 (sseintprefix): Likewise.
9952 (bcstscalarsuff): Likewise.
9953 (xtg_mode): Likewise.
9954 (VI12HF_AVX512VL): New mode_iterator.
9955 (VF_AVX512FP16): Likewise.
9957 (VIHF_256): Likewise.
9958 (VIHF_AVX512BW): Likewise.
9959 (V16_256): Likewise.
9960 (V32_512): Likewise.
9961 (sseintmodesuffix): New mode_attr.
9962 (sse): Add scalar and vector HFmodes.
9963 (ssescalarmode): Add vector HFmode mapping.
9964 (ssescalarmodesuffix): Add sh suffix for HFmode.
9965 (*<sse>_vm<insn><mode>3): Use VFH_128.
9966 (*<sse>_vm<multdiv_mnemonic><mode>3): Likewise.
9967 (*ieee_<ieee_maxmin><mode>3): Likewise.
9968 (<avx512>_blendm<mode>): New define_insn.
9969 (vec_setv8hf): New define_expand.
9970 (vec_set<mode>_0): New define_insn for HF vector set.
9971 (*avx512fp16_movsh): Likewise.
9972 (avx512fp16_movsh): Likewise.
9973 (vec_extract_lo_v32hi): Rename to ...
9974 (vec_extract_lo_<mode>): ... this, and adjust to allow HF
9976 (vec_extract_hi_v32hi): Likewise.
9977 (vec_extract_hi_<mode>): Likewise.
9978 (vec_extract_lo_v16hi): Likewise.
9979 (vec_extract_lo_<mode>): Likewise.
9980 (vec_extract_hi_v16hi): Likewise.
9981 (vec_extract_hi_<mode>): Likewise.
9982 (vec_set_hi_v16hi): Likewise.
9983 (vec_set_hi_<mode>): Likewise.
9984 (vec_set_lo_v16hi): Likewise.
9985 (vec_set_lo_<mode>): Likewise.
9986 (*vec_extract<mode>_0): New define_insn_and_split for HF
9988 (*vec_extracthf): New define_insn.
9989 (VEC_EXTRACT_MODE): Add HF vector modes.
9990 (PINSR_MODE): Add V8HF.
9991 (sse2p4_1): Likewise.
9992 (pinsr_evex_isa): Likewise.
9993 (<sse2p4_1>_pinsr<ssemodesuffix>): Adjust to support
9994 insert for V8HFmode.
9995 (pbroadcast_evex_isa): Add HF vector modes.
9996 (AVX2_VEC_DUP_MODE): Likewise.
9997 (VEC_INIT_MODE): Likewise.
9998 (VEC_INIT_HALF_MODE): Likewise.
9999 (avx2_pbroadcast<mode>): Adjust to support HF vector mode
10001 (avx2_pbroadcast<mode>_1): Likewise.
10002 (<avx512>_vec_dup<mode>_1): Likewise.
10003 (<avx512>_vec_dup<mode><mask_name>): Likewise.
10004 (<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>):
10007 2021-09-08 Guo, Xuepeng <xuepeng.guo@intel.com>
10008 H.J. Lu <hongjiu.lu@intel.com>
10009 Liu Hongtao <hongtao.liu@intel.com>
10010 Wang Hongyu <hongyu.wang@intel.com>
10011 Xu Dianhong <dianhong.xu@intel.com>
10013 * common/config/i386/cpuinfo.h (get_available_features):
10014 Detect FEATURE_AVX512FP16.
10015 * common/config/i386/i386-common.c
10016 (OPTION_MASK_ISA_AVX512FP16_SET,
10017 OPTION_MASK_ISA_AVX512FP16_UNSET,
10018 OPTION_MASK_ISA2_AVX512FP16_SET,
10019 OPTION_MASK_ISA2_AVX512FP16_UNSET): New.
10020 (OPTION_MASK_ISA2_AVX512BW_UNSET,
10021 OPTION_MASK_ISA2_AVX512BF16_UNSET): Add AVX512FP16.
10022 (ix86_handle_option): Handle -mavx512fp16.
10023 * common/config/i386/i386-cpuinfo.h (enum processor_features):
10024 Add FEATURE_AVX512FP16.
10025 * common/config/i386/i386-isas.h: Add entry for AVX512FP16.
10026 * config.gcc: Add avx512fp16intrin.h.
10027 * config/i386/avx512fp16intrin.h: New intrinsic header.
10028 * config/i386/cpuid.h: Add bit_AVX512FP16.
10029 * config/i386/i386-builtin-types.def: (FLOAT16): New primitive type.
10030 * config/i386/i386-builtins.c: Support _Float16 type for i386
10032 (ix86_register_float16_builtin_type): New function.
10033 (ix86_float16_type_node): New.
10034 * config/i386/i386-c.c (ix86_target_macros_internal): Define
10036 * config/i386/i386-expand.c (ix86_expand_branch): Support
10038 (ix86_prepare_fp_compare_args): Adjust TARGET_SSE_MATH &&
10039 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
10040 (ix86_expand_fp_movcc): Ditto.
10041 * config/i386/i386-isa.def: Add PTA define for AVX512FP16.
10042 * config/i386/i386-options.c (isa2_opts): Add -mavx512fp16.
10043 (ix86_valid_target_attribute_inner_p): Add avx512fp16 attribute.
10044 * config/i386/i386.c (ix86_get_ssemov): Use
10045 vmovdqu16/vmovw/vmovsh for HFmode/HImode scalar or vector.
10046 (ix86_get_excess_precision): Use
10047 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when TARGET_AVX512FP16
10049 (sse_store_index): Use SFmode cost for HFmode cost.
10050 (inline_memory_move_cost): Add HFmode, and perfer SSE cost over
10051 GPR cost for HFmode.
10052 (ix86_hard_regno_mode_ok): Allow HImode in sse register.
10053 (ix86_mangle_type): Add manlging for _Float16 type.
10054 (inline_secondary_memory_needed): No memory is needed for
10055 16bit movement between gpr and sse reg under
10057 (ix86_multiplication_cost): Adjust TARGET_SSE_MATH &&
10058 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
10059 (ix86_division_cost): Ditto.
10060 (ix86_rtx_costs): Ditto.
10061 (ix86_add_stmt_cost): Ditto.
10062 (ix86_optab_supported_p): Ditto.
10063 * config/i386/i386.h (VALID_AVX512F_SCALAR_MODE): Add HFmode.
10064 (SSE_FLOAT_MODE_SSEMATH_OR_HF_P): Add HFmode.
10065 (PTA_SAPPHIRERAPIDS): Add PTA_AVX512FP16.
10066 * config/i386/i386.md (mode): Add HFmode.
10067 (MODE_SIZE): Add HFmode.
10068 (isa): Add avx512fp16.
10069 (enabled): Handle avx512fp16.
10070 (ssemodesuffix): Add sh suffix for HFmode.
10071 (comm): Add mult, div.
10072 (plusminusmultdiv): New code iterator.
10073 (insn): Add mult, div.
10074 (*movhf_internal): Adjust for avx512fp16 instruction.
10075 (*movhi_internal): Ditto.
10076 (*cmpi<unord>hf): New define_insn for HFmode.
10077 (*ieee_s<ieee_maxmin>hf3): Likewise.
10078 (extendhf<mode>2): Likewise.
10079 (trunc<mode>hf2): Likewise.
10080 (float<floatunssuffix><mode>hf2): Likewise.
10081 (*<insn>hf): Likewise.
10082 (cbranchhf4): New expander.
10083 (movhfcc): Likewise.
10084 (<insn>hf3): Likewise.
10085 (mulhf3): Likewise.
10086 (divhf3): Likewise.
10087 * config/i386/i386.opt: Add mavx512fp16.
10088 * config/i386/immintrin.h: Include avx512fp16intrin.h.
10089 * doc/invoke.texi: Add mavx512fp16.
10090 * doc/extend.texi: Add avx512fp16 Usage Notes.
10092 2021-09-08 liuhongt <hongtao.liu@intel.com>
10094 * common.opt: Support -fexcess-precision=16.
10095 * config/aarch64/aarch64.c (aarch64_excess_precision): Return
10096 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when
10097 EXCESS_PRECISION_TYPE_FLOAT16.
10098 * config/arm/arm.c (arm_excess_precision): Ditto.
10099 * config/i386/i386.c (ix86_get_excess_precision): Ditto.
10100 * config/m68k/m68k.c (m68k_excess_precision): Issue an error
10101 when EXCESS_PRECISION_TYPE_FLOAT16.
10102 * config/s390/s390.c (s390_excess_precision): Ditto.
10103 * coretypes.h (enum excess_precision_type): Add
10104 EXCESS_PRECISION_TYPE_FLOAT16.
10105 * doc/tm.texi (TARGET_C_EXCESS_PRECISION): Update documents.
10106 * doc/tm.texi.in (TARGET_C_EXCESS_PRECISION): Ditto.
10107 * doc/extend.texi (Half-Precision): Document
10108 -fexcess-precision=16.
10109 * flag-types.h (enum excess_precision): Add
10110 EXCESS_PRECISION_FLOAT16.
10111 * target.def (excess_precision): Update document.
10112 * tree.c (excess_precision_type): Set excess_precision_type to
10113 EXCESS_PRECISION_FLOAT16 when -fexcess-precision=16.
10115 2021-09-08 liuhongt <hongtao.liu@intel.com>
10117 * doc/extend.texi: (@node Floating Types): Adjust the wording.
10118 (@node Half-Precision): Ditto.
10120 2021-09-07 Takayuki 'January June' Suwa <jjsuwa_sys3175@yahoo.co.jp>
10123 * config/xtensa/xtensa.c (xtensa_emit_move_sequence): Add
10124 'CONST_INT_P (src)' to the condition of the block that tries to
10125 eliminate literal when loading integer contant.
10127 2021-09-07 David Faust <david.faust@oracle.com>
10129 * doc/extend.texi (BPF Type Attributes) New node.
10130 Document new preserve_access_index attribute.
10131 Document new preserve_access_index builtin.
10132 * doc/invoke.texi: Document -mco-re and -mno-co-re options.
10134 2021-09-07 David Faust <david.faust@oracle.com>
10136 * config/bpf/bpf.c: Adjust includes.
10137 (bpf_handle_preserve_access_index_attribute): New function.
10138 (bpf_attribute_table): Use it here.
10139 (bpf_builtins): Add BPF_BUILTIN_PRESERVE_ACCESS_INDEX.
10140 (bpf_option_override): Handle "-mco-re" option.
10141 (bpf_asm_init_sections): New.
10142 (TARGET_ASM_INIT_SECTIONS): Redefine.
10143 (bpf_file_end): New.
10144 (TARGET_ASM_FILE_END): Redefine.
10145 (bpf_init_builtins): Add "__builtin_preserve_access_index".
10146 (bpf_core_compute, bpf_core_get_index): New.
10147 (is_attr_preserve_access): New.
10148 (bpf_expand_builtin): Handle new builtins.
10149 (bpf_core_newdecl, bpf_core_is_maybe_aggregate_access): New.
10150 (bpf_core_walk): New.
10151 (bpf_resolve_overloaded_builtin): New.
10152 (TARGET_RESOLVE_OVERLOADED_BUILTIN): Redefine.
10153 (handle_attr): New.
10154 (pass_bpf_core_attr): New RTL pass.
10155 * config/bpf/bpf-passes.def: New file.
10156 * config/bpf/bpf-protos.h (make_pass_bpf_core_attr): New.
10157 * config/bpf/coreout.c: New file.
10158 * config/bpf/coreout.h: Likewise.
10159 * config/bpf/t-bpf (TM_H): Add $(srcdir)/config/bpf/coreout.h.
10160 (coreout.o): New rule.
10161 (PASSES_EXTRA): Add $(srcdir)/config/bpf/bpf-passes.def.
10162 * config.gcc (bpf): Add coreout.h to extra_headers.
10163 Add coreout.o to extra_objs.
10164 Add $(srcdir)/config/bpf/coreout.c to target_gtfiles.
10166 2021-09-07 David Faust <david.faust@oracle.com>
10168 * btfout.c (get_btf_id): Function is no longer static.
10169 * ctfc.h: Expose it here.
10171 2021-09-07 David Faust <david.faust@oracle.com>
10173 * ctfc.c (ctf_lookup_tree_type): New function.
10174 * ctfc.h: Likewise.
10176 2021-09-07 David Faust <david.faust@oracle.com>
10178 * ctfc.c (ctf_dtd_lookup): Function is no longer static.
10179 * ctfc.h: Analogous change.
10181 2021-09-07 David Faust <david.faust@oracle.com>
10183 * dwarf2out.c (lookup_type_die): Function is no longer static.
10184 * dwarf2out.h: Expose it here.
10186 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
10188 * dwarf2ctf.c (ctf_debug_finalize): Make it static.
10189 (ctf_debug_early_finish): New definition.
10190 (ctf_debug_finish): Likewise.
10191 * dwarf2ctf.h (ctf_debug_finalize): Remove declaration.
10192 (ctf_debug_early_finish): New declaration.
10193 (ctf_debug_finish): Likewise.
10194 * dwarf2out.c (dwarf2out_finish): Invoke ctf_debug_finish.
10195 (dwarf2out_early_finish): Invoke ctf_debug_early_finish.
10197 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
10199 * config/bpf/bpf.c (bpf_option_override): For BPF backend, disable LTO
10200 support when compiling for CO-RE.
10201 * config/bpf/bpf.opt: Add new command line option -mco-re.
10203 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
10205 * flag-types.h (enum debug_info_type): Add new enum
10206 DINFO_TYPE_BTF_WITH_CORE.
10207 (BTF_WITH_CORE_DEBUG): New bitmask.
10208 * flags.h (btf_with_core_debuginfo_p): New declaration.
10209 * opts.c (btf_with_core_debuginfo_p): New definition.
10211 2021-09-07 Jason Merrill <jason@redhat.com>
10213 * tree.h (error_operand_p): Change to inline function.
10215 2021-09-07 Aldy Hernandez <aldyh@redhat.com>
10217 * tree-ssa-threadedge.c (forwarder_block_p): Rename to...
10218 (empty_block_with_phis_p): ...this.
10219 (potentially_threadable_block): Same.
10220 (jump_threader::thread_through_normal_block): Same.
10222 2021-09-07 Eric Botcazou <ebotcazou@adacore.com>
10225 * dwarf2out.c (mark_base_types): New overloaded function.
10226 (dwarf2out_early_finish): Invoke it on the COMDAT type list as well
10227 as the compilation unit, and call move_marked_base_types afterward.
10229 2021-09-07 H.J. Lu <hjl.tools@gmail.com>
10232 * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
10234 (ix86_expand_vector_convert_uns_vsivsf): Likewise.
10236 2021-09-07 Richard Biener <rguenther@suse.de>
10238 PR tree-optimization/102226
10239 * tree-vect-loop.c (vect_transform_cycle_phi): Record
10240 the converted value for the epilogue PHI use.
10242 2021-09-07 Martin Liska <mliska@suse.cz>
10244 PR gcov-profile/80223
10245 * ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
10246 options, do not inline when no_profile_instrument_function
10247 attributes are different in early inliner. It's fine to inline
10248 it after PGO instrumentation.
10250 2021-09-07 Richard Biener <rguenther@suse.de>
10252 PR tree-optimization/101555
10253 * tree-ssa-pre.c (translate_vuse_through_block): Do not
10254 perform an alias walk to determine the validity of the
10255 mem at the start of the block which is already guaranteed
10256 by means of prune_clobbered_mems.
10257 (phi_translate_1): Pass edge to translate_vuse_through_block.
10259 2021-09-07 Xionghu Luo <luoxhu@linux.ibm.com>
10262 * config/rs6000/rs6000.md (fmod<mode>3): New define_expand.
10263 (remainder<mode>3): Likewise.
10265 2021-09-07 YunQiang Su <yunqiang.su@cipunited.com>
10267 * config/mips/mips.c (mips_file_start): add .module for
10270 2021-09-06 Roger Sayle <roger@nextmovesoftware.com>
10272 * wide-int.cc (wi::clz): Reorder tests to ensure the result
10273 is zero for all negative values.
10275 2021-09-06 Tobias Burnus <tobias@codesourcery.com>
10277 * doc/invoke.texi (-foffload-options): Fix @opindex.
10279 2021-09-06 H.J. Lu <hjl.tools@gmail.com>
10282 * config/i386/i386-expand.c (ix86_split_xorsign): Use operands[2].
10283 * config/i386/i386.md (@xorsign<mode>3_1): Add non-destructive
10284 source alternative for AVX.
10286 2021-09-06 liuhongt <hongtao.liu@intel.com>
10288 PR middle-end/102182
10289 * optabs.c (expand_fix): Add from1 to avoid from being
10292 2021-09-06 Eric Botcazou <ebotcazou@adacore.com>
10294 * dwarf2out.c (modified_type_die): Deal with all array types earlier
10295 and use local variable consistently throughout the function.
10297 2021-09-06 Jakub Jelinek <jakub@redhat.com>
10299 PR tree-optimization/102207
10300 * match.pd: Don't demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
10301 were promoted from signed to wider unsigned type.
10303 2021-09-06 Andrew Pinski <apinski@marvell.com>
10305 PR tree-optimization/63184
10306 * match.pd: Add simplification of pointer_diff of two pointer_plus
10307 with addr_expr in the first operand of each pointer_plus.
10308 Add simplificatoin of ne/eq of two pointer_plus with addr_expr
10309 in the first operand of each pointer_plus.
10311 2021-09-06 Richard Biener <rguenther@suse.de>
10313 PR tree-optimization/102176
10314 * tree-vect-slp.c (vect_slp_gather_vectorized_scalar_stmts):
10316 (vect_bb_slp_scalar_cost): Use the computed set of
10317 vectorized scalar stmts instead of relying on the out-of-date
10318 and not accurate PURE_SLP_STMT.
10319 (vect_bb_vectorization_profitable_p): Compute the set
10320 of vectorized scalar stmts.
10322 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
10324 * gimple-range-path.cc (path_range_query::range_of_stmt): Remove
10325 GIMPLE_COND special casing.
10326 (path_range_query::range_defined_in_block): Use range_of_stmt
10327 instead of calling fold_range directly.
10329 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
10331 * gimple-range-path.cc (path_range_query::range_of_expr): Set
10332 m_undefined_path when appropriate.
10333 (path_range_query::internal_range_of_expr): Copy from range_of_expr.
10334 (path_range_query::unreachable_path_p): New.
10335 (path_range_query::precompute_ranges): Set m_undefined_path.
10336 * gimple-range-path.h (path_range_query::unreachable_path_p): New.
10337 (path_range_query::internal_range_of_expr): New.
10338 * tree-ssa-threadbackward.c (back_threader::find_taken_edge_cond):
10339 Use unreachable_path_p.
10341 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
10343 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
10344 Remove argument and call find_taken_edge.
10345 (back_threader::resolve_phi): Do not calculate taken edge before
10346 calling maybe_register_path.
10347 (back_threader::find_paths_to_names): Same.
10349 2021-09-05 Jeff Law <jlaw@localhost.localdomain>
10351 * config/h8300/h8300.md (QHSI2 mode iterator): New mode iterator.
10352 * config/h8300/testcompare.md (store_c): Update name, use new
10354 (store_neg_c, store_shifted_c): New patterns.
10356 2021-09-03 Segher Boessenkool <segher@kernel.crashing.org>
10359 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): On ELFv2 use r11
10360 instead of r12 for CR save, in all cases.
10362 2021-09-03 Andrew Pinski <apinski@marvell.com>
10364 * config/aarch64/aarch64-sve-builtins.cc (register_vector_type):
10365 Handle error_mark_node as the type of the type_decl.
10367 2021-09-03 Andrew Pinski <apinski@marvell.com>
10369 * config/aarch64/aarch64-builtins.c (struct aarch64_simd_type_info):
10371 (aarch64_simd_types): Likewise.
10372 (aarch64_simd_intOI_type_node): Likewise.
10373 (aarch64_simd_intCI_type_node): Likewise.
10374 (aarch64_simd_intXI_type_node): Likewise.
10375 * config/aarch64/aarch64.h (aarch64_fp16_type_node): Likewise.
10376 (aarch64_fp16_ptr_type_node): Likewise.
10377 (aarch64_bf16_type_node): Likewise.
10378 (aarch64_bf16_ptr_type_node): Likewise.
10380 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
10382 * range-op.cc (operator_minus::op1_op2_relation_effect): Abstract
10384 (minus_op1_op2_relation_effect): ...here.
10385 (class operator_pointer_diff): New.
10386 (operator_pointer_diff::op1_op2_relation_effect): Call
10387 minus_op1_op2_relation_effect.
10388 (integral_table::integral_table): Add entry for POINTER_DIFF_EXPR.
10390 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
10392 * tree-ssa-threadbackward.c (back_threader::thread_through_all_blocks):
10393 Add may_peel_loop_headers.
10394 (back_threader_registry::thread_through_all_blocks): Same.
10395 (try_thread_blocks): Pass may_peel_loop_headers argument.
10396 (pass_early_thread_jumps::execute): Same.
10398 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
10400 * tree-ssa-threadedge.c (has_phis_p): New.
10401 (forwarder_block_p): New.
10402 (potentially_threadable_block): Call forwarder_block_p.
10403 (jump_threader::thread_around_empty_blocks): Call has_phis_p.
10404 (jump_threader::thread_through_normal_block): Call
10407 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
10409 * tree-ssa-threadbackward.c (back_threader::dump): New.
10410 (back_threader::debug): New.
10411 (back_threader_profitability::profitable_path_p): Dump blocks
10412 even if we are bailing early.
10414 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
10416 * tree-ssa-threadupdate.c (cancel_thread): New.
10417 (jump_thread_path_registry::thread_block_1): Use cancel_thread.
10418 (jump_thread_path_registry::mark_threaded_blocks): Same.
10419 (jump_thread_path_registry::register_jump_thread): Same.
10421 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
10423 * tree-ssa-threadedge.c (jt_state::push): Only call methods for
10424 which objects are available.
10425 (jt_state::pop): Same.
10426 (jt_state::register_equiv): Same.
10427 (jt_state::register_equivs_on_edge): Same.
10429 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
10431 * tree-ssa-threadedge.c (jump_threader::thread_across_edge):
10432 Move pop until after a thread is registered.
10434 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
10436 * tree-ssa-threadupdate.c (debug): New.
10438 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
10440 * gimple-range-trace.cc (push_dump_file::push_dump_file): New.
10441 (push_dump_file::~push_dump_file): New.
10442 (dump_ranger): Change dump_file temporarily while dumping
10444 * gimple-range-trace.h (class push_dump_file): New.
10446 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
10448 * gimple-range-trace.cc (debug_seed_ranger): Remove static.
10449 (dump_ranger): Dump function name.
10451 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
10453 * gimple-range-path.cc (path_range_query::range_defined_in_block):
10454 Adjust for non-null.
10455 (path_range_query::adjust_for_non_null_uses): New.
10456 (path_range_query::precompute_ranges): Call
10457 adjust_for_non_null_uses.
10458 * gimple-range-path.h: Add m_non_null and
10459 adjust_for_non_null_uses.
10461 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
10463 * gimple-range-path.cc (path_range_query::dump): Dump path
10465 (path_range_query::precompute_ranges): Dump entire path.
10467 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
10469 * value-relation.cc (relation_oracle::debug): New.
10470 * value-relation.h (relation_oracle::debug): New.
10472 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
10474 * tree-ssa-loop-ch.c: Remove unnecessary include file.
10476 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
10478 * gimple-range-fold.cc (fold_using_range::postfold_gcond_edges):
10479 Skip statements with no defining BB.
10480 * gimple-range-path.cc (path_range_query::range_defined_in_block):
10481 Do not get confused by statements with no defining BB.
10483 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
10485 * gimple-range-fold.cc (adjust_imagpart_expr): Move from
10486 gimple_range_adjustment. Add support for constants.
10487 (adjust_realpart_expr): New.
10488 (gimple_range_adjustment): Move IMAGPART_EXPR code to
10489 adjust_imagpart_expr.
10490 * range-op.cc (integral_table::integral_table): Add entry for
10493 2021-09-03 Jakub Jelinek <jakub@redhat.com>
10495 * omp-expand.c (expand_omp_atomic_pipeline): Use
10496 IFN_ATOMIC_COMPARE_EXCHANGE instead of
10497 BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_? so that memory order
10500 2021-09-03 Jakub Jelinek <jakub@redhat.com>
10503 * tree.h (DECL_FIELD_ABI_IGNORED): Changed into rvalue only macro
10504 that is false if DECL_BIT_FIELD.
10505 (SET_DECL_FIELD_ABI_IGNORED, DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD,
10506 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD): Define.
10507 * tree-streamer-out.c (pack_ts_decl_common_value_fields): For
10508 DECL_BIT_FIELD stream DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead
10509 of DECL_FIELD_ABI_IGNORED.
10510 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
10511 SET_DECL_FIELD_ABI_IGNORED instead of writing to
10512 DECL_FIELD_ABI_IGNORED and for DECL_BIT_FIELD use
10513 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead.
10514 * lto-streamer-out.c (hash_tree): For DECL_BIT_FIELD hash
10515 DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead of DECL_FIELD_ABI_IGNORED.
10517 2021-09-03 liuhongt <hongtao.liu@intel.com>
10520 * config/i386/amxbf16intrin.h : Remove macro check for __AMX_BF16__.
10521 * config/i386/amxint8intrin.h : Remove macro check for __AMX_INT8__.
10522 * config/i386/amxtileintrin.h : Remove macro check for __AMX_TILE__.
10524 2021-09-02 Martin Sebor <msebor@redhat.com>
10526 PR tree-optimization/17506
10528 * tree-ssa-uninit.c (warn_uninit): Remove conditional guarding note.
10530 2021-09-02 Richard Biener <rguenther@suse.de>
10532 * tree-ssa-loop-im.c (fill_always_executed_in_1): Refine
10533 fix for PR78185 and continue processing when leaving
10534 finite inner loops.
10536 2021-09-02 Jakub Jelinek <jakub@redhat.com>
10538 PR tree-optimization/99591
10539 * match.pd: Demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
10542 2021-09-02 Richard Biener <rguenther@suse.de>
10545 2021-09-02 Richard Biener <rguenther@suse.de>
10547 PR tree-optimization/102155
10548 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
10549 over a part of the RPO array and do not recurse here.
10550 Dump blocks marked as always executed.
10551 (fill_always_executed_in): Walk over the RPO array and
10552 process loops whose header we run into.
10553 (loop_invariant_motion_in_fun): Compute the first RPO
10554 using rev_post_order_and_mark_dfs_back_seme in iteration
10555 order and pass that to fill_always_executed_in.
10557 2021-09-02 liuhongt <hongtao.liu@intel.com>
10559 * config/i386/i386-modes.def (FLOAT_MODE): Define ieee HFmode.
10560 * config/i386/i386.c (enum x86_64_reg_class): Add
10561 X86_64_SSEHF_CLASS.
10562 (merge_classes): Handle X86_64_SSEHF_CLASS.
10563 (examine_argument): Ditto.
10564 (construct_container): Ditto.
10565 (classify_argument): Ditto, and set HFmode/HCmode to
10566 X86_64_SSEHF_CLASS.
10567 (function_value_32): Return _FLoat16/Complex Float16 by
10569 (function_value_64): Return _Float16/Complex Float16 by SSE
10571 (ix86_print_operand): Handle CONST_DOUBLE HFmode.
10572 (ix86_secondary_reload): Require gpr as intermediate register
10573 to store _Float16 from sse register when sse4 is not
10575 (ix86_libgcc_floating_mode_supported_p): Enable _FLoat16 under
10577 (ix86_scalar_mode_supported_p): Ditto.
10578 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Defined.
10579 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add HFmode.
10580 (VALID_INT_MODE_P): Add HFmode and HCmode.
10581 * config/i386/i386.md (*pushhf_rex64): New define_insn.
10583 (*movhf_internal): Ditto.
10584 * doc/extend.texi (Half-Precision Floating Point): Documemt
10587 2021-09-02 Richard Biener <rguenther@suse.de>
10589 PR tree-optimization/102155
10590 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
10591 over a part of the RPO array and do not recurse here.
10592 Dump blocks marked as always executed.
10593 (fill_always_executed_in): Walk over the RPO array and
10594 process loops whose header we run into.
10595 (loop_invariant_motion_in_fun): Compute the first RPO
10596 using rev_post_order_and_mark_dfs_back_seme in iteration
10597 order and pass that to fill_always_executed_in.
10599 2021-09-02 YunQiang Su <syq@debian.org>
10602 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
10604 * config/mips/mips.c (mips_module_isa_name): New.
10605 mips_file_start: add .module mipsREV to all asm output
10607 2021-09-01 Jeff Law <jlaw@localhost.localdomain>
10609 PR tree-optimization/102152
10610 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Reduce a vector
10611 comparison to a scalar comparison before calling
10612 update_stmt_if_modified.
10614 2021-09-01 Andrew Pinski <apinski@marvell.com>
10617 * config/aarch64/aarch64.c (aarch64_expand_setmem):
10618 Check STRICT_ALIGNMENT before creating an overlapping
10621 2021-09-01 Martin Sebor <msebor@redhat.com>
10623 * gimple-ssa-warn-access.cc (get_size_range): Add argument.
10624 (check_access): Pass additional argument.
10625 (check_memop_access): Remove template and make a member function.
10626 (maybe_check_dealloc_call): Make a pass_waccess member function.
10627 (class pass_waccess): Add, rename, and remove members.
10628 (pass_waccess::pass_waccess): Adjust to name change.
10629 (pass_waccess::~pass_waccess): Same.
10630 (check_alloca): Make a member function.
10631 (check_alloc_size_call): Same.
10632 (check_strcat): Same.
10633 (check_strncat): Same.
10634 (check_stxcpy): Same.
10635 (check_stxncpy): Same.
10636 (check_strncmp): Same.
10637 (maybe_warn_rdwr_sizes): Rename...
10638 (pass_waccess::maybe_check_access_sizes): ...to this.
10639 (pass_waccess::check_call): Adjust to name changes.
10640 (pass_waccess::maybe_check_dealloc_call): Make a pass_waccess member
10642 (pass_waccess::execute): Adjust to name changes.
10643 * gimple-ssa-warn-access.h (check_memop_access): Remove.
10644 * pointer-query.cc (access_ref::phi): Handle null pointer.
10645 (access_ref::inform_access): Same.
10646 (pointer_query::put_ref): Modify a cached value, not a copy of it.
10647 (pointer_query::dump): New function.
10648 (compute_objsize_r): Avoid overwriting access_ref::bndrng. Cache
10650 * pointer-query.h (pointer_query::dump): Declare.
10651 * tree-ssa-strlen.c (get_range): Simplify. Use function query.
10652 (dump_strlen_info): Use function query.
10653 (printf_strlen_execute): Factor code out into pointer_query::put_ref.
10655 2021-09-01 Thomas Schwinge <thomas@codesourcery.com>
10657 * tree.c (walk_tree_1) <OMP_CLAUSE>: Simplify.
10659 2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
10661 * doc/extend.texi: Document unavailable attribute.
10662 * print-tree.c (print_node): Handle unavailable attribute.
10663 * tree-core.h (struct tree_base): Add a bit to carry unavailability.
10664 * tree.c (error_unavailable_use): New.
10665 * tree.h (TREE_UNAVAILABLE): New.
10666 (error_unavailable_use): New.
10668 2021-09-01 Jakub Jelinek <jakub@redhat.com>
10670 PR tree-optimization/102124
10671 * tree-vect-patterns.c (vect_recog_widen_op_pattern): For ORIG_CODE
10672 MINUS_EXPR, if itype is unsigned with smaller precision than type,
10673 add an extra cast to signed variant of itype to ensure sign-extension.
10675 2021-09-01 Martin Liska <mliska@suse.cz>
10677 * graph.c (draw_cfg_node_succ_edges): Do not color fallthru
10678 edges and rather use colors for TRUE and FALSE edges.
10680 2021-09-01 Richard Biener <rguenther@suse.de>
10682 PR tree-optimization/93491
10683 * tree-ssa-pre.c (compute_avail): Set BB_MAY_NOTRETURN
10684 after processing the stmt itself. Do not consider
10685 pure functions possibly not returning. Properly avoid
10686 adding possibly trapping calls to EXP_GEN when there's
10687 a preceeding possibly not returning call.
10688 * tree-ssa-sccvn.c (vn_reference_may_trap): Conservatively
10691 2021-09-01 Richard Biener <rguenther@suse.de>
10693 PR tree-optimization/102139
10694 * tree-vectorizer.h (vec_base_alignments): Adjust hash-map
10695 type to record a std::pair of the stmt-info and the innermost
10697 (dr_vec_info::group): New member.
10698 * tree-vect-data-refs.c (vect_record_base_alignment): Adjust.
10699 (vect_compute_data_ref_alignment): Verify the recorded
10700 base alignment can be used.
10701 (data_ref_pair): Remove.
10702 (dr_group_sort_cmp): Adjust.
10703 (vect_analyze_data_ref_accesses): Store the group-ID in the
10704 dr_vec_info and operate on a vector of dr_vec_infos.
10706 2021-09-01 YunQiang Su <yunqiang.su@cipunited.com>
10708 * read-md.c (md_reader::handle_enum): support value assignation.
10709 * doc/md.texi: record define_c_enum value assignation support.
10711 2021-09-01 Jakub Jelinek <jakub@redhat.com>
10713 PR tree-optimization/102141
10714 * gimple-ssa-store-merging.c (bswap_view_convert): Add BEFORE
10715 argument. If false, emit stmts after gsi instead of before, and
10717 (bswap_replace): Adjust callers. When converting output of bswap,
10718 emit VIEW_CONVERT prepratation stmts after a copy of gsi instead
10721 2021-09-01 liuhongt <hongtao.liu@intel.com>
10723 * emit-rtl.c (validate_subreg): Get rid of all float-int
10726 2021-09-01 liuhongt <hongtao.liu@intel.com>
10729 2021-08-30 liuhongt <hongtao.liu@intel.com>
10731 * expmed.c (extract_bit_field_1): Make sure we're playing with
10732 integral modes before call extract_integral_bit_field.
10733 (extract_integral_bit_field): Add a parameter of type
10734 scalar_int_mode which corresponds to of tmode.
10735 And call extract_and_convert_fixed_bit_field instead of
10736 extract_fixed_bit_field and convert_extracted_bit_field.
10737 (extract_and_convert_fixed_bit_field): New function, it's a
10738 combination of extract_fixed_bit_field and
10739 convert_extracted_bit_field.
10741 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
10743 * tree.c (walk_tree_1) <OMP_CLAUSE_TILE>: Handle three operands.
10745 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
10747 * omp-general.h (omp_is_reference): Rename to...
10748 (omp_privatize_by_reference): ... this. Adjust all users...
10749 * omp-general.c: ... here, ...
10750 * gimplify.c: ... here, ...
10751 * omp-expand.c: ... here, ...
10752 * omp-low.c: ... here.
10754 2021-08-31 Martin Sebor <msebor@redhat.com>
10756 * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Test
10757 pointer element for equality to zero, not that of the cotaining
10760 2021-08-31 Martin Sebor <msebor@redhat.com>
10762 * gcc-rich-location.h (gcc_rich_location): Make ctor explicit.
10764 2021-08-31 Martin Sebor <msebor@redhat.com>
10766 * function.h (function): Add comments.
10767 (get_range_query): Same. Add attribute returns nonnull.
10769 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
10771 * expr.c (convert_modes): Don't use subreg_promoted_mode on a
10772 SUBREG if it can't be guaranteed to a SUBREG_PROMOTED_VAR_P set.
10773 Instead use the standard (safer) is_a <scalar_int_mode> idiom.
10775 2021-08-31 Jeff Law <jlaw@localhost.localdomain>
10777 * config.gcc (cris-*-elf, cris-*-none): Remove dbxelf.h from
10779 (m32r-*-elf, m32rle-*-elf, m32r-*-linux): Likewise.
10780 (mn10300-*-*, am33_2.0-*-linux*): Likewise.
10781 (xtensa*-*-elf, xtensa*-*-linux, xtensa*-*-uclinux): Likewise.
10782 (m32c-*-elf*, m32c-*-rtems*): Likewise.
10783 * config/cris/cris.h (DBX_NO_XREFS): Remove.
10784 (DBX_CONTIN_LENGTH, DBX_CONTIN_CHAR): Likewise.
10785 * config/m32r/m32r.h (DBXOUT_SOURCE_LINE): Likewise.
10786 (DBX_DEBUGGING_INFO, DBX_CONTIN_LENGTH): Likewise.
10787 * config/mn10300/mn10300.h (DEFAULT_GDB_EXTENSIONS): Likewise.
10788 * config/mn10300/linux.h (DBX_REGISTER_NAMES): Likewise.
10790 2021-08-31 Marcel Vollweiler <marcel@codesourcery.com>
10792 * gimplify.c (gimplify_scan_omp_clauses): Error handling. 'ancestor' only
10793 allowed on target constructs and only with particular other clauses.
10794 * omp-expand.c (expand_omp_target): Output of 'sorry, not supported' if
10795 'ancestor' is used.
10796 * omp-low.c (check_omp_nesting_restrictions): Error handling. No nested OpenMP
10797 structs when 'ancestor' is used.
10798 (scan_omp_1_stmt): No usage of OpenMP runtime routines in a target region when
10799 'ancestor' is used.
10800 * tree-pretty-print.c (dump_omp_clause): Append 'ancestor'.
10801 * tree.h (OMP_CLAUSE_DEVICE_ANCESTOR): Define macro.
10803 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
10805 * expr.c (convert_modes): Preserve SUBREG_PROMOTED_VAR_P when
10806 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
10808 * simplify-rtx.c (simplify_unary_operation_1) [SIGN_EXTEND]:
10809 Likewise, preserve SUBREG_PROMOTED_VAR_P when creating a (wider)
10810 partial subreg from a SUBREG_PROMOTED_VAR_P subreg. Generate
10811 SIGN_EXTEND of the SUBREG_REG when a subreg would be paradoxical.
10812 [ZERO_EXTEND]: Likewise, preserve SUBREG_PROMOTED_VAR_P when
10813 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
10814 subreg. Generate ZERO_EXTEND of the SUBREG_REG when a subreg
10815 would be paradoxical.
10817 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
10819 * combine.c (combine_simplify_rtx): Avoid converting an explicit
10820 TRUNCATE into a lowpart SUBREG on !TRULY_NOOP_TRUNCATION targets.
10821 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10823 2021-08-31 Richard Biener <rguenther@suse.de>
10825 PR tree-optimization/102142
10826 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Fix
10827 condition under which to unset the visited flag.
10829 2021-08-31 Richard Biener <rguenther@suse.de>
10831 PR middle-end/102129
10832 * tree-ssa-ter.c (find_replaceable_in_bb): Do not move
10833 possibly trapping expressions across calls.
10835 2021-08-31 Jakub Jelinek <jakub@redhat.com>
10837 PR tree-optimization/102134
10838 * tree-ssa-ccp.c (bit_value_binop) <case RSHIFT_EXPR>: If sgn is
10839 UNSIGNED and r1val | r1mask has MSB set, ensure lzcount doesn't
10842 2021-08-31 Andrew Pinski <apinski@marvell.com>
10845 * collect-utils.c (setup_signals): New declaration.
10846 * collect-utils.h (setup_signals): New function.
10847 * collect2.c (handler): Delete.
10848 (main): Instead of manually setting up the signals,
10849 just call setup_signals.
10850 * lto-wrapper.c (main): Likewise.
10852 2021-08-31 Andrew Pinski <apinski@marvell.com>
10855 * config/i386/i386-protos.h (x86_output_aligned_bss):
10856 Change align argument to unsigned type.
10857 (x86_elf_aligned_decl_common): Likewise.
10858 * config/i386/i386.c (x86_elf_aligned_decl_common): Likewise.
10859 (x86_output_aligned_bss): Likewise.
10861 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
10863 * config/mips/mips.c (mips_module_isa_name): New.
10864 mips_file_start: add .module mipsREV to all asm output
10866 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
10868 * config/mips/mips.h (struct mips_cpu_info): define enum mips_isa;
10869 use enum instead of int for 'isa' member.
10870 * config.gcc, config/mips/mips.c, config/mips/mips-cpus.def,
10871 config/mips/netbsd.h: replace hardcoded numbers with enum.
10873 2021-08-31 liuhongt <hongtao.liu@intel.com>
10875 * config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
10876 define_split to define_insn_and_split.
10877 (*avx2_eq<mode>3): Removed.
10878 (<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
10879 (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
10880 (*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
10882 (*avx2_gt<mode>3): Removed.
10883 (<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
10884 define_insn to define_expand, and adjust pattern.
10885 (UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.
10887 2021-08-30 David Malcolm <dmalcolm@redhat.com>
10890 * Makefile.in (ANALYZER_OBJS): Add analyzer/call-info.o.
10892 2021-08-30 Jason Merrill <jason@redhat.com>
10894 * doc/invoke.texi: Document -Wmissing-requires.
10896 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
10898 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Remove
10899 TARGET_EXTRA_BUILTINS guard.
10901 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
10903 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Change
10904 initialization of V2DI_type_node and unsigned_V2DI_type_node.
10906 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
10908 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Use the new
10909 decl when new_builtins_are_live.
10910 * config/rs6000/rs6000-builtin-new.def (__builtin_cfstring): New
10913 2021-08-30 Pat Haugen <pthaugen@linux.ibm.com>
10915 * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add
10916 OPTION_MASK_P10_FUSION_2STORE.
10917 (POWERPC_MASKS): Likewise.
10918 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
10919 store fusion for Power10.
10920 (is_fusable_store): New.
10921 (power10_sched_reorder): Likewise.
10922 (rs6000_sched_reorder): Do Power10 specific reordering.
10923 (rs6000_sched_reorder2): Likewise.
10924 * config/rs6000/rs6000.opt: Add new option.
10926 2021-08-30 Richard Biener <rguenther@suse.de>
10928 PR tree-optimization/102128
10929 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
10930 Move scanning for if-converted scalar code to the caller
10931 and instead delay clearing the visited flag for profitable
10933 (vect_slp_region): Cost all subgraphs before scheduling.
10934 For if-converted BB vectorization scan for scalar COND_EXPRs
10935 and do not vectorize if any found and the cost model is
10938 2021-08-30 Richard Biener <rguenther@suse.de>
10940 * common.opt (fexceptions): Mark
10941 EnabledBy(fnon-call-exceptions).
10942 * doc/invoke.texi (fnon-call-exceptions): Document this
10943 enables -fexceptions.
10945 2021-08-30 Sebastian Huber <sebastian.huber@embedded-brains.de>
10947 * tsystem.h (abort): Define abort() if inhibit_libc is defined and it
10948 is not already defined.
10950 2021-08-30 liuhongt <hongtao.liu@intel.com>
10952 * expmed.c (extract_bit_field_1): Make sure we're playing with
10953 integral modes before call extract_integral_bit_field.
10954 (extract_integral_bit_field): Add a parameter of type
10955 scalar_int_mode which corresponds to of tmode.
10956 And call extract_and_convert_fixed_bit_field instead of
10957 extract_fixed_bit_field and convert_extracted_bit_field.
10958 (extract_and_convert_fixed_bit_field): New function, it's a
10959 combination of extract_fixed_bit_field and
10960 convert_extracted_bit_field.
10962 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
10964 * config/darwin.c (darwin_libc_has_function): Do not run
10965 the checks for x86 or modern Darwin. Make sure that there
10966 is a value set for darwin_macosx_version_min before testing.
10968 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
10970 * config/i386/darwin.h (CLEAR_INSN_CACHE): New.
10972 2021-08-28 Jan Hubicka <hubicka@ucw.cz>
10974 * ipa-modref-tree.h (modref_access_node::merge): Break out
10975 logic combining offsets and logic merging ranges to ...
10976 (modref_access_node::combined_offsets): ... here
10977 (modref_access_node::update2): ... here
10978 (modref_access_node::closer_pair_p): New member function.
10979 (modref_access_node::forced_merge): New member function.
10980 (modre_ref_node::insert): Do merging when table is full.
10982 2021-08-28 YunQiang Su <yunqiang.su@cipunited.com>
10985 * config.gcc: MIPS: use N64 ABI by default if the triple end
10986 with -gnuabi64, which is used by Debian since 2013.
10988 2021-08-28 Alexandre Oliva <oliva@adacore.com>
10990 * ipa-modref.c (analyze_function): Skip debug stmts.
10991 * tree-inline.c (estimate_num_insn): Consider builtins even
10992 without a cgraph_node.
10994 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
10996 * config/h8300/bitfield.md (cstore<mode>4): Remove expander.
10997 * config/h8300/h8300.c (h8300_expand_branch): Remove function.
10998 * config/h8300/h8300-protos.h (h8300_expadn_branch): Remove prototype.
10999 * config/h8300/h8300.md (eqne): New code iterator.
11000 (geultu, geultu_to_c): Similarly.
11001 * config/h8300/testcompare.md (cstore<mode>4): Dummy expander.
11002 (store_c_<mode>, store_c_i_<mode>): New define_insn_and_splits
11003 (cmp<mode>_c): New pattern
11005 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
11007 * tree-ssa-dom.c (reduce_vector_comparison_to_scalar_comparison): New
11009 (dom_opt_dom_walker::optimize_stmt): Use it.
11011 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
11013 * config/darwin.c (finalize_ctors): Add a section-start linker-
11015 (finalize_dtors): Likewise.
11016 * config/darwin.h (MIN_LD64_INIT_TERM_START_LABELS): New.
11018 2021-08-27 Bill Schmidt <wschmidt@linux.ibm.com>
11020 * config/rs6000/rs6000-call.c (rs6000-builtins.h): New #include.
11021 (rs6000_init_builtins): Call rs6000_init_generated_builtins. Skip the
11022 old initialization logic when new builtins are enabled.
11023 * config/rs6000/rs6000-gen-builtins.c (write_decls): Rename
11024 rs6000_autoinit_builtins to rs6000_init_generated_builtins.
11025 (write_init_file): Likewise.
11027 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
11029 * configure.ac (darwin2[[0-9]]* | darwin19*): Alter use of
11030 gcc_GAS_CHECK_FEATURE to remove an extraneous parameter.
11031 (amdgcn-* | gcn-*) Likewise.
11033 2021-08-27 Anthony Sharp <anthonysharp15@gmail.com>
11035 * symbol-summary.h: Added missing template keyword.
11037 2021-08-27 Richard Biener <rguenther@suse.de>
11039 PR tree-optimization/45178
11040 * tree-ssa-dce.c (find_obviously_necessary_stmts): For
11041 infinite loops without exit do not mark control dependent
11042 edges of the latch necessary.
11044 2021-08-27 konglin1 <lingling.kong@intel.com>
11047 * config/i386/sse.md: (<avx512>scattersi<mode>): Add mask operand to
11049 (<avx512>scattersi<mode>): Likewise.
11050 (*avx512f_scattersi<VI48F:mode>): Merge mask operand to set_dest.
11051 (*avx512f_scatterdi<VI48F:mode>): Likewise
11053 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
11055 * config/rs6000/rs6000.c (rs6000_builtin_md_vectorized_function): Add
11056 support for built-in functions MISC_BUILTIN_DIVWE, MISC_BUILTIN_DIVWEU,
11057 MISC_BUILTIN_DIVDE, MISC_BUILTIN_DIVDEU, P10_BUILTIN_CFUGED,
11058 P10_BUILTIN_CNTLZDM, P10_BUILTIN_CNTTZDM, P10_BUILTIN_PDEPD and
11059 P10_BUILTIN_PEXTD on Power10.
11061 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
11063 * config/rs6000/rs6000-call.c (builtin_function_type): Add unsigned
11064 signedness for some Power10 bifs.
11066 2021-08-27 David Edelsohn <dje.gcc@gmail.com>
11069 * config/rs6000/rs6000.c (rs6000_adjust_field_align): Use
11070 computed alignment if the entire struct has attribute packed.
11072 2021-08-27 liuhongt <hongtao.liu@intel.com>
11076 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
11077 IX86_BUILTIN_SHUFPD512, IX86_BUILTIN_SHUFPS512,
11078 IX86_BUILTIN_SHUFPD256, IX86_BUILTIN_SHUFPS,
11079 IX86_BUILTIN_SHUFPS256.
11080 (ix86_masked_all_ones): New function.
11082 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
11084 * config/i386/i386.md (*btr<mode>_1): Call force_reg unconditionally.
11085 (conditional moves with memory inputs splitters): Ditto.
11086 * config/i386/sse.md (one_cmpl<mode>2): Simplify.
11088 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
11090 * ipa-modref-tree.h (modref_access_node::try_merge_with): Restart
11091 search after merging.
11093 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
11095 * config/rs6000/rs6000-overload.def: Add remaining overloads.
11097 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
11099 * config/rs6000/rs6000-builtin-new.def: Add cell stanza.
11101 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
11103 * config/rs6000/rs6000-builtin-new.def: Add ieee128-hw, dfp,
11104 crypto, and htm stanzas.
11106 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
11108 * config/rs6000/rs6000-builtin-new.def: Add mma stanza.
11110 2021-08-26 Martin Sebor <msebor@redhat.com>
11112 * tree-ssa-uninit.c (warn_uninit): Refactor and simplify.
11113 (warn_uninit_phi_uses): Remove argument from calls to warn_uninit.
11114 (warn_uninitialized_vars): Same. Reduce visibility of locals.
11115 (warn_uninitialized_phi): Same.
11117 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
11119 * tree-ssa-ccp.c (get_individual_bits): Helper function to
11120 extract the individual bits from a widest_int constant (mask).
11121 (gray_code_bit_flips): New read-only table for effiently
11122 enumerating permutations/combinations of bits.
11123 (bit_value_binop) [LROTATE_EXPR, RROTATE_EXPR]: Handle rotates
11124 by unknown counts that are guaranteed less than the target
11125 precision and four or fewer unknown bits by enumeration.
11126 [LSHIFT_EXPR, RSHIFT_EXPR]: Likewise, also handle shifts by
11127 enumeration under the same conditions. Handle remaining
11128 shifts as a mask based upon the minimum possible shift value.
11130 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
11131 Richard Biener <rguenther@suse.de>
11133 * match.pd (shift transformations): Remove a redundant
11134 !POINTER_TYPE_P check.
11136 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
11139 * config/i386/i386.md (cmove reg-reg move elimination peephole2s):
11140 Set all_regs to true in the call to replace_rtx.
11142 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
11144 * ipa-modref-tree.c (test_insert_search_collapse): Update test.
11145 * ipa-modref-tree.h (modref_base_node::insert): Be smarter when
11146 hiting --param modref-max-refs limit.
11147 (modref_tree:insert_base): Be smarter when hitting
11148 --param modref-max-bases limit. Add new parameter REF.
11149 (modref_tree:insert): Update.
11150 (modref_tree:merge): Update.
11151 * ipa-modref.c (read_modref_records): Update.
11153 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
11155 * params.opt: (modref-max-adjustments): Add full stop.
11157 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
11159 * ipa-modref-tree.h (modref_ref_node::verify): New member
11161 (modref_ref_node::insert): Use it.
11162 (modref_ref_node::try_mere_with): Fix off by one error.
11164 2021-08-26 Martin Liska <mliska@suse.cz>
11165 Stefan Kneifel <stefan.kneifel@bluewin.ch>
11167 * cgraph.h (create_version_clone_with_body): Add new parameter.
11168 * cgraphclones.c: Likewise.
11169 * multiple_target.c (create_dispatcher_calls): Do not use
11171 (create_target_clone): Likewise here.
11173 2021-08-26 Jonathan Yong <10walls@gmail.com>
11175 * doc/extend.texi: Add note about reserved priorities
11176 to the constructor attribute.
11178 2021-08-25 Martin Sebor <msebor@redhat.com>
11180 * gimple-range-cache.cc (ssa_global_cache::dump): Avoid printing
11181 range table header alone.
11182 * gimple-range.cc (gimple_ranger::export_global_ranges): Same.
11184 2021-08-25 Jan Hubicka <hubicka@ucw.cz>
11186 * doc/invoke.texi: Document --param modref-max-adjustments.
11187 * ipa-modref-tree.c (test_insert_search_collapse): Update.
11188 (test_merge): Update.
11189 * ipa-modref-tree.h (struct modref_access_node): Add adjustments;
11190 (modref_access_node::operator==): Fix handling of access ranges.
11191 (modref_access_node::contains): Constify parameter; handle also
11192 mismatched parm offsets.
11193 (modref_access_node::update): New function.
11194 (modref_access_node::merge): New function.
11195 (unspecified_modref_access_node): Update constructor.
11196 (modref_ref_node::insert_access): Add record_adjustments parameter;
11198 (modref_ref_node::try_merge_with): New private function.
11199 (modref_tree::insert): New record_adjustments parameter.
11200 (modref_tree::merge): New record_adjustments parameter.
11201 (modref_tree::copy_from): Update.
11202 * ipa-modref.c (dump_access): Dump adjustments field.
11203 (get_access): Update constructor.
11204 (record_access): Update call of insert.
11205 (record_access_lto): Update call of insert.
11206 (merge_call_side_effects): Add record_adjustments parameter.
11207 (get_access_for_fnspec): Update.
11208 (process_fnspec): Update.
11209 (analyze_call): Update.
11210 (analyze_function): Update.
11211 (read_modref_records): Update.
11212 (ipa_merge_modref_summary_after_inlining): Update.
11213 (propagate_unknown_call): Update.
11214 (modref_propagate_in_scc): Update.
11215 * params.opt (param-max-modref-adjustments=): New.
11217 2021-08-25 Michael Meissner <meissner@linux.ibm.com>
11219 * config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
11221 (xxspltiw_v4si): Use vecperm type attribute.
11222 (xxspltiw_v4si_inst): Use vecperm type attribute.
11223 (xxspltiw_v4sf_inst): Likewise.
11224 (xxspltidp_v2df): Use vecperm type attribute. Use
11225 UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
11226 (xxspltidp_v2df_inst): Likewise.
11227 (xxsplti32dx_v4si): Use vecperm type attribute.
11228 (xxsplti32dx_v4si_inst): Likewise.
11229 (xxsplti32dx_v4sf_inst): Likewise.
11230 (xxblend_<mode>): Likewise.
11231 (xxpermx): Likewise.
11232 (xxpermx_inst): Likewise.
11233 (xxeval): Likewise.
11235 2021-08-25 Lewis Hyatt <lhyatt@gmail.com>
11238 * coretypes.h (typedef diagnostic_input_charset_callback): Declare.
11239 * diagnostic.c (diagnostic_initialize_input_context): New function.
11240 * diagnostic.h (diagnostic_initialize_input_context): Declare.
11241 * input.c (default_charset_callback): New function.
11242 (file_cache::initialize_input_context): New function.
11243 (file_cache_slot::create): Added ability to convert the input
11244 according to the input context.
11245 (file_cache::file_cache): Initialize the new input context.
11246 (class file_cache_slot): Added new m_alloc_offset member.
11247 (file_cache_slot::file_cache_slot): Initialize the new member.
11248 (file_cache_slot::~file_cache_slot): Handle potentially offset buffer.
11249 (file_cache_slot::maybe_grow): Likewise.
11250 (file_cache_slot::needs_read_p): Handle NULL fp, which is now possible.
11251 (file_cache_slot::get_next_line): Likewise.
11252 * input.h (class file_cache): Added input context member.
11254 2021-08-25 Richard Biener <rguenther@suse.de>
11256 PR tree-optimization/102046
11257 * tree-vect-slp.c (vect_build_slp_tree_2): Conservatively
11258 update ->any_pattern when swapping operands.
11260 2021-08-25 Hongyu Wang <hongyu.wang@intel.com>
11263 * config/i386/i386.c (ix86_live_on_entry): Adjust comment.
11264 (ix86_decompose_address): Remove retval check for ASHIFT,
11265 allow non-canonical zero extend if AND mask covers ASHIFT
11267 (ix86_legitimate_address_p): Adjust condition for decompose.
11268 (ix86_rtx_costs): Adjust cost for lea with non-canonical
11270 Co-Authored by: Uros Bizjak <ubizjak@gmail.com>
11272 2021-08-25 Jiufu Guo <guojiufu@linux.ibm.com>
11274 PR tree-optimization/101145
11275 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
11277 (number_of_iterations_lt): Invoke above function.
11278 (adjust_cond_for_loop_until_wrap):
11279 Merge to number_of_iterations_until_wrap.
11280 (number_of_iterations_cond): Update invokes for
11281 adjust_cond_for_loop_until_wrap and number_of_iterations_lt.
11283 2021-08-25 konglin1 <lingling.kong@intel.com>
11286 * config/i386/avx512dqintrin.h (_mm512_fpclass_ps_mask): Fix
11287 macro define in O0.
11288 (_mm512_mask_fpclass_ps_mask): Ditto.
11290 2021-08-25 Kewen Lin <linkw@linux.ibm.com>
11292 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Remove.
11293 (vec_unpacku_hi_v8hi): Likewise.
11294 (vec_unpacku_lo_v16qi): Likewise.
11295 (vec_unpacku_lo_v8hi): Likewise.
11296 (vec_unpacku_hi_<VP_small_lc>): New define_expand.
11297 (vec_unpacku_lo_<VP_small_lc>): Likewise.
11299 2021-08-24 David Edelsohn <dje.gcc@gmail.com>
11301 * config/rs6000/aix.h (SYSTEM_IMPLICIT_EXTERN_C): Delete.
11302 * config/rs6000/aix71.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
11303 * config/rs6000/aix72.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
11304 * config/rs6000/aix73.h (TARGET_AIX_VERSION): Increase to 73.
11306 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
11308 PR middle-end/102031
11309 * simplify-rtx.c (simplify_truncation): When comparing precisions
11310 use "subreg_prec" variable, not "subreg_mode".
11312 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
11314 * config/rs6000/rs6000-builtin-new.def: Add power10 and power10-64
11317 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
11319 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Initialize
11320 various pointer type nodes.
11321 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add enum
11322 values for various pointer types.
11323 (ptr_V16QI_type_node): New macro.
11324 (ptr_V1TI_type_node): New macro.
11325 (ptr_V2DI_type_node): New macro.
11326 (ptr_V2DF_type_node): New macro.
11327 (ptr_V4SI_type_node): New macro.
11328 (ptr_V4SF_type_node): New macro.
11329 (ptr_V8HI_type_node): New macro.
11330 (ptr_unsigned_V16QI_type_node): New macro.
11331 (ptr_unsigned_V1TI_type_node): New macro.
11332 (ptr_unsigned_V8HI_type_node): New macro.
11333 (ptr_unsigned_V4SI_type_node): New macro.
11334 (ptr_unsigned_V2DI_type_node): New macro.
11335 (ptr_bool_V16QI_type_node): New macro.
11336 (ptr_bool_V8HI_type_node): New macro.
11337 (ptr_bool_V4SI_type_node): New macro.
11338 (ptr_bool_V2DI_type_node): New macro.
11339 (ptr_bool_V1TI_type_node): New macro.
11340 (ptr_pixel_type_node): New macro.
11341 (ptr_intQI_type_node): New macro.
11342 (ptr_uintQI_type_node): New macro.
11343 (ptr_intHI_type_node): New macro.
11344 (ptr_uintHI_type_node): New macro.
11345 (ptr_intSI_type_node): New macro.
11346 (ptr_uintSI_type_node): New macro.
11347 (ptr_intDI_type_node): New macro.
11348 (ptr_uintDI_type_node): New macro.
11349 (ptr_intTI_type_node): New macro.
11350 (ptr_uintTI_type_node): New macro.
11351 (ptr_long_integer_type_node): New macro.
11352 (ptr_long_unsigned_type_node): New macro.
11353 (ptr_float_type_node): New macro.
11354 (ptr_double_type_node): New macro.
11355 (ptr_long_double_type_node): New macro.
11356 (ptr_dfloat64_type_node): New macro.
11357 (ptr_dfloat128_type_node): New macro.
11358 (ptr_ieee128_type_node): New macro.
11359 (ptr_ibm128_type_node): New macro.
11360 (ptr_vector_pair_type_node): New macro.
11361 (ptr_vector_quad_type_node): New macro.
11362 (ptr_long_long_integer_type_node): New macro.
11363 (ptr_long_long_unsigned_type_node): New macro.
11365 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
11367 * config/rs6000/rs6000-builtin-new.def: Add power9-vector, power9,
11368 and power9-64 stanzas.
11370 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
11371 Tom de Vries <tdevries@suse.de>
11373 * config.gcc (nvptx-*-*): Define {c,c++}_target_objs.
11374 * config/nvptx/nvptx-protos.h (nvptx_cpu_cpp_builtins): Prototype.
11375 * config/nvptx/nvptx.h (TARGET_CPU_CPP_BUILTINS): Implement with
11376 a call to the new nvptx_cpu_cpp_builtins function in nvptx-c.c.
11377 * config/nvptx/t-nvptx (nvptx-c.o): New rule.
11378 * config/nvptx/nvptx-c.c: New source file.
11379 (nvptx_cpu_cpp_builtins): Move implementation here.
11381 2021-08-24 Martin Sebor <msebor@redhat.com>
11383 PR middle-end/101600
11384 PR middle-end/101977
11385 * gimple-ssa-warn-access.cc (maybe_warn_for_bound): Tighten up
11386 the phrasing of a warning.
11387 (check_access): Use the remaining size after subtracting any offset
11388 rather than the whole object size.
11389 * pointer-query.cc (access_ref::get_ref): Clear BASE0 flag if it's
11390 clear for any nonnull PHI argument.
11391 (compute_objsize): Clear argument.
11393 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
11395 * config/rs6000/rs6000-builtin-new.def: Add power8-vector stanza.
11397 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
11399 * config/rs6000/rs6000-builtin-new.def: Add power7 and power7-64
11402 2021-08-24 Andrew MacLeod <amacleod@redhat.com>
11404 * value-relation.cc (rr_transitive_table): New.
11405 (relation_transitive): New.
11406 (value_relation::swap): Remove.
11407 (value_relation::apply_transitive): New.
11408 (relation_oracle::relation_oracle): Allocate a new tmp bitmap.
11409 (relation_oracle::register_relation): Call register_transitives.
11410 (relation_oracle::register_transitives): New.
11411 * value-relation.h (relation_oracle): Add new temporary bitmap and
11414 2021-08-24 H.J. Lu <hjl.tools@gmail.com>
11417 * config/i386/i386-expand.c (ix86_expand_vector_move): Broadcast
11418 from integer to a pseudo vector register.
11420 2021-08-24 Richard Biener <rguenther@suse.de>
11422 PR tree-optimization/100089
11423 * tree-vectorizer.h (vect_slp_bb): Rename to ...
11424 (vect_slp_if_converted_bb): ... this and get the original
11425 loop as new argument.
11426 * tree-vectorizer.c (try_vectorize_loop_1): Revert previous fix,
11427 pass original loop to vect_slp_if_converted_bb.
11428 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
11429 If orig_loop was passed scan the not vectorized stmts
11430 for COND_EXPRs and force not profitable if found.
11431 (vect_slp_region): Pass down all SLP instances to costing
11432 if orig_loop was specified.
11433 (vect_slp_bbs): Pass through orig_loop.
11434 (vect_slp_bb): Rename to ...
11435 (vect_slp_if_converted_bb): ... this and get the original
11436 loop as new argument.
11437 (vect_slp_function): Adjust.
11439 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
11442 * config/arm/arm.md (attribute arch): Add fix_vlldm.
11443 (arch_enabled): Use it.
11444 * config/arm/vfp.md (lazy_store_multiple_insn): Add alternative to
11445 use when erratum mitigation is needed.
11447 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
11450 * config/arm/arm.opt (mfix-cmse-cve-2021-35465): New option.
11451 * doc/invoke.texi (Arm Options): Document it.
11452 * config/arm/arm-cpus.in (quirk_vlldm): New feature bit.
11453 (ALL_QUIRKS): Add quirk_vlldm.
11454 (cortex-m33): Add quirk_vlldm.
11455 (cortex-m35p, cortex-m55): Likewise.
11456 * config/arm/arm.c (arm_option_override): Enable fix_vlldm if
11457 targetting an affected CPU and not explicitly controlled on
11460 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
11462 * config/arm/vfp.md (lazy_store_multiple_insn): Rewrite as valid RTL.
11463 (lazy_load_multiple_insn): Likewise.
11465 2021-08-24 liuhongt <hongtao.liu@intel.com>
11468 * config/i386/sse.md (<avx512>_vternlog<mode><sd_maskz_name>):
11469 Enable avx512 embedded broadcast.
11470 (*<avx512>_vternlog<mode>_all): Ditto.
11471 (<avx512>_vternlog<mode>_mask): Ditto.
11473 2021-08-24 liuhongt <hongtao.liu@intel.com>
11476 * config/i386/i386.c (ix86_rtx_costs): Define cost for
11478 * config/i386/i386.h (STRIP_UNARY): New macro.
11479 * config/i386/predicates.md (reg_or_notreg_operand): New
11481 * config/i386/sse.md (*<avx512>_vternlog<mode>_all): New define_insn.
11482 (*<avx512>_vternlog<mode>_1): New pre_reload
11483 define_insn_and_split.
11484 (*<avx512>_vternlog<mode>_2): Ditto.
11485 (*<avx512>_vternlog<mode>_3): Ditto.
11486 (any_logic1,any_logic2): New code iterator.
11487 (logic_op): New code attribute.
11488 (ternlogsuffix): Extend to VNxDF and VNxSF.
11490 2021-08-24 Richard Biener <rguenther@suse.de>
11492 * doc/invoke.texi (vect-inner-loop-cost-factor): Adjust.
11493 * params.opt (--param vect-inner-loop-cost-factor): Adjust
11495 * tree-vect-loop.c (vect_analyze_loop_form): Initialize
11496 inner_loop_cost_factor to the minimum of the estimated number
11497 of iterations of the inner loop and vect-inner-loop-cost-factor.
11499 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
11500 Richard Biener <rguenther@suse.de>
11502 * config/i386/i386-features.c (compute_convert_gain): Provide
11503 more accurate values for CONST_INT, when optimizing for size.
11504 * config/i386/i386.c (COSTS_N_BYTES): Move definition from here...
11505 * config/i386/i386.h (COSTS_N_BYTES): to here.
11507 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
11508 Jakub Jelinek <jakub@redhat.com>
11510 PR middle-end/102029
11511 * match.pd (shift transformations): Add an additional check for
11512 !POINTER_TYPE_P in the recently added left shift transformation.
11514 2021-08-24 liuhongt <hongtao.liu@intel.com>
11516 PR tree-optimization/100089
11517 * tree-vectorizer.c (try_vectorize_loop_1): Disable slp in
11518 loop vectorizer when cost model is very-cheap.
11520 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
11522 * config/rs6000/rs6000-gen-builtins.c (parse_bif_entry): Don't call
11523 asprintf, which is not available on AIX.
11525 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
11527 * config.gcc (target_gtfiles): Add ./rs6000-builtins.h.
11528 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Set.
11530 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
11532 * config.gcc (powerpc*-*-*): Add rs6000-builtins.o to extra_objs.
11533 * config/rs6000/rs6000-gen-builtins.c (main): Close init_file
11535 * config/rs6000/t-rs6000 (rs6000-gen-builtins.o): New target.
11536 (rbtree.o): Likewise.
11537 (rs6000-gen-builtins): Likewise.
11538 (rs6000-builtins.c): Likewise.
11539 (rs6000-builtins.h): Likewise.
11540 (rs6000.o): Add dependency.
11541 (EXTRA_HEADERS): Add rs6000-vecdefines.h.
11542 (rs6000-vecdefines.h): New target.
11543 (rs6000-builtins.o): Likewise.
11544 (rs6000-call.o): Add rs6000-builtins.h as a dependency.
11545 (rs6000-c.o): Likewise.
11547 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
11550 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace):
11551 Diagnose buffer overrun.
11552 (safe_inc_pos): Fix overrun detection.
11553 (match_identifier): Diagnose buffer overrun.
11554 (match_integer): Likewise.
11555 (match_to_right_bracket): Likewise.
11557 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
11559 * ipa-modref-tree.h (modref_access_node::range_info_useful_p):
11560 Improve range compare.
11561 (modref_access_node::contains): New member function.
11562 (modref_access_node::search): Remove.
11563 (modref_access_node::insert): Be smarter about subaccesses.
11565 2021-08-23 Thomas Schwinge <thomas@codesourcery.com>
11567 * config/i386/i386-options.c (ix86_omp_device_kind_arch_isa)
11568 <omp_device_arch> [ACCEL_COMPILER]: Match "intel_mic".
11569 * config/i386/t-omp-device (omp-device-properties-i386) <arch>:
11572 2021-08-23 Jeff Law <jlaw@localhost.localdomain>
11574 * config/h8300/h8300-protos.h (h8300_expand_epilogue): Add new
11576 * config/h8300/jumpcall.md (call, call_value): Restrict to
11577 !SIBLING_CALL_P cases.
11578 (subcall, sibcall_value): New patterns & expanders.
11579 * config/h8300/proepi.md (epilogue): Pass new argument to
11580 h8300_expand_epilogue.
11581 (sibcall_epilogue): New expander.
11582 * config/h8300/h8300.c (h8300_expand_epilogue): Handle sibcall
11584 (h8300_ok_for_sibcall_p): New function.
11585 (TARGET_FUNCTION_OK_FOR_SIBCALL): define.
11587 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
11589 * simplify-rtx.c (simplify_unary_operation_1): [TRUNCATE]:
11590 Handle case where the operand is already the desired mode.
11592 2021-08-23 Richard Biener <rguenther@suse.de>
11595 * tree-ssa-structalias.c (ipa_pta_execute): Check in_other_partition
11596 in addition to has_gimple_body.
11598 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
11600 PR middle-end/101949
11601 * ipa-modref.c (analyze_ssa_name_flags): Fix merging of
11604 2021-08-23 Martin Liska <mliska@suse.cz>
11606 * doc/invoke.texi: Put the option out of -mxl-mode-app-model
11609 2021-08-23 Richard Biener <rguenther@suse.de>
11611 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
11612 Properly scale the inner loop cost only once.
11614 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
11616 * tree-ssa-ccp.c (bit_value_binop) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]:
11617 Provide bounds for unsigned (and signed with non-negative operands)
11618 division and modulus.
11620 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
11622 * simplify-rtx.c (simplify_truncation): Generalize simplification
11623 of (truncate:A (subreg:B X)).
11624 (simplify_unary_operation_1) [FLOAT_TRUNCATE, FLOAT_EXTEND,
11625 SIGN_EXTEND, ZERO_EXTEND]: Handle cases where the operand
11626 already has the desired machine mode.
11627 (test_scalar_int_ops): Add tests that useless extensions and
11628 truncations are optimized away.
11629 (test_scalar_int_ext_ops): New self-test function to confirm
11630 that truncations of extensions are correctly simplified.
11631 (test_scalar_int_ext_ops2): New self-test function to check
11632 truncations of truncations, extensions of extensions, and
11633 truncations of extensions.
11634 (test_scalar_ops): Call the above two functions with a
11635 representative sampling of integer machine modes.
11637 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
11639 * match.pd (shift transformations): Change the sign of an
11640 LSHIFT_EXPR if it reduces the number of explicit conversions.
11642 2021-08-23 Jakub Jelinek <jakub@redhat.com>
11644 PR tree-optimization/86723
11645 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
11646 cast64_to_32 argument, set *cast64_to_32 to false, unless n is
11647 non-memory permutation of 64-bit src which only has bytes of
11648 0 or [5..8] and n->range is 4.
11649 (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
11650 find_bswap_or_nop_finalize caller, support bswap with some bytes
11651 zeroed, as long as at least two bytes are not zeroed.
11652 (bswap_replace): Add mask argument and handle masking of bswap
11654 (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
11655 caller, punt if cast64_to_32 or mask is not all ones.
11656 (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
11657 caller, for now punt if cast64_to_32.
11659 2021-08-23 Richard Biener <rguenther@suse.de>
11661 PR tree-optimization/79334
11662 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Record
11663 a type also for COMPONENT_REFs.
11664 (vn_reference_may_trap): Check ARRAY_REF with constant index
11665 against the array domain.
11667 2021-08-23 liuhongt <hongtao.liu@intel.com>
11670 * config/i386/sse.md (*avx512f_pshufb_truncv8hiv8qi_1): Add
11671 TARGET_AVX512BW to condition.
11673 2021-08-23 Jakub Jelinek <jakub@redhat.com>
11676 * dwarf2out.c (gen_variable_die): Add DW_AT_location for global
11677 register variables already during early_dwarf if possible.
11679 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
11681 * config/arm/arm_mve.h: Fix __arm_vctp16q return type.
11683 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
11686 * config/arm/arm.opt: Fix typo.
11687 * config/arm/t-rmprofile: Fix typo.
11689 2021-08-23 Jakub Jelinek <jakub@redhat.com>
11691 * tree.h (OMP_CLAUSE_GRAINSIZE_STRICT): Define.
11692 (OMP_CLAUSE_NUM_TASKS_STRICT): Define.
11693 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_GRAINSIZE,
11694 case OMP_CLAUSE_NUM_TASKS>: Print strict: modifier.
11695 * omp-expand.c (expand_task_call): Use GOMP_TASK_FLAG_STRICT in iflags
11696 if either grainsize or num_tasks clause has the strict modifier.
11698 2021-08-23 Martin Liska <mliska@suse.cz>
11700 * dbgcnt.def (DEBUG_COUNTER): New counter.
11701 * gimple.c (gimple_call_arg_flags): Use it in IPA PTA.
11703 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
11705 * ipa-modref.c (analyze_ssa_name_flags): Improve handling of return slot.
11707 2021-08-23 Xi Ruoyao <xry111@mengyan1223.wang>
11710 * config/mips/mips-protos.h (mips_msa_output_shift_immediate):
11712 * config/mips/mips.c (mips_msa_output_shift_immediate): New
11714 * config/mips/mips-msa.md (vashl<mode>3, vashr<mode>3,
11715 vlshr<mode>3): Call it.
11717 2021-08-22 Jan Hubicka <hubicka@ucw.cz>
11718 Martin Liska <mliska@suse.cz>
11720 PR middle-end/101949
11721 * ipa-modref.c (analyze_ssa_name_flags): Indirect call implies
11724 2021-08-21 Dragan Mladjenovic <OT_Dragan.Mladjenovic@mediatek.com>
11726 * config/mips/mips.c (mips_function_rodata_section,
11727 TARGET_ASM_FUNCTION_RODATA_SECTION): Removed.
11729 2021-08-21 John David Anglin <danglin@gcc.gnu.org>
11731 * config/pa/pa.c (pa_asm_output_aligned_common): Remove warning.
11733 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
11735 * configure.ac (thread-local storage support): Remove tls_first_major
11736 and tls_first_minor. Use "$conftest_s" to check support.
11737 * configure: Regenerate.
11739 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
11741 * configure.ac: Fixup formatting.
11743 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
11745 * acinclude.m4 (gcc_GAS_CHECK_FEATURE): Remove third argument and ...
11746 * configure.ac: ... update all callers.
11748 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
11751 * acinclude.m4 (_gcc_COMPUTE_GAS_VERSION, _gcc_GAS_VERSION_GTE_IFELSE)
11752 (gcc_GAS_VERSION_GTE_IFELSE): Remove.
11753 (gcc_GAS_CHECK_FEATURE): Do not handle in-tree case specially.
11754 * configure.ac: Remove gcc_cv_gas_major_version, gcc_cv_gas_minor_version.
11755 Remove remaining checks for in-tree assembler.
11756 * configure: Regenerate.
11758 2021-08-20 Jeff Law <jlaw@localhost.localdomain>
11760 * config/h8300/h8300.c (shift_alg_hi): Improve arithmetic shift right
11761 by 15 bits for H8/300H and H8/S. Improve logical shifts by 12
11763 (shift_alg_si): Improve arithmetic right shift by 28-30 bits for
11764 H8/300H. Improve arithmetic shift right by 15 bits for H8/S.
11765 Improve logical shifts by 27 bits for H8/S.
11766 (get_shift_alg): Corresponding changes.
11767 (h8300_option_override): Revert to loops for -Os when profitable.
11769 2021-08-20 Richard Biener <rguenther@suse.de>
11771 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
11773 (vect_analyze_data_ref_accesses): Likewise. Assign the BB
11774 index as group_id when dataref_groups were not computed.
11775 * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
11776 we advace to the next BB.
11778 2021-08-20 Jakub Jelinek <jakub@redhat.com>
11780 * omp-builtins.def (BUILT_IN_GOMP_WARNING, BUILT_IN_GOMP_ERROR): New
11783 2021-08-20 Martin Liska <mliska@suse.cz>
11785 PR gcov-profile/89961
11786 * gcov.c (make_gcov_file_name): Rewrite using std::string.
11787 (mangle_name): Simplify, do not used the second argument.
11788 (strip_extention): New function.
11789 (get_md5sum): Likewise.
11790 (get_gcov_intermediate_filename): Handle properly -p and -x
11792 (output_gcov_file): Use string type.
11793 (generate_results): Likewise.
11794 (md5sum_to_hex): Remove.
11796 2021-08-20 Michael Meissner <meissner@linux.ibm.com>
11798 * config/rs6000/altivec.md (UNSPEC_XXEVAL): Move to vsx.md.
11799 (UNSPEC_XXSPLTIW): Move to vsx.md.
11800 (UNSPEC_XXSPLTID): Move to vsx.md.
11801 (UNSPEC_XXSPLTI32DX): Move to vsx.md.
11802 (UNSPEC_XXBLEND): Move to vsx.md.
11803 (UNSPEC_XXPERMX): Move to vsx.md.
11804 (VM3): Move to vsx.md.
11805 (VM3_char): Move to vsx.md.
11806 (xxspltiw_v4si): Move to vsx.md.
11807 (xxspltiw_v4sf): Move to vsx.md.
11808 (xxspltiw_v4sf_inst): Move to vsx.md.
11809 (xxspltidp_v2df): Move to vsx.md.
11810 (xxspltidp_v2df_inst): Move to vsx.md.
11811 (xxsplti32dx_v4si_inst): Move to vsx.md.
11812 (xxsplti32dx_v4sf): Move to vsx.md.
11813 (xxsplti32dx_v4sf_inst): Move to vsx.md.
11814 (xxblend_<mode>): Move to vsx.md.
11815 (xxpermx): Move to vsx.md.
11816 (xxpermx_inst): Move to vsx.md.
11817 * config/rs6000/vsx.md (UNSPEC_XXEVAL): Move from altivec.md.
11818 (UNSPEC_XXSPLTIW): Move from altivec.md.
11819 (UNSPEC_XXSPLTID): Move from altivec.md.
11820 (UNSPEC_XXSPLTI32DX): Move from altivec.md.
11821 (UNSPEC_XXBLEND): Move from altivec.md.
11822 (UNSPEC_XXPERMX): Move from altivec.md.
11823 (VM3): Move from altivec.md.
11824 (VM3_char): Move from altivec.md.
11825 (xxspltiw_v4si): Move from altivec.md.
11826 (xxspltiw_v4sf): Move from altivec.md.
11827 (xxspltiw_v4sf_inst): Move from altivec.md.
11828 (xxspltidp_v2df): Move from altivec.md.
11829 (xxspltidp_v2df_inst): Move from altivec.md.
11830 (xxsplti32dx_v4si_inst): Move from altivec.md.
11831 (xxsplti32dx_v4sf): Move from altivec.md.
11832 (xxsplti32dx_v4sf_inst): Move from altivec.md.
11833 (xxblend_<mode>): Move from altivec.md.
11834 (xxpermx): Move from altivec.md.
11835 (xxpermx_inst): Move from altivec.md.
11837 2021-08-19 Roger Sayle <roger@nextmovesoftware.com>
11839 * tree-vect-generic.c (expand_vector_operations_1): Use either
11840 gimplify_build1 or gimplify_build2 instead of gimple_build_assign
11841 when constructing scalar splat expressions.
11843 2021-08-19 Peter Bergner <bergner@linux.ibm.com>
11846 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Cast
11847 pointer to __vector_pair *.
11849 2021-08-19 Martin Sebor <msebor@redhat.com>
11851 * gimple-range.cc: Add comments.
11852 * gimple-range.h: Same.
11854 2021-08-19 Martin Sebor <msebor@redhat.com>
11856 PR middle-end/101984
11857 * gimple-ssa-warn-access.cc (pass_waccess::execute): Also call
11860 2021-08-19 Jeff Law <jlaw@localhost.localdomain>
11862 * config.gcc (h8300-*-elf*): Do not include dbxelf.h.
11863 (h8300-*-linux*, v850-*-rtems*, v850*-elf*): Likewise.
11864 * config/v850/v850.h (DEFAULT_GDB_EXTENSIONS): Remove.
11866 2021-08-19 Jakub Jelinek <jakub@redhat.com>
11868 PR middle-end/101950
11869 * optabs.c (expand_clrsb_using_clz): New function.
11870 (expand_unop): Use it as another clrsb expansion fallback.
11872 2021-08-19 liuhongt <hongtao.liu@intel.com>
11875 2021-07-28 liuhongt <hongtao.liu@intel.com>
11878 * config/i386/i386.h (processor_costs): Add new member
11880 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
11881 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
11882 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
11883 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
11884 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
11885 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
11886 generic_cost, core_cost): Initialize integer_to_sse same value
11888 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
11889 * config/i386/i386.c (ix86_builtin_vectorization_cost):
11890 Use integer_to_sse instead of sse_op to calculate the cost of
11893 2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
11895 * config.gcc: Include rpath.opt for Darwin.
11896 * config/darwin.h (DRIVER_SELF_SPECS): Handle -rpath.
11898 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
11900 PR bootstrap/101959
11901 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
11904 2021-08-18 Jonathan Wright <jonathan.wright@arm.com>
11906 * config/aarch64/arm_neon.h (vld3_lane_f64): Use float RTL
11907 pattern and type cast.
11908 (vld4_lane_f32): Use float RTL pattern.
11909 (vld4q_lane_f64): Use float type cast.
11911 2021-08-18 Jan Hubicka <hubicka@ucw.cz>
11913 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Check also
11916 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
11918 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor): Extend.
11919 (test_map_of_type_with_ctor_and_dtor_expand): Add function.
11920 (hash_map_tests_c_tests): Call it.
11922 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
11924 * ggc.h (enum ggc_collect): New.
11925 (ggc_collect): Use it.
11926 * ggc-page.c: Adjust.
11927 * ggc-common.c: Likewise.
11928 * ggc-tests.c: Likewise.
11929 * read-rtl-function.c: Likewise.
11930 * selftest-run-tests.c: Likewise.
11931 * doc/gty.texi (Invoking the garbage collector): Likewise.
11933 2021-08-18 liuhongt <hongtao.liu@intel.com>
11936 * config/i386/i386.h (TARGET_V2DF_REDUCTION_PREFER_HADDPD):
11938 * config/i386/sse.md (*sse3_haddv2df3_low): Add
11939 TARGET_V2DF_REDUCTION_PREFER_HADDPD.
11940 (*sse3_hsubv2df3_low): Ditto.
11941 * config/i386/x86-tune.def
11942 (X86_TUNE_V2DF_REDUCTION_PREFER_HADDPD): New tune.
11944 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
11946 * gimple-range-gori.cc (gori_compute::gori_compute): Enable tracing.
11947 (gori_compute::compute_operand_range): Add tracing.
11948 (gori_compute::logical_combine): Ditto.
11949 (gori_compute::compute_logical_operands): Ditto.
11950 (gori_compute::compute_operand1_range): Ditto.
11951 (gori_compute::compute_operand2_range): Ditto.
11952 (gori_compute::outgoing_edge_range_p): Ditto.
11953 * gimple-range-gori.h (class gori_compute): Add range_tracer.
11955 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
11957 * flag-types.h (enum evrp_mode): Adjust evrp-mode values.
11958 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Relocate from.
11959 * gimple-range-trace.h (DEBUG_RANGE_CACHE): Here.
11960 * params.opt (--param=evrp-mode): Adjust options.
11962 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
11964 * Makefile.in (OBJS): Add gimple-range-trace.o.
11965 * gimple-range-cache.h (enable_new_values): Remove unused prototype.
11966 * gimple-range-fold.cc: Adjust headers.
11967 * gimple-range-trace.cc: New.
11968 * gimple-range-trace.h: New.
11969 * gimple-range.cc (gimple_ranger::gimple_ranger): Enable tracer.
11970 (gimple_ranger::range_of_expr): Add tracing.
11971 (gimple_ranger::range_on_entry): Ditto.
11972 (gimple_ranger::range_on_exit): Ditto.
11973 (gimple_ranger::range_on_edge): Ditto.
11974 (gimple_ranger::fold_range_internal): Ditto.
11975 (gimple_ranger::dump_bb): Do not calculate edge range twice.
11976 (trace_ranger::*): Remove.
11977 (enable_ranger): Never create a trace_ranger.
11978 (debug_seed_ranger): Move to gimple-range-trace.cc.
11979 (dump_ranger): Ditto.
11980 (debug_ranger): Ditto.
11981 * gimple-range.h: Include gimple-range-trace.h.
11982 (range_on_entry, range_on_exit): No longer virtual.
11983 (class trace_ranger): Remove.
11984 (DEBUG_RANGE_CACHE): Move to gimple-range-trace.h.
11986 2021-08-17 Martin Sebor <msebor@redhat.com>
11988 PR middle-end/101854
11989 * builtins.c (expand_builtin_alloca): Move warning code to check_alloca
11990 in gimple-ssa-warn-access.cc.
11991 * calls.c (alloc_max_size): Move code to check_alloca.
11992 (get_size_range): Move to pointer-query.cc.
11993 (maybe_warn_alloc_args_overflow): Move to gimple-ssa-warn-access.cc.
11994 (get_attr_nonstring_decl): Move to tree.c.
11995 (fntype_argno_type): Move to gimple-ssa-warn-access.cc.
11996 (append_attrname): Same.
11997 (maybe_warn_rdwr_sizes): Same.
11998 (initialize_argument_information): Move code to
11999 gimple-ssa-warn-access.cc.
12000 * calls.h (maybe_warn_alloc_args_overflow): Move to
12001 gimple-ssa-warn-access.h.
12002 (get_attr_nonstring_decl): Move to tree.h.
12003 (maybe_warn_nonstring_arg): Move to gimple-ssa-warn-access.h.
12004 (enum size_range_flags): Move to pointer-query.h.
12005 (get_size_range): Same.
12006 * gimple-ssa-warn-access.cc (has_location): Remove unused overload
12007 to avoid Clang -Wunused-function.
12008 (get_size_range): Declare static.
12009 (maybe_emit_free_warning): Rename...
12010 (maybe_check_dealloc_call): ...to this for consistency.
12011 (class pass_waccess): Add members.
12012 (pass_waccess::~pass_waccess): Defined.
12013 (alloc_max_size): Move here from calls.c.
12014 (maybe_warn_alloc_args_overflow): Same.
12015 (check_alloca): New function.
12016 (check_alloc_size_call): New function.
12017 (check_strncat): Handle another warning flag.
12018 (pass_waccess::check_builtin): Handle alloca.
12019 (fntype_argno_type): Move here from calls.c.
12020 (append_attrname): Same.
12021 (maybe_warn_rdwr_sizes): Same.
12022 (pass_waccess::check_call): Define.
12023 (check_nonstring_args): New function.
12024 (pass_waccess::check): Call new member functions.
12025 (pass_waccess::execute): Enable ranger.
12026 * gimple-ssa-warn-access.h (get_size_range): Move here from calls.h.
12027 (maybe_warn_nonstring_arg): Same.
12028 * gimple-ssa-warn-restrict.c: Remove #include.
12029 * pointer-query.cc (get_size_range): Move here from calls.c.
12030 * pointer-query.h (enum size_range_flags): Same.
12031 (get_size_range): Same.
12032 * tree.c (get_attr_nonstring_decl): Move here from calls.c.
12033 * tree.h (get_attr_nonstring_decl): Move here from calls.h.
12035 2021-08-17 Thomas Schwinge <thomas@codesourcery.com>
12037 * ggc.h (ggc_collect): Add 'force_collect' parameter.
12038 * ggc-page.c (ggc_collect): Use that one instead of global
12039 'ggc_force_collect'. Adjust all users.
12040 * doc/gty.texi (Invoking the garbage collector): Update.
12041 * ggc-internal.h (ggc_force_collect): Remove.
12042 * ggc-common.c (ggc_force_collect): Likewise.
12043 * selftest.h (forcibly_ggc_collect): Remove.
12044 * ggc-tests.c (selftest::forcibly_ggc_collect): Likewise.
12045 * read-rtl-function.c (test_loading_labels): Adjust.
12046 * selftest-run-tests.c (run_tests): Likewise.
12048 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
12050 * config/darwin.c (darwin_file_end): Reset and reclaim the
12051 section names table at the end of compile.
12053 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
12056 * config.in: Regenerate.
12057 * config/i386/darwin.h (EXTRA_ASM_OPTS): New
12058 (ASM_SPEC): Pass options to disable branch shortening where
12060 * configure: Regenerate.
12061 * configure.ac: Detect versions of 'as' that support the
12062 optimisation which has the bug.
12064 2021-08-17 Richard Biener <rguenther@suse.de>
12066 * optabs-query.c (supports_vec_gather_load_p): Also check
12068 (supports_vec_scatter_store_p): Likewise.
12069 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Fall
12070 back to masked variants if non-masked are not supported.
12071 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
12072 When we need to use masked gather/scatter but do not have
12073 a mask set up a constant true one.
12074 * tree-vect-stmts.c (vect_check_scalar_mask): Also allow
12075 non-SSA_NAME masks.
12077 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
12079 * tree-ssa-ccp.c (bit_value_binop) [MINUS_EXPR]: Use same
12080 algorithm as PLUS_EXPR to improve subtraction bit bounds.
12081 [POINTER_DIFF_EXPR]: Treat as synonymous with MINUS_EXPR.
12083 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
12085 * tree-ssa-ccp.c (bit_value_mult_const): New helper function to
12086 calculate the mask-value pair result of a multiplication by an
12088 (bit_value_binop) [MULT_EXPR]: Call it from here for
12089 multiplications by (sparse) non-negative constants.
12091 2021-08-17 Christophe Lyon <christophe.lyon@foss.st.com>
12094 * config.gcc (gcc_cv_initfini_array): Leave undefined for
12095 uclinuxfdpiceabi targets.
12097 2021-08-17 Alexandre Oliva <oliva@adacore.com>
12099 * tree-inline.c (maybe_move_debug_stmts_to_successors): Don't
12100 reverse debug stmts.
12102 2021-08-17 Alexandre Oliva <oliva@adacore.com>
12104 * tree-cfg.c (dump_function_to_file): Use fun, not cfun.
12106 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
12108 * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete.
12109 (__LD4Q_LANE_FUNC): Likewise.
12110 (vld4_lane_u8): Define without macro.
12111 (vld4_lane_u16): Likewise.
12112 (vld4_lane_u32): Likewise.
12113 (vld4_lane_u64): Likewise.
12114 (vld4_lane_s8): Likewise.
12115 (vld4_lane_s16): Likewise.
12116 (vld4_lane_s32): Likewise.
12117 (vld4_lane_s64): Likewise.
12118 (vld4_lane_f16): Likewise.
12119 (vld4_lane_f32): Likewise.
12120 (vld4_lane_f64): Likewise.
12121 (vld4_lane_p8): Likewise.
12122 (vld4_lane_p16): Likewise.
12123 (vld4_lane_p64): Likewise.
12124 (vld4q_lane_u8): Likewise.
12125 (vld4q_lane_u16): Likewise.
12126 (vld4q_lane_u32): Likewise.
12127 (vld4q_lane_u64): Likewise.
12128 (vld4q_lane_s8): Likewise.
12129 (vld4q_lane_s16): Likewise.
12130 (vld4q_lane_s32): Likewise.
12131 (vld4q_lane_s64): Likewise.
12132 (vld4q_lane_f16): Likewise.
12133 (vld4q_lane_f32): Likewise.
12134 (vld4q_lane_f64): Likewise.
12135 (vld4q_lane_p8): Likewise.
12136 (vld4q_lane_p16): Likewise.
12137 (vld4q_lane_p64): Likewise.
12138 (vld4_lane_bf16): Likewise.
12139 (vld4q_lane_bf16): Likewise.
12141 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
12143 * config/aarch64/arm_neon.h (__LD3_LANE_FUNC): Delete.
12144 (__LD3Q_LANE_FUNC): Delete.
12145 (vld3_lane_u8): Define without macro.
12146 (vld3_lane_u16): Likewise.
12147 (vld3_lane_u32): Likewise.
12148 (vld3_lane_u64): Likewise.
12149 (vld3_lane_s8): Likewise.
12150 (vld3_lane_s16): Likewise.
12151 (vld3_lane_s32): Likewise.
12152 (vld3_lane_s64): Likewise.
12153 (vld3_lane_f16): Likewise.
12154 (vld3_lane_f32): Likewise.
12155 (vld3_lane_f64): Likewise.
12156 (vld3_lane_p8): Likewise.
12157 (vld3_lane_p16): Likewise.
12158 (vld3_lane_p64): Likewise.
12159 (vld3q_lane_u8): Likewise.
12160 (vld3q_lane_u16): Likewise.
12161 (vld3q_lane_u32): Likewise.
12162 (vld3q_lane_u64): Likewise.
12163 (vld3q_lane_s8): Likewise.
12164 (vld3q_lane_s16): Likewise.
12165 (vld3q_lane_s32): Likewise.
12166 (vld3q_lane_s64): Likewise.
12167 (vld3q_lane_f16): Likewise.
12168 (vld3q_lane_f32): Likewise.
12169 (vld3q_lane_f64): Likewise.
12170 (vld3q_lane_p8): Likewise.
12171 (vld3q_lane_p16): Likewise.
12172 (vld3q_lane_p64): Likewise.
12173 (vld3_lane_bf16): Likewise.
12174 (vld3q_lane_bf16): Likewise.
12176 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
12178 * config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Delete.
12179 (__LD2Q_LANE_FUNC): Likewise.
12180 (vld2_lane_u8): Define without macro.
12181 (vld2_lane_u16): Likewise.
12182 (vld2_lane_u32): Likewise.
12183 (vld2_lane_u64): Likewise.
12184 (vld2_lane_s8): Likewise.
12185 (vld2_lane_s16): Likewise.
12186 (vld2_lane_s32): Likewise.
12187 (vld2_lane_s64): Likewise.
12188 (vld2_lane_f16): Likewise.
12189 (vld2_lane_f32): Likewise.
12190 (vld2_lane_f64): Likewise.
12191 (vld2_lane_p8): Likewise.
12192 (vld2_lane_p16): Likewise.
12193 (vld2_lane_p64): Likewise.
12194 (vld2q_lane_u8): Likewise.
12195 (vld2q_lane_u16): Likewise.
12196 (vld2q_lane_u32): Likewise.
12197 (vld2q_lane_u64): Likewise.
12198 (vld2q_lane_s8): Likewise.
12199 (vld2q_lane_s16): Likewise.
12200 (vld2q_lane_s32): Likewise.
12201 (vld2q_lane_s64): Likewise.
12202 (vld2q_lane_f16): Likewise.
12203 (vld2q_lane_f32): Likewise.
12204 (vld2q_lane_f64): Likewise.
12205 (vld2q_lane_p8): Likewise.
12206 (vld2q_lane_p16): Likewise.
12207 (vld2q_lane_p64): Likewise.
12208 (vld2_lane_bf16): Likewise.
12209 (vld2q_lane_bf16): Likewise.
12211 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
12213 * haifa-sched.c (advance_one_cycle): Output more context-synchronization
12216 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
12218 * haifa-sched.c (enum rfs_decision, rfs_str): Add RFS_AUTOPREF.
12219 (rank_for_schedule): Use it.
12221 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
12223 PR rtl-optimization/91598
12224 * haifa-sched.c (autopref_rank_for_schedule): Prioritize "irrelevant"
12225 insns after memory reads and before memory writes.
12227 2021-08-17 Alistair_Lee <alistair.lee@arm.com>
12229 * rtl.h (CONST_VECTOR_P): New macro.
12230 * config/aarch64/aarch64.c (aarch64_get_sve_pred_bits): Use RTL
12231 code testing macros.
12232 (aarch64_ptrue_all_mode): Likewise.
12233 (aarch64_expand_mov_immediate): Likewise.
12234 (aarch64_const_vec_all_in_range_p): Likewise.
12235 (aarch64_rtx_costs): Likewise.
12236 (aarch64_legitimate_constant_p): Likewise.
12237 (aarch64_simd_valid_immediate): Likewise.
12238 (aarch64_simd_make_constant): Likewise.
12239 (aarch64_convert_mult_to_shift): Likewise.
12240 (aarch64_expand_sve_vec_perm): Likewise.
12241 (aarch64_vec_fpconst_pow_of_2): Likewise.
12243 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
12245 PR tree-optimization/101938
12246 * range-op.cc (operator_abs::op1_range): Special case
12247 -TYPE_MIN_VALUE for flag_wrapv.
12249 2021-08-17 Kewen Lin <linkw@linux.ibm.com>
12251 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Add the cost for
12254 2021-08-17 Jakub Jelinek <jakub@redhat.com>
12256 * tree.def (OMP_SCOPE): New tree code.
12257 * tree.h (OMP_SCOPE_BODY, OMP_SCOPE_CLAUSES): Define.
12258 * tree-nested.c (convert_nonlocal_reference_stmt,
12259 convert_local_reference_stmt, convert_gimple_call): Handle
12261 * tree-pretty-print.c (dump_generic_node): Handle OMP_SCOPE.
12262 * gimple.def (GIMPLE_OMP_SCOPE): New gimple code.
12263 * gimple.c (gimple_build_omp_scope): New function.
12264 (gimple_copy): Handle GIMPLE_OMP_SCOPE.
12265 * gimple.h (gimple_build_omp_scope): Declare.
12266 (gimple_has_substatements): Handle GIMPLE_OMP_SCOPE.
12267 (gimple_omp_scope_clauses, gimple_omp_scope_clauses_ptr,
12268 gimple_omp_scope_set_clauses): New inline functions.
12269 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_SCOPE.
12270 * gimple-pretty-print.c (dump_gimple_omp_scope): New function.
12271 (pp_gimple_stmt_1): Handle GIMPLE_OMP_SCOPE.
12272 * gimple-walk.c (walk_gimple_stmt): Likewise.
12273 * gimple-low.c (lower_stmt): Likewise.
12274 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
12275 (gimplify_scan_omp_clauses): For task reductions, handle OMP_SCOPE
12276 like ORT_WORKSHARE constructs. Adjust diagnostics for %<scope%>
12277 allowing task reductions. Reject inscan reductions on scope.
12278 (omp_find_stores_stmt): Handle GIMPLE_OMP_SCOPE.
12279 (gimplify_omp_workshare, gimplify_expr): Handle OMP_SCOPE.
12280 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_SCOPE.
12281 (estimate_num_insns): Likewise.
12282 * omp-low.c (build_outer_var_ref): Look through GIMPLE_OMP_SCOPE
12283 contexts if var isn't privatized there.
12284 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_SCOPE.
12285 (scan_omp_1_stmt): Likewise.
12286 (maybe_add_implicit_barrier_cancel): Look through outer
12288 (lower_omp_scope): New function.
12289 (lower_omp_task_reductions): Handle OMP_SCOPE.
12290 (lower_omp_1): Handle GIMPLE_OMP_SCOPE.
12291 (diagnose_sb_1, diagnose_sb_2): Likewise.
12292 * omp-expand.c (expand_omp_single): Support also GIMPLE_OMP_SCOPE.
12293 (expand_omp): Handle GIMPLE_OMP_SCOPE.
12294 (omp_make_gimple_edges): Likewise.
12295 * omp-builtins.def (BUILT_IN_GOMP_SCOPE_START): New built-in.
12297 2021-08-17 Richard Biener <rguenther@suse.de>
12299 PR tree-optimization/101925
12300 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Set
12301 reverse on COMPONENT_REF and ARRAY_REF according to
12302 what reverse_storage_order_for_component_p does.
12303 (vn_reference_eq): Compare reversed on reference ops.
12304 (reverse_storage_order_for_component_p): New overload.
12305 (vn_reference_lookup_3): Check reverse_storage_order_for_component_p
12306 on the reference looked up.
12308 2021-08-17 Jeff Law <jlaw@localhost.localdomain>
12310 * config/h8300/h8300.c (shift_alg_si): Avoid loops for most SImode
12311 shifts on the H8/S.
12312 (h8300_option_override): Use loops on H8/S more often when optimizing
12314 (get_shift_alg): Handle new "special" cases on H8/S. Simplify
12315 accordingly. Handle various arithmetic right shifts with special
12316 sequences that we couldn't handle before.
12318 2021-08-16 Jeff Law <jlaw@localhost.localdomain>
12320 * config.gcc (rl78-*-elf*): Do not include dbxelf.h.
12322 2021-08-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
12324 * config/sparc/rtemself.h (SPARC_GCOV_TYPE_SIZE): Define.
12325 * config/sparc/sparc.c (sparc_gcov_type_size): New.
12326 (TARGET_GCOV_TYPE_SIZE): Redefine if SPARC_GCOV_TYPE_SIZE is defined.
12327 * coverage.c (get_gcov_type): Use targetm.gcov_type_size().
12328 * doc/tm.texi (TARGET_GCOV_TYPE_SIZE): Add hook under "Misc".
12329 * doc/tm.texi.in: Regenerate.
12330 * target.def (gcov_type_size): New target hook.
12331 * targhooks.c (default_gcov_type_size): New.
12332 * targhooks.h (default_gcov_type_size): Declare.
12333 * tree-profile.c (gimple_gen_edge_profiler): Use precision of
12335 (gimple_gen_time_profiler): Likewise.
12337 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
12339 * dwarf2out.c (add_scalar_info): Deal with DW_AT_data_bit_offset.
12341 2021-08-16 Tobias Burnus <tobias@codesourcery.com>
12343 PR middle-end/101931
12344 * omp-low.c (omp_runtime_api_call): Update for routines
12345 added in the meanwhile.
12347 2021-08-16 Martin Liska <mliska@suse.cz>
12349 PR tree-optimization/100393
12350 * tree-switch-conversion.c (group_cluster::dump): Use
12351 get_comparison_count.
12352 (jump_table_cluster::find_jump_tables): Pre-compute number of
12353 comparisons and then decrement it. Cache also max_ratio.
12354 (jump_table_cluster::can_be_handled): Change signature.
12355 * tree-switch-conversion.h (get_comparison_count): New.
12357 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
12359 * dwarf2out.c (add_data_member_location_attribute): Use GNAT
12360 encodings only when -fgnat-encodings=all is specified.
12361 (add_bound_info): Likewise.
12362 (add_byte_size_attribute): Likewise.
12363 (gen_member_die): Likewise.
12365 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
12367 * omp-oacc-neuter-broadcast.cc
12368 (execute_omp_oacc_neuter_broadcast): Plug 'par' memory leak.
12370 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
12372 * omp-oacc-neuter-broadcast.cc
12373 (execute_omp_oacc_neuter_broadcast): Clarify memory management for
12376 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
12378 * omp-oacc-neuter-broadcast.cc (field_map): Move variable into...
12379 (execute_omp_oacc_neuter_broadcast): ... here.
12380 (install_var_field, build_receiver_ref, build_sender_ref): Take
12381 'field_map_t *' parameter. Adjust all users.
12382 (worker_single_copy, neuter_worker_single): Take a
12383 'record_field_map_t *' parameter. Adjust all users.
12385 2021-08-16 liuhongt <hongtao.liu@intel.com>
12388 * config/i386/i386.md (ldexp<mode>3): Force operands[1] to
12391 2021-08-16 Martin Liska <mliska@suse.cz>
12394 * multiple_target.c (create_dispatcher_calls): Make default
12395 function local only if it is a definition.
12397 2021-08-16 Martin Liska <mliska@suse.cz>
12400 * ipa-icf-gimple.c (func_checker::compare_ssa_name): Do not
12401 consider equal SSA_NAMEs when one is a param.
12403 2021-08-16 liuhongt <hongtao.liu@intel.com>
12406 * config/i386/i386-expand.c (ix86_expand_vec_perm_vpermt2):
12407 Support vpermi2b for V32QI/V16QImode.
12408 (ix86_extract_perm_from_pool_constant): New function.
12409 (ix86_expand_vec_one_operand_perm_avx512): Support
12410 vpermw/vpermb under TARGET_AVX512BW/TARGET_AVX512VBMI.
12411 (expand_vec_perm_1): Adjust comments for upper.
12412 * config/i386/i386-protos.h (ix86_extract_perm_from_pool_constant):
12414 * config/i386/predicates.md (permvar_truncate_operand): New predicate.
12415 (pshufb_truncv4siv4hi_operand): Ditto.
12416 (pshufb_truncv8hiv8qi_operand): Ditto.
12417 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1):
12418 New pre_reload define_insn_and_split.
12419 (*avx512f_permvar_truncv8siv8hi_1): Ditto.
12420 (*avx512f_vpermvar_truncv8div8si_1): Ditto.
12421 (*avx512f_permvar_truncv32hiv32qi_1): Ditto.
12422 (*avx512f_permvar_truncv16hiv16qi_1): Ditto.
12423 (*avx512f_permvar_truncv4div4si_1): Ditto.
12424 (*avx512f_pshufb_truncv8hiv8qi_1): Ditto.
12425 (*avx512f_pshufb_truncv4siv4hi_1): Ditto.
12426 (*avx512f_pshufd_truncv2div2si_1): Ditto.
12428 2021-08-16 Kito Cheng <kito.cheng@sifive.com>
12430 * config/riscv/multilib-generator: Support code model option for
12432 * doc/install.texi: Add document of new option for
12433 --with-multilib-generator.
12435 2021-08-15 Clément Chigot <clement.chigot@atos.net>
12437 * config/rs6000/rs6000.c (xcoff_tls_exec_model_detected): New.
12438 (rs6000_legitimize_tls_address_aix): Use it.
12439 (rs6000_xcoff_file_end): Add ".ref __tls_get_addr" when
12440 xcoff_tls_exec_model_detected is true.
12442 2021-08-15 Jeff Law <jlaw@localhost.localdomain>
12444 * config/h8300/h8300.c (shift_alg_si): Retune H8/300H shifts
12445 to allow a bit more code growth, saving many dozens of cycles.
12446 (h8300_option_override): Adjus shift_alg_si if optimizing for
12448 (get_shift_alg): Use special + inline shifts for residuals
12451 2021-08-14 Stafford Horne <shorne@gmail.com>
12454 * config/or1k/or1k-opts.h: New file.
12455 * config/or1k/or1k.c (or1k_legitimize_address_1, print_reloc):
12456 Support generating gotha relocations if -mcmodel=large is
12458 * config/or1k/or1k.h (TARGET_CMODEL_SMALL, TARGET_CMODEL_LARGE):
12460 * config/or1k/or1k.opt (mcmodel=): New option.
12461 * doc/invoke.texi (OpenRISC Options): Document mcmodel.
12463 2021-08-14 Martin Sebor <msebor@redhat.com>
12465 PR middle-end/101791
12466 * gimple-ssa-warn-access.cc (new_delete_mismatch_p): Use new argument
12467 to valid_new_delete_pair_p.
12468 * tree.c (valid_new_delete_pair_p): Add argument.
12469 * tree.h (valid_new_delete_pair_p): Same.
12471 2021-08-14 Jakub Jelinek <jakub@redhat.com>
12474 * config/i386/i386-expand.c (expand_vec_perm_broadcast_1)
12475 <case E_V64QImode>: For this mode assert
12476 !TARGET_AVX512BW || d->perm[0] rather than !TARGET_AVX2 || d->perm[0].
12478 2021-08-13 Michael Meissner <meissner@linux.ibm.com>
12481 * config/rs6000/altivec.md (xxeval): Use register_predicate
12482 instead of altivec_register_predicate.
12484 2021-08-13 Martin Sebor <msebor@redhat.com>
12486 PR middle-end/101734
12487 * tree-ssa-uninit.c (maybe_warn_read_write_only): New function.
12488 (maybe_warn_operand): Call it.
12490 2021-08-13 Martin Liska <mliska@suse.cz>
12493 * attribs.c (decl_attributes): Make naked functions "noipa"
12496 2021-08-13 Martin Liska <mliska@suse.cz>
12499 * symtab.c (symtab_node::noninterposable_alias): Do not create
12500 local aliases for target_clone functions as the clonning pass
12503 2021-08-13 Martin Liska <mliska@suse.cz>
12505 * opts.c (LIVE_PATCHING_OPTION): Define.
12506 (control_options_for_live_patching): Use it in error messages.
12508 2021-08-13 Jan Hubicka <hubicka@ucw.cz>
12510 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOREAD.
12511 (implicit_const_eaf_flags, implicit_pure_eaf_flags,
12512 ignore_stores_eaf_flags): New constants.
12513 (remove_useless_eaf_flags): New function.
12514 (eaf_flags_useful_p): Use it.
12515 (deref_flags): Add EAF_NOT_RETURNED if flag is unused;
12517 (modref_lattice::init): Add EAF_NOREAD.
12518 (modref_lattice::add_escape_point): Do not reacord escape point if
12520 (modref_lattice::merge): EAF_NOESCAPE implies EAF_NODIRECTESCAPE;
12521 use remove_useless_eaf_flags.
12522 (modref_lattice::merge_deref): Use ignore_stores_eaf_flags.
12523 (modref_lattice::merge_direct_load): Add EAF_NOREAD
12524 (analyze_ssa_name_flags): Fix handling EAF_NOT_RETURNED
12525 (analyze_parms): Use remove_useless_eaf_flags.
12526 (ipa_merge_modref_summary_after_inlining): Use ignore_stores_eaf_flags.
12527 (modref_merge_call_site_flags): Add caller and ecf_flags parameter;
12528 use remove_useless_eaf_flags.
12529 (modref_propagate_flags_in_scc): Update.
12530 * ipa-modref.h: Turn eaf_flags_t back to char.
12531 * tree-core.h (EAF_NOT_RETURNED): Fix.
12532 (EAF_NOREAD): New constant
12533 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1): Check for
12535 * tree-ssa-structalias.c (handle_rhs_call): Handle new flags.
12536 (handle_pure_call): Likewise.
12538 2021-08-12 Jakub Jelinek <jakub@redhat.com>
12540 * tree.def (OMP_MASKED): New tree code.
12541 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_FILTER.
12542 * tree.h (OMP_MASKED_BODY, OMP_MASKED_CLAUSES, OMP_MASKED_COMBINED,
12543 OMP_CLAUSE_FILTER_EXPR): Define.
12544 * tree.c (omp_clause_num_ops): Add OMP_CLAUSE_FILTER entry.
12545 (omp_clause_code_name): Likewise.
12546 (walk_tree_1): Handle OMP_CLAUSE_FILTER.
12547 * tree-nested.c (convert_nonlocal_omp_clauses,
12548 convert_local_omp_clauses): Handle OMP_CLAUSE_FILTER.
12549 (convert_nonlocal_reference_stmt, convert_local_reference_stmt,
12550 convert_gimple_call): Handle GIMPLE_OMP_MASTER.
12551 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_FILTER.
12552 (dump_generic_node): Handle OMP_MASTER.
12553 * gimple.def (GIMPLE_OMP_MASKED): New gimple code.
12554 * gimple.c (gimple_build_omp_masked): New function.
12555 (gimple_copy): Handle GIMPLE_OMP_MASKED.
12556 * gimple.h (gimple_build_omp_masked): Declare.
12557 (gimple_has_substatements): Handle GIMPLE_OMP_MASKED.
12558 (gimple_omp_masked_clauses, gimple_omp_masked_clauses_ptr,
12559 gimple_omp_masked_set_clauses): New inline functions.
12560 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_MASKED.
12561 * gimple-pretty-print.c (dump_gimple_omp_masked): New function.
12562 (pp_gimple_stmt_1): Handle GIMPLE_OMP_MASKED.
12563 * gimple-walk.c (walk_gimple_stmt): Likewise.
12564 * gimple-low.c (lower_stmt): Likewise.
12565 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
12566 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_FILTER. For clauses
12567 that take one expression rather than decl or constant, force
12568 gimplification of that into a SSA_NAME or temporary unless min
12570 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_FILTER.
12571 (gimplify_expr): Handle OMP_MASKED.
12572 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_MASKED.
12573 (estimate_num_insns): Likewise.
12574 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_FILTER.
12575 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_MASKED. Adjust
12576 diagnostics for existence of masked construct.
12577 (scan_omp_1_stmt, lower_omp_master, lower_omp_1, diagnose_sb_1,
12578 diagnose_sb_2): Handle GIMPLE_OMP_MASKED.
12579 * omp-expand.c (expand_omp_synch, expand_omp, omp_make_gimple_edges):
12582 2021-08-12 Uroš Bizjak <ubizjak@gmail.com>
12585 * config/i386/i386.md (avx512f_scalef<mode>2): New insn pattern.
12586 (ldexp<mode>3): Use avx512f_scalef<mode>2.
12587 (UNSPEC_SCALEF): Move from ...
12588 * config/i386/sse.md (UNSPEC_SCALEF): ... here.
12590 2021-08-12 Jan Hubicka <hubicka@ucw.cz>
12592 * ipa-split.c (consider_split): Fix condition testing void functions.
12594 2021-08-12 Aldy Hernandez <aldyh@redhat.com>
12596 * doc/invoke.texi: Remove docs for threader-mode param.
12597 * flag-types.h (enum threader_mode): Remove.
12598 * params.opt: Remove threader-mode param.
12599 * tree-ssa-threadbackward.c (class back_threader): Remove
12600 path_is_unreachable_p.
12601 Make find_paths private.
12602 Add maybe_thread and thread_through_all_blocks.
12603 Remove reference marker for m_registry.
12604 Remove reference marker for m_profit.
12605 (back_threader::back_threader): Adjust for registry and profit not
12607 (dump_path): Move down.
12608 (debug): Move down.
12609 (class thread_jumps): Remove.
12610 (class back_threader_registry): Remove m_all_paths.
12612 (thread_jumps::thread_through_all_blocks): Move to back_threader
12614 (fsm_find_thread_path): Remove
12615 (back_threader::maybe_thread): New.
12616 (back_threader::thread_through_all_blocks): Move from
12618 (back_threader_registry::back_threader_registry): Remove
12620 (back_threader_registry::~back_threader_registry): Remove.
12621 (thread_jumps::find_taken_edge): Remove.
12622 (thread_jumps::check_subpath_and_update_thread_path): Remove.
12623 (thread_jumps::maybe_register_path): Remove.
12624 (thread_jumps::handle_phi): Remove.
12625 (handle_assignment_p): Remove.
12626 (thread_jumps::handle_assignment): Remove.
12627 (thread_jumps::fsm_find_control_statement_thread_paths): Remove.
12628 (thread_jumps::find_jump_threads_backwards): Remove.
12629 (thread_jumps::find_jump_threads_backwards_with_ranger): Remove.
12630 (try_thread_blocks): Rename find_jump_threads_backwards to
12632 (pass_early_thread_jumps::execute): Same.
12634 2021-08-12 Tobias Burnus <tobias@codesourcery.com>
12636 * tree-core.h (omp_clause_proc_bind_kind): Add
12637 OMP_CLAUSE_PROC_BIND_PRIMARY.
12638 * tree-pretty-print.c (dump_omp_clause): Add TODO comment to
12639 change 'master' to 'primary' in proc_bind for OpenMP 5.1.
12641 2021-08-12 Claudiu Zissulescu <claziss@synopsys.com>
12643 * common/config/arc/arc-common.c (arc_option_init_struct): Remove
12644 fno-common reference.
12645 * config/arc/arc.c (arc_override_options): Remove overriding of
12648 2021-08-12 Jakub Jelinek <jakub@redhat.com>
12651 * config/i386/i386-expand.c (ix86_expand_vec_one_operand_perm_avx512):
12652 If d->testing_p, return true after performing checks instead of
12653 actually expanding the insn.
12654 (expand_vec_perm_broadcast_1): Handle V32HImode - assert
12655 !TARGET_AVX512BW and return false.
12657 2021-08-12 Eric Botcazou <ebotcazou@gcc.gnu.org>
12659 * configure.ac (PE linker --disable-dynamicbase support): New check.
12660 * configure: Regenerate.
12661 * config.in: Likewise.
12662 * config/i386/mingw32.h (LINK_SPEC_DISABLE_DYNAMICBASE): New define.
12663 (LINK_SPEC): Use it.
12664 * config/i386/mingw-w64.h (LINK_SPEC_DISABLE_DYNAMICBASE): Likewise.
12665 (LINK_SPEC): Likewise.
12667 2021-08-12 liuhongt <hongtao.liu@intel.com>
12670 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_2): New
12671 post_reload define_insn_and_split.
12672 (*avx512bw_zero_extendv32qiv32hi2_2): Ditto.
12673 (*sse4_1_zero_extendv8qiv8hi2_4): Ditto.
12674 (*avx512f_zero_extendv16hiv16si2_2): Ditto.
12675 (*avx2_zero_extendv8hiv8si2_2): Ditto.
12676 (*sse4_1_zero_extendv4hiv4si2_4): Ditto.
12677 (*avx512f_zero_extendv8siv8di2_2): Ditto.
12678 (*avx2_zero_extendv4siv4di2_2): Ditto.
12679 (*sse4_1_zero_extendv2siv2di2_4): Ditto.
12680 (VI248_256, VI248_512, VI148_512, VI148_256, VI148_128): New
12683 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
12685 * config/rs6000/rs6000-builtin-new.def: Add always, power5, and
12688 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
12690 * config/rs6000/rs6000-builtin-new.def: Add vsx stanza.
12692 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
12694 * config/rs6000/rs6000-builtin-new.def: Finish altivec stanza.
12695 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
12696 initialization of pcvoid_type_node here...
12697 (altivec_init_builtins): ...from here.
12698 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add
12699 RS6000_BTI_const_ptr_void.
12700 (pcvoid_type_node): New macro.
12702 2021-08-11 Richard Biener <rguenther@suse.de>
12705 * tree-ssa-forwprop.c (pass_forwprop::execute): Do not decompose
12706 hard-register accesses.
12708 2021-08-11 Richard Biener <rguenther@suse.de>
12710 * tree-ssa-operands.c (operands_scanner::get_expr_operands):
12711 Do not look at COMPONENT_REF FIELD_DECLs TREE_THIS_VOLATILE
12712 to determine has_volatile_ops.
12714 2021-08-11 Eric Botcazou <ebotcazou@gcc.gnu.org>
12716 * cfgexpand.c (expand_used_vars): Reuse attribs local variable.
12718 2021-08-11 Jan Hubicka <hubicka@ucw.cz>
12719 Alexandre Oliva <oliva@adacore.com>
12721 * ipa-modref.c (modref_lattice::dump): Fix escape_point's min_flags
12723 (modref_lattice::merge_deref): Fix handling of indirect scape points.
12724 (update_escape_summary_1): Likewise.
12725 (update_escape_summary): Likewise.
12726 (ipa_merge_modref_summary_after_inlining): Likewise.
12728 2021-08-11 Richard Biener <rguenther@suse.de>
12730 PR middle-end/101858
12731 * fold-const.c (fold_binary_loc): Guard simplification
12732 of X < (cast) (1 << Y) to integer types.
12734 2021-08-11 Richard Biener <rguenther@suse.de>
12736 PR tree-optimization/101861
12737 * tree-vect-stmts.c (vectorizable_load): Fix error in
12738 previous change with regard to gather vectorization.
12740 2021-08-11 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
12743 * config/arm/arm_neon.h (vdup_n_s8): Replace call to builtin
12745 (vdup_n_s16): Likewise.
12746 (vdup_n_s32): Likewise.
12747 (vdup_n_s64): Likewise.
12748 (vdup_n_u8): Likewise.
12749 (vdup_n_u16): Likewise.
12750 (vdup_n_u32): Likewise.
12751 (vdup_n_u64): Likewise.
12752 (vdup_n_p8): Likewise.
12753 (vdup_n_p16): Likewise.
12754 (vdup_n_p64): Likewise.
12755 (vdup_n_f16): Likewise.
12756 (vdup_n_f32): Likewise.
12757 (vdupq_n_s8): Likewise.
12758 (vdupq_n_s16): Likewise.
12759 (vdupq_n_s32): Likewise.
12760 (vdupq_n_s64): Likewise.
12761 (vdupq_n_u8): Likewise.
12762 (vdupq_n_u16): Likewise.
12763 (vdupq_n_u32): Likewise.
12764 (vdupq_n_u64): Likewise.
12765 (vdupq_n_p8): Likewise.
12766 (vdupq_n_p16): Likewise.
12767 (vdupq_n_p64): Likewise.
12768 (vdupq_n_f16): Likewise.
12769 (vdupq_n_f32): Likewise.
12770 (vmov_n_s8): Replace call to builtin with call to corresponding
12772 (vmov_n_s16): Likewise.
12773 (vmov_n_s32): Likewise.
12774 (vmov_n_s64): Likewise.
12775 (vmov_n_u8): Likewise.
12776 (vmov_n_u16): Likewise.
12777 (vmov_n_u32): Likewise.
12778 (vmov_n_u64): Likewise.
12779 (vmov_n_p8): Likewise.
12780 (vmov_n_p16): Likewise.
12781 (vmov_n_f16): Likewise.
12782 (vmov_n_f32): Likewise.
12783 (vmovq_n_s8): Likewise.
12784 (vmovq_n_s16): Likewise.
12785 (vmovq_n_s32): Likewise.
12786 (vmovq_n_s64): Likewise.
12787 (vmovq_n_u8): Likewise.
12788 (vmovq_n_u16): Likewise.
12789 (vmovq_n_u32): Likewise.
12790 (vmovq_n_u64): Likewise.
12791 (vmovq_n_p8): Likewise.
12792 (vmovq_n_p16): Likewise.
12793 (vmovq_n_f16): Likewise.
12794 (vmovq_n_f32): Likewise.
12795 * config/arm/arm_neon_builtins.def: Remove entries for vdup_n.
12797 2021-08-11 liuhongt <hongtao.liu@intel.com>
12800 * config/i386/i386.md (ldexp<mode>3): Extend to vscalefs[sd]
12801 when TARGET_AVX512F and TARGET_SSE_MATH.
12803 2021-08-10 Jakub Jelinek <jakub@redhat.com>
12806 * config/i386/i386-expand.c (expand_vec_perm_even_odd): Return false
12807 for V32HImode if !TARGET_AVX512BW.
12808 (ix86_vectorize_vec_perm_const) <case E_V32HImode, case E_V64QImode>:
12809 If !TARGET_AVX512BW and TARGET_AVX512F and d.testing_p, don't fail
12810 early, but actually check the permutation.
12812 2021-08-10 Richard Biener <rguenther@suse.de>
12814 PR tree-optimization/101809
12815 * tree-vect-stmts.c (get_load_store_type): Allow emulated
12816 gathers with offset vector nunits being a constant multiple
12817 of the data vector nunits.
12818 (vect_get_gather_scatter_ops): Use the appropriate nunits
12819 for the offset vector defs.
12820 (vectorizable_store): Adjust call to
12821 vect_get_gather_scatter_ops.
12822 (vectorizable_load): Likewise. Handle the case of less
12823 offset vectors than data vectors.
12825 2021-08-10 Jakub Jelinek <jakub@redhat.com>
12828 * config/i386/sse.md (*avx512f_shuf_<shuffletype>64x2_1<mask_name>_1,
12829 *avx512f_shuf_<shuffletype>32x4_1<mask_name>_1): New define_insn
12832 2021-08-10 Richard Biener <rguenther@suse.de>
12834 PR tree-optimization/101801
12835 PR tree-optimization/101819
12836 * tree-vectorizer.h (vect_emulated_vector_p): Declare.
12837 * tree-vect-loop.c (vect_emulated_vector_p): New function.
12838 (vectorizable_reduction): Re-instantiate a check for emulated
12840 * tree-vect-stmts.c (vectorizable_shift): Likewise.
12841 (vectorizable_operation): Likewise. Cost emulated vector
12842 operations according to the scalar sequence synthesized by
12845 2021-08-10 Richard Biener <rguenther@suse.de>
12847 PR middle-end/101824
12848 * tree-nested.c (get_frame_field): Mark the COMPONENT_REF as
12849 volatile in case the variable was.
12851 2021-08-10 H.J. Lu <hjl.tools@gmail.com>
12854 * config/i386/constraints.md (BC): Document for integer SSE
12855 constant all bits set operand.
12856 (BF): New constraint for const floating-point all bits set
12858 * config/i386/i386.c (standard_sse_constant_p): Likewise.
12859 (standard_sse_constant_opcode): Likewise.
12860 * config/i386/sse.md (sseconstm1): New mode attribute.
12861 (mov<mode>_internal): Replace BC with <sseconstm1>.
12863 2021-08-10 liuhongt <hongtao.liu@intel.com>
12865 * config/i386/sse.md (cond_<insn><mode>): New expander.
12866 (VI248_AVX512VLBW): New mode iterator.
12867 * config/i386/predicates.md
12868 (nonimmediate_or_const_vec_dup_operand): New predicate.
12870 2021-08-09 Andrew MacLeod <amacleod@redhat.com>
12872 PR tree-optimization/101741
12873 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Check
12874 type of parameter for toupper/tolower.
12876 2021-08-09 Martin Jambor <mjambor@suse.cz>
12878 PR testsuite/101654
12879 * ipa-prop.c (propagate_controlled_uses): Removed a spurious space.
12881 2021-08-09 Pat Haugen <pthaugen@linux.ibm.com>
12883 * config/rs6000/rs6000.c (is_load_insn1): Verify destination is a
12885 (is_store_insn1): Verify source is a register.
12887 2021-08-09 Uroš Bizjak <ubizjak@gmail.com>
12890 * config/i386/mmx.md (<any_logic:code>v2sf3):
12891 Rename from *mmx_<any_logic:code>v2sf3
12893 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
12895 * config/nvptx/nvptx.c: Cross-reference parts adapted in
12896 'gcc/omp-oacc-neuter-broadcast.cc'.
12897 * omp-low.c: Likewise.
12898 * omp-oacc-neuter-broadcast.cc: Cross-reference parts adapted from
12901 2021-08-09 Julian Brown <julian@codesourcery.com>
12902 Kwok Cheung Yeung <kcy@codesourcery.com>
12903 Thomas Schwinge <thomas@codesourcery.com>
12905 * config/gcn/gcn.c (gcn_init_builtins): Override decls for
12906 BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START,
12907 BUILT_IN_GOACC_SINGLE_COPY_END and BUILT_IN_GOACC_BARRIER.
12908 (gcn_goacc_validate_dims): Turn on worker partitioning unconditionally.
12909 (gcn_fork_join): Update comment.
12910 * config/gcn/gcn.opt (flag_worker_partitioning): Remove.
12911 (macc_experimental_workers): Remove unused option.
12913 2021-08-09 Julian Brown <julian@codesourcery.com>
12914 Nathan Sidwell <nathan@codesourcery.com> (via 'gcc/config/nvptx/nvptx.c' master)
12915 Kwok Cheung Yeung <kcy@codesourcery.com>
12916 Thomas Schwinge <thomas@codesourcery.com>
12918 * Makefile.in (OBJS): Add omp-oacc-neuter-broadcast.o.
12919 * doc/tm.texi.in (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD):
12920 Add documentation hook.
12921 * doc/tm.texi: Regenerate.
12922 * omp-oacc-neuter-broadcast.cc: New file.
12923 * omp-builtins.def (BUILT_IN_GOACC_BARRIER)
12924 (BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START)
12925 (BUILT_IN_GOACC_SINGLE_COPY_END): New builtins.
12926 * passes.def (pass_omp_oacc_neuter_broadcast): Add pass.
12927 * target.def (goacc.create_worker_broadcast_record): Add target
12929 * tree-pass.h (make_pass_omp_oacc_neuter_broadcast): Add
12931 * config/gcn/gcn-protos.h (gcn_goacc_adjust_propagation_record):
12932 Rename prototype to...
12933 (gcn_goacc_create_worker_broadcast_record): ... this.
12934 * config/gcn/gcn-tree.c (gcn_goacc_adjust_propagation_record): Rename
12936 (gcn_goacc_create_worker_broadcast_record): ... this.
12937 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_PROPAGATION_RECORD):
12939 (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD): ... this.
12941 2021-08-09 Tejas Belagod <tejas.belagod@arm.com>
12944 * config/aarch64/aarch64-simd.md (vlshr<mode>3, vashr<mode>3): Use
12945 the right iterator.
12947 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
12949 * Makefile.in (GTFILES): Remove '$(srcdir)/omp-offload.c'.
12951 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
12953 * builtins.def (DEF_GOACC_BUILTIN, DEF_GOMP_BUILTIN): Don't
12954 consider '-foffload-abi'.
12955 * common.opt (-foffload-abi): Remove 'Var', 'Init'.
12956 * opts.c (common_handle_option) <-foffload-abi> [ACCEL_COMPILER]:
12959 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
12961 * optc-gen.awk: Sanity check that 'Init' doesn't appear without
12964 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
12966 * omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE): Remove.
12968 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
12970 * doc/gty.texi (Files): Update.
12972 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
12974 * doc/gty.texi (Files): Fix GTY header file example.
12976 2021-08-09 Roger Sayle <roger@nextmovesoftware.com>
12978 * tree-ssa-ccp.c (value_mask_to_min_max): Helper function to
12979 determine the upper and lower bounds from a mask-value pair.
12980 (bit_value_unop) [ABS_EXPR, ABSU_EXPR]: Add support for
12981 absolute value and unsigned absolute value expressions.
12982 (bit_value_binop): Initialize *VAL's precision.
12983 [LT_EXPR, LE_EXPR]: Use value_mask_to_min_max to determine
12984 upper and lower bounds of operands. Add LE_EXPR/GE_EXPR
12985 support when the operands are unknown but potentially equal.
12986 [MIN_EXPR, MAX_EXPR]: Support minimum/maximum expressions.
12988 2021-08-09 Bin Cheng <bin.cheng@linux.alibaba.com>
12990 * config/aarch64/aarch64.md
12991 (*extend<SHORT:mode><GPI:mode>2_aarch64): Use %<GPI:w>0.
12993 2021-08-08 Sergei Trofimovich <siarheit@google.com>
12995 * lra-constraints.c: Fix s/otput/output/ typo.
12997 2021-08-06 Martin Sebor <msebor@redhat.com>
12999 * builtins.c (expand_builtin_memchr): Move to gimple-ssa-warn-access.cc.
13000 (expand_builtin_strcat): Same.
13001 (expand_builtin_stpncpy): Same.
13002 (expand_builtin_strncat): Same.
13003 (check_read_access): Same.
13004 (check_memop_access): Same.
13005 (expand_builtin_strlen): Move checks to gimple-ssa-warn-access.cc.
13006 (expand_builtin_strnlen): Same.
13007 (expand_builtin_memcpy): Same.
13008 (expand_builtin_memmove): Same.
13009 (expand_builtin_mempcpy): Same.
13010 (expand_builtin_strcpy): Same.
13011 (expand_builtin_strcpy_args): Same.
13012 (expand_builtin_stpcpy_1): Same.
13013 (expand_builtin_strncpy): Same.
13014 (expand_builtin_memset): Same.
13015 (expand_builtin_bzero): Same.
13016 (expand_builtin_strcmp): Same.
13017 (expand_builtin_strncmp): Same.
13018 (expand_builtin): Remove handlers.
13019 (fold_builtin_strlen): Add a comment.
13020 * builtins.h (check_access): Move to gimple-ssa-warn-access.cc.
13021 * calls.c (maybe_warn_nonstring_arg): Same.
13022 * diagnostic-spec.c (nowarn_spec_t::nowarn_spec_t): Add warning option.
13023 * gimple-fold.c (gimple_fold_builtin_strcpy): Pass argument to callee.
13024 (gimple_fold_builtin_stpcpy): Same.
13025 * gimple-ssa-warn-access.cc (has_location): New function.
13026 (get_location): Same.
13027 (get_callee_fndecl): Same.
13028 (call_nargs): Same.
13030 (warn_string_no_nul): Define.
13031 (unterminated_array): Same.
13032 (check_nul_terminated_array): Same.
13033 (maybe_warn_nonstring_arg): Same.
13034 (maybe_warn_for_bound): Same.
13035 (warn_for_access): Same.
13036 (check_access): Same.
13037 (check_memop_access): Same.
13038 (check_read_access): Same.
13039 (warn_dealloc_offset): Use helper functions.
13040 (maybe_emit_free_warning): Same.
13041 (class pass_waccess): Add members.
13042 (check_strcat): New function.
13043 (check_strncat): New function.
13044 (check_stxcpy): New function.
13045 (check_stxncpy): New function.
13046 (check_strncmp): New function.
13047 (pass_waccess::check_builtin): New function.
13048 (pass_waccess::check): Call it.
13049 * gimple-ssa-warn-access.h (warn_string_no_nul): Move here from
13051 (maybe_warn_for_bound): Same.
13052 (check_access): Same.
13053 (check_memop_access): Same.
13054 (check_read_access): Same.
13055 * pointer-query.h (struct access_data): Define a ctor overload.
13057 2021-08-06 Richard Biener <rguenther@suse.de>
13059 PR tree-optimization/101801
13060 * tree-vectorizer.h (vect_worthwhile_without_simd_p): Rename...
13061 (vect_can_vectorize_without_simd_p): ... to this.
13062 * tree-vect-loop.c (vect_worthwhile_without_simd_p): Rename...
13063 (vect_can_vectorize_without_simd_p): ... to this and fold
13064 in vect_min_worthwhile_factor.
13065 (vect_min_worthwhile_factor): Remove.
13066 (vectorizable_reduction): Adjust and remove the cost part.
13067 * tree-vect-stmts.c (vectorizable_shift): Likewise.
13068 (vectorizable_operation): Likewise.
13070 2021-08-06 Uroš Bizjak <ubizjak@gmail.com>
13073 * config/i386/i386.md (cmove reg-to-reg move elimination peephole2s):
13074 Add general_gr_operand predicate to operand 3.
13076 2021-08-06 Roger Sayle <roger@nextmovesoftware.com>
13078 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern): Use
13079 CFN_BUILT_IN_CLRSB* instead of BUILT_IN_CLRSB* for consistency.
13081 2021-08-06 Tamar Christina <tamar.christina@arm.com>
13083 * config/aarch64/aarch64-sve-builtins.cc (register_svpattern,
13084 register_svprfop): Pass vec<> by pointer.
13085 * langhooks-def.h (lhd_simulate_enum_decl): Likewise.
13086 * langhooks.c (lhd_simulate_enum_decl): Likewise.
13087 * langhooks.h (struct lang_hooks_for_types): Likewise.
13089 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
13091 * config/aarch64/arm_neon.h (vst1_bf16_x2): Use
13092 __builtin_memcpy instead of constructing an additional
13093 __builtin_aarch64_simd_oi one vector at a time.
13094 (vst1q_bf16_x2): Likewise.
13095 (vst1_bf16_x3): Use __builtin_memcpy instead of constructing
13096 an additional __builtin_aarch64_simd_ci one vector at a time.
13097 (vst1q_bf16_x3): Likewise.
13098 (vst1_bf16_x4): Use __builtin_memcpy instead of a union.
13099 (vst1q_bf16_x4): Likewise.
13100 (vst2_bf16): Use __builtin_memcpy instead of constructing an
13101 additional __builtin_aarch64_simd_oi one vector at a time.
13102 (vst2q_bf16): Likewise.
13103 (vst3_bf16): Use __builtin_memcpy instead of constructing an
13104 additional __builtin_aarch64_simd_ci mode one vector at a
13106 (vst3q_bf16): Likewise.
13107 (vst4_bf16): Use __builtin_memcpy instead of constructing an
13108 additional __builtin_aarch64_simd_xi one vector at a time.
13109 (vst4q_bf16): Likewise.
13111 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
13113 * config/aarch64/arm_neon.h (__ST2_LANE_FUNC): Delete.
13114 (__ST2Q_LANE_FUNC): Delete.
13115 (vst2_lane_f16): Use __builtin_memcpy to copy vector
13116 structure instead of constructing __builtin_aarch64_simd_oi
13117 one vector at a time.
13118 (vst2_lane_f32): Likewise.
13119 (vst2_lane_f64): Likewise.
13120 (vst2_lane_p8): Likewise.
13121 (vst2_lane_p16): Likewise.
13122 (vst2_lane_p64): Likewise.
13123 (vst2_lane_s8): Likewise.
13124 (vst2_lane_s16): Likewise.
13125 (vst2_lane_s32): Likewise.
13126 (vst2_lane_s64): Likewise.
13127 (vst2_lane_u8): Likewise.
13128 (vst2_lane_u16): Likewise.
13129 (vst2_lane_u32): Likewise.
13130 (vst2_lane_u64): Likewise.
13131 (vst2_lane_bf16): Likewise.
13132 (vst2q_lane_f16): Use __builtin_memcpy to copy vector
13133 structure instead of using a union.
13134 (vst2q_lane_f32): Likewise.
13135 (vst2q_lane_f64): Likewise.
13136 (vst2q_lane_p8): Likewise.
13137 (vst2q_lane_p16): Likewise.
13138 (vst2q_lane_p64): Likewise.
13139 (vst2q_lane_s8): Likewise.
13140 (vst2q_lane_s16): Likewise.
13141 (vst2q_lane_s32): Likewise.
13142 (vst2q_lane_s64): Likewise.
13143 (vst2q_lane_u8): Likewise.
13144 (vst2q_lane_u16): Likewise.
13145 (vst2q_lane_u32): Likewise.
13146 (vst2q_lane_u64): Likewise.
13147 (vst2q_lane_bf16): Likewise.
13149 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
13151 * config/aarch64/arm_neon.h (__ST3_LANE_FUNC): Delete.
13152 (__ST3Q_LANE_FUNC): Delete.
13153 (vst3_lane_f16): Use __builtin_memcpy to copy vector
13154 structure instead of constructing __builtin_aarch64_simd_ci
13155 one vector at a time.
13156 (vst3_lane_f32): Likewise.
13157 (vst3_lane_f64): Likewise.
13158 (vst3_lane_p8): Likewise.
13159 (vst3_lane_p16): Likewise.
13160 (vst3_lane_p64): Likewise.
13161 (vst3_lane_s8): Likewise.
13162 (vst3_lane_s16): Likewise.
13163 (vst3_lane_s32): Likewise.
13164 (vst3_lane_s64): Likewise.
13165 (vst3_lane_u8): Likewise.
13166 (vst3_lane_u16): Likewise.
13167 (vst3_lane_u32): Likewise.
13168 (vst3_lane_u64): Likewise.
13169 (vst3_lane_bf16): Likewise.
13170 (vst3q_lane_f16): Use __builtin_memcpy to copy vector
13171 structure instead of using a union.
13172 (vst3q_lane_f32): Likewise.
13173 (vst3q_lane_f64): Likewise.
13174 (vst3q_lane_p8): Likewise.
13175 (vst3q_lane_p16): Likewise.
13176 (vst3q_lane_p64): Likewise.
13177 (vst3q_lane_s8): Likewise.
13178 (vst3q_lane_s16): Likewise.
13179 (vst3q_lane_s32): Likewise.
13180 (vst3q_lane_s64): Likewise.
13181 (vst3q_lane_u8): Likewise.
13182 (vst3q_lane_u16): Likewise.
13183 (vst3q_lane_u32): Likewise.
13184 (vst3q_lane_u64): Likewise.
13185 (vst3q_lane_bf16): Likewise.
13187 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
13189 * config/aarch64/arm_neon.h (__ST4_LANE_FUNC): Delete.
13190 (__ST4Q_LANE_FUNC): Delete.
13191 (vst4_lane_f16): Use __builtin_memcpy to copy vector
13192 structure instead of constructing __builtin_aarch64_simd_xi
13193 one vector at a time.
13194 (vst4_lane_f32): Likewise.
13195 (vst4_lane_f64): Likewise.
13196 (vst4_lane_p8): Likewise.
13197 (vst4_lane_p16): Likewise.
13198 (vst4_lane_p64): Likewise.
13199 (vst4_lane_s8): Likewise.
13200 (vst4_lane_s16): Likewise.
13201 (vst4_lane_s32): Likewise.
13202 (vst4_lane_s64): Likewise.
13203 (vst4_lane_u8): Likewise.
13204 (vst4_lane_u16): Likewise.
13205 (vst4_lane_u32): Likewise.
13206 (vst4_lane_u64): Likewise.
13207 (vst4_lane_bf16): Likewise.
13208 (vst4q_lane_f16): Use __builtin_memcpy to copy vector
13209 structure instead of using a union.
13210 (vst4q_lane_f32): Likewise.
13211 (vst4q_lane_f64): Likewise.
13212 (vst4q_lane_p8): Likewise.
13213 (vst4q_lane_p16): Likewise.
13214 (vst4q_lane_p64): Likewise.
13215 (vst4q_lane_s8): Likewise.
13216 (vst4q_lane_s16): Likewise.
13217 (vst4q_lane_s32): Likewise.
13218 (vst4q_lane_s64): Likewise.
13219 (vst4q_lane_u8): Likewise.
13220 (vst4q_lane_u16): Likewise.
13221 (vst4q_lane_u32): Likewise.
13222 (vst4q_lane_u64): Likewise.
13223 (vst4q_lane_bf16): Likewise.
13225 2021-08-06 Martin Liska <mliska@suse.cz>
13227 * config/rs6000/rs6000.c (rs6000_option_override_internal): When
13228 a target option is restored, it can have
13229 rs6000_long_double_type_size set to FLOAT_PRECISION_TFmode
13230 and error should not be emitted.
13232 2021-08-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
13234 * gcov-io.h (gcov_write): Declare.
13235 * gcov-io.c (gcov_write): New.
13236 (gcov_write_counter): Remove.
13237 (gcov_write_tag_length): Likewise.
13238 (gcov_write_summary): Replace gcov_write_tag_length() with calls to
13239 gcov_write_unsigned().
13240 * doc/invoke.texi (fprofile-info-section): Mention
13241 __gcov_info_to_gdca().
13243 2021-08-06 Martin Sebor <msebor@redhat.com>
13245 * dominance.c (prune_bbs_to_update_dominators): Adjust by-value vec
13246 arguments to by-reference.
13247 (iterate_fix_dominators): Same.
13248 * dominance.h (iterate_fix_dominators): Same.
13249 * ipa-prop.h: Call auto_vec::to_vec_legacy.
13250 * tree-data-ref.c (dump_data_dependence_relation): Adjust by-value vec
13251 arguments to by-reference.
13252 (debug_data_dependence_relation): Same.
13253 (dump_data_dependence_relations): Same.
13254 * tree-data-ref.h (debug_data_dependence_relation): Same.
13255 (dump_data_dependence_relations): Same.
13256 * tree-predcom.c (dump_chains): Same.
13257 (initialize_root_vars_lm): Same.
13258 (determine_unroll_factor): Same.
13259 (replace_phis_by_defined_names): Same.
13260 (insert_init_seqs): Same.
13261 (pcom_worker::tree_predictive_commoning_loop): Call
13262 auto_vec::to_vec_legacy.
13263 * tree-ssa-pre.c (insert_into_preds_of_block): Adjust by-value vec
13264 arguments to by-reference.
13265 * tree-ssa-threadbackward.c (populate_worklist): Same.
13266 (back_threader::resolve_def): Same.
13267 * tree-vect-data-refs.c (vect_check_nonzero_value): Same.
13268 (vect_enhance_data_refs_alignment): Same.
13269 (vect_check_lower_bound): Same.
13270 (vect_prune_runtime_alias_test_list): Same.
13271 (vect_permute_store_chain): Same.
13272 * tree-vect-slp-patterns.c (vect_normalize_conj_loc): Same.
13273 * tree-vect-stmts.c (vect_create_vectorized_demotion_stmts): Same.
13274 * tree-vectorizer.h (vect_permute_store_chain): Same.
13275 * vec.c (test_init): New function.
13276 (vec_c_tests): Call new function.
13277 * vec.h (vec): Declare ctors, dtor, and assignment.
13278 (auto_vec::vec_to_legacy): New function.
13279 (vec::copy): Adjust initialization.
13281 2021-08-05 H.J. Lu <hjl.tools@gmail.com>
13284 * config/i386/i386.c (ix86_can_inline_p): Ignore MASK_80387 if
13285 callee only uses GPRs.
13286 * config/i386/ia32intrin.h: Revert commit 5463cee2770.
13287 * config/i386/serializeintrin.h: Revert commit 71958f740f1.
13288 * config/i386/x86gprintrin.h: Add
13289 #pragma GCC target("general-regs-only") and #pragma GCC pop_options
13290 to disable non-GPR ISAs.
13292 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
13294 PR middle-end/101787
13295 * doc/md.texi (cond_ashl, cond_ashr, cond_lshr): Document.
13297 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
13299 * tree-vectorizer.h (vect_is_store_elt_extraction, vect_is_reduction)
13300 (vect_reduc_type, vect_embedded_comparison_type, vect_comparison_type)
13301 (vect_is_extending_load, vect_is_integer_truncation): New functions,
13302 moved from aarch64.c but given different names.
13303 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction)
13304 (aarch64_is_reduction, aarch64_reduc_type)
13305 (aarch64_embedded_comparison_type, aarch64_comparison_type)
13306 (aarch64_extending_load_p, aarch64_integer_truncation_p): Delete
13307 in favor of the above. Update callers accordingly.
13309 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
13312 * config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
13313 writing .cpu directive in asm output.
13314 * config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
13315 (arm_last_printed_arch_string): Delete.
13316 (arm_last-printed_fpu_string): Delete.
13317 (arm_configure_build_target): If use of floating-point/SIMD is
13318 disabled, remove all fp/simd related features from the target ISA.
13319 (last_arm_targ_options): New variable.
13320 (arm_print_asm_arch_directives): Add new parameters. Change order
13321 of emitted directives and handle all cases here.
13322 (arm_file_start): Always call arm_print_asm_arch_directives, move
13323 all generation of .arch/.arch_extension here.
13324 (arm_file_end): Call arm_print_asm_arch.
13325 (arm_declare_function_name): Call arm_print_asm_arch_directives
13326 instead of printing .arch/.fpu directives directly.
13328 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
13330 * config/arm/arm.c (arm_configure_build_target): Don't call
13331 arm_option_reconfigure_globals.
13332 (arm_option_restore): Call arm_option_reconfigure_globals after
13333 reconfiguring the target.
13334 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
13336 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
13338 * config/arm/arm.c (arm_configure_build_target): Ensure the target's
13339 arch_name is always set.
13341 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
13343 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
13344 of vec_select high-half from being added into Neon subtract
13347 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
13349 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
13350 of vec_select high-half from being added into Neon add cost.
13352 2021-08-05 Kewen Lin <linkw@linux.ibm.com>
13354 * cfgloop.h (loops_list::loops_list): Add one optional argument
13355 root and adjust accordingly, update loop tree walking and factor
13357 * cfgloop.c (loops_list::walk_loop_tree): ... this. New function.
13359 2021-08-05 Eric Botcazou <ebotcazou@gcc.gnu.org>
13361 PR tree-optimization/101626
13362 * tree-sra.c (propagate_subaccesses_from_rhs): Do not set the
13363 reverse scalar storage order on a pointer or vector component.
13365 2021-08-05 liuhongt <hongtao.liu@intel.com>
13367 * config/i386/sse.md (cond_<code><mode>): New expander.
13369 2021-08-05 liuhongt <hongtao.liu@intel.com>
13371 * config/i386/sse.md (cond_<code><mode>): New expander.
13373 2021-08-05 liuhongt <hongtao.liu@intel.com>
13375 * config/i386/sse.md (cond_<code><mode>): New expander.
13377 2021-08-04 David Malcolm <dmalcolm@redhat.com>
13380 * Makefile.in (ANALYZER_OBJS): Add analyzer/region-model-asm.o.
13382 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
13385 * config/i386/i386.h (STORE_MAX_PIECES): Allow 16/32/64 bytes
13386 only if TARGET_INTER_UNIT_MOVES_TO_VEC is true.
13388 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
13391 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
13392 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
13393 data with SSE register from one memory location to another.
13395 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
13397 * config/s390/s390.c (expand_perm_with_vpdi): New function.
13398 (vectorize_vec_perm_const_1): Call expand_perm_with_vpdi.
13399 * config/s390/vector.md (*vpdi1<mode>, @vpdi1<mode>): Enable a
13400 parameterized expander.
13401 (*vpdi4<mode>, @vpdi4<mode>): Likewise.
13403 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
13405 * config/s390/s390.c (MAX_VECT_LEN): Define macro.
13406 (struct expand_vec_perm_d): Define struct.
13407 (expand_perm_with_merge): New function.
13408 (vectorize_vec_perm_const_1): New function.
13409 (s390_vectorize_vec_perm_const): New function.
13410 (TARGET_VECTORIZE_VEC_PERM_CONST): Define target macro.
13412 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
13414 * config/s390/vector.md (V_HW_64): Remove mode iterator.
13415 (*vec_load_pair<mode>): Use V_HW_2 instead of V_HW_64.
13416 * config/s390/vx-builtins.md
13417 (vec_scatter_element<V_HW_2:mode>_SI): Use V_HW_2 instead of
13420 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
13422 * config/s390/s390.md (UNSPEC_VEC_PERMI): Remove constant
13424 * config/s390/vector.md (*vpdi1<mode>, *vpdi4<mode>): New pattern
13426 * config/s390/vx-builtins.md (*vec_permi<mode>): Emit generic rtx
13427 instead of an unspec.
13429 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
13431 * config/s390/s390-modes.def: Add more vector modes to support
13432 concatenation of two vectors.
13433 * config/s390/s390-protos.h (s390_expand_merge_perm_const): Add
13435 (s390_expand_merge): Likewise.
13436 * config/s390/s390.c (s390_expand_merge_perm_const): New function.
13437 (s390_expand_merge): New function.
13438 * config/s390/s390.md (UNSPEC_VEC_MERGEH, UNSPEC_VEC_MERGEL):
13439 Remove constant definitions.
13440 * config/s390/vector.md (V_HW_2): Add mode iterators.
13441 (VI_HW_4, V_HW_4): Rename VI_HW_4 to V_HW_4.
13442 (vec_2x_nelts, vec_2x_wide): New mode attributes.
13443 (*vmrhb, *vmrlb, *vmrhh, *vmrlh, *vmrhf, *vmrlf, *vmrhg, *vmrlg):
13444 New pattern definitions.
13445 (vec_widen_umult_lo_<mode>, vec_widen_umult_hi_<mode>)
13446 (vec_widen_smult_lo_<mode>, vec_widen_smult_hi_<mode>)
13447 (vec_unpacks_lo_v4sf, vec_unpacks_hi_v4sf, vec_unpacks_lo_v2df)
13448 (vec_unpacks_hi_v2df): Adjust expanders to emit non-unspec RTX for
13450 * config/s390/vx-builtins.md (V_HW_4): Remove mode iterator. Now
13452 (vec_mergeh<mode>, vec_mergel<mode>): Use s390_expand_merge to
13453 emit vec merge pattern.
13455 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
13457 * config/aarch64/aarch64.c (aarch64_strip_extend_vec_half):
13459 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent cost of
13460 vec_select high-half from being added into Neon multiply
13462 * rtlanal.c (vec_series_highpart_p): Define.
13463 * rtlanal.h (vec_series_highpart_p): Declare.
13465 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
13467 * config/aarch64/aarch64.c (aarch64_strip_duplicate_vec_elt):
13469 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent
13470 vec_select cost from being added into Neon multiply cost.
13472 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
13474 * tree-vect-loop.c (vect_better_loop_vinfo_p): Detect cases in
13475 which old_loop_vinfo is an epilogue loop that handles a constant
13476 number of iterations.
13478 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
13480 * tree-vect-loop.c (vect_analyze_loop): Print a dump message
13481 when a reanalyzed loop fails to be cheaper than the current
13484 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
13486 * config/aarch64/aarch64.c: Fix a typo.
13488 2021-08-04 Vincent Lefèvre <vincent-gcc@vinc17.net>
13490 PR gcov-profile/101773
13491 * gcov-io.c (gcov_close): Check return code of a fclose.
13493 2021-08-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
13496 * dwarf2out.c (dwarf2out_assembly_start): Emit a dummy
13497 .file statement when needed.
13499 2021-08-04 Richard Biener <rguenther@suse.de>
13501 * tree-vect-data-refs.c (vect_check_gather_scatter):
13502 Include widening conversions only when the result is
13503 still handed by native gather or the current offset
13504 size not already matches the data size.
13505 Also succeed analysis in case there's no native support,
13506 noted by a IFN_LAST ifn and a NULL decl.
13507 (vect_analyze_data_refs): Always consider gathers.
13508 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
13509 Test for no IFN gather rather than decl gather.
13510 * tree-vect-stmts.c (vect_model_load_cost): Pass in the
13511 gather-scatter info and cost emulated gathers accordingly.
13512 (vect_truncate_gather_scatter_offset): Properly test for
13514 (vect_use_strided_gather_scatters_p): Likewise.
13515 (get_load_store_type): Handle emulated gathers and its
13517 (vectorizable_load): Likewise. Emulate them by extracting
13518 scalar offsets, doing scalar loads and a vector construct.
13520 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
13523 * expr.c (op_by_pieces_d::op_by_pieces_d): Add a max_pieces
13524 argument to set m_max_size.
13525 (move_by_pieces_d): Pass MOVE_MAX_PIECES to op_by_pieces_d.
13526 (store_by_pieces_d): Pass STORE_MAX_PIECES to op_by_pieces_d.
13527 (compare_by_pieces_d): Pass COMPARE_MAX_PIECES to op_by_pieces_d.
13529 2021-08-04 Roger Sayle <roger@nextmovesoftware.com>
13530 Marc Glisse <marc.glisse@inria.fr>
13532 * match.pd (bit_ior, bit_xor): Canonicalize (X*C1)|(X*C2) and
13533 (X*C1)^(X*C2) as X*(C1+C2), and related variants, using
13534 tree_nonzero_bits to ensure that operands are bit-wise disjoint.
13536 2021-08-04 Richard Biener <rguenther@suse.de>
13538 * tree-ssa-forwprop.c (pass_forwprop::execute): Split
13539 out code to decompose vector loads ...
13540 (optimize_vector_load): ... here. Generalize it to
13541 handle intermediate widening and TARGET_MEM_REF loads
13542 and apply it to loads with a supported vector mode as well.
13544 2021-08-04 Richard Biener <rguenther@suse.de>
13546 PR tree-optimization/101756
13547 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Make sure
13548 the result of the reduction epilogue is compatible to the original
13551 2021-08-04 liuhongt <hongtao.liu@intel.com>
13554 * config/i386/i386.md (peephole2): Refine predicate from
13555 register_operand to general_reg_operand.
13557 2021-08-04 Aldy Hernandez <aldyh@redhat.com>
13559 * gimple-range-path.h (path_range_query::dump): Mark override.
13561 2021-08-04 Richard Biener <rguenther@suse.de>
13563 PR tree-optimization/101769
13564 * tree-tailcall.c (eliminate_tail_call): Add the created loop
13565 for the first recursion and return it via the new output parameter.
13566 (optimize_tail_call): Pass through new output param.
13567 (tree_optimize_tail_calls_1): After creating all latches,
13568 add the created loop to the loop tree. Do not mark loops for fixup.
13570 2021-08-04 Martin Liska <mliska@suse.cz>
13572 * doc/invoke.texi: Document threader-mode param.
13574 2021-08-04 liuhongt <hongtao.liu@intel.com>
13576 * config/i386/sse.md (cond_fma<mode>): New expander.
13577 (cond_fms<mode>): Ditto.
13578 (cond_fnma<mode>): Ditto.
13579 (cond_fnms<mode>): Ditto.
13581 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
13583 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode>): Use && instead of &.
13585 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
13587 * config/rs6000/constraints.md: Remove "e" from the list of available
13588 constraint characters.
13590 2021-08-03 Eugene Rozenfeld <erozen@microsoft.com>
13592 PR gcov-profile/71672
13593 * auto-profile.c (afdo_indirect_call): Fix setup of the historgram value for indirect calls.
13595 2021-08-03 Paul A. Clarke <pc@us.ibm.com>
13597 * config/rs6000/smmintrin.h (_mm_minpos_epu16): New.
13599 2021-08-03 H.J. Lu <hjl.tools@gmail.com>
13601 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): In 64-bit mode,
13602 try XMM31 to avoid vzeroupper.
13604 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
13606 * doc/invoke.texi: Document -mtune=neoverse-512tvb and
13607 -mcpu=neoverse-512tvb.
13608 * config/aarch64/aarch64-cores.def (neoverse-512tvb): New entry.
13609 * config/aarch64/aarch64-tune.md: Regenerate.
13610 * config/aarch64/aarch64.c (neoverse512tvb_sve_vector_cost)
13611 (neoverse512tvb_sve_issue_info, neoverse512tvb_vec_issue_info)
13612 (neoverse512tvb_vector_cost, neoverse512tvb_tunings): New structures.
13613 (aarch64_adjust_body_cost_sve): Handle -mtune=neoverse-512tvb.
13614 (aarch64_adjust_body_cost): Likewise.
13616 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
13618 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Only
13619 record issue information for operations that occur in the
13622 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
13624 * config/aarch64/aarch64.c (aarch64_multiply_add_p): Add a vec_flags
13625 parameter. Detect cases in which an Advanced SIMD MLA would almost
13626 certainly require a MOV.
13627 (aarch64_count_ops): Update accordingly.
13629 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
13631 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction): New
13632 function, split out from...
13633 (aarch64_detect_vector_stmt_subtype): ...here.
13634 (aarch64_add_stmt_cost): Treat extracting element 0 as free.
13636 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
13638 * config/aarch64/aarch64-protos.h (sve_vec_cost):
13639 Add gather_load_x32_cost and gather_load_x64_cost.
13640 * config/aarch64/aarch64.c (generic_sve_vector_cost)
13641 (a64fx_sve_vector_cost, neoversev1_sve_vector_cost): Update
13642 accordingly, using the values given by the scalar_load * number
13643 of elements calculation that we used previously.
13644 (aarch64_detect_vector_stmt_subtype): Use the new fields.
13646 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
13648 * config/aarch64/aarch64.c (aarch64_adjust_body_cost_sve): New
13649 function, split out from...
13650 (aarch64_adjust_body_cost): ...here.
13652 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
13654 * config/aarch64/fractional-cost.h: New file.
13655 * config/aarch64/aarch64.c: Include <algorithm> (indirectly)
13656 and cost_fraction.h.
13657 (vec_cost_fraction): New typedef.
13658 (aarch64_detect_scalar_stmt_subtype): Use it for statement costs.
13659 (aarch64_detect_vector_stmt_subtype): Likewise.
13660 (aarch64_sve_adjust_stmt_cost, aarch64_adjust_stmt_cost): Likewise.
13661 (aarch64_estimate_min_cycles_per_iter): Use vec_cost_fraction
13663 (aarch64_adjust_body_cost): Likewise.
13664 (aarch64_test_cost_fraction): New function.
13665 (aarch64_run_selftests): Call it.
13667 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
13669 * config/aarch64/aarch64-protos.h (tune_params::sve_width): Turn
13671 * config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Update
13673 (aarch64_estimated_poly_value): Likewise. Use the least significant
13674 set bit for the minimum and likely values. Use the most significant
13675 set bit for the maximum value.
13677 2021-08-03 liuhongt <hongtao.liu@intel.com>
13679 * config/i386/sse.md (cond_<insn><mode>): New expander.
13680 (cond_mul<mode>): Ditto.
13682 2021-08-03 Kewen Lin <linkw@linux.ibm.com>
13684 * tree-cfg.c (move_sese_region_to_fn): Fix typos on dloop.
13686 2021-08-03 liuhongt <hongtao.liu@intel.com>
13688 * config/i386/sse.md (cond_<insn><mode>):New expander.
13689 (cond_mul<mode>): Ditto.
13690 (cond_div<mode>): Ditto.
13692 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
13694 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Also
13695 check stack_realign_needed for stack realignment.
13696 (ix86_legitimate_constant_p): Always allow CONST_WIDE_INT smaller
13697 than the largest integer supported by vector register.
13698 * config/i386/i386.h (MAX_MOVE_MAX): New. Set to 64.
13699 (MOVE_MAX): Set to bytes of the largest integer supported by
13701 (STORE_MAX_PIECES): New.
13703 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
13705 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
13706 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
13707 data from one memory location to another.
13709 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
13711 PR middle-end/90773
13712 * config/i386/i386.c (TARGET_GEN_MEMSET_SCRATCH_RTX): New.
13714 2021-08-02 Aldy Hernandez <aldyh@redhat.com>
13716 PR tree-optimization/101724
13717 * params.opt: Remove --param=threader-iterative.
13718 * tree-ssa-threadbackward.c (pass_thread_jumps::execute): Remove
13721 2021-08-02 Tom de Vries <tdevries@suse.de>
13723 PR middle-end/101665
13724 * doc/extend.texi (nonnull attribute): Improve documentation.
13726 2021-08-02 Andrew Pinski <apinski@marvell.com>
13728 PR rtl-optimization/101683
13729 * rtlanal.c (may_trap_p_1): Handle UNSIGNED_FIX.
13731 2021-08-02 Roger Sayle <roger@nextmovesoftware.com>
13733 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern):
13734 Renamed from cond_removal_in_popcount_clz_ctz_pattern.
13735 Add support for BSWAP, FFS, PARITY and CLRSB builtins.
13736 (tree_ssa_phiop_worker): Update call to function above.
13738 2021-08-01 H.J. Lu <hjl.tools@gmail.com>
13741 * config/i386/i386.md (bsr_rex64_1_zext): New.
13742 (combine splitter for constant - clzll): Replace gen_bsr_rex64_1
13743 with gen_bsr_rex64_1_zext.
13745 2021-07-31 Jakub Jelinek <jakub@redhat.com>
13748 * config/i386/i386.md (bsr_rex64_1, bsr_1, bsr_zext_1): New
13749 define_insn patterns.
13750 (*bsr_rex64_2, *bsr_2): New define_insn_and_split patterns.
13751 Add combine splitters for constant - clz.
13752 (clz<mode>2): Use a temporary pseudo for bsr result.
13754 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
13756 * config/rs6000/smmintrin.h (_mm_floor_pd, _mm_floor_ps,
13757 _mm_floor_sd, _mm_floor_ss): New.
13759 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
13761 * config/rs6000/smmintrin.h (_mm_ceil_pd, _mm_ceil_ps,
13762 _mm_ceil_sd, _mm_ceil_ss): New.
13764 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
13766 * config/rs6000/smmintrin.h (_mm_blend_pd, _mm_blendv_pd,
13767 _mm_blend_ps, _mm_blendv_ps): New.
13769 2021-07-30 Roger Sayle <roger@nextmovesoftware.com>
13770 Uroš Bizjak <ubizjak@gmail.com>
13772 * config/i386/i386.md (*dec_cmov<mode>): New define_insn_and_split
13773 to generate a conditional move using the carry flag after sub $1.
13774 (peephole2): Eliminate a register-to-register move by inverting
13775 the condition of a conditional move.
13777 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
13779 * config/mmix/mmix.md ("call", "call_value", "*call_real")
13780 ("*call_value_real"): Don't generate rtx mentioning the generic
13781 operands 1 and 2 to "call", and similarly for "call_value".
13782 * config/mmix/mmix.c (mmix_print_operand_punct_valid_p)
13783 (mmix_print_operand): Use '!' instead of 'p'.
13785 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
13787 * doc/md.texi (call): Correct information about operand 2.
13788 * config/mmix/mmix.md ("call", "call_value"): Remove fixed FIXMEs.
13790 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
13792 * range-op.cc (operator_trunc_mod::wi_fold): Fold constants.
13794 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
13796 * range-op.cc (operator_div::wi_fold): Return UNDEFINED for [0, 0] divisor.
13798 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
13800 * gimple-range-cache.cc (*::set_bb_range): Change const basic_block to
13801 const_basic_block..
13802 (*::get_bb_range): Ditto.
13803 (*::bb_range_p): Ditto.
13804 * gimple-range-cache.h: Change prototypes.
13806 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
13808 PR middle-end/90773
13809 * builtins.c (builtin_memcpy_read_str): Change the mode argument
13810 from scalar_int_mode to fixed_size_mode.
13811 (builtin_strncpy_read_str): Likewise.
13812 (gen_memset_value_from_prev): New function.
13813 (builtin_memset_read_str): Change the mode argument from
13814 scalar_int_mode to fixed_size_mode. Use gen_memset_value_from_prev
13815 and support CONST_VECTOR.
13816 (builtin_memset_gen_str): Likewise.
13817 (try_store_by_multiple_pieces): Use by_pieces_constfn to declare
13819 * builtins.h (builtin_strncpy_read_str): Replace scalar_int_mode
13820 with fixed_size_mode.
13821 (builtin_memset_read_str): Likewise.
13822 * expr.c (widest_int_mode_for_size): Renamed to ...
13823 (widest_fixed_size_mode_for_size): Add a bool argument to
13824 indicate if QI vector mode can be used.
13825 (by_pieces_ninsns): Call widest_fixed_size_mode_for_size
13826 instead of widest_int_mode_for_size.
13827 (pieces_addr::adjust): Change the mode argument from
13828 scalar_int_mode to fixed_size_mode.
13829 (op_by_pieces_d): Make m_len read-only. Add a bool member,
13830 m_qi_vector_mode, to indicate that QI vector mode can be used.
13831 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
13832 initialize m_qi_vector_mode. Call widest_fixed_size_mode_for_size
13833 instead of widest_int_mode_for_size.
13834 (op_by_pieces_d::get_usable_mode): Change the mode argument from
13835 scalar_int_mode to fixed_size_mode. Call
13836 widest_fixed_size_mode_for_size instead of
13837 widest_int_mode_for_size.
13838 (op_by_pieces_d::smallest_fixed_size_mode_for_size): New member
13839 function to return the smallest integer or QI vector mode.
13840 (op_by_pieces_d::run): Call widest_fixed_size_mode_for_size
13841 instead of widest_int_mode_for_size. Call
13842 smallest_fixed_size_mode_for_size instead of
13843 smallest_int_mode_for_size.
13844 (store_by_pieces_d::store_by_pieces_d): Add a bool argument to
13845 indicate that QI vector mode can be used and pass it to
13846 op_by_pieces_d::op_by_pieces_d.
13847 (can_store_by_pieces): Call widest_fixed_size_mode_for_size
13848 instead of widest_int_mode_for_size. Pass memsetp to
13849 widest_fixed_size_mode_for_size to support QI vector mode.
13850 Allow all CONST_VECTORs for memset if vec_duplicate is supported.
13851 (store_by_pieces): Pass memsetp to
13852 store_by_pieces_d::store_by_pieces_d.
13853 (clear_by_pieces_1): Removed.
13854 (clear_by_pieces): Replace clear_by_pieces_1 with
13855 builtin_memset_read_str and pass true to store_by_pieces_d to
13856 support vector mode broadcast.
13857 (string_cst_read_str): Change the mode argument from
13858 scalar_int_mode to fixed_size_mode.
13859 * expr.h (by_pieces_constfn): Change scalar_int_mode to
13861 (by_pieces_prev): Likewise.
13862 * rtl.h (lowpart_subreg_regno): New.
13863 * rtlanal.c (lowpart_subreg_regno): New. A wrapper around
13864 simplify_subreg_regno.
13865 * target.def (gen_memset_scratch_rtx): New hook.
13866 * doc/tm.texi.in: Add TARGET_GEN_MEMSET_SCRATCH_RTX.
13867 * doc/tm.texi: Regenerated.
13869 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
13872 * config/mips/mips.c (mips_atomic_assign_expand_fenv): Use
13873 TARGET_EXPR instead of MODIFY_EXPR.
13875 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
13878 * config/mips/mips-protos.h (mips_expand_vec_cmp_expr): Declare.
13879 * config/mips/mips.c (mips_expand_vec_cmp_expr): New function.
13880 * config/mips/mips-msa.md (vec_cmp<MSA:mode><mode_i>): New
13882 (vec_cmpu<IMSA:mode><mode_i>): New expander.
13884 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
13887 * config/i386/i386-options.c (ix86_option_override_internal):
13888 Don't enable LZCNT/POPCNT if they have been disabled explicitly.
13890 2021-07-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
13893 * config/arm/arm_neon.h (vld1_p64): Replace call to builtin by
13894 explicitly dereferencing __a.
13895 (vld1_s64): Likewise.
13896 (vld1_u64): Likewise.
13897 * config/arm/arm_neon_builtins.def (vld1): Remove entry for di
13898 and change to VAR13.
13900 2021-07-30 Aldy Hernandez <aldyh@redhat.com>
13902 * gimple-loop-versioning.cc (lv_dom_walker::lv_dom_walker): Remove
13903 use of m_range_analyzer.
13904 (loop_versioning::lv_dom_walker::before_dom_children): Same.
13905 (loop_versioning::lv_dom_walker::after_dom_children): Remove.
13906 (loop_versioning::prune_loop_conditions): Replace vr_values use
13907 with range_query interface.
13908 (pass_loop_versioning::execute): Use ranger.
13910 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
13913 * ipa-devirt.c (ipa_odr_read_section): Compare the precision of
13914 enum values, and emit a warning if they mismatch.
13916 2021-07-30 Kewen Lin <linkw@linux.ibm.com>
13918 * cfgloop.h (as_const): New function.
13919 (class loop_iterator): Rename to ...
13920 (class loops_list): ... this.
13921 (loop_iterator::next): Rename to ...
13922 (loops_list::Iter::fill_curr_loop): ... this and adjust.
13923 (loop_iterator::loop_iterator): Rename to ...
13924 (loops_list::loops_list): ... this and adjust.
13925 (loops_list::Iter): New class.
13926 (loops_list::iterator): New type.
13927 (loops_list::const_iterator): New type.
13928 (loops_list::begin): New function.
13929 (loops_list::end): Likewise.
13930 (loops_list::begin const): Likewise.
13931 (loops_list::end const): Likewise.
13932 (FOR_EACH_LOOP): Remove.
13933 (FOR_EACH_LOOP_FN): Remove.
13934 * cfgloop.c (flow_loops_dump): Adjust FOR_EACH_LOOP* with range-based
13935 for loop with loops_list instance.
13936 (sort_sibling_loops): Likewise.
13937 (disambiguate_loops_with_multiple_latches): Likewise.
13938 (verify_loop_structure): Likewise.
13939 * cfgloopmanip.c (create_preheaders): Likewise.
13940 (force_single_succ_latches): Likewise.
13941 * config/aarch64/falkor-tag-collision-avoidance.c
13942 (execute_tag_collision_avoidance): Likewise.
13943 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Likewise.
13944 * config/s390/s390.c (s390_adjust_loops): Likewise.
13945 * doc/loop.texi: Likewise.
13946 * gimple-loop-interchange.cc (pass_linterchange::execute): Likewise.
13947 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
13948 * gimple-loop-versioning.cc (loop_versioning::analyze_blocks): Likewise.
13949 (loop_versioning::make_versioning_decisions): Likewise.
13950 * gimple-ssa-split-paths.c (split_paths): Likewise.
13951 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
13952 * graphite.c (canonicalize_loop_form): Likewise.
13953 (graphite_transform_loops): Likewise.
13954 * ipa-fnsummary.c (analyze_function_body): Likewise.
13955 * ipa-pure-const.c (analyze_function): Likewise.
13956 * loop-doloop.c (doloop_optimize_loops): Likewise.
13957 * loop-init.c (loop_optimizer_finalize): Likewise.
13958 (fix_loop_structure): Likewise.
13959 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
13960 (move_loop_invariants): Likewise.
13961 * loop-unroll.c (decide_unrolling): Likewise.
13962 (unroll_loops): Likewise.
13963 * modulo-sched.c (sms_schedule): Likewise.
13964 * predict.c (predict_loops): Likewise.
13965 (pass_profile::execute): Likewise.
13966 * profile.c (branch_prob): Likewise.
13967 * sel-sched-ir.c (sel_finish_pipelining): Likewise.
13968 (sel_find_rgns): Likewise.
13969 * tree-cfg.c (replace_loop_annotate): Likewise.
13970 (replace_uses_by): Likewise.
13971 (move_sese_region_to_fn): Likewise.
13972 * tree-if-conv.c (pass_if_conversion::execute): Likewise.
13973 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
13974 * tree-parloops.c (parallelize_loops): Likewise.
13975 * tree-predcom.c (tree_predictive_commoning): Likewise.
13976 * tree-scalar-evolution.c (scev_initialize): Likewise.
13977 (scev_reset): Likewise.
13978 * tree-ssa-dce.c (find_obviously_necessary_stmts): Likewise.
13979 * tree-ssa-live.c (remove_unused_locals): Likewise.
13980 * tree-ssa-loop-ch.c (ch_base::copy_headers): Likewise.
13981 * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
13982 (tree_ssa_lim_initialize): Likewise.
13983 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Likewise.
13984 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize): Likewise.
13985 * tree-ssa-loop-manip.c (get_loops_exits): Likewise.
13986 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations): Likewise.
13987 (free_numbers_of_iterations_estimates): Likewise.
13988 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Likewise.
13989 * tree-ssa-loop-split.c (tree_ssa_split_loops): Likewise.
13990 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Likewise.
13991 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
13992 (pass_scev_cprop::execute): Likewise.
13993 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
13994 * tree-ssa-sccvn.c (do_rpo_vn): Likewise.
13995 * tree-ssa-threadupdate.c
13996 (jump_thread_path_registry::thread_through_all_blocks): Likewise.
13997 * tree-vectorizer.c (vectorize_loops): Likewise.
13998 * tree-vrp.c (vrp_asserts::find_assert_locations): Likewise.
14000 2021-07-29 Hans-Peter Nilsson <hp@bitrange.com>
14002 * config/mmix/mmix.c (mmix_function_arg_1): Avoid
14003 generating a VOIDmode register for e.g the
14004 function_arg_info::end_marker.
14006 2021-07-29 Jeff Law <jeffreyalaw@gmail.com>
14008 * config/h8300/h8300-modes.def: Add CCZ, CCV and CCC, drop CCZNV.
14009 * config/h8300/h8300.md (H8cc mode iterator): Add CCZ.
14010 (cc mode_attr): Similarly.
14011 (ccz subst_attr): Similarly.
14012 * config/h8300/jumpcall.md: Add new patterns for branch-on-bit.
14013 * config/h8300/testcompare.md: Remove various cc0 based patterns
14014 that had been commented out. Add pattern to set CCZ from a bit
14017 2021-07-29 Thomas Schwinge <thomas@codesourcery.com>
14018 Julian Brown <julian@codesourcery.com>
14019 Kwok Cheung Yeung <kcy@codesourcery.com>
14021 * omp-offload.c (oacc_loop_xform_head_tail, oacc_loop_process):
14022 'update_stmt' after modification.
14023 (pass_oacc_loop_designation): New function, extracted out of...
14024 (pass_oacc_device_lower): ... this.
14025 (pass_data_oacc_loop_designation, pass_oacc_loop_designation)
14026 (make_pass_oacc_loop_designation): New
14027 * passes.def: Add it.
14028 * tree-parloops.c (create_parallel_loop): Adjust.
14029 * tree-pass.h (make_pass_oacc_loop_designation): New.
14031 2021-07-29 Aldy Hernandez <aldyh@redhat.com>
14033 * flag-types.h (enum threader_mode): New.
14034 * params.opt: Add entry for --param=threader-mode.
14035 * tree-ssa-threadbackward.c (THREADER_ITERATIVE_MODE): New.
14036 (class back_threader): New.
14037 (back_threader::back_threader): New.
14038 (back_threader::~back_threader): New.
14039 (back_threader::maybe_register_path): New.
14040 (back_threader::find_taken_edge): New.
14041 (back_threader::find_taken_edge_switch): New.
14042 (back_threader::find_taken_edge_cond): New.
14043 (back_threader::resolve_def): New.
14044 (back_threader::resolve_phi): New.
14045 (back_threader::find_paths_to_names): New.
14046 (back_threader::find_paths): New.
14049 (thread_jumps::find_jump_threads_backwards): Call ranger threader.
14050 (thread_jumps::find_jump_threads_backwards_with_ranger): New.
14051 (pass_thread_jumps::execute): Abstract out code...
14052 (try_thread_blocks): ...here.
14053 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
14054 Abstract out threading candidate code to...
14055 (single_succ_to_potentially_threadable_block): ...here.
14056 * tree-ssa-threadedge.h (single_succ_to_potentially_threadable_block):
14058 * tree-ssa-threadupdate.c (register_jump_thread): Return boolean.
14059 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
14060 Return bool from register_jump_thread.
14062 2021-07-29 Andreas Krebbel <krebbel@linux.ibm.com>
14064 * target.def: in0 and in1 do not need to be registers.
14065 * doc/tm.texi: Regenerate.
14067 2021-07-29 liuhongt <hongtao.liu@intel.com>
14070 * config/i386/i386.c (ix86_widen_mult_cost): New function.
14071 (ix86_add_stmt_cost): Use ix86_widen_mult_cost for
14074 2021-07-29 Jiufu Guo <guojiufu@linux.ibm.com>
14077 * config/rs6000/rs6000.c (TARGET_PREFERRED_DOLOOP_MODE): New hook.
14078 (rs6000_preferred_doloop_mode): New hook.
14079 * doc/tm.texi: Regenerate.
14080 * doc/tm.texi.in: Add hook preferred_doloop_mode.
14081 * target.def (preferred_doloop_mode): New hook.
14082 * targhooks.c (default_preferred_doloop_mode): New hook.
14083 * targhooks.h (default_preferred_doloop_mode): New hook.
14084 * tree-ssa-loop-ivopts.c (compute_doloop_base_on_mode): New function.
14085 (add_iv_candidate_for_doloop): Call targetm.preferred_doloop_mode
14086 and compute_doloop_base_on_mode.
14088 2021-07-28 Martin Sebor <msebor@redhat.com>
14090 PR middle-end/101494
14091 * tree-ssa-uninit.c (maybe_warn_operand): Correct object offset
14092 and size computation.
14094 2021-07-28 Martin Sebor <msebor@redhat.com>
14096 PR middle-end/101601
14097 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Remove
14099 Handle pointers to functions.
14101 2021-07-28 Martin Sebor <msebor@redhat.com>
14103 * Makefile.in (OBJS): Add gimple-ssa-warn-access.o and pointer-query.o.
14104 * attribs.h (fndecl_dealloc_argno): Move fndecl_dealloc_argno to tree.h.
14105 * builtins.c (compute_objsize_r): Move to pointer-query.cc.
14106 (access_ref::access_ref): Same.
14107 (access_ref::phi): Same.
14108 (access_ref::get_ref): Same.
14109 (access_ref::size_remaining): Same.
14110 (access_ref::offset_in_range): Same.
14111 (access_ref::add_offset): Same.
14112 (access_ref::inform_access): Same.
14113 (ssa_name_limit_t::visit_phi): Same.
14114 (ssa_name_limit_t::leave_phi): Same.
14115 (ssa_name_limit_t::next): Same.
14116 (ssa_name_limit_t::next_phi): Same.
14117 (ssa_name_limit_t::~ssa_name_limit_t): Same.
14118 (pointer_query::pointer_query): Same.
14119 (pointer_query::get_ref): Same.
14120 (pointer_query::put_ref): Same.
14121 (pointer_query::flush_cache): Same.
14122 (warn_string_no_nul): Move to gimple-ssa-warn-access.cc.
14123 (check_nul_terminated_array): Same.
14124 (unterminated_array): Same.
14125 (maybe_warn_for_bound): Same.
14126 (check_read_access): Same.
14127 (warn_for_access): Same.
14128 (get_size_range): Same.
14129 (check_access): Same.
14130 (gimple_call_alloc_size): Move to tree.c.
14131 (gimple_parm_array_size): Move to pointer-query.cc.
14132 (get_offset_range): Same.
14133 (gimple_call_return_array): Same.
14134 (handle_min_max_size): Same.
14135 (handle_array_ref): Same.
14136 (handle_mem_ref): Same.
14137 (compute_objsize): Same.
14138 (gimple_call_alloc_p): Move to gimple-ssa-warn-access.cc.
14139 (call_dealloc_argno): Same.
14140 (fndecl_dealloc_argno): Same.
14141 (new_delete_mismatch_p): Same.
14142 (matching_alloc_calls_p): Same.
14143 (warn_dealloc_offset): Same.
14144 (maybe_emit_free_warning): Same.
14145 * builtins.h (check_nul_terminated_array): Move to
14146 gimple-ssa-warn-access.h.
14147 (check_nul_terminated_array): Same.
14148 (warn_string_no_nul): Same.
14149 (unterminated_array): Same.
14150 (class ssa_name_limit_t): Same.
14151 (class pointer_query): Same.
14152 (struct access_ref): Same.
14153 (class range_query): Same.
14154 (struct access_data): Same.
14155 (gimple_call_alloc_size): Same.
14156 (gimple_parm_array_size): Same.
14157 (compute_objsize): Same.
14158 (class access_data): Same.
14159 (maybe_emit_free_warning): Same.
14160 * calls.c (initialize_argument_information): Remove call to
14161 maybe_emit_free_warning.
14162 * gimple-array-bounds.cc: Include new header..
14163 * gimple-fold.c: Same.
14164 * gimple-ssa-sprintf.c: Same.
14165 * gimple-ssa-warn-restrict.c: Same.
14166 * passes.def: Add pass_warn_access.
14167 * tree-pass.h (make_pass_warn_access): Declare.
14168 * tree-ssa-strlen.c: Include new headers.
14169 * tree.c (fndecl_dealloc_argno): Move here from builtins.c.
14170 * tree.h (fndecl_dealloc_argno): Move here from attribs.h.
14171 * gimple-ssa-warn-access.cc: New file.
14172 * gimple-ssa-warn-access.h: New file.
14173 * pointer-query.cc: New file.
14174 * pointer-query.h: New file.
14176 2021-07-28 Jakub Jelinek <jakub@redhat.com>
14178 PR middle-end/101624
14179 * ubsan.c (maybe_instrument_pointer_overflow,
14180 instrument_object_size): Only test DECL_REGISTER on VAR_DECLs,
14181 PARM_DECLs or RESULT_DECLs.
14182 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
14184 2021-07-28 Jakub Jelinek <jakub@redhat.com>
14186 PR middle-end/101642
14187 * match.pd (bswap16 (x) == bswap16 (y)): Cast both operands
14188 to type of bswap16 for comparison.
14189 (bswap16 (x) == cst): Cast bswap16 operand to type of cst.
14191 2021-07-28 Richard Biener <rguenther@suse.de>
14193 PR tree-optimization/101615
14194 * tree-vect-slp.c (vect_optimize_slp): Materialize permutes
14195 at CTOR SLP graph entries.
14197 2021-07-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14199 * config/aarch64/aarch64.md (*extend<SHORT:mode><GPI:mode>2_aarch64):
14200 Add "r,w" alternative.
14202 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
14205 * config/i386/i386.c (ix86_avx_u128_mode_needed): Don't set
14206 AVX_U128_DIRTY when all bits are zero.
14208 2021-07-28 Richard Biener <rguenther@suse.de>
14210 PR tree-optimization/101615
14211 * tree-vect-slp.c (vect_optimize_slp): Pre-existing vector
14212 external nodes cannot be permuted so make them perm_out 0.
14214 2021-07-28 Andrew Stubbs <ams@codesourcery.com>
14217 * config.in: Regenerate.
14218 * config/gcn/gcn-hsa.h (A_FIJI): New define.
14219 (A_900): New define.
14220 (A_906): New define.
14221 (A_908): New define.
14222 (ASM_SPEC): Use A_FIJI, A_900, A_906 and A_908.
14223 * config/gcn/gcn.c (output_file_start): Adjust attributes according
14224 to the assembler capabilities.
14225 * config/gcn/mkoffload.c (main): Likewise.
14226 * configure: Regenerate.
14227 * configure.ac: Add tests for LLVM assembler attribute features.
14229 2021-07-28 Andrew MacLeod <amacleod@redhat.com>
14231 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Check for
14232 cond_false and cond_true on branches.
14234 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
14236 * config/aarch64/aarch64.c (aarch64_gen_adjusted_ldpstp): use
14239 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
14241 * alias.c (init_alias_analysis): Don't skip prologue/epilogue.
14243 2021-07-28 Jakub Jelinek <jakub@redhat.com>
14246 * config/i386/sse.md (vashr<mode>3): Split into vashrv8di3 expander
14247 and vashrv4di3 expander, where the latter requires just TARGET_AVX2
14248 and has special !TARGET_AVX512VL expansion.
14249 (vashrv2di3<mask_name>): Rename to ...
14250 (vashrv2di3): ... this. Change condition to TARGET_XOP || TARGET_AVX2
14251 and add special !TARGET_XOP && !TARGET_AVX512VL expansion.
14253 2021-07-28 Martin Uecker <muecker@gwdg.de>
14255 * calls.c (maybe_warn_rdwr_sizes): Correct argument
14256 numbers in warning that were switched.
14258 2021-07-28 Kewen Lin <linkw@linux.ibm.com>
14260 PR tree-optimization/101596
14261 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Fix wrong check
14262 by using new_type's precision instead.
14264 2021-07-28 liuhongt <hongtao.liu@intel.com>
14267 * config/i386/i386.h (processor_costs): Add new member
14269 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
14270 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
14271 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
14272 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
14273 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
14274 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
14275 generic_cost, core_cost): Initialize integer_to_sse same value
14277 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
14278 * config/i386/i386.c (ix86_builtin_vectorization_cost):
14279 Use integer_to_sse instead of sse_op to calculate the cost of
14282 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
14284 * config/rs6000/rs6000-gen-builtins.c (write_ovld_static_init): New
14286 (write_init_file): Call write_ovld_static_init.
14288 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
14290 * config/rs6000/rs6000-gen-builtins.c (write_bif_static_init): New
14292 (write_init_file): Call write_bif_static_init.
14294 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
14296 * config/rs6000/rs6000-gen-builtins.c (typemap): New struct.
14297 (TYPE_MAP_SIZE): New macro.
14298 (type_map): New initialized variable.
14299 (typemap_cmp): New function.
14300 (write_type_node): Likewise.
14301 (write_fntype_init): Implement.
14303 2021-07-27 Martin Sebor <msebor@redhat.com>
14305 PR tree-optimization/101584
14306 * tree-ssa-uninit.c (builtin_call_nomodifying_p): New function.
14307 (check_defs): Call it.
14309 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
14311 * tree-ssa-dom.c (dom_jump_threader_simplifier):
14312 Put avail_exprs_stack in the class, instead of passing it to
14313 jump_threader_simplifier.
14314 (dom_jump_threader_simplifier::simplify): Add state argument.
14315 (dom_opt_dom_walker): Add state.
14316 (pass_dominator::execute): Pass state to threader.
14317 (dom_opt_dom_walker::before_dom_children): Use state.
14318 * tree-ssa-threadedge.c (jump_threader::jump_threader): Replace
14319 arguments by state.
14320 (jump_threader::record_temporary_equivalences_from_phis):
14321 Register equivalences through the state variable.
14322 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
14323 Record ranges in a statement through the state variable.
14324 (jump_threader::simplify_control_stmt_condition): Pass state to
14326 (jump_threader::simplify_control_stmt_condition_1): Same.
14327 (jump_threader::thread_around_empty_blocks): Remove obsolete
14329 (jump_threader::thread_through_normal_block): Record equivalences
14330 on edge through the state variable.
14331 (jump_threader::thread_across_edge): Abstract state pushing.
14332 (jt_state::jt_state): New.
14333 (jt_state::push): New.
14334 (jt_state::pop): New.
14335 (jt_state::register_equiv): New.
14336 (jt_state::record_ranges_from_stmt): New.
14337 (jt_state::register_equivs_on_edge): New.
14338 (jump_threader_simplifier::jump_threader_simplifier): Move from
14340 (jump_threader_simplifier::simplify): Add state argument.
14341 * tree-ssa-threadedge.h (class jt_state): New.
14342 (class jump_threader): Add state to constructor.
14343 (class jump_threader_simplifier): Add state to simplify. Remove
14344 avail_exprs_stack from class.
14345 * tree-vrp.c (vrp_jump_threader_simplifier::simplify): Add state
14347 (vrp_jump_threader::vrp_jump_threader): Add state.
14348 (vrp_jump_threader::~vrp_jump_threader): Cleanup state.
14350 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
14352 * Makefile.in (OBJS): Add gimple-range-path.o.
14353 * gimple-range-path.cc: New file.
14354 * gimple-range-path.h: New file.
14356 2021-07-27 Jonathan Wright <jonathan.wright@arm.com>
14358 * config/aarch64/aarch64-simd.md: Push sign/zero-extension
14359 inside vec_duplicate for all patterns.
14360 * simplify-rtx.c (simplify_context::simplify_unary_operation_1):
14361 Push sign/zero-extension inside vec_duplicate.
14363 2021-07-27 Richard Biener <rguenther@suse.de>
14365 PR tree-optimization/101573
14366 * tree-ssa-uninit.c (warn_uninit_phi_uses): New function
14367 looking at uninitialized PHI arg defs in some constrained cases.
14368 (warn_uninitialized_vars): Call it.
14369 (execute_early_warn_uninitialized): Calculate dominators.
14371 2021-07-27 Richard Biener <rguenther@suse.de>
14373 PR tree-optimization/39821
14374 * tree-vect-stmts.c (vect_model_promotion_demotion_cost): Use
14375 vector_stmt for widening arithmetic.
14376 (vectorizable_conversion): Adjust.
14378 2021-07-27 Martin Jambor <mjambor@suse.cz>
14380 * cgraph.h (ipa_replace_map): New field force_load_ref.
14381 * ipa-prop.h (ipa_param_descriptor): Reduce precision of move_cost,
14382 aded new flag load_dereferenced, adjusted comments.
14383 (ipa_get_param_dereferenced): New function.
14384 (ipa_set_param_dereferenced): Likewise.
14385 * cgraphclones.c (cgraph_node::create_virtual_clone): Follow it.
14386 * ipa-cp.c: Include gimple.h.
14387 (ipcp_discover_new_direct_edges): Take into account dereferenced flag.
14388 (get_replacement_map): New parameter force_load_ref, set the
14389 appropriate flag in ipa_replace_map if set.
14390 (struct symbol_and_index_together): New type.
14391 (adjust_refs_in_act_callers): New function.
14392 (adjust_references_in_caller): Likewise.
14393 (create_specialized_node): When appropriate, call
14394 adjust_references_in_caller and force only load references.
14395 * ipa-prop.c (load_from_dereferenced_name): New function.
14396 (ipa_analyze_controlled_uses): Also detect loads from a
14397 dereference, harden testing of call statements.
14398 (ipa_write_node_info): Stream the dereferenced flag.
14399 (ipa_read_node_info): Likewise.
14400 (ipa_set_jf_constant): Also create refdesc when jump function
14401 references a variable.
14402 (cgraph_node_for_jfunc): Rename to symtab_node_for_jfunc, work
14403 also on references of variables and return a symtab_node. Adjust
14405 (propagate_controlled_uses): Also remove references to VAR_DECLs.
14407 2021-07-27 Jakub Jelinek <jakub@redhat.com>
14409 PR middle-end/101586
14410 * gimple-fold.c (clear_padding_type): Ignore FIELD_DECLs with byte
14411 positions above or equal to sz except for diagnostics of flexible
14414 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
14416 PR tree-optimization/78888
14417 * gimple-range-fold.cc (get_letter_range): New.
14418 (fold_using_range::range_of_builtin_call): Call get_letter_range.
14420 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
14422 PR tree-optimization/78888
14423 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Add cases
14424 for CFN_BUILT_IN_TOUPPER and CFN_BUILT_IN_TOLOWER.
14426 2021-07-26 Roger Sayle <roger@nextmovesoftware.com>
14427 Marc Glisse <marc.glisse@inria.fr>
14429 * match.pd (rotate): Simplify equality/inequality of rotations.
14430 (bswap): Simplify equality/inequality tests of byte swapping.
14432 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
14434 * range-op.cc (operator_bitwise_xor::op1_op2_relation_effect):
14437 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
14439 * range-op.cc (operator_lshift::fold_range): Pass rel to
14440 base class fold_range.
14441 (operator_rshift::fold_range): Same.
14443 2021-07-26 Ashimida <ashimida@linux.alibaba.com>
14446 * toplev.h (min_align_loops_log): Remove declaration.
14447 (min_align_jumps_log, min_align_labels_log): Likewise.
14448 (min_align_functions_log): Likewise.
14450 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
14452 * tree-vrp.c (vrp_simplify_cond_using_ranges): Rename vr_values
14454 (execute_vrp): Abstract out simplification of conditionals...
14455 (simplify_casted_conds): ...here.
14457 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
14459 * gimple-array-bounds.cc (array_bounds_checker::get_value_range):
14460 Add gimple argument.
14461 (array_bounds_checker::check_array_ref): Same.
14462 (array_bounds_checker::check_addr_expr): Same.
14463 (array_bounds_checker::check_array_bounds): Pass statement to
14464 check_array_bounds and check_addr_expr.
14465 * gimple-array-bounds.h (check_array_bounds): Add gimple argument.
14466 (check_addr_expr): Same.
14467 (get_value_range): Same.
14469 2021-07-26 Tamar Christina <tamar.christina@arm.com>
14471 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): Rename to..
14472 (sdot_prod, udot_prod): ... This.
14473 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): Merged
14475 (<sur>dot_prod<vsi2qi>): ... this.
14476 (aarch64_<sur>dot_lane<vsi2qi>, aarch64_<sur>dot_laneq<vsi2qi>):
14477 Change operands order.
14478 (<sur>sadv16qi): Use new operands order.
14479 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32,
14480 vdotq_s32): Use new RTL ordering.
14482 2021-07-26 Tamar Christina <tamar.christina@arm.com>
14484 * config/aarch64/aarch64-builtins.c (TYPES_TERNOP_SUSS,
14485 aarch64_types_ternop_suss_qualifiers): New.
14486 * config/aarch64/aarch64-simd-builtins.def (usdot_prod): Use it.
14487 * config/aarch64/aarch64-simd.md (usdot_prod<vsi2qi>): Re-organize RTL.
14488 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Use it.
14490 2021-07-23 Jakub Jelinek <jakub@redhat.com>
14492 PR rtl-optimization/101562
14493 * expmed.c (store_integral_bit_field): Only use movstrict_optab
14494 if the operand isn't paradoxical.
14496 2021-07-23 Aldy Hernandez <aldyh@redhat.com>
14498 * gimple-array-bounds.h (class array_bounds_checker): Change
14499 ranges type to range_query.
14501 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
14503 * config/aarch64/arm_neon.h (vst1_s64_x2): Use
14504 __builtin_memcpy instead of constructing
14505 __builtin_aarch64_simd_oi one vector at a time.
14506 (vst1_u64_x2): Likewise.
14507 (vst1_f64_x2): Likewise.
14508 (vst1_s8_x2): Likewise.
14509 (vst1_p8_x2): Likewise.
14510 (vst1_s16_x2): Likewise.
14511 (vst1_p16_x2): Likewise.
14512 (vst1_s32_x2): Likewise.
14513 (vst1_u8_x2): Likewise.
14514 (vst1_u16_x2): Likewise.
14515 (vst1_u32_x2): Likewise.
14516 (vst1_f16_x2): Likewise.
14517 (vst1_f32_x2): Likewise.
14518 (vst1_p64_x2): Likewise.
14519 (vst1q_s8_x2): Likewise.
14520 (vst1q_p8_x2): Likewise.
14521 (vst1q_s16_x2): Likewise.
14522 (vst1q_p16_x2): Likewise.
14523 (vst1q_s32_x2): Likewise.
14524 (vst1q_s64_x2): Likewise.
14525 (vst1q_u8_x2): Likewise.
14526 (vst1q_u16_x2): Likewise.
14527 (vst1q_u32_x2): Likewise.
14528 (vst1q_u64_x2): Likewise.
14529 (vst1q_f16_x2): Likewise.
14530 (vst1q_f32_x2): Likewise.
14531 (vst1q_f64_x2): Likewise.
14532 (vst1q_p64_x2): Likewise.
14534 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
14536 * config/aarch64/arm_neon.h (vst1_s64_x3): Use
14537 __builtin_memcpy instead of constructing
14538 __builtin_aarch64_simd_ci one vector at a time.
14539 (vst1_u64_x3): Likewise.
14540 (vst1_f64_x3): Likewise.
14541 (vst1_s8_x3): Likewise.
14542 (vst1_p8_x3): Likewise.
14543 (vst1_s16_x3): Likewise.
14544 (vst1_p16_x3): Likewise.
14545 (vst1_s32_x3): Likewise.
14546 (vst1_u8_x3): Likewise.
14547 (vst1_u16_x3): Likewise.
14548 (vst1_u32_x3): Likewise.
14549 (vst1_f16_x3): Likewise.
14550 (vst1_f32_x3): Likewise.
14551 (vst1_p64_x3): Likewise.
14552 (vst1q_s8_x3): Likewise.
14553 (vst1q_p8_x3): Likewise.
14554 (vst1q_s16_x3): Likewise.
14555 (vst1q_p16_x3): Likewise.
14556 (vst1q_s32_x3): Likewise.
14557 (vst1q_s64_x3): Likewise.
14558 (vst1q_u8_x3): Likewise.
14559 (vst1q_u16_x3): Likewise.
14560 (vst1q_u32_x3): Likewise.
14561 (vst1q_u64_x3): Likewise.
14562 (vst1q_f16_x3): Likewise.
14563 (vst1q_f32_x3): Likewise.
14564 (vst1q_f64_x3): Likewise.
14565 (vst1q_p64_x3): Likewise.
14567 2021-07-23 H.J. Lu <hjl.tools@gmail.com>
14570 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): Don't return
14571 hard register when LRA is in progress.
14573 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
14575 * config/aarch64/arm_neon.h (vst1_s8_x4): Use
14576 __builtin_memcpy instead of using a union.
14577 (vst1q_s8_x4): Likewise.
14578 (vst1_s16_x4): Likewise.
14579 (vst1q_s16_x4): Likewise.
14580 (vst1_s32_x4): Likewise.
14581 (vst1q_s32_x4): Likewise.
14582 (vst1_u8_x4): Likewise.
14583 (vst1q_u8_x4): Likewise.
14584 (vst1_u16_x4): Likewise.
14585 (vst1q_u16_x4): Likewise.
14586 (vst1_u32_x4): Likewise.
14587 (vst1q_u32_x4): Likewise.
14588 (vst1_f16_x4): Likewise.
14589 (vst1q_f16_x4): Likewise.
14590 (vst1_f32_x4): Likewise.
14591 (vst1q_f32_x4): Likewise.
14592 (vst1_p8_x4): Likewise.
14593 (vst1q_p8_x4): Likewise.
14594 (vst1_p16_x4): Likewise.
14595 (vst1q_p16_x4): Likewise.
14596 (vst1_s64_x4): Likewise.
14597 (vst1_u64_x4): Likewise.
14598 (vst1_p64_x4): Likewise.
14599 (vst1q_s64_x4): Likewise.
14600 (vst1q_u64_x4): Likewise.
14601 (vst1q_p64_x4): Likewise.
14602 (vst1_f64_x4): Likewise.
14603 (vst1q_f64_x4): Likewise.
14605 2021-07-23 Jonathan Wrightt <jonathan.wright@arm.com>
14607 * config/aarch64/arm_neon.h (vst2_s64): Use __builtin_memcpy
14608 instead of constructing __builtin_aarch64_simd_oi one vector
14610 (vst2_u64): Likewise.
14611 (vst2_f64): Likewise.
14612 (vst2_s8): Likewise.
14613 (vst2_p8): Likewise.
14614 (vst2_s16): Likewise.
14615 (vst2_p16): Likewise.
14616 (vst2_s32): Likewise.
14617 (vst2_u8): Likewise.
14618 (vst2_u16): Likewise.
14619 (vst2_u32): Likewise.
14620 (vst2_f16): Likewise.
14621 (vst2_f32): Likewise.
14622 (vst2_p64): Likewise.
14623 (vst2q_s8): Likewise.
14624 (vst2q_p8): Likewise.
14625 (vst2q_s16): Likewise.
14626 (vst2q_p16): Likewise.
14627 (vst2q_s32): Likewise.
14628 (vst2q_s64): Likewise.
14629 (vst2q_u8): Likewise.
14630 (vst2q_u16): Likewise.
14631 (vst2q_u32): Likewise.
14632 (vst2q_u64): Likewise.
14633 (vst2q_f16): Likewise.
14634 (vst2q_f32): Likewise.
14635 (vst2q_f64): Likewise.
14636 (vst2q_p64): Likewise.
14638 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
14640 * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy
14641 instead of constructing __builtin_aarch64_simd_ci one vector
14643 (vst3_u64): Likewise.
14644 (vst3_f64): Likewise.
14645 (vst3_s8): Likewise.
14646 (vst3_p8): Likewise.
14647 (vst3_s16): Likewise.
14648 (vst3_p16): Likewise.
14649 (vst3_s32): Likewise.
14650 (vst3_u8): Likewise.
14651 (vst3_u16): Likewise.
14652 (vst3_u32): Likewise.
14653 (vst3_f16): Likewise.
14654 (vst3_f32): Likewise.
14655 (vst3_p64): Likewise.
14656 (vst3q_s8): Likewise.
14657 (vst3q_p8): Likewise.
14658 (vst3q_s16): Likewise.
14659 (vst3q_p16): Likewise.
14660 (vst3q_s32): Likewise.
14661 (vst3q_s64): Likewise.
14662 (vst3q_u8): Likewise.
14663 (vst3q_u16): Likewise.
14664 (vst3q_u32): Likewise.
14665 (vst3q_u64): Likewise.
14666 (vst3q_f16): Likewise.
14667 (vst3q_f32): Likewise.
14668 (vst3q_f64): Likewise.
14669 (vst3q_p64): Likewise.
14671 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
14673 * config/aarch64/arm_neon.h (vst4_s64): Use __builtin_memcpy
14674 instead of constructing __builtin_aarch64_simd_xi one vector
14676 (vst4_u64): Likewise.
14677 (vst4_f64): Likewise.
14678 (vst4_s8): Likewise.
14679 (vst4_p8): Likewise.
14680 (vst4_s16): Likewise.
14681 (vst4_p16): Likewise.
14682 (vst4_s32): Likewise.
14683 (vst4_u8): Likewise.
14684 (vst4_u16): Likewise.
14685 (vst4_u32): Likewise.
14686 (vst4_f16): Likewise.
14687 (vst4_f32): Likewise.
14688 (vst4_p64): Likewise.
14689 (vst4q_s8): Likewise.
14690 (vst4q_p8): Likewise.
14691 (vst4q_s16): Likewise.
14692 (vst4q_p16): Likewise.
14693 (vst4q_s32): Likewise.
14694 (vst4q_s64): Likewise.
14695 (vst4q_u8): Likewise.
14696 (vst4q_u16): Likewise.
14697 (vst4q_u32): Likewise.
14698 (vst4q_u64): Likewise.
14699 (vst4q_f16): Likewise.
14700 (vst4q_f32): Likewise.
14701 (vst4q_f64): Likewise.
14702 (vst4q_p64): Likewise.
14704 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
14706 * config/aarch64/arm_neon.h (vtbx4_s8): Use __builtin_memcpy
14707 instead of constructing __builtin_aarch64_simd_oi one vector
14709 (vtbx4_u8): Likewise.
14710 (vtbx4_p8): Likewise.
14712 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
14714 * config/aarch64/arm_neon.h (vtbl3_s8): Use __builtin_memcpy
14715 instead of constructing __builtin_aarch64_simd_oi one vector
14717 (vtbl3_u8): Likewise.
14718 (vtbl3_p8): Likewise.
14719 (vtbl4_s8): Likewise.
14720 (vtbl4_u8): Likewise.
14721 (vtbl4_p8): Likewise.
14723 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
14725 * config/aarch64/arm_neon.h (vqtbx2_s8): Use __builtin_memcpy
14726 instead of constructing __builtin_aarch64_simd_oi one vector
14728 (vqtbx2_u8): Likewise.
14729 (vqtbx2_p8): Likewise.
14730 (vqtbx2q_s8): Likewise.
14731 (vqtbx2q_u8): Likewise.
14732 (vqtbx2q_p8): Likewise.
14733 (vqtbx3_s8): Use __builtin_memcpy instead of constructing
14734 __builtin_aarch64_simd_ci one vector at a time.
14735 (vqtbx3_u8): Likewise.
14736 (vqtbx3_p8): Likewise.
14737 (vqtbx3q_s8): Likewise.
14738 (vqtbx3q_u8): Likewise.
14739 (vqtbx3q_p8): Likewise.
14740 (vqtbx4_s8): Use __builtin_memcpy instead of constructing
14741 __builtin_aarch64_simd_xi one vector at a time.
14742 (vqtbx4_u8): Likewise.
14743 (vqtbx4_p8): Likewise.
14744 (vqtbx4q_s8): Likewise.
14745 (vqtbx4q_u8): Likewise.
14746 (vqtbx4q_p8): Likewise.
14748 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
14750 * config/aarch64/arm_neon.h (vqtbl2_s8): Use __builtin_memcpy
14751 instead of constructing __builtin_aarch64_simd_oi one vector
14753 (vqtbl2_u8): Likewise.
14754 (vqtbl2_p8): Likewise.
14755 (vqtbl2q_s8): Likewise.
14756 (vqtbl2q_u8): Likewise.
14757 (vqtbl2q_p8): Likewise.
14758 (vqtbl3_s8): Use __builtin_memcpy instead of constructing
14759 __builtin_aarch64_simd_ci one vector at a time.
14760 (vqtbl3_u8): Likewise.
14761 (vqtbl3_p8): Likewise.
14762 (vqtbl3q_s8): Likewise.
14763 (vqtbl3q_u8): Likewise.
14764 (vqtbl3q_p8): Likewise.
14765 (vqtbl4_s8): Use __builtin_memcpy instead of constructing
14766 __builtin_aarch64_simd_xi one vector at a time.
14767 (vqtbl4_u8): Likewise.
14768 (vqtbl4_p8): Likewise.
14769 (vqtbl4q_s8): Likewise.
14770 (vqtbl4q_u8): Likewise.
14771 (vqtbl4q_p8): Likewise.
14773 2021-07-23 Haochen Gui <guihaoc@gcc.gnu.org>
14776 * config/rs6000/rs6000.md (cstore<mode>4): Fix wrong fall through.
14778 2021-07-22 Andrew Pinski <apinski@marvell.com>
14780 PR tree-optimization/10153
14781 * tree-tailcall.c (create_tailcall_accumulator):
14782 Don't call fold_convert as the type should be correct already.
14783 (tree_optimize_tail_calls_1): Use build_{one,zero}_cst instead
14784 of integer_{one,zero}_node for the call of create_tailcall_accumulator.
14786 2021-07-22 Aldy Hernandez <aldyh@redhat.com>
14788 * gimple-range-cache.cc (non_null_ref::adjust_range): Replace
14789 varying_p check for null/non-null check.
14791 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
14793 PR tree-optimization/101511
14794 * value-relation.cc (relation_oracle::query_relation): Check if ssa1
14795 is in ssa2's equiv set, and don't trap if so.
14797 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
14799 PR tree-optimization/101497
14800 * gimple-range-fold.cc (fold_using_range::range_of_cond_expr): Check
14803 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
14805 PR tree-optimization/101496
14806 * vr-values.c (simplify_using_ranges::fold_cond): Call range_of_stmt
14807 first, then vrp_visit_cond_Stmt.
14809 2021-07-22 liuhongt <hongtao.liu@intel.com>
14811 * config/i386/i386-expand.c
14812 (ix86_broadcast_from_integer_constant): Rename to ..
14813 (ix86_broadcast_from_constant): .. this, and extend it to
14815 (ix86_expand_vector_move): Extend to float mode.
14816 * config/i386/i386-features.c
14817 (replace_constant_pool_with_broadcast): Remove.
14818 (remove_partial_avx_dependency_gate): Ditto.
14819 (constant_pool_broadcast): Ditto.
14820 (class pass_constant_pool_broadcast): Ditto.
14821 (make_pass_constant_pool_broadcast): Ditto.
14822 (remove_partial_avx_dependency): Adjust gate.
14823 * config/i386/i386-passes.def: Remove pass_constant_pool_broadcast.
14824 * config/i386/i386-protos.h
14825 (make_pass_constant_pool_broadcast): Remove.
14827 2021-07-22 liuhongt <hongtao.liu@intel.com>
14829 * config/i386/constraints.md (Wb): New constraint.
14831 * config/i386/i386.md (*ashlhi3_1): Extend to avx512 mask
14833 (*ashlqi3_1): Ditto.
14834 (*<insn><mode>3_1): Split to ..
14835 (*ashr<mode>3_1): this, ...
14836 (*lshr<mode>3_1): and this, also extend this pattern to avx512
14838 (*<insn><mode>3_1): Split to ..
14839 (*ashr<mode>3_1): this, ...
14840 (*lshrqi3_1): and this, also extend this pattern to avx512
14842 (*lshrhi3_1): And this, also extend this pattern to avx512
14844 * config/i386/sse.md (k<code><mode>): New define_split after
14845 it to convert generic shift pattern to mask shift ones.
14847 2021-07-21 Thomas Schwinge <thomas@codesourcery.com>
14848 Joseph Myers <joseph@codesourcery.com>
14849 Cesar Philippidis <cesar@codesourcery.com>
14851 * tree-core.h (omp_clause_code): Add 'OMP_CLAUSE_NOHOST'.
14852 * tree.c (omp_clause_num_ops, omp_clause_code_name, walk_tree_1):
14854 * tree-pretty-print.c (dump_omp_clause): Likewise.
14855 * omp-general.c (oacc_verify_routine_clauses): Likewise.
14856 * gimplify.c (gimplify_scan_omp_clauses)
14857 (gimplify_adjust_omp_clauses): Likewise.
14858 * tree-nested.c (convert_nonlocal_omp_clauses)
14859 (convert_local_omp_clauses): Likewise.
14860 * omp-low.c (scan_sharing_clauses): Likewise.
14861 * omp-offload.c (execute_oacc_device_lower): Update.
14863 2021-07-21 Martin Sebor <msebor@redhat.com>
14865 * tree-ssa-alias.c (walk_aliased_vdefs_1): Fix typos in a comment.
14867 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
14869 * config/rs6000/rs6000-gen-builtins.c (write_init_bif_table):
14872 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
14874 * config/rs6000/rs6000-gen-builtins.c (write_fntype): New
14876 (write_fntype_init): New stub function.
14877 (write_init_bif_table): Likewise.
14878 (write_init_ovld_table): New function.
14879 (write_init_file): Implement.
14881 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
14883 * config/rs6000/rs6000-gen-builtins.c
14884 (write_autogenerated_header): New function.
14885 (write_decls): Likewise.
14886 (write_extern_fntype): New callback function.
14887 (write_header_file): Implement.
14889 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
14891 * config/rs6000/rs6000-gen-builtins.c (write_defines_file):
14894 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
14896 * config/rs6000/rs6000-gen-builtins.c (complete_vector_type): New
14898 (complete_base_type): Likewise.
14899 (construct_fntype_id): Likewise.
14900 (parse_bif_entry): Call contruct_fntype_id.
14901 (parse_ovld_entry): Likewise.
14903 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
14905 * config/rs6000/rs6000-gen-builtins.c (ovld_stanza): New struct.
14906 (MAXOVLDSTANZAS): New macro.
14907 (ovld_stanzas): New variable.
14908 (curr_ovld_stanza): Likewise.
14909 (MAXOVLDS): New macro.
14910 (ovlddata): New struct.
14911 (ovlds): New variable.
14912 (curr_ovld): Likewise.
14913 (max_ovld_args): Likewise.
14914 (parse_ovld_entry): New function.
14915 (parse_ovld_stanza): Likewise.
14916 (parse_ovld): Implement.
14918 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
14920 * config/rs6000/rs6000-gen-builtins.c (parse_bif_attrs):
14923 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
14925 * config/rs6000/rs6000-gen-builtins.c (parse_args): New function.
14926 (parse_prototype): Implement.
14928 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
14930 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): New enum.
14931 (curr_bif_stanza): New variable.
14932 (stanza_entry): New struct.
14933 (stanza_map): New initialized variable.
14934 (enable_string): Likewise.
14935 (fnkinds): New enum.
14936 (typelist): New struct.
14937 (attrinfo): Likewise.
14938 (MAXRESTROPNDS): New macro.
14939 (prototype): New struct.
14940 (MAXBIFS): New macro.
14941 (bifdata): New struct.
14942 (bifs): New variable.
14943 (curr_bif): Likewise.
14944 (bif_order): Likewise.
14945 (bif_index): Likewise.
14946 (fatal): New function.
14947 (stanza_name_to_stanza): Likewise.
14948 (parse_bif_attrs): New stub function.
14949 (parse_prototype): Likewise.
14950 (parse_bif_entry): New function.
14951 (parse_bif_stanza): Likewise.
14952 (parse_bif): Implement.
14953 (set_bif_order): New function.
14954 (create_bif_order): Implement.
14956 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
14958 * config/rs6000/rs6000-gen-builtins.c (rbtree.h): New #include.
14959 (num_bifs): New variable.
14960 (num_ovld_stanzas): Likewise.
14961 (num_ovlds): Likewise.
14962 (parse_codes): New enum.
14963 (bif_rbt): New variable.
14964 (ovld_rbt): Likewise.
14965 (fntype_rbt): Likewise.
14966 (bifo_rbt): Likewise.
14967 (parse_bif): New stub function.
14968 (create_bif_order): Likewise.
14969 (parse_ovld): Likewise.
14970 (write_header_file): Likewise.
14971 (write_init_file): Likewise.
14972 (write_defines_file): Likewise.
14973 (delete_output_files): New function.
14976 2021-07-21 H.J. Lu <hjl.tools@gmail.com>
14979 * config/i386/i386-builtin.def: Remove OPTION_MASK_ISA_SSE4_2
14980 from CRC32 _builtin functions.
14982 2021-07-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
14984 * coverage.c (build_gcov_info_var_registration): Mark the object placed
14985 in the linker set as referenced so that it does not get optimized away.
14987 2021-07-21 Kito Cheng <kito.cheng@sifive.com>
14990 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
14992 * config.gcc (riscv*-*-*): Detect which python is available.
14994 2021-07-21 Jakub Jelinek <jakub@redhat.com>
14996 PR middle-end/101535
14997 * gimplify.c (omp_check_private): Properly skip ORT_TARGET_DATA
14998 contexts in which decl isn't privatized and for ORT_TARGET return
14999 false if decl is mapped.
15001 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
15003 * gimple-loop-jam.c: Include tree-ssa-sccvn.h.
15004 (tree_loop_unroll_and_jam): Run value-numbering on a loop that
15005 has been successfully unrolled.
15007 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
15009 * tree-ssa-loop-manip.c (determine_exit_conditions): Return a null
15010 exit condition if no tail loop is needed, and if the original exit
15011 condition should therefore be kept as-is.
15012 (tree_transform_and_unroll_loop): Handle that case here too.
15014 2021-07-21 Kewen Lin <linkw@linux.ibm.com>
15016 * tree-data-ref.c (free_dependence_relations): Adjust to pass vec
15018 (free_data_refs): Likewise.
15019 * tree-data-ref.h (free_dependence_relations): Likewise.
15020 (free_data_refs): Likewise.
15021 * tree-predcom.c (struct chain): Use auto_vec instead of vec for
15023 (struct component): Likewise.
15024 (pcom_worker::pcom_worker): Adjust for auto_vec and renaming changes.
15025 (pcom_worker::~pcom_worker): Likewise.
15026 (pcom_worker::release_chain): Adjust as auto_vec changes.
15027 (pcom_worker::loop): Rename to ...
15028 (pcom_worker::m_loop): ... this.
15029 (pcom_worker::datarefs): Rename to ...
15030 (pcom_worker::m_datarefs): ... this. Use auto_vec instead of vec.
15031 (pcom_worker::dependences): Rename to ...
15032 (pcom_worker::m_dependences): ... this. Use auto_vec instead of vec.
15033 (pcom_worker::chains): Rename to ...
15034 (pcom_worker::m_chains): ... this. Use auto_vec instead of vec.
15035 (pcom_worker::looparound_phis): Rename to ...
15036 (pcom_worker::m_looparound_phis): ... this. Use auto_vec instead of
15038 (pcom_worker::cache): Rename to ...
15039 (pcom_worker::m_cache): ... this. Use auto_vec instead of vec.
15040 (pcom_worker::release_chain): Adjust for auto_vec changes.
15041 (pcom_worker::release_chains): Adjust for auto_vec and renaming
15043 (release_component): Remove.
15044 (release_components): Adjust for release_component removal.
15045 (component_of): Adjust to use vec.
15046 (merge_comps): Likewise.
15047 (pcom_worker::aff_combination_dr_offset): Adjust for renaming changes.
15048 (pcom_worker::determine_offset): Likewise.
15049 (class comp_ptrs): Remove.
15050 (pcom_worker::split_data_refs_to_components): Adjust for renaming
15051 changes, for comp_ptrs removal with auto_vec.
15052 (pcom_worker::suitable_component_p): Adjust for renaming changes.
15053 (pcom_worker::filter_suitable_components): Adjust for release_component
15055 (pcom_worker::valid_initializer_p): Adjust for renaming changes.
15056 (pcom_worker::find_looparound_phi): Likewise.
15057 (pcom_worker::add_looparound_copies): Likewise.
15058 (pcom_worker::determine_roots_comp): Likewise.
15059 (pcom_worker::single_nonlooparound_use): Likewise.
15060 (pcom_worker::execute_pred_commoning_chain): Likewise.
15061 (pcom_worker::execute_pred_commoning): Likewise.
15062 (pcom_worker::try_combine_chains): Likewise.
15063 (pcom_worker::prepare_initializers_chain): Likewise.
15064 (pcom_worker::prepare_initializers): Likewise.
15065 (pcom_worker::prepare_finalizers_chain): Likewise.
15066 (pcom_worker::prepare_finalizers): Likewise.
15067 (pcom_worker::tree_predictive_commoning_loop): Likewise.
15069 2021-07-20 Martin Sebor <msebor@redhat.com>
15071 PR middle-end/101397
15072 * builtins.c (gimple_call_return_array): Add argument. Correct
15073 offsets for memchr, mempcpy, stpcpy, and stpncpy.
15074 (compute_objsize_r): Adjust offset computation for argument returning
15077 2021-07-20 Martin Sebor <msebor@redhat.com>
15079 PR middle-end/101300
15080 * tree-ssa-uninit.c (check_defs): Handle UBSAN built-ins.
15082 2021-07-20 Jeff Law <jlaw@localhost.localdomain>
15084 * function.c (assign_parm_setup_block): Use adjust_address instead
15085 of change_address to preserve MEM_EXPR and friends.
15087 2021-07-20 Martin Sebor <msebor@redhat.com>
15089 * cfgloop.h (single_likely_exit): Adjust by-value argument to
15090 by-const-reference.
15091 * cfgloopanal.c (single_likely_exit): Same.
15092 * cgraph.h (struct cgraph_node): Same.
15093 * cgraphclones.c (cgraph_node::create_virtual_clone): Same.
15094 * genautomata.c (merge_states): Same.
15095 * genextract.c (VEC_char_to_string): Same.
15096 * genmatch.c (dt_node::gen_kids_1): Same.
15097 (walk_captures): Adjust by-value argument to by-reference.
15098 * gimple-ssa-store-merging.c (check_no_overlap): Adjust by-value argument
15099 to by-const-reference.
15100 * gimple.c (gimple_build_call_vec): Same.
15101 (gimple_build_call_internal_vec): Same.
15102 (gimple_build_switch): Same.
15103 (sort_case_labels): Same.
15104 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
15106 * gimple.h (gimple_build_call_vec): Adjust by-value argument to
15107 by-const-reference.
15108 (gimple_build_call_internal_vec): Same.
15109 (gimple_build_switch): Same.
15110 (sort_case_labels): Same.
15111 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
15113 * haifa-sched.c (calc_priorities): Adjust by-value argument to
15114 by-const-reference.
15115 (sched_init_luids): Same.
15116 (haifa_init_h_i_d): Same.
15117 * ipa-cp.c (ipa_get_indirect_edge_target_1): Same.
15118 (adjust_callers_for_value_intersection): Adjust by-value argument to
15120 (find_more_scalar_values_for_callers_subset): Adjust by-value argument to
15121 by-const-reference.
15122 (find_more_contexts_for_caller_subset): Same.
15123 (find_aggregate_values_for_callers_subset): Same.
15124 (copy_useful_known_contexts): Same.
15125 * ipa-fnsummary.c (remap_edge_summaries): Same.
15126 (remap_freqcounting_predicate): Same.
15127 * ipa-inline.c (add_new_edges_to_heap): Adjust by-value argument to
15129 * ipa-predicate.c (predicate::remap_after_inlining): Adjust by-value argument
15130 to by-const-reference.
15131 * ipa-predicate.h (predicate::remap_after_inlining): Same.
15132 * ipa-prop.c (ipa_find_agg_cst_for_param): Same.
15133 * ipa-prop.h (ipa_find_agg_cst_for_param): Same.
15134 * ira-build.c (ira_loop_tree_body_rev_postorder): Same.
15135 * read-rtl.c (add_overload_instance): Same.
15136 * rtl.h (native_decode_rtx): Same.
15137 (native_decode_vector_rtx): Same.
15138 * sched-int.h (sched_init_luids): Same.
15139 (haifa_init_h_i_d): Same.
15140 * simplify-rtx.c (native_decode_vector_rtx): Same.
15141 (native_decode_rtx): Same.
15142 * tree-call-cdce.c (gen_shrink_wrap_conditions): Same.
15143 (shrink_wrap_one_built_in_call_with_conds): Same.
15144 (shrink_wrap_conditional_dead_built_in_calls): Same.
15145 * tree-data-ref.c (create_runtime_alias_checks): Same.
15146 (compute_all_dependences): Same.
15147 * tree-data-ref.h (compute_all_dependences): Same.
15148 (create_runtime_alias_checks): Same.
15149 (index_in_loop_nest): Same.
15150 * tree-if-conv.c (mask_exists): Same.
15151 * tree-loop-distribution.c (class loop_distribution): Same.
15152 (loop_distribution::create_rdg_vertices): Same.
15153 (dump_rdg_partitions): Same.
15154 (debug_rdg_partitions): Same.
15155 (partition_contains_all_rw): Same.
15156 (loop_distribution::distribute_loop): Same.
15157 * tree-parloops.c (oacc_entry_exit_ok_1): Same.
15158 (oacc_entry_exit_single_gang): Same.
15159 * tree-ssa-loop-im.c (hoist_memory_references): Same.
15160 (loop_suitable_for_sm): Same.
15161 * tree-ssa-loop-niter.c (bound_index): Same.
15162 * tree-ssa-reassoc.c (update_ops): Same.
15163 (swap_ops_for_binary_stmt): Same.
15164 (rewrite_expr_tree): Same.
15165 (rewrite_expr_tree_parallel): Same.
15166 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Same.
15167 * tree-ssa-sccvn.h (ao_ref_init_from_vn_reference): Same.
15168 * tree-ssa-structalias.c (process_all_all_constraints): Same.
15169 (make_constraints_to): Same.
15170 (handle_lhs_call): Same.
15171 (find_func_aliases_for_builtin_call): Same.
15172 (sort_fieldstack): Same.
15173 (check_for_overlaps): Same.
15174 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Same.
15175 (vect_create_cond_for_unequal_addrs): Same.
15176 (vect_create_cond_for_lower_bounds): Same.
15177 (vect_create_cond_for_alias_checks): Same.
15178 * tree-vect-slp-patterns.c (vect_validate_multiplication): Same.
15179 * tree-vect-slp.c (vect_analyze_slp_instance): Same.
15180 (vect_make_slp_decision): Same.
15181 (vect_slp_bbs): Same.
15182 (duplicate_and_interleave): Same.
15183 (vect_transform_slp_perm_load): Same.
15184 (vect_schedule_slp): Same.
15185 * tree-vectorizer.h (vect_transform_slp_perm_load): Same.
15186 (vect_schedule_slp): Same.
15187 (duplicate_and_interleave): Same.
15188 * tree.c (build_vector_from_ctor): Same.
15189 (build_vector): Same.
15190 (check_vector_cst): Same.
15191 (check_vector_cst_duplicate): Same.
15192 (check_vector_cst_fill): Same.
15193 (check_vector_cst_stepped): Same.
15194 * tree.h (build_vector_from_ctor): Same.
15196 2021-07-20 Jakub Jelinek <jakub@redhat.com>
15199 * config/rs6000/rs6000-protos.h (easy_altivec_constant): Change return
15200 type from bool to int.
15201 * config/rs6000/rs6000.c (vspltis_constant): Fix up handling the
15202 EASY_VECTOR_MSB case if either step or copies is not 1.
15203 (vspltis_shifted): Fix comment typo.
15204 (easy_altivec_constant): Change return type from bool to int, instead
15205 of returning true return byte size of the element mode that should be
15206 used to synthetize the constant.
15207 * config/rs6000/predicates.md (easy_vector_constant_msb): Require
15208 that vspltis_shifted is 0, handle the case where easy_altivec_constant
15209 assumes using different vector mode from CONST_VECTOR's mode.
15210 * config/rs6000/altivec.md (easy_vector_constant_msb splitter): Use
15211 easy_altivec_constant to determine mode in which -1 >> -1 should be
15212 performed, use rs6000_expand_vector_init instead of gen_vec_initv4sisi.
15214 2021-07-20 Richard Biener <rguenther@suse.de>
15217 * dwarf2out.h (dwarf_file_data): Add key member.
15218 * dwarf2out.c (dwarf_file_hasher::equal): Compare key.
15219 (dwarf_file_hasher::hash): Hash key.
15220 (lookup_filename): Remap the filename and store it in the
15221 filename member of dwarf_file_data when creating a new
15223 (file_name_acquire): Do not remap the filename again.
15224 (maybe_emit_file): Likewise.
15226 2021-07-20 Jonathan Wright <jonathan.wright@arm.com>
15228 * config/aarch64/aarch64-simd-builtins.def: Use two variant
15229 generators for all TBL/TBX intrinsics and rename to
15230 consistent forms: qtbl[1234] or qtbx[1234].
15231 * config/aarch64/aarch64-simd.md (aarch64_tbl1<mode>):
15233 (aarch64_qtbl1<mode>): This.
15234 (aarch64_tbx1<mode>): Rename to...
15235 (aarch64_qtbx1<mode>): This.
15236 (aarch64_tbl2v16qi): Delete.
15237 (aarch64_tbl3<mode>): Rename to...
15238 (aarch64_qtbl2<mode>): This.
15239 (aarch64_tbx4<mode>): Rename to...
15240 (aarch64_qtbx2<mode>): This.
15241 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_1): Use
15242 renamed qtbl1 and qtbl2 RTL patterns.
15243 * config/aarch64/arm_neon.h (vqtbl1_p8): Use renamed qtbl1
15245 (vqtbl1_s8): Likewise.
15246 (vqtbl1_u8): Likewise.
15247 (vqtbl1q_p8): Likewise.
15248 (vqtbl1q_s8): Likewise.
15249 (vqtbl1q_u8): Likewise.
15250 (vqtbx1_s8): Use renamed qtbx1 RTL pattern.
15251 (vqtbx1_u8): Likewise.
15252 (vqtbx1_p8): Likewise.
15253 (vqtbx1q_s8): Likewise.
15254 (vqtbx1q_u8): Likewise.
15255 (vqtbx1q_p8): Likewise.
15256 (vtbl1_s8): Use renamed qtbl1 RTL pattern.
15257 (vtbl1_u8): Likewise.
15258 (vtbl1_p8): Likewise.
15259 (vtbl2_s8): Likewise
15260 (vtbl2_u8): Likewise.
15261 (vtbl2_p8): Likewise.
15262 (vtbl3_s8): Use renamed qtbl2 RTL pattern.
15263 (vtbl3_u8): Likewise.
15264 (vtbl3_p8): Likewise.
15265 (vtbl4_s8): Likewise.
15266 (vtbl4_u8): Likewise.
15267 (vtbl4_p8): Likewise.
15268 (vtbx2_s8): Use renamed qtbx2 RTL pattern.
15269 (vtbx2_u8): Likewise.
15270 (vtbx2_p8): Likewise.
15271 (vqtbl2_s8): Use renamed qtbl2 RTL pattern.
15272 (vqtbl2_u8): Likewise.
15273 (vqtbl2_p8): Likewise.
15274 (vqtbl2q_s8): Likewise.
15275 (vqtbl2q_u8): Likewise.
15276 (vqtbl2q_p8): Likewise.
15277 (vqtbx2_s8): Use renamed qtbx2 RTL pattern.
15278 (vqtbx2_u8): Likewise.
15279 (vqtbx2_p8): Likewise.
15280 (vqtbx2q_s8): Likewise.
15281 (vqtbx2q_u8): Likewise.
15282 (vqtbx2q_p8): Likewise.
15283 (vtbx4_s8): Likewise.
15284 (vtbx4_u8): Likewise.
15285 (vtbx4_p8): Likewise.
15287 2021-07-20 Uroš Bizjak <ubizjak@gmail.com>
15290 * config/i386/sync.md (define_peephole2 atomic_storedi_fpu):
15292 (define_peephole2 atomic_loaddi_fpu): Ditto.
15294 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
15296 * config.gcc (riscv*-*-*): Detect which python is available.
15298 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
15300 * config/rs6000/vsx.md (mulhs_<mode>): Rename to...
15301 (smul<mode>3_highpart): ... this.
15302 (mulhu_<mode>): Rename to...
15303 (umul<mode>3_highpart): ... this.
15304 * config/rs6000/rs6000-builtin.def (MULHS_V2DI, MULHS_V4SI,
15305 MULHU_V2DI, MULHU_V4SI): Adjust.
15307 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
15309 PR tree-optimization/100696
15310 * internal-fn.c (first_commutative_argument): Add info for IFN_MULH.
15311 * internal-fn.def (IFN_MULH): New internal function.
15312 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Add support to
15313 recog normal multiply highpart as IFN_MULH.
15314 * config/i386/i386.c (ix86_add_stmt_cost): Adjust for combined
15317 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
15319 * config/elfos.h (CTF_DEBUGGING_INFO): New definition.
15320 (BTF_DEBUGGING_INFO): Likewise.
15321 * doc/tm.texi.in: Document the new macros.
15322 * doc/tm.texi: Regenerated.
15323 * toplev.c: Guard initialization of debug hooks.
15325 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
15327 * flags.h (ctf_debuginfo_p): New function declaration.
15328 * opts.c (ctf_debuginfo_p): New function definition.
15330 2021-07-19 Andrew Stubbs <ams@codesourcery.com>
15333 * config/gcn/gcn-hsa.h (DRIVER_SELF_SPECS): New.
15334 (ASM_SPEC): Set -mattr for xnack and sram-ecc.
15335 * config/gcn/gcn-opts.h (enum sram_ecc_type): New.
15336 * config/gcn/gcn-valu.md: Add a warning comment.
15337 * config/gcn/gcn.c (gcn_option_override): Add "sorry" for -mxnack.
15338 (output_file_start): Add xnack and sram-ecc state to ".amdgcn_target".
15339 * config/gcn/gcn.md: Add a warning comment.
15340 * config/gcn/gcn.opt: Add -mxnack and -msram-ecc.
15341 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): Remove
15343 (EF_AMDGPU_XNACK): New.
15344 (EF_AMDGPU_SRAM_ECC): New.
15346 (copy_early_debug_info): Use elf_flags.
15347 (main): Handle -mxnack and -msram-ecc options.
15348 * doc/invoke.texi: Document -mxnack and -msram-ecc.
15350 2021-07-19 Andrew Pinski <apinski@marvell.com>
15353 * config/aarch64/aarch64.md (csneg3_uxtw_insn): Rename to ...
15354 (*cs<neg_not_cs>3_uxtw_insn4): and extend to NEG_NOT.
15356 2021-07-19 Richard Biener <rguenther@suse.de>
15358 PR tree-optimization/101505
15359 * tree-vect-patterns.c (vect_determine_precisions): Walk
15360 PHIs also for loop vectorization.
15362 2021-07-19 Richard Biener <rguenther@suse.de>
15364 * gimple.h (gimple_expr_type): Remove.
15365 * doc/gimple.texi: Remove gimple_expr_type documentation.
15367 2021-07-19 Richard Biener <rguenther@suse.de>
15369 * tree-ssa-sccvn.c (vn_reference_eq): Handle NULL vr->type.
15370 (ao_ref_init_from_vn_reference): Likewise.
15371 (fully_constant_reference): Likewise.
15372 (vn_reference_lookup_call): Do not set vr->type to random
15374 * tree-ssa-pre.c (compute_avail): Do not try to PRE calls
15376 * tree-vect-generic.c (expand_vector_piecewise): Pass in
15377 whether we expanded parallel.
15378 (expand_vector_parallel): Adjust.
15379 (expand_vector_addition): Likewise.
15380 (expand_vector_comparison): Likewise.
15381 (expand_vector_operation): Likewise.
15382 (expand_vector_scalar_condition): Likewise.
15383 (expand_vector_conversion): Likewise.
15385 2021-07-19 Richard Biener <rguenther@suse.de>
15387 * tree-vrp.c (register_edge_assert_for_2): Use the
15389 (vrp_folder::fold_predicate_in): Likewise.
15390 * vr-values.c (gimple_assign_nonzero_p): Likewise.
15391 (vr_values::extract_range_from_comparison): Likewise.
15392 (vr_values::extract_range_from_ubsan_builtin): Use the
15393 type of the first operand.
15394 (vr_values::extract_range_basic): Push down type
15395 computation, use the appropriate LHS.
15396 (vr_values::extract_range_from_assignment): Use the
15399 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
15402 * common/config/i386/i386-common.c (ix86_handle_option): For
15403 -mgeneral-regs-only, enable the GPR only instructions which are
15404 enabled implicitly by SSE ISAs unless they have been disabled
15407 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
15410 * config/i386/i386.c (ix86_check_avx_upper_stores): Moved before
15411 ix86_avx_u128_mode_needed.
15412 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY if callee
15413 returns AVX register.
15415 2021-07-17 Jan Hubicka <hubicka@ucw.cz>
15417 * tree-ssa-structalias.c (handle_rhs_call): Support EAF_NOT_RETURNED.
15418 (handle_const_call): Liekise
15419 (handle_pure_call): Liekise
15421 2021-07-17 Andrew MacLeod <amacleod@redhat.com>
15423 PR tree-optimization/96542
15424 * range-op.cc (range_operator::wi_fold_in_parts): New.
15425 (range_operator::fold_range): Call wi_fold_in_parts.
15426 (operator_lshift::wi_fold): Fix broken lshift by [0,0].
15427 * range-op.h (wi_fold_in_parts): Add prototype.
15429 2021-07-16 David Malcolm <dmalcolm@redhat.com>
15431 * doc/analyzer.texi: Add __analyzer_dump_state.
15433 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
15435 * config/rs6000/rbtree.c: New file.
15436 * config/rs6000/rbtree.h: New file.
15438 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
15440 * config/rs6000/rs6000-gen-builtins.c (restriction): New enum.
15441 (typeinfo): Add restr field.
15442 (match_bracketed_pair): New function.
15443 (match_const_restriction): Implement.
15445 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
15447 * config/rs6000/rs6000-gen-builtins.c (match_basetype): Implement.
15449 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
15451 * config/rs6000/rs6000-gen-builtins.c (void_status): New enum.
15452 (basetype): Likewise.
15453 (typeinfo): Likewise.
15454 (handle_pointer): New function.
15455 (match_basetype): New stub function.
15456 (match_const_restriction): Likewise.
15457 (match_type): New function.
15459 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
15461 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace): New
15463 (advance_line): Likewise.
15464 (safe_inc_pos): Likewise.
15465 (match_identifier): Likewise.
15466 (match_integer): Likewise.
15467 (match_to_right_bracket): Likewise.
15469 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
15471 * config/rs6000/rs6000-gen-builtins.c (bif_file): New variable.
15472 (ovld_file): Likewise.
15473 (header_file): Likewise.
15474 (init_file): Likewise.
15475 (defines_file): Likewise.
15476 (pgm_path): Likewise.
15477 (bif_path): Likewise.
15478 (ovld_path): Likewise.
15479 (header_path): Likewise.
15480 (init_path): Likewise.
15481 (defines_path): Likewise.
15482 (LINELEN): New macro.
15483 (linebuf): New variable.
15487 (bif_diag): New function.
15488 (ovld_diag): Likewise.
15490 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
15492 * config/rs6000/rs6000-builtin-new.def: New.
15493 * config/rs6000/rs6000-overload.def: New.
15495 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
15497 * config/rs6000/rs6000-gen-builtins.c: New.
15499 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
15501 * Makefile.in (EXTRA_GTYPE_DEPS): New variable.
15502 (s-gtype): Depend on EXTRA_GTYPE_DEPS.
15503 * gengtype-state.c (state_writer::write_state_file_list): Add a
15504 parameter to the fileslist expression for the number of build
15506 (read_state_files_list): Detect build headers and strip the
15507 initial "./" or ".\" from their names.
15508 * gengtype.c (build_headers): New global variable.
15509 (num_build_headers): Likewise.
15510 (open_base_files): Emit #include for each build header.
15511 (main): Detect and count build headers.
15512 * gengtype.h (build_headers): New extern variable.
15513 (num_build_headers): Likewise.
15515 2021-07-16 Richard Biener <rguenther@suse.de>
15517 * gimple-ssa-store-merging.c (verify_symbolic_number_p): Use
15518 the type of the LHS.
15519 (find_bswap_or_nop_1): Likewise.
15520 (find_bswap_or_nop): Likewise.
15521 * tree-vectorizer.h (vect_get_smallest_scalar_type): Adjust
15523 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
15524 Remove unused parameters, pass in the scalar type. Fix
15525 internal store function handling.
15526 * tree-vect-stmts.c (vect_analyze_stmt): Remove assert.
15527 (vect_get_vector_types_for_stmt): Move down check for
15528 existing vector stmt after we've determined a scalar type.
15529 Pass down the used scalar type to vect_get_smallest_scalar_type.
15530 * tree-vect-generic.c (expand_vector_condition): Use
15531 the type of the LHS.
15532 (expand_vector_scalar_condition): Likewise.
15533 (expand_vector_operations_1): Likewise.
15534 * tree-vect-patterns.c (vect_widened_op_tree): Likewise.
15535 (vect_recog_dot_prod_pattern): Likewise.
15536 (vect_recog_sad_pattern): Likewise.
15537 (vect_recog_widen_op_pattern): Likewise.
15538 (vect_recog_widen_sum_pattern): Likewise.
15539 (vect_recog_mixed_size_cond_pattern): Likewise.
15541 2021-07-16 Jan Hubicka <hubicka@ucw.cz>
15543 * ipa-modref.c (struct escape_entry): Use eaf_fleags_t.
15544 (dump_eaf_flags): Dump EAF_NOT_RETURNED
15545 (eaf_flags_useful_p): Use eaf_fleags_t; handle const functions
15546 and EAF_NOT_RETURNED.
15547 (modref_summary::useful_p): Likewise.
15548 (modref_summary_lto::useful_p): Likewise.
15549 (struct) modref_summary_lto: Use eaf_fleags_t.
15550 (deref_flags): Handle EAF_NOT_RETURNED.
15551 (struct escape_point): Use min_flags.
15552 (modref_lattice::init): Add EAF_NOT_RETURNED.
15553 (merge_call_lhs_flags): Ignore EAF_NOT_RETURNED functions
15554 (analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
15556 (analyze_parms): Also analyze const functions; update conition on
15558 (modref_write): Update streaming.
15559 (read_section): Update streaming.
15560 (remap_arg_flags): Use eaf_flags_t.
15561 (modref_merge_call_site_flags): Hanlde EAF_NOT_RETURNED.
15562 * ipa-modref.h: (eaf_flags_t): New typedef.
15563 (struct modref_summary): Use eaf_flags_t.
15564 * tree-core.h (EAF_NOT_RETURNED): New constant.
15566 2021-07-16 Richard Biener <rguenther@suse.de>
15568 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Use
15569 the type of the LHS.
15570 (gimple_assign_nonnegative_warnv_p): Likewise.
15571 (gimple_call_nonnegative_warnv_p): Likewise. Return false
15572 if the call has no LHS.
15573 * gimple.c (gimple_could_trap_p_1): Use the type of the LHS.
15574 * tree-eh.c (stmt_could_throw_1_p): Likewise.
15575 * tree-inline.c (insert_init_stmt): Likewise.
15576 * tree-ssa-loop-niter.c (get_val_for): Likewise.
15577 * tree-outof-ssa.c (ssa_is_replaceable_p): Use the type of
15579 * tree-ssa-sccvn.c (init_vn_nary_op_from_stmt): Take a
15580 gassign *. Use the type of the lhs.
15581 (vn_nary_op_lookup_stmt): Adjust.
15582 (vn_nary_op_insert_stmt): Likewise.
15584 2021-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
15586 * config/s390/predicates.md (bras_sym_operand): Accept all
15587 functions in 64-bit mode, use UNSPEC_PLT31.
15588 (larl_operand): Use UNSPEC_PLT31.
15589 * config/s390/s390.c (s390_loadrelative_operand_p): Likewise.
15590 (legitimize_pic_address): Likewise.
15591 (s390_emit_tls_call_insn): Mark __tls_get_offset as function,
15593 (s390_delegitimize_address): Use UNSPEC_PLT31.
15594 (s390_output_addr_const_extra): Likewise.
15595 (print_operand): Add @PLT to TLS calls, handle %K.
15596 (s390_function_profiler): Mark __fentry__/_mcount as function,
15597 use %K, use UNSPEC_PLT31.
15598 (s390_output_mi_thunk): Use only UNSPEC_GOT, use %K.
15599 (s390_emit_call): Use UNSPEC_PLT31.
15600 (s390_emit_tpf_eh_return): Mark __tpf_eh_return as function.
15601 * config/s390/s390.md (UNSPEC_PLT31): Rename from UNSPEC_PLT.
15602 (*movdi_64): Use %K.
15603 (reload_base_64): Likewise.
15604 (*sibcall_brc): Likewise.
15605 (*sibcall_brcl): Likewise.
15606 (*sibcall_value_brc): Likewise.
15607 (*sibcall_value_brcl): Likewise.
15609 (*brasl): Likewise.
15610 (*bras_r): Likewise.
15611 (*brasl_r): Likewise.
15612 (*bras_tls): Likewise.
15613 (*brasl_tls): Likewise.
15614 (main_base_64): Likewise.
15615 (reload_base_64): Likewise.
15616 (@split_stack_call<mode>): Likewise.
15618 2021-07-16 Richard Biener <rguenther@suse.de>
15620 PR tree-optimization/101467
15621 * tree-vect-stmts.c (vect_gen_while): Properly guard
15622 make_temp_ssa_name usage.
15624 2021-07-16 Cooper Qu <cooper.qu@linux.alibaba.com>
15626 * config.gcc: Don't use forked print-sysroot-suffix.sh and
15627 t-sysroot-suffix for C-SKY.
15628 * config/csky/print-sysroot-suffix.sh: Delete.
15629 * config/csky/t-csky-linux: Delete.
15630 * config/csky/t-sysroot-suffix: Define MULTILIB_DIRNAMES
15631 instead of CSKY_MULTILIB_DIRNAMES.
15633 2021-07-16 Richard Biener <rguenther@suse.de>
15635 * tree-vect-loop.c (vect_transform_cycle_phi): Correct sign
15636 conversion issues with the partial reduction of the reused
15637 vector accumulator.
15639 2021-07-16 Richard Biener <rguenther@suse.de>
15641 * config/i386/i386-options.c (ix86_option_override_internal): Set
15642 param_vect_partial_vector_usage to zero if not set.
15644 2021-07-15 Uroš Bizjak <ubizjak@gmail.com>
15647 * config/i386/i386.h (VALID_SSE_REG_MODE): Add TDmode.
15648 (VALID_INT_MODE_P): Add SDmode and DDmode.
15649 Add TDmode for TARGET_64BIT.
15650 (VALID_DFP_MODE_P): Remove.
15651 * config/i386/i386.c (ix86_hard_regno_mode_ok):
15652 Do not use VALID_DFP_MODE_P.
15654 2021-07-15 Andrew MacLeod <amacleod@redhat.com>
15656 * gimple-range-fold.cc (adjust_pointer_diff_expr): Use
15658 (fold_using_range::fold_stmt): Ditto.
15659 (fold_using_range::range_of_range_op): Ditto.
15660 (fold_using_range::range_of_phi): Ditto.
15661 (fold_using_range::range_of_call): Ditto.
15662 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
15663 (fold_using_range::range_of_builtin_call): Ditto.
15664 (fold_using_range::range_of_cond_expr): Ditto.
15665 * gimple-range-fold.h (gimple_range_type): New.
15667 2021-07-15 Martin Sebor <msebor@redhat.com>
15669 PR middle-end/97027
15670 * tree-ssa-strlen.c (handle_assign): New function.
15671 (maybe_warn_overflow): Add argument.
15672 (nonzero_bytes_for_type): New function.
15673 (count_nonzero_bytes): Handle more tree types. Call
15674 nonzero_bytes_for_tye.
15675 (count_nonzero_bytes): Handle types.
15676 (handle_store): Handle stores from function calls.
15677 (strlen_check_and_optimize_call): Move code to handle_assign. Call
15678 it for assignments from function calls.
15680 2021-07-15 David Malcolm <dmalcolm@redhat.com>
15685 * doc/invoke.texi: Add -Wanalyzer-use-of-uninitialized-value.
15687 2021-07-15 David Malcolm <dmalcolm@redhat.com>
15689 * doc/invoke.texi (-fdump-analyzer-exploded-paths): New.
15691 2021-07-15 Martin Sebor <msebor@redhat.com>
15695 * fold-const.c (operand_compare::operand_equal_p): Handle OEP_DECL_NAME.
15696 (operand_compare::verify_hash_value): Same.
15697 * tree-core.h (OEP_DECL_NAME): New.
15699 2021-07-15 Martin Jambor <mjambor@suse.cz>
15701 * profile-count.h (profile_count::value): Change the return type to
15703 * gimple-pretty-print.c (dump_gimple_bb_header): Adjust print
15705 * tree-cfg.c (dump_function_to_file): Likewise.
15707 2021-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
15710 * config/rs6000/rs6000-p8swap.c (has_part_mult): New.
15711 (rs6000_analyze_swaps): Insns containing a subreg of a mult are
15714 2021-07-15 Richard Biener <rguenther@suse.de>
15716 * tree-vectorizer.h (vect_gen_while): Match up with
15717 vect_gen_while_not.
15718 * tree-vect-stmts.c (vect_gen_while): Adjust API to that
15719 of vect_gen_while_not.
15720 (vect_gen_while_not): Adjust.
15721 * tree-vect-loop-manip.c (vect_set_loop_controls_directly): Likewise.
15723 2021-07-15 Aldy Hernandez <aldyh@redhat.com>
15725 * gimple-range-cache.cc (non_null_ref::adjust_range): New.
15726 (ranger_cache::range_of_def): Call adjust_range.
15727 (ranger_cache::entry_range): Same.
15728 * gimple-range-cache.h (non_null_ref::adjust_range): New.
15729 * gimple-range.cc (gimple_ranger::range_of_expr): Call
15731 (gimple_ranger::range_on_entry): Same.
15733 2021-07-15 Tamar Christina <tamar.christina@arm.com>
15736 2021-07-14 Tamar Christina <tamar.christina@arm.com>
15738 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
15740 2021-07-15 Tamar Christina <tamar.christina@arm.com>
15743 2021-07-14 Tamar Christina <tamar.christina@arm.com>
15745 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
15746 (sdot_prod, udot_prod): ...These.
15747 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
15748 (aarch64_<sur>dot<vsi2qi>): Rename to...
15749 (<sur>dot_prod<vsi2qi>): ...This.
15750 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
15753 2021-07-15 Jakub Jelinek <jakub@redhat.com>
15755 PR middle-end/101437
15756 * gimplify.c (gimplify_expr): Throw away volatile reads from empty
15757 types even if they have non-BLKmode TYPE_MODE.
15759 2021-07-15 Richard Biener <rguenther@suse.de>
15762 * gcc.c (process_command): Process -gtoggle like process_options
15763 would after parsing options.
15765 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
15767 * cfgexpand.c (expand_asm_loc): Adjust.
15768 (expand_asm_stmt): Likewise.
15769 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Likewise.
15770 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
15771 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
15772 * config/avr/avr.c (avr_md_asm_adjust): Likewise.
15773 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
15774 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
15775 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
15776 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
15777 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
15778 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
15779 * config/s390/s390.c (s390_md_asm_adjust): Likewise.
15780 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
15781 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
15782 * doc/tm.texi: Regenerate.
15783 * target.def: Add location argument to md_asm_adjust.
15785 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
15787 * tree-diagnostic.c (diagnostic_report_current_function): Use the
15788 diagnostic's location, not input_location.
15790 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
15792 * cfgexpand.c (tree_conflicts_with_clobbers_p): Pass location to
15794 (expand_asm_stmt): Likewise.
15796 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
15798 * config/rs6000/rs6000.c (adjacent_mem_locations): Return the lower
15799 addressed memory rtx, if any.
15800 (rs6000_split_multireg_move): Fix code formatting.
15801 Handle MMA build built-ins with operands in adjacent memory locations.
15803 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
15805 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Move to later
15808 2021-07-14 Jason Merrill <jason@redhat.com>
15810 * sel-sched-ir.h (get_all_loop_exits): Use auto_vec.
15812 2021-07-14 Jason Merrill <jason@redhat.com>
15814 * doc/invoke.texi: -fdelete-dead-exceptions is on by default for
15817 2021-07-14 Tamar Christina <tamar.christina@arm.com>
15819 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
15820 Remove erroneous line.
15822 2021-07-14 Andrew MacLeod <amacleod@redhat.com>
15824 * params.opt (param_evrp_mode): Change default.
15826 2021-07-14 Tamar Christina <tamar.christina@arm.com>
15828 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
15829 (sdot_prod, udot_prod): ...These.
15830 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
15831 (aarch64_<sur>dot<vsi2qi>): Rename to...
15832 (<sur>dot_prod<vsi2qi>): ...This.
15833 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
15836 2021-07-14 Tamar Christina <tamar.christina@arm.com>
15838 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
15840 2021-07-14 Tamar Christina <tamar.christina@arm.com>
15842 * doc/sourcebuild.texi (arm_v8_2a_i8mm_neon_hw): Document.
15844 2021-07-14 Tamar Christina <tamar.christina@arm.com>
15846 * config/arm/neon.md (usdot_prod<vsi2qi>): New.
15848 2021-07-14 Tamar Christina <tamar.christina@arm.com>
15850 * config/aarch64/aarch64-simd.md (aarch64_usdot<vsi2qi>): Rename to...
15851 (usdot_prod<vsi2qi>): ... This.
15852 * config/aarch64/aarch64-simd-builtins.def (usdot): Rename to...
15853 (usdot_prod): ...This.
15854 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Likewise.
15855 * config/aarch64/aarch64-sve.md (@aarch64_<sur>dot_prod<vsi2qi>):
15857 (@<sur>dot_prod<vsi2qi>): ...This.
15858 * config/aarch64/aarch64-sve-builtins-base.cc
15859 (svusdot_impl::expand): Use it.
15861 2021-07-14 Tamar Christina <tamar.christina@arm.com>
15863 * optabs.def (usdot_prod_optab): New.
15864 * doc/md.texi: Document it and clarify other dot prod optabs.
15865 * optabs-tree.h (enum optab_subtype): Add optab_vector_mixed_sign.
15866 * optabs-tree.c (optab_for_tree_code): Support usdot_prod_optab.
15867 * optabs.c (expand_widen_pattern_expr): Likewise.
15868 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
15869 * tree-vect-loop.c (vectorizable_reduction): Query dot-product kind.
15870 * tree-vect-patterns.c (vect_supportable_direct_optab_p): Take optional
15872 (vect_widened_op_tree): Optionally ignore
15874 (vect_recog_dot_prod_pattern): Support usdot_prod_optab.
15876 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
15879 * config/i386/driver-i386.c (host_detect_local_cpu): Check
15880 "arch [32|64]" and "tune [32|64]" for 32-bit and 64-bit codegen.
15881 Enable UINTR only for 64-bit codegen.
15882 * config/i386/i386-options.c
15883 (ix86_option_override_internal::DEF_PTA): Skip PTA_UINTR if not
15885 * config/i386/i386.h (ARCH_ARG): New.
15886 (CC1_CPU_SPEC): Pass "[arch|tune] 32" for 32-bit codegen and
15887 "[arch|tune] 64" for 64-bit codegen.
15889 2021-07-14 Richard Biener <rguenther@suse.de>
15891 PR tree-optimization/101445
15892 * tree-vect-stmts.c (vectorizable_load): Do the gap adjustment
15893 of the IV in the correct direction for negative stride
15896 2021-07-14 Jakub Jelinek <jakub@redhat.com>
15899 * godump.c (godump_str_hash): New type.
15900 (godump_container::pot_dummy_types): Use string_hash instead of
15901 ptr_hash in the hash_set.
15903 2021-07-14 Richard Biener <rguenther@suse.de>
15905 * tree-vect-loop.c (vect_find_reusable_accumulator): Handle
15906 vector types where the old vector type has a multiple of
15907 the new vector type elements.
15908 (vect_create_partial_epilog): New function, split out from...
15909 (vect_create_epilog_for_reduction): ... here.
15910 (vect_transform_cycle_phi): Reduce the re-used accumulator
15911 to the new vector type.
15913 2021-07-14 Alexandre Oliva <oliva@adacore.com>
15915 * tree-ssa-alias.c (attr_fnspec::verify): Fix index in
15916 non-'t'-sized arg check.
15918 2021-07-14 Alexandre Oliva <oliva@adacore.com>
15920 * tree-cfg.c (cleanup_dead_labels_eh): Update
15921 post_landing_pad label upon change of landing pad block's
15923 (cleanup_dead_labels): Check that a removed label is not that
15926 2021-07-13 Jonathan Wright <jonathan.wright@arm.com>
15928 * combine.c (combine_simplify_rtx): Add vec_select -> subreg
15930 * config/aarch64/aarch64.md (*zero_extend<SHORT:mode><GPI:mode>2_aarch64):
15931 Add Neon to general purpose register case for zero-extend
15933 * config/arm/vfp.md (*arm_movsi_vfp): Remove "*" from *t -> r
15934 case to prevent some cases opting to go through memory.
15935 * cse.c (fold_rtx): Add vec_select -> subreg simplification.
15936 * rtl.c (rtvec_series_p): Define predicate to determine
15937 whether a vector contains a linear series of integers.
15938 * rtl.h (rtvec_series_p): Define.
15939 * rtlanal.c (vec_series_lowpart_p): Define predicate to
15940 determine if a vector selection is equivalent to the low part
15942 * rtlanal.h (vec_series_lowpart_p): Define.
15943 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
15944 Add vec_select -> subreg simplification.
15946 2021-07-13 Paul A. Clarke <pc@us.ibm.com>
15948 * config/rs6000/smmintrin.h (_mm_testz_si128, _mm_testc_si128,
15949 _mm_testnzc_si128, _mm_test_all_ones, _mm_test_all_zeros,
15950 _mm_test_mix_ones_zeros): New.
15952 2021-07-13 Roger Sayle <roger@nextmovesoftware.com>
15953 Richard Biener <rguenther@suse.de>
15955 * gimple.c (gimple_could_trap_p_1): Make S argument a
15956 "const gimple*". Preserve constness in call to
15957 gimple_asm_volatile_p.
15958 (gimple_could_trap_p): Make S argument a "const gimple*".
15959 * gimple.h (gimple_could_trap_p_1, gimple_could_trap_p):
15960 Update function prototypes.
15962 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
15964 * tree-vectorizer.h (vect_reusable_accumulator): New structure.
15965 (_loop_vec_info::main_loop_edge): New field.
15966 (_loop_vec_info::skip_main_loop_edge): Likewise.
15967 (_loop_vec_info::skip_this_loop_edge): Likewise.
15968 (_loop_vec_info::reusable_accumulators): Likewise.
15969 (_stmt_vec_info::reduc_scalar_results): Likewise.
15970 (_stmt_vec_info::reused_accumulator): Likewise.
15971 (vect_get_main_loop_result): Declare.
15972 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize
15973 reduc_scalar_inputs.
15974 (vec_info::free_stmt_vec_info): Free reduc_scalar_inputs.
15975 * tree-vect-loop-manip.c (vect_get_main_loop_result): New function.
15976 (vect_do_peeling): Fill an epilogue loop's main_loop_edge,
15977 skip_main_loop_edge and skip_this_loop_edge fields.
15978 * tree-vect-loop.c (INCLUDE_ALGORITHM): Define.
15979 (vect_emit_reduction_init_stmts): New function.
15980 (get_initial_def_for_reduction): Use it.
15981 (get_initial_defs_for_reduction): Likewise. Change the vinfo
15982 parameter to a loop_vec_info.
15983 (vect_create_epilog_for_reduction): Store the scalar results
15984 in the reduc_info. If an epilogue loop is reusing an accumulator
15985 from the main loop, and if the epilogue loop can also be skipped,
15986 try to place the reduction code in the join block. Record
15987 accumulators that could potentially be reused by epilogue loops.
15988 (vect_transform_cycle_phi): When vectorizing epilogue loops,
15989 try to reuse accumulators from the main loop. Record the initial
15990 value in reduc_info for non-SLP reductions too.
15992 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
15994 * tree-vect-loop.c (get_initial_def_for_reduction): Remove
15995 adjustment handling. Take the neutral value as an argument,
15996 in place of the code argument.
15997 (vect_transform_cycle_phi): Update accordingly. Handle the
15998 initial values of cond reductions separately from code reductions.
15999 Choose the adjustment here rather than in
16000 get_initial_def_for_reduction. Sink the splat of vec_initial_def.
16002 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
16004 * tree-vect-loop.c (neutral_op_for_slp_reduction): Replace with...
16005 (neutral_op_for_reduction): ...this, providing a more general
16007 (vect_create_epilog_for_reduction): Update accordingly.
16008 (vectorizable_reduction): Likewise.
16009 (vect_transform_cycle_phi): Likewise.
16011 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
16013 * tree-vect-loop.c (get_initial_def_for_reduction): Take the
16014 reduc_info instead of the original stmt_vec_info.
16015 (vect_transform_cycle_phi): Update accordingly.
16017 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
16019 * tree-vect-loop.c (get_initial_defs_for_reduction): Take the
16020 reduc_info as an additional parameter.
16021 (vect_transform_cycle_phi): Update accordingly.
16023 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
16025 * tree-vectorizer.h: Include tree-ssa-operands.h.
16026 (vect_phi_initial_value): New function.
16027 * tree-vect-loop.c (neutral_op_for_slp_reduction): Use it.
16028 (get_initial_defs_for_reduction, info_for_reduction): Likewise.
16029 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
16030 (vect_transform_cycle_phi, vectorizable_induction): Likewise.
16032 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
16034 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
16035 the phi results to vectype after creating them. Remove later
16036 conversion code that thus becomes redundant.
16038 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
16040 * tree-vect-loop.c (vect_create_epilog_for_reduction): Replace
16041 the new_phis vector with a reduc_inputs vector. Combine handling
16042 of reduction chains and ncopies > 1.
16044 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
16046 * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
16047 scalar_results to group_size elements after reducing down from
16048 N*group_size elements. Construct an array_slice of the live-out
16049 stmts and assert that there is one stmt per scalar result.
16051 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
16053 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
16054 nested_in_vect_loop and use double_reduc everywhere. Remove dead
16055 assignment to "loop".
16057 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
16059 * internal-fn.c (vectorized_internal_fn_supported_p): Handle
16060 vector types first. For scalar types, consider both the preferred
16061 vector mode and the alternative vector modes.
16062 * optabs-query.c (can_vec_mask_load_store_p): Use the same
16063 structure as above, in particular using related_vector_mode
16064 for modes provided by autovectorize_vector_modes.
16066 2021-07-13 Jakub Jelinek <jakub@redhat.com>
16067 Richard Biener <rguenther@suse.de>
16069 PR tree-optimization/101419
16070 * tree-pass.h (PROP_objsz): Define.
16071 (make_pass_early_object_sizes): Declare.
16072 * passes.def (pass_all_early_optimizations): Rename pass_object_sizes
16073 there to pass_early_object_sizes, drop parameter.
16074 (pass_all_optimizations): Move pass_object_sizes right after pass_ccp,
16075 drop parameter, move pass_post_ipa_warn right after that.
16076 * tree-object-size.c (pass_object_sizes::execute): Rename to...
16077 (object_sizes_execute): ... this. Add insert_min_max_p argument.
16078 (pass_data_object_sizes): Move after object_sizes_execute.
16079 (pass_object_sizes): Likewise. In execute method call
16080 object_sizes_execute, drop set_pass_param method and insert_min_max_p
16081 non-static data member and its initializer in the ctor.
16082 (pass_data_early_object_sizes, pass_early_object_sizes,
16083 make_pass_early_object_sizes): New.
16084 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Use
16085 (cfun->curr_properties & PROP_objsz) instead of cfun->after_inlining.
16087 2021-07-13 Kito Cheng <kito.cheng@sifive.com>
16090 * config/riscv/constraints.md ("S"): Update description and remove
16092 * doc/md.texi (Machine Constraints): Document the 'S' constraints
16095 2021-07-13 Richard Biener <rguenther@suse.de>
16098 2021-07-12 Richard Biener <rguenther@suse.de>
16100 * tree-vect-slp.c (vect_slp_region): Show the number of
16101 SLP graph entries in the optimization message.
16103 2021-07-13 Michael Meissner <meissner@linux.ibm.com>
16105 * config/rs6000/altivec.md (xxspltiw_v4sf): Change local variable
16107 * config/rs6000/rs6000-protos.h (rs6000_const_f32_to_i32): Change
16108 return type to long.
16109 * config/rs6000/rs6000.c (rs6000_const_f32_to_i32): Change return
16112 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
16114 * gimple-range-fold.cc (fold_using_range::range_of_builtin_ubsan_call):
16115 Query relation between the 2 operands and use it.
16117 2021-07-12 Sergei Trofimovich <siarheit@google.com>
16119 * doc/cfg.texi: Fix s/ei_safe_safe/ei_safe_edge/ typo.
16121 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
16124 * config/i386/predicates.md (vec_setm_sse41_operand):
16125 Rename from vec_setm_operand.
16126 (vec_setm_avx2_operand): New predicate.
16127 * config/i386/sse.md (vec_set<V_128:mode>): Use V_128 mode iterator.
16128 Use vec_setm_sse41_operand as operand 2 predicate.
16129 (vec_set<V_256_512:mode): New expander.
16130 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_sse41_operand
16131 as operand 2 predicate.
16133 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
16135 PR tree-optimization/101335
16136 * range-op.cc (operator_cast::lhs_op1_relation): Delete.
16138 2021-07-12 Andrew Pinski <apinski@marvell.com>
16140 * tree-ssa-phiopt.c (match_simplify_replacement): Move
16141 insert of the sequence before the movement of the
16142 statement. Check if to see if the statement is used
16143 outside of the original phi to see if we should move it.
16145 2021-07-12 Richard Biener <rguenther@suse.de>
16147 * dump-context.h (debug_dump_context::debug_dump_context):
16148 Add FILE * parameter defaulted to stderr.
16149 * dumpfile.c (debug_dump_context::debug_dump_context): Adjust.
16150 * tree-vect-slp.c (dot_slp_tree): New functions.
16152 2021-07-12 Richard Biener <rguenther@suse.de>
16154 PR tree-optimization/101373
16155 * tree-ssa-pre.c (prune_clobbered_mems): Also prune trapping
16156 references when the BB may not return.
16157 (compute_avail): Pass in the function we're working on and
16158 replace cfun references with it. Externally throwing
16159 const calls also possibly terminate the function.
16160 (pass_pre::execute): Pass down the function we're working on.
16161 * gcse.c (compute_hash_table_work): Externally throwing
16162 const/pure calls also need record_last_mem_set_info.
16163 * postreload-gcse.c (record_opr_changes): Looping or externally
16164 throwing const/pure calls also need record_last_mem_set_info.
16166 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
16168 * recog.c (memory_address_addr_space_p): Change the type to bool.
16169 Return true/false instead of 1/0.
16170 (offsettable_memref_p): Ditto.
16171 (offsettable_nonstrict_memref_p): Ditto.
16172 (offsettable_address_addr_space_p): Ditto.
16173 Change the type of addressp indirect function to bool.
16174 * recog.h (memory_address_addr_space_p): Change the type to bool.
16175 (strict_memory_address_addr_space_p): Ditto.
16176 (offsettable_memref_p): Ditto.
16177 (offsettable_nonstrict_memref_p): Ditto.
16178 (offsettable_address_addr_space_p): Ditto.
16179 * reload.c (maybe_memory_address_addr_space_p): Ditto.
16180 (strict_memory_address_addr_space_p): Change the type to bool.
16181 Return true/false instead of 1/0.
16182 (maybe_memory_address_addr_space_p): Change the type to bool.
16184 2021-07-12 Richard Biener <rguenther@suse.de>
16186 * tree-vect-slp.c (vect_slp_region): Show the number of
16187 SLP graph entries in the optimization message.
16189 2021-07-12 Richard Biener <rguenther@suse.de>
16191 PR tree-optimization/101394
16192 * tree-ssa-pre.c (do_pre_regular_insertion): Avoid inserting
16193 copies from abnormals for a full redundancy.
16195 2021-07-12 Richard Biener <rguenther@suse.de>
16197 PR middle-end/101423
16198 * gimple.c (gimple_could_trap_p_1): Internal function calls
16200 * tree-eh.c (tree_could_trap_p): Likewise.
16202 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
16205 * config/arm/arm_neon.h (vmul_n_u32): Replace call to builtin with
16207 (vmulq_n_u32): Likewise.
16208 (vmul_n_f32): Gate __a * __b on __FAST_MATH__.
16209 (vmulq_n_f32): Likewise.
16210 (vmul_n_f16): Likewise.
16211 (vmulq_n_f16): Likewise.
16213 2021-07-12 Martin Liska <mliska@suse.cz>
16215 PR sanitizer/101425
16216 * gcc.c (check_offload_target_name): Call
16217 candidates_list_and_hint only if we have a candidate.
16219 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
16222 * config/arm/neon.md (vec_init): Move to ...
16223 * config/arm/vec-common.md (vec_init): ... here.
16224 Change the pattern's mode to VDQX and gate it on VALID_MVE_MODE.
16226 2021-07-12 Roger Sayle <roger@nextmovesoftware.com>
16228 PR tree-optimization/101403
16229 * match.pd ((T)bswap(X)>>C): Correctly handle cases where
16230 signedness of the shift is not the same as the signedness of
16231 the type extension.
16233 2021-07-09 Roger Sayle <roger@nextmovesoftware.com>
16234 Uroš Bizjak <ubizjak@gmail.com>
16236 * config/i386/i386.md (*divmodsi4_const): Optimize SImode
16237 divmod of a constant numerator with new define_insn_and_split.
16239 2021-07-09 Iain Sandoe <iain@sandoe.co.uk>
16242 * config/i386/i386-expand.c (ix86_expand_call): If a call is
16243 to a non-local-binding, or local but to a public symbol, then
16244 assume that it might be indirected via the lazy symbol binder.
16245 Mark R10 and R10 as clobbered in that case.
16247 2021-07-09 Eric Botcazou <ebotcazou@adacore.com>
16250 * gcc.c (ASM_DEBUG_DWARF_OPTION): Set again to --gdwarf2 in
16251 the case where HAVE_AS_WORKING_DWARF_N_FLAG is not defined
16252 and HAVE_LD_BROKEN_PE_DWARF5 is defined.
16254 2021-07-09 Uroš Bizjak <ubizjak@gmail.com>
16256 * config/i386/i386.md (*udivmodsi4_pow2_zext_1): Limit the
16257 log2 range of operands[3] to [1,31].
16258 (*udivmodsi4_pow2_zext_2): Ditto. Correct insn RTX pattern.
16260 2021-07-09 Sergei Trofimovich <siarheit@google.com>
16262 * doc/md.texi: Don't split @smallexample in multiple @groups.
16264 2021-07-09 Sergei Trofimovich <siarheit@google.com>
16266 * doc/md.texi: Add missing 'see' word.
16268 2021-07-09 Andrew Pinski <apinski@marvell.com>
16270 * tree-ssa-phiopt.c (phiopt_early_allow): Change arguments
16271 to take sequence and gimple_match_op. Accept the case where
16272 op is a SSA_NAME and one statement in the sequence.
16273 Also allow constants.
16274 (gimple_simplify_phiopt): Always pass a sequence to resimplify.
16275 Update call to phiopt_early_allow. Discard the sequence if not
16278 2021-07-09 Xi Ruoyao <xry111@mengyan1223.wang>
16283 * config/mips/mips.c (mips_const_insns): Use MSA_SUPPORTED_MODE_P
16284 instead of ISA_HAS_MSA.
16285 (mips_expand_vec_unpack): Likewise.
16286 (mips_expand_vector_init): Likewise.
16288 2021-07-09 Kewen Lin <linkw@linux.ibm.com>
16290 * config/rs6000/vsx.md (mods_<mode>): Rename to...
16291 (mod<mode>3): ... this.
16292 (modu_<mode>): Rename to...
16293 (umod<mode>3): ... this.
16294 * config/rs6000/rs6000-builtin.def (MODS_V2DI, MODS_V4SI, MODU_V2DI,
16295 MODU_V4SI): Adjust.
16297 2021-07-08 Jeff Law <jeffreyalaw@gmail.com>
16299 * config/h8300/shiftrotate.md (variable shifts): Expose condition
16300 code handling for the test before the loop.
16302 2021-07-08 Martin Jambor <mjambor@suse.cz>
16305 * ipa-sra.c (class isra_call_summary): New member
16306 m_before_any_store, initialize it in the constructor.
16307 (isra_call_summary::dump): Dump the new field.
16308 (ipa_sra_call_summaries::duplicate): Copy it.
16309 (process_scan_results): Set it.
16310 (isra_write_edge_summary): Stream it.
16311 (isra_read_edge_summary): Likewise.
16312 (param_splitting_across_edge): Only override
16313 safe_to_import_accesses if m_before_any_store is set.
16315 2021-07-08 Martin Sebor <msebor@redhat.com>
16317 PR bootstrap/101374
16318 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref):
16319 Use Object Size Type 0 instead of 1.
16321 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
16323 * tree-vect-loop.c (vectorizable_reduction): Remove always-true
16326 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
16328 * match.pd: Simplify an extend-operate-truncate sequence involving
16331 2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
16332 Richard Biener <rguenther@suse.de>
16334 PR tree-optimization/40210
16335 * match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
16336 (x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
16337 when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
16339 2021-07-08 Uroš Bizjak <ubizjak@gmail.com>
16342 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
16344 * config/i386/mmx.md (V_32): New mode iterator.
16345 (mov<V_32:mode>): Use V_32 mode iterator.
16346 (*mov<V_32:mode>_internal): Ditto.
16347 (*push<V_32:mode>2_rex64): Ditto.
16348 (*push<V_32:mode>2): Ditto.
16349 (movmisalign<V_32:mode>): Ditto.
16350 (mmx_<any_shiftrt:insn>v1si3): New insn pattern.
16351 (sse4_1_<any_extend:code>v2qiv2hi2): Ditto.
16352 (vec_unpacks_lo_v4qi): New expander.
16353 (vec_unpacks_hi_v4qi): Ditto.
16354 (vec_unpacku_lo_v4qi): Ditto.
16355 (vec_unpacku_hi_v4qi): Ditto.
16356 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add V1SImode.
16357 (VALID_INT_MODE_P): Ditto.
16359 2021-07-08 Michael Meissner <meissner@linux.ibm.com>
16362 * config/rs6000/rs6000.md (udivti3): New insn.
16363 (divti3): New insn.
16364 (umodti3): New insn.
16365 (modti3): New insn.
16367 2021-07-07 Martin Sebor <msebor@redhat.com>
16369 PR tree-optimization/100137
16370 PR tree-optimization/99121
16371 PR tree-optimization/97027
16372 * builtins.c (access_ref::access_ref): Also set offmax.
16373 (access_ref::offset_in_range): Define new function.
16374 (access_ref::add_offset): Set offmax.
16375 (access_ref::inform_access): Handle access_none.
16376 (handle_mem_ref): Clear ostype.
16377 (compute_objsize_r): Handle ASSERT_EXPR.
16378 * builtins.h (struct access_ref): Add offmax member.
16379 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Use
16380 compute_objsize() and simplify.
16382 2021-07-07 Peter Bergner <bergner@linux.ibm.com>
16384 * config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
16385 and VSX_BUILTIN_STXVP.
16387 2021-07-07 Martin Sebor <msebor@redhat.com>
16390 * config/aarch64/aarch64.c (aarch64_simd_lane_bounds): Remove
16391 a stray %K from error_at() missed in r12-2088.
16393 2021-07-07 Richard Biener <rguenther@suse.de>
16395 PR tree-optimization/99728
16396 * tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
16398 (mem_refs_may_alias_p): Add assert we handled aggregate
16400 (sm_seq_valid_bb): Give up when running into aggregate copies.
16401 (ref_indep_loop_p): Handle aggregate copies as never
16402 being invariant themselves but allow other refs to be
16403 disambiguated against them.
16404 (can_sm_ref_p): Do not try to apply store-motion to aggregate
16407 2021-07-06 Indu Bhagat <indu.bhagat@oracle.com>
16410 * dwarf2ctf.c (ctf_get_AT_data_member_location): Multiply by 8 to get
16413 2021-07-06 Martin Sebor <msebor@redhat.com>
16415 * gimple-pretty-print.c (percent_G_format): Remove.
16416 * tree-diagnostic.c (default_tree_printer): Remove calls.
16417 * tree-pretty-print.c (percent_K_format): Remove.
16418 * tree-pretty-print.h (percent_K_format): Remove.
16420 2021-07-06 Martin Sebor <msebor@redhat.com>
16422 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_builtin):
16423 Remove %K and use error_at.
16424 (aarch64_expand_fcmla_builtin): Same.
16425 (aarch64_expand_builtin_tme): Same.
16426 (aarch64_expand_builtin_memtag): Same.
16427 * config/arm/arm-builtins.c (arm_expand_acle_builtin): Same.
16428 (arm_expand_builtin): Same.
16429 * config/arm/arm.c (bounds_check): Same.
16431 2021-07-06 Martin Sebor <msebor@redhat.com>
16433 * builtins.c (warn_string_no_nul): Remove %G.
16434 (maybe_warn_for_bound): Same.
16435 (warn_for_access): Same.
16436 (check_access): Same.
16437 (check_strncat_sizes): Same.
16438 (expand_builtin_strncat): Same.
16439 (expand_builtin_strncmp): Same.
16440 (expand_builtin): Same.
16441 (expand_builtin_object_size): Same.
16442 (warn_dealloc_offset): Same.
16443 (maybe_emit_free_warning): Same.
16444 * calls.c (maybe_warn_alloc_args_overflow): Same.
16445 (maybe_warn_nonstring_arg): Same.
16446 (maybe_warn_rdwr_sizes): Same.
16447 * expr.c (expand_expr_real_1): Remove %K.
16448 * gimple-fold.c (gimple_fold_builtin_strncpy): Remove %G.
16449 (gimple_fold_builtin_strncat): Same.
16450 * gimple-ssa-sprintf.c (format_directive): Same.
16451 (handle_printf_call): Same.
16452 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Same.
16453 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
16454 (maybe_diag_access_bounds): Same. Call gimple_location.
16455 (check_bounds_or_overlap): Same.
16456 * trans-mem.c (ipa_tm_scan_irr_block): Remove %K. Simplify.
16457 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Remove %G.
16458 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
16459 (maybe_diag_stxncpy_trunc): Same.
16460 (handle_builtin_stxncpy_strncat): Same.
16461 (maybe_warn_pointless_strcmp): Same.
16462 * tree-ssa-uninit.c (maybe_warn_operand): Same.
16464 2021-07-06 Uroš Bizjak <ubizjak@gmail.com>
16467 * config/i386/predicates.md (vec_setm_operand): Enable
16468 register_operand for TARGET_SSE4_1.
16469 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_operand
16470 as operand 2 predicate. Call ix86_expand_vector_set_var
16471 for non-constant index operand.
16472 (vec_setv4qi): Use vec_setm_mmx_operand as operand 2 predicate.
16473 Call ix86_expand_vector_set_var for non-constant index operand.
16475 2021-07-06 Jeff Law <jeffreyalaw@gmail.com>
16477 * config/h8300/jumpcall.md (*branch): When possible, generate
16478 the comparison in CCZN mode.
16479 * config/h8300/predicates.md (simple_memory_operand): Reject all
16480 auto-increment addressing modes.
16482 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
16484 PR bootstrap/100246
16485 * config/i386/i386.h (struct stringop_algs): Define a CTOR for
16488 2021-07-06 Richard Biener <rguenther@suse.de>
16490 * doc/md.texi (vec_fmaddsub<mode>4): Document.
16491 (vec_fmsubadd<mode>4): Likewise.
16492 * optabs.def (vec_fmaddsub$a4): Add.
16493 (vec_fmsubadd$a4): Likewise.
16494 * internal-fn.def (IFN_VEC_FMADDSUB): Add.
16495 (IFN_VEC_FMSUBADD): Likewise.
16496 * tree-vect-slp-patterns.c (addsub_pattern::recognize):
16497 Refactor to handle IFN_VEC_FMADDSUB and IFN_VEC_FMSUBADD.
16498 (addsub_pattern::build): Likewise.
16499 * tree-vect-slp.c (vect_optimize_slp): CFN_VEC_FMADDSUB
16500 and CFN_VEC_FMSUBADD are not transparent for permutes.
16501 * config/i386/sse.md (vec_fmaddsub<mode>4): New expander.
16502 (vec_fmsubadd<mode>4): Likewise.
16504 2021-07-06 Richard Biener <rguenther@suse.de>
16506 * doc/invoke.texi (fmove-loop-stores): Document.
16507 * common.opt (fmove-loop-stores): New option.
16508 * opts.c (default_options_table): Enable -fmove-loop-stores
16509 at -O1 but not -Og.
16510 * tree-ssa-loop-im.c (pass_lim::execute): Pass
16511 flag_move_loop_stores instead of true to
16512 loop_invariant_motion_in_fun.
16514 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
16516 * doc/install.texi: Document --with-dsymutil.
16518 2021-07-06 Andrew Pinski <apinski@marvell.com>
16520 PR tree-optimization/101256
16521 * dbgcnt.def (phiopt_edge_range): New counter.
16522 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
16523 Check to make sure the new name is defined in the same
16524 bb as the conditional before duplicating range info.
16525 Also add debug counter.
16527 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
16529 PR rtl-optimization/100328
16530 * config/i386/i386-options.c (ix86_option_override_internal):
16531 Set param_ira_consider_dup_in_all_alts to 0.
16533 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
16535 PR rtl-optimization/100328
16536 * doc/invoke.texi (ira-consider-dup-in-all-alts): Document new
16538 * ira.c (ira_get_dup_out_num): Adjust as parameter
16539 param_ira_consider_dup_in_all_alts.
16540 * params.opt (ira-consider-dup-in-all-alts): New.
16541 * ira-conflicts.c (process_regs_for_copy): Add one parameter
16542 single_input_op_has_cstr_p.
16543 (get_freq_for_shuffle_copy): New function.
16544 (add_insn_allocno_copies): Adjust as single_input_op_has_cstr_p.
16545 * ira-int.h (ira_get_dup_out_num): Add one bool parameter.
16547 2021-07-05 Jeff Law <jeffreyalaw@gmail.com>
16549 * config/h8300/shiftrotate.md (shift-by-variable patterns): Update to
16550 generate condition code aware RTL directly.
16552 2021-07-05 Andrew Pinski <apinski@marvell.com>
16554 PR tree-optimization/101039
16555 * match.pd (A CMP 0 ? A : -A): New patterns.
16556 * tree-ssa-phiopt.c (abs_replacement): Delete function.
16557 (tree_ssa_phiopt_worker): Don't call abs_replacement.
16558 Update comment about abs_replacement.
16560 2021-07-05 Andrew Pinski <apinski@marvell.com>
16562 * tree-ssa-phiopt.c (gimple_simplify_phiopt):
16563 If "A ? B : C" fails to simplify, try "(!A) ? C : B".
16565 2021-07-05 Andrew Pinski <apinski@marvell.com>
16567 * tree-ssa-phiopt.c (match_simplify_replacement):
16568 Add early_p argument. Call gimple_simplify_phiopt
16569 instead of gimple_simplify.
16570 (tree_ssa_phiopt_worker): Update call to
16571 match_simplify_replacement and allow unconditionally.
16572 (phiopt_early_allow): New function.
16573 (gimple_simplify_phiopt): New function.
16575 2021-07-05 Andrew Pinski <apinski@marvell.com>
16577 PR middle-end/101237
16578 * fold-const.c (negate_expr_p): Remove call to element_mode
16579 and TREE_MODE/TREE_TYPE when calling HONOR_SIGNED_ZEROS,
16580 HONOR_SIGN_DEPENDENT_ROUNDING, and HONOR_SNANS.
16581 (fold_negate_expr_1): Likewise.
16582 (const_unop): Likewise.
16583 (fold_cond_expr_with_comparison): Likewise.
16584 (fold_binary_loc): Likewise.
16585 (fold_ternary_loc): Likewise.
16586 (tree_call_nonnegative_warnv_p): Likewise.
16587 * match.pd (-(A + B) -> (-B) - A): Likewise.
16589 2021-07-05 Iain Sandoe <iain@sandoe.co.uk>
16591 * configure.ac: Handle --with-dsymutil in the same way as we
16592 do for the assembler and linker. (DEFAULT_DSYMUTIL): New.
16593 Extract the type and version for the dsymutil configured or
16594 found by the default searches.
16595 * config.in: Regenerated.
16596 * configure: Regenerated.
16597 * collect2.c (do_dsymutil): Handle locating dsymutil in the
16598 same way as for the assembler and linker.
16599 * config/darwin.h (DSYMUTIL): Delete.
16600 * gcc.c: Report a configured dsymutil correctly.
16601 * exec-tool.in: Allow for dsymutil.
16603 2021-07-05 Uroš Bizjak <ubizjak@gmail.com>
16605 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
16606 Handle V4QI and V2HI modes.
16607 (expand_vec_perm_blend): Allow 4-byte vector modes with TARGET_SSE4_1.
16608 Handle V4QI mode. Emit mmx_pblendvb32 for 4-byte modes.
16609 (expand_vec_perm_pshufb): Rewrite to use switch statemets.
16610 Handle 4-byte dual operands with TARGET_XOP and single operands
16611 with TARGET_SSSE3. Emit mmx_ppermv32 for TARGET_XOP and
16612 mmx_pshufbv4qi3 for TARGET_SSSE3.
16613 (expand_vec_perm_pblendv): Allow 4-byte vector modes with TARGET_SSE4_1.
16614 (expand_vec_perm_interleave2): Allow 4-byte vector modes.
16615 (expand_vec_perm_pshufb2): Allow 4-byte vector modes with TARGET_SSSE3.
16616 (expand_vec_perm_even_odd_1): Handle V4QI mode.
16617 (expand_vec_perm_broadcast_1): Handle V4QI mode.
16618 (ix86_vectorize_vec_perm_const): Handle V4QI mode.
16619 * config/i386/mmx.md (mmx_ppermv32): New insn pattern.
16620 (mmx_pshufbv4qi3): Ditto.
16621 (*mmx_pblendw32): Ditto.
16622 (*mmx_pblendw64): Rename from *mmx_pblendw.
16623 (mmx_punpckhbw_low): New insn_and_split pattern.
16624 (mmx_punpcklbw_low): Ditto.
16626 2021-07-05 Richard Biener <rguenther@suse.de>
16628 * tree-vect-loop-manip.c (vect_loop_versioning): Do not
16629 set LOOP_C_INFINITE on the vectorized loop.
16631 2021-07-05 Richard Biener <rguenther@suse.de>
16633 PR middle-end/101291
16634 * cfgloopmanip.c (loop_version): Set the loop copy of the
16635 versioned loop to the new loop.
16637 2021-07-04 Iain Sandoe <iain@sandoe.co.uk>
16640 * config.gcc: Ensure that Darwin biarch definitions are
16641 added before i386.h.
16642 * config/i386/darwin.h (TARGET_64BIT): Remove.
16643 (PR80556_WORKAROUND): New.
16644 (REAL_LIBGCC_SPEC): Amend to use PR80556_WORKAROUND.
16645 (DARWIN_SUBARCH_SPEC): New.
16646 * config/i386/darwin32-biarch.h (TARGET_64BIT_DEFAULT,
16647 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
16648 (REAL_LIBGCC_SPEC): Remove.
16649 * config/i386/darwin64-biarch.h (TARGET_64BIT_DEFAULT,
16650 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
16651 (REAL_LIBGCC_SPEC): Remove.
16653 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
16655 PR middle-end/101294
16656 * expr.c (store_constructor): Don't use vec_duplicate on vector.
16658 2021-07-02 Martin Sebor <msebor@redhat.com>
16660 PR middle-end/98871
16661 PR middle-end/98512
16662 * diagnostic.c (get_any_inlining_info): New.
16663 (update_effective_level_from_pragmas): Handle inlining context.
16664 (diagnostic_enabled): Same.
16665 (diagnostic_report_diagnostic): Same.
16666 * diagnostic.h (struct diagnostic_info): Add ctor.
16667 (struct diagnostic_context): Add new member.
16668 * tree-diagnostic.c (set_inlining_locations): New.
16669 (tree_diagnostics_defaults): Set new callback pointer.
16671 2021-07-02 Peter Bergner <bergner@linux.ibm.com>
16673 * config/rs6000/rs6000-builtin.def (BU_MMA_PAIR_LD, BU_MMA_PAIR_ST):
16675 (__builtin_vsx_lxvp, __builtin_vsx_stxvp): New built-ins.
16676 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Expand
16677 lxvp and stxvp built-ins.
16678 (mma_init_builtins): Handle lxvp and stxvp built-ins.
16679 (builtin_function_type): Likewise.
16680 * doc/extend.texi (__builtin_vsx_lxvp, __builtin_mma_stxvp): Document.
16682 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
16684 * config/h8300/h8300-protos.h (compute_a_shift_cc): Accept
16685 additional argument for the code.
16686 * config/h8300/h8300.c (compute_a_shift_cc): Accept additional
16687 argument for the code. Just return if the ZN bits are useful or
16688 not rather than the old style CC_* enums.
16689 * config/h8300/shiftrotate.md (shiftqi_noscratch): Move before
16690 more generic shiftqi patterns.
16691 (shifthi_noscratch, shiftsi_noscratch): Similarly.
16692 (shiftqi_noscratch_set_flags): New pattern.
16693 (shifthi_noscratch_set_flags, shiftsi_noscratch_set_flags): Likewise.
16695 2021-07-02 Andrew MacLeod <amacleod@redhat.com>
16697 PR tree-optimization/101223
16698 * range-op.cc (build_lt): Add -1 for signed values.
16699 (built_gt): Subtract -1 for signed values.
16701 2021-07-02 David Faust <david.faust@oracle.com>
16703 * btfout.c (get_btf_kind): Support BTF_KIND_FLOAT.
16704 (btf_asm_type): Likewise.
16706 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
16708 * config/h8300/h8300-protos.h (output_a_shift): Make first argument
16709 an array of rtx rather than a pointer to rtx. Add code argument.
16710 (compute_a_shift_length): Similarly.
16711 * config/h8300/h8300.c (h8300_shift_costs): Adjust now that the
16712 shift itself isn't an operand. Create dummy operand[0] to carry
16713 a mode and pass a suitable rtx code to compute_a_shift_length.
16714 (get_shift_alg): Adjust operand number of clobber in output templates.
16715 (output_a_shift): Make first argument an array of rtx rather than
16716 a pointer to rtx. Add code argument for the type of shift.
16717 Adjust now that the shift itself is no longer an operand.
16718 (compute_a_shift_length): Similarly.
16719 * config/h8300/shiftrotate.md (shiftqi, shifthi, shiftsi): Use an
16720 iterator rather than nshift_operator.
16721 (shiftqi_noscratch, shifthi_noscratch, shiftsi_noscratch): Likewise.
16722 (shiftqi_clobber_flags): Adjust to API changes in output_a_shift
16723 and compute_a_shift_length.
16724 (shiftqi_noscratch_clobber_flags): Likewise.
16725 (shifthi_noscratch_clobber_flags): Likewise.
16726 (shiftsi_noscratch_clobber_flags): Likewise.
16728 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
16731 * config/darwin.h (DSYMUTIL_SPEC): Do not try to run
16732 dsymutil for BTF/CTF.
16734 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
16737 * config/darwin.h (CTF_INFO_SECTION_NAME): Update the
16738 segment to include BTF.
16739 (BTF_INFO_SECTION_NAME): New.
16741 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
16743 * config/m32r/m32r-protos.h (call_operand): Adjust return type.
16744 (small_data_operand, memreg_operand, small_insn_p): Likewise.
16745 * config/m32r/m32r.c (call_operand): Adjust return type.
16746 (small_data_operand, memreg_operand): Likewise.
16748 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
16750 * config/frv/frv-protos.h (integer_register_operand): Adjust return
16752 (frv_load_operand, gpr_or_fpr_operand, gpr_no_subreg_operand): Likewise.
16753 (fpr_or_int6_operand, gpr_or_int_operand); Likewise.
16754 (gpr_or_int12_operand, gpr_or_int10_operand); Likewise.
16755 (move_source_operand, move_destination_operand): Likewise.
16756 (condexec_source_operand, condexec_dest_operand): Likewise.
16757 (lr_operand, gpr_or_memory_operand, fpr_or_memory_operand): Likewise.
16758 (reg_or_0_operand, fcc_operand, icc_operand, cc_operand): Likewise.
16759 (fcr_operand, icr_operand, cr_operand, call_operand): Likewise.
16760 (fpr_operand, even_reg_operand, odd_reg_operand): Likewise.
16761 (even_gpr_operand, odd_gpr_operand, quad_fpr_operand): Likewise.
16762 (even_fpr_operand, odd_fpr_operand): Likewise.
16763 (dbl_memory_one_insn_operand, dbl_memory_two_insn_operand): Likewise.
16764 (int12_operand, int6_operand, int5_operand, uint5_operand): Likewise.
16765 (uint4_operand, uint1_operand, int_2word_operand): Likewise
16766 (upper_int16_operand, uint16_operand, symbolic_operand): Likewise.
16767 (relational_operator, float_relational_operator): Likewise.
16768 (ccr_eqne_operator, minmax_operator): Likewise.
16769 (condexec_si_binary_operator, condexec_si_media_operator): Likewise.
16770 (condexec_si_divide_operator, condexec_si_unary_operator): Likewise.
16771 (condexec_sf_conv_operator, condexec_sf_add_operator): Likewise.
16772 (intop_compare_operator, acc_operand, even_acc_operand): Likewise.
16773 (quad_acc_operand, accg_operand): Likewise.
16775 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
16777 * config/stormy16/stormy16-protos.h (xstormy16_below_100_symbol): Change
16778 return type to a bool.
16779 (nonimmediate_nonstack_operand): Likewise.
16780 (xstormy16_splittable_below100_operand): Likewise.
16781 * config/stormy16/stormy16.c (xstormy16_below_100_symbol): Fix
16783 (xstormy16_splittable_below100_operand): Likewise.
16785 2021-07-02 Richard Biener <rguenther@suse.de>
16787 PR tree-optimization/101293
16788 * tree-ssa-loop-im.c (mem_ref_hasher::equal): Compare MEM_REF bases
16789 with combined offsets.
16790 (gather_mem_refs_stmt): Hash MEM_REFs as if their offset were
16791 combined with the rest of the offset.
16793 2021-07-02 Eric Botcazou <ebotcazou@adacore.com>
16795 * config/i386/i386.c (asm_preferred_eh_data_format): Always use the
16796 PIC encodings for PE-COFF targets.
16798 2021-07-02 Jakub Jelinek <jakub@redhat.com>
16801 * config/i386/i386-expand.c (ix86_broadcast_from_integer_constant):
16802 Return nullptr for TImode inner mode.
16804 2021-07-02 Richard Biener <rguenther@suse.de>
16806 PR tree-optimization/101280
16807 PR tree-optimization/101173
16808 * gimple-loop-interchange.cc
16809 (tree_loop_interchange::valid_data_dependences): Properly
16810 guard all dependence checks with DDR_REVERSED_P or its
16813 2021-07-02 Hongyu Wang <hongyu.wang@intel.com>
16815 * config/i386/i386-expand.c (ix86_expand_builtin):
16816 Add branch to clear odata when ZF is set for asedecenc_expand
16817 and wideaesdecenc_expand.
16819 2021-07-02 Eugene Rozenfeld <erozen@microsoft.com>
16821 * config/i386/gcc-auto-profile: regenerate
16823 2021-07-02 liuhongt <hongtao.liu@intel.com>
16825 * config/i386/sse.md (trunc<mode><pmov_dst_4>2): Refined to ..
16826 (trunc<mode><pmov_dst_4_lower>2): this.
16828 2021-07-01 David Malcolm <dmalcolm@redhat.com>
16830 * diagnostic.h (diagnostic_context::m_file_cache): New field.
16831 * input.c (class fcache): Rename to...
16832 (class file_cache_slot): ...this, making most members private and
16833 prefixing fields with "m_".
16834 (file_cache_slot::get_file_path): New accessor.
16835 (file_cache_slot::get_use_count): New accessor.
16836 (file_cache_slot::missing_trailing_newline_p): New accessor.
16837 (file_cache_slot::inc_use_count): New.
16838 (fcache_buffer_size): Move to...
16839 (file_cache_slot::buffer_size): ...here.
16840 (fcache_line_record_size): Move to...
16841 (file_cache_slot::line_record_size): ...here.
16842 (fcache_tab): Delete, in favor of global_dc->m_file_cache.
16843 (fcache_tab_size): Move to file_cache::num_file_slots.
16844 (diagnostic_file_cache_init): Update for move of fcache_tab
16845 to global_dc->m_file_cache.
16846 (diagnostic_file_cache_fini): Likewise.
16847 (lookup_file_in_cache_tab): Convert to...
16848 (file_cache::lookup_file): ...this.
16849 (diagnostics_file_cache_forcibly_evict_file): Update for move of
16850 fcache_tab to global_dc->m_file_cache, moving most of
16851 implementation to...
16852 (file_cache::forcibly_evict_file): ...this new function and...
16853 (file_cache_slot::evict): ...this new function.
16854 (evicted_cache_tab_entry): Convert to...
16855 (file_cache::evicted_cache_tab_entry): ...this.
16856 (add_file_to_cache_tab): Convert to...
16857 (file_cache::add_file): ...this, moving bulk of implementation
16859 (file_cache_slot::create): ..this new function.
16860 (file_cache::file_cache): New.
16861 (file_cache::~file_cache): New.
16862 (lookup_or_add_file_to_cache_tab): Convert to...
16863 (file_cache::lookup_or_add_file): ..this new function.
16864 (fcache::fcache): Rename to...
16865 (file_cache_slot::file_cache_slot): ...this, adding "m_" prefixes
16867 (fcache::~fcache): Rename to...
16868 (file_cache_slot::~file_cache_slot): ...this, adding "m_" prefixes
16870 (needs_read): Convert to...
16871 (file_cache_slot::needs_read_p): ...this.
16872 (needs_grow): Convert to...
16873 (file_cache_slot::needs_grow_p): ...this.
16874 (maybe_grow): Convert to...
16875 (file_cache_slot::maybe_grow): ...this.
16876 (read_data): Convert to...
16877 (file_cache_slot::read_data): ...this.
16878 (maybe_read_data): Convert to...
16879 (file_cache_slot::maybe_read_data): ...this.
16880 (get_next_line): Convert to...
16881 (file_cache_slot::get_next_line): ...this.
16882 (goto_next_line): Convert to...
16883 (file_cache_slot::goto_next_line): ...this.
16884 (read_line_num): Convert to...
16885 (file_cache_slot::read_line_num): ...this.
16886 (location_get_source_line): Update for moving of globals to
16887 global_dc->m_file_cache.
16888 (location_missing_trailing_newline): Likewise.
16889 * input.h (class file_cache_slot): New forward decl.
16890 (class file_cache): New.
16892 2021-07-01 Michael Meissner <meissner@linux.ibm.com>
16894 * config/rs6000/rs6000.c (rs6000_maybe_emit_fp_cmove): Add IEEE
16895 128-bit floating point conditional move support.
16896 (have_compare_and_set_mask): Add IEEE 128-bit floating point
16898 * config/rs6000/rs6000.md (mov<mode>cc, IEEE128 iterator): New insn.
16899 (mov<mode>cc_p10, IEEE128 iterator): New insn.
16900 (mov<mode>cc_invert_p10, IEEE128 iterator): New insn.
16901 (fpmask<mode>, IEEE128 iterator): New insn.
16902 (xxsel<mode>, IEEE128 iterator): New insn.
16904 2021-07-01 Iain Sandoe <iain@sandoe.co.uk>
16907 * config/darwin.h (CTF_INFO_SECTION_NAME): New.
16909 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
16911 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
16913 * config/i386/i386-protos.h (ix86_expand_vector_init_duplicate):
16915 * config/i386/sse.md (INT_BROADCAST_MODE): New mode iterator.
16916 (vec_duplicate<mode>): New expander.
16918 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
16921 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
16923 (ix86_byte_broadcast): New function.
16924 (ix86_convert_const_wide_int_to_broadcast): Likewise.
16925 (ix86_expand_move): Convert CONST_WIDE_INT to broadcast if mode
16926 size is 16 bytes or bigger.
16927 (ix86_broadcast_from_integer_constant): New function.
16928 (ix86_expand_vector_move): Convert CONST_WIDE_INT and CONST_VECTOR
16929 to broadcast if mode size is 16 bytes or bigger.
16930 * config/i386/i386-protos.h (ix86_gen_scratch_sse_rtx): New
16932 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): New function.
16934 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
16936 * config/i386/predicates.md (ix86_endbr_immediate_operand):
16937 Return true/false instead of 1/0.
16938 (movq_parallel): Ditto.
16940 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
16942 * recog.c (general_operand): Return true/false instead of 1/0.
16943 (register_operand): Ditto.
16944 (immediate_operand): Ditto.
16945 (const_int_operand): Ditto.
16946 (const_scalar_int_operand): Ditto.
16947 (const_double_operand): Ditto.
16948 (push_operand): Ditto.
16949 (pop_operand): Ditto.
16950 (memory_operand): Ditto.
16951 (indirect_operand): Ditto.
16953 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
16955 * genpreds.c (write_predicate_subfunction):
16956 Change the type of written subfunction to bool.
16957 (write_one_predicate_function):
16958 Change the type of written function to bool.
16959 (write_tm_preds_h): Ditto.
16960 * recog.h (*insn_operand_predicate_fn): Change the type to bool.
16961 * recog.c (general_operand): Change the type to bool.
16962 (address_operand): Ditto.
16963 (register_operand): Ditto.
16964 (pmode_register_operand): Ditto.
16965 (scratch_operand): Ditto.
16966 (immediate_operand): Ditto.
16967 (const_int_operand): Ditto.
16968 (const_scalar_int_operand): Ditto.
16969 (const_double_operand): Ditto.
16970 (nonimmediate_operand): Ditto.
16971 (nonmemory_operand): Ditto.
16972 (push_operand): Ditto.
16973 (pop_operand): Ditto.
16974 (memory_operand): Ditto.
16975 (indirect_operand): Ditto.
16976 (ordered_comparison_operator): Ditto.
16977 (comparison_operator): Ditto.
16978 * config/i386/i386-expand.c (ix86_expand_sse_cmp):
16979 Change the type of indirect predicate function to bool.
16980 * config/rs6000/rs6000.c (easy_vector_constant):
16981 Change the type to bool.
16982 * config/mips/mips-protos.h (m16_based_address_p):
16983 Change the type of operand 3 to bool.
16985 2021-07-01 Richard Biener <rguenther@suse.de>
16987 PR tree-optimization/101280
16988 PR tree-optimization/101173
16989 * gimple-loop-interchange.cc
16990 (tree_loop_interchange::valid_data_dependences): Revert
16991 previous change and instead correctly handle DDR_REVERSED_P
16994 2021-07-01 Richard Biener <rguenther@suse.de>
16996 PR tree-optimization/101278
16997 * tree-ssa-dse.c (dse_classify_store): First check for
16998 uses, then ignore stmt for chaining purposes.
17000 2021-07-01 Richard Biener <rguenther@suse.de>
17002 PR tree-optimization/100778
17003 * tree-vect-slp.c (vect_schedule_slp_node): Do not place trapping
17004 vectorized ops ahead of their scalar BB.
17006 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
17009 * config/i386/i386.md (*nabs<dwi>2_doubleword):
17010 New insn_and_split pattern.
17011 (*nabs<dwi>2_1): Ditto.
17012 * config/i386/i386-features.c
17013 (general_scalar_chain::compute_convert_gain):
17014 Handle (NEG (ABS (...))) RTX. Rewrite src code
17015 scanner as switch statement.
17016 (general_scalar_chain::convert_insn):
17017 Handle (NEG (ABS (...))) RTX.
17018 (general_scalar_to_vector_candidate_p):
17019 Detect (NEG (ABS (...))) RTX. Reorder case statements
17020 for (AND (NOT (...) ...)) fallthrough.
17022 2021-07-01 Richard Biener <rguenther@suse.de>
17024 PR tree-optimization/101178
17025 * tree-vect-slp.c (slpg_vertex::materialize): Remove.
17026 (slpg::perm_in): Add.
17027 (slpg::get_perm_in): Remove.
17028 (slpg::get_perm_materialized): Add.
17029 (vect_optimize_slp): Handle VEC_PERM nodes more optimally
17030 during permute propagation and materialization.
17032 2021-07-01 Jakub Jelinek <jakub@redhat.com>
17035 * dwarf2out.c (loc_list_from_tree_1): Handle COMPOUND_LITERAL_EXPR.
17037 2021-07-01 Jakub Jelinek <jakub@redhat.com>
17039 PR middle-end/94366
17040 * omp-low.c (lower_rec_input_clauses): Rename is_fp_and_or to
17041 is_truth_op, set it for TRUTH_*IF_EXPR regardless of new_var's type,
17042 use boolean_type_node instead of integer_type_node as NE_EXPR type.
17043 (lower_reduction_clauses): Likewise.
17045 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
17047 * config/gcn/gcn.c: Include dwarf2.h.
17048 (gcn_addr_space_debug): New function.
17049 (TARGET_ADDR_SPACE_DEBUG): New hook.
17051 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
17053 * common/config/gcn/gcn-common.c
17054 (gcn_option_optimization_table): Change OPT_fomit_frame_pointer to -O3.
17055 * config/gcn/gcn.c (gcn_expand_prologue): Prefer the frame pointer
17057 (gcn_expand_prologue): Prefer the frame pointer when emitting CFI.
17058 (gcn_frame_pointer_rqd): New function.
17059 (TARGET_FRAME_POINTER_REQUIRED): New hook.
17061 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
17063 * config/gcn/gcn.c (move_callee_saved_registers): Emit CFI notes for
17064 prologue register saves.
17065 (gcn_debug_unwind_info): Use UI_DWARF2.
17066 (gcn_dwarf_register_number): Map DWARF_LINK_REGISTER to DWARF PC.
17067 (gcn_dwarf_register_span): DWARF_LINK_REGISTER doesn't span.
17068 * config/gcn/gcn.h: (DWARF_FRAME_RETURN_COLUMN): New define.
17069 (DWARF_LINK_REGISTER): New define.
17070 (FIRST_PSEUDO_REGISTER): Increment.
17071 (FIXED_REGISTERS): Add entry for DWARF_LINK_REGISTER.
17072 (CALL_USED_REGISTERS): Likewise.
17073 (REGISTER_NAMES): Likewise.
17075 2021-06-30 Richard Biener <rguenther@suse.de>
17077 PR tree-optimization/101267
17078 * tree-vect-stmts.c (vect_check_scalar_mask): Adjust
17079 API and use SLP compatible interface of vect_is_simple_use.
17080 Reject not vectorized SLP defs for callers that do not support
17082 (vect_check_store_rhs): Handle masked stores and pass down
17083 the appropriate operator index.
17084 (vectorizable_call): Adjust.
17085 (vectorizable_store): Likewise.
17086 (vectorizable_load): Likewise. Handle SLP pecularity of
17088 (vect_is_simple_use): Remove special-casing of masked stores.
17090 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
17092 * common.opt (foffload): Remove help as Driver only.
17093 * gcc.c (display_help): Add -foffload.
17095 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
17097 * gcc.c (close_at_file, execute): Replace alloca by XALLOCAVEC.
17098 (check_offload_target_name): Fix splitting OFFLOAD_TARGETS into
17099 a candidate list; better inform no offload target is configured
17100 and fix hint extraction when passed target is not '\0' at [len].
17101 * common.opt (foffload): Add tailing '.'.
17102 (foffload-options): Likewise; fix flag name in the help string.
17104 2021-06-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
17107 * config/arm/arm_neon.h: Move vabs intrinsics before vcage_f32.
17108 (vcage_f32): Gate comparison on __FAST_MATH__.
17109 (vcageq_f32): Likewise.
17110 (vcale_f32): Likewise.
17111 (vcaleq_f32): Likewise.
17112 (vcagt_f32): Likewise.
17113 (vcagtq_f32): Likewise.
17114 (vcalt_f32): Likewise.
17115 (vcaltq_f32): Likewise.
17116 (vcage_f16): Likewise.
17117 (vcageq_f16): Likewise.
17118 (vcale_f16): Likewise.
17119 (vcaleq_f16): Likewise.
17120 (vcagt_f16): Likewise.
17121 (vcagtq_f16): Likewise.
17122 (vcalt_f16): Likewise.
17123 (vcaltq_f16): Likewise.
17125 2021-06-30 Richard Biener <rguenther@suse.de>
17127 PR tree-optimization/101264
17128 * tree-vect-slp.c (vect_optimize_slp): Propagate the
17129 computed perm_in to all "any" permute successors
17130 we cannot de-duplicate immediately.
17132 2021-06-30 liuhongt <hongtao.liu@intel.com>
17135 * config/i386/sse.md
17136 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>):
17138 (avx512f_sfixupimm<mode><maskz_scalar_name><round_saeonly_name>):
17140 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>"): Refined.
17141 * config/i386/subst.md (maskz_scalar): New define_subst.
17142 (maskz_scalar_name): New subst_attr.
17143 (maskz_scalar_op5): Ditto.
17144 (round_saeonly_maskz_scalar_op5): Ditto.
17145 (round_saeonly_maskz_scalar_operand5): Ditto.
17147 2021-06-30 David Edelsohn <dje.gcc@gmail.com>
17149 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
17150 Increase code CSECT alignment to at least 32 bytes.
17151 * config/rs6000/xcoff.h (TEXT_SECTION_ASM_OP): Add 32 byte
17152 alignment designation.
17154 2021-06-29 Sergei Trofimovich <siarheit@google.com>
17156 * doc/generic.texi: Fix s/net yet/not yet/ typo.
17158 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
17160 PR tree-optimization/101254
17161 * range-op.cc (operator_minus::op1_op2_relation_effect): Check for
17162 wrapping/non-wrapping when setting the result range.
17164 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
17166 * value-query.cc (gimple_range_global): Allow phis.
17168 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
17170 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): Use stmt.
17171 (simplify_using_ranges::op_with_boolean_value_range_p): Add a
17172 statement for location context.
17173 (check_for_binary_op_overflow): Ditto.
17174 (simplify_using_ranges::get_vr_for_comparison): Ditto.
17175 (simplify_using_ranges::compare_name_with_value): Ditto.
17176 (simplify_using_ranges::compare_names): Ditto.
17177 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Ditto.
17178 (simplify_using_ranges::simplify_truth_ops_using_ranges): Ditto.
17179 (simplify_using_ranges::simplify_min_or_max_using_ranges): Ditto.
17180 (simplify_using_ranges::simplify_internal_call_using_ranges): Ditto.
17181 (simplify_using_ranges::two_valued_val_range_p): Ditto.
17182 (simplify_using_ranges::simplify): Ditto.
17183 * vr-values.h: Adjust prototypes.
17185 2021-06-29 Uroš Bizjak <ubizjak@gmail.com>
17188 * config/i386/mmx.md (vec_addsubv2sf3): New insn pattern.
17190 2021-06-29 Julian Brown <julian@codesourcery.com>
17192 * config/gcn/gcn.c (gcn_init_libfuncs): New function.
17193 (TARGET_INIT_LIBFUNCS): Define target hook using above function.
17194 * config/gcn/gcn.h (UNITS_PER_WORD): Define to 8 for IN_LIBGCC2, 4
17196 (LIBGCC2_UNITS_PER_WORD, BITS_PER_WORD): Remove definitions.
17197 (MAX_FIXED_MODE_SIZE): Change to 128.
17199 2021-06-29 Julian Brown <julian@codesourcery.com>
17201 * config/gcn/gcn.md (UNSPEC_FLBIT_INT): New unspec constant.
17202 (s_mnemonic): Add clrsb.
17203 (gcn_flbit<mode>_int): Add insn pattern for SImode/DImode.
17204 (clrsb<mode>2): Add expander for SImode/DImode.
17206 2021-06-29 Julian Brown <julian@codesourcery.com>
17208 * config/gcn/gcn.md (<su>mulsidi3, <su>mulsidi3_reg, <su>mulsidi3_imm,
17209 muldi3): Add patterns.
17211 2021-06-29 Julian Brown <julian@codesourcery.com>
17213 * config/gcn/gcn.md (<su>mulsi3_highpart): Change to expander.
17214 (<su>mulsi3_highpart_reg, <su>mulsi3_highpart_imm): New patterns.
17216 2021-06-29 Julian Brown <julian@codesourcery.com>
17218 * config/gcn/gcn.md (mulsi3): Make s_mulk_i32 variant clobber SCC.
17220 2021-06-29 Joseph Myers <joseph@codesourcery.com>
17222 * btfout.c, ctfout.c: Include "memmodel.h".
17224 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
17226 * gcc.c (check_offload_target_name): Cast len argument to
17227 %q.*s to 'int'; avoid -Wstringop-truncation warning.
17229 2021-06-29 Richard Biener <rguenther@suse.de>
17231 * tree-vect-slp.c (vect_optimize_slp): Forward propagate
17232 to "any" permute nodes and relax "any" permute proapgation
17233 during iterative backward propagation.
17235 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
17238 * common.opt (-foffload=): Update description.
17239 (-foffload-options=): New.
17240 * doc/invoke.texi (C Language Options): Document
17241 -foffload and -foffload-options.
17242 * gcc.c (check_offload_target_name): New, split off from
17243 handle_foffload_option.
17244 (check_foffload_target_names): New.
17245 (handle_foffload_option): Handle -foffload=default.
17246 (driver_handle_option): Update for -foffload-options.
17247 * lto-opts.c (lto_write_options): Use -foffload-options
17248 instead of -foffload.
17249 * lto-wrapper.c (merge_and_complain, append_offload_options):
17251 * opts.c (common_handle_option): Likewise.
17253 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
17255 * doc/invoke.texi (C Language Options): Sort options
17256 alphabetically in optlist and also the description itself.
17257 Remove leftover -fallow-single-precision from and add missing
17258 -fgnu-tm to the optlist.
17260 2021-06-29 Richard Biener <rguenther@suse.de>
17262 * tree-vect-slp.c (slpg_vertex::visited): Remove.
17263 (vect_slp_perms_eq): Handle -1 permutes.
17264 (vect_optimize_slp): Rewrite permute propagation.
17266 2021-06-29 Jakub Jelinek <jakub@redhat.com>
17269 * match.pd ((intptr_t)x eq/ne CST to x eq/ne (typeof x) CST): Don't
17270 perform the optimization in GENERIC when sanitizing and x has a
17273 2021-06-29 Richard Biener <rguenther@suse.de>
17275 PR tree-optimization/101242
17276 * tree-vect-slp.c (vect_slp_build_vertices): Force-add
17277 PHIs with not represented initial values as leafs.
17279 2021-06-29 Jan-Benedict Glaw <jbglaw@getslash.de>
17281 * config/pdp11/pdp11.h (ASM_OUTPUT_SKIP): Fix signedness warning.
17282 * config/pdp11/pdp11.c (pdp11_asm_print_operand_punct_valid_p): Remove
17283 "register" keyword.
17284 (pdp11_initial_elimination_offset) Remove unused variable.
17285 (pdp11_cmp_length) Ditto.
17286 (pdp11_insn_cost): Ditto, and fix signedness warning.
17288 2021-06-29 David Edelsohn <dje.gcc@gmail.com>
17290 * btfout.c: Include tm_p.h.
17293 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
17295 * config/bpf/bpf.c (bpf_expand_prologue): Do not mark insns as
17297 (bpf_expand_epilogue): Likewise.
17298 * config/bpf/bpf.h (DWARF2_FRAME_INFO): Define to 0.
17299 Do not define DBX_DEBUGGING_INFO.
17301 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
17303 * doc/invoke.texi: Document the CTF and BTF debug info options.
17305 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
17306 David Faust <david.faust@oracle.com>
17307 Jose E. Marchesi <jose.marchesi@oracle.com>
17308 Weimin Pan <weimin.pan@oracle.com>
17310 * Makefile.in: Add ctfc.*, ctfout.c and btfout.c files to
17311 GTFILES. Add new object files.
17312 * common.opt: Add CTF and BTF debug info options.
17313 * btfout.c: New file.
17314 * ctfc.c: Likewise.
17315 * ctfc.h: Likewise.
17316 * ctfout.c: Likewise.
17317 * dwarf2ctf.c: Likewise.
17318 * dwarf2ctf.h: Likewise.
17319 * dwarf2cfi.c (dwarf2out_do_frame): Acknowledge CTF_DEBUG and
17321 * dwarf2out.c (dwarf2out_source_line): Likewise.
17322 (dwarf2out_finish): Skip emitting DWARF if CTF or BTF are to
17324 (debug_format_do_cu): New function.
17325 (dwarf2out_early_finish): Traverse DIEs and emit CTF/BTF for
17327 Include dwarf2ctf.c.
17328 * final.c (dwarf2_debug_info_emitted_p): Acknowledge DWARF-based debug
17330 * flag-types.h (enum debug_info_type): Add CTF_DEBUG and BTF_DEBUG.
17331 (CTF_DEBUG): New bitmask.
17332 (BTF_DEBUG): Likewise.
17333 (enum ctf_debug_info_levels): New enum.
17334 * gengtype.c (open_base_files): Handle ctfc.h.
17335 (main): Handle uint32_t type.
17336 * flags.h (btf_debuginfo_p): New definition.
17337 (dwarf_based_debuginfo_p): Likewise.
17338 * opts.c (debug_type_names): Add entries for CTF and BTF.
17339 (btf_debuginfo_p): New function.
17340 (dwarf_based_debuginfo_p): Likewise.
17341 (common_handle_option): Handle -gctfN and -gbtf options.
17342 (set_debug_level): Set CTF_DEBUG, BTF_DEBUG whenever appropriate.
17343 * toplev.c (process_options): Inform the user and ignore -gctfLEVEL if
17346 2021-06-28 Jose E. Marchesi <jose.marchesi@oracle.com>
17348 * dwarf2out.c (AT_class): Function is no longer static.
17349 (AT_int): Likewise.
17350 (AT_unsigned): Likewise.
17351 (AT_loc): Likewise.
17352 (get_AT): Likewise.
17353 (get_AT_string): Likewise.
17354 (get_AT_flag): Likewise.
17355 (get_AT_unsigned): Likewise.
17356 (get_AT_ref): Likewise.
17357 (new_die_raw): Likewise.
17358 (lookup_decl_die): Likewise.
17359 (base_type_die): Likewise.
17360 (add_name_attribute): Likewise.
17361 (add_AT_int): Likewise.
17362 (add_AT_unsigned): Likewise.
17363 (add_AT_loc): Likewise.
17364 (dw_get_die_tag): New function.
17365 (dw_get_die_child): Likewise.
17366 (dw_get_die_sib): Likewise.
17367 (struct dwarf_file_data): Move from here to dwarf2out.h
17368 (struct dw_attr_struct): Likewise.
17369 * dwarf2out.h: Analogous changes.
17371 2021-06-28 Martin Jambor <mjambor@suse.cz>
17374 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
17375 members m_dead_stmts and m_dead_ssas.
17376 * ipa-param-manipulation.c
17377 (ipa_param_body_adjustments::mark_dead_statements): New function.
17378 (ipa_param_body_adjustments::common_initialization): Call it on
17379 all removed but not split parameters.
17380 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
17382 (ipa_param_body_adjustments::modify_call_stmt): Remove arguments that
17384 * tree-inline.c (remap_gimple_stmt): Do not copy dead statements, reset
17385 dead debug statements.
17386 (copy_phis_for_bb): Do not copy dead PHI nodes.
17388 2021-06-28 Martin Jambor <mjambor@suse.cz>
17391 * symtab-clones.h (clone_info): Removed member param_adjustments.
17392 * ipa-param-manipulation.h: Adjust initial comment to reflect how we
17393 deal with pass-through splits now.
17394 (ipa_param_performed_split): Removed.
17395 (ipa_param_adjustments::modify_call): Adjusted parameters.
17396 (class ipa_param_body_adjustments): Adjusted parameters of
17397 register_replacement, modify_gimple_stmt and modify_call_stmt.
17398 (ipa_verify_edge_has_no_modifications): Declare.
17399 (ipa_edge_modifications_finalize): Declare.
17400 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Remove
17401 performed_splits processing, pas only edge to padjs->modify_call,
17402 check that call arguments were not modified if they should not have
17404 * cgraphclones.c (cgraph_node::create_clone): Do not copy performed
17406 * ipa-param-manipulation.c (struct pass_through_split_map): New type.
17407 (ipa_edge_modification_info): Likewise.
17408 (ipa_edge_modification_sum): Likewise.
17409 (ipa_edge_modifications): New edge summary.
17410 (ipa_verify_edge_has_no_modifications): New function.
17411 (transitive_split_p): Removed.
17412 (transitive_split_map): Likewise.
17413 (init_transitive_splits): Likewise.
17414 (ipa_param_adjustments::modify_call): Adjusted to use the new edge
17415 summary instead of performed_splits.
17416 (ipa_param_body_adjustments::register_replacement): Drop dummy
17417 parameter, set base_index of the created ipa_param_body_replacement.
17418 (phi_arg_will_live_p): New function.
17419 (ipa_param_body_adjustments::common_initialization): Do not create
17420 IPA_SRA dummy decls.
17421 (simple_tree_swap_info): Removed.
17422 (remap_split_decl_to_dummy): Likewise.
17423 (record_argument_state_1): New function.
17424 (record_argument_state): Likewise.
17425 (ipa_param_body_adjustments::modify_call_stmt): New parameter
17426 orig_stmt. Do not work with dummy decls, save necessary info about
17427 changes to ipa_edge_modifications.
17428 (ipa_param_body_adjustments::modify_gimple_stmt): New parameter
17429 orig_stmt, pass it to modify_call_stmt.
17430 (ipa_param_body_adjustments::modify_cfun_body): Adjust call to
17431 modify_gimple_stmt.
17432 (ipa_edge_modifications_finalize): New function.
17433 * tree-inline.c (remap_gimple_stmt): Pass original statement to
17434 modify_gimple_stmt.
17435 (copy_phis_for_bb): Do not copy dead PHI nodes.
17436 (expand_call_inline): Do not remap performed_splits.
17437 (update_clone_info): Likewise.
17438 * toplev.c: Include ipa-param-manipulation.h.
17439 (toplev::finalize): Call ipa_edge_modifications_finalize.
17441 2021-06-28 Andrew Pinski <apinski@marvell.com>
17443 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Duplicate range
17444 info if we're the only things setting the target PHI.
17445 (value_replacement): Don't duplicate range here.
17446 (minmax_replacement): Likewise.
17448 2021-06-28 Richard Biener <rguenther@suse.de>
17450 PR tree-optimization/101229
17451 * gimple-walk.c (gimple_walk_op): Handle PHIs.
17453 2021-06-28 Martin Liska <mliska@suse.cz>
17455 * config/v850/v850.c (construct_dispose_instruction): Allocate
17457 (construct_prepare_instruction): Likewise.
17459 2021-06-28 Martin Liska <mliska@suse.cz>
17461 * config/v850/v850.c (v850_option_override): Build default
17463 (v850_can_inline_p): New. Allow MASK_PROLOG_FUNCTION to be
17464 ignored for inlining.
17465 (TARGET_CAN_INLINE_P): New.
17467 2021-06-28 Richard Biener <rguenther@suse.de>
17469 PR tree-optimization/101207
17470 * tree-vect-slp.c (vect_optimize_slp): Do BB reduction
17471 permute eliding for load permutations properly.
17473 2021-06-28 Richard Biener <rguenther@suse.de>
17475 PR tree-optimization/101173
17476 * gimple-loop-interchange.cc
17477 (tree_loop_interchange::valid_data_dependences): Disallow outer
17478 loop dependence distance of zero.
17480 2021-06-28 liuhongt <hongtao.liu@intel.com>
17483 * config/i386/sse.md (*avx_cmp<mode>3_lt): New
17484 define_insn_and_split.
17485 (*avx_cmp<mode>3_ltint): Ditto.
17486 (*avx2_pcmp<mode>3_3): Ditto.
17487 (*avx2_pcmp<mode>3_4): Ditto.
17488 (*avx2_pcmp<mode>3_5): Ditto.
17490 2021-06-28 liuhongt <hongtao.liu@intel.com>
17492 * config/i386/i386-builtin.def (IX86_BUILTIN_BLENDVPD256,
17493 IX86_BUILTIN_BLENDVPS256, IX86_BUILTIN_PBLENDVB256,
17494 IX86_BUILTIN_BLENDVPD, IX86_BUILTIN_BLENDVPS,
17495 IX86_BUILTIN_PBLENDVB128): Replace icode with
17497 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold blendv
17499 * config/i386/sse.md (*<sse4_1_avx2>_pblendvb_lt_subreg_not):
17500 New pre_reload splitter.
17502 2021-06-27 Andrew Pinski <apinski@marvell.com>
17504 PR middle-end/101230
17505 * fold-const.c (fold_ternary_loc): Check
17506 the return value of invert_tree_comparison.
17508 2021-06-27 David Edelsohn <dje.gcc@gmail.com>
17510 * config.gcc: Add SPDX License Identifier.
17511 (powerpc-ibm-aix789): Default to aix73.h.
17512 (powerpc-ibm-aix7.2.*.*): New stanza.
17513 * config/rs6000/aix72.h: Add SPDX License Identifier.
17514 * config/rs6000/aix73.h: New file.
17516 2021-06-26 Jason Merrill <jason@redhat.com>
17518 * except.c: #include "dwarf2.h" instead of "dwarf2out.h".
17520 2021-06-26 Andrew Pinski <apinski@marvell.com>
17522 * genmatch.c (lower_cond): Copy for_subst_vec
17523 for the simplify also.
17524 (lower): Swap the order for lower_for and lower_cond.
17526 2021-06-26 Andrew Pinski <apinski@marvell.com>
17528 * tree-ssa-phiopt.c (match_simplify_replacement): Reset
17529 flow senatitive info on the moved ssa set.
17531 2021-06-26 Andrew Pinski <apinski@marvell.com>
17533 * fold-const.c (fold_cond_expr_with_comparison):
17534 Exand arg0 into comp_code, arg00, and arg01.
17535 (fold_ternary_loc): Use invert_tree_comparison
17536 instead of fold_invert_truthvalue for the case
17537 where we have A CMP B ? C : A.
17539 2021-06-25 Martin Sebor <msebor@redhat.com>
17541 PR middle-end/101216
17542 * calls.c (maybe_warn_rdwr_sizes): Use the no_warning constant.
17544 2021-06-25 Jeff Law <jeffreyalaw@gmail.com>
17546 * config/h8300/h8300.c (select_cc_mode): Handle ASHIFTRT and LSHIFTRT.
17548 2021-06-25 Richard Biener <rguenther@suse.de>
17550 PR tree-optimization/101202
17551 * tree-vect-slp.c (vect_optimize_slp): Explicitely handle
17554 2021-06-25 Richard Biener <rguenther@suse.de>
17556 * tree-vect-slp-patterns.c (addsub_pattern::build): Copy
17557 STMT_VINFO_REDUC_DEF from the original representative.
17559 2021-06-25 Martin Sebor <msebor@redhat.com>
17561 * builtins.c (warn_string_no_nul): Replace uses of TREE_NO_WARNING,
17562 gimple_no_warning_p and gimple_set_no_warning with
17563 warning_suppressed_p, and suppress_warning.
17565 (maybe_warn_for_bound): Same.
17566 (warn_for_access): Same.
17567 (check_access): Same.
17568 (expand_builtin_strncmp): Same.
17569 (fold_builtin_varargs): Same.
17570 * calls.c (maybe_warn_nonstring_arg): Same.
17571 (maybe_warn_rdwr_sizes): Same.
17572 * cfgexpand.c (expand_call_stmt): Same.
17573 * cgraphunit.c (check_global_declaration): Same.
17574 * fold-const.c (fold_undefer_overflow_warnings): Same.
17575 (fold_truth_not_expr): Same.
17576 (fold_unary_loc): Same.
17577 (fold_checksum_tree): Same.
17578 * gimple-array-bounds.cc (array_bounds_checker::check_array_ref): Same.
17579 (array_bounds_checker::check_mem_ref): Same.
17580 (array_bounds_checker::check_addr_expr): Same.
17581 (array_bounds_checker::check_array_bounds): Same.
17582 * gimple-expr.c (copy_var_decl): Same.
17583 * gimple-fold.c (gimple_fold_builtin_strcpy): Same.
17584 (gimple_fold_builtin_strncat): Same.
17585 (gimple_fold_builtin_stxcpy_chk): Same.
17586 (gimple_fold_builtin_stpcpy): Same.
17587 (gimple_fold_builtin_sprintf): Same.
17588 (fold_stmt_1): Same.
17589 * gimple-ssa-isolate-paths.c (diag_returned_locals): Same.
17590 * gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Same.
17591 * gimple-ssa-sprintf.c (handle_printf_call): Same.
17592 * gimple-ssa-store-merging.c (imm_store_chain_info::output_merged_store): Same.
17593 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
17594 * gimple-ssa-warn-restrict.h: Adjust declarations.
17595 (maybe_diag_access_bounds): Replace uses of TREE_NO_WARNING,
17596 gimple_no_warning_p and gimple_set_no_warning with
17597 warning_suppressed_p, and suppress_warning.
17598 (check_call): Same.
17599 (check_bounds_or_overlap): Same.
17600 * gimple.c (gimple_build_call_from_tree): Same.
17601 * gimplify.c (gimplify_return_expr): Same.
17602 (gimplify_cond_expr): Same.
17603 (gimplify_modify_expr_complex_part): Same.
17604 (gimplify_modify_expr): Same.
17605 (gimple_push_cleanup): Same.
17606 (gimplify_expr): Same.
17607 * omp-expand.c (expand_omp_for_generic): Same.
17608 (expand_omp_taskloop_for_outer): Same.
17609 * omp-low.c (lower_rec_input_clauses): Same.
17610 (lower_lastprivate_clauses): Same.
17611 (lower_send_clauses): Same.
17612 (lower_omp_target): Same.
17613 * tree-cfg.c (pass_warn_function_return::execute): Same.
17614 * tree-complex.c (create_one_component_var): Same.
17615 * tree-inline.c (remap_gimple_op_r): Same.
17616 (copy_tree_body_r): Same.
17617 (declare_return_variable): Same.
17618 (expand_call_inline): Same.
17619 * tree-nested.c (lookup_field_for_decl): Same.
17620 * tree-sra.c (create_access_replacement): Same.
17621 (generate_subtree_copies): Same.
17622 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Same.
17623 * tree-ssa-forwprop.c (combine_cond_expr_cond): Same.
17624 * tree-ssa-loop-ch.c (ch_base::copy_headers): Same.
17625 * tree-ssa-loop-im.c (execute_sm): Same.
17626 * tree-ssa-phiopt.c (cond_store_replacement): Same.
17627 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
17628 (handle_builtin_strcpy): Same.
17629 (maybe_diag_stxncpy_trunc): Same.
17630 (handle_builtin_stxncpy_strncat): Same.
17631 (handle_builtin_strcat): Same.
17632 * tree-ssa-uninit.c (get_no_uninit_warning): Same.
17633 (set_no_uninit_warning): Same.
17634 (uninit_undefined_value_p): Same.
17635 (warn_uninit): Same.
17636 (maybe_warn_operand): Same.
17637 * tree-vrp.c (compare_values_warnv): Same.
17638 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr): Same.
17639 (test_for_singularity): Same.
17640 * gimple.h (warning_suppressed_p): New function.
17641 (suppress_warning): Same.
17642 (copy_no_warning): Same.
17643 (gimple_set_block): Call gimple_set_location.
17644 (gimple_set_location): Call copy_warning.
17646 2021-06-25 Martin Sebor <msebor@redhat.com>
17648 * tree.h (warning_suppressed_at, copy_warning,
17649 warning_suppressed_p, suppress_warning): New functions.
17651 2021-06-25 Martin Sebor <msebor@redhat.com>
17653 * Makefile.in (OBJS-libcommon): Add diagnostic-spec.o.
17654 * gengtype.c (open_base_files): Add diagnostic-spec.h.
17655 * diagnostic-spec.c: New file.
17656 * diagnostic-spec.h: New file.
17657 * tree.h (no_warning, all_warnings, suppress_warning_at): New
17659 * warning-control.cc: New file.
17661 2021-06-25 liuhongt <hongtao.liu@intel.com>
17664 * config/i386/i386.c (x86_order_regs_for_local_alloc):
17667 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
17669 PR tree-optimization/101189
17670 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Pass
17671 LHS range of condition to postfold routine.
17672 (fold_using_range::postfold_gcond_edges): Only process the TRUE or
17673 FALSE edge if the LHS range supports it being taken.
17674 * gimple-range-fold.h (postfold_gcond_edges): Add range parameter.
17676 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
17678 * value-relation.cc (equiv_oracle::dump): Do not dump NULL blocks.
17679 (relation_oracle::find_relation_block): Check correct bitmap.
17680 (relation_oracle::dump): Do not dump NULL blocks.
17682 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
17684 * gimple-range-cache.cc (ranger_cache::propagate_cache): Call
17685 range_on_edge instead of manually calculating.
17687 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
17689 * range-op.cc: Fix comment.
17691 2021-06-24 Uroš Bizjak <ubizjak@gmail.com>
17694 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
17695 Handle V8QI and V4HI modes.
17696 * config/i386/mmx.md (sse4_1_<any_extend:code>v4qiv4hi2):
17698 (sse4_1_<any_extend:code>v4qiv4hi2): Ditto.
17699 (mmxpackmode): New mode attribute.
17700 (vec_pack_trunc_<mmxpackmode:mode>): New expander.
17701 (mmxunpackmode): New mode attribute.
17702 (vec_unpacks_lo_<mmxunpackmode:mode>): New expander.
17703 (vec_unpacks_hi_<mmxunpackmode:mode>): Ditto.
17704 (vec_unpacku_lo_<mmxunpackmode:mode>): Ditto.
17705 (vec_unpacku_hi_<mmxunpackmode:mode>): Ditto.
17706 * config/i386/i386.md (extsuffix): Move from ...
17707 * config/i386/sse.md: ... here.
17709 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
17711 * dwarf2out.c (dwarf2out_assembly_start): Emit .file 0 marker here..
17712 (dwarf2out_finish): ...instead of here.
17714 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
17716 * configure.ac (--gdwarf-5 option): Use objdump instead of readelf.
17717 (working --gdwarf-4/--gdwarf-5 for all sources): Likewise.
17718 (--gdwarf-4 not refusing generated .debug_line): Adjust for Windows.
17719 * configure: Regenerate.
17721 2021-06-24 Richard Biener <rguenther@suse.de>
17723 * config/i386/sse.md (vec_addsubv4df3, vec_addsubv2df3,
17724 vec_addsubv8sf3, vec_addsubv4sf3): Merge into ...
17725 (vec_addsub<mode>3): ... using a new addsub_cst mode attribute.
17727 2021-06-24 Richard Biener <rguenther@suse.de>
17729 * config/i386/sse.md (avx_addsubv4df3): Rename to
17731 (avx_addsubv8sf3): Rename to vec_addsubv8sf3.
17732 (sse3_addsubv2df3): Rename to vec_addsubv2df3.
17733 (sse3_addsubv4sf3): Rename to vec_addsubv4sf3.
17734 * config/i386/i386-builtin.def: Adjust.
17735 * internal-fn.def (VEC_ADDSUB): New internal optab fn.
17736 * optabs.def (vec_addsub_optab): New optab.
17737 * tree-vect-slp-patterns.c (class addsub_pattern): New.
17738 (slp_patterns): Add addsub_pattern.
17739 * tree-vect-slp.c (vect_optimize_slp): Disable propagation
17740 across CFN_VEC_ADDSUB.
17741 * tree-vectorizer.h (vect_pattern::vect_pattern): Make
17743 * doc/md.texi (vec_addsub<mode>3): Document.
17745 2021-06-24 Jakub Jelinek <jakub@redhat.com>
17747 PR middle-end/101170
17748 * df-scan.c (df_ref_record): For paradoxical big-endian SUBREGs
17749 where regno + subreg_regno_offset wraps around use 0 as starting
17752 2021-06-24 Jakub Jelinek <jakub@redhat.com>
17754 PR middle-end/101172
17755 * stor-layout.c (finish_bitfield_representative): If nextf has
17756 error_mark_node type, set repr type to error_mark_node too.
17758 2021-06-24 Ilya Leoshkevich <iii@linux.ibm.com>
17760 * config/s390/s390.c (s390_function_profiler): Ignore labelno
17762 * config/s390/s390.h (NO_PROFILE_COUNTERS): Define.
17764 2021-06-24 Richard Biener <rguenther@suse.de>
17766 * tree-vect-slp.c (vect_optimize_slp): Do not propagate
17767 across operations that have different semantics on different
17770 2021-06-24 Jakub Jelinek <jakub@redhat.com>
17772 * tree.h (OMP_CLAUSE_MAP_IN_REDUCTION): Document meaning for OpenMP.
17773 * gimplify.c (gimplify_scan_omp_clauses): For OpenMP map clauses
17774 with OMP_CLAUSE_MAP_IN_REDUCTION flag partially defer gimplification
17775 of non-decl OMP_CLAUSE_DECL. For OMP_CLAUSE_IN_REDUCTION on
17776 OMP_TARGET user outer_ctx instead of ctx for placeholders and
17777 initializer/combiner gimplification.
17778 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_MAP_IN_REDUCTION
17779 on target constructs.
17780 (lower_rec_input_clauses): Likewise.
17781 (lower_omp_target): Likewise.
17782 * omp-expand.c (expand_omp_target): Temporarily ignore nowait clause
17783 on target if in_reduction is present.
17785 2021-06-24 Kewen Lin <linkw@linux.ibm.com>
17787 * tree-predcom.c (class pcom_worker): New class.
17788 (release_chain): Renamed to...
17789 (pcom_worker::release_chain): ...this.
17790 (release_chains): Renamed to...
17791 (pcom_worker::release_chains): ...this.
17792 (aff_combination_dr_offset): Renamed to...
17793 (pcom_worker::aff_combination_dr_offset): ...this.
17794 (determine_offset): Renamed to...
17795 (pcom_worker::determine_offset): ...this.
17796 (class comp_ptrs): New class.
17797 (split_data_refs_to_components): Renamed to...
17798 (pcom_worker::split_data_refs_to_components): ...this,
17799 and update with class comp_ptrs.
17800 (suitable_component_p): Renamed to...
17801 (pcom_worker::suitable_component_p): ...this.
17802 (filter_suitable_components): Renamed to...
17803 (pcom_worker::filter_suitable_components): ...this.
17804 (valid_initializer_p): Renamed to...
17805 (pcom_worker::valid_initializer_p): ...this.
17806 (find_looparound_phi): Renamed to...
17807 (pcom_worker::find_looparound_phi): ...this.
17808 (add_looparound_copies): Renamed to...
17809 (pcom_worker::add_looparound_copies): ...this.
17810 (determine_roots_comp): Renamed to...
17811 (pcom_worker::determine_roots_comp): ...this.
17812 (determine_roots): Renamed to...
17813 (pcom_worker::determine_roots): ...this.
17814 (single_nonlooparound_use): Renamed to...
17815 (pcom_worker::single_nonlooparound_use): ...this.
17816 (remove_stmt): Renamed to...
17817 (pcom_worker::remove_stmt): ...this.
17818 (execute_pred_commoning_chain): Renamed to...
17819 (pcom_worker::execute_pred_commoning_chain): ...this.
17820 (execute_pred_commoning): Renamed to...
17821 (pcom_worker::execute_pred_commoning): ...this.
17822 (struct epcc_data): New member worker.
17823 (execute_pred_commoning_cbck): Call execute_pred_commoning
17824 with pcom_worker pointer.
17825 (find_use_stmt): Renamed to...
17826 (pcom_worker::find_use_stmt): ...this.
17827 (find_associative_operation_root): Renamed to...
17828 (pcom_worker::find_associative_operation_root): ...this.
17829 (find_common_use_stmt): Renamed to...
17830 (pcom_worker::find_common_use_stmt): ...this.
17831 (combinable_refs_p): Renamed to...
17832 (pcom_worker::combinable_refs_p): ...this.
17833 (reassociate_to_the_same_stmt): Renamed to...
17834 (pcom_worker::reassociate_to_the_same_stmt): ...this.
17835 (stmt_combining_refs): Renamed to...
17836 (pcom_worker::stmt_combining_refs): ...this.
17837 (combine_chains): Renamed to...
17838 (pcom_worker::combine_chains): ...this.
17839 (try_combine_chains): Renamed to...
17840 (pcom_worker::try_combine_chains): ...this.
17841 (prepare_initializers_chain): Renamed to...
17842 (pcom_worker::prepare_initializers_chain): ...this.
17843 (prepare_initializers): Renamed to...
17844 (pcom_worker::prepare_initializers): ...this.
17845 (prepare_finalizers_chain): Renamed to...
17846 (pcom_worker::prepare_finalizers_chain): ...this.
17847 (prepare_finalizers): Renamed to...
17848 (pcom_worker::prepare_finalizers): ...this.
17849 (tree_predictive_commoning_loop): Renamed to...
17850 (pcom_worker::tree_predictive_commoning_loop): ...this, adjust
17851 some calls and remove some cleanup code.
17852 (tree_predictive_commoning): Adjusted to use pcom_worker instance.
17853 (static variable looparound_phis): Remove.
17854 (static variable name_expansions): Remove.
17856 2021-06-24 Richard Biener <rguenther@suse.de>
17858 * tree-vect-slp.c (slpg_vertex): New struct.
17859 (vect_slp_build_vertices): Adjust.
17860 (vect_optimize_slp): Likewise. Maintain an outgoing permute
17861 and a materialized one.
17863 2021-06-24 Richard Biener <rguenther@suse.de>
17865 PR tree-optimization/101105
17866 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
17867 Only ignore steps when they are equal or scalar order is preserved.
17869 2021-06-24 liuhongt <hongtao.liu@intel.com>
17872 * config/i386/i386-expand.c (ix86_expand_vec_interleave):
17873 Adjust comments for ix86_expand_vecop_qihi2.
17874 (ix86_expand_vecmul_qihi): Renamed to ..
17875 (ix86_expand_vecop_qihi2): Adjust function prototype to
17876 support shift operation, add static to definition.
17877 (ix86_expand_vec_shift_qihi_constant): Add static to definition.
17878 (ix86_expand_vecop_qihi): Call ix86_expand_vecop_qihi2 and
17879 ix86_expand_vec_shift_qihi_constant.
17880 * config/i386/i386-protos.h (ix86_expand_vecmul_qihi): Deleted.
17881 (ix86_expand_vec_shift_qihi_constant): Deleted.
17882 * config/i386/sse.md (VI12_256_512_AVX512VL): New mode
17884 (mulv8qi3): Call ix86_expand_vecop_qihi directly, add
17885 condition TARGET_64BIT.
17886 (mul<mode>3): Ditto.
17887 (<insn><mode>3): Ditto.
17888 (vlshr<mode>3): Extend to support avx512 vlshr.
17889 (v<insn><mode>3): New expander for
17891 (v<insn>v8qi3): Ditto.
17892 (vashrv8hi3<mask_name>): Renamed to ..
17893 (vashr<mode>3): And extend to support V16QImode for avx512.
17894 (vashrv16qi3): Deleted.
17895 (vashrv2di3<mask_name>): Extend expander to support avx512
17898 2021-06-23 Dimitar Dimitrov <dimitar@dinux.eu>
17900 * doc/lto.texi (Design Overview): Update that slim objects are
17903 2021-06-23 Aaron Sawdey <acsawdey@linux.ibm.com>
17905 * config/rs6000/rs6000-cpus.def: Take OPTION_MASK_PCREL_OPT out
17906 of OTHER_POWER10_MASKS so it will not be enabled by default.
17908 2021-06-23 Richard Biener <rguenther@suse.de>
17909 Martin Jambor <mjambor@suse.cz>
17911 * tree-inline.c (setup_one_parameter): Set TREE_READONLY of the
17912 param replacement unconditionally. Adjust comment.
17914 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
17916 * Makefile.in (OBJS): Add gimple-range-fold.o
17917 * gimple-range-fold.cc: New.
17918 * gimple-range-fold.h: New.
17919 * gimple-range-gori.cc (gimple_range_calc_op1): Move to here.
17920 (gimple_range_calc_op2): Ditto.
17921 * gimple-range-gori.h: Move prototypes to here.
17922 * gimple-range.cc: Adjust include files.
17923 (fur_source:fur_source): Relocate to gimple-range-fold.cc.
17924 (fur_source::get_operand): Ditto.
17925 (fur_source::get_phi_operand): Ditto.
17926 (fur_source::query_relation): Ditto.
17927 (fur_source::register_relation): Ditto.
17928 (class fur_edge): Ditto.
17929 (fur_edge::fur_edge): Ditto.
17930 (fur_edge::get_operand): Ditto.
17931 (fur_edge::get_phi_operand): Ditto.
17932 (fur_stmt::fur_stmt): Ditto.
17933 (fur_stmt::get_operand): Ditto.
17934 (fur_stmt::get_phi_operand): Ditto.
17935 (fur_stmt::query_relation): Ditto.
17936 (class fur_depend): Relocate to gimple-range-fold.h.
17937 (fur_depend::fur_depend): Relocate to gimple-range-fold.cc.
17938 (fur_depend::register_relation): Ditto.
17939 (fur_depend::register_relation): Ditto.
17940 (class fur_list): Ditto.
17941 (fur_list::fur_list): Ditto.
17942 (fur_list::get_operand): Ditto.
17943 (fur_list::get_phi_operand): Ditto.
17944 (fold_range): Ditto.
17945 (adjust_pointer_diff_expr): Ditto.
17946 (gimple_range_adjustment): Ditto.
17947 (gimple_range_base_of_assignment): Ditto.
17948 (gimple_range_operand1): Ditto.
17949 (gimple_range_operand2): Ditto.
17950 (gimple_range_calc_op1): Relocate to gimple-range-gori.cc.
17951 (gimple_range_calc_op2): Ditto.
17952 (fold_using_range::fold_stmt): Relocate to gimple-range-fold.cc.
17953 (fold_using_range::range_of_range_op): Ditto.
17954 (fold_using_range::range_of_address): Ditto.
17955 (fold_using_range::range_of_phi): Ditto.
17956 (fold_using_range::range_of_call): Ditto.
17957 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
17958 (fold_using_range::range_of_builtin_call): Ditto.
17959 (fold_using_range::range_of_cond_expr): Ditto.
17960 (fold_using_range::range_of_ssa_name_with_loop_info): Ditto.
17961 (fold_using_range::relation_fold_and_or): Ditto.
17962 (fold_using_range::postfold_gcond_edges): Ditto.
17963 * gimple-range.h: Add gimple-range-fold.h to include files. Change
17964 GIMPLE_RANGE_STMT_H to GIMPLE_RANGE_H.
17965 (gimple_range_handler): Relocate to gimple-range-fold.h.
17966 (gimple_range_ssa_p): Ditto.
17967 (range_compatible_p): Ditto.
17968 (class fur_source): Ditto.
17969 (class fur_stmt): Ditto.
17970 (class fold_using_range): Ditto.
17971 (gimple_range_calc_op1): Relocate to gimple-range-gori.h
17972 (gimple_range_calc_op2): Ditto.
17974 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
17976 PR tree-optimization/101148
17977 PR tree-optimization/101014
17978 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
17979 (ranger_cache::~ranger_cache): Adjust.
17980 (ranger_cache::block_range): Check if propagation disallowed.
17981 (ranger_cache::propagate_cache): Disallow propagation if new value
17982 can't be stored properly.
17983 * gimple-range-cache.h (ranger_cache::m_propfail): New member.
17985 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
17987 * gimple-range-cache.cc (class ssa_block_ranges): Adjust prototype.
17988 (sbr_vector::set_bb_range): Return true.
17989 (class sbr_sparse_bitmap): Adjust.
17990 (sbr_sparse_bitmap::set_bb_range): Return value.
17991 (block_range_cache::set_bb_range): Return value.
17992 (ranger_cache::propagate_cache): Use return value to print msg.
17993 * gimple-range-cache.h (class block_range_cache): Adjust.
17995 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
17997 * gimple-range.cc (dump_bb): Use range_on_edge from the cache.
17999 2021-06-23 Jeff Law <jeffreyalaw@gmail.com>
18001 * config/h8300/logical.md (<code><mode>3<ccnz>): Use <cczn>
18002 so this pattern can be used for test/compare removal. Pass
18003 current insn to compute_logical_op_length and output_logical_op.
18004 * config/h8300/h8300.c (compute_logical_op_cc): Remove.
18005 (h8300_and_costs): Add argument to compute_logical_op_length.
18006 (output_logical_op): Add new argument. Use it to determine if the
18007 condition codes are used and adjust the output accordingly.
18008 (compute_logical_op_length): Add new argument and update length
18009 computations when condition codes are used.
18010 * config/h8300/h8300-protos.h (compute_logical_op_length): Update
18012 (output_logical_op): Likewise.
18014 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
18017 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
18018 Handle 64bit modes for TARGET_XOP. Use indirect gen_* functions.
18019 * config/i386/mmx.md (mmx_ppermv64): New insn pattern.
18020 * config/i386/i386.md (unspec): Move UNSPEC_XOP_PERMUTE from ...
18021 * config/i386/sse.md (unspec): ... here.
18023 2021-06-23 Martin Liska <mliska@suse.cz>
18026 * optc-save-gen.awk: Put back arm_fp16_format to
18029 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
18032 * config/i386/i386.md (bsr_rex64): Add zero-flag setting RTX.
18035 (clz<mode>2): Update RTX pattern for additions.
18037 2021-06-23 Jakub Jelinek <jakub@redhat.com>
18039 PR middle-end/101167
18040 * omp-low.c (lower_omp_regimplify_p): Regimplify also PARM_DECLs
18041 and RESULT_DECLs that have DECL_HAS_VALUE_EXPR_P set.
18043 2021-06-22 Sergei Trofimovich <siarheit@google.com>
18045 * doc/rtl.texi: drop unbalanced parenthesis.
18047 2021-06-22 Richard Biener <rguenther@suse.de>
18049 PR middle-end/101156
18050 * gimplify.c (gimplify_expr): Remove premature incorrect
18053 2021-06-22 Jakub Jelinek <jakub@redhat.com>
18055 PR tree-optimization/101159
18056 * tree-vect-patterns.c (vect_recog_popcount_pattern): Fix some
18059 2021-06-22 Jakub Jelinek <jakub@redhat.com>
18061 PR middle-end/101160
18062 * function.c (assign_parms): For decl_result with TYPE_EMPTY_P type
18063 clear crtl->return_rtx instead of keeping it referencing a pseudo.
18065 2021-06-22 Jakub Jelinek <jakub@redhat.com>
18066 Andrew Pinski <apinski@marvell.com>
18068 PR tree-optimization/101162
18069 * fold-const.c (range_check_type): Handle OFFSET_TYPE like pointer
18072 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
18074 * range-op.cc (range_relational_tests): New.
18075 (range_op_tests): Call range_relational_tests.
18077 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
18079 * range-op.cc (operator_cast::lhs_op1_relation): New.
18080 (operator_identity::lhs_op1_relation): Mew.
18082 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
18084 * range-op.cc (operator_minus::op1_op2_relation_effect): New.
18086 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
18088 * range-op.cc (operator_plus::lhs_op1_relation): New.
18089 (operator_plus::lhs_op2_relation): New.
18091 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
18093 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a
18094 relation_oracle if dominators exist.
18095 (ranger_cache::~ranger_cache): Dispose of oracle.
18096 (ranger_cache::dump_bb): Dump oracle.
18097 * gimple-range.cc (fur_source::fur_source): New.
18098 (fur_source::get_operand): Use mmeber query.
18099 (fur_source::get_phi_operand): Use member_query.
18100 (fur_source::query_relation): New.
18101 (fur_source::register_dependency): Delete.
18102 (fur_source::register_relation): New.
18103 (fur_edge::fur_edge): Adjust.
18104 (fur_edge::get_phi_operand): Fix comment.
18105 (fur_edge::query): Delete.
18106 (fur_stmt::fur_stmt): Adjust.
18107 (fur_stmt::query): Delete.
18108 (fur_depend::fur_depend): Adjust.
18109 (fur_depend::register_relation): New.
18110 (fur_depend::register_relation): New.
18111 (fur_list::fur_list): Adjust.
18112 (fur_list::get_operand): Use member query.
18113 (fold_using_range::range_of_range_op): Process and query relations.
18114 (fold_using_range::range_of_address): Adjust dependency call.
18115 (fold_using_range::range_of_phi): Ditto.
18116 (gimple_ranger::gimple_ranger): New. Use ranger_ache oracle.
18117 (fold_using_range::relation_fold_and_or): New.
18118 (fold_using_range::postfold_gcond_edges): New.
18119 * gimple-range.h (class gimple_ranger): Adjust.
18120 (class fur_source): Adjust members.
18121 (class fur_stmt): Ditto.
18122 (class fold_using_range): Ditto.
18124 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
18126 * range-op.cc (range_operator::wi_fold): Apply relation effect.
18127 (range_operator::fold_range): Adjust and apply relation effect.
18128 (*::fold_range): Add relation parameters.
18129 (*::op1_range): Ditto.
18130 (*::op2_range): Ditto.
18131 (range_operator::lhs_op1_relation): New.
18132 (range_operator::lhs_op2_relation): New.
18133 (range_operator::op1_op2_relation): New.
18134 (range_operator::op1_op2_relation_effect): New.
18135 (relop_early_resolve): New.
18136 (operator_equal::op1_op2_relation): New.
18137 (operator_equal::fold_range): Call relop_early_resolve.
18138 (operator_not_equal::op1_op2_relation): New.
18139 (operator_not_equal::fold_range): Call relop_early_resolve.
18140 (operator_lt::op1_op2_relation): New.
18141 (operator_lt::fold_range): Call relop_early_resolve.
18142 (operator_le::op1_op2_relation): New.
18143 (operator_le::fold_range): Call relop_early_resolve.
18144 (operator_gt::op1_op2_relation): New.
18145 (operator_gt::fold_range): Call relop_early_resolve.
18146 (operator_ge::op1_op2_relation): New.
18147 (operator_ge::fold_range): Call relop_early_resolve.
18148 * range-op.h (class range_operator): Adjust parameters and methods.
18150 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
18152 * Makefile.in (OBJS): Add value-relation.o.
18153 * gimple-range.h: Adjust include files.
18154 * tree-data-ref.c: Adjust include file order.
18155 * value-query.cc (range_query::get_value_range): Default to no oracle.
18156 (range_query::query_relation): New.
18157 (range_query::query_relation): New.
18158 * value-query.h (class range_query): Adjust.
18159 * value-relation.cc: New.
18160 * value-relation.h: New.
18162 2021-06-22 Richard Biener <rguenther@suse.de>
18164 PR tree-optimization/101151
18165 * tree-ssa-sink.c (statement_sink_location): Expand irreducible
18168 2021-06-22 Jojo R <rjiejie@linux.alibaba.com>
18170 * config/riscv/riscv.c (thead_c906_tune_info): New.
18171 (riscv_tune_info_table): Use new tune.
18173 2021-06-22 Richard Biener <rguenther@suse.de>
18175 PR tree-optimization/101158
18176 * tree-vect-slp.c (vect_build_slp_tree_1): Move same operand
18177 checking after checking for matching operation.
18179 2021-06-22 Richard Biener <rguenther@suse.de>
18181 PR tree-optimization/101159
18182 * tree-vect-patterns.c (vect_recog_popcount_pattern): Add
18183 missing NULL vectype check.
18185 2021-06-22 Richard Biener <rguenther@suse.de>
18187 PR tree-optimization/101154
18188 * tree-vect-slp.c (vect_build_slp_tree_2): Fix out-of-bound access.
18190 2021-06-22 Jakub Jelinek <jakub@redhat.com>
18193 * config/i386/i386-protos.h (ix86_last_zero_store_uid): Declare.
18194 * config/i386/i386-expand.c (ix86_last_zero_store_uid): New variable.
18195 * config/i386/i386.c (ix86_expand_prologue): Clear it.
18196 * config/i386/i386.md (peephole2s for 1/2/4 stores of const0_rtx):
18197 Remove "" from match_operand. Emit new insns using emit_move_insn and
18198 set ix86_last_zero_store_uid to INSN_UID of the last store.
18199 Add peephole2s for 1/2/4 stores of const0_rtx following previous
18202 2021-06-22 Martin Liska <mliska@suse.cz>
18204 * auto-profile.c (AUTO_PROFILE_VERSION): Bump as string format
18207 2021-06-22 Martin Liska <mliska@suse.cz>
18209 * gcov-io.h: Remove padding entries.
18211 2021-06-22 liuhongt <hongtao.liu@intel.com>
18213 PR tree-optimization/97770
18214 * tree-vect-patterns.c (vect_recog_popcount_pattern):
18216 (vect_recog_func vect_vect_recog_func_ptrs): Add new pattern.
18218 2021-06-22 liuhongt <hongtao.liu@intel.com>
18221 * config/i386/i386-builtin.def (BDESC): Adjust builtin name.
18222 * config/i386/sse.md (<avx512>_expand<mode>_mask): Rename to ..
18223 (expand<mode>_mask): this ..
18224 (*expand<mode>_mask): New pre_reload splitter to transform
18225 v{,p}expand* to vmov* when mask is zero, all ones, or has all
18226 ones in it's lower part, otherwise still generate
18229 2021-06-22 liuhongt <hongtao.liu@intel.com>
18232 * config/i386/i386-expand.c
18233 (ix86_expand_special_args_builtin): Keep constm1_operand only
18234 if it satisfies insn's operand predicate.
18236 2021-06-21 Jason Merrill <jason@redhat.com>
18239 * df-scan.c (df_ref_record): Check that regno < endregno.
18240 * function.c (assign_parms, expand_function_end): Do nothing with a
18241 TYPE_EMPTY_P result.
18243 2021-06-21 Richard Biener <rguenther@suse.de>
18245 PR tree-optimization/101120
18246 * tree-vect-data-refs.c (bump_vector_ptr): Fold the
18248 * tree-vect-slp.c (vect_transform_slp_perm_load): Add
18249 DR chain DCE capability.
18250 * tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
18251 * tree-vect-stmts.c (vectorizable_load): Remove unused
18252 loads in the DR chain for SLP.
18254 2021-06-21 Jakub Jelinek <jakub@redhat.com>
18256 PR inline-asm/100785
18257 * gimplify.c (gimplify_asm_expr): Don't diagnose errors if
18258 output or input operands were already error_mark_node.
18259 * cfgexpand.c (expand_asm_stmt): If errors are emitted,
18260 remove all inputs, outputs and clobbers from the asm and
18261 set template to "".
18263 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
18265 * config/arm/arm_neon.h (vceq_s8): Replace builtin with __a == __b.
18266 (vceq_s16): Likewise.
18267 (vceq_s32): Likewise.
18268 (vceq_u8): Likewise.
18269 (vceq_u16): Likewise.
18270 (vceq_u32): Likewise.
18271 (vceq_p8): Likewise.
18272 (vceqq_s8): Likewise.
18273 (vceqq_s16): Likewise.
18274 (vceqq_s32): Likewise.
18275 (vceqq_u8): Likewise.
18276 (vceqq_u16): Likewise.
18277 (vceqq_u32): Likewise.
18278 (vceqq_p8): Likewise.
18279 (vceq_f32): Gate __a == __b on __FAST_MATH__.
18280 (vceqq_f32): Likewise.
18281 (vceq_f16): Likewise.
18282 (vceqq_f16): Likewise.
18284 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
18287 * config/arm/iterators.md (NEON_VACMP): Remove.
18288 * config/arm/neon.md (neon_vca<cmp_op><mode>): Use GLTE instead of GTGE
18290 (neon_vca<cmp_op><mode>_insn): Likewise.
18291 (neon_vca<cmp_op_unsp><mode>_insn_unspec): Use NEON_VAGLTE instead of
18294 2021-06-21 Richard Biener <rguenther@suse.de>
18296 PR tree-optimization/101121
18297 * tree-vect-slp.c (vect_build_slp_tree_2): To not fail fatally
18298 when we just lack a stmt with the desired op when doing permutation.
18299 (vect_build_slp_tree): When caching a failed SLP build attempt
18300 assert that at least one lane is marked as not matching.
18302 2021-06-21 liuhongt <hongtao.liu@intel.com>
18305 * config/i386/i386.md: (*anddi_1): Disparage slightly the mask
18306 register alternative.
18307 (*and<mode>_1): Ditto.
18309 (*andn<mode>_1): Ditto.
18310 (*<code><mode>_1): Ditto.
18311 (*<code>qi_1): Ditto.
18312 (*one_cmpl<mode>2_1): Ditto.
18313 (*one_cmplsi2_1_zext): Ditto.
18314 (*one_cmplqi2_1): Ditto.
18315 * config/i386/i386.c (x86_order_regs_for_local_alloc): Change
18316 the order of mask registers to be before general registers.
18318 2021-06-21 Roger Sayle <roger@nextmovesoftware.com>
18321 * config/i386/i386.md: New define_peephole2s to shrink writing
18322 1, 2 or 4 consecutive zeros to memory when optimizing for size.
18324 2021-06-18 Jeff Law <jeffreyalaw@gmail.com>
18326 * config/h8300/h8300.c (h8300_select_cc_mode): Handle SYMBOL_REF.
18327 * config/h8300/logical.md (<code><mode>3 logcial expander): Generate
18328 more efficient code when the source can be trivially simplified.
18330 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
18332 * gimple-range-cache.cc (ranger_cache::range_of_def): Calculate
18333 a range if global is not available.
18334 (ranger_cache::entry_range): Fallback to range_of_def.
18335 * gimple-range-cache.h (range_of_def): Adjust prototype.
18337 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
18339 PR tree-optimization/101014
18340 * gimple-range-cache.cc (ranger_cache::ranger_cache): Remove poor
18342 (ranger_cache::~ranger_cache): Ditto.
18343 (ranger_cache::enable_new_values): Delete.
18344 (ranger_cache::push_poor_value): Delete.
18345 (ranger_cache::range_of_def): Remove poor value processing.
18346 (ranger_cache::entry_range): Ditto.
18347 (ranger_cache::fill_block_cache): Ditto.
18348 * gimple-range-cache.h (class ranger_cache): Remove poor value members.
18349 * gimple-range.cc (gimple_ranger::range_of_expr): Remove call.
18350 * gimple-range.h (class gimple_ranger): Adjust.
18352 2021-06-18 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
18355 * common/config/arm/arm-common.c (arm_canon_arch_option_1): New function
18356 derived from arm_canon_arch.
18357 (arm_canon_arch_option): Call it.
18358 (arm_canon_arch_multilib_option): New function.
18359 * config/arm/arm-cpus.in (IGNORE_FOR_MULTILIB): New fgroup.
18360 * config/arm/arm.h (arm_canon_arch_multilib_option): New prototype.
18361 (CANON_ARCH_MULTILIB_SPEC_FUNCTION): New macro.
18362 (MULTILIB_ARCH_CANONICAL_SPECS): New macro.
18363 (DRIVER_SELF_SPECS): Add MULTILIB_ARCH_CANONICAL_SPECS.
18364 * config/arm/arm.opt (mlibarch): New option.
18365 * config/arm/t-rmprofile (MULTILIB_MATCHES): For armv8*-m, replace use
18366 of march on RHS with mlibarch.
18368 2021-06-18 Marcel Vollweiler <marcel@codesourcery.com>
18370 * config.in: Regenerate.
18371 * config/gcn/gcn.c (print_operand_address): Fix for global_load assembler
18373 * configure: Regenerate.
18374 * configure.ac: Fix for global_load assembler functions.
18376 2021-06-18 Richard Biener <rguenther@suse.de>
18378 PR tree-optimization/101112
18379 * tree-vect-slp.c (vect_slp_linearize_chain): Fix condition
18380 to lookup a pattern stmt def.
18382 2021-06-18 Jakub Jelinek <jakub@redhat.com>
18384 PR middle-end/101062
18385 * stor-layout.c (finish_bitfield_layout): Don't add bitfield
18386 representatives in QUAL_UNION_TYPE.
18388 2021-06-18 Andrew Pinski <apinski@marvell.com>
18390 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
18391 Add counting of how many times it is done.
18392 (factor_out_conditional_conversion): Likewise.
18393 (match_simplify_replacement): Likewise.
18394 (value_replacement): Likewise.
18395 (spaceship_replacement): Likewise.
18396 (cond_store_replacement): Likewise.
18397 (cond_if_else_store_replacement_1): Likewise.
18398 (hoist_adjacent_loads): Likewise.
18400 2021-06-18 Andrew Pinski <apinski@marvell.com>
18402 * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
18403 types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and CNONJ_EXPR.
18404 (verify_gimple_assign_binary): Reject point and offset types on
18405 MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
18406 FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
18407 FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.
18409 2021-06-18 Michael Meissner <meissner@linux.ibm.com>
18411 * config/rs6000/rs6000.c (rs6000_emit_minmax): Add support for ISA
18412 3.1 IEEE 128-bit floating point xsmaxcqp/xsmincqp instructions.
18413 * config/rs6000/rs6000.md (s<minmax><mode>3, IEEE128 iterator):
18416 2021-06-17 Aaron Sawdey <acsawdey@linux.ibm.com>
18418 * config/rs6000/genfusion.pl (gen_logical_addsubf): Add
18419 earlyclobber to alts 0/1.
18420 (gen_addadd): Add earlyclobber to alts 0/1.
18421 * config/rs6000/fusion.md: Regenerate file.
18423 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
18425 * cfgloopanal.c (get_loop_hot_path): Make path an auto_vec.
18427 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
18429 * gimple-range-cache.cc: Comment cleanups.
18430 * gimple-range-gori.cc: Comment cleanups.
18431 * gimple-range.cc: Comment/spacing cleanups
18432 * value-range.h: Comment cleanups.
18434 2021-06-17 H.J. Lu <hjl.tools@gmail.com>
18437 * calls.c (expand_call): Replace PUSH_ARGS with
18438 targetm.calls.push_argument (0).
18439 (emit_library_call_value_1): Likewise.
18440 * defaults.h (PUSH_ARGS): Removed.
18441 (PUSH_ARGS_REVERSED): Replace PUSH_ARGS with
18442 targetm.calls.push_argument (0).
18443 * expr.c (block_move_libcall_safe_for_call_parm): Likewise.
18444 (emit_push_insn): Pass the number bytes to push to
18445 targetm.calls.push_argument and pass 0 if ARGS_ADDR is 0.
18446 * hooks.c (hook_bool_uint_true): New.
18447 * hooks.h (hook_bool_uint_true): Likewise.
18448 * rtlanal.c (nonzero_bits1): Replace PUSH_ARGS with
18449 targetm.calls.push_argument (0).
18450 * target.def (push_argument): Add a targetm.calls hook.
18451 * targhooks.c (default_push_argument): New.
18452 * targhooks.h (default_push_argument): Likewise.
18453 * config/bpf/bpf.h (PUSH_ARGS): Removed.
18454 * config/cr16/cr16.c (TARGET_PUSH_ARGUMENT): New.
18455 * config/cr16/cr16.h (PUSH_ARGS): Removed.
18456 * config/i386/i386.c (ix86_push_argument): New.
18457 (TARGET_PUSH_ARGUMENT): Likewise.
18458 * config/i386/i386.h (PUSH_ARGS): Removed.
18459 * config/m32c/m32c.c (TARGET_PUSH_ARGUMENT): New.
18460 * config/m32c/m32c.h (PUSH_ARGS): Removed.
18461 * config/nios2/nios2.h (PUSH_ARGS): Likewise.
18462 * config/pru/pru.h (PUSH_ARGS): Likewise.
18463 * doc/tm.texi.in: Remove PUSH_ARGS documentation. Add
18464 TARGET_PUSH_ARGUMENT hook.
18465 * doc/tm.texi: Regenerated.
18467 2021-06-17 Uroš Bizjak <ubizjak@gmail.com>
18470 * config/i386/i386-expand.c (expand_vector_set_var):
18471 Handle V2FS mode remapping. Pass TARGET_MMX_WITH_SSE to
18472 ix86_expand_vector_init_duplicate.
18473 (ix86_expand_vector_init_duplicate): Emit insv_1 for
18474 QImode for !TARGET_PARTIAL_REG_STALL.
18475 * config/i386/predicates.md (vec_setm_mmx_operand): New predicate.
18476 * config/i386/mmx.md (vec_setv2sf): Use vec_setm_mmx_operand
18477 as operand 2 predicate. Call ix86_expand_vector_set_var
18478 for non-constant index operand.
18479 (vec_setv2si): Ditto.
18480 (vec_setv4hi): Ditto.
18481 (vec_setv8qi): ditto.
18483 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
18485 PR tree-optimization/100790
18486 * gimple-range.cc (range_of_builtin_call): Cleanup clz and ctz
18489 2021-06-17 Martin Liska <mliska@suse.cz>
18491 * doc/invoke.texi: Use consistently -O1 instead of -O.
18493 2021-06-17 Martin Liska <mliska@suse.cz>
18495 * gcov-io.h: Update documentation entry about string format.
18497 2021-06-17 Marius Hillenbrand <mhillen@linux.ibm.com>
18500 * config/s390/vecintrin.h (vec_doublee): Fix to use
18501 __builtin_s390_vflls.
18502 (vec_floate): Fix to use __builtin_s390_vflrd.
18504 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
18506 * dominance.c (get_dominated_to_depth): Return auto_vec<basic_block>.
18507 * dominance.h (get_dominated_to_depth): Likewise.
18508 (get_all_dominated_blocks): Likewise.
18509 * cfgcleanup.c (delete_unreachable_blocks): Adjust.
18510 * gcse.c (hoist_code): Likewise.
18511 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
18512 * tree-parloops.c (oacc_entry_exit_ok): Likewise.
18513 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Likewise.
18514 * tree-ssa-phiprop.c (pass_phiprop::execute): Likewise.
18516 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
18518 * dominance.c (get_dominated_by_region): Return auto_vec<basic_block>.
18519 * dominance.h (get_dominated_by_region): Likewise.
18520 * tree-cfg.c (gimple_duplicate_sese_region): Adjust.
18521 (gimple_duplicate_sese_tail): Likewise.
18522 (move_sese_region_to_fn): Likewise.
18524 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
18526 * dominance.c (get_dominated_by): Return auto_vec<basic_block>.
18527 * dominance.h (get_dominated_by): Likewise.
18528 * auto-profile.c (afdo_find_equiv_class): Adjust.
18529 * cfgloopmanip.c (duplicate_loop_to_header_edge): Likewise.
18530 * loop-unroll.c (unroll_loop_runtime_iterations): Likewise.
18531 * tree-cfg.c (test_linear_chain): Likewise.
18532 (test_diamond): Likewise.
18534 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
18536 * cfgloop.h (get_loop_hot_path): Return auto_vec<basic_block>.
18537 * cfgloopanal.c (get_loop_hot_path): Likewise.
18538 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
18540 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
18542 * cgraph.c (cgraph_node::collect_callers): Return
18543 auto_vec<cgraph_edge *>.
18544 * cgraph.h (cgraph_node::collect_callers): Likewise.
18545 * ipa-cp.c (create_specialized_node): Adjust.
18546 (decide_about_value): Likewise.
18547 (decide_whether_version_node): Likewise.
18548 * ipa-sra.c (process_isra_node_results): Likewise.
18550 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
18552 * vec.h (vl_ptr>::using_auto_storage): Handle null m_vec.
18553 (auto_vec<T, 0>::auto_vec): Define move constructor, and delete copy
18555 (auto_vec<T, 0>::operator=): Define move assignment and delete copy
18558 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
18560 * gimple-range.cc (debug_seed_ranger): New.
18561 (dump_ranger): New.
18562 (debug_ranger): New.
18564 2021-06-17 Richard Biener <rguenther@suse.de>
18566 PR tree-optimization/54400
18567 * tree-vectorizer.h (enum slp_instance_kind): Add
18568 slp_inst_kind_bb_reduc.
18569 (reduction_fn_for_scalar_code): Declare.
18570 * tree-vect-data-refs.c (vect_slp_analyze_instance_dependence):
18571 Check SLP_INSTANCE_KIND instead of looking at the
18573 (vect_slp_analyze_instance_alignment): Likewise.
18574 * tree-vect-loop.c (reduction_fn_for_scalar_code): Export.
18575 * tree-vect-slp.c (vect_slp_linearize_chain): Split out
18576 chain linearization from vect_build_slp_tree_2 and generalize
18577 for the use of BB reduction vectorization.
18578 (vect_build_slp_tree_2): Adjust accordingly.
18579 (vect_optimize_slp): Elide permutes at the root of BB reduction
18581 (vectorizable_bb_reduc_epilogue): New function.
18582 (vect_slp_prune_covered_roots): Likewise.
18583 (vect_slp_analyze_operations): Use them.
18584 (vect_slp_check_for_constructors): Recognize associatable
18585 chains for BB reduction vectorization.
18586 (vectorize_slp_instance_root_stmt): Generate code for the
18587 BB reduction epilogue.
18589 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
18591 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Check with
18593 (gori_compute::may_recompute_p): New.
18594 (gori_compute::outgoing_edge_range_p): Perform recomputations.
18595 * gimple-range-gori.h (class gori_compute): Add prototype.
18597 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
18599 * gimple-range-cache.cc (ranger_cache::range_on_edge): Always return
18600 true when a range can be calculated.
18601 * gimple-range.cc (gimple_ranger::dump_bb): Check has_edge_range_p.
18603 2021-06-16 Martin Sebor <msebor@redhat.com>
18605 * doc/invoke.texi (-Wmismatched-dealloc, -Wmismatched-new-delete):
18606 Correct documented defaults.
18608 2021-06-16 Andrew MacLeod <amacleod@redhat.com>
18610 * gimple-range-cache.cc (ranger_cache::ranger_cache): Initialize
18611 m_new_value_p directly.
18613 2021-06-16 Uroš Bizjak <ubizjak@gmail.com>
18616 * config/i386/i386-expand.c (expand_vec_perm_2perm_pblendv):
18617 Handle 64bit modes for TARGET_SSE4_1.
18618 (expand_vec_perm_pshufb2): Handle 64bit modes for TARGET_SSSE3.
18619 (expand_vec_perm_even_odd_pack): Handle V4HI mode.
18620 (expand_vec_perm_even_odd_1) <case E_V4HImode>: Expand via
18621 expand_vec_perm_pshufb2 for TARGET_SSSE3 and via
18622 expand_vec_perm_even_odd_pack for TARGET_SSE4_1.
18623 * config/i386/mmx.md (mmx_packusdw): New insn pattern.
18625 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
18627 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn<mode>):
18628 Change to an expander that emits the correct instruction
18629 depending on endianness.
18630 (aarch64_<sur><addsub>hn<mode>_insn_le): Define.
18631 (aarch64_<sur><addsub>hn<mode>_insn_be): Define.
18633 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
18635 * config/aarch64/aarch64-simd-builtins.def: Split generator
18636 for aarch64_<su>qmovn builtins into scalar and vector
18638 * config/aarch64/aarch64-simd.md (aarch64_<su>qmovn<mode>_insn_le):
18640 (aarch64_<su>qmovn<mode>_insn_be): Define.
18641 (aarch64_<su>qmovn<mode>): Split into scalar and vector
18642 variants. Change vector variant to an expander that emits the
18643 correct instruction depending on endianness.
18645 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
18647 * config/aarch64/aarch64-simd-builtins.def: Split generator
18648 for aarch64_sqmovun builtins into scalar and vector variants.
18649 * config/aarch64/aarch64-simd.md (aarch64_sqmovun<mode>):
18650 Split into scalar and vector variants. Change vector variant
18651 to an expander that emits the correct instruction depending
18653 (aarch64_sqmovun<mode>_insn_le): Define.
18654 (aarch64_sqmovun<mode>_insn_be): Define.
18656 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
18658 * config/aarch64/aarch64-simd.md (aarch64_xtn<mode>_insn_le):
18659 Define - modeling zero-high-half semantics.
18660 (aarch64_xtn<mode>): Change to an expander that emits the
18661 appropriate instruction depending on endianness.
18662 (aarch64_xtn<mode>_insn_be): Define - modeling zero-high-half
18664 (aarch64_xtn2<mode>_le): Rename to...
18665 (aarch64_xtn2<mode>_insn_le): This.
18666 (aarch64_xtn2<mode>_be): Rename to...
18667 (aarch64_xtn2<mode>_insn_be): This.
18668 (vec_pack_trunc_<mode>): Emit truncation instruction instead
18670 * config/aarch64/iterators.md (Vnarrowd): Add Vnarrowd mode
18671 attribute iterator.
18673 2021-06-16 Martin Jambor <mjambor@suse.cz>
18675 PR tree-optimization/100453
18676 * tree-sra.c (create_access): Disqualify any const candidates
18677 which are written to.
18678 (sra_modify_expr): Do not store sub-replacements back to a const base.
18679 (handle_unscalarized_data_in_subtree): Likewise.
18680 (sra_modify_assign): Likewise. Earlier, use TREE_READONLy test
18681 instead of constant_decl_p.
18683 2021-06-16 Jakub Jelinek <jakub@redhat.com>
18685 PR middle-end/101062
18686 * stor-layout.c (finish_bitfield_representative): For fields in unions
18687 assume nextf is always NULL.
18688 (finish_bitfield_layout): Compute bit field representatives also in
18689 unions, but handle it as if each bitfield was the only field in the
18692 2021-06-16 Richard Biener <rguenther@suse.de>
18694 PR tree-optimization/101088
18695 * tree-ssa-loop-im.c (sm_seq_valid_bb): Only look for
18696 supported refs on edges. Do not assert same ref but
18697 different kind stores are unsuported but mark them so.
18698 (hoist_memory_references): Only look for supported refs
18701 2021-06-16 Roger Sayle <roger@nextmovesoftware.com>
18703 PR rtl-optimization/46235
18704 * config/i386/i386.md: New define_split for bt followed by cmov.
18705 (*bt<mode>_setcqi): New define_insn_and_split for bt followed by setc.
18706 (*bt<mode>_setncqi): New define_insn_and_split for bt then setnc.
18707 (*bt<mode>_setnc<mode>): New define_insn_and_split for bt followed
18708 by setnc with zero extension.
18710 2021-06-16 Richard Biener <rguenther@suse.de>
18712 PR tree-optimization/101083
18713 * tree-vect-slp.c (vect_slp_build_two_operator_nodes): Get
18714 vectype as argument.
18715 (vect_build_slp_tree_2): Adjust.
18717 2021-06-15 Martin Sebor <msebor@redhat.com>
18719 PR middle-end/100876
18720 * builtins.c: (gimple_call_return_array): Account for size_t
18721 mangling as either unsigned int or unsigned long
18723 2021-06-15 Jeff Law <jeffreyalaw@gmail.com>
18725 * compare-elim.c (try_eliminate_compare): Run DCE to clean things
18726 up before eliminating comparisons.
18728 2021-06-15 Aldy Hernandez <aldyh@redhat.com>
18730 * range-op.cc (operator_bitwise_or::wi_fold): Make sure
18731 nonzero|X is nonzero.
18732 (range_op_bitwise_and_tests): Add tests for above.
18734 2021-06-15 Carl Love <cel@us.ibm.com>
18737 * config/rs6000/rs6000-builtin.def (VCMPEQUT): Fix the ICODE for the
18739 (VRLQ, VSLQ, VSRQ, VSRAQ): Remove unused BU_P10_OVERLOAD_2
18742 2021-06-15 Tobias Burnus <tobias@codesourcery.com>
18745 * gimplify.c (enum gimplify_defaultmap_kind): Add GDMK_SCALAR_TARGET.
18746 (struct gimplify_omp_ctx): Extend defaultmap array by one.
18747 (new_omp_context): Init defaultmap[GDMK_SCALAR_TARGET].
18748 (omp_notice_variable): Update type classification for Fortran.
18749 (gimplify_scan_omp_clauses): Update calls for new argument; handle
18750 GDMK_SCALAR_TARGET; for Fortran, GDMK_POINTER avoid GOVD_MAP_0LEN_ARRAY.
18751 * langhooks-def.h (lhd_omp_scalar_p): Add 'ptr_ok' argument.
18752 * langhooks.c (lhd_omp_scalar_p): Likewise.
18753 (LANG_HOOKS_OMP_ALLOCATABLE_P, LANG_HOOKS_OMP_SCALAR_TARGET_P): New.
18754 (LANG_HOOKS_DECLS): Add them.
18755 * langhooks.h (struct lang_hooks_for_decls): Add new hooks, update
18756 omp_scalar_p pointer type to include the new bool argument.
18758 2021-06-15 David Malcolm <dmalcolm@redhat.com>
18760 * doc/analyzer.texi
18761 (Special Functions for Debugging the Analyzer): Add
18762 __analyzer_dump_capacity.
18764 2021-06-15 Jakub Jelinek <jakub@redhat.com>
18767 * expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
18768 case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.
18770 2021-06-15 Richard Biener <rguenther@suse.de>
18772 * cfgloopanal.c (mark_irreducible_loops): Use a dominance
18773 check to identify loop latches.
18774 * cfgloop.c (verify_loop_structure): Likewise.
18775 * loop-init.c (apply_loop_flags): Allow marked irreducible
18776 regions even with multiple latches.
18777 * predict.c (rebuild_frequencies): Simplify.
18779 2021-06-15 Richard Biener <rguenther@suse.de>
18781 * tree-ssa-threadupdate.c
18782 (jump_thread_path_registry::mark_threaded_blocks): Assert we
18783 have marked irreducible regions.
18785 2021-06-14 Martin Sebor <msebor@redhat.com>
18788 * builtins.c (gimple_call_return_array): Check for attribute fn spec.
18789 Handle calls to placement new.
18790 (ndecl_dealloc_argno): Avoid placement delete.
18792 2021-06-14 Peter Bergner <bergner@linux.ibm.com>
18795 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
18796 create_tmp_reg_or_ssa_name().
18798 2021-06-14 Andrew MacLeod <amacleod@redhat.com>
18800 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
18801 (ranger_cache::enable_new_values): Set to specified value and
18802 return the old value.
18803 (ranger_cache::disable_new_values): Delete.
18804 (ranger_cache::fill_block_cache): Disable non 1st order derived
18806 * gimple-range-cache.h (ranger_cache): Adjust prototypes.
18807 * gimple-range.cc (gimple_ranger::range_of_expr): Adjust.
18809 2021-06-14 Uroš Bizjak <ubizjak@gmail.com>
18812 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
18813 Return true early when testing with V2HImode.
18814 * config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.
18816 2021-06-14 Christophe Lyon <christophe.lyon@linaro.org>
18818 * config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
18819 (mve_vec_unpack<US>_hi_<mode>): New pattern.
18820 (@mve_vec_pack_trunc_lo_<mode>): New pattern.
18821 (mve_vmovntq_<supf><mode>): Prefix with '@'.
18822 * config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
18824 (vec_unpack<US>_lo_<mode>): Likewise.
18825 (vec_pack_trunc_<mode>): Rename to
18826 neon_quad_vec_pack_trunc_<mode>.
18827 * config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
18829 (vec_unpack<US>_lo_<mode>): New.
18830 (vec_pack_trunc_<mode>): New.
18832 2021-06-14 Richard Biener <rguenther@suse.de>
18834 PR tree-optimization/100934
18835 * tree-ssa-dom.c (pass_dominator::execute): Properly
18836 mark irreducible regions.
18838 2021-06-14 Martin Liska <mliska@suse.cz>
18840 * doc/invoke.texi: Put r{...} on the same line as @item.
18842 2021-06-14 Martin Liska <mliska@suse.cz>
18844 * doc/invoke.texi: Add missing newline.
18846 2021-06-14 Martin Liska <mliska@suse.cz>
18848 * doc/invoke.texi: Remove '+' charasters.
18850 2021-06-14 Claudiu Zissulescu <claziss@synopsys.com>
18852 * config.gcc (arc): Add support for with_cpu option.
18853 * config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.
18855 2021-06-14 Richard Biener <rguenther@suse.de>
18857 PR tree-optimization/101031
18858 * tree-ssa-strlen.c (maybe_invalidate): Increment max_size
18859 instead of size when accounting for a possibly string
18862 2021-06-14 Martin Liska <mliska@suse.cz>
18864 * gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].
18866 2021-06-14 Aldy Hernandez <aldyh@redhat.com>
18868 * value-query.cc (gimple_range_global): Call get_range_global
18869 if called after inlining.
18871 2021-06-13 Uroš Bizjak <ubizjak@gmail.com>
18874 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
18875 Emit constant permutation insn directly from here.
18877 2021-06-13 Trevor Saunders <tbsaunde@tbsaunde.org>
18879 * attribs.c (find_attribute_namespace): Iterate over vec<> with
18881 * auto-profile.c (afdo_find_equiv_class): Likewise.
18882 * gcc.c (do_specs_vec): Likewise.
18883 (do_spec_1): Likewise.
18884 (driver::set_up_specs): Likewise.
18885 * gimple-loop-jam.c (any_access_function_variant_p): Likewise.
18886 * gimple-ssa-store-merging.c (compatible_load_p): Likewise.
18887 (imm_store_chain_info::try_coalesce_bswap): Likewise.
18888 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
18889 (get_location_for_stmts): Likewise.
18890 * graphite-poly.c (print_iteration_domains): Likewise.
18891 (free_poly_bb): Likewise.
18892 (remove_gbbs_in_scop): Likewise.
18893 (free_scop): Likewise.
18894 (dump_gbb_cases): Likewise.
18895 (dump_gbb_conditions): Likewise.
18896 (print_pdrs): Likewise.
18897 (print_scop): Likewise.
18898 * ifcvt.c (cond_move_process_if_block): Likewise.
18899 * lower-subreg.c (decompose_multiword_subregs): Likewise.
18900 * regcprop.c (pass_cprop_hardreg::execute): Likewise.
18901 * sanopt.c (sanitize_rewrite_addressable_params): Likewise.
18902 * sel-sched-dump.c (dump_insn_vector): Likewise.
18903 * store-motion.c (store_ops_ok): Likewise.
18904 (store_killed_in_insn): Likewise.
18905 * timevar.c (timer::named_items::print): Likewise.
18906 * tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
18907 (cleanup_tree_cfg_noloop): Likewise.
18908 * tree-data-ref.c (dump_data_references): Likewise.
18909 (print_dir_vectors): Likewise.
18910 (print_dist_vectors): Likewise.
18911 (dump_data_dependence_relations): Likewise.
18912 (dump_dist_dir_vectors): Likewise.
18913 (dump_ddrs): Likewise.
18914 (create_runtime_alias_checks): Likewise.
18915 (free_subscripts): Likewise.
18916 (save_dist_v): Likewise.
18917 (save_dir_v): Likewise.
18918 (invariant_access_functions): Likewise.
18919 (same_access_functions): Likewise.
18920 (access_functions_are_affine_or_constant_p): Likewise.
18921 (find_data_references_in_stmt): Likewise.
18922 (graphite_find_data_references_in_stmt): Likewise.
18923 (free_dependence_relations): Likewise.
18924 (free_data_refs): Likewise.
18925 * tree-inline.c (copy_debug_stmts): Likewise.
18926 * tree-into-ssa.c (dump_currdefs): Likewise.
18927 (rewrite_update_phi_arguments): Likewise.
18928 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
18929 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
18931 (vect_slp_analyze_node_dependences): Likewise.
18932 (vect_slp_analyze_instance_dependence): Likewise.
18933 (vect_record_base_alignments): Likewise.
18934 (vect_get_peeling_costs_all_drs): Likewise.
18935 (vect_peeling_supportable): Likewise.
18936 * tree-vectorizer.c (vec_info::~vec_info): Likewise.
18937 (vec_info::free_stmt_vec_infos): Likewise.
18939 2021-06-13 Jeff Law <jeffreyalaw@gmail.com>
18941 * config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
18942 (andqi3_1<cczn>): Removed.
18943 (<ors>qi3_1): Do not split for IOR/XOR a single bit.
18944 (H8/SX bit logicals): Split out from other patterns.
18945 * config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
18946 mulqihi3_const_clobber_flags.
18947 (mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly
18949 2021-06-13 H.J. Lu <hjl.tools@gmail.com>
18952 * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
18953 to true if red zone is used.
18954 (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
18955 ix86_red_zone_used.
18956 * config/i386/i386.h (machine_function): Add red_zone_used.
18957 (ix86_red_zone_size): Removed.
18958 (ix86_red_zone_used): New.
18959 * config/i386/i386.md (peephole2 patterns): Replace
18960 ix86_red_zone_size with ix86_red_zone_used.
18962 2021-06-12 Jason Merrill <jason@redhat.com>
18964 * doc/extend.texi (unused variable attribute): Applies to
18965 structure fields as well.
18967 2021-06-12 Eugene Rozenfeld <erozen@microsoft.com>
18969 * auto-profile.c (read_profile): fix a typo in an error string
18971 2021-06-11 Thomas Schwinge <thomas@codesourcery.com>
18973 * tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
18975 * tree-pretty-print.c (dump_omp_clauses): Update.
18976 (dump_generic_node) <OMP_CLAUSE>: Use it.
18978 2021-06-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
18981 * config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
18982 int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
18983 the polymorphic variants matching code.
18984 (__arm_vld1q_z): Likewise.
18985 (__arm_vld2q): Likewise.
18986 (__arm_vld4q): Likewise.
18987 (__arm_vldrbq_gather_offset): Likewise.
18988 (__arm_vldrbq_gather_offset_z): Likewise.
18990 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
18992 PR tree-optimization/96392
18993 * fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.
18995 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
18997 PR tree-optimization/96392
18998 * fold-const.c (fold_real_zero_addition_p): Take both arguments
18999 of the addition or subtraction, not just the zero. Use this
19000 other argument in tests for signaling NaNs and signed zeros.
19001 (tree_expr_maybe_real_minus_zero_p): New predicate.
19002 * fold-const.h (fold_real_zero_addition_p): Update prototype.
19003 (tree_expr_maybe_real_minus_zero_p): New function prototype.
19004 * match.pd: Update calls to fold_real_zero_addition_p.
19005 Replace HONOR_NANS with tree_expr_maybe_nan_p.
19006 Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
19007 Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
19008 * tree-ssa-reassoc.c (eliminate_using_constants): Update
19009 call to fold_real_zero_addition_p.
19011 2021-06-11 Richard Biener <rguenther@suse.de>
19013 PR tree-optimization/101025
19014 * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
19015 all refs that require dependence checking.
19017 2021-06-11 Richard Biener <rguenther@suse.de>
19019 PR tree-optimization/101028
19020 * tree-vect-slp.c (vect_build_slp_tree_2): When SLP
19021 reassoc discovery fails fatally, mark appropriate lanes
19024 2021-06-11 Richard Biener <rguenther@suse.de>
19026 PR tree-optimization/101026
19027 * tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
19028 have a representative for the associated chain nodes.
19030 2021-06-11 Jakub Jelinek <jakub@redhat.com>
19032 PR rtl-optimization/101008
19033 * simplify-rtx.c (relational_result): New function.
19034 (simplify_logical_relational_operation,
19035 simplify_relational_operation): Use it.
19037 2021-06-11 Jakub Jelinek <jakub@redhat.com>
19040 * config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.
19042 2021-06-11 Uroš Bizjak <ubizjak@gmail.com>
19045 * config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
19046 false if the permutation can be implemented with constant
19047 permutation instruction in wider mode.
19048 (canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
19049 Handle V8QImode and V4HImode.
19051 2021-06-11 Martin Liska <mliska@suse.cz>
19053 PR gcov-profile/100788
19054 * common.opt: Add new option.
19055 * coverage.c (coverage_begin_function): Emit warning instead on
19056 the internal compiler error.
19057 * doc/invoke.texi: Document the option.
19058 * toplev.c (process_options): Enable it by default.
19060 2021-06-11 Richard Biener <rguenther@suse.de>
19062 PR middle-end/101009
19063 * tree-data-ref.c (build_classic_dist_vector_1): Make sure
19064 to set *init_b to true when we encounter a constant equal
19066 (compute_affine_dependence): Also dump the actual DR_REF.
19068 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
19070 PR tree-optimization/100984
19071 * gimple-ssa-evrp.c (ssa_equiv_stack): Use auto_vec for
19072 replacements table.
19073 (ssa_equiv_stack::~ssa_equiv_stack): Remove.
19075 2021-06-11 Kewen Lin <linkw@linux.ibm.com>
19077 * config/rs6000/rs6000.md
19078 (floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
19079 define_insn_and_split.
19081 2021-06-11 Richard Biener <rguenther@suse.de>
19083 * tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
19084 to sort operands of the associative chain.
19086 2021-06-11 Richard Biener <rguenther@suse.de>
19088 * system.h (gcc_stablesort_r): Declare.
19089 * sort.cc (gcc_sort_r): Support stable sort.
19090 (gcc_stablesort_r): Define.
19091 * vec.h (vec<>::stablesort): Add.
19093 2021-06-10 Uroš Bizjak <ubizjak@gmail.com>
19096 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
19097 Handle V2SF mode. Emit SHUFPS to fixup unpack-high for V2SF mode.
19098 (expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
19099 (expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
19100 (expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
19101 (expand_vec_perm_interleave2): Handle 64bit modes.
19102 (expand_vec_perm_even_odd_pack): Handle V8QI mode.
19103 (expand_vec_perm_even_odd_1): Ditto.
19104 (ix86_vectorize_vec_perm_const): Ditto.
19105 * config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
19106 * config/i386/sse.md: ... here.
19107 * config/i386/mmx.md (*vec_interleave_lowv2sf):
19108 New insn_and_split pattern.
19109 (*vec_interleave_highv2sf): Ditto.
19110 (mmx_pshufbv8qi3): New insn pattern.
19111 (*mmx_pblendw): Ditto.
19113 2021-06-10 Peter Bergner <bergner@linux.ibm.com>
19115 * config/rs6000/rs6000-builtin.def (build_pair): New built-in.
19116 (build_acc): Likewise.
19117 * config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
19118 source operands in little-endian mode.
19119 (rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
19120 (mma_init_builtins): Likewise.
19121 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
19122 ordering for the MMA assemble and build source operands.
19123 * doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
19125 (__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
19128 2021-06-10 Jeff Law <jeffreyalaw@gmail.com>
19130 * config/h8300/h8300.c (select_cc_mode): Handle MEM. Use
19132 * config/h8300/extensions.md: Replace _clobber_flags patterns
19135 2021-06-10 Robin Dapp <rdapp@linux.ibm.com>
19137 * config/s390/vector.md (vcond_mask_<mode><mode>): Change to
19138 (vcond_mask_<mode><tointvec>): this.
19140 2021-06-10 Andrew Stubbs <ams@codesourcery.com>
19141 Thomas Schwinge <thomas@codesourcery.com>
19143 * omp-builtins.def (BUILT_IN_GOACC_ENTER_EXIT_DATA): Split into...
19144 (BUILT_IN_GOACC_ENTER_DATA, BUILT_IN_GOACC_EXIT_DATA): ... these.
19145 * gimple.h (enum gf_mask): Split
19146 'GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA' into
19147 'GF_OMP_TARGET_KIND_OACC_ENTER_DATA' and
19148 'GF_OMP_TARGET_KIND_OACC_EXIT_DATA'.
19149 (is_gimple_omp_oacc): Update.
19150 * gimple-pretty-print.c (dump_gimple_omp_target): Likewise.
19151 * gimplify.c (gimplify_omp_target_update): Likewise.
19152 * omp-expand.c (expand_omp_target, build_omp_regions_1)
19153 (omp_make_gimple_edges): Likewise.
19154 * omp-low.c (check_omp_nesting_restrictions, lower_omp_target):
19157 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
19159 * value-query.cc (value_query::value_on_edge): Rename name to
19161 (range_query::range_on_edge): Same.
19162 (range_query::value_of_expr): Same.
19163 (range_query::value_on_edge): Same.
19164 * value-query.h (class value_query): Same.
19165 (class range_query): Same.
19167 2021-06-10 Richard Biener <rguenther@suse.de>
19169 PR tree-optimization/101003
19170 * tree-vect-slp.c (vect_build_slp_tree_2): Appropriately
19171 use the pattern stmt defs when linearizing a chain.
19173 2021-06-10 Jakub Jelinek <jakub@redhat.com>
19176 * ifcvt.c (noce_get_alt_condition, noce_try_abs): Use
19177 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
19179 2021-06-10 Clement Chigot <clement.chigot@atos.net>
19181 * config/rs6000/aix71.h (ASM_CPU_SPEC): Add Power10 directive.
19182 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
19184 2021-06-09 Andrew Pinski <apinski@marvell.com>
19186 PR tree-optimization/100925
19187 * match.pd (a ? CST1 : CST2): Limit transformations
19188 that would produce a negative to integeral types only.
19189 Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.
19191 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
19194 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
19196 * doc/tm.texi: Correctly update.
19198 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
19200 * doc/tm.texi: Correctly update.
19202 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
19205 * doc/tm.texi.in (Trampolines): Add a missing blank line.
19207 2021-06-09 Paul Eggert <eggert@cs.ucla.edu>
19210 * doc/invoke.texi (Code Gen Options); Document that -fno-trampolines
19211 and -ftrampolines work only with Ada.
19212 * doc/tm.texi.in (Trampolines): Likewise.
19213 * doc/tm.texi: Regenerated.
19215 2021-06-09 Carl Love <cel@us.ibm.com>
19217 * config/rs6000/altivec.h (vec_signextll, vec_signexti, vec_signextq):
19218 Add define for new builtins.
19219 * config/rs6000/altivec.md(altivec_vreveti2): Add define_expand.
19220 * config/rs6000/rs6000-builtin.def (VSIGNEXTI, VSIGNEXTLL): Add
19221 overloaded builtin definitions.
19222 (VSIGNEXTSB2W, VSIGNEXTSH2W, VSIGNEXTSB2D, VSIGNEXTSH2D,VSIGNEXTSW2D,
19223 VSIGNEXTSD2Q): Add builtin expansions.
19224 (SIGNEXT): Add P10 overload definition.
19225 * config/rs6000/rs6000-call.c (P9V_BUILTIN_VEC_VSIGNEXTI, P9V_BUILTIN_VEC_VSIGNEXTLL,
19226 P10_BUILTIN_VEC_SIGNEXT): Add overloaded argument definitions.
19227 * config/rs6000/vsx.md (vsx_sign_extend_v2di_v1ti): Add define_insn.
19228 (vsignextend_v2di_v1ti, vsignextend_qi_<mode>, vsignextend_hi_<mode>,
19229 vsignextend_si_v2di)[VIlong]: Add define_expand.
19230 Make define_insn vsx_sign_extend_si_v2di visible.
19231 * doc/extend.texi: Add documentation for the vec_signexti,
19232 vec_signextll builtins and vec_signextq.
19234 2021-06-09 Carl Love <cel@us.ibm.com>
19236 * config/rs6000/rs6000.c (__fixkfti, __fixunskfti, __floattikf,
19237 __floatuntikf): Names changed to __fixkfti_sw, __fixunskfti_sw,
19238 __floattikf_sw, __floatuntikf_sw respectively.
19239 * config/rs6000/rs6000.md (floatti<mode>2, floatunsti<mode>2,
19240 fix_trunc<mode>ti2, fixuns_trunc<mode>ti2): Add
19241 define_insn for mode IEEE 128.
19243 2021-06-09 Carl Love <cel@us.ibm.com>
19245 * config/rs6000/altivec.md (altivec_vslq, altivec_vsrq):
19246 Rename to altivec_vslq_<mode>, altivec_vsrq_<mode>, mode VEC_TI.
19247 * config/rs6000/vector.md (VEC_TI): Was named VSX_TI in vsx.md.
19248 (vashlv1ti3): Change to vashl<mode>3, mode VEC_TI.
19249 (vlshrv1ti3): Change to vlshr<mode>3, mode VEC_TI.
19250 * config/rs6000/vsx.md (VSX_TI): Remove define_mode_iterator. Update
19251 uses of VSX_TI to VEC_TI.
19253 2021-06-09 Carl Love <cel@us.ibm.com>
19255 * config/rs6000/dfp.md (floattitd2, fixtdti2): New define_insns.
19257 2021-06-09 Carl Love <cel@us.ibm.com>
19259 * config/rs6000/altivec.h (vec_dive, vec_mod): Add define for new
19261 * config/rs6000/altivec.md (UNSPEC_VMULEUD, UNSPEC_VMULESD,
19262 UNSPEC_VMULOUD, UNSPEC_VMULOSD): New unspecs.
19263 (altivec_eqv1ti, altivec_gtv1ti, altivec_gtuv1ti, altivec_vmuleud,
19264 altivec_vmuloud, altivec_vmulesd, altivec_vmulosd, altivec_vrlq,
19265 altivec_vrlqmi, altivec_vrlqmi_inst, altivec_vrlqnm,
19266 altivec_vrlqnm_inst, altivec_vslq, altivec_vsrq, altivec_vsraq,
19267 altivec_vcmpequt_p, altivec_vcmpgtst_p, altivec_vcmpgtut_p): New
19269 (vec_widen_umult_even_v2di, vec_widen_smult_even_v2di,
19270 vec_widen_umult_odd_v2di, vec_widen_smult_odd_v2di, altivec_vrlqmi,
19271 altivec_vrlqnm): New define_expands.
19272 * config/rs6000/rs6000-builtin.def (VCMPEQUT_P, VCMPGTST_P,
19273 VCMPGTUT_P): Add macro expansions.
19274 (BU_P10V_AV_P): Add builtin predicate definition.
19275 (VCMPGTUT, VCMPGTST, VCMPEQUT, CMPNET, CMPGE_1TI,
19276 CMPGE_U1TI, CMPLE_1TI, CMPLE_U1TI, VNOR_V1TI_UNS, VNOR_V1TI, VCMPNET_P,
19277 VCMPAET_P, VMULEUD, VMULESD, VMULOUD, VMULOSD, VRLQ,
19278 VSLQ, VSRQ, VSRAQ, VRLQNM, DIV_V1TI, UDIV_V1TI, DIVES_V1TI, DIVEU_V1TI,
19279 MODS_V1TI, MODU_V1TI, VRLQMI): New macro expansions.
19280 (VRLQ, VSLQ, VSRQ, VSRAQ, DIVE, MOD): New overload expansions.
19281 * config/rs6000/rs6000-call.c (P10_BUILTIN_VCMPEQUT,
19282 P10V_BUILTIN_CMPGE_1TI, P10V_BUILTIN_CMPGE_U1TI,
19283 P10V_BUILTIN_VCMPGTUT, P10V_BUILTIN_VCMPGTST,
19284 P10V_BUILTIN_CMPLE_1TI, P10V_BUILTIN_VCMPLE_U1TI,
19285 P10V_BUILTIN_DIV_V1TI, P10V_BUILTIN_UDIV_V1TI,
19286 P10V_BUILTIN_VMULESD, P10V_BUILTIN_VMULEUD,
19287 P10V_BUILTIN_VMULOSD, P10V_BUILTIN_VMULOUD,
19288 P10V_BUILTIN_VNOR_V1TI, P10V_BUILTIN_VNOR_V1TI_UNS,
19289 P10V_BUILTIN_VRLQ, P10V_BUILTIN_VRLQMI,
19290 P10V_BUILTIN_VRLQNM, P10V_BUILTIN_VSLQ,
19291 P10V_BUILTIN_VSRQ, P10V_BUILTIN_VSRAQ,
19292 P10V_BUILTIN_VCMPGTUT_P, P10V_BUILTIN_VCMPGTST_P,
19293 P10V_BUILTIN_VCMPEQUT_P, P10V_BUILTIN_VCMPGTUT_P,
19294 P10V_BUILTIN_VCMPGTST_P, P10V_BUILTIN_CMPNET,
19295 P10V_BUILTIN_VCMPNET_P, P10V_BUILTIN_VCMPAET_P,
19296 P10V_BUILTIN_DIVES_V1TI, P10V_BUILTIN_MODS_V1TI,
19297 P10V_BUILTIN_MODU_V1TI):
19298 New overloaded definitions.
19299 (rs6000_gimple_fold_builtin) [P10V_BUILTIN_VCMPEQUT,
19300 P10V_BUILTIN_CMPNET, P10V_BUILTIN_CMPGE_1TI,
19301 P10V_BUILTIN_CMPGE_U1TI, P10V_BUILTIN_VCMPGTUT,
19302 P10V_BUILTIN_VCMPGTST, P10V_BUILTIN_CMPLE_1TI,
19303 P10V_BUILTIN_CMPLE_U1TI]: New case statements.
19304 (rs6000_init_builtins) [bool_V1TI_type_node, int_ftype_int_v1ti_v1ti]:
19306 (altivec_init_builtins): New E_V1TImode case statement.
19307 (builtin_function_type)[P10_BUILTIN_128BIT_VMULEUD,
19308 P10_BUILTIN_128BIT_VMULOUD, P10_BUILTIN_128BIT_DIVEU_V1TI,
19309 P10_BUILTIN_128BIT_MODU_V1TI, P10_BUILTIN_CMPGE_U1TI,
19310 P10_BUILTIN_VCMPGTUT, P10_BUILTIN_VCMPEQUT]: New case statements.
19311 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute) [E_TImode,
19312 E_V1TImode]: New case statements.
19313 * config/rs6000/rs6000.h (rs6000_builtin_type_index): New enum
19314 value RS6000_BTI_bool_V1TI.
19315 * config/rs6000/vector.md (vector_gtv1ti,vector_nltv1ti,
19316 vector_gtuv1ti, vector_nltuv1ti, vector_ngtv1ti, vector_ngtuv1ti,
19317 vector_eq_v1ti_p, vector_ne_v1ti_p, vector_ae_v1ti_p,
19318 vector_gt_v1ti_p, vector_gtu_v1ti_p, vrotlv1ti3, vashlv1ti3,
19319 vlshrv1ti3, vashrv1ti3): New define_expands.
19320 * config/rs6000/vsx.md (UNSPEC_VSX_DIVSQ, UNSPEC_VSX_DIVUQ,
19321 UNSPEC_VSX_DIVESQ, UNSPEC_VSX_DIVEUQ, UNSPEC_VSX_MODSQ,
19322 UNSPEC_VSX_MODUQ): New unspecs.
19323 (mulv2di3, vsx_div_v1ti, vsx_udiv_v1ti, vsx_dives_v1ti,
19324 vsx_diveu_v1ti, vsx_mods_v1ti, vsx_modu_v1ti, xxswapd_v1ti): New
19326 (vcmpnet): New define_expand.
19327 * doc/extend.texi: Add documentation for the new builtins vec_rl,
19328 vec_rlmi, vec_rlnm, vec_sl, vec_sr, vec_sra, vec_mule, vec_mulo,
19329 vec_div, vec_dive, vec_mod, vec_cmpeq, vec_cmpne, vec_cmpgt, vec_cmplt,
19330 vec_cmpge, vec_cmple, vec_all_eq, vec_all_ne, vec_all_gt, vec_all_lt,
19331 vec_all_ge, vec_all_le, vec_any_eq, vec_any_ne, vec_any_gt, vec_any_lt,
19332 vec_any_ge, vec_any_le.
19334 2021-06-09 Carl Love <cel@us.ibm.com>
19336 * config/rs6000/altivec.md (altivec_vrl<VI_char>mi): Fix
19337 bug in argument generation.
19339 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
19341 * config/arm/iterators.md (<supf>): Remove VCLZQ_U, VCLZQ_S.
19343 * config/arm/mve.md (mve_vclzq_<supf><mode>): Add '@' prefix,
19344 remove <supf> iterator.
19345 (mve_vclzq_u<mode>): New.
19346 * config/arm/neon.md (clz<mode>2): Rename to neon_vclz<mode>.
19347 (neon_vclz<mode): Move to ...
19348 * config/arm/unspecs.md (VCLZQ_U, VCLZQ_S): Remove.
19349 * config/arm/vec-common.md: ... here. Add support for MVE.
19351 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
19353 * config/arm/mve.md (mve_vhaddq_<supf><mode>): Prefix with '@'.
19354 (@mve_vrhaddq_<supf><mode): Likewise.
19355 * config/arm/neon.md (neon_v<r>hadd<sup><mode>): Likewise.
19356 * config/arm/vec-common.md (avg<mode>3_floor, uavg<mode>3_floor)
19357 (avg<mode>3_ceil", uavg<mode>3_ceil): New patterns.
19359 2021-06-09 imba-tjd <109224573@qq.com>
19361 * doc/invoke.texi: Fix typo.
19363 2021-06-09 Roger Sayle <roger@nextmovesoftware.com>
19365 PR middle-end/53267
19366 * fold-const-call.c (fold_const_call_sss) [CASE_CFN_FMOD]:
19367 Support evaluation of fmod/fmodf/fmodl at compile-time.
19369 2021-06-09 Richard Biener <rguenther@suse.de>
19371 PR tree-optimization/100981
19372 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
19373 gimple_get_lhs to also handle calls.
19374 * tree-vect-slp-patterns.c (complex_pattern::build): Transfer
19377 2021-06-09 Richard Biener <rguenther@suse.de>
19379 PR tree-optimization/97832
19380 * tree-vectorizer.h (_slp_tree::failed): New.
19381 * tree-vect-slp.c (_slp_tree::_slp_tree): Initialize
19383 (_slp_tree::~_slp_tree): Free failed.
19384 (vect_build_slp_tree): Retain failed nodes and record
19385 matches in them, copying that back out when running
19386 into a cached fail. Dump start and end of discovery.
19387 (dt_sort_cmp): New.
19388 (vect_build_slp_tree_2): Handle associatable chains
19389 together doing more aggressive operand swapping.
19391 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
19394 * config.gcc (gcc_cv_initfini_array): Set to yes for Linux and
19396 * doc/install.texi: Require glibc 2.1 and binutils 2.12 for
19397 Linux and GNU targets.
19399 2021-06-09 Richard Biener <rguenther@suse.de>
19401 * tree-vect-stmts.c (vect_is_simple_use): Always get dt
19404 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
19406 * config/arc/arc.md (loop_end): Change it to
19407 define_insn_and_split.
19409 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
19411 * config/arc/arc.md (maddhisi4): Use VMAC2H instruction.
19412 (machi): New pattern.
19413 (umaddhisi4): Use VMAC2HU instruction.
19414 (umachi): New pattern.
19416 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
19418 * config/arc/arc-protos.h (arc_split_move_p): New prototype.
19419 * config/arc/arc.c (arc_split_move_p): New function.
19420 (arc_split_move): Clean up.
19421 * config/arc/arc.md (movdi_insn): Clean up, use arc_split_move_p.
19422 (movdf_insn): Likewise.
19423 * config/arc/simdext.md (mov<VWH>_insn): Likewise.
19425 2021-06-09 Uroš Bizjak <ubizjak@gmail.com>
19428 * config/i386/i386.c (print_operand_address_as): Rename "no_rip"
19429 argument to "raw". Do not emit segment overrides when "raw" is true.
19431 2021-06-09 Martin Liska <mliska@suse.cz>
19433 * doc/gcov.texi: Create a proper JSON files.
19434 * doc/invoke.texi: Remove dots in order to make it a valid
19437 2021-06-09 Xionghu Luo <luoxhu@linux.ibm.com>
19439 * config/rs6000/rs6000-p8swap.c (pattern_is_rotate64): New.
19440 (insn_is_load_p): Use pattern_is_rotate64.
19441 (insn_is_swap_p): Likewise.
19442 (quad_aligned_load_p): Likewise.
19443 (const_load_sequence_p): Likewise.
19444 (replace_swapped_aligned_load): Likewise.
19445 (recombine_lvx_pattern): Likewise.
19446 (recombine_stvx_pattern): Likewise.
19448 2021-06-09 Andrew MacLeod <amacleod@redhat.com>
19450 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Use a
19451 fur_stmt source record.
19452 * gimple-range.cc (fur_source::get_operand): Generic range query.
19453 (fur_source::get_phi_operand): New.
19454 (fur_source::register_dependency): New.
19455 (fur_source::query): New.
19456 (class fur_edge): New. Edge source for operands.
19457 (fur_edge::fur_edge): New.
19458 (fur_edge::get_operand): New.
19459 (fur_edge::get_phi_operand): New.
19460 (fur_edge::query): New.
19461 (fur_stmt::fur_stmt): New.
19462 (fur_stmt::get_operand): New.
19463 (fur_stmt::get_phi_operand): New.
19464 (fur_stmt::query): New.
19465 (class fur_depend): New. Statement source and process dependencies.
19466 (fur_depend::fur_depend): New.
19467 (fur_depend::register_dependency): New.
19468 (class fur_list): New. List source for operands.
19469 (fur_list::fur_list): New.
19470 (fur_list::get_operand): New.
19471 (fur_list::get_phi_operand): New.
19472 (fold_range): New. Instantiate appropriate fur_source class and fold.
19473 (fold_using_range::range_of_range_op): Use new API.
19474 (fold_using_range::range_of_address): Ditto.
19475 (fold_using_range::range_of_phi): Ditto.
19476 (imple_ranger::fold_range_internal): Use fur_depend class.
19477 (fold_using_range::range_of_ssa_name_with_loop_info): Use new API.
19478 * gimple-range.h (class fur_source): Now a base class.
19479 (class fur_stmt): New.
19480 (fold_range): New prototypes.
19481 (fur_source::fur_source): Delete.
19483 2021-06-08 Andrew Pinski <apinski@marvell.com>
19485 PR tree-optimization/25290
19486 * tree-ssa-phiopt.c (xor_replacement): Delete.
19487 (tree_ssa_phiopt_worker): Delete use of xor_replacement.
19488 (match_simplify_replacement): Allow one cheap preparation
19489 statement that can be moved to before the if.
19491 2021-06-08 Pat Haugen <pthaugen@linux.ibm.com>
19493 * config/rs6000/power10.md (power10-fused-load, power10-fused-store,
19494 power10-fused_alu, power10-fused-vec, power10-fused-branch): New.
19496 2021-06-08 Jeff Law <jeffreyalaw@gmail.com>
19498 * config/h8300/logical.md (andqi3_1): Move BCLR case into define_insn_and_split.
19499 Create length attribute on define_insn_and_split. Only split for cases which we
19501 (andqi3_1<cczn>): Renamed from andqi3_1_clobber_flags. Only handle AND here and
19502 fix length computation.
19503 (b<code><mode>msx): Combine QImode and HImode H8/SX patterns using iterator.
19505 2021-06-08 Richard Biener <rguenther@suse.de>
19507 PR tree-optimization/100923
19508 * tree-ssa-sccvn.c (valueize_refs_1): Take a pointer to
19509 the operand vector to be valueized.
19510 (valueize_refs): Likewise.
19511 (valueize_shared_reference_ops_from_ref): Adjust.
19512 (valueize_shared_reference_ops_from_call): Likewise.
19513 (vn_reference_lookup_3): Likewise.
19514 (vn_reference_lookup_pieces): Likewise. Re-valueize
19515 with honoring availability when we are about to create
19516 the ao_ref and valueized before.
19517 (vn_reference_lookup): Likewise.
19518 (vn_reference_insert_pieces): Adjust.
19520 2021-06-08 Richard Biener <rguenther@suse.de>
19522 * tree-vectorizer.h (_slp_instance::root_stmt): Change to...
19523 (_slp_instance::root_stmts): ... a vector.
19524 (SLP_INSTANCE_ROOT_STMT): Rename to ...
19525 (SLP_INSTANCE_ROOT_STMTS): ... this.
19526 (slp_root::root): Change to...
19527 (slp_root::roots): ... a vector.
19528 (slp_root::slp_root): Adjust.
19529 * tree-vect-slp.c (_slp_instance::location): Adjust.
19530 (vect_free_slp_instance): Release the root stmt vector.
19531 (vect_build_slp_instance): Adjust.
19532 (vect_analyze_slp): Likewise.
19533 (_bb_vec_info::~_bb_vec_info): Likewise.
19534 (vect_slp_analyze_operations): Likewise.
19535 (vect_bb_vectorization_profitable_p): Likewise. Adjust
19536 costs for the root stmt.
19537 (vect_slp_check_for_constructors): Gather all BIT_INSERT_EXPRs
19539 (vect_slp_analyze_bb_1): Simplify by marking all root stmts
19541 (vectorize_slp_instance_root_stmt): Adjust.
19542 (vect_schedule_slp): Likewise.
19544 2021-06-08 Aldy Hernandez <aldyh@redhat.com>
19546 * gimple-ssa-evrp.c (class ssa_equiv_stack): New.
19547 (ssa_equiv_stack::ssa_equiv_stack): New.
19548 (ssa_equiv_stack::~ssa_equiv_stack): New.
19549 (ssa_equiv_stack::enter): New.
19550 (ssa_equiv_stack::leave): New.
19551 (ssa_equiv_stack::push_replacement): New.
19552 (ssa_equiv_stack::get_replacement): New.
19553 (is_pointer_ssa): New.
19554 (class pointer_equiv_analyzer): New.
19555 (pointer_equiv_analyzer::pointer_equiv_analyzer): New.
19556 (pointer_equiv_analyzer::~pointer_equiv_analyzer): New.
19557 (pointer_equiv_analyzer::set_global_equiv): New.
19558 (pointer_equiv_analyzer::set_cond_equiv): New.
19559 (pointer_equiv_analyzer::get_equiv): New.
19560 (pointer_equiv_analyzer::enter): New.
19561 (pointer_equiv_analyzer::leave): New.
19562 (pointer_equiv_analyzer::get_equiv_expr): New.
19563 (pta_valueize): New.
19564 (pointer_equiv_analyzer::visit_stmt): New.
19565 (pointer_equiv_analyzer::visit_edge): New.
19566 (hybrid_folder::value_of_expr): Call PTA.
19567 (hybrid_folder::value_on_edge): Same.
19568 (hybrid_folder::pre_fold_bb): New.
19569 (hybrid_folder::post_fold_bb): New.
19570 (hybrid_folder::pre_fold_stmt): New.
19571 (rvrp_folder::pre_fold_bb): New.
19572 (rvrp_folder::post_fold_bb): New.
19573 (rvrp_folder::pre_fold_stmt): New.
19574 (rvrp_folder::value_of_expr): Call PTA.
19575 (rvrp_folder::value_on_edge): Same.
19577 2021-06-08 Jakub Jelinek <jakub@redhat.com>
19580 * tree-inline.c (copy_tree_body_r): For OMP_CLAUSE_DEPEND don't
19581 check TREE_CODE if OMP_CLAUSE_DECL is NULL.
19583 2021-06-08 Richard Biener <rguenther@suse.de>
19585 PR middle-end/100951
19586 * tree-vect-generic.c (expand_vector_piecewise): Build a
19587 VECTOR_CST if all elements are constant.
19588 (expand_vector_condition): Likewise.
19589 (lower_vec_perm): Likewise.
19590 (expand_vector_conversion): Likewise.
19592 2021-06-08 Martin Liska <mliska@suse.cz>
19594 * doc/invoke.texi: Document new param evrp-sparse-threshold.
19596 2021-06-08 Martin Liska <mliska@suse.cz>
19598 * genautomata.c (create_automata): Fix typo.
19600 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
19602 PR tree-optimization/100794
19603 * tree-predcom.c (tree_predictive_commoning_loop): Add parameter
19604 allow_unroll_p and only allow unrolling when it's true.
19605 (tree_predictive_commoning): Add parameter allow_unroll_p and
19607 (run_tree_predictive_commoning): Likewise.
19608 (pass_predcom::gate): Check flag_tree_loop_vectorize and
19609 global_options_set.x_flag_predictive_commoning.
19610 (pass_predcom::execute): Adjust for allow_unroll_p.
19612 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
19614 * tree-predcom.c (execute_pred_commoning): Remove update_ssa call.
19615 (tree_predictive_commoning_loop): Factor some cleanup stuffs into
19616 lambda function cleanup, remove scev_reset call, and adjust return
19618 (tree_predictive_commoning): Adjust for different changed values,
19619 only set flag TODO_update_ssa_only_virtuals if changed.
19620 (pass_data pass_data_predcom): Remove TODO_update_ssa_only_virtuals
19621 from todo_flags_finish.
19623 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
19625 * gimple-range-cache.cc (class sbr_sparse_bitmap): New.
19626 (sbr_sparse_bitmap::sbr_sparse_bitmap): New.
19627 (sbr_sparse_bitmap::bitmap_set_quad): New.
19628 (sbr_sparse_bitmap::bitmap_get_quad): New.
19629 (sbr_sparse_bitmap::set_bb_range): New.
19630 (sbr_sparse_bitmap::get_bb_range): New.
19631 (sbr_sparse_bitmap::bb_range_p): New.
19632 (block_range_cache::block_range_cache): initialize bitmap obstack.
19633 (block_range_cache::~block_range_cache): Destruct obstack.
19634 (block_range_cache::set_bb_range): Decide when to utilze the
19635 sparse on entry cache.
19636 * gimple-range-cache.h (block_range_cache): Add bitmap obstack.
19637 * params.opt (-param=evrp-sparse-threshold): New.
19639 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
19641 * bitmap.c (bitmap_set_aligned_chunk): New.
19642 (bitmap_get_aligned_chunk): New.
19643 (test_aligned_chunk): New.
19644 (bitmap_c_tests): Call test_aligned_chunk.
19645 * bitmap.h (bitmap_set_aligned_chunk, bitmap_get_aligned_chunk): New.
19647 2021-06-07 Uroš Bizjak <ubizjak@gmail.com>
19650 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
19652 (ix86_expand_vector_init_one_nonzero): Ditto.
19653 (ix86_expand_vector_init_one_var): Ditto.
19654 (ix86_expand_vector_init_general): Ditto.
19655 * config/i386/mmx.md (vec_initv4qiqi): New expander.
19657 2021-06-07 Jeff Law <jeffreyalaw@gmail.com>
19659 * config/h8300/movepush.md: Change most _clobber_flags
19660 patterns to instead use <cczn> subst.
19661 (movsi_cczn): New pattern with usable CC cases split out.
19662 (movsi_h8sx_cczn): Likewise.
19664 2021-06-07 Martin Liska <mliska@suse.cz>
19666 * common/common-target.def: Split long lines and replace them
19668 * target.def: Likewise.
19669 * doc/tm.texi: Re-generated.
19671 2021-06-07 Jakub Jelinek <jakub@redhat.com>
19674 * fold-const.c (fold_read_from_vector): Return NULL if trying to
19675 read from a CONSTRUCTOR with vector type elements.
19677 2021-06-07 Jakub Jelinek <jakub@redhat.com>
19679 PR middle-end/100898
19680 * tree-inline.c (copy_bb): Only use gimple_call_arg_ptr if memcpy
19681 should copy any arguments. Don't call gimple_call_num_args
19682 on id->call_stmt or call_stmt more than once.
19684 2021-06-07 liuhongt <hongtao.liu@intel.com>
19687 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): Refine
19689 (<insn>v4siv4di2): Delete constraints for define_expand.
19691 2021-06-07 liuhongt <hongtao.liu@intel.com>
19694 * config/i386/i386-expand.c (ix86_expand_builtin): Remove
19695 assignment of cfun->machine->has_explicit_vzeroupper.
19696 * config/i386/i386-features.c
19697 (ix86_add_reg_usage_to_vzerouppers): Delete.
19698 (ix86_add_reg_usage_to_vzeroupper): Ditto.
19699 (rest_of_handle_insert_vzeroupper): Remove
19700 ix86_add_reg_usage_to_vzerouppers, add df_analyze at the end
19702 (gate): Remove cfun->machine->has_explicit_vzeroupper.
19703 * config/i386/i386-protos.h (ix86_expand_avx_vzeroupper):
19705 * config/i386/i386.c (ix86_insn_callee_abi): New function.
19706 (ix86_initialize_callee_abi): Ditto.
19707 (ix86_expand_avx_vzeroupper): Ditto.
19708 (ix86_hard_regno_call_part_clobbered): Adjust for vzeroupper
19710 (TARGET_INSN_CALLEE_ABI): Define as ix86_insn_callee_abi.
19711 (ix86_emit_mode_set): Call ix86_expand_avx_vzeroupper
19713 * config/i386/i386.h (struct GTY(()) machine_function): Delete
19714 has_explicit_vzeroupper.
19715 * config/i386/i386.md (enum unspec): New member
19717 (ABI_DEFAULT,ABI_VZEROUPPER,ABI_UNKNOWN): New
19718 define_constants for insn callee abi index.
19719 * config/i386/predicates.md (vzeroupper_pattern): Adjust.
19720 * config/i386/sse.md (UNSPECV_VZEROUPPER): Deleted.
19721 (avx_vzeroupper): Call ix86_expand_avx_vzeroupper.
19722 (*avx_vzeroupper): Rename to ..
19723 (avx_vzeroupper_callee_abi): .. this, and adjust pattern as
19724 call_insn which has a special vzeroupper ABI.
19725 (*avx_vzeroupper_1): Deleted.
19727 2021-06-07 liuhongt <hongtao.liu@intel.com>
19730 * df-scan.c (df_get_call_refs): When call_insn is a fake call,
19731 it won't use stack pointer reg.
19732 * final.c (leaf_function_p): When call_insn is a fake call, it
19733 won't affect caller as a leaf function.
19734 * reg-stack.c (callee_clobbers_any_stack_reg): New.
19735 (subst_stack_regs): When call_insn doesn't clobber any stack
19736 reg, don't clear the arguments.
19737 * rtl.c (shallow_copy_rtx): Don't clear flag used when orig is
19739 * shrink-wrap.c (requires_stack_frame_p): No need for stack
19740 frame for a fake call.
19741 * rtl.h (FAKE_CALL_P): New macro.
19743 2021-06-06 Eric Botcazou <ebotcazou@adacore.com>
19745 * config/sparc/sparc-protos.h (order_regs_for_local_alloc): Rename
19747 (sparc_order_regs_for_local_alloc): ...this.
19748 (sparc_leaf_reg_remap): Declare.
19749 * config/sparc/sparc.h (ADJUST_REG_ALLOC_ORDER): Adjust.
19750 (LEAF_REG_REMAP): Reimplement as call to sparc_leaf_reg_remap.
19751 * config/sparc/sparc.c (leaf_reg_remap): Delete.
19752 (order_regs_for_local_alloc): Rename to...
19753 (sparc_order_regs_for_local_alloc): ...this.
19754 (sparc_leaf_reg_remap): New function.
19755 (sparc_conditional_register_usage): Do not modify leaf_reg_remap.
19757 2021-06-06 David Edelsohn <dje.gcc@gmail.com>
19759 * config/rs6000/rs6000.c (rs6000_xcoff_asm_output_aligned_decl_common):
19760 Use assemble_name to output BSS section name.
19762 2021-06-06 Uroš Bizjak <ubizjak@gmail.com>
19764 * config/i386/constraints.md (Bs):
19765 Remove boolean operators from match_test RTX.
19768 (M): Use "mode" variable instead of GET_MODE (op) in match_test RTX.
19771 2021-06-06 Martin Liska <mliska@suse.cz>
19773 * doc/extend.texi: Add missing @headitem.
19774 * doc/invoke.texi: Likewise.
19775 * doc/objc.texi: Likewise.
19777 2021-06-06 Martin Liska <mliska@suse.cz>
19779 * genhooks.c (emit_findices): Remove unused function.
19780 (emit_documentation): Do not call emit_findices
19781 and do not search for @Fcode directives.
19783 2021-06-06 Martin Liska <mliska@suse.cz>
19785 * doc/invoke.texi: Remove extra character.
19787 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
19789 * config/sh/sh.md (doloop_end_split): Fix empty split condition.
19791 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
19793 * config/sparc/sparc.md (*snedi<W:mode>_zero_vis3,
19794 *neg_snedi<W:mode>_zero_subxc, *plus_snedi<W:mode>_zero,
19795 *plus_plus_snedi<W:mode>_zero, *minus_snedi<W:mode>_zero,
19796 *minus_minus_snedi<W:mode>_zero): Fix empty split condition.
19798 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
19800 * config/or1k/or1k.md (*movdi): Fix empty split condition.
19802 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
19804 * config/mips/mips.md (<anonymous>, bswapsi2, bswapdi2): Fix empty
19807 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
19809 * config/m68k/m68k.md (*zero_extend_inc, *zero_extend_dec,
19810 *zero_extendsidi2): Fix empty split condition.
19812 2021-06-05 Jeff Law <jeffreyalaw@gmail.com>
19814 * config/h8300/addsub.md: Fix split condition in define_insn_and_split
19816 * config/h8300/bitfield.md: Likewise.
19817 * config/h8300/combiner.md: Likewise.
19818 * config/h8300/divmod.md: Likewise.
19819 * config/h8300/extensions.md: Likewise.
19820 * config/h8300/jumpcall.md: Likewise.
19821 * config/h8300/movepush.md: Likewise.
19822 * config/h8300/multiply.md: Likewise.
19823 * config/h8300/other.md: Likewise.
19824 * config/h8300/shiftrotate.md: Likewise.
19825 * config/h8300/logical.md: Likewise. Fix split pattern to use
19826 code iterator that somehow slipped through.
19828 2021-06-04 Tobias Burnus <tobias@codesourcery.com>
19830 PR middle-end/100905
19831 * tree-nested.c (convert_nonlocal_omp_clauses,
19832 convert_local_omp_clauses): Handle OMP_CLAUSE_BIND.
19834 2021-06-04 Martin Sebor <msebor@redhat.com>
19836 PR middle-end/100732
19837 * gimple-fold.c (gimple_fold_builtin_sprintf): Avoid folding calls
19838 with either source or destination argument of invalid type.
19839 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Avoid checking
19840 calls with arguments of invalid type.
19842 2021-06-04 Martin Sebor <msebor@redhat.com>
19844 * attribs.c (init_attr_rdwr_indices): Use VLA bounds in the expected
19846 (attr_access::vla_bounds): Also handle VLA bounds.
19848 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
19850 * config/i386/predicates.md (GOT_memory_operand):
19851 Implement using match_code RTXes.
19852 (GOT32_symbol_operand): Ditto.
19854 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
19857 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
19859 (ix86_expand_vector_init_general): Ditto.
19860 Use SImode instead of word_mode for logic operations
19861 when GET_MODE_SIZE (mode) < UNITS_PER_WORD.
19862 (expand_vec_perm_even_odd_1): Assert that V2HI mode should be
19863 implemented by expand_vec_perm_1.
19864 (expand_vec_perm_broadcast_1): Assert that V2HI and V4HI modes
19865 should be implemented using standard shuffle patterns.
19866 (ix86_vectorize_vec_perm_const): Handle V2HImode. Add V4HI and
19867 V2HI modes to modes, implementable with shuffle for one operand.
19868 * config/i386/mmx.md (*punpckwd): New insn_and_split pattern.
19869 (*pshufw_1): New insn pattern.
19870 (*vec_dupv2hi): Ditto.
19871 (vec_initv2hihi): New expander.
19873 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
19875 * config/arm/vfp.md (no_literal_pool_df_immediate,
19876 no_literal_pool_sf_immediate): Fix empty split condition.
19878 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
19880 * config/i386/i386.md (*load_tp_x32_zext, *add_tp_x32_zext,
19881 *tls_dynamic_gnu2_combine_32): Fix empty split condition.
19882 * config/i386/sse.md (*<sse2_avx2>_pmovmskb_lt,
19883 *<sse2_avx2>_pmovmskb_zext_lt, *sse2_pmovmskb_ext_lt,
19884 *<sse4_1_avx2>_pblendvb_lt): Likewise.
19886 2021-06-04 Jakub Jelinek <jakub@redhat.com>
19889 * config/i386/i386-expand.c (ix86_expand_vector_init): Handle
19890 concatenation from half-sized modes with TImode elements.
19892 2021-06-04 Claudiu Zissulescu <claziss@synopsys.com>
19894 * config/arc/arc.c (arc_override_options): Disable millicode
19895 thunks when RF16 is on.
19897 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
19899 * config/rs6000/rs6000.h (PROMOTE_MODE): Remove.
19901 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
19903 * config/rs6000/rs6000-call.c (rs6000_promote_function_mode):
19904 Replace PROMOTE_MODE marco with its content.
19906 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
19908 * config/cris/cris.md (*addi_reload): Fix empty split condition.
19910 2021-06-03 Jim Wilson <jimw@sifive.com>
19912 * config.gcc (riscv*-*-*): If --with-riscv-attribute not used,
19913 turn it on for all riscv targets.
19915 2021-06-03 Uroš Bizjak <ubizjak@gmail.com>
19918 * config/i386/i386-expand.c (ix86_expand_vector_set):
19919 Handle V2HI and V4QI modes.
19920 (ix86_expand_vector_extract): Ditto.
19921 * config/i386/mmx.md (*pinsrw): New insn pattern.
19924 (*pextrw_zext): Ditto.
19926 (*pextrb_zext): Ditto.
19927 (vec_setv2hi): New expander.
19928 (vec_extractv2hihi): Ditto.
19929 (vec_setv4qi): Ditto.
19930 (vec_extractv4qiqi): Ditto.
19931 (vec_setv8qi): Enable only for TARGET_SSE4_1.
19932 (vec_extractv8qiqi): Ditto.
19934 2021-06-03 Aaron Sawdey <acsawdey@linux.ibm.com>
19936 * config/rs6000/genfusion.pl (gen_logical_addsubf): Fix input
19937 order to subf instruction.
19938 * config/rs6000/fusion.md: Regenerate.
19940 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
19942 * calls.c (get_size_range): Use range_of_expr instead of
19943 determine_value_range.
19944 * tree-affine.c (expr_to_aff_combination): Same.
19945 * tree-data-ref.c (split_constant_offset): Same.
19946 * tree-vrp.c (determine_value_range_1): Remove.
19947 (determine_value_range): Remove.
19948 * tree-vrp.h (determine_value_range): Remove.
19950 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
19952 * function-tests.c (test_ranges): Call gimple_range_tests.
19953 * gimple-range-cache.cc (ranger_cache::range_of_expr): Pass stmt
19955 * gimple-range.cc (fur_source::get_operand): Do not call
19956 get_tree_range or gimple_range_global.
19958 (get_tree_range): Move to value-query.cc.
19959 Call get_arith_expr_range.
19960 (gimple_ranger::range_of_expr): Add argument to get_tree_range.
19961 Include gimple-range-tests.cc.
19962 * gimple-range.h (fold_range): Add argument.
19963 (get_tree_range): Remove.
19964 * selftest.h (gimple_range_tests): New.
19965 * value-query.cc (global_range_query::range_of_expr): Add
19967 (range_query::get_tree_range): Move from gimple-range.cc.
19968 * value-query.h (class range_query): Add get_tree_range and
19969 get_arith_expr_range. Make fur_source a friend.
19970 * vr-values.c (vr_values::range_of_expr): Pass stmt to
19972 * gimple-range-tests.cc: New file.
19974 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
19976 * gimple-range.cc (gimple_ranger::export_global_ranges): Call
19977 update_global_range.
19978 * value-query.cc (update_global_range): New.
19979 * value-query.h (update_global_range): New.
19981 2021-06-03 David Malcolm <dmalcolm@redhat.com>
19983 * diagnostic-show-locus.c (diagnostic_show_locus): Don't reject
19984 printing the same location twice if there are fix-it hints,
19985 multiple locations, or a label.
19987 2021-06-03 Andre Vieira <andre.simoesdiasvieira@arm.com>
19989 * tree-vect-loop.c (vect_transform_loop): Use main loop's various'
19990 thresholds to narrow the upper bound on epilogue iterations.
19992 2021-06-03 Christophe Lyon <christophe.lyon@linaro.org>
19994 * config/arm/mve.md (mve_vabsq_f<mode>): Use 'abs' instead of unspec.
19995 (mve_vabsq_s<mode>): Likewise.
19996 * config/arm/neon.md (abs<mode>2): Rename to neon_abs<mode>2.
19997 * config/arm/unspecs.md (VABSQ_F, VABSQ_S): Delete.
19998 * config/arm/vec-common.md (neg<mode>2): Rename to
19999 <absneg_str><mode>2.
20001 2021-06-03 Claudiu Zissulescu <claziss@synopsys.com>
20003 * common/config/arc/arc-common.c (arc_option_optimization_table):
20004 Remove malign-call.
20005 * config/arc/arc.c (arc_unalign_branch_p): Remove unused function.
20006 * config/arc/arc.h (TARGET_MIXED_CODE): Remove macro.
20007 (INDEX_REG_CLASS): Only refer to GENERAL_REGS.
20008 * config/arc/arc.md (abssi2_mixed): Remove pattern.
20009 * config/arc/arc.opt (munalign-prob-threshold): Mark it obsolete.
20010 (malign-call): Likewise.
20011 (mmixed-code): Likewise.
20012 * doc/invoke.texi (ARC): Update doc.
20014 2021-06-03 Martin Liska <mliska@suse.cz>
20016 * common.opt: Use proper Enum values.
20017 * opts.c (COVERAGE_SANITIZER_OPT): Remove.
20018 (parse_sanitizer_options): Handle only sanitizer_opts.
20019 (common_handle_option): Just assign value.
20021 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
20024 * tree-inline.c (inline_forbidden_p): Remove test on return type.
20026 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
20028 * dwarf2out.c (loc_list_from_tree_1) <FUNCTION_DECL>: Also generate
20029 DW_OP_GNU_variable_value referencing an existing DIE at file scope.
20030 (type_byte_size): Inline into...
20031 (add_byte_size_attribute): ...this and call add_scalar_info.
20033 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
20035 * dwarf2out.c (mem_loc_descriptor) <UDIV>: Fix typo.
20036 (typed_binop_from_tree): New function.
20037 (loc_list_from_tree_1) <EXACT_DIV_EXPR>: For an unsigned type,
20038 turn a divide by a power of 2 into a shift.
20039 <CEIL_DIV_EXPR>: For an unsigned type, use a signed divide if the
20040 size of the mode is lower than DWARF2_ADDR_SIZE; otherwise, do a
20041 typed divide by calling typed_binop_from_tree.
20043 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
20045 * dwarf2out.c (scompare_loc_descriptor): Fix head comment.
20046 (is_handled_procedure_type): Likewise.
20047 (struct loc_descr_context): Add strict_signedness field.
20048 (resolve_args_picking_1): Deal with DW_OP_[GNU_]deref_type,
20049 DW_OP_[GNU_]convert and DW_OP_[GNU_]reinterpret.
20050 (resolve_args_picking): Minor tweak.
20051 (function_to_dwarf_procedure): Initialize strict_signedness field.
20052 (type_byte_size): Likewise.
20053 (field_byte_offset): Likewise.
20054 (gen_descr_array_type_die): Likewise.
20055 (gen_variant_part): Likewise.
20056 (loc_list_from_tree_1) <CALL_EXPR>: Tidy up and set strict_signedness
20057 to true when a context is present before evaluating the arguments.
20058 <COND_EXPR>: Do not generate a useless comparison with zero.
20059 When dereferencing an address, if strict_signedness is true and the
20060 type is small and signed, use DW_OP_deref_type to do the dereference
20061 and then DW_OP_convert to convert back to the generic type.
20063 2021-06-03 Jakub Jelinek <jakub@redhat.com>
20066 * tree-inline.c (copy_tree_body_r): Handle iterators on
20067 OMP_CLAUSE_AFFINITY or OMP_CLAUSE_DEPEND.
20069 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
20071 * config/arc/arc.md (*bbit_di): Remove.
20073 2021-06-02 Christoph Muellner <cmuellner@gcc.gnu.org>
20075 PR rtl-optimization/100264
20076 * ree.c (get_sub_rtx): Ignore SET expressions without register
20077 destinations and remove assertion, as it is not valid anymore
20078 with this new behaviour.
20079 (merge_def_and_ext): Eliminate destination check for register
20080 as such SET expressions can't occur anymore.
20081 (combine_reaching_defs): Likewise.
20083 2021-06-02 Jakub Jelinek <jakub@redhat.com>
20086 * config/xtensa/xtensa.h (LEAF_REG_REMAP): Cast REGNO to int to avoid
20087 -Wtype-limits warnings.
20088 (DWARF_FRAME_REGISTER): Rewrite into ternary operator with addition
20089 in operands to avoid -Wsign-compare warnings.
20091 2021-06-02 Pat Haugen <pthaugen@linux.ibm.com>
20093 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): Use
20096 2021-06-02 Vineet Gupta <vgupta@synopsys.com>
20098 * config/arc/arc.h (TARGET_CPU_DEFAULT): Change to hs38_linux.
20100 2021-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
20102 * config/s390/s390.md(*ashrdi3_31<setcc><cconly>): Use a single
20104 * config/s390/subst.md(cconly_subst): Use a single constraint
20105 in (match_scratch).
20107 2021-06-02 Martin Liska <mliska@suse.cz>
20109 * ipa-icf.h: Use auto_vec for memory_access_types.
20111 2021-06-02 Jeff Law <jeffreyalaw@gmail.com>
20113 * config/h8300/h8300-protos.h (compute_a_shift_length): Drop unused
20114 argument from prototype.
20115 (output_logical_op): Add rtx_code argument.
20116 (compute_logical_op_length): Likewise.
20117 * config/h8300/h8300.c (h8300_and_costs): Pass additional argument
20118 to compute_a_shift_length.
20119 (output_logical_op); New argument with the rtx code rather than
20120 extracting it from an operand. Handle QImode too.
20121 (compute_logical_op_length): Similary.
20122 (compute_a_shift_length): Drop unused argument.
20123 * config/h8300/h8300.md (logicals): New code iterator.
20124 * config/h8300/logical.md (<code><mode>3 expander): Combine
20125 the "and" expander with the "ior"/"xor" expander.
20126 (bclr<mode>msx): Combine the QI/HI mode patterns.
20127 (<logical><mode>3 insns): Use code iterator rather than match_operator.
20128 Handle QImode as well. Update call to output_logical_op and
20129 compute_logical_op_length to pass in rtx_code
20130 Fix split condition on all define_insn_and_split patterns.
20131 (one_cmpl<mode>2<cczn>): Use <cczn> to support both clobbering
20132 the flags and setting ZN via existing define_subst.
20133 * config/h8300/shiftrotate.md: Drop unused argument from
20134 calls to compute_a_shift_length.
20135 Signed-off-by: Jeff Law <jeffreyalaw@gmail.com>
20137 2021-06-01 Andrew Pinski <apinski@marvell.com>
20139 PR tree-optimization/25290
20140 * tree-ssa-phiopt.c (match_simplify_replacement):
20142 (tree_ssa_phiopt_worker): Use match_simplify_replacement.
20143 (two_value_replacement): Change the comment about
20144 conditional_replacement.
20145 (conditional_replacement): Delete.
20147 2021-06-01 Andrew Pinski <apinski@marvell.com>
20149 PR tree-optimization/95481
20150 * tree-tailcall.c (find_tail_calls): Handle empty typed
20153 2021-06-01 Andrew Pinski <apinski@marvell.com>
20155 * gimplify.c (zero_sized_field_decl): Delete
20156 (zero_sized_type): Delete
20157 (gimplify_init_ctor_eval): Use is_empty_type instead
20158 of zero_sized_field_decl.
20159 (gimplify_modify_expr): Use is_empty_type instead of
20162 2021-06-01 Jason Merrill <jason@redhat.com>
20165 * tree.h (CALL_FROM_NEW_OR_DELETE_P): Adjust comment.
20167 2021-06-01 Jason Merrill <jason@redhat.com>
20170 * diagnostic.h (warning_enabled_at): Declare.
20171 * diagnostic.c (diagnostic_enabled): Factor out from...
20172 (diagnostic_report_diagnostic): ...here.
20173 (warning_enabled_at): New.
20175 2021-06-01 Aldy Hernandez <aldyh@redhat.com>
20177 * gimple-ssa-evrp.c: Enable exporting of global ranges.
20179 2021-06-01 Martin Liska <mliska@suse.cz>
20182 * doc/invoke.texi: Mention that -fgcse-after-reload
20183 is enabled with -O3.
20185 2021-06-01 liuhongt <hongtao.liu@intel.com>
20187 PR tree-optimization/98365
20188 * tree-if-conv.c (strip_nop_cond_scalar_reduction): New function.
20189 (is_cond_scalar_reduction): Handle nop_expr in cond scalar reduction.
20190 (convert_scalar_cond_reduction): Ditto.
20191 (predicate_scalar_phi): Ditto.
20193 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
20195 PR tree-optimization/100781
20196 * gimple-range-cache.cc (ranger_cache::ranger_cache): Enable new
20197 value calculation by default.
20198 (ranger_cache::enable_new_values): New.
20199 (ranger_cache::disable_new_values): New.
20200 (ranger_cache::push_poor_value): Check if new values are allowed.
20201 * gimple-range-cache.h (class ranger_cache): New member/methods.
20202 * gimple-range.cc (gimple_ranger::range_of_expr): Check for debug
20203 statement, and disable/renable new value calculation.
20205 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
20207 * gimple-range-cache.cc (ranger_cache::ssa_range_in_bb): Delete.
20208 (ranger_cache::range_of_def): New.
20209 (ranger_cache::entry_range): New.
20210 (ranger_cache::exit_range): New.
20211 (ranger_cache::range_of_expr): Adjust.
20212 (ranger_cache::range_on_edge): Adjust.
20213 (ranger_cache::propagate_cache): Call exit_range directly.
20214 * gimple-range-cache.h (class ranger_cache): Adjust.
20216 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
20218 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust for
20219 gori_compute being a member rather than base class.
20220 dervied call to member call.
20221 (ranger_cache::dump): No longer dump gori_map.
20222 (ranger_cache::dump_bb): New.
20223 (ranger_cache::get_non_stale_global_range): Adjust for gori_compute
20224 being a member rather than base class.
20225 (ranger_cache::set_global_range): Ditto.
20226 (ranger_cache::ssa_range_in_bb): Ditto.
20227 (ranger_cache::range_of_expr): New.
20228 (ranger_cache::range_on_edge): New.
20229 (ranger_cache::block_range): Adjust for gori_computes. Debug changes.
20230 (ranger_cache::propagate_cache): Adjust debugging output.
20231 (ranger_cache::fill_block_cache): Adjust for gori_computes. Debug
20233 * gimple-range-cache.h (class ranger_cache): Make gori_compute a
20234 member, and inherit from range_query instead.
20235 (ranger_cache::dump_bb): New. split from dump.
20236 * gimple-range-gori.cc (gori_compute::ssa_range_in_bb): Delete.
20237 (gori_compute::expr_range_at_stmt): Delete.
20238 (gori_compute::compute_name_range_op): Delete.
20239 (gori_compute::compute_operand_range_switch): Add fur_source.
20240 (gori_compute::compute_operand_range): Add fur_source param, inline
20241 old compute_name_range_op and optimize_logical_operands.
20242 (struct tf_range): Delete.
20243 (gori_compute::logical_combine): Adjust
20244 (gori_compute::optimize_logical_operands): Delete.
20245 (gori_compute::compute_logical_operands_in_chain): Delete.
20246 (gori_compute::compute_logical_operands): Adjust.
20247 (gori_compute::compute_operand1_range): Adjust to fur_source.
20248 (gori_compute::compute_operand2_range): Ditto.
20249 (gori_compute::compute_operand1_and_operand2_range): Ditto.
20250 (gori_compute::outgoing_edge_range_p): Add range_query parameter,
20251 and adjust to fur_source.
20252 * gimple-range-gori.h (class gori_compute): Simplify and adjust to
20253 range_query and fur_source.
20254 * gimple-range.cc (gimple_ranger::range_on_edge): Query range_on_edge
20255 from the ranger_cache..
20256 (gimple_ranger::fold_range_internal): Adjust to base class change of
20258 (gimple_ranger::dump_bb): Adjust dump.
20259 * gimple-range.h (gimple_ranger):export gori computes object.
20261 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
20263 PR tree-optimization/100774
20264 * gimple-range-cache.cc (ranger_cache::get_non_stale_global_range):
20265 Constant values are also not stale.
20266 (ranger_cache::set_global_range): Range invariant values should also
20267 have the correct timestamp.
20269 2021-05-31 Martin Liska <mliska@suse.cz>
20271 * tree-streamer-in.c (unpack_ts_function_decl_value_fields):
20272 Unpack FUNCTION_DECL_DECL_TYPE.
20273 * tree-streamer-out.c (pack_ts_function_decl_value_fields):
20274 Stream FUNCTION_DECL_DECL_TYPE instead of
20275 DECL_IS_OPERATOR_NEW_P.
20276 * tree.h (set_function_decl_type): Use FUNCTION_DECL_DECL_TYPE
20278 (DECL_IS_OPERATOR_NEW_P): Likewise.
20279 (DECL_IS_OPERATOR_DELETE_P): Likewise.
20280 (DECL_LAMBDA_FUNCTION_P): Likewise.
20282 2021-05-31 Richard Biener <rguenther@suse.de>
20285 * internal-fn.c (expand_SHUFFLEVECTOR): Define.
20286 * internal-fn.def (SHUFFLEVECTOR): New.
20287 * internal-fn.h (expand_SHUFFLEVECTOR): Declare.
20288 * doc/extend.texi: Document __builtin_shufflevector.
20290 2021-05-31 Peter Bergner <bergner@linux.ibm.com>
20293 * config/rs6000/predicates.md(mma_assemble_input_operand): Allow
20294 indexed form addresses.
20296 2021-05-29 Jeff Law <jlaw@tachyum.com>
20298 * config/h8300/h8300.c (h8300_emit_stack_adjustment): Drop unused
20299 parameter. Call callers fixed.
20301 (output_plussi): Add FALLTHRU markers.
20302 (h8300_shift_needs_scratch_p): Add gcc_unreachable marker.
20304 2021-05-29 Jakub Jelinek <jakub@redhat.com>
20306 PR middle-end/99928
20307 * gimplify.c (gimplify_scan_omp_clauses): For taskloop simd
20308 combined with parallel, make sure to add shared clause to
20309 parallel for explicit linear clause.
20311 2021-05-29 Aldy Hernandez <aldyh@redhat.com>
20313 PR tree-optimization/100787
20314 * gimple-ssa-evrp.c: Disable exporting of global ranges.
20316 2021-05-28 Jason Merrill <jason@redhat.com>
20318 * tree-iterator.h (struct tree_stmt_iterator): Add operator++,
20319 operator--, operator*, operator==, and operator!=.
20320 (class tsi_range): New.
20322 2021-05-28 Richard Biener <rguenther@suse.de>
20324 PR tree-optimization/100778
20325 * tree-vect-slp.c (vect_build_slp_tree_1): Prevent possibly
20326 trapping ops in different BBs.
20328 2021-05-28 Richard Biener <rguenther@suse.de>
20331 * tree-inline.c (copy_bb): When processing __builtin_va_arg_pack
20332 copy fntype from original call.
20334 2021-05-28 Martin Liska <mliska@suse.cz>
20336 PR gcov-profile/100751
20337 * doc/gcov.texi: Revert partially a hunk that was wrong.
20339 2021-05-28 Cooper Qu <cooper.qu@linux.alibaba.com>
20341 * config/csky/csky-linux-elf.h (HAVE_sync_compare_and_swapqi):
20343 (HAVE_sync_compare_and_swaphi): Likewise.
20344 (HAVE_sync_compare_and_swapsi): Likewise.
20346 2021-05-28 Jakub Jelinek <jakub@redhat.com>
20348 PR middle-end/99928
20349 * tree.h (OMP_CLAUSE_MAP_IMPLICIT): Define.
20351 2021-05-28 Tobias Burnus <tobias@codesourcery.com>
20353 * gimplify.c (gimplify_omp_affinity): New.
20354 (gimplify_scan_omp_clauses): Call it; remove affinity clause afterwards.
20355 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_AFFINITY.
20356 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_AFFINITY.
20357 * tree.c (omp_clause_num_ops, omp_clause_code_name): Add clause.
20358 (walk_tree_1): Handle OMP_CLAUSE_AFFINITY.
20360 2021-05-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
20361 Richard Biener <rguenther@suse.de>
20363 * match.pd <popcount & / + pattern matching>:
20364 When generating popcount directly fails, try doing it in two halves.
20366 2021-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
20368 * Makefile.in (generated_files): Add gimple-match.c and
20371 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
20373 * gensupport.c (alter_predicate_for_insn): Handle MATCH_DUP.
20375 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
20377 * gensupport.c (alter_constraints): Add MATCH_SCRATCH case.
20379 2021-05-28 Kewen Lin <linkw@linux.ibm.com>
20381 PR tree-optimization/99398
20382 * tree-ssa-forwprop.c (simplify_permutation): Optimize some cases
20383 where the fed operands are CTOR/CST and propagated through
20384 VIEW_CONVERT_EXPR. Call vec_perm_indices::new_shrunk_vector.
20385 * vec-perm-indices.c (vec_perm_indices::new_shrunk_vector): New
20387 * vec-perm-indices.h (vec_perm_indices::new_shrunk_vector): New
20390 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
20392 * config/i386/mmx.md (addv2sf3): Do not call
20393 ix86_fixup_binary_operands_no_copy.
20396 (<smaxmin:code>v2sf3): Ditto.
20397 (<plusminus:insn><MMXMODEI:mode>3): Ditto.
20398 (<plusminus:insn><VI_32:mode>3): Remove expander.
20399 (<plusminus:insn><VI_32:mode>3): Rename from
20400 "*<plusminus:insn><VI_32:mode>3".
20401 (mulv4hi): Do not call ix86_fixup_binary_operands_no_copy.
20402 (mulv2hi3): Remove expander.
20403 (mulv2hi3): Rename from *mulv2hi3.
20404 (<s>mulv2hi3_highpart): Remove expander.
20405 (<s>mulv2hi3_highpart): Rename from *<s>mulv2hi3_highpart.
20406 (<smaxmin:code><MMXMODE14:mode>3): Rename from
20407 "*mmx_<smaxmin:code><MMXMODE14:mode>3".
20408 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Remove expander.
20409 (SMAXMIN_MMXMODEI): Remove mode iterator.
20410 (<smaxmin:code>v4hi3): New expander.
20411 (<smaxmin:code>v4qi3): Rename from *<smaxmin:code>v4qi3.
20412 (<smaxmin:code>v2hi3): Rename from *<smaxmin:code>v2hi3.
20413 (<smaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
20414 (SMAXMIN_VI_32): Remove mode iterator.
20415 (<umaxmin:code><MMXMODE24:mode>3): Rename from
20416 "*mmx_<umaxmin:code><MMXMODE24:mode>3".
20417 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Remove expander.
20418 (UMAXMIN_MMXMODEI): Remove mode iterator.
20419 (<umaxmin:code>v8qi3): New expander.
20420 (<umaxmin:code>v4qi3): Rename from *<umaxmin:code>v4qi3.
20421 (<umaxmin:code>v2hi3): Rename from *<umaxmin:code>v2hi3.
20422 (<umaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
20423 (UMAXMIN_VI_32): Remove mode iterator.
20424 (<any_shift:insn>v2hi3): Remove expander.
20425 (<any_shift:insn>v2hi3): Rename from *<any_shift:insn>v2hi3.
20426 (<any_logic:code><MMXMODEI:mode>3): Do not call
20427 ix86_fixup_binary_operands_no_copy.
20428 (<any_logic:code><VI_32:mode>3): Remove expander.
20429 (<any_logic:code><VI_32:mode>3): Rename from
20430 "*<any_logic:code><VI_32:mode>3".
20431 (uavg<mode>3_ceil): Do not call ix86_fixup_binary_operands_no_copy.
20432 * config/i386/sse.md (div<VF2:mode>3): Do not call
20433 ix86_fixup_binary_operands_no_copy.
20434 (div<VF1:mode>3): Ditto.
20435 (<maxmin:code><VI8_AVX2_AVX512F:mode>3): Ditto.
20436 (smulhrsv4hi3): Ditto.
20437 (smulhrsv2hi3): Ditto.
20439 2021-05-27 Martin Sebor <msebor@redhat.com>
20441 * ggc.h (gt_ggc_mx): Add overloads for all integers.
20443 * hash-map.h (class hash_map): Add pch_nx_helper overloads for all
20445 (hash_map::operator==): New function.
20447 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
20450 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
20451 For TARGET_XOP bypass SSE comparisons for all supported vector modes.
20452 * config/i386/mmx.md (*xop_maskcmp<MMXMODEI:mode>3): New insn pattern.
20453 (*xop_maskcmp<VI_32:mode>3): Ditto.
20454 (*xop_maskcmp_uns<MMXMODEI:mode>3): Ditto.
20455 (*xop_maskcmp_uns<VI_32:mode>3): Ditto.
20457 2021-05-27 Richard Earnshaw <rearnsha@arm.com>
20460 * config/arm/arm.c (arm_configure_build_target): Remove parameter
20461 opts_set, directly check opts parameters for being non-null.
20462 (arm_option_restore): Update call to arm_configure_build_target.
20463 (arm_option_override): Likewise.
20464 (arm_can_inline_p): Likewise.
20465 (arm_valid_target_attribute_tree): Likewise.
20466 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
20467 * config/arm/arm-protos.h (arm_configure_build_target): Adjust
20470 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
20472 * vr-values.c (simplify_conversion_using_ranges): Use
20473 get_range_query instead of get_global_range_query.
20475 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
20477 * gimple-range.cc (get_range_global): Move to value-query.cc.
20478 (gimple_range_global): Same.
20479 (get_global_range_query): Same.
20480 (global_range_query::range_of_expr): Same.
20481 * gimple-range.h (class global_range_query): Move to
20483 (gimple_range_global): Same.
20484 * tree-ssanames.c (get_range_info): Move to value-query.cc.
20485 (get_ptr_nonnull): Same.
20486 * tree-ssanames.h (get_range_info): Remove.
20487 (get_ptr_nonnull): Remove.
20488 * value-query.cc (get_ssa_name_range_info): Move from
20490 (get_ssa_name_ptr_info_nonnull): Same.
20491 (get_range_global): Move from gimple-range.cc.
20492 (gimple_range_global): Same.
20493 (get_global_range_query): Same.
20494 (global_range_query::range_of_expr): Same.
20495 * value-query.h (class global_range_query): Move from
20497 (gimple_range_global): Same.
20499 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
20502 * config/i386/mmx.md (uavgv4qi3_ceil): New insn pattern.
20503 (uavgv2hi3_ceil): Ditto.
20505 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
20508 * doc/extend.texi (scalar_storage_order): Rephrase slightly.
20510 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
20512 * tree-ssanames.c (get_range_info): Merge both copies of
20513 get_range_info into one that works with irange.
20514 * tree-ssanames.h (get_range_info): Remove version that works on
20517 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
20519 * builtins.c (check_nul_terminated_array): Convert to get_range_query.
20520 (expand_builtin_strnlen): Same.
20521 (determine_block_size): Same.
20522 * fold-const.c (expr_not_equal_to): Same.
20523 * gimple-fold.c (size_must_be_zero_p): Same.
20524 * gimple-match-head.c: Include gimple-range.h.
20525 * gimple-pretty-print.c (dump_ssaname_info): Convert to get_range_query.
20526 * gimple-ssa-warn-restrict.c
20527 (builtin_memref::extend_offset_range): Same.
20528 * graphite-sese-to-poly.c (add_param_constraints): Same.
20529 * internal-fn.c (get_min_precision): Same.
20530 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): Same.
20531 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Same.
20533 * tree-data-ref.c (split_constant_offset): Same.
20534 (dr_step_indicator): Same.
20535 * tree-dfa.c (get_ref_base_and_extent): Same.
20536 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
20537 * tree-ssa-loop-niter.c (refine_value_range_using_guard): Same.
20538 (determine_value_range): Same.
20539 (record_nonwrapping_iv): Same.
20540 (infer_loop_bounds_from_signedness): Same.
20541 (scev_var_range_cant_overflow): Same.
20542 * tree-ssa-phiopt.c (two_value_replacement): Same.
20543 * tree-ssa-pre.c (insert_into_preds_of_block): Same.
20544 * tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Same.
20545 * tree-ssa-strlen.c (handle_builtin_stxncpy_strncat): Same.
20547 (dump_strlen_info): Same.
20548 (set_strlen_range): Same.
20549 (maybe_diag_stxncpy_trunc): Same.
20550 (get_len_or_size): Same.
20551 (handle_integral_assign): Same.
20552 * tree-ssa-structalias.c (find_what_p_points_to): Same.
20553 * tree-ssa-uninit.c (find_var_cmp_const): Same.
20554 * tree-switch-conversion.c (bit_test_cluster::emit): Same.
20555 * tree-vect-patterns.c (vect_get_range_info): Same.
20556 (vect_recog_divmod_pattern): Same.
20557 * tree-vrp.c (intersect_range_with_nonzero_bits): Same.
20558 (register_edge_assert_for_2): Same.
20559 (determine_value_range_1): Same.
20560 * tree.c (get_range_pos_neg): Same.
20561 * vr-values.c (vr_values::get_lattice_entry): Same.
20562 (vr_values::update_value_range): Same.
20563 (simplify_conversion_using_ranges): Same.
20565 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
20567 * gimple-ssa-warn-alloca.c (alloca_call_type): Use
20568 get_range_query instead of query argument.
20569 (pass_walloca::execute): Enable and disable global ranger.
20571 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
20573 * gimple-ssa-evrp.c (rvrp_folder::rvrp_folder): Call
20575 (rvrp_folder::~rvrp_folder): Call disable_ranger.
20576 (hybrid_folder::hybrid_folder): Call enable_ranger.
20577 (hybrid_folder::~hybrid_folder): Call disable_ranger.
20579 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
20581 * function.c (allocate_struct_function): Set cfun->x_range_query.
20582 * function.h (struct function): Declare x_range_query.
20583 (get_range_query): New.
20584 (get_global_range_query): New.
20585 * gimple-range-cache.cc (ssa_global_cache::ssa_global_cache):
20586 Remove call to safe_grow_cleared.
20587 * gimple-range.cc (get_range_global): New.
20588 (gimple_range_global): Move from gimple-range.h.
20589 (get_global_range_query): New.
20590 (global_range_query::range_of_expr): New.
20591 (enable_ranger): New.
20592 (disable_ranger): New.
20593 * gimple-range.h (gimple_range_global): Move to gimple-range.cc.
20594 (class global_range_query): New.
20595 (enable_ranger): New.
20596 (disable_ranger): New.
20597 * gimple-ssa-evrp.c (evrp_folder::~evrp_folder): Rename
20598 dump_all_value_ranges to dump.
20599 * tree-vrp.c (vrp_prop::finalize): Same.
20600 * value-query.cc (range_query::dump): New.
20601 * value-query.h (range_query::dump): New.
20602 * vr-values.c (vr_values::dump_all_value_ranges): Rename to...
20603 (vr_values::dump): ...this.
20604 * vr-values.h (class vr_values): Rename dump_all_value_ranges to
20605 dump and make virtual.
20607 2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
20609 * config/i386/i386.c (ix86_autovectorize_vector_modes):
20610 Add V4QImode and V16QImode for TARGET_SSE2.
20611 * doc/sourcebuild.texi (Vector-specific attributes):
20612 Add vect64 and vect32 description.
20614 2021-05-26 Bernd Edlinger <bernd.edlinger@hotmail.de>
20616 * gimple-range-gori.cc (range_def_chain::register_dependency):
20617 Resize m_def_chain when needed.
20619 2021-05-26 Christophe Lyon <christophe.lyon@linaro.org>
20621 * config/arm/mve.md (mve_vaddvq_<supf><mode>): Prefix with '@'.
20622 * config/arm/neon.md (reduc_plus_scal_<mode>): Move to ..
20623 * config/arm/vec-common.md: .. here. Add support for MVE.
20625 2021-05-26 Jakub Jelinek <jakub@redhat.com>
20627 * config/epiphany/epiphany.c (epiphany_print_operand_address): Remove
20629 * config/microblaze/microblaze.c (microblaze_legitimize_address,
20631 microblaze_option_override, print_operand): Likewise.
20632 * config/microblaze/microblaze.md (call_internal_plt,
20633 call_value_intern_plt, call_value_intern): Likewise.
20634 * config/arm/aout.h (ASM_OUTPUT_ALIGN): Likewise.
20635 * config/iq2000/iq2000.md (call_internal1, call_value_internal1,
20636 call_value_multiple_internal1): Likewise.
20637 * config/bfin/bfin.c (symbolic_reference_mentioned_p): Likewise.
20639 2021-05-26 Jan-Benedict Glaw <jbglaw@lug-owl.de>
20641 * config/arc/arc.c (arc_address_cost, arc_print_operand_address,
20642 arc_ccfsm_advance, symbolic_reference_mentioned_p,
20643 arc_raw_symbolic_reference_mentioned_p): Remove register
20646 2021-05-26 Jakub Jelinek <jakub@redhat.com>
20649 * omp-low.c: Include omp-offload.h.
20650 (create_omp_child_function): If current_function_decl has
20651 "omp declare target" attribute and is_gimple_omp_offloaded,
20652 remove that attribute from the copy of attribute list and
20653 add "omp target entrypoint" attribute instead.
20654 (lower_omp_target): Mark .omp_data_sizes.* and .omp_data_kinds.*
20655 variables for offloading if in omp_maybe_offloaded_ctx.
20656 * omp-offload.c (pass_omp_target_link::execute): Nullify second
20657 argument to GOMP_target_data_ext in offloaded code.
20659 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
20661 * config/csky/csky.c (csky_can_change_mode_class): Delete.
20662 For csky, HF/SF mode use the low bits of VREGS.
20664 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
20666 * gimplify.c (gimplify_decl_expr): Do not clear TREE_READONLY on a
20667 DECL which is a reference for OMP.
20669 2021-05-26 Martin Liska <mliska@suse.cz>
20671 PR gcov-profile/100751
20672 * doc/gcov.texi: Document that __gcov_dump can be called just
20673 once and that __gcov_reset resets run-time counters.
20675 2021-05-26 Martin Liska <mliska@suse.cz>
20677 * doc/install.texi: Port relevant part from install-old.texi
20678 and re-generate list of CPUs and systems.
20680 2021-05-26 Martin Liska <mliska@suse.cz>
20682 * Makefile.in: Remove it.
20683 * doc/include/fdl.texi: Update next/previous chapters.
20684 * doc/install.texi: Likewise.
20685 * doc/install-old.texi: Removed.
20687 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
20689 * config/csky/csky.c (ck810_legitimate_index_p): Support
20690 "base + index" with DF mode.
20691 * config/csky/constraints.md ("Y"): New constraint for memory operands
20692 without index register.
20693 * config/csky/csky_insn_fpuv2.md (fpuv3_movdf): Use "Y" instead of "m"
20694 when mov between memory and general registers, and lower their priority.
20695 * config/csky/csky_insn_fpuv3.md (fpuv2_movdf): Likewise.
20697 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
20699 * config/csky/csky.c (TARGET_PROMOTE_PROTOTYPES): Delete.
20701 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
20703 * config/csky/csky.md (untyped_call): Emit clobber for return
20704 registers to mark them used.
20706 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
20708 * config/csky/csky.md (cskyv2_sextend_ldbs): New.
20710 2021-05-26 Andrew Pinski <apinski@marvell.com>
20712 * match.pd (x < 0 ? ~y : y): New patterns.
20714 2021-05-26 Andrew Pinski <apinski@marvell.com>
20716 * match.pd (A?CST1:CST2): Add simplifcations for A?0:+-1, A?+-1:0,
20717 A?POW2:0 and A?0:POW2.
20719 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
20721 * gimple-range-gori.cc (class logical_stmt_cache): Delete
20722 (logical_stmt_cache::logical_stmt_cache ): Delete.
20723 (logical_stmt_cache::~logical_stmt_cache): Delete.
20724 (logical_stmt_cache::cache_entry::dump): Delete.
20725 (logical_stmt_cache::get_range): Delete.
20726 (logical_stmt_cache::cached_name ): Delete.
20727 (logical_stmt_cache::same_cached_name): Delete.
20728 (logical_stmt_cache::cacheable_p): Delete.
20729 (logical_stmt_cache::slot_diagnostics ): Delete.
20730 (logical_stmt_cache::dump): Delete.
20731 (gori_compute_cache::gori_compute_cache): Delete.
20732 (gori_compute_cache::~gori_compute_cache): Delete.
20733 (gori_compute_cache::compute_operand_range): Delete.
20734 (gori_compute_cache::cache_stmt): Delete.
20735 * gimple-range-gori.h (gori_compute::compute_operand_range): Remove
20737 (class gori_compute_cache): Delete.
20739 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
20741 * gimple-range.cc (fold_using_range::range_of_range_op): Use m_gori
20743 (fold_using_range::range_of_address): Adjust.
20744 (fold_using_range::range_of_phi): Adjust.
20745 * gimple-range.h (class fur_source): Adjust.
20746 (fur_source::fur_source): Adjust.
20748 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
20750 * gimple-range-gori.cc (gori_compute::expr_range_at_stmt): Rename
20751 from expr_range_in_bb and adjust.
20752 (gori_compute::compute_name_range_op): Adjust.
20753 (gori_compute::optimize_logical_operands): Adjust.
20754 (gori_compute::compute_logical_operands_in_chain): Adjust.
20755 (gori_compute::compute_operand1_range): Adjust.
20756 (gori_compute::compute_operand2_range): Adjust.
20757 (ori_compute_cache::cache_stmt): Adjust.
20758 * gimple-range-gori.h (gori_compute): Rename prototype.
20760 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
20762 * gimple-range.cc (gimple_ranger::range_of_expr): Non-null should be
20763 checked only after range_of_stmt, not range_on_entry.
20764 (gimple_ranger::range_on_entry): Check for non-null in any
20765 predecessor block, if it is not already non-null.
20766 (gimple_ranger::range_on_exit): DOnt check for non-null after
20767 range on entry call.
20768 (gimple_ranger::dump_bb): New. Split from dump.
20769 (gimple_ranger::dump): Adjust.
20770 * gimple-range.h (class gimple_ranger): Adjust.
20772 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
20774 * gimple-range-cache.cc (struct range_timestamp): Delete.
20775 (class temporal_cache): Adjust.
20776 (temporal_cache::get_timestamp): Delete.
20777 (temporal_cache::set_dependency): Delete.
20778 (temporal_cache::temporal_value): Adjust.
20779 (temporal_cache::current_p): Take dependencies as params.
20780 (temporal_cache::set_timestamp): Adjust.
20781 (temporal_cache::set_always_current): Adjust.
20782 (ranger_cache::get_non_stale_global_range): Adjust.
20783 (ranger_cache::register_dependency): Delete.
20784 * gimple-range-cache.h (class range_cache): Adjust.
20786 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
20788 * gimple-range-gori.cc (range_def_chain::range_def_chain): init
20790 (range_def_chain::~range_def_chain): Dispose of obstack rather than
20791 each individual bitmap.
20792 (range_def_chain::set_import): New.
20793 (range_def_chain::get_imports): New.
20794 (range_def_chain::chain_import_p): New.
20795 (range_def_chain::register_dependency): Rename from build_def_chain
20797 (range_def_chain::def_chain_in_bitmap_p): New.
20798 (range_def_chain::add_def_chain_to_bitmap): New.
20799 (range_def_chain::has_def_chain): Just check first depenedence.
20800 (range_def_chain::get_def_chain): Process imports, use generic
20801 register_dependency routine.
20802 (range_def_chain::dump): New.
20803 (gori_map::gori_map): Allocate import list.
20804 (gori_map::~gori_map): Release imports.
20805 (gori_map::exports): Check for past allocated block size.
20806 (gori_map::imports): New.
20807 (gori_map::def_chain_in_export_p): Delete.
20808 (gori_map::is_import_p): New.
20809 (gori_map::maybe_add_gori): Handle imports.
20810 (gori_map::dump): Adjust output, add imports.
20811 (gori_compute::has_edge_range_p): Remove def_chain_in_export call.
20812 (gori_export_iterator::gori_export_iterator): New.
20813 (gori_export_iterator::next): New.
20814 (gori_export_iterator::get_name): New.
20815 * gimple-range-gori.h (range_def_chain): Add imports and direct
20816 dependecies via struct rdc.
20817 (range_def_chain::depend1): New.
20818 (range_def_chain::depend2): New.
20819 (class gori_map): Adjust.
20820 (FOR_EACH_GORI_IMPORT_NAME): New.
20821 (FOR_EACH_GORI_EXPORT_NAME): New.
20822 (class gori_export_iterator): New.
20824 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
20826 * gimple-range-cache.cc (ranger_cache::ranger_cache): Move initial
20827 export cache filling to here.
20828 * gimple-range-gori.cc (gori_compute::gori_compute) : From Here.
20830 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
20832 * gimple-range-gori.cc (range_def_chain): Move to gimple-range-gori.h.
20833 (gori_map): Move to gimple-range-gori.h.
20834 (gori_compute::gori_compute): Adjust.
20835 (gori_compute::~gori_compute): Delete.
20836 (gori_compute::compute_operand_range_switch): Adjust.
20837 (gori_compute::compute_operand_range): Adjust.
20838 (gori_compute::compute_logical_operands): Adjust.
20839 (gori_compute::has_edge_range_p ): Adjust.
20840 (gori_compute::set_range_invariant): Delete.
20841 (gori_compute::dump): Adjust.
20842 (gori_compute::outgoing_edge_range_p): Adjust.
20843 * gimple-range-gori.h (class range_def_chain): Relocate here.
20844 (class gori_map): Relocate here.
20845 (class gori_compute): Inherit from gori_map, and adjust.
20847 2021-05-25 Aldy Hernandez <aldyh@redhat.com>
20849 * value-range.cc (range_tests_legacy): Use
20850 build_nonstandard_integer_type instead of int and short.
20852 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
20854 * gimplify.c (gimplify_decl_expr): Clear TREE_READONLY on the DECL
20855 when really creating an initialization statement for it.
20857 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
20859 * tree-inline.c (setup_one_parameter): Fix thinko in new condition.
20861 2021-05-25 Kito Cheng <kito.cheng@sifive.com>
20863 * config/riscv/riscv.h (ASM_SPEC): Pass -mno-relax.
20865 2021-05-25 Martin Liska <mliska@suse.cz>
20867 PR tree-optimization/92860
20869 * optc-save-gen.awk: Remove exceptions.
20871 2021-05-25 Martin Liska <mliska@suse.cz>
20873 * asan.h (sanitize_coverage_p): New function.
20874 * doc/extend.texi: Document it.
20875 * fold-const.c (fold_range_test): Use sanitize_flags_p
20876 instead of flag_sanitize_coverage.
20877 (fold_truth_andor): Likewise.
20878 * sancov.c: Likewise.
20879 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
20880 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Handle
20881 -fsanitize-coverage when inlining.
20883 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
20885 * config/csky/csky-modes.def : Fix copyright.
20887 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
20889 * config/csky/csky-modes.def : Amend copyright.
20890 * config/csky/csky_insn_fpuv2.md : Likewise.
20891 * config/csky/csky_insn_fpuv3.md : Likewise.
20893 2021-05-25 Richard Biener <rguenther@suse.de>
20895 PR middle-end/100727
20896 * calls.c (initialize_argument_information): Explicitely test
20897 for WITH_SIZE_EXPR.
20898 * gimple-expr.c (mark_addressable): Skip outer WITH_SIZE_EXPR.
20900 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
20902 * config/csky/csky.h (FRAME_POINTER_REGNUM): Use
20903 HARD_FRAME_POINTER_REGNUM and FRAME_POINTER_REGNUM instead of
20904 the signle definition. The signle definition may not work well
20905 at simplify_subreg_regno().
20906 (HARD_FRAME_POINTER_REGNUM): New.
20907 (ELIMINABLE_REGS): Add for HARD_FRAME_POINTER_REGNUM.
20908 * config/csky/csky.c (get_csky_live_regs, csky_can_eliminate,
20909 csky_initial_elimination_offset, csky_expand_prologue,
20910 csky_expand_epilogue): Add for HARD_FRAME_POINTER_REGNUM.
20912 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
20914 * config/csky/csky.c (csky_option_override):
20915 Init csky_arch_isa_features[] in advance, so TARGET_DSP
20916 and TARGET_DIV can be set well.
20918 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
20920 * config/csky/constraints.md ("l", "h"): Delete.
20921 * config/csky/csky.h (reg_class, REG_CLASS_NAMES,
20922 REG_CLASS_CONTENTS): Delete LO_REGS and HI_REGS.
20923 * config/csky/csky.c (regno_reg_classm,
20924 csky_secondary_reload, csky_register_move_cost):
20925 Use HILO_REGS instead of LO_REGS and HI_REGS.
20927 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
20929 * config/csky/constraints.md ("W"): New constriant for mem operand
20930 with base reg, index register.
20931 ("Q"): Renamed and modified "csky_valid_fpuv2_mem_operand" to
20932 "csky_valid_mem_constraint_operand" to deal with both "Q" and "W"
20934 ("Dv"): New constraint for const double value that can be used at
20936 * config/csky/csky-modes.def (HFmode): New mode.
20937 * config/csky/csky-protos.h (csky_valid_fpuv2_mem_operand): Rename
20938 to "csky_valid_mem_constraint_operand" and support new constraint
20940 (csky_get_movedouble_length): New.
20941 (fpuv3_output_move): New.
20942 (fpuv3_const_double): New.
20943 * config/csky/csky.c (csky_option_override): New arch CK860 with fpv3.
20944 (decompose_csky_address): Refine.
20945 (csky_print_operand): New "CONST_DOUBLE" operand.
20946 (csky_output_move): Support fpv3 instructions.
20947 (csky_get_movedouble_length): New.
20948 (fpuv3_output_move): New.
20949 (fpuv3_const_double): New.
20950 (csky_emit_compare): Cover float comparsion.
20951 (csky_emit_compare_float): Refine.
20952 (csky_vaild_fpuv2_mem_operand): Rename to
20953 "csky_valid_mem_constraint_operand" and support new constraint "W".
20954 (ck860_rtx_costs): New.
20955 (csky_rtx_costs): Add the cost calculation of CK860.
20956 (regno_reg_class): New vregs for fpuv3.
20957 (csky_dbx_regno): Likewise.
20958 (csky_cpu_cpp_builtins): New builtin macro for fpuv3.
20959 (csky_conditional_register_usage): Suporrot fpuv3.
20960 (csky_dwarf_register_span): Suporrot fpuv3.
20961 (csky_init_builtins, csky_mangle_type): Support "__fp16" type.
20962 (ck810_legitimate_index_p): Support fp16.
20963 * config/csky/csky.h (TARGET_TLS): ADD CK860.
20964 (CSKY_VREG_P, CSKY_VREG_LO_P, CSKY_VREG_HI_P): Support fpuv3.
20965 (TARGET_SINGLE_FPU): Support fpuv3.
20966 (TARGET_SUPPORT_FPV3): New.
20967 (FIRST_PSEUDO_REGISTER): Change to 202 to hold the new fpuv3 registers.
20968 (FIXED_REGISTERS, CALL_REALLY_USED_REGISTERS, REGISTER_NAMES,
20969 REG_CLASS_CONTENTS): Support fpuv3.
20970 * config/csky/csky.md (movsf): Move to cksy_insn_fpu.md and refine.
20971 (csky_movsf_fpv2): Likewise.
20972 (ck801_movsf): Likewise.
20973 (csky_movsf): Likewise.
20975 (csky_movdf_fpv2): Likewise.
20976 (ck801_movdf): Likewise.
20977 (csky_movdf): Likewise.
20978 (movsicc): Refine. Use "comparison_operatior" instead of
20979 "ordered_comparison_operatior".
20980 (addsicc): Likewise.
20981 (CSKY_FIRST_VFP3_REGNUM, CSKY_LAST_VFP3_REGNUM): New constant.
20982 (call_value_internal_vh): New.
20983 * config/csky/csky_cores.def (CK860): New arch and cpu.
20988 * config/csky/csky_insn_fpu.md: Refactor. Separate all float patterns
20989 into emit-patterns and match-patterns, remain the emit-patterns here,
20990 and move the match-patterns to csky_insn_fpuv2.md or
20991 csky_insn_fpuv3.md.
20992 * config/csky/csky_insn_fpuv2.md: New file for fpuv2 instructions.
20993 * config/csky/csky_insn_fpuv3.md: New file and new patterns for fpuv3
20995 * config/csky/csky_isa.def (fcr): New.
21000 (CK860): New definition for ck860.
21001 * config/csky/csky_tables.opt (ck860): New processors ck860,
21002 ck860f. And new arch ck860.
21007 * config/csky/predicates.md (csky_float_comparsion_operator): Delete
21008 "geu", "gtu", "leu", "ltu", which will never appear at float comparison.
21009 * config/csky/t-csky-elf: Support 860.
21010 * config/csky/t-csky-linux: Likewise.
21011 * doc/md.texi: Add "Q" and "W" constraints for C-SKY.
21013 2021-05-24 Aaron Sawdey <acsawdey@linux.ibm.com>
21015 * config/rs6000/genfusion.pl (gen_logical_addsubf): Refactor to
21016 add generation of logical-add and add-logical fusion pairs.
21017 * config/rs6000/rs6000-cpus.def: Add new fusion to ISA 3.1 mask
21019 * config/rs6000/rs6000.c (rs6000_option_override_internal): Turn on
21020 logical-add and add-logical fusion by default.
21021 * config/rs6000/rs6000.opt: Add -mpower10-fusion-logical-add and
21022 -mpower10-fusion-add-logical options.
21023 * config/rs6000/fusion.md: Regenerate file.
21025 2021-05-24 Aldy Hernandez <aldyh@redhat.com>
21027 * value-range.cc (irange::legacy_equal_p): Check type when
21028 comparing VR_VARYING types.
21029 (range_tests_legacy): Test comparing VARYING ranges of different
21032 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
21034 * config/aarch64/aarch64.c (neoversen1_tunings):
21035 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
21037 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
21039 * config/aarch64/aarch64.c (aarch64_classify_symbol): Use GOT for
21040 extern weak symbols. Limit symbol offsets for non-GOT symbols with
21043 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
21045 * config/arm/neon.md (vec_load_lanesxi<mode>)
21046 (vec_store_lanexoi<mode>): Move ...
21047 * config/arm/vec-common.md: here.
21049 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
21051 * config/arm/neon.md (vec_load_lanesoi<mode>)
21052 (vec_store_lanesoi<mode>): Move ...
21053 * config/arm/vec-common.md: here.
21055 2021-05-24 liuhongt <hongtao.liu@intel.com>
21058 * config/i386/i386.c (ix86_gimple_fold_builtin): Replacing
21059 stmt with GIMPLE_NOP when lhs doesn't exist.
21061 2021-05-23 Uroš Bizjak <ubizjak@gmail.com>
21064 * config/i386/mmx.md (*push<VI_32:mode>2_rex64):
21065 New instruction pattern.
21066 (*push<VI_32:mode>2): Ditto.
21067 (push splitter for SSE registers): New splitter.
21069 2021-05-23 Andrew Pinski <apinski@marvell.com>
21071 * match.pd ((A & C) != 0 ? D : 0): Limit to non pointer types.
21073 2021-05-22 Aaron Sawdey <acsawdey@linux.ibm.com>
21075 * config/rs6000/genfusion.pl (gen_addadd): Fix incorrect attr types.
21076 * config/rs6000/fusion.md: Regenerate file.
21078 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
21080 * config/rs6000/genfusion.pl (gen_addadd): New function.
21081 * config/rs6000/fusion.md: Regenerate file.
21082 * config/rs6000/rs6000-cpus.def: Add
21083 OPTION_MASK_P10_FUSION_2ADD to masks.
21084 * config/rs6000/rs6000.c (rs6000_option_override_internal):
21085 Handle default value of OPTION_MASK_P10_FUSION_2ADD.
21086 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2add.
21088 2021-05-21 Jakub Jelinek <jakub@redhat.com>
21090 PR middle-end/99928
21091 * tree.h (OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET): Define.
21092 * gimplify.c (enum gimplify_omp_var_data): Fix up
21093 GOVD_MAP_HAS_ATTACHMENTS value, add GOVD_FIRSTPRIVATE_IMPLICIT.
21094 (omp_lastprivate_for_combined_outer_constructs): If combined target
21095 has GOVD_FIRSTPRIVATE_IMPLICIT set for the decl, change it to
21096 GOVD_MAP | GOVD_SEEN.
21097 (gimplify_scan_omp_clauses): Set GOVD_FIRSTPRIVATE_IMPLICIT for
21098 firstprivate clauses with OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT.
21099 (gimplify_adjust_omp_clauses): For firstprivate clauses with
21100 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT either clear that bit and
21101 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET too, or remove it and
21102 let it be replaced by implicit map clause.
21104 2021-05-21 Jakub Jelinek <jakub@redhat.com>
21106 PR middle-end/99928
21107 * gimplify.c (omp_lastprivate_for_combined_outer_constructs): New
21109 (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LASTPRIVATE>: Use it.
21110 (gimplify_omp_for): Likewise.
21112 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
21114 PR middle-end/90115
21115 * omp-low.c (oacc_privatization_candidate_p): Reject 'static',
21116 'external' in blocks.
21118 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
21120 PR middle-end/90115
21121 * flag-types.h (enum openacc_privatization): New.
21122 * params.opt (-param=openacc-privatization): New.
21123 * doc/invoke.texi (openacc-privatization): Document it.
21124 * omp-general.h (get_openacc_privatization_dump_flags): New
21126 * omp-low.c (oacc_privatization_candidate_p): Add diagnostics.
21127 * omp-offload.c (execute_oacc_device_lower)
21128 <IFN_UNIQUE_OACC_PRIVATE>: Re-work diagnostics.
21129 * target.def (goacc.adjust_private_decl): Add 'location_t'
21131 * doc/tm.texi: Regenerate.
21132 * config/gcn/gcn-protos.h (gcn_goacc_adjust_private_decl): Adjust.
21133 * config/gcn/gcn-tree.c (gcn_goacc_adjust_private_decl): Likewise.
21134 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl):
21135 Likewise. Preserve it for...
21136 (nvptx_goacc_expand_var_decl): ... use here.
21138 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
21140 * doc/sourcebuild.texi (Other attributes): Document '__OPTIMIZE__'
21143 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
21145 PR middle-end/90115
21146 * omp-low.c (oacc_privatization_candidate_p): New function.
21147 (oacc_privatization_scan_clause_chain)
21148 (oacc_privatization_scan_decl_chain): Use it. Also
21149 'gcc_checking_assert' that we're not seeing duplicates.
21151 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
21153 PR middle-end/90115
21154 * omp-offload.c (execute_oacc_device_lower): Skip processing if no
21157 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
21159 PR middle-end/90115
21160 * omp-offload.c (execute_oacc_device_lower): Explain.
21162 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
21164 PR middle-end/90115
21165 * omp-offload.c (execute_oacc_device_lower)
21166 <IFN_UNIQUE_OACC_PRIVATE>: Diagnose and handle for 'level == -1'
21168 * internal-fn.c (expand_UNIQUE): Don't expect
21169 'IFN_UNIQUE_OACC_PRIVATE'.
21171 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
21173 PR middle-end/90115
21174 * omp-low.c (lower_omp_for): Don't evaluate OpenMP 'for' clauses.
21176 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
21178 PR middle-end/90115
21179 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl)
21180 (nvptx_goacc_expand_var_decl): Tighten.
21182 2021-05-21 Julian Brown <julian@codesourcery.com>
21183 Chung-Lin Tang <cltang@codesourcery.com>
21184 Thomas Schwinge <thomas@codesourcery.com>
21186 PR middle-end/90115
21187 * doc/tm.texi.in (TARGET_GOACC_EXPAND_VAR_DECL)
21188 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Add documentation hooks.
21189 * doc/tm.texi: Regenerate.
21190 * expr.c (expand_expr_real_1): Expand decls using the
21191 expand_var_decl OpenACC hook if defined.
21192 * internal-fn.c (expand_UNIQUE): Handle IFN_UNIQUE_OACC_PRIVATE.
21193 * internal-fn.h (IFN_UNIQUE_CODES): Add OACC_PRIVATE.
21194 * omp-low.c (omp_context): Add oacc_privatization_candidates
21196 (lower_oacc_reductions): Add PRIVATE_MARKER parameter. Insert
21198 (lower_oacc_head_tail): Add PRIVATE_MARKER parameter. Modify
21199 private marker's gimple call arguments, and pass it to
21200 lower_oacc_reductions.
21201 (oacc_privatization_scan_clause_chain)
21202 (oacc_privatization_scan_decl_chain, lower_oacc_private_marker):
21204 (lower_omp_for, lower_omp_target, lower_omp_1): Use these.
21205 * omp-offload.c (convert.h): Include.
21206 (oacc_loop_xform_head_tail): Treat private-variable markers like
21207 fork/join when transforming head/tail sequences.
21208 (struct var_decl_rewrite_info): Add struct.
21209 (oacc_rewrite_var_decl, is_sync_builtin_call): New functions.
21210 (execute_oacc_device_lower): Support rewriting gang-private
21211 variables using target hook, and fix up addr_expr and var_decl
21213 * target.def (adjust_private_decl, expand_var_decl): New hooks.
21214 * config/gcn/gcn-protos.h (gcn_goacc_adjust_gangprivate_decl):
21216 (gcn_goacc_adjust_private_decl): ...this.
21217 * config/gcn/gcn-tree.c (gcn_goacc_adjust_gangprivate_decl):
21219 (gcn_goacc_adjust_private_decl): ...this. Add LEVEL parameter.
21220 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_GANGPRIVATE_DECL): Rename
21221 definition using gcn_goacc_adjust_gangprivate_decl...
21222 (TARGET_GOACC_ADJUST_PRIVATE_DECL): ...to this, using
21223 gcn_goacc_adjust_private_decl.
21224 * config/nvptx/nvptx.c (tree-pretty-print.h): Include.
21225 (gang_private_shared_size): New global variable.
21226 (gang_private_shared_align): Likewise.
21227 (gang_private_shared_sym): Likewise.
21228 (gang_private_shared_hmap): Likewise.
21229 (nvptx_option_override): Initialize these.
21230 (nvptx_file_end): Output gang_private_shared_sym.
21231 (nvptx_goacc_adjust_private_decl, nvptx_goacc_expand_var_decl):
21233 (nvptx_set_current_function): Clear gang_private_shared_hmap.
21234 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Define hook.
21235 (TARGET_GOACC_EXPAND_VAR_DECL): Likewise.
21237 2021-05-21 H.J. Lu <hjl.tools@gmail.com>
21239 * config/i386/i386-modes.def (MAX_BITSIZE_MODE_ANY_INT): Removed.
21241 2021-05-21 Richard Biener <rguenther@suse.de>
21242 H.J. Lu <hjl.tools@gmail.com>
21244 PR middle-end/90773
21245 * expr.c (expand_constructor): Elide expand_constructor if
21246 move by pieces is preferred.
21248 2021-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21250 * config/aarch64/aarch64-builtins.c (aarch64_call_properties):
21251 Take a flag and mode value as arguments.
21252 (aarch64_modifies_global_state_p): Likewise.
21253 (aarch64_reads_global_state_p): Likewise.
21254 (aarch64_could_trap_p): Likewise.
21255 (aarch64_get_attributes): Likewise.
21256 (aarch64_init_simd_builtins): Adjust callsite of above.
21257 (aarch64_init_fcmla_laneq_builtins): Use aarch64_get_attributes to get
21258 function attributes to apply to builtins.
21259 (aarch64_init_crc32_builtins): Likewise.
21260 (aarch64_init_builtin_rsqrt): Likewise.
21262 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
21264 * config/rs6000/rs6000.md (define_attr "type"): Add types for fusion.
21265 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10): Use new fusion types.
21266 (gen_2logical): Use new fusion types.
21267 * config/rs6000/fusion.md: Regenerate.
21269 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
21272 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
21273 Handle V4QI and V2HI modes.
21274 (ix86_expand_sse_movcc): Ditto.
21275 * config/i386/mmx.md (*<sat_plusminus:insn><VI_32:mode>3):
21276 New instruction pattern.
21277 (*eq<VI_32:mode>3): Ditto.
21278 (*gt<VI_32:mode>3): Ditto.
21279 (*xop_pcmov_<VI_32:mode>): Ditto.
21280 (mmx_pblendvb32): Ditto.
21281 (mmx_pblendvb64): Rename from mmx_pblendvb.
21282 (vec_cmp<VI_32:mode><VI_32:mode>): New expander.
21283 (vec_cmpu<VI_32:mode><VI_32:mode>): Ditto.
21284 (vcond<VI_32:mode><VI_32:mode>): Ditto.
21285 (vcondu<VI_32:mode><VI_32:mode>): Ditto.
21286 (vcond_mask_<VI_32:mode><VI_32:mode>): Ditto.
21288 2021-05-21 Jakub Jelinek <jakub@redhat.com>
21290 PR tree-optimization/94589
21291 * tree-ssa-phiopt.c (spaceship_replacement): For integral rhs1 and
21292 rhs2, treat x <= 4 equivalently to x < 5 etc. In cmp1 and cmp2 (if
21293 not the same as cmp3) treat <= the same as < and >= the same as >.
21294 Don't require that cond2_phi_edge is true edge, instead take
21295 false/true edges into account based on cmp1/cmp2 comparison kinds.
21297 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
21300 * config/i386/mmx.md (SMAXMIN_MMXMODEI): New mode iterator.
21301 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Macroize expander
21302 from <smaxmin:code>v4hi3> and <smaxmin:code><MMXMODE14:mode>3
21303 using SMAXMIN_MMXMODEI mode iterator.
21304 (*<smaxmin:code>v4qi3): New insn pattern.
21305 (*<smaxmin:code>v2hi3): Ditto.
21306 (SMAXMIN_VI_32): New mode iterator.
21307 (<smaxmin:code><SMAXMIN_VI_32>mode3): New expander.
21308 (UMAXMIN_MMXMODEI): New mode iterator.
21309 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Macroize expander
21310 from <umaxmin:code>v8qi3> and <umaxmin:code><MMXMODE24:mode>3
21311 using UMAXMIN_MMXMODEI mode iterator.
21312 (*<umaxmin:code>v4qi3): New insn pattern.
21313 (*<umaxmin:code>v2hi3): Ditto.
21314 (UMAXMIN_VI_32): New mode iterator.
21315 (<umaxmin:code><UMAXMIN_VI_32>mode3): New expander.
21316 (abs<VI_32:mode>2): New insn pattern.
21317 (ssse3_abs<MMXMODEI:mode>2, abs<MMXMODEI:mode>2): Move from ...
21318 * config/i386/sse.md: ... here.
21320 2021-05-20 Clement Chigot <clement.chigot@atos.net>
21321 David Edelsohn <dje.gcc@gmail.com>
21323 * collect2.c (scan_prog_file): Issue non-fatal warning for
21326 2021-05-20 Jonathan Wakely <jwakely@redhat.com>
21328 * doc/invoke.texi (-Wno-c++11-extensions)
21329 (-Wno-c++14-extensions, -Wno-c++17-extensions)
21330 (-Wno-c++20-extensions, -Wno-c++23-extensions): Document
21333 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
21335 * config/c6x/c6x.c (c6x_output_file_unwind): Use dwarf_debuginfo_p.
21336 * config/darwin.c (darwin_override_options): Likewise.
21337 * config/i386/cygming.h (DBX_REGISTER_NUMBER): Likewise.
21338 * config/i386/darwin.h (DBX_REGISTER_NUMBER): Likewise.
21339 (DWARF2_FRAME_REG_OUT): Likewise.
21340 * config/mips/mips.c (mips_output_filename): Likewise.
21341 * config/rs6000/rs6000.c (rs6000_xcoff_declare_function_name):
21343 (rs6000_dbx_register_number): Likewise.
21344 * dbxout.c: Include flags.h.
21345 * dwarf2cfi.c (cfi_label_required_p): Likewise.
21346 (dwarf2out_do_frame): Likewise.
21347 * except.c: Include flags.h.
21348 * final.c (dwarf2_debug_info_emitted_p): Likewise.
21349 (final_scan_insn_1): Likewise.
21350 * flags.h (dwarf_debuginfo_p): New function declaration.
21351 * opts.c (dwarf_debuginfo_p): New function definition.
21352 * targhooks.c (default_debug_unwind_info): Use dwarf_debuginfo_p.
21353 * toplev.c (process_options): Likewise.
21355 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
21357 * common.opt: Change type to support bitmasks.
21358 * flag-types.h (enum debug_info_type): Rename enumerator constants.
21359 (NO_DEBUG): New bitmask.
21360 (DBX_DEBUG): Likewise.
21361 (DWARF2_DEBUG): Likewise.
21362 (XCOFF_DEBUG): Likewise.
21363 (VMS_DEBUG): Likewise.
21364 (VMS_AND_DWARF2_DEBUG): Likewise.
21365 * flags.h (debug_set_to_format): New function declaration.
21366 (debug_set_count): Likewise.
21367 (debug_set_names): Likewise.
21368 * opts.c (debug_type_masks): Array of bitmasks for debug formats.
21369 (debug_set_to_format): New function definition.
21370 (debug_set_count): Likewise.
21371 (debug_set_names): Likewise.
21372 (set_debug_level): Update access to debug_type_names.
21373 * toplev.c: Likewise.
21375 2021-05-20 Martin Sebor <msebor@redhat.com>
21377 PR middle-end/100684
21378 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Handle C++ lambda.
21380 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
21383 * config/i386/i386.md (isa): Remove x64_bmi.
21384 (enabled): Remove x64_bmi.
21385 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
21386 Remove general register alternative.
21387 (*andnot<VI_32:mode>3): Ditto.
21388 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
21389 (*<any_logic:code><VI_32:mode>3): Ditto.
21391 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
21393 * config/arm/arm.c: Include head files tree-vectorizer.h and
21396 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
21399 * config/i386/mmx.md (Yv_Yw): Revert adding V4QI and V2HI modes.
21400 (*<plusminus:insn><VI_32:mode>3): Use Yw instad of <Yv_Yw> constrint.
21401 (<s>mulv4hi3_highpart): New expander.
21402 (*<s>mulv2hi3_highpart): New insn pattern.
21403 (<s>mulv2hi3_higpart): New expander.
21404 (*<any_shift:insn>v2hi3): New insn pattern.
21405 (<any_shift:insn>v2hi3): New expander.
21406 * config/i386/sse.md (smulhrsv2hi3): New expander.
21407 (*smulhrsv2hi3): New insn pattern.
21409 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
21411 * doc/invoke.texi (vect-inner-loop-cost-factor): Document new
21413 * params.opt (vect-inner-loop-cost-factor): New.
21414 * targhooks.c (default_add_stmt_cost): Replace hardcoded factor
21415 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR, include head file
21416 tree-vectorizer.h and its required ones.
21417 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Replace
21418 hardcoded factor 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR.
21419 * config/arm/arm.c (arm_add_stmt_cost): Likewise.
21420 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
21421 * config/rs6000/rs6000.c (rs6000_add_stmt_cost): Likewise.
21422 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
21424 (_loop_vec_info::_loop_vec_info): Init inner_loop_cost_factor.
21425 * tree-vectorizer.h (_loop_vec_info): Add inner_loop_cost_factor.
21426 (LOOP_VINFO_INNER_LOOP_COST_FACTOR): New macro.
21428 2021-05-20 Christophe Lyon <christophe.lyon@linaro.org>
21429 Torbjörn Svensson <torbjorn.svensson@st.com>
21432 * doc/cpp.texi (Common Predefined Macros): Document __FILE_NAME__.
21434 2021-05-20 Jakub Jelinek <jakub@redhat.com>
21436 PR middle-end/99928
21437 * gimplify.c (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LINEAR>: For
21438 explicit linear clause when combined with target, make it map(tofrom:)
21439 instead of no clause or firstprivate.
21441 2021-05-20 Jakub Jelinek <jakub@redhat.com>
21443 PR tree-optimization/94589
21444 * match.pd ((X & Y) == X -> (X & ~Y) == 0): Simplify even in presence
21445 of integral conversions.
21447 2021-05-19 Andrew MacLeod <amacleod@redhat.com>
21449 * gimple-range.cc (fur_source::get_operand): New.
21450 (gimple_range_fold): Delete.
21451 (fold_using_range::fold_stmt): Move from gimple_ranger::calc_stmt.
21452 (fold_using_range::range_of_range_op): Move from gimple_ranger.
21453 (fold_using_range::range_of_address): Ditto.
21454 (fold_using_range::range_of_phi): Ditto.
21455 (fold_using_range::range_of_call): Ditto.
21456 (fold_using_range::range_of_builtin_ubsan_call): Move from
21457 range_of_builtin_ubsan_call.
21458 (fold_using_range::range_of_builtin_call): Move from
21459 range_of_builtin_call.
21460 (gimple_ranger::range_of_builtin_call): Delete.
21461 (fold_using_range::range_of_cond_expr): Move from gimple_ranger.
21462 (gimple_ranger::fold_range_internal): New.
21463 (gimple_ranger::range_of_stmt): Use new fold_using_range API.
21464 (fold_using_range::range_of_ssa_name_with_loop_info): Move from
21465 gimple_ranger. Improve ranges of SSA_NAMES when possible.
21466 * gimple-range.h (gimple_ranger): Remove various range_of routines.
21467 (class fur_source): New.
21468 (class fold_using_range): New.
21469 (fur_source::fur_source): New.
21471 * vr-values.c (vr_values::extract_range_basic): Use fold_using_range
21472 instead of range_of_builtin_call.
21474 2021-05-19 Jonathan Wakely <jwakely@redhat.com>
21476 * doc/cpp.texi (Common Predefined Macros): Update documentation
21477 for the __GXX_EXPERIMENTAL_CXX0X__ macro.
21479 2021-05-19 Alex Coplan <alex.coplan@arm.com>
21482 * config/arm/arm.md (nonsecure_call_internal): Always ensure
21483 callee's address is in a register.
21485 2021-05-19 Geng Qi <gengqi@linux.alibaba.com>
21487 * common/config/riscv/riscv-common.c
21488 (riscv_subset_list::parsing_subset_version): Properly parse the letter
21490 (riscv_subset_list::parse_std_ext,
21491 riscv_subset_list::parse_multiletter_ext): To handle errors generated
21492 in riscv_subset_list::parsing_subset_version.
21494 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
21496 * config/aarch64/aarch64-simd.md: Use "neon_move_narrow_q"
21497 type attribute in patterns generating XTN(2).
21499 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
21501 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_pack_trunc_<mode>):
21502 Remove as duplicate of...
21503 (aarch64_xtn<mode>): This.
21504 (aarch64_xtn2<mode>_le): Move position in file.
21505 (aarch64_xtn2<mode>_be): Move position in file.
21506 (aarch64_xtn2<mode>): Move position in file.
21507 (vec_pack_trunc_<mode>): Define as an expander.
21509 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
21511 * config/aarch64/aarch64-simd-builtins.def: Split builtin
21512 generation for aarch64_<sur>q<r>shr<u>n_n<mode> pattern into
21513 separate scalar and vector generators.
21514 * config/aarch64/aarch64-simd.md
21515 (aarch64_<sur>q<r>shr<u>n_n<mode>): Define as an expander and
21517 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_le): This and...
21518 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_be): This.
21519 * config/aarch64/iterators.md: Define SD_HSDI iterator.
21521 2021-05-19 Jonathn Wright <jonathan.wright@arm.com>
21523 * config/aarch64/aarch64-simd.md: Use UNSPEC_SQXTUN instead
21525 * config/aarch64/iterators.md: Remove UNSPEC_SQXTUN2.
21527 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
21529 * config/aarch64/aarch64-simd.md (aarch64_<sur>q<r>shr<u>n2_n<mode>):
21530 Implement as an expand emitting a big/little endian
21531 instruction pattern.
21532 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_le): Define.
21533 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_be): Define.
21535 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
21537 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn2<mode>):
21538 Implement as an expand emitting a big/little endian
21539 instruction pattern.
21540 (aarch64_<sur><addsub>hn2<mode>_insn_le): Define.
21541 (aarch64_<sur><addsub>hn2<mode>_insn_be): Define.
21542 * config/aarch64/iterators.md: Remove UNSPEC_[R]ADDHN2 and
21543 UNSPEC_[R]SUBHN2 unspecs and ADDSUBHN2 iterator.
21545 2021-05-19 Richard Biener <rguenther@suse.de>
21547 PR middle-end/100672
21548 * fold-const.c (fold_negate_expr_1): Use element_precision.
21549 (negate_expr_p): Likewise.
21551 2021-05-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
21553 * config/aarch64/iterators.md (SVE_PRED_LOAD): New iterator.
21554 (pred_load): New int attribute.
21555 * config/aarch64/aarch64-sve.md
21556 (aarch64_load_<ANY_EXTEND:optab><SVE_HSDI:mode><SVE_PARTIAL_I:mode>): Use
21557 SVE_PRED_LOAD enum iterator and corresponding pred_load attribute.
21558 * config/aarch64/aarch64-sve-builtins-base.cc (expand): Update call to
21559 code_for_aarch64_load.
21561 2021-05-19 Richard Biener <rguenther@suse.de>
21563 * cfgexpand.c (discover_nonconstant_array_refs_r): Make
21564 sure TARGET_MEM_REF bases are expanded as memory.
21565 * tree-ssa-operands.c (operands_scanner::get_tmr_operands):
21566 Do not mark TARGET_MEM_REF bases addressable.
21567 * tree-ssa.c (non_rewritable_mem_ref_base): Handle
21568 TARGET_MEM_REF bases as never rewritable.
21569 * gimple-walk.c (walk_stmt_load_store_addr_ops): Do not
21570 walk TARGET_MEM_REF bases as address-takens.
21571 * tree-ssa-dce.c (ref_may_be_aliased): Handle TARGET_MEM_REF.
21573 2021-05-19 Richard Biener <rguenther@suse.de>
21575 * builtins.c (get_object_alignment_1): Strip outer
21577 * tree-dfa.c (get_ref_base_and_extent): Handle outer
21578 WITH_SIZE_EXPR for size processing and process the
21580 * tree-ssa-alias.c (ao_ref_base_alias_set): Strip
21581 outer WITH_SIZE_EXPR.
21582 (ao_ref_base_alias_ptr_type): Likewise.
21583 (refs_may_alias_p_2): Allow WITH_SIZE_EXPR in ref->ref
21584 and handle that accordingly, stripping it for the
21585 core alias workers.
21586 * tree.c (get_base_address): Handle WITH_SIZE_EXPR by
21587 looking through it instead of returning NULL.
21589 2021-05-19 Jakub Jelinek <jakub@redhat.com>
21591 PR middle-end/100576
21592 * builtins.c (check_read_access): Convert bound to size_type_node if
21595 2021-05-19 Richard Biener <rguenther@suse.de>
21597 * tree-cfg.c (verify_types_in_gimple_min_lval): Inline...
21598 (verify_types_in_gimple_reference): ... here. Sanitize.
21599 (verify_gimple_call): Verify references in LHS and arguments.
21600 (verify_gimple_assign_single): Reject WITH_SIZE_EXPR.
21602 2021-05-19 Uroš Bizjak <ubizjak@gmail.com>
21604 * config/i386/i386.h (VALID_INT_MODE_P):
21605 Add V8QI, V4HI and V2SI modes for TARGET_64BIT.
21606 * config/i386/i386.md (isa): Add x64_bmi.
21607 (enabled): Handle x64_bmi.
21608 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
21609 Add alternative using 64bit general registers.
21610 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
21612 2021-05-19 Jakub Jelinek <jakub@redhat.com>
21614 PR middle-end/99928
21615 * tree.h (OMP_MASTER_COMBINED): Define.
21616 * gimplify.c (gimplify_scan_omp_clauses): Rewrite lastprivate
21617 handling for outer combined/composite constructs to a loop.
21618 Handle lastprivate on combined target.
21619 (gimplify_expr): Formatting fix.
21621 2021-05-19 Xionghu Luo <luoxhu@linux.ibm.com>
21623 * passes.def: Add sink_code pass before store_merging.
21624 * tree-ssa-sink.c (pass_sink_code:clone): New.
21626 2021-05-18 Bill Schmidt <wschmidt@linux.ibm.com>
21628 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Remove call to
21629 rs6000_special_adjust_field_align_p.
21630 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
21631 * config/rs6000/rs6000-call.c (rs6000_function_arg_boundary):
21632 Remove ABI warning.
21633 (rs6000_function_arg): Likewise.
21634 * config/rs6000/rs6000-protos.h
21635 (rs6000_special_adjust_field_align_p): Remove prototype.
21636 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
21638 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Remove call to
21639 rs6000_special_adjust_field_align_p.
21641 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
21644 * config/i386/i386.h (VALID_SSE2_REG_MODE):
21645 Add V4QI and V2HI modes.
21646 (VALID_INT_MODE_P): Ditto.
21647 * config/i386/mmx.md (VI_32): New mode iterator.
21648 (mmxvecsize): Handle V4QI and V2HI.
21650 (mov<VI_32:mode>): New expander.
21651 (*mov<mode>_internal): New insn pattern.
21652 (movmisalign<VI_32:mode>): New expander.
21653 (neg<VI_32:mode>): New expander.
21654 (<plusminus:insn><VI_32:mode>3): New expander.
21655 (*<plusminus:insn><VI_32:mode>3): New insn pattern.
21656 (mulv2hi3): New expander.
21657 (*mulv2hi3): New insn pattern.
21658 (one_cmpl<VI_32:mode>2): New expander.
21659 (*andnot<VI_32:mode>3): New insn pattern.
21660 (<any_logic:code><VI_32:mode>3): New expander.
21661 (*<any_logic:code><VI_32:mode>3): New insn pattern.
21663 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
21665 * config/i386/sse.md (<any_extend:insn>v4qiv4di2):
21666 Fix a mode mismatch with operand 1.
21668 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
21671 * config/i386/i386-expand.c (split_double_mode): Return
21672 temporary register when simplify_gen_subreg fails with
21673 the high half od the paradoxical subreg.
21675 2021-05-18 Richard Biener <rguenther@suse.de>
21677 * cfgexpand.c (expand_one_var): Pass in forced_stack_var
21678 and honor it when expanding.
21679 (expand_used_vars_for_block): Pass through forced_stack_var.
21680 (expand_used_vars): Likewise.
21681 (discover_nonconstant_array_refs_r): Set bits in
21682 forced_stack_vars instead of marking vars TREE_ADDRESSABLE.
21683 (avoid_type_punning_on_regs): Likewise.
21684 (discover_nonconstant_array_refs): Likewise.
21685 (pass_expand::execute): Create and pass down forced_stack_var
21686 bitmap. For parameters and returns temporarily set
21687 TREE_ADDRESSABLE when expand_function_start.
21689 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
21691 * doc/sourcebuild.texi: Document 'dg-note'.
21693 2021-05-18 Tobias Burnus <tobias@codesourcery.com>
21696 * configure: Regenerate.
21697 * configure.ac (BUILD_CFLAG, BUILD_CXXFLAGS): Add $(CFLAGS-$@).
21699 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
21701 * gimple.h (is_gimple_omp_oacc): Tighten.
21702 * omp-low.c (check_omp_nesting_restrictions): Adjust.
21704 2021-05-18 Richard Biener <rguenther@suse.de>
21706 * tree-ssa-operands.c (mark_address_taken): Simplify.
21708 2021-05-18 Martin Liska <mliska@suse.cz>
21710 * config/gcn/mkoffload.c (STR): Redefine.
21711 * config/i386/intelmic-mkoffload.c (STR): Likewise.
21712 * config/nvptx/mkoffload.c (STR): Likewise.
21714 2021-05-18 Martin Liska <mliska@suse.cz>
21716 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
21717 Use startswith function instead of strncmp.
21718 * common/config/bfin/bfin-common.c (bfin_handle_option): Likewise.
21719 * common/config/riscv/riscv-common.c (riscv_subset_list::parse): Likewise.
21720 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): Likewise.
21721 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Likewise.
21722 * config/alpha/alpha.c (alpha_elf_section_type_flags): Likewise.
21723 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
21724 * config/arm/arm.c (arm_file_start): Likewise.
21725 (arm_valid_target_attribute_rec): Likewise.
21726 (thumb1_md_asm_adjust): Likewise.
21727 * config/arm/driver-arm.c (host_detect_local_cpu): Likewise.
21728 * config/avr/avr.c (STR_PREFIX_P): Likewise.
21729 (avr_set_current_function): Likewise.
21730 (avr_handle_addr_attribute): Likewise.
21731 (avr_asm_output_aligned_decl_common): Likewise.
21732 (avr_asm_named_section): Likewise.
21733 (avr_section_type_flags): Likewise.
21734 (avr_asm_select_section): Likewise.
21735 * config/c6x/c6x.c (c6x_in_small_data_p): Likewise.
21736 (c6x_section_type_flags): Likewise.
21737 * config/darwin-c.c (darwin_cfstring_ref_p): Likewise.
21738 (darwin_objc_declare_unresolved_class_reference): Likewise.
21739 (darwin_objc_declare_class_definition): Likewise.
21740 * config/darwin.c (indirect_data): Likewise.
21741 (darwin_encode_section_info): Likewise.
21742 (darwin_objc2_section): Likewise.
21743 (darwin_objc1_section): Likewise.
21744 (machopic_select_section): Likewise.
21745 (darwin_globalize_label): Likewise.
21746 (darwin_label_is_anonymous_local_objc_name): Likewise.
21747 (darwin_asm_named_section): Likewise.
21748 (darwin_asm_output_dwarf_offset): Likewise.
21749 * config/frv/frv.c (frv_string_begins_with): Likewise.
21750 (frv_in_small_data_p): Likewise.
21751 * config/gcn/mkoffload.c (STR): Likewise.
21753 * config/i386/i386-builtins.c (get_builtin_code_for_version): Likewise.
21754 * config/i386/i386-options.c (ix86_option_override_internal): Likewise.
21755 * config/i386/i386.c (x86_64_elf_section_type_flags): Likewise.
21756 (ix86_md_asm_adjust): Likewise.
21757 * config/i386/intelmic-mkoffload.c (STR): Likewise.
21758 * config/i386/winnt.c (i386_pe_asm_named_section): Likewise.
21759 (i386_pe_file_end): Likewise.
21760 * config/ia64/ia64.c (ia64_in_small_data_p): Likewise.
21761 (ia64_section_type_flags): Likewise.
21762 * config/mips/driver-native.c (host_detect_local_cpu): Likewise.
21763 * config/mips/mips.c (mips_handle_interrupt_attr): Likewise.
21764 (mips16_stub_function_p): Likewise.
21765 (mips_function_rodata_section): Likewise.
21766 * config/msp430/msp430.c (msp430_mcu_name): Likewise.
21767 (msp430_function_section): Likewise.
21768 (msp430_section_type_flags): Likewise.
21769 (msp430_expand_helper): Likewise.
21770 * config/nios2/nios2.c (nios2_small_section_name_p): Likewise.
21771 (nios2_valid_target_attribute_rec): Likewise.
21772 * config/nvptx/mkoffload.c (process): Likewise.
21774 * config/pa/som.h: Likewise.
21775 * config/pdp11/pdp11.c (pdp11_output_ident): Likewise.
21776 * config/riscv/riscv.c (riscv_elf_select_rtx_section): Likewise.
21777 * config/rs6000/rs6000.c (VTABLE_NAME_P): Likewise.
21778 (rs6000_inner_target_options): Likewise.
21779 * config/s390/driver-native.c (s390_host_detect_local_cpu): Likewise.
21780 * config/sparc/driver-sparc.c (host_detect_local_cpu): Likewise.
21781 * config/vax/vax.c (vax_output_int_move): Likewise.
21782 * config/vms/vms-ld.c (startswith): Likewise.
21783 (process_args): Likewise.
21785 * config/vms/vms.c: Likewise.
21787 2021-05-18 Jakub Jelinek <jakub@redhat.com>
21789 PR rtl-optimization/100590
21790 * regcprop.c (copyprop_hardreg_forward_1): Only DCE dead sets if
21791 they are NONJUMP_INSN_P.
21793 2021-05-18 Jakub Jelinek <jakub@redhat.com>
21796 * function.c (push_dummy_function): Set DECL_ARTIFICIAL and
21797 DECL_ASSEMBLER_NAME on the fn_decl.
21799 2021-05-18 Jakub Jelinek <jakub@redhat.com>
21801 PR tree-optimization/94589
21802 * tree-ssa-phiopt.c (spaceship_replacement): Pattern match
21803 phi result used in (res & ~1) == 0 comparison as res >= 0 as
21804 res == 2 would be UB with -ffinite-math-only.
21806 2021-05-18 Martin Liska <mliska@suse.cz>
21808 * Makefile.in: genversion.o should depend on DATESTAMP.
21810 2021-05-18 Claudiu Zissulescu <claziss@synopsys.com>
21812 * config/arc/simdext.md (negv2si2): Remove round bracket.
21814 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
21816 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Define
21817 _Bool as macro expanding to _Bool.
21819 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
21822 * tree.c (build_reference_type_for_mode)
21823 (build_pointer_type_for_mode): Pick pointer mode if MODE argument
21825 (build_reference_type, build_pointer_type): Invoke
21826 build_*_type_for_mode with VOIDmode.
21828 2021-05-17 Andrew MacLeod <amacleod@redhat.com>
21830 PR tree-optimization/100512
21831 * gimple-range-cache.cc (ranger_cache::set_global_range): Mark const
21832 and non-zero pointer ranges as invariant.
21833 * gimple-range.cc (gimple_ranger::range_of_stmt): Remove pointer
21834 processing from here.
21836 2021-05-17 Tom de Vries <tdevries@suse.de>
21839 * config/nvptx/nvptx-protos.h (nvptx_output_atomic_insn): Declare
21840 * config/nvptx/nvptx.c (nvptx_output_barrier)
21841 (nvptx_output_atomic_insn): New function.
21842 (nvptx_print_operand): Add support for 'B'.
21843 * config/nvptx/nvptx.md: Use nvptx_output_atomic_insn for atomic
21846 2021-05-17 Aldy Hernandez <aldyh@redhat.com>
21848 PR tree-optimization/100349
21849 * vr-values.c (bounds_of_var_in_loop): Bail if scev returns
21852 2021-05-17 Tamar Christina <tamar.christina@arm.com>
21854 * config/aarch64/driver-aarch64.c (DEFAULT_ARCH): New.
21855 (host_detect_local_cpu): Use it.
21857 2021-05-17 Martin Liska <mliska@suse.cz>
21859 * doc/invoke.texi: Add 2 missing dots.
21861 2021-05-17 Marius Hillenbrand <mhillen@linux.ibm.com>
21863 PR bootstrap/100552
21864 * configure.ac: Replace pattern substitution with call to sed.
21865 * configure: Regenerate.
21867 2021-05-17 Richard Biener <rguenther@suse.de>
21869 PR middle-end/100582
21870 * tree.c (array_at_struct_end_p): Get to the base of the
21871 reference before looking for the underlying decl.
21873 2021-05-17 Joern Rennecke <joern.rennecke@embecosm.com>
21875 * genoutput.c (validate_insn_alternatives) Make "wrong number of
21876 alternatives" message more specific, and remove assumption on where
21879 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
21881 * config/arm/iterators.md (V16): New iterator.
21882 (VH_cvtto): New iterator.
21883 (v_cmp_result): Added V4HF and V8HF support.
21884 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result>): Use VDQWH.
21885 (vcond<mode><mode>): Likewise.
21886 (vcond_mask_<mode><v_cmp_result>): Likewise.
21887 (vcond<VH_cvtto><mode>): New expander.
21889 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
21891 * config/arm/arm-protos.h (arm_expand_vector_compare): Update
21893 * config/arm/arm.c (arm_expand_vector_compare): Add support for
21895 (arm_expand_vcond): Likewise.
21896 * config/arm/iterators.md (supf): Remove VCMPNEQ_S, VCMPEQQ_S,
21897 VCMPEQQ_N_S, VCMPNEQ_N_S.
21898 (VCMPNEQ, VCMPEQQ, VCMPEQQ_N, VCMPNEQ_N): Remove.
21899 * config/arm/mve.md (@mve_vcmp<mve_cmp_op>q_<mode>): Add '@' prefix.
21900 (@mve_vcmp<mve_cmp_op>q_f<mode>): Likewise.
21901 (@mve_vcmp<mve_cmp_op>q_n_f<mode>): Likewise.
21902 (@mve_vpselq_<supf><mode>): Likewise.
21903 (@mve_vpselq_f<mode>"): Likewise.
21904 * config/arm/neon.md (vec_cmp<mode><v_cmp_result): Enable for MVE
21905 and move to vec-common.md.
21906 (vec_cmpu<mode><mode>): Likewise.
21907 (vcond<mode><mode>): Likewise.
21908 (vcond<V_cvtto><mode>): Likewise.
21909 (vcondu<mode><v_cmp_result>): Likewise.
21910 (vcond_mask_<mode><v_cmp_result>): Likewise.
21911 * config/arm/unspecs.md (VCMPNEQ_U, VCMPNEQ_S, VCMPEQQ_S)
21912 (VCMPEQQ_N_S, VCMPNEQ_N_S, VCMPEQQ_U, CMPEQQ_N_U, VCMPNEQ_N_U)
21913 (VCMPGEQ_N_S, VCMPGEQ_S, VCMPGTQ_N_S, VCMPGTQ_S, VCMPLEQ_N_S)
21914 (VCMPLEQ_S, VCMPLTQ_N_S, VCMPLTQ_S, VCMPCSQ_N_U, VCMPCSQ_U)
21915 (VCMPHIQ_N_U, VCMPHIQ_U): Remove.
21916 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result): Moved
21918 (vec_cmpu<mode><mode>): Likewise.
21919 (vcond<mode><mode>): Likewise.
21920 (vcond<V_cvtto><mode>): Likewise.
21921 (vcondu<mode><v_cmp_result>): Likewise.
21922 (vcond_mask_<mode><v_cmp_result>): Likewise. Added unsafe math
21925 2021-05-17 liuhongt <hongtao.liu@intel.com>
21928 * config/i386/i386.c (ix86_gimple_fold_builtin): Use
21929 gsi_insert_seq_before instead.
21931 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
21933 * doc/sourcebuild.texi (arm_qbit_ok): Rename into...
21934 (arm_sat_ok): ...this.
21936 2021-05-17 Martin Liska <mliska@suse.cz>
21938 * lto-wrapper.c (merge_flto_options): Factor out a new function.
21939 (merge_and_complain): Use it.
21940 (run_gcc): Merge also linker command line -flto=foo argument
21943 2021-05-16 Christophe Lyon <christophe.lyon@linaro.org>
21945 * config/arm/arm.h (CPP_SPEC): Remove error message about
21946 -mlittle-endian/-mbig-endian conflict.
21948 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
21950 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
21951 __ROP_PROTECT__ if -mrop-protect is selected.
21953 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
21955 * config/rs6000/rs6000-internal.h (rs6000_stack): Add
21956 rop_hash_save_offset and rop_hash_size.
21957 * config/rs6000/rs6000-logue.c (rs6000_stack_info): Compute
21958 rop_hash_size and rop_hash_save_offset.
21959 (debug_stack_info): Dump rop_hash_save_offset and rop_hash_size.
21960 (rs6000_emit_prologue): Emit hashst[p] in prologue.
21961 (rs6000_emit_epilogue): Emit hashchk[p] in epilogue.
21962 * config/rs6000/rs6000.md (unspec): Add UNSPEC_HASHST and
21964 (hashst): New define_insn.
21965 (hashchk): Likewise.
21967 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
21969 * config/rs6000/rs6000.c (rs6000_option_override_internal):
21970 Disable shrink wrap when inserting ROP-protect instructions.
21971 * config/rs6000/rs6000.opt (mrop-protect): New option.
21972 (mprivileged): Likewise.
21973 * doc/invoke.texi: Document mrop-protect and mprivileged.
21975 2021-05-15 Hans-Peter Nilsson <hp@axis.com>
21977 * reorg.c (fill_slots_from_thread): Reinstate code typoed out in
21980 2021-05-15 Martin Jambor <mjambor@suse.cz>
21983 2021-05-13 Martin Jambor <mjambor@suse.cz>
21985 PR tree-optimization/100453
21986 * tree-sra.c (sra_modify_assign): All const base accesses do not
21987 need refreshing, not just those from decl_pool.
21988 (sra_modify_assign): Do not refresh into a const base decl.
21990 2021-05-15 Jakub Jelinek <jakub@redhat.com>
21992 PR rtl-optimization/100342
21993 * regcprop.c (copy_value): When copying a source reg in a wider
21994 mode than it has recorded for the value, adjust recorded destination
21995 mode too or punt if !REG_CAN_CHANGE_MODE_P.
21997 2021-05-14 Jason Merrill <jason@redhat.com>
21999 * intl.h: Add comments.
22001 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22003 * config/aarch64/aarch64-simd.md
22004 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Split into...
22005 (aarch64_sqdmlsl2_lane<mode>_internal): ... This...
22006 (aarch64_sqdmlal2_lane<mode>_internal): ... And this.
22007 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Split into ...
22008 (aarch64_sqdmlsl2_laneq<mode>_internal): ... This...
22009 (aarch64_sqdmlal2_laneq<mode>_internal): ... And this.
22010 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>_internal): Split into...
22011 (aarch64_sqdmlsl2_n<mode>_internal): ... This...
22012 (aarch64_sqdmlal2_n<mode>_internal): ... And this.
22014 2021-05-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22017 * config/arm/arm_neon.h (vtst_s8): Replace call to vtst builtin with it's
22018 boolean logic equivalent.
22019 (vtst_s16): Likewise.
22020 (vtst_s32): Likewise.
22021 (vtst_u8): Likewise.
22022 (vtst_u16): Likewise.
22023 (vtst_u32): Likewise.
22024 (vtst_p8): Likewise.
22025 (vtst_p16): Likewise.
22026 (vtstq_s8): Likewise.
22027 (vtstq_s16): Likewise.
22028 (vtstq_s32): Likewise.
22029 (vtstq_u8): Likewise.
22030 (vtstq_u16): Likewise.
22031 (vtstq_u32): Likewise.
22032 (vtstq_p8): Likewise.
22033 (vtstq_p16): Likewise.
22034 * config/arm/arm_neon_builtins.def: Remove entry for vtst.
22035 * config/arm/neon.md (neon_vtst<mode>): Remove pattern.
22037 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22039 * config/aarch64/aarch64-simd.md (aarch64_sqdmlal2<mode>): Merge into...
22040 (aarch64_sqdml<SBINQOPS:as>l2<mode>): ... This.
22041 (aarch64_sqdmlsl2<mode>): Delete.
22042 (aarch64_sqdmlal2_lane<mode>): Merge this...
22043 (aarch64_sqdmlsl2_lane<mode>): ... And this...
22044 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>): ... Into this.
22045 (aarch64_sqdmlal2_laneq<mode>): Merge this...
22046 (aarch64_sqdmlsl2_laneq<mode>): ... And this...
22047 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>): ... Into this.
22048 (aarch64_sqdmlal2_n<mode>): Merge this...
22049 (aarch64_sqdmlsl2_n<mode>): ... And this...
22050 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>): ... Into this.
22052 2021-05-13 Martin Sebor <msebor@redhat.com>
22054 PR middle-end/100574
22055 * builtins.c (access_ref::get_ref): Improve detection of PHIs with
22056 all null arguments.
22058 2021-05-13 Martin Sebor <msebor@redhat.com>
22060 PR tree-optimization/93100
22061 PR middle-end/98583
22062 * tree-ssa-uninit.c (check_defs): Exclude intrinsic functions that
22063 don't modify referenced objects.
22065 2021-05-13 Martin Jambor <mjambor@suse.cz>
22067 PR tree-optimization/100453
22068 * tree-sra.c (sra_modify_assign): All const base accesses do not
22069 need refreshing, not just those from decl_pool.
22070 (sra_modify_assign): Do not refresh into a const base decl.
22072 2021-05-13 Martin Liska <mliska@suse.cz>
22074 * tree-ssa-dom.c: Remove m_simplifier.
22076 2021-05-13 Richard Earnshaw <rearnsha@arm.com>
22079 * config/arm/arm.c (arm_canonicalize_comparison): Correctly
22080 canonicalize DImode inequality comparisons against the
22081 maximum integral value.
22083 2021-05-13 Jakub Jelinek <jakub@redhat.com>
22085 PR tree-optimization/98856
22086 * config/i386/i386.c (ix86_shift_rotate_cost): Add CODE argument.
22087 Expect V2DI and V4DI arithmetic right shifts to be emulated.
22088 (ix86_rtx_costs, ix86_add_stmt_cost): Adjust ix86_shift_rotate_cost
22090 * config/i386/i386-expand.c (expand_vec_perm_2perm_interleave,
22091 expand_vec_perm_2perm_pblendv): New functions.
22092 (ix86_expand_vec_perm_const_1): Use them.
22093 * config/i386/sse.md (ashr<mode>3<mask_name>): Rename to ...
22094 (<mask_codefor>ashr<mode>3<mask_name>): ... this.
22095 (ashr<mode>3): New define_expand with VI248_AVX512BW iterator.
22096 (ashrv4di3): New define_expand.
22097 (ashrv2di3): Change condition to TARGET_SSE2, handle !TARGET_XOP
22098 and !TARGET_AVX512VL expansion.
22100 2021-05-13 Uroš Bizjak <ubizjak@gmail.com>
22103 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Force mode
22104 sizes < 16 to a register when constructing vpcmov pattern.
22105 * config/i386/mmx.md (*xop_pcmov_<mode>): Use MMXMODE124 mode.
22107 2021-05-13 Martin Liska <mliska@suse.cz>
22109 * gcov-io.c (gcov_write_block): Remove.
22110 (gcov_write_words): Likewise.
22111 (gcov_read_words): Re-implement using gcov_read_bytes.
22112 (gcov_allocate): Remove.
22113 (GCOV_BLOCK_SIZE): Likewise.
22114 (struct gcov_var): Remove most of the fields.
22115 (gcov_position): Implement with ftell.
22116 (gcov_rewrite): Remove setting of start and offset fields.
22117 (from_file): Re-format.
22118 (gcov_open): Remove setbuf call. It should not be needed.
22119 (gcov_close): Remove internal buffer handling.
22120 (gcov_magic): Use __builtin_bswap32.
22121 (gcov_write_counter): Use directly gcov_write_unsigned.
22122 (gcov_write_string): Use direct fwrite and do not round
22124 (gcov_seek): Use directly fseek.
22125 (gcov_write_tag): Use gcov_write_unsigned directly.
22126 (gcov_write_length): Likewise.
22127 (gcov_write_tag_length): Likewise.
22128 (gcov_read_bytes): Use directly fread.
22129 (gcov_read_unsigned): Use gcov_read_words.
22130 (gcov_read_counter): Likewise.
22131 (gcov_read_string): Use gcov_read_bytes.
22132 * gcov-io.h (GCOV_WORD_SIZE): Adjust to reflect
22133 that size is not in bytes, but words (4B).
22134 (GCOV_TAG_FUNCTION_LENGTH): Likewise.
22135 (GCOV_TAG_ARCS_LENGTH): Likewise.
22136 (GCOV_TAG_ARCS_NUM): Likewise.
22137 (GCOV_TAG_COUNTER_LENGTH): Likewise.
22138 (GCOV_TAG_COUNTER_NUM): Likewise.
22139 (GCOV_TAG_SUMMARY_LENGTH): Likewise.
22141 2021-05-13 liuhongt <hongtao.liu@intel.com>
22144 * config/i386/sse.md (ssedoublevecmode): Add attribute for
22145 V64QI/V32HI/V16SI/V4DI.
22146 (ssehalfvecmode): Add attribute for V2DI/V2DF.
22147 (*vec_concatv4si_0): Extend to VI124_128.
22148 (*vec_concat<mode>_0): New pre-reload splitter.
22149 * config/i386/predicates.md (movq_parallel): New predicate.
22151 2021-05-13 Alexandre Oliva <oliva@adacore.com>
22153 * targhooks.c (default_zero_call_used_regs): Retry using
22154 successfully-zeroed registers as sources.
22156 2021-05-12 Tobias Burnus <tobias@codesourcery.com>
22158 * omp-low.c (finish_taskreg_scan): Use the proper detach decl.
22160 2021-05-12 Aldy Hernandez <aldyh@redhat.com>
22163 * gimple-range.cc (range_of_builtin_call): Skip out on
22164 processing __builtin_clz when varying.
22166 2021-05-12 Tom de Vries <tdevries@suse.de>
22169 * config/nvptx/nvptx-opts.h (enum ptx_version): New enum.
22170 * config/nvptx/nvptx.c (nvptx_file_start): Print .version according
22171 to ptx_version_option.
22172 * config/nvptx/nvptx.h (TARGET_PTX_6_3): Define.
22173 * config/nvptx/nvptx.md (define_insn "nvptx_shuffle<mode>")
22174 (define_insn "nvptx_vote_ballot"): Use sync variant for
22176 * config/nvptx/nvptx.opt (ptx_version): Add enum.
22177 (mptx): Add option.
22178 * doc/invoke.texi (Nvidia PTX Options): Add mptx item.
22180 2021-05-12 Richard Biener <rguenther@suse.de>
22182 PR tree-optimization/100566
22183 * tree-ssa-sccvn.c (dominated_by_p_w_unex): Properly handle
22184 allow_back for all edge queries.
22186 2021-05-12 liuhongt <hongtao.liu@intel.com>
22189 * config/i386/sse.md (<sse4_1_avx2>_pblendvb): Add
22190 splitters for pblendvb of NOT mask register.
22192 2021-05-12 Richard Biener <rguenther@suse.de>
22194 PR tree-optimization/100519
22195 * tree-ssa-reassoc.c (can_associate_p): Split into...
22196 (can_associate_op_p): ... this
22197 (can_associate_type_p): ... and this.
22198 (is_reassociable_op): Call can_associate_op_p.
22199 (break_up_subtract_bb): Call the appropriate predicates.
22200 (reassociate_bb): Likewise.
22202 2021-05-12 Martin Liska <mliska@suse.cz>
22204 * lto-wrapper.c (merge_and_complain): Merge -flto=arg options.
22205 (run_gcc): Use -flto argument detection for merged
22208 2021-05-12 Martin Liska <mliska@suse.cz>
22210 * lto-wrapper.c (print_lto_docs_link): New function.
22211 (run_gcc): Print warning about missing job server detection
22212 after we know NR of partitions. Do the same for -flto{,=1}.
22213 * opts.c (get_option_html_page): Support -flto option.
22215 2021-05-12 Martin Liska <mliska@suse.cz>
22217 * lto-wrapper.c (get_options_from_collect_gcc_options): Change
22219 (append_option): Remove.
22220 (find_option): Rework to use the vector type.
22221 (remove_option): Remove.
22222 (merge_and_complain): Use vectors for cl_decoded_option data
22224 (append_compiler_options): Likewise.
22225 (append_diag_options): Likewise.
22226 (append_linker_options): Likewise.
22227 (append_offload_options): Likewise.
22228 (compile_offload_image): Likewise.
22229 (compile_images_for_offload_targets): Likewise.
22230 (find_and_merge_options): Likewise.
22231 (run_gcc): Likewise.
22233 2021-05-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
22236 * dwarf2out.c (dwarf2out_finish): Set
22237 have_multiple_function_sections with multi-range text_section.
22239 2021-05-12 Martin Liska <mliska@suse.cz>
22241 PR bootstrap/100560
22242 * Makefile.in: Remove version.h from linker command line.
22244 2021-05-12 Richard Biener <rguenther@suse.de>
22246 PR middle-end/100547
22247 * rtl.h (rtvec_alloc): Make argument size_t.
22248 * rtl.c (rtvec_alloc): Verify the count is less than INT_MAX.
22250 2021-05-12 Jakub Jelinek <jakub@redhat.com>
22252 PR middle-end/100508
22253 * cfgexpand.c (expand_debug_expr): For DEBUG_EXPR_DECL with vector
22254 type, don't reuse DECL_RTL if it has different mode, instead force
22255 creation of a new DEBUG_EXPR.
22257 2021-05-12 Jakub Jelinek <jakub@redhat.com>
22258 Marc Glisse <marc.glisse@inria.fr>
22260 PR tree-optimization/94589
22261 * match.pd ((X & Y) == X -> (X & ~Y) == 0,
22262 (X | Y) == Y -> (X & ~Y) == 0): New GIMPLE simplifications.
22264 2021-05-12 Uroš Bizjak <ubizjak@gmail.com>
22267 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Handle V2SF mode.
22268 * config/i386/mmx.md (MMXMODE124): New mode iterator.
22270 (mmxintvecmode): New mode attribute.
22271 (mmxintvecmodelower): Ditto.
22272 (*mmx_maskcmpv2sf3_comm): New insn pattern.
22273 (*mmx_maskcmpv2sf3): Ditto.
22274 (vec_cmpv2sfv2si): New expander.
22275 (vcond<V2FI:mode>v2si): Ditto.
22276 (mmx_vlendvps): New insn pattern.
22277 (vcond<MMXMODE124:mode><MMXMODEI:mode>): Also handle V2SFmode.
22278 (vcondu<MMXMODE124:mode><MMXMODEI:mode>): Ditto.
22279 (vcond_mask_<mode><mmxintvecmodelower>): Ditto.
22281 2021-05-11 Martin Sebor <msebor@redhat.com>
22283 PR middle-end/21433
22284 * expr.c (expand_expr_real_1): Replace unreachable code with an assert.
22286 2021-05-11 Richard Biener <rguenther@suse.de>
22288 * gimple-fold.c (gimple_fold_call): Do not call
22289 maybe_fold_reference on call arguments or the static chain.
22290 (fold_stmt_1): Do not call maybe_fold_reference on GIMPLE_ASM
22293 2021-05-11 Martin Liska <mliska@suse.cz>
22295 * builtins.def (DEF_HSAIL_BUILTIN): Remove.
22296 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
22297 (DEF_HSAIL_SAT_BUILTIN): Likewise.
22298 (DEF_HSAIL_INTR_BUILTIN): Likewise.
22299 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
22300 * doc/frontends.texi: Remove BRIG.
22301 * doc/install.texi: Likewise.
22302 * doc/invoke.texi: Likewise.
22303 * doc/standards.texi: Likewise.
22304 * brig-builtins.def: Removed.
22305 * brig/ChangeLog: Removed.
22306 * brig/Make-lang.in: Removed.
22307 * brig/brig-builtins.h: Removed.
22308 * brig/brig-c.h: Removed.
22309 * brig/brig-lang.c: Removed.
22310 * brig/brigfrontend/brig-arg-block-handler.cc: Removed.
22311 * brig/brigfrontend/brig-atomic-inst-handler.cc: Removed.
22312 * brig/brigfrontend/brig-basic-inst-handler.cc: Removed.
22313 * brig/brigfrontend/brig-branch-inst-handler.cc: Removed.
22314 * brig/brigfrontend/brig-cmp-inst-handler.cc: Removed.
22315 * brig/brigfrontend/brig-code-entry-handler.cc: Removed.
22316 * brig/brigfrontend/brig-code-entry-handler.h: Removed.
22317 * brig/brigfrontend/brig-comment-handler.cc: Removed.
22318 * brig/brigfrontend/brig-control-handler.cc: Removed.
22319 * brig/brigfrontend/brig-copy-move-inst-handler.cc: Removed.
22320 * brig/brigfrontend/brig-cvt-inst-handler.cc: Removed.
22321 * brig/brigfrontend/brig-fbarrier-handler.cc: Removed.
22322 * brig/brigfrontend/brig-function-handler.cc: Removed.
22323 * brig/brigfrontend/brig-function.cc: Removed.
22324 * brig/brigfrontend/brig-function.h: Removed.
22325 * brig/brigfrontend/brig-inst-mod-handler.cc: Removed.
22326 * brig/brigfrontend/brig-label-handler.cc: Removed.
22327 * brig/brigfrontend/brig-lane-inst-handler.cc: Removed.
22328 * brig/brigfrontend/brig-machine.c: Removed.
22329 * brig/brigfrontend/brig-machine.h: Removed.
22330 * brig/brigfrontend/brig-mem-inst-handler.cc: Removed.
22331 * brig/brigfrontend/brig-module-handler.cc: Removed.
22332 * brig/brigfrontend/brig-queue-inst-handler.cc: Removed.
22333 * brig/brigfrontend/brig-seg-inst-handler.cc: Removed.
22334 * brig/brigfrontend/brig-signal-inst-handler.cc: Removed.
22335 * brig/brigfrontend/brig-to-generic.cc: Removed.
22336 * brig/brigfrontend/brig-to-generic.h: Removed.
22337 * brig/brigfrontend/brig-util.cc: Removed.
22338 * brig/brigfrontend/brig-util.h: Removed.
22339 * brig/brigfrontend/brig-variable-handler.cc: Removed.
22340 * brig/brigfrontend/hsa-brig-format.h: Removed.
22341 * brig/brigfrontend/phsa.h: Removed.
22342 * brig/brigspec.c: Removed.
22343 * brig/config-lang.in: Removed.
22344 * brig/gccbrig.texi: Removed.
22345 * brig/lang-specs.h: Removed.
22346 * brig/lang.opt: Removed.
22348 2021-05-11 Richard Biener <rguenther@suse.de>
22351 * ipa-param-manipulation.c
22352 (ipa_param_body_adjustments::modify_call_stmt): Avoid
22353 altering SSA_NAME_DEF_STMT by adjusting the calls LHS
22354 via gimple_call_lhs_ptr.
22356 2021-05-11 Alex Coplan <alex.coplan@arm.com>
22359 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear):
22360 Avoid emitting CFA adjusts on the sp if we have the fp.
22362 2021-05-11 Richard Sandiford <richard.sandiford@arm.com>
22364 * config/aarch64/iterators.md (VMUL_CHANGE_NLANES): Delete.
22365 (VMULD): New iterator.
22366 (VCOND): Handle V4HF and V8HF.
22367 (VCONQ): Fix entry for V2SF.
22368 * config/aarch64/aarch64-simd.md (mul_lane<mode>3): Use VMULD
22369 instead of VMUL. Use a 64-bit vector mode for the indexed operand.
22370 (*aarch64_mul3_elt_<vswap_width_name><mode>): Merge with...
22371 (mul_laneq<mode>3): ...this define_insn. Use VMUL instead of VDQSF.
22372 Use a 128-bit vector mode for the indexed operand. Use stype for
22373 the scheduling type.
22375 2021-05-11 Richard Biener <rguenther@suse.de>
22377 * gimple-fold.c (maybe_fold_reference): Only return
22378 is_gimple_min_invariant values.
22380 2021-05-11 Richard Biener <rguenther@suse.de>
22382 PR middle-end/100509
22383 * gimple-fold.c (fold_gimple_assign): Only call
22384 get_symbol_constant_value on register type symbols.
22386 2021-05-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
22387 Joe Ramsay <joe.ramsay@arm.com>
22390 * config/arm/arm_mve.h (__arm_vstrwq_scatter_offset): Fix wrong arguments.
22391 (__arm_vcmpneq): Remove duplicate definition.
22392 (__arm_vstrwq_scatter_offset_p): Likewise.
22393 (__arm_vmaxq_x): Likewise.
22394 (__arm_vmlsdavaq): Likewise.
22395 (__arm_vmlsdavaxq): Likewise.
22396 (__arm_vmlsdavq_p): Likewise.
22397 (__arm_vmlsdavxq_p): Likewise.
22398 (__arm_vrmlaldavhaq): Likewise.
22399 (__arm_vstrbq_p): Likewise.
22400 (__arm_vstrbq_scatter_offset): Likewise.
22401 (__arm_vstrbq_scatter_offset_p): Likewise.
22402 (__arm_vstrdq_scatter_offset): Likewise.
22403 (__arm_vstrdq_scatter_offset_p): Likewise.
22404 (__arm_vstrdq_scatter_shifted_offset): Likewise.
22405 (__arm_vstrdq_scatter_shifted_offset_p): Likewise.
22407 2021-05-11 Jakub Jelinek <jakub@redhat.com>
22409 PR middle-end/100471
22410 * omp-low.c (lower_omp_task_reductions): For OMP_TASKLOOP, if data
22411 is 0, bypass the reduction loop including
22412 GOMP_taskgroup_reduction_unregister call.
22414 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
22416 * config/rs6000/rs6000.c (struct rs6000_cost_data): New member
22417 costing_for_scalar.
22418 (rs6000_density_test): Early return if costing_for_scalar is true.
22419 (rs6000_init_cost): Init costing_for_scalar of rs6000_cost_data.
22421 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
22423 * doc/tm.texi: Regenerated.
22424 * target.def (init_cost): Add new parameter costing_for_scalar.
22425 * targhooks.c (default_init_cost): Adjust for new parameter.
22426 * targhooks.h (default_init_cost): Likewise.
22427 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
22428 (vect_compute_single_scalar_iteration_cost): Likewise.
22429 (vect_analyze_loop_2): Likewise.
22430 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
22431 (vect_bb_vectorization_profitable_p): Likewise.
22432 * tree-vectorizer.h (init_cost): Likewise.
22433 * config/aarch64/aarch64.c (aarch64_init_cost): Likewise.
22434 * config/i386/i386.c (ix86_init_cost): Likewise.
22435 * config/rs6000/rs6000.c (rs6000_init_cost): Likewise.
22437 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
22439 * config/rs6000/rs6000.c (rs6000_vect_nonmem): Renamed to
22440 vect_nonmem and moved into...
22441 (struct rs6000_cost_data): ...here.
22442 (rs6000_init_cost): Use vect_nonmem of cost_data instead.
22443 (rs6000_add_stmt_cost): Likewise.
22444 (rs6000_finish_cost): Likewise.
22446 2021-05-10 Eric Botcazou <ebotcazou@adacore.com>
22448 * range-op.cc (get_bool_state): Adjust head comment.
22449 (operator_not_equal::op1_range): Fix comment.
22450 (operator_bitwise_xor::op1_range): Remove call to gcc_unreachable.
22452 2021-05-10 Martin Sebor <msebor@redhat.com>
22454 PR middle-end/100425
22455 PR middle-end/100510
22456 * gimple-ssa-warn-alloca.c (pass_walloca::firast_time_p): Rename...
22457 (pass_walloca::xlimit_certain_p): ...to this.
22458 (pass_walloca::gate): Execute for any kind of handled warning.
22459 (pass_walloca::execute): Avoid issuing "maybe" and "unbounded"
22460 warnings when xlimit_certain_p is set.
22462 2021-05-10 Pat Haugen <pthaugen@linux.ibm.com>
22464 * config/rs6000/rs6000.c (rs6000_ira_change_pseudo_allocno_class):
22465 Return ALTIVEC_REGS if that is best_class.
22466 (rs6000_compute_pressure_classes): Add ALTIVEC_REGS.
22468 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
22470 * config/arm/arm.h (CPP_SPEC): Remove error message about
22473 2021-05-10 Martin Jambor <mjambor@suse.cz>
22475 * ipa-prop.h (IPA_NODE_REF): Removed.
22476 (IPA_NODE_REF_GET_CREATE): Likewise.
22477 (IPA_EDGE_REF): Likewise.
22478 (IPA_EDGE_REF_GET_CREATE): Likewise.
22479 (IS_VALID_JUMP_FUNC_INDEX): Likewise.
22480 * ipa-cp.c (print_all_lattices): Replaced IPA_NODE_REF with a direct
22481 use of ipa_node_params_sum.
22482 (ipcp_versionable_function_p): Likewise.
22483 (push_node_to_stack): Likewise.
22484 (pop_node_from_stack): Likewise.
22485 (set_single_call_flag): Replaced two IPA_NODE_REF with one single
22486 direct use of ipa_node_params_sum.
22487 (initialize_node_lattices): Replaced IPA_NODE_REF with a direct use of
22488 ipa_node_params_sum.
22489 (ipa_context_from_jfunc): Replaced IPA_EDGE_REF with a direct use of
22491 (ipcp_verify_propagated_values): Replaced IPA_NODE_REF with a direct
22492 use of ipa_node_params_sum.
22493 (self_recursively_generated_p): Likewise.
22494 (propagate_scalar_across_jump_function): Likewise.
22495 (propagate_context_across_jump_function): Replaced IPA_EDGE_REF with a
22496 direct use of ipa_edge_args_sum, moved the lookup after the early
22497 exit. Replaced IPA_NODE_REF with a direct use of ipa_node_params_sum.
22498 (propagate_bits_across_jump_function): Replaced IPA_NODE_REF with
22499 direct uses of ipa_node_params_sum.
22500 (propagate_vr_across_jump_function): Likewise.
22501 (propagate_aggregate_lattice): Likewise.
22502 (propagate_aggs_across_jump_function): Likewise.
22503 (propagate_constants_across_call): Likewise, also replaced
22504 IPA_EDGE_REF with a direct use of ipa_edge_args_sum.
22505 (good_cloning_opportunity_p): Replaced IPA_NODE_REF with a direct use
22506 of ipa_node_params_sum.
22507 (estimate_local_effects): Likewise.
22508 (add_all_node_vals_to_toposort): Likewise.
22509 (propagate_constants_topo): Likewise.
22510 (ipcp_propagate_stage): Likewise.
22511 (ipcp_discover_new_direct_edges): Likewise.
22512 (calls_same_node_or_its_all_contexts_clone_p): Likewise.
22513 (cgraph_edge_brings_value_p): Likewise (in both overloaded functions).
22514 (get_info_about_necessary_edges): Likewise.
22515 (want_remove_some_param_p): Likewise.
22516 (create_specialized_node): Likewise.
22517 (self_recursive_pass_through_p): Likewise.
22518 (self_recursive_agg_pass_through_p): Likewise.
22519 (find_more_scalar_values_for_callers_subset): Likewise and also
22520 replaced IPA_EDGE_REF with direct uses of ipa_edge_args_sum, in one
22521 case replacing two of those with a single query.
22522 (find_more_contexts_for_caller_subset): Likewise for the
22523 ipa_polymorphic_call_context overload.
22524 (intersect_aggregates_with_edge): Replaced IPA_EDGE_REF with a direct
22525 use of ipa_edge_args_sum. Replaced IPA_NODE_REF with direct uses of
22526 ipa_node_params_sum.
22527 (find_aggregate_values_for_callers_subset): Likewise, also reusing
22528 results of ipa_edge_args_sum->get.
22529 (cgraph_edge_brings_all_scalars_for_node): Replaced IPA_NODE_REF with
22530 direct uses of ipa_node_params_sum, replaced IPA_EDGE_REF with a
22531 direct use of ipa_edge_args_sum.
22532 (cgraph_edge_brings_all_agg_vals_for_node): Likewise, moved node
22533 summary query after the early exit and reused the result later.
22534 (decide_about_value): Replaced IPA_NODE_REF with a direct use of
22535 ipa_node_params_sum.
22536 (decide_whether_version_node): Likewise. Removed re-querying for
22537 summaries after cloning.
22538 (spread_undeadness): Replaced IPA_NODE_REF with a direct use of
22539 ipa_node_params_sum.
22540 (has_undead_caller_from_outside_scc_p): Likewise, reusing results of
22542 (identify_dead_nodes): Likewise.
22543 (ipcp_store_bits_results): Replaced IPA_NODE_REF with direct uses of
22544 ipa_node_params_sum.
22545 (ipcp_store_vr_results): Likewise.
22546 * ipa-fnsummary.c (evaluate_properties_for_edge): Likewise.
22547 (ipa_fn_summary_t::duplicate): Likewise.
22548 (analyze_function_body): Likewise.
22549 (estimate_calls_size_and_time): Likewise.
22550 (ipa_cached_call_context::duplicate_from): Likewise.
22551 (ipa_call_context::equal_to): Likewise.
22552 (remap_edge_params): Likewise.
22553 (ipa_merge_fn_summary_after_inlining): Likewise.
22554 (inline_read_section): Likewise.
22555 * ipa-icf.c (sem_function::param_used_p): Likewise.
22556 * ipa-modref.c (compute_parm_map): Likewise.
22557 (compute_parm_map): Replaced IPA_EDGE_REF with a direct use of
22559 (get_access_for_fnspec): Replaced IPA_NODE_REF with a direct use of
22560 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
22562 * ipa-profile.c (check_argument_count): Likewise.
22563 * ipa-prop.c (ipa_alloc_node_params): Replaced IPA_NODE_REF_GET_CREATE
22564 with a direct use of ipa_node_params_sum.
22565 (ipa_initialize_node_params): Likewise.
22566 (ipa_print_node_jump_functions_for_edge): Replaced IPA_EDGE_REF with a
22567 direct use of ipa_edge_args_sum and reused the query result.
22568 (ipa_compute_jump_functions_for_edge): Replaced IPA_NODE_REF with a
22569 direct use of ipa_node_params_sum and replaced IPA_EDGE_REF with a
22570 direct use of ipa_edge_args_sum.
22571 (ipa_note_param_call): Replaced IPA_NODE_REF with a direct use of
22572 ipa_node_params_sum and reused the result of the query.
22573 (ipa_analyze_node): Likewise.
22574 (ipa_analyze_controlled_uses): Replaced IPA_NODE_REF with a direct use
22575 of ipa_node_params_sum.
22576 (update_jump_functions_after_inlining): Replaced IPA_EDGE_REF with
22577 direct uses of ipa_edge_args_sum.
22578 (update_indirect_edges_after_inlining): Replaced IPA_NODE_REF with
22579 direct uses of ipa_node_params_sum and replaced IPA_EDGE_REF with a
22580 direct use of ipa_edge_args_sum. Removed superficial re-querying the
22582 (propagate_controlled_uses): Replaced IPA_NODE_REF with direct uses of
22583 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
22585 (ipa_propagate_indirect_call_infos): Replaced IPA_EDGE_REF with a
22586 direct use of ipa_edge_args_sum.
22587 (ipa_edge_args_sum_t::duplicate): Replaced IPA_NODE_REF with a direct
22588 use of ipa_node_params_sum.
22589 (ipa_print_node_params): Likewise.
22590 (ipa_write_node_info): Likewise and also replaced IPA_EDGE_REF with
22591 direct uses of ipa_edge_args_sum.
22592 (ipa_read_edge_info): Replaced IPA_EDGE_REF with a direct use of
22594 (ipa_read_node_info): Replaced IPA_NODE_REF with a direct use of
22595 ipa_node_params_sum.
22596 (ipa_prop_write_jump_functions): Likewise. Move variable node to the
22597 scopes where it is used.
22599 2021-05-10 Uroš Bizjak <ubizjak@gmail.com>
22601 * config/i386/i386-expand.c (ix86_expand_sse_movcc)
22602 <case E_V2SImode>: Force op_true to register.
22604 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
22606 * config/arm/iterators.md (MVE_FP_COMPARISONS): New.
22607 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_f<mode>)
22608 (mve_vcmp<mve_cmp_op>q_n_f<mode>): New, merge all vcmp_*f*
22610 (mve_vcmpeqq_f<mode>, mve_vcmpeqq_n_f<mode>, mve_vcmpgeq_f<mode>)
22611 (mve_vcmpgeq_n_f<mode>, mve_vcmpgtq_f<mode>)
22612 (mve_vcmpgtq_n_f<mode>, mve_vcmpleq_f<mode>)
22613 (mve_vcmpleq_n_f<mode>, mve_vcmpltq_f<mode>)
22614 (mve_vcmpltq_n_f<mode>, mve_vcmpneq_f<mode>)
22615 (mve_vcmpneq_n_f<mode>): Remove.
22616 * config/arm/unspecs.md (VCMPEQQ_F, VCMPEQQ_N_F, VCMPGEQ_F)
22617 (VCMPGEQ_N_F, VCMPGTQ_F, VCMPGTQ_N_F, VCMPLEQ_F, VCMPLEQ_N_F)
22618 (VCMPLTQ_F, VCMPLTQ_N_F, VCMPNEQ_F, VCMPNEQ_N_F): Remove.
22620 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
22622 * config/arm/iterators.md (MVE_COMPARISONS): New.
22624 (mve_cmp_type): New.
22625 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_<mode>): New, merge all
22627 (mve_vcmpneq_<mode>, mve_vcmpcsq_n_<mode>, mve_vcmpcsq_<mode>)
22628 (mve_vcmpeqq_n_<mode>, mve_vcmpeqq_<mode>, mve_vcmpgeq_n_<mode>)
22629 (mve_vcmpgeq_<mode>, mve_vcmpgtq_n_<mode>, mve_vcmpgtq_<mode>)
22630 (mve_vcmphiq_n_<mode>, mve_vcmphiq_<mode>, mve_vcmpleq_n_<mode>)
22631 (mve_vcmpleq_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
22632 (mve_vcmpneq_n_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
22633 (mve_vcmpneq_n_<mode>): Remove.
22635 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
22637 * config/arm/arm_mve.h (__arm_vcmp*): Remove 's' suffix.
22638 * config/arm/arm_mve_builtins.def (vcmp*): Remove 's' suffix.
22639 * config/arm/mve.md (mve_vcmp*): Remove 's' suffix in pattern
22642 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
22644 * config/arm/arm_mve_builtins.def (vcmpneq_u): Remove.
22645 (vcmpneq_n_u): Likewise.
22646 (vcmpeqq_u,): Likewise.
22647 (vcmpeqq_n_u): Likewise.
22648 * config/arm/iterators.md (supf): Remove VCMPNEQ_U, VCMPEQQ_U,
22649 VCMPEQQ_N_U and VCMPNEQ_N_U.
22650 * config/arm/mve.md (mve_vcmpneq): Remove <supf> iteration.
22651 (mve_vcmpeqq_n): Likewise.
22652 (mve_vcmpeqq): Likewise.
22653 (mve_vcmpneq_n): Likewise.
22655 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
22657 * config/arm/arm_mve.h (__arm_vcmpeq*u*, __arm_vcmpne*u*): Call
22658 the 's' version of the builtin.
22660 2021-05-10 Richard Biener <rguenther@suse.de>
22662 PR tree-optimization/100492
22663 * tree-loop-distribution.c (find_seed_stmts_for_distribution):
22664 Find nothing when the loop contains an irreducible region.
22666 2021-05-10 Richard Biener <rguenther@suse.de>
22668 PR middle-end/100464
22670 * gimple-fold.c (canonicalize_constructor_val): Do not set
22673 2021-05-10 Richard Biener <rguenther@suse.de>
22675 PR tree-optimization/100434
22676 * tree-ssa-dse.c (initialize_ao_ref_for_dse): Handle
22678 (dse_optimize_stmt): Handle call LHS by dropping the
22679 LHS or the whole call if it doesn't have other
22681 (pass_dse::execute): Adjust.
22683 2021-05-10 Martin Liska <mliska@suse.cz>
22685 * Makefile.in: Add missing genversion rule.
22687 2021-05-10 Alex Coplan <alex.coplan@arm.com>
22690 * config/arm/mve.md (*mve_mov<mode>): Simplify output code. Use
22691 vldrw.u32 and vstrw.32 for V2D[IF]mode loads and stores.
22693 2021-05-10 Martin Liska <mliska@suse.cz>
22695 * builtins.c (is_builtin_name): Use startswith
22696 function instead of strncmp.
22697 * collect2.c (main): Likewise.
22698 (has_lto_section): Likewise.
22699 (scan_libraries): Likewise.
22700 * coverage.c (coverage_checksum_string): Likewise.
22701 (coverage_init): Likewise.
22702 * dwarf2out.c (is_cxx): Likewise.
22703 (gen_compile_unit_die): Likewise.
22704 * gcc-ar.c (main): Likewise.
22705 * gcc.c (init_spec): Likewise.
22706 (read_specs): Likewise.
22707 (execute): Likewise.
22708 (check_live_switch): Likewise.
22709 * genattrtab.c (write_attr_case): Likewise.
22710 (IS_ATTR_GROUP): Likewise.
22711 * gencfn-macros.c (main): Likewise.
22712 * gengtype.c (type_for_name): Likewise.
22713 (gen_rtx_next): Likewise.
22714 (get_file_langdir): Likewise.
22715 (write_local): Likewise.
22716 * genmatch.c (get_operator): Likewise.
22717 (get_operand_type): Likewise.
22718 (expr::gen_transform): Likewise.
22719 * genoutput.c (validate_optab_operands): Likewise.
22720 * incpath.c (add_sysroot_to_chain): Likewise.
22721 * langhooks.c (lang_GNU_C): Likewise.
22722 (lang_GNU_CXX): Likewise.
22723 (lang_GNU_Fortran): Likewise.
22724 (lang_GNU_OBJC): Likewise.
22725 * lto-wrapper.c (run_gcc): Likewise.
22726 * omp-general.c (omp_max_simt_vf): Likewise.
22727 * omp-low.c (omp_runtime_api_call): Likewise.
22728 * opts-common.c (parse_options_from_collect_gcc_options): Likewise.
22729 * read-rtl-function.c (function_reader::read_rtx_operand_r): Likewise.
22730 * real.c (real_from_string): Likewise.
22731 * selftest.c (assert_str_startswith): Likewise.
22732 * timevar.c (timer::validate_phases): Likewise.
22733 * tree.c (get_file_function_name): Likewise.
22734 * ubsan.c (ubsan_use_new_style_p): Likewise.
22735 * varasm.c (default_function_rodata_section): Likewise.
22736 (incorporeal_function_p): Likewise.
22737 (default_section_type_flags): Likewise.
22738 * system.h (startswith): Define startswith.
22740 2021-05-10 Martin Liska <mliska@suse.cz>
22742 * bitmap.h (class auto_bitmap): Remove
22743 __cplusplus >= 201103.
22744 * config/aarch64/aarch64.c: Likewise.
22745 * gimple-ssa-store-merging.c (store_immediate_info::store_immediate_info):
22747 * sbitmap.h: Likewise.
22749 2021-05-10 Martin Liska <mliska@suse.cz>
22751 * Makefile.in: Rename gcov-iov to genversion and depend
22752 on version.h (instead of gcov-iov.h).
22753 * gcov-io.h: Include version.h instread of gcov-iov.h.
22754 * gengtype-state.c (read_state_version): Likewise.
22755 * gcov-iov.c: Moved to...
22756 * genversion.c: ...here.
22757 * lto-streamer.h (LTO_major_version): Define it with
22759 * version.c: Removed.
22760 * version.h: Removed.
22762 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
22764 * config/arc/arc.md (UNSPEC_ARC_DMPYWH): Define.
22765 * config/arc/simdext.md (VCT): Add predicates for iterator
22768 (voptab): Likewise.
22769 (vec_widen_<V_US>mult_hi_v4hi): Change pattern predicate.
22770 (<voptab>v2si3): New patterns.
22772 (reduc_plus_scal_v4hi): Likewise.
22773 (reduc_plus_scal_v2si): Likewise.
22774 (vec_duplicatev2si): Likewise.
22775 (vec_duplicatev4hi): Likewise.
22777 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
22779 * config/arc/simdext.md: Format and cleanup file.
22781 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
22783 * config/arc/simdext.md (movmisalignv2hi): Allow misaligned access
22784 only when munaligned-access option is on.
22785 (movmisalign<mode>): Likewise.
22787 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
22789 * common/config/arc/arc-common.c (arc_handle_option): Remove dot
22791 * config/arc/arc.c (arc_reorg): Remove underscore from string.
22793 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
22795 * config/arc/arc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
22796 (CTZ_DEFINED_VALUE_AT_ZERO): Likewise.
22797 * config/arc/arc.md (clrsbsi2): Cleanup pattern.
22798 (norm_f): Likewise.
22801 (clzsi2): Use fls instruction when available.
22802 (arc_clzsi2): Likewise.
22804 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
22806 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add r26 and r27.
22808 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
22810 * doc/extend.texi (__builtin_arc_sr): Swap arguments.
22812 2021-05-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
22814 PR middle-end/100467
22815 * toplev.c (compile_file): Call insn_locations_init before
22816 targetm.asm_out.code_end.
22818 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
22821 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
22823 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
22825 2021-05-07 Jakub Jelinek <jakub@redhat.com>
22826 Andrew Stubbs <amd@codesourcery.com>
22829 * builtins.c (try_store_by_multiple_pieces): Use force_operand for
22830 emit_move_insn operands.
22832 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
22834 * cfgexpand.c (expand_gimple_basic_block): Do not inherit a current
22835 location for the outgoing edges of an empty block.
22836 * dwarf2out.c (add_subscript_info): Retrieve the bounds and index
22837 type by means of the get_array_descr_info langhook, if it is set and
22838 returns true. Remove obsolete code dealing with unnamed subtypes.
22840 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
22842 * gimple-range-cache.cc (ssa_block_ranges): Virtualize.
22843 (sbr_vector): Renamed from ssa_block_cache.
22844 (sbr_vector::sbr_vector): Allocate from obstack abd initialize.
22845 (ssa_block_ranges::~ssa_block_ranges): Remove.
22846 (sbr_vector::set_bb_range): Use varying and undefined cached values.
22847 (ssa_block_ranges::set_bb_varying): Remove.
22848 (sbr_vector::get_bb_range): Adjust assert.
22849 (sbr_vector::bb_range_p): Adjust assert.
22850 (~block_range_cache): No freeing loop required.
22851 (block_range_cache::get_block_ranges): Remove.
22852 (block_range_cache::set_bb_range): Inline get_block_ranges.
22853 (block_range_cache::set_bb_varying): Remove.
22854 * gimple-range-cache.h (set_bb_varying): Remove prototype.
22855 * value-range.h (irange_allocator::get_memory): New.
22857 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
22859 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Search
22860 dominator tree is available and requested.
22861 (ranger_cache::ssa_range_in_bb): Don't search dom tree here.
22862 (ranger_cache::fill_block_cache): Don't search dom tree here either.
22863 * gimple-range-cache.h (non_null_deref_p): Add dom_search param.
22865 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
22867 * gimple-range.cc (gimple_ranger::range_on_exit): Handle block with
22868 only PHI nodes better.
22870 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
22872 * gimple-range-edge.h (gimple_outgoing_range): Rename from
22874 (gcond_edge_range): Export prototype.
22875 * gimple-range-edge.cc (gcond_edge_range): New.
22876 (gimple_outgoing_range::edge_range_p): Use gcond_edge_range.
22877 * gimple-range-gori.h (gori_compute): Use gimple_outgoing_range.
22879 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
22881 * gimple-range-edge.cc (outgoing_range::calc_switch_ranges): Compute
22882 default range into a temp and allocate only what is needed.
22884 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
22886 * range-op.cc (operator_trunc_mod::wi_fold): x % 0 is UNDEFINED.
22888 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
22890 * gimple-range.h (gimple_range_global): Pick up parameter initial
22891 values, and use-before defined locals are UNDEFINED.
22893 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
22895 * doc/extend.texi (scalar_storage_order): Mention effect on pointer
22897 * tree.h (reverse_storage_order_for_component_p): Return false if
22898 the type is a pointer.
22900 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
22902 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
22904 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
22907 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
22908 Handle V8QI, V4HI and V2SI modes.
22909 * config/i386/mmx.md (mmx_pblendvb): New insn pattern.
22910 * config/i386/sse.md (unspec): Move UNSPEC_BLENDV ...
22911 * config/i386/i386.md (unspec): ... here.
22913 2021-05-07 Tobias Burnus <tobias@codesourcery.com>
22914 Tom de Vries <tdevries@suse.de>
22916 * omp-low.c (lower_rec_simd_input_clauses): Set max_vf = 1 if
22917 a truth_value_p reduction variable is nonintegral.
22919 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
22922 * config/i386/i386-expand.c (ix86_use_mask_cmp_p):
22923 Return false for mode sizes < 16.
22925 2021-05-07 Jakub Jelinek <jakub@redhat.com>
22928 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn.
22930 2021-05-06 Martin Jambor <mjambor@suse.cz>
22932 * ipa-sra.c (ipa_sra_dump_all_summaries): Dump edge summaries even
22933 when there is no function summary.
22934 (ipa_sra_summarize_function): produce edge summaries even when
22937 2021-05-06 Tom Tromey <tom@tromey.com>
22939 * godump.c (string_hash_eq): Remove.
22940 (go_finish): Use htab_eq_string.
22942 2021-05-06 Tom Tromey <tom@tromey.com>
22944 * gengtype-state.c (read_state): Use htab_eq_string.
22945 (string_eq): Remove.
22947 2021-05-06 Tom Tromey <tom@tromey.com>
22949 * gensupport.c (htab_eq_string): Remove.
22951 2021-05-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
22954 * debug.h (gcc_debug_hooks): Add set_ignored_loc function pointer.
22955 * dwarf2out.h (dw_fde_node::ignored_debug): New data item.
22956 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Add dummy
22957 set_ignored_loc callbacks.
22958 * debug.c (do_nothing_debug_hooks): Likewise.
22959 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
22960 * dwarf2out.c (text_section_used, cold_text_section_used): Remove.
22961 (in_text_section_p, last_text_label, last_cold_label,
22962 switch_text_ranges, switch_cold_ranges): New data items.
22963 (dwarf2out_note_section_used): Remove.
22964 (dwarf2out_begin_prologue): Set fde->ignored_debug and
22966 (mark_ignored_debug_section): New helper function.
22967 (dwarf2out_end_epilogue, dwarf2out_switch_text_section): Call
22968 mark_ignored_debug_section.
22969 (dwarf2_debug_hooks): Use dwarf2out_set_ignored_loc.
22970 (dwarf2_lineno_debug_hooks): Use dummy for set_ignored_loc.
22971 (size_of_aranges): Adjust formula for multi-part text ranges size.
22972 (output_aranges): Output multi-part text ranges.
22973 (dwarf2out_set_ignored_loc): New callback function.
22974 (dwarf2out_finish): Output multi-part text ranges.
22975 (dwarf2out_c_finalize): Clear new data items.
22976 * final.c (final_start_function_1): Call set_ignored_loc callback.
22977 (final_scan_insn_1): Likewise.
22978 * ggc-page.c (gt_ggc_mx): New helper function.
22979 * stringpool.c (gt_pch_nx): Likewise.
22981 2021-05-06 Richard Biener <rguenther@suse.de>
22983 * timevar.def (TV_TREE_INSERT_PHI_NODES): Remove.
22984 (TV_TREE_SSA_REWRITE_BLOCKS): Likewise.
22985 (TV_TREE_INTO_SSA): New.
22986 * tree-into-ssa.c (insert_phi_nodes): Do not account separately.
22987 (rewrite_blocks): Likewise.
22988 (pass_data_build_ssa): Account to TV_TREE_INTO_SSA.
22990 2021-05-06 Jakub Jelinek <jakub@redhat.com>
22992 * tree-ssa-phiopt.c (value_replacement, minmax_replacement,
22993 abs_replacement, xor_replacement,
22994 cond_removal_in_popcount_clz_ctz_pattern,
22995 replace_phi_edge_with_variable): Change type of phi argument from
22996 gimple * to gphi *.
22998 2021-05-06 Richard Biener <rguenther@suse.de>
23000 * tree-ssa-loop-split.c (split_loop): Delay updating SSA form.
23001 Output an opt-info message.
23002 (do_split_loop_on_cond): Likewise.
23003 (tree_ssa_split_loops): Update SSA form here.
23005 2021-05-06 Richard Biener <rguenther@suse.de>
23007 * tree-inline.c (tree_function_versioning): Fix DECL_BY_REFERENCE
23008 return variable removal.
23010 2021-05-06 Marius Hillenbrand <mhillen@linux.ibm.com>
23012 * config/s390/s390-builtins.def (O_M5, O1_M5, ...): Remove unused macros.
23013 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
23014 (s390_vec_permi_dbl, s390_vpdi): Use the O3_U2 type for the immediate
23016 * config/s390/s390.c (s390_const_operand_ok): Remove unused
23019 2021-05-06 Jakub Jelinek <jakub@redhat.com>
23021 PR tree-optimization/94589
23022 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Call
23023 spaceship_replacement.
23024 (cond_only_block_p, spaceship_replacement): New functions.
23026 2021-05-06 Richard Biener <rguenther@suse.de>
23029 * tree-emutls.c (gen_emutls_addr): Pass in whether we're
23030 dealing with a debug use and only query existing addresses
23032 (lower_emutls_1): Avoid splitting out addresses for debug
23033 stmts, reset the debug stmt when we fail to find existing
23035 (lower_emutls_phi_arg): Set wi.stmt.
23037 2021-05-06 Christoph Muellner <cmuellner@gcc.gnu.org>
23040 * config/riscv/riscv.c (riscv_block_move_loop): Use cbranch helper.
23041 * config/riscv/riscv.md (cbranch<mode>4): Generate helpers.
23042 (stack_protect_test): Use cbranch helper.
23044 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
23047 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
23048 always return the establisher frame for __builtin_frame_address (0).
23050 2021-05-05 Ivan Sorokin <vanyacpp@gmail.com>
23053 * config/i386/i386-builtins.c (ix86_cpu_model_type_node): New.
23054 (ix86_cpu_model_var): Likewise.
23055 (ix86_cpu_features2_type_node): Likewise.
23056 (ix86_cpu_features2_var): Likewise.
23057 (fold_builtin_cpu): Cache __cpu_model and __cpu_features2 with
23060 2021-05-05 Martin Sebor <msebor@redhat.com>
23062 * passes.def (pass_warn_printf): Run after SSA.
23064 2021-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23066 * config/arm/neon.md (neon_vtst_combine<mode>): New pattern.
23067 * config/arm/predicates.md (minus_one_operand): New predicate.
23069 2021-05-05 Jeff Law <jlaw@tachyum.com>
23071 * config/avr/avr.md: Remove references to CC_STATUS_INIT.
23073 2021-05-05 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
23075 PR rtl-optimization/100263
23076 * postreload.c (move2add_valid_value_p): Ensure register can
23079 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
23081 PR rtl-optimization/100411
23082 * cfgcleanup.c (try_crossjump_to_edge): Also skip end of prologue
23083 and beginning of function markers.
23085 2021-05-05 Jeff Law <jlaw@tachyum.com>
23087 * config/cr16/cr16.h (NOTICE_UPDATE_CC): Remove.
23088 * config/cr16/cr16.c (notice_update_cc): Remove.
23089 * config/cr16/cr16-protos.h (notice_update_cc): Remove.
23091 2021-05-05 Uroš Bizjak <ubizjak@gmail.com>
23094 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
23095 Handle V8QI, V4HI and V2SI modes.
23096 * config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
23097 (ix86_build_signbit_mask): Ditto.
23098 * config/i386/mmx.md (MMXMODE14): New mode iterator.
23099 (<smaxmin:code><MMXMODE14:mode>3): New expander.
23100 (*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
23101 (<umaxmin:code><MMXMODE24:mode>3): New expander.
23102 (*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
23103 (vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
23104 (vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
23105 (vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
23106 (vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
23107 (vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
23109 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
23111 * dwarf2out.c (loc_list_from_tree_1) <DECL>: During early DWARF, do
23112 not expand the VALUE_EXPR of variables put in the non-local frame.
23113 * gimplify.c (gimplify_type_sizes) <RECORD_TYPE>: If the type is not
23114 to be ignored for debug info, ensure its variable offsets are not.
23116 2021-05-05 Richard Biener <rguenther@suse.de>
23118 PR tree-optimization/79333
23119 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
23120 Fold stmt following SSA edges.
23122 2021-05-05 Richard Biener <rguenther@suse.de>
23124 PR middle-end/100394
23125 * calls.c (expand_call): Preserve possibly throwing calls.
23126 * cfgexpand.c (expand_call_stmt): When a call can throw signal
23127 RTL expansion there are side-effects.
23128 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Simplify,
23129 mark all possibly throwing stmts necessary unless we can elide
23131 * tree-ssa-dse.c (pass_dse::execute): Preserve exceptions unless
23132 -fdelete-dead-exceptions.
23133 * tree.h (DECL_PURE_P): Add note about exceptions.
23135 2021-05-05 Alexandre Oliva <oliva@adacore.com>
23137 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Make it
23140 2021-05-04 David Edelsohn <dje.gcc@gmail.com>
23142 * config/rs6000/rs6000-call.c (rs6000_output_mi_thunk): Use
23143 get_fnname_from_decl for name of thunk.
23144 * config/rs6000/rs6000.c (rs6000_declare_alias): Use assemble_name
23145 and ASM_OUTPUT_LABEL.
23146 (rs6000_xcoff_declare_function_name): Use assemble_name and
23148 (rs6000_xcoff_declare_object_name): Use ASM_OUTPUT_LABEL.
23149 (rs6000_xcoff_encode_section_info): Don't add mapping class
23150 for aliases. Always add [DS] mapping class to primary
23152 (rs6000_asm_weaken_decl): Don't explicitly add [DS].
23154 2021-05-04 Martin Sebor <msebor@redhat.com>
23156 PR middle-end/100307
23157 * builtins.c (compute_objsize_r): Clear base0 for pointers.
23159 2021-05-04 Jeff Law <jlaw@tachyum.com>
23161 * config/bfin/bfin.h (NOTICE_UPDATE_CC): Remove.
23163 2021-05-04 Segher Boessenkool <segher@kernel.crashing.org>
23165 * caller-save.c: Remove CC0.
23166 * cfgcleanup.c: Remove CC0.
23167 * cfgrtl.c: Remove CC0.
23168 * combine.c: Remove CC0.
23169 * compare-elim.c: Remove CC0.
23170 * conditions.h: Remove CC0.
23171 * config/h8300/h8300.h: Remove CC0.
23172 * config/h8300/h8300-protos.h: Remove CC0.
23173 * config/h8300/peepholes.md: Remove CC0.
23174 * config/i386/x86-tune-sched.c: Remove CC0.
23175 * config/m68k/m68k.c: Remove CC0.
23176 * config/rl78/rl78.c: Remove CC0.
23177 * config/sparc/sparc.c: Remove CC0.
23178 * config/xtensa/xtensa.c: Remove CC0.
23179 (gen_conditional_move): Use pc_rtx instead of cc0_rtx in a piece of
23180 RTL where that is used as a placeholder only.
23181 * cprop.c: Remove CC0.
23182 * cse.c: Remove CC0.
23183 * cselib.c: Remove CC0.
23184 * df-problems.c: Remove CC0.
23185 * df-scan.c: Remove CC0.
23186 * doc/md.texi: Remove CC0. Adjust an example.
23187 * doc/rtl.texi: Remove CC0. Adjust an example.
23188 * doc/tm.texi: Regenerate.
23189 * doc/tm.texi.in: Remove CC0.
23190 * emit-rtl.c: Remove CC0.
23191 * final.c: Remove CC0.
23192 * fwprop.c: Remove CC0.
23193 * gcse-common.c: Remove CC0.
23194 * gcse.c: Remove CC0.
23195 * genattrtab.c: Remove CC0.
23196 * genconfig.c: Remove CC0.
23197 * genemit.c: Remove CC0.
23198 * genextract.c: Remove CC0.
23199 * gengenrtl.c: Remove CC0.
23200 * genrecog.c: Remove CC0.
23201 * haifa-sched.c: Remove CC0.
23202 * ifcvt.c: Remove CC0.
23203 * ira-costs.c: Remove CC0.
23204 * ira.c: Remove CC0.
23205 * jump.c: Remove CC0.
23206 * loop-invariant.c: Remove CC0.
23207 * lra-constraints.c: Remove CC0.
23208 * lra-eliminations.c: Remove CC0.
23209 * optabs.c: Remove CC0.
23210 * postreload-gcse.c: Remove CC0.
23211 * postreload.c: Remove CC0.
23212 * print-rtl.c: Remove CC0.
23213 * read-rtl-function.c: Remove CC0.
23214 * reg-notes.def: Remove CC0.
23215 * reg-stack.c: Remove CC0.
23216 * reginfo.c: Remove CC0.
23217 * regrename.c: Remove CC0.
23218 * reload.c: Remove CC0.
23219 * reload1.c: Remove CC0.
23220 * reorg.c: Remove CC0.
23221 * resource.c: Remove CC0.
23222 * rtl.c: Remove CC0.
23223 * rtl.def: Remove CC0.
23224 * rtl.h: Remove CC0.
23225 * rtlanal.c: Remove CC0.
23226 * sched-deps.c: Remove CC0.
23227 * sched-rgn.c: Remove CC0.
23228 * shrink-wrap.c: Remove CC0.
23229 * simplify-rtx.c: Remove CC0.
23230 * system.h: Remove CC0. Poison NOTICE_UPDATE_CC, CC_STATUS_MDEP_INIT,
23231 CC_STATUS_MDEP, and CC_STATUS.
23232 * target.def: Remove CC0.
23233 * valtrack.c: Remove CC0.
23234 * var-tracking.c: Remove CC0.
23236 2021-05-04 Richard Biener <rguenther@suse.de>
23238 PR tree-optimization/100414
23239 * tree-ssa-phiopt.c (get_non_trapping): Do not compute dominance
23241 (tree_ssa_phiopt_worker): But unconditionally here.
23243 2021-05-04 Tobias Burnus <tobias@codesourcery.com>
23245 * omp-low.c (lower_rec_input_clauses, lower_reduction_clauses): Handle
23246 && and || with floating-point and complex arguments.
23248 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
23250 * tree-inline.c (insert_debug_decl_map): Delete.
23251 (copy_debug_stmt): Minor tweak.
23252 (setup_one_parameter): Do not use a variable if the value is either
23253 a read-only DECL or a non-addressable local variable in the caller.
23254 In this case, insert the debug-only variable in the map manually.
23255 (expand_call_inline): Do not generate a CLOBBER for these values.
23256 * tree-inline.h (debug_map): Minor tweak.
23258 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
23260 * builtins.c (builtin_with_linkage_p): Return true for stp[n]cpy.
23261 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Tidy up.
23263 2021-05-04 Richard Biener <rguenther@suse.de>
23265 PR tree-optimization/100329
23266 * tree-ssa-reassoc.c (can_reassociate_p): Do not reassociate
23268 (insert_stmt_after): Assert we're not running into asm goto.
23270 2021-05-04 Richard Biener <rguenther@suse.de>
23272 PR tree-optimization/100398
23273 * tree-ssa-dse.c (pass_dse::execute): Preserve control
23276 2021-05-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23278 * builtins.c (try_store_by_multiple_pieces): Fix constfun's prototype.
23280 2021-05-04 Alexandre Oliva <oliva@adacore.com>
23282 * builtins.c (try_store_by_multiple_pieces): New.
23283 (expand_builtin_memset_args): Use it. If target_char_cast
23284 fails, proceed as for non-constant val. Pass len's ctz to...
23285 * expr.c (clear_storage_hints): ... this. Try store by
23286 multiple pieces after setmem.
23287 (clear_storage): Adjust.
23288 * expr.h (clear_storage_hints): Likewise.
23289 (try_store_by_multiple_pieces): Declare.
23290 * passes.def: Replace the last copy_prop with ccp.
23292 2021-05-03 Tom de Vries <tdevries@suse.de>
23295 * omp-low.c (lower_rec_input_clauses): Disable SIMT for user-defined
23298 2021-05-03 Richard Biener <rguenther@suse.de>
23300 * tree-ssa-dse.c (dse_classify_store): Track two PHI defs.
23302 2021-05-03 Richard Biener <rguenther@suse.de>
23304 * tree-ssa-dse.c: Do not include domwalk.h but cfganal.h.
23305 (dse_dom_walker): Remove.
23306 (dse_dom_walker::dse_optimize_stmt): Rename...
23307 (dse_optimize_stmt): ... to this, pass in live_bytes sbitmap.
23308 (dse_dom_walker::before_dom_children): Inline ...
23309 (pass_dse::execute): ... here. Perform a reverse program
23312 2021-05-03 H.J. Lu <hjl.tools@gmail.com>
23315 * configure: Regenerated.
23317 2021-05-03 Ilya Leoshkevich <iii@linux.ibm.com>
23320 * config/s390/s390.c (s390_hard_fp_reg_p): New function.
23321 (s390_md_asm_adjust): Handle hard registers.
23323 2021-05-03 Jakub Jelinek <jakub@redhat.com>
23325 PR tree-optimization/100382
23326 * tree-ssa-dse.c: Include tree-eh.h.
23327 (dse_dom_walker::before_dom_children): Don't remove stmts if
23328 stmt_unremovable_because_of_non_call_eh_p is true.
23330 2021-05-02 David Edelsohn <dje.gcc@gmail.com>
23332 * varasm.c (compute_reloc_for_var): Split out from...
23333 (get_variable_section): Use it.
23334 * output.h (compute_reloc_for_var): Declare.
23335 * config/rs6000/rs6000-protos.h
23336 (rs6000_xcoff_asm_output_aligned_decl_common): Change alignment to
23338 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address_aix):
23339 Don't append storage mapping class to symbol.
23340 (rs6000_xcoff_asm_named_section): Add BS and UL mapping classes.
23341 Don't convert TLS BSS to common.
23342 (rs6000_xcoff_unique_section): Don't fall back to select_secton.
23343 (rs6000_xcoff_section_type_flags): Add SECTION_BSS if DECL is
23345 (rs6000_xcoff_asm_globalize_decl_name): Don't strip storage
23347 (rs6000_xcoff_asm_output_aligned_decl_common): Align is unsigned int.
23348 If align is 0 from TLS class, use the same rules as varasm.c
23349 If not common, switch to BSS section manually.
23350 If common, emit appropriate comm or lcomm directive.
23351 (rs6000_xcoff_encode_section_info): Add logic to append all
23352 storage mapping classes.
23353 (rs6000_asm_weaken_decl): Adjust for qualname symbols.
23354 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
23355 rs6000_xcoff_asm_output_aligned_decl_common.
23356 (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
23357 rs6000_xcoff_asm_output_aligned_decl_common.
23358 (ASM_OUTPUT_TLS_COMMON): Use
23359 rs6000_xcoff_asm_output_aligned_decl_common.
23361 2021-05-02 Jakub Jelinek <jakub@redhat.com>
23364 * config/nvptx/nvptx.c (nvptx_sese_pseudo): Use nullptr instead of 0
23365 as first argument of pseudo_node_t constructors.
23367 2021-05-02 Jakub Jelinek <jakub@redhat.com>
23370 * config/i386/t-i386 (TM_H): Add $(srcdir)/config/i386/i386-isa.def.
23372 2021-05-01 Aldy Hernandez <aldyh@redhat.com>
23374 * value-range.cc (DEFINE_INT_RANGE_GC_STUBS): Remove.
23375 (gt_pch_nx (int_range<1> *&)): New.
23376 (gt_ggc_mx (int_range<1> *&)): New.
23377 * value-range.h (class irange): Add GTY support for
23380 2021-05-01 Geng Qi <gengqi@linux.alibaba.com>
23382 * doc/options.texi (Negative): Change either or to both and.
23384 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
23386 * config/aarch64/aarch64-simd-builtins.def: Add
23387 float_ml[as][q]_laneq builtin generator macros.
23388 * config/aarch64/aarch64-simd.md (mul_laneq<mode>3): Define.
23389 (aarch64_float_mla_laneq<mode>): Define.
23390 (aarch64_float_mls_laneq<mode>): Define.
23391 * config/aarch64/arm_neon.h (vmla_laneq_f32): Use RTL builtin
23392 instead of GCC vector extensions.
23393 (vmlaq_laneq_f32): Likewise.
23394 (vmls_laneq_f32): Likewise.
23395 (vmlsq_laneq_f32): Likewise.
23397 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
23399 * config/aarch64/aarch64-simd-builtins.def: Add
23400 float_ml[as]_lane builtin generator macros.
23401 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt<mode>):
23403 (mul_lane<mode>3): This, and re-order arguments.
23404 (aarch64_float_mla_lane<mode>): Define.
23405 (aarch64_float_mls_lane<mode>): Define.
23406 * config/aarch64/arm_neon.h (vmla_lane_f32): Use RTL builtin
23407 instead of GCC vector extensions.
23408 (vmlaq_lane_f32): Likewise.
23409 (vmls_lane_f32): Likewise.
23410 (vmlsq_lane_f32): Likewise.
23412 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
23414 * config/aarch64/aarch64-simd-builtins.def: Add float_ml[as]
23415 builtin generator macros.
23416 * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>):
23418 (aarch64_float_mls<mode>): Define.
23419 * config/aarch64/arm_neon.h (vmla_f32): Use RTL builtin
23420 instead of relying on GCC vector extensions.
23421 (vmla_f64): Likewise.
23422 (vmlaq_f32): Likewise.
23423 (vmlaq_f64): Likewise.
23424 (vmls_f32): Likewise.
23425 (vmls_f64): Likewise.
23426 (vmlsq_f32): Likewise.
23427 (vmlsq_f64): Likewise.
23428 * config/aarch64/iterators.md: Define VDQF_DF mode iterator.
23430 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
23432 * config/aarch64/aarch64-simd-builtins.def: Add
23433 float_ml[as]_n_builtin generator macros.
23434 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt_from_dup<mode>):
23436 (mul_n<mode>3): This, and re-order arguments.
23437 (aarch64_float_mla_n<mode>): Define.
23438 (aarch64_float_mls_n<mode>): Define.
23439 * config/aarch64/arm_neon.h (vmla_n_f32): Use RTL builtin
23440 instead of inline asm.
23441 (vmlaq_n_f32): Likewise.
23442 (vmls_n_f32): Likewise.
23443 (vmlsq_n_f32): Likewise.
23445 2021-04-30 Jonathan Wright <joanthan.wright@arm.com>
23447 * config/aarch64/aarch64-simd-builtins.def: Add pmull[2]
23448 builtin generator macros.
23449 * config/aarch64/aarch64-simd.md (aarch64_pmullv8qi): Define.
23450 (aarch64_pmull_hiv16qi_insn): Define.
23451 (aarch64_pmull_hiv16qi): Define.
23452 * config/aarch64/arm_neon.h (vmull_high_p8): Use RTL builtin
23453 instead of inline asm.
23454 (vmull_p8): Likewise.
23456 2021-04-30 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
23458 * config/avr/avr.md: Adjust peepholes to match and
23459 generate parallels with clobber of REG_CC.
23460 (mov<mode>_insn): Rename to mov<mode>_insn_split.
23461 (*mov<mode>_insn): Rename to mov<mode>_insn.
23463 2021-04-30 David Edelsohn <dje.gcc@gmail.com>
23465 * varasm.c (use_blocks_for_decl_p): Don't use section anchors
23466 for VAR_DECLs if -fdata-sections enabled.
23468 2021-04-30 Michael Meissner <meissner@linux.ibm.com>
23470 PR bootstrap/100327
23471 * config/rs6000/rs6000.c
23472 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Define.
23473 (rs6000_libgcc_floating_mode_supported_p): New target hook.
23475 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
23477 * tree-ssa-threadbackward.c (class thread_jumps): Split out code
23479 (class back_threader_registry): ...to here...
23480 (class back_threader_profitability): ...and here...
23481 (thread_jumps::thread_through_all_blocks): Remove argument.
23482 (back_threader_registry::back_threader_registry): New.
23483 (back_threader_registry::~back_threader_registry): New.
23484 (back_threader_registry::thread_through_all_blocks): New.
23485 (thread_jumps::profitable_jump_thread_path): Move from here...
23486 (back_threader_profitability::profitable_path_p): ...to here.
23487 (thread_jumps::find_taken_edge): New.
23488 (thread_jumps::convert_and_register_current_path): Move...
23489 (back_threader_registry::register_path): ...to here.
23490 (thread_jumps::register_jump_thread_path_if_profitable): Move...
23491 (thread_jumps::maybe_register_path): ...to here.
23492 (thread_jumps::handle_phi): Call find_taken_edge and
23493 maybe_register_path.
23494 (thread_jumps::handle_assignment): Same.
23495 (thread_jumps::fsm_find_control_statement_thread_paths): Remove
23496 tree argument to handle_phi and handle_assignment.
23497 (thread_jumps::find_jump_threads_backwards): Set m_name. Remove
23498 set of m_speed_p and m_max_threaded_paths.
23499 (pass_thread_jumps::execute): Remove second argument from
23500 find_jump_threads_backwards.
23501 (pass_early_thread_jumps::execute): Same.
23503 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
23505 * tree-ssa-dom.c (class dom_jump_threader_simplifier): New.
23506 (class dom_opt_dom_walker): Initialize some class variables.
23507 (pass_dominator::execute): Pass evrp_range_analyzer and
23508 dom_jump_threader_simplifier to dom_opt_dom_walker.
23509 Adjust for some functions moving into classes.
23510 (simplify_stmt_for_jump_threading): Adjust and move to...
23511 (jump_threader_simplifier::simplify): ...here.
23512 (dom_opt_dom_walker::before_dom_children): Adjust for
23513 m_evrp_range_analyzer.
23514 (dom_opt_dom_walker::after_dom_children): Remove x_vr_values hack.
23515 (test_for_singularity): Place in dom_opt_dom_walker class.
23516 (dom_opt_dom_walker::optimize_stmt): The argument
23517 evrp_range_analyzer is now a class field.
23518 * tree-ssa-threadbackward.c (class thread_jumps): Add m_registry.
23519 (thread_jumps::thread_through_all_blocks): New.
23520 (thread_jumps::convert_and_register_current_path): Use m_registry.
23521 (pass_thread_jumps::execute): Adjust for thread_through_all_blocks
23522 being in the threader class.
23523 (pass_early_thread_jumps::execute): Same.
23524 * tree-ssa-threadedge.c (threadedge_initialize_values): Move...
23525 (jump_threader::jump_threader): ...here.
23526 (threadedge_finalize_values): Move...
23527 (jump_threader::~jump_threader): ...here.
23528 (jump_threader::remove_jump_threads_including): New.
23529 (jump_threader::thread_through_all_blocks): New.
23530 (record_temporary_equivalences_from_phis): Move...
23531 (jump_threader::record_temporary_equivalences_from_phis): ...here.
23532 (record_temporary_equivalences_from_stmts_at_dest): Move...
23533 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
23535 (simplify_control_stmt_condition_1): Move to jump_threader class.
23536 (simplify_control_stmt_condition): Move...
23537 (jump_threader::simplify_control_stmt_condition): ...here.
23538 (thread_around_empty_blocks): Move...
23539 (jump_threader::thread_around_empty_blocks): ...here.
23540 (thread_through_normal_block): Move...
23541 (jump_threader::thread_through_normal_block): ...here.
23542 (thread_across_edge): Move...
23543 (jump_threader::thread_across_edge): ...here.
23544 (thread_outgoing_edges): Move...
23545 (jump_threader::thread_outgoing_edges): ...here.
23546 * tree-ssa-threadedge.h: Move externally facing functings...
23547 (class jump_threader): ...here...
23548 (class jump_threader_simplifier): ...and here.
23549 * tree-ssa-threadupdate.c (struct redirection_data): Remove comment.
23550 (jump_thread_path_allocator::jump_thread_path_allocator): New.
23551 (jump_thread_path_allocator::~jump_thread_path_allocator): New.
23552 (jump_thread_path_allocator::allocate_thread_edge): New.
23553 (jump_thread_path_allocator::allocate_thread_path): New.
23554 (jump_thread_path_registry::jump_thread_path_registry): New.
23555 (jump_thread_path_registry::~jump_thread_path_registry): New.
23556 (jump_thread_path_registry::allocate_thread_edge): New.
23557 (jump_thread_path_registry::allocate_thread_path): New.
23558 (dump_jump_thread_path): Make extern.
23559 (debug (const vec<jump_thread_edge *> &path)): New.
23560 (struct removed_edges): Move to tree-ssa-threadupdate.h.
23561 (struct thread_stats_d): Remove.
23562 (remove_ctrl_stmt_and_useless_edges): Make static.
23563 (lookup_redirection_data): Move...
23564 (jump_thread_path_registry::lookup_redirection_data): ...here.
23565 (ssa_redirect_edges): Make static.
23566 (thread_block_1): Move...
23567 (jump_thread_path_registry::thread_block_1): ...here.
23568 (thread_block): Move...
23569 (jump_thread_path_registry::thread_block): ...here.
23570 (thread_through_loop_header): Move...
23571 (jump_thread_path_registry::thread_through_loop_header): ...here.
23572 (mark_threaded_blocks): Move...
23573 (jump_thread_path_registry::mark_threaded_blocks): ...here.
23574 (debug_path): Move...
23575 (jump_thread_path_registry::debug_path): ...here.
23576 (debug_all_paths): Move...
23577 (jump_thread_path_registry::dump): ..here.
23578 (rewire_first_differing_edge): Move...
23579 (jump_thread_path_registry::rewire_first_differing_edge): ...here.
23580 (adjust_paths_after_duplication): Move...
23581 (jump_thread_path_registry::adjust_paths_after_duplication): ...here.
23582 (duplicate_thread_path): Move...
23583 (jump_thread_path_registry::duplicate_thread_path): ..here.
23584 (remove_jump_threads_including): Move...
23585 (jump_thread_path_registry::remove_jump_threads_including): ...here.
23586 (thread_through_all_blocks): Move to...
23587 (jump_thread_path_registry::thread_through_all_blocks): ...here.
23588 (delete_jump_thread_path): Remove.
23589 (register_jump_thread): Move...
23590 (jump_thread_path_registry::register_jump_thread): ...here.
23591 * tree-ssa-threadupdate.h: Move externally facing functions...
23592 (class jump_thread_path_allocator): ...here...
23593 (class jump_thread_path_registry): ...and here.
23594 (thread_through_all_blocks): Remove.
23595 (struct removed_edges): New.
23596 (register_jump_thread): Remove.
23597 (remove_jump_threads_including): Remove.
23598 (delete_jump_thread_path): Remove.
23599 (remove_ctrl_stmt_and_useless_edges): Remove.
23600 (free_dom_edge_info): New prototype.
23601 * tree-vrp.c: Remove x_vr_values hack.
23602 (class vrp_jump_threader_simplifier): New.
23603 (vrp_jump_threader_simplifier::simplify): New.
23604 (vrp_jump_threader::vrp_jump_threader): Adjust method signature.
23605 Remove m_dummy_cond.
23606 Instantiate m_simplifier and m_threader.
23607 (vrp_jump_threader::thread_through_all_blocks): New.
23608 (vrp_jump_threader::simplify_stmt): Remove.
23609 (vrp_jump_threader::after_dom_children): Do not set m_dummy_cond.
23610 Remove x_vr_values hack.
23611 (execute_vrp): Adjust for thread_through_all_blocks being in a
23614 2021-04-30 Christophe Lyon <christophe.lyon@linaro.org>
23616 * genflags.c (gen_insn): Print failed expansion string.
23618 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
23620 * expr.c (alignment_for_piecewise_move): Call mode_for_size
23621 without limit to MAX_FIXED_MODE_SIZE.
23623 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
23625 PR middle-end/90773
23626 * builtins.c (builtin_memset_gen_str): Don't use return from
23627 simplify_gen_subreg.
23629 2021-04-30 Uroš Bizjak <ubizjak@gmail.com>
23632 * config/i386/i386.md (*add<mode>3_carry_0r): New insn pattern.
23633 (*addsi3_carry_zext_0r): Ditto.
23634 (*sub<mode>3_carry_0): Ditto.
23635 (*subsi3_carry_zext_0r): Ditto.
23636 * config/i386/predicates.md (ix86_carry_flag_unset_operator):
23638 * config/i386/i386.c (ix86_rtx_costs) <case PLUS, case MINUS>:
23639 Also consider ix86_carry_flag_unset_operator to calculate
23640 the cost of adc/sbb insn.
23642 2021-04-30 Roman Zhuykov <zhroma@ispras.ru>
23644 PR rtl-optimization/100225
23645 PR rtl-optimization/84878
23646 * modulo-sched.c (sms_schedule): Use note_stores to skip loops
23647 where we have an instruction which touches (writes) any hard
23648 register from df->regular_block_artificial_uses set.
23649 Allow not-single-set instruction only right before basic block
23652 2021-04-30 Geng Qi <gengqi@linux.alibaba.com>
23654 * config/riscv/riscv.opt (march=,mabi=): Negative itself.
23656 2021-04-30 LevyHsu <admin@levyhsu.com>
23658 * config/riscv/riscv.c (riscv_min_arithmetic_precision): New.
23659 * config/riscv/riscv.h (TARGET_MIN_ARITHMETIC_PRECISION): New.
23660 * config/riscv/riscv.md (addv<mode>4, uaddv<mode>4): New.
23661 (subv<mode>4, usubv<mode>4, mulv<mode>4, umulv<mode>4): New.
23663 2021-04-29 Alexandre Oliva <oliva@adacore.com>
23665 * config.gcc: Merged x86 and x86_64 cpu_type-setting cases.
23667 2021-04-29 Alexandre Oliva <oliva@adacore.com>
23669 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Rename to...
23670 (ASM_OUTPUT_MAX_SKIP_ALIGN): ... this. Enclose in do/while(0).
23671 * config/i386/i386.c: Adjust.
23672 * config/i386/i386.md: Adjust.
23673 * config/i386/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Drop.
23674 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
23675 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
23676 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
23677 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
23678 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
23679 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
23680 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
23681 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
23682 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
23683 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
23685 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
23687 * config/i386/i386-expand.c (ix86_expand_int_compare):
23688 Swap operands of GTU and LEU comparison to emit carry flag comparison.
23689 * config/i386/i386.md (*add<mode>3_carry_0): Change insn
23690 predicate to allow more combine opportunities with memory operands.
23691 (*sub<mode>3_carry_0): Ditto.
23693 2021-04-29 Richard Sandiford <richard.sandiford@arm.com>
23695 PR rtl-optimization/100303
23696 * rtl-ssa/accesses.cc (function_info::make_use_available): Take a
23697 boolean that indicates whether the use will only be used in
23698 debug instructions. Treat it in the same way that existing
23699 cross-EBB debug references would be handled if so.
23700 (function_info::make_uses_available): Likewise.
23701 * rtl-ssa/functions.h (function_info::make_uses_available): Update
23702 prototype accordingly.
23703 (function_info::make_uses_available): Likewise.
23704 * fwprop.c (try_fwprop_subst): Update call accordingly.
23706 2021-04-29 Jeff Law <jlaw@tachyum.com>
23708 * config/nios2/nios2-protos.h (nios2_fpu_insn_enabled): Move outside
23711 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
23712 Richard Biener <rguenther@suse.de>
23715 * config/i386/i386-builtin.def (IX86_BUILTIN_MASKLOADPD)
23716 (IX86_BUILTIN_MASKLOADPS, IX86_BUILTIN_MASKLOADPD256)
23717 (IX86_BUILTIN_MASKLOADPS256, IX86_BUILTIN_MASKLOADD)
23718 (IX86_BUILTIN_MASKLOADQ, IX86_BUILTIN_MASKLOADD256)
23719 (IX86_BUILTIN_MASKLOADQ256): Move from SPECIAL_ARGS
23720 to PURE_ARGS category.
23721 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
23722 Handle PURE_ARGS category.
23723 * config/i386/i386-expand.c (ix86_expand_builtin): Ditto.
23725 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
23727 * configure.ac: Check for the presence of sys/locking.h header and
23728 for whether _LK_LOCK is supported by _locking.
23729 * configure: Regenerate.
23730 * config.in: Likewise.
23731 * gcov-io.h: Define GCOV_LOCKED_WITH_LOCKING if HOST_HAS_LK_LOCK.
23732 * gcov-io.c (gcov_open): Add support for GCOV_LOCKED_WITH_LOCKING.
23733 * system.h: Include <sys/locking.h> if HAVE_SYS_LOCKING_H.
23735 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
23737 * config/i386/predicates.md (fcmov_comparison_operator):
23738 Do not check for trivial FP comparison operator.
23739 <case GEU, case LTU>: Allow CCGZmode.
23740 <case GTU, case LEU>: Do not allow CCCmode.
23741 (ix86_comparison_operator) <case GTU, case LEU>: Allow only CCmode.
23742 (ix86_carry_flag_operator): Match only LTU and UNLT code.
23743 Do not check for trivial FP comparison operator. Allow CCGZmode.
23745 2021-04-29 Tom de Vries <tdevries@suse.de>
23747 * omp-expand.c (expand_omp_simd): Add step_orig, and replace uses of
23748 fd->loop.step by either step or orig_step.
23750 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
23752 * config/sparc/sparc.c (gen_load_pcrel_sym): Delete.
23753 (load_got_register): Do the PIC dance here.
23754 (sparc_legitimize_tls_address): Simplify.
23755 (sparc_emit_probe_stack_range): Likewise.
23756 (sparc32_initialize_trampoline): Likewise.
23757 (sparc64_initialize_trampoline): Likewise.
23758 * config/sparc/sparc.md (load_pcrel_sym<P:mode>): Add @ marker.
23759 (probe_stack_range<P:mode>): Likewise.
23760 (flush<P:mode>): Likewise.
23761 (tgd_hi22<P:mode>): Likewise.
23762 (tgd_lo10<P:mode>): Likewise.
23763 (tgd_add<P:mode>): Likewise.
23764 (tgd_call<P:mode>): Likewise.
23765 (tldm_hi22<P:mode>): Likewise.
23766 (tldm_lo10<P:mode>): Likewise.
23767 (tldm_add<P:mode>): Likewise.
23768 (tldm_call<P:mode>): Likewise.
23769 (tldo_hix22<P:mode>): Likewise.
23770 (tldo_lox10<P:mode>): Likewise.
23771 (tldo_add<P:mode>): Likewise.
23772 (tie_hi22<P:mode>): Likewise.
23773 (tie_lo10<P:mode>): Likewise.
23774 (tie_add<P:mode>): Likewise.
23775 (tle_hix22<P:mode>): Likewise.
23776 (tle_lox10<P:mode>): Likewise.
23777 (stack_protect_setsi): Rename to...
23778 (stack_protect_set32): ...this.
23779 (stack_protect_setdi): Rename to...
23780 (stack_protect_set64): ...this.
23781 (stack_protect_set): Adjust calls to above.
23782 (stack_protect_testsi): Rename to...
23783 (stack_protect_test32): ...this.
23784 (stack_protect_testdi): Rename to...
23785 (stack_protect_test64): ...this.
23786 (stack_protect_test): Adjust calls to above.
23788 2021-04-29 H.J. Lu <hjl.tools@gmail.com>
23790 PR middle-end/90773
23791 * builtins.c (builtin_memcpy_read_str): Add a dummy argument.
23792 (builtin_strncpy_read_str): Likewise.
23793 (builtin_memset_read_str): Add an argument for the previous RTL
23794 information and generate the new RTL from the previous RTL info.
23795 (builtin_memset_gen_str): Likewise.
23796 * builtins.h (builtin_strncpy_read_str): Update the prototype.
23797 (builtin_memset_read_str): Likewise.
23798 * expr.c (by_pieces_ninsns): If targetm.overlap_op_by_pieces_p()
23799 returns true, round up size and alignment to the widest integer
23800 mode for maximum size.
23801 (pieces_addr::adjust): Add a pointer to by_pieces_prev argument
23802 and pass it to m_constfn.
23803 (op_by_pieces_d): Add m_push and m_overlap_op_by_pieces.
23804 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
23805 initialize m_push. Initialize m_overlap_op_by_pieces with
23806 targetm.overlap_op_by_pieces_p ().
23807 (op_by_pieces_d::run): Pass the previous RTL information to
23808 pieces_addr::adjust and generate overlapping operations if
23809 m_overlap_op_by_pieces is true.
23811 (move_by_pieces_d::move_by_pieces_d): Updated for op_by_pieces_d
23813 (store_by_pieces_d::store_by_pieces_d): Updated for op_by_pieces_d
23815 (can_store_by_pieces): Use by_pieces_constfn on constfun.
23816 (store_by_pieces): Use by_pieces_constfn on constfun. Updated
23817 for op_by_pieces_d change.
23818 (clear_by_pieces_1): Add a dummy argument.
23819 (clear_by_pieces): Updated for op_by_pieces_d change.
23820 (compare_by_pieces_d::compare_by_pieces_d): Likewise.
23821 (string_cst_read_str): Add a dummy argument.
23822 * expr.h (by_pieces_constfn): Add a dummy argument.
23823 (by_pieces_prev): New.
23824 * target.def (overlap_op_by_pieces_p): New target hook.
23825 * config/i386/i386.c (TARGET_OVERLAP_OP_BY_PIECES_P): New.
23826 * doc/tm.texi.in: Add TARGET_OVERLAP_OP_BY_PIECES_P.
23827 * doc/tm.texi: Regenerated.
23829 2021-04-29 Richard Biener <rguenther@suse.de>
23831 PR tree-optimization/100253
23832 * tree-vect-stmts.c (vectorizable_load): Do not assume
23833 element alignment when DR_MISALIGNMENT is -1.
23834 (vectorizable_store): Likewise.
23836 2021-04-29 Jakub Jelinek <jakub@redhat.com>
23839 * config/aarch64/aarch64.c (aarch64_add_offset_1_temporaries): Use
23840 absu_hwi instead of abs_hwi.
23842 2021-04-29 Richard Biener <rguenther@suse.de>
23844 PR middle-end/38474
23845 * tree-ssa-structalias.c (add_graph_edge): Avoid direct
23846 forwarding when indirect forwarding through ESCAPED
23849 2021-04-29 Tom de Vries <tdevries@suse.de>
23852 * internal-fn.c (expand_GOMP_SIMT_ENTER_ALLOC)
23853 (expand_GOMP_SIMT_LAST_LANE, expand_GOMP_SIMT_ORDERED_PRED)
23854 (expand_GOMP_SIMT_VOTE_ANY, expand_GOMP_SIMT_XCHG_BFLY)
23855 (expand_GOMP_SIMT_XCHG_IDX): Ensure target is assigned to.
23857 2021-04-29 Richard Biener <rguenther@suse.de>
23859 PR tree-optimization/99912
23860 * tree-ssa-dse.c (dse_dom_walker::m_need_cfg_cleanup): New.
23861 (dse_dom_walker::todo): Likewise.
23862 (dse_dom_walker::dse_optimize_stmt): Move VDEF check to the
23864 (dse_dom_walker::before_dom_children): Remove trivially
23865 dead SSA defs and schedule CFG cleanup if we removed all
23867 (pass_dse::execute): Get TODO as computed by the DOM walker
23868 and return it. Wipe dominator info earlier.
23870 2021-04-29 Richard Biener <rguenther@suse.de>
23873 * ipa-prop.c (ipcp_modif_dom_walker::before_dom_children):
23874 Track blocks to cleanup EH in new m_need_eh_cleanup.
23875 (ipcp_modif_dom_walker::cleanup_eh): New.
23876 (ipcp_transform_function): Release dominator info before
23879 2021-04-29 Martin Sebor <msebor@redhat.com>
23881 PR middle-end/100250
23882 * attribs.c (attr_access::array_as_string): Avoid dereferencing
23883 a pointer when it's null.
23885 2021-04-29 Martin Sebor <msebor@redhat.com>
23887 * Makefile.in (OBJS): Add ipa-free-lang-data.o.
23888 * ipa-free-lang-data.cc: New file.
23889 * tree.c: Move pass free_lang_data to file above.
23890 (build_array_type_1): Declare extern.
23891 * tree.h (build_array_type_1): Declare.
23893 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
23895 * config/aarch64/aarch64-simd-builtins.def: Modify comment to
23896 make consistent with updated RTL pattern.
23897 * config/aarch64/aarch64-simd.md (aarch64_<sur>qmovn<mode>):
23898 Implement using ss_truncate and us_truncate rather than
23900 * config/aarch64/iterators.md: Remove redundant unspecs and
23901 iterator: UNSPEC_[SU]QXTN and SUQMOVN respectively.
23903 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
23905 * config/aarch64/arm_acle.h (__attribute__): Make intrinsic
23906 attributes consistent with those defined in arm_neon.h.
23908 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
23910 * config/aarch64/arm_fp16.h (__attribute__): Make intrinsic
23911 attributes consistent with those defined in arm_neon.h.
23913 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
23915 * config/aarch64/aarch64-simd-builtins.def: Add
23916 float_trunc_rodd builtin generator macros.
23917 * config/aarch64/aarch64-simd.md (aarch64_float_trunc_rodd_df):
23919 (aarch64_float_trunc_rodd_lo_v2sf): Define.
23920 (aarch64_float_trunc_rodd_hi_v4sf_le): Define.
23921 (aarch64_float_trunc_rodd_hi_v4sf_be): Define.
23922 (aarch64_float_trunc_rodd_hi_v4sf): Define.
23923 * config/aarch64/arm_neon.h (vcvtx_f32_f64): Use RTL builtin
23924 instead of inline asm.
23925 (vcvtx_high_f32_f64): Likewise.
23926 (vcvtxd_f32_f64): Likewise.
23927 * config/aarch64/iterators.md: Add FCVTXN unspec.
23929 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
23931 * config/aarch64/aarch64-simd-builtins.def: Add tbx1 builtin
23933 * config/aarch64/aarch64-simd.md (aarch64_tbx1<mode>):
23935 * config/aarch64/arm_neon.h (vqtbx1_s8): USE RTL builtin
23936 instead of inline asm.
23937 (vqtbx1_u8): Likewise.
23938 (vqtbx1_p8): Likewise.
23939 (vqtbx1q_s8): Likewise.
23940 (vqtbx1q_u8): Likewise.
23941 (vqtbx1q_p8): Likewise.
23942 (vtbx2_s8): Likewise.
23943 (vtbx2_u8): Likewise.
23944 (vtbx2_p8): Likewise.
23946 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
23948 * config/aarch64/aarch64-simd-builtins.def: Add tbl1 builtin
23950 * config/aarch64/arm_neon.h (vqtbl1_p8): Use RTL builtin
23951 instead of inline asm.
23952 (vqtbl1_s8): Likewise.
23953 (vqtbl1_u8): Likewise.
23954 (vqtbl1q_p8): Likewise.
23955 (vqtbl1q_s8): Likewise.
23956 (vqtbl1q_u8): Likewise.
23957 (vtbl1_s8): Likewise.
23958 (vtbl1_u8): Likewise.
23959 (vtbl1_p8): Likewise.
23960 (vtbl2_s8): Likewise.
23961 (vtbl2_u8): Likewise.
23962 (vtbl2_p8): Likewise.
23964 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
23966 * config/aarch64/aarch64-simd-builtins.def: Add polynomial
23967 ssri_n buitin generator macro.
23968 * config/aarch64/arm_neon.h (vsri_n_p8): Use RTL builtin
23969 instead of inline asm.
23970 (vsri_n_p16): Likewise.
23971 (vsri_n_p64): Likewise.
23972 (vsriq_n_p8): Likewise.
23973 (vsriq_n_p16): Likewise.
23974 (vsriq_n_p64): Likewise.
23976 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
23978 * config/aarch64/aarch64-simd-builtins.def: Use VALLP mode
23979 iterator for polynomial ssli_n builtin generator macro.
23980 * config/aarch64/arm_neon.h (vsli_n_p8): Use RTL builtin
23981 instead of inline asm.
23982 (vsli_n_p16): Likewise.
23983 (vsliq_n_p8): Likewise.
23984 (vsliq_n_p16): Likewise.
23985 * config/aarch64/iterators.md: Define VALLP mode iterator.
23987 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
23989 * config/aarch64/aarch64-simd-builtins.def: Use VDQV_L
23990 iterator to generate [su]adalp RTL builtins.
23991 * config/aarch64/aarch64-simd.md: Use VDQV_L iterator in
23992 [su]adalp RTL pattern.
23993 * config/aarch64/arm_neon.h (vpadal_s32): Use RTL builtin
23994 instead of inline asm.
23995 (vpadal_u32): Likewise.
23997 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
23999 * config/aarch64/aarch64-simd-builtins.def: Add [su]addlp
24000 builtin generator macros.
24001 * config/aarch64/aarch64-simd.md (aarch64_<su>addlp<mode>):
24003 * config/aarch64/arm_neon.h (vpaddl_s8): Use RTL builtin
24004 instead of inline asm.
24005 (vpaddl_s16): Likewise.
24006 (vpaddl_s32): Likewise.
24007 (vpaddl_u8): Likewise.
24008 (vpaddl_u16): Likewise.
24009 (vpaddl_u32): Likewise.
24010 (vpaddlq_s8): Likewise.
24011 (vpaddlq_s16): Likewise.
24012 (vpaddlq_s32): Likewise.
24013 (vpaddlq_u8): Likewise.
24014 (vpaddlq_u16): Likewise.
24015 (vpaddlq_u32): Liwewise.
24016 * config/aarch64/iterators.md: Define [SU]ADDLP unspecs with
24017 appropriate attributes.
24019 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
24021 * config/aarch64/aarch64-simd-builtins.def: Use VDQ_I iterator
24022 for aarch64_addp<mode> builtin macro generator.
24023 * config/aarch64/aarch64-simd.md: Use VDQ_I iterator in
24024 aarch64_addp<mode> RTL pattern.
24025 * config/aarch64/arm_neon.h (vpaddq_s8): Use RTL builtin
24026 instead of inline asm.
24027 (vpaddq_s16): Likewise.
24028 (vpaddq_s32): Likewise.
24029 (vpaddq_s64): Likewise.
24030 (vpaddq_u8): Likewise.
24031 (vpaddq_u16): Likewise.
24032 (vpaddq_u32): Likewise.
24033 (vpaddq_u64): Likewise.
24035 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
24037 * config/aarch64/aarch64-simd-builtins.def: Add sq[r]dmulh_n
24038 builtin generator macros.
24039 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_n<mode>):
24041 * config/aarch64/arm_neon.h (vqdmulh_n_s16): Use RTL builtin
24042 instead of inline asm.
24043 (vqdmulh_n_s32): Likewise.
24044 (vqdmulhq_n_s16): Likewise.
24045 (vqdmulhq_n_s32): Likewise.
24046 (vqrdmulh_n_s16): Likewise.
24047 (vqrdmulh_n_s32): Likewise.
24048 (vqrdmulhq_n_s16): Likewise.
24049 (vqrdmulhq_n_s32): Likewise.
24051 2021-04-28 Tobias Burnus <tobias@codesourcery.com>
24053 * doc/install.texi (--enable-offload-defaulted): Document.
24055 2021-04-28 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
24057 * config/avr/avr-dimode.md: Turn existing patterns into
24058 define_insn_and_split style patterns where the splitter
24059 adds a clobber of the condition code register. Drop "cc"
24060 attribute. Add new patterns to match output of
24062 * config/avr/avr-fixed.md: Likewise.
24063 * config/avr/avr.c (cc_reg_rtx): New.
24064 (avr_parallel_insn_from_insns): Adjust insn count
24065 for removal of set of cc0.
24066 (avr_is_casesi_sequence): Likewise.
24067 (avr_casei_sequence_check_operands): Likewise.
24068 (avr_optimize_casesi): Likewise. Also insert
24069 new insns after jump_insn.
24070 (avr_pass_casesi::avr_rest_of_handle_casesi): Adjust
24071 for removal of set of cc0.
24072 (avr_init_expanders): Initialize cc_reg_rtx.
24073 (avr_regno_reg_class): Handle REG_CC.
24074 (cond_string): Remove usage of CC_OVERFLOW_UNUSABLE.
24075 (avr_notice_update_cc): Remove function.
24076 (ret_cond_branch): Remove usage of CC_OVERFLOW_UNUSABLE.
24077 (compare_condition): Adjust for PARALLEL with
24079 (out_shift_with_cnt): Likewise.
24080 (ashlhi3_out): Likewise.
24081 (ashrhi3_out): Likewise.
24082 (lshrhi3_out): Likewise.
24083 (avr_class_max_nregs): Return single reg for REG_CC.
24084 (avr_compare_pattern): Check for REG_CC instead
24086 (avr_reorg_remove_redundant_compare): Likewise.
24087 (avr_reorg):Adjust for PARALLEL with REG_CC clobber.
24088 (avr_hard_regno_nregs): Return single reg for REG_CC.
24089 (avr_hard_regno_mode_ok): Allow only CCmode for REG_CC.
24090 (avr_md_asm_adjust): Clobber REG_CC.
24091 (TARGET_HARD_REGNO_NREGS): Define.
24092 (TARGET_CLASS_MAX_NREGS): Define.
24093 (TARGET_MD_ASM_ADJUST): Define.
24094 * config/avr/avr.h (FIRST_PSEUDO_REGISTER): Adjust
24096 (enum reg_class): Add CC_REG class.
24097 (NOTICE_UPDATE_CC): Remove.
24098 (CC_OVERFLOW_UNUSABLE): Remove.
24099 (CC_NO_CARRY): Remove.
24100 * config/avr/avr.md: Turn existing patterns into
24101 define_insn_and_split style patterns where the splitter
24102 adds a clobber of the condition code register. Drop "cc"
24103 attribute. Add new patterns to match output of
24105 (sez): Remove unused pattern.
24107 2021-04-28 Richard Earnshaw <rearnsha@arm.com>
24110 * config/arm/arm.c (arm_hard_regno_mode_ok): Only allow VPR to be
24113 2021-04-28 Richard Sandiford <richard.sandiford@arm.com>
24116 * config/aarch64/constraints.md (Utq): Require the address to
24117 be valid for both the element mode and for V2DImode.
24119 2021-04-28 Jakub Jelinek <jakub@redhat.com>
24120 Tobias Burnus <tobias@codesourcery.com>
24122 * configure.ac (OFFLOAD_DEFAULTED): AC_DEFINE if offload-defaulted.
24123 * gcc.c (process_command): New variable.
24124 (driver::maybe_putenv_OFFLOAD_TARGETS): If OFFLOAD_DEFAULTED,
24125 set it if -foffload is defaulted.
24126 * lto-wrapper.c (OFFLOAD_TARGET_DEFAULT_ENV): Define.
24127 (compile_offload_image): If OFFLOAD_DEFAULTED and
24128 OFFLOAD_TARGET_DEFAULT is in the environment, don't fail
24129 if corresponding mkoffload can't be found.
24130 (compile_images_for_offload_targets): Likewise. Free and clear
24131 offload_names if no valid offload is found.
24132 * config.in: Regenerate.
24133 * configure: Regenerate.
24135 2021-04-28 Richard Biener <rguenther@suse.de>
24137 PR tree-optimization/100292
24138 * tree-vect-generic.c (expand_vector_condition): Do not fold
24141 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
24143 * config/rs6000/aix.h (SUBTARGET_DRIVER_SELF_SPECS): New.
24144 * config/rs6000/aix64.opt (m64): New.
24147 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
24149 * config/vax/vax.c (print_operand_address, vax_address_cost_1)
24150 (index_term_p): Handle ASHIFT too.
24152 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
24154 * config/vax/builtins.md (jbb<ccss>i<mode>): Remove operand #3.
24155 (sync_lock_test_and_set<mode>): Adjust accordingly.
24156 (sync_lock_release<mode>): Likewise.
24158 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
24160 * config/vax/vax-protos.h (adjacent_operands_p): Remove
24162 * config/vax/vax.c (adjacent_operands_p): Remove.
24164 2021-04-27 Maciej W. Rozycki <macro@linux-mips.org>
24166 * ifcvt.c (dead_or_predicable) [!IFCVT_MODIFY_TESTS]: Fall
24167 through to the non-conditional execution case if getting the
24168 condition for conditional execution has failed.
24170 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
24172 PR middle-end/100284
24173 * gimple.c (gimple_could_trap_p_1): Remove VEC_COND_EXPR test.
24174 * tree-eh.c (operation_could_trap_p): Handle VEC_COND_EXPR rather
24175 than asserting on it.
24177 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
24179 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): Protect
24180 with TARGET_AIX_OS.
24182 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
24185 * calls.c (precompute_register_parameters): Additionally test
24186 targetm.precompute_tls_p to pre-compute argument.
24187 * config/rs6000/aix.h (TARGET_PRECOMPUTE_TLS_P): Define.
24188 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): New.
24189 * target.def (precompute_tls_p): New.
24190 * doc/tm.texi.in (TARGET_PRECOMPUTE_TLS_P): Add hook documentation.
24191 * doc/tm.texi: Regenerated.
24193 2021-04-27 Jakub Jelinek <jakub@redhat.com>
24196 * config/aarch64/aarch64.c (aarch64_print_operand): Cast -UINTVAL
24197 back to HOST_WIDE_INT.
24199 2021-04-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
24202 * simplify-rtx.c (simplify_context::simplify_subreg): Check the
24203 memory alignment for the outer mode.
24205 2021-04-27 H.J. Lu <hjl.tools@gmail.com>
24207 PR middle-end/90773
24208 * expr.c (op_by_pieces_d::get_usable_mode): New member function.
24209 (op_by_pieces_d::run): Cange a while loop to a do-while loop.
24211 2021-04-27 Alex Coplan <alex.coplan@arm.com>
24214 * config/arm/arm.c (arm_split_compare_and_swap): Fix up codegen
24215 with negative immediates: ensure we expand cbranchsi4_scratch
24216 correctly and ensure we satisfy its constraints.
24217 * config/arm/sync.md
24218 (@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): Don't
24219 attempt to tie two output operands together with constraints;
24220 collapse two alternatives.
24221 (@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
24222 * config/arm/thumb1.md (cbranchsi4_neg_late): New.
24224 2021-04-27 Jakub Jelinek <jakub@redhat.com>
24227 * config/aarch64/predicates.md (aarch64_sub_immediate,
24228 aarch64_plus_immediate): Use -UINTVAL instead of -INTVAL.
24229 * config/aarch64/aarch64.md (casesi, rotl<mode>3): Likewise.
24230 * config/aarch64/aarch64.c (aarch64_print_operand,
24231 aarch64_split_atomic_op, aarch64_expand_subvti): Likewise.
24233 2021-04-27 Jakub Jelinek <jakub@redhat.com>
24235 PR tree-optimization/100239
24236 * tree-vect-generic.c (lower_vec_perm): Don't accept constant
24237 permutations with all indices from the first zero element as vec_shl.
24239 2021-04-27 Jakub Jelinek <jakub@redhat.com>
24241 PR rtl-optimization/100254
24242 * cfgcleanup.c (outgoing_edges_match): Check REG_EH_REGION on
24243 last1 and last2 insns rather than BB_END (bb1) and BB_END (bb2) insns.
24245 2021-04-27 Richard Biener <rguenther@suse.de>
24247 PR tree-optimization/99912
24248 * passes.def: Add comment about new TODO_remove_unused_locals.
24249 * tree-stdarg.c (pass_data_stdarg): Run TODO_remove_unused_locals
24252 2021-04-27 Richard Biener <rguenther@suse.de>
24254 PR tree-optimization/99912
24255 * passes.def (pass_all_optimizations): Add pass_dse before
24256 the first pass_dce, move the first pass_dse before the
24257 pass_dce following pass_pre.
24259 2021-04-27 Jakub Jelinek <jakub@redhat.com>
24261 PR tree-optimization/95527
24262 * generic-match-head.c: Include tm.h.
24263 * gimple-match-head.c: Include tm.h.
24264 * match.pd (CLZ == INTEGER_CST): Don't use
24265 #ifdef CLZ_DEFINED_VALUE_AT_ZERO, only test CLZ_DEFINED_VALUE_AT_ZERO
24266 if clz == CFN_CLZ. Add missing val declaration.
24267 (CTZ cmp CST): New simplifications.
24269 2021-04-27 Jakub Jelinek <jakub@redhat.com>
24271 PR tree-optimization/96696
24272 * expr.c (expand_expr_divmod): New function.
24273 (expand_expr_real_2) <case TRUNC_DIV_EXPR>: Use it for truncations and
24274 divisions. Formatting fixes.
24275 <case MULT_EXPR>: Optimize x / y * y as x - x % y if the latter is
24278 2021-04-27 Martin Jambor <mjambor@suse.cz>
24281 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
24282 If removing a call statement LHS SSA name, release it.
24284 2021-04-27 Richard Earnshaw <rearnsha@arm.com>
24287 * config/arm/arm.c (THUMB2_WORK_REGS): Check PIC_OFFSET_TABLE_REGNUM
24288 is valid before including it in the mask.
24290 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
24293 * config/aarch64/aarch64.c (aarch64_comp_type_attributes): Handle
24296 2021-04-27 Richard Biener <rguenther@suse.de>
24298 PR tree-optimization/100051
24299 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Add
24300 disambiguator based on access size vs. decl size.
24302 2021-04-27 Richard Biener <rguenther@suse.de>
24304 PR tree-optimization/100278
24305 * tree-ssa-pre.c (compute_avail): Give up when we cannot
24306 adjust TBAA beacuse of mismatching bases.
24308 2021-04-27 Jakub Jelinek <jakub@redhat.com>
24311 * config/i386/i386.md (*<insn><mode>3_mask, *<insn><mode>3_mask_1):
24312 For any_rotate define_insn_split and following splitters, use
24313 SWI iterator instead of SWI48.
24315 2021-04-27 Richard Biener <rguenther@suse.de>
24317 PR tree-optimization/99776
24318 * match.pd (bit_field_ref (ctor)): Relax element extract
24319 type compatibility checks.
24321 2021-04-27 Cui,Lili <lili.cui@intel.com>
24323 * common/config/i386/i386-common.c (processor_names):
24324 Sync processor_names with processor_type.
24325 * config/i386/i386-options.c (processor_cost_table):
24326 Sync processor_cost_table with processor_type.
24328 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
24330 * value-range.cc (irange::irange_set_1bit_anti_range): Add assert.
24331 (irange::set): Call irange_set_1bit_anti_range for handling all
24332 1-bit ranges. Fall through on ~[MIN,MAX].
24334 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
24336 * value-range.cc (irange::legacy_num_pairs): Remove.
24337 (irange::invert): Change gcc_assert to gcc_checking_assert.
24338 * value-range.h (irange::num_pairs): Adjust for a cached
24339 num_pairs(). Also, rename all gcc_assert's to
24340 gcc_checking_assert's.
24342 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
24344 * value-range.cc (irange::operator=): Set m_kind.
24345 (irange::copy_to_legacy): Handle varying and undefined sources
24346 as a legacy copy since they can be easily copied.
24347 (irange::irange_set): Set m_kind.
24348 (irange::irange_set_anti_range): Same.
24349 (irange::set): Rename normalize_min_max to normalize_kind.
24350 (irange::verify_range): Adjust for multi-ranges having the
24352 (irange::irange_union): Set m_kind.
24353 (irange::irange_intersect): Same.
24354 (irange::invert): Same.
24355 * value-range.h (irange::kind): Always return m_kind.
24356 (irange::varying_p): Rename to...
24357 (irange::varying_comptaible_p): ...this.
24358 (irange::undefined_p): Only look at m_kind.
24359 (irange::irange): Always set VR_UNDEFINED if applicable.
24360 (irange::set_undefined): Always set VR_UNDEFINED.
24361 (irange::set_varying): Always set m_kind to VR_VARYING.
24362 (irange::normalize_min_max): Rename to...
24363 (irange::normalize_kind): ...this.
24365 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
24367 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
24368 Adjust for constant_p including varying_p.
24369 * tree-vrp.c (vrp_prop::finalize): Same.
24370 (determine_value_range): Same.
24371 * vr-values.c (vr_values::range_of_expr): Same.
24372 * value-range.cc (irange::symbolic_p): Do not check varying_p.
24373 (irange::constant_p): Same.
24375 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
24377 * value-range.cc (irange::legacy_lower_bound): Replace
24378 !undefined_p check with num_ranges > 0.
24379 (irange::legacy_upper_bound): Same.
24380 * value-range.h (irange::type): Same.
24381 (irange::lower_bound): Same.
24382 (irange::upper_bound): Same.
24384 2021-04-26 Richard Biener <rguenther@suse.de>
24386 PR tree-optimization/99956
24387 * gimple-loop-interchange.cc (compute_access_stride):
24388 Try instantiating the access in a shallower loop nest
24389 if instantiating failed.
24390 (compute_access_strides): Pass adjustable loop_nest
24391 to compute_access_stride.
24393 2021-04-26 Christophe Lyon <christophe.lyon@linaro.org>
24395 * doc/sourcebuild.texi (arm_cmse_hw): Document.
24397 2021-04-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24399 * config/aarch64/iterators.md (vwcore): Handle V4BF, V8BF.
24401 2021-04-26 Thomas Schwinge <thomas@codesourcery.com>
24402 Nathan Sidwell <nathan@codesourcery.com>
24403 Tom de Vries <vries@codesourcery.com>
24404 Julian Brown <julian@codesourcery.com>
24405 Kwok Cheung Yeung <kcy@codesourcery.com>
24407 * omp-offload.c (oacc_validate_dims): Implement
24408 '-Wopenacc-parallelism'.
24409 * doc/invoke.texi (-Wopenacc-parallelism): Document.
24411 2021-04-26 Richard Biener <rguenther@suse.de>
24413 * tree-cfg.h (gimplify_build1): Remove.
24414 (gimplify_build2): Likewise.
24415 (gimplify_build3): Likewise.
24416 * tree-cfg.c (gimplify_build1): Move to tree-vect-generic.c.
24417 (gimplify_build2): Likewise.
24418 (gimplify_build3): Likewise.
24419 * tree-vect-generic.c (gimplify_build1): Move from tree-cfg.c.
24421 (gimplify_build2): Likewise.
24422 (gimplify_build3): Likewise.
24423 (tree_vec_extract): Use resimplify with following SSA edges.
24424 (expand_vector_parallel): Avoid passing NULL size/bitpos
24425 to tree_vec_extract.
24426 * expr.c (store_constructor): Deal with zero-element CTORs.
24427 * match.pd (bit_field_ref <vector CTOR>): Make sure to
24428 produce vector constants when possible.
24430 2021-04-26 Richard Biener <rguenther@suse.de>
24432 * tree-complex.c: Include gimple-fold.h.
24433 (expand_complex_addition): Use gimple_build.
24434 (expand_complex_multiplication_components): Likewise.
24435 (expand_complex_multiplication): Likewise.
24436 (expand_complex_div_straight): Likewise.
24437 (expand_complex_div_wide): Likewise.
24438 (expand_complex_division): Likewise.
24439 (expand_complex_conjugate): Likewise.
24440 (expand_complex_comparison): Likewise.
24442 2021-04-26 Richard Biener <rguenther@suse.de>
24444 * tree-ssa-phiopt.c (two_value_replacement): Remove use
24445 of legacy gimplify_buildN API.
24447 2021-04-26 Richard Biener <rguenther@suse.de>
24449 PR tree-optimization/99473
24450 * tree-ssa-phiopt.c (cond_store_replacement): Handle all
24453 2021-04-26 Richard Biener <rguenther@suse.de>
24455 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin):
24456 Use replace_call_with_value.
24458 2021-04-26 Richard Biener <rguenther@suse.de>
24460 * tree-ssa-propagate.h (valid_gimple_rhs_p): Remove.
24461 (update_gimple_call): Likewise.
24462 (update_call_from_tree): Likewise.
24463 * tree-ssa-propagate.c (valid_gimple_rhs_p): Remove.
24464 (valid_gimple_call_p): Likewise.
24465 (move_ssa_defining_stmt_for_defs): Likewise.
24466 (finish_update_gimple_call): Likewise.
24467 (update_gimple_call): Likewise.
24468 (update_call_from_tree): Likewise.
24469 (propagate_tree_value_into_stmt): Use replace_call_with_value.
24470 * gimple-fold.h (update_gimple_call): Declare.
24471 * gimple-fold.c (valid_gimple_rhs_p): Move here from
24472 tree-ssa-propagate.c.
24473 (update_gimple_call): Likewise.
24474 (valid_gimple_call_p): Likewise.
24475 (finish_update_gimple_call): Likewise, and simplify.
24476 (gimplify_and_update_call_from_tree): Implement
24477 update_call_from_tree functionality, avoid excessive
24478 push/pop_gimplify_context.
24479 (gimple_fold_builtin): Use only gimplify_and_update_call_from_tree.
24480 (gimple_fold_call): Likewise.
24481 * gimple-ssa-sprintf.c (try_substitute_return_value): Likewise.
24482 * tree-ssa-ccp.c (ccp_folder::fold_stmt): Likewise.
24483 (pass_fold_builtins::execute): Likewise.
24484 (optimize_stack_restore): Use replace_call_with_value.
24485 * tree-cfg.c (fold_loop_internal_call): Likewise.
24486 * tree-ssa-dce.c (maybe_optimize_arith_overflow): Use
24487 only gimplify_and_update_call_from_tree.
24488 * tree-ssa-strlen.c (handle_builtin_strlen): Likewise.
24489 (handle_builtin_strchr): Likewise.
24490 * tsan.c: Include gimple-fold.h instead of tree-ssa-propagate.h.
24492 2021-04-26 Jakub Jelinek <jakub@redhat.com>
24495 * vmsdbgout.c (ASM_OUTPUT_DEBUG_STRING, vmsdbgout_begin_block,
24496 vmsdbgout_end_block, lookup_filename, vmsdbgout_source_line): Remove
24499 2021-04-25 liuhongt <hongtao.liu@intel.com>
24502 * config/i386/i386-builtin.def (BDESC): Change the icode of
24503 the following builtins to CODE_FOR_nothing.
24504 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
24505 IX86_BUILTIN_PCMPEQB128, IX86_BUILTIN_PCMPEQW128,
24506 IX86_BUILTIN_PCMPEQD128, IX86_BUILTIN_PCMPEQQ,
24507 IX86_BUILTIN_PCMPEQB256, IX86_BUILTIN_PCMPEQW256,
24508 IX86_BUILTIN_PCMPEQD256, IX86_BUILTIN_PCMPEQQ256,
24509 IX86_BUILTIN_PCMPGTB128, IX86_BUILTIN_PCMPGTW128,
24510 IX86_BUILTIN_PCMPGTD128, IX86_BUILTIN_PCMPGTQ,
24511 IX86_BUILTIN_PCMPGTB256, IX86_BUILTIN_PCMPGTW256,
24512 IX86_BUILTIN_PCMPGTD256, IX86_BUILTIN_PCMPGTQ256.
24513 * config/i386/sse.md (avx2_eq<mode>3): Deleted.
24514 (sse2_eq<mode>3): Ditto.
24515 (sse4_1_eqv2di3): Ditto.
24516 (sse2_gt<mode>3): Rename to ..
24517 (*sse2_gt<mode>3): .. this.
24519 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
24522 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
24525 * config/darwin.c (darwin_binds_local_p): Assume that any
24526 public symbol might be interposed for PIC code. Update function
24527 header comment to reflect current Darwin capability.
24529 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
24532 * config/darwin.c (darwin_binds_local_p): Assume that any
24533 public symbol might be interposed for PIC code. Update function
24534 header comment to reflect current Darwin capability.
24536 2021-04-24 Richard Sandiford <richard.sandiford@arm.com>
24538 * doc/sourcebuild.texi: Document no-opts and any-opts target
24541 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
24543 * config/bpf/bpf.h (ASM_OUTPUT_ALIGNED_BSS): Use .type and .lcomm.
24545 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
24547 * config/bpf/bpf.h (FUNCTION_BOUNDARY): Set to 64.
24549 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
24552 * config/i386/i386-options.c (ix86_option_override_internal):
24553 Error out when -m96bit-long-double is used with 64bit targets.
24554 * config/i386/i386.md (*pushxf_rounded): Remove pattern.
24556 2021-04-23 Martin Liska <mliska@suse.cz>
24558 * lto-wrapper.c: Remove FIXME about usage of
24559 hardware_concurrency. The function is not on par with
24562 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
24565 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
24566 Copy operand 3 to operand 4. Use sse_reg_operand
24567 as operand 3 predicate.
24568 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): Ditto.
24569 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
24570 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
24571 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2):
24572 Copy operand 1 to operand 0.
24573 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): Ditto.
24574 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
24575 (LDX_ATOMIC/LDX_ATOMIC FP store peephole2 with mem blockage): Ditto.
24577 2021-04-23 Alex Coplan <alex.coplan@arm.com>
24579 PR rtl-optimization/100230
24580 * early-remat.c (early_remat::sort_candidates): Use delete[]
24581 instead of delete for array allocated with new[].
24583 2021-04-23 Richard Biener <rguenther@suse.de>
24585 * genmatch.c (lower_cond): Remove VEC_COND_EXPR special-casing.
24586 (capture_info::capture_info): Likewise.
24587 (capture_info::walk_match): Likewise.
24588 (expr::gen_transform): Likewise.
24589 (dt_simplify::gen_1): Likewise.
24590 * gimple-match-head.c (maybe_resimplify_conditional_op):
24591 Remove VEC_COND_EXPR special-casing.
24592 (gimple_simplify): Likewise.
24593 * gimple.c (gimple_could_trap_p_1): Adjust.
24594 * tree-ssa-pre.c (compute_avail): Allow VEC_COND_EXPR
24595 to participate in PRE.
24597 2021-04-23 Richard Biener <rguenther@suse.de>
24599 * cfganal.c (connect_infinite_loops_to_exit): First call
24600 add_noreturn_fake_exit_edges.
24601 * ipa-sra.c (process_scan_results): Do not call the now redundant
24602 add_noreturn_fake_exit_edges.
24603 * predict.c (tree_estimate_probability): Likewise.
24604 (rebuild_frequencies): Likewise.
24605 * store-motion.c (one_store_motion_pass): Likewise.
24607 2021-04-23 Richard Biener <rguenther@suse.de>
24609 PR tree-optimization/100222
24610 * predict.c (pass_profile::execute): Remove redundant call to
24611 mark_irreducible_loops.
24612 (report_predictor_hitrates): Likewise.
24614 2021-04-23 Richard Biener <rguenther@suse.de>
24616 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Avoid
24617 valid_gimple_rhs_p by instead gimplifying to one.
24619 2021-04-23 Richard Biener <rguenther@suse.de>
24621 PR tree-optimization/99971
24622 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
24623 Always use TBAA for loads.
24625 2021-04-23 liuhongt <hongtao.liu@intel.com>
24628 * config/i386/i386-options.c (ix86_option_override_internal):
24629 Clear MASK_AVX256_SPLIT_UNALIGNED_LOAD/STORE in x_target_flags
24630 when X86_TUNE_AVX256_UNALIGNED_LOAD/STORE_OPTIMAL is enabled
24631 by target attribute.
24633 2021-04-23 David Edelsohn <dje.gcc@gmail.com>
24635 * config/rs6000/aix71.h (PREFERRED_DEBUGGING_TYPE): Change to
24637 * config/rs6000/aix72.h (PREFERRED_DEBUGGING_TYPE): Same.
24639 2021-04-22 David Edelsohn <dje.gcc@gmail.com>
24641 * config.gcc (powerpc-ibm-aix6.*): Remove.
24642 * config/rs6000/aix61.h: Delete.
24644 2021-04-22 Martin Liska <mliska@suse.cz>
24646 PR testsuite/100159
24647 PR testsuite/100192
24648 * builtins.c (expand_builtin): Fix typos and missing comments.
24649 * dwarf2out.c (gen_subprogram_die): Likewise.
24650 (gen_struct_or_union_type_die): Likewise.
24652 2021-04-22 Uroš Bizjak <ubizjak@gmail.com>
24655 * config/i386/i386-expand.c (ix86_expand_convert_uns_sidf_sse):
24656 Remove the sign with FE_DOWNWARD, where x - x = -0.0.
24658 2021-04-21 Iain Sandoe <iain@sandoe.co.uk>
24660 * config/i386/darwin.h (TARGET_64BIT): Remove definition
24661 based on TARGET_ISA_64BIT.
24662 (TARGET_64BIT_P): Remove definition based on
24663 TARGET_ISA_64BIT_P().
24665 2021-04-21 Martin Liska <mliska@suse.cz>
24668 2021-04-21 Martin Liska <mliska@suse.cz>
24670 * lto-wrapper.c (cpuset_popcount): Remove.
24671 (init_num_threads): Remove and use hardware_concurrency.
24673 2021-04-21 Martin Liska <mliska@suse.cz>
24676 * main.c (main): Call toplev::finalize in CHECKING_P mode.
24677 * ipa-modref.c (ipa_modref_c_finalize): summaries are NULL
24678 when incremental LTO linking happens.
24680 2021-04-21 Martin Liska <mliska@suse.cz>
24682 * lto-wrapper.c (run_gcc): When -flto=jobserver is used, but the
24683 makeserver cannot be detected, then use -flto=N fallback.
24685 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
24687 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): When cross-compiling,
24688 default to yes for aarch64-linux-gnu.
24689 * configure: Regenerate.
24691 2021-04-21 Martin Liska <mliska@suse.cz>
24693 * lto-wrapper.c (cpuset_popcount): Remove.
24694 (init_num_threads): Remove and use hardware_concurrency.
24696 2021-04-21 Martin Liska <mliska@suse.cz>
24698 * config/i386/i386.c: Remove superfluous || TARGET_MACHO
24699 which remains to be '(... || 0)' and clang complains about it.
24700 * dwarf2out.c (AT_vms_delta): Declare conditionally.
24701 (add_AT_vms_delta): Likewise.
24702 * tree.c (fld_simplified_type): Use rather more common pattern
24703 for disabling of something (#if 0).
24704 (get_tree_code_name): Likewise.
24705 (verify_type_variant): Likewise.
24707 2021-04-21 Martin Liska <mliska@suse.cz>
24709 * config/i386/i386-expand.c (decide_alignment): Use newly named
24710 macro TARGET_CPU_P.
24711 * config/i386/i386.c (ix86_decompose_address): Likewise.
24712 (ix86_address_cost): Likewise.
24713 (ix86_lea_outperforms): Likewise.
24714 (ix86_avoid_lea_for_addr): Likewise.
24715 (ix86_add_stmt_cost): Likewise.
24716 * config/i386/i386.h (TARGET_*): Remove.
24717 (TARGET_CPU_P): New macro.
24718 * config/i386/i386.md: Use newly named macro TARGET_CPU_P.
24719 * config/i386/x86-tune-sched-atom.c (do_reorder_for_imul): Likewise.
24720 (swap_top_of_ready_list): Likewise.
24721 (ix86_atom_sched_reorder): Likewise.
24722 * config/i386/x86-tune-sched-bd.c (ix86_bd_has_dispatch): Likewise.
24723 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Likewise.
24725 2021-04-21 Martin Liska <mliska@suse.cz>
24727 * config/i386/i386-options.c (TARGET_EXPLICIT_NO_SAHF_P):
24729 (SET_TARGET_NO_SAHF): Likewise.
24730 (TARGET_EXPLICIT_PREFETCH_SSE_P): Likewise.
24731 (SET_TARGET_PREFETCH_SSE): Likewise.
24732 (TARGET_EXPLICIT_NO_TUNE_P): Likewise.
24733 (SET_TARGET_NO_TUNE): Likewise.
24734 (TARGET_EXPLICIT_NO_80387_P): Likewise.
24735 (SET_TARGET_NO_80387): Likewise.
24737 * config/i386/i386.h (TARGET_*): Remove.
24738 * opth-gen.awk: Generate new used macros.
24740 2021-04-21 Martin Liska <mliska@suse.cz>
24742 * config/i386/i386.h (PTA_*): Remove.
24743 (enum pta_flag): New.
24744 (DEF_PTA): Generate PTA_* values from i386-isa.def.
24745 * config/i386/i386-isa.def: New file.
24747 2021-04-21 Alex Coplan <alex.coplan@arm.com>
24750 * config/aarch64/aarch64-bti-insert.c (aarch64_bti_j_insn_p): New.
24751 (rest_of_insert_bti): Avoid inserting duplicate bti j insns for
24752 jump table targets.
24754 2021-04-21 H.J. Lu <hjl.tools@gmail.com>
24756 * config.gcc: Install mwaitintrin.h for i[34567]86-*-* and
24757 x86_64-*-* targets.
24758 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_MWAIT_SET):
24760 (OPTION_MASK_ISA2_MWAIT_UNSET): Likewise.
24761 (ix86_handle_option): Handle -mmwait.
24762 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
24763 Replace OPTION_MASK_ISA_SSE3 with OPTION_MASK_ISA2_MWAIT on
24764 __builtin_ia32_monitor and __builtin_ia32_mwait.
24765 * config/i386/i386-options.c (isa2_opts): Add -mmwait.
24766 (ix86_valid_target_attribute_inner_p): Likewise.
24767 (ix86_option_override_internal): Enable mwait/monitor
24768 instructions for -msse3.
24769 * config/i386/i386.h (TARGET_MWAIT): New.
24770 (TARGET_MWAIT_P): Likewise.
24771 * config/i386/i386.opt: Add -mmwait.
24772 * config/i386/mwaitintrin.h: New file.
24773 * config/i386/pmmintrin.h: Include <mwaitintrin.h>.
24774 * config/i386/sse.md (sse3_mwait): Replace TARGET_SSE3 with
24776 (@sse3_monitor_<mode>): Likewise.
24777 * config/i386/x86gprintrin.h: Include <mwaitintrin.h>.
24778 * doc/extend.texi: Document mwait target attribute.
24779 * doc/invoke.texi: Document -mmwait.
24781 2021-04-21 Martin Liska <mliska@suse.cz>
24783 * config/i386/i386-options.c (DEF_ENUM): Remove it.
24784 * config/i386/i386-opts.h (DEF_ENUM): Likewise.
24785 * config/i386/stringop.def (DEF_ENUM): Likewise.
24787 2021-04-21 Martin Liska <mliska@suse.cz>
24789 * tree-cfg.c (gimple_verify_flow_info): Use qD instead
24790 of print_generic_expr.
24792 2021-04-21 Jakub Jelinek <jakub@redhat.com>
24794 PR rtl-optimization/100148
24795 * cprop.c (constprop_register): Use next_nondebug_insn instead of
24798 2021-04-21 Martin Liska <mliska@suse.cz>
24801 * cgraphunit.c (cgraph_node::analyze): Remove duplicate
24802 free_dominance_info calls.
24804 2021-04-21 Richard Biener <rguenther@suse.de>
24806 * gimple-fold.c (maybe_fold_reference): Remove is_lhs
24807 parameter (and assume it to be false).
24808 (fold_gimple_assign): Adjust, remove all callers of
24809 maybe_fold_reference calling it with is_lhs true.
24810 (gimple_fold_call): Likewise.
24811 (fold_stmt_1): Likewise.
24813 2021-04-21 Richard Biener <rguenther@suse.de>
24815 * fold-const.c (pedantic_non_lvalue_loc): Remove.
24816 (fold_binary_loc): Adjust.
24817 (fold_ternary_loc): Likewise.
24819 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
24821 PR middle-end/100130
24822 * varasm.c (get_block_for_decl): Make sure that any use of the
24823 retain attribute matches the section's retain flag.
24824 (switch_to_section): Check for retain mismatches even when
24825 changing sections, but do not warn if the given decl is the
24826 section's named.decl.
24827 (output_object_block): Pass the first decl in the block (if any)
24828 to switch_to_section.
24830 2021-04-20 H.J. Lu <hjl.tools@gmail.com>
24832 * config/i386/i386-c.c (ix86_target_macros_internal): Define
24833 __CRC32__ for -mcrc32.
24834 * config/i386/i386-options.c (ix86_option_override_internal):
24835 Enable crc32 instruction for -msse4.2.
24836 * config/i386/i386.md (sse4_2_crc32<mode>): Remove TARGET_SSE4_2
24838 (sse4_2_crc32di): Likewise.
24839 * config/i386/ia32intrin.h: Use crc32 target option for CRC32
24842 2021-04-20 Segher Boessenkool <segher@kernel.crashing.org>
24845 * config/rs6000/rs6000.c (rs6000_machine_from_flags): Do not consider
24848 2021-04-20 Martin Liska <mliska@suse.cz>
24850 * doc/invoke.texi: Fix typo.
24851 * params.opt: Likewise.
24853 2021-04-20 Martin Liska <mliska@suse.cz>
24855 * doc/invoke.texi: Document new param.
24857 2021-04-19 Andrew MacLeod <amacleod@redhat.com>
24859 PR tree-optimization/100081
24860 * gimple-range-cache.h (ranger_cache): Inherit from gori_compute
24861 rather than gori_compute_cache.
24862 * gimple-range-gori.cc (is_gimple_logical_p): Move to top of file.
24863 (range_def_chain::m_logical_depth): New member.
24864 (range_def_chain::range_def_chain): Initialize m_logical_depth.
24865 (range_def_chain::get_def_chain): Don't build defchains through more
24866 than LOGICAL_LIMIT logical expressions.
24867 * params.opt (param_ranger_logical_depth): New.
24869 2021-04-19 Richard Earnshaw <rearnsha@arm.com>
24872 * config/arm/arm.c (arm_configure_build_target): Do not strip
24873 extended FPU/SIMD feature bits from the target ISA when -mfpu
24874 is specified (partial revert of r11-8168).
24876 2021-04-19 Thomas Schwinge <thomas@codesourcery.com>
24878 * params.opt (-param=openacc-kernels=): Add.
24879 * omp-oacc-kernels-decompose.cc
24880 (pass_omp_oacc_kernels_decompose::gate): Use it.
24881 * doc/invoke.texi (-fopenacc-kernels=@var{mode}): Move...
24882 (--param): ... here, 'openacc-kernels'.
24884 2021-04-19 Martin Liska <mliska@suse.cz>
24887 * gengtype.c (finish_root_table): Align function arguments
24888 in between declaration and definition.
24890 2021-04-19 Eric Botcazou <ebotcazou@adacore.com>
24892 * config/i386/winnt.c (i386_pe_seh_cold_init): Properly deal with
24893 frames larger than the SEH maximum frame size.
24895 2021-04-18 Segher Boessenkool <segher@kernel.crashing.org>
24897 PR rtl-optimization/99927
24898 * combine.c (distribute_notes) [REG_UNUSED]: If the register already
24899 is dead, just drop it.
24901 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
24904 * config/i386/winnt-d.c (TARGET_D_TEMPLATES_ALWAYS_COMDAT): Define.
24905 * doc/tm.texi: Regenerate.
24906 * doc/tm.texi.in (D language and ABI): Add @hook for
24907 TARGET_D_TEMPLATES_ALWAYS_COMDAT.
24909 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
24911 * config/darwin-d.c (darwin_d_handle_target_object_format): New
24913 (darwin_d_register_target_info): New function.
24914 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
24915 * config/dragonfly-d.c (dragonfly_d_handle_target_object_format): New
24917 (dragonfly_d_register_target_info): New function.
24918 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
24919 * config/freebsd-d.c (freebsd_d_handle_target_object_format): New
24921 (freebsd_d_register_target_info): New function.
24922 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
24923 * config/glibc-d.c (glibc_d_handle_target_object_format): New
24925 (glibc_d_register_target_info): New function.
24926 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
24927 * config/i386/i386-d.c (ix86_d_handle_target_object_format): New
24929 (ix86_d_register_target_info): Add ix86_d_handle_target_object_format
24930 as handler for objectFormat key.
24931 * config/i386/winnt-d.c (winnt_d_handle_target_object_format): New
24933 (winnt_d_register_target_info): New function.
24934 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
24935 * config/netbsd-d.c (netbsd_d_handle_target_object_format): New
24937 (netbsd_d_register_target_info): New function.
24938 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
24939 * config/openbsd-d.c (openbsd_d_handle_target_object_format): New
24941 (openbsd_d_register_target_info): New function.
24942 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
24943 * config/pa/pa-d.c (pa_d_handle_target_object_format): New function.
24944 (pa_d_register_target_info): Add pa_d_handle_target_object_format as
24945 handler for objectFormat key.
24946 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_object_format): New
24948 (rs6000_d_register_target_info): Add
24949 rs6000_d_handle_target_object_format as handler for objectFormat key.
24950 * config/sol2-d.c (solaris_d_handle_target_object_format): New
24952 (solaris_d_register_target_info): New function.
24953 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
24955 2021-04-16 Jakub Jelinek <jakub@redhat.com>
24958 * config/aarch64/aarch64.c (aarch64_function_arg_alignment): Change
24959 abi_break argument from bool * to unsigned *, store there the pre-GCC 9
24961 (aarch64_layout_arg, aarch64_gimplify_va_arg_expr): Adjust callers.
24962 (aarch64_function_arg_regno_p): Likewise. Only emit -Wpsabi note if
24963 the old and new alignment after applying MIN/MAX to it is different.
24965 2021-04-16 Tamar Christina <tamar.christina@arm.com>
24968 * config/aarch64/aarch64-sve.md (@aarch64_sve_trn1_conv<mode>): New.
24969 * config/aarch64/aarch64.c (aarch64_expand_sve_const_pred_trn): Use new
24971 * config/aarch64/iterators.md (UNSPEC_TRN1_CONV): New.
24973 2021-04-16 Bill Schmidt <wschmidt@linux.ibm.com>
24975 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Revise
24976 this section and its subsections.
24978 2021-04-16 Jakub Jelinek <jakub@redhat.com>
24981 * config/aarch64/aarch64.md (*neg_asr_si2_extr, *extrsi5_insn_di): New
24982 define_insn patterns.
24984 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
24986 PR rtl-optimization/98689
24987 * reg-notes.def (UNTYPED_CALL): New note.
24988 * combine.c (distribute_notes): Handle it.
24989 * emit-rtl.c (try_split): Likewise.
24990 * rtlanal.c (rtx_properties::try_to_add_insn): Likewise. Assume
24991 that calls with the note implicitly set all return value registers.
24992 * builtins.c (expand_builtin_apply): Add a REG_UNTYPED_CALL
24995 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
24997 PR rtl-optimization/99596
24998 * rtlanal.c (rtx_properties::try_to_add_insn): Don't add global
24999 register accesses for const calls. Assume that pure functions
25000 can only read from global registers. Ignore cases in which
25001 the stack pointer has been marked global.
25003 2021-04-16 Jakub Jelinek <jakub@redhat.com>
25006 * tree-vect-loop.c (vect_transform_loop): Don't remove just
25007 dead scalar .MASK_LOAD calls, but also dead .COND_* calls - replace
25008 them by their last argument.
25010 2021-04-15 Martin Liska <mliska@suse.cz>
25012 * doc/invoke.texi: Other params don't use it, remove it.
25014 2021-04-15 Richard Biener <rguenther@suse.de>
25016 * gimple-builder.h: Add deprecation note.
25018 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
25021 * attribs.h (restrict_type_identity_attributes_to): Declare.
25022 * attribs.c (restrict_type_identity_attributes_to): New function.
25024 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
25027 * attribs.h (affects_type_identity_attributes): Declare.
25028 * attribs.c (remove_attributes_matching): New function.
25029 (affects_type_identity_attributes): Likewise.
25031 2021-04-15 Jakub Jelinek <jakub@redhat.com>
25034 * config/aarch64/aarch64.md (*<LOGICAL:optab>_<SHIFT:optab><mode>3):
25035 Add combine splitters for *<LOGICAL:optab>_ashl<mode>3 with
25036 ZERO_EXTEND, SIGN_EXTEND or AND.
25038 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
25040 PR rtl-optimization/99929
25041 * rtl.h (same_vector_encodings_p): New function.
25042 * cse.c (exp_equiv_p): Check that CONST_VECTORs have the same encoding.
25043 * cselib.c (rtx_equal_for_cselib_1): Likewise.
25044 * jump.c (rtx_renumbered_equal_p): Likewise.
25045 * lra-constraints.c (operands_match_p): Likewise.
25046 * reload.c (operands_match_p): Likewise.
25047 * rtl.c (rtx_equal_p_cb, rtx_equal_p): Likewise.
25049 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
25051 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Print
25052 more information about variable-length CONST_VECTORs.
25054 2021-04-14 Vladimir N. Makarov <vmakarov@redhat.com>
25056 PR rtl-optimization/100066
25057 * lra-constraints.c (split_reg): Check paradoxical_subreg_p for
25058 ordered modes when choosing splitting mode for hard reg.
25060 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
25063 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector_sel):
25065 (aarch64_expand_sve_const_vector): Use it for nelts_per_pattern==2.
25067 2021-04-14 Andreas Krebbel <krebbel@linux.ibm.com>
25069 * config/s390/s390-builtins.def (O_M5, O_M12, ...): Add new macros
25070 for mask operand types.
25071 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
25072 (s390_vec_permi_dbl, s390_vpdi): Use the M5 type for the immediate
25074 (s390_vec_msum_u128, s390_vmslg): Use the M12 type for the
25076 * config/s390/s390.c (s390_const_operand_ok): Check the new
25077 operand types and generate a list of valid values.
25079 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
25081 * doc/tm.texi: Regenerate.
25082 * doc/tm.texi.in (D language and ABI): Add @hook for
25083 TARGET_D_REGISTER_OS_TARGET_INFO.
25085 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
25087 * config/aarch64/aarch64-d.c (aarch64_d_handle_target_float_abi): New
25089 (aarch64_d_register_target_info): New function.
25090 * config/aarch64/aarch64-protos.h (aarch64_d_register_target_info):
25092 * config/aarch64/aarch64.h (TARGET_D_REGISTER_CPU_TARGET_INFO):
25094 * config/arm/arm-d.c (arm_d_handle_target_float_abi): New function.
25095 (arm_d_register_target_info): New function.
25096 * config/arm/arm-protos.h (arm_d_register_target_info): Declare.
25097 * config/arm/arm.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
25098 * config/i386/i386-d.c (ix86_d_handle_target_float_abi): New function.
25099 (ix86_d_register_target_info): New function.
25100 * config/i386/i386-protos.h (ix86_d_register_target_info): Declare.
25101 * config/i386/i386.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
25102 * config/mips/mips-d.c (mips_d_handle_target_float_abi): New function.
25103 (mips_d_register_target_info): New function.
25104 * config/mips/mips-protos.h (mips_d_register_target_info): Declare.
25105 * config/mips/mips.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
25106 * config/pa/pa-d.c (pa_d_handle_target_float_abi): New function.
25107 (pa_d_register_target_info): New function.
25108 * config/pa/pa-protos.h (pa_d_register_target_info): Declare.
25109 * config/pa/pa.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
25110 * config/riscv/riscv-d.c (riscv_d_handle_target_float_abi): New
25112 (riscv_d_register_target_info): New function.
25113 * config/riscv/riscv-protos.h (riscv_d_register_target_info): Declare.
25114 * config/riscv/riscv.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
25115 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_float_abi): New
25117 (rs6000_d_register_target_info): New function.
25118 * config/rs6000/rs6000-protos.h (rs6000_d_register_target_info):
25120 * config/rs6000/rs6000.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
25121 * config/s390/s390-d.c (s390_d_handle_target_float_abi): New function.
25122 (s390_d_register_target_info): New function.
25123 * config/s390/s390-protos.h (s390_d_register_target_info): Declare.
25124 * config/s390/s390.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
25125 * config/sparc/sparc-d.c (sparc_d_handle_target_float_abi): New
25127 (sparc_d_register_target_info): New function.
25128 * config/sparc/sparc-protos.h (sparc_d_register_target_info): Declare.
25129 * config/sparc/sparc.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
25130 * doc/tm.texi: Regenerate.
25131 * doc/tm.texi.in (D language and ABI): Add @hook for
25132 TARGET_D_REGISTER_CPU_TARGET_INFO.
25134 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
25136 * config/i386/i386-d.c (ix86_d_has_stdcall_convention): New function.
25137 * config/i386/i386-protos.h (ix86_d_has_stdcall_convention): Declare.
25138 * config/i386/i386.h (TARGET_D_HAS_STDCALL_CONVENTION): Define.
25139 * doc/tm.texi: Regenerate.
25140 * doc/tm.texi.in (D language and ABI): Add @hook for
25141 TARGET_D_HAS_STDCALL_CONVENTION.
25143 2021-04-14 Richard Biener <rguenther@suse.de>
25145 * tree-cfg.c (verify_gimple_assign_ternary): Verify that
25146 VEC_COND_EXPRs have a gimple_val condition.
25147 * tree-ssa-propagate.c (valid_gimple_rhs_p): VEC_COND_EXPR
25148 can no longer have a GENERIC condition.
25150 2021-04-14 Richard Earnshaw <rearnsha@arm.com>
25153 * config/arm/arm.c (arm_configure_build_target): Strip isa_all_fpbits
25154 from the isa_delta when -mfpu has been used.
25155 (arm_options_perform_arch_sanity_checks): It's the architecture that
25156 lacks an FPU not the processor.
25158 2021-04-13 Richard Biener <rguenther@suse.de>
25160 PR tree-optimization/100053
25161 * tree-ssa-sccvn.c (vn_nary_op_get_predicated_value): Do
25162 not use optimistic dominance queries for backedges to validate
25164 (dominated_by_p_w_unex): Add parameter to ignore executable
25165 state on backedges.
25166 (rpo_elim::eliminate_avail): Adjust.
25168 2021-04-13 Jakub Jelinek <jakub@redhat.com>
25171 * config/aarch64/aarch64.md (*aarch64_bfxil<mode>_extr,
25172 *aarch64_bfxilsi_extrdi): New define_insn patterns.
25174 2021-04-13 Jakub Jelinek <jakub@redhat.com>
25177 * simplify-rtx.c (simplify_immed_subreg): For MODE_COMPOSITE_P
25178 outermode, return NULL if the result doesn't encode back to the
25179 original byte sequence.
25180 (simplify_gen_subreg): Don't create SUBREGs from constants to
25181 MODE_COMPOSITE_P outermode.
25183 2021-04-12 Jakub Jelinek <jakub@redhat.com>
25185 PR rtl-optimization/99905
25186 * combine.c (expand_compound_operation): If pos + len > modewidth,
25187 perform the right shift by pos in inner_mode and then convert to mode,
25188 instead of trying to simplify a shift of rtx with inner_mode by pos
25189 as if it was a shift in mode.
25191 2021-04-12 Jakub Jelinek <jakub@redhat.com>
25194 * combine.c (simplify_and_const_int_1): Don't optimize varop
25195 away if it has side-effects.
25197 2021-04-12 Martin Liska <mliska@suse.cz>
25199 * doc/extend.texi: Escape @smallexample content.
25201 2021-04-12 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
25203 * config/s390/s390.md ("*movdi_31", "*movdi_64"): Add
25204 alternative in order to load a DFP zero.
25206 2021-04-12 Martin Liska <mliska@suse.cz>
25208 * doc/extend.texi: Be more precise in documentation
25209 of symver attribute.
25211 2021-04-12 Martin Liska <mliska@suse.cz>
25214 * gimplify.c (gimplify_expr): Right now, we unpoison all
25215 variables before a goto <dest>. We should not do it if we are
25218 2021-04-12 Cui,Lili <lili.cui@intel.com>
25220 * common/config/i386/cpuinfo.h (get_intel_cpu): Handle
25222 * common/config/i386/i386-common.c (processor_names): Add
25224 (processor_alias_table): Add rocketlake.
25225 * common/config/i386/i386-cpuinfo.h (processor_subtypes): Add
25226 INTEL_COREI7_ROCKETLAKE.
25227 * config.gcc: Add -march=rocketlake.
25228 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
25230 * config/i386/i386-options.c (m_ROCKETLAKE) : Define.
25231 (processor_cost_table): Add rocketlake cost.
25232 * config/i386/i386.h (ix86_size_cost) : Define
25234 (processor_type) : Add PROCESSOR_ROCKETLAKE.
25235 (PTA_ROCKETLAKE): Ditto.
25236 * doc/extend.texi: Add rocketlake.
25237 * doc/invoke.texi: Add rocketlake.
25239 2021-04-12 Cui,Lili <lili.cui@intel.com>
25241 * config/i386/i386.h (PTA_ALDERLAKE): Change alderlake ISA list.
25242 * config/i386/i386-options.c (m_CORE_AVX2): Add m_ALDERLAKE.
25243 * common/config/i386/cpuinfo.h (get_intel_cpu): Add AlderLake model.
25244 * doc/invoke.texi: Change alderlake ISA list.
25246 2021-04-11 Hafiz Abid Qadeer <abidh@codesourcery.com>
25248 PR middle-end/98088
25249 * omp-expand.c (expand_oacc_collapse_init): Update condition in
25252 2021-04-10 H.J. Lu <hjl.tools@gmail.com>
25255 * config/i386/serializeintrin.h (_serialize): Defined as macro.
25257 2021-04-10 Jakub Jelinek <jakub@redhat.com>
25260 * expr.c (expand_expr_addr_expr_1): Test is_global_var rather than
25261 just TREE_STATIC on COMPOUND_LITERAL_EXPR_DECLs.
25263 2021-04-10 Jakub Jelinek <jakub@redhat.com>
25265 PR middle-end/99989
25266 * gimple-ssa-warn-alloca.c
25267 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit to
25268 0 with integer precision unconditionally.
25270 2021-04-10 Jakub Jelinek <jakub@redhat.com>
25272 PR rtl-optimization/98601
25273 * rtlanal.c (rtx_addr_can_trap_p_1): Allow in assert unknown size
25274 not just for BLKmode, but also for VOIDmode. For STRICT_ALIGNMENT
25275 unaligned_mems handle VOIDmode like BLKmode.
25277 2021-04-10 Jan Hubicka <hubicka@ucw.cz>
25280 * tree.c (free_lang_data_in_decl): Do not release body of
25281 declare_variant_alt.
25283 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
25285 * config/aarch64/aarch64.c (aarch64_option_restore): If the
25286 architecture was specified explicitly and the tuning wasn't,
25287 tune for the architecture rather than the configured default CPU.
25289 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
25291 * config/aarch64/aarch64.md (tlsdesc_small_sve_<mode>): Use X30
25292 as the temporary register.
25294 2021-04-09 Martin Liska <mliska@suse.cz>
25296 * doc/extend.texi: Move non-target attributes on the top level.
25298 2021-04-09 Martin Liska <mliska@suse.cz>
25300 * doc/invoke.texi: Document minimum and maximum value of the
25301 argument for both supported compression algorithms.
25303 2021-04-08 David Edelsohn <dje.gcc@gmail.com>
25305 * config/rs6000/rs6000.c (rs6000_xcoff_select_section): Select
25306 TLS BSS before TLS data.
25307 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .comm.
25309 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
25311 * doc/sourcebuild.texi (stdint_types_mbig_endian): Document.
25313 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
25315 * match.pd: Extend vec_cond folds to handle shifts.
25317 2021-04-08 Maciej W. Rozycki <macro@orcam.me.uk>
25319 * config/vax/vax.md: Fix comment for `*bit<mode>' pattern's
25322 2021-04-08 Alex Coplan <alex.coplan@arm.com>
25325 * config/arm/iterators.md (MVE_vecs): New.
25326 (V_elem): Also handle V2DF.
25327 * config/arm/mve.md (*mve_mov<mode>): Rename to ...
25328 (*mve_vdup<mode>): ... this. Remove second alternative since
25329 vec_duplicate of const_int is not canonical RTL, and we don't
25330 want to match symbol_refs.
25331 (*mve_vec_duplicate<mode>): Delete (pattern is redundant).
25333 2021-04-08 Xionghu Luo <luoxhu@linux.ibm.com>
25335 * fold-const.c (fold_single_bit_test): Fix typo.
25336 * print-rtl.c (print_rtx_insn_vec): Call print_rtl_single
25339 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
25341 PR tree-optimization/97513
25342 * tree-vect-slp.c (vect_add_slp_permutation): New function,
25344 (vectorizable_slp_permutation): ...here. Detect cases in which
25345 all VEC_PERM_EXPRs are guaranteed to have the same stepped
25346 permute vector and only generate one permute vector for that case.
25347 Extend that case to handle variable-length vectors.
25349 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
25351 PR tree-optimization/99873
25352 * tree-vect-slp.c (vect_slp_prefer_store_lanes_p): New function.
25353 (vect_build_slp_instance): Don't split store groups that could
25354 use IFN_STORE_LANES.
25356 2021-04-07 Jakub Jelinek <jakub@redhat.com>
25359 * varasm.c (output_constant_pool_contents): Don't strip name encoding
25360 from XSTR (desc->sym, 0) or from label before passing those to
25363 2021-04-07 Richard Biener <rguenther@suse.de>
25365 PR tree-optimization/99954
25366 * tree-loop-distribution.c: Include tree-affine.h.
25367 (generate_memcpy_builtin): Try using tree-affine to prove
25369 (loop_distribution::classify_builtin_ldst): Always classify
25372 2021-04-07 Richard Biener <rguenther@suse.de>
25374 PR tree-optimization/99947
25375 * tree-vect-loop.c (vectorizable_induction): Pre-allocate
25376 steps vector to avoid pushing elements from the reallocated
25379 2021-04-07 Richard Biener <rguenther@suse.de>
25381 * tree-ssa-sccvn.h (print_vn_reference_ops): Declare.
25382 * tree-ssa-pre.c (print_pre_expr): Factor out VN reference operand
25384 * tree-ssa-sccvn.c (print_vn_reference_ops): ... into this new
25386 (debug_vn_reference_ops): New.
25388 2021-04-07 Bin Cheng <bin.cheng@linux.alibaba.com>
25390 PR tree-optimization/98736
25391 * tree-loop-distribution.c
25392 * (loop_distribution::bb_top_order_init):
25393 Compute RPO with programing order preserved by calling function
25394 rev_post_order_and_mark_dfs_back_seme.
25396 2021-04-06 Vladimir N. Makarov <vmakarov@redhat.com>
25399 * lra-constraints.c (split_reg): Don't check paradoxical_subreg_p.
25400 * lra-lives.c (clear_sparseset_regnos, regnos_in_sparseset_p): New
25402 (process_bb_lives): Don't update biggest mode of hard reg for
25403 implicit in multi-register group. Use the new functions for
25404 updating dead_set and unused_set by register notes.
25406 2021-04-06 Xianmiao Qu <xianmiao_qu@c-sky.com>
25408 * config/csky/csky_pipeline_ck802.md : Use insn reservation name
25411 2021-04-06 H.J. Lu <hjl.tools@gmail.com>
25413 * config/i386/x86-tune-costs.h (skylake_memcpy): Updated.
25414 (skylake_memset): Likewise.
25415 (skylake_cost): Change CLEAR_RATIO to 17.
25416 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
25417 Replace m_CANNONLAKE, m_ICELAKE_CLIENT, m_ICELAKE_SERVER,
25418 m_TIGERLAKE and m_SAPPHIRERAPIDS with m_SKYLAKE and m_CORE_AVX512.
25420 2021-04-06 Richard Biener <rguenther@suse.de>
25422 PR tree-optimization/99880
25423 * tree-vect-loop.c (maybe_set_vectorized_backedge_value): Only
25424 set vectorized defs of relevant PHIs.
25426 2021-04-06 Richard Biener <rguenther@suse.de>
25428 PR tree-optimization/99924
25429 * tree-vect-slp.c (vect_bb_partition_graph_r): Do not mark
25430 nodes w/o scalar stmts as visited.
25432 2021-04-06 Alex Coplan <alex.coplan@arm.com>
25435 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Also use base
25436 PCS for [su]fix_optab.
25438 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
25440 * config/darwin.c (machopic_legitimize_pic_address): Check
25441 that the current pic register is one of the hard reg set
25442 before setting liveness.
25444 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
25446 * config/darwin.c (machopic_legitimize_pic_address): Fix
25447 whitespace, remove unused code.
25449 2021-04-03 Jakub Jelinek <jakub@redhat.com>
25451 PR tree-optimization/99882
25452 * gimple-ssa-store-merging.c (bswap_view_convert): Handle val with
25455 2021-04-03 Jakub Jelinek <jakub@redhat.com>
25457 PR rtl-optimization/99863
25458 * dse.c (replace_read): Drop regs_live argument. Instead of
25459 regs_live, use store_insn->fixed_regs_live if non-NULL,
25460 otherwise punt if insns sequence clobbers or sets any hard
25463 2021-04-03 Jakub Jelinek <jakub@redhat.com>
25466 * targhooks.h (default_print_patchable_function_entry_1): Declare.
25467 * targhooks.c (default_print_patchable_function_entry_1): New function,
25468 copied from default_print_patchable_function_entry with an added flags
25470 (default_print_patchable_function_entry): Rewritten into a small
25471 wrapper around default_print_patchable_function_entry_1.
25472 * config/rs6000/rs6000.c (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
25474 (rs6000_print_patchable_function_entry): New function.
25476 2021-04-02 Eric Botcazou <ebotcazou@adacore.com>
25478 * doc/invoke.texi (fdelete-dead-exceptions): Minor tweak.
25480 2021-04-01 Jason Merrill <jason@redhat.com>
25483 * common.opt: Document v15 and v16.
25485 2021-04-01 Richard Biener <rguenther@suse.de>
25487 PR tree-optimization/99863
25488 * gimplify.c (gimplify_init_constructor): Recompute vector
25491 2021-04-01 Jakub Jelinek <jakub@redhat.com>
25493 * doc/extend.texi (symver attribute): Fix up syntax errors
25496 2021-04-01 Jakub Jelinek <jakub@redhat.com>
25498 PR tree-optimization/96573
25499 * gimple-ssa-store-merging.c (init_symbolic_number): Handle
25500 also pointer types.
25502 2021-04-01 Richard Biener <rguenther@suse.de>
25504 PR tree-optimization/99856
25505 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Promote
25506 precision to vector element precision.
25508 2021-04-01 Martin Jambor <mjambor@suse.cz>
25510 PR tree-optimization/97009
25511 * tree-sra.c (access_or_its_child_written): New function.
25512 (propagate_subaccesses_from_rhs): Use it instead of a simple grp_write
25515 2021-03-31 Jan Hubicka <hubicka@ucw.cz>
25518 * cif-code.def (USES_COMDAT_LOCAL): Make CIF_FINAL_NORMAL.
25520 2021-03-31 Pat Haugen <pthaugen@linux.ibm.com>
25523 * config/rs6000/altivec.md (xxspltiw_v4si, xxspltiw_v4sf_inst,
25524 xxspltidp_v2df_inst, xxsplti32dx_v4si_inst, xxsplti32dx_v4sf_inst,
25525 xxblend_<mode>, xxpermx_inst, xxeval): Mark prefixed.
25526 * config/rs6000/mma.md (mma_<vvi4i4i8>, mma_<avvi4i4i8>,
25527 mma_<vvi4i4i2>, mma_<avvi4i4i2>, mma_<vvi4i4>, mma_<avvi4i4>,
25528 mma_<pvi4i2>, mma_<apvi4i2>, mma_<vvi4i4i4>, mma_<avvi4i4i4>):
25530 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Adjust test.
25531 * config/rs6000/rs6000.md (define_attr "maybe_prefixed"): New.
25532 (define_attr "prefixed"): Update initializer.
25534 2021-03-31 Jakub Jelinek <jakub@redhat.com>
25537 * dwarf2out.c (debug_ranges_dwo_section): New variable.
25538 (DW_RANGES_IDX_SKELETON): Define.
25539 (struct dw_ranges): Add begin_entry and end_entry members.
25540 (DEBUG_DWO_RNGLISTS_SECTION): Define.
25541 (add_ranges_num): Adjust r initializer for addition of *_entry
25543 (add_ranges_by_labels): For -gsplit-dwarf and force_direct,
25544 set idx to DW_RANGES_IDX_SKELETON.
25545 (use_distinct_base_address_for_range): New function.
25546 (index_rnglists): Don't set r->idx if it is equal to
25547 DW_RANGES_IDX_SKELETON. Initialize r->begin_entry and
25548 r->end_entry for -gsplit-dwarf if those will be needed by
25550 (output_rnglists): Add DWO argument. If true, switch to
25551 debug_ranges_dwo_section rather than debug_ranges_section.
25552 Adjust l1/l2 label indexes. Only output the offset table when
25553 dwo is true and don't include in there the skeleton range
25554 entry if present. For -gsplit-dwarf, skip ranges that belong
25555 to the other rnglists section. Change return type from void
25556 to bool and return true if there are any range entries for
25557 the other section. For dwarf_split_debug_info use
25558 DW_RLE_startx_endx, DW_RLE_startx_length and DW_RLE_base_addressx
25559 entries instead of DW_RLE_start_end, DW_RLE_start_length and
25560 DW_RLE_base_address. Use use_distinct_base_address_for_range.
25561 (init_sections_and_labels): Initialize debug_ranges_dwo_section
25562 if -gsplit-dwarf and DWARF >= 5. Adjust ranges_section_label
25563 and range_base_label indexes.
25564 (dwarf2out_finish): Call index_rnglists earlier before finalizing
25565 .debug_addr. Never emit DW_AT_rnglists_base attribute. For
25566 -gsplit-dwarf and DWARF >= 5 call output_rnglists up to twice
25567 with different dwo arguments.
25568 (dwarf2out_c_finalize): Clear debug_ranges_dwo_section.
25570 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
25572 PR tree-optimization/98268
25573 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Call
25574 recompute_tree_invariant_for_addr_expr after successfully
25575 folding a TARGET_MEM_REF that occurs inside an ADDR_EXPR.
25577 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
25579 PR tree-optimization/99726
25580 * tree-data-ref.c (create_intersect_range_checks_index): Bail
25581 out if there is more than one access function SCEV for the loop
25584 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
25586 PR rtl-optimization/97141
25587 PR rtl-optimization/98726
25588 * emit-rtl.c (valid_for_const_vector_p): Return true for
25590 * rtx-vector-builder.h (rtx_vector_builder::step): Return a
25591 poly_wide_int instead of a wide_int.
25592 (rtx_vector_builder::apply_set): Take a poly_wide_int instead
25594 * rtx-vector-builder.c (rtx_vector_builder::apply_set): Likewise.
25595 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Return
25596 false for CONST_VECTORs that cannot be forced to memory.
25597 * config/aarch64/aarch64-simd.md (mov<mode>): If a CONST_VECTOR
25598 is too complex to force to memory, build it up from individual
25601 2021-03-31 Jan Hubicka <jh@suse.cz>
25604 * cgraph.c (cgraph_node::release_body): Fix overactive check.
25606 2021-03-31 Christophe Lyon <christophe.lyon@linaro.org>
25609 * config/arm/vec-common.md (mul<mode>3): Disable on iwMMXT, expect
25612 2021-03-31 H.J. Lu <hjl.tools@gmail.com>
25614 * config/i386/i386-expand.c (expand_set_or_cpymem_via_rep):
25615 For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, don't convert QImode
25617 (decide_alg): For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, use
25618 "rep movsb/stosb" only for known sizes.
25619 * config/i386/i386-options.c (processor_cost_table): Use Ice
25620 Lake cost for Cannon Lake, Ice Lake, Tiger Lake, Sapphire
25621 Rapids and Alder Lake.
25622 * config/i386/i386.h (TARGET_PREFER_KNOWN_REP_MOVSB_STOSB): New.
25623 * config/i386/x86-tune-costs.h (icelake_memcpy): New.
25624 (icelake_memset): Likewise.
25625 (icelake_cost): Likewise.
25626 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
25629 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
25632 * config/aarch64/aarch64.c
25633 (aarch64_vectorize_preferred_vector_alignment): Query the size
25634 of the provided SVE vector; do not assume that all SVE vectors
25635 have the same size.
25637 2021-03-31 Jan Hubicka <jh@suse.cz>
25640 * cgraph.c (cgraph_node::release_body): Remove all callers and
25642 * cgraphclones.c (cgraph_node::materialize_clone): Do not do it here.
25643 * cgraphunit.c (cgraph_node::expand): And here.
25645 2021-03-31 Martin Liska <mliska@suse.cz>
25647 * ipa-modref.c (analyze_ssa_name_flags): Fix coding style
25648 and one negated condition.
25650 2021-03-31 Jakub Jelinek <jakub@redhat.com>
25651 Richard Sandiford <richard.sandiford@arm.com>
25654 * config/aarch64/aarch64.md (*add<mode>3_poly_1): Swap Uai and Uav
25655 constraints on operands[2] and similarly 0 and rk constraints
25656 on operands[1] corresponding to that.
25658 2021-03-31 Jakub Jelinek <jakub@redhat.com>
25661 * configure.ac (HAVE_LD_BROKEN_PE_DWARF5): New AC_DEFINE if PECOFF
25662 linker doesn't support DWARF sections new in DWARF5.
25663 * config/i386/i386-options.c (ix86_option_override_internal): Default
25664 to dwarf_version 4 if HAVE_LD_BROKEN_PE_DWARF5 for TARGET_PECOFF
25666 * config.in: Regenerated.
25667 * configure: Regenerated.
25669 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25672 * config/aarch64/aarch64.c (aarch64_analyze_loop_vinfo): Check for
25673 available issue_info before using it.
25675 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25678 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): Do not allow zero
25681 2021-03-30 Xionghu Luo <luoxhu@linux.ibm.com>
25684 * config/rs6000/altivec.md (altivec_lvsl_reg): Change to ...
25685 (altivec_lvsl_reg_<mode>): ... this.
25686 (altivec_lvsr_reg): Change to ...
25687 (altivec_lvsr_reg_<mode>): ... this.
25688 * config/rs6000/predicates.md (vec_set_index_operand): New.
25689 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
25690 Enable 32bit variable vec_insert for all TARGET_VSX.
25691 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
25692 Enable 32bit variable vec_insert for p9 and above.
25693 (rs6000_expand_vector_set_var_p8): Rename to ...
25694 (rs6000_expand_vector_set_var_p7): ... this.
25695 (rs6000_expand_vector_set): Use TARGET_VSX and adjust assert
25697 * config/rs6000/vector.md (vec_set<mode>): Use vec_set_index_operand.
25698 * config/rs6000/vsx.md (xl_len_r): Use gen_altivec_lvsl_reg_di and
25699 gen_altivec_lvsr_reg_di.
25701 2021-03-30 H.J. Lu <hjl.tools@gmail.com>
25704 * config/i386/ia32intrin.h (__rdtsc): Defined as macro.
25705 (__rdtscp): Likewise.
25707 2021-03-30 Tamar Christina <tamar.christina@arm.com>
25709 PR tree-optimization/99825
25710 * tree-vect-slp-patterns.c (vect_check_evenodd_blend):
25711 Reject non-mult 2 lanes.
25713 2021-03-30 Richard Earnshaw <rearnsha@arm.com>
25716 * config/arm/arm.c (arm_file_start): Fix emission of
25717 Tag_ABI_VFP_args attribute.
25719 2021-03-30 Richard Biener <rguenther@suse.de>
25721 PR tree-optimization/99824
25722 * stor-layout.c (set_min_and_max_values_for_integral_type):
25723 Assert the precision is within the bounds of
25724 WIDE_INT_MAX_PRECISION.
25725 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
25726 the outermost component ref only to lower the access size
25727 and initialize that from the access type.
25729 2021-03-30 Richard Sandiford <richard.sandiford@arm.com>
25732 * config/aarch64/aarch64.md (mov<mode>): Pass multi-instruction
25733 CONST_INTs to aarch64_expand_mov_immediate when called after RA.
25735 2021-03-30 Mihailo Stojanovic <mihailo.stojanovic@typhoon-hil.com>
25737 * config/aarch64/aarch64.md
25738 (<optab>_trunc<fcvt_target><GPI:mode>2): Set the "arch"
25739 attribute to disambiguate between SIMD and FP variants of the
25742 2021-03-29 Jan Hubicka <hubicka@ucw.cz>
25744 * ipa-modref.c (merge_call_lhs_flags): Correct handling of deref.
25745 (analyze_ssa_name_flags): Fix typo in comment.
25747 2021-03-29 Alex Coplan <alex.coplan@arm.com>
25750 * config/aarch64/aarch64-sve-builtins.cc
25751 (function_builder::add_function): Add placeholder_p argument, use
25752 placeholder decls if this is set.
25753 (function_builder::add_unique_function): Instead of conditionally adding
25754 direct overloads, unconditionally add either a direct overload or a
25756 (function_builder::add_overloaded_function): Set placeholder_p if we're
25757 using C++ overloads. Use the obstack for string storage instead
25758 of relying on the tree nodes.
25759 (function_builder::add_overloaded_functions): Don't return early for
25760 m_direct_overloads: we need to add placeholders.
25761 * config/aarch64/aarch64-sve-builtins.h
25762 (function_builder::add_function): Add placeholder_p argument.
25764 2021-03-29 Richard Biener <rguenther@suse.de>
25766 PR tree-optimization/99807
25767 * tree-vect-slp.c (vect_slp_analyze_node_operations_1): Move
25768 assert below VEC_PERM handling.
25770 2021-03-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25773 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>): Use
25774 aarch64_simd_or_scalar_imm_zero to match zeroes. Remove pattern
25775 matching const_int 0.
25776 (move_lo_quad_internal_be_<mode>): Likewise.
25777 (move_lo_quad_<mode>): Update for the above.
25778 * config/aarch64/iterators.md (VQ_2E): Delete.
25780 2021-03-29 Jakub Jelinek <jakub@redhat.com>
25782 PR tree-optimization/99777
25783 * fold-const.c (extract_muldiv_1): For conversions, punt on casts from
25784 types other than scalar integral types.
25786 2021-03-28 David Edelsohn <dje.gcc@gmail.com>
25788 * config/rs6000/rs6000.c (rs6000_output_dwarf_dtprel): Do not add
25789 XCOFF TLS reloc decorations.
25791 2021-03-28 Gerald Pfeifer <gerald@pfeifer.com>
25793 * doc/analyzer.texi (Analyzer Internals): Update link to
25794 "A Memory Model for Static Analysis of C Programs".
25796 2021-03-26 David Edelsohn <dje.gcc@gmail.com>
25798 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Call function.
25799 * config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align):
25801 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align): New.
25802 (rs6000_special_round_type_align): Recursively check innermost first
25805 2021-03-26 Jakub Jelinek <jakub@redhat.com>
25808 * dwarf2out.h (struct dw_fde_node): Add rule18 member.
25809 * dwarf2cfi.c (dwarf2out_frame_debug_expr): When handling (set hfp sp)
25810 assignment with drap_reg active, queue reg save for hfp with offset 0
25811 and flush queued reg saves. When handling a push with rule18,
25812 defer queueing reg save for hfp and just assert the offset is 0.
25813 (scan_trace): Assert that fde->rule18 is false.
25815 2021-03-26 Vladimir Makarov <vmakarov@redhat.com>
25818 * ira-costs.c (record_reg_classes): Put case with
25819 CT_RELAXED_MEMORY adjacent to one with CT_MEMORY.
25820 * ira.c (ira_setup_alts): Ditto.
25821 * lra-constraints.c (process_alt_operands): Ditto.
25822 * recog.c (asm_operand_ok): Ditto.
25823 * reload.c (find_reloads): Ditto.
25825 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
25827 * config/aarch64/aarch64-protos.h
25828 (cpu_addrcost_table::post_modify_ld3_st3): New member variable.
25829 (cpu_addrcost_table::post_modify_ld4_st4): Likewise.
25830 * config/aarch64/aarch64.c (generic_addrcost_table): Update
25831 accordingly, using the same costs as for post_modify.
25832 (exynosm1_addrcost_table, xgene1_addrcost_table): Likewise.
25833 (thunderx2t99_addrcost_table, thunderx3t110_addrcost_table):
25834 (tsv110_addrcost_table, qdf24xx_addrcost_table): Likewise.
25835 (a64fx_addrcost_table): Likewise.
25836 (neoversev1_addrcost_table): New.
25837 (neoversev1_tunings): Use neoversev1_addrcost_table.
25838 (aarch64_address_cost): Use the new post_modify costs for CImode
25841 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
25843 * config/aarch64/aarch64.opt
25844 (-param=aarch64-loop-vect-issue-rate-niters=): New parameter.
25845 * doc/invoke.texi: Document it.
25846 * config/aarch64/aarch64-protos.h (aarch64_base_vec_issue_info)
25847 (aarch64_scalar_vec_issue_info, aarch64_simd_vec_issue_info)
25848 (aarch64_advsimd_vec_issue_info, aarch64_sve_vec_issue_info)
25849 (aarch64_vec_issue_info): New structures.
25850 (cpu_vector_cost): Write comments above the variables rather
25852 (cpu_vector_cost::issue_info): New member variable.
25853 * config/aarch64/aarch64.c: Include gimple-pretty-print.h
25854 and tree-ssa-loop-niter.h.
25855 (generic_vector_cost, a64fx_vector_cost, qdf24xx_vector_cost)
25856 (thunderx_vector_cost, tsv110_vector_cost, cortexa57_vector_cost)
25857 (exynosm1_vector_cost, xgene1_vector_cost, thunderx2t99_vector_cost)
25858 (thunderx3t110_vector_cost): Initialize issue_info to null.
25859 (neoversev1_scalar_issue_info, neoversev1_advsimd_issue_info)
25860 (neoversev1_sve_issue_info, neoversev1_vec_issue_info): New structures.
25861 (neoversev1_vector_cost): Use them.
25862 (aarch64_vec_op_count, aarch64_sve_op_count): New structures.
25863 (aarch64_vector_costs::saw_sve_only_op): New member variable.
25864 (aarch64_vector_costs::num_vector_iterations): Likewise.
25865 (aarch64_vector_costs::scalar_ops): Likewise.
25866 (aarch64_vector_costs::advsimd_ops): Likewise.
25867 (aarch64_vector_costs::sve_ops): Likewise.
25868 (aarch64_vector_costs::seen_loads): Likewise.
25869 (aarch64_simd_vec_costs_for_flags): New function.
25870 (aarch64_analyze_loop_vinfo): Initialize num_vector_iterations.
25871 Count the number of predicate operations required by SVE WHILE
25873 (aarch64_comparison_type, aarch64_multiply_add_p): New functions.
25874 (aarch64_sve_only_stmt_p, aarch64_in_loop_reduction_latency): Likewise.
25875 (aarch64_count_ops): Likewise.
25876 (aarch64_add_stmt_cost): Record whether see an SVE operation
25877 that cannot currently be implementing using Advanced SIMD.
25878 Record issue information about the scalar, Advanced SIMD
25879 and (where relevant) SVE versions of a loop.
25880 (aarch64_vec_op_count::dump): New function.
25881 (aarch64_sve_op_count::dump): Likewise.
25882 (aarch64_estimate_min_cycles_per_iter): Likewise.
25883 (aarch64_adjust_body_cost): If issue information is available,
25884 try to compare the issue rates of the various loop implementations
25885 and increase or decrease the vector body cost accordingly.
25887 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
25889 * config/aarch64/aarch64.c (aarch64_detect_vector_stmt_subtype):
25890 Assume a zero cost for induction phis.
25892 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
25894 * config/aarch64/aarch64.c (aarch64_embedded_comparison_type): New
25896 (aarch64_adjust_stmt_cost): Add the costs of embedded scalar and
25897 vector comparisons.
25899 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
25901 * config/aarch64/aarch64.c (aarch64_detect_scalar_stmt_subtype):
25903 (aarch64_add_stmt_cost): Call it.
25905 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
25907 * config/aarch64/aarch64-tuning-flags.def (matched_vector_throughput):
25908 New tuning parameter.
25909 * config/aarch64/aarch64.c (neoversev1_tunings): Use it.
25910 (aarch64_estimated_sve_vq): New function.
25911 (aarch64_vector_costs::analyzed_vinfo): New member variable.
25912 (aarch64_vector_costs::is_loop): Likewise.
25913 (aarch64_vector_costs::unrolled_advsimd_niters): Likewise.
25914 (aarch64_vector_costs::unrolled_advsimd_stmts): Likewise.
25915 (aarch64_record_potential_advsimd_unrolling): New function.
25916 (aarch64_analyze_loop_vinfo, aarch64_analyze_bb_vinfo): Likewise.
25917 (aarch64_add_stmt_cost): Call aarch64_analyze_loop_vinfo or
25918 aarch64_analyze_bb_vinfo on the first use of a costs structure.
25919 Detect whether we're vectorizing a loop for SVE that might be
25920 completely unrolled if it used Advanced SIMD instead.
25921 (aarch64_adjust_body_cost_for_latency): New function.
25922 (aarch64_finish_cost): Call it.
25924 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
25926 * config/aarch64/aarch64.c (aarch64_vector_costs): New structure.
25927 (aarch64_init_cost): New function.
25928 (aarch64_add_stmt_cost): Use aarch64_vector_costs instead of
25929 the default unsigned[3].
25930 (aarch64_finish_cost, aarch64_destroy_cost_data): New functions.
25931 (TARGET_VECTORIZE_INIT_COST): Override.
25932 (TARGET_VECTORIZE_FINISH_COST): Likewise.
25933 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
25935 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
25937 * config/aarch64/aarch64.c (neoversev1_advsimd_vector_cost)
25938 (neoversev1_sve_vector_cost): New cost structures.
25939 (neoversev1_vector_cost): Likewise.
25940 (neoversev1_tunings): Use them. Enable use_new_vector_costs.
25942 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
25944 * config/aarch64/aarch64-protos.h
25945 (sve_vec_cost::scatter_store_elt_cost): New member variable.
25946 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
25947 accordingly, taking the cost from the cost of a scalar_store.
25948 (a64fx_sve_vector_cost): Likewise.
25949 (aarch64_detect_vector_stmt_subtype): Detect scatter stores.
25951 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
25953 * config/aarch64/aarch64-protos.h
25954 (simd_vec_cost::store_elt_extra_cost): New member variable.
25955 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
25956 accordingly, using the vec_to_scalar cost for the new field.
25957 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
25958 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
25959 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
25960 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
25961 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
25962 (thunderx3t110_advsimd_vector_cost): Likewise.
25963 (aarch64_detect_vector_stmt_subtype): Detect single-element stores.
25965 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
25967 * config/aarch64/aarch64-protos.h (simd_vec_cost::ld2_st2_permute_cost)
25968 (simd_vec_cost::ld3_st3_permute_cost): New member variables.
25969 (simd_vec_cost::ld4_st4_permute_cost): Likewise.
25970 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
25971 accordingly, using zero for the new costs.
25972 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
25973 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
25974 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
25975 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
25976 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
25977 (thunderx3t110_advsimd_vector_cost): Likewise.
25978 (aarch64_ld234_st234_vectors): New function.
25979 (aarch64_adjust_stmt_cost): Likewise.
25980 (aarch64_add_stmt_cost): Call aarch64_adjust_stmt_cost if using
25981 the new vector costs.
25983 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
25985 * config/aarch64/aarch64-protos.h (sve_vec_cost): Turn into a
25986 derived class of simd_vec_cost. Add information about CLAST[AB]
25987 and FADDA instructions.
25988 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
25989 accordingly, using the vec_to_scalar costs for the new fields.
25990 (a64fx_sve_vector_cost): Likewise.
25991 (aarch64_reduc_type): New function.
25992 (aarch64_sve_in_loop_reduction_latency): Likewise.
25993 (aarch64_detect_vector_stmt_subtype): Take a vinfo parameter.
25994 Use aarch64_sve_in_loop_reduction_latency to handle SVE reductions
25995 that occur in the loop body.
25996 (aarch64_add_stmt_cost): Update call accordingly.
25998 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
26000 * config/aarch64/aarch64-tuning-flags.def (use_new_vector_costs):
26002 * config/aarch64/aarch64-protos.h (simd_vec_cost): Put comments
26003 above the fields rather than to the right.
26004 (simd_vec_cost::reduc_i8_cost): New member variable.
26005 (simd_vec_cost::reduc_i16_cost): Likewise.
26006 (simd_vec_cost::reduc_i32_cost): Likewise.
26007 (simd_vec_cost::reduc_i64_cost): Likewise.
26008 (simd_vec_cost::reduc_f16_cost): Likewise.
26009 (simd_vec_cost::reduc_f32_cost): Likewise.
26010 (simd_vec_cost::reduc_f64_cost): Likewise.
26011 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
26012 accordingly, using the vec_to_scalar_cost for the new fields.
26013 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
26014 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
26015 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
26016 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
26017 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
26018 (thunderx3t110_advsimd_vector_cost): Likewise.
26019 (aarch64_use_new_vector_costs_p): New function.
26020 (aarch64_simd_vec_costs): New function, split out from...
26021 (aarch64_builtin_vectorization_cost): ...here.
26022 (aarch64_is_reduction): New function.
26023 (aarch64_detect_vector_stmt_subtype): Likewise.
26024 (aarch64_add_stmt_cost): Call aarch64_detect_vector_stmt_subtype if
26025 using the new vector costs.
26027 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
26030 * tree-emutls.c (get_emutls_init_templ_addr): Mark initializer of weak
26031 TLS declarations as public.
26033 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
26035 * config/aarch64/aarch64-d.c (IN_TARGET_CODE): Define.
26036 * config/arm/arm-d.c (IN_TARGET_CODE): Likewise.
26037 * config/i386/i386-d.c (IN_TARGET_CODE): Likewise.
26038 * config/mips/mips-d.c (IN_TARGET_CODE): Likewise.
26039 * config/pa/pa-d.c (IN_TARGET_CODE): Likewise.
26040 * config/riscv/riscv-d.c (IN_TARGET_CODE): Likewise.
26041 * config/rs6000/rs6000-d.c (IN_TARGET_CODE): Likewise.
26042 * config/s390/s390-d.c (IN_TARGET_CODE): Likewise.
26043 * config/sparc/sparc-d.c (IN_TARGET_CODE): Likewise.
26045 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
26048 * config.gcc (*-*-cygwin*): Add winnt-d.o
26049 (*-*-mingw*): Likewise.
26050 * config/i386/cygwin.h (EXTRA_TARGET_D_OS_VERSIONS): New macro.
26051 * config/i386/mingw32.h (EXTRA_TARGET_D_OS_VERSIONS): Likewise.
26052 * config/i386/t-cygming: Add winnt-d.o.
26053 * config/i386/winnt-d.c: New file.
26055 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
26057 * config/freebsd-d.c: Include memmodel.h.
26059 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
26062 * config.gcc (*-*-openbsd*): Add openbsd-d.o.
26063 * config/t-openbsd: Add openbsd-d.o.
26064 * config/openbsd-d.c: New file.
26066 2021-03-25 Stam Markianos-Wright <stam.markianos-wright@arm.com>
26068 PR tree-optimization/96974
26069 * tree-vect-stmts.c (vect_get_vector_types_for_stmt): Replace assert
26070 with graceful exit.
26072 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
26075 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
26079 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
26080 always_inline in system headers.
26082 2021-03-25 Kewen Lin <linkw@linux.ibm.com>
26084 * tree-vect-loop.c (vect_model_reduction_cost): Init inside_cost.
26086 2021-03-25 Jakub Jelinek <jakub@redhat.com>
26089 * tree-core.h (enum operand_equal_flag): Add OEP_ADDRESS_OF_SAME_FIELD.
26090 * fold-const.c (operand_compare::operand_equal_p): Don't compare
26091 field offsets if OEP_ADDRESS_OF_SAME_FIELD.
26093 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
26097 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
26098 always_inline in system headers.
26100 2021-03-25 Richard Biener <rguenther@suse.de>
26102 PR tree-optimization/99746
26103 * tree-vect-slp-patterns.c (complex_pattern::build): Do not mark
26104 the scalar stmt as patterned. Instead set up required things
26107 2021-03-25 Xionghu Luo <luoxhu@linux.ibm.com>
26109 * config/rs6000/rs6000.c (power8_costs): Change l2 cache
26112 2021-03-24 Martin Liska <mliska@suse.cz>
26115 * common/config/i386/i386-common.c (ARRAY_SIZE): Fix off-by-one
26117 * config/i386/i386-options.c (ix86_option_override_internal):
26118 Add run-time assert.
26120 2021-03-24 Martin Jambor <mjambor@suse.cz>
26123 * ipa-cp.c (initialize_node_lattices): Mark as bottom all
26124 parameters with unknown type.
26125 (ipacp_value_safe_for_type): New function.
26126 (propagate_vals_across_arith_jfunc): Verify that the constant type
26127 can be used for a type of the formal parameter.
26128 (propagate_vals_across_ancestor): Likewise.
26129 (propagate_scalar_across_jump_function): Likewise. Pass the type
26130 also to propagate_vals_across_ancestor.
26132 2021-03-24 Christophe Lyon <christophe.lyon@linaro.org>
26135 * config/arm/mve.md (movmisalign<mode>_mve_store): Use Ux
26137 (movmisalign<mode>_mve_load): Likewise.
26139 2021-03-24 Jakub Jelinek <jakub@redhat.com>
26142 * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
26143 movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.
26145 2021-03-24 Alexandre Oliva <oliva@adacore.com>
26147 * doc/sourcebuild.texi (sysconf): New effective target.
26149 2021-03-24 Alexandre Oliva <oliva@adacore.com>
26151 * config/i386/predicates.md (reg_or_const_vec_operand): New.
26152 * config/i386/sse.md (ssse3_pshufbv8qi3): Add an expander for
26153 the now *-prefixed insn_and_split, turn the splitter const vec
26154 into an input for the insn, making it an ignored immediate for
26155 non-split cases, and loaded into the scratch register
26158 2021-03-23 Vladimir N. Makarov <vmakarov@redhat.com>
26161 * config/aarch64/constraints.md (Utq, UOb, UOh, UOw, UOd, UOty):
26162 Use define_relaxed_memory_constraint for them.
26164 2021-03-23 Iain Sandoe <iain@sandoe.co.uk>
26167 * config/host-darwin.c (darwin_gt_pch_use_address): Add a
26168 colon to the diagnostic message.
26170 2021-03-23 Ilya Leoshkevich <iii@linux.ibm.com>
26172 * fwprop.c (fwprop_propagation::fwprop_propagation): Look at
26174 (try_fwprop_subst_note): Use set_info instead of insn_info.
26175 (try_fwprop_subst_pattern): Likewise.
26176 (try_fwprop_subst_notes): Likewise.
26177 (try_fwprop_subst): Likewise.
26178 (forward_propagate_subreg): Likewise.
26179 (forward_propagate_and_simplify): Likewise.
26180 (forward_propagate_into): Likewise.
26181 * rtl-ssa/accesses.h (set_info::single_nondebug_use) New
26183 (set_info::single_nondebug_insn_use): Likewise.
26184 (set_info::single_phi_use): Likewise.
26185 * rtl-ssa/member-fns.inl (set_info::single_nondebug_use) New
26187 (set_info::single_nondebug_insn_use): Likewise.
26188 (set_info::single_phi_use): Likewise.
26190 2021-03-23 Christophe Lyon <christophe.lyon@linaro.org>
26192 * doc/sourcebuild.texi (arm_dsp_ok, arm_dsp): Document.
26194 2021-03-23 Jakub Jelinek <jakub@redhat.com>
26197 * config/aarch64/aarch64.c (aarch64_add_offset): Tell
26198 expand_mult to perform an unsigned rather than a signed
26201 2021-03-23 H.J. Lu <hjl.tools@gmail.com>
26204 * config/i386/cpuid.h (__cpuid): Add __volatile__.
26205 (__cpuid_count): Likewise.
26207 2021-03-23 Richard Biener <rguenther@suse.de>
26209 PR tree-optimization/99721
26210 * tree-vect-slp.c (vect_slp_analyze_node_operations):
26211 Make sure we can schedule the node.
26213 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
26215 * config/riscv/riscv.c (riscv_subword): Take endianness into
26216 account when calculating the byte offset.
26218 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
26220 * config/riscv/predicates.md (subreg_lowpart_operator): New predicate
26221 * config/riscv/riscv.md (*addsi3_extended2, *subsi3_extended2)
26222 (*negsi2_extended2, *mulsi3_extended2, *<optab>si3_mask)
26223 (*<optab>si3_mask_1, *<optab>di3_mask, *<optab>di3_mask_1)
26224 (*<optab>si3_extend_mask, *<optab>si3_extend_mask_1): Use
26225 new predicate "subreg_lowpart_operator"
26227 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
26229 * config/riscv/riscv.c (riscv_swap_instruction): New function
26230 to byteswap an SImode rtx containing an instruction.
26231 (riscv_trampoline_init): Byteswap the generated instructions
26234 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
26236 * common/config/riscv/riscv-common.c
26237 (TARGET_DEFAULT_TARGET_FLAGS): Set default endianness.
26238 * config.gcc (riscv32be-*, riscv64be-*): Set
26239 TARGET_BIG_ENDIAN_DEFAULT to 1.
26240 * config/riscv/elf.h (LINK_SPEC): Change -melf* value
26241 depending on default endianness.
26242 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
26243 * config/riscv/linux.h (LINK_SPEC): Likewise.
26244 * config/riscv/riscv.c (TARGET_DEFAULT_TARGET_FLAGS): Set
26245 default endianness.
26246 * config/riscv/riscv.h (DEFAULT_ENDIAN_SPEC): New macro.
26248 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
26250 * config/riscv/elf.h (LINK_SPEC): Pass linker endianness flag.
26251 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
26252 * config/riscv/linux.h (LINK_SPEC): Likewise.
26253 * config/riscv/riscv.h (ASM_SPEC): Pass -mbig-endian and
26255 (BYTES_BIG_ENDIAN): Handle big endian.
26256 (WORDS_BIG_ENDIAN): Define to BYTES_BIG_ENDIAN.
26257 * config/riscv/riscv.opt (-mbig-endian, -mlittle-endian): New
26259 * doc/invoke.texi (-mbig-endian, -mlittle-endian): Document.
26261 2021-03-23 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
26263 * regcprop.c (find_oldest_value_reg): Ask target whether
26264 different mode is fine for replacement register.
26266 2021-03-23 Aldy Hernandez <aldyh@redhat.com>
26268 PR tree-optimization/99296
26269 * value-range.cc (irange::irange_set_1bit_anti_range): New.
26270 (irange::irange_set_anti_range): Call irange_set_1bit_anti_range
26271 * value-range.h (irange::irange_set_1bit_anti_range): New.
26273 2021-03-22 Vladimir N. Makarov <vmakarov@redhat.com>
26276 * config/aarch64/constraints.md (UtQ): Use
26277 define_relaxed_memory_constraint for it.
26278 * doc/md.texi (define_relaxed_memory_constraint): Describe it.
26279 * genoutput.c (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
26280 * genpreds.c (constraint_data): Add bitfield is_relaxed_memory.
26281 (have_relaxed_memory_constraints): New static var.
26282 (relaxed_memory_start, relaxed_memory_end): Ditto.
26283 (add_constraint): Add arg is_relaxed_memory. Check name for
26284 relaxed memory. Set up is_relaxed_memory in constraint_data and
26285 have_relaxed_memory_constraints. Adjust calls.
26286 (choose_enum_order): Process relaxed memory.
26287 (write_tm_preds_h): Ditto.
26288 (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
26289 * gensupport.c (process_rtx): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
26290 * ira-costs.c (record_reg_classes): Process CT_RELAXED_MEMORY.
26291 * ira-lives.c (single_reg_class): Use
26292 insn_extra_relaxed_memory_constraint.
26293 * ira.c (ira_setup_alts): CT_RELAXED_MEMORY.
26294 * lra-constraints.c (valid_address_p): Use
26295 insn_extra_relaxed_memory_constraint instead of other memory
26297 (process_alt_operands): Process CT_RELAXED_MEMORY.
26298 (curr_insn_transform): Use insn_extra_relaxed_memory_constraint.
26299 * recog.c (asm_operand_ok, preprocess_constraints): Process
26301 * reload.c (find_reloads): Ditto.
26302 * rtl.def (DEFINE_RELAXED_MEMORY_CONSTRAINT): New.
26303 * stmt.c (parse_input_constraint): Use
26304 insn_extra_relaxed_memory_constraint.
26306 2021-03-22 Segher Boessenkool <segher@kernel.crashing.org>
26309 * ubsan.c (ubsan_instrument_float_cast): Don't test for unordered if
26312 2021-03-22 Alex Coplan <alex.coplan@arm.com>
26315 * config/arm/arm-protos.h (neon_make_constant): Add generate
26316 argument to guard emitting insns, default to true.
26317 * config/arm/arm.c (arm_legitimate_constant_p_1): Reject
26318 CONST_VECTORs which neon_make_constant can't handle.
26319 (neon_vdup_constant): Add generate argument, avoid emitting
26320 insns if it's not set.
26321 (neon_make_constant): Plumb new generate argument through.
26322 * config/arm/constraints.md (Ui): New. Use it...
26323 * config/arm/mve.md (*mve_mov<mode>): ... here.
26324 * config/arm/vec-common.md (movv8hf): Use neon_make_constant to
26325 synthesize constants.
26327 2021-03-22 Richard Biener <rguenther@suse.de>
26329 * debug.h: Add deprecation warning.
26331 2021-03-22 Richard Biener <rguenther@suse.de>
26333 PR tree-optimization/99694
26334 * tree-ssa-sccvn.c (visit_phi): Ignore edges with the
26337 2021-03-22 Kito Cheng <kito.cheng@sifive.com>
26340 * config/riscv/riscv.c (riscv_expand_block_move): Get RTL value
26341 after type checking.
26343 2021-03-22 Jakub Jelinek <jakub@redhat.com>
26347 * dwarf2out.c (get_full_len): Use get_precision rather than
26349 (add_const_value_attribute): Make sure add_AT_wide argument has
26350 precision prec rather than some very wide one.
26352 2021-03-22 Kewen Lin <linkw@linux.ibm.com>
26354 * config/rs6000/rs6000.md (*rotldi3_insert_sf,
26355 *mov<SFDF:mode><SFDF2:mode>cc_p9, floatsi<mode>2_lfiwax,
26356 floatsi<mode>2_lfiwax_mem, floatunssi<mode>2_lfiwzx,
26357 floatunssi<mode>2_lfiwzx_mem, *floatsidf2_internal,
26358 *floatunssidf2_internal, fix_trunc<mode>si2_stfiwx,
26359 fix_trunc<mode>si2_internal, fixuns_trunc<mode>si2_stfiwx,
26360 *round32<mode>2_fprs, *roundu32<mode>2_fprs,
26361 *fix_trunc<mode>si2_internal): Fix empty split condition.
26362 * config/rs6000/vsx.md (*vsx_le_undo_permute_<mode>,
26363 vsx_reduc_<VEC_reduc_name>_v2df, vsx_reduc_<VEC_reduc_name>_v4sf,
26364 *vsx_reduc_<VEC_reduc_name>_v2df_scalar,
26365 *vsx_reduc_<VEC_reduc_name>_v4sf_scalar): Likewise.
26367 2021-03-22 Xionghu Luo <luoxhu@linux.ibm.com>
26370 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
26371 Convert idx to DImode.
26372 (rs6000_expand_vector_set_var_p8): Likewise.
26374 2021-03-21 Jakub Jelinek <jakub@redhat.com>
26377 * dwarf2out.c (insert_float): Change return type from void to
26378 unsigned, handle GET_MODE_SIZE (mode) == 2 and return element size.
26379 (mem_loc_descriptor, loc_descriptor, add_const_value_attribute):
26382 2021-03-20 H.J. Lu <hjl.tools@gmail.com>
26385 * config/i386/i386.c (construct_container): Check cfun != NULL
26386 before accessing silent_p.
26388 2021-03-20 Ahamed Husni <ahamedhusni73@gmail.com>
26390 * asan.c: Fix typos in comments.
26392 2021-03-20 Vladimir N. Makarov <vmakarov@redhat.com>
26394 PR rtl-optimization/99680
26395 * lra-constraints.c (skip_contraint_modifiers): Rename to skip_constraint_modifiers.
26396 (process_address_1): Check empty constraint before using
26399 2021-03-19 Pat Haugen <pthaugen@linux.ibm.com>
26401 * config/rs6000/rs6000.c (power10_cost): New.
26402 (rs6000_option_override_internal): Set Power10 costs.
26403 (rs6000_issue_rate): Set Power10 issue rate.
26404 * config/rs6000/power10.md: Rewrite for Power10.
26406 2021-03-19 Vladimir N. Makarov <vmakarov@redhat.com>
26409 * lra-constraints.c (process_address_1): Don't use unknown
26410 constraint for address constraint.
26412 2021-03-19 Iain Sandoe <iain@sandoe.co.uk>
26415 * config.gcc (powerpc-*-darwin8): Delete the reference to
26416 the now removed darwin8.h.
26418 2021-03-19 Olivier Hainque <hainque@adacore.com>
26421 * config/vxworksae.h (VX_CPU_PREFIX): Define.
26423 2021-03-19 John David Anglin <danglin@gcc.gnu.org>
26425 * config/pa/pa.c (import_milli): Use memcpy instead of strncpy.
26427 2021-03-19 Tamar Christina <tamar.christina@arm.com>
26429 PR tree-optimization/99656
26430 * tree-vect-slp-patterns.c (linear_loads_p,
26431 complex_add_pattern::matches, is_eq_or_top,
26432 vect_validate_multiplication, complex_mul_pattern::matches,
26433 complex_fms_pattern::matches): Remove complex_perm_kinds_t.
26434 * tree-vectorizer.h: (complex_load_perm_t): Removed.
26435 (slp_tree_to_load_perm_map_t): Use complex_perm_kinds_t instead of
26436 complex_load_perm_t.
26438 2021-03-19 H.J. Lu <hjl.tools@gmail.com>
26441 * config/i386/i386-options.c (ix86_init_machine_status): Set
26443 * config/i386/i386.c (init_cumulative_args): Set silent_p to
26445 (construct_container): Return early for return and argument
26446 errors if silent_p is true.
26447 * config/i386/i386.h (machine_function): Add silent_p.
26449 2021-03-19 Jakub Jelinek <jakub@redhat.com>
26452 * config/arm/constraints.md (Ds): New constraint.
26453 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): Use w,Ds
26454 constraint instead of w,Dm.
26456 2021-03-19 Andrew Stubbs <ams@codesourcery.com>
26458 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Fix quotes
26461 2021-03-19 Eric Botcazou <ebotcazou@adacore.com>
26463 PR middle-end/99641
26464 * fold-const.c (native_encode_initializer) <CONSTRUCTOR>: For an
26465 array type, do the computation of the current position in sizetype.
26467 2021-03-18 Vladimir N. Makarov <vmakarov@redhat.com>
26470 * lra-constraints.c (process_address_1): Use lookup_constraint
26471 only for a single constraint.
26473 2021-03-18 Martin Sebor <msebor@redhat.com>
26475 PR middle-end/99502
26476 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): Rename...
26477 (inbounds_memaccess_p): ...to this. Check the ending offset of
26478 the accessed member.
26480 2021-03-18 Andrew Stubbs <ams@codesourcery.com>
26482 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Add %< and
26483 %> quote markers to error messages.
26484 (gcn_goacc_validate_dims): Likewise.
26485 (gcn_conditional_register_usage): Remove exclaimation mark from error
26487 (gcn_vectorize_vec_perm_const): Ensure perm is fully uninitialized.
26489 2021-03-18 Jan Hubicka <hubicka@ucw.cz>
26491 * config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
26494 2021-03-18 Sinan Lin <sinan@isrc.iscas.ac.cn>
26495 Kito Cheng <kito.cheng@sifive.com>
26497 * config/riscv/riscv.c (riscv_block_move_straight): Change type
26498 to unsigned HOST_WIDE_INT for parameter and local variable with
26499 HOST_WIDE_INT type.
26500 (riscv_adjust_block_mem): Ditto.
26501 (riscv_block_move_loop): Ditto.
26502 (riscv_expand_block_move): Ditto.
26504 2021-03-18 Nick Clifton <nickc@redhat.com>
26506 * config/v850/v850.c (construct_restore_jr): Increase static
26508 (construct_save_jarl): Likewise.
26509 * config/v850/v850.h (DWARF2_DEBUGGING_INFO): Define.
26511 2021-03-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26513 * config/aarch64/aarch64.c (aarch64_adjust_generic_arch_tuning): Define.
26514 (aarch64_override_options_internal): Use it.
26515 (generic_tunings): Add AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS to
26518 2021-03-17 Sandra Loosemore <sandra@codesourcery.com>
26520 * config/nios2/nios2.c (nios2_custom_check_insns): Clean up
26521 error message format issues.
26522 (nios2_option_override): Likewise.
26523 (nios2_expand_fpu_builtin): Likewise.
26524 (nios2_init_custom_builtins): Adjust to avoid bogus strncpy
26525 truncation warning.
26526 (nios2_expand_custom_builtin): More error message format fixes.
26527 (nios2_expand_rdwrctl_builtin): Likewise.
26528 (nios2_expand_rdprs_builtin): Likewise.
26529 (nios2_expand_eni_builtin): Likewise.
26530 (nios2_expand_builtin): Likewise.
26531 (nios2_register_custom_code): Likewise.
26532 (nios2_valid_target_attribute_rec): Likewise.
26533 (nios2_add_insn_asm): Fix uninitialized variable warning.
26535 2021-03-17 Jan Hubicka <jh@suse.cz>
26537 * config/i386/x86-tune-costs.h (struct processor_costs): Update costs
26538 of gather to match reality.
26539 * config/i386/x86-tune.def (X86_TUNE_USE_GATHER): Enable for znver3.
26541 2021-03-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26543 * config/aarch64/aarch64-builtins.c (aarch64_expand_rng_builtin): Use EQ
26544 to compare against CC_REG rather than NE.
26546 2021-03-17 H.J. Lu <hjl.tools@gmail.com>
26549 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support
26550 inline assembly statements.
26551 (ix86_print_operand): Update 'P' handling for -fno-plt.
26553 2021-03-17 Tamar Christina <tamar.christina@arm.com>
26556 * config/aarch64/aarch64.c
26557 (aarch64_simd_clone_compute_vecsize_and_simdlen): Remove unused var.
26559 2021-03-16 Segher Boessenkool <segher@kernel.crashing.org>
26562 * config/rs6000/predicates.md (branch_comparison_operator): Allow
26563 ordered and unordered for CCFPmode, if flag_finite_math_only.
26565 2021-03-16 Jakub Jelinek <jakub@redhat.com>
26568 * config/i386/i386-expand.c (ix86_split_lea_for_addr): Emit a MULT
26569 rather than ASHIFT.
26570 * config/i386/i386.md (mult by 1248 into ashift): New splitter.
26572 2021-03-16 Martin Liska <mliska@suse.cz>
26575 * optc-save-gen.awk: Add flag_ipa_ra to exceptions for
26576 cl_optimization_compare function.
26578 2021-03-16 Ilya Leoshkevich <iii@linux.ibm.com>
26580 * config/s390/s390.c (f_constraint_p): Treat "fv" constraints
26583 2021-03-16 Jakub Jelinek <jakub@redhat.com>
26586 * config/i386/i386.h (struct machine_function): Add
26587 has_explicit_vzeroupper bitfield.
26588 * config/i386/i386-expand.c (ix86_expand_builtin): Set
26589 cfun->machine->has_explicit_vzeroupper when expanding
26590 IX86_BUILTIN_VZEROUPPER.
26591 * config/i386/i386-features.c (rest_of_handle_insert_vzeroupper):
26592 Do the mode switching only when TARGET_VZEROUPPER, expensive
26593 optimizations turned on and not optimizing for size.
26594 (pass_insert_vzeroupper::gate): Enable even when
26595 cfun->machine->has_explicit_vzeroupper is set.
26597 2021-03-16 Jakub Jelinek <jakub@redhat.com>
26600 * config/aarch64/aarch64.c
26601 (aarch64_simd_clone_compute_vecsize_and_simdlen): If not a function
26602 definition, walk TYPE_ARG_TYPES list if non-NULL for argument types
26603 instead of DECL_ARGUMENTS. Ignore types for uniform arguments.
26605 2021-03-15 Richard Biener <rguenther@suse.de>
26607 PR tree-optimization/98834
26608 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle missing
26609 subsetting by truncating the access size.
26611 2021-03-15 Jan Hubicka <hubicka@ucw.cz>
26613 * config/i386/i386-options.c (processor_cost_table): Add znver3_cost.
26614 * config/i386/x86-tune-costs.h (znver3_cost): New gobal variable; copy
26617 2021-03-15 Martin Liska <mliska@suse.cz>
26619 * spellcheck.c: Add missing comma in initialization.
26621 2021-03-14 Uroš Bizjak <ubizjak@gmail.com>
26623 * config/i386/sse.md (*vec_extract<mode>): Merge alternative 0 with
26624 alternative 2 and alternative 1 with alternative 3 using
26625 YW register constraint.
26626 (*vec_extract<PEXTR_MODE12:mode>_zext): Merge alternatives
26627 using YW register constraint.
26628 (*vec_extractv16qi_zext): Ditto.
26629 (*vec_extractv4si): Merge alternatives 4 and 5
26630 using Yw register constraint.
26631 (*ssse3_palignr<mode>_perm): Use Yw instead of v for alternative 3.
26633 2021-03-13 Martin Sebor <msebor@redhat.com>
26635 PR tree-optimization/99489
26636 * builtins.c (gimple_call_alloc_size): Fail gracefully when argument
26637 is not a call statement.
26639 2021-03-13 Jakub Jelinek <jakub@redhat.com>
26641 PR tree-optimization/99544
26642 * match.pd (X + (X << C) -> X * (1 + (1 << C))): Don't simplify
26643 if for vector types multiplication can't be done in type's mode.
26645 2021-03-12 Eric Botcazou <ebotcazou@adacore.com>
26648 * config/sparc/constraints.md (w): Rename to...
26649 (W): ... this and ditch previous implementation.
26650 * config/sparc/sparc.md (*movdi_insn_sp64): Replace W with m.
26651 (*movdf_insn_sp64): Likewise.
26652 (*mov<VM64:mode>_insn_sp64): Likewise.
26653 * config/sparc/sync.md (*atomic_compare_and_swap<mode>_1): Replace
26655 (atomic_compare_and_swap_leon3_1): Likewise.
26656 (*atomic_compare_and_swapdi_v8plus): Likewise.
26657 * config/sparc/sparc.c (memory_ok_for_ldd): Remove useless test on
26658 architecture and add missing address validity check during LRA.
26660 2021-03-12 Tobias Burnus <tobias@codesourcery.com>
26663 * gimplify.c (omp_add_variable): Handle NULL_TREE as size
26664 occuring for assumed-size arrays in use_device_{ptr,addr}.
26666 2021-03-12 Jakub Jelinek <jakub@redhat.com>
26669 * config/i386/constraints.md (YW): New internal constraint.
26670 * config/i386/sse.md (v_Yw): Add V4TI, V2TI, V1TI and TI cases.
26671 (*<sse2_avx2>_<insn><mode>3<mask_name>,
26672 *<sse2_avx2>_uavg<mode>3<mask_name>, *abs<mode>2,
26673 *<s>mul<mode>3_highpart<mask_name>): Use <v_Yw> instead of v in
26675 (<sse2_avx2>_psadbw): Use YW instead of v in constraints.
26676 (*avx2_pmaddwd, *sse2_pmaddwd, *<code>v8hi3, *<code>v16qi3,
26677 avx2_pmaddubsw256, ssse3_pmaddubsw128): Merge last two alternatives
26678 into one, use Yw instead of former x,v.
26679 (ashr<mode>3, <insn><mode>3): Use <v_Yw> instead of x in constraints of
26680 the last alternative.
26681 (<sse2_avx2>_packsswb<mask_name>, <sse2_avx2>_packssdw<mask_name>,
26682 <sse2_avx2>_packuswb<mask_name>, <sse4_1_avx2>_packusdw<mask_name>,
26683 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, <ssse3_avx2>_palignr<mode>,
26684 <ssse3_avx2>_pshufb<mode>3<mask_name>): Merge last two alternatives
26685 into one, use <v_Yw> instead of former x,v.
26686 (avx2_interleave_highv32qi<mask_name>,
26687 vec_interleave_highv16qi<mask_name>): Use Yw instead of v in
26688 constraints. Add && <mask_avx512bw_condition> to condition.
26689 (avx2_interleave_lowv32qi<mask_name>,
26690 vec_interleave_lowv16qi<mask_name>,
26691 avx2_interleave_highv16hi<mask_name>,
26692 vec_interleave_highv8hi<mask_name>,
26693 avx2_interleave_lowv16hi<mask_name>, vec_interleave_lowv8hi<mask_name>,
26694 avx2_pshuflw_1<mask_name>, sse2_pshuflw_1<mask_name>,
26695 avx2_pshufhw_1<mask_name>, sse2_pshufhw_1<mask_name>,
26696 avx2_<code>v16qiv16hi2<mask_name>, sse4_1_<code>v8qiv8hi2<mask_name>,
26697 *sse4_1_<code>v8qiv8hi2<mask_name>_1, <sse2_avx2>_<insn><mode>3): Use
26698 Yw instead of v in constraints.
26699 * config/i386/mmx.md (Yv_Yw): New define_mode_attr.
26700 (*mmx_<insn><mode>3, mmx_ashr<mode>3, mmx_<insn><mode>3): Use <Yv_Yw>
26701 instead of Yv in constraints.
26702 (*mmx_<insn><mode>3, *mmx_mulv4hi3, *mmx_smulv4hi3_highpart,
26703 *mmx_umulv4hi3_highpart, *mmx_pmaddwd, *mmx_<code>v4hi3,
26704 *mmx_<code>v8qi3, mmx_pack<s_trunsuffix>swb, mmx_packssdw,
26705 mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
26706 *mmx_uavgv8qi3, *mmx_uavgv4hi3, mmx_psadbw): Use Yw instead of Yv in
26708 (*mmx_pinsrw, *mmx_pinsrb, *mmx_pextrw, *mmx_pextrw_zext, *mmx_pextrb,
26709 *mmx_pextrb_zext): Use YW instead of Yv in constraints.
26710 (*mmx_eq<mode>3, mmx_gt<mode>3): Use x instead of Yv in constraints.
26711 (mmx_andnot<mode>3, *mmx_<code><mode>3): Split last alternative into
26712 two, one with just x, another isa avx512vl with v.
26714 2021-03-12 Martin Liska <mliska@suse.cz>
26716 * doc/invoke.texi: Add missing param documentation.
26718 2021-03-11 David Malcolm <dmalcolm@redhat.com>
26721 * Makefile.in (ANALYZER_OBJS): Add analyzer/feasible-graph.o and
26722 analyzer/trimmed-graph.o.
26723 * doc/analyzer.texi (Analyzer Paths): Rewrite description of
26724 feasibility checking to reflect new implementation.
26725 * doc/invoke.texi (-fdump-analyzer-feasibility): Document new
26727 * shortest-paths.h (shortest_paths::get_shortest_distance): New.
26729 2021-03-11 David Malcolm <dmalcolm@redhat.com>
26731 * digraph.cc (selftest::test_shortest_paths): Update
26732 shortest_paths init for new param. Add test of
26733 SPS_TO_GIVEN_TARGET.
26734 * shortest-paths.h (enum shortest_path_sense): New.
26735 (shortest_paths::shortest_paths): Add "sense" param.
26736 Update for renamings. Generalize to use "sense" param.
26737 (shortest_paths::get_shortest_path): Rename param.
26738 (shortest_paths::m_sense): New field.
26739 (shortest_paths::m_prev): Rename...
26740 (shortest_paths::m_best_edge): ...to this.
26741 (shortest_paths::get_shortest_path): Update for renamings.
26742 Conditionalize flipping of path on sense of traversal.
26744 2021-03-11 David Malcolm <dmalcolm@redhat.com>
26746 * digraph.cc (selftest::test_shortest_paths): Add test coverage
26747 for paths from B and C.
26748 * shortest-paths.h (shortest_paths::shortest_paths): Handle
26749 unreachable nodes, rather than asserting.
26751 2021-03-11 David Edelsohn <dje.gcc@gmail.com>
26754 * config/rs6000/rs6000.c (rs6000_xcoff_file_start): Don't create
26755 xcoff_tbss_section_name.
26756 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .lcomm.
26757 * xcoffout.c (xcoff_tbss_section_name): Delete.
26758 * xcoffout.h (xcoff_tbss_section_name): Delete.
26760 2021-03-11 Richard Biener <rguenther@suse.de>
26762 PR tree-optimization/99523
26763 * tree-cfg.c (dump_function_to_file): Dump SSA names
26764 w/o identifier to the decls section as well, not only those
26765 without a VAR_DECL.
26767 2021-03-11 Jakub Jelinek <jakub@redhat.com>
26770 * ipa-icf-gimple.c (func_checker::compare_gimple_call): For internal
26771 function calls with lhs fail if the lhs don't have compatible types.
26773 2021-03-11 Hans-Peter Nilsson <hp@axis.com>
26775 * config/cris/cris.h (HARD_FRAME_POINTER_REGNUM): Define.
26776 Change FRAME_POINTER_REGNUM to correspond to a new faked
26777 register faked_fp, part of GENNONACR_REGS like faked_ap.
26778 (CRIS_FAKED_REGS_CONTENTS): New helper macro.
26779 (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS, CALL_USED_REGISTERS):
26780 (REG_ALLOC_ORDER, REG_CLASS_CONTENTS, REGNO_OK_FOR_BASE_P)
26781 (ELIMINABLE_REGS, REGISTER_NAMES): Adjust accordingly.
26782 * config/cris/cris.md (CRIS_FP_REGNUM): Renumber to new faked
26784 (CRIS_REAL_FP_REGNUM): New constant.
26785 * config/cris/cris.c (cris_reg_saved_in_regsave_area): Check
26786 for HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM.
26787 (cris_initial_elimination_offset): Handle elimination changes
26788 to HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM
26789 and add one from FRAME_POINTER_REGNUM to
26790 HARD_FRAME_POINTER_REGNUM.
26791 (cris_expand_prologue, cris_expand_epilogue): Emit code for
26792 hard_frame_pointer_rtx instead of frame_pointer_rtx.
26794 2021-03-10 David Edelsohn <dje.gcc@gmail.com>
26797 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Add check for DCmode.
26798 * config/rs6000/rs6000.c (rs6000_special_round_type_align): Same.
26800 2021-03-10 Vladimir N. Makarov <vmakarov@redhat.com>
26803 * lra-constraints.c (process_address_1): Don't check unknown
26804 constraint, use X for empty constraint.
26806 2021-03-10 Alex Coplan <alex.coplan@arm.com>
26808 * config/aarch64/aarch64.c (aarch64_vfp_is_call_or_return_candidate):
26809 Fix typo in comment describing "is_ha" argument.
26811 2021-03-10 John David Anglin <danglin@gcc.gnu.org>
26813 * doc/sourcebuild.texi: Document LRA target selector.
26815 2021-03-10 David Malcolm <dmalcolm@redhat.com>
26817 * doc/ux.texi: Add subsection contrasting interactive versus
26818 batch usage of GCC.
26820 2021-03-10 Joel Hutton <joel.hutton@arm.com>
26823 * tree-vect-stmts.c (vectorizable_store): Fix scatter store mask
26825 (vectorizable_load): Fix gather load mask check condition.
26827 2021-03-10 Richard Biener <rguenther@suse.de>
26829 PR tree-optimization/99510
26830 * tree.c (check_aligned_type): Check that the candidate
26831 has TYPE_USER_ALIGN set instead of matching with the
26834 2021-03-10 Eric Botcazou <ebotcazou@adacore.com>
26836 * config/sparc/sparc.c (sparc_regmode_natural_size): Return 4 for
26837 float and vector integer modes only if the mode is not larger.
26839 2021-03-10 Hans-Peter Nilsson <hp@axis.com>
26841 * config/cris/cris.h (DWARF_FRAME_REGISTERS): Define.
26843 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
26845 * ira.c (ira_setup_alts, ira_get_dup_out_num): Process digital
26847 * ira-lives.c (single_reg_class): Ditto.
26849 2021-03-09 Sebastian Huber <sebastian.huber@embedded-brains.de>
26851 * config.gcc (aarch64-*-rtems*): Include general rtems.h after
26852 the architecture-specific rtems.h.
26853 (aarch64-*-rtems*): Likewise.
26854 (arm*-*-rtems*): Likewise.
26855 (epiphany-*-rtems*): Likewise.
26856 (riscv*-*-rtems*): Likewise.
26858 2021-03-09 Jakub Jelinek <jakub@redhat.com>
26860 PR tree-optimization/99305
26861 * tree-ssa-phiopt.c (conditional_replacement): Test integer_pow2p
26862 before integer_all_onesp instead of vice versa.
26864 2021-03-09 Richard Earnshaw <rearnsha@arm.com>
26866 * common/config/arm/arm-common.c (arm_config_default): Change type
26867 of 'i' to unsigned.
26869 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
26872 * lra-constraints.c (process_address_1): Process constraint 'g'
26873 separately and digital constraints containing more one digit.
26875 2021-03-09 Nick Clifton <nickc@redhat.com>
26877 * config/rx/rx.h (DBX_DEBUGGING_INFO): Define.
26878 (DWARF"_DEBUGGING_INFO): Define.
26880 2021-03-09 Eric Botcazou <ebotcazou@adacore.com>
26883 * calls.c (initialize_argument_information): When the argument
26884 is passed by reference, do not make a copy in a thunk only if
26885 the argument is already in memory. Remove redundant test for
26886 the case of callee copy.
26888 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
26891 * lra-constraints.c (process_address_1): Process 0..9 constraints
26892 in process_address_1.
26894 2021-03-09 Andreas Krebbel <krebbel@linux.ibm.com>
26896 * config/s390/s390.c (struct s390_processor processor_table):
26897 Binutils name string must not be empty.
26899 2021-03-09 Claudiu Zissulescu <claziss@synopsys.com>
26901 * config/arc/arc.c (arc_attr_type): Remove function.
26903 2021-03-09 Martin Liska <mliska@suse.cz>
26906 * config/i386/i386-options.c (ix86_option_override_internal):
26907 Set isa_flags for OPTS argument and not for the global
26910 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
26912 * config/rs6000/predicates.md (ds_form_mem_operand): Check
26915 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
26918 * config/rs6000/predicates.md (ds_form_mem_operand) New
26920 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10) Use
26921 ds_form_mem_operand in ld/lwa patterns.
26922 * config/rs6000/fusion.md: Regenerate file.
26924 2021-03-08 Martin Sebor <msebor@redhat.com>
26926 PR middle-end/98266
26927 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): New function.
26928 (array_bounds_checker::check_array_bounds): Call it.
26930 2021-03-08 Martin Sebor <msebor@redhat.com>
26932 PR middle-end/97631
26933 * tree-ssa-strlen.c (maybe_warn_overflow): Test rawmem.
26934 (handle_builtin_stxncpy_strncat): Rename locals. Determine
26935 destination size from allocation calls. Issue a more appropriate
26937 (handle_builtin_memcpy): Pass true as rawmem to maybe_warn_overflow.
26938 (handle_builtin_memset): Same.
26940 2021-03-08 Peter Bergner <bergner@linux.ibm.com>
26943 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_permute): Add an assert
26944 to ensure we do not have an Altivec style address.
26945 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): Disable if passed
26946 an Altivec style address.
26947 (*vsx_le_perm_store_<mode>): Likewise.
26948 (splitters after *vsx_le_perm_store_<mode>): Likewise.
26949 (vsx_load_<mode>): Disable special expander if passed an Altivec
26951 (vsx_store_<mode>): Likewise.
26953 2021-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26956 * config/aarch64/predicates.md (aarch64_simd_shift_imm_vec_qi): Define.
26957 (aarch64_simd_shift_imm_vec_hi): Likewise.
26958 (aarch64_simd_shift_imm_vec_si): Likewise.
26959 (aarch64_simd_shift_imm_vec_di): Likewise.
26960 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le): Use
26961 predicate from above.
26962 (aarch64_shrn<mode>_insn_be): Likewise.
26963 (aarch64_rshrn<mode>_insn_le): Likewise.
26964 (aarch64_rshrn<mode>_insn_be): Likewise.
26965 (aarch64_shrn2<mode>_insn_le): Likewise.
26966 (aarch64_shrn2<mode>_insn_be): Likewise.
26967 (aarch64_rshrn2<mode>_insn_le): Likewise.
26968 (aarch64_rshrn2<mode>_insn_be): Likewise.
26970 2021-03-08 Vladimir N. Makarov <vmakarov@redhat.com>
26973 * lra-constraints.c (skip_contraint_modifiers): New function.
26974 (process_address_1): Use it before lookup_constraint call.
26976 2021-03-08 Martin Liska <mliska@suse.cz>
26979 * config/i386/i386-options.c (ix86_option_override_internal):
26980 Enable UINTR and HRESET for -march that supports it.
26982 2021-03-08 Ilya Leoshkevich <iii@linux.ibm.com>
26984 * config/s390/s390.c (f_constraint_p): New function.
26985 (s390_md_asm_adjust): Implement TARGET_MD_ASM_ADJUST.
26986 (TARGET_MD_ASM_ADJUST): Likewise.
26988 2021-03-08 Tobias Burnus <tobias@codesourcery.com>
26991 * tree-nested.c (convert_local_reference_stmt): Avoid calling
26992 lookup_field_for_decl for Fortran module (= namespace context).
26994 2021-03-08 Andreas Krebbel <krebbel@linux.ibm.com>
26996 * config/s390/s390.c (s390_expand_vec_compare): Implement <0
26997 comparison with arithmetic right shift.
26998 (s390_expand_vcond): No need for a force_reg anymore.
26999 s390_vec_compare will do it.
27000 * config/s390/vector.md ("vec_cmp<mode><tointvec>"): Accept also
27001 immediate operands.
27003 2021-03-07 Jakub Jelinek <jakub@redhat.com>
27006 * config/i386/constraints.md (Yw): Use SSE_REGS if TARGET_SSE
27007 but TARGET_AVX512BW or TARGET_AVX512VL is not set. Adjust description
27009 * config/i386/sse.md (v_Yw): New define_mode_attr.
27010 (*<insn><mode>3, *mul<mode>3<mask_name>, *avx2_<code><mode>3,
27011 *sse4_1_<code><mode>3<mask_name>): Use <v_Yw> instead of v
27013 * config/i386/mmx.md (mmx_pshufw_1, *vec_dupv4hi): Use Yw instead of
27014 xYw in constraints.
27016 2021-03-06 Julian Brown <julian@codesourcery.com>
27018 * tree-pretty-print.c (dump_generic_node): Emit non-generic
27019 address space info for aggregates.
27021 2021-03-06 Hans-Peter Nilsson <hp@axis.com>
27023 * config/cris/cris.h (MAX_FIXED_MODE_SIZE): Don't define.
27025 2021-03-05 Jakub Jelinek <jakub@redhat.com>
27027 PR middle-end/99322
27028 * tree-cfg.c (bb_to_omp_idx): New variable.
27029 (execute_build_cfg): Release the bb_to_omp_idx vector after
27030 cleanup_tree_cfg returns.
27031 (handle_abnormal_edges): Remove bb_to_omp_idx argument, adjust
27032 for bb_to_omp_idx being a vec<int> instead of pointer to array
27034 (make_edges): Remove bb_to_omp_idx local variable, don't pass
27035 it to handle_abnormal_edges, adjust for bb_to_omp_idx being a
27036 vec<int> instead of pointer to array of ints and don't free/release
27038 (remove_bb): When removing a bb and placing forced label somewhere
27039 else, ensure it is put into the same OpenMP region during cfg
27040 pass if possible or to entry successor as fallback. Unregister
27041 bb from bb_to_omp_idx.
27043 2021-03-05 Vladimir N. Makarov <vmakarov@redhat.com>
27046 * lra-constraints.c (process_address_1): Skip decomposing address
27047 for asm insn operand with unknown constraint.
27049 2021-03-05 Martin Jambor <mjambor@suse.cz>
27052 * cgraph.c (cgraph_edge::set_call_stmt): Do not update all
27053 corresponding speculative edges if we are about to resolve
27054 sepculation. Make edge direct (and so resolve speculations) before
27055 removing it from call_site_hash.
27056 (cgraph_edge::make_direct): Relax the initial assert to allow calling
27057 the function on speculative direct edges.
27059 2021-03-05 Eric Botcazou <ebotcazou@adacore.com>
27061 PR rtl-optimization/99376
27062 * rtlanal.c (nonzero_bits1) <arithmetic operators>: If the number
27063 of low-order zero bits is too large, set the result to 0 directly.
27065 2021-03-04 Jakub Jelinek <jakub@redhat.com>
27067 PR middle-end/93235
27068 * expmed.c (store_bit_field_using_insv): Return false of xop0 is a
27069 SUBREG and a SUBREG to op_mode can't be created.
27071 2021-03-04 Alex Coplan <alex.coplan@arm.com>
27074 * config/aarch64/aarch64-sve-builtins.cc
27075 (function_resolver::require_vector_type): Handle error_mark_node.
27077 2021-03-04 Ilya Leoshkevich <iii@linux.ibm.com>
27079 * cfgexpand.c (expand_asm_loc): Pass new parameter.
27080 (expand_asm_stmt): Likewise.
27081 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Add new
27083 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
27084 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
27085 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
27086 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
27087 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
27088 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
27089 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
27090 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
27091 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
27092 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
27093 * doc/tm.texi (md_asm_adjust): Likewise.
27094 * target.def (md_asm_adjust): Likewise.
27096 2021-03-04 Richard Biener <rguenther@suse.de>
27098 PR middle-end/97855
27099 * tree-pretty-print.c: Poison pp_printf.
27100 (dump_decl_name): Avoid use of pp_printf.
27101 (dump_block_node): Likewise.
27102 (dump_generic_node): Likewise.
27104 2021-03-04 Martin Sebor <msebor@redhat.com>
27106 PR middle-end/96963
27107 PR middle-end/94655
27108 * builtins.c (handle_array_ref): New helper.
27109 (handle_mem_ref): New helper.
27110 (compute_objsize_r): Factor out ARRAY_REF and MEM_REF handling
27111 into new helper functions. Correct a workaround for vectorized
27114 2021-03-03 Pat Haugen <pthaugen@linux.ibm.com>
27116 * config/rs6000/dfp.md (extendddtd2, trunctddd2, *cmp<mode>_internal1,
27117 floatditd2, ftrunc<mode>2, fix<mode>di2, dfp_ddedpd_<mode>,
27118 dfp_denbcd_<mode>, dfp_dxex_<mode>, dfp_diex_<mode>,
27119 *dfp_sgnfcnc_<mode>, dfp_dscli_<mode>, dfp_dscri_<mode>): Update size
27120 attribute for Power10.
27121 * config/rs6000/mma.md (*movoo): Likewise.
27122 * config/rs6000/rs6000.md (define_attr "size"): Add 256.
27123 (define_mode_attr bits): Add DD/TD modes.
27124 * config/rs6000/sync.md (load_quadpti, store_quadpti, load_lockedpti,
27125 store_conditionalpti): Update size attribute for Power10.
27127 2021-03-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
27130 * config/sparc/t-sparc (tree-ssanames.o-warn): Don't error for
27131 -Wuninitialized, -Wmaybe-uninitialized.
27132 (wide-int.o-warn): Likewise.
27134 2021-03-03 Richard Earnshaw <rearnsha@arm.com>
27136 * common/config/arm/arm-common.c: Include configargs.h.
27137 (arm_config_default): New function.
27138 (arm_target_mode): Renamed from arm_target_thumb_only. Handle
27139 processors that do not support Thumb. Take into account the
27140 --with-mode configuration setting for selecting the default.
27141 * config/arm/arm.h (OPTION_DEFAULT_SPECS): Remove entry for 'mode'.
27142 (TARGET_MODE_SPEC_FUNCTIONS): Update for function name change.
27144 2021-03-03 Martin Liska <mliska@suse.cz>
27146 PR gcov-profile/97461
27147 * gcov-io.h (GCOV_PREALLOCATED_KVP): Remove.
27149 2021-03-03 Eric Botcazou <ebotcazou@adacore.com>
27152 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
27153 point back the hard frame pointer to its default location when the
27154 frame is larger than SEH_MAX_FRAME_SIZE.
27156 2021-03-03 Jakub Jelinek <jakub@redhat.com>
27159 * config/i386/predicates.md (logic_operator): New define_predicate.
27160 * config/i386/i386.md (mov + mem using comm arith peephole2):
27161 Punt if operands[1] is EXT_REX_SSE_REGNO_P, AVX512BW is not enabled
27162 and the inner mode is [QH]Imode.
27164 2021-03-03 Jakub Jelinek <jakub@redhat.com>
27167 * dwarf2out.c (dw_loc_list_struct): Add end_entry member.
27168 (new_loc_list): Clear end_entry.
27169 (output_loc_list): Only use DW_LLE_startx_length for -gsplit-dwarf
27170 if HAVE_AS_LEB128, otherwise use DW_LLE_startx_endx. Fix comment
27172 (index_location_lists): For dwarf_version >= 5 without HAVE_AS_LEB128,
27173 initialize also end_entry.
27175 2021-03-03 Jakub Jelinek <jakub@redhat.com>
27178 * cfgrtl.c (fixup_partitions): When changing some bbs from hot to cold
27179 partitions, if in non-layout mode after reorder_blocks also move
27180 affected blocks to ensure a single partition transition.
27182 2021-03-03 Jason Merrill <jason@redhat.com>
27185 * cgraphunit.c (process_function_and_variable_attributes): Don't
27186 warn about flatten on an alias if the target also has it.
27187 * cgraph.h (symtab_node::get_alias_target_tree): New.
27189 2021-03-02 David Edelsohn <dje.gcc@gmail.com>
27191 * config/rs6000/rs6000.md (tls_get_tpointer_internal): Prepend
27192 period to symbol name.
27193 (tls_get_addr_internal<mode>): Same.
27195 2021-03-02 David Malcolm <dmalcolm@redhat.com>
27198 * diagnostic-show-locus.c
27199 (selftest::test_one_liner_many_fixits_2): Fix accidental usage of
27202 2021-03-02 Martin Sebor <msebor@redhat.com>
27204 PR middle-end/99276
27205 * builtins.c (warn_for_access): Remove stray warning text.
27207 2021-03-02 Martin Sebor <msebor@redhat.com>
27209 PR middle-end/99295
27210 * doc/extend.texi (attribute malloc): Reword and clarify nonaliasing
27213 2021-03-02 Jakub Jelinek <jakub@redhat.com>
27216 * dwarf2out.c (output_macinfo_op): Use DW_MACRO_*_str* even with
27217 -gdwarf-5 -gstrict-dwarf. For -gsplit-dwarf -gdwarf-5 use
27218 DW_MACRO_*_strx instead of DW_MACRO_*_strp. Handle
27219 DW_MACRO_define_strx and DW_MACRO_undef_strx.
27220 (save_macinfo_strings): Use DW_MACRO_*_str* even with
27221 -gdwarf-5 -gstrict-dwarf. Handle DW_MACRO_define_strx and
27222 DW_MACRO_undef_strx.
27224 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
27226 * config/s390/s390-builtin-types.def (BT_FN_V4SF_V8HI_UINT): New
27228 (BT_FN_V8HI_V8HI_UINT): Likewise.
27229 (BT_FN_V8HI_V4SF_V4SF_UINT): Likewise.
27230 * config/s390/s390-builtins.def (B_NNPA): New macro definition.
27231 (s390_vclfnhs, s390_vclfnls, s390_vcrnfs, s390_vcfn, s390_vcnf):
27232 New builtin definitions.
27233 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Bump
27234 vector extension version.
27235 * config/s390/s390.c (s390_expand_builtin): Check if builtins are
27236 available with current -march level.
27237 * config/s390/s390.md (UNSPEC_NNPA_VCLFNHS_V8HI)
27238 (UNSPEC_NNPA_VCLFNLS_V8HI, UNSPEC_NNPA_VCRNFS_V8HI)
27239 (UNSPEC_NNPA_VCFN_V8HI, UNSPEC_NNPA_VCNF_V8HI): New constants.
27240 * config/s390/vecintrin.h (vec_extend_to_fp32_hi): New macro.
27241 (vec_extend_to_fp32_lo): Likewise.
27242 (vec_round_from_fp32): Likewise.
27243 (vec_convert_to_fp16): Likewise.
27244 (vec_convert_from_fp16): Likewise.
27245 * config/s390/vx-builtins.md (vclfnhs_v8hi): New insn pattern.
27246 (vclfnls_v8hi): Likewise.
27247 (vcrnfs_v8hi): Likewise.
27248 (vcfn_v8hi): Likewise.
27249 (vcnf_v8hi): Likewise.
27251 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
27253 * common/config/s390/s390-common.c (processor_flags_table): New entry.
27254 * config.gcc: Enable arch14 for --with-arch and --with-tune.
27255 * config/s390/driver-native.c (s390_host_detect_local_cpu): Pick
27256 arch14 for unknown CPU models.
27257 * config/s390/s390-opts.h (enum processor_type): Add PROCESSOR_ARCH14.
27258 * config/s390/s390.c (s390_issue_rate): Add case for PROCESSOR_ARCH14.
27259 (s390_get_sched_attrmask): Likewise.
27260 (s390_get_unit_mask): Likewise.
27261 * config/s390/s390.h (enum processor_flags): Add PF_NNPA and PF_ARCH14.
27262 (TARGET_CPU_ARCH14, TARGET_CPU_ARCH14_P, TARGET_CPU_NNPA)
27263 (TARGET_CPU_NNPA_P, TARGET_ARCH14, TARGET_ARCH14_P, TARGET_NNPA)
27264 (TARGET_NNPA_P): New macro definitions.
27265 * config/s390/s390.md ("cpu_facility", "enabled"): Add arch14 and nnpa.
27266 * config/s390/s390.opt: Add PROCESSOR_ARCH14.
27268 2021-03-02 Jakub Jelinek <jakub@redhat.com>
27270 PR middle-end/95757
27271 * tree-vrp.c (register_edge_assert_for): Remove superfluous ()s around
27272 condition. Call register_edge_assert_for_1 for == 0, != 0, == 1 and
27273 != 1 comparisons if name is lhs of a comparison.
27275 2021-03-01 Iain Sandoe <iain@sandoe.co.uk>
27279 * config/darwin-protos.h (darwin_should_restore_cfa_state): New.
27280 * config/darwin.c (darwin_should_restore_cfa_state): New.
27281 * config/darwin.h (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New.
27282 * doc/tm.texi: Regenerated.
27283 * doc/tm.texi.in: Document TARGET_ASM_SHOULD_RESTORE_CFA_STATE.
27284 * dwarf2cfi.c (connect_traces): If the target requests, restore
27285 the CFA expression after a DW_CFA_restore.
27286 * target.def (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New hook.
27288 2021-03-01 Martin Liska <mliska@suse.cz>
27291 * optc-save-gen.awk: Add 4 more exceptions.
27293 2021-03-01 Nathan Sidwell <nathan@acm.org>
27296 * tree.h (TYPE_ALIGN_RAW): New accessor.
27297 (TYPE_ALIGN): Use it.
27299 2021-03-01 Jan Hubicka <jh@suse.cz>
27302 * ipa-fnsummary.c (compute_fn_summary): Fix sanity check.
27304 2021-03-01 Eric Botcazou <ebotcazou@adacore.com>
27307 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
27308 point the hard frame pointer to the SSE register save area instead
27309 of the general register save area. Perform only minimal adjustment
27310 for small frames if it is initially not correctly aligned.
27311 (ix86_expand_prologue): Remove early saves for a SEH target.
27312 * config/i386/winnt.c (struct seh_frame_state): Document constraint.
27314 2021-02-28 Jakub Jelinek <jakub@redhat.com>
27317 * ipa.c (symbol_table::remove_unreachable_nodes): Fix a comment
27318 typo - referneced -> referenced.
27319 * tree.c (component_ref_size): Fix comment typo -
27320 refernce -> reference.
27321 * tree-ssa-alias.c (access_path_may_continue_p): Fix comment typo -
27322 traling -> trailing.
27323 (aliasing_component_refs_p): Fix comment typos -
27324 refernce -> reference and refernece -> reference and
27325 traling -> trailing.
27326 (nonoverlapping_refs_since_match_p): Fix comment typo -
27327 referneces -> references.
27328 * doc/invoke.texi (--param modref-max-bases): Fix a typo -
27329 referneces -> references.
27331 2021-02-27 Iain Sandoe <iain@sandoe.co.uk>
27333 * config/host-darwin.c (darwin_gt_pch_use_address): Modify
27334 diagnostic message to avoid use of a contraction and format
27337 2021-02-27 Jakub Jelinek <jakub@redhat.com>
27340 * gcse.c (gcse_or_cprop_is_too_expensive): Use %wu instead of
27341 HOST_WIDE_INT_PRINT_UNSIGNED in warning format string.
27342 * ipa-devirt.c (ipa_odr_read_section): Use %wd instead of
27343 HOST_WIDE_INT_PRINT_DEC in inform format string. Fix comment
27346 2021-02-26 Richard Biener <rguenther@suse.de>
27348 PR middle-end/99281
27349 * expr.c (store_field): For calls with return-slot optimization
27350 and addressable return type expand the store directly.
27352 2021-02-26 Richard Biener <rguenther@suse.de>
27355 * builtins.c (warn_string_no_nul): Fix diagnostic formatting.
27357 2021-02-26 Peter Bergner <bergner@linux.ibm.com>
27360 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Replace assert
27363 2021-02-26 Aaron Sawdey <acsawdey@linux.ibm.com>
27365 * config.gcc: Add rs6000-pcrel-opt.o.
27366 * config/rs6000/rs6000-pcrel-opt.c: New file.
27367 * config/rs6000/pcrel-opt.md: New file.
27368 * config/rs6000/predicates.md: Add d_form_memory predicate.
27369 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_PCREL_OPT.
27370 * config/rs6000/rs6000-passes.def: Add pass_pcrel_opt.
27371 * config/rs6000/rs6000-protos.h: Add reg_to_non_prefixed(),
27372 pcrel_opt_valid_mem_p(), output_pcrel_opt_reloc(),
27373 and make_pass_pcrel_opt().
27374 * config/rs6000/rs6000.c (reg_to_non_prefixed): Make global.
27375 (rs6000_option_override_internal): Add pcrel-opt.
27376 (rs6000_delegitimize_address): Support pcrel-opt.
27377 (rs6000_opt_masks): Add pcrel-opt.
27378 (pcrel_opt_valid_mem_p): New function.
27379 (reg_to_non_prefixed): Make global.
27380 (rs6000_asm_output_opcode): Reset prepend_p_to_next_insn.
27381 (output_pcrel_opt_reloc): New function.
27382 * config/rs6000/rs6000.md (loads_extern_addr): New attr.
27383 (pcrel_extern_addr): Set loads_extern_addr.
27384 Add include for pcrel-opt.md.
27385 * config/rs6000/rs6000.opt: Add -mpcrel-opt.
27386 * config/rs6000/t-rs6000: Add rules for pcrel-opt.c and
27389 2021-02-26 YunQiang Su <yunqiang.su@cipunited.com>
27392 * config/mips/mips.c (mips_expand_ext_as_unaligned_load):
27393 If TARGET_64BIT and dest is SUBREG, we check the width, if it
27394 equal to SImode, we use SImode operation, just like what we are
27397 2021-02-26 Marek Polacek <polacek@redhat.com>
27399 * builtins.c (warn_for_access): Fix typos.
27401 2021-02-25 Iain Sandoe <iain@sandoe.co.uk>
27403 * config/aarch64/aarch64.md (<optab>_rol<mode>3): Add a '#'
27404 mark in front of the immediate quantity.
27405 (<optab>_rolsi3_uxtw): Likewise.
27407 2021-02-25 Richard Earnshaw <rearnsha@arm.com>
27410 * config/arm/thumb2.md (nonsecure_call_reg_thumb2_fpcxt): New pattern.
27411 (nonsecure_call_value_reg_thumb2_fpcxt): Likewise.
27412 (nonsecure_call_reg_thumb2): Restrict to using r4 for the callee
27413 address and disable when the FPCXT is not available.
27414 (nonsecure_call_value_reg_thumb2): Likewise.
27416 2021-02-25 Nathan Sidwell <nathan@acm.org>
27419 * doc/invoke.texi (flang-info-module-cmi): Renamed option.
27421 2021-02-25 Tamar Christina <tamar.christina@arm.com>
27423 * tree-vect-slp.c (optimize_load_redistribution_1): Abort on NULL nodes.
27425 2021-02-25 Richard Biener <rguenther@suse.de>
27427 PR tree-optimization/99253
27428 * tree-vect-loop.c (check_reduction_path): First compute
27429 code, then verify out-of-loop uses.
27431 2021-02-25 Jakub Jelinek <jakub@redhat.com>
27434 * match.pd ((T)(A) + CST -> (T)(A + CST)): Add :s to convert.
27436 2021-02-25 Jakub Jelinek <jakub@redhat.com>
27438 PR tree-optimization/80635
27439 * tree-vrp.c (vrp_simplify_cond_using_ranges): Also handle
27440 VIEW_CONVERT_EXPR if modes are the same, innerop is integral and
27441 has mode precision.
27443 2021-02-25 Richard Biener <rguenther@suse.de>
27445 * tree-vect-slp.c (optimize_load_redistribution_1): Delay
27446 load_map population.
27447 (vect_match_slp_patterns_2): Revert part of last change.
27448 (vect_analyze_slp): Do not interleave optimize_load_redistribution
27449 with pattern detection but do it afterwards. Dump the
27450 whole SLP graph after pattern recognition and load
27451 redistribution optimization finished.
27453 2021-02-24 Jakub Jelinek <jakub@redhat.com>
27456 * omp-low.c (struct omp_context): Add teams_nested_p and
27457 nonteams_nested_p members.
27458 (scan_omp_target): Diagnose teams nested inside of target with other
27459 directives strictly nested inside of the same target.
27460 (check_omp_nesting_restrictions): Set ctx->teams_nested_p or
27461 ctx->nonteams_nested_p as needed.
27463 2021-02-24 Vladimir N. Makarov <vmakarov@redhat.com>
27465 PR inline-asm/99123
27466 * lra-constraints.c (uses_hard_regs_p): Don't use decompose_mem_address.
27468 2021-02-24 Hans-Peter Nilsson <hp@axis.com>
27470 * config/cris/cris.c (cris_expand_prologue): Set
27471 current_function_static_stack_size, if flag_stack_usage_info.
27473 2021-02-24 Pat Haugen <pthaugen@linux.ibm.com>
27475 * config/rs6000/rs6000.c (next_insn_prefixed_p): Rename.
27476 (rs6000_final_prescan_insn): Adjust.
27477 (rs6000_asm_output_opcode): Likewise.
27479 2021-02-24 Martin Sebor <msebor@redhat.com>
27481 PR middle-end/97172
27482 * attribs.c (attr_access::free_lang_data): Clear attribute arg spec
27483 from function arguments.
27485 2021-02-24 Tamar Christina <tamar.christina@arm.com>
27487 PR tree-optimization/99220
27488 * tree-vect-slp.c (optimize_load_redistribution_1): Remove
27489 node from cache when it's about to be deleted.
27491 2021-02-24 Jakub Jelinek <jakub@redhat.com>
27493 PR tree-optimization/99225
27494 * fold-const.c (fold_binary_loc) <case NE_EXPR>: In (x & (1 << y)) != 0
27495 to ((x >> y) & 1) != 0 simplifications use build_one_cst instead of
27496 build_int_cst (..., 1). Formatting fixes.
27498 2021-02-24 Tamar Christina <tamar.christina@arm.com>
27500 PR tree-optimization/99149
27501 * tree-vect-slp-patterns.c (vect_detect_pair_op): Don't recreate the
27503 (vect_slp_reset_pattern): Remove.
27504 (complex_fma_pattern::matches): Remove call to vect_slp_reset_pattern.
27505 (complex_mul_pattern::build, complex_fma_pattern::build,
27506 complex_fms_pattern::build): Fix ref counts.
27507 * tree-vect-slp.c (vect_free_slp_tree): Undo SLP only pattern relevancy
27508 when node is being deleted.
27509 (vect_match_slp_patterns_2): Correct result of cache hit on patterns.
27510 (vect_schedule_slp): Invalidate SLP_TREE_REPRESENTATIVE of removed
27512 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize value.
27514 2021-02-24 Matthias Klose <doko@ubuntu.com>
27517 2020-12-07 Matthias Klose <doko@ubuntu.com>
27519 * genextract.c (print_header): Undefine ENABLE_RTL_CHECKING
27520 and ENABLE_RTL_FLAG_CHECKING.
27522 2021-02-24 Richard Biener <rguenther@suse.de>
27525 * builtins.c (fold_builtin_next_arg): Avoid NULL arg.
27527 2021-02-23 Peter Bergner <bergner@linux.ibm.com>
27529 * config/rs6000/mma.md (mma_assemble_pair): Rename from this...
27530 (vsx_assemble_pair): ...to this.
27531 (*mma_assemble_pair): Rename from this...
27532 (*vsx_assemble_pair): ...to this.
27533 (mma_disassemble_pair): Rename from this...
27534 (vsx_disassemble_pair): ...to this.
27535 (*mma_disassemble_pair): Rename from this...
27536 (*vsx_disassemble_pair): ...to this.
27537 * config/rs6000/rs6000-builtin.def (BU_MMA_V2, BU_MMA_V3,
27538 BU_COMPAT): New macros.
27539 (mma_assemble_pair): Rename from this...
27540 (vsx_assemble_pair): ...to this.
27541 (mma_disassemble_pair): Rename from this...
27542 (vsx_disassemble_pair): ...to this.
27543 (mma_assemble_pair): New compatibility built-in.
27544 (mma_disassemble_pair): Likewise.
27545 * config/rs6000/rs6000-call.c (struct builtin_compatibility): New.
27546 (RS6000_BUILTIN_COMPAT): Define.
27547 (bdesc_compat): New.
27548 (mma_expand_builtin): Use VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
27549 (rs6000_gimple_fold_mma_builtin): Use MMA_BUILTIN_DISASSEMBLE_PAIR
27550 and VSX_BUILTIN_ASSEMBLE_PAIR.
27551 (rs6000_init_builtins): Register compatibility built-ins.
27552 (mma_init_builtins): Use VSX_BUILTIN_ASSEMBLE_PAIR,
27553 VSX_BUILTIN_ASSEMBLE_PAIR_INTERNAL, VSX_BUILTIN_DISASSEMBLE_PAIR and
27554 VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
27555 * doc/extend.texi (__builtin_mma_assemble_pair): Rename from this...
27556 (__builtin_vsx_assemble_pair): ...to this.
27557 (__builtin_mma_disassemble_pair): Rename from this...
27558 (__builtin_vsx_disassemble_pair): ...to this.
27560 2021-02-23 Martin Liska <mliska@suse.cz>
27563 * ipa-icf.c (sem_variable::merge): Do not merge 2 variables
27564 with different alignment. That leads to an invalid red zone
27565 size allocated in runtime.
27567 2021-02-23 Jakub Jelinek <jakub@redhat.com>
27569 PR tree-optimization/99204
27570 * fold-const.c (fold_read_from_constant_string): Check that
27571 tree_fits_uhwi_p (index) rather than just that index is INTEGER_CST.
27573 2021-02-23 Segher Boessenkool <segher@kernel.crashing.org>
27574 Kewen Lin <linkw@gcc.gnu.org>
27576 * config/rs6000/rs6000.md (*rotl<mode>3_insert_3): Renamed to...
27577 (rotl<mode>3_insert_3): ...this.
27578 (plus_ior_xor): New code_iterator.
27579 (define_split for GPR rl*imi): New splitter.
27580 * config/rs6000/vsx.md (vsx_init_v4si): Use gen_rotldi3_insert_3
27581 for integer merging.
27583 2021-02-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27585 * config/aarch64/aarch64-tuning-flags.def (cse_sve_vl_constants):
27587 * config/aarch64/aarch64.md (add<mode>3): Force CONST_POLY_INT immediates
27588 into a register when the above is enabled.
27589 * config/aarch64/aarch64.c (neoversev1_tunings):
27590 AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
27591 (aarch64_rtx_costs): Use AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
27593 2021-02-22 Hans-Peter Nilsson <hp@axis.com>
27595 * config/cris/cris.c (cris_print_operand) <'T'>: Change
27596 valid operand from is now an addi mult-value to shift-value.
27597 * config/cris/cris.md (*addi): Change expression of scaled
27598 operand from mult to ashift.
27599 * config/cris/cris.md (*addi_reload): New insn_and_split.
27601 2021-02-22 John David Anglin <danglin@gcc.gnu.org>
27604 * config/pa/pa.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Define as
27605 hook_bool_const_tree_hwi_hwi_const_tree_true.
27606 (pa_asm_output_mi_thunk): Add support for nonzero vcall_offset.
27608 2021-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
27610 PR rtl-optimization/98791
27611 * ira-conflicts.c (process_regs_for_copy): Don't create allocno copies
27612 for unordered modes.
27614 2021-02-22 Martin Liska <mliska@suse.cz>
27616 * tree-inline.c (inline_forbidden_p): Set
27617 inline_forbidden_reason.
27619 2021-02-22 Richard Biener <rguenther@suse.de>
27621 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Dump
27624 2021-02-22 Richard Biener <rguenther@suse.de>
27626 PR tree-optimization/99165
27627 * gimple-ssa-store-merging.c (pass_store_merging::process_store):
27628 Accumulate changed to ret.
27630 2021-02-21 Uros Bizjak <ubizjak@gmail.com>
27633 2020-12-09 Uroš Bizjak <ubizjak@gmail.com>
27635 * config/i386/i386.h (REG_ALLOC_ORDER): Remove
27637 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
27640 * config/s390/vector.md (trunctf<DFP_ALL:mode>2_vr): New
27642 (trunctf<DFP_ALL:mode>2): Likewise.
27643 (trunctdtf2_vr): Likewise.
27644 (trunctdtf2): Likewise.
27645 (extend<DFP_ALL:mode>tf2_vr): Likewise.
27646 (extend<DFP_ALL:mode>tf2): Likewise.
27647 (extendtftd2_vr): Likewise.
27648 (extendtftd2): Likewise.
27650 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
27652 * config/s390/vector.md (*fprx2_to_tf): Rename to fprx2_to_tf,
27653 add memory alternative.
27654 (tf_to_fprx2): New pattern.
27656 2021-02-19 Martin Sebor <msebor@redhat.com>
27659 * attribs.c (init_attr_rdwr_indices): Guard vblist use.
27660 (attr_access::free_lang_data): Remove a spurious test.
27662 2021-02-19 Nathan Sidwell <nathan@acm.org>
27664 * doc/invoke.texi (flang-info-module-read): Document.
27666 2021-02-19 Martin Liska <mliska@suse.cz>
27668 PR translation/99167
27669 * params.opt: Fix typo.
27671 2021-02-19 Richard Biener <rguenther@suse.de>
27673 PR middle-end/99122
27674 * tree-inline.c (inline_forbidden_p): Do not inline functions
27675 with VLA arguments or return value.
27677 2021-02-19 Jakub Jelinek <jakub@redhat.com>
27680 * config/arm/arm.md (*stack_protect_combined_set_insn,
27681 *stack_protect_combined_test_insn): If force_const_mem result
27682 is not valid general operand, force its address into the destination
27685 2021-02-19 Jakub Jelinek <jakub@redhat.com>
27688 * tree-cfg.c (gimple_merge_blocks): If bb a starts with eh landing
27689 pad or non-local label, put FORCED_LABELs from bb b after that label
27690 rather than before it.
27692 2021-02-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
27695 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3): Use
27696 expand_vector_broadcast' to emit the vec_duplicate operand.
27698 2021-02-18 Vladimir N. Makarov <vmakarov@redhat.com>
27700 PR rtl-optimization/96264
27701 * lra-remat.c (reg_overlap_for_remat_p): Check also output insn
27704 2021-02-18 H.J. Lu <hjl.tools@gmail.com>
27707 * varasm.c (get_section): Replace SUPPORTS_SHF_GNU_RETAIN with
27708 looking up the retain attribute.
27709 (resolve_unique_section): Likewise.
27710 (get_variable_section): Likewise.
27711 (switch_to_section): Likewise. Warn when a symbol without the
27712 retain attribute and a symbol with the retain attribute are
27713 placed in the section with the same name, instead of the used
27715 * doc/extend.texi: Document the "retain" attribute.
27717 2021-02-18 Nathan Sidwell <nathan@acm.org>
27720 * doc/invoke.texi (flang-info-include-translate): Document header
27723 2021-02-18 Richard Biener <rguenther@suse.de>
27725 PR middle-end/99122
27726 * ipa-fnsummary.c (analyze_function_body): Set
27727 CIF_FUNCTION_NOT_INLINABLE for VLA parameter calls.
27728 * tree-inline.c (insert_init_debug_bind): Pass NULL for
27729 error_mark_node values.
27730 (force_value_to_type): Do not build V_C_Es for WITH_SIZE_EXPR
27732 (setup_one_parameter): Delay force_value_to_type until when
27735 2021-02-18 Hans-Peter Nilsson <hp@axis.com>
27737 PR tree-optimization/99142
27738 * match.pd (clz cmp 0): Gate replacement on single_use of clz result.
27740 2021-02-18 Jakub Jelinek <jakub@redhat.com>
27742 * wide-int-bitmask.h (wide_int_bitmask::wide_int_bitmask (),
27743 wide_int_bitmask::wide_int_bitmask (uint64_t),
27744 wide_int_bitmask::wide_int_bitmask (uint64_t, uint64_t),
27745 wide_int_bitmask::operator ~ () const,
27746 wide_int_bitmask::operator | (wide_int_bitmask) const,
27747 wide_int_bitmask::operator & (wide_int_bitmask) const): Use constexpr
27749 * config/i386/i386.h (PTA_3DNOW, PTA_3DNOW_A, PTA_64BIT, PTA_ABM,
27750 PTA_AES, PTA_AVX, PTA_BMI, PTA_CX16, PTA_F16C, PTA_FMA, PTA_FMA4,
27751 PTA_FSGSBASE, PTA_LWP, PTA_LZCNT, PTA_MMX, PTA_MOVBE, PTA_NO_SAHF,
27752 PTA_PCLMUL, PTA_POPCNT, PTA_PREFETCH_SSE, PTA_RDRND, PTA_SSE, PTA_SSE2,
27753 PTA_SSE3, PTA_SSE4_1, PTA_SSE4_2, PTA_SSE4A, PTA_SSSE3, PTA_TBM,
27754 PTA_XOP, PTA_AVX2, PTA_BMI2, PTA_RTM, PTA_HLE, PTA_PRFCHW, PTA_RDSEED,
27755 PTA_ADX, PTA_FXSR, PTA_XSAVE, PTA_XSAVEOPT, PTA_AVX512F, PTA_AVX512ER,
27756 PTA_AVX512PF, PTA_AVX512CD, PTA_NO_TUNE, PTA_SHA, PTA_PREFETCHWT1,
27757 PTA_CLFLUSHOPT, PTA_XSAVEC, PTA_XSAVES, PTA_AVX512DQ, PTA_AVX512BW,
27758 PTA_AVX512VL, PTA_AVX512IFMA, PTA_AVX512VBMI, PTA_CLWB, PTA_MWAITX,
27759 PTA_CLZERO, PTA_NO_80387, PTA_PKU, PTA_AVX5124VNNIW, PTA_AVX5124FMAPS,
27760 PTA_AVX512VPOPCNTDQ, PTA_SGX, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES,
27761 PTA_AVX512VBMI2, PTA_VPCLMULQDQ, PTA_AVX512BITALG, PTA_RDPID,
27762 PTA_PCONFIG, PTA_WBNOINVD, PTA_AVX512VP2INTERSECT, PTA_PTWRITE,
27763 PTA_AVX512BF16, PTA_WAITPKG, PTA_MOVDIRI, PTA_MOVDIR64B, PTA_ENQCMD,
27764 PTA_CLDEMOTE, PTA_SERIALIZE, PTA_TSXLDTRK, PTA_AMX_TILE, PTA_AMX_INT8,
27765 PTA_AMX_BF16, PTA_UINTR, PTA_HRESET, PTA_KL, PTA_WIDEKL, PTA_AVXVNNI,
27766 PTA_X86_64_BASELINE, PTA_X86_64_V2, PTA_X86_64_V3, PTA_X86_64_V4,
27767 PTA_CORE2, PTA_NEHALEM, PTA_WESTMERE, PTA_SANDYBRIDGE, PTA_IVYBRIDGE,
27768 PTA_HASWELL, PTA_BROADWELL, PTA_SKYLAKE, PTA_SKYLAKE_AVX512,
27769 PTA_CASCADELAKE, PTA_COOPERLAKE, PTA_CANNONLAKE, PTA_ICELAKE_CLIENT,
27770 PTA_ICELAKE_SERVER, PTA_TIGERLAKE, PTA_SAPPHIRERAPIDS, PTA_ALDERLAKE,
27771 PTA_KNL, PTA_BONNELL, PTA_SILVERMONT, PTA_GOLDMONT, PTA_GOLDMONT_PLUS,
27772 PTA_TREMONT, PTA_KNM): Use constexpr instead of const.
27774 2021-02-18 Jakub Jelinek <jakub@redhat.com>
27776 PR middle-end/99109
27777 * gimple-array-bounds.cc (build_zero_elt_array_type): Rename to ...
27778 (build_printable_array_type): ... this. Add nelts argument. For
27779 overaligned eltype, use TYPE_MAIN_VARIANT (eltype) instead. If
27780 nelts, call build_array_type_nelts.
27781 (array_bounds_checker::check_mem_ref): Use build_printable_array_type
27782 instead of build_zero_elt_array_type and build_array_type_nelts.
27784 2021-02-18 Jakub Jelinek <jakub@redhat.com>
27787 * config/i386/i386.c (distance_non_agu_define): Don't call
27788 extract_insn_cached here.
27789 (ix86_lea_outperforms): Save and restore recog_data around call
27790 to distance_non_agu_define and distance_agu_use.
27791 (ix86_ok_to_clobber_flags): Remove.
27792 (ix86_avoid_lea_for_add): Don't call ix86_ok_to_clobber_flags.
27793 (ix86_avoid_lea_for_addr): Likewise. Adjust function comment.
27794 * config/i386/i386.md (*lea<mode>): Change from define_insn_and_split
27795 into define_insn. Move the splitting to define_peephole2 and
27796 check there using peep2_regno_dead_p if FLAGS_REG is dead.
27798 2021-02-17 Julian Brown <julian@codesourcery.com>
27800 * gimplify.c (gimplify_scan_omp_clauses): Handle ATTACH_DETACH
27803 2021-02-17 Xi Ruoyao <xry111@mengyan1223.wang>
27806 * config/mips/mips.c (mips_symbol_insns): Do not use
27807 MSA_SUPPORTED_MODE_P if mode is MAX_MACHINE_MODE.
27809 2021-02-16 Vladimir N. Makarov <vmakarov@redhat.com>
27811 PR inline-asm/98096
27812 * stmt.c (resolve_operand_name_1): Take inout operands into account
27813 for access to labels by names.
27814 * doc/extend.texi: Describe counting operands for accessing labels.
27816 2021-02-16 Richard Biener <rguenther@suse.de>
27818 PR tree-optimization/38474
27819 * tree-ssa-structalias.c (variable_info::address_taken): New.
27820 (new_var_info): Initialize address_taken.
27821 (process_constraint): Set address_taken.
27822 (solve_constraints): Use the new address_taken flag rather
27823 than is_reg_var for sorting variables.
27824 (dump_constraint): Dump the variable number if the name
27827 2021-02-16 Jakub Jelinek <jakub@redhat.com>
27830 * tree-vect-stmts.c (vectorizable_simd_clone_call): For num_calls != 1
27831 multiply by 4096 and for inbranch by 8192.
27832 * config/i386/i386.c (ix86_simd_clone_usable): For TARGET_AVX512F,
27833 return 3, 2 or 1 for mangle letters 'b', 'c' or 'd'.
27835 2021-02-15 Maya Rashish <coypu@sdf.org>
27837 * config/aarch64/aarch64.c (aarch64_init_builtins):
27838 Call SUBTARGET_INIT_BUILTINS.
27840 2021-02-15 Peter Bergner <bergner@linux.ibm.com>
27842 PR rtl-optimization/98872
27843 * init-regs.c (initialize_uninitialized_regs): Skip initialization
27844 if CONST0_RTX is NULL.
27846 2021-02-15 Richard Sandiford <richard.sandiford@arm.com>
27848 PR rtl-optimization/98863
27849 * rtl-ssa/functions.h (function_info::bb_live_out_info): Delete.
27850 (function_info::build_info): Turn into a declaration, moving the
27851 definition to internals.h.
27852 (function_info::bb_walker): Declare.
27853 (function_info::create_reg_use): Likewise.
27854 (function_info::calculate_potential_phi_regs): Take a build_info
27856 (function_info::place_phis, function_info::create_ebbs): Declare.
27857 (function_info::calculate_ebb_live_in_for_debug): Likewise.
27858 (function_info::populate_backedge_phis): Delete.
27859 (function_info::start_block, function_info::end_block): Declare.
27860 (function_info::populate_phi_inputs): Delete.
27861 (function_info::m_potential_phi_regs): Move information to build_info.
27862 * rtl-ssa/internals.h: New file.
27863 (function_info::bb_phi_info): New class.
27864 (function_info::build_info): Moved from functions.h.
27865 Add a constructor and destructor.
27866 (function_info::build_info::ebb_use): Delete.
27867 (function_info::build_info::ebb_def): Likewise.
27868 (function_info::build_info::bb_live_out): Likewise.
27869 (function_info::build_info::tmp_ebb_live_in_for_debug): New variable.
27870 (function_info::build_info::potential_phi_regs): Likewise.
27871 (function_info::build_info::potential_phi_regs_for_debug): Likewise.
27872 (function_info::build_info::ebb_def_regs): Likewise.
27873 (function_info::build_info::bb_phis): Likewise.
27874 (function_info::build_info::bb_mem_live_out): Likewise.
27875 (function_info::build_info::bb_to_rpo): Likewise.
27876 (function_info::build_info::def_stack): Likewise.
27877 (function_info::build_info::old_def_stack_limit): Likewise.
27878 * rtl-ssa/internals.inl (function_info::build_info::record_reg_def):
27879 Remove the regno argument. Push the previous definition onto the
27880 definition stack where necessary.
27881 * rtl-ssa/accesses.cc: Include internals.h.
27882 * rtl-ssa/changes.cc: Likewise.
27883 * rtl-ssa/blocks.cc: Likewise.
27884 (function_info::build_info::build_info): Define.
27885 (function_info::build_info::~build_info): Likewise.
27886 (function_info::bb_walker): New class.
27887 (function_info::bb_walker::bb_walker): Define.
27888 (function_info::add_live_out_use): Convert a logarithmic-complexity
27889 test into a linear one. Allow the same definition to be passed
27891 (function_info::calculate_potential_phi_regs): Moved from
27892 functions.cc. Take a build_info parameter and store the
27893 information there instead.
27894 (function_info::place_phis): New function.
27895 (function_info::add_entry_block_defs): Update call to record_reg_def.
27896 (function_info::calculate_ebb_live_in_for_debug): New function.
27897 (function_info::add_phi_nodes): Use bb_phis to decide which
27898 registers need phi nodes and initialize ebb_def_regs accordingly.
27899 Do not add degenerate phis here.
27900 (function_info::add_artificial_accesses): Use create_reg_use.
27901 Assert that all definitions are listed in the DF LR sets.
27902 Update call to record_reg_def.
27903 (function_info::record_block_live_out): Record live-out register
27904 values in the phis of successor blocks. Use the live-out set
27905 when processing the last block in an EBB, instead of always
27906 using the live-in sets of successor blocks. AND the live sets
27907 with the set of registers that have been defined in the EBB,
27908 rather than with all potential phi registers. Cope correctly
27909 with branches back to the start of the current EBB.
27910 (function_info::start_block): New function.
27911 (function_info::end_block): Likewise.
27912 (function_info::populate_phi_inputs): Likewise.
27913 (function_info::create_ebbs): Likewise.
27914 (function_info::process_all_blocks): Rewrite into a multi-phase
27916 * rtl-ssa/functions.cc: Include internals.h.
27917 (function_info::calculate_potential_phi_regs): Move to blocks.cc.
27918 (function_info::init_function_data): Remove caller.
27919 * rtl-ssa/insns.cc: Include internals.h
27920 (function_info::create_reg_use): New function. Lazily any
27921 degenerate phis needed by the linear RPO view.
27922 (function_info::record_use): Use create_reg_use. When processing
27923 debug uses, use potential_phi_regs and test it before checking
27924 whether the register is live on entry to the current EBB. Lazily
27925 calculate ebb_live_in_for_debug.
27926 (function_info::record_call_clobbers): Update call to record_reg_def.
27927 (function_info::record_def): Likewise.
27929 2021-02-15 Martin Liska <mliska@suse.cz>
27931 * toplev.c (init_asm_output): Free output of
27932 gen_command_line_string function.
27933 (process_options): Likewise.
27935 2021-02-15 Martin Liska <mliska@suse.cz>
27937 * params.opt: Add 2 missing Param keywords.
27939 2021-02-15 Eric Botcazou <ebotcazou@adacore.com>
27941 * df-core.c (df_worklist_dataflow_doublequeue): Use proper cast.
27943 2021-02-15 Jakub Jelinek <jakub@redhat.com>
27945 PR tree-optimization/99079
27946 * match.pd (A % (pow2pcst << N) -> A & ((pow2pcst << N) - 1)): Remove
27947 useless tree_nop_conversion_p (type, TREE_TYPE (@3)) check. Instead
27948 require both type and TREE_TYPE (@1) to be integral types and either
27949 type having smaller or equal precision, or TREE_TYPE (@1) being
27950 unsigned type, or type being signed type. If TREE_TYPE (@1)
27951 doesn't have wrapping overflow, perform the subtraction of one in
27954 2021-02-14 Jan Hubicka <hubicka@ucw.cz>
27955 Richard Biener <rguether@suse.de>
27958 * ipa-reference.c (ipa_init): Only conditinally initialize
27959 reference_vars_to_consider.
27960 (propagate): Conditionally deninitialize reference_vars_to_consider.
27961 (ipa_reference_write_optimization_summary): Sanity check that
27962 reference_vars_to_consider is not allocated.
27964 2021-02-13 Levy Hsu <admin@levyhsu.com>
27967 * config/riscv/riscv-shorten-memrefs.c (pass_shorten_memrefs): Add
27968 extend parameter to get_si_mem_base_reg declaration.
27969 (get_si_mem_base_reg): Add extend parameter. Set it.
27970 (analyze): Pass extend arg to get_si_mem_base_reg.
27971 (transform): Likewise. Use it when rewriting mems.
27972 * config/riscv/riscv.c (riscv_legitimize_move): Check for subword
27973 loads and emit sign/zero extending load followed by subreg move.
27975 2021-02-13 Jim Wilson <jimw@sifive.com>
27978 * config/riscv/riscv.c (riscv_compressed_lw_address_p): Drop early
27979 exit when !reload_completed. Only perform check for compressed reg
27980 if reload_completed.
27981 (riscv_rtx_costs): In MEM case, when optimizing for size and
27982 shorten memrefs, if not compressible, then increase cost.
27984 2021-02-13 Jakub Jelinek <jakub@redhat.com>
27986 PR rtl-optimization/98439
27987 * recog.c (pass_split_before_regstack::gate): Enable even when
27988 pass_split_before_sched2 is enabled if -fselective-scheduling2 is
27991 2021-02-13 Jakub Jelinek <jakub@redhat.com>
27994 * config/i386/mmx.md (*mmx_pshufd_1): Add a combine splitter for
27995 swap of V2SImode elements in memory into DImode memory rotate by 32.
27997 2021-02-12 Martin Sebor <msebor@redhat.com>
27999 * tree-pretty-print.c (print_generic_expr_to_str): Update comment.
28001 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
28003 * rtl-ssa/accesses.cc (function_info::make_use_available): Use
28004 m_temp_obstack rather than m_obstack to allocate the temporary use.
28006 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
28008 * df-problems.c (df_lr_bb_local_compute): Treat partial definitions
28009 as read-modify operations.
28011 2021-02-12 Richard Biener <rguenther@suse.de>
28013 PR middle-end/38474
28014 * ipa-fnsummary.c (unmodified_parm_1): Only walk when
28015 fbi->aa_walk_budget is bigger than zero. Update
28016 fbi->aa_walk_budget.
28017 (param_change_prob): Likewise.
28018 * ipa-prop.c (detect_type_change_from_memory_writes):
28019 Properly account walk_aliased_vdefs.
28020 (parm_preserved_before_stmt_p): Canonicalize updates.
28021 (parm_ref_data_preserved_p): Likewise.
28022 (parm_ref_data_pass_through_p): Likewise.
28023 (determine_known_aggregate_parts): Account own alias queries.
28025 2021-02-12 Martin Liska <mliska@suse.cz>
28027 * opts-common.c (decode_cmdline_option): Release werror_arg.
28028 * opts.c (gen_producer_string): Release output of
28029 gen_command_line_string.
28031 2021-02-12 Richard Biener <rguenther@suse.de>
28033 PR tree-optimization/38474
28034 * params.opt (-param=max-store-chains-to-track=): New param.
28035 (-param=max-stores-to-track=): Likewise.
28036 * doc/invoke.texi (max-store-chains-to-track): Document.
28037 (max-stores-to-track): Likewise.
28038 * gimple-ssa-store-merging.c (pass_store_merging::m_n_chains):
28040 (pass_store_merging::m_n_stores): Likewise.
28041 (pass_store_merging::terminate_and_process_chain): Update
28042 m_n_stores and m_n_chains.
28043 (pass_store_merging::process_store): Likewise. Terminate
28044 oldest chains if the number of stores or chains get too large.
28045 (imm_store_chain_info::terminate_and_process_chain): Dump
28048 2021-02-11 Eric Botcazou <ebotcazou@adacore.com>
28050 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
28051 the cold section, emit a nop before the directive if the previous
28052 active instruction can throw.
28054 2021-02-11 Peter Bergner <bergner@linux.ibm.com>
28057 * config/rs6000/predicates.md (mma_assemble_input_operand): Restrict
28058 memory addresses that are legal for quad word accesses.
28060 2021-02-11 Andrea Corallo <andrea.corallo@arm.com>
28063 * config/arm/thumb2.md (*doloop_end_internal): Generate
28064 alternative sequence to handle long range branches.
28066 2021-02-11 Joel Hutton <joel.hutton@arm.com>
28068 PR tree-optimization/98772
28069 * optabs-tree.c (supportable_half_widening_operation): New function
28070 to check for supportable V8QI->V8HI widening patterns.
28071 * optabs-tree.h (supportable_half_widening_operation): New function.
28072 * tree-vect-stmts.c (vect_create_half_widening_stmts): New function
28073 to create promotion stmts for V8QI->V8HI widening patterns.
28074 (vectorizable_conversion): Add case for V8QI->V8HI.
28076 2021-02-11 Richard Biener <rguenther@suse.de>
28078 * sparseset.h (SPARSESET_ELT_BITS): Remove.
28079 (SPARSESET_ELT_TYPE): Use unsigned int.
28080 * fwprop.c: Do not include sparseset.h.
28082 2021-02-10 Jakub Jelinek <jakub@redhat.com>
28085 * varasm.c (declare_weak): For -fsyntax-only, allow even
28086 TREE_ASM_WRITTEN function decls.
28088 2021-02-10 Jakub Jelinek <jakub@redhat.com>
28091 * config/i386/sse.md (fix<fixunssuffix>_truncv2sfv2di2,
28092 <insn>v8qiv8hi2, <insn>v8qiv8si2, <insn>v4qiv4si2, <insn>v4hiv4si2,
28093 <insn>v8qiv8di2, <insn>v4qiv4di2, <insn>v2qiv2di2, <insn>v4hiv4di2,
28094 <insn>v2hiv2di2, <insn>v2siv2di2): Force operands[1] into REG before
28095 calling simplify_gen_subreg on it.
28097 2021-02-10 Martin Liska <mliska@suse.cz>
28099 * config/nvptx/nvptx.c (nvptx_option_override): Use
28100 flag_patchable_function_entry instead of the removed
28101 function_entry_patch_area_size.
28103 2021-02-10 Martin Liska <mliska@suse.cz>
28105 PR tree-optimization/99002
28106 PR tree-optimization/99026
28107 * gimple-if-to-switch.cc (if_chain::is_beneficial): Fix memory
28108 leak when adjacent cases are merged.
28109 * tree-switch-conversion.c (switch_decision_tree::analyze_switch_statement): Use
28111 (make_pass_lower_switch): Remove trailing whitespace.
28112 * tree-switch-conversion.h (release_clusters): New.
28114 2021-02-10 Richard Biener <rguenther@suse.de>
28116 PR rtl-optimization/99054
28117 * cfgrtl.c (rtl-optimization/99054): Return an auto_vec.
28118 (fixup_partitions): Adjust.
28119 (rtl_verify_edges): Likewise.
28121 2021-02-10 Jakub Jelinek <jakub@redhat.com>
28123 PR middle-end/99007
28124 * gimplify.c (gimplify_scan_omp_clauses): For MEM_REF on reductions,
28125 temporarily disable gimplify_ctxp->into_ssa around gimplify_expr
28128 2021-02-10 Richard Biener <rguenther@suse.de>
28131 * ipa-pure-const.c (propagate_malloc): Use an auto_vec<>
28134 2021-02-10 Richard Biener <rguenther@suse.de>
28136 PR tree-optimization/99024
28137 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Only
28138 clear loop->aux if it is associated with the destroyed loop_vinfo.
28140 2021-02-10 Martin Liska <mliska@suse.cz>
28142 PR tree-optimization/99002
28143 * gimple-if-to-switch.cc (find_conditions): Fix memory leak
28146 2021-02-10 Martin Liska <mliska@suse.cz>
28149 * ipa-icf.c (sem_item::add_reference): Fix memory leak when
28150 a reference exists.
28152 2021-02-10 Jakub Jelinek <jakub@redhat.com>
28155 * dwarf2out.c (prune_unused_types_walk): Mark DW_TAG_variable DIEs
28156 at class scope for DWARF5+.
28158 2021-02-09 Eric Botcazou <ebotcazou@adacore.com>
28160 PR rtl-optimization/96015
28161 * reorg.c (skip_consecutive_labels): Minor comment tweaks.
28162 (relax_delay_slots): When deleting a jump to the next active
28163 instruction over a barrier, first delete the barrier if the
28164 jump is the only way to reach the target label.
28166 2021-02-09 Andre Vieira <andre.simoesdiasvieira@arm.com>
28168 * config/aarch64/aarch64-cost-tables.h: Add entries for vect.mul.
28169 * config/aarch64/aarch64.c (aarch64_rtx_mult_cost): Use vect.mul for
28170 vector multiplies and vect.alu for SSRA.
28171 * config/arm/aarch-common-protos.h (struct vector_cost_table): Define
28172 vect.mul cost field.
28173 * config/arm/aarch-cost-tables.h: Add entries for vect.mul.
28174 * config/arm/arm.c: Likewise.
28176 2021-02-09 Richard Biener <rguenther@suse.de>
28178 PR tree-optimization/98863
28179 * tree-ssa-sccvn.h (vn_avail::next_undo): Add.
28180 * tree-ssa-sccvn.c (last_pushed_avail): New global.
28181 (rpo_elim::eliminate_push_avail): Chain pushed avails.
28182 (unwind_state::avail_top): Add.
28183 (do_unwind): Rewrite unwinding of avail entries.
28184 (do_rpo_vn): Initialize last_pushed_avail and
28185 avail_top of the undo state.
28187 2021-02-09 Jakub Jelinek <jakub@redhat.com>
28189 PR middle-end/99004
28190 * calls.c (maybe_warn_rdwr_sizes): Change s0 and s1 type from
28191 const char * to char * and free those pointers after use.
28193 2021-02-09 Richard Biener <rguenther@suse.de>
28195 PR tree-optimization/99017
28196 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Allow
28197 zero vector cost entries.
28199 2021-02-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
28201 PR middle-end/98974
28202 * tree-vect-stmts.c (vectorizable_condition): Remove shadow vec_num
28203 parameter in vectorizable_condition.
28205 2021-02-08 Richard Biener <rguenther@suse.de>
28208 * tree.c (walk_tree_1): Walk VECTOR_CST elements.
28210 2021-02-08 Martin Liska <mliska@suse.cz>
28213 * cfgexpand.c (pass_expand::execute): Parse per-function option
28214 flag_patchable_function_entry and use it.
28215 * common.opt: Remove function_entry_patch_area_size and
28216 function_entry_patch_area_start global variables.
28217 * opts.c (parse_and_check_patch_area): New function.
28218 (common_handle_option): Use it.
28219 * opts.h (parse_and_check_patch_area): New function.
28220 * toplev.c (process_options): Parse and use
28221 function_entry_patch_area_size.
28223 2021-02-08 Martin Sebor <msebor@redhat.com>
28225 * doc/extend.texi (attribute malloc): Correct typos.
28227 2021-02-05 Nathan Sidwell <nathan@acm.org>
28230 * gcc.c (driver::maybe_run_linker): Check for input file
28231 accessibility if not linking.
28233 2021-02-05 Richard Biener <rguenther@suse.de>
28235 PR tree-optimization/98855
28236 * tree-vectorizer.h (add_stmt_cost): New overload.
28237 * tree-vect-slp.c (li_cost_vec_cmp): New.
28238 (vect_bb_slp_scalar_cost): Cost individual loop regions
28239 separately. Account for the scalar instance root stmt.
28241 2021-02-05 Tom de Vries <tdevries@suse.de>
28244 * tree-switch-conversion.c (jump_table_cluster::emit): Add loc
28246 (bit_test_cluster::emit): Reuse location_t for newly created
28248 (switch_decision_tree::try_switch_expansion): Preserve
28250 * tree-switch-conversion.h: Change function signatures.
28252 2021-02-05 Jakub Jelinek <jakub@redhat.com>
28255 * config/i386/i386-options.c (m_NONE, m_ALL): Define.
28256 * config/i386/x86-tune.def (X86_TUNE_BRANCH_PREDICTION_HINTS,
28257 X86_TUNE_PROMOTE_QI_REGS): Use m_NONE instead of 0U.
28258 (X86_TUNE_QIMODE_MATH): Use m_ALL instead of ~0U.
28260 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28262 * config/aarch64/aarch64-simd-builtins.def (get_high): Define builtin.
28263 * config/aarch64/aarch64-simd.md (aarch64_get_high<mode>): Define.
28264 * config/aarch64/arm_neon.h (__GET_HIGH): Delete.
28265 (vget_high_f16): Reimplement using new builtin.
28266 (vget_high_f32): Likewise.
28267 (vget_high_f64): Likewise.
28268 (vget_high_p8): Likewise.
28269 (vget_high_p16): Likewise.
28270 (vget_high_p64): Likewise.
28271 (vget_high_s8): Likewise.
28272 (vget_high_s16): Likewise.
28273 (vget_high_s32): Likewise.
28274 (vget_high_s64): Likewise.
28275 (vget_high_u8): Likewise.
28276 (vget_high_u16): Likewise.
28277 (vget_high_u32): Likewise.
28278 (vget_high_u64): Likewise.
28280 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28282 * config/aarch64/aarch64-simd-builtins.def (get_low): Define builtin.
28283 * config/aarch64/aarch64-simd.md (aarch64_get_low<mode>): Define.
28284 * config/aarch64/arm_neon.h (__GET_LOW): Delete.
28285 (vget_low_f16): Reimplement using new builtin.
28286 (vget_low_f32): Likewise.
28287 (vget_low_f64): Likewise.
28288 (vget_low_p8): Likewise.
28289 (vget_low_p16): Likewise.
28290 (vget_low_p64): Likewise.
28291 (vget_low_s8): Likewise.
28292 (vget_low_s16): Likewise.
28293 (vget_low_s32): Likewise.
28294 (vget_low_s64): Likewise.
28295 (vget_low_u8): Likewise.
28296 (vget_low_u16): Likewise.
28297 (vget_low_u32): Likewise.
28298 (vget_low_u64): Likewise.
28300 2021-02-05 Kito Cheng <kito.cheng@sifive.com>
28302 * gcc.c (print_multilib_info): Check all required argument is provided
28305 2021-02-05 liuhongt <hongtao.liu@intel.com>
28308 * config/i386/i386-expand.c (ix86_expand_sse_cmp): Don't
28309 generate integer mask comparison for 128/256-bits vector when
28310 op_true/op_false is NULL_RTX or CONSTM1_RTX/CONST0_RTX. Also
28311 delete redundant !maskcmp condition.
28312 (ix86_expand_int_vec_cmp): Ditto but no redundant deletion
28314 (ix86_expand_sse_movcc): Delete definition of maskcmp, add the
28315 condition directly to if (maskcmp), add extra check for
28316 cmpmode, it should be MODE_INT.
28317 (ix86_expand_fp_vec_cmp): Pass NULL to ix86_expand_sse_cmp's
28318 parameters op_true/op_false.
28319 (ix86_use_mask_cmp_p): New.
28321 2021-02-05 liuhongt <hongtao.liu@intel.com>
28324 * config/i386/x86-tune.def (X86_TUNE_AVX256_UNALIGNED_LOAD_OPTIMAL):
28325 Remove m_GENERIC from ~list.
28326 (X86_TUNE_AVX256_UNALIGNED_STORE_OPTIMAL): Ditto.
28328 2021-02-04 David Malcolm <dmalcolm@redhat.com>
28331 * diagnostic-show-locus.c (compatible_locations_p): Require
28332 locations in the same macro map to be either both from the
28333 macro definition, or both from the macro arguments.
28335 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
28337 * config/aarch64/aarch64-simd-builtins.def: Add
28338 [su]mull_hi_lane[q] builtin generator macros.
28339 * config/aarch64/aarch64-simd.md
28340 (aarch64_<su>mull_hi_lane<mode>_insn): Define.
28341 (aarch64_<su>mull_hi_lane<mode>): Define.
28342 (aarch64_<su>mull_hi_laneq<mode>_insn): Define.
28343 (aarch64_<su>mull_hi_laneq<mode>): Define.
28344 * config/aarch64/arm_neon.h (vmull_high_lane_s16): Use RTL
28345 builtin instead of inline asm.
28346 (vmull_high_lane_s32): Likewise.
28347 (vmull_high_lane_u16): Likewise.
28348 (vmull_high_lane_u32): Likewise.
28349 (vmull_high_laneq_s16): Likewise.
28350 (vmull_high_laneq_s32): Likewise.
28351 (vmull_high_laneq_u16): Likewise.
28352 (vmull_high_laneq_u32): Liekwise.
28354 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
28356 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_hi_n
28357 builtin generator macros.
28358 * config/aarch64/aarch64-simd.md
28359 (aarch64_<su>mull_hi_n<mode>_insn): Define.
28360 (aarch64_<su>mull_hi_n<mode>): Define.
28361 * config/aarch64/arm_neon.h (vmull_high_n_s16): Use RTL builtin
28362 instead of inline asm.
28363 (vmull_high_n_s32): Likewise.
28364 (vmull_high_n_u16): Likewise.
28365 (vmull_high_n_u32): Likewise.
28367 2021-02-04 Richard Biener <rguenther@suse.de>
28369 PR tree-optimization/98855
28370 * tree-vect-loop.c (vectorizable_phi): Do not cost
28371 single-argument PHIs.
28372 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Likewise.
28373 * tree-vect-stmts.c (vectorizable_bswap): Also perform
28374 costing for SLP operation.
28376 2021-02-04 Martin Liska <mliska@suse.cz>
28378 * doc/extend.texi: Mention -mprefer-vector-width in target
28381 2021-02-03 Martin Sebor <msebor@redhat.com>
28383 PR tree-optimization/98937
28384 * tree-ssa-strlen.c (strlen_dom_walker::~strlen_dom_walker): Define.
28385 Flush pointer_query cache.
28387 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
28389 * config/rs6000/genfusion.pl (gen_2logical): Add missing
28390 fixes based on patch review.
28391 * config/rs6000/fusion.md: Regenerate file.
28393 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
28395 * config/rs6000/t-rs6000: Comment out auto generation of
28398 2021-02-03 Andrew Stubbs <ams@codesourcery.com>
28400 * config/gcn/gcn-opts.h (enum processor_type): Add PROCESSOR_GFX908.
28401 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Add gfx908.
28402 (output_file_start): Add gfx908.
28403 * config/gcn/gcn.opt (gpu_type): Add gfx908.
28404 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Add march=gfx908.
28405 (MULTILIB_DIRNAMES): Add gfx908.
28406 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): New define.
28407 (main): Recognize gfx908.
28408 * config/gcn/t-omp-device: Add gfx908.
28410 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
28412 * config/aarch64/aarch64-simd-builtins.def: Add
28413 [su]mlsl_hi_lane[q] builtin macro generators.
28414 * config/aarch64/aarch64-simd.md
28415 (aarch64_<su>mlsl_hi_lane<mode>_insn): Define.
28416 (aarch64_<su>mlsl_hi_lane<mode>): Define.
28417 (aarch64_<su>mlsl_hi_laneq<mode>_insn): Define.
28418 (aarch64_<su>mlsl_hi_laneq<mode>): Define.
28419 * config/aarch64/arm_neon.h (vmlsl_high_lane_s16): Use RTL
28420 builtin instead of inline asm.
28421 (vmlsl_high_lane_s32): Likewise.
28422 (vmlsl_high_lane_u16): Likewise.
28423 (vmlsl_high_lane_u32): Likewise.
28424 (vmlsl_high_laneq_s16): Likewise.
28425 (vmlsl_high_laneq_s32): Likewise.
28426 (vmlsl_high_laneq_u16): Likewise.
28427 (vmlsl_high_laneq_u32): Likewise.
28428 (vmlal_high_laneq_u32): Likewise.
28430 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
28432 * config/aarch64/aarch64-simd-builtins.def: Add
28433 [su]mlal_hi_lane[q] builtin generator macros.
28434 * config/aarch64/aarch64-simd.md
28435 (aarch64_<su>mlal_hi_lane<mode>_insn): Define.
28436 (aarch64_<su>mlal_hi_lane<mode>): Define.
28437 (aarch64_<su>mlal_hi_laneq<mode>_insn): Define.
28438 (aarch64_<su>mlal_hi_laneq<mode>): Define.
28439 * config/aarch64/arm_neon.h (vmlal_high_lane_s16): Use RTL
28440 builtin instead of inline asm.
28441 (vmlal_high_lane_s32): Likewise.
28442 (vmlal_high_lane_u16): Likewise.
28443 (vmlal_high_lane_u32): Likewise.
28444 (vmlal_high_laneq_s16): Likewise.
28445 (vmlal_high_laneq_s32): Likewise.
28446 (vmlal_high_laneq_u16): Likewise.
28447 (vmlal_high_laneq_u32): Likewise.
28449 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
28451 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_hi_n
28452 builtin generator macros.
28453 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_hi_n<mode>_insn):
28455 (aarch64_<su>mlsl_hi_n<mode>): Define.
28456 * config/aarch64/arm_neon.h (vmlsl_high_n_s16): Use RTL builtin
28457 instead of inline asm.
28458 (vmlsl_high_n_s32): Likewise.
28459 (vmlsl_high_n_u16): Likewise.
28460 (vmlsl_high_n_u32): Likewise.
28462 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
28464 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_hi_n
28465 builtin generator macros.
28466 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_hi_n<mode>_insn):
28468 (aarch64_<su>mlal_hi_n<mode>): Define.
28469 * config/aarch64/arm_neon.h (vmlal_high_n_s16): Use RTL builtin
28470 instead of inline asm.
28471 (vmlal_high_n_s32): Likewise.
28472 (vmlal_high_n_u16): Likewise.
28473 (vmlal_high_n_u32): Likewise.
28475 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
28477 * config/aarch64/aarch64-simd-builtins.def: Add RTL builtin
28479 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal_hi<mode>):
28481 (aarch64_<su>mlal_hi<mode>_insn): This.
28482 (aarch64_<su>mlal_hi<mode>): Define.
28483 * config/aarch64/arm_neon.h (vmlal_high_s8): Use RTL builtin
28484 instead of inline asm.
28485 (vmlal_high_s16): Likewise.
28486 (vmlal_high_s32): Likewise.
28487 (vmlal_high_u8): Likewise.
28488 (vmlal_high_u16): Likewise.
28489 (vmlal_high_u32): Likewise.
28491 2021-02-03 Ilya Leoshkevich <iii@linux.ibm.com>
28493 * lra-spills.c (remove_pseudos): Call lra_update_insn_recog_data()
28494 after calling alter_subreg() on a (mem).
28496 2021-02-03 Martin Liska <mliska@suse.cz>
28499 * lto-streamer-out.c (produce_lto_section): Fill up missing
28501 * lto-streamer.h (struct lto_section): Add _padding field.
28503 2021-02-03 Richard Biener <rguenther@suse.de>
28505 * lto-streamer.c (lto_get_section_name): Free temporary
28507 * tree-loop-distribution.c
28508 (loop_distribution::merge_dep_scc_partitions): Free edge data.
28510 2021-02-03 Jakub Jelinek <jakub@redhat.com>
28512 PR middle-end/97487
28513 * ifcvt.c (noce_can_force_operand): New function.
28514 (noce_emit_move_insn): Use it.
28515 (noce_try_sign_mask): Likewise. Formatting fix.
28517 2021-02-03 Jakub Jelinek <jakub@redhat.com>
28519 PR middle-end/97971
28520 * lra-constraints.c (process_alt_operands): For inline asm, don't call
28521 fatal_insn, but instead return false.
28523 2021-02-03 Jakub Jelinek <jakub@redhat.com>
28525 PR tree-optimization/98287
28526 * config/i386/mmx.md (<insn><mode>3): For shifts don't enable expander
28529 2021-02-03 Tamar Christina <tamar.christina@arm.com>
28531 PR tree-optimization/98928
28532 * tree-vect-loop.c (vect_analyze_loop_2): Change
28533 STMT_VINFO_SLP_VECT_ONLY to STMT_VINFO_SLP_VECT_ONLY_PATTERN.
28534 * tree-vect-slp-patterns.c (complex_pattern::build): Likewise.
28535 * tree-vectorizer.h (STMT_VINFO_SLP_VECT_ONLY_PATTERN): New.
28536 (class _stmt_vec_info): Add slp_vect_pattern_only_p.
28538 2021-02-02 Richard Biener <rguenther@suse.de>
28540 * gimple-loop-interchange.cc (prepare_data_references):
28542 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
28543 * tree-ssa-loop-im.c (hoist_memory_references): Likewise.
28544 * tree-vect-stmts.c (vectorizable_condition): Do not
28546 (vectorizable_comparison): Likewise.
28548 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28550 * config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
28551 * config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
28552 * config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
28553 (vrsqrteq_u32): Likewise.
28555 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28557 * config/aarch64/aarch64-simd-builtins.def (sqxtun2): Define builtin.
28558 * config/aarch64/aarch64-simd.md (aarch64_sqxtun2<mode>_le): Define.
28559 (aarch64_sqxtun2<mode>_be): Likewise.
28560 (aarch64_sqxtun2<mode>): Likewise.
28561 * config/aarch64/arm_neon.h (vqmovun_high_s16): Reimplement using builtin.
28562 (vqmovun_high_s32): Likewise.
28563 (vqmovun_high_s64): Likewise.
28564 * config/aarch64/iterators.md (UNSPEC_SQXTUN2): Define.
28566 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28568 * config/aarch64/aarch64-simd-builtins.def (bfdot_lane, bfdot_laneq): Use
28570 (bfmlalb_lane, bfmlalt_lane, bfmlalb_lane_q, bfmlalt_lane_q): Use FP flags.
28572 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28574 * config/aarch64/aarch64-simd-builtins.def (fcmla_lane0, fcmla_lane90,
28575 fcmla_lane180, fcmla_lane270, fcmlaq_lane0, fcmlaq_lane90, fcmlaq_lane180,
28576 fcmlaq_lane270, scvtf, ucvtf, fcvtzs, fcvtzu, scvtfsi, scvtfdi, ucvtfsi,
28577 ucvtfdi, fcvtzshf, fcvtzuhf, fmlal_lane_low, fmlsl_lane_low,
28578 fmlal_laneq_low, fmlsl_laneq_low, fmlalq_lane_low, fmlslq_lane_low,
28579 fmlalq_laneq_low, fmlslq_laneq_low, fmlal_lane_high, fmlsl_lane_high,
28580 fmlal_laneq_high, fmlsl_laneq_high, fmlalq_lane_high, fmlslq_lane_high,
28581 fmlalq_laneq_high, fmlslq_laneq_high): Use FP flags.
28583 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28585 * config/aarch64/aarch64-builtins.c (FLAG_LOAD): Define.
28586 * config/aarch64/aarch64-simd-builtins.def (ld1x2, ld2, ld3, ld4, ld2r,
28587 ld3r, ld4r, ld1, ld1x3, ld1x4): Use LOAD flags.
28589 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28591 * config/aarch64/aarch64-simd-builtins.def (combine, zip1, zip2,
28592 uzp1, uzp2, trn1, trn2, simd_bsl): Use AUTO_FP flags.
28594 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28596 * config/aarch64/aarch64-simd-builtins.def (clrsb, clz, ctz, popcount,
28597 vec_smult_lane_, vec_smlal_lane_, vec_smult_laneq_, vec_smlal_laneq_,
28598 vec_umult_lane_, vec_umlal_lane_, vec_umult_laneq_, vec_umlal_laneq_,
28599 ashl, sshl, ushl, srshl, urshl, sdot_lane, udot_lane, sdot_laneq,
28600 udot_laneq, usdot_lane, usdot_laneq, sudot_lane, sudot_laneq, ashr,
28601 ashr_simd, lshr, lshr_simd, srshr_n, urshr_n, ssra_n, usra_n, srsra_n,
28602 ursra_n, sshll_n, ushll_n, sshll2_n, ushll2_n, ssri_n, usri_n, ssli_n,
28603 ssli_n, usli_n, bswap, rbit, simd_bsl, eor3q, rax1q, xarq, bcaxq): Use
28604 NONE builtin flags.
28606 2021-02-02 Jakub Jelinek <jakub@redhat.com>
28608 PR tree-optimization/98848
28609 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
28610 STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
28612 2021-02-02 Kito Cheng <kito.cheng@sifive.com>
28615 * expr.c: Check mode before calling store_expr.
28617 2021-02-02 Christophe Lyon <christophe.lyon@linaro.org>
28619 * config/arm/iterators.md (supf): Remove VORNQ_S and VORNQ_U.
28621 * config/arm/mve.md (mve_vornq_s<mode>): New entry for vorn
28622 instruction using expression ior.
28623 (mve_vornq_u<mode>): New expander.
28624 (mve_vornq_f<mode>): Use ior code instead of unspec.
28625 * config/arm/unspecs.md (VORNQ_S, VORNQ_U, VORNQ_F): Remove.
28627 2021-02-02 Alexandre Oliva <oliva@adacore.com>
28629 * tree-nested.c (convert_nonlocal_reference_op): Move
28630 current_function_decl restore after re-gimplification.
28631 (convert_local_reference_op): Likewise.
28633 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28635 * config/aarch64/aarch64-simd-builtins.def (rshrn, rshrn2):
28637 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>_insn_le):
28639 (aarch64_rshrn<mode>_insn_be): Likewise.
28640 (aarch64_rshrn<mode>): Likewise.
28641 (aarch64_rshrn2<mode>_insn_le): Likewise.
28642 (aarch64_rshrn2<mode>_insn_be): Likewise.
28643 (aarch64_rshrn2<mode>): Likewise.
28644 * config/aarch64/aarch64.md (unspec): Add UNSPEC_RSHRN.
28645 * config/aarch64/arm_neon.h (vrshrn_high_n_s16): Reimplement
28647 (vrshrn_high_n_s32): Likewise.
28648 (vrshrn_high_n_s64): Likewise.
28649 (vrshrn_high_n_u16): Likewise.
28650 (vrshrn_high_n_u32): Likewise.
28651 (vrshrn_high_n_u64): Likewise.
28652 (vrshrn_n_s16): Likewise.
28653 (vrshrn_n_s32): Likewise.
28654 (vrshrn_n_s64): Likewise.
28655 (vrshrn_n_u16): Likewise.
28656 (vrshrn_n_u32): Likewise.
28657 (vrshrn_n_u64): Likewise.
28659 2021-02-01 Sergei Trofimovich <siarheit@google.com>
28661 PR tree-optimization/98499
28662 * ipa-modref.c (analyze_ssa_name_flags): treat RVO
28663 conservatively and assume all possible side-effects.
28665 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28667 * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi,
28668 vec_unpacku_hi_): Define builtins.
28669 * config/aarch64/arm_neon.h (vmovl_high_s8): Reimplement using
28671 (vmovl_high_s16): Likewise.
28672 (vmovl_high_s32): Likewise.
28673 (vmovl_high_u8): Likewise.
28674 (vmovl_high_u16): Likewise.
28675 (vmovl_high_u32): Likewise.
28677 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28679 * config/aarch64/aarch64-simd-builtins.def (sabdl, uabdl):
28681 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl<mode>): New
28683 * config/aarch64/aarch64.md (unspec): Define UNSPEC_SABDL,
28685 * config/aarch64/arm_neon.h (vabdl_s8): Reimplemet using
28687 (vabdl_s16): Likewise.
28688 (vabdl_s32): Likewise.
28689 (vabdl_u8): Likewise.
28690 (vabdl_u16): Likewise.
28691 (vabdl_u32): Likewise.
28692 * config/aarch64/iterators.md (ABDL): New int iterator.
28693 (sur): Handle UNSPEC_SABDL, UNSPEC_UABDL.
28695 2021-02-01 Martin Sebor <msebor@redhat.com>
28697 * tree.h (BLOCK_VARS): Add comment.
28698 (BLOCK_SUBBLOCKS): Same.
28699 (BLOCK_SUPERCONTEXT): Same.
28700 (BLOCK_ABSTRACT_ORIGIN): Same.
28701 (inlined_function_outer_scope_p): Same.
28703 2021-02-01 Martin Sebor <msebor@redhat.com>
28705 PR middle-end/97172
28706 * attribs.c (attr_access::free_lang_data): Define new function.
28707 * attribs.h (attr_access::free_lang_data): Declare new function.
28709 2021-02-01 Richard Biener <rguenther@suse.de>
28711 * vec.h (auto_vec::auto_vec): Add memory stat parameters
28713 * bitmap.h (auto_bitmap::auto_bitmap): Likewise.
28715 2021-02-01 Tamar Christina <tamar.christina@arm.com>
28717 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>,
28718 aarch64_<su>mlsl<mode>, aarch64_<su>mlsl_n<mode>): Flip mult operands.
28720 2021-02-01 Richard Biener <rguenther@suse.de>
28722 PR rtl-optimization/98863
28723 * config/i386/i386-features.c (convert_scalars_to_vector):
28724 Set DF_RD_PRUNE_DEAD_DEFS.
28726 2021-01-31 Eric Botcazou <ebotcazou@adacore.com>
28728 * system.h (SIZE_MAX): Define if not already defined.
28730 2021-01-30 Aaron Sawdey <acsawdey@linux.ibm.com>
28732 * config/rs6000/genfusion.pl (gen_2logical): New function to
28733 generate patterns for logical-logical fusion.
28734 * config/rs6000/fusion.md: Regenerated patterns.
28735 * config/rs6000/rs6000-cpus.def: Add
28736 OPTION_MASK_P10_FUSION_2LOGICAL.
28737 * config/rs6000/rs6000.c (rs6000_option_override_internal):
28738 Enable logical-logical fusion for p10.
28739 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2logical.
28741 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
28743 * config/rs6000/rs6000.opt: Add periods to new AIX options.
28745 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
28747 * config/rs6000/rs6000.opt (mabi=vec-extabi): New.
28748 (mabi=vec-default): New.
28749 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
28750 __EXTABI__ for AIX Vector extended ABI.
28751 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print AIX Vector
28753 (conditional_register_usage): If AIX vec_extabi enabled, vs20-vs31
28755 * doc/invoke.texi (PowerPC mabi): Add AIX vec-extabi and vec-default.
28757 2021-01-30 Jakub Jelinek <jakub@redhat.com>
28759 * config/i386/i386-features.c (remove_partial_avx_dependency): Clear
28760 DF_DEFER_INSN_RESCAN after calling df_process_deferred_rescans.
28762 2021-01-29 Vladimir N. Makarov <vmakarov@redhat.com>
28765 * lra-constraints.c (in_class_p): Don't narrow class only for REG
28768 2021-01-29 Will Schmidt <will_schmidt@vnet.ibm.com>
28770 * config/rs6000/rs6000-call.c (rs6000_expand_binup_builtin): Add
28771 clauses for CODE_FOR_vsx_xvcvuxddp_scale and
28772 CODE_FOR_vsx_xvcvsxddp_scale to the parameter checking code.
28774 2021-01-29 Andrew MacLeod <amacleod@redhat.com>
28776 PR tree-optimization/98866
28777 * gimple-range-gori.h (gori_compute:set_range_invariant): New.
28778 * gimple-range-gori.cc (gori_map::set_range_invariant): New.
28779 (gori_map::m_maybe_invariant): Rename from all_outgoing.
28780 (gori_map::gori_map): Rename all_outgoing to m_maybe_invariant.
28781 (gori_map::is_export_p): Ditto.
28782 (gori_map::calculate_gori): Ditto.
28783 (gori_compute::set_range_invariant): New.
28784 * gimple-range.cc (gimple_ranger::range_of_stmt): Set range
28785 invariant for pointers evaluating to [1, +INF].
28787 2021-01-29 Richard Biener <rguenther@suse.de>
28789 PR rtl-optimization/98863
28790 * config/i386/i386-features.c (remove_partial_avx_dependency):
28791 Do not perform DF analysis.
28792 (pass_data_remove_partial_avx_dependency): Remove
28795 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
28797 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_n
28798 builtin generator macros.
28799 * config/aarch64/aarch64-simd.md (aarch64_<su>mull_n<mode>):
28801 * config/aarch64/arm_neon.h (vmull_n_s16): Use RTL builtin
28802 instead of inline asm.
28803 (vmull_n_s32): Likewise.
28804 (vmull_n_u16): Likewise.
28805 (vmull_n_u32): Likewise.
28807 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28809 * config/aarch64/aarch64-simd-builtins.def (sabdl2, uabdl2):
28811 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
28813 (aarch64_<sur>abdl2<mode>): ... This.
28814 (<sur>sadv16qi): Adjust use of above.
28815 * config/aarch64/arm_neon.h (vabdl_high_s8): Reimplement using
28817 (vabdl_high_s16): Likewise.
28818 (vabdl_high_s32): Likewise.
28819 (vabdl_high_u8): Likewise.
28820 (vabdl_high_u16): Likewise.
28821 (vabdl_high_u32): Likewise.
28823 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28825 * config/aarch64/aarch64-simd-builtins.def (sabal2): Define
28827 (uabal2): Likewise.
28828 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal2<mode>): New
28830 * config/aarch64/aarch64.md (unspec): Add UNSPEC_SABAL2 and
28832 * config/aarch64/arm_neon.h (vabal_high_s8): Reimplement using
28834 (vabal_high_s16): Likewise.
28835 (vabal_high_s32): Likewise.
28836 (vabal_high_u8): Likewise.
28837 (vabal_high_u16): Likewise.
28838 (vabal_high_u32): Likewise.
28839 * config/aarch64/iterators.md (ABAL2): New mode iterator.
28840 (sur): Handle UNSPEC_SABAL2, UNSPEC_UABAL2.
28842 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28844 * config/aarch64/aarch64-simd-builtins.def (sabal): Define
28847 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal<mode>_4):
28849 (aarch64_<sur>abal<mode>): ... This
28850 (<sur>sadv16qi): Adust use of the above.
28851 * config/aarch64/arm_neon.h (vabal_s8): Reimplement using
28853 (vabal_s16): Likewise.
28854 (vabal_s32): Likewise.
28855 (vabal_u8): Likewise.
28856 (vabal_u16): Likewise.
28857 (vabal_u32): Likewise.
28859 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28861 * config/aarch64/aarch64-simd-builtins.def (saddlv, uaddlv):
28863 * config/aarch64/aarch64-simd.md (aarch64_<su>addlv<mode>):
28865 * config/aarch64/arm_neon.h (vaddlv_s8): Reimplement using
28867 (vaddlv_s16): Likewise.
28868 (vaddlv_u8): Likewise.
28869 (vaddlv_u16): Likewise.
28870 (vaddlvq_s8): Likewise.
28871 (vaddlvq_s16): Likewise.
28872 (vaddlvq_s32): Likewise.
28873 (vaddlvq_u8): Likewise.
28874 (vaddlvq_u16): Likewise.
28875 (vaddlvq_u32): Likewise.
28876 (vaddlv_s32): Likewise.
28877 (vaddlv_u32): Likewise.
28878 * config/aarch64/iterators.md (VDQV_L): New mode iterator.
28879 (unspec): Add UNSPEC_SADDLV, UNSPEC_UADDLV.
28880 (Vwstype): New mode attribute.
28882 (VWIDE_S): Likewise.
28883 (USADDLV): New int iterator.
28884 (su): Handle UNSPEC_SADDLV, UNSPEC_UADDLV.
28886 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
28888 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_lane[q]
28889 builtin generator macros.
28890 * config/aarch64/aarch64-simd.md (aarch64_vec_<su>mlsl_lane<Qlane>):
28892 * config/aarch64/arm_neon.h (vmlsl_lane_s16): Use RTL builtin
28893 instead of inline asm.
28894 (vmlsl_lane_s32): Likewise.
28895 (vmlsl_lane_u16): Likewise.
28896 (vmlsl_lane_u32): Likewise.
28897 (vmlsl_laneq_s16): Likewise.
28898 (vmlsl_laneq_s32): Likewise.
28899 (vmlsl_laneq_u16): Likewise.
28900 (vmlsl_laneq_u32): Likewise.
28902 2021-01-29 Richard Biener <rguenther@suse.de>
28904 * doc/invoke.texi (--param max-gcse-memory): Document unit
28906 * gcse.c (gcse_or_cprop_is_too_expensive): Adjust.
28907 * params.opt (--param max-gcse-memory): Adjust default and
28908 document unit of size.
28910 2021-01-29 Richard Biener <rguenther@suse.de>
28912 PR rtl-optimization/98863
28913 * gcse.c (gcse_or_cprop_is_too_expensive): Use unsigned
28914 HOST_WIDE_INT for the memory estimate.
28916 2021-01-29 Bin Cheng <bin.cheng@linux.alibaba.com>
28917 Richard Biener <rguenther@suse.de>
28919 PR tree-optimization/97627
28920 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
28921 Do not analyze fake edges.
28923 2021-01-29 Richard Biener <rguenther@suse.de>
28925 PR rtl-optimization/98144
28926 * df.h (df_mir_bb_info): Add con_visited member.
28927 * df-problems.c (df_mir_alloc): Initialize con_visited,
28928 do not fully populate IN and OUT.
28929 (df_mir_reset): Likewise.
28930 (df_mir_confluence_0): Set con_visited.
28931 (df_mir_confluence_n): Properly handle implicitely
28932 fully populated IN and OUT as designated by con_visited
28933 and update con_visited accordingly.
28935 2021-01-29 Jakub Jelinek <jakub@redhat.com>
28938 * config/arm/vec-common.md (mve_vshlq_<supf><mode>,
28939 vashl<mode>3, vashr<mode>3, vlshr<mode>3): Add
28940 && !TARGET_REALLY_IWMMXT to conditions.
28942 2021-01-29 Jakub Jelinek <jakub@redhat.com>
28945 * cfgbuild.c (find_bb_boundaries): Reset debug_insn when seeing
28948 2021-01-28 Marek Polacek <polacek@redhat.com>
28951 * stor-layout.c (finalize_type_size): If we reset TYPE_USER_ALIGN in
28952 the main variant, maybe reset it in its variants too.
28953 * tree.c (check_base_type): Return true only if TYPE_USER_ALIGN match.
28954 (check_aligned_type): Check if TYPE_USER_ALIGN match.
28956 2021-01-28 Christophe Lyon <christophe.lyon@linaro.org>
28959 * config/arm/arm.c (arm_rtx_costs_internal): Adjust cost of vector
28960 of constant zero for comparisons.
28962 2021-01-28 Michael Meissner <meissner@linux.ibm.com>
28964 * config/rs6000/rs6000.c (rs6000_mangle_decl_assembler_name): Add
28965 support for mapping built-in function names for long double
28966 built-in functions if long double is IEEE 128-bit.
28968 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
28970 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_n
28971 builtin generator macros.
28972 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_n<mode>):
28974 * config/aarch64/arm_neon.h (vmlsl_n_s16): Use RTL builtin
28975 instead of inline asm.
28976 (vmlsl_n_s32): Likewise.
28977 (vmlsl_n_u16): Likewise.
28978 (vmlsl_n_u32): Likewise.
28980 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
28982 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_n
28983 builtin generator macros.
28984 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>):
28986 * config/aarch64/arm_neon.h (vmlal_n_s16): Use RTL builtin
28987 instead of inline asm.
28988 (vmlal_n_s32): Likewise.
28989 (vmlal_n_u16): Likewise.
28990 (vmlal_n_u32): Likewise.
28992 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28994 * config/aarch64/aarch64-simd-builtins.def (shrn2): Define
28996 * config/aarch64/aarch64-simd.md (aarch64_shrn2<mode>_insn_le):
28998 (aarch64_shrn2<mode>_insn_be): Likewise.
28999 (aarch64_shrn2<mode>): Likewise.
29000 * config/aarch64/arm_neon.h (vshrn_high_n_s16): Reimlplement
29002 (vshrn_high_n_s32): Likewise.
29003 (vshrn_high_n_s64): Likewise.
29004 (vshrn_high_n_u16): Likewise.
29005 (vshrn_high_n_u32): Likewise.
29006 (vshrn_high_n_u64): Likewise.
29008 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29010 * config/aarch64/aarch64-simd-builtins.def (shrn): Define
29012 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le):
29014 (aarch64_shrn<mode>_insn_be): Likewise.
29015 (aarch64_shrn<mode>): Likewise.
29016 * config/aarch64/arm_neon.h (vshrn_n_s16): Reimplement using
29018 (vshrn_n_s32): Likewise.
29019 (vshrn_n_s64): Likewise.
29020 (vshrn_n_u16): Likewise.
29021 (vshrn_n_u32): Likewise.
29022 (vshrn_n_u64): Likewise.
29023 * config/aarch64/iterators.md (vn_mode): New mode attribute.
29025 2021-01-28 Richard Biener <rguenther@suse.de>
29027 PR rtl-optimization/80960
29028 * dse.c (check_mem_read_rtx): Call get_addr on the
29031 2021-01-28 Xionghu Luo <luoxhu@linux.ibm.com>
29032 David Edelsohn <dje.gcc@gmail.com>
29035 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
29036 Don't generate VIEW_CONVERT_EXPR for fcode ALTIVEC_BUILTIN_VEC_INSERT
29038 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
29040 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Remove the
29041 wrapper call rs6000_expand_vector_set_var for cleanup. Call
29042 rs6000_expand_vector_set_var_p9 and rs6000_expand_vector_set_var_p8
29044 (rs6000_expand_vector_set_var): Delete.
29045 (rs6000_expand_vector_set_var_p9): Make static.
29046 (rs6000_expand_vector_set_var_p8): Make static.
29048 2021-01-28 Xing GUO <higuoxing@gmail.com>
29050 * common/config/riscv/riscv-common.c
29051 (riscv_subset_list::parsing_subset_version): Fix -march option parsing
29052 when `p` extension exists.
29054 2021-01-27 Vladimir N. Makarov <vmakarov@redhat.com>
29056 PR rtl-optimization/97684
29057 * ira.c (ira): Call ira_set_pseudo_classes before
29058 update_equiv_regs when it is necessary.
29060 2021-01-27 Jakub Jelinek <jakub@redhat.com>
29063 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Use
29064 %w0, %w1 and %2 instead of %0, %1 and %2.
29066 2021-01-27 Aaron Sawdey <acsawdey@linux.ibm.com>
29068 * config/rs6000/genfusion.pl: New script to generate
29069 define_insn_and_split patterns so combine can arrange fused
29070 instructions next to each other.
29071 * config/rs6000/fusion.md: New file, generated fused instruction
29072 patterns for combine.
29073 * config/rs6000/predicates.md (const_m1_to_1_operand): New predicate.
29074 (non_update_memory_operand): New predicate.
29075 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_P10_FUSION and
29076 OPTION_MASK_P10_FUSION_LD_CMPI to ISA_3_1_MASKS_SERVER and
29078 * config/rs6000/rs6000-protos.h (address_is_non_pfx_d_or_x): Add
29080 * config/rs6000/rs6000.c (rs6000_option_override_internal):
29081 Automatically set OPTION_MASK_P10_FUSION and
29082 OPTION_MASK_P10_FUSION_LD_CMPI if target is power10.
29083 (rs600_opt_masks): Allow -mpower10-fusion
29084 in function attributes.
29085 (address_is_non_pfx_d_or_x): New function.
29086 * config/rs6000/rs6000.h: Add MASK_P10_FUSION.
29087 * config/rs6000/rs6000.md: Include fusion.md.
29088 * config/rs6000/rs6000.opt: Add -mpower10-fusion
29089 and -mpower10-fusion-ld-cmpi.
29090 * config/rs6000/t-rs6000: Add dependencies involving fusion.md.
29092 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
29094 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal
29095 builtin generator macros.
29096 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal<mode>):
29098 (aarch64_<su>mlal<mode>): This.
29099 * config/aarch64/arm_neon.h (vmlal_s8): Use RTL builtin
29100 instead of inline asm.
29101 (vmlal_s16): Likewise.
29102 (vmlal_s32): Likewise.
29103 (vmlal_u8): Likewise.
29104 (vmlal_u16): Likewise.
29105 (vmlal_u32): Likewise.
29107 2021-01-27 Richard Biener <rguenther@suse.de>
29109 PR tree-optimization/98854
29110 * tree-vect-slp.c (vect_build_slp_tree_2): Also build
29111 PHIs from scalars when the number of CTORs matches the
29112 number of children.
29114 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
29116 * config/aarch64/aarch64-simd-builtins.def: Add mls_n builtin
29118 * config/aarch64/aarch64-simd.md (*aarch64_mls_elt_merge<mode>):
29120 (aarch64_mls_n<mode>): This.
29121 * config/aarch64/arm_neon.h (vmls_n_s16): Use RTL builtin
29123 (vmls_n_s32): Likewise.
29124 (vmls_n_u16): Likewise.
29125 (vmls_n_u32): Likewise.
29126 (vmlsq_n_s16): Likewise.
29127 (vmlsq_n_s32): Likewise.
29128 (vmlsq_n_u16): Likewise.
29129 (vmlsq_n_u32): Likewise.
29131 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
29133 * config/aarch64/aarch64-simd-builtins.def: Add mls builtin
29135 * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather
29137 (vmls_s16): Likewise.
29138 (vmls_s32): Likewise.
29139 (vmls_u8): Likewise.
29140 (vmls_u16): Likewise.
29141 (vmls_u32): Likewise.
29142 (vmlsq_s8): Likewise.
29143 (vmlsq_s16): Likewise.
29144 (vmlsq_s32): Likewise.
29145 (vmlsq_u8): Likewise.
29146 (vmlsq_u16): Likewise.
29147 (vmlsq_u32): Likewise.
29149 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
29151 * config/aarch64/aarch64-simd-builtins.def: Add mla_n builtin
29153 * config/aarch64/aarch64-simd.md (*aarch64_mla_elt_merge<mode>):
29155 (aarch64_mla_n<mode>): This.
29156 * config/aarch64/arm_neon.h (vmla_n_s16): Use RTL builtin
29158 (vmla_n_s32): Likewise.
29159 (vmla_n_u16): Likewise.
29160 (vmla_n_u32): Likewise.
29161 (vmlaq_n_s16): Likewise.
29162 (vmlaq_n_s32): Likewise.
29163 (vmlaq_n_u16): Likewise.
29164 (vmlaq_n_u32): Likewise.
29166 2021-01-27 liuhongt <hongtao.liu@intel.com>
29169 * config/i386/sse.md (sse2_gt<mode>3): Drop !TARGET_XOP in condition.
29170 (*sse2_eq<mode>3): Ditto.
29172 2021-01-27 Jakub Jelinek <jakub@redhat.com>
29174 * tree-pass.h (PROP_trees): Rename to ...
29175 (PROP_gimple): ... this.
29176 * cfgexpand.c (pass_data_expand): Replace PROP_trees with PROP_gimple.
29177 * passes.c (execute_function_dump, execute_function_todo,
29178 execute_one_ipa_transform_pass, execute_one_pass): Likewise.
29179 * varpool.c (ctor_for_folding): Likewise.
29181 2021-01-27 Jakub Jelinek <jakub@redhat.com>
29183 PR tree-optimization/97260
29184 * varpool.c: Include tree-pass.h.
29185 (ctor_for_folding): In GENERIC return DECL_INITIAL for TREE_READONLY
29186 non-TREE_SIDE_EFFECTS automatic variables.
29188 2021-01-26 Paul Fee <paul.f.fee@gmail.com>
29190 * doc/cpp.texi (__cplusplus): Document value for -std=c++23
29192 * doc/invoke.texi: Document -std=c++23 and -std=gnu++23.
29193 * dwarf2out.c (highest_c_language): Recognise C++20 and C++23.
29194 (gen_compile_unit_die): Recognise C++23.
29196 2021-01-26 Jakub Jelinek <jakub@redhat.com>
29199 * dwarf2asm.c (dw2_assemble_integer): Cast DWARF2_ADDR_SIZE to int
29202 2021-01-26 Jakub Jelinek <jakub@redhat.com>
29205 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
29206 Use UINTVAL (shft_amnt) and UINTVAL (mask) instead of INTVAL (shft_amnt)
29207 and INTVAL (mask). Add && INTVAL (mask) > 0 condition.
29209 2021-01-26 Richard Biener <rguenther@suse.de>
29211 * gimple-pretty-print.c (dump_binary_rhs): Handle
29212 VEC_WIDEN_{PLUS,MINUS}_{LO,HI}_EXPR.
29214 2021-01-26 Richard Biener <rguenther@suse.de>
29216 PR middle-end/98726
29217 * tree.h (vector_cst_int_elt): Remove.
29218 * tree.c (vector_cst_int_elt): Use poly_wide_int for computations,
29221 2021-01-26 Andrew Stubbs <ams@codesourcery.com>
29223 * config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
29224 for V64DFmode min/max reductions.
29226 2021-01-26 Jakub Jelinek <jakub@redhat.com>
29228 * dwarf2asm.c (dw2_assemble_integer): Handle size twice as large
29229 as DWARF2_ADDR_SIZE if x is not a scalar int by emitting it as
29230 two halves, one with x and the other with const0_rtx, ordered
29231 depending on endianity.
29233 2021-01-26 Alexandre Oliva <oliva@adacore.com>
29235 * gimplify.c (gimplify_decl_expr): Skip asan marking calls for
29236 temporaries not seen in binding block, and not about to be
29237 added as gimple variables.
29239 2021-01-25 Martin Sebor <msebor@redhat.com>
29242 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust warning text.
29244 2021-01-25 Martin Liska <mliska@suse.cz>
29246 * value-prof.c (get_nth_most_common_value): Use %s instead
29249 2021-01-25 Jakub Jelinek <jakub@redhat.com>
29252 * configure.ac (HAVE_AS_GDWARF_5_DEBUG_FLAG): Only define if
29253 readelf -wi is able to read the emitted .debug_info back.
29254 * configure: Regenerated.
29256 2021-01-25 Martin Liska <mliska@suse.cz>
29258 PR gcov-profile/98739
29259 * common.opt: Add missing sign symbol.
29260 * value-prof.c (get_nth_most_common_value): Restore handling
29261 of PROFILE_REPRODUCIBILITY_PARALLEL_RUNS and
29262 PROFILE_REPRODUCIBILITY_MULTITHREADED.
29264 2021-01-25 Richard Biener <rguenther@suse.de>
29266 PR middle-end/98807
29267 * tree.c (vector_element_bits): Always use precision of
29268 the element type for boolean vectors.
29270 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
29272 * config/rtems.h (STARTFILE_SPEC): Remove qnolinkcmds.
29273 (ENDFILE_SPEC): Evaluate qnolinkcmds.
29275 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
29277 * config/rtems.h (STARTFILE_SPEC): Remove nostdlib and
29278 nostartfiles handling since this is already done by
29279 LINK_COMMAND_SPEC. Evaluate qnolinkcmds.
29280 (ENDFILE_SPEC): Remove nostdlib and nostartfiles handling since this
29281 is already done by LINK_COMMAND_SPEC.
29282 (LIB_SPECS): Remove nostdlib and nodefaultlibs handling since
29283 this is already done by LINK_COMMAND_SPEC. Remove qnolinkcmds
29286 2021-01-25 Jakub Jelinek <jakub@redhat.com>
29289 * fold-const-call.c (host_size_t_cst_p): Renamed to ...
29290 (size_t_cst_p): ... this. Check and store unsigned HOST_WIDE_INT
29291 value rather than host size_t.
29292 (fold_const_call): Change type of s2 from size_t to
29293 unsigned HOST_WIDE_INT. Use size_t_cst_p instead of
29294 host_size_t_cst_p. For strncmp calls, pass MIN (s2, SIZE_MAX)
29295 instead of s2 as last argument.
29297 2021-01-25 Tamar Christina <tamar.christina@arm.com>
29299 * config/arm/iterators.md (rotsplit1, rotsplit2, conj_op, fcmac1,
29300 VCMLA_OP, VCMUL_OP): New.
29301 * config/arm/mve.md (mve_vcmlaq<mve_rot><mode>): Support vec_dup 0.
29302 * config/arm/neon.md (cmul<conj_op><mode>3): New.
29303 * config/arm/unspecs.md (UNSPEC_VCMLA_CONJ, UNSPEC_VCMLA180_CONJ,
29304 UNSPEC_VCMUL_CONJ): New.
29305 * config/arm/vec-common.md (cmul<conj_op><mode>3, arm_vcmla<rot><mode>,
29306 cml<fcmac1><conj_op><mode>4): New.
29308 2021-01-23 Jakub Jelinek <jakub@redhat.com>
29311 * config/rs6000/mmintrin.h (__m64): Add __may_alias__ attribute.
29313 2021-01-22 Jonathan Wright <jonathan.wright@arm.com>
29315 * config/aarch64/aarch64-simd-builtins.def: Add mla builtin
29317 * config/aarch64/arm_neon.h (vmla_s8): Use RTL builtin rather
29319 (vmla_s16): Likewise.
29320 (vmla_s32): Likewise.
29321 (vmla_u8): Likewise.
29322 (vmla_u16): Likewise.
29323 (vmla_u32): Likewise.
29324 (vmlaq_s8): Likewise.
29325 (vmlaq_s16): Likewise.
29326 (vmlaq_s32): Likewise.
29327 (vmlaq_u8): Likewise.
29328 (vmlaq_u16): Likewise.
29329 (vmlaq_u32): Likewise.
29331 2021-01-22 David Malcolm <dmalcolm@redhat.com>
29333 * doc/invoke.texi (GCC_EXTRA_DIAGNOSTIC_OUTPUT): Add @findex
29336 2021-01-22 Jakub Jelinek <jakub@redhat.com>
29339 * dwarf2out.c (output_file_names): For -gdwarf-5, if there are no
29340 filenames to emit, still emit the required 0 index directory and
29341 filename entries that match DW_AT_comp_dir and DW_AT_name of the
29344 2021-01-22 Marek Polacek <polacek@redhat.com>
29347 * doc/invoke.texi: Update C++ ABI Version 15 description.
29349 2021-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29351 PR tree-optimization/98766
29352 * tree-ssa-math-opts.c (convert_mult_to_fma): Use maybe_le when
29353 comparing against type size with param_avoid_fma_max_bits.
29355 2021-01-22 Richard Biener <rguenther@suse.de>
29357 PR middle-end/98793
29358 * tree.c (vector_element_bits): Key single-bit bool vector on
29359 integer mode rather than not vector mode.
29361 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
29364 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
29365 Generate ARRAY_REF(VIEW_CONVERT_EXPR) for P8 and later
29367 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): Update
29368 to call different path for P8 and P9.
29369 (rs6000_expand_vector_set_var_p9): New function.
29370 (rs6000_expand_vector_set_var_p8): New function.
29372 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
29376 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
29377 Ajdust variable index vec_insert from address dereference to
29378 ARRAY_REF(VIEW_CONVERT_EXPR) tree expression.
29379 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
29381 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): New function.
29383 2021-01-22 Martin Liska <mliska@suse.cz>
29385 PR gcov-profile/98739
29386 * profile.c (compute_value_histograms): Drop time profile for
29387 -fprofile-reproducible=multithreaded.
29389 2021-01-22 Nathan Sidwell <nathan@acm.org>
29391 * gcc.c (process_command): Don't check OPT_SPECIAL_input_file
29394 2021-01-22 Richard Biener <rguenther@suse.de>
29396 PR middle-end/98773
29397 * tree-data-ref.c (initalize_matrix_A): Revert previous
29398 change, retaining failing on HOST_WIDE_INT_MIN CHREC_RIGHT.
29400 2021-01-22 Jakub Jelinek <jakub@redhat.com>
29402 PR tree-optimization/90248
29403 * match.pd (X cmp 0.0 ? 1.0 : -1.0 -> copysign(1, +-X),
29404 X cmp 0.0 ? -1.0 : +1.0 -> copysign(1, -+X)): Remove
29406 (X * (X cmp 0.0 ? 1.0 : -1.0) -> +-abs(X),
29407 X * (X cmp 0.0 ? -1.0 : 1.0) -> +-abs(X)): New simplifications.
29409 2021-01-22 Jakub Jelinek <jakub@redhat.com>
29411 PR tree-optimization/98255
29412 * tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
29413 extend index - low_bound from sizetype's precision rather than index
29415 (get_addr_base_and_unit_offset_1): Likewise.
29416 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
29417 * gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
29419 2021-01-22 Richard Biener <rguenther@suse.de>
29421 PR tree-optimization/98786
29422 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Avoid
29423 adding new uses of abnormals. Verify we deal with a conditional
29426 2021-01-22 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
29429 * optc-save-gen.awk: Add arm_fp16_format to checked_options.
29431 2021-01-22 liuhongt <hongtao.liu@intel.com>
29435 * config/i386/sse.md (VI_128_256): New mode iterator.
29436 (*avx_cmp<mode>3_1, *avx_cmp<mode>3_2, *avx_cmp<mode>3_3,
29437 *avx_cmp<mode>3_4, *avx2_eq<mode>3, *avx2_pcmp<mode>3_1,
29438 *avx2_pcmp<mode>3_2, *avx2_gt<mode>3): New
29439 define_insn_and_split to lower avx512 vector comparison to avx
29440 version when dest is vector.
29441 (*<avx512>_cmp<mode>3,*<avx512>_cmp<mode>3,*<avx512>_ucmp<mode>3):
29442 define_insn_and_split for negating the comparison result.
29443 * config/i386/predicates.md (float_vector_all_ones_operand):
29445 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Use
29446 general NOT operator without UNSPEC_MASKOP.
29448 2021-01-21 Vladimir N. Makarov <vmakarov@redhat.com>
29450 PR rtl-optimization/98777
29451 * lra-int.h (lra_pmode_pseudo): New extern.
29452 * lra.c (lra_pmode_pseudo): New global.
29454 * lra-eliminations.c (eliminate_regs_in_insn): Use it.
29456 2021-01-21 Ilya Leoshkevich <iii@linux.ibm.com>
29458 * fwprop.c (fwprop_propagation::classify_result): Allow
29459 (subreg (mem)) simplifications.
29461 2021-01-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29463 * config/aarch64/aarch64-simd.md (aarch64_sqdml<SBINQOPS:as>l<mode>):
29465 (aarch64_sqdmlal<mode>): ... This...
29466 (aarch64_sqdmlsl<mode>): ... And this.
29467 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Split into...
29468 (aarch64_sqdmlal_lane<mode>): ... This...
29469 (aarch64_sqdmlsl_lane<mode>): ... And this.
29470 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Split into...
29471 (aarch64_sqdmlsl_laneq<mode>): ... This...
29472 (aarch64_sqdmlal_laneq<mode>): ... And this.
29473 (aarch64_sqdml<SBINQOPS:as>l_n<mode>): Split into...
29474 (aarch64_sqdmlsl_n<mode>): ... This...
29475 (aarch64_sqdmlal_n<mode>): ... And this.
29476 (aarch64_sqdml<SBINQOPS:as>l2<mode>_internal): Split into...
29477 (aarch64_sqdmlal2<mode>_internal): ... This...
29478 (aarch64_sqdmlsl2<mode>_internal): ... And this.
29480 2021-01-21 Christophe Lyon <christophe.lyon@linaro.org>
29482 * config/arm/arm_mve.h (__arm_vcmpneq_s8): Fix return type.
29484 2021-01-21 Andrea Corallo <andrea.corallo@arm.com>
29487 * doc/sourcebuild.texi (arm_thumb2_no_arm_v8_1_lob): Document.
29489 2021-01-21 liuhongt <hongtao.liu@intel.com>
29491 PR rtl-optimization/98694
29492 * regcprop.c (copy_value): If SRC had been assigned a mode
29493 narrower than the copy, we can't link DEST into the chain even
29494 they have same hard_regno_nregs(i.e. HImode/SImode in i386
29497 2021-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29499 * config/aarch64/aarch64-simd.md (aarch64_get_lane<mode>):
29500 Convert to define_insn_and_split. Split into simple move when moving
29503 2021-01-20 Segher Boessenkool <segher@kernel.crashing.org>
29505 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Change assert.
29506 Adjust comment. Simplify code.
29508 2021-01-20 Jakub Jelinek <jakub@redhat.com>
29511 * dwarf2out.c (reset_indirect_string): Also reset indirect strings
29512 with DW_FORM_line_strp form.
29513 (prune_unused_types_update_strings): Don't add into debug_str_hash
29514 indirect strings with DW_FORM_line_strp form.
29515 (adjust_name_comp_dir): New function.
29516 (dwarf2out_finish): Call it on CU DIEs after resetting
29517 debug_line_str_hash.
29519 2021-01-20 Vladimir N. Makarov <vmakarov@redhat.com>
29521 PR rtl-optimization/98722
29522 * lra-eliminations.c (eliminate_regs_in_insn): Check that target
29523 has no 3-op add insn to transform insns containing two pluses.
29525 2021-01-20 Richard Biener <rguenther@suse.de>
29527 * hwint.h (add_hwi): New function.
29528 (mul_hwi): Likewise.
29529 * tree-data-ref.c (initialize_matrix_A): Properly translate
29530 tree constants and avoid HOST_WIDE_INT_MIN.
29531 (lambda_matrix_row_add): Avoid undefined integer overflow
29532 and return true on such overflow.
29533 (lambda_matrix_right_hermite): Handle overflow from
29534 lambda_matrix_row_add gracefully. Simplify previous fix.
29535 (analyze_subscript_affine_affine): Likewise.
29537 2021-01-20 Eugene Rozenfeld <erozen@microsoft.com>
29539 PR tree-optimization/96674
29540 * match.pd: New patterns: x < y || y == XXX_MIN --> x <= y - 1
29541 x >= y && y != XXX_MIN --> x > y - 1
29543 2021-01-20 Richard Sandiford <richard.sandiford@arm.com>
29545 PR tree-optimization/98535
29546 * tree-vect-slp.c (duplicate_and_interleave): Use quick_grow_cleared.
29547 If the high and low permutes are the same, remove the high permutes
29548 from the working set and only continue with the low ones.
29550 2021-01-20 Jakub Jelinek <jakub@redhat.com>
29552 PR tree-optimization/98721
29553 * builtins.c (access_ref::inform_access): Don't assume
29554 SSA_NAME_IDENTIFIER must be non-NULL. Print messages about
29555 object whenever allocfn is NULL, rather than only when DECL_P
29556 is true. Use %qE instead of %qD for that. Formatting fixes.
29558 2021-01-20 Richard Biener <rguenther@suse.de>
29560 PR tree-optimization/98758
29561 * tree-data-ref.c (int_divides_p): Use lambda_int arguments.
29562 (lambda_matrix_right_hermite): Avoid undefinedness with
29563 signed integer abs and multiplication.
29564 (analyze_subscript_affine_affine): Use lambda_int.
29566 2021-01-20 David Malcolm <dmalcolm@redhat.com>
29569 * dwarf2out.c (output_line_info): Rename static variable
29570 "generation", moving it out of the function to...
29571 (output_line_info_generation): New.
29572 (init_sections_and_labels): Likewise, renaming the variable to...
29573 (init_sections_and_labels_generation): New.
29574 (dwarf2out_c_finalize): Reset the new variables.
29576 2021-01-19 Martin Sebor <msebor@redhat.com>
29578 PR middle-end/98664
29579 * tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
29580 all functions, even if they're not declared artificial or inline.
29581 * tree.c (tree_inlined_location): Use macro expansion location
29582 only if scope traversal fails to expose one.
29584 2021-01-19 Richard Sandiford <richard.sandiford@arm.com>
29586 PR rtl-optimization/92294
29587 * alias.c (compare_base_symbol_refs): Take an extra parameter
29588 and add the distance between two symbols to it. Enshrine in
29589 comments that -1 means "either 0 or 1, but we can't tell
29590 which at compile time".
29591 (memrefs_conflict_p): Update call accordingly.
29592 (rtx_equal_for_memref_p): Likewise. Take the distance between symbols
29595 2021-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29597 * config/aarch64/aarch64-simd-builtins.def (sqshl, uqshl,
29598 sqrshl, uqrshl, sqadd, uqadd, sqsub, uqsub, suqadd, usqadd, sqmovn,
29599 uqmovn, sqxtn2, uqxtn2, sqabs, sqneg, sqdmlal, sqdmlsl, sqdmlal_lane,
29600 sqdmlsl_lane, sqdmlal_laneq, sqdmlsl_laneq, sqdmlal_n, sqdmlsl_n,
29601 sqdmlal2, sqdmlsl2, sqdmlal2_lane, sqdmlsl2_lane, sqdmlal2_laneq,
29602 sqdmlsl2_laneq, sqdmlal2_n, sqdmlsl2_n, sqdmull, sqdmull_lane,
29603 sqdmull_laneq, sqdmull_n, sqdmull2, sqdmull2_lane, sqdmull2_laneq,
29604 sqdmull2_n, sqdmulh, sqrdmulh, sqdmulh_lane, sqdmulh_laneq,
29605 sqrdmulh_lane, sqrdmulh_laneq, sqshrun_n, sqrshrun_n, sqshrn_n,
29606 uqshrn_n, sqrshrn_n, uqrshrn_n, sqshlu_n, sqshl_n, uqshl_n, sqrdmlah,
29607 sqrdmlsh, sqrdmlah_lane, sqrdmlsh_lane, sqrdmlah_laneq, sqrdmlsh_laneq,
29608 sqmovun): Use NONE flags.
29610 2021-01-19 Richard Biener <rguenther@suse.de>
29613 * ipa-modref.c (analyze_stmt): Only record a summary for a
29616 2021-01-19 Richard Biener <rguenther@suse.de>
29618 PR middle-end/98638
29619 * tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
29621 2021-01-19 Daniel Hellstrom <daniel@gaisler.com>
29623 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
29624 built-in define __FIX_LEON3FT_TN0018.
29626 2021-01-19 Richard Biener <rguenther@suse.de>
29629 * tree-inline.c (tree_function_versioning): Set input_location
29630 to UNKNOWN_LOCATION throughout the function.
29632 2021-01-19 Tobias Burnus <tobias@codesourcery.com>
29635 * omp-low.c (lower_omp_target): Handle nonpointer is_device_ptr.
29637 2021-01-19 Martin Jambor <mjambor@suse.cz>
29640 * ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
29641 whether non-call exceptions allow removal of a statement.
29642 (isra_analyze_call): Pass the appropriate function to
29643 ssa_name_only_returned_p.
29645 2021-01-19 Geng Qi <gengqi@linux.alibaba.com>
29647 * config/riscv/arch-canonicalize (longext_sort): New function for
29648 sorting 'multi-letter'.
29649 * config/riscv/multilib-generator: Adjusting the loop of 'alt' in
29650 'alts'. The 'arch' may not be the first of 'alts'.
29651 (_expand_combination): Add underline for the 'ext' without '*'.
29652 This is because, a single-letter extension can always be treated well
29653 with a '_' prefix, but it cannot be separated out if it is appended
29656 2021-01-18 Vladimir N. Makarov <vmakarov@redhat.com>
29659 * ira.c (ira): Skip abnormal critical edge splitting.
29661 2021-01-18 Jakub Jelinek <jakub@redhat.com>
29663 PR tree-optimization/98727
29664 * tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
29665 second .MUL_OVERFLOW operand for signed multiplication with overflow
29666 checking if the second operand of multiplication is not constant.
29668 2021-01-18 David Edelsohn <dje.gcc@gmail.com>
29670 * doc/invoke.texi (-gdwarf): TPF defaults to version 2 and AIX
29671 defaults to version 4.
29673 2021-01-18 David Malcolm <dmalcolm@redhat.com>
29675 * attribs.h (fndecl_dealloc_argno): New decl.
29676 * builtins.c (call_dealloc_argno): Split out second half of
29678 (fndecl_dealloc_argno): New.
29679 * doc/extend.texi (Common Function Attributes): Document the
29680 interaction between the analyzer and the malloc attribute.
29681 * doc/invoke.texi (Static Analyzer Options): Likewise.
29683 2021-01-17 David Edelsohn <dje.gcc@gmail.com>
29685 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Override
29686 dwarf_version to 4.
29687 * config/rs6000/aix72.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
29689 2021-01-17 Martin Jambor <mjambor@suse.cz>
29692 * cgraph.c (clone_of_p): Check also former_clone_of as we climb
29695 2021-01-17 Mark Wielaard <mark@klomp.org>
29697 * common.opt (gdwarf-): Init(5).
29698 * doc/invoke.texi (-gdwarf): Document default to 5.
29700 2021-01-16 Kwok Cheung Yeung <kcy@codesourcery.com>
29702 * builtin-types.def
29703 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
29705 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
29706 ...this. Add extra argument.
29707 * gimplify.c (omp_default_clause): Ensure that event handle is
29708 firstprivate in a task region.
29709 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_DETACH.
29710 (gimplify_adjust_omp_clauses): Likewise.
29711 * omp-builtins.def (BUILT_IN_GOMP_TASK): Change function type to
29712 BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR.
29713 * omp-expand.c (expand_task_call): Add GOMP_TASK_FLAG_DETACH to flags
29714 if detach clause specified. Add detach argument when generating
29716 * omp-low.c (scan_sharing_clauses): Setup data environment for detach
29718 (finish_taskreg_scan): Move field for variable containing the event
29719 handle to the front of the struct.
29720 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_DETACH. Fix
29722 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
29723 OMP_CLAUSE_DETACH clause.
29724 (convert_local_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
29725 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_DETACH.
29726 * tree.c (omp_clause_num_ops): Add entry for OMP_CLAUSE_DETACH.
29728 (omp_clause_code_name): Add entry for OMP_CLAUSE_DETACH. Fix
29730 (walk_tree_1): Handle OMP_CLAUSE_DETACH.
29732 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
29734 * config/nios2/t-rtems: Reset all MULTILIB_* variables. Shorten
29735 multilib directory names. Use MULTILIB_REQUIRED instead of
29736 MULTILIB_EXCEPTIONS. Add -mhw-mul -mhw-mulx -mhw-div
29737 -mcustom-fpu-cfg=fph2 multilib.
29739 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
29741 * config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
29742 (nios2_init_fpu_configs): Provide register values for new
29743 -mcustom-fpu-cfg=fph2 option variant.
29744 * doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
29747 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
29749 * config/nios2/nios2.c (nios2_custom_check_insns): Remove
29750 custom instruction warnings.
29752 2021-01-16 Jakub Jelinek <jakub@redhat.com>
29754 PR tree-optimization/96669
29755 * match.pd ((CST << x) & 1 -> x == 0): New simplification.
29757 2021-01-16 Jakub Jelinek <jakub@redhat.com>
29759 PR tree-optimization/96271
29760 * passes.def: Pass false argument to first two pass_cd_dce
29761 instances and true to last instance. Add comment that
29762 last instance rewrites no longer addressed locals.
29763 * tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
29765 (pass_cd_dce::set_pass_param): New method.
29766 (pass_cd_dce::execute): Return TODO_update_address_taken from
29767 last cd_dce instance.
29769 2021-01-15 Carl Love <cel@us.ibm.com>
29771 * config/rs6000/altivec.h (vec_mulh, vec_div, vec_dive, vec_mod):
29773 * config/rs6000/altivec.md (VIlong): Move define to file vsx.md.
29774 * config/rs6000/rs6000-builtin.def (DIVES_V4SI, DIVES_V2DI,
29775 DIVEU_V4SI, DIVEU_V2DI, DIVS_V4SI, DIVS_V2DI, DIVU_V4SI,
29776 DIVU_V2DI, MODS_V2DI, MODS_V4SI, MODU_V2DI, MODU_V4SI,
29777 MULHS_V2DI, MULHS_V4SI, MULHU_V2DI, MULHU_V4SI, MULLD_V2DI):
29778 Add builtin define.
29779 (MULH, DIVE, MOD): Add new BU_P10_OVERLOAD_2 definitions.
29780 * config/rs6000/rs6000-call.c (VSX_BUILTIN_VEC_DIV,
29781 VSX_BUILTIN_VEC_DIVE, P10_BUILTIN_VEC_MOD, P10_BUILTIN_VEC_MULH):
29782 New overloaded definitions.
29783 (builtin_function_type) [P10V_BUILTIN_DIVEU_V4SI,
29784 P10V_BUILTIN_DIVEU_V2DI, P10V_BUILTIN_DIVU_V4SI,
29785 P10V_BUILTIN_DIVU_V2DI, P10V_BUILTIN_MODU_V2DI,
29786 P10V_BUILTIN_MODU_V4SI, P10V_BUILTIN_MULHU_V2DI,
29787 P10V_BUILTIN_MULHU_V4SI]: Add case
29788 statement for builtins.
29789 * config/rs6000/rs6000.md (bits): Add new attribute sizes V4SI, V2DI.
29790 * config/rs6000/vsx.md (VIlong): Moved from config/rs6000/altivec.md.
29791 (UNSPEC_VDIVES, UNSPEC_VDIVEU): New unspec definitions.
29792 (vsx_mul_v2di): Add if TARGET_POWER10 statement.
29793 (vsx_udiv_v2di): Add if TARGET_POWER10 statement.
29794 (dives_<mode>, diveu_<mode>, div<mode>3, uvdiv<mode>3,
29795 mods_<mode>, modu_<mode>, mulhs_<mode>, mulhu_<mode>, mulv2di3):
29796 Add define_insn, mode is VIlong.
29797 * doc/extend.texi (vec_mulh, vec_mul, vec_div, vec_dive, vec_mod):
29798 Add builtin descriptions.
29800 2021-01-15 Eric Botcazou <ebotcazou@adacore.com>
29802 * final.c (final_start_function_1): Reset force_source_line.
29804 2021-01-15 Jakub Jelinek <jakub@redhat.com>
29806 PR tree-optimization/96669
29807 * match.pd (((1 << A) & 1) != 0 -> A == 0,
29808 ((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
29809 possibly different power of two constants and to right shift too.
29811 2021-01-15 Jakub Jelinek <jakub@redhat.com>
29813 PR tree-optimization/96681
29814 * match.pd ((x < 0) ^ (y < 0) to (x ^ y) < 0): New simplification.
29815 ((x >= 0) ^ (y >= 0) to (x ^ y) < 0): Likewise.
29816 ((x < 0) ^ (y >= 0) to (x ^ y) >= 0): Likewise.
29817 ((x >= 0) ^ (y < 0) to (x ^ y) >= 0): Likewise.
29819 2021-01-15 Alexandre Oliva <oliva@adacore.com>
29821 * opts.c (gen_command_line_string): Exclude -dumpbase-ext.
29823 2021-01-15 Tamar Christina <tamar.christina@arm.com>
29825 * config/aarch64/aarch64-simd.md (cml<fcmac1><conj_op><mode>4,
29826 cmul<conj_op><mode>3): New.
29827 * config/aarch64/iterators.md (UNSPEC_FCMUL,
29828 UNSPEC_FCMUL180, UNSPEC_FCMLA_CONJ, UNSPEC_FCMLA180_CONJ,
29829 UNSPEC_CMLA_CONJ, UNSPEC_CMLA180_CONJ, UNSPEC_CMUL, UNSPEC_CMUL180,
29830 FCMLA_OP, FCMUL_OP, conj_op, rotsplit1, rotsplit2, fcmac1, sve_rot1,
29831 sve_rot2, SVE2_INT_CMLA_OP, SVE2_INT_CMUL_OP, SVE2_INT_CADD_OP): New.
29832 (rot): Add UNSPEC_FCMUL, UNSPEC_FCMUL180.
29833 (rot_op): Renamed to conj_op.
29834 * config/aarch64/aarch64-sve.md (cml<fcmac1><conj_op><mode>4,
29835 cmul<conj_op><mode>3): New.
29836 * config/aarch64/aarch64-sve2.md (cml<fcmac1><conj_op><mode>4,
29837 cmul<conj_op><mode>3): New.
29839 2021-01-15 David Malcolm <dmalcolm@redhat.com>
29843 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
29844 Escape the tempfile name when constructing the expected output.
29846 2021-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29848 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlsl_hi<mode>):
29850 (aarch64_<su>mlsl_hi<mode>): ... This.
29851 (aarch64_<su>mlsl_hi<mode>): Define.
29852 (*aarch64_<su>mlsl<mode): Rename to...
29853 (aarch64_<su>mlsl<mode): ... This.
29854 * config/aarch64/aarch64-simd-builtins.def (smlsl, umlsl,
29855 smlsl_hi, umlsl_hi): Define builtins.
29856 * config/aarch64/arm_neon.h (vmlsl_high_s8, vmlsl_high_s16,
29857 vmlsl_high_s32, vmlsl_high_u8, vmlsl_high_u16, vmlsl_high_u32,
29858 vmlsl_s8, vmlsl_s16, vmlsl_s32, vmlsl_u8,
29859 vmlsl_u16, vmlsl_u32): Reimplement with builtins.
29861 2021-01-15 Uroš Bizjak <ubizjak@gmail.com>
29863 * config/i386/i386-c.c (ix86_target_macros):
29864 Use cpp_define_formatted for __SIZEOF_FLOAT80__ definition.
29866 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
29869 * config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
29870 * Makefile.in (RTL_SSA_H): New variable.
29871 * config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
29872 * config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
29873 * config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
29875 * config/aarch64/aarch64-cc-fusion.cc: New file.
29877 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
29879 * recog.h (insn_change_watermark::~insn_change_watermark): Avoid
29880 calling cancel_changes for changes that no longer exist.
29882 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
29884 * rtl-ssa/functions.h (function_info::ref_defs): Rename to...
29885 (function_info::reg_defs): ...this.
29886 * rtl-ssa/member-fns.inl (function_info::ref_defs): Rename to...
29887 (function_info::reg_defs): ...this.
29889 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
29892 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
29894 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
29897 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
29900 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
29902 2021-01-15 Richard Biener <rguenther@suse.de>
29904 PR tree-optimization/96376
29905 * tree-vect-stmts.c (get_load_store_type): Disregard alignment
29906 for VMAT_INVARIANT.
29908 2021-01-15 Martin Liska <mliska@suse.cz>
29910 * doc/install.texi: Document that some tests need pytest module.
29911 * doc/sourcebuild.texi: Likewise.
29913 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
29916 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
29918 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
29920 * config/arm/mve.md (mve_vshrq_n_s<mode>_imm): New entry.
29921 (mve_vshrq_n_u<mode>_imm): Likewise.
29922 * config/arm/neon.md (vashr<mode>3, vlshr<mode>3): Move to ...
29923 * config/arm/vec-common.md: ... here.
29925 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
29927 * config/arm/mve.md (mve_vshlq_<supf><mode>): Move to
29929 * config/arm/neon.md (vashl<mode>3): Delete.
29930 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): New.
29931 (vasl<mode>3): New expander.
29933 2021-01-15 Richard Biener <rguenther@suse.de>
29935 PR tree-optimization/98685
29936 * tree-vect-slp.c (vect_schedule_slp_node): Refactor handling
29937 of vector extern defs.
29939 2021-01-14 David Malcolm <dmalcolm@redhat.com>
29942 * diagnostic.c (diagnostic_kind_text): Break out this array
29944 (diagnostic_build_prefix): ...here.
29945 (fancy_abort): Detect when diagnostic_initialize has not yet been
29946 called and fall back to a minimal implementation of printing the
29947 ICE, rather than segfaulting in internal_error.
29949 2021-01-14 David Malcolm <dmalcolm@redhat.com>
29951 * diagnostic.c (diagnostic_initialize): Eliminate
29952 parseable_fixits_p in favor of initializing extra_output_kind from
29953 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
29954 (convert_column_unit): New function, split out from...
29955 (diagnostic_converted_column): ...this.
29956 (print_parseable_fixits): Add "column_unit" and "tabstop" params.
29957 Use them to call convert_column_unit on the column values.
29958 (diagnostic_report_diagnostic): Eliminate conditional on
29959 parseable_fixits_p in favor of a switch statement on
29960 extra_output_kind, passing the appropriate values to the new
29961 params of print_parseable_fixits.
29962 (selftest::test_print_parseable_fixits_none): Update for new
29963 params of print_parseable_fixits.
29964 (selftest::test_print_parseable_fixits_insert): Likewise.
29965 (selftest::test_print_parseable_fixits_remove): Likewise.
29966 (selftest::test_print_parseable_fixits_replace): Likewise.
29967 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
29969 (selftest::diagnostic_c_tests): Call it.
29970 * diagnostic.h (enum diagnostics_extra_output_kind): New.
29971 (diagnostic_context::parseable_fixits_p): Delete field in favor
29973 (diagnostic_context::extra_output_kind): ...this new field.
29974 * doc/invoke.texi (Environment Variables): Add
29975 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
29976 * opts.c (common_handle_option): Update handling of
29977 OPT_fdiagnostics_parseable_fixits for change to diagnostic_context
29980 2021-01-14 Tamar Christina <tamar.christina@arm.com>
29982 * tree-vect-slp-patterns.c (class complex_operations_pattern,
29983 complex_operations_pattern::matches,
29984 complex_operations_pattern::recognize,
29985 complex_operations_pattern::build): New.
29986 (slp_patterns): Use it.
29988 2021-01-14 Tamar Christina <tamar.christina@arm.com>
29990 * internal-fn.def (COMPLEX_FMS, COMPLEX_FMS_CONJ): New.
29991 * optabs.def (cmls_optab, cmls_conj_optab): New.
29992 * doc/md.texi: Document them.
29993 * tree-vect-slp-patterns.c (class complex_fms_pattern,
29994 complex_fms_pattern::matches, complex_fms_pattern::recognize,
29995 complex_fms_pattern::build): New.
29997 2021-01-14 Tamar Christina <tamar.christina@arm.com>
29999 * internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
30000 * optabs.def (cmla_optab, cmla_conj_optab): New.
30001 * doc/md.texi: Document them.
30002 * tree-vect-slp-patterns.c (vect_match_call_p,
30003 class complex_fma_pattern, vect_slp_reset_pattern,
30004 complex_fma_pattern::matches, complex_fma_pattern::recognize,
30005 complex_fma_pattern::build): New.
30007 2021-01-14 Tamar Christina <tamar.christina@arm.com>
30009 * internal-fn.def (COMPLEX_MUL, COMPLEX_MUL_CONJ): New.
30010 * optabs.def (cmul_optab, cmul_conj_optab): New.
30011 * doc/md.texi: Document them.
30012 * tree-vect-slp-patterns.c (vect_match_call_complex_mla,
30013 vect_normalize_conj_loc, is_eq_or_top, vect_validate_multiplication,
30014 vect_build_combine_node, class complex_mul_pattern,
30015 complex_mul_pattern::matches, complex_mul_pattern::recognize,
30016 complex_mul_pattern::build): New.
30018 2021-01-14 Tamar Christina <tamar.christina@arm.com>
30020 * tree-vect-slp.c (optimize_load_redistribution_1): New.
30021 (optimize_load_redistribution, vect_is_slp_load_node): New.
30022 (vect_match_slp_patterns): Use it.
30024 2021-01-14 Tamar Christina <tamar.christina@arm.com>
30026 * tree-vect-slp-patterns.c (complex_add_pattern::build):
30029 2021-01-14 Thomas Schwinge <thomas@codesourcery.com>
30031 * config/gcn/mkoffload.c (main): Create an offload image only in
30032 64-bit configurations.
30034 2021-01-14 H.J. Lu <hjl.tools@gmail.com>
30037 * config/i386/i386-options.c (ix86_option_override_internal):
30038 Issue an error for -fcf-protection with CF_BRANCH when compiling
30039 for 32-bit non-TARGET_CMOV targets.
30041 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
30044 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p):
30045 Remove declaration and initialization of shadow variable "ret".
30046 (ix86_option_override_internal): Remove delcaration of
30047 shadow variable "i". Redeclare shadowed variable to unsigned.
30048 * common/config/i386/i386-common.c (pta_size): Redeclare to unsigned.
30049 * config/i386/i386-builtins.c (get_builtin_code_for_version):
30050 Update for redeclaration.
30051 * config/i386/i386.h (pta_size): Ditto.
30053 2021-01-14 Richard Biener <rguenther@suse.de>
30055 PR tree-optimization/98674
30056 * tree-data-ref.c (base_supports_access_fn_components_p): New.
30057 (initialize_data_dependence_relation): For two bases without
30058 possible access fns resort to type size equality when determining
30059 shape compatibility.
30061 2021-01-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
30064 * config/arm/arm_neon.h: Replace calls to __builtin_vcge* by
30065 <=, >= operators in vcle and vcge intrinsics respectively.
30066 * config/arm/arm_neon_builtins.def: Remove entry for
30069 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
30072 * config/i386/i386-options.c (ix86_function_specific_save):
30073 Remove redundant assignment to opts->x_ix86_branch_cost.
30074 * config/i386/i386.c (ix86_prefetch_sse):
30075 Rename from x86_prefetch_sse. Update all uses.
30076 * config/i386/i386.h: Update for rename.
30077 * config/i386/i386-options.h: Ditto.
30079 2021-01-14 Jakub Jelinek <jakub@redhat.com>
30082 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3,
30083 *sse4_1_zero_extendv4hiv4si2_3, *sse4_1_zero_extendv2siv2di2_3):
30084 Use Bm instead of m for non-avx. Add isa attribute.
30086 2021-01-14 Jakub Jelinek <jakub@redhat.com>
30088 PR tree-optimization/96688
30089 * match.pd (~(X >> Y) -> ~X >> Y): New simplification if
30090 ~X can be simplified.
30092 2021-01-14 Richard Sandiford <richard.sandiford@arm.com>
30094 * tree-vect-stmts.c (vect_model_load_cost): Account for unused
30095 IFN_LOAD_LANES results.
30097 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30099 * config/aarch64/aarch64-simd.md (aarch64_<su>xtl<mode>):
30101 (aarch64_xtn<mode>): Likewise.
30102 * config/aarch64/aarch64-simd-builtins.def (sxtl, uxtl, xtn):
30105 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
30107 (vmovl_s16): Likewise.
30108 (vmovl_s32): Likewise.
30109 (vmovl_u8): Likewise.
30110 (vmovl_u16): Likewise.
30111 (vmovl_u32): Likewise.
30112 (vmovn_s16): Likewise.
30113 (vmovn_s32): Likewise.
30114 (vmovn_s64): Likewise.
30115 (vmovn_u16): Likewise.
30116 (vmovn_u32): Likewise.
30117 (vmovn_u64): Likewise.
30119 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30121 * config/aarch64/aarch64-simd.md (aarch64_<su>qxtn2<mode>_le):
30123 (aarch64_<su>qxtn2<mode>_be): Likewise.
30124 (aarch64_<su>qxtn2<mode>): Likewise.
30125 * config/aarch64/aarch64-simd-builtins.def (sqxtn2, uqxtn2):
30127 * config/aarch64/iterators.md (SAT_TRUNC): Define code_iterator.
30128 (su): Handle ss_truncate and us_truncate.
30129 * config/aarch64/arm_neon.h (vqmovn_high_s16): Reimplement using
30131 (vqmovn_high_s32): Likewise.
30132 (vqmovn_high_s64): Likewise.
30133 (vqmovn_high_u16): Likewise.
30134 (vqmovn_high_u32): Likewise.
30135 (vqmovn_high_u64): Likewise.
30137 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30139 * config/aarch64/aarch64-simd.md (aarch64_xtn2<mode>_le):
30141 (aarch64_xtn2<mode>_be): Likewise.
30142 (aarch64_xtn2<mode>): Likewise.
30143 * config/aarch64/aarch64-simd-builtins.def (xtn2): Define
30145 * config/aarch64/arm_neon.h (vmovn_high_s16): Reimplement using
30147 (vmovn_high_s32): Likewise.
30148 (vmovn_high_s64): Likewise.
30149 (vmovn_high_u16): Likewise.
30150 (vmovn_high_u32): Likewise.
30151 (vmovn_high_u64): Likewise.
30153 2021-01-13 Stafford Horne <shorne@gmail.com>
30155 * config/or1k/or1k.h (ASM_PREFERRED_EH_DATA_FORMAT): New macro.
30157 2021-01-13 Stafford Horne <shorne@gmail.com>
30159 * config/or1k/linux.h (TARGET_ASM_FILE_END): Define macro.
30161 2021-01-13 Stafford Horne <shorne@gmail.com>
30163 * config/or1k/or1k.h (TARGET_CPU_CPP_BUILTINS): Add builtin
30164 define for __or1k_hard_float__.
30166 2021-01-13 Stafford Horne <shorne@gmail.com>
30168 * config/or1k/or1k.h (NO_PROFILE_COUNTERS): Define as 1.
30169 (PROFILE_HOOK): Define to call _mcount.
30170 (FUNCTION_PROFILER): Change from abort to no-op.
30172 2021-01-13 Jakub Jelinek <jakub@redhat.com>
30174 PR tree-optimization/96691
30175 * match.pd ((~X | C) ^ D -> (X | C) ^ (~D ^ C),
30176 (~X & C) ^ D -> (X & C) ^ (D ^ C)): New simplifications if
30177 (~D ^ C) or (D ^ C) can be simplified.
30179 2021-01-13 Richard Biener <rguenther@suse.de>
30181 PR tree-optimization/92645
30182 * match.pd (BIT_FIELD_REF to conversion): Delay canonicalization
30183 until after vector lowering.
30185 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
30187 * config/aarch64/aarch64-sve.md (fnma<mode>4): Extend from SVE_FULL_I
30189 (@aarch64_pred_fnma<mode>, cond_fnma<mode>, *cond_fnma<mode>_2)
30190 (*cond_fnma<mode>_4, *cond_fnma<mode>_any): Likewise.
30192 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
30194 * config/aarch64/aarch64-sve.md (fma<mode>4): Extend from SVE_FULL_I
30196 (@aarch64_pred_fma<mode>, cond_fma<mode>, *cond_fma<mode>_2)
30197 (*cond_fma<mode>_4, *cond_fma<mode>_any): Likewise.
30199 2021-01-13 Richard Biener <rguenther@suse.de>
30201 PR tree-optimization/92645
30202 * tree-vect-slp.c (vect_build_slp_tree_1): Relax supported
30203 BIT_FIELD_REF argument.
30204 (vect_build_slp_tree_2): Record the desired vector type
30205 on the external vector def.
30206 (vectorizable_slp_permutation): Handle required punning
30207 of existing vector defs.
30209 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
30211 * rtl-ssa/accesses.h (def_lookup): Fix order of comparison results.
30213 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
30215 * config/sh/sh.md (movsf_ie): Remove operands[2] test.
30217 2021-01-13 Samuel Thibault <samuel.thibault@ens-lyon.org>
30219 * config.gcc [$target == *-*-gnu*]: Enable
30220 'default_gnu_indirect_function'.
30222 2021-01-13 Jakub Jelinek <jakub@redhat.com>
30225 * optabs.c (expand_vec_perm_const): Don't force v0 and v1 into
30226 registers before calling targetm.vectorize.vec_perm_const, only after
30228 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const): Handle
30229 two argument permutation when one operand is zero vector and only
30230 after that force operands into registers.
30231 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_1): New
30232 define_insn_and_split pattern.
30233 (*avx512bw_zero_extendv32qiv32hi2_1): Likewise.
30234 (*avx512f_zero_extendv16hiv16si2_1): Likewise.
30235 (*avx2_zero_extendv8hiv8si2_1): Likewise.
30236 (*avx512f_zero_extendv8siv8di2_1): Likewise.
30237 (*avx2_zero_extendv4siv4di2_1): Likewise.
30238 * config/mips/mips.c (mips_vectorize_vec_perm_const): Force operands
30240 * config/arm/arm.c (arm_vectorize_vec_perm_const): Likewise.
30241 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Likewise.
30242 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const): Likewise.
30243 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const): Likewise.
30244 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const): Likewise.
30245 * config/gcn/gcn.c (gcn_vectorize_vec_perm_const): Likewise. Use std::swap.
30247 2021-01-13 Martin Liska <mliska@suse.cz>
30249 PR tree-optimization/98455
30250 * gimple-if-to-switch.cc (condition_info::record_phi_mapping):
30251 Record also virtual PHIs.
30252 (pass_if_to_switch::execute): Return TODO_cleanup_cfg only
30255 2021-01-13 Jonathan Wakely <jwakely@redhat.com>
30257 * doc/invoke.texi (C++ Modules): Fix typos.
30259 2021-01-13 Richard Biener <rguenther@suse.de>
30261 PR tree-optimization/98640
30262 * tree-ssa-sccvn.c (visit_nary_op): Do not try to
30263 handle plus or minus from a truncated operand to be
30266 2021-01-13 Jakub Jelinek <jakub@redhat.com>
30269 * config/i386/i386.md (*btr<mode>_1, *btr<mode>_2): New
30270 define_insn_and_split patterns.
30271 (splitter after *btr<mode>_2): New splitter.
30273 2021-01-13 Martin Liska <mliska@suse.cz>
30276 * cgraphunit.c (analyze_functions): Remove dead code.
30278 2021-01-13 Qian Jianhua <qianjh@cn.fujitsu.com>
30280 * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
30281 * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
30282 (a64fx_regmove_cost, a64fx_vector_cost): New.
30283 (a64fx_tunings): Use the new added cost tables.
30285 2021-01-13 Jakub Jelinek <jakub@redhat.com>
30288 * config/i386/predicates.md (pmovzx_parallel): New predicate.
30289 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): New
30290 define_insn_and_split pattern.
30291 (*sse4_1_zero_extendv4hiv4si2_3): Likewise.
30292 (*sse4_1_zero_extendv2siv2di2_3): Likewise.
30294 2021-01-13 Julian Brown <julian@codesourcery.com>
30296 * config/gcn/gcn.c (gcn_conditional_register_usage): Remove dead code
30297 to fix v0 register.
30299 2021-01-13 Julian Brown <julian@codesourcery.com>
30301 * config/gcn/gcn.c (gcn_md_reorg): Fix case where EXEC reg is live
30304 2021-01-13 Julian Brown <julian@codesourcery.com>
30306 * config/gcn/gcn-valu.md (recip<mode>2<exec>, recip<mode>2): Use unspec
30307 for reciprocal-approximation instructions.
30308 (div<mode>3): Use fused multiply-accumulate operations for reciprocal
30309 refinement and division result.
30310 * config/gcn/gcn.md (UNSPEC_RCP): New unspec constant.
30312 2021-01-13 Julian Brown <julian@codesourcery.com>
30314 * config/gcn/gcn-valu.md (subdf): Rename to...
30317 2021-01-12 Martin Liska <mliska@suse.cz>
30319 * gcov.c (source_info::debug): Fix printf format for 32-bit hosts.
30321 2021-01-12 Andrea Corallo <andrea.corallo@arm.com>
30323 * function-abi.h: Fix typo.
30325 2021-01-12 Christophe Lyon <christophe.lyon@linaro.org>
30329 * config/arm/arm.h (ARM_HAVE_NEON_V8QI_LDST): New macro.
30330 (ARM_HAVE_NEON_V16QI_LDST, ARM_HAVE_NEON_V4HI_LDST): Likewise.
30331 (ARM_HAVE_NEON_V8HI_LDST, ARM_HAVE_NEON_V2SI_LDST): Likewise.
30332 (ARM_HAVE_NEON_V4SI_LDST, ARM_HAVE_NEON_V4HF_LDST): Likewise.
30333 (ARM_HAVE_NEON_V8HF_LDST, ARM_HAVE_NEON_V4BF_LDST): Likewise.
30334 (ARM_HAVE_NEON_V8BF_LDST, ARM_HAVE_NEON_V2SF_LDST): Likewise.
30335 (ARM_HAVE_NEON_V4SF_LDST, ARM_HAVE_NEON_DI_LDST): Likewise.
30336 (ARM_HAVE_NEON_V2DI_LDST): Likewise.
30337 (ARM_HAVE_V8QI_LDST, ARM_HAVE_V16QI_LDST): Likewise.
30338 (ARM_HAVE_V4HI_LDST, ARM_HAVE_V8HI_LDST): Likewise.
30339 (ARM_HAVE_V2SI_LDST, ARM_HAVE_V4SI_LDST, ARM_HAVE_V4HF_LDST): Likewise.
30340 (ARM_HAVE_V8HF_LDST, ARM_HAVE_V4BF_LDST, ARM_HAVE_V8BF_LDST): Likewise.
30341 (ARM_HAVE_V2SF_LDST, ARM_HAVE_V4SF_LDST, ARM_HAVE_DI_LDST): Likewise.
30342 (ARM_HAVE_V2DI_LDST): Likewise.
30343 * config/arm/mve.md (*movmisalign<mode>_mve_store): New pattern.
30344 (*movmisalign<mode>_mve_load): New pattern.
30345 * config/arm/neon.md (movmisalign<mode>): Move to ...
30346 * config/arm/vec-common.md: ... here.
30348 2021-01-12 Vladimir N. Makarov <vmakarov@redhat.com>
30351 * lra-eliminations.c (eliminate_regs_in_insn): Add transformation
30352 of pattern 'plus (plus (hard reg, const), pseudo)'.
30354 2021-01-12 Richard Biener <rguenther@suse.de>
30356 PR tree-optimization/98550
30357 * tree-vect-slp.c (vect_record_max_nunits): Check whether
30358 the group size is a multiple of the vector element count.
30359 (vect_build_slp_tree_1): When we need to fail because
30360 the vector type choosen causes unrolling do so lazily
30361 without affecting matches only at the end to guide group splitting.
30363 2021-01-12 Martin Liska <mliska@suse.cz>
30366 * optc-save-gen.awk: Compare also n_target_save vars with
30369 2021-01-12 Martin Liska <mliska@suse.cz>
30371 * gcov.c (source_info::debug): New.
30372 (print_usage): Add --debug (-D) option.
30373 (process_args): Likewise.
30374 (generate_results): Call src->debug after
30375 accumulate_line_counts.
30376 (read_graph_file): Properly assign id for EXIT_BLOCK.
30377 * profile.c (branch_prob): Dump function body before it is
30380 2021-01-12 Jakub Jelinek <jakub@redhat.com>
30382 PR tree-optimization/98629
30383 * tree-ssa-math-opts.c (arith_overflow_check_p): Don't update use_stmt
30384 unless returning non-zero.
30386 2021-01-12 Jakub Jelinek <jakub@redhat.com>
30388 PR tree-optimization/95731
30389 * tree-ssa-reassoc.c (optimize_range_tests_cmp_bitwise): Also optimize
30390 x < 0 && y < 0 && z < 0 into (x | y | z) < 0 for signed x, y, z.
30391 (optimize_range_tests): Call optimize_range_tests_cmp_bitwise
30392 only after optimize_range_tests_var_bound.
30394 2021-01-12 Jakub Jelinek <jakub@redhat.com>
30396 * configure.ac: Ensure c/Make-lang.in comes first in @all_lang_makefrags@.
30397 * configure: Regenerated.
30399 2021-01-12 liuhongt <hongtao.liu@intel.com>
30402 * config/i386/i386-builtins.h (BUILTIN_DESC_SWAP_OPERANDS):
30404 * config/i386/i386-expand.c (ix86_expand_sse_comi): Delete
30407 2021-01-12 Alexandre Oliva <oliva@adacore.com>
30409 * ssa-iterators.h (end_imm_use_stmt_traverse): Forward
30411 (auto_end_imm_use_stmt_traverse): New struct.
30412 (FOR_EACH_IMM_USE_STMT): Use it.
30413 (BREAK_FROM_IMM_USE_STMT, RETURN_FROM_IMM_USE_STMT): Remove,
30415 * gimple-ssa-strength-reduction.c: ... here, ...
30416 * graphite-scop-detection.c: ... here, ...
30417 * ipa-modref.c, ipa-pure-const.c, ipa-sra.c: ... here, ...
30418 * tree-predcom.c, tree-ssa-ccp.c: ... here, ...
30419 * tree-ssa-dce.c, tree-ssa-dse.c: ... here, ...
30420 * tree-ssa-loop-ivopts.c, tree-ssa-math-opts.c: ... here, ...
30421 * tree-ssa-phiprop.c, tree-ssa.c: ... here, ...
30422 * tree-vect-slp.c: ... and here, ...
30423 * doc/tree-ssa.texi: ... and the example here.
30425 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
30427 * config/aarch64/aarch64-sve.md (sdiv_pow2<mode>3): Extend from
30428 SVE_FULL_I to SVE_I. Generate an UNSPEC_PRED_X.
30429 (*sdiv_pow2<mode>3): New pattern.
30430 (@cond_<sve_int_op><mode>): Extend from SVE_FULL_I to SVE_I.
30431 Wrap the ASRD in an UNSPEC_PRED_X.
30432 (*cond_<sve_int_op><mode>_2): Likewise. Replace the UNSPEC_PRED_X
30433 predicate with a constant PTRUE, if it isn't already.
30434 (*cond_<sve_int_op><mode>_z): Replace with...
30435 (*cond_<sve_int_op><mode>_any): ...this new pattern.
30437 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
30439 * config/aarch64/aarch64-sve.md (*cond_bic<mode>_2): Extend from
30440 SVE_FULL_I to SVE_I.
30441 (*cond_bic<mode>_any): Likewise.
30443 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
30445 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart)
30446 (@aarch64_pred_<MUL_HIGHPART:optab><mode>): Extend from SVE_FULL_I
30449 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
30451 * config/aarch64/aarch64-sve.md (<su>abd<mode>_3): Extend from
30452 SVE_FULL_I to SVE_I.
30453 (*aarch64_cond_<su>abd<mode>_2): Likewise.
30454 (*aarch64_cond_<su>abd<mode>_any): Likewise.
30455 (@aarch64_pred_<su>abd<mode>): Likewise. Use UNSPEC_PRED_X
30456 for the max and min but not for the minus.
30457 (*aarch64_cond_<su>abd<mode>_3): New pattern.
30459 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
30461 * config/aarch64/iterators.md (SVE_24I): New iterator.
30462 * config/aarch64/aarch64-sve.md (*aarch64_adr<mode>_shift): Extend from
30463 SVE_FULL_SDI to SVE_24I. Use containers rather than elements.
30465 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
30467 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_BINARY:optab><mode>)
30468 (*cond_<SVE_INT_BINARY:optab><mode>_2): Extend from SVE_FULL_I
30470 (*cond_<SVE_INT_BINARY:optab><mode>_3): Likewise.
30471 (*cond_<SVE_INT_BINARY:optab><mode>_any): Likewise.
30472 (*cond_<SVE_INT_BINARY:optab><mode>_2_const): Likewise.
30473 (*cond_<SVE_INT_BINARY:optab><mode>_any_const): Likewise.
30475 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
30477 * config/aarch64/aarch64-sve.md (<SVE_INT_BINARY_IMM:optab><mode>3)
30478 (@aarch64_pred_<SVE_INT_BINARY_IMM:optab><mode>)
30479 (*post_ra_<SVE_INT_BINARY_IMM:optab><mode>3): Extend from SVE_FULL_I
30482 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
30484 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3)
30485 (v<ASHIFT:optab><mode>3, @aarch64_pred_<optab><mode>)
30486 (*post_ra_v<ASHIFT:optab><mode>3): Extend from SVE_FULL_I to SVE_I.
30488 2021-01-11 Martin Liska <mliska@suse.cz>
30491 * symtab-clones.h (clone_info::release): Release
30492 symtab::m_clones with ggc_delete as it's a GGC memory.
30494 2021-01-11 Matthias Klose <doko@ubuntu.com>
30496 * Makefile.in (LINK_PROGRESS): Show the link target.
30498 2021-01-11 Richard Biener <rguenther@suse.de>
30500 PR tree-optimization/91403
30501 * tree-vect-data-refs.c (vect_analyze_group_access_1): Cap
30502 single-element interleaving group size at 4096 elements.
30504 2021-01-11 Richard Biener <rguenther@suse.de>
30506 PR tree-optimization/98526
30507 * tree-vect-loop.c (vect_model_reduction_cost): Remove costing
30508 of the actual reduction op for the regular case.
30509 (vectorizable_reduction): Cost the stmts
30510 vect_transform_reduction produces here.
30512 2021-01-11 Andreas Krebbel <krebbel@linux.ibm.com>
30514 * tree-ssa-forwprop.c (simplify_vector_constructor): For
30515 big-endian, use UNPACK[_FLOAT]_HI.
30517 2021-01-11 Tamar Christina <tamar.christina@arm.com>
30519 * tree-vect-slp-patterns.c (class complex_pattern,
30520 class complex_add_pattern): Add parameters to matches.
30521 (complex_add_pattern::build): Free memory.
30522 (complex_add_pattern::matches): Move validation end of match.
30523 (complex_add_pattern::recognize): Likewise.
30525 2021-01-11 Tamar Christina <tamar.christina@arm.com>
30527 * tree-vect-slp-patterns.c (linear_loads_p): Fix externals.
30529 2021-01-11 Tamar Christina <tamar.christina@arm.com>
30531 * tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
30533 2021-01-11 Jakub Jelinek <jakub@redhat.com>
30535 PR tree-optimization/95867
30536 * tree-ssa-math-opts.h: New header.
30537 * tree-ssa-math-opts.c: Include tree-ssa-math-opts.h.
30538 (powi_as_mults): No longer static. Use build_one_cst instead of
30539 build_real. Formatting fix.
30540 * tree-ssa-reassoc.c: Include tree-ssa-math-opts.h.
30541 (attempt_builtin_powi): Handle multiplication reassociation without
30542 powi_fndecl using powi_as_mults.
30543 (reassociate_bb): For integral types don't require
30544 -funsafe-math-optimizations to call attempt_builtin_powi.
30546 2021-01-11 Jakub Jelinek <jakub@redhat.com>
30548 PR tree-optimization/95852
30549 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): Change
30550 mul_stmts parameter type to vec<gimple *> &. Before cond_stmt
30551 allow in the bb any of the stmts in that vector, div_stmt and
30552 up to 3 cast stmts.
30553 (arith_cast_equal_p): New function.
30554 (arith_overflow_check_p): Add cast_stmt argument, handle signed
30555 multiply overflow checks.
30556 (match_arith_overflow): Adjust caller. Handle signed multiply
30559 2021-01-11 Jakub Jelinek <jakub@redhat.com>
30561 PR tree-optimization/95852
30562 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): New function.
30563 (uaddsub_overflow_check_p): Renamed to ...
30564 (arith_overflow_check_p): ... this. Handle also multiplication
30565 with overflow check.
30566 (match_uaddsub_overflow): Renamed to ...
30567 (match_arith_overflow): ... this. Add cfg_changed argument. Handle
30568 also multiplication with overflow check. Adjust function comment.
30569 (math_opts_dom_walker::after_dom_children): Adjust callers. Call
30570 match_arith_overflow also for MULT_EXPR.
30572 2021-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30574 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
30575 __builtin_convertvector.
30576 (vmovl_s16): Likewise.
30577 (vmovl_s32): Likewise.
30578 (vmovl_u8): Likewise.
30579 (vmovl_u16): Likewise.
30580 (vmovl_u32): Likewise.
30581 (vmovn_s16): Likewise.
30582 (vmovn_s32): Likewise.
30583 (vmovn_s64): Likewise.
30584 (vmovn_u16): Likewise.
30585 (vmovn_u32): Likewise.
30586 (vmovn_u64): Likewise.
30588 2021-01-11 Martin Liska <mliska@suse.cz>
30590 * gimple-if-to-switch.cc (struct condition_info): Use auto_var.
30591 (if_chain::is_beneficial): Delete clusters
30592 (find_conditions): Make second argument of conditions_in_bbs a
30593 pointer so that we control over it's lifetime.
30594 (pass_if_to_switch::execute): Delete them.
30596 2021-01-11 Kewen Lin <linkw@linux.ibm.com>
30598 * ira.c (move_unallocated_pseudos): Check other_reg and skip if
30601 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
30603 * config/vax/vax.md (cc): Remove mode attribute.
30604 (subst_<cc>, subst_f<cc>): Rename to...
30605 (subst_<mode>, subst_f<VAXccnz:mode>): ... these respectively.
30606 (*cbranch<VAXint:mode>4_<VAXcc:mode>): Update for `cc' removal.
30607 (*cbranch<VAXfp:mode>4_<VAXccnz:mode>): Likewise.
30608 (*branch_<mode>, *branch_<mode>_reversed): Likewise.
30610 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
30612 * config/vax/vax.md (subst_f<cc>): Add mode to operands and
30613 `const_double_zero'.
30615 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
30617 * config/pdp11/pdp11.md (PDPfp): New mode iterator.
30618 (fcc_cc, fcc_ccnz): Use it. Add mode to `const_double_zero' and
30621 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
30623 * genemit.c (gen_exp) <CONST_DOUBLE>: Handle `const_double_zero'
30625 * read-rtl.c (rtx_reader::read_rtx_code): Handle machine mode
30626 with `const_double_zero'.
30627 * doc/rtl.texi (Constant Expression Types): Document it.
30629 2021-01-09 Jakub Jelinek <jakub@redhat.com>
30632 * tree-cfg.c (verify_gimple_assign_binary): Allow lhs of
30633 POINTER_DIFF_EXPR to be any integral type.
30635 2021-01-09 Jakub Jelinek <jakub@redhat.com>
30637 PR rtl-optimization/98603
30638 * function.c (instantiate_virtual_regs_in_insn): For asm goto
30639 with impossible constraints, drop all SETs, CLOBBERs, drop PARALLEL
30640 if any, set ASM_OPERANDS mode to VOIDmode and change
30641 ASM_OPERANDS_OUTPUT_CONSTRAINT and ASM_OPERANDS_OUTPUT_IDX.
30643 2021-01-09 Alexandre Oliva <oliva@gnu.org>
30646 * final.c (notice_source_line): Narrow down the condition to
30647 skip a line-0 marker.
30649 2021-01-08 Sergei Trofimovich <siarheit@google.com>
30651 * ipa-modref.c (merge_call_side_effects): Fix
30652 linebreak split by reordering two print calls.
30654 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
30656 * config/s390/vector.md (*tf_to_fprx2_0): Rename from
30657 "*mov_tf_to_fprx2_0" for consistency, fix constraint.
30658 (*tf_to_fprx2_1): Rename from "*mov_tf_to_fprx2_1" for
30659 consistency, fix constraint.
30661 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
30663 * config/s390/s390-c.c (s390_def_or_undef_macro): Accept
30664 callables instead of mask values.
30665 (struct target_flag_set_p): New predicate.
30666 (s390_cpu_cpp_builtins_internal): Define or undefine
30667 __LONG_DOUBLE_VX__ macro.
30669 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
30672 * config/i386/i386.c (x86_function_profiler): Use R10 and R11
30673 to call mcount in large model with PIC for NO_PROFILE_COUNTERS
30676 2021-01-08 Richard Biener <rguenther@suse.de>
30678 * tree-ssa-sccvn.c (pass_fre::execute): Reset the SCEV hash table.
30680 2021-01-08 Richard Biener <rguenther@suse.de>
30682 * tree-vect-slp.c (scalar_stmts_to_slp_tree_map_t): Fix.
30683 (vect_build_slp_tree): On cache hit release the matched
30684 scalar stmts vector.
30685 * tree-vect-stmts.c (vectorizable_store): Properly free
30686 vec_oprnds before possibly gathering them again.
30688 2021-01-08 Richard Biener <rguenther@suse.de>
30690 PR tree-optimization/98544
30691 * tree-vect-slp.c (vect_optimize_slp): Always materialize
30692 permutes at a permute node.
30694 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
30697 * config/i386/i386.c (x86_function_profiler): Use R10 to call
30698 mcount in large model. Sorry for large model with PIC.
30700 2021-01-08 Jakub Jelinek <jakub@redhat.com>
30703 * config/i386/i386.opt (ix86_cmodel, ix86_incoming_stack_boundary_arg,
30704 ix86_pmode, ix86_preferred_stack_boundary_arg, ix86_regparm,
30705 ix86_veclibabi_type): Remove x_ prefix, use TargetVariable instead of
30706 TargetSave and initialize for variables with enum types.
30707 (mfentry, mstack-protector-guard-reg=, mstack-protector-guard-offset=,
30708 mstack-protector-guard-symbol=): Add Save.
30709 * config/i386/i386-options.c (ix86_function_specific_save,
30710 ix86_function_specific_restore): Don't save or restore x_ix86_cmodel,
30711 x_ix86_incoming_stack_boundary_arg, x_ix86_pmode,
30712 x_ix86_preferred_stack_boundary_arg, x_ix86_regparm,
30713 x_ix86_veclibabi_type.
30715 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
30717 * config/aarch64/aarch64-sve.md (*cnot<mode>): Extend from
30718 SVE_FULL_I to SVE_I.
30719 (*cond_cnot<mode>_2, *cond_cnot<mode>_any): Likewise.
30721 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
30723 * config/aarch64/aarch64-sve.md (*cond_uxt<mode>_2): Extend from
30724 SVE_FULL_I to SVE_I.
30725 (*cond_uxt<mode>_any): Likewise.
30727 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30729 * config/aarch64/iterators.md (Vwhalf): New iterator.
30730 * config/aarch64/aarch64-simd.md (aarch64_<sur>adalp<mode>_3):
30732 (aarch64_<sur>adalp<mode>): ... This. Make more
30734 (<sur>sadv16qi): Adjust callsite of the above.
30735 * config/aarch64/aarch64-simd-builtins.def (sadalp, uadalp): New
30737 * config/aarch64/arm_neon.h (vpadal_s8): Reimplement using
30739 (vpadal_s16): Likewise.
30740 (vpadal_u8): Likewise.
30741 (vpadal_u16): Likewise.
30742 (vpadalq_s8): Likewise.
30743 (vpadalq_s16): Likewise.
30744 (vpadalq_s32): Likewise.
30745 (vpadalq_u8): Likewise.
30746 (vpadalq_u16): Likewise.
30747 (vpadalq_u32): Likewise.
30749 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30751 * config/aarch64/aarch64-simd.md (aarch64_<su>abd<mode>_3):
30753 (aarch64_<su>abd<mode>): ... This.
30754 (<sur>sadv16qi): Adjust callsite of the above.
30755 * config/aarch64/aarch64-simd-builtins.def (sabd, uabd): Define
30757 * config/aarch64/arm_neon.h (vabd_s8): Reimplement using
30759 (vabd_s16): Likewise.
30760 (vabd_s32): Likewise.
30761 (vabd_u8): Likewise.
30762 (vabd_u16): Likewise.
30763 (vabd_u32): Likewise.
30764 (vabdq_s8): Likewise.
30765 (vabdq_s16): Likewise.
30766 (vabdq_s32): Likewise.
30767 (vabdq_u8): Likewise.
30768 (vabdq_u16): Likewise.
30769 (vabdq_u32): Likewise.
30771 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30773 * config/aarch64/aarch64-simd-builtins.def (saba, uaba): Define
30775 * config/aarch64/arm_neon.h (vaba_s8): Implement using builtin.
30776 (vaba_s16): Likewise.
30777 (vaba_s32): Likewise.
30778 (vaba_u8): Likewise.
30779 (vaba_u16): Likewise.
30780 (vaba_u32): Likewise.
30781 (vabaq_s8): Likewise.
30782 (vabaq_s16): Likewise.
30783 (vabaq_s32): Likewise.
30784 (vabaq_u8): Likewise.
30785 (vabaq_u16): Likewise.
30786 (vabaq_u32): Likewise.
30788 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30790 * config/aarch64/aarch64-simd.md (aba<mode>_3): Rename to...
30791 (aarch64_<su>aba<mode>): ... This. Handle uaba as well.
30792 Change RTL pattern to match.
30794 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
30796 * common/config/riscv/riscv-common.c (riscv_current_subset_list): New.
30797 * config/riscv/riscv-c.c (riscv-subset.h): New.
30798 (INCLUDE_STRING): Define.
30799 (riscv_cpu_cpp_builtins): Add new style architecture extension
30801 * config/riscv/riscv-subset.h (riscv_subset_list::begin): New.
30802 (riscv_subset_list::end): New.
30803 (riscv_current_subset_list): New.
30805 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
30807 * common/config/riscv/riscv-common.c (RISCV_DONT_CARE_VERSION):
30808 Move to riscv-subset.h.
30809 (struct riscv_subset_t): Ditto.
30810 (class riscv_subset_list): Ditto.
30811 * config/riscv/riscv-subset.h (RISCV_DONT_CARE_VERSION): Move
30812 from riscv-common.c.
30813 (struct riscv_subset_t): Ditto.
30814 (class riscv_subset_list): Ditto.
30815 * config/riscv/t-riscv ($(common_out_file)): Add file
30818 2021-01-07 Jakub Jelinek <jakub@redhat.com>
30821 * config/i386/i386.md (*bmi_blsi_<mode>_cmp, *bmi_blsi_<mode>_ccno):
30822 New define_insn patterns.
30824 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
30826 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_UNARY:optab><mode>)
30827 (*cond_<SVE_INT_UNARY:optab><mode>_2): Extend from SVE_FULL_I to SVE_I.
30828 (*cond_<SVE_INT_UNARY:optab><mode>_any): Likewise.
30830 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
30832 PR tree-optimization/98560
30833 * internal-fn.def (IFN_VCONDU, IFN_VCONDEQ): Use type vec_cond.
30834 * internal-fn.c (vec_cond_mask_direct): Get the data mode from
30836 (vec_cond_direct): Likewise argument 2.
30837 (vec_condu_direct, vec_condeq_direct): Delete.
30838 (expand_vect_cond_optab_fn): Rename to...
30839 (expand_vec_cond_optab_fn): ...this, replacing old macro.
30840 (expand_vec_condu_optab_fn, expand_vec_condeq_optab_fn): Delete.
30841 (expand_vect_cond_mask_optab_fn): Rename to...
30842 (expand_vec_cond_mask_optab_fn): ...this, replacing old macro.
30843 (direct_vec_cond_mask_optab_supported_p): Treat the optab as a
30845 (direct_vec_cond_optab_supported_p): Likewise.
30846 (direct_vec_condu_optab_supported_p): Delete.
30847 (direct_vec_condeq_optab_supported_p): Delete.
30848 * gimple-isel.cc: Include internal-fn.h.
30849 (gimple_expand_vec_cond_expr): Check that IFN_VCONDEQ is supported
30852 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
30854 PR tree-optimization/98560
30855 * gimple-isel.cc (gimple_expand_vec_cond_expr): If we fail to use
30856 IFN_VCOND{,U,EQ}, fall back on IFN_VCOND_MASK.
30858 2021-01-07 Uroš Bizjak <ubizjak@gmail.com>
30860 * config/i386/i386.md (insn): Merge from plusminus_insn, shift_insn,
30861 rotate_insn and optab code attributes.
30862 Update all uses to merged code attribute.
30863 * config/i386/sse.md: Update all uses to merged code attribute.
30864 * config/i386/mmx.md: Update all uses to merged code attribute.
30866 2021-01-07 Jakub Jelinek <jakub@redhat.com>
30868 PR tree-optimization/98568
30869 * gimple-ssa-store-merging.c (bswap_view_convert): New function.
30870 (bswap_replace): Use it.
30872 2021-01-06 Vladimir N. Makarov <vmakarov@redhat.com>
30874 PR rtl-optimization/97978
30875 * lra-int.h (lra_hard_reg_split_p): New external.
30876 * lra.c (lra_hard_reg_split_p): New global.
30877 (lra): Set up lra_hard_reg_split_p after splitting a hard reg.
30878 * lra-assigns.c (lra_assign): Don't check allocation correctness
30879 after hard reg splitting.
30881 2021-01-06 Martin Sebor <msebor@redhat.com>
30884 * builtins.c (new_delete_mismatch_p): New overload.
30885 (new_delete_mismatch_p (tree, tree)): Call it.
30887 2021-01-06 Alexandre Oliva <oliva@adacore.com>
30889 * Makefile.in (T_GLIMITS_H): New.
30890 (stmp-int-hdrs): Depend on it, use it.
30891 * config/t-vxworks (T_GLIMITS_H): Override it.
30892 (vxw-glimits.h): New.
30894 2021-01-06 Richard Biener <rguenther@suse.de>
30896 PR tree-optimization/98513
30897 * value-range.cc (intersect_ranges): Compare the upper bounds
30898 for the expected relation.
30900 2021-01-06 Gerald Pfeifer <gerald@pfeifer.com>
30903 2020-12-28 Gerald Pfeifer <gerald@pfeifer.com>
30905 * doc/standards.texi (HSAIL): Remove section.
30907 2021-01-05 Samuel Thibault <samuel.thibault@ens-lyon.org>
30909 * configure: Re-generate.
30911 2021-01-05 Jakub Jelinek <jakub@redhat.com>
30913 * doc/invoke.texi (-std=c++20): Adjust for the publication of
30914 ISO 14882:2020 standard.
30915 * doc/standards.texi: Likewise.
30917 2021-01-05 Jakub Jelinek <jakub@redhat.com>
30919 PR tree-optimization/94802
30920 * expr.h (maybe_optimize_sub_cmp_0): Declare.
30921 * expr.c: Include tree-pretty-print.h and flags.h.
30922 (maybe_optimize_sub_cmp_0): New function.
30923 (do_store_flag): Use it.
30924 * cfgexpand.c (expand_gimple_cond): Likewise.
30926 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
30928 * mux-utils.h (pointer_mux::m_ptr): Tweak description of contents.
30929 * rtlanal.c (simple_regno_set): Tweak description to clarify the
30932 2021-01-05 Richard Biener <rguenther@suse.de>
30934 PR tree-optimization/98516
30935 * tree-vect-slp.c (vect_optimize_slp): Permute the incoming
30936 lanes when materializing on a VEC_PERM node.
30937 (vectorizable_slp_permutation): Dump the permute properly.
30939 2021-01-05 Richard Biener <rguenther@suse.de>
30941 * tree-vect-slp.c (vect_slp_region): Move debug counter
30942 to cover individual subgraphs.
30944 2021-01-05 Richard Biener <rguenther@suse.de>
30946 PR tree-optimization/98428
30947 * tree-vect-slp.c (vect_build_slp_tree_1): Properly reject
30948 vector lane extracts for loop vectorization.
30950 2021-01-05 Jakub Jelinek <jakub@redhat.com>
30952 PR tree-optimization/98514
30953 * tree-ssa-reassoc.c (bb_rank): Change type from long * to
30955 (operand_rank): Change type from hash_map<tree, long> to
30956 hash_map<tree, int64_t>.
30957 (phi_rank): Change return type from long to int64_t.
30958 (loop_carried_phi): Change block_rank variable type from long to
30960 (propagate_rank): Change return type, rank parameter type and
30961 op_rank variable type from long to int64_t.
30962 (find_operand_rank): Change return type from long to int64_t
30963 and change slot variable type from long * to int64_t *.
30964 (insert_operand_rank): Change rank parameter type from long to
30966 (get_rank): Change return type and rank variable type from long to
30967 int64_t. Use PRId64 instead of ld to print the rank.
30968 (init_reassoc): Change rank variable type from long to int64_t
30969 and adjust correspondingly bb_rank and operand_rank initialization.
30971 2021-01-05 Jakub Jelinek <jakub@redhat.com>
30973 PR tree-optimization/96928
30974 * tree-ssa-phiopt.c (xor_replacement): New function.
30975 (tree_ssa_phiopt_worker): Call it.
30977 2021-01-05 Jakub Jelinek <jakub@redhat.com>
30979 PR tree-optimization/96930
30980 * match.pd ((A / (1 << B)) -> (A >> B)): If A is extended
30981 from narrower value which has the same type as 1 << B, perform
30982 the right shift on the narrower value followed by extension.
30984 2021-01-05 Jakub Jelinek <jakub@redhat.com>
30986 PR tree-optimization/96239
30987 * gimple-ssa-store-merging.c (maybe_optimize_vector_constructor): New
30989 (get_status_for_store_merging): Don't return BB_INVALID for blocks
30990 with potential bswap optimizable CONSTRUCTORs.
30991 (pass_store_merging::execute): Optimize vector CONSTRUCTORs with bswap
30994 2021-01-05 Richard Biener <rguenther@suse.de>
30996 PR tree-optimization/98381
30997 * tree.c (vector_element_bits): Properly compute bool vector
30999 * tree-vect-loop.c (vectorizable_live_operation): Properly
31000 compute the last lane bit offset.
31002 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
31005 * config/i386/sse.md (sse_cvtps2pi): Redefine as define_insn_and_split.
31006 Clear the top 64 bytes of the input XMM register.
31007 (sse_cvttps2pi): Ditto.
31009 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
31012 * config/i386/xopintrin.h (_mm256_cmov_si256): New.
31014 2021-01-05 H.J. Lu <hjl.tools@gmail.com>
31017 * config/i386/xmmintrin.h (_mm_extract_pi16): Cast to unsigned
31020 2021-01-05 Claudiu Zissulescu <claziss@synopsys.com>
31022 * config/arc/arc.md (maddsidi4_split): Use ACC_REG_FIRST.
31023 (umaddsidi4_split): Likewise.
31025 2021-01-05 liuhongt <hongtao.liu@intel.com>
31028 * config/i386/sse.md (*sse2_pmovskb_zexthisi): New
31029 define_insn_and_split for zero_extend of subreg HI of pmovskb
31031 (*sse2_pmovskb_zexthisi): Add new combine splitters for
31032 zero_extend of not of subreg HI of pmovskb result.
31034 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
31037 * explow.c (convert_memory_address_addr_space_1): Handle UNSPECs
31039 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate): Use
31040 convert_memory_address to convert symbolic immediates to ptr_mode
31041 before forcing them to memory.
31043 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
31045 PR rtl-optimization/97144
31046 * recog.c (constrain_operands): Initialize matching_operand
31047 for each alternative, rather than only doing it once.
31049 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
31051 PR rtl-optimization/98403
31052 * rtl-ssa/changes.cc (function_info::finalize_new_accesses): Explain
31053 why we don't remove call clobbers.
31054 (function_info::apply_changes_to_insn): Don't attempt to add
31055 call clobbers here.
31057 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
31059 PR tree-optimization/98371
31060 * tree-vect-loop.c (vect_reanalyze_as_main_loop): New function.
31061 (vect_analyze_loop): If an epilogue loop appears to be cheaper
31062 than the main loop, re-analyze it as a main loop before adopting
31065 2021-01-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
31068 * configure.ac (NETLIBS): Determine using AX_LIB_SOCKET_NSL.
31069 * aclocal.m4, configure: Regenerate.
31070 * Makefile.in (NETLIBS): Define.
31071 (BACKEND): Remove $(CODYLIB).
31073 2021-01-05 Jakub Jelinek <jakub@redhat.com>
31075 PR rtl-optimization/98334
31076 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
31077 Optimize (X - 1) * Y + Y to X * Y or (X + 1) * Y - Y to X * Y.
31079 2021-01-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
31081 * tree-inline.c (expand_call_inline): Restore input_location.
31082 Return result from recursive call.
31084 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
31086 PR tree-optimization/95401
31087 * config/aarch64/aarch64-sve-builtins.cc
31088 (gimple_folder::load_store_cookie): Use bits rather than bytes
31089 for the alignment argument to IFN_MASK_LOAD and IFN_MASK_STORE.
31090 * gimple-fold.c (gimple_fold_mask_load_store_mem_ref): Likewise.
31091 * tree-vect-stmts.c (vectorizable_store): Likewise.
31092 (vectorizable_load): Likewise.
31094 2021-01-04 Richard Biener <rguenther@suse.de>
31096 PR tree-optimization/98308
31097 * tree-vect-stmts.c (vectorizable_load): Set invariant mask
31100 2021-01-04 Jakub Jelinek <jakub@redhat.com>
31102 PR tree-optimization/95771
31103 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Handle types
31104 with precision smaller than int's precision and types with precision
31105 twice as large as long long. Formatting fixes.
31107 2021-01-04 Richard Biener <rguenther@suse.de>
31109 PR tree-optimization/98464
31110 * tree-ssa-sccvn.c (vn_valueize_for_srt): Rename from ...
31111 (vn_valueize_wrapper): ... this. Temporarily adjust vn_context_bb.
31112 (process_bb): Adjust.
31114 2021-01-04 Matthew Malcomson <matthew.malcomson@arm.com>
31117 * doc/invoke.texi (-fsanitize=address): Fix wording describing
31118 clash with -fsanitize=hwaddress.
31120 2021-01-04 Richard Biener <rguenther@suse.de>
31122 PR tree-optimization/98282
31123 * tree-ssa-sccvn.c (vn_get_stmt_kind): Classify tcc_reference on
31124 invariants as VN_NARY.
31126 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
31129 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Accept
31130 aarch64_simd_reg_or_zero for operand 2. Use the combinez patterns
31131 to handle zero operands.
31133 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
31135 * config/aarch64/aarch64.c (offset_6bit_signed_scaled_p): New function.
31136 (offset_6bit_unsigned_scaled_p): Fix typo in comment.
31137 (aarch64_sve_prefetch_operand_p): Accept MUL VLs in the range
31140 2021-01-04 Richard Biener <rguenther@suse.de>
31142 PR tree-optimization/98393
31143 * tree-vect-slp.c (vect_build_slp_tree): Properly zero matches
31144 when hitting the limit.
31146 2021-01-04 Richard Biener <rguenther@suse.de>
31148 PR tree-optimization/98291
31149 * tree-vect-loop.c (vectorizable_reduction): Bypass
31150 associativity check for SLP reductions with VF 1.
31152 2021-01-04 Jakub Jelinek <jakub@redhat.com>
31154 PR tree-optimization/96782
31155 * match.pd (x == ~x -> false, x != ~x -> true): New simplifications.
31157 2021-01-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
31159 * collect-utils.c (collect_execute): Check dumppfx.
31160 * collect2.c (maybe_run_lto_and_relink, do_link): Pass atsuffix
31161 to collect_execute.
31162 (do_link): Add new parameter atsuffix.
31163 (main): Handle -dumpdir option. Skip one argument for
31164 -o, -isystem and -B options.
31165 * gcc.c (make_at_file): New helper function.
31166 (close_at_file): Use it.
31168 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
31170 * config/darwin.h (MIN_LD64_NO_COAL_SECTS): Adjust.
31171 Amend handling for LD64_VERSION fallback defaults.
31173 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
31175 * config.gcc: Compute default version information
31176 from the configured target. Likewise defaults for
31178 * config/darwin10.h: Removed.
31179 * config/darwin12.h: Removed.
31180 * config/darwin9.h: Removed.
31181 * config/rs6000/darwin8.h: Removed.
31183 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
31185 * config/darwin9.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
31187 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
31189 * config/darwin9.h (STACK_CHECK_STATIC_BUILTIN): Move from here..
31190 * config/darwin.h (STACK_CHECK_STATIC_BUILTIN): .. to here.
31192 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
31194 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move from
31196 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): ... to here.
31198 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
31200 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move the spec
31201 for the Darwin10 unwinder stub from here ...
31202 * config/darwin.h (LINK_COMMAND_SPEC_A): ... to here.
31204 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
31206 * config/darwin.h (DSYMUTIL_SPEC): Default to DWARF
31207 (ASM_DEBUG_SPEC):Only define if the assembler supports
31209 (PREFERRED_DEBUGGING_TYPE): Default to DWARF.
31210 (DARWIN_PREFER_DWARF): Define.
31211 * config/darwin9.h (PREFERRED_DEBUGGING_TYPE): Remove.
31212 (DARWIN_PREFER_DWARF): Likewise
31213 (DSYMUTIL_SPEC): Likewise.
31214 (COLLECT_RUN_DSYMUTIL): Likewise.
31215 (ASM_DEBUG_SPEC): Likewise.
31216 (ASM_DEBUG_OPTION_SPEC): Likewise.
31218 2021-01-02 Jan Hubicka <jh@suse.cz>
31220 * cfg.c (free_block): ggc_free bb.
31222 2021-01-01 Jakub Jelinek <jakub@redhat.com>
31224 * gcc.c (process_command): Update copyright notice dates.
31225 * gcov-dump.c (print_version): Ditto.
31226 * gcov.c (print_version): Ditto.
31227 * gcov-tool.c (print_version): Ditto.
31228 * gengtype.c (create_file): Ditto.
31229 * doc/cpp.texi: Bump @copying's copyright year.
31230 * doc/cppinternals.texi: Ditto.
31231 * doc/gcc.texi: Ditto.
31232 * doc/gccint.texi: Ditto.
31233 * doc/gcov.texi: Ditto.
31234 * doc/install.texi: Ditto.
31235 * doc/invoke.texi: Ditto.
31237 2021-01-01 Jakub Jelinek <jakub@redhat.com>
31239 * ChangeLog-2020: Rotate ChangeLog. New file.
31242 Copyright (C) 2021 Free Software Foundation, Inc.
31244 Copying and distribution of this file, with or without modification,
31245 are permitted in any medium without royalty provided the copyright
31246 notice and this notice are preserved.