1 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
3 * ctfc.c (ctfc_delete_container): Free hash table contents.
5 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
7 * toplev.c (process_options): Do not warn for GNU GIMPLE.
9 2021-10-07 Siddhesh Poyarekar <siddhesh@gotplt.org>
11 * tree-object-size.c (addr_object_size,
12 compute_builtin_object_size): Drop PDECL and POFF arguments.
13 (addr_object_size): Adjust calls.
14 * tree-object-size.h (compute_builtin_object_size): Drop PDECL
17 2021-10-07 Roger Sayle <roger@nextmovesoftware.com>
19 * rtl.def (SMUL_HIGHPART, UMUL_HIGHPART): New RTX codes for
20 representing signed and unsigned high-part multiplication resp.
21 * simplify-rtx.c (simplify_binary_operation_1) [SMUL_HIGHPART,
22 UMUL_HIGHPART]: Simplify high-part multiplications by zero.
23 [SS_PLUS, US_PLUS, SS_MINUS, US_MINUS, SS_MULT, US_MULT,
24 SS_DIV, US_DIV]: Similar simplifications for saturating
26 (simplify_const_binary_operation) [SS_PLUS, US_PLUS, SS_MINUS,
27 US_MINUS, SS_MULT, US_MULT, SMUL_HIGHPART, UMUL_HIGHPART]:
28 Implement compile-time evaluation for constant operands.
29 * dwarf2out.c (mem_loc_descriptor): Skip SMUL_HIGHPART and
31 * doc/rtl.texi (smul_highpart, umul_highpart): Document RTX codes.
32 * doc/md.texi (smul@var{m}3_highpart, umul@var{m3}_highpart):
33 Mention the new smul_highpart and umul_highpart RTX codes.
34 * doc/invoke.texi: Silence @xref "compilation" warnings.
36 2021-10-07 Martin Jambor <mjambor@suse.cz>
39 * ipa-prop.c (ipa_edge_args_sum_t::duplicate): Also handle the
40 case when the source reference description corresponds to a
41 referance taken in a function src->caller is inlined to.
43 2021-10-07 Jan Hubicka <hubicka@ucw.cz>
46 * ipa-modref-tree.h (modref_access_node::contains_p): Handle offsets
48 (modref_access_node::try_merge_with): Add sanity check that there
49 are no redundant entries in the list.
51 2021-10-07 Richard Biener <rguenther@suse.de>
53 PR tree-optimization/102608
54 * tree-ssa-sccvn.c (visit_stmt): Drop .DEFERRED_INIT to
57 2021-10-07 Martin Liska <mliska@suse.cz>
59 * toplev.c (toplev::main): Make
60 save_opt_decoded_options a pointer type
63 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
65 * config/gcn/gcn-valu.md (gather<mode>_insn_2offsets<exec>): Apply
66 HAVE_GCN_ASM_GLOBAL_LOAD_FIXED.
67 (scatter<mode>_insn_2offsets<exec_scatter>): Likewise.
69 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
71 * config/gcn/gcn-hsa.h (SRAMOPT): Include the whole option string.
72 Adjust for new -msram-ecc=any behaviour.
73 (ASM_SPEC): Adjust -mxnack and -msram-ecc usage.
74 * config/gcn/gcn.c (output_file_start): Implement -msram-ecc=any.
75 * config/gcn/mkoffload.c (EF_AMDGPU_XNACK): Rename to ...
76 (EF_AMDGPU_XNACK_V3): ... this.
77 (EF_AMDGPU_SRAM_ECC): Rename to ...
78 (EF_AMDGPU_SRAM_ECC_V3): ... this.
79 (EF_AMDGPU_FEATURE_XNACK_V4): New.
80 (EF_AMDGPU_FEATURE_XNACK_UNSUPPORTED_V4): New.
81 (EF_AMDGPU_FEATURE_XNACK_ANY_V4): New.
82 (EF_AMDGPU_FEATURE_XNACK_OFF_V4): New.
83 (EF_AMDGPU_FEATURE_XNACK_ON_V4): New.
84 (EF_AMDGPU_FEATURE_SRAMECC_V4): New.
85 (EF_AMDGPU_FEATURE_SRAMECC_UNSUPPORTED_V4): New.
86 (EF_AMDGPU_FEATURE_SRAMECC_ANY_V4): New.
87 (EF_AMDGPU_FEATURE_SRAMECC_OFF_V4): New.
88 (EF_AMDGPU_FEATURE_SRAMECC_ON_V4): New.
92 (SET_SRAM_ECC_ON): New.
93 (SET_SRAM_ECC_ANY): New.
94 (SET_SRAM_ECC_OFF): New.
95 (TEST_SRAM_ECC_ANY): New.
96 (TEST_SRAM_ECC_ON): New.
97 (main): Implement HSACOv4 and -msram-ecc=any.
99 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
101 * config.in: Regenerate.
102 * config/gcn/gcn-hsa.h (X_FIJI): New macro.
106 (A_FIJI): Rename to ...
108 (A_900): Rename to ...
110 (A_906): Rename to ...
112 (A_908): Rename to ...
114 (SRAMOPT): New macro.
115 (ASM_SPEC): Adjust xnack option usage.
116 * config/gcn/gcn.c (output_file_start): Adjust amdgcn_target usage.
117 * configure: Regenerate.
118 * configure.ac: Detect LLVM assembler dialect.
120 2021-10-07 Richard Biener <rguenther@suse.de>
122 * tree-pretty-print.c (dump_generic_node): Do not elide
123 printing '&' when dumping with -gimple.
125 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
127 * gimple-range-cache.cc (non_null_ref::adjust_range): Call new
129 * gimple-range-fold.cc (adjust_pointer_diff_expr): Ditto.
130 (adjust_imagpart_expr): Ditto.
131 * value-range.cc (irange::irange_intersect): Call new routine if
132 RHS is a single pair.
133 (irange::intersect): New wide_int version.
134 * value-range.h (class irange): New prototype.
136 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
138 * gimple-range-edge.cc (gimple_outgoing_range::gimple_outgoing_range):
139 Add parameter to limit size when recognizing switches.
140 (gimple_outgoing_range::edge_range_p): Check size limit.
141 * gimple-range-edge.h (gimple_outgoing_range): Add size field.
142 * gimple-range-gori.cc (gori_map::calculate_gori): Ignore switches
143 that exceed the size limit.
144 (gori_compute::gori_compute): Add initializer.
145 * params.opt (evrp-switch-limit): New.
146 * doc/invoke.texi: Update docs.
148 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
150 * value-range.h (irange::set_varying): Use TYPE_MIN_VALUE and
151 TYPE_MAX_VALUE instead of creating new trees when possible.
153 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
155 * gimple-range-cache.cc (non_null_ref::adjust_range): Check for
156 zero and non-zero more efficently.
158 2021-10-06 Richard Biener <rguenther@suse.de>
161 * dumpfile.h (TDF_GIMPLE_VAL): New.
162 (dump_flag): Re-order and adjust TDF_* flags. Make
163 the enum uint32_t. Use std::underlying_type in the
165 (optgroup_flag): Likewise for the operator overloads.
166 * tree-pretty-print.c (dump_generic_node): Wrap ADDR_EXPR
167 in _Literal if TDF_GIMPLE_VAL.
168 * gimple-pretty-print.c (dump_gimple_assign): Add
169 TDF_GIMPLE_VAL to flags when dumping operands where only
170 is_gimple_val are allowed.
171 (dump_gimple_cond): Likewise.
173 2021-10-06 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
175 * gimple-isel.cc (gimple_expand_vec_cond_expr): Remove redundant if
178 2021-10-05 qing zhao <qing.zhao@oracle.com>
181 * gimplify.c (gimplify_decl_expr): Not add initialization for an
182 auto variable when it has been initialized by frontend.
184 2021-10-05 Aldy Hernandez <aldyh@redhat.com>
186 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
189 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
191 * common/config/avr/avr-common.c (avr_handle_option): Mark
192 argument as ATTRIBUTE_UNUSED.
194 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
196 * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC):
197 Undefine before redefinition.
199 2021-10-05 Richard Biener <rguenther@suse.de>
201 * toplev.c (no_backend): Remove global var.
202 (process_options): Pass in no_backend, move post_options
203 langhook call to toplev::main.
204 (do_compile): Pass in no_backend, move process_options call
206 (toplev::run_self_tests): Check no_backend at the caller.
207 (toplev::main): Call post_options and process_options
208 split out from do_compile, do self-tests only if
209 no_backend is initialized.
211 2021-10-05 Richard Biener <rguenther@suse.de>
213 * tree-cfg.c (dump_function_to_file): Dump the UID of the
214 function as part of the name when requested.
215 * tree-pretty-print.c (dump_function_name): Dump the UID when
216 requested and the langhook produced the actual name.
218 2021-10-05 Richard Biener <rguenther@suse.de>
222 * internal-fn.c (expand_DEFERRED_INIT): Fall back to
223 zero-initialization as last resort, use the constant
224 size as given by the DEFERRED_INIT argument to build
227 2021-10-04 Marek Polacek <polacek@redhat.com>
230 * doc/invoke.texi: Document -Warray-compare.
232 2021-10-04 Richard Biener <rguenther@suse.de>
234 * gimplify.c (is_var_need_auto_init): DECL_HARD_REGISTER
235 variables are not to be initialized.
237 2021-10-04 Richard Biener <rguenther@suse.de>
239 * expr.h (non_mem_decl_p): Declare.
240 (mem_ref_refers_to_non_mem_p): Likewise.
241 * expr.c (non_mem_decl_p): Export.
242 (mem_ref_refers_to_non_mem_p): Likewise.
243 * internal-fn.c (expand_DEFERRED_INIT): Do not expand the LHS
244 but check the base with mem_ref_refers_to_non_mem_p
247 2021-10-04 Richard Biener <rguenther@suse.de>
249 PR tree-optimization/102570
250 * tree-ssa-sccvn.h (vn_reference_op_struct): Document
251 we are using clique for the internal function code.
252 * tree-ssa-sccvn.c (vn_reference_op_eq): Compare the
253 internal function code.
254 (print_vn_reference_ops): Print the internal function code.
255 (vn_reference_op_compute_hash): Hash it.
256 (copy_reference_ops_from_call): Record it.
257 (visit_stmt): Remove the restriction around internal function
259 (fully_constant_vn_reference_p): Use fold_const_call and handle
261 (vn_reference_eq): Compare call return types.
262 * tree-ssa-pre.c (create_expression_by_pieces): Handle
263 generating calls to internal functions.
264 (compute_avail): Remove the restriction around internal function
267 2021-10-04 Aldy Hernandez <aldyh@redhat.com>
269 PR tree-optimization/102560
270 * gimple-ssa-warn-alloca.c (alloca_call_type): Remove static
271 marker for invalid_range.
273 2021-10-04 Richard Biener <rguenther@suse.de>
276 * internal-fn.c (expand_DEFERRED_INIT): Guard register
277 initialization path an avoid initializing VLA registers
280 2021-10-04 Eric Botcazou <ebotcazou@adacore.com>
282 * config/rs6000/vxworks.h (TARGET_INIT_LIBFUNCS): Delete.
284 2021-10-03 Martin Liska <mliska@suse.cz>
286 * toplev.c (toplev::main): Check opt_index if it is a part
289 2021-10-02 Aldy Hernandez <aldyh@redhat.com>
291 PR tree-optimization/102563
292 * range-op.cc (operator_lshift::op1_range): Do not clobber
295 2021-10-02 Martin Liska <mliska@suse.cz>
297 * toplev.c (toplev::main): save_decoded_options[0] is program
298 name and so it should be skipped.
300 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
302 PR tree-optimization/102546
303 * range-op.cc (operator_lshift::op1_range): Teach range-ops that
304 X << Y is non-zero implies X is also non-zero.
306 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
308 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
310 * config/aarch64/aarch64-tune.md: Regenerate.
311 * doc/invoke.texi: Update docs.
313 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
315 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
317 * config/aarch64/aarch64-tune.md: Regenerate.
318 * doc/invoke.texi: Update docs.
320 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
322 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
324 * config/aarch64/aarch64-tune.md: Regenerate.
325 * doc/invoke.texi: Update docs.
327 2021-10-01 Martin Sebor <msebor@redhat.com>
330 * doc/invoke.texi (-Waddress): Update.
331 * gengtype.c (write_types): Avoid -Waddress.
332 * poly-int.h (POLY_SET_COEFF): Avoid using null.
334 2021-10-01 John David Anglin <danglin@gcc.gnu.org>
337 * config/pa/pa.c (pa_option_override): Default to dwarf version 4
340 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
342 * config/aarch64/aarch64.h (AARCH64_FL_V9): Update value.
344 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
346 * gimple-range-path.cc (path_range_query::compute_ranges): Use
348 * gimple-range-path.h (class path_range_query): Remove shadowed
350 (path_range_query::get_path_oracle): New.
352 2021-10-01 Jakub Jelinek <jakub@redhat.com>
353 Richard Biener <rguenther@suse.de>
356 * doc/invoke.texi (-fsanitize=integer-divide-by-zero): Remove
357 INT_MIN / -1 division detection from here ...
358 (-fsanitize=signed-integer-overflow): ... and add it here.
360 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
362 * config/aarch64/aarch64-arches.def (AARCH64_ARCH): Added
364 * config/aarch64/aarch64.h (AARCH64_FL_V9): New.
365 (AARCH64_FL_FOR_ARCH9): New flags for Armv9-A.
366 (AARCH64_ISA_V9): New ISA flag.
367 * doc/invoke.texi: Update docs.
369 2021-10-01 Martin Liska <mliska@suse.cz>
371 * toplev.c (toplev::main): Save decoded optimization options.
372 * toplev.h (save_opt_decoded_options): New.
373 * doc/extend.texi: Be more clear about optimize and target
376 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
378 * explow.c: Include langhooks.h.
379 (set_stack_check_libfunc): Build a proper function type.
381 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
384 * config/i386/i386.c (legitimate_pic_address_disp_p): For PE-COFF do
385 not return true for external weak function symbols in medium model.
387 2021-10-01 Jakub Jelinek <jakub@redhat.com>
389 * tree.h (OMP_CLAUSE_ORDER_REPRODUCIBLE): Define.
390 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_ORDER>: Print
391 reproducible: for OMP_CLAUSE_ORDER_REPRODUCIBLE.
392 * omp-general.c (omp_extract_for_data): If OMP_CLAUSE_ORDER is seen
393 without OMP_CLAUSE_ORDER_UNCONSTRAINED, overwrite sched_kind to
394 OMP_CLAUSE_SCHEDULE_STATIC.
396 2021-10-01 Richard Biener <rguenther@suse.de>
399 * tree-inline.c (setup_one_parameter): Avoid substituting
400 an invariant into contexts where a GIMPLE register is not valid.
402 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
404 * config/arm/arm-cpus.in: Add Cortex-R52+ CPU.
405 * config/arm/arm-tables.opt: Regenerate.
406 * config/arm/arm-tune.md: Regenerate.
407 * doc/invoke.texi: Update docs.
409 2021-09-30 Uroš Bizjak <ubizjak@gmail.com>
412 * config/i386/i386.md
413 (sign_extend:WIDE (any_logic:NARROW (memory, immediate)) splitters):
416 2021-09-30 Tobias Burnus <tobias@codesourcery.com>
418 * omp-low.c (omp_runtime_api_call): Add omp_aligned_{,c}alloc and
419 omp_{c,re}alloc, fix omp_alloc/omp_free.
421 2021-09-30 Martin Liska <mliska@suse.cz>
423 * defaults.h (ASM_OUTPUT_ASCII): Do not hide global variable
424 asm_out_file and stream directly to MYFILE.
426 2021-09-30 Richard Biener <rguenther@suse.de>
428 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
429 Restore and fix condition under which we apply npeel to
430 the DRs misalignment value.
432 2021-09-30 Richard Biener <rguenther@suse.de>
434 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
435 Fix npeel check for variable amount of peeling.
437 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
439 * lto-wrapper.c (run_gcc): Plug snprintf overflow.
441 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
443 * gimple-range.cc (gimple_ranger::debug): New.
444 * gimple-range.h (class gimple_ranger): Add debug.
446 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
449 * tree-vrp.c (hybrid_threader::~hybrid_threader): Free m_query.
451 2021-09-29 Indu Bhagat <indu.bhagat@oracle.com>
454 * btfout.c (GTY): Add GTY (()) albeit for cosmetic only purpose.
455 (btf_finalize): Empty the hash_map btf_var_ids.
457 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
459 * tree-vrp.c (thread_through_all_blocks): Return bool.
460 (execute_vrp_threader): Return TODO_* flags.
461 (pass_data_vrp_threader): Set todo_flags_finish to 0.
463 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
465 * timevar.def (TV_TREE_VRP_THREADER): New.
466 * tree-vrp.c: Use TV_TREE_VRP_THREADER for VRP threader pass.
468 2021-09-29 David Faust <david.faust@oracle.com>
470 * config.gcc (bpf-*-*): Do not overwrite extra_headers.
472 2021-09-29 Jonathan Wright <jonathan.wright@arm.com>
474 * config/aarch64/aarch64-builtins.c (TYPES_BINOP_PPU): Define
475 new type qualifier enum.
476 (TYPES_TERNOP_SSSU): Likewise.
477 (TYPES_TERNOP_PPPU): Likewise.
478 * config/aarch64/aarch64-simd-builtins.def: Define PPU, SSU,
479 PPPU and SSSU builtin generator macros for qtbl1 and qtbx1
481 * config/aarch64/arm_neon.h (vqtbl1_p8): Use type-qualified
482 builtin and remove casts.
483 (vqtbl1_s8): Likewise.
484 (vqtbl1q_p8): Likewise.
485 (vqtbl1q_s8): Likewise.
486 (vqtbx1_s8): Likewise.
487 (vqtbx1_p8): Likewise.
488 (vqtbx1q_s8): Likewise.
489 (vqtbx1q_p8): Likewise.
490 (vtbl1_p8): Likewise.
491 (vtbl2_p8): Likewise.
492 (vtbx2_p8): Likewise.
494 2021-09-29 Richard Biener <rguenther@suse.de>
496 * tree-vect-data-refs.c (vect_dr_misalign_for_aligned_access):
498 (vect_update_misalignment_for_peel): Use it to update
499 misaligned to the value necessary for an aligned access.
500 (vect_get_peeling_costs_all_drs): Likewise.
501 (vect_enhance_data_refs_alignment): Likewise.
503 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
505 * config/aarch64/aarch64.c (aarch64_expand_cpymem): Count number of
506 emitted operations and adjust heuristic for code size.
508 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
510 * config/aarch64/aarch64.c (aarch64_expand_setmem): Count number of
511 emitted operations and adjust heuristic for code size.
513 2021-09-29 Jakub Jelinek <jakub@redhat.com>
516 * gimplify.c (gimplify_scan_omp_clauses): Use omp_check_private even
517 in OMP_SCOPE clauses, not just on worksharing construct clauses.
519 2021-09-28 Geng Qi <gengqi@linux.alibaba.com>
521 * config/riscv/riscv.md (mulv<mode>4): Call gen_smul<mode>3_highpart.
522 (<u>mulditi3): Call <su>muldi3_highpart.
523 (<u>muldi3_highpart): Rename to <su>muldi3_highpart.
524 (<u>mulsidi3): Call <su>mulsi3_highpart.
525 (<u>mulsi3_highpart): Rename to <su>mulsi3_highpart.
527 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
529 * config/darwin.h (DSYMUTIL_SPEC): Recognize D sources.
531 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
533 * config/rs6000/darwin.h (FIXED_R13): Add for PPC64.
534 (FIRST_SAVED_GP_REGNO): Save from R13 even when it is one
537 2021-09-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
539 * config/aarch64/aarch64.h (AARCH64_FL_LS64): Define
540 (AARCH64_FL_V8_7): Likewise.
541 (AARCH64_FL_FOR_ARCH8_7): Likewise.
542 * config/aarch64/aarch64-arches.def (armv8.7-a): Define.
543 * config/aarch64/aarch64-option-extensions.def (ls64): Define.
544 * doc/invoke.texi: Document the above.
546 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
548 * dbgcnt.c (dbg_cnt_counter): New.
549 * dbgcnt.h (dbg_cnt_counter): New.
550 * dumpfile.c (dump_options): Add entry for TDF_THREADING.
551 * dumpfile.h (enum dump_flag): Add TDF_THREADING.
552 * gimple-range-path.cc (DEBUG_SOLVER): Use TDF_THREADING.
553 * tree-ssa-threadupdate.c (dump_jump_thread_path): Dump out
556 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
558 * cfgcleanup.c (pass_jump::execute): Check
559 flag_expensive_optimizations.
560 (pass_jump_after_combine::gate): Same.
561 * doc/invoke.texi (-fthread-jumps): Enable for -O1.
562 * opts.c (default_options_table): Enable -fthread-jumps at -O1.
563 * tree-ssa-threadupdate.c
564 (fwd_jt_path_registry::remove_jump_threads_including): Bail unless
567 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
569 * tree-ssa-reassoc.c (biased_names): New global.
570 (propagate_bias_p): New function.
571 (loop_carried_phi): Remove.
572 (propagate_rank): Propagate bias along single uses.
573 (get_rank): Update biased_names when needed.
575 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
577 * passes.def (pass_reassoc): Rename parameter to early_p.
578 * tree-ssa-reassoc.c (reassoc_bias_loop_carried_phi_ranks_p):
580 (phi_rank): Don't bias loop-carried phi ranks
581 before vectorization pass.
582 (execute_reassoc): Add bias_loop_carried_phi_ranks_p parameter.
583 (pass_reassoc::pass_reassoc): Add bias_loop_carried_phi_ranks_p
585 (pass_reassoc::set_param): Set bias_loop_carried_phi_ranks_p
587 (pass_reassoc::execute): Pass bias_loop_carried_phi_ranks_p to
589 (pass_reassoc::bias_loop_carried_phi_ranks_p): New member.
591 2021-09-28 Jakub Jelinek <jakub@redhat.com>
594 * config/i386/i386.c (standard_80387_constant_p): Don't recognize
595 special 80387 instruction XFmode constants if flag_rounding_math.
597 2021-09-28 Richard Biener <rguenther@suse.de>
599 PR tree-optimization/100112
600 * tree-ssa-sccvn.c (visit_reference_op_load): Record the
601 referece into the hashtable twice in case last_vuse is
602 different from the original vuse on the stmt.
604 2021-09-28 Jakub Jelinek <jakub@redhat.com>
607 * gimplify.c (gimplify_adjust_omp_clauses_1): Don't call the
608 omp_finish_clause langhook on implicitly added OMP_CLAUSE_PRIVATE
609 clauses on SIMD constructs.
611 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
613 PR tree-optimization/102511
614 * gimple-range-path.cc (path_range_query::range_on_path_entry):
615 Return VARYING when nothing found.
617 2021-09-28 Hongyu Wang <hongyu.wang@intel.com>
620 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): Add
622 (VALID_SSE2_REG_VHF_MODE): Add V4HFmode and V2HFmode.
623 (VALID_MMX_REG_MODE): Add V4HFmode.
624 (SSE_REG_MODE_P): Replace VALID_AVX512FP16_REG_MODE with
625 vector mode condition.
626 * config/i386/i386.c (classify_argument): Parse V4HF/V2HF
628 (function_arg_32): Add V4HFmode.
629 (function_arg_advance_32): Likewise.
630 * config/i386/i386.md (mode): Add V4HF/V2HF.
631 (MODE_SIZE): Likewise.
632 * config/i386/mmx.md (MMXMODE): Add V4HF mode.
633 (V_32): Add V2HF mode.
634 (VHF_32_64): New mode iterator.
635 (*mov<mode>_internal): Adjust sse alternatives to support
637 (*mov<mode>_internal): Adjust sse alternatives to support
639 (<insn><mode>3): New define_insn for add/sub/mul/div.
641 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
643 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Check
645 (pass_early_thread_jumps::gate): Same.
646 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
647 Return if !flag_thread_jumps.
648 * tree-ssa-threadupdate.c
649 (jt_path_registry::register_jump_thread): Assert that
650 flag_thread_jumps is true.
652 2021-09-28 liuhongt <hongtao.liu@intel.com>
655 (simplify_context::simplify_binary_operation_1): Relax
656 condition of simplifying (vec_concat:M (vec_select op0
657 index0)(vec_select op1 index1)) to allow different modes
658 between op0 and M, but have same inner mode.
660 2021-09-28 liuhongt <hongtao.liu@intel.com>
662 * config/i386/i386-expand.c (emit_reduc_half): Handle
663 V8HF/V16HF/V32HFmode.
664 * config/i386/sse.md (REDUC_SSE_PLUS_MODE): Add V8HF.
665 (REDUC_SSE_SMINMAX_MODE): Ditto.
666 (REDUC_PLUS_MODE): Add V16HF and V32HF.
667 (REDUC_SMINMAX_MODE): Ditto.
669 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
671 * gimple-range-path.cc
672 (path_range_query::precompute_ranges_in_block): Rename to...
673 (path_range_query::compute_ranges_in_block): ...this.
674 (path_range_query::precompute_ranges): Rename to...
675 (path_range_query::compute_ranges): ...this.
676 (path_range_query::precompute_relations): Rename to...
677 (path_range_query::compute_relations): ...this.
678 (path_range_query::precompute_phi_relations): Rename to...
679 (path_range_query::compute_phi_relations): ...this.
680 * gimple-range-path.h: Rename precompute* to compute*.
681 * tree-ssa-threadbackward.c
682 (back_threader::find_taken_edge_switch): Same.
683 (back_threader::find_taken_edge_cond): Same.
684 * tree-ssa-threadedge.c
685 (hybrid_jt_simplifier::compute_ranges_from_state): Same.
686 (hybrid_jt_state::register_equivs_stmt): Inline...
687 * tree-ssa-threadedge.h: ...here.
689 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
691 * tree-vrp.c (lhs_of_dominating_assert): Remove.
692 (class vrp_jt_state): Remove.
693 (class vrp_jt_simplifier): Remove.
694 (vrp_jt_simplifier::simplify): Remove.
695 (class vrp_jump_threader): Remove.
696 (vrp_jump_threader::vrp_jump_threader): Remove.
697 (vrp_jump_threader::~vrp_jump_threader): Remove.
698 (vrp_jump_threader::before_dom_children): Remove.
699 (vrp_jump_threader::after_dom_children): Remove.
701 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
703 * passes.def (pass_vrp_threader): New.
704 * tree-pass.h (make_pass_vrp_threader): Add make_pass_vrp_threader.
705 * tree-ssa-threadedge.c (hybrid_jt_state::register_equivs_stmt): New.
706 (hybrid_jt_simplifier::hybrid_jt_simplifier): New.
707 (hybrid_jt_simplifier::simplify): New.
708 (hybrid_jt_simplifier::compute_ranges_from_state): New.
709 * tree-ssa-threadedge.h (class hybrid_jt_state): New.
710 (class hybrid_jt_simplifier): New.
711 * tree-vrp.c (execute_vrp): Remove ASSERT_EXPR based jump
713 (class hybrid_threader): New.
714 (hybrid_threader::hybrid_threader): New.
715 (hybrid_threader::~hybrid_threader): New.
716 (hybrid_threader::before_dom_children): New.
717 (hybrid_threader::after_dom_children): New.
718 (execute_vrp_threader): New.
719 (class pass_vrp_threader): New.
720 (make_pass_vrp_threader): New.
722 2021-09-27 Martin Liska <mliska@suse.cz>
724 * output.h (enum section_flag): New.
725 (SECTION_FORGET): Remove.
726 (SECTION_ENTSIZE): Make it (1UL << 8) - 1.
727 (SECTION_STYLE_MASK): Define it based on other enum
729 * varasm.c (switch_to_section): Remove unused handling of
732 2021-09-27 Martin Liska <mliska@suse.cz>
734 * common.opt: Add new variable flag_default_complex_method.
735 * opts.c (finish_options): Handle flags related to
736 x_flag_complex_method.
737 * toplev.c (process_options): Remove option handling related
738 to flag_complex_method.
740 2021-09-27 Richard Biener <rguenther@suse.de>
743 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid using
744 type_for_size, instead use int_mode_for_size.
746 2021-09-27 Andrew Pinski <apinski@marvell.com>
749 * gimplify.c (gimplify_save_expr): Return early
750 if the type of val is error_mark_node.
752 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
754 * tree-ssanames.c (ssa_name_has_boolean_range): Use
757 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
759 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
761 * tree-ssa-dom.c (cprop_operand): Convert to range_query API.
762 (cprop_into_stmt): Same.
763 (dom_opt_dom_walker::optimize_stmt): Same.
765 2021-09-27 Richard Biener <rguenther@suse.de>
767 PR tree-optimization/97351
768 PR tree-optimization/97352
769 PR tree-optimization/82426
770 * tree-vectorizer.h (dr_misalignment): Add vector type
772 (aligned_access_p): Likewise.
773 (known_alignment_for_access_p): Likewise.
774 (vect_supportable_dr_alignment): Likewise.
775 (vect_known_alignment_in_bytes): Likewise. Refactor.
776 (DR_MISALIGNMENT): Remove.
777 (vect_update_shared_vectype): Likewise.
778 * tree-vect-data-refs.c (dr_misalignment): Refactor, handle
779 a vector type with larger alignment requirement and apply
780 the negative step adjustment here.
781 (vect_calculate_target_alignment): Remove.
782 (vect_compute_data_ref_alignment): Get explicit vector type
783 argument, do not apply a negative step alignment adjustment
785 (vect_slp_analyze_node_alignment): Re-analyze alignment
786 when we re-visit the DR with a bigger desired alignment but
787 keep more precise results from smaller alignments.
788 * tree-vect-slp.c (vect_update_shared_vectype): Remove.
789 (vect_slp_analyze_node_operations_1): Do not update the
790 shared vector type on stmts.
791 * tree-vect-stmts.c (vect_analyze_stmt): Push/pop the
792 vector type of an SLP node to the representative stmt-info.
793 (vect_transform_stmt): Likewise.
795 2021-09-27 liuhongt <hongtao.liu@intel.com>
798 2021-09-09 liuhongt <hongtao.liu@intel.com>
801 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
802 (reduc_plus_scal_v4sf): .. this, New define_expand.
803 (reduc_plus_scal_v2df): .. and this, New define_expand.
805 2021-09-26 liuhongt <hongtao.liu@intel.com>
807 * doc/extend.texi (Half-Precision): Remove storage only
808 description for _Float16 w/o avx512fp16.
810 2021-09-25 Dimitar Dimitrov <dimitar@dinux.eu>
812 * config/pru/constraints.md (Rrio): New constraint.
813 * config/pru/predicates.md (regio_operand): New predicate.
814 * config/pru/pru-pragma.c (pru_register_pragmas): Register
815 the __regio_symbol address space.
816 * config/pru/pru-protos.h (pru_symref2ioregno): Declaration.
817 * config/pru/pru.c (pru_symref2ioregno): New helper function.
818 (pru_legitimate_address_p): Remove.
819 (pru_addr_space_legitimate_address_p): Use the address space
821 (pru_nongeneric_pointer_addrspace): New helper function.
822 (pru_insert_attributes): New function to validate __regio_symbol
824 (TARGET_INSERT_ATTRIBUTES): New macro.
825 (TARGET_LEGITIMATE_ADDRESS_P): Remove.
826 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): New macro.
827 * config/pru/pru.h (enum reg_class): Add REGIO_REGS class.
828 * config/pru/pru.md (*regio_readsi): New pattern to read I/O
830 (*regio_nozext_writesi): New pattern to write to I/O registers.
831 (*regio_zext_write_r30<EQS0:mode>): Ditto.
832 * doc/extend.texi: Document the new PRU Named Address Space.
834 2021-09-24 Patrick Palka <ppalka@redhat.com>
838 * real.c (encode_ieee_double): Avoid unwanted sign extension.
839 (encode_ieee_quad): Likewise.
841 2021-09-24 Vladimir Makarov <vmakarov@redhat.com>
843 PR rtl-optimization/102147
844 * ira-build.c (ira_conflict_vector_profitable_p): Make
845 profitability calculation independent of host compiler pointer and
848 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
850 * gimple-range-path.cc (path_range_query::path_range_query):
851 Move debugging header...
852 (path_range_query::precompute_ranges): ...here.
853 (path_range_query::internal_range_of_expr): Do not call
854 range_on_path_entry if NAME is defined in the current block.
856 2021-09-24 Richard Biener <rguenther@suse.de>
858 * cfghooks.c (verify_flow_info): Verify unallocated BB and
859 edge flags are not set.
861 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
863 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
865 (jt_path_registry::register_jump_thread): Call
866 cancel_invalid_paths.
867 * tree-ssa-threadupdate.h (class jt_path_registry): Add
868 cancel_invalid_paths.
870 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
872 PR tree-optimization/102400
873 * tree-ssa-sccvn.c (vn_reference_insert_pieces): Initialize
874 result_vdef to zero value.
876 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
878 PR tree-optimization/102451
879 * tree-ssa-dse.c (delete_dead_or_redundant_call): Record bb of stmt
882 2021-09-24 Hongyu Wang <hongyu.wang@intel.com>
884 * config/i386/sse.md (cond_<insn><mode>): Extend to support
886 (cond_mul<mode>): Likewise.
887 (cond_div<mode>): Likewise.
888 (cond_<code><mode>): Likewise.
889 (cond_fma<mode>): Likewise.
890 (cond_fms<mode>): Likewise.
891 (cond_fnma<mode>): Likewise.
892 (cond_fnms<mode>): Likewise.
894 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
896 PR tree-optimization/102463
897 * gimple-range-fold.cc (fold_using_range::relation_fold_and_or): If
898 there is no range-ops handler, don't look for a relation.
900 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
902 * gimple-range-cache.cc (ranger_cache::ranger_cache): Take
903 non-executable_edge flag as parameter.
904 * gimple-range-cache.h (ranger_cache): Adjust prototype.
905 * gimple-range-gori.cc (gori_compute::gori_compute): Take
906 non-executable_edge flag as parameter.
907 (gori_compute::outgoing_edge_range_p): Check new flag.
908 * gimple-range-gori.h (gori_compute): Adjust prototype.
909 * gimple-range.cc (gimple_ranger::gimple_ranger): Create new flag.
910 (gimple_ranger::range_on_edge): Check new flag.
911 * gimple-range.h (gimple_ranger::non_executable_edge_flag): New.
912 * gimple-ssa-evrp.c (rvrp_folder): Pass ranger flag to simplifer.
913 (hybrid_folder::hybrid_folder): Set ranger non-executable flag value.
914 (hybrid_folder::fold_stmt): Set flag value in the simplifer.
915 * vr-values.c (simplify_using_ranges::set_and_propagate_unexecutable):
916 Use not_executable flag if provided inmstead of EDGE_EXECUTABLE.
917 (simplify_using_ranges::simplify_switch_using_ranges): Clear
918 EDGE_EXECUTABLE like it originally did.
919 (simplify_using_ranges::cleanup_edges_and_switches): Clear any
920 NON_EXECUTABLE flags.
921 (simplify_using_ranges::simplify_using_ranges): Adjust.
922 * vr-values.h (class simplify_using_ranges): Adjust.
923 (simplify_using_ranges::set_range_query): Add non-executable flag param.
925 2021-09-23 Bill Schmidt <wschmidt@linux.ibm.com>
928 * config/rs6000/rs6000-call.c (rs6000_aggregate_candidate): Detect
929 zero-width bit fields and return indicator.
930 (rs6000_discover_homogeneous_aggregate): Diagnose when the
931 presence of a zero-width bit field changes parameter passing in
934 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
936 * gimple-range-fold.cc (fold_using_range::range_of_phi):
937 Remove dominator check.
939 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
941 * gimple-range-path.cc (path_range_query::precompute_relations):
942 Hoist edge calculations before using EDGE_SUCC.
944 2021-09-23 Jonathan Wakely <jwakely@redhat.com>
946 * configure.ac: Fix --with-multilib-list description.
947 * configure: Regenerate.
949 2021-09-23 Richard Biener <rguenther@suse.de>
951 PR tree-optimization/102448
952 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
953 Clear alignment info copied from DR_PTR_INFO.
955 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
957 * config/i386/i386-expand.c (ix86_use_mask_cmp_p): Enable
959 * config/i386/sse.md (sseintvecmodelower): Add HF vector modes.
960 (<avx512>_store<mode>_mask): Extend to support HF vector modes.
961 (vec_cmp<mode><avx512fmaskmodelower>): Likewise.
962 (vcond_mask_<mode><avx512fmaskmodelower>): Likewise.
963 (vcond<mode><mode>): New expander.
964 (vcond<mode><sseintvecmodelower>): Likewise.
965 (vcond<sseintvecmodelower><mode>): Likewise.
966 (vcondu<mode><sseintvecmodelower>): Likewise.
968 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
970 * config/i386/sse.md (extend<ssePHmodelower><mode>2):
972 (extendv4hf<mode>2): Likewise.
973 (extendv2hfv2df2): Likewise.
974 (trunc<mode><ssePHmodelower>2): Likewise.
975 (avx512fp16_vcvt<castmode>2ph_<mode>): Rename to ...
976 (trunc<mode>v4hf2): ... this, and drop constraints.
977 (avx512fp16_vcvtpd2ph_v2df): Rename to ...
978 (truncv2dfv2hf2): ... this, and likewise.
980 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
982 * config/i386/sse.md (float<floatunssuffix><mode><ssePHmodelower>2):
984 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>):
986 (float<floatunssuffix><mode>v4hf2): ... this, and drop constraints.
987 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Rename to ...
988 (float<floatunssuffix>v2div2hf2): ... this, and likewise.
990 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
992 * config/i386/i386.md (fix<fixunssuffix>_trunchf<mode>2): New expander.
993 (fixuns_trunchfhi2): Likewise.
994 (*fixuns_trunchfsi2zext): New define_insn.
995 * config/i386/sse.md (ssePHmodelower): New mode_attr.
996 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
997 New expander for same element vector fix_truncate.
998 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
999 Likewise for V4HF to V4SI/V4DI fix_truncate.
1000 (fix<fixunssuffix>_truncv2hfv2di2):
1001 Likeise for V2HF to V2DI fix_truncate.
1003 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
1005 * config/i386/i386.md (<code>hf3): New expander.
1007 2021-09-23 liuhongt <hongtao.liu@intel.com>
1009 * config/i386/sse.md (FMAMODEM): extend to handle FP16.
1010 (VFH_SF_AVX512VL): Extend to handle HFmode.
1011 (VF_SF_AVX512VL): Deleted.
1013 2021-09-23 liuhongt <hongtao.liu@intel.com>
1015 * config/i386/i386.md (rinthf2): New expander.
1016 (nearbyinthf2): New expander.
1018 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
1020 * tree-ssa-dom.c (class dom_jump_threader_simplifier): Rename...
1021 (class dom_jt_state): ...this and provide virtual overrides.
1022 (dom_jt_state::register_equiv): New.
1023 (class dom_jt_simplifier): Rename from
1024 dom_jump_threader_simplifier.
1025 (dom_jump_threader_simplifier::simplify): Rename...
1026 (dom_jt_simplifier::simplify): ...to this.
1027 (pass_dominator::execute): Use dom_jt_simplifier and
1029 * tree-ssa-threadedge.c (jump_threader::jump_threader):
1031 (jt_state::register_equivs_stmt): Abstract out...
1032 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
1034 (jump_threader::thread_around_empty_blocks): Update state.
1035 (jump_threader::thread_through_normal_block): Same.
1036 (jt_state::jt_state): Remove.
1037 (jt_state::push): Remove pass specific bits. Keep block vector
1039 (jt_state::append_path): New.
1040 (jt_state::pop): Remove pass specific bits.
1041 (jt_state::register_equiv): Same.
1042 (jt_state::record_ranges_from_stmt): Same.
1043 (jt_state::register_equivs_on_edge): Same. Rename...
1044 (jt_state::register_equivs_edge): ...to this.
1045 (jt_state::dump): New.
1046 (jt_state::debug): New.
1047 (jump_threader_simplifier::simplify): Remove.
1048 (jt_state::get_path): New.
1049 * tree-ssa-threadedge.h (class jt_simplifier): Make into a base
1050 class. Expose common functionality as virtual methods.
1051 (class jump_threader_simplifier): Same. Rename...
1052 (class jt_simplifier): ...to this.
1053 * tree-vrp.c (class vrp_jump_threader_simplifier): Rename...
1054 (class vrp_jt_simplifier): ...to this. Provide pass specific
1056 (class vrp_jt_state): New.
1057 (vrp_jump_threader_simplifier::simplify): Rename...
1058 (vrp_jt_simplifier::simplify): ...to this. Inline code from
1059 what used to be the base class.
1060 (vrp_jump_threader::vrp_jump_threader): Use vrp_jt_state and
1063 2021-09-22 Tobias Burnus <tobias@codesourcery.com>
1066 * doc/invoke.texi (-Wno-missing-include-dirs.): Document Fortran
1069 2021-09-22 Roger Sayle <roger@nextmovesoftware.com>
1070 Richard Biener <rguenther@suse.de>
1072 * match.pd (negation simplifications): Implement some negation
1073 folding transformations from fold-const.c's fold_negate_expr.
1074 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Add a SIMPLIFY
1075 argument, to control whether the op should be simplified prior
1076 to looking up/assigning a value number.
1077 (vn_nary_build_or_lookup): Update call to vn_nary_build_or_lookup_1.
1078 (vn_nary_simplify): Likewise.
1079 (visit_nary_op): Likewise, but when constructing a NEGATE_EXPR
1080 now call vn_nary_build_or_lookup_1 disabling simplification.
1082 2021-09-22 Jiufu Guo <guojiufu@linux.ibm.com>
1084 PR tree-optimization/102087
1085 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
1086 Update bound/cmp/control for niter.
1088 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
1090 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
1091 Move check for non-empty BB here.
1092 (fur_source::register_outgoing_edges): ...from here.
1094 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
1096 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
1097 Remove call to improve_range_with_equivs.
1098 (path_range_query::improve_range_with_equivs): Remove
1099 * gimple-range-path.h: Remove improve_range_with_equivs.
1101 2021-09-22 dianhong xu <dianhong.xu@intel.com>
1103 * config/i386/avx512fp16intrin.h:
1104 (_mm512_mask_blend_ph): New intrinsic.
1105 (_mm512_permutex2var_ph): Ditto.
1106 (_mm512_permutexvar_ph): Ditto.
1107 * config/i386/avx512fp16vlintrin.h:
1108 (_mm256_mask_blend_ph): New intrinsic.
1109 (_mm256_permutex2var_ph): Ditto.
1110 (_mm256_permutexvar_ph): Ditto.
1111 (_mm_mask_blend_ph): Ditto.
1112 (_mm_permutex2var_ph): Ditto.
1113 (_mm_permutexvar_ph): Ditto.
1115 2021-09-22 dianhong xu <dianhong.xu@intel.com>
1117 * config/i386/avx512fp16intrin.h: Add new intrinsics.
1118 (_mm512_conj_pch): New intrinsic.
1119 (_mm512_mask_conj_pch): Ditto.
1120 (_mm512_maskz_conj_pch): Ditto.
1121 * config/i386/avx512fp16vlintrin.h: Add new intrinsics.
1122 (_mm256_conj_pch): New intrinsic.
1123 (_mm256_mask_conj_pch): Ditto.
1124 (_mm256_maskz_conj_pch): Ditto.
1125 (_mm_conj_pch): Ditto.
1126 (_mm_mask_conj_pch): Ditto.
1127 (_mm_maskz_conj_pch): Ditto.
1129 2021-09-22 dianhong xu <dianhong.xu@intel.com>
1131 * config/i386/avx512fp16intrin.h (_MM512_REDUCE_OP): New macro
1132 (_mm512_reduce_add_ph): New intrinsic.
1133 (_mm512_reduce_mul_ph): Ditto.
1134 (_mm512_reduce_min_ph): Ditto.
1135 (_mm512_reduce_max_ph): Ditto.
1136 * config/i386/avx512fp16vlintrin.h
1137 (_MM256_REDUCE_OP/_MM_REDUCE_OP): New macro.
1138 (_mm256_reduce_add_ph): New intrinsic.
1139 (_mm256_reduce_mul_ph): Ditto.
1140 (_mm256_reduce_min_ph): Ditto.
1141 (_mm256_reduce_max_ph): Ditto.
1142 (_mm_reduce_add_ph): Ditto.
1143 (_mm_reduce_mul_ph): Ditto.
1144 (_mm_reduce_min_ph): Ditto.
1145 (_mm_reduce_max_ph): Ditto.
1147 2021-09-22 dianhong xu <dianhong.xu@intel.com>
1149 * config/i386/avx512fp16intrin.h (__m512h_u, __m256h_u,
1150 __m128h_u): New typedef.
1151 (_mm512_load_ph): New intrinsic.
1152 (_mm256_load_ph): Ditto.
1153 (_mm_load_ph): Ditto.
1154 (_mm512_loadu_ph): Ditto.
1155 (_mm256_loadu_ph): Ditto.
1156 (_mm_loadu_ph): Ditto.
1157 (_mm512_store_ph): Ditto.
1158 (_mm256_store_ph): Ditto.
1159 (_mm_store_ph): Ditto.
1160 (_mm512_storeu_ph): Ditto.
1161 (_mm256_storeu_ph): Ditto.
1162 (_mm_storeu_ph): Ditto.
1163 (_mm512_abs_ph): Ditto.
1164 * config/i386/avx512fp16vlintrin.h
1165 (_mm_abs_ph): Ditto.
1166 (_mm256_abs_ph): Ditto.
1168 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
1170 * config/s390/tpf.md (prologue_tpf, epilogue_tpf): Add cc clobber.
1172 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
1175 * config/s390/s390.c (s390_expand_insv): Emit a normal move if it
1176 is actually a full copy of the source operand into the target.
1177 Don't emit a strict low part move if source and target mode match.
1179 2021-09-22 Jakub Jelinek <jakub@redhat.com>
1181 PR middle-end/102415
1182 * omp-expand.c (expand_omp_single): If region->exit is NULL,
1183 assert region->entry is GIMPLE_OMP_SCOPE region and return.
1185 2021-09-22 Jakub Jelinek <jakub@redhat.com>
1187 * tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
1188 * tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
1189 arguments from 2 to 3.
1190 * tree-pretty-print.c (dump_omp_clause): Print allocator() around
1191 allocate clause allocator and print align if present.
1192 * omp-low.c (scan_sharing_clauses): Force allocate_map entry even
1193 for omp_default_mem_alloc if align modifier is present. If align
1194 modifier is present, use TREE_LIST to encode both allocator and
1196 (lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
1197 Handle align modifier on allocator clause if present.
1199 2021-09-22 liuhongt <hongtao.liu@intel.com>
1201 * config/i386/i386.md (define_attr "isa"): Add
1203 (define_attr "enabled"): Correspond fma_or_avx512vl to
1204 TARGET_FMA || TARGET_AVX512VL.
1205 * config/i386/mmx.md (fmav2sf4): Extend to AVX512 fma.
1210 2021-09-22 liuhongt <hongtao.liu@intel.com>
1212 * config/i386/i386.md (cstorehf3): New define_expand.
1214 2021-09-22 liuhongt <hongtao.liu@intel.com>
1216 * config/i386/i386.md (<rounding_insn>hf2): New expander.
1217 (sse4_1_round<mode>2): Extend from MODEF to MODEFH.
1218 * config/i386/sse.md (*sse4_1_round<ssescalarmodesuffix>):
1219 Extend from VF_128 to VFH_128.
1221 2021-09-22 liuhongt <hongtao.liu@intel.com>
1223 * config/i386/i386-features.c (i386-features.c): Handle
1225 * config/i386/i386.md (sqrthf2): New expander.
1226 (*sqrthf2): New define_insn.
1227 * config/i386/sse.md
1228 (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
1231 2021-09-22 liuhongt <hongtao.liu@intel.com>
1233 * config/i386/avx512fp16intrin.h (_mm_mask_fcmadd_sch):
1235 (_mm_mask3_fcmadd_sch): Likewise.
1236 (_mm_maskz_fcmadd_sch): Likewise.
1237 (_mm_fcmadd_sch): Likewise.
1238 (_mm_mask_fmadd_sch): Likewise.
1239 (_mm_mask3_fmadd_sch): Likewise.
1240 (_mm_maskz_fmadd_sch): Likewise.
1241 (_mm_fmadd_sch): Likewise.
1242 (_mm_mask_fcmadd_round_sch): Likewise.
1243 (_mm_mask3_fcmadd_round_sch): Likewise.
1244 (_mm_maskz_fcmadd_round_sch): Likewise.
1245 (_mm_fcmadd_round_sch): Likewise.
1246 (_mm_mask_fmadd_round_sch): Likewise.
1247 (_mm_mask3_fmadd_round_sch): Likewise.
1248 (_mm_maskz_fmadd_round_sch): Likewise.
1249 (_mm_fmadd_round_sch): Likewise.
1250 (_mm_fcmul_sch): Likewise.
1251 (_mm_mask_fcmul_sch): Likewise.
1252 (_mm_maskz_fcmul_sch): Likewise.
1253 (_mm_fmul_sch): Likewise.
1254 (_mm_mask_fmul_sch): Likewise.
1255 (_mm_maskz_fmul_sch): Likewise.
1256 (_mm_fcmul_round_sch): Likewise.
1257 (_mm_mask_fcmul_round_sch): Likewise.
1258 (_mm_maskz_fcmul_round_sch): Likewise.
1259 (_mm_fmul_round_sch): Likewise.
1260 (_mm_mask_fmul_round_sch): Likewise.
1261 (_mm_maskz_fmul_round_sch): Likewise.
1262 * config/i386/i386-builtin.def: Add corresponding new builtins.
1263 * config/i386/sse.md
1264 (avx512fp16_fmaddcsh_v8hf_maskz<round_expand_name>): New expander.
1265 (avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>): Ditto.
1266 (avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>):
1268 (avx512fp16_<complexopname>sh_v8hf_mask<round_name>): Ditto.
1269 (avx512fp16_<complexopname>sh_v8hf<mask_scalarc_name><round_scalarcz_name>):
1271 * config/i386/subst.md (mask_scalarcz_name): New.
1272 (mask_scalarc_name): Ditto.
1273 (mask_scalarc_operand3): Ditto.
1274 (mask_scalarcz_operand4): Ditto.
1275 (round_scalarcz_name): Ditto.
1276 (round_scalarc_mask_operand3): Ditto.
1277 (round_scalarcz_mask_operand4): Ditto.
1278 (round_scalarc_mask_op3): Ditto.
1279 (round_scalarcz_mask_op4): Ditto.
1280 (round_scalarcz_constraint): Ditto.
1281 (round_scalarcz_nimm_predicate): Ditto.
1282 (mask_scalarcz): Ditto.
1283 (mask_scalarc): Ditto.
1284 (round_scalarcz): Ditto.
1286 2021-09-22 liuhongt <hongtao.liu@intel.com>
1288 * config/i386/avx512fp16intrin.h (_mm512_fcmadd_pch):
1290 (_mm512_mask_fcmadd_pch): Likewise.
1291 (_mm512_mask3_fcmadd_pch): Likewise.
1292 (_mm512_maskz_fcmadd_pch): Likewise.
1293 (_mm512_fmadd_pch): Likewise.
1294 (_mm512_mask_fmadd_pch): Likewise.
1295 (_mm512_mask3_fmadd_pch): Likewise.
1296 (_mm512_maskz_fmadd_pch): Likewise.
1297 (_mm512_fcmadd_round_pch): Likewise.
1298 (_mm512_mask_fcmadd_round_pch): Likewise.
1299 (_mm512_mask3_fcmadd_round_pch): Likewise.
1300 (_mm512_maskz_fcmadd_round_pch): Likewise.
1301 (_mm512_fmadd_round_pch): Likewise.
1302 (_mm512_mask_fmadd_round_pch): Likewise.
1303 (_mm512_mask3_fmadd_round_pch): Likewise.
1304 (_mm512_maskz_fmadd_round_pch): Likewise.
1305 (_mm512_fcmul_pch): Likewise.
1306 (_mm512_mask_fcmul_pch): Likewise.
1307 (_mm512_maskz_fcmul_pch): Likewise.
1308 (_mm512_fmul_pch): Likewise.
1309 (_mm512_mask_fmul_pch): Likewise.
1310 (_mm512_maskz_fmul_pch): Likewise.
1311 (_mm512_fcmul_round_pch): Likewise.
1312 (_mm512_mask_fcmul_round_pch): Likewise.
1313 (_mm512_maskz_fcmul_round_pch): Likewise.
1314 (_mm512_fmul_round_pch): Likewise.
1315 (_mm512_mask_fmul_round_pch): Likewise.
1316 (_mm512_maskz_fmul_round_pch): Likewise.
1317 * config/i386/avx512fp16vlintrin.h (_mm_fmadd_pch):
1319 (_mm_mask_fmadd_pch): Likewise.
1320 (_mm_mask3_fmadd_pch): Likewise.
1321 (_mm_maskz_fmadd_pch): Likewise.
1322 (_mm256_fmadd_pch): Likewise.
1323 (_mm256_mask_fmadd_pch): Likewise.
1324 (_mm256_mask3_fmadd_pch): Likewise.
1325 (_mm256_maskz_fmadd_pch): Likewise.
1326 (_mm_fcmadd_pch): Likewise.
1327 (_mm_mask_fcmadd_pch): Likewise.
1328 (_mm_mask3_fcmadd_pch): Likewise.
1329 (_mm_maskz_fcmadd_pch): Likewise.
1330 (_mm256_fcmadd_pch): Likewise.
1331 (_mm256_mask_fcmadd_pch): Likewise.
1332 (_mm256_mask3_fcmadd_pch): Likewise.
1333 (_mm256_maskz_fcmadd_pch): Likewise.
1334 (_mm_fmul_pch): Likewise.
1335 (_mm_mask_fmul_pch): Likewise.
1336 (_mm_maskz_fmul_pch): Likewise.
1337 (_mm256_fmul_pch): Likewise.
1338 (_mm256_mask_fmul_pch): Likewise.
1339 (_mm256_maskz_fmul_pch): Likewise.
1340 (_mm_fcmul_pch): Likewise.
1341 (_mm_mask_fcmul_pch): Likewise.
1342 (_mm_maskz_fcmul_pch): Likewise.
1343 (_mm256_fcmul_pch): Likewise.
1344 (_mm256_mask_fcmul_pch): Likewise.
1345 (_mm256_maskz_fcmul_pch): Likewise.
1346 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF_V8HF_V8HF,
1347 V8HF_FTYPE_V16HF_V16HF_V16HF, V16HF_FTYPE_V16HF_V16HF_V16HF_UQI,
1348 V32HF_FTYPE_V32HF_V32HF_V32HF_INT,
1349 V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT): Add new builtin types.
1350 * config/i386/i386-builtin.def: Add new builtins.
1351 * config/i386/i386-expand.c: Handle new builtin types.
1352 * config/i386/subst.md (SUBST_CV): New.
1353 (maskc_name): Ditto.
1354 (maskc_operand3): Ditto.
1356 (sdc_maskz_name): Ditto.
1357 (sdc_mask_op4): Ditto.
1358 (sdc_mask_op5): Ditto.
1359 (sdc_mask_mode512bit_condition): Ditto.
1361 (round_maskc_operand3): Ditto.
1362 (round_sdc_mask_operand4): Ditto.
1363 (round_maskc_op3): Ditto.
1364 (round_sdc_mask_op4): Ditto.
1365 (round_saeonly_sdc_mask_operand5): Ditto.
1366 * config/i386/sse.md (unspec): Add complex fma unspecs.
1367 (avx512fmaskcmode): New.
1368 (UNSPEC_COMPLEX_F_C_MA): Ditto.
1369 (UNSPEC_COMPLEX_F_C_MUL): Ditto.
1370 (complexopname): Ditto.
1371 (<avx512>_fmaddc_<mode>_maskz<round_expand_name>): New expander.
1372 (<avx512>_fcmaddc_<mode>_maskz<round_expand_name>): Ditto.
1373 (fma_<complexopname>_<mode><sdc_maskz_name><round_name>): New
1375 (<avx512>_<complexopname>_<mode>_mask<round_name>): Ditto.
1376 (<avx512>_<complexopname>_<mode><maskc_name><round_name>): Ditto.
1378 2021-09-22 Kewen Lin <linkw@linux.ibm.com>
1380 * config/rs6000/rs6000.opt (rs6000-density-pct-threshold,
1381 rs6000-density-size-threshold, rs6000-density-penalty,
1382 rs6000-density-load-pct-threshold,
1383 rs6000-density-load-num-threshold): New parameter.
1384 * config/rs6000/rs6000.c (rs6000_density_test): Adjust with
1385 corresponding parameters.
1387 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1389 * gimple-range-path.cc (path_range_query::defined_outside_path):
1391 (path_range_query::range_on_path_entry): New.
1392 (path_range_query::internal_range_of_expr): Resolve unknowns
1394 (path_range_query::improve_range_with_equivs): New.
1395 (path_range_query::ssa_range_in_phi): Resolve unknowns with
1397 * gimple-range-path.h (class path_range_query): Add
1398 defined_outside_path, range_on_path_entry, and
1399 improve_range_with_equivs.
1401 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1403 * gimple-range-path.cc (path_range_query::add_to_imports): New.
1404 (path_range_query::add_copies_to_imports): New.
1405 (path_range_query::precompute_ranges): Call
1406 add_copies_to_imports.
1407 * gimple-range-path.h (class path_range_query): Add prototypes
1408 for add_copies_to_imports and add_to_imports.
1410 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1412 * gimple-range-path.cc (path_range_query::range_defined_in_block):
1413 Remove useless code.
1415 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1417 * gimple-range-fold.h (class fur_source): Make oracle protected.
1418 * gimple-range-path.cc (path_range_query::path_range_query): Add
1419 resolve argument. Initialize oracle.
1420 (path_range_query::~path_range_query): Delete oracle.
1421 (path_range_query::range_of_stmt): Adapt to use relations.
1422 (path_range_query::precompute_ranges): Pre-compute relations.
1423 (class jt_fur_source): New
1424 (jt_fur_source::jt_fur_source): New.
1425 (jt_fur_source::register_relation): New.
1426 (jt_fur_source::query_relation): New.
1427 (path_range_query::precompute_relations): New.
1428 (path_range_query::precompute_phi_relations): New.
1429 * gimple-range-path.h (path_range_query): Add resolve argument.
1430 Add oracle, precompute_relations, precompute_phi_relations.
1431 * tree-ssa-threadbackward.c (back_threader::back_threader): Pass
1432 resolve argument to solver.
1434 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1436 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
1437 Rename postfold_gcond_edges to register_outgoing_edges and
1439 (fold_using_range::postfold_gcond_edges): Rename...
1440 (fur_source::register_outgoing_edges): ...to this.
1441 * gimple-range-fold.h (postfold_gcond_edges): Rename to
1442 register_outgoing_edges and move to fur_source.
1444 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1446 * gimple-range-fold.cc (fold_using_range::range_of_phi): Check
1447 dom_info_available_p.
1449 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1451 * gimple-range-cache.cc (non_null_ref::non_null_ref): Use create
1452 and quick_grow_cleared instead of safe_grow_cleared.
1454 2021-09-21 Thomas Schwinge <thomas@codesourcery.com>
1457 * omp-oacc-neuter-broadcast.cc (oacc_do_neutering): Evaluate
1460 2021-09-21 Richard Earnshaw <rearnsha@arm.com>
1462 * configure.ac: Detect when the assembler supports new-style
1463 architecture extensions.
1464 * common/config/arm/arm-common.c (arm_rewrite_mcpu): Return
1465 the full CPU string if the assembler can grok it.
1466 (arm_rewrite_march): Likewise but for the architecture.
1467 * config.in: Regenerate.
1468 * configure: Regenerate.
1470 2021-09-21 Richard Biener <rguenther@suse.de>
1472 PR tree-optimization/102421
1473 * tree-vect-loop.c (vect_dissolve_slp_only_groups): Copy and
1474 adjust alignment info.
1476 2021-09-21 Kewen Lin <linkw@linux.ibm.com>
1478 * ipa-fnsummary.c (ipa_fn_summary_write): Remove inconsistent
1479 bitfield stream out.
1481 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
1483 * gimple-range-fold.cc (fold_using_range::range_of_phi): Ignore
1484 undefined edges, apply an equivalence if appropriate.
1485 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Return
1486 UNDEFINED if EDGE_EXECUTABLE is not set.
1487 * gimple-range.cc (gimple_ranger::gimple_ranger): Set all edges
1488 as EXECUTABLE upon startup.
1489 (gimple_ranger::range_on_edge): Return UNDEFINED for edges without
1490 EDGE_EXECUTABLE set.
1491 * vr-values.c (set_and_propagate_unexecutable): New.
1492 (simplify_using_ranges::fold_cond): Call set_and_propagate.
1493 (simplify_using_ranges::simplify_switch_using_ranges): Ditto.
1494 * vr-values.h: Add prototype.
1496 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
1498 * value-relation.cc (equiv_oracle::register_initial_def): New.
1499 (equiv_oracle::register_relation): Call register_initial_def.
1500 (equiv_oracle::add_equiv_to_block): New. Split register_relation.
1501 (relation_oracle::register_stmt): Check def block of PHI arguments.
1502 * value-relation.h (equiv_oracle): Add new prototypes.
1504 2021-09-20 Matthias Kretz <m.kretz@gsi.de>
1506 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
1507 Define __RECIPROCAL_MATH__, __NO_SIGNED_ZEROS__,
1508 __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__, and
1509 __ROUNDING_MATH__ according to their corresponding flags.
1510 * doc/cpp.texi: Document __RECIPROCAL_MATH__,
1511 __NO_SIGNED_ZEROS__, __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__,
1512 and __ROUNDING_MATH__.
1514 2021-09-20 Richard Biener <rguenther@suse.de>
1516 * tree-vect-stmts.c (vectorizable_load): Use the vectype
1519 2021-09-20 Richard Biener <rguenther@suse.de>
1521 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
1522 Do not compute alignment of the vectorized access here.
1524 2021-09-20 Richard Biener <rguenther@suse.de>
1526 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
1527 Store -1 for runtime alias peeling iterations.
1529 2021-09-20 Richard Biener <rguenther@suse.de>
1531 * config.gcc: Obsolete hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11*.
1533 2021-09-20 Thomas Schwinge <thomas@codesourcery.com>
1535 * input.c (string_concat_db::record_string_concatenation)
1536 (string_concat_db::get_string_concatenation): Skip for
1537 'RESERVED_LOCATION_P'.
1539 2021-09-20 Richard Biener <rguenther@suse.de>
1541 PR tree-optimization/65206
1542 * tree-data-ref.h (struct data_reference): Add alt_indices,
1544 * tree-data-ref.c (free_data_ref): Release alt_indices.
1545 (dr_analyze_indices): Work on struct indices and get DR_REF as tree.
1546 (create_data_ref): Adjust.
1547 (initialize_data_dependence_relation): Split into head
1548 and tail. When the base objects fail to match up try
1549 again with pointer-based analysis of indices.
1550 * tree-vectorizer.c (vec_info_shared::check_datarefs): Do
1551 not compare the lazily computed alternate set of indices.
1553 2021-09-20 Iain Sandoe <iain@sandoe.co.uk>
1555 * gcc.c: Test for execute OK when we find the
1556 programs for assembler linker and dsymutil and those
1557 were specified at configure-time.
1559 2021-09-19 Martin Sebor <msebor@redhat.com>
1561 PR middle-end/102403
1562 * gimple-predicate-analysis.cc (predicate::init_from_control_deps):
1563 Correct a function pre/postcondition.
1565 2021-09-19 Martin Sebor <msebor@redhat.com>
1567 PR middle-end/102243
1568 * tree-ssa-strlen.c (get_range): Handle null cfun.
1570 2021-09-19 Iain Sandoe <iain@sandoe.co.uk>
1572 * config/darwin.h (LINK_COMMAND_SPEC_A): Use Darwin10
1573 unwinder shim as a convenience library.
1575 2021-09-19 Andrew Pinski <apinski@marvell.com>
1577 * doc/install.texi: Add note about
1578 binutils 2.35 is required for LTO usage.
1580 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
1582 * tree-ssa-threadbackward.c
1583 (back_threader_registry::register_path): Use push_edge.
1584 * tree-ssa-threadedge.c
1585 (jump_threader::thread_around_empty_blocks): Same.
1586 (jump_threader::thread_through_normal_block): Same.
1587 (jump_threader::thread_across_edge): Same. Also, use auto_bitmap.
1589 * tree-ssa-threadupdate.c
1590 (jt_path_registry::allocate_thread_edge): Remove.
1591 (jt_path_registry::push_edge): New.
1592 (dump_jump_thread_path): Make static.
1593 * tree-ssa-threadupdate.h (allocate_thread_edge): Remove.
1596 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
1598 * gimple-range-path.cc (path_range_query::path_range_query): Add
1600 (path_range_query::dump): Remove extern declaration of dump_ranger.
1601 * gimple-range-trace.cc (dump_ranger): Add DEBUG_FUNCTION marker.
1602 * gimple-range-trace.h (dump_ranger): Add prototype.
1604 2021-09-19 John Ericson <git@JohnEricson.me>
1606 * gcc.c (find_a_program): New function, factored out of...
1607 (find_a_file): Here.
1608 (execute): Use find_a_program when looking for programs rather
1611 2021-09-19 Matwey V. Kornilov <matwey.kornilov@gmail.com>
1613 * config/avr/avr-mcus.def: Add atmega324pb.
1614 * doc/avr-mmcu.texi: Corresponding changes.
1616 2021-09-19 Roger Sayle <roger@nextmovesoftware.com>
1619 * match.pd (cmp @0 REAL_CST@1): When @0 is also REAL_CST, apply
1620 the same transformations as to @1. For comparisons against NaN,
1621 don't check HONOR_SNANS but confirm that neither operand is a
1624 2021-09-19 Benjamin Peterson <benjamin@locrian.net>
1626 * attribs.c (make_unique_name): Delete.
1627 * attribs.h (make_unique_name): Delete.
1629 2021-09-19 Andrew Pinski <apinski@marvell.com>
1631 * lra-constraints.c (check_and_process_move): Assert
1632 that dclass and sclass are greater than or equal to NO_REGS.
1634 2021-09-18 Jakub Jelinek <jakub@redhat.com>
1636 * tree.h (OMP_CLAUSE_ORDER_UNCONSTRAINED): Define.
1637 * tree-pretty-print.c (dump_omp_clause): Print unconstrained:
1638 for OMP_CLAUSE_ORDER_UNCONSTRAINED.
1640 2021-09-18 liuhongt <hongtao.liu@intel.com>
1642 * config/i386/i386-features.c (remove_partial_avx_dependency):
1643 Restrict TARGET_USE_VECTOR_FP_CONVERTS and
1644 TARGET_USE_VECTOR_CONVERTS to conversion instructions only.
1646 2021-09-18 Jakub Jelinek <jakub@redhat.com>
1648 * gimplify.c (omp_default_clause): For C/C++ default({,first}private),
1649 if file/namespace scope variable doesn't have predetermined sharing,
1650 treat it as if there was default(none).
1652 2021-09-18 liuhongt <hongtao.liu@intel.com>
1654 * config/i386/avx512fp16intrin.h (_mm_fmadd_sh):
1656 (_mm_mask_fmadd_sh): Likewise.
1657 (_mm_mask3_fmadd_sh): Likewise.
1658 (_mm_maskz_fmadd_sh): Likewise.
1659 (_mm_fmadd_round_sh): Likewise.
1660 (_mm_mask_fmadd_round_sh): Likewise.
1661 (_mm_mask3_fmadd_round_sh): Likewise.
1662 (_mm_maskz_fmadd_round_sh): Likewise.
1663 (_mm_fnmadd_sh): Likewise.
1664 (_mm_mask_fnmadd_sh): Likewise.
1665 (_mm_mask3_fnmadd_sh): Likewise.
1666 (_mm_maskz_fnmadd_sh): Likewise.
1667 (_mm_fnmadd_round_sh): Likewise.
1668 (_mm_mask_fnmadd_round_sh): Likewise.
1669 (_mm_mask3_fnmadd_round_sh): Likewise.
1670 (_mm_maskz_fnmadd_round_sh): Likewise.
1671 (_mm_fmsub_sh): Likewise.
1672 (_mm_mask_fmsub_sh): Likewise.
1673 (_mm_mask3_fmsub_sh): Likewise.
1674 (_mm_maskz_fmsub_sh): Likewise.
1675 (_mm_fmsub_round_sh): Likewise.
1676 (_mm_mask_fmsub_round_sh): Likewise.
1677 (_mm_mask3_fmsub_round_sh): Likewise.
1678 (_mm_maskz_fmsub_round_sh): Likewise.
1679 (_mm_fnmsub_sh): Likewise.
1680 (_mm_mask_fnmsub_sh): Likewise.
1681 (_mm_mask3_fnmsub_sh): Likewise.
1682 (_mm_maskz_fnmsub_sh): Likewise.
1683 (_mm_fnmsub_round_sh): Likewise.
1684 (_mm_mask_fnmsub_round_sh): Likewise.
1685 (_mm_mask3_fnmsub_round_sh): Likewise.
1686 (_mm_maskz_fnmsub_round_sh): Likewise.
1687 * config/i386/i386-builtin-types.def
1688 (V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT): New builtin type.
1689 * config/i386/i386-builtin.def: Add new builtins.
1690 * config/i386/i386-expand.c: Handle new builtin type.
1691 * config/i386/sse.md (fmai_vmfmadd_<mode><round_name>):
1692 Ajdust to support FP16.
1693 (fmai_vmfmsub_<mode><round_name>): Ditto.
1694 (fmai_vmfnmadd_<mode><round_name>): Ditto.
1695 (fmai_vmfnmsub_<mode><round_name>): Ditto.
1696 (*fmai_fmadd_<mode>): Ditto.
1697 (*fmai_fmsub_<mode>): Ditto.
1698 (*fmai_fnmadd_<mode><round_name>): Ditto.
1699 (*fmai_fnmsub_<mode><round_name>): Ditto.
1700 (avx512f_vmfmadd_<mode>_mask<round_name>): Ditto.
1701 (avx512f_vmfmadd_<mode>_mask3<round_name>): Ditto.
1702 (avx512f_vmfmadd_<mode>_maskz<round_expand_name>): Ditto.
1703 (avx512f_vmfmadd_<mode>_maskz_1<round_name>): Ditto.
1704 (*avx512f_vmfmsub_<mode>_mask<round_name>): Ditto.
1705 (avx512f_vmfmsub_<mode>_mask3<round_name>): Ditto.
1706 (*avx512f_vmfmsub_<mode>_maskz_1<round_name>): Ditto.
1707 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
1708 (*avx512f_vmfnmsub_<mode>_mask3<round_name>): Ditto.
1709 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
1710 (*avx512f_vmfnmadd_<mode>_mask<round_name>): Renamed to ...
1711 (avx512f_vmfnmadd_<mode>_mask<round_name>) ... this, and
1712 adjust to support FP16.
1713 (avx512f_vmfnmadd_<mode>_mask3<round_name>): Ditto.
1714 (avx512f_vmfnmadd_<mode>_maskz_1<round_name>): Ditto.
1715 (avx512f_vmfnmadd_<mode>_maskz<round_expand_name>): New
1718 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
1720 * config/i386/sse.md (avx512fmaskmodelower): Extend to support
1722 (maskload<mode><avx512fmaskmodelower>): Ditto.
1723 (maskstore<mode><avx512fmaskmodelower>): Ditto.
1725 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
1727 * config/i386/i386-expand.c (ix86_expand_fp_absneg_operator):
1729 (ix86_expand_copysign): Ditto.
1730 (ix86_expand_xorsign): Ditto.
1731 * config/i386/i386.c (ix86_build_const_vector): Handle HF vector
1733 (ix86_build_signbit_mask): Ditto.
1734 (ix86_can_change_mode_class): Ditto.
1735 * config/i386/i386.md
1736 (SSEMODEF): Add HFmode.
1737 (ssevecmodef): Ditto.
1738 (<code>hf2): New define_expand.
1739 (*<code>hf2_1): New define_insn_and_split.
1740 (copysign<mode>): Extend to support HFmode under AVX512FP16.
1741 (xorsign<mode>): Ditto.
1742 * config/i386/sse.md (VFB): New mode iterator.
1743 (VFB_128_256): Ditto.
1745 (sseintvecmode2): Support HF vector mode.
1746 (<code><mode>2): Use new mode iterator.
1747 (*<code><mode>2): Ditto.
1748 (copysign<mode>3): Ditto.
1749 (xorsign<mode>3): Ditto.
1750 (<code><mode>3<mask_name>): Ditto.
1751 (<code><mode>3<mask_name>): Ditto.
1752 (<sse>_andnot<mode>3<mask_name>): Adjust for HF vector mode.
1753 (<sse>_andnot<mode>3<mask_name>): Ditto.
1754 (*<code><mode>3<mask_name>): Ditto.
1755 (*<code><mode>3<mask_name>): Ditto.
1757 2021-09-18 liuhongt <hongtao.liu@intel.com>
1759 * config/i386/avx512fp16intrin.h (_mm512_mask_fmadd_ph):
1761 (_mm512_mask3_fmadd_ph): Likewise.
1762 (_mm512_maskz_fmadd_ph): Likewise.
1763 (_mm512_fmadd_round_ph): Likewise.
1764 (_mm512_mask_fmadd_round_ph): Likewise.
1765 (_mm512_mask3_fmadd_round_ph): Likewise.
1766 (_mm512_maskz_fmadd_round_ph): Likewise.
1767 (_mm512_fnmadd_ph): Likewise.
1768 (_mm512_mask_fnmadd_ph): Likewise.
1769 (_mm512_mask3_fnmadd_ph): Likewise.
1770 (_mm512_maskz_fnmadd_ph): Likewise.
1771 (_mm512_fnmadd_round_ph): Likewise.
1772 (_mm512_mask_fnmadd_round_ph): Likewise.
1773 (_mm512_mask3_fnmadd_round_ph): Likewise.
1774 (_mm512_maskz_fnmadd_round_ph): Likewise.
1775 (_mm512_fmsub_ph): Likewise.
1776 (_mm512_mask_fmsub_ph): Likewise.
1777 (_mm512_mask3_fmsub_ph): Likewise.
1778 (_mm512_maskz_fmsub_ph): Likewise.
1779 (_mm512_fmsub_round_ph): Likewise.
1780 (_mm512_mask_fmsub_round_ph): Likewise.
1781 (_mm512_mask3_fmsub_round_ph): Likewise.
1782 (_mm512_maskz_fmsub_round_ph): Likewise.
1783 (_mm512_fnmsub_ph): Likewise.
1784 (_mm512_mask_fnmsub_ph): Likewise.
1785 (_mm512_mask3_fnmsub_ph): Likewise.
1786 (_mm512_maskz_fnmsub_ph): Likewise.
1787 (_mm512_fnmsub_round_ph): Likewise.
1788 (_mm512_mask_fnmsub_round_ph): Likewise.
1789 (_mm512_mask3_fnmsub_round_ph): Likewise.
1790 (_mm512_maskz_fnmsub_round_ph): Likewise.
1791 * config/i386/avx512fp16vlintrin.h (_mm256_fmadd_ph):
1793 (_mm256_mask_fmadd_ph): Likewise.
1794 (_mm256_mask3_fmadd_ph): Likewise.
1795 (_mm256_maskz_fmadd_ph): Likewise.
1796 (_mm_fmadd_ph): Likewise.
1797 (_mm_mask_fmadd_ph): Likewise.
1798 (_mm_mask3_fmadd_ph): Likewise.
1799 (_mm_maskz_fmadd_ph): Likewise.
1800 (_mm256_fnmadd_ph): Likewise.
1801 (_mm256_mask_fnmadd_ph): Likewise.
1802 (_mm256_mask3_fnmadd_ph): Likewise.
1803 (_mm256_maskz_fnmadd_ph): Likewise.
1804 (_mm_fnmadd_ph): Likewise.
1805 (_mm_mask_fnmadd_ph): Likewise.
1806 (_mm_mask3_fnmadd_ph): Likewise.
1807 (_mm_maskz_fnmadd_ph): Likewise.
1808 (_mm256_fmsub_ph): Likewise.
1809 (_mm256_mask_fmsub_ph): Likewise.
1810 (_mm256_mask3_fmsub_ph): Likewise.
1811 (_mm256_maskz_fmsub_ph): Likewise.
1812 (_mm_fmsub_ph): Likewise.
1813 (_mm_mask_fmsub_ph): Likewise.
1814 (_mm_mask3_fmsub_ph): Likewise.
1815 (_mm_maskz_fmsub_ph): Likewise.
1816 (_mm256_fnmsub_ph): Likewise.
1817 (_mm256_mask_fnmsub_ph): Likewise.
1818 (_mm256_mask3_fnmsub_ph): Likewise.
1819 (_mm256_maskz_fnmsub_ph): Likewise.
1820 (_mm_fnmsub_ph): Likewise.
1821 (_mm_mask_fnmsub_ph): Likewise.
1822 (_mm_mask3_fnmsub_ph): Likewise.
1823 (_mm_maskz_fnmsub_ph): Likewise.
1824 * config/i386/i386-builtin.def: Add corresponding new builtins.
1825 * config/i386/sse.md
1826 (<avx512>_fmadd_<mode>_maskz<round_expand_name>): Adjust to
1827 support HF vector modes.
1828 (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
1830 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
1831 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
1832 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
1833 (<avx512>_fmadd_<mode>_mask<round_name>): Ditto.
1834 (<avx512>_fmadd_<mode>_mask3<round_name>): Ditto.
1835 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Ditto.
1836 (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>):
1838 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
1839 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
1840 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
1841 (<avx512>_fmsub_<mode>_mask<round_name>): Ditto.
1842 (<avx512>_fmsub_<mode>_mask3<round_name>): Ditto.
1843 (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>):
1845 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
1846 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
1847 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
1848 (<avx512>_fnmadd_<mode>_mask<round_name>): Ditto.
1849 (<avx512>_fnmadd_<mode>_mask3<round_name>): Ditto.
1850 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Ditto.
1851 (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>):
1853 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
1854 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
1855 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
1856 (<avx512>_fnmsub_<mode>_mask<round_name>): Ditto.
1857 (<avx512>_fnmsub_<mode>_mask3<round_name>): Ditto.
1859 2021-09-18 liuhongt <hongtao.liu@intel.com>
1861 * config/i386/avx512fp16intrin.h (_mm512_fmaddsub_ph):
1863 (_mm512_mask_fmaddsub_ph): Likewise.
1864 (_mm512_mask3_fmaddsub_ph): Likewise.
1865 (_mm512_maskz_fmaddsub_ph): Likewise.
1866 (_mm512_fmaddsub_round_ph): Likewise.
1867 (_mm512_mask_fmaddsub_round_ph): Likewise.
1868 (_mm512_mask3_fmaddsub_round_ph): Likewise.
1869 (_mm512_maskz_fmaddsub_round_ph): Likewise.
1870 (_mm512_mask_fmsubadd_ph): Likewise.
1871 (_mm512_mask3_fmsubadd_ph): Likewise.
1872 (_mm512_maskz_fmsubadd_ph): Likewise.
1873 (_mm512_fmsubadd_round_ph): Likewise.
1874 (_mm512_mask_fmsubadd_round_ph): Likewise.
1875 (_mm512_mask3_fmsubadd_round_ph): Likewise.
1876 (_mm512_maskz_fmsubadd_round_ph): Likewise.
1877 * config/i386/avx512fp16vlintrin.h (_mm256_fmaddsub_ph):
1879 (_mm256_mask_fmaddsub_ph): Likewise.
1880 (_mm256_mask3_fmaddsub_ph): Likewise.
1881 (_mm256_maskz_fmaddsub_ph): Likewise.
1882 (_mm_fmaddsub_ph): Likewise.
1883 (_mm_mask_fmaddsub_ph): Likewise.
1884 (_mm_mask3_fmaddsub_ph): Likewise.
1885 (_mm_maskz_fmaddsub_ph): Likewise.
1886 (_mm256_fmsubadd_ph): Likewise.
1887 (_mm256_mask_fmsubadd_ph): Likewise.
1888 (_mm256_mask3_fmsubadd_ph): Likewise.
1889 (_mm256_maskz_fmsubadd_ph): Likewise.
1890 (_mm_fmsubadd_ph): Likewise.
1891 (_mm_mask_fmsubadd_ph): Likewise.
1892 (_mm_mask3_fmsubadd_ph): Likewise.
1893 (_mm_maskz_fmsubadd_ph): Likewise.
1894 * config/i386/i386-builtin.def: Add corresponding new builtins.
1895 * config/i386/sse.md (VFH_SF_AVX512VL): New mode iterator.
1896 * (<avx512>_fmsubadd_<mode>_maskz<round_expand_name>): New expander.
1897 * (<avx512>_fmaddsub_<mode>_maskz<round_expand_name>): Use
1899 * (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
1901 * (<avx512>_fmaddsub_<mode>_mask<round_name>): Ditto.
1902 * (<avx512>_fmaddsub_<mode>_mask3<round_name>): Ditto.
1903 * (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
1905 * (<avx512>_fmsubadd_<mode>_mask<round_name>): Ditto.
1906 * (<avx512>_fmsubadd_<mode>_mask3<round_name>): Ditto.
1908 2021-09-18 liuhongt <hongtao.liu@intel.com>
1911 * config/i386/i386.c (ix86_print_operand): Handle
1912 V8HF/V16HF/V32HFmode.
1913 * config/i386/i386.h (VALID_BCST_MODE_P): Add HFmode.
1914 * config/i386/sse.md (avx512bcst): Remove.
1916 2021-09-17 Martin Sebor <msebor@redhat.com>
1918 * Makefile.in (OBJS): Add gimple-predicate-analysis.o.
1919 * tree-ssa-uninit.c (max_phi_args): Move to gimple-predicate-analysis.
1920 (MASK_SET_BIT, MASK_TEST_BIT, MASK_EMPTY): Same.
1921 (check_defs): Add comment.
1922 (can_skip_redundant_opnd): Update comment.
1923 (compute_uninit_opnds_pos): Adjust to namespace change.
1924 (find_pdom): Move to gimple-predicate-analysis.cc.
1926 (struct uninit_undef_val_t): New.
1927 (is_non_loop_exit_postdominating): Move to gimple-predicate-analysis.cc.
1928 (find_control_equiv_block): Same.
1929 (MAX_NUM_CHAINS, MAX_CHAIN_LEN, MAX_POSTDOM_CHECK): Same.
1930 (MAX_SWITCH_CASES): Same.
1931 (compute_control_dep_chain): Same.
1932 (find_uninit_use): Use predicate analyzer.
1933 (struct pred_info): Move to gimple-predicate-analysis.
1934 (convert_control_dep_chain_into_preds): Same.
1935 (find_predicates): Same.
1936 (collect_phi_def_edges): Same.
1937 (warn_uninitialized_phi): Use predicate analyzer.
1938 (find_def_preds): Move to gimple-predicate-analysis.
1939 (dump_pred_info): Same.
1940 (dump_pred_chain): Same.
1941 (dump_predicates): Same.
1942 (destroy_predicate_vecs): Remove.
1943 (execute_late_warn_uninitialized): New.
1944 (get_cmp_code): Move to gimple-predicate-analysis.
1945 (is_value_included_in): Same.
1946 (value_sat_pred_p): Same.
1947 (find_matching_predicate_in_rest_chains): Same.
1948 (is_use_properly_guarded): Same.
1949 (prune_uninit_phi_opnds): Same.
1950 (find_var_cmp_const): Same.
1951 (use_pred_not_overlap_with_undef_path_pred): Same.
1952 (pred_equal_p): Same.
1953 (is_neq_relop_p): Same.
1954 (is_neq_zero_form_p): Same.
1955 (pred_expr_equal_p): Same.
1956 (is_pred_expr_subset_of): Same.
1957 (is_pred_chain_subset_of): Same.
1958 (is_included_in): Same.
1959 (is_superset_of): Same.
1961 (simplify_pred): Same.
1962 (simplify_preds_2): Same.
1963 (simplify_preds_3): Same.
1964 (simplify_preds_4): Same.
1965 (simplify_preds): Same.
1967 (push_to_worklist): Same.
1968 (get_pred_info_from_cmp): Same.
1969 (is_degenerated_phi): Same.
1970 (normalize_one_pred_1): Same.
1971 (normalize_one_pred): Same.
1972 (normalize_one_pred_chain): Same.
1973 (normalize_preds): Same.
1974 (can_one_predicate_be_invalidated_p): Same.
1975 (can_chain_union_be_invalidated_p): Same.
1976 (uninit_uses_cannot_happen): Same.
1977 (pass_late_warn_uninitialized::execute): Define.
1978 * gimple-predicate-analysis.cc: New file.
1979 * gimple-predicate-analysis.h: New file.
1981 2021-09-17 Julian Brown <julian@codesourcery.com>
1983 * config/gcn/gcn.c (gimple.h): Include.
1984 (gcn_fork_join): Emit barrier for worker-level joins.
1985 * omp-oacc-neuter-broadcast.cc (find_local_vars_to_propagate): Add
1986 writes_gang_private bitmap parameter. Set bit for blocks
1987 containing gang-private variable writes.
1988 (worker_single_simple): Don't emit barrier after predicated block.
1989 (worker_single_copy): Don't emit barrier if we're not broadcasting
1990 anything and the block contains no gang-private writes.
1991 (neuter_worker_single): Don't predicate blocks that only contain
1992 NOPs or internal marker functions. Pass has_gang_private_write
1993 argument to worker_single_copy.
1994 (oacc_do_neutering): Add writes_gang_private bitmap handling.
1996 2021-09-17 Julian Brown <julian@codesourcery.com>
1998 * config/gcn/gcn-protos.h
1999 (gcn_goacc_create_worker_broadcast_record): Update prototype.
2000 * config/gcn/gcn-tree.c (gcn_goacc_get_worker_red_decl): Use
2001 preallocated block of LDS memory. Do not cache/share decls for
2002 reduction temporaries between invocations.
2003 (gcn_goacc_reduction_teardown): Unshare VAR on second use.
2004 (gcn_goacc_create_worker_broadcast_record): Add OFFSET parameter
2005 and return temporary LDS space at that offset. Return pointer in
2007 * config/gcn/gcn.c (acc_lds_size, gang_private_hwm, lds_allocs):
2009 (ACC_LDS_SIZE): Define as acc_lds_size.
2010 (gcn_init_machine_status): Don't initialise lds_allocated,
2011 lds_allocs, reduc_decls fields of machine function struct.
2012 (gcn_option_override): Handle default size for gang-private
2013 variables and -mgang-private-size option.
2014 (gcn_expand_prologue): Use LDS_SIZE instead of LDS_SIZE-1 when
2015 initialising M0_REG.
2016 (gcn_shared_mem_layout): New function.
2017 (gcn_print_lds_decl): Update comment. Use global lds_allocs map and
2018 gang_private_hwm variable.
2019 (TARGET_GOACC_SHARED_MEM_LAYOUT): Define target hook.
2020 * config/gcn/gcn.h (machine_function): Remove lds_allocated,
2021 lds_allocs, reduc_decls. Add reduction_base, reduction_limit.
2022 * config/gcn/gcn.opt (gang_private_size_opt): New global.
2023 (mgang-private-size=): New option.
2024 * doc/tm.texi.in (TARGET_GOACC_SHARED_MEM_LAYOUT): Place
2026 * doc/tm.texi: Regenerate.
2027 * omp-oacc-neuter-broadcast.cc (targhooks.h, diagnostic-core.h):
2029 (build_sender_ref): Handle sender_decl being pointer.
2030 (worker_single_copy): Add PLACEMENT and ISOLATE_BROADCASTS
2031 parameters. Pass placement argument to
2032 create_worker_broadcast_record hook invocations. Handle
2033 sender_decl being pointer and isolate_broadcasts inserting extra
2035 (blk_offset_map_t): Add typedef.
2036 (neuter_worker_single): Add BLK_OFFSET_MAP parameter. Pass
2037 preallocated range to worker_single_copy call.
2038 (dfs_broadcast_reachable_1): New function.
2039 (idx_decl_pair_t, used_range_vec_t): New typedefs.
2040 (sort_size_descending): New function.
2041 (addr_range): New class.
2042 (splay_tree_compare_addr_range, splay_tree_free_key)
2043 (first_fit_range, merge_ranges_1, merge_ranges): New functions.
2044 (execute_omp_oacc_neuter_broadcast): Rename to...
2045 (oacc_do_neutering): ... this. Add BOUNDS_LO, BOUNDS_HI
2046 parameters. Arrange layout of shared memory for broadcast
2048 (execute_omp_oacc_neuter_broadcast): New function.
2049 (pass_omp_oacc_neuter_broadcast::gate): Remove num_workers==1
2050 handling from here. Enable pass for all OpenACC routines in order
2051 to call shared memory-layout hook.
2052 * target.def (create_worker_broadcast_record): Add OFFSET
2054 (shared_mem_layout): New hook.
2056 2021-09-17 Julian Brown <julian@codesourcery.com>
2057 Thomas Schwinge <thomas@codesourcery.com>
2059 * omp-oacc-neuter-broadcast.cc
2060 (pass_omp_oacc_neuter_broadcast::gate): Disable if num_workers is
2062 (execute_omp_oacc_neuter_broadcast): Adjust.
2064 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
2066 * value-relation.cc (class equiv_chain): Move to header file.
2067 (path_oracle::path_oracle): New.
2068 (path_oracle::~path_oracle): New.
2069 (path_oracle::register_relation): New.
2070 (path_oracle::query_relation): New.
2071 (path_oracle::reset_path): New.
2072 (path_oracle::dump): New.
2073 * value-relation.h (class equiv_chain): Move to here.
2074 (class path_oracle): New.
2076 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
2078 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a DOM
2080 * gimple-range-fold.cc (fur_depend::register_relation): Use
2081 register_stmt/edge routines.
2082 * value-relation.cc (equiv_chain::find): Relocate from equiv_oracle.
2083 (equiv_oracle::equiv_oracle): Create self equivalence cache.
2084 (equiv_oracle::~equiv_oracle): Release same.
2085 (equiv_oracle::equiv_set): Return entry from self equiv cache if there
2086 are no equivalences.
2087 (equiv_oracle::find_equiv_block): Move list find to equiv_chain.
2088 (equiv_oracle::register_relation): Rename from register_equiv.
2089 (relation_chain_head::find_relation): Relocate from dom_oracle.
2090 (relation_oracle::register_stmt): New.
2091 (relation_oracle::register_edge): New.
2092 (dom_oracle::*): Rename from relation_oracle.
2093 (dom_oracle::register_relation): Adjust to call equiv_oracle.
2094 (dom_oracle::set_one_relation): Split from register_relation.
2095 (dom_oracle::register_transitives): Consolidate 2 methods.
2096 (dom_oracle::find_relation_block): Move core to relation_chain.
2097 (dom_oracle::query_relation): Rename from find_relation_dom and adjust.
2098 * value-relation.h (class relation_oracle): New pure virtual base.
2099 (class equiv_oracle): Inherit from relation_oracle and adjust.
2100 (class dom_oracle): Rename from old relation_oracle and adjust.
2102 2021-09-17 Martin Sebor <msebor@redhat.com>
2104 PR middle-end/102200
2105 * pointer-query.cc (access_ref::inform_access): Handle MIN/MAX_EXPR.
2106 (handle_min_max_size): Change argument. Store original SSA_NAME for
2107 operands to potentially distinct (sub)objects.
2108 (compute_objsize_r): Adjust call to the above.
2110 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
2112 * config/rs6000/rs6000.c (rs6000-builtins.h): New include.
2113 (rs6000_new_builtin_vectorized_function): New function.
2114 (rs6000_new_builtin_md_vectorized_function): Likewise.
2115 (rs6000_builtin_vectorized_function): Call
2116 rs6000_new_builtin_vectorized_function.
2117 (rs6000_builtin_md_vectorized_function): Call
2118 rs6000_new_builtin_md_vectorized_function.
2120 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
2122 * config/rs6000/rs6000-builtin-new.def (ASSEMBLE_ACC): Add mmaint flag.
2123 (ASSEMBLE_PAIR): Likewise.
2124 (BUILD_ACC): Likewise.
2125 (DISASSEMBLE_ACC): Likewise.
2126 (DISASSEMBLE_PAIR): Likewise.
2127 (PMXVBF16GER2): Likewise.
2128 (PMXVBF16GER2NN): Likewise.
2129 (PMXVBF16GER2NP): Likewise.
2130 (PMXVBF16GER2PN): Likewise.
2131 (PMXVBF16GER2PP): Likewise.
2132 (PMXVF16GER2): Likewise.
2133 (PMXVF16GER2NN): Likewise.
2134 (PMXVF16GER2NP): Likewise.
2135 (PMXVF16GER2PN): Likewise.
2136 (PMXVF16GER2PP): Likewise.
2137 (PMXVF32GER): Likewise.
2138 (PMXVF32GERNN): Likewise.
2139 (PMXVF32GERNP): Likewise.
2140 (PMXVF32GERPN): Likewise.
2141 (PMXVF32GERPP): Likewise.
2142 (PMXVF64GER): Likewise.
2143 (PMXVF64GERNN): Likewise.
2144 (PMXVF64GERNP): Likewise.
2145 (PMXVF64GERPN): Likewise.
2146 (PMXVF64GERPP): Likewise.
2147 (PMXVI16GER2): Likewise.
2148 (PMXVI16GER2PP): Likewise.
2149 (PMXVI16GER2S): Likewise.
2150 (PMXVI16GER2SPP): Likewise.
2151 (PMXVI4GER8): Likewise.
2152 (PMXVI4GER8PP): Likewise.
2153 (PMXVI8GER4): Likewise.
2154 (PMXVI8GER4PP): Likewise.
2155 (PMXVI8GER4SPP): Likewise.
2156 (XVBF16GER2): Likewise.
2157 (XVBF16GER2NN): Likewise.
2158 (XVBF16GER2NP): Likewise.
2159 (XVBF16GER2PN): Likewise.
2160 (XVBF16GER2PP): Likewise.
2161 (XVF16GER2): Likewise.
2162 (XVF16GER2NN): Likewise.
2163 (XVF16GER2NP): Likewise.
2164 (XVF16GER2PN): Likewise.
2165 (XVF16GER2PP): Likewise.
2166 (XVF32GER): Likewise.
2167 (XVF32GERNN): Likewise.
2168 (XVF32GERNP): Likewise.
2169 (XVF32GERPN): Likewise.
2170 (XVF32GERPP): Likewise.
2171 (XVF64GER): Likewise.
2172 (XVF64GERNN): Likewise.
2173 (XVF64GERNP): Likewise.
2174 (XVF64GERPN): Likewise.
2175 (XVF64GERPP): Likewise.
2176 (XVI16GER2): Likewise.
2177 (XVI16GER2PP): Likewise.
2178 (XVI16GER2S): Likewise.
2179 (XVI16GER2SPP): Likewise.
2180 (XVI4GER8): Likewise.
2181 (XVI4GER8PP): Likewise.
2182 (XVI8GER4): Likewise.
2183 (XVI8GER4PP): Likewise.
2184 (XVI8GER4SPP): Likewise.
2185 (XXMFACC): Likewise.
2186 (XXMTACC): Likewise.
2187 (XXSETACCZ): Likewise.
2188 (ASSEMBLE_PAIR_V): Likewise.
2189 (BUILD_PAIR): Likewise.
2190 (DISASSEMBLE_PAIR_V): Likewise.
2193 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_mma_builtin):
2194 Handle RS6000_BIF_LXVP and RS6000_BIF_STXVP.
2195 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add ismmaint.
2196 (parse_bif_attrs): Handle ismmaint.
2197 (write_decls): Add bif_mmaint_bit and bif_is_mmaint.
2198 (write_bif_static_init): Handle ismmaint.
2200 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
2202 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin): New
2204 (rs6000_gimple_fold_builtin): Call rs6000_gimple_fold_new_builtin.
2205 (rs6000_new_builtin_valid_without_lhs): New function.
2206 (rs6000_gimple_fold_new_mma_builtin): Likewise.
2207 (rs6000_gimple_fold_new_builtin): Likewise.
2209 2021-09-17 Thomas Schwinge <thomas@codesourcery.com>
2211 * hash-table.h (hash_table<Descriptor, Lazy, Allocator>::expand):
2212 Destruct stale Value objects.
2213 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
2216 2021-09-17 Roger Sayle <roger@nextmovesoftware.com>
2219 * match.pd (shift optimizations): Disable recent sign-changing
2220 optimization for shifts by zero, these will be folded later.
2222 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
2224 * config/rs6000/rs6000-builtin-new.def (__builtin_mffsl): Move from
2225 [power9] to [always].
2227 2021-09-17 Richard Biener <rguenther@suse.de>
2229 * tree-vect-stmts.c (vectorizable_load): Do not frob
2232 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
2234 * config/i386/i386-features.c (remove_partial_avx_dependency):
2235 Also check TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY and
2236 and TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY before generating
2238 * config/i386/i386.h (TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY):
2240 (TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
2241 * config/i386/i386.md (SSE FP to FP splitters): Replace
2242 TARGET_SSE_PARTIAL_REG_DEPENDENCY with
2243 TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY.
2244 (SSE INT to FP splitter): Replace TARGET_SSE_PARTIAL_REG_DEPENDENCY
2245 with TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY.
2246 * config/i386/x86-tune.def
2247 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): New.
2248 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
2250 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
2253 * config/i386/i386-features.c (remove_partial_avx_dependency):
2254 Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
2255 before generating vxorps.
2257 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
2259 * config/i386/i386-options.c (processor_cost_table): Use
2260 tremont_cost for Tremont.
2261 * config/i386/x86-tune-costs.h (tremont_memcpy): New.
2262 (tremont_memset): Likewise.
2263 (tremont_cost): Likewise.
2264 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
2267 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
2269 * common/config/i386/i386-common.c: Use Haswell scheduling model
2271 * config/i386/i386.c (ix86_sched_init_global): Prepare for Tremont
2273 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Tremont
2275 (ix86_adjust_cost): Handle Tremont.
2276 * config/i386/x86-tune.def (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY):
2278 (X86_TUNE_USE_LEAVE): Likewise.
2279 (X86_TUNE_PUSH_MEMORY): Likewise.
2280 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
2281 (X86_TUNE_USE_CLTD): Likewise.
2282 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
2283 (X86_TUNE_AVOID_MFENCE): Likewise.
2284 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
2285 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
2286 (X86_TUNE_ACCUMULATE_OUTGOING_ARGS): Disable for Tremont.
2287 (X86_TUNE_FOUR_JUMP_LIMIT): Likewise.
2288 (X86_TUNE_OPT_AGU): Likewise.
2289 (X86_TUNE_AVOID_LEA_FOR_ADDR): Likewise.
2290 (X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE): Likewise.
2291 (X86_TUNE_EXPAND_ABS): Likewise.
2292 (X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS): Likewise.
2293 (X86_TUNE_SLOW_PSHUFB): Likewise.
2295 2021-09-17 Eric Botcazou <ebotcazou@adacore.com>
2297 PR rtl-optimization/102306
2298 * combine.c (try_combine): Abort the combination if we are about to
2299 duplicate volatile references.
2301 2021-09-17 liuhongt <hongtao.liu@intel.com>
2303 * config/i386/avx512fp16intrin.h (_mm_undefined_ph):
2305 (_mm256_undefined_ph): Likewise.
2306 (_mm512_undefined_ph): Likewise.
2307 (_mm_cvtsh_h): Likewise.
2308 (_mm256_cvtsh_h): Likewise.
2309 (_mm512_cvtsh_h): Likewise.
2310 (_mm512_castph_ps): Likewise.
2311 (_mm512_castph_pd): Likewise.
2312 (_mm512_castph_si512): Likewise.
2313 (_mm512_castph512_ph128): Likewise.
2314 (_mm512_castph512_ph256): Likewise.
2315 (_mm512_castph128_ph512): Likewise.
2316 (_mm512_castph256_ph512): Likewise.
2317 (_mm512_zextph128_ph512): Likewise.
2318 (_mm512_zextph256_ph512): Likewise.
2319 (_mm512_castps_ph): Likewise.
2320 (_mm512_castpd_ph): Likewise.
2321 (_mm512_castsi512_ph): Likewise.
2322 * config/i386/avx512fp16vlintrin.h (_mm_castph_ps):
2324 (_mm256_castph_ps): Likewise.
2325 (_mm_castph_pd): Likewise.
2326 (_mm256_castph_pd): Likewise.
2327 (_mm_castph_si128): Likewise.
2328 (_mm256_castph_si256): Likewise.
2329 (_mm_castps_ph): Likewise.
2330 (_mm256_castps_ph): Likewise.
2331 (_mm_castpd_ph): Likewise.
2332 (_mm256_castpd_ph): Likewise.
2333 (_mm_castsi128_ph): Likewise.
2334 (_mm256_castsi256_ph): Likewise.
2335 (_mm256_castph256_ph128): Likewise.
2336 (_mm256_castph128_ph256): Likewise.
2337 (_mm256_zextph128_ph256): Likewise.
2339 2021-09-17 liuhongt <hongtao.liu@intel.com>
2341 * config/i386/avx512fp16intrin.h (_mm_cvtsh_ss):
2343 (_mm_mask_cvtsh_ss): Likewise.
2344 (_mm_maskz_cvtsh_ss): Likewise.
2345 (_mm_cvtsh_sd): Likewise.
2346 (_mm_mask_cvtsh_sd): Likewise.
2347 (_mm_maskz_cvtsh_sd): Likewise.
2348 (_mm_cvt_roundsh_ss): Likewise.
2349 (_mm_mask_cvt_roundsh_ss): Likewise.
2350 (_mm_maskz_cvt_roundsh_ss): Likewise.
2351 (_mm_cvt_roundsh_sd): Likewise.
2352 (_mm_mask_cvt_roundsh_sd): Likewise.
2353 (_mm_maskz_cvt_roundsh_sd): Likewise.
2354 (_mm_cvtss_sh): Likewise.
2355 (_mm_mask_cvtss_sh): Likewise.
2356 (_mm_maskz_cvtss_sh): Likewise.
2357 (_mm_cvtsd_sh): Likewise.
2358 (_mm_mask_cvtsd_sh): Likewise.
2359 (_mm_maskz_cvtsd_sh): Likewise.
2360 (_mm_cvt_roundss_sh): Likewise.
2361 (_mm_mask_cvt_roundss_sh): Likewise.
2362 (_mm_maskz_cvt_roundss_sh): Likewise.
2363 (_mm_cvt_roundsd_sh): Likewise.
2364 (_mm_mask_cvt_roundsd_sh): Likewise.
2365 (_mm_maskz_cvt_roundsd_sh): Likewise.
2366 * config/i386/i386-builtin-types.def
2367 (V8HF_FTYPE_V2DF_V8HF_V8HF_UQI_INT,
2368 V8HF_FTYPE_V4SF_V8HF_V8HF_UQI_INT,
2369 V2DF_FTYPE_V8HF_V2DF_V2DF_UQI_INT,
2370 V4SF_FTYPE_V8HF_V4SF_V4SF_UQI_INT): Add new builtin types.
2371 * config/i386/i386-builtin.def: Add corrresponding new builtins.
2372 * config/i386/i386-expand.c: Handle new builtin types.
2373 * config/i386/sse.md (VF48_128): New mode iterator.
2374 (avx512fp16_vcvtsh2<ssescalarmodesuffix><mask_scalar_name><round_saeonly_scalar_name>):
2376 (avx512fp16_vcvt<ssescalarmodesuffix>2sh<mask_scalar_name><round_scalar_name>):
2379 2021-09-17 liuhongt <hongtao.liu@intel.com>
2381 * config/i386/avx512fp16intrin.h (_mm512_cvtph_pd):
2383 (_mm512_mask_cvtph_pd): Likewise.
2384 (_mm512_maskz_cvtph_pd): Likewise.
2385 (_mm512_cvt_roundph_pd): Likewise.
2386 (_mm512_mask_cvt_roundph_pd): Likewise.
2387 (_mm512_maskz_cvt_roundph_pd): Likewise.
2388 (_mm512_cvtxph_ps): Likewise.
2389 (_mm512_mask_cvtxph_ps): Likewise.
2390 (_mm512_maskz_cvtxph_ps): Likewise.
2391 (_mm512_cvtx_roundph_ps): Likewise.
2392 (_mm512_mask_cvtx_roundph_ps): Likewise.
2393 (_mm512_maskz_cvtx_roundph_ps): Likewise.
2394 (_mm512_cvtxps_ph): Likewise.
2395 (_mm512_mask_cvtxps_ph): Likewise.
2396 (_mm512_maskz_cvtxps_ph): Likewise.
2397 (_mm512_cvtx_roundps_ph): Likewise.
2398 (_mm512_mask_cvtx_roundps_ph): Likewise.
2399 (_mm512_maskz_cvtx_roundps_ph): Likewise.
2400 (_mm512_cvtpd_ph): Likewise.
2401 (_mm512_mask_cvtpd_ph): Likewise.
2402 (_mm512_maskz_cvtpd_ph): Likewise.
2403 (_mm512_cvt_roundpd_ph): Likewise.
2404 (_mm512_mask_cvt_roundpd_ph): Likewise.
2405 (_mm512_maskz_cvt_roundpd_ph): Likewise.
2406 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_pd):
2408 (_mm_mask_cvtph_pd): Likewise.
2409 (_mm_maskz_cvtph_pd): Likewise.
2410 (_mm256_cvtph_pd): Likewise.
2411 (_mm256_mask_cvtph_pd): Likewise.
2412 (_mm256_maskz_cvtph_pd): Likewise.
2413 (_mm_cvtxph_ps): Likewise.
2414 (_mm_mask_cvtxph_ps): Likewise.
2415 (_mm_maskz_cvtxph_ps): Likewise.
2416 (_mm256_cvtxph_ps): Likewise.
2417 (_mm256_mask_cvtxph_ps): Likewise.
2418 (_mm256_maskz_cvtxph_ps): Likewise.
2419 (_mm_cvtxps_ph): Likewise.
2420 (_mm_mask_cvtxps_ph): Likewise.
2421 (_mm_maskz_cvtxps_ph): Likewise.
2422 (_mm256_cvtxps_ph): Likewise.
2423 (_mm256_mask_cvtxps_ph): Likewise.
2424 (_mm256_maskz_cvtxps_ph): Likewise.
2425 (_mm_cvtpd_ph): Likewise.
2426 (_mm_mask_cvtpd_ph): Likewise.
2427 (_mm_maskz_cvtpd_ph): Likewise.
2428 (_mm256_cvtpd_ph): Likewise.
2429 (_mm256_mask_cvtpd_ph): Likewise.
2430 (_mm256_maskz_cvtpd_ph): Likewise.
2431 * config/i386/i386-builtin.def: Add corresponding new builtins.
2432 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2433 * config/i386/i386-expand.c: Handle new builtin types.
2434 * config/i386/sse.md
2435 (VF4_128_8_256): New.
2436 (VF48H_AVX512VL): Ditto.
2437 (ssePHmode): Add HF vector modes.
2438 (castmode): Add new convertable modes.
2441 (avx512fp16_vcvt<castmode>2ph_<mode><mask_name><round_name>): Ditto.
2442 (avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
2443 (*avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
2444 (avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
2445 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
2446 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask_1): Ditto.
2447 (avx512fp16_float_extend_ph<mode>2<mask_name><round_saeonly_name>):
2449 (avx512fp16_float_extend_ph<mode>2<mask_name>): Ditto.
2450 (*avx512fp16_float_extend_ph<mode>2_load<mask_name>): Ditto.
2451 (avx512fp16_float_extend_phv2df2<mask_name>): Ditto.
2452 (*avx512fp16_float_extend_phv2df2_load<mask_name>): Ditto.
2454 2021-09-17 liuhongt <hongtao.liu@intel.com>
2456 * config/i386/avx512fp16intrin.h (_mm_cvttsh_i32):
2458 (_mm_cvttsh_u32): Likewise.
2459 (_mm_cvtt_roundsh_i32): Likewise.
2460 (_mm_cvtt_roundsh_u32): Likewise.
2461 (_mm_cvttsh_i64): Likewise.
2462 (_mm_cvttsh_u64): Likewise.
2463 (_mm_cvtt_roundsh_i64): Likewise.
2464 (_mm_cvtt_roundsh_u64): Likewise.
2465 * config/i386/i386-builtin.def: Add corresponding new builtins.
2466 * config/i386/sse.md
2467 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<round_saeonly_name>):
2470 2021-09-17 liuhongt <hongtao.liu@intel.com>
2472 * config/i386/avx512fp16intrin.h (_mm512_cvttph_epi32):
2474 (_mm512_mask_cvttph_epi32): Likewise.
2475 (_mm512_maskz_cvttph_epi32): Likewise.
2476 (_mm512_cvtt_roundph_epi32): Likewise.
2477 (_mm512_mask_cvtt_roundph_epi32): Likewise.
2478 (_mm512_maskz_cvtt_roundph_epi32): Likewise.
2479 (_mm512_cvttph_epu32): Likewise.
2480 (_mm512_mask_cvttph_epu32): Likewise.
2481 (_mm512_maskz_cvttph_epu32): Likewise.
2482 (_mm512_cvtt_roundph_epu32): Likewise.
2483 (_mm512_mask_cvtt_roundph_epu32): Likewise.
2484 (_mm512_maskz_cvtt_roundph_epu32): Likewise.
2485 (_mm512_cvttph_epi64): Likewise.
2486 (_mm512_mask_cvttph_epi64): Likewise.
2487 (_mm512_maskz_cvttph_epi64): Likewise.
2488 (_mm512_cvtt_roundph_epi64): Likewise.
2489 (_mm512_mask_cvtt_roundph_epi64): Likewise.
2490 (_mm512_maskz_cvtt_roundph_epi64): Likewise.
2491 (_mm512_cvttph_epu64): Likewise.
2492 (_mm512_mask_cvttph_epu64): Likewise.
2493 (_mm512_maskz_cvttph_epu64): Likewise.
2494 (_mm512_cvtt_roundph_epu64): Likewise.
2495 (_mm512_mask_cvtt_roundph_epu64): Likewise.
2496 (_mm512_maskz_cvtt_roundph_epu64): Likewise.
2497 (_mm512_cvttph_epi16): Likewise.
2498 (_mm512_mask_cvttph_epi16): Likewise.
2499 (_mm512_maskz_cvttph_epi16): Likewise.
2500 (_mm512_cvtt_roundph_epi16): Likewise.
2501 (_mm512_mask_cvtt_roundph_epi16): Likewise.
2502 (_mm512_maskz_cvtt_roundph_epi16): Likewise.
2503 (_mm512_cvttph_epu16): Likewise.
2504 (_mm512_mask_cvttph_epu16): Likewise.
2505 (_mm512_maskz_cvttph_epu16): Likewise.
2506 (_mm512_cvtt_roundph_epu16): Likewise.
2507 (_mm512_mask_cvtt_roundph_epu16): Likewise.
2508 (_mm512_maskz_cvtt_roundph_epu16): Likewise.
2509 * config/i386/avx512fp16vlintrin.h (_mm_cvttph_epi32):
2511 (_mm_mask_cvttph_epi32): Likewise.
2512 (_mm_maskz_cvttph_epi32): Likewise.
2513 (_mm256_cvttph_epi32): Likewise.
2514 (_mm256_mask_cvttph_epi32): Likewise.
2515 (_mm256_maskz_cvttph_epi32): Likewise.
2516 (_mm_cvttph_epu32): Likewise.
2517 (_mm_mask_cvttph_epu32): Likewise.
2518 (_mm_maskz_cvttph_epu32): Likewise.
2519 (_mm256_cvttph_epu32): Likewise.
2520 (_mm256_mask_cvttph_epu32): Likewise.
2521 (_mm256_maskz_cvttph_epu32): Likewise.
2522 (_mm_cvttph_epi64): Likewise.
2523 (_mm_mask_cvttph_epi64): Likewise.
2524 (_mm_maskz_cvttph_epi64): Likewise.
2525 (_mm256_cvttph_epi64): Likewise.
2526 (_mm256_mask_cvttph_epi64): Likewise.
2527 (_mm256_maskz_cvttph_epi64): Likewise.
2528 (_mm_cvttph_epu64): Likewise.
2529 (_mm_mask_cvttph_epu64): Likewise.
2530 (_mm_maskz_cvttph_epu64): Likewise.
2531 (_mm256_cvttph_epu64): Likewise.
2532 (_mm256_mask_cvttph_epu64): Likewise.
2533 (_mm256_maskz_cvttph_epu64): Likewise.
2534 (_mm_cvttph_epi16): Likewise.
2535 (_mm_mask_cvttph_epi16): Likewise.
2536 (_mm_maskz_cvttph_epi16): Likewise.
2537 (_mm256_cvttph_epi16): Likewise.
2538 (_mm256_mask_cvttph_epi16): Likewise.
2539 (_mm256_maskz_cvttph_epi16): Likewise.
2540 (_mm_cvttph_epu16): Likewise.
2541 (_mm_mask_cvttph_epu16): Likewise.
2542 (_mm_maskz_cvttph_epu16): Likewise.
2543 (_mm256_cvttph_epu16): Likewise.
2544 (_mm256_mask_cvttph_epu16): Likewise.
2545 (_mm256_maskz_cvttph_epu16): Likewise.
2546 * config/i386/i386-builtin.def: Add new builtins.
2547 * config/i386/sse.md
2548 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name><round_saeonly_name>):
2550 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name>): Ditto.
2551 (*avx512fp16_fix<fixunssuffix>_trunc<mode>2_load<mask_name>): Ditto.
2552 (avx512fp16_fix<fixunssuffix>_truncv2di2<mask_name>): Ditto.
2553 (avx512fp16_fix<fixunssuffix>_truncv2di2_load<mask_name>): Ditto.
2555 2021-09-17 liuhongt <hongtao.liu@intel.com>
2557 * config/i386/avx512fp16intrin.h (_mm_cvtsh_i32): New intrinsic.
2558 (_mm_cvtsh_u32): Likewise.
2559 (_mm_cvt_roundsh_i32): Likewise.
2560 (_mm_cvt_roundsh_u32): Likewise.
2561 (_mm_cvtsh_i64): Likewise.
2562 (_mm_cvtsh_u64): Likewise.
2563 (_mm_cvt_roundsh_i64): Likewise.
2564 (_mm_cvt_roundsh_u64): Likewise.
2565 (_mm_cvti32_sh): Likewise.
2566 (_mm_cvtu32_sh): Likewise.
2567 (_mm_cvt_roundi32_sh): Likewise.
2568 (_mm_cvt_roundu32_sh): Likewise.
2569 (_mm_cvti64_sh): Likewise.
2570 (_mm_cvtu64_sh): Likewise.
2571 (_mm_cvt_roundi64_sh): Likewise.
2572 (_mm_cvt_roundu64_sh): Likewise.
2573 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2574 * config/i386/i386-builtin.def: Add corresponding new builtins.
2575 * config/i386/i386-expand.c (ix86_expand_round_builtin):
2576 Handle new builtin types.
2577 * config/i386/sse.md
2578 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix><round_name>):
2580 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix>_2): Likewise.
2581 (avx512fp16_vcvt<floatsuffix>si2sh<rex64namesuffix><round_name>): Likewise.
2583 2021-09-16 Bill Schmidt <wschmidt@linux.ibm.com>
2585 * config/rs6000/rs6000-c.c (rs6000-builtins.h): New include.
2586 (altivec_resolve_new_overloaded_builtin): New forward decl.
2587 (rs6000_new_builtin_type_compatible): New function.
2588 (altivec_resolve_overloaded_builtin): Call
2589 altivec_resolve_new_overloaded_builtin.
2590 (altivec_build_new_resolved_builtin): New function.
2591 (altivec_resolve_new_overloaded_builtin): Likewise.
2592 * config/rs6000/rs6000-call.c (rs6000_new_builtin_is_supported):
2594 * config/rs6000/rs6000-gen-builtins.c (write_decls): Remove _p from
2595 name of rs6000_new_builtin_is_supported.
2597 2021-09-16 Uroš Bizjak <ubizjak@gmail.com>
2599 * config/i386/i386-protos.h (ix86_decompose_address):
2600 Change return type to bool.
2601 * config/i386/i386.c (ix86_decompose_address): Ditto.
2603 2021-09-16 Tobias Burnus <tobias@codesourcery.com>
2606 * config/rs6000/t-rs6000 (build/rs6000-gen-builtins.o, build/rbtree.o):
2607 Added 'build/' to target, use build/%.o rule.
2608 (build/rs6000-gen-builtins$(build_exeext)): Add 'build/' and
2609 '$(build_exeext)' to target and 'build/' for the *.o files.
2610 (rs6000-builtins.c): Update for those changes; run rs6000-gen-builtins
2613 2021-09-16 Martin Jambor <mjambor@suse.cz>
2615 * cgraph.c (cgraph_node::dump): Do not check caller count sums if
2616 the body has been removed. Remove trailing whitespace.
2618 2021-09-16 Richard Biener <rguenther@suse.de>
2620 PR middle-end/102360
2621 * internal-fn.c (expand_DEFERRED_INIT): Make pattern-init
2622 of non-memory more robust.
2624 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2626 * config/sparc/sparc-opts.h (enum sparc_processor_type): Add LEON5
2627 * config/sparc/sparc.c (struct processor_costs): Add LEON5 costs
2628 (leon5_adjust_cost): Increase cost of store with data dependency
2629 on ALU instruction and FPU anti-dependencies.
2630 (sparc_option_override): Add LEON5 costs
2631 (sparc_adjust_cost): Add LEON5 cost adjustments
2632 * config/sparc/sparc.h: Add LEON5
2633 * config/sparc/sparc.md: Include LEON5 scheduling information
2634 * config/sparc/sparc.opt: Add LEON5
2635 * doc/invoke.texi: Add LEON5
2636 * config/sparc/leon5.md: New file.
2638 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2640 * config/sparc/sparc.md (stack_protect_set32): Add NOP to prevent
2641 sensitive sequence for B2BST errata workaround.
2643 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2645 * config/sparc/sparc.c (sparc_do_work_around_errata): Do not begin
2646 functions with atomic instruction in the UT700 errata workaround.
2648 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2650 * config/sparc/sparc.c (next_active_non_empty_insn): New function
2651 that returns next active non empty assembly instruction.
2652 (sparc_do_work_around_errata): Use new function.
2654 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2656 * config/sparc/sparc.c (store_insn_p): Add predicate for store
2658 (load_insn_p): Add predicate for load attributes.
2659 (sparc_do_work_around_errata): Use new predicates.
2661 2021-09-16 Andreas Larsson <andreas@gaisler.com>
2663 * config/sparc/sparc.c (dump_target_flag_bits): Print bit names for
2666 2021-09-16 Martin Liska <mliska@suse.cz>
2668 * config/mips/netbsd.h: Fix typo in name of a macro.
2670 2021-09-16 liuhongt <hongtao.liu@intel.com>
2672 PR middle-end/102080
2673 * match.pd: Check mask type when doing cond_op related gimple
2675 * tree.c (is_truth_type_for): New function.
2676 * tree.h (is_truth_type_for): New declaration.
2678 2021-09-16 liuhongt <hongtao.liu@intel.com>
2680 * config/i386/avx512fp16intrin.h (_mm512_cvtepi32_ph): New
2682 (_mm512_mask_cvtepi32_ph): Likewise.
2683 (_mm512_maskz_cvtepi32_ph): Likewise.
2684 (_mm512_cvt_roundepi32_ph): Likewise.
2685 (_mm512_mask_cvt_roundepi32_ph): Likewise.
2686 (_mm512_maskz_cvt_roundepi32_ph): Likewise.
2687 (_mm512_cvtepu32_ph): Likewise.
2688 (_mm512_mask_cvtepu32_ph): Likewise.
2689 (_mm512_maskz_cvtepu32_ph): Likewise.
2690 (_mm512_cvt_roundepu32_ph): Likewise.
2691 (_mm512_mask_cvt_roundepu32_ph): Likewise.
2692 (_mm512_maskz_cvt_roundepu32_ph): Likewise.
2693 (_mm512_cvtepi64_ph): Likewise.
2694 (_mm512_mask_cvtepi64_ph): Likewise.
2695 (_mm512_maskz_cvtepi64_ph): Likewise.
2696 (_mm512_cvt_roundepi64_ph): Likewise.
2697 (_mm512_mask_cvt_roundepi64_ph): Likewise.
2698 (_mm512_maskz_cvt_roundepi64_ph): Likewise.
2699 (_mm512_cvtepu64_ph): Likewise.
2700 (_mm512_mask_cvtepu64_ph): Likewise.
2701 (_mm512_maskz_cvtepu64_ph): Likewise.
2702 (_mm512_cvt_roundepu64_ph): Likewise.
2703 (_mm512_mask_cvt_roundepu64_ph): Likewise.
2704 (_mm512_maskz_cvt_roundepu64_ph): Likewise.
2705 (_mm512_cvtepi16_ph): Likewise.
2706 (_mm512_mask_cvtepi16_ph): Likewise.
2707 (_mm512_maskz_cvtepi16_ph): Likewise.
2708 (_mm512_cvt_roundepi16_ph): Likewise.
2709 (_mm512_mask_cvt_roundepi16_ph): Likewise.
2710 (_mm512_maskz_cvt_roundepi16_ph): Likewise.
2711 (_mm512_cvtepu16_ph): Likewise.
2712 (_mm512_mask_cvtepu16_ph): Likewise.
2713 (_mm512_maskz_cvtepu16_ph): Likewise.
2714 (_mm512_cvt_roundepu16_ph): Likewise.
2715 (_mm512_mask_cvt_roundepu16_ph): Likewise.
2716 (_mm512_maskz_cvt_roundepu16_ph): Likewise.
2717 * config/i386/avx512fp16vlintrin.h (_mm_cvtepi32_ph): New
2719 (_mm_mask_cvtepi32_ph): Likewise.
2720 (_mm_maskz_cvtepi32_ph): Likewise.
2721 (_mm256_cvtepi32_ph): Likewise.
2722 (_mm256_mask_cvtepi32_ph): Likewise.
2723 (_mm256_maskz_cvtepi32_ph): Likewise.
2724 (_mm_cvtepu32_ph): Likewise.
2725 (_mm_mask_cvtepu32_ph): Likewise.
2726 (_mm_maskz_cvtepu32_ph): Likewise.
2727 (_mm256_cvtepu32_ph): Likewise.
2728 (_mm256_mask_cvtepu32_ph): Likewise.
2729 (_mm256_maskz_cvtepu32_ph): Likewise.
2730 (_mm_cvtepi64_ph): Likewise.
2731 (_mm_mask_cvtepi64_ph): Likewise.
2732 (_mm_maskz_cvtepi64_ph): Likewise.
2733 (_mm256_cvtepi64_ph): Likewise.
2734 (_mm256_mask_cvtepi64_ph): Likewise.
2735 (_mm256_maskz_cvtepi64_ph): Likewise.
2736 (_mm_cvtepu64_ph): Likewise.
2737 (_mm_mask_cvtepu64_ph): Likewise.
2738 (_mm_maskz_cvtepu64_ph): Likewise.
2739 (_mm256_cvtepu64_ph): Likewise.
2740 (_mm256_mask_cvtepu64_ph): Likewise.
2741 (_mm256_maskz_cvtepu64_ph): Likewise.
2742 (_mm_cvtepi16_ph): Likewise.
2743 (_mm_mask_cvtepi16_ph): Likewise.
2744 (_mm_maskz_cvtepi16_ph): Likewise.
2745 (_mm256_cvtepi16_ph): Likewise.
2746 (_mm256_mask_cvtepi16_ph): Likewise.
2747 (_mm256_maskz_cvtepi16_ph): Likewise.
2748 (_mm_cvtepu16_ph): Likewise.
2749 (_mm_mask_cvtepu16_ph): Likewise.
2750 (_mm_maskz_cvtepu16_ph): Likewise.
2751 (_mm256_cvtepu16_ph): Likewise.
2752 (_mm256_mask_cvtepu16_ph): Likewise.
2753 (_mm256_maskz_cvtepu16_ph): Likewise.
2754 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2755 * config/i386/i386-builtin.def: Add corresponding new builtins.
2756 * config/i386/i386-expand.c
2757 (ix86_expand_args_builtin): Handle new builtin types.
2758 (ix86_expand_round_builtin): Ditto.
2759 * config/i386/i386-modes.def: Declare V2HF and V6HF.
2760 * config/i386/sse.md (VI2H_AVX512VL): New.
2762 (sseintvecmode): Add HF vector modes.
2763 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode><mask_name><round_name>):
2765 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
2766 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
2767 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
2768 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
2769 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask_1): Ditto.
2770 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
2771 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
2772 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
2773 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
2774 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask_1): Ditto.
2775 * config/i386/subst.md (round_qq2phsuff): New subst_attr.
2777 2021-09-16 liuhongt <hongtao.liu@intel.com>
2779 * config/i386/avx512fp16intrin.h (_mm512_cvtph_epi32):
2781 (_mm512_mask_cvtph_epi32): Likewise.
2782 (_mm512_maskz_cvtph_epi32): Likewise.
2783 (_mm512_cvt_roundph_epi32): Likewise.
2784 (_mm512_mask_cvt_roundph_epi32): Likewise.
2785 (_mm512_maskz_cvt_roundph_epi32): Likewise.
2786 (_mm512_cvtph_epu32): Likewise.
2787 (_mm512_mask_cvtph_epu32): Likewise.
2788 (_mm512_maskz_cvtph_epu32): Likewise.
2789 (_mm512_cvt_roundph_epu32): Likewise.
2790 (_mm512_mask_cvt_roundph_epu32): Likewise.
2791 (_mm512_maskz_cvt_roundph_epu32): Likewise.
2792 (_mm512_cvtph_epi64): Likewise.
2793 (_mm512_mask_cvtph_epi64): Likewise.
2794 (_mm512_maskz_cvtph_epi64): Likewise.
2795 (_mm512_cvt_roundph_epi64): Likewise.
2796 (_mm512_mask_cvt_roundph_epi64): Likewise.
2797 (_mm512_maskz_cvt_roundph_epi64): Likewise.
2798 (_mm512_cvtph_epu64): Likewise.
2799 (_mm512_mask_cvtph_epu64): Likewise.
2800 (_mm512_maskz_cvtph_epu64): Likewise.
2801 (_mm512_cvt_roundph_epu64): Likewise.
2802 (_mm512_mask_cvt_roundph_epu64): Likewise.
2803 (_mm512_maskz_cvt_roundph_epu64): Likewise.
2804 (_mm512_cvtph_epi16): Likewise.
2805 (_mm512_mask_cvtph_epi16): Likewise.
2806 (_mm512_maskz_cvtph_epi16): Likewise.
2807 (_mm512_cvt_roundph_epi16): Likewise.
2808 (_mm512_mask_cvt_roundph_epi16): Likewise.
2809 (_mm512_maskz_cvt_roundph_epi16): Likewise.
2810 (_mm512_cvtph_epu16): Likewise.
2811 (_mm512_mask_cvtph_epu16): Likewise.
2812 (_mm512_maskz_cvtph_epu16): Likewise.
2813 (_mm512_cvt_roundph_epu16): Likewise.
2814 (_mm512_mask_cvt_roundph_epu16): Likewise.
2815 (_mm512_maskz_cvt_roundph_epu16): Likewise.
2816 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_epi32):
2818 (_mm_mask_cvtph_epi32): Likewise.
2819 (_mm_maskz_cvtph_epi32): Likewise.
2820 (_mm256_cvtph_epi32): Likewise.
2821 (_mm256_mask_cvtph_epi32): Likewise.
2822 (_mm256_maskz_cvtph_epi32): Likewise.
2823 (_mm_cvtph_epu32): Likewise.
2824 (_mm_mask_cvtph_epu32): Likewise.
2825 (_mm_maskz_cvtph_epu32): Likewise.
2826 (_mm256_cvtph_epu32): Likewise.
2827 (_mm256_mask_cvtph_epu32): Likewise.
2828 (_mm256_maskz_cvtph_epu32): Likewise.
2829 (_mm_cvtph_epi64): Likewise.
2830 (_mm_mask_cvtph_epi64): Likewise.
2831 (_mm_maskz_cvtph_epi64): Likewise.
2832 (_mm256_cvtph_epi64): Likewise.
2833 (_mm256_mask_cvtph_epi64): Likewise.
2834 (_mm256_maskz_cvtph_epi64): Likewise.
2835 (_mm_cvtph_epu64): Likewise.
2836 (_mm_mask_cvtph_epu64): Likewise.
2837 (_mm_maskz_cvtph_epu64): Likewise.
2838 (_mm256_cvtph_epu64): Likewise.
2839 (_mm256_mask_cvtph_epu64): Likewise.
2840 (_mm256_maskz_cvtph_epu64): Likewise.
2841 (_mm_cvtph_epi16): Likewise.
2842 (_mm_mask_cvtph_epi16): Likewise.
2843 (_mm_maskz_cvtph_epi16): Likewise.
2844 (_mm256_cvtph_epi16): Likewise.
2845 (_mm256_mask_cvtph_epi16): Likewise.
2846 (_mm256_maskz_cvtph_epi16): Likewise.
2847 (_mm_cvtph_epu16): Likewise.
2848 (_mm_mask_cvtph_epu16): Likewise.
2849 (_mm_maskz_cvtph_epu16): Likewise.
2850 (_mm256_cvtph_epu16): Likewise.
2851 (_mm256_mask_cvtph_epu16): Likewise.
2852 (_mm256_maskz_cvtph_epu16): Likewise.
2853 * config/i386/i386-builtin-types.def: Add new builtin types.
2854 * config/i386/i386-builtin.def: Add new builtins.
2855 * config/i386/i386-expand.c
2856 (ix86_expand_args_builtin): Handle new builtin types.
2857 (ix86_expand_round_builtin): Ditto.
2858 * config/i386/sse.md (sseintconvert): New.
2860 (UNSPEC_US_FIX_NOTRUNC): Ditto.
2861 (sseintconvertsignprefix): Ditto.
2862 (avx512fp16_vcvtph2<sseintconvertsignprefix><sseintconvert>_<mode><mask_name><round_name>):
2865 2021-09-16 liuhongt <hongtao.liu@intel.com>
2867 * config/i386/avx512fp16intrin.h: (_mm_cvtsi16_si128):
2869 (_mm_cvtsi128_si16): Likewise.
2870 (_mm_mask_load_sh): Likewise.
2871 (_mm_maskz_load_sh): Likewise.
2872 (_mm_mask_store_sh): Likewise.
2873 (_mm_move_sh): Likewise.
2874 (_mm_mask_move_sh): Likewise.
2875 (_mm_maskz_move_sh): Likewise.
2876 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2877 * config/i386/i386-builtin.def: Add corresponding new builtins.
2878 * config/i386/i386-expand.c
2879 (ix86_expand_special_args_builtin): Handle new builtin types.
2880 (ix86_expand_vector_init_one_nonzero): Adjust for FP16 target.
2881 * config/i386/sse.md (VI2F): New mode iterator.
2882 (vec_set<mode>_0): Use new mode iterator.
2883 (avx512f_mov<ssescalarmodelower>_mask): Adjust for HF vector mode.
2884 (avx512f_store<mode>_mask): Ditto.
2886 2021-09-16 Kewen Lin <linkw@linux.ibm.com>
2888 * config/rs6000/rs6000.opt (-mtoc-fusion): Remove.
2890 2021-09-15 David Edelsohn <dje.gcc@gmail.com>
2892 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
2893 Proceed if no symbol summary or the symbol alias flag is false.
2895 2021-09-15 Jakub Jelinek <jakub@redhat.com>
2899 * varasm.c (output_constructor_regular_field): Instead of assertion
2900 that array_size_for_constructor result is equal to size of
2901 TREE_TYPE (local->val) in bytes, assert that the type size is greater
2902 or equal to array_size_for_constructor result and use type size as
2905 2021-09-15 Martin Liska <mliska@suse.cz>
2908 * config/i386/vxworks.h: Use new macro TARGET_CPU_P.
2910 2021-09-15 Martin Liska <mliska@suse.cz>
2913 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
2914 Check that we have a symbol summary for a symbol.
2916 2021-09-15 Richard Biener <rguenther@suse.de>
2919 * config/rs6000/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
2920 to inherit from elfos.h
2922 2021-09-15 liuhongt <hongtao.liu@intel.com>
2925 * config/i386/i386-expand.c
2926 (ix86_expand_vector_init_interleave): Use puncklwd to pack 2
2928 (ix86_expand_vector_set): Use blendw instead of pinsrw.
2929 * config/i386/i386.c (ix86_can_change_mode_class): Adjust for
2930 AVX512FP16 which supports 16bit vector load.
2931 * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>):
2933 (avx512bw_interleave_high<mode><mask_name>): .. this, and
2934 extend to V32HFmode.
2935 (avx2_interleave_highv16hi<mask_name>): Rename to ..
2936 (avx2_interleave_high<mode><mask_name>): .. this, and extend
2938 (vec_interleave_highv8hi<mask_name>): Rename to ..
2939 (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode.
2940 (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>):
2942 (<mask_codefor>avx512bw_interleave_low<mode><mask_name>):
2943 this, and extend to V32HFmode.
2944 (avx2_interleave_lowv16hi<mask_name>): Rename to ..
2945 (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode.
2946 (vec_interleave_lowv8hi<mask_name>): Rename to ..
2947 (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode.
2948 (sse4_1_pblendw): Rename to ..
2949 (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode.
2950 (avx2_pblendph): New define_expand.
2951 (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use
2952 sseintmodesuffix instead of ssemodesuffix.
2953 (blendsuf): New mode attr.
2955 2021-09-15 Richard Biener <rguenther@suse.de>
2957 * tree-vectorizer.h (dr_misalignment): Move out of line.
2958 (dr_target_alignment): New.
2959 (DR_TARGET_ALIGNMENT): Wrap dr_target_alignment.
2960 (set_dr_target_alignment): New.
2961 (SET_DR_TARGET_ALIGNMENT): Wrap set_dr_target_alignment.
2962 * tree-vect-data-refs.c (dr_misalignment): Compute and
2963 return the group members misalignment.
2964 (vect_compute_data_ref_alignment): Use SET_DR_TARGET_ALIGNMENT.
2965 (vect_analyze_data_refs_alignment): Compute alignment only
2966 for the first element of a DR group.
2967 (vect_slp_analyze_node_alignment): Likewise.
2969 2021-09-15 Hongyu Wang <hongyu.wang@intel.com>
2971 * config/i386/avx512fp16intrin.h: Adjust all builtin calls.
2972 * config/i386/avx512fp16vlintrin.h: Likewise.
2973 * config/i386/i386-builtin.def: Adjust builtin name and
2974 enumeration to match AVX512F style.
2976 2021-09-15 Richard Biener <rguenther@suse.de>
2978 PR tree-optimization/102318
2979 * tree-vect-loop.c (vect_transform_cycle_phi): Revert
2980 previous change and do the mode conversion separately from
2981 the sign conversion.
2983 2021-09-15 Hongtao Liu <hongtao.liu@intel.com>
2984 Peter Cordes <peter@cordes.ca>
2987 * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
2988 (*vec_extract<mode><ssescalarmodelower>_valign): Output
2989 vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
2992 2021-09-15 Richard Biener <rguenther@suse.de>
2994 * config.gcc: Remove vax-*-openbsd* configuration.
2996 2021-09-15 Richard Biener <rguenther@suse.de>
2998 * config.gcc: Remove m68k-openbsd.
3000 2021-09-15 Max Filippov <jcmvbkbc@gmail.com>
3003 * config/xtensa/t-xtensa (TM_H): Add include/xtensa-config.h.
3005 2021-09-14 Peter Bergner <bergner@linux.ibm.com>
3007 * config/rs6000/mma.md (unspec): Delete UNSPEC_MMA_XXSETACCZ.
3008 (unspecv): Add UNSPECV_MMA_XXSETACCZ.
3009 (*mma_xxsetaccz): Delete.
3010 (mma_xxsetaccz): Change to define_insn. Remove operand 1.
3011 Use UNSPECV_MMA_XXSETACCZ. Update comment.
3012 * config/rs6000/rs6000.c (rs6000_rtx_costs): Use UNSPECV_MMA_XXSETACCZ.
3014 2021-09-14 Iain Sandoe <iain@sandoe.co.uk>
3016 * Makefile.in: Remove variables related to applying no-PIE
3017 to the exes on $build.
3018 * configure: Regenerate.
3019 * configure.ac: Remove configuration related to applying
3020 no-PIE to the exes on $build.
3022 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
3024 * config/arc/arc.md (doloop_end): Add missing mode.
3025 (loop_end): Likewise.
3027 2021-09-14 Jakub Jelinek <jakub@redhat.com>
3029 * gimplify.c (goa_stabilize_expr): Add depth argument, propagate
3030 it to recursive calls, for depth above 7 just gimplify or return.
3031 Perform a test even for MODIFY_EXPR, ADDR_EXPR, COMPOUND_EXPR with
3032 __builtin_clear_padding and TARGET_EXPR.
3033 (gimplify_omp_atomic): Adjust goa_stabilize_expr callers.
3035 2021-09-14 liuhongt <hongtao.liu@intel.com>
3037 * config/i386/avx512fp16intrin.h (_mm_fpclass_sh_mask):
3039 (_mm_mask_fpclass_sh_mask): Likewise.
3040 (_mm512_mask_fpclass_ph_mask): Likewise.
3041 (_mm512_fpclass_ph_mask): Likewise.
3042 (_mm_getexp_sh): Likewise.
3043 (_mm_mask_getexp_sh): Likewise.
3044 (_mm_maskz_getexp_sh): Likewise.
3045 (_mm512_getexp_ph): Likewise.
3046 (_mm512_mask_getexp_ph): Likewise.
3047 (_mm512_maskz_getexp_ph): Likewise.
3048 (_mm_getexp_round_sh): Likewise.
3049 (_mm_mask_getexp_round_sh): Likewise.
3050 (_mm_maskz_getexp_round_sh): Likewise.
3051 (_mm512_getexp_round_ph): Likewise.
3052 (_mm512_mask_getexp_round_ph): Likewise.
3053 (_mm512_maskz_getexp_round_ph): Likewise.
3054 (_mm_getmant_sh): Likewise.
3055 (_mm_mask_getmant_sh): Likewise.
3056 (_mm_maskz_getmant_sh): Likewise.
3057 (_mm512_getmant_ph): Likewise.
3058 (_mm512_mask_getmant_ph): Likewise.
3059 (_mm512_maskz_getmant_ph): Likewise.
3060 (_mm_getmant_round_sh): Likewise.
3061 (_mm_mask_getmant_round_sh): Likewise.
3062 (_mm_maskz_getmant_round_sh): Likewise.
3063 (_mm512_getmant_round_ph): Likewise.
3064 (_mm512_mask_getmant_round_ph): Likewise.
3065 (_mm512_maskz_getmant_round_ph): Likewise.
3066 * config/i386/avx512fp16vlintrin.h (_mm_mask_fpclass_ph_mask):
3068 (_mm_fpclass_ph_mask): Likewise.
3069 (_mm256_mask_fpclass_ph_mask): Likewise.
3070 (_mm256_fpclass_ph_mask): Likewise.
3071 (_mm256_getexp_ph): Likewise.
3072 (_mm256_mask_getexp_ph): Likewise.
3073 (_mm256_maskz_getexp_ph): Likewise.
3074 (_mm_getexp_ph): Likewise.
3075 (_mm_mask_getexp_ph): Likewise.
3076 (_mm_maskz_getexp_ph): Likewise.
3077 (_mm256_getmant_ph): Likewise.
3078 (_mm256_mask_getmant_ph): Likewise.
3079 (_mm256_maskz_getmant_ph): Likewise.
3080 (_mm_getmant_ph): Likewise.
3081 (_mm_mask_getmant_ph): Likewise.
3082 (_mm_maskz_getmant_ph): Likewise.
3083 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3084 * config/i386/i386-builtin.def: Add corresponding new builtins.
3085 * config/i386/i386-expand.c
3086 (ix86_expand_args_builtin): Handle new builtin types.
3087 (ix86_expand_round_builtin): Ditto.
3088 * config/i386/sse.md (vecmemsuffix): Add HF vector modes.
3089 (<avx512>_getexp<mode><mask_name><round_saeonly_name>): Adjust
3090 to support HF vector modes.
3091 (avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name):
3093 (avx512dq_fpclass<mode><mask_scalar_merge_name>): Ditto.
3094 (avx512dq_vmfpclass<mode><mask_scalar_merge_name>): Ditto.
3095 (<avx512>_getmant<mode><mask_name><round_saeonly_name>): Ditto.
3096 (avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3099 2021-09-14 liuhongt <hongtao.liu@intel.com>
3101 * config/i386/avx512fp16intrin.h (_mm512_reduce_ph):
3103 (_mm512_mask_reduce_ph): Likewise.
3104 (_mm512_maskz_reduce_ph): Likewise.
3105 (_mm512_reduce_round_ph): Likewise.
3106 (_mm512_mask_reduce_round_ph): Likewise.
3107 (_mm512_maskz_reduce_round_ph): Likewise.
3108 (_mm_reduce_sh): Likewise.
3109 (_mm_mask_reduce_sh): Likewise.
3110 (_mm_maskz_reduce_sh): Likewise.
3111 (_mm_reduce_round_sh): Likewise.
3112 (_mm_mask_reduce_round_sh): Likewise.
3113 (_mm_maskz_reduce_round_sh): Likewise.
3114 (_mm512_roundscale_ph): Likewise.
3115 (_mm512_mask_roundscale_ph): Likewise.
3116 (_mm512_maskz_roundscale_ph): Likewise.
3117 (_mm512_roundscale_round_ph): Likewise.
3118 (_mm512_mask_roundscale_round_ph): Likewise.
3119 (_mm512_maskz_roundscale_round_ph): Likewise.
3120 (_mm_roundscale_sh): Likewise.
3121 (_mm_mask_roundscale_sh): Likewise.
3122 (_mm_maskz_roundscale_sh): Likewise.
3123 (_mm_roundscale_round_sh): Likewise.
3124 (_mm_mask_roundscale_round_sh): Likewise.
3125 (_mm_maskz_roundscale_round_sh): Likewise.
3126 * config/i386/avx512fp16vlintrin.h: (_mm_reduce_ph):
3128 (_mm_mask_reduce_ph): Likewise.
3129 (_mm_maskz_reduce_ph): Likewise.
3130 (_mm256_reduce_ph): Likewise.
3131 (_mm256_mask_reduce_ph): Likewise.
3132 (_mm256_maskz_reduce_ph): Likewise.
3133 (_mm_roundscale_ph): Likewise.
3134 (_mm_mask_roundscale_ph): Likewise.
3135 (_mm_maskz_roundscale_ph): Likewise.
3136 (_mm256_roundscale_ph): Likewise.
3137 (_mm256_mask_roundscale_ph): Likewise.
3138 (_mm256_maskz_roundscale_ph): Likewise.
3139 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3140 * config/i386/i386-builtin.def: Add corresponding new builtins.
3141 * config/i386/i386-expand.c
3142 (ix86_expand_args_builtin): Handle new builtin types.
3143 (ix86_expand_round_builtin): Ditto.
3144 * config/i386/sse.md (<mask_codefor>reducep<mode><mask_name>):
3146 (<mask_codefor>reducep<mode><mask_name><round_saeonly_name>):
3147 ... this, and adjust for round operands.
3148 (reduces<mode><mask_scalar_name>): Likewise, with ...
3149 (reduces<mode><mask_scalar_name><round_saeonly_scalar_name):
3151 (<avx512>_rndscale<mode><mask_name><round_saeonly_name>):
3152 Adjust for HF vector modes.
3153 (avx512f_rndscale<mode><mask_scalar_name><round_saeonly_scalar_name>):
3155 (*avx512f_rndscale<mode><round_saeonly_name>): Ditto.
3157 2021-09-14 liuhongt <hongtao.liu@intel.com>
3159 * config/i386/avx512fp16intrin.h: (_mm512_rcp_ph):
3161 (_mm512_mask_rcp_ph): Likewise.
3162 (_mm512_maskz_rcp_ph): Likewise.
3163 (_mm_rcp_sh): Likewise.
3164 (_mm_mask_rcp_sh): Likewise.
3165 (_mm_maskz_rcp_sh): Likewise.
3166 (_mm512_scalef_ph): Likewise.
3167 (_mm512_mask_scalef_ph): Likewise.
3168 (_mm512_maskz_scalef_ph): Likewise.
3169 (_mm512_scalef_round_ph): Likewise.
3170 (_mm512_mask_scalef_round_ph): Likewise.
3171 (_mm512_maskz_scalef_round_ph): Likewise.
3172 (_mm_scalef_sh): Likewise.
3173 (_mm_mask_scalef_sh): Likewise.
3174 (_mm_maskz_scalef_sh): Likewise.
3175 (_mm_scalef_round_sh): Likewise.
3176 (_mm_mask_scalef_round_sh): Likewise.
3177 (_mm_maskz_scalef_round_sh): Likewise.
3178 * config/i386/avx512fp16vlintrin.h (_mm_rcp_ph):
3180 (_mm256_rcp_ph): Likewise.
3181 (_mm_mask_rcp_ph): Likewise.
3182 (_mm256_mask_rcp_ph): Likewise.
3183 (_mm_maskz_rcp_ph): Likewise.
3184 (_mm256_maskz_rcp_ph): Likewise.
3185 (_mm_scalef_ph): Likewise.
3186 (_mm256_scalef_ph): Likewise.
3187 (_mm_mask_scalef_ph): Likewise.
3188 (_mm256_mask_scalef_ph): Likewise.
3189 (_mm_maskz_scalef_ph): Likewise.
3190 (_mm256_maskz_scalef_ph): Likewise.
3191 * config/i386/i386-builtin.def: Add new builtins.
3192 * config/i386/sse.md (VFH_AVX512VL): New.
3193 (avx512fp16_rcp<mode>2<mask_name>): Ditto.
3194 (avx512fp16_vmrcpv8hf2<mask_scalar_name>): Ditto.
3195 (avx512f_vmscalef<mode><mask_scalar_name><round_scalar_name>):
3196 Adjust to support HF vector modes.
3197 (<avx512>_scalef<mode><mask_name><round_name>): Ditto.
3199 2021-09-14 liuhongt <hongtao.liu@intel.com>
3201 * config/i386/avx512fp16intrin.h: (_mm512_sqrt_ph):
3203 (_mm512_mask_sqrt_ph): Likewise.
3204 (_mm512_maskz_sqrt_ph): Likewise.
3205 (_mm512_sqrt_round_ph): Likewise.
3206 (_mm512_mask_sqrt_round_ph): Likewise.
3207 (_mm512_maskz_sqrt_round_ph): Likewise.
3208 (_mm512_rsqrt_ph): Likewise.
3209 (_mm512_mask_rsqrt_ph): Likewise.
3210 (_mm512_maskz_rsqrt_ph): Likewise.
3211 (_mm_rsqrt_sh): Likewise.
3212 (_mm_mask_rsqrt_sh): Likewise.
3213 (_mm_maskz_rsqrt_sh): Likewise.
3214 (_mm_sqrt_sh): Likewise.
3215 (_mm_mask_sqrt_sh): Likewise.
3216 (_mm_maskz_sqrt_sh): Likewise.
3217 (_mm_sqrt_round_sh): Likewise.
3218 (_mm_mask_sqrt_round_sh): Likewise.
3219 (_mm_maskz_sqrt_round_sh): Likewise.
3220 * config/i386/avx512fp16vlintrin.h (_mm_sqrt_ph): New intrinsic.
3221 (_mm256_sqrt_ph): Likewise.
3222 (_mm_mask_sqrt_ph): Likewise.
3223 (_mm256_mask_sqrt_ph): Likewise.
3224 (_mm_maskz_sqrt_ph): Likewise.
3225 (_mm256_maskz_sqrt_ph): Likewise.
3226 (_mm_rsqrt_ph): Likewise.
3227 (_mm256_rsqrt_ph): Likewise.
3228 (_mm_mask_rsqrt_ph): Likewise.
3229 (_mm256_mask_rsqrt_ph): Likewise.
3230 (_mm_maskz_rsqrt_ph): Likewise.
3231 (_mm256_maskz_rsqrt_ph): Likewise.
3232 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3233 * config/i386/i386-builtin.def: Add corresponding new builtins.
3234 * config/i386/i386-expand.c
3235 (ix86_expand_args_builtin): Handle new builtins.
3236 (ix86_expand_round_builtin): Ditto.
3237 * config/i386/sse.md (VF_AVX512FP16VL): New.
3238 (sqrt<mode>2): Adjust for HF vector modes.
3239 (<sse>_sqrt<mode>2<mask_name><round_name>): Likewise.
3240 (<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
3242 (<sse>_rsqrt<mode>2<mask_name>): New.
3243 (avx512fp16_vmrsqrtv8hf2<mask_scalar_name>): Likewise.
3245 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
3248 * diagnostic-spec.c (warning_suppressed_at, copy_warning): Handle
3249 'RESERVED_LOCATION_P' locations.
3250 * warning-control.cc (get_nowarn_spec, suppress_warning)
3251 (copy_warning): Likewise.
3253 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
3255 * diagnostic-spec.h (typedef xint_hash_t): Use 'location_t' instead of...
3256 (typedef key_type_t): ... this. Remove.
3257 (nowarn_map): Document.
3258 * diagnostic-spec.c (nowarn_map): Likewise.
3259 * warning-control.cc (convert_to_key): Evolve functions into...
3260 (get_location): ... these. Adjust all users.
3262 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
3264 * warning-control.cc (copy_warning): Remove 'nowarn_map' setup.
3266 2021-09-13 Jason Merrill <jason@redhat.com>
3268 * params.opt: Add destructive-interference-size and
3269 constructive-interference-size.
3270 * doc/invoke.texi: Document them.
3271 * config/aarch64/aarch64.c (aarch64_override_options_internal):
3273 * config/arm/arm.c (arm_option_override): Set them.
3274 * config/i386/i386-options.c (ix86_option_override_internal):
3277 2021-09-13 Martin Liska <mliska@suse.cz>
3278 H.J. Lu <hjl.tools@gmail.com>
3281 * common/config/i386/cpuinfo.h (cpu_indicator_init): Add support
3282 for x86-64 micro levels for __builtin_cpu_supports.
3283 * common/config/i386/i386-cpuinfo.h (enum feature_priority):
3284 Add priorities for the micro-arch levels.
3285 (enum processor_features): Add new features.
3286 * common/config/i386/i386-isas.h: Add micro-arch features.
3287 * config/i386/i386-builtins.c (get_builtin_code_for_version):
3288 Support the micro-arch levels by callsing
3289 __builtin_cpu_supports.
3290 * doc/extend.texi: Document that the levels are support by
3291 __builtin_cpu_supports.
3293 2021-09-13 Andrew Pinski <apinski@marvell.com>
3296 * config/aarch64/aarch64-builtins.c (aarch64_fold_builtin_lane_check):
3298 (aarch64_general_fold_builtin): Handle AARCH64_SIMD_BUILTIN_LANE_CHECK.
3299 (aarch64_general_gimple_fold_builtin): Likewise.
3301 2021-09-13 Andrew Pinski <apinski@marvell.com>
3303 * config.gcc: Add m32r-*-linux* and m32rle-*-linux*
3304 to the Unsupported targets list.
3305 Remove support for m32r-*-linux* and m32rle-*-linux*.
3306 * config/m32r/linux.h: Removed.
3307 * config/m32r/t-linux: Removed.
3309 2021-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3312 * config/aarch64/aarch64.c (aarch64_classify_address): Don't allow
3313 register index for SVE predicate modes.
3315 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
3317 * tree-ssa-threadbackward.c
3318 (back_threader_profitability::profitable_path_p): Remove FSM
3320 (back_threader_registry::register_path): Same.
3321 * tree-ssa-threadedge.c
3322 (jump_threader::simplify_control_stmt_condition): Same.
3323 * tree-ssa-threadupdate.c (jt_path_registry::jt_path_registry):
3324 Add backedge_threads argument.
3325 (fwd_jt_path_registry::fwd_jt_path_registry): Pass
3326 backedge_threads argument.
3327 (back_jt_path_registry::back_jt_path_registry): Same.
3328 (dump_jump_thread_path): Adjust for FSM removal.
3329 (back_jt_path_registry::rewire_first_differing_edge): Same.
3330 (back_jt_path_registry::adjust_paths_after_duplication): Same.
3331 (back_jt_path_registry::update_cfg): Same.
3332 (jt_path_registry::register_jump_thread): Same.
3333 * tree-ssa-threadupdate.h (enum jump_thread_edge_type): Remove
3335 (class back_jt_path_registry): Add backedge_threads to
3338 2021-09-13 Martin Liska <mliska@suse.cz>
3341 * asan.h (sanitize_coverage_p): Handle when fn == NULL.
3343 2021-09-13 H.J. Lu <hjl.tools@gmail.com>
3346 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): New.
3347 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
3348 (MOVE_MAX): Check TARGET_AVX256_MOVE_BY_PIECES and
3349 TARGET_AVX256_STORE_BY_PIECES instead of
3350 TARGET_AVX256_SPLIT_UNALIGNED_LOAD and
3351 TARGET_AVX256_SPLIT_UNALIGNED_STORE.
3352 (STORE_MAX_PIECES): Check TARGET_AVX256_STORE_BY_PIECES instead
3353 of TARGET_AVX256_SPLIT_UNALIGNED_STORE.
3354 * config/i386/x86-tune.def (X86_TUNE_AVX256_MOVE_BY_PIECES): New.
3355 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
3357 2021-09-13 liuhongt <hongtao.liu@intel.com>
3360 * expmed.c (extract_bit_field_using_extv): Use
3361 gen_lowpart_if_possible instead of gen_lowpart to avoid ICE.
3363 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
3365 * Makefile.in (OBJS): Add value-pointer-equiv.o.
3366 * gimple-ssa-evrp.c (class ssa_equiv_stack): Move to
3367 value-pointer-equiv.*.
3368 (ssa_equiv_stack::ssa_equiv_stack): Same.
3369 (ssa_equiv_stack::enter): Same.
3370 (ssa_equiv_stack::leave): Same.
3371 (ssa_equiv_stack::push_replacement): Same.
3372 (ssa_equiv_stack::get_replacement): Same.
3373 (is_pointer_ssa): Same.
3374 (class pointer_equiv_analyzer): Same.
3375 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
3376 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Same.
3377 (pointer_equiv_analyzer::set_global_equiv): Same.
3378 (pointer_equiv_analyzer::set_cond_equiv): Same.
3379 (pointer_equiv_analyzer::get_equiv): Same.
3380 (pointer_equiv_analyzer::enter): Same.
3381 (pointer_equiv_analyzer::leave): Same.
3382 (pointer_equiv_analyzer::get_equiv_expr): Same.
3383 (pta_valueize): Same.
3384 (pointer_equiv_analyzer::visit_stmt): Same.
3385 (pointer_equiv_analyzer::visit_edge): Same.
3386 (hybrid_folder::value_of_expr): Same.
3387 (hybrid_folder::value_on_edge): Same.
3388 * value-pointer-equiv.cc: New file.
3389 * value-pointer-equiv.h: New file.
3391 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
3394 * gimple-fold.c (gimple_fold_builtin_memory_op): Allow folding
3395 memcpy if the size is not more than MOVE_MAX * MOVE_RATIO.
3397 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
3400 * config/arm/arm.md (movmisaligndi): New define_expand.
3401 * config/arm/vec-common.md (movmisalign<mode>): Iterate over VDQ mode.
3403 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
3406 * emit-rtl.c (gen_highpart): Use adjust_address to handle
3407 MEM rather than calling simplify_gen_subreg.
3409 2021-09-13 Jan-Benedict Glaw <jbglaw@ług-owl.de>
3411 * config/alpha/vms.h (INIT_CUMULATIVE_ARGS): Wrap multi-statment
3412 define into a block.
3414 2021-09-13 Richard Biener <rguenther@suse.de>
3416 * config/darwin.h (DARWIN_PREFER_DWARF): Do not define.
3417 * config/i386/darwin.h (PREFERRED_DEBUGGING_TYPE): Do not
3418 change based on DARWIN_PREFER_DWARF not being defined.
3420 2021-09-13 Richard Biener <rguenther@suse.de>
3422 * config/i386/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
3423 to inherit from elfos.h
3425 2021-09-13 Richard Biener <rguenther@suse.de>
3427 * config.gcc: Add cr16-*-* to the list of obsoleted targets.
3429 2021-09-13 Richard Biener <rguenther@suse.de>
3431 * config/avr/elf.h (PREFERRED_DEBUGGING_TYPE): Remove
3432 override, pick up DWARF2_DEBUG define from elfos.h
3434 2021-09-13 Richard Biener <rguenther@suse.de>
3436 * config/rx/rx.h (PREFERRED_DEBUGGING_TYPE): Always define to
3439 2021-09-13 Richard Biener <rguenther@suse.de>
3441 * config/alpha/vms.h (PREFERRED_DEBUGGING_TYPE): Define to
3444 2021-09-13 Richard Biener <rguenther@suse.de>
3446 * config/i386/cygming.h: Always default to DWARF2 debugging.
3447 Do not define DBX_DEBUGGING_INFO, that's done via dbxcoff.h
3449 * doc/install.texi: Document binutils 2.16 as minimum
3450 requirement for mingw.
3452 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
3454 * config/rs6000/rs6000.c (struct rs6000_cost_data): New members
3455 nstmts, nloads and extra_ctor_cost.
3456 (rs6000_density_test): Add load density related heuristics. Do
3457 extra costing on vector construction statements if need.
3458 (rs6000_init_cost): Init new members.
3459 (rs6000_update_target_cost_per_stmt): New function.
3460 (rs6000_add_stmt_cost): Factor vect_nonmem hunk out to function
3461 rs6000_update_target_cost_per_stmt and call it.
3463 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
3465 * config/rs6000/rs6000.c (struct rs6000_cost_data): Remove typedef.
3466 (rs6000_init_cost): Adjust.
3468 2021-09-13 liuhongt <hongtao.liu@intel.com>
3470 * config/i386/i386.md: (UNSPEC_COPYSIGN): Remove.
3471 (UNSPEC_XORSIGN): Ditto.
3473 2021-09-12 Roger Sayle <roger@nextmovesoftware.com>
3475 * expr.c (convert_move): Preserve SUBREG_PROMOTED_VAR_P when
3476 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
3479 2021-09-11 Aldy Hernandez <aldyh@redhat.com>
3481 * tree-ssa-threadbackward.c (class back_threader_registry): Use
3482 back_jt_path_registry.
3483 * tree-ssa-threadedge.c (jump_threader::jump_threader): Use
3484 fwd_jt_path_registry.
3485 * tree-ssa-threadedge.h (class jump_threader): Same..
3486 * tree-ssa-threadupdate.c
3487 (jump_thread_path_registry::jump_thread_path_registry): Rename...
3488 (jt_path_registry::jt_path_registry): ...to this.
3489 (jump_thread_path_registry::~jump_thread_path_registry): Rename...
3490 (jt_path_registry::~jt_path_registry): ...this.
3491 (fwd_jt_path_registry::fwd_jt_path_registry): New.
3492 (fwd_jt_path_registry::~fwd_jt_path_registry): New.
3493 (jump_thread_path_registry::allocate_thread_edge): Rename...
3494 (jt_path_registry::allocate_thread_edge): ...to this.
3495 (jump_thread_path_registry::allocate_thread_path): Rename...
3496 (jt_path_registry::allocate_thread_path): ...to this.
3497 (jump_thread_path_registry::lookup_redirection_data): Rename...
3498 (fwd_jt_path_registry::lookup_redirection_data): ...to this.
3499 (jump_thread_path_registry::thread_block_1): Rename...
3500 (fwd_jt_path_registry::thread_block_1): ...to this.
3501 (jump_thread_path_registry::thread_block): Rename...
3502 (fwd_jt_path_registry::thread_block): ...to this.
3503 (jt_path_registry::thread_through_loop_header): Rename...
3504 (fwd_jt_path_registry::thread_through_loop_header): ...to this.
3505 (jump_thread_path_registry::mark_threaded_blocks): Rename...
3506 (fwd_jt_path_registry::mark_threaded_blocks): ...to this.
3507 (jump_thread_path_registry::debug_path): Rename...
3508 (jt_path_registry::debug_path): ...to this.
3509 (jump_thread_path_registry::dump): Rename...
3510 (jt_path_registry::debug): ...to this.
3511 (jump_thread_path_registry::rewire_first_differing_edge): Rename...
3512 (back_jt_path_registry::rewire_first_differing_edge): ...to this.
3513 (jump_thread_path_registry::adjust_paths_after_duplication): Rename...
3514 (back_jt_path_registry::adjust_paths_after_duplication): ...to this.
3515 (jump_thread_path_registry::duplicate_thread_path): Rename...
3516 (back_jt_path_registry::duplicate_thread_path): ...to this. Also,
3517 drop ill-formed candidates.
3518 (jump_thread_path_registry::remove_jump_threads_including): Rename...
3519 (fwd_jt_path_registry::remove_jump_threads_including): ...to this.
3520 (jt_path_registry::thread_through_all_blocks): New.
3521 (back_jt_path_registry::update_cfg): New.
3522 (fwd_jt_path_registry::update_cfg): New.
3523 (jump_thread_path_registry::register_jump_thread): Rename...
3524 (jt_path_registry::register_jump_thread): ...to this.
3525 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
3527 (class jt_path_registry): ...here.
3528 (class fwd_jt_path_registry): New.
3529 (class back_jt_path_registry): New.
3531 2021-09-10 liuhongt <hongtao.liu@intel.com>
3534 2021-09-01 liuhongt <hongtao.liu@intel.com>
3536 * emit-rtl.c (validate_subreg): Get rid of all float-int
3539 2021-09-10 Jakub Jelinek <jakub@redhat.com>
3541 * tree-core.h (enum omp_memory_order): Add OMP_MEMORY_ORDER_MASK,
3542 OMP_FAIL_MEMORY_ORDER_UNSPECIFIED, OMP_FAIL_MEMORY_ORDER_RELAXED,
3543 OMP_FAIL_MEMORY_ORDER_ACQUIRE, OMP_FAIL_MEMORY_ORDER_RELEASE,
3544 OMP_FAIL_MEMORY_ORDER_ACQ_REL, OMP_FAIL_MEMORY_ORDER_SEQ_CST and
3545 OMP_FAIL_MEMORY_ORDER_MASK enumerators.
3546 (OMP_FAIL_MEMORY_ORDER_SHIFT): Define.
3547 * gimple-pretty-print.c (dump_gimple_omp_atomic_load,
3548 dump_gimple_omp_atomic_store): Print [weak] for weak atomic
3550 * gimple.h (enum gf_mask): Change GF_OMP_ATOMIC_MEMORY_ORDER
3551 to 6-bit mask, adjust GF_OMP_ATOMIC_NEED_VALUE value and add
3553 (gimple_omp_atomic_weak_p, gimple_omp_atomic_set_weak): New inline
3555 * tree.h (OMP_ATOMIC_WEAK): Define.
3556 * tree-pretty-print.c (dump_omp_atomic_memory_order): Adjust for
3557 fail memory order being encoded in the same enum and also print
3558 fail clause if present.
3559 (dump_generic_node): Print weak clause if OMP_ATOMIC_WEAK.
3560 * gimplify.c (goa_stabilize_expr): Add target_expr and rhs arguments,
3561 handle pre_p == NULL case as a test mode that only returns value
3562 but doesn't change gimplify nor change anything otherwise, adjust
3563 recursive calls, add MODIFY_EXPR, ADDR_EXPR, COND_EXPR, TARGET_EXPR
3564 and CALL_EXPR handling, adjust COMPOUND_EXPR handling for
3565 __builtin_clear_padding calls, for !rhs gimplify as lvalue rather
3567 (gimplify_omp_atomic): Adjust goa_stabilize_expr caller. Handle
3568 COND_EXPR rhs. Set weak flag on gimple load/store for
3570 * omp-expand.c (omp_memory_order_to_fail_memmodel): New function.
3571 (omp_memory_order_to_memmodel): Adjust for fail clause encoded
3573 (expand_omp_atomic_cas): New function.
3574 (expand_omp_atomic_pipeline): Use omp_memory_order_to_fail_memmodel
3576 (expand_omp_atomic): Attempt to optimize atomic compare and exchange
3577 using expand_omp_atomic_cas.
3579 2021-09-10 Aldy Hernandez <aldyh@redhat.com>
3580 Michael Matz <matz@suse.de>
3582 * tree-pass.h (PROP_loop_opts_done): New.
3583 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
3584 Intersect with global range.
3585 * tree-ssa-loop.c (tree_ssa_loop_done): Set PROP_loop_opts_done.
3586 * tree-ssa-threadbackward.c
3587 (back_threader_profitability::profitable_path_p): Disable
3588 threading through latches until after loop optimizations have run.
3590 2021-09-10 David Faust <david.faust@oracle.com>
3592 * doc/invoke.texi: Document BPF -mcpu, -mjmpext, -mjmp32 and -malu32
3595 2021-09-10 David Faust <david.faust@oracle.com>
3597 * config/bpf/bpf-opts.h (bpf_isa_version): New enum.
3598 * config/bpf/bpf-protos.h (bpf_expand_cbranch): New.
3599 * config/bpf/bpf.c (bpf_option_override): Handle -mcpu option.
3600 (bpf_expand_cbranch): New function.
3601 * config/bpf/bpf.md (AM mode iterator): Conditionalize support for SI
3603 (zero_extendsidi2): Only use mov32 instruction if it is available.
3604 (SIM mode iterator): Conditionalize support for SI mode.
3605 (JM mode iterator): New.
3606 (cbranchdi4): Update name, use new JM iterator. Use bpf_expand_cbranch.
3607 (*branch_on_di): Update name, use new JM iterator.
3608 * config/bpf/bpf.opt: (mjmpext): New option.
3612 (bpf_isa): New enum.
3614 2021-09-10 David Faust <david.faust@oracle.com>
3616 * config/bpf/bpf.md (zero_extendhidi2): Add new output template
3617 for register-to-register extensions.
3618 (zero_extendqidi2): Likewise.
3620 2021-09-10 Richard Biener <rguenther@suse.de>
3622 PR middle-end/102273
3623 * internal-fn.c (expand_DEFERRED_INIT): Always expand non-SSA vars.
3625 2021-09-10 Richard Biener <rguenther@suse.de>
3627 PR middle-end/102269
3628 * gimplify.c (is_var_need_auto_init): Empty types do not need
3631 2021-09-10 Richard Biener <rguenther@suse.de>
3633 * configure.ac (--with-stabs): Remove.
3634 * configure: Regenerate.
3635 * doc/install.texi: Remove --with-stabs documentation.
3637 2021-09-10 liuhongt <hongtao.liu@intel.com>
3639 * config/i386/avx512fp16intrin.h: (_mm512_cmp_ph_mask):
3641 (_mm512_mask_cmp_ph_mask): Likewise.
3642 (_mm512_cmp_round_ph_mask): Likewise.
3643 (_mm512_mask_cmp_round_ph_mask): Likewise.
3644 (_mm_cmp_sh_mask): Likewise.
3645 (_mm_mask_cmp_sh_mask): Likewise.
3646 (_mm_cmp_round_sh_mask): Likewise.
3647 (_mm_mask_cmp_round_sh_mask): Likewise.
3648 (_mm_comieq_sh): Likewise.
3649 (_mm_comilt_sh): Likewise.
3650 (_mm_comile_sh): Likewise.
3651 (_mm_comigt_sh): Likewise.
3652 (_mm_comige_sh): Likewise.
3653 (_mm_comineq_sh): Likewise.
3654 (_mm_ucomieq_sh): Likewise.
3655 (_mm_ucomilt_sh): Likewise.
3656 (_mm_ucomile_sh): Likewise.
3657 (_mm_ucomigt_sh): Likewise.
3658 (_mm_ucomige_sh): Likewise.
3659 (_mm_ucomineq_sh): Likewise.
3660 (_mm_comi_round_sh): Likewise.
3661 (_mm_comi_sh): Likewise.
3662 * config/i386/avx512fp16vlintrin.h (_mm_cmp_ph_mask): New intrinsic.
3663 (_mm_mask_cmp_ph_mask): Likewise.
3664 (_mm256_cmp_ph_mask): Likewise.
3665 (_mm256_mask_cmp_ph_mask): Likewise.
3666 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3667 * config/i386/i386-builtin.def: Add corresponding new builtins.
3668 * config/i386/i386-expand.c
3669 (ix86_expand_args_builtin): Handle new builtin types.
3670 (ix86_expand_round_builtin): Ditto.
3671 * config/i386/i386.md (ssevecmode): Add HF mode.
3672 (MODEFH): New mode iterator.
3673 * config/i386/sse.md
3674 (V48H_AVX512VL): New mode iterator to support HF vector modes.
3675 Ajdust corresponding description.
3676 (ssecmpintprefix): New.
3677 (VI12_AVX512VL): Adjust to support HF vector modes.
3678 (cmp_imm_predicate): Likewise.
3679 (<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
3681 (avx512f_vmcmp<mode>3<round_saeonly_name>): Likewise.
3682 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Likewise.
3683 (<sse>_<unord>comi<round_saeonly_name>): Likewise.
3685 2021-09-10 liuhongt <hongtao.liu@intel.com>
3687 * config/i386/avx512fp16intrin.h: (_mm512_max_ph): New intrinsic.
3688 (_mm512_mask_max_ph): Likewise.
3689 (_mm512_maskz_max_ph): Likewise.
3690 (_mm512_min_ph): Likewise.
3691 (_mm512_mask_min_ph): Likewise.
3692 (_mm512_maskz_min_ph): Likewise.
3693 (_mm512_max_round_ph): Likewise.
3694 (_mm512_mask_max_round_ph): Likewise.
3695 (_mm512_maskz_max_round_ph): Likewise.
3696 (_mm512_min_round_ph): Likewise.
3697 (_mm512_mask_min_round_ph): Likewise.
3698 (_mm512_maskz_min_round_ph): Likewise.
3699 (_mm_max_sh): Likewise.
3700 (_mm_mask_max_sh): Likewise.
3701 (_mm_maskz_max_sh): Likewise.
3702 (_mm_min_sh): Likewise.
3703 (_mm_mask_min_sh): Likewise.
3704 (_mm_maskz_min_sh): Likewise.
3705 (_mm_max_round_sh): Likewise.
3706 (_mm_mask_max_round_sh): Likewise.
3707 (_mm_maskz_max_round_sh): Likewise.
3708 (_mm_min_round_sh): Likewise.
3709 (_mm_mask_min_round_sh): Likewise.
3710 (_mm_maskz_min_round_sh): Likewise.
3711 * config/i386/avx512fp16vlintrin.h (_mm_max_ph): New intrinsic.
3712 (_mm256_max_ph): Likewise.
3713 (_mm_mask_max_ph): Likewise.
3714 (_mm256_mask_max_ph): Likewise.
3715 (_mm_maskz_max_ph): Likewise.
3716 (_mm256_maskz_max_ph): Likewise.
3717 (_mm_min_ph): Likewise.
3718 (_mm256_min_ph): Likewise.
3719 (_mm_mask_min_ph): Likewise.
3720 (_mm256_mask_min_ph): Likewise.
3721 (_mm_maskz_min_ph): Likewise.
3722 (_mm256_maskz_min_ph): Likewise.
3723 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3724 * config/i386/i386-builtin.def: Add corresponding new builtins.
3725 * config/i386/i386-expand.c
3726 (ix86_expand_args_builtin): Handle new builtin types.
3727 * config/i386/sse.md
3728 (<code><mode>3<mask_name><round_saeonly_name>): Adjust to
3729 support HF vector modes.
3730 (*<code><mode>3<mask_name><round_saeonly_name>): Likewise.
3731 (ieee_<ieee_maxmin><mode>3<mask_name><round_saeonly_name>):
3733 (<sse>_vm<code><mode>3<mask_scalar_name><round_saeonly_scalar_name>):
3735 * config/i386/subst.md (round_saeonly_mode512bit_condition):
3736 Adjust for HF vector modes.
3738 2021-09-10 Liu, Hongtao <hongtao.liu@intel.com>
3740 * config/i386/avx512fp16intrin.h (_mm_add_sh): New intrinsic.
3741 (_mm_mask_add_sh): Likewise.
3742 (_mm_maskz_add_sh): Likewise.
3743 (_mm_sub_sh): Likewise.
3744 (_mm_mask_sub_sh): Likewise.
3745 (_mm_maskz_sub_sh): Likewise.
3746 (_mm_mul_sh): Likewise.
3747 (_mm_mask_mul_sh): Likewise.
3748 (_mm_maskz_mul_sh): Likewise.
3749 (_mm_div_sh): Likewise.
3750 (_mm_mask_div_sh): Likewise.
3751 (_mm_maskz_div_sh): Likewise.
3752 (_mm_add_round_sh): Likewise.
3753 (_mm_mask_add_round_sh): Likewise.
3754 (_mm_maskz_add_round_sh): Likewise.
3755 (_mm_sub_round_sh): Likewise.
3756 (_mm_mask_sub_round_sh): Likewise.
3757 (_mm_maskz_sub_round_sh): Likewise.
3758 (_mm_mul_round_sh): Likewise.
3759 (_mm_mask_mul_round_sh): Likewise.
3760 (_mm_maskz_mul_round_sh): Likewise.
3761 (_mm_div_round_sh): Likewise.
3762 (_mm_mask_div_round_sh): Likewise.
3763 (_mm_maskz_div_round_sh): Likewise.
3764 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3765 * config/i386/i386-builtin.def: Add corresponding new builtins.
3766 * config/i386/i386-expand.c
3767 (ix86_expand_round_builtin): Handle new builtins.
3768 * config/i386/sse.md (VF_128): Change description.
3769 (<sse>_vm<plusminus_insn><mode>3<mask_scalar_name><round_scalar_name>):
3770 Adjust to support HF vector modes.
3771 (<sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name><round_scalar_name>):
3774 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
3776 * config/i386/i386-expand.c
3777 (ix86_avx256_split_vector_move_misalign): Handle V16HF mode.
3778 * config/i386/i386.c
3779 (ix86_preferred_simd_mode): Handle HF mode.
3780 * config/i386/sse.md (V_256H): New mode iterator.
3781 (avx_vextractf128<mode>): Use it.
3782 (VEC_INIT_MODE): Align vector HFmode condition to vector
3783 HImodes since there're no real HF instruction used.
3784 (VEC_INIT_HALF_MODE): Ditto.
3786 (VIHF_AVX512BW): Ditto.
3787 (*vec_extracthf): Ditto.
3788 (VEC_EXTRACT_MODE): Ditto.
3790 2021-09-10 Richard Biener <rguenther@suse.de>
3793 * config/dbx.h: Remove.
3794 * config/dbxcoff.h: Do not define PREFERRED_DEBUGGING_TYPE.
3795 * config/lynx.h: Likewise.
3797 2021-09-10 liuhongt <hongtao.liu@intel.com>
3799 * config/i386/i386-expand.c (ix86_expand_copysign): Expand
3800 right into ANDNOT + AND + IOR, using paradoxical subregs.
3801 (ix86_split_copysign_const): Remove.
3802 (ix86_split_copysign_var): Ditto.
3803 * config/i386/i386-protos.h (ix86_split_copysign_const): Dotto.
3804 (ix86_split_copysign_var): Ditto.
3805 * config/i386/i386.md (@copysign<mode>3_const): Ditto.
3806 (@copysign<mode>3_var): Ditto.
3808 2021-09-09 qing zhao <qing.zhao@oracle.com>
3810 * builtins.c (expand_builtin_memset): Make external visible.
3811 * builtins.h (expand_builtin_memset): Declare extern.
3812 * common.opt (ftrivial-auto-var-init=): New option.
3813 * doc/extend.texi: Document the uninitialized attribute.
3814 * doc/invoke.texi: Document -ftrivial-auto-var-init.
3815 * flag-types.h (enum auto_init_type): New enumerated type
3817 * gimple-fold.c (clear_padding_type): Add one new parameter.
3818 (clear_padding_union): Likewise.
3819 (clear_padding_emit_loop): Likewise.
3820 (clear_type_padding_in_mask): Likewise.
3821 (gimple_fold_builtin_clear_padding): Handle this new parameter.
3822 * gimplify.c (gimple_add_init_for_auto_var): New function.
3823 (gimple_add_padding_init_for_auto_var): New function.
3824 (is_var_need_auto_init): New function.
3825 (gimplify_decl_expr): Add initialization to automatic variables per
3827 (gimplify_call_expr): Add one new parameter for call to
3828 __builtin_clear_padding.
3829 (gimplify_init_constructor): Add padding initialization in the end.
3830 * internal-fn.c (INIT_PATTERN_VALUE): New macro.
3831 (expand_DEFERRED_INIT): New function.
3832 * internal-fn.def (DEFERRED_INIT): New internal function.
3833 * tree-cfg.c (verify_gimple_call): Verify calls to .DEFERRED_INIT.
3834 * tree-sra.c (generate_subtree_deferred_init): New function.
3835 (scan_function): Avoid setting cannot_scalarize_away_bitmap for
3836 calls to .DEFERRED_INIT.
3837 (sra_modify_deferred_init): New function.
3838 (sra_modify_function_body): Handle calls to DEFERRED_INIT specially.
3839 * tree-ssa-structalias.c (find_func_aliases_for_call): Likewise.
3840 * tree-ssa-uninit.c (warn_uninit): Handle calls to DEFERRED_INIT
3842 (check_defs): Likewise.
3843 (warn_uninitialized_vars): Likewise.
3844 * tree-ssa.c (ssa_undefined_value_p): Likewise.
3845 * tree.c (build_common_builtin_nodes): Build tree node for
3846 BUILT_IN_CLEAR_PADDING when needed.
3848 2021-09-09 Richard Biener <rguenther@suse.de>
3850 * tree-ssa-loop-im.c (fill_always_executed_in_1): Walk
3853 2021-09-09 Richard Biener <rguenther@suse.de>
3855 * tree-ssa-loop-im.c (fill_always_executed_in_1): Integrate
3856 DOM walk from get_loop_body_in_dom_order using a worklist
3859 2021-09-09 liuhongt <hongtao.liu@intel.com>
3861 * config.gcc: Add avx512fp16vlintrin.h.
3862 * config/i386/avx512fp16intrin.h: (_mm512_add_ph): New intrinsic.
3863 (_mm512_mask_add_ph): Likewise.
3864 (_mm512_maskz_add_ph): Likewise.
3865 (_mm512_sub_ph): Likewise.
3866 (_mm512_mask_sub_ph): Likewise.
3867 (_mm512_maskz_sub_ph): Likewise.
3868 (_mm512_mul_ph): Likewise.
3869 (_mm512_mask_mul_ph): Likewise.
3870 (_mm512_maskz_mul_ph): Likewise.
3871 (_mm512_div_ph): Likewise.
3872 (_mm512_mask_div_ph): Likewise.
3873 (_mm512_maskz_div_ph): Likewise.
3874 (_mm512_add_round_ph): Likewise.
3875 (_mm512_mask_add_round_ph): Likewise.
3876 (_mm512_maskz_add_round_ph): Likewise.
3877 (_mm512_sub_round_ph): Likewise.
3878 (_mm512_mask_sub_round_ph): Likewise.
3879 (_mm512_maskz_sub_round_ph): Likewise.
3880 (_mm512_mul_round_ph): Likewise.
3881 (_mm512_mask_mul_round_ph): Likewise.
3882 (_mm512_maskz_mul_round_ph): Likewise.
3883 (_mm512_div_round_ph): Likewise.
3884 (_mm512_mask_div_round_ph): Likewise.
3885 (_mm512_maskz_div_round_ph): Likewise.
3886 * config/i386/avx512fp16vlintrin.h: New header.
3887 * config/i386/i386-builtin-types.def (V16HF, V8HF, V32HF):
3888 Add new builtin types.
3889 * config/i386/i386-builtin.def: Add corresponding builtins.
3890 * config/i386/i386-expand.c
3891 (ix86_expand_args_builtin): Handle new builtin types.
3892 (ix86_expand_round_builtin): Likewise.
3893 * config/i386/immintrin.h: Include avx512fp16vlintrin.h
3894 * config/i386/sse.md (VFH): New mode_iterator.
3896 (avx512fmaskmode): Add HF vector modes.
3897 (avx512fmaskhalfmode): Likewise.
3898 (<plusminus_insn><mode>3<mask_name><round_name>): Adjust to for
3900 (*<plusminus_insn><mode>3<mask_name><round_name>): Likewise.
3901 (mul<mode>3<mask_name><round_name>): Likewise.
3902 (*mul<mode>3<mask_name><round_name>): Likewise.
3903 (div<mode>3): Likewise.
3904 (<sse>_div<mode>3<mask_name><round_name>): Likewise.
3905 * config/i386/subst.md (SUBST_V): Add HF vector modes.
3906 (SUBST_A): Likewise.
3907 (round_mode512bit_condition): Adjust for V32HFmode.
3909 2021-09-09 liuhongt <hongtao.liu@intel.com>
3912 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
3913 (reduc_plus_scal_v4sf): .. this, New define_expand.
3914 (reduc_plus_scal_v2df): .. and this, New define_expand.
3916 2021-09-09 liuhongt <hongtao.liu@intel.com>
3919 * config/i386/sse.md (*vec_extract<mode><ssescalarmodelower>_valign):
3922 2021-09-08 Jonathan Wakely <jwakely@redhat.com>
3925 * doc/trouble.texi (Copy Assignment): Fix description of
3926 behaviour and fix code in example.
3928 2021-09-08 Segher Boessenkool <segher@kernel.crashing.org>
3931 * config/rs6000/rs6000-logue.c (rs6000_emit_epilogue): For ELFv2 use
3932 r11 instead of r12 for restoring CR.
3934 2021-09-08 Jakub Jelinek <jakub@redhat.com>
3935 liuhongt <hongtao.liu@intel.com>
3938 * config/i386/i386.md (@xorsign<mode>3_1): Remove.
3939 * config/i386/i386-expand.c (ix86_expand_xorsign): Expand right away
3940 into AND with mask and XOR, using paradoxical subregs.
3941 (ix86_split_xorsign): Remove.
3942 * config/i386/i386-protos.h (ix86_split_xorsign): Remove.
3944 2021-09-08 Di Zhao <dizhao@os.amperecomputing.com>
3946 * tree-ssa-sccvn.c (vn_nary_op_insert_into): fix result compare
3948 2021-09-08 Jakub Jelinek <jakub@redhat.com>
3951 * config/i386/i386.md (xorsign<mode>3): If operands[1] is equal to
3952 operands[2], emit abs<mode>2 instead.
3953 (@xorsign<mode>3_1): Add early-clobbers for output operand, enable
3954 first alternative even for avx, add another alternative with
3955 =&Yv <- 0, Yv, Yvm constraints.
3956 * config/i386/i386-expand.c (ix86_split_xorsign): If op0 is equal
3957 to op1, emit vpandn instead.
3959 2021-09-08 liuhongt <hongtao.liu@intel.com>
3961 * config/i386/avx512fp16intrin.h (_mm_set_ph): New intrinsic.
3962 (_mm256_set_ph): Likewise.
3963 (_mm512_set_ph): Likewise.
3964 (_mm_setr_ph): Likewise.
3965 (_mm256_setr_ph): Likewise.
3966 (_mm512_setr_ph): Likewise.
3967 (_mm_set1_ph): Likewise.
3968 (_mm256_set1_ph): Likewise.
3969 (_mm512_set1_ph): Likewise.
3970 (_mm_setzero_ph): Likewise.
3971 (_mm256_setzero_ph): Likewise.
3972 (_mm512_setzero_ph): Likewise.
3973 (_mm_set_sh): Likewise.
3974 (_mm_load_sh): Likewise.
3975 (_mm_store_sh): Likewise.
3976 * config/i386/i386-builtin-types.def (V8HF): New type.
3977 (DEF_FUNCTION_TYPE (V8HF, V8HI)): New builtin function type
3978 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
3979 Support vector HFmodes.
3980 (ix86_expand_vector_init_one_nonzero): Likewise.
3981 (ix86_expand_vector_init_one_var): Likewise.
3982 (ix86_expand_vector_init_interleave): Likewise.
3983 (ix86_expand_vector_init_general): Likewise.
3984 (ix86_expand_vector_set): Likewise.
3985 (ix86_expand_vector_extract): Likewise.
3986 (ix86_expand_vector_init_concat): Likewise.
3987 (ix86_expand_sse_movcc): Handle vector HFmodes.
3988 (ix86_expand_vector_set_var): Ditto.
3989 * config/i386/i386-modes.def: Add HF vector modes in comment.
3990 * config/i386/i386.c (classify_argument): Add HF vector modes.
3991 (ix86_hard_regno_mode_ok): Allow HF vector modes for AVX512FP16.
3992 (ix86_vector_mode_supported_p): Likewise.
3993 (ix86_set_reg_reg_cost): Handle vector HFmode.
3994 (ix86_get_ssemov): Handle vector HFmode.
3995 (function_arg_advance_64): Pass unamed V16HFmode and V32HFmode
3997 (function_arg_advance_32): Pass V8HF/V16HF/V32HF by sse reg for 32bit
3999 (function_arg_advance_32): Ditto.
4000 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): New.
4001 (VALID_AVX256_REG_OR_OI_MODE): Rename to ..
4002 (VALID_AVX256_REG_OR_OI_VHF_MODE): .. this, and add V16HF.
4003 (VALID_SSE2_REG_VHF_MODE): New.
4004 (VALID_AVX512VL_128_REG_MODE): Add V8HF and TImode.
4005 (SSE_REG_MODE_P): Add vector HFmode.
4006 * config/i386/i386.md (mode): Add HF vector modes.
4007 (MODE_SIZE): Likewise.
4008 (ssemodesuffix): Add ph suffix for HF vector modes.
4009 * config/i386/sse.md (VFH_128): New mode iterator.
4010 (VMOVE): Adjust for HF vector modes.
4012 (V_256_512): Likewise.
4014 (avx512fmaskmode): Likewise.
4015 (shuffletype): Likewise.
4016 (sseinsnmode): Likewise.
4017 (ssedoublevecmode): Likewise.
4018 (ssehalfvecmode): Likewise.
4019 (ssehalfvecmodelower): Likewise.
4020 (ssePScmode): Likewise.
4021 (ssescalarmode): Likewise.
4022 (ssescalarmodelower): Likewise.
4023 (sseintprefix): Likewise.
4025 (bcstscalarsuff): Likewise.
4026 (xtg_mode): Likewise.
4027 (VI12HF_AVX512VL): New mode_iterator.
4028 (VF_AVX512FP16): Likewise.
4030 (VIHF_256): Likewise.
4031 (VIHF_AVX512BW): Likewise.
4032 (V16_256): Likewise.
4033 (V32_512): Likewise.
4034 (sseintmodesuffix): New mode_attr.
4035 (sse): Add scalar and vector HFmodes.
4036 (ssescalarmode): Add vector HFmode mapping.
4037 (ssescalarmodesuffix): Add sh suffix for HFmode.
4038 (*<sse>_vm<insn><mode>3): Use VFH_128.
4039 (*<sse>_vm<multdiv_mnemonic><mode>3): Likewise.
4040 (*ieee_<ieee_maxmin><mode>3): Likewise.
4041 (<avx512>_blendm<mode>): New define_insn.
4042 (vec_setv8hf): New define_expand.
4043 (vec_set<mode>_0): New define_insn for HF vector set.
4044 (*avx512fp16_movsh): Likewise.
4045 (avx512fp16_movsh): Likewise.
4046 (vec_extract_lo_v32hi): Rename to ...
4047 (vec_extract_lo_<mode>): ... this, and adjust to allow HF
4049 (vec_extract_hi_v32hi): Likewise.
4050 (vec_extract_hi_<mode>): Likewise.
4051 (vec_extract_lo_v16hi): Likewise.
4052 (vec_extract_lo_<mode>): Likewise.
4053 (vec_extract_hi_v16hi): Likewise.
4054 (vec_extract_hi_<mode>): Likewise.
4055 (vec_set_hi_v16hi): Likewise.
4056 (vec_set_hi_<mode>): Likewise.
4057 (vec_set_lo_v16hi): Likewise.
4058 (vec_set_lo_<mode>): Likewise.
4059 (*vec_extract<mode>_0): New define_insn_and_split for HF
4061 (*vec_extracthf): New define_insn.
4062 (VEC_EXTRACT_MODE): Add HF vector modes.
4063 (PINSR_MODE): Add V8HF.
4064 (sse2p4_1): Likewise.
4065 (pinsr_evex_isa): Likewise.
4066 (<sse2p4_1>_pinsr<ssemodesuffix>): Adjust to support
4067 insert for V8HFmode.
4068 (pbroadcast_evex_isa): Add HF vector modes.
4069 (AVX2_VEC_DUP_MODE): Likewise.
4070 (VEC_INIT_MODE): Likewise.
4071 (VEC_INIT_HALF_MODE): Likewise.
4072 (avx2_pbroadcast<mode>): Adjust to support HF vector mode
4074 (avx2_pbroadcast<mode>_1): Likewise.
4075 (<avx512>_vec_dup<mode>_1): Likewise.
4076 (<avx512>_vec_dup<mode><mask_name>): Likewise.
4077 (<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>):
4080 2021-09-08 Guo, Xuepeng <xuepeng.guo@intel.com>
4081 H.J. Lu <hongjiu.lu@intel.com>
4082 Liu Hongtao <hongtao.liu@intel.com>
4083 Wang Hongyu <hongyu.wang@intel.com>
4084 Xu Dianhong <dianhong.xu@intel.com>
4086 * common/config/i386/cpuinfo.h (get_available_features):
4087 Detect FEATURE_AVX512FP16.
4088 * common/config/i386/i386-common.c
4089 (OPTION_MASK_ISA_AVX512FP16_SET,
4090 OPTION_MASK_ISA_AVX512FP16_UNSET,
4091 OPTION_MASK_ISA2_AVX512FP16_SET,
4092 OPTION_MASK_ISA2_AVX512FP16_UNSET): New.
4093 (OPTION_MASK_ISA2_AVX512BW_UNSET,
4094 OPTION_MASK_ISA2_AVX512BF16_UNSET): Add AVX512FP16.
4095 (ix86_handle_option): Handle -mavx512fp16.
4096 * common/config/i386/i386-cpuinfo.h (enum processor_features):
4097 Add FEATURE_AVX512FP16.
4098 * common/config/i386/i386-isas.h: Add entry for AVX512FP16.
4099 * config.gcc: Add avx512fp16intrin.h.
4100 * config/i386/avx512fp16intrin.h: New intrinsic header.
4101 * config/i386/cpuid.h: Add bit_AVX512FP16.
4102 * config/i386/i386-builtin-types.def: (FLOAT16): New primitive type.
4103 * config/i386/i386-builtins.c: Support _Float16 type for i386
4105 (ix86_register_float16_builtin_type): New function.
4106 (ix86_float16_type_node): New.
4107 * config/i386/i386-c.c (ix86_target_macros_internal): Define
4109 * config/i386/i386-expand.c (ix86_expand_branch): Support
4111 (ix86_prepare_fp_compare_args): Adjust TARGET_SSE_MATH &&
4112 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
4113 (ix86_expand_fp_movcc): Ditto.
4114 * config/i386/i386-isa.def: Add PTA define for AVX512FP16.
4115 * config/i386/i386-options.c (isa2_opts): Add -mavx512fp16.
4116 (ix86_valid_target_attribute_inner_p): Add avx512fp16 attribute.
4117 * config/i386/i386.c (ix86_get_ssemov): Use
4118 vmovdqu16/vmovw/vmovsh for HFmode/HImode scalar or vector.
4119 (ix86_get_excess_precision): Use
4120 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when TARGET_AVX512FP16
4122 (sse_store_index): Use SFmode cost for HFmode cost.
4123 (inline_memory_move_cost): Add HFmode, and perfer SSE cost over
4124 GPR cost for HFmode.
4125 (ix86_hard_regno_mode_ok): Allow HImode in sse register.
4126 (ix86_mangle_type): Add manlging for _Float16 type.
4127 (inline_secondary_memory_needed): No memory is needed for
4128 16bit movement between gpr and sse reg under
4130 (ix86_multiplication_cost): Adjust TARGET_SSE_MATH &&
4131 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
4132 (ix86_division_cost): Ditto.
4133 (ix86_rtx_costs): Ditto.
4134 (ix86_add_stmt_cost): Ditto.
4135 (ix86_optab_supported_p): Ditto.
4136 * config/i386/i386.h (VALID_AVX512F_SCALAR_MODE): Add HFmode.
4137 (SSE_FLOAT_MODE_SSEMATH_OR_HF_P): Add HFmode.
4138 (PTA_SAPPHIRERAPIDS): Add PTA_AVX512FP16.
4139 * config/i386/i386.md (mode): Add HFmode.
4140 (MODE_SIZE): Add HFmode.
4141 (isa): Add avx512fp16.
4142 (enabled): Handle avx512fp16.
4143 (ssemodesuffix): Add sh suffix for HFmode.
4144 (comm): Add mult, div.
4145 (plusminusmultdiv): New code iterator.
4146 (insn): Add mult, div.
4147 (*movhf_internal): Adjust for avx512fp16 instruction.
4148 (*movhi_internal): Ditto.
4149 (*cmpi<unord>hf): New define_insn for HFmode.
4150 (*ieee_s<ieee_maxmin>hf3): Likewise.
4151 (extendhf<mode>2): Likewise.
4152 (trunc<mode>hf2): Likewise.
4153 (float<floatunssuffix><mode>hf2): Likewise.
4154 (*<insn>hf): Likewise.
4155 (cbranchhf4): New expander.
4156 (movhfcc): Likewise.
4157 (<insn>hf3): Likewise.
4160 * config/i386/i386.opt: Add mavx512fp16.
4161 * config/i386/immintrin.h: Include avx512fp16intrin.h.
4162 * doc/invoke.texi: Add mavx512fp16.
4163 * doc/extend.texi: Add avx512fp16 Usage Notes.
4165 2021-09-08 liuhongt <hongtao.liu@intel.com>
4167 * common.opt: Support -fexcess-precision=16.
4168 * config/aarch64/aarch64.c (aarch64_excess_precision): Return
4169 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when
4170 EXCESS_PRECISION_TYPE_FLOAT16.
4171 * config/arm/arm.c (arm_excess_precision): Ditto.
4172 * config/i386/i386.c (ix86_get_excess_precision): Ditto.
4173 * config/m68k/m68k.c (m68k_excess_precision): Issue an error
4174 when EXCESS_PRECISION_TYPE_FLOAT16.
4175 * config/s390/s390.c (s390_excess_precision): Ditto.
4176 * coretypes.h (enum excess_precision_type): Add
4177 EXCESS_PRECISION_TYPE_FLOAT16.
4178 * doc/tm.texi (TARGET_C_EXCESS_PRECISION): Update documents.
4179 * doc/tm.texi.in (TARGET_C_EXCESS_PRECISION): Ditto.
4180 * doc/extend.texi (Half-Precision): Document
4181 -fexcess-precision=16.
4182 * flag-types.h (enum excess_precision): Add
4183 EXCESS_PRECISION_FLOAT16.
4184 * target.def (excess_precision): Update document.
4185 * tree.c (excess_precision_type): Set excess_precision_type to
4186 EXCESS_PRECISION_FLOAT16 when -fexcess-precision=16.
4188 2021-09-08 liuhongt <hongtao.liu@intel.com>
4190 * doc/extend.texi: (@node Floating Types): Adjust the wording.
4191 (@node Half-Precision): Ditto.
4193 2021-09-07 Takayuki 'January June' Suwa <jjsuwa_sys3175@yahoo.co.jp>
4196 * config/xtensa/xtensa.c (xtensa_emit_move_sequence): Add
4197 'CONST_INT_P (src)' to the condition of the block that tries to
4198 eliminate literal when loading integer contant.
4200 2021-09-07 David Faust <david.faust@oracle.com>
4202 * doc/extend.texi (BPF Type Attributes) New node.
4203 Document new preserve_access_index attribute.
4204 Document new preserve_access_index builtin.
4205 * doc/invoke.texi: Document -mco-re and -mno-co-re options.
4207 2021-09-07 David Faust <david.faust@oracle.com>
4209 * config/bpf/bpf.c: Adjust includes.
4210 (bpf_handle_preserve_access_index_attribute): New function.
4211 (bpf_attribute_table): Use it here.
4212 (bpf_builtins): Add BPF_BUILTIN_PRESERVE_ACCESS_INDEX.
4213 (bpf_option_override): Handle "-mco-re" option.
4214 (bpf_asm_init_sections): New.
4215 (TARGET_ASM_INIT_SECTIONS): Redefine.
4216 (bpf_file_end): New.
4217 (TARGET_ASM_FILE_END): Redefine.
4218 (bpf_init_builtins): Add "__builtin_preserve_access_index".
4219 (bpf_core_compute, bpf_core_get_index): New.
4220 (is_attr_preserve_access): New.
4221 (bpf_expand_builtin): Handle new builtins.
4222 (bpf_core_newdecl, bpf_core_is_maybe_aggregate_access): New.
4223 (bpf_core_walk): New.
4224 (bpf_resolve_overloaded_builtin): New.
4225 (TARGET_RESOLVE_OVERLOADED_BUILTIN): Redefine.
4227 (pass_bpf_core_attr): New RTL pass.
4228 * config/bpf/bpf-passes.def: New file.
4229 * config/bpf/bpf-protos.h (make_pass_bpf_core_attr): New.
4230 * config/bpf/coreout.c: New file.
4231 * config/bpf/coreout.h: Likewise.
4232 * config/bpf/t-bpf (TM_H): Add $(srcdir)/config/bpf/coreout.h.
4233 (coreout.o): New rule.
4234 (PASSES_EXTRA): Add $(srcdir)/config/bpf/bpf-passes.def.
4235 * config.gcc (bpf): Add coreout.h to extra_headers.
4236 Add coreout.o to extra_objs.
4237 Add $(srcdir)/config/bpf/coreout.c to target_gtfiles.
4239 2021-09-07 David Faust <david.faust@oracle.com>
4241 * btfout.c (get_btf_id): Function is no longer static.
4242 * ctfc.h: Expose it here.
4244 2021-09-07 David Faust <david.faust@oracle.com>
4246 * ctfc.c (ctf_lookup_tree_type): New function.
4249 2021-09-07 David Faust <david.faust@oracle.com>
4251 * ctfc.c (ctf_dtd_lookup): Function is no longer static.
4252 * ctfc.h: Analogous change.
4254 2021-09-07 David Faust <david.faust@oracle.com>
4256 * dwarf2out.c (lookup_type_die): Function is no longer static.
4257 * dwarf2out.h: Expose it here.
4259 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
4261 * dwarf2ctf.c (ctf_debug_finalize): Make it static.
4262 (ctf_debug_early_finish): New definition.
4263 (ctf_debug_finish): Likewise.
4264 * dwarf2ctf.h (ctf_debug_finalize): Remove declaration.
4265 (ctf_debug_early_finish): New declaration.
4266 (ctf_debug_finish): Likewise.
4267 * dwarf2out.c (dwarf2out_finish): Invoke ctf_debug_finish.
4268 (dwarf2out_early_finish): Invoke ctf_debug_early_finish.
4270 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
4272 * config/bpf/bpf.c (bpf_option_override): For BPF backend, disable LTO
4273 support when compiling for CO-RE.
4274 * config/bpf/bpf.opt: Add new command line option -mco-re.
4276 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
4278 * flag-types.h (enum debug_info_type): Add new enum
4279 DINFO_TYPE_BTF_WITH_CORE.
4280 (BTF_WITH_CORE_DEBUG): New bitmask.
4281 * flags.h (btf_with_core_debuginfo_p): New declaration.
4282 * opts.c (btf_with_core_debuginfo_p): New definition.
4284 2021-09-07 Jason Merrill <jason@redhat.com>
4286 * tree.h (error_operand_p): Change to inline function.
4288 2021-09-07 Aldy Hernandez <aldyh@redhat.com>
4290 * tree-ssa-threadedge.c (forwarder_block_p): Rename to...
4291 (empty_block_with_phis_p): ...this.
4292 (potentially_threadable_block): Same.
4293 (jump_threader::thread_through_normal_block): Same.
4295 2021-09-07 Eric Botcazou <ebotcazou@adacore.com>
4298 * dwarf2out.c (mark_base_types): New overloaded function.
4299 (dwarf2out_early_finish): Invoke it on the COMDAT type list as well
4300 as the compilation unit, and call move_marked_base_types afterward.
4302 2021-09-07 H.J. Lu <hjl.tools@gmail.com>
4305 * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
4307 (ix86_expand_vector_convert_uns_vsivsf): Likewise.
4309 2021-09-07 Richard Biener <rguenther@suse.de>
4311 PR tree-optimization/102226
4312 * tree-vect-loop.c (vect_transform_cycle_phi): Record
4313 the converted value for the epilogue PHI use.
4315 2021-09-07 Martin Liska <mliska@suse.cz>
4317 PR gcov-profile/80223
4318 * ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
4319 options, do not inline when no_profile_instrument_function
4320 attributes are different in early inliner. It's fine to inline
4321 it after PGO instrumentation.
4323 2021-09-07 Richard Biener <rguenther@suse.de>
4325 PR tree-optimization/101555
4326 * tree-ssa-pre.c (translate_vuse_through_block): Do not
4327 perform an alias walk to determine the validity of the
4328 mem at the start of the block which is already guaranteed
4329 by means of prune_clobbered_mems.
4330 (phi_translate_1): Pass edge to translate_vuse_through_block.
4332 2021-09-07 Xionghu Luo <luoxhu@linux.ibm.com>
4335 * config/rs6000/rs6000.md (fmod<mode>3): New define_expand.
4336 (remainder<mode>3): Likewise.
4338 2021-09-07 YunQiang Su <yunqiang.su@cipunited.com>
4340 * config/mips/mips.c (mips_file_start): add .module for
4343 2021-09-06 Roger Sayle <roger@nextmovesoftware.com>
4345 * wide-int.cc (wi::clz): Reorder tests to ensure the result
4346 is zero for all negative values.
4348 2021-09-06 Tobias Burnus <tobias@codesourcery.com>
4350 * doc/invoke.texi (-foffload-options): Fix @opindex.
4352 2021-09-06 H.J. Lu <hjl.tools@gmail.com>
4355 * config/i386/i386-expand.c (ix86_split_xorsign): Use operands[2].
4356 * config/i386/i386.md (@xorsign<mode>3_1): Add non-destructive
4357 source alternative for AVX.
4359 2021-09-06 liuhongt <hongtao.liu@intel.com>
4361 PR middle-end/102182
4362 * optabs.c (expand_fix): Add from1 to avoid from being
4365 2021-09-06 Eric Botcazou <ebotcazou@adacore.com>
4367 * dwarf2out.c (modified_type_die): Deal with all array types earlier
4368 and use local variable consistently throughout the function.
4370 2021-09-06 Jakub Jelinek <jakub@redhat.com>
4372 PR tree-optimization/102207
4373 * match.pd: Don't demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
4374 were promoted from signed to wider unsigned type.
4376 2021-09-06 Andrew Pinski <apinski@marvell.com>
4378 PR tree-optimization/63184
4379 * match.pd: Add simplification of pointer_diff of two pointer_plus
4380 with addr_expr in the first operand of each pointer_plus.
4381 Add simplificatoin of ne/eq of two pointer_plus with addr_expr
4382 in the first operand of each pointer_plus.
4384 2021-09-06 Richard Biener <rguenther@suse.de>
4386 PR tree-optimization/102176
4387 * tree-vect-slp.c (vect_slp_gather_vectorized_scalar_stmts):
4389 (vect_bb_slp_scalar_cost): Use the computed set of
4390 vectorized scalar stmts instead of relying on the out-of-date
4391 and not accurate PURE_SLP_STMT.
4392 (vect_bb_vectorization_profitable_p): Compute the set
4393 of vectorized scalar stmts.
4395 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
4397 * gimple-range-path.cc (path_range_query::range_of_stmt): Remove
4398 GIMPLE_COND special casing.
4399 (path_range_query::range_defined_in_block): Use range_of_stmt
4400 instead of calling fold_range directly.
4402 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
4404 * gimple-range-path.cc (path_range_query::range_of_expr): Set
4405 m_undefined_path when appropriate.
4406 (path_range_query::internal_range_of_expr): Copy from range_of_expr.
4407 (path_range_query::unreachable_path_p): New.
4408 (path_range_query::precompute_ranges): Set m_undefined_path.
4409 * gimple-range-path.h (path_range_query::unreachable_path_p): New.
4410 (path_range_query::internal_range_of_expr): New.
4411 * tree-ssa-threadbackward.c (back_threader::find_taken_edge_cond):
4412 Use unreachable_path_p.
4414 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
4416 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
4417 Remove argument and call find_taken_edge.
4418 (back_threader::resolve_phi): Do not calculate taken edge before
4419 calling maybe_register_path.
4420 (back_threader::find_paths_to_names): Same.
4422 2021-09-05 Jeff Law <jlaw@localhost.localdomain>
4424 * config/h8300/h8300.md (QHSI2 mode iterator): New mode iterator.
4425 * config/h8300/testcompare.md (store_c): Update name, use new
4427 (store_neg_c, store_shifted_c): New patterns.
4429 2021-09-03 Segher Boessenkool <segher@kernel.crashing.org>
4432 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): On ELFv2 use r11
4433 instead of r12 for CR save, in all cases.
4435 2021-09-03 Andrew Pinski <apinski@marvell.com>
4437 * config/aarch64/aarch64-sve-builtins.cc (register_vector_type):
4438 Handle error_mark_node as the type of the type_decl.
4440 2021-09-03 Andrew Pinski <apinski@marvell.com>
4442 * config/aarch64/aarch64-builtins.c (struct aarch64_simd_type_info):
4444 (aarch64_simd_types): Likewise.
4445 (aarch64_simd_intOI_type_node): Likewise.
4446 (aarch64_simd_intCI_type_node): Likewise.
4447 (aarch64_simd_intXI_type_node): Likewise.
4448 * config/aarch64/aarch64.h (aarch64_fp16_type_node): Likewise.
4449 (aarch64_fp16_ptr_type_node): Likewise.
4450 (aarch64_bf16_type_node): Likewise.
4451 (aarch64_bf16_ptr_type_node): Likewise.
4453 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4455 * range-op.cc (operator_minus::op1_op2_relation_effect): Abstract
4457 (minus_op1_op2_relation_effect): ...here.
4458 (class operator_pointer_diff): New.
4459 (operator_pointer_diff::op1_op2_relation_effect): Call
4460 minus_op1_op2_relation_effect.
4461 (integral_table::integral_table): Add entry for POINTER_DIFF_EXPR.
4463 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4465 * tree-ssa-threadbackward.c (back_threader::thread_through_all_blocks):
4466 Add may_peel_loop_headers.
4467 (back_threader_registry::thread_through_all_blocks): Same.
4468 (try_thread_blocks): Pass may_peel_loop_headers argument.
4469 (pass_early_thread_jumps::execute): Same.
4471 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4473 * tree-ssa-threadedge.c (has_phis_p): New.
4474 (forwarder_block_p): New.
4475 (potentially_threadable_block): Call forwarder_block_p.
4476 (jump_threader::thread_around_empty_blocks): Call has_phis_p.
4477 (jump_threader::thread_through_normal_block): Call
4480 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4482 * tree-ssa-threadbackward.c (back_threader::dump): New.
4483 (back_threader::debug): New.
4484 (back_threader_profitability::profitable_path_p): Dump blocks
4485 even if we are bailing early.
4487 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4489 * tree-ssa-threadupdate.c (cancel_thread): New.
4490 (jump_thread_path_registry::thread_block_1): Use cancel_thread.
4491 (jump_thread_path_registry::mark_threaded_blocks): Same.
4492 (jump_thread_path_registry::register_jump_thread): Same.
4494 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4496 * tree-ssa-threadedge.c (jt_state::push): Only call methods for
4497 which objects are available.
4498 (jt_state::pop): Same.
4499 (jt_state::register_equiv): Same.
4500 (jt_state::register_equivs_on_edge): Same.
4502 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4504 * tree-ssa-threadedge.c (jump_threader::thread_across_edge):
4505 Move pop until after a thread is registered.
4507 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4509 * tree-ssa-threadupdate.c (debug): New.
4511 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4513 * gimple-range-trace.cc (push_dump_file::push_dump_file): New.
4514 (push_dump_file::~push_dump_file): New.
4515 (dump_ranger): Change dump_file temporarily while dumping
4517 * gimple-range-trace.h (class push_dump_file): New.
4519 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4521 * gimple-range-trace.cc (debug_seed_ranger): Remove static.
4522 (dump_ranger): Dump function name.
4524 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4526 * gimple-range-path.cc (path_range_query::range_defined_in_block):
4527 Adjust for non-null.
4528 (path_range_query::adjust_for_non_null_uses): New.
4529 (path_range_query::precompute_ranges): Call
4530 adjust_for_non_null_uses.
4531 * gimple-range-path.h: Add m_non_null and
4532 adjust_for_non_null_uses.
4534 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4536 * gimple-range-path.cc (path_range_query::dump): Dump path
4538 (path_range_query::precompute_ranges): Dump entire path.
4540 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4542 * value-relation.cc (relation_oracle::debug): New.
4543 * value-relation.h (relation_oracle::debug): New.
4545 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4547 * tree-ssa-loop-ch.c: Remove unnecessary include file.
4549 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4551 * gimple-range-fold.cc (fold_using_range::postfold_gcond_edges):
4552 Skip statements with no defining BB.
4553 * gimple-range-path.cc (path_range_query::range_defined_in_block):
4554 Do not get confused by statements with no defining BB.
4556 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4558 * gimple-range-fold.cc (adjust_imagpart_expr): Move from
4559 gimple_range_adjustment. Add support for constants.
4560 (adjust_realpart_expr): New.
4561 (gimple_range_adjustment): Move IMAGPART_EXPR code to
4562 adjust_imagpart_expr.
4563 * range-op.cc (integral_table::integral_table): Add entry for
4566 2021-09-03 Jakub Jelinek <jakub@redhat.com>
4568 * omp-expand.c (expand_omp_atomic_pipeline): Use
4569 IFN_ATOMIC_COMPARE_EXCHANGE instead of
4570 BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_? so that memory order
4573 2021-09-03 Jakub Jelinek <jakub@redhat.com>
4576 * tree.h (DECL_FIELD_ABI_IGNORED): Changed into rvalue only macro
4577 that is false if DECL_BIT_FIELD.
4578 (SET_DECL_FIELD_ABI_IGNORED, DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD,
4579 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD): Define.
4580 * tree-streamer-out.c (pack_ts_decl_common_value_fields): For
4581 DECL_BIT_FIELD stream DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead
4582 of DECL_FIELD_ABI_IGNORED.
4583 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
4584 SET_DECL_FIELD_ABI_IGNORED instead of writing to
4585 DECL_FIELD_ABI_IGNORED and for DECL_BIT_FIELD use
4586 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead.
4587 * lto-streamer-out.c (hash_tree): For DECL_BIT_FIELD hash
4588 DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead of DECL_FIELD_ABI_IGNORED.
4590 2021-09-03 liuhongt <hongtao.liu@intel.com>
4593 * config/i386/amxbf16intrin.h : Remove macro check for __AMX_BF16__.
4594 * config/i386/amxint8intrin.h : Remove macro check for __AMX_INT8__.
4595 * config/i386/amxtileintrin.h : Remove macro check for __AMX_TILE__.
4597 2021-09-02 Martin Sebor <msebor@redhat.com>
4599 PR tree-optimization/17506
4601 * tree-ssa-uninit.c (warn_uninit): Remove conditional guarding note.
4603 2021-09-02 Richard Biener <rguenther@suse.de>
4605 * tree-ssa-loop-im.c (fill_always_executed_in_1): Refine
4606 fix for PR78185 and continue processing when leaving
4609 2021-09-02 Jakub Jelinek <jakub@redhat.com>
4611 PR tree-optimization/99591
4612 * match.pd: Demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
4615 2021-09-02 Richard Biener <rguenther@suse.de>
4618 2021-09-02 Richard Biener <rguenther@suse.de>
4620 PR tree-optimization/102155
4621 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
4622 over a part of the RPO array and do not recurse here.
4623 Dump blocks marked as always executed.
4624 (fill_always_executed_in): Walk over the RPO array and
4625 process loops whose header we run into.
4626 (loop_invariant_motion_in_fun): Compute the first RPO
4627 using rev_post_order_and_mark_dfs_back_seme in iteration
4628 order and pass that to fill_always_executed_in.
4630 2021-09-02 liuhongt <hongtao.liu@intel.com>
4632 * config/i386/i386-modes.def (FLOAT_MODE): Define ieee HFmode.
4633 * config/i386/i386.c (enum x86_64_reg_class): Add
4635 (merge_classes): Handle X86_64_SSEHF_CLASS.
4636 (examine_argument): Ditto.
4637 (construct_container): Ditto.
4638 (classify_argument): Ditto, and set HFmode/HCmode to
4640 (function_value_32): Return _FLoat16/Complex Float16 by
4642 (function_value_64): Return _Float16/Complex Float16 by SSE
4644 (ix86_print_operand): Handle CONST_DOUBLE HFmode.
4645 (ix86_secondary_reload): Require gpr as intermediate register
4646 to store _Float16 from sse register when sse4 is not
4648 (ix86_libgcc_floating_mode_supported_p): Enable _FLoat16 under
4650 (ix86_scalar_mode_supported_p): Ditto.
4651 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Defined.
4652 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add HFmode.
4653 (VALID_INT_MODE_P): Add HFmode and HCmode.
4654 * config/i386/i386.md (*pushhf_rex64): New define_insn.
4656 (*movhf_internal): Ditto.
4657 * doc/extend.texi (Half-Precision Floating Point): Documemt
4660 2021-09-02 Richard Biener <rguenther@suse.de>
4662 PR tree-optimization/102155
4663 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
4664 over a part of the RPO array and do not recurse here.
4665 Dump blocks marked as always executed.
4666 (fill_always_executed_in): Walk over the RPO array and
4667 process loops whose header we run into.
4668 (loop_invariant_motion_in_fun): Compute the first RPO
4669 using rev_post_order_and_mark_dfs_back_seme in iteration
4670 order and pass that to fill_always_executed_in.
4672 2021-09-02 YunQiang Su <syq@debian.org>
4675 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
4677 * config/mips/mips.c (mips_module_isa_name): New.
4678 mips_file_start: add .module mipsREV to all asm output
4680 2021-09-01 Jeff Law <jlaw@localhost.localdomain>
4682 PR tree-optimization/102152
4683 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Reduce a vector
4684 comparison to a scalar comparison before calling
4685 update_stmt_if_modified.
4687 2021-09-01 Andrew Pinski <apinski@marvell.com>
4690 * config/aarch64/aarch64.c (aarch64_expand_setmem):
4691 Check STRICT_ALIGNMENT before creating an overlapping
4694 2021-09-01 Martin Sebor <msebor@redhat.com>
4696 * gimple-ssa-warn-access.cc (get_size_range): Add argument.
4697 (check_access): Pass additional argument.
4698 (check_memop_access): Remove template and make a member function.
4699 (maybe_check_dealloc_call): Make a pass_waccess member function.
4700 (class pass_waccess): Add, rename, and remove members.
4701 (pass_waccess::pass_waccess): Adjust to name change.
4702 (pass_waccess::~pass_waccess): Same.
4703 (check_alloca): Make a member function.
4704 (check_alloc_size_call): Same.
4705 (check_strcat): Same.
4706 (check_strncat): Same.
4707 (check_stxcpy): Same.
4708 (check_stxncpy): Same.
4709 (check_strncmp): Same.
4710 (maybe_warn_rdwr_sizes): Rename...
4711 (pass_waccess::maybe_check_access_sizes): ...to this.
4712 (pass_waccess::check_call): Adjust to name changes.
4713 (pass_waccess::maybe_check_dealloc_call): Make a pass_waccess member
4715 (pass_waccess::execute): Adjust to name changes.
4716 * gimple-ssa-warn-access.h (check_memop_access): Remove.
4717 * pointer-query.cc (access_ref::phi): Handle null pointer.
4718 (access_ref::inform_access): Same.
4719 (pointer_query::put_ref): Modify a cached value, not a copy of it.
4720 (pointer_query::dump): New function.
4721 (compute_objsize_r): Avoid overwriting access_ref::bndrng. Cache
4723 * pointer-query.h (pointer_query::dump): Declare.
4724 * tree-ssa-strlen.c (get_range): Simplify. Use function query.
4725 (dump_strlen_info): Use function query.
4726 (printf_strlen_execute): Factor code out into pointer_query::put_ref.
4728 2021-09-01 Thomas Schwinge <thomas@codesourcery.com>
4730 * tree.c (walk_tree_1) <OMP_CLAUSE>: Simplify.
4732 2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
4734 * doc/extend.texi: Document unavailable attribute.
4735 * print-tree.c (print_node): Handle unavailable attribute.
4736 * tree-core.h (struct tree_base): Add a bit to carry unavailability.
4737 * tree.c (error_unavailable_use): New.
4738 * tree.h (TREE_UNAVAILABLE): New.
4739 (error_unavailable_use): New.
4741 2021-09-01 Jakub Jelinek <jakub@redhat.com>
4743 PR tree-optimization/102124
4744 * tree-vect-patterns.c (vect_recog_widen_op_pattern): For ORIG_CODE
4745 MINUS_EXPR, if itype is unsigned with smaller precision than type,
4746 add an extra cast to signed variant of itype to ensure sign-extension.
4748 2021-09-01 Martin Liska <mliska@suse.cz>
4750 * graph.c (draw_cfg_node_succ_edges): Do not color fallthru
4751 edges and rather use colors for TRUE and FALSE edges.
4753 2021-09-01 Richard Biener <rguenther@suse.de>
4755 PR tree-optimization/93491
4756 * tree-ssa-pre.c (compute_avail): Set BB_MAY_NOTRETURN
4757 after processing the stmt itself. Do not consider
4758 pure functions possibly not returning. Properly avoid
4759 adding possibly trapping calls to EXP_GEN when there's
4760 a preceeding possibly not returning call.
4761 * tree-ssa-sccvn.c (vn_reference_may_trap): Conservatively
4764 2021-09-01 Richard Biener <rguenther@suse.de>
4766 PR tree-optimization/102139
4767 * tree-vectorizer.h (vec_base_alignments): Adjust hash-map
4768 type to record a std::pair of the stmt-info and the innermost
4770 (dr_vec_info::group): New member.
4771 * tree-vect-data-refs.c (vect_record_base_alignment): Adjust.
4772 (vect_compute_data_ref_alignment): Verify the recorded
4773 base alignment can be used.
4774 (data_ref_pair): Remove.
4775 (dr_group_sort_cmp): Adjust.
4776 (vect_analyze_data_ref_accesses): Store the group-ID in the
4777 dr_vec_info and operate on a vector of dr_vec_infos.
4779 2021-09-01 YunQiang Su <yunqiang.su@cipunited.com>
4781 * read-md.c (md_reader::handle_enum): support value assignation.
4782 * doc/md.texi: record define_c_enum value assignation support.
4784 2021-09-01 Jakub Jelinek <jakub@redhat.com>
4786 PR tree-optimization/102141
4787 * gimple-ssa-store-merging.c (bswap_view_convert): Add BEFORE
4788 argument. If false, emit stmts after gsi instead of before, and
4790 (bswap_replace): Adjust callers. When converting output of bswap,
4791 emit VIEW_CONVERT prepratation stmts after a copy of gsi instead
4794 2021-09-01 liuhongt <hongtao.liu@intel.com>
4796 * emit-rtl.c (validate_subreg): Get rid of all float-int
4799 2021-09-01 liuhongt <hongtao.liu@intel.com>
4802 2021-08-30 liuhongt <hongtao.liu@intel.com>
4804 * expmed.c (extract_bit_field_1): Make sure we're playing with
4805 integral modes before call extract_integral_bit_field.
4806 (extract_integral_bit_field): Add a parameter of type
4807 scalar_int_mode which corresponds to of tmode.
4808 And call extract_and_convert_fixed_bit_field instead of
4809 extract_fixed_bit_field and convert_extracted_bit_field.
4810 (extract_and_convert_fixed_bit_field): New function, it's a
4811 combination of extract_fixed_bit_field and
4812 convert_extracted_bit_field.
4814 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
4816 * tree.c (walk_tree_1) <OMP_CLAUSE_TILE>: Handle three operands.
4818 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
4820 * omp-general.h (omp_is_reference): Rename to...
4821 (omp_privatize_by_reference): ... this. Adjust all users...
4822 * omp-general.c: ... here, ...
4823 * gimplify.c: ... here, ...
4824 * omp-expand.c: ... here, ...
4825 * omp-low.c: ... here.
4827 2021-08-31 Martin Sebor <msebor@redhat.com>
4829 * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Test
4830 pointer element for equality to zero, not that of the cotaining
4833 2021-08-31 Martin Sebor <msebor@redhat.com>
4835 * gcc-rich-location.h (gcc_rich_location): Make ctor explicit.
4837 2021-08-31 Martin Sebor <msebor@redhat.com>
4839 * function.h (function): Add comments.
4840 (get_range_query): Same. Add attribute returns nonnull.
4842 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
4844 * expr.c (convert_modes): Don't use subreg_promoted_mode on a
4845 SUBREG if it can't be guaranteed to a SUBREG_PROMOTED_VAR_P set.
4846 Instead use the standard (safer) is_a <scalar_int_mode> idiom.
4848 2021-08-31 Jeff Law <jlaw@localhost.localdomain>
4850 * config.gcc (cris-*-elf, cris-*-none): Remove dbxelf.h from
4852 (m32r-*-elf, m32rle-*-elf, m32r-*-linux): Likewise.
4853 (mn10300-*-*, am33_2.0-*-linux*): Likewise.
4854 (xtensa*-*-elf, xtensa*-*-linux, xtensa*-*-uclinux): Likewise.
4855 (m32c-*-elf*, m32c-*-rtems*): Likewise.
4856 * config/cris/cris.h (DBX_NO_XREFS): Remove.
4857 (DBX_CONTIN_LENGTH, DBX_CONTIN_CHAR): Likewise.
4858 * config/m32r/m32r.h (DBXOUT_SOURCE_LINE): Likewise.
4859 (DBX_DEBUGGING_INFO, DBX_CONTIN_LENGTH): Likewise.
4860 * config/mn10300/mn10300.h (DEFAULT_GDB_EXTENSIONS): Likewise.
4861 * config/mn10300/linux.h (DBX_REGISTER_NAMES): Likewise.
4863 2021-08-31 Marcel Vollweiler <marcel@codesourcery.com>
4865 * gimplify.c (gimplify_scan_omp_clauses): Error handling. 'ancestor' only
4866 allowed on target constructs and only with particular other clauses.
4867 * omp-expand.c (expand_omp_target): Output of 'sorry, not supported' if
4869 * omp-low.c (check_omp_nesting_restrictions): Error handling. No nested OpenMP
4870 structs when 'ancestor' is used.
4871 (scan_omp_1_stmt): No usage of OpenMP runtime routines in a target region when
4873 * tree-pretty-print.c (dump_omp_clause): Append 'ancestor'.
4874 * tree.h (OMP_CLAUSE_DEVICE_ANCESTOR): Define macro.
4876 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
4878 * expr.c (convert_modes): Preserve SUBREG_PROMOTED_VAR_P when
4879 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
4881 * simplify-rtx.c (simplify_unary_operation_1) [SIGN_EXTEND]:
4882 Likewise, preserve SUBREG_PROMOTED_VAR_P when creating a (wider)
4883 partial subreg from a SUBREG_PROMOTED_VAR_P subreg. Generate
4884 SIGN_EXTEND of the SUBREG_REG when a subreg would be paradoxical.
4885 [ZERO_EXTEND]: Likewise, preserve SUBREG_PROMOTED_VAR_P when
4886 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
4887 subreg. Generate ZERO_EXTEND of the SUBREG_REG when a subreg
4888 would be paradoxical.
4890 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
4892 * combine.c (combine_simplify_rtx): Avoid converting an explicit
4893 TRUNCATE into a lowpart SUBREG on !TRULY_NOOP_TRUNCATION targets.
4894 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
4896 2021-08-31 Richard Biener <rguenther@suse.de>
4898 PR tree-optimization/102142
4899 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Fix
4900 condition under which to unset the visited flag.
4902 2021-08-31 Richard Biener <rguenther@suse.de>
4904 PR middle-end/102129
4905 * tree-ssa-ter.c (find_replaceable_in_bb): Do not move
4906 possibly trapping expressions across calls.
4908 2021-08-31 Jakub Jelinek <jakub@redhat.com>
4910 PR tree-optimization/102134
4911 * tree-ssa-ccp.c (bit_value_binop) <case RSHIFT_EXPR>: If sgn is
4912 UNSIGNED and r1val | r1mask has MSB set, ensure lzcount doesn't
4915 2021-08-31 Andrew Pinski <apinski@marvell.com>
4918 * collect-utils.c (setup_signals): New declaration.
4919 * collect-utils.h (setup_signals): New function.
4920 * collect2.c (handler): Delete.
4921 (main): Instead of manually setting up the signals,
4922 just call setup_signals.
4923 * lto-wrapper.c (main): Likewise.
4925 2021-08-31 Andrew Pinski <apinski@marvell.com>
4928 * config/i386/i386-protos.h (x86_output_aligned_bss):
4929 Change align argument to unsigned type.
4930 (x86_elf_aligned_decl_common): Likewise.
4931 * config/i386/i386.c (x86_elf_aligned_decl_common): Likewise.
4932 (x86_output_aligned_bss): Likewise.
4934 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
4936 * config/mips/mips.c (mips_module_isa_name): New.
4937 mips_file_start: add .module mipsREV to all asm output
4939 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
4941 * config/mips/mips.h (struct mips_cpu_info): define enum mips_isa;
4942 use enum instead of int for 'isa' member.
4943 * config.gcc, config/mips/mips.c, config/mips/mips-cpus.def,
4944 config/mips/netbsd.h: replace hardcoded numbers with enum.
4946 2021-08-31 liuhongt <hongtao.liu@intel.com>
4948 * config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
4949 define_split to define_insn_and_split.
4950 (*avx2_eq<mode>3): Removed.
4951 (<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
4952 (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
4953 (*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
4955 (*avx2_gt<mode>3): Removed.
4956 (<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
4957 define_insn to define_expand, and adjust pattern.
4958 (UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.
4960 2021-08-30 David Malcolm <dmalcolm@redhat.com>
4963 * Makefile.in (ANALYZER_OBJS): Add analyzer/call-info.o.
4965 2021-08-30 Jason Merrill <jason@redhat.com>
4967 * doc/invoke.texi: Document -Wmissing-requires.
4969 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
4971 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Remove
4972 TARGET_EXTRA_BUILTINS guard.
4974 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
4976 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Change
4977 initialization of V2DI_type_node and unsigned_V2DI_type_node.
4979 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
4981 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Use the new
4982 decl when new_builtins_are_live.
4983 * config/rs6000/rs6000-builtin-new.def (__builtin_cfstring): New
4986 2021-08-30 Pat Haugen <pthaugen@linux.ibm.com>
4988 * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add
4989 OPTION_MASK_P10_FUSION_2STORE.
4990 (POWERPC_MASKS): Likewise.
4991 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
4992 store fusion for Power10.
4993 (is_fusable_store): New.
4994 (power10_sched_reorder): Likewise.
4995 (rs6000_sched_reorder): Do Power10 specific reordering.
4996 (rs6000_sched_reorder2): Likewise.
4997 * config/rs6000/rs6000.opt: Add new option.
4999 2021-08-30 Richard Biener <rguenther@suse.de>
5001 PR tree-optimization/102128
5002 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
5003 Move scanning for if-converted scalar code to the caller
5004 and instead delay clearing the visited flag for profitable
5006 (vect_slp_region): Cost all subgraphs before scheduling.
5007 For if-converted BB vectorization scan for scalar COND_EXPRs
5008 and do not vectorize if any found and the cost model is
5011 2021-08-30 Richard Biener <rguenther@suse.de>
5013 * common.opt (fexceptions): Mark
5014 EnabledBy(fnon-call-exceptions).
5015 * doc/invoke.texi (fnon-call-exceptions): Document this
5016 enables -fexceptions.
5018 2021-08-30 Sebastian Huber <sebastian.huber@embedded-brains.de>
5020 * tsystem.h (abort): Define abort() if inhibit_libc is defined and it
5021 is not already defined.
5023 2021-08-30 liuhongt <hongtao.liu@intel.com>
5025 * expmed.c (extract_bit_field_1): Make sure we're playing with
5026 integral modes before call extract_integral_bit_field.
5027 (extract_integral_bit_field): Add a parameter of type
5028 scalar_int_mode which corresponds to of tmode.
5029 And call extract_and_convert_fixed_bit_field instead of
5030 extract_fixed_bit_field and convert_extracted_bit_field.
5031 (extract_and_convert_fixed_bit_field): New function, it's a
5032 combination of extract_fixed_bit_field and
5033 convert_extracted_bit_field.
5035 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
5037 * config/darwin.c (darwin_libc_has_function): Do not run
5038 the checks for x86 or modern Darwin. Make sure that there
5039 is a value set for darwin_macosx_version_min before testing.
5041 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
5043 * config/i386/darwin.h (CLEAR_INSN_CACHE): New.
5045 2021-08-28 Jan Hubicka <hubicka@ucw.cz>
5047 * ipa-modref-tree.h (modref_access_node::merge): Break out
5048 logic combining offsets and logic merging ranges to ...
5049 (modref_access_node::combined_offsets): ... here
5050 (modref_access_node::update2): ... here
5051 (modref_access_node::closer_pair_p): New member function.
5052 (modref_access_node::forced_merge): New member function.
5053 (modre_ref_node::insert): Do merging when table is full.
5055 2021-08-28 YunQiang Su <yunqiang.su@cipunited.com>
5058 * config.gcc: MIPS: use N64 ABI by default if the triple end
5059 with -gnuabi64, which is used by Debian since 2013.
5061 2021-08-28 Alexandre Oliva <oliva@adacore.com>
5063 * ipa-modref.c (analyze_function): Skip debug stmts.
5064 * tree-inline.c (estimate_num_insn): Consider builtins even
5065 without a cgraph_node.
5067 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
5069 * config/h8300/bitfield.md (cstore<mode>4): Remove expander.
5070 * config/h8300/h8300.c (h8300_expand_branch): Remove function.
5071 * config/h8300/h8300-protos.h (h8300_expadn_branch): Remove prototype.
5072 * config/h8300/h8300.md (eqne): New code iterator.
5073 (geultu, geultu_to_c): Similarly.
5074 * config/h8300/testcompare.md (cstore<mode>4): Dummy expander.
5075 (store_c_<mode>, store_c_i_<mode>): New define_insn_and_splits
5076 (cmp<mode>_c): New pattern
5078 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
5080 * tree-ssa-dom.c (reduce_vector_comparison_to_scalar_comparison): New
5082 (dom_opt_dom_walker::optimize_stmt): Use it.
5084 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
5086 * config/darwin.c (finalize_ctors): Add a section-start linker-
5088 (finalize_dtors): Likewise.
5089 * config/darwin.h (MIN_LD64_INIT_TERM_START_LABELS): New.
5091 2021-08-27 Bill Schmidt <wschmidt@linux.ibm.com>
5093 * config/rs6000/rs6000-call.c (rs6000-builtins.h): New #include.
5094 (rs6000_init_builtins): Call rs6000_init_generated_builtins. Skip the
5095 old initialization logic when new builtins are enabled.
5096 * config/rs6000/rs6000-gen-builtins.c (write_decls): Rename
5097 rs6000_autoinit_builtins to rs6000_init_generated_builtins.
5098 (write_init_file): Likewise.
5100 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
5102 * configure.ac (darwin2[[0-9]]* | darwin19*): Alter use of
5103 gcc_GAS_CHECK_FEATURE to remove an extraneous parameter.
5104 (amdgcn-* | gcn-*) Likewise.
5106 2021-08-27 Anthony Sharp <anthonysharp15@gmail.com>
5108 * symbol-summary.h: Added missing template keyword.
5110 2021-08-27 Richard Biener <rguenther@suse.de>
5112 PR tree-optimization/45178
5113 * tree-ssa-dce.c (find_obviously_necessary_stmts): For
5114 infinite loops without exit do not mark control dependent
5115 edges of the latch necessary.
5117 2021-08-27 konglin1 <lingling.kong@intel.com>
5120 * config/i386/sse.md: (<avx512>scattersi<mode>): Add mask operand to
5122 (<avx512>scattersi<mode>): Likewise.
5123 (*avx512f_scattersi<VI48F:mode>): Merge mask operand to set_dest.
5124 (*avx512f_scatterdi<VI48F:mode>): Likewise
5126 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
5128 * config/rs6000/rs6000.c (rs6000_builtin_md_vectorized_function): Add
5129 support for built-in functions MISC_BUILTIN_DIVWE, MISC_BUILTIN_DIVWEU,
5130 MISC_BUILTIN_DIVDE, MISC_BUILTIN_DIVDEU, P10_BUILTIN_CFUGED,
5131 P10_BUILTIN_CNTLZDM, P10_BUILTIN_CNTTZDM, P10_BUILTIN_PDEPD and
5132 P10_BUILTIN_PEXTD on Power10.
5134 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
5136 * config/rs6000/rs6000-call.c (builtin_function_type): Add unsigned
5137 signedness for some Power10 bifs.
5139 2021-08-27 David Edelsohn <dje.gcc@gmail.com>
5142 * config/rs6000/rs6000.c (rs6000_adjust_field_align): Use
5143 computed alignment if the entire struct has attribute packed.
5145 2021-08-27 liuhongt <hongtao.liu@intel.com>
5149 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
5150 IX86_BUILTIN_SHUFPD512, IX86_BUILTIN_SHUFPS512,
5151 IX86_BUILTIN_SHUFPD256, IX86_BUILTIN_SHUFPS,
5152 IX86_BUILTIN_SHUFPS256.
5153 (ix86_masked_all_ones): New function.
5155 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
5157 * config/i386/i386.md (*btr<mode>_1): Call force_reg unconditionally.
5158 (conditional moves with memory inputs splitters): Ditto.
5159 * config/i386/sse.md (one_cmpl<mode>2): Simplify.
5161 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
5163 * ipa-modref-tree.h (modref_access_node::try_merge_with): Restart
5164 search after merging.
5166 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
5168 * config/rs6000/rs6000-overload.def: Add remaining overloads.
5170 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
5172 * config/rs6000/rs6000-builtin-new.def: Add cell stanza.
5174 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
5176 * config/rs6000/rs6000-builtin-new.def: Add ieee128-hw, dfp,
5177 crypto, and htm stanzas.
5179 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
5181 * config/rs6000/rs6000-builtin-new.def: Add mma stanza.
5183 2021-08-26 Martin Sebor <msebor@redhat.com>
5185 * tree-ssa-uninit.c (warn_uninit): Refactor and simplify.
5186 (warn_uninit_phi_uses): Remove argument from calls to warn_uninit.
5187 (warn_uninitialized_vars): Same. Reduce visibility of locals.
5188 (warn_uninitialized_phi): Same.
5190 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
5192 * tree-ssa-ccp.c (get_individual_bits): Helper function to
5193 extract the individual bits from a widest_int constant (mask).
5194 (gray_code_bit_flips): New read-only table for effiently
5195 enumerating permutations/combinations of bits.
5196 (bit_value_binop) [LROTATE_EXPR, RROTATE_EXPR]: Handle rotates
5197 by unknown counts that are guaranteed less than the target
5198 precision and four or fewer unknown bits by enumeration.
5199 [LSHIFT_EXPR, RSHIFT_EXPR]: Likewise, also handle shifts by
5200 enumeration under the same conditions. Handle remaining
5201 shifts as a mask based upon the minimum possible shift value.
5203 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
5204 Richard Biener <rguenther@suse.de>
5206 * match.pd (shift transformations): Remove a redundant
5207 !POINTER_TYPE_P check.
5209 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
5212 * config/i386/i386.md (cmove reg-reg move elimination peephole2s):
5213 Set all_regs to true in the call to replace_rtx.
5215 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
5217 * ipa-modref-tree.c (test_insert_search_collapse): Update test.
5218 * ipa-modref-tree.h (modref_base_node::insert): Be smarter when
5219 hiting --param modref-max-refs limit.
5220 (modref_tree:insert_base): Be smarter when hitting
5221 --param modref-max-bases limit. Add new parameter REF.
5222 (modref_tree:insert): Update.
5223 (modref_tree:merge): Update.
5224 * ipa-modref.c (read_modref_records): Update.
5226 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
5228 * params.opt: (modref-max-adjustments): Add full stop.
5230 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
5232 * ipa-modref-tree.h (modref_ref_node::verify): New member
5234 (modref_ref_node::insert): Use it.
5235 (modref_ref_node::try_mere_with): Fix off by one error.
5237 2021-08-26 Martin Liska <mliska@suse.cz>
5238 Stefan Kneifel <stefan.kneifel@bluewin.ch>
5240 * cgraph.h (create_version_clone_with_body): Add new parameter.
5241 * cgraphclones.c: Likewise.
5242 * multiple_target.c (create_dispatcher_calls): Do not use
5244 (create_target_clone): Likewise here.
5246 2021-08-26 Jonathan Yong <10walls@gmail.com>
5248 * doc/extend.texi: Add note about reserved priorities
5249 to the constructor attribute.
5251 2021-08-25 Martin Sebor <msebor@redhat.com>
5253 * gimple-range-cache.cc (ssa_global_cache::dump): Avoid printing
5254 range table header alone.
5255 * gimple-range.cc (gimple_ranger::export_global_ranges): Same.
5257 2021-08-25 Jan Hubicka <hubicka@ucw.cz>
5259 * doc/invoke.texi: Document --param modref-max-adjustments.
5260 * ipa-modref-tree.c (test_insert_search_collapse): Update.
5261 (test_merge): Update.
5262 * ipa-modref-tree.h (struct modref_access_node): Add adjustments;
5263 (modref_access_node::operator==): Fix handling of access ranges.
5264 (modref_access_node::contains): Constify parameter; handle also
5265 mismatched parm offsets.
5266 (modref_access_node::update): New function.
5267 (modref_access_node::merge): New function.
5268 (unspecified_modref_access_node): Update constructor.
5269 (modref_ref_node::insert_access): Add record_adjustments parameter;
5271 (modref_ref_node::try_merge_with): New private function.
5272 (modref_tree::insert): New record_adjustments parameter.
5273 (modref_tree::merge): New record_adjustments parameter.
5274 (modref_tree::copy_from): Update.
5275 * ipa-modref.c (dump_access): Dump adjustments field.
5276 (get_access): Update constructor.
5277 (record_access): Update call of insert.
5278 (record_access_lto): Update call of insert.
5279 (merge_call_side_effects): Add record_adjustments parameter.
5280 (get_access_for_fnspec): Update.
5281 (process_fnspec): Update.
5282 (analyze_call): Update.
5283 (analyze_function): Update.
5284 (read_modref_records): Update.
5285 (ipa_merge_modref_summary_after_inlining): Update.
5286 (propagate_unknown_call): Update.
5287 (modref_propagate_in_scc): Update.
5288 * params.opt (param-max-modref-adjustments=): New.
5290 2021-08-25 Michael Meissner <meissner@linux.ibm.com>
5292 * config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
5294 (xxspltiw_v4si): Use vecperm type attribute.
5295 (xxspltiw_v4si_inst): Use vecperm type attribute.
5296 (xxspltiw_v4sf_inst): Likewise.
5297 (xxspltidp_v2df): Use vecperm type attribute. Use
5298 UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
5299 (xxspltidp_v2df_inst): Likewise.
5300 (xxsplti32dx_v4si): Use vecperm type attribute.
5301 (xxsplti32dx_v4si_inst): Likewise.
5302 (xxsplti32dx_v4sf_inst): Likewise.
5303 (xxblend_<mode>): Likewise.
5304 (xxpermx): Likewise.
5305 (xxpermx_inst): Likewise.
5308 2021-08-25 Lewis Hyatt <lhyatt@gmail.com>
5311 * coretypes.h (typedef diagnostic_input_charset_callback): Declare.
5312 * diagnostic.c (diagnostic_initialize_input_context): New function.
5313 * diagnostic.h (diagnostic_initialize_input_context): Declare.
5314 * input.c (default_charset_callback): New function.
5315 (file_cache::initialize_input_context): New function.
5316 (file_cache_slot::create): Added ability to convert the input
5317 according to the input context.
5318 (file_cache::file_cache): Initialize the new input context.
5319 (class file_cache_slot): Added new m_alloc_offset member.
5320 (file_cache_slot::file_cache_slot): Initialize the new member.
5321 (file_cache_slot::~file_cache_slot): Handle potentially offset buffer.
5322 (file_cache_slot::maybe_grow): Likewise.
5323 (file_cache_slot::needs_read_p): Handle NULL fp, which is now possible.
5324 (file_cache_slot::get_next_line): Likewise.
5325 * input.h (class file_cache): Added input context member.
5327 2021-08-25 Richard Biener <rguenther@suse.de>
5329 PR tree-optimization/102046
5330 * tree-vect-slp.c (vect_build_slp_tree_2): Conservatively
5331 update ->any_pattern when swapping operands.
5333 2021-08-25 Hongyu Wang <hongyu.wang@intel.com>
5336 * config/i386/i386.c (ix86_live_on_entry): Adjust comment.
5337 (ix86_decompose_address): Remove retval check for ASHIFT,
5338 allow non-canonical zero extend if AND mask covers ASHIFT
5340 (ix86_legitimate_address_p): Adjust condition for decompose.
5341 (ix86_rtx_costs): Adjust cost for lea with non-canonical
5343 Co-Authored by: Uros Bizjak <ubizjak@gmail.com>
5345 2021-08-25 Jiufu Guo <guojiufu@linux.ibm.com>
5347 PR tree-optimization/101145
5348 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
5350 (number_of_iterations_lt): Invoke above function.
5351 (adjust_cond_for_loop_until_wrap):
5352 Merge to number_of_iterations_until_wrap.
5353 (number_of_iterations_cond): Update invokes for
5354 adjust_cond_for_loop_until_wrap and number_of_iterations_lt.
5356 2021-08-25 konglin1 <lingling.kong@intel.com>
5359 * config/i386/avx512dqintrin.h (_mm512_fpclass_ps_mask): Fix
5361 (_mm512_mask_fpclass_ps_mask): Ditto.
5363 2021-08-25 Kewen Lin <linkw@linux.ibm.com>
5365 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Remove.
5366 (vec_unpacku_hi_v8hi): Likewise.
5367 (vec_unpacku_lo_v16qi): Likewise.
5368 (vec_unpacku_lo_v8hi): Likewise.
5369 (vec_unpacku_hi_<VP_small_lc>): New define_expand.
5370 (vec_unpacku_lo_<VP_small_lc>): Likewise.
5372 2021-08-24 David Edelsohn <dje.gcc@gmail.com>
5374 * config/rs6000/aix.h (SYSTEM_IMPLICIT_EXTERN_C): Delete.
5375 * config/rs6000/aix71.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
5376 * config/rs6000/aix72.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
5377 * config/rs6000/aix73.h (TARGET_AIX_VERSION): Increase to 73.
5379 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5381 PR middle-end/102031
5382 * simplify-rtx.c (simplify_truncation): When comparing precisions
5383 use "subreg_prec" variable, not "subreg_mode".
5385 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5387 * config/rs6000/rs6000-builtin-new.def: Add power10 and power10-64
5390 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5392 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Initialize
5393 various pointer type nodes.
5394 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add enum
5395 values for various pointer types.
5396 (ptr_V16QI_type_node): New macro.
5397 (ptr_V1TI_type_node): New macro.
5398 (ptr_V2DI_type_node): New macro.
5399 (ptr_V2DF_type_node): New macro.
5400 (ptr_V4SI_type_node): New macro.
5401 (ptr_V4SF_type_node): New macro.
5402 (ptr_V8HI_type_node): New macro.
5403 (ptr_unsigned_V16QI_type_node): New macro.
5404 (ptr_unsigned_V1TI_type_node): New macro.
5405 (ptr_unsigned_V8HI_type_node): New macro.
5406 (ptr_unsigned_V4SI_type_node): New macro.
5407 (ptr_unsigned_V2DI_type_node): New macro.
5408 (ptr_bool_V16QI_type_node): New macro.
5409 (ptr_bool_V8HI_type_node): New macro.
5410 (ptr_bool_V4SI_type_node): New macro.
5411 (ptr_bool_V2DI_type_node): New macro.
5412 (ptr_bool_V1TI_type_node): New macro.
5413 (ptr_pixel_type_node): New macro.
5414 (ptr_intQI_type_node): New macro.
5415 (ptr_uintQI_type_node): New macro.
5416 (ptr_intHI_type_node): New macro.
5417 (ptr_uintHI_type_node): New macro.
5418 (ptr_intSI_type_node): New macro.
5419 (ptr_uintSI_type_node): New macro.
5420 (ptr_intDI_type_node): New macro.
5421 (ptr_uintDI_type_node): New macro.
5422 (ptr_intTI_type_node): New macro.
5423 (ptr_uintTI_type_node): New macro.
5424 (ptr_long_integer_type_node): New macro.
5425 (ptr_long_unsigned_type_node): New macro.
5426 (ptr_float_type_node): New macro.
5427 (ptr_double_type_node): New macro.
5428 (ptr_long_double_type_node): New macro.
5429 (ptr_dfloat64_type_node): New macro.
5430 (ptr_dfloat128_type_node): New macro.
5431 (ptr_ieee128_type_node): New macro.
5432 (ptr_ibm128_type_node): New macro.
5433 (ptr_vector_pair_type_node): New macro.
5434 (ptr_vector_quad_type_node): New macro.
5435 (ptr_long_long_integer_type_node): New macro.
5436 (ptr_long_long_unsigned_type_node): New macro.
5438 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5440 * config/rs6000/rs6000-builtin-new.def: Add power9-vector, power9,
5441 and power9-64 stanzas.
5443 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5444 Tom de Vries <tdevries@suse.de>
5446 * config.gcc (nvptx-*-*): Define {c,c++}_target_objs.
5447 * config/nvptx/nvptx-protos.h (nvptx_cpu_cpp_builtins): Prototype.
5448 * config/nvptx/nvptx.h (TARGET_CPU_CPP_BUILTINS): Implement with
5449 a call to the new nvptx_cpu_cpp_builtins function in nvptx-c.c.
5450 * config/nvptx/t-nvptx (nvptx-c.o): New rule.
5451 * config/nvptx/nvptx-c.c: New source file.
5452 (nvptx_cpu_cpp_builtins): Move implementation here.
5454 2021-08-24 Martin Sebor <msebor@redhat.com>
5456 PR middle-end/101600
5457 PR middle-end/101977
5458 * gimple-ssa-warn-access.cc (maybe_warn_for_bound): Tighten up
5459 the phrasing of a warning.
5460 (check_access): Use the remaining size after subtracting any offset
5461 rather than the whole object size.
5462 * pointer-query.cc (access_ref::get_ref): Clear BASE0 flag if it's
5463 clear for any nonnull PHI argument.
5464 (compute_objsize): Clear argument.
5466 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5468 * config/rs6000/rs6000-builtin-new.def: Add power8-vector stanza.
5470 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5472 * config/rs6000/rs6000-builtin-new.def: Add power7 and power7-64
5475 2021-08-24 Andrew MacLeod <amacleod@redhat.com>
5477 * value-relation.cc (rr_transitive_table): New.
5478 (relation_transitive): New.
5479 (value_relation::swap): Remove.
5480 (value_relation::apply_transitive): New.
5481 (relation_oracle::relation_oracle): Allocate a new tmp bitmap.
5482 (relation_oracle::register_relation): Call register_transitives.
5483 (relation_oracle::register_transitives): New.
5484 * value-relation.h (relation_oracle): Add new temporary bitmap and
5487 2021-08-24 H.J. Lu <hjl.tools@gmail.com>
5490 * config/i386/i386-expand.c (ix86_expand_vector_move): Broadcast
5491 from integer to a pseudo vector register.
5493 2021-08-24 Richard Biener <rguenther@suse.de>
5495 PR tree-optimization/100089
5496 * tree-vectorizer.h (vect_slp_bb): Rename to ...
5497 (vect_slp_if_converted_bb): ... this and get the original
5498 loop as new argument.
5499 * tree-vectorizer.c (try_vectorize_loop_1): Revert previous fix,
5500 pass original loop to vect_slp_if_converted_bb.
5501 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
5502 If orig_loop was passed scan the not vectorized stmts
5503 for COND_EXPRs and force not profitable if found.
5504 (vect_slp_region): Pass down all SLP instances to costing
5505 if orig_loop was specified.
5506 (vect_slp_bbs): Pass through orig_loop.
5507 (vect_slp_bb): Rename to ...
5508 (vect_slp_if_converted_bb): ... this and get the original
5509 loop as new argument.
5510 (vect_slp_function): Adjust.
5512 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
5515 * config/arm/arm.md (attribute arch): Add fix_vlldm.
5516 (arch_enabled): Use it.
5517 * config/arm/vfp.md (lazy_store_multiple_insn): Add alternative to
5518 use when erratum mitigation is needed.
5520 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
5523 * config/arm/arm.opt (mfix-cmse-cve-2021-35465): New option.
5524 * doc/invoke.texi (Arm Options): Document it.
5525 * config/arm/arm-cpus.in (quirk_vlldm): New feature bit.
5526 (ALL_QUIRKS): Add quirk_vlldm.
5527 (cortex-m33): Add quirk_vlldm.
5528 (cortex-m35p, cortex-m55): Likewise.
5529 * config/arm/arm.c (arm_option_override): Enable fix_vlldm if
5530 targetting an affected CPU and not explicitly controlled on
5533 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
5535 * config/arm/vfp.md (lazy_store_multiple_insn): Rewrite as valid RTL.
5536 (lazy_load_multiple_insn): Likewise.
5538 2021-08-24 liuhongt <hongtao.liu@intel.com>
5541 * config/i386/sse.md (<avx512>_vternlog<mode><sd_maskz_name>):
5542 Enable avx512 embedded broadcast.
5543 (*<avx512>_vternlog<mode>_all): Ditto.
5544 (<avx512>_vternlog<mode>_mask): Ditto.
5546 2021-08-24 liuhongt <hongtao.liu@intel.com>
5549 * config/i386/i386.c (ix86_rtx_costs): Define cost for
5551 * config/i386/i386.h (STRIP_UNARY): New macro.
5552 * config/i386/predicates.md (reg_or_notreg_operand): New
5554 * config/i386/sse.md (*<avx512>_vternlog<mode>_all): New define_insn.
5555 (*<avx512>_vternlog<mode>_1): New pre_reload
5556 define_insn_and_split.
5557 (*<avx512>_vternlog<mode>_2): Ditto.
5558 (*<avx512>_vternlog<mode>_3): Ditto.
5559 (any_logic1,any_logic2): New code iterator.
5560 (logic_op): New code attribute.
5561 (ternlogsuffix): Extend to VNxDF and VNxSF.
5563 2021-08-24 Richard Biener <rguenther@suse.de>
5565 * doc/invoke.texi (vect-inner-loop-cost-factor): Adjust.
5566 * params.opt (--param vect-inner-loop-cost-factor): Adjust
5568 * tree-vect-loop.c (vect_analyze_loop_form): Initialize
5569 inner_loop_cost_factor to the minimum of the estimated number
5570 of iterations of the inner loop and vect-inner-loop-cost-factor.
5572 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5573 Richard Biener <rguenther@suse.de>
5575 * config/i386/i386-features.c (compute_convert_gain): Provide
5576 more accurate values for CONST_INT, when optimizing for size.
5577 * config/i386/i386.c (COSTS_N_BYTES): Move definition from here...
5578 * config/i386/i386.h (COSTS_N_BYTES): to here.
5580 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5581 Jakub Jelinek <jakub@redhat.com>
5583 PR middle-end/102029
5584 * match.pd (shift transformations): Add an additional check for
5585 !POINTER_TYPE_P in the recently added left shift transformation.
5587 2021-08-24 liuhongt <hongtao.liu@intel.com>
5589 PR tree-optimization/100089
5590 * tree-vectorizer.c (try_vectorize_loop_1): Disable slp in
5591 loop vectorizer when cost model is very-cheap.
5593 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5595 * config/rs6000/rs6000-gen-builtins.c (parse_bif_entry): Don't call
5596 asprintf, which is not available on AIX.
5598 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5600 * config.gcc (target_gtfiles): Add ./rs6000-builtins.h.
5601 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Set.
5603 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5605 * config.gcc (powerpc*-*-*): Add rs6000-builtins.o to extra_objs.
5606 * config/rs6000/rs6000-gen-builtins.c (main): Close init_file
5608 * config/rs6000/t-rs6000 (rs6000-gen-builtins.o): New target.
5609 (rbtree.o): Likewise.
5610 (rs6000-gen-builtins): Likewise.
5611 (rs6000-builtins.c): Likewise.
5612 (rs6000-builtins.h): Likewise.
5613 (rs6000.o): Add dependency.
5614 (EXTRA_HEADERS): Add rs6000-vecdefines.h.
5615 (rs6000-vecdefines.h): New target.
5616 (rs6000-builtins.o): Likewise.
5617 (rs6000-call.o): Add rs6000-builtins.h as a dependency.
5618 (rs6000-c.o): Likewise.
5620 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5623 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace):
5624 Diagnose buffer overrun.
5625 (safe_inc_pos): Fix overrun detection.
5626 (match_identifier): Diagnose buffer overrun.
5627 (match_integer): Likewise.
5628 (match_to_right_bracket): Likewise.
5630 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
5632 * ipa-modref-tree.h (modref_access_node::range_info_useful_p):
5633 Improve range compare.
5634 (modref_access_node::contains): New member function.
5635 (modref_access_node::search): Remove.
5636 (modref_access_node::insert): Be smarter about subaccesses.
5638 2021-08-23 Thomas Schwinge <thomas@codesourcery.com>
5640 * config/i386/i386-options.c (ix86_omp_device_kind_arch_isa)
5641 <omp_device_arch> [ACCEL_COMPILER]: Match "intel_mic".
5642 * config/i386/t-omp-device (omp-device-properties-i386) <arch>:
5645 2021-08-23 Jeff Law <jlaw@localhost.localdomain>
5647 * config/h8300/h8300-protos.h (h8300_expand_epilogue): Add new
5649 * config/h8300/jumpcall.md (call, call_value): Restrict to
5650 !SIBLING_CALL_P cases.
5651 (subcall, sibcall_value): New patterns & expanders.
5652 * config/h8300/proepi.md (epilogue): Pass new argument to
5653 h8300_expand_epilogue.
5654 (sibcall_epilogue): New expander.
5655 * config/h8300/h8300.c (h8300_expand_epilogue): Handle sibcall
5657 (h8300_ok_for_sibcall_p): New function.
5658 (TARGET_FUNCTION_OK_FOR_SIBCALL): define.
5660 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5662 * simplify-rtx.c (simplify_unary_operation_1): [TRUNCATE]:
5663 Handle case where the operand is already the desired mode.
5665 2021-08-23 Richard Biener <rguenther@suse.de>
5668 * tree-ssa-structalias.c (ipa_pta_execute): Check in_other_partition
5669 in addition to has_gimple_body.
5671 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
5673 PR middle-end/101949
5674 * ipa-modref.c (analyze_ssa_name_flags): Fix merging of
5677 2021-08-23 Martin Liska <mliska@suse.cz>
5679 * doc/invoke.texi: Put the option out of -mxl-mode-app-model
5682 2021-08-23 Richard Biener <rguenther@suse.de>
5684 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5685 Properly scale the inner loop cost only once.
5687 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5689 * tree-ssa-ccp.c (bit_value_binop) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]:
5690 Provide bounds for unsigned (and signed with non-negative operands)
5691 division and modulus.
5693 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5695 * simplify-rtx.c (simplify_truncation): Generalize simplification
5696 of (truncate:A (subreg:B X)).
5697 (simplify_unary_operation_1) [FLOAT_TRUNCATE, FLOAT_EXTEND,
5698 SIGN_EXTEND, ZERO_EXTEND]: Handle cases where the operand
5699 already has the desired machine mode.
5700 (test_scalar_int_ops): Add tests that useless extensions and
5701 truncations are optimized away.
5702 (test_scalar_int_ext_ops): New self-test function to confirm
5703 that truncations of extensions are correctly simplified.
5704 (test_scalar_int_ext_ops2): New self-test function to check
5705 truncations of truncations, extensions of extensions, and
5706 truncations of extensions.
5707 (test_scalar_ops): Call the above two functions with a
5708 representative sampling of integer machine modes.
5710 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5712 * match.pd (shift transformations): Change the sign of an
5713 LSHIFT_EXPR if it reduces the number of explicit conversions.
5715 2021-08-23 Jakub Jelinek <jakub@redhat.com>
5717 PR tree-optimization/86723
5718 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
5719 cast64_to_32 argument, set *cast64_to_32 to false, unless n is
5720 non-memory permutation of 64-bit src which only has bytes of
5721 0 or [5..8] and n->range is 4.
5722 (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
5723 find_bswap_or_nop_finalize caller, support bswap with some bytes
5724 zeroed, as long as at least two bytes are not zeroed.
5725 (bswap_replace): Add mask argument and handle masking of bswap
5727 (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
5728 caller, punt if cast64_to_32 or mask is not all ones.
5729 (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
5730 caller, for now punt if cast64_to_32.
5732 2021-08-23 Richard Biener <rguenther@suse.de>
5734 PR tree-optimization/79334
5735 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Record
5736 a type also for COMPONENT_REFs.
5737 (vn_reference_may_trap): Check ARRAY_REF with constant index
5738 against the array domain.
5740 2021-08-23 liuhongt <hongtao.liu@intel.com>
5743 * config/i386/sse.md (*avx512f_pshufb_truncv8hiv8qi_1): Add
5744 TARGET_AVX512BW to condition.
5746 2021-08-23 Jakub Jelinek <jakub@redhat.com>
5749 * dwarf2out.c (gen_variable_die): Add DW_AT_location for global
5750 register variables already during early_dwarf if possible.
5752 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
5754 * config/arm/arm_mve.h: Fix __arm_vctp16q return type.
5756 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
5759 * config/arm/arm.opt: Fix typo.
5760 * config/arm/t-rmprofile: Fix typo.
5762 2021-08-23 Jakub Jelinek <jakub@redhat.com>
5764 * tree.h (OMP_CLAUSE_GRAINSIZE_STRICT): Define.
5765 (OMP_CLAUSE_NUM_TASKS_STRICT): Define.
5766 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_GRAINSIZE,
5767 case OMP_CLAUSE_NUM_TASKS>: Print strict: modifier.
5768 * omp-expand.c (expand_task_call): Use GOMP_TASK_FLAG_STRICT in iflags
5769 if either grainsize or num_tasks clause has the strict modifier.
5771 2021-08-23 Martin Liska <mliska@suse.cz>
5773 * dbgcnt.def (DEBUG_COUNTER): New counter.
5774 * gimple.c (gimple_call_arg_flags): Use it in IPA PTA.
5776 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
5778 * ipa-modref.c (analyze_ssa_name_flags): Improve handling of return slot.
5780 2021-08-23 Xi Ruoyao <xry111@mengyan1223.wang>
5783 * config/mips/mips-protos.h (mips_msa_output_shift_immediate):
5785 * config/mips/mips.c (mips_msa_output_shift_immediate): New
5787 * config/mips/mips-msa.md (vashl<mode>3, vashr<mode>3,
5788 vlshr<mode>3): Call it.
5790 2021-08-22 Jan Hubicka <hubicka@ucw.cz>
5791 Martin Liska <mliska@suse.cz>
5793 PR middle-end/101949
5794 * ipa-modref.c (analyze_ssa_name_flags): Indirect call implies
5797 2021-08-21 Dragan Mladjenovic <OT_Dragan.Mladjenovic@mediatek.com>
5799 * config/mips/mips.c (mips_function_rodata_section,
5800 TARGET_ASM_FUNCTION_RODATA_SECTION): Removed.
5802 2021-08-21 John David Anglin <danglin@gcc.gnu.org>
5804 * config/pa/pa.c (pa_asm_output_aligned_common): Remove warning.
5806 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5808 * configure.ac (thread-local storage support): Remove tls_first_major
5809 and tls_first_minor. Use "$conftest_s" to check support.
5810 * configure: Regenerate.
5812 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5814 * configure.ac: Fixup formatting.
5816 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5818 * acinclude.m4 (gcc_GAS_CHECK_FEATURE): Remove third argument and ...
5819 * configure.ac: ... update all callers.
5821 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5824 * acinclude.m4 (_gcc_COMPUTE_GAS_VERSION, _gcc_GAS_VERSION_GTE_IFELSE)
5825 (gcc_GAS_VERSION_GTE_IFELSE): Remove.
5826 (gcc_GAS_CHECK_FEATURE): Do not handle in-tree case specially.
5827 * configure.ac: Remove gcc_cv_gas_major_version, gcc_cv_gas_minor_version.
5828 Remove remaining checks for in-tree assembler.
5829 * configure: Regenerate.
5831 2021-08-20 Jeff Law <jlaw@localhost.localdomain>
5833 * config/h8300/h8300.c (shift_alg_hi): Improve arithmetic shift right
5834 by 15 bits for H8/300H and H8/S. Improve logical shifts by 12
5836 (shift_alg_si): Improve arithmetic right shift by 28-30 bits for
5837 H8/300H. Improve arithmetic shift right by 15 bits for H8/S.
5838 Improve logical shifts by 27 bits for H8/S.
5839 (get_shift_alg): Corresponding changes.
5840 (h8300_option_override): Revert to loops for -Os when profitable.
5842 2021-08-20 Richard Biener <rguenther@suse.de>
5844 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
5846 (vect_analyze_data_ref_accesses): Likewise. Assign the BB
5847 index as group_id when dataref_groups were not computed.
5848 * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
5849 we advace to the next BB.
5851 2021-08-20 Jakub Jelinek <jakub@redhat.com>
5853 * omp-builtins.def (BUILT_IN_GOMP_WARNING, BUILT_IN_GOMP_ERROR): New
5856 2021-08-20 Martin Liska <mliska@suse.cz>
5858 PR gcov-profile/89961
5859 * gcov.c (make_gcov_file_name): Rewrite using std::string.
5860 (mangle_name): Simplify, do not used the second argument.
5861 (strip_extention): New function.
5862 (get_md5sum): Likewise.
5863 (get_gcov_intermediate_filename): Handle properly -p and -x
5865 (output_gcov_file): Use string type.
5866 (generate_results): Likewise.
5867 (md5sum_to_hex): Remove.
5869 2021-08-20 Michael Meissner <meissner@linux.ibm.com>
5871 * config/rs6000/altivec.md (UNSPEC_XXEVAL): Move to vsx.md.
5872 (UNSPEC_XXSPLTIW): Move to vsx.md.
5873 (UNSPEC_XXSPLTID): Move to vsx.md.
5874 (UNSPEC_XXSPLTI32DX): Move to vsx.md.
5875 (UNSPEC_XXBLEND): Move to vsx.md.
5876 (UNSPEC_XXPERMX): Move to vsx.md.
5877 (VM3): Move to vsx.md.
5878 (VM3_char): Move to vsx.md.
5879 (xxspltiw_v4si): Move to vsx.md.
5880 (xxspltiw_v4sf): Move to vsx.md.
5881 (xxspltiw_v4sf_inst): Move to vsx.md.
5882 (xxspltidp_v2df): Move to vsx.md.
5883 (xxspltidp_v2df_inst): Move to vsx.md.
5884 (xxsplti32dx_v4si_inst): Move to vsx.md.
5885 (xxsplti32dx_v4sf): Move to vsx.md.
5886 (xxsplti32dx_v4sf_inst): Move to vsx.md.
5887 (xxblend_<mode>): Move to vsx.md.
5888 (xxpermx): Move to vsx.md.
5889 (xxpermx_inst): Move to vsx.md.
5890 * config/rs6000/vsx.md (UNSPEC_XXEVAL): Move from altivec.md.
5891 (UNSPEC_XXSPLTIW): Move from altivec.md.
5892 (UNSPEC_XXSPLTID): Move from altivec.md.
5893 (UNSPEC_XXSPLTI32DX): Move from altivec.md.
5894 (UNSPEC_XXBLEND): Move from altivec.md.
5895 (UNSPEC_XXPERMX): Move from altivec.md.
5896 (VM3): Move from altivec.md.
5897 (VM3_char): Move from altivec.md.
5898 (xxspltiw_v4si): Move from altivec.md.
5899 (xxspltiw_v4sf): Move from altivec.md.
5900 (xxspltiw_v4sf_inst): Move from altivec.md.
5901 (xxspltidp_v2df): Move from altivec.md.
5902 (xxspltidp_v2df_inst): Move from altivec.md.
5903 (xxsplti32dx_v4si_inst): Move from altivec.md.
5904 (xxsplti32dx_v4sf): Move from altivec.md.
5905 (xxsplti32dx_v4sf_inst): Move from altivec.md.
5906 (xxblend_<mode>): Move from altivec.md.
5907 (xxpermx): Move from altivec.md.
5908 (xxpermx_inst): Move from altivec.md.
5910 2021-08-19 Roger Sayle <roger@nextmovesoftware.com>
5912 * tree-vect-generic.c (expand_vector_operations_1): Use either
5913 gimplify_build1 or gimplify_build2 instead of gimple_build_assign
5914 when constructing scalar splat expressions.
5916 2021-08-19 Peter Bergner <bergner@linux.ibm.com>
5919 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Cast
5920 pointer to __vector_pair *.
5922 2021-08-19 Martin Sebor <msebor@redhat.com>
5924 * gimple-range.cc: Add comments.
5925 * gimple-range.h: Same.
5927 2021-08-19 Martin Sebor <msebor@redhat.com>
5929 PR middle-end/101984
5930 * gimple-ssa-warn-access.cc (pass_waccess::execute): Also call
5933 2021-08-19 Jeff Law <jlaw@localhost.localdomain>
5935 * config.gcc (h8300-*-elf*): Do not include dbxelf.h.
5936 (h8300-*-linux*, v850-*-rtems*, v850*-elf*): Likewise.
5937 * config/v850/v850.h (DEFAULT_GDB_EXTENSIONS): Remove.
5939 2021-08-19 Jakub Jelinek <jakub@redhat.com>
5941 PR middle-end/101950
5942 * optabs.c (expand_clrsb_using_clz): New function.
5943 (expand_unop): Use it as another clrsb expansion fallback.
5945 2021-08-19 liuhongt <hongtao.liu@intel.com>
5948 2021-07-28 liuhongt <hongtao.liu@intel.com>
5951 * config/i386/i386.h (processor_costs): Add new member
5953 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
5954 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
5955 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
5956 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
5957 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
5958 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
5959 generic_cost, core_cost): Initialize integer_to_sse same value
5961 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
5962 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5963 Use integer_to_sse instead of sse_op to calculate the cost of
5966 2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
5968 * config.gcc: Include rpath.opt for Darwin.
5969 * config/darwin.h (DRIVER_SELF_SPECS): Handle -rpath.
5971 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
5974 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
5977 2021-08-18 Jonathan Wright <jonathan.wright@arm.com>
5979 * config/aarch64/arm_neon.h (vld3_lane_f64): Use float RTL
5980 pattern and type cast.
5981 (vld4_lane_f32): Use float RTL pattern.
5982 (vld4q_lane_f64): Use float type cast.
5984 2021-08-18 Jan Hubicka <hubicka@ucw.cz>
5986 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Check also
5989 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
5991 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor): Extend.
5992 (test_map_of_type_with_ctor_and_dtor_expand): Add function.
5993 (hash_map_tests_c_tests): Call it.
5995 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
5997 * ggc.h (enum ggc_collect): New.
5998 (ggc_collect): Use it.
5999 * ggc-page.c: Adjust.
6000 * ggc-common.c: Likewise.
6001 * ggc-tests.c: Likewise.
6002 * read-rtl-function.c: Likewise.
6003 * selftest-run-tests.c: Likewise.
6004 * doc/gty.texi (Invoking the garbage collector): Likewise.
6006 2021-08-18 liuhongt <hongtao.liu@intel.com>
6009 * config/i386/i386.h (TARGET_V2DF_REDUCTION_PREFER_HADDPD):
6011 * config/i386/sse.md (*sse3_haddv2df3_low): Add
6012 TARGET_V2DF_REDUCTION_PREFER_HADDPD.
6013 (*sse3_hsubv2df3_low): Ditto.
6014 * config/i386/x86-tune.def
6015 (X86_TUNE_V2DF_REDUCTION_PREFER_HADDPD): New tune.
6017 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
6019 * gimple-range-gori.cc (gori_compute::gori_compute): Enable tracing.
6020 (gori_compute::compute_operand_range): Add tracing.
6021 (gori_compute::logical_combine): Ditto.
6022 (gori_compute::compute_logical_operands): Ditto.
6023 (gori_compute::compute_operand1_range): Ditto.
6024 (gori_compute::compute_operand2_range): Ditto.
6025 (gori_compute::outgoing_edge_range_p): Ditto.
6026 * gimple-range-gori.h (class gori_compute): Add range_tracer.
6028 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
6030 * flag-types.h (enum evrp_mode): Adjust evrp-mode values.
6031 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Relocate from.
6032 * gimple-range-trace.h (DEBUG_RANGE_CACHE): Here.
6033 * params.opt (--param=evrp-mode): Adjust options.
6035 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
6037 * Makefile.in (OBJS): Add gimple-range-trace.o.
6038 * gimple-range-cache.h (enable_new_values): Remove unused prototype.
6039 * gimple-range-fold.cc: Adjust headers.
6040 * gimple-range-trace.cc: New.
6041 * gimple-range-trace.h: New.
6042 * gimple-range.cc (gimple_ranger::gimple_ranger): Enable tracer.
6043 (gimple_ranger::range_of_expr): Add tracing.
6044 (gimple_ranger::range_on_entry): Ditto.
6045 (gimple_ranger::range_on_exit): Ditto.
6046 (gimple_ranger::range_on_edge): Ditto.
6047 (gimple_ranger::fold_range_internal): Ditto.
6048 (gimple_ranger::dump_bb): Do not calculate edge range twice.
6049 (trace_ranger::*): Remove.
6050 (enable_ranger): Never create a trace_ranger.
6051 (debug_seed_ranger): Move to gimple-range-trace.cc.
6052 (dump_ranger): Ditto.
6053 (debug_ranger): Ditto.
6054 * gimple-range.h: Include gimple-range-trace.h.
6055 (range_on_entry, range_on_exit): No longer virtual.
6056 (class trace_ranger): Remove.
6057 (DEBUG_RANGE_CACHE): Move to gimple-range-trace.h.
6059 2021-08-17 Martin Sebor <msebor@redhat.com>
6061 PR middle-end/101854
6062 * builtins.c (expand_builtin_alloca): Move warning code to check_alloca
6063 in gimple-ssa-warn-access.cc.
6064 * calls.c (alloc_max_size): Move code to check_alloca.
6065 (get_size_range): Move to pointer-query.cc.
6066 (maybe_warn_alloc_args_overflow): Move to gimple-ssa-warn-access.cc.
6067 (get_attr_nonstring_decl): Move to tree.c.
6068 (fntype_argno_type): Move to gimple-ssa-warn-access.cc.
6069 (append_attrname): Same.
6070 (maybe_warn_rdwr_sizes): Same.
6071 (initialize_argument_information): Move code to
6072 gimple-ssa-warn-access.cc.
6073 * calls.h (maybe_warn_alloc_args_overflow): Move to
6074 gimple-ssa-warn-access.h.
6075 (get_attr_nonstring_decl): Move to tree.h.
6076 (maybe_warn_nonstring_arg): Move to gimple-ssa-warn-access.h.
6077 (enum size_range_flags): Move to pointer-query.h.
6078 (get_size_range): Same.
6079 * gimple-ssa-warn-access.cc (has_location): Remove unused overload
6080 to avoid Clang -Wunused-function.
6081 (get_size_range): Declare static.
6082 (maybe_emit_free_warning): Rename...
6083 (maybe_check_dealloc_call): ...to this for consistency.
6084 (class pass_waccess): Add members.
6085 (pass_waccess::~pass_waccess): Defined.
6086 (alloc_max_size): Move here from calls.c.
6087 (maybe_warn_alloc_args_overflow): Same.
6088 (check_alloca): New function.
6089 (check_alloc_size_call): New function.
6090 (check_strncat): Handle another warning flag.
6091 (pass_waccess::check_builtin): Handle alloca.
6092 (fntype_argno_type): Move here from calls.c.
6093 (append_attrname): Same.
6094 (maybe_warn_rdwr_sizes): Same.
6095 (pass_waccess::check_call): Define.
6096 (check_nonstring_args): New function.
6097 (pass_waccess::check): Call new member functions.
6098 (pass_waccess::execute): Enable ranger.
6099 * gimple-ssa-warn-access.h (get_size_range): Move here from calls.h.
6100 (maybe_warn_nonstring_arg): Same.
6101 * gimple-ssa-warn-restrict.c: Remove #include.
6102 * pointer-query.cc (get_size_range): Move here from calls.c.
6103 * pointer-query.h (enum size_range_flags): Same.
6104 (get_size_range): Same.
6105 * tree.c (get_attr_nonstring_decl): Move here from calls.c.
6106 * tree.h (get_attr_nonstring_decl): Move here from calls.h.
6108 2021-08-17 Thomas Schwinge <thomas@codesourcery.com>
6110 * ggc.h (ggc_collect): Add 'force_collect' parameter.
6111 * ggc-page.c (ggc_collect): Use that one instead of global
6112 'ggc_force_collect'. Adjust all users.
6113 * doc/gty.texi (Invoking the garbage collector): Update.
6114 * ggc-internal.h (ggc_force_collect): Remove.
6115 * ggc-common.c (ggc_force_collect): Likewise.
6116 * selftest.h (forcibly_ggc_collect): Remove.
6117 * ggc-tests.c (selftest::forcibly_ggc_collect): Likewise.
6118 * read-rtl-function.c (test_loading_labels): Adjust.
6119 * selftest-run-tests.c (run_tests): Likewise.
6121 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
6123 * config/darwin.c (darwin_file_end): Reset and reclaim the
6124 section names table at the end of compile.
6126 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
6129 * config.in: Regenerate.
6130 * config/i386/darwin.h (EXTRA_ASM_OPTS): New
6131 (ASM_SPEC): Pass options to disable branch shortening where
6133 * configure: Regenerate.
6134 * configure.ac: Detect versions of 'as' that support the
6135 optimisation which has the bug.
6137 2021-08-17 Richard Biener <rguenther@suse.de>
6139 * optabs-query.c (supports_vec_gather_load_p): Also check
6141 (supports_vec_scatter_store_p): Likewise.
6142 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Fall
6143 back to masked variants if non-masked are not supported.
6144 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
6145 When we need to use masked gather/scatter but do not have
6146 a mask set up a constant true one.
6147 * tree-vect-stmts.c (vect_check_scalar_mask): Also allow
6150 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
6152 * tree-ssa-ccp.c (bit_value_binop) [MINUS_EXPR]: Use same
6153 algorithm as PLUS_EXPR to improve subtraction bit bounds.
6154 [POINTER_DIFF_EXPR]: Treat as synonymous with MINUS_EXPR.
6156 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
6158 * tree-ssa-ccp.c (bit_value_mult_const): New helper function to
6159 calculate the mask-value pair result of a multiplication by an
6161 (bit_value_binop) [MULT_EXPR]: Call it from here for
6162 multiplications by (sparse) non-negative constants.
6164 2021-08-17 Christophe Lyon <christophe.lyon@foss.st.com>
6167 * config.gcc (gcc_cv_initfini_array): Leave undefined for
6168 uclinuxfdpiceabi targets.
6170 2021-08-17 Alexandre Oliva <oliva@adacore.com>
6172 * tree-inline.c (maybe_move_debug_stmts_to_successors): Don't
6173 reverse debug stmts.
6175 2021-08-17 Alexandre Oliva <oliva@adacore.com>
6177 * tree-cfg.c (dump_function_to_file): Use fun, not cfun.
6179 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
6181 * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete.
6182 (__LD4Q_LANE_FUNC): Likewise.
6183 (vld4_lane_u8): Define without macro.
6184 (vld4_lane_u16): Likewise.
6185 (vld4_lane_u32): Likewise.
6186 (vld4_lane_u64): Likewise.
6187 (vld4_lane_s8): Likewise.
6188 (vld4_lane_s16): Likewise.
6189 (vld4_lane_s32): Likewise.
6190 (vld4_lane_s64): Likewise.
6191 (vld4_lane_f16): Likewise.
6192 (vld4_lane_f32): Likewise.
6193 (vld4_lane_f64): Likewise.
6194 (vld4_lane_p8): Likewise.
6195 (vld4_lane_p16): Likewise.
6196 (vld4_lane_p64): Likewise.
6197 (vld4q_lane_u8): Likewise.
6198 (vld4q_lane_u16): Likewise.
6199 (vld4q_lane_u32): Likewise.
6200 (vld4q_lane_u64): Likewise.
6201 (vld4q_lane_s8): Likewise.
6202 (vld4q_lane_s16): Likewise.
6203 (vld4q_lane_s32): Likewise.
6204 (vld4q_lane_s64): Likewise.
6205 (vld4q_lane_f16): Likewise.
6206 (vld4q_lane_f32): Likewise.
6207 (vld4q_lane_f64): Likewise.
6208 (vld4q_lane_p8): Likewise.
6209 (vld4q_lane_p16): Likewise.
6210 (vld4q_lane_p64): Likewise.
6211 (vld4_lane_bf16): Likewise.
6212 (vld4q_lane_bf16): Likewise.
6214 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
6216 * config/aarch64/arm_neon.h (__LD3_LANE_FUNC): Delete.
6217 (__LD3Q_LANE_FUNC): Delete.
6218 (vld3_lane_u8): Define without macro.
6219 (vld3_lane_u16): Likewise.
6220 (vld3_lane_u32): Likewise.
6221 (vld3_lane_u64): Likewise.
6222 (vld3_lane_s8): Likewise.
6223 (vld3_lane_s16): Likewise.
6224 (vld3_lane_s32): Likewise.
6225 (vld3_lane_s64): Likewise.
6226 (vld3_lane_f16): Likewise.
6227 (vld3_lane_f32): Likewise.
6228 (vld3_lane_f64): Likewise.
6229 (vld3_lane_p8): Likewise.
6230 (vld3_lane_p16): Likewise.
6231 (vld3_lane_p64): Likewise.
6232 (vld3q_lane_u8): Likewise.
6233 (vld3q_lane_u16): Likewise.
6234 (vld3q_lane_u32): Likewise.
6235 (vld3q_lane_u64): Likewise.
6236 (vld3q_lane_s8): Likewise.
6237 (vld3q_lane_s16): Likewise.
6238 (vld3q_lane_s32): Likewise.
6239 (vld3q_lane_s64): Likewise.
6240 (vld3q_lane_f16): Likewise.
6241 (vld3q_lane_f32): Likewise.
6242 (vld3q_lane_f64): Likewise.
6243 (vld3q_lane_p8): Likewise.
6244 (vld3q_lane_p16): Likewise.
6245 (vld3q_lane_p64): Likewise.
6246 (vld3_lane_bf16): Likewise.
6247 (vld3q_lane_bf16): Likewise.
6249 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
6251 * config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Delete.
6252 (__LD2Q_LANE_FUNC): Likewise.
6253 (vld2_lane_u8): Define without macro.
6254 (vld2_lane_u16): Likewise.
6255 (vld2_lane_u32): Likewise.
6256 (vld2_lane_u64): Likewise.
6257 (vld2_lane_s8): Likewise.
6258 (vld2_lane_s16): Likewise.
6259 (vld2_lane_s32): Likewise.
6260 (vld2_lane_s64): Likewise.
6261 (vld2_lane_f16): Likewise.
6262 (vld2_lane_f32): Likewise.
6263 (vld2_lane_f64): Likewise.
6264 (vld2_lane_p8): Likewise.
6265 (vld2_lane_p16): Likewise.
6266 (vld2_lane_p64): Likewise.
6267 (vld2q_lane_u8): Likewise.
6268 (vld2q_lane_u16): Likewise.
6269 (vld2q_lane_u32): Likewise.
6270 (vld2q_lane_u64): Likewise.
6271 (vld2q_lane_s8): Likewise.
6272 (vld2q_lane_s16): Likewise.
6273 (vld2q_lane_s32): Likewise.
6274 (vld2q_lane_s64): Likewise.
6275 (vld2q_lane_f16): Likewise.
6276 (vld2q_lane_f32): Likewise.
6277 (vld2q_lane_f64): Likewise.
6278 (vld2q_lane_p8): Likewise.
6279 (vld2q_lane_p16): Likewise.
6280 (vld2q_lane_p64): Likewise.
6281 (vld2_lane_bf16): Likewise.
6282 (vld2q_lane_bf16): Likewise.
6284 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
6286 * haifa-sched.c (advance_one_cycle): Output more context-synchronization
6289 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
6291 * haifa-sched.c (enum rfs_decision, rfs_str): Add RFS_AUTOPREF.
6292 (rank_for_schedule): Use it.
6294 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
6296 PR rtl-optimization/91598
6297 * haifa-sched.c (autopref_rank_for_schedule): Prioritize "irrelevant"
6298 insns after memory reads and before memory writes.
6300 2021-08-17 Alistair_Lee <alistair.lee@arm.com>
6302 * rtl.h (CONST_VECTOR_P): New macro.
6303 * config/aarch64/aarch64.c (aarch64_get_sve_pred_bits): Use RTL
6304 code testing macros.
6305 (aarch64_ptrue_all_mode): Likewise.
6306 (aarch64_expand_mov_immediate): Likewise.
6307 (aarch64_const_vec_all_in_range_p): Likewise.
6308 (aarch64_rtx_costs): Likewise.
6309 (aarch64_legitimate_constant_p): Likewise.
6310 (aarch64_simd_valid_immediate): Likewise.
6311 (aarch64_simd_make_constant): Likewise.
6312 (aarch64_convert_mult_to_shift): Likewise.
6313 (aarch64_expand_sve_vec_perm): Likewise.
6314 (aarch64_vec_fpconst_pow_of_2): Likewise.
6316 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
6318 PR tree-optimization/101938
6319 * range-op.cc (operator_abs::op1_range): Special case
6320 -TYPE_MIN_VALUE for flag_wrapv.
6322 2021-08-17 Kewen Lin <linkw@linux.ibm.com>
6324 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Add the cost for
6327 2021-08-17 Jakub Jelinek <jakub@redhat.com>
6329 * tree.def (OMP_SCOPE): New tree code.
6330 * tree.h (OMP_SCOPE_BODY, OMP_SCOPE_CLAUSES): Define.
6331 * tree-nested.c (convert_nonlocal_reference_stmt,
6332 convert_local_reference_stmt, convert_gimple_call): Handle
6334 * tree-pretty-print.c (dump_generic_node): Handle OMP_SCOPE.
6335 * gimple.def (GIMPLE_OMP_SCOPE): New gimple code.
6336 * gimple.c (gimple_build_omp_scope): New function.
6337 (gimple_copy): Handle GIMPLE_OMP_SCOPE.
6338 * gimple.h (gimple_build_omp_scope): Declare.
6339 (gimple_has_substatements): Handle GIMPLE_OMP_SCOPE.
6340 (gimple_omp_scope_clauses, gimple_omp_scope_clauses_ptr,
6341 gimple_omp_scope_set_clauses): New inline functions.
6342 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_SCOPE.
6343 * gimple-pretty-print.c (dump_gimple_omp_scope): New function.
6344 (pp_gimple_stmt_1): Handle GIMPLE_OMP_SCOPE.
6345 * gimple-walk.c (walk_gimple_stmt): Likewise.
6346 * gimple-low.c (lower_stmt): Likewise.
6347 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
6348 (gimplify_scan_omp_clauses): For task reductions, handle OMP_SCOPE
6349 like ORT_WORKSHARE constructs. Adjust diagnostics for %<scope%>
6350 allowing task reductions. Reject inscan reductions on scope.
6351 (omp_find_stores_stmt): Handle GIMPLE_OMP_SCOPE.
6352 (gimplify_omp_workshare, gimplify_expr): Handle OMP_SCOPE.
6353 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_SCOPE.
6354 (estimate_num_insns): Likewise.
6355 * omp-low.c (build_outer_var_ref): Look through GIMPLE_OMP_SCOPE
6356 contexts if var isn't privatized there.
6357 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_SCOPE.
6358 (scan_omp_1_stmt): Likewise.
6359 (maybe_add_implicit_barrier_cancel): Look through outer
6361 (lower_omp_scope): New function.
6362 (lower_omp_task_reductions): Handle OMP_SCOPE.
6363 (lower_omp_1): Handle GIMPLE_OMP_SCOPE.
6364 (diagnose_sb_1, diagnose_sb_2): Likewise.
6365 * omp-expand.c (expand_omp_single): Support also GIMPLE_OMP_SCOPE.
6366 (expand_omp): Handle GIMPLE_OMP_SCOPE.
6367 (omp_make_gimple_edges): Likewise.
6368 * omp-builtins.def (BUILT_IN_GOMP_SCOPE_START): New built-in.
6370 2021-08-17 Richard Biener <rguenther@suse.de>
6372 PR tree-optimization/101925
6373 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Set
6374 reverse on COMPONENT_REF and ARRAY_REF according to
6375 what reverse_storage_order_for_component_p does.
6376 (vn_reference_eq): Compare reversed on reference ops.
6377 (reverse_storage_order_for_component_p): New overload.
6378 (vn_reference_lookup_3): Check reverse_storage_order_for_component_p
6379 on the reference looked up.
6381 2021-08-17 Jeff Law <jlaw@localhost.localdomain>
6383 * config/h8300/h8300.c (shift_alg_si): Avoid loops for most SImode
6385 (h8300_option_override): Use loops on H8/S more often when optimizing
6387 (get_shift_alg): Handle new "special" cases on H8/S. Simplify
6388 accordingly. Handle various arithmetic right shifts with special
6389 sequences that we couldn't handle before.
6391 2021-08-16 Jeff Law <jlaw@localhost.localdomain>
6393 * config.gcc (rl78-*-elf*): Do not include dbxelf.h.
6395 2021-08-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
6397 * config/sparc/rtemself.h (SPARC_GCOV_TYPE_SIZE): Define.
6398 * config/sparc/sparc.c (sparc_gcov_type_size): New.
6399 (TARGET_GCOV_TYPE_SIZE): Redefine if SPARC_GCOV_TYPE_SIZE is defined.
6400 * coverage.c (get_gcov_type): Use targetm.gcov_type_size().
6401 * doc/tm.texi (TARGET_GCOV_TYPE_SIZE): Add hook under "Misc".
6402 * doc/tm.texi.in: Regenerate.
6403 * target.def (gcov_type_size): New target hook.
6404 * targhooks.c (default_gcov_type_size): New.
6405 * targhooks.h (default_gcov_type_size): Declare.
6406 * tree-profile.c (gimple_gen_edge_profiler): Use precision of
6408 (gimple_gen_time_profiler): Likewise.
6410 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
6412 * dwarf2out.c (add_scalar_info): Deal with DW_AT_data_bit_offset.
6414 2021-08-16 Tobias Burnus <tobias@codesourcery.com>
6416 PR middle-end/101931
6417 * omp-low.c (omp_runtime_api_call): Update for routines
6418 added in the meanwhile.
6420 2021-08-16 Martin Liska <mliska@suse.cz>
6422 PR tree-optimization/100393
6423 * tree-switch-conversion.c (group_cluster::dump): Use
6424 get_comparison_count.
6425 (jump_table_cluster::find_jump_tables): Pre-compute number of
6426 comparisons and then decrement it. Cache also max_ratio.
6427 (jump_table_cluster::can_be_handled): Change signature.
6428 * tree-switch-conversion.h (get_comparison_count): New.
6430 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
6432 * dwarf2out.c (add_data_member_location_attribute): Use GNAT
6433 encodings only when -fgnat-encodings=all is specified.
6434 (add_bound_info): Likewise.
6435 (add_byte_size_attribute): Likewise.
6436 (gen_member_die): Likewise.
6438 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
6440 * omp-oacc-neuter-broadcast.cc
6441 (execute_omp_oacc_neuter_broadcast): Plug 'par' memory leak.
6443 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
6445 * omp-oacc-neuter-broadcast.cc
6446 (execute_omp_oacc_neuter_broadcast): Clarify memory management for
6449 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
6451 * omp-oacc-neuter-broadcast.cc (field_map): Move variable into...
6452 (execute_omp_oacc_neuter_broadcast): ... here.
6453 (install_var_field, build_receiver_ref, build_sender_ref): Take
6454 'field_map_t *' parameter. Adjust all users.
6455 (worker_single_copy, neuter_worker_single): Take a
6456 'record_field_map_t *' parameter. Adjust all users.
6458 2021-08-16 liuhongt <hongtao.liu@intel.com>
6461 * config/i386/i386.md (ldexp<mode>3): Force operands[1] to
6464 2021-08-16 Martin Liska <mliska@suse.cz>
6467 * multiple_target.c (create_dispatcher_calls): Make default
6468 function local only if it is a definition.
6470 2021-08-16 Martin Liska <mliska@suse.cz>
6473 * ipa-icf-gimple.c (func_checker::compare_ssa_name): Do not
6474 consider equal SSA_NAMEs when one is a param.
6476 2021-08-16 liuhongt <hongtao.liu@intel.com>
6479 * config/i386/i386-expand.c (ix86_expand_vec_perm_vpermt2):
6480 Support vpermi2b for V32QI/V16QImode.
6481 (ix86_extract_perm_from_pool_constant): New function.
6482 (ix86_expand_vec_one_operand_perm_avx512): Support
6483 vpermw/vpermb under TARGET_AVX512BW/TARGET_AVX512VBMI.
6484 (expand_vec_perm_1): Adjust comments for upper.
6485 * config/i386/i386-protos.h (ix86_extract_perm_from_pool_constant):
6487 * config/i386/predicates.md (permvar_truncate_operand): New predicate.
6488 (pshufb_truncv4siv4hi_operand): Ditto.
6489 (pshufb_truncv8hiv8qi_operand): Ditto.
6490 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1):
6491 New pre_reload define_insn_and_split.
6492 (*avx512f_permvar_truncv8siv8hi_1): Ditto.
6493 (*avx512f_vpermvar_truncv8div8si_1): Ditto.
6494 (*avx512f_permvar_truncv32hiv32qi_1): Ditto.
6495 (*avx512f_permvar_truncv16hiv16qi_1): Ditto.
6496 (*avx512f_permvar_truncv4div4si_1): Ditto.
6497 (*avx512f_pshufb_truncv8hiv8qi_1): Ditto.
6498 (*avx512f_pshufb_truncv4siv4hi_1): Ditto.
6499 (*avx512f_pshufd_truncv2div2si_1): Ditto.
6501 2021-08-16 Kito Cheng <kito.cheng@sifive.com>
6503 * config/riscv/multilib-generator: Support code model option for
6505 * doc/install.texi: Add document of new option for
6506 --with-multilib-generator.
6508 2021-08-15 Clément Chigot <clement.chigot@atos.net>
6510 * config/rs6000/rs6000.c (xcoff_tls_exec_model_detected): New.
6511 (rs6000_legitimize_tls_address_aix): Use it.
6512 (rs6000_xcoff_file_end): Add ".ref __tls_get_addr" when
6513 xcoff_tls_exec_model_detected is true.
6515 2021-08-15 Jeff Law <jlaw@localhost.localdomain>
6517 * config/h8300/h8300.c (shift_alg_si): Retune H8/300H shifts
6518 to allow a bit more code growth, saving many dozens of cycles.
6519 (h8300_option_override): Adjus shift_alg_si if optimizing for
6521 (get_shift_alg): Use special + inline shifts for residuals
6524 2021-08-14 Stafford Horne <shorne@gmail.com>
6527 * config/or1k/or1k-opts.h: New file.
6528 * config/or1k/or1k.c (or1k_legitimize_address_1, print_reloc):
6529 Support generating gotha relocations if -mcmodel=large is
6531 * config/or1k/or1k.h (TARGET_CMODEL_SMALL, TARGET_CMODEL_LARGE):
6533 * config/or1k/or1k.opt (mcmodel=): New option.
6534 * doc/invoke.texi (OpenRISC Options): Document mcmodel.
6536 2021-08-14 Martin Sebor <msebor@redhat.com>
6538 PR middle-end/101791
6539 * gimple-ssa-warn-access.cc (new_delete_mismatch_p): Use new argument
6540 to valid_new_delete_pair_p.
6541 * tree.c (valid_new_delete_pair_p): Add argument.
6542 * tree.h (valid_new_delete_pair_p): Same.
6544 2021-08-14 Jakub Jelinek <jakub@redhat.com>
6547 * config/i386/i386-expand.c (expand_vec_perm_broadcast_1)
6548 <case E_V64QImode>: For this mode assert
6549 !TARGET_AVX512BW || d->perm[0] rather than !TARGET_AVX2 || d->perm[0].
6551 2021-08-13 Michael Meissner <meissner@linux.ibm.com>
6554 * config/rs6000/altivec.md (xxeval): Use register_predicate
6555 instead of altivec_register_predicate.
6557 2021-08-13 Martin Sebor <msebor@redhat.com>
6559 PR middle-end/101734
6560 * tree-ssa-uninit.c (maybe_warn_read_write_only): New function.
6561 (maybe_warn_operand): Call it.
6563 2021-08-13 Martin Liska <mliska@suse.cz>
6566 * attribs.c (decl_attributes): Make naked functions "noipa"
6569 2021-08-13 Martin Liska <mliska@suse.cz>
6572 * symtab.c (symtab_node::noninterposable_alias): Do not create
6573 local aliases for target_clone functions as the clonning pass
6576 2021-08-13 Martin Liska <mliska@suse.cz>
6578 * opts.c (LIVE_PATCHING_OPTION): Define.
6579 (control_options_for_live_patching): Use it in error messages.
6581 2021-08-13 Jan Hubicka <hubicka@ucw.cz>
6583 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOREAD.
6584 (implicit_const_eaf_flags, implicit_pure_eaf_flags,
6585 ignore_stores_eaf_flags): New constants.
6586 (remove_useless_eaf_flags): New function.
6587 (eaf_flags_useful_p): Use it.
6588 (deref_flags): Add EAF_NOT_RETURNED if flag is unused;
6590 (modref_lattice::init): Add EAF_NOREAD.
6591 (modref_lattice::add_escape_point): Do not reacord escape point if
6593 (modref_lattice::merge): EAF_NOESCAPE implies EAF_NODIRECTESCAPE;
6594 use remove_useless_eaf_flags.
6595 (modref_lattice::merge_deref): Use ignore_stores_eaf_flags.
6596 (modref_lattice::merge_direct_load): Add EAF_NOREAD
6597 (analyze_ssa_name_flags): Fix handling EAF_NOT_RETURNED
6598 (analyze_parms): Use remove_useless_eaf_flags.
6599 (ipa_merge_modref_summary_after_inlining): Use ignore_stores_eaf_flags.
6600 (modref_merge_call_site_flags): Add caller and ecf_flags parameter;
6601 use remove_useless_eaf_flags.
6602 (modref_propagate_flags_in_scc): Update.
6603 * ipa-modref.h: Turn eaf_flags_t back to char.
6604 * tree-core.h (EAF_NOT_RETURNED): Fix.
6605 (EAF_NOREAD): New constant
6606 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1): Check for
6608 * tree-ssa-structalias.c (handle_rhs_call): Handle new flags.
6609 (handle_pure_call): Likewise.
6611 2021-08-12 Jakub Jelinek <jakub@redhat.com>
6613 * tree.def (OMP_MASKED): New tree code.
6614 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_FILTER.
6615 * tree.h (OMP_MASKED_BODY, OMP_MASKED_CLAUSES, OMP_MASKED_COMBINED,
6616 OMP_CLAUSE_FILTER_EXPR): Define.
6617 * tree.c (omp_clause_num_ops): Add OMP_CLAUSE_FILTER entry.
6618 (omp_clause_code_name): Likewise.
6619 (walk_tree_1): Handle OMP_CLAUSE_FILTER.
6620 * tree-nested.c (convert_nonlocal_omp_clauses,
6621 convert_local_omp_clauses): Handle OMP_CLAUSE_FILTER.
6622 (convert_nonlocal_reference_stmt, convert_local_reference_stmt,
6623 convert_gimple_call): Handle GIMPLE_OMP_MASTER.
6624 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_FILTER.
6625 (dump_generic_node): Handle OMP_MASTER.
6626 * gimple.def (GIMPLE_OMP_MASKED): New gimple code.
6627 * gimple.c (gimple_build_omp_masked): New function.
6628 (gimple_copy): Handle GIMPLE_OMP_MASKED.
6629 * gimple.h (gimple_build_omp_masked): Declare.
6630 (gimple_has_substatements): Handle GIMPLE_OMP_MASKED.
6631 (gimple_omp_masked_clauses, gimple_omp_masked_clauses_ptr,
6632 gimple_omp_masked_set_clauses): New inline functions.
6633 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_MASKED.
6634 * gimple-pretty-print.c (dump_gimple_omp_masked): New function.
6635 (pp_gimple_stmt_1): Handle GIMPLE_OMP_MASKED.
6636 * gimple-walk.c (walk_gimple_stmt): Likewise.
6637 * gimple-low.c (lower_stmt): Likewise.
6638 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
6639 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_FILTER. For clauses
6640 that take one expression rather than decl or constant, force
6641 gimplification of that into a SSA_NAME or temporary unless min
6643 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_FILTER.
6644 (gimplify_expr): Handle OMP_MASKED.
6645 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_MASKED.
6646 (estimate_num_insns): Likewise.
6647 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_FILTER.
6648 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_MASKED. Adjust
6649 diagnostics for existence of masked construct.
6650 (scan_omp_1_stmt, lower_omp_master, lower_omp_1, diagnose_sb_1,
6651 diagnose_sb_2): Handle GIMPLE_OMP_MASKED.
6652 * omp-expand.c (expand_omp_synch, expand_omp, omp_make_gimple_edges):
6655 2021-08-12 Uroš Bizjak <ubizjak@gmail.com>
6658 * config/i386/i386.md (avx512f_scalef<mode>2): New insn pattern.
6659 (ldexp<mode>3): Use avx512f_scalef<mode>2.
6660 (UNSPEC_SCALEF): Move from ...
6661 * config/i386/sse.md (UNSPEC_SCALEF): ... here.
6663 2021-08-12 Jan Hubicka <hubicka@ucw.cz>
6665 * ipa-split.c (consider_split): Fix condition testing void functions.
6667 2021-08-12 Aldy Hernandez <aldyh@redhat.com>
6669 * doc/invoke.texi: Remove docs for threader-mode param.
6670 * flag-types.h (enum threader_mode): Remove.
6671 * params.opt: Remove threader-mode param.
6672 * tree-ssa-threadbackward.c (class back_threader): Remove
6673 path_is_unreachable_p.
6674 Make find_paths private.
6675 Add maybe_thread and thread_through_all_blocks.
6676 Remove reference marker for m_registry.
6677 Remove reference marker for m_profit.
6678 (back_threader::back_threader): Adjust for registry and profit not
6680 (dump_path): Move down.
6682 (class thread_jumps): Remove.
6683 (class back_threader_registry): Remove m_all_paths.
6685 (thread_jumps::thread_through_all_blocks): Move to back_threader
6687 (fsm_find_thread_path): Remove
6688 (back_threader::maybe_thread): New.
6689 (back_threader::thread_through_all_blocks): Move from
6691 (back_threader_registry::back_threader_registry): Remove
6693 (back_threader_registry::~back_threader_registry): Remove.
6694 (thread_jumps::find_taken_edge): Remove.
6695 (thread_jumps::check_subpath_and_update_thread_path): Remove.
6696 (thread_jumps::maybe_register_path): Remove.
6697 (thread_jumps::handle_phi): Remove.
6698 (handle_assignment_p): Remove.
6699 (thread_jumps::handle_assignment): Remove.
6700 (thread_jumps::fsm_find_control_statement_thread_paths): Remove.
6701 (thread_jumps::find_jump_threads_backwards): Remove.
6702 (thread_jumps::find_jump_threads_backwards_with_ranger): Remove.
6703 (try_thread_blocks): Rename find_jump_threads_backwards to
6705 (pass_early_thread_jumps::execute): Same.
6707 2021-08-12 Tobias Burnus <tobias@codesourcery.com>
6709 * tree-core.h (omp_clause_proc_bind_kind): Add
6710 OMP_CLAUSE_PROC_BIND_PRIMARY.
6711 * tree-pretty-print.c (dump_omp_clause): Add TODO comment to
6712 change 'master' to 'primary' in proc_bind for OpenMP 5.1.
6714 2021-08-12 Claudiu Zissulescu <claziss@synopsys.com>
6716 * common/config/arc/arc-common.c (arc_option_init_struct): Remove
6717 fno-common reference.
6718 * config/arc/arc.c (arc_override_options): Remove overriding of
6721 2021-08-12 Jakub Jelinek <jakub@redhat.com>
6724 * config/i386/i386-expand.c (ix86_expand_vec_one_operand_perm_avx512):
6725 If d->testing_p, return true after performing checks instead of
6726 actually expanding the insn.
6727 (expand_vec_perm_broadcast_1): Handle V32HImode - assert
6728 !TARGET_AVX512BW and return false.
6730 2021-08-12 Eric Botcazou <ebotcazou@gcc.gnu.org>
6732 * configure.ac (PE linker --disable-dynamicbase support): New check.
6733 * configure: Regenerate.
6734 * config.in: Likewise.
6735 * config/i386/mingw32.h (LINK_SPEC_DISABLE_DYNAMICBASE): New define.
6736 (LINK_SPEC): Use it.
6737 * config/i386/mingw-w64.h (LINK_SPEC_DISABLE_DYNAMICBASE): Likewise.
6738 (LINK_SPEC): Likewise.
6740 2021-08-12 liuhongt <hongtao.liu@intel.com>
6743 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_2): New
6744 post_reload define_insn_and_split.
6745 (*avx512bw_zero_extendv32qiv32hi2_2): Ditto.
6746 (*sse4_1_zero_extendv8qiv8hi2_4): Ditto.
6747 (*avx512f_zero_extendv16hiv16si2_2): Ditto.
6748 (*avx2_zero_extendv8hiv8si2_2): Ditto.
6749 (*sse4_1_zero_extendv4hiv4si2_4): Ditto.
6750 (*avx512f_zero_extendv8siv8di2_2): Ditto.
6751 (*avx2_zero_extendv4siv4di2_2): Ditto.
6752 (*sse4_1_zero_extendv2siv2di2_4): Ditto.
6753 (VI248_256, VI248_512, VI148_512, VI148_256, VI148_128): New
6756 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
6758 * config/rs6000/rs6000-builtin-new.def: Add always, power5, and
6761 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
6763 * config/rs6000/rs6000-builtin-new.def: Add vsx stanza.
6765 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
6767 * config/rs6000/rs6000-builtin-new.def: Finish altivec stanza.
6768 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
6769 initialization of pcvoid_type_node here...
6770 (altivec_init_builtins): ...from here.
6771 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add
6772 RS6000_BTI_const_ptr_void.
6773 (pcvoid_type_node): New macro.
6775 2021-08-11 Richard Biener <rguenther@suse.de>
6778 * tree-ssa-forwprop.c (pass_forwprop::execute): Do not decompose
6779 hard-register accesses.
6781 2021-08-11 Richard Biener <rguenther@suse.de>
6783 * tree-ssa-operands.c (operands_scanner::get_expr_operands):
6784 Do not look at COMPONENT_REF FIELD_DECLs TREE_THIS_VOLATILE
6785 to determine has_volatile_ops.
6787 2021-08-11 Eric Botcazou <ebotcazou@gcc.gnu.org>
6789 * cfgexpand.c (expand_used_vars): Reuse attribs local variable.
6791 2021-08-11 Jan Hubicka <hubicka@ucw.cz>
6792 Alexandre Oliva <oliva@adacore.com>
6794 * ipa-modref.c (modref_lattice::dump): Fix escape_point's min_flags
6796 (modref_lattice::merge_deref): Fix handling of indirect scape points.
6797 (update_escape_summary_1): Likewise.
6798 (update_escape_summary): Likewise.
6799 (ipa_merge_modref_summary_after_inlining): Likewise.
6801 2021-08-11 Richard Biener <rguenther@suse.de>
6803 PR middle-end/101858
6804 * fold-const.c (fold_binary_loc): Guard simplification
6805 of X < (cast) (1 << Y) to integer types.
6807 2021-08-11 Richard Biener <rguenther@suse.de>
6809 PR tree-optimization/101861
6810 * tree-vect-stmts.c (vectorizable_load): Fix error in
6811 previous change with regard to gather vectorization.
6813 2021-08-11 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
6816 * config/arm/arm_neon.h (vdup_n_s8): Replace call to builtin
6818 (vdup_n_s16): Likewise.
6819 (vdup_n_s32): Likewise.
6820 (vdup_n_s64): Likewise.
6821 (vdup_n_u8): Likewise.
6822 (vdup_n_u16): Likewise.
6823 (vdup_n_u32): Likewise.
6824 (vdup_n_u64): Likewise.
6825 (vdup_n_p8): Likewise.
6826 (vdup_n_p16): Likewise.
6827 (vdup_n_p64): Likewise.
6828 (vdup_n_f16): Likewise.
6829 (vdup_n_f32): Likewise.
6830 (vdupq_n_s8): Likewise.
6831 (vdupq_n_s16): Likewise.
6832 (vdupq_n_s32): Likewise.
6833 (vdupq_n_s64): Likewise.
6834 (vdupq_n_u8): Likewise.
6835 (vdupq_n_u16): Likewise.
6836 (vdupq_n_u32): Likewise.
6837 (vdupq_n_u64): Likewise.
6838 (vdupq_n_p8): Likewise.
6839 (vdupq_n_p16): Likewise.
6840 (vdupq_n_p64): Likewise.
6841 (vdupq_n_f16): Likewise.
6842 (vdupq_n_f32): Likewise.
6843 (vmov_n_s8): Replace call to builtin with call to corresponding
6845 (vmov_n_s16): Likewise.
6846 (vmov_n_s32): Likewise.
6847 (vmov_n_s64): Likewise.
6848 (vmov_n_u8): Likewise.
6849 (vmov_n_u16): Likewise.
6850 (vmov_n_u32): Likewise.
6851 (vmov_n_u64): Likewise.
6852 (vmov_n_p8): Likewise.
6853 (vmov_n_p16): Likewise.
6854 (vmov_n_f16): Likewise.
6855 (vmov_n_f32): Likewise.
6856 (vmovq_n_s8): Likewise.
6857 (vmovq_n_s16): Likewise.
6858 (vmovq_n_s32): Likewise.
6859 (vmovq_n_s64): Likewise.
6860 (vmovq_n_u8): Likewise.
6861 (vmovq_n_u16): Likewise.
6862 (vmovq_n_u32): Likewise.
6863 (vmovq_n_u64): Likewise.
6864 (vmovq_n_p8): Likewise.
6865 (vmovq_n_p16): Likewise.
6866 (vmovq_n_f16): Likewise.
6867 (vmovq_n_f32): Likewise.
6868 * config/arm/arm_neon_builtins.def: Remove entries for vdup_n.
6870 2021-08-11 liuhongt <hongtao.liu@intel.com>
6873 * config/i386/i386.md (ldexp<mode>3): Extend to vscalefs[sd]
6874 when TARGET_AVX512F and TARGET_SSE_MATH.
6876 2021-08-10 Jakub Jelinek <jakub@redhat.com>
6879 * config/i386/i386-expand.c (expand_vec_perm_even_odd): Return false
6880 for V32HImode if !TARGET_AVX512BW.
6881 (ix86_vectorize_vec_perm_const) <case E_V32HImode, case E_V64QImode>:
6882 If !TARGET_AVX512BW and TARGET_AVX512F and d.testing_p, don't fail
6883 early, but actually check the permutation.
6885 2021-08-10 Richard Biener <rguenther@suse.de>
6887 PR tree-optimization/101809
6888 * tree-vect-stmts.c (get_load_store_type): Allow emulated
6889 gathers with offset vector nunits being a constant multiple
6890 of the data vector nunits.
6891 (vect_get_gather_scatter_ops): Use the appropriate nunits
6892 for the offset vector defs.
6893 (vectorizable_store): Adjust call to
6894 vect_get_gather_scatter_ops.
6895 (vectorizable_load): Likewise. Handle the case of less
6896 offset vectors than data vectors.
6898 2021-08-10 Jakub Jelinek <jakub@redhat.com>
6901 * config/i386/sse.md (*avx512f_shuf_<shuffletype>64x2_1<mask_name>_1,
6902 *avx512f_shuf_<shuffletype>32x4_1<mask_name>_1): New define_insn
6905 2021-08-10 Richard Biener <rguenther@suse.de>
6907 PR tree-optimization/101801
6908 PR tree-optimization/101819
6909 * tree-vectorizer.h (vect_emulated_vector_p): Declare.
6910 * tree-vect-loop.c (vect_emulated_vector_p): New function.
6911 (vectorizable_reduction): Re-instantiate a check for emulated
6913 * tree-vect-stmts.c (vectorizable_shift): Likewise.
6914 (vectorizable_operation): Likewise. Cost emulated vector
6915 operations according to the scalar sequence synthesized by
6918 2021-08-10 Richard Biener <rguenther@suse.de>
6920 PR middle-end/101824
6921 * tree-nested.c (get_frame_field): Mark the COMPONENT_REF as
6922 volatile in case the variable was.
6924 2021-08-10 H.J. Lu <hjl.tools@gmail.com>
6927 * config/i386/constraints.md (BC): Document for integer SSE
6928 constant all bits set operand.
6929 (BF): New constraint for const floating-point all bits set
6931 * config/i386/i386.c (standard_sse_constant_p): Likewise.
6932 (standard_sse_constant_opcode): Likewise.
6933 * config/i386/sse.md (sseconstm1): New mode attribute.
6934 (mov<mode>_internal): Replace BC with <sseconstm1>.
6936 2021-08-10 liuhongt <hongtao.liu@intel.com>
6938 * config/i386/sse.md (cond_<insn><mode>): New expander.
6939 (VI248_AVX512VLBW): New mode iterator.
6940 * config/i386/predicates.md
6941 (nonimmediate_or_const_vec_dup_operand): New predicate.
6943 2021-08-09 Andrew MacLeod <amacleod@redhat.com>
6945 PR tree-optimization/101741
6946 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Check
6947 type of parameter for toupper/tolower.
6949 2021-08-09 Martin Jambor <mjambor@suse.cz>
6952 * ipa-prop.c (propagate_controlled_uses): Removed a spurious space.
6954 2021-08-09 Pat Haugen <pthaugen@linux.ibm.com>
6956 * config/rs6000/rs6000.c (is_load_insn1): Verify destination is a
6958 (is_store_insn1): Verify source is a register.
6960 2021-08-09 Uroš Bizjak <ubizjak@gmail.com>
6963 * config/i386/mmx.md (<any_logic:code>v2sf3):
6964 Rename from *mmx_<any_logic:code>v2sf3
6966 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6968 * config/nvptx/nvptx.c: Cross-reference parts adapted in
6969 'gcc/omp-oacc-neuter-broadcast.cc'.
6970 * omp-low.c: Likewise.
6971 * omp-oacc-neuter-broadcast.cc: Cross-reference parts adapted from
6974 2021-08-09 Julian Brown <julian@codesourcery.com>
6975 Kwok Cheung Yeung <kcy@codesourcery.com>
6976 Thomas Schwinge <thomas@codesourcery.com>
6978 * config/gcn/gcn.c (gcn_init_builtins): Override decls for
6979 BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START,
6980 BUILT_IN_GOACC_SINGLE_COPY_END and BUILT_IN_GOACC_BARRIER.
6981 (gcn_goacc_validate_dims): Turn on worker partitioning unconditionally.
6982 (gcn_fork_join): Update comment.
6983 * config/gcn/gcn.opt (flag_worker_partitioning): Remove.
6984 (macc_experimental_workers): Remove unused option.
6986 2021-08-09 Julian Brown <julian@codesourcery.com>
6987 Nathan Sidwell <nathan@codesourcery.com> (via 'gcc/config/nvptx/nvptx.c' master)
6988 Kwok Cheung Yeung <kcy@codesourcery.com>
6989 Thomas Schwinge <thomas@codesourcery.com>
6991 * Makefile.in (OBJS): Add omp-oacc-neuter-broadcast.o.
6992 * doc/tm.texi.in (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD):
6993 Add documentation hook.
6994 * doc/tm.texi: Regenerate.
6995 * omp-oacc-neuter-broadcast.cc: New file.
6996 * omp-builtins.def (BUILT_IN_GOACC_BARRIER)
6997 (BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START)
6998 (BUILT_IN_GOACC_SINGLE_COPY_END): New builtins.
6999 * passes.def (pass_omp_oacc_neuter_broadcast): Add pass.
7000 * target.def (goacc.create_worker_broadcast_record): Add target
7002 * tree-pass.h (make_pass_omp_oacc_neuter_broadcast): Add
7004 * config/gcn/gcn-protos.h (gcn_goacc_adjust_propagation_record):
7005 Rename prototype to...
7006 (gcn_goacc_create_worker_broadcast_record): ... this.
7007 * config/gcn/gcn-tree.c (gcn_goacc_adjust_propagation_record): Rename
7009 (gcn_goacc_create_worker_broadcast_record): ... this.
7010 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_PROPAGATION_RECORD):
7012 (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD): ... this.
7014 2021-08-09 Tejas Belagod <tejas.belagod@arm.com>
7017 * config/aarch64/aarch64-simd.md (vlshr<mode>3, vashr<mode>3): Use
7020 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7022 * Makefile.in (GTFILES): Remove '$(srcdir)/omp-offload.c'.
7024 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7026 * builtins.def (DEF_GOACC_BUILTIN, DEF_GOMP_BUILTIN): Don't
7027 consider '-foffload-abi'.
7028 * common.opt (-foffload-abi): Remove 'Var', 'Init'.
7029 * opts.c (common_handle_option) <-foffload-abi> [ACCEL_COMPILER]:
7032 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7034 * optc-gen.awk: Sanity check that 'Init' doesn't appear without
7037 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7039 * omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE): Remove.
7041 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7043 * doc/gty.texi (Files): Update.
7045 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7047 * doc/gty.texi (Files): Fix GTY header file example.
7049 2021-08-09 Roger Sayle <roger@nextmovesoftware.com>
7051 * tree-ssa-ccp.c (value_mask_to_min_max): Helper function to
7052 determine the upper and lower bounds from a mask-value pair.
7053 (bit_value_unop) [ABS_EXPR, ABSU_EXPR]: Add support for
7054 absolute value and unsigned absolute value expressions.
7055 (bit_value_binop): Initialize *VAL's precision.
7056 [LT_EXPR, LE_EXPR]: Use value_mask_to_min_max to determine
7057 upper and lower bounds of operands. Add LE_EXPR/GE_EXPR
7058 support when the operands are unknown but potentially equal.
7059 [MIN_EXPR, MAX_EXPR]: Support minimum/maximum expressions.
7061 2021-08-09 Bin Cheng <bin.cheng@linux.alibaba.com>
7063 * config/aarch64/aarch64.md
7064 (*extend<SHORT:mode><GPI:mode>2_aarch64): Use %<GPI:w>0.
7066 2021-08-08 Sergei Trofimovich <siarheit@google.com>
7068 * lra-constraints.c: Fix s/otput/output/ typo.
7070 2021-08-06 Martin Sebor <msebor@redhat.com>
7072 * builtins.c (expand_builtin_memchr): Move to gimple-ssa-warn-access.cc.
7073 (expand_builtin_strcat): Same.
7074 (expand_builtin_stpncpy): Same.
7075 (expand_builtin_strncat): Same.
7076 (check_read_access): Same.
7077 (check_memop_access): Same.
7078 (expand_builtin_strlen): Move checks to gimple-ssa-warn-access.cc.
7079 (expand_builtin_strnlen): Same.
7080 (expand_builtin_memcpy): Same.
7081 (expand_builtin_memmove): Same.
7082 (expand_builtin_mempcpy): Same.
7083 (expand_builtin_strcpy): Same.
7084 (expand_builtin_strcpy_args): Same.
7085 (expand_builtin_stpcpy_1): Same.
7086 (expand_builtin_strncpy): Same.
7087 (expand_builtin_memset): Same.
7088 (expand_builtin_bzero): Same.
7089 (expand_builtin_strcmp): Same.
7090 (expand_builtin_strncmp): Same.
7091 (expand_builtin): Remove handlers.
7092 (fold_builtin_strlen): Add a comment.
7093 * builtins.h (check_access): Move to gimple-ssa-warn-access.cc.
7094 * calls.c (maybe_warn_nonstring_arg): Same.
7095 * diagnostic-spec.c (nowarn_spec_t::nowarn_spec_t): Add warning option.
7096 * gimple-fold.c (gimple_fold_builtin_strcpy): Pass argument to callee.
7097 (gimple_fold_builtin_stpcpy): Same.
7098 * gimple-ssa-warn-access.cc (has_location): New function.
7099 (get_location): Same.
7100 (get_callee_fndecl): Same.
7103 (warn_string_no_nul): Define.
7104 (unterminated_array): Same.
7105 (check_nul_terminated_array): Same.
7106 (maybe_warn_nonstring_arg): Same.
7107 (maybe_warn_for_bound): Same.
7108 (warn_for_access): Same.
7109 (check_access): Same.
7110 (check_memop_access): Same.
7111 (check_read_access): Same.
7112 (warn_dealloc_offset): Use helper functions.
7113 (maybe_emit_free_warning): Same.
7114 (class pass_waccess): Add members.
7115 (check_strcat): New function.
7116 (check_strncat): New function.
7117 (check_stxcpy): New function.
7118 (check_stxncpy): New function.
7119 (check_strncmp): New function.
7120 (pass_waccess::check_builtin): New function.
7121 (pass_waccess::check): Call it.
7122 * gimple-ssa-warn-access.h (warn_string_no_nul): Move here from
7124 (maybe_warn_for_bound): Same.
7125 (check_access): Same.
7126 (check_memop_access): Same.
7127 (check_read_access): Same.
7128 * pointer-query.h (struct access_data): Define a ctor overload.
7130 2021-08-06 Richard Biener <rguenther@suse.de>
7132 PR tree-optimization/101801
7133 * tree-vectorizer.h (vect_worthwhile_without_simd_p): Rename...
7134 (vect_can_vectorize_without_simd_p): ... to this.
7135 * tree-vect-loop.c (vect_worthwhile_without_simd_p): Rename...
7136 (vect_can_vectorize_without_simd_p): ... to this and fold
7137 in vect_min_worthwhile_factor.
7138 (vect_min_worthwhile_factor): Remove.
7139 (vectorizable_reduction): Adjust and remove the cost part.
7140 * tree-vect-stmts.c (vectorizable_shift): Likewise.
7141 (vectorizable_operation): Likewise.
7143 2021-08-06 Uroš Bizjak <ubizjak@gmail.com>
7146 * config/i386/i386.md (cmove reg-to-reg move elimination peephole2s):
7147 Add general_gr_operand predicate to operand 3.
7149 2021-08-06 Roger Sayle <roger@nextmovesoftware.com>
7151 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern): Use
7152 CFN_BUILT_IN_CLRSB* instead of BUILT_IN_CLRSB* for consistency.
7154 2021-08-06 Tamar Christina <tamar.christina@arm.com>
7156 * config/aarch64/aarch64-sve-builtins.cc (register_svpattern,
7157 register_svprfop): Pass vec<> by pointer.
7158 * langhooks-def.h (lhd_simulate_enum_decl): Likewise.
7159 * langhooks.c (lhd_simulate_enum_decl): Likewise.
7160 * langhooks.h (struct lang_hooks_for_types): Likewise.
7162 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
7164 * config/aarch64/arm_neon.h (vst1_bf16_x2): Use
7165 __builtin_memcpy instead of constructing an additional
7166 __builtin_aarch64_simd_oi one vector at a time.
7167 (vst1q_bf16_x2): Likewise.
7168 (vst1_bf16_x3): Use __builtin_memcpy instead of constructing
7169 an additional __builtin_aarch64_simd_ci one vector at a time.
7170 (vst1q_bf16_x3): Likewise.
7171 (vst1_bf16_x4): Use __builtin_memcpy instead of a union.
7172 (vst1q_bf16_x4): Likewise.
7173 (vst2_bf16): Use __builtin_memcpy instead of constructing an
7174 additional __builtin_aarch64_simd_oi one vector at a time.
7175 (vst2q_bf16): Likewise.
7176 (vst3_bf16): Use __builtin_memcpy instead of constructing an
7177 additional __builtin_aarch64_simd_ci mode one vector at a
7179 (vst3q_bf16): Likewise.
7180 (vst4_bf16): Use __builtin_memcpy instead of constructing an
7181 additional __builtin_aarch64_simd_xi one vector at a time.
7182 (vst4q_bf16): Likewise.
7184 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
7186 * config/aarch64/arm_neon.h (__ST2_LANE_FUNC): Delete.
7187 (__ST2Q_LANE_FUNC): Delete.
7188 (vst2_lane_f16): Use __builtin_memcpy to copy vector
7189 structure instead of constructing __builtin_aarch64_simd_oi
7190 one vector at a time.
7191 (vst2_lane_f32): Likewise.
7192 (vst2_lane_f64): Likewise.
7193 (vst2_lane_p8): Likewise.
7194 (vst2_lane_p16): Likewise.
7195 (vst2_lane_p64): Likewise.
7196 (vst2_lane_s8): Likewise.
7197 (vst2_lane_s16): Likewise.
7198 (vst2_lane_s32): Likewise.
7199 (vst2_lane_s64): Likewise.
7200 (vst2_lane_u8): Likewise.
7201 (vst2_lane_u16): Likewise.
7202 (vst2_lane_u32): Likewise.
7203 (vst2_lane_u64): Likewise.
7204 (vst2_lane_bf16): Likewise.
7205 (vst2q_lane_f16): Use __builtin_memcpy to copy vector
7206 structure instead of using a union.
7207 (vst2q_lane_f32): Likewise.
7208 (vst2q_lane_f64): Likewise.
7209 (vst2q_lane_p8): Likewise.
7210 (vst2q_lane_p16): Likewise.
7211 (vst2q_lane_p64): Likewise.
7212 (vst2q_lane_s8): Likewise.
7213 (vst2q_lane_s16): Likewise.
7214 (vst2q_lane_s32): Likewise.
7215 (vst2q_lane_s64): Likewise.
7216 (vst2q_lane_u8): Likewise.
7217 (vst2q_lane_u16): Likewise.
7218 (vst2q_lane_u32): Likewise.
7219 (vst2q_lane_u64): Likewise.
7220 (vst2q_lane_bf16): Likewise.
7222 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
7224 * config/aarch64/arm_neon.h (__ST3_LANE_FUNC): Delete.
7225 (__ST3Q_LANE_FUNC): Delete.
7226 (vst3_lane_f16): Use __builtin_memcpy to copy vector
7227 structure instead of constructing __builtin_aarch64_simd_ci
7228 one vector at a time.
7229 (vst3_lane_f32): Likewise.
7230 (vst3_lane_f64): Likewise.
7231 (vst3_lane_p8): Likewise.
7232 (vst3_lane_p16): Likewise.
7233 (vst3_lane_p64): Likewise.
7234 (vst3_lane_s8): Likewise.
7235 (vst3_lane_s16): Likewise.
7236 (vst3_lane_s32): Likewise.
7237 (vst3_lane_s64): Likewise.
7238 (vst3_lane_u8): Likewise.
7239 (vst3_lane_u16): Likewise.
7240 (vst3_lane_u32): Likewise.
7241 (vst3_lane_u64): Likewise.
7242 (vst3_lane_bf16): Likewise.
7243 (vst3q_lane_f16): Use __builtin_memcpy to copy vector
7244 structure instead of using a union.
7245 (vst3q_lane_f32): Likewise.
7246 (vst3q_lane_f64): Likewise.
7247 (vst3q_lane_p8): Likewise.
7248 (vst3q_lane_p16): Likewise.
7249 (vst3q_lane_p64): Likewise.
7250 (vst3q_lane_s8): Likewise.
7251 (vst3q_lane_s16): Likewise.
7252 (vst3q_lane_s32): Likewise.
7253 (vst3q_lane_s64): Likewise.
7254 (vst3q_lane_u8): Likewise.
7255 (vst3q_lane_u16): Likewise.
7256 (vst3q_lane_u32): Likewise.
7257 (vst3q_lane_u64): Likewise.
7258 (vst3q_lane_bf16): Likewise.
7260 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
7262 * config/aarch64/arm_neon.h (__ST4_LANE_FUNC): Delete.
7263 (__ST4Q_LANE_FUNC): Delete.
7264 (vst4_lane_f16): Use __builtin_memcpy to copy vector
7265 structure instead of constructing __builtin_aarch64_simd_xi
7266 one vector at a time.
7267 (vst4_lane_f32): Likewise.
7268 (vst4_lane_f64): Likewise.
7269 (vst4_lane_p8): Likewise.
7270 (vst4_lane_p16): Likewise.
7271 (vst4_lane_p64): Likewise.
7272 (vst4_lane_s8): Likewise.
7273 (vst4_lane_s16): Likewise.
7274 (vst4_lane_s32): Likewise.
7275 (vst4_lane_s64): Likewise.
7276 (vst4_lane_u8): Likewise.
7277 (vst4_lane_u16): Likewise.
7278 (vst4_lane_u32): Likewise.
7279 (vst4_lane_u64): Likewise.
7280 (vst4_lane_bf16): Likewise.
7281 (vst4q_lane_f16): Use __builtin_memcpy to copy vector
7282 structure instead of using a union.
7283 (vst4q_lane_f32): Likewise.
7284 (vst4q_lane_f64): Likewise.
7285 (vst4q_lane_p8): Likewise.
7286 (vst4q_lane_p16): Likewise.
7287 (vst4q_lane_p64): Likewise.
7288 (vst4q_lane_s8): Likewise.
7289 (vst4q_lane_s16): Likewise.
7290 (vst4q_lane_s32): Likewise.
7291 (vst4q_lane_s64): Likewise.
7292 (vst4q_lane_u8): Likewise.
7293 (vst4q_lane_u16): Likewise.
7294 (vst4q_lane_u32): Likewise.
7295 (vst4q_lane_u64): Likewise.
7296 (vst4q_lane_bf16): Likewise.
7298 2021-08-06 Martin Liska <mliska@suse.cz>
7300 * config/rs6000/rs6000.c (rs6000_option_override_internal): When
7301 a target option is restored, it can have
7302 rs6000_long_double_type_size set to FLOAT_PRECISION_TFmode
7303 and error should not be emitted.
7305 2021-08-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
7307 * gcov-io.h (gcov_write): Declare.
7308 * gcov-io.c (gcov_write): New.
7309 (gcov_write_counter): Remove.
7310 (gcov_write_tag_length): Likewise.
7311 (gcov_write_summary): Replace gcov_write_tag_length() with calls to
7312 gcov_write_unsigned().
7313 * doc/invoke.texi (fprofile-info-section): Mention
7314 __gcov_info_to_gdca().
7316 2021-08-06 Martin Sebor <msebor@redhat.com>
7318 * dominance.c (prune_bbs_to_update_dominators): Adjust by-value vec
7319 arguments to by-reference.
7320 (iterate_fix_dominators): Same.
7321 * dominance.h (iterate_fix_dominators): Same.
7322 * ipa-prop.h: Call auto_vec::to_vec_legacy.
7323 * tree-data-ref.c (dump_data_dependence_relation): Adjust by-value vec
7324 arguments to by-reference.
7325 (debug_data_dependence_relation): Same.
7326 (dump_data_dependence_relations): Same.
7327 * tree-data-ref.h (debug_data_dependence_relation): Same.
7328 (dump_data_dependence_relations): Same.
7329 * tree-predcom.c (dump_chains): Same.
7330 (initialize_root_vars_lm): Same.
7331 (determine_unroll_factor): Same.
7332 (replace_phis_by_defined_names): Same.
7333 (insert_init_seqs): Same.
7334 (pcom_worker::tree_predictive_commoning_loop): Call
7335 auto_vec::to_vec_legacy.
7336 * tree-ssa-pre.c (insert_into_preds_of_block): Adjust by-value vec
7337 arguments to by-reference.
7338 * tree-ssa-threadbackward.c (populate_worklist): Same.
7339 (back_threader::resolve_def): Same.
7340 * tree-vect-data-refs.c (vect_check_nonzero_value): Same.
7341 (vect_enhance_data_refs_alignment): Same.
7342 (vect_check_lower_bound): Same.
7343 (vect_prune_runtime_alias_test_list): Same.
7344 (vect_permute_store_chain): Same.
7345 * tree-vect-slp-patterns.c (vect_normalize_conj_loc): Same.
7346 * tree-vect-stmts.c (vect_create_vectorized_demotion_stmts): Same.
7347 * tree-vectorizer.h (vect_permute_store_chain): Same.
7348 * vec.c (test_init): New function.
7349 (vec_c_tests): Call new function.
7350 * vec.h (vec): Declare ctors, dtor, and assignment.
7351 (auto_vec::vec_to_legacy): New function.
7352 (vec::copy): Adjust initialization.
7354 2021-08-05 H.J. Lu <hjl.tools@gmail.com>
7357 * config/i386/i386.c (ix86_can_inline_p): Ignore MASK_80387 if
7358 callee only uses GPRs.
7359 * config/i386/ia32intrin.h: Revert commit 5463cee2770.
7360 * config/i386/serializeintrin.h: Revert commit 71958f740f1.
7361 * config/i386/x86gprintrin.h: Add
7362 #pragma GCC target("general-regs-only") and #pragma GCC pop_options
7363 to disable non-GPR ISAs.
7365 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
7367 PR middle-end/101787
7368 * doc/md.texi (cond_ashl, cond_ashr, cond_lshr): Document.
7370 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
7372 * tree-vectorizer.h (vect_is_store_elt_extraction, vect_is_reduction)
7373 (vect_reduc_type, vect_embedded_comparison_type, vect_comparison_type)
7374 (vect_is_extending_load, vect_is_integer_truncation): New functions,
7375 moved from aarch64.c but given different names.
7376 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction)
7377 (aarch64_is_reduction, aarch64_reduc_type)
7378 (aarch64_embedded_comparison_type, aarch64_comparison_type)
7379 (aarch64_extending_load_p, aarch64_integer_truncation_p): Delete
7380 in favor of the above. Update callers accordingly.
7382 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
7385 * config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
7386 writing .cpu directive in asm output.
7387 * config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
7388 (arm_last_printed_arch_string): Delete.
7389 (arm_last-printed_fpu_string): Delete.
7390 (arm_configure_build_target): If use of floating-point/SIMD is
7391 disabled, remove all fp/simd related features from the target ISA.
7392 (last_arm_targ_options): New variable.
7393 (arm_print_asm_arch_directives): Add new parameters. Change order
7394 of emitted directives and handle all cases here.
7395 (arm_file_start): Always call arm_print_asm_arch_directives, move
7396 all generation of .arch/.arch_extension here.
7397 (arm_file_end): Call arm_print_asm_arch.
7398 (arm_declare_function_name): Call arm_print_asm_arch_directives
7399 instead of printing .arch/.fpu directives directly.
7401 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
7403 * config/arm/arm.c (arm_configure_build_target): Don't call
7404 arm_option_reconfigure_globals.
7405 (arm_option_restore): Call arm_option_reconfigure_globals after
7406 reconfiguring the target.
7407 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
7409 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
7411 * config/arm/arm.c (arm_configure_build_target): Ensure the target's
7412 arch_name is always set.
7414 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
7416 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
7417 of vec_select high-half from being added into Neon subtract
7420 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
7422 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
7423 of vec_select high-half from being added into Neon add cost.
7425 2021-08-05 Kewen Lin <linkw@linux.ibm.com>
7427 * cfgloop.h (loops_list::loops_list): Add one optional argument
7428 root and adjust accordingly, update loop tree walking and factor
7430 * cfgloop.c (loops_list::walk_loop_tree): ... this. New function.
7432 2021-08-05 Eric Botcazou <ebotcazou@gcc.gnu.org>
7434 PR tree-optimization/101626
7435 * tree-sra.c (propagate_subaccesses_from_rhs): Do not set the
7436 reverse scalar storage order on a pointer or vector component.
7438 2021-08-05 liuhongt <hongtao.liu@intel.com>
7440 * config/i386/sse.md (cond_<code><mode>): New expander.
7442 2021-08-05 liuhongt <hongtao.liu@intel.com>
7444 * config/i386/sse.md (cond_<code><mode>): New expander.
7446 2021-08-05 liuhongt <hongtao.liu@intel.com>
7448 * config/i386/sse.md (cond_<code><mode>): New expander.
7450 2021-08-04 David Malcolm <dmalcolm@redhat.com>
7453 * Makefile.in (ANALYZER_OBJS): Add analyzer/region-model-asm.o.
7455 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
7458 * config/i386/i386.h (STORE_MAX_PIECES): Allow 16/32/64 bytes
7459 only if TARGET_INTER_UNIT_MOVES_TO_VEC is true.
7461 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
7464 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
7465 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
7466 data with SSE register from one memory location to another.
7468 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7470 * config/s390/s390.c (expand_perm_with_vpdi): New function.
7471 (vectorize_vec_perm_const_1): Call expand_perm_with_vpdi.
7472 * config/s390/vector.md (*vpdi1<mode>, @vpdi1<mode>): Enable a
7473 parameterized expander.
7474 (*vpdi4<mode>, @vpdi4<mode>): Likewise.
7476 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7478 * config/s390/s390.c (MAX_VECT_LEN): Define macro.
7479 (struct expand_vec_perm_d): Define struct.
7480 (expand_perm_with_merge): New function.
7481 (vectorize_vec_perm_const_1): New function.
7482 (s390_vectorize_vec_perm_const): New function.
7483 (TARGET_VECTORIZE_VEC_PERM_CONST): Define target macro.
7485 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7487 * config/s390/vector.md (V_HW_64): Remove mode iterator.
7488 (*vec_load_pair<mode>): Use V_HW_2 instead of V_HW_64.
7489 * config/s390/vx-builtins.md
7490 (vec_scatter_element<V_HW_2:mode>_SI): Use V_HW_2 instead of
7493 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7495 * config/s390/s390.md (UNSPEC_VEC_PERMI): Remove constant
7497 * config/s390/vector.md (*vpdi1<mode>, *vpdi4<mode>): New pattern
7499 * config/s390/vx-builtins.md (*vec_permi<mode>): Emit generic rtx
7500 instead of an unspec.
7502 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7504 * config/s390/s390-modes.def: Add more vector modes to support
7505 concatenation of two vectors.
7506 * config/s390/s390-protos.h (s390_expand_merge_perm_const): Add
7508 (s390_expand_merge): Likewise.
7509 * config/s390/s390.c (s390_expand_merge_perm_const): New function.
7510 (s390_expand_merge): New function.
7511 * config/s390/s390.md (UNSPEC_VEC_MERGEH, UNSPEC_VEC_MERGEL):
7512 Remove constant definitions.
7513 * config/s390/vector.md (V_HW_2): Add mode iterators.
7514 (VI_HW_4, V_HW_4): Rename VI_HW_4 to V_HW_4.
7515 (vec_2x_nelts, vec_2x_wide): New mode attributes.
7516 (*vmrhb, *vmrlb, *vmrhh, *vmrlh, *vmrhf, *vmrlf, *vmrhg, *vmrlg):
7517 New pattern definitions.
7518 (vec_widen_umult_lo_<mode>, vec_widen_umult_hi_<mode>)
7519 (vec_widen_smult_lo_<mode>, vec_widen_smult_hi_<mode>)
7520 (vec_unpacks_lo_v4sf, vec_unpacks_hi_v4sf, vec_unpacks_lo_v2df)
7521 (vec_unpacks_hi_v2df): Adjust expanders to emit non-unspec RTX for
7523 * config/s390/vx-builtins.md (V_HW_4): Remove mode iterator. Now
7525 (vec_mergeh<mode>, vec_mergel<mode>): Use s390_expand_merge to
7526 emit vec merge pattern.
7528 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
7530 * config/aarch64/aarch64.c (aarch64_strip_extend_vec_half):
7532 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent cost of
7533 vec_select high-half from being added into Neon multiply
7535 * rtlanal.c (vec_series_highpart_p): Define.
7536 * rtlanal.h (vec_series_highpart_p): Declare.
7538 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
7540 * config/aarch64/aarch64.c (aarch64_strip_duplicate_vec_elt):
7542 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent
7543 vec_select cost from being added into Neon multiply cost.
7545 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
7547 * tree-vect-loop.c (vect_better_loop_vinfo_p): Detect cases in
7548 which old_loop_vinfo is an epilogue loop that handles a constant
7549 number of iterations.
7551 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
7553 * tree-vect-loop.c (vect_analyze_loop): Print a dump message
7554 when a reanalyzed loop fails to be cheaper than the current
7557 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
7559 * config/aarch64/aarch64.c: Fix a typo.
7561 2021-08-04 Vincent Lefèvre <vincent-gcc@vinc17.net>
7563 PR gcov-profile/101773
7564 * gcov-io.c (gcov_close): Check return code of a fclose.
7566 2021-08-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
7569 * dwarf2out.c (dwarf2out_assembly_start): Emit a dummy
7570 .file statement when needed.
7572 2021-08-04 Richard Biener <rguenther@suse.de>
7574 * tree-vect-data-refs.c (vect_check_gather_scatter):
7575 Include widening conversions only when the result is
7576 still handed by native gather or the current offset
7577 size not already matches the data size.
7578 Also succeed analysis in case there's no native support,
7579 noted by a IFN_LAST ifn and a NULL decl.
7580 (vect_analyze_data_refs): Always consider gathers.
7581 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
7582 Test for no IFN gather rather than decl gather.
7583 * tree-vect-stmts.c (vect_model_load_cost): Pass in the
7584 gather-scatter info and cost emulated gathers accordingly.
7585 (vect_truncate_gather_scatter_offset): Properly test for
7587 (vect_use_strided_gather_scatters_p): Likewise.
7588 (get_load_store_type): Handle emulated gathers and its
7590 (vectorizable_load): Likewise. Emulate them by extracting
7591 scalar offsets, doing scalar loads and a vector construct.
7593 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
7596 * expr.c (op_by_pieces_d::op_by_pieces_d): Add a max_pieces
7597 argument to set m_max_size.
7598 (move_by_pieces_d): Pass MOVE_MAX_PIECES to op_by_pieces_d.
7599 (store_by_pieces_d): Pass STORE_MAX_PIECES to op_by_pieces_d.
7600 (compare_by_pieces_d): Pass COMPARE_MAX_PIECES to op_by_pieces_d.
7602 2021-08-04 Roger Sayle <roger@nextmovesoftware.com>
7603 Marc Glisse <marc.glisse@inria.fr>
7605 * match.pd (bit_ior, bit_xor): Canonicalize (X*C1)|(X*C2) and
7606 (X*C1)^(X*C2) as X*(C1+C2), and related variants, using
7607 tree_nonzero_bits to ensure that operands are bit-wise disjoint.
7609 2021-08-04 Richard Biener <rguenther@suse.de>
7611 * tree-ssa-forwprop.c (pass_forwprop::execute): Split
7612 out code to decompose vector loads ...
7613 (optimize_vector_load): ... here. Generalize it to
7614 handle intermediate widening and TARGET_MEM_REF loads
7615 and apply it to loads with a supported vector mode as well.
7617 2021-08-04 Richard Biener <rguenther@suse.de>
7619 PR tree-optimization/101756
7620 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Make sure
7621 the result of the reduction epilogue is compatible to the original
7624 2021-08-04 liuhongt <hongtao.liu@intel.com>
7627 * config/i386/i386.md (peephole2): Refine predicate from
7628 register_operand to general_reg_operand.
7630 2021-08-04 Aldy Hernandez <aldyh@redhat.com>
7632 * gimple-range-path.h (path_range_query::dump): Mark override.
7634 2021-08-04 Richard Biener <rguenther@suse.de>
7636 PR tree-optimization/101769
7637 * tree-tailcall.c (eliminate_tail_call): Add the created loop
7638 for the first recursion and return it via the new output parameter.
7639 (optimize_tail_call): Pass through new output param.
7640 (tree_optimize_tail_calls_1): After creating all latches,
7641 add the created loop to the loop tree. Do not mark loops for fixup.
7643 2021-08-04 Martin Liska <mliska@suse.cz>
7645 * doc/invoke.texi: Document threader-mode param.
7647 2021-08-04 liuhongt <hongtao.liu@intel.com>
7649 * config/i386/sse.md (cond_fma<mode>): New expander.
7650 (cond_fms<mode>): Ditto.
7651 (cond_fnma<mode>): Ditto.
7652 (cond_fnms<mode>): Ditto.
7654 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
7656 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode>): Use && instead of &.
7658 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
7660 * config/rs6000/constraints.md: Remove "e" from the list of available
7661 constraint characters.
7663 2021-08-03 Eugene Rozenfeld <erozen@microsoft.com>
7665 PR gcov-profile/71672
7666 * auto-profile.c (afdo_indirect_call): Fix setup of the historgram value for indirect calls.
7668 2021-08-03 Paul A. Clarke <pc@us.ibm.com>
7670 * config/rs6000/smmintrin.h (_mm_minpos_epu16): New.
7672 2021-08-03 H.J. Lu <hjl.tools@gmail.com>
7674 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): In 64-bit mode,
7675 try XMM31 to avoid vzeroupper.
7677 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7679 * doc/invoke.texi: Document -mtune=neoverse-512tvb and
7680 -mcpu=neoverse-512tvb.
7681 * config/aarch64/aarch64-cores.def (neoverse-512tvb): New entry.
7682 * config/aarch64/aarch64-tune.md: Regenerate.
7683 * config/aarch64/aarch64.c (neoverse512tvb_sve_vector_cost)
7684 (neoverse512tvb_sve_issue_info, neoverse512tvb_vec_issue_info)
7685 (neoverse512tvb_vector_cost, neoverse512tvb_tunings): New structures.
7686 (aarch64_adjust_body_cost_sve): Handle -mtune=neoverse-512tvb.
7687 (aarch64_adjust_body_cost): Likewise.
7689 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7691 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Only
7692 record issue information for operations that occur in the
7695 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7697 * config/aarch64/aarch64.c (aarch64_multiply_add_p): Add a vec_flags
7698 parameter. Detect cases in which an Advanced SIMD MLA would almost
7699 certainly require a MOV.
7700 (aarch64_count_ops): Update accordingly.
7702 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7704 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction): New
7705 function, split out from...
7706 (aarch64_detect_vector_stmt_subtype): ...here.
7707 (aarch64_add_stmt_cost): Treat extracting element 0 as free.
7709 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7711 * config/aarch64/aarch64-protos.h (sve_vec_cost):
7712 Add gather_load_x32_cost and gather_load_x64_cost.
7713 * config/aarch64/aarch64.c (generic_sve_vector_cost)
7714 (a64fx_sve_vector_cost, neoversev1_sve_vector_cost): Update
7715 accordingly, using the values given by the scalar_load * number
7716 of elements calculation that we used previously.
7717 (aarch64_detect_vector_stmt_subtype): Use the new fields.
7719 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7721 * config/aarch64/aarch64.c (aarch64_adjust_body_cost_sve): New
7722 function, split out from...
7723 (aarch64_adjust_body_cost): ...here.
7725 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7727 * config/aarch64/fractional-cost.h: New file.
7728 * config/aarch64/aarch64.c: Include <algorithm> (indirectly)
7729 and cost_fraction.h.
7730 (vec_cost_fraction): New typedef.
7731 (aarch64_detect_scalar_stmt_subtype): Use it for statement costs.
7732 (aarch64_detect_vector_stmt_subtype): Likewise.
7733 (aarch64_sve_adjust_stmt_cost, aarch64_adjust_stmt_cost): Likewise.
7734 (aarch64_estimate_min_cycles_per_iter): Use vec_cost_fraction
7736 (aarch64_adjust_body_cost): Likewise.
7737 (aarch64_test_cost_fraction): New function.
7738 (aarch64_run_selftests): Call it.
7740 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7742 * config/aarch64/aarch64-protos.h (tune_params::sve_width): Turn
7744 * config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Update
7746 (aarch64_estimated_poly_value): Likewise. Use the least significant
7747 set bit for the minimum and likely values. Use the most significant
7748 set bit for the maximum value.
7750 2021-08-03 liuhongt <hongtao.liu@intel.com>
7752 * config/i386/sse.md (cond_<insn><mode>): New expander.
7753 (cond_mul<mode>): Ditto.
7755 2021-08-03 Kewen Lin <linkw@linux.ibm.com>
7757 * tree-cfg.c (move_sese_region_to_fn): Fix typos on dloop.
7759 2021-08-03 liuhongt <hongtao.liu@intel.com>
7761 * config/i386/sse.md (cond_<insn><mode>):New expander.
7762 (cond_mul<mode>): Ditto.
7763 (cond_div<mode>): Ditto.
7765 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
7767 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Also
7768 check stack_realign_needed for stack realignment.
7769 (ix86_legitimate_constant_p): Always allow CONST_WIDE_INT smaller
7770 than the largest integer supported by vector register.
7771 * config/i386/i386.h (MAX_MOVE_MAX): New. Set to 64.
7772 (MOVE_MAX): Set to bytes of the largest integer supported by
7774 (STORE_MAX_PIECES): New.
7776 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
7778 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
7779 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
7780 data from one memory location to another.
7782 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
7785 * config/i386/i386.c (TARGET_GEN_MEMSET_SCRATCH_RTX): New.
7787 2021-08-02 Aldy Hernandez <aldyh@redhat.com>
7789 PR tree-optimization/101724
7790 * params.opt: Remove --param=threader-iterative.
7791 * tree-ssa-threadbackward.c (pass_thread_jumps::execute): Remove
7794 2021-08-02 Tom de Vries <tdevries@suse.de>
7796 PR middle-end/101665
7797 * doc/extend.texi (nonnull attribute): Improve documentation.
7799 2021-08-02 Andrew Pinski <apinski@marvell.com>
7801 PR rtl-optimization/101683
7802 * rtlanal.c (may_trap_p_1): Handle UNSIGNED_FIX.
7804 2021-08-02 Roger Sayle <roger@nextmovesoftware.com>
7806 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern):
7807 Renamed from cond_removal_in_popcount_clz_ctz_pattern.
7808 Add support for BSWAP, FFS, PARITY and CLRSB builtins.
7809 (tree_ssa_phiop_worker): Update call to function above.
7811 2021-08-01 H.J. Lu <hjl.tools@gmail.com>
7814 * config/i386/i386.md (bsr_rex64_1_zext): New.
7815 (combine splitter for constant - clzll): Replace gen_bsr_rex64_1
7816 with gen_bsr_rex64_1_zext.
7818 2021-07-31 Jakub Jelinek <jakub@redhat.com>
7821 * config/i386/i386.md (bsr_rex64_1, bsr_1, bsr_zext_1): New
7822 define_insn patterns.
7823 (*bsr_rex64_2, *bsr_2): New define_insn_and_split patterns.
7824 Add combine splitters for constant - clz.
7825 (clz<mode>2): Use a temporary pseudo for bsr result.
7827 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
7829 * config/rs6000/smmintrin.h (_mm_floor_pd, _mm_floor_ps,
7830 _mm_floor_sd, _mm_floor_ss): New.
7832 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
7834 * config/rs6000/smmintrin.h (_mm_ceil_pd, _mm_ceil_ps,
7835 _mm_ceil_sd, _mm_ceil_ss): New.
7837 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
7839 * config/rs6000/smmintrin.h (_mm_blend_pd, _mm_blendv_pd,
7840 _mm_blend_ps, _mm_blendv_ps): New.
7842 2021-07-30 Roger Sayle <roger@nextmovesoftware.com>
7843 Uroš Bizjak <ubizjak@gmail.com>
7845 * config/i386/i386.md (*dec_cmov<mode>): New define_insn_and_split
7846 to generate a conditional move using the carry flag after sub $1.
7847 (peephole2): Eliminate a register-to-register move by inverting
7848 the condition of a conditional move.
7850 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
7852 * config/mmix/mmix.md ("call", "call_value", "*call_real")
7853 ("*call_value_real"): Don't generate rtx mentioning the generic
7854 operands 1 and 2 to "call", and similarly for "call_value".
7855 * config/mmix/mmix.c (mmix_print_operand_punct_valid_p)
7856 (mmix_print_operand): Use '!' instead of 'p'.
7858 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
7860 * doc/md.texi (call): Correct information about operand 2.
7861 * config/mmix/mmix.md ("call", "call_value"): Remove fixed FIXMEs.
7863 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
7865 * range-op.cc (operator_trunc_mod::wi_fold): Fold constants.
7867 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
7869 * range-op.cc (operator_div::wi_fold): Return UNDEFINED for [0, 0] divisor.
7871 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
7873 * gimple-range-cache.cc (*::set_bb_range): Change const basic_block to
7875 (*::get_bb_range): Ditto.
7876 (*::bb_range_p): Ditto.
7877 * gimple-range-cache.h: Change prototypes.
7879 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
7882 * builtins.c (builtin_memcpy_read_str): Change the mode argument
7883 from scalar_int_mode to fixed_size_mode.
7884 (builtin_strncpy_read_str): Likewise.
7885 (gen_memset_value_from_prev): New function.
7886 (builtin_memset_read_str): Change the mode argument from
7887 scalar_int_mode to fixed_size_mode. Use gen_memset_value_from_prev
7888 and support CONST_VECTOR.
7889 (builtin_memset_gen_str): Likewise.
7890 (try_store_by_multiple_pieces): Use by_pieces_constfn to declare
7892 * builtins.h (builtin_strncpy_read_str): Replace scalar_int_mode
7893 with fixed_size_mode.
7894 (builtin_memset_read_str): Likewise.
7895 * expr.c (widest_int_mode_for_size): Renamed to ...
7896 (widest_fixed_size_mode_for_size): Add a bool argument to
7897 indicate if QI vector mode can be used.
7898 (by_pieces_ninsns): Call widest_fixed_size_mode_for_size
7899 instead of widest_int_mode_for_size.
7900 (pieces_addr::adjust): Change the mode argument from
7901 scalar_int_mode to fixed_size_mode.
7902 (op_by_pieces_d): Make m_len read-only. Add a bool member,
7903 m_qi_vector_mode, to indicate that QI vector mode can be used.
7904 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
7905 initialize m_qi_vector_mode. Call widest_fixed_size_mode_for_size
7906 instead of widest_int_mode_for_size.
7907 (op_by_pieces_d::get_usable_mode): Change the mode argument from
7908 scalar_int_mode to fixed_size_mode. Call
7909 widest_fixed_size_mode_for_size instead of
7910 widest_int_mode_for_size.
7911 (op_by_pieces_d::smallest_fixed_size_mode_for_size): New member
7912 function to return the smallest integer or QI vector mode.
7913 (op_by_pieces_d::run): Call widest_fixed_size_mode_for_size
7914 instead of widest_int_mode_for_size. Call
7915 smallest_fixed_size_mode_for_size instead of
7916 smallest_int_mode_for_size.
7917 (store_by_pieces_d::store_by_pieces_d): Add a bool argument to
7918 indicate that QI vector mode can be used and pass it to
7919 op_by_pieces_d::op_by_pieces_d.
7920 (can_store_by_pieces): Call widest_fixed_size_mode_for_size
7921 instead of widest_int_mode_for_size. Pass memsetp to
7922 widest_fixed_size_mode_for_size to support QI vector mode.
7923 Allow all CONST_VECTORs for memset if vec_duplicate is supported.
7924 (store_by_pieces): Pass memsetp to
7925 store_by_pieces_d::store_by_pieces_d.
7926 (clear_by_pieces_1): Removed.
7927 (clear_by_pieces): Replace clear_by_pieces_1 with
7928 builtin_memset_read_str and pass true to store_by_pieces_d to
7929 support vector mode broadcast.
7930 (string_cst_read_str): Change the mode argument from
7931 scalar_int_mode to fixed_size_mode.
7932 * expr.h (by_pieces_constfn): Change scalar_int_mode to
7934 (by_pieces_prev): Likewise.
7935 * rtl.h (lowpart_subreg_regno): New.
7936 * rtlanal.c (lowpart_subreg_regno): New. A wrapper around
7937 simplify_subreg_regno.
7938 * target.def (gen_memset_scratch_rtx): New hook.
7939 * doc/tm.texi.in: Add TARGET_GEN_MEMSET_SCRATCH_RTX.
7940 * doc/tm.texi: Regenerated.
7942 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
7945 * config/mips/mips.c (mips_atomic_assign_expand_fenv): Use
7946 TARGET_EXPR instead of MODIFY_EXPR.
7948 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
7951 * config/mips/mips-protos.h (mips_expand_vec_cmp_expr): Declare.
7952 * config/mips/mips.c (mips_expand_vec_cmp_expr): New function.
7953 * config/mips/mips-msa.md (vec_cmp<MSA:mode><mode_i>): New
7955 (vec_cmpu<IMSA:mode><mode_i>): New expander.
7957 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
7960 * config/i386/i386-options.c (ix86_option_override_internal):
7961 Don't enable LZCNT/POPCNT if they have been disabled explicitly.
7963 2021-07-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
7966 * config/arm/arm_neon.h (vld1_p64): Replace call to builtin by
7967 explicitly dereferencing __a.
7968 (vld1_s64): Likewise.
7969 (vld1_u64): Likewise.
7970 * config/arm/arm_neon_builtins.def (vld1): Remove entry for di
7971 and change to VAR13.
7973 2021-07-30 Aldy Hernandez <aldyh@redhat.com>
7975 * gimple-loop-versioning.cc (lv_dom_walker::lv_dom_walker): Remove
7976 use of m_range_analyzer.
7977 (loop_versioning::lv_dom_walker::before_dom_children): Same.
7978 (loop_versioning::lv_dom_walker::after_dom_children): Remove.
7979 (loop_versioning::prune_loop_conditions): Replace vr_values use
7980 with range_query interface.
7981 (pass_loop_versioning::execute): Use ranger.
7983 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
7986 * ipa-devirt.c (ipa_odr_read_section): Compare the precision of
7987 enum values, and emit a warning if they mismatch.
7989 2021-07-30 Kewen Lin <linkw@linux.ibm.com>
7991 * cfgloop.h (as_const): New function.
7992 (class loop_iterator): Rename to ...
7993 (class loops_list): ... this.
7994 (loop_iterator::next): Rename to ...
7995 (loops_list::Iter::fill_curr_loop): ... this and adjust.
7996 (loop_iterator::loop_iterator): Rename to ...
7997 (loops_list::loops_list): ... this and adjust.
7998 (loops_list::Iter): New class.
7999 (loops_list::iterator): New type.
8000 (loops_list::const_iterator): New type.
8001 (loops_list::begin): New function.
8002 (loops_list::end): Likewise.
8003 (loops_list::begin const): Likewise.
8004 (loops_list::end const): Likewise.
8005 (FOR_EACH_LOOP): Remove.
8006 (FOR_EACH_LOOP_FN): Remove.
8007 * cfgloop.c (flow_loops_dump): Adjust FOR_EACH_LOOP* with range-based
8008 for loop with loops_list instance.
8009 (sort_sibling_loops): Likewise.
8010 (disambiguate_loops_with_multiple_latches): Likewise.
8011 (verify_loop_structure): Likewise.
8012 * cfgloopmanip.c (create_preheaders): Likewise.
8013 (force_single_succ_latches): Likewise.
8014 * config/aarch64/falkor-tag-collision-avoidance.c
8015 (execute_tag_collision_avoidance): Likewise.
8016 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Likewise.
8017 * config/s390/s390.c (s390_adjust_loops): Likewise.
8018 * doc/loop.texi: Likewise.
8019 * gimple-loop-interchange.cc (pass_linterchange::execute): Likewise.
8020 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
8021 * gimple-loop-versioning.cc (loop_versioning::analyze_blocks): Likewise.
8022 (loop_versioning::make_versioning_decisions): Likewise.
8023 * gimple-ssa-split-paths.c (split_paths): Likewise.
8024 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
8025 * graphite.c (canonicalize_loop_form): Likewise.
8026 (graphite_transform_loops): Likewise.
8027 * ipa-fnsummary.c (analyze_function_body): Likewise.
8028 * ipa-pure-const.c (analyze_function): Likewise.
8029 * loop-doloop.c (doloop_optimize_loops): Likewise.
8030 * loop-init.c (loop_optimizer_finalize): Likewise.
8031 (fix_loop_structure): Likewise.
8032 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
8033 (move_loop_invariants): Likewise.
8034 * loop-unroll.c (decide_unrolling): Likewise.
8035 (unroll_loops): Likewise.
8036 * modulo-sched.c (sms_schedule): Likewise.
8037 * predict.c (predict_loops): Likewise.
8038 (pass_profile::execute): Likewise.
8039 * profile.c (branch_prob): Likewise.
8040 * sel-sched-ir.c (sel_finish_pipelining): Likewise.
8041 (sel_find_rgns): Likewise.
8042 * tree-cfg.c (replace_loop_annotate): Likewise.
8043 (replace_uses_by): Likewise.
8044 (move_sese_region_to_fn): Likewise.
8045 * tree-if-conv.c (pass_if_conversion::execute): Likewise.
8046 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
8047 * tree-parloops.c (parallelize_loops): Likewise.
8048 * tree-predcom.c (tree_predictive_commoning): Likewise.
8049 * tree-scalar-evolution.c (scev_initialize): Likewise.
8050 (scev_reset): Likewise.
8051 * tree-ssa-dce.c (find_obviously_necessary_stmts): Likewise.
8052 * tree-ssa-live.c (remove_unused_locals): Likewise.
8053 * tree-ssa-loop-ch.c (ch_base::copy_headers): Likewise.
8054 * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
8055 (tree_ssa_lim_initialize): Likewise.
8056 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Likewise.
8057 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize): Likewise.
8058 * tree-ssa-loop-manip.c (get_loops_exits): Likewise.
8059 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations): Likewise.
8060 (free_numbers_of_iterations_estimates): Likewise.
8061 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Likewise.
8062 * tree-ssa-loop-split.c (tree_ssa_split_loops): Likewise.
8063 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Likewise.
8064 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
8065 (pass_scev_cprop::execute): Likewise.
8066 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
8067 * tree-ssa-sccvn.c (do_rpo_vn): Likewise.
8068 * tree-ssa-threadupdate.c
8069 (jump_thread_path_registry::thread_through_all_blocks): Likewise.
8070 * tree-vectorizer.c (vectorize_loops): Likewise.
8071 * tree-vrp.c (vrp_asserts::find_assert_locations): Likewise.
8073 2021-07-29 Hans-Peter Nilsson <hp@bitrange.com>
8075 * config/mmix/mmix.c (mmix_function_arg_1): Avoid
8076 generating a VOIDmode register for e.g the
8077 function_arg_info::end_marker.
8079 2021-07-29 Jeff Law <jeffreyalaw@gmail.com>
8081 * config/h8300/h8300-modes.def: Add CCZ, CCV and CCC, drop CCZNV.
8082 * config/h8300/h8300.md (H8cc mode iterator): Add CCZ.
8083 (cc mode_attr): Similarly.
8084 (ccz subst_attr): Similarly.
8085 * config/h8300/jumpcall.md: Add new patterns for branch-on-bit.
8086 * config/h8300/testcompare.md: Remove various cc0 based patterns
8087 that had been commented out. Add pattern to set CCZ from a bit
8090 2021-07-29 Thomas Schwinge <thomas@codesourcery.com>
8091 Julian Brown <julian@codesourcery.com>
8092 Kwok Cheung Yeung <kcy@codesourcery.com>
8094 * omp-offload.c (oacc_loop_xform_head_tail, oacc_loop_process):
8095 'update_stmt' after modification.
8096 (pass_oacc_loop_designation): New function, extracted out of...
8097 (pass_oacc_device_lower): ... this.
8098 (pass_data_oacc_loop_designation, pass_oacc_loop_designation)
8099 (make_pass_oacc_loop_designation): New
8100 * passes.def: Add it.
8101 * tree-parloops.c (create_parallel_loop): Adjust.
8102 * tree-pass.h (make_pass_oacc_loop_designation): New.
8104 2021-07-29 Aldy Hernandez <aldyh@redhat.com>
8106 * flag-types.h (enum threader_mode): New.
8107 * params.opt: Add entry for --param=threader-mode.
8108 * tree-ssa-threadbackward.c (THREADER_ITERATIVE_MODE): New.
8109 (class back_threader): New.
8110 (back_threader::back_threader): New.
8111 (back_threader::~back_threader): New.
8112 (back_threader::maybe_register_path): New.
8113 (back_threader::find_taken_edge): New.
8114 (back_threader::find_taken_edge_switch): New.
8115 (back_threader::find_taken_edge_cond): New.
8116 (back_threader::resolve_def): New.
8117 (back_threader::resolve_phi): New.
8118 (back_threader::find_paths_to_names): New.
8119 (back_threader::find_paths): New.
8122 (thread_jumps::find_jump_threads_backwards): Call ranger threader.
8123 (thread_jumps::find_jump_threads_backwards_with_ranger): New.
8124 (pass_thread_jumps::execute): Abstract out code...
8125 (try_thread_blocks): ...here.
8126 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
8127 Abstract out threading candidate code to...
8128 (single_succ_to_potentially_threadable_block): ...here.
8129 * tree-ssa-threadedge.h (single_succ_to_potentially_threadable_block):
8131 * tree-ssa-threadupdate.c (register_jump_thread): Return boolean.
8132 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
8133 Return bool from register_jump_thread.
8135 2021-07-29 Andreas Krebbel <krebbel@linux.ibm.com>
8137 * target.def: in0 and in1 do not need to be registers.
8138 * doc/tm.texi: Regenerate.
8140 2021-07-29 liuhongt <hongtao.liu@intel.com>
8143 * config/i386/i386.c (ix86_widen_mult_cost): New function.
8144 (ix86_add_stmt_cost): Use ix86_widen_mult_cost for
8147 2021-07-29 Jiufu Guo <guojiufu@linux.ibm.com>
8150 * config/rs6000/rs6000.c (TARGET_PREFERRED_DOLOOP_MODE): New hook.
8151 (rs6000_preferred_doloop_mode): New hook.
8152 * doc/tm.texi: Regenerate.
8153 * doc/tm.texi.in: Add hook preferred_doloop_mode.
8154 * target.def (preferred_doloop_mode): New hook.
8155 * targhooks.c (default_preferred_doloop_mode): New hook.
8156 * targhooks.h (default_preferred_doloop_mode): New hook.
8157 * tree-ssa-loop-ivopts.c (compute_doloop_base_on_mode): New function.
8158 (add_iv_candidate_for_doloop): Call targetm.preferred_doloop_mode
8159 and compute_doloop_base_on_mode.
8161 2021-07-28 Martin Sebor <msebor@redhat.com>
8163 PR middle-end/101494
8164 * tree-ssa-uninit.c (maybe_warn_operand): Correct object offset
8165 and size computation.
8167 2021-07-28 Martin Sebor <msebor@redhat.com>
8169 PR middle-end/101601
8170 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Remove
8172 Handle pointers to functions.
8174 2021-07-28 Martin Sebor <msebor@redhat.com>
8176 * Makefile.in (OBJS): Add gimple-ssa-warn-access.o and pointer-query.o.
8177 * attribs.h (fndecl_dealloc_argno): Move fndecl_dealloc_argno to tree.h.
8178 * builtins.c (compute_objsize_r): Move to pointer-query.cc.
8179 (access_ref::access_ref): Same.
8180 (access_ref::phi): Same.
8181 (access_ref::get_ref): Same.
8182 (access_ref::size_remaining): Same.
8183 (access_ref::offset_in_range): Same.
8184 (access_ref::add_offset): Same.
8185 (access_ref::inform_access): Same.
8186 (ssa_name_limit_t::visit_phi): Same.
8187 (ssa_name_limit_t::leave_phi): Same.
8188 (ssa_name_limit_t::next): Same.
8189 (ssa_name_limit_t::next_phi): Same.
8190 (ssa_name_limit_t::~ssa_name_limit_t): Same.
8191 (pointer_query::pointer_query): Same.
8192 (pointer_query::get_ref): Same.
8193 (pointer_query::put_ref): Same.
8194 (pointer_query::flush_cache): Same.
8195 (warn_string_no_nul): Move to gimple-ssa-warn-access.cc.
8196 (check_nul_terminated_array): Same.
8197 (unterminated_array): Same.
8198 (maybe_warn_for_bound): Same.
8199 (check_read_access): Same.
8200 (warn_for_access): Same.
8201 (get_size_range): Same.
8202 (check_access): Same.
8203 (gimple_call_alloc_size): Move to tree.c.
8204 (gimple_parm_array_size): Move to pointer-query.cc.
8205 (get_offset_range): Same.
8206 (gimple_call_return_array): Same.
8207 (handle_min_max_size): Same.
8208 (handle_array_ref): Same.
8209 (handle_mem_ref): Same.
8210 (compute_objsize): Same.
8211 (gimple_call_alloc_p): Move to gimple-ssa-warn-access.cc.
8212 (call_dealloc_argno): Same.
8213 (fndecl_dealloc_argno): Same.
8214 (new_delete_mismatch_p): Same.
8215 (matching_alloc_calls_p): Same.
8216 (warn_dealloc_offset): Same.
8217 (maybe_emit_free_warning): Same.
8218 * builtins.h (check_nul_terminated_array): Move to
8219 gimple-ssa-warn-access.h.
8220 (check_nul_terminated_array): Same.
8221 (warn_string_no_nul): Same.
8222 (unterminated_array): Same.
8223 (class ssa_name_limit_t): Same.
8224 (class pointer_query): Same.
8225 (struct access_ref): Same.
8226 (class range_query): Same.
8227 (struct access_data): Same.
8228 (gimple_call_alloc_size): Same.
8229 (gimple_parm_array_size): Same.
8230 (compute_objsize): Same.
8231 (class access_data): Same.
8232 (maybe_emit_free_warning): Same.
8233 * calls.c (initialize_argument_information): Remove call to
8234 maybe_emit_free_warning.
8235 * gimple-array-bounds.cc: Include new header..
8236 * gimple-fold.c: Same.
8237 * gimple-ssa-sprintf.c: Same.
8238 * gimple-ssa-warn-restrict.c: Same.
8239 * passes.def: Add pass_warn_access.
8240 * tree-pass.h (make_pass_warn_access): Declare.
8241 * tree-ssa-strlen.c: Include new headers.
8242 * tree.c (fndecl_dealloc_argno): Move here from builtins.c.
8243 * tree.h (fndecl_dealloc_argno): Move here from attribs.h.
8244 * gimple-ssa-warn-access.cc: New file.
8245 * gimple-ssa-warn-access.h: New file.
8246 * pointer-query.cc: New file.
8247 * pointer-query.h: New file.
8249 2021-07-28 Jakub Jelinek <jakub@redhat.com>
8251 PR middle-end/101624
8252 * ubsan.c (maybe_instrument_pointer_overflow,
8253 instrument_object_size): Only test DECL_REGISTER on VAR_DECLs,
8254 PARM_DECLs or RESULT_DECLs.
8255 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
8257 2021-07-28 Jakub Jelinek <jakub@redhat.com>
8259 PR middle-end/101642
8260 * match.pd (bswap16 (x) == bswap16 (y)): Cast both operands
8261 to type of bswap16 for comparison.
8262 (bswap16 (x) == cst): Cast bswap16 operand to type of cst.
8264 2021-07-28 Richard Biener <rguenther@suse.de>
8266 PR tree-optimization/101615
8267 * tree-vect-slp.c (vect_optimize_slp): Materialize permutes
8268 at CTOR SLP graph entries.
8270 2021-07-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8272 * config/aarch64/aarch64.md (*extend<SHORT:mode><GPI:mode>2_aarch64):
8273 Add "r,w" alternative.
8275 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
8278 * config/i386/i386.c (ix86_avx_u128_mode_needed): Don't set
8279 AVX_U128_DIRTY when all bits are zero.
8281 2021-07-28 Richard Biener <rguenther@suse.de>
8283 PR tree-optimization/101615
8284 * tree-vect-slp.c (vect_optimize_slp): Pre-existing vector
8285 external nodes cannot be permuted so make them perm_out 0.
8287 2021-07-28 Andrew Stubbs <ams@codesourcery.com>
8290 * config.in: Regenerate.
8291 * config/gcn/gcn-hsa.h (A_FIJI): New define.
8292 (A_900): New define.
8293 (A_906): New define.
8294 (A_908): New define.
8295 (ASM_SPEC): Use A_FIJI, A_900, A_906 and A_908.
8296 * config/gcn/gcn.c (output_file_start): Adjust attributes according
8297 to the assembler capabilities.
8298 * config/gcn/mkoffload.c (main): Likewise.
8299 * configure: Regenerate.
8300 * configure.ac: Add tests for LLVM assembler attribute features.
8302 2021-07-28 Andrew MacLeod <amacleod@redhat.com>
8304 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Check for
8305 cond_false and cond_true on branches.
8307 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
8309 * config/aarch64/aarch64.c (aarch64_gen_adjusted_ldpstp): use
8312 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
8314 * alias.c (init_alias_analysis): Don't skip prologue/epilogue.
8316 2021-07-28 Jakub Jelinek <jakub@redhat.com>
8319 * config/i386/sse.md (vashr<mode>3): Split into vashrv8di3 expander
8320 and vashrv4di3 expander, where the latter requires just TARGET_AVX2
8321 and has special !TARGET_AVX512VL expansion.
8322 (vashrv2di3<mask_name>): Rename to ...
8323 (vashrv2di3): ... this. Change condition to TARGET_XOP || TARGET_AVX2
8324 and add special !TARGET_XOP && !TARGET_AVX512VL expansion.
8326 2021-07-28 Martin Uecker <muecker@gwdg.de>
8328 * calls.c (maybe_warn_rdwr_sizes): Correct argument
8329 numbers in warning that were switched.
8331 2021-07-28 Kewen Lin <linkw@linux.ibm.com>
8333 PR tree-optimization/101596
8334 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Fix wrong check
8335 by using new_type's precision instead.
8337 2021-07-28 liuhongt <hongtao.liu@intel.com>
8340 * config/i386/i386.h (processor_costs): Add new member
8342 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
8343 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
8344 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
8345 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
8346 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
8347 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
8348 generic_cost, core_cost): Initialize integer_to_sse same value
8350 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
8351 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8352 Use integer_to_sse instead of sse_op to calculate the cost of
8355 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
8357 * config/rs6000/rs6000-gen-builtins.c (write_ovld_static_init): New
8359 (write_init_file): Call write_ovld_static_init.
8361 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
8363 * config/rs6000/rs6000-gen-builtins.c (write_bif_static_init): New
8365 (write_init_file): Call write_bif_static_init.
8367 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
8369 * config/rs6000/rs6000-gen-builtins.c (typemap): New struct.
8370 (TYPE_MAP_SIZE): New macro.
8371 (type_map): New initialized variable.
8372 (typemap_cmp): New function.
8373 (write_type_node): Likewise.
8374 (write_fntype_init): Implement.
8376 2021-07-27 Martin Sebor <msebor@redhat.com>
8378 PR tree-optimization/101584
8379 * tree-ssa-uninit.c (builtin_call_nomodifying_p): New function.
8380 (check_defs): Call it.
8382 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
8384 * tree-ssa-dom.c (dom_jump_threader_simplifier):
8385 Put avail_exprs_stack in the class, instead of passing it to
8386 jump_threader_simplifier.
8387 (dom_jump_threader_simplifier::simplify): Add state argument.
8388 (dom_opt_dom_walker): Add state.
8389 (pass_dominator::execute): Pass state to threader.
8390 (dom_opt_dom_walker::before_dom_children): Use state.
8391 * tree-ssa-threadedge.c (jump_threader::jump_threader): Replace
8393 (jump_threader::record_temporary_equivalences_from_phis):
8394 Register equivalences through the state variable.
8395 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
8396 Record ranges in a statement through the state variable.
8397 (jump_threader::simplify_control_stmt_condition): Pass state to
8399 (jump_threader::simplify_control_stmt_condition_1): Same.
8400 (jump_threader::thread_around_empty_blocks): Remove obsolete
8402 (jump_threader::thread_through_normal_block): Record equivalences
8403 on edge through the state variable.
8404 (jump_threader::thread_across_edge): Abstract state pushing.
8405 (jt_state::jt_state): New.
8406 (jt_state::push): New.
8407 (jt_state::pop): New.
8408 (jt_state::register_equiv): New.
8409 (jt_state::record_ranges_from_stmt): New.
8410 (jt_state::register_equivs_on_edge): New.
8411 (jump_threader_simplifier::jump_threader_simplifier): Move from
8413 (jump_threader_simplifier::simplify): Add state argument.
8414 * tree-ssa-threadedge.h (class jt_state): New.
8415 (class jump_threader): Add state to constructor.
8416 (class jump_threader_simplifier): Add state to simplify. Remove
8417 avail_exprs_stack from class.
8418 * tree-vrp.c (vrp_jump_threader_simplifier::simplify): Add state
8420 (vrp_jump_threader::vrp_jump_threader): Add state.
8421 (vrp_jump_threader::~vrp_jump_threader): Cleanup state.
8423 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
8425 * Makefile.in (OBJS): Add gimple-range-path.o.
8426 * gimple-range-path.cc: New file.
8427 * gimple-range-path.h: New file.
8429 2021-07-27 Jonathan Wright <jonathan.wright@arm.com>
8431 * config/aarch64/aarch64-simd.md: Push sign/zero-extension
8432 inside vec_duplicate for all patterns.
8433 * simplify-rtx.c (simplify_context::simplify_unary_operation_1):
8434 Push sign/zero-extension inside vec_duplicate.
8436 2021-07-27 Richard Biener <rguenther@suse.de>
8438 PR tree-optimization/101573
8439 * tree-ssa-uninit.c (warn_uninit_phi_uses): New function
8440 looking at uninitialized PHI arg defs in some constrained cases.
8441 (warn_uninitialized_vars): Call it.
8442 (execute_early_warn_uninitialized): Calculate dominators.
8444 2021-07-27 Richard Biener <rguenther@suse.de>
8446 PR tree-optimization/39821
8447 * tree-vect-stmts.c (vect_model_promotion_demotion_cost): Use
8448 vector_stmt for widening arithmetic.
8449 (vectorizable_conversion): Adjust.
8451 2021-07-27 Martin Jambor <mjambor@suse.cz>
8453 * cgraph.h (ipa_replace_map): New field force_load_ref.
8454 * ipa-prop.h (ipa_param_descriptor): Reduce precision of move_cost,
8455 aded new flag load_dereferenced, adjusted comments.
8456 (ipa_get_param_dereferenced): New function.
8457 (ipa_set_param_dereferenced): Likewise.
8458 * cgraphclones.c (cgraph_node::create_virtual_clone): Follow it.
8459 * ipa-cp.c: Include gimple.h.
8460 (ipcp_discover_new_direct_edges): Take into account dereferenced flag.
8461 (get_replacement_map): New parameter force_load_ref, set the
8462 appropriate flag in ipa_replace_map if set.
8463 (struct symbol_and_index_together): New type.
8464 (adjust_refs_in_act_callers): New function.
8465 (adjust_references_in_caller): Likewise.
8466 (create_specialized_node): When appropriate, call
8467 adjust_references_in_caller and force only load references.
8468 * ipa-prop.c (load_from_dereferenced_name): New function.
8469 (ipa_analyze_controlled_uses): Also detect loads from a
8470 dereference, harden testing of call statements.
8471 (ipa_write_node_info): Stream the dereferenced flag.
8472 (ipa_read_node_info): Likewise.
8473 (ipa_set_jf_constant): Also create refdesc when jump function
8474 references a variable.
8475 (cgraph_node_for_jfunc): Rename to symtab_node_for_jfunc, work
8476 also on references of variables and return a symtab_node. Adjust
8478 (propagate_controlled_uses): Also remove references to VAR_DECLs.
8480 2021-07-27 Jakub Jelinek <jakub@redhat.com>
8482 PR middle-end/101586
8483 * gimple-fold.c (clear_padding_type): Ignore FIELD_DECLs with byte
8484 positions above or equal to sz except for diagnostics of flexible
8487 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
8489 PR tree-optimization/78888
8490 * gimple-range-fold.cc (get_letter_range): New.
8491 (fold_using_range::range_of_builtin_call): Call get_letter_range.
8493 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
8495 PR tree-optimization/78888
8496 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Add cases
8497 for CFN_BUILT_IN_TOUPPER and CFN_BUILT_IN_TOLOWER.
8499 2021-07-26 Roger Sayle <roger@nextmovesoftware.com>
8500 Marc Glisse <marc.glisse@inria.fr>
8502 * match.pd (rotate): Simplify equality/inequality of rotations.
8503 (bswap): Simplify equality/inequality tests of byte swapping.
8505 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8507 * range-op.cc (operator_bitwise_xor::op1_op2_relation_effect):
8510 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8512 * range-op.cc (operator_lshift::fold_range): Pass rel to
8513 base class fold_range.
8514 (operator_rshift::fold_range): Same.
8516 2021-07-26 Ashimida <ashimida@linux.alibaba.com>
8519 * toplev.h (min_align_loops_log): Remove declaration.
8520 (min_align_jumps_log, min_align_labels_log): Likewise.
8521 (min_align_functions_log): Likewise.
8523 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8525 * tree-vrp.c (vrp_simplify_cond_using_ranges): Rename vr_values
8527 (execute_vrp): Abstract out simplification of conditionals...
8528 (simplify_casted_conds): ...here.
8530 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8532 * gimple-array-bounds.cc (array_bounds_checker::get_value_range):
8533 Add gimple argument.
8534 (array_bounds_checker::check_array_ref): Same.
8535 (array_bounds_checker::check_addr_expr): Same.
8536 (array_bounds_checker::check_array_bounds): Pass statement to
8537 check_array_bounds and check_addr_expr.
8538 * gimple-array-bounds.h (check_array_bounds): Add gimple argument.
8539 (check_addr_expr): Same.
8540 (get_value_range): Same.
8542 2021-07-26 Tamar Christina <tamar.christina@arm.com>
8544 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): Rename to..
8545 (sdot_prod, udot_prod): ... This.
8546 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): Merged
8548 (<sur>dot_prod<vsi2qi>): ... this.
8549 (aarch64_<sur>dot_lane<vsi2qi>, aarch64_<sur>dot_laneq<vsi2qi>):
8550 Change operands order.
8551 (<sur>sadv16qi): Use new operands order.
8552 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32,
8553 vdotq_s32): Use new RTL ordering.
8555 2021-07-26 Tamar Christina <tamar.christina@arm.com>
8557 * config/aarch64/aarch64-builtins.c (TYPES_TERNOP_SUSS,
8558 aarch64_types_ternop_suss_qualifiers): New.
8559 * config/aarch64/aarch64-simd-builtins.def (usdot_prod): Use it.
8560 * config/aarch64/aarch64-simd.md (usdot_prod<vsi2qi>): Re-organize RTL.
8561 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Use it.
8563 2021-07-23 Jakub Jelinek <jakub@redhat.com>
8565 PR rtl-optimization/101562
8566 * expmed.c (store_integral_bit_field): Only use movstrict_optab
8567 if the operand isn't paradoxical.
8569 2021-07-23 Aldy Hernandez <aldyh@redhat.com>
8571 * gimple-array-bounds.h (class array_bounds_checker): Change
8572 ranges type to range_query.
8574 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8576 * config/aarch64/arm_neon.h (vst1_s64_x2): Use
8577 __builtin_memcpy instead of constructing
8578 __builtin_aarch64_simd_oi one vector at a time.
8579 (vst1_u64_x2): Likewise.
8580 (vst1_f64_x2): Likewise.
8581 (vst1_s8_x2): Likewise.
8582 (vst1_p8_x2): Likewise.
8583 (vst1_s16_x2): Likewise.
8584 (vst1_p16_x2): Likewise.
8585 (vst1_s32_x2): Likewise.
8586 (vst1_u8_x2): Likewise.
8587 (vst1_u16_x2): Likewise.
8588 (vst1_u32_x2): Likewise.
8589 (vst1_f16_x2): Likewise.
8590 (vst1_f32_x2): Likewise.
8591 (vst1_p64_x2): Likewise.
8592 (vst1q_s8_x2): Likewise.
8593 (vst1q_p8_x2): Likewise.
8594 (vst1q_s16_x2): Likewise.
8595 (vst1q_p16_x2): Likewise.
8596 (vst1q_s32_x2): Likewise.
8597 (vst1q_s64_x2): Likewise.
8598 (vst1q_u8_x2): Likewise.
8599 (vst1q_u16_x2): Likewise.
8600 (vst1q_u32_x2): Likewise.
8601 (vst1q_u64_x2): Likewise.
8602 (vst1q_f16_x2): Likewise.
8603 (vst1q_f32_x2): Likewise.
8604 (vst1q_f64_x2): Likewise.
8605 (vst1q_p64_x2): Likewise.
8607 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8609 * config/aarch64/arm_neon.h (vst1_s64_x3): Use
8610 __builtin_memcpy instead of constructing
8611 __builtin_aarch64_simd_ci one vector at a time.
8612 (vst1_u64_x3): Likewise.
8613 (vst1_f64_x3): Likewise.
8614 (vst1_s8_x3): Likewise.
8615 (vst1_p8_x3): Likewise.
8616 (vst1_s16_x3): Likewise.
8617 (vst1_p16_x3): Likewise.
8618 (vst1_s32_x3): Likewise.
8619 (vst1_u8_x3): Likewise.
8620 (vst1_u16_x3): Likewise.
8621 (vst1_u32_x3): Likewise.
8622 (vst1_f16_x3): Likewise.
8623 (vst1_f32_x3): Likewise.
8624 (vst1_p64_x3): Likewise.
8625 (vst1q_s8_x3): Likewise.
8626 (vst1q_p8_x3): Likewise.
8627 (vst1q_s16_x3): Likewise.
8628 (vst1q_p16_x3): Likewise.
8629 (vst1q_s32_x3): Likewise.
8630 (vst1q_s64_x3): Likewise.
8631 (vst1q_u8_x3): Likewise.
8632 (vst1q_u16_x3): Likewise.
8633 (vst1q_u32_x3): Likewise.
8634 (vst1q_u64_x3): Likewise.
8635 (vst1q_f16_x3): Likewise.
8636 (vst1q_f32_x3): Likewise.
8637 (vst1q_f64_x3): Likewise.
8638 (vst1q_p64_x3): Likewise.
8640 2021-07-23 H.J. Lu <hjl.tools@gmail.com>
8643 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): Don't return
8644 hard register when LRA is in progress.
8646 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8648 * config/aarch64/arm_neon.h (vst1_s8_x4): Use
8649 __builtin_memcpy instead of using a union.
8650 (vst1q_s8_x4): Likewise.
8651 (vst1_s16_x4): Likewise.
8652 (vst1q_s16_x4): Likewise.
8653 (vst1_s32_x4): Likewise.
8654 (vst1q_s32_x4): Likewise.
8655 (vst1_u8_x4): Likewise.
8656 (vst1q_u8_x4): Likewise.
8657 (vst1_u16_x4): Likewise.
8658 (vst1q_u16_x4): Likewise.
8659 (vst1_u32_x4): Likewise.
8660 (vst1q_u32_x4): Likewise.
8661 (vst1_f16_x4): Likewise.
8662 (vst1q_f16_x4): Likewise.
8663 (vst1_f32_x4): Likewise.
8664 (vst1q_f32_x4): Likewise.
8665 (vst1_p8_x4): Likewise.
8666 (vst1q_p8_x4): Likewise.
8667 (vst1_p16_x4): Likewise.
8668 (vst1q_p16_x4): Likewise.
8669 (vst1_s64_x4): Likewise.
8670 (vst1_u64_x4): Likewise.
8671 (vst1_p64_x4): Likewise.
8672 (vst1q_s64_x4): Likewise.
8673 (vst1q_u64_x4): Likewise.
8674 (vst1q_p64_x4): Likewise.
8675 (vst1_f64_x4): Likewise.
8676 (vst1q_f64_x4): Likewise.
8678 2021-07-23 Jonathan Wrightt <jonathan.wright@arm.com>
8680 * config/aarch64/arm_neon.h (vst2_s64): Use __builtin_memcpy
8681 instead of constructing __builtin_aarch64_simd_oi one vector
8683 (vst2_u64): Likewise.
8684 (vst2_f64): Likewise.
8685 (vst2_s8): Likewise.
8686 (vst2_p8): Likewise.
8687 (vst2_s16): Likewise.
8688 (vst2_p16): Likewise.
8689 (vst2_s32): Likewise.
8690 (vst2_u8): Likewise.
8691 (vst2_u16): Likewise.
8692 (vst2_u32): Likewise.
8693 (vst2_f16): Likewise.
8694 (vst2_f32): Likewise.
8695 (vst2_p64): Likewise.
8696 (vst2q_s8): Likewise.
8697 (vst2q_p8): Likewise.
8698 (vst2q_s16): Likewise.
8699 (vst2q_p16): Likewise.
8700 (vst2q_s32): Likewise.
8701 (vst2q_s64): Likewise.
8702 (vst2q_u8): Likewise.
8703 (vst2q_u16): Likewise.
8704 (vst2q_u32): Likewise.
8705 (vst2q_u64): Likewise.
8706 (vst2q_f16): Likewise.
8707 (vst2q_f32): Likewise.
8708 (vst2q_f64): Likewise.
8709 (vst2q_p64): Likewise.
8711 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8713 * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy
8714 instead of constructing __builtin_aarch64_simd_ci one vector
8716 (vst3_u64): Likewise.
8717 (vst3_f64): Likewise.
8718 (vst3_s8): Likewise.
8719 (vst3_p8): Likewise.
8720 (vst3_s16): Likewise.
8721 (vst3_p16): Likewise.
8722 (vst3_s32): Likewise.
8723 (vst3_u8): Likewise.
8724 (vst3_u16): Likewise.
8725 (vst3_u32): Likewise.
8726 (vst3_f16): Likewise.
8727 (vst3_f32): Likewise.
8728 (vst3_p64): Likewise.
8729 (vst3q_s8): Likewise.
8730 (vst3q_p8): Likewise.
8731 (vst3q_s16): Likewise.
8732 (vst3q_p16): Likewise.
8733 (vst3q_s32): Likewise.
8734 (vst3q_s64): Likewise.
8735 (vst3q_u8): Likewise.
8736 (vst3q_u16): Likewise.
8737 (vst3q_u32): Likewise.
8738 (vst3q_u64): Likewise.
8739 (vst3q_f16): Likewise.
8740 (vst3q_f32): Likewise.
8741 (vst3q_f64): Likewise.
8742 (vst3q_p64): Likewise.
8744 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8746 * config/aarch64/arm_neon.h (vst4_s64): Use __builtin_memcpy
8747 instead of constructing __builtin_aarch64_simd_xi one vector
8749 (vst4_u64): Likewise.
8750 (vst4_f64): Likewise.
8751 (vst4_s8): Likewise.
8752 (vst4_p8): Likewise.
8753 (vst4_s16): Likewise.
8754 (vst4_p16): Likewise.
8755 (vst4_s32): Likewise.
8756 (vst4_u8): Likewise.
8757 (vst4_u16): Likewise.
8758 (vst4_u32): Likewise.
8759 (vst4_f16): Likewise.
8760 (vst4_f32): Likewise.
8761 (vst4_p64): Likewise.
8762 (vst4q_s8): Likewise.
8763 (vst4q_p8): Likewise.
8764 (vst4q_s16): Likewise.
8765 (vst4q_p16): Likewise.
8766 (vst4q_s32): Likewise.
8767 (vst4q_s64): Likewise.
8768 (vst4q_u8): Likewise.
8769 (vst4q_u16): Likewise.
8770 (vst4q_u32): Likewise.
8771 (vst4q_u64): Likewise.
8772 (vst4q_f16): Likewise.
8773 (vst4q_f32): Likewise.
8774 (vst4q_f64): Likewise.
8775 (vst4q_p64): Likewise.
8777 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8779 * config/aarch64/arm_neon.h (vtbx4_s8): Use __builtin_memcpy
8780 instead of constructing __builtin_aarch64_simd_oi one vector
8782 (vtbx4_u8): Likewise.
8783 (vtbx4_p8): Likewise.
8785 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8787 * config/aarch64/arm_neon.h (vtbl3_s8): Use __builtin_memcpy
8788 instead of constructing __builtin_aarch64_simd_oi one vector
8790 (vtbl3_u8): Likewise.
8791 (vtbl3_p8): Likewise.
8792 (vtbl4_s8): Likewise.
8793 (vtbl4_u8): Likewise.
8794 (vtbl4_p8): Likewise.
8796 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8798 * config/aarch64/arm_neon.h (vqtbx2_s8): Use __builtin_memcpy
8799 instead of constructing __builtin_aarch64_simd_oi one vector
8801 (vqtbx2_u8): Likewise.
8802 (vqtbx2_p8): Likewise.
8803 (vqtbx2q_s8): Likewise.
8804 (vqtbx2q_u8): Likewise.
8805 (vqtbx2q_p8): Likewise.
8806 (vqtbx3_s8): Use __builtin_memcpy instead of constructing
8807 __builtin_aarch64_simd_ci one vector at a time.
8808 (vqtbx3_u8): Likewise.
8809 (vqtbx3_p8): Likewise.
8810 (vqtbx3q_s8): Likewise.
8811 (vqtbx3q_u8): Likewise.
8812 (vqtbx3q_p8): Likewise.
8813 (vqtbx4_s8): Use __builtin_memcpy instead of constructing
8814 __builtin_aarch64_simd_xi one vector at a time.
8815 (vqtbx4_u8): Likewise.
8816 (vqtbx4_p8): Likewise.
8817 (vqtbx4q_s8): Likewise.
8818 (vqtbx4q_u8): Likewise.
8819 (vqtbx4q_p8): Likewise.
8821 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8823 * config/aarch64/arm_neon.h (vqtbl2_s8): Use __builtin_memcpy
8824 instead of constructing __builtin_aarch64_simd_oi one vector
8826 (vqtbl2_u8): Likewise.
8827 (vqtbl2_p8): Likewise.
8828 (vqtbl2q_s8): Likewise.
8829 (vqtbl2q_u8): Likewise.
8830 (vqtbl2q_p8): Likewise.
8831 (vqtbl3_s8): Use __builtin_memcpy instead of constructing
8832 __builtin_aarch64_simd_ci one vector at a time.
8833 (vqtbl3_u8): Likewise.
8834 (vqtbl3_p8): Likewise.
8835 (vqtbl3q_s8): Likewise.
8836 (vqtbl3q_u8): Likewise.
8837 (vqtbl3q_p8): Likewise.
8838 (vqtbl4_s8): Use __builtin_memcpy instead of constructing
8839 __builtin_aarch64_simd_xi one vector at a time.
8840 (vqtbl4_u8): Likewise.
8841 (vqtbl4_p8): Likewise.
8842 (vqtbl4q_s8): Likewise.
8843 (vqtbl4q_u8): Likewise.
8844 (vqtbl4q_p8): Likewise.
8846 2021-07-23 Haochen Gui <guihaoc@gcc.gnu.org>
8849 * config/rs6000/rs6000.md (cstore<mode>4): Fix wrong fall through.
8851 2021-07-22 Andrew Pinski <apinski@marvell.com>
8853 PR tree-optimization/10153
8854 * tree-tailcall.c (create_tailcall_accumulator):
8855 Don't call fold_convert as the type should be correct already.
8856 (tree_optimize_tail_calls_1): Use build_{one,zero}_cst instead
8857 of integer_{one,zero}_node for the call of create_tailcall_accumulator.
8859 2021-07-22 Aldy Hernandez <aldyh@redhat.com>
8861 * gimple-range-cache.cc (non_null_ref::adjust_range): Replace
8862 varying_p check for null/non-null check.
8864 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
8866 PR tree-optimization/101511
8867 * value-relation.cc (relation_oracle::query_relation): Check if ssa1
8868 is in ssa2's equiv set, and don't trap if so.
8870 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
8872 PR tree-optimization/101497
8873 * gimple-range-fold.cc (fold_using_range::range_of_cond_expr): Check
8876 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
8878 PR tree-optimization/101496
8879 * vr-values.c (simplify_using_ranges::fold_cond): Call range_of_stmt
8880 first, then vrp_visit_cond_Stmt.
8882 2021-07-22 liuhongt <hongtao.liu@intel.com>
8884 * config/i386/i386-expand.c
8885 (ix86_broadcast_from_integer_constant): Rename to ..
8886 (ix86_broadcast_from_constant): .. this, and extend it to
8888 (ix86_expand_vector_move): Extend to float mode.
8889 * config/i386/i386-features.c
8890 (replace_constant_pool_with_broadcast): Remove.
8891 (remove_partial_avx_dependency_gate): Ditto.
8892 (constant_pool_broadcast): Ditto.
8893 (class pass_constant_pool_broadcast): Ditto.
8894 (make_pass_constant_pool_broadcast): Ditto.
8895 (remove_partial_avx_dependency): Adjust gate.
8896 * config/i386/i386-passes.def: Remove pass_constant_pool_broadcast.
8897 * config/i386/i386-protos.h
8898 (make_pass_constant_pool_broadcast): Remove.
8900 2021-07-22 liuhongt <hongtao.liu@intel.com>
8902 * config/i386/constraints.md (Wb): New constraint.
8904 * config/i386/i386.md (*ashlhi3_1): Extend to avx512 mask
8906 (*ashlqi3_1): Ditto.
8907 (*<insn><mode>3_1): Split to ..
8908 (*ashr<mode>3_1): this, ...
8909 (*lshr<mode>3_1): and this, also extend this pattern to avx512
8911 (*<insn><mode>3_1): Split to ..
8912 (*ashr<mode>3_1): this, ...
8913 (*lshrqi3_1): and this, also extend this pattern to avx512
8915 (*lshrhi3_1): And this, also extend this pattern to avx512
8917 * config/i386/sse.md (k<code><mode>): New define_split after
8918 it to convert generic shift pattern to mask shift ones.
8920 2021-07-21 Thomas Schwinge <thomas@codesourcery.com>
8921 Joseph Myers <joseph@codesourcery.com>
8922 Cesar Philippidis <cesar@codesourcery.com>
8924 * tree-core.h (omp_clause_code): Add 'OMP_CLAUSE_NOHOST'.
8925 * tree.c (omp_clause_num_ops, omp_clause_code_name, walk_tree_1):
8927 * tree-pretty-print.c (dump_omp_clause): Likewise.
8928 * omp-general.c (oacc_verify_routine_clauses): Likewise.
8929 * gimplify.c (gimplify_scan_omp_clauses)
8930 (gimplify_adjust_omp_clauses): Likewise.
8931 * tree-nested.c (convert_nonlocal_omp_clauses)
8932 (convert_local_omp_clauses): Likewise.
8933 * omp-low.c (scan_sharing_clauses): Likewise.
8934 * omp-offload.c (execute_oacc_device_lower): Update.
8936 2021-07-21 Martin Sebor <msebor@redhat.com>
8938 * tree-ssa-alias.c (walk_aliased_vdefs_1): Fix typos in a comment.
8940 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8942 * config/rs6000/rs6000-gen-builtins.c (write_init_bif_table):
8945 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8947 * config/rs6000/rs6000-gen-builtins.c (write_fntype): New
8949 (write_fntype_init): New stub function.
8950 (write_init_bif_table): Likewise.
8951 (write_init_ovld_table): New function.
8952 (write_init_file): Implement.
8954 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8956 * config/rs6000/rs6000-gen-builtins.c
8957 (write_autogenerated_header): New function.
8958 (write_decls): Likewise.
8959 (write_extern_fntype): New callback function.
8960 (write_header_file): Implement.
8962 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8964 * config/rs6000/rs6000-gen-builtins.c (write_defines_file):
8967 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8969 * config/rs6000/rs6000-gen-builtins.c (complete_vector_type): New
8971 (complete_base_type): Likewise.
8972 (construct_fntype_id): Likewise.
8973 (parse_bif_entry): Call contruct_fntype_id.
8974 (parse_ovld_entry): Likewise.
8976 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8978 * config/rs6000/rs6000-gen-builtins.c (ovld_stanza): New struct.
8979 (MAXOVLDSTANZAS): New macro.
8980 (ovld_stanzas): New variable.
8981 (curr_ovld_stanza): Likewise.
8982 (MAXOVLDS): New macro.
8983 (ovlddata): New struct.
8984 (ovlds): New variable.
8985 (curr_ovld): Likewise.
8986 (max_ovld_args): Likewise.
8987 (parse_ovld_entry): New function.
8988 (parse_ovld_stanza): Likewise.
8989 (parse_ovld): Implement.
8991 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8993 * config/rs6000/rs6000-gen-builtins.c (parse_bif_attrs):
8996 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8998 * config/rs6000/rs6000-gen-builtins.c (parse_args): New function.
8999 (parse_prototype): Implement.
9001 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9003 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): New enum.
9004 (curr_bif_stanza): New variable.
9005 (stanza_entry): New struct.
9006 (stanza_map): New initialized variable.
9007 (enable_string): Likewise.
9008 (fnkinds): New enum.
9009 (typelist): New struct.
9010 (attrinfo): Likewise.
9011 (MAXRESTROPNDS): New macro.
9012 (prototype): New struct.
9013 (MAXBIFS): New macro.
9014 (bifdata): New struct.
9015 (bifs): New variable.
9016 (curr_bif): Likewise.
9017 (bif_order): Likewise.
9018 (bif_index): Likewise.
9019 (fatal): New function.
9020 (stanza_name_to_stanza): Likewise.
9021 (parse_bif_attrs): New stub function.
9022 (parse_prototype): Likewise.
9023 (parse_bif_entry): New function.
9024 (parse_bif_stanza): Likewise.
9025 (parse_bif): Implement.
9026 (set_bif_order): New function.
9027 (create_bif_order): Implement.
9029 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9031 * config/rs6000/rs6000-gen-builtins.c (rbtree.h): New #include.
9032 (num_bifs): New variable.
9033 (num_ovld_stanzas): Likewise.
9034 (num_ovlds): Likewise.
9035 (parse_codes): New enum.
9036 (bif_rbt): New variable.
9037 (ovld_rbt): Likewise.
9038 (fntype_rbt): Likewise.
9039 (bifo_rbt): Likewise.
9040 (parse_bif): New stub function.
9041 (create_bif_order): Likewise.
9042 (parse_ovld): Likewise.
9043 (write_header_file): Likewise.
9044 (write_init_file): Likewise.
9045 (write_defines_file): Likewise.
9046 (delete_output_files): New function.
9049 2021-07-21 H.J. Lu <hjl.tools@gmail.com>
9052 * config/i386/i386-builtin.def: Remove OPTION_MASK_ISA_SSE4_2
9053 from CRC32 _builtin functions.
9055 2021-07-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
9057 * coverage.c (build_gcov_info_var_registration): Mark the object placed
9058 in the linker set as referenced so that it does not get optimized away.
9060 2021-07-21 Kito Cheng <kito.cheng@sifive.com>
9063 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
9065 * config.gcc (riscv*-*-*): Detect which python is available.
9067 2021-07-21 Jakub Jelinek <jakub@redhat.com>
9069 PR middle-end/101535
9070 * gimplify.c (omp_check_private): Properly skip ORT_TARGET_DATA
9071 contexts in which decl isn't privatized and for ORT_TARGET return
9072 false if decl is mapped.
9074 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
9076 * gimple-loop-jam.c: Include tree-ssa-sccvn.h.
9077 (tree_loop_unroll_and_jam): Run value-numbering on a loop that
9078 has been successfully unrolled.
9080 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
9082 * tree-ssa-loop-manip.c (determine_exit_conditions): Return a null
9083 exit condition if no tail loop is needed, and if the original exit
9084 condition should therefore be kept as-is.
9085 (tree_transform_and_unroll_loop): Handle that case here too.
9087 2021-07-21 Kewen Lin <linkw@linux.ibm.com>
9089 * tree-data-ref.c (free_dependence_relations): Adjust to pass vec
9091 (free_data_refs): Likewise.
9092 * tree-data-ref.h (free_dependence_relations): Likewise.
9093 (free_data_refs): Likewise.
9094 * tree-predcom.c (struct chain): Use auto_vec instead of vec for
9096 (struct component): Likewise.
9097 (pcom_worker::pcom_worker): Adjust for auto_vec and renaming changes.
9098 (pcom_worker::~pcom_worker): Likewise.
9099 (pcom_worker::release_chain): Adjust as auto_vec changes.
9100 (pcom_worker::loop): Rename to ...
9101 (pcom_worker::m_loop): ... this.
9102 (pcom_worker::datarefs): Rename to ...
9103 (pcom_worker::m_datarefs): ... this. Use auto_vec instead of vec.
9104 (pcom_worker::dependences): Rename to ...
9105 (pcom_worker::m_dependences): ... this. Use auto_vec instead of vec.
9106 (pcom_worker::chains): Rename to ...
9107 (pcom_worker::m_chains): ... this. Use auto_vec instead of vec.
9108 (pcom_worker::looparound_phis): Rename to ...
9109 (pcom_worker::m_looparound_phis): ... this. Use auto_vec instead of
9111 (pcom_worker::cache): Rename to ...
9112 (pcom_worker::m_cache): ... this. Use auto_vec instead of vec.
9113 (pcom_worker::release_chain): Adjust for auto_vec changes.
9114 (pcom_worker::release_chains): Adjust for auto_vec and renaming
9116 (release_component): Remove.
9117 (release_components): Adjust for release_component removal.
9118 (component_of): Adjust to use vec.
9119 (merge_comps): Likewise.
9120 (pcom_worker::aff_combination_dr_offset): Adjust for renaming changes.
9121 (pcom_worker::determine_offset): Likewise.
9122 (class comp_ptrs): Remove.
9123 (pcom_worker::split_data_refs_to_components): Adjust for renaming
9124 changes, for comp_ptrs removal with auto_vec.
9125 (pcom_worker::suitable_component_p): Adjust for renaming changes.
9126 (pcom_worker::filter_suitable_components): Adjust for release_component
9128 (pcom_worker::valid_initializer_p): Adjust for renaming changes.
9129 (pcom_worker::find_looparound_phi): Likewise.
9130 (pcom_worker::add_looparound_copies): Likewise.
9131 (pcom_worker::determine_roots_comp): Likewise.
9132 (pcom_worker::single_nonlooparound_use): Likewise.
9133 (pcom_worker::execute_pred_commoning_chain): Likewise.
9134 (pcom_worker::execute_pred_commoning): Likewise.
9135 (pcom_worker::try_combine_chains): Likewise.
9136 (pcom_worker::prepare_initializers_chain): Likewise.
9137 (pcom_worker::prepare_initializers): Likewise.
9138 (pcom_worker::prepare_finalizers_chain): Likewise.
9139 (pcom_worker::prepare_finalizers): Likewise.
9140 (pcom_worker::tree_predictive_commoning_loop): Likewise.
9142 2021-07-20 Martin Sebor <msebor@redhat.com>
9144 PR middle-end/101397
9145 * builtins.c (gimple_call_return_array): Add argument. Correct
9146 offsets for memchr, mempcpy, stpcpy, and stpncpy.
9147 (compute_objsize_r): Adjust offset computation for argument returning
9150 2021-07-20 Martin Sebor <msebor@redhat.com>
9152 PR middle-end/101300
9153 * tree-ssa-uninit.c (check_defs): Handle UBSAN built-ins.
9155 2021-07-20 Jeff Law <jlaw@localhost.localdomain>
9157 * function.c (assign_parm_setup_block): Use adjust_address instead
9158 of change_address to preserve MEM_EXPR and friends.
9160 2021-07-20 Martin Sebor <msebor@redhat.com>
9162 * cfgloop.h (single_likely_exit): Adjust by-value argument to
9164 * cfgloopanal.c (single_likely_exit): Same.
9165 * cgraph.h (struct cgraph_node): Same.
9166 * cgraphclones.c (cgraph_node::create_virtual_clone): Same.
9167 * genautomata.c (merge_states): Same.
9168 * genextract.c (VEC_char_to_string): Same.
9169 * genmatch.c (dt_node::gen_kids_1): Same.
9170 (walk_captures): Adjust by-value argument to by-reference.
9171 * gimple-ssa-store-merging.c (check_no_overlap): Adjust by-value argument
9172 to by-const-reference.
9173 * gimple.c (gimple_build_call_vec): Same.
9174 (gimple_build_call_internal_vec): Same.
9175 (gimple_build_switch): Same.
9176 (sort_case_labels): Same.
9177 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
9179 * gimple.h (gimple_build_call_vec): Adjust by-value argument to
9181 (gimple_build_call_internal_vec): Same.
9182 (gimple_build_switch): Same.
9183 (sort_case_labels): Same.
9184 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
9186 * haifa-sched.c (calc_priorities): Adjust by-value argument to
9188 (sched_init_luids): Same.
9189 (haifa_init_h_i_d): Same.
9190 * ipa-cp.c (ipa_get_indirect_edge_target_1): Same.
9191 (adjust_callers_for_value_intersection): Adjust by-value argument to
9193 (find_more_scalar_values_for_callers_subset): Adjust by-value argument to
9195 (find_more_contexts_for_caller_subset): Same.
9196 (find_aggregate_values_for_callers_subset): Same.
9197 (copy_useful_known_contexts): Same.
9198 * ipa-fnsummary.c (remap_edge_summaries): Same.
9199 (remap_freqcounting_predicate): Same.
9200 * ipa-inline.c (add_new_edges_to_heap): Adjust by-value argument to
9202 * ipa-predicate.c (predicate::remap_after_inlining): Adjust by-value argument
9203 to by-const-reference.
9204 * ipa-predicate.h (predicate::remap_after_inlining): Same.
9205 * ipa-prop.c (ipa_find_agg_cst_for_param): Same.
9206 * ipa-prop.h (ipa_find_agg_cst_for_param): Same.
9207 * ira-build.c (ira_loop_tree_body_rev_postorder): Same.
9208 * read-rtl.c (add_overload_instance): Same.
9209 * rtl.h (native_decode_rtx): Same.
9210 (native_decode_vector_rtx): Same.
9211 * sched-int.h (sched_init_luids): Same.
9212 (haifa_init_h_i_d): Same.
9213 * simplify-rtx.c (native_decode_vector_rtx): Same.
9214 (native_decode_rtx): Same.
9215 * tree-call-cdce.c (gen_shrink_wrap_conditions): Same.
9216 (shrink_wrap_one_built_in_call_with_conds): Same.
9217 (shrink_wrap_conditional_dead_built_in_calls): Same.
9218 * tree-data-ref.c (create_runtime_alias_checks): Same.
9219 (compute_all_dependences): Same.
9220 * tree-data-ref.h (compute_all_dependences): Same.
9221 (create_runtime_alias_checks): Same.
9222 (index_in_loop_nest): Same.
9223 * tree-if-conv.c (mask_exists): Same.
9224 * tree-loop-distribution.c (class loop_distribution): Same.
9225 (loop_distribution::create_rdg_vertices): Same.
9226 (dump_rdg_partitions): Same.
9227 (debug_rdg_partitions): Same.
9228 (partition_contains_all_rw): Same.
9229 (loop_distribution::distribute_loop): Same.
9230 * tree-parloops.c (oacc_entry_exit_ok_1): Same.
9231 (oacc_entry_exit_single_gang): Same.
9232 * tree-ssa-loop-im.c (hoist_memory_references): Same.
9233 (loop_suitable_for_sm): Same.
9234 * tree-ssa-loop-niter.c (bound_index): Same.
9235 * tree-ssa-reassoc.c (update_ops): Same.
9236 (swap_ops_for_binary_stmt): Same.
9237 (rewrite_expr_tree): Same.
9238 (rewrite_expr_tree_parallel): Same.
9239 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Same.
9240 * tree-ssa-sccvn.h (ao_ref_init_from_vn_reference): Same.
9241 * tree-ssa-structalias.c (process_all_all_constraints): Same.
9242 (make_constraints_to): Same.
9243 (handle_lhs_call): Same.
9244 (find_func_aliases_for_builtin_call): Same.
9245 (sort_fieldstack): Same.
9246 (check_for_overlaps): Same.
9247 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Same.
9248 (vect_create_cond_for_unequal_addrs): Same.
9249 (vect_create_cond_for_lower_bounds): Same.
9250 (vect_create_cond_for_alias_checks): Same.
9251 * tree-vect-slp-patterns.c (vect_validate_multiplication): Same.
9252 * tree-vect-slp.c (vect_analyze_slp_instance): Same.
9253 (vect_make_slp_decision): Same.
9254 (vect_slp_bbs): Same.
9255 (duplicate_and_interleave): Same.
9256 (vect_transform_slp_perm_load): Same.
9257 (vect_schedule_slp): Same.
9258 * tree-vectorizer.h (vect_transform_slp_perm_load): Same.
9259 (vect_schedule_slp): Same.
9260 (duplicate_and_interleave): Same.
9261 * tree.c (build_vector_from_ctor): Same.
9262 (build_vector): Same.
9263 (check_vector_cst): Same.
9264 (check_vector_cst_duplicate): Same.
9265 (check_vector_cst_fill): Same.
9266 (check_vector_cst_stepped): Same.
9267 * tree.h (build_vector_from_ctor): Same.
9269 2021-07-20 Jakub Jelinek <jakub@redhat.com>
9272 * config/rs6000/rs6000-protos.h (easy_altivec_constant): Change return
9273 type from bool to int.
9274 * config/rs6000/rs6000.c (vspltis_constant): Fix up handling the
9275 EASY_VECTOR_MSB case if either step or copies is not 1.
9276 (vspltis_shifted): Fix comment typo.
9277 (easy_altivec_constant): Change return type from bool to int, instead
9278 of returning true return byte size of the element mode that should be
9279 used to synthetize the constant.
9280 * config/rs6000/predicates.md (easy_vector_constant_msb): Require
9281 that vspltis_shifted is 0, handle the case where easy_altivec_constant
9282 assumes using different vector mode from CONST_VECTOR's mode.
9283 * config/rs6000/altivec.md (easy_vector_constant_msb splitter): Use
9284 easy_altivec_constant to determine mode in which -1 >> -1 should be
9285 performed, use rs6000_expand_vector_init instead of gen_vec_initv4sisi.
9287 2021-07-20 Richard Biener <rguenther@suse.de>
9290 * dwarf2out.h (dwarf_file_data): Add key member.
9291 * dwarf2out.c (dwarf_file_hasher::equal): Compare key.
9292 (dwarf_file_hasher::hash): Hash key.
9293 (lookup_filename): Remap the filename and store it in the
9294 filename member of dwarf_file_data when creating a new
9296 (file_name_acquire): Do not remap the filename again.
9297 (maybe_emit_file): Likewise.
9299 2021-07-20 Jonathan Wright <jonathan.wright@arm.com>
9301 * config/aarch64/aarch64-simd-builtins.def: Use two variant
9302 generators for all TBL/TBX intrinsics and rename to
9303 consistent forms: qtbl[1234] or qtbx[1234].
9304 * config/aarch64/aarch64-simd.md (aarch64_tbl1<mode>):
9306 (aarch64_qtbl1<mode>): This.
9307 (aarch64_tbx1<mode>): Rename to...
9308 (aarch64_qtbx1<mode>): This.
9309 (aarch64_tbl2v16qi): Delete.
9310 (aarch64_tbl3<mode>): Rename to...
9311 (aarch64_qtbl2<mode>): This.
9312 (aarch64_tbx4<mode>): Rename to...
9313 (aarch64_qtbx2<mode>): This.
9314 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_1): Use
9315 renamed qtbl1 and qtbl2 RTL patterns.
9316 * config/aarch64/arm_neon.h (vqtbl1_p8): Use renamed qtbl1
9318 (vqtbl1_s8): Likewise.
9319 (vqtbl1_u8): Likewise.
9320 (vqtbl1q_p8): Likewise.
9321 (vqtbl1q_s8): Likewise.
9322 (vqtbl1q_u8): Likewise.
9323 (vqtbx1_s8): Use renamed qtbx1 RTL pattern.
9324 (vqtbx1_u8): Likewise.
9325 (vqtbx1_p8): Likewise.
9326 (vqtbx1q_s8): Likewise.
9327 (vqtbx1q_u8): Likewise.
9328 (vqtbx1q_p8): Likewise.
9329 (vtbl1_s8): Use renamed qtbl1 RTL pattern.
9330 (vtbl1_u8): Likewise.
9331 (vtbl1_p8): Likewise.
9332 (vtbl2_s8): Likewise
9333 (vtbl2_u8): Likewise.
9334 (vtbl2_p8): Likewise.
9335 (vtbl3_s8): Use renamed qtbl2 RTL pattern.
9336 (vtbl3_u8): Likewise.
9337 (vtbl3_p8): Likewise.
9338 (vtbl4_s8): Likewise.
9339 (vtbl4_u8): Likewise.
9340 (vtbl4_p8): Likewise.
9341 (vtbx2_s8): Use renamed qtbx2 RTL pattern.
9342 (vtbx2_u8): Likewise.
9343 (vtbx2_p8): Likewise.
9344 (vqtbl2_s8): Use renamed qtbl2 RTL pattern.
9345 (vqtbl2_u8): Likewise.
9346 (vqtbl2_p8): Likewise.
9347 (vqtbl2q_s8): Likewise.
9348 (vqtbl2q_u8): Likewise.
9349 (vqtbl2q_p8): Likewise.
9350 (vqtbx2_s8): Use renamed qtbx2 RTL pattern.
9351 (vqtbx2_u8): Likewise.
9352 (vqtbx2_p8): Likewise.
9353 (vqtbx2q_s8): Likewise.
9354 (vqtbx2q_u8): Likewise.
9355 (vqtbx2q_p8): Likewise.
9356 (vtbx4_s8): Likewise.
9357 (vtbx4_u8): Likewise.
9358 (vtbx4_p8): Likewise.
9360 2021-07-20 Uroš Bizjak <ubizjak@gmail.com>
9363 * config/i386/sync.md (define_peephole2 atomic_storedi_fpu):
9365 (define_peephole2 atomic_loaddi_fpu): Ditto.
9367 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
9369 * config.gcc (riscv*-*-*): Detect which python is available.
9371 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
9373 * config/rs6000/vsx.md (mulhs_<mode>): Rename to...
9374 (smul<mode>3_highpart): ... this.
9375 (mulhu_<mode>): Rename to...
9376 (umul<mode>3_highpart): ... this.
9377 * config/rs6000/rs6000-builtin.def (MULHS_V2DI, MULHS_V4SI,
9378 MULHU_V2DI, MULHU_V4SI): Adjust.
9380 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
9382 PR tree-optimization/100696
9383 * internal-fn.c (first_commutative_argument): Add info for IFN_MULH.
9384 * internal-fn.def (IFN_MULH): New internal function.
9385 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Add support to
9386 recog normal multiply highpart as IFN_MULH.
9387 * config/i386/i386.c (ix86_add_stmt_cost): Adjust for combined
9390 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
9392 * config/elfos.h (CTF_DEBUGGING_INFO): New definition.
9393 (BTF_DEBUGGING_INFO): Likewise.
9394 * doc/tm.texi.in: Document the new macros.
9395 * doc/tm.texi: Regenerated.
9396 * toplev.c: Guard initialization of debug hooks.
9398 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
9400 * flags.h (ctf_debuginfo_p): New function declaration.
9401 * opts.c (ctf_debuginfo_p): New function definition.
9403 2021-07-19 Andrew Stubbs <ams@codesourcery.com>
9406 * config/gcn/gcn-hsa.h (DRIVER_SELF_SPECS): New.
9407 (ASM_SPEC): Set -mattr for xnack and sram-ecc.
9408 * config/gcn/gcn-opts.h (enum sram_ecc_type): New.
9409 * config/gcn/gcn-valu.md: Add a warning comment.
9410 * config/gcn/gcn.c (gcn_option_override): Add "sorry" for -mxnack.
9411 (output_file_start): Add xnack and sram-ecc state to ".amdgcn_target".
9412 * config/gcn/gcn.md: Add a warning comment.
9413 * config/gcn/gcn.opt: Add -mxnack and -msram-ecc.
9414 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): Remove
9416 (EF_AMDGPU_XNACK): New.
9417 (EF_AMDGPU_SRAM_ECC): New.
9419 (copy_early_debug_info): Use elf_flags.
9420 (main): Handle -mxnack and -msram-ecc options.
9421 * doc/invoke.texi: Document -mxnack and -msram-ecc.
9423 2021-07-19 Andrew Pinski <apinski@marvell.com>
9426 * config/aarch64/aarch64.md (csneg3_uxtw_insn): Rename to ...
9427 (*cs<neg_not_cs>3_uxtw_insn4): and extend to NEG_NOT.
9429 2021-07-19 Richard Biener <rguenther@suse.de>
9431 PR tree-optimization/101505
9432 * tree-vect-patterns.c (vect_determine_precisions): Walk
9433 PHIs also for loop vectorization.
9435 2021-07-19 Richard Biener <rguenther@suse.de>
9437 * gimple.h (gimple_expr_type): Remove.
9438 * doc/gimple.texi: Remove gimple_expr_type documentation.
9440 2021-07-19 Richard Biener <rguenther@suse.de>
9442 * tree-ssa-sccvn.c (vn_reference_eq): Handle NULL vr->type.
9443 (ao_ref_init_from_vn_reference): Likewise.
9444 (fully_constant_reference): Likewise.
9445 (vn_reference_lookup_call): Do not set vr->type to random
9447 * tree-ssa-pre.c (compute_avail): Do not try to PRE calls
9449 * tree-vect-generic.c (expand_vector_piecewise): Pass in
9450 whether we expanded parallel.
9451 (expand_vector_parallel): Adjust.
9452 (expand_vector_addition): Likewise.
9453 (expand_vector_comparison): Likewise.
9454 (expand_vector_operation): Likewise.
9455 (expand_vector_scalar_condition): Likewise.
9456 (expand_vector_conversion): Likewise.
9458 2021-07-19 Richard Biener <rguenther@suse.de>
9460 * tree-vrp.c (register_edge_assert_for_2): Use the
9462 (vrp_folder::fold_predicate_in): Likewise.
9463 * vr-values.c (gimple_assign_nonzero_p): Likewise.
9464 (vr_values::extract_range_from_comparison): Likewise.
9465 (vr_values::extract_range_from_ubsan_builtin): Use the
9466 type of the first operand.
9467 (vr_values::extract_range_basic): Push down type
9468 computation, use the appropriate LHS.
9469 (vr_values::extract_range_from_assignment): Use the
9472 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
9475 * common/config/i386/i386-common.c (ix86_handle_option): For
9476 -mgeneral-regs-only, enable the GPR only instructions which are
9477 enabled implicitly by SSE ISAs unless they have been disabled
9480 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
9483 * config/i386/i386.c (ix86_check_avx_upper_stores): Moved before
9484 ix86_avx_u128_mode_needed.
9485 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY if callee
9486 returns AVX register.
9488 2021-07-17 Jan Hubicka <hubicka@ucw.cz>
9490 * tree-ssa-structalias.c (handle_rhs_call): Support EAF_NOT_RETURNED.
9491 (handle_const_call): Liekise
9492 (handle_pure_call): Liekise
9494 2021-07-17 Andrew MacLeod <amacleod@redhat.com>
9496 PR tree-optimization/96542
9497 * range-op.cc (range_operator::wi_fold_in_parts): New.
9498 (range_operator::fold_range): Call wi_fold_in_parts.
9499 (operator_lshift::wi_fold): Fix broken lshift by [0,0].
9500 * range-op.h (wi_fold_in_parts): Add prototype.
9502 2021-07-16 David Malcolm <dmalcolm@redhat.com>
9504 * doc/analyzer.texi: Add __analyzer_dump_state.
9506 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9508 * config/rs6000/rbtree.c: New file.
9509 * config/rs6000/rbtree.h: New file.
9511 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9513 * config/rs6000/rs6000-gen-builtins.c (restriction): New enum.
9514 (typeinfo): Add restr field.
9515 (match_bracketed_pair): New function.
9516 (match_const_restriction): Implement.
9518 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9520 * config/rs6000/rs6000-gen-builtins.c (match_basetype): Implement.
9522 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9524 * config/rs6000/rs6000-gen-builtins.c (void_status): New enum.
9525 (basetype): Likewise.
9526 (typeinfo): Likewise.
9527 (handle_pointer): New function.
9528 (match_basetype): New stub function.
9529 (match_const_restriction): Likewise.
9530 (match_type): New function.
9532 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9534 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace): New
9536 (advance_line): Likewise.
9537 (safe_inc_pos): Likewise.
9538 (match_identifier): Likewise.
9539 (match_integer): Likewise.
9540 (match_to_right_bracket): Likewise.
9542 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9544 * config/rs6000/rs6000-gen-builtins.c (bif_file): New variable.
9545 (ovld_file): Likewise.
9546 (header_file): Likewise.
9547 (init_file): Likewise.
9548 (defines_file): Likewise.
9549 (pgm_path): Likewise.
9550 (bif_path): Likewise.
9551 (ovld_path): Likewise.
9552 (header_path): Likewise.
9553 (init_path): Likewise.
9554 (defines_path): Likewise.
9555 (LINELEN): New macro.
9556 (linebuf): New variable.
9560 (bif_diag): New function.
9561 (ovld_diag): Likewise.
9563 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9565 * config/rs6000/rs6000-builtin-new.def: New.
9566 * config/rs6000/rs6000-overload.def: New.
9568 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9570 * config/rs6000/rs6000-gen-builtins.c: New.
9572 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9574 * Makefile.in (EXTRA_GTYPE_DEPS): New variable.
9575 (s-gtype): Depend on EXTRA_GTYPE_DEPS.
9576 * gengtype-state.c (state_writer::write_state_file_list): Add a
9577 parameter to the fileslist expression for the number of build
9579 (read_state_files_list): Detect build headers and strip the
9580 initial "./" or ".\" from their names.
9581 * gengtype.c (build_headers): New global variable.
9582 (num_build_headers): Likewise.
9583 (open_base_files): Emit #include for each build header.
9584 (main): Detect and count build headers.
9585 * gengtype.h (build_headers): New extern variable.
9586 (num_build_headers): Likewise.
9588 2021-07-16 Richard Biener <rguenther@suse.de>
9590 * gimple-ssa-store-merging.c (verify_symbolic_number_p): Use
9591 the type of the LHS.
9592 (find_bswap_or_nop_1): Likewise.
9593 (find_bswap_or_nop): Likewise.
9594 * tree-vectorizer.h (vect_get_smallest_scalar_type): Adjust
9596 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
9597 Remove unused parameters, pass in the scalar type. Fix
9598 internal store function handling.
9599 * tree-vect-stmts.c (vect_analyze_stmt): Remove assert.
9600 (vect_get_vector_types_for_stmt): Move down check for
9601 existing vector stmt after we've determined a scalar type.
9602 Pass down the used scalar type to vect_get_smallest_scalar_type.
9603 * tree-vect-generic.c (expand_vector_condition): Use
9604 the type of the LHS.
9605 (expand_vector_scalar_condition): Likewise.
9606 (expand_vector_operations_1): Likewise.
9607 * tree-vect-patterns.c (vect_widened_op_tree): Likewise.
9608 (vect_recog_dot_prod_pattern): Likewise.
9609 (vect_recog_sad_pattern): Likewise.
9610 (vect_recog_widen_op_pattern): Likewise.
9611 (vect_recog_widen_sum_pattern): Likewise.
9612 (vect_recog_mixed_size_cond_pattern): Likewise.
9614 2021-07-16 Jan Hubicka <hubicka@ucw.cz>
9616 * ipa-modref.c (struct escape_entry): Use eaf_fleags_t.
9617 (dump_eaf_flags): Dump EAF_NOT_RETURNED
9618 (eaf_flags_useful_p): Use eaf_fleags_t; handle const functions
9619 and EAF_NOT_RETURNED.
9620 (modref_summary::useful_p): Likewise.
9621 (modref_summary_lto::useful_p): Likewise.
9622 (struct) modref_summary_lto: Use eaf_fleags_t.
9623 (deref_flags): Handle EAF_NOT_RETURNED.
9624 (struct escape_point): Use min_flags.
9625 (modref_lattice::init): Add EAF_NOT_RETURNED.
9626 (merge_call_lhs_flags): Ignore EAF_NOT_RETURNED functions
9627 (analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
9629 (analyze_parms): Also analyze const functions; update conition on
9631 (modref_write): Update streaming.
9632 (read_section): Update streaming.
9633 (remap_arg_flags): Use eaf_flags_t.
9634 (modref_merge_call_site_flags): Hanlde EAF_NOT_RETURNED.
9635 * ipa-modref.h: (eaf_flags_t): New typedef.
9636 (struct modref_summary): Use eaf_flags_t.
9637 * tree-core.h (EAF_NOT_RETURNED): New constant.
9639 2021-07-16 Richard Biener <rguenther@suse.de>
9641 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Use
9642 the type of the LHS.
9643 (gimple_assign_nonnegative_warnv_p): Likewise.
9644 (gimple_call_nonnegative_warnv_p): Likewise. Return false
9645 if the call has no LHS.
9646 * gimple.c (gimple_could_trap_p_1): Use the type of the LHS.
9647 * tree-eh.c (stmt_could_throw_1_p): Likewise.
9648 * tree-inline.c (insert_init_stmt): Likewise.
9649 * tree-ssa-loop-niter.c (get_val_for): Likewise.
9650 * tree-outof-ssa.c (ssa_is_replaceable_p): Use the type of
9652 * tree-ssa-sccvn.c (init_vn_nary_op_from_stmt): Take a
9653 gassign *. Use the type of the lhs.
9654 (vn_nary_op_lookup_stmt): Adjust.
9655 (vn_nary_op_insert_stmt): Likewise.
9657 2021-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
9659 * config/s390/predicates.md (bras_sym_operand): Accept all
9660 functions in 64-bit mode, use UNSPEC_PLT31.
9661 (larl_operand): Use UNSPEC_PLT31.
9662 * config/s390/s390.c (s390_loadrelative_operand_p): Likewise.
9663 (legitimize_pic_address): Likewise.
9664 (s390_emit_tls_call_insn): Mark __tls_get_offset as function,
9666 (s390_delegitimize_address): Use UNSPEC_PLT31.
9667 (s390_output_addr_const_extra): Likewise.
9668 (print_operand): Add @PLT to TLS calls, handle %K.
9669 (s390_function_profiler): Mark __fentry__/_mcount as function,
9670 use %K, use UNSPEC_PLT31.
9671 (s390_output_mi_thunk): Use only UNSPEC_GOT, use %K.
9672 (s390_emit_call): Use UNSPEC_PLT31.
9673 (s390_emit_tpf_eh_return): Mark __tpf_eh_return as function.
9674 * config/s390/s390.md (UNSPEC_PLT31): Rename from UNSPEC_PLT.
9675 (*movdi_64): Use %K.
9676 (reload_base_64): Likewise.
9677 (*sibcall_brc): Likewise.
9678 (*sibcall_brcl): Likewise.
9679 (*sibcall_value_brc): Likewise.
9680 (*sibcall_value_brcl): Likewise.
9683 (*bras_r): Likewise.
9684 (*brasl_r): Likewise.
9685 (*bras_tls): Likewise.
9686 (*brasl_tls): Likewise.
9687 (main_base_64): Likewise.
9688 (reload_base_64): Likewise.
9689 (@split_stack_call<mode>): Likewise.
9691 2021-07-16 Richard Biener <rguenther@suse.de>
9693 PR tree-optimization/101467
9694 * tree-vect-stmts.c (vect_gen_while): Properly guard
9695 make_temp_ssa_name usage.
9697 2021-07-16 Cooper Qu <cooper.qu@linux.alibaba.com>
9699 * config.gcc: Don't use forked print-sysroot-suffix.sh and
9700 t-sysroot-suffix for C-SKY.
9701 * config/csky/print-sysroot-suffix.sh: Delete.
9702 * config/csky/t-csky-linux: Delete.
9703 * config/csky/t-sysroot-suffix: Define MULTILIB_DIRNAMES
9704 instead of CSKY_MULTILIB_DIRNAMES.
9706 2021-07-16 Richard Biener <rguenther@suse.de>
9708 * tree-vect-loop.c (vect_transform_cycle_phi): Correct sign
9709 conversion issues with the partial reduction of the reused
9712 2021-07-16 Richard Biener <rguenther@suse.de>
9714 * config/i386/i386-options.c (ix86_option_override_internal): Set
9715 param_vect_partial_vector_usage to zero if not set.
9717 2021-07-15 Uroš Bizjak <ubizjak@gmail.com>
9720 * config/i386/i386.h (VALID_SSE_REG_MODE): Add TDmode.
9721 (VALID_INT_MODE_P): Add SDmode and DDmode.
9722 Add TDmode for TARGET_64BIT.
9723 (VALID_DFP_MODE_P): Remove.
9724 * config/i386/i386.c (ix86_hard_regno_mode_ok):
9725 Do not use VALID_DFP_MODE_P.
9727 2021-07-15 Andrew MacLeod <amacleod@redhat.com>
9729 * gimple-range-fold.cc (adjust_pointer_diff_expr): Use
9731 (fold_using_range::fold_stmt): Ditto.
9732 (fold_using_range::range_of_range_op): Ditto.
9733 (fold_using_range::range_of_phi): Ditto.
9734 (fold_using_range::range_of_call): Ditto.
9735 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
9736 (fold_using_range::range_of_builtin_call): Ditto.
9737 (fold_using_range::range_of_cond_expr): Ditto.
9738 * gimple-range-fold.h (gimple_range_type): New.
9740 2021-07-15 Martin Sebor <msebor@redhat.com>
9743 * tree-ssa-strlen.c (handle_assign): New function.
9744 (maybe_warn_overflow): Add argument.
9745 (nonzero_bytes_for_type): New function.
9746 (count_nonzero_bytes): Handle more tree types. Call
9747 nonzero_bytes_for_tye.
9748 (count_nonzero_bytes): Handle types.
9749 (handle_store): Handle stores from function calls.
9750 (strlen_check_and_optimize_call): Move code to handle_assign. Call
9751 it for assignments from function calls.
9753 2021-07-15 David Malcolm <dmalcolm@redhat.com>
9758 * doc/invoke.texi: Add -Wanalyzer-use-of-uninitialized-value.
9760 2021-07-15 David Malcolm <dmalcolm@redhat.com>
9762 * doc/invoke.texi (-fdump-analyzer-exploded-paths): New.
9764 2021-07-15 Martin Sebor <msebor@redhat.com>
9768 * fold-const.c (operand_compare::operand_equal_p): Handle OEP_DECL_NAME.
9769 (operand_compare::verify_hash_value): Same.
9770 * tree-core.h (OEP_DECL_NAME): New.
9772 2021-07-15 Martin Jambor <mjambor@suse.cz>
9774 * profile-count.h (profile_count::value): Change the return type to
9776 * gimple-pretty-print.c (dump_gimple_bb_header): Adjust print
9778 * tree-cfg.c (dump_function_to_file): Likewise.
9780 2021-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
9783 * config/rs6000/rs6000-p8swap.c (has_part_mult): New.
9784 (rs6000_analyze_swaps): Insns containing a subreg of a mult are
9787 2021-07-15 Richard Biener <rguenther@suse.de>
9789 * tree-vectorizer.h (vect_gen_while): Match up with
9791 * tree-vect-stmts.c (vect_gen_while): Adjust API to that
9792 of vect_gen_while_not.
9793 (vect_gen_while_not): Adjust.
9794 * tree-vect-loop-manip.c (vect_set_loop_controls_directly): Likewise.
9796 2021-07-15 Aldy Hernandez <aldyh@redhat.com>
9798 * gimple-range-cache.cc (non_null_ref::adjust_range): New.
9799 (ranger_cache::range_of_def): Call adjust_range.
9800 (ranger_cache::entry_range): Same.
9801 * gimple-range-cache.h (non_null_ref::adjust_range): New.
9802 * gimple-range.cc (gimple_ranger::range_of_expr): Call
9804 (gimple_ranger::range_on_entry): Same.
9806 2021-07-15 Tamar Christina <tamar.christina@arm.com>
9809 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9811 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
9813 2021-07-15 Tamar Christina <tamar.christina@arm.com>
9816 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9818 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
9819 (sdot_prod, udot_prod): ...These.
9820 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
9821 (aarch64_<sur>dot<vsi2qi>): Rename to...
9822 (<sur>dot_prod<vsi2qi>): ...This.
9823 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
9826 2021-07-15 Jakub Jelinek <jakub@redhat.com>
9828 PR middle-end/101437
9829 * gimplify.c (gimplify_expr): Throw away volatile reads from empty
9830 types even if they have non-BLKmode TYPE_MODE.
9832 2021-07-15 Richard Biener <rguenther@suse.de>
9835 * gcc.c (process_command): Process -gtoggle like process_options
9836 would after parsing options.
9838 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
9840 * cfgexpand.c (expand_asm_loc): Adjust.
9841 (expand_asm_stmt): Likewise.
9842 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Likewise.
9843 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
9844 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
9845 * config/avr/avr.c (avr_md_asm_adjust): Likewise.
9846 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
9847 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
9848 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
9849 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
9850 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
9851 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
9852 * config/s390/s390.c (s390_md_asm_adjust): Likewise.
9853 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
9854 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
9855 * doc/tm.texi: Regenerate.
9856 * target.def: Add location argument to md_asm_adjust.
9858 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
9860 * tree-diagnostic.c (diagnostic_report_current_function): Use the
9861 diagnostic's location, not input_location.
9863 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
9865 * cfgexpand.c (tree_conflicts_with_clobbers_p): Pass location to
9867 (expand_asm_stmt): Likewise.
9869 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
9871 * config/rs6000/rs6000.c (adjacent_mem_locations): Return the lower
9872 addressed memory rtx, if any.
9873 (rs6000_split_multireg_move): Fix code formatting.
9874 Handle MMA build built-ins with operands in adjacent memory locations.
9876 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
9878 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Move to later
9881 2021-07-14 Jason Merrill <jason@redhat.com>
9883 * sel-sched-ir.h (get_all_loop_exits): Use auto_vec.
9885 2021-07-14 Jason Merrill <jason@redhat.com>
9887 * doc/invoke.texi: -fdelete-dead-exceptions is on by default for
9890 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9892 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
9893 Remove erroneous line.
9895 2021-07-14 Andrew MacLeod <amacleod@redhat.com>
9897 * params.opt (param_evrp_mode): Change default.
9899 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9901 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
9902 (sdot_prod, udot_prod): ...These.
9903 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
9904 (aarch64_<sur>dot<vsi2qi>): Rename to...
9905 (<sur>dot_prod<vsi2qi>): ...This.
9906 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
9909 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9911 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
9913 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9915 * doc/sourcebuild.texi (arm_v8_2a_i8mm_neon_hw): Document.
9917 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9919 * config/arm/neon.md (usdot_prod<vsi2qi>): New.
9921 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9923 * config/aarch64/aarch64-simd.md (aarch64_usdot<vsi2qi>): Rename to...
9924 (usdot_prod<vsi2qi>): ... This.
9925 * config/aarch64/aarch64-simd-builtins.def (usdot): Rename to...
9926 (usdot_prod): ...This.
9927 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Likewise.
9928 * config/aarch64/aarch64-sve.md (@aarch64_<sur>dot_prod<vsi2qi>):
9930 (@<sur>dot_prod<vsi2qi>): ...This.
9931 * config/aarch64/aarch64-sve-builtins-base.cc
9932 (svusdot_impl::expand): Use it.
9934 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9936 * optabs.def (usdot_prod_optab): New.
9937 * doc/md.texi: Document it and clarify other dot prod optabs.
9938 * optabs-tree.h (enum optab_subtype): Add optab_vector_mixed_sign.
9939 * optabs-tree.c (optab_for_tree_code): Support usdot_prod_optab.
9940 * optabs.c (expand_widen_pattern_expr): Likewise.
9941 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
9942 * tree-vect-loop.c (vectorizable_reduction): Query dot-product kind.
9943 * tree-vect-patterns.c (vect_supportable_direct_optab_p): Take optional
9945 (vect_widened_op_tree): Optionally ignore
9947 (vect_recog_dot_prod_pattern): Support usdot_prod_optab.
9949 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
9952 * config/i386/driver-i386.c (host_detect_local_cpu): Check
9953 "arch [32|64]" and "tune [32|64]" for 32-bit and 64-bit codegen.
9954 Enable UINTR only for 64-bit codegen.
9955 * config/i386/i386-options.c
9956 (ix86_option_override_internal::DEF_PTA): Skip PTA_UINTR if not
9958 * config/i386/i386.h (ARCH_ARG): New.
9959 (CC1_CPU_SPEC): Pass "[arch|tune] 32" for 32-bit codegen and
9960 "[arch|tune] 64" for 64-bit codegen.
9962 2021-07-14 Richard Biener <rguenther@suse.de>
9964 PR tree-optimization/101445
9965 * tree-vect-stmts.c (vectorizable_load): Do the gap adjustment
9966 of the IV in the correct direction for negative stride
9969 2021-07-14 Jakub Jelinek <jakub@redhat.com>
9972 * godump.c (godump_str_hash): New type.
9973 (godump_container::pot_dummy_types): Use string_hash instead of
9974 ptr_hash in the hash_set.
9976 2021-07-14 Richard Biener <rguenther@suse.de>
9978 * tree-vect-loop.c (vect_find_reusable_accumulator): Handle
9979 vector types where the old vector type has a multiple of
9980 the new vector type elements.
9981 (vect_create_partial_epilog): New function, split out from...
9982 (vect_create_epilog_for_reduction): ... here.
9983 (vect_transform_cycle_phi): Reduce the re-used accumulator
9984 to the new vector type.
9986 2021-07-14 Alexandre Oliva <oliva@adacore.com>
9988 * tree-ssa-alias.c (attr_fnspec::verify): Fix index in
9989 non-'t'-sized arg check.
9991 2021-07-14 Alexandre Oliva <oliva@adacore.com>
9993 * tree-cfg.c (cleanup_dead_labels_eh): Update
9994 post_landing_pad label upon change of landing pad block's
9996 (cleanup_dead_labels): Check that a removed label is not that
9999 2021-07-13 Jonathan Wright <jonathan.wright@arm.com>
10001 * combine.c (combine_simplify_rtx): Add vec_select -> subreg
10003 * config/aarch64/aarch64.md (*zero_extend<SHORT:mode><GPI:mode>2_aarch64):
10004 Add Neon to general purpose register case for zero-extend
10006 * config/arm/vfp.md (*arm_movsi_vfp): Remove "*" from *t -> r
10007 case to prevent some cases opting to go through memory.
10008 * cse.c (fold_rtx): Add vec_select -> subreg simplification.
10009 * rtl.c (rtvec_series_p): Define predicate to determine
10010 whether a vector contains a linear series of integers.
10011 * rtl.h (rtvec_series_p): Define.
10012 * rtlanal.c (vec_series_lowpart_p): Define predicate to
10013 determine if a vector selection is equivalent to the low part
10015 * rtlanal.h (vec_series_lowpart_p): Define.
10016 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
10017 Add vec_select -> subreg simplification.
10019 2021-07-13 Paul A. Clarke <pc@us.ibm.com>
10021 * config/rs6000/smmintrin.h (_mm_testz_si128, _mm_testc_si128,
10022 _mm_testnzc_si128, _mm_test_all_ones, _mm_test_all_zeros,
10023 _mm_test_mix_ones_zeros): New.
10025 2021-07-13 Roger Sayle <roger@nextmovesoftware.com>
10026 Richard Biener <rguenther@suse.de>
10028 * gimple.c (gimple_could_trap_p_1): Make S argument a
10029 "const gimple*". Preserve constness in call to
10030 gimple_asm_volatile_p.
10031 (gimple_could_trap_p): Make S argument a "const gimple*".
10032 * gimple.h (gimple_could_trap_p_1, gimple_could_trap_p):
10033 Update function prototypes.
10035 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10037 * tree-vectorizer.h (vect_reusable_accumulator): New structure.
10038 (_loop_vec_info::main_loop_edge): New field.
10039 (_loop_vec_info::skip_main_loop_edge): Likewise.
10040 (_loop_vec_info::skip_this_loop_edge): Likewise.
10041 (_loop_vec_info::reusable_accumulators): Likewise.
10042 (_stmt_vec_info::reduc_scalar_results): Likewise.
10043 (_stmt_vec_info::reused_accumulator): Likewise.
10044 (vect_get_main_loop_result): Declare.
10045 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize
10046 reduc_scalar_inputs.
10047 (vec_info::free_stmt_vec_info): Free reduc_scalar_inputs.
10048 * tree-vect-loop-manip.c (vect_get_main_loop_result): New function.
10049 (vect_do_peeling): Fill an epilogue loop's main_loop_edge,
10050 skip_main_loop_edge and skip_this_loop_edge fields.
10051 * tree-vect-loop.c (INCLUDE_ALGORITHM): Define.
10052 (vect_emit_reduction_init_stmts): New function.
10053 (get_initial_def_for_reduction): Use it.
10054 (get_initial_defs_for_reduction): Likewise. Change the vinfo
10055 parameter to a loop_vec_info.
10056 (vect_create_epilog_for_reduction): Store the scalar results
10057 in the reduc_info. If an epilogue loop is reusing an accumulator
10058 from the main loop, and if the epilogue loop can also be skipped,
10059 try to place the reduction code in the join block. Record
10060 accumulators that could potentially be reused by epilogue loops.
10061 (vect_transform_cycle_phi): When vectorizing epilogue loops,
10062 try to reuse accumulators from the main loop. Record the initial
10063 value in reduc_info for non-SLP reductions too.
10065 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10067 * tree-vect-loop.c (get_initial_def_for_reduction): Remove
10068 adjustment handling. Take the neutral value as an argument,
10069 in place of the code argument.
10070 (vect_transform_cycle_phi): Update accordingly. Handle the
10071 initial values of cond reductions separately from code reductions.
10072 Choose the adjustment here rather than in
10073 get_initial_def_for_reduction. Sink the splat of vec_initial_def.
10075 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10077 * tree-vect-loop.c (neutral_op_for_slp_reduction): Replace with...
10078 (neutral_op_for_reduction): ...this, providing a more general
10080 (vect_create_epilog_for_reduction): Update accordingly.
10081 (vectorizable_reduction): Likewise.
10082 (vect_transform_cycle_phi): Likewise.
10084 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10086 * tree-vect-loop.c (get_initial_def_for_reduction): Take the
10087 reduc_info instead of the original stmt_vec_info.
10088 (vect_transform_cycle_phi): Update accordingly.
10090 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10092 * tree-vect-loop.c (get_initial_defs_for_reduction): Take the
10093 reduc_info as an additional parameter.
10094 (vect_transform_cycle_phi): Update accordingly.
10096 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10098 * tree-vectorizer.h: Include tree-ssa-operands.h.
10099 (vect_phi_initial_value): New function.
10100 * tree-vect-loop.c (neutral_op_for_slp_reduction): Use it.
10101 (get_initial_defs_for_reduction, info_for_reduction): Likewise.
10102 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
10103 (vect_transform_cycle_phi, vectorizable_induction): Likewise.
10105 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10107 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
10108 the phi results to vectype after creating them. Remove later
10109 conversion code that thus becomes redundant.
10111 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10113 * tree-vect-loop.c (vect_create_epilog_for_reduction): Replace
10114 the new_phis vector with a reduc_inputs vector. Combine handling
10115 of reduction chains and ncopies > 1.
10117 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10119 * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
10120 scalar_results to group_size elements after reducing down from
10121 N*group_size elements. Construct an array_slice of the live-out
10122 stmts and assert that there is one stmt per scalar result.
10124 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10126 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
10127 nested_in_vect_loop and use double_reduc everywhere. Remove dead
10128 assignment to "loop".
10130 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10132 * internal-fn.c (vectorized_internal_fn_supported_p): Handle
10133 vector types first. For scalar types, consider both the preferred
10134 vector mode and the alternative vector modes.
10135 * optabs-query.c (can_vec_mask_load_store_p): Use the same
10136 structure as above, in particular using related_vector_mode
10137 for modes provided by autovectorize_vector_modes.
10139 2021-07-13 Jakub Jelinek <jakub@redhat.com>
10140 Richard Biener <rguenther@suse.de>
10142 PR tree-optimization/101419
10143 * tree-pass.h (PROP_objsz): Define.
10144 (make_pass_early_object_sizes): Declare.
10145 * passes.def (pass_all_early_optimizations): Rename pass_object_sizes
10146 there to pass_early_object_sizes, drop parameter.
10147 (pass_all_optimizations): Move pass_object_sizes right after pass_ccp,
10148 drop parameter, move pass_post_ipa_warn right after that.
10149 * tree-object-size.c (pass_object_sizes::execute): Rename to...
10150 (object_sizes_execute): ... this. Add insert_min_max_p argument.
10151 (pass_data_object_sizes): Move after object_sizes_execute.
10152 (pass_object_sizes): Likewise. In execute method call
10153 object_sizes_execute, drop set_pass_param method and insert_min_max_p
10154 non-static data member and its initializer in the ctor.
10155 (pass_data_early_object_sizes, pass_early_object_sizes,
10156 make_pass_early_object_sizes): New.
10157 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Use
10158 (cfun->curr_properties & PROP_objsz) instead of cfun->after_inlining.
10160 2021-07-13 Kito Cheng <kito.cheng@sifive.com>
10163 * config/riscv/constraints.md ("S"): Update description and remove
10165 * doc/md.texi (Machine Constraints): Document the 'S' constraints
10168 2021-07-13 Richard Biener <rguenther@suse.de>
10171 2021-07-12 Richard Biener <rguenther@suse.de>
10173 * tree-vect-slp.c (vect_slp_region): Show the number of
10174 SLP graph entries in the optimization message.
10176 2021-07-13 Michael Meissner <meissner@linux.ibm.com>
10178 * config/rs6000/altivec.md (xxspltiw_v4sf): Change local variable
10180 * config/rs6000/rs6000-protos.h (rs6000_const_f32_to_i32): Change
10181 return type to long.
10182 * config/rs6000/rs6000.c (rs6000_const_f32_to_i32): Change return
10185 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
10187 * gimple-range-fold.cc (fold_using_range::range_of_builtin_ubsan_call):
10188 Query relation between the 2 operands and use it.
10190 2021-07-12 Sergei Trofimovich <siarheit@google.com>
10192 * doc/cfg.texi: Fix s/ei_safe_safe/ei_safe_edge/ typo.
10194 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
10197 * config/i386/predicates.md (vec_setm_sse41_operand):
10198 Rename from vec_setm_operand.
10199 (vec_setm_avx2_operand): New predicate.
10200 * config/i386/sse.md (vec_set<V_128:mode>): Use V_128 mode iterator.
10201 Use vec_setm_sse41_operand as operand 2 predicate.
10202 (vec_set<V_256_512:mode): New expander.
10203 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_sse41_operand
10204 as operand 2 predicate.
10206 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
10208 PR tree-optimization/101335
10209 * range-op.cc (operator_cast::lhs_op1_relation): Delete.
10211 2021-07-12 Andrew Pinski <apinski@marvell.com>
10213 * tree-ssa-phiopt.c (match_simplify_replacement): Move
10214 insert of the sequence before the movement of the
10215 statement. Check if to see if the statement is used
10216 outside of the original phi to see if we should move it.
10218 2021-07-12 Richard Biener <rguenther@suse.de>
10220 * dump-context.h (debug_dump_context::debug_dump_context):
10221 Add FILE * parameter defaulted to stderr.
10222 * dumpfile.c (debug_dump_context::debug_dump_context): Adjust.
10223 * tree-vect-slp.c (dot_slp_tree): New functions.
10225 2021-07-12 Richard Biener <rguenther@suse.de>
10227 PR tree-optimization/101373
10228 * tree-ssa-pre.c (prune_clobbered_mems): Also prune trapping
10229 references when the BB may not return.
10230 (compute_avail): Pass in the function we're working on and
10231 replace cfun references with it. Externally throwing
10232 const calls also possibly terminate the function.
10233 (pass_pre::execute): Pass down the function we're working on.
10234 * gcse.c (compute_hash_table_work): Externally throwing
10235 const/pure calls also need record_last_mem_set_info.
10236 * postreload-gcse.c (record_opr_changes): Looping or externally
10237 throwing const/pure calls also need record_last_mem_set_info.
10239 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
10241 * recog.c (memory_address_addr_space_p): Change the type to bool.
10242 Return true/false instead of 1/0.
10243 (offsettable_memref_p): Ditto.
10244 (offsettable_nonstrict_memref_p): Ditto.
10245 (offsettable_address_addr_space_p): Ditto.
10246 Change the type of addressp indirect function to bool.
10247 * recog.h (memory_address_addr_space_p): Change the type to bool.
10248 (strict_memory_address_addr_space_p): Ditto.
10249 (offsettable_memref_p): Ditto.
10250 (offsettable_nonstrict_memref_p): Ditto.
10251 (offsettable_address_addr_space_p): Ditto.
10252 * reload.c (maybe_memory_address_addr_space_p): Ditto.
10253 (strict_memory_address_addr_space_p): Change the type to bool.
10254 Return true/false instead of 1/0.
10255 (maybe_memory_address_addr_space_p): Change the type to bool.
10257 2021-07-12 Richard Biener <rguenther@suse.de>
10259 * tree-vect-slp.c (vect_slp_region): Show the number of
10260 SLP graph entries in the optimization message.
10262 2021-07-12 Richard Biener <rguenther@suse.de>
10264 PR tree-optimization/101394
10265 * tree-ssa-pre.c (do_pre_regular_insertion): Avoid inserting
10266 copies from abnormals for a full redundancy.
10268 2021-07-12 Richard Biener <rguenther@suse.de>
10270 PR middle-end/101423
10271 * gimple.c (gimple_could_trap_p_1): Internal function calls
10273 * tree-eh.c (tree_could_trap_p): Likewise.
10275 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
10278 * config/arm/arm_neon.h (vmul_n_u32): Replace call to builtin with
10280 (vmulq_n_u32): Likewise.
10281 (vmul_n_f32): Gate __a * __b on __FAST_MATH__.
10282 (vmulq_n_f32): Likewise.
10283 (vmul_n_f16): Likewise.
10284 (vmulq_n_f16): Likewise.
10286 2021-07-12 Martin Liska <mliska@suse.cz>
10288 PR sanitizer/101425
10289 * gcc.c (check_offload_target_name): Call
10290 candidates_list_and_hint only if we have a candidate.
10292 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
10295 * config/arm/neon.md (vec_init): Move to ...
10296 * config/arm/vec-common.md (vec_init): ... here.
10297 Change the pattern's mode to VDQX and gate it on VALID_MVE_MODE.
10299 2021-07-12 Roger Sayle <roger@nextmovesoftware.com>
10301 PR tree-optimization/101403
10302 * match.pd ((T)bswap(X)>>C): Correctly handle cases where
10303 signedness of the shift is not the same as the signedness of
10304 the type extension.
10306 2021-07-09 Roger Sayle <roger@nextmovesoftware.com>
10307 Uroš Bizjak <ubizjak@gmail.com>
10309 * config/i386/i386.md (*divmodsi4_const): Optimize SImode
10310 divmod of a constant numerator with new define_insn_and_split.
10312 2021-07-09 Iain Sandoe <iain@sandoe.co.uk>
10315 * config/i386/i386-expand.c (ix86_expand_call): If a call is
10316 to a non-local-binding, or local but to a public symbol, then
10317 assume that it might be indirected via the lazy symbol binder.
10318 Mark R10 and R10 as clobbered in that case.
10320 2021-07-09 Eric Botcazou <ebotcazou@adacore.com>
10323 * gcc.c (ASM_DEBUG_DWARF_OPTION): Set again to --gdwarf2 in
10324 the case where HAVE_AS_WORKING_DWARF_N_FLAG is not defined
10325 and HAVE_LD_BROKEN_PE_DWARF5 is defined.
10327 2021-07-09 Uroš Bizjak <ubizjak@gmail.com>
10329 * config/i386/i386.md (*udivmodsi4_pow2_zext_1): Limit the
10330 log2 range of operands[3] to [1,31].
10331 (*udivmodsi4_pow2_zext_2): Ditto. Correct insn RTX pattern.
10333 2021-07-09 Sergei Trofimovich <siarheit@google.com>
10335 * doc/md.texi: Don't split @smallexample in multiple @groups.
10337 2021-07-09 Sergei Trofimovich <siarheit@google.com>
10339 * doc/md.texi: Add missing 'see' word.
10341 2021-07-09 Andrew Pinski <apinski@marvell.com>
10343 * tree-ssa-phiopt.c (phiopt_early_allow): Change arguments
10344 to take sequence and gimple_match_op. Accept the case where
10345 op is a SSA_NAME and one statement in the sequence.
10346 Also allow constants.
10347 (gimple_simplify_phiopt): Always pass a sequence to resimplify.
10348 Update call to phiopt_early_allow. Discard the sequence if not
10351 2021-07-09 Xi Ruoyao <xry111@mengyan1223.wang>
10356 * config/mips/mips.c (mips_const_insns): Use MSA_SUPPORTED_MODE_P
10357 instead of ISA_HAS_MSA.
10358 (mips_expand_vec_unpack): Likewise.
10359 (mips_expand_vector_init): Likewise.
10361 2021-07-09 Kewen Lin <linkw@linux.ibm.com>
10363 * config/rs6000/vsx.md (mods_<mode>): Rename to...
10364 (mod<mode>3): ... this.
10365 (modu_<mode>): Rename to...
10366 (umod<mode>3): ... this.
10367 * config/rs6000/rs6000-builtin.def (MODS_V2DI, MODS_V4SI, MODU_V2DI,
10368 MODU_V4SI): Adjust.
10370 2021-07-08 Jeff Law <jeffreyalaw@gmail.com>
10372 * config/h8300/shiftrotate.md (variable shifts): Expose condition
10373 code handling for the test before the loop.
10375 2021-07-08 Martin Jambor <mjambor@suse.cz>
10378 * ipa-sra.c (class isra_call_summary): New member
10379 m_before_any_store, initialize it in the constructor.
10380 (isra_call_summary::dump): Dump the new field.
10381 (ipa_sra_call_summaries::duplicate): Copy it.
10382 (process_scan_results): Set it.
10383 (isra_write_edge_summary): Stream it.
10384 (isra_read_edge_summary): Likewise.
10385 (param_splitting_across_edge): Only override
10386 safe_to_import_accesses if m_before_any_store is set.
10388 2021-07-08 Martin Sebor <msebor@redhat.com>
10390 PR bootstrap/101374
10391 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref):
10392 Use Object Size Type 0 instead of 1.
10394 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
10396 * tree-vect-loop.c (vectorizable_reduction): Remove always-true
10399 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
10401 * match.pd: Simplify an extend-operate-truncate sequence involving
10404 2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
10405 Richard Biener <rguenther@suse.de>
10407 PR tree-optimization/40210
10408 * match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
10409 (x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
10410 when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
10412 2021-07-08 Uroš Bizjak <ubizjak@gmail.com>
10415 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
10417 * config/i386/mmx.md (V_32): New mode iterator.
10418 (mov<V_32:mode>): Use V_32 mode iterator.
10419 (*mov<V_32:mode>_internal): Ditto.
10420 (*push<V_32:mode>2_rex64): Ditto.
10421 (*push<V_32:mode>2): Ditto.
10422 (movmisalign<V_32:mode>): Ditto.
10423 (mmx_<any_shiftrt:insn>v1si3): New insn pattern.
10424 (sse4_1_<any_extend:code>v2qiv2hi2): Ditto.
10425 (vec_unpacks_lo_v4qi): New expander.
10426 (vec_unpacks_hi_v4qi): Ditto.
10427 (vec_unpacku_lo_v4qi): Ditto.
10428 (vec_unpacku_hi_v4qi): Ditto.
10429 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add V1SImode.
10430 (VALID_INT_MODE_P): Ditto.
10432 2021-07-08 Michael Meissner <meissner@linux.ibm.com>
10435 * config/rs6000/rs6000.md (udivti3): New insn.
10436 (divti3): New insn.
10437 (umodti3): New insn.
10438 (modti3): New insn.
10440 2021-07-07 Martin Sebor <msebor@redhat.com>
10442 PR tree-optimization/100137
10443 PR tree-optimization/99121
10444 PR tree-optimization/97027
10445 * builtins.c (access_ref::access_ref): Also set offmax.
10446 (access_ref::offset_in_range): Define new function.
10447 (access_ref::add_offset): Set offmax.
10448 (access_ref::inform_access): Handle access_none.
10449 (handle_mem_ref): Clear ostype.
10450 (compute_objsize_r): Handle ASSERT_EXPR.
10451 * builtins.h (struct access_ref): Add offmax member.
10452 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Use
10453 compute_objsize() and simplify.
10455 2021-07-07 Peter Bergner <bergner@linux.ibm.com>
10457 * config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
10458 and VSX_BUILTIN_STXVP.
10460 2021-07-07 Martin Sebor <msebor@redhat.com>
10463 * config/aarch64/aarch64.c (aarch64_simd_lane_bounds): Remove
10464 a stray %K from error_at() missed in r12-2088.
10466 2021-07-07 Richard Biener <rguenther@suse.de>
10468 PR tree-optimization/99728
10469 * tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
10471 (mem_refs_may_alias_p): Add assert we handled aggregate
10473 (sm_seq_valid_bb): Give up when running into aggregate copies.
10474 (ref_indep_loop_p): Handle aggregate copies as never
10475 being invariant themselves but allow other refs to be
10476 disambiguated against them.
10477 (can_sm_ref_p): Do not try to apply store-motion to aggregate
10480 2021-07-06 Indu Bhagat <indu.bhagat@oracle.com>
10483 * dwarf2ctf.c (ctf_get_AT_data_member_location): Multiply by 8 to get
10486 2021-07-06 Martin Sebor <msebor@redhat.com>
10488 * gimple-pretty-print.c (percent_G_format): Remove.
10489 * tree-diagnostic.c (default_tree_printer): Remove calls.
10490 * tree-pretty-print.c (percent_K_format): Remove.
10491 * tree-pretty-print.h (percent_K_format): Remove.
10493 2021-07-06 Martin Sebor <msebor@redhat.com>
10495 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_builtin):
10496 Remove %K and use error_at.
10497 (aarch64_expand_fcmla_builtin): Same.
10498 (aarch64_expand_builtin_tme): Same.
10499 (aarch64_expand_builtin_memtag): Same.
10500 * config/arm/arm-builtins.c (arm_expand_acle_builtin): Same.
10501 (arm_expand_builtin): Same.
10502 * config/arm/arm.c (bounds_check): Same.
10504 2021-07-06 Martin Sebor <msebor@redhat.com>
10506 * builtins.c (warn_string_no_nul): Remove %G.
10507 (maybe_warn_for_bound): Same.
10508 (warn_for_access): Same.
10509 (check_access): Same.
10510 (check_strncat_sizes): Same.
10511 (expand_builtin_strncat): Same.
10512 (expand_builtin_strncmp): Same.
10513 (expand_builtin): Same.
10514 (expand_builtin_object_size): Same.
10515 (warn_dealloc_offset): Same.
10516 (maybe_emit_free_warning): Same.
10517 * calls.c (maybe_warn_alloc_args_overflow): Same.
10518 (maybe_warn_nonstring_arg): Same.
10519 (maybe_warn_rdwr_sizes): Same.
10520 * expr.c (expand_expr_real_1): Remove %K.
10521 * gimple-fold.c (gimple_fold_builtin_strncpy): Remove %G.
10522 (gimple_fold_builtin_strncat): Same.
10523 * gimple-ssa-sprintf.c (format_directive): Same.
10524 (handle_printf_call): Same.
10525 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Same.
10526 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
10527 (maybe_diag_access_bounds): Same. Call gimple_location.
10528 (check_bounds_or_overlap): Same.
10529 * trans-mem.c (ipa_tm_scan_irr_block): Remove %K. Simplify.
10530 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Remove %G.
10531 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
10532 (maybe_diag_stxncpy_trunc): Same.
10533 (handle_builtin_stxncpy_strncat): Same.
10534 (maybe_warn_pointless_strcmp): Same.
10535 * tree-ssa-uninit.c (maybe_warn_operand): Same.
10537 2021-07-06 Uroš Bizjak <ubizjak@gmail.com>
10540 * config/i386/predicates.md (vec_setm_operand): Enable
10541 register_operand for TARGET_SSE4_1.
10542 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_operand
10543 as operand 2 predicate. Call ix86_expand_vector_set_var
10544 for non-constant index operand.
10545 (vec_setv4qi): Use vec_setm_mmx_operand as operand 2 predicate.
10546 Call ix86_expand_vector_set_var for non-constant index operand.
10548 2021-07-06 Jeff Law <jeffreyalaw@gmail.com>
10550 * config/h8300/jumpcall.md (*branch): When possible, generate
10551 the comparison in CCZN mode.
10552 * config/h8300/predicates.md (simple_memory_operand): Reject all
10553 auto-increment addressing modes.
10555 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
10557 PR bootstrap/100246
10558 * config/i386/i386.h (struct stringop_algs): Define a CTOR for
10561 2021-07-06 Richard Biener <rguenther@suse.de>
10563 * doc/md.texi (vec_fmaddsub<mode>4): Document.
10564 (vec_fmsubadd<mode>4): Likewise.
10565 * optabs.def (vec_fmaddsub$a4): Add.
10566 (vec_fmsubadd$a4): Likewise.
10567 * internal-fn.def (IFN_VEC_FMADDSUB): Add.
10568 (IFN_VEC_FMSUBADD): Likewise.
10569 * tree-vect-slp-patterns.c (addsub_pattern::recognize):
10570 Refactor to handle IFN_VEC_FMADDSUB and IFN_VEC_FMSUBADD.
10571 (addsub_pattern::build): Likewise.
10572 * tree-vect-slp.c (vect_optimize_slp): CFN_VEC_FMADDSUB
10573 and CFN_VEC_FMSUBADD are not transparent for permutes.
10574 * config/i386/sse.md (vec_fmaddsub<mode>4): New expander.
10575 (vec_fmsubadd<mode>4): Likewise.
10577 2021-07-06 Richard Biener <rguenther@suse.de>
10579 * doc/invoke.texi (fmove-loop-stores): Document.
10580 * common.opt (fmove-loop-stores): New option.
10581 * opts.c (default_options_table): Enable -fmove-loop-stores
10582 at -O1 but not -Og.
10583 * tree-ssa-loop-im.c (pass_lim::execute): Pass
10584 flag_move_loop_stores instead of true to
10585 loop_invariant_motion_in_fun.
10587 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
10589 * doc/install.texi: Document --with-dsymutil.
10591 2021-07-06 Andrew Pinski <apinski@marvell.com>
10593 PR tree-optimization/101256
10594 * dbgcnt.def (phiopt_edge_range): New counter.
10595 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
10596 Check to make sure the new name is defined in the same
10597 bb as the conditional before duplicating range info.
10598 Also add debug counter.
10600 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
10602 PR rtl-optimization/100328
10603 * config/i386/i386-options.c (ix86_option_override_internal):
10604 Set param_ira_consider_dup_in_all_alts to 0.
10606 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
10608 PR rtl-optimization/100328
10609 * doc/invoke.texi (ira-consider-dup-in-all-alts): Document new
10611 * ira.c (ira_get_dup_out_num): Adjust as parameter
10612 param_ira_consider_dup_in_all_alts.
10613 * params.opt (ira-consider-dup-in-all-alts): New.
10614 * ira-conflicts.c (process_regs_for_copy): Add one parameter
10615 single_input_op_has_cstr_p.
10616 (get_freq_for_shuffle_copy): New function.
10617 (add_insn_allocno_copies): Adjust as single_input_op_has_cstr_p.
10618 * ira-int.h (ira_get_dup_out_num): Add one bool parameter.
10620 2021-07-05 Jeff Law <jeffreyalaw@gmail.com>
10622 * config/h8300/shiftrotate.md (shift-by-variable patterns): Update to
10623 generate condition code aware RTL directly.
10625 2021-07-05 Andrew Pinski <apinski@marvell.com>
10627 PR tree-optimization/101039
10628 * match.pd (A CMP 0 ? A : -A): New patterns.
10629 * tree-ssa-phiopt.c (abs_replacement): Delete function.
10630 (tree_ssa_phiopt_worker): Don't call abs_replacement.
10631 Update comment about abs_replacement.
10633 2021-07-05 Andrew Pinski <apinski@marvell.com>
10635 * tree-ssa-phiopt.c (gimple_simplify_phiopt):
10636 If "A ? B : C" fails to simplify, try "(!A) ? C : B".
10638 2021-07-05 Andrew Pinski <apinski@marvell.com>
10640 * tree-ssa-phiopt.c (match_simplify_replacement):
10641 Add early_p argument. Call gimple_simplify_phiopt
10642 instead of gimple_simplify.
10643 (tree_ssa_phiopt_worker): Update call to
10644 match_simplify_replacement and allow unconditionally.
10645 (phiopt_early_allow): New function.
10646 (gimple_simplify_phiopt): New function.
10648 2021-07-05 Andrew Pinski <apinski@marvell.com>
10650 PR middle-end/101237
10651 * fold-const.c (negate_expr_p): Remove call to element_mode
10652 and TREE_MODE/TREE_TYPE when calling HONOR_SIGNED_ZEROS,
10653 HONOR_SIGN_DEPENDENT_ROUNDING, and HONOR_SNANS.
10654 (fold_negate_expr_1): Likewise.
10655 (const_unop): Likewise.
10656 (fold_cond_expr_with_comparison): Likewise.
10657 (fold_binary_loc): Likewise.
10658 (fold_ternary_loc): Likewise.
10659 (tree_call_nonnegative_warnv_p): Likewise.
10660 * match.pd (-(A + B) -> (-B) - A): Likewise.
10662 2021-07-05 Iain Sandoe <iain@sandoe.co.uk>
10664 * configure.ac: Handle --with-dsymutil in the same way as we
10665 do for the assembler and linker. (DEFAULT_DSYMUTIL): New.
10666 Extract the type and version for the dsymutil configured or
10667 found by the default searches.
10668 * config.in: Regenerated.
10669 * configure: Regenerated.
10670 * collect2.c (do_dsymutil): Handle locating dsymutil in the
10671 same way as for the assembler and linker.
10672 * config/darwin.h (DSYMUTIL): Delete.
10673 * gcc.c: Report a configured dsymutil correctly.
10674 * exec-tool.in: Allow for dsymutil.
10676 2021-07-05 Uroš Bizjak <ubizjak@gmail.com>
10678 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
10679 Handle V4QI and V2HI modes.
10680 (expand_vec_perm_blend): Allow 4-byte vector modes with TARGET_SSE4_1.
10681 Handle V4QI mode. Emit mmx_pblendvb32 for 4-byte modes.
10682 (expand_vec_perm_pshufb): Rewrite to use switch statemets.
10683 Handle 4-byte dual operands with TARGET_XOP and single operands
10684 with TARGET_SSSE3. Emit mmx_ppermv32 for TARGET_XOP and
10685 mmx_pshufbv4qi3 for TARGET_SSSE3.
10686 (expand_vec_perm_pblendv): Allow 4-byte vector modes with TARGET_SSE4_1.
10687 (expand_vec_perm_interleave2): Allow 4-byte vector modes.
10688 (expand_vec_perm_pshufb2): Allow 4-byte vector modes with TARGET_SSSE3.
10689 (expand_vec_perm_even_odd_1): Handle V4QI mode.
10690 (expand_vec_perm_broadcast_1): Handle V4QI mode.
10691 (ix86_vectorize_vec_perm_const): Handle V4QI mode.
10692 * config/i386/mmx.md (mmx_ppermv32): New insn pattern.
10693 (mmx_pshufbv4qi3): Ditto.
10694 (*mmx_pblendw32): Ditto.
10695 (*mmx_pblendw64): Rename from *mmx_pblendw.
10696 (mmx_punpckhbw_low): New insn_and_split pattern.
10697 (mmx_punpcklbw_low): Ditto.
10699 2021-07-05 Richard Biener <rguenther@suse.de>
10701 * tree-vect-loop-manip.c (vect_loop_versioning): Do not
10702 set LOOP_C_INFINITE on the vectorized loop.
10704 2021-07-05 Richard Biener <rguenther@suse.de>
10706 PR middle-end/101291
10707 * cfgloopmanip.c (loop_version): Set the loop copy of the
10708 versioned loop to the new loop.
10710 2021-07-04 Iain Sandoe <iain@sandoe.co.uk>
10713 * config.gcc: Ensure that Darwin biarch definitions are
10714 added before i386.h.
10715 * config/i386/darwin.h (TARGET_64BIT): Remove.
10716 (PR80556_WORKAROUND): New.
10717 (REAL_LIBGCC_SPEC): Amend to use PR80556_WORKAROUND.
10718 (DARWIN_SUBARCH_SPEC): New.
10719 * config/i386/darwin32-biarch.h (TARGET_64BIT_DEFAULT,
10720 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
10721 (REAL_LIBGCC_SPEC): Remove.
10722 * config/i386/darwin64-biarch.h (TARGET_64BIT_DEFAULT,
10723 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
10724 (REAL_LIBGCC_SPEC): Remove.
10726 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
10728 PR middle-end/101294
10729 * expr.c (store_constructor): Don't use vec_duplicate on vector.
10731 2021-07-02 Martin Sebor <msebor@redhat.com>
10733 PR middle-end/98871
10734 PR middle-end/98512
10735 * diagnostic.c (get_any_inlining_info): New.
10736 (update_effective_level_from_pragmas): Handle inlining context.
10737 (diagnostic_enabled): Same.
10738 (diagnostic_report_diagnostic): Same.
10739 * diagnostic.h (struct diagnostic_info): Add ctor.
10740 (struct diagnostic_context): Add new member.
10741 * tree-diagnostic.c (set_inlining_locations): New.
10742 (tree_diagnostics_defaults): Set new callback pointer.
10744 2021-07-02 Peter Bergner <bergner@linux.ibm.com>
10746 * config/rs6000/rs6000-builtin.def (BU_MMA_PAIR_LD, BU_MMA_PAIR_ST):
10748 (__builtin_vsx_lxvp, __builtin_vsx_stxvp): New built-ins.
10749 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Expand
10750 lxvp and stxvp built-ins.
10751 (mma_init_builtins): Handle lxvp and stxvp built-ins.
10752 (builtin_function_type): Likewise.
10753 * doc/extend.texi (__builtin_vsx_lxvp, __builtin_mma_stxvp): Document.
10755 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10757 * config/h8300/h8300-protos.h (compute_a_shift_cc): Accept
10758 additional argument for the code.
10759 * config/h8300/h8300.c (compute_a_shift_cc): Accept additional
10760 argument for the code. Just return if the ZN bits are useful or
10761 not rather than the old style CC_* enums.
10762 * config/h8300/shiftrotate.md (shiftqi_noscratch): Move before
10763 more generic shiftqi patterns.
10764 (shifthi_noscratch, shiftsi_noscratch): Similarly.
10765 (shiftqi_noscratch_set_flags): New pattern.
10766 (shifthi_noscratch_set_flags, shiftsi_noscratch_set_flags): Likewise.
10768 2021-07-02 Andrew MacLeod <amacleod@redhat.com>
10770 PR tree-optimization/101223
10771 * range-op.cc (build_lt): Add -1 for signed values.
10772 (built_gt): Subtract -1 for signed values.
10774 2021-07-02 David Faust <david.faust@oracle.com>
10776 * btfout.c (get_btf_kind): Support BTF_KIND_FLOAT.
10777 (btf_asm_type): Likewise.
10779 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10781 * config/h8300/h8300-protos.h (output_a_shift): Make first argument
10782 an array of rtx rather than a pointer to rtx. Add code argument.
10783 (compute_a_shift_length): Similarly.
10784 * config/h8300/h8300.c (h8300_shift_costs): Adjust now that the
10785 shift itself isn't an operand. Create dummy operand[0] to carry
10786 a mode and pass a suitable rtx code to compute_a_shift_length.
10787 (get_shift_alg): Adjust operand number of clobber in output templates.
10788 (output_a_shift): Make first argument an array of rtx rather than
10789 a pointer to rtx. Add code argument for the type of shift.
10790 Adjust now that the shift itself is no longer an operand.
10791 (compute_a_shift_length): Similarly.
10792 * config/h8300/shiftrotate.md (shiftqi, shifthi, shiftsi): Use an
10793 iterator rather than nshift_operator.
10794 (shiftqi_noscratch, shifthi_noscratch, shiftsi_noscratch): Likewise.
10795 (shiftqi_clobber_flags): Adjust to API changes in output_a_shift
10796 and compute_a_shift_length.
10797 (shiftqi_noscratch_clobber_flags): Likewise.
10798 (shifthi_noscratch_clobber_flags): Likewise.
10799 (shiftsi_noscratch_clobber_flags): Likewise.
10801 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
10804 * config/darwin.h (DSYMUTIL_SPEC): Do not try to run
10805 dsymutil for BTF/CTF.
10807 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
10810 * config/darwin.h (CTF_INFO_SECTION_NAME): Update the
10811 segment to include BTF.
10812 (BTF_INFO_SECTION_NAME): New.
10814 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10816 * config/m32r/m32r-protos.h (call_operand): Adjust return type.
10817 (small_data_operand, memreg_operand, small_insn_p): Likewise.
10818 * config/m32r/m32r.c (call_operand): Adjust return type.
10819 (small_data_operand, memreg_operand): Likewise.
10821 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10823 * config/frv/frv-protos.h (integer_register_operand): Adjust return
10825 (frv_load_operand, gpr_or_fpr_operand, gpr_no_subreg_operand): Likewise.
10826 (fpr_or_int6_operand, gpr_or_int_operand); Likewise.
10827 (gpr_or_int12_operand, gpr_or_int10_operand); Likewise.
10828 (move_source_operand, move_destination_operand): Likewise.
10829 (condexec_source_operand, condexec_dest_operand): Likewise.
10830 (lr_operand, gpr_or_memory_operand, fpr_or_memory_operand): Likewise.
10831 (reg_or_0_operand, fcc_operand, icc_operand, cc_operand): Likewise.
10832 (fcr_operand, icr_operand, cr_operand, call_operand): Likewise.
10833 (fpr_operand, even_reg_operand, odd_reg_operand): Likewise.
10834 (even_gpr_operand, odd_gpr_operand, quad_fpr_operand): Likewise.
10835 (even_fpr_operand, odd_fpr_operand): Likewise.
10836 (dbl_memory_one_insn_operand, dbl_memory_two_insn_operand): Likewise.
10837 (int12_operand, int6_operand, int5_operand, uint5_operand): Likewise.
10838 (uint4_operand, uint1_operand, int_2word_operand): Likewise
10839 (upper_int16_operand, uint16_operand, symbolic_operand): Likewise.
10840 (relational_operator, float_relational_operator): Likewise.
10841 (ccr_eqne_operator, minmax_operator): Likewise.
10842 (condexec_si_binary_operator, condexec_si_media_operator): Likewise.
10843 (condexec_si_divide_operator, condexec_si_unary_operator): Likewise.
10844 (condexec_sf_conv_operator, condexec_sf_add_operator): Likewise.
10845 (intop_compare_operator, acc_operand, even_acc_operand): Likewise.
10846 (quad_acc_operand, accg_operand): Likewise.
10848 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10850 * config/stormy16/stormy16-protos.h (xstormy16_below_100_symbol): Change
10851 return type to a bool.
10852 (nonimmediate_nonstack_operand): Likewise.
10853 (xstormy16_splittable_below100_operand): Likewise.
10854 * config/stormy16/stormy16.c (xstormy16_below_100_symbol): Fix
10856 (xstormy16_splittable_below100_operand): Likewise.
10858 2021-07-02 Richard Biener <rguenther@suse.de>
10860 PR tree-optimization/101293
10861 * tree-ssa-loop-im.c (mem_ref_hasher::equal): Compare MEM_REF bases
10862 with combined offsets.
10863 (gather_mem_refs_stmt): Hash MEM_REFs as if their offset were
10864 combined with the rest of the offset.
10866 2021-07-02 Eric Botcazou <ebotcazou@adacore.com>
10868 * config/i386/i386.c (asm_preferred_eh_data_format): Always use the
10869 PIC encodings for PE-COFF targets.
10871 2021-07-02 Jakub Jelinek <jakub@redhat.com>
10874 * config/i386/i386-expand.c (ix86_broadcast_from_integer_constant):
10875 Return nullptr for TImode inner mode.
10877 2021-07-02 Richard Biener <rguenther@suse.de>
10879 PR tree-optimization/101280
10880 PR tree-optimization/101173
10881 * gimple-loop-interchange.cc
10882 (tree_loop_interchange::valid_data_dependences): Properly
10883 guard all dependence checks with DDR_REVERSED_P or its
10886 2021-07-02 Hongyu Wang <hongyu.wang@intel.com>
10888 * config/i386/i386-expand.c (ix86_expand_builtin):
10889 Add branch to clear odata when ZF is set for asedecenc_expand
10890 and wideaesdecenc_expand.
10892 2021-07-02 Eugene Rozenfeld <erozen@microsoft.com>
10894 * config/i386/gcc-auto-profile: regenerate
10896 2021-07-02 liuhongt <hongtao.liu@intel.com>
10898 * config/i386/sse.md (trunc<mode><pmov_dst_4>2): Refined to ..
10899 (trunc<mode><pmov_dst_4_lower>2): this.
10901 2021-07-01 David Malcolm <dmalcolm@redhat.com>
10903 * diagnostic.h (diagnostic_context::m_file_cache): New field.
10904 * input.c (class fcache): Rename to...
10905 (class file_cache_slot): ...this, making most members private and
10906 prefixing fields with "m_".
10907 (file_cache_slot::get_file_path): New accessor.
10908 (file_cache_slot::get_use_count): New accessor.
10909 (file_cache_slot::missing_trailing_newline_p): New accessor.
10910 (file_cache_slot::inc_use_count): New.
10911 (fcache_buffer_size): Move to...
10912 (file_cache_slot::buffer_size): ...here.
10913 (fcache_line_record_size): Move to...
10914 (file_cache_slot::line_record_size): ...here.
10915 (fcache_tab): Delete, in favor of global_dc->m_file_cache.
10916 (fcache_tab_size): Move to file_cache::num_file_slots.
10917 (diagnostic_file_cache_init): Update for move of fcache_tab
10918 to global_dc->m_file_cache.
10919 (diagnostic_file_cache_fini): Likewise.
10920 (lookup_file_in_cache_tab): Convert to...
10921 (file_cache::lookup_file): ...this.
10922 (diagnostics_file_cache_forcibly_evict_file): Update for move of
10923 fcache_tab to global_dc->m_file_cache, moving most of
10924 implementation to...
10925 (file_cache::forcibly_evict_file): ...this new function and...
10926 (file_cache_slot::evict): ...this new function.
10927 (evicted_cache_tab_entry): Convert to...
10928 (file_cache::evicted_cache_tab_entry): ...this.
10929 (add_file_to_cache_tab): Convert to...
10930 (file_cache::add_file): ...this, moving bulk of implementation
10932 (file_cache_slot::create): ..this new function.
10933 (file_cache::file_cache): New.
10934 (file_cache::~file_cache): New.
10935 (lookup_or_add_file_to_cache_tab): Convert to...
10936 (file_cache::lookup_or_add_file): ..this new function.
10937 (fcache::fcache): Rename to...
10938 (file_cache_slot::file_cache_slot): ...this, adding "m_" prefixes
10940 (fcache::~fcache): Rename to...
10941 (file_cache_slot::~file_cache_slot): ...this, adding "m_" prefixes
10943 (needs_read): Convert to...
10944 (file_cache_slot::needs_read_p): ...this.
10945 (needs_grow): Convert to...
10946 (file_cache_slot::needs_grow_p): ...this.
10947 (maybe_grow): Convert to...
10948 (file_cache_slot::maybe_grow): ...this.
10949 (read_data): Convert to...
10950 (file_cache_slot::read_data): ...this.
10951 (maybe_read_data): Convert to...
10952 (file_cache_slot::maybe_read_data): ...this.
10953 (get_next_line): Convert to...
10954 (file_cache_slot::get_next_line): ...this.
10955 (goto_next_line): Convert to...
10956 (file_cache_slot::goto_next_line): ...this.
10957 (read_line_num): Convert to...
10958 (file_cache_slot::read_line_num): ...this.
10959 (location_get_source_line): Update for moving of globals to
10960 global_dc->m_file_cache.
10961 (location_missing_trailing_newline): Likewise.
10962 * input.h (class file_cache_slot): New forward decl.
10963 (class file_cache): New.
10965 2021-07-01 Michael Meissner <meissner@linux.ibm.com>
10967 * config/rs6000/rs6000.c (rs6000_maybe_emit_fp_cmove): Add IEEE
10968 128-bit floating point conditional move support.
10969 (have_compare_and_set_mask): Add IEEE 128-bit floating point
10971 * config/rs6000/rs6000.md (mov<mode>cc, IEEE128 iterator): New insn.
10972 (mov<mode>cc_p10, IEEE128 iterator): New insn.
10973 (mov<mode>cc_invert_p10, IEEE128 iterator): New insn.
10974 (fpmask<mode>, IEEE128 iterator): New insn.
10975 (xxsel<mode>, IEEE128 iterator): New insn.
10977 2021-07-01 Iain Sandoe <iain@sandoe.co.uk>
10980 * config/darwin.h (CTF_INFO_SECTION_NAME): New.
10982 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
10984 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
10986 * config/i386/i386-protos.h (ix86_expand_vector_init_duplicate):
10988 * config/i386/sse.md (INT_BROADCAST_MODE): New mode iterator.
10989 (vec_duplicate<mode>): New expander.
10991 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
10994 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
10996 (ix86_byte_broadcast): New function.
10997 (ix86_convert_const_wide_int_to_broadcast): Likewise.
10998 (ix86_expand_move): Convert CONST_WIDE_INT to broadcast if mode
10999 size is 16 bytes or bigger.
11000 (ix86_broadcast_from_integer_constant): New function.
11001 (ix86_expand_vector_move): Convert CONST_WIDE_INT and CONST_VECTOR
11002 to broadcast if mode size is 16 bytes or bigger.
11003 * config/i386/i386-protos.h (ix86_gen_scratch_sse_rtx): New
11005 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): New function.
11007 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
11009 * config/i386/predicates.md (ix86_endbr_immediate_operand):
11010 Return true/false instead of 1/0.
11011 (movq_parallel): Ditto.
11013 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
11015 * recog.c (general_operand): Return true/false instead of 1/0.
11016 (register_operand): Ditto.
11017 (immediate_operand): Ditto.
11018 (const_int_operand): Ditto.
11019 (const_scalar_int_operand): Ditto.
11020 (const_double_operand): Ditto.
11021 (push_operand): Ditto.
11022 (pop_operand): Ditto.
11023 (memory_operand): Ditto.
11024 (indirect_operand): Ditto.
11026 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
11028 * genpreds.c (write_predicate_subfunction):
11029 Change the type of written subfunction to bool.
11030 (write_one_predicate_function):
11031 Change the type of written function to bool.
11032 (write_tm_preds_h): Ditto.
11033 * recog.h (*insn_operand_predicate_fn): Change the type to bool.
11034 * recog.c (general_operand): Change the type to bool.
11035 (address_operand): Ditto.
11036 (register_operand): Ditto.
11037 (pmode_register_operand): Ditto.
11038 (scratch_operand): Ditto.
11039 (immediate_operand): Ditto.
11040 (const_int_operand): Ditto.
11041 (const_scalar_int_operand): Ditto.
11042 (const_double_operand): Ditto.
11043 (nonimmediate_operand): Ditto.
11044 (nonmemory_operand): Ditto.
11045 (push_operand): Ditto.
11046 (pop_operand): Ditto.
11047 (memory_operand): Ditto.
11048 (indirect_operand): Ditto.
11049 (ordered_comparison_operator): Ditto.
11050 (comparison_operator): Ditto.
11051 * config/i386/i386-expand.c (ix86_expand_sse_cmp):
11052 Change the type of indirect predicate function to bool.
11053 * config/rs6000/rs6000.c (easy_vector_constant):
11054 Change the type to bool.
11055 * config/mips/mips-protos.h (m16_based_address_p):
11056 Change the type of operand 3 to bool.
11058 2021-07-01 Richard Biener <rguenther@suse.de>
11060 PR tree-optimization/101280
11061 PR tree-optimization/101173
11062 * gimple-loop-interchange.cc
11063 (tree_loop_interchange::valid_data_dependences): Revert
11064 previous change and instead correctly handle DDR_REVERSED_P
11067 2021-07-01 Richard Biener <rguenther@suse.de>
11069 PR tree-optimization/101278
11070 * tree-ssa-dse.c (dse_classify_store): First check for
11071 uses, then ignore stmt for chaining purposes.
11073 2021-07-01 Richard Biener <rguenther@suse.de>
11075 PR tree-optimization/100778
11076 * tree-vect-slp.c (vect_schedule_slp_node): Do not place trapping
11077 vectorized ops ahead of their scalar BB.
11079 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
11082 * config/i386/i386.md (*nabs<dwi>2_doubleword):
11083 New insn_and_split pattern.
11084 (*nabs<dwi>2_1): Ditto.
11085 * config/i386/i386-features.c
11086 (general_scalar_chain::compute_convert_gain):
11087 Handle (NEG (ABS (...))) RTX. Rewrite src code
11088 scanner as switch statement.
11089 (general_scalar_chain::convert_insn):
11090 Handle (NEG (ABS (...))) RTX.
11091 (general_scalar_to_vector_candidate_p):
11092 Detect (NEG (ABS (...))) RTX. Reorder case statements
11093 for (AND (NOT (...) ...)) fallthrough.
11095 2021-07-01 Richard Biener <rguenther@suse.de>
11097 PR tree-optimization/101178
11098 * tree-vect-slp.c (slpg_vertex::materialize): Remove.
11099 (slpg::perm_in): Add.
11100 (slpg::get_perm_in): Remove.
11101 (slpg::get_perm_materialized): Add.
11102 (vect_optimize_slp): Handle VEC_PERM nodes more optimally
11103 during permute propagation and materialization.
11105 2021-07-01 Jakub Jelinek <jakub@redhat.com>
11108 * dwarf2out.c (loc_list_from_tree_1): Handle COMPOUND_LITERAL_EXPR.
11110 2021-07-01 Jakub Jelinek <jakub@redhat.com>
11112 PR middle-end/94366
11113 * omp-low.c (lower_rec_input_clauses): Rename is_fp_and_or to
11114 is_truth_op, set it for TRUTH_*IF_EXPR regardless of new_var's type,
11115 use boolean_type_node instead of integer_type_node as NE_EXPR type.
11116 (lower_reduction_clauses): Likewise.
11118 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
11120 * config/gcn/gcn.c: Include dwarf2.h.
11121 (gcn_addr_space_debug): New function.
11122 (TARGET_ADDR_SPACE_DEBUG): New hook.
11124 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
11126 * common/config/gcn/gcn-common.c
11127 (gcn_option_optimization_table): Change OPT_fomit_frame_pointer to -O3.
11128 * config/gcn/gcn.c (gcn_expand_prologue): Prefer the frame pointer
11130 (gcn_expand_prologue): Prefer the frame pointer when emitting CFI.
11131 (gcn_frame_pointer_rqd): New function.
11132 (TARGET_FRAME_POINTER_REQUIRED): New hook.
11134 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
11136 * config/gcn/gcn.c (move_callee_saved_registers): Emit CFI notes for
11137 prologue register saves.
11138 (gcn_debug_unwind_info): Use UI_DWARF2.
11139 (gcn_dwarf_register_number): Map DWARF_LINK_REGISTER to DWARF PC.
11140 (gcn_dwarf_register_span): DWARF_LINK_REGISTER doesn't span.
11141 * config/gcn/gcn.h: (DWARF_FRAME_RETURN_COLUMN): New define.
11142 (DWARF_LINK_REGISTER): New define.
11143 (FIRST_PSEUDO_REGISTER): Increment.
11144 (FIXED_REGISTERS): Add entry for DWARF_LINK_REGISTER.
11145 (CALL_USED_REGISTERS): Likewise.
11146 (REGISTER_NAMES): Likewise.
11148 2021-06-30 Richard Biener <rguenther@suse.de>
11150 PR tree-optimization/101267
11151 * tree-vect-stmts.c (vect_check_scalar_mask): Adjust
11152 API and use SLP compatible interface of vect_is_simple_use.
11153 Reject not vectorized SLP defs for callers that do not support
11155 (vect_check_store_rhs): Handle masked stores and pass down
11156 the appropriate operator index.
11157 (vectorizable_call): Adjust.
11158 (vectorizable_store): Likewise.
11159 (vectorizable_load): Likewise. Handle SLP pecularity of
11161 (vect_is_simple_use): Remove special-casing of masked stores.
11163 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
11165 * common.opt (foffload): Remove help as Driver only.
11166 * gcc.c (display_help): Add -foffload.
11168 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
11170 * gcc.c (close_at_file, execute): Replace alloca by XALLOCAVEC.
11171 (check_offload_target_name): Fix splitting OFFLOAD_TARGETS into
11172 a candidate list; better inform no offload target is configured
11173 and fix hint extraction when passed target is not '\0' at [len].
11174 * common.opt (foffload): Add tailing '.'.
11175 (foffload-options): Likewise; fix flag name in the help string.
11177 2021-06-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
11180 * config/arm/arm_neon.h: Move vabs intrinsics before vcage_f32.
11181 (vcage_f32): Gate comparison on __FAST_MATH__.
11182 (vcageq_f32): Likewise.
11183 (vcale_f32): Likewise.
11184 (vcaleq_f32): Likewise.
11185 (vcagt_f32): Likewise.
11186 (vcagtq_f32): Likewise.
11187 (vcalt_f32): Likewise.
11188 (vcaltq_f32): Likewise.
11189 (vcage_f16): Likewise.
11190 (vcageq_f16): Likewise.
11191 (vcale_f16): Likewise.
11192 (vcaleq_f16): Likewise.
11193 (vcagt_f16): Likewise.
11194 (vcagtq_f16): Likewise.
11195 (vcalt_f16): Likewise.
11196 (vcaltq_f16): Likewise.
11198 2021-06-30 Richard Biener <rguenther@suse.de>
11200 PR tree-optimization/101264
11201 * tree-vect-slp.c (vect_optimize_slp): Propagate the
11202 computed perm_in to all "any" permute successors
11203 we cannot de-duplicate immediately.
11205 2021-06-30 liuhongt <hongtao.liu@intel.com>
11208 * config/i386/sse.md
11209 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>):
11211 (avx512f_sfixupimm<mode><maskz_scalar_name><round_saeonly_name>):
11213 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>"): Refined.
11214 * config/i386/subst.md (maskz_scalar): New define_subst.
11215 (maskz_scalar_name): New subst_attr.
11216 (maskz_scalar_op5): Ditto.
11217 (round_saeonly_maskz_scalar_op5): Ditto.
11218 (round_saeonly_maskz_scalar_operand5): Ditto.
11220 2021-06-30 David Edelsohn <dje.gcc@gmail.com>
11222 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
11223 Increase code CSECT alignment to at least 32 bytes.
11224 * config/rs6000/xcoff.h (TEXT_SECTION_ASM_OP): Add 32 byte
11225 alignment designation.
11227 2021-06-29 Sergei Trofimovich <siarheit@google.com>
11229 * doc/generic.texi: Fix s/net yet/not yet/ typo.
11231 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
11233 PR tree-optimization/101254
11234 * range-op.cc (operator_minus::op1_op2_relation_effect): Check for
11235 wrapping/non-wrapping when setting the result range.
11237 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
11239 * value-query.cc (gimple_range_global): Allow phis.
11241 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
11243 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): Use stmt.
11244 (simplify_using_ranges::op_with_boolean_value_range_p): Add a
11245 statement for location context.
11246 (check_for_binary_op_overflow): Ditto.
11247 (simplify_using_ranges::get_vr_for_comparison): Ditto.
11248 (simplify_using_ranges::compare_name_with_value): Ditto.
11249 (simplify_using_ranges::compare_names): Ditto.
11250 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Ditto.
11251 (simplify_using_ranges::simplify_truth_ops_using_ranges): Ditto.
11252 (simplify_using_ranges::simplify_min_or_max_using_ranges): Ditto.
11253 (simplify_using_ranges::simplify_internal_call_using_ranges): Ditto.
11254 (simplify_using_ranges::two_valued_val_range_p): Ditto.
11255 (simplify_using_ranges::simplify): Ditto.
11256 * vr-values.h: Adjust prototypes.
11258 2021-06-29 Uroš Bizjak <ubizjak@gmail.com>
11261 * config/i386/mmx.md (vec_addsubv2sf3): New insn pattern.
11263 2021-06-29 Julian Brown <julian@codesourcery.com>
11265 * config/gcn/gcn.c (gcn_init_libfuncs): New function.
11266 (TARGET_INIT_LIBFUNCS): Define target hook using above function.
11267 * config/gcn/gcn.h (UNITS_PER_WORD): Define to 8 for IN_LIBGCC2, 4
11269 (LIBGCC2_UNITS_PER_WORD, BITS_PER_WORD): Remove definitions.
11270 (MAX_FIXED_MODE_SIZE): Change to 128.
11272 2021-06-29 Julian Brown <julian@codesourcery.com>
11274 * config/gcn/gcn.md (UNSPEC_FLBIT_INT): New unspec constant.
11275 (s_mnemonic): Add clrsb.
11276 (gcn_flbit<mode>_int): Add insn pattern for SImode/DImode.
11277 (clrsb<mode>2): Add expander for SImode/DImode.
11279 2021-06-29 Julian Brown <julian@codesourcery.com>
11281 * config/gcn/gcn.md (<su>mulsidi3, <su>mulsidi3_reg, <su>mulsidi3_imm,
11282 muldi3): Add patterns.
11284 2021-06-29 Julian Brown <julian@codesourcery.com>
11286 * config/gcn/gcn.md (<su>mulsi3_highpart): Change to expander.
11287 (<su>mulsi3_highpart_reg, <su>mulsi3_highpart_imm): New patterns.
11289 2021-06-29 Julian Brown <julian@codesourcery.com>
11291 * config/gcn/gcn.md (mulsi3): Make s_mulk_i32 variant clobber SCC.
11293 2021-06-29 Joseph Myers <joseph@codesourcery.com>
11295 * btfout.c, ctfout.c: Include "memmodel.h".
11297 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
11299 * gcc.c (check_offload_target_name): Cast len argument to
11300 %q.*s to 'int'; avoid -Wstringop-truncation warning.
11302 2021-06-29 Richard Biener <rguenther@suse.de>
11304 * tree-vect-slp.c (vect_optimize_slp): Forward propagate
11305 to "any" permute nodes and relax "any" permute proapgation
11306 during iterative backward propagation.
11308 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
11311 * common.opt (-foffload=): Update description.
11312 (-foffload-options=): New.
11313 * doc/invoke.texi (C Language Options): Document
11314 -foffload and -foffload-options.
11315 * gcc.c (check_offload_target_name): New, split off from
11316 handle_foffload_option.
11317 (check_foffload_target_names): New.
11318 (handle_foffload_option): Handle -foffload=default.
11319 (driver_handle_option): Update for -foffload-options.
11320 * lto-opts.c (lto_write_options): Use -foffload-options
11321 instead of -foffload.
11322 * lto-wrapper.c (merge_and_complain, append_offload_options):
11324 * opts.c (common_handle_option): Likewise.
11326 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
11328 * doc/invoke.texi (C Language Options): Sort options
11329 alphabetically in optlist and also the description itself.
11330 Remove leftover -fallow-single-precision from and add missing
11331 -fgnu-tm to the optlist.
11333 2021-06-29 Richard Biener <rguenther@suse.de>
11335 * tree-vect-slp.c (slpg_vertex::visited): Remove.
11336 (vect_slp_perms_eq): Handle -1 permutes.
11337 (vect_optimize_slp): Rewrite permute propagation.
11339 2021-06-29 Jakub Jelinek <jakub@redhat.com>
11342 * match.pd ((intptr_t)x eq/ne CST to x eq/ne (typeof x) CST): Don't
11343 perform the optimization in GENERIC when sanitizing and x has a
11346 2021-06-29 Richard Biener <rguenther@suse.de>
11348 PR tree-optimization/101242
11349 * tree-vect-slp.c (vect_slp_build_vertices): Force-add
11350 PHIs with not represented initial values as leafs.
11352 2021-06-29 Jan-Benedict Glaw <jbglaw@getslash.de>
11354 * config/pdp11/pdp11.h (ASM_OUTPUT_SKIP): Fix signedness warning.
11355 * config/pdp11/pdp11.c (pdp11_asm_print_operand_punct_valid_p): Remove
11356 "register" keyword.
11357 (pdp11_initial_elimination_offset) Remove unused variable.
11358 (pdp11_cmp_length) Ditto.
11359 (pdp11_insn_cost): Ditto, and fix signedness warning.
11361 2021-06-29 David Edelsohn <dje.gcc@gmail.com>
11363 * btfout.c: Include tm_p.h.
11366 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
11368 * config/bpf/bpf.c (bpf_expand_prologue): Do not mark insns as
11370 (bpf_expand_epilogue): Likewise.
11371 * config/bpf/bpf.h (DWARF2_FRAME_INFO): Define to 0.
11372 Do not define DBX_DEBUGGING_INFO.
11374 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
11376 * doc/invoke.texi: Document the CTF and BTF debug info options.
11378 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
11379 David Faust <david.faust@oracle.com>
11380 Jose E. Marchesi <jose.marchesi@oracle.com>
11381 Weimin Pan <weimin.pan@oracle.com>
11383 * Makefile.in: Add ctfc.*, ctfout.c and btfout.c files to
11384 GTFILES. Add new object files.
11385 * common.opt: Add CTF and BTF debug info options.
11386 * btfout.c: New file.
11387 * ctfc.c: Likewise.
11388 * ctfc.h: Likewise.
11389 * ctfout.c: Likewise.
11390 * dwarf2ctf.c: Likewise.
11391 * dwarf2ctf.h: Likewise.
11392 * dwarf2cfi.c (dwarf2out_do_frame): Acknowledge CTF_DEBUG and
11394 * dwarf2out.c (dwarf2out_source_line): Likewise.
11395 (dwarf2out_finish): Skip emitting DWARF if CTF or BTF are to
11397 (debug_format_do_cu): New function.
11398 (dwarf2out_early_finish): Traverse DIEs and emit CTF/BTF for
11400 Include dwarf2ctf.c.
11401 * final.c (dwarf2_debug_info_emitted_p): Acknowledge DWARF-based debug
11403 * flag-types.h (enum debug_info_type): Add CTF_DEBUG and BTF_DEBUG.
11404 (CTF_DEBUG): New bitmask.
11405 (BTF_DEBUG): Likewise.
11406 (enum ctf_debug_info_levels): New enum.
11407 * gengtype.c (open_base_files): Handle ctfc.h.
11408 (main): Handle uint32_t type.
11409 * flags.h (btf_debuginfo_p): New definition.
11410 (dwarf_based_debuginfo_p): Likewise.
11411 * opts.c (debug_type_names): Add entries for CTF and BTF.
11412 (btf_debuginfo_p): New function.
11413 (dwarf_based_debuginfo_p): Likewise.
11414 (common_handle_option): Handle -gctfN and -gbtf options.
11415 (set_debug_level): Set CTF_DEBUG, BTF_DEBUG whenever appropriate.
11416 * toplev.c (process_options): Inform the user and ignore -gctfLEVEL if
11419 2021-06-28 Jose E. Marchesi <jose.marchesi@oracle.com>
11421 * dwarf2out.c (AT_class): Function is no longer static.
11422 (AT_int): Likewise.
11423 (AT_unsigned): Likewise.
11424 (AT_loc): Likewise.
11425 (get_AT): Likewise.
11426 (get_AT_string): Likewise.
11427 (get_AT_flag): Likewise.
11428 (get_AT_unsigned): Likewise.
11429 (get_AT_ref): Likewise.
11430 (new_die_raw): Likewise.
11431 (lookup_decl_die): Likewise.
11432 (base_type_die): Likewise.
11433 (add_name_attribute): Likewise.
11434 (add_AT_int): Likewise.
11435 (add_AT_unsigned): Likewise.
11436 (add_AT_loc): Likewise.
11437 (dw_get_die_tag): New function.
11438 (dw_get_die_child): Likewise.
11439 (dw_get_die_sib): Likewise.
11440 (struct dwarf_file_data): Move from here to dwarf2out.h
11441 (struct dw_attr_struct): Likewise.
11442 * dwarf2out.h: Analogous changes.
11444 2021-06-28 Martin Jambor <mjambor@suse.cz>
11447 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
11448 members m_dead_stmts and m_dead_ssas.
11449 * ipa-param-manipulation.c
11450 (ipa_param_body_adjustments::mark_dead_statements): New function.
11451 (ipa_param_body_adjustments::common_initialization): Call it on
11452 all removed but not split parameters.
11453 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
11455 (ipa_param_body_adjustments::modify_call_stmt): Remove arguments that
11457 * tree-inline.c (remap_gimple_stmt): Do not copy dead statements, reset
11458 dead debug statements.
11459 (copy_phis_for_bb): Do not copy dead PHI nodes.
11461 2021-06-28 Martin Jambor <mjambor@suse.cz>
11464 * symtab-clones.h (clone_info): Removed member param_adjustments.
11465 * ipa-param-manipulation.h: Adjust initial comment to reflect how we
11466 deal with pass-through splits now.
11467 (ipa_param_performed_split): Removed.
11468 (ipa_param_adjustments::modify_call): Adjusted parameters.
11469 (class ipa_param_body_adjustments): Adjusted parameters of
11470 register_replacement, modify_gimple_stmt and modify_call_stmt.
11471 (ipa_verify_edge_has_no_modifications): Declare.
11472 (ipa_edge_modifications_finalize): Declare.
11473 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Remove
11474 performed_splits processing, pas only edge to padjs->modify_call,
11475 check that call arguments were not modified if they should not have
11477 * cgraphclones.c (cgraph_node::create_clone): Do not copy performed
11479 * ipa-param-manipulation.c (struct pass_through_split_map): New type.
11480 (ipa_edge_modification_info): Likewise.
11481 (ipa_edge_modification_sum): Likewise.
11482 (ipa_edge_modifications): New edge summary.
11483 (ipa_verify_edge_has_no_modifications): New function.
11484 (transitive_split_p): Removed.
11485 (transitive_split_map): Likewise.
11486 (init_transitive_splits): Likewise.
11487 (ipa_param_adjustments::modify_call): Adjusted to use the new edge
11488 summary instead of performed_splits.
11489 (ipa_param_body_adjustments::register_replacement): Drop dummy
11490 parameter, set base_index of the created ipa_param_body_replacement.
11491 (phi_arg_will_live_p): New function.
11492 (ipa_param_body_adjustments::common_initialization): Do not create
11493 IPA_SRA dummy decls.
11494 (simple_tree_swap_info): Removed.
11495 (remap_split_decl_to_dummy): Likewise.
11496 (record_argument_state_1): New function.
11497 (record_argument_state): Likewise.
11498 (ipa_param_body_adjustments::modify_call_stmt): New parameter
11499 orig_stmt. Do not work with dummy decls, save necessary info about
11500 changes to ipa_edge_modifications.
11501 (ipa_param_body_adjustments::modify_gimple_stmt): New parameter
11502 orig_stmt, pass it to modify_call_stmt.
11503 (ipa_param_body_adjustments::modify_cfun_body): Adjust call to
11504 modify_gimple_stmt.
11505 (ipa_edge_modifications_finalize): New function.
11506 * tree-inline.c (remap_gimple_stmt): Pass original statement to
11507 modify_gimple_stmt.
11508 (copy_phis_for_bb): Do not copy dead PHI nodes.
11509 (expand_call_inline): Do not remap performed_splits.
11510 (update_clone_info): Likewise.
11511 * toplev.c: Include ipa-param-manipulation.h.
11512 (toplev::finalize): Call ipa_edge_modifications_finalize.
11514 2021-06-28 Andrew Pinski <apinski@marvell.com>
11516 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Duplicate range
11517 info if we're the only things setting the target PHI.
11518 (value_replacement): Don't duplicate range here.
11519 (minmax_replacement): Likewise.
11521 2021-06-28 Richard Biener <rguenther@suse.de>
11523 PR tree-optimization/101229
11524 * gimple-walk.c (gimple_walk_op): Handle PHIs.
11526 2021-06-28 Martin Liska <mliska@suse.cz>
11528 * config/v850/v850.c (construct_dispose_instruction): Allocate
11530 (construct_prepare_instruction): Likewise.
11532 2021-06-28 Martin Liska <mliska@suse.cz>
11534 * config/v850/v850.c (v850_option_override): Build default
11536 (v850_can_inline_p): New. Allow MASK_PROLOG_FUNCTION to be
11537 ignored for inlining.
11538 (TARGET_CAN_INLINE_P): New.
11540 2021-06-28 Richard Biener <rguenther@suse.de>
11542 PR tree-optimization/101207
11543 * tree-vect-slp.c (vect_optimize_slp): Do BB reduction
11544 permute eliding for load permutations properly.
11546 2021-06-28 Richard Biener <rguenther@suse.de>
11548 PR tree-optimization/101173
11549 * gimple-loop-interchange.cc
11550 (tree_loop_interchange::valid_data_dependences): Disallow outer
11551 loop dependence distance of zero.
11553 2021-06-28 liuhongt <hongtao.liu@intel.com>
11556 * config/i386/sse.md (*avx_cmp<mode>3_lt): New
11557 define_insn_and_split.
11558 (*avx_cmp<mode>3_ltint): Ditto.
11559 (*avx2_pcmp<mode>3_3): Ditto.
11560 (*avx2_pcmp<mode>3_4): Ditto.
11561 (*avx2_pcmp<mode>3_5): Ditto.
11563 2021-06-28 liuhongt <hongtao.liu@intel.com>
11565 * config/i386/i386-builtin.def (IX86_BUILTIN_BLENDVPD256,
11566 IX86_BUILTIN_BLENDVPS256, IX86_BUILTIN_PBLENDVB256,
11567 IX86_BUILTIN_BLENDVPD, IX86_BUILTIN_BLENDVPS,
11568 IX86_BUILTIN_PBLENDVB128): Replace icode with
11570 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold blendv
11572 * config/i386/sse.md (*<sse4_1_avx2>_pblendvb_lt_subreg_not):
11573 New pre_reload splitter.
11575 2021-06-27 Andrew Pinski <apinski@marvell.com>
11577 PR middle-end/101230
11578 * fold-const.c (fold_ternary_loc): Check
11579 the return value of invert_tree_comparison.
11581 2021-06-27 David Edelsohn <dje.gcc@gmail.com>
11583 * config.gcc: Add SPDX License Identifier.
11584 (powerpc-ibm-aix789): Default to aix73.h.
11585 (powerpc-ibm-aix7.2.*.*): New stanza.
11586 * config/rs6000/aix72.h: Add SPDX License Identifier.
11587 * config/rs6000/aix73.h: New file.
11589 2021-06-26 Jason Merrill <jason@redhat.com>
11591 * except.c: #include "dwarf2.h" instead of "dwarf2out.h".
11593 2021-06-26 Andrew Pinski <apinski@marvell.com>
11595 * genmatch.c (lower_cond): Copy for_subst_vec
11596 for the simplify also.
11597 (lower): Swap the order for lower_for and lower_cond.
11599 2021-06-26 Andrew Pinski <apinski@marvell.com>
11601 * tree-ssa-phiopt.c (match_simplify_replacement): Reset
11602 flow senatitive info on the moved ssa set.
11604 2021-06-26 Andrew Pinski <apinski@marvell.com>
11606 * fold-const.c (fold_cond_expr_with_comparison):
11607 Exand arg0 into comp_code, arg00, and arg01.
11608 (fold_ternary_loc): Use invert_tree_comparison
11609 instead of fold_invert_truthvalue for the case
11610 where we have A CMP B ? C : A.
11612 2021-06-25 Martin Sebor <msebor@redhat.com>
11614 PR middle-end/101216
11615 * calls.c (maybe_warn_rdwr_sizes): Use the no_warning constant.
11617 2021-06-25 Jeff Law <jeffreyalaw@gmail.com>
11619 * config/h8300/h8300.c (select_cc_mode): Handle ASHIFTRT and LSHIFTRT.
11621 2021-06-25 Richard Biener <rguenther@suse.de>
11623 PR tree-optimization/101202
11624 * tree-vect-slp.c (vect_optimize_slp): Explicitely handle
11627 2021-06-25 Richard Biener <rguenther@suse.de>
11629 * tree-vect-slp-patterns.c (addsub_pattern::build): Copy
11630 STMT_VINFO_REDUC_DEF from the original representative.
11632 2021-06-25 Martin Sebor <msebor@redhat.com>
11634 * builtins.c (warn_string_no_nul): Replace uses of TREE_NO_WARNING,
11635 gimple_no_warning_p and gimple_set_no_warning with
11636 warning_suppressed_p, and suppress_warning.
11638 (maybe_warn_for_bound): Same.
11639 (warn_for_access): Same.
11640 (check_access): Same.
11641 (expand_builtin_strncmp): Same.
11642 (fold_builtin_varargs): Same.
11643 * calls.c (maybe_warn_nonstring_arg): Same.
11644 (maybe_warn_rdwr_sizes): Same.
11645 * cfgexpand.c (expand_call_stmt): Same.
11646 * cgraphunit.c (check_global_declaration): Same.
11647 * fold-const.c (fold_undefer_overflow_warnings): Same.
11648 (fold_truth_not_expr): Same.
11649 (fold_unary_loc): Same.
11650 (fold_checksum_tree): Same.
11651 * gimple-array-bounds.cc (array_bounds_checker::check_array_ref): Same.
11652 (array_bounds_checker::check_mem_ref): Same.
11653 (array_bounds_checker::check_addr_expr): Same.
11654 (array_bounds_checker::check_array_bounds): Same.
11655 * gimple-expr.c (copy_var_decl): Same.
11656 * gimple-fold.c (gimple_fold_builtin_strcpy): Same.
11657 (gimple_fold_builtin_strncat): Same.
11658 (gimple_fold_builtin_stxcpy_chk): Same.
11659 (gimple_fold_builtin_stpcpy): Same.
11660 (gimple_fold_builtin_sprintf): Same.
11661 (fold_stmt_1): Same.
11662 * gimple-ssa-isolate-paths.c (diag_returned_locals): Same.
11663 * gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Same.
11664 * gimple-ssa-sprintf.c (handle_printf_call): Same.
11665 * gimple-ssa-store-merging.c (imm_store_chain_info::output_merged_store): Same.
11666 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
11667 * gimple-ssa-warn-restrict.h: Adjust declarations.
11668 (maybe_diag_access_bounds): Replace uses of TREE_NO_WARNING,
11669 gimple_no_warning_p and gimple_set_no_warning with
11670 warning_suppressed_p, and suppress_warning.
11671 (check_call): Same.
11672 (check_bounds_or_overlap): Same.
11673 * gimple.c (gimple_build_call_from_tree): Same.
11674 * gimplify.c (gimplify_return_expr): Same.
11675 (gimplify_cond_expr): Same.
11676 (gimplify_modify_expr_complex_part): Same.
11677 (gimplify_modify_expr): Same.
11678 (gimple_push_cleanup): Same.
11679 (gimplify_expr): Same.
11680 * omp-expand.c (expand_omp_for_generic): Same.
11681 (expand_omp_taskloop_for_outer): Same.
11682 * omp-low.c (lower_rec_input_clauses): Same.
11683 (lower_lastprivate_clauses): Same.
11684 (lower_send_clauses): Same.
11685 (lower_omp_target): Same.
11686 * tree-cfg.c (pass_warn_function_return::execute): Same.
11687 * tree-complex.c (create_one_component_var): Same.
11688 * tree-inline.c (remap_gimple_op_r): Same.
11689 (copy_tree_body_r): Same.
11690 (declare_return_variable): Same.
11691 (expand_call_inline): Same.
11692 * tree-nested.c (lookup_field_for_decl): Same.
11693 * tree-sra.c (create_access_replacement): Same.
11694 (generate_subtree_copies): Same.
11695 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Same.
11696 * tree-ssa-forwprop.c (combine_cond_expr_cond): Same.
11697 * tree-ssa-loop-ch.c (ch_base::copy_headers): Same.
11698 * tree-ssa-loop-im.c (execute_sm): Same.
11699 * tree-ssa-phiopt.c (cond_store_replacement): Same.
11700 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
11701 (handle_builtin_strcpy): Same.
11702 (maybe_diag_stxncpy_trunc): Same.
11703 (handle_builtin_stxncpy_strncat): Same.
11704 (handle_builtin_strcat): Same.
11705 * tree-ssa-uninit.c (get_no_uninit_warning): Same.
11706 (set_no_uninit_warning): Same.
11707 (uninit_undefined_value_p): Same.
11708 (warn_uninit): Same.
11709 (maybe_warn_operand): Same.
11710 * tree-vrp.c (compare_values_warnv): Same.
11711 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr): Same.
11712 (test_for_singularity): Same.
11713 * gimple.h (warning_suppressed_p): New function.
11714 (suppress_warning): Same.
11715 (copy_no_warning): Same.
11716 (gimple_set_block): Call gimple_set_location.
11717 (gimple_set_location): Call copy_warning.
11719 2021-06-25 Martin Sebor <msebor@redhat.com>
11721 * tree.h (warning_suppressed_at, copy_warning,
11722 warning_suppressed_p, suppress_warning): New functions.
11724 2021-06-25 Martin Sebor <msebor@redhat.com>
11726 * Makefile.in (OBJS-libcommon): Add diagnostic-spec.o.
11727 * gengtype.c (open_base_files): Add diagnostic-spec.h.
11728 * diagnostic-spec.c: New file.
11729 * diagnostic-spec.h: New file.
11730 * tree.h (no_warning, all_warnings, suppress_warning_at): New
11732 * warning-control.cc: New file.
11734 2021-06-25 liuhongt <hongtao.liu@intel.com>
11737 * config/i386/i386.c (x86_order_regs_for_local_alloc):
11740 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11742 PR tree-optimization/101189
11743 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Pass
11744 LHS range of condition to postfold routine.
11745 (fold_using_range::postfold_gcond_edges): Only process the TRUE or
11746 FALSE edge if the LHS range supports it being taken.
11747 * gimple-range-fold.h (postfold_gcond_edges): Add range parameter.
11749 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11751 * value-relation.cc (equiv_oracle::dump): Do not dump NULL blocks.
11752 (relation_oracle::find_relation_block): Check correct bitmap.
11753 (relation_oracle::dump): Do not dump NULL blocks.
11755 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11757 * gimple-range-cache.cc (ranger_cache::propagate_cache): Call
11758 range_on_edge instead of manually calculating.
11760 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11762 * range-op.cc: Fix comment.
11764 2021-06-24 Uroš Bizjak <ubizjak@gmail.com>
11767 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
11768 Handle V8QI and V4HI modes.
11769 * config/i386/mmx.md (sse4_1_<any_extend:code>v4qiv4hi2):
11771 (sse4_1_<any_extend:code>v4qiv4hi2): Ditto.
11772 (mmxpackmode): New mode attribute.
11773 (vec_pack_trunc_<mmxpackmode:mode>): New expander.
11774 (mmxunpackmode): New mode attribute.
11775 (vec_unpacks_lo_<mmxunpackmode:mode>): New expander.
11776 (vec_unpacks_hi_<mmxunpackmode:mode>): Ditto.
11777 (vec_unpacku_lo_<mmxunpackmode:mode>): Ditto.
11778 (vec_unpacku_hi_<mmxunpackmode:mode>): Ditto.
11779 * config/i386/i386.md (extsuffix): Move from ...
11780 * config/i386/sse.md: ... here.
11782 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
11784 * dwarf2out.c (dwarf2out_assembly_start): Emit .file 0 marker here..
11785 (dwarf2out_finish): ...instead of here.
11787 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
11789 * configure.ac (--gdwarf-5 option): Use objdump instead of readelf.
11790 (working --gdwarf-4/--gdwarf-5 for all sources): Likewise.
11791 (--gdwarf-4 not refusing generated .debug_line): Adjust for Windows.
11792 * configure: Regenerate.
11794 2021-06-24 Richard Biener <rguenther@suse.de>
11796 * config/i386/sse.md (vec_addsubv4df3, vec_addsubv2df3,
11797 vec_addsubv8sf3, vec_addsubv4sf3): Merge into ...
11798 (vec_addsub<mode>3): ... using a new addsub_cst mode attribute.
11800 2021-06-24 Richard Biener <rguenther@suse.de>
11802 * config/i386/sse.md (avx_addsubv4df3): Rename to
11804 (avx_addsubv8sf3): Rename to vec_addsubv8sf3.
11805 (sse3_addsubv2df3): Rename to vec_addsubv2df3.
11806 (sse3_addsubv4sf3): Rename to vec_addsubv4sf3.
11807 * config/i386/i386-builtin.def: Adjust.
11808 * internal-fn.def (VEC_ADDSUB): New internal optab fn.
11809 * optabs.def (vec_addsub_optab): New optab.
11810 * tree-vect-slp-patterns.c (class addsub_pattern): New.
11811 (slp_patterns): Add addsub_pattern.
11812 * tree-vect-slp.c (vect_optimize_slp): Disable propagation
11813 across CFN_VEC_ADDSUB.
11814 * tree-vectorizer.h (vect_pattern::vect_pattern): Make
11816 * doc/md.texi (vec_addsub<mode>3): Document.
11818 2021-06-24 Jakub Jelinek <jakub@redhat.com>
11820 PR middle-end/101170
11821 * df-scan.c (df_ref_record): For paradoxical big-endian SUBREGs
11822 where regno + subreg_regno_offset wraps around use 0 as starting
11825 2021-06-24 Jakub Jelinek <jakub@redhat.com>
11827 PR middle-end/101172
11828 * stor-layout.c (finish_bitfield_representative): If nextf has
11829 error_mark_node type, set repr type to error_mark_node too.
11831 2021-06-24 Ilya Leoshkevich <iii@linux.ibm.com>
11833 * config/s390/s390.c (s390_function_profiler): Ignore labelno
11835 * config/s390/s390.h (NO_PROFILE_COUNTERS): Define.
11837 2021-06-24 Richard Biener <rguenther@suse.de>
11839 * tree-vect-slp.c (vect_optimize_slp): Do not propagate
11840 across operations that have different semantics on different
11843 2021-06-24 Jakub Jelinek <jakub@redhat.com>
11845 * tree.h (OMP_CLAUSE_MAP_IN_REDUCTION): Document meaning for OpenMP.
11846 * gimplify.c (gimplify_scan_omp_clauses): For OpenMP map clauses
11847 with OMP_CLAUSE_MAP_IN_REDUCTION flag partially defer gimplification
11848 of non-decl OMP_CLAUSE_DECL. For OMP_CLAUSE_IN_REDUCTION on
11849 OMP_TARGET user outer_ctx instead of ctx for placeholders and
11850 initializer/combiner gimplification.
11851 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_MAP_IN_REDUCTION
11852 on target constructs.
11853 (lower_rec_input_clauses): Likewise.
11854 (lower_omp_target): Likewise.
11855 * omp-expand.c (expand_omp_target): Temporarily ignore nowait clause
11856 on target if in_reduction is present.
11858 2021-06-24 Kewen Lin <linkw@linux.ibm.com>
11860 * tree-predcom.c (class pcom_worker): New class.
11861 (release_chain): Renamed to...
11862 (pcom_worker::release_chain): ...this.
11863 (release_chains): Renamed to...
11864 (pcom_worker::release_chains): ...this.
11865 (aff_combination_dr_offset): Renamed to...
11866 (pcom_worker::aff_combination_dr_offset): ...this.
11867 (determine_offset): Renamed to...
11868 (pcom_worker::determine_offset): ...this.
11869 (class comp_ptrs): New class.
11870 (split_data_refs_to_components): Renamed to...
11871 (pcom_worker::split_data_refs_to_components): ...this,
11872 and update with class comp_ptrs.
11873 (suitable_component_p): Renamed to...
11874 (pcom_worker::suitable_component_p): ...this.
11875 (filter_suitable_components): Renamed to...
11876 (pcom_worker::filter_suitable_components): ...this.
11877 (valid_initializer_p): Renamed to...
11878 (pcom_worker::valid_initializer_p): ...this.
11879 (find_looparound_phi): Renamed to...
11880 (pcom_worker::find_looparound_phi): ...this.
11881 (add_looparound_copies): Renamed to...
11882 (pcom_worker::add_looparound_copies): ...this.
11883 (determine_roots_comp): Renamed to...
11884 (pcom_worker::determine_roots_comp): ...this.
11885 (determine_roots): Renamed to...
11886 (pcom_worker::determine_roots): ...this.
11887 (single_nonlooparound_use): Renamed to...
11888 (pcom_worker::single_nonlooparound_use): ...this.
11889 (remove_stmt): Renamed to...
11890 (pcom_worker::remove_stmt): ...this.
11891 (execute_pred_commoning_chain): Renamed to...
11892 (pcom_worker::execute_pred_commoning_chain): ...this.
11893 (execute_pred_commoning): Renamed to...
11894 (pcom_worker::execute_pred_commoning): ...this.
11895 (struct epcc_data): New member worker.
11896 (execute_pred_commoning_cbck): Call execute_pred_commoning
11897 with pcom_worker pointer.
11898 (find_use_stmt): Renamed to...
11899 (pcom_worker::find_use_stmt): ...this.
11900 (find_associative_operation_root): Renamed to...
11901 (pcom_worker::find_associative_operation_root): ...this.
11902 (find_common_use_stmt): Renamed to...
11903 (pcom_worker::find_common_use_stmt): ...this.
11904 (combinable_refs_p): Renamed to...
11905 (pcom_worker::combinable_refs_p): ...this.
11906 (reassociate_to_the_same_stmt): Renamed to...
11907 (pcom_worker::reassociate_to_the_same_stmt): ...this.
11908 (stmt_combining_refs): Renamed to...
11909 (pcom_worker::stmt_combining_refs): ...this.
11910 (combine_chains): Renamed to...
11911 (pcom_worker::combine_chains): ...this.
11912 (try_combine_chains): Renamed to...
11913 (pcom_worker::try_combine_chains): ...this.
11914 (prepare_initializers_chain): Renamed to...
11915 (pcom_worker::prepare_initializers_chain): ...this.
11916 (prepare_initializers): Renamed to...
11917 (pcom_worker::prepare_initializers): ...this.
11918 (prepare_finalizers_chain): Renamed to...
11919 (pcom_worker::prepare_finalizers_chain): ...this.
11920 (prepare_finalizers): Renamed to...
11921 (pcom_worker::prepare_finalizers): ...this.
11922 (tree_predictive_commoning_loop): Renamed to...
11923 (pcom_worker::tree_predictive_commoning_loop): ...this, adjust
11924 some calls and remove some cleanup code.
11925 (tree_predictive_commoning): Adjusted to use pcom_worker instance.
11926 (static variable looparound_phis): Remove.
11927 (static variable name_expansions): Remove.
11929 2021-06-24 Richard Biener <rguenther@suse.de>
11931 * tree-vect-slp.c (slpg_vertex): New struct.
11932 (vect_slp_build_vertices): Adjust.
11933 (vect_optimize_slp): Likewise. Maintain an outgoing permute
11934 and a materialized one.
11936 2021-06-24 Richard Biener <rguenther@suse.de>
11938 PR tree-optimization/101105
11939 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
11940 Only ignore steps when they are equal or scalar order is preserved.
11942 2021-06-24 liuhongt <hongtao.liu@intel.com>
11945 * config/i386/i386-expand.c (ix86_expand_vec_interleave):
11946 Adjust comments for ix86_expand_vecop_qihi2.
11947 (ix86_expand_vecmul_qihi): Renamed to ..
11948 (ix86_expand_vecop_qihi2): Adjust function prototype to
11949 support shift operation, add static to definition.
11950 (ix86_expand_vec_shift_qihi_constant): Add static to definition.
11951 (ix86_expand_vecop_qihi): Call ix86_expand_vecop_qihi2 and
11952 ix86_expand_vec_shift_qihi_constant.
11953 * config/i386/i386-protos.h (ix86_expand_vecmul_qihi): Deleted.
11954 (ix86_expand_vec_shift_qihi_constant): Deleted.
11955 * config/i386/sse.md (VI12_256_512_AVX512VL): New mode
11957 (mulv8qi3): Call ix86_expand_vecop_qihi directly, add
11958 condition TARGET_64BIT.
11959 (mul<mode>3): Ditto.
11960 (<insn><mode>3): Ditto.
11961 (vlshr<mode>3): Extend to support avx512 vlshr.
11962 (v<insn><mode>3): New expander for
11964 (v<insn>v8qi3): Ditto.
11965 (vashrv8hi3<mask_name>): Renamed to ..
11966 (vashr<mode>3): And extend to support V16QImode for avx512.
11967 (vashrv16qi3): Deleted.
11968 (vashrv2di3<mask_name>): Extend expander to support avx512
11971 2021-06-23 Dimitar Dimitrov <dimitar@dinux.eu>
11973 * doc/lto.texi (Design Overview): Update that slim objects are
11976 2021-06-23 Aaron Sawdey <acsawdey@linux.ibm.com>
11978 * config/rs6000/rs6000-cpus.def: Take OPTION_MASK_PCREL_OPT out
11979 of OTHER_POWER10_MASKS so it will not be enabled by default.
11981 2021-06-23 Richard Biener <rguenther@suse.de>
11982 Martin Jambor <mjambor@suse.cz>
11984 * tree-inline.c (setup_one_parameter): Set TREE_READONLY of the
11985 param replacement unconditionally. Adjust comment.
11987 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11989 * Makefile.in (OBJS): Add gimple-range-fold.o
11990 * gimple-range-fold.cc: New.
11991 * gimple-range-fold.h: New.
11992 * gimple-range-gori.cc (gimple_range_calc_op1): Move to here.
11993 (gimple_range_calc_op2): Ditto.
11994 * gimple-range-gori.h: Move prototypes to here.
11995 * gimple-range.cc: Adjust include files.
11996 (fur_source:fur_source): Relocate to gimple-range-fold.cc.
11997 (fur_source::get_operand): Ditto.
11998 (fur_source::get_phi_operand): Ditto.
11999 (fur_source::query_relation): Ditto.
12000 (fur_source::register_relation): Ditto.
12001 (class fur_edge): Ditto.
12002 (fur_edge::fur_edge): Ditto.
12003 (fur_edge::get_operand): Ditto.
12004 (fur_edge::get_phi_operand): Ditto.
12005 (fur_stmt::fur_stmt): Ditto.
12006 (fur_stmt::get_operand): Ditto.
12007 (fur_stmt::get_phi_operand): Ditto.
12008 (fur_stmt::query_relation): Ditto.
12009 (class fur_depend): Relocate to gimple-range-fold.h.
12010 (fur_depend::fur_depend): Relocate to gimple-range-fold.cc.
12011 (fur_depend::register_relation): Ditto.
12012 (fur_depend::register_relation): Ditto.
12013 (class fur_list): Ditto.
12014 (fur_list::fur_list): Ditto.
12015 (fur_list::get_operand): Ditto.
12016 (fur_list::get_phi_operand): Ditto.
12017 (fold_range): Ditto.
12018 (adjust_pointer_diff_expr): Ditto.
12019 (gimple_range_adjustment): Ditto.
12020 (gimple_range_base_of_assignment): Ditto.
12021 (gimple_range_operand1): Ditto.
12022 (gimple_range_operand2): Ditto.
12023 (gimple_range_calc_op1): Relocate to gimple-range-gori.cc.
12024 (gimple_range_calc_op2): Ditto.
12025 (fold_using_range::fold_stmt): Relocate to gimple-range-fold.cc.
12026 (fold_using_range::range_of_range_op): Ditto.
12027 (fold_using_range::range_of_address): Ditto.
12028 (fold_using_range::range_of_phi): Ditto.
12029 (fold_using_range::range_of_call): Ditto.
12030 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
12031 (fold_using_range::range_of_builtin_call): Ditto.
12032 (fold_using_range::range_of_cond_expr): Ditto.
12033 (fold_using_range::range_of_ssa_name_with_loop_info): Ditto.
12034 (fold_using_range::relation_fold_and_or): Ditto.
12035 (fold_using_range::postfold_gcond_edges): Ditto.
12036 * gimple-range.h: Add gimple-range-fold.h to include files. Change
12037 GIMPLE_RANGE_STMT_H to GIMPLE_RANGE_H.
12038 (gimple_range_handler): Relocate to gimple-range-fold.h.
12039 (gimple_range_ssa_p): Ditto.
12040 (range_compatible_p): Ditto.
12041 (class fur_source): Ditto.
12042 (class fur_stmt): Ditto.
12043 (class fold_using_range): Ditto.
12044 (gimple_range_calc_op1): Relocate to gimple-range-gori.h
12045 (gimple_range_calc_op2): Ditto.
12047 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
12049 PR tree-optimization/101148
12050 PR tree-optimization/101014
12051 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
12052 (ranger_cache::~ranger_cache): Adjust.
12053 (ranger_cache::block_range): Check if propagation disallowed.
12054 (ranger_cache::propagate_cache): Disallow propagation if new value
12055 can't be stored properly.
12056 * gimple-range-cache.h (ranger_cache::m_propfail): New member.
12058 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
12060 * gimple-range-cache.cc (class ssa_block_ranges): Adjust prototype.
12061 (sbr_vector::set_bb_range): Return true.
12062 (class sbr_sparse_bitmap): Adjust.
12063 (sbr_sparse_bitmap::set_bb_range): Return value.
12064 (block_range_cache::set_bb_range): Return value.
12065 (ranger_cache::propagate_cache): Use return value to print msg.
12066 * gimple-range-cache.h (class block_range_cache): Adjust.
12068 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
12070 * gimple-range.cc (dump_bb): Use range_on_edge from the cache.
12072 2021-06-23 Jeff Law <jeffreyalaw@gmail.com>
12074 * config/h8300/logical.md (<code><mode>3<ccnz>): Use <cczn>
12075 so this pattern can be used for test/compare removal. Pass
12076 current insn to compute_logical_op_length and output_logical_op.
12077 * config/h8300/h8300.c (compute_logical_op_cc): Remove.
12078 (h8300_and_costs): Add argument to compute_logical_op_length.
12079 (output_logical_op): Add new argument. Use it to determine if the
12080 condition codes are used and adjust the output accordingly.
12081 (compute_logical_op_length): Add new argument and update length
12082 computations when condition codes are used.
12083 * config/h8300/h8300-protos.h (compute_logical_op_length): Update
12085 (output_logical_op): Likewise.
12087 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
12090 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
12091 Handle 64bit modes for TARGET_XOP. Use indirect gen_* functions.
12092 * config/i386/mmx.md (mmx_ppermv64): New insn pattern.
12093 * config/i386/i386.md (unspec): Move UNSPEC_XOP_PERMUTE from ...
12094 * config/i386/sse.md (unspec): ... here.
12096 2021-06-23 Martin Liska <mliska@suse.cz>
12099 * optc-save-gen.awk: Put back arm_fp16_format to
12102 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
12105 * config/i386/i386.md (bsr_rex64): Add zero-flag setting RTX.
12108 (clz<mode>2): Update RTX pattern for additions.
12110 2021-06-23 Jakub Jelinek <jakub@redhat.com>
12112 PR middle-end/101167
12113 * omp-low.c (lower_omp_regimplify_p): Regimplify also PARM_DECLs
12114 and RESULT_DECLs that have DECL_HAS_VALUE_EXPR_P set.
12116 2021-06-22 Sergei Trofimovich <siarheit@google.com>
12118 * doc/rtl.texi: drop unbalanced parenthesis.
12120 2021-06-22 Richard Biener <rguenther@suse.de>
12122 PR middle-end/101156
12123 * gimplify.c (gimplify_expr): Remove premature incorrect
12126 2021-06-22 Jakub Jelinek <jakub@redhat.com>
12128 PR tree-optimization/101159
12129 * tree-vect-patterns.c (vect_recog_popcount_pattern): Fix some
12132 2021-06-22 Jakub Jelinek <jakub@redhat.com>
12134 PR middle-end/101160
12135 * function.c (assign_parms): For decl_result with TYPE_EMPTY_P type
12136 clear crtl->return_rtx instead of keeping it referencing a pseudo.
12138 2021-06-22 Jakub Jelinek <jakub@redhat.com>
12139 Andrew Pinski <apinski@marvell.com>
12141 PR tree-optimization/101162
12142 * fold-const.c (range_check_type): Handle OFFSET_TYPE like pointer
12145 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12147 * range-op.cc (range_relational_tests): New.
12148 (range_op_tests): Call range_relational_tests.
12150 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12152 * range-op.cc (operator_cast::lhs_op1_relation): New.
12153 (operator_identity::lhs_op1_relation): Mew.
12155 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12157 * range-op.cc (operator_minus::op1_op2_relation_effect): New.
12159 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12161 * range-op.cc (operator_plus::lhs_op1_relation): New.
12162 (operator_plus::lhs_op2_relation): New.
12164 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12166 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a
12167 relation_oracle if dominators exist.
12168 (ranger_cache::~ranger_cache): Dispose of oracle.
12169 (ranger_cache::dump_bb): Dump oracle.
12170 * gimple-range.cc (fur_source::fur_source): New.
12171 (fur_source::get_operand): Use mmeber query.
12172 (fur_source::get_phi_operand): Use member_query.
12173 (fur_source::query_relation): New.
12174 (fur_source::register_dependency): Delete.
12175 (fur_source::register_relation): New.
12176 (fur_edge::fur_edge): Adjust.
12177 (fur_edge::get_phi_operand): Fix comment.
12178 (fur_edge::query): Delete.
12179 (fur_stmt::fur_stmt): Adjust.
12180 (fur_stmt::query): Delete.
12181 (fur_depend::fur_depend): Adjust.
12182 (fur_depend::register_relation): New.
12183 (fur_depend::register_relation): New.
12184 (fur_list::fur_list): Adjust.
12185 (fur_list::get_operand): Use member query.
12186 (fold_using_range::range_of_range_op): Process and query relations.
12187 (fold_using_range::range_of_address): Adjust dependency call.
12188 (fold_using_range::range_of_phi): Ditto.
12189 (gimple_ranger::gimple_ranger): New. Use ranger_ache oracle.
12190 (fold_using_range::relation_fold_and_or): New.
12191 (fold_using_range::postfold_gcond_edges): New.
12192 * gimple-range.h (class gimple_ranger): Adjust.
12193 (class fur_source): Adjust members.
12194 (class fur_stmt): Ditto.
12195 (class fold_using_range): Ditto.
12197 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12199 * range-op.cc (range_operator::wi_fold): Apply relation effect.
12200 (range_operator::fold_range): Adjust and apply relation effect.
12201 (*::fold_range): Add relation parameters.
12202 (*::op1_range): Ditto.
12203 (*::op2_range): Ditto.
12204 (range_operator::lhs_op1_relation): New.
12205 (range_operator::lhs_op2_relation): New.
12206 (range_operator::op1_op2_relation): New.
12207 (range_operator::op1_op2_relation_effect): New.
12208 (relop_early_resolve): New.
12209 (operator_equal::op1_op2_relation): New.
12210 (operator_equal::fold_range): Call relop_early_resolve.
12211 (operator_not_equal::op1_op2_relation): New.
12212 (operator_not_equal::fold_range): Call relop_early_resolve.
12213 (operator_lt::op1_op2_relation): New.
12214 (operator_lt::fold_range): Call relop_early_resolve.
12215 (operator_le::op1_op2_relation): New.
12216 (operator_le::fold_range): Call relop_early_resolve.
12217 (operator_gt::op1_op2_relation): New.
12218 (operator_gt::fold_range): Call relop_early_resolve.
12219 (operator_ge::op1_op2_relation): New.
12220 (operator_ge::fold_range): Call relop_early_resolve.
12221 * range-op.h (class range_operator): Adjust parameters and methods.
12223 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12225 * Makefile.in (OBJS): Add value-relation.o.
12226 * gimple-range.h: Adjust include files.
12227 * tree-data-ref.c: Adjust include file order.
12228 * value-query.cc (range_query::get_value_range): Default to no oracle.
12229 (range_query::query_relation): New.
12230 (range_query::query_relation): New.
12231 * value-query.h (class range_query): Adjust.
12232 * value-relation.cc: New.
12233 * value-relation.h: New.
12235 2021-06-22 Richard Biener <rguenther@suse.de>
12237 PR tree-optimization/101151
12238 * tree-ssa-sink.c (statement_sink_location): Expand irreducible
12241 2021-06-22 Jojo R <rjiejie@linux.alibaba.com>
12243 * config/riscv/riscv.c (thead_c906_tune_info): New.
12244 (riscv_tune_info_table): Use new tune.
12246 2021-06-22 Richard Biener <rguenther@suse.de>
12248 PR tree-optimization/101158
12249 * tree-vect-slp.c (vect_build_slp_tree_1): Move same operand
12250 checking after checking for matching operation.
12252 2021-06-22 Richard Biener <rguenther@suse.de>
12254 PR tree-optimization/101159
12255 * tree-vect-patterns.c (vect_recog_popcount_pattern): Add
12256 missing NULL vectype check.
12258 2021-06-22 Richard Biener <rguenther@suse.de>
12260 PR tree-optimization/101154
12261 * tree-vect-slp.c (vect_build_slp_tree_2): Fix out-of-bound access.
12263 2021-06-22 Jakub Jelinek <jakub@redhat.com>
12266 * config/i386/i386-protos.h (ix86_last_zero_store_uid): Declare.
12267 * config/i386/i386-expand.c (ix86_last_zero_store_uid): New variable.
12268 * config/i386/i386.c (ix86_expand_prologue): Clear it.
12269 * config/i386/i386.md (peephole2s for 1/2/4 stores of const0_rtx):
12270 Remove "" from match_operand. Emit new insns using emit_move_insn and
12271 set ix86_last_zero_store_uid to INSN_UID of the last store.
12272 Add peephole2s for 1/2/4 stores of const0_rtx following previous
12275 2021-06-22 Martin Liska <mliska@suse.cz>
12277 * auto-profile.c (AUTO_PROFILE_VERSION): Bump as string format
12280 2021-06-22 Martin Liska <mliska@suse.cz>
12282 * gcov-io.h: Remove padding entries.
12284 2021-06-22 liuhongt <hongtao.liu@intel.com>
12286 PR tree-optimization/97770
12287 * tree-vect-patterns.c (vect_recog_popcount_pattern):
12289 (vect_recog_func vect_vect_recog_func_ptrs): Add new pattern.
12291 2021-06-22 liuhongt <hongtao.liu@intel.com>
12294 * config/i386/i386-builtin.def (BDESC): Adjust builtin name.
12295 * config/i386/sse.md (<avx512>_expand<mode>_mask): Rename to ..
12296 (expand<mode>_mask): this ..
12297 (*expand<mode>_mask): New pre_reload splitter to transform
12298 v{,p}expand* to vmov* when mask is zero, all ones, or has all
12299 ones in it's lower part, otherwise still generate
12302 2021-06-22 liuhongt <hongtao.liu@intel.com>
12305 * config/i386/i386-expand.c
12306 (ix86_expand_special_args_builtin): Keep constm1_operand only
12307 if it satisfies insn's operand predicate.
12309 2021-06-21 Jason Merrill <jason@redhat.com>
12312 * df-scan.c (df_ref_record): Check that regno < endregno.
12313 * function.c (assign_parms, expand_function_end): Do nothing with a
12314 TYPE_EMPTY_P result.
12316 2021-06-21 Richard Biener <rguenther@suse.de>
12318 PR tree-optimization/101120
12319 * tree-vect-data-refs.c (bump_vector_ptr): Fold the
12321 * tree-vect-slp.c (vect_transform_slp_perm_load): Add
12322 DR chain DCE capability.
12323 * tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
12324 * tree-vect-stmts.c (vectorizable_load): Remove unused
12325 loads in the DR chain for SLP.
12327 2021-06-21 Jakub Jelinek <jakub@redhat.com>
12329 PR inline-asm/100785
12330 * gimplify.c (gimplify_asm_expr): Don't diagnose errors if
12331 output or input operands were already error_mark_node.
12332 * cfgexpand.c (expand_asm_stmt): If errors are emitted,
12333 remove all inputs, outputs and clobbers from the asm and
12334 set template to "".
12336 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
12338 * config/arm/arm_neon.h (vceq_s8): Replace builtin with __a == __b.
12339 (vceq_s16): Likewise.
12340 (vceq_s32): Likewise.
12341 (vceq_u8): Likewise.
12342 (vceq_u16): Likewise.
12343 (vceq_u32): Likewise.
12344 (vceq_p8): Likewise.
12345 (vceqq_s8): Likewise.
12346 (vceqq_s16): Likewise.
12347 (vceqq_s32): Likewise.
12348 (vceqq_u8): Likewise.
12349 (vceqq_u16): Likewise.
12350 (vceqq_u32): Likewise.
12351 (vceqq_p8): Likewise.
12352 (vceq_f32): Gate __a == __b on __FAST_MATH__.
12353 (vceqq_f32): Likewise.
12354 (vceq_f16): Likewise.
12355 (vceqq_f16): Likewise.
12357 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
12360 * config/arm/iterators.md (NEON_VACMP): Remove.
12361 * config/arm/neon.md (neon_vca<cmp_op><mode>): Use GLTE instead of GTGE
12363 (neon_vca<cmp_op><mode>_insn): Likewise.
12364 (neon_vca<cmp_op_unsp><mode>_insn_unspec): Use NEON_VAGLTE instead of
12367 2021-06-21 Richard Biener <rguenther@suse.de>
12369 PR tree-optimization/101121
12370 * tree-vect-slp.c (vect_build_slp_tree_2): To not fail fatally
12371 when we just lack a stmt with the desired op when doing permutation.
12372 (vect_build_slp_tree): When caching a failed SLP build attempt
12373 assert that at least one lane is marked as not matching.
12375 2021-06-21 liuhongt <hongtao.liu@intel.com>
12378 * config/i386/i386.md: (*anddi_1): Disparage slightly the mask
12379 register alternative.
12380 (*and<mode>_1): Ditto.
12382 (*andn<mode>_1): Ditto.
12383 (*<code><mode>_1): Ditto.
12384 (*<code>qi_1): Ditto.
12385 (*one_cmpl<mode>2_1): Ditto.
12386 (*one_cmplsi2_1_zext): Ditto.
12387 (*one_cmplqi2_1): Ditto.
12388 * config/i386/i386.c (x86_order_regs_for_local_alloc): Change
12389 the order of mask registers to be before general registers.
12391 2021-06-21 Roger Sayle <roger@nextmovesoftware.com>
12394 * config/i386/i386.md: New define_peephole2s to shrink writing
12395 1, 2 or 4 consecutive zeros to memory when optimizing for size.
12397 2021-06-18 Jeff Law <jeffreyalaw@gmail.com>
12399 * config/h8300/h8300.c (h8300_select_cc_mode): Handle SYMBOL_REF.
12400 * config/h8300/logical.md (<code><mode>3 logcial expander): Generate
12401 more efficient code when the source can be trivially simplified.
12403 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
12405 * gimple-range-cache.cc (ranger_cache::range_of_def): Calculate
12406 a range if global is not available.
12407 (ranger_cache::entry_range): Fallback to range_of_def.
12408 * gimple-range-cache.h (range_of_def): Adjust prototype.
12410 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
12412 PR tree-optimization/101014
12413 * gimple-range-cache.cc (ranger_cache::ranger_cache): Remove poor
12415 (ranger_cache::~ranger_cache): Ditto.
12416 (ranger_cache::enable_new_values): Delete.
12417 (ranger_cache::push_poor_value): Delete.
12418 (ranger_cache::range_of_def): Remove poor value processing.
12419 (ranger_cache::entry_range): Ditto.
12420 (ranger_cache::fill_block_cache): Ditto.
12421 * gimple-range-cache.h (class ranger_cache): Remove poor value members.
12422 * gimple-range.cc (gimple_ranger::range_of_expr): Remove call.
12423 * gimple-range.h (class gimple_ranger): Adjust.
12425 2021-06-18 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
12428 * common/config/arm/arm-common.c (arm_canon_arch_option_1): New function
12429 derived from arm_canon_arch.
12430 (arm_canon_arch_option): Call it.
12431 (arm_canon_arch_multilib_option): New function.
12432 * config/arm/arm-cpus.in (IGNORE_FOR_MULTILIB): New fgroup.
12433 * config/arm/arm.h (arm_canon_arch_multilib_option): New prototype.
12434 (CANON_ARCH_MULTILIB_SPEC_FUNCTION): New macro.
12435 (MULTILIB_ARCH_CANONICAL_SPECS): New macro.
12436 (DRIVER_SELF_SPECS): Add MULTILIB_ARCH_CANONICAL_SPECS.
12437 * config/arm/arm.opt (mlibarch): New option.
12438 * config/arm/t-rmprofile (MULTILIB_MATCHES): For armv8*-m, replace use
12439 of march on RHS with mlibarch.
12441 2021-06-18 Marcel Vollweiler <marcel@codesourcery.com>
12443 * config.in: Regenerate.
12444 * config/gcn/gcn.c (print_operand_address): Fix for global_load assembler
12446 * configure: Regenerate.
12447 * configure.ac: Fix for global_load assembler functions.
12449 2021-06-18 Richard Biener <rguenther@suse.de>
12451 PR tree-optimization/101112
12452 * tree-vect-slp.c (vect_slp_linearize_chain): Fix condition
12453 to lookup a pattern stmt def.
12455 2021-06-18 Jakub Jelinek <jakub@redhat.com>
12457 PR middle-end/101062
12458 * stor-layout.c (finish_bitfield_layout): Don't add bitfield
12459 representatives in QUAL_UNION_TYPE.
12461 2021-06-18 Andrew Pinski <apinski@marvell.com>
12463 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
12464 Add counting of how many times it is done.
12465 (factor_out_conditional_conversion): Likewise.
12466 (match_simplify_replacement): Likewise.
12467 (value_replacement): Likewise.
12468 (spaceship_replacement): Likewise.
12469 (cond_store_replacement): Likewise.
12470 (cond_if_else_store_replacement_1): Likewise.
12471 (hoist_adjacent_loads): Likewise.
12473 2021-06-18 Andrew Pinski <apinski@marvell.com>
12475 * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
12476 types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and CNONJ_EXPR.
12477 (verify_gimple_assign_binary): Reject point and offset types on
12478 MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
12479 FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
12480 FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.
12482 2021-06-18 Michael Meissner <meissner@linux.ibm.com>
12484 * config/rs6000/rs6000.c (rs6000_emit_minmax): Add support for ISA
12485 3.1 IEEE 128-bit floating point xsmaxcqp/xsmincqp instructions.
12486 * config/rs6000/rs6000.md (s<minmax><mode>3, IEEE128 iterator):
12489 2021-06-17 Aaron Sawdey <acsawdey@linux.ibm.com>
12491 * config/rs6000/genfusion.pl (gen_logical_addsubf): Add
12492 earlyclobber to alts 0/1.
12493 (gen_addadd): Add earlyclobber to alts 0/1.
12494 * config/rs6000/fusion.md: Regenerate file.
12496 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12498 * cfgloopanal.c (get_loop_hot_path): Make path an auto_vec.
12500 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
12502 * gimple-range-cache.cc: Comment cleanups.
12503 * gimple-range-gori.cc: Comment cleanups.
12504 * gimple-range.cc: Comment/spacing cleanups
12505 * value-range.h: Comment cleanups.
12507 2021-06-17 H.J. Lu <hjl.tools@gmail.com>
12510 * calls.c (expand_call): Replace PUSH_ARGS with
12511 targetm.calls.push_argument (0).
12512 (emit_library_call_value_1): Likewise.
12513 * defaults.h (PUSH_ARGS): Removed.
12514 (PUSH_ARGS_REVERSED): Replace PUSH_ARGS with
12515 targetm.calls.push_argument (0).
12516 * expr.c (block_move_libcall_safe_for_call_parm): Likewise.
12517 (emit_push_insn): Pass the number bytes to push to
12518 targetm.calls.push_argument and pass 0 if ARGS_ADDR is 0.
12519 * hooks.c (hook_bool_uint_true): New.
12520 * hooks.h (hook_bool_uint_true): Likewise.
12521 * rtlanal.c (nonzero_bits1): Replace PUSH_ARGS with
12522 targetm.calls.push_argument (0).
12523 * target.def (push_argument): Add a targetm.calls hook.
12524 * targhooks.c (default_push_argument): New.
12525 * targhooks.h (default_push_argument): Likewise.
12526 * config/bpf/bpf.h (PUSH_ARGS): Removed.
12527 * config/cr16/cr16.c (TARGET_PUSH_ARGUMENT): New.
12528 * config/cr16/cr16.h (PUSH_ARGS): Removed.
12529 * config/i386/i386.c (ix86_push_argument): New.
12530 (TARGET_PUSH_ARGUMENT): Likewise.
12531 * config/i386/i386.h (PUSH_ARGS): Removed.
12532 * config/m32c/m32c.c (TARGET_PUSH_ARGUMENT): New.
12533 * config/m32c/m32c.h (PUSH_ARGS): Removed.
12534 * config/nios2/nios2.h (PUSH_ARGS): Likewise.
12535 * config/pru/pru.h (PUSH_ARGS): Likewise.
12536 * doc/tm.texi.in: Remove PUSH_ARGS documentation. Add
12537 TARGET_PUSH_ARGUMENT hook.
12538 * doc/tm.texi: Regenerated.
12540 2021-06-17 Uroš Bizjak <ubizjak@gmail.com>
12543 * config/i386/i386-expand.c (expand_vector_set_var):
12544 Handle V2FS mode remapping. Pass TARGET_MMX_WITH_SSE to
12545 ix86_expand_vector_init_duplicate.
12546 (ix86_expand_vector_init_duplicate): Emit insv_1 for
12547 QImode for !TARGET_PARTIAL_REG_STALL.
12548 * config/i386/predicates.md (vec_setm_mmx_operand): New predicate.
12549 * config/i386/mmx.md (vec_setv2sf): Use vec_setm_mmx_operand
12550 as operand 2 predicate. Call ix86_expand_vector_set_var
12551 for non-constant index operand.
12552 (vec_setv2si): Ditto.
12553 (vec_setv4hi): Ditto.
12554 (vec_setv8qi): ditto.
12556 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
12558 PR tree-optimization/100790
12559 * gimple-range.cc (range_of_builtin_call): Cleanup clz and ctz
12562 2021-06-17 Martin Liska <mliska@suse.cz>
12564 * doc/invoke.texi: Use consistently -O1 instead of -O.
12566 2021-06-17 Martin Liska <mliska@suse.cz>
12568 * gcov-io.h: Update documentation entry about string format.
12570 2021-06-17 Marius Hillenbrand <mhillen@linux.ibm.com>
12573 * config/s390/vecintrin.h (vec_doublee): Fix to use
12574 __builtin_s390_vflls.
12575 (vec_floate): Fix to use __builtin_s390_vflrd.
12577 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12579 * dominance.c (get_dominated_to_depth): Return auto_vec<basic_block>.
12580 * dominance.h (get_dominated_to_depth): Likewise.
12581 (get_all_dominated_blocks): Likewise.
12582 * cfgcleanup.c (delete_unreachable_blocks): Adjust.
12583 * gcse.c (hoist_code): Likewise.
12584 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
12585 * tree-parloops.c (oacc_entry_exit_ok): Likewise.
12586 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Likewise.
12587 * tree-ssa-phiprop.c (pass_phiprop::execute): Likewise.
12589 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12591 * dominance.c (get_dominated_by_region): Return auto_vec<basic_block>.
12592 * dominance.h (get_dominated_by_region): Likewise.
12593 * tree-cfg.c (gimple_duplicate_sese_region): Adjust.
12594 (gimple_duplicate_sese_tail): Likewise.
12595 (move_sese_region_to_fn): Likewise.
12597 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12599 * dominance.c (get_dominated_by): Return auto_vec<basic_block>.
12600 * dominance.h (get_dominated_by): Likewise.
12601 * auto-profile.c (afdo_find_equiv_class): Adjust.
12602 * cfgloopmanip.c (duplicate_loop_to_header_edge): Likewise.
12603 * loop-unroll.c (unroll_loop_runtime_iterations): Likewise.
12604 * tree-cfg.c (test_linear_chain): Likewise.
12605 (test_diamond): Likewise.
12607 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12609 * cfgloop.h (get_loop_hot_path): Return auto_vec<basic_block>.
12610 * cfgloopanal.c (get_loop_hot_path): Likewise.
12611 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
12613 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12615 * cgraph.c (cgraph_node::collect_callers): Return
12616 auto_vec<cgraph_edge *>.
12617 * cgraph.h (cgraph_node::collect_callers): Likewise.
12618 * ipa-cp.c (create_specialized_node): Adjust.
12619 (decide_about_value): Likewise.
12620 (decide_whether_version_node): Likewise.
12621 * ipa-sra.c (process_isra_node_results): Likewise.
12623 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12625 * vec.h (vl_ptr>::using_auto_storage): Handle null m_vec.
12626 (auto_vec<T, 0>::auto_vec): Define move constructor, and delete copy
12628 (auto_vec<T, 0>::operator=): Define move assignment and delete copy
12631 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
12633 * gimple-range.cc (debug_seed_ranger): New.
12634 (dump_ranger): New.
12635 (debug_ranger): New.
12637 2021-06-17 Richard Biener <rguenther@suse.de>
12639 PR tree-optimization/54400
12640 * tree-vectorizer.h (enum slp_instance_kind): Add
12641 slp_inst_kind_bb_reduc.
12642 (reduction_fn_for_scalar_code): Declare.
12643 * tree-vect-data-refs.c (vect_slp_analyze_instance_dependence):
12644 Check SLP_INSTANCE_KIND instead of looking at the
12646 (vect_slp_analyze_instance_alignment): Likewise.
12647 * tree-vect-loop.c (reduction_fn_for_scalar_code): Export.
12648 * tree-vect-slp.c (vect_slp_linearize_chain): Split out
12649 chain linearization from vect_build_slp_tree_2 and generalize
12650 for the use of BB reduction vectorization.
12651 (vect_build_slp_tree_2): Adjust accordingly.
12652 (vect_optimize_slp): Elide permutes at the root of BB reduction
12654 (vectorizable_bb_reduc_epilogue): New function.
12655 (vect_slp_prune_covered_roots): Likewise.
12656 (vect_slp_analyze_operations): Use them.
12657 (vect_slp_check_for_constructors): Recognize associatable
12658 chains for BB reduction vectorization.
12659 (vectorize_slp_instance_root_stmt): Generate code for the
12660 BB reduction epilogue.
12662 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
12664 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Check with
12666 (gori_compute::may_recompute_p): New.
12667 (gori_compute::outgoing_edge_range_p): Perform recomputations.
12668 * gimple-range-gori.h (class gori_compute): Add prototype.
12670 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
12672 * gimple-range-cache.cc (ranger_cache::range_on_edge): Always return
12673 true when a range can be calculated.
12674 * gimple-range.cc (gimple_ranger::dump_bb): Check has_edge_range_p.
12676 2021-06-16 Martin Sebor <msebor@redhat.com>
12678 * doc/invoke.texi (-Wmismatched-dealloc, -Wmismatched-new-delete):
12679 Correct documented defaults.
12681 2021-06-16 Andrew MacLeod <amacleod@redhat.com>
12683 * gimple-range-cache.cc (ranger_cache::ranger_cache): Initialize
12684 m_new_value_p directly.
12686 2021-06-16 Uroš Bizjak <ubizjak@gmail.com>
12689 * config/i386/i386-expand.c (expand_vec_perm_2perm_pblendv):
12690 Handle 64bit modes for TARGET_SSE4_1.
12691 (expand_vec_perm_pshufb2): Handle 64bit modes for TARGET_SSSE3.
12692 (expand_vec_perm_even_odd_pack): Handle V4HI mode.
12693 (expand_vec_perm_even_odd_1) <case E_V4HImode>: Expand via
12694 expand_vec_perm_pshufb2 for TARGET_SSSE3 and via
12695 expand_vec_perm_even_odd_pack for TARGET_SSE4_1.
12696 * config/i386/mmx.md (mmx_packusdw): New insn pattern.
12698 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12700 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn<mode>):
12701 Change to an expander that emits the correct instruction
12702 depending on endianness.
12703 (aarch64_<sur><addsub>hn<mode>_insn_le): Define.
12704 (aarch64_<sur><addsub>hn<mode>_insn_be): Define.
12706 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12708 * config/aarch64/aarch64-simd-builtins.def: Split generator
12709 for aarch64_<su>qmovn builtins into scalar and vector
12711 * config/aarch64/aarch64-simd.md (aarch64_<su>qmovn<mode>_insn_le):
12713 (aarch64_<su>qmovn<mode>_insn_be): Define.
12714 (aarch64_<su>qmovn<mode>): Split into scalar and vector
12715 variants. Change vector variant to an expander that emits the
12716 correct instruction depending on endianness.
12718 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12720 * config/aarch64/aarch64-simd-builtins.def: Split generator
12721 for aarch64_sqmovun builtins into scalar and vector variants.
12722 * config/aarch64/aarch64-simd.md (aarch64_sqmovun<mode>):
12723 Split into scalar and vector variants. Change vector variant
12724 to an expander that emits the correct instruction depending
12726 (aarch64_sqmovun<mode>_insn_le): Define.
12727 (aarch64_sqmovun<mode>_insn_be): Define.
12729 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12731 * config/aarch64/aarch64-simd.md (aarch64_xtn<mode>_insn_le):
12732 Define - modeling zero-high-half semantics.
12733 (aarch64_xtn<mode>): Change to an expander that emits the
12734 appropriate instruction depending on endianness.
12735 (aarch64_xtn<mode>_insn_be): Define - modeling zero-high-half
12737 (aarch64_xtn2<mode>_le): Rename to...
12738 (aarch64_xtn2<mode>_insn_le): This.
12739 (aarch64_xtn2<mode>_be): Rename to...
12740 (aarch64_xtn2<mode>_insn_be): This.
12741 (vec_pack_trunc_<mode>): Emit truncation instruction instead
12743 * config/aarch64/iterators.md (Vnarrowd): Add Vnarrowd mode
12744 attribute iterator.
12746 2021-06-16 Martin Jambor <mjambor@suse.cz>
12748 PR tree-optimization/100453
12749 * tree-sra.c (create_access): Disqualify any const candidates
12750 which are written to.
12751 (sra_modify_expr): Do not store sub-replacements back to a const base.
12752 (handle_unscalarized_data_in_subtree): Likewise.
12753 (sra_modify_assign): Likewise. Earlier, use TREE_READONLy test
12754 instead of constant_decl_p.
12756 2021-06-16 Jakub Jelinek <jakub@redhat.com>
12758 PR middle-end/101062
12759 * stor-layout.c (finish_bitfield_representative): For fields in unions
12760 assume nextf is always NULL.
12761 (finish_bitfield_layout): Compute bit field representatives also in
12762 unions, but handle it as if each bitfield was the only field in the
12765 2021-06-16 Richard Biener <rguenther@suse.de>
12767 PR tree-optimization/101088
12768 * tree-ssa-loop-im.c (sm_seq_valid_bb): Only look for
12769 supported refs on edges. Do not assert same ref but
12770 different kind stores are unsuported but mark them so.
12771 (hoist_memory_references): Only look for supported refs
12774 2021-06-16 Roger Sayle <roger@nextmovesoftware.com>
12776 PR rtl-optimization/46235
12777 * config/i386/i386.md: New define_split for bt followed by cmov.
12778 (*bt<mode>_setcqi): New define_insn_and_split for bt followed by setc.
12779 (*bt<mode>_setncqi): New define_insn_and_split for bt then setnc.
12780 (*bt<mode>_setnc<mode>): New define_insn_and_split for bt followed
12781 by setnc with zero extension.
12783 2021-06-16 Richard Biener <rguenther@suse.de>
12785 PR tree-optimization/101083
12786 * tree-vect-slp.c (vect_slp_build_two_operator_nodes): Get
12787 vectype as argument.
12788 (vect_build_slp_tree_2): Adjust.
12790 2021-06-15 Martin Sebor <msebor@redhat.com>
12792 PR middle-end/100876
12793 * builtins.c: (gimple_call_return_array): Account for size_t
12794 mangling as either unsigned int or unsigned long
12796 2021-06-15 Jeff Law <jeffreyalaw@gmail.com>
12798 * compare-elim.c (try_eliminate_compare): Run DCE to clean things
12799 up before eliminating comparisons.
12801 2021-06-15 Aldy Hernandez <aldyh@redhat.com>
12803 * range-op.cc (operator_bitwise_or::wi_fold): Make sure
12804 nonzero|X is nonzero.
12805 (range_op_bitwise_and_tests): Add tests for above.
12807 2021-06-15 Carl Love <cel@us.ibm.com>
12810 * config/rs6000/rs6000-builtin.def (VCMPEQUT): Fix the ICODE for the
12812 (VRLQ, VSLQ, VSRQ, VSRAQ): Remove unused BU_P10_OVERLOAD_2
12815 2021-06-15 Tobias Burnus <tobias@codesourcery.com>
12818 * gimplify.c (enum gimplify_defaultmap_kind): Add GDMK_SCALAR_TARGET.
12819 (struct gimplify_omp_ctx): Extend defaultmap array by one.
12820 (new_omp_context): Init defaultmap[GDMK_SCALAR_TARGET].
12821 (omp_notice_variable): Update type classification for Fortran.
12822 (gimplify_scan_omp_clauses): Update calls for new argument; handle
12823 GDMK_SCALAR_TARGET; for Fortran, GDMK_POINTER avoid GOVD_MAP_0LEN_ARRAY.
12824 * langhooks-def.h (lhd_omp_scalar_p): Add 'ptr_ok' argument.
12825 * langhooks.c (lhd_omp_scalar_p): Likewise.
12826 (LANG_HOOKS_OMP_ALLOCATABLE_P, LANG_HOOKS_OMP_SCALAR_TARGET_P): New.
12827 (LANG_HOOKS_DECLS): Add them.
12828 * langhooks.h (struct lang_hooks_for_decls): Add new hooks, update
12829 omp_scalar_p pointer type to include the new bool argument.
12831 2021-06-15 David Malcolm <dmalcolm@redhat.com>
12833 * doc/analyzer.texi
12834 (Special Functions for Debugging the Analyzer): Add
12835 __analyzer_dump_capacity.
12837 2021-06-15 Jakub Jelinek <jakub@redhat.com>
12840 * expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
12841 case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.
12843 2021-06-15 Richard Biener <rguenther@suse.de>
12845 * cfgloopanal.c (mark_irreducible_loops): Use a dominance
12846 check to identify loop latches.
12847 * cfgloop.c (verify_loop_structure): Likewise.
12848 * loop-init.c (apply_loop_flags): Allow marked irreducible
12849 regions even with multiple latches.
12850 * predict.c (rebuild_frequencies): Simplify.
12852 2021-06-15 Richard Biener <rguenther@suse.de>
12854 * tree-ssa-threadupdate.c
12855 (jump_thread_path_registry::mark_threaded_blocks): Assert we
12856 have marked irreducible regions.
12858 2021-06-14 Martin Sebor <msebor@redhat.com>
12861 * builtins.c (gimple_call_return_array): Check for attribute fn spec.
12862 Handle calls to placement new.
12863 (ndecl_dealloc_argno): Avoid placement delete.
12865 2021-06-14 Peter Bergner <bergner@linux.ibm.com>
12868 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
12869 create_tmp_reg_or_ssa_name().
12871 2021-06-14 Andrew MacLeod <amacleod@redhat.com>
12873 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
12874 (ranger_cache::enable_new_values): Set to specified value and
12875 return the old value.
12876 (ranger_cache::disable_new_values): Delete.
12877 (ranger_cache::fill_block_cache): Disable non 1st order derived
12879 * gimple-range-cache.h (ranger_cache): Adjust prototypes.
12880 * gimple-range.cc (gimple_ranger::range_of_expr): Adjust.
12882 2021-06-14 Uroš Bizjak <ubizjak@gmail.com>
12885 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
12886 Return true early when testing with V2HImode.
12887 * config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.
12889 2021-06-14 Christophe Lyon <christophe.lyon@linaro.org>
12891 * config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
12892 (mve_vec_unpack<US>_hi_<mode>): New pattern.
12893 (@mve_vec_pack_trunc_lo_<mode>): New pattern.
12894 (mve_vmovntq_<supf><mode>): Prefix with '@'.
12895 * config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
12897 (vec_unpack<US>_lo_<mode>): Likewise.
12898 (vec_pack_trunc_<mode>): Rename to
12899 neon_quad_vec_pack_trunc_<mode>.
12900 * config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
12902 (vec_unpack<US>_lo_<mode>): New.
12903 (vec_pack_trunc_<mode>): New.
12905 2021-06-14 Richard Biener <rguenther@suse.de>
12907 PR tree-optimization/100934
12908 * tree-ssa-dom.c (pass_dominator::execute): Properly
12909 mark irreducible regions.
12911 2021-06-14 Martin Liska <mliska@suse.cz>
12913 * doc/invoke.texi: Put r{...} on the same line as @item.
12915 2021-06-14 Martin Liska <mliska@suse.cz>
12917 * doc/invoke.texi: Add missing newline.
12919 2021-06-14 Martin Liska <mliska@suse.cz>
12921 * doc/invoke.texi: Remove '+' charasters.
12923 2021-06-14 Claudiu Zissulescu <claziss@synopsys.com>
12925 * config.gcc (arc): Add support for with_cpu option.
12926 * config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.
12928 2021-06-14 Richard Biener <rguenther@suse.de>
12930 PR tree-optimization/101031
12931 * tree-ssa-strlen.c (maybe_invalidate): Increment max_size
12932 instead of size when accounting for a possibly string
12935 2021-06-14 Martin Liska <mliska@suse.cz>
12937 * gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].
12939 2021-06-14 Aldy Hernandez <aldyh@redhat.com>
12941 * value-query.cc (gimple_range_global): Call get_range_global
12942 if called after inlining.
12944 2021-06-13 Uroš Bizjak <ubizjak@gmail.com>
12947 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
12948 Emit constant permutation insn directly from here.
12950 2021-06-13 Trevor Saunders <tbsaunde@tbsaunde.org>
12952 * attribs.c (find_attribute_namespace): Iterate over vec<> with
12954 * auto-profile.c (afdo_find_equiv_class): Likewise.
12955 * gcc.c (do_specs_vec): Likewise.
12956 (do_spec_1): Likewise.
12957 (driver::set_up_specs): Likewise.
12958 * gimple-loop-jam.c (any_access_function_variant_p): Likewise.
12959 * gimple-ssa-store-merging.c (compatible_load_p): Likewise.
12960 (imm_store_chain_info::try_coalesce_bswap): Likewise.
12961 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
12962 (get_location_for_stmts): Likewise.
12963 * graphite-poly.c (print_iteration_domains): Likewise.
12964 (free_poly_bb): Likewise.
12965 (remove_gbbs_in_scop): Likewise.
12966 (free_scop): Likewise.
12967 (dump_gbb_cases): Likewise.
12968 (dump_gbb_conditions): Likewise.
12969 (print_pdrs): Likewise.
12970 (print_scop): Likewise.
12971 * ifcvt.c (cond_move_process_if_block): Likewise.
12972 * lower-subreg.c (decompose_multiword_subregs): Likewise.
12973 * regcprop.c (pass_cprop_hardreg::execute): Likewise.
12974 * sanopt.c (sanitize_rewrite_addressable_params): Likewise.
12975 * sel-sched-dump.c (dump_insn_vector): Likewise.
12976 * store-motion.c (store_ops_ok): Likewise.
12977 (store_killed_in_insn): Likewise.
12978 * timevar.c (timer::named_items::print): Likewise.
12979 * tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
12980 (cleanup_tree_cfg_noloop): Likewise.
12981 * tree-data-ref.c (dump_data_references): Likewise.
12982 (print_dir_vectors): Likewise.
12983 (print_dist_vectors): Likewise.
12984 (dump_data_dependence_relations): Likewise.
12985 (dump_dist_dir_vectors): Likewise.
12986 (dump_ddrs): Likewise.
12987 (create_runtime_alias_checks): Likewise.
12988 (free_subscripts): Likewise.
12989 (save_dist_v): Likewise.
12990 (save_dir_v): Likewise.
12991 (invariant_access_functions): Likewise.
12992 (same_access_functions): Likewise.
12993 (access_functions_are_affine_or_constant_p): Likewise.
12994 (find_data_references_in_stmt): Likewise.
12995 (graphite_find_data_references_in_stmt): Likewise.
12996 (free_dependence_relations): Likewise.
12997 (free_data_refs): Likewise.
12998 * tree-inline.c (copy_debug_stmts): Likewise.
12999 * tree-into-ssa.c (dump_currdefs): Likewise.
13000 (rewrite_update_phi_arguments): Likewise.
13001 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
13002 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
13004 (vect_slp_analyze_node_dependences): Likewise.
13005 (vect_slp_analyze_instance_dependence): Likewise.
13006 (vect_record_base_alignments): Likewise.
13007 (vect_get_peeling_costs_all_drs): Likewise.
13008 (vect_peeling_supportable): Likewise.
13009 * tree-vectorizer.c (vec_info::~vec_info): Likewise.
13010 (vec_info::free_stmt_vec_infos): Likewise.
13012 2021-06-13 Jeff Law <jeffreyalaw@gmail.com>
13014 * config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
13015 (andqi3_1<cczn>): Removed.
13016 (<ors>qi3_1): Do not split for IOR/XOR a single bit.
13017 (H8/SX bit logicals): Split out from other patterns.
13018 * config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
13019 mulqihi3_const_clobber_flags.
13020 (mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly
13022 2021-06-13 H.J. Lu <hjl.tools@gmail.com>
13025 * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
13026 to true if red zone is used.
13027 (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
13028 ix86_red_zone_used.
13029 * config/i386/i386.h (machine_function): Add red_zone_used.
13030 (ix86_red_zone_size): Removed.
13031 (ix86_red_zone_used): New.
13032 * config/i386/i386.md (peephole2 patterns): Replace
13033 ix86_red_zone_size with ix86_red_zone_used.
13035 2021-06-12 Jason Merrill <jason@redhat.com>
13037 * doc/extend.texi (unused variable attribute): Applies to
13038 structure fields as well.
13040 2021-06-12 Eugene Rozenfeld <erozen@microsoft.com>
13042 * auto-profile.c (read_profile): fix a typo in an error string
13044 2021-06-11 Thomas Schwinge <thomas@codesourcery.com>
13046 * tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
13048 * tree-pretty-print.c (dump_omp_clauses): Update.
13049 (dump_generic_node) <OMP_CLAUSE>: Use it.
13051 2021-06-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
13054 * config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
13055 int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
13056 the polymorphic variants matching code.
13057 (__arm_vld1q_z): Likewise.
13058 (__arm_vld2q): Likewise.
13059 (__arm_vld4q): Likewise.
13060 (__arm_vldrbq_gather_offset): Likewise.
13061 (__arm_vldrbq_gather_offset_z): Likewise.
13063 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
13065 PR tree-optimization/96392
13066 * fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.
13068 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
13070 PR tree-optimization/96392
13071 * fold-const.c (fold_real_zero_addition_p): Take both arguments
13072 of the addition or subtraction, not just the zero. Use this
13073 other argument in tests for signaling NaNs and signed zeros.
13074 (tree_expr_maybe_real_minus_zero_p): New predicate.
13075 * fold-const.h (fold_real_zero_addition_p): Update prototype.
13076 (tree_expr_maybe_real_minus_zero_p): New function prototype.
13077 * match.pd: Update calls to fold_real_zero_addition_p.
13078 Replace HONOR_NANS with tree_expr_maybe_nan_p.
13079 Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
13080 Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
13081 * tree-ssa-reassoc.c (eliminate_using_constants): Update
13082 call to fold_real_zero_addition_p.
13084 2021-06-11 Richard Biener <rguenther@suse.de>
13086 PR tree-optimization/101025
13087 * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
13088 all refs that require dependence checking.
13090 2021-06-11 Richard Biener <rguenther@suse.de>
13092 PR tree-optimization/101028
13093 * tree-vect-slp.c (vect_build_slp_tree_2): When SLP
13094 reassoc discovery fails fatally, mark appropriate lanes
13097 2021-06-11 Richard Biener <rguenther@suse.de>
13099 PR tree-optimization/101026
13100 * tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
13101 have a representative for the associated chain nodes.
13103 2021-06-11 Jakub Jelinek <jakub@redhat.com>
13105 PR rtl-optimization/101008
13106 * simplify-rtx.c (relational_result): New function.
13107 (simplify_logical_relational_operation,
13108 simplify_relational_operation): Use it.
13110 2021-06-11 Jakub Jelinek <jakub@redhat.com>
13113 * config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.
13115 2021-06-11 Uroš Bizjak <ubizjak@gmail.com>
13118 * config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
13119 false if the permutation can be implemented with constant
13120 permutation instruction in wider mode.
13121 (canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
13122 Handle V8QImode and V4HImode.
13124 2021-06-11 Martin Liska <mliska@suse.cz>
13126 PR gcov-profile/100788
13127 * common.opt: Add new option.
13128 * coverage.c (coverage_begin_function): Emit warning instead on
13129 the internal compiler error.
13130 * doc/invoke.texi: Document the option.
13131 * toplev.c (process_options): Enable it by default.
13133 2021-06-11 Richard Biener <rguenther@suse.de>
13135 PR middle-end/101009
13136 * tree-data-ref.c (build_classic_dist_vector_1): Make sure
13137 to set *init_b to true when we encounter a constant equal
13139 (compute_affine_dependence): Also dump the actual DR_REF.
13141 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
13143 PR tree-optimization/100984
13144 * gimple-ssa-evrp.c (ssa_equiv_stack): Use auto_vec for
13145 replacements table.
13146 (ssa_equiv_stack::~ssa_equiv_stack): Remove.
13148 2021-06-11 Kewen Lin <linkw@linux.ibm.com>
13150 * config/rs6000/rs6000.md
13151 (floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
13152 define_insn_and_split.
13154 2021-06-11 Richard Biener <rguenther@suse.de>
13156 * tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
13157 to sort operands of the associative chain.
13159 2021-06-11 Richard Biener <rguenther@suse.de>
13161 * system.h (gcc_stablesort_r): Declare.
13162 * sort.cc (gcc_sort_r): Support stable sort.
13163 (gcc_stablesort_r): Define.
13164 * vec.h (vec<>::stablesort): Add.
13166 2021-06-10 Uroš Bizjak <ubizjak@gmail.com>
13169 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
13170 Handle V2SF mode. Emit SHUFPS to fixup unpack-high for V2SF mode.
13171 (expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
13172 (expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
13173 (expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
13174 (expand_vec_perm_interleave2): Handle 64bit modes.
13175 (expand_vec_perm_even_odd_pack): Handle V8QI mode.
13176 (expand_vec_perm_even_odd_1): Ditto.
13177 (ix86_vectorize_vec_perm_const): Ditto.
13178 * config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
13179 * config/i386/sse.md: ... here.
13180 * config/i386/mmx.md (*vec_interleave_lowv2sf):
13181 New insn_and_split pattern.
13182 (*vec_interleave_highv2sf): Ditto.
13183 (mmx_pshufbv8qi3): New insn pattern.
13184 (*mmx_pblendw): Ditto.
13186 2021-06-10 Peter Bergner <bergner@linux.ibm.com>
13188 * config/rs6000/rs6000-builtin.def (build_pair): New built-in.
13189 (build_acc): Likewise.
13190 * config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
13191 source operands in little-endian mode.
13192 (rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
13193 (mma_init_builtins): Likewise.
13194 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
13195 ordering for the MMA assemble and build source operands.
13196 * doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
13198 (__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
13201 2021-06-10 Jeff Law <jeffreyalaw@gmail.com>
13203 * config/h8300/h8300.c (select_cc_mode): Handle MEM. Use
13205 * config/h8300/extensions.md: Replace _clobber_flags patterns
13208 2021-06-10 Robin Dapp <rdapp@linux.ibm.com>
13210 * config/s390/vector.md (vcond_mask_<mode><mode>): Change to
13211 (vcond_mask_<mode><tointvec>): this.
13213 2021-06-10 Andrew Stubbs <ams@codesourcery.com>
13214 Thomas Schwinge <thomas@codesourcery.com>
13216 * omp-builtins.def (BUILT_IN_GOACC_ENTER_EXIT_DATA): Split into...
13217 (BUILT_IN_GOACC_ENTER_DATA, BUILT_IN_GOACC_EXIT_DATA): ... these.
13218 * gimple.h (enum gf_mask): Split
13219 'GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA' into
13220 'GF_OMP_TARGET_KIND_OACC_ENTER_DATA' and
13221 'GF_OMP_TARGET_KIND_OACC_EXIT_DATA'.
13222 (is_gimple_omp_oacc): Update.
13223 * gimple-pretty-print.c (dump_gimple_omp_target): Likewise.
13224 * gimplify.c (gimplify_omp_target_update): Likewise.
13225 * omp-expand.c (expand_omp_target, build_omp_regions_1)
13226 (omp_make_gimple_edges): Likewise.
13227 * omp-low.c (check_omp_nesting_restrictions, lower_omp_target):
13230 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
13232 * value-query.cc (value_query::value_on_edge): Rename name to
13234 (range_query::range_on_edge): Same.
13235 (range_query::value_of_expr): Same.
13236 (range_query::value_on_edge): Same.
13237 * value-query.h (class value_query): Same.
13238 (class range_query): Same.
13240 2021-06-10 Richard Biener <rguenther@suse.de>
13242 PR tree-optimization/101003
13243 * tree-vect-slp.c (vect_build_slp_tree_2): Appropriately
13244 use the pattern stmt defs when linearizing a chain.
13246 2021-06-10 Jakub Jelinek <jakub@redhat.com>
13249 * ifcvt.c (noce_get_alt_condition, noce_try_abs): Use
13250 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
13252 2021-06-10 Clement Chigot <clement.chigot@atos.net>
13254 * config/rs6000/aix71.h (ASM_CPU_SPEC): Add Power10 directive.
13255 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
13257 2021-06-09 Andrew Pinski <apinski@marvell.com>
13259 PR tree-optimization/100925
13260 * match.pd (a ? CST1 : CST2): Limit transformations
13261 that would produce a negative to integeral types only.
13262 Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.
13264 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
13267 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
13269 * doc/tm.texi: Correctly update.
13271 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
13273 * doc/tm.texi: Correctly update.
13275 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
13278 * doc/tm.texi.in (Trampolines): Add a missing blank line.
13280 2021-06-09 Paul Eggert <eggert@cs.ucla.edu>
13283 * doc/invoke.texi (Code Gen Options); Document that -fno-trampolines
13284 and -ftrampolines work only with Ada.
13285 * doc/tm.texi.in (Trampolines): Likewise.
13286 * doc/tm.texi: Regenerated.
13288 2021-06-09 Carl Love <cel@us.ibm.com>
13290 * config/rs6000/altivec.h (vec_signextll, vec_signexti, vec_signextq):
13291 Add define for new builtins.
13292 * config/rs6000/altivec.md(altivec_vreveti2): Add define_expand.
13293 * config/rs6000/rs6000-builtin.def (VSIGNEXTI, VSIGNEXTLL): Add
13294 overloaded builtin definitions.
13295 (VSIGNEXTSB2W, VSIGNEXTSH2W, VSIGNEXTSB2D, VSIGNEXTSH2D,VSIGNEXTSW2D,
13296 VSIGNEXTSD2Q): Add builtin expansions.
13297 (SIGNEXT): Add P10 overload definition.
13298 * config/rs6000/rs6000-call.c (P9V_BUILTIN_VEC_VSIGNEXTI, P9V_BUILTIN_VEC_VSIGNEXTLL,
13299 P10_BUILTIN_VEC_SIGNEXT): Add overloaded argument definitions.
13300 * config/rs6000/vsx.md (vsx_sign_extend_v2di_v1ti): Add define_insn.
13301 (vsignextend_v2di_v1ti, vsignextend_qi_<mode>, vsignextend_hi_<mode>,
13302 vsignextend_si_v2di)[VIlong]: Add define_expand.
13303 Make define_insn vsx_sign_extend_si_v2di visible.
13304 * doc/extend.texi: Add documentation for the vec_signexti,
13305 vec_signextll builtins and vec_signextq.
13307 2021-06-09 Carl Love <cel@us.ibm.com>
13309 * config/rs6000/rs6000.c (__fixkfti, __fixunskfti, __floattikf,
13310 __floatuntikf): Names changed to __fixkfti_sw, __fixunskfti_sw,
13311 __floattikf_sw, __floatuntikf_sw respectively.
13312 * config/rs6000/rs6000.md (floatti<mode>2, floatunsti<mode>2,
13313 fix_trunc<mode>ti2, fixuns_trunc<mode>ti2): Add
13314 define_insn for mode IEEE 128.
13316 2021-06-09 Carl Love <cel@us.ibm.com>
13318 * config/rs6000/altivec.md (altivec_vslq, altivec_vsrq):
13319 Rename to altivec_vslq_<mode>, altivec_vsrq_<mode>, mode VEC_TI.
13320 * config/rs6000/vector.md (VEC_TI): Was named VSX_TI in vsx.md.
13321 (vashlv1ti3): Change to vashl<mode>3, mode VEC_TI.
13322 (vlshrv1ti3): Change to vlshr<mode>3, mode VEC_TI.
13323 * config/rs6000/vsx.md (VSX_TI): Remove define_mode_iterator. Update
13324 uses of VSX_TI to VEC_TI.
13326 2021-06-09 Carl Love <cel@us.ibm.com>
13328 * config/rs6000/dfp.md (floattitd2, fixtdti2): New define_insns.
13330 2021-06-09 Carl Love <cel@us.ibm.com>
13332 * config/rs6000/altivec.h (vec_dive, vec_mod): Add define for new
13334 * config/rs6000/altivec.md (UNSPEC_VMULEUD, UNSPEC_VMULESD,
13335 UNSPEC_VMULOUD, UNSPEC_VMULOSD): New unspecs.
13336 (altivec_eqv1ti, altivec_gtv1ti, altivec_gtuv1ti, altivec_vmuleud,
13337 altivec_vmuloud, altivec_vmulesd, altivec_vmulosd, altivec_vrlq,
13338 altivec_vrlqmi, altivec_vrlqmi_inst, altivec_vrlqnm,
13339 altivec_vrlqnm_inst, altivec_vslq, altivec_vsrq, altivec_vsraq,
13340 altivec_vcmpequt_p, altivec_vcmpgtst_p, altivec_vcmpgtut_p): New
13342 (vec_widen_umult_even_v2di, vec_widen_smult_even_v2di,
13343 vec_widen_umult_odd_v2di, vec_widen_smult_odd_v2di, altivec_vrlqmi,
13344 altivec_vrlqnm): New define_expands.
13345 * config/rs6000/rs6000-builtin.def (VCMPEQUT_P, VCMPGTST_P,
13346 VCMPGTUT_P): Add macro expansions.
13347 (BU_P10V_AV_P): Add builtin predicate definition.
13348 (VCMPGTUT, VCMPGTST, VCMPEQUT, CMPNET, CMPGE_1TI,
13349 CMPGE_U1TI, CMPLE_1TI, CMPLE_U1TI, VNOR_V1TI_UNS, VNOR_V1TI, VCMPNET_P,
13350 VCMPAET_P, VMULEUD, VMULESD, VMULOUD, VMULOSD, VRLQ,
13351 VSLQ, VSRQ, VSRAQ, VRLQNM, DIV_V1TI, UDIV_V1TI, DIVES_V1TI, DIVEU_V1TI,
13352 MODS_V1TI, MODU_V1TI, VRLQMI): New macro expansions.
13353 (VRLQ, VSLQ, VSRQ, VSRAQ, DIVE, MOD): New overload expansions.
13354 * config/rs6000/rs6000-call.c (P10_BUILTIN_VCMPEQUT,
13355 P10V_BUILTIN_CMPGE_1TI, P10V_BUILTIN_CMPGE_U1TI,
13356 P10V_BUILTIN_VCMPGTUT, P10V_BUILTIN_VCMPGTST,
13357 P10V_BUILTIN_CMPLE_1TI, P10V_BUILTIN_VCMPLE_U1TI,
13358 P10V_BUILTIN_DIV_V1TI, P10V_BUILTIN_UDIV_V1TI,
13359 P10V_BUILTIN_VMULESD, P10V_BUILTIN_VMULEUD,
13360 P10V_BUILTIN_VMULOSD, P10V_BUILTIN_VMULOUD,
13361 P10V_BUILTIN_VNOR_V1TI, P10V_BUILTIN_VNOR_V1TI_UNS,
13362 P10V_BUILTIN_VRLQ, P10V_BUILTIN_VRLQMI,
13363 P10V_BUILTIN_VRLQNM, P10V_BUILTIN_VSLQ,
13364 P10V_BUILTIN_VSRQ, P10V_BUILTIN_VSRAQ,
13365 P10V_BUILTIN_VCMPGTUT_P, P10V_BUILTIN_VCMPGTST_P,
13366 P10V_BUILTIN_VCMPEQUT_P, P10V_BUILTIN_VCMPGTUT_P,
13367 P10V_BUILTIN_VCMPGTST_P, P10V_BUILTIN_CMPNET,
13368 P10V_BUILTIN_VCMPNET_P, P10V_BUILTIN_VCMPAET_P,
13369 P10V_BUILTIN_DIVES_V1TI, P10V_BUILTIN_MODS_V1TI,
13370 P10V_BUILTIN_MODU_V1TI):
13371 New overloaded definitions.
13372 (rs6000_gimple_fold_builtin) [P10V_BUILTIN_VCMPEQUT,
13373 P10V_BUILTIN_CMPNET, P10V_BUILTIN_CMPGE_1TI,
13374 P10V_BUILTIN_CMPGE_U1TI, P10V_BUILTIN_VCMPGTUT,
13375 P10V_BUILTIN_VCMPGTST, P10V_BUILTIN_CMPLE_1TI,
13376 P10V_BUILTIN_CMPLE_U1TI]: New case statements.
13377 (rs6000_init_builtins) [bool_V1TI_type_node, int_ftype_int_v1ti_v1ti]:
13379 (altivec_init_builtins): New E_V1TImode case statement.
13380 (builtin_function_type)[P10_BUILTIN_128BIT_VMULEUD,
13381 P10_BUILTIN_128BIT_VMULOUD, P10_BUILTIN_128BIT_DIVEU_V1TI,
13382 P10_BUILTIN_128BIT_MODU_V1TI, P10_BUILTIN_CMPGE_U1TI,
13383 P10_BUILTIN_VCMPGTUT, P10_BUILTIN_VCMPEQUT]: New case statements.
13384 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute) [E_TImode,
13385 E_V1TImode]: New case statements.
13386 * config/rs6000/rs6000.h (rs6000_builtin_type_index): New enum
13387 value RS6000_BTI_bool_V1TI.
13388 * config/rs6000/vector.md (vector_gtv1ti,vector_nltv1ti,
13389 vector_gtuv1ti, vector_nltuv1ti, vector_ngtv1ti, vector_ngtuv1ti,
13390 vector_eq_v1ti_p, vector_ne_v1ti_p, vector_ae_v1ti_p,
13391 vector_gt_v1ti_p, vector_gtu_v1ti_p, vrotlv1ti3, vashlv1ti3,
13392 vlshrv1ti3, vashrv1ti3): New define_expands.
13393 * config/rs6000/vsx.md (UNSPEC_VSX_DIVSQ, UNSPEC_VSX_DIVUQ,
13394 UNSPEC_VSX_DIVESQ, UNSPEC_VSX_DIVEUQ, UNSPEC_VSX_MODSQ,
13395 UNSPEC_VSX_MODUQ): New unspecs.
13396 (mulv2di3, vsx_div_v1ti, vsx_udiv_v1ti, vsx_dives_v1ti,
13397 vsx_diveu_v1ti, vsx_mods_v1ti, vsx_modu_v1ti, xxswapd_v1ti): New
13399 (vcmpnet): New define_expand.
13400 * doc/extend.texi: Add documentation for the new builtins vec_rl,
13401 vec_rlmi, vec_rlnm, vec_sl, vec_sr, vec_sra, vec_mule, vec_mulo,
13402 vec_div, vec_dive, vec_mod, vec_cmpeq, vec_cmpne, vec_cmpgt, vec_cmplt,
13403 vec_cmpge, vec_cmple, vec_all_eq, vec_all_ne, vec_all_gt, vec_all_lt,
13404 vec_all_ge, vec_all_le, vec_any_eq, vec_any_ne, vec_any_gt, vec_any_lt,
13405 vec_any_ge, vec_any_le.
13407 2021-06-09 Carl Love <cel@us.ibm.com>
13409 * config/rs6000/altivec.md (altivec_vrl<VI_char>mi): Fix
13410 bug in argument generation.
13412 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
13414 * config/arm/iterators.md (<supf>): Remove VCLZQ_U, VCLZQ_S.
13416 * config/arm/mve.md (mve_vclzq_<supf><mode>): Add '@' prefix,
13417 remove <supf> iterator.
13418 (mve_vclzq_u<mode>): New.
13419 * config/arm/neon.md (clz<mode>2): Rename to neon_vclz<mode>.
13420 (neon_vclz<mode): Move to ...
13421 * config/arm/unspecs.md (VCLZQ_U, VCLZQ_S): Remove.
13422 * config/arm/vec-common.md: ... here. Add support for MVE.
13424 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
13426 * config/arm/mve.md (mve_vhaddq_<supf><mode>): Prefix with '@'.
13427 (@mve_vrhaddq_<supf><mode): Likewise.
13428 * config/arm/neon.md (neon_v<r>hadd<sup><mode>): Likewise.
13429 * config/arm/vec-common.md (avg<mode>3_floor, uavg<mode>3_floor)
13430 (avg<mode>3_ceil", uavg<mode>3_ceil): New patterns.
13432 2021-06-09 imba-tjd <109224573@qq.com>
13434 * doc/invoke.texi: Fix typo.
13436 2021-06-09 Roger Sayle <roger@nextmovesoftware.com>
13438 PR middle-end/53267
13439 * fold-const-call.c (fold_const_call_sss) [CASE_CFN_FMOD]:
13440 Support evaluation of fmod/fmodf/fmodl at compile-time.
13442 2021-06-09 Richard Biener <rguenther@suse.de>
13444 PR tree-optimization/100981
13445 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
13446 gimple_get_lhs to also handle calls.
13447 * tree-vect-slp-patterns.c (complex_pattern::build): Transfer
13450 2021-06-09 Richard Biener <rguenther@suse.de>
13452 PR tree-optimization/97832
13453 * tree-vectorizer.h (_slp_tree::failed): New.
13454 * tree-vect-slp.c (_slp_tree::_slp_tree): Initialize
13456 (_slp_tree::~_slp_tree): Free failed.
13457 (vect_build_slp_tree): Retain failed nodes and record
13458 matches in them, copying that back out when running
13459 into a cached fail. Dump start and end of discovery.
13460 (dt_sort_cmp): New.
13461 (vect_build_slp_tree_2): Handle associatable chains
13462 together doing more aggressive operand swapping.
13464 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
13467 * config.gcc (gcc_cv_initfini_array): Set to yes for Linux and
13469 * doc/install.texi: Require glibc 2.1 and binutils 2.12 for
13470 Linux and GNU targets.
13472 2021-06-09 Richard Biener <rguenther@suse.de>
13474 * tree-vect-stmts.c (vect_is_simple_use): Always get dt
13477 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
13479 * config/arc/arc.md (loop_end): Change it to
13480 define_insn_and_split.
13482 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
13484 * config/arc/arc.md (maddhisi4): Use VMAC2H instruction.
13485 (machi): New pattern.
13486 (umaddhisi4): Use VMAC2HU instruction.
13487 (umachi): New pattern.
13489 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
13491 * config/arc/arc-protos.h (arc_split_move_p): New prototype.
13492 * config/arc/arc.c (arc_split_move_p): New function.
13493 (arc_split_move): Clean up.
13494 * config/arc/arc.md (movdi_insn): Clean up, use arc_split_move_p.
13495 (movdf_insn): Likewise.
13496 * config/arc/simdext.md (mov<VWH>_insn): Likewise.
13498 2021-06-09 Uroš Bizjak <ubizjak@gmail.com>
13501 * config/i386/i386.c (print_operand_address_as): Rename "no_rip"
13502 argument to "raw". Do not emit segment overrides when "raw" is true.
13504 2021-06-09 Martin Liska <mliska@suse.cz>
13506 * doc/gcov.texi: Create a proper JSON files.
13507 * doc/invoke.texi: Remove dots in order to make it a valid
13510 2021-06-09 Xionghu Luo <luoxhu@linux.ibm.com>
13512 * config/rs6000/rs6000-p8swap.c (pattern_is_rotate64): New.
13513 (insn_is_load_p): Use pattern_is_rotate64.
13514 (insn_is_swap_p): Likewise.
13515 (quad_aligned_load_p): Likewise.
13516 (const_load_sequence_p): Likewise.
13517 (replace_swapped_aligned_load): Likewise.
13518 (recombine_lvx_pattern): Likewise.
13519 (recombine_stvx_pattern): Likewise.
13521 2021-06-09 Andrew MacLeod <amacleod@redhat.com>
13523 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Use a
13524 fur_stmt source record.
13525 * gimple-range.cc (fur_source::get_operand): Generic range query.
13526 (fur_source::get_phi_operand): New.
13527 (fur_source::register_dependency): New.
13528 (fur_source::query): New.
13529 (class fur_edge): New. Edge source for operands.
13530 (fur_edge::fur_edge): New.
13531 (fur_edge::get_operand): New.
13532 (fur_edge::get_phi_operand): New.
13533 (fur_edge::query): New.
13534 (fur_stmt::fur_stmt): New.
13535 (fur_stmt::get_operand): New.
13536 (fur_stmt::get_phi_operand): New.
13537 (fur_stmt::query): New.
13538 (class fur_depend): New. Statement source and process dependencies.
13539 (fur_depend::fur_depend): New.
13540 (fur_depend::register_dependency): New.
13541 (class fur_list): New. List source for operands.
13542 (fur_list::fur_list): New.
13543 (fur_list::get_operand): New.
13544 (fur_list::get_phi_operand): New.
13545 (fold_range): New. Instantiate appropriate fur_source class and fold.
13546 (fold_using_range::range_of_range_op): Use new API.
13547 (fold_using_range::range_of_address): Ditto.
13548 (fold_using_range::range_of_phi): Ditto.
13549 (imple_ranger::fold_range_internal): Use fur_depend class.
13550 (fold_using_range::range_of_ssa_name_with_loop_info): Use new API.
13551 * gimple-range.h (class fur_source): Now a base class.
13552 (class fur_stmt): New.
13553 (fold_range): New prototypes.
13554 (fur_source::fur_source): Delete.
13556 2021-06-08 Andrew Pinski <apinski@marvell.com>
13558 PR tree-optimization/25290
13559 * tree-ssa-phiopt.c (xor_replacement): Delete.
13560 (tree_ssa_phiopt_worker): Delete use of xor_replacement.
13561 (match_simplify_replacement): Allow one cheap preparation
13562 statement that can be moved to before the if.
13564 2021-06-08 Pat Haugen <pthaugen@linux.ibm.com>
13566 * config/rs6000/power10.md (power10-fused-load, power10-fused-store,
13567 power10-fused_alu, power10-fused-vec, power10-fused-branch): New.
13569 2021-06-08 Jeff Law <jeffreyalaw@gmail.com>
13571 * config/h8300/logical.md (andqi3_1): Move BCLR case into define_insn_and_split.
13572 Create length attribute on define_insn_and_split. Only split for cases which we
13574 (andqi3_1<cczn>): Renamed from andqi3_1_clobber_flags. Only handle AND here and
13575 fix length computation.
13576 (b<code><mode>msx): Combine QImode and HImode H8/SX patterns using iterator.
13578 2021-06-08 Richard Biener <rguenther@suse.de>
13580 PR tree-optimization/100923
13581 * tree-ssa-sccvn.c (valueize_refs_1): Take a pointer to
13582 the operand vector to be valueized.
13583 (valueize_refs): Likewise.
13584 (valueize_shared_reference_ops_from_ref): Adjust.
13585 (valueize_shared_reference_ops_from_call): Likewise.
13586 (vn_reference_lookup_3): Likewise.
13587 (vn_reference_lookup_pieces): Likewise. Re-valueize
13588 with honoring availability when we are about to create
13589 the ao_ref and valueized before.
13590 (vn_reference_lookup): Likewise.
13591 (vn_reference_insert_pieces): Adjust.
13593 2021-06-08 Richard Biener <rguenther@suse.de>
13595 * tree-vectorizer.h (_slp_instance::root_stmt): Change to...
13596 (_slp_instance::root_stmts): ... a vector.
13597 (SLP_INSTANCE_ROOT_STMT): Rename to ...
13598 (SLP_INSTANCE_ROOT_STMTS): ... this.
13599 (slp_root::root): Change to...
13600 (slp_root::roots): ... a vector.
13601 (slp_root::slp_root): Adjust.
13602 * tree-vect-slp.c (_slp_instance::location): Adjust.
13603 (vect_free_slp_instance): Release the root stmt vector.
13604 (vect_build_slp_instance): Adjust.
13605 (vect_analyze_slp): Likewise.
13606 (_bb_vec_info::~_bb_vec_info): Likewise.
13607 (vect_slp_analyze_operations): Likewise.
13608 (vect_bb_vectorization_profitable_p): Likewise. Adjust
13609 costs for the root stmt.
13610 (vect_slp_check_for_constructors): Gather all BIT_INSERT_EXPRs
13612 (vect_slp_analyze_bb_1): Simplify by marking all root stmts
13614 (vectorize_slp_instance_root_stmt): Adjust.
13615 (vect_schedule_slp): Likewise.
13617 2021-06-08 Aldy Hernandez <aldyh@redhat.com>
13619 * gimple-ssa-evrp.c (class ssa_equiv_stack): New.
13620 (ssa_equiv_stack::ssa_equiv_stack): New.
13621 (ssa_equiv_stack::~ssa_equiv_stack): New.
13622 (ssa_equiv_stack::enter): New.
13623 (ssa_equiv_stack::leave): New.
13624 (ssa_equiv_stack::push_replacement): New.
13625 (ssa_equiv_stack::get_replacement): New.
13626 (is_pointer_ssa): New.
13627 (class pointer_equiv_analyzer): New.
13628 (pointer_equiv_analyzer::pointer_equiv_analyzer): New.
13629 (pointer_equiv_analyzer::~pointer_equiv_analyzer): New.
13630 (pointer_equiv_analyzer::set_global_equiv): New.
13631 (pointer_equiv_analyzer::set_cond_equiv): New.
13632 (pointer_equiv_analyzer::get_equiv): New.
13633 (pointer_equiv_analyzer::enter): New.
13634 (pointer_equiv_analyzer::leave): New.
13635 (pointer_equiv_analyzer::get_equiv_expr): New.
13636 (pta_valueize): New.
13637 (pointer_equiv_analyzer::visit_stmt): New.
13638 (pointer_equiv_analyzer::visit_edge): New.
13639 (hybrid_folder::value_of_expr): Call PTA.
13640 (hybrid_folder::value_on_edge): Same.
13641 (hybrid_folder::pre_fold_bb): New.
13642 (hybrid_folder::post_fold_bb): New.
13643 (hybrid_folder::pre_fold_stmt): New.
13644 (rvrp_folder::pre_fold_bb): New.
13645 (rvrp_folder::post_fold_bb): New.
13646 (rvrp_folder::pre_fold_stmt): New.
13647 (rvrp_folder::value_of_expr): Call PTA.
13648 (rvrp_folder::value_on_edge): Same.
13650 2021-06-08 Jakub Jelinek <jakub@redhat.com>
13653 * tree-inline.c (copy_tree_body_r): For OMP_CLAUSE_DEPEND don't
13654 check TREE_CODE if OMP_CLAUSE_DECL is NULL.
13656 2021-06-08 Richard Biener <rguenther@suse.de>
13658 PR middle-end/100951
13659 * tree-vect-generic.c (expand_vector_piecewise): Build a
13660 VECTOR_CST if all elements are constant.
13661 (expand_vector_condition): Likewise.
13662 (lower_vec_perm): Likewise.
13663 (expand_vector_conversion): Likewise.
13665 2021-06-08 Martin Liska <mliska@suse.cz>
13667 * doc/invoke.texi: Document new param evrp-sparse-threshold.
13669 2021-06-08 Martin Liska <mliska@suse.cz>
13671 * genautomata.c (create_automata): Fix typo.
13673 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
13675 PR tree-optimization/100794
13676 * tree-predcom.c (tree_predictive_commoning_loop): Add parameter
13677 allow_unroll_p and only allow unrolling when it's true.
13678 (tree_predictive_commoning): Add parameter allow_unroll_p and
13680 (run_tree_predictive_commoning): Likewise.
13681 (pass_predcom::gate): Check flag_tree_loop_vectorize and
13682 global_options_set.x_flag_predictive_commoning.
13683 (pass_predcom::execute): Adjust for allow_unroll_p.
13685 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
13687 * tree-predcom.c (execute_pred_commoning): Remove update_ssa call.
13688 (tree_predictive_commoning_loop): Factor some cleanup stuffs into
13689 lambda function cleanup, remove scev_reset call, and adjust return
13691 (tree_predictive_commoning): Adjust for different changed values,
13692 only set flag TODO_update_ssa_only_virtuals if changed.
13693 (pass_data pass_data_predcom): Remove TODO_update_ssa_only_virtuals
13694 from todo_flags_finish.
13696 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
13698 * gimple-range-cache.cc (class sbr_sparse_bitmap): New.
13699 (sbr_sparse_bitmap::sbr_sparse_bitmap): New.
13700 (sbr_sparse_bitmap::bitmap_set_quad): New.
13701 (sbr_sparse_bitmap::bitmap_get_quad): New.
13702 (sbr_sparse_bitmap::set_bb_range): New.
13703 (sbr_sparse_bitmap::get_bb_range): New.
13704 (sbr_sparse_bitmap::bb_range_p): New.
13705 (block_range_cache::block_range_cache): initialize bitmap obstack.
13706 (block_range_cache::~block_range_cache): Destruct obstack.
13707 (block_range_cache::set_bb_range): Decide when to utilze the
13708 sparse on entry cache.
13709 * gimple-range-cache.h (block_range_cache): Add bitmap obstack.
13710 * params.opt (-param=evrp-sparse-threshold): New.
13712 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
13714 * bitmap.c (bitmap_set_aligned_chunk): New.
13715 (bitmap_get_aligned_chunk): New.
13716 (test_aligned_chunk): New.
13717 (bitmap_c_tests): Call test_aligned_chunk.
13718 * bitmap.h (bitmap_set_aligned_chunk, bitmap_get_aligned_chunk): New.
13720 2021-06-07 Uroš Bizjak <ubizjak@gmail.com>
13723 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
13725 (ix86_expand_vector_init_one_nonzero): Ditto.
13726 (ix86_expand_vector_init_one_var): Ditto.
13727 (ix86_expand_vector_init_general): Ditto.
13728 * config/i386/mmx.md (vec_initv4qiqi): New expander.
13730 2021-06-07 Jeff Law <jeffreyalaw@gmail.com>
13732 * config/h8300/movepush.md: Change most _clobber_flags
13733 patterns to instead use <cczn> subst.
13734 (movsi_cczn): New pattern with usable CC cases split out.
13735 (movsi_h8sx_cczn): Likewise.
13737 2021-06-07 Martin Liska <mliska@suse.cz>
13739 * common/common-target.def: Split long lines and replace them
13741 * target.def: Likewise.
13742 * doc/tm.texi: Re-generated.
13744 2021-06-07 Jakub Jelinek <jakub@redhat.com>
13747 * fold-const.c (fold_read_from_vector): Return NULL if trying to
13748 read from a CONSTRUCTOR with vector type elements.
13750 2021-06-07 Jakub Jelinek <jakub@redhat.com>
13752 PR middle-end/100898
13753 * tree-inline.c (copy_bb): Only use gimple_call_arg_ptr if memcpy
13754 should copy any arguments. Don't call gimple_call_num_args
13755 on id->call_stmt or call_stmt more than once.
13757 2021-06-07 liuhongt <hongtao.liu@intel.com>
13760 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): Refine
13762 (<insn>v4siv4di2): Delete constraints for define_expand.
13764 2021-06-07 liuhongt <hongtao.liu@intel.com>
13767 * config/i386/i386-expand.c (ix86_expand_builtin): Remove
13768 assignment of cfun->machine->has_explicit_vzeroupper.
13769 * config/i386/i386-features.c
13770 (ix86_add_reg_usage_to_vzerouppers): Delete.
13771 (ix86_add_reg_usage_to_vzeroupper): Ditto.
13772 (rest_of_handle_insert_vzeroupper): Remove
13773 ix86_add_reg_usage_to_vzerouppers, add df_analyze at the end
13775 (gate): Remove cfun->machine->has_explicit_vzeroupper.
13776 * config/i386/i386-protos.h (ix86_expand_avx_vzeroupper):
13778 * config/i386/i386.c (ix86_insn_callee_abi): New function.
13779 (ix86_initialize_callee_abi): Ditto.
13780 (ix86_expand_avx_vzeroupper): Ditto.
13781 (ix86_hard_regno_call_part_clobbered): Adjust for vzeroupper
13783 (TARGET_INSN_CALLEE_ABI): Define as ix86_insn_callee_abi.
13784 (ix86_emit_mode_set): Call ix86_expand_avx_vzeroupper
13786 * config/i386/i386.h (struct GTY(()) machine_function): Delete
13787 has_explicit_vzeroupper.
13788 * config/i386/i386.md (enum unspec): New member
13790 (ABI_DEFAULT,ABI_VZEROUPPER,ABI_UNKNOWN): New
13791 define_constants for insn callee abi index.
13792 * config/i386/predicates.md (vzeroupper_pattern): Adjust.
13793 * config/i386/sse.md (UNSPECV_VZEROUPPER): Deleted.
13794 (avx_vzeroupper): Call ix86_expand_avx_vzeroupper.
13795 (*avx_vzeroupper): Rename to ..
13796 (avx_vzeroupper_callee_abi): .. this, and adjust pattern as
13797 call_insn which has a special vzeroupper ABI.
13798 (*avx_vzeroupper_1): Deleted.
13800 2021-06-07 liuhongt <hongtao.liu@intel.com>
13803 * df-scan.c (df_get_call_refs): When call_insn is a fake call,
13804 it won't use stack pointer reg.
13805 * final.c (leaf_function_p): When call_insn is a fake call, it
13806 won't affect caller as a leaf function.
13807 * reg-stack.c (callee_clobbers_any_stack_reg): New.
13808 (subst_stack_regs): When call_insn doesn't clobber any stack
13809 reg, don't clear the arguments.
13810 * rtl.c (shallow_copy_rtx): Don't clear flag used when orig is
13812 * shrink-wrap.c (requires_stack_frame_p): No need for stack
13813 frame for a fake call.
13814 * rtl.h (FAKE_CALL_P): New macro.
13816 2021-06-06 Eric Botcazou <ebotcazou@adacore.com>
13818 * config/sparc/sparc-protos.h (order_regs_for_local_alloc): Rename
13820 (sparc_order_regs_for_local_alloc): ...this.
13821 (sparc_leaf_reg_remap): Declare.
13822 * config/sparc/sparc.h (ADJUST_REG_ALLOC_ORDER): Adjust.
13823 (LEAF_REG_REMAP): Reimplement as call to sparc_leaf_reg_remap.
13824 * config/sparc/sparc.c (leaf_reg_remap): Delete.
13825 (order_regs_for_local_alloc): Rename to...
13826 (sparc_order_regs_for_local_alloc): ...this.
13827 (sparc_leaf_reg_remap): New function.
13828 (sparc_conditional_register_usage): Do not modify leaf_reg_remap.
13830 2021-06-06 David Edelsohn <dje.gcc@gmail.com>
13832 * config/rs6000/rs6000.c (rs6000_xcoff_asm_output_aligned_decl_common):
13833 Use assemble_name to output BSS section name.
13835 2021-06-06 Uroš Bizjak <ubizjak@gmail.com>
13837 * config/i386/constraints.md (Bs):
13838 Remove boolean operators from match_test RTX.
13841 (M): Use "mode" variable instead of GET_MODE (op) in match_test RTX.
13844 2021-06-06 Martin Liska <mliska@suse.cz>
13846 * doc/extend.texi: Add missing @headitem.
13847 * doc/invoke.texi: Likewise.
13848 * doc/objc.texi: Likewise.
13850 2021-06-06 Martin Liska <mliska@suse.cz>
13852 * genhooks.c (emit_findices): Remove unused function.
13853 (emit_documentation): Do not call emit_findices
13854 and do not search for @Fcode directives.
13856 2021-06-06 Martin Liska <mliska@suse.cz>
13858 * doc/invoke.texi: Remove extra character.
13860 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13862 * config/sh/sh.md (doloop_end_split): Fix empty split condition.
13864 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13866 * config/sparc/sparc.md (*snedi<W:mode>_zero_vis3,
13867 *neg_snedi<W:mode>_zero_subxc, *plus_snedi<W:mode>_zero,
13868 *plus_plus_snedi<W:mode>_zero, *minus_snedi<W:mode>_zero,
13869 *minus_minus_snedi<W:mode>_zero): Fix empty split condition.
13871 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13873 * config/or1k/or1k.md (*movdi): Fix empty split condition.
13875 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13877 * config/mips/mips.md (<anonymous>, bswapsi2, bswapdi2): Fix empty
13880 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13882 * config/m68k/m68k.md (*zero_extend_inc, *zero_extend_dec,
13883 *zero_extendsidi2): Fix empty split condition.
13885 2021-06-05 Jeff Law <jeffreyalaw@gmail.com>
13887 * config/h8300/addsub.md: Fix split condition in define_insn_and_split
13889 * config/h8300/bitfield.md: Likewise.
13890 * config/h8300/combiner.md: Likewise.
13891 * config/h8300/divmod.md: Likewise.
13892 * config/h8300/extensions.md: Likewise.
13893 * config/h8300/jumpcall.md: Likewise.
13894 * config/h8300/movepush.md: Likewise.
13895 * config/h8300/multiply.md: Likewise.
13896 * config/h8300/other.md: Likewise.
13897 * config/h8300/shiftrotate.md: Likewise.
13898 * config/h8300/logical.md: Likewise. Fix split pattern to use
13899 code iterator that somehow slipped through.
13901 2021-06-04 Tobias Burnus <tobias@codesourcery.com>
13903 PR middle-end/100905
13904 * tree-nested.c (convert_nonlocal_omp_clauses,
13905 convert_local_omp_clauses): Handle OMP_CLAUSE_BIND.
13907 2021-06-04 Martin Sebor <msebor@redhat.com>
13909 PR middle-end/100732
13910 * gimple-fold.c (gimple_fold_builtin_sprintf): Avoid folding calls
13911 with either source or destination argument of invalid type.
13912 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Avoid checking
13913 calls with arguments of invalid type.
13915 2021-06-04 Martin Sebor <msebor@redhat.com>
13917 * attribs.c (init_attr_rdwr_indices): Use VLA bounds in the expected
13919 (attr_access::vla_bounds): Also handle VLA bounds.
13921 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
13923 * config/i386/predicates.md (GOT_memory_operand):
13924 Implement using match_code RTXes.
13925 (GOT32_symbol_operand): Ditto.
13927 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
13930 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
13932 (ix86_expand_vector_init_general): Ditto.
13933 Use SImode instead of word_mode for logic operations
13934 when GET_MODE_SIZE (mode) < UNITS_PER_WORD.
13935 (expand_vec_perm_even_odd_1): Assert that V2HI mode should be
13936 implemented by expand_vec_perm_1.
13937 (expand_vec_perm_broadcast_1): Assert that V2HI and V4HI modes
13938 should be implemented using standard shuffle patterns.
13939 (ix86_vectorize_vec_perm_const): Handle V2HImode. Add V4HI and
13940 V2HI modes to modes, implementable with shuffle for one operand.
13941 * config/i386/mmx.md (*punpckwd): New insn_and_split pattern.
13942 (*pshufw_1): New insn pattern.
13943 (*vec_dupv2hi): Ditto.
13944 (vec_initv2hihi): New expander.
13946 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
13948 * config/arm/vfp.md (no_literal_pool_df_immediate,
13949 no_literal_pool_sf_immediate): Fix empty split condition.
13951 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
13953 * config/i386/i386.md (*load_tp_x32_zext, *add_tp_x32_zext,
13954 *tls_dynamic_gnu2_combine_32): Fix empty split condition.
13955 * config/i386/sse.md (*<sse2_avx2>_pmovmskb_lt,
13956 *<sse2_avx2>_pmovmskb_zext_lt, *sse2_pmovmskb_ext_lt,
13957 *<sse4_1_avx2>_pblendvb_lt): Likewise.
13959 2021-06-04 Jakub Jelinek <jakub@redhat.com>
13962 * config/i386/i386-expand.c (ix86_expand_vector_init): Handle
13963 concatenation from half-sized modes with TImode elements.
13965 2021-06-04 Claudiu Zissulescu <claziss@synopsys.com>
13967 * config/arc/arc.c (arc_override_options): Disable millicode
13968 thunks when RF16 is on.
13970 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
13972 * config/rs6000/rs6000.h (PROMOTE_MODE): Remove.
13974 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
13976 * config/rs6000/rs6000-call.c (rs6000_promote_function_mode):
13977 Replace PROMOTE_MODE marco with its content.
13979 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
13981 * config/cris/cris.md (*addi_reload): Fix empty split condition.
13983 2021-06-03 Jim Wilson <jimw@sifive.com>
13985 * config.gcc (riscv*-*-*): If --with-riscv-attribute not used,
13986 turn it on for all riscv targets.
13988 2021-06-03 Uroš Bizjak <ubizjak@gmail.com>
13991 * config/i386/i386-expand.c (ix86_expand_vector_set):
13992 Handle V2HI and V4QI modes.
13993 (ix86_expand_vector_extract): Ditto.
13994 * config/i386/mmx.md (*pinsrw): New insn pattern.
13997 (*pextrw_zext): Ditto.
13999 (*pextrb_zext): Ditto.
14000 (vec_setv2hi): New expander.
14001 (vec_extractv2hihi): Ditto.
14002 (vec_setv4qi): Ditto.
14003 (vec_extractv4qiqi): Ditto.
14004 (vec_setv8qi): Enable only for TARGET_SSE4_1.
14005 (vec_extractv8qiqi): Ditto.
14007 2021-06-03 Aaron Sawdey <acsawdey@linux.ibm.com>
14009 * config/rs6000/genfusion.pl (gen_logical_addsubf): Fix input
14010 order to subf instruction.
14011 * config/rs6000/fusion.md: Regenerate.
14013 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
14015 * calls.c (get_size_range): Use range_of_expr instead of
14016 determine_value_range.
14017 * tree-affine.c (expr_to_aff_combination): Same.
14018 * tree-data-ref.c (split_constant_offset): Same.
14019 * tree-vrp.c (determine_value_range_1): Remove.
14020 (determine_value_range): Remove.
14021 * tree-vrp.h (determine_value_range): Remove.
14023 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
14025 * function-tests.c (test_ranges): Call gimple_range_tests.
14026 * gimple-range-cache.cc (ranger_cache::range_of_expr): Pass stmt
14028 * gimple-range.cc (fur_source::get_operand): Do not call
14029 get_tree_range or gimple_range_global.
14031 (get_tree_range): Move to value-query.cc.
14032 Call get_arith_expr_range.
14033 (gimple_ranger::range_of_expr): Add argument to get_tree_range.
14034 Include gimple-range-tests.cc.
14035 * gimple-range.h (fold_range): Add argument.
14036 (get_tree_range): Remove.
14037 * selftest.h (gimple_range_tests): New.
14038 * value-query.cc (global_range_query::range_of_expr): Add
14040 (range_query::get_tree_range): Move from gimple-range.cc.
14041 * value-query.h (class range_query): Add get_tree_range and
14042 get_arith_expr_range. Make fur_source a friend.
14043 * vr-values.c (vr_values::range_of_expr): Pass stmt to
14045 * gimple-range-tests.cc: New file.
14047 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
14049 * gimple-range.cc (gimple_ranger::export_global_ranges): Call
14050 update_global_range.
14051 * value-query.cc (update_global_range): New.
14052 * value-query.h (update_global_range): New.
14054 2021-06-03 David Malcolm <dmalcolm@redhat.com>
14056 * diagnostic-show-locus.c (diagnostic_show_locus): Don't reject
14057 printing the same location twice if there are fix-it hints,
14058 multiple locations, or a label.
14060 2021-06-03 Andre Vieira <andre.simoesdiasvieira@arm.com>
14062 * tree-vect-loop.c (vect_transform_loop): Use main loop's various'
14063 thresholds to narrow the upper bound on epilogue iterations.
14065 2021-06-03 Christophe Lyon <christophe.lyon@linaro.org>
14067 * config/arm/mve.md (mve_vabsq_f<mode>): Use 'abs' instead of unspec.
14068 (mve_vabsq_s<mode>): Likewise.
14069 * config/arm/neon.md (abs<mode>2): Rename to neon_abs<mode>2.
14070 * config/arm/unspecs.md (VABSQ_F, VABSQ_S): Delete.
14071 * config/arm/vec-common.md (neg<mode>2): Rename to
14072 <absneg_str><mode>2.
14074 2021-06-03 Claudiu Zissulescu <claziss@synopsys.com>
14076 * common/config/arc/arc-common.c (arc_option_optimization_table):
14077 Remove malign-call.
14078 * config/arc/arc.c (arc_unalign_branch_p): Remove unused function.
14079 * config/arc/arc.h (TARGET_MIXED_CODE): Remove macro.
14080 (INDEX_REG_CLASS): Only refer to GENERAL_REGS.
14081 * config/arc/arc.md (abssi2_mixed): Remove pattern.
14082 * config/arc/arc.opt (munalign-prob-threshold): Mark it obsolete.
14083 (malign-call): Likewise.
14084 (mmixed-code): Likewise.
14085 * doc/invoke.texi (ARC): Update doc.
14087 2021-06-03 Martin Liska <mliska@suse.cz>
14089 * common.opt: Use proper Enum values.
14090 * opts.c (COVERAGE_SANITIZER_OPT): Remove.
14091 (parse_sanitizer_options): Handle only sanitizer_opts.
14092 (common_handle_option): Just assign value.
14094 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
14097 * tree-inline.c (inline_forbidden_p): Remove test on return type.
14099 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
14101 * dwarf2out.c (loc_list_from_tree_1) <FUNCTION_DECL>: Also generate
14102 DW_OP_GNU_variable_value referencing an existing DIE at file scope.
14103 (type_byte_size): Inline into...
14104 (add_byte_size_attribute): ...this and call add_scalar_info.
14106 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
14108 * dwarf2out.c (mem_loc_descriptor) <UDIV>: Fix typo.
14109 (typed_binop_from_tree): New function.
14110 (loc_list_from_tree_1) <EXACT_DIV_EXPR>: For an unsigned type,
14111 turn a divide by a power of 2 into a shift.
14112 <CEIL_DIV_EXPR>: For an unsigned type, use a signed divide if the
14113 size of the mode is lower than DWARF2_ADDR_SIZE; otherwise, do a
14114 typed divide by calling typed_binop_from_tree.
14116 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
14118 * dwarf2out.c (scompare_loc_descriptor): Fix head comment.
14119 (is_handled_procedure_type): Likewise.
14120 (struct loc_descr_context): Add strict_signedness field.
14121 (resolve_args_picking_1): Deal with DW_OP_[GNU_]deref_type,
14122 DW_OP_[GNU_]convert and DW_OP_[GNU_]reinterpret.
14123 (resolve_args_picking): Minor tweak.
14124 (function_to_dwarf_procedure): Initialize strict_signedness field.
14125 (type_byte_size): Likewise.
14126 (field_byte_offset): Likewise.
14127 (gen_descr_array_type_die): Likewise.
14128 (gen_variant_part): Likewise.
14129 (loc_list_from_tree_1) <CALL_EXPR>: Tidy up and set strict_signedness
14130 to true when a context is present before evaluating the arguments.
14131 <COND_EXPR>: Do not generate a useless comparison with zero.
14132 When dereferencing an address, if strict_signedness is true and the
14133 type is small and signed, use DW_OP_deref_type to do the dereference
14134 and then DW_OP_convert to convert back to the generic type.
14136 2021-06-03 Jakub Jelinek <jakub@redhat.com>
14139 * tree-inline.c (copy_tree_body_r): Handle iterators on
14140 OMP_CLAUSE_AFFINITY or OMP_CLAUSE_DEPEND.
14142 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
14144 * config/arc/arc.md (*bbit_di): Remove.
14146 2021-06-02 Christoph Muellner <cmuellner@gcc.gnu.org>
14148 PR rtl-optimization/100264
14149 * ree.c (get_sub_rtx): Ignore SET expressions without register
14150 destinations and remove assertion, as it is not valid anymore
14151 with this new behaviour.
14152 (merge_def_and_ext): Eliminate destination check for register
14153 as such SET expressions can't occur anymore.
14154 (combine_reaching_defs): Likewise.
14156 2021-06-02 Jakub Jelinek <jakub@redhat.com>
14159 * config/xtensa/xtensa.h (LEAF_REG_REMAP): Cast REGNO to int to avoid
14160 -Wtype-limits warnings.
14161 (DWARF_FRAME_REGISTER): Rewrite into ternary operator with addition
14162 in operands to avoid -Wsign-compare warnings.
14164 2021-06-02 Pat Haugen <pthaugen@linux.ibm.com>
14166 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): Use
14169 2021-06-02 Vineet Gupta <vgupta@synopsys.com>
14171 * config/arc/arc.h (TARGET_CPU_DEFAULT): Change to hs38_linux.
14173 2021-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
14175 * config/s390/s390.md(*ashrdi3_31<setcc><cconly>): Use a single
14177 * config/s390/subst.md(cconly_subst): Use a single constraint
14178 in (match_scratch).
14180 2021-06-02 Martin Liska <mliska@suse.cz>
14182 * ipa-icf.h: Use auto_vec for memory_access_types.
14184 2021-06-02 Jeff Law <jeffreyalaw@gmail.com>
14186 * config/h8300/h8300-protos.h (compute_a_shift_length): Drop unused
14187 argument from prototype.
14188 (output_logical_op): Add rtx_code argument.
14189 (compute_logical_op_length): Likewise.
14190 * config/h8300/h8300.c (h8300_and_costs): Pass additional argument
14191 to compute_a_shift_length.
14192 (output_logical_op); New argument with the rtx code rather than
14193 extracting it from an operand. Handle QImode too.
14194 (compute_logical_op_length): Similary.
14195 (compute_a_shift_length): Drop unused argument.
14196 * config/h8300/h8300.md (logicals): New code iterator.
14197 * config/h8300/logical.md (<code><mode>3 expander): Combine
14198 the "and" expander with the "ior"/"xor" expander.
14199 (bclr<mode>msx): Combine the QI/HI mode patterns.
14200 (<logical><mode>3 insns): Use code iterator rather than match_operator.
14201 Handle QImode as well. Update call to output_logical_op and
14202 compute_logical_op_length to pass in rtx_code
14203 Fix split condition on all define_insn_and_split patterns.
14204 (one_cmpl<mode>2<cczn>): Use <cczn> to support both clobbering
14205 the flags and setting ZN via existing define_subst.
14206 * config/h8300/shiftrotate.md: Drop unused argument from
14207 calls to compute_a_shift_length.
14208 Signed-off-by: Jeff Law <jeffreyalaw@gmail.com>
14210 2021-06-01 Andrew Pinski <apinski@marvell.com>
14212 PR tree-optimization/25290
14213 * tree-ssa-phiopt.c (match_simplify_replacement):
14215 (tree_ssa_phiopt_worker): Use match_simplify_replacement.
14216 (two_value_replacement): Change the comment about
14217 conditional_replacement.
14218 (conditional_replacement): Delete.
14220 2021-06-01 Andrew Pinski <apinski@marvell.com>
14222 PR tree-optimization/95481
14223 * tree-tailcall.c (find_tail_calls): Handle empty typed
14226 2021-06-01 Andrew Pinski <apinski@marvell.com>
14228 * gimplify.c (zero_sized_field_decl): Delete
14229 (zero_sized_type): Delete
14230 (gimplify_init_ctor_eval): Use is_empty_type instead
14231 of zero_sized_field_decl.
14232 (gimplify_modify_expr): Use is_empty_type instead of
14235 2021-06-01 Jason Merrill <jason@redhat.com>
14238 * tree.h (CALL_FROM_NEW_OR_DELETE_P): Adjust comment.
14240 2021-06-01 Jason Merrill <jason@redhat.com>
14243 * diagnostic.h (warning_enabled_at): Declare.
14244 * diagnostic.c (diagnostic_enabled): Factor out from...
14245 (diagnostic_report_diagnostic): ...here.
14246 (warning_enabled_at): New.
14248 2021-06-01 Aldy Hernandez <aldyh@redhat.com>
14250 * gimple-ssa-evrp.c: Enable exporting of global ranges.
14252 2021-06-01 Martin Liska <mliska@suse.cz>
14255 * doc/invoke.texi: Mention that -fgcse-after-reload
14256 is enabled with -O3.
14258 2021-06-01 liuhongt <hongtao.liu@intel.com>
14260 PR tree-optimization/98365
14261 * tree-if-conv.c (strip_nop_cond_scalar_reduction): New function.
14262 (is_cond_scalar_reduction): Handle nop_expr in cond scalar reduction.
14263 (convert_scalar_cond_reduction): Ditto.
14264 (predicate_scalar_phi): Ditto.
14266 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
14268 PR tree-optimization/100781
14269 * gimple-range-cache.cc (ranger_cache::ranger_cache): Enable new
14270 value calculation by default.
14271 (ranger_cache::enable_new_values): New.
14272 (ranger_cache::disable_new_values): New.
14273 (ranger_cache::push_poor_value): Check if new values are allowed.
14274 * gimple-range-cache.h (class ranger_cache): New member/methods.
14275 * gimple-range.cc (gimple_ranger::range_of_expr): Check for debug
14276 statement, and disable/renable new value calculation.
14278 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
14280 * gimple-range-cache.cc (ranger_cache::ssa_range_in_bb): Delete.
14281 (ranger_cache::range_of_def): New.
14282 (ranger_cache::entry_range): New.
14283 (ranger_cache::exit_range): New.
14284 (ranger_cache::range_of_expr): Adjust.
14285 (ranger_cache::range_on_edge): Adjust.
14286 (ranger_cache::propagate_cache): Call exit_range directly.
14287 * gimple-range-cache.h (class ranger_cache): Adjust.
14289 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
14291 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust for
14292 gori_compute being a member rather than base class.
14293 dervied call to member call.
14294 (ranger_cache::dump): No longer dump gori_map.
14295 (ranger_cache::dump_bb): New.
14296 (ranger_cache::get_non_stale_global_range): Adjust for gori_compute
14297 being a member rather than base class.
14298 (ranger_cache::set_global_range): Ditto.
14299 (ranger_cache::ssa_range_in_bb): Ditto.
14300 (ranger_cache::range_of_expr): New.
14301 (ranger_cache::range_on_edge): New.
14302 (ranger_cache::block_range): Adjust for gori_computes. Debug changes.
14303 (ranger_cache::propagate_cache): Adjust debugging output.
14304 (ranger_cache::fill_block_cache): Adjust for gori_computes. Debug
14306 * gimple-range-cache.h (class ranger_cache): Make gori_compute a
14307 member, and inherit from range_query instead.
14308 (ranger_cache::dump_bb): New. split from dump.
14309 * gimple-range-gori.cc (gori_compute::ssa_range_in_bb): Delete.
14310 (gori_compute::expr_range_at_stmt): Delete.
14311 (gori_compute::compute_name_range_op): Delete.
14312 (gori_compute::compute_operand_range_switch): Add fur_source.
14313 (gori_compute::compute_operand_range): Add fur_source param, inline
14314 old compute_name_range_op and optimize_logical_operands.
14315 (struct tf_range): Delete.
14316 (gori_compute::logical_combine): Adjust
14317 (gori_compute::optimize_logical_operands): Delete.
14318 (gori_compute::compute_logical_operands_in_chain): Delete.
14319 (gori_compute::compute_logical_operands): Adjust.
14320 (gori_compute::compute_operand1_range): Adjust to fur_source.
14321 (gori_compute::compute_operand2_range): Ditto.
14322 (gori_compute::compute_operand1_and_operand2_range): Ditto.
14323 (gori_compute::outgoing_edge_range_p): Add range_query parameter,
14324 and adjust to fur_source.
14325 * gimple-range-gori.h (class gori_compute): Simplify and adjust to
14326 range_query and fur_source.
14327 * gimple-range.cc (gimple_ranger::range_on_edge): Query range_on_edge
14328 from the ranger_cache..
14329 (gimple_ranger::fold_range_internal): Adjust to base class change of
14331 (gimple_ranger::dump_bb): Adjust dump.
14332 * gimple-range.h (gimple_ranger):export gori computes object.
14334 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
14336 PR tree-optimization/100774
14337 * gimple-range-cache.cc (ranger_cache::get_non_stale_global_range):
14338 Constant values are also not stale.
14339 (ranger_cache::set_global_range): Range invariant values should also
14340 have the correct timestamp.
14342 2021-05-31 Martin Liska <mliska@suse.cz>
14344 * tree-streamer-in.c (unpack_ts_function_decl_value_fields):
14345 Unpack FUNCTION_DECL_DECL_TYPE.
14346 * tree-streamer-out.c (pack_ts_function_decl_value_fields):
14347 Stream FUNCTION_DECL_DECL_TYPE instead of
14348 DECL_IS_OPERATOR_NEW_P.
14349 * tree.h (set_function_decl_type): Use FUNCTION_DECL_DECL_TYPE
14351 (DECL_IS_OPERATOR_NEW_P): Likewise.
14352 (DECL_IS_OPERATOR_DELETE_P): Likewise.
14353 (DECL_LAMBDA_FUNCTION_P): Likewise.
14355 2021-05-31 Richard Biener <rguenther@suse.de>
14358 * internal-fn.c (expand_SHUFFLEVECTOR): Define.
14359 * internal-fn.def (SHUFFLEVECTOR): New.
14360 * internal-fn.h (expand_SHUFFLEVECTOR): Declare.
14361 * doc/extend.texi: Document __builtin_shufflevector.
14363 2021-05-31 Peter Bergner <bergner@linux.ibm.com>
14366 * config/rs6000/predicates.md(mma_assemble_input_operand): Allow
14367 indexed form addresses.
14369 2021-05-29 Jeff Law <jlaw@tachyum.com>
14371 * config/h8300/h8300.c (h8300_emit_stack_adjustment): Drop unused
14372 parameter. Call callers fixed.
14374 (output_plussi): Add FALLTHRU markers.
14375 (h8300_shift_needs_scratch_p): Add gcc_unreachable marker.
14377 2021-05-29 Jakub Jelinek <jakub@redhat.com>
14379 PR middle-end/99928
14380 * gimplify.c (gimplify_scan_omp_clauses): For taskloop simd
14381 combined with parallel, make sure to add shared clause to
14382 parallel for explicit linear clause.
14384 2021-05-29 Aldy Hernandez <aldyh@redhat.com>
14386 PR tree-optimization/100787
14387 * gimple-ssa-evrp.c: Disable exporting of global ranges.
14389 2021-05-28 Jason Merrill <jason@redhat.com>
14391 * tree-iterator.h (struct tree_stmt_iterator): Add operator++,
14392 operator--, operator*, operator==, and operator!=.
14393 (class tsi_range): New.
14395 2021-05-28 Richard Biener <rguenther@suse.de>
14397 PR tree-optimization/100778
14398 * tree-vect-slp.c (vect_build_slp_tree_1): Prevent possibly
14399 trapping ops in different BBs.
14401 2021-05-28 Richard Biener <rguenther@suse.de>
14404 * tree-inline.c (copy_bb): When processing __builtin_va_arg_pack
14405 copy fntype from original call.
14407 2021-05-28 Martin Liska <mliska@suse.cz>
14409 PR gcov-profile/100751
14410 * doc/gcov.texi: Revert partially a hunk that was wrong.
14412 2021-05-28 Cooper Qu <cooper.qu@linux.alibaba.com>
14414 * config/csky/csky-linux-elf.h (HAVE_sync_compare_and_swapqi):
14416 (HAVE_sync_compare_and_swaphi): Likewise.
14417 (HAVE_sync_compare_and_swapsi): Likewise.
14419 2021-05-28 Jakub Jelinek <jakub@redhat.com>
14421 PR middle-end/99928
14422 * tree.h (OMP_CLAUSE_MAP_IMPLICIT): Define.
14424 2021-05-28 Tobias Burnus <tobias@codesourcery.com>
14426 * gimplify.c (gimplify_omp_affinity): New.
14427 (gimplify_scan_omp_clauses): Call it; remove affinity clause afterwards.
14428 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_AFFINITY.
14429 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_AFFINITY.
14430 * tree.c (omp_clause_num_ops, omp_clause_code_name): Add clause.
14431 (walk_tree_1): Handle OMP_CLAUSE_AFFINITY.
14433 2021-05-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
14434 Richard Biener <rguenther@suse.de>
14436 * match.pd <popcount & / + pattern matching>:
14437 When generating popcount directly fails, try doing it in two halves.
14439 2021-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
14441 * Makefile.in (generated_files): Add gimple-match.c and
14444 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
14446 * gensupport.c (alter_predicate_for_insn): Handle MATCH_DUP.
14448 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
14450 * gensupport.c (alter_constraints): Add MATCH_SCRATCH case.
14452 2021-05-28 Kewen Lin <linkw@linux.ibm.com>
14454 PR tree-optimization/99398
14455 * tree-ssa-forwprop.c (simplify_permutation): Optimize some cases
14456 where the fed operands are CTOR/CST and propagated through
14457 VIEW_CONVERT_EXPR. Call vec_perm_indices::new_shrunk_vector.
14458 * vec-perm-indices.c (vec_perm_indices::new_shrunk_vector): New
14460 * vec-perm-indices.h (vec_perm_indices::new_shrunk_vector): New
14463 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
14465 * config/i386/mmx.md (addv2sf3): Do not call
14466 ix86_fixup_binary_operands_no_copy.
14469 (<smaxmin:code>v2sf3): Ditto.
14470 (<plusminus:insn><MMXMODEI:mode>3): Ditto.
14471 (<plusminus:insn><VI_32:mode>3): Remove expander.
14472 (<plusminus:insn><VI_32:mode>3): Rename from
14473 "*<plusminus:insn><VI_32:mode>3".
14474 (mulv4hi): Do not call ix86_fixup_binary_operands_no_copy.
14475 (mulv2hi3): Remove expander.
14476 (mulv2hi3): Rename from *mulv2hi3.
14477 (<s>mulv2hi3_highpart): Remove expander.
14478 (<s>mulv2hi3_highpart): Rename from *<s>mulv2hi3_highpart.
14479 (<smaxmin:code><MMXMODE14:mode>3): Rename from
14480 "*mmx_<smaxmin:code><MMXMODE14:mode>3".
14481 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Remove expander.
14482 (SMAXMIN_MMXMODEI): Remove mode iterator.
14483 (<smaxmin:code>v4hi3): New expander.
14484 (<smaxmin:code>v4qi3): Rename from *<smaxmin:code>v4qi3.
14485 (<smaxmin:code>v2hi3): Rename from *<smaxmin:code>v2hi3.
14486 (<smaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
14487 (SMAXMIN_VI_32): Remove mode iterator.
14488 (<umaxmin:code><MMXMODE24:mode>3): Rename from
14489 "*mmx_<umaxmin:code><MMXMODE24:mode>3".
14490 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Remove expander.
14491 (UMAXMIN_MMXMODEI): Remove mode iterator.
14492 (<umaxmin:code>v8qi3): New expander.
14493 (<umaxmin:code>v4qi3): Rename from *<umaxmin:code>v4qi3.
14494 (<umaxmin:code>v2hi3): Rename from *<umaxmin:code>v2hi3.
14495 (<umaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
14496 (UMAXMIN_VI_32): Remove mode iterator.
14497 (<any_shift:insn>v2hi3): Remove expander.
14498 (<any_shift:insn>v2hi3): Rename from *<any_shift:insn>v2hi3.
14499 (<any_logic:code><MMXMODEI:mode>3): Do not call
14500 ix86_fixup_binary_operands_no_copy.
14501 (<any_logic:code><VI_32:mode>3): Remove expander.
14502 (<any_logic:code><VI_32:mode>3): Rename from
14503 "*<any_logic:code><VI_32:mode>3".
14504 (uavg<mode>3_ceil): Do not call ix86_fixup_binary_operands_no_copy.
14505 * config/i386/sse.md (div<VF2:mode>3): Do not call
14506 ix86_fixup_binary_operands_no_copy.
14507 (div<VF1:mode>3): Ditto.
14508 (<maxmin:code><VI8_AVX2_AVX512F:mode>3): Ditto.
14509 (smulhrsv4hi3): Ditto.
14510 (smulhrsv2hi3): Ditto.
14512 2021-05-27 Martin Sebor <msebor@redhat.com>
14514 * ggc.h (gt_ggc_mx): Add overloads for all integers.
14516 * hash-map.h (class hash_map): Add pch_nx_helper overloads for all
14518 (hash_map::operator==): New function.
14520 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
14523 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
14524 For TARGET_XOP bypass SSE comparisons for all supported vector modes.
14525 * config/i386/mmx.md (*xop_maskcmp<MMXMODEI:mode>3): New insn pattern.
14526 (*xop_maskcmp<VI_32:mode>3): Ditto.
14527 (*xop_maskcmp_uns<MMXMODEI:mode>3): Ditto.
14528 (*xop_maskcmp_uns<VI_32:mode>3): Ditto.
14530 2021-05-27 Richard Earnshaw <rearnsha@arm.com>
14533 * config/arm/arm.c (arm_configure_build_target): Remove parameter
14534 opts_set, directly check opts parameters for being non-null.
14535 (arm_option_restore): Update call to arm_configure_build_target.
14536 (arm_option_override): Likewise.
14537 (arm_can_inline_p): Likewise.
14538 (arm_valid_target_attribute_tree): Likewise.
14539 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
14540 * config/arm/arm-protos.h (arm_configure_build_target): Adjust
14543 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
14545 * vr-values.c (simplify_conversion_using_ranges): Use
14546 get_range_query instead of get_global_range_query.
14548 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
14550 * gimple-range.cc (get_range_global): Move to value-query.cc.
14551 (gimple_range_global): Same.
14552 (get_global_range_query): Same.
14553 (global_range_query::range_of_expr): Same.
14554 * gimple-range.h (class global_range_query): Move to
14556 (gimple_range_global): Same.
14557 * tree-ssanames.c (get_range_info): Move to value-query.cc.
14558 (get_ptr_nonnull): Same.
14559 * tree-ssanames.h (get_range_info): Remove.
14560 (get_ptr_nonnull): Remove.
14561 * value-query.cc (get_ssa_name_range_info): Move from
14563 (get_ssa_name_ptr_info_nonnull): Same.
14564 (get_range_global): Move from gimple-range.cc.
14565 (gimple_range_global): Same.
14566 (get_global_range_query): Same.
14567 (global_range_query::range_of_expr): Same.
14568 * value-query.h (class global_range_query): Move from
14570 (gimple_range_global): Same.
14572 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
14575 * config/i386/mmx.md (uavgv4qi3_ceil): New insn pattern.
14576 (uavgv2hi3_ceil): Ditto.
14578 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
14581 * doc/extend.texi (scalar_storage_order): Rephrase slightly.
14583 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14585 * tree-ssanames.c (get_range_info): Merge both copies of
14586 get_range_info into one that works with irange.
14587 * tree-ssanames.h (get_range_info): Remove version that works on
14590 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14592 * builtins.c (check_nul_terminated_array): Convert to get_range_query.
14593 (expand_builtin_strnlen): Same.
14594 (determine_block_size): Same.
14595 * fold-const.c (expr_not_equal_to): Same.
14596 * gimple-fold.c (size_must_be_zero_p): Same.
14597 * gimple-match-head.c: Include gimple-range.h.
14598 * gimple-pretty-print.c (dump_ssaname_info): Convert to get_range_query.
14599 * gimple-ssa-warn-restrict.c
14600 (builtin_memref::extend_offset_range): Same.
14601 * graphite-sese-to-poly.c (add_param_constraints): Same.
14602 * internal-fn.c (get_min_precision): Same.
14603 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): Same.
14604 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Same.
14606 * tree-data-ref.c (split_constant_offset): Same.
14607 (dr_step_indicator): Same.
14608 * tree-dfa.c (get_ref_base_and_extent): Same.
14609 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
14610 * tree-ssa-loop-niter.c (refine_value_range_using_guard): Same.
14611 (determine_value_range): Same.
14612 (record_nonwrapping_iv): Same.
14613 (infer_loop_bounds_from_signedness): Same.
14614 (scev_var_range_cant_overflow): Same.
14615 * tree-ssa-phiopt.c (two_value_replacement): Same.
14616 * tree-ssa-pre.c (insert_into_preds_of_block): Same.
14617 * tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Same.
14618 * tree-ssa-strlen.c (handle_builtin_stxncpy_strncat): Same.
14620 (dump_strlen_info): Same.
14621 (set_strlen_range): Same.
14622 (maybe_diag_stxncpy_trunc): Same.
14623 (get_len_or_size): Same.
14624 (handle_integral_assign): Same.
14625 * tree-ssa-structalias.c (find_what_p_points_to): Same.
14626 * tree-ssa-uninit.c (find_var_cmp_const): Same.
14627 * tree-switch-conversion.c (bit_test_cluster::emit): Same.
14628 * tree-vect-patterns.c (vect_get_range_info): Same.
14629 (vect_recog_divmod_pattern): Same.
14630 * tree-vrp.c (intersect_range_with_nonzero_bits): Same.
14631 (register_edge_assert_for_2): Same.
14632 (determine_value_range_1): Same.
14633 * tree.c (get_range_pos_neg): Same.
14634 * vr-values.c (vr_values::get_lattice_entry): Same.
14635 (vr_values::update_value_range): Same.
14636 (simplify_conversion_using_ranges): Same.
14638 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14640 * gimple-ssa-warn-alloca.c (alloca_call_type): Use
14641 get_range_query instead of query argument.
14642 (pass_walloca::execute): Enable and disable global ranger.
14644 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14646 * gimple-ssa-evrp.c (rvrp_folder::rvrp_folder): Call
14648 (rvrp_folder::~rvrp_folder): Call disable_ranger.
14649 (hybrid_folder::hybrid_folder): Call enable_ranger.
14650 (hybrid_folder::~hybrid_folder): Call disable_ranger.
14652 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14654 * function.c (allocate_struct_function): Set cfun->x_range_query.
14655 * function.h (struct function): Declare x_range_query.
14656 (get_range_query): New.
14657 (get_global_range_query): New.
14658 * gimple-range-cache.cc (ssa_global_cache::ssa_global_cache):
14659 Remove call to safe_grow_cleared.
14660 * gimple-range.cc (get_range_global): New.
14661 (gimple_range_global): Move from gimple-range.h.
14662 (get_global_range_query): New.
14663 (global_range_query::range_of_expr): New.
14664 (enable_ranger): New.
14665 (disable_ranger): New.
14666 * gimple-range.h (gimple_range_global): Move to gimple-range.cc.
14667 (class global_range_query): New.
14668 (enable_ranger): New.
14669 (disable_ranger): New.
14670 * gimple-ssa-evrp.c (evrp_folder::~evrp_folder): Rename
14671 dump_all_value_ranges to dump.
14672 * tree-vrp.c (vrp_prop::finalize): Same.
14673 * value-query.cc (range_query::dump): New.
14674 * value-query.h (range_query::dump): New.
14675 * vr-values.c (vr_values::dump_all_value_ranges): Rename to...
14676 (vr_values::dump): ...this.
14677 * vr-values.h (class vr_values): Rename dump_all_value_ranges to
14678 dump and make virtual.
14680 2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
14682 * config/i386/i386.c (ix86_autovectorize_vector_modes):
14683 Add V4QImode and V16QImode for TARGET_SSE2.
14684 * doc/sourcebuild.texi (Vector-specific attributes):
14685 Add vect64 and vect32 description.
14687 2021-05-26 Bernd Edlinger <bernd.edlinger@hotmail.de>
14689 * gimple-range-gori.cc (range_def_chain::register_dependency):
14690 Resize m_def_chain when needed.
14692 2021-05-26 Christophe Lyon <christophe.lyon@linaro.org>
14694 * config/arm/mve.md (mve_vaddvq_<supf><mode>): Prefix with '@'.
14695 * config/arm/neon.md (reduc_plus_scal_<mode>): Move to ..
14696 * config/arm/vec-common.md: .. here. Add support for MVE.
14698 2021-05-26 Jakub Jelinek <jakub@redhat.com>
14700 * config/epiphany/epiphany.c (epiphany_print_operand_address): Remove
14702 * config/microblaze/microblaze.c (microblaze_legitimize_address,
14704 microblaze_option_override, print_operand): Likewise.
14705 * config/microblaze/microblaze.md (call_internal_plt,
14706 call_value_intern_plt, call_value_intern): Likewise.
14707 * config/arm/aout.h (ASM_OUTPUT_ALIGN): Likewise.
14708 * config/iq2000/iq2000.md (call_internal1, call_value_internal1,
14709 call_value_multiple_internal1): Likewise.
14710 * config/bfin/bfin.c (symbolic_reference_mentioned_p): Likewise.
14712 2021-05-26 Jan-Benedict Glaw <jbglaw@lug-owl.de>
14714 * config/arc/arc.c (arc_address_cost, arc_print_operand_address,
14715 arc_ccfsm_advance, symbolic_reference_mentioned_p,
14716 arc_raw_symbolic_reference_mentioned_p): Remove register
14719 2021-05-26 Jakub Jelinek <jakub@redhat.com>
14722 * omp-low.c: Include omp-offload.h.
14723 (create_omp_child_function): If current_function_decl has
14724 "omp declare target" attribute and is_gimple_omp_offloaded,
14725 remove that attribute from the copy of attribute list and
14726 add "omp target entrypoint" attribute instead.
14727 (lower_omp_target): Mark .omp_data_sizes.* and .omp_data_kinds.*
14728 variables for offloading if in omp_maybe_offloaded_ctx.
14729 * omp-offload.c (pass_omp_target_link::execute): Nullify second
14730 argument to GOMP_target_data_ext in offloaded code.
14732 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14734 * config/csky/csky.c (csky_can_change_mode_class): Delete.
14735 For csky, HF/SF mode use the low bits of VREGS.
14737 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
14739 * gimplify.c (gimplify_decl_expr): Do not clear TREE_READONLY on a
14740 DECL which is a reference for OMP.
14742 2021-05-26 Martin Liska <mliska@suse.cz>
14744 PR gcov-profile/100751
14745 * doc/gcov.texi: Document that __gcov_dump can be called just
14746 once and that __gcov_reset resets run-time counters.
14748 2021-05-26 Martin Liska <mliska@suse.cz>
14750 * doc/install.texi: Port relevant part from install-old.texi
14751 and re-generate list of CPUs and systems.
14753 2021-05-26 Martin Liska <mliska@suse.cz>
14755 * Makefile.in: Remove it.
14756 * doc/include/fdl.texi: Update next/previous chapters.
14757 * doc/install.texi: Likewise.
14758 * doc/install-old.texi: Removed.
14760 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14762 * config/csky/csky.c (ck810_legitimate_index_p): Support
14763 "base + index" with DF mode.
14764 * config/csky/constraints.md ("Y"): New constraint for memory operands
14765 without index register.
14766 * config/csky/csky_insn_fpuv2.md (fpuv3_movdf): Use "Y" instead of "m"
14767 when mov between memory and general registers, and lower their priority.
14768 * config/csky/csky_insn_fpuv3.md (fpuv2_movdf): Likewise.
14770 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14772 * config/csky/csky.c (TARGET_PROMOTE_PROTOTYPES): Delete.
14774 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14776 * config/csky/csky.md (untyped_call): Emit clobber for return
14777 registers to mark them used.
14779 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14781 * config/csky/csky.md (cskyv2_sextend_ldbs): New.
14783 2021-05-26 Andrew Pinski <apinski@marvell.com>
14785 * match.pd (x < 0 ? ~y : y): New patterns.
14787 2021-05-26 Andrew Pinski <apinski@marvell.com>
14789 * match.pd (A?CST1:CST2): Add simplifcations for A?0:+-1, A?+-1:0,
14790 A?POW2:0 and A?0:POW2.
14792 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14794 * gimple-range-gori.cc (class logical_stmt_cache): Delete
14795 (logical_stmt_cache::logical_stmt_cache ): Delete.
14796 (logical_stmt_cache::~logical_stmt_cache): Delete.
14797 (logical_stmt_cache::cache_entry::dump): Delete.
14798 (logical_stmt_cache::get_range): Delete.
14799 (logical_stmt_cache::cached_name ): Delete.
14800 (logical_stmt_cache::same_cached_name): Delete.
14801 (logical_stmt_cache::cacheable_p): Delete.
14802 (logical_stmt_cache::slot_diagnostics ): Delete.
14803 (logical_stmt_cache::dump): Delete.
14804 (gori_compute_cache::gori_compute_cache): Delete.
14805 (gori_compute_cache::~gori_compute_cache): Delete.
14806 (gori_compute_cache::compute_operand_range): Delete.
14807 (gori_compute_cache::cache_stmt): Delete.
14808 * gimple-range-gori.h (gori_compute::compute_operand_range): Remove
14810 (class gori_compute_cache): Delete.
14812 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14814 * gimple-range.cc (fold_using_range::range_of_range_op): Use m_gori
14816 (fold_using_range::range_of_address): Adjust.
14817 (fold_using_range::range_of_phi): Adjust.
14818 * gimple-range.h (class fur_source): Adjust.
14819 (fur_source::fur_source): Adjust.
14821 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14823 * gimple-range-gori.cc (gori_compute::expr_range_at_stmt): Rename
14824 from expr_range_in_bb and adjust.
14825 (gori_compute::compute_name_range_op): Adjust.
14826 (gori_compute::optimize_logical_operands): Adjust.
14827 (gori_compute::compute_logical_operands_in_chain): Adjust.
14828 (gori_compute::compute_operand1_range): Adjust.
14829 (gori_compute::compute_operand2_range): Adjust.
14830 (ori_compute_cache::cache_stmt): Adjust.
14831 * gimple-range-gori.h (gori_compute): Rename prototype.
14833 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14835 * gimple-range.cc (gimple_ranger::range_of_expr): Non-null should be
14836 checked only after range_of_stmt, not range_on_entry.
14837 (gimple_ranger::range_on_entry): Check for non-null in any
14838 predecessor block, if it is not already non-null.
14839 (gimple_ranger::range_on_exit): DOnt check for non-null after
14840 range on entry call.
14841 (gimple_ranger::dump_bb): New. Split from dump.
14842 (gimple_ranger::dump): Adjust.
14843 * gimple-range.h (class gimple_ranger): Adjust.
14845 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14847 * gimple-range-cache.cc (struct range_timestamp): Delete.
14848 (class temporal_cache): Adjust.
14849 (temporal_cache::get_timestamp): Delete.
14850 (temporal_cache::set_dependency): Delete.
14851 (temporal_cache::temporal_value): Adjust.
14852 (temporal_cache::current_p): Take dependencies as params.
14853 (temporal_cache::set_timestamp): Adjust.
14854 (temporal_cache::set_always_current): Adjust.
14855 (ranger_cache::get_non_stale_global_range): Adjust.
14856 (ranger_cache::register_dependency): Delete.
14857 * gimple-range-cache.h (class range_cache): Adjust.
14859 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14861 * gimple-range-gori.cc (range_def_chain::range_def_chain): init
14863 (range_def_chain::~range_def_chain): Dispose of obstack rather than
14864 each individual bitmap.
14865 (range_def_chain::set_import): New.
14866 (range_def_chain::get_imports): New.
14867 (range_def_chain::chain_import_p): New.
14868 (range_def_chain::register_dependency): Rename from build_def_chain
14870 (range_def_chain::def_chain_in_bitmap_p): New.
14871 (range_def_chain::add_def_chain_to_bitmap): New.
14872 (range_def_chain::has_def_chain): Just check first depenedence.
14873 (range_def_chain::get_def_chain): Process imports, use generic
14874 register_dependency routine.
14875 (range_def_chain::dump): New.
14876 (gori_map::gori_map): Allocate import list.
14877 (gori_map::~gori_map): Release imports.
14878 (gori_map::exports): Check for past allocated block size.
14879 (gori_map::imports): New.
14880 (gori_map::def_chain_in_export_p): Delete.
14881 (gori_map::is_import_p): New.
14882 (gori_map::maybe_add_gori): Handle imports.
14883 (gori_map::dump): Adjust output, add imports.
14884 (gori_compute::has_edge_range_p): Remove def_chain_in_export call.
14885 (gori_export_iterator::gori_export_iterator): New.
14886 (gori_export_iterator::next): New.
14887 (gori_export_iterator::get_name): New.
14888 * gimple-range-gori.h (range_def_chain): Add imports and direct
14889 dependecies via struct rdc.
14890 (range_def_chain::depend1): New.
14891 (range_def_chain::depend2): New.
14892 (class gori_map): Adjust.
14893 (FOR_EACH_GORI_IMPORT_NAME): New.
14894 (FOR_EACH_GORI_EXPORT_NAME): New.
14895 (class gori_export_iterator): New.
14897 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14899 * gimple-range-cache.cc (ranger_cache::ranger_cache): Move initial
14900 export cache filling to here.
14901 * gimple-range-gori.cc (gori_compute::gori_compute) : From Here.
14903 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14905 * gimple-range-gori.cc (range_def_chain): Move to gimple-range-gori.h.
14906 (gori_map): Move to gimple-range-gori.h.
14907 (gori_compute::gori_compute): Adjust.
14908 (gori_compute::~gori_compute): Delete.
14909 (gori_compute::compute_operand_range_switch): Adjust.
14910 (gori_compute::compute_operand_range): Adjust.
14911 (gori_compute::compute_logical_operands): Adjust.
14912 (gori_compute::has_edge_range_p ): Adjust.
14913 (gori_compute::set_range_invariant): Delete.
14914 (gori_compute::dump): Adjust.
14915 (gori_compute::outgoing_edge_range_p): Adjust.
14916 * gimple-range-gori.h (class range_def_chain): Relocate here.
14917 (class gori_map): Relocate here.
14918 (class gori_compute): Inherit from gori_map, and adjust.
14920 2021-05-25 Aldy Hernandez <aldyh@redhat.com>
14922 * value-range.cc (range_tests_legacy): Use
14923 build_nonstandard_integer_type instead of int and short.
14925 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
14927 * gimplify.c (gimplify_decl_expr): Clear TREE_READONLY on the DECL
14928 when really creating an initialization statement for it.
14930 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
14932 * tree-inline.c (setup_one_parameter): Fix thinko in new condition.
14934 2021-05-25 Kito Cheng <kito.cheng@sifive.com>
14936 * config/riscv/riscv.h (ASM_SPEC): Pass -mno-relax.
14938 2021-05-25 Martin Liska <mliska@suse.cz>
14940 PR tree-optimization/92860
14942 * optc-save-gen.awk: Remove exceptions.
14944 2021-05-25 Martin Liska <mliska@suse.cz>
14946 * asan.h (sanitize_coverage_p): New function.
14947 * doc/extend.texi: Document it.
14948 * fold-const.c (fold_range_test): Use sanitize_flags_p
14949 instead of flag_sanitize_coverage.
14950 (fold_truth_andor): Likewise.
14951 * sancov.c: Likewise.
14952 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
14953 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Handle
14954 -fsanitize-coverage when inlining.
14956 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
14958 * config/csky/csky-modes.def : Fix copyright.
14960 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
14962 * config/csky/csky-modes.def : Amend copyright.
14963 * config/csky/csky_insn_fpuv2.md : Likewise.
14964 * config/csky/csky_insn_fpuv3.md : Likewise.
14966 2021-05-25 Richard Biener <rguenther@suse.de>
14968 PR middle-end/100727
14969 * calls.c (initialize_argument_information): Explicitely test
14970 for WITH_SIZE_EXPR.
14971 * gimple-expr.c (mark_addressable): Skip outer WITH_SIZE_EXPR.
14973 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14975 * config/csky/csky.h (FRAME_POINTER_REGNUM): Use
14976 HARD_FRAME_POINTER_REGNUM and FRAME_POINTER_REGNUM instead of
14977 the signle definition. The signle definition may not work well
14978 at simplify_subreg_regno().
14979 (HARD_FRAME_POINTER_REGNUM): New.
14980 (ELIMINABLE_REGS): Add for HARD_FRAME_POINTER_REGNUM.
14981 * config/csky/csky.c (get_csky_live_regs, csky_can_eliminate,
14982 csky_initial_elimination_offset, csky_expand_prologue,
14983 csky_expand_epilogue): Add for HARD_FRAME_POINTER_REGNUM.
14985 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14987 * config/csky/csky.c (csky_option_override):
14988 Init csky_arch_isa_features[] in advance, so TARGET_DSP
14989 and TARGET_DIV can be set well.
14991 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14993 * config/csky/constraints.md ("l", "h"): Delete.
14994 * config/csky/csky.h (reg_class, REG_CLASS_NAMES,
14995 REG_CLASS_CONTENTS): Delete LO_REGS and HI_REGS.
14996 * config/csky/csky.c (regno_reg_classm,
14997 csky_secondary_reload, csky_register_move_cost):
14998 Use HILO_REGS instead of LO_REGS and HI_REGS.
15000 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
15002 * config/csky/constraints.md ("W"): New constriant for mem operand
15003 with base reg, index register.
15004 ("Q"): Renamed and modified "csky_valid_fpuv2_mem_operand" to
15005 "csky_valid_mem_constraint_operand" to deal with both "Q" and "W"
15007 ("Dv"): New constraint for const double value that can be used at
15009 * config/csky/csky-modes.def (HFmode): New mode.
15010 * config/csky/csky-protos.h (csky_valid_fpuv2_mem_operand): Rename
15011 to "csky_valid_mem_constraint_operand" and support new constraint
15013 (csky_get_movedouble_length): New.
15014 (fpuv3_output_move): New.
15015 (fpuv3_const_double): New.
15016 * config/csky/csky.c (csky_option_override): New arch CK860 with fpv3.
15017 (decompose_csky_address): Refine.
15018 (csky_print_operand): New "CONST_DOUBLE" operand.
15019 (csky_output_move): Support fpv3 instructions.
15020 (csky_get_movedouble_length): New.
15021 (fpuv3_output_move): New.
15022 (fpuv3_const_double): New.
15023 (csky_emit_compare): Cover float comparsion.
15024 (csky_emit_compare_float): Refine.
15025 (csky_vaild_fpuv2_mem_operand): Rename to
15026 "csky_valid_mem_constraint_operand" and support new constraint "W".
15027 (ck860_rtx_costs): New.
15028 (csky_rtx_costs): Add the cost calculation of CK860.
15029 (regno_reg_class): New vregs for fpuv3.
15030 (csky_dbx_regno): Likewise.
15031 (csky_cpu_cpp_builtins): New builtin macro for fpuv3.
15032 (csky_conditional_register_usage): Suporrot fpuv3.
15033 (csky_dwarf_register_span): Suporrot fpuv3.
15034 (csky_init_builtins, csky_mangle_type): Support "__fp16" type.
15035 (ck810_legitimate_index_p): Support fp16.
15036 * config/csky/csky.h (TARGET_TLS): ADD CK860.
15037 (CSKY_VREG_P, CSKY_VREG_LO_P, CSKY_VREG_HI_P): Support fpuv3.
15038 (TARGET_SINGLE_FPU): Support fpuv3.
15039 (TARGET_SUPPORT_FPV3): New.
15040 (FIRST_PSEUDO_REGISTER): Change to 202 to hold the new fpuv3 registers.
15041 (FIXED_REGISTERS, CALL_REALLY_USED_REGISTERS, REGISTER_NAMES,
15042 REG_CLASS_CONTENTS): Support fpuv3.
15043 * config/csky/csky.md (movsf): Move to cksy_insn_fpu.md and refine.
15044 (csky_movsf_fpv2): Likewise.
15045 (ck801_movsf): Likewise.
15046 (csky_movsf): Likewise.
15048 (csky_movdf_fpv2): Likewise.
15049 (ck801_movdf): Likewise.
15050 (csky_movdf): Likewise.
15051 (movsicc): Refine. Use "comparison_operatior" instead of
15052 "ordered_comparison_operatior".
15053 (addsicc): Likewise.
15054 (CSKY_FIRST_VFP3_REGNUM, CSKY_LAST_VFP3_REGNUM): New constant.
15055 (call_value_internal_vh): New.
15056 * config/csky/csky_cores.def (CK860): New arch and cpu.
15061 * config/csky/csky_insn_fpu.md: Refactor. Separate all float patterns
15062 into emit-patterns and match-patterns, remain the emit-patterns here,
15063 and move the match-patterns to csky_insn_fpuv2.md or
15064 csky_insn_fpuv3.md.
15065 * config/csky/csky_insn_fpuv2.md: New file for fpuv2 instructions.
15066 * config/csky/csky_insn_fpuv3.md: New file and new patterns for fpuv3
15068 * config/csky/csky_isa.def (fcr): New.
15073 (CK860): New definition for ck860.
15074 * config/csky/csky_tables.opt (ck860): New processors ck860,
15075 ck860f. And new arch ck860.
15080 * config/csky/predicates.md (csky_float_comparsion_operator): Delete
15081 "geu", "gtu", "leu", "ltu", which will never appear at float comparison.
15082 * config/csky/t-csky-elf: Support 860.
15083 * config/csky/t-csky-linux: Likewise.
15084 * doc/md.texi: Add "Q" and "W" constraints for C-SKY.
15086 2021-05-24 Aaron Sawdey <acsawdey@linux.ibm.com>
15088 * config/rs6000/genfusion.pl (gen_logical_addsubf): Refactor to
15089 add generation of logical-add and add-logical fusion pairs.
15090 * config/rs6000/rs6000-cpus.def: Add new fusion to ISA 3.1 mask
15092 * config/rs6000/rs6000.c (rs6000_option_override_internal): Turn on
15093 logical-add and add-logical fusion by default.
15094 * config/rs6000/rs6000.opt: Add -mpower10-fusion-logical-add and
15095 -mpower10-fusion-add-logical options.
15096 * config/rs6000/fusion.md: Regenerate file.
15098 2021-05-24 Aldy Hernandez <aldyh@redhat.com>
15100 * value-range.cc (irange::legacy_equal_p): Check type when
15101 comparing VR_VARYING types.
15102 (range_tests_legacy): Test comparing VARYING ranges of different
15105 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
15107 * config/aarch64/aarch64.c (neoversen1_tunings):
15108 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
15110 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
15112 * config/aarch64/aarch64.c (aarch64_classify_symbol): Use GOT for
15113 extern weak symbols. Limit symbol offsets for non-GOT symbols with
15116 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
15118 * config/arm/neon.md (vec_load_lanesxi<mode>)
15119 (vec_store_lanexoi<mode>): Move ...
15120 * config/arm/vec-common.md: here.
15122 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
15124 * config/arm/neon.md (vec_load_lanesoi<mode>)
15125 (vec_store_lanesoi<mode>): Move ...
15126 * config/arm/vec-common.md: here.
15128 2021-05-24 liuhongt <hongtao.liu@intel.com>
15131 * config/i386/i386.c (ix86_gimple_fold_builtin): Replacing
15132 stmt with GIMPLE_NOP when lhs doesn't exist.
15134 2021-05-23 Uroš Bizjak <ubizjak@gmail.com>
15137 * config/i386/mmx.md (*push<VI_32:mode>2_rex64):
15138 New instruction pattern.
15139 (*push<VI_32:mode>2): Ditto.
15140 (push splitter for SSE registers): New splitter.
15142 2021-05-23 Andrew Pinski <apinski@marvell.com>
15144 * match.pd ((A & C) != 0 ? D : 0): Limit to non pointer types.
15146 2021-05-22 Aaron Sawdey <acsawdey@linux.ibm.com>
15148 * config/rs6000/genfusion.pl (gen_addadd): Fix incorrect attr types.
15149 * config/rs6000/fusion.md: Regenerate file.
15151 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
15153 * config/rs6000/genfusion.pl (gen_addadd): New function.
15154 * config/rs6000/fusion.md: Regenerate file.
15155 * config/rs6000/rs6000-cpus.def: Add
15156 OPTION_MASK_P10_FUSION_2ADD to masks.
15157 * config/rs6000/rs6000.c (rs6000_option_override_internal):
15158 Handle default value of OPTION_MASK_P10_FUSION_2ADD.
15159 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2add.
15161 2021-05-21 Jakub Jelinek <jakub@redhat.com>
15163 PR middle-end/99928
15164 * tree.h (OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET): Define.
15165 * gimplify.c (enum gimplify_omp_var_data): Fix up
15166 GOVD_MAP_HAS_ATTACHMENTS value, add GOVD_FIRSTPRIVATE_IMPLICIT.
15167 (omp_lastprivate_for_combined_outer_constructs): If combined target
15168 has GOVD_FIRSTPRIVATE_IMPLICIT set for the decl, change it to
15169 GOVD_MAP | GOVD_SEEN.
15170 (gimplify_scan_omp_clauses): Set GOVD_FIRSTPRIVATE_IMPLICIT for
15171 firstprivate clauses with OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT.
15172 (gimplify_adjust_omp_clauses): For firstprivate clauses with
15173 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT either clear that bit and
15174 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET too, or remove it and
15175 let it be replaced by implicit map clause.
15177 2021-05-21 Jakub Jelinek <jakub@redhat.com>
15179 PR middle-end/99928
15180 * gimplify.c (omp_lastprivate_for_combined_outer_constructs): New
15182 (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LASTPRIVATE>: Use it.
15183 (gimplify_omp_for): Likewise.
15185 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15187 PR middle-end/90115
15188 * omp-low.c (oacc_privatization_candidate_p): Reject 'static',
15189 'external' in blocks.
15191 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15193 PR middle-end/90115
15194 * flag-types.h (enum openacc_privatization): New.
15195 * params.opt (-param=openacc-privatization): New.
15196 * doc/invoke.texi (openacc-privatization): Document it.
15197 * omp-general.h (get_openacc_privatization_dump_flags): New
15199 * omp-low.c (oacc_privatization_candidate_p): Add diagnostics.
15200 * omp-offload.c (execute_oacc_device_lower)
15201 <IFN_UNIQUE_OACC_PRIVATE>: Re-work diagnostics.
15202 * target.def (goacc.adjust_private_decl): Add 'location_t'
15204 * doc/tm.texi: Regenerate.
15205 * config/gcn/gcn-protos.h (gcn_goacc_adjust_private_decl): Adjust.
15206 * config/gcn/gcn-tree.c (gcn_goacc_adjust_private_decl): Likewise.
15207 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl):
15208 Likewise. Preserve it for...
15209 (nvptx_goacc_expand_var_decl): ... use here.
15211 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15213 * doc/sourcebuild.texi (Other attributes): Document '__OPTIMIZE__'
15216 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15218 PR middle-end/90115
15219 * omp-low.c (oacc_privatization_candidate_p): New function.
15220 (oacc_privatization_scan_clause_chain)
15221 (oacc_privatization_scan_decl_chain): Use it. Also
15222 'gcc_checking_assert' that we're not seeing duplicates.
15224 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15226 PR middle-end/90115
15227 * omp-offload.c (execute_oacc_device_lower): Skip processing if no
15230 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15232 PR middle-end/90115
15233 * omp-offload.c (execute_oacc_device_lower): Explain.
15235 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15237 PR middle-end/90115
15238 * omp-offload.c (execute_oacc_device_lower)
15239 <IFN_UNIQUE_OACC_PRIVATE>: Diagnose and handle for 'level == -1'
15241 * internal-fn.c (expand_UNIQUE): Don't expect
15242 'IFN_UNIQUE_OACC_PRIVATE'.
15244 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15246 PR middle-end/90115
15247 * omp-low.c (lower_omp_for): Don't evaluate OpenMP 'for' clauses.
15249 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15251 PR middle-end/90115
15252 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl)
15253 (nvptx_goacc_expand_var_decl): Tighten.
15255 2021-05-21 Julian Brown <julian@codesourcery.com>
15256 Chung-Lin Tang <cltang@codesourcery.com>
15257 Thomas Schwinge <thomas@codesourcery.com>
15259 PR middle-end/90115
15260 * doc/tm.texi.in (TARGET_GOACC_EXPAND_VAR_DECL)
15261 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Add documentation hooks.
15262 * doc/tm.texi: Regenerate.
15263 * expr.c (expand_expr_real_1): Expand decls using the
15264 expand_var_decl OpenACC hook if defined.
15265 * internal-fn.c (expand_UNIQUE): Handle IFN_UNIQUE_OACC_PRIVATE.
15266 * internal-fn.h (IFN_UNIQUE_CODES): Add OACC_PRIVATE.
15267 * omp-low.c (omp_context): Add oacc_privatization_candidates
15269 (lower_oacc_reductions): Add PRIVATE_MARKER parameter. Insert
15271 (lower_oacc_head_tail): Add PRIVATE_MARKER parameter. Modify
15272 private marker's gimple call arguments, and pass it to
15273 lower_oacc_reductions.
15274 (oacc_privatization_scan_clause_chain)
15275 (oacc_privatization_scan_decl_chain, lower_oacc_private_marker):
15277 (lower_omp_for, lower_omp_target, lower_omp_1): Use these.
15278 * omp-offload.c (convert.h): Include.
15279 (oacc_loop_xform_head_tail): Treat private-variable markers like
15280 fork/join when transforming head/tail sequences.
15281 (struct var_decl_rewrite_info): Add struct.
15282 (oacc_rewrite_var_decl, is_sync_builtin_call): New functions.
15283 (execute_oacc_device_lower): Support rewriting gang-private
15284 variables using target hook, and fix up addr_expr and var_decl
15286 * target.def (adjust_private_decl, expand_var_decl): New hooks.
15287 * config/gcn/gcn-protos.h (gcn_goacc_adjust_gangprivate_decl):
15289 (gcn_goacc_adjust_private_decl): ...this.
15290 * config/gcn/gcn-tree.c (gcn_goacc_adjust_gangprivate_decl):
15292 (gcn_goacc_adjust_private_decl): ...this. Add LEVEL parameter.
15293 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_GANGPRIVATE_DECL): Rename
15294 definition using gcn_goacc_adjust_gangprivate_decl...
15295 (TARGET_GOACC_ADJUST_PRIVATE_DECL): ...to this, using
15296 gcn_goacc_adjust_private_decl.
15297 * config/nvptx/nvptx.c (tree-pretty-print.h): Include.
15298 (gang_private_shared_size): New global variable.
15299 (gang_private_shared_align): Likewise.
15300 (gang_private_shared_sym): Likewise.
15301 (gang_private_shared_hmap): Likewise.
15302 (nvptx_option_override): Initialize these.
15303 (nvptx_file_end): Output gang_private_shared_sym.
15304 (nvptx_goacc_adjust_private_decl, nvptx_goacc_expand_var_decl):
15306 (nvptx_set_current_function): Clear gang_private_shared_hmap.
15307 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Define hook.
15308 (TARGET_GOACC_EXPAND_VAR_DECL): Likewise.
15310 2021-05-21 H.J. Lu <hjl.tools@gmail.com>
15312 * config/i386/i386-modes.def (MAX_BITSIZE_MODE_ANY_INT): Removed.
15314 2021-05-21 Richard Biener <rguenther@suse.de>
15315 H.J. Lu <hjl.tools@gmail.com>
15317 PR middle-end/90773
15318 * expr.c (expand_constructor): Elide expand_constructor if
15319 move by pieces is preferred.
15321 2021-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15323 * config/aarch64/aarch64-builtins.c (aarch64_call_properties):
15324 Take a flag and mode value as arguments.
15325 (aarch64_modifies_global_state_p): Likewise.
15326 (aarch64_reads_global_state_p): Likewise.
15327 (aarch64_could_trap_p): Likewise.
15328 (aarch64_get_attributes): Likewise.
15329 (aarch64_init_simd_builtins): Adjust callsite of above.
15330 (aarch64_init_fcmla_laneq_builtins): Use aarch64_get_attributes to get
15331 function attributes to apply to builtins.
15332 (aarch64_init_crc32_builtins): Likewise.
15333 (aarch64_init_builtin_rsqrt): Likewise.
15335 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
15337 * config/rs6000/rs6000.md (define_attr "type"): Add types for fusion.
15338 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10): Use new fusion types.
15339 (gen_2logical): Use new fusion types.
15340 * config/rs6000/fusion.md: Regenerate.
15342 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
15345 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
15346 Handle V4QI and V2HI modes.
15347 (ix86_expand_sse_movcc): Ditto.
15348 * config/i386/mmx.md (*<sat_plusminus:insn><VI_32:mode>3):
15349 New instruction pattern.
15350 (*eq<VI_32:mode>3): Ditto.
15351 (*gt<VI_32:mode>3): Ditto.
15352 (*xop_pcmov_<VI_32:mode>): Ditto.
15353 (mmx_pblendvb32): Ditto.
15354 (mmx_pblendvb64): Rename from mmx_pblendvb.
15355 (vec_cmp<VI_32:mode><VI_32:mode>): New expander.
15356 (vec_cmpu<VI_32:mode><VI_32:mode>): Ditto.
15357 (vcond<VI_32:mode><VI_32:mode>): Ditto.
15358 (vcondu<VI_32:mode><VI_32:mode>): Ditto.
15359 (vcond_mask_<VI_32:mode><VI_32:mode>): Ditto.
15361 2021-05-21 Jakub Jelinek <jakub@redhat.com>
15363 PR tree-optimization/94589
15364 * tree-ssa-phiopt.c (spaceship_replacement): For integral rhs1 and
15365 rhs2, treat x <= 4 equivalently to x < 5 etc. In cmp1 and cmp2 (if
15366 not the same as cmp3) treat <= the same as < and >= the same as >.
15367 Don't require that cond2_phi_edge is true edge, instead take
15368 false/true edges into account based on cmp1/cmp2 comparison kinds.
15370 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
15373 * config/i386/mmx.md (SMAXMIN_MMXMODEI): New mode iterator.
15374 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Macroize expander
15375 from <smaxmin:code>v4hi3> and <smaxmin:code><MMXMODE14:mode>3
15376 using SMAXMIN_MMXMODEI mode iterator.
15377 (*<smaxmin:code>v4qi3): New insn pattern.
15378 (*<smaxmin:code>v2hi3): Ditto.
15379 (SMAXMIN_VI_32): New mode iterator.
15380 (<smaxmin:code><SMAXMIN_VI_32>mode3): New expander.
15381 (UMAXMIN_MMXMODEI): New mode iterator.
15382 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Macroize expander
15383 from <umaxmin:code>v8qi3> and <umaxmin:code><MMXMODE24:mode>3
15384 using UMAXMIN_MMXMODEI mode iterator.
15385 (*<umaxmin:code>v4qi3): New insn pattern.
15386 (*<umaxmin:code>v2hi3): Ditto.
15387 (UMAXMIN_VI_32): New mode iterator.
15388 (<umaxmin:code><UMAXMIN_VI_32>mode3): New expander.
15389 (abs<VI_32:mode>2): New insn pattern.
15390 (ssse3_abs<MMXMODEI:mode>2, abs<MMXMODEI:mode>2): Move from ...
15391 * config/i386/sse.md: ... here.
15393 2021-05-20 Clement Chigot <clement.chigot@atos.net>
15394 David Edelsohn <dje.gcc@gmail.com>
15396 * collect2.c (scan_prog_file): Issue non-fatal warning for
15399 2021-05-20 Jonathan Wakely <jwakely@redhat.com>
15401 * doc/invoke.texi (-Wno-c++11-extensions)
15402 (-Wno-c++14-extensions, -Wno-c++17-extensions)
15403 (-Wno-c++20-extensions, -Wno-c++23-extensions): Document
15406 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
15408 * config/c6x/c6x.c (c6x_output_file_unwind): Use dwarf_debuginfo_p.
15409 * config/darwin.c (darwin_override_options): Likewise.
15410 * config/i386/cygming.h (DBX_REGISTER_NUMBER): Likewise.
15411 * config/i386/darwin.h (DBX_REGISTER_NUMBER): Likewise.
15412 (DWARF2_FRAME_REG_OUT): Likewise.
15413 * config/mips/mips.c (mips_output_filename): Likewise.
15414 * config/rs6000/rs6000.c (rs6000_xcoff_declare_function_name):
15416 (rs6000_dbx_register_number): Likewise.
15417 * dbxout.c: Include flags.h.
15418 * dwarf2cfi.c (cfi_label_required_p): Likewise.
15419 (dwarf2out_do_frame): Likewise.
15420 * except.c: Include flags.h.
15421 * final.c (dwarf2_debug_info_emitted_p): Likewise.
15422 (final_scan_insn_1): Likewise.
15423 * flags.h (dwarf_debuginfo_p): New function declaration.
15424 * opts.c (dwarf_debuginfo_p): New function definition.
15425 * targhooks.c (default_debug_unwind_info): Use dwarf_debuginfo_p.
15426 * toplev.c (process_options): Likewise.
15428 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
15430 * common.opt: Change type to support bitmasks.
15431 * flag-types.h (enum debug_info_type): Rename enumerator constants.
15432 (NO_DEBUG): New bitmask.
15433 (DBX_DEBUG): Likewise.
15434 (DWARF2_DEBUG): Likewise.
15435 (XCOFF_DEBUG): Likewise.
15436 (VMS_DEBUG): Likewise.
15437 (VMS_AND_DWARF2_DEBUG): Likewise.
15438 * flags.h (debug_set_to_format): New function declaration.
15439 (debug_set_count): Likewise.
15440 (debug_set_names): Likewise.
15441 * opts.c (debug_type_masks): Array of bitmasks for debug formats.
15442 (debug_set_to_format): New function definition.
15443 (debug_set_count): Likewise.
15444 (debug_set_names): Likewise.
15445 (set_debug_level): Update access to debug_type_names.
15446 * toplev.c: Likewise.
15448 2021-05-20 Martin Sebor <msebor@redhat.com>
15450 PR middle-end/100684
15451 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Handle C++ lambda.
15453 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
15456 * config/i386/i386.md (isa): Remove x64_bmi.
15457 (enabled): Remove x64_bmi.
15458 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
15459 Remove general register alternative.
15460 (*andnot<VI_32:mode>3): Ditto.
15461 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
15462 (*<any_logic:code><VI_32:mode>3): Ditto.
15464 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
15466 * config/arm/arm.c: Include head files tree-vectorizer.h and
15469 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
15472 * config/i386/mmx.md (Yv_Yw): Revert adding V4QI and V2HI modes.
15473 (*<plusminus:insn><VI_32:mode>3): Use Yw instad of <Yv_Yw> constrint.
15474 (<s>mulv4hi3_highpart): New expander.
15475 (*<s>mulv2hi3_highpart): New insn pattern.
15476 (<s>mulv2hi3_higpart): New expander.
15477 (*<any_shift:insn>v2hi3): New insn pattern.
15478 (<any_shift:insn>v2hi3): New expander.
15479 * config/i386/sse.md (smulhrsv2hi3): New expander.
15480 (*smulhrsv2hi3): New insn pattern.
15482 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
15484 * doc/invoke.texi (vect-inner-loop-cost-factor): Document new
15486 * params.opt (vect-inner-loop-cost-factor): New.
15487 * targhooks.c (default_add_stmt_cost): Replace hardcoded factor
15488 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR, include head file
15489 tree-vectorizer.h and its required ones.
15490 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Replace
15491 hardcoded factor 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR.
15492 * config/arm/arm.c (arm_add_stmt_cost): Likewise.
15493 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
15494 * config/rs6000/rs6000.c (rs6000_add_stmt_cost): Likewise.
15495 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
15497 (_loop_vec_info::_loop_vec_info): Init inner_loop_cost_factor.
15498 * tree-vectorizer.h (_loop_vec_info): Add inner_loop_cost_factor.
15499 (LOOP_VINFO_INNER_LOOP_COST_FACTOR): New macro.
15501 2021-05-20 Christophe Lyon <christophe.lyon@linaro.org>
15502 Torbjörn Svensson <torbjorn.svensson@st.com>
15505 * doc/cpp.texi (Common Predefined Macros): Document __FILE_NAME__.
15507 2021-05-20 Jakub Jelinek <jakub@redhat.com>
15509 PR middle-end/99928
15510 * gimplify.c (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LINEAR>: For
15511 explicit linear clause when combined with target, make it map(tofrom:)
15512 instead of no clause or firstprivate.
15514 2021-05-20 Jakub Jelinek <jakub@redhat.com>
15516 PR tree-optimization/94589
15517 * match.pd ((X & Y) == X -> (X & ~Y) == 0): Simplify even in presence
15518 of integral conversions.
15520 2021-05-19 Andrew MacLeod <amacleod@redhat.com>
15522 * gimple-range.cc (fur_source::get_operand): New.
15523 (gimple_range_fold): Delete.
15524 (fold_using_range::fold_stmt): Move from gimple_ranger::calc_stmt.
15525 (fold_using_range::range_of_range_op): Move from gimple_ranger.
15526 (fold_using_range::range_of_address): Ditto.
15527 (fold_using_range::range_of_phi): Ditto.
15528 (fold_using_range::range_of_call): Ditto.
15529 (fold_using_range::range_of_builtin_ubsan_call): Move from
15530 range_of_builtin_ubsan_call.
15531 (fold_using_range::range_of_builtin_call): Move from
15532 range_of_builtin_call.
15533 (gimple_ranger::range_of_builtin_call): Delete.
15534 (fold_using_range::range_of_cond_expr): Move from gimple_ranger.
15535 (gimple_ranger::fold_range_internal): New.
15536 (gimple_ranger::range_of_stmt): Use new fold_using_range API.
15537 (fold_using_range::range_of_ssa_name_with_loop_info): Move from
15538 gimple_ranger. Improve ranges of SSA_NAMES when possible.
15539 * gimple-range.h (gimple_ranger): Remove various range_of routines.
15540 (class fur_source): New.
15541 (class fold_using_range): New.
15542 (fur_source::fur_source): New.
15544 * vr-values.c (vr_values::extract_range_basic): Use fold_using_range
15545 instead of range_of_builtin_call.
15547 2021-05-19 Jonathan Wakely <jwakely@redhat.com>
15549 * doc/cpp.texi (Common Predefined Macros): Update documentation
15550 for the __GXX_EXPERIMENTAL_CXX0X__ macro.
15552 2021-05-19 Alex Coplan <alex.coplan@arm.com>
15555 * config/arm/arm.md (nonsecure_call_internal): Always ensure
15556 callee's address is in a register.
15558 2021-05-19 Geng Qi <gengqi@linux.alibaba.com>
15560 * common/config/riscv/riscv-common.c
15561 (riscv_subset_list::parsing_subset_version): Properly parse the letter
15563 (riscv_subset_list::parse_std_ext,
15564 riscv_subset_list::parse_multiletter_ext): To handle errors generated
15565 in riscv_subset_list::parsing_subset_version.
15567 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15569 * config/aarch64/aarch64-simd.md: Use "neon_move_narrow_q"
15570 type attribute in patterns generating XTN(2).
15572 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15574 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_pack_trunc_<mode>):
15575 Remove as duplicate of...
15576 (aarch64_xtn<mode>): This.
15577 (aarch64_xtn2<mode>_le): Move position in file.
15578 (aarch64_xtn2<mode>_be): Move position in file.
15579 (aarch64_xtn2<mode>): Move position in file.
15580 (vec_pack_trunc_<mode>): Define as an expander.
15582 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15584 * config/aarch64/aarch64-simd-builtins.def: Split builtin
15585 generation for aarch64_<sur>q<r>shr<u>n_n<mode> pattern into
15586 separate scalar and vector generators.
15587 * config/aarch64/aarch64-simd.md
15588 (aarch64_<sur>q<r>shr<u>n_n<mode>): Define as an expander and
15590 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_le): This and...
15591 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_be): This.
15592 * config/aarch64/iterators.md: Define SD_HSDI iterator.
15594 2021-05-19 Jonathn Wright <jonathan.wright@arm.com>
15596 * config/aarch64/aarch64-simd.md: Use UNSPEC_SQXTUN instead
15598 * config/aarch64/iterators.md: Remove UNSPEC_SQXTUN2.
15600 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15602 * config/aarch64/aarch64-simd.md (aarch64_<sur>q<r>shr<u>n2_n<mode>):
15603 Implement as an expand emitting a big/little endian
15604 instruction pattern.
15605 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_le): Define.
15606 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_be): Define.
15608 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15610 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn2<mode>):
15611 Implement as an expand emitting a big/little endian
15612 instruction pattern.
15613 (aarch64_<sur><addsub>hn2<mode>_insn_le): Define.
15614 (aarch64_<sur><addsub>hn2<mode>_insn_be): Define.
15615 * config/aarch64/iterators.md: Remove UNSPEC_[R]ADDHN2 and
15616 UNSPEC_[R]SUBHN2 unspecs and ADDSUBHN2 iterator.
15618 2021-05-19 Richard Biener <rguenther@suse.de>
15620 PR middle-end/100672
15621 * fold-const.c (fold_negate_expr_1): Use element_precision.
15622 (negate_expr_p): Likewise.
15624 2021-05-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
15626 * config/aarch64/iterators.md (SVE_PRED_LOAD): New iterator.
15627 (pred_load): New int attribute.
15628 * config/aarch64/aarch64-sve.md
15629 (aarch64_load_<ANY_EXTEND:optab><SVE_HSDI:mode><SVE_PARTIAL_I:mode>): Use
15630 SVE_PRED_LOAD enum iterator and corresponding pred_load attribute.
15631 * config/aarch64/aarch64-sve-builtins-base.cc (expand): Update call to
15632 code_for_aarch64_load.
15634 2021-05-19 Richard Biener <rguenther@suse.de>
15636 * cfgexpand.c (discover_nonconstant_array_refs_r): Make
15637 sure TARGET_MEM_REF bases are expanded as memory.
15638 * tree-ssa-operands.c (operands_scanner::get_tmr_operands):
15639 Do not mark TARGET_MEM_REF bases addressable.
15640 * tree-ssa.c (non_rewritable_mem_ref_base): Handle
15641 TARGET_MEM_REF bases as never rewritable.
15642 * gimple-walk.c (walk_stmt_load_store_addr_ops): Do not
15643 walk TARGET_MEM_REF bases as address-takens.
15644 * tree-ssa-dce.c (ref_may_be_aliased): Handle TARGET_MEM_REF.
15646 2021-05-19 Richard Biener <rguenther@suse.de>
15648 * builtins.c (get_object_alignment_1): Strip outer
15650 * tree-dfa.c (get_ref_base_and_extent): Handle outer
15651 WITH_SIZE_EXPR for size processing and process the
15653 * tree-ssa-alias.c (ao_ref_base_alias_set): Strip
15654 outer WITH_SIZE_EXPR.
15655 (ao_ref_base_alias_ptr_type): Likewise.
15656 (refs_may_alias_p_2): Allow WITH_SIZE_EXPR in ref->ref
15657 and handle that accordingly, stripping it for the
15658 core alias workers.
15659 * tree.c (get_base_address): Handle WITH_SIZE_EXPR by
15660 looking through it instead of returning NULL.
15662 2021-05-19 Jakub Jelinek <jakub@redhat.com>
15664 PR middle-end/100576
15665 * builtins.c (check_read_access): Convert bound to size_type_node if
15668 2021-05-19 Richard Biener <rguenther@suse.de>
15670 * tree-cfg.c (verify_types_in_gimple_min_lval): Inline...
15671 (verify_types_in_gimple_reference): ... here. Sanitize.
15672 (verify_gimple_call): Verify references in LHS and arguments.
15673 (verify_gimple_assign_single): Reject WITH_SIZE_EXPR.
15675 2021-05-19 Uroš Bizjak <ubizjak@gmail.com>
15677 * config/i386/i386.h (VALID_INT_MODE_P):
15678 Add V8QI, V4HI and V2SI modes for TARGET_64BIT.
15679 * config/i386/i386.md (isa): Add x64_bmi.
15680 (enabled): Handle x64_bmi.
15681 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
15682 Add alternative using 64bit general registers.
15683 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
15685 2021-05-19 Jakub Jelinek <jakub@redhat.com>
15687 PR middle-end/99928
15688 * tree.h (OMP_MASTER_COMBINED): Define.
15689 * gimplify.c (gimplify_scan_omp_clauses): Rewrite lastprivate
15690 handling for outer combined/composite constructs to a loop.
15691 Handle lastprivate on combined target.
15692 (gimplify_expr): Formatting fix.
15694 2021-05-19 Xionghu Luo <luoxhu@linux.ibm.com>
15696 * passes.def: Add sink_code pass before store_merging.
15697 * tree-ssa-sink.c (pass_sink_code:clone): New.
15699 2021-05-18 Bill Schmidt <wschmidt@linux.ibm.com>
15701 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Remove call to
15702 rs6000_special_adjust_field_align_p.
15703 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
15704 * config/rs6000/rs6000-call.c (rs6000_function_arg_boundary):
15705 Remove ABI warning.
15706 (rs6000_function_arg): Likewise.
15707 * config/rs6000/rs6000-protos.h
15708 (rs6000_special_adjust_field_align_p): Remove prototype.
15709 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
15711 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Remove call to
15712 rs6000_special_adjust_field_align_p.
15714 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
15717 * config/i386/i386.h (VALID_SSE2_REG_MODE):
15718 Add V4QI and V2HI modes.
15719 (VALID_INT_MODE_P): Ditto.
15720 * config/i386/mmx.md (VI_32): New mode iterator.
15721 (mmxvecsize): Handle V4QI and V2HI.
15723 (mov<VI_32:mode>): New expander.
15724 (*mov<mode>_internal): New insn pattern.
15725 (movmisalign<VI_32:mode>): New expander.
15726 (neg<VI_32:mode>): New expander.
15727 (<plusminus:insn><VI_32:mode>3): New expander.
15728 (*<plusminus:insn><VI_32:mode>3): New insn pattern.
15729 (mulv2hi3): New expander.
15730 (*mulv2hi3): New insn pattern.
15731 (one_cmpl<VI_32:mode>2): New expander.
15732 (*andnot<VI_32:mode>3): New insn pattern.
15733 (<any_logic:code><VI_32:mode>3): New expander.
15734 (*<any_logic:code><VI_32:mode>3): New insn pattern.
15736 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
15738 * config/i386/sse.md (<any_extend:insn>v4qiv4di2):
15739 Fix a mode mismatch with operand 1.
15741 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
15744 * config/i386/i386-expand.c (split_double_mode): Return
15745 temporary register when simplify_gen_subreg fails with
15746 the high half od the paradoxical subreg.
15748 2021-05-18 Richard Biener <rguenther@suse.de>
15750 * cfgexpand.c (expand_one_var): Pass in forced_stack_var
15751 and honor it when expanding.
15752 (expand_used_vars_for_block): Pass through forced_stack_var.
15753 (expand_used_vars): Likewise.
15754 (discover_nonconstant_array_refs_r): Set bits in
15755 forced_stack_vars instead of marking vars TREE_ADDRESSABLE.
15756 (avoid_type_punning_on_regs): Likewise.
15757 (discover_nonconstant_array_refs): Likewise.
15758 (pass_expand::execute): Create and pass down forced_stack_var
15759 bitmap. For parameters and returns temporarily set
15760 TREE_ADDRESSABLE when expand_function_start.
15762 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
15764 * doc/sourcebuild.texi: Document 'dg-note'.
15766 2021-05-18 Tobias Burnus <tobias@codesourcery.com>
15769 * configure: Regenerate.
15770 * configure.ac (BUILD_CFLAG, BUILD_CXXFLAGS): Add $(CFLAGS-$@).
15772 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
15774 * gimple.h (is_gimple_omp_oacc): Tighten.
15775 * omp-low.c (check_omp_nesting_restrictions): Adjust.
15777 2021-05-18 Richard Biener <rguenther@suse.de>
15779 * tree-ssa-operands.c (mark_address_taken): Simplify.
15781 2021-05-18 Martin Liska <mliska@suse.cz>
15783 * config/gcn/mkoffload.c (STR): Redefine.
15784 * config/i386/intelmic-mkoffload.c (STR): Likewise.
15785 * config/nvptx/mkoffload.c (STR): Likewise.
15787 2021-05-18 Martin Liska <mliska@suse.cz>
15789 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
15790 Use startswith function instead of strncmp.
15791 * common/config/bfin/bfin-common.c (bfin_handle_option): Likewise.
15792 * common/config/riscv/riscv-common.c (riscv_subset_list::parse): Likewise.
15793 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): Likewise.
15794 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Likewise.
15795 * config/alpha/alpha.c (alpha_elf_section_type_flags): Likewise.
15796 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
15797 * config/arm/arm.c (arm_file_start): Likewise.
15798 (arm_valid_target_attribute_rec): Likewise.
15799 (thumb1_md_asm_adjust): Likewise.
15800 * config/arm/driver-arm.c (host_detect_local_cpu): Likewise.
15801 * config/avr/avr.c (STR_PREFIX_P): Likewise.
15802 (avr_set_current_function): Likewise.
15803 (avr_handle_addr_attribute): Likewise.
15804 (avr_asm_output_aligned_decl_common): Likewise.
15805 (avr_asm_named_section): Likewise.
15806 (avr_section_type_flags): Likewise.
15807 (avr_asm_select_section): Likewise.
15808 * config/c6x/c6x.c (c6x_in_small_data_p): Likewise.
15809 (c6x_section_type_flags): Likewise.
15810 * config/darwin-c.c (darwin_cfstring_ref_p): Likewise.
15811 (darwin_objc_declare_unresolved_class_reference): Likewise.
15812 (darwin_objc_declare_class_definition): Likewise.
15813 * config/darwin.c (indirect_data): Likewise.
15814 (darwin_encode_section_info): Likewise.
15815 (darwin_objc2_section): Likewise.
15816 (darwin_objc1_section): Likewise.
15817 (machopic_select_section): Likewise.
15818 (darwin_globalize_label): Likewise.
15819 (darwin_label_is_anonymous_local_objc_name): Likewise.
15820 (darwin_asm_named_section): Likewise.
15821 (darwin_asm_output_dwarf_offset): Likewise.
15822 * config/frv/frv.c (frv_string_begins_with): Likewise.
15823 (frv_in_small_data_p): Likewise.
15824 * config/gcn/mkoffload.c (STR): Likewise.
15826 * config/i386/i386-builtins.c (get_builtin_code_for_version): Likewise.
15827 * config/i386/i386-options.c (ix86_option_override_internal): Likewise.
15828 * config/i386/i386.c (x86_64_elf_section_type_flags): Likewise.
15829 (ix86_md_asm_adjust): Likewise.
15830 * config/i386/intelmic-mkoffload.c (STR): Likewise.
15831 * config/i386/winnt.c (i386_pe_asm_named_section): Likewise.
15832 (i386_pe_file_end): Likewise.
15833 * config/ia64/ia64.c (ia64_in_small_data_p): Likewise.
15834 (ia64_section_type_flags): Likewise.
15835 * config/mips/driver-native.c (host_detect_local_cpu): Likewise.
15836 * config/mips/mips.c (mips_handle_interrupt_attr): Likewise.
15837 (mips16_stub_function_p): Likewise.
15838 (mips_function_rodata_section): Likewise.
15839 * config/msp430/msp430.c (msp430_mcu_name): Likewise.
15840 (msp430_function_section): Likewise.
15841 (msp430_section_type_flags): Likewise.
15842 (msp430_expand_helper): Likewise.
15843 * config/nios2/nios2.c (nios2_small_section_name_p): Likewise.
15844 (nios2_valid_target_attribute_rec): Likewise.
15845 * config/nvptx/mkoffload.c (process): Likewise.
15847 * config/pa/som.h: Likewise.
15848 * config/pdp11/pdp11.c (pdp11_output_ident): Likewise.
15849 * config/riscv/riscv.c (riscv_elf_select_rtx_section): Likewise.
15850 * config/rs6000/rs6000.c (VTABLE_NAME_P): Likewise.
15851 (rs6000_inner_target_options): Likewise.
15852 * config/s390/driver-native.c (s390_host_detect_local_cpu): Likewise.
15853 * config/sparc/driver-sparc.c (host_detect_local_cpu): Likewise.
15854 * config/vax/vax.c (vax_output_int_move): Likewise.
15855 * config/vms/vms-ld.c (startswith): Likewise.
15856 (process_args): Likewise.
15858 * config/vms/vms.c: Likewise.
15860 2021-05-18 Jakub Jelinek <jakub@redhat.com>
15862 PR rtl-optimization/100590
15863 * regcprop.c (copyprop_hardreg_forward_1): Only DCE dead sets if
15864 they are NONJUMP_INSN_P.
15866 2021-05-18 Jakub Jelinek <jakub@redhat.com>
15869 * function.c (push_dummy_function): Set DECL_ARTIFICIAL and
15870 DECL_ASSEMBLER_NAME on the fn_decl.
15872 2021-05-18 Jakub Jelinek <jakub@redhat.com>
15874 PR tree-optimization/94589
15875 * tree-ssa-phiopt.c (spaceship_replacement): Pattern match
15876 phi result used in (res & ~1) == 0 comparison as res >= 0 as
15877 res == 2 would be UB with -ffinite-math-only.
15879 2021-05-18 Martin Liska <mliska@suse.cz>
15881 * Makefile.in: genversion.o should depend on DATESTAMP.
15883 2021-05-18 Claudiu Zissulescu <claziss@synopsys.com>
15885 * config/arc/simdext.md (negv2si2): Remove round bracket.
15887 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
15889 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Define
15890 _Bool as macro expanding to _Bool.
15892 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
15895 * tree.c (build_reference_type_for_mode)
15896 (build_pointer_type_for_mode): Pick pointer mode if MODE argument
15898 (build_reference_type, build_pointer_type): Invoke
15899 build_*_type_for_mode with VOIDmode.
15901 2021-05-17 Andrew MacLeod <amacleod@redhat.com>
15903 PR tree-optimization/100512
15904 * gimple-range-cache.cc (ranger_cache::set_global_range): Mark const
15905 and non-zero pointer ranges as invariant.
15906 * gimple-range.cc (gimple_ranger::range_of_stmt): Remove pointer
15907 processing from here.
15909 2021-05-17 Tom de Vries <tdevries@suse.de>
15912 * config/nvptx/nvptx-protos.h (nvptx_output_atomic_insn): Declare
15913 * config/nvptx/nvptx.c (nvptx_output_barrier)
15914 (nvptx_output_atomic_insn): New function.
15915 (nvptx_print_operand): Add support for 'B'.
15916 * config/nvptx/nvptx.md: Use nvptx_output_atomic_insn for atomic
15919 2021-05-17 Aldy Hernandez <aldyh@redhat.com>
15921 PR tree-optimization/100349
15922 * vr-values.c (bounds_of_var_in_loop): Bail if scev returns
15925 2021-05-17 Tamar Christina <tamar.christina@arm.com>
15927 * config/aarch64/driver-aarch64.c (DEFAULT_ARCH): New.
15928 (host_detect_local_cpu): Use it.
15930 2021-05-17 Martin Liska <mliska@suse.cz>
15932 * doc/invoke.texi: Add 2 missing dots.
15934 2021-05-17 Marius Hillenbrand <mhillen@linux.ibm.com>
15936 PR bootstrap/100552
15937 * configure.ac: Replace pattern substitution with call to sed.
15938 * configure: Regenerate.
15940 2021-05-17 Richard Biener <rguenther@suse.de>
15942 PR middle-end/100582
15943 * tree.c (array_at_struct_end_p): Get to the base of the
15944 reference before looking for the underlying decl.
15946 2021-05-17 Joern Rennecke <joern.rennecke@embecosm.com>
15948 * genoutput.c (validate_insn_alternatives) Make "wrong number of
15949 alternatives" message more specific, and remove assumption on where
15952 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
15954 * config/arm/iterators.md (V16): New iterator.
15955 (VH_cvtto): New iterator.
15956 (v_cmp_result): Added V4HF and V8HF support.
15957 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result>): Use VDQWH.
15958 (vcond<mode><mode>): Likewise.
15959 (vcond_mask_<mode><v_cmp_result>): Likewise.
15960 (vcond<VH_cvtto><mode>): New expander.
15962 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
15964 * config/arm/arm-protos.h (arm_expand_vector_compare): Update
15966 * config/arm/arm.c (arm_expand_vector_compare): Add support for
15968 (arm_expand_vcond): Likewise.
15969 * config/arm/iterators.md (supf): Remove VCMPNEQ_S, VCMPEQQ_S,
15970 VCMPEQQ_N_S, VCMPNEQ_N_S.
15971 (VCMPNEQ, VCMPEQQ, VCMPEQQ_N, VCMPNEQ_N): Remove.
15972 * config/arm/mve.md (@mve_vcmp<mve_cmp_op>q_<mode>): Add '@' prefix.
15973 (@mve_vcmp<mve_cmp_op>q_f<mode>): Likewise.
15974 (@mve_vcmp<mve_cmp_op>q_n_f<mode>): Likewise.
15975 (@mve_vpselq_<supf><mode>): Likewise.
15976 (@mve_vpselq_f<mode>"): Likewise.
15977 * config/arm/neon.md (vec_cmp<mode><v_cmp_result): Enable for MVE
15978 and move to vec-common.md.
15979 (vec_cmpu<mode><mode>): Likewise.
15980 (vcond<mode><mode>): Likewise.
15981 (vcond<V_cvtto><mode>): Likewise.
15982 (vcondu<mode><v_cmp_result>): Likewise.
15983 (vcond_mask_<mode><v_cmp_result>): Likewise.
15984 * config/arm/unspecs.md (VCMPNEQ_U, VCMPNEQ_S, VCMPEQQ_S)
15985 (VCMPEQQ_N_S, VCMPNEQ_N_S, VCMPEQQ_U, CMPEQQ_N_U, VCMPNEQ_N_U)
15986 (VCMPGEQ_N_S, VCMPGEQ_S, VCMPGTQ_N_S, VCMPGTQ_S, VCMPLEQ_N_S)
15987 (VCMPLEQ_S, VCMPLTQ_N_S, VCMPLTQ_S, VCMPCSQ_N_U, VCMPCSQ_U)
15988 (VCMPHIQ_N_U, VCMPHIQ_U): Remove.
15989 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result): Moved
15991 (vec_cmpu<mode><mode>): Likewise.
15992 (vcond<mode><mode>): Likewise.
15993 (vcond<V_cvtto><mode>): Likewise.
15994 (vcondu<mode><v_cmp_result>): Likewise.
15995 (vcond_mask_<mode><v_cmp_result>): Likewise. Added unsafe math
15998 2021-05-17 liuhongt <hongtao.liu@intel.com>
16001 * config/i386/i386.c (ix86_gimple_fold_builtin): Use
16002 gsi_insert_seq_before instead.
16004 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
16006 * doc/sourcebuild.texi (arm_qbit_ok): Rename into...
16007 (arm_sat_ok): ...this.
16009 2021-05-17 Martin Liska <mliska@suse.cz>
16011 * lto-wrapper.c (merge_flto_options): Factor out a new function.
16012 (merge_and_complain): Use it.
16013 (run_gcc): Merge also linker command line -flto=foo argument
16016 2021-05-16 Christophe Lyon <christophe.lyon@linaro.org>
16018 * config/arm/arm.h (CPP_SPEC): Remove error message about
16019 -mlittle-endian/-mbig-endian conflict.
16021 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
16023 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
16024 __ROP_PROTECT__ if -mrop-protect is selected.
16026 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
16028 * config/rs6000/rs6000-internal.h (rs6000_stack): Add
16029 rop_hash_save_offset and rop_hash_size.
16030 * config/rs6000/rs6000-logue.c (rs6000_stack_info): Compute
16031 rop_hash_size and rop_hash_save_offset.
16032 (debug_stack_info): Dump rop_hash_save_offset and rop_hash_size.
16033 (rs6000_emit_prologue): Emit hashst[p] in prologue.
16034 (rs6000_emit_epilogue): Emit hashchk[p] in epilogue.
16035 * config/rs6000/rs6000.md (unspec): Add UNSPEC_HASHST and
16037 (hashst): New define_insn.
16038 (hashchk): Likewise.
16040 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
16042 * config/rs6000/rs6000.c (rs6000_option_override_internal):
16043 Disable shrink wrap when inserting ROP-protect instructions.
16044 * config/rs6000/rs6000.opt (mrop-protect): New option.
16045 (mprivileged): Likewise.
16046 * doc/invoke.texi: Document mrop-protect and mprivileged.
16048 2021-05-15 Hans-Peter Nilsson <hp@axis.com>
16050 * reorg.c (fill_slots_from_thread): Reinstate code typoed out in
16053 2021-05-15 Martin Jambor <mjambor@suse.cz>
16056 2021-05-13 Martin Jambor <mjambor@suse.cz>
16058 PR tree-optimization/100453
16059 * tree-sra.c (sra_modify_assign): All const base accesses do not
16060 need refreshing, not just those from decl_pool.
16061 (sra_modify_assign): Do not refresh into a const base decl.
16063 2021-05-15 Jakub Jelinek <jakub@redhat.com>
16065 PR rtl-optimization/100342
16066 * regcprop.c (copy_value): When copying a source reg in a wider
16067 mode than it has recorded for the value, adjust recorded destination
16068 mode too or punt if !REG_CAN_CHANGE_MODE_P.
16070 2021-05-14 Jason Merrill <jason@redhat.com>
16072 * intl.h: Add comments.
16074 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16076 * config/aarch64/aarch64-simd.md
16077 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Split into...
16078 (aarch64_sqdmlsl2_lane<mode>_internal): ... This...
16079 (aarch64_sqdmlal2_lane<mode>_internal): ... And this.
16080 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Split into ...
16081 (aarch64_sqdmlsl2_laneq<mode>_internal): ... This...
16082 (aarch64_sqdmlal2_laneq<mode>_internal): ... And this.
16083 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>_internal): Split into...
16084 (aarch64_sqdmlsl2_n<mode>_internal): ... This...
16085 (aarch64_sqdmlal2_n<mode>_internal): ... And this.
16087 2021-05-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16090 * config/arm/arm_neon.h (vtst_s8): Replace call to vtst builtin with it's
16091 boolean logic equivalent.
16092 (vtst_s16): Likewise.
16093 (vtst_s32): Likewise.
16094 (vtst_u8): Likewise.
16095 (vtst_u16): Likewise.
16096 (vtst_u32): Likewise.
16097 (vtst_p8): Likewise.
16098 (vtst_p16): Likewise.
16099 (vtstq_s8): Likewise.
16100 (vtstq_s16): Likewise.
16101 (vtstq_s32): Likewise.
16102 (vtstq_u8): Likewise.
16103 (vtstq_u16): Likewise.
16104 (vtstq_u32): Likewise.
16105 (vtstq_p8): Likewise.
16106 (vtstq_p16): Likewise.
16107 * config/arm/arm_neon_builtins.def: Remove entry for vtst.
16108 * config/arm/neon.md (neon_vtst<mode>): Remove pattern.
16110 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16112 * config/aarch64/aarch64-simd.md (aarch64_sqdmlal2<mode>): Merge into...
16113 (aarch64_sqdml<SBINQOPS:as>l2<mode>): ... This.
16114 (aarch64_sqdmlsl2<mode>): Delete.
16115 (aarch64_sqdmlal2_lane<mode>): Merge this...
16116 (aarch64_sqdmlsl2_lane<mode>): ... And this...
16117 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>): ... Into this.
16118 (aarch64_sqdmlal2_laneq<mode>): Merge this...
16119 (aarch64_sqdmlsl2_laneq<mode>): ... And this...
16120 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>): ... Into this.
16121 (aarch64_sqdmlal2_n<mode>): Merge this...
16122 (aarch64_sqdmlsl2_n<mode>): ... And this...
16123 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>): ... Into this.
16125 2021-05-13 Martin Sebor <msebor@redhat.com>
16127 PR middle-end/100574
16128 * builtins.c (access_ref::get_ref): Improve detection of PHIs with
16129 all null arguments.
16131 2021-05-13 Martin Sebor <msebor@redhat.com>
16133 PR tree-optimization/93100
16134 PR middle-end/98583
16135 * tree-ssa-uninit.c (check_defs): Exclude intrinsic functions that
16136 don't modify referenced objects.
16138 2021-05-13 Martin Jambor <mjambor@suse.cz>
16140 PR tree-optimization/100453
16141 * tree-sra.c (sra_modify_assign): All const base accesses do not
16142 need refreshing, not just those from decl_pool.
16143 (sra_modify_assign): Do not refresh into a const base decl.
16145 2021-05-13 Martin Liska <mliska@suse.cz>
16147 * tree-ssa-dom.c: Remove m_simplifier.
16149 2021-05-13 Richard Earnshaw <rearnsha@arm.com>
16152 * config/arm/arm.c (arm_canonicalize_comparison): Correctly
16153 canonicalize DImode inequality comparisons against the
16154 maximum integral value.
16156 2021-05-13 Jakub Jelinek <jakub@redhat.com>
16158 PR tree-optimization/98856
16159 * config/i386/i386.c (ix86_shift_rotate_cost): Add CODE argument.
16160 Expect V2DI and V4DI arithmetic right shifts to be emulated.
16161 (ix86_rtx_costs, ix86_add_stmt_cost): Adjust ix86_shift_rotate_cost
16163 * config/i386/i386-expand.c (expand_vec_perm_2perm_interleave,
16164 expand_vec_perm_2perm_pblendv): New functions.
16165 (ix86_expand_vec_perm_const_1): Use them.
16166 * config/i386/sse.md (ashr<mode>3<mask_name>): Rename to ...
16167 (<mask_codefor>ashr<mode>3<mask_name>): ... this.
16168 (ashr<mode>3): New define_expand with VI248_AVX512BW iterator.
16169 (ashrv4di3): New define_expand.
16170 (ashrv2di3): Change condition to TARGET_SSE2, handle !TARGET_XOP
16171 and !TARGET_AVX512VL expansion.
16173 2021-05-13 Uroš Bizjak <ubizjak@gmail.com>
16176 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Force mode
16177 sizes < 16 to a register when constructing vpcmov pattern.
16178 * config/i386/mmx.md (*xop_pcmov_<mode>): Use MMXMODE124 mode.
16180 2021-05-13 Martin Liska <mliska@suse.cz>
16182 * gcov-io.c (gcov_write_block): Remove.
16183 (gcov_write_words): Likewise.
16184 (gcov_read_words): Re-implement using gcov_read_bytes.
16185 (gcov_allocate): Remove.
16186 (GCOV_BLOCK_SIZE): Likewise.
16187 (struct gcov_var): Remove most of the fields.
16188 (gcov_position): Implement with ftell.
16189 (gcov_rewrite): Remove setting of start and offset fields.
16190 (from_file): Re-format.
16191 (gcov_open): Remove setbuf call. It should not be needed.
16192 (gcov_close): Remove internal buffer handling.
16193 (gcov_magic): Use __builtin_bswap32.
16194 (gcov_write_counter): Use directly gcov_write_unsigned.
16195 (gcov_write_string): Use direct fwrite and do not round
16197 (gcov_seek): Use directly fseek.
16198 (gcov_write_tag): Use gcov_write_unsigned directly.
16199 (gcov_write_length): Likewise.
16200 (gcov_write_tag_length): Likewise.
16201 (gcov_read_bytes): Use directly fread.
16202 (gcov_read_unsigned): Use gcov_read_words.
16203 (gcov_read_counter): Likewise.
16204 (gcov_read_string): Use gcov_read_bytes.
16205 * gcov-io.h (GCOV_WORD_SIZE): Adjust to reflect
16206 that size is not in bytes, but words (4B).
16207 (GCOV_TAG_FUNCTION_LENGTH): Likewise.
16208 (GCOV_TAG_ARCS_LENGTH): Likewise.
16209 (GCOV_TAG_ARCS_NUM): Likewise.
16210 (GCOV_TAG_COUNTER_LENGTH): Likewise.
16211 (GCOV_TAG_COUNTER_NUM): Likewise.
16212 (GCOV_TAG_SUMMARY_LENGTH): Likewise.
16214 2021-05-13 liuhongt <hongtao.liu@intel.com>
16217 * config/i386/sse.md (ssedoublevecmode): Add attribute for
16218 V64QI/V32HI/V16SI/V4DI.
16219 (ssehalfvecmode): Add attribute for V2DI/V2DF.
16220 (*vec_concatv4si_0): Extend to VI124_128.
16221 (*vec_concat<mode>_0): New pre-reload splitter.
16222 * config/i386/predicates.md (movq_parallel): New predicate.
16224 2021-05-13 Alexandre Oliva <oliva@adacore.com>
16226 * targhooks.c (default_zero_call_used_regs): Retry using
16227 successfully-zeroed registers as sources.
16229 2021-05-12 Tobias Burnus <tobias@codesourcery.com>
16231 * omp-low.c (finish_taskreg_scan): Use the proper detach decl.
16233 2021-05-12 Aldy Hernandez <aldyh@redhat.com>
16236 * gimple-range.cc (range_of_builtin_call): Skip out on
16237 processing __builtin_clz when varying.
16239 2021-05-12 Tom de Vries <tdevries@suse.de>
16242 * config/nvptx/nvptx-opts.h (enum ptx_version): New enum.
16243 * config/nvptx/nvptx.c (nvptx_file_start): Print .version according
16244 to ptx_version_option.
16245 * config/nvptx/nvptx.h (TARGET_PTX_6_3): Define.
16246 * config/nvptx/nvptx.md (define_insn "nvptx_shuffle<mode>")
16247 (define_insn "nvptx_vote_ballot"): Use sync variant for
16249 * config/nvptx/nvptx.opt (ptx_version): Add enum.
16250 (mptx): Add option.
16251 * doc/invoke.texi (Nvidia PTX Options): Add mptx item.
16253 2021-05-12 Richard Biener <rguenther@suse.de>
16255 PR tree-optimization/100566
16256 * tree-ssa-sccvn.c (dominated_by_p_w_unex): Properly handle
16257 allow_back for all edge queries.
16259 2021-05-12 liuhongt <hongtao.liu@intel.com>
16262 * config/i386/sse.md (<sse4_1_avx2>_pblendvb): Add
16263 splitters for pblendvb of NOT mask register.
16265 2021-05-12 Richard Biener <rguenther@suse.de>
16267 PR tree-optimization/100519
16268 * tree-ssa-reassoc.c (can_associate_p): Split into...
16269 (can_associate_op_p): ... this
16270 (can_associate_type_p): ... and this.
16271 (is_reassociable_op): Call can_associate_op_p.
16272 (break_up_subtract_bb): Call the appropriate predicates.
16273 (reassociate_bb): Likewise.
16275 2021-05-12 Martin Liska <mliska@suse.cz>
16277 * lto-wrapper.c (merge_and_complain): Merge -flto=arg options.
16278 (run_gcc): Use -flto argument detection for merged
16281 2021-05-12 Martin Liska <mliska@suse.cz>
16283 * lto-wrapper.c (print_lto_docs_link): New function.
16284 (run_gcc): Print warning about missing job server detection
16285 after we know NR of partitions. Do the same for -flto{,=1}.
16286 * opts.c (get_option_html_page): Support -flto option.
16288 2021-05-12 Martin Liska <mliska@suse.cz>
16290 * lto-wrapper.c (get_options_from_collect_gcc_options): Change
16292 (append_option): Remove.
16293 (find_option): Rework to use the vector type.
16294 (remove_option): Remove.
16295 (merge_and_complain): Use vectors for cl_decoded_option data
16297 (append_compiler_options): Likewise.
16298 (append_diag_options): Likewise.
16299 (append_linker_options): Likewise.
16300 (append_offload_options): Likewise.
16301 (compile_offload_image): Likewise.
16302 (compile_images_for_offload_targets): Likewise.
16303 (find_and_merge_options): Likewise.
16304 (run_gcc): Likewise.
16306 2021-05-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
16309 * dwarf2out.c (dwarf2out_finish): Set
16310 have_multiple_function_sections with multi-range text_section.
16312 2021-05-12 Martin Liska <mliska@suse.cz>
16314 PR bootstrap/100560
16315 * Makefile.in: Remove version.h from linker command line.
16317 2021-05-12 Richard Biener <rguenther@suse.de>
16319 PR middle-end/100547
16320 * rtl.h (rtvec_alloc): Make argument size_t.
16321 * rtl.c (rtvec_alloc): Verify the count is less than INT_MAX.
16323 2021-05-12 Jakub Jelinek <jakub@redhat.com>
16325 PR middle-end/100508
16326 * cfgexpand.c (expand_debug_expr): For DEBUG_EXPR_DECL with vector
16327 type, don't reuse DECL_RTL if it has different mode, instead force
16328 creation of a new DEBUG_EXPR.
16330 2021-05-12 Jakub Jelinek <jakub@redhat.com>
16331 Marc Glisse <marc.glisse@inria.fr>
16333 PR tree-optimization/94589
16334 * match.pd ((X & Y) == X -> (X & ~Y) == 0,
16335 (X | Y) == Y -> (X & ~Y) == 0): New GIMPLE simplifications.
16337 2021-05-12 Uroš Bizjak <ubizjak@gmail.com>
16340 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Handle V2SF mode.
16341 * config/i386/mmx.md (MMXMODE124): New mode iterator.
16343 (mmxintvecmode): New mode attribute.
16344 (mmxintvecmodelower): Ditto.
16345 (*mmx_maskcmpv2sf3_comm): New insn pattern.
16346 (*mmx_maskcmpv2sf3): Ditto.
16347 (vec_cmpv2sfv2si): New expander.
16348 (vcond<V2FI:mode>v2si): Ditto.
16349 (mmx_vlendvps): New insn pattern.
16350 (vcond<MMXMODE124:mode><MMXMODEI:mode>): Also handle V2SFmode.
16351 (vcondu<MMXMODE124:mode><MMXMODEI:mode>): Ditto.
16352 (vcond_mask_<mode><mmxintvecmodelower>): Ditto.
16354 2021-05-11 Martin Sebor <msebor@redhat.com>
16356 PR middle-end/21433
16357 * expr.c (expand_expr_real_1): Replace unreachable code with an assert.
16359 2021-05-11 Richard Biener <rguenther@suse.de>
16361 * gimple-fold.c (gimple_fold_call): Do not call
16362 maybe_fold_reference on call arguments or the static chain.
16363 (fold_stmt_1): Do not call maybe_fold_reference on GIMPLE_ASM
16366 2021-05-11 Martin Liska <mliska@suse.cz>
16368 * builtins.def (DEF_HSAIL_BUILTIN): Remove.
16369 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
16370 (DEF_HSAIL_SAT_BUILTIN): Likewise.
16371 (DEF_HSAIL_INTR_BUILTIN): Likewise.
16372 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
16373 * doc/frontends.texi: Remove BRIG.
16374 * doc/install.texi: Likewise.
16375 * doc/invoke.texi: Likewise.
16376 * doc/standards.texi: Likewise.
16377 * brig-builtins.def: Removed.
16378 * brig/ChangeLog: Removed.
16379 * brig/Make-lang.in: Removed.
16380 * brig/brig-builtins.h: Removed.
16381 * brig/brig-c.h: Removed.
16382 * brig/brig-lang.c: Removed.
16383 * brig/brigfrontend/brig-arg-block-handler.cc: Removed.
16384 * brig/brigfrontend/brig-atomic-inst-handler.cc: Removed.
16385 * brig/brigfrontend/brig-basic-inst-handler.cc: Removed.
16386 * brig/brigfrontend/brig-branch-inst-handler.cc: Removed.
16387 * brig/brigfrontend/brig-cmp-inst-handler.cc: Removed.
16388 * brig/brigfrontend/brig-code-entry-handler.cc: Removed.
16389 * brig/brigfrontend/brig-code-entry-handler.h: Removed.
16390 * brig/brigfrontend/brig-comment-handler.cc: Removed.
16391 * brig/brigfrontend/brig-control-handler.cc: Removed.
16392 * brig/brigfrontend/brig-copy-move-inst-handler.cc: Removed.
16393 * brig/brigfrontend/brig-cvt-inst-handler.cc: Removed.
16394 * brig/brigfrontend/brig-fbarrier-handler.cc: Removed.
16395 * brig/brigfrontend/brig-function-handler.cc: Removed.
16396 * brig/brigfrontend/brig-function.cc: Removed.
16397 * brig/brigfrontend/brig-function.h: Removed.
16398 * brig/brigfrontend/brig-inst-mod-handler.cc: Removed.
16399 * brig/brigfrontend/brig-label-handler.cc: Removed.
16400 * brig/brigfrontend/brig-lane-inst-handler.cc: Removed.
16401 * brig/brigfrontend/brig-machine.c: Removed.
16402 * brig/brigfrontend/brig-machine.h: Removed.
16403 * brig/brigfrontend/brig-mem-inst-handler.cc: Removed.
16404 * brig/brigfrontend/brig-module-handler.cc: Removed.
16405 * brig/brigfrontend/brig-queue-inst-handler.cc: Removed.
16406 * brig/brigfrontend/brig-seg-inst-handler.cc: Removed.
16407 * brig/brigfrontend/brig-signal-inst-handler.cc: Removed.
16408 * brig/brigfrontend/brig-to-generic.cc: Removed.
16409 * brig/brigfrontend/brig-to-generic.h: Removed.
16410 * brig/brigfrontend/brig-util.cc: Removed.
16411 * brig/brigfrontend/brig-util.h: Removed.
16412 * brig/brigfrontend/brig-variable-handler.cc: Removed.
16413 * brig/brigfrontend/hsa-brig-format.h: Removed.
16414 * brig/brigfrontend/phsa.h: Removed.
16415 * brig/brigspec.c: Removed.
16416 * brig/config-lang.in: Removed.
16417 * brig/gccbrig.texi: Removed.
16418 * brig/lang-specs.h: Removed.
16419 * brig/lang.opt: Removed.
16421 2021-05-11 Richard Biener <rguenther@suse.de>
16424 * ipa-param-manipulation.c
16425 (ipa_param_body_adjustments::modify_call_stmt): Avoid
16426 altering SSA_NAME_DEF_STMT by adjusting the calls LHS
16427 via gimple_call_lhs_ptr.
16429 2021-05-11 Alex Coplan <alex.coplan@arm.com>
16432 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear):
16433 Avoid emitting CFA adjusts on the sp if we have the fp.
16435 2021-05-11 Richard Sandiford <richard.sandiford@arm.com>
16437 * config/aarch64/iterators.md (VMUL_CHANGE_NLANES): Delete.
16438 (VMULD): New iterator.
16439 (VCOND): Handle V4HF and V8HF.
16440 (VCONQ): Fix entry for V2SF.
16441 * config/aarch64/aarch64-simd.md (mul_lane<mode>3): Use VMULD
16442 instead of VMUL. Use a 64-bit vector mode for the indexed operand.
16443 (*aarch64_mul3_elt_<vswap_width_name><mode>): Merge with...
16444 (mul_laneq<mode>3): ...this define_insn. Use VMUL instead of VDQSF.
16445 Use a 128-bit vector mode for the indexed operand. Use stype for
16446 the scheduling type.
16448 2021-05-11 Richard Biener <rguenther@suse.de>
16450 * gimple-fold.c (maybe_fold_reference): Only return
16451 is_gimple_min_invariant values.
16453 2021-05-11 Richard Biener <rguenther@suse.de>
16455 PR middle-end/100509
16456 * gimple-fold.c (fold_gimple_assign): Only call
16457 get_symbol_constant_value on register type symbols.
16459 2021-05-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
16460 Joe Ramsay <joe.ramsay@arm.com>
16463 * config/arm/arm_mve.h (__arm_vstrwq_scatter_offset): Fix wrong arguments.
16464 (__arm_vcmpneq): Remove duplicate definition.
16465 (__arm_vstrwq_scatter_offset_p): Likewise.
16466 (__arm_vmaxq_x): Likewise.
16467 (__arm_vmlsdavaq): Likewise.
16468 (__arm_vmlsdavaxq): Likewise.
16469 (__arm_vmlsdavq_p): Likewise.
16470 (__arm_vmlsdavxq_p): Likewise.
16471 (__arm_vrmlaldavhaq): Likewise.
16472 (__arm_vstrbq_p): Likewise.
16473 (__arm_vstrbq_scatter_offset): Likewise.
16474 (__arm_vstrbq_scatter_offset_p): Likewise.
16475 (__arm_vstrdq_scatter_offset): Likewise.
16476 (__arm_vstrdq_scatter_offset_p): Likewise.
16477 (__arm_vstrdq_scatter_shifted_offset): Likewise.
16478 (__arm_vstrdq_scatter_shifted_offset_p): Likewise.
16480 2021-05-11 Jakub Jelinek <jakub@redhat.com>
16482 PR middle-end/100471
16483 * omp-low.c (lower_omp_task_reductions): For OMP_TASKLOOP, if data
16484 is 0, bypass the reduction loop including
16485 GOMP_taskgroup_reduction_unregister call.
16487 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
16489 * config/rs6000/rs6000.c (struct rs6000_cost_data): New member
16490 costing_for_scalar.
16491 (rs6000_density_test): Early return if costing_for_scalar is true.
16492 (rs6000_init_cost): Init costing_for_scalar of rs6000_cost_data.
16494 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
16496 * doc/tm.texi: Regenerated.
16497 * target.def (init_cost): Add new parameter costing_for_scalar.
16498 * targhooks.c (default_init_cost): Adjust for new parameter.
16499 * targhooks.h (default_init_cost): Likewise.
16500 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
16501 (vect_compute_single_scalar_iteration_cost): Likewise.
16502 (vect_analyze_loop_2): Likewise.
16503 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
16504 (vect_bb_vectorization_profitable_p): Likewise.
16505 * tree-vectorizer.h (init_cost): Likewise.
16506 * config/aarch64/aarch64.c (aarch64_init_cost): Likewise.
16507 * config/i386/i386.c (ix86_init_cost): Likewise.
16508 * config/rs6000/rs6000.c (rs6000_init_cost): Likewise.
16510 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
16512 * config/rs6000/rs6000.c (rs6000_vect_nonmem): Renamed to
16513 vect_nonmem and moved into...
16514 (struct rs6000_cost_data): ...here.
16515 (rs6000_init_cost): Use vect_nonmem of cost_data instead.
16516 (rs6000_add_stmt_cost): Likewise.
16517 (rs6000_finish_cost): Likewise.
16519 2021-05-10 Eric Botcazou <ebotcazou@adacore.com>
16521 * range-op.cc (get_bool_state): Adjust head comment.
16522 (operator_not_equal::op1_range): Fix comment.
16523 (operator_bitwise_xor::op1_range): Remove call to gcc_unreachable.
16525 2021-05-10 Martin Sebor <msebor@redhat.com>
16527 PR middle-end/100425
16528 PR middle-end/100510
16529 * gimple-ssa-warn-alloca.c (pass_walloca::firast_time_p): Rename...
16530 (pass_walloca::xlimit_certain_p): ...to this.
16531 (pass_walloca::gate): Execute for any kind of handled warning.
16532 (pass_walloca::execute): Avoid issuing "maybe" and "unbounded"
16533 warnings when xlimit_certain_p is set.
16535 2021-05-10 Pat Haugen <pthaugen@linux.ibm.com>
16537 * config/rs6000/rs6000.c (rs6000_ira_change_pseudo_allocno_class):
16538 Return ALTIVEC_REGS if that is best_class.
16539 (rs6000_compute_pressure_classes): Add ALTIVEC_REGS.
16541 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16543 * config/arm/arm.h (CPP_SPEC): Remove error message about
16546 2021-05-10 Martin Jambor <mjambor@suse.cz>
16548 * ipa-prop.h (IPA_NODE_REF): Removed.
16549 (IPA_NODE_REF_GET_CREATE): Likewise.
16550 (IPA_EDGE_REF): Likewise.
16551 (IPA_EDGE_REF_GET_CREATE): Likewise.
16552 (IS_VALID_JUMP_FUNC_INDEX): Likewise.
16553 * ipa-cp.c (print_all_lattices): Replaced IPA_NODE_REF with a direct
16554 use of ipa_node_params_sum.
16555 (ipcp_versionable_function_p): Likewise.
16556 (push_node_to_stack): Likewise.
16557 (pop_node_from_stack): Likewise.
16558 (set_single_call_flag): Replaced two IPA_NODE_REF with one single
16559 direct use of ipa_node_params_sum.
16560 (initialize_node_lattices): Replaced IPA_NODE_REF with a direct use of
16561 ipa_node_params_sum.
16562 (ipa_context_from_jfunc): Replaced IPA_EDGE_REF with a direct use of
16564 (ipcp_verify_propagated_values): Replaced IPA_NODE_REF with a direct
16565 use of ipa_node_params_sum.
16566 (self_recursively_generated_p): Likewise.
16567 (propagate_scalar_across_jump_function): Likewise.
16568 (propagate_context_across_jump_function): Replaced IPA_EDGE_REF with a
16569 direct use of ipa_edge_args_sum, moved the lookup after the early
16570 exit. Replaced IPA_NODE_REF with a direct use of ipa_node_params_sum.
16571 (propagate_bits_across_jump_function): Replaced IPA_NODE_REF with
16572 direct uses of ipa_node_params_sum.
16573 (propagate_vr_across_jump_function): Likewise.
16574 (propagate_aggregate_lattice): Likewise.
16575 (propagate_aggs_across_jump_function): Likewise.
16576 (propagate_constants_across_call): Likewise, also replaced
16577 IPA_EDGE_REF with a direct use of ipa_edge_args_sum.
16578 (good_cloning_opportunity_p): Replaced IPA_NODE_REF with a direct use
16579 of ipa_node_params_sum.
16580 (estimate_local_effects): Likewise.
16581 (add_all_node_vals_to_toposort): Likewise.
16582 (propagate_constants_topo): Likewise.
16583 (ipcp_propagate_stage): Likewise.
16584 (ipcp_discover_new_direct_edges): Likewise.
16585 (calls_same_node_or_its_all_contexts_clone_p): Likewise.
16586 (cgraph_edge_brings_value_p): Likewise (in both overloaded functions).
16587 (get_info_about_necessary_edges): Likewise.
16588 (want_remove_some_param_p): Likewise.
16589 (create_specialized_node): Likewise.
16590 (self_recursive_pass_through_p): Likewise.
16591 (self_recursive_agg_pass_through_p): Likewise.
16592 (find_more_scalar_values_for_callers_subset): Likewise and also
16593 replaced IPA_EDGE_REF with direct uses of ipa_edge_args_sum, in one
16594 case replacing two of those with a single query.
16595 (find_more_contexts_for_caller_subset): Likewise for the
16596 ipa_polymorphic_call_context overload.
16597 (intersect_aggregates_with_edge): Replaced IPA_EDGE_REF with a direct
16598 use of ipa_edge_args_sum. Replaced IPA_NODE_REF with direct uses of
16599 ipa_node_params_sum.
16600 (find_aggregate_values_for_callers_subset): Likewise, also reusing
16601 results of ipa_edge_args_sum->get.
16602 (cgraph_edge_brings_all_scalars_for_node): Replaced IPA_NODE_REF with
16603 direct uses of ipa_node_params_sum, replaced IPA_EDGE_REF with a
16604 direct use of ipa_edge_args_sum.
16605 (cgraph_edge_brings_all_agg_vals_for_node): Likewise, moved node
16606 summary query after the early exit and reused the result later.
16607 (decide_about_value): Replaced IPA_NODE_REF with a direct use of
16608 ipa_node_params_sum.
16609 (decide_whether_version_node): Likewise. Removed re-querying for
16610 summaries after cloning.
16611 (spread_undeadness): Replaced IPA_NODE_REF with a direct use of
16612 ipa_node_params_sum.
16613 (has_undead_caller_from_outside_scc_p): Likewise, reusing results of
16615 (identify_dead_nodes): Likewise.
16616 (ipcp_store_bits_results): Replaced IPA_NODE_REF with direct uses of
16617 ipa_node_params_sum.
16618 (ipcp_store_vr_results): Likewise.
16619 * ipa-fnsummary.c (evaluate_properties_for_edge): Likewise.
16620 (ipa_fn_summary_t::duplicate): Likewise.
16621 (analyze_function_body): Likewise.
16622 (estimate_calls_size_and_time): Likewise.
16623 (ipa_cached_call_context::duplicate_from): Likewise.
16624 (ipa_call_context::equal_to): Likewise.
16625 (remap_edge_params): Likewise.
16626 (ipa_merge_fn_summary_after_inlining): Likewise.
16627 (inline_read_section): Likewise.
16628 * ipa-icf.c (sem_function::param_used_p): Likewise.
16629 * ipa-modref.c (compute_parm_map): Likewise.
16630 (compute_parm_map): Replaced IPA_EDGE_REF with a direct use of
16632 (get_access_for_fnspec): Replaced IPA_NODE_REF with a direct use of
16633 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
16635 * ipa-profile.c (check_argument_count): Likewise.
16636 * ipa-prop.c (ipa_alloc_node_params): Replaced IPA_NODE_REF_GET_CREATE
16637 with a direct use of ipa_node_params_sum.
16638 (ipa_initialize_node_params): Likewise.
16639 (ipa_print_node_jump_functions_for_edge): Replaced IPA_EDGE_REF with a
16640 direct use of ipa_edge_args_sum and reused the query result.
16641 (ipa_compute_jump_functions_for_edge): Replaced IPA_NODE_REF with a
16642 direct use of ipa_node_params_sum and replaced IPA_EDGE_REF with a
16643 direct use of ipa_edge_args_sum.
16644 (ipa_note_param_call): Replaced IPA_NODE_REF with a direct use of
16645 ipa_node_params_sum and reused the result of the query.
16646 (ipa_analyze_node): Likewise.
16647 (ipa_analyze_controlled_uses): Replaced IPA_NODE_REF with a direct use
16648 of ipa_node_params_sum.
16649 (update_jump_functions_after_inlining): Replaced IPA_EDGE_REF with
16650 direct uses of ipa_edge_args_sum.
16651 (update_indirect_edges_after_inlining): Replaced IPA_NODE_REF with
16652 direct uses of ipa_node_params_sum and replaced IPA_EDGE_REF with a
16653 direct use of ipa_edge_args_sum. Removed superficial re-querying the
16655 (propagate_controlled_uses): Replaced IPA_NODE_REF with direct uses of
16656 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
16658 (ipa_propagate_indirect_call_infos): Replaced IPA_EDGE_REF with a
16659 direct use of ipa_edge_args_sum.
16660 (ipa_edge_args_sum_t::duplicate): Replaced IPA_NODE_REF with a direct
16661 use of ipa_node_params_sum.
16662 (ipa_print_node_params): Likewise.
16663 (ipa_write_node_info): Likewise and also replaced IPA_EDGE_REF with
16664 direct uses of ipa_edge_args_sum.
16665 (ipa_read_edge_info): Replaced IPA_EDGE_REF with a direct use of
16667 (ipa_read_node_info): Replaced IPA_NODE_REF with a direct use of
16668 ipa_node_params_sum.
16669 (ipa_prop_write_jump_functions): Likewise. Move variable node to the
16670 scopes where it is used.
16672 2021-05-10 Uroš Bizjak <ubizjak@gmail.com>
16674 * config/i386/i386-expand.c (ix86_expand_sse_movcc)
16675 <case E_V2SImode>: Force op_true to register.
16677 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16679 * config/arm/iterators.md (MVE_FP_COMPARISONS): New.
16680 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_f<mode>)
16681 (mve_vcmp<mve_cmp_op>q_n_f<mode>): New, merge all vcmp_*f*
16683 (mve_vcmpeqq_f<mode>, mve_vcmpeqq_n_f<mode>, mve_vcmpgeq_f<mode>)
16684 (mve_vcmpgeq_n_f<mode>, mve_vcmpgtq_f<mode>)
16685 (mve_vcmpgtq_n_f<mode>, mve_vcmpleq_f<mode>)
16686 (mve_vcmpleq_n_f<mode>, mve_vcmpltq_f<mode>)
16687 (mve_vcmpltq_n_f<mode>, mve_vcmpneq_f<mode>)
16688 (mve_vcmpneq_n_f<mode>): Remove.
16689 * config/arm/unspecs.md (VCMPEQQ_F, VCMPEQQ_N_F, VCMPGEQ_F)
16690 (VCMPGEQ_N_F, VCMPGTQ_F, VCMPGTQ_N_F, VCMPLEQ_F, VCMPLEQ_N_F)
16691 (VCMPLTQ_F, VCMPLTQ_N_F, VCMPNEQ_F, VCMPNEQ_N_F): Remove.
16693 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16695 * config/arm/iterators.md (MVE_COMPARISONS): New.
16697 (mve_cmp_type): New.
16698 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_<mode>): New, merge all
16700 (mve_vcmpneq_<mode>, mve_vcmpcsq_n_<mode>, mve_vcmpcsq_<mode>)
16701 (mve_vcmpeqq_n_<mode>, mve_vcmpeqq_<mode>, mve_vcmpgeq_n_<mode>)
16702 (mve_vcmpgeq_<mode>, mve_vcmpgtq_n_<mode>, mve_vcmpgtq_<mode>)
16703 (mve_vcmphiq_n_<mode>, mve_vcmphiq_<mode>, mve_vcmpleq_n_<mode>)
16704 (mve_vcmpleq_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
16705 (mve_vcmpneq_n_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
16706 (mve_vcmpneq_n_<mode>): Remove.
16708 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16710 * config/arm/arm_mve.h (__arm_vcmp*): Remove 's' suffix.
16711 * config/arm/arm_mve_builtins.def (vcmp*): Remove 's' suffix.
16712 * config/arm/mve.md (mve_vcmp*): Remove 's' suffix in pattern
16715 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16717 * config/arm/arm_mve_builtins.def (vcmpneq_u): Remove.
16718 (vcmpneq_n_u): Likewise.
16719 (vcmpeqq_u,): Likewise.
16720 (vcmpeqq_n_u): Likewise.
16721 * config/arm/iterators.md (supf): Remove VCMPNEQ_U, VCMPEQQ_U,
16722 VCMPEQQ_N_U and VCMPNEQ_N_U.
16723 * config/arm/mve.md (mve_vcmpneq): Remove <supf> iteration.
16724 (mve_vcmpeqq_n): Likewise.
16725 (mve_vcmpeqq): Likewise.
16726 (mve_vcmpneq_n): Likewise.
16728 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16730 * config/arm/arm_mve.h (__arm_vcmpeq*u*, __arm_vcmpne*u*): Call
16731 the 's' version of the builtin.
16733 2021-05-10 Richard Biener <rguenther@suse.de>
16735 PR tree-optimization/100492
16736 * tree-loop-distribution.c (find_seed_stmts_for_distribution):
16737 Find nothing when the loop contains an irreducible region.
16739 2021-05-10 Richard Biener <rguenther@suse.de>
16741 PR middle-end/100464
16743 * gimple-fold.c (canonicalize_constructor_val): Do not set
16746 2021-05-10 Richard Biener <rguenther@suse.de>
16748 PR tree-optimization/100434
16749 * tree-ssa-dse.c (initialize_ao_ref_for_dse): Handle
16751 (dse_optimize_stmt): Handle call LHS by dropping the
16752 LHS or the whole call if it doesn't have other
16754 (pass_dse::execute): Adjust.
16756 2021-05-10 Martin Liska <mliska@suse.cz>
16758 * Makefile.in: Add missing genversion rule.
16760 2021-05-10 Alex Coplan <alex.coplan@arm.com>
16763 * config/arm/mve.md (*mve_mov<mode>): Simplify output code. Use
16764 vldrw.u32 and vstrw.32 for V2D[IF]mode loads and stores.
16766 2021-05-10 Martin Liska <mliska@suse.cz>
16768 * builtins.c (is_builtin_name): Use startswith
16769 function instead of strncmp.
16770 * collect2.c (main): Likewise.
16771 (has_lto_section): Likewise.
16772 (scan_libraries): Likewise.
16773 * coverage.c (coverage_checksum_string): Likewise.
16774 (coverage_init): Likewise.
16775 * dwarf2out.c (is_cxx): Likewise.
16776 (gen_compile_unit_die): Likewise.
16777 * gcc-ar.c (main): Likewise.
16778 * gcc.c (init_spec): Likewise.
16779 (read_specs): Likewise.
16780 (execute): Likewise.
16781 (check_live_switch): Likewise.
16782 * genattrtab.c (write_attr_case): Likewise.
16783 (IS_ATTR_GROUP): Likewise.
16784 * gencfn-macros.c (main): Likewise.
16785 * gengtype.c (type_for_name): Likewise.
16786 (gen_rtx_next): Likewise.
16787 (get_file_langdir): Likewise.
16788 (write_local): Likewise.
16789 * genmatch.c (get_operator): Likewise.
16790 (get_operand_type): Likewise.
16791 (expr::gen_transform): Likewise.
16792 * genoutput.c (validate_optab_operands): Likewise.
16793 * incpath.c (add_sysroot_to_chain): Likewise.
16794 * langhooks.c (lang_GNU_C): Likewise.
16795 (lang_GNU_CXX): Likewise.
16796 (lang_GNU_Fortran): Likewise.
16797 (lang_GNU_OBJC): Likewise.
16798 * lto-wrapper.c (run_gcc): Likewise.
16799 * omp-general.c (omp_max_simt_vf): Likewise.
16800 * omp-low.c (omp_runtime_api_call): Likewise.
16801 * opts-common.c (parse_options_from_collect_gcc_options): Likewise.
16802 * read-rtl-function.c (function_reader::read_rtx_operand_r): Likewise.
16803 * real.c (real_from_string): Likewise.
16804 * selftest.c (assert_str_startswith): Likewise.
16805 * timevar.c (timer::validate_phases): Likewise.
16806 * tree.c (get_file_function_name): Likewise.
16807 * ubsan.c (ubsan_use_new_style_p): Likewise.
16808 * varasm.c (default_function_rodata_section): Likewise.
16809 (incorporeal_function_p): Likewise.
16810 (default_section_type_flags): Likewise.
16811 * system.h (startswith): Define startswith.
16813 2021-05-10 Martin Liska <mliska@suse.cz>
16815 * bitmap.h (class auto_bitmap): Remove
16816 __cplusplus >= 201103.
16817 * config/aarch64/aarch64.c: Likewise.
16818 * gimple-ssa-store-merging.c (store_immediate_info::store_immediate_info):
16820 * sbitmap.h: Likewise.
16822 2021-05-10 Martin Liska <mliska@suse.cz>
16824 * Makefile.in: Rename gcov-iov to genversion and depend
16825 on version.h (instead of gcov-iov.h).
16826 * gcov-io.h: Include version.h instread of gcov-iov.h.
16827 * gengtype-state.c (read_state_version): Likewise.
16828 * gcov-iov.c: Moved to...
16829 * genversion.c: ...here.
16830 * lto-streamer.h (LTO_major_version): Define it with
16832 * version.c: Removed.
16833 * version.h: Removed.
16835 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16837 * config/arc/arc.md (UNSPEC_ARC_DMPYWH): Define.
16838 * config/arc/simdext.md (VCT): Add predicates for iterator
16841 (voptab): Likewise.
16842 (vec_widen_<V_US>mult_hi_v4hi): Change pattern predicate.
16843 (<voptab>v2si3): New patterns.
16845 (reduc_plus_scal_v4hi): Likewise.
16846 (reduc_plus_scal_v2si): Likewise.
16847 (vec_duplicatev2si): Likewise.
16848 (vec_duplicatev4hi): Likewise.
16850 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16852 * config/arc/simdext.md: Format and cleanup file.
16854 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16856 * config/arc/simdext.md (movmisalignv2hi): Allow misaligned access
16857 only when munaligned-access option is on.
16858 (movmisalign<mode>): Likewise.
16860 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16862 * common/config/arc/arc-common.c (arc_handle_option): Remove dot
16864 * config/arc/arc.c (arc_reorg): Remove underscore from string.
16866 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16868 * config/arc/arc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
16869 (CTZ_DEFINED_VALUE_AT_ZERO): Likewise.
16870 * config/arc/arc.md (clrsbsi2): Cleanup pattern.
16871 (norm_f): Likewise.
16874 (clzsi2): Use fls instruction when available.
16875 (arc_clzsi2): Likewise.
16877 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16879 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add r26 and r27.
16881 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16883 * doc/extend.texi (__builtin_arc_sr): Swap arguments.
16885 2021-05-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
16887 PR middle-end/100467
16888 * toplev.c (compile_file): Call insn_locations_init before
16889 targetm.asm_out.code_end.
16891 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
16894 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
16896 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
16898 2021-05-07 Jakub Jelinek <jakub@redhat.com>
16899 Andrew Stubbs <amd@codesourcery.com>
16902 * builtins.c (try_store_by_multiple_pieces): Use force_operand for
16903 emit_move_insn operands.
16905 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
16907 * cfgexpand.c (expand_gimple_basic_block): Do not inherit a current
16908 location for the outgoing edges of an empty block.
16909 * dwarf2out.c (add_subscript_info): Retrieve the bounds and index
16910 type by means of the get_array_descr_info langhook, if it is set and
16911 returns true. Remove obsolete code dealing with unnamed subtypes.
16913 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16915 * gimple-range-cache.cc (ssa_block_ranges): Virtualize.
16916 (sbr_vector): Renamed from ssa_block_cache.
16917 (sbr_vector::sbr_vector): Allocate from obstack abd initialize.
16918 (ssa_block_ranges::~ssa_block_ranges): Remove.
16919 (sbr_vector::set_bb_range): Use varying and undefined cached values.
16920 (ssa_block_ranges::set_bb_varying): Remove.
16921 (sbr_vector::get_bb_range): Adjust assert.
16922 (sbr_vector::bb_range_p): Adjust assert.
16923 (~block_range_cache): No freeing loop required.
16924 (block_range_cache::get_block_ranges): Remove.
16925 (block_range_cache::set_bb_range): Inline get_block_ranges.
16926 (block_range_cache::set_bb_varying): Remove.
16927 * gimple-range-cache.h (set_bb_varying): Remove prototype.
16928 * value-range.h (irange_allocator::get_memory): New.
16930 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16932 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Search
16933 dominator tree is available and requested.
16934 (ranger_cache::ssa_range_in_bb): Don't search dom tree here.
16935 (ranger_cache::fill_block_cache): Don't search dom tree here either.
16936 * gimple-range-cache.h (non_null_deref_p): Add dom_search param.
16938 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16940 * gimple-range.cc (gimple_ranger::range_on_exit): Handle block with
16941 only PHI nodes better.
16943 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16945 * gimple-range-edge.h (gimple_outgoing_range): Rename from
16947 (gcond_edge_range): Export prototype.
16948 * gimple-range-edge.cc (gcond_edge_range): New.
16949 (gimple_outgoing_range::edge_range_p): Use gcond_edge_range.
16950 * gimple-range-gori.h (gori_compute): Use gimple_outgoing_range.
16952 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16954 * gimple-range-edge.cc (outgoing_range::calc_switch_ranges): Compute
16955 default range into a temp and allocate only what is needed.
16957 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16959 * range-op.cc (operator_trunc_mod::wi_fold): x % 0 is UNDEFINED.
16961 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16963 * gimple-range.h (gimple_range_global): Pick up parameter initial
16964 values, and use-before defined locals are UNDEFINED.
16966 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
16968 * doc/extend.texi (scalar_storage_order): Mention effect on pointer
16970 * tree.h (reverse_storage_order_for_component_p): Return false if
16971 the type is a pointer.
16973 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
16975 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
16977 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
16980 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
16981 Handle V8QI, V4HI and V2SI modes.
16982 * config/i386/mmx.md (mmx_pblendvb): New insn pattern.
16983 * config/i386/sse.md (unspec): Move UNSPEC_BLENDV ...
16984 * config/i386/i386.md (unspec): ... here.
16986 2021-05-07 Tobias Burnus <tobias@codesourcery.com>
16987 Tom de Vries <tdevries@suse.de>
16989 * omp-low.c (lower_rec_simd_input_clauses): Set max_vf = 1 if
16990 a truth_value_p reduction variable is nonintegral.
16992 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
16995 * config/i386/i386-expand.c (ix86_use_mask_cmp_p):
16996 Return false for mode sizes < 16.
16998 2021-05-07 Jakub Jelinek <jakub@redhat.com>
17001 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn.
17003 2021-05-06 Martin Jambor <mjambor@suse.cz>
17005 * ipa-sra.c (ipa_sra_dump_all_summaries): Dump edge summaries even
17006 when there is no function summary.
17007 (ipa_sra_summarize_function): produce edge summaries even when
17010 2021-05-06 Tom Tromey <tom@tromey.com>
17012 * godump.c (string_hash_eq): Remove.
17013 (go_finish): Use htab_eq_string.
17015 2021-05-06 Tom Tromey <tom@tromey.com>
17017 * gengtype-state.c (read_state): Use htab_eq_string.
17018 (string_eq): Remove.
17020 2021-05-06 Tom Tromey <tom@tromey.com>
17022 * gensupport.c (htab_eq_string): Remove.
17024 2021-05-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
17027 * debug.h (gcc_debug_hooks): Add set_ignored_loc function pointer.
17028 * dwarf2out.h (dw_fde_node::ignored_debug): New data item.
17029 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Add dummy
17030 set_ignored_loc callbacks.
17031 * debug.c (do_nothing_debug_hooks): Likewise.
17032 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
17033 * dwarf2out.c (text_section_used, cold_text_section_used): Remove.
17034 (in_text_section_p, last_text_label, last_cold_label,
17035 switch_text_ranges, switch_cold_ranges): New data items.
17036 (dwarf2out_note_section_used): Remove.
17037 (dwarf2out_begin_prologue): Set fde->ignored_debug and
17039 (mark_ignored_debug_section): New helper function.
17040 (dwarf2out_end_epilogue, dwarf2out_switch_text_section): Call
17041 mark_ignored_debug_section.
17042 (dwarf2_debug_hooks): Use dwarf2out_set_ignored_loc.
17043 (dwarf2_lineno_debug_hooks): Use dummy for set_ignored_loc.
17044 (size_of_aranges): Adjust formula for multi-part text ranges size.
17045 (output_aranges): Output multi-part text ranges.
17046 (dwarf2out_set_ignored_loc): New callback function.
17047 (dwarf2out_finish): Output multi-part text ranges.
17048 (dwarf2out_c_finalize): Clear new data items.
17049 * final.c (final_start_function_1): Call set_ignored_loc callback.
17050 (final_scan_insn_1): Likewise.
17051 * ggc-page.c (gt_ggc_mx): New helper function.
17052 * stringpool.c (gt_pch_nx): Likewise.
17054 2021-05-06 Richard Biener <rguenther@suse.de>
17056 * timevar.def (TV_TREE_INSERT_PHI_NODES): Remove.
17057 (TV_TREE_SSA_REWRITE_BLOCKS): Likewise.
17058 (TV_TREE_INTO_SSA): New.
17059 * tree-into-ssa.c (insert_phi_nodes): Do not account separately.
17060 (rewrite_blocks): Likewise.
17061 (pass_data_build_ssa): Account to TV_TREE_INTO_SSA.
17063 2021-05-06 Jakub Jelinek <jakub@redhat.com>
17065 * tree-ssa-phiopt.c (value_replacement, minmax_replacement,
17066 abs_replacement, xor_replacement,
17067 cond_removal_in_popcount_clz_ctz_pattern,
17068 replace_phi_edge_with_variable): Change type of phi argument from
17069 gimple * to gphi *.
17071 2021-05-06 Richard Biener <rguenther@suse.de>
17073 * tree-ssa-loop-split.c (split_loop): Delay updating SSA form.
17074 Output an opt-info message.
17075 (do_split_loop_on_cond): Likewise.
17076 (tree_ssa_split_loops): Update SSA form here.
17078 2021-05-06 Richard Biener <rguenther@suse.de>
17080 * tree-inline.c (tree_function_versioning): Fix DECL_BY_REFERENCE
17081 return variable removal.
17083 2021-05-06 Marius Hillenbrand <mhillen@linux.ibm.com>
17085 * config/s390/s390-builtins.def (O_M5, O1_M5, ...): Remove unused macros.
17086 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
17087 (s390_vec_permi_dbl, s390_vpdi): Use the O3_U2 type for the immediate
17089 * config/s390/s390.c (s390_const_operand_ok): Remove unused
17092 2021-05-06 Jakub Jelinek <jakub@redhat.com>
17094 PR tree-optimization/94589
17095 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Call
17096 spaceship_replacement.
17097 (cond_only_block_p, spaceship_replacement): New functions.
17099 2021-05-06 Richard Biener <rguenther@suse.de>
17102 * tree-emutls.c (gen_emutls_addr): Pass in whether we're
17103 dealing with a debug use and only query existing addresses
17105 (lower_emutls_1): Avoid splitting out addresses for debug
17106 stmts, reset the debug stmt when we fail to find existing
17108 (lower_emutls_phi_arg): Set wi.stmt.
17110 2021-05-06 Christoph Muellner <cmuellner@gcc.gnu.org>
17113 * config/riscv/riscv.c (riscv_block_move_loop): Use cbranch helper.
17114 * config/riscv/riscv.md (cbranch<mode>4): Generate helpers.
17115 (stack_protect_test): Use cbranch helper.
17117 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
17120 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
17121 always return the establisher frame for __builtin_frame_address (0).
17123 2021-05-05 Ivan Sorokin <vanyacpp@gmail.com>
17126 * config/i386/i386-builtins.c (ix86_cpu_model_type_node): New.
17127 (ix86_cpu_model_var): Likewise.
17128 (ix86_cpu_features2_type_node): Likewise.
17129 (ix86_cpu_features2_var): Likewise.
17130 (fold_builtin_cpu): Cache __cpu_model and __cpu_features2 with
17133 2021-05-05 Martin Sebor <msebor@redhat.com>
17135 * passes.def (pass_warn_printf): Run after SSA.
17137 2021-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17139 * config/arm/neon.md (neon_vtst_combine<mode>): New pattern.
17140 * config/arm/predicates.md (minus_one_operand): New predicate.
17142 2021-05-05 Jeff Law <jlaw@tachyum.com>
17144 * config/avr/avr.md: Remove references to CC_STATUS_INIT.
17146 2021-05-05 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
17148 PR rtl-optimization/100263
17149 * postreload.c (move2add_valid_value_p): Ensure register can
17152 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
17154 PR rtl-optimization/100411
17155 * cfgcleanup.c (try_crossjump_to_edge): Also skip end of prologue
17156 and beginning of function markers.
17158 2021-05-05 Jeff Law <jlaw@tachyum.com>
17160 * config/cr16/cr16.h (NOTICE_UPDATE_CC): Remove.
17161 * config/cr16/cr16.c (notice_update_cc): Remove.
17162 * config/cr16/cr16-protos.h (notice_update_cc): Remove.
17164 2021-05-05 Uroš Bizjak <ubizjak@gmail.com>
17167 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
17168 Handle V8QI, V4HI and V2SI modes.
17169 * config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
17170 (ix86_build_signbit_mask): Ditto.
17171 * config/i386/mmx.md (MMXMODE14): New mode iterator.
17172 (<smaxmin:code><MMXMODE14:mode>3): New expander.
17173 (*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
17174 (<umaxmin:code><MMXMODE24:mode>3): New expander.
17175 (*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
17176 (vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
17177 (vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
17178 (vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
17179 (vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
17180 (vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
17182 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
17184 * dwarf2out.c (loc_list_from_tree_1) <DECL>: During early DWARF, do
17185 not expand the VALUE_EXPR of variables put in the non-local frame.
17186 * gimplify.c (gimplify_type_sizes) <RECORD_TYPE>: If the type is not
17187 to be ignored for debug info, ensure its variable offsets are not.
17189 2021-05-05 Richard Biener <rguenther@suse.de>
17191 PR tree-optimization/79333
17192 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
17193 Fold stmt following SSA edges.
17195 2021-05-05 Richard Biener <rguenther@suse.de>
17197 PR middle-end/100394
17198 * calls.c (expand_call): Preserve possibly throwing calls.
17199 * cfgexpand.c (expand_call_stmt): When a call can throw signal
17200 RTL expansion there are side-effects.
17201 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Simplify,
17202 mark all possibly throwing stmts necessary unless we can elide
17204 * tree-ssa-dse.c (pass_dse::execute): Preserve exceptions unless
17205 -fdelete-dead-exceptions.
17206 * tree.h (DECL_PURE_P): Add note about exceptions.
17208 2021-05-05 Alexandre Oliva <oliva@adacore.com>
17210 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Make it
17213 2021-05-04 David Edelsohn <dje.gcc@gmail.com>
17215 * config/rs6000/rs6000-call.c (rs6000_output_mi_thunk): Use
17216 get_fnname_from_decl for name of thunk.
17217 * config/rs6000/rs6000.c (rs6000_declare_alias): Use assemble_name
17218 and ASM_OUTPUT_LABEL.
17219 (rs6000_xcoff_declare_function_name): Use assemble_name and
17221 (rs6000_xcoff_declare_object_name): Use ASM_OUTPUT_LABEL.
17222 (rs6000_xcoff_encode_section_info): Don't add mapping class
17223 for aliases. Always add [DS] mapping class to primary
17225 (rs6000_asm_weaken_decl): Don't explicitly add [DS].
17227 2021-05-04 Martin Sebor <msebor@redhat.com>
17229 PR middle-end/100307
17230 * builtins.c (compute_objsize_r): Clear base0 for pointers.
17232 2021-05-04 Jeff Law <jlaw@tachyum.com>
17234 * config/bfin/bfin.h (NOTICE_UPDATE_CC): Remove.
17236 2021-05-04 Segher Boessenkool <segher@kernel.crashing.org>
17238 * caller-save.c: Remove CC0.
17239 * cfgcleanup.c: Remove CC0.
17240 * cfgrtl.c: Remove CC0.
17241 * combine.c: Remove CC0.
17242 * compare-elim.c: Remove CC0.
17243 * conditions.h: Remove CC0.
17244 * config/h8300/h8300.h: Remove CC0.
17245 * config/h8300/h8300-protos.h: Remove CC0.
17246 * config/h8300/peepholes.md: Remove CC0.
17247 * config/i386/x86-tune-sched.c: Remove CC0.
17248 * config/m68k/m68k.c: Remove CC0.
17249 * config/rl78/rl78.c: Remove CC0.
17250 * config/sparc/sparc.c: Remove CC0.
17251 * config/xtensa/xtensa.c: Remove CC0.
17252 (gen_conditional_move): Use pc_rtx instead of cc0_rtx in a piece of
17253 RTL where that is used as a placeholder only.
17254 * cprop.c: Remove CC0.
17255 * cse.c: Remove CC0.
17256 * cselib.c: Remove CC0.
17257 * df-problems.c: Remove CC0.
17258 * df-scan.c: Remove CC0.
17259 * doc/md.texi: Remove CC0. Adjust an example.
17260 * doc/rtl.texi: Remove CC0. Adjust an example.
17261 * doc/tm.texi: Regenerate.
17262 * doc/tm.texi.in: Remove CC0.
17263 * emit-rtl.c: Remove CC0.
17264 * final.c: Remove CC0.
17265 * fwprop.c: Remove CC0.
17266 * gcse-common.c: Remove CC0.
17267 * gcse.c: Remove CC0.
17268 * genattrtab.c: Remove CC0.
17269 * genconfig.c: Remove CC0.
17270 * genemit.c: Remove CC0.
17271 * genextract.c: Remove CC0.
17272 * gengenrtl.c: Remove CC0.
17273 * genrecog.c: Remove CC0.
17274 * haifa-sched.c: Remove CC0.
17275 * ifcvt.c: Remove CC0.
17276 * ira-costs.c: Remove CC0.
17277 * ira.c: Remove CC0.
17278 * jump.c: Remove CC0.
17279 * loop-invariant.c: Remove CC0.
17280 * lra-constraints.c: Remove CC0.
17281 * lra-eliminations.c: Remove CC0.
17282 * optabs.c: Remove CC0.
17283 * postreload-gcse.c: Remove CC0.
17284 * postreload.c: Remove CC0.
17285 * print-rtl.c: Remove CC0.
17286 * read-rtl-function.c: Remove CC0.
17287 * reg-notes.def: Remove CC0.
17288 * reg-stack.c: Remove CC0.
17289 * reginfo.c: Remove CC0.
17290 * regrename.c: Remove CC0.
17291 * reload.c: Remove CC0.
17292 * reload1.c: Remove CC0.
17293 * reorg.c: Remove CC0.
17294 * resource.c: Remove CC0.
17295 * rtl.c: Remove CC0.
17296 * rtl.def: Remove CC0.
17297 * rtl.h: Remove CC0.
17298 * rtlanal.c: Remove CC0.
17299 * sched-deps.c: Remove CC0.
17300 * sched-rgn.c: Remove CC0.
17301 * shrink-wrap.c: Remove CC0.
17302 * simplify-rtx.c: Remove CC0.
17303 * system.h: Remove CC0. Poison NOTICE_UPDATE_CC, CC_STATUS_MDEP_INIT,
17304 CC_STATUS_MDEP, and CC_STATUS.
17305 * target.def: Remove CC0.
17306 * valtrack.c: Remove CC0.
17307 * var-tracking.c: Remove CC0.
17309 2021-05-04 Richard Biener <rguenther@suse.de>
17311 PR tree-optimization/100414
17312 * tree-ssa-phiopt.c (get_non_trapping): Do not compute dominance
17314 (tree_ssa_phiopt_worker): But unconditionally here.
17316 2021-05-04 Tobias Burnus <tobias@codesourcery.com>
17318 * omp-low.c (lower_rec_input_clauses, lower_reduction_clauses): Handle
17319 && and || with floating-point and complex arguments.
17321 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
17323 * tree-inline.c (insert_debug_decl_map): Delete.
17324 (copy_debug_stmt): Minor tweak.
17325 (setup_one_parameter): Do not use a variable if the value is either
17326 a read-only DECL or a non-addressable local variable in the caller.
17327 In this case, insert the debug-only variable in the map manually.
17328 (expand_call_inline): Do not generate a CLOBBER for these values.
17329 * tree-inline.h (debug_map): Minor tweak.
17331 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
17333 * builtins.c (builtin_with_linkage_p): Return true for stp[n]cpy.
17334 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Tidy up.
17336 2021-05-04 Richard Biener <rguenther@suse.de>
17338 PR tree-optimization/100329
17339 * tree-ssa-reassoc.c (can_reassociate_p): Do not reassociate
17341 (insert_stmt_after): Assert we're not running into asm goto.
17343 2021-05-04 Richard Biener <rguenther@suse.de>
17345 PR tree-optimization/100398
17346 * tree-ssa-dse.c (pass_dse::execute): Preserve control
17349 2021-05-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17351 * builtins.c (try_store_by_multiple_pieces): Fix constfun's prototype.
17353 2021-05-04 Alexandre Oliva <oliva@adacore.com>
17355 * builtins.c (try_store_by_multiple_pieces): New.
17356 (expand_builtin_memset_args): Use it. If target_char_cast
17357 fails, proceed as for non-constant val. Pass len's ctz to...
17358 * expr.c (clear_storage_hints): ... this. Try store by
17359 multiple pieces after setmem.
17360 (clear_storage): Adjust.
17361 * expr.h (clear_storage_hints): Likewise.
17362 (try_store_by_multiple_pieces): Declare.
17363 * passes.def: Replace the last copy_prop with ccp.
17365 2021-05-03 Tom de Vries <tdevries@suse.de>
17368 * omp-low.c (lower_rec_input_clauses): Disable SIMT for user-defined
17371 2021-05-03 Richard Biener <rguenther@suse.de>
17373 * tree-ssa-dse.c (dse_classify_store): Track two PHI defs.
17375 2021-05-03 Richard Biener <rguenther@suse.de>
17377 * tree-ssa-dse.c: Do not include domwalk.h but cfganal.h.
17378 (dse_dom_walker): Remove.
17379 (dse_dom_walker::dse_optimize_stmt): Rename...
17380 (dse_optimize_stmt): ... to this, pass in live_bytes sbitmap.
17381 (dse_dom_walker::before_dom_children): Inline ...
17382 (pass_dse::execute): ... here. Perform a reverse program
17385 2021-05-03 H.J. Lu <hjl.tools@gmail.com>
17388 * configure: Regenerated.
17390 2021-05-03 Ilya Leoshkevich <iii@linux.ibm.com>
17393 * config/s390/s390.c (s390_hard_fp_reg_p): New function.
17394 (s390_md_asm_adjust): Handle hard registers.
17396 2021-05-03 Jakub Jelinek <jakub@redhat.com>
17398 PR tree-optimization/100382
17399 * tree-ssa-dse.c: Include tree-eh.h.
17400 (dse_dom_walker::before_dom_children): Don't remove stmts if
17401 stmt_unremovable_because_of_non_call_eh_p is true.
17403 2021-05-02 David Edelsohn <dje.gcc@gmail.com>
17405 * varasm.c (compute_reloc_for_var): Split out from...
17406 (get_variable_section): Use it.
17407 * output.h (compute_reloc_for_var): Declare.
17408 * config/rs6000/rs6000-protos.h
17409 (rs6000_xcoff_asm_output_aligned_decl_common): Change alignment to
17411 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address_aix):
17412 Don't append storage mapping class to symbol.
17413 (rs6000_xcoff_asm_named_section): Add BS and UL mapping classes.
17414 Don't convert TLS BSS to common.
17415 (rs6000_xcoff_unique_section): Don't fall back to select_secton.
17416 (rs6000_xcoff_section_type_flags): Add SECTION_BSS if DECL is
17418 (rs6000_xcoff_asm_globalize_decl_name): Don't strip storage
17420 (rs6000_xcoff_asm_output_aligned_decl_common): Align is unsigned int.
17421 If align is 0 from TLS class, use the same rules as varasm.c
17422 If not common, switch to BSS section manually.
17423 If common, emit appropriate comm or lcomm directive.
17424 (rs6000_xcoff_encode_section_info): Add logic to append all
17425 storage mapping classes.
17426 (rs6000_asm_weaken_decl): Adjust for qualname symbols.
17427 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
17428 rs6000_xcoff_asm_output_aligned_decl_common.
17429 (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
17430 rs6000_xcoff_asm_output_aligned_decl_common.
17431 (ASM_OUTPUT_TLS_COMMON): Use
17432 rs6000_xcoff_asm_output_aligned_decl_common.
17434 2021-05-02 Jakub Jelinek <jakub@redhat.com>
17437 * config/nvptx/nvptx.c (nvptx_sese_pseudo): Use nullptr instead of 0
17438 as first argument of pseudo_node_t constructors.
17440 2021-05-02 Jakub Jelinek <jakub@redhat.com>
17443 * config/i386/t-i386 (TM_H): Add $(srcdir)/config/i386/i386-isa.def.
17445 2021-05-01 Aldy Hernandez <aldyh@redhat.com>
17447 * value-range.cc (DEFINE_INT_RANGE_GC_STUBS): Remove.
17448 (gt_pch_nx (int_range<1> *&)): New.
17449 (gt_ggc_mx (int_range<1> *&)): New.
17450 * value-range.h (class irange): Add GTY support for
17453 2021-05-01 Geng Qi <gengqi@linux.alibaba.com>
17455 * doc/options.texi (Negative): Change either or to both and.
17457 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17459 * config/aarch64/aarch64-simd-builtins.def: Add
17460 float_ml[as][q]_laneq builtin generator macros.
17461 * config/aarch64/aarch64-simd.md (mul_laneq<mode>3): Define.
17462 (aarch64_float_mla_laneq<mode>): Define.
17463 (aarch64_float_mls_laneq<mode>): Define.
17464 * config/aarch64/arm_neon.h (vmla_laneq_f32): Use RTL builtin
17465 instead of GCC vector extensions.
17466 (vmlaq_laneq_f32): Likewise.
17467 (vmls_laneq_f32): Likewise.
17468 (vmlsq_laneq_f32): Likewise.
17470 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17472 * config/aarch64/aarch64-simd-builtins.def: Add
17473 float_ml[as]_lane builtin generator macros.
17474 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt<mode>):
17476 (mul_lane<mode>3): This, and re-order arguments.
17477 (aarch64_float_mla_lane<mode>): Define.
17478 (aarch64_float_mls_lane<mode>): Define.
17479 * config/aarch64/arm_neon.h (vmla_lane_f32): Use RTL builtin
17480 instead of GCC vector extensions.
17481 (vmlaq_lane_f32): Likewise.
17482 (vmls_lane_f32): Likewise.
17483 (vmlsq_lane_f32): Likewise.
17485 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17487 * config/aarch64/aarch64-simd-builtins.def: Add float_ml[as]
17488 builtin generator macros.
17489 * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>):
17491 (aarch64_float_mls<mode>): Define.
17492 * config/aarch64/arm_neon.h (vmla_f32): Use RTL builtin
17493 instead of relying on GCC vector extensions.
17494 (vmla_f64): Likewise.
17495 (vmlaq_f32): Likewise.
17496 (vmlaq_f64): Likewise.
17497 (vmls_f32): Likewise.
17498 (vmls_f64): Likewise.
17499 (vmlsq_f32): Likewise.
17500 (vmlsq_f64): Likewise.
17501 * config/aarch64/iterators.md: Define VDQF_DF mode iterator.
17503 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17505 * config/aarch64/aarch64-simd-builtins.def: Add
17506 float_ml[as]_n_builtin generator macros.
17507 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt_from_dup<mode>):
17509 (mul_n<mode>3): This, and re-order arguments.
17510 (aarch64_float_mla_n<mode>): Define.
17511 (aarch64_float_mls_n<mode>): Define.
17512 * config/aarch64/arm_neon.h (vmla_n_f32): Use RTL builtin
17513 instead of inline asm.
17514 (vmlaq_n_f32): Likewise.
17515 (vmls_n_f32): Likewise.
17516 (vmlsq_n_f32): Likewise.
17518 2021-04-30 Jonathan Wright <joanthan.wright@arm.com>
17520 * config/aarch64/aarch64-simd-builtins.def: Add pmull[2]
17521 builtin generator macros.
17522 * config/aarch64/aarch64-simd.md (aarch64_pmullv8qi): Define.
17523 (aarch64_pmull_hiv16qi_insn): Define.
17524 (aarch64_pmull_hiv16qi): Define.
17525 * config/aarch64/arm_neon.h (vmull_high_p8): Use RTL builtin
17526 instead of inline asm.
17527 (vmull_p8): Likewise.
17529 2021-04-30 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
17531 * config/avr/avr.md: Adjust peepholes to match and
17532 generate parallels with clobber of REG_CC.
17533 (mov<mode>_insn): Rename to mov<mode>_insn_split.
17534 (*mov<mode>_insn): Rename to mov<mode>_insn.
17536 2021-04-30 David Edelsohn <dje.gcc@gmail.com>
17538 * varasm.c (use_blocks_for_decl_p): Don't use section anchors
17539 for VAR_DECLs if -fdata-sections enabled.
17541 2021-04-30 Michael Meissner <meissner@linux.ibm.com>
17543 PR bootstrap/100327
17544 * config/rs6000/rs6000.c
17545 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Define.
17546 (rs6000_libgcc_floating_mode_supported_p): New target hook.
17548 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
17550 * tree-ssa-threadbackward.c (class thread_jumps): Split out code
17552 (class back_threader_registry): ...to here...
17553 (class back_threader_profitability): ...and here...
17554 (thread_jumps::thread_through_all_blocks): Remove argument.
17555 (back_threader_registry::back_threader_registry): New.
17556 (back_threader_registry::~back_threader_registry): New.
17557 (back_threader_registry::thread_through_all_blocks): New.
17558 (thread_jumps::profitable_jump_thread_path): Move from here...
17559 (back_threader_profitability::profitable_path_p): ...to here.
17560 (thread_jumps::find_taken_edge): New.
17561 (thread_jumps::convert_and_register_current_path): Move...
17562 (back_threader_registry::register_path): ...to here.
17563 (thread_jumps::register_jump_thread_path_if_profitable): Move...
17564 (thread_jumps::maybe_register_path): ...to here.
17565 (thread_jumps::handle_phi): Call find_taken_edge and
17566 maybe_register_path.
17567 (thread_jumps::handle_assignment): Same.
17568 (thread_jumps::fsm_find_control_statement_thread_paths): Remove
17569 tree argument to handle_phi and handle_assignment.
17570 (thread_jumps::find_jump_threads_backwards): Set m_name. Remove
17571 set of m_speed_p and m_max_threaded_paths.
17572 (pass_thread_jumps::execute): Remove second argument from
17573 find_jump_threads_backwards.
17574 (pass_early_thread_jumps::execute): Same.
17576 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
17578 * tree-ssa-dom.c (class dom_jump_threader_simplifier): New.
17579 (class dom_opt_dom_walker): Initialize some class variables.
17580 (pass_dominator::execute): Pass evrp_range_analyzer and
17581 dom_jump_threader_simplifier to dom_opt_dom_walker.
17582 Adjust for some functions moving into classes.
17583 (simplify_stmt_for_jump_threading): Adjust and move to...
17584 (jump_threader_simplifier::simplify): ...here.
17585 (dom_opt_dom_walker::before_dom_children): Adjust for
17586 m_evrp_range_analyzer.
17587 (dom_opt_dom_walker::after_dom_children): Remove x_vr_values hack.
17588 (test_for_singularity): Place in dom_opt_dom_walker class.
17589 (dom_opt_dom_walker::optimize_stmt): The argument
17590 evrp_range_analyzer is now a class field.
17591 * tree-ssa-threadbackward.c (class thread_jumps): Add m_registry.
17592 (thread_jumps::thread_through_all_blocks): New.
17593 (thread_jumps::convert_and_register_current_path): Use m_registry.
17594 (pass_thread_jumps::execute): Adjust for thread_through_all_blocks
17595 being in the threader class.
17596 (pass_early_thread_jumps::execute): Same.
17597 * tree-ssa-threadedge.c (threadedge_initialize_values): Move...
17598 (jump_threader::jump_threader): ...here.
17599 (threadedge_finalize_values): Move...
17600 (jump_threader::~jump_threader): ...here.
17601 (jump_threader::remove_jump_threads_including): New.
17602 (jump_threader::thread_through_all_blocks): New.
17603 (record_temporary_equivalences_from_phis): Move...
17604 (jump_threader::record_temporary_equivalences_from_phis): ...here.
17605 (record_temporary_equivalences_from_stmts_at_dest): Move...
17606 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
17608 (simplify_control_stmt_condition_1): Move to jump_threader class.
17609 (simplify_control_stmt_condition): Move...
17610 (jump_threader::simplify_control_stmt_condition): ...here.
17611 (thread_around_empty_blocks): Move...
17612 (jump_threader::thread_around_empty_blocks): ...here.
17613 (thread_through_normal_block): Move...
17614 (jump_threader::thread_through_normal_block): ...here.
17615 (thread_across_edge): Move...
17616 (jump_threader::thread_across_edge): ...here.
17617 (thread_outgoing_edges): Move...
17618 (jump_threader::thread_outgoing_edges): ...here.
17619 * tree-ssa-threadedge.h: Move externally facing functings...
17620 (class jump_threader): ...here...
17621 (class jump_threader_simplifier): ...and here.
17622 * tree-ssa-threadupdate.c (struct redirection_data): Remove comment.
17623 (jump_thread_path_allocator::jump_thread_path_allocator): New.
17624 (jump_thread_path_allocator::~jump_thread_path_allocator): New.
17625 (jump_thread_path_allocator::allocate_thread_edge): New.
17626 (jump_thread_path_allocator::allocate_thread_path): New.
17627 (jump_thread_path_registry::jump_thread_path_registry): New.
17628 (jump_thread_path_registry::~jump_thread_path_registry): New.
17629 (jump_thread_path_registry::allocate_thread_edge): New.
17630 (jump_thread_path_registry::allocate_thread_path): New.
17631 (dump_jump_thread_path): Make extern.
17632 (debug (const vec<jump_thread_edge *> &path)): New.
17633 (struct removed_edges): Move to tree-ssa-threadupdate.h.
17634 (struct thread_stats_d): Remove.
17635 (remove_ctrl_stmt_and_useless_edges): Make static.
17636 (lookup_redirection_data): Move...
17637 (jump_thread_path_registry::lookup_redirection_data): ...here.
17638 (ssa_redirect_edges): Make static.
17639 (thread_block_1): Move...
17640 (jump_thread_path_registry::thread_block_1): ...here.
17641 (thread_block): Move...
17642 (jump_thread_path_registry::thread_block): ...here.
17643 (thread_through_loop_header): Move...
17644 (jump_thread_path_registry::thread_through_loop_header): ...here.
17645 (mark_threaded_blocks): Move...
17646 (jump_thread_path_registry::mark_threaded_blocks): ...here.
17647 (debug_path): Move...
17648 (jump_thread_path_registry::debug_path): ...here.
17649 (debug_all_paths): Move...
17650 (jump_thread_path_registry::dump): ..here.
17651 (rewire_first_differing_edge): Move...
17652 (jump_thread_path_registry::rewire_first_differing_edge): ...here.
17653 (adjust_paths_after_duplication): Move...
17654 (jump_thread_path_registry::adjust_paths_after_duplication): ...here.
17655 (duplicate_thread_path): Move...
17656 (jump_thread_path_registry::duplicate_thread_path): ..here.
17657 (remove_jump_threads_including): Move...
17658 (jump_thread_path_registry::remove_jump_threads_including): ...here.
17659 (thread_through_all_blocks): Move to...
17660 (jump_thread_path_registry::thread_through_all_blocks): ...here.
17661 (delete_jump_thread_path): Remove.
17662 (register_jump_thread): Move...
17663 (jump_thread_path_registry::register_jump_thread): ...here.
17664 * tree-ssa-threadupdate.h: Move externally facing functions...
17665 (class jump_thread_path_allocator): ...here...
17666 (class jump_thread_path_registry): ...and here.
17667 (thread_through_all_blocks): Remove.
17668 (struct removed_edges): New.
17669 (register_jump_thread): Remove.
17670 (remove_jump_threads_including): Remove.
17671 (delete_jump_thread_path): Remove.
17672 (remove_ctrl_stmt_and_useless_edges): Remove.
17673 (free_dom_edge_info): New prototype.
17674 * tree-vrp.c: Remove x_vr_values hack.
17675 (class vrp_jump_threader_simplifier): New.
17676 (vrp_jump_threader_simplifier::simplify): New.
17677 (vrp_jump_threader::vrp_jump_threader): Adjust method signature.
17678 Remove m_dummy_cond.
17679 Instantiate m_simplifier and m_threader.
17680 (vrp_jump_threader::thread_through_all_blocks): New.
17681 (vrp_jump_threader::simplify_stmt): Remove.
17682 (vrp_jump_threader::after_dom_children): Do not set m_dummy_cond.
17683 Remove x_vr_values hack.
17684 (execute_vrp): Adjust for thread_through_all_blocks being in a
17687 2021-04-30 Christophe Lyon <christophe.lyon@linaro.org>
17689 * genflags.c (gen_insn): Print failed expansion string.
17691 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
17693 * expr.c (alignment_for_piecewise_move): Call mode_for_size
17694 without limit to MAX_FIXED_MODE_SIZE.
17696 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
17698 PR middle-end/90773
17699 * builtins.c (builtin_memset_gen_str): Don't use return from
17700 simplify_gen_subreg.
17702 2021-04-30 Uroš Bizjak <ubizjak@gmail.com>
17705 * config/i386/i386.md (*add<mode>3_carry_0r): New insn pattern.
17706 (*addsi3_carry_zext_0r): Ditto.
17707 (*sub<mode>3_carry_0): Ditto.
17708 (*subsi3_carry_zext_0r): Ditto.
17709 * config/i386/predicates.md (ix86_carry_flag_unset_operator):
17711 * config/i386/i386.c (ix86_rtx_costs) <case PLUS, case MINUS>:
17712 Also consider ix86_carry_flag_unset_operator to calculate
17713 the cost of adc/sbb insn.
17715 2021-04-30 Roman Zhuykov <zhroma@ispras.ru>
17717 PR rtl-optimization/100225
17718 PR rtl-optimization/84878
17719 * modulo-sched.c (sms_schedule): Use note_stores to skip loops
17720 where we have an instruction which touches (writes) any hard
17721 register from df->regular_block_artificial_uses set.
17722 Allow not-single-set instruction only right before basic block
17725 2021-04-30 Geng Qi <gengqi@linux.alibaba.com>
17727 * config/riscv/riscv.opt (march=,mabi=): Negative itself.
17729 2021-04-30 LevyHsu <admin@levyhsu.com>
17731 * config/riscv/riscv.c (riscv_min_arithmetic_precision): New.
17732 * config/riscv/riscv.h (TARGET_MIN_ARITHMETIC_PRECISION): New.
17733 * config/riscv/riscv.md (addv<mode>4, uaddv<mode>4): New.
17734 (subv<mode>4, usubv<mode>4, mulv<mode>4, umulv<mode>4): New.
17736 2021-04-29 Alexandre Oliva <oliva@adacore.com>
17738 * config.gcc: Merged x86 and x86_64 cpu_type-setting cases.
17740 2021-04-29 Alexandre Oliva <oliva@adacore.com>
17742 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Rename to...
17743 (ASM_OUTPUT_MAX_SKIP_ALIGN): ... this. Enclose in do/while(0).
17744 * config/i386/i386.c: Adjust.
17745 * config/i386/i386.md: Adjust.
17746 * config/i386/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Drop.
17747 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17748 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17749 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17750 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17751 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17752 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17753 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17754 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17755 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17756 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
17758 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
17760 * config/i386/i386-expand.c (ix86_expand_int_compare):
17761 Swap operands of GTU and LEU comparison to emit carry flag comparison.
17762 * config/i386/i386.md (*add<mode>3_carry_0): Change insn
17763 predicate to allow more combine opportunities with memory operands.
17764 (*sub<mode>3_carry_0): Ditto.
17766 2021-04-29 Richard Sandiford <richard.sandiford@arm.com>
17768 PR rtl-optimization/100303
17769 * rtl-ssa/accesses.cc (function_info::make_use_available): Take a
17770 boolean that indicates whether the use will only be used in
17771 debug instructions. Treat it in the same way that existing
17772 cross-EBB debug references would be handled if so.
17773 (function_info::make_uses_available): Likewise.
17774 * rtl-ssa/functions.h (function_info::make_uses_available): Update
17775 prototype accordingly.
17776 (function_info::make_uses_available): Likewise.
17777 * fwprop.c (try_fwprop_subst): Update call accordingly.
17779 2021-04-29 Jeff Law <jlaw@tachyum.com>
17781 * config/nios2/nios2-protos.h (nios2_fpu_insn_enabled): Move outside
17784 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
17785 Richard Biener <rguenther@suse.de>
17788 * config/i386/i386-builtin.def (IX86_BUILTIN_MASKLOADPD)
17789 (IX86_BUILTIN_MASKLOADPS, IX86_BUILTIN_MASKLOADPD256)
17790 (IX86_BUILTIN_MASKLOADPS256, IX86_BUILTIN_MASKLOADD)
17791 (IX86_BUILTIN_MASKLOADQ, IX86_BUILTIN_MASKLOADD256)
17792 (IX86_BUILTIN_MASKLOADQ256): Move from SPECIAL_ARGS
17793 to PURE_ARGS category.
17794 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
17795 Handle PURE_ARGS category.
17796 * config/i386/i386-expand.c (ix86_expand_builtin): Ditto.
17798 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
17800 * configure.ac: Check for the presence of sys/locking.h header and
17801 for whether _LK_LOCK is supported by _locking.
17802 * configure: Regenerate.
17803 * config.in: Likewise.
17804 * gcov-io.h: Define GCOV_LOCKED_WITH_LOCKING if HOST_HAS_LK_LOCK.
17805 * gcov-io.c (gcov_open): Add support for GCOV_LOCKED_WITH_LOCKING.
17806 * system.h: Include <sys/locking.h> if HAVE_SYS_LOCKING_H.
17808 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
17810 * config/i386/predicates.md (fcmov_comparison_operator):
17811 Do not check for trivial FP comparison operator.
17812 <case GEU, case LTU>: Allow CCGZmode.
17813 <case GTU, case LEU>: Do not allow CCCmode.
17814 (ix86_comparison_operator) <case GTU, case LEU>: Allow only CCmode.
17815 (ix86_carry_flag_operator): Match only LTU and UNLT code.
17816 Do not check for trivial FP comparison operator. Allow CCGZmode.
17818 2021-04-29 Tom de Vries <tdevries@suse.de>
17820 * omp-expand.c (expand_omp_simd): Add step_orig, and replace uses of
17821 fd->loop.step by either step or orig_step.
17823 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
17825 * config/sparc/sparc.c (gen_load_pcrel_sym): Delete.
17826 (load_got_register): Do the PIC dance here.
17827 (sparc_legitimize_tls_address): Simplify.
17828 (sparc_emit_probe_stack_range): Likewise.
17829 (sparc32_initialize_trampoline): Likewise.
17830 (sparc64_initialize_trampoline): Likewise.
17831 * config/sparc/sparc.md (load_pcrel_sym<P:mode>): Add @ marker.
17832 (probe_stack_range<P:mode>): Likewise.
17833 (flush<P:mode>): Likewise.
17834 (tgd_hi22<P:mode>): Likewise.
17835 (tgd_lo10<P:mode>): Likewise.
17836 (tgd_add<P:mode>): Likewise.
17837 (tgd_call<P:mode>): Likewise.
17838 (tldm_hi22<P:mode>): Likewise.
17839 (tldm_lo10<P:mode>): Likewise.
17840 (tldm_add<P:mode>): Likewise.
17841 (tldm_call<P:mode>): Likewise.
17842 (tldo_hix22<P:mode>): Likewise.
17843 (tldo_lox10<P:mode>): Likewise.
17844 (tldo_add<P:mode>): Likewise.
17845 (tie_hi22<P:mode>): Likewise.
17846 (tie_lo10<P:mode>): Likewise.
17847 (tie_add<P:mode>): Likewise.
17848 (tle_hix22<P:mode>): Likewise.
17849 (tle_lox10<P:mode>): Likewise.
17850 (stack_protect_setsi): Rename to...
17851 (stack_protect_set32): ...this.
17852 (stack_protect_setdi): Rename to...
17853 (stack_protect_set64): ...this.
17854 (stack_protect_set): Adjust calls to above.
17855 (stack_protect_testsi): Rename to...
17856 (stack_protect_test32): ...this.
17857 (stack_protect_testdi): Rename to...
17858 (stack_protect_test64): ...this.
17859 (stack_protect_test): Adjust calls to above.
17861 2021-04-29 H.J. Lu <hjl.tools@gmail.com>
17863 PR middle-end/90773
17864 * builtins.c (builtin_memcpy_read_str): Add a dummy argument.
17865 (builtin_strncpy_read_str): Likewise.
17866 (builtin_memset_read_str): Add an argument for the previous RTL
17867 information and generate the new RTL from the previous RTL info.
17868 (builtin_memset_gen_str): Likewise.
17869 * builtins.h (builtin_strncpy_read_str): Update the prototype.
17870 (builtin_memset_read_str): Likewise.
17871 * expr.c (by_pieces_ninsns): If targetm.overlap_op_by_pieces_p()
17872 returns true, round up size and alignment to the widest integer
17873 mode for maximum size.
17874 (pieces_addr::adjust): Add a pointer to by_pieces_prev argument
17875 and pass it to m_constfn.
17876 (op_by_pieces_d): Add m_push and m_overlap_op_by_pieces.
17877 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
17878 initialize m_push. Initialize m_overlap_op_by_pieces with
17879 targetm.overlap_op_by_pieces_p ().
17880 (op_by_pieces_d::run): Pass the previous RTL information to
17881 pieces_addr::adjust and generate overlapping operations if
17882 m_overlap_op_by_pieces is true.
17884 (move_by_pieces_d::move_by_pieces_d): Updated for op_by_pieces_d
17886 (store_by_pieces_d::store_by_pieces_d): Updated for op_by_pieces_d
17888 (can_store_by_pieces): Use by_pieces_constfn on constfun.
17889 (store_by_pieces): Use by_pieces_constfn on constfun. Updated
17890 for op_by_pieces_d change.
17891 (clear_by_pieces_1): Add a dummy argument.
17892 (clear_by_pieces): Updated for op_by_pieces_d change.
17893 (compare_by_pieces_d::compare_by_pieces_d): Likewise.
17894 (string_cst_read_str): Add a dummy argument.
17895 * expr.h (by_pieces_constfn): Add a dummy argument.
17896 (by_pieces_prev): New.
17897 * target.def (overlap_op_by_pieces_p): New target hook.
17898 * config/i386/i386.c (TARGET_OVERLAP_OP_BY_PIECES_P): New.
17899 * doc/tm.texi.in: Add TARGET_OVERLAP_OP_BY_PIECES_P.
17900 * doc/tm.texi: Regenerated.
17902 2021-04-29 Richard Biener <rguenther@suse.de>
17904 PR tree-optimization/100253
17905 * tree-vect-stmts.c (vectorizable_load): Do not assume
17906 element alignment when DR_MISALIGNMENT is -1.
17907 (vectorizable_store): Likewise.
17909 2021-04-29 Jakub Jelinek <jakub@redhat.com>
17912 * config/aarch64/aarch64.c (aarch64_add_offset_1_temporaries): Use
17913 absu_hwi instead of abs_hwi.
17915 2021-04-29 Richard Biener <rguenther@suse.de>
17917 PR middle-end/38474
17918 * tree-ssa-structalias.c (add_graph_edge): Avoid direct
17919 forwarding when indirect forwarding through ESCAPED
17922 2021-04-29 Tom de Vries <tdevries@suse.de>
17925 * internal-fn.c (expand_GOMP_SIMT_ENTER_ALLOC)
17926 (expand_GOMP_SIMT_LAST_LANE, expand_GOMP_SIMT_ORDERED_PRED)
17927 (expand_GOMP_SIMT_VOTE_ANY, expand_GOMP_SIMT_XCHG_BFLY)
17928 (expand_GOMP_SIMT_XCHG_IDX): Ensure target is assigned to.
17930 2021-04-29 Richard Biener <rguenther@suse.de>
17932 PR tree-optimization/99912
17933 * tree-ssa-dse.c (dse_dom_walker::m_need_cfg_cleanup): New.
17934 (dse_dom_walker::todo): Likewise.
17935 (dse_dom_walker::dse_optimize_stmt): Move VDEF check to the
17937 (dse_dom_walker::before_dom_children): Remove trivially
17938 dead SSA defs and schedule CFG cleanup if we removed all
17940 (pass_dse::execute): Get TODO as computed by the DOM walker
17941 and return it. Wipe dominator info earlier.
17943 2021-04-29 Richard Biener <rguenther@suse.de>
17946 * ipa-prop.c (ipcp_modif_dom_walker::before_dom_children):
17947 Track blocks to cleanup EH in new m_need_eh_cleanup.
17948 (ipcp_modif_dom_walker::cleanup_eh): New.
17949 (ipcp_transform_function): Release dominator info before
17952 2021-04-29 Martin Sebor <msebor@redhat.com>
17954 PR middle-end/100250
17955 * attribs.c (attr_access::array_as_string): Avoid dereferencing
17956 a pointer when it's null.
17958 2021-04-29 Martin Sebor <msebor@redhat.com>
17960 * Makefile.in (OBJS): Add ipa-free-lang-data.o.
17961 * ipa-free-lang-data.cc: New file.
17962 * tree.c: Move pass free_lang_data to file above.
17963 (build_array_type_1): Declare extern.
17964 * tree.h (build_array_type_1): Declare.
17966 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17968 * config/aarch64/aarch64-simd-builtins.def: Modify comment to
17969 make consistent with updated RTL pattern.
17970 * config/aarch64/aarch64-simd.md (aarch64_<sur>qmovn<mode>):
17971 Implement using ss_truncate and us_truncate rather than
17973 * config/aarch64/iterators.md: Remove redundant unspecs and
17974 iterator: UNSPEC_[SU]QXTN and SUQMOVN respectively.
17976 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17978 * config/aarch64/arm_acle.h (__attribute__): Make intrinsic
17979 attributes consistent with those defined in arm_neon.h.
17981 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17983 * config/aarch64/arm_fp16.h (__attribute__): Make intrinsic
17984 attributes consistent with those defined in arm_neon.h.
17986 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17988 * config/aarch64/aarch64-simd-builtins.def: Add
17989 float_trunc_rodd builtin generator macros.
17990 * config/aarch64/aarch64-simd.md (aarch64_float_trunc_rodd_df):
17992 (aarch64_float_trunc_rodd_lo_v2sf): Define.
17993 (aarch64_float_trunc_rodd_hi_v4sf_le): Define.
17994 (aarch64_float_trunc_rodd_hi_v4sf_be): Define.
17995 (aarch64_float_trunc_rodd_hi_v4sf): Define.
17996 * config/aarch64/arm_neon.h (vcvtx_f32_f64): Use RTL builtin
17997 instead of inline asm.
17998 (vcvtx_high_f32_f64): Likewise.
17999 (vcvtxd_f32_f64): Likewise.
18000 * config/aarch64/iterators.md: Add FCVTXN unspec.
18002 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18004 * config/aarch64/aarch64-simd-builtins.def: Add tbx1 builtin
18006 * config/aarch64/aarch64-simd.md (aarch64_tbx1<mode>):
18008 * config/aarch64/arm_neon.h (vqtbx1_s8): USE RTL builtin
18009 instead of inline asm.
18010 (vqtbx1_u8): Likewise.
18011 (vqtbx1_p8): Likewise.
18012 (vqtbx1q_s8): Likewise.
18013 (vqtbx1q_u8): Likewise.
18014 (vqtbx1q_p8): Likewise.
18015 (vtbx2_s8): Likewise.
18016 (vtbx2_u8): Likewise.
18017 (vtbx2_p8): Likewise.
18019 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18021 * config/aarch64/aarch64-simd-builtins.def: Add tbl1 builtin
18023 * config/aarch64/arm_neon.h (vqtbl1_p8): Use RTL builtin
18024 instead of inline asm.
18025 (vqtbl1_s8): Likewise.
18026 (vqtbl1_u8): Likewise.
18027 (vqtbl1q_p8): Likewise.
18028 (vqtbl1q_s8): Likewise.
18029 (vqtbl1q_u8): Likewise.
18030 (vtbl1_s8): Likewise.
18031 (vtbl1_u8): Likewise.
18032 (vtbl1_p8): Likewise.
18033 (vtbl2_s8): Likewise.
18034 (vtbl2_u8): Likewise.
18035 (vtbl2_p8): Likewise.
18037 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18039 * config/aarch64/aarch64-simd-builtins.def: Add polynomial
18040 ssri_n buitin generator macro.
18041 * config/aarch64/arm_neon.h (vsri_n_p8): Use RTL builtin
18042 instead of inline asm.
18043 (vsri_n_p16): Likewise.
18044 (vsri_n_p64): Likewise.
18045 (vsriq_n_p8): Likewise.
18046 (vsriq_n_p16): Likewise.
18047 (vsriq_n_p64): Likewise.
18049 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18051 * config/aarch64/aarch64-simd-builtins.def: Use VALLP mode
18052 iterator for polynomial ssli_n builtin generator macro.
18053 * config/aarch64/arm_neon.h (vsli_n_p8): Use RTL builtin
18054 instead of inline asm.
18055 (vsli_n_p16): Likewise.
18056 (vsliq_n_p8): Likewise.
18057 (vsliq_n_p16): Likewise.
18058 * config/aarch64/iterators.md: Define VALLP mode iterator.
18060 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18062 * config/aarch64/aarch64-simd-builtins.def: Use VDQV_L
18063 iterator to generate [su]adalp RTL builtins.
18064 * config/aarch64/aarch64-simd.md: Use VDQV_L iterator in
18065 [su]adalp RTL pattern.
18066 * config/aarch64/arm_neon.h (vpadal_s32): Use RTL builtin
18067 instead of inline asm.
18068 (vpadal_u32): Likewise.
18070 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18072 * config/aarch64/aarch64-simd-builtins.def: Add [su]addlp
18073 builtin generator macros.
18074 * config/aarch64/aarch64-simd.md (aarch64_<su>addlp<mode>):
18076 * config/aarch64/arm_neon.h (vpaddl_s8): Use RTL builtin
18077 instead of inline asm.
18078 (vpaddl_s16): Likewise.
18079 (vpaddl_s32): Likewise.
18080 (vpaddl_u8): Likewise.
18081 (vpaddl_u16): Likewise.
18082 (vpaddl_u32): Likewise.
18083 (vpaddlq_s8): Likewise.
18084 (vpaddlq_s16): Likewise.
18085 (vpaddlq_s32): Likewise.
18086 (vpaddlq_u8): Likewise.
18087 (vpaddlq_u16): Likewise.
18088 (vpaddlq_u32): Liwewise.
18089 * config/aarch64/iterators.md: Define [SU]ADDLP unspecs with
18090 appropriate attributes.
18092 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18094 * config/aarch64/aarch64-simd-builtins.def: Use VDQ_I iterator
18095 for aarch64_addp<mode> builtin macro generator.
18096 * config/aarch64/aarch64-simd.md: Use VDQ_I iterator in
18097 aarch64_addp<mode> RTL pattern.
18098 * config/aarch64/arm_neon.h (vpaddq_s8): Use RTL builtin
18099 instead of inline asm.
18100 (vpaddq_s16): Likewise.
18101 (vpaddq_s32): Likewise.
18102 (vpaddq_s64): Likewise.
18103 (vpaddq_u8): Likewise.
18104 (vpaddq_u16): Likewise.
18105 (vpaddq_u32): Likewise.
18106 (vpaddq_u64): Likewise.
18108 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18110 * config/aarch64/aarch64-simd-builtins.def: Add sq[r]dmulh_n
18111 builtin generator macros.
18112 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_n<mode>):
18114 * config/aarch64/arm_neon.h (vqdmulh_n_s16): Use RTL builtin
18115 instead of inline asm.
18116 (vqdmulh_n_s32): Likewise.
18117 (vqdmulhq_n_s16): Likewise.
18118 (vqdmulhq_n_s32): Likewise.
18119 (vqrdmulh_n_s16): Likewise.
18120 (vqrdmulh_n_s32): Likewise.
18121 (vqrdmulhq_n_s16): Likewise.
18122 (vqrdmulhq_n_s32): Likewise.
18124 2021-04-28 Tobias Burnus <tobias@codesourcery.com>
18126 * doc/install.texi (--enable-offload-defaulted): Document.
18128 2021-04-28 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
18130 * config/avr/avr-dimode.md: Turn existing patterns into
18131 define_insn_and_split style patterns where the splitter
18132 adds a clobber of the condition code register. Drop "cc"
18133 attribute. Add new patterns to match output of
18135 * config/avr/avr-fixed.md: Likewise.
18136 * config/avr/avr.c (cc_reg_rtx): New.
18137 (avr_parallel_insn_from_insns): Adjust insn count
18138 for removal of set of cc0.
18139 (avr_is_casesi_sequence): Likewise.
18140 (avr_casei_sequence_check_operands): Likewise.
18141 (avr_optimize_casesi): Likewise. Also insert
18142 new insns after jump_insn.
18143 (avr_pass_casesi::avr_rest_of_handle_casesi): Adjust
18144 for removal of set of cc0.
18145 (avr_init_expanders): Initialize cc_reg_rtx.
18146 (avr_regno_reg_class): Handle REG_CC.
18147 (cond_string): Remove usage of CC_OVERFLOW_UNUSABLE.
18148 (avr_notice_update_cc): Remove function.
18149 (ret_cond_branch): Remove usage of CC_OVERFLOW_UNUSABLE.
18150 (compare_condition): Adjust for PARALLEL with
18152 (out_shift_with_cnt): Likewise.
18153 (ashlhi3_out): Likewise.
18154 (ashrhi3_out): Likewise.
18155 (lshrhi3_out): Likewise.
18156 (avr_class_max_nregs): Return single reg for REG_CC.
18157 (avr_compare_pattern): Check for REG_CC instead
18159 (avr_reorg_remove_redundant_compare): Likewise.
18160 (avr_reorg):Adjust for PARALLEL with REG_CC clobber.
18161 (avr_hard_regno_nregs): Return single reg for REG_CC.
18162 (avr_hard_regno_mode_ok): Allow only CCmode for REG_CC.
18163 (avr_md_asm_adjust): Clobber REG_CC.
18164 (TARGET_HARD_REGNO_NREGS): Define.
18165 (TARGET_CLASS_MAX_NREGS): Define.
18166 (TARGET_MD_ASM_ADJUST): Define.
18167 * config/avr/avr.h (FIRST_PSEUDO_REGISTER): Adjust
18169 (enum reg_class): Add CC_REG class.
18170 (NOTICE_UPDATE_CC): Remove.
18171 (CC_OVERFLOW_UNUSABLE): Remove.
18172 (CC_NO_CARRY): Remove.
18173 * config/avr/avr.md: Turn existing patterns into
18174 define_insn_and_split style patterns where the splitter
18175 adds a clobber of the condition code register. Drop "cc"
18176 attribute. Add new patterns to match output of
18178 (sez): Remove unused pattern.
18180 2021-04-28 Richard Earnshaw <rearnsha@arm.com>
18183 * config/arm/arm.c (arm_hard_regno_mode_ok): Only allow VPR to be
18186 2021-04-28 Richard Sandiford <richard.sandiford@arm.com>
18189 * config/aarch64/constraints.md (Utq): Require the address to
18190 be valid for both the element mode and for V2DImode.
18192 2021-04-28 Jakub Jelinek <jakub@redhat.com>
18193 Tobias Burnus <tobias@codesourcery.com>
18195 * configure.ac (OFFLOAD_DEFAULTED): AC_DEFINE if offload-defaulted.
18196 * gcc.c (process_command): New variable.
18197 (driver::maybe_putenv_OFFLOAD_TARGETS): If OFFLOAD_DEFAULTED,
18198 set it if -foffload is defaulted.
18199 * lto-wrapper.c (OFFLOAD_TARGET_DEFAULT_ENV): Define.
18200 (compile_offload_image): If OFFLOAD_DEFAULTED and
18201 OFFLOAD_TARGET_DEFAULT is in the environment, don't fail
18202 if corresponding mkoffload can't be found.
18203 (compile_images_for_offload_targets): Likewise. Free and clear
18204 offload_names if no valid offload is found.
18205 * config.in: Regenerate.
18206 * configure: Regenerate.
18208 2021-04-28 Richard Biener <rguenther@suse.de>
18210 PR tree-optimization/100292
18211 * tree-vect-generic.c (expand_vector_condition): Do not fold
18214 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
18216 * config/rs6000/aix.h (SUBTARGET_DRIVER_SELF_SPECS): New.
18217 * config/rs6000/aix64.opt (m64): New.
18220 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
18222 * config/vax/vax.c (print_operand_address, vax_address_cost_1)
18223 (index_term_p): Handle ASHIFT too.
18225 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
18227 * config/vax/builtins.md (jbb<ccss>i<mode>): Remove operand #3.
18228 (sync_lock_test_and_set<mode>): Adjust accordingly.
18229 (sync_lock_release<mode>): Likewise.
18231 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
18233 * config/vax/vax-protos.h (adjacent_operands_p): Remove
18235 * config/vax/vax.c (adjacent_operands_p): Remove.
18237 2021-04-27 Maciej W. Rozycki <macro@linux-mips.org>
18239 * ifcvt.c (dead_or_predicable) [!IFCVT_MODIFY_TESTS]: Fall
18240 through to the non-conditional execution case if getting the
18241 condition for conditional execution has failed.
18243 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
18245 PR middle-end/100284
18246 * gimple.c (gimple_could_trap_p_1): Remove VEC_COND_EXPR test.
18247 * tree-eh.c (operation_could_trap_p): Handle VEC_COND_EXPR rather
18248 than asserting on it.
18250 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
18252 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): Protect
18253 with TARGET_AIX_OS.
18255 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
18258 * calls.c (precompute_register_parameters): Additionally test
18259 targetm.precompute_tls_p to pre-compute argument.
18260 * config/rs6000/aix.h (TARGET_PRECOMPUTE_TLS_P): Define.
18261 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): New.
18262 * target.def (precompute_tls_p): New.
18263 * doc/tm.texi.in (TARGET_PRECOMPUTE_TLS_P): Add hook documentation.
18264 * doc/tm.texi: Regenerated.
18266 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18269 * config/aarch64/aarch64.c (aarch64_print_operand): Cast -UINTVAL
18270 back to HOST_WIDE_INT.
18272 2021-04-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
18275 * simplify-rtx.c (simplify_context::simplify_subreg): Check the
18276 memory alignment for the outer mode.
18278 2021-04-27 H.J. Lu <hjl.tools@gmail.com>
18280 PR middle-end/90773
18281 * expr.c (op_by_pieces_d::get_usable_mode): New member function.
18282 (op_by_pieces_d::run): Cange a while loop to a do-while loop.
18284 2021-04-27 Alex Coplan <alex.coplan@arm.com>
18287 * config/arm/arm.c (arm_split_compare_and_swap): Fix up codegen
18288 with negative immediates: ensure we expand cbranchsi4_scratch
18289 correctly and ensure we satisfy its constraints.
18290 * config/arm/sync.md
18291 (@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): Don't
18292 attempt to tie two output operands together with constraints;
18293 collapse two alternatives.
18294 (@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
18295 * config/arm/thumb1.md (cbranchsi4_neg_late): New.
18297 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18300 * config/aarch64/predicates.md (aarch64_sub_immediate,
18301 aarch64_plus_immediate): Use -UINTVAL instead of -INTVAL.
18302 * config/aarch64/aarch64.md (casesi, rotl<mode>3): Likewise.
18303 * config/aarch64/aarch64.c (aarch64_print_operand,
18304 aarch64_split_atomic_op, aarch64_expand_subvti): Likewise.
18306 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18308 PR tree-optimization/100239
18309 * tree-vect-generic.c (lower_vec_perm): Don't accept constant
18310 permutations with all indices from the first zero element as vec_shl.
18312 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18314 PR rtl-optimization/100254
18315 * cfgcleanup.c (outgoing_edges_match): Check REG_EH_REGION on
18316 last1 and last2 insns rather than BB_END (bb1) and BB_END (bb2) insns.
18318 2021-04-27 Richard Biener <rguenther@suse.de>
18320 PR tree-optimization/99912
18321 * passes.def: Add comment about new TODO_remove_unused_locals.
18322 * tree-stdarg.c (pass_data_stdarg): Run TODO_remove_unused_locals
18325 2021-04-27 Richard Biener <rguenther@suse.de>
18327 PR tree-optimization/99912
18328 * passes.def (pass_all_optimizations): Add pass_dse before
18329 the first pass_dce, move the first pass_dse before the
18330 pass_dce following pass_pre.
18332 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18334 PR tree-optimization/95527
18335 * generic-match-head.c: Include tm.h.
18336 * gimple-match-head.c: Include tm.h.
18337 * match.pd (CLZ == INTEGER_CST): Don't use
18338 #ifdef CLZ_DEFINED_VALUE_AT_ZERO, only test CLZ_DEFINED_VALUE_AT_ZERO
18339 if clz == CFN_CLZ. Add missing val declaration.
18340 (CTZ cmp CST): New simplifications.
18342 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18344 PR tree-optimization/96696
18345 * expr.c (expand_expr_divmod): New function.
18346 (expand_expr_real_2) <case TRUNC_DIV_EXPR>: Use it for truncations and
18347 divisions. Formatting fixes.
18348 <case MULT_EXPR>: Optimize x / y * y as x - x % y if the latter is
18351 2021-04-27 Martin Jambor <mjambor@suse.cz>
18354 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
18355 If removing a call statement LHS SSA name, release it.
18357 2021-04-27 Richard Earnshaw <rearnsha@arm.com>
18360 * config/arm/arm.c (THUMB2_WORK_REGS): Check PIC_OFFSET_TABLE_REGNUM
18361 is valid before including it in the mask.
18363 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
18366 * config/aarch64/aarch64.c (aarch64_comp_type_attributes): Handle
18369 2021-04-27 Richard Biener <rguenther@suse.de>
18371 PR tree-optimization/100051
18372 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Add
18373 disambiguator based on access size vs. decl size.
18375 2021-04-27 Richard Biener <rguenther@suse.de>
18377 PR tree-optimization/100278
18378 * tree-ssa-pre.c (compute_avail): Give up when we cannot
18379 adjust TBAA beacuse of mismatching bases.
18381 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18384 * config/i386/i386.md (*<insn><mode>3_mask, *<insn><mode>3_mask_1):
18385 For any_rotate define_insn_split and following splitters, use
18386 SWI iterator instead of SWI48.
18388 2021-04-27 Richard Biener <rguenther@suse.de>
18390 PR tree-optimization/99776
18391 * match.pd (bit_field_ref (ctor)): Relax element extract
18392 type compatibility checks.
18394 2021-04-27 Cui,Lili <lili.cui@intel.com>
18396 * common/config/i386/i386-common.c (processor_names):
18397 Sync processor_names with processor_type.
18398 * config/i386/i386-options.c (processor_cost_table):
18399 Sync processor_cost_table with processor_type.
18401 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18403 * value-range.cc (irange::irange_set_1bit_anti_range): Add assert.
18404 (irange::set): Call irange_set_1bit_anti_range for handling all
18405 1-bit ranges. Fall through on ~[MIN,MAX].
18407 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18409 * value-range.cc (irange::legacy_num_pairs): Remove.
18410 (irange::invert): Change gcc_assert to gcc_checking_assert.
18411 * value-range.h (irange::num_pairs): Adjust for a cached
18412 num_pairs(). Also, rename all gcc_assert's to
18413 gcc_checking_assert's.
18415 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18417 * value-range.cc (irange::operator=): Set m_kind.
18418 (irange::copy_to_legacy): Handle varying and undefined sources
18419 as a legacy copy since they can be easily copied.
18420 (irange::irange_set): Set m_kind.
18421 (irange::irange_set_anti_range): Same.
18422 (irange::set): Rename normalize_min_max to normalize_kind.
18423 (irange::verify_range): Adjust for multi-ranges having the
18425 (irange::irange_union): Set m_kind.
18426 (irange::irange_intersect): Same.
18427 (irange::invert): Same.
18428 * value-range.h (irange::kind): Always return m_kind.
18429 (irange::varying_p): Rename to...
18430 (irange::varying_comptaible_p): ...this.
18431 (irange::undefined_p): Only look at m_kind.
18432 (irange::irange): Always set VR_UNDEFINED if applicable.
18433 (irange::set_undefined): Always set VR_UNDEFINED.
18434 (irange::set_varying): Always set m_kind to VR_VARYING.
18435 (irange::normalize_min_max): Rename to...
18436 (irange::normalize_kind): ...this.
18438 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18440 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
18441 Adjust for constant_p including varying_p.
18442 * tree-vrp.c (vrp_prop::finalize): Same.
18443 (determine_value_range): Same.
18444 * vr-values.c (vr_values::range_of_expr): Same.
18445 * value-range.cc (irange::symbolic_p): Do not check varying_p.
18446 (irange::constant_p): Same.
18448 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18450 * value-range.cc (irange::legacy_lower_bound): Replace
18451 !undefined_p check with num_ranges > 0.
18452 (irange::legacy_upper_bound): Same.
18453 * value-range.h (irange::type): Same.
18454 (irange::lower_bound): Same.
18455 (irange::upper_bound): Same.
18457 2021-04-26 Richard Biener <rguenther@suse.de>
18459 PR tree-optimization/99956
18460 * gimple-loop-interchange.cc (compute_access_stride):
18461 Try instantiating the access in a shallower loop nest
18462 if instantiating failed.
18463 (compute_access_strides): Pass adjustable loop_nest
18464 to compute_access_stride.
18466 2021-04-26 Christophe Lyon <christophe.lyon@linaro.org>
18468 * doc/sourcebuild.texi (arm_cmse_hw): Document.
18470 2021-04-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18472 * config/aarch64/iterators.md (vwcore): Handle V4BF, V8BF.
18474 2021-04-26 Thomas Schwinge <thomas@codesourcery.com>
18475 Nathan Sidwell <nathan@codesourcery.com>
18476 Tom de Vries <vries@codesourcery.com>
18477 Julian Brown <julian@codesourcery.com>
18478 Kwok Cheung Yeung <kcy@codesourcery.com>
18480 * omp-offload.c (oacc_validate_dims): Implement
18481 '-Wopenacc-parallelism'.
18482 * doc/invoke.texi (-Wopenacc-parallelism): Document.
18484 2021-04-26 Richard Biener <rguenther@suse.de>
18486 * tree-cfg.h (gimplify_build1): Remove.
18487 (gimplify_build2): Likewise.
18488 (gimplify_build3): Likewise.
18489 * tree-cfg.c (gimplify_build1): Move to tree-vect-generic.c.
18490 (gimplify_build2): Likewise.
18491 (gimplify_build3): Likewise.
18492 * tree-vect-generic.c (gimplify_build1): Move from tree-cfg.c.
18494 (gimplify_build2): Likewise.
18495 (gimplify_build3): Likewise.
18496 (tree_vec_extract): Use resimplify with following SSA edges.
18497 (expand_vector_parallel): Avoid passing NULL size/bitpos
18498 to tree_vec_extract.
18499 * expr.c (store_constructor): Deal with zero-element CTORs.
18500 * match.pd (bit_field_ref <vector CTOR>): Make sure to
18501 produce vector constants when possible.
18503 2021-04-26 Richard Biener <rguenther@suse.de>
18505 * tree-complex.c: Include gimple-fold.h.
18506 (expand_complex_addition): Use gimple_build.
18507 (expand_complex_multiplication_components): Likewise.
18508 (expand_complex_multiplication): Likewise.
18509 (expand_complex_div_straight): Likewise.
18510 (expand_complex_div_wide): Likewise.
18511 (expand_complex_division): Likewise.
18512 (expand_complex_conjugate): Likewise.
18513 (expand_complex_comparison): Likewise.
18515 2021-04-26 Richard Biener <rguenther@suse.de>
18517 * tree-ssa-phiopt.c (two_value_replacement): Remove use
18518 of legacy gimplify_buildN API.
18520 2021-04-26 Richard Biener <rguenther@suse.de>
18522 PR tree-optimization/99473
18523 * tree-ssa-phiopt.c (cond_store_replacement): Handle all
18526 2021-04-26 Richard Biener <rguenther@suse.de>
18528 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin):
18529 Use replace_call_with_value.
18531 2021-04-26 Richard Biener <rguenther@suse.de>
18533 * tree-ssa-propagate.h (valid_gimple_rhs_p): Remove.
18534 (update_gimple_call): Likewise.
18535 (update_call_from_tree): Likewise.
18536 * tree-ssa-propagate.c (valid_gimple_rhs_p): Remove.
18537 (valid_gimple_call_p): Likewise.
18538 (move_ssa_defining_stmt_for_defs): Likewise.
18539 (finish_update_gimple_call): Likewise.
18540 (update_gimple_call): Likewise.
18541 (update_call_from_tree): Likewise.
18542 (propagate_tree_value_into_stmt): Use replace_call_with_value.
18543 * gimple-fold.h (update_gimple_call): Declare.
18544 * gimple-fold.c (valid_gimple_rhs_p): Move here from
18545 tree-ssa-propagate.c.
18546 (update_gimple_call): Likewise.
18547 (valid_gimple_call_p): Likewise.
18548 (finish_update_gimple_call): Likewise, and simplify.
18549 (gimplify_and_update_call_from_tree): Implement
18550 update_call_from_tree functionality, avoid excessive
18551 push/pop_gimplify_context.
18552 (gimple_fold_builtin): Use only gimplify_and_update_call_from_tree.
18553 (gimple_fold_call): Likewise.
18554 * gimple-ssa-sprintf.c (try_substitute_return_value): Likewise.
18555 * tree-ssa-ccp.c (ccp_folder::fold_stmt): Likewise.
18556 (pass_fold_builtins::execute): Likewise.
18557 (optimize_stack_restore): Use replace_call_with_value.
18558 * tree-cfg.c (fold_loop_internal_call): Likewise.
18559 * tree-ssa-dce.c (maybe_optimize_arith_overflow): Use
18560 only gimplify_and_update_call_from_tree.
18561 * tree-ssa-strlen.c (handle_builtin_strlen): Likewise.
18562 (handle_builtin_strchr): Likewise.
18563 * tsan.c: Include gimple-fold.h instead of tree-ssa-propagate.h.
18565 2021-04-26 Jakub Jelinek <jakub@redhat.com>
18568 * vmsdbgout.c (ASM_OUTPUT_DEBUG_STRING, vmsdbgout_begin_block,
18569 vmsdbgout_end_block, lookup_filename, vmsdbgout_source_line): Remove
18572 2021-04-25 liuhongt <hongtao.liu@intel.com>
18575 * config/i386/i386-builtin.def (BDESC): Change the icode of
18576 the following builtins to CODE_FOR_nothing.
18577 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
18578 IX86_BUILTIN_PCMPEQB128, IX86_BUILTIN_PCMPEQW128,
18579 IX86_BUILTIN_PCMPEQD128, IX86_BUILTIN_PCMPEQQ,
18580 IX86_BUILTIN_PCMPEQB256, IX86_BUILTIN_PCMPEQW256,
18581 IX86_BUILTIN_PCMPEQD256, IX86_BUILTIN_PCMPEQQ256,
18582 IX86_BUILTIN_PCMPGTB128, IX86_BUILTIN_PCMPGTW128,
18583 IX86_BUILTIN_PCMPGTD128, IX86_BUILTIN_PCMPGTQ,
18584 IX86_BUILTIN_PCMPGTB256, IX86_BUILTIN_PCMPGTW256,
18585 IX86_BUILTIN_PCMPGTD256, IX86_BUILTIN_PCMPGTQ256.
18586 * config/i386/sse.md (avx2_eq<mode>3): Deleted.
18587 (sse2_eq<mode>3): Ditto.
18588 (sse4_1_eqv2di3): Ditto.
18589 (sse2_gt<mode>3): Rename to ..
18590 (*sse2_gt<mode>3): .. this.
18592 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
18595 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
18598 * config/darwin.c (darwin_binds_local_p): Assume that any
18599 public symbol might be interposed for PIC code. Update function
18600 header comment to reflect current Darwin capability.
18602 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
18605 * config/darwin.c (darwin_binds_local_p): Assume that any
18606 public symbol might be interposed for PIC code. Update function
18607 header comment to reflect current Darwin capability.
18609 2021-04-24 Richard Sandiford <richard.sandiford@arm.com>
18611 * doc/sourcebuild.texi: Document no-opts and any-opts target
18614 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
18616 * config/bpf/bpf.h (ASM_OUTPUT_ALIGNED_BSS): Use .type and .lcomm.
18618 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
18620 * config/bpf/bpf.h (FUNCTION_BOUNDARY): Set to 64.
18622 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
18625 * config/i386/i386-options.c (ix86_option_override_internal):
18626 Error out when -m96bit-long-double is used with 64bit targets.
18627 * config/i386/i386.md (*pushxf_rounded): Remove pattern.
18629 2021-04-23 Martin Liska <mliska@suse.cz>
18631 * lto-wrapper.c: Remove FIXME about usage of
18632 hardware_concurrency. The function is not on par with
18635 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
18638 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
18639 Copy operand 3 to operand 4. Use sse_reg_operand
18640 as operand 3 predicate.
18641 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): Ditto.
18642 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
18643 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
18644 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2):
18645 Copy operand 1 to operand 0.
18646 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): Ditto.
18647 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
18648 (LDX_ATOMIC/LDX_ATOMIC FP store peephole2 with mem blockage): Ditto.
18650 2021-04-23 Alex Coplan <alex.coplan@arm.com>
18652 PR rtl-optimization/100230
18653 * early-remat.c (early_remat::sort_candidates): Use delete[]
18654 instead of delete for array allocated with new[].
18656 2021-04-23 Richard Biener <rguenther@suse.de>
18658 * genmatch.c (lower_cond): Remove VEC_COND_EXPR special-casing.
18659 (capture_info::capture_info): Likewise.
18660 (capture_info::walk_match): Likewise.
18661 (expr::gen_transform): Likewise.
18662 (dt_simplify::gen_1): Likewise.
18663 * gimple-match-head.c (maybe_resimplify_conditional_op):
18664 Remove VEC_COND_EXPR special-casing.
18665 (gimple_simplify): Likewise.
18666 * gimple.c (gimple_could_trap_p_1): Adjust.
18667 * tree-ssa-pre.c (compute_avail): Allow VEC_COND_EXPR
18668 to participate in PRE.
18670 2021-04-23 Richard Biener <rguenther@suse.de>
18672 * cfganal.c (connect_infinite_loops_to_exit): First call
18673 add_noreturn_fake_exit_edges.
18674 * ipa-sra.c (process_scan_results): Do not call the now redundant
18675 add_noreturn_fake_exit_edges.
18676 * predict.c (tree_estimate_probability): Likewise.
18677 (rebuild_frequencies): Likewise.
18678 * store-motion.c (one_store_motion_pass): Likewise.
18680 2021-04-23 Richard Biener <rguenther@suse.de>
18682 PR tree-optimization/100222
18683 * predict.c (pass_profile::execute): Remove redundant call to
18684 mark_irreducible_loops.
18685 (report_predictor_hitrates): Likewise.
18687 2021-04-23 Richard Biener <rguenther@suse.de>
18689 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Avoid
18690 valid_gimple_rhs_p by instead gimplifying to one.
18692 2021-04-23 Richard Biener <rguenther@suse.de>
18694 PR tree-optimization/99971
18695 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
18696 Always use TBAA for loads.
18698 2021-04-23 liuhongt <hongtao.liu@intel.com>
18701 * config/i386/i386-options.c (ix86_option_override_internal):
18702 Clear MASK_AVX256_SPLIT_UNALIGNED_LOAD/STORE in x_target_flags
18703 when X86_TUNE_AVX256_UNALIGNED_LOAD/STORE_OPTIMAL is enabled
18704 by target attribute.
18706 2021-04-23 David Edelsohn <dje.gcc@gmail.com>
18708 * config/rs6000/aix71.h (PREFERRED_DEBUGGING_TYPE): Change to
18710 * config/rs6000/aix72.h (PREFERRED_DEBUGGING_TYPE): Same.
18712 2021-04-22 David Edelsohn <dje.gcc@gmail.com>
18714 * config.gcc (powerpc-ibm-aix6.*): Remove.
18715 * config/rs6000/aix61.h: Delete.
18717 2021-04-22 Martin Liska <mliska@suse.cz>
18719 PR testsuite/100159
18720 PR testsuite/100192
18721 * builtins.c (expand_builtin): Fix typos and missing comments.
18722 * dwarf2out.c (gen_subprogram_die): Likewise.
18723 (gen_struct_or_union_type_die): Likewise.
18725 2021-04-22 Uroš Bizjak <ubizjak@gmail.com>
18728 * config/i386/i386-expand.c (ix86_expand_convert_uns_sidf_sse):
18729 Remove the sign with FE_DOWNWARD, where x - x = -0.0.
18731 2021-04-21 Iain Sandoe <iain@sandoe.co.uk>
18733 * config/i386/darwin.h (TARGET_64BIT): Remove definition
18734 based on TARGET_ISA_64BIT.
18735 (TARGET_64BIT_P): Remove definition based on
18736 TARGET_ISA_64BIT_P().
18738 2021-04-21 Martin Liska <mliska@suse.cz>
18741 2021-04-21 Martin Liska <mliska@suse.cz>
18743 * lto-wrapper.c (cpuset_popcount): Remove.
18744 (init_num_threads): Remove and use hardware_concurrency.
18746 2021-04-21 Martin Liska <mliska@suse.cz>
18749 * main.c (main): Call toplev::finalize in CHECKING_P mode.
18750 * ipa-modref.c (ipa_modref_c_finalize): summaries are NULL
18751 when incremental LTO linking happens.
18753 2021-04-21 Martin Liska <mliska@suse.cz>
18755 * lto-wrapper.c (run_gcc): When -flto=jobserver is used, but the
18756 makeserver cannot be detected, then use -flto=N fallback.
18758 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
18760 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): When cross-compiling,
18761 default to yes for aarch64-linux-gnu.
18762 * configure: Regenerate.
18764 2021-04-21 Martin Liska <mliska@suse.cz>
18766 * lto-wrapper.c (cpuset_popcount): Remove.
18767 (init_num_threads): Remove and use hardware_concurrency.
18769 2021-04-21 Martin Liska <mliska@suse.cz>
18771 * config/i386/i386.c: Remove superfluous || TARGET_MACHO
18772 which remains to be '(... || 0)' and clang complains about it.
18773 * dwarf2out.c (AT_vms_delta): Declare conditionally.
18774 (add_AT_vms_delta): Likewise.
18775 * tree.c (fld_simplified_type): Use rather more common pattern
18776 for disabling of something (#if 0).
18777 (get_tree_code_name): Likewise.
18778 (verify_type_variant): Likewise.
18780 2021-04-21 Martin Liska <mliska@suse.cz>
18782 * config/i386/i386-expand.c (decide_alignment): Use newly named
18783 macro TARGET_CPU_P.
18784 * config/i386/i386.c (ix86_decompose_address): Likewise.
18785 (ix86_address_cost): Likewise.
18786 (ix86_lea_outperforms): Likewise.
18787 (ix86_avoid_lea_for_addr): Likewise.
18788 (ix86_add_stmt_cost): Likewise.
18789 * config/i386/i386.h (TARGET_*): Remove.
18790 (TARGET_CPU_P): New macro.
18791 * config/i386/i386.md: Use newly named macro TARGET_CPU_P.
18792 * config/i386/x86-tune-sched-atom.c (do_reorder_for_imul): Likewise.
18793 (swap_top_of_ready_list): Likewise.
18794 (ix86_atom_sched_reorder): Likewise.
18795 * config/i386/x86-tune-sched-bd.c (ix86_bd_has_dispatch): Likewise.
18796 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Likewise.
18798 2021-04-21 Martin Liska <mliska@suse.cz>
18800 * config/i386/i386-options.c (TARGET_EXPLICIT_NO_SAHF_P):
18802 (SET_TARGET_NO_SAHF): Likewise.
18803 (TARGET_EXPLICIT_PREFETCH_SSE_P): Likewise.
18804 (SET_TARGET_PREFETCH_SSE): Likewise.
18805 (TARGET_EXPLICIT_NO_TUNE_P): Likewise.
18806 (SET_TARGET_NO_TUNE): Likewise.
18807 (TARGET_EXPLICIT_NO_80387_P): Likewise.
18808 (SET_TARGET_NO_80387): Likewise.
18810 * config/i386/i386.h (TARGET_*): Remove.
18811 * opth-gen.awk: Generate new used macros.
18813 2021-04-21 Martin Liska <mliska@suse.cz>
18815 * config/i386/i386.h (PTA_*): Remove.
18816 (enum pta_flag): New.
18817 (DEF_PTA): Generate PTA_* values from i386-isa.def.
18818 * config/i386/i386-isa.def: New file.
18820 2021-04-21 Alex Coplan <alex.coplan@arm.com>
18823 * config/aarch64/aarch64-bti-insert.c (aarch64_bti_j_insn_p): New.
18824 (rest_of_insert_bti): Avoid inserting duplicate bti j insns for
18825 jump table targets.
18827 2021-04-21 H.J. Lu <hjl.tools@gmail.com>
18829 * config.gcc: Install mwaitintrin.h for i[34567]86-*-* and
18830 x86_64-*-* targets.
18831 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_MWAIT_SET):
18833 (OPTION_MASK_ISA2_MWAIT_UNSET): Likewise.
18834 (ix86_handle_option): Handle -mmwait.
18835 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
18836 Replace OPTION_MASK_ISA_SSE3 with OPTION_MASK_ISA2_MWAIT on
18837 __builtin_ia32_monitor and __builtin_ia32_mwait.
18838 * config/i386/i386-options.c (isa2_opts): Add -mmwait.
18839 (ix86_valid_target_attribute_inner_p): Likewise.
18840 (ix86_option_override_internal): Enable mwait/monitor
18841 instructions for -msse3.
18842 * config/i386/i386.h (TARGET_MWAIT): New.
18843 (TARGET_MWAIT_P): Likewise.
18844 * config/i386/i386.opt: Add -mmwait.
18845 * config/i386/mwaitintrin.h: New file.
18846 * config/i386/pmmintrin.h: Include <mwaitintrin.h>.
18847 * config/i386/sse.md (sse3_mwait): Replace TARGET_SSE3 with
18849 (@sse3_monitor_<mode>): Likewise.
18850 * config/i386/x86gprintrin.h: Include <mwaitintrin.h>.
18851 * doc/extend.texi: Document mwait target attribute.
18852 * doc/invoke.texi: Document -mmwait.
18854 2021-04-21 Martin Liska <mliska@suse.cz>
18856 * config/i386/i386-options.c (DEF_ENUM): Remove it.
18857 * config/i386/i386-opts.h (DEF_ENUM): Likewise.
18858 * config/i386/stringop.def (DEF_ENUM): Likewise.
18860 2021-04-21 Martin Liska <mliska@suse.cz>
18862 * tree-cfg.c (gimple_verify_flow_info): Use qD instead
18863 of print_generic_expr.
18865 2021-04-21 Jakub Jelinek <jakub@redhat.com>
18867 PR rtl-optimization/100148
18868 * cprop.c (constprop_register): Use next_nondebug_insn instead of
18871 2021-04-21 Martin Liska <mliska@suse.cz>
18874 * cgraphunit.c (cgraph_node::analyze): Remove duplicate
18875 free_dominance_info calls.
18877 2021-04-21 Richard Biener <rguenther@suse.de>
18879 * gimple-fold.c (maybe_fold_reference): Remove is_lhs
18880 parameter (and assume it to be false).
18881 (fold_gimple_assign): Adjust, remove all callers of
18882 maybe_fold_reference calling it with is_lhs true.
18883 (gimple_fold_call): Likewise.
18884 (fold_stmt_1): Likewise.
18886 2021-04-21 Richard Biener <rguenther@suse.de>
18888 * fold-const.c (pedantic_non_lvalue_loc): Remove.
18889 (fold_binary_loc): Adjust.
18890 (fold_ternary_loc): Likewise.
18892 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
18894 PR middle-end/100130
18895 * varasm.c (get_block_for_decl): Make sure that any use of the
18896 retain attribute matches the section's retain flag.
18897 (switch_to_section): Check for retain mismatches even when
18898 changing sections, but do not warn if the given decl is the
18899 section's named.decl.
18900 (output_object_block): Pass the first decl in the block (if any)
18901 to switch_to_section.
18903 2021-04-20 H.J. Lu <hjl.tools@gmail.com>
18905 * config/i386/i386-c.c (ix86_target_macros_internal): Define
18906 __CRC32__ for -mcrc32.
18907 * config/i386/i386-options.c (ix86_option_override_internal):
18908 Enable crc32 instruction for -msse4.2.
18909 * config/i386/i386.md (sse4_2_crc32<mode>): Remove TARGET_SSE4_2
18911 (sse4_2_crc32di): Likewise.
18912 * config/i386/ia32intrin.h: Use crc32 target option for CRC32
18915 2021-04-20 Segher Boessenkool <segher@kernel.crashing.org>
18918 * config/rs6000/rs6000.c (rs6000_machine_from_flags): Do not consider
18921 2021-04-20 Martin Liska <mliska@suse.cz>
18923 * doc/invoke.texi: Fix typo.
18924 * params.opt: Likewise.
18926 2021-04-20 Martin Liska <mliska@suse.cz>
18928 * doc/invoke.texi: Document new param.
18930 2021-04-19 Andrew MacLeod <amacleod@redhat.com>
18932 PR tree-optimization/100081
18933 * gimple-range-cache.h (ranger_cache): Inherit from gori_compute
18934 rather than gori_compute_cache.
18935 * gimple-range-gori.cc (is_gimple_logical_p): Move to top of file.
18936 (range_def_chain::m_logical_depth): New member.
18937 (range_def_chain::range_def_chain): Initialize m_logical_depth.
18938 (range_def_chain::get_def_chain): Don't build defchains through more
18939 than LOGICAL_LIMIT logical expressions.
18940 * params.opt (param_ranger_logical_depth): New.
18942 2021-04-19 Richard Earnshaw <rearnsha@arm.com>
18945 * config/arm/arm.c (arm_configure_build_target): Do not strip
18946 extended FPU/SIMD feature bits from the target ISA when -mfpu
18947 is specified (partial revert of r11-8168).
18949 2021-04-19 Thomas Schwinge <thomas@codesourcery.com>
18951 * params.opt (-param=openacc-kernels=): Add.
18952 * omp-oacc-kernels-decompose.cc
18953 (pass_omp_oacc_kernels_decompose::gate): Use it.
18954 * doc/invoke.texi (-fopenacc-kernels=@var{mode}): Move...
18955 (--param): ... here, 'openacc-kernels'.
18957 2021-04-19 Martin Liska <mliska@suse.cz>
18960 * gengtype.c (finish_root_table): Align function arguments
18961 in between declaration and definition.
18963 2021-04-19 Eric Botcazou <ebotcazou@adacore.com>
18965 * config/i386/winnt.c (i386_pe_seh_cold_init): Properly deal with
18966 frames larger than the SEH maximum frame size.
18968 2021-04-18 Segher Boessenkool <segher@kernel.crashing.org>
18970 PR rtl-optimization/99927
18971 * combine.c (distribute_notes) [REG_UNUSED]: If the register already
18972 is dead, just drop it.
18974 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
18977 * config/i386/winnt-d.c (TARGET_D_TEMPLATES_ALWAYS_COMDAT): Define.
18978 * doc/tm.texi: Regenerate.
18979 * doc/tm.texi.in (D language and ABI): Add @hook for
18980 TARGET_D_TEMPLATES_ALWAYS_COMDAT.
18982 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
18984 * config/darwin-d.c (darwin_d_handle_target_object_format): New
18986 (darwin_d_register_target_info): New function.
18987 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18988 * config/dragonfly-d.c (dragonfly_d_handle_target_object_format): New
18990 (dragonfly_d_register_target_info): New function.
18991 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18992 * config/freebsd-d.c (freebsd_d_handle_target_object_format): New
18994 (freebsd_d_register_target_info): New function.
18995 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18996 * config/glibc-d.c (glibc_d_handle_target_object_format): New
18998 (glibc_d_register_target_info): New function.
18999 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19000 * config/i386/i386-d.c (ix86_d_handle_target_object_format): New
19002 (ix86_d_register_target_info): Add ix86_d_handle_target_object_format
19003 as handler for objectFormat key.
19004 * config/i386/winnt-d.c (winnt_d_handle_target_object_format): New
19006 (winnt_d_register_target_info): New function.
19007 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19008 * config/netbsd-d.c (netbsd_d_handle_target_object_format): New
19010 (netbsd_d_register_target_info): New function.
19011 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19012 * config/openbsd-d.c (openbsd_d_handle_target_object_format): New
19014 (openbsd_d_register_target_info): New function.
19015 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19016 * config/pa/pa-d.c (pa_d_handle_target_object_format): New function.
19017 (pa_d_register_target_info): Add pa_d_handle_target_object_format as
19018 handler for objectFormat key.
19019 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_object_format): New
19021 (rs6000_d_register_target_info): Add
19022 rs6000_d_handle_target_object_format as handler for objectFormat key.
19023 * config/sol2-d.c (solaris_d_handle_target_object_format): New
19025 (solaris_d_register_target_info): New function.
19026 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19028 2021-04-16 Jakub Jelinek <jakub@redhat.com>
19031 * config/aarch64/aarch64.c (aarch64_function_arg_alignment): Change
19032 abi_break argument from bool * to unsigned *, store there the pre-GCC 9
19034 (aarch64_layout_arg, aarch64_gimplify_va_arg_expr): Adjust callers.
19035 (aarch64_function_arg_regno_p): Likewise. Only emit -Wpsabi note if
19036 the old and new alignment after applying MIN/MAX to it is different.
19038 2021-04-16 Tamar Christina <tamar.christina@arm.com>
19041 * config/aarch64/aarch64-sve.md (@aarch64_sve_trn1_conv<mode>): New.
19042 * config/aarch64/aarch64.c (aarch64_expand_sve_const_pred_trn): Use new
19044 * config/aarch64/iterators.md (UNSPEC_TRN1_CONV): New.
19046 2021-04-16 Bill Schmidt <wschmidt@linux.ibm.com>
19048 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Revise
19049 this section and its subsections.
19051 2021-04-16 Jakub Jelinek <jakub@redhat.com>
19054 * config/aarch64/aarch64.md (*neg_asr_si2_extr, *extrsi5_insn_di): New
19055 define_insn patterns.
19057 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
19059 PR rtl-optimization/98689
19060 * reg-notes.def (UNTYPED_CALL): New note.
19061 * combine.c (distribute_notes): Handle it.
19062 * emit-rtl.c (try_split): Likewise.
19063 * rtlanal.c (rtx_properties::try_to_add_insn): Likewise. Assume
19064 that calls with the note implicitly set all return value registers.
19065 * builtins.c (expand_builtin_apply): Add a REG_UNTYPED_CALL
19068 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
19070 PR rtl-optimization/99596
19071 * rtlanal.c (rtx_properties::try_to_add_insn): Don't add global
19072 register accesses for const calls. Assume that pure functions
19073 can only read from global registers. Ignore cases in which
19074 the stack pointer has been marked global.
19076 2021-04-16 Jakub Jelinek <jakub@redhat.com>
19079 * tree-vect-loop.c (vect_transform_loop): Don't remove just
19080 dead scalar .MASK_LOAD calls, but also dead .COND_* calls - replace
19081 them by their last argument.
19083 2021-04-15 Martin Liska <mliska@suse.cz>
19085 * doc/invoke.texi: Other params don't use it, remove it.
19087 2021-04-15 Richard Biener <rguenther@suse.de>
19089 * gimple-builder.h: Add deprecation note.
19091 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
19094 * attribs.h (restrict_type_identity_attributes_to): Declare.
19095 * attribs.c (restrict_type_identity_attributes_to): New function.
19097 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
19100 * attribs.h (affects_type_identity_attributes): Declare.
19101 * attribs.c (remove_attributes_matching): New function.
19102 (affects_type_identity_attributes): Likewise.
19104 2021-04-15 Jakub Jelinek <jakub@redhat.com>
19107 * config/aarch64/aarch64.md (*<LOGICAL:optab>_<SHIFT:optab><mode>3):
19108 Add combine splitters for *<LOGICAL:optab>_ashl<mode>3 with
19109 ZERO_EXTEND, SIGN_EXTEND or AND.
19111 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
19113 PR rtl-optimization/99929
19114 * rtl.h (same_vector_encodings_p): New function.
19115 * cse.c (exp_equiv_p): Check that CONST_VECTORs have the same encoding.
19116 * cselib.c (rtx_equal_for_cselib_1): Likewise.
19117 * jump.c (rtx_renumbered_equal_p): Likewise.
19118 * lra-constraints.c (operands_match_p): Likewise.
19119 * reload.c (operands_match_p): Likewise.
19120 * rtl.c (rtx_equal_p_cb, rtx_equal_p): Likewise.
19122 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
19124 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Print
19125 more information about variable-length CONST_VECTORs.
19127 2021-04-14 Vladimir N. Makarov <vmakarov@redhat.com>
19129 PR rtl-optimization/100066
19130 * lra-constraints.c (split_reg): Check paradoxical_subreg_p for
19131 ordered modes when choosing splitting mode for hard reg.
19133 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
19136 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector_sel):
19138 (aarch64_expand_sve_const_vector): Use it for nelts_per_pattern==2.
19140 2021-04-14 Andreas Krebbel <krebbel@linux.ibm.com>
19142 * config/s390/s390-builtins.def (O_M5, O_M12, ...): Add new macros
19143 for mask operand types.
19144 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
19145 (s390_vec_permi_dbl, s390_vpdi): Use the M5 type for the immediate
19147 (s390_vec_msum_u128, s390_vmslg): Use the M12 type for the
19149 * config/s390/s390.c (s390_const_operand_ok): Check the new
19150 operand types and generate a list of valid values.
19152 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
19154 * doc/tm.texi: Regenerate.
19155 * doc/tm.texi.in (D language and ABI): Add @hook for
19156 TARGET_D_REGISTER_OS_TARGET_INFO.
19158 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
19160 * config/aarch64/aarch64-d.c (aarch64_d_handle_target_float_abi): New
19162 (aarch64_d_register_target_info): New function.
19163 * config/aarch64/aarch64-protos.h (aarch64_d_register_target_info):
19165 * config/aarch64/aarch64.h (TARGET_D_REGISTER_CPU_TARGET_INFO):
19167 * config/arm/arm-d.c (arm_d_handle_target_float_abi): New function.
19168 (arm_d_register_target_info): New function.
19169 * config/arm/arm-protos.h (arm_d_register_target_info): Declare.
19170 * config/arm/arm.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19171 * config/i386/i386-d.c (ix86_d_handle_target_float_abi): New function.
19172 (ix86_d_register_target_info): New function.
19173 * config/i386/i386-protos.h (ix86_d_register_target_info): Declare.
19174 * config/i386/i386.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19175 * config/mips/mips-d.c (mips_d_handle_target_float_abi): New function.
19176 (mips_d_register_target_info): New function.
19177 * config/mips/mips-protos.h (mips_d_register_target_info): Declare.
19178 * config/mips/mips.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19179 * config/pa/pa-d.c (pa_d_handle_target_float_abi): New function.
19180 (pa_d_register_target_info): New function.
19181 * config/pa/pa-protos.h (pa_d_register_target_info): Declare.
19182 * config/pa/pa.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19183 * config/riscv/riscv-d.c (riscv_d_handle_target_float_abi): New
19185 (riscv_d_register_target_info): New function.
19186 * config/riscv/riscv-protos.h (riscv_d_register_target_info): Declare.
19187 * config/riscv/riscv.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19188 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_float_abi): New
19190 (rs6000_d_register_target_info): New function.
19191 * config/rs6000/rs6000-protos.h (rs6000_d_register_target_info):
19193 * config/rs6000/rs6000.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19194 * config/s390/s390-d.c (s390_d_handle_target_float_abi): New function.
19195 (s390_d_register_target_info): New function.
19196 * config/s390/s390-protos.h (s390_d_register_target_info): Declare.
19197 * config/s390/s390.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19198 * config/sparc/sparc-d.c (sparc_d_handle_target_float_abi): New
19200 (sparc_d_register_target_info): New function.
19201 * config/sparc/sparc-protos.h (sparc_d_register_target_info): Declare.
19202 * config/sparc/sparc.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19203 * doc/tm.texi: Regenerate.
19204 * doc/tm.texi.in (D language and ABI): Add @hook for
19205 TARGET_D_REGISTER_CPU_TARGET_INFO.
19207 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
19209 * config/i386/i386-d.c (ix86_d_has_stdcall_convention): New function.
19210 * config/i386/i386-protos.h (ix86_d_has_stdcall_convention): Declare.
19211 * config/i386/i386.h (TARGET_D_HAS_STDCALL_CONVENTION): Define.
19212 * doc/tm.texi: Regenerate.
19213 * doc/tm.texi.in (D language and ABI): Add @hook for
19214 TARGET_D_HAS_STDCALL_CONVENTION.
19216 2021-04-14 Richard Biener <rguenther@suse.de>
19218 * tree-cfg.c (verify_gimple_assign_ternary): Verify that
19219 VEC_COND_EXPRs have a gimple_val condition.
19220 * tree-ssa-propagate.c (valid_gimple_rhs_p): VEC_COND_EXPR
19221 can no longer have a GENERIC condition.
19223 2021-04-14 Richard Earnshaw <rearnsha@arm.com>
19226 * config/arm/arm.c (arm_configure_build_target): Strip isa_all_fpbits
19227 from the isa_delta when -mfpu has been used.
19228 (arm_options_perform_arch_sanity_checks): It's the architecture that
19229 lacks an FPU not the processor.
19231 2021-04-13 Richard Biener <rguenther@suse.de>
19233 PR tree-optimization/100053
19234 * tree-ssa-sccvn.c (vn_nary_op_get_predicated_value): Do
19235 not use optimistic dominance queries for backedges to validate
19237 (dominated_by_p_w_unex): Add parameter to ignore executable
19238 state on backedges.
19239 (rpo_elim::eliminate_avail): Adjust.
19241 2021-04-13 Jakub Jelinek <jakub@redhat.com>
19244 * config/aarch64/aarch64.md (*aarch64_bfxil<mode>_extr,
19245 *aarch64_bfxilsi_extrdi): New define_insn patterns.
19247 2021-04-13 Jakub Jelinek <jakub@redhat.com>
19250 * simplify-rtx.c (simplify_immed_subreg): For MODE_COMPOSITE_P
19251 outermode, return NULL if the result doesn't encode back to the
19252 original byte sequence.
19253 (simplify_gen_subreg): Don't create SUBREGs from constants to
19254 MODE_COMPOSITE_P outermode.
19256 2021-04-12 Jakub Jelinek <jakub@redhat.com>
19258 PR rtl-optimization/99905
19259 * combine.c (expand_compound_operation): If pos + len > modewidth,
19260 perform the right shift by pos in inner_mode and then convert to mode,
19261 instead of trying to simplify a shift of rtx with inner_mode by pos
19262 as if it was a shift in mode.
19264 2021-04-12 Jakub Jelinek <jakub@redhat.com>
19267 * combine.c (simplify_and_const_int_1): Don't optimize varop
19268 away if it has side-effects.
19270 2021-04-12 Martin Liska <mliska@suse.cz>
19272 * doc/extend.texi: Escape @smallexample content.
19274 2021-04-12 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
19276 * config/s390/s390.md ("*movdi_31", "*movdi_64"): Add
19277 alternative in order to load a DFP zero.
19279 2021-04-12 Martin Liska <mliska@suse.cz>
19281 * doc/extend.texi: Be more precise in documentation
19282 of symver attribute.
19284 2021-04-12 Martin Liska <mliska@suse.cz>
19287 * gimplify.c (gimplify_expr): Right now, we unpoison all
19288 variables before a goto <dest>. We should not do it if we are
19291 2021-04-12 Cui,Lili <lili.cui@intel.com>
19293 * common/config/i386/cpuinfo.h (get_intel_cpu): Handle
19295 * common/config/i386/i386-common.c (processor_names): Add
19297 (processor_alias_table): Add rocketlake.
19298 * common/config/i386/i386-cpuinfo.h (processor_subtypes): Add
19299 INTEL_COREI7_ROCKETLAKE.
19300 * config.gcc: Add -march=rocketlake.
19301 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19303 * config/i386/i386-options.c (m_ROCKETLAKE) : Define.
19304 (processor_cost_table): Add rocketlake cost.
19305 * config/i386/i386.h (ix86_size_cost) : Define
19307 (processor_type) : Add PROCESSOR_ROCKETLAKE.
19308 (PTA_ROCKETLAKE): Ditto.
19309 * doc/extend.texi: Add rocketlake.
19310 * doc/invoke.texi: Add rocketlake.
19312 2021-04-12 Cui,Lili <lili.cui@intel.com>
19314 * config/i386/i386.h (PTA_ALDERLAKE): Change alderlake ISA list.
19315 * config/i386/i386-options.c (m_CORE_AVX2): Add m_ALDERLAKE.
19316 * common/config/i386/cpuinfo.h (get_intel_cpu): Add AlderLake model.
19317 * doc/invoke.texi: Change alderlake ISA list.
19319 2021-04-11 Hafiz Abid Qadeer <abidh@codesourcery.com>
19321 PR middle-end/98088
19322 * omp-expand.c (expand_oacc_collapse_init): Update condition in
19325 2021-04-10 H.J. Lu <hjl.tools@gmail.com>
19328 * config/i386/serializeintrin.h (_serialize): Defined as macro.
19330 2021-04-10 Jakub Jelinek <jakub@redhat.com>
19333 * expr.c (expand_expr_addr_expr_1): Test is_global_var rather than
19334 just TREE_STATIC on COMPOUND_LITERAL_EXPR_DECLs.
19336 2021-04-10 Jakub Jelinek <jakub@redhat.com>
19338 PR middle-end/99989
19339 * gimple-ssa-warn-alloca.c
19340 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit to
19341 0 with integer precision unconditionally.
19343 2021-04-10 Jakub Jelinek <jakub@redhat.com>
19345 PR rtl-optimization/98601
19346 * rtlanal.c (rtx_addr_can_trap_p_1): Allow in assert unknown size
19347 not just for BLKmode, but also for VOIDmode. For STRICT_ALIGNMENT
19348 unaligned_mems handle VOIDmode like BLKmode.
19350 2021-04-10 Jan Hubicka <hubicka@ucw.cz>
19353 * tree.c (free_lang_data_in_decl): Do not release body of
19354 declare_variant_alt.
19356 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
19358 * config/aarch64/aarch64.c (aarch64_option_restore): If the
19359 architecture was specified explicitly and the tuning wasn't,
19360 tune for the architecture rather than the configured default CPU.
19362 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
19364 * config/aarch64/aarch64.md (tlsdesc_small_sve_<mode>): Use X30
19365 as the temporary register.
19367 2021-04-09 Martin Liska <mliska@suse.cz>
19369 * doc/extend.texi: Move non-target attributes on the top level.
19371 2021-04-09 Martin Liska <mliska@suse.cz>
19373 * doc/invoke.texi: Document minimum and maximum value of the
19374 argument for both supported compression algorithms.
19376 2021-04-08 David Edelsohn <dje.gcc@gmail.com>
19378 * config/rs6000/rs6000.c (rs6000_xcoff_select_section): Select
19379 TLS BSS before TLS data.
19380 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .comm.
19382 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
19384 * doc/sourcebuild.texi (stdint_types_mbig_endian): Document.
19386 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
19388 * match.pd: Extend vec_cond folds to handle shifts.
19390 2021-04-08 Maciej W. Rozycki <macro@orcam.me.uk>
19392 * config/vax/vax.md: Fix comment for `*bit<mode>' pattern's
19395 2021-04-08 Alex Coplan <alex.coplan@arm.com>
19398 * config/arm/iterators.md (MVE_vecs): New.
19399 (V_elem): Also handle V2DF.
19400 * config/arm/mve.md (*mve_mov<mode>): Rename to ...
19401 (*mve_vdup<mode>): ... this. Remove second alternative since
19402 vec_duplicate of const_int is not canonical RTL, and we don't
19403 want to match symbol_refs.
19404 (*mve_vec_duplicate<mode>): Delete (pattern is redundant).
19406 2021-04-08 Xionghu Luo <luoxhu@linux.ibm.com>
19408 * fold-const.c (fold_single_bit_test): Fix typo.
19409 * print-rtl.c (print_rtx_insn_vec): Call print_rtl_single
19412 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
19414 PR tree-optimization/97513
19415 * tree-vect-slp.c (vect_add_slp_permutation): New function,
19417 (vectorizable_slp_permutation): ...here. Detect cases in which
19418 all VEC_PERM_EXPRs are guaranteed to have the same stepped
19419 permute vector and only generate one permute vector for that case.
19420 Extend that case to handle variable-length vectors.
19422 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
19424 PR tree-optimization/99873
19425 * tree-vect-slp.c (vect_slp_prefer_store_lanes_p): New function.
19426 (vect_build_slp_instance): Don't split store groups that could
19427 use IFN_STORE_LANES.
19429 2021-04-07 Jakub Jelinek <jakub@redhat.com>
19432 * varasm.c (output_constant_pool_contents): Don't strip name encoding
19433 from XSTR (desc->sym, 0) or from label before passing those to
19436 2021-04-07 Richard Biener <rguenther@suse.de>
19438 PR tree-optimization/99954
19439 * tree-loop-distribution.c: Include tree-affine.h.
19440 (generate_memcpy_builtin): Try using tree-affine to prove
19442 (loop_distribution::classify_builtin_ldst): Always classify
19445 2021-04-07 Richard Biener <rguenther@suse.de>
19447 PR tree-optimization/99947
19448 * tree-vect-loop.c (vectorizable_induction): Pre-allocate
19449 steps vector to avoid pushing elements from the reallocated
19452 2021-04-07 Richard Biener <rguenther@suse.de>
19454 * tree-ssa-sccvn.h (print_vn_reference_ops): Declare.
19455 * tree-ssa-pre.c (print_pre_expr): Factor out VN reference operand
19457 * tree-ssa-sccvn.c (print_vn_reference_ops): ... into this new
19459 (debug_vn_reference_ops): New.
19461 2021-04-07 Bin Cheng <bin.cheng@linux.alibaba.com>
19463 PR tree-optimization/98736
19464 * tree-loop-distribution.c
19465 * (loop_distribution::bb_top_order_init):
19466 Compute RPO with programing order preserved by calling function
19467 rev_post_order_and_mark_dfs_back_seme.
19469 2021-04-06 Vladimir N. Makarov <vmakarov@redhat.com>
19472 * lra-constraints.c (split_reg): Don't check paradoxical_subreg_p.
19473 * lra-lives.c (clear_sparseset_regnos, regnos_in_sparseset_p): New
19475 (process_bb_lives): Don't update biggest mode of hard reg for
19476 implicit in multi-register group. Use the new functions for
19477 updating dead_set and unused_set by register notes.
19479 2021-04-06 Xianmiao Qu <xianmiao_qu@c-sky.com>
19481 * config/csky/csky_pipeline_ck802.md : Use insn reservation name
19484 2021-04-06 H.J. Lu <hjl.tools@gmail.com>
19486 * config/i386/x86-tune-costs.h (skylake_memcpy): Updated.
19487 (skylake_memset): Likewise.
19488 (skylake_cost): Change CLEAR_RATIO to 17.
19489 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
19490 Replace m_CANNONLAKE, m_ICELAKE_CLIENT, m_ICELAKE_SERVER,
19491 m_TIGERLAKE and m_SAPPHIRERAPIDS with m_SKYLAKE and m_CORE_AVX512.
19493 2021-04-06 Richard Biener <rguenther@suse.de>
19495 PR tree-optimization/99880
19496 * tree-vect-loop.c (maybe_set_vectorized_backedge_value): Only
19497 set vectorized defs of relevant PHIs.
19499 2021-04-06 Richard Biener <rguenther@suse.de>
19501 PR tree-optimization/99924
19502 * tree-vect-slp.c (vect_bb_partition_graph_r): Do not mark
19503 nodes w/o scalar stmts as visited.
19505 2021-04-06 Alex Coplan <alex.coplan@arm.com>
19508 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Also use base
19509 PCS for [su]fix_optab.
19511 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
19513 * config/darwin.c (machopic_legitimize_pic_address): Check
19514 that the current pic register is one of the hard reg set
19515 before setting liveness.
19517 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
19519 * config/darwin.c (machopic_legitimize_pic_address): Fix
19520 whitespace, remove unused code.
19522 2021-04-03 Jakub Jelinek <jakub@redhat.com>
19524 PR tree-optimization/99882
19525 * gimple-ssa-store-merging.c (bswap_view_convert): Handle val with
19528 2021-04-03 Jakub Jelinek <jakub@redhat.com>
19530 PR rtl-optimization/99863
19531 * dse.c (replace_read): Drop regs_live argument. Instead of
19532 regs_live, use store_insn->fixed_regs_live if non-NULL,
19533 otherwise punt if insns sequence clobbers or sets any hard
19536 2021-04-03 Jakub Jelinek <jakub@redhat.com>
19539 * targhooks.h (default_print_patchable_function_entry_1): Declare.
19540 * targhooks.c (default_print_patchable_function_entry_1): New function,
19541 copied from default_print_patchable_function_entry with an added flags
19543 (default_print_patchable_function_entry): Rewritten into a small
19544 wrapper around default_print_patchable_function_entry_1.
19545 * config/rs6000/rs6000.c (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
19547 (rs6000_print_patchable_function_entry): New function.
19549 2021-04-02 Eric Botcazou <ebotcazou@adacore.com>
19551 * doc/invoke.texi (fdelete-dead-exceptions): Minor tweak.
19553 2021-04-01 Jason Merrill <jason@redhat.com>
19556 * common.opt: Document v15 and v16.
19558 2021-04-01 Richard Biener <rguenther@suse.de>
19560 PR tree-optimization/99863
19561 * gimplify.c (gimplify_init_constructor): Recompute vector
19564 2021-04-01 Jakub Jelinek <jakub@redhat.com>
19566 * doc/extend.texi (symver attribute): Fix up syntax errors
19569 2021-04-01 Jakub Jelinek <jakub@redhat.com>
19571 PR tree-optimization/96573
19572 * gimple-ssa-store-merging.c (init_symbolic_number): Handle
19573 also pointer types.
19575 2021-04-01 Richard Biener <rguenther@suse.de>
19577 PR tree-optimization/99856
19578 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Promote
19579 precision to vector element precision.
19581 2021-04-01 Martin Jambor <mjambor@suse.cz>
19583 PR tree-optimization/97009
19584 * tree-sra.c (access_or_its_child_written): New function.
19585 (propagate_subaccesses_from_rhs): Use it instead of a simple grp_write
19588 2021-03-31 Jan Hubicka <hubicka@ucw.cz>
19591 * cif-code.def (USES_COMDAT_LOCAL): Make CIF_FINAL_NORMAL.
19593 2021-03-31 Pat Haugen <pthaugen@linux.ibm.com>
19596 * config/rs6000/altivec.md (xxspltiw_v4si, xxspltiw_v4sf_inst,
19597 xxspltidp_v2df_inst, xxsplti32dx_v4si_inst, xxsplti32dx_v4sf_inst,
19598 xxblend_<mode>, xxpermx_inst, xxeval): Mark prefixed.
19599 * config/rs6000/mma.md (mma_<vvi4i4i8>, mma_<avvi4i4i8>,
19600 mma_<vvi4i4i2>, mma_<avvi4i4i2>, mma_<vvi4i4>, mma_<avvi4i4>,
19601 mma_<pvi4i2>, mma_<apvi4i2>, mma_<vvi4i4i4>, mma_<avvi4i4i4>):
19603 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Adjust test.
19604 * config/rs6000/rs6000.md (define_attr "maybe_prefixed"): New.
19605 (define_attr "prefixed"): Update initializer.
19607 2021-03-31 Jakub Jelinek <jakub@redhat.com>
19610 * dwarf2out.c (debug_ranges_dwo_section): New variable.
19611 (DW_RANGES_IDX_SKELETON): Define.
19612 (struct dw_ranges): Add begin_entry and end_entry members.
19613 (DEBUG_DWO_RNGLISTS_SECTION): Define.
19614 (add_ranges_num): Adjust r initializer for addition of *_entry
19616 (add_ranges_by_labels): For -gsplit-dwarf and force_direct,
19617 set idx to DW_RANGES_IDX_SKELETON.
19618 (use_distinct_base_address_for_range): New function.
19619 (index_rnglists): Don't set r->idx if it is equal to
19620 DW_RANGES_IDX_SKELETON. Initialize r->begin_entry and
19621 r->end_entry for -gsplit-dwarf if those will be needed by
19623 (output_rnglists): Add DWO argument. If true, switch to
19624 debug_ranges_dwo_section rather than debug_ranges_section.
19625 Adjust l1/l2 label indexes. Only output the offset table when
19626 dwo is true and don't include in there the skeleton range
19627 entry if present. For -gsplit-dwarf, skip ranges that belong
19628 to the other rnglists section. Change return type from void
19629 to bool and return true if there are any range entries for
19630 the other section. For dwarf_split_debug_info use
19631 DW_RLE_startx_endx, DW_RLE_startx_length and DW_RLE_base_addressx
19632 entries instead of DW_RLE_start_end, DW_RLE_start_length and
19633 DW_RLE_base_address. Use use_distinct_base_address_for_range.
19634 (init_sections_and_labels): Initialize debug_ranges_dwo_section
19635 if -gsplit-dwarf and DWARF >= 5. Adjust ranges_section_label
19636 and range_base_label indexes.
19637 (dwarf2out_finish): Call index_rnglists earlier before finalizing
19638 .debug_addr. Never emit DW_AT_rnglists_base attribute. For
19639 -gsplit-dwarf and DWARF >= 5 call output_rnglists up to twice
19640 with different dwo arguments.
19641 (dwarf2out_c_finalize): Clear debug_ranges_dwo_section.
19643 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19645 PR tree-optimization/98268
19646 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Call
19647 recompute_tree_invariant_for_addr_expr after successfully
19648 folding a TARGET_MEM_REF that occurs inside an ADDR_EXPR.
19650 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19652 PR tree-optimization/99726
19653 * tree-data-ref.c (create_intersect_range_checks_index): Bail
19654 out if there is more than one access function SCEV for the loop
19657 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19659 PR rtl-optimization/97141
19660 PR rtl-optimization/98726
19661 * emit-rtl.c (valid_for_const_vector_p): Return true for
19663 * rtx-vector-builder.h (rtx_vector_builder::step): Return a
19664 poly_wide_int instead of a wide_int.
19665 (rtx_vector_builder::apply_set): Take a poly_wide_int instead
19667 * rtx-vector-builder.c (rtx_vector_builder::apply_set): Likewise.
19668 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Return
19669 false for CONST_VECTORs that cannot be forced to memory.
19670 * config/aarch64/aarch64-simd.md (mov<mode>): If a CONST_VECTOR
19671 is too complex to force to memory, build it up from individual
19674 2021-03-31 Jan Hubicka <jh@suse.cz>
19677 * cgraph.c (cgraph_node::release_body): Fix overactive check.
19679 2021-03-31 Christophe Lyon <christophe.lyon@linaro.org>
19682 * config/arm/vec-common.md (mul<mode>3): Disable on iwMMXT, expect
19685 2021-03-31 H.J. Lu <hjl.tools@gmail.com>
19687 * config/i386/i386-expand.c (expand_set_or_cpymem_via_rep):
19688 For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, don't convert QImode
19690 (decide_alg): For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, use
19691 "rep movsb/stosb" only for known sizes.
19692 * config/i386/i386-options.c (processor_cost_table): Use Ice
19693 Lake cost for Cannon Lake, Ice Lake, Tiger Lake, Sapphire
19694 Rapids and Alder Lake.
19695 * config/i386/i386.h (TARGET_PREFER_KNOWN_REP_MOVSB_STOSB): New.
19696 * config/i386/x86-tune-costs.h (icelake_memcpy): New.
19697 (icelake_memset): Likewise.
19698 (icelake_cost): Likewise.
19699 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
19702 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19705 * config/aarch64/aarch64.c
19706 (aarch64_vectorize_preferred_vector_alignment): Query the size
19707 of the provided SVE vector; do not assume that all SVE vectors
19708 have the same size.
19710 2021-03-31 Jan Hubicka <jh@suse.cz>
19713 * cgraph.c (cgraph_node::release_body): Remove all callers and
19715 * cgraphclones.c (cgraph_node::materialize_clone): Do not do it here.
19716 * cgraphunit.c (cgraph_node::expand): And here.
19718 2021-03-31 Martin Liska <mliska@suse.cz>
19720 * ipa-modref.c (analyze_ssa_name_flags): Fix coding style
19721 and one negated condition.
19723 2021-03-31 Jakub Jelinek <jakub@redhat.com>
19724 Richard Sandiford <richard.sandiford@arm.com>
19727 * config/aarch64/aarch64.md (*add<mode>3_poly_1): Swap Uai and Uav
19728 constraints on operands[2] and similarly 0 and rk constraints
19729 on operands[1] corresponding to that.
19731 2021-03-31 Jakub Jelinek <jakub@redhat.com>
19734 * configure.ac (HAVE_LD_BROKEN_PE_DWARF5): New AC_DEFINE if PECOFF
19735 linker doesn't support DWARF sections new in DWARF5.
19736 * config/i386/i386-options.c (ix86_option_override_internal): Default
19737 to dwarf_version 4 if HAVE_LD_BROKEN_PE_DWARF5 for TARGET_PECOFF
19739 * config.in: Regenerated.
19740 * configure: Regenerated.
19742 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19745 * config/aarch64/aarch64.c (aarch64_analyze_loop_vinfo): Check for
19746 available issue_info before using it.
19748 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19751 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): Do not allow zero
19754 2021-03-30 Xionghu Luo <luoxhu@linux.ibm.com>
19757 * config/rs6000/altivec.md (altivec_lvsl_reg): Change to ...
19758 (altivec_lvsl_reg_<mode>): ... this.
19759 (altivec_lvsr_reg): Change to ...
19760 (altivec_lvsr_reg_<mode>): ... this.
19761 * config/rs6000/predicates.md (vec_set_index_operand): New.
19762 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
19763 Enable 32bit variable vec_insert for all TARGET_VSX.
19764 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
19765 Enable 32bit variable vec_insert for p9 and above.
19766 (rs6000_expand_vector_set_var_p8): Rename to ...
19767 (rs6000_expand_vector_set_var_p7): ... this.
19768 (rs6000_expand_vector_set): Use TARGET_VSX and adjust assert
19770 * config/rs6000/vector.md (vec_set<mode>): Use vec_set_index_operand.
19771 * config/rs6000/vsx.md (xl_len_r): Use gen_altivec_lvsl_reg_di and
19772 gen_altivec_lvsr_reg_di.
19774 2021-03-30 H.J. Lu <hjl.tools@gmail.com>
19777 * config/i386/ia32intrin.h (__rdtsc): Defined as macro.
19778 (__rdtscp): Likewise.
19780 2021-03-30 Tamar Christina <tamar.christina@arm.com>
19782 PR tree-optimization/99825
19783 * tree-vect-slp-patterns.c (vect_check_evenodd_blend):
19784 Reject non-mult 2 lanes.
19786 2021-03-30 Richard Earnshaw <rearnsha@arm.com>
19789 * config/arm/arm.c (arm_file_start): Fix emission of
19790 Tag_ABI_VFP_args attribute.
19792 2021-03-30 Richard Biener <rguenther@suse.de>
19794 PR tree-optimization/99824
19795 * stor-layout.c (set_min_and_max_values_for_integral_type):
19796 Assert the precision is within the bounds of
19797 WIDE_INT_MAX_PRECISION.
19798 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
19799 the outermost component ref only to lower the access size
19800 and initialize that from the access type.
19802 2021-03-30 Richard Sandiford <richard.sandiford@arm.com>
19805 * config/aarch64/aarch64.md (mov<mode>): Pass multi-instruction
19806 CONST_INTs to aarch64_expand_mov_immediate when called after RA.
19808 2021-03-30 Mihailo Stojanovic <mihailo.stojanovic@typhoon-hil.com>
19810 * config/aarch64/aarch64.md
19811 (<optab>_trunc<fcvt_target><GPI:mode>2): Set the "arch"
19812 attribute to disambiguate between SIMD and FP variants of the
19815 2021-03-29 Jan Hubicka <hubicka@ucw.cz>
19817 * ipa-modref.c (merge_call_lhs_flags): Correct handling of deref.
19818 (analyze_ssa_name_flags): Fix typo in comment.
19820 2021-03-29 Alex Coplan <alex.coplan@arm.com>
19823 * config/aarch64/aarch64-sve-builtins.cc
19824 (function_builder::add_function): Add placeholder_p argument, use
19825 placeholder decls if this is set.
19826 (function_builder::add_unique_function): Instead of conditionally adding
19827 direct overloads, unconditionally add either a direct overload or a
19829 (function_builder::add_overloaded_function): Set placeholder_p if we're
19830 using C++ overloads. Use the obstack for string storage instead
19831 of relying on the tree nodes.
19832 (function_builder::add_overloaded_functions): Don't return early for
19833 m_direct_overloads: we need to add placeholders.
19834 * config/aarch64/aarch64-sve-builtins.h
19835 (function_builder::add_function): Add placeholder_p argument.
19837 2021-03-29 Richard Biener <rguenther@suse.de>
19839 PR tree-optimization/99807
19840 * tree-vect-slp.c (vect_slp_analyze_node_operations_1): Move
19841 assert below VEC_PERM handling.
19843 2021-03-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19846 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>): Use
19847 aarch64_simd_or_scalar_imm_zero to match zeroes. Remove pattern
19848 matching const_int 0.
19849 (move_lo_quad_internal_be_<mode>): Likewise.
19850 (move_lo_quad_<mode>): Update for the above.
19851 * config/aarch64/iterators.md (VQ_2E): Delete.
19853 2021-03-29 Jakub Jelinek <jakub@redhat.com>
19855 PR tree-optimization/99777
19856 * fold-const.c (extract_muldiv_1): For conversions, punt on casts from
19857 types other than scalar integral types.
19859 2021-03-28 David Edelsohn <dje.gcc@gmail.com>
19861 * config/rs6000/rs6000.c (rs6000_output_dwarf_dtprel): Do not add
19862 XCOFF TLS reloc decorations.
19864 2021-03-28 Gerald Pfeifer <gerald@pfeifer.com>
19866 * doc/analyzer.texi (Analyzer Internals): Update link to
19867 "A Memory Model for Static Analysis of C Programs".
19869 2021-03-26 David Edelsohn <dje.gcc@gmail.com>
19871 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Call function.
19872 * config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align):
19874 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align): New.
19875 (rs6000_special_round_type_align): Recursively check innermost first
19878 2021-03-26 Jakub Jelinek <jakub@redhat.com>
19881 * dwarf2out.h (struct dw_fde_node): Add rule18 member.
19882 * dwarf2cfi.c (dwarf2out_frame_debug_expr): When handling (set hfp sp)
19883 assignment with drap_reg active, queue reg save for hfp with offset 0
19884 and flush queued reg saves. When handling a push with rule18,
19885 defer queueing reg save for hfp and just assert the offset is 0.
19886 (scan_trace): Assert that fde->rule18 is false.
19888 2021-03-26 Vladimir Makarov <vmakarov@redhat.com>
19891 * ira-costs.c (record_reg_classes): Put case with
19892 CT_RELAXED_MEMORY adjacent to one with CT_MEMORY.
19893 * ira.c (ira_setup_alts): Ditto.
19894 * lra-constraints.c (process_alt_operands): Ditto.
19895 * recog.c (asm_operand_ok): Ditto.
19896 * reload.c (find_reloads): Ditto.
19898 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19900 * config/aarch64/aarch64-protos.h
19901 (cpu_addrcost_table::post_modify_ld3_st3): New member variable.
19902 (cpu_addrcost_table::post_modify_ld4_st4): Likewise.
19903 * config/aarch64/aarch64.c (generic_addrcost_table): Update
19904 accordingly, using the same costs as for post_modify.
19905 (exynosm1_addrcost_table, xgene1_addrcost_table): Likewise.
19906 (thunderx2t99_addrcost_table, thunderx3t110_addrcost_table):
19907 (tsv110_addrcost_table, qdf24xx_addrcost_table): Likewise.
19908 (a64fx_addrcost_table): Likewise.
19909 (neoversev1_addrcost_table): New.
19910 (neoversev1_tunings): Use neoversev1_addrcost_table.
19911 (aarch64_address_cost): Use the new post_modify costs for CImode
19914 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19916 * config/aarch64/aarch64.opt
19917 (-param=aarch64-loop-vect-issue-rate-niters=): New parameter.
19918 * doc/invoke.texi: Document it.
19919 * config/aarch64/aarch64-protos.h (aarch64_base_vec_issue_info)
19920 (aarch64_scalar_vec_issue_info, aarch64_simd_vec_issue_info)
19921 (aarch64_advsimd_vec_issue_info, aarch64_sve_vec_issue_info)
19922 (aarch64_vec_issue_info): New structures.
19923 (cpu_vector_cost): Write comments above the variables rather
19925 (cpu_vector_cost::issue_info): New member variable.
19926 * config/aarch64/aarch64.c: Include gimple-pretty-print.h
19927 and tree-ssa-loop-niter.h.
19928 (generic_vector_cost, a64fx_vector_cost, qdf24xx_vector_cost)
19929 (thunderx_vector_cost, tsv110_vector_cost, cortexa57_vector_cost)
19930 (exynosm1_vector_cost, xgene1_vector_cost, thunderx2t99_vector_cost)
19931 (thunderx3t110_vector_cost): Initialize issue_info to null.
19932 (neoversev1_scalar_issue_info, neoversev1_advsimd_issue_info)
19933 (neoversev1_sve_issue_info, neoversev1_vec_issue_info): New structures.
19934 (neoversev1_vector_cost): Use them.
19935 (aarch64_vec_op_count, aarch64_sve_op_count): New structures.
19936 (aarch64_vector_costs::saw_sve_only_op): New member variable.
19937 (aarch64_vector_costs::num_vector_iterations): Likewise.
19938 (aarch64_vector_costs::scalar_ops): Likewise.
19939 (aarch64_vector_costs::advsimd_ops): Likewise.
19940 (aarch64_vector_costs::sve_ops): Likewise.
19941 (aarch64_vector_costs::seen_loads): Likewise.
19942 (aarch64_simd_vec_costs_for_flags): New function.
19943 (aarch64_analyze_loop_vinfo): Initialize num_vector_iterations.
19944 Count the number of predicate operations required by SVE WHILE
19946 (aarch64_comparison_type, aarch64_multiply_add_p): New functions.
19947 (aarch64_sve_only_stmt_p, aarch64_in_loop_reduction_latency): Likewise.
19948 (aarch64_count_ops): Likewise.
19949 (aarch64_add_stmt_cost): Record whether see an SVE operation
19950 that cannot currently be implementing using Advanced SIMD.
19951 Record issue information about the scalar, Advanced SIMD
19952 and (where relevant) SVE versions of a loop.
19953 (aarch64_vec_op_count::dump): New function.
19954 (aarch64_sve_op_count::dump): Likewise.
19955 (aarch64_estimate_min_cycles_per_iter): Likewise.
19956 (aarch64_adjust_body_cost): If issue information is available,
19957 try to compare the issue rates of the various loop implementations
19958 and increase or decrease the vector body cost accordingly.
19960 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19962 * config/aarch64/aarch64.c (aarch64_detect_vector_stmt_subtype):
19963 Assume a zero cost for induction phis.
19965 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19967 * config/aarch64/aarch64.c (aarch64_embedded_comparison_type): New
19969 (aarch64_adjust_stmt_cost): Add the costs of embedded scalar and
19970 vector comparisons.
19972 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19974 * config/aarch64/aarch64.c (aarch64_detect_scalar_stmt_subtype):
19976 (aarch64_add_stmt_cost): Call it.
19978 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19980 * config/aarch64/aarch64-tuning-flags.def (matched_vector_throughput):
19981 New tuning parameter.
19982 * config/aarch64/aarch64.c (neoversev1_tunings): Use it.
19983 (aarch64_estimated_sve_vq): New function.
19984 (aarch64_vector_costs::analyzed_vinfo): New member variable.
19985 (aarch64_vector_costs::is_loop): Likewise.
19986 (aarch64_vector_costs::unrolled_advsimd_niters): Likewise.
19987 (aarch64_vector_costs::unrolled_advsimd_stmts): Likewise.
19988 (aarch64_record_potential_advsimd_unrolling): New function.
19989 (aarch64_analyze_loop_vinfo, aarch64_analyze_bb_vinfo): Likewise.
19990 (aarch64_add_stmt_cost): Call aarch64_analyze_loop_vinfo or
19991 aarch64_analyze_bb_vinfo on the first use of a costs structure.
19992 Detect whether we're vectorizing a loop for SVE that might be
19993 completely unrolled if it used Advanced SIMD instead.
19994 (aarch64_adjust_body_cost_for_latency): New function.
19995 (aarch64_finish_cost): Call it.
19997 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19999 * config/aarch64/aarch64.c (aarch64_vector_costs): New structure.
20000 (aarch64_init_cost): New function.
20001 (aarch64_add_stmt_cost): Use aarch64_vector_costs instead of
20002 the default unsigned[3].
20003 (aarch64_finish_cost, aarch64_destroy_cost_data): New functions.
20004 (TARGET_VECTORIZE_INIT_COST): Override.
20005 (TARGET_VECTORIZE_FINISH_COST): Likewise.
20006 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
20008 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20010 * config/aarch64/aarch64.c (neoversev1_advsimd_vector_cost)
20011 (neoversev1_sve_vector_cost): New cost structures.
20012 (neoversev1_vector_cost): Likewise.
20013 (neoversev1_tunings): Use them. Enable use_new_vector_costs.
20015 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20017 * config/aarch64/aarch64-protos.h
20018 (sve_vec_cost::scatter_store_elt_cost): New member variable.
20019 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
20020 accordingly, taking the cost from the cost of a scalar_store.
20021 (a64fx_sve_vector_cost): Likewise.
20022 (aarch64_detect_vector_stmt_subtype): Detect scatter stores.
20024 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20026 * config/aarch64/aarch64-protos.h
20027 (simd_vec_cost::store_elt_extra_cost): New member variable.
20028 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
20029 accordingly, using the vec_to_scalar cost for the new field.
20030 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
20031 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
20032 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
20033 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
20034 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
20035 (thunderx3t110_advsimd_vector_cost): Likewise.
20036 (aarch64_detect_vector_stmt_subtype): Detect single-element stores.
20038 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20040 * config/aarch64/aarch64-protos.h (simd_vec_cost::ld2_st2_permute_cost)
20041 (simd_vec_cost::ld3_st3_permute_cost): New member variables.
20042 (simd_vec_cost::ld4_st4_permute_cost): Likewise.
20043 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
20044 accordingly, using zero for the new costs.
20045 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
20046 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
20047 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
20048 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
20049 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
20050 (thunderx3t110_advsimd_vector_cost): Likewise.
20051 (aarch64_ld234_st234_vectors): New function.
20052 (aarch64_adjust_stmt_cost): Likewise.
20053 (aarch64_add_stmt_cost): Call aarch64_adjust_stmt_cost if using
20054 the new vector costs.
20056 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20058 * config/aarch64/aarch64-protos.h (sve_vec_cost): Turn into a
20059 derived class of simd_vec_cost. Add information about CLAST[AB]
20060 and FADDA instructions.
20061 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
20062 accordingly, using the vec_to_scalar costs for the new fields.
20063 (a64fx_sve_vector_cost): Likewise.
20064 (aarch64_reduc_type): New function.
20065 (aarch64_sve_in_loop_reduction_latency): Likewise.
20066 (aarch64_detect_vector_stmt_subtype): Take a vinfo parameter.
20067 Use aarch64_sve_in_loop_reduction_latency to handle SVE reductions
20068 that occur in the loop body.
20069 (aarch64_add_stmt_cost): Update call accordingly.
20071 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20073 * config/aarch64/aarch64-tuning-flags.def (use_new_vector_costs):
20075 * config/aarch64/aarch64-protos.h (simd_vec_cost): Put comments
20076 above the fields rather than to the right.
20077 (simd_vec_cost::reduc_i8_cost): New member variable.
20078 (simd_vec_cost::reduc_i16_cost): Likewise.
20079 (simd_vec_cost::reduc_i32_cost): Likewise.
20080 (simd_vec_cost::reduc_i64_cost): Likewise.
20081 (simd_vec_cost::reduc_f16_cost): Likewise.
20082 (simd_vec_cost::reduc_f32_cost): Likewise.
20083 (simd_vec_cost::reduc_f64_cost): Likewise.
20084 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
20085 accordingly, using the vec_to_scalar_cost for the new fields.
20086 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
20087 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
20088 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
20089 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
20090 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
20091 (thunderx3t110_advsimd_vector_cost): Likewise.
20092 (aarch64_use_new_vector_costs_p): New function.
20093 (aarch64_simd_vec_costs): New function, split out from...
20094 (aarch64_builtin_vectorization_cost): ...here.
20095 (aarch64_is_reduction): New function.
20096 (aarch64_detect_vector_stmt_subtype): Likewise.
20097 (aarch64_add_stmt_cost): Call aarch64_detect_vector_stmt_subtype if
20098 using the new vector costs.
20100 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
20103 * tree-emutls.c (get_emutls_init_templ_addr): Mark initializer of weak
20104 TLS declarations as public.
20106 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
20108 * config/aarch64/aarch64-d.c (IN_TARGET_CODE): Define.
20109 * config/arm/arm-d.c (IN_TARGET_CODE): Likewise.
20110 * config/i386/i386-d.c (IN_TARGET_CODE): Likewise.
20111 * config/mips/mips-d.c (IN_TARGET_CODE): Likewise.
20112 * config/pa/pa-d.c (IN_TARGET_CODE): Likewise.
20113 * config/riscv/riscv-d.c (IN_TARGET_CODE): Likewise.
20114 * config/rs6000/rs6000-d.c (IN_TARGET_CODE): Likewise.
20115 * config/s390/s390-d.c (IN_TARGET_CODE): Likewise.
20116 * config/sparc/sparc-d.c (IN_TARGET_CODE): Likewise.
20118 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
20121 * config.gcc (*-*-cygwin*): Add winnt-d.o
20122 (*-*-mingw*): Likewise.
20123 * config/i386/cygwin.h (EXTRA_TARGET_D_OS_VERSIONS): New macro.
20124 * config/i386/mingw32.h (EXTRA_TARGET_D_OS_VERSIONS): Likewise.
20125 * config/i386/t-cygming: Add winnt-d.o.
20126 * config/i386/winnt-d.c: New file.
20128 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
20130 * config/freebsd-d.c: Include memmodel.h.
20132 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
20135 * config.gcc (*-*-openbsd*): Add openbsd-d.o.
20136 * config/t-openbsd: Add openbsd-d.o.
20137 * config/openbsd-d.c: New file.
20139 2021-03-25 Stam Markianos-Wright <stam.markianos-wright@arm.com>
20141 PR tree-optimization/96974
20142 * tree-vect-stmts.c (vect_get_vector_types_for_stmt): Replace assert
20143 with graceful exit.
20145 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
20148 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
20152 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
20153 always_inline in system headers.
20155 2021-03-25 Kewen Lin <linkw@linux.ibm.com>
20157 * tree-vect-loop.c (vect_model_reduction_cost): Init inside_cost.
20159 2021-03-25 Jakub Jelinek <jakub@redhat.com>
20162 * tree-core.h (enum operand_equal_flag): Add OEP_ADDRESS_OF_SAME_FIELD.
20163 * fold-const.c (operand_compare::operand_equal_p): Don't compare
20164 field offsets if OEP_ADDRESS_OF_SAME_FIELD.
20166 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
20170 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
20171 always_inline in system headers.
20173 2021-03-25 Richard Biener <rguenther@suse.de>
20175 PR tree-optimization/99746
20176 * tree-vect-slp-patterns.c (complex_pattern::build): Do not mark
20177 the scalar stmt as patterned. Instead set up required things
20180 2021-03-25 Xionghu Luo <luoxhu@linux.ibm.com>
20182 * config/rs6000/rs6000.c (power8_costs): Change l2 cache
20185 2021-03-24 Martin Liska <mliska@suse.cz>
20188 * common/config/i386/i386-common.c (ARRAY_SIZE): Fix off-by-one
20190 * config/i386/i386-options.c (ix86_option_override_internal):
20191 Add run-time assert.
20193 2021-03-24 Martin Jambor <mjambor@suse.cz>
20196 * ipa-cp.c (initialize_node_lattices): Mark as bottom all
20197 parameters with unknown type.
20198 (ipacp_value_safe_for_type): New function.
20199 (propagate_vals_across_arith_jfunc): Verify that the constant type
20200 can be used for a type of the formal parameter.
20201 (propagate_vals_across_ancestor): Likewise.
20202 (propagate_scalar_across_jump_function): Likewise. Pass the type
20203 also to propagate_vals_across_ancestor.
20205 2021-03-24 Christophe Lyon <christophe.lyon@linaro.org>
20208 * config/arm/mve.md (movmisalign<mode>_mve_store): Use Ux
20210 (movmisalign<mode>_mve_load): Likewise.
20212 2021-03-24 Jakub Jelinek <jakub@redhat.com>
20215 * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
20216 movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.
20218 2021-03-24 Alexandre Oliva <oliva@adacore.com>
20220 * doc/sourcebuild.texi (sysconf): New effective target.
20222 2021-03-24 Alexandre Oliva <oliva@adacore.com>
20224 * config/i386/predicates.md (reg_or_const_vec_operand): New.
20225 * config/i386/sse.md (ssse3_pshufbv8qi3): Add an expander for
20226 the now *-prefixed insn_and_split, turn the splitter const vec
20227 into an input for the insn, making it an ignored immediate for
20228 non-split cases, and loaded into the scratch register
20231 2021-03-23 Vladimir N. Makarov <vmakarov@redhat.com>
20234 * config/aarch64/constraints.md (Utq, UOb, UOh, UOw, UOd, UOty):
20235 Use define_relaxed_memory_constraint for them.
20237 2021-03-23 Iain Sandoe <iain@sandoe.co.uk>
20240 * config/host-darwin.c (darwin_gt_pch_use_address): Add a
20241 colon to the diagnostic message.
20243 2021-03-23 Ilya Leoshkevich <iii@linux.ibm.com>
20245 * fwprop.c (fwprop_propagation::fwprop_propagation): Look at
20247 (try_fwprop_subst_note): Use set_info instead of insn_info.
20248 (try_fwprop_subst_pattern): Likewise.
20249 (try_fwprop_subst_notes): Likewise.
20250 (try_fwprop_subst): Likewise.
20251 (forward_propagate_subreg): Likewise.
20252 (forward_propagate_and_simplify): Likewise.
20253 (forward_propagate_into): Likewise.
20254 * rtl-ssa/accesses.h (set_info::single_nondebug_use) New
20256 (set_info::single_nondebug_insn_use): Likewise.
20257 (set_info::single_phi_use): Likewise.
20258 * rtl-ssa/member-fns.inl (set_info::single_nondebug_use) New
20260 (set_info::single_nondebug_insn_use): Likewise.
20261 (set_info::single_phi_use): Likewise.
20263 2021-03-23 Christophe Lyon <christophe.lyon@linaro.org>
20265 * doc/sourcebuild.texi (arm_dsp_ok, arm_dsp): Document.
20267 2021-03-23 Jakub Jelinek <jakub@redhat.com>
20270 * config/aarch64/aarch64.c (aarch64_add_offset): Tell
20271 expand_mult to perform an unsigned rather than a signed
20274 2021-03-23 H.J. Lu <hjl.tools@gmail.com>
20277 * config/i386/cpuid.h (__cpuid): Add __volatile__.
20278 (__cpuid_count): Likewise.
20280 2021-03-23 Richard Biener <rguenther@suse.de>
20282 PR tree-optimization/99721
20283 * tree-vect-slp.c (vect_slp_analyze_node_operations):
20284 Make sure we can schedule the node.
20286 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20288 * config/riscv/riscv.c (riscv_subword): Take endianness into
20289 account when calculating the byte offset.
20291 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20293 * config/riscv/predicates.md (subreg_lowpart_operator): New predicate
20294 * config/riscv/riscv.md (*addsi3_extended2, *subsi3_extended2)
20295 (*negsi2_extended2, *mulsi3_extended2, *<optab>si3_mask)
20296 (*<optab>si3_mask_1, *<optab>di3_mask, *<optab>di3_mask_1)
20297 (*<optab>si3_extend_mask, *<optab>si3_extend_mask_1): Use
20298 new predicate "subreg_lowpart_operator"
20300 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20302 * config/riscv/riscv.c (riscv_swap_instruction): New function
20303 to byteswap an SImode rtx containing an instruction.
20304 (riscv_trampoline_init): Byteswap the generated instructions
20307 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20309 * common/config/riscv/riscv-common.c
20310 (TARGET_DEFAULT_TARGET_FLAGS): Set default endianness.
20311 * config.gcc (riscv32be-*, riscv64be-*): Set
20312 TARGET_BIG_ENDIAN_DEFAULT to 1.
20313 * config/riscv/elf.h (LINK_SPEC): Change -melf* value
20314 depending on default endianness.
20315 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
20316 * config/riscv/linux.h (LINK_SPEC): Likewise.
20317 * config/riscv/riscv.c (TARGET_DEFAULT_TARGET_FLAGS): Set
20318 default endianness.
20319 * config/riscv/riscv.h (DEFAULT_ENDIAN_SPEC): New macro.
20321 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20323 * config/riscv/elf.h (LINK_SPEC): Pass linker endianness flag.
20324 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
20325 * config/riscv/linux.h (LINK_SPEC): Likewise.
20326 * config/riscv/riscv.h (ASM_SPEC): Pass -mbig-endian and
20328 (BYTES_BIG_ENDIAN): Handle big endian.
20329 (WORDS_BIG_ENDIAN): Define to BYTES_BIG_ENDIAN.
20330 * config/riscv/riscv.opt (-mbig-endian, -mlittle-endian): New
20332 * doc/invoke.texi (-mbig-endian, -mlittle-endian): Document.
20334 2021-03-23 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
20336 * regcprop.c (find_oldest_value_reg): Ask target whether
20337 different mode is fine for replacement register.
20339 2021-03-23 Aldy Hernandez <aldyh@redhat.com>
20341 PR tree-optimization/99296
20342 * value-range.cc (irange::irange_set_1bit_anti_range): New.
20343 (irange::irange_set_anti_range): Call irange_set_1bit_anti_range
20344 * value-range.h (irange::irange_set_1bit_anti_range): New.
20346 2021-03-22 Vladimir N. Makarov <vmakarov@redhat.com>
20349 * config/aarch64/constraints.md (UtQ): Use
20350 define_relaxed_memory_constraint for it.
20351 * doc/md.texi (define_relaxed_memory_constraint): Describe it.
20352 * genoutput.c (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
20353 * genpreds.c (constraint_data): Add bitfield is_relaxed_memory.
20354 (have_relaxed_memory_constraints): New static var.
20355 (relaxed_memory_start, relaxed_memory_end): Ditto.
20356 (add_constraint): Add arg is_relaxed_memory. Check name for
20357 relaxed memory. Set up is_relaxed_memory in constraint_data and
20358 have_relaxed_memory_constraints. Adjust calls.
20359 (choose_enum_order): Process relaxed memory.
20360 (write_tm_preds_h): Ditto.
20361 (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
20362 * gensupport.c (process_rtx): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
20363 * ira-costs.c (record_reg_classes): Process CT_RELAXED_MEMORY.
20364 * ira-lives.c (single_reg_class): Use
20365 insn_extra_relaxed_memory_constraint.
20366 * ira.c (ira_setup_alts): CT_RELAXED_MEMORY.
20367 * lra-constraints.c (valid_address_p): Use
20368 insn_extra_relaxed_memory_constraint instead of other memory
20370 (process_alt_operands): Process CT_RELAXED_MEMORY.
20371 (curr_insn_transform): Use insn_extra_relaxed_memory_constraint.
20372 * recog.c (asm_operand_ok, preprocess_constraints): Process
20374 * reload.c (find_reloads): Ditto.
20375 * rtl.def (DEFINE_RELAXED_MEMORY_CONSTRAINT): New.
20376 * stmt.c (parse_input_constraint): Use
20377 insn_extra_relaxed_memory_constraint.
20379 2021-03-22 Segher Boessenkool <segher@kernel.crashing.org>
20382 * ubsan.c (ubsan_instrument_float_cast): Don't test for unordered if
20385 2021-03-22 Alex Coplan <alex.coplan@arm.com>
20388 * config/arm/arm-protos.h (neon_make_constant): Add generate
20389 argument to guard emitting insns, default to true.
20390 * config/arm/arm.c (arm_legitimate_constant_p_1): Reject
20391 CONST_VECTORs which neon_make_constant can't handle.
20392 (neon_vdup_constant): Add generate argument, avoid emitting
20393 insns if it's not set.
20394 (neon_make_constant): Plumb new generate argument through.
20395 * config/arm/constraints.md (Ui): New. Use it...
20396 * config/arm/mve.md (*mve_mov<mode>): ... here.
20397 * config/arm/vec-common.md (movv8hf): Use neon_make_constant to
20398 synthesize constants.
20400 2021-03-22 Richard Biener <rguenther@suse.de>
20402 * debug.h: Add deprecation warning.
20404 2021-03-22 Richard Biener <rguenther@suse.de>
20406 PR tree-optimization/99694
20407 * tree-ssa-sccvn.c (visit_phi): Ignore edges with the
20410 2021-03-22 Kito Cheng <kito.cheng@sifive.com>
20413 * config/riscv/riscv.c (riscv_expand_block_move): Get RTL value
20414 after type checking.
20416 2021-03-22 Jakub Jelinek <jakub@redhat.com>
20420 * dwarf2out.c (get_full_len): Use get_precision rather than
20422 (add_const_value_attribute): Make sure add_AT_wide argument has
20423 precision prec rather than some very wide one.
20425 2021-03-22 Kewen Lin <linkw@linux.ibm.com>
20427 * config/rs6000/rs6000.md (*rotldi3_insert_sf,
20428 *mov<SFDF:mode><SFDF2:mode>cc_p9, floatsi<mode>2_lfiwax,
20429 floatsi<mode>2_lfiwax_mem, floatunssi<mode>2_lfiwzx,
20430 floatunssi<mode>2_lfiwzx_mem, *floatsidf2_internal,
20431 *floatunssidf2_internal, fix_trunc<mode>si2_stfiwx,
20432 fix_trunc<mode>si2_internal, fixuns_trunc<mode>si2_stfiwx,
20433 *round32<mode>2_fprs, *roundu32<mode>2_fprs,
20434 *fix_trunc<mode>si2_internal): Fix empty split condition.
20435 * config/rs6000/vsx.md (*vsx_le_undo_permute_<mode>,
20436 vsx_reduc_<VEC_reduc_name>_v2df, vsx_reduc_<VEC_reduc_name>_v4sf,
20437 *vsx_reduc_<VEC_reduc_name>_v2df_scalar,
20438 *vsx_reduc_<VEC_reduc_name>_v4sf_scalar): Likewise.
20440 2021-03-22 Xionghu Luo <luoxhu@linux.ibm.com>
20443 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
20444 Convert idx to DImode.
20445 (rs6000_expand_vector_set_var_p8): Likewise.
20447 2021-03-21 Jakub Jelinek <jakub@redhat.com>
20450 * dwarf2out.c (insert_float): Change return type from void to
20451 unsigned, handle GET_MODE_SIZE (mode) == 2 and return element size.
20452 (mem_loc_descriptor, loc_descriptor, add_const_value_attribute):
20455 2021-03-20 H.J. Lu <hjl.tools@gmail.com>
20458 * config/i386/i386.c (construct_container): Check cfun != NULL
20459 before accessing silent_p.
20461 2021-03-20 Ahamed Husni <ahamedhusni73@gmail.com>
20463 * asan.c: Fix typos in comments.
20465 2021-03-20 Vladimir N. Makarov <vmakarov@redhat.com>
20467 PR rtl-optimization/99680
20468 * lra-constraints.c (skip_contraint_modifiers): Rename to skip_constraint_modifiers.
20469 (process_address_1): Check empty constraint before using
20472 2021-03-19 Pat Haugen <pthaugen@linux.ibm.com>
20474 * config/rs6000/rs6000.c (power10_cost): New.
20475 (rs6000_option_override_internal): Set Power10 costs.
20476 (rs6000_issue_rate): Set Power10 issue rate.
20477 * config/rs6000/power10.md: Rewrite for Power10.
20479 2021-03-19 Vladimir N. Makarov <vmakarov@redhat.com>
20482 * lra-constraints.c (process_address_1): Don't use unknown
20483 constraint for address constraint.
20485 2021-03-19 Iain Sandoe <iain@sandoe.co.uk>
20488 * config.gcc (powerpc-*-darwin8): Delete the reference to
20489 the now removed darwin8.h.
20491 2021-03-19 Olivier Hainque <hainque@adacore.com>
20494 * config/vxworksae.h (VX_CPU_PREFIX): Define.
20496 2021-03-19 John David Anglin <danglin@gcc.gnu.org>
20498 * config/pa/pa.c (import_milli): Use memcpy instead of strncpy.
20500 2021-03-19 Tamar Christina <tamar.christina@arm.com>
20502 PR tree-optimization/99656
20503 * tree-vect-slp-patterns.c (linear_loads_p,
20504 complex_add_pattern::matches, is_eq_or_top,
20505 vect_validate_multiplication, complex_mul_pattern::matches,
20506 complex_fms_pattern::matches): Remove complex_perm_kinds_t.
20507 * tree-vectorizer.h: (complex_load_perm_t): Removed.
20508 (slp_tree_to_load_perm_map_t): Use complex_perm_kinds_t instead of
20509 complex_load_perm_t.
20511 2021-03-19 H.J. Lu <hjl.tools@gmail.com>
20514 * config/i386/i386-options.c (ix86_init_machine_status): Set
20516 * config/i386/i386.c (init_cumulative_args): Set silent_p to
20518 (construct_container): Return early for return and argument
20519 errors if silent_p is true.
20520 * config/i386/i386.h (machine_function): Add silent_p.
20522 2021-03-19 Jakub Jelinek <jakub@redhat.com>
20525 * config/arm/constraints.md (Ds): New constraint.
20526 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): Use w,Ds
20527 constraint instead of w,Dm.
20529 2021-03-19 Andrew Stubbs <ams@codesourcery.com>
20531 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Fix quotes
20534 2021-03-19 Eric Botcazou <ebotcazou@adacore.com>
20536 PR middle-end/99641
20537 * fold-const.c (native_encode_initializer) <CONSTRUCTOR>: For an
20538 array type, do the computation of the current position in sizetype.
20540 2021-03-18 Vladimir N. Makarov <vmakarov@redhat.com>
20543 * lra-constraints.c (process_address_1): Use lookup_constraint
20544 only for a single constraint.
20546 2021-03-18 Martin Sebor <msebor@redhat.com>
20548 PR middle-end/99502
20549 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): Rename...
20550 (inbounds_memaccess_p): ...to this. Check the ending offset of
20551 the accessed member.
20553 2021-03-18 Andrew Stubbs <ams@codesourcery.com>
20555 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Add %< and
20556 %> quote markers to error messages.
20557 (gcn_goacc_validate_dims): Likewise.
20558 (gcn_conditional_register_usage): Remove exclaimation mark from error
20560 (gcn_vectorize_vec_perm_const): Ensure perm is fully uninitialized.
20562 2021-03-18 Jan Hubicka <hubicka@ucw.cz>
20564 * config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
20567 2021-03-18 Sinan Lin <sinan@isrc.iscas.ac.cn>
20568 Kito Cheng <kito.cheng@sifive.com>
20570 * config/riscv/riscv.c (riscv_block_move_straight): Change type
20571 to unsigned HOST_WIDE_INT for parameter and local variable with
20572 HOST_WIDE_INT type.
20573 (riscv_adjust_block_mem): Ditto.
20574 (riscv_block_move_loop): Ditto.
20575 (riscv_expand_block_move): Ditto.
20577 2021-03-18 Nick Clifton <nickc@redhat.com>
20579 * config/v850/v850.c (construct_restore_jr): Increase static
20581 (construct_save_jarl): Likewise.
20582 * config/v850/v850.h (DWARF2_DEBUGGING_INFO): Define.
20584 2021-03-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20586 * config/aarch64/aarch64.c (aarch64_adjust_generic_arch_tuning): Define.
20587 (aarch64_override_options_internal): Use it.
20588 (generic_tunings): Add AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS to
20591 2021-03-17 Sandra Loosemore <sandra@codesourcery.com>
20593 * config/nios2/nios2.c (nios2_custom_check_insns): Clean up
20594 error message format issues.
20595 (nios2_option_override): Likewise.
20596 (nios2_expand_fpu_builtin): Likewise.
20597 (nios2_init_custom_builtins): Adjust to avoid bogus strncpy
20598 truncation warning.
20599 (nios2_expand_custom_builtin): More error message format fixes.
20600 (nios2_expand_rdwrctl_builtin): Likewise.
20601 (nios2_expand_rdprs_builtin): Likewise.
20602 (nios2_expand_eni_builtin): Likewise.
20603 (nios2_expand_builtin): Likewise.
20604 (nios2_register_custom_code): Likewise.
20605 (nios2_valid_target_attribute_rec): Likewise.
20606 (nios2_add_insn_asm): Fix uninitialized variable warning.
20608 2021-03-17 Jan Hubicka <jh@suse.cz>
20610 * config/i386/x86-tune-costs.h (struct processor_costs): Update costs
20611 of gather to match reality.
20612 * config/i386/x86-tune.def (X86_TUNE_USE_GATHER): Enable for znver3.
20614 2021-03-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20616 * config/aarch64/aarch64-builtins.c (aarch64_expand_rng_builtin): Use EQ
20617 to compare against CC_REG rather than NE.
20619 2021-03-17 H.J. Lu <hjl.tools@gmail.com>
20622 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support
20623 inline assembly statements.
20624 (ix86_print_operand): Update 'P' handling for -fno-plt.
20626 2021-03-17 Tamar Christina <tamar.christina@arm.com>
20629 * config/aarch64/aarch64.c
20630 (aarch64_simd_clone_compute_vecsize_and_simdlen): Remove unused var.
20632 2021-03-16 Segher Boessenkool <segher@kernel.crashing.org>
20635 * config/rs6000/predicates.md (branch_comparison_operator): Allow
20636 ordered and unordered for CCFPmode, if flag_finite_math_only.
20638 2021-03-16 Jakub Jelinek <jakub@redhat.com>
20641 * config/i386/i386-expand.c (ix86_split_lea_for_addr): Emit a MULT
20642 rather than ASHIFT.
20643 * config/i386/i386.md (mult by 1248 into ashift): New splitter.
20645 2021-03-16 Martin Liska <mliska@suse.cz>
20648 * optc-save-gen.awk: Add flag_ipa_ra to exceptions for
20649 cl_optimization_compare function.
20651 2021-03-16 Ilya Leoshkevich <iii@linux.ibm.com>
20653 * config/s390/s390.c (f_constraint_p): Treat "fv" constraints
20656 2021-03-16 Jakub Jelinek <jakub@redhat.com>
20659 * config/i386/i386.h (struct machine_function): Add
20660 has_explicit_vzeroupper bitfield.
20661 * config/i386/i386-expand.c (ix86_expand_builtin): Set
20662 cfun->machine->has_explicit_vzeroupper when expanding
20663 IX86_BUILTIN_VZEROUPPER.
20664 * config/i386/i386-features.c (rest_of_handle_insert_vzeroupper):
20665 Do the mode switching only when TARGET_VZEROUPPER, expensive
20666 optimizations turned on and not optimizing for size.
20667 (pass_insert_vzeroupper::gate): Enable even when
20668 cfun->machine->has_explicit_vzeroupper is set.
20670 2021-03-16 Jakub Jelinek <jakub@redhat.com>
20673 * config/aarch64/aarch64.c
20674 (aarch64_simd_clone_compute_vecsize_and_simdlen): If not a function
20675 definition, walk TYPE_ARG_TYPES list if non-NULL for argument types
20676 instead of DECL_ARGUMENTS. Ignore types for uniform arguments.
20678 2021-03-15 Richard Biener <rguenther@suse.de>
20680 PR tree-optimization/98834
20681 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle missing
20682 subsetting by truncating the access size.
20684 2021-03-15 Jan Hubicka <hubicka@ucw.cz>
20686 * config/i386/i386-options.c (processor_cost_table): Add znver3_cost.
20687 * config/i386/x86-tune-costs.h (znver3_cost): New gobal variable; copy
20690 2021-03-15 Martin Liska <mliska@suse.cz>
20692 * spellcheck.c: Add missing comma in initialization.
20694 2021-03-14 Uroš Bizjak <ubizjak@gmail.com>
20696 * config/i386/sse.md (*vec_extract<mode>): Merge alternative 0 with
20697 alternative 2 and alternative 1 with alternative 3 using
20698 YW register constraint.
20699 (*vec_extract<PEXTR_MODE12:mode>_zext): Merge alternatives
20700 using YW register constraint.
20701 (*vec_extractv16qi_zext): Ditto.
20702 (*vec_extractv4si): Merge alternatives 4 and 5
20703 using Yw register constraint.
20704 (*ssse3_palignr<mode>_perm): Use Yw instead of v for alternative 3.
20706 2021-03-13 Martin Sebor <msebor@redhat.com>
20708 PR tree-optimization/99489
20709 * builtins.c (gimple_call_alloc_size): Fail gracefully when argument
20710 is not a call statement.
20712 2021-03-13 Jakub Jelinek <jakub@redhat.com>
20714 PR tree-optimization/99544
20715 * match.pd (X + (X << C) -> X * (1 + (1 << C))): Don't simplify
20716 if for vector types multiplication can't be done in type's mode.
20718 2021-03-12 Eric Botcazou <ebotcazou@adacore.com>
20721 * config/sparc/constraints.md (w): Rename to...
20722 (W): ... this and ditch previous implementation.
20723 * config/sparc/sparc.md (*movdi_insn_sp64): Replace W with m.
20724 (*movdf_insn_sp64): Likewise.
20725 (*mov<VM64:mode>_insn_sp64): Likewise.
20726 * config/sparc/sync.md (*atomic_compare_and_swap<mode>_1): Replace
20728 (atomic_compare_and_swap_leon3_1): Likewise.
20729 (*atomic_compare_and_swapdi_v8plus): Likewise.
20730 * config/sparc/sparc.c (memory_ok_for_ldd): Remove useless test on
20731 architecture and add missing address validity check during LRA.
20733 2021-03-12 Tobias Burnus <tobias@codesourcery.com>
20736 * gimplify.c (omp_add_variable): Handle NULL_TREE as size
20737 occuring for assumed-size arrays in use_device_{ptr,addr}.
20739 2021-03-12 Jakub Jelinek <jakub@redhat.com>
20742 * config/i386/constraints.md (YW): New internal constraint.
20743 * config/i386/sse.md (v_Yw): Add V4TI, V2TI, V1TI and TI cases.
20744 (*<sse2_avx2>_<insn><mode>3<mask_name>,
20745 *<sse2_avx2>_uavg<mode>3<mask_name>, *abs<mode>2,
20746 *<s>mul<mode>3_highpart<mask_name>): Use <v_Yw> instead of v in
20748 (<sse2_avx2>_psadbw): Use YW instead of v in constraints.
20749 (*avx2_pmaddwd, *sse2_pmaddwd, *<code>v8hi3, *<code>v16qi3,
20750 avx2_pmaddubsw256, ssse3_pmaddubsw128): Merge last two alternatives
20751 into one, use Yw instead of former x,v.
20752 (ashr<mode>3, <insn><mode>3): Use <v_Yw> instead of x in constraints of
20753 the last alternative.
20754 (<sse2_avx2>_packsswb<mask_name>, <sse2_avx2>_packssdw<mask_name>,
20755 <sse2_avx2>_packuswb<mask_name>, <sse4_1_avx2>_packusdw<mask_name>,
20756 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, <ssse3_avx2>_palignr<mode>,
20757 <ssse3_avx2>_pshufb<mode>3<mask_name>): Merge last two alternatives
20758 into one, use <v_Yw> instead of former x,v.
20759 (avx2_interleave_highv32qi<mask_name>,
20760 vec_interleave_highv16qi<mask_name>): Use Yw instead of v in
20761 constraints. Add && <mask_avx512bw_condition> to condition.
20762 (avx2_interleave_lowv32qi<mask_name>,
20763 vec_interleave_lowv16qi<mask_name>,
20764 avx2_interleave_highv16hi<mask_name>,
20765 vec_interleave_highv8hi<mask_name>,
20766 avx2_interleave_lowv16hi<mask_name>, vec_interleave_lowv8hi<mask_name>,
20767 avx2_pshuflw_1<mask_name>, sse2_pshuflw_1<mask_name>,
20768 avx2_pshufhw_1<mask_name>, sse2_pshufhw_1<mask_name>,
20769 avx2_<code>v16qiv16hi2<mask_name>, sse4_1_<code>v8qiv8hi2<mask_name>,
20770 *sse4_1_<code>v8qiv8hi2<mask_name>_1, <sse2_avx2>_<insn><mode>3): Use
20771 Yw instead of v in constraints.
20772 * config/i386/mmx.md (Yv_Yw): New define_mode_attr.
20773 (*mmx_<insn><mode>3, mmx_ashr<mode>3, mmx_<insn><mode>3): Use <Yv_Yw>
20774 instead of Yv in constraints.
20775 (*mmx_<insn><mode>3, *mmx_mulv4hi3, *mmx_smulv4hi3_highpart,
20776 *mmx_umulv4hi3_highpart, *mmx_pmaddwd, *mmx_<code>v4hi3,
20777 *mmx_<code>v8qi3, mmx_pack<s_trunsuffix>swb, mmx_packssdw,
20778 mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
20779 *mmx_uavgv8qi3, *mmx_uavgv4hi3, mmx_psadbw): Use Yw instead of Yv in
20781 (*mmx_pinsrw, *mmx_pinsrb, *mmx_pextrw, *mmx_pextrw_zext, *mmx_pextrb,
20782 *mmx_pextrb_zext): Use YW instead of Yv in constraints.
20783 (*mmx_eq<mode>3, mmx_gt<mode>3): Use x instead of Yv in constraints.
20784 (mmx_andnot<mode>3, *mmx_<code><mode>3): Split last alternative into
20785 two, one with just x, another isa avx512vl with v.
20787 2021-03-12 Martin Liska <mliska@suse.cz>
20789 * doc/invoke.texi: Add missing param documentation.
20791 2021-03-11 David Malcolm <dmalcolm@redhat.com>
20794 * Makefile.in (ANALYZER_OBJS): Add analyzer/feasible-graph.o and
20795 analyzer/trimmed-graph.o.
20796 * doc/analyzer.texi (Analyzer Paths): Rewrite description of
20797 feasibility checking to reflect new implementation.
20798 * doc/invoke.texi (-fdump-analyzer-feasibility): Document new
20800 * shortest-paths.h (shortest_paths::get_shortest_distance): New.
20802 2021-03-11 David Malcolm <dmalcolm@redhat.com>
20804 * digraph.cc (selftest::test_shortest_paths): Update
20805 shortest_paths init for new param. Add test of
20806 SPS_TO_GIVEN_TARGET.
20807 * shortest-paths.h (enum shortest_path_sense): New.
20808 (shortest_paths::shortest_paths): Add "sense" param.
20809 Update for renamings. Generalize to use "sense" param.
20810 (shortest_paths::get_shortest_path): Rename param.
20811 (shortest_paths::m_sense): New field.
20812 (shortest_paths::m_prev): Rename...
20813 (shortest_paths::m_best_edge): ...to this.
20814 (shortest_paths::get_shortest_path): Update for renamings.
20815 Conditionalize flipping of path on sense of traversal.
20817 2021-03-11 David Malcolm <dmalcolm@redhat.com>
20819 * digraph.cc (selftest::test_shortest_paths): Add test coverage
20820 for paths from B and C.
20821 * shortest-paths.h (shortest_paths::shortest_paths): Handle
20822 unreachable nodes, rather than asserting.
20824 2021-03-11 David Edelsohn <dje.gcc@gmail.com>
20827 * config/rs6000/rs6000.c (rs6000_xcoff_file_start): Don't create
20828 xcoff_tbss_section_name.
20829 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .lcomm.
20830 * xcoffout.c (xcoff_tbss_section_name): Delete.
20831 * xcoffout.h (xcoff_tbss_section_name): Delete.
20833 2021-03-11 Richard Biener <rguenther@suse.de>
20835 PR tree-optimization/99523
20836 * tree-cfg.c (dump_function_to_file): Dump SSA names
20837 w/o identifier to the decls section as well, not only those
20838 without a VAR_DECL.
20840 2021-03-11 Jakub Jelinek <jakub@redhat.com>
20843 * ipa-icf-gimple.c (func_checker::compare_gimple_call): For internal
20844 function calls with lhs fail if the lhs don't have compatible types.
20846 2021-03-11 Hans-Peter Nilsson <hp@axis.com>
20848 * config/cris/cris.h (HARD_FRAME_POINTER_REGNUM): Define.
20849 Change FRAME_POINTER_REGNUM to correspond to a new faked
20850 register faked_fp, part of GENNONACR_REGS like faked_ap.
20851 (CRIS_FAKED_REGS_CONTENTS): New helper macro.
20852 (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS, CALL_USED_REGISTERS):
20853 (REG_ALLOC_ORDER, REG_CLASS_CONTENTS, REGNO_OK_FOR_BASE_P)
20854 (ELIMINABLE_REGS, REGISTER_NAMES): Adjust accordingly.
20855 * config/cris/cris.md (CRIS_FP_REGNUM): Renumber to new faked
20857 (CRIS_REAL_FP_REGNUM): New constant.
20858 * config/cris/cris.c (cris_reg_saved_in_regsave_area): Check
20859 for HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM.
20860 (cris_initial_elimination_offset): Handle elimination changes
20861 to HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM
20862 and add one from FRAME_POINTER_REGNUM to
20863 HARD_FRAME_POINTER_REGNUM.
20864 (cris_expand_prologue, cris_expand_epilogue): Emit code for
20865 hard_frame_pointer_rtx instead of frame_pointer_rtx.
20867 2021-03-10 David Edelsohn <dje.gcc@gmail.com>
20870 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Add check for DCmode.
20871 * config/rs6000/rs6000.c (rs6000_special_round_type_align): Same.
20873 2021-03-10 Vladimir N. Makarov <vmakarov@redhat.com>
20876 * lra-constraints.c (process_address_1): Don't check unknown
20877 constraint, use X for empty constraint.
20879 2021-03-10 Alex Coplan <alex.coplan@arm.com>
20881 * config/aarch64/aarch64.c (aarch64_vfp_is_call_or_return_candidate):
20882 Fix typo in comment describing "is_ha" argument.
20884 2021-03-10 John David Anglin <danglin@gcc.gnu.org>
20886 * doc/sourcebuild.texi: Document LRA target selector.
20888 2021-03-10 David Malcolm <dmalcolm@redhat.com>
20890 * doc/ux.texi: Add subsection contrasting interactive versus
20891 batch usage of GCC.
20893 2021-03-10 Joel Hutton <joel.hutton@arm.com>
20896 * tree-vect-stmts.c (vectorizable_store): Fix scatter store mask
20898 (vectorizable_load): Fix gather load mask check condition.
20900 2021-03-10 Richard Biener <rguenther@suse.de>
20902 PR tree-optimization/99510
20903 * tree.c (check_aligned_type): Check that the candidate
20904 has TYPE_USER_ALIGN set instead of matching with the
20907 2021-03-10 Eric Botcazou <ebotcazou@adacore.com>
20909 * config/sparc/sparc.c (sparc_regmode_natural_size): Return 4 for
20910 float and vector integer modes only if the mode is not larger.
20912 2021-03-10 Hans-Peter Nilsson <hp@axis.com>
20914 * config/cris/cris.h (DWARF_FRAME_REGISTERS): Define.
20916 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
20918 * ira.c (ira_setup_alts, ira_get_dup_out_num): Process digital
20920 * ira-lives.c (single_reg_class): Ditto.
20922 2021-03-09 Sebastian Huber <sebastian.huber@embedded-brains.de>
20924 * config.gcc (aarch64-*-rtems*): Include general rtems.h after
20925 the architecture-specific rtems.h.
20926 (aarch64-*-rtems*): Likewise.
20927 (arm*-*-rtems*): Likewise.
20928 (epiphany-*-rtems*): Likewise.
20929 (riscv*-*-rtems*): Likewise.
20931 2021-03-09 Jakub Jelinek <jakub@redhat.com>
20933 PR tree-optimization/99305
20934 * tree-ssa-phiopt.c (conditional_replacement): Test integer_pow2p
20935 before integer_all_onesp instead of vice versa.
20937 2021-03-09 Richard Earnshaw <rearnsha@arm.com>
20939 * common/config/arm/arm-common.c (arm_config_default): Change type
20940 of 'i' to unsigned.
20942 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
20945 * lra-constraints.c (process_address_1): Process constraint 'g'
20946 separately and digital constraints containing more one digit.
20948 2021-03-09 Nick Clifton <nickc@redhat.com>
20950 * config/rx/rx.h (DBX_DEBUGGING_INFO): Define.
20951 (DWARF"_DEBUGGING_INFO): Define.
20953 2021-03-09 Eric Botcazou <ebotcazou@adacore.com>
20956 * calls.c (initialize_argument_information): When the argument
20957 is passed by reference, do not make a copy in a thunk only if
20958 the argument is already in memory. Remove redundant test for
20959 the case of callee copy.
20961 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
20964 * lra-constraints.c (process_address_1): Process 0..9 constraints
20965 in process_address_1.
20967 2021-03-09 Andreas Krebbel <krebbel@linux.ibm.com>
20969 * config/s390/s390.c (struct s390_processor processor_table):
20970 Binutils name string must not be empty.
20972 2021-03-09 Claudiu Zissulescu <claziss@synopsys.com>
20974 * config/arc/arc.c (arc_attr_type): Remove function.
20976 2021-03-09 Martin Liska <mliska@suse.cz>
20979 * config/i386/i386-options.c (ix86_option_override_internal):
20980 Set isa_flags for OPTS argument and not for the global
20983 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
20985 * config/rs6000/predicates.md (ds_form_mem_operand): Check
20988 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
20991 * config/rs6000/predicates.md (ds_form_mem_operand) New
20993 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10) Use
20994 ds_form_mem_operand in ld/lwa patterns.
20995 * config/rs6000/fusion.md: Regenerate file.
20997 2021-03-08 Martin Sebor <msebor@redhat.com>
20999 PR middle-end/98266
21000 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): New function.
21001 (array_bounds_checker::check_array_bounds): Call it.
21003 2021-03-08 Martin Sebor <msebor@redhat.com>
21005 PR middle-end/97631
21006 * tree-ssa-strlen.c (maybe_warn_overflow): Test rawmem.
21007 (handle_builtin_stxncpy_strncat): Rename locals. Determine
21008 destination size from allocation calls. Issue a more appropriate
21010 (handle_builtin_memcpy): Pass true as rawmem to maybe_warn_overflow.
21011 (handle_builtin_memset): Same.
21013 2021-03-08 Peter Bergner <bergner@linux.ibm.com>
21016 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_permute): Add an assert
21017 to ensure we do not have an Altivec style address.
21018 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): Disable if passed
21019 an Altivec style address.
21020 (*vsx_le_perm_store_<mode>): Likewise.
21021 (splitters after *vsx_le_perm_store_<mode>): Likewise.
21022 (vsx_load_<mode>): Disable special expander if passed an Altivec
21024 (vsx_store_<mode>): Likewise.
21026 2021-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21029 * config/aarch64/predicates.md (aarch64_simd_shift_imm_vec_qi): Define.
21030 (aarch64_simd_shift_imm_vec_hi): Likewise.
21031 (aarch64_simd_shift_imm_vec_si): Likewise.
21032 (aarch64_simd_shift_imm_vec_di): Likewise.
21033 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le): Use
21034 predicate from above.
21035 (aarch64_shrn<mode>_insn_be): Likewise.
21036 (aarch64_rshrn<mode>_insn_le): Likewise.
21037 (aarch64_rshrn<mode>_insn_be): Likewise.
21038 (aarch64_shrn2<mode>_insn_le): Likewise.
21039 (aarch64_shrn2<mode>_insn_be): Likewise.
21040 (aarch64_rshrn2<mode>_insn_le): Likewise.
21041 (aarch64_rshrn2<mode>_insn_be): Likewise.
21043 2021-03-08 Vladimir N. Makarov <vmakarov@redhat.com>
21046 * lra-constraints.c (skip_contraint_modifiers): New function.
21047 (process_address_1): Use it before lookup_constraint call.
21049 2021-03-08 Martin Liska <mliska@suse.cz>
21052 * config/i386/i386-options.c (ix86_option_override_internal):
21053 Enable UINTR and HRESET for -march that supports it.
21055 2021-03-08 Ilya Leoshkevich <iii@linux.ibm.com>
21057 * config/s390/s390.c (f_constraint_p): New function.
21058 (s390_md_asm_adjust): Implement TARGET_MD_ASM_ADJUST.
21059 (TARGET_MD_ASM_ADJUST): Likewise.
21061 2021-03-08 Tobias Burnus <tobias@codesourcery.com>
21064 * tree-nested.c (convert_local_reference_stmt): Avoid calling
21065 lookup_field_for_decl for Fortran module (= namespace context).
21067 2021-03-08 Andreas Krebbel <krebbel@linux.ibm.com>
21069 * config/s390/s390.c (s390_expand_vec_compare): Implement <0
21070 comparison with arithmetic right shift.
21071 (s390_expand_vcond): No need for a force_reg anymore.
21072 s390_vec_compare will do it.
21073 * config/s390/vector.md ("vec_cmp<mode><tointvec>"): Accept also
21074 immediate operands.
21076 2021-03-07 Jakub Jelinek <jakub@redhat.com>
21079 * config/i386/constraints.md (Yw): Use SSE_REGS if TARGET_SSE
21080 but TARGET_AVX512BW or TARGET_AVX512VL is not set. Adjust description
21082 * config/i386/sse.md (v_Yw): New define_mode_attr.
21083 (*<insn><mode>3, *mul<mode>3<mask_name>, *avx2_<code><mode>3,
21084 *sse4_1_<code><mode>3<mask_name>): Use <v_Yw> instead of v
21086 * config/i386/mmx.md (mmx_pshufw_1, *vec_dupv4hi): Use Yw instead of
21087 xYw in constraints.
21089 2021-03-06 Julian Brown <julian@codesourcery.com>
21091 * tree-pretty-print.c (dump_generic_node): Emit non-generic
21092 address space info for aggregates.
21094 2021-03-06 Hans-Peter Nilsson <hp@axis.com>
21096 * config/cris/cris.h (MAX_FIXED_MODE_SIZE): Don't define.
21098 2021-03-05 Jakub Jelinek <jakub@redhat.com>
21100 PR middle-end/99322
21101 * tree-cfg.c (bb_to_omp_idx): New variable.
21102 (execute_build_cfg): Release the bb_to_omp_idx vector after
21103 cleanup_tree_cfg returns.
21104 (handle_abnormal_edges): Remove bb_to_omp_idx argument, adjust
21105 for bb_to_omp_idx being a vec<int> instead of pointer to array
21107 (make_edges): Remove bb_to_omp_idx local variable, don't pass
21108 it to handle_abnormal_edges, adjust for bb_to_omp_idx being a
21109 vec<int> instead of pointer to array of ints and don't free/release
21111 (remove_bb): When removing a bb and placing forced label somewhere
21112 else, ensure it is put into the same OpenMP region during cfg
21113 pass if possible or to entry successor as fallback. Unregister
21114 bb from bb_to_omp_idx.
21116 2021-03-05 Vladimir N. Makarov <vmakarov@redhat.com>
21119 * lra-constraints.c (process_address_1): Skip decomposing address
21120 for asm insn operand with unknown constraint.
21122 2021-03-05 Martin Jambor <mjambor@suse.cz>
21125 * cgraph.c (cgraph_edge::set_call_stmt): Do not update all
21126 corresponding speculative edges if we are about to resolve
21127 sepculation. Make edge direct (and so resolve speculations) before
21128 removing it from call_site_hash.
21129 (cgraph_edge::make_direct): Relax the initial assert to allow calling
21130 the function on speculative direct edges.
21132 2021-03-05 Eric Botcazou <ebotcazou@adacore.com>
21134 PR rtl-optimization/99376
21135 * rtlanal.c (nonzero_bits1) <arithmetic operators>: If the number
21136 of low-order zero bits is too large, set the result to 0 directly.
21138 2021-03-04 Jakub Jelinek <jakub@redhat.com>
21140 PR middle-end/93235
21141 * expmed.c (store_bit_field_using_insv): Return false of xop0 is a
21142 SUBREG and a SUBREG to op_mode can't be created.
21144 2021-03-04 Alex Coplan <alex.coplan@arm.com>
21147 * config/aarch64/aarch64-sve-builtins.cc
21148 (function_resolver::require_vector_type): Handle error_mark_node.
21150 2021-03-04 Ilya Leoshkevich <iii@linux.ibm.com>
21152 * cfgexpand.c (expand_asm_loc): Pass new parameter.
21153 (expand_asm_stmt): Likewise.
21154 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Add new
21156 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
21157 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
21158 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
21159 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
21160 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
21161 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
21162 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
21163 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
21164 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
21165 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
21166 * doc/tm.texi (md_asm_adjust): Likewise.
21167 * target.def (md_asm_adjust): Likewise.
21169 2021-03-04 Richard Biener <rguenther@suse.de>
21171 PR middle-end/97855
21172 * tree-pretty-print.c: Poison pp_printf.
21173 (dump_decl_name): Avoid use of pp_printf.
21174 (dump_block_node): Likewise.
21175 (dump_generic_node): Likewise.
21177 2021-03-04 Martin Sebor <msebor@redhat.com>
21179 PR middle-end/96963
21180 PR middle-end/94655
21181 * builtins.c (handle_array_ref): New helper.
21182 (handle_mem_ref): New helper.
21183 (compute_objsize_r): Factor out ARRAY_REF and MEM_REF handling
21184 into new helper functions. Correct a workaround for vectorized
21187 2021-03-03 Pat Haugen <pthaugen@linux.ibm.com>
21189 * config/rs6000/dfp.md (extendddtd2, trunctddd2, *cmp<mode>_internal1,
21190 floatditd2, ftrunc<mode>2, fix<mode>di2, dfp_ddedpd_<mode>,
21191 dfp_denbcd_<mode>, dfp_dxex_<mode>, dfp_diex_<mode>,
21192 *dfp_sgnfcnc_<mode>, dfp_dscli_<mode>, dfp_dscri_<mode>): Update size
21193 attribute for Power10.
21194 * config/rs6000/mma.md (*movoo): Likewise.
21195 * config/rs6000/rs6000.md (define_attr "size"): Add 256.
21196 (define_mode_attr bits): Add DD/TD modes.
21197 * config/rs6000/sync.md (load_quadpti, store_quadpti, load_lockedpti,
21198 store_conditionalpti): Update size attribute for Power10.
21200 2021-03-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21203 * config/sparc/t-sparc (tree-ssanames.o-warn): Don't error for
21204 -Wuninitialized, -Wmaybe-uninitialized.
21205 (wide-int.o-warn): Likewise.
21207 2021-03-03 Richard Earnshaw <rearnsha@arm.com>
21209 * common/config/arm/arm-common.c: Include configargs.h.
21210 (arm_config_default): New function.
21211 (arm_target_mode): Renamed from arm_target_thumb_only. Handle
21212 processors that do not support Thumb. Take into account the
21213 --with-mode configuration setting for selecting the default.
21214 * config/arm/arm.h (OPTION_DEFAULT_SPECS): Remove entry for 'mode'.
21215 (TARGET_MODE_SPEC_FUNCTIONS): Update for function name change.
21217 2021-03-03 Martin Liska <mliska@suse.cz>
21219 PR gcov-profile/97461
21220 * gcov-io.h (GCOV_PREALLOCATED_KVP): Remove.
21222 2021-03-03 Eric Botcazou <ebotcazou@adacore.com>
21225 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
21226 point back the hard frame pointer to its default location when the
21227 frame is larger than SEH_MAX_FRAME_SIZE.
21229 2021-03-03 Jakub Jelinek <jakub@redhat.com>
21232 * config/i386/predicates.md (logic_operator): New define_predicate.
21233 * config/i386/i386.md (mov + mem using comm arith peephole2):
21234 Punt if operands[1] is EXT_REX_SSE_REGNO_P, AVX512BW is not enabled
21235 and the inner mode is [QH]Imode.
21237 2021-03-03 Jakub Jelinek <jakub@redhat.com>
21240 * dwarf2out.c (dw_loc_list_struct): Add end_entry member.
21241 (new_loc_list): Clear end_entry.
21242 (output_loc_list): Only use DW_LLE_startx_length for -gsplit-dwarf
21243 if HAVE_AS_LEB128, otherwise use DW_LLE_startx_endx. Fix comment
21245 (index_location_lists): For dwarf_version >= 5 without HAVE_AS_LEB128,
21246 initialize also end_entry.
21248 2021-03-03 Jakub Jelinek <jakub@redhat.com>
21251 * cfgrtl.c (fixup_partitions): When changing some bbs from hot to cold
21252 partitions, if in non-layout mode after reorder_blocks also move
21253 affected blocks to ensure a single partition transition.
21255 2021-03-03 Jason Merrill <jason@redhat.com>
21258 * cgraphunit.c (process_function_and_variable_attributes): Don't
21259 warn about flatten on an alias if the target also has it.
21260 * cgraph.h (symtab_node::get_alias_target_tree): New.
21262 2021-03-02 David Edelsohn <dje.gcc@gmail.com>
21264 * config/rs6000/rs6000.md (tls_get_tpointer_internal): Prepend
21265 period to symbol name.
21266 (tls_get_addr_internal<mode>): Same.
21268 2021-03-02 David Malcolm <dmalcolm@redhat.com>
21271 * diagnostic-show-locus.c
21272 (selftest::test_one_liner_many_fixits_2): Fix accidental usage of
21275 2021-03-02 Martin Sebor <msebor@redhat.com>
21277 PR middle-end/99276
21278 * builtins.c (warn_for_access): Remove stray warning text.
21280 2021-03-02 Martin Sebor <msebor@redhat.com>
21282 PR middle-end/99295
21283 * doc/extend.texi (attribute malloc): Reword and clarify nonaliasing
21286 2021-03-02 Jakub Jelinek <jakub@redhat.com>
21289 * dwarf2out.c (output_macinfo_op): Use DW_MACRO_*_str* even with
21290 -gdwarf-5 -gstrict-dwarf. For -gsplit-dwarf -gdwarf-5 use
21291 DW_MACRO_*_strx instead of DW_MACRO_*_strp. Handle
21292 DW_MACRO_define_strx and DW_MACRO_undef_strx.
21293 (save_macinfo_strings): Use DW_MACRO_*_str* even with
21294 -gdwarf-5 -gstrict-dwarf. Handle DW_MACRO_define_strx and
21295 DW_MACRO_undef_strx.
21297 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
21299 * config/s390/s390-builtin-types.def (BT_FN_V4SF_V8HI_UINT): New
21301 (BT_FN_V8HI_V8HI_UINT): Likewise.
21302 (BT_FN_V8HI_V4SF_V4SF_UINT): Likewise.
21303 * config/s390/s390-builtins.def (B_NNPA): New macro definition.
21304 (s390_vclfnhs, s390_vclfnls, s390_vcrnfs, s390_vcfn, s390_vcnf):
21305 New builtin definitions.
21306 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Bump
21307 vector extension version.
21308 * config/s390/s390.c (s390_expand_builtin): Check if builtins are
21309 available with current -march level.
21310 * config/s390/s390.md (UNSPEC_NNPA_VCLFNHS_V8HI)
21311 (UNSPEC_NNPA_VCLFNLS_V8HI, UNSPEC_NNPA_VCRNFS_V8HI)
21312 (UNSPEC_NNPA_VCFN_V8HI, UNSPEC_NNPA_VCNF_V8HI): New constants.
21313 * config/s390/vecintrin.h (vec_extend_to_fp32_hi): New macro.
21314 (vec_extend_to_fp32_lo): Likewise.
21315 (vec_round_from_fp32): Likewise.
21316 (vec_convert_to_fp16): Likewise.
21317 (vec_convert_from_fp16): Likewise.
21318 * config/s390/vx-builtins.md (vclfnhs_v8hi): New insn pattern.
21319 (vclfnls_v8hi): Likewise.
21320 (vcrnfs_v8hi): Likewise.
21321 (vcfn_v8hi): Likewise.
21322 (vcnf_v8hi): Likewise.
21324 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
21326 * common/config/s390/s390-common.c (processor_flags_table): New entry.
21327 * config.gcc: Enable arch14 for --with-arch and --with-tune.
21328 * config/s390/driver-native.c (s390_host_detect_local_cpu): Pick
21329 arch14 for unknown CPU models.
21330 * config/s390/s390-opts.h (enum processor_type): Add PROCESSOR_ARCH14.
21331 * config/s390/s390.c (s390_issue_rate): Add case for PROCESSOR_ARCH14.
21332 (s390_get_sched_attrmask): Likewise.
21333 (s390_get_unit_mask): Likewise.
21334 * config/s390/s390.h (enum processor_flags): Add PF_NNPA and PF_ARCH14.
21335 (TARGET_CPU_ARCH14, TARGET_CPU_ARCH14_P, TARGET_CPU_NNPA)
21336 (TARGET_CPU_NNPA_P, TARGET_ARCH14, TARGET_ARCH14_P, TARGET_NNPA)
21337 (TARGET_NNPA_P): New macro definitions.
21338 * config/s390/s390.md ("cpu_facility", "enabled"): Add arch14 and nnpa.
21339 * config/s390/s390.opt: Add PROCESSOR_ARCH14.
21341 2021-03-02 Jakub Jelinek <jakub@redhat.com>
21343 PR middle-end/95757
21344 * tree-vrp.c (register_edge_assert_for): Remove superfluous ()s around
21345 condition. Call register_edge_assert_for_1 for == 0, != 0, == 1 and
21346 != 1 comparisons if name is lhs of a comparison.
21348 2021-03-01 Iain Sandoe <iain@sandoe.co.uk>
21352 * config/darwin-protos.h (darwin_should_restore_cfa_state): New.
21353 * config/darwin.c (darwin_should_restore_cfa_state): New.
21354 * config/darwin.h (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New.
21355 * doc/tm.texi: Regenerated.
21356 * doc/tm.texi.in: Document TARGET_ASM_SHOULD_RESTORE_CFA_STATE.
21357 * dwarf2cfi.c (connect_traces): If the target requests, restore
21358 the CFA expression after a DW_CFA_restore.
21359 * target.def (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New hook.
21361 2021-03-01 Martin Liska <mliska@suse.cz>
21364 * optc-save-gen.awk: Add 4 more exceptions.
21366 2021-03-01 Nathan Sidwell <nathan@acm.org>
21369 * tree.h (TYPE_ALIGN_RAW): New accessor.
21370 (TYPE_ALIGN): Use it.
21372 2021-03-01 Jan Hubicka <jh@suse.cz>
21375 * ipa-fnsummary.c (compute_fn_summary): Fix sanity check.
21377 2021-03-01 Eric Botcazou <ebotcazou@adacore.com>
21380 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
21381 point the hard frame pointer to the SSE register save area instead
21382 of the general register save area. Perform only minimal adjustment
21383 for small frames if it is initially not correctly aligned.
21384 (ix86_expand_prologue): Remove early saves for a SEH target.
21385 * config/i386/winnt.c (struct seh_frame_state): Document constraint.
21387 2021-02-28 Jakub Jelinek <jakub@redhat.com>
21390 * ipa.c (symbol_table::remove_unreachable_nodes): Fix a comment
21391 typo - referneced -> referenced.
21392 * tree.c (component_ref_size): Fix comment typo -
21393 refernce -> reference.
21394 * tree-ssa-alias.c (access_path_may_continue_p): Fix comment typo -
21395 traling -> trailing.
21396 (aliasing_component_refs_p): Fix comment typos -
21397 refernce -> reference and refernece -> reference and
21398 traling -> trailing.
21399 (nonoverlapping_refs_since_match_p): Fix comment typo -
21400 referneces -> references.
21401 * doc/invoke.texi (--param modref-max-bases): Fix a typo -
21402 referneces -> references.
21404 2021-02-27 Iain Sandoe <iain@sandoe.co.uk>
21406 * config/host-darwin.c (darwin_gt_pch_use_address): Modify
21407 diagnostic message to avoid use of a contraction and format
21410 2021-02-27 Jakub Jelinek <jakub@redhat.com>
21413 * gcse.c (gcse_or_cprop_is_too_expensive): Use %wu instead of
21414 HOST_WIDE_INT_PRINT_UNSIGNED in warning format string.
21415 * ipa-devirt.c (ipa_odr_read_section): Use %wd instead of
21416 HOST_WIDE_INT_PRINT_DEC in inform format string. Fix comment
21419 2021-02-26 Richard Biener <rguenther@suse.de>
21421 PR middle-end/99281
21422 * expr.c (store_field): For calls with return-slot optimization
21423 and addressable return type expand the store directly.
21425 2021-02-26 Richard Biener <rguenther@suse.de>
21428 * builtins.c (warn_string_no_nul): Fix diagnostic formatting.
21430 2021-02-26 Peter Bergner <bergner@linux.ibm.com>
21433 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Replace assert
21436 2021-02-26 Aaron Sawdey <acsawdey@linux.ibm.com>
21438 * config.gcc: Add rs6000-pcrel-opt.o.
21439 * config/rs6000/rs6000-pcrel-opt.c: New file.
21440 * config/rs6000/pcrel-opt.md: New file.
21441 * config/rs6000/predicates.md: Add d_form_memory predicate.
21442 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_PCREL_OPT.
21443 * config/rs6000/rs6000-passes.def: Add pass_pcrel_opt.
21444 * config/rs6000/rs6000-protos.h: Add reg_to_non_prefixed(),
21445 pcrel_opt_valid_mem_p(), output_pcrel_opt_reloc(),
21446 and make_pass_pcrel_opt().
21447 * config/rs6000/rs6000.c (reg_to_non_prefixed): Make global.
21448 (rs6000_option_override_internal): Add pcrel-opt.
21449 (rs6000_delegitimize_address): Support pcrel-opt.
21450 (rs6000_opt_masks): Add pcrel-opt.
21451 (pcrel_opt_valid_mem_p): New function.
21452 (reg_to_non_prefixed): Make global.
21453 (rs6000_asm_output_opcode): Reset prepend_p_to_next_insn.
21454 (output_pcrel_opt_reloc): New function.
21455 * config/rs6000/rs6000.md (loads_extern_addr): New attr.
21456 (pcrel_extern_addr): Set loads_extern_addr.
21457 Add include for pcrel-opt.md.
21458 * config/rs6000/rs6000.opt: Add -mpcrel-opt.
21459 * config/rs6000/t-rs6000: Add rules for pcrel-opt.c and
21462 2021-02-26 YunQiang Su <yunqiang.su@cipunited.com>
21465 * config/mips/mips.c (mips_expand_ext_as_unaligned_load):
21466 If TARGET_64BIT and dest is SUBREG, we check the width, if it
21467 equal to SImode, we use SImode operation, just like what we are
21470 2021-02-26 Marek Polacek <polacek@redhat.com>
21472 * builtins.c (warn_for_access): Fix typos.
21474 2021-02-25 Iain Sandoe <iain@sandoe.co.uk>
21476 * config/aarch64/aarch64.md (<optab>_rol<mode>3): Add a '#'
21477 mark in front of the immediate quantity.
21478 (<optab>_rolsi3_uxtw): Likewise.
21480 2021-02-25 Richard Earnshaw <rearnsha@arm.com>
21483 * config/arm/thumb2.md (nonsecure_call_reg_thumb2_fpcxt): New pattern.
21484 (nonsecure_call_value_reg_thumb2_fpcxt): Likewise.
21485 (nonsecure_call_reg_thumb2): Restrict to using r4 for the callee
21486 address and disable when the FPCXT is not available.
21487 (nonsecure_call_value_reg_thumb2): Likewise.
21489 2021-02-25 Nathan Sidwell <nathan@acm.org>
21492 * doc/invoke.texi (flang-info-module-cmi): Renamed option.
21494 2021-02-25 Tamar Christina <tamar.christina@arm.com>
21496 * tree-vect-slp.c (optimize_load_redistribution_1): Abort on NULL nodes.
21498 2021-02-25 Richard Biener <rguenther@suse.de>
21500 PR tree-optimization/99253
21501 * tree-vect-loop.c (check_reduction_path): First compute
21502 code, then verify out-of-loop uses.
21504 2021-02-25 Jakub Jelinek <jakub@redhat.com>
21507 * match.pd ((T)(A) + CST -> (T)(A + CST)): Add :s to convert.
21509 2021-02-25 Jakub Jelinek <jakub@redhat.com>
21511 PR tree-optimization/80635
21512 * tree-vrp.c (vrp_simplify_cond_using_ranges): Also handle
21513 VIEW_CONVERT_EXPR if modes are the same, innerop is integral and
21514 has mode precision.
21516 2021-02-25 Richard Biener <rguenther@suse.de>
21518 * tree-vect-slp.c (optimize_load_redistribution_1): Delay
21519 load_map population.
21520 (vect_match_slp_patterns_2): Revert part of last change.
21521 (vect_analyze_slp): Do not interleave optimize_load_redistribution
21522 with pattern detection but do it afterwards. Dump the
21523 whole SLP graph after pattern recognition and load
21524 redistribution optimization finished.
21526 2021-02-24 Jakub Jelinek <jakub@redhat.com>
21529 * omp-low.c (struct omp_context): Add teams_nested_p and
21530 nonteams_nested_p members.
21531 (scan_omp_target): Diagnose teams nested inside of target with other
21532 directives strictly nested inside of the same target.
21533 (check_omp_nesting_restrictions): Set ctx->teams_nested_p or
21534 ctx->nonteams_nested_p as needed.
21536 2021-02-24 Vladimir N. Makarov <vmakarov@redhat.com>
21538 PR inline-asm/99123
21539 * lra-constraints.c (uses_hard_regs_p): Don't use decompose_mem_address.
21541 2021-02-24 Hans-Peter Nilsson <hp@axis.com>
21543 * config/cris/cris.c (cris_expand_prologue): Set
21544 current_function_static_stack_size, if flag_stack_usage_info.
21546 2021-02-24 Pat Haugen <pthaugen@linux.ibm.com>
21548 * config/rs6000/rs6000.c (next_insn_prefixed_p): Rename.
21549 (rs6000_final_prescan_insn): Adjust.
21550 (rs6000_asm_output_opcode): Likewise.
21552 2021-02-24 Martin Sebor <msebor@redhat.com>
21554 PR middle-end/97172
21555 * attribs.c (attr_access::free_lang_data): Clear attribute arg spec
21556 from function arguments.
21558 2021-02-24 Tamar Christina <tamar.christina@arm.com>
21560 PR tree-optimization/99220
21561 * tree-vect-slp.c (optimize_load_redistribution_1): Remove
21562 node from cache when it's about to be deleted.
21564 2021-02-24 Jakub Jelinek <jakub@redhat.com>
21566 PR tree-optimization/99225
21567 * fold-const.c (fold_binary_loc) <case NE_EXPR>: In (x & (1 << y)) != 0
21568 to ((x >> y) & 1) != 0 simplifications use build_one_cst instead of
21569 build_int_cst (..., 1). Formatting fixes.
21571 2021-02-24 Tamar Christina <tamar.christina@arm.com>
21573 PR tree-optimization/99149
21574 * tree-vect-slp-patterns.c (vect_detect_pair_op): Don't recreate the
21576 (vect_slp_reset_pattern): Remove.
21577 (complex_fma_pattern::matches): Remove call to vect_slp_reset_pattern.
21578 (complex_mul_pattern::build, complex_fma_pattern::build,
21579 complex_fms_pattern::build): Fix ref counts.
21580 * tree-vect-slp.c (vect_free_slp_tree): Undo SLP only pattern relevancy
21581 when node is being deleted.
21582 (vect_match_slp_patterns_2): Correct result of cache hit on patterns.
21583 (vect_schedule_slp): Invalidate SLP_TREE_REPRESENTATIVE of removed
21585 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize value.
21587 2021-02-24 Matthias Klose <doko@ubuntu.com>
21590 2020-12-07 Matthias Klose <doko@ubuntu.com>
21592 * genextract.c (print_header): Undefine ENABLE_RTL_CHECKING
21593 and ENABLE_RTL_FLAG_CHECKING.
21595 2021-02-24 Richard Biener <rguenther@suse.de>
21598 * builtins.c (fold_builtin_next_arg): Avoid NULL arg.
21600 2021-02-23 Peter Bergner <bergner@linux.ibm.com>
21602 * config/rs6000/mma.md (mma_assemble_pair): Rename from this...
21603 (vsx_assemble_pair): ...to this.
21604 (*mma_assemble_pair): Rename from this...
21605 (*vsx_assemble_pair): ...to this.
21606 (mma_disassemble_pair): Rename from this...
21607 (vsx_disassemble_pair): ...to this.
21608 (*mma_disassemble_pair): Rename from this...
21609 (*vsx_disassemble_pair): ...to this.
21610 * config/rs6000/rs6000-builtin.def (BU_MMA_V2, BU_MMA_V3,
21611 BU_COMPAT): New macros.
21612 (mma_assemble_pair): Rename from this...
21613 (vsx_assemble_pair): ...to this.
21614 (mma_disassemble_pair): Rename from this...
21615 (vsx_disassemble_pair): ...to this.
21616 (mma_assemble_pair): New compatibility built-in.
21617 (mma_disassemble_pair): Likewise.
21618 * config/rs6000/rs6000-call.c (struct builtin_compatibility): New.
21619 (RS6000_BUILTIN_COMPAT): Define.
21620 (bdesc_compat): New.
21621 (mma_expand_builtin): Use VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
21622 (rs6000_gimple_fold_mma_builtin): Use MMA_BUILTIN_DISASSEMBLE_PAIR
21623 and VSX_BUILTIN_ASSEMBLE_PAIR.
21624 (rs6000_init_builtins): Register compatibility built-ins.
21625 (mma_init_builtins): Use VSX_BUILTIN_ASSEMBLE_PAIR,
21626 VSX_BUILTIN_ASSEMBLE_PAIR_INTERNAL, VSX_BUILTIN_DISASSEMBLE_PAIR and
21627 VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
21628 * doc/extend.texi (__builtin_mma_assemble_pair): Rename from this...
21629 (__builtin_vsx_assemble_pair): ...to this.
21630 (__builtin_mma_disassemble_pair): Rename from this...
21631 (__builtin_vsx_disassemble_pair): ...to this.
21633 2021-02-23 Martin Liska <mliska@suse.cz>
21636 * ipa-icf.c (sem_variable::merge): Do not merge 2 variables
21637 with different alignment. That leads to an invalid red zone
21638 size allocated in runtime.
21640 2021-02-23 Jakub Jelinek <jakub@redhat.com>
21642 PR tree-optimization/99204
21643 * fold-const.c (fold_read_from_constant_string): Check that
21644 tree_fits_uhwi_p (index) rather than just that index is INTEGER_CST.
21646 2021-02-23 Segher Boessenkool <segher@kernel.crashing.org>
21647 Kewen Lin <linkw@gcc.gnu.org>
21649 * config/rs6000/rs6000.md (*rotl<mode>3_insert_3): Renamed to...
21650 (rotl<mode>3_insert_3): ...this.
21651 (plus_ior_xor): New code_iterator.
21652 (define_split for GPR rl*imi): New splitter.
21653 * config/rs6000/vsx.md (vsx_init_v4si): Use gen_rotldi3_insert_3
21654 for integer merging.
21656 2021-02-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21658 * config/aarch64/aarch64-tuning-flags.def (cse_sve_vl_constants):
21660 * config/aarch64/aarch64.md (add<mode>3): Force CONST_POLY_INT immediates
21661 into a register when the above is enabled.
21662 * config/aarch64/aarch64.c (neoversev1_tunings):
21663 AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
21664 (aarch64_rtx_costs): Use AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
21666 2021-02-22 Hans-Peter Nilsson <hp@axis.com>
21668 * config/cris/cris.c (cris_print_operand) <'T'>: Change
21669 valid operand from is now an addi mult-value to shift-value.
21670 * config/cris/cris.md (*addi): Change expression of scaled
21671 operand from mult to ashift.
21672 * config/cris/cris.md (*addi_reload): New insn_and_split.
21674 2021-02-22 John David Anglin <danglin@gcc.gnu.org>
21677 * config/pa/pa.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Define as
21678 hook_bool_const_tree_hwi_hwi_const_tree_true.
21679 (pa_asm_output_mi_thunk): Add support for nonzero vcall_offset.
21681 2021-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
21683 PR rtl-optimization/98791
21684 * ira-conflicts.c (process_regs_for_copy): Don't create allocno copies
21685 for unordered modes.
21687 2021-02-22 Martin Liska <mliska@suse.cz>
21689 * tree-inline.c (inline_forbidden_p): Set
21690 inline_forbidden_reason.
21692 2021-02-22 Richard Biener <rguenther@suse.de>
21694 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Dump
21697 2021-02-22 Richard Biener <rguenther@suse.de>
21699 PR tree-optimization/99165
21700 * gimple-ssa-store-merging.c (pass_store_merging::process_store):
21701 Accumulate changed to ret.
21703 2021-02-21 Uros Bizjak <ubizjak@gmail.com>
21706 2020-12-09 Uroš Bizjak <ubizjak@gmail.com>
21708 * config/i386/i386.h (REG_ALLOC_ORDER): Remove
21710 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
21713 * config/s390/vector.md (trunctf<DFP_ALL:mode>2_vr): New
21715 (trunctf<DFP_ALL:mode>2): Likewise.
21716 (trunctdtf2_vr): Likewise.
21717 (trunctdtf2): Likewise.
21718 (extend<DFP_ALL:mode>tf2_vr): Likewise.
21719 (extend<DFP_ALL:mode>tf2): Likewise.
21720 (extendtftd2_vr): Likewise.
21721 (extendtftd2): Likewise.
21723 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
21725 * config/s390/vector.md (*fprx2_to_tf): Rename to fprx2_to_tf,
21726 add memory alternative.
21727 (tf_to_fprx2): New pattern.
21729 2021-02-19 Martin Sebor <msebor@redhat.com>
21732 * attribs.c (init_attr_rdwr_indices): Guard vblist use.
21733 (attr_access::free_lang_data): Remove a spurious test.
21735 2021-02-19 Nathan Sidwell <nathan@acm.org>
21737 * doc/invoke.texi (flang-info-module-read): Document.
21739 2021-02-19 Martin Liska <mliska@suse.cz>
21741 PR translation/99167
21742 * params.opt: Fix typo.
21744 2021-02-19 Richard Biener <rguenther@suse.de>
21746 PR middle-end/99122
21747 * tree-inline.c (inline_forbidden_p): Do not inline functions
21748 with VLA arguments or return value.
21750 2021-02-19 Jakub Jelinek <jakub@redhat.com>
21753 * config/arm/arm.md (*stack_protect_combined_set_insn,
21754 *stack_protect_combined_test_insn): If force_const_mem result
21755 is not valid general operand, force its address into the destination
21758 2021-02-19 Jakub Jelinek <jakub@redhat.com>
21761 * tree-cfg.c (gimple_merge_blocks): If bb a starts with eh landing
21762 pad or non-local label, put FORCED_LABELs from bb b after that label
21763 rather than before it.
21765 2021-02-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
21768 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3): Use
21769 expand_vector_broadcast' to emit the vec_duplicate operand.
21771 2021-02-18 Vladimir N. Makarov <vmakarov@redhat.com>
21773 PR rtl-optimization/96264
21774 * lra-remat.c (reg_overlap_for_remat_p): Check also output insn
21777 2021-02-18 H.J. Lu <hjl.tools@gmail.com>
21780 * varasm.c (get_section): Replace SUPPORTS_SHF_GNU_RETAIN with
21781 looking up the retain attribute.
21782 (resolve_unique_section): Likewise.
21783 (get_variable_section): Likewise.
21784 (switch_to_section): Likewise. Warn when a symbol without the
21785 retain attribute and a symbol with the retain attribute are
21786 placed in the section with the same name, instead of the used
21788 * doc/extend.texi: Document the "retain" attribute.
21790 2021-02-18 Nathan Sidwell <nathan@acm.org>
21793 * doc/invoke.texi (flang-info-include-translate): Document header
21796 2021-02-18 Richard Biener <rguenther@suse.de>
21798 PR middle-end/99122
21799 * ipa-fnsummary.c (analyze_function_body): Set
21800 CIF_FUNCTION_NOT_INLINABLE for VLA parameter calls.
21801 * tree-inline.c (insert_init_debug_bind): Pass NULL for
21802 error_mark_node values.
21803 (force_value_to_type): Do not build V_C_Es for WITH_SIZE_EXPR
21805 (setup_one_parameter): Delay force_value_to_type until when
21808 2021-02-18 Hans-Peter Nilsson <hp@axis.com>
21810 PR tree-optimization/99142
21811 * match.pd (clz cmp 0): Gate replacement on single_use of clz result.
21813 2021-02-18 Jakub Jelinek <jakub@redhat.com>
21815 * wide-int-bitmask.h (wide_int_bitmask::wide_int_bitmask (),
21816 wide_int_bitmask::wide_int_bitmask (uint64_t),
21817 wide_int_bitmask::wide_int_bitmask (uint64_t, uint64_t),
21818 wide_int_bitmask::operator ~ () const,
21819 wide_int_bitmask::operator | (wide_int_bitmask) const,
21820 wide_int_bitmask::operator & (wide_int_bitmask) const): Use constexpr
21822 * config/i386/i386.h (PTA_3DNOW, PTA_3DNOW_A, PTA_64BIT, PTA_ABM,
21823 PTA_AES, PTA_AVX, PTA_BMI, PTA_CX16, PTA_F16C, PTA_FMA, PTA_FMA4,
21824 PTA_FSGSBASE, PTA_LWP, PTA_LZCNT, PTA_MMX, PTA_MOVBE, PTA_NO_SAHF,
21825 PTA_PCLMUL, PTA_POPCNT, PTA_PREFETCH_SSE, PTA_RDRND, PTA_SSE, PTA_SSE2,
21826 PTA_SSE3, PTA_SSE4_1, PTA_SSE4_2, PTA_SSE4A, PTA_SSSE3, PTA_TBM,
21827 PTA_XOP, PTA_AVX2, PTA_BMI2, PTA_RTM, PTA_HLE, PTA_PRFCHW, PTA_RDSEED,
21828 PTA_ADX, PTA_FXSR, PTA_XSAVE, PTA_XSAVEOPT, PTA_AVX512F, PTA_AVX512ER,
21829 PTA_AVX512PF, PTA_AVX512CD, PTA_NO_TUNE, PTA_SHA, PTA_PREFETCHWT1,
21830 PTA_CLFLUSHOPT, PTA_XSAVEC, PTA_XSAVES, PTA_AVX512DQ, PTA_AVX512BW,
21831 PTA_AVX512VL, PTA_AVX512IFMA, PTA_AVX512VBMI, PTA_CLWB, PTA_MWAITX,
21832 PTA_CLZERO, PTA_NO_80387, PTA_PKU, PTA_AVX5124VNNIW, PTA_AVX5124FMAPS,
21833 PTA_AVX512VPOPCNTDQ, PTA_SGX, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES,
21834 PTA_AVX512VBMI2, PTA_VPCLMULQDQ, PTA_AVX512BITALG, PTA_RDPID,
21835 PTA_PCONFIG, PTA_WBNOINVD, PTA_AVX512VP2INTERSECT, PTA_PTWRITE,
21836 PTA_AVX512BF16, PTA_WAITPKG, PTA_MOVDIRI, PTA_MOVDIR64B, PTA_ENQCMD,
21837 PTA_CLDEMOTE, PTA_SERIALIZE, PTA_TSXLDTRK, PTA_AMX_TILE, PTA_AMX_INT8,
21838 PTA_AMX_BF16, PTA_UINTR, PTA_HRESET, PTA_KL, PTA_WIDEKL, PTA_AVXVNNI,
21839 PTA_X86_64_BASELINE, PTA_X86_64_V2, PTA_X86_64_V3, PTA_X86_64_V4,
21840 PTA_CORE2, PTA_NEHALEM, PTA_WESTMERE, PTA_SANDYBRIDGE, PTA_IVYBRIDGE,
21841 PTA_HASWELL, PTA_BROADWELL, PTA_SKYLAKE, PTA_SKYLAKE_AVX512,
21842 PTA_CASCADELAKE, PTA_COOPERLAKE, PTA_CANNONLAKE, PTA_ICELAKE_CLIENT,
21843 PTA_ICELAKE_SERVER, PTA_TIGERLAKE, PTA_SAPPHIRERAPIDS, PTA_ALDERLAKE,
21844 PTA_KNL, PTA_BONNELL, PTA_SILVERMONT, PTA_GOLDMONT, PTA_GOLDMONT_PLUS,
21845 PTA_TREMONT, PTA_KNM): Use constexpr instead of const.
21847 2021-02-18 Jakub Jelinek <jakub@redhat.com>
21849 PR middle-end/99109
21850 * gimple-array-bounds.cc (build_zero_elt_array_type): Rename to ...
21851 (build_printable_array_type): ... this. Add nelts argument. For
21852 overaligned eltype, use TYPE_MAIN_VARIANT (eltype) instead. If
21853 nelts, call build_array_type_nelts.
21854 (array_bounds_checker::check_mem_ref): Use build_printable_array_type
21855 instead of build_zero_elt_array_type and build_array_type_nelts.
21857 2021-02-18 Jakub Jelinek <jakub@redhat.com>
21860 * config/i386/i386.c (distance_non_agu_define): Don't call
21861 extract_insn_cached here.
21862 (ix86_lea_outperforms): Save and restore recog_data around call
21863 to distance_non_agu_define and distance_agu_use.
21864 (ix86_ok_to_clobber_flags): Remove.
21865 (ix86_avoid_lea_for_add): Don't call ix86_ok_to_clobber_flags.
21866 (ix86_avoid_lea_for_addr): Likewise. Adjust function comment.
21867 * config/i386/i386.md (*lea<mode>): Change from define_insn_and_split
21868 into define_insn. Move the splitting to define_peephole2 and
21869 check there using peep2_regno_dead_p if FLAGS_REG is dead.
21871 2021-02-17 Julian Brown <julian@codesourcery.com>
21873 * gimplify.c (gimplify_scan_omp_clauses): Handle ATTACH_DETACH
21876 2021-02-17 Xi Ruoyao <xry111@mengyan1223.wang>
21879 * config/mips/mips.c (mips_symbol_insns): Do not use
21880 MSA_SUPPORTED_MODE_P if mode is MAX_MACHINE_MODE.
21882 2021-02-16 Vladimir N. Makarov <vmakarov@redhat.com>
21884 PR inline-asm/98096
21885 * stmt.c (resolve_operand_name_1): Take inout operands into account
21886 for access to labels by names.
21887 * doc/extend.texi: Describe counting operands for accessing labels.
21889 2021-02-16 Richard Biener <rguenther@suse.de>
21891 PR tree-optimization/38474
21892 * tree-ssa-structalias.c (variable_info::address_taken): New.
21893 (new_var_info): Initialize address_taken.
21894 (process_constraint): Set address_taken.
21895 (solve_constraints): Use the new address_taken flag rather
21896 than is_reg_var for sorting variables.
21897 (dump_constraint): Dump the variable number if the name
21900 2021-02-16 Jakub Jelinek <jakub@redhat.com>
21903 * tree-vect-stmts.c (vectorizable_simd_clone_call): For num_calls != 1
21904 multiply by 4096 and for inbranch by 8192.
21905 * config/i386/i386.c (ix86_simd_clone_usable): For TARGET_AVX512F,
21906 return 3, 2 or 1 for mangle letters 'b', 'c' or 'd'.
21908 2021-02-15 Maya Rashish <coypu@sdf.org>
21910 * config/aarch64/aarch64.c (aarch64_init_builtins):
21911 Call SUBTARGET_INIT_BUILTINS.
21913 2021-02-15 Peter Bergner <bergner@linux.ibm.com>
21915 PR rtl-optimization/98872
21916 * init-regs.c (initialize_uninitialized_regs): Skip initialization
21917 if CONST0_RTX is NULL.
21919 2021-02-15 Richard Sandiford <richard.sandiford@arm.com>
21921 PR rtl-optimization/98863
21922 * rtl-ssa/functions.h (function_info::bb_live_out_info): Delete.
21923 (function_info::build_info): Turn into a declaration, moving the
21924 definition to internals.h.
21925 (function_info::bb_walker): Declare.
21926 (function_info::create_reg_use): Likewise.
21927 (function_info::calculate_potential_phi_regs): Take a build_info
21929 (function_info::place_phis, function_info::create_ebbs): Declare.
21930 (function_info::calculate_ebb_live_in_for_debug): Likewise.
21931 (function_info::populate_backedge_phis): Delete.
21932 (function_info::start_block, function_info::end_block): Declare.
21933 (function_info::populate_phi_inputs): Delete.
21934 (function_info::m_potential_phi_regs): Move information to build_info.
21935 * rtl-ssa/internals.h: New file.
21936 (function_info::bb_phi_info): New class.
21937 (function_info::build_info): Moved from functions.h.
21938 Add a constructor and destructor.
21939 (function_info::build_info::ebb_use): Delete.
21940 (function_info::build_info::ebb_def): Likewise.
21941 (function_info::build_info::bb_live_out): Likewise.
21942 (function_info::build_info::tmp_ebb_live_in_for_debug): New variable.
21943 (function_info::build_info::potential_phi_regs): Likewise.
21944 (function_info::build_info::potential_phi_regs_for_debug): Likewise.
21945 (function_info::build_info::ebb_def_regs): Likewise.
21946 (function_info::build_info::bb_phis): Likewise.
21947 (function_info::build_info::bb_mem_live_out): Likewise.
21948 (function_info::build_info::bb_to_rpo): Likewise.
21949 (function_info::build_info::def_stack): Likewise.
21950 (function_info::build_info::old_def_stack_limit): Likewise.
21951 * rtl-ssa/internals.inl (function_info::build_info::record_reg_def):
21952 Remove the regno argument. Push the previous definition onto the
21953 definition stack where necessary.
21954 * rtl-ssa/accesses.cc: Include internals.h.
21955 * rtl-ssa/changes.cc: Likewise.
21956 * rtl-ssa/blocks.cc: Likewise.
21957 (function_info::build_info::build_info): Define.
21958 (function_info::build_info::~build_info): Likewise.
21959 (function_info::bb_walker): New class.
21960 (function_info::bb_walker::bb_walker): Define.
21961 (function_info::add_live_out_use): Convert a logarithmic-complexity
21962 test into a linear one. Allow the same definition to be passed
21964 (function_info::calculate_potential_phi_regs): Moved from
21965 functions.cc. Take a build_info parameter and store the
21966 information there instead.
21967 (function_info::place_phis): New function.
21968 (function_info::add_entry_block_defs): Update call to record_reg_def.
21969 (function_info::calculate_ebb_live_in_for_debug): New function.
21970 (function_info::add_phi_nodes): Use bb_phis to decide which
21971 registers need phi nodes and initialize ebb_def_regs accordingly.
21972 Do not add degenerate phis here.
21973 (function_info::add_artificial_accesses): Use create_reg_use.
21974 Assert that all definitions are listed in the DF LR sets.
21975 Update call to record_reg_def.
21976 (function_info::record_block_live_out): Record live-out register
21977 values in the phis of successor blocks. Use the live-out set
21978 when processing the last block in an EBB, instead of always
21979 using the live-in sets of successor blocks. AND the live sets
21980 with the set of registers that have been defined in the EBB,
21981 rather than with all potential phi registers. Cope correctly
21982 with branches back to the start of the current EBB.
21983 (function_info::start_block): New function.
21984 (function_info::end_block): Likewise.
21985 (function_info::populate_phi_inputs): Likewise.
21986 (function_info::create_ebbs): Likewise.
21987 (function_info::process_all_blocks): Rewrite into a multi-phase
21989 * rtl-ssa/functions.cc: Include internals.h.
21990 (function_info::calculate_potential_phi_regs): Move to blocks.cc.
21991 (function_info::init_function_data): Remove caller.
21992 * rtl-ssa/insns.cc: Include internals.h
21993 (function_info::create_reg_use): New function. Lazily any
21994 degenerate phis needed by the linear RPO view.
21995 (function_info::record_use): Use create_reg_use. When processing
21996 debug uses, use potential_phi_regs and test it before checking
21997 whether the register is live on entry to the current EBB. Lazily
21998 calculate ebb_live_in_for_debug.
21999 (function_info::record_call_clobbers): Update call to record_reg_def.
22000 (function_info::record_def): Likewise.
22002 2021-02-15 Martin Liska <mliska@suse.cz>
22004 * toplev.c (init_asm_output): Free output of
22005 gen_command_line_string function.
22006 (process_options): Likewise.
22008 2021-02-15 Martin Liska <mliska@suse.cz>
22010 * params.opt: Add 2 missing Param keywords.
22012 2021-02-15 Eric Botcazou <ebotcazou@adacore.com>
22014 * df-core.c (df_worklist_dataflow_doublequeue): Use proper cast.
22016 2021-02-15 Jakub Jelinek <jakub@redhat.com>
22018 PR tree-optimization/99079
22019 * match.pd (A % (pow2pcst << N) -> A & ((pow2pcst << N) - 1)): Remove
22020 useless tree_nop_conversion_p (type, TREE_TYPE (@3)) check. Instead
22021 require both type and TREE_TYPE (@1) to be integral types and either
22022 type having smaller or equal precision, or TREE_TYPE (@1) being
22023 unsigned type, or type being signed type. If TREE_TYPE (@1)
22024 doesn't have wrapping overflow, perform the subtraction of one in
22027 2021-02-14 Jan Hubicka <hubicka@ucw.cz>
22028 Richard Biener <rguether@suse.de>
22031 * ipa-reference.c (ipa_init): Only conditinally initialize
22032 reference_vars_to_consider.
22033 (propagate): Conditionally deninitialize reference_vars_to_consider.
22034 (ipa_reference_write_optimization_summary): Sanity check that
22035 reference_vars_to_consider is not allocated.
22037 2021-02-13 Levy Hsu <admin@levyhsu.com>
22040 * config/riscv/riscv-shorten-memrefs.c (pass_shorten_memrefs): Add
22041 extend parameter to get_si_mem_base_reg declaration.
22042 (get_si_mem_base_reg): Add extend parameter. Set it.
22043 (analyze): Pass extend arg to get_si_mem_base_reg.
22044 (transform): Likewise. Use it when rewriting mems.
22045 * config/riscv/riscv.c (riscv_legitimize_move): Check for subword
22046 loads and emit sign/zero extending load followed by subreg move.
22048 2021-02-13 Jim Wilson <jimw@sifive.com>
22051 * config/riscv/riscv.c (riscv_compressed_lw_address_p): Drop early
22052 exit when !reload_completed. Only perform check for compressed reg
22053 if reload_completed.
22054 (riscv_rtx_costs): In MEM case, when optimizing for size and
22055 shorten memrefs, if not compressible, then increase cost.
22057 2021-02-13 Jakub Jelinek <jakub@redhat.com>
22059 PR rtl-optimization/98439
22060 * recog.c (pass_split_before_regstack::gate): Enable even when
22061 pass_split_before_sched2 is enabled if -fselective-scheduling2 is
22064 2021-02-13 Jakub Jelinek <jakub@redhat.com>
22067 * config/i386/mmx.md (*mmx_pshufd_1): Add a combine splitter for
22068 swap of V2SImode elements in memory into DImode memory rotate by 32.
22070 2021-02-12 Martin Sebor <msebor@redhat.com>
22072 * tree-pretty-print.c (print_generic_expr_to_str): Update comment.
22074 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
22076 * rtl-ssa/accesses.cc (function_info::make_use_available): Use
22077 m_temp_obstack rather than m_obstack to allocate the temporary use.
22079 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
22081 * df-problems.c (df_lr_bb_local_compute): Treat partial definitions
22082 as read-modify operations.
22084 2021-02-12 Richard Biener <rguenther@suse.de>
22086 PR middle-end/38474
22087 * ipa-fnsummary.c (unmodified_parm_1): Only walk when
22088 fbi->aa_walk_budget is bigger than zero. Update
22089 fbi->aa_walk_budget.
22090 (param_change_prob): Likewise.
22091 * ipa-prop.c (detect_type_change_from_memory_writes):
22092 Properly account walk_aliased_vdefs.
22093 (parm_preserved_before_stmt_p): Canonicalize updates.
22094 (parm_ref_data_preserved_p): Likewise.
22095 (parm_ref_data_pass_through_p): Likewise.
22096 (determine_known_aggregate_parts): Account own alias queries.
22098 2021-02-12 Martin Liska <mliska@suse.cz>
22100 * opts-common.c (decode_cmdline_option): Release werror_arg.
22101 * opts.c (gen_producer_string): Release output of
22102 gen_command_line_string.
22104 2021-02-12 Richard Biener <rguenther@suse.de>
22106 PR tree-optimization/38474
22107 * params.opt (-param=max-store-chains-to-track=): New param.
22108 (-param=max-stores-to-track=): Likewise.
22109 * doc/invoke.texi (max-store-chains-to-track): Document.
22110 (max-stores-to-track): Likewise.
22111 * gimple-ssa-store-merging.c (pass_store_merging::m_n_chains):
22113 (pass_store_merging::m_n_stores): Likewise.
22114 (pass_store_merging::terminate_and_process_chain): Update
22115 m_n_stores and m_n_chains.
22116 (pass_store_merging::process_store): Likewise. Terminate
22117 oldest chains if the number of stores or chains get too large.
22118 (imm_store_chain_info::terminate_and_process_chain): Dump
22121 2021-02-11 Eric Botcazou <ebotcazou@adacore.com>
22123 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
22124 the cold section, emit a nop before the directive if the previous
22125 active instruction can throw.
22127 2021-02-11 Peter Bergner <bergner@linux.ibm.com>
22130 * config/rs6000/predicates.md (mma_assemble_input_operand): Restrict
22131 memory addresses that are legal for quad word accesses.
22133 2021-02-11 Andrea Corallo <andrea.corallo@arm.com>
22136 * config/arm/thumb2.md (*doloop_end_internal): Generate
22137 alternative sequence to handle long range branches.
22139 2021-02-11 Joel Hutton <joel.hutton@arm.com>
22141 PR tree-optimization/98772
22142 * optabs-tree.c (supportable_half_widening_operation): New function
22143 to check for supportable V8QI->V8HI widening patterns.
22144 * optabs-tree.h (supportable_half_widening_operation): New function.
22145 * tree-vect-stmts.c (vect_create_half_widening_stmts): New function
22146 to create promotion stmts for V8QI->V8HI widening patterns.
22147 (vectorizable_conversion): Add case for V8QI->V8HI.
22149 2021-02-11 Richard Biener <rguenther@suse.de>
22151 * sparseset.h (SPARSESET_ELT_BITS): Remove.
22152 (SPARSESET_ELT_TYPE): Use unsigned int.
22153 * fwprop.c: Do not include sparseset.h.
22155 2021-02-10 Jakub Jelinek <jakub@redhat.com>
22158 * varasm.c (declare_weak): For -fsyntax-only, allow even
22159 TREE_ASM_WRITTEN function decls.
22161 2021-02-10 Jakub Jelinek <jakub@redhat.com>
22164 * config/i386/sse.md (fix<fixunssuffix>_truncv2sfv2di2,
22165 <insn>v8qiv8hi2, <insn>v8qiv8si2, <insn>v4qiv4si2, <insn>v4hiv4si2,
22166 <insn>v8qiv8di2, <insn>v4qiv4di2, <insn>v2qiv2di2, <insn>v4hiv4di2,
22167 <insn>v2hiv2di2, <insn>v2siv2di2): Force operands[1] into REG before
22168 calling simplify_gen_subreg on it.
22170 2021-02-10 Martin Liska <mliska@suse.cz>
22172 * config/nvptx/nvptx.c (nvptx_option_override): Use
22173 flag_patchable_function_entry instead of the removed
22174 function_entry_patch_area_size.
22176 2021-02-10 Martin Liska <mliska@suse.cz>
22178 PR tree-optimization/99002
22179 PR tree-optimization/99026
22180 * gimple-if-to-switch.cc (if_chain::is_beneficial): Fix memory
22181 leak when adjacent cases are merged.
22182 * tree-switch-conversion.c (switch_decision_tree::analyze_switch_statement): Use
22184 (make_pass_lower_switch): Remove trailing whitespace.
22185 * tree-switch-conversion.h (release_clusters): New.
22187 2021-02-10 Richard Biener <rguenther@suse.de>
22189 PR rtl-optimization/99054
22190 * cfgrtl.c (rtl-optimization/99054): Return an auto_vec.
22191 (fixup_partitions): Adjust.
22192 (rtl_verify_edges): Likewise.
22194 2021-02-10 Jakub Jelinek <jakub@redhat.com>
22196 PR middle-end/99007
22197 * gimplify.c (gimplify_scan_omp_clauses): For MEM_REF on reductions,
22198 temporarily disable gimplify_ctxp->into_ssa around gimplify_expr
22201 2021-02-10 Richard Biener <rguenther@suse.de>
22204 * ipa-pure-const.c (propagate_malloc): Use an auto_vec<>
22207 2021-02-10 Richard Biener <rguenther@suse.de>
22209 PR tree-optimization/99024
22210 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Only
22211 clear loop->aux if it is associated with the destroyed loop_vinfo.
22213 2021-02-10 Martin Liska <mliska@suse.cz>
22215 PR tree-optimization/99002
22216 * gimple-if-to-switch.cc (find_conditions): Fix memory leak
22219 2021-02-10 Martin Liska <mliska@suse.cz>
22222 * ipa-icf.c (sem_item::add_reference): Fix memory leak when
22223 a reference exists.
22225 2021-02-10 Jakub Jelinek <jakub@redhat.com>
22228 * dwarf2out.c (prune_unused_types_walk): Mark DW_TAG_variable DIEs
22229 at class scope for DWARF5+.
22231 2021-02-09 Eric Botcazou <ebotcazou@adacore.com>
22233 PR rtl-optimization/96015
22234 * reorg.c (skip_consecutive_labels): Minor comment tweaks.
22235 (relax_delay_slots): When deleting a jump to the next active
22236 instruction over a barrier, first delete the barrier if the
22237 jump is the only way to reach the target label.
22239 2021-02-09 Andre Vieira <andre.simoesdiasvieira@arm.com>
22241 * config/aarch64/aarch64-cost-tables.h: Add entries for vect.mul.
22242 * config/aarch64/aarch64.c (aarch64_rtx_mult_cost): Use vect.mul for
22243 vector multiplies and vect.alu for SSRA.
22244 * config/arm/aarch-common-protos.h (struct vector_cost_table): Define
22245 vect.mul cost field.
22246 * config/arm/aarch-cost-tables.h: Add entries for vect.mul.
22247 * config/arm/arm.c: Likewise.
22249 2021-02-09 Richard Biener <rguenther@suse.de>
22251 PR tree-optimization/98863
22252 * tree-ssa-sccvn.h (vn_avail::next_undo): Add.
22253 * tree-ssa-sccvn.c (last_pushed_avail): New global.
22254 (rpo_elim::eliminate_push_avail): Chain pushed avails.
22255 (unwind_state::avail_top): Add.
22256 (do_unwind): Rewrite unwinding of avail entries.
22257 (do_rpo_vn): Initialize last_pushed_avail and
22258 avail_top of the undo state.
22260 2021-02-09 Jakub Jelinek <jakub@redhat.com>
22262 PR middle-end/99004
22263 * calls.c (maybe_warn_rdwr_sizes): Change s0 and s1 type from
22264 const char * to char * and free those pointers after use.
22266 2021-02-09 Richard Biener <rguenther@suse.de>
22268 PR tree-optimization/99017
22269 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Allow
22270 zero vector cost entries.
22272 2021-02-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
22274 PR middle-end/98974
22275 * tree-vect-stmts.c (vectorizable_condition): Remove shadow vec_num
22276 parameter in vectorizable_condition.
22278 2021-02-08 Richard Biener <rguenther@suse.de>
22281 * tree.c (walk_tree_1): Walk VECTOR_CST elements.
22283 2021-02-08 Martin Liska <mliska@suse.cz>
22286 * cfgexpand.c (pass_expand::execute): Parse per-function option
22287 flag_patchable_function_entry and use it.
22288 * common.opt: Remove function_entry_patch_area_size and
22289 function_entry_patch_area_start global variables.
22290 * opts.c (parse_and_check_patch_area): New function.
22291 (common_handle_option): Use it.
22292 * opts.h (parse_and_check_patch_area): New function.
22293 * toplev.c (process_options): Parse and use
22294 function_entry_patch_area_size.
22296 2021-02-08 Martin Sebor <msebor@redhat.com>
22298 * doc/extend.texi (attribute malloc): Correct typos.
22300 2021-02-05 Nathan Sidwell <nathan@acm.org>
22303 * gcc.c (driver::maybe_run_linker): Check for input file
22304 accessibility if not linking.
22306 2021-02-05 Richard Biener <rguenther@suse.de>
22308 PR tree-optimization/98855
22309 * tree-vectorizer.h (add_stmt_cost): New overload.
22310 * tree-vect-slp.c (li_cost_vec_cmp): New.
22311 (vect_bb_slp_scalar_cost): Cost individual loop regions
22312 separately. Account for the scalar instance root stmt.
22314 2021-02-05 Tom de Vries <tdevries@suse.de>
22317 * tree-switch-conversion.c (jump_table_cluster::emit): Add loc
22319 (bit_test_cluster::emit): Reuse location_t for newly created
22321 (switch_decision_tree::try_switch_expansion): Preserve
22323 * tree-switch-conversion.h: Change function signatures.
22325 2021-02-05 Jakub Jelinek <jakub@redhat.com>
22328 * config/i386/i386-options.c (m_NONE, m_ALL): Define.
22329 * config/i386/x86-tune.def (X86_TUNE_BRANCH_PREDICTION_HINTS,
22330 X86_TUNE_PROMOTE_QI_REGS): Use m_NONE instead of 0U.
22331 (X86_TUNE_QIMODE_MATH): Use m_ALL instead of ~0U.
22333 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22335 * config/aarch64/aarch64-simd-builtins.def (get_high): Define builtin.
22336 * config/aarch64/aarch64-simd.md (aarch64_get_high<mode>): Define.
22337 * config/aarch64/arm_neon.h (__GET_HIGH): Delete.
22338 (vget_high_f16): Reimplement using new builtin.
22339 (vget_high_f32): Likewise.
22340 (vget_high_f64): Likewise.
22341 (vget_high_p8): Likewise.
22342 (vget_high_p16): Likewise.
22343 (vget_high_p64): Likewise.
22344 (vget_high_s8): Likewise.
22345 (vget_high_s16): Likewise.
22346 (vget_high_s32): Likewise.
22347 (vget_high_s64): Likewise.
22348 (vget_high_u8): Likewise.
22349 (vget_high_u16): Likewise.
22350 (vget_high_u32): Likewise.
22351 (vget_high_u64): Likewise.
22353 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22355 * config/aarch64/aarch64-simd-builtins.def (get_low): Define builtin.
22356 * config/aarch64/aarch64-simd.md (aarch64_get_low<mode>): Define.
22357 * config/aarch64/arm_neon.h (__GET_LOW): Delete.
22358 (vget_low_f16): Reimplement using new builtin.
22359 (vget_low_f32): Likewise.
22360 (vget_low_f64): Likewise.
22361 (vget_low_p8): Likewise.
22362 (vget_low_p16): Likewise.
22363 (vget_low_p64): Likewise.
22364 (vget_low_s8): Likewise.
22365 (vget_low_s16): Likewise.
22366 (vget_low_s32): Likewise.
22367 (vget_low_s64): Likewise.
22368 (vget_low_u8): Likewise.
22369 (vget_low_u16): Likewise.
22370 (vget_low_u32): Likewise.
22371 (vget_low_u64): Likewise.
22373 2021-02-05 Kito Cheng <kito.cheng@sifive.com>
22375 * gcc.c (print_multilib_info): Check all required argument is provided
22378 2021-02-05 liuhongt <hongtao.liu@intel.com>
22381 * config/i386/i386-expand.c (ix86_expand_sse_cmp): Don't
22382 generate integer mask comparison for 128/256-bits vector when
22383 op_true/op_false is NULL_RTX or CONSTM1_RTX/CONST0_RTX. Also
22384 delete redundant !maskcmp condition.
22385 (ix86_expand_int_vec_cmp): Ditto but no redundant deletion
22387 (ix86_expand_sse_movcc): Delete definition of maskcmp, add the
22388 condition directly to if (maskcmp), add extra check for
22389 cmpmode, it should be MODE_INT.
22390 (ix86_expand_fp_vec_cmp): Pass NULL to ix86_expand_sse_cmp's
22391 parameters op_true/op_false.
22392 (ix86_use_mask_cmp_p): New.
22394 2021-02-05 liuhongt <hongtao.liu@intel.com>
22397 * config/i386/x86-tune.def (X86_TUNE_AVX256_UNALIGNED_LOAD_OPTIMAL):
22398 Remove m_GENERIC from ~list.
22399 (X86_TUNE_AVX256_UNALIGNED_STORE_OPTIMAL): Ditto.
22401 2021-02-04 David Malcolm <dmalcolm@redhat.com>
22404 * diagnostic-show-locus.c (compatible_locations_p): Require
22405 locations in the same macro map to be either both from the
22406 macro definition, or both from the macro arguments.
22408 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
22410 * config/aarch64/aarch64-simd-builtins.def: Add
22411 [su]mull_hi_lane[q] builtin generator macros.
22412 * config/aarch64/aarch64-simd.md
22413 (aarch64_<su>mull_hi_lane<mode>_insn): Define.
22414 (aarch64_<su>mull_hi_lane<mode>): Define.
22415 (aarch64_<su>mull_hi_laneq<mode>_insn): Define.
22416 (aarch64_<su>mull_hi_laneq<mode>): Define.
22417 * config/aarch64/arm_neon.h (vmull_high_lane_s16): Use RTL
22418 builtin instead of inline asm.
22419 (vmull_high_lane_s32): Likewise.
22420 (vmull_high_lane_u16): Likewise.
22421 (vmull_high_lane_u32): Likewise.
22422 (vmull_high_laneq_s16): Likewise.
22423 (vmull_high_laneq_s32): Likewise.
22424 (vmull_high_laneq_u16): Likewise.
22425 (vmull_high_laneq_u32): Liekwise.
22427 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
22429 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_hi_n
22430 builtin generator macros.
22431 * config/aarch64/aarch64-simd.md
22432 (aarch64_<su>mull_hi_n<mode>_insn): Define.
22433 (aarch64_<su>mull_hi_n<mode>): Define.
22434 * config/aarch64/arm_neon.h (vmull_high_n_s16): Use RTL builtin
22435 instead of inline asm.
22436 (vmull_high_n_s32): Likewise.
22437 (vmull_high_n_u16): Likewise.
22438 (vmull_high_n_u32): Likewise.
22440 2021-02-04 Richard Biener <rguenther@suse.de>
22442 PR tree-optimization/98855
22443 * tree-vect-loop.c (vectorizable_phi): Do not cost
22444 single-argument PHIs.
22445 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Likewise.
22446 * tree-vect-stmts.c (vectorizable_bswap): Also perform
22447 costing for SLP operation.
22449 2021-02-04 Martin Liska <mliska@suse.cz>
22451 * doc/extend.texi: Mention -mprefer-vector-width in target
22454 2021-02-03 Martin Sebor <msebor@redhat.com>
22456 PR tree-optimization/98937
22457 * tree-ssa-strlen.c (strlen_dom_walker::~strlen_dom_walker): Define.
22458 Flush pointer_query cache.
22460 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
22462 * config/rs6000/genfusion.pl (gen_2logical): Add missing
22463 fixes based on patch review.
22464 * config/rs6000/fusion.md: Regenerate file.
22466 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
22468 * config/rs6000/t-rs6000: Comment out auto generation of
22471 2021-02-03 Andrew Stubbs <ams@codesourcery.com>
22473 * config/gcn/gcn-opts.h (enum processor_type): Add PROCESSOR_GFX908.
22474 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Add gfx908.
22475 (output_file_start): Add gfx908.
22476 * config/gcn/gcn.opt (gpu_type): Add gfx908.
22477 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Add march=gfx908.
22478 (MULTILIB_DIRNAMES): Add gfx908.
22479 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): New define.
22480 (main): Recognize gfx908.
22481 * config/gcn/t-omp-device: Add gfx908.
22483 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22485 * config/aarch64/aarch64-simd-builtins.def: Add
22486 [su]mlsl_hi_lane[q] builtin macro generators.
22487 * config/aarch64/aarch64-simd.md
22488 (aarch64_<su>mlsl_hi_lane<mode>_insn): Define.
22489 (aarch64_<su>mlsl_hi_lane<mode>): Define.
22490 (aarch64_<su>mlsl_hi_laneq<mode>_insn): Define.
22491 (aarch64_<su>mlsl_hi_laneq<mode>): Define.
22492 * config/aarch64/arm_neon.h (vmlsl_high_lane_s16): Use RTL
22493 builtin instead of inline asm.
22494 (vmlsl_high_lane_s32): Likewise.
22495 (vmlsl_high_lane_u16): Likewise.
22496 (vmlsl_high_lane_u32): Likewise.
22497 (vmlsl_high_laneq_s16): Likewise.
22498 (vmlsl_high_laneq_s32): Likewise.
22499 (vmlsl_high_laneq_u16): Likewise.
22500 (vmlsl_high_laneq_u32): Likewise.
22501 (vmlal_high_laneq_u32): Likewise.
22503 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22505 * config/aarch64/aarch64-simd-builtins.def: Add
22506 [su]mlal_hi_lane[q] builtin generator macros.
22507 * config/aarch64/aarch64-simd.md
22508 (aarch64_<su>mlal_hi_lane<mode>_insn): Define.
22509 (aarch64_<su>mlal_hi_lane<mode>): Define.
22510 (aarch64_<su>mlal_hi_laneq<mode>_insn): Define.
22511 (aarch64_<su>mlal_hi_laneq<mode>): Define.
22512 * config/aarch64/arm_neon.h (vmlal_high_lane_s16): Use RTL
22513 builtin instead of inline asm.
22514 (vmlal_high_lane_s32): Likewise.
22515 (vmlal_high_lane_u16): Likewise.
22516 (vmlal_high_lane_u32): Likewise.
22517 (vmlal_high_laneq_s16): Likewise.
22518 (vmlal_high_laneq_s32): Likewise.
22519 (vmlal_high_laneq_u16): Likewise.
22520 (vmlal_high_laneq_u32): Likewise.
22522 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22524 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_hi_n
22525 builtin generator macros.
22526 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_hi_n<mode>_insn):
22528 (aarch64_<su>mlsl_hi_n<mode>): Define.
22529 * config/aarch64/arm_neon.h (vmlsl_high_n_s16): Use RTL builtin
22530 instead of inline asm.
22531 (vmlsl_high_n_s32): Likewise.
22532 (vmlsl_high_n_u16): Likewise.
22533 (vmlsl_high_n_u32): Likewise.
22535 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22537 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_hi_n
22538 builtin generator macros.
22539 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_hi_n<mode>_insn):
22541 (aarch64_<su>mlal_hi_n<mode>): Define.
22542 * config/aarch64/arm_neon.h (vmlal_high_n_s16): Use RTL builtin
22543 instead of inline asm.
22544 (vmlal_high_n_s32): Likewise.
22545 (vmlal_high_n_u16): Likewise.
22546 (vmlal_high_n_u32): Likewise.
22548 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22550 * config/aarch64/aarch64-simd-builtins.def: Add RTL builtin
22552 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal_hi<mode>):
22554 (aarch64_<su>mlal_hi<mode>_insn): This.
22555 (aarch64_<su>mlal_hi<mode>): Define.
22556 * config/aarch64/arm_neon.h (vmlal_high_s8): Use RTL builtin
22557 instead of inline asm.
22558 (vmlal_high_s16): Likewise.
22559 (vmlal_high_s32): Likewise.
22560 (vmlal_high_u8): Likewise.
22561 (vmlal_high_u16): Likewise.
22562 (vmlal_high_u32): Likewise.
22564 2021-02-03 Ilya Leoshkevich <iii@linux.ibm.com>
22566 * lra-spills.c (remove_pseudos): Call lra_update_insn_recog_data()
22567 after calling alter_subreg() on a (mem).
22569 2021-02-03 Martin Liska <mliska@suse.cz>
22572 * lto-streamer-out.c (produce_lto_section): Fill up missing
22574 * lto-streamer.h (struct lto_section): Add _padding field.
22576 2021-02-03 Richard Biener <rguenther@suse.de>
22578 * lto-streamer.c (lto_get_section_name): Free temporary
22580 * tree-loop-distribution.c
22581 (loop_distribution::merge_dep_scc_partitions): Free edge data.
22583 2021-02-03 Jakub Jelinek <jakub@redhat.com>
22585 PR middle-end/97487
22586 * ifcvt.c (noce_can_force_operand): New function.
22587 (noce_emit_move_insn): Use it.
22588 (noce_try_sign_mask): Likewise. Formatting fix.
22590 2021-02-03 Jakub Jelinek <jakub@redhat.com>
22592 PR middle-end/97971
22593 * lra-constraints.c (process_alt_operands): For inline asm, don't call
22594 fatal_insn, but instead return false.
22596 2021-02-03 Jakub Jelinek <jakub@redhat.com>
22598 PR tree-optimization/98287
22599 * config/i386/mmx.md (<insn><mode>3): For shifts don't enable expander
22602 2021-02-03 Tamar Christina <tamar.christina@arm.com>
22604 PR tree-optimization/98928
22605 * tree-vect-loop.c (vect_analyze_loop_2): Change
22606 STMT_VINFO_SLP_VECT_ONLY to STMT_VINFO_SLP_VECT_ONLY_PATTERN.
22607 * tree-vect-slp-patterns.c (complex_pattern::build): Likewise.
22608 * tree-vectorizer.h (STMT_VINFO_SLP_VECT_ONLY_PATTERN): New.
22609 (class _stmt_vec_info): Add slp_vect_pattern_only_p.
22611 2021-02-02 Richard Biener <rguenther@suse.de>
22613 * gimple-loop-interchange.cc (prepare_data_references):
22615 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
22616 * tree-ssa-loop-im.c (hoist_memory_references): Likewise.
22617 * tree-vect-stmts.c (vectorizable_condition): Do not
22619 (vectorizable_comparison): Likewise.
22621 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22623 * config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
22624 * config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
22625 * config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
22626 (vrsqrteq_u32): Likewise.
22628 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22630 * config/aarch64/aarch64-simd-builtins.def (sqxtun2): Define builtin.
22631 * config/aarch64/aarch64-simd.md (aarch64_sqxtun2<mode>_le): Define.
22632 (aarch64_sqxtun2<mode>_be): Likewise.
22633 (aarch64_sqxtun2<mode>): Likewise.
22634 * config/aarch64/arm_neon.h (vqmovun_high_s16): Reimplement using builtin.
22635 (vqmovun_high_s32): Likewise.
22636 (vqmovun_high_s64): Likewise.
22637 * config/aarch64/iterators.md (UNSPEC_SQXTUN2): Define.
22639 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22641 * config/aarch64/aarch64-simd-builtins.def (bfdot_lane, bfdot_laneq): Use
22643 (bfmlalb_lane, bfmlalt_lane, bfmlalb_lane_q, bfmlalt_lane_q): Use FP flags.
22645 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22647 * config/aarch64/aarch64-simd-builtins.def (fcmla_lane0, fcmla_lane90,
22648 fcmla_lane180, fcmla_lane270, fcmlaq_lane0, fcmlaq_lane90, fcmlaq_lane180,
22649 fcmlaq_lane270, scvtf, ucvtf, fcvtzs, fcvtzu, scvtfsi, scvtfdi, ucvtfsi,
22650 ucvtfdi, fcvtzshf, fcvtzuhf, fmlal_lane_low, fmlsl_lane_low,
22651 fmlal_laneq_low, fmlsl_laneq_low, fmlalq_lane_low, fmlslq_lane_low,
22652 fmlalq_laneq_low, fmlslq_laneq_low, fmlal_lane_high, fmlsl_lane_high,
22653 fmlal_laneq_high, fmlsl_laneq_high, fmlalq_lane_high, fmlslq_lane_high,
22654 fmlalq_laneq_high, fmlslq_laneq_high): Use FP flags.
22656 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22658 * config/aarch64/aarch64-builtins.c (FLAG_LOAD): Define.
22659 * config/aarch64/aarch64-simd-builtins.def (ld1x2, ld2, ld3, ld4, ld2r,
22660 ld3r, ld4r, ld1, ld1x3, ld1x4): Use LOAD flags.
22662 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22664 * config/aarch64/aarch64-simd-builtins.def (combine, zip1, zip2,
22665 uzp1, uzp2, trn1, trn2, simd_bsl): Use AUTO_FP flags.
22667 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22669 * config/aarch64/aarch64-simd-builtins.def (clrsb, clz, ctz, popcount,
22670 vec_smult_lane_, vec_smlal_lane_, vec_smult_laneq_, vec_smlal_laneq_,
22671 vec_umult_lane_, vec_umlal_lane_, vec_umult_laneq_, vec_umlal_laneq_,
22672 ashl, sshl, ushl, srshl, urshl, sdot_lane, udot_lane, sdot_laneq,
22673 udot_laneq, usdot_lane, usdot_laneq, sudot_lane, sudot_laneq, ashr,
22674 ashr_simd, lshr, lshr_simd, srshr_n, urshr_n, ssra_n, usra_n, srsra_n,
22675 ursra_n, sshll_n, ushll_n, sshll2_n, ushll2_n, ssri_n, usri_n, ssli_n,
22676 ssli_n, usli_n, bswap, rbit, simd_bsl, eor3q, rax1q, xarq, bcaxq): Use
22677 NONE builtin flags.
22679 2021-02-02 Jakub Jelinek <jakub@redhat.com>
22681 PR tree-optimization/98848
22682 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
22683 STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
22685 2021-02-02 Kito Cheng <kito.cheng@sifive.com>
22688 * expr.c: Check mode before calling store_expr.
22690 2021-02-02 Christophe Lyon <christophe.lyon@linaro.org>
22692 * config/arm/iterators.md (supf): Remove VORNQ_S and VORNQ_U.
22694 * config/arm/mve.md (mve_vornq_s<mode>): New entry for vorn
22695 instruction using expression ior.
22696 (mve_vornq_u<mode>): New expander.
22697 (mve_vornq_f<mode>): Use ior code instead of unspec.
22698 * config/arm/unspecs.md (VORNQ_S, VORNQ_U, VORNQ_F): Remove.
22700 2021-02-02 Alexandre Oliva <oliva@adacore.com>
22702 * tree-nested.c (convert_nonlocal_reference_op): Move
22703 current_function_decl restore after re-gimplification.
22704 (convert_local_reference_op): Likewise.
22706 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22708 * config/aarch64/aarch64-simd-builtins.def (rshrn, rshrn2):
22710 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>_insn_le):
22712 (aarch64_rshrn<mode>_insn_be): Likewise.
22713 (aarch64_rshrn<mode>): Likewise.
22714 (aarch64_rshrn2<mode>_insn_le): Likewise.
22715 (aarch64_rshrn2<mode>_insn_be): Likewise.
22716 (aarch64_rshrn2<mode>): Likewise.
22717 * config/aarch64/aarch64.md (unspec): Add UNSPEC_RSHRN.
22718 * config/aarch64/arm_neon.h (vrshrn_high_n_s16): Reimplement
22720 (vrshrn_high_n_s32): Likewise.
22721 (vrshrn_high_n_s64): Likewise.
22722 (vrshrn_high_n_u16): Likewise.
22723 (vrshrn_high_n_u32): Likewise.
22724 (vrshrn_high_n_u64): Likewise.
22725 (vrshrn_n_s16): Likewise.
22726 (vrshrn_n_s32): Likewise.
22727 (vrshrn_n_s64): Likewise.
22728 (vrshrn_n_u16): Likewise.
22729 (vrshrn_n_u32): Likewise.
22730 (vrshrn_n_u64): Likewise.
22732 2021-02-01 Sergei Trofimovich <siarheit@google.com>
22734 PR tree-optimization/98499
22735 * ipa-modref.c (analyze_ssa_name_flags): treat RVO
22736 conservatively and assume all possible side-effects.
22738 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22740 * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi,
22741 vec_unpacku_hi_): Define builtins.
22742 * config/aarch64/arm_neon.h (vmovl_high_s8): Reimplement using
22744 (vmovl_high_s16): Likewise.
22745 (vmovl_high_s32): Likewise.
22746 (vmovl_high_u8): Likewise.
22747 (vmovl_high_u16): Likewise.
22748 (vmovl_high_u32): Likewise.
22750 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22752 * config/aarch64/aarch64-simd-builtins.def (sabdl, uabdl):
22754 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl<mode>): New
22756 * config/aarch64/aarch64.md (unspec): Define UNSPEC_SABDL,
22758 * config/aarch64/arm_neon.h (vabdl_s8): Reimplemet using
22760 (vabdl_s16): Likewise.
22761 (vabdl_s32): Likewise.
22762 (vabdl_u8): Likewise.
22763 (vabdl_u16): Likewise.
22764 (vabdl_u32): Likewise.
22765 * config/aarch64/iterators.md (ABDL): New int iterator.
22766 (sur): Handle UNSPEC_SABDL, UNSPEC_UABDL.
22768 2021-02-01 Martin Sebor <msebor@redhat.com>
22770 * tree.h (BLOCK_VARS): Add comment.
22771 (BLOCK_SUBBLOCKS): Same.
22772 (BLOCK_SUPERCONTEXT): Same.
22773 (BLOCK_ABSTRACT_ORIGIN): Same.
22774 (inlined_function_outer_scope_p): Same.
22776 2021-02-01 Martin Sebor <msebor@redhat.com>
22778 PR middle-end/97172
22779 * attribs.c (attr_access::free_lang_data): Define new function.
22780 * attribs.h (attr_access::free_lang_data): Declare new function.
22782 2021-02-01 Richard Biener <rguenther@suse.de>
22784 * vec.h (auto_vec::auto_vec): Add memory stat parameters
22786 * bitmap.h (auto_bitmap::auto_bitmap): Likewise.
22788 2021-02-01 Tamar Christina <tamar.christina@arm.com>
22790 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>,
22791 aarch64_<su>mlsl<mode>, aarch64_<su>mlsl_n<mode>): Flip mult operands.
22793 2021-02-01 Richard Biener <rguenther@suse.de>
22795 PR rtl-optimization/98863
22796 * config/i386/i386-features.c (convert_scalars_to_vector):
22797 Set DF_RD_PRUNE_DEAD_DEFS.
22799 2021-01-31 Eric Botcazou <ebotcazou@adacore.com>
22801 * system.h (SIZE_MAX): Define if not already defined.
22803 2021-01-30 Aaron Sawdey <acsawdey@linux.ibm.com>
22805 * config/rs6000/genfusion.pl (gen_2logical): New function to
22806 generate patterns for logical-logical fusion.
22807 * config/rs6000/fusion.md: Regenerated patterns.
22808 * config/rs6000/rs6000-cpus.def: Add
22809 OPTION_MASK_P10_FUSION_2LOGICAL.
22810 * config/rs6000/rs6000.c (rs6000_option_override_internal):
22811 Enable logical-logical fusion for p10.
22812 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2logical.
22814 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
22816 * config/rs6000/rs6000.opt: Add periods to new AIX options.
22818 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
22820 * config/rs6000/rs6000.opt (mabi=vec-extabi): New.
22821 (mabi=vec-default): New.
22822 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
22823 __EXTABI__ for AIX Vector extended ABI.
22824 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print AIX Vector
22826 (conditional_register_usage): If AIX vec_extabi enabled, vs20-vs31
22828 * doc/invoke.texi (PowerPC mabi): Add AIX vec-extabi and vec-default.
22830 2021-01-30 Jakub Jelinek <jakub@redhat.com>
22832 * config/i386/i386-features.c (remove_partial_avx_dependency): Clear
22833 DF_DEFER_INSN_RESCAN after calling df_process_deferred_rescans.
22835 2021-01-29 Vladimir N. Makarov <vmakarov@redhat.com>
22838 * lra-constraints.c (in_class_p): Don't narrow class only for REG
22841 2021-01-29 Will Schmidt <will_schmidt@vnet.ibm.com>
22843 * config/rs6000/rs6000-call.c (rs6000_expand_binup_builtin): Add
22844 clauses for CODE_FOR_vsx_xvcvuxddp_scale and
22845 CODE_FOR_vsx_xvcvsxddp_scale to the parameter checking code.
22847 2021-01-29 Andrew MacLeod <amacleod@redhat.com>
22849 PR tree-optimization/98866
22850 * gimple-range-gori.h (gori_compute:set_range_invariant): New.
22851 * gimple-range-gori.cc (gori_map::set_range_invariant): New.
22852 (gori_map::m_maybe_invariant): Rename from all_outgoing.
22853 (gori_map::gori_map): Rename all_outgoing to m_maybe_invariant.
22854 (gori_map::is_export_p): Ditto.
22855 (gori_map::calculate_gori): Ditto.
22856 (gori_compute::set_range_invariant): New.
22857 * gimple-range.cc (gimple_ranger::range_of_stmt): Set range
22858 invariant for pointers evaluating to [1, +INF].
22860 2021-01-29 Richard Biener <rguenther@suse.de>
22862 PR rtl-optimization/98863
22863 * config/i386/i386-features.c (remove_partial_avx_dependency):
22864 Do not perform DF analysis.
22865 (pass_data_remove_partial_avx_dependency): Remove
22868 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
22870 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_n
22871 builtin generator macros.
22872 * config/aarch64/aarch64-simd.md (aarch64_<su>mull_n<mode>):
22874 * config/aarch64/arm_neon.h (vmull_n_s16): Use RTL builtin
22875 instead of inline asm.
22876 (vmull_n_s32): Likewise.
22877 (vmull_n_u16): Likewise.
22878 (vmull_n_u32): Likewise.
22880 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22882 * config/aarch64/aarch64-simd-builtins.def (sabdl2, uabdl2):
22884 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
22886 (aarch64_<sur>abdl2<mode>): ... This.
22887 (<sur>sadv16qi): Adjust use of above.
22888 * config/aarch64/arm_neon.h (vabdl_high_s8): Reimplement using
22890 (vabdl_high_s16): Likewise.
22891 (vabdl_high_s32): Likewise.
22892 (vabdl_high_u8): Likewise.
22893 (vabdl_high_u16): Likewise.
22894 (vabdl_high_u32): Likewise.
22896 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22898 * config/aarch64/aarch64-simd-builtins.def (sabal2): Define
22900 (uabal2): Likewise.
22901 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal2<mode>): New
22903 * config/aarch64/aarch64.md (unspec): Add UNSPEC_SABAL2 and
22905 * config/aarch64/arm_neon.h (vabal_high_s8): Reimplement using
22907 (vabal_high_s16): Likewise.
22908 (vabal_high_s32): Likewise.
22909 (vabal_high_u8): Likewise.
22910 (vabal_high_u16): Likewise.
22911 (vabal_high_u32): Likewise.
22912 * config/aarch64/iterators.md (ABAL2): New mode iterator.
22913 (sur): Handle UNSPEC_SABAL2, UNSPEC_UABAL2.
22915 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22917 * config/aarch64/aarch64-simd-builtins.def (sabal): Define
22920 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal<mode>_4):
22922 (aarch64_<sur>abal<mode>): ... This
22923 (<sur>sadv16qi): Adust use of the above.
22924 * config/aarch64/arm_neon.h (vabal_s8): Reimplement using
22926 (vabal_s16): Likewise.
22927 (vabal_s32): Likewise.
22928 (vabal_u8): Likewise.
22929 (vabal_u16): Likewise.
22930 (vabal_u32): Likewise.
22932 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22934 * config/aarch64/aarch64-simd-builtins.def (saddlv, uaddlv):
22936 * config/aarch64/aarch64-simd.md (aarch64_<su>addlv<mode>):
22938 * config/aarch64/arm_neon.h (vaddlv_s8): Reimplement using
22940 (vaddlv_s16): Likewise.
22941 (vaddlv_u8): Likewise.
22942 (vaddlv_u16): Likewise.
22943 (vaddlvq_s8): Likewise.
22944 (vaddlvq_s16): Likewise.
22945 (vaddlvq_s32): Likewise.
22946 (vaddlvq_u8): Likewise.
22947 (vaddlvq_u16): Likewise.
22948 (vaddlvq_u32): Likewise.
22949 (vaddlv_s32): Likewise.
22950 (vaddlv_u32): Likewise.
22951 * config/aarch64/iterators.md (VDQV_L): New mode iterator.
22952 (unspec): Add UNSPEC_SADDLV, UNSPEC_UADDLV.
22953 (Vwstype): New mode attribute.
22955 (VWIDE_S): Likewise.
22956 (USADDLV): New int iterator.
22957 (su): Handle UNSPEC_SADDLV, UNSPEC_UADDLV.
22959 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
22961 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_lane[q]
22962 builtin generator macros.
22963 * config/aarch64/aarch64-simd.md (aarch64_vec_<su>mlsl_lane<Qlane>):
22965 * config/aarch64/arm_neon.h (vmlsl_lane_s16): Use RTL builtin
22966 instead of inline asm.
22967 (vmlsl_lane_s32): Likewise.
22968 (vmlsl_lane_u16): Likewise.
22969 (vmlsl_lane_u32): Likewise.
22970 (vmlsl_laneq_s16): Likewise.
22971 (vmlsl_laneq_s32): Likewise.
22972 (vmlsl_laneq_u16): Likewise.
22973 (vmlsl_laneq_u32): Likewise.
22975 2021-01-29 Richard Biener <rguenther@suse.de>
22977 * doc/invoke.texi (--param max-gcse-memory): Document unit
22979 * gcse.c (gcse_or_cprop_is_too_expensive): Adjust.
22980 * params.opt (--param max-gcse-memory): Adjust default and
22981 document unit of size.
22983 2021-01-29 Richard Biener <rguenther@suse.de>
22985 PR rtl-optimization/98863
22986 * gcse.c (gcse_or_cprop_is_too_expensive): Use unsigned
22987 HOST_WIDE_INT for the memory estimate.
22989 2021-01-29 Bin Cheng <bin.cheng@linux.alibaba.com>
22990 Richard Biener <rguenther@suse.de>
22992 PR tree-optimization/97627
22993 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
22994 Do not analyze fake edges.
22996 2021-01-29 Richard Biener <rguenther@suse.de>
22998 PR rtl-optimization/98144
22999 * df.h (df_mir_bb_info): Add con_visited member.
23000 * df-problems.c (df_mir_alloc): Initialize con_visited,
23001 do not fully populate IN and OUT.
23002 (df_mir_reset): Likewise.
23003 (df_mir_confluence_0): Set con_visited.
23004 (df_mir_confluence_n): Properly handle implicitely
23005 fully populated IN and OUT as designated by con_visited
23006 and update con_visited accordingly.
23008 2021-01-29 Jakub Jelinek <jakub@redhat.com>
23011 * config/arm/vec-common.md (mve_vshlq_<supf><mode>,
23012 vashl<mode>3, vashr<mode>3, vlshr<mode>3): Add
23013 && !TARGET_REALLY_IWMMXT to conditions.
23015 2021-01-29 Jakub Jelinek <jakub@redhat.com>
23018 * cfgbuild.c (find_bb_boundaries): Reset debug_insn when seeing
23021 2021-01-28 Marek Polacek <polacek@redhat.com>
23024 * stor-layout.c (finalize_type_size): If we reset TYPE_USER_ALIGN in
23025 the main variant, maybe reset it in its variants too.
23026 * tree.c (check_base_type): Return true only if TYPE_USER_ALIGN match.
23027 (check_aligned_type): Check if TYPE_USER_ALIGN match.
23029 2021-01-28 Christophe Lyon <christophe.lyon@linaro.org>
23032 * config/arm/arm.c (arm_rtx_costs_internal): Adjust cost of vector
23033 of constant zero for comparisons.
23035 2021-01-28 Michael Meissner <meissner@linux.ibm.com>
23037 * config/rs6000/rs6000.c (rs6000_mangle_decl_assembler_name): Add
23038 support for mapping built-in function names for long double
23039 built-in functions if long double is IEEE 128-bit.
23041 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
23043 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_n
23044 builtin generator macros.
23045 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_n<mode>):
23047 * config/aarch64/arm_neon.h (vmlsl_n_s16): Use RTL builtin
23048 instead of inline asm.
23049 (vmlsl_n_s32): Likewise.
23050 (vmlsl_n_u16): Likewise.
23051 (vmlsl_n_u32): Likewise.
23053 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
23055 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_n
23056 builtin generator macros.
23057 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>):
23059 * config/aarch64/arm_neon.h (vmlal_n_s16): Use RTL builtin
23060 instead of inline asm.
23061 (vmlal_n_s32): Likewise.
23062 (vmlal_n_u16): Likewise.
23063 (vmlal_n_u32): Likewise.
23065 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23067 * config/aarch64/aarch64-simd-builtins.def (shrn2): Define
23069 * config/aarch64/aarch64-simd.md (aarch64_shrn2<mode>_insn_le):
23071 (aarch64_shrn2<mode>_insn_be): Likewise.
23072 (aarch64_shrn2<mode>): Likewise.
23073 * config/aarch64/arm_neon.h (vshrn_high_n_s16): Reimlplement
23075 (vshrn_high_n_s32): Likewise.
23076 (vshrn_high_n_s64): Likewise.
23077 (vshrn_high_n_u16): Likewise.
23078 (vshrn_high_n_u32): Likewise.
23079 (vshrn_high_n_u64): Likewise.
23081 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23083 * config/aarch64/aarch64-simd-builtins.def (shrn): Define
23085 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le):
23087 (aarch64_shrn<mode>_insn_be): Likewise.
23088 (aarch64_shrn<mode>): Likewise.
23089 * config/aarch64/arm_neon.h (vshrn_n_s16): Reimplement using
23091 (vshrn_n_s32): Likewise.
23092 (vshrn_n_s64): Likewise.
23093 (vshrn_n_u16): Likewise.
23094 (vshrn_n_u32): Likewise.
23095 (vshrn_n_u64): Likewise.
23096 * config/aarch64/iterators.md (vn_mode): New mode attribute.
23098 2021-01-28 Richard Biener <rguenther@suse.de>
23100 PR rtl-optimization/80960
23101 * dse.c (check_mem_read_rtx): Call get_addr on the
23104 2021-01-28 Xionghu Luo <luoxhu@linux.ibm.com>
23105 David Edelsohn <dje.gcc@gmail.com>
23108 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
23109 Don't generate VIEW_CONVERT_EXPR for fcode ALTIVEC_BUILTIN_VEC_INSERT
23111 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
23113 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Remove the
23114 wrapper call rs6000_expand_vector_set_var for cleanup. Call
23115 rs6000_expand_vector_set_var_p9 and rs6000_expand_vector_set_var_p8
23117 (rs6000_expand_vector_set_var): Delete.
23118 (rs6000_expand_vector_set_var_p9): Make static.
23119 (rs6000_expand_vector_set_var_p8): Make static.
23121 2021-01-28 Xing GUO <higuoxing@gmail.com>
23123 * common/config/riscv/riscv-common.c
23124 (riscv_subset_list::parsing_subset_version): Fix -march option parsing
23125 when `p` extension exists.
23127 2021-01-27 Vladimir N. Makarov <vmakarov@redhat.com>
23129 PR rtl-optimization/97684
23130 * ira.c (ira): Call ira_set_pseudo_classes before
23131 update_equiv_regs when it is necessary.
23133 2021-01-27 Jakub Jelinek <jakub@redhat.com>
23136 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Use
23137 %w0, %w1 and %2 instead of %0, %1 and %2.
23139 2021-01-27 Aaron Sawdey <acsawdey@linux.ibm.com>
23141 * config/rs6000/genfusion.pl: New script to generate
23142 define_insn_and_split patterns so combine can arrange fused
23143 instructions next to each other.
23144 * config/rs6000/fusion.md: New file, generated fused instruction
23145 patterns for combine.
23146 * config/rs6000/predicates.md (const_m1_to_1_operand): New predicate.
23147 (non_update_memory_operand): New predicate.
23148 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_P10_FUSION and
23149 OPTION_MASK_P10_FUSION_LD_CMPI to ISA_3_1_MASKS_SERVER and
23151 * config/rs6000/rs6000-protos.h (address_is_non_pfx_d_or_x): Add
23153 * config/rs6000/rs6000.c (rs6000_option_override_internal):
23154 Automatically set OPTION_MASK_P10_FUSION and
23155 OPTION_MASK_P10_FUSION_LD_CMPI if target is power10.
23156 (rs600_opt_masks): Allow -mpower10-fusion
23157 in function attributes.
23158 (address_is_non_pfx_d_or_x): New function.
23159 * config/rs6000/rs6000.h: Add MASK_P10_FUSION.
23160 * config/rs6000/rs6000.md: Include fusion.md.
23161 * config/rs6000/rs6000.opt: Add -mpower10-fusion
23162 and -mpower10-fusion-ld-cmpi.
23163 * config/rs6000/t-rs6000: Add dependencies involving fusion.md.
23165 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
23167 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal
23168 builtin generator macros.
23169 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal<mode>):
23171 (aarch64_<su>mlal<mode>): This.
23172 * config/aarch64/arm_neon.h (vmlal_s8): Use RTL builtin
23173 instead of inline asm.
23174 (vmlal_s16): Likewise.
23175 (vmlal_s32): Likewise.
23176 (vmlal_u8): Likewise.
23177 (vmlal_u16): Likewise.
23178 (vmlal_u32): Likewise.
23180 2021-01-27 Richard Biener <rguenther@suse.de>
23182 PR tree-optimization/98854
23183 * tree-vect-slp.c (vect_build_slp_tree_2): Also build
23184 PHIs from scalars when the number of CTORs matches the
23185 number of children.
23187 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
23189 * config/aarch64/aarch64-simd-builtins.def: Add mls_n builtin
23191 * config/aarch64/aarch64-simd.md (*aarch64_mls_elt_merge<mode>):
23193 (aarch64_mls_n<mode>): This.
23194 * config/aarch64/arm_neon.h (vmls_n_s16): Use RTL builtin
23196 (vmls_n_s32): Likewise.
23197 (vmls_n_u16): Likewise.
23198 (vmls_n_u32): Likewise.
23199 (vmlsq_n_s16): Likewise.
23200 (vmlsq_n_s32): Likewise.
23201 (vmlsq_n_u16): Likewise.
23202 (vmlsq_n_u32): Likewise.
23204 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
23206 * config/aarch64/aarch64-simd-builtins.def: Add mls builtin
23208 * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather
23210 (vmls_s16): Likewise.
23211 (vmls_s32): Likewise.
23212 (vmls_u8): Likewise.
23213 (vmls_u16): Likewise.
23214 (vmls_u32): Likewise.
23215 (vmlsq_s8): Likewise.
23216 (vmlsq_s16): Likewise.
23217 (vmlsq_s32): Likewise.
23218 (vmlsq_u8): Likewise.
23219 (vmlsq_u16): Likewise.
23220 (vmlsq_u32): Likewise.
23222 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
23224 * config/aarch64/aarch64-simd-builtins.def: Add mla_n builtin
23226 * config/aarch64/aarch64-simd.md (*aarch64_mla_elt_merge<mode>):
23228 (aarch64_mla_n<mode>): This.
23229 * config/aarch64/arm_neon.h (vmla_n_s16): Use RTL builtin
23231 (vmla_n_s32): Likewise.
23232 (vmla_n_u16): Likewise.
23233 (vmla_n_u32): Likewise.
23234 (vmlaq_n_s16): Likewise.
23235 (vmlaq_n_s32): Likewise.
23236 (vmlaq_n_u16): Likewise.
23237 (vmlaq_n_u32): Likewise.
23239 2021-01-27 liuhongt <hongtao.liu@intel.com>
23242 * config/i386/sse.md (sse2_gt<mode>3): Drop !TARGET_XOP in condition.
23243 (*sse2_eq<mode>3): Ditto.
23245 2021-01-27 Jakub Jelinek <jakub@redhat.com>
23247 * tree-pass.h (PROP_trees): Rename to ...
23248 (PROP_gimple): ... this.
23249 * cfgexpand.c (pass_data_expand): Replace PROP_trees with PROP_gimple.
23250 * passes.c (execute_function_dump, execute_function_todo,
23251 execute_one_ipa_transform_pass, execute_one_pass): Likewise.
23252 * varpool.c (ctor_for_folding): Likewise.
23254 2021-01-27 Jakub Jelinek <jakub@redhat.com>
23256 PR tree-optimization/97260
23257 * varpool.c: Include tree-pass.h.
23258 (ctor_for_folding): In GENERIC return DECL_INITIAL for TREE_READONLY
23259 non-TREE_SIDE_EFFECTS automatic variables.
23261 2021-01-26 Paul Fee <paul.f.fee@gmail.com>
23263 * doc/cpp.texi (__cplusplus): Document value for -std=c++23
23265 * doc/invoke.texi: Document -std=c++23 and -std=gnu++23.
23266 * dwarf2out.c (highest_c_language): Recognise C++20 and C++23.
23267 (gen_compile_unit_die): Recognise C++23.
23269 2021-01-26 Jakub Jelinek <jakub@redhat.com>
23272 * dwarf2asm.c (dw2_assemble_integer): Cast DWARF2_ADDR_SIZE to int
23275 2021-01-26 Jakub Jelinek <jakub@redhat.com>
23278 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
23279 Use UINTVAL (shft_amnt) and UINTVAL (mask) instead of INTVAL (shft_amnt)
23280 and INTVAL (mask). Add && INTVAL (mask) > 0 condition.
23282 2021-01-26 Richard Biener <rguenther@suse.de>
23284 * gimple-pretty-print.c (dump_binary_rhs): Handle
23285 VEC_WIDEN_{PLUS,MINUS}_{LO,HI}_EXPR.
23287 2021-01-26 Richard Biener <rguenther@suse.de>
23289 PR middle-end/98726
23290 * tree.h (vector_cst_int_elt): Remove.
23291 * tree.c (vector_cst_int_elt): Use poly_wide_int for computations,
23294 2021-01-26 Andrew Stubbs <ams@codesourcery.com>
23296 * config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
23297 for V64DFmode min/max reductions.
23299 2021-01-26 Jakub Jelinek <jakub@redhat.com>
23301 * dwarf2asm.c (dw2_assemble_integer): Handle size twice as large
23302 as DWARF2_ADDR_SIZE if x is not a scalar int by emitting it as
23303 two halves, one with x and the other with const0_rtx, ordered
23304 depending on endianity.
23306 2021-01-26 Alexandre Oliva <oliva@adacore.com>
23308 * gimplify.c (gimplify_decl_expr): Skip asan marking calls for
23309 temporaries not seen in binding block, and not about to be
23310 added as gimple variables.
23312 2021-01-25 Martin Sebor <msebor@redhat.com>
23315 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust warning text.
23317 2021-01-25 Martin Liska <mliska@suse.cz>
23319 * value-prof.c (get_nth_most_common_value): Use %s instead
23322 2021-01-25 Jakub Jelinek <jakub@redhat.com>
23325 * configure.ac (HAVE_AS_GDWARF_5_DEBUG_FLAG): Only define if
23326 readelf -wi is able to read the emitted .debug_info back.
23327 * configure: Regenerated.
23329 2021-01-25 Martin Liska <mliska@suse.cz>
23331 PR gcov-profile/98739
23332 * common.opt: Add missing sign symbol.
23333 * value-prof.c (get_nth_most_common_value): Restore handling
23334 of PROFILE_REPRODUCIBILITY_PARALLEL_RUNS and
23335 PROFILE_REPRODUCIBILITY_MULTITHREADED.
23337 2021-01-25 Richard Biener <rguenther@suse.de>
23339 PR middle-end/98807
23340 * tree.c (vector_element_bits): Always use precision of
23341 the element type for boolean vectors.
23343 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
23345 * config/rtems.h (STARTFILE_SPEC): Remove qnolinkcmds.
23346 (ENDFILE_SPEC): Evaluate qnolinkcmds.
23348 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
23350 * config/rtems.h (STARTFILE_SPEC): Remove nostdlib and
23351 nostartfiles handling since this is already done by
23352 LINK_COMMAND_SPEC. Evaluate qnolinkcmds.
23353 (ENDFILE_SPEC): Remove nostdlib and nostartfiles handling since this
23354 is already done by LINK_COMMAND_SPEC.
23355 (LIB_SPECS): Remove nostdlib and nodefaultlibs handling since
23356 this is already done by LINK_COMMAND_SPEC. Remove qnolinkcmds
23359 2021-01-25 Jakub Jelinek <jakub@redhat.com>
23362 * fold-const-call.c (host_size_t_cst_p): Renamed to ...
23363 (size_t_cst_p): ... this. Check and store unsigned HOST_WIDE_INT
23364 value rather than host size_t.
23365 (fold_const_call): Change type of s2 from size_t to
23366 unsigned HOST_WIDE_INT. Use size_t_cst_p instead of
23367 host_size_t_cst_p. For strncmp calls, pass MIN (s2, SIZE_MAX)
23368 instead of s2 as last argument.
23370 2021-01-25 Tamar Christina <tamar.christina@arm.com>
23372 * config/arm/iterators.md (rotsplit1, rotsplit2, conj_op, fcmac1,
23373 VCMLA_OP, VCMUL_OP): New.
23374 * config/arm/mve.md (mve_vcmlaq<mve_rot><mode>): Support vec_dup 0.
23375 * config/arm/neon.md (cmul<conj_op><mode>3): New.
23376 * config/arm/unspecs.md (UNSPEC_VCMLA_CONJ, UNSPEC_VCMLA180_CONJ,
23377 UNSPEC_VCMUL_CONJ): New.
23378 * config/arm/vec-common.md (cmul<conj_op><mode>3, arm_vcmla<rot><mode>,
23379 cml<fcmac1><conj_op><mode>4): New.
23381 2021-01-23 Jakub Jelinek <jakub@redhat.com>
23384 * config/rs6000/mmintrin.h (__m64): Add __may_alias__ attribute.
23386 2021-01-22 Jonathan Wright <jonathan.wright@arm.com>
23388 * config/aarch64/aarch64-simd-builtins.def: Add mla builtin
23390 * config/aarch64/arm_neon.h (vmla_s8): Use RTL builtin rather
23392 (vmla_s16): Likewise.
23393 (vmla_s32): Likewise.
23394 (vmla_u8): Likewise.
23395 (vmla_u16): Likewise.
23396 (vmla_u32): Likewise.
23397 (vmlaq_s8): Likewise.
23398 (vmlaq_s16): Likewise.
23399 (vmlaq_s32): Likewise.
23400 (vmlaq_u8): Likewise.
23401 (vmlaq_u16): Likewise.
23402 (vmlaq_u32): Likewise.
23404 2021-01-22 David Malcolm <dmalcolm@redhat.com>
23406 * doc/invoke.texi (GCC_EXTRA_DIAGNOSTIC_OUTPUT): Add @findex
23409 2021-01-22 Jakub Jelinek <jakub@redhat.com>
23412 * dwarf2out.c (output_file_names): For -gdwarf-5, if there are no
23413 filenames to emit, still emit the required 0 index directory and
23414 filename entries that match DW_AT_comp_dir and DW_AT_name of the
23417 2021-01-22 Marek Polacek <polacek@redhat.com>
23420 * doc/invoke.texi: Update C++ ABI Version 15 description.
23422 2021-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23424 PR tree-optimization/98766
23425 * tree-ssa-math-opts.c (convert_mult_to_fma): Use maybe_le when
23426 comparing against type size with param_avoid_fma_max_bits.
23428 2021-01-22 Richard Biener <rguenther@suse.de>
23430 PR middle-end/98793
23431 * tree.c (vector_element_bits): Key single-bit bool vector on
23432 integer mode rather than not vector mode.
23434 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
23437 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
23438 Generate ARRAY_REF(VIEW_CONVERT_EXPR) for P8 and later
23440 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): Update
23441 to call different path for P8 and P9.
23442 (rs6000_expand_vector_set_var_p9): New function.
23443 (rs6000_expand_vector_set_var_p8): New function.
23445 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
23449 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
23450 Ajdust variable index vec_insert from address dereference to
23451 ARRAY_REF(VIEW_CONVERT_EXPR) tree expression.
23452 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
23454 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): New function.
23456 2021-01-22 Martin Liska <mliska@suse.cz>
23458 PR gcov-profile/98739
23459 * profile.c (compute_value_histograms): Drop time profile for
23460 -fprofile-reproducible=multithreaded.
23462 2021-01-22 Nathan Sidwell <nathan@acm.org>
23464 * gcc.c (process_command): Don't check OPT_SPECIAL_input_file
23467 2021-01-22 Richard Biener <rguenther@suse.de>
23469 PR middle-end/98773
23470 * tree-data-ref.c (initalize_matrix_A): Revert previous
23471 change, retaining failing on HOST_WIDE_INT_MIN CHREC_RIGHT.
23473 2021-01-22 Jakub Jelinek <jakub@redhat.com>
23475 PR tree-optimization/90248
23476 * match.pd (X cmp 0.0 ? 1.0 : -1.0 -> copysign(1, +-X),
23477 X cmp 0.0 ? -1.0 : +1.0 -> copysign(1, -+X)): Remove
23479 (X * (X cmp 0.0 ? 1.0 : -1.0) -> +-abs(X),
23480 X * (X cmp 0.0 ? -1.0 : 1.0) -> +-abs(X)): New simplifications.
23482 2021-01-22 Jakub Jelinek <jakub@redhat.com>
23484 PR tree-optimization/98255
23485 * tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
23486 extend index - low_bound from sizetype's precision rather than index
23488 (get_addr_base_and_unit_offset_1): Likewise.
23489 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
23490 * gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
23492 2021-01-22 Richard Biener <rguenther@suse.de>
23494 PR tree-optimization/98786
23495 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Avoid
23496 adding new uses of abnormals. Verify we deal with a conditional
23499 2021-01-22 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23502 * optc-save-gen.awk: Add arm_fp16_format to checked_options.
23504 2021-01-22 liuhongt <hongtao.liu@intel.com>
23508 * config/i386/sse.md (VI_128_256): New mode iterator.
23509 (*avx_cmp<mode>3_1, *avx_cmp<mode>3_2, *avx_cmp<mode>3_3,
23510 *avx_cmp<mode>3_4, *avx2_eq<mode>3, *avx2_pcmp<mode>3_1,
23511 *avx2_pcmp<mode>3_2, *avx2_gt<mode>3): New
23512 define_insn_and_split to lower avx512 vector comparison to avx
23513 version when dest is vector.
23514 (*<avx512>_cmp<mode>3,*<avx512>_cmp<mode>3,*<avx512>_ucmp<mode>3):
23515 define_insn_and_split for negating the comparison result.
23516 * config/i386/predicates.md (float_vector_all_ones_operand):
23518 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Use
23519 general NOT operator without UNSPEC_MASKOP.
23521 2021-01-21 Vladimir N. Makarov <vmakarov@redhat.com>
23523 PR rtl-optimization/98777
23524 * lra-int.h (lra_pmode_pseudo): New extern.
23525 * lra.c (lra_pmode_pseudo): New global.
23527 * lra-eliminations.c (eliminate_regs_in_insn): Use it.
23529 2021-01-21 Ilya Leoshkevich <iii@linux.ibm.com>
23531 * fwprop.c (fwprop_propagation::classify_result): Allow
23532 (subreg (mem)) simplifications.
23534 2021-01-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23536 * config/aarch64/aarch64-simd.md (aarch64_sqdml<SBINQOPS:as>l<mode>):
23538 (aarch64_sqdmlal<mode>): ... This...
23539 (aarch64_sqdmlsl<mode>): ... And this.
23540 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Split into...
23541 (aarch64_sqdmlal_lane<mode>): ... This...
23542 (aarch64_sqdmlsl_lane<mode>): ... And this.
23543 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Split into...
23544 (aarch64_sqdmlsl_laneq<mode>): ... This...
23545 (aarch64_sqdmlal_laneq<mode>): ... And this.
23546 (aarch64_sqdml<SBINQOPS:as>l_n<mode>): Split into...
23547 (aarch64_sqdmlsl_n<mode>): ... This...
23548 (aarch64_sqdmlal_n<mode>): ... And this.
23549 (aarch64_sqdml<SBINQOPS:as>l2<mode>_internal): Split into...
23550 (aarch64_sqdmlal2<mode>_internal): ... This...
23551 (aarch64_sqdmlsl2<mode>_internal): ... And this.
23553 2021-01-21 Christophe Lyon <christophe.lyon@linaro.org>
23555 * config/arm/arm_mve.h (__arm_vcmpneq_s8): Fix return type.
23557 2021-01-21 Andrea Corallo <andrea.corallo@arm.com>
23560 * doc/sourcebuild.texi (arm_thumb2_no_arm_v8_1_lob): Document.
23562 2021-01-21 liuhongt <hongtao.liu@intel.com>
23564 PR rtl-optimization/98694
23565 * regcprop.c (copy_value): If SRC had been assigned a mode
23566 narrower than the copy, we can't link DEST into the chain even
23567 they have same hard_regno_nregs(i.e. HImode/SImode in i386
23570 2021-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23572 * config/aarch64/aarch64-simd.md (aarch64_get_lane<mode>):
23573 Convert to define_insn_and_split. Split into simple move when moving
23576 2021-01-20 Segher Boessenkool <segher@kernel.crashing.org>
23578 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Change assert.
23579 Adjust comment. Simplify code.
23581 2021-01-20 Jakub Jelinek <jakub@redhat.com>
23584 * dwarf2out.c (reset_indirect_string): Also reset indirect strings
23585 with DW_FORM_line_strp form.
23586 (prune_unused_types_update_strings): Don't add into debug_str_hash
23587 indirect strings with DW_FORM_line_strp form.
23588 (adjust_name_comp_dir): New function.
23589 (dwarf2out_finish): Call it on CU DIEs after resetting
23590 debug_line_str_hash.
23592 2021-01-20 Vladimir N. Makarov <vmakarov@redhat.com>
23594 PR rtl-optimization/98722
23595 * lra-eliminations.c (eliminate_regs_in_insn): Check that target
23596 has no 3-op add insn to transform insns containing two pluses.
23598 2021-01-20 Richard Biener <rguenther@suse.de>
23600 * hwint.h (add_hwi): New function.
23601 (mul_hwi): Likewise.
23602 * tree-data-ref.c (initialize_matrix_A): Properly translate
23603 tree constants and avoid HOST_WIDE_INT_MIN.
23604 (lambda_matrix_row_add): Avoid undefined integer overflow
23605 and return true on such overflow.
23606 (lambda_matrix_right_hermite): Handle overflow from
23607 lambda_matrix_row_add gracefully. Simplify previous fix.
23608 (analyze_subscript_affine_affine): Likewise.
23610 2021-01-20 Eugene Rozenfeld <erozen@microsoft.com>
23612 PR tree-optimization/96674
23613 * match.pd: New patterns: x < y || y == XXX_MIN --> x <= y - 1
23614 x >= y && y != XXX_MIN --> x > y - 1
23616 2021-01-20 Richard Sandiford <richard.sandiford@arm.com>
23618 PR tree-optimization/98535
23619 * tree-vect-slp.c (duplicate_and_interleave): Use quick_grow_cleared.
23620 If the high and low permutes are the same, remove the high permutes
23621 from the working set and only continue with the low ones.
23623 2021-01-20 Jakub Jelinek <jakub@redhat.com>
23625 PR tree-optimization/98721
23626 * builtins.c (access_ref::inform_access): Don't assume
23627 SSA_NAME_IDENTIFIER must be non-NULL. Print messages about
23628 object whenever allocfn is NULL, rather than only when DECL_P
23629 is true. Use %qE instead of %qD for that. Formatting fixes.
23631 2021-01-20 Richard Biener <rguenther@suse.de>
23633 PR tree-optimization/98758
23634 * tree-data-ref.c (int_divides_p): Use lambda_int arguments.
23635 (lambda_matrix_right_hermite): Avoid undefinedness with
23636 signed integer abs and multiplication.
23637 (analyze_subscript_affine_affine): Use lambda_int.
23639 2021-01-20 David Malcolm <dmalcolm@redhat.com>
23642 * dwarf2out.c (output_line_info): Rename static variable
23643 "generation", moving it out of the function to...
23644 (output_line_info_generation): New.
23645 (init_sections_and_labels): Likewise, renaming the variable to...
23646 (init_sections_and_labels_generation): New.
23647 (dwarf2out_c_finalize): Reset the new variables.
23649 2021-01-19 Martin Sebor <msebor@redhat.com>
23651 PR middle-end/98664
23652 * tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
23653 all functions, even if they're not declared artificial or inline.
23654 * tree.c (tree_inlined_location): Use macro expansion location
23655 only if scope traversal fails to expose one.
23657 2021-01-19 Richard Sandiford <richard.sandiford@arm.com>
23659 PR rtl-optimization/92294
23660 * alias.c (compare_base_symbol_refs): Take an extra parameter
23661 and add the distance between two symbols to it. Enshrine in
23662 comments that -1 means "either 0 or 1, but we can't tell
23663 which at compile time".
23664 (memrefs_conflict_p): Update call accordingly.
23665 (rtx_equal_for_memref_p): Likewise. Take the distance between symbols
23668 2021-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23670 * config/aarch64/aarch64-simd-builtins.def (sqshl, uqshl,
23671 sqrshl, uqrshl, sqadd, uqadd, sqsub, uqsub, suqadd, usqadd, sqmovn,
23672 uqmovn, sqxtn2, uqxtn2, sqabs, sqneg, sqdmlal, sqdmlsl, sqdmlal_lane,
23673 sqdmlsl_lane, sqdmlal_laneq, sqdmlsl_laneq, sqdmlal_n, sqdmlsl_n,
23674 sqdmlal2, sqdmlsl2, sqdmlal2_lane, sqdmlsl2_lane, sqdmlal2_laneq,
23675 sqdmlsl2_laneq, sqdmlal2_n, sqdmlsl2_n, sqdmull, sqdmull_lane,
23676 sqdmull_laneq, sqdmull_n, sqdmull2, sqdmull2_lane, sqdmull2_laneq,
23677 sqdmull2_n, sqdmulh, sqrdmulh, sqdmulh_lane, sqdmulh_laneq,
23678 sqrdmulh_lane, sqrdmulh_laneq, sqshrun_n, sqrshrun_n, sqshrn_n,
23679 uqshrn_n, sqrshrn_n, uqrshrn_n, sqshlu_n, sqshl_n, uqshl_n, sqrdmlah,
23680 sqrdmlsh, sqrdmlah_lane, sqrdmlsh_lane, sqrdmlah_laneq, sqrdmlsh_laneq,
23681 sqmovun): Use NONE flags.
23683 2021-01-19 Richard Biener <rguenther@suse.de>
23686 * ipa-modref.c (analyze_stmt): Only record a summary for a
23689 2021-01-19 Richard Biener <rguenther@suse.de>
23691 PR middle-end/98638
23692 * tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
23694 2021-01-19 Daniel Hellstrom <daniel@gaisler.com>
23696 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
23697 built-in define __FIX_LEON3FT_TN0018.
23699 2021-01-19 Richard Biener <rguenther@suse.de>
23702 * tree-inline.c (tree_function_versioning): Set input_location
23703 to UNKNOWN_LOCATION throughout the function.
23705 2021-01-19 Tobias Burnus <tobias@codesourcery.com>
23708 * omp-low.c (lower_omp_target): Handle nonpointer is_device_ptr.
23710 2021-01-19 Martin Jambor <mjambor@suse.cz>
23713 * ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
23714 whether non-call exceptions allow removal of a statement.
23715 (isra_analyze_call): Pass the appropriate function to
23716 ssa_name_only_returned_p.
23718 2021-01-19 Geng Qi <gengqi@linux.alibaba.com>
23720 * config/riscv/arch-canonicalize (longext_sort): New function for
23721 sorting 'multi-letter'.
23722 * config/riscv/multilib-generator: Adjusting the loop of 'alt' in
23723 'alts'. The 'arch' may not be the first of 'alts'.
23724 (_expand_combination): Add underline for the 'ext' without '*'.
23725 This is because, a single-letter extension can always be treated well
23726 with a '_' prefix, but it cannot be separated out if it is appended
23729 2021-01-18 Vladimir N. Makarov <vmakarov@redhat.com>
23732 * ira.c (ira): Skip abnormal critical edge splitting.
23734 2021-01-18 Jakub Jelinek <jakub@redhat.com>
23736 PR tree-optimization/98727
23737 * tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
23738 second .MUL_OVERFLOW operand for signed multiplication with overflow
23739 checking if the second operand of multiplication is not constant.
23741 2021-01-18 David Edelsohn <dje.gcc@gmail.com>
23743 * doc/invoke.texi (-gdwarf): TPF defaults to version 2 and AIX
23744 defaults to version 4.
23746 2021-01-18 David Malcolm <dmalcolm@redhat.com>
23748 * attribs.h (fndecl_dealloc_argno): New decl.
23749 * builtins.c (call_dealloc_argno): Split out second half of
23751 (fndecl_dealloc_argno): New.
23752 * doc/extend.texi (Common Function Attributes): Document the
23753 interaction between the analyzer and the malloc attribute.
23754 * doc/invoke.texi (Static Analyzer Options): Likewise.
23756 2021-01-17 David Edelsohn <dje.gcc@gmail.com>
23758 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Override
23759 dwarf_version to 4.
23760 * config/rs6000/aix72.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
23762 2021-01-17 Martin Jambor <mjambor@suse.cz>
23765 * cgraph.c (clone_of_p): Check also former_clone_of as we climb
23768 2021-01-17 Mark Wielaard <mark@klomp.org>
23770 * common.opt (gdwarf-): Init(5).
23771 * doc/invoke.texi (-gdwarf): Document default to 5.
23773 2021-01-16 Kwok Cheung Yeung <kcy@codesourcery.com>
23775 * builtin-types.def
23776 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
23778 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
23779 ...this. Add extra argument.
23780 * gimplify.c (omp_default_clause): Ensure that event handle is
23781 firstprivate in a task region.
23782 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_DETACH.
23783 (gimplify_adjust_omp_clauses): Likewise.
23784 * omp-builtins.def (BUILT_IN_GOMP_TASK): Change function type to
23785 BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR.
23786 * omp-expand.c (expand_task_call): Add GOMP_TASK_FLAG_DETACH to flags
23787 if detach clause specified. Add detach argument when generating
23789 * omp-low.c (scan_sharing_clauses): Setup data environment for detach
23791 (finish_taskreg_scan): Move field for variable containing the event
23792 handle to the front of the struct.
23793 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_DETACH. Fix
23795 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
23796 OMP_CLAUSE_DETACH clause.
23797 (convert_local_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
23798 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_DETACH.
23799 * tree.c (omp_clause_num_ops): Add entry for OMP_CLAUSE_DETACH.
23801 (omp_clause_code_name): Add entry for OMP_CLAUSE_DETACH. Fix
23803 (walk_tree_1): Handle OMP_CLAUSE_DETACH.
23805 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
23807 * config/nios2/t-rtems: Reset all MULTILIB_* variables. Shorten
23808 multilib directory names. Use MULTILIB_REQUIRED instead of
23809 MULTILIB_EXCEPTIONS. Add -mhw-mul -mhw-mulx -mhw-div
23810 -mcustom-fpu-cfg=fph2 multilib.
23812 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
23814 * config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
23815 (nios2_init_fpu_configs): Provide register values for new
23816 -mcustom-fpu-cfg=fph2 option variant.
23817 * doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
23820 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
23822 * config/nios2/nios2.c (nios2_custom_check_insns): Remove
23823 custom instruction warnings.
23825 2021-01-16 Jakub Jelinek <jakub@redhat.com>
23827 PR tree-optimization/96669
23828 * match.pd ((CST << x) & 1 -> x == 0): New simplification.
23830 2021-01-16 Jakub Jelinek <jakub@redhat.com>
23832 PR tree-optimization/96271
23833 * passes.def: Pass false argument to first two pass_cd_dce
23834 instances and true to last instance. Add comment that
23835 last instance rewrites no longer addressed locals.
23836 * tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
23838 (pass_cd_dce::set_pass_param): New method.
23839 (pass_cd_dce::execute): Return TODO_update_address_taken from
23840 last cd_dce instance.
23842 2021-01-15 Carl Love <cel@us.ibm.com>
23844 * config/rs6000/altivec.h (vec_mulh, vec_div, vec_dive, vec_mod):
23846 * config/rs6000/altivec.md (VIlong): Move define to file vsx.md.
23847 * config/rs6000/rs6000-builtin.def (DIVES_V4SI, DIVES_V2DI,
23848 DIVEU_V4SI, DIVEU_V2DI, DIVS_V4SI, DIVS_V2DI, DIVU_V4SI,
23849 DIVU_V2DI, MODS_V2DI, MODS_V4SI, MODU_V2DI, MODU_V4SI,
23850 MULHS_V2DI, MULHS_V4SI, MULHU_V2DI, MULHU_V4SI, MULLD_V2DI):
23851 Add builtin define.
23852 (MULH, DIVE, MOD): Add new BU_P10_OVERLOAD_2 definitions.
23853 * config/rs6000/rs6000-call.c (VSX_BUILTIN_VEC_DIV,
23854 VSX_BUILTIN_VEC_DIVE, P10_BUILTIN_VEC_MOD, P10_BUILTIN_VEC_MULH):
23855 New overloaded definitions.
23856 (builtin_function_type) [P10V_BUILTIN_DIVEU_V4SI,
23857 P10V_BUILTIN_DIVEU_V2DI, P10V_BUILTIN_DIVU_V4SI,
23858 P10V_BUILTIN_DIVU_V2DI, P10V_BUILTIN_MODU_V2DI,
23859 P10V_BUILTIN_MODU_V4SI, P10V_BUILTIN_MULHU_V2DI,
23860 P10V_BUILTIN_MULHU_V4SI]: Add case
23861 statement for builtins.
23862 * config/rs6000/rs6000.md (bits): Add new attribute sizes V4SI, V2DI.
23863 * config/rs6000/vsx.md (VIlong): Moved from config/rs6000/altivec.md.
23864 (UNSPEC_VDIVES, UNSPEC_VDIVEU): New unspec definitions.
23865 (vsx_mul_v2di): Add if TARGET_POWER10 statement.
23866 (vsx_udiv_v2di): Add if TARGET_POWER10 statement.
23867 (dives_<mode>, diveu_<mode>, div<mode>3, uvdiv<mode>3,
23868 mods_<mode>, modu_<mode>, mulhs_<mode>, mulhu_<mode>, mulv2di3):
23869 Add define_insn, mode is VIlong.
23870 * doc/extend.texi (vec_mulh, vec_mul, vec_div, vec_dive, vec_mod):
23871 Add builtin descriptions.
23873 2021-01-15 Eric Botcazou <ebotcazou@adacore.com>
23875 * final.c (final_start_function_1): Reset force_source_line.
23877 2021-01-15 Jakub Jelinek <jakub@redhat.com>
23879 PR tree-optimization/96669
23880 * match.pd (((1 << A) & 1) != 0 -> A == 0,
23881 ((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
23882 possibly different power of two constants and to right shift too.
23884 2021-01-15 Jakub Jelinek <jakub@redhat.com>
23886 PR tree-optimization/96681
23887 * match.pd ((x < 0) ^ (y < 0) to (x ^ y) < 0): New simplification.
23888 ((x >= 0) ^ (y >= 0) to (x ^ y) < 0): Likewise.
23889 ((x < 0) ^ (y >= 0) to (x ^ y) >= 0): Likewise.
23890 ((x >= 0) ^ (y < 0) to (x ^ y) >= 0): Likewise.
23892 2021-01-15 Alexandre Oliva <oliva@adacore.com>
23894 * opts.c (gen_command_line_string): Exclude -dumpbase-ext.
23896 2021-01-15 Tamar Christina <tamar.christina@arm.com>
23898 * config/aarch64/aarch64-simd.md (cml<fcmac1><conj_op><mode>4,
23899 cmul<conj_op><mode>3): New.
23900 * config/aarch64/iterators.md (UNSPEC_FCMUL,
23901 UNSPEC_FCMUL180, UNSPEC_FCMLA_CONJ, UNSPEC_FCMLA180_CONJ,
23902 UNSPEC_CMLA_CONJ, UNSPEC_CMLA180_CONJ, UNSPEC_CMUL, UNSPEC_CMUL180,
23903 FCMLA_OP, FCMUL_OP, conj_op, rotsplit1, rotsplit2, fcmac1, sve_rot1,
23904 sve_rot2, SVE2_INT_CMLA_OP, SVE2_INT_CMUL_OP, SVE2_INT_CADD_OP): New.
23905 (rot): Add UNSPEC_FCMUL, UNSPEC_FCMUL180.
23906 (rot_op): Renamed to conj_op.
23907 * config/aarch64/aarch64-sve.md (cml<fcmac1><conj_op><mode>4,
23908 cmul<conj_op><mode>3): New.
23909 * config/aarch64/aarch64-sve2.md (cml<fcmac1><conj_op><mode>4,
23910 cmul<conj_op><mode>3): New.
23912 2021-01-15 David Malcolm <dmalcolm@redhat.com>
23916 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
23917 Escape the tempfile name when constructing the expected output.
23919 2021-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23921 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlsl_hi<mode>):
23923 (aarch64_<su>mlsl_hi<mode>): ... This.
23924 (aarch64_<su>mlsl_hi<mode>): Define.
23925 (*aarch64_<su>mlsl<mode): Rename to...
23926 (aarch64_<su>mlsl<mode): ... This.
23927 * config/aarch64/aarch64-simd-builtins.def (smlsl, umlsl,
23928 smlsl_hi, umlsl_hi): Define builtins.
23929 * config/aarch64/arm_neon.h (vmlsl_high_s8, vmlsl_high_s16,
23930 vmlsl_high_s32, vmlsl_high_u8, vmlsl_high_u16, vmlsl_high_u32,
23931 vmlsl_s8, vmlsl_s16, vmlsl_s32, vmlsl_u8,
23932 vmlsl_u16, vmlsl_u32): Reimplement with builtins.
23934 2021-01-15 Uroš Bizjak <ubizjak@gmail.com>
23936 * config/i386/i386-c.c (ix86_target_macros):
23937 Use cpp_define_formatted for __SIZEOF_FLOAT80__ definition.
23939 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
23942 * config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
23943 * Makefile.in (RTL_SSA_H): New variable.
23944 * config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
23945 * config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
23946 * config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
23948 * config/aarch64/aarch64-cc-fusion.cc: New file.
23950 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
23952 * recog.h (insn_change_watermark::~insn_change_watermark): Avoid
23953 calling cancel_changes for changes that no longer exist.
23955 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
23957 * rtl-ssa/functions.h (function_info::ref_defs): Rename to...
23958 (function_info::reg_defs): ...this.
23959 * rtl-ssa/member-fns.inl (function_info::ref_defs): Rename to...
23960 (function_info::reg_defs): ...this.
23962 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23965 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
23967 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23970 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23973 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
23975 2021-01-15 Richard Biener <rguenther@suse.de>
23977 PR tree-optimization/96376
23978 * tree-vect-stmts.c (get_load_store_type): Disregard alignment
23979 for VMAT_INVARIANT.
23981 2021-01-15 Martin Liska <mliska@suse.cz>
23983 * doc/install.texi: Document that some tests need pytest module.
23984 * doc/sourcebuild.texi: Likewise.
23986 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23989 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
23991 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23993 * config/arm/mve.md (mve_vshrq_n_s<mode>_imm): New entry.
23994 (mve_vshrq_n_u<mode>_imm): Likewise.
23995 * config/arm/neon.md (vashr<mode>3, vlshr<mode>3): Move to ...
23996 * config/arm/vec-common.md: ... here.
23998 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
24000 * config/arm/mve.md (mve_vshlq_<supf><mode>): Move to
24002 * config/arm/neon.md (vashl<mode>3): Delete.
24003 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): New.
24004 (vasl<mode>3): New expander.
24006 2021-01-15 Richard Biener <rguenther@suse.de>
24008 PR tree-optimization/98685
24009 * tree-vect-slp.c (vect_schedule_slp_node): Refactor handling
24010 of vector extern defs.
24012 2021-01-14 David Malcolm <dmalcolm@redhat.com>
24015 * diagnostic.c (diagnostic_kind_text): Break out this array
24017 (diagnostic_build_prefix): ...here.
24018 (fancy_abort): Detect when diagnostic_initialize has not yet been
24019 called and fall back to a minimal implementation of printing the
24020 ICE, rather than segfaulting in internal_error.
24022 2021-01-14 David Malcolm <dmalcolm@redhat.com>
24024 * diagnostic.c (diagnostic_initialize): Eliminate
24025 parseable_fixits_p in favor of initializing extra_output_kind from
24026 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
24027 (convert_column_unit): New function, split out from...
24028 (diagnostic_converted_column): ...this.
24029 (print_parseable_fixits): Add "column_unit" and "tabstop" params.
24030 Use them to call convert_column_unit on the column values.
24031 (diagnostic_report_diagnostic): Eliminate conditional on
24032 parseable_fixits_p in favor of a switch statement on
24033 extra_output_kind, passing the appropriate values to the new
24034 params of print_parseable_fixits.
24035 (selftest::test_print_parseable_fixits_none): Update for new
24036 params of print_parseable_fixits.
24037 (selftest::test_print_parseable_fixits_insert): Likewise.
24038 (selftest::test_print_parseable_fixits_remove): Likewise.
24039 (selftest::test_print_parseable_fixits_replace): Likewise.
24040 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
24042 (selftest::diagnostic_c_tests): Call it.
24043 * diagnostic.h (enum diagnostics_extra_output_kind): New.
24044 (diagnostic_context::parseable_fixits_p): Delete field in favor
24046 (diagnostic_context::extra_output_kind): ...this new field.
24047 * doc/invoke.texi (Environment Variables): Add
24048 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
24049 * opts.c (common_handle_option): Update handling of
24050 OPT_fdiagnostics_parseable_fixits for change to diagnostic_context
24053 2021-01-14 Tamar Christina <tamar.christina@arm.com>
24055 * tree-vect-slp-patterns.c (class complex_operations_pattern,
24056 complex_operations_pattern::matches,
24057 complex_operations_pattern::recognize,
24058 complex_operations_pattern::build): New.
24059 (slp_patterns): Use it.
24061 2021-01-14 Tamar Christina <tamar.christina@arm.com>
24063 * internal-fn.def (COMPLEX_FMS, COMPLEX_FMS_CONJ): New.
24064 * optabs.def (cmls_optab, cmls_conj_optab): New.
24065 * doc/md.texi: Document them.
24066 * tree-vect-slp-patterns.c (class complex_fms_pattern,
24067 complex_fms_pattern::matches, complex_fms_pattern::recognize,
24068 complex_fms_pattern::build): New.
24070 2021-01-14 Tamar Christina <tamar.christina@arm.com>
24072 * internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
24073 * optabs.def (cmla_optab, cmla_conj_optab): New.
24074 * doc/md.texi: Document them.
24075 * tree-vect-slp-patterns.c (vect_match_call_p,
24076 class complex_fma_pattern, vect_slp_reset_pattern,
24077 complex_fma_pattern::matches, complex_fma_pattern::recognize,
24078 complex_fma_pattern::build): New.
24080 2021-01-14 Tamar Christina <tamar.christina@arm.com>
24082 * internal-fn.def (COMPLEX_MUL, COMPLEX_MUL_CONJ): New.
24083 * optabs.def (cmul_optab, cmul_conj_optab): New.
24084 * doc/md.texi: Document them.
24085 * tree-vect-slp-patterns.c (vect_match_call_complex_mla,
24086 vect_normalize_conj_loc, is_eq_or_top, vect_validate_multiplication,
24087 vect_build_combine_node, class complex_mul_pattern,
24088 complex_mul_pattern::matches, complex_mul_pattern::recognize,
24089 complex_mul_pattern::build): New.
24091 2021-01-14 Tamar Christina <tamar.christina@arm.com>
24093 * tree-vect-slp.c (optimize_load_redistribution_1): New.
24094 (optimize_load_redistribution, vect_is_slp_load_node): New.
24095 (vect_match_slp_patterns): Use it.
24097 2021-01-14 Tamar Christina <tamar.christina@arm.com>
24099 * tree-vect-slp-patterns.c (complex_add_pattern::build):
24102 2021-01-14 Thomas Schwinge <thomas@codesourcery.com>
24104 * config/gcn/mkoffload.c (main): Create an offload image only in
24105 64-bit configurations.
24107 2021-01-14 H.J. Lu <hjl.tools@gmail.com>
24110 * config/i386/i386-options.c (ix86_option_override_internal):
24111 Issue an error for -fcf-protection with CF_BRANCH when compiling
24112 for 32-bit non-TARGET_CMOV targets.
24114 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
24117 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p):
24118 Remove declaration and initialization of shadow variable "ret".
24119 (ix86_option_override_internal): Remove delcaration of
24120 shadow variable "i". Redeclare shadowed variable to unsigned.
24121 * common/config/i386/i386-common.c (pta_size): Redeclare to unsigned.
24122 * config/i386/i386-builtins.c (get_builtin_code_for_version):
24123 Update for redeclaration.
24124 * config/i386/i386.h (pta_size): Ditto.
24126 2021-01-14 Richard Biener <rguenther@suse.de>
24128 PR tree-optimization/98674
24129 * tree-data-ref.c (base_supports_access_fn_components_p): New.
24130 (initialize_data_dependence_relation): For two bases without
24131 possible access fns resort to type size equality when determining
24132 shape compatibility.
24134 2021-01-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24137 * config/arm/arm_neon.h: Replace calls to __builtin_vcge* by
24138 <=, >= operators in vcle and vcge intrinsics respectively.
24139 * config/arm/arm_neon_builtins.def: Remove entry for
24142 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
24145 * config/i386/i386-options.c (ix86_function_specific_save):
24146 Remove redundant assignment to opts->x_ix86_branch_cost.
24147 * config/i386/i386.c (ix86_prefetch_sse):
24148 Rename from x86_prefetch_sse. Update all uses.
24149 * config/i386/i386.h: Update for rename.
24150 * config/i386/i386-options.h: Ditto.
24152 2021-01-14 Jakub Jelinek <jakub@redhat.com>
24155 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3,
24156 *sse4_1_zero_extendv4hiv4si2_3, *sse4_1_zero_extendv2siv2di2_3):
24157 Use Bm instead of m for non-avx. Add isa attribute.
24159 2021-01-14 Jakub Jelinek <jakub@redhat.com>
24161 PR tree-optimization/96688
24162 * match.pd (~(X >> Y) -> ~X >> Y): New simplification if
24163 ~X can be simplified.
24165 2021-01-14 Richard Sandiford <richard.sandiford@arm.com>
24167 * tree-vect-stmts.c (vect_model_load_cost): Account for unused
24168 IFN_LOAD_LANES results.
24170 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24172 * config/aarch64/aarch64-simd.md (aarch64_<su>xtl<mode>):
24174 (aarch64_xtn<mode>): Likewise.
24175 * config/aarch64/aarch64-simd-builtins.def (sxtl, uxtl, xtn):
24178 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
24180 (vmovl_s16): Likewise.
24181 (vmovl_s32): Likewise.
24182 (vmovl_u8): Likewise.
24183 (vmovl_u16): Likewise.
24184 (vmovl_u32): Likewise.
24185 (vmovn_s16): Likewise.
24186 (vmovn_s32): Likewise.
24187 (vmovn_s64): Likewise.
24188 (vmovn_u16): Likewise.
24189 (vmovn_u32): Likewise.
24190 (vmovn_u64): Likewise.
24192 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24194 * config/aarch64/aarch64-simd.md (aarch64_<su>qxtn2<mode>_le):
24196 (aarch64_<su>qxtn2<mode>_be): Likewise.
24197 (aarch64_<su>qxtn2<mode>): Likewise.
24198 * config/aarch64/aarch64-simd-builtins.def (sqxtn2, uqxtn2):
24200 * config/aarch64/iterators.md (SAT_TRUNC): Define code_iterator.
24201 (su): Handle ss_truncate and us_truncate.
24202 * config/aarch64/arm_neon.h (vqmovn_high_s16): Reimplement using
24204 (vqmovn_high_s32): Likewise.
24205 (vqmovn_high_s64): Likewise.
24206 (vqmovn_high_u16): Likewise.
24207 (vqmovn_high_u32): Likewise.
24208 (vqmovn_high_u64): Likewise.
24210 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24212 * config/aarch64/aarch64-simd.md (aarch64_xtn2<mode>_le):
24214 (aarch64_xtn2<mode>_be): Likewise.
24215 (aarch64_xtn2<mode>): Likewise.
24216 * config/aarch64/aarch64-simd-builtins.def (xtn2): Define
24218 * config/aarch64/arm_neon.h (vmovn_high_s16): Reimplement using
24220 (vmovn_high_s32): Likewise.
24221 (vmovn_high_s64): Likewise.
24222 (vmovn_high_u16): Likewise.
24223 (vmovn_high_u32): Likewise.
24224 (vmovn_high_u64): Likewise.
24226 2021-01-13 Stafford Horne <shorne@gmail.com>
24228 * config/or1k/or1k.h (ASM_PREFERRED_EH_DATA_FORMAT): New macro.
24230 2021-01-13 Stafford Horne <shorne@gmail.com>
24232 * config/or1k/linux.h (TARGET_ASM_FILE_END): Define macro.
24234 2021-01-13 Stafford Horne <shorne@gmail.com>
24236 * config/or1k/or1k.h (TARGET_CPU_CPP_BUILTINS): Add builtin
24237 define for __or1k_hard_float__.
24239 2021-01-13 Stafford Horne <shorne@gmail.com>
24241 * config/or1k/or1k.h (NO_PROFILE_COUNTERS): Define as 1.
24242 (PROFILE_HOOK): Define to call _mcount.
24243 (FUNCTION_PROFILER): Change from abort to no-op.
24245 2021-01-13 Jakub Jelinek <jakub@redhat.com>
24247 PR tree-optimization/96691
24248 * match.pd ((~X | C) ^ D -> (X | C) ^ (~D ^ C),
24249 (~X & C) ^ D -> (X & C) ^ (D ^ C)): New simplifications if
24250 (~D ^ C) or (D ^ C) can be simplified.
24252 2021-01-13 Richard Biener <rguenther@suse.de>
24254 PR tree-optimization/92645
24255 * match.pd (BIT_FIELD_REF to conversion): Delay canonicalization
24256 until after vector lowering.
24258 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
24260 * config/aarch64/aarch64-sve.md (fnma<mode>4): Extend from SVE_FULL_I
24262 (@aarch64_pred_fnma<mode>, cond_fnma<mode>, *cond_fnma<mode>_2)
24263 (*cond_fnma<mode>_4, *cond_fnma<mode>_any): Likewise.
24265 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
24267 * config/aarch64/aarch64-sve.md (fma<mode>4): Extend from SVE_FULL_I
24269 (@aarch64_pred_fma<mode>, cond_fma<mode>, *cond_fma<mode>_2)
24270 (*cond_fma<mode>_4, *cond_fma<mode>_any): Likewise.
24272 2021-01-13 Richard Biener <rguenther@suse.de>
24274 PR tree-optimization/92645
24275 * tree-vect-slp.c (vect_build_slp_tree_1): Relax supported
24276 BIT_FIELD_REF argument.
24277 (vect_build_slp_tree_2): Record the desired vector type
24278 on the external vector def.
24279 (vectorizable_slp_permutation): Handle required punning
24280 of existing vector defs.
24282 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
24284 * rtl-ssa/accesses.h (def_lookup): Fix order of comparison results.
24286 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
24288 * config/sh/sh.md (movsf_ie): Remove operands[2] test.
24290 2021-01-13 Samuel Thibault <samuel.thibault@ens-lyon.org>
24292 * config.gcc [$target == *-*-gnu*]: Enable
24293 'default_gnu_indirect_function'.
24295 2021-01-13 Jakub Jelinek <jakub@redhat.com>
24298 * optabs.c (expand_vec_perm_const): Don't force v0 and v1 into
24299 registers before calling targetm.vectorize.vec_perm_const, only after
24301 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const): Handle
24302 two argument permutation when one operand is zero vector and only
24303 after that force operands into registers.
24304 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_1): New
24305 define_insn_and_split pattern.
24306 (*avx512bw_zero_extendv32qiv32hi2_1): Likewise.
24307 (*avx512f_zero_extendv16hiv16si2_1): Likewise.
24308 (*avx2_zero_extendv8hiv8si2_1): Likewise.
24309 (*avx512f_zero_extendv8siv8di2_1): Likewise.
24310 (*avx2_zero_extendv4siv4di2_1): Likewise.
24311 * config/mips/mips.c (mips_vectorize_vec_perm_const): Force operands
24313 * config/arm/arm.c (arm_vectorize_vec_perm_const): Likewise.
24314 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Likewise.
24315 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const): Likewise.
24316 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const): Likewise.
24317 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const): Likewise.
24318 * config/gcn/gcn.c (gcn_vectorize_vec_perm_const): Likewise. Use std::swap.
24320 2021-01-13 Martin Liska <mliska@suse.cz>
24322 PR tree-optimization/98455
24323 * gimple-if-to-switch.cc (condition_info::record_phi_mapping):
24324 Record also virtual PHIs.
24325 (pass_if_to_switch::execute): Return TODO_cleanup_cfg only
24328 2021-01-13 Jonathan Wakely <jwakely@redhat.com>
24330 * doc/invoke.texi (C++ Modules): Fix typos.
24332 2021-01-13 Richard Biener <rguenther@suse.de>
24334 PR tree-optimization/98640
24335 * tree-ssa-sccvn.c (visit_nary_op): Do not try to
24336 handle plus or minus from a truncated operand to be
24339 2021-01-13 Jakub Jelinek <jakub@redhat.com>
24342 * config/i386/i386.md (*btr<mode>_1, *btr<mode>_2): New
24343 define_insn_and_split patterns.
24344 (splitter after *btr<mode>_2): New splitter.
24346 2021-01-13 Martin Liska <mliska@suse.cz>
24349 * cgraphunit.c (analyze_functions): Remove dead code.
24351 2021-01-13 Qian Jianhua <qianjh@cn.fujitsu.com>
24353 * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
24354 * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
24355 (a64fx_regmove_cost, a64fx_vector_cost): New.
24356 (a64fx_tunings): Use the new added cost tables.
24358 2021-01-13 Jakub Jelinek <jakub@redhat.com>
24361 * config/i386/predicates.md (pmovzx_parallel): New predicate.
24362 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): New
24363 define_insn_and_split pattern.
24364 (*sse4_1_zero_extendv4hiv4si2_3): Likewise.
24365 (*sse4_1_zero_extendv2siv2di2_3): Likewise.
24367 2021-01-13 Julian Brown <julian@codesourcery.com>
24369 * config/gcn/gcn.c (gcn_conditional_register_usage): Remove dead code
24370 to fix v0 register.
24372 2021-01-13 Julian Brown <julian@codesourcery.com>
24374 * config/gcn/gcn.c (gcn_md_reorg): Fix case where EXEC reg is live
24377 2021-01-13 Julian Brown <julian@codesourcery.com>
24379 * config/gcn/gcn-valu.md (recip<mode>2<exec>, recip<mode>2): Use unspec
24380 for reciprocal-approximation instructions.
24381 (div<mode>3): Use fused multiply-accumulate operations for reciprocal
24382 refinement and division result.
24383 * config/gcn/gcn.md (UNSPEC_RCP): New unspec constant.
24385 2021-01-13 Julian Brown <julian@codesourcery.com>
24387 * config/gcn/gcn-valu.md (subdf): Rename to...
24390 2021-01-12 Martin Liska <mliska@suse.cz>
24392 * gcov.c (source_info::debug): Fix printf format for 32-bit hosts.
24394 2021-01-12 Andrea Corallo <andrea.corallo@arm.com>
24396 * function-abi.h: Fix typo.
24398 2021-01-12 Christophe Lyon <christophe.lyon@linaro.org>
24402 * config/arm/arm.h (ARM_HAVE_NEON_V8QI_LDST): New macro.
24403 (ARM_HAVE_NEON_V16QI_LDST, ARM_HAVE_NEON_V4HI_LDST): Likewise.
24404 (ARM_HAVE_NEON_V8HI_LDST, ARM_HAVE_NEON_V2SI_LDST): Likewise.
24405 (ARM_HAVE_NEON_V4SI_LDST, ARM_HAVE_NEON_V4HF_LDST): Likewise.
24406 (ARM_HAVE_NEON_V8HF_LDST, ARM_HAVE_NEON_V4BF_LDST): Likewise.
24407 (ARM_HAVE_NEON_V8BF_LDST, ARM_HAVE_NEON_V2SF_LDST): Likewise.
24408 (ARM_HAVE_NEON_V4SF_LDST, ARM_HAVE_NEON_DI_LDST): Likewise.
24409 (ARM_HAVE_NEON_V2DI_LDST): Likewise.
24410 (ARM_HAVE_V8QI_LDST, ARM_HAVE_V16QI_LDST): Likewise.
24411 (ARM_HAVE_V4HI_LDST, ARM_HAVE_V8HI_LDST): Likewise.
24412 (ARM_HAVE_V2SI_LDST, ARM_HAVE_V4SI_LDST, ARM_HAVE_V4HF_LDST): Likewise.
24413 (ARM_HAVE_V8HF_LDST, ARM_HAVE_V4BF_LDST, ARM_HAVE_V8BF_LDST): Likewise.
24414 (ARM_HAVE_V2SF_LDST, ARM_HAVE_V4SF_LDST, ARM_HAVE_DI_LDST): Likewise.
24415 (ARM_HAVE_V2DI_LDST): Likewise.
24416 * config/arm/mve.md (*movmisalign<mode>_mve_store): New pattern.
24417 (*movmisalign<mode>_mve_load): New pattern.
24418 * config/arm/neon.md (movmisalign<mode>): Move to ...
24419 * config/arm/vec-common.md: ... here.
24421 2021-01-12 Vladimir N. Makarov <vmakarov@redhat.com>
24424 * lra-eliminations.c (eliminate_regs_in_insn): Add transformation
24425 of pattern 'plus (plus (hard reg, const), pseudo)'.
24427 2021-01-12 Richard Biener <rguenther@suse.de>
24429 PR tree-optimization/98550
24430 * tree-vect-slp.c (vect_record_max_nunits): Check whether
24431 the group size is a multiple of the vector element count.
24432 (vect_build_slp_tree_1): When we need to fail because
24433 the vector type choosen causes unrolling do so lazily
24434 without affecting matches only at the end to guide group splitting.
24436 2021-01-12 Martin Liska <mliska@suse.cz>
24439 * optc-save-gen.awk: Compare also n_target_save vars with
24442 2021-01-12 Martin Liska <mliska@suse.cz>
24444 * gcov.c (source_info::debug): New.
24445 (print_usage): Add --debug (-D) option.
24446 (process_args): Likewise.
24447 (generate_results): Call src->debug after
24448 accumulate_line_counts.
24449 (read_graph_file): Properly assign id for EXIT_BLOCK.
24450 * profile.c (branch_prob): Dump function body before it is
24453 2021-01-12 Jakub Jelinek <jakub@redhat.com>
24455 PR tree-optimization/98629
24456 * tree-ssa-math-opts.c (arith_overflow_check_p): Don't update use_stmt
24457 unless returning non-zero.
24459 2021-01-12 Jakub Jelinek <jakub@redhat.com>
24461 PR tree-optimization/95731
24462 * tree-ssa-reassoc.c (optimize_range_tests_cmp_bitwise): Also optimize
24463 x < 0 && y < 0 && z < 0 into (x | y | z) < 0 for signed x, y, z.
24464 (optimize_range_tests): Call optimize_range_tests_cmp_bitwise
24465 only after optimize_range_tests_var_bound.
24467 2021-01-12 Jakub Jelinek <jakub@redhat.com>
24469 * configure.ac: Ensure c/Make-lang.in comes first in @all_lang_makefrags@.
24470 * configure: Regenerated.
24472 2021-01-12 liuhongt <hongtao.liu@intel.com>
24475 * config/i386/i386-builtins.h (BUILTIN_DESC_SWAP_OPERANDS):
24477 * config/i386/i386-expand.c (ix86_expand_sse_comi): Delete
24480 2021-01-12 Alexandre Oliva <oliva@adacore.com>
24482 * ssa-iterators.h (end_imm_use_stmt_traverse): Forward
24484 (auto_end_imm_use_stmt_traverse): New struct.
24485 (FOR_EACH_IMM_USE_STMT): Use it.
24486 (BREAK_FROM_IMM_USE_STMT, RETURN_FROM_IMM_USE_STMT): Remove,
24488 * gimple-ssa-strength-reduction.c: ... here, ...
24489 * graphite-scop-detection.c: ... here, ...
24490 * ipa-modref.c, ipa-pure-const.c, ipa-sra.c: ... here, ...
24491 * tree-predcom.c, tree-ssa-ccp.c: ... here, ...
24492 * tree-ssa-dce.c, tree-ssa-dse.c: ... here, ...
24493 * tree-ssa-loop-ivopts.c, tree-ssa-math-opts.c: ... here, ...
24494 * tree-ssa-phiprop.c, tree-ssa.c: ... here, ...
24495 * tree-vect-slp.c: ... and here, ...
24496 * doc/tree-ssa.texi: ... and the example here.
24498 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24500 * config/aarch64/aarch64-sve.md (sdiv_pow2<mode>3): Extend from
24501 SVE_FULL_I to SVE_I. Generate an UNSPEC_PRED_X.
24502 (*sdiv_pow2<mode>3): New pattern.
24503 (@cond_<sve_int_op><mode>): Extend from SVE_FULL_I to SVE_I.
24504 Wrap the ASRD in an UNSPEC_PRED_X.
24505 (*cond_<sve_int_op><mode>_2): Likewise. Replace the UNSPEC_PRED_X
24506 predicate with a constant PTRUE, if it isn't already.
24507 (*cond_<sve_int_op><mode>_z): Replace with...
24508 (*cond_<sve_int_op><mode>_any): ...this new pattern.
24510 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24512 * config/aarch64/aarch64-sve.md (*cond_bic<mode>_2): Extend from
24513 SVE_FULL_I to SVE_I.
24514 (*cond_bic<mode>_any): Likewise.
24516 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24518 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart)
24519 (@aarch64_pred_<MUL_HIGHPART:optab><mode>): Extend from SVE_FULL_I
24522 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24524 * config/aarch64/aarch64-sve.md (<su>abd<mode>_3): Extend from
24525 SVE_FULL_I to SVE_I.
24526 (*aarch64_cond_<su>abd<mode>_2): Likewise.
24527 (*aarch64_cond_<su>abd<mode>_any): Likewise.
24528 (@aarch64_pred_<su>abd<mode>): Likewise. Use UNSPEC_PRED_X
24529 for the max and min but not for the minus.
24530 (*aarch64_cond_<su>abd<mode>_3): New pattern.
24532 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24534 * config/aarch64/iterators.md (SVE_24I): New iterator.
24535 * config/aarch64/aarch64-sve.md (*aarch64_adr<mode>_shift): Extend from
24536 SVE_FULL_SDI to SVE_24I. Use containers rather than elements.
24538 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24540 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_BINARY:optab><mode>)
24541 (*cond_<SVE_INT_BINARY:optab><mode>_2): Extend from SVE_FULL_I
24543 (*cond_<SVE_INT_BINARY:optab><mode>_3): Likewise.
24544 (*cond_<SVE_INT_BINARY:optab><mode>_any): Likewise.
24545 (*cond_<SVE_INT_BINARY:optab><mode>_2_const): Likewise.
24546 (*cond_<SVE_INT_BINARY:optab><mode>_any_const): Likewise.
24548 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24550 * config/aarch64/aarch64-sve.md (<SVE_INT_BINARY_IMM:optab><mode>3)
24551 (@aarch64_pred_<SVE_INT_BINARY_IMM:optab><mode>)
24552 (*post_ra_<SVE_INT_BINARY_IMM:optab><mode>3): Extend from SVE_FULL_I
24555 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24557 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3)
24558 (v<ASHIFT:optab><mode>3, @aarch64_pred_<optab><mode>)
24559 (*post_ra_v<ASHIFT:optab><mode>3): Extend from SVE_FULL_I to SVE_I.
24561 2021-01-11 Martin Liska <mliska@suse.cz>
24564 * symtab-clones.h (clone_info::release): Release
24565 symtab::m_clones with ggc_delete as it's a GGC memory.
24567 2021-01-11 Matthias Klose <doko@ubuntu.com>
24569 * Makefile.in (LINK_PROGRESS): Show the link target.
24571 2021-01-11 Richard Biener <rguenther@suse.de>
24573 PR tree-optimization/91403
24574 * tree-vect-data-refs.c (vect_analyze_group_access_1): Cap
24575 single-element interleaving group size at 4096 elements.
24577 2021-01-11 Richard Biener <rguenther@suse.de>
24579 PR tree-optimization/98526
24580 * tree-vect-loop.c (vect_model_reduction_cost): Remove costing
24581 of the actual reduction op for the regular case.
24582 (vectorizable_reduction): Cost the stmts
24583 vect_transform_reduction produces here.
24585 2021-01-11 Andreas Krebbel <krebbel@linux.ibm.com>
24587 * tree-ssa-forwprop.c (simplify_vector_constructor): For
24588 big-endian, use UNPACK[_FLOAT]_HI.
24590 2021-01-11 Tamar Christina <tamar.christina@arm.com>
24592 * tree-vect-slp-patterns.c (class complex_pattern,
24593 class complex_add_pattern): Add parameters to matches.
24594 (complex_add_pattern::build): Free memory.
24595 (complex_add_pattern::matches): Move validation end of match.
24596 (complex_add_pattern::recognize): Likewise.
24598 2021-01-11 Tamar Christina <tamar.christina@arm.com>
24600 * tree-vect-slp-patterns.c (linear_loads_p): Fix externals.
24602 2021-01-11 Tamar Christina <tamar.christina@arm.com>
24604 * tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
24606 2021-01-11 Jakub Jelinek <jakub@redhat.com>
24608 PR tree-optimization/95867
24609 * tree-ssa-math-opts.h: New header.
24610 * tree-ssa-math-opts.c: Include tree-ssa-math-opts.h.
24611 (powi_as_mults): No longer static. Use build_one_cst instead of
24612 build_real. Formatting fix.
24613 * tree-ssa-reassoc.c: Include tree-ssa-math-opts.h.
24614 (attempt_builtin_powi): Handle multiplication reassociation without
24615 powi_fndecl using powi_as_mults.
24616 (reassociate_bb): For integral types don't require
24617 -funsafe-math-optimizations to call attempt_builtin_powi.
24619 2021-01-11 Jakub Jelinek <jakub@redhat.com>
24621 PR tree-optimization/95852
24622 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): Change
24623 mul_stmts parameter type to vec<gimple *> &. Before cond_stmt
24624 allow in the bb any of the stmts in that vector, div_stmt and
24625 up to 3 cast stmts.
24626 (arith_cast_equal_p): New function.
24627 (arith_overflow_check_p): Add cast_stmt argument, handle signed
24628 multiply overflow checks.
24629 (match_arith_overflow): Adjust caller. Handle signed multiply
24632 2021-01-11 Jakub Jelinek <jakub@redhat.com>
24634 PR tree-optimization/95852
24635 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): New function.
24636 (uaddsub_overflow_check_p): Renamed to ...
24637 (arith_overflow_check_p): ... this. Handle also multiplication
24638 with overflow check.
24639 (match_uaddsub_overflow): Renamed to ...
24640 (match_arith_overflow): ... this. Add cfg_changed argument. Handle
24641 also multiplication with overflow check. Adjust function comment.
24642 (math_opts_dom_walker::after_dom_children): Adjust callers. Call
24643 match_arith_overflow also for MULT_EXPR.
24645 2021-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24647 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
24648 __builtin_convertvector.
24649 (vmovl_s16): Likewise.
24650 (vmovl_s32): Likewise.
24651 (vmovl_u8): Likewise.
24652 (vmovl_u16): Likewise.
24653 (vmovl_u32): Likewise.
24654 (vmovn_s16): Likewise.
24655 (vmovn_s32): Likewise.
24656 (vmovn_s64): Likewise.
24657 (vmovn_u16): Likewise.
24658 (vmovn_u32): Likewise.
24659 (vmovn_u64): Likewise.
24661 2021-01-11 Martin Liska <mliska@suse.cz>
24663 * gimple-if-to-switch.cc (struct condition_info): Use auto_var.
24664 (if_chain::is_beneficial): Delete clusters
24665 (find_conditions): Make second argument of conditions_in_bbs a
24666 pointer so that we control over it's lifetime.
24667 (pass_if_to_switch::execute): Delete them.
24669 2021-01-11 Kewen Lin <linkw@linux.ibm.com>
24671 * ira.c (move_unallocated_pseudos): Check other_reg and skip if
24674 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24676 * config/vax/vax.md (cc): Remove mode attribute.
24677 (subst_<cc>, subst_f<cc>): Rename to...
24678 (subst_<mode>, subst_f<VAXccnz:mode>): ... these respectively.
24679 (*cbranch<VAXint:mode>4_<VAXcc:mode>): Update for `cc' removal.
24680 (*cbranch<VAXfp:mode>4_<VAXccnz:mode>): Likewise.
24681 (*branch_<mode>, *branch_<mode>_reversed): Likewise.
24683 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24685 * config/vax/vax.md (subst_f<cc>): Add mode to operands and
24686 `const_double_zero'.
24688 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24690 * config/pdp11/pdp11.md (PDPfp): New mode iterator.
24691 (fcc_cc, fcc_ccnz): Use it. Add mode to `const_double_zero' and
24694 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24696 * genemit.c (gen_exp) <CONST_DOUBLE>: Handle `const_double_zero'
24698 * read-rtl.c (rtx_reader::read_rtx_code): Handle machine mode
24699 with `const_double_zero'.
24700 * doc/rtl.texi (Constant Expression Types): Document it.
24702 2021-01-09 Jakub Jelinek <jakub@redhat.com>
24705 * tree-cfg.c (verify_gimple_assign_binary): Allow lhs of
24706 POINTER_DIFF_EXPR to be any integral type.
24708 2021-01-09 Jakub Jelinek <jakub@redhat.com>
24710 PR rtl-optimization/98603
24711 * function.c (instantiate_virtual_regs_in_insn): For asm goto
24712 with impossible constraints, drop all SETs, CLOBBERs, drop PARALLEL
24713 if any, set ASM_OPERANDS mode to VOIDmode and change
24714 ASM_OPERANDS_OUTPUT_CONSTRAINT and ASM_OPERANDS_OUTPUT_IDX.
24716 2021-01-09 Alexandre Oliva <oliva@gnu.org>
24719 * final.c (notice_source_line): Narrow down the condition to
24720 skip a line-0 marker.
24722 2021-01-08 Sergei Trofimovich <siarheit@google.com>
24724 * ipa-modref.c (merge_call_side_effects): Fix
24725 linebreak split by reordering two print calls.
24727 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
24729 * config/s390/vector.md (*tf_to_fprx2_0): Rename from
24730 "*mov_tf_to_fprx2_0" for consistency, fix constraint.
24731 (*tf_to_fprx2_1): Rename from "*mov_tf_to_fprx2_1" for
24732 consistency, fix constraint.
24734 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
24736 * config/s390/s390-c.c (s390_def_or_undef_macro): Accept
24737 callables instead of mask values.
24738 (struct target_flag_set_p): New predicate.
24739 (s390_cpu_cpp_builtins_internal): Define or undefine
24740 __LONG_DOUBLE_VX__ macro.
24742 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
24745 * config/i386/i386.c (x86_function_profiler): Use R10 and R11
24746 to call mcount in large model with PIC for NO_PROFILE_COUNTERS
24749 2021-01-08 Richard Biener <rguenther@suse.de>
24751 * tree-ssa-sccvn.c (pass_fre::execute): Reset the SCEV hash table.
24753 2021-01-08 Richard Biener <rguenther@suse.de>
24755 * tree-vect-slp.c (scalar_stmts_to_slp_tree_map_t): Fix.
24756 (vect_build_slp_tree): On cache hit release the matched
24757 scalar stmts vector.
24758 * tree-vect-stmts.c (vectorizable_store): Properly free
24759 vec_oprnds before possibly gathering them again.
24761 2021-01-08 Richard Biener <rguenther@suse.de>
24763 PR tree-optimization/98544
24764 * tree-vect-slp.c (vect_optimize_slp): Always materialize
24765 permutes at a permute node.
24767 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
24770 * config/i386/i386.c (x86_function_profiler): Use R10 to call
24771 mcount in large model. Sorry for large model with PIC.
24773 2021-01-08 Jakub Jelinek <jakub@redhat.com>
24776 * config/i386/i386.opt (ix86_cmodel, ix86_incoming_stack_boundary_arg,
24777 ix86_pmode, ix86_preferred_stack_boundary_arg, ix86_regparm,
24778 ix86_veclibabi_type): Remove x_ prefix, use TargetVariable instead of
24779 TargetSave and initialize for variables with enum types.
24780 (mfentry, mstack-protector-guard-reg=, mstack-protector-guard-offset=,
24781 mstack-protector-guard-symbol=): Add Save.
24782 * config/i386/i386-options.c (ix86_function_specific_save,
24783 ix86_function_specific_restore): Don't save or restore x_ix86_cmodel,
24784 x_ix86_incoming_stack_boundary_arg, x_ix86_pmode,
24785 x_ix86_preferred_stack_boundary_arg, x_ix86_regparm,
24786 x_ix86_veclibabi_type.
24788 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
24790 * config/aarch64/aarch64-sve.md (*cnot<mode>): Extend from
24791 SVE_FULL_I to SVE_I.
24792 (*cond_cnot<mode>_2, *cond_cnot<mode>_any): Likewise.
24794 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
24796 * config/aarch64/aarch64-sve.md (*cond_uxt<mode>_2): Extend from
24797 SVE_FULL_I to SVE_I.
24798 (*cond_uxt<mode>_any): Likewise.
24800 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24802 * config/aarch64/iterators.md (Vwhalf): New iterator.
24803 * config/aarch64/aarch64-simd.md (aarch64_<sur>adalp<mode>_3):
24805 (aarch64_<sur>adalp<mode>): ... This. Make more
24807 (<sur>sadv16qi): Adjust callsite of the above.
24808 * config/aarch64/aarch64-simd-builtins.def (sadalp, uadalp): New
24810 * config/aarch64/arm_neon.h (vpadal_s8): Reimplement using
24812 (vpadal_s16): Likewise.
24813 (vpadal_u8): Likewise.
24814 (vpadal_u16): Likewise.
24815 (vpadalq_s8): Likewise.
24816 (vpadalq_s16): Likewise.
24817 (vpadalq_s32): Likewise.
24818 (vpadalq_u8): Likewise.
24819 (vpadalq_u16): Likewise.
24820 (vpadalq_u32): Likewise.
24822 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24824 * config/aarch64/aarch64-simd.md (aarch64_<su>abd<mode>_3):
24826 (aarch64_<su>abd<mode>): ... This.
24827 (<sur>sadv16qi): Adjust callsite of the above.
24828 * config/aarch64/aarch64-simd-builtins.def (sabd, uabd): Define
24830 * config/aarch64/arm_neon.h (vabd_s8): Reimplement using
24832 (vabd_s16): Likewise.
24833 (vabd_s32): Likewise.
24834 (vabd_u8): Likewise.
24835 (vabd_u16): Likewise.
24836 (vabd_u32): Likewise.
24837 (vabdq_s8): Likewise.
24838 (vabdq_s16): Likewise.
24839 (vabdq_s32): Likewise.
24840 (vabdq_u8): Likewise.
24841 (vabdq_u16): Likewise.
24842 (vabdq_u32): Likewise.
24844 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24846 * config/aarch64/aarch64-simd-builtins.def (saba, uaba): Define
24848 * config/aarch64/arm_neon.h (vaba_s8): Implement using builtin.
24849 (vaba_s16): Likewise.
24850 (vaba_s32): Likewise.
24851 (vaba_u8): Likewise.
24852 (vaba_u16): Likewise.
24853 (vaba_u32): Likewise.
24854 (vabaq_s8): Likewise.
24855 (vabaq_s16): Likewise.
24856 (vabaq_s32): Likewise.
24857 (vabaq_u8): Likewise.
24858 (vabaq_u16): Likewise.
24859 (vabaq_u32): Likewise.
24861 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24863 * config/aarch64/aarch64-simd.md (aba<mode>_3): Rename to...
24864 (aarch64_<su>aba<mode>): ... This. Handle uaba as well.
24865 Change RTL pattern to match.
24867 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
24869 * common/config/riscv/riscv-common.c (riscv_current_subset_list): New.
24870 * config/riscv/riscv-c.c (riscv-subset.h): New.
24871 (INCLUDE_STRING): Define.
24872 (riscv_cpu_cpp_builtins): Add new style architecture extension
24874 * config/riscv/riscv-subset.h (riscv_subset_list::begin): New.
24875 (riscv_subset_list::end): New.
24876 (riscv_current_subset_list): New.
24878 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
24880 * common/config/riscv/riscv-common.c (RISCV_DONT_CARE_VERSION):
24881 Move to riscv-subset.h.
24882 (struct riscv_subset_t): Ditto.
24883 (class riscv_subset_list): Ditto.
24884 * config/riscv/riscv-subset.h (RISCV_DONT_CARE_VERSION): Move
24885 from riscv-common.c.
24886 (struct riscv_subset_t): Ditto.
24887 (class riscv_subset_list): Ditto.
24888 * config/riscv/t-riscv ($(common_out_file)): Add file
24891 2021-01-07 Jakub Jelinek <jakub@redhat.com>
24894 * config/i386/i386.md (*bmi_blsi_<mode>_cmp, *bmi_blsi_<mode>_ccno):
24895 New define_insn patterns.
24897 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
24899 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_UNARY:optab><mode>)
24900 (*cond_<SVE_INT_UNARY:optab><mode>_2): Extend from SVE_FULL_I to SVE_I.
24901 (*cond_<SVE_INT_UNARY:optab><mode>_any): Likewise.
24903 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
24905 PR tree-optimization/98560
24906 * internal-fn.def (IFN_VCONDU, IFN_VCONDEQ): Use type vec_cond.
24907 * internal-fn.c (vec_cond_mask_direct): Get the data mode from
24909 (vec_cond_direct): Likewise argument 2.
24910 (vec_condu_direct, vec_condeq_direct): Delete.
24911 (expand_vect_cond_optab_fn): Rename to...
24912 (expand_vec_cond_optab_fn): ...this, replacing old macro.
24913 (expand_vec_condu_optab_fn, expand_vec_condeq_optab_fn): Delete.
24914 (expand_vect_cond_mask_optab_fn): Rename to...
24915 (expand_vec_cond_mask_optab_fn): ...this, replacing old macro.
24916 (direct_vec_cond_mask_optab_supported_p): Treat the optab as a
24918 (direct_vec_cond_optab_supported_p): Likewise.
24919 (direct_vec_condu_optab_supported_p): Delete.
24920 (direct_vec_condeq_optab_supported_p): Delete.
24921 * gimple-isel.cc: Include internal-fn.h.
24922 (gimple_expand_vec_cond_expr): Check that IFN_VCONDEQ is supported
24925 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
24927 PR tree-optimization/98560
24928 * gimple-isel.cc (gimple_expand_vec_cond_expr): If we fail to use
24929 IFN_VCOND{,U,EQ}, fall back on IFN_VCOND_MASK.
24931 2021-01-07 Uroš Bizjak <ubizjak@gmail.com>
24933 * config/i386/i386.md (insn): Merge from plusminus_insn, shift_insn,
24934 rotate_insn and optab code attributes.
24935 Update all uses to merged code attribute.
24936 * config/i386/sse.md: Update all uses to merged code attribute.
24937 * config/i386/mmx.md: Update all uses to merged code attribute.
24939 2021-01-07 Jakub Jelinek <jakub@redhat.com>
24941 PR tree-optimization/98568
24942 * gimple-ssa-store-merging.c (bswap_view_convert): New function.
24943 (bswap_replace): Use it.
24945 2021-01-06 Vladimir N. Makarov <vmakarov@redhat.com>
24947 PR rtl-optimization/97978
24948 * lra-int.h (lra_hard_reg_split_p): New external.
24949 * lra.c (lra_hard_reg_split_p): New global.
24950 (lra): Set up lra_hard_reg_split_p after splitting a hard reg.
24951 * lra-assigns.c (lra_assign): Don't check allocation correctness
24952 after hard reg splitting.
24954 2021-01-06 Martin Sebor <msebor@redhat.com>
24957 * builtins.c (new_delete_mismatch_p): New overload.
24958 (new_delete_mismatch_p (tree, tree)): Call it.
24960 2021-01-06 Alexandre Oliva <oliva@adacore.com>
24962 * Makefile.in (T_GLIMITS_H): New.
24963 (stmp-int-hdrs): Depend on it, use it.
24964 * config/t-vxworks (T_GLIMITS_H): Override it.
24965 (vxw-glimits.h): New.
24967 2021-01-06 Richard Biener <rguenther@suse.de>
24969 PR tree-optimization/98513
24970 * value-range.cc (intersect_ranges): Compare the upper bounds
24971 for the expected relation.
24973 2021-01-06 Gerald Pfeifer <gerald@pfeifer.com>
24976 2020-12-28 Gerald Pfeifer <gerald@pfeifer.com>
24978 * doc/standards.texi (HSAIL): Remove section.
24980 2021-01-05 Samuel Thibault <samuel.thibault@ens-lyon.org>
24982 * configure: Re-generate.
24984 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24986 * doc/invoke.texi (-std=c++20): Adjust for the publication of
24987 ISO 14882:2020 standard.
24988 * doc/standards.texi: Likewise.
24990 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24992 PR tree-optimization/94802
24993 * expr.h (maybe_optimize_sub_cmp_0): Declare.
24994 * expr.c: Include tree-pretty-print.h and flags.h.
24995 (maybe_optimize_sub_cmp_0): New function.
24996 (do_store_flag): Use it.
24997 * cfgexpand.c (expand_gimple_cond): Likewise.
24999 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
25001 * mux-utils.h (pointer_mux::m_ptr): Tweak description of contents.
25002 * rtlanal.c (simple_regno_set): Tweak description to clarify the
25005 2021-01-05 Richard Biener <rguenther@suse.de>
25007 PR tree-optimization/98516
25008 * tree-vect-slp.c (vect_optimize_slp): Permute the incoming
25009 lanes when materializing on a VEC_PERM node.
25010 (vectorizable_slp_permutation): Dump the permute properly.
25012 2021-01-05 Richard Biener <rguenther@suse.de>
25014 * tree-vect-slp.c (vect_slp_region): Move debug counter
25015 to cover individual subgraphs.
25017 2021-01-05 Richard Biener <rguenther@suse.de>
25019 PR tree-optimization/98428
25020 * tree-vect-slp.c (vect_build_slp_tree_1): Properly reject
25021 vector lane extracts for loop vectorization.
25023 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25025 PR tree-optimization/98514
25026 * tree-ssa-reassoc.c (bb_rank): Change type from long * to
25028 (operand_rank): Change type from hash_map<tree, long> to
25029 hash_map<tree, int64_t>.
25030 (phi_rank): Change return type from long to int64_t.
25031 (loop_carried_phi): Change block_rank variable type from long to
25033 (propagate_rank): Change return type, rank parameter type and
25034 op_rank variable type from long to int64_t.
25035 (find_operand_rank): Change return type from long to int64_t
25036 and change slot variable type from long * to int64_t *.
25037 (insert_operand_rank): Change rank parameter type from long to
25039 (get_rank): Change return type and rank variable type from long to
25040 int64_t. Use PRId64 instead of ld to print the rank.
25041 (init_reassoc): Change rank variable type from long to int64_t
25042 and adjust correspondingly bb_rank and operand_rank initialization.
25044 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25046 PR tree-optimization/96928
25047 * tree-ssa-phiopt.c (xor_replacement): New function.
25048 (tree_ssa_phiopt_worker): Call it.
25050 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25052 PR tree-optimization/96930
25053 * match.pd ((A / (1 << B)) -> (A >> B)): If A is extended
25054 from narrower value which has the same type as 1 << B, perform
25055 the right shift on the narrower value followed by extension.
25057 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25059 PR tree-optimization/96239
25060 * gimple-ssa-store-merging.c (maybe_optimize_vector_constructor): New
25062 (get_status_for_store_merging): Don't return BB_INVALID for blocks
25063 with potential bswap optimizable CONSTRUCTORs.
25064 (pass_store_merging::execute): Optimize vector CONSTRUCTORs with bswap
25067 2021-01-05 Richard Biener <rguenther@suse.de>
25069 PR tree-optimization/98381
25070 * tree.c (vector_element_bits): Properly compute bool vector
25072 * tree-vect-loop.c (vectorizable_live_operation): Properly
25073 compute the last lane bit offset.
25075 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
25078 * config/i386/sse.md (sse_cvtps2pi): Redefine as define_insn_and_split.
25079 Clear the top 64 bytes of the input XMM register.
25080 (sse_cvttps2pi): Ditto.
25082 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
25085 * config/i386/xopintrin.h (_mm256_cmov_si256): New.
25087 2021-01-05 H.J. Lu <hjl.tools@gmail.com>
25090 * config/i386/xmmintrin.h (_mm_extract_pi16): Cast to unsigned
25093 2021-01-05 Claudiu Zissulescu <claziss@synopsys.com>
25095 * config/arc/arc.md (maddsidi4_split): Use ACC_REG_FIRST.
25096 (umaddsidi4_split): Likewise.
25098 2021-01-05 liuhongt <hongtao.liu@intel.com>
25101 * config/i386/sse.md (*sse2_pmovskb_zexthisi): New
25102 define_insn_and_split for zero_extend of subreg HI of pmovskb
25104 (*sse2_pmovskb_zexthisi): Add new combine splitters for
25105 zero_extend of not of subreg HI of pmovskb result.
25107 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
25110 * explow.c (convert_memory_address_addr_space_1): Handle UNSPECs
25112 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate): Use
25113 convert_memory_address to convert symbolic immediates to ptr_mode
25114 before forcing them to memory.
25116 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
25118 PR rtl-optimization/97144
25119 * recog.c (constrain_operands): Initialize matching_operand
25120 for each alternative, rather than only doing it once.
25122 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
25124 PR rtl-optimization/98403
25125 * rtl-ssa/changes.cc (function_info::finalize_new_accesses): Explain
25126 why we don't remove call clobbers.
25127 (function_info::apply_changes_to_insn): Don't attempt to add
25128 call clobbers here.
25130 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
25132 PR tree-optimization/98371
25133 * tree-vect-loop.c (vect_reanalyze_as_main_loop): New function.
25134 (vect_analyze_loop): If an epilogue loop appears to be cheaper
25135 than the main loop, re-analyze it as a main loop before adopting
25138 2021-01-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
25141 * configure.ac (NETLIBS): Determine using AX_LIB_SOCKET_NSL.
25142 * aclocal.m4, configure: Regenerate.
25143 * Makefile.in (NETLIBS): Define.
25144 (BACKEND): Remove $(CODYLIB).
25146 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25148 PR rtl-optimization/98334
25149 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
25150 Optimize (X - 1) * Y + Y to X * Y or (X + 1) * Y - Y to X * Y.
25152 2021-01-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
25154 * tree-inline.c (expand_call_inline): Restore input_location.
25155 Return result from recursive call.
25157 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
25159 PR tree-optimization/95401
25160 * config/aarch64/aarch64-sve-builtins.cc
25161 (gimple_folder::load_store_cookie): Use bits rather than bytes
25162 for the alignment argument to IFN_MASK_LOAD and IFN_MASK_STORE.
25163 * gimple-fold.c (gimple_fold_mask_load_store_mem_ref): Likewise.
25164 * tree-vect-stmts.c (vectorizable_store): Likewise.
25165 (vectorizable_load): Likewise.
25167 2021-01-04 Richard Biener <rguenther@suse.de>
25169 PR tree-optimization/98308
25170 * tree-vect-stmts.c (vectorizable_load): Set invariant mask
25173 2021-01-04 Jakub Jelinek <jakub@redhat.com>
25175 PR tree-optimization/95771
25176 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Handle types
25177 with precision smaller than int's precision and types with precision
25178 twice as large as long long. Formatting fixes.
25180 2021-01-04 Richard Biener <rguenther@suse.de>
25182 PR tree-optimization/98464
25183 * tree-ssa-sccvn.c (vn_valueize_for_srt): Rename from ...
25184 (vn_valueize_wrapper): ... this. Temporarily adjust vn_context_bb.
25185 (process_bb): Adjust.
25187 2021-01-04 Matthew Malcomson <matthew.malcomson@arm.com>
25190 * doc/invoke.texi (-fsanitize=address): Fix wording describing
25191 clash with -fsanitize=hwaddress.
25193 2021-01-04 Richard Biener <rguenther@suse.de>
25195 PR tree-optimization/98282
25196 * tree-ssa-sccvn.c (vn_get_stmt_kind): Classify tcc_reference on
25197 invariants as VN_NARY.
25199 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
25202 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Accept
25203 aarch64_simd_reg_or_zero for operand 2. Use the combinez patterns
25204 to handle zero operands.
25206 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
25208 * config/aarch64/aarch64.c (offset_6bit_signed_scaled_p): New function.
25209 (offset_6bit_unsigned_scaled_p): Fix typo in comment.
25210 (aarch64_sve_prefetch_operand_p): Accept MUL VLs in the range
25213 2021-01-04 Richard Biener <rguenther@suse.de>
25215 PR tree-optimization/98393
25216 * tree-vect-slp.c (vect_build_slp_tree): Properly zero matches
25217 when hitting the limit.
25219 2021-01-04 Richard Biener <rguenther@suse.de>
25221 PR tree-optimization/98291
25222 * tree-vect-loop.c (vectorizable_reduction): Bypass
25223 associativity check for SLP reductions with VF 1.
25225 2021-01-04 Jakub Jelinek <jakub@redhat.com>
25227 PR tree-optimization/96782
25228 * match.pd (x == ~x -> false, x != ~x -> true): New simplifications.
25230 2021-01-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
25232 * collect-utils.c (collect_execute): Check dumppfx.
25233 * collect2.c (maybe_run_lto_and_relink, do_link): Pass atsuffix
25234 to collect_execute.
25235 (do_link): Add new parameter atsuffix.
25236 (main): Handle -dumpdir option. Skip one argument for
25237 -o, -isystem and -B options.
25238 * gcc.c (make_at_file): New helper function.
25239 (close_at_file): Use it.
25241 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25243 * config/darwin.h (MIN_LD64_NO_COAL_SECTS): Adjust.
25244 Amend handling for LD64_VERSION fallback defaults.
25246 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25248 * config.gcc: Compute default version information
25249 from the configured target. Likewise defaults for
25251 * config/darwin10.h: Removed.
25252 * config/darwin12.h: Removed.
25253 * config/darwin9.h: Removed.
25254 * config/rs6000/darwin8.h: Removed.
25256 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25258 * config/darwin9.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
25260 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25262 * config/darwin9.h (STACK_CHECK_STATIC_BUILTIN): Move from here..
25263 * config/darwin.h (STACK_CHECK_STATIC_BUILTIN): .. to here.
25265 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25267 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move from
25269 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): ... to here.
25271 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25273 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move the spec
25274 for the Darwin10 unwinder stub from here ...
25275 * config/darwin.h (LINK_COMMAND_SPEC_A): ... to here.
25277 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25279 * config/darwin.h (DSYMUTIL_SPEC): Default to DWARF
25280 (ASM_DEBUG_SPEC):Only define if the assembler supports
25282 (PREFERRED_DEBUGGING_TYPE): Default to DWARF.
25283 (DARWIN_PREFER_DWARF): Define.
25284 * config/darwin9.h (PREFERRED_DEBUGGING_TYPE): Remove.
25285 (DARWIN_PREFER_DWARF): Likewise
25286 (DSYMUTIL_SPEC): Likewise.
25287 (COLLECT_RUN_DSYMUTIL): Likewise.
25288 (ASM_DEBUG_SPEC): Likewise.
25289 (ASM_DEBUG_OPTION_SPEC): Likewise.
25291 2021-01-02 Jan Hubicka <jh@suse.cz>
25293 * cfg.c (free_block): ggc_free bb.
25295 2021-01-01 Jakub Jelinek <jakub@redhat.com>
25297 * gcc.c (process_command): Update copyright notice dates.
25298 * gcov-dump.c (print_version): Ditto.
25299 * gcov.c (print_version): Ditto.
25300 * gcov-tool.c (print_version): Ditto.
25301 * gengtype.c (create_file): Ditto.
25302 * doc/cpp.texi: Bump @copying's copyright year.
25303 * doc/cppinternals.texi: Ditto.
25304 * doc/gcc.texi: Ditto.
25305 * doc/gccint.texi: Ditto.
25306 * doc/gcov.texi: Ditto.
25307 * doc/install.texi: Ditto.
25308 * doc/invoke.texi: Ditto.
25310 2021-01-01 Jakub Jelinek <jakub@redhat.com>
25312 * ChangeLog-2020: Rotate ChangeLog. New file.
25315 Copyright (C) 2021 Free Software Foundation, Inc.
25317 Copying and distribution of this file, with or without modification,
25318 are permitted in any medium without royalty provided the copyright
25319 notice and this notice are preserved.