1 2021-10-12 Bill Schmidt <wschmidt@linux.ibm.com>
4 * config/rs6000/altivec.h (vec_cpsgn): Swap operand order.
5 * config/rs6000/rs6000-overload.def (VEC_COPYSIGN): Use SKIP to
6 avoid generating an automatic #define of vec_cpsgn. Use the
7 correct built-in for V4SFmode that doesn't depend on VSX.
9 2021-10-12 Uroš Bizjak <ubizjak@gmail.com>
13 * config/i386/i386.md (*add<mode>_1_slp): Rewrite as
14 define_insn_and_split pattern. Add alternative 1 and split it
15 post reload to insert operand 1 into the low part of operand 0.
16 (*sub<mode>_1_slp): Ditto.
17 (*and<mode>_1_slp): Ditto.
18 (*<any_or:code><mode>_1_slp): Ditto.
19 (*ashl<mode>3_1_slp): Ditto.
20 (*<any_shiftrt:insn><mode>3_1_slp): Ditto.
21 (*<any_rotate:insn><mode>3_1_slp): Ditto.
22 (*neg<mode>_1_slp): New insn_and_split pattern.
23 (*one_cmpl<mode>_1_slp): Ditto.
25 2021-10-12 David Edelsohn <dje.gcc@gmail.com>
27 * doc/install.texi: Update MinGW and mingw-64 Binaries
30 2021-10-12 Daniel Le Duc Khoi Nguyen <greenrecyclebin@gmail.com>
32 * doc/extend.texi (Common Variable Attributes): Fix typos in
33 alloc_size documentation.
35 2021-10-12 Richard Biener <rguenther@suse.de>
37 PR tree-optimization/102696
38 * tree-vect-slp.c (vect_build_slp_tree_2): Properly mark
39 the tree fatally failed when we reject a BIT_FIELD_REF.
41 2021-10-12 Richard Biener <rguenther@suse.de>
43 PR tree-optimization/102572
44 * tree-vect-stmts.c (vect_build_gather_load_calls): When
45 gathering the vectorized defs for the mask pass in the
46 desired mask vector type so invariants will be handled
49 2021-10-12 Tamar Christina <tamar.christina@arm.com>
51 * config/aarch64/aarch64-sve.md (*fcm<cmp_op><mode>_bic_combine,
52 *fcm<cmp_op><mode>_nor_combine, *fcmuo<mode>_bic_combine,
53 *fcmuo<mode>_nor_combine): New.
55 2021-10-12 Eric Botcazou <ebotcazou@adacore.com>
58 * config/sparc/sparc-modes.def (OI): New integer mode.
60 2021-10-12 Jakub Jelinek <jakub@redhat.com>
62 * gimple-fold.h (clear_padding_type_may_have_padding_p): Declare.
63 * gimple-fold.c (clear_padding_type_may_have_padding_p): No longer
66 2021-10-12 Jakub Jelinek <jakub@redhat.com>
68 * tree-vectorizer.h (loop_cost_model): New function.
69 (unlimited_cost_model): Use it.
70 * tree-vect-loop.c (vect_analyze_loop_costing): Use loop_cost_model
71 call instead of flag_vect_cost_model.
72 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
73 (vect_prune_runtime_alias_test_list): Likewise. Also use it instead
74 of flag_simd_cost_model.
76 2021-10-12 liuhongt <hongtao.liu@intel.com>
79 * config/i386/i386-expand.c (emit_reduc_half): Handle
81 * config/i386/mmx.md (reduc_<code>_scal_v4qi): New expander.
82 (reduc_plus_scal_v4qi): Ditto.
84 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
86 * config/rs6000/smmintrin.h (_mm_cmpeq_epi64, _mm_cmpgt_epi64,
87 _mm_mullo_epi32, _mm_mul_epi32, _mm_packus_epi32): New.
88 * config/rs6000/nmmintrin.h: Copy from i386, tweak to suit.
90 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
92 * config/rs6000/smmintrin.h (_mm_cvtepi8_epi16, _mm_cvtepi8_epi32,
93 _mm_cvtepi8_epi64, _mm_cvtepi16_epi32, _mm_cvtepi16_epi64,
94 _mm_cvtepi32_epi64, _mm_cvtepu8_epi16, _mm_cvtepu8_epi32,
95 _mm_cvtepu8_epi64, _mm_cvtepu16_epi32, _mm_cvtepu16_epi64,
96 _mm_cvtepu32_epi64): New.
98 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
100 * config/rs6000/smmintrin.h (_mm_test_all_zeros,
101 _mm_test_all_ones, _mm_test_mix_ones_zeros): Rewrite as macro.
103 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
105 * config/rs6000/smmintrin.h (_mm_min_epi8, _mm_min_epu16,
106 _mm_min_epi32, _mm_min_epu32, _mm_max_epi8, _mm_max_epu16,
107 _mm_max_epi32, _mm_max_epu32): New.
109 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
111 * ipa-modref-tree.h (struct modref_access_node): Revert
113 (struct modref_ref_node): Likewise.
115 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
117 * ipa-modref-tree.h (modref_tree::global_access_p): New member
120 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
121 ignore_stores_eaf_flags): Move to ipa-modref.h
122 (remove_useless_eaf_flags): Remove early exit on NOCLOBBER.
123 (modref_summary::global_memory_read_p): New member function.
124 (modref_summary::global_memory_written_p): New member function.
125 * ipa-modref.h (modref_summary::global_memory_read_p,
126 modref_summary::global_memory_written_p): Declare.
127 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
128 ignore_stores_eaf_flags): move here.
129 * tree-ssa-structalias.c: Include ipa-modref-tree.h, ipa-modref.h
131 (handle_rhs_call): Rewrite.
132 (handle_call_arg): New function.
133 (determine_global_memory_access): New function.
134 (handle_const_call): Remove
135 (handle_pure_call): Remove
136 (find_func_aliases_for_call): Update use of handle_rhs_call.
137 (compute_points_to_sets): Handle global memory acccesses
140 2021-10-11 Diane Meirowitz <diane.meirowitz@oracle.com>
142 * doc/invoke.texi: Add link to UndefinedBehaviorSanitizer
143 documentation, mention UBSAN_OPTIONS, similar to what is done
144 for AddressSanitizer.
146 2021-10-11 Richard Biener <rguenther@suse.de>
149 * internal-fn.c (expand_DEFERRED_INIT): Check for mode
150 availability before building an integer type for storage
153 2021-10-11 Richard Biener <rguenther@suse.de>
156 * gimple.c (gimple_call_fnspec): Do not mark operator new/delete
159 2021-10-11 Martin Liska <mliska@suse.cz>
161 * common.opt: Remove Init(2) for some options.
162 * toplev.c (process_options): Do not use AUTODETECT_VALUE, but
163 use rather OPTION_SET_P.
165 2021-10-11 Martin Liska <mliska@suse.cz>
167 * common.opt: Remove usage of IRA_REGION_AUTODETECT.
168 * flag-types.h (enum ira_region): Likewise.
169 * toplev.c (process_options): Use OPTION_SET_P instead of
170 IRA_REGION_AUTODETECT.
172 2021-10-11 Jakub Jelinek <jakub@redhat.com>
174 * omp-low.c (omp_runtime_api_call): Handle omp_get_max_teams,
175 omp_[sg]et_teams_thread_limit and omp_set_num_teams.
177 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
179 * config/s390/s390-protos.h (s390_rawmemchr): Add prototype.
180 * config/s390/s390.c (s390_rawmemchr): New function.
181 * config/s390/s390.md (rawmemchr<SINT:mode>): New expander.
182 * config/s390/vector.md (@vec_vfees<mode>): Basically a copy of
183 the pattern vfees<mode> from vx-builtins.md.
184 * config/s390/vx-builtins.md (*vfees<mode>): Remove.
186 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
188 * builtins.c (get_memory_rtx): Change to external linkage.
189 * builtins.h (get_memory_rtx): Add function prototype.
190 * doc/md.texi (rawmemchr<mode>): Document.
191 * internal-fn.c (expand_RAWMEMCHR): Define.
192 * internal-fn.def (RAWMEMCHR): Add.
193 * optabs.def (rawmemchr_optab): Add.
194 * tree-loop-distribution.c (find_single_drs): Change return code
195 behaviour by also returning true if no single store was found
197 (loop_distribution::classify_partition): Respect the new return
198 code behaviour of function find_single_drs.
199 (loop_distribution::execute): Call new function
200 transform_reduction_loop in order to replace rawmemchr or strlen
201 like loops by calls into builtins.
202 (generate_reduction_builtin_1): New function.
203 (generate_rawmemchr_builtin): New function.
204 (generate_strlen_builtin_1): New function.
205 (generate_strlen_builtin): New function.
206 (generate_strlen_builtin_using_rawmemchr): New function.
207 (reduction_var_overflows_first): New function.
208 (determine_reduction_stmt_1): New function.
209 (determine_reduction_stmt): New function.
210 (loop_distribution::transform_reduction_loop): New function.
212 2021-10-11 Martin Liska <mliska@suse.cz>
214 * tree.c (cl_option_hasher::hash): Use cl_optimization_hash
215 and remove legacy hashing code.
217 2021-10-11 Kito Cheng <kito.cheng@sifive.com>
220 * builtins.c (maybe_emit_call_builtin___clear_cache): Allow
221 CONST_INT for BEGIN and END, and use gcc_assert rather than
224 2021-10-10 Jakub Jelinek <jakub@redhat.com>
227 * var-tracking.c (add_stores): For cselib_sp_derived_value_p values
228 use MO_VAL_SET if loc is not sp.
230 2021-10-10 Andrew Pinski <apinski@marvell.com>
232 PR tree-optimization/102622
233 * match.pd: Swap the order of a?pow2cst:0 and a?-1:0 transformations.
234 Swap the order of a?0:pow2cst and a?0:-1 transformations.
236 2021-10-09 liuhongt <hongtao.liu@intel.com>
239 * config/i386/i386-expand.c (ix86_valid_mask_cmp_mode): Handle
241 (ix86_use_mask_cmp_p): Ditto.
242 (ix86_expand_sse_movcc): Ditto.
243 * config/i386/i386.md (setcc_hf_mask): New define_insn.
245 (UNSPEC_MOVCC_MASK): New unspec.
246 * config/i386/sse.md (UNSPEC_PCMP): Move to i386.md.
248 2021-10-08 Vladimir N. Makarov <vmakarov@redhat.com>
250 PR rtl-optimization/102627
251 * lra-constraints.c (split_reg): Use at least natural mode of hard reg.
253 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
255 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Grow
258 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
260 * value-range.cc (irange::debug): New.
261 * value-range.h (irange::debug): New.
263 2021-10-08 Richard Sandiford <richard.sandiford@arm.com>
265 PR tree-optimization/102385
266 * predict.h (change_edge_frequency): Declare.
267 * predict.c (change_edge_frequency): New function.
268 * tree-ssa-loop-manip.h (tree_transform_and_unroll_loop): Remove
270 (tree_unroll_loop): Likewise.
271 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Update accordingly.
272 * tree-predcom.c (pcom_worker::tree_predictive_commoning_loop):
274 * tree-ssa-loop-prefetch.c (loop_prefetch_arrays): Likewise.
275 * tree-ssa-loop-manip.c (tree_unroll_loop): Likewise.
276 (tree_transform_and_unroll_loop): Likewise. Use single_dom_exit
277 to retrieve the exit edges. Make all the old profile update code
278 conditional on !single_loop_p -- the case it was written for --
279 and use a different approach for the single-loop case.
281 2021-10-08 Martin Liska <mliska@suse.cz>
283 * config/alpha/alpha.c (alpha_option_override): Use new macro
285 * config/arc/arc.c (arc_override_options): Likewise.
286 * config/arm/arm.c (arm_option_override): Likewise.
287 * config/bfin/bfin.c (bfin_load_pic_reg): Likewise.
288 * config/c6x/c6x.c (c6x_option_override): Likewise.
289 * config/csky/csky.c: Likewise.
290 * config/darwin.c (darwin_override_options): Likewise.
291 * config/frv/frv.c (frv_option_override): Likewise.
292 * config/i386/djgpp.h: Likewise.
293 * config/i386/i386.c (ix86_stack_protect_guard): Likewise.
294 (ix86_max_noce_ifcvt_seq_cost): Likewise.
295 * config/ia64/ia64.c (ia64_option_override): Likewise.
296 (ia64_override_options_after_change): Likewise.
297 * config/m32c/m32c.c (m32c_option_override): Likewise.
298 * config/m32r/m32r.c (m32r_init): Likewise.
299 * config/m68k/m68k.c (m68k_option_override): Likewise.
300 * config/microblaze/microblaze.c (microblaze_option_override): Likewise.
301 * config/mips/mips.c (mips_option_override): Likewise.
302 * config/nios2/nios2.c (nios2_option_override): Likewise.
303 * config/nvptx/nvptx.c (nvptx_option_override): Likewise.
304 * config/pa/pa.c (pa_option_override): Likewise.
305 * config/riscv/riscv.c (riscv_option_override): Likewise.
306 * config/rs6000/aix71.h: Likewise.
307 * config/rs6000/aix72.h: Likewise.
308 * config/rs6000/aix73.h: Likewise.
309 * config/rs6000/rs6000.c (darwin_rs6000_override_options): Likewise.
310 (rs6000_override_options_after_change): Likewise.
311 (rs6000_linux64_override_options): Likewise.
312 (glibc_supports_ieee_128bit): Likewise.
313 (rs6000_option_override_internal): Likewise.
314 (rs6000_file_start): Likewise.
315 (rs6000_darwin_file_start): Likewise.
316 * config/rs6000/rtems.h: Likewise.
317 * config/rs6000/sysv4.h: Likewise.
318 * config/rs6000/vxworks.h (SUB3TARGET_OVERRIDE_OPTIONS): Likewise.
319 * config/s390/s390.c (s390_option_override): Likewise.
320 * config/sh/linux.h: Likewise.
321 * config/sh/netbsd-elf.h (while): Likewise.
322 * config/sh/sh.c (sh_option_override): Likewise.
323 * config/sol2.c (solaris_override_options): Likewise.
324 * config/sparc/sparc.c (sparc_option_override): Likewise.
325 * config/tilegx/tilegx.c (tilegx_option_override): Likewise.
326 * config/visium/visium.c (visium_option_override): Likewise.
327 * config/vxworks.c (vxworks_override_options): Likewise.
328 * lto-opts.c (lto_write_options): Likewise.
329 * omp-expand.c (expand_omp_simd): Likewise.
330 * omp-general.c (omp_max_vf): Likewise.
331 * omp-offload.c (oacc_xform_loop): Likewise.
332 * opts.h (OPTION_SET_P): Likewise.
333 * targhooks.c (default_max_noce_ifcvt_seq_cost): Likewise.
334 * toplev.c (process_options): Likewise.
335 * tree-predcom.c: Likewise.
336 * tree-sra.c (analyze_all_variable_accesses): Likewise.
338 2021-10-08 liuhongt <hongtao.liu@intel.com>
341 * config/i386/i386.c (ix86_optab_supported_p):
342 Return true for HFmode.
343 * match.pd: Simplify (_Float16) ceil ((double) x) to
344 __builtin_ceilf16 (a) when a is _Float16 type and
345 direct_internal_fn_supported_p.
347 2021-10-08 liuhongt <hongtao.liu@intel.com>
350 * config/i386/i386-expand.c (emit_reduc_half): Hanlde V4HImode.
351 * config/i386/mmx.md (reduc_plus_scal_v4hi): New.
352 (reduc_<code>_scal_v4hi): New.
354 2021-10-08 liuhongt <hongtao.liu@intel.com>
356 * common.opt (ftree-vectorize): Add Var(flag_tree_vectorize).
357 * doc/invoke.texi (Options That Control Optimization): Update
359 * opts.c (default_options_table): Enable auto-vectorization at
360 O2 with very-cheap cost model.
361 (finish_options): Use cheap cost model for
362 explicit -ftree{,-loop}-vectorize.
364 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
366 * ctfc.c (ctfc_delete_container): Free hash table contents.
368 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
370 * toplev.c (process_options): Do not warn for GNU GIMPLE.
372 2021-10-07 Siddhesh Poyarekar <siddhesh@gotplt.org>
374 * tree-object-size.c (addr_object_size,
375 compute_builtin_object_size): Drop PDECL and POFF arguments.
376 (addr_object_size): Adjust calls.
377 * tree-object-size.h (compute_builtin_object_size): Drop PDECL
380 2021-10-07 Roger Sayle <roger@nextmovesoftware.com>
382 * rtl.def (SMUL_HIGHPART, UMUL_HIGHPART): New RTX codes for
383 representing signed and unsigned high-part multiplication resp.
384 * simplify-rtx.c (simplify_binary_operation_1) [SMUL_HIGHPART,
385 UMUL_HIGHPART]: Simplify high-part multiplications by zero.
386 [SS_PLUS, US_PLUS, SS_MINUS, US_MINUS, SS_MULT, US_MULT,
387 SS_DIV, US_DIV]: Similar simplifications for saturating
389 (simplify_const_binary_operation) [SS_PLUS, US_PLUS, SS_MINUS,
390 US_MINUS, SS_MULT, US_MULT, SMUL_HIGHPART, UMUL_HIGHPART]:
391 Implement compile-time evaluation for constant operands.
392 * dwarf2out.c (mem_loc_descriptor): Skip SMUL_HIGHPART and
394 * doc/rtl.texi (smul_highpart, umul_highpart): Document RTX codes.
395 * doc/md.texi (smul@var{m}3_highpart, umul@var{m3}_highpart):
396 Mention the new smul_highpart and umul_highpart RTX codes.
397 * doc/invoke.texi: Silence @xref "compilation" warnings.
399 2021-10-07 Martin Jambor <mjambor@suse.cz>
402 * ipa-prop.c (ipa_edge_args_sum_t::duplicate): Also handle the
403 case when the source reference description corresponds to a
404 referance taken in a function src->caller is inlined to.
406 2021-10-07 Jan Hubicka <hubicka@ucw.cz>
409 * ipa-modref-tree.h (modref_access_node::contains_p): Handle offsets
411 (modref_access_node::try_merge_with): Add sanity check that there
412 are no redundant entries in the list.
414 2021-10-07 Richard Biener <rguenther@suse.de>
416 PR tree-optimization/102608
417 * tree-ssa-sccvn.c (visit_stmt): Drop .DEFERRED_INIT to
420 2021-10-07 Martin Liska <mliska@suse.cz>
422 * toplev.c (toplev::main): Make
423 save_opt_decoded_options a pointer type
424 * toplev.h: Likewise.
426 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
428 * config/gcn/gcn-valu.md (gather<mode>_insn_2offsets<exec>): Apply
429 HAVE_GCN_ASM_GLOBAL_LOAD_FIXED.
430 (scatter<mode>_insn_2offsets<exec_scatter>): Likewise.
432 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
434 * config/gcn/gcn-hsa.h (SRAMOPT): Include the whole option string.
435 Adjust for new -msram-ecc=any behaviour.
436 (ASM_SPEC): Adjust -mxnack and -msram-ecc usage.
437 * config/gcn/gcn.c (output_file_start): Implement -msram-ecc=any.
438 * config/gcn/mkoffload.c (EF_AMDGPU_XNACK): Rename to ...
439 (EF_AMDGPU_XNACK_V3): ... this.
440 (EF_AMDGPU_SRAM_ECC): Rename to ...
441 (EF_AMDGPU_SRAM_ECC_V3): ... this.
442 (EF_AMDGPU_FEATURE_XNACK_V4): New.
443 (EF_AMDGPU_FEATURE_XNACK_UNSUPPORTED_V4): New.
444 (EF_AMDGPU_FEATURE_XNACK_ANY_V4): New.
445 (EF_AMDGPU_FEATURE_XNACK_OFF_V4): New.
446 (EF_AMDGPU_FEATURE_XNACK_ON_V4): New.
447 (EF_AMDGPU_FEATURE_SRAMECC_V4): New.
448 (EF_AMDGPU_FEATURE_SRAMECC_UNSUPPORTED_V4): New.
449 (EF_AMDGPU_FEATURE_SRAMECC_ANY_V4): New.
450 (EF_AMDGPU_FEATURE_SRAMECC_OFF_V4): New.
451 (EF_AMDGPU_FEATURE_SRAMECC_ON_V4): New.
453 (SET_XNACK_OFF): New.
455 (SET_SRAM_ECC_ON): New.
456 (SET_SRAM_ECC_ANY): New.
457 (SET_SRAM_ECC_OFF): New.
458 (TEST_SRAM_ECC_ANY): New.
459 (TEST_SRAM_ECC_ON): New.
460 (main): Implement HSACOv4 and -msram-ecc=any.
462 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
464 * config.in: Regenerate.
465 * config/gcn/gcn-hsa.h (X_FIJI): New macro.
469 (A_FIJI): Rename to ...
471 (A_900): Rename to ...
473 (A_906): Rename to ...
475 (A_908): Rename to ...
477 (SRAMOPT): New macro.
478 (ASM_SPEC): Adjust xnack option usage.
479 * config/gcn/gcn.c (output_file_start): Adjust amdgcn_target usage.
480 * configure: Regenerate.
481 * configure.ac: Detect LLVM assembler dialect.
483 2021-10-07 Richard Biener <rguenther@suse.de>
485 * tree-pretty-print.c (dump_generic_node): Do not elide
486 printing '&' when dumping with -gimple.
488 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
490 * gimple-range-cache.cc (non_null_ref::adjust_range): Call new
492 * gimple-range-fold.cc (adjust_pointer_diff_expr): Ditto.
493 (adjust_imagpart_expr): Ditto.
494 * value-range.cc (irange::irange_intersect): Call new routine if
495 RHS is a single pair.
496 (irange::intersect): New wide_int version.
497 * value-range.h (class irange): New prototype.
499 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
501 * gimple-range-edge.cc (gimple_outgoing_range::gimple_outgoing_range):
502 Add parameter to limit size when recognizing switches.
503 (gimple_outgoing_range::edge_range_p): Check size limit.
504 * gimple-range-edge.h (gimple_outgoing_range): Add size field.
505 * gimple-range-gori.cc (gori_map::calculate_gori): Ignore switches
506 that exceed the size limit.
507 (gori_compute::gori_compute): Add initializer.
508 * params.opt (evrp-switch-limit): New.
509 * doc/invoke.texi: Update docs.
511 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
513 * value-range.h (irange::set_varying): Use TYPE_MIN_VALUE and
514 TYPE_MAX_VALUE instead of creating new trees when possible.
516 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
518 * gimple-range-cache.cc (non_null_ref::adjust_range): Check for
519 zero and non-zero more efficently.
521 2021-10-06 Richard Biener <rguenther@suse.de>
524 * dumpfile.h (TDF_GIMPLE_VAL): New.
525 (dump_flag): Re-order and adjust TDF_* flags. Make
526 the enum uint32_t. Use std::underlying_type in the
528 (optgroup_flag): Likewise for the operator overloads.
529 * tree-pretty-print.c (dump_generic_node): Wrap ADDR_EXPR
530 in _Literal if TDF_GIMPLE_VAL.
531 * gimple-pretty-print.c (dump_gimple_assign): Add
532 TDF_GIMPLE_VAL to flags when dumping operands where only
533 is_gimple_val are allowed.
534 (dump_gimple_cond): Likewise.
536 2021-10-06 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
538 * gimple-isel.cc (gimple_expand_vec_cond_expr): Remove redundant if
541 2021-10-05 qing zhao <qing.zhao@oracle.com>
544 * gimplify.c (gimplify_decl_expr): Not add initialization for an
545 auto variable when it has been initialized by frontend.
547 2021-10-05 Aldy Hernandez <aldyh@redhat.com>
549 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
552 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
554 * common/config/avr/avr-common.c (avr_handle_option): Mark
555 argument as ATTRIBUTE_UNUSED.
557 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
559 * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC):
560 Undefine before redefinition.
562 2021-10-05 Richard Biener <rguenther@suse.de>
564 * toplev.c (no_backend): Remove global var.
565 (process_options): Pass in no_backend, move post_options
566 langhook call to toplev::main.
567 (do_compile): Pass in no_backend, move process_options call
569 (toplev::run_self_tests): Check no_backend at the caller.
570 (toplev::main): Call post_options and process_options
571 split out from do_compile, do self-tests only if
572 no_backend is initialized.
574 2021-10-05 Richard Biener <rguenther@suse.de>
576 * tree-cfg.c (dump_function_to_file): Dump the UID of the
577 function as part of the name when requested.
578 * tree-pretty-print.c (dump_function_name): Dump the UID when
579 requested and the langhook produced the actual name.
581 2021-10-05 Richard Biener <rguenther@suse.de>
585 * internal-fn.c (expand_DEFERRED_INIT): Fall back to
586 zero-initialization as last resort, use the constant
587 size as given by the DEFERRED_INIT argument to build
590 2021-10-04 Marek Polacek <polacek@redhat.com>
593 * doc/invoke.texi: Document -Warray-compare.
595 2021-10-04 Richard Biener <rguenther@suse.de>
597 * gimplify.c (is_var_need_auto_init): DECL_HARD_REGISTER
598 variables are not to be initialized.
600 2021-10-04 Richard Biener <rguenther@suse.de>
602 * expr.h (non_mem_decl_p): Declare.
603 (mem_ref_refers_to_non_mem_p): Likewise.
604 * expr.c (non_mem_decl_p): Export.
605 (mem_ref_refers_to_non_mem_p): Likewise.
606 * internal-fn.c (expand_DEFERRED_INIT): Do not expand the LHS
607 but check the base with mem_ref_refers_to_non_mem_p
610 2021-10-04 Richard Biener <rguenther@suse.de>
612 PR tree-optimization/102570
613 * tree-ssa-sccvn.h (vn_reference_op_struct): Document
614 we are using clique for the internal function code.
615 * tree-ssa-sccvn.c (vn_reference_op_eq): Compare the
616 internal function code.
617 (print_vn_reference_ops): Print the internal function code.
618 (vn_reference_op_compute_hash): Hash it.
619 (copy_reference_ops_from_call): Record it.
620 (visit_stmt): Remove the restriction around internal function
622 (fully_constant_vn_reference_p): Use fold_const_call and handle
624 (vn_reference_eq): Compare call return types.
625 * tree-ssa-pre.c (create_expression_by_pieces): Handle
626 generating calls to internal functions.
627 (compute_avail): Remove the restriction around internal function
630 2021-10-04 Aldy Hernandez <aldyh@redhat.com>
632 PR tree-optimization/102560
633 * gimple-ssa-warn-alloca.c (alloca_call_type): Remove static
634 marker for invalid_range.
636 2021-10-04 Richard Biener <rguenther@suse.de>
639 * internal-fn.c (expand_DEFERRED_INIT): Guard register
640 initialization path an avoid initializing VLA registers
643 2021-10-04 Eric Botcazou <ebotcazou@adacore.com>
645 * config/rs6000/vxworks.h (TARGET_INIT_LIBFUNCS): Delete.
647 2021-10-03 Martin Liska <mliska@suse.cz>
649 * toplev.c (toplev::main): Check opt_index if it is a part
652 2021-10-02 Aldy Hernandez <aldyh@redhat.com>
654 PR tree-optimization/102563
655 * range-op.cc (operator_lshift::op1_range): Do not clobber
658 2021-10-02 Martin Liska <mliska@suse.cz>
660 * toplev.c (toplev::main): save_decoded_options[0] is program
661 name and so it should be skipped.
663 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
665 PR tree-optimization/102546
666 * range-op.cc (operator_lshift::op1_range): Teach range-ops that
667 X << Y is non-zero implies X is also non-zero.
669 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
671 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
673 * config/aarch64/aarch64-tune.md: Regenerate.
674 * doc/invoke.texi: Update docs.
676 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
678 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
680 * config/aarch64/aarch64-tune.md: Regenerate.
681 * doc/invoke.texi: Update docs.
683 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
685 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
687 * config/aarch64/aarch64-tune.md: Regenerate.
688 * doc/invoke.texi: Update docs.
690 2021-10-01 Martin Sebor <msebor@redhat.com>
693 * doc/invoke.texi (-Waddress): Update.
694 * gengtype.c (write_types): Avoid -Waddress.
695 * poly-int.h (POLY_SET_COEFF): Avoid using null.
697 2021-10-01 John David Anglin <danglin@gcc.gnu.org>
700 * config/pa/pa.c (pa_option_override): Default to dwarf version 4
703 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
705 * config/aarch64/aarch64.h (AARCH64_FL_V9): Update value.
707 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
709 * gimple-range-path.cc (path_range_query::compute_ranges): Use
711 * gimple-range-path.h (class path_range_query): Remove shadowed
713 (path_range_query::get_path_oracle): New.
715 2021-10-01 Jakub Jelinek <jakub@redhat.com>
716 Richard Biener <rguenther@suse.de>
719 * doc/invoke.texi (-fsanitize=integer-divide-by-zero): Remove
720 INT_MIN / -1 division detection from here ...
721 (-fsanitize=signed-integer-overflow): ... and add it here.
723 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
725 * config/aarch64/aarch64-arches.def (AARCH64_ARCH): Added
727 * config/aarch64/aarch64.h (AARCH64_FL_V9): New.
728 (AARCH64_FL_FOR_ARCH9): New flags for Armv9-A.
729 (AARCH64_ISA_V9): New ISA flag.
730 * doc/invoke.texi: Update docs.
732 2021-10-01 Martin Liska <mliska@suse.cz>
734 * toplev.c (toplev::main): Save decoded optimization options.
735 * toplev.h (save_opt_decoded_options): New.
736 * doc/extend.texi: Be more clear about optimize and target
739 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
741 * explow.c: Include langhooks.h.
742 (set_stack_check_libfunc): Build a proper function type.
744 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
747 * config/i386/i386.c (legitimate_pic_address_disp_p): For PE-COFF do
748 not return true for external weak function symbols in medium model.
750 2021-10-01 Jakub Jelinek <jakub@redhat.com>
752 * tree.h (OMP_CLAUSE_ORDER_REPRODUCIBLE): Define.
753 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_ORDER>: Print
754 reproducible: for OMP_CLAUSE_ORDER_REPRODUCIBLE.
755 * omp-general.c (omp_extract_for_data): If OMP_CLAUSE_ORDER is seen
756 without OMP_CLAUSE_ORDER_UNCONSTRAINED, overwrite sched_kind to
757 OMP_CLAUSE_SCHEDULE_STATIC.
759 2021-10-01 Richard Biener <rguenther@suse.de>
762 * tree-inline.c (setup_one_parameter): Avoid substituting
763 an invariant into contexts where a GIMPLE register is not valid.
765 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
767 * config/arm/arm-cpus.in: Add Cortex-R52+ CPU.
768 * config/arm/arm-tables.opt: Regenerate.
769 * config/arm/arm-tune.md: Regenerate.
770 * doc/invoke.texi: Update docs.
772 2021-09-30 Uroš Bizjak <ubizjak@gmail.com>
775 * config/i386/i386.md
776 (sign_extend:WIDE (any_logic:NARROW (memory, immediate)) splitters):
779 2021-09-30 Tobias Burnus <tobias@codesourcery.com>
781 * omp-low.c (omp_runtime_api_call): Add omp_aligned_{,c}alloc and
782 omp_{c,re}alloc, fix omp_alloc/omp_free.
784 2021-09-30 Martin Liska <mliska@suse.cz>
786 * defaults.h (ASM_OUTPUT_ASCII): Do not hide global variable
787 asm_out_file and stream directly to MYFILE.
789 2021-09-30 Richard Biener <rguenther@suse.de>
791 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
792 Restore and fix condition under which we apply npeel to
793 the DRs misalignment value.
795 2021-09-30 Richard Biener <rguenther@suse.de>
797 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
798 Fix npeel check for variable amount of peeling.
800 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
802 * lto-wrapper.c (run_gcc): Plug snprintf overflow.
804 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
806 * gimple-range.cc (gimple_ranger::debug): New.
807 * gimple-range.h (class gimple_ranger): Add debug.
809 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
812 * tree-vrp.c (hybrid_threader::~hybrid_threader): Free m_query.
814 2021-09-29 Indu Bhagat <indu.bhagat@oracle.com>
817 * btfout.c (GTY): Add GTY (()) albeit for cosmetic only purpose.
818 (btf_finalize): Empty the hash_map btf_var_ids.
820 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
822 * tree-vrp.c (thread_through_all_blocks): Return bool.
823 (execute_vrp_threader): Return TODO_* flags.
824 (pass_data_vrp_threader): Set todo_flags_finish to 0.
826 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
828 * timevar.def (TV_TREE_VRP_THREADER): New.
829 * tree-vrp.c: Use TV_TREE_VRP_THREADER for VRP threader pass.
831 2021-09-29 David Faust <david.faust@oracle.com>
833 * config.gcc (bpf-*-*): Do not overwrite extra_headers.
835 2021-09-29 Jonathan Wright <jonathan.wright@arm.com>
837 * config/aarch64/aarch64-builtins.c (TYPES_BINOP_PPU): Define
838 new type qualifier enum.
839 (TYPES_TERNOP_SSSU): Likewise.
840 (TYPES_TERNOP_PPPU): Likewise.
841 * config/aarch64/aarch64-simd-builtins.def: Define PPU, SSU,
842 PPPU and SSSU builtin generator macros for qtbl1 and qtbx1
844 * config/aarch64/arm_neon.h (vqtbl1_p8): Use type-qualified
845 builtin and remove casts.
846 (vqtbl1_s8): Likewise.
847 (vqtbl1q_p8): Likewise.
848 (vqtbl1q_s8): Likewise.
849 (vqtbx1_s8): Likewise.
850 (vqtbx1_p8): Likewise.
851 (vqtbx1q_s8): Likewise.
852 (vqtbx1q_p8): Likewise.
853 (vtbl1_p8): Likewise.
854 (vtbl2_p8): Likewise.
855 (vtbx2_p8): Likewise.
857 2021-09-29 Richard Biener <rguenther@suse.de>
859 * tree-vect-data-refs.c (vect_dr_misalign_for_aligned_access):
861 (vect_update_misalignment_for_peel): Use it to update
862 misaligned to the value necessary for an aligned access.
863 (vect_get_peeling_costs_all_drs): Likewise.
864 (vect_enhance_data_refs_alignment): Likewise.
866 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
868 * config/aarch64/aarch64.c (aarch64_expand_cpymem): Count number of
869 emitted operations and adjust heuristic for code size.
871 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
873 * config/aarch64/aarch64.c (aarch64_expand_setmem): Count number of
874 emitted operations and adjust heuristic for code size.
876 2021-09-29 Jakub Jelinek <jakub@redhat.com>
879 * gimplify.c (gimplify_scan_omp_clauses): Use omp_check_private even
880 in OMP_SCOPE clauses, not just on worksharing construct clauses.
882 2021-09-28 Geng Qi <gengqi@linux.alibaba.com>
884 * config/riscv/riscv.md (mulv<mode>4): Call gen_smul<mode>3_highpart.
885 (<u>mulditi3): Call <su>muldi3_highpart.
886 (<u>muldi3_highpart): Rename to <su>muldi3_highpart.
887 (<u>mulsidi3): Call <su>mulsi3_highpart.
888 (<u>mulsi3_highpart): Rename to <su>mulsi3_highpart.
890 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
892 * config/darwin.h (DSYMUTIL_SPEC): Recognize D sources.
894 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
896 * config/rs6000/darwin.h (FIXED_R13): Add for PPC64.
897 (FIRST_SAVED_GP_REGNO): Save from R13 even when it is one
900 2021-09-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
902 * config/aarch64/aarch64.h (AARCH64_FL_LS64): Define
903 (AARCH64_FL_V8_7): Likewise.
904 (AARCH64_FL_FOR_ARCH8_7): Likewise.
905 * config/aarch64/aarch64-arches.def (armv8.7-a): Define.
906 * config/aarch64/aarch64-option-extensions.def (ls64): Define.
907 * doc/invoke.texi: Document the above.
909 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
911 * dbgcnt.c (dbg_cnt_counter): New.
912 * dbgcnt.h (dbg_cnt_counter): New.
913 * dumpfile.c (dump_options): Add entry for TDF_THREADING.
914 * dumpfile.h (enum dump_flag): Add TDF_THREADING.
915 * gimple-range-path.cc (DEBUG_SOLVER): Use TDF_THREADING.
916 * tree-ssa-threadupdate.c (dump_jump_thread_path): Dump out
919 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
921 * cfgcleanup.c (pass_jump::execute): Check
922 flag_expensive_optimizations.
923 (pass_jump_after_combine::gate): Same.
924 * doc/invoke.texi (-fthread-jumps): Enable for -O1.
925 * opts.c (default_options_table): Enable -fthread-jumps at -O1.
926 * tree-ssa-threadupdate.c
927 (fwd_jt_path_registry::remove_jump_threads_including): Bail unless
930 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
932 * tree-ssa-reassoc.c (biased_names): New global.
933 (propagate_bias_p): New function.
934 (loop_carried_phi): Remove.
935 (propagate_rank): Propagate bias along single uses.
936 (get_rank): Update biased_names when needed.
938 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
940 * passes.def (pass_reassoc): Rename parameter to early_p.
941 * tree-ssa-reassoc.c (reassoc_bias_loop_carried_phi_ranks_p):
943 (phi_rank): Don't bias loop-carried phi ranks
944 before vectorization pass.
945 (execute_reassoc): Add bias_loop_carried_phi_ranks_p parameter.
946 (pass_reassoc::pass_reassoc): Add bias_loop_carried_phi_ranks_p
948 (pass_reassoc::set_param): Set bias_loop_carried_phi_ranks_p
950 (pass_reassoc::execute): Pass bias_loop_carried_phi_ranks_p to
952 (pass_reassoc::bias_loop_carried_phi_ranks_p): New member.
954 2021-09-28 Jakub Jelinek <jakub@redhat.com>
957 * config/i386/i386.c (standard_80387_constant_p): Don't recognize
958 special 80387 instruction XFmode constants if flag_rounding_math.
960 2021-09-28 Richard Biener <rguenther@suse.de>
962 PR tree-optimization/100112
963 * tree-ssa-sccvn.c (visit_reference_op_load): Record the
964 referece into the hashtable twice in case last_vuse is
965 different from the original vuse on the stmt.
967 2021-09-28 Jakub Jelinek <jakub@redhat.com>
970 * gimplify.c (gimplify_adjust_omp_clauses_1): Don't call the
971 omp_finish_clause langhook on implicitly added OMP_CLAUSE_PRIVATE
972 clauses on SIMD constructs.
974 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
976 PR tree-optimization/102511
977 * gimple-range-path.cc (path_range_query::range_on_path_entry):
978 Return VARYING when nothing found.
980 2021-09-28 Hongyu Wang <hongyu.wang@intel.com>
983 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): Add
985 (VALID_SSE2_REG_VHF_MODE): Add V4HFmode and V2HFmode.
986 (VALID_MMX_REG_MODE): Add V4HFmode.
987 (SSE_REG_MODE_P): Replace VALID_AVX512FP16_REG_MODE with
988 vector mode condition.
989 * config/i386/i386.c (classify_argument): Parse V4HF/V2HF
991 (function_arg_32): Add V4HFmode.
992 (function_arg_advance_32): Likewise.
993 * config/i386/i386.md (mode): Add V4HF/V2HF.
994 (MODE_SIZE): Likewise.
995 * config/i386/mmx.md (MMXMODE): Add V4HF mode.
996 (V_32): Add V2HF mode.
997 (VHF_32_64): New mode iterator.
998 (*mov<mode>_internal): Adjust sse alternatives to support
1000 (*mov<mode>_internal): Adjust sse alternatives to support
1002 (<insn><mode>3): New define_insn for add/sub/mul/div.
1004 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
1006 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Check
1008 (pass_early_thread_jumps::gate): Same.
1009 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
1010 Return if !flag_thread_jumps.
1011 * tree-ssa-threadupdate.c
1012 (jt_path_registry::register_jump_thread): Assert that
1013 flag_thread_jumps is true.
1015 2021-09-28 liuhongt <hongtao.liu@intel.com>
1018 (simplify_context::simplify_binary_operation_1): Relax
1019 condition of simplifying (vec_concat:M (vec_select op0
1020 index0)(vec_select op1 index1)) to allow different modes
1021 between op0 and M, but have same inner mode.
1023 2021-09-28 liuhongt <hongtao.liu@intel.com>
1025 * config/i386/i386-expand.c (emit_reduc_half): Handle
1026 V8HF/V16HF/V32HFmode.
1027 * config/i386/sse.md (REDUC_SSE_PLUS_MODE): Add V8HF.
1028 (REDUC_SSE_SMINMAX_MODE): Ditto.
1029 (REDUC_PLUS_MODE): Add V16HF and V32HF.
1030 (REDUC_SMINMAX_MODE): Ditto.
1032 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
1034 * gimple-range-path.cc
1035 (path_range_query::precompute_ranges_in_block): Rename to...
1036 (path_range_query::compute_ranges_in_block): ...this.
1037 (path_range_query::precompute_ranges): Rename to...
1038 (path_range_query::compute_ranges): ...this.
1039 (path_range_query::precompute_relations): Rename to...
1040 (path_range_query::compute_relations): ...this.
1041 (path_range_query::precompute_phi_relations): Rename to...
1042 (path_range_query::compute_phi_relations): ...this.
1043 * gimple-range-path.h: Rename precompute* to compute*.
1044 * tree-ssa-threadbackward.c
1045 (back_threader::find_taken_edge_switch): Same.
1046 (back_threader::find_taken_edge_cond): Same.
1047 * tree-ssa-threadedge.c
1048 (hybrid_jt_simplifier::compute_ranges_from_state): Same.
1049 (hybrid_jt_state::register_equivs_stmt): Inline...
1050 * tree-ssa-threadedge.h: ...here.
1052 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
1054 * tree-vrp.c (lhs_of_dominating_assert): Remove.
1055 (class vrp_jt_state): Remove.
1056 (class vrp_jt_simplifier): Remove.
1057 (vrp_jt_simplifier::simplify): Remove.
1058 (class vrp_jump_threader): Remove.
1059 (vrp_jump_threader::vrp_jump_threader): Remove.
1060 (vrp_jump_threader::~vrp_jump_threader): Remove.
1061 (vrp_jump_threader::before_dom_children): Remove.
1062 (vrp_jump_threader::after_dom_children): Remove.
1064 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
1066 * passes.def (pass_vrp_threader): New.
1067 * tree-pass.h (make_pass_vrp_threader): Add make_pass_vrp_threader.
1068 * tree-ssa-threadedge.c (hybrid_jt_state::register_equivs_stmt): New.
1069 (hybrid_jt_simplifier::hybrid_jt_simplifier): New.
1070 (hybrid_jt_simplifier::simplify): New.
1071 (hybrid_jt_simplifier::compute_ranges_from_state): New.
1072 * tree-ssa-threadedge.h (class hybrid_jt_state): New.
1073 (class hybrid_jt_simplifier): New.
1074 * tree-vrp.c (execute_vrp): Remove ASSERT_EXPR based jump
1076 (class hybrid_threader): New.
1077 (hybrid_threader::hybrid_threader): New.
1078 (hybrid_threader::~hybrid_threader): New.
1079 (hybrid_threader::before_dom_children): New.
1080 (hybrid_threader::after_dom_children): New.
1081 (execute_vrp_threader): New.
1082 (class pass_vrp_threader): New.
1083 (make_pass_vrp_threader): New.
1085 2021-09-27 Martin Liska <mliska@suse.cz>
1087 * output.h (enum section_flag): New.
1088 (SECTION_FORGET): Remove.
1089 (SECTION_ENTSIZE): Make it (1UL << 8) - 1.
1090 (SECTION_STYLE_MASK): Define it based on other enum
1092 * varasm.c (switch_to_section): Remove unused handling of
1095 2021-09-27 Martin Liska <mliska@suse.cz>
1097 * common.opt: Add new variable flag_default_complex_method.
1098 * opts.c (finish_options): Handle flags related to
1099 x_flag_complex_method.
1100 * toplev.c (process_options): Remove option handling related
1101 to flag_complex_method.
1103 2021-09-27 Richard Biener <rguenther@suse.de>
1105 PR middle-end/102450
1106 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid using
1107 type_for_size, instead use int_mode_for_size.
1109 2021-09-27 Andrew Pinski <apinski@marvell.com>
1112 * gimplify.c (gimplify_save_expr): Return early
1113 if the type of val is error_mark_node.
1115 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
1117 * tree-ssanames.c (ssa_name_has_boolean_range): Use
1120 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
1122 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
1123 vrp_visit_cond_stmt.
1124 * tree-ssa-dom.c (cprop_operand): Convert to range_query API.
1125 (cprop_into_stmt): Same.
1126 (dom_opt_dom_walker::optimize_stmt): Same.
1128 2021-09-27 Richard Biener <rguenther@suse.de>
1130 PR tree-optimization/97351
1131 PR tree-optimization/97352
1132 PR tree-optimization/82426
1133 * tree-vectorizer.h (dr_misalignment): Add vector type
1135 (aligned_access_p): Likewise.
1136 (known_alignment_for_access_p): Likewise.
1137 (vect_supportable_dr_alignment): Likewise.
1138 (vect_known_alignment_in_bytes): Likewise. Refactor.
1139 (DR_MISALIGNMENT): Remove.
1140 (vect_update_shared_vectype): Likewise.
1141 * tree-vect-data-refs.c (dr_misalignment): Refactor, handle
1142 a vector type with larger alignment requirement and apply
1143 the negative step adjustment here.
1144 (vect_calculate_target_alignment): Remove.
1145 (vect_compute_data_ref_alignment): Get explicit vector type
1146 argument, do not apply a negative step alignment adjustment
1148 (vect_slp_analyze_node_alignment): Re-analyze alignment
1149 when we re-visit the DR with a bigger desired alignment but
1150 keep more precise results from smaller alignments.
1151 * tree-vect-slp.c (vect_update_shared_vectype): Remove.
1152 (vect_slp_analyze_node_operations_1): Do not update the
1153 shared vector type on stmts.
1154 * tree-vect-stmts.c (vect_analyze_stmt): Push/pop the
1155 vector type of an SLP node to the representative stmt-info.
1156 (vect_transform_stmt): Likewise.
1158 2021-09-27 liuhongt <hongtao.liu@intel.com>
1161 2021-09-09 liuhongt <hongtao.liu@intel.com>
1164 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
1165 (reduc_plus_scal_v4sf): .. this, New define_expand.
1166 (reduc_plus_scal_v2df): .. and this, New define_expand.
1168 2021-09-26 liuhongt <hongtao.liu@intel.com>
1170 * doc/extend.texi (Half-Precision): Remove storage only
1171 description for _Float16 w/o avx512fp16.
1173 2021-09-25 Dimitar Dimitrov <dimitar@dinux.eu>
1175 * config/pru/constraints.md (Rrio): New constraint.
1176 * config/pru/predicates.md (regio_operand): New predicate.
1177 * config/pru/pru-pragma.c (pru_register_pragmas): Register
1178 the __regio_symbol address space.
1179 * config/pru/pru-protos.h (pru_symref2ioregno): Declaration.
1180 * config/pru/pru.c (pru_symref2ioregno): New helper function.
1181 (pru_legitimate_address_p): Remove.
1182 (pru_addr_space_legitimate_address_p): Use the address space
1184 (pru_nongeneric_pointer_addrspace): New helper function.
1185 (pru_insert_attributes): New function to validate __regio_symbol
1187 (TARGET_INSERT_ATTRIBUTES): New macro.
1188 (TARGET_LEGITIMATE_ADDRESS_P): Remove.
1189 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): New macro.
1190 * config/pru/pru.h (enum reg_class): Add REGIO_REGS class.
1191 * config/pru/pru.md (*regio_readsi): New pattern to read I/O
1193 (*regio_nozext_writesi): New pattern to write to I/O registers.
1194 (*regio_zext_write_r30<EQS0:mode>): Ditto.
1195 * doc/extend.texi: Document the new PRU Named Address Space.
1197 2021-09-24 Patrick Palka <ppalka@redhat.com>
1201 * real.c (encode_ieee_double): Avoid unwanted sign extension.
1202 (encode_ieee_quad): Likewise.
1204 2021-09-24 Vladimir Makarov <vmakarov@redhat.com>
1206 PR rtl-optimization/102147
1207 * ira-build.c (ira_conflict_vector_profitable_p): Make
1208 profitability calculation independent of host compiler pointer and
1211 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
1213 * gimple-range-path.cc (path_range_query::path_range_query):
1214 Move debugging header...
1215 (path_range_query::precompute_ranges): ...here.
1216 (path_range_query::internal_range_of_expr): Do not call
1217 range_on_path_entry if NAME is defined in the current block.
1219 2021-09-24 Richard Biener <rguenther@suse.de>
1221 * cfghooks.c (verify_flow_info): Verify unallocated BB and
1222 edge flags are not set.
1224 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
1226 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
1228 (jt_path_registry::register_jump_thread): Call
1229 cancel_invalid_paths.
1230 * tree-ssa-threadupdate.h (class jt_path_registry): Add
1231 cancel_invalid_paths.
1233 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
1235 PR tree-optimization/102400
1236 * tree-ssa-sccvn.c (vn_reference_insert_pieces): Initialize
1237 result_vdef to zero value.
1239 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
1241 PR tree-optimization/102451
1242 * tree-ssa-dse.c (delete_dead_or_redundant_call): Record bb of stmt
1245 2021-09-24 Hongyu Wang <hongyu.wang@intel.com>
1247 * config/i386/sse.md (cond_<insn><mode>): Extend to support
1249 (cond_mul<mode>): Likewise.
1250 (cond_div<mode>): Likewise.
1251 (cond_<code><mode>): Likewise.
1252 (cond_fma<mode>): Likewise.
1253 (cond_fms<mode>): Likewise.
1254 (cond_fnma<mode>): Likewise.
1255 (cond_fnms<mode>): Likewise.
1257 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
1259 PR tree-optimization/102463
1260 * gimple-range-fold.cc (fold_using_range::relation_fold_and_or): If
1261 there is no range-ops handler, don't look for a relation.
1263 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
1265 * gimple-range-cache.cc (ranger_cache::ranger_cache): Take
1266 non-executable_edge flag as parameter.
1267 * gimple-range-cache.h (ranger_cache): Adjust prototype.
1268 * gimple-range-gori.cc (gori_compute::gori_compute): Take
1269 non-executable_edge flag as parameter.
1270 (gori_compute::outgoing_edge_range_p): Check new flag.
1271 * gimple-range-gori.h (gori_compute): Adjust prototype.
1272 * gimple-range.cc (gimple_ranger::gimple_ranger): Create new flag.
1273 (gimple_ranger::range_on_edge): Check new flag.
1274 * gimple-range.h (gimple_ranger::non_executable_edge_flag): New.
1275 * gimple-ssa-evrp.c (rvrp_folder): Pass ranger flag to simplifer.
1276 (hybrid_folder::hybrid_folder): Set ranger non-executable flag value.
1277 (hybrid_folder::fold_stmt): Set flag value in the simplifer.
1278 * vr-values.c (simplify_using_ranges::set_and_propagate_unexecutable):
1279 Use not_executable flag if provided inmstead of EDGE_EXECUTABLE.
1280 (simplify_using_ranges::simplify_switch_using_ranges): Clear
1281 EDGE_EXECUTABLE like it originally did.
1282 (simplify_using_ranges::cleanup_edges_and_switches): Clear any
1283 NON_EXECUTABLE flags.
1284 (simplify_using_ranges::simplify_using_ranges): Adjust.
1285 * vr-values.h (class simplify_using_ranges): Adjust.
1286 (simplify_using_ranges::set_range_query): Add non-executable flag param.
1288 2021-09-23 Bill Schmidt <wschmidt@linux.ibm.com>
1291 * config/rs6000/rs6000-call.c (rs6000_aggregate_candidate): Detect
1292 zero-width bit fields and return indicator.
1293 (rs6000_discover_homogeneous_aggregate): Diagnose when the
1294 presence of a zero-width bit field changes parameter passing in
1297 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
1299 * gimple-range-fold.cc (fold_using_range::range_of_phi):
1300 Remove dominator check.
1302 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
1304 * gimple-range-path.cc (path_range_query::precompute_relations):
1305 Hoist edge calculations before using EDGE_SUCC.
1307 2021-09-23 Jonathan Wakely <jwakely@redhat.com>
1309 * configure.ac: Fix --with-multilib-list description.
1310 * configure: Regenerate.
1312 2021-09-23 Richard Biener <rguenther@suse.de>
1314 PR tree-optimization/102448
1315 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
1316 Clear alignment info copied from DR_PTR_INFO.
1318 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
1320 * config/i386/i386-expand.c (ix86_use_mask_cmp_p): Enable
1322 * config/i386/sse.md (sseintvecmodelower): Add HF vector modes.
1323 (<avx512>_store<mode>_mask): Extend to support HF vector modes.
1324 (vec_cmp<mode><avx512fmaskmodelower>): Likewise.
1325 (vcond_mask_<mode><avx512fmaskmodelower>): Likewise.
1326 (vcond<mode><mode>): New expander.
1327 (vcond<mode><sseintvecmodelower>): Likewise.
1328 (vcond<sseintvecmodelower><mode>): Likewise.
1329 (vcondu<mode><sseintvecmodelower>): Likewise.
1331 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
1333 * config/i386/sse.md (extend<ssePHmodelower><mode>2):
1335 (extendv4hf<mode>2): Likewise.
1336 (extendv2hfv2df2): Likewise.
1337 (trunc<mode><ssePHmodelower>2): Likewise.
1338 (avx512fp16_vcvt<castmode>2ph_<mode>): Rename to ...
1339 (trunc<mode>v4hf2): ... this, and drop constraints.
1340 (avx512fp16_vcvtpd2ph_v2df): Rename to ...
1341 (truncv2dfv2hf2): ... this, and likewise.
1343 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
1345 * config/i386/sse.md (float<floatunssuffix><mode><ssePHmodelower>2):
1347 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>):
1349 (float<floatunssuffix><mode>v4hf2): ... this, and drop constraints.
1350 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Rename to ...
1351 (float<floatunssuffix>v2div2hf2): ... this, and likewise.
1353 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
1355 * config/i386/i386.md (fix<fixunssuffix>_trunchf<mode>2): New expander.
1356 (fixuns_trunchfhi2): Likewise.
1357 (*fixuns_trunchfsi2zext): New define_insn.
1358 * config/i386/sse.md (ssePHmodelower): New mode_attr.
1359 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
1360 New expander for same element vector fix_truncate.
1361 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
1362 Likewise for V4HF to V4SI/V4DI fix_truncate.
1363 (fix<fixunssuffix>_truncv2hfv2di2):
1364 Likeise for V2HF to V2DI fix_truncate.
1366 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
1368 * config/i386/i386.md (<code>hf3): New expander.
1370 2021-09-23 liuhongt <hongtao.liu@intel.com>
1372 * config/i386/sse.md (FMAMODEM): extend to handle FP16.
1373 (VFH_SF_AVX512VL): Extend to handle HFmode.
1374 (VF_SF_AVX512VL): Deleted.
1376 2021-09-23 liuhongt <hongtao.liu@intel.com>
1378 * config/i386/i386.md (rinthf2): New expander.
1379 (nearbyinthf2): New expander.
1381 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
1383 * tree-ssa-dom.c (class dom_jump_threader_simplifier): Rename...
1384 (class dom_jt_state): ...this and provide virtual overrides.
1385 (dom_jt_state::register_equiv): New.
1386 (class dom_jt_simplifier): Rename from
1387 dom_jump_threader_simplifier.
1388 (dom_jump_threader_simplifier::simplify): Rename...
1389 (dom_jt_simplifier::simplify): ...to this.
1390 (pass_dominator::execute): Use dom_jt_simplifier and
1392 * tree-ssa-threadedge.c (jump_threader::jump_threader):
1394 (jt_state::register_equivs_stmt): Abstract out...
1395 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
1397 (jump_threader::thread_around_empty_blocks): Update state.
1398 (jump_threader::thread_through_normal_block): Same.
1399 (jt_state::jt_state): Remove.
1400 (jt_state::push): Remove pass specific bits. Keep block vector
1402 (jt_state::append_path): New.
1403 (jt_state::pop): Remove pass specific bits.
1404 (jt_state::register_equiv): Same.
1405 (jt_state::record_ranges_from_stmt): Same.
1406 (jt_state::register_equivs_on_edge): Same. Rename...
1407 (jt_state::register_equivs_edge): ...to this.
1408 (jt_state::dump): New.
1409 (jt_state::debug): New.
1410 (jump_threader_simplifier::simplify): Remove.
1411 (jt_state::get_path): New.
1412 * tree-ssa-threadedge.h (class jt_simplifier): Make into a base
1413 class. Expose common functionality as virtual methods.
1414 (class jump_threader_simplifier): Same. Rename...
1415 (class jt_simplifier): ...to this.
1416 * tree-vrp.c (class vrp_jump_threader_simplifier): Rename...
1417 (class vrp_jt_simplifier): ...to this. Provide pass specific
1419 (class vrp_jt_state): New.
1420 (vrp_jump_threader_simplifier::simplify): Rename...
1421 (vrp_jt_simplifier::simplify): ...to this. Inline code from
1422 what used to be the base class.
1423 (vrp_jump_threader::vrp_jump_threader): Use vrp_jt_state and
1426 2021-09-22 Tobias Burnus <tobias@codesourcery.com>
1429 * doc/invoke.texi (-Wno-missing-include-dirs.): Document Fortran
1432 2021-09-22 Roger Sayle <roger@nextmovesoftware.com>
1433 Richard Biener <rguenther@suse.de>
1435 * match.pd (negation simplifications): Implement some negation
1436 folding transformations from fold-const.c's fold_negate_expr.
1437 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Add a SIMPLIFY
1438 argument, to control whether the op should be simplified prior
1439 to looking up/assigning a value number.
1440 (vn_nary_build_or_lookup): Update call to vn_nary_build_or_lookup_1.
1441 (vn_nary_simplify): Likewise.
1442 (visit_nary_op): Likewise, but when constructing a NEGATE_EXPR
1443 now call vn_nary_build_or_lookup_1 disabling simplification.
1445 2021-09-22 Jiufu Guo <guojiufu@linux.ibm.com>
1447 PR tree-optimization/102087
1448 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
1449 Update bound/cmp/control for niter.
1451 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
1453 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
1454 Move check for non-empty BB here.
1455 (fur_source::register_outgoing_edges): ...from here.
1457 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
1459 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
1460 Remove call to improve_range_with_equivs.
1461 (path_range_query::improve_range_with_equivs): Remove
1462 * gimple-range-path.h: Remove improve_range_with_equivs.
1464 2021-09-22 dianhong xu <dianhong.xu@intel.com>
1466 * config/i386/avx512fp16intrin.h:
1467 (_mm512_mask_blend_ph): New intrinsic.
1468 (_mm512_permutex2var_ph): Ditto.
1469 (_mm512_permutexvar_ph): Ditto.
1470 * config/i386/avx512fp16vlintrin.h:
1471 (_mm256_mask_blend_ph): New intrinsic.
1472 (_mm256_permutex2var_ph): Ditto.
1473 (_mm256_permutexvar_ph): Ditto.
1474 (_mm_mask_blend_ph): Ditto.
1475 (_mm_permutex2var_ph): Ditto.
1476 (_mm_permutexvar_ph): Ditto.
1478 2021-09-22 dianhong xu <dianhong.xu@intel.com>
1480 * config/i386/avx512fp16intrin.h: Add new intrinsics.
1481 (_mm512_conj_pch): New intrinsic.
1482 (_mm512_mask_conj_pch): Ditto.
1483 (_mm512_maskz_conj_pch): Ditto.
1484 * config/i386/avx512fp16vlintrin.h: Add new intrinsics.
1485 (_mm256_conj_pch): New intrinsic.
1486 (_mm256_mask_conj_pch): Ditto.
1487 (_mm256_maskz_conj_pch): Ditto.
1488 (_mm_conj_pch): Ditto.
1489 (_mm_mask_conj_pch): Ditto.
1490 (_mm_maskz_conj_pch): Ditto.
1492 2021-09-22 dianhong xu <dianhong.xu@intel.com>
1494 * config/i386/avx512fp16intrin.h (_MM512_REDUCE_OP): New macro
1495 (_mm512_reduce_add_ph): New intrinsic.
1496 (_mm512_reduce_mul_ph): Ditto.
1497 (_mm512_reduce_min_ph): Ditto.
1498 (_mm512_reduce_max_ph): Ditto.
1499 * config/i386/avx512fp16vlintrin.h
1500 (_MM256_REDUCE_OP/_MM_REDUCE_OP): New macro.
1501 (_mm256_reduce_add_ph): New intrinsic.
1502 (_mm256_reduce_mul_ph): Ditto.
1503 (_mm256_reduce_min_ph): Ditto.
1504 (_mm256_reduce_max_ph): Ditto.
1505 (_mm_reduce_add_ph): Ditto.
1506 (_mm_reduce_mul_ph): Ditto.
1507 (_mm_reduce_min_ph): Ditto.
1508 (_mm_reduce_max_ph): Ditto.
1510 2021-09-22 dianhong xu <dianhong.xu@intel.com>
1512 * config/i386/avx512fp16intrin.h (__m512h_u, __m256h_u,
1513 __m128h_u): New typedef.
1514 (_mm512_load_ph): New intrinsic.
1515 (_mm256_load_ph): Ditto.
1516 (_mm_load_ph): Ditto.
1517 (_mm512_loadu_ph): Ditto.
1518 (_mm256_loadu_ph): Ditto.
1519 (_mm_loadu_ph): Ditto.
1520 (_mm512_store_ph): Ditto.
1521 (_mm256_store_ph): Ditto.
1522 (_mm_store_ph): Ditto.
1523 (_mm512_storeu_ph): Ditto.
1524 (_mm256_storeu_ph): Ditto.
1525 (_mm_storeu_ph): Ditto.
1526 (_mm512_abs_ph): Ditto.
1527 * config/i386/avx512fp16vlintrin.h
1528 (_mm_abs_ph): Ditto.
1529 (_mm256_abs_ph): Ditto.
1531 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
1533 * config/s390/tpf.md (prologue_tpf, epilogue_tpf): Add cc clobber.
1535 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
1538 * config/s390/s390.c (s390_expand_insv): Emit a normal move if it
1539 is actually a full copy of the source operand into the target.
1540 Don't emit a strict low part move if source and target mode match.
1542 2021-09-22 Jakub Jelinek <jakub@redhat.com>
1544 PR middle-end/102415
1545 * omp-expand.c (expand_omp_single): If region->exit is NULL,
1546 assert region->entry is GIMPLE_OMP_SCOPE region and return.
1548 2021-09-22 Jakub Jelinek <jakub@redhat.com>
1550 * tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
1551 * tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
1552 arguments from 2 to 3.
1553 * tree-pretty-print.c (dump_omp_clause): Print allocator() around
1554 allocate clause allocator and print align if present.
1555 * omp-low.c (scan_sharing_clauses): Force allocate_map entry even
1556 for omp_default_mem_alloc if align modifier is present. If align
1557 modifier is present, use TREE_LIST to encode both allocator and
1559 (lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
1560 Handle align modifier on allocator clause if present.
1562 2021-09-22 liuhongt <hongtao.liu@intel.com>
1564 * config/i386/i386.md (define_attr "isa"): Add
1566 (define_attr "enabled"): Correspond fma_or_avx512vl to
1567 TARGET_FMA || TARGET_AVX512VL.
1568 * config/i386/mmx.md (fmav2sf4): Extend to AVX512 fma.
1573 2021-09-22 liuhongt <hongtao.liu@intel.com>
1575 * config/i386/i386.md (cstorehf3): New define_expand.
1577 2021-09-22 liuhongt <hongtao.liu@intel.com>
1579 * config/i386/i386.md (<rounding_insn>hf2): New expander.
1580 (sse4_1_round<mode>2): Extend from MODEF to MODEFH.
1581 * config/i386/sse.md (*sse4_1_round<ssescalarmodesuffix>):
1582 Extend from VF_128 to VFH_128.
1584 2021-09-22 liuhongt <hongtao.liu@intel.com>
1586 * config/i386/i386-features.c (i386-features.c): Handle
1588 * config/i386/i386.md (sqrthf2): New expander.
1589 (*sqrthf2): New define_insn.
1590 * config/i386/sse.md
1591 (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
1594 2021-09-22 liuhongt <hongtao.liu@intel.com>
1596 * config/i386/avx512fp16intrin.h (_mm_mask_fcmadd_sch):
1598 (_mm_mask3_fcmadd_sch): Likewise.
1599 (_mm_maskz_fcmadd_sch): Likewise.
1600 (_mm_fcmadd_sch): Likewise.
1601 (_mm_mask_fmadd_sch): Likewise.
1602 (_mm_mask3_fmadd_sch): Likewise.
1603 (_mm_maskz_fmadd_sch): Likewise.
1604 (_mm_fmadd_sch): Likewise.
1605 (_mm_mask_fcmadd_round_sch): Likewise.
1606 (_mm_mask3_fcmadd_round_sch): Likewise.
1607 (_mm_maskz_fcmadd_round_sch): Likewise.
1608 (_mm_fcmadd_round_sch): Likewise.
1609 (_mm_mask_fmadd_round_sch): Likewise.
1610 (_mm_mask3_fmadd_round_sch): Likewise.
1611 (_mm_maskz_fmadd_round_sch): Likewise.
1612 (_mm_fmadd_round_sch): Likewise.
1613 (_mm_fcmul_sch): Likewise.
1614 (_mm_mask_fcmul_sch): Likewise.
1615 (_mm_maskz_fcmul_sch): Likewise.
1616 (_mm_fmul_sch): Likewise.
1617 (_mm_mask_fmul_sch): Likewise.
1618 (_mm_maskz_fmul_sch): Likewise.
1619 (_mm_fcmul_round_sch): Likewise.
1620 (_mm_mask_fcmul_round_sch): Likewise.
1621 (_mm_maskz_fcmul_round_sch): Likewise.
1622 (_mm_fmul_round_sch): Likewise.
1623 (_mm_mask_fmul_round_sch): Likewise.
1624 (_mm_maskz_fmul_round_sch): Likewise.
1625 * config/i386/i386-builtin.def: Add corresponding new builtins.
1626 * config/i386/sse.md
1627 (avx512fp16_fmaddcsh_v8hf_maskz<round_expand_name>): New expander.
1628 (avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>): Ditto.
1629 (avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>):
1631 (avx512fp16_<complexopname>sh_v8hf_mask<round_name>): Ditto.
1632 (avx512fp16_<complexopname>sh_v8hf<mask_scalarc_name><round_scalarcz_name>):
1634 * config/i386/subst.md (mask_scalarcz_name): New.
1635 (mask_scalarc_name): Ditto.
1636 (mask_scalarc_operand3): Ditto.
1637 (mask_scalarcz_operand4): Ditto.
1638 (round_scalarcz_name): Ditto.
1639 (round_scalarc_mask_operand3): Ditto.
1640 (round_scalarcz_mask_operand4): Ditto.
1641 (round_scalarc_mask_op3): Ditto.
1642 (round_scalarcz_mask_op4): Ditto.
1643 (round_scalarcz_constraint): Ditto.
1644 (round_scalarcz_nimm_predicate): Ditto.
1645 (mask_scalarcz): Ditto.
1646 (mask_scalarc): Ditto.
1647 (round_scalarcz): Ditto.
1649 2021-09-22 liuhongt <hongtao.liu@intel.com>
1651 * config/i386/avx512fp16intrin.h (_mm512_fcmadd_pch):
1653 (_mm512_mask_fcmadd_pch): Likewise.
1654 (_mm512_mask3_fcmadd_pch): Likewise.
1655 (_mm512_maskz_fcmadd_pch): Likewise.
1656 (_mm512_fmadd_pch): Likewise.
1657 (_mm512_mask_fmadd_pch): Likewise.
1658 (_mm512_mask3_fmadd_pch): Likewise.
1659 (_mm512_maskz_fmadd_pch): Likewise.
1660 (_mm512_fcmadd_round_pch): Likewise.
1661 (_mm512_mask_fcmadd_round_pch): Likewise.
1662 (_mm512_mask3_fcmadd_round_pch): Likewise.
1663 (_mm512_maskz_fcmadd_round_pch): Likewise.
1664 (_mm512_fmadd_round_pch): Likewise.
1665 (_mm512_mask_fmadd_round_pch): Likewise.
1666 (_mm512_mask3_fmadd_round_pch): Likewise.
1667 (_mm512_maskz_fmadd_round_pch): Likewise.
1668 (_mm512_fcmul_pch): Likewise.
1669 (_mm512_mask_fcmul_pch): Likewise.
1670 (_mm512_maskz_fcmul_pch): Likewise.
1671 (_mm512_fmul_pch): Likewise.
1672 (_mm512_mask_fmul_pch): Likewise.
1673 (_mm512_maskz_fmul_pch): Likewise.
1674 (_mm512_fcmul_round_pch): Likewise.
1675 (_mm512_mask_fcmul_round_pch): Likewise.
1676 (_mm512_maskz_fcmul_round_pch): Likewise.
1677 (_mm512_fmul_round_pch): Likewise.
1678 (_mm512_mask_fmul_round_pch): Likewise.
1679 (_mm512_maskz_fmul_round_pch): Likewise.
1680 * config/i386/avx512fp16vlintrin.h (_mm_fmadd_pch):
1682 (_mm_mask_fmadd_pch): Likewise.
1683 (_mm_mask3_fmadd_pch): Likewise.
1684 (_mm_maskz_fmadd_pch): Likewise.
1685 (_mm256_fmadd_pch): Likewise.
1686 (_mm256_mask_fmadd_pch): Likewise.
1687 (_mm256_mask3_fmadd_pch): Likewise.
1688 (_mm256_maskz_fmadd_pch): Likewise.
1689 (_mm_fcmadd_pch): Likewise.
1690 (_mm_mask_fcmadd_pch): Likewise.
1691 (_mm_mask3_fcmadd_pch): Likewise.
1692 (_mm_maskz_fcmadd_pch): Likewise.
1693 (_mm256_fcmadd_pch): Likewise.
1694 (_mm256_mask_fcmadd_pch): Likewise.
1695 (_mm256_mask3_fcmadd_pch): Likewise.
1696 (_mm256_maskz_fcmadd_pch): Likewise.
1697 (_mm_fmul_pch): Likewise.
1698 (_mm_mask_fmul_pch): Likewise.
1699 (_mm_maskz_fmul_pch): Likewise.
1700 (_mm256_fmul_pch): Likewise.
1701 (_mm256_mask_fmul_pch): Likewise.
1702 (_mm256_maskz_fmul_pch): Likewise.
1703 (_mm_fcmul_pch): Likewise.
1704 (_mm_mask_fcmul_pch): Likewise.
1705 (_mm_maskz_fcmul_pch): Likewise.
1706 (_mm256_fcmul_pch): Likewise.
1707 (_mm256_mask_fcmul_pch): Likewise.
1708 (_mm256_maskz_fcmul_pch): Likewise.
1709 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF_V8HF_V8HF,
1710 V8HF_FTYPE_V16HF_V16HF_V16HF, V16HF_FTYPE_V16HF_V16HF_V16HF_UQI,
1711 V32HF_FTYPE_V32HF_V32HF_V32HF_INT,
1712 V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT): Add new builtin types.
1713 * config/i386/i386-builtin.def: Add new builtins.
1714 * config/i386/i386-expand.c: Handle new builtin types.
1715 * config/i386/subst.md (SUBST_CV): New.
1716 (maskc_name): Ditto.
1717 (maskc_operand3): Ditto.
1719 (sdc_maskz_name): Ditto.
1720 (sdc_mask_op4): Ditto.
1721 (sdc_mask_op5): Ditto.
1722 (sdc_mask_mode512bit_condition): Ditto.
1724 (round_maskc_operand3): Ditto.
1725 (round_sdc_mask_operand4): Ditto.
1726 (round_maskc_op3): Ditto.
1727 (round_sdc_mask_op4): Ditto.
1728 (round_saeonly_sdc_mask_operand5): Ditto.
1729 * config/i386/sse.md (unspec): Add complex fma unspecs.
1730 (avx512fmaskcmode): New.
1731 (UNSPEC_COMPLEX_F_C_MA): Ditto.
1732 (UNSPEC_COMPLEX_F_C_MUL): Ditto.
1733 (complexopname): Ditto.
1734 (<avx512>_fmaddc_<mode>_maskz<round_expand_name>): New expander.
1735 (<avx512>_fcmaddc_<mode>_maskz<round_expand_name>): Ditto.
1736 (fma_<complexopname>_<mode><sdc_maskz_name><round_name>): New
1738 (<avx512>_<complexopname>_<mode>_mask<round_name>): Ditto.
1739 (<avx512>_<complexopname>_<mode><maskc_name><round_name>): Ditto.
1741 2021-09-22 Kewen Lin <linkw@linux.ibm.com>
1743 * config/rs6000/rs6000.opt (rs6000-density-pct-threshold,
1744 rs6000-density-size-threshold, rs6000-density-penalty,
1745 rs6000-density-load-pct-threshold,
1746 rs6000-density-load-num-threshold): New parameter.
1747 * config/rs6000/rs6000.c (rs6000_density_test): Adjust with
1748 corresponding parameters.
1750 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1752 * gimple-range-path.cc (path_range_query::defined_outside_path):
1754 (path_range_query::range_on_path_entry): New.
1755 (path_range_query::internal_range_of_expr): Resolve unknowns
1757 (path_range_query::improve_range_with_equivs): New.
1758 (path_range_query::ssa_range_in_phi): Resolve unknowns with
1760 * gimple-range-path.h (class path_range_query): Add
1761 defined_outside_path, range_on_path_entry, and
1762 improve_range_with_equivs.
1764 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1766 * gimple-range-path.cc (path_range_query::add_to_imports): New.
1767 (path_range_query::add_copies_to_imports): New.
1768 (path_range_query::precompute_ranges): Call
1769 add_copies_to_imports.
1770 * gimple-range-path.h (class path_range_query): Add prototypes
1771 for add_copies_to_imports and add_to_imports.
1773 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1775 * gimple-range-path.cc (path_range_query::range_defined_in_block):
1776 Remove useless code.
1778 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1780 * gimple-range-fold.h (class fur_source): Make oracle protected.
1781 * gimple-range-path.cc (path_range_query::path_range_query): Add
1782 resolve argument. Initialize oracle.
1783 (path_range_query::~path_range_query): Delete oracle.
1784 (path_range_query::range_of_stmt): Adapt to use relations.
1785 (path_range_query::precompute_ranges): Pre-compute relations.
1786 (class jt_fur_source): New
1787 (jt_fur_source::jt_fur_source): New.
1788 (jt_fur_source::register_relation): New.
1789 (jt_fur_source::query_relation): New.
1790 (path_range_query::precompute_relations): New.
1791 (path_range_query::precompute_phi_relations): New.
1792 * gimple-range-path.h (path_range_query): Add resolve argument.
1793 Add oracle, precompute_relations, precompute_phi_relations.
1794 * tree-ssa-threadbackward.c (back_threader::back_threader): Pass
1795 resolve argument to solver.
1797 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1799 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
1800 Rename postfold_gcond_edges to register_outgoing_edges and
1802 (fold_using_range::postfold_gcond_edges): Rename...
1803 (fur_source::register_outgoing_edges): ...to this.
1804 * gimple-range-fold.h (postfold_gcond_edges): Rename to
1805 register_outgoing_edges and move to fur_source.
1807 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1809 * gimple-range-fold.cc (fold_using_range::range_of_phi): Check
1810 dom_info_available_p.
1812 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1814 * gimple-range-cache.cc (non_null_ref::non_null_ref): Use create
1815 and quick_grow_cleared instead of safe_grow_cleared.
1817 2021-09-21 Thomas Schwinge <thomas@codesourcery.com>
1820 * omp-oacc-neuter-broadcast.cc (oacc_do_neutering): Evaluate
1823 2021-09-21 Richard Earnshaw <rearnsha@arm.com>
1825 * configure.ac: Detect when the assembler supports new-style
1826 architecture extensions.
1827 * common/config/arm/arm-common.c (arm_rewrite_mcpu): Return
1828 the full CPU string if the assembler can grok it.
1829 (arm_rewrite_march): Likewise but for the architecture.
1830 * config.in: Regenerate.
1831 * configure: Regenerate.
1833 2021-09-21 Richard Biener <rguenther@suse.de>
1835 PR tree-optimization/102421
1836 * tree-vect-loop.c (vect_dissolve_slp_only_groups): Copy and
1837 adjust alignment info.
1839 2021-09-21 Kewen Lin <linkw@linux.ibm.com>
1841 * ipa-fnsummary.c (ipa_fn_summary_write): Remove inconsistent
1842 bitfield stream out.
1844 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
1846 * gimple-range-fold.cc (fold_using_range::range_of_phi): Ignore
1847 undefined edges, apply an equivalence if appropriate.
1848 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Return
1849 UNDEFINED if EDGE_EXECUTABLE is not set.
1850 * gimple-range.cc (gimple_ranger::gimple_ranger): Set all edges
1851 as EXECUTABLE upon startup.
1852 (gimple_ranger::range_on_edge): Return UNDEFINED for edges without
1853 EDGE_EXECUTABLE set.
1854 * vr-values.c (set_and_propagate_unexecutable): New.
1855 (simplify_using_ranges::fold_cond): Call set_and_propagate.
1856 (simplify_using_ranges::simplify_switch_using_ranges): Ditto.
1857 * vr-values.h: Add prototype.
1859 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
1861 * value-relation.cc (equiv_oracle::register_initial_def): New.
1862 (equiv_oracle::register_relation): Call register_initial_def.
1863 (equiv_oracle::add_equiv_to_block): New. Split register_relation.
1864 (relation_oracle::register_stmt): Check def block of PHI arguments.
1865 * value-relation.h (equiv_oracle): Add new prototypes.
1867 2021-09-20 Matthias Kretz <m.kretz@gsi.de>
1869 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
1870 Define __RECIPROCAL_MATH__, __NO_SIGNED_ZEROS__,
1871 __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__, and
1872 __ROUNDING_MATH__ according to their corresponding flags.
1873 * doc/cpp.texi: Document __RECIPROCAL_MATH__,
1874 __NO_SIGNED_ZEROS__, __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__,
1875 and __ROUNDING_MATH__.
1877 2021-09-20 Richard Biener <rguenther@suse.de>
1879 * tree-vect-stmts.c (vectorizable_load): Use the vectype
1882 2021-09-20 Richard Biener <rguenther@suse.de>
1884 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
1885 Do not compute alignment of the vectorized access here.
1887 2021-09-20 Richard Biener <rguenther@suse.de>
1889 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
1890 Store -1 for runtime alias peeling iterations.
1892 2021-09-20 Richard Biener <rguenther@suse.de>
1894 * config.gcc: Obsolete hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11*.
1896 2021-09-20 Thomas Schwinge <thomas@codesourcery.com>
1898 * input.c (string_concat_db::record_string_concatenation)
1899 (string_concat_db::get_string_concatenation): Skip for
1900 'RESERVED_LOCATION_P'.
1902 2021-09-20 Richard Biener <rguenther@suse.de>
1904 PR tree-optimization/65206
1905 * tree-data-ref.h (struct data_reference): Add alt_indices,
1907 * tree-data-ref.c (free_data_ref): Release alt_indices.
1908 (dr_analyze_indices): Work on struct indices and get DR_REF as tree.
1909 (create_data_ref): Adjust.
1910 (initialize_data_dependence_relation): Split into head
1911 and tail. When the base objects fail to match up try
1912 again with pointer-based analysis of indices.
1913 * tree-vectorizer.c (vec_info_shared::check_datarefs): Do
1914 not compare the lazily computed alternate set of indices.
1916 2021-09-20 Iain Sandoe <iain@sandoe.co.uk>
1918 * gcc.c: Test for execute OK when we find the
1919 programs for assembler linker and dsymutil and those
1920 were specified at configure-time.
1922 2021-09-19 Martin Sebor <msebor@redhat.com>
1924 PR middle-end/102403
1925 * gimple-predicate-analysis.cc (predicate::init_from_control_deps):
1926 Correct a function pre/postcondition.
1928 2021-09-19 Martin Sebor <msebor@redhat.com>
1930 PR middle-end/102243
1931 * tree-ssa-strlen.c (get_range): Handle null cfun.
1933 2021-09-19 Iain Sandoe <iain@sandoe.co.uk>
1935 * config/darwin.h (LINK_COMMAND_SPEC_A): Use Darwin10
1936 unwinder shim as a convenience library.
1938 2021-09-19 Andrew Pinski <apinski@marvell.com>
1940 * doc/install.texi: Add note about
1941 binutils 2.35 is required for LTO usage.
1943 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
1945 * tree-ssa-threadbackward.c
1946 (back_threader_registry::register_path): Use push_edge.
1947 * tree-ssa-threadedge.c
1948 (jump_threader::thread_around_empty_blocks): Same.
1949 (jump_threader::thread_through_normal_block): Same.
1950 (jump_threader::thread_across_edge): Same. Also, use auto_bitmap.
1952 * tree-ssa-threadupdate.c
1953 (jt_path_registry::allocate_thread_edge): Remove.
1954 (jt_path_registry::push_edge): New.
1955 (dump_jump_thread_path): Make static.
1956 * tree-ssa-threadupdate.h (allocate_thread_edge): Remove.
1959 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
1961 * gimple-range-path.cc (path_range_query::path_range_query): Add
1963 (path_range_query::dump): Remove extern declaration of dump_ranger.
1964 * gimple-range-trace.cc (dump_ranger): Add DEBUG_FUNCTION marker.
1965 * gimple-range-trace.h (dump_ranger): Add prototype.
1967 2021-09-19 John Ericson <git@JohnEricson.me>
1969 * gcc.c (find_a_program): New function, factored out of...
1970 (find_a_file): Here.
1971 (execute): Use find_a_program when looking for programs rather
1974 2021-09-19 Matwey V. Kornilov <matwey.kornilov@gmail.com>
1976 * config/avr/avr-mcus.def: Add atmega324pb.
1977 * doc/avr-mmcu.texi: Corresponding changes.
1979 2021-09-19 Roger Sayle <roger@nextmovesoftware.com>
1982 * match.pd (cmp @0 REAL_CST@1): When @0 is also REAL_CST, apply
1983 the same transformations as to @1. For comparisons against NaN,
1984 don't check HONOR_SNANS but confirm that neither operand is a
1987 2021-09-19 Benjamin Peterson <benjamin@locrian.net>
1989 * attribs.c (make_unique_name): Delete.
1990 * attribs.h (make_unique_name): Delete.
1992 2021-09-19 Andrew Pinski <apinski@marvell.com>
1994 * lra-constraints.c (check_and_process_move): Assert
1995 that dclass and sclass are greater than or equal to NO_REGS.
1997 2021-09-18 Jakub Jelinek <jakub@redhat.com>
1999 * tree.h (OMP_CLAUSE_ORDER_UNCONSTRAINED): Define.
2000 * tree-pretty-print.c (dump_omp_clause): Print unconstrained:
2001 for OMP_CLAUSE_ORDER_UNCONSTRAINED.
2003 2021-09-18 liuhongt <hongtao.liu@intel.com>
2005 * config/i386/i386-features.c (remove_partial_avx_dependency):
2006 Restrict TARGET_USE_VECTOR_FP_CONVERTS and
2007 TARGET_USE_VECTOR_CONVERTS to conversion instructions only.
2009 2021-09-18 Jakub Jelinek <jakub@redhat.com>
2011 * gimplify.c (omp_default_clause): For C/C++ default({,first}private),
2012 if file/namespace scope variable doesn't have predetermined sharing,
2013 treat it as if there was default(none).
2015 2021-09-18 liuhongt <hongtao.liu@intel.com>
2017 * config/i386/avx512fp16intrin.h (_mm_fmadd_sh):
2019 (_mm_mask_fmadd_sh): Likewise.
2020 (_mm_mask3_fmadd_sh): Likewise.
2021 (_mm_maskz_fmadd_sh): Likewise.
2022 (_mm_fmadd_round_sh): Likewise.
2023 (_mm_mask_fmadd_round_sh): Likewise.
2024 (_mm_mask3_fmadd_round_sh): Likewise.
2025 (_mm_maskz_fmadd_round_sh): Likewise.
2026 (_mm_fnmadd_sh): Likewise.
2027 (_mm_mask_fnmadd_sh): Likewise.
2028 (_mm_mask3_fnmadd_sh): Likewise.
2029 (_mm_maskz_fnmadd_sh): Likewise.
2030 (_mm_fnmadd_round_sh): Likewise.
2031 (_mm_mask_fnmadd_round_sh): Likewise.
2032 (_mm_mask3_fnmadd_round_sh): Likewise.
2033 (_mm_maskz_fnmadd_round_sh): Likewise.
2034 (_mm_fmsub_sh): Likewise.
2035 (_mm_mask_fmsub_sh): Likewise.
2036 (_mm_mask3_fmsub_sh): Likewise.
2037 (_mm_maskz_fmsub_sh): Likewise.
2038 (_mm_fmsub_round_sh): Likewise.
2039 (_mm_mask_fmsub_round_sh): Likewise.
2040 (_mm_mask3_fmsub_round_sh): Likewise.
2041 (_mm_maskz_fmsub_round_sh): Likewise.
2042 (_mm_fnmsub_sh): Likewise.
2043 (_mm_mask_fnmsub_sh): Likewise.
2044 (_mm_mask3_fnmsub_sh): Likewise.
2045 (_mm_maskz_fnmsub_sh): Likewise.
2046 (_mm_fnmsub_round_sh): Likewise.
2047 (_mm_mask_fnmsub_round_sh): Likewise.
2048 (_mm_mask3_fnmsub_round_sh): Likewise.
2049 (_mm_maskz_fnmsub_round_sh): Likewise.
2050 * config/i386/i386-builtin-types.def
2051 (V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT): New builtin type.
2052 * config/i386/i386-builtin.def: Add new builtins.
2053 * config/i386/i386-expand.c: Handle new builtin type.
2054 * config/i386/sse.md (fmai_vmfmadd_<mode><round_name>):
2055 Ajdust to support FP16.
2056 (fmai_vmfmsub_<mode><round_name>): Ditto.
2057 (fmai_vmfnmadd_<mode><round_name>): Ditto.
2058 (fmai_vmfnmsub_<mode><round_name>): Ditto.
2059 (*fmai_fmadd_<mode>): Ditto.
2060 (*fmai_fmsub_<mode>): Ditto.
2061 (*fmai_fnmadd_<mode><round_name>): Ditto.
2062 (*fmai_fnmsub_<mode><round_name>): Ditto.
2063 (avx512f_vmfmadd_<mode>_mask<round_name>): Ditto.
2064 (avx512f_vmfmadd_<mode>_mask3<round_name>): Ditto.
2065 (avx512f_vmfmadd_<mode>_maskz<round_expand_name>): Ditto.
2066 (avx512f_vmfmadd_<mode>_maskz_1<round_name>): Ditto.
2067 (*avx512f_vmfmsub_<mode>_mask<round_name>): Ditto.
2068 (avx512f_vmfmsub_<mode>_mask3<round_name>): Ditto.
2069 (*avx512f_vmfmsub_<mode>_maskz_1<round_name>): Ditto.
2070 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
2071 (*avx512f_vmfnmsub_<mode>_mask3<round_name>): Ditto.
2072 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
2073 (*avx512f_vmfnmadd_<mode>_mask<round_name>): Renamed to ...
2074 (avx512f_vmfnmadd_<mode>_mask<round_name>) ... this, and
2075 adjust to support FP16.
2076 (avx512f_vmfnmadd_<mode>_mask3<round_name>): Ditto.
2077 (avx512f_vmfnmadd_<mode>_maskz_1<round_name>): Ditto.
2078 (avx512f_vmfnmadd_<mode>_maskz<round_expand_name>): New
2081 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
2083 * config/i386/sse.md (avx512fmaskmodelower): Extend to support
2085 (maskload<mode><avx512fmaskmodelower>): Ditto.
2086 (maskstore<mode><avx512fmaskmodelower>): Ditto.
2088 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
2090 * config/i386/i386-expand.c (ix86_expand_fp_absneg_operator):
2092 (ix86_expand_copysign): Ditto.
2093 (ix86_expand_xorsign): Ditto.
2094 * config/i386/i386.c (ix86_build_const_vector): Handle HF vector
2096 (ix86_build_signbit_mask): Ditto.
2097 (ix86_can_change_mode_class): Ditto.
2098 * config/i386/i386.md
2099 (SSEMODEF): Add HFmode.
2100 (ssevecmodef): Ditto.
2101 (<code>hf2): New define_expand.
2102 (*<code>hf2_1): New define_insn_and_split.
2103 (copysign<mode>): Extend to support HFmode under AVX512FP16.
2104 (xorsign<mode>): Ditto.
2105 * config/i386/sse.md (VFB): New mode iterator.
2106 (VFB_128_256): Ditto.
2108 (sseintvecmode2): Support HF vector mode.
2109 (<code><mode>2): Use new mode iterator.
2110 (*<code><mode>2): Ditto.
2111 (copysign<mode>3): Ditto.
2112 (xorsign<mode>3): Ditto.
2113 (<code><mode>3<mask_name>): Ditto.
2114 (<code><mode>3<mask_name>): Ditto.
2115 (<sse>_andnot<mode>3<mask_name>): Adjust for HF vector mode.
2116 (<sse>_andnot<mode>3<mask_name>): Ditto.
2117 (*<code><mode>3<mask_name>): Ditto.
2118 (*<code><mode>3<mask_name>): Ditto.
2120 2021-09-18 liuhongt <hongtao.liu@intel.com>
2122 * config/i386/avx512fp16intrin.h (_mm512_mask_fmadd_ph):
2124 (_mm512_mask3_fmadd_ph): Likewise.
2125 (_mm512_maskz_fmadd_ph): Likewise.
2126 (_mm512_fmadd_round_ph): Likewise.
2127 (_mm512_mask_fmadd_round_ph): Likewise.
2128 (_mm512_mask3_fmadd_round_ph): Likewise.
2129 (_mm512_maskz_fmadd_round_ph): Likewise.
2130 (_mm512_fnmadd_ph): Likewise.
2131 (_mm512_mask_fnmadd_ph): Likewise.
2132 (_mm512_mask3_fnmadd_ph): Likewise.
2133 (_mm512_maskz_fnmadd_ph): Likewise.
2134 (_mm512_fnmadd_round_ph): Likewise.
2135 (_mm512_mask_fnmadd_round_ph): Likewise.
2136 (_mm512_mask3_fnmadd_round_ph): Likewise.
2137 (_mm512_maskz_fnmadd_round_ph): Likewise.
2138 (_mm512_fmsub_ph): Likewise.
2139 (_mm512_mask_fmsub_ph): Likewise.
2140 (_mm512_mask3_fmsub_ph): Likewise.
2141 (_mm512_maskz_fmsub_ph): Likewise.
2142 (_mm512_fmsub_round_ph): Likewise.
2143 (_mm512_mask_fmsub_round_ph): Likewise.
2144 (_mm512_mask3_fmsub_round_ph): Likewise.
2145 (_mm512_maskz_fmsub_round_ph): Likewise.
2146 (_mm512_fnmsub_ph): Likewise.
2147 (_mm512_mask_fnmsub_ph): Likewise.
2148 (_mm512_mask3_fnmsub_ph): Likewise.
2149 (_mm512_maskz_fnmsub_ph): Likewise.
2150 (_mm512_fnmsub_round_ph): Likewise.
2151 (_mm512_mask_fnmsub_round_ph): Likewise.
2152 (_mm512_mask3_fnmsub_round_ph): Likewise.
2153 (_mm512_maskz_fnmsub_round_ph): Likewise.
2154 * config/i386/avx512fp16vlintrin.h (_mm256_fmadd_ph):
2156 (_mm256_mask_fmadd_ph): Likewise.
2157 (_mm256_mask3_fmadd_ph): Likewise.
2158 (_mm256_maskz_fmadd_ph): Likewise.
2159 (_mm_fmadd_ph): Likewise.
2160 (_mm_mask_fmadd_ph): Likewise.
2161 (_mm_mask3_fmadd_ph): Likewise.
2162 (_mm_maskz_fmadd_ph): Likewise.
2163 (_mm256_fnmadd_ph): Likewise.
2164 (_mm256_mask_fnmadd_ph): Likewise.
2165 (_mm256_mask3_fnmadd_ph): Likewise.
2166 (_mm256_maskz_fnmadd_ph): Likewise.
2167 (_mm_fnmadd_ph): Likewise.
2168 (_mm_mask_fnmadd_ph): Likewise.
2169 (_mm_mask3_fnmadd_ph): Likewise.
2170 (_mm_maskz_fnmadd_ph): Likewise.
2171 (_mm256_fmsub_ph): Likewise.
2172 (_mm256_mask_fmsub_ph): Likewise.
2173 (_mm256_mask3_fmsub_ph): Likewise.
2174 (_mm256_maskz_fmsub_ph): Likewise.
2175 (_mm_fmsub_ph): Likewise.
2176 (_mm_mask_fmsub_ph): Likewise.
2177 (_mm_mask3_fmsub_ph): Likewise.
2178 (_mm_maskz_fmsub_ph): Likewise.
2179 (_mm256_fnmsub_ph): Likewise.
2180 (_mm256_mask_fnmsub_ph): Likewise.
2181 (_mm256_mask3_fnmsub_ph): Likewise.
2182 (_mm256_maskz_fnmsub_ph): Likewise.
2183 (_mm_fnmsub_ph): Likewise.
2184 (_mm_mask_fnmsub_ph): Likewise.
2185 (_mm_mask3_fnmsub_ph): Likewise.
2186 (_mm_maskz_fnmsub_ph): Likewise.
2187 * config/i386/i386-builtin.def: Add corresponding new builtins.
2188 * config/i386/sse.md
2189 (<avx512>_fmadd_<mode>_maskz<round_expand_name>): Adjust to
2190 support HF vector modes.
2191 (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
2193 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
2194 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
2195 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
2196 (<avx512>_fmadd_<mode>_mask<round_name>): Ditto.
2197 (<avx512>_fmadd_<mode>_mask3<round_name>): Ditto.
2198 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Ditto.
2199 (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>):
2201 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
2202 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
2203 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
2204 (<avx512>_fmsub_<mode>_mask<round_name>): Ditto.
2205 (<avx512>_fmsub_<mode>_mask3<round_name>): Ditto.
2206 (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>):
2208 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
2209 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
2210 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
2211 (<avx512>_fnmadd_<mode>_mask<round_name>): Ditto.
2212 (<avx512>_fnmadd_<mode>_mask3<round_name>): Ditto.
2213 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Ditto.
2214 (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>):
2216 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
2217 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
2218 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
2219 (<avx512>_fnmsub_<mode>_mask<round_name>): Ditto.
2220 (<avx512>_fnmsub_<mode>_mask3<round_name>): Ditto.
2222 2021-09-18 liuhongt <hongtao.liu@intel.com>
2224 * config/i386/avx512fp16intrin.h (_mm512_fmaddsub_ph):
2226 (_mm512_mask_fmaddsub_ph): Likewise.
2227 (_mm512_mask3_fmaddsub_ph): Likewise.
2228 (_mm512_maskz_fmaddsub_ph): Likewise.
2229 (_mm512_fmaddsub_round_ph): Likewise.
2230 (_mm512_mask_fmaddsub_round_ph): Likewise.
2231 (_mm512_mask3_fmaddsub_round_ph): Likewise.
2232 (_mm512_maskz_fmaddsub_round_ph): Likewise.
2233 (_mm512_mask_fmsubadd_ph): Likewise.
2234 (_mm512_mask3_fmsubadd_ph): Likewise.
2235 (_mm512_maskz_fmsubadd_ph): Likewise.
2236 (_mm512_fmsubadd_round_ph): Likewise.
2237 (_mm512_mask_fmsubadd_round_ph): Likewise.
2238 (_mm512_mask3_fmsubadd_round_ph): Likewise.
2239 (_mm512_maskz_fmsubadd_round_ph): Likewise.
2240 * config/i386/avx512fp16vlintrin.h (_mm256_fmaddsub_ph):
2242 (_mm256_mask_fmaddsub_ph): Likewise.
2243 (_mm256_mask3_fmaddsub_ph): Likewise.
2244 (_mm256_maskz_fmaddsub_ph): Likewise.
2245 (_mm_fmaddsub_ph): Likewise.
2246 (_mm_mask_fmaddsub_ph): Likewise.
2247 (_mm_mask3_fmaddsub_ph): Likewise.
2248 (_mm_maskz_fmaddsub_ph): Likewise.
2249 (_mm256_fmsubadd_ph): Likewise.
2250 (_mm256_mask_fmsubadd_ph): Likewise.
2251 (_mm256_mask3_fmsubadd_ph): Likewise.
2252 (_mm256_maskz_fmsubadd_ph): Likewise.
2253 (_mm_fmsubadd_ph): Likewise.
2254 (_mm_mask_fmsubadd_ph): Likewise.
2255 (_mm_mask3_fmsubadd_ph): Likewise.
2256 (_mm_maskz_fmsubadd_ph): Likewise.
2257 * config/i386/i386-builtin.def: Add corresponding new builtins.
2258 * config/i386/sse.md (VFH_SF_AVX512VL): New mode iterator.
2259 * (<avx512>_fmsubadd_<mode>_maskz<round_expand_name>): New expander.
2260 * (<avx512>_fmaddsub_<mode>_maskz<round_expand_name>): Use
2262 * (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
2264 * (<avx512>_fmaddsub_<mode>_mask<round_name>): Ditto.
2265 * (<avx512>_fmaddsub_<mode>_mask3<round_name>): Ditto.
2266 * (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
2268 * (<avx512>_fmsubadd_<mode>_mask<round_name>): Ditto.
2269 * (<avx512>_fmsubadd_<mode>_mask3<round_name>): Ditto.
2271 2021-09-18 liuhongt <hongtao.liu@intel.com>
2274 * config/i386/i386.c (ix86_print_operand): Handle
2275 V8HF/V16HF/V32HFmode.
2276 * config/i386/i386.h (VALID_BCST_MODE_P): Add HFmode.
2277 * config/i386/sse.md (avx512bcst): Remove.
2279 2021-09-17 Martin Sebor <msebor@redhat.com>
2281 * Makefile.in (OBJS): Add gimple-predicate-analysis.o.
2282 * tree-ssa-uninit.c (max_phi_args): Move to gimple-predicate-analysis.
2283 (MASK_SET_BIT, MASK_TEST_BIT, MASK_EMPTY): Same.
2284 (check_defs): Add comment.
2285 (can_skip_redundant_opnd): Update comment.
2286 (compute_uninit_opnds_pos): Adjust to namespace change.
2287 (find_pdom): Move to gimple-predicate-analysis.cc.
2289 (struct uninit_undef_val_t): New.
2290 (is_non_loop_exit_postdominating): Move to gimple-predicate-analysis.cc.
2291 (find_control_equiv_block): Same.
2292 (MAX_NUM_CHAINS, MAX_CHAIN_LEN, MAX_POSTDOM_CHECK): Same.
2293 (MAX_SWITCH_CASES): Same.
2294 (compute_control_dep_chain): Same.
2295 (find_uninit_use): Use predicate analyzer.
2296 (struct pred_info): Move to gimple-predicate-analysis.
2297 (convert_control_dep_chain_into_preds): Same.
2298 (find_predicates): Same.
2299 (collect_phi_def_edges): Same.
2300 (warn_uninitialized_phi): Use predicate analyzer.
2301 (find_def_preds): Move to gimple-predicate-analysis.
2302 (dump_pred_info): Same.
2303 (dump_pred_chain): Same.
2304 (dump_predicates): Same.
2305 (destroy_predicate_vecs): Remove.
2306 (execute_late_warn_uninitialized): New.
2307 (get_cmp_code): Move to gimple-predicate-analysis.
2308 (is_value_included_in): Same.
2309 (value_sat_pred_p): Same.
2310 (find_matching_predicate_in_rest_chains): Same.
2311 (is_use_properly_guarded): Same.
2312 (prune_uninit_phi_opnds): Same.
2313 (find_var_cmp_const): Same.
2314 (use_pred_not_overlap_with_undef_path_pred): Same.
2315 (pred_equal_p): Same.
2316 (is_neq_relop_p): Same.
2317 (is_neq_zero_form_p): Same.
2318 (pred_expr_equal_p): Same.
2319 (is_pred_expr_subset_of): Same.
2320 (is_pred_chain_subset_of): Same.
2321 (is_included_in): Same.
2322 (is_superset_of): Same.
2324 (simplify_pred): Same.
2325 (simplify_preds_2): Same.
2326 (simplify_preds_3): Same.
2327 (simplify_preds_4): Same.
2328 (simplify_preds): Same.
2330 (push_to_worklist): Same.
2331 (get_pred_info_from_cmp): Same.
2332 (is_degenerated_phi): Same.
2333 (normalize_one_pred_1): Same.
2334 (normalize_one_pred): Same.
2335 (normalize_one_pred_chain): Same.
2336 (normalize_preds): Same.
2337 (can_one_predicate_be_invalidated_p): Same.
2338 (can_chain_union_be_invalidated_p): Same.
2339 (uninit_uses_cannot_happen): Same.
2340 (pass_late_warn_uninitialized::execute): Define.
2341 * gimple-predicate-analysis.cc: New file.
2342 * gimple-predicate-analysis.h: New file.
2344 2021-09-17 Julian Brown <julian@codesourcery.com>
2346 * config/gcn/gcn.c (gimple.h): Include.
2347 (gcn_fork_join): Emit barrier for worker-level joins.
2348 * omp-oacc-neuter-broadcast.cc (find_local_vars_to_propagate): Add
2349 writes_gang_private bitmap parameter. Set bit for blocks
2350 containing gang-private variable writes.
2351 (worker_single_simple): Don't emit barrier after predicated block.
2352 (worker_single_copy): Don't emit barrier if we're not broadcasting
2353 anything and the block contains no gang-private writes.
2354 (neuter_worker_single): Don't predicate blocks that only contain
2355 NOPs or internal marker functions. Pass has_gang_private_write
2356 argument to worker_single_copy.
2357 (oacc_do_neutering): Add writes_gang_private bitmap handling.
2359 2021-09-17 Julian Brown <julian@codesourcery.com>
2361 * config/gcn/gcn-protos.h
2362 (gcn_goacc_create_worker_broadcast_record): Update prototype.
2363 * config/gcn/gcn-tree.c (gcn_goacc_get_worker_red_decl): Use
2364 preallocated block of LDS memory. Do not cache/share decls for
2365 reduction temporaries between invocations.
2366 (gcn_goacc_reduction_teardown): Unshare VAR on second use.
2367 (gcn_goacc_create_worker_broadcast_record): Add OFFSET parameter
2368 and return temporary LDS space at that offset. Return pointer in
2370 * config/gcn/gcn.c (acc_lds_size, gang_private_hwm, lds_allocs):
2372 (ACC_LDS_SIZE): Define as acc_lds_size.
2373 (gcn_init_machine_status): Don't initialise lds_allocated,
2374 lds_allocs, reduc_decls fields of machine function struct.
2375 (gcn_option_override): Handle default size for gang-private
2376 variables and -mgang-private-size option.
2377 (gcn_expand_prologue): Use LDS_SIZE instead of LDS_SIZE-1 when
2378 initialising M0_REG.
2379 (gcn_shared_mem_layout): New function.
2380 (gcn_print_lds_decl): Update comment. Use global lds_allocs map and
2381 gang_private_hwm variable.
2382 (TARGET_GOACC_SHARED_MEM_LAYOUT): Define target hook.
2383 * config/gcn/gcn.h (machine_function): Remove lds_allocated,
2384 lds_allocs, reduc_decls. Add reduction_base, reduction_limit.
2385 * config/gcn/gcn.opt (gang_private_size_opt): New global.
2386 (mgang-private-size=): New option.
2387 * doc/tm.texi.in (TARGET_GOACC_SHARED_MEM_LAYOUT): Place
2389 * doc/tm.texi: Regenerate.
2390 * omp-oacc-neuter-broadcast.cc (targhooks.h, diagnostic-core.h):
2392 (build_sender_ref): Handle sender_decl being pointer.
2393 (worker_single_copy): Add PLACEMENT and ISOLATE_BROADCASTS
2394 parameters. Pass placement argument to
2395 create_worker_broadcast_record hook invocations. Handle
2396 sender_decl being pointer and isolate_broadcasts inserting extra
2398 (blk_offset_map_t): Add typedef.
2399 (neuter_worker_single): Add BLK_OFFSET_MAP parameter. Pass
2400 preallocated range to worker_single_copy call.
2401 (dfs_broadcast_reachable_1): New function.
2402 (idx_decl_pair_t, used_range_vec_t): New typedefs.
2403 (sort_size_descending): New function.
2404 (addr_range): New class.
2405 (splay_tree_compare_addr_range, splay_tree_free_key)
2406 (first_fit_range, merge_ranges_1, merge_ranges): New functions.
2407 (execute_omp_oacc_neuter_broadcast): Rename to...
2408 (oacc_do_neutering): ... this. Add BOUNDS_LO, BOUNDS_HI
2409 parameters. Arrange layout of shared memory for broadcast
2411 (execute_omp_oacc_neuter_broadcast): New function.
2412 (pass_omp_oacc_neuter_broadcast::gate): Remove num_workers==1
2413 handling from here. Enable pass for all OpenACC routines in order
2414 to call shared memory-layout hook.
2415 * target.def (create_worker_broadcast_record): Add OFFSET
2417 (shared_mem_layout): New hook.
2419 2021-09-17 Julian Brown <julian@codesourcery.com>
2420 Thomas Schwinge <thomas@codesourcery.com>
2422 * omp-oacc-neuter-broadcast.cc
2423 (pass_omp_oacc_neuter_broadcast::gate): Disable if num_workers is
2425 (execute_omp_oacc_neuter_broadcast): Adjust.
2427 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
2429 * value-relation.cc (class equiv_chain): Move to header file.
2430 (path_oracle::path_oracle): New.
2431 (path_oracle::~path_oracle): New.
2432 (path_oracle::register_relation): New.
2433 (path_oracle::query_relation): New.
2434 (path_oracle::reset_path): New.
2435 (path_oracle::dump): New.
2436 * value-relation.h (class equiv_chain): Move to here.
2437 (class path_oracle): New.
2439 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
2441 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a DOM
2443 * gimple-range-fold.cc (fur_depend::register_relation): Use
2444 register_stmt/edge routines.
2445 * value-relation.cc (equiv_chain::find): Relocate from equiv_oracle.
2446 (equiv_oracle::equiv_oracle): Create self equivalence cache.
2447 (equiv_oracle::~equiv_oracle): Release same.
2448 (equiv_oracle::equiv_set): Return entry from self equiv cache if there
2449 are no equivalences.
2450 (equiv_oracle::find_equiv_block): Move list find to equiv_chain.
2451 (equiv_oracle::register_relation): Rename from register_equiv.
2452 (relation_chain_head::find_relation): Relocate from dom_oracle.
2453 (relation_oracle::register_stmt): New.
2454 (relation_oracle::register_edge): New.
2455 (dom_oracle::*): Rename from relation_oracle.
2456 (dom_oracle::register_relation): Adjust to call equiv_oracle.
2457 (dom_oracle::set_one_relation): Split from register_relation.
2458 (dom_oracle::register_transitives): Consolidate 2 methods.
2459 (dom_oracle::find_relation_block): Move core to relation_chain.
2460 (dom_oracle::query_relation): Rename from find_relation_dom and adjust.
2461 * value-relation.h (class relation_oracle): New pure virtual base.
2462 (class equiv_oracle): Inherit from relation_oracle and adjust.
2463 (class dom_oracle): Rename from old relation_oracle and adjust.
2465 2021-09-17 Martin Sebor <msebor@redhat.com>
2467 PR middle-end/102200
2468 * pointer-query.cc (access_ref::inform_access): Handle MIN/MAX_EXPR.
2469 (handle_min_max_size): Change argument. Store original SSA_NAME for
2470 operands to potentially distinct (sub)objects.
2471 (compute_objsize_r): Adjust call to the above.
2473 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
2475 * config/rs6000/rs6000.c (rs6000-builtins.h): New include.
2476 (rs6000_new_builtin_vectorized_function): New function.
2477 (rs6000_new_builtin_md_vectorized_function): Likewise.
2478 (rs6000_builtin_vectorized_function): Call
2479 rs6000_new_builtin_vectorized_function.
2480 (rs6000_builtin_md_vectorized_function): Call
2481 rs6000_new_builtin_md_vectorized_function.
2483 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
2485 * config/rs6000/rs6000-builtin-new.def (ASSEMBLE_ACC): Add mmaint flag.
2486 (ASSEMBLE_PAIR): Likewise.
2487 (BUILD_ACC): Likewise.
2488 (DISASSEMBLE_ACC): Likewise.
2489 (DISASSEMBLE_PAIR): Likewise.
2490 (PMXVBF16GER2): Likewise.
2491 (PMXVBF16GER2NN): Likewise.
2492 (PMXVBF16GER2NP): Likewise.
2493 (PMXVBF16GER2PN): Likewise.
2494 (PMXVBF16GER2PP): Likewise.
2495 (PMXVF16GER2): Likewise.
2496 (PMXVF16GER2NN): Likewise.
2497 (PMXVF16GER2NP): Likewise.
2498 (PMXVF16GER2PN): Likewise.
2499 (PMXVF16GER2PP): Likewise.
2500 (PMXVF32GER): Likewise.
2501 (PMXVF32GERNN): Likewise.
2502 (PMXVF32GERNP): Likewise.
2503 (PMXVF32GERPN): Likewise.
2504 (PMXVF32GERPP): Likewise.
2505 (PMXVF64GER): Likewise.
2506 (PMXVF64GERNN): Likewise.
2507 (PMXVF64GERNP): Likewise.
2508 (PMXVF64GERPN): Likewise.
2509 (PMXVF64GERPP): Likewise.
2510 (PMXVI16GER2): Likewise.
2511 (PMXVI16GER2PP): Likewise.
2512 (PMXVI16GER2S): Likewise.
2513 (PMXVI16GER2SPP): Likewise.
2514 (PMXVI4GER8): Likewise.
2515 (PMXVI4GER8PP): Likewise.
2516 (PMXVI8GER4): Likewise.
2517 (PMXVI8GER4PP): Likewise.
2518 (PMXVI8GER4SPP): Likewise.
2519 (XVBF16GER2): Likewise.
2520 (XVBF16GER2NN): Likewise.
2521 (XVBF16GER2NP): Likewise.
2522 (XVBF16GER2PN): Likewise.
2523 (XVBF16GER2PP): Likewise.
2524 (XVF16GER2): Likewise.
2525 (XVF16GER2NN): Likewise.
2526 (XVF16GER2NP): Likewise.
2527 (XVF16GER2PN): Likewise.
2528 (XVF16GER2PP): Likewise.
2529 (XVF32GER): Likewise.
2530 (XVF32GERNN): Likewise.
2531 (XVF32GERNP): Likewise.
2532 (XVF32GERPN): Likewise.
2533 (XVF32GERPP): Likewise.
2534 (XVF64GER): Likewise.
2535 (XVF64GERNN): Likewise.
2536 (XVF64GERNP): Likewise.
2537 (XVF64GERPN): Likewise.
2538 (XVF64GERPP): Likewise.
2539 (XVI16GER2): Likewise.
2540 (XVI16GER2PP): Likewise.
2541 (XVI16GER2S): Likewise.
2542 (XVI16GER2SPP): Likewise.
2543 (XVI4GER8): Likewise.
2544 (XVI4GER8PP): Likewise.
2545 (XVI8GER4): Likewise.
2546 (XVI8GER4PP): Likewise.
2547 (XVI8GER4SPP): Likewise.
2548 (XXMFACC): Likewise.
2549 (XXMTACC): Likewise.
2550 (XXSETACCZ): Likewise.
2551 (ASSEMBLE_PAIR_V): Likewise.
2552 (BUILD_PAIR): Likewise.
2553 (DISASSEMBLE_PAIR_V): Likewise.
2556 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_mma_builtin):
2557 Handle RS6000_BIF_LXVP and RS6000_BIF_STXVP.
2558 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add ismmaint.
2559 (parse_bif_attrs): Handle ismmaint.
2560 (write_decls): Add bif_mmaint_bit and bif_is_mmaint.
2561 (write_bif_static_init): Handle ismmaint.
2563 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
2565 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin): New
2567 (rs6000_gimple_fold_builtin): Call rs6000_gimple_fold_new_builtin.
2568 (rs6000_new_builtin_valid_without_lhs): New function.
2569 (rs6000_gimple_fold_new_mma_builtin): Likewise.
2570 (rs6000_gimple_fold_new_builtin): Likewise.
2572 2021-09-17 Thomas Schwinge <thomas@codesourcery.com>
2574 * hash-table.h (hash_table<Descriptor, Lazy, Allocator>::expand):
2575 Destruct stale Value objects.
2576 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
2579 2021-09-17 Roger Sayle <roger@nextmovesoftware.com>
2582 * match.pd (shift optimizations): Disable recent sign-changing
2583 optimization for shifts by zero, these will be folded later.
2585 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
2587 * config/rs6000/rs6000-builtin-new.def (__builtin_mffsl): Move from
2588 [power9] to [always].
2590 2021-09-17 Richard Biener <rguenther@suse.de>
2592 * tree-vect-stmts.c (vectorizable_load): Do not frob
2595 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
2597 * config/i386/i386-features.c (remove_partial_avx_dependency):
2598 Also check TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY and
2599 and TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY before generating
2601 * config/i386/i386.h (TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY):
2603 (TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
2604 * config/i386/i386.md (SSE FP to FP splitters): Replace
2605 TARGET_SSE_PARTIAL_REG_DEPENDENCY with
2606 TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY.
2607 (SSE INT to FP splitter): Replace TARGET_SSE_PARTIAL_REG_DEPENDENCY
2608 with TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY.
2609 * config/i386/x86-tune.def
2610 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): New.
2611 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
2613 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
2616 * config/i386/i386-features.c (remove_partial_avx_dependency):
2617 Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
2618 before generating vxorps.
2620 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
2622 * config/i386/i386-options.c (processor_cost_table): Use
2623 tremont_cost for Tremont.
2624 * config/i386/x86-tune-costs.h (tremont_memcpy): New.
2625 (tremont_memset): Likewise.
2626 (tremont_cost): Likewise.
2627 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
2630 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
2632 * common/config/i386/i386-common.c: Use Haswell scheduling model
2634 * config/i386/i386.c (ix86_sched_init_global): Prepare for Tremont
2636 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Tremont
2638 (ix86_adjust_cost): Handle Tremont.
2639 * config/i386/x86-tune.def (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY):
2641 (X86_TUNE_USE_LEAVE): Likewise.
2642 (X86_TUNE_PUSH_MEMORY): Likewise.
2643 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
2644 (X86_TUNE_USE_CLTD): Likewise.
2645 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
2646 (X86_TUNE_AVOID_MFENCE): Likewise.
2647 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
2648 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
2649 (X86_TUNE_ACCUMULATE_OUTGOING_ARGS): Disable for Tremont.
2650 (X86_TUNE_FOUR_JUMP_LIMIT): Likewise.
2651 (X86_TUNE_OPT_AGU): Likewise.
2652 (X86_TUNE_AVOID_LEA_FOR_ADDR): Likewise.
2653 (X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE): Likewise.
2654 (X86_TUNE_EXPAND_ABS): Likewise.
2655 (X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS): Likewise.
2656 (X86_TUNE_SLOW_PSHUFB): Likewise.
2658 2021-09-17 Eric Botcazou <ebotcazou@adacore.com>
2660 PR rtl-optimization/102306
2661 * combine.c (try_combine): Abort the combination if we are about to
2662 duplicate volatile references.
2664 2021-09-17 liuhongt <hongtao.liu@intel.com>
2666 * config/i386/avx512fp16intrin.h (_mm_undefined_ph):
2668 (_mm256_undefined_ph): Likewise.
2669 (_mm512_undefined_ph): Likewise.
2670 (_mm_cvtsh_h): Likewise.
2671 (_mm256_cvtsh_h): Likewise.
2672 (_mm512_cvtsh_h): Likewise.
2673 (_mm512_castph_ps): Likewise.
2674 (_mm512_castph_pd): Likewise.
2675 (_mm512_castph_si512): Likewise.
2676 (_mm512_castph512_ph128): Likewise.
2677 (_mm512_castph512_ph256): Likewise.
2678 (_mm512_castph128_ph512): Likewise.
2679 (_mm512_castph256_ph512): Likewise.
2680 (_mm512_zextph128_ph512): Likewise.
2681 (_mm512_zextph256_ph512): Likewise.
2682 (_mm512_castps_ph): Likewise.
2683 (_mm512_castpd_ph): Likewise.
2684 (_mm512_castsi512_ph): Likewise.
2685 * config/i386/avx512fp16vlintrin.h (_mm_castph_ps):
2687 (_mm256_castph_ps): Likewise.
2688 (_mm_castph_pd): Likewise.
2689 (_mm256_castph_pd): Likewise.
2690 (_mm_castph_si128): Likewise.
2691 (_mm256_castph_si256): Likewise.
2692 (_mm_castps_ph): Likewise.
2693 (_mm256_castps_ph): Likewise.
2694 (_mm_castpd_ph): Likewise.
2695 (_mm256_castpd_ph): Likewise.
2696 (_mm_castsi128_ph): Likewise.
2697 (_mm256_castsi256_ph): Likewise.
2698 (_mm256_castph256_ph128): Likewise.
2699 (_mm256_castph128_ph256): Likewise.
2700 (_mm256_zextph128_ph256): Likewise.
2702 2021-09-17 liuhongt <hongtao.liu@intel.com>
2704 * config/i386/avx512fp16intrin.h (_mm_cvtsh_ss):
2706 (_mm_mask_cvtsh_ss): Likewise.
2707 (_mm_maskz_cvtsh_ss): Likewise.
2708 (_mm_cvtsh_sd): Likewise.
2709 (_mm_mask_cvtsh_sd): Likewise.
2710 (_mm_maskz_cvtsh_sd): Likewise.
2711 (_mm_cvt_roundsh_ss): Likewise.
2712 (_mm_mask_cvt_roundsh_ss): Likewise.
2713 (_mm_maskz_cvt_roundsh_ss): Likewise.
2714 (_mm_cvt_roundsh_sd): Likewise.
2715 (_mm_mask_cvt_roundsh_sd): Likewise.
2716 (_mm_maskz_cvt_roundsh_sd): Likewise.
2717 (_mm_cvtss_sh): Likewise.
2718 (_mm_mask_cvtss_sh): Likewise.
2719 (_mm_maskz_cvtss_sh): Likewise.
2720 (_mm_cvtsd_sh): Likewise.
2721 (_mm_mask_cvtsd_sh): Likewise.
2722 (_mm_maskz_cvtsd_sh): Likewise.
2723 (_mm_cvt_roundss_sh): Likewise.
2724 (_mm_mask_cvt_roundss_sh): Likewise.
2725 (_mm_maskz_cvt_roundss_sh): Likewise.
2726 (_mm_cvt_roundsd_sh): Likewise.
2727 (_mm_mask_cvt_roundsd_sh): Likewise.
2728 (_mm_maskz_cvt_roundsd_sh): Likewise.
2729 * config/i386/i386-builtin-types.def
2730 (V8HF_FTYPE_V2DF_V8HF_V8HF_UQI_INT,
2731 V8HF_FTYPE_V4SF_V8HF_V8HF_UQI_INT,
2732 V2DF_FTYPE_V8HF_V2DF_V2DF_UQI_INT,
2733 V4SF_FTYPE_V8HF_V4SF_V4SF_UQI_INT): Add new builtin types.
2734 * config/i386/i386-builtin.def: Add corrresponding new builtins.
2735 * config/i386/i386-expand.c: Handle new builtin types.
2736 * config/i386/sse.md (VF48_128): New mode iterator.
2737 (avx512fp16_vcvtsh2<ssescalarmodesuffix><mask_scalar_name><round_saeonly_scalar_name>):
2739 (avx512fp16_vcvt<ssescalarmodesuffix>2sh<mask_scalar_name><round_scalar_name>):
2742 2021-09-17 liuhongt <hongtao.liu@intel.com>
2744 * config/i386/avx512fp16intrin.h (_mm512_cvtph_pd):
2746 (_mm512_mask_cvtph_pd): Likewise.
2747 (_mm512_maskz_cvtph_pd): Likewise.
2748 (_mm512_cvt_roundph_pd): Likewise.
2749 (_mm512_mask_cvt_roundph_pd): Likewise.
2750 (_mm512_maskz_cvt_roundph_pd): Likewise.
2751 (_mm512_cvtxph_ps): Likewise.
2752 (_mm512_mask_cvtxph_ps): Likewise.
2753 (_mm512_maskz_cvtxph_ps): Likewise.
2754 (_mm512_cvtx_roundph_ps): Likewise.
2755 (_mm512_mask_cvtx_roundph_ps): Likewise.
2756 (_mm512_maskz_cvtx_roundph_ps): Likewise.
2757 (_mm512_cvtxps_ph): Likewise.
2758 (_mm512_mask_cvtxps_ph): Likewise.
2759 (_mm512_maskz_cvtxps_ph): Likewise.
2760 (_mm512_cvtx_roundps_ph): Likewise.
2761 (_mm512_mask_cvtx_roundps_ph): Likewise.
2762 (_mm512_maskz_cvtx_roundps_ph): Likewise.
2763 (_mm512_cvtpd_ph): Likewise.
2764 (_mm512_mask_cvtpd_ph): Likewise.
2765 (_mm512_maskz_cvtpd_ph): Likewise.
2766 (_mm512_cvt_roundpd_ph): Likewise.
2767 (_mm512_mask_cvt_roundpd_ph): Likewise.
2768 (_mm512_maskz_cvt_roundpd_ph): Likewise.
2769 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_pd):
2771 (_mm_mask_cvtph_pd): Likewise.
2772 (_mm_maskz_cvtph_pd): Likewise.
2773 (_mm256_cvtph_pd): Likewise.
2774 (_mm256_mask_cvtph_pd): Likewise.
2775 (_mm256_maskz_cvtph_pd): Likewise.
2776 (_mm_cvtxph_ps): Likewise.
2777 (_mm_mask_cvtxph_ps): Likewise.
2778 (_mm_maskz_cvtxph_ps): Likewise.
2779 (_mm256_cvtxph_ps): Likewise.
2780 (_mm256_mask_cvtxph_ps): Likewise.
2781 (_mm256_maskz_cvtxph_ps): Likewise.
2782 (_mm_cvtxps_ph): Likewise.
2783 (_mm_mask_cvtxps_ph): Likewise.
2784 (_mm_maskz_cvtxps_ph): Likewise.
2785 (_mm256_cvtxps_ph): Likewise.
2786 (_mm256_mask_cvtxps_ph): Likewise.
2787 (_mm256_maskz_cvtxps_ph): Likewise.
2788 (_mm_cvtpd_ph): Likewise.
2789 (_mm_mask_cvtpd_ph): Likewise.
2790 (_mm_maskz_cvtpd_ph): Likewise.
2791 (_mm256_cvtpd_ph): Likewise.
2792 (_mm256_mask_cvtpd_ph): Likewise.
2793 (_mm256_maskz_cvtpd_ph): Likewise.
2794 * config/i386/i386-builtin.def: Add corresponding new builtins.
2795 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2796 * config/i386/i386-expand.c: Handle new builtin types.
2797 * config/i386/sse.md
2798 (VF4_128_8_256): New.
2799 (VF48H_AVX512VL): Ditto.
2800 (ssePHmode): Add HF vector modes.
2801 (castmode): Add new convertable modes.
2804 (avx512fp16_vcvt<castmode>2ph_<mode><mask_name><round_name>): Ditto.
2805 (avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
2806 (*avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
2807 (avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
2808 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
2809 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask_1): Ditto.
2810 (avx512fp16_float_extend_ph<mode>2<mask_name><round_saeonly_name>):
2812 (avx512fp16_float_extend_ph<mode>2<mask_name>): Ditto.
2813 (*avx512fp16_float_extend_ph<mode>2_load<mask_name>): Ditto.
2814 (avx512fp16_float_extend_phv2df2<mask_name>): Ditto.
2815 (*avx512fp16_float_extend_phv2df2_load<mask_name>): Ditto.
2817 2021-09-17 liuhongt <hongtao.liu@intel.com>
2819 * config/i386/avx512fp16intrin.h (_mm_cvttsh_i32):
2821 (_mm_cvttsh_u32): Likewise.
2822 (_mm_cvtt_roundsh_i32): Likewise.
2823 (_mm_cvtt_roundsh_u32): Likewise.
2824 (_mm_cvttsh_i64): Likewise.
2825 (_mm_cvttsh_u64): Likewise.
2826 (_mm_cvtt_roundsh_i64): Likewise.
2827 (_mm_cvtt_roundsh_u64): Likewise.
2828 * config/i386/i386-builtin.def: Add corresponding new builtins.
2829 * config/i386/sse.md
2830 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<round_saeonly_name>):
2833 2021-09-17 liuhongt <hongtao.liu@intel.com>
2835 * config/i386/avx512fp16intrin.h (_mm512_cvttph_epi32):
2837 (_mm512_mask_cvttph_epi32): Likewise.
2838 (_mm512_maskz_cvttph_epi32): Likewise.
2839 (_mm512_cvtt_roundph_epi32): Likewise.
2840 (_mm512_mask_cvtt_roundph_epi32): Likewise.
2841 (_mm512_maskz_cvtt_roundph_epi32): Likewise.
2842 (_mm512_cvttph_epu32): Likewise.
2843 (_mm512_mask_cvttph_epu32): Likewise.
2844 (_mm512_maskz_cvttph_epu32): Likewise.
2845 (_mm512_cvtt_roundph_epu32): Likewise.
2846 (_mm512_mask_cvtt_roundph_epu32): Likewise.
2847 (_mm512_maskz_cvtt_roundph_epu32): Likewise.
2848 (_mm512_cvttph_epi64): Likewise.
2849 (_mm512_mask_cvttph_epi64): Likewise.
2850 (_mm512_maskz_cvttph_epi64): Likewise.
2851 (_mm512_cvtt_roundph_epi64): Likewise.
2852 (_mm512_mask_cvtt_roundph_epi64): Likewise.
2853 (_mm512_maskz_cvtt_roundph_epi64): Likewise.
2854 (_mm512_cvttph_epu64): Likewise.
2855 (_mm512_mask_cvttph_epu64): Likewise.
2856 (_mm512_maskz_cvttph_epu64): Likewise.
2857 (_mm512_cvtt_roundph_epu64): Likewise.
2858 (_mm512_mask_cvtt_roundph_epu64): Likewise.
2859 (_mm512_maskz_cvtt_roundph_epu64): Likewise.
2860 (_mm512_cvttph_epi16): Likewise.
2861 (_mm512_mask_cvttph_epi16): Likewise.
2862 (_mm512_maskz_cvttph_epi16): Likewise.
2863 (_mm512_cvtt_roundph_epi16): Likewise.
2864 (_mm512_mask_cvtt_roundph_epi16): Likewise.
2865 (_mm512_maskz_cvtt_roundph_epi16): Likewise.
2866 (_mm512_cvttph_epu16): Likewise.
2867 (_mm512_mask_cvttph_epu16): Likewise.
2868 (_mm512_maskz_cvttph_epu16): Likewise.
2869 (_mm512_cvtt_roundph_epu16): Likewise.
2870 (_mm512_mask_cvtt_roundph_epu16): Likewise.
2871 (_mm512_maskz_cvtt_roundph_epu16): Likewise.
2872 * config/i386/avx512fp16vlintrin.h (_mm_cvttph_epi32):
2874 (_mm_mask_cvttph_epi32): Likewise.
2875 (_mm_maskz_cvttph_epi32): Likewise.
2876 (_mm256_cvttph_epi32): Likewise.
2877 (_mm256_mask_cvttph_epi32): Likewise.
2878 (_mm256_maskz_cvttph_epi32): Likewise.
2879 (_mm_cvttph_epu32): Likewise.
2880 (_mm_mask_cvttph_epu32): Likewise.
2881 (_mm_maskz_cvttph_epu32): Likewise.
2882 (_mm256_cvttph_epu32): Likewise.
2883 (_mm256_mask_cvttph_epu32): Likewise.
2884 (_mm256_maskz_cvttph_epu32): Likewise.
2885 (_mm_cvttph_epi64): Likewise.
2886 (_mm_mask_cvttph_epi64): Likewise.
2887 (_mm_maskz_cvttph_epi64): Likewise.
2888 (_mm256_cvttph_epi64): Likewise.
2889 (_mm256_mask_cvttph_epi64): Likewise.
2890 (_mm256_maskz_cvttph_epi64): Likewise.
2891 (_mm_cvttph_epu64): Likewise.
2892 (_mm_mask_cvttph_epu64): Likewise.
2893 (_mm_maskz_cvttph_epu64): Likewise.
2894 (_mm256_cvttph_epu64): Likewise.
2895 (_mm256_mask_cvttph_epu64): Likewise.
2896 (_mm256_maskz_cvttph_epu64): Likewise.
2897 (_mm_cvttph_epi16): Likewise.
2898 (_mm_mask_cvttph_epi16): Likewise.
2899 (_mm_maskz_cvttph_epi16): Likewise.
2900 (_mm256_cvttph_epi16): Likewise.
2901 (_mm256_mask_cvttph_epi16): Likewise.
2902 (_mm256_maskz_cvttph_epi16): Likewise.
2903 (_mm_cvttph_epu16): Likewise.
2904 (_mm_mask_cvttph_epu16): Likewise.
2905 (_mm_maskz_cvttph_epu16): Likewise.
2906 (_mm256_cvttph_epu16): Likewise.
2907 (_mm256_mask_cvttph_epu16): Likewise.
2908 (_mm256_maskz_cvttph_epu16): Likewise.
2909 * config/i386/i386-builtin.def: Add new builtins.
2910 * config/i386/sse.md
2911 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name><round_saeonly_name>):
2913 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name>): Ditto.
2914 (*avx512fp16_fix<fixunssuffix>_trunc<mode>2_load<mask_name>): Ditto.
2915 (avx512fp16_fix<fixunssuffix>_truncv2di2<mask_name>): Ditto.
2916 (avx512fp16_fix<fixunssuffix>_truncv2di2_load<mask_name>): Ditto.
2918 2021-09-17 liuhongt <hongtao.liu@intel.com>
2920 * config/i386/avx512fp16intrin.h (_mm_cvtsh_i32): New intrinsic.
2921 (_mm_cvtsh_u32): Likewise.
2922 (_mm_cvt_roundsh_i32): Likewise.
2923 (_mm_cvt_roundsh_u32): Likewise.
2924 (_mm_cvtsh_i64): Likewise.
2925 (_mm_cvtsh_u64): Likewise.
2926 (_mm_cvt_roundsh_i64): Likewise.
2927 (_mm_cvt_roundsh_u64): Likewise.
2928 (_mm_cvti32_sh): Likewise.
2929 (_mm_cvtu32_sh): Likewise.
2930 (_mm_cvt_roundi32_sh): Likewise.
2931 (_mm_cvt_roundu32_sh): Likewise.
2932 (_mm_cvti64_sh): Likewise.
2933 (_mm_cvtu64_sh): Likewise.
2934 (_mm_cvt_roundi64_sh): Likewise.
2935 (_mm_cvt_roundu64_sh): Likewise.
2936 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2937 * config/i386/i386-builtin.def: Add corresponding new builtins.
2938 * config/i386/i386-expand.c (ix86_expand_round_builtin):
2939 Handle new builtin types.
2940 * config/i386/sse.md
2941 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix><round_name>):
2943 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix>_2): Likewise.
2944 (avx512fp16_vcvt<floatsuffix>si2sh<rex64namesuffix><round_name>): Likewise.
2946 2021-09-16 Bill Schmidt <wschmidt@linux.ibm.com>
2948 * config/rs6000/rs6000-c.c (rs6000-builtins.h): New include.
2949 (altivec_resolve_new_overloaded_builtin): New forward decl.
2950 (rs6000_new_builtin_type_compatible): New function.
2951 (altivec_resolve_overloaded_builtin): Call
2952 altivec_resolve_new_overloaded_builtin.
2953 (altivec_build_new_resolved_builtin): New function.
2954 (altivec_resolve_new_overloaded_builtin): Likewise.
2955 * config/rs6000/rs6000-call.c (rs6000_new_builtin_is_supported):
2957 * config/rs6000/rs6000-gen-builtins.c (write_decls): Remove _p from
2958 name of rs6000_new_builtin_is_supported.
2960 2021-09-16 Uroš Bizjak <ubizjak@gmail.com>
2962 * config/i386/i386-protos.h (ix86_decompose_address):
2963 Change return type to bool.
2964 * config/i386/i386.c (ix86_decompose_address): Ditto.
2966 2021-09-16 Tobias Burnus <tobias@codesourcery.com>
2969 * config/rs6000/t-rs6000 (build/rs6000-gen-builtins.o, build/rbtree.o):
2970 Added 'build/' to target, use build/%.o rule.
2971 (build/rs6000-gen-builtins$(build_exeext)): Add 'build/' and
2972 '$(build_exeext)' to target and 'build/' for the *.o files.
2973 (rs6000-builtins.c): Update for those changes; run rs6000-gen-builtins
2976 2021-09-16 Martin Jambor <mjambor@suse.cz>
2978 * cgraph.c (cgraph_node::dump): Do not check caller count sums if
2979 the body has been removed. Remove trailing whitespace.
2981 2021-09-16 Richard Biener <rguenther@suse.de>
2983 PR middle-end/102360
2984 * internal-fn.c (expand_DEFERRED_INIT): Make pattern-init
2985 of non-memory more robust.
2987 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2989 * config/sparc/sparc-opts.h (enum sparc_processor_type): Add LEON5
2990 * config/sparc/sparc.c (struct processor_costs): Add LEON5 costs
2991 (leon5_adjust_cost): Increase cost of store with data dependency
2992 on ALU instruction and FPU anti-dependencies.
2993 (sparc_option_override): Add LEON5 costs
2994 (sparc_adjust_cost): Add LEON5 cost adjustments
2995 * config/sparc/sparc.h: Add LEON5
2996 * config/sparc/sparc.md: Include LEON5 scheduling information
2997 * config/sparc/sparc.opt: Add LEON5
2998 * doc/invoke.texi: Add LEON5
2999 * config/sparc/leon5.md: New file.
3001 2021-09-16 Daniel Cederman <cederman@gaisler.com>
3003 * config/sparc/sparc.md (stack_protect_set32): Add NOP to prevent
3004 sensitive sequence for B2BST errata workaround.
3006 2021-09-16 Daniel Cederman <cederman@gaisler.com>
3008 * config/sparc/sparc.c (sparc_do_work_around_errata): Do not begin
3009 functions with atomic instruction in the UT700 errata workaround.
3011 2021-09-16 Daniel Cederman <cederman@gaisler.com>
3013 * config/sparc/sparc.c (next_active_non_empty_insn): New function
3014 that returns next active non empty assembly instruction.
3015 (sparc_do_work_around_errata): Use new function.
3017 2021-09-16 Daniel Cederman <cederman@gaisler.com>
3019 * config/sparc/sparc.c (store_insn_p): Add predicate for store
3021 (load_insn_p): Add predicate for load attributes.
3022 (sparc_do_work_around_errata): Use new predicates.
3024 2021-09-16 Andreas Larsson <andreas@gaisler.com>
3026 * config/sparc/sparc.c (dump_target_flag_bits): Print bit names for
3029 2021-09-16 Martin Liska <mliska@suse.cz>
3031 * config/mips/netbsd.h: Fix typo in name of a macro.
3033 2021-09-16 liuhongt <hongtao.liu@intel.com>
3035 PR middle-end/102080
3036 * match.pd: Check mask type when doing cond_op related gimple
3038 * tree.c (is_truth_type_for): New function.
3039 * tree.h (is_truth_type_for): New declaration.
3041 2021-09-16 liuhongt <hongtao.liu@intel.com>
3043 * config/i386/avx512fp16intrin.h (_mm512_cvtepi32_ph): New
3045 (_mm512_mask_cvtepi32_ph): Likewise.
3046 (_mm512_maskz_cvtepi32_ph): Likewise.
3047 (_mm512_cvt_roundepi32_ph): Likewise.
3048 (_mm512_mask_cvt_roundepi32_ph): Likewise.
3049 (_mm512_maskz_cvt_roundepi32_ph): Likewise.
3050 (_mm512_cvtepu32_ph): Likewise.
3051 (_mm512_mask_cvtepu32_ph): Likewise.
3052 (_mm512_maskz_cvtepu32_ph): Likewise.
3053 (_mm512_cvt_roundepu32_ph): Likewise.
3054 (_mm512_mask_cvt_roundepu32_ph): Likewise.
3055 (_mm512_maskz_cvt_roundepu32_ph): Likewise.
3056 (_mm512_cvtepi64_ph): Likewise.
3057 (_mm512_mask_cvtepi64_ph): Likewise.
3058 (_mm512_maskz_cvtepi64_ph): Likewise.
3059 (_mm512_cvt_roundepi64_ph): Likewise.
3060 (_mm512_mask_cvt_roundepi64_ph): Likewise.
3061 (_mm512_maskz_cvt_roundepi64_ph): Likewise.
3062 (_mm512_cvtepu64_ph): Likewise.
3063 (_mm512_mask_cvtepu64_ph): Likewise.
3064 (_mm512_maskz_cvtepu64_ph): Likewise.
3065 (_mm512_cvt_roundepu64_ph): Likewise.
3066 (_mm512_mask_cvt_roundepu64_ph): Likewise.
3067 (_mm512_maskz_cvt_roundepu64_ph): Likewise.
3068 (_mm512_cvtepi16_ph): Likewise.
3069 (_mm512_mask_cvtepi16_ph): Likewise.
3070 (_mm512_maskz_cvtepi16_ph): Likewise.
3071 (_mm512_cvt_roundepi16_ph): Likewise.
3072 (_mm512_mask_cvt_roundepi16_ph): Likewise.
3073 (_mm512_maskz_cvt_roundepi16_ph): Likewise.
3074 (_mm512_cvtepu16_ph): Likewise.
3075 (_mm512_mask_cvtepu16_ph): Likewise.
3076 (_mm512_maskz_cvtepu16_ph): Likewise.
3077 (_mm512_cvt_roundepu16_ph): Likewise.
3078 (_mm512_mask_cvt_roundepu16_ph): Likewise.
3079 (_mm512_maskz_cvt_roundepu16_ph): Likewise.
3080 * config/i386/avx512fp16vlintrin.h (_mm_cvtepi32_ph): New
3082 (_mm_mask_cvtepi32_ph): Likewise.
3083 (_mm_maskz_cvtepi32_ph): Likewise.
3084 (_mm256_cvtepi32_ph): Likewise.
3085 (_mm256_mask_cvtepi32_ph): Likewise.
3086 (_mm256_maskz_cvtepi32_ph): Likewise.
3087 (_mm_cvtepu32_ph): Likewise.
3088 (_mm_mask_cvtepu32_ph): Likewise.
3089 (_mm_maskz_cvtepu32_ph): Likewise.
3090 (_mm256_cvtepu32_ph): Likewise.
3091 (_mm256_mask_cvtepu32_ph): Likewise.
3092 (_mm256_maskz_cvtepu32_ph): Likewise.
3093 (_mm_cvtepi64_ph): Likewise.
3094 (_mm_mask_cvtepi64_ph): Likewise.
3095 (_mm_maskz_cvtepi64_ph): Likewise.
3096 (_mm256_cvtepi64_ph): Likewise.
3097 (_mm256_mask_cvtepi64_ph): Likewise.
3098 (_mm256_maskz_cvtepi64_ph): Likewise.
3099 (_mm_cvtepu64_ph): Likewise.
3100 (_mm_mask_cvtepu64_ph): Likewise.
3101 (_mm_maskz_cvtepu64_ph): Likewise.
3102 (_mm256_cvtepu64_ph): Likewise.
3103 (_mm256_mask_cvtepu64_ph): Likewise.
3104 (_mm256_maskz_cvtepu64_ph): Likewise.
3105 (_mm_cvtepi16_ph): Likewise.
3106 (_mm_mask_cvtepi16_ph): Likewise.
3107 (_mm_maskz_cvtepi16_ph): Likewise.
3108 (_mm256_cvtepi16_ph): Likewise.
3109 (_mm256_mask_cvtepi16_ph): Likewise.
3110 (_mm256_maskz_cvtepi16_ph): Likewise.
3111 (_mm_cvtepu16_ph): Likewise.
3112 (_mm_mask_cvtepu16_ph): Likewise.
3113 (_mm_maskz_cvtepu16_ph): Likewise.
3114 (_mm256_cvtepu16_ph): Likewise.
3115 (_mm256_mask_cvtepu16_ph): Likewise.
3116 (_mm256_maskz_cvtepu16_ph): Likewise.
3117 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3118 * config/i386/i386-builtin.def: Add corresponding new builtins.
3119 * config/i386/i386-expand.c
3120 (ix86_expand_args_builtin): Handle new builtin types.
3121 (ix86_expand_round_builtin): Ditto.
3122 * config/i386/i386-modes.def: Declare V2HF and V6HF.
3123 * config/i386/sse.md (VI2H_AVX512VL): New.
3125 (sseintvecmode): Add HF vector modes.
3126 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode><mask_name><round_name>):
3128 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
3129 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
3130 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
3131 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
3132 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask_1): Ditto.
3133 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
3134 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
3135 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
3136 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
3137 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask_1): Ditto.
3138 * config/i386/subst.md (round_qq2phsuff): New subst_attr.
3140 2021-09-16 liuhongt <hongtao.liu@intel.com>
3142 * config/i386/avx512fp16intrin.h (_mm512_cvtph_epi32):
3144 (_mm512_mask_cvtph_epi32): Likewise.
3145 (_mm512_maskz_cvtph_epi32): Likewise.
3146 (_mm512_cvt_roundph_epi32): Likewise.
3147 (_mm512_mask_cvt_roundph_epi32): Likewise.
3148 (_mm512_maskz_cvt_roundph_epi32): Likewise.
3149 (_mm512_cvtph_epu32): Likewise.
3150 (_mm512_mask_cvtph_epu32): Likewise.
3151 (_mm512_maskz_cvtph_epu32): Likewise.
3152 (_mm512_cvt_roundph_epu32): Likewise.
3153 (_mm512_mask_cvt_roundph_epu32): Likewise.
3154 (_mm512_maskz_cvt_roundph_epu32): Likewise.
3155 (_mm512_cvtph_epi64): Likewise.
3156 (_mm512_mask_cvtph_epi64): Likewise.
3157 (_mm512_maskz_cvtph_epi64): Likewise.
3158 (_mm512_cvt_roundph_epi64): Likewise.
3159 (_mm512_mask_cvt_roundph_epi64): Likewise.
3160 (_mm512_maskz_cvt_roundph_epi64): Likewise.
3161 (_mm512_cvtph_epu64): Likewise.
3162 (_mm512_mask_cvtph_epu64): Likewise.
3163 (_mm512_maskz_cvtph_epu64): Likewise.
3164 (_mm512_cvt_roundph_epu64): Likewise.
3165 (_mm512_mask_cvt_roundph_epu64): Likewise.
3166 (_mm512_maskz_cvt_roundph_epu64): Likewise.
3167 (_mm512_cvtph_epi16): Likewise.
3168 (_mm512_mask_cvtph_epi16): Likewise.
3169 (_mm512_maskz_cvtph_epi16): Likewise.
3170 (_mm512_cvt_roundph_epi16): Likewise.
3171 (_mm512_mask_cvt_roundph_epi16): Likewise.
3172 (_mm512_maskz_cvt_roundph_epi16): Likewise.
3173 (_mm512_cvtph_epu16): Likewise.
3174 (_mm512_mask_cvtph_epu16): Likewise.
3175 (_mm512_maskz_cvtph_epu16): Likewise.
3176 (_mm512_cvt_roundph_epu16): Likewise.
3177 (_mm512_mask_cvt_roundph_epu16): Likewise.
3178 (_mm512_maskz_cvt_roundph_epu16): Likewise.
3179 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_epi32):
3181 (_mm_mask_cvtph_epi32): Likewise.
3182 (_mm_maskz_cvtph_epi32): Likewise.
3183 (_mm256_cvtph_epi32): Likewise.
3184 (_mm256_mask_cvtph_epi32): Likewise.
3185 (_mm256_maskz_cvtph_epi32): Likewise.
3186 (_mm_cvtph_epu32): Likewise.
3187 (_mm_mask_cvtph_epu32): Likewise.
3188 (_mm_maskz_cvtph_epu32): Likewise.
3189 (_mm256_cvtph_epu32): Likewise.
3190 (_mm256_mask_cvtph_epu32): Likewise.
3191 (_mm256_maskz_cvtph_epu32): Likewise.
3192 (_mm_cvtph_epi64): Likewise.
3193 (_mm_mask_cvtph_epi64): Likewise.
3194 (_mm_maskz_cvtph_epi64): Likewise.
3195 (_mm256_cvtph_epi64): Likewise.
3196 (_mm256_mask_cvtph_epi64): Likewise.
3197 (_mm256_maskz_cvtph_epi64): Likewise.
3198 (_mm_cvtph_epu64): Likewise.
3199 (_mm_mask_cvtph_epu64): Likewise.
3200 (_mm_maskz_cvtph_epu64): Likewise.
3201 (_mm256_cvtph_epu64): Likewise.
3202 (_mm256_mask_cvtph_epu64): Likewise.
3203 (_mm256_maskz_cvtph_epu64): Likewise.
3204 (_mm_cvtph_epi16): Likewise.
3205 (_mm_mask_cvtph_epi16): Likewise.
3206 (_mm_maskz_cvtph_epi16): Likewise.
3207 (_mm256_cvtph_epi16): Likewise.
3208 (_mm256_mask_cvtph_epi16): Likewise.
3209 (_mm256_maskz_cvtph_epi16): Likewise.
3210 (_mm_cvtph_epu16): Likewise.
3211 (_mm_mask_cvtph_epu16): Likewise.
3212 (_mm_maskz_cvtph_epu16): Likewise.
3213 (_mm256_cvtph_epu16): Likewise.
3214 (_mm256_mask_cvtph_epu16): Likewise.
3215 (_mm256_maskz_cvtph_epu16): Likewise.
3216 * config/i386/i386-builtin-types.def: Add new builtin types.
3217 * config/i386/i386-builtin.def: Add new builtins.
3218 * config/i386/i386-expand.c
3219 (ix86_expand_args_builtin): Handle new builtin types.
3220 (ix86_expand_round_builtin): Ditto.
3221 * config/i386/sse.md (sseintconvert): New.
3223 (UNSPEC_US_FIX_NOTRUNC): Ditto.
3224 (sseintconvertsignprefix): Ditto.
3225 (avx512fp16_vcvtph2<sseintconvertsignprefix><sseintconvert>_<mode><mask_name><round_name>):
3228 2021-09-16 liuhongt <hongtao.liu@intel.com>
3230 * config/i386/avx512fp16intrin.h: (_mm_cvtsi16_si128):
3232 (_mm_cvtsi128_si16): Likewise.
3233 (_mm_mask_load_sh): Likewise.
3234 (_mm_maskz_load_sh): Likewise.
3235 (_mm_mask_store_sh): Likewise.
3236 (_mm_move_sh): Likewise.
3237 (_mm_mask_move_sh): Likewise.
3238 (_mm_maskz_move_sh): Likewise.
3239 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3240 * config/i386/i386-builtin.def: Add corresponding new builtins.
3241 * config/i386/i386-expand.c
3242 (ix86_expand_special_args_builtin): Handle new builtin types.
3243 (ix86_expand_vector_init_one_nonzero): Adjust for FP16 target.
3244 * config/i386/sse.md (VI2F): New mode iterator.
3245 (vec_set<mode>_0): Use new mode iterator.
3246 (avx512f_mov<ssescalarmodelower>_mask): Adjust for HF vector mode.
3247 (avx512f_store<mode>_mask): Ditto.
3249 2021-09-16 Kewen Lin <linkw@linux.ibm.com>
3251 * config/rs6000/rs6000.opt (-mtoc-fusion): Remove.
3253 2021-09-15 David Edelsohn <dje.gcc@gmail.com>
3255 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
3256 Proceed if no symbol summary or the symbol alias flag is false.
3258 2021-09-15 Jakub Jelinek <jakub@redhat.com>
3262 * varasm.c (output_constructor_regular_field): Instead of assertion
3263 that array_size_for_constructor result is equal to size of
3264 TREE_TYPE (local->val) in bytes, assert that the type size is greater
3265 or equal to array_size_for_constructor result and use type size as
3268 2021-09-15 Martin Liska <mliska@suse.cz>
3271 * config/i386/vxworks.h: Use new macro TARGET_CPU_P.
3273 2021-09-15 Martin Liska <mliska@suse.cz>
3276 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
3277 Check that we have a symbol summary for a symbol.
3279 2021-09-15 Richard Biener <rguenther@suse.de>
3282 * config/rs6000/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
3283 to inherit from elfos.h
3285 2021-09-15 liuhongt <hongtao.liu@intel.com>
3288 * config/i386/i386-expand.c
3289 (ix86_expand_vector_init_interleave): Use puncklwd to pack 2
3291 (ix86_expand_vector_set): Use blendw instead of pinsrw.
3292 * config/i386/i386.c (ix86_can_change_mode_class): Adjust for
3293 AVX512FP16 which supports 16bit vector load.
3294 * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>):
3296 (avx512bw_interleave_high<mode><mask_name>): .. this, and
3297 extend to V32HFmode.
3298 (avx2_interleave_highv16hi<mask_name>): Rename to ..
3299 (avx2_interleave_high<mode><mask_name>): .. this, and extend
3301 (vec_interleave_highv8hi<mask_name>): Rename to ..
3302 (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode.
3303 (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>):
3305 (<mask_codefor>avx512bw_interleave_low<mode><mask_name>):
3306 this, and extend to V32HFmode.
3307 (avx2_interleave_lowv16hi<mask_name>): Rename to ..
3308 (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode.
3309 (vec_interleave_lowv8hi<mask_name>): Rename to ..
3310 (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode.
3311 (sse4_1_pblendw): Rename to ..
3312 (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode.
3313 (avx2_pblendph): New define_expand.
3314 (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use
3315 sseintmodesuffix instead of ssemodesuffix.
3316 (blendsuf): New mode attr.
3318 2021-09-15 Richard Biener <rguenther@suse.de>
3320 * tree-vectorizer.h (dr_misalignment): Move out of line.
3321 (dr_target_alignment): New.
3322 (DR_TARGET_ALIGNMENT): Wrap dr_target_alignment.
3323 (set_dr_target_alignment): New.
3324 (SET_DR_TARGET_ALIGNMENT): Wrap set_dr_target_alignment.
3325 * tree-vect-data-refs.c (dr_misalignment): Compute and
3326 return the group members misalignment.
3327 (vect_compute_data_ref_alignment): Use SET_DR_TARGET_ALIGNMENT.
3328 (vect_analyze_data_refs_alignment): Compute alignment only
3329 for the first element of a DR group.
3330 (vect_slp_analyze_node_alignment): Likewise.
3332 2021-09-15 Hongyu Wang <hongyu.wang@intel.com>
3334 * config/i386/avx512fp16intrin.h: Adjust all builtin calls.
3335 * config/i386/avx512fp16vlintrin.h: Likewise.
3336 * config/i386/i386-builtin.def: Adjust builtin name and
3337 enumeration to match AVX512F style.
3339 2021-09-15 Richard Biener <rguenther@suse.de>
3341 PR tree-optimization/102318
3342 * tree-vect-loop.c (vect_transform_cycle_phi): Revert
3343 previous change and do the mode conversion separately from
3344 the sign conversion.
3346 2021-09-15 Hongtao Liu <hongtao.liu@intel.com>
3347 Peter Cordes <peter@cordes.ca>
3350 * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
3351 (*vec_extract<mode><ssescalarmodelower>_valign): Output
3352 vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
3355 2021-09-15 Richard Biener <rguenther@suse.de>
3357 * config.gcc: Remove vax-*-openbsd* configuration.
3359 2021-09-15 Richard Biener <rguenther@suse.de>
3361 * config.gcc: Remove m68k-openbsd.
3363 2021-09-15 Max Filippov <jcmvbkbc@gmail.com>
3366 * config/xtensa/t-xtensa (TM_H): Add include/xtensa-config.h.
3368 2021-09-14 Peter Bergner <bergner@linux.ibm.com>
3370 * config/rs6000/mma.md (unspec): Delete UNSPEC_MMA_XXSETACCZ.
3371 (unspecv): Add UNSPECV_MMA_XXSETACCZ.
3372 (*mma_xxsetaccz): Delete.
3373 (mma_xxsetaccz): Change to define_insn. Remove operand 1.
3374 Use UNSPECV_MMA_XXSETACCZ. Update comment.
3375 * config/rs6000/rs6000.c (rs6000_rtx_costs): Use UNSPECV_MMA_XXSETACCZ.
3377 2021-09-14 Iain Sandoe <iain@sandoe.co.uk>
3379 * Makefile.in: Remove variables related to applying no-PIE
3380 to the exes on $build.
3381 * configure: Regenerate.
3382 * configure.ac: Remove configuration related to applying
3383 no-PIE to the exes on $build.
3385 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
3387 * config/arc/arc.md (doloop_end): Add missing mode.
3388 (loop_end): Likewise.
3390 2021-09-14 Jakub Jelinek <jakub@redhat.com>
3392 * gimplify.c (goa_stabilize_expr): Add depth argument, propagate
3393 it to recursive calls, for depth above 7 just gimplify or return.
3394 Perform a test even for MODIFY_EXPR, ADDR_EXPR, COMPOUND_EXPR with
3395 __builtin_clear_padding and TARGET_EXPR.
3396 (gimplify_omp_atomic): Adjust goa_stabilize_expr callers.
3398 2021-09-14 liuhongt <hongtao.liu@intel.com>
3400 * config/i386/avx512fp16intrin.h (_mm_fpclass_sh_mask):
3402 (_mm_mask_fpclass_sh_mask): Likewise.
3403 (_mm512_mask_fpclass_ph_mask): Likewise.
3404 (_mm512_fpclass_ph_mask): Likewise.
3405 (_mm_getexp_sh): Likewise.
3406 (_mm_mask_getexp_sh): Likewise.
3407 (_mm_maskz_getexp_sh): Likewise.
3408 (_mm512_getexp_ph): Likewise.
3409 (_mm512_mask_getexp_ph): Likewise.
3410 (_mm512_maskz_getexp_ph): Likewise.
3411 (_mm_getexp_round_sh): Likewise.
3412 (_mm_mask_getexp_round_sh): Likewise.
3413 (_mm_maskz_getexp_round_sh): Likewise.
3414 (_mm512_getexp_round_ph): Likewise.
3415 (_mm512_mask_getexp_round_ph): Likewise.
3416 (_mm512_maskz_getexp_round_ph): Likewise.
3417 (_mm_getmant_sh): Likewise.
3418 (_mm_mask_getmant_sh): Likewise.
3419 (_mm_maskz_getmant_sh): Likewise.
3420 (_mm512_getmant_ph): Likewise.
3421 (_mm512_mask_getmant_ph): Likewise.
3422 (_mm512_maskz_getmant_ph): Likewise.
3423 (_mm_getmant_round_sh): Likewise.
3424 (_mm_mask_getmant_round_sh): Likewise.
3425 (_mm_maskz_getmant_round_sh): Likewise.
3426 (_mm512_getmant_round_ph): Likewise.
3427 (_mm512_mask_getmant_round_ph): Likewise.
3428 (_mm512_maskz_getmant_round_ph): Likewise.
3429 * config/i386/avx512fp16vlintrin.h (_mm_mask_fpclass_ph_mask):
3431 (_mm_fpclass_ph_mask): Likewise.
3432 (_mm256_mask_fpclass_ph_mask): Likewise.
3433 (_mm256_fpclass_ph_mask): Likewise.
3434 (_mm256_getexp_ph): Likewise.
3435 (_mm256_mask_getexp_ph): Likewise.
3436 (_mm256_maskz_getexp_ph): Likewise.
3437 (_mm_getexp_ph): Likewise.
3438 (_mm_mask_getexp_ph): Likewise.
3439 (_mm_maskz_getexp_ph): Likewise.
3440 (_mm256_getmant_ph): Likewise.
3441 (_mm256_mask_getmant_ph): Likewise.
3442 (_mm256_maskz_getmant_ph): Likewise.
3443 (_mm_getmant_ph): Likewise.
3444 (_mm_mask_getmant_ph): Likewise.
3445 (_mm_maskz_getmant_ph): Likewise.
3446 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3447 * config/i386/i386-builtin.def: Add corresponding new builtins.
3448 * config/i386/i386-expand.c
3449 (ix86_expand_args_builtin): Handle new builtin types.
3450 (ix86_expand_round_builtin): Ditto.
3451 * config/i386/sse.md (vecmemsuffix): Add HF vector modes.
3452 (<avx512>_getexp<mode><mask_name><round_saeonly_name>): Adjust
3453 to support HF vector modes.
3454 (avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name):
3456 (avx512dq_fpclass<mode><mask_scalar_merge_name>): Ditto.
3457 (avx512dq_vmfpclass<mode><mask_scalar_merge_name>): Ditto.
3458 (<avx512>_getmant<mode><mask_name><round_saeonly_name>): Ditto.
3459 (avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3462 2021-09-14 liuhongt <hongtao.liu@intel.com>
3464 * config/i386/avx512fp16intrin.h (_mm512_reduce_ph):
3466 (_mm512_mask_reduce_ph): Likewise.
3467 (_mm512_maskz_reduce_ph): Likewise.
3468 (_mm512_reduce_round_ph): Likewise.
3469 (_mm512_mask_reduce_round_ph): Likewise.
3470 (_mm512_maskz_reduce_round_ph): Likewise.
3471 (_mm_reduce_sh): Likewise.
3472 (_mm_mask_reduce_sh): Likewise.
3473 (_mm_maskz_reduce_sh): Likewise.
3474 (_mm_reduce_round_sh): Likewise.
3475 (_mm_mask_reduce_round_sh): Likewise.
3476 (_mm_maskz_reduce_round_sh): Likewise.
3477 (_mm512_roundscale_ph): Likewise.
3478 (_mm512_mask_roundscale_ph): Likewise.
3479 (_mm512_maskz_roundscale_ph): Likewise.
3480 (_mm512_roundscale_round_ph): Likewise.
3481 (_mm512_mask_roundscale_round_ph): Likewise.
3482 (_mm512_maskz_roundscale_round_ph): Likewise.
3483 (_mm_roundscale_sh): Likewise.
3484 (_mm_mask_roundscale_sh): Likewise.
3485 (_mm_maskz_roundscale_sh): Likewise.
3486 (_mm_roundscale_round_sh): Likewise.
3487 (_mm_mask_roundscale_round_sh): Likewise.
3488 (_mm_maskz_roundscale_round_sh): Likewise.
3489 * config/i386/avx512fp16vlintrin.h: (_mm_reduce_ph):
3491 (_mm_mask_reduce_ph): Likewise.
3492 (_mm_maskz_reduce_ph): Likewise.
3493 (_mm256_reduce_ph): Likewise.
3494 (_mm256_mask_reduce_ph): Likewise.
3495 (_mm256_maskz_reduce_ph): Likewise.
3496 (_mm_roundscale_ph): Likewise.
3497 (_mm_mask_roundscale_ph): Likewise.
3498 (_mm_maskz_roundscale_ph): Likewise.
3499 (_mm256_roundscale_ph): Likewise.
3500 (_mm256_mask_roundscale_ph): Likewise.
3501 (_mm256_maskz_roundscale_ph): Likewise.
3502 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3503 * config/i386/i386-builtin.def: Add corresponding new builtins.
3504 * config/i386/i386-expand.c
3505 (ix86_expand_args_builtin): Handle new builtin types.
3506 (ix86_expand_round_builtin): Ditto.
3507 * config/i386/sse.md (<mask_codefor>reducep<mode><mask_name>):
3509 (<mask_codefor>reducep<mode><mask_name><round_saeonly_name>):
3510 ... this, and adjust for round operands.
3511 (reduces<mode><mask_scalar_name>): Likewise, with ...
3512 (reduces<mode><mask_scalar_name><round_saeonly_scalar_name):
3514 (<avx512>_rndscale<mode><mask_name><round_saeonly_name>):
3515 Adjust for HF vector modes.
3516 (avx512f_rndscale<mode><mask_scalar_name><round_saeonly_scalar_name>):
3518 (*avx512f_rndscale<mode><round_saeonly_name>): Ditto.
3520 2021-09-14 liuhongt <hongtao.liu@intel.com>
3522 * config/i386/avx512fp16intrin.h: (_mm512_rcp_ph):
3524 (_mm512_mask_rcp_ph): Likewise.
3525 (_mm512_maskz_rcp_ph): Likewise.
3526 (_mm_rcp_sh): Likewise.
3527 (_mm_mask_rcp_sh): Likewise.
3528 (_mm_maskz_rcp_sh): Likewise.
3529 (_mm512_scalef_ph): Likewise.
3530 (_mm512_mask_scalef_ph): Likewise.
3531 (_mm512_maskz_scalef_ph): Likewise.
3532 (_mm512_scalef_round_ph): Likewise.
3533 (_mm512_mask_scalef_round_ph): Likewise.
3534 (_mm512_maskz_scalef_round_ph): Likewise.
3535 (_mm_scalef_sh): Likewise.
3536 (_mm_mask_scalef_sh): Likewise.
3537 (_mm_maskz_scalef_sh): Likewise.
3538 (_mm_scalef_round_sh): Likewise.
3539 (_mm_mask_scalef_round_sh): Likewise.
3540 (_mm_maskz_scalef_round_sh): Likewise.
3541 * config/i386/avx512fp16vlintrin.h (_mm_rcp_ph):
3543 (_mm256_rcp_ph): Likewise.
3544 (_mm_mask_rcp_ph): Likewise.
3545 (_mm256_mask_rcp_ph): Likewise.
3546 (_mm_maskz_rcp_ph): Likewise.
3547 (_mm256_maskz_rcp_ph): Likewise.
3548 (_mm_scalef_ph): Likewise.
3549 (_mm256_scalef_ph): Likewise.
3550 (_mm_mask_scalef_ph): Likewise.
3551 (_mm256_mask_scalef_ph): Likewise.
3552 (_mm_maskz_scalef_ph): Likewise.
3553 (_mm256_maskz_scalef_ph): Likewise.
3554 * config/i386/i386-builtin.def: Add new builtins.
3555 * config/i386/sse.md (VFH_AVX512VL): New.
3556 (avx512fp16_rcp<mode>2<mask_name>): Ditto.
3557 (avx512fp16_vmrcpv8hf2<mask_scalar_name>): Ditto.
3558 (avx512f_vmscalef<mode><mask_scalar_name><round_scalar_name>):
3559 Adjust to support HF vector modes.
3560 (<avx512>_scalef<mode><mask_name><round_name>): Ditto.
3562 2021-09-14 liuhongt <hongtao.liu@intel.com>
3564 * config/i386/avx512fp16intrin.h: (_mm512_sqrt_ph):
3566 (_mm512_mask_sqrt_ph): Likewise.
3567 (_mm512_maskz_sqrt_ph): Likewise.
3568 (_mm512_sqrt_round_ph): Likewise.
3569 (_mm512_mask_sqrt_round_ph): Likewise.
3570 (_mm512_maskz_sqrt_round_ph): Likewise.
3571 (_mm512_rsqrt_ph): Likewise.
3572 (_mm512_mask_rsqrt_ph): Likewise.
3573 (_mm512_maskz_rsqrt_ph): Likewise.
3574 (_mm_rsqrt_sh): Likewise.
3575 (_mm_mask_rsqrt_sh): Likewise.
3576 (_mm_maskz_rsqrt_sh): Likewise.
3577 (_mm_sqrt_sh): Likewise.
3578 (_mm_mask_sqrt_sh): Likewise.
3579 (_mm_maskz_sqrt_sh): Likewise.
3580 (_mm_sqrt_round_sh): Likewise.
3581 (_mm_mask_sqrt_round_sh): Likewise.
3582 (_mm_maskz_sqrt_round_sh): Likewise.
3583 * config/i386/avx512fp16vlintrin.h (_mm_sqrt_ph): New intrinsic.
3584 (_mm256_sqrt_ph): Likewise.
3585 (_mm_mask_sqrt_ph): Likewise.
3586 (_mm256_mask_sqrt_ph): Likewise.
3587 (_mm_maskz_sqrt_ph): Likewise.
3588 (_mm256_maskz_sqrt_ph): Likewise.
3589 (_mm_rsqrt_ph): Likewise.
3590 (_mm256_rsqrt_ph): Likewise.
3591 (_mm_mask_rsqrt_ph): Likewise.
3592 (_mm256_mask_rsqrt_ph): Likewise.
3593 (_mm_maskz_rsqrt_ph): Likewise.
3594 (_mm256_maskz_rsqrt_ph): Likewise.
3595 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3596 * config/i386/i386-builtin.def: Add corresponding new builtins.
3597 * config/i386/i386-expand.c
3598 (ix86_expand_args_builtin): Handle new builtins.
3599 (ix86_expand_round_builtin): Ditto.
3600 * config/i386/sse.md (VF_AVX512FP16VL): New.
3601 (sqrt<mode>2): Adjust for HF vector modes.
3602 (<sse>_sqrt<mode>2<mask_name><round_name>): Likewise.
3603 (<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
3605 (<sse>_rsqrt<mode>2<mask_name>): New.
3606 (avx512fp16_vmrsqrtv8hf2<mask_scalar_name>): Likewise.
3608 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
3611 * diagnostic-spec.c (warning_suppressed_at, copy_warning): Handle
3612 'RESERVED_LOCATION_P' locations.
3613 * warning-control.cc (get_nowarn_spec, suppress_warning)
3614 (copy_warning): Likewise.
3616 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
3618 * diagnostic-spec.h (typedef xint_hash_t): Use 'location_t' instead of...
3619 (typedef key_type_t): ... this. Remove.
3620 (nowarn_map): Document.
3621 * diagnostic-spec.c (nowarn_map): Likewise.
3622 * warning-control.cc (convert_to_key): Evolve functions into...
3623 (get_location): ... these. Adjust all users.
3625 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
3627 * warning-control.cc (copy_warning): Remove 'nowarn_map' setup.
3629 2021-09-13 Jason Merrill <jason@redhat.com>
3631 * params.opt: Add destructive-interference-size and
3632 constructive-interference-size.
3633 * doc/invoke.texi: Document them.
3634 * config/aarch64/aarch64.c (aarch64_override_options_internal):
3636 * config/arm/arm.c (arm_option_override): Set them.
3637 * config/i386/i386-options.c (ix86_option_override_internal):
3640 2021-09-13 Martin Liska <mliska@suse.cz>
3641 H.J. Lu <hjl.tools@gmail.com>
3644 * common/config/i386/cpuinfo.h (cpu_indicator_init): Add support
3645 for x86-64 micro levels for __builtin_cpu_supports.
3646 * common/config/i386/i386-cpuinfo.h (enum feature_priority):
3647 Add priorities for the micro-arch levels.
3648 (enum processor_features): Add new features.
3649 * common/config/i386/i386-isas.h: Add micro-arch features.
3650 * config/i386/i386-builtins.c (get_builtin_code_for_version):
3651 Support the micro-arch levels by callsing
3652 __builtin_cpu_supports.
3653 * doc/extend.texi: Document that the levels are support by
3654 __builtin_cpu_supports.
3656 2021-09-13 Andrew Pinski <apinski@marvell.com>
3659 * config/aarch64/aarch64-builtins.c (aarch64_fold_builtin_lane_check):
3661 (aarch64_general_fold_builtin): Handle AARCH64_SIMD_BUILTIN_LANE_CHECK.
3662 (aarch64_general_gimple_fold_builtin): Likewise.
3664 2021-09-13 Andrew Pinski <apinski@marvell.com>
3666 * config.gcc: Add m32r-*-linux* and m32rle-*-linux*
3667 to the Unsupported targets list.
3668 Remove support for m32r-*-linux* and m32rle-*-linux*.
3669 * config/m32r/linux.h: Removed.
3670 * config/m32r/t-linux: Removed.
3672 2021-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3675 * config/aarch64/aarch64.c (aarch64_classify_address): Don't allow
3676 register index for SVE predicate modes.
3678 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
3680 * tree-ssa-threadbackward.c
3681 (back_threader_profitability::profitable_path_p): Remove FSM
3683 (back_threader_registry::register_path): Same.
3684 * tree-ssa-threadedge.c
3685 (jump_threader::simplify_control_stmt_condition): Same.
3686 * tree-ssa-threadupdate.c (jt_path_registry::jt_path_registry):
3687 Add backedge_threads argument.
3688 (fwd_jt_path_registry::fwd_jt_path_registry): Pass
3689 backedge_threads argument.
3690 (back_jt_path_registry::back_jt_path_registry): Same.
3691 (dump_jump_thread_path): Adjust for FSM removal.
3692 (back_jt_path_registry::rewire_first_differing_edge): Same.
3693 (back_jt_path_registry::adjust_paths_after_duplication): Same.
3694 (back_jt_path_registry::update_cfg): Same.
3695 (jt_path_registry::register_jump_thread): Same.
3696 * tree-ssa-threadupdate.h (enum jump_thread_edge_type): Remove
3698 (class back_jt_path_registry): Add backedge_threads to
3701 2021-09-13 Martin Liska <mliska@suse.cz>
3704 * asan.h (sanitize_coverage_p): Handle when fn == NULL.
3706 2021-09-13 H.J. Lu <hjl.tools@gmail.com>
3709 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): New.
3710 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
3711 (MOVE_MAX): Check TARGET_AVX256_MOVE_BY_PIECES and
3712 TARGET_AVX256_STORE_BY_PIECES instead of
3713 TARGET_AVX256_SPLIT_UNALIGNED_LOAD and
3714 TARGET_AVX256_SPLIT_UNALIGNED_STORE.
3715 (STORE_MAX_PIECES): Check TARGET_AVX256_STORE_BY_PIECES instead
3716 of TARGET_AVX256_SPLIT_UNALIGNED_STORE.
3717 * config/i386/x86-tune.def (X86_TUNE_AVX256_MOVE_BY_PIECES): New.
3718 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
3720 2021-09-13 liuhongt <hongtao.liu@intel.com>
3723 * expmed.c (extract_bit_field_using_extv): Use
3724 gen_lowpart_if_possible instead of gen_lowpart to avoid ICE.
3726 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
3728 * Makefile.in (OBJS): Add value-pointer-equiv.o.
3729 * gimple-ssa-evrp.c (class ssa_equiv_stack): Move to
3730 value-pointer-equiv.*.
3731 (ssa_equiv_stack::ssa_equiv_stack): Same.
3732 (ssa_equiv_stack::enter): Same.
3733 (ssa_equiv_stack::leave): Same.
3734 (ssa_equiv_stack::push_replacement): Same.
3735 (ssa_equiv_stack::get_replacement): Same.
3736 (is_pointer_ssa): Same.
3737 (class pointer_equiv_analyzer): Same.
3738 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
3739 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Same.
3740 (pointer_equiv_analyzer::set_global_equiv): Same.
3741 (pointer_equiv_analyzer::set_cond_equiv): Same.
3742 (pointer_equiv_analyzer::get_equiv): Same.
3743 (pointer_equiv_analyzer::enter): Same.
3744 (pointer_equiv_analyzer::leave): Same.
3745 (pointer_equiv_analyzer::get_equiv_expr): Same.
3746 (pta_valueize): Same.
3747 (pointer_equiv_analyzer::visit_stmt): Same.
3748 (pointer_equiv_analyzer::visit_edge): Same.
3749 (hybrid_folder::value_of_expr): Same.
3750 (hybrid_folder::value_on_edge): Same.
3751 * value-pointer-equiv.cc: New file.
3752 * value-pointer-equiv.h: New file.
3754 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
3757 * gimple-fold.c (gimple_fold_builtin_memory_op): Allow folding
3758 memcpy if the size is not more than MOVE_MAX * MOVE_RATIO.
3760 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
3763 * config/arm/arm.md (movmisaligndi): New define_expand.
3764 * config/arm/vec-common.md (movmisalign<mode>): Iterate over VDQ mode.
3766 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
3769 * emit-rtl.c (gen_highpart): Use adjust_address to handle
3770 MEM rather than calling simplify_gen_subreg.
3772 2021-09-13 Jan-Benedict Glaw <jbglaw@ług-owl.de>
3774 * config/alpha/vms.h (INIT_CUMULATIVE_ARGS): Wrap multi-statment
3775 define into a block.
3777 2021-09-13 Richard Biener <rguenther@suse.de>
3779 * config/darwin.h (DARWIN_PREFER_DWARF): Do not define.
3780 * config/i386/darwin.h (PREFERRED_DEBUGGING_TYPE): Do not
3781 change based on DARWIN_PREFER_DWARF not being defined.
3783 2021-09-13 Richard Biener <rguenther@suse.de>
3785 * config/i386/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
3786 to inherit from elfos.h
3788 2021-09-13 Richard Biener <rguenther@suse.de>
3790 * config.gcc: Add cr16-*-* to the list of obsoleted targets.
3792 2021-09-13 Richard Biener <rguenther@suse.de>
3794 * config/avr/elf.h (PREFERRED_DEBUGGING_TYPE): Remove
3795 override, pick up DWARF2_DEBUG define from elfos.h
3797 2021-09-13 Richard Biener <rguenther@suse.de>
3799 * config/rx/rx.h (PREFERRED_DEBUGGING_TYPE): Always define to
3802 2021-09-13 Richard Biener <rguenther@suse.de>
3804 * config/alpha/vms.h (PREFERRED_DEBUGGING_TYPE): Define to
3807 2021-09-13 Richard Biener <rguenther@suse.de>
3809 * config/i386/cygming.h: Always default to DWARF2 debugging.
3810 Do not define DBX_DEBUGGING_INFO, that's done via dbxcoff.h
3812 * doc/install.texi: Document binutils 2.16 as minimum
3813 requirement for mingw.
3815 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
3817 * config/rs6000/rs6000.c (struct rs6000_cost_data): New members
3818 nstmts, nloads and extra_ctor_cost.
3819 (rs6000_density_test): Add load density related heuristics. Do
3820 extra costing on vector construction statements if need.
3821 (rs6000_init_cost): Init new members.
3822 (rs6000_update_target_cost_per_stmt): New function.
3823 (rs6000_add_stmt_cost): Factor vect_nonmem hunk out to function
3824 rs6000_update_target_cost_per_stmt and call it.
3826 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
3828 * config/rs6000/rs6000.c (struct rs6000_cost_data): Remove typedef.
3829 (rs6000_init_cost): Adjust.
3831 2021-09-13 liuhongt <hongtao.liu@intel.com>
3833 * config/i386/i386.md: (UNSPEC_COPYSIGN): Remove.
3834 (UNSPEC_XORSIGN): Ditto.
3836 2021-09-12 Roger Sayle <roger@nextmovesoftware.com>
3838 * expr.c (convert_move): Preserve SUBREG_PROMOTED_VAR_P when
3839 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
3842 2021-09-11 Aldy Hernandez <aldyh@redhat.com>
3844 * tree-ssa-threadbackward.c (class back_threader_registry): Use
3845 back_jt_path_registry.
3846 * tree-ssa-threadedge.c (jump_threader::jump_threader): Use
3847 fwd_jt_path_registry.
3848 * tree-ssa-threadedge.h (class jump_threader): Same..
3849 * tree-ssa-threadupdate.c
3850 (jump_thread_path_registry::jump_thread_path_registry): Rename...
3851 (jt_path_registry::jt_path_registry): ...to this.
3852 (jump_thread_path_registry::~jump_thread_path_registry): Rename...
3853 (jt_path_registry::~jt_path_registry): ...this.
3854 (fwd_jt_path_registry::fwd_jt_path_registry): New.
3855 (fwd_jt_path_registry::~fwd_jt_path_registry): New.
3856 (jump_thread_path_registry::allocate_thread_edge): Rename...
3857 (jt_path_registry::allocate_thread_edge): ...to this.
3858 (jump_thread_path_registry::allocate_thread_path): Rename...
3859 (jt_path_registry::allocate_thread_path): ...to this.
3860 (jump_thread_path_registry::lookup_redirection_data): Rename...
3861 (fwd_jt_path_registry::lookup_redirection_data): ...to this.
3862 (jump_thread_path_registry::thread_block_1): Rename...
3863 (fwd_jt_path_registry::thread_block_1): ...to this.
3864 (jump_thread_path_registry::thread_block): Rename...
3865 (fwd_jt_path_registry::thread_block): ...to this.
3866 (jt_path_registry::thread_through_loop_header): Rename...
3867 (fwd_jt_path_registry::thread_through_loop_header): ...to this.
3868 (jump_thread_path_registry::mark_threaded_blocks): Rename...
3869 (fwd_jt_path_registry::mark_threaded_blocks): ...to this.
3870 (jump_thread_path_registry::debug_path): Rename...
3871 (jt_path_registry::debug_path): ...to this.
3872 (jump_thread_path_registry::dump): Rename...
3873 (jt_path_registry::debug): ...to this.
3874 (jump_thread_path_registry::rewire_first_differing_edge): Rename...
3875 (back_jt_path_registry::rewire_first_differing_edge): ...to this.
3876 (jump_thread_path_registry::adjust_paths_after_duplication): Rename...
3877 (back_jt_path_registry::adjust_paths_after_duplication): ...to this.
3878 (jump_thread_path_registry::duplicate_thread_path): Rename...
3879 (back_jt_path_registry::duplicate_thread_path): ...to this. Also,
3880 drop ill-formed candidates.
3881 (jump_thread_path_registry::remove_jump_threads_including): Rename...
3882 (fwd_jt_path_registry::remove_jump_threads_including): ...to this.
3883 (jt_path_registry::thread_through_all_blocks): New.
3884 (back_jt_path_registry::update_cfg): New.
3885 (fwd_jt_path_registry::update_cfg): New.
3886 (jump_thread_path_registry::register_jump_thread): Rename...
3887 (jt_path_registry::register_jump_thread): ...to this.
3888 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
3890 (class jt_path_registry): ...here.
3891 (class fwd_jt_path_registry): New.
3892 (class back_jt_path_registry): New.
3894 2021-09-10 liuhongt <hongtao.liu@intel.com>
3897 2021-09-01 liuhongt <hongtao.liu@intel.com>
3899 * emit-rtl.c (validate_subreg): Get rid of all float-int
3902 2021-09-10 Jakub Jelinek <jakub@redhat.com>
3904 * tree-core.h (enum omp_memory_order): Add OMP_MEMORY_ORDER_MASK,
3905 OMP_FAIL_MEMORY_ORDER_UNSPECIFIED, OMP_FAIL_MEMORY_ORDER_RELAXED,
3906 OMP_FAIL_MEMORY_ORDER_ACQUIRE, OMP_FAIL_MEMORY_ORDER_RELEASE,
3907 OMP_FAIL_MEMORY_ORDER_ACQ_REL, OMP_FAIL_MEMORY_ORDER_SEQ_CST and
3908 OMP_FAIL_MEMORY_ORDER_MASK enumerators.
3909 (OMP_FAIL_MEMORY_ORDER_SHIFT): Define.
3910 * gimple-pretty-print.c (dump_gimple_omp_atomic_load,
3911 dump_gimple_omp_atomic_store): Print [weak] for weak atomic
3913 * gimple.h (enum gf_mask): Change GF_OMP_ATOMIC_MEMORY_ORDER
3914 to 6-bit mask, adjust GF_OMP_ATOMIC_NEED_VALUE value and add
3916 (gimple_omp_atomic_weak_p, gimple_omp_atomic_set_weak): New inline
3918 * tree.h (OMP_ATOMIC_WEAK): Define.
3919 * tree-pretty-print.c (dump_omp_atomic_memory_order): Adjust for
3920 fail memory order being encoded in the same enum and also print
3921 fail clause if present.
3922 (dump_generic_node): Print weak clause if OMP_ATOMIC_WEAK.
3923 * gimplify.c (goa_stabilize_expr): Add target_expr and rhs arguments,
3924 handle pre_p == NULL case as a test mode that only returns value
3925 but doesn't change gimplify nor change anything otherwise, adjust
3926 recursive calls, add MODIFY_EXPR, ADDR_EXPR, COND_EXPR, TARGET_EXPR
3927 and CALL_EXPR handling, adjust COMPOUND_EXPR handling for
3928 __builtin_clear_padding calls, for !rhs gimplify as lvalue rather
3930 (gimplify_omp_atomic): Adjust goa_stabilize_expr caller. Handle
3931 COND_EXPR rhs. Set weak flag on gimple load/store for
3933 * omp-expand.c (omp_memory_order_to_fail_memmodel): New function.
3934 (omp_memory_order_to_memmodel): Adjust for fail clause encoded
3936 (expand_omp_atomic_cas): New function.
3937 (expand_omp_atomic_pipeline): Use omp_memory_order_to_fail_memmodel
3939 (expand_omp_atomic): Attempt to optimize atomic compare and exchange
3940 using expand_omp_atomic_cas.
3942 2021-09-10 Aldy Hernandez <aldyh@redhat.com>
3943 Michael Matz <matz@suse.de>
3945 * tree-pass.h (PROP_loop_opts_done): New.
3946 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
3947 Intersect with global range.
3948 * tree-ssa-loop.c (tree_ssa_loop_done): Set PROP_loop_opts_done.
3949 * tree-ssa-threadbackward.c
3950 (back_threader_profitability::profitable_path_p): Disable
3951 threading through latches until after loop optimizations have run.
3953 2021-09-10 David Faust <david.faust@oracle.com>
3955 * doc/invoke.texi: Document BPF -mcpu, -mjmpext, -mjmp32 and -malu32
3958 2021-09-10 David Faust <david.faust@oracle.com>
3960 * config/bpf/bpf-opts.h (bpf_isa_version): New enum.
3961 * config/bpf/bpf-protos.h (bpf_expand_cbranch): New.
3962 * config/bpf/bpf.c (bpf_option_override): Handle -mcpu option.
3963 (bpf_expand_cbranch): New function.
3964 * config/bpf/bpf.md (AM mode iterator): Conditionalize support for SI
3966 (zero_extendsidi2): Only use mov32 instruction if it is available.
3967 (SIM mode iterator): Conditionalize support for SI mode.
3968 (JM mode iterator): New.
3969 (cbranchdi4): Update name, use new JM iterator. Use bpf_expand_cbranch.
3970 (*branch_on_di): Update name, use new JM iterator.
3971 * config/bpf/bpf.opt: (mjmpext): New option.
3975 (bpf_isa): New enum.
3977 2021-09-10 David Faust <david.faust@oracle.com>
3979 * config/bpf/bpf.md (zero_extendhidi2): Add new output template
3980 for register-to-register extensions.
3981 (zero_extendqidi2): Likewise.
3983 2021-09-10 Richard Biener <rguenther@suse.de>
3985 PR middle-end/102273
3986 * internal-fn.c (expand_DEFERRED_INIT): Always expand non-SSA vars.
3988 2021-09-10 Richard Biener <rguenther@suse.de>
3990 PR middle-end/102269
3991 * gimplify.c (is_var_need_auto_init): Empty types do not need
3994 2021-09-10 Richard Biener <rguenther@suse.de>
3996 * configure.ac (--with-stabs): Remove.
3997 * configure: Regenerate.
3998 * doc/install.texi: Remove --with-stabs documentation.
4000 2021-09-10 liuhongt <hongtao.liu@intel.com>
4002 * config/i386/avx512fp16intrin.h: (_mm512_cmp_ph_mask):
4004 (_mm512_mask_cmp_ph_mask): Likewise.
4005 (_mm512_cmp_round_ph_mask): Likewise.
4006 (_mm512_mask_cmp_round_ph_mask): Likewise.
4007 (_mm_cmp_sh_mask): Likewise.
4008 (_mm_mask_cmp_sh_mask): Likewise.
4009 (_mm_cmp_round_sh_mask): Likewise.
4010 (_mm_mask_cmp_round_sh_mask): Likewise.
4011 (_mm_comieq_sh): Likewise.
4012 (_mm_comilt_sh): Likewise.
4013 (_mm_comile_sh): Likewise.
4014 (_mm_comigt_sh): Likewise.
4015 (_mm_comige_sh): Likewise.
4016 (_mm_comineq_sh): Likewise.
4017 (_mm_ucomieq_sh): Likewise.
4018 (_mm_ucomilt_sh): Likewise.
4019 (_mm_ucomile_sh): Likewise.
4020 (_mm_ucomigt_sh): Likewise.
4021 (_mm_ucomige_sh): Likewise.
4022 (_mm_ucomineq_sh): Likewise.
4023 (_mm_comi_round_sh): Likewise.
4024 (_mm_comi_sh): Likewise.
4025 * config/i386/avx512fp16vlintrin.h (_mm_cmp_ph_mask): New intrinsic.
4026 (_mm_mask_cmp_ph_mask): Likewise.
4027 (_mm256_cmp_ph_mask): Likewise.
4028 (_mm256_mask_cmp_ph_mask): Likewise.
4029 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
4030 * config/i386/i386-builtin.def: Add corresponding new builtins.
4031 * config/i386/i386-expand.c
4032 (ix86_expand_args_builtin): Handle new builtin types.
4033 (ix86_expand_round_builtin): Ditto.
4034 * config/i386/i386.md (ssevecmode): Add HF mode.
4035 (MODEFH): New mode iterator.
4036 * config/i386/sse.md
4037 (V48H_AVX512VL): New mode iterator to support HF vector modes.
4038 Ajdust corresponding description.
4039 (ssecmpintprefix): New.
4040 (VI12_AVX512VL): Adjust to support HF vector modes.
4041 (cmp_imm_predicate): Likewise.
4042 (<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
4044 (avx512f_vmcmp<mode>3<round_saeonly_name>): Likewise.
4045 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Likewise.
4046 (<sse>_<unord>comi<round_saeonly_name>): Likewise.
4048 2021-09-10 liuhongt <hongtao.liu@intel.com>
4050 * config/i386/avx512fp16intrin.h: (_mm512_max_ph): New intrinsic.
4051 (_mm512_mask_max_ph): Likewise.
4052 (_mm512_maskz_max_ph): Likewise.
4053 (_mm512_min_ph): Likewise.
4054 (_mm512_mask_min_ph): Likewise.
4055 (_mm512_maskz_min_ph): Likewise.
4056 (_mm512_max_round_ph): Likewise.
4057 (_mm512_mask_max_round_ph): Likewise.
4058 (_mm512_maskz_max_round_ph): Likewise.
4059 (_mm512_min_round_ph): Likewise.
4060 (_mm512_mask_min_round_ph): Likewise.
4061 (_mm512_maskz_min_round_ph): Likewise.
4062 (_mm_max_sh): Likewise.
4063 (_mm_mask_max_sh): Likewise.
4064 (_mm_maskz_max_sh): Likewise.
4065 (_mm_min_sh): Likewise.
4066 (_mm_mask_min_sh): Likewise.
4067 (_mm_maskz_min_sh): Likewise.
4068 (_mm_max_round_sh): Likewise.
4069 (_mm_mask_max_round_sh): Likewise.
4070 (_mm_maskz_max_round_sh): Likewise.
4071 (_mm_min_round_sh): Likewise.
4072 (_mm_mask_min_round_sh): Likewise.
4073 (_mm_maskz_min_round_sh): Likewise.
4074 * config/i386/avx512fp16vlintrin.h (_mm_max_ph): New intrinsic.
4075 (_mm256_max_ph): Likewise.
4076 (_mm_mask_max_ph): Likewise.
4077 (_mm256_mask_max_ph): Likewise.
4078 (_mm_maskz_max_ph): Likewise.
4079 (_mm256_maskz_max_ph): Likewise.
4080 (_mm_min_ph): Likewise.
4081 (_mm256_min_ph): Likewise.
4082 (_mm_mask_min_ph): Likewise.
4083 (_mm256_mask_min_ph): Likewise.
4084 (_mm_maskz_min_ph): Likewise.
4085 (_mm256_maskz_min_ph): Likewise.
4086 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
4087 * config/i386/i386-builtin.def: Add corresponding new builtins.
4088 * config/i386/i386-expand.c
4089 (ix86_expand_args_builtin): Handle new builtin types.
4090 * config/i386/sse.md
4091 (<code><mode>3<mask_name><round_saeonly_name>): Adjust to
4092 support HF vector modes.
4093 (*<code><mode>3<mask_name><round_saeonly_name>): Likewise.
4094 (ieee_<ieee_maxmin><mode>3<mask_name><round_saeonly_name>):
4096 (<sse>_vm<code><mode>3<mask_scalar_name><round_saeonly_scalar_name>):
4098 * config/i386/subst.md (round_saeonly_mode512bit_condition):
4099 Adjust for HF vector modes.
4101 2021-09-10 Liu, Hongtao <hongtao.liu@intel.com>
4103 * config/i386/avx512fp16intrin.h (_mm_add_sh): New intrinsic.
4104 (_mm_mask_add_sh): Likewise.
4105 (_mm_maskz_add_sh): Likewise.
4106 (_mm_sub_sh): Likewise.
4107 (_mm_mask_sub_sh): Likewise.
4108 (_mm_maskz_sub_sh): Likewise.
4109 (_mm_mul_sh): Likewise.
4110 (_mm_mask_mul_sh): Likewise.
4111 (_mm_maskz_mul_sh): Likewise.
4112 (_mm_div_sh): Likewise.
4113 (_mm_mask_div_sh): Likewise.
4114 (_mm_maskz_div_sh): Likewise.
4115 (_mm_add_round_sh): Likewise.
4116 (_mm_mask_add_round_sh): Likewise.
4117 (_mm_maskz_add_round_sh): Likewise.
4118 (_mm_sub_round_sh): Likewise.
4119 (_mm_mask_sub_round_sh): Likewise.
4120 (_mm_maskz_sub_round_sh): Likewise.
4121 (_mm_mul_round_sh): Likewise.
4122 (_mm_mask_mul_round_sh): Likewise.
4123 (_mm_maskz_mul_round_sh): Likewise.
4124 (_mm_div_round_sh): Likewise.
4125 (_mm_mask_div_round_sh): Likewise.
4126 (_mm_maskz_div_round_sh): Likewise.
4127 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
4128 * config/i386/i386-builtin.def: Add corresponding new builtins.
4129 * config/i386/i386-expand.c
4130 (ix86_expand_round_builtin): Handle new builtins.
4131 * config/i386/sse.md (VF_128): Change description.
4132 (<sse>_vm<plusminus_insn><mode>3<mask_scalar_name><round_scalar_name>):
4133 Adjust to support HF vector modes.
4134 (<sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name><round_scalar_name>):
4137 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
4139 * config/i386/i386-expand.c
4140 (ix86_avx256_split_vector_move_misalign): Handle V16HF mode.
4141 * config/i386/i386.c
4142 (ix86_preferred_simd_mode): Handle HF mode.
4143 * config/i386/sse.md (V_256H): New mode iterator.
4144 (avx_vextractf128<mode>): Use it.
4145 (VEC_INIT_MODE): Align vector HFmode condition to vector
4146 HImodes since there're no real HF instruction used.
4147 (VEC_INIT_HALF_MODE): Ditto.
4149 (VIHF_AVX512BW): Ditto.
4150 (*vec_extracthf): Ditto.
4151 (VEC_EXTRACT_MODE): Ditto.
4153 2021-09-10 Richard Biener <rguenther@suse.de>
4156 * config/dbx.h: Remove.
4157 * config/dbxcoff.h: Do not define PREFERRED_DEBUGGING_TYPE.
4158 * config/lynx.h: Likewise.
4160 2021-09-10 liuhongt <hongtao.liu@intel.com>
4162 * config/i386/i386-expand.c (ix86_expand_copysign): Expand
4163 right into ANDNOT + AND + IOR, using paradoxical subregs.
4164 (ix86_split_copysign_const): Remove.
4165 (ix86_split_copysign_var): Ditto.
4166 * config/i386/i386-protos.h (ix86_split_copysign_const): Dotto.
4167 (ix86_split_copysign_var): Ditto.
4168 * config/i386/i386.md (@copysign<mode>3_const): Ditto.
4169 (@copysign<mode>3_var): Ditto.
4171 2021-09-09 qing zhao <qing.zhao@oracle.com>
4173 * builtins.c (expand_builtin_memset): Make external visible.
4174 * builtins.h (expand_builtin_memset): Declare extern.
4175 * common.opt (ftrivial-auto-var-init=): New option.
4176 * doc/extend.texi: Document the uninitialized attribute.
4177 * doc/invoke.texi: Document -ftrivial-auto-var-init.
4178 * flag-types.h (enum auto_init_type): New enumerated type
4180 * gimple-fold.c (clear_padding_type): Add one new parameter.
4181 (clear_padding_union): Likewise.
4182 (clear_padding_emit_loop): Likewise.
4183 (clear_type_padding_in_mask): Likewise.
4184 (gimple_fold_builtin_clear_padding): Handle this new parameter.
4185 * gimplify.c (gimple_add_init_for_auto_var): New function.
4186 (gimple_add_padding_init_for_auto_var): New function.
4187 (is_var_need_auto_init): New function.
4188 (gimplify_decl_expr): Add initialization to automatic variables per
4190 (gimplify_call_expr): Add one new parameter for call to
4191 __builtin_clear_padding.
4192 (gimplify_init_constructor): Add padding initialization in the end.
4193 * internal-fn.c (INIT_PATTERN_VALUE): New macro.
4194 (expand_DEFERRED_INIT): New function.
4195 * internal-fn.def (DEFERRED_INIT): New internal function.
4196 * tree-cfg.c (verify_gimple_call): Verify calls to .DEFERRED_INIT.
4197 * tree-sra.c (generate_subtree_deferred_init): New function.
4198 (scan_function): Avoid setting cannot_scalarize_away_bitmap for
4199 calls to .DEFERRED_INIT.
4200 (sra_modify_deferred_init): New function.
4201 (sra_modify_function_body): Handle calls to DEFERRED_INIT specially.
4202 * tree-ssa-structalias.c (find_func_aliases_for_call): Likewise.
4203 * tree-ssa-uninit.c (warn_uninit): Handle calls to DEFERRED_INIT
4205 (check_defs): Likewise.
4206 (warn_uninitialized_vars): Likewise.
4207 * tree-ssa.c (ssa_undefined_value_p): Likewise.
4208 * tree.c (build_common_builtin_nodes): Build tree node for
4209 BUILT_IN_CLEAR_PADDING when needed.
4211 2021-09-09 Richard Biener <rguenther@suse.de>
4213 * tree-ssa-loop-im.c (fill_always_executed_in_1): Walk
4216 2021-09-09 Richard Biener <rguenther@suse.de>
4218 * tree-ssa-loop-im.c (fill_always_executed_in_1): Integrate
4219 DOM walk from get_loop_body_in_dom_order using a worklist
4222 2021-09-09 liuhongt <hongtao.liu@intel.com>
4224 * config.gcc: Add avx512fp16vlintrin.h.
4225 * config/i386/avx512fp16intrin.h: (_mm512_add_ph): New intrinsic.
4226 (_mm512_mask_add_ph): Likewise.
4227 (_mm512_maskz_add_ph): Likewise.
4228 (_mm512_sub_ph): Likewise.
4229 (_mm512_mask_sub_ph): Likewise.
4230 (_mm512_maskz_sub_ph): Likewise.
4231 (_mm512_mul_ph): Likewise.
4232 (_mm512_mask_mul_ph): Likewise.
4233 (_mm512_maskz_mul_ph): Likewise.
4234 (_mm512_div_ph): Likewise.
4235 (_mm512_mask_div_ph): Likewise.
4236 (_mm512_maskz_div_ph): Likewise.
4237 (_mm512_add_round_ph): Likewise.
4238 (_mm512_mask_add_round_ph): Likewise.
4239 (_mm512_maskz_add_round_ph): Likewise.
4240 (_mm512_sub_round_ph): Likewise.
4241 (_mm512_mask_sub_round_ph): Likewise.
4242 (_mm512_maskz_sub_round_ph): Likewise.
4243 (_mm512_mul_round_ph): Likewise.
4244 (_mm512_mask_mul_round_ph): Likewise.
4245 (_mm512_maskz_mul_round_ph): Likewise.
4246 (_mm512_div_round_ph): Likewise.
4247 (_mm512_mask_div_round_ph): Likewise.
4248 (_mm512_maskz_div_round_ph): Likewise.
4249 * config/i386/avx512fp16vlintrin.h: New header.
4250 * config/i386/i386-builtin-types.def (V16HF, V8HF, V32HF):
4251 Add new builtin types.
4252 * config/i386/i386-builtin.def: Add corresponding builtins.
4253 * config/i386/i386-expand.c
4254 (ix86_expand_args_builtin): Handle new builtin types.
4255 (ix86_expand_round_builtin): Likewise.
4256 * config/i386/immintrin.h: Include avx512fp16vlintrin.h
4257 * config/i386/sse.md (VFH): New mode_iterator.
4259 (avx512fmaskmode): Add HF vector modes.
4260 (avx512fmaskhalfmode): Likewise.
4261 (<plusminus_insn><mode>3<mask_name><round_name>): Adjust to for
4263 (*<plusminus_insn><mode>3<mask_name><round_name>): Likewise.
4264 (mul<mode>3<mask_name><round_name>): Likewise.
4265 (*mul<mode>3<mask_name><round_name>): Likewise.
4266 (div<mode>3): Likewise.
4267 (<sse>_div<mode>3<mask_name><round_name>): Likewise.
4268 * config/i386/subst.md (SUBST_V): Add HF vector modes.
4269 (SUBST_A): Likewise.
4270 (round_mode512bit_condition): Adjust for V32HFmode.
4272 2021-09-09 liuhongt <hongtao.liu@intel.com>
4275 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
4276 (reduc_plus_scal_v4sf): .. this, New define_expand.
4277 (reduc_plus_scal_v2df): .. and this, New define_expand.
4279 2021-09-09 liuhongt <hongtao.liu@intel.com>
4282 * config/i386/sse.md (*vec_extract<mode><ssescalarmodelower>_valign):
4285 2021-09-08 Jonathan Wakely <jwakely@redhat.com>
4288 * doc/trouble.texi (Copy Assignment): Fix description of
4289 behaviour and fix code in example.
4291 2021-09-08 Segher Boessenkool <segher@kernel.crashing.org>
4294 * config/rs6000/rs6000-logue.c (rs6000_emit_epilogue): For ELFv2 use
4295 r11 instead of r12 for restoring CR.
4297 2021-09-08 Jakub Jelinek <jakub@redhat.com>
4298 liuhongt <hongtao.liu@intel.com>
4301 * config/i386/i386.md (@xorsign<mode>3_1): Remove.
4302 * config/i386/i386-expand.c (ix86_expand_xorsign): Expand right away
4303 into AND with mask and XOR, using paradoxical subregs.
4304 (ix86_split_xorsign): Remove.
4305 * config/i386/i386-protos.h (ix86_split_xorsign): Remove.
4307 2021-09-08 Di Zhao <dizhao@os.amperecomputing.com>
4309 * tree-ssa-sccvn.c (vn_nary_op_insert_into): fix result compare
4311 2021-09-08 Jakub Jelinek <jakub@redhat.com>
4314 * config/i386/i386.md (xorsign<mode>3): If operands[1] is equal to
4315 operands[2], emit abs<mode>2 instead.
4316 (@xorsign<mode>3_1): Add early-clobbers for output operand, enable
4317 first alternative even for avx, add another alternative with
4318 =&Yv <- 0, Yv, Yvm constraints.
4319 * config/i386/i386-expand.c (ix86_split_xorsign): If op0 is equal
4320 to op1, emit vpandn instead.
4322 2021-09-08 liuhongt <hongtao.liu@intel.com>
4324 * config/i386/avx512fp16intrin.h (_mm_set_ph): New intrinsic.
4325 (_mm256_set_ph): Likewise.
4326 (_mm512_set_ph): Likewise.
4327 (_mm_setr_ph): Likewise.
4328 (_mm256_setr_ph): Likewise.
4329 (_mm512_setr_ph): Likewise.
4330 (_mm_set1_ph): Likewise.
4331 (_mm256_set1_ph): Likewise.
4332 (_mm512_set1_ph): Likewise.
4333 (_mm_setzero_ph): Likewise.
4334 (_mm256_setzero_ph): Likewise.
4335 (_mm512_setzero_ph): Likewise.
4336 (_mm_set_sh): Likewise.
4337 (_mm_load_sh): Likewise.
4338 (_mm_store_sh): Likewise.
4339 * config/i386/i386-builtin-types.def (V8HF): New type.
4340 (DEF_FUNCTION_TYPE (V8HF, V8HI)): New builtin function type
4341 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
4342 Support vector HFmodes.
4343 (ix86_expand_vector_init_one_nonzero): Likewise.
4344 (ix86_expand_vector_init_one_var): Likewise.
4345 (ix86_expand_vector_init_interleave): Likewise.
4346 (ix86_expand_vector_init_general): Likewise.
4347 (ix86_expand_vector_set): Likewise.
4348 (ix86_expand_vector_extract): Likewise.
4349 (ix86_expand_vector_init_concat): Likewise.
4350 (ix86_expand_sse_movcc): Handle vector HFmodes.
4351 (ix86_expand_vector_set_var): Ditto.
4352 * config/i386/i386-modes.def: Add HF vector modes in comment.
4353 * config/i386/i386.c (classify_argument): Add HF vector modes.
4354 (ix86_hard_regno_mode_ok): Allow HF vector modes for AVX512FP16.
4355 (ix86_vector_mode_supported_p): Likewise.
4356 (ix86_set_reg_reg_cost): Handle vector HFmode.
4357 (ix86_get_ssemov): Handle vector HFmode.
4358 (function_arg_advance_64): Pass unamed V16HFmode and V32HFmode
4360 (function_arg_advance_32): Pass V8HF/V16HF/V32HF by sse reg for 32bit
4362 (function_arg_advance_32): Ditto.
4363 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): New.
4364 (VALID_AVX256_REG_OR_OI_MODE): Rename to ..
4365 (VALID_AVX256_REG_OR_OI_VHF_MODE): .. this, and add V16HF.
4366 (VALID_SSE2_REG_VHF_MODE): New.
4367 (VALID_AVX512VL_128_REG_MODE): Add V8HF and TImode.
4368 (SSE_REG_MODE_P): Add vector HFmode.
4369 * config/i386/i386.md (mode): Add HF vector modes.
4370 (MODE_SIZE): Likewise.
4371 (ssemodesuffix): Add ph suffix for HF vector modes.
4372 * config/i386/sse.md (VFH_128): New mode iterator.
4373 (VMOVE): Adjust for HF vector modes.
4375 (V_256_512): Likewise.
4377 (avx512fmaskmode): Likewise.
4378 (shuffletype): Likewise.
4379 (sseinsnmode): Likewise.
4380 (ssedoublevecmode): Likewise.
4381 (ssehalfvecmode): Likewise.
4382 (ssehalfvecmodelower): Likewise.
4383 (ssePScmode): Likewise.
4384 (ssescalarmode): Likewise.
4385 (ssescalarmodelower): Likewise.
4386 (sseintprefix): Likewise.
4388 (bcstscalarsuff): Likewise.
4389 (xtg_mode): Likewise.
4390 (VI12HF_AVX512VL): New mode_iterator.
4391 (VF_AVX512FP16): Likewise.
4393 (VIHF_256): Likewise.
4394 (VIHF_AVX512BW): Likewise.
4395 (V16_256): Likewise.
4396 (V32_512): Likewise.
4397 (sseintmodesuffix): New mode_attr.
4398 (sse): Add scalar and vector HFmodes.
4399 (ssescalarmode): Add vector HFmode mapping.
4400 (ssescalarmodesuffix): Add sh suffix for HFmode.
4401 (*<sse>_vm<insn><mode>3): Use VFH_128.
4402 (*<sse>_vm<multdiv_mnemonic><mode>3): Likewise.
4403 (*ieee_<ieee_maxmin><mode>3): Likewise.
4404 (<avx512>_blendm<mode>): New define_insn.
4405 (vec_setv8hf): New define_expand.
4406 (vec_set<mode>_0): New define_insn for HF vector set.
4407 (*avx512fp16_movsh): Likewise.
4408 (avx512fp16_movsh): Likewise.
4409 (vec_extract_lo_v32hi): Rename to ...
4410 (vec_extract_lo_<mode>): ... this, and adjust to allow HF
4412 (vec_extract_hi_v32hi): Likewise.
4413 (vec_extract_hi_<mode>): Likewise.
4414 (vec_extract_lo_v16hi): Likewise.
4415 (vec_extract_lo_<mode>): Likewise.
4416 (vec_extract_hi_v16hi): Likewise.
4417 (vec_extract_hi_<mode>): Likewise.
4418 (vec_set_hi_v16hi): Likewise.
4419 (vec_set_hi_<mode>): Likewise.
4420 (vec_set_lo_v16hi): Likewise.
4421 (vec_set_lo_<mode>): Likewise.
4422 (*vec_extract<mode>_0): New define_insn_and_split for HF
4424 (*vec_extracthf): New define_insn.
4425 (VEC_EXTRACT_MODE): Add HF vector modes.
4426 (PINSR_MODE): Add V8HF.
4427 (sse2p4_1): Likewise.
4428 (pinsr_evex_isa): Likewise.
4429 (<sse2p4_1>_pinsr<ssemodesuffix>): Adjust to support
4430 insert for V8HFmode.
4431 (pbroadcast_evex_isa): Add HF vector modes.
4432 (AVX2_VEC_DUP_MODE): Likewise.
4433 (VEC_INIT_MODE): Likewise.
4434 (VEC_INIT_HALF_MODE): Likewise.
4435 (avx2_pbroadcast<mode>): Adjust to support HF vector mode
4437 (avx2_pbroadcast<mode>_1): Likewise.
4438 (<avx512>_vec_dup<mode>_1): Likewise.
4439 (<avx512>_vec_dup<mode><mask_name>): Likewise.
4440 (<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>):
4443 2021-09-08 Guo, Xuepeng <xuepeng.guo@intel.com>
4444 H.J. Lu <hongjiu.lu@intel.com>
4445 Liu Hongtao <hongtao.liu@intel.com>
4446 Wang Hongyu <hongyu.wang@intel.com>
4447 Xu Dianhong <dianhong.xu@intel.com>
4449 * common/config/i386/cpuinfo.h (get_available_features):
4450 Detect FEATURE_AVX512FP16.
4451 * common/config/i386/i386-common.c
4452 (OPTION_MASK_ISA_AVX512FP16_SET,
4453 OPTION_MASK_ISA_AVX512FP16_UNSET,
4454 OPTION_MASK_ISA2_AVX512FP16_SET,
4455 OPTION_MASK_ISA2_AVX512FP16_UNSET): New.
4456 (OPTION_MASK_ISA2_AVX512BW_UNSET,
4457 OPTION_MASK_ISA2_AVX512BF16_UNSET): Add AVX512FP16.
4458 (ix86_handle_option): Handle -mavx512fp16.
4459 * common/config/i386/i386-cpuinfo.h (enum processor_features):
4460 Add FEATURE_AVX512FP16.
4461 * common/config/i386/i386-isas.h: Add entry for AVX512FP16.
4462 * config.gcc: Add avx512fp16intrin.h.
4463 * config/i386/avx512fp16intrin.h: New intrinsic header.
4464 * config/i386/cpuid.h: Add bit_AVX512FP16.
4465 * config/i386/i386-builtin-types.def: (FLOAT16): New primitive type.
4466 * config/i386/i386-builtins.c: Support _Float16 type for i386
4468 (ix86_register_float16_builtin_type): New function.
4469 (ix86_float16_type_node): New.
4470 * config/i386/i386-c.c (ix86_target_macros_internal): Define
4472 * config/i386/i386-expand.c (ix86_expand_branch): Support
4474 (ix86_prepare_fp_compare_args): Adjust TARGET_SSE_MATH &&
4475 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
4476 (ix86_expand_fp_movcc): Ditto.
4477 * config/i386/i386-isa.def: Add PTA define for AVX512FP16.
4478 * config/i386/i386-options.c (isa2_opts): Add -mavx512fp16.
4479 (ix86_valid_target_attribute_inner_p): Add avx512fp16 attribute.
4480 * config/i386/i386.c (ix86_get_ssemov): Use
4481 vmovdqu16/vmovw/vmovsh for HFmode/HImode scalar or vector.
4482 (ix86_get_excess_precision): Use
4483 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when TARGET_AVX512FP16
4485 (sse_store_index): Use SFmode cost for HFmode cost.
4486 (inline_memory_move_cost): Add HFmode, and perfer SSE cost over
4487 GPR cost for HFmode.
4488 (ix86_hard_regno_mode_ok): Allow HImode in sse register.
4489 (ix86_mangle_type): Add manlging for _Float16 type.
4490 (inline_secondary_memory_needed): No memory is needed for
4491 16bit movement between gpr and sse reg under
4493 (ix86_multiplication_cost): Adjust TARGET_SSE_MATH &&
4494 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
4495 (ix86_division_cost): Ditto.
4496 (ix86_rtx_costs): Ditto.
4497 (ix86_add_stmt_cost): Ditto.
4498 (ix86_optab_supported_p): Ditto.
4499 * config/i386/i386.h (VALID_AVX512F_SCALAR_MODE): Add HFmode.
4500 (SSE_FLOAT_MODE_SSEMATH_OR_HF_P): Add HFmode.
4501 (PTA_SAPPHIRERAPIDS): Add PTA_AVX512FP16.
4502 * config/i386/i386.md (mode): Add HFmode.
4503 (MODE_SIZE): Add HFmode.
4504 (isa): Add avx512fp16.
4505 (enabled): Handle avx512fp16.
4506 (ssemodesuffix): Add sh suffix for HFmode.
4507 (comm): Add mult, div.
4508 (plusminusmultdiv): New code iterator.
4509 (insn): Add mult, div.
4510 (*movhf_internal): Adjust for avx512fp16 instruction.
4511 (*movhi_internal): Ditto.
4512 (*cmpi<unord>hf): New define_insn for HFmode.
4513 (*ieee_s<ieee_maxmin>hf3): Likewise.
4514 (extendhf<mode>2): Likewise.
4515 (trunc<mode>hf2): Likewise.
4516 (float<floatunssuffix><mode>hf2): Likewise.
4517 (*<insn>hf): Likewise.
4518 (cbranchhf4): New expander.
4519 (movhfcc): Likewise.
4520 (<insn>hf3): Likewise.
4523 * config/i386/i386.opt: Add mavx512fp16.
4524 * config/i386/immintrin.h: Include avx512fp16intrin.h.
4525 * doc/invoke.texi: Add mavx512fp16.
4526 * doc/extend.texi: Add avx512fp16 Usage Notes.
4528 2021-09-08 liuhongt <hongtao.liu@intel.com>
4530 * common.opt: Support -fexcess-precision=16.
4531 * config/aarch64/aarch64.c (aarch64_excess_precision): Return
4532 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when
4533 EXCESS_PRECISION_TYPE_FLOAT16.
4534 * config/arm/arm.c (arm_excess_precision): Ditto.
4535 * config/i386/i386.c (ix86_get_excess_precision): Ditto.
4536 * config/m68k/m68k.c (m68k_excess_precision): Issue an error
4537 when EXCESS_PRECISION_TYPE_FLOAT16.
4538 * config/s390/s390.c (s390_excess_precision): Ditto.
4539 * coretypes.h (enum excess_precision_type): Add
4540 EXCESS_PRECISION_TYPE_FLOAT16.
4541 * doc/tm.texi (TARGET_C_EXCESS_PRECISION): Update documents.
4542 * doc/tm.texi.in (TARGET_C_EXCESS_PRECISION): Ditto.
4543 * doc/extend.texi (Half-Precision): Document
4544 -fexcess-precision=16.
4545 * flag-types.h (enum excess_precision): Add
4546 EXCESS_PRECISION_FLOAT16.
4547 * target.def (excess_precision): Update document.
4548 * tree.c (excess_precision_type): Set excess_precision_type to
4549 EXCESS_PRECISION_FLOAT16 when -fexcess-precision=16.
4551 2021-09-08 liuhongt <hongtao.liu@intel.com>
4553 * doc/extend.texi: (@node Floating Types): Adjust the wording.
4554 (@node Half-Precision): Ditto.
4556 2021-09-07 Takayuki 'January June' Suwa <jjsuwa_sys3175@yahoo.co.jp>
4559 * config/xtensa/xtensa.c (xtensa_emit_move_sequence): Add
4560 'CONST_INT_P (src)' to the condition of the block that tries to
4561 eliminate literal when loading integer contant.
4563 2021-09-07 David Faust <david.faust@oracle.com>
4565 * doc/extend.texi (BPF Type Attributes) New node.
4566 Document new preserve_access_index attribute.
4567 Document new preserve_access_index builtin.
4568 * doc/invoke.texi: Document -mco-re and -mno-co-re options.
4570 2021-09-07 David Faust <david.faust@oracle.com>
4572 * config/bpf/bpf.c: Adjust includes.
4573 (bpf_handle_preserve_access_index_attribute): New function.
4574 (bpf_attribute_table): Use it here.
4575 (bpf_builtins): Add BPF_BUILTIN_PRESERVE_ACCESS_INDEX.
4576 (bpf_option_override): Handle "-mco-re" option.
4577 (bpf_asm_init_sections): New.
4578 (TARGET_ASM_INIT_SECTIONS): Redefine.
4579 (bpf_file_end): New.
4580 (TARGET_ASM_FILE_END): Redefine.
4581 (bpf_init_builtins): Add "__builtin_preserve_access_index".
4582 (bpf_core_compute, bpf_core_get_index): New.
4583 (is_attr_preserve_access): New.
4584 (bpf_expand_builtin): Handle new builtins.
4585 (bpf_core_newdecl, bpf_core_is_maybe_aggregate_access): New.
4586 (bpf_core_walk): New.
4587 (bpf_resolve_overloaded_builtin): New.
4588 (TARGET_RESOLVE_OVERLOADED_BUILTIN): Redefine.
4590 (pass_bpf_core_attr): New RTL pass.
4591 * config/bpf/bpf-passes.def: New file.
4592 * config/bpf/bpf-protos.h (make_pass_bpf_core_attr): New.
4593 * config/bpf/coreout.c: New file.
4594 * config/bpf/coreout.h: Likewise.
4595 * config/bpf/t-bpf (TM_H): Add $(srcdir)/config/bpf/coreout.h.
4596 (coreout.o): New rule.
4597 (PASSES_EXTRA): Add $(srcdir)/config/bpf/bpf-passes.def.
4598 * config.gcc (bpf): Add coreout.h to extra_headers.
4599 Add coreout.o to extra_objs.
4600 Add $(srcdir)/config/bpf/coreout.c to target_gtfiles.
4602 2021-09-07 David Faust <david.faust@oracle.com>
4604 * btfout.c (get_btf_id): Function is no longer static.
4605 * ctfc.h: Expose it here.
4607 2021-09-07 David Faust <david.faust@oracle.com>
4609 * ctfc.c (ctf_lookup_tree_type): New function.
4612 2021-09-07 David Faust <david.faust@oracle.com>
4614 * ctfc.c (ctf_dtd_lookup): Function is no longer static.
4615 * ctfc.h: Analogous change.
4617 2021-09-07 David Faust <david.faust@oracle.com>
4619 * dwarf2out.c (lookup_type_die): Function is no longer static.
4620 * dwarf2out.h: Expose it here.
4622 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
4624 * dwarf2ctf.c (ctf_debug_finalize): Make it static.
4625 (ctf_debug_early_finish): New definition.
4626 (ctf_debug_finish): Likewise.
4627 * dwarf2ctf.h (ctf_debug_finalize): Remove declaration.
4628 (ctf_debug_early_finish): New declaration.
4629 (ctf_debug_finish): Likewise.
4630 * dwarf2out.c (dwarf2out_finish): Invoke ctf_debug_finish.
4631 (dwarf2out_early_finish): Invoke ctf_debug_early_finish.
4633 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
4635 * config/bpf/bpf.c (bpf_option_override): For BPF backend, disable LTO
4636 support when compiling for CO-RE.
4637 * config/bpf/bpf.opt: Add new command line option -mco-re.
4639 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
4641 * flag-types.h (enum debug_info_type): Add new enum
4642 DINFO_TYPE_BTF_WITH_CORE.
4643 (BTF_WITH_CORE_DEBUG): New bitmask.
4644 * flags.h (btf_with_core_debuginfo_p): New declaration.
4645 * opts.c (btf_with_core_debuginfo_p): New definition.
4647 2021-09-07 Jason Merrill <jason@redhat.com>
4649 * tree.h (error_operand_p): Change to inline function.
4651 2021-09-07 Aldy Hernandez <aldyh@redhat.com>
4653 * tree-ssa-threadedge.c (forwarder_block_p): Rename to...
4654 (empty_block_with_phis_p): ...this.
4655 (potentially_threadable_block): Same.
4656 (jump_threader::thread_through_normal_block): Same.
4658 2021-09-07 Eric Botcazou <ebotcazou@adacore.com>
4661 * dwarf2out.c (mark_base_types): New overloaded function.
4662 (dwarf2out_early_finish): Invoke it on the COMDAT type list as well
4663 as the compilation unit, and call move_marked_base_types afterward.
4665 2021-09-07 H.J. Lu <hjl.tools@gmail.com>
4668 * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
4670 (ix86_expand_vector_convert_uns_vsivsf): Likewise.
4672 2021-09-07 Richard Biener <rguenther@suse.de>
4674 PR tree-optimization/102226
4675 * tree-vect-loop.c (vect_transform_cycle_phi): Record
4676 the converted value for the epilogue PHI use.
4678 2021-09-07 Martin Liska <mliska@suse.cz>
4680 PR gcov-profile/80223
4681 * ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
4682 options, do not inline when no_profile_instrument_function
4683 attributes are different in early inliner. It's fine to inline
4684 it after PGO instrumentation.
4686 2021-09-07 Richard Biener <rguenther@suse.de>
4688 PR tree-optimization/101555
4689 * tree-ssa-pre.c (translate_vuse_through_block): Do not
4690 perform an alias walk to determine the validity of the
4691 mem at the start of the block which is already guaranteed
4692 by means of prune_clobbered_mems.
4693 (phi_translate_1): Pass edge to translate_vuse_through_block.
4695 2021-09-07 Xionghu Luo <luoxhu@linux.ibm.com>
4698 * config/rs6000/rs6000.md (fmod<mode>3): New define_expand.
4699 (remainder<mode>3): Likewise.
4701 2021-09-07 YunQiang Su <yunqiang.su@cipunited.com>
4703 * config/mips/mips.c (mips_file_start): add .module for
4706 2021-09-06 Roger Sayle <roger@nextmovesoftware.com>
4708 * wide-int.cc (wi::clz): Reorder tests to ensure the result
4709 is zero for all negative values.
4711 2021-09-06 Tobias Burnus <tobias@codesourcery.com>
4713 * doc/invoke.texi (-foffload-options): Fix @opindex.
4715 2021-09-06 H.J. Lu <hjl.tools@gmail.com>
4718 * config/i386/i386-expand.c (ix86_split_xorsign): Use operands[2].
4719 * config/i386/i386.md (@xorsign<mode>3_1): Add non-destructive
4720 source alternative for AVX.
4722 2021-09-06 liuhongt <hongtao.liu@intel.com>
4724 PR middle-end/102182
4725 * optabs.c (expand_fix): Add from1 to avoid from being
4728 2021-09-06 Eric Botcazou <ebotcazou@adacore.com>
4730 * dwarf2out.c (modified_type_die): Deal with all array types earlier
4731 and use local variable consistently throughout the function.
4733 2021-09-06 Jakub Jelinek <jakub@redhat.com>
4735 PR tree-optimization/102207
4736 * match.pd: Don't demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
4737 were promoted from signed to wider unsigned type.
4739 2021-09-06 Andrew Pinski <apinski@marvell.com>
4741 PR tree-optimization/63184
4742 * match.pd: Add simplification of pointer_diff of two pointer_plus
4743 with addr_expr in the first operand of each pointer_plus.
4744 Add simplificatoin of ne/eq of two pointer_plus with addr_expr
4745 in the first operand of each pointer_plus.
4747 2021-09-06 Richard Biener <rguenther@suse.de>
4749 PR tree-optimization/102176
4750 * tree-vect-slp.c (vect_slp_gather_vectorized_scalar_stmts):
4752 (vect_bb_slp_scalar_cost): Use the computed set of
4753 vectorized scalar stmts instead of relying on the out-of-date
4754 and not accurate PURE_SLP_STMT.
4755 (vect_bb_vectorization_profitable_p): Compute the set
4756 of vectorized scalar stmts.
4758 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
4760 * gimple-range-path.cc (path_range_query::range_of_stmt): Remove
4761 GIMPLE_COND special casing.
4762 (path_range_query::range_defined_in_block): Use range_of_stmt
4763 instead of calling fold_range directly.
4765 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
4767 * gimple-range-path.cc (path_range_query::range_of_expr): Set
4768 m_undefined_path when appropriate.
4769 (path_range_query::internal_range_of_expr): Copy from range_of_expr.
4770 (path_range_query::unreachable_path_p): New.
4771 (path_range_query::precompute_ranges): Set m_undefined_path.
4772 * gimple-range-path.h (path_range_query::unreachable_path_p): New.
4773 (path_range_query::internal_range_of_expr): New.
4774 * tree-ssa-threadbackward.c (back_threader::find_taken_edge_cond):
4775 Use unreachable_path_p.
4777 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
4779 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
4780 Remove argument and call find_taken_edge.
4781 (back_threader::resolve_phi): Do not calculate taken edge before
4782 calling maybe_register_path.
4783 (back_threader::find_paths_to_names): Same.
4785 2021-09-05 Jeff Law <jlaw@localhost.localdomain>
4787 * config/h8300/h8300.md (QHSI2 mode iterator): New mode iterator.
4788 * config/h8300/testcompare.md (store_c): Update name, use new
4790 (store_neg_c, store_shifted_c): New patterns.
4792 2021-09-03 Segher Boessenkool <segher@kernel.crashing.org>
4795 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): On ELFv2 use r11
4796 instead of r12 for CR save, in all cases.
4798 2021-09-03 Andrew Pinski <apinski@marvell.com>
4800 * config/aarch64/aarch64-sve-builtins.cc (register_vector_type):
4801 Handle error_mark_node as the type of the type_decl.
4803 2021-09-03 Andrew Pinski <apinski@marvell.com>
4805 * config/aarch64/aarch64-builtins.c (struct aarch64_simd_type_info):
4807 (aarch64_simd_types): Likewise.
4808 (aarch64_simd_intOI_type_node): Likewise.
4809 (aarch64_simd_intCI_type_node): Likewise.
4810 (aarch64_simd_intXI_type_node): Likewise.
4811 * config/aarch64/aarch64.h (aarch64_fp16_type_node): Likewise.
4812 (aarch64_fp16_ptr_type_node): Likewise.
4813 (aarch64_bf16_type_node): Likewise.
4814 (aarch64_bf16_ptr_type_node): Likewise.
4816 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4818 * range-op.cc (operator_minus::op1_op2_relation_effect): Abstract
4820 (minus_op1_op2_relation_effect): ...here.
4821 (class operator_pointer_diff): New.
4822 (operator_pointer_diff::op1_op2_relation_effect): Call
4823 minus_op1_op2_relation_effect.
4824 (integral_table::integral_table): Add entry for POINTER_DIFF_EXPR.
4826 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4828 * tree-ssa-threadbackward.c (back_threader::thread_through_all_blocks):
4829 Add may_peel_loop_headers.
4830 (back_threader_registry::thread_through_all_blocks): Same.
4831 (try_thread_blocks): Pass may_peel_loop_headers argument.
4832 (pass_early_thread_jumps::execute): Same.
4834 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4836 * tree-ssa-threadedge.c (has_phis_p): New.
4837 (forwarder_block_p): New.
4838 (potentially_threadable_block): Call forwarder_block_p.
4839 (jump_threader::thread_around_empty_blocks): Call has_phis_p.
4840 (jump_threader::thread_through_normal_block): Call
4843 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4845 * tree-ssa-threadbackward.c (back_threader::dump): New.
4846 (back_threader::debug): New.
4847 (back_threader_profitability::profitable_path_p): Dump blocks
4848 even if we are bailing early.
4850 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4852 * tree-ssa-threadupdate.c (cancel_thread): New.
4853 (jump_thread_path_registry::thread_block_1): Use cancel_thread.
4854 (jump_thread_path_registry::mark_threaded_blocks): Same.
4855 (jump_thread_path_registry::register_jump_thread): Same.
4857 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4859 * tree-ssa-threadedge.c (jt_state::push): Only call methods for
4860 which objects are available.
4861 (jt_state::pop): Same.
4862 (jt_state::register_equiv): Same.
4863 (jt_state::register_equivs_on_edge): Same.
4865 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4867 * tree-ssa-threadedge.c (jump_threader::thread_across_edge):
4868 Move pop until after a thread is registered.
4870 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4872 * tree-ssa-threadupdate.c (debug): New.
4874 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4876 * gimple-range-trace.cc (push_dump_file::push_dump_file): New.
4877 (push_dump_file::~push_dump_file): New.
4878 (dump_ranger): Change dump_file temporarily while dumping
4880 * gimple-range-trace.h (class push_dump_file): New.
4882 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4884 * gimple-range-trace.cc (debug_seed_ranger): Remove static.
4885 (dump_ranger): Dump function name.
4887 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4889 * gimple-range-path.cc (path_range_query::range_defined_in_block):
4890 Adjust for non-null.
4891 (path_range_query::adjust_for_non_null_uses): New.
4892 (path_range_query::precompute_ranges): Call
4893 adjust_for_non_null_uses.
4894 * gimple-range-path.h: Add m_non_null and
4895 adjust_for_non_null_uses.
4897 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4899 * gimple-range-path.cc (path_range_query::dump): Dump path
4901 (path_range_query::precompute_ranges): Dump entire path.
4903 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4905 * value-relation.cc (relation_oracle::debug): New.
4906 * value-relation.h (relation_oracle::debug): New.
4908 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4910 * tree-ssa-loop-ch.c: Remove unnecessary include file.
4912 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4914 * gimple-range-fold.cc (fold_using_range::postfold_gcond_edges):
4915 Skip statements with no defining BB.
4916 * gimple-range-path.cc (path_range_query::range_defined_in_block):
4917 Do not get confused by statements with no defining BB.
4919 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4921 * gimple-range-fold.cc (adjust_imagpart_expr): Move from
4922 gimple_range_adjustment. Add support for constants.
4923 (adjust_realpart_expr): New.
4924 (gimple_range_adjustment): Move IMAGPART_EXPR code to
4925 adjust_imagpart_expr.
4926 * range-op.cc (integral_table::integral_table): Add entry for
4929 2021-09-03 Jakub Jelinek <jakub@redhat.com>
4931 * omp-expand.c (expand_omp_atomic_pipeline): Use
4932 IFN_ATOMIC_COMPARE_EXCHANGE instead of
4933 BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_? so that memory order
4936 2021-09-03 Jakub Jelinek <jakub@redhat.com>
4939 * tree.h (DECL_FIELD_ABI_IGNORED): Changed into rvalue only macro
4940 that is false if DECL_BIT_FIELD.
4941 (SET_DECL_FIELD_ABI_IGNORED, DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD,
4942 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD): Define.
4943 * tree-streamer-out.c (pack_ts_decl_common_value_fields): For
4944 DECL_BIT_FIELD stream DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead
4945 of DECL_FIELD_ABI_IGNORED.
4946 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
4947 SET_DECL_FIELD_ABI_IGNORED instead of writing to
4948 DECL_FIELD_ABI_IGNORED and for DECL_BIT_FIELD use
4949 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead.
4950 * lto-streamer-out.c (hash_tree): For DECL_BIT_FIELD hash
4951 DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead of DECL_FIELD_ABI_IGNORED.
4953 2021-09-03 liuhongt <hongtao.liu@intel.com>
4956 * config/i386/amxbf16intrin.h : Remove macro check for __AMX_BF16__.
4957 * config/i386/amxint8intrin.h : Remove macro check for __AMX_INT8__.
4958 * config/i386/amxtileintrin.h : Remove macro check for __AMX_TILE__.
4960 2021-09-02 Martin Sebor <msebor@redhat.com>
4962 PR tree-optimization/17506
4964 * tree-ssa-uninit.c (warn_uninit): Remove conditional guarding note.
4966 2021-09-02 Richard Biener <rguenther@suse.de>
4968 * tree-ssa-loop-im.c (fill_always_executed_in_1): Refine
4969 fix for PR78185 and continue processing when leaving
4972 2021-09-02 Jakub Jelinek <jakub@redhat.com>
4974 PR tree-optimization/99591
4975 * match.pd: Demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
4978 2021-09-02 Richard Biener <rguenther@suse.de>
4981 2021-09-02 Richard Biener <rguenther@suse.de>
4983 PR tree-optimization/102155
4984 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
4985 over a part of the RPO array and do not recurse here.
4986 Dump blocks marked as always executed.
4987 (fill_always_executed_in): Walk over the RPO array and
4988 process loops whose header we run into.
4989 (loop_invariant_motion_in_fun): Compute the first RPO
4990 using rev_post_order_and_mark_dfs_back_seme in iteration
4991 order and pass that to fill_always_executed_in.
4993 2021-09-02 liuhongt <hongtao.liu@intel.com>
4995 * config/i386/i386-modes.def (FLOAT_MODE): Define ieee HFmode.
4996 * config/i386/i386.c (enum x86_64_reg_class): Add
4998 (merge_classes): Handle X86_64_SSEHF_CLASS.
4999 (examine_argument): Ditto.
5000 (construct_container): Ditto.
5001 (classify_argument): Ditto, and set HFmode/HCmode to
5003 (function_value_32): Return _FLoat16/Complex Float16 by
5005 (function_value_64): Return _Float16/Complex Float16 by SSE
5007 (ix86_print_operand): Handle CONST_DOUBLE HFmode.
5008 (ix86_secondary_reload): Require gpr as intermediate register
5009 to store _Float16 from sse register when sse4 is not
5011 (ix86_libgcc_floating_mode_supported_p): Enable _FLoat16 under
5013 (ix86_scalar_mode_supported_p): Ditto.
5014 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Defined.
5015 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add HFmode.
5016 (VALID_INT_MODE_P): Add HFmode and HCmode.
5017 * config/i386/i386.md (*pushhf_rex64): New define_insn.
5019 (*movhf_internal): Ditto.
5020 * doc/extend.texi (Half-Precision Floating Point): Documemt
5023 2021-09-02 Richard Biener <rguenther@suse.de>
5025 PR tree-optimization/102155
5026 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
5027 over a part of the RPO array and do not recurse here.
5028 Dump blocks marked as always executed.
5029 (fill_always_executed_in): Walk over the RPO array and
5030 process loops whose header we run into.
5031 (loop_invariant_motion_in_fun): Compute the first RPO
5032 using rev_post_order_and_mark_dfs_back_seme in iteration
5033 order and pass that to fill_always_executed_in.
5035 2021-09-02 YunQiang Su <syq@debian.org>
5038 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
5040 * config/mips/mips.c (mips_module_isa_name): New.
5041 mips_file_start: add .module mipsREV to all asm output
5043 2021-09-01 Jeff Law <jlaw@localhost.localdomain>
5045 PR tree-optimization/102152
5046 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Reduce a vector
5047 comparison to a scalar comparison before calling
5048 update_stmt_if_modified.
5050 2021-09-01 Andrew Pinski <apinski@marvell.com>
5053 * config/aarch64/aarch64.c (aarch64_expand_setmem):
5054 Check STRICT_ALIGNMENT before creating an overlapping
5057 2021-09-01 Martin Sebor <msebor@redhat.com>
5059 * gimple-ssa-warn-access.cc (get_size_range): Add argument.
5060 (check_access): Pass additional argument.
5061 (check_memop_access): Remove template and make a member function.
5062 (maybe_check_dealloc_call): Make a pass_waccess member function.
5063 (class pass_waccess): Add, rename, and remove members.
5064 (pass_waccess::pass_waccess): Adjust to name change.
5065 (pass_waccess::~pass_waccess): Same.
5066 (check_alloca): Make a member function.
5067 (check_alloc_size_call): Same.
5068 (check_strcat): Same.
5069 (check_strncat): Same.
5070 (check_stxcpy): Same.
5071 (check_stxncpy): Same.
5072 (check_strncmp): Same.
5073 (maybe_warn_rdwr_sizes): Rename...
5074 (pass_waccess::maybe_check_access_sizes): ...to this.
5075 (pass_waccess::check_call): Adjust to name changes.
5076 (pass_waccess::maybe_check_dealloc_call): Make a pass_waccess member
5078 (pass_waccess::execute): Adjust to name changes.
5079 * gimple-ssa-warn-access.h (check_memop_access): Remove.
5080 * pointer-query.cc (access_ref::phi): Handle null pointer.
5081 (access_ref::inform_access): Same.
5082 (pointer_query::put_ref): Modify a cached value, not a copy of it.
5083 (pointer_query::dump): New function.
5084 (compute_objsize_r): Avoid overwriting access_ref::bndrng. Cache
5086 * pointer-query.h (pointer_query::dump): Declare.
5087 * tree-ssa-strlen.c (get_range): Simplify. Use function query.
5088 (dump_strlen_info): Use function query.
5089 (printf_strlen_execute): Factor code out into pointer_query::put_ref.
5091 2021-09-01 Thomas Schwinge <thomas@codesourcery.com>
5093 * tree.c (walk_tree_1) <OMP_CLAUSE>: Simplify.
5095 2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
5097 * doc/extend.texi: Document unavailable attribute.
5098 * print-tree.c (print_node): Handle unavailable attribute.
5099 * tree-core.h (struct tree_base): Add a bit to carry unavailability.
5100 * tree.c (error_unavailable_use): New.
5101 * tree.h (TREE_UNAVAILABLE): New.
5102 (error_unavailable_use): New.
5104 2021-09-01 Jakub Jelinek <jakub@redhat.com>
5106 PR tree-optimization/102124
5107 * tree-vect-patterns.c (vect_recog_widen_op_pattern): For ORIG_CODE
5108 MINUS_EXPR, if itype is unsigned with smaller precision than type,
5109 add an extra cast to signed variant of itype to ensure sign-extension.
5111 2021-09-01 Martin Liska <mliska@suse.cz>
5113 * graph.c (draw_cfg_node_succ_edges): Do not color fallthru
5114 edges and rather use colors for TRUE and FALSE edges.
5116 2021-09-01 Richard Biener <rguenther@suse.de>
5118 PR tree-optimization/93491
5119 * tree-ssa-pre.c (compute_avail): Set BB_MAY_NOTRETURN
5120 after processing the stmt itself. Do not consider
5121 pure functions possibly not returning. Properly avoid
5122 adding possibly trapping calls to EXP_GEN when there's
5123 a preceeding possibly not returning call.
5124 * tree-ssa-sccvn.c (vn_reference_may_trap): Conservatively
5127 2021-09-01 Richard Biener <rguenther@suse.de>
5129 PR tree-optimization/102139
5130 * tree-vectorizer.h (vec_base_alignments): Adjust hash-map
5131 type to record a std::pair of the stmt-info and the innermost
5133 (dr_vec_info::group): New member.
5134 * tree-vect-data-refs.c (vect_record_base_alignment): Adjust.
5135 (vect_compute_data_ref_alignment): Verify the recorded
5136 base alignment can be used.
5137 (data_ref_pair): Remove.
5138 (dr_group_sort_cmp): Adjust.
5139 (vect_analyze_data_ref_accesses): Store the group-ID in the
5140 dr_vec_info and operate on a vector of dr_vec_infos.
5142 2021-09-01 YunQiang Su <yunqiang.su@cipunited.com>
5144 * read-md.c (md_reader::handle_enum): support value assignation.
5145 * doc/md.texi: record define_c_enum value assignation support.
5147 2021-09-01 Jakub Jelinek <jakub@redhat.com>
5149 PR tree-optimization/102141
5150 * gimple-ssa-store-merging.c (bswap_view_convert): Add BEFORE
5151 argument. If false, emit stmts after gsi instead of before, and
5153 (bswap_replace): Adjust callers. When converting output of bswap,
5154 emit VIEW_CONVERT prepratation stmts after a copy of gsi instead
5157 2021-09-01 liuhongt <hongtao.liu@intel.com>
5159 * emit-rtl.c (validate_subreg): Get rid of all float-int
5162 2021-09-01 liuhongt <hongtao.liu@intel.com>
5165 2021-08-30 liuhongt <hongtao.liu@intel.com>
5167 * expmed.c (extract_bit_field_1): Make sure we're playing with
5168 integral modes before call extract_integral_bit_field.
5169 (extract_integral_bit_field): Add a parameter of type
5170 scalar_int_mode which corresponds to of tmode.
5171 And call extract_and_convert_fixed_bit_field instead of
5172 extract_fixed_bit_field and convert_extracted_bit_field.
5173 (extract_and_convert_fixed_bit_field): New function, it's a
5174 combination of extract_fixed_bit_field and
5175 convert_extracted_bit_field.
5177 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
5179 * tree.c (walk_tree_1) <OMP_CLAUSE_TILE>: Handle three operands.
5181 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
5183 * omp-general.h (omp_is_reference): Rename to...
5184 (omp_privatize_by_reference): ... this. Adjust all users...
5185 * omp-general.c: ... here, ...
5186 * gimplify.c: ... here, ...
5187 * omp-expand.c: ... here, ...
5188 * omp-low.c: ... here.
5190 2021-08-31 Martin Sebor <msebor@redhat.com>
5192 * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Test
5193 pointer element for equality to zero, not that of the cotaining
5196 2021-08-31 Martin Sebor <msebor@redhat.com>
5198 * gcc-rich-location.h (gcc_rich_location): Make ctor explicit.
5200 2021-08-31 Martin Sebor <msebor@redhat.com>
5202 * function.h (function): Add comments.
5203 (get_range_query): Same. Add attribute returns nonnull.
5205 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
5207 * expr.c (convert_modes): Don't use subreg_promoted_mode on a
5208 SUBREG if it can't be guaranteed to a SUBREG_PROMOTED_VAR_P set.
5209 Instead use the standard (safer) is_a <scalar_int_mode> idiom.
5211 2021-08-31 Jeff Law <jlaw@localhost.localdomain>
5213 * config.gcc (cris-*-elf, cris-*-none): Remove dbxelf.h from
5215 (m32r-*-elf, m32rle-*-elf, m32r-*-linux): Likewise.
5216 (mn10300-*-*, am33_2.0-*-linux*): Likewise.
5217 (xtensa*-*-elf, xtensa*-*-linux, xtensa*-*-uclinux): Likewise.
5218 (m32c-*-elf*, m32c-*-rtems*): Likewise.
5219 * config/cris/cris.h (DBX_NO_XREFS): Remove.
5220 (DBX_CONTIN_LENGTH, DBX_CONTIN_CHAR): Likewise.
5221 * config/m32r/m32r.h (DBXOUT_SOURCE_LINE): Likewise.
5222 (DBX_DEBUGGING_INFO, DBX_CONTIN_LENGTH): Likewise.
5223 * config/mn10300/mn10300.h (DEFAULT_GDB_EXTENSIONS): Likewise.
5224 * config/mn10300/linux.h (DBX_REGISTER_NAMES): Likewise.
5226 2021-08-31 Marcel Vollweiler <marcel@codesourcery.com>
5228 * gimplify.c (gimplify_scan_omp_clauses): Error handling. 'ancestor' only
5229 allowed on target constructs and only with particular other clauses.
5230 * omp-expand.c (expand_omp_target): Output of 'sorry, not supported' if
5232 * omp-low.c (check_omp_nesting_restrictions): Error handling. No nested OpenMP
5233 structs when 'ancestor' is used.
5234 (scan_omp_1_stmt): No usage of OpenMP runtime routines in a target region when
5236 * tree-pretty-print.c (dump_omp_clause): Append 'ancestor'.
5237 * tree.h (OMP_CLAUSE_DEVICE_ANCESTOR): Define macro.
5239 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
5241 * expr.c (convert_modes): Preserve SUBREG_PROMOTED_VAR_P when
5242 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
5244 * simplify-rtx.c (simplify_unary_operation_1) [SIGN_EXTEND]:
5245 Likewise, preserve SUBREG_PROMOTED_VAR_P when creating a (wider)
5246 partial subreg from a SUBREG_PROMOTED_VAR_P subreg. Generate
5247 SIGN_EXTEND of the SUBREG_REG when a subreg would be paradoxical.
5248 [ZERO_EXTEND]: Likewise, preserve SUBREG_PROMOTED_VAR_P when
5249 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
5250 subreg. Generate ZERO_EXTEND of the SUBREG_REG when a subreg
5251 would be paradoxical.
5253 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
5255 * combine.c (combine_simplify_rtx): Avoid converting an explicit
5256 TRUNCATE into a lowpart SUBREG on !TRULY_NOOP_TRUNCATION targets.
5257 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
5259 2021-08-31 Richard Biener <rguenther@suse.de>
5261 PR tree-optimization/102142
5262 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Fix
5263 condition under which to unset the visited flag.
5265 2021-08-31 Richard Biener <rguenther@suse.de>
5267 PR middle-end/102129
5268 * tree-ssa-ter.c (find_replaceable_in_bb): Do not move
5269 possibly trapping expressions across calls.
5271 2021-08-31 Jakub Jelinek <jakub@redhat.com>
5273 PR tree-optimization/102134
5274 * tree-ssa-ccp.c (bit_value_binop) <case RSHIFT_EXPR>: If sgn is
5275 UNSIGNED and r1val | r1mask has MSB set, ensure lzcount doesn't
5278 2021-08-31 Andrew Pinski <apinski@marvell.com>
5281 * collect-utils.c (setup_signals): New declaration.
5282 * collect-utils.h (setup_signals): New function.
5283 * collect2.c (handler): Delete.
5284 (main): Instead of manually setting up the signals,
5285 just call setup_signals.
5286 * lto-wrapper.c (main): Likewise.
5288 2021-08-31 Andrew Pinski <apinski@marvell.com>
5291 * config/i386/i386-protos.h (x86_output_aligned_bss):
5292 Change align argument to unsigned type.
5293 (x86_elf_aligned_decl_common): Likewise.
5294 * config/i386/i386.c (x86_elf_aligned_decl_common): Likewise.
5295 (x86_output_aligned_bss): Likewise.
5297 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
5299 * config/mips/mips.c (mips_module_isa_name): New.
5300 mips_file_start: add .module mipsREV to all asm output
5302 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
5304 * config/mips/mips.h (struct mips_cpu_info): define enum mips_isa;
5305 use enum instead of int for 'isa' member.
5306 * config.gcc, config/mips/mips.c, config/mips/mips-cpus.def,
5307 config/mips/netbsd.h: replace hardcoded numbers with enum.
5309 2021-08-31 liuhongt <hongtao.liu@intel.com>
5311 * config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
5312 define_split to define_insn_and_split.
5313 (*avx2_eq<mode>3): Removed.
5314 (<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
5315 (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
5316 (*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
5318 (*avx2_gt<mode>3): Removed.
5319 (<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
5320 define_insn to define_expand, and adjust pattern.
5321 (UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.
5323 2021-08-30 David Malcolm <dmalcolm@redhat.com>
5326 * Makefile.in (ANALYZER_OBJS): Add analyzer/call-info.o.
5328 2021-08-30 Jason Merrill <jason@redhat.com>
5330 * doc/invoke.texi: Document -Wmissing-requires.
5332 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
5334 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Remove
5335 TARGET_EXTRA_BUILTINS guard.
5337 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
5339 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Change
5340 initialization of V2DI_type_node and unsigned_V2DI_type_node.
5342 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
5344 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Use the new
5345 decl when new_builtins_are_live.
5346 * config/rs6000/rs6000-builtin-new.def (__builtin_cfstring): New
5349 2021-08-30 Pat Haugen <pthaugen@linux.ibm.com>
5351 * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add
5352 OPTION_MASK_P10_FUSION_2STORE.
5353 (POWERPC_MASKS): Likewise.
5354 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
5355 store fusion for Power10.
5356 (is_fusable_store): New.
5357 (power10_sched_reorder): Likewise.
5358 (rs6000_sched_reorder): Do Power10 specific reordering.
5359 (rs6000_sched_reorder2): Likewise.
5360 * config/rs6000/rs6000.opt: Add new option.
5362 2021-08-30 Richard Biener <rguenther@suse.de>
5364 PR tree-optimization/102128
5365 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
5366 Move scanning for if-converted scalar code to the caller
5367 and instead delay clearing the visited flag for profitable
5369 (vect_slp_region): Cost all subgraphs before scheduling.
5370 For if-converted BB vectorization scan for scalar COND_EXPRs
5371 and do not vectorize if any found and the cost model is
5374 2021-08-30 Richard Biener <rguenther@suse.de>
5376 * common.opt (fexceptions): Mark
5377 EnabledBy(fnon-call-exceptions).
5378 * doc/invoke.texi (fnon-call-exceptions): Document this
5379 enables -fexceptions.
5381 2021-08-30 Sebastian Huber <sebastian.huber@embedded-brains.de>
5383 * tsystem.h (abort): Define abort() if inhibit_libc is defined and it
5384 is not already defined.
5386 2021-08-30 liuhongt <hongtao.liu@intel.com>
5388 * expmed.c (extract_bit_field_1): Make sure we're playing with
5389 integral modes before call extract_integral_bit_field.
5390 (extract_integral_bit_field): Add a parameter of type
5391 scalar_int_mode which corresponds to of tmode.
5392 And call extract_and_convert_fixed_bit_field instead of
5393 extract_fixed_bit_field and convert_extracted_bit_field.
5394 (extract_and_convert_fixed_bit_field): New function, it's a
5395 combination of extract_fixed_bit_field and
5396 convert_extracted_bit_field.
5398 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
5400 * config/darwin.c (darwin_libc_has_function): Do not run
5401 the checks for x86 or modern Darwin. Make sure that there
5402 is a value set for darwin_macosx_version_min before testing.
5404 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
5406 * config/i386/darwin.h (CLEAR_INSN_CACHE): New.
5408 2021-08-28 Jan Hubicka <hubicka@ucw.cz>
5410 * ipa-modref-tree.h (modref_access_node::merge): Break out
5411 logic combining offsets and logic merging ranges to ...
5412 (modref_access_node::combined_offsets): ... here
5413 (modref_access_node::update2): ... here
5414 (modref_access_node::closer_pair_p): New member function.
5415 (modref_access_node::forced_merge): New member function.
5416 (modre_ref_node::insert): Do merging when table is full.
5418 2021-08-28 YunQiang Su <yunqiang.su@cipunited.com>
5421 * config.gcc: MIPS: use N64 ABI by default if the triple end
5422 with -gnuabi64, which is used by Debian since 2013.
5424 2021-08-28 Alexandre Oliva <oliva@adacore.com>
5426 * ipa-modref.c (analyze_function): Skip debug stmts.
5427 * tree-inline.c (estimate_num_insn): Consider builtins even
5428 without a cgraph_node.
5430 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
5432 * config/h8300/bitfield.md (cstore<mode>4): Remove expander.
5433 * config/h8300/h8300.c (h8300_expand_branch): Remove function.
5434 * config/h8300/h8300-protos.h (h8300_expadn_branch): Remove prototype.
5435 * config/h8300/h8300.md (eqne): New code iterator.
5436 (geultu, geultu_to_c): Similarly.
5437 * config/h8300/testcompare.md (cstore<mode>4): Dummy expander.
5438 (store_c_<mode>, store_c_i_<mode>): New define_insn_and_splits
5439 (cmp<mode>_c): New pattern
5441 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
5443 * tree-ssa-dom.c (reduce_vector_comparison_to_scalar_comparison): New
5445 (dom_opt_dom_walker::optimize_stmt): Use it.
5447 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
5449 * config/darwin.c (finalize_ctors): Add a section-start linker-
5451 (finalize_dtors): Likewise.
5452 * config/darwin.h (MIN_LD64_INIT_TERM_START_LABELS): New.
5454 2021-08-27 Bill Schmidt <wschmidt@linux.ibm.com>
5456 * config/rs6000/rs6000-call.c (rs6000-builtins.h): New #include.
5457 (rs6000_init_builtins): Call rs6000_init_generated_builtins. Skip the
5458 old initialization logic when new builtins are enabled.
5459 * config/rs6000/rs6000-gen-builtins.c (write_decls): Rename
5460 rs6000_autoinit_builtins to rs6000_init_generated_builtins.
5461 (write_init_file): Likewise.
5463 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
5465 * configure.ac (darwin2[[0-9]]* | darwin19*): Alter use of
5466 gcc_GAS_CHECK_FEATURE to remove an extraneous parameter.
5467 (amdgcn-* | gcn-*) Likewise.
5469 2021-08-27 Anthony Sharp <anthonysharp15@gmail.com>
5471 * symbol-summary.h: Added missing template keyword.
5473 2021-08-27 Richard Biener <rguenther@suse.de>
5475 PR tree-optimization/45178
5476 * tree-ssa-dce.c (find_obviously_necessary_stmts): For
5477 infinite loops without exit do not mark control dependent
5478 edges of the latch necessary.
5480 2021-08-27 konglin1 <lingling.kong@intel.com>
5483 * config/i386/sse.md: (<avx512>scattersi<mode>): Add mask operand to
5485 (<avx512>scattersi<mode>): Likewise.
5486 (*avx512f_scattersi<VI48F:mode>): Merge mask operand to set_dest.
5487 (*avx512f_scatterdi<VI48F:mode>): Likewise
5489 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
5491 * config/rs6000/rs6000.c (rs6000_builtin_md_vectorized_function): Add
5492 support for built-in functions MISC_BUILTIN_DIVWE, MISC_BUILTIN_DIVWEU,
5493 MISC_BUILTIN_DIVDE, MISC_BUILTIN_DIVDEU, P10_BUILTIN_CFUGED,
5494 P10_BUILTIN_CNTLZDM, P10_BUILTIN_CNTTZDM, P10_BUILTIN_PDEPD and
5495 P10_BUILTIN_PEXTD on Power10.
5497 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
5499 * config/rs6000/rs6000-call.c (builtin_function_type): Add unsigned
5500 signedness for some Power10 bifs.
5502 2021-08-27 David Edelsohn <dje.gcc@gmail.com>
5505 * config/rs6000/rs6000.c (rs6000_adjust_field_align): Use
5506 computed alignment if the entire struct has attribute packed.
5508 2021-08-27 liuhongt <hongtao.liu@intel.com>
5512 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
5513 IX86_BUILTIN_SHUFPD512, IX86_BUILTIN_SHUFPS512,
5514 IX86_BUILTIN_SHUFPD256, IX86_BUILTIN_SHUFPS,
5515 IX86_BUILTIN_SHUFPS256.
5516 (ix86_masked_all_ones): New function.
5518 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
5520 * config/i386/i386.md (*btr<mode>_1): Call force_reg unconditionally.
5521 (conditional moves with memory inputs splitters): Ditto.
5522 * config/i386/sse.md (one_cmpl<mode>2): Simplify.
5524 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
5526 * ipa-modref-tree.h (modref_access_node::try_merge_with): Restart
5527 search after merging.
5529 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
5531 * config/rs6000/rs6000-overload.def: Add remaining overloads.
5533 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
5535 * config/rs6000/rs6000-builtin-new.def: Add cell stanza.
5537 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
5539 * config/rs6000/rs6000-builtin-new.def: Add ieee128-hw, dfp,
5540 crypto, and htm stanzas.
5542 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
5544 * config/rs6000/rs6000-builtin-new.def: Add mma stanza.
5546 2021-08-26 Martin Sebor <msebor@redhat.com>
5548 * tree-ssa-uninit.c (warn_uninit): Refactor and simplify.
5549 (warn_uninit_phi_uses): Remove argument from calls to warn_uninit.
5550 (warn_uninitialized_vars): Same. Reduce visibility of locals.
5551 (warn_uninitialized_phi): Same.
5553 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
5555 * tree-ssa-ccp.c (get_individual_bits): Helper function to
5556 extract the individual bits from a widest_int constant (mask).
5557 (gray_code_bit_flips): New read-only table for effiently
5558 enumerating permutations/combinations of bits.
5559 (bit_value_binop) [LROTATE_EXPR, RROTATE_EXPR]: Handle rotates
5560 by unknown counts that are guaranteed less than the target
5561 precision and four or fewer unknown bits by enumeration.
5562 [LSHIFT_EXPR, RSHIFT_EXPR]: Likewise, also handle shifts by
5563 enumeration under the same conditions. Handle remaining
5564 shifts as a mask based upon the minimum possible shift value.
5566 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
5567 Richard Biener <rguenther@suse.de>
5569 * match.pd (shift transformations): Remove a redundant
5570 !POINTER_TYPE_P check.
5572 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
5575 * config/i386/i386.md (cmove reg-reg move elimination peephole2s):
5576 Set all_regs to true in the call to replace_rtx.
5578 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
5580 * ipa-modref-tree.c (test_insert_search_collapse): Update test.
5581 * ipa-modref-tree.h (modref_base_node::insert): Be smarter when
5582 hiting --param modref-max-refs limit.
5583 (modref_tree:insert_base): Be smarter when hitting
5584 --param modref-max-bases limit. Add new parameter REF.
5585 (modref_tree:insert): Update.
5586 (modref_tree:merge): Update.
5587 * ipa-modref.c (read_modref_records): Update.
5589 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
5591 * params.opt: (modref-max-adjustments): Add full stop.
5593 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
5595 * ipa-modref-tree.h (modref_ref_node::verify): New member
5597 (modref_ref_node::insert): Use it.
5598 (modref_ref_node::try_mere_with): Fix off by one error.
5600 2021-08-26 Martin Liska <mliska@suse.cz>
5601 Stefan Kneifel <stefan.kneifel@bluewin.ch>
5603 * cgraph.h (create_version_clone_with_body): Add new parameter.
5604 * cgraphclones.c: Likewise.
5605 * multiple_target.c (create_dispatcher_calls): Do not use
5607 (create_target_clone): Likewise here.
5609 2021-08-26 Jonathan Yong <10walls@gmail.com>
5611 * doc/extend.texi: Add note about reserved priorities
5612 to the constructor attribute.
5614 2021-08-25 Martin Sebor <msebor@redhat.com>
5616 * gimple-range-cache.cc (ssa_global_cache::dump): Avoid printing
5617 range table header alone.
5618 * gimple-range.cc (gimple_ranger::export_global_ranges): Same.
5620 2021-08-25 Jan Hubicka <hubicka@ucw.cz>
5622 * doc/invoke.texi: Document --param modref-max-adjustments.
5623 * ipa-modref-tree.c (test_insert_search_collapse): Update.
5624 (test_merge): Update.
5625 * ipa-modref-tree.h (struct modref_access_node): Add adjustments;
5626 (modref_access_node::operator==): Fix handling of access ranges.
5627 (modref_access_node::contains): Constify parameter; handle also
5628 mismatched parm offsets.
5629 (modref_access_node::update): New function.
5630 (modref_access_node::merge): New function.
5631 (unspecified_modref_access_node): Update constructor.
5632 (modref_ref_node::insert_access): Add record_adjustments parameter;
5634 (modref_ref_node::try_merge_with): New private function.
5635 (modref_tree::insert): New record_adjustments parameter.
5636 (modref_tree::merge): New record_adjustments parameter.
5637 (modref_tree::copy_from): Update.
5638 * ipa-modref.c (dump_access): Dump adjustments field.
5639 (get_access): Update constructor.
5640 (record_access): Update call of insert.
5641 (record_access_lto): Update call of insert.
5642 (merge_call_side_effects): Add record_adjustments parameter.
5643 (get_access_for_fnspec): Update.
5644 (process_fnspec): Update.
5645 (analyze_call): Update.
5646 (analyze_function): Update.
5647 (read_modref_records): Update.
5648 (ipa_merge_modref_summary_after_inlining): Update.
5649 (propagate_unknown_call): Update.
5650 (modref_propagate_in_scc): Update.
5651 * params.opt (param-max-modref-adjustments=): New.
5653 2021-08-25 Michael Meissner <meissner@linux.ibm.com>
5655 * config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
5657 (xxspltiw_v4si): Use vecperm type attribute.
5658 (xxspltiw_v4si_inst): Use vecperm type attribute.
5659 (xxspltiw_v4sf_inst): Likewise.
5660 (xxspltidp_v2df): Use vecperm type attribute. Use
5661 UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
5662 (xxspltidp_v2df_inst): Likewise.
5663 (xxsplti32dx_v4si): Use vecperm type attribute.
5664 (xxsplti32dx_v4si_inst): Likewise.
5665 (xxsplti32dx_v4sf_inst): Likewise.
5666 (xxblend_<mode>): Likewise.
5667 (xxpermx): Likewise.
5668 (xxpermx_inst): Likewise.
5671 2021-08-25 Lewis Hyatt <lhyatt@gmail.com>
5674 * coretypes.h (typedef diagnostic_input_charset_callback): Declare.
5675 * diagnostic.c (diagnostic_initialize_input_context): New function.
5676 * diagnostic.h (diagnostic_initialize_input_context): Declare.
5677 * input.c (default_charset_callback): New function.
5678 (file_cache::initialize_input_context): New function.
5679 (file_cache_slot::create): Added ability to convert the input
5680 according to the input context.
5681 (file_cache::file_cache): Initialize the new input context.
5682 (class file_cache_slot): Added new m_alloc_offset member.
5683 (file_cache_slot::file_cache_slot): Initialize the new member.
5684 (file_cache_slot::~file_cache_slot): Handle potentially offset buffer.
5685 (file_cache_slot::maybe_grow): Likewise.
5686 (file_cache_slot::needs_read_p): Handle NULL fp, which is now possible.
5687 (file_cache_slot::get_next_line): Likewise.
5688 * input.h (class file_cache): Added input context member.
5690 2021-08-25 Richard Biener <rguenther@suse.de>
5692 PR tree-optimization/102046
5693 * tree-vect-slp.c (vect_build_slp_tree_2): Conservatively
5694 update ->any_pattern when swapping operands.
5696 2021-08-25 Hongyu Wang <hongyu.wang@intel.com>
5699 * config/i386/i386.c (ix86_live_on_entry): Adjust comment.
5700 (ix86_decompose_address): Remove retval check for ASHIFT,
5701 allow non-canonical zero extend if AND mask covers ASHIFT
5703 (ix86_legitimate_address_p): Adjust condition for decompose.
5704 (ix86_rtx_costs): Adjust cost for lea with non-canonical
5706 Co-Authored by: Uros Bizjak <ubizjak@gmail.com>
5708 2021-08-25 Jiufu Guo <guojiufu@linux.ibm.com>
5710 PR tree-optimization/101145
5711 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
5713 (number_of_iterations_lt): Invoke above function.
5714 (adjust_cond_for_loop_until_wrap):
5715 Merge to number_of_iterations_until_wrap.
5716 (number_of_iterations_cond): Update invokes for
5717 adjust_cond_for_loop_until_wrap and number_of_iterations_lt.
5719 2021-08-25 konglin1 <lingling.kong@intel.com>
5722 * config/i386/avx512dqintrin.h (_mm512_fpclass_ps_mask): Fix
5724 (_mm512_mask_fpclass_ps_mask): Ditto.
5726 2021-08-25 Kewen Lin <linkw@linux.ibm.com>
5728 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Remove.
5729 (vec_unpacku_hi_v8hi): Likewise.
5730 (vec_unpacku_lo_v16qi): Likewise.
5731 (vec_unpacku_lo_v8hi): Likewise.
5732 (vec_unpacku_hi_<VP_small_lc>): New define_expand.
5733 (vec_unpacku_lo_<VP_small_lc>): Likewise.
5735 2021-08-24 David Edelsohn <dje.gcc@gmail.com>
5737 * config/rs6000/aix.h (SYSTEM_IMPLICIT_EXTERN_C): Delete.
5738 * config/rs6000/aix71.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
5739 * config/rs6000/aix72.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
5740 * config/rs6000/aix73.h (TARGET_AIX_VERSION): Increase to 73.
5742 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5744 PR middle-end/102031
5745 * simplify-rtx.c (simplify_truncation): When comparing precisions
5746 use "subreg_prec" variable, not "subreg_mode".
5748 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5750 * config/rs6000/rs6000-builtin-new.def: Add power10 and power10-64
5753 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5755 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Initialize
5756 various pointer type nodes.
5757 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add enum
5758 values for various pointer types.
5759 (ptr_V16QI_type_node): New macro.
5760 (ptr_V1TI_type_node): New macro.
5761 (ptr_V2DI_type_node): New macro.
5762 (ptr_V2DF_type_node): New macro.
5763 (ptr_V4SI_type_node): New macro.
5764 (ptr_V4SF_type_node): New macro.
5765 (ptr_V8HI_type_node): New macro.
5766 (ptr_unsigned_V16QI_type_node): New macro.
5767 (ptr_unsigned_V1TI_type_node): New macro.
5768 (ptr_unsigned_V8HI_type_node): New macro.
5769 (ptr_unsigned_V4SI_type_node): New macro.
5770 (ptr_unsigned_V2DI_type_node): New macro.
5771 (ptr_bool_V16QI_type_node): New macro.
5772 (ptr_bool_V8HI_type_node): New macro.
5773 (ptr_bool_V4SI_type_node): New macro.
5774 (ptr_bool_V2DI_type_node): New macro.
5775 (ptr_bool_V1TI_type_node): New macro.
5776 (ptr_pixel_type_node): New macro.
5777 (ptr_intQI_type_node): New macro.
5778 (ptr_uintQI_type_node): New macro.
5779 (ptr_intHI_type_node): New macro.
5780 (ptr_uintHI_type_node): New macro.
5781 (ptr_intSI_type_node): New macro.
5782 (ptr_uintSI_type_node): New macro.
5783 (ptr_intDI_type_node): New macro.
5784 (ptr_uintDI_type_node): New macro.
5785 (ptr_intTI_type_node): New macro.
5786 (ptr_uintTI_type_node): New macro.
5787 (ptr_long_integer_type_node): New macro.
5788 (ptr_long_unsigned_type_node): New macro.
5789 (ptr_float_type_node): New macro.
5790 (ptr_double_type_node): New macro.
5791 (ptr_long_double_type_node): New macro.
5792 (ptr_dfloat64_type_node): New macro.
5793 (ptr_dfloat128_type_node): New macro.
5794 (ptr_ieee128_type_node): New macro.
5795 (ptr_ibm128_type_node): New macro.
5796 (ptr_vector_pair_type_node): New macro.
5797 (ptr_vector_quad_type_node): New macro.
5798 (ptr_long_long_integer_type_node): New macro.
5799 (ptr_long_long_unsigned_type_node): New macro.
5801 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5803 * config/rs6000/rs6000-builtin-new.def: Add power9-vector, power9,
5804 and power9-64 stanzas.
5806 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5807 Tom de Vries <tdevries@suse.de>
5809 * config.gcc (nvptx-*-*): Define {c,c++}_target_objs.
5810 * config/nvptx/nvptx-protos.h (nvptx_cpu_cpp_builtins): Prototype.
5811 * config/nvptx/nvptx.h (TARGET_CPU_CPP_BUILTINS): Implement with
5812 a call to the new nvptx_cpu_cpp_builtins function in nvptx-c.c.
5813 * config/nvptx/t-nvptx (nvptx-c.o): New rule.
5814 * config/nvptx/nvptx-c.c: New source file.
5815 (nvptx_cpu_cpp_builtins): Move implementation here.
5817 2021-08-24 Martin Sebor <msebor@redhat.com>
5819 PR middle-end/101600
5820 PR middle-end/101977
5821 * gimple-ssa-warn-access.cc (maybe_warn_for_bound): Tighten up
5822 the phrasing of a warning.
5823 (check_access): Use the remaining size after subtracting any offset
5824 rather than the whole object size.
5825 * pointer-query.cc (access_ref::get_ref): Clear BASE0 flag if it's
5826 clear for any nonnull PHI argument.
5827 (compute_objsize): Clear argument.
5829 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5831 * config/rs6000/rs6000-builtin-new.def: Add power8-vector stanza.
5833 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5835 * config/rs6000/rs6000-builtin-new.def: Add power7 and power7-64
5838 2021-08-24 Andrew MacLeod <amacleod@redhat.com>
5840 * value-relation.cc (rr_transitive_table): New.
5841 (relation_transitive): New.
5842 (value_relation::swap): Remove.
5843 (value_relation::apply_transitive): New.
5844 (relation_oracle::relation_oracle): Allocate a new tmp bitmap.
5845 (relation_oracle::register_relation): Call register_transitives.
5846 (relation_oracle::register_transitives): New.
5847 * value-relation.h (relation_oracle): Add new temporary bitmap and
5850 2021-08-24 H.J. Lu <hjl.tools@gmail.com>
5853 * config/i386/i386-expand.c (ix86_expand_vector_move): Broadcast
5854 from integer to a pseudo vector register.
5856 2021-08-24 Richard Biener <rguenther@suse.de>
5858 PR tree-optimization/100089
5859 * tree-vectorizer.h (vect_slp_bb): Rename to ...
5860 (vect_slp_if_converted_bb): ... this and get the original
5861 loop as new argument.
5862 * tree-vectorizer.c (try_vectorize_loop_1): Revert previous fix,
5863 pass original loop to vect_slp_if_converted_bb.
5864 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
5865 If orig_loop was passed scan the not vectorized stmts
5866 for COND_EXPRs and force not profitable if found.
5867 (vect_slp_region): Pass down all SLP instances to costing
5868 if orig_loop was specified.
5869 (vect_slp_bbs): Pass through orig_loop.
5870 (vect_slp_bb): Rename to ...
5871 (vect_slp_if_converted_bb): ... this and get the original
5872 loop as new argument.
5873 (vect_slp_function): Adjust.
5875 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
5878 * config/arm/arm.md (attribute arch): Add fix_vlldm.
5879 (arch_enabled): Use it.
5880 * config/arm/vfp.md (lazy_store_multiple_insn): Add alternative to
5881 use when erratum mitigation is needed.
5883 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
5886 * config/arm/arm.opt (mfix-cmse-cve-2021-35465): New option.
5887 * doc/invoke.texi (Arm Options): Document it.
5888 * config/arm/arm-cpus.in (quirk_vlldm): New feature bit.
5889 (ALL_QUIRKS): Add quirk_vlldm.
5890 (cortex-m33): Add quirk_vlldm.
5891 (cortex-m35p, cortex-m55): Likewise.
5892 * config/arm/arm.c (arm_option_override): Enable fix_vlldm if
5893 targetting an affected CPU and not explicitly controlled on
5896 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
5898 * config/arm/vfp.md (lazy_store_multiple_insn): Rewrite as valid RTL.
5899 (lazy_load_multiple_insn): Likewise.
5901 2021-08-24 liuhongt <hongtao.liu@intel.com>
5904 * config/i386/sse.md (<avx512>_vternlog<mode><sd_maskz_name>):
5905 Enable avx512 embedded broadcast.
5906 (*<avx512>_vternlog<mode>_all): Ditto.
5907 (<avx512>_vternlog<mode>_mask): Ditto.
5909 2021-08-24 liuhongt <hongtao.liu@intel.com>
5912 * config/i386/i386.c (ix86_rtx_costs): Define cost for
5914 * config/i386/i386.h (STRIP_UNARY): New macro.
5915 * config/i386/predicates.md (reg_or_notreg_operand): New
5917 * config/i386/sse.md (*<avx512>_vternlog<mode>_all): New define_insn.
5918 (*<avx512>_vternlog<mode>_1): New pre_reload
5919 define_insn_and_split.
5920 (*<avx512>_vternlog<mode>_2): Ditto.
5921 (*<avx512>_vternlog<mode>_3): Ditto.
5922 (any_logic1,any_logic2): New code iterator.
5923 (logic_op): New code attribute.
5924 (ternlogsuffix): Extend to VNxDF and VNxSF.
5926 2021-08-24 Richard Biener <rguenther@suse.de>
5928 * doc/invoke.texi (vect-inner-loop-cost-factor): Adjust.
5929 * params.opt (--param vect-inner-loop-cost-factor): Adjust
5931 * tree-vect-loop.c (vect_analyze_loop_form): Initialize
5932 inner_loop_cost_factor to the minimum of the estimated number
5933 of iterations of the inner loop and vect-inner-loop-cost-factor.
5935 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5936 Richard Biener <rguenther@suse.de>
5938 * config/i386/i386-features.c (compute_convert_gain): Provide
5939 more accurate values for CONST_INT, when optimizing for size.
5940 * config/i386/i386.c (COSTS_N_BYTES): Move definition from here...
5941 * config/i386/i386.h (COSTS_N_BYTES): to here.
5943 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5944 Jakub Jelinek <jakub@redhat.com>
5946 PR middle-end/102029
5947 * match.pd (shift transformations): Add an additional check for
5948 !POINTER_TYPE_P in the recently added left shift transformation.
5950 2021-08-24 liuhongt <hongtao.liu@intel.com>
5952 PR tree-optimization/100089
5953 * tree-vectorizer.c (try_vectorize_loop_1): Disable slp in
5954 loop vectorizer when cost model is very-cheap.
5956 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5958 * config/rs6000/rs6000-gen-builtins.c (parse_bif_entry): Don't call
5959 asprintf, which is not available on AIX.
5961 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5963 * config.gcc (target_gtfiles): Add ./rs6000-builtins.h.
5964 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Set.
5966 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5968 * config.gcc (powerpc*-*-*): Add rs6000-builtins.o to extra_objs.
5969 * config/rs6000/rs6000-gen-builtins.c (main): Close init_file
5971 * config/rs6000/t-rs6000 (rs6000-gen-builtins.o): New target.
5972 (rbtree.o): Likewise.
5973 (rs6000-gen-builtins): Likewise.
5974 (rs6000-builtins.c): Likewise.
5975 (rs6000-builtins.h): Likewise.
5976 (rs6000.o): Add dependency.
5977 (EXTRA_HEADERS): Add rs6000-vecdefines.h.
5978 (rs6000-vecdefines.h): New target.
5979 (rs6000-builtins.o): Likewise.
5980 (rs6000-call.o): Add rs6000-builtins.h as a dependency.
5981 (rs6000-c.o): Likewise.
5983 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5986 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace):
5987 Diagnose buffer overrun.
5988 (safe_inc_pos): Fix overrun detection.
5989 (match_identifier): Diagnose buffer overrun.
5990 (match_integer): Likewise.
5991 (match_to_right_bracket): Likewise.
5993 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
5995 * ipa-modref-tree.h (modref_access_node::range_info_useful_p):
5996 Improve range compare.
5997 (modref_access_node::contains): New member function.
5998 (modref_access_node::search): Remove.
5999 (modref_access_node::insert): Be smarter about subaccesses.
6001 2021-08-23 Thomas Schwinge <thomas@codesourcery.com>
6003 * config/i386/i386-options.c (ix86_omp_device_kind_arch_isa)
6004 <omp_device_arch> [ACCEL_COMPILER]: Match "intel_mic".
6005 * config/i386/t-omp-device (omp-device-properties-i386) <arch>:
6008 2021-08-23 Jeff Law <jlaw@localhost.localdomain>
6010 * config/h8300/h8300-protos.h (h8300_expand_epilogue): Add new
6012 * config/h8300/jumpcall.md (call, call_value): Restrict to
6013 !SIBLING_CALL_P cases.
6014 (subcall, sibcall_value): New patterns & expanders.
6015 * config/h8300/proepi.md (epilogue): Pass new argument to
6016 h8300_expand_epilogue.
6017 (sibcall_epilogue): New expander.
6018 * config/h8300/h8300.c (h8300_expand_epilogue): Handle sibcall
6020 (h8300_ok_for_sibcall_p): New function.
6021 (TARGET_FUNCTION_OK_FOR_SIBCALL): define.
6023 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
6025 * simplify-rtx.c (simplify_unary_operation_1): [TRUNCATE]:
6026 Handle case where the operand is already the desired mode.
6028 2021-08-23 Richard Biener <rguenther@suse.de>
6031 * tree-ssa-structalias.c (ipa_pta_execute): Check in_other_partition
6032 in addition to has_gimple_body.
6034 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
6036 PR middle-end/101949
6037 * ipa-modref.c (analyze_ssa_name_flags): Fix merging of
6040 2021-08-23 Martin Liska <mliska@suse.cz>
6042 * doc/invoke.texi: Put the option out of -mxl-mode-app-model
6045 2021-08-23 Richard Biener <rguenther@suse.de>
6047 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6048 Properly scale the inner loop cost only once.
6050 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
6052 * tree-ssa-ccp.c (bit_value_binop) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]:
6053 Provide bounds for unsigned (and signed with non-negative operands)
6054 division and modulus.
6056 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
6058 * simplify-rtx.c (simplify_truncation): Generalize simplification
6059 of (truncate:A (subreg:B X)).
6060 (simplify_unary_operation_1) [FLOAT_TRUNCATE, FLOAT_EXTEND,
6061 SIGN_EXTEND, ZERO_EXTEND]: Handle cases where the operand
6062 already has the desired machine mode.
6063 (test_scalar_int_ops): Add tests that useless extensions and
6064 truncations are optimized away.
6065 (test_scalar_int_ext_ops): New self-test function to confirm
6066 that truncations of extensions are correctly simplified.
6067 (test_scalar_int_ext_ops2): New self-test function to check
6068 truncations of truncations, extensions of extensions, and
6069 truncations of extensions.
6070 (test_scalar_ops): Call the above two functions with a
6071 representative sampling of integer machine modes.
6073 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
6075 * match.pd (shift transformations): Change the sign of an
6076 LSHIFT_EXPR if it reduces the number of explicit conversions.
6078 2021-08-23 Jakub Jelinek <jakub@redhat.com>
6080 PR tree-optimization/86723
6081 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
6082 cast64_to_32 argument, set *cast64_to_32 to false, unless n is
6083 non-memory permutation of 64-bit src which only has bytes of
6084 0 or [5..8] and n->range is 4.
6085 (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
6086 find_bswap_or_nop_finalize caller, support bswap with some bytes
6087 zeroed, as long as at least two bytes are not zeroed.
6088 (bswap_replace): Add mask argument and handle masking of bswap
6090 (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
6091 caller, punt if cast64_to_32 or mask is not all ones.
6092 (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
6093 caller, for now punt if cast64_to_32.
6095 2021-08-23 Richard Biener <rguenther@suse.de>
6097 PR tree-optimization/79334
6098 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Record
6099 a type also for COMPONENT_REFs.
6100 (vn_reference_may_trap): Check ARRAY_REF with constant index
6101 against the array domain.
6103 2021-08-23 liuhongt <hongtao.liu@intel.com>
6106 * config/i386/sse.md (*avx512f_pshufb_truncv8hiv8qi_1): Add
6107 TARGET_AVX512BW to condition.
6109 2021-08-23 Jakub Jelinek <jakub@redhat.com>
6112 * dwarf2out.c (gen_variable_die): Add DW_AT_location for global
6113 register variables already during early_dwarf if possible.
6115 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
6117 * config/arm/arm_mve.h: Fix __arm_vctp16q return type.
6119 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
6122 * config/arm/arm.opt: Fix typo.
6123 * config/arm/t-rmprofile: Fix typo.
6125 2021-08-23 Jakub Jelinek <jakub@redhat.com>
6127 * tree.h (OMP_CLAUSE_GRAINSIZE_STRICT): Define.
6128 (OMP_CLAUSE_NUM_TASKS_STRICT): Define.
6129 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_GRAINSIZE,
6130 case OMP_CLAUSE_NUM_TASKS>: Print strict: modifier.
6131 * omp-expand.c (expand_task_call): Use GOMP_TASK_FLAG_STRICT in iflags
6132 if either grainsize or num_tasks clause has the strict modifier.
6134 2021-08-23 Martin Liska <mliska@suse.cz>
6136 * dbgcnt.def (DEBUG_COUNTER): New counter.
6137 * gimple.c (gimple_call_arg_flags): Use it in IPA PTA.
6139 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
6141 * ipa-modref.c (analyze_ssa_name_flags): Improve handling of return slot.
6143 2021-08-23 Xi Ruoyao <xry111@mengyan1223.wang>
6146 * config/mips/mips-protos.h (mips_msa_output_shift_immediate):
6148 * config/mips/mips.c (mips_msa_output_shift_immediate): New
6150 * config/mips/mips-msa.md (vashl<mode>3, vashr<mode>3,
6151 vlshr<mode>3): Call it.
6153 2021-08-22 Jan Hubicka <hubicka@ucw.cz>
6154 Martin Liska <mliska@suse.cz>
6156 PR middle-end/101949
6157 * ipa-modref.c (analyze_ssa_name_flags): Indirect call implies
6160 2021-08-21 Dragan Mladjenovic <OT_Dragan.Mladjenovic@mediatek.com>
6162 * config/mips/mips.c (mips_function_rodata_section,
6163 TARGET_ASM_FUNCTION_RODATA_SECTION): Removed.
6165 2021-08-21 John David Anglin <danglin@gcc.gnu.org>
6167 * config/pa/pa.c (pa_asm_output_aligned_common): Remove warning.
6169 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
6171 * configure.ac (thread-local storage support): Remove tls_first_major
6172 and tls_first_minor. Use "$conftest_s" to check support.
6173 * configure: Regenerate.
6175 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
6177 * configure.ac: Fixup formatting.
6179 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
6181 * acinclude.m4 (gcc_GAS_CHECK_FEATURE): Remove third argument and ...
6182 * configure.ac: ... update all callers.
6184 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
6187 * acinclude.m4 (_gcc_COMPUTE_GAS_VERSION, _gcc_GAS_VERSION_GTE_IFELSE)
6188 (gcc_GAS_VERSION_GTE_IFELSE): Remove.
6189 (gcc_GAS_CHECK_FEATURE): Do not handle in-tree case specially.
6190 * configure.ac: Remove gcc_cv_gas_major_version, gcc_cv_gas_minor_version.
6191 Remove remaining checks for in-tree assembler.
6192 * configure: Regenerate.
6194 2021-08-20 Jeff Law <jlaw@localhost.localdomain>
6196 * config/h8300/h8300.c (shift_alg_hi): Improve arithmetic shift right
6197 by 15 bits for H8/300H and H8/S. Improve logical shifts by 12
6199 (shift_alg_si): Improve arithmetic right shift by 28-30 bits for
6200 H8/300H. Improve arithmetic shift right by 15 bits for H8/S.
6201 Improve logical shifts by 27 bits for H8/S.
6202 (get_shift_alg): Corresponding changes.
6203 (h8300_option_override): Revert to loops for -Os when profitable.
6205 2021-08-20 Richard Biener <rguenther@suse.de>
6207 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
6209 (vect_analyze_data_ref_accesses): Likewise. Assign the BB
6210 index as group_id when dataref_groups were not computed.
6211 * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
6212 we advace to the next BB.
6214 2021-08-20 Jakub Jelinek <jakub@redhat.com>
6216 * omp-builtins.def (BUILT_IN_GOMP_WARNING, BUILT_IN_GOMP_ERROR): New
6219 2021-08-20 Martin Liska <mliska@suse.cz>
6221 PR gcov-profile/89961
6222 * gcov.c (make_gcov_file_name): Rewrite using std::string.
6223 (mangle_name): Simplify, do not used the second argument.
6224 (strip_extention): New function.
6225 (get_md5sum): Likewise.
6226 (get_gcov_intermediate_filename): Handle properly -p and -x
6228 (output_gcov_file): Use string type.
6229 (generate_results): Likewise.
6230 (md5sum_to_hex): Remove.
6232 2021-08-20 Michael Meissner <meissner@linux.ibm.com>
6234 * config/rs6000/altivec.md (UNSPEC_XXEVAL): Move to vsx.md.
6235 (UNSPEC_XXSPLTIW): Move to vsx.md.
6236 (UNSPEC_XXSPLTID): Move to vsx.md.
6237 (UNSPEC_XXSPLTI32DX): Move to vsx.md.
6238 (UNSPEC_XXBLEND): Move to vsx.md.
6239 (UNSPEC_XXPERMX): Move to vsx.md.
6240 (VM3): Move to vsx.md.
6241 (VM3_char): Move to vsx.md.
6242 (xxspltiw_v4si): Move to vsx.md.
6243 (xxspltiw_v4sf): Move to vsx.md.
6244 (xxspltiw_v4sf_inst): Move to vsx.md.
6245 (xxspltidp_v2df): Move to vsx.md.
6246 (xxspltidp_v2df_inst): Move to vsx.md.
6247 (xxsplti32dx_v4si_inst): Move to vsx.md.
6248 (xxsplti32dx_v4sf): Move to vsx.md.
6249 (xxsplti32dx_v4sf_inst): Move to vsx.md.
6250 (xxblend_<mode>): Move to vsx.md.
6251 (xxpermx): Move to vsx.md.
6252 (xxpermx_inst): Move to vsx.md.
6253 * config/rs6000/vsx.md (UNSPEC_XXEVAL): Move from altivec.md.
6254 (UNSPEC_XXSPLTIW): Move from altivec.md.
6255 (UNSPEC_XXSPLTID): Move from altivec.md.
6256 (UNSPEC_XXSPLTI32DX): Move from altivec.md.
6257 (UNSPEC_XXBLEND): Move from altivec.md.
6258 (UNSPEC_XXPERMX): Move from altivec.md.
6259 (VM3): Move from altivec.md.
6260 (VM3_char): Move from altivec.md.
6261 (xxspltiw_v4si): Move from altivec.md.
6262 (xxspltiw_v4sf): Move from altivec.md.
6263 (xxspltiw_v4sf_inst): Move from altivec.md.
6264 (xxspltidp_v2df): Move from altivec.md.
6265 (xxspltidp_v2df_inst): Move from altivec.md.
6266 (xxsplti32dx_v4si_inst): Move from altivec.md.
6267 (xxsplti32dx_v4sf): Move from altivec.md.
6268 (xxsplti32dx_v4sf_inst): Move from altivec.md.
6269 (xxblend_<mode>): Move from altivec.md.
6270 (xxpermx): Move from altivec.md.
6271 (xxpermx_inst): Move from altivec.md.
6273 2021-08-19 Roger Sayle <roger@nextmovesoftware.com>
6275 * tree-vect-generic.c (expand_vector_operations_1): Use either
6276 gimplify_build1 or gimplify_build2 instead of gimple_build_assign
6277 when constructing scalar splat expressions.
6279 2021-08-19 Peter Bergner <bergner@linux.ibm.com>
6282 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Cast
6283 pointer to __vector_pair *.
6285 2021-08-19 Martin Sebor <msebor@redhat.com>
6287 * gimple-range.cc: Add comments.
6288 * gimple-range.h: Same.
6290 2021-08-19 Martin Sebor <msebor@redhat.com>
6292 PR middle-end/101984
6293 * gimple-ssa-warn-access.cc (pass_waccess::execute): Also call
6296 2021-08-19 Jeff Law <jlaw@localhost.localdomain>
6298 * config.gcc (h8300-*-elf*): Do not include dbxelf.h.
6299 (h8300-*-linux*, v850-*-rtems*, v850*-elf*): Likewise.
6300 * config/v850/v850.h (DEFAULT_GDB_EXTENSIONS): Remove.
6302 2021-08-19 Jakub Jelinek <jakub@redhat.com>
6304 PR middle-end/101950
6305 * optabs.c (expand_clrsb_using_clz): New function.
6306 (expand_unop): Use it as another clrsb expansion fallback.
6308 2021-08-19 liuhongt <hongtao.liu@intel.com>
6311 2021-07-28 liuhongt <hongtao.liu@intel.com>
6314 * config/i386/i386.h (processor_costs): Add new member
6316 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
6317 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
6318 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
6319 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
6320 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
6321 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
6322 generic_cost, core_cost): Initialize integer_to_sse same value
6324 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
6325 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6326 Use integer_to_sse instead of sse_op to calculate the cost of
6329 2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
6331 * config.gcc: Include rpath.opt for Darwin.
6332 * config/darwin.h (DRIVER_SELF_SPECS): Handle -rpath.
6334 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
6337 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
6340 2021-08-18 Jonathan Wright <jonathan.wright@arm.com>
6342 * config/aarch64/arm_neon.h (vld3_lane_f64): Use float RTL
6343 pattern and type cast.
6344 (vld4_lane_f32): Use float RTL pattern.
6345 (vld4q_lane_f64): Use float type cast.
6347 2021-08-18 Jan Hubicka <hubicka@ucw.cz>
6349 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Check also
6352 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
6354 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor): Extend.
6355 (test_map_of_type_with_ctor_and_dtor_expand): Add function.
6356 (hash_map_tests_c_tests): Call it.
6358 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
6360 * ggc.h (enum ggc_collect): New.
6361 (ggc_collect): Use it.
6362 * ggc-page.c: Adjust.
6363 * ggc-common.c: Likewise.
6364 * ggc-tests.c: Likewise.
6365 * read-rtl-function.c: Likewise.
6366 * selftest-run-tests.c: Likewise.
6367 * doc/gty.texi (Invoking the garbage collector): Likewise.
6369 2021-08-18 liuhongt <hongtao.liu@intel.com>
6372 * config/i386/i386.h (TARGET_V2DF_REDUCTION_PREFER_HADDPD):
6374 * config/i386/sse.md (*sse3_haddv2df3_low): Add
6375 TARGET_V2DF_REDUCTION_PREFER_HADDPD.
6376 (*sse3_hsubv2df3_low): Ditto.
6377 * config/i386/x86-tune.def
6378 (X86_TUNE_V2DF_REDUCTION_PREFER_HADDPD): New tune.
6380 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
6382 * gimple-range-gori.cc (gori_compute::gori_compute): Enable tracing.
6383 (gori_compute::compute_operand_range): Add tracing.
6384 (gori_compute::logical_combine): Ditto.
6385 (gori_compute::compute_logical_operands): Ditto.
6386 (gori_compute::compute_operand1_range): Ditto.
6387 (gori_compute::compute_operand2_range): Ditto.
6388 (gori_compute::outgoing_edge_range_p): Ditto.
6389 * gimple-range-gori.h (class gori_compute): Add range_tracer.
6391 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
6393 * flag-types.h (enum evrp_mode): Adjust evrp-mode values.
6394 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Relocate from.
6395 * gimple-range-trace.h (DEBUG_RANGE_CACHE): Here.
6396 * params.opt (--param=evrp-mode): Adjust options.
6398 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
6400 * Makefile.in (OBJS): Add gimple-range-trace.o.
6401 * gimple-range-cache.h (enable_new_values): Remove unused prototype.
6402 * gimple-range-fold.cc: Adjust headers.
6403 * gimple-range-trace.cc: New.
6404 * gimple-range-trace.h: New.
6405 * gimple-range.cc (gimple_ranger::gimple_ranger): Enable tracer.
6406 (gimple_ranger::range_of_expr): Add tracing.
6407 (gimple_ranger::range_on_entry): Ditto.
6408 (gimple_ranger::range_on_exit): Ditto.
6409 (gimple_ranger::range_on_edge): Ditto.
6410 (gimple_ranger::fold_range_internal): Ditto.
6411 (gimple_ranger::dump_bb): Do not calculate edge range twice.
6412 (trace_ranger::*): Remove.
6413 (enable_ranger): Never create a trace_ranger.
6414 (debug_seed_ranger): Move to gimple-range-trace.cc.
6415 (dump_ranger): Ditto.
6416 (debug_ranger): Ditto.
6417 * gimple-range.h: Include gimple-range-trace.h.
6418 (range_on_entry, range_on_exit): No longer virtual.
6419 (class trace_ranger): Remove.
6420 (DEBUG_RANGE_CACHE): Move to gimple-range-trace.h.
6422 2021-08-17 Martin Sebor <msebor@redhat.com>
6424 PR middle-end/101854
6425 * builtins.c (expand_builtin_alloca): Move warning code to check_alloca
6426 in gimple-ssa-warn-access.cc.
6427 * calls.c (alloc_max_size): Move code to check_alloca.
6428 (get_size_range): Move to pointer-query.cc.
6429 (maybe_warn_alloc_args_overflow): Move to gimple-ssa-warn-access.cc.
6430 (get_attr_nonstring_decl): Move to tree.c.
6431 (fntype_argno_type): Move to gimple-ssa-warn-access.cc.
6432 (append_attrname): Same.
6433 (maybe_warn_rdwr_sizes): Same.
6434 (initialize_argument_information): Move code to
6435 gimple-ssa-warn-access.cc.
6436 * calls.h (maybe_warn_alloc_args_overflow): Move to
6437 gimple-ssa-warn-access.h.
6438 (get_attr_nonstring_decl): Move to tree.h.
6439 (maybe_warn_nonstring_arg): Move to gimple-ssa-warn-access.h.
6440 (enum size_range_flags): Move to pointer-query.h.
6441 (get_size_range): Same.
6442 * gimple-ssa-warn-access.cc (has_location): Remove unused overload
6443 to avoid Clang -Wunused-function.
6444 (get_size_range): Declare static.
6445 (maybe_emit_free_warning): Rename...
6446 (maybe_check_dealloc_call): ...to this for consistency.
6447 (class pass_waccess): Add members.
6448 (pass_waccess::~pass_waccess): Defined.
6449 (alloc_max_size): Move here from calls.c.
6450 (maybe_warn_alloc_args_overflow): Same.
6451 (check_alloca): New function.
6452 (check_alloc_size_call): New function.
6453 (check_strncat): Handle another warning flag.
6454 (pass_waccess::check_builtin): Handle alloca.
6455 (fntype_argno_type): Move here from calls.c.
6456 (append_attrname): Same.
6457 (maybe_warn_rdwr_sizes): Same.
6458 (pass_waccess::check_call): Define.
6459 (check_nonstring_args): New function.
6460 (pass_waccess::check): Call new member functions.
6461 (pass_waccess::execute): Enable ranger.
6462 * gimple-ssa-warn-access.h (get_size_range): Move here from calls.h.
6463 (maybe_warn_nonstring_arg): Same.
6464 * gimple-ssa-warn-restrict.c: Remove #include.
6465 * pointer-query.cc (get_size_range): Move here from calls.c.
6466 * pointer-query.h (enum size_range_flags): Same.
6467 (get_size_range): Same.
6468 * tree.c (get_attr_nonstring_decl): Move here from calls.c.
6469 * tree.h (get_attr_nonstring_decl): Move here from calls.h.
6471 2021-08-17 Thomas Schwinge <thomas@codesourcery.com>
6473 * ggc.h (ggc_collect): Add 'force_collect' parameter.
6474 * ggc-page.c (ggc_collect): Use that one instead of global
6475 'ggc_force_collect'. Adjust all users.
6476 * doc/gty.texi (Invoking the garbage collector): Update.
6477 * ggc-internal.h (ggc_force_collect): Remove.
6478 * ggc-common.c (ggc_force_collect): Likewise.
6479 * selftest.h (forcibly_ggc_collect): Remove.
6480 * ggc-tests.c (selftest::forcibly_ggc_collect): Likewise.
6481 * read-rtl-function.c (test_loading_labels): Adjust.
6482 * selftest-run-tests.c (run_tests): Likewise.
6484 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
6486 * config/darwin.c (darwin_file_end): Reset and reclaim the
6487 section names table at the end of compile.
6489 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
6492 * config.in: Regenerate.
6493 * config/i386/darwin.h (EXTRA_ASM_OPTS): New
6494 (ASM_SPEC): Pass options to disable branch shortening where
6496 * configure: Regenerate.
6497 * configure.ac: Detect versions of 'as' that support the
6498 optimisation which has the bug.
6500 2021-08-17 Richard Biener <rguenther@suse.de>
6502 * optabs-query.c (supports_vec_gather_load_p): Also check
6504 (supports_vec_scatter_store_p): Likewise.
6505 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Fall
6506 back to masked variants if non-masked are not supported.
6507 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
6508 When we need to use masked gather/scatter but do not have
6509 a mask set up a constant true one.
6510 * tree-vect-stmts.c (vect_check_scalar_mask): Also allow
6513 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
6515 * tree-ssa-ccp.c (bit_value_binop) [MINUS_EXPR]: Use same
6516 algorithm as PLUS_EXPR to improve subtraction bit bounds.
6517 [POINTER_DIFF_EXPR]: Treat as synonymous with MINUS_EXPR.
6519 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
6521 * tree-ssa-ccp.c (bit_value_mult_const): New helper function to
6522 calculate the mask-value pair result of a multiplication by an
6524 (bit_value_binop) [MULT_EXPR]: Call it from here for
6525 multiplications by (sparse) non-negative constants.
6527 2021-08-17 Christophe Lyon <christophe.lyon@foss.st.com>
6530 * config.gcc (gcc_cv_initfini_array): Leave undefined for
6531 uclinuxfdpiceabi targets.
6533 2021-08-17 Alexandre Oliva <oliva@adacore.com>
6535 * tree-inline.c (maybe_move_debug_stmts_to_successors): Don't
6536 reverse debug stmts.
6538 2021-08-17 Alexandre Oliva <oliva@adacore.com>
6540 * tree-cfg.c (dump_function_to_file): Use fun, not cfun.
6542 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
6544 * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete.
6545 (__LD4Q_LANE_FUNC): Likewise.
6546 (vld4_lane_u8): Define without macro.
6547 (vld4_lane_u16): Likewise.
6548 (vld4_lane_u32): Likewise.
6549 (vld4_lane_u64): Likewise.
6550 (vld4_lane_s8): Likewise.
6551 (vld4_lane_s16): Likewise.
6552 (vld4_lane_s32): Likewise.
6553 (vld4_lane_s64): Likewise.
6554 (vld4_lane_f16): Likewise.
6555 (vld4_lane_f32): Likewise.
6556 (vld4_lane_f64): Likewise.
6557 (vld4_lane_p8): Likewise.
6558 (vld4_lane_p16): Likewise.
6559 (vld4_lane_p64): Likewise.
6560 (vld4q_lane_u8): Likewise.
6561 (vld4q_lane_u16): Likewise.
6562 (vld4q_lane_u32): Likewise.
6563 (vld4q_lane_u64): Likewise.
6564 (vld4q_lane_s8): Likewise.
6565 (vld4q_lane_s16): Likewise.
6566 (vld4q_lane_s32): Likewise.
6567 (vld4q_lane_s64): Likewise.
6568 (vld4q_lane_f16): Likewise.
6569 (vld4q_lane_f32): Likewise.
6570 (vld4q_lane_f64): Likewise.
6571 (vld4q_lane_p8): Likewise.
6572 (vld4q_lane_p16): Likewise.
6573 (vld4q_lane_p64): Likewise.
6574 (vld4_lane_bf16): Likewise.
6575 (vld4q_lane_bf16): Likewise.
6577 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
6579 * config/aarch64/arm_neon.h (__LD3_LANE_FUNC): Delete.
6580 (__LD3Q_LANE_FUNC): Delete.
6581 (vld3_lane_u8): Define without macro.
6582 (vld3_lane_u16): Likewise.
6583 (vld3_lane_u32): Likewise.
6584 (vld3_lane_u64): Likewise.
6585 (vld3_lane_s8): Likewise.
6586 (vld3_lane_s16): Likewise.
6587 (vld3_lane_s32): Likewise.
6588 (vld3_lane_s64): Likewise.
6589 (vld3_lane_f16): Likewise.
6590 (vld3_lane_f32): Likewise.
6591 (vld3_lane_f64): Likewise.
6592 (vld3_lane_p8): Likewise.
6593 (vld3_lane_p16): Likewise.
6594 (vld3_lane_p64): Likewise.
6595 (vld3q_lane_u8): Likewise.
6596 (vld3q_lane_u16): Likewise.
6597 (vld3q_lane_u32): Likewise.
6598 (vld3q_lane_u64): Likewise.
6599 (vld3q_lane_s8): Likewise.
6600 (vld3q_lane_s16): Likewise.
6601 (vld3q_lane_s32): Likewise.
6602 (vld3q_lane_s64): Likewise.
6603 (vld3q_lane_f16): Likewise.
6604 (vld3q_lane_f32): Likewise.
6605 (vld3q_lane_f64): Likewise.
6606 (vld3q_lane_p8): Likewise.
6607 (vld3q_lane_p16): Likewise.
6608 (vld3q_lane_p64): Likewise.
6609 (vld3_lane_bf16): Likewise.
6610 (vld3q_lane_bf16): Likewise.
6612 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
6614 * config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Delete.
6615 (__LD2Q_LANE_FUNC): Likewise.
6616 (vld2_lane_u8): Define without macro.
6617 (vld2_lane_u16): Likewise.
6618 (vld2_lane_u32): Likewise.
6619 (vld2_lane_u64): Likewise.
6620 (vld2_lane_s8): Likewise.
6621 (vld2_lane_s16): Likewise.
6622 (vld2_lane_s32): Likewise.
6623 (vld2_lane_s64): Likewise.
6624 (vld2_lane_f16): Likewise.
6625 (vld2_lane_f32): Likewise.
6626 (vld2_lane_f64): Likewise.
6627 (vld2_lane_p8): Likewise.
6628 (vld2_lane_p16): Likewise.
6629 (vld2_lane_p64): Likewise.
6630 (vld2q_lane_u8): Likewise.
6631 (vld2q_lane_u16): Likewise.
6632 (vld2q_lane_u32): Likewise.
6633 (vld2q_lane_u64): Likewise.
6634 (vld2q_lane_s8): Likewise.
6635 (vld2q_lane_s16): Likewise.
6636 (vld2q_lane_s32): Likewise.
6637 (vld2q_lane_s64): Likewise.
6638 (vld2q_lane_f16): Likewise.
6639 (vld2q_lane_f32): Likewise.
6640 (vld2q_lane_f64): Likewise.
6641 (vld2q_lane_p8): Likewise.
6642 (vld2q_lane_p16): Likewise.
6643 (vld2q_lane_p64): Likewise.
6644 (vld2_lane_bf16): Likewise.
6645 (vld2q_lane_bf16): Likewise.
6647 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
6649 * haifa-sched.c (advance_one_cycle): Output more context-synchronization
6652 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
6654 * haifa-sched.c (enum rfs_decision, rfs_str): Add RFS_AUTOPREF.
6655 (rank_for_schedule): Use it.
6657 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
6659 PR rtl-optimization/91598
6660 * haifa-sched.c (autopref_rank_for_schedule): Prioritize "irrelevant"
6661 insns after memory reads and before memory writes.
6663 2021-08-17 Alistair_Lee <alistair.lee@arm.com>
6665 * rtl.h (CONST_VECTOR_P): New macro.
6666 * config/aarch64/aarch64.c (aarch64_get_sve_pred_bits): Use RTL
6667 code testing macros.
6668 (aarch64_ptrue_all_mode): Likewise.
6669 (aarch64_expand_mov_immediate): Likewise.
6670 (aarch64_const_vec_all_in_range_p): Likewise.
6671 (aarch64_rtx_costs): Likewise.
6672 (aarch64_legitimate_constant_p): Likewise.
6673 (aarch64_simd_valid_immediate): Likewise.
6674 (aarch64_simd_make_constant): Likewise.
6675 (aarch64_convert_mult_to_shift): Likewise.
6676 (aarch64_expand_sve_vec_perm): Likewise.
6677 (aarch64_vec_fpconst_pow_of_2): Likewise.
6679 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
6681 PR tree-optimization/101938
6682 * range-op.cc (operator_abs::op1_range): Special case
6683 -TYPE_MIN_VALUE for flag_wrapv.
6685 2021-08-17 Kewen Lin <linkw@linux.ibm.com>
6687 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Add the cost for
6690 2021-08-17 Jakub Jelinek <jakub@redhat.com>
6692 * tree.def (OMP_SCOPE): New tree code.
6693 * tree.h (OMP_SCOPE_BODY, OMP_SCOPE_CLAUSES): Define.
6694 * tree-nested.c (convert_nonlocal_reference_stmt,
6695 convert_local_reference_stmt, convert_gimple_call): Handle
6697 * tree-pretty-print.c (dump_generic_node): Handle OMP_SCOPE.
6698 * gimple.def (GIMPLE_OMP_SCOPE): New gimple code.
6699 * gimple.c (gimple_build_omp_scope): New function.
6700 (gimple_copy): Handle GIMPLE_OMP_SCOPE.
6701 * gimple.h (gimple_build_omp_scope): Declare.
6702 (gimple_has_substatements): Handle GIMPLE_OMP_SCOPE.
6703 (gimple_omp_scope_clauses, gimple_omp_scope_clauses_ptr,
6704 gimple_omp_scope_set_clauses): New inline functions.
6705 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_SCOPE.
6706 * gimple-pretty-print.c (dump_gimple_omp_scope): New function.
6707 (pp_gimple_stmt_1): Handle GIMPLE_OMP_SCOPE.
6708 * gimple-walk.c (walk_gimple_stmt): Likewise.
6709 * gimple-low.c (lower_stmt): Likewise.
6710 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
6711 (gimplify_scan_omp_clauses): For task reductions, handle OMP_SCOPE
6712 like ORT_WORKSHARE constructs. Adjust diagnostics for %<scope%>
6713 allowing task reductions. Reject inscan reductions on scope.
6714 (omp_find_stores_stmt): Handle GIMPLE_OMP_SCOPE.
6715 (gimplify_omp_workshare, gimplify_expr): Handle OMP_SCOPE.
6716 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_SCOPE.
6717 (estimate_num_insns): Likewise.
6718 * omp-low.c (build_outer_var_ref): Look through GIMPLE_OMP_SCOPE
6719 contexts if var isn't privatized there.
6720 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_SCOPE.
6721 (scan_omp_1_stmt): Likewise.
6722 (maybe_add_implicit_barrier_cancel): Look through outer
6724 (lower_omp_scope): New function.
6725 (lower_omp_task_reductions): Handle OMP_SCOPE.
6726 (lower_omp_1): Handle GIMPLE_OMP_SCOPE.
6727 (diagnose_sb_1, diagnose_sb_2): Likewise.
6728 * omp-expand.c (expand_omp_single): Support also GIMPLE_OMP_SCOPE.
6729 (expand_omp): Handle GIMPLE_OMP_SCOPE.
6730 (omp_make_gimple_edges): Likewise.
6731 * omp-builtins.def (BUILT_IN_GOMP_SCOPE_START): New built-in.
6733 2021-08-17 Richard Biener <rguenther@suse.de>
6735 PR tree-optimization/101925
6736 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Set
6737 reverse on COMPONENT_REF and ARRAY_REF according to
6738 what reverse_storage_order_for_component_p does.
6739 (vn_reference_eq): Compare reversed on reference ops.
6740 (reverse_storage_order_for_component_p): New overload.
6741 (vn_reference_lookup_3): Check reverse_storage_order_for_component_p
6742 on the reference looked up.
6744 2021-08-17 Jeff Law <jlaw@localhost.localdomain>
6746 * config/h8300/h8300.c (shift_alg_si): Avoid loops for most SImode
6748 (h8300_option_override): Use loops on H8/S more often when optimizing
6750 (get_shift_alg): Handle new "special" cases on H8/S. Simplify
6751 accordingly. Handle various arithmetic right shifts with special
6752 sequences that we couldn't handle before.
6754 2021-08-16 Jeff Law <jlaw@localhost.localdomain>
6756 * config.gcc (rl78-*-elf*): Do not include dbxelf.h.
6758 2021-08-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
6760 * config/sparc/rtemself.h (SPARC_GCOV_TYPE_SIZE): Define.
6761 * config/sparc/sparc.c (sparc_gcov_type_size): New.
6762 (TARGET_GCOV_TYPE_SIZE): Redefine if SPARC_GCOV_TYPE_SIZE is defined.
6763 * coverage.c (get_gcov_type): Use targetm.gcov_type_size().
6764 * doc/tm.texi (TARGET_GCOV_TYPE_SIZE): Add hook under "Misc".
6765 * doc/tm.texi.in: Regenerate.
6766 * target.def (gcov_type_size): New target hook.
6767 * targhooks.c (default_gcov_type_size): New.
6768 * targhooks.h (default_gcov_type_size): Declare.
6769 * tree-profile.c (gimple_gen_edge_profiler): Use precision of
6771 (gimple_gen_time_profiler): Likewise.
6773 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
6775 * dwarf2out.c (add_scalar_info): Deal with DW_AT_data_bit_offset.
6777 2021-08-16 Tobias Burnus <tobias@codesourcery.com>
6779 PR middle-end/101931
6780 * omp-low.c (omp_runtime_api_call): Update for routines
6781 added in the meanwhile.
6783 2021-08-16 Martin Liska <mliska@suse.cz>
6785 PR tree-optimization/100393
6786 * tree-switch-conversion.c (group_cluster::dump): Use
6787 get_comparison_count.
6788 (jump_table_cluster::find_jump_tables): Pre-compute number of
6789 comparisons and then decrement it. Cache also max_ratio.
6790 (jump_table_cluster::can_be_handled): Change signature.
6791 * tree-switch-conversion.h (get_comparison_count): New.
6793 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
6795 * dwarf2out.c (add_data_member_location_attribute): Use GNAT
6796 encodings only when -fgnat-encodings=all is specified.
6797 (add_bound_info): Likewise.
6798 (add_byte_size_attribute): Likewise.
6799 (gen_member_die): Likewise.
6801 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
6803 * omp-oacc-neuter-broadcast.cc
6804 (execute_omp_oacc_neuter_broadcast): Plug 'par' memory leak.
6806 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
6808 * omp-oacc-neuter-broadcast.cc
6809 (execute_omp_oacc_neuter_broadcast): Clarify memory management for
6812 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
6814 * omp-oacc-neuter-broadcast.cc (field_map): Move variable into...
6815 (execute_omp_oacc_neuter_broadcast): ... here.
6816 (install_var_field, build_receiver_ref, build_sender_ref): Take
6817 'field_map_t *' parameter. Adjust all users.
6818 (worker_single_copy, neuter_worker_single): Take a
6819 'record_field_map_t *' parameter. Adjust all users.
6821 2021-08-16 liuhongt <hongtao.liu@intel.com>
6824 * config/i386/i386.md (ldexp<mode>3): Force operands[1] to
6827 2021-08-16 Martin Liska <mliska@suse.cz>
6830 * multiple_target.c (create_dispatcher_calls): Make default
6831 function local only if it is a definition.
6833 2021-08-16 Martin Liska <mliska@suse.cz>
6836 * ipa-icf-gimple.c (func_checker::compare_ssa_name): Do not
6837 consider equal SSA_NAMEs when one is a param.
6839 2021-08-16 liuhongt <hongtao.liu@intel.com>
6842 * config/i386/i386-expand.c (ix86_expand_vec_perm_vpermt2):
6843 Support vpermi2b for V32QI/V16QImode.
6844 (ix86_extract_perm_from_pool_constant): New function.
6845 (ix86_expand_vec_one_operand_perm_avx512): Support
6846 vpermw/vpermb under TARGET_AVX512BW/TARGET_AVX512VBMI.
6847 (expand_vec_perm_1): Adjust comments for upper.
6848 * config/i386/i386-protos.h (ix86_extract_perm_from_pool_constant):
6850 * config/i386/predicates.md (permvar_truncate_operand): New predicate.
6851 (pshufb_truncv4siv4hi_operand): Ditto.
6852 (pshufb_truncv8hiv8qi_operand): Ditto.
6853 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1):
6854 New pre_reload define_insn_and_split.
6855 (*avx512f_permvar_truncv8siv8hi_1): Ditto.
6856 (*avx512f_vpermvar_truncv8div8si_1): Ditto.
6857 (*avx512f_permvar_truncv32hiv32qi_1): Ditto.
6858 (*avx512f_permvar_truncv16hiv16qi_1): Ditto.
6859 (*avx512f_permvar_truncv4div4si_1): Ditto.
6860 (*avx512f_pshufb_truncv8hiv8qi_1): Ditto.
6861 (*avx512f_pshufb_truncv4siv4hi_1): Ditto.
6862 (*avx512f_pshufd_truncv2div2si_1): Ditto.
6864 2021-08-16 Kito Cheng <kito.cheng@sifive.com>
6866 * config/riscv/multilib-generator: Support code model option for
6868 * doc/install.texi: Add document of new option for
6869 --with-multilib-generator.
6871 2021-08-15 Clément Chigot <clement.chigot@atos.net>
6873 * config/rs6000/rs6000.c (xcoff_tls_exec_model_detected): New.
6874 (rs6000_legitimize_tls_address_aix): Use it.
6875 (rs6000_xcoff_file_end): Add ".ref __tls_get_addr" when
6876 xcoff_tls_exec_model_detected is true.
6878 2021-08-15 Jeff Law <jlaw@localhost.localdomain>
6880 * config/h8300/h8300.c (shift_alg_si): Retune H8/300H shifts
6881 to allow a bit more code growth, saving many dozens of cycles.
6882 (h8300_option_override): Adjus shift_alg_si if optimizing for
6884 (get_shift_alg): Use special + inline shifts for residuals
6887 2021-08-14 Stafford Horne <shorne@gmail.com>
6890 * config/or1k/or1k-opts.h: New file.
6891 * config/or1k/or1k.c (or1k_legitimize_address_1, print_reloc):
6892 Support generating gotha relocations if -mcmodel=large is
6894 * config/or1k/or1k.h (TARGET_CMODEL_SMALL, TARGET_CMODEL_LARGE):
6896 * config/or1k/or1k.opt (mcmodel=): New option.
6897 * doc/invoke.texi (OpenRISC Options): Document mcmodel.
6899 2021-08-14 Martin Sebor <msebor@redhat.com>
6901 PR middle-end/101791
6902 * gimple-ssa-warn-access.cc (new_delete_mismatch_p): Use new argument
6903 to valid_new_delete_pair_p.
6904 * tree.c (valid_new_delete_pair_p): Add argument.
6905 * tree.h (valid_new_delete_pair_p): Same.
6907 2021-08-14 Jakub Jelinek <jakub@redhat.com>
6910 * config/i386/i386-expand.c (expand_vec_perm_broadcast_1)
6911 <case E_V64QImode>: For this mode assert
6912 !TARGET_AVX512BW || d->perm[0] rather than !TARGET_AVX2 || d->perm[0].
6914 2021-08-13 Michael Meissner <meissner@linux.ibm.com>
6917 * config/rs6000/altivec.md (xxeval): Use register_predicate
6918 instead of altivec_register_predicate.
6920 2021-08-13 Martin Sebor <msebor@redhat.com>
6922 PR middle-end/101734
6923 * tree-ssa-uninit.c (maybe_warn_read_write_only): New function.
6924 (maybe_warn_operand): Call it.
6926 2021-08-13 Martin Liska <mliska@suse.cz>
6929 * attribs.c (decl_attributes): Make naked functions "noipa"
6932 2021-08-13 Martin Liska <mliska@suse.cz>
6935 * symtab.c (symtab_node::noninterposable_alias): Do not create
6936 local aliases for target_clone functions as the clonning pass
6939 2021-08-13 Martin Liska <mliska@suse.cz>
6941 * opts.c (LIVE_PATCHING_OPTION): Define.
6942 (control_options_for_live_patching): Use it in error messages.
6944 2021-08-13 Jan Hubicka <hubicka@ucw.cz>
6946 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOREAD.
6947 (implicit_const_eaf_flags, implicit_pure_eaf_flags,
6948 ignore_stores_eaf_flags): New constants.
6949 (remove_useless_eaf_flags): New function.
6950 (eaf_flags_useful_p): Use it.
6951 (deref_flags): Add EAF_NOT_RETURNED if flag is unused;
6953 (modref_lattice::init): Add EAF_NOREAD.
6954 (modref_lattice::add_escape_point): Do not reacord escape point if
6956 (modref_lattice::merge): EAF_NOESCAPE implies EAF_NODIRECTESCAPE;
6957 use remove_useless_eaf_flags.
6958 (modref_lattice::merge_deref): Use ignore_stores_eaf_flags.
6959 (modref_lattice::merge_direct_load): Add EAF_NOREAD
6960 (analyze_ssa_name_flags): Fix handling EAF_NOT_RETURNED
6961 (analyze_parms): Use remove_useless_eaf_flags.
6962 (ipa_merge_modref_summary_after_inlining): Use ignore_stores_eaf_flags.
6963 (modref_merge_call_site_flags): Add caller and ecf_flags parameter;
6964 use remove_useless_eaf_flags.
6965 (modref_propagate_flags_in_scc): Update.
6966 * ipa-modref.h: Turn eaf_flags_t back to char.
6967 * tree-core.h (EAF_NOT_RETURNED): Fix.
6968 (EAF_NOREAD): New constant
6969 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1): Check for
6971 * tree-ssa-structalias.c (handle_rhs_call): Handle new flags.
6972 (handle_pure_call): Likewise.
6974 2021-08-12 Jakub Jelinek <jakub@redhat.com>
6976 * tree.def (OMP_MASKED): New tree code.
6977 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_FILTER.
6978 * tree.h (OMP_MASKED_BODY, OMP_MASKED_CLAUSES, OMP_MASKED_COMBINED,
6979 OMP_CLAUSE_FILTER_EXPR): Define.
6980 * tree.c (omp_clause_num_ops): Add OMP_CLAUSE_FILTER entry.
6981 (omp_clause_code_name): Likewise.
6982 (walk_tree_1): Handle OMP_CLAUSE_FILTER.
6983 * tree-nested.c (convert_nonlocal_omp_clauses,
6984 convert_local_omp_clauses): Handle OMP_CLAUSE_FILTER.
6985 (convert_nonlocal_reference_stmt, convert_local_reference_stmt,
6986 convert_gimple_call): Handle GIMPLE_OMP_MASTER.
6987 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_FILTER.
6988 (dump_generic_node): Handle OMP_MASTER.
6989 * gimple.def (GIMPLE_OMP_MASKED): New gimple code.
6990 * gimple.c (gimple_build_omp_masked): New function.
6991 (gimple_copy): Handle GIMPLE_OMP_MASKED.
6992 * gimple.h (gimple_build_omp_masked): Declare.
6993 (gimple_has_substatements): Handle GIMPLE_OMP_MASKED.
6994 (gimple_omp_masked_clauses, gimple_omp_masked_clauses_ptr,
6995 gimple_omp_masked_set_clauses): New inline functions.
6996 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_MASKED.
6997 * gimple-pretty-print.c (dump_gimple_omp_masked): New function.
6998 (pp_gimple_stmt_1): Handle GIMPLE_OMP_MASKED.
6999 * gimple-walk.c (walk_gimple_stmt): Likewise.
7000 * gimple-low.c (lower_stmt): Likewise.
7001 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
7002 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_FILTER. For clauses
7003 that take one expression rather than decl or constant, force
7004 gimplification of that into a SSA_NAME or temporary unless min
7006 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_FILTER.
7007 (gimplify_expr): Handle OMP_MASKED.
7008 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_MASKED.
7009 (estimate_num_insns): Likewise.
7010 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_FILTER.
7011 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_MASKED. Adjust
7012 diagnostics for existence of masked construct.
7013 (scan_omp_1_stmt, lower_omp_master, lower_omp_1, diagnose_sb_1,
7014 diagnose_sb_2): Handle GIMPLE_OMP_MASKED.
7015 * omp-expand.c (expand_omp_synch, expand_omp, omp_make_gimple_edges):
7018 2021-08-12 Uroš Bizjak <ubizjak@gmail.com>
7021 * config/i386/i386.md (avx512f_scalef<mode>2): New insn pattern.
7022 (ldexp<mode>3): Use avx512f_scalef<mode>2.
7023 (UNSPEC_SCALEF): Move from ...
7024 * config/i386/sse.md (UNSPEC_SCALEF): ... here.
7026 2021-08-12 Jan Hubicka <hubicka@ucw.cz>
7028 * ipa-split.c (consider_split): Fix condition testing void functions.
7030 2021-08-12 Aldy Hernandez <aldyh@redhat.com>
7032 * doc/invoke.texi: Remove docs for threader-mode param.
7033 * flag-types.h (enum threader_mode): Remove.
7034 * params.opt: Remove threader-mode param.
7035 * tree-ssa-threadbackward.c (class back_threader): Remove
7036 path_is_unreachable_p.
7037 Make find_paths private.
7038 Add maybe_thread and thread_through_all_blocks.
7039 Remove reference marker for m_registry.
7040 Remove reference marker for m_profit.
7041 (back_threader::back_threader): Adjust for registry and profit not
7043 (dump_path): Move down.
7045 (class thread_jumps): Remove.
7046 (class back_threader_registry): Remove m_all_paths.
7048 (thread_jumps::thread_through_all_blocks): Move to back_threader
7050 (fsm_find_thread_path): Remove
7051 (back_threader::maybe_thread): New.
7052 (back_threader::thread_through_all_blocks): Move from
7054 (back_threader_registry::back_threader_registry): Remove
7056 (back_threader_registry::~back_threader_registry): Remove.
7057 (thread_jumps::find_taken_edge): Remove.
7058 (thread_jumps::check_subpath_and_update_thread_path): Remove.
7059 (thread_jumps::maybe_register_path): Remove.
7060 (thread_jumps::handle_phi): Remove.
7061 (handle_assignment_p): Remove.
7062 (thread_jumps::handle_assignment): Remove.
7063 (thread_jumps::fsm_find_control_statement_thread_paths): Remove.
7064 (thread_jumps::find_jump_threads_backwards): Remove.
7065 (thread_jumps::find_jump_threads_backwards_with_ranger): Remove.
7066 (try_thread_blocks): Rename find_jump_threads_backwards to
7068 (pass_early_thread_jumps::execute): Same.
7070 2021-08-12 Tobias Burnus <tobias@codesourcery.com>
7072 * tree-core.h (omp_clause_proc_bind_kind): Add
7073 OMP_CLAUSE_PROC_BIND_PRIMARY.
7074 * tree-pretty-print.c (dump_omp_clause): Add TODO comment to
7075 change 'master' to 'primary' in proc_bind for OpenMP 5.1.
7077 2021-08-12 Claudiu Zissulescu <claziss@synopsys.com>
7079 * common/config/arc/arc-common.c (arc_option_init_struct): Remove
7080 fno-common reference.
7081 * config/arc/arc.c (arc_override_options): Remove overriding of
7084 2021-08-12 Jakub Jelinek <jakub@redhat.com>
7087 * config/i386/i386-expand.c (ix86_expand_vec_one_operand_perm_avx512):
7088 If d->testing_p, return true after performing checks instead of
7089 actually expanding the insn.
7090 (expand_vec_perm_broadcast_1): Handle V32HImode - assert
7091 !TARGET_AVX512BW and return false.
7093 2021-08-12 Eric Botcazou <ebotcazou@gcc.gnu.org>
7095 * configure.ac (PE linker --disable-dynamicbase support): New check.
7096 * configure: Regenerate.
7097 * config.in: Likewise.
7098 * config/i386/mingw32.h (LINK_SPEC_DISABLE_DYNAMICBASE): New define.
7099 (LINK_SPEC): Use it.
7100 * config/i386/mingw-w64.h (LINK_SPEC_DISABLE_DYNAMICBASE): Likewise.
7101 (LINK_SPEC): Likewise.
7103 2021-08-12 liuhongt <hongtao.liu@intel.com>
7106 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_2): New
7107 post_reload define_insn_and_split.
7108 (*avx512bw_zero_extendv32qiv32hi2_2): Ditto.
7109 (*sse4_1_zero_extendv8qiv8hi2_4): Ditto.
7110 (*avx512f_zero_extendv16hiv16si2_2): Ditto.
7111 (*avx2_zero_extendv8hiv8si2_2): Ditto.
7112 (*sse4_1_zero_extendv4hiv4si2_4): Ditto.
7113 (*avx512f_zero_extendv8siv8di2_2): Ditto.
7114 (*avx2_zero_extendv4siv4di2_2): Ditto.
7115 (*sse4_1_zero_extendv2siv2di2_4): Ditto.
7116 (VI248_256, VI248_512, VI148_512, VI148_256, VI148_128): New
7119 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
7121 * config/rs6000/rs6000-builtin-new.def: Add always, power5, and
7124 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
7126 * config/rs6000/rs6000-builtin-new.def: Add vsx stanza.
7128 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
7130 * config/rs6000/rs6000-builtin-new.def: Finish altivec stanza.
7131 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
7132 initialization of pcvoid_type_node here...
7133 (altivec_init_builtins): ...from here.
7134 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add
7135 RS6000_BTI_const_ptr_void.
7136 (pcvoid_type_node): New macro.
7138 2021-08-11 Richard Biener <rguenther@suse.de>
7141 * tree-ssa-forwprop.c (pass_forwprop::execute): Do not decompose
7142 hard-register accesses.
7144 2021-08-11 Richard Biener <rguenther@suse.de>
7146 * tree-ssa-operands.c (operands_scanner::get_expr_operands):
7147 Do not look at COMPONENT_REF FIELD_DECLs TREE_THIS_VOLATILE
7148 to determine has_volatile_ops.
7150 2021-08-11 Eric Botcazou <ebotcazou@gcc.gnu.org>
7152 * cfgexpand.c (expand_used_vars): Reuse attribs local variable.
7154 2021-08-11 Jan Hubicka <hubicka@ucw.cz>
7155 Alexandre Oliva <oliva@adacore.com>
7157 * ipa-modref.c (modref_lattice::dump): Fix escape_point's min_flags
7159 (modref_lattice::merge_deref): Fix handling of indirect scape points.
7160 (update_escape_summary_1): Likewise.
7161 (update_escape_summary): Likewise.
7162 (ipa_merge_modref_summary_after_inlining): Likewise.
7164 2021-08-11 Richard Biener <rguenther@suse.de>
7166 PR middle-end/101858
7167 * fold-const.c (fold_binary_loc): Guard simplification
7168 of X < (cast) (1 << Y) to integer types.
7170 2021-08-11 Richard Biener <rguenther@suse.de>
7172 PR tree-optimization/101861
7173 * tree-vect-stmts.c (vectorizable_load): Fix error in
7174 previous change with regard to gather vectorization.
7176 2021-08-11 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
7179 * config/arm/arm_neon.h (vdup_n_s8): Replace call to builtin
7181 (vdup_n_s16): Likewise.
7182 (vdup_n_s32): Likewise.
7183 (vdup_n_s64): Likewise.
7184 (vdup_n_u8): Likewise.
7185 (vdup_n_u16): Likewise.
7186 (vdup_n_u32): Likewise.
7187 (vdup_n_u64): Likewise.
7188 (vdup_n_p8): Likewise.
7189 (vdup_n_p16): Likewise.
7190 (vdup_n_p64): Likewise.
7191 (vdup_n_f16): Likewise.
7192 (vdup_n_f32): Likewise.
7193 (vdupq_n_s8): Likewise.
7194 (vdupq_n_s16): Likewise.
7195 (vdupq_n_s32): Likewise.
7196 (vdupq_n_s64): Likewise.
7197 (vdupq_n_u8): Likewise.
7198 (vdupq_n_u16): Likewise.
7199 (vdupq_n_u32): Likewise.
7200 (vdupq_n_u64): Likewise.
7201 (vdupq_n_p8): Likewise.
7202 (vdupq_n_p16): Likewise.
7203 (vdupq_n_p64): Likewise.
7204 (vdupq_n_f16): Likewise.
7205 (vdupq_n_f32): Likewise.
7206 (vmov_n_s8): Replace call to builtin with call to corresponding
7208 (vmov_n_s16): Likewise.
7209 (vmov_n_s32): Likewise.
7210 (vmov_n_s64): Likewise.
7211 (vmov_n_u8): Likewise.
7212 (vmov_n_u16): Likewise.
7213 (vmov_n_u32): Likewise.
7214 (vmov_n_u64): Likewise.
7215 (vmov_n_p8): Likewise.
7216 (vmov_n_p16): Likewise.
7217 (vmov_n_f16): Likewise.
7218 (vmov_n_f32): Likewise.
7219 (vmovq_n_s8): Likewise.
7220 (vmovq_n_s16): Likewise.
7221 (vmovq_n_s32): Likewise.
7222 (vmovq_n_s64): Likewise.
7223 (vmovq_n_u8): Likewise.
7224 (vmovq_n_u16): Likewise.
7225 (vmovq_n_u32): Likewise.
7226 (vmovq_n_u64): Likewise.
7227 (vmovq_n_p8): Likewise.
7228 (vmovq_n_p16): Likewise.
7229 (vmovq_n_f16): Likewise.
7230 (vmovq_n_f32): Likewise.
7231 * config/arm/arm_neon_builtins.def: Remove entries for vdup_n.
7233 2021-08-11 liuhongt <hongtao.liu@intel.com>
7236 * config/i386/i386.md (ldexp<mode>3): Extend to vscalefs[sd]
7237 when TARGET_AVX512F and TARGET_SSE_MATH.
7239 2021-08-10 Jakub Jelinek <jakub@redhat.com>
7242 * config/i386/i386-expand.c (expand_vec_perm_even_odd): Return false
7243 for V32HImode if !TARGET_AVX512BW.
7244 (ix86_vectorize_vec_perm_const) <case E_V32HImode, case E_V64QImode>:
7245 If !TARGET_AVX512BW and TARGET_AVX512F and d.testing_p, don't fail
7246 early, but actually check the permutation.
7248 2021-08-10 Richard Biener <rguenther@suse.de>
7250 PR tree-optimization/101809
7251 * tree-vect-stmts.c (get_load_store_type): Allow emulated
7252 gathers with offset vector nunits being a constant multiple
7253 of the data vector nunits.
7254 (vect_get_gather_scatter_ops): Use the appropriate nunits
7255 for the offset vector defs.
7256 (vectorizable_store): Adjust call to
7257 vect_get_gather_scatter_ops.
7258 (vectorizable_load): Likewise. Handle the case of less
7259 offset vectors than data vectors.
7261 2021-08-10 Jakub Jelinek <jakub@redhat.com>
7264 * config/i386/sse.md (*avx512f_shuf_<shuffletype>64x2_1<mask_name>_1,
7265 *avx512f_shuf_<shuffletype>32x4_1<mask_name>_1): New define_insn
7268 2021-08-10 Richard Biener <rguenther@suse.de>
7270 PR tree-optimization/101801
7271 PR tree-optimization/101819
7272 * tree-vectorizer.h (vect_emulated_vector_p): Declare.
7273 * tree-vect-loop.c (vect_emulated_vector_p): New function.
7274 (vectorizable_reduction): Re-instantiate a check for emulated
7276 * tree-vect-stmts.c (vectorizable_shift): Likewise.
7277 (vectorizable_operation): Likewise. Cost emulated vector
7278 operations according to the scalar sequence synthesized by
7281 2021-08-10 Richard Biener <rguenther@suse.de>
7283 PR middle-end/101824
7284 * tree-nested.c (get_frame_field): Mark the COMPONENT_REF as
7285 volatile in case the variable was.
7287 2021-08-10 H.J. Lu <hjl.tools@gmail.com>
7290 * config/i386/constraints.md (BC): Document for integer SSE
7291 constant all bits set operand.
7292 (BF): New constraint for const floating-point all bits set
7294 * config/i386/i386.c (standard_sse_constant_p): Likewise.
7295 (standard_sse_constant_opcode): Likewise.
7296 * config/i386/sse.md (sseconstm1): New mode attribute.
7297 (mov<mode>_internal): Replace BC with <sseconstm1>.
7299 2021-08-10 liuhongt <hongtao.liu@intel.com>
7301 * config/i386/sse.md (cond_<insn><mode>): New expander.
7302 (VI248_AVX512VLBW): New mode iterator.
7303 * config/i386/predicates.md
7304 (nonimmediate_or_const_vec_dup_operand): New predicate.
7306 2021-08-09 Andrew MacLeod <amacleod@redhat.com>
7308 PR tree-optimization/101741
7309 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Check
7310 type of parameter for toupper/tolower.
7312 2021-08-09 Martin Jambor <mjambor@suse.cz>
7315 * ipa-prop.c (propagate_controlled_uses): Removed a spurious space.
7317 2021-08-09 Pat Haugen <pthaugen@linux.ibm.com>
7319 * config/rs6000/rs6000.c (is_load_insn1): Verify destination is a
7321 (is_store_insn1): Verify source is a register.
7323 2021-08-09 Uroš Bizjak <ubizjak@gmail.com>
7326 * config/i386/mmx.md (<any_logic:code>v2sf3):
7327 Rename from *mmx_<any_logic:code>v2sf3
7329 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7331 * config/nvptx/nvptx.c: Cross-reference parts adapted in
7332 'gcc/omp-oacc-neuter-broadcast.cc'.
7333 * omp-low.c: Likewise.
7334 * omp-oacc-neuter-broadcast.cc: Cross-reference parts adapted from
7337 2021-08-09 Julian Brown <julian@codesourcery.com>
7338 Kwok Cheung Yeung <kcy@codesourcery.com>
7339 Thomas Schwinge <thomas@codesourcery.com>
7341 * config/gcn/gcn.c (gcn_init_builtins): Override decls for
7342 BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START,
7343 BUILT_IN_GOACC_SINGLE_COPY_END and BUILT_IN_GOACC_BARRIER.
7344 (gcn_goacc_validate_dims): Turn on worker partitioning unconditionally.
7345 (gcn_fork_join): Update comment.
7346 * config/gcn/gcn.opt (flag_worker_partitioning): Remove.
7347 (macc_experimental_workers): Remove unused option.
7349 2021-08-09 Julian Brown <julian@codesourcery.com>
7350 Nathan Sidwell <nathan@codesourcery.com> (via 'gcc/config/nvptx/nvptx.c' master)
7351 Kwok Cheung Yeung <kcy@codesourcery.com>
7352 Thomas Schwinge <thomas@codesourcery.com>
7354 * Makefile.in (OBJS): Add omp-oacc-neuter-broadcast.o.
7355 * doc/tm.texi.in (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD):
7356 Add documentation hook.
7357 * doc/tm.texi: Regenerate.
7358 * omp-oacc-neuter-broadcast.cc: New file.
7359 * omp-builtins.def (BUILT_IN_GOACC_BARRIER)
7360 (BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START)
7361 (BUILT_IN_GOACC_SINGLE_COPY_END): New builtins.
7362 * passes.def (pass_omp_oacc_neuter_broadcast): Add pass.
7363 * target.def (goacc.create_worker_broadcast_record): Add target
7365 * tree-pass.h (make_pass_omp_oacc_neuter_broadcast): Add
7367 * config/gcn/gcn-protos.h (gcn_goacc_adjust_propagation_record):
7368 Rename prototype to...
7369 (gcn_goacc_create_worker_broadcast_record): ... this.
7370 * config/gcn/gcn-tree.c (gcn_goacc_adjust_propagation_record): Rename
7372 (gcn_goacc_create_worker_broadcast_record): ... this.
7373 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_PROPAGATION_RECORD):
7375 (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD): ... this.
7377 2021-08-09 Tejas Belagod <tejas.belagod@arm.com>
7380 * config/aarch64/aarch64-simd.md (vlshr<mode>3, vashr<mode>3): Use
7383 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7385 * Makefile.in (GTFILES): Remove '$(srcdir)/omp-offload.c'.
7387 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7389 * builtins.def (DEF_GOACC_BUILTIN, DEF_GOMP_BUILTIN): Don't
7390 consider '-foffload-abi'.
7391 * common.opt (-foffload-abi): Remove 'Var', 'Init'.
7392 * opts.c (common_handle_option) <-foffload-abi> [ACCEL_COMPILER]:
7395 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7397 * optc-gen.awk: Sanity check that 'Init' doesn't appear without
7400 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7402 * omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE): Remove.
7404 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7406 * doc/gty.texi (Files): Update.
7408 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
7410 * doc/gty.texi (Files): Fix GTY header file example.
7412 2021-08-09 Roger Sayle <roger@nextmovesoftware.com>
7414 * tree-ssa-ccp.c (value_mask_to_min_max): Helper function to
7415 determine the upper and lower bounds from a mask-value pair.
7416 (bit_value_unop) [ABS_EXPR, ABSU_EXPR]: Add support for
7417 absolute value and unsigned absolute value expressions.
7418 (bit_value_binop): Initialize *VAL's precision.
7419 [LT_EXPR, LE_EXPR]: Use value_mask_to_min_max to determine
7420 upper and lower bounds of operands. Add LE_EXPR/GE_EXPR
7421 support when the operands are unknown but potentially equal.
7422 [MIN_EXPR, MAX_EXPR]: Support minimum/maximum expressions.
7424 2021-08-09 Bin Cheng <bin.cheng@linux.alibaba.com>
7426 * config/aarch64/aarch64.md
7427 (*extend<SHORT:mode><GPI:mode>2_aarch64): Use %<GPI:w>0.
7429 2021-08-08 Sergei Trofimovich <siarheit@google.com>
7431 * lra-constraints.c: Fix s/otput/output/ typo.
7433 2021-08-06 Martin Sebor <msebor@redhat.com>
7435 * builtins.c (expand_builtin_memchr): Move to gimple-ssa-warn-access.cc.
7436 (expand_builtin_strcat): Same.
7437 (expand_builtin_stpncpy): Same.
7438 (expand_builtin_strncat): Same.
7439 (check_read_access): Same.
7440 (check_memop_access): Same.
7441 (expand_builtin_strlen): Move checks to gimple-ssa-warn-access.cc.
7442 (expand_builtin_strnlen): Same.
7443 (expand_builtin_memcpy): Same.
7444 (expand_builtin_memmove): Same.
7445 (expand_builtin_mempcpy): Same.
7446 (expand_builtin_strcpy): Same.
7447 (expand_builtin_strcpy_args): Same.
7448 (expand_builtin_stpcpy_1): Same.
7449 (expand_builtin_strncpy): Same.
7450 (expand_builtin_memset): Same.
7451 (expand_builtin_bzero): Same.
7452 (expand_builtin_strcmp): Same.
7453 (expand_builtin_strncmp): Same.
7454 (expand_builtin): Remove handlers.
7455 (fold_builtin_strlen): Add a comment.
7456 * builtins.h (check_access): Move to gimple-ssa-warn-access.cc.
7457 * calls.c (maybe_warn_nonstring_arg): Same.
7458 * diagnostic-spec.c (nowarn_spec_t::nowarn_spec_t): Add warning option.
7459 * gimple-fold.c (gimple_fold_builtin_strcpy): Pass argument to callee.
7460 (gimple_fold_builtin_stpcpy): Same.
7461 * gimple-ssa-warn-access.cc (has_location): New function.
7462 (get_location): Same.
7463 (get_callee_fndecl): Same.
7466 (warn_string_no_nul): Define.
7467 (unterminated_array): Same.
7468 (check_nul_terminated_array): Same.
7469 (maybe_warn_nonstring_arg): Same.
7470 (maybe_warn_for_bound): Same.
7471 (warn_for_access): Same.
7472 (check_access): Same.
7473 (check_memop_access): Same.
7474 (check_read_access): Same.
7475 (warn_dealloc_offset): Use helper functions.
7476 (maybe_emit_free_warning): Same.
7477 (class pass_waccess): Add members.
7478 (check_strcat): New function.
7479 (check_strncat): New function.
7480 (check_stxcpy): New function.
7481 (check_stxncpy): New function.
7482 (check_strncmp): New function.
7483 (pass_waccess::check_builtin): New function.
7484 (pass_waccess::check): Call it.
7485 * gimple-ssa-warn-access.h (warn_string_no_nul): Move here from
7487 (maybe_warn_for_bound): Same.
7488 (check_access): Same.
7489 (check_memop_access): Same.
7490 (check_read_access): Same.
7491 * pointer-query.h (struct access_data): Define a ctor overload.
7493 2021-08-06 Richard Biener <rguenther@suse.de>
7495 PR tree-optimization/101801
7496 * tree-vectorizer.h (vect_worthwhile_without_simd_p): Rename...
7497 (vect_can_vectorize_without_simd_p): ... to this.
7498 * tree-vect-loop.c (vect_worthwhile_without_simd_p): Rename...
7499 (vect_can_vectorize_without_simd_p): ... to this and fold
7500 in vect_min_worthwhile_factor.
7501 (vect_min_worthwhile_factor): Remove.
7502 (vectorizable_reduction): Adjust and remove the cost part.
7503 * tree-vect-stmts.c (vectorizable_shift): Likewise.
7504 (vectorizable_operation): Likewise.
7506 2021-08-06 Uroš Bizjak <ubizjak@gmail.com>
7509 * config/i386/i386.md (cmove reg-to-reg move elimination peephole2s):
7510 Add general_gr_operand predicate to operand 3.
7512 2021-08-06 Roger Sayle <roger@nextmovesoftware.com>
7514 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern): Use
7515 CFN_BUILT_IN_CLRSB* instead of BUILT_IN_CLRSB* for consistency.
7517 2021-08-06 Tamar Christina <tamar.christina@arm.com>
7519 * config/aarch64/aarch64-sve-builtins.cc (register_svpattern,
7520 register_svprfop): Pass vec<> by pointer.
7521 * langhooks-def.h (lhd_simulate_enum_decl): Likewise.
7522 * langhooks.c (lhd_simulate_enum_decl): Likewise.
7523 * langhooks.h (struct lang_hooks_for_types): Likewise.
7525 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
7527 * config/aarch64/arm_neon.h (vst1_bf16_x2): Use
7528 __builtin_memcpy instead of constructing an additional
7529 __builtin_aarch64_simd_oi one vector at a time.
7530 (vst1q_bf16_x2): Likewise.
7531 (vst1_bf16_x3): Use __builtin_memcpy instead of constructing
7532 an additional __builtin_aarch64_simd_ci one vector at a time.
7533 (vst1q_bf16_x3): Likewise.
7534 (vst1_bf16_x4): Use __builtin_memcpy instead of a union.
7535 (vst1q_bf16_x4): Likewise.
7536 (vst2_bf16): Use __builtin_memcpy instead of constructing an
7537 additional __builtin_aarch64_simd_oi one vector at a time.
7538 (vst2q_bf16): Likewise.
7539 (vst3_bf16): Use __builtin_memcpy instead of constructing an
7540 additional __builtin_aarch64_simd_ci mode one vector at a
7542 (vst3q_bf16): Likewise.
7543 (vst4_bf16): Use __builtin_memcpy instead of constructing an
7544 additional __builtin_aarch64_simd_xi one vector at a time.
7545 (vst4q_bf16): Likewise.
7547 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
7549 * config/aarch64/arm_neon.h (__ST2_LANE_FUNC): Delete.
7550 (__ST2Q_LANE_FUNC): Delete.
7551 (vst2_lane_f16): Use __builtin_memcpy to copy vector
7552 structure instead of constructing __builtin_aarch64_simd_oi
7553 one vector at a time.
7554 (vst2_lane_f32): Likewise.
7555 (vst2_lane_f64): Likewise.
7556 (vst2_lane_p8): Likewise.
7557 (vst2_lane_p16): Likewise.
7558 (vst2_lane_p64): Likewise.
7559 (vst2_lane_s8): Likewise.
7560 (vst2_lane_s16): Likewise.
7561 (vst2_lane_s32): Likewise.
7562 (vst2_lane_s64): Likewise.
7563 (vst2_lane_u8): Likewise.
7564 (vst2_lane_u16): Likewise.
7565 (vst2_lane_u32): Likewise.
7566 (vst2_lane_u64): Likewise.
7567 (vst2_lane_bf16): Likewise.
7568 (vst2q_lane_f16): Use __builtin_memcpy to copy vector
7569 structure instead of using a union.
7570 (vst2q_lane_f32): Likewise.
7571 (vst2q_lane_f64): Likewise.
7572 (vst2q_lane_p8): Likewise.
7573 (vst2q_lane_p16): Likewise.
7574 (vst2q_lane_p64): Likewise.
7575 (vst2q_lane_s8): Likewise.
7576 (vst2q_lane_s16): Likewise.
7577 (vst2q_lane_s32): Likewise.
7578 (vst2q_lane_s64): Likewise.
7579 (vst2q_lane_u8): Likewise.
7580 (vst2q_lane_u16): Likewise.
7581 (vst2q_lane_u32): Likewise.
7582 (vst2q_lane_u64): Likewise.
7583 (vst2q_lane_bf16): Likewise.
7585 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
7587 * config/aarch64/arm_neon.h (__ST3_LANE_FUNC): Delete.
7588 (__ST3Q_LANE_FUNC): Delete.
7589 (vst3_lane_f16): Use __builtin_memcpy to copy vector
7590 structure instead of constructing __builtin_aarch64_simd_ci
7591 one vector at a time.
7592 (vst3_lane_f32): Likewise.
7593 (vst3_lane_f64): Likewise.
7594 (vst3_lane_p8): Likewise.
7595 (vst3_lane_p16): Likewise.
7596 (vst3_lane_p64): Likewise.
7597 (vst3_lane_s8): Likewise.
7598 (vst3_lane_s16): Likewise.
7599 (vst3_lane_s32): Likewise.
7600 (vst3_lane_s64): Likewise.
7601 (vst3_lane_u8): Likewise.
7602 (vst3_lane_u16): Likewise.
7603 (vst3_lane_u32): Likewise.
7604 (vst3_lane_u64): Likewise.
7605 (vst3_lane_bf16): Likewise.
7606 (vst3q_lane_f16): Use __builtin_memcpy to copy vector
7607 structure instead of using a union.
7608 (vst3q_lane_f32): Likewise.
7609 (vst3q_lane_f64): Likewise.
7610 (vst3q_lane_p8): Likewise.
7611 (vst3q_lane_p16): Likewise.
7612 (vst3q_lane_p64): Likewise.
7613 (vst3q_lane_s8): Likewise.
7614 (vst3q_lane_s16): Likewise.
7615 (vst3q_lane_s32): Likewise.
7616 (vst3q_lane_s64): Likewise.
7617 (vst3q_lane_u8): Likewise.
7618 (vst3q_lane_u16): Likewise.
7619 (vst3q_lane_u32): Likewise.
7620 (vst3q_lane_u64): Likewise.
7621 (vst3q_lane_bf16): Likewise.
7623 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
7625 * config/aarch64/arm_neon.h (__ST4_LANE_FUNC): Delete.
7626 (__ST4Q_LANE_FUNC): Delete.
7627 (vst4_lane_f16): Use __builtin_memcpy to copy vector
7628 structure instead of constructing __builtin_aarch64_simd_xi
7629 one vector at a time.
7630 (vst4_lane_f32): Likewise.
7631 (vst4_lane_f64): Likewise.
7632 (vst4_lane_p8): Likewise.
7633 (vst4_lane_p16): Likewise.
7634 (vst4_lane_p64): Likewise.
7635 (vst4_lane_s8): Likewise.
7636 (vst4_lane_s16): Likewise.
7637 (vst4_lane_s32): Likewise.
7638 (vst4_lane_s64): Likewise.
7639 (vst4_lane_u8): Likewise.
7640 (vst4_lane_u16): Likewise.
7641 (vst4_lane_u32): Likewise.
7642 (vst4_lane_u64): Likewise.
7643 (vst4_lane_bf16): Likewise.
7644 (vst4q_lane_f16): Use __builtin_memcpy to copy vector
7645 structure instead of using a union.
7646 (vst4q_lane_f32): Likewise.
7647 (vst4q_lane_f64): Likewise.
7648 (vst4q_lane_p8): Likewise.
7649 (vst4q_lane_p16): Likewise.
7650 (vst4q_lane_p64): Likewise.
7651 (vst4q_lane_s8): Likewise.
7652 (vst4q_lane_s16): Likewise.
7653 (vst4q_lane_s32): Likewise.
7654 (vst4q_lane_s64): Likewise.
7655 (vst4q_lane_u8): Likewise.
7656 (vst4q_lane_u16): Likewise.
7657 (vst4q_lane_u32): Likewise.
7658 (vst4q_lane_u64): Likewise.
7659 (vst4q_lane_bf16): Likewise.
7661 2021-08-06 Martin Liska <mliska@suse.cz>
7663 * config/rs6000/rs6000.c (rs6000_option_override_internal): When
7664 a target option is restored, it can have
7665 rs6000_long_double_type_size set to FLOAT_PRECISION_TFmode
7666 and error should not be emitted.
7668 2021-08-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
7670 * gcov-io.h (gcov_write): Declare.
7671 * gcov-io.c (gcov_write): New.
7672 (gcov_write_counter): Remove.
7673 (gcov_write_tag_length): Likewise.
7674 (gcov_write_summary): Replace gcov_write_tag_length() with calls to
7675 gcov_write_unsigned().
7676 * doc/invoke.texi (fprofile-info-section): Mention
7677 __gcov_info_to_gdca().
7679 2021-08-06 Martin Sebor <msebor@redhat.com>
7681 * dominance.c (prune_bbs_to_update_dominators): Adjust by-value vec
7682 arguments to by-reference.
7683 (iterate_fix_dominators): Same.
7684 * dominance.h (iterate_fix_dominators): Same.
7685 * ipa-prop.h: Call auto_vec::to_vec_legacy.
7686 * tree-data-ref.c (dump_data_dependence_relation): Adjust by-value vec
7687 arguments to by-reference.
7688 (debug_data_dependence_relation): Same.
7689 (dump_data_dependence_relations): Same.
7690 * tree-data-ref.h (debug_data_dependence_relation): Same.
7691 (dump_data_dependence_relations): Same.
7692 * tree-predcom.c (dump_chains): Same.
7693 (initialize_root_vars_lm): Same.
7694 (determine_unroll_factor): Same.
7695 (replace_phis_by_defined_names): Same.
7696 (insert_init_seqs): Same.
7697 (pcom_worker::tree_predictive_commoning_loop): Call
7698 auto_vec::to_vec_legacy.
7699 * tree-ssa-pre.c (insert_into_preds_of_block): Adjust by-value vec
7700 arguments to by-reference.
7701 * tree-ssa-threadbackward.c (populate_worklist): Same.
7702 (back_threader::resolve_def): Same.
7703 * tree-vect-data-refs.c (vect_check_nonzero_value): Same.
7704 (vect_enhance_data_refs_alignment): Same.
7705 (vect_check_lower_bound): Same.
7706 (vect_prune_runtime_alias_test_list): Same.
7707 (vect_permute_store_chain): Same.
7708 * tree-vect-slp-patterns.c (vect_normalize_conj_loc): Same.
7709 * tree-vect-stmts.c (vect_create_vectorized_demotion_stmts): Same.
7710 * tree-vectorizer.h (vect_permute_store_chain): Same.
7711 * vec.c (test_init): New function.
7712 (vec_c_tests): Call new function.
7713 * vec.h (vec): Declare ctors, dtor, and assignment.
7714 (auto_vec::vec_to_legacy): New function.
7715 (vec::copy): Adjust initialization.
7717 2021-08-05 H.J. Lu <hjl.tools@gmail.com>
7720 * config/i386/i386.c (ix86_can_inline_p): Ignore MASK_80387 if
7721 callee only uses GPRs.
7722 * config/i386/ia32intrin.h: Revert commit 5463cee2770.
7723 * config/i386/serializeintrin.h: Revert commit 71958f740f1.
7724 * config/i386/x86gprintrin.h: Add
7725 #pragma GCC target("general-regs-only") and #pragma GCC pop_options
7726 to disable non-GPR ISAs.
7728 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
7730 PR middle-end/101787
7731 * doc/md.texi (cond_ashl, cond_ashr, cond_lshr): Document.
7733 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
7735 * tree-vectorizer.h (vect_is_store_elt_extraction, vect_is_reduction)
7736 (vect_reduc_type, vect_embedded_comparison_type, vect_comparison_type)
7737 (vect_is_extending_load, vect_is_integer_truncation): New functions,
7738 moved from aarch64.c but given different names.
7739 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction)
7740 (aarch64_is_reduction, aarch64_reduc_type)
7741 (aarch64_embedded_comparison_type, aarch64_comparison_type)
7742 (aarch64_extending_load_p, aarch64_integer_truncation_p): Delete
7743 in favor of the above. Update callers accordingly.
7745 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
7748 * config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
7749 writing .cpu directive in asm output.
7750 * config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
7751 (arm_last_printed_arch_string): Delete.
7752 (arm_last-printed_fpu_string): Delete.
7753 (arm_configure_build_target): If use of floating-point/SIMD is
7754 disabled, remove all fp/simd related features from the target ISA.
7755 (last_arm_targ_options): New variable.
7756 (arm_print_asm_arch_directives): Add new parameters. Change order
7757 of emitted directives and handle all cases here.
7758 (arm_file_start): Always call arm_print_asm_arch_directives, move
7759 all generation of .arch/.arch_extension here.
7760 (arm_file_end): Call arm_print_asm_arch.
7761 (arm_declare_function_name): Call arm_print_asm_arch_directives
7762 instead of printing .arch/.fpu directives directly.
7764 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
7766 * config/arm/arm.c (arm_configure_build_target): Don't call
7767 arm_option_reconfigure_globals.
7768 (arm_option_restore): Call arm_option_reconfigure_globals after
7769 reconfiguring the target.
7770 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
7772 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
7774 * config/arm/arm.c (arm_configure_build_target): Ensure the target's
7775 arch_name is always set.
7777 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
7779 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
7780 of vec_select high-half from being added into Neon subtract
7783 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
7785 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
7786 of vec_select high-half from being added into Neon add cost.
7788 2021-08-05 Kewen Lin <linkw@linux.ibm.com>
7790 * cfgloop.h (loops_list::loops_list): Add one optional argument
7791 root and adjust accordingly, update loop tree walking and factor
7793 * cfgloop.c (loops_list::walk_loop_tree): ... this. New function.
7795 2021-08-05 Eric Botcazou <ebotcazou@gcc.gnu.org>
7797 PR tree-optimization/101626
7798 * tree-sra.c (propagate_subaccesses_from_rhs): Do not set the
7799 reverse scalar storage order on a pointer or vector component.
7801 2021-08-05 liuhongt <hongtao.liu@intel.com>
7803 * config/i386/sse.md (cond_<code><mode>): New expander.
7805 2021-08-05 liuhongt <hongtao.liu@intel.com>
7807 * config/i386/sse.md (cond_<code><mode>): New expander.
7809 2021-08-05 liuhongt <hongtao.liu@intel.com>
7811 * config/i386/sse.md (cond_<code><mode>): New expander.
7813 2021-08-04 David Malcolm <dmalcolm@redhat.com>
7816 * Makefile.in (ANALYZER_OBJS): Add analyzer/region-model-asm.o.
7818 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
7821 * config/i386/i386.h (STORE_MAX_PIECES): Allow 16/32/64 bytes
7822 only if TARGET_INTER_UNIT_MOVES_TO_VEC is true.
7824 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
7827 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
7828 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
7829 data with SSE register from one memory location to another.
7831 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7833 * config/s390/s390.c (expand_perm_with_vpdi): New function.
7834 (vectorize_vec_perm_const_1): Call expand_perm_with_vpdi.
7835 * config/s390/vector.md (*vpdi1<mode>, @vpdi1<mode>): Enable a
7836 parameterized expander.
7837 (*vpdi4<mode>, @vpdi4<mode>): Likewise.
7839 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7841 * config/s390/s390.c (MAX_VECT_LEN): Define macro.
7842 (struct expand_vec_perm_d): Define struct.
7843 (expand_perm_with_merge): New function.
7844 (vectorize_vec_perm_const_1): New function.
7845 (s390_vectorize_vec_perm_const): New function.
7846 (TARGET_VECTORIZE_VEC_PERM_CONST): Define target macro.
7848 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7850 * config/s390/vector.md (V_HW_64): Remove mode iterator.
7851 (*vec_load_pair<mode>): Use V_HW_2 instead of V_HW_64.
7852 * config/s390/vx-builtins.md
7853 (vec_scatter_element<V_HW_2:mode>_SI): Use V_HW_2 instead of
7856 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7858 * config/s390/s390.md (UNSPEC_VEC_PERMI): Remove constant
7860 * config/s390/vector.md (*vpdi1<mode>, *vpdi4<mode>): New pattern
7862 * config/s390/vx-builtins.md (*vec_permi<mode>): Emit generic rtx
7863 instead of an unspec.
7865 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7867 * config/s390/s390-modes.def: Add more vector modes to support
7868 concatenation of two vectors.
7869 * config/s390/s390-protos.h (s390_expand_merge_perm_const): Add
7871 (s390_expand_merge): Likewise.
7872 * config/s390/s390.c (s390_expand_merge_perm_const): New function.
7873 (s390_expand_merge): New function.
7874 * config/s390/s390.md (UNSPEC_VEC_MERGEH, UNSPEC_VEC_MERGEL):
7875 Remove constant definitions.
7876 * config/s390/vector.md (V_HW_2): Add mode iterators.
7877 (VI_HW_4, V_HW_4): Rename VI_HW_4 to V_HW_4.
7878 (vec_2x_nelts, vec_2x_wide): New mode attributes.
7879 (*vmrhb, *vmrlb, *vmrhh, *vmrlh, *vmrhf, *vmrlf, *vmrhg, *vmrlg):
7880 New pattern definitions.
7881 (vec_widen_umult_lo_<mode>, vec_widen_umult_hi_<mode>)
7882 (vec_widen_smult_lo_<mode>, vec_widen_smult_hi_<mode>)
7883 (vec_unpacks_lo_v4sf, vec_unpacks_hi_v4sf, vec_unpacks_lo_v2df)
7884 (vec_unpacks_hi_v2df): Adjust expanders to emit non-unspec RTX for
7886 * config/s390/vx-builtins.md (V_HW_4): Remove mode iterator. Now
7888 (vec_mergeh<mode>, vec_mergel<mode>): Use s390_expand_merge to
7889 emit vec merge pattern.
7891 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
7893 * config/aarch64/aarch64.c (aarch64_strip_extend_vec_half):
7895 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent cost of
7896 vec_select high-half from being added into Neon multiply
7898 * rtlanal.c (vec_series_highpart_p): Define.
7899 * rtlanal.h (vec_series_highpart_p): Declare.
7901 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
7903 * config/aarch64/aarch64.c (aarch64_strip_duplicate_vec_elt):
7905 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent
7906 vec_select cost from being added into Neon multiply cost.
7908 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
7910 * tree-vect-loop.c (vect_better_loop_vinfo_p): Detect cases in
7911 which old_loop_vinfo is an epilogue loop that handles a constant
7912 number of iterations.
7914 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
7916 * tree-vect-loop.c (vect_analyze_loop): Print a dump message
7917 when a reanalyzed loop fails to be cheaper than the current
7920 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
7922 * config/aarch64/aarch64.c: Fix a typo.
7924 2021-08-04 Vincent Lefèvre <vincent-gcc@vinc17.net>
7926 PR gcov-profile/101773
7927 * gcov-io.c (gcov_close): Check return code of a fclose.
7929 2021-08-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
7932 * dwarf2out.c (dwarf2out_assembly_start): Emit a dummy
7933 .file statement when needed.
7935 2021-08-04 Richard Biener <rguenther@suse.de>
7937 * tree-vect-data-refs.c (vect_check_gather_scatter):
7938 Include widening conversions only when the result is
7939 still handed by native gather or the current offset
7940 size not already matches the data size.
7941 Also succeed analysis in case there's no native support,
7942 noted by a IFN_LAST ifn and a NULL decl.
7943 (vect_analyze_data_refs): Always consider gathers.
7944 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
7945 Test for no IFN gather rather than decl gather.
7946 * tree-vect-stmts.c (vect_model_load_cost): Pass in the
7947 gather-scatter info and cost emulated gathers accordingly.
7948 (vect_truncate_gather_scatter_offset): Properly test for
7950 (vect_use_strided_gather_scatters_p): Likewise.
7951 (get_load_store_type): Handle emulated gathers and its
7953 (vectorizable_load): Likewise. Emulate them by extracting
7954 scalar offsets, doing scalar loads and a vector construct.
7956 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
7959 * expr.c (op_by_pieces_d::op_by_pieces_d): Add a max_pieces
7960 argument to set m_max_size.
7961 (move_by_pieces_d): Pass MOVE_MAX_PIECES to op_by_pieces_d.
7962 (store_by_pieces_d): Pass STORE_MAX_PIECES to op_by_pieces_d.
7963 (compare_by_pieces_d): Pass COMPARE_MAX_PIECES to op_by_pieces_d.
7965 2021-08-04 Roger Sayle <roger@nextmovesoftware.com>
7966 Marc Glisse <marc.glisse@inria.fr>
7968 * match.pd (bit_ior, bit_xor): Canonicalize (X*C1)|(X*C2) and
7969 (X*C1)^(X*C2) as X*(C1+C2), and related variants, using
7970 tree_nonzero_bits to ensure that operands are bit-wise disjoint.
7972 2021-08-04 Richard Biener <rguenther@suse.de>
7974 * tree-ssa-forwprop.c (pass_forwprop::execute): Split
7975 out code to decompose vector loads ...
7976 (optimize_vector_load): ... here. Generalize it to
7977 handle intermediate widening and TARGET_MEM_REF loads
7978 and apply it to loads with a supported vector mode as well.
7980 2021-08-04 Richard Biener <rguenther@suse.de>
7982 PR tree-optimization/101756
7983 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Make sure
7984 the result of the reduction epilogue is compatible to the original
7987 2021-08-04 liuhongt <hongtao.liu@intel.com>
7990 * config/i386/i386.md (peephole2): Refine predicate from
7991 register_operand to general_reg_operand.
7993 2021-08-04 Aldy Hernandez <aldyh@redhat.com>
7995 * gimple-range-path.h (path_range_query::dump): Mark override.
7997 2021-08-04 Richard Biener <rguenther@suse.de>
7999 PR tree-optimization/101769
8000 * tree-tailcall.c (eliminate_tail_call): Add the created loop
8001 for the first recursion and return it via the new output parameter.
8002 (optimize_tail_call): Pass through new output param.
8003 (tree_optimize_tail_calls_1): After creating all latches,
8004 add the created loop to the loop tree. Do not mark loops for fixup.
8006 2021-08-04 Martin Liska <mliska@suse.cz>
8008 * doc/invoke.texi: Document threader-mode param.
8010 2021-08-04 liuhongt <hongtao.liu@intel.com>
8012 * config/i386/sse.md (cond_fma<mode>): New expander.
8013 (cond_fms<mode>): Ditto.
8014 (cond_fnma<mode>): Ditto.
8015 (cond_fnms<mode>): Ditto.
8017 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
8019 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode>): Use && instead of &.
8021 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
8023 * config/rs6000/constraints.md: Remove "e" from the list of available
8024 constraint characters.
8026 2021-08-03 Eugene Rozenfeld <erozen@microsoft.com>
8028 PR gcov-profile/71672
8029 * auto-profile.c (afdo_indirect_call): Fix setup of the historgram value for indirect calls.
8031 2021-08-03 Paul A. Clarke <pc@us.ibm.com>
8033 * config/rs6000/smmintrin.h (_mm_minpos_epu16): New.
8035 2021-08-03 H.J. Lu <hjl.tools@gmail.com>
8037 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): In 64-bit mode,
8038 try XMM31 to avoid vzeroupper.
8040 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8042 * doc/invoke.texi: Document -mtune=neoverse-512tvb and
8043 -mcpu=neoverse-512tvb.
8044 * config/aarch64/aarch64-cores.def (neoverse-512tvb): New entry.
8045 * config/aarch64/aarch64-tune.md: Regenerate.
8046 * config/aarch64/aarch64.c (neoverse512tvb_sve_vector_cost)
8047 (neoverse512tvb_sve_issue_info, neoverse512tvb_vec_issue_info)
8048 (neoverse512tvb_vector_cost, neoverse512tvb_tunings): New structures.
8049 (aarch64_adjust_body_cost_sve): Handle -mtune=neoverse-512tvb.
8050 (aarch64_adjust_body_cost): Likewise.
8052 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8054 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Only
8055 record issue information for operations that occur in the
8058 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8060 * config/aarch64/aarch64.c (aarch64_multiply_add_p): Add a vec_flags
8061 parameter. Detect cases in which an Advanced SIMD MLA would almost
8062 certainly require a MOV.
8063 (aarch64_count_ops): Update accordingly.
8065 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8067 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction): New
8068 function, split out from...
8069 (aarch64_detect_vector_stmt_subtype): ...here.
8070 (aarch64_add_stmt_cost): Treat extracting element 0 as free.
8072 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8074 * config/aarch64/aarch64-protos.h (sve_vec_cost):
8075 Add gather_load_x32_cost and gather_load_x64_cost.
8076 * config/aarch64/aarch64.c (generic_sve_vector_cost)
8077 (a64fx_sve_vector_cost, neoversev1_sve_vector_cost): Update
8078 accordingly, using the values given by the scalar_load * number
8079 of elements calculation that we used previously.
8080 (aarch64_detect_vector_stmt_subtype): Use the new fields.
8082 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8084 * config/aarch64/aarch64.c (aarch64_adjust_body_cost_sve): New
8085 function, split out from...
8086 (aarch64_adjust_body_cost): ...here.
8088 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8090 * config/aarch64/fractional-cost.h: New file.
8091 * config/aarch64/aarch64.c: Include <algorithm> (indirectly)
8092 and cost_fraction.h.
8093 (vec_cost_fraction): New typedef.
8094 (aarch64_detect_scalar_stmt_subtype): Use it for statement costs.
8095 (aarch64_detect_vector_stmt_subtype): Likewise.
8096 (aarch64_sve_adjust_stmt_cost, aarch64_adjust_stmt_cost): Likewise.
8097 (aarch64_estimate_min_cycles_per_iter): Use vec_cost_fraction
8099 (aarch64_adjust_body_cost): Likewise.
8100 (aarch64_test_cost_fraction): New function.
8101 (aarch64_run_selftests): Call it.
8103 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
8105 * config/aarch64/aarch64-protos.h (tune_params::sve_width): Turn
8107 * config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Update
8109 (aarch64_estimated_poly_value): Likewise. Use the least significant
8110 set bit for the minimum and likely values. Use the most significant
8111 set bit for the maximum value.
8113 2021-08-03 liuhongt <hongtao.liu@intel.com>
8115 * config/i386/sse.md (cond_<insn><mode>): New expander.
8116 (cond_mul<mode>): Ditto.
8118 2021-08-03 Kewen Lin <linkw@linux.ibm.com>
8120 * tree-cfg.c (move_sese_region_to_fn): Fix typos on dloop.
8122 2021-08-03 liuhongt <hongtao.liu@intel.com>
8124 * config/i386/sse.md (cond_<insn><mode>):New expander.
8125 (cond_mul<mode>): Ditto.
8126 (cond_div<mode>): Ditto.
8128 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
8130 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Also
8131 check stack_realign_needed for stack realignment.
8132 (ix86_legitimate_constant_p): Always allow CONST_WIDE_INT smaller
8133 than the largest integer supported by vector register.
8134 * config/i386/i386.h (MAX_MOVE_MAX): New. Set to 64.
8135 (MOVE_MAX): Set to bytes of the largest integer supported by
8137 (STORE_MAX_PIECES): New.
8139 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
8141 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
8142 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
8143 data from one memory location to another.
8145 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
8148 * config/i386/i386.c (TARGET_GEN_MEMSET_SCRATCH_RTX): New.
8150 2021-08-02 Aldy Hernandez <aldyh@redhat.com>
8152 PR tree-optimization/101724
8153 * params.opt: Remove --param=threader-iterative.
8154 * tree-ssa-threadbackward.c (pass_thread_jumps::execute): Remove
8157 2021-08-02 Tom de Vries <tdevries@suse.de>
8159 PR middle-end/101665
8160 * doc/extend.texi (nonnull attribute): Improve documentation.
8162 2021-08-02 Andrew Pinski <apinski@marvell.com>
8164 PR rtl-optimization/101683
8165 * rtlanal.c (may_trap_p_1): Handle UNSIGNED_FIX.
8167 2021-08-02 Roger Sayle <roger@nextmovesoftware.com>
8169 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern):
8170 Renamed from cond_removal_in_popcount_clz_ctz_pattern.
8171 Add support for BSWAP, FFS, PARITY and CLRSB builtins.
8172 (tree_ssa_phiop_worker): Update call to function above.
8174 2021-08-01 H.J. Lu <hjl.tools@gmail.com>
8177 * config/i386/i386.md (bsr_rex64_1_zext): New.
8178 (combine splitter for constant - clzll): Replace gen_bsr_rex64_1
8179 with gen_bsr_rex64_1_zext.
8181 2021-07-31 Jakub Jelinek <jakub@redhat.com>
8184 * config/i386/i386.md (bsr_rex64_1, bsr_1, bsr_zext_1): New
8185 define_insn patterns.
8186 (*bsr_rex64_2, *bsr_2): New define_insn_and_split patterns.
8187 Add combine splitters for constant - clz.
8188 (clz<mode>2): Use a temporary pseudo for bsr result.
8190 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
8192 * config/rs6000/smmintrin.h (_mm_floor_pd, _mm_floor_ps,
8193 _mm_floor_sd, _mm_floor_ss): New.
8195 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
8197 * config/rs6000/smmintrin.h (_mm_ceil_pd, _mm_ceil_ps,
8198 _mm_ceil_sd, _mm_ceil_ss): New.
8200 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
8202 * config/rs6000/smmintrin.h (_mm_blend_pd, _mm_blendv_pd,
8203 _mm_blend_ps, _mm_blendv_ps): New.
8205 2021-07-30 Roger Sayle <roger@nextmovesoftware.com>
8206 Uroš Bizjak <ubizjak@gmail.com>
8208 * config/i386/i386.md (*dec_cmov<mode>): New define_insn_and_split
8209 to generate a conditional move using the carry flag after sub $1.
8210 (peephole2): Eliminate a register-to-register move by inverting
8211 the condition of a conditional move.
8213 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
8215 * config/mmix/mmix.md ("call", "call_value", "*call_real")
8216 ("*call_value_real"): Don't generate rtx mentioning the generic
8217 operands 1 and 2 to "call", and similarly for "call_value".
8218 * config/mmix/mmix.c (mmix_print_operand_punct_valid_p)
8219 (mmix_print_operand): Use '!' instead of 'p'.
8221 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
8223 * doc/md.texi (call): Correct information about operand 2.
8224 * config/mmix/mmix.md ("call", "call_value"): Remove fixed FIXMEs.
8226 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
8228 * range-op.cc (operator_trunc_mod::wi_fold): Fold constants.
8230 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
8232 * range-op.cc (operator_div::wi_fold): Return UNDEFINED for [0, 0] divisor.
8234 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
8236 * gimple-range-cache.cc (*::set_bb_range): Change const basic_block to
8238 (*::get_bb_range): Ditto.
8239 (*::bb_range_p): Ditto.
8240 * gimple-range-cache.h: Change prototypes.
8242 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
8245 * builtins.c (builtin_memcpy_read_str): Change the mode argument
8246 from scalar_int_mode to fixed_size_mode.
8247 (builtin_strncpy_read_str): Likewise.
8248 (gen_memset_value_from_prev): New function.
8249 (builtin_memset_read_str): Change the mode argument from
8250 scalar_int_mode to fixed_size_mode. Use gen_memset_value_from_prev
8251 and support CONST_VECTOR.
8252 (builtin_memset_gen_str): Likewise.
8253 (try_store_by_multiple_pieces): Use by_pieces_constfn to declare
8255 * builtins.h (builtin_strncpy_read_str): Replace scalar_int_mode
8256 with fixed_size_mode.
8257 (builtin_memset_read_str): Likewise.
8258 * expr.c (widest_int_mode_for_size): Renamed to ...
8259 (widest_fixed_size_mode_for_size): Add a bool argument to
8260 indicate if QI vector mode can be used.
8261 (by_pieces_ninsns): Call widest_fixed_size_mode_for_size
8262 instead of widest_int_mode_for_size.
8263 (pieces_addr::adjust): Change the mode argument from
8264 scalar_int_mode to fixed_size_mode.
8265 (op_by_pieces_d): Make m_len read-only. Add a bool member,
8266 m_qi_vector_mode, to indicate that QI vector mode can be used.
8267 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
8268 initialize m_qi_vector_mode. Call widest_fixed_size_mode_for_size
8269 instead of widest_int_mode_for_size.
8270 (op_by_pieces_d::get_usable_mode): Change the mode argument from
8271 scalar_int_mode to fixed_size_mode. Call
8272 widest_fixed_size_mode_for_size instead of
8273 widest_int_mode_for_size.
8274 (op_by_pieces_d::smallest_fixed_size_mode_for_size): New member
8275 function to return the smallest integer or QI vector mode.
8276 (op_by_pieces_d::run): Call widest_fixed_size_mode_for_size
8277 instead of widest_int_mode_for_size. Call
8278 smallest_fixed_size_mode_for_size instead of
8279 smallest_int_mode_for_size.
8280 (store_by_pieces_d::store_by_pieces_d): Add a bool argument to
8281 indicate that QI vector mode can be used and pass it to
8282 op_by_pieces_d::op_by_pieces_d.
8283 (can_store_by_pieces): Call widest_fixed_size_mode_for_size
8284 instead of widest_int_mode_for_size. Pass memsetp to
8285 widest_fixed_size_mode_for_size to support QI vector mode.
8286 Allow all CONST_VECTORs for memset if vec_duplicate is supported.
8287 (store_by_pieces): Pass memsetp to
8288 store_by_pieces_d::store_by_pieces_d.
8289 (clear_by_pieces_1): Removed.
8290 (clear_by_pieces): Replace clear_by_pieces_1 with
8291 builtin_memset_read_str and pass true to store_by_pieces_d to
8292 support vector mode broadcast.
8293 (string_cst_read_str): Change the mode argument from
8294 scalar_int_mode to fixed_size_mode.
8295 * expr.h (by_pieces_constfn): Change scalar_int_mode to
8297 (by_pieces_prev): Likewise.
8298 * rtl.h (lowpart_subreg_regno): New.
8299 * rtlanal.c (lowpart_subreg_regno): New. A wrapper around
8300 simplify_subreg_regno.
8301 * target.def (gen_memset_scratch_rtx): New hook.
8302 * doc/tm.texi.in: Add TARGET_GEN_MEMSET_SCRATCH_RTX.
8303 * doc/tm.texi: Regenerated.
8305 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
8308 * config/mips/mips.c (mips_atomic_assign_expand_fenv): Use
8309 TARGET_EXPR instead of MODIFY_EXPR.
8311 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
8314 * config/mips/mips-protos.h (mips_expand_vec_cmp_expr): Declare.
8315 * config/mips/mips.c (mips_expand_vec_cmp_expr): New function.
8316 * config/mips/mips-msa.md (vec_cmp<MSA:mode><mode_i>): New
8318 (vec_cmpu<IMSA:mode><mode_i>): New expander.
8320 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
8323 * config/i386/i386-options.c (ix86_option_override_internal):
8324 Don't enable LZCNT/POPCNT if they have been disabled explicitly.
8326 2021-07-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
8329 * config/arm/arm_neon.h (vld1_p64): Replace call to builtin by
8330 explicitly dereferencing __a.
8331 (vld1_s64): Likewise.
8332 (vld1_u64): Likewise.
8333 * config/arm/arm_neon_builtins.def (vld1): Remove entry for di
8334 and change to VAR13.
8336 2021-07-30 Aldy Hernandez <aldyh@redhat.com>
8338 * gimple-loop-versioning.cc (lv_dom_walker::lv_dom_walker): Remove
8339 use of m_range_analyzer.
8340 (loop_versioning::lv_dom_walker::before_dom_children): Same.
8341 (loop_versioning::lv_dom_walker::after_dom_children): Remove.
8342 (loop_versioning::prune_loop_conditions): Replace vr_values use
8343 with range_query interface.
8344 (pass_loop_versioning::execute): Use ranger.
8346 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
8349 * ipa-devirt.c (ipa_odr_read_section): Compare the precision of
8350 enum values, and emit a warning if they mismatch.
8352 2021-07-30 Kewen Lin <linkw@linux.ibm.com>
8354 * cfgloop.h (as_const): New function.
8355 (class loop_iterator): Rename to ...
8356 (class loops_list): ... this.
8357 (loop_iterator::next): Rename to ...
8358 (loops_list::Iter::fill_curr_loop): ... this and adjust.
8359 (loop_iterator::loop_iterator): Rename to ...
8360 (loops_list::loops_list): ... this and adjust.
8361 (loops_list::Iter): New class.
8362 (loops_list::iterator): New type.
8363 (loops_list::const_iterator): New type.
8364 (loops_list::begin): New function.
8365 (loops_list::end): Likewise.
8366 (loops_list::begin const): Likewise.
8367 (loops_list::end const): Likewise.
8368 (FOR_EACH_LOOP): Remove.
8369 (FOR_EACH_LOOP_FN): Remove.
8370 * cfgloop.c (flow_loops_dump): Adjust FOR_EACH_LOOP* with range-based
8371 for loop with loops_list instance.
8372 (sort_sibling_loops): Likewise.
8373 (disambiguate_loops_with_multiple_latches): Likewise.
8374 (verify_loop_structure): Likewise.
8375 * cfgloopmanip.c (create_preheaders): Likewise.
8376 (force_single_succ_latches): Likewise.
8377 * config/aarch64/falkor-tag-collision-avoidance.c
8378 (execute_tag_collision_avoidance): Likewise.
8379 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Likewise.
8380 * config/s390/s390.c (s390_adjust_loops): Likewise.
8381 * doc/loop.texi: Likewise.
8382 * gimple-loop-interchange.cc (pass_linterchange::execute): Likewise.
8383 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
8384 * gimple-loop-versioning.cc (loop_versioning::analyze_blocks): Likewise.
8385 (loop_versioning::make_versioning_decisions): Likewise.
8386 * gimple-ssa-split-paths.c (split_paths): Likewise.
8387 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
8388 * graphite.c (canonicalize_loop_form): Likewise.
8389 (graphite_transform_loops): Likewise.
8390 * ipa-fnsummary.c (analyze_function_body): Likewise.
8391 * ipa-pure-const.c (analyze_function): Likewise.
8392 * loop-doloop.c (doloop_optimize_loops): Likewise.
8393 * loop-init.c (loop_optimizer_finalize): Likewise.
8394 (fix_loop_structure): Likewise.
8395 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
8396 (move_loop_invariants): Likewise.
8397 * loop-unroll.c (decide_unrolling): Likewise.
8398 (unroll_loops): Likewise.
8399 * modulo-sched.c (sms_schedule): Likewise.
8400 * predict.c (predict_loops): Likewise.
8401 (pass_profile::execute): Likewise.
8402 * profile.c (branch_prob): Likewise.
8403 * sel-sched-ir.c (sel_finish_pipelining): Likewise.
8404 (sel_find_rgns): Likewise.
8405 * tree-cfg.c (replace_loop_annotate): Likewise.
8406 (replace_uses_by): Likewise.
8407 (move_sese_region_to_fn): Likewise.
8408 * tree-if-conv.c (pass_if_conversion::execute): Likewise.
8409 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
8410 * tree-parloops.c (parallelize_loops): Likewise.
8411 * tree-predcom.c (tree_predictive_commoning): Likewise.
8412 * tree-scalar-evolution.c (scev_initialize): Likewise.
8413 (scev_reset): Likewise.
8414 * tree-ssa-dce.c (find_obviously_necessary_stmts): Likewise.
8415 * tree-ssa-live.c (remove_unused_locals): Likewise.
8416 * tree-ssa-loop-ch.c (ch_base::copy_headers): Likewise.
8417 * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
8418 (tree_ssa_lim_initialize): Likewise.
8419 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Likewise.
8420 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize): Likewise.
8421 * tree-ssa-loop-manip.c (get_loops_exits): Likewise.
8422 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations): Likewise.
8423 (free_numbers_of_iterations_estimates): Likewise.
8424 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Likewise.
8425 * tree-ssa-loop-split.c (tree_ssa_split_loops): Likewise.
8426 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Likewise.
8427 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
8428 (pass_scev_cprop::execute): Likewise.
8429 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
8430 * tree-ssa-sccvn.c (do_rpo_vn): Likewise.
8431 * tree-ssa-threadupdate.c
8432 (jump_thread_path_registry::thread_through_all_blocks): Likewise.
8433 * tree-vectorizer.c (vectorize_loops): Likewise.
8434 * tree-vrp.c (vrp_asserts::find_assert_locations): Likewise.
8436 2021-07-29 Hans-Peter Nilsson <hp@bitrange.com>
8438 * config/mmix/mmix.c (mmix_function_arg_1): Avoid
8439 generating a VOIDmode register for e.g the
8440 function_arg_info::end_marker.
8442 2021-07-29 Jeff Law <jeffreyalaw@gmail.com>
8444 * config/h8300/h8300-modes.def: Add CCZ, CCV and CCC, drop CCZNV.
8445 * config/h8300/h8300.md (H8cc mode iterator): Add CCZ.
8446 (cc mode_attr): Similarly.
8447 (ccz subst_attr): Similarly.
8448 * config/h8300/jumpcall.md: Add new patterns for branch-on-bit.
8449 * config/h8300/testcompare.md: Remove various cc0 based patterns
8450 that had been commented out. Add pattern to set CCZ from a bit
8453 2021-07-29 Thomas Schwinge <thomas@codesourcery.com>
8454 Julian Brown <julian@codesourcery.com>
8455 Kwok Cheung Yeung <kcy@codesourcery.com>
8457 * omp-offload.c (oacc_loop_xform_head_tail, oacc_loop_process):
8458 'update_stmt' after modification.
8459 (pass_oacc_loop_designation): New function, extracted out of...
8460 (pass_oacc_device_lower): ... this.
8461 (pass_data_oacc_loop_designation, pass_oacc_loop_designation)
8462 (make_pass_oacc_loop_designation): New
8463 * passes.def: Add it.
8464 * tree-parloops.c (create_parallel_loop): Adjust.
8465 * tree-pass.h (make_pass_oacc_loop_designation): New.
8467 2021-07-29 Aldy Hernandez <aldyh@redhat.com>
8469 * flag-types.h (enum threader_mode): New.
8470 * params.opt: Add entry for --param=threader-mode.
8471 * tree-ssa-threadbackward.c (THREADER_ITERATIVE_MODE): New.
8472 (class back_threader): New.
8473 (back_threader::back_threader): New.
8474 (back_threader::~back_threader): New.
8475 (back_threader::maybe_register_path): New.
8476 (back_threader::find_taken_edge): New.
8477 (back_threader::find_taken_edge_switch): New.
8478 (back_threader::find_taken_edge_cond): New.
8479 (back_threader::resolve_def): New.
8480 (back_threader::resolve_phi): New.
8481 (back_threader::find_paths_to_names): New.
8482 (back_threader::find_paths): New.
8485 (thread_jumps::find_jump_threads_backwards): Call ranger threader.
8486 (thread_jumps::find_jump_threads_backwards_with_ranger): New.
8487 (pass_thread_jumps::execute): Abstract out code...
8488 (try_thread_blocks): ...here.
8489 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
8490 Abstract out threading candidate code to...
8491 (single_succ_to_potentially_threadable_block): ...here.
8492 * tree-ssa-threadedge.h (single_succ_to_potentially_threadable_block):
8494 * tree-ssa-threadupdate.c (register_jump_thread): Return boolean.
8495 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
8496 Return bool from register_jump_thread.
8498 2021-07-29 Andreas Krebbel <krebbel@linux.ibm.com>
8500 * target.def: in0 and in1 do not need to be registers.
8501 * doc/tm.texi: Regenerate.
8503 2021-07-29 liuhongt <hongtao.liu@intel.com>
8506 * config/i386/i386.c (ix86_widen_mult_cost): New function.
8507 (ix86_add_stmt_cost): Use ix86_widen_mult_cost for
8510 2021-07-29 Jiufu Guo <guojiufu@linux.ibm.com>
8513 * config/rs6000/rs6000.c (TARGET_PREFERRED_DOLOOP_MODE): New hook.
8514 (rs6000_preferred_doloop_mode): New hook.
8515 * doc/tm.texi: Regenerate.
8516 * doc/tm.texi.in: Add hook preferred_doloop_mode.
8517 * target.def (preferred_doloop_mode): New hook.
8518 * targhooks.c (default_preferred_doloop_mode): New hook.
8519 * targhooks.h (default_preferred_doloop_mode): New hook.
8520 * tree-ssa-loop-ivopts.c (compute_doloop_base_on_mode): New function.
8521 (add_iv_candidate_for_doloop): Call targetm.preferred_doloop_mode
8522 and compute_doloop_base_on_mode.
8524 2021-07-28 Martin Sebor <msebor@redhat.com>
8526 PR middle-end/101494
8527 * tree-ssa-uninit.c (maybe_warn_operand): Correct object offset
8528 and size computation.
8530 2021-07-28 Martin Sebor <msebor@redhat.com>
8532 PR middle-end/101601
8533 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Remove
8535 Handle pointers to functions.
8537 2021-07-28 Martin Sebor <msebor@redhat.com>
8539 * Makefile.in (OBJS): Add gimple-ssa-warn-access.o and pointer-query.o.
8540 * attribs.h (fndecl_dealloc_argno): Move fndecl_dealloc_argno to tree.h.
8541 * builtins.c (compute_objsize_r): Move to pointer-query.cc.
8542 (access_ref::access_ref): Same.
8543 (access_ref::phi): Same.
8544 (access_ref::get_ref): Same.
8545 (access_ref::size_remaining): Same.
8546 (access_ref::offset_in_range): Same.
8547 (access_ref::add_offset): Same.
8548 (access_ref::inform_access): Same.
8549 (ssa_name_limit_t::visit_phi): Same.
8550 (ssa_name_limit_t::leave_phi): Same.
8551 (ssa_name_limit_t::next): Same.
8552 (ssa_name_limit_t::next_phi): Same.
8553 (ssa_name_limit_t::~ssa_name_limit_t): Same.
8554 (pointer_query::pointer_query): Same.
8555 (pointer_query::get_ref): Same.
8556 (pointer_query::put_ref): Same.
8557 (pointer_query::flush_cache): Same.
8558 (warn_string_no_nul): Move to gimple-ssa-warn-access.cc.
8559 (check_nul_terminated_array): Same.
8560 (unterminated_array): Same.
8561 (maybe_warn_for_bound): Same.
8562 (check_read_access): Same.
8563 (warn_for_access): Same.
8564 (get_size_range): Same.
8565 (check_access): Same.
8566 (gimple_call_alloc_size): Move to tree.c.
8567 (gimple_parm_array_size): Move to pointer-query.cc.
8568 (get_offset_range): Same.
8569 (gimple_call_return_array): Same.
8570 (handle_min_max_size): Same.
8571 (handle_array_ref): Same.
8572 (handle_mem_ref): Same.
8573 (compute_objsize): Same.
8574 (gimple_call_alloc_p): Move to gimple-ssa-warn-access.cc.
8575 (call_dealloc_argno): Same.
8576 (fndecl_dealloc_argno): Same.
8577 (new_delete_mismatch_p): Same.
8578 (matching_alloc_calls_p): Same.
8579 (warn_dealloc_offset): Same.
8580 (maybe_emit_free_warning): Same.
8581 * builtins.h (check_nul_terminated_array): Move to
8582 gimple-ssa-warn-access.h.
8583 (check_nul_terminated_array): Same.
8584 (warn_string_no_nul): Same.
8585 (unterminated_array): Same.
8586 (class ssa_name_limit_t): Same.
8587 (class pointer_query): Same.
8588 (struct access_ref): Same.
8589 (class range_query): Same.
8590 (struct access_data): Same.
8591 (gimple_call_alloc_size): Same.
8592 (gimple_parm_array_size): Same.
8593 (compute_objsize): Same.
8594 (class access_data): Same.
8595 (maybe_emit_free_warning): Same.
8596 * calls.c (initialize_argument_information): Remove call to
8597 maybe_emit_free_warning.
8598 * gimple-array-bounds.cc: Include new header..
8599 * gimple-fold.c: Same.
8600 * gimple-ssa-sprintf.c: Same.
8601 * gimple-ssa-warn-restrict.c: Same.
8602 * passes.def: Add pass_warn_access.
8603 * tree-pass.h (make_pass_warn_access): Declare.
8604 * tree-ssa-strlen.c: Include new headers.
8605 * tree.c (fndecl_dealloc_argno): Move here from builtins.c.
8606 * tree.h (fndecl_dealloc_argno): Move here from attribs.h.
8607 * gimple-ssa-warn-access.cc: New file.
8608 * gimple-ssa-warn-access.h: New file.
8609 * pointer-query.cc: New file.
8610 * pointer-query.h: New file.
8612 2021-07-28 Jakub Jelinek <jakub@redhat.com>
8614 PR middle-end/101624
8615 * ubsan.c (maybe_instrument_pointer_overflow,
8616 instrument_object_size): Only test DECL_REGISTER on VAR_DECLs,
8617 PARM_DECLs or RESULT_DECLs.
8618 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
8620 2021-07-28 Jakub Jelinek <jakub@redhat.com>
8622 PR middle-end/101642
8623 * match.pd (bswap16 (x) == bswap16 (y)): Cast both operands
8624 to type of bswap16 for comparison.
8625 (bswap16 (x) == cst): Cast bswap16 operand to type of cst.
8627 2021-07-28 Richard Biener <rguenther@suse.de>
8629 PR tree-optimization/101615
8630 * tree-vect-slp.c (vect_optimize_slp): Materialize permutes
8631 at CTOR SLP graph entries.
8633 2021-07-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8635 * config/aarch64/aarch64.md (*extend<SHORT:mode><GPI:mode>2_aarch64):
8636 Add "r,w" alternative.
8638 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
8641 * config/i386/i386.c (ix86_avx_u128_mode_needed): Don't set
8642 AVX_U128_DIRTY when all bits are zero.
8644 2021-07-28 Richard Biener <rguenther@suse.de>
8646 PR tree-optimization/101615
8647 * tree-vect-slp.c (vect_optimize_slp): Pre-existing vector
8648 external nodes cannot be permuted so make them perm_out 0.
8650 2021-07-28 Andrew Stubbs <ams@codesourcery.com>
8653 * config.in: Regenerate.
8654 * config/gcn/gcn-hsa.h (A_FIJI): New define.
8655 (A_900): New define.
8656 (A_906): New define.
8657 (A_908): New define.
8658 (ASM_SPEC): Use A_FIJI, A_900, A_906 and A_908.
8659 * config/gcn/gcn.c (output_file_start): Adjust attributes according
8660 to the assembler capabilities.
8661 * config/gcn/mkoffload.c (main): Likewise.
8662 * configure: Regenerate.
8663 * configure.ac: Add tests for LLVM assembler attribute features.
8665 2021-07-28 Andrew MacLeod <amacleod@redhat.com>
8667 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Check for
8668 cond_false and cond_true on branches.
8670 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
8672 * config/aarch64/aarch64.c (aarch64_gen_adjusted_ldpstp): use
8675 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
8677 * alias.c (init_alias_analysis): Don't skip prologue/epilogue.
8679 2021-07-28 Jakub Jelinek <jakub@redhat.com>
8682 * config/i386/sse.md (vashr<mode>3): Split into vashrv8di3 expander
8683 and vashrv4di3 expander, where the latter requires just TARGET_AVX2
8684 and has special !TARGET_AVX512VL expansion.
8685 (vashrv2di3<mask_name>): Rename to ...
8686 (vashrv2di3): ... this. Change condition to TARGET_XOP || TARGET_AVX2
8687 and add special !TARGET_XOP && !TARGET_AVX512VL expansion.
8689 2021-07-28 Martin Uecker <muecker@gwdg.de>
8691 * calls.c (maybe_warn_rdwr_sizes): Correct argument
8692 numbers in warning that were switched.
8694 2021-07-28 Kewen Lin <linkw@linux.ibm.com>
8696 PR tree-optimization/101596
8697 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Fix wrong check
8698 by using new_type's precision instead.
8700 2021-07-28 liuhongt <hongtao.liu@intel.com>
8703 * config/i386/i386.h (processor_costs): Add new member
8705 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
8706 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
8707 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
8708 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
8709 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
8710 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
8711 generic_cost, core_cost): Initialize integer_to_sse same value
8713 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
8714 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8715 Use integer_to_sse instead of sse_op to calculate the cost of
8718 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
8720 * config/rs6000/rs6000-gen-builtins.c (write_ovld_static_init): New
8722 (write_init_file): Call write_ovld_static_init.
8724 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
8726 * config/rs6000/rs6000-gen-builtins.c (write_bif_static_init): New
8728 (write_init_file): Call write_bif_static_init.
8730 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
8732 * config/rs6000/rs6000-gen-builtins.c (typemap): New struct.
8733 (TYPE_MAP_SIZE): New macro.
8734 (type_map): New initialized variable.
8735 (typemap_cmp): New function.
8736 (write_type_node): Likewise.
8737 (write_fntype_init): Implement.
8739 2021-07-27 Martin Sebor <msebor@redhat.com>
8741 PR tree-optimization/101584
8742 * tree-ssa-uninit.c (builtin_call_nomodifying_p): New function.
8743 (check_defs): Call it.
8745 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
8747 * tree-ssa-dom.c (dom_jump_threader_simplifier):
8748 Put avail_exprs_stack in the class, instead of passing it to
8749 jump_threader_simplifier.
8750 (dom_jump_threader_simplifier::simplify): Add state argument.
8751 (dom_opt_dom_walker): Add state.
8752 (pass_dominator::execute): Pass state to threader.
8753 (dom_opt_dom_walker::before_dom_children): Use state.
8754 * tree-ssa-threadedge.c (jump_threader::jump_threader): Replace
8756 (jump_threader::record_temporary_equivalences_from_phis):
8757 Register equivalences through the state variable.
8758 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
8759 Record ranges in a statement through the state variable.
8760 (jump_threader::simplify_control_stmt_condition): Pass state to
8762 (jump_threader::simplify_control_stmt_condition_1): Same.
8763 (jump_threader::thread_around_empty_blocks): Remove obsolete
8765 (jump_threader::thread_through_normal_block): Record equivalences
8766 on edge through the state variable.
8767 (jump_threader::thread_across_edge): Abstract state pushing.
8768 (jt_state::jt_state): New.
8769 (jt_state::push): New.
8770 (jt_state::pop): New.
8771 (jt_state::register_equiv): New.
8772 (jt_state::record_ranges_from_stmt): New.
8773 (jt_state::register_equivs_on_edge): New.
8774 (jump_threader_simplifier::jump_threader_simplifier): Move from
8776 (jump_threader_simplifier::simplify): Add state argument.
8777 * tree-ssa-threadedge.h (class jt_state): New.
8778 (class jump_threader): Add state to constructor.
8779 (class jump_threader_simplifier): Add state to simplify. Remove
8780 avail_exprs_stack from class.
8781 * tree-vrp.c (vrp_jump_threader_simplifier::simplify): Add state
8783 (vrp_jump_threader::vrp_jump_threader): Add state.
8784 (vrp_jump_threader::~vrp_jump_threader): Cleanup state.
8786 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
8788 * Makefile.in (OBJS): Add gimple-range-path.o.
8789 * gimple-range-path.cc: New file.
8790 * gimple-range-path.h: New file.
8792 2021-07-27 Jonathan Wright <jonathan.wright@arm.com>
8794 * config/aarch64/aarch64-simd.md: Push sign/zero-extension
8795 inside vec_duplicate for all patterns.
8796 * simplify-rtx.c (simplify_context::simplify_unary_operation_1):
8797 Push sign/zero-extension inside vec_duplicate.
8799 2021-07-27 Richard Biener <rguenther@suse.de>
8801 PR tree-optimization/101573
8802 * tree-ssa-uninit.c (warn_uninit_phi_uses): New function
8803 looking at uninitialized PHI arg defs in some constrained cases.
8804 (warn_uninitialized_vars): Call it.
8805 (execute_early_warn_uninitialized): Calculate dominators.
8807 2021-07-27 Richard Biener <rguenther@suse.de>
8809 PR tree-optimization/39821
8810 * tree-vect-stmts.c (vect_model_promotion_demotion_cost): Use
8811 vector_stmt for widening arithmetic.
8812 (vectorizable_conversion): Adjust.
8814 2021-07-27 Martin Jambor <mjambor@suse.cz>
8816 * cgraph.h (ipa_replace_map): New field force_load_ref.
8817 * ipa-prop.h (ipa_param_descriptor): Reduce precision of move_cost,
8818 aded new flag load_dereferenced, adjusted comments.
8819 (ipa_get_param_dereferenced): New function.
8820 (ipa_set_param_dereferenced): Likewise.
8821 * cgraphclones.c (cgraph_node::create_virtual_clone): Follow it.
8822 * ipa-cp.c: Include gimple.h.
8823 (ipcp_discover_new_direct_edges): Take into account dereferenced flag.
8824 (get_replacement_map): New parameter force_load_ref, set the
8825 appropriate flag in ipa_replace_map if set.
8826 (struct symbol_and_index_together): New type.
8827 (adjust_refs_in_act_callers): New function.
8828 (adjust_references_in_caller): Likewise.
8829 (create_specialized_node): When appropriate, call
8830 adjust_references_in_caller and force only load references.
8831 * ipa-prop.c (load_from_dereferenced_name): New function.
8832 (ipa_analyze_controlled_uses): Also detect loads from a
8833 dereference, harden testing of call statements.
8834 (ipa_write_node_info): Stream the dereferenced flag.
8835 (ipa_read_node_info): Likewise.
8836 (ipa_set_jf_constant): Also create refdesc when jump function
8837 references a variable.
8838 (cgraph_node_for_jfunc): Rename to symtab_node_for_jfunc, work
8839 also on references of variables and return a symtab_node. Adjust
8841 (propagate_controlled_uses): Also remove references to VAR_DECLs.
8843 2021-07-27 Jakub Jelinek <jakub@redhat.com>
8845 PR middle-end/101586
8846 * gimple-fold.c (clear_padding_type): Ignore FIELD_DECLs with byte
8847 positions above or equal to sz except for diagnostics of flexible
8850 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
8852 PR tree-optimization/78888
8853 * gimple-range-fold.cc (get_letter_range): New.
8854 (fold_using_range::range_of_builtin_call): Call get_letter_range.
8856 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
8858 PR tree-optimization/78888
8859 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Add cases
8860 for CFN_BUILT_IN_TOUPPER and CFN_BUILT_IN_TOLOWER.
8862 2021-07-26 Roger Sayle <roger@nextmovesoftware.com>
8863 Marc Glisse <marc.glisse@inria.fr>
8865 * match.pd (rotate): Simplify equality/inequality of rotations.
8866 (bswap): Simplify equality/inequality tests of byte swapping.
8868 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8870 * range-op.cc (operator_bitwise_xor::op1_op2_relation_effect):
8873 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8875 * range-op.cc (operator_lshift::fold_range): Pass rel to
8876 base class fold_range.
8877 (operator_rshift::fold_range): Same.
8879 2021-07-26 Ashimida <ashimida@linux.alibaba.com>
8882 * toplev.h (min_align_loops_log): Remove declaration.
8883 (min_align_jumps_log, min_align_labels_log): Likewise.
8884 (min_align_functions_log): Likewise.
8886 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8888 * tree-vrp.c (vrp_simplify_cond_using_ranges): Rename vr_values
8890 (execute_vrp): Abstract out simplification of conditionals...
8891 (simplify_casted_conds): ...here.
8893 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8895 * gimple-array-bounds.cc (array_bounds_checker::get_value_range):
8896 Add gimple argument.
8897 (array_bounds_checker::check_array_ref): Same.
8898 (array_bounds_checker::check_addr_expr): Same.
8899 (array_bounds_checker::check_array_bounds): Pass statement to
8900 check_array_bounds and check_addr_expr.
8901 * gimple-array-bounds.h (check_array_bounds): Add gimple argument.
8902 (check_addr_expr): Same.
8903 (get_value_range): Same.
8905 2021-07-26 Tamar Christina <tamar.christina@arm.com>
8907 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): Rename to..
8908 (sdot_prod, udot_prod): ... This.
8909 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): Merged
8911 (<sur>dot_prod<vsi2qi>): ... this.
8912 (aarch64_<sur>dot_lane<vsi2qi>, aarch64_<sur>dot_laneq<vsi2qi>):
8913 Change operands order.
8914 (<sur>sadv16qi): Use new operands order.
8915 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32,
8916 vdotq_s32): Use new RTL ordering.
8918 2021-07-26 Tamar Christina <tamar.christina@arm.com>
8920 * config/aarch64/aarch64-builtins.c (TYPES_TERNOP_SUSS,
8921 aarch64_types_ternop_suss_qualifiers): New.
8922 * config/aarch64/aarch64-simd-builtins.def (usdot_prod): Use it.
8923 * config/aarch64/aarch64-simd.md (usdot_prod<vsi2qi>): Re-organize RTL.
8924 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Use it.
8926 2021-07-23 Jakub Jelinek <jakub@redhat.com>
8928 PR rtl-optimization/101562
8929 * expmed.c (store_integral_bit_field): Only use movstrict_optab
8930 if the operand isn't paradoxical.
8932 2021-07-23 Aldy Hernandez <aldyh@redhat.com>
8934 * gimple-array-bounds.h (class array_bounds_checker): Change
8935 ranges type to range_query.
8937 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8939 * config/aarch64/arm_neon.h (vst1_s64_x2): Use
8940 __builtin_memcpy instead of constructing
8941 __builtin_aarch64_simd_oi one vector at a time.
8942 (vst1_u64_x2): Likewise.
8943 (vst1_f64_x2): Likewise.
8944 (vst1_s8_x2): Likewise.
8945 (vst1_p8_x2): Likewise.
8946 (vst1_s16_x2): Likewise.
8947 (vst1_p16_x2): Likewise.
8948 (vst1_s32_x2): Likewise.
8949 (vst1_u8_x2): Likewise.
8950 (vst1_u16_x2): Likewise.
8951 (vst1_u32_x2): Likewise.
8952 (vst1_f16_x2): Likewise.
8953 (vst1_f32_x2): Likewise.
8954 (vst1_p64_x2): Likewise.
8955 (vst1q_s8_x2): Likewise.
8956 (vst1q_p8_x2): Likewise.
8957 (vst1q_s16_x2): Likewise.
8958 (vst1q_p16_x2): Likewise.
8959 (vst1q_s32_x2): Likewise.
8960 (vst1q_s64_x2): Likewise.
8961 (vst1q_u8_x2): Likewise.
8962 (vst1q_u16_x2): Likewise.
8963 (vst1q_u32_x2): Likewise.
8964 (vst1q_u64_x2): Likewise.
8965 (vst1q_f16_x2): Likewise.
8966 (vst1q_f32_x2): Likewise.
8967 (vst1q_f64_x2): Likewise.
8968 (vst1q_p64_x2): Likewise.
8970 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8972 * config/aarch64/arm_neon.h (vst1_s64_x3): Use
8973 __builtin_memcpy instead of constructing
8974 __builtin_aarch64_simd_ci one vector at a time.
8975 (vst1_u64_x3): Likewise.
8976 (vst1_f64_x3): Likewise.
8977 (vst1_s8_x3): Likewise.
8978 (vst1_p8_x3): Likewise.
8979 (vst1_s16_x3): Likewise.
8980 (vst1_p16_x3): Likewise.
8981 (vst1_s32_x3): Likewise.
8982 (vst1_u8_x3): Likewise.
8983 (vst1_u16_x3): Likewise.
8984 (vst1_u32_x3): Likewise.
8985 (vst1_f16_x3): Likewise.
8986 (vst1_f32_x3): Likewise.
8987 (vst1_p64_x3): Likewise.
8988 (vst1q_s8_x3): Likewise.
8989 (vst1q_p8_x3): Likewise.
8990 (vst1q_s16_x3): Likewise.
8991 (vst1q_p16_x3): Likewise.
8992 (vst1q_s32_x3): Likewise.
8993 (vst1q_s64_x3): Likewise.
8994 (vst1q_u8_x3): Likewise.
8995 (vst1q_u16_x3): Likewise.
8996 (vst1q_u32_x3): Likewise.
8997 (vst1q_u64_x3): Likewise.
8998 (vst1q_f16_x3): Likewise.
8999 (vst1q_f32_x3): Likewise.
9000 (vst1q_f64_x3): Likewise.
9001 (vst1q_p64_x3): Likewise.
9003 2021-07-23 H.J. Lu <hjl.tools@gmail.com>
9006 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): Don't return
9007 hard register when LRA is in progress.
9009 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9011 * config/aarch64/arm_neon.h (vst1_s8_x4): Use
9012 __builtin_memcpy instead of using a union.
9013 (vst1q_s8_x4): Likewise.
9014 (vst1_s16_x4): Likewise.
9015 (vst1q_s16_x4): Likewise.
9016 (vst1_s32_x4): Likewise.
9017 (vst1q_s32_x4): Likewise.
9018 (vst1_u8_x4): Likewise.
9019 (vst1q_u8_x4): Likewise.
9020 (vst1_u16_x4): Likewise.
9021 (vst1q_u16_x4): Likewise.
9022 (vst1_u32_x4): Likewise.
9023 (vst1q_u32_x4): Likewise.
9024 (vst1_f16_x4): Likewise.
9025 (vst1q_f16_x4): Likewise.
9026 (vst1_f32_x4): Likewise.
9027 (vst1q_f32_x4): Likewise.
9028 (vst1_p8_x4): Likewise.
9029 (vst1q_p8_x4): Likewise.
9030 (vst1_p16_x4): Likewise.
9031 (vst1q_p16_x4): Likewise.
9032 (vst1_s64_x4): Likewise.
9033 (vst1_u64_x4): Likewise.
9034 (vst1_p64_x4): Likewise.
9035 (vst1q_s64_x4): Likewise.
9036 (vst1q_u64_x4): Likewise.
9037 (vst1q_p64_x4): Likewise.
9038 (vst1_f64_x4): Likewise.
9039 (vst1q_f64_x4): Likewise.
9041 2021-07-23 Jonathan Wrightt <jonathan.wright@arm.com>
9043 * config/aarch64/arm_neon.h (vst2_s64): Use __builtin_memcpy
9044 instead of constructing __builtin_aarch64_simd_oi one vector
9046 (vst2_u64): Likewise.
9047 (vst2_f64): Likewise.
9048 (vst2_s8): Likewise.
9049 (vst2_p8): Likewise.
9050 (vst2_s16): Likewise.
9051 (vst2_p16): Likewise.
9052 (vst2_s32): Likewise.
9053 (vst2_u8): Likewise.
9054 (vst2_u16): Likewise.
9055 (vst2_u32): Likewise.
9056 (vst2_f16): Likewise.
9057 (vst2_f32): Likewise.
9058 (vst2_p64): Likewise.
9059 (vst2q_s8): Likewise.
9060 (vst2q_p8): Likewise.
9061 (vst2q_s16): Likewise.
9062 (vst2q_p16): Likewise.
9063 (vst2q_s32): Likewise.
9064 (vst2q_s64): Likewise.
9065 (vst2q_u8): Likewise.
9066 (vst2q_u16): Likewise.
9067 (vst2q_u32): Likewise.
9068 (vst2q_u64): Likewise.
9069 (vst2q_f16): Likewise.
9070 (vst2q_f32): Likewise.
9071 (vst2q_f64): Likewise.
9072 (vst2q_p64): Likewise.
9074 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9076 * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy
9077 instead of constructing __builtin_aarch64_simd_ci one vector
9079 (vst3_u64): Likewise.
9080 (vst3_f64): Likewise.
9081 (vst3_s8): Likewise.
9082 (vst3_p8): Likewise.
9083 (vst3_s16): Likewise.
9084 (vst3_p16): Likewise.
9085 (vst3_s32): Likewise.
9086 (vst3_u8): Likewise.
9087 (vst3_u16): Likewise.
9088 (vst3_u32): Likewise.
9089 (vst3_f16): Likewise.
9090 (vst3_f32): Likewise.
9091 (vst3_p64): Likewise.
9092 (vst3q_s8): Likewise.
9093 (vst3q_p8): Likewise.
9094 (vst3q_s16): Likewise.
9095 (vst3q_p16): Likewise.
9096 (vst3q_s32): Likewise.
9097 (vst3q_s64): Likewise.
9098 (vst3q_u8): Likewise.
9099 (vst3q_u16): Likewise.
9100 (vst3q_u32): Likewise.
9101 (vst3q_u64): Likewise.
9102 (vst3q_f16): Likewise.
9103 (vst3q_f32): Likewise.
9104 (vst3q_f64): Likewise.
9105 (vst3q_p64): Likewise.
9107 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9109 * config/aarch64/arm_neon.h (vst4_s64): Use __builtin_memcpy
9110 instead of constructing __builtin_aarch64_simd_xi one vector
9112 (vst4_u64): Likewise.
9113 (vst4_f64): Likewise.
9114 (vst4_s8): Likewise.
9115 (vst4_p8): Likewise.
9116 (vst4_s16): Likewise.
9117 (vst4_p16): Likewise.
9118 (vst4_s32): Likewise.
9119 (vst4_u8): Likewise.
9120 (vst4_u16): Likewise.
9121 (vst4_u32): Likewise.
9122 (vst4_f16): Likewise.
9123 (vst4_f32): Likewise.
9124 (vst4_p64): Likewise.
9125 (vst4q_s8): Likewise.
9126 (vst4q_p8): Likewise.
9127 (vst4q_s16): Likewise.
9128 (vst4q_p16): Likewise.
9129 (vst4q_s32): Likewise.
9130 (vst4q_s64): Likewise.
9131 (vst4q_u8): Likewise.
9132 (vst4q_u16): Likewise.
9133 (vst4q_u32): Likewise.
9134 (vst4q_u64): Likewise.
9135 (vst4q_f16): Likewise.
9136 (vst4q_f32): Likewise.
9137 (vst4q_f64): Likewise.
9138 (vst4q_p64): Likewise.
9140 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9142 * config/aarch64/arm_neon.h (vtbx4_s8): Use __builtin_memcpy
9143 instead of constructing __builtin_aarch64_simd_oi one vector
9145 (vtbx4_u8): Likewise.
9146 (vtbx4_p8): Likewise.
9148 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9150 * config/aarch64/arm_neon.h (vtbl3_s8): Use __builtin_memcpy
9151 instead of constructing __builtin_aarch64_simd_oi one vector
9153 (vtbl3_u8): Likewise.
9154 (vtbl3_p8): Likewise.
9155 (vtbl4_s8): Likewise.
9156 (vtbl4_u8): Likewise.
9157 (vtbl4_p8): Likewise.
9159 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9161 * config/aarch64/arm_neon.h (vqtbx2_s8): Use __builtin_memcpy
9162 instead of constructing __builtin_aarch64_simd_oi one vector
9164 (vqtbx2_u8): Likewise.
9165 (vqtbx2_p8): Likewise.
9166 (vqtbx2q_s8): Likewise.
9167 (vqtbx2q_u8): Likewise.
9168 (vqtbx2q_p8): Likewise.
9169 (vqtbx3_s8): Use __builtin_memcpy instead of constructing
9170 __builtin_aarch64_simd_ci one vector at a time.
9171 (vqtbx3_u8): Likewise.
9172 (vqtbx3_p8): Likewise.
9173 (vqtbx3q_s8): Likewise.
9174 (vqtbx3q_u8): Likewise.
9175 (vqtbx3q_p8): Likewise.
9176 (vqtbx4_s8): Use __builtin_memcpy instead of constructing
9177 __builtin_aarch64_simd_xi one vector at a time.
9178 (vqtbx4_u8): Likewise.
9179 (vqtbx4_p8): Likewise.
9180 (vqtbx4q_s8): Likewise.
9181 (vqtbx4q_u8): Likewise.
9182 (vqtbx4q_p8): Likewise.
9184 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
9186 * config/aarch64/arm_neon.h (vqtbl2_s8): Use __builtin_memcpy
9187 instead of constructing __builtin_aarch64_simd_oi one vector
9189 (vqtbl2_u8): Likewise.
9190 (vqtbl2_p8): Likewise.
9191 (vqtbl2q_s8): Likewise.
9192 (vqtbl2q_u8): Likewise.
9193 (vqtbl2q_p8): Likewise.
9194 (vqtbl3_s8): Use __builtin_memcpy instead of constructing
9195 __builtin_aarch64_simd_ci one vector at a time.
9196 (vqtbl3_u8): Likewise.
9197 (vqtbl3_p8): Likewise.
9198 (vqtbl3q_s8): Likewise.
9199 (vqtbl3q_u8): Likewise.
9200 (vqtbl3q_p8): Likewise.
9201 (vqtbl4_s8): Use __builtin_memcpy instead of constructing
9202 __builtin_aarch64_simd_xi one vector at a time.
9203 (vqtbl4_u8): Likewise.
9204 (vqtbl4_p8): Likewise.
9205 (vqtbl4q_s8): Likewise.
9206 (vqtbl4q_u8): Likewise.
9207 (vqtbl4q_p8): Likewise.
9209 2021-07-23 Haochen Gui <guihaoc@gcc.gnu.org>
9212 * config/rs6000/rs6000.md (cstore<mode>4): Fix wrong fall through.
9214 2021-07-22 Andrew Pinski <apinski@marvell.com>
9216 PR tree-optimization/10153
9217 * tree-tailcall.c (create_tailcall_accumulator):
9218 Don't call fold_convert as the type should be correct already.
9219 (tree_optimize_tail_calls_1): Use build_{one,zero}_cst instead
9220 of integer_{one,zero}_node for the call of create_tailcall_accumulator.
9222 2021-07-22 Aldy Hernandez <aldyh@redhat.com>
9224 * gimple-range-cache.cc (non_null_ref::adjust_range): Replace
9225 varying_p check for null/non-null check.
9227 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
9229 PR tree-optimization/101511
9230 * value-relation.cc (relation_oracle::query_relation): Check if ssa1
9231 is in ssa2's equiv set, and don't trap if so.
9233 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
9235 PR tree-optimization/101497
9236 * gimple-range-fold.cc (fold_using_range::range_of_cond_expr): Check
9239 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
9241 PR tree-optimization/101496
9242 * vr-values.c (simplify_using_ranges::fold_cond): Call range_of_stmt
9243 first, then vrp_visit_cond_Stmt.
9245 2021-07-22 liuhongt <hongtao.liu@intel.com>
9247 * config/i386/i386-expand.c
9248 (ix86_broadcast_from_integer_constant): Rename to ..
9249 (ix86_broadcast_from_constant): .. this, and extend it to
9251 (ix86_expand_vector_move): Extend to float mode.
9252 * config/i386/i386-features.c
9253 (replace_constant_pool_with_broadcast): Remove.
9254 (remove_partial_avx_dependency_gate): Ditto.
9255 (constant_pool_broadcast): Ditto.
9256 (class pass_constant_pool_broadcast): Ditto.
9257 (make_pass_constant_pool_broadcast): Ditto.
9258 (remove_partial_avx_dependency): Adjust gate.
9259 * config/i386/i386-passes.def: Remove pass_constant_pool_broadcast.
9260 * config/i386/i386-protos.h
9261 (make_pass_constant_pool_broadcast): Remove.
9263 2021-07-22 liuhongt <hongtao.liu@intel.com>
9265 * config/i386/constraints.md (Wb): New constraint.
9267 * config/i386/i386.md (*ashlhi3_1): Extend to avx512 mask
9269 (*ashlqi3_1): Ditto.
9270 (*<insn><mode>3_1): Split to ..
9271 (*ashr<mode>3_1): this, ...
9272 (*lshr<mode>3_1): and this, also extend this pattern to avx512
9274 (*<insn><mode>3_1): Split to ..
9275 (*ashr<mode>3_1): this, ...
9276 (*lshrqi3_1): and this, also extend this pattern to avx512
9278 (*lshrhi3_1): And this, also extend this pattern to avx512
9280 * config/i386/sse.md (k<code><mode>): New define_split after
9281 it to convert generic shift pattern to mask shift ones.
9283 2021-07-21 Thomas Schwinge <thomas@codesourcery.com>
9284 Joseph Myers <joseph@codesourcery.com>
9285 Cesar Philippidis <cesar@codesourcery.com>
9287 * tree-core.h (omp_clause_code): Add 'OMP_CLAUSE_NOHOST'.
9288 * tree.c (omp_clause_num_ops, omp_clause_code_name, walk_tree_1):
9290 * tree-pretty-print.c (dump_omp_clause): Likewise.
9291 * omp-general.c (oacc_verify_routine_clauses): Likewise.
9292 * gimplify.c (gimplify_scan_omp_clauses)
9293 (gimplify_adjust_omp_clauses): Likewise.
9294 * tree-nested.c (convert_nonlocal_omp_clauses)
9295 (convert_local_omp_clauses): Likewise.
9296 * omp-low.c (scan_sharing_clauses): Likewise.
9297 * omp-offload.c (execute_oacc_device_lower): Update.
9299 2021-07-21 Martin Sebor <msebor@redhat.com>
9301 * tree-ssa-alias.c (walk_aliased_vdefs_1): Fix typos in a comment.
9303 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9305 * config/rs6000/rs6000-gen-builtins.c (write_init_bif_table):
9308 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9310 * config/rs6000/rs6000-gen-builtins.c (write_fntype): New
9312 (write_fntype_init): New stub function.
9313 (write_init_bif_table): Likewise.
9314 (write_init_ovld_table): New function.
9315 (write_init_file): Implement.
9317 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9319 * config/rs6000/rs6000-gen-builtins.c
9320 (write_autogenerated_header): New function.
9321 (write_decls): Likewise.
9322 (write_extern_fntype): New callback function.
9323 (write_header_file): Implement.
9325 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9327 * config/rs6000/rs6000-gen-builtins.c (write_defines_file):
9330 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9332 * config/rs6000/rs6000-gen-builtins.c (complete_vector_type): New
9334 (complete_base_type): Likewise.
9335 (construct_fntype_id): Likewise.
9336 (parse_bif_entry): Call contruct_fntype_id.
9337 (parse_ovld_entry): Likewise.
9339 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9341 * config/rs6000/rs6000-gen-builtins.c (ovld_stanza): New struct.
9342 (MAXOVLDSTANZAS): New macro.
9343 (ovld_stanzas): New variable.
9344 (curr_ovld_stanza): Likewise.
9345 (MAXOVLDS): New macro.
9346 (ovlddata): New struct.
9347 (ovlds): New variable.
9348 (curr_ovld): Likewise.
9349 (max_ovld_args): Likewise.
9350 (parse_ovld_entry): New function.
9351 (parse_ovld_stanza): Likewise.
9352 (parse_ovld): Implement.
9354 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9356 * config/rs6000/rs6000-gen-builtins.c (parse_bif_attrs):
9359 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9361 * config/rs6000/rs6000-gen-builtins.c (parse_args): New function.
9362 (parse_prototype): Implement.
9364 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9366 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): New enum.
9367 (curr_bif_stanza): New variable.
9368 (stanza_entry): New struct.
9369 (stanza_map): New initialized variable.
9370 (enable_string): Likewise.
9371 (fnkinds): New enum.
9372 (typelist): New struct.
9373 (attrinfo): Likewise.
9374 (MAXRESTROPNDS): New macro.
9375 (prototype): New struct.
9376 (MAXBIFS): New macro.
9377 (bifdata): New struct.
9378 (bifs): New variable.
9379 (curr_bif): Likewise.
9380 (bif_order): Likewise.
9381 (bif_index): Likewise.
9382 (fatal): New function.
9383 (stanza_name_to_stanza): Likewise.
9384 (parse_bif_attrs): New stub function.
9385 (parse_prototype): Likewise.
9386 (parse_bif_entry): New function.
9387 (parse_bif_stanza): Likewise.
9388 (parse_bif): Implement.
9389 (set_bif_order): New function.
9390 (create_bif_order): Implement.
9392 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
9394 * config/rs6000/rs6000-gen-builtins.c (rbtree.h): New #include.
9395 (num_bifs): New variable.
9396 (num_ovld_stanzas): Likewise.
9397 (num_ovlds): Likewise.
9398 (parse_codes): New enum.
9399 (bif_rbt): New variable.
9400 (ovld_rbt): Likewise.
9401 (fntype_rbt): Likewise.
9402 (bifo_rbt): Likewise.
9403 (parse_bif): New stub function.
9404 (create_bif_order): Likewise.
9405 (parse_ovld): Likewise.
9406 (write_header_file): Likewise.
9407 (write_init_file): Likewise.
9408 (write_defines_file): Likewise.
9409 (delete_output_files): New function.
9412 2021-07-21 H.J. Lu <hjl.tools@gmail.com>
9415 * config/i386/i386-builtin.def: Remove OPTION_MASK_ISA_SSE4_2
9416 from CRC32 _builtin functions.
9418 2021-07-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
9420 * coverage.c (build_gcov_info_var_registration): Mark the object placed
9421 in the linker set as referenced so that it does not get optimized away.
9423 2021-07-21 Kito Cheng <kito.cheng@sifive.com>
9426 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
9428 * config.gcc (riscv*-*-*): Detect which python is available.
9430 2021-07-21 Jakub Jelinek <jakub@redhat.com>
9432 PR middle-end/101535
9433 * gimplify.c (omp_check_private): Properly skip ORT_TARGET_DATA
9434 contexts in which decl isn't privatized and for ORT_TARGET return
9435 false if decl is mapped.
9437 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
9439 * gimple-loop-jam.c: Include tree-ssa-sccvn.h.
9440 (tree_loop_unroll_and_jam): Run value-numbering on a loop that
9441 has been successfully unrolled.
9443 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
9445 * tree-ssa-loop-manip.c (determine_exit_conditions): Return a null
9446 exit condition if no tail loop is needed, and if the original exit
9447 condition should therefore be kept as-is.
9448 (tree_transform_and_unroll_loop): Handle that case here too.
9450 2021-07-21 Kewen Lin <linkw@linux.ibm.com>
9452 * tree-data-ref.c (free_dependence_relations): Adjust to pass vec
9454 (free_data_refs): Likewise.
9455 * tree-data-ref.h (free_dependence_relations): Likewise.
9456 (free_data_refs): Likewise.
9457 * tree-predcom.c (struct chain): Use auto_vec instead of vec for
9459 (struct component): Likewise.
9460 (pcom_worker::pcom_worker): Adjust for auto_vec and renaming changes.
9461 (pcom_worker::~pcom_worker): Likewise.
9462 (pcom_worker::release_chain): Adjust as auto_vec changes.
9463 (pcom_worker::loop): Rename to ...
9464 (pcom_worker::m_loop): ... this.
9465 (pcom_worker::datarefs): Rename to ...
9466 (pcom_worker::m_datarefs): ... this. Use auto_vec instead of vec.
9467 (pcom_worker::dependences): Rename to ...
9468 (pcom_worker::m_dependences): ... this. Use auto_vec instead of vec.
9469 (pcom_worker::chains): Rename to ...
9470 (pcom_worker::m_chains): ... this. Use auto_vec instead of vec.
9471 (pcom_worker::looparound_phis): Rename to ...
9472 (pcom_worker::m_looparound_phis): ... this. Use auto_vec instead of
9474 (pcom_worker::cache): Rename to ...
9475 (pcom_worker::m_cache): ... this. Use auto_vec instead of vec.
9476 (pcom_worker::release_chain): Adjust for auto_vec changes.
9477 (pcom_worker::release_chains): Adjust for auto_vec and renaming
9479 (release_component): Remove.
9480 (release_components): Adjust for release_component removal.
9481 (component_of): Adjust to use vec.
9482 (merge_comps): Likewise.
9483 (pcom_worker::aff_combination_dr_offset): Adjust for renaming changes.
9484 (pcom_worker::determine_offset): Likewise.
9485 (class comp_ptrs): Remove.
9486 (pcom_worker::split_data_refs_to_components): Adjust for renaming
9487 changes, for comp_ptrs removal with auto_vec.
9488 (pcom_worker::suitable_component_p): Adjust for renaming changes.
9489 (pcom_worker::filter_suitable_components): Adjust for release_component
9491 (pcom_worker::valid_initializer_p): Adjust for renaming changes.
9492 (pcom_worker::find_looparound_phi): Likewise.
9493 (pcom_worker::add_looparound_copies): Likewise.
9494 (pcom_worker::determine_roots_comp): Likewise.
9495 (pcom_worker::single_nonlooparound_use): Likewise.
9496 (pcom_worker::execute_pred_commoning_chain): Likewise.
9497 (pcom_worker::execute_pred_commoning): Likewise.
9498 (pcom_worker::try_combine_chains): Likewise.
9499 (pcom_worker::prepare_initializers_chain): Likewise.
9500 (pcom_worker::prepare_initializers): Likewise.
9501 (pcom_worker::prepare_finalizers_chain): Likewise.
9502 (pcom_worker::prepare_finalizers): Likewise.
9503 (pcom_worker::tree_predictive_commoning_loop): Likewise.
9505 2021-07-20 Martin Sebor <msebor@redhat.com>
9507 PR middle-end/101397
9508 * builtins.c (gimple_call_return_array): Add argument. Correct
9509 offsets for memchr, mempcpy, stpcpy, and stpncpy.
9510 (compute_objsize_r): Adjust offset computation for argument returning
9513 2021-07-20 Martin Sebor <msebor@redhat.com>
9515 PR middle-end/101300
9516 * tree-ssa-uninit.c (check_defs): Handle UBSAN built-ins.
9518 2021-07-20 Jeff Law <jlaw@localhost.localdomain>
9520 * function.c (assign_parm_setup_block): Use adjust_address instead
9521 of change_address to preserve MEM_EXPR and friends.
9523 2021-07-20 Martin Sebor <msebor@redhat.com>
9525 * cfgloop.h (single_likely_exit): Adjust by-value argument to
9527 * cfgloopanal.c (single_likely_exit): Same.
9528 * cgraph.h (struct cgraph_node): Same.
9529 * cgraphclones.c (cgraph_node::create_virtual_clone): Same.
9530 * genautomata.c (merge_states): Same.
9531 * genextract.c (VEC_char_to_string): Same.
9532 * genmatch.c (dt_node::gen_kids_1): Same.
9533 (walk_captures): Adjust by-value argument to by-reference.
9534 * gimple-ssa-store-merging.c (check_no_overlap): Adjust by-value argument
9535 to by-const-reference.
9536 * gimple.c (gimple_build_call_vec): Same.
9537 (gimple_build_call_internal_vec): Same.
9538 (gimple_build_switch): Same.
9539 (sort_case_labels): Same.
9540 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
9542 * gimple.h (gimple_build_call_vec): Adjust by-value argument to
9544 (gimple_build_call_internal_vec): Same.
9545 (gimple_build_switch): Same.
9546 (sort_case_labels): Same.
9547 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
9549 * haifa-sched.c (calc_priorities): Adjust by-value argument to
9551 (sched_init_luids): Same.
9552 (haifa_init_h_i_d): Same.
9553 * ipa-cp.c (ipa_get_indirect_edge_target_1): Same.
9554 (adjust_callers_for_value_intersection): Adjust by-value argument to
9556 (find_more_scalar_values_for_callers_subset): Adjust by-value argument to
9558 (find_more_contexts_for_caller_subset): Same.
9559 (find_aggregate_values_for_callers_subset): Same.
9560 (copy_useful_known_contexts): Same.
9561 * ipa-fnsummary.c (remap_edge_summaries): Same.
9562 (remap_freqcounting_predicate): Same.
9563 * ipa-inline.c (add_new_edges_to_heap): Adjust by-value argument to
9565 * ipa-predicate.c (predicate::remap_after_inlining): Adjust by-value argument
9566 to by-const-reference.
9567 * ipa-predicate.h (predicate::remap_after_inlining): Same.
9568 * ipa-prop.c (ipa_find_agg_cst_for_param): Same.
9569 * ipa-prop.h (ipa_find_agg_cst_for_param): Same.
9570 * ira-build.c (ira_loop_tree_body_rev_postorder): Same.
9571 * read-rtl.c (add_overload_instance): Same.
9572 * rtl.h (native_decode_rtx): Same.
9573 (native_decode_vector_rtx): Same.
9574 * sched-int.h (sched_init_luids): Same.
9575 (haifa_init_h_i_d): Same.
9576 * simplify-rtx.c (native_decode_vector_rtx): Same.
9577 (native_decode_rtx): Same.
9578 * tree-call-cdce.c (gen_shrink_wrap_conditions): Same.
9579 (shrink_wrap_one_built_in_call_with_conds): Same.
9580 (shrink_wrap_conditional_dead_built_in_calls): Same.
9581 * tree-data-ref.c (create_runtime_alias_checks): Same.
9582 (compute_all_dependences): Same.
9583 * tree-data-ref.h (compute_all_dependences): Same.
9584 (create_runtime_alias_checks): Same.
9585 (index_in_loop_nest): Same.
9586 * tree-if-conv.c (mask_exists): Same.
9587 * tree-loop-distribution.c (class loop_distribution): Same.
9588 (loop_distribution::create_rdg_vertices): Same.
9589 (dump_rdg_partitions): Same.
9590 (debug_rdg_partitions): Same.
9591 (partition_contains_all_rw): Same.
9592 (loop_distribution::distribute_loop): Same.
9593 * tree-parloops.c (oacc_entry_exit_ok_1): Same.
9594 (oacc_entry_exit_single_gang): Same.
9595 * tree-ssa-loop-im.c (hoist_memory_references): Same.
9596 (loop_suitable_for_sm): Same.
9597 * tree-ssa-loop-niter.c (bound_index): Same.
9598 * tree-ssa-reassoc.c (update_ops): Same.
9599 (swap_ops_for_binary_stmt): Same.
9600 (rewrite_expr_tree): Same.
9601 (rewrite_expr_tree_parallel): Same.
9602 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Same.
9603 * tree-ssa-sccvn.h (ao_ref_init_from_vn_reference): Same.
9604 * tree-ssa-structalias.c (process_all_all_constraints): Same.
9605 (make_constraints_to): Same.
9606 (handle_lhs_call): Same.
9607 (find_func_aliases_for_builtin_call): Same.
9608 (sort_fieldstack): Same.
9609 (check_for_overlaps): Same.
9610 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Same.
9611 (vect_create_cond_for_unequal_addrs): Same.
9612 (vect_create_cond_for_lower_bounds): Same.
9613 (vect_create_cond_for_alias_checks): Same.
9614 * tree-vect-slp-patterns.c (vect_validate_multiplication): Same.
9615 * tree-vect-slp.c (vect_analyze_slp_instance): Same.
9616 (vect_make_slp_decision): Same.
9617 (vect_slp_bbs): Same.
9618 (duplicate_and_interleave): Same.
9619 (vect_transform_slp_perm_load): Same.
9620 (vect_schedule_slp): Same.
9621 * tree-vectorizer.h (vect_transform_slp_perm_load): Same.
9622 (vect_schedule_slp): Same.
9623 (duplicate_and_interleave): Same.
9624 * tree.c (build_vector_from_ctor): Same.
9625 (build_vector): Same.
9626 (check_vector_cst): Same.
9627 (check_vector_cst_duplicate): Same.
9628 (check_vector_cst_fill): Same.
9629 (check_vector_cst_stepped): Same.
9630 * tree.h (build_vector_from_ctor): Same.
9632 2021-07-20 Jakub Jelinek <jakub@redhat.com>
9635 * config/rs6000/rs6000-protos.h (easy_altivec_constant): Change return
9636 type from bool to int.
9637 * config/rs6000/rs6000.c (vspltis_constant): Fix up handling the
9638 EASY_VECTOR_MSB case if either step or copies is not 1.
9639 (vspltis_shifted): Fix comment typo.
9640 (easy_altivec_constant): Change return type from bool to int, instead
9641 of returning true return byte size of the element mode that should be
9642 used to synthetize the constant.
9643 * config/rs6000/predicates.md (easy_vector_constant_msb): Require
9644 that vspltis_shifted is 0, handle the case where easy_altivec_constant
9645 assumes using different vector mode from CONST_VECTOR's mode.
9646 * config/rs6000/altivec.md (easy_vector_constant_msb splitter): Use
9647 easy_altivec_constant to determine mode in which -1 >> -1 should be
9648 performed, use rs6000_expand_vector_init instead of gen_vec_initv4sisi.
9650 2021-07-20 Richard Biener <rguenther@suse.de>
9653 * dwarf2out.h (dwarf_file_data): Add key member.
9654 * dwarf2out.c (dwarf_file_hasher::equal): Compare key.
9655 (dwarf_file_hasher::hash): Hash key.
9656 (lookup_filename): Remap the filename and store it in the
9657 filename member of dwarf_file_data when creating a new
9659 (file_name_acquire): Do not remap the filename again.
9660 (maybe_emit_file): Likewise.
9662 2021-07-20 Jonathan Wright <jonathan.wright@arm.com>
9664 * config/aarch64/aarch64-simd-builtins.def: Use two variant
9665 generators for all TBL/TBX intrinsics and rename to
9666 consistent forms: qtbl[1234] or qtbx[1234].
9667 * config/aarch64/aarch64-simd.md (aarch64_tbl1<mode>):
9669 (aarch64_qtbl1<mode>): This.
9670 (aarch64_tbx1<mode>): Rename to...
9671 (aarch64_qtbx1<mode>): This.
9672 (aarch64_tbl2v16qi): Delete.
9673 (aarch64_tbl3<mode>): Rename to...
9674 (aarch64_qtbl2<mode>): This.
9675 (aarch64_tbx4<mode>): Rename to...
9676 (aarch64_qtbx2<mode>): This.
9677 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_1): Use
9678 renamed qtbl1 and qtbl2 RTL patterns.
9679 * config/aarch64/arm_neon.h (vqtbl1_p8): Use renamed qtbl1
9681 (vqtbl1_s8): Likewise.
9682 (vqtbl1_u8): Likewise.
9683 (vqtbl1q_p8): Likewise.
9684 (vqtbl1q_s8): Likewise.
9685 (vqtbl1q_u8): Likewise.
9686 (vqtbx1_s8): Use renamed qtbx1 RTL pattern.
9687 (vqtbx1_u8): Likewise.
9688 (vqtbx1_p8): Likewise.
9689 (vqtbx1q_s8): Likewise.
9690 (vqtbx1q_u8): Likewise.
9691 (vqtbx1q_p8): Likewise.
9692 (vtbl1_s8): Use renamed qtbl1 RTL pattern.
9693 (vtbl1_u8): Likewise.
9694 (vtbl1_p8): Likewise.
9695 (vtbl2_s8): Likewise
9696 (vtbl2_u8): Likewise.
9697 (vtbl2_p8): Likewise.
9698 (vtbl3_s8): Use renamed qtbl2 RTL pattern.
9699 (vtbl3_u8): Likewise.
9700 (vtbl3_p8): Likewise.
9701 (vtbl4_s8): Likewise.
9702 (vtbl4_u8): Likewise.
9703 (vtbl4_p8): Likewise.
9704 (vtbx2_s8): Use renamed qtbx2 RTL pattern.
9705 (vtbx2_u8): Likewise.
9706 (vtbx2_p8): Likewise.
9707 (vqtbl2_s8): Use renamed qtbl2 RTL pattern.
9708 (vqtbl2_u8): Likewise.
9709 (vqtbl2_p8): Likewise.
9710 (vqtbl2q_s8): Likewise.
9711 (vqtbl2q_u8): Likewise.
9712 (vqtbl2q_p8): Likewise.
9713 (vqtbx2_s8): Use renamed qtbx2 RTL pattern.
9714 (vqtbx2_u8): Likewise.
9715 (vqtbx2_p8): Likewise.
9716 (vqtbx2q_s8): Likewise.
9717 (vqtbx2q_u8): Likewise.
9718 (vqtbx2q_p8): Likewise.
9719 (vtbx4_s8): Likewise.
9720 (vtbx4_u8): Likewise.
9721 (vtbx4_p8): Likewise.
9723 2021-07-20 Uroš Bizjak <ubizjak@gmail.com>
9726 * config/i386/sync.md (define_peephole2 atomic_storedi_fpu):
9728 (define_peephole2 atomic_loaddi_fpu): Ditto.
9730 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
9732 * config.gcc (riscv*-*-*): Detect which python is available.
9734 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
9736 * config/rs6000/vsx.md (mulhs_<mode>): Rename to...
9737 (smul<mode>3_highpart): ... this.
9738 (mulhu_<mode>): Rename to...
9739 (umul<mode>3_highpart): ... this.
9740 * config/rs6000/rs6000-builtin.def (MULHS_V2DI, MULHS_V4SI,
9741 MULHU_V2DI, MULHU_V4SI): Adjust.
9743 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
9745 PR tree-optimization/100696
9746 * internal-fn.c (first_commutative_argument): Add info for IFN_MULH.
9747 * internal-fn.def (IFN_MULH): New internal function.
9748 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Add support to
9749 recog normal multiply highpart as IFN_MULH.
9750 * config/i386/i386.c (ix86_add_stmt_cost): Adjust for combined
9753 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
9755 * config/elfos.h (CTF_DEBUGGING_INFO): New definition.
9756 (BTF_DEBUGGING_INFO): Likewise.
9757 * doc/tm.texi.in: Document the new macros.
9758 * doc/tm.texi: Regenerated.
9759 * toplev.c: Guard initialization of debug hooks.
9761 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
9763 * flags.h (ctf_debuginfo_p): New function declaration.
9764 * opts.c (ctf_debuginfo_p): New function definition.
9766 2021-07-19 Andrew Stubbs <ams@codesourcery.com>
9769 * config/gcn/gcn-hsa.h (DRIVER_SELF_SPECS): New.
9770 (ASM_SPEC): Set -mattr for xnack and sram-ecc.
9771 * config/gcn/gcn-opts.h (enum sram_ecc_type): New.
9772 * config/gcn/gcn-valu.md: Add a warning comment.
9773 * config/gcn/gcn.c (gcn_option_override): Add "sorry" for -mxnack.
9774 (output_file_start): Add xnack and sram-ecc state to ".amdgcn_target".
9775 * config/gcn/gcn.md: Add a warning comment.
9776 * config/gcn/gcn.opt: Add -mxnack and -msram-ecc.
9777 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): Remove
9779 (EF_AMDGPU_XNACK): New.
9780 (EF_AMDGPU_SRAM_ECC): New.
9782 (copy_early_debug_info): Use elf_flags.
9783 (main): Handle -mxnack and -msram-ecc options.
9784 * doc/invoke.texi: Document -mxnack and -msram-ecc.
9786 2021-07-19 Andrew Pinski <apinski@marvell.com>
9789 * config/aarch64/aarch64.md (csneg3_uxtw_insn): Rename to ...
9790 (*cs<neg_not_cs>3_uxtw_insn4): and extend to NEG_NOT.
9792 2021-07-19 Richard Biener <rguenther@suse.de>
9794 PR tree-optimization/101505
9795 * tree-vect-patterns.c (vect_determine_precisions): Walk
9796 PHIs also for loop vectorization.
9798 2021-07-19 Richard Biener <rguenther@suse.de>
9800 * gimple.h (gimple_expr_type): Remove.
9801 * doc/gimple.texi: Remove gimple_expr_type documentation.
9803 2021-07-19 Richard Biener <rguenther@suse.de>
9805 * tree-ssa-sccvn.c (vn_reference_eq): Handle NULL vr->type.
9806 (ao_ref_init_from_vn_reference): Likewise.
9807 (fully_constant_reference): Likewise.
9808 (vn_reference_lookup_call): Do not set vr->type to random
9810 * tree-ssa-pre.c (compute_avail): Do not try to PRE calls
9812 * tree-vect-generic.c (expand_vector_piecewise): Pass in
9813 whether we expanded parallel.
9814 (expand_vector_parallel): Adjust.
9815 (expand_vector_addition): Likewise.
9816 (expand_vector_comparison): Likewise.
9817 (expand_vector_operation): Likewise.
9818 (expand_vector_scalar_condition): Likewise.
9819 (expand_vector_conversion): Likewise.
9821 2021-07-19 Richard Biener <rguenther@suse.de>
9823 * tree-vrp.c (register_edge_assert_for_2): Use the
9825 (vrp_folder::fold_predicate_in): Likewise.
9826 * vr-values.c (gimple_assign_nonzero_p): Likewise.
9827 (vr_values::extract_range_from_comparison): Likewise.
9828 (vr_values::extract_range_from_ubsan_builtin): Use the
9829 type of the first operand.
9830 (vr_values::extract_range_basic): Push down type
9831 computation, use the appropriate LHS.
9832 (vr_values::extract_range_from_assignment): Use the
9835 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
9838 * common/config/i386/i386-common.c (ix86_handle_option): For
9839 -mgeneral-regs-only, enable the GPR only instructions which are
9840 enabled implicitly by SSE ISAs unless they have been disabled
9843 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
9846 * config/i386/i386.c (ix86_check_avx_upper_stores): Moved before
9847 ix86_avx_u128_mode_needed.
9848 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY if callee
9849 returns AVX register.
9851 2021-07-17 Jan Hubicka <hubicka@ucw.cz>
9853 * tree-ssa-structalias.c (handle_rhs_call): Support EAF_NOT_RETURNED.
9854 (handle_const_call): Liekise
9855 (handle_pure_call): Liekise
9857 2021-07-17 Andrew MacLeod <amacleod@redhat.com>
9859 PR tree-optimization/96542
9860 * range-op.cc (range_operator::wi_fold_in_parts): New.
9861 (range_operator::fold_range): Call wi_fold_in_parts.
9862 (operator_lshift::wi_fold): Fix broken lshift by [0,0].
9863 * range-op.h (wi_fold_in_parts): Add prototype.
9865 2021-07-16 David Malcolm <dmalcolm@redhat.com>
9867 * doc/analyzer.texi: Add __analyzer_dump_state.
9869 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9871 * config/rs6000/rbtree.c: New file.
9872 * config/rs6000/rbtree.h: New file.
9874 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9876 * config/rs6000/rs6000-gen-builtins.c (restriction): New enum.
9877 (typeinfo): Add restr field.
9878 (match_bracketed_pair): New function.
9879 (match_const_restriction): Implement.
9881 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9883 * config/rs6000/rs6000-gen-builtins.c (match_basetype): Implement.
9885 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9887 * config/rs6000/rs6000-gen-builtins.c (void_status): New enum.
9888 (basetype): Likewise.
9889 (typeinfo): Likewise.
9890 (handle_pointer): New function.
9891 (match_basetype): New stub function.
9892 (match_const_restriction): Likewise.
9893 (match_type): New function.
9895 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9897 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace): New
9899 (advance_line): Likewise.
9900 (safe_inc_pos): Likewise.
9901 (match_identifier): Likewise.
9902 (match_integer): Likewise.
9903 (match_to_right_bracket): Likewise.
9905 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9907 * config/rs6000/rs6000-gen-builtins.c (bif_file): New variable.
9908 (ovld_file): Likewise.
9909 (header_file): Likewise.
9910 (init_file): Likewise.
9911 (defines_file): Likewise.
9912 (pgm_path): Likewise.
9913 (bif_path): Likewise.
9914 (ovld_path): Likewise.
9915 (header_path): Likewise.
9916 (init_path): Likewise.
9917 (defines_path): Likewise.
9918 (LINELEN): New macro.
9919 (linebuf): New variable.
9923 (bif_diag): New function.
9924 (ovld_diag): Likewise.
9926 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9928 * config/rs6000/rs6000-builtin-new.def: New.
9929 * config/rs6000/rs6000-overload.def: New.
9931 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9933 * config/rs6000/rs6000-gen-builtins.c: New.
9935 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9937 * Makefile.in (EXTRA_GTYPE_DEPS): New variable.
9938 (s-gtype): Depend on EXTRA_GTYPE_DEPS.
9939 * gengtype-state.c (state_writer::write_state_file_list): Add a
9940 parameter to the fileslist expression for the number of build
9942 (read_state_files_list): Detect build headers and strip the
9943 initial "./" or ".\" from their names.
9944 * gengtype.c (build_headers): New global variable.
9945 (num_build_headers): Likewise.
9946 (open_base_files): Emit #include for each build header.
9947 (main): Detect and count build headers.
9948 * gengtype.h (build_headers): New extern variable.
9949 (num_build_headers): Likewise.
9951 2021-07-16 Richard Biener <rguenther@suse.de>
9953 * gimple-ssa-store-merging.c (verify_symbolic_number_p): Use
9954 the type of the LHS.
9955 (find_bswap_or_nop_1): Likewise.
9956 (find_bswap_or_nop): Likewise.
9957 * tree-vectorizer.h (vect_get_smallest_scalar_type): Adjust
9959 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
9960 Remove unused parameters, pass in the scalar type. Fix
9961 internal store function handling.
9962 * tree-vect-stmts.c (vect_analyze_stmt): Remove assert.
9963 (vect_get_vector_types_for_stmt): Move down check for
9964 existing vector stmt after we've determined a scalar type.
9965 Pass down the used scalar type to vect_get_smallest_scalar_type.
9966 * tree-vect-generic.c (expand_vector_condition): Use
9967 the type of the LHS.
9968 (expand_vector_scalar_condition): Likewise.
9969 (expand_vector_operations_1): Likewise.
9970 * tree-vect-patterns.c (vect_widened_op_tree): Likewise.
9971 (vect_recog_dot_prod_pattern): Likewise.
9972 (vect_recog_sad_pattern): Likewise.
9973 (vect_recog_widen_op_pattern): Likewise.
9974 (vect_recog_widen_sum_pattern): Likewise.
9975 (vect_recog_mixed_size_cond_pattern): Likewise.
9977 2021-07-16 Jan Hubicka <hubicka@ucw.cz>
9979 * ipa-modref.c (struct escape_entry): Use eaf_fleags_t.
9980 (dump_eaf_flags): Dump EAF_NOT_RETURNED
9981 (eaf_flags_useful_p): Use eaf_fleags_t; handle const functions
9982 and EAF_NOT_RETURNED.
9983 (modref_summary::useful_p): Likewise.
9984 (modref_summary_lto::useful_p): Likewise.
9985 (struct) modref_summary_lto: Use eaf_fleags_t.
9986 (deref_flags): Handle EAF_NOT_RETURNED.
9987 (struct escape_point): Use min_flags.
9988 (modref_lattice::init): Add EAF_NOT_RETURNED.
9989 (merge_call_lhs_flags): Ignore EAF_NOT_RETURNED functions
9990 (analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
9992 (analyze_parms): Also analyze const functions; update conition on
9994 (modref_write): Update streaming.
9995 (read_section): Update streaming.
9996 (remap_arg_flags): Use eaf_flags_t.
9997 (modref_merge_call_site_flags): Hanlde EAF_NOT_RETURNED.
9998 * ipa-modref.h: (eaf_flags_t): New typedef.
9999 (struct modref_summary): Use eaf_flags_t.
10000 * tree-core.h (EAF_NOT_RETURNED): New constant.
10002 2021-07-16 Richard Biener <rguenther@suse.de>
10004 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Use
10005 the type of the LHS.
10006 (gimple_assign_nonnegative_warnv_p): Likewise.
10007 (gimple_call_nonnegative_warnv_p): Likewise. Return false
10008 if the call has no LHS.
10009 * gimple.c (gimple_could_trap_p_1): Use the type of the LHS.
10010 * tree-eh.c (stmt_could_throw_1_p): Likewise.
10011 * tree-inline.c (insert_init_stmt): Likewise.
10012 * tree-ssa-loop-niter.c (get_val_for): Likewise.
10013 * tree-outof-ssa.c (ssa_is_replaceable_p): Use the type of
10015 * tree-ssa-sccvn.c (init_vn_nary_op_from_stmt): Take a
10016 gassign *. Use the type of the lhs.
10017 (vn_nary_op_lookup_stmt): Adjust.
10018 (vn_nary_op_insert_stmt): Likewise.
10020 2021-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
10022 * config/s390/predicates.md (bras_sym_operand): Accept all
10023 functions in 64-bit mode, use UNSPEC_PLT31.
10024 (larl_operand): Use UNSPEC_PLT31.
10025 * config/s390/s390.c (s390_loadrelative_operand_p): Likewise.
10026 (legitimize_pic_address): Likewise.
10027 (s390_emit_tls_call_insn): Mark __tls_get_offset as function,
10029 (s390_delegitimize_address): Use UNSPEC_PLT31.
10030 (s390_output_addr_const_extra): Likewise.
10031 (print_operand): Add @PLT to TLS calls, handle %K.
10032 (s390_function_profiler): Mark __fentry__/_mcount as function,
10033 use %K, use UNSPEC_PLT31.
10034 (s390_output_mi_thunk): Use only UNSPEC_GOT, use %K.
10035 (s390_emit_call): Use UNSPEC_PLT31.
10036 (s390_emit_tpf_eh_return): Mark __tpf_eh_return as function.
10037 * config/s390/s390.md (UNSPEC_PLT31): Rename from UNSPEC_PLT.
10038 (*movdi_64): Use %K.
10039 (reload_base_64): Likewise.
10040 (*sibcall_brc): Likewise.
10041 (*sibcall_brcl): Likewise.
10042 (*sibcall_value_brc): Likewise.
10043 (*sibcall_value_brcl): Likewise.
10045 (*brasl): Likewise.
10046 (*bras_r): Likewise.
10047 (*brasl_r): Likewise.
10048 (*bras_tls): Likewise.
10049 (*brasl_tls): Likewise.
10050 (main_base_64): Likewise.
10051 (reload_base_64): Likewise.
10052 (@split_stack_call<mode>): Likewise.
10054 2021-07-16 Richard Biener <rguenther@suse.de>
10056 PR tree-optimization/101467
10057 * tree-vect-stmts.c (vect_gen_while): Properly guard
10058 make_temp_ssa_name usage.
10060 2021-07-16 Cooper Qu <cooper.qu@linux.alibaba.com>
10062 * config.gcc: Don't use forked print-sysroot-suffix.sh and
10063 t-sysroot-suffix for C-SKY.
10064 * config/csky/print-sysroot-suffix.sh: Delete.
10065 * config/csky/t-csky-linux: Delete.
10066 * config/csky/t-sysroot-suffix: Define MULTILIB_DIRNAMES
10067 instead of CSKY_MULTILIB_DIRNAMES.
10069 2021-07-16 Richard Biener <rguenther@suse.de>
10071 * tree-vect-loop.c (vect_transform_cycle_phi): Correct sign
10072 conversion issues with the partial reduction of the reused
10073 vector accumulator.
10075 2021-07-16 Richard Biener <rguenther@suse.de>
10077 * config/i386/i386-options.c (ix86_option_override_internal): Set
10078 param_vect_partial_vector_usage to zero if not set.
10080 2021-07-15 Uroš Bizjak <ubizjak@gmail.com>
10083 * config/i386/i386.h (VALID_SSE_REG_MODE): Add TDmode.
10084 (VALID_INT_MODE_P): Add SDmode and DDmode.
10085 Add TDmode for TARGET_64BIT.
10086 (VALID_DFP_MODE_P): Remove.
10087 * config/i386/i386.c (ix86_hard_regno_mode_ok):
10088 Do not use VALID_DFP_MODE_P.
10090 2021-07-15 Andrew MacLeod <amacleod@redhat.com>
10092 * gimple-range-fold.cc (adjust_pointer_diff_expr): Use
10094 (fold_using_range::fold_stmt): Ditto.
10095 (fold_using_range::range_of_range_op): Ditto.
10096 (fold_using_range::range_of_phi): Ditto.
10097 (fold_using_range::range_of_call): Ditto.
10098 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
10099 (fold_using_range::range_of_builtin_call): Ditto.
10100 (fold_using_range::range_of_cond_expr): Ditto.
10101 * gimple-range-fold.h (gimple_range_type): New.
10103 2021-07-15 Martin Sebor <msebor@redhat.com>
10105 PR middle-end/97027
10106 * tree-ssa-strlen.c (handle_assign): New function.
10107 (maybe_warn_overflow): Add argument.
10108 (nonzero_bytes_for_type): New function.
10109 (count_nonzero_bytes): Handle more tree types. Call
10110 nonzero_bytes_for_tye.
10111 (count_nonzero_bytes): Handle types.
10112 (handle_store): Handle stores from function calls.
10113 (strlen_check_and_optimize_call): Move code to handle_assign. Call
10114 it for assignments from function calls.
10116 2021-07-15 David Malcolm <dmalcolm@redhat.com>
10121 * doc/invoke.texi: Add -Wanalyzer-use-of-uninitialized-value.
10123 2021-07-15 David Malcolm <dmalcolm@redhat.com>
10125 * doc/invoke.texi (-fdump-analyzer-exploded-paths): New.
10127 2021-07-15 Martin Sebor <msebor@redhat.com>
10131 * fold-const.c (operand_compare::operand_equal_p): Handle OEP_DECL_NAME.
10132 (operand_compare::verify_hash_value): Same.
10133 * tree-core.h (OEP_DECL_NAME): New.
10135 2021-07-15 Martin Jambor <mjambor@suse.cz>
10137 * profile-count.h (profile_count::value): Change the return type to
10139 * gimple-pretty-print.c (dump_gimple_bb_header): Adjust print
10141 * tree-cfg.c (dump_function_to_file): Likewise.
10143 2021-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
10146 * config/rs6000/rs6000-p8swap.c (has_part_mult): New.
10147 (rs6000_analyze_swaps): Insns containing a subreg of a mult are
10150 2021-07-15 Richard Biener <rguenther@suse.de>
10152 * tree-vectorizer.h (vect_gen_while): Match up with
10153 vect_gen_while_not.
10154 * tree-vect-stmts.c (vect_gen_while): Adjust API to that
10155 of vect_gen_while_not.
10156 (vect_gen_while_not): Adjust.
10157 * tree-vect-loop-manip.c (vect_set_loop_controls_directly): Likewise.
10159 2021-07-15 Aldy Hernandez <aldyh@redhat.com>
10161 * gimple-range-cache.cc (non_null_ref::adjust_range): New.
10162 (ranger_cache::range_of_def): Call adjust_range.
10163 (ranger_cache::entry_range): Same.
10164 * gimple-range-cache.h (non_null_ref::adjust_range): New.
10165 * gimple-range.cc (gimple_ranger::range_of_expr): Call
10167 (gimple_ranger::range_on_entry): Same.
10169 2021-07-15 Tamar Christina <tamar.christina@arm.com>
10172 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10174 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
10176 2021-07-15 Tamar Christina <tamar.christina@arm.com>
10179 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10181 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
10182 (sdot_prod, udot_prod): ...These.
10183 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
10184 (aarch64_<sur>dot<vsi2qi>): Rename to...
10185 (<sur>dot_prod<vsi2qi>): ...This.
10186 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
10189 2021-07-15 Jakub Jelinek <jakub@redhat.com>
10191 PR middle-end/101437
10192 * gimplify.c (gimplify_expr): Throw away volatile reads from empty
10193 types even if they have non-BLKmode TYPE_MODE.
10195 2021-07-15 Richard Biener <rguenther@suse.de>
10198 * gcc.c (process_command): Process -gtoggle like process_options
10199 would after parsing options.
10201 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
10203 * cfgexpand.c (expand_asm_loc): Adjust.
10204 (expand_asm_stmt): Likewise.
10205 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Likewise.
10206 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
10207 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
10208 * config/avr/avr.c (avr_md_asm_adjust): Likewise.
10209 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
10210 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
10211 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
10212 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
10213 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
10214 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
10215 * config/s390/s390.c (s390_md_asm_adjust): Likewise.
10216 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
10217 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
10218 * doc/tm.texi: Regenerate.
10219 * target.def: Add location argument to md_asm_adjust.
10221 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
10223 * tree-diagnostic.c (diagnostic_report_current_function): Use the
10224 diagnostic's location, not input_location.
10226 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
10228 * cfgexpand.c (tree_conflicts_with_clobbers_p): Pass location to
10230 (expand_asm_stmt): Likewise.
10232 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
10234 * config/rs6000/rs6000.c (adjacent_mem_locations): Return the lower
10235 addressed memory rtx, if any.
10236 (rs6000_split_multireg_move): Fix code formatting.
10237 Handle MMA build built-ins with operands in adjacent memory locations.
10239 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
10241 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Move to later
10244 2021-07-14 Jason Merrill <jason@redhat.com>
10246 * sel-sched-ir.h (get_all_loop_exits): Use auto_vec.
10248 2021-07-14 Jason Merrill <jason@redhat.com>
10250 * doc/invoke.texi: -fdelete-dead-exceptions is on by default for
10253 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10255 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
10256 Remove erroneous line.
10258 2021-07-14 Andrew MacLeod <amacleod@redhat.com>
10260 * params.opt (param_evrp_mode): Change default.
10262 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10264 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
10265 (sdot_prod, udot_prod): ...These.
10266 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
10267 (aarch64_<sur>dot<vsi2qi>): Rename to...
10268 (<sur>dot_prod<vsi2qi>): ...This.
10269 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
10272 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10274 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
10276 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10278 * doc/sourcebuild.texi (arm_v8_2a_i8mm_neon_hw): Document.
10280 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10282 * config/arm/neon.md (usdot_prod<vsi2qi>): New.
10284 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10286 * config/aarch64/aarch64-simd.md (aarch64_usdot<vsi2qi>): Rename to...
10287 (usdot_prod<vsi2qi>): ... This.
10288 * config/aarch64/aarch64-simd-builtins.def (usdot): Rename to...
10289 (usdot_prod): ...This.
10290 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Likewise.
10291 * config/aarch64/aarch64-sve.md (@aarch64_<sur>dot_prod<vsi2qi>):
10293 (@<sur>dot_prod<vsi2qi>): ...This.
10294 * config/aarch64/aarch64-sve-builtins-base.cc
10295 (svusdot_impl::expand): Use it.
10297 2021-07-14 Tamar Christina <tamar.christina@arm.com>
10299 * optabs.def (usdot_prod_optab): New.
10300 * doc/md.texi: Document it and clarify other dot prod optabs.
10301 * optabs-tree.h (enum optab_subtype): Add optab_vector_mixed_sign.
10302 * optabs-tree.c (optab_for_tree_code): Support usdot_prod_optab.
10303 * optabs.c (expand_widen_pattern_expr): Likewise.
10304 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
10305 * tree-vect-loop.c (vectorizable_reduction): Query dot-product kind.
10306 * tree-vect-patterns.c (vect_supportable_direct_optab_p): Take optional
10308 (vect_widened_op_tree): Optionally ignore
10310 (vect_recog_dot_prod_pattern): Support usdot_prod_optab.
10312 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
10315 * config/i386/driver-i386.c (host_detect_local_cpu): Check
10316 "arch [32|64]" and "tune [32|64]" for 32-bit and 64-bit codegen.
10317 Enable UINTR only for 64-bit codegen.
10318 * config/i386/i386-options.c
10319 (ix86_option_override_internal::DEF_PTA): Skip PTA_UINTR if not
10321 * config/i386/i386.h (ARCH_ARG): New.
10322 (CC1_CPU_SPEC): Pass "[arch|tune] 32" for 32-bit codegen and
10323 "[arch|tune] 64" for 64-bit codegen.
10325 2021-07-14 Richard Biener <rguenther@suse.de>
10327 PR tree-optimization/101445
10328 * tree-vect-stmts.c (vectorizable_load): Do the gap adjustment
10329 of the IV in the correct direction for negative stride
10332 2021-07-14 Jakub Jelinek <jakub@redhat.com>
10335 * godump.c (godump_str_hash): New type.
10336 (godump_container::pot_dummy_types): Use string_hash instead of
10337 ptr_hash in the hash_set.
10339 2021-07-14 Richard Biener <rguenther@suse.de>
10341 * tree-vect-loop.c (vect_find_reusable_accumulator): Handle
10342 vector types where the old vector type has a multiple of
10343 the new vector type elements.
10344 (vect_create_partial_epilog): New function, split out from...
10345 (vect_create_epilog_for_reduction): ... here.
10346 (vect_transform_cycle_phi): Reduce the re-used accumulator
10347 to the new vector type.
10349 2021-07-14 Alexandre Oliva <oliva@adacore.com>
10351 * tree-ssa-alias.c (attr_fnspec::verify): Fix index in
10352 non-'t'-sized arg check.
10354 2021-07-14 Alexandre Oliva <oliva@adacore.com>
10356 * tree-cfg.c (cleanup_dead_labels_eh): Update
10357 post_landing_pad label upon change of landing pad block's
10359 (cleanup_dead_labels): Check that a removed label is not that
10362 2021-07-13 Jonathan Wright <jonathan.wright@arm.com>
10364 * combine.c (combine_simplify_rtx): Add vec_select -> subreg
10366 * config/aarch64/aarch64.md (*zero_extend<SHORT:mode><GPI:mode>2_aarch64):
10367 Add Neon to general purpose register case for zero-extend
10369 * config/arm/vfp.md (*arm_movsi_vfp): Remove "*" from *t -> r
10370 case to prevent some cases opting to go through memory.
10371 * cse.c (fold_rtx): Add vec_select -> subreg simplification.
10372 * rtl.c (rtvec_series_p): Define predicate to determine
10373 whether a vector contains a linear series of integers.
10374 * rtl.h (rtvec_series_p): Define.
10375 * rtlanal.c (vec_series_lowpart_p): Define predicate to
10376 determine if a vector selection is equivalent to the low part
10378 * rtlanal.h (vec_series_lowpart_p): Define.
10379 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
10380 Add vec_select -> subreg simplification.
10382 2021-07-13 Paul A. Clarke <pc@us.ibm.com>
10384 * config/rs6000/smmintrin.h (_mm_testz_si128, _mm_testc_si128,
10385 _mm_testnzc_si128, _mm_test_all_ones, _mm_test_all_zeros,
10386 _mm_test_mix_ones_zeros): New.
10388 2021-07-13 Roger Sayle <roger@nextmovesoftware.com>
10389 Richard Biener <rguenther@suse.de>
10391 * gimple.c (gimple_could_trap_p_1): Make S argument a
10392 "const gimple*". Preserve constness in call to
10393 gimple_asm_volatile_p.
10394 (gimple_could_trap_p): Make S argument a "const gimple*".
10395 * gimple.h (gimple_could_trap_p_1, gimple_could_trap_p):
10396 Update function prototypes.
10398 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10400 * tree-vectorizer.h (vect_reusable_accumulator): New structure.
10401 (_loop_vec_info::main_loop_edge): New field.
10402 (_loop_vec_info::skip_main_loop_edge): Likewise.
10403 (_loop_vec_info::skip_this_loop_edge): Likewise.
10404 (_loop_vec_info::reusable_accumulators): Likewise.
10405 (_stmt_vec_info::reduc_scalar_results): Likewise.
10406 (_stmt_vec_info::reused_accumulator): Likewise.
10407 (vect_get_main_loop_result): Declare.
10408 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize
10409 reduc_scalar_inputs.
10410 (vec_info::free_stmt_vec_info): Free reduc_scalar_inputs.
10411 * tree-vect-loop-manip.c (vect_get_main_loop_result): New function.
10412 (vect_do_peeling): Fill an epilogue loop's main_loop_edge,
10413 skip_main_loop_edge and skip_this_loop_edge fields.
10414 * tree-vect-loop.c (INCLUDE_ALGORITHM): Define.
10415 (vect_emit_reduction_init_stmts): New function.
10416 (get_initial_def_for_reduction): Use it.
10417 (get_initial_defs_for_reduction): Likewise. Change the vinfo
10418 parameter to a loop_vec_info.
10419 (vect_create_epilog_for_reduction): Store the scalar results
10420 in the reduc_info. If an epilogue loop is reusing an accumulator
10421 from the main loop, and if the epilogue loop can also be skipped,
10422 try to place the reduction code in the join block. Record
10423 accumulators that could potentially be reused by epilogue loops.
10424 (vect_transform_cycle_phi): When vectorizing epilogue loops,
10425 try to reuse accumulators from the main loop. Record the initial
10426 value in reduc_info for non-SLP reductions too.
10428 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10430 * tree-vect-loop.c (get_initial_def_for_reduction): Remove
10431 adjustment handling. Take the neutral value as an argument,
10432 in place of the code argument.
10433 (vect_transform_cycle_phi): Update accordingly. Handle the
10434 initial values of cond reductions separately from code reductions.
10435 Choose the adjustment here rather than in
10436 get_initial_def_for_reduction. Sink the splat of vec_initial_def.
10438 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10440 * tree-vect-loop.c (neutral_op_for_slp_reduction): Replace with...
10441 (neutral_op_for_reduction): ...this, providing a more general
10443 (vect_create_epilog_for_reduction): Update accordingly.
10444 (vectorizable_reduction): Likewise.
10445 (vect_transform_cycle_phi): Likewise.
10447 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10449 * tree-vect-loop.c (get_initial_def_for_reduction): Take the
10450 reduc_info instead of the original stmt_vec_info.
10451 (vect_transform_cycle_phi): Update accordingly.
10453 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10455 * tree-vect-loop.c (get_initial_defs_for_reduction): Take the
10456 reduc_info as an additional parameter.
10457 (vect_transform_cycle_phi): Update accordingly.
10459 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10461 * tree-vectorizer.h: Include tree-ssa-operands.h.
10462 (vect_phi_initial_value): New function.
10463 * tree-vect-loop.c (neutral_op_for_slp_reduction): Use it.
10464 (get_initial_defs_for_reduction, info_for_reduction): Likewise.
10465 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
10466 (vect_transform_cycle_phi, vectorizable_induction): Likewise.
10468 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10470 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
10471 the phi results to vectype after creating them. Remove later
10472 conversion code that thus becomes redundant.
10474 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10476 * tree-vect-loop.c (vect_create_epilog_for_reduction): Replace
10477 the new_phis vector with a reduc_inputs vector. Combine handling
10478 of reduction chains and ncopies > 1.
10480 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10482 * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
10483 scalar_results to group_size elements after reducing down from
10484 N*group_size elements. Construct an array_slice of the live-out
10485 stmts and assert that there is one stmt per scalar result.
10487 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10489 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
10490 nested_in_vect_loop and use double_reduc everywhere. Remove dead
10491 assignment to "loop".
10493 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10495 * internal-fn.c (vectorized_internal_fn_supported_p): Handle
10496 vector types first. For scalar types, consider both the preferred
10497 vector mode and the alternative vector modes.
10498 * optabs-query.c (can_vec_mask_load_store_p): Use the same
10499 structure as above, in particular using related_vector_mode
10500 for modes provided by autovectorize_vector_modes.
10502 2021-07-13 Jakub Jelinek <jakub@redhat.com>
10503 Richard Biener <rguenther@suse.de>
10505 PR tree-optimization/101419
10506 * tree-pass.h (PROP_objsz): Define.
10507 (make_pass_early_object_sizes): Declare.
10508 * passes.def (pass_all_early_optimizations): Rename pass_object_sizes
10509 there to pass_early_object_sizes, drop parameter.
10510 (pass_all_optimizations): Move pass_object_sizes right after pass_ccp,
10511 drop parameter, move pass_post_ipa_warn right after that.
10512 * tree-object-size.c (pass_object_sizes::execute): Rename to...
10513 (object_sizes_execute): ... this. Add insert_min_max_p argument.
10514 (pass_data_object_sizes): Move after object_sizes_execute.
10515 (pass_object_sizes): Likewise. In execute method call
10516 object_sizes_execute, drop set_pass_param method and insert_min_max_p
10517 non-static data member and its initializer in the ctor.
10518 (pass_data_early_object_sizes, pass_early_object_sizes,
10519 make_pass_early_object_sizes): New.
10520 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Use
10521 (cfun->curr_properties & PROP_objsz) instead of cfun->after_inlining.
10523 2021-07-13 Kito Cheng <kito.cheng@sifive.com>
10526 * config/riscv/constraints.md ("S"): Update description and remove
10528 * doc/md.texi (Machine Constraints): Document the 'S' constraints
10531 2021-07-13 Richard Biener <rguenther@suse.de>
10534 2021-07-12 Richard Biener <rguenther@suse.de>
10536 * tree-vect-slp.c (vect_slp_region): Show the number of
10537 SLP graph entries in the optimization message.
10539 2021-07-13 Michael Meissner <meissner@linux.ibm.com>
10541 * config/rs6000/altivec.md (xxspltiw_v4sf): Change local variable
10543 * config/rs6000/rs6000-protos.h (rs6000_const_f32_to_i32): Change
10544 return type to long.
10545 * config/rs6000/rs6000.c (rs6000_const_f32_to_i32): Change return
10548 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
10550 * gimple-range-fold.cc (fold_using_range::range_of_builtin_ubsan_call):
10551 Query relation between the 2 operands and use it.
10553 2021-07-12 Sergei Trofimovich <siarheit@google.com>
10555 * doc/cfg.texi: Fix s/ei_safe_safe/ei_safe_edge/ typo.
10557 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
10560 * config/i386/predicates.md (vec_setm_sse41_operand):
10561 Rename from vec_setm_operand.
10562 (vec_setm_avx2_operand): New predicate.
10563 * config/i386/sse.md (vec_set<V_128:mode>): Use V_128 mode iterator.
10564 Use vec_setm_sse41_operand as operand 2 predicate.
10565 (vec_set<V_256_512:mode): New expander.
10566 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_sse41_operand
10567 as operand 2 predicate.
10569 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
10571 PR tree-optimization/101335
10572 * range-op.cc (operator_cast::lhs_op1_relation): Delete.
10574 2021-07-12 Andrew Pinski <apinski@marvell.com>
10576 * tree-ssa-phiopt.c (match_simplify_replacement): Move
10577 insert of the sequence before the movement of the
10578 statement. Check if to see if the statement is used
10579 outside of the original phi to see if we should move it.
10581 2021-07-12 Richard Biener <rguenther@suse.de>
10583 * dump-context.h (debug_dump_context::debug_dump_context):
10584 Add FILE * parameter defaulted to stderr.
10585 * dumpfile.c (debug_dump_context::debug_dump_context): Adjust.
10586 * tree-vect-slp.c (dot_slp_tree): New functions.
10588 2021-07-12 Richard Biener <rguenther@suse.de>
10590 PR tree-optimization/101373
10591 * tree-ssa-pre.c (prune_clobbered_mems): Also prune trapping
10592 references when the BB may not return.
10593 (compute_avail): Pass in the function we're working on and
10594 replace cfun references with it. Externally throwing
10595 const calls also possibly terminate the function.
10596 (pass_pre::execute): Pass down the function we're working on.
10597 * gcse.c (compute_hash_table_work): Externally throwing
10598 const/pure calls also need record_last_mem_set_info.
10599 * postreload-gcse.c (record_opr_changes): Looping or externally
10600 throwing const/pure calls also need record_last_mem_set_info.
10602 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
10604 * recog.c (memory_address_addr_space_p): Change the type to bool.
10605 Return true/false instead of 1/0.
10606 (offsettable_memref_p): Ditto.
10607 (offsettable_nonstrict_memref_p): Ditto.
10608 (offsettable_address_addr_space_p): Ditto.
10609 Change the type of addressp indirect function to bool.
10610 * recog.h (memory_address_addr_space_p): Change the type to bool.
10611 (strict_memory_address_addr_space_p): Ditto.
10612 (offsettable_memref_p): Ditto.
10613 (offsettable_nonstrict_memref_p): Ditto.
10614 (offsettable_address_addr_space_p): Ditto.
10615 * reload.c (maybe_memory_address_addr_space_p): Ditto.
10616 (strict_memory_address_addr_space_p): Change the type to bool.
10617 Return true/false instead of 1/0.
10618 (maybe_memory_address_addr_space_p): Change the type to bool.
10620 2021-07-12 Richard Biener <rguenther@suse.de>
10622 * tree-vect-slp.c (vect_slp_region): Show the number of
10623 SLP graph entries in the optimization message.
10625 2021-07-12 Richard Biener <rguenther@suse.de>
10627 PR tree-optimization/101394
10628 * tree-ssa-pre.c (do_pre_regular_insertion): Avoid inserting
10629 copies from abnormals for a full redundancy.
10631 2021-07-12 Richard Biener <rguenther@suse.de>
10633 PR middle-end/101423
10634 * gimple.c (gimple_could_trap_p_1): Internal function calls
10636 * tree-eh.c (tree_could_trap_p): Likewise.
10638 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
10641 * config/arm/arm_neon.h (vmul_n_u32): Replace call to builtin with
10643 (vmulq_n_u32): Likewise.
10644 (vmul_n_f32): Gate __a * __b on __FAST_MATH__.
10645 (vmulq_n_f32): Likewise.
10646 (vmul_n_f16): Likewise.
10647 (vmulq_n_f16): Likewise.
10649 2021-07-12 Martin Liska <mliska@suse.cz>
10651 PR sanitizer/101425
10652 * gcc.c (check_offload_target_name): Call
10653 candidates_list_and_hint only if we have a candidate.
10655 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
10658 * config/arm/neon.md (vec_init): Move to ...
10659 * config/arm/vec-common.md (vec_init): ... here.
10660 Change the pattern's mode to VDQX and gate it on VALID_MVE_MODE.
10662 2021-07-12 Roger Sayle <roger@nextmovesoftware.com>
10664 PR tree-optimization/101403
10665 * match.pd ((T)bswap(X)>>C): Correctly handle cases where
10666 signedness of the shift is not the same as the signedness of
10667 the type extension.
10669 2021-07-09 Roger Sayle <roger@nextmovesoftware.com>
10670 Uroš Bizjak <ubizjak@gmail.com>
10672 * config/i386/i386.md (*divmodsi4_const): Optimize SImode
10673 divmod of a constant numerator with new define_insn_and_split.
10675 2021-07-09 Iain Sandoe <iain@sandoe.co.uk>
10678 * config/i386/i386-expand.c (ix86_expand_call): If a call is
10679 to a non-local-binding, or local but to a public symbol, then
10680 assume that it might be indirected via the lazy symbol binder.
10681 Mark R10 and R10 as clobbered in that case.
10683 2021-07-09 Eric Botcazou <ebotcazou@adacore.com>
10686 * gcc.c (ASM_DEBUG_DWARF_OPTION): Set again to --gdwarf2 in
10687 the case where HAVE_AS_WORKING_DWARF_N_FLAG is not defined
10688 and HAVE_LD_BROKEN_PE_DWARF5 is defined.
10690 2021-07-09 Uroš Bizjak <ubizjak@gmail.com>
10692 * config/i386/i386.md (*udivmodsi4_pow2_zext_1): Limit the
10693 log2 range of operands[3] to [1,31].
10694 (*udivmodsi4_pow2_zext_2): Ditto. Correct insn RTX pattern.
10696 2021-07-09 Sergei Trofimovich <siarheit@google.com>
10698 * doc/md.texi: Don't split @smallexample in multiple @groups.
10700 2021-07-09 Sergei Trofimovich <siarheit@google.com>
10702 * doc/md.texi: Add missing 'see' word.
10704 2021-07-09 Andrew Pinski <apinski@marvell.com>
10706 * tree-ssa-phiopt.c (phiopt_early_allow): Change arguments
10707 to take sequence and gimple_match_op. Accept the case where
10708 op is a SSA_NAME and one statement in the sequence.
10709 Also allow constants.
10710 (gimple_simplify_phiopt): Always pass a sequence to resimplify.
10711 Update call to phiopt_early_allow. Discard the sequence if not
10714 2021-07-09 Xi Ruoyao <xry111@mengyan1223.wang>
10719 * config/mips/mips.c (mips_const_insns): Use MSA_SUPPORTED_MODE_P
10720 instead of ISA_HAS_MSA.
10721 (mips_expand_vec_unpack): Likewise.
10722 (mips_expand_vector_init): Likewise.
10724 2021-07-09 Kewen Lin <linkw@linux.ibm.com>
10726 * config/rs6000/vsx.md (mods_<mode>): Rename to...
10727 (mod<mode>3): ... this.
10728 (modu_<mode>): Rename to...
10729 (umod<mode>3): ... this.
10730 * config/rs6000/rs6000-builtin.def (MODS_V2DI, MODS_V4SI, MODU_V2DI,
10731 MODU_V4SI): Adjust.
10733 2021-07-08 Jeff Law <jeffreyalaw@gmail.com>
10735 * config/h8300/shiftrotate.md (variable shifts): Expose condition
10736 code handling for the test before the loop.
10738 2021-07-08 Martin Jambor <mjambor@suse.cz>
10741 * ipa-sra.c (class isra_call_summary): New member
10742 m_before_any_store, initialize it in the constructor.
10743 (isra_call_summary::dump): Dump the new field.
10744 (ipa_sra_call_summaries::duplicate): Copy it.
10745 (process_scan_results): Set it.
10746 (isra_write_edge_summary): Stream it.
10747 (isra_read_edge_summary): Likewise.
10748 (param_splitting_across_edge): Only override
10749 safe_to_import_accesses if m_before_any_store is set.
10751 2021-07-08 Martin Sebor <msebor@redhat.com>
10753 PR bootstrap/101374
10754 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref):
10755 Use Object Size Type 0 instead of 1.
10757 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
10759 * tree-vect-loop.c (vectorizable_reduction): Remove always-true
10762 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
10764 * match.pd: Simplify an extend-operate-truncate sequence involving
10767 2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
10768 Richard Biener <rguenther@suse.de>
10770 PR tree-optimization/40210
10771 * match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
10772 (x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
10773 when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
10775 2021-07-08 Uroš Bizjak <ubizjak@gmail.com>
10778 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
10780 * config/i386/mmx.md (V_32): New mode iterator.
10781 (mov<V_32:mode>): Use V_32 mode iterator.
10782 (*mov<V_32:mode>_internal): Ditto.
10783 (*push<V_32:mode>2_rex64): Ditto.
10784 (*push<V_32:mode>2): Ditto.
10785 (movmisalign<V_32:mode>): Ditto.
10786 (mmx_<any_shiftrt:insn>v1si3): New insn pattern.
10787 (sse4_1_<any_extend:code>v2qiv2hi2): Ditto.
10788 (vec_unpacks_lo_v4qi): New expander.
10789 (vec_unpacks_hi_v4qi): Ditto.
10790 (vec_unpacku_lo_v4qi): Ditto.
10791 (vec_unpacku_hi_v4qi): Ditto.
10792 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add V1SImode.
10793 (VALID_INT_MODE_P): Ditto.
10795 2021-07-08 Michael Meissner <meissner@linux.ibm.com>
10798 * config/rs6000/rs6000.md (udivti3): New insn.
10799 (divti3): New insn.
10800 (umodti3): New insn.
10801 (modti3): New insn.
10803 2021-07-07 Martin Sebor <msebor@redhat.com>
10805 PR tree-optimization/100137
10806 PR tree-optimization/99121
10807 PR tree-optimization/97027
10808 * builtins.c (access_ref::access_ref): Also set offmax.
10809 (access_ref::offset_in_range): Define new function.
10810 (access_ref::add_offset): Set offmax.
10811 (access_ref::inform_access): Handle access_none.
10812 (handle_mem_ref): Clear ostype.
10813 (compute_objsize_r): Handle ASSERT_EXPR.
10814 * builtins.h (struct access_ref): Add offmax member.
10815 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Use
10816 compute_objsize() and simplify.
10818 2021-07-07 Peter Bergner <bergner@linux.ibm.com>
10820 * config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
10821 and VSX_BUILTIN_STXVP.
10823 2021-07-07 Martin Sebor <msebor@redhat.com>
10826 * config/aarch64/aarch64.c (aarch64_simd_lane_bounds): Remove
10827 a stray %K from error_at() missed in r12-2088.
10829 2021-07-07 Richard Biener <rguenther@suse.de>
10831 PR tree-optimization/99728
10832 * tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
10834 (mem_refs_may_alias_p): Add assert we handled aggregate
10836 (sm_seq_valid_bb): Give up when running into aggregate copies.
10837 (ref_indep_loop_p): Handle aggregate copies as never
10838 being invariant themselves but allow other refs to be
10839 disambiguated against them.
10840 (can_sm_ref_p): Do not try to apply store-motion to aggregate
10843 2021-07-06 Indu Bhagat <indu.bhagat@oracle.com>
10846 * dwarf2ctf.c (ctf_get_AT_data_member_location): Multiply by 8 to get
10849 2021-07-06 Martin Sebor <msebor@redhat.com>
10851 * gimple-pretty-print.c (percent_G_format): Remove.
10852 * tree-diagnostic.c (default_tree_printer): Remove calls.
10853 * tree-pretty-print.c (percent_K_format): Remove.
10854 * tree-pretty-print.h (percent_K_format): Remove.
10856 2021-07-06 Martin Sebor <msebor@redhat.com>
10858 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_builtin):
10859 Remove %K and use error_at.
10860 (aarch64_expand_fcmla_builtin): Same.
10861 (aarch64_expand_builtin_tme): Same.
10862 (aarch64_expand_builtin_memtag): Same.
10863 * config/arm/arm-builtins.c (arm_expand_acle_builtin): Same.
10864 (arm_expand_builtin): Same.
10865 * config/arm/arm.c (bounds_check): Same.
10867 2021-07-06 Martin Sebor <msebor@redhat.com>
10869 * builtins.c (warn_string_no_nul): Remove %G.
10870 (maybe_warn_for_bound): Same.
10871 (warn_for_access): Same.
10872 (check_access): Same.
10873 (check_strncat_sizes): Same.
10874 (expand_builtin_strncat): Same.
10875 (expand_builtin_strncmp): Same.
10876 (expand_builtin): Same.
10877 (expand_builtin_object_size): Same.
10878 (warn_dealloc_offset): Same.
10879 (maybe_emit_free_warning): Same.
10880 * calls.c (maybe_warn_alloc_args_overflow): Same.
10881 (maybe_warn_nonstring_arg): Same.
10882 (maybe_warn_rdwr_sizes): Same.
10883 * expr.c (expand_expr_real_1): Remove %K.
10884 * gimple-fold.c (gimple_fold_builtin_strncpy): Remove %G.
10885 (gimple_fold_builtin_strncat): Same.
10886 * gimple-ssa-sprintf.c (format_directive): Same.
10887 (handle_printf_call): Same.
10888 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Same.
10889 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
10890 (maybe_diag_access_bounds): Same. Call gimple_location.
10891 (check_bounds_or_overlap): Same.
10892 * trans-mem.c (ipa_tm_scan_irr_block): Remove %K. Simplify.
10893 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Remove %G.
10894 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
10895 (maybe_diag_stxncpy_trunc): Same.
10896 (handle_builtin_stxncpy_strncat): Same.
10897 (maybe_warn_pointless_strcmp): Same.
10898 * tree-ssa-uninit.c (maybe_warn_operand): Same.
10900 2021-07-06 Uroš Bizjak <ubizjak@gmail.com>
10903 * config/i386/predicates.md (vec_setm_operand): Enable
10904 register_operand for TARGET_SSE4_1.
10905 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_operand
10906 as operand 2 predicate. Call ix86_expand_vector_set_var
10907 for non-constant index operand.
10908 (vec_setv4qi): Use vec_setm_mmx_operand as operand 2 predicate.
10909 Call ix86_expand_vector_set_var for non-constant index operand.
10911 2021-07-06 Jeff Law <jeffreyalaw@gmail.com>
10913 * config/h8300/jumpcall.md (*branch): When possible, generate
10914 the comparison in CCZN mode.
10915 * config/h8300/predicates.md (simple_memory_operand): Reject all
10916 auto-increment addressing modes.
10918 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
10920 PR bootstrap/100246
10921 * config/i386/i386.h (struct stringop_algs): Define a CTOR for
10924 2021-07-06 Richard Biener <rguenther@suse.de>
10926 * doc/md.texi (vec_fmaddsub<mode>4): Document.
10927 (vec_fmsubadd<mode>4): Likewise.
10928 * optabs.def (vec_fmaddsub$a4): Add.
10929 (vec_fmsubadd$a4): Likewise.
10930 * internal-fn.def (IFN_VEC_FMADDSUB): Add.
10931 (IFN_VEC_FMSUBADD): Likewise.
10932 * tree-vect-slp-patterns.c (addsub_pattern::recognize):
10933 Refactor to handle IFN_VEC_FMADDSUB and IFN_VEC_FMSUBADD.
10934 (addsub_pattern::build): Likewise.
10935 * tree-vect-slp.c (vect_optimize_slp): CFN_VEC_FMADDSUB
10936 and CFN_VEC_FMSUBADD are not transparent for permutes.
10937 * config/i386/sse.md (vec_fmaddsub<mode>4): New expander.
10938 (vec_fmsubadd<mode>4): Likewise.
10940 2021-07-06 Richard Biener <rguenther@suse.de>
10942 * doc/invoke.texi (fmove-loop-stores): Document.
10943 * common.opt (fmove-loop-stores): New option.
10944 * opts.c (default_options_table): Enable -fmove-loop-stores
10945 at -O1 but not -Og.
10946 * tree-ssa-loop-im.c (pass_lim::execute): Pass
10947 flag_move_loop_stores instead of true to
10948 loop_invariant_motion_in_fun.
10950 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
10952 * doc/install.texi: Document --with-dsymutil.
10954 2021-07-06 Andrew Pinski <apinski@marvell.com>
10956 PR tree-optimization/101256
10957 * dbgcnt.def (phiopt_edge_range): New counter.
10958 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
10959 Check to make sure the new name is defined in the same
10960 bb as the conditional before duplicating range info.
10961 Also add debug counter.
10963 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
10965 PR rtl-optimization/100328
10966 * config/i386/i386-options.c (ix86_option_override_internal):
10967 Set param_ira_consider_dup_in_all_alts to 0.
10969 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
10971 PR rtl-optimization/100328
10972 * doc/invoke.texi (ira-consider-dup-in-all-alts): Document new
10974 * ira.c (ira_get_dup_out_num): Adjust as parameter
10975 param_ira_consider_dup_in_all_alts.
10976 * params.opt (ira-consider-dup-in-all-alts): New.
10977 * ira-conflicts.c (process_regs_for_copy): Add one parameter
10978 single_input_op_has_cstr_p.
10979 (get_freq_for_shuffle_copy): New function.
10980 (add_insn_allocno_copies): Adjust as single_input_op_has_cstr_p.
10981 * ira-int.h (ira_get_dup_out_num): Add one bool parameter.
10983 2021-07-05 Jeff Law <jeffreyalaw@gmail.com>
10985 * config/h8300/shiftrotate.md (shift-by-variable patterns): Update to
10986 generate condition code aware RTL directly.
10988 2021-07-05 Andrew Pinski <apinski@marvell.com>
10990 PR tree-optimization/101039
10991 * match.pd (A CMP 0 ? A : -A): New patterns.
10992 * tree-ssa-phiopt.c (abs_replacement): Delete function.
10993 (tree_ssa_phiopt_worker): Don't call abs_replacement.
10994 Update comment about abs_replacement.
10996 2021-07-05 Andrew Pinski <apinski@marvell.com>
10998 * tree-ssa-phiopt.c (gimple_simplify_phiopt):
10999 If "A ? B : C" fails to simplify, try "(!A) ? C : B".
11001 2021-07-05 Andrew Pinski <apinski@marvell.com>
11003 * tree-ssa-phiopt.c (match_simplify_replacement):
11004 Add early_p argument. Call gimple_simplify_phiopt
11005 instead of gimple_simplify.
11006 (tree_ssa_phiopt_worker): Update call to
11007 match_simplify_replacement and allow unconditionally.
11008 (phiopt_early_allow): New function.
11009 (gimple_simplify_phiopt): New function.
11011 2021-07-05 Andrew Pinski <apinski@marvell.com>
11013 PR middle-end/101237
11014 * fold-const.c (negate_expr_p): Remove call to element_mode
11015 and TREE_MODE/TREE_TYPE when calling HONOR_SIGNED_ZEROS,
11016 HONOR_SIGN_DEPENDENT_ROUNDING, and HONOR_SNANS.
11017 (fold_negate_expr_1): Likewise.
11018 (const_unop): Likewise.
11019 (fold_cond_expr_with_comparison): Likewise.
11020 (fold_binary_loc): Likewise.
11021 (fold_ternary_loc): Likewise.
11022 (tree_call_nonnegative_warnv_p): Likewise.
11023 * match.pd (-(A + B) -> (-B) - A): Likewise.
11025 2021-07-05 Iain Sandoe <iain@sandoe.co.uk>
11027 * configure.ac: Handle --with-dsymutil in the same way as we
11028 do for the assembler and linker. (DEFAULT_DSYMUTIL): New.
11029 Extract the type and version for the dsymutil configured or
11030 found by the default searches.
11031 * config.in: Regenerated.
11032 * configure: Regenerated.
11033 * collect2.c (do_dsymutil): Handle locating dsymutil in the
11034 same way as for the assembler and linker.
11035 * config/darwin.h (DSYMUTIL): Delete.
11036 * gcc.c: Report a configured dsymutil correctly.
11037 * exec-tool.in: Allow for dsymutil.
11039 2021-07-05 Uroš Bizjak <ubizjak@gmail.com>
11041 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
11042 Handle V4QI and V2HI modes.
11043 (expand_vec_perm_blend): Allow 4-byte vector modes with TARGET_SSE4_1.
11044 Handle V4QI mode. Emit mmx_pblendvb32 for 4-byte modes.
11045 (expand_vec_perm_pshufb): Rewrite to use switch statemets.
11046 Handle 4-byte dual operands with TARGET_XOP and single operands
11047 with TARGET_SSSE3. Emit mmx_ppermv32 for TARGET_XOP and
11048 mmx_pshufbv4qi3 for TARGET_SSSE3.
11049 (expand_vec_perm_pblendv): Allow 4-byte vector modes with TARGET_SSE4_1.
11050 (expand_vec_perm_interleave2): Allow 4-byte vector modes.
11051 (expand_vec_perm_pshufb2): Allow 4-byte vector modes with TARGET_SSSE3.
11052 (expand_vec_perm_even_odd_1): Handle V4QI mode.
11053 (expand_vec_perm_broadcast_1): Handle V4QI mode.
11054 (ix86_vectorize_vec_perm_const): Handle V4QI mode.
11055 * config/i386/mmx.md (mmx_ppermv32): New insn pattern.
11056 (mmx_pshufbv4qi3): Ditto.
11057 (*mmx_pblendw32): Ditto.
11058 (*mmx_pblendw64): Rename from *mmx_pblendw.
11059 (mmx_punpckhbw_low): New insn_and_split pattern.
11060 (mmx_punpcklbw_low): Ditto.
11062 2021-07-05 Richard Biener <rguenther@suse.de>
11064 * tree-vect-loop-manip.c (vect_loop_versioning): Do not
11065 set LOOP_C_INFINITE on the vectorized loop.
11067 2021-07-05 Richard Biener <rguenther@suse.de>
11069 PR middle-end/101291
11070 * cfgloopmanip.c (loop_version): Set the loop copy of the
11071 versioned loop to the new loop.
11073 2021-07-04 Iain Sandoe <iain@sandoe.co.uk>
11076 * config.gcc: Ensure that Darwin biarch definitions are
11077 added before i386.h.
11078 * config/i386/darwin.h (TARGET_64BIT): Remove.
11079 (PR80556_WORKAROUND): New.
11080 (REAL_LIBGCC_SPEC): Amend to use PR80556_WORKAROUND.
11081 (DARWIN_SUBARCH_SPEC): New.
11082 * config/i386/darwin32-biarch.h (TARGET_64BIT_DEFAULT,
11083 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
11084 (REAL_LIBGCC_SPEC): Remove.
11085 * config/i386/darwin64-biarch.h (TARGET_64BIT_DEFAULT,
11086 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
11087 (REAL_LIBGCC_SPEC): Remove.
11089 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
11091 PR middle-end/101294
11092 * expr.c (store_constructor): Don't use vec_duplicate on vector.
11094 2021-07-02 Martin Sebor <msebor@redhat.com>
11096 PR middle-end/98871
11097 PR middle-end/98512
11098 * diagnostic.c (get_any_inlining_info): New.
11099 (update_effective_level_from_pragmas): Handle inlining context.
11100 (diagnostic_enabled): Same.
11101 (diagnostic_report_diagnostic): Same.
11102 * diagnostic.h (struct diagnostic_info): Add ctor.
11103 (struct diagnostic_context): Add new member.
11104 * tree-diagnostic.c (set_inlining_locations): New.
11105 (tree_diagnostics_defaults): Set new callback pointer.
11107 2021-07-02 Peter Bergner <bergner@linux.ibm.com>
11109 * config/rs6000/rs6000-builtin.def (BU_MMA_PAIR_LD, BU_MMA_PAIR_ST):
11111 (__builtin_vsx_lxvp, __builtin_vsx_stxvp): New built-ins.
11112 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Expand
11113 lxvp and stxvp built-ins.
11114 (mma_init_builtins): Handle lxvp and stxvp built-ins.
11115 (builtin_function_type): Likewise.
11116 * doc/extend.texi (__builtin_vsx_lxvp, __builtin_mma_stxvp): Document.
11118 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
11120 * config/h8300/h8300-protos.h (compute_a_shift_cc): Accept
11121 additional argument for the code.
11122 * config/h8300/h8300.c (compute_a_shift_cc): Accept additional
11123 argument for the code. Just return if the ZN bits are useful or
11124 not rather than the old style CC_* enums.
11125 * config/h8300/shiftrotate.md (shiftqi_noscratch): Move before
11126 more generic shiftqi patterns.
11127 (shifthi_noscratch, shiftsi_noscratch): Similarly.
11128 (shiftqi_noscratch_set_flags): New pattern.
11129 (shifthi_noscratch_set_flags, shiftsi_noscratch_set_flags): Likewise.
11131 2021-07-02 Andrew MacLeod <amacleod@redhat.com>
11133 PR tree-optimization/101223
11134 * range-op.cc (build_lt): Add -1 for signed values.
11135 (built_gt): Subtract -1 for signed values.
11137 2021-07-02 David Faust <david.faust@oracle.com>
11139 * btfout.c (get_btf_kind): Support BTF_KIND_FLOAT.
11140 (btf_asm_type): Likewise.
11142 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
11144 * config/h8300/h8300-protos.h (output_a_shift): Make first argument
11145 an array of rtx rather than a pointer to rtx. Add code argument.
11146 (compute_a_shift_length): Similarly.
11147 * config/h8300/h8300.c (h8300_shift_costs): Adjust now that the
11148 shift itself isn't an operand. Create dummy operand[0] to carry
11149 a mode and pass a suitable rtx code to compute_a_shift_length.
11150 (get_shift_alg): Adjust operand number of clobber in output templates.
11151 (output_a_shift): Make first argument an array of rtx rather than
11152 a pointer to rtx. Add code argument for the type of shift.
11153 Adjust now that the shift itself is no longer an operand.
11154 (compute_a_shift_length): Similarly.
11155 * config/h8300/shiftrotate.md (shiftqi, shifthi, shiftsi): Use an
11156 iterator rather than nshift_operator.
11157 (shiftqi_noscratch, shifthi_noscratch, shiftsi_noscratch): Likewise.
11158 (shiftqi_clobber_flags): Adjust to API changes in output_a_shift
11159 and compute_a_shift_length.
11160 (shiftqi_noscratch_clobber_flags): Likewise.
11161 (shifthi_noscratch_clobber_flags): Likewise.
11162 (shiftsi_noscratch_clobber_flags): Likewise.
11164 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
11167 * config/darwin.h (DSYMUTIL_SPEC): Do not try to run
11168 dsymutil for BTF/CTF.
11170 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
11173 * config/darwin.h (CTF_INFO_SECTION_NAME): Update the
11174 segment to include BTF.
11175 (BTF_INFO_SECTION_NAME): New.
11177 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
11179 * config/m32r/m32r-protos.h (call_operand): Adjust return type.
11180 (small_data_operand, memreg_operand, small_insn_p): Likewise.
11181 * config/m32r/m32r.c (call_operand): Adjust return type.
11182 (small_data_operand, memreg_operand): Likewise.
11184 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
11186 * config/frv/frv-protos.h (integer_register_operand): Adjust return
11188 (frv_load_operand, gpr_or_fpr_operand, gpr_no_subreg_operand): Likewise.
11189 (fpr_or_int6_operand, gpr_or_int_operand); Likewise.
11190 (gpr_or_int12_operand, gpr_or_int10_operand); Likewise.
11191 (move_source_operand, move_destination_operand): Likewise.
11192 (condexec_source_operand, condexec_dest_operand): Likewise.
11193 (lr_operand, gpr_or_memory_operand, fpr_or_memory_operand): Likewise.
11194 (reg_or_0_operand, fcc_operand, icc_operand, cc_operand): Likewise.
11195 (fcr_operand, icr_operand, cr_operand, call_operand): Likewise.
11196 (fpr_operand, even_reg_operand, odd_reg_operand): Likewise.
11197 (even_gpr_operand, odd_gpr_operand, quad_fpr_operand): Likewise.
11198 (even_fpr_operand, odd_fpr_operand): Likewise.
11199 (dbl_memory_one_insn_operand, dbl_memory_two_insn_operand): Likewise.
11200 (int12_operand, int6_operand, int5_operand, uint5_operand): Likewise.
11201 (uint4_operand, uint1_operand, int_2word_operand): Likewise
11202 (upper_int16_operand, uint16_operand, symbolic_operand): Likewise.
11203 (relational_operator, float_relational_operator): Likewise.
11204 (ccr_eqne_operator, minmax_operator): Likewise.
11205 (condexec_si_binary_operator, condexec_si_media_operator): Likewise.
11206 (condexec_si_divide_operator, condexec_si_unary_operator): Likewise.
11207 (condexec_sf_conv_operator, condexec_sf_add_operator): Likewise.
11208 (intop_compare_operator, acc_operand, even_acc_operand): Likewise.
11209 (quad_acc_operand, accg_operand): Likewise.
11211 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
11213 * config/stormy16/stormy16-protos.h (xstormy16_below_100_symbol): Change
11214 return type to a bool.
11215 (nonimmediate_nonstack_operand): Likewise.
11216 (xstormy16_splittable_below100_operand): Likewise.
11217 * config/stormy16/stormy16.c (xstormy16_below_100_symbol): Fix
11219 (xstormy16_splittable_below100_operand): Likewise.
11221 2021-07-02 Richard Biener <rguenther@suse.de>
11223 PR tree-optimization/101293
11224 * tree-ssa-loop-im.c (mem_ref_hasher::equal): Compare MEM_REF bases
11225 with combined offsets.
11226 (gather_mem_refs_stmt): Hash MEM_REFs as if their offset were
11227 combined with the rest of the offset.
11229 2021-07-02 Eric Botcazou <ebotcazou@adacore.com>
11231 * config/i386/i386.c (asm_preferred_eh_data_format): Always use the
11232 PIC encodings for PE-COFF targets.
11234 2021-07-02 Jakub Jelinek <jakub@redhat.com>
11237 * config/i386/i386-expand.c (ix86_broadcast_from_integer_constant):
11238 Return nullptr for TImode inner mode.
11240 2021-07-02 Richard Biener <rguenther@suse.de>
11242 PR tree-optimization/101280
11243 PR tree-optimization/101173
11244 * gimple-loop-interchange.cc
11245 (tree_loop_interchange::valid_data_dependences): Properly
11246 guard all dependence checks with DDR_REVERSED_P or its
11249 2021-07-02 Hongyu Wang <hongyu.wang@intel.com>
11251 * config/i386/i386-expand.c (ix86_expand_builtin):
11252 Add branch to clear odata when ZF is set for asedecenc_expand
11253 and wideaesdecenc_expand.
11255 2021-07-02 Eugene Rozenfeld <erozen@microsoft.com>
11257 * config/i386/gcc-auto-profile: regenerate
11259 2021-07-02 liuhongt <hongtao.liu@intel.com>
11261 * config/i386/sse.md (trunc<mode><pmov_dst_4>2): Refined to ..
11262 (trunc<mode><pmov_dst_4_lower>2): this.
11264 2021-07-01 David Malcolm <dmalcolm@redhat.com>
11266 * diagnostic.h (diagnostic_context::m_file_cache): New field.
11267 * input.c (class fcache): Rename to...
11268 (class file_cache_slot): ...this, making most members private and
11269 prefixing fields with "m_".
11270 (file_cache_slot::get_file_path): New accessor.
11271 (file_cache_slot::get_use_count): New accessor.
11272 (file_cache_slot::missing_trailing_newline_p): New accessor.
11273 (file_cache_slot::inc_use_count): New.
11274 (fcache_buffer_size): Move to...
11275 (file_cache_slot::buffer_size): ...here.
11276 (fcache_line_record_size): Move to...
11277 (file_cache_slot::line_record_size): ...here.
11278 (fcache_tab): Delete, in favor of global_dc->m_file_cache.
11279 (fcache_tab_size): Move to file_cache::num_file_slots.
11280 (diagnostic_file_cache_init): Update for move of fcache_tab
11281 to global_dc->m_file_cache.
11282 (diagnostic_file_cache_fini): Likewise.
11283 (lookup_file_in_cache_tab): Convert to...
11284 (file_cache::lookup_file): ...this.
11285 (diagnostics_file_cache_forcibly_evict_file): Update for move of
11286 fcache_tab to global_dc->m_file_cache, moving most of
11287 implementation to...
11288 (file_cache::forcibly_evict_file): ...this new function and...
11289 (file_cache_slot::evict): ...this new function.
11290 (evicted_cache_tab_entry): Convert to...
11291 (file_cache::evicted_cache_tab_entry): ...this.
11292 (add_file_to_cache_tab): Convert to...
11293 (file_cache::add_file): ...this, moving bulk of implementation
11295 (file_cache_slot::create): ..this new function.
11296 (file_cache::file_cache): New.
11297 (file_cache::~file_cache): New.
11298 (lookup_or_add_file_to_cache_tab): Convert to...
11299 (file_cache::lookup_or_add_file): ..this new function.
11300 (fcache::fcache): Rename to...
11301 (file_cache_slot::file_cache_slot): ...this, adding "m_" prefixes
11303 (fcache::~fcache): Rename to...
11304 (file_cache_slot::~file_cache_slot): ...this, adding "m_" prefixes
11306 (needs_read): Convert to...
11307 (file_cache_slot::needs_read_p): ...this.
11308 (needs_grow): Convert to...
11309 (file_cache_slot::needs_grow_p): ...this.
11310 (maybe_grow): Convert to...
11311 (file_cache_slot::maybe_grow): ...this.
11312 (read_data): Convert to...
11313 (file_cache_slot::read_data): ...this.
11314 (maybe_read_data): Convert to...
11315 (file_cache_slot::maybe_read_data): ...this.
11316 (get_next_line): Convert to...
11317 (file_cache_slot::get_next_line): ...this.
11318 (goto_next_line): Convert to...
11319 (file_cache_slot::goto_next_line): ...this.
11320 (read_line_num): Convert to...
11321 (file_cache_slot::read_line_num): ...this.
11322 (location_get_source_line): Update for moving of globals to
11323 global_dc->m_file_cache.
11324 (location_missing_trailing_newline): Likewise.
11325 * input.h (class file_cache_slot): New forward decl.
11326 (class file_cache): New.
11328 2021-07-01 Michael Meissner <meissner@linux.ibm.com>
11330 * config/rs6000/rs6000.c (rs6000_maybe_emit_fp_cmove): Add IEEE
11331 128-bit floating point conditional move support.
11332 (have_compare_and_set_mask): Add IEEE 128-bit floating point
11334 * config/rs6000/rs6000.md (mov<mode>cc, IEEE128 iterator): New insn.
11335 (mov<mode>cc_p10, IEEE128 iterator): New insn.
11336 (mov<mode>cc_invert_p10, IEEE128 iterator): New insn.
11337 (fpmask<mode>, IEEE128 iterator): New insn.
11338 (xxsel<mode>, IEEE128 iterator): New insn.
11340 2021-07-01 Iain Sandoe <iain@sandoe.co.uk>
11343 * config/darwin.h (CTF_INFO_SECTION_NAME): New.
11345 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
11347 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
11349 * config/i386/i386-protos.h (ix86_expand_vector_init_duplicate):
11351 * config/i386/sse.md (INT_BROADCAST_MODE): New mode iterator.
11352 (vec_duplicate<mode>): New expander.
11354 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
11357 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
11359 (ix86_byte_broadcast): New function.
11360 (ix86_convert_const_wide_int_to_broadcast): Likewise.
11361 (ix86_expand_move): Convert CONST_WIDE_INT to broadcast if mode
11362 size is 16 bytes or bigger.
11363 (ix86_broadcast_from_integer_constant): New function.
11364 (ix86_expand_vector_move): Convert CONST_WIDE_INT and CONST_VECTOR
11365 to broadcast if mode size is 16 bytes or bigger.
11366 * config/i386/i386-protos.h (ix86_gen_scratch_sse_rtx): New
11368 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): New function.
11370 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
11372 * config/i386/predicates.md (ix86_endbr_immediate_operand):
11373 Return true/false instead of 1/0.
11374 (movq_parallel): Ditto.
11376 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
11378 * recog.c (general_operand): Return true/false instead of 1/0.
11379 (register_operand): Ditto.
11380 (immediate_operand): Ditto.
11381 (const_int_operand): Ditto.
11382 (const_scalar_int_operand): Ditto.
11383 (const_double_operand): Ditto.
11384 (push_operand): Ditto.
11385 (pop_operand): Ditto.
11386 (memory_operand): Ditto.
11387 (indirect_operand): Ditto.
11389 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
11391 * genpreds.c (write_predicate_subfunction):
11392 Change the type of written subfunction to bool.
11393 (write_one_predicate_function):
11394 Change the type of written function to bool.
11395 (write_tm_preds_h): Ditto.
11396 * recog.h (*insn_operand_predicate_fn): Change the type to bool.
11397 * recog.c (general_operand): Change the type to bool.
11398 (address_operand): Ditto.
11399 (register_operand): Ditto.
11400 (pmode_register_operand): Ditto.
11401 (scratch_operand): Ditto.
11402 (immediate_operand): Ditto.
11403 (const_int_operand): Ditto.
11404 (const_scalar_int_operand): Ditto.
11405 (const_double_operand): Ditto.
11406 (nonimmediate_operand): Ditto.
11407 (nonmemory_operand): Ditto.
11408 (push_operand): Ditto.
11409 (pop_operand): Ditto.
11410 (memory_operand): Ditto.
11411 (indirect_operand): Ditto.
11412 (ordered_comparison_operator): Ditto.
11413 (comparison_operator): Ditto.
11414 * config/i386/i386-expand.c (ix86_expand_sse_cmp):
11415 Change the type of indirect predicate function to bool.
11416 * config/rs6000/rs6000.c (easy_vector_constant):
11417 Change the type to bool.
11418 * config/mips/mips-protos.h (m16_based_address_p):
11419 Change the type of operand 3 to bool.
11421 2021-07-01 Richard Biener <rguenther@suse.de>
11423 PR tree-optimization/101280
11424 PR tree-optimization/101173
11425 * gimple-loop-interchange.cc
11426 (tree_loop_interchange::valid_data_dependences): Revert
11427 previous change and instead correctly handle DDR_REVERSED_P
11430 2021-07-01 Richard Biener <rguenther@suse.de>
11432 PR tree-optimization/101278
11433 * tree-ssa-dse.c (dse_classify_store): First check for
11434 uses, then ignore stmt for chaining purposes.
11436 2021-07-01 Richard Biener <rguenther@suse.de>
11438 PR tree-optimization/100778
11439 * tree-vect-slp.c (vect_schedule_slp_node): Do not place trapping
11440 vectorized ops ahead of their scalar BB.
11442 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
11445 * config/i386/i386.md (*nabs<dwi>2_doubleword):
11446 New insn_and_split pattern.
11447 (*nabs<dwi>2_1): Ditto.
11448 * config/i386/i386-features.c
11449 (general_scalar_chain::compute_convert_gain):
11450 Handle (NEG (ABS (...))) RTX. Rewrite src code
11451 scanner as switch statement.
11452 (general_scalar_chain::convert_insn):
11453 Handle (NEG (ABS (...))) RTX.
11454 (general_scalar_to_vector_candidate_p):
11455 Detect (NEG (ABS (...))) RTX. Reorder case statements
11456 for (AND (NOT (...) ...)) fallthrough.
11458 2021-07-01 Richard Biener <rguenther@suse.de>
11460 PR tree-optimization/101178
11461 * tree-vect-slp.c (slpg_vertex::materialize): Remove.
11462 (slpg::perm_in): Add.
11463 (slpg::get_perm_in): Remove.
11464 (slpg::get_perm_materialized): Add.
11465 (vect_optimize_slp): Handle VEC_PERM nodes more optimally
11466 during permute propagation and materialization.
11468 2021-07-01 Jakub Jelinek <jakub@redhat.com>
11471 * dwarf2out.c (loc_list_from_tree_1): Handle COMPOUND_LITERAL_EXPR.
11473 2021-07-01 Jakub Jelinek <jakub@redhat.com>
11475 PR middle-end/94366
11476 * omp-low.c (lower_rec_input_clauses): Rename is_fp_and_or to
11477 is_truth_op, set it for TRUTH_*IF_EXPR regardless of new_var's type,
11478 use boolean_type_node instead of integer_type_node as NE_EXPR type.
11479 (lower_reduction_clauses): Likewise.
11481 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
11483 * config/gcn/gcn.c: Include dwarf2.h.
11484 (gcn_addr_space_debug): New function.
11485 (TARGET_ADDR_SPACE_DEBUG): New hook.
11487 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
11489 * common/config/gcn/gcn-common.c
11490 (gcn_option_optimization_table): Change OPT_fomit_frame_pointer to -O3.
11491 * config/gcn/gcn.c (gcn_expand_prologue): Prefer the frame pointer
11493 (gcn_expand_prologue): Prefer the frame pointer when emitting CFI.
11494 (gcn_frame_pointer_rqd): New function.
11495 (TARGET_FRAME_POINTER_REQUIRED): New hook.
11497 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
11499 * config/gcn/gcn.c (move_callee_saved_registers): Emit CFI notes for
11500 prologue register saves.
11501 (gcn_debug_unwind_info): Use UI_DWARF2.
11502 (gcn_dwarf_register_number): Map DWARF_LINK_REGISTER to DWARF PC.
11503 (gcn_dwarf_register_span): DWARF_LINK_REGISTER doesn't span.
11504 * config/gcn/gcn.h: (DWARF_FRAME_RETURN_COLUMN): New define.
11505 (DWARF_LINK_REGISTER): New define.
11506 (FIRST_PSEUDO_REGISTER): Increment.
11507 (FIXED_REGISTERS): Add entry for DWARF_LINK_REGISTER.
11508 (CALL_USED_REGISTERS): Likewise.
11509 (REGISTER_NAMES): Likewise.
11511 2021-06-30 Richard Biener <rguenther@suse.de>
11513 PR tree-optimization/101267
11514 * tree-vect-stmts.c (vect_check_scalar_mask): Adjust
11515 API and use SLP compatible interface of vect_is_simple_use.
11516 Reject not vectorized SLP defs for callers that do not support
11518 (vect_check_store_rhs): Handle masked stores and pass down
11519 the appropriate operator index.
11520 (vectorizable_call): Adjust.
11521 (vectorizable_store): Likewise.
11522 (vectorizable_load): Likewise. Handle SLP pecularity of
11524 (vect_is_simple_use): Remove special-casing of masked stores.
11526 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
11528 * common.opt (foffload): Remove help as Driver only.
11529 * gcc.c (display_help): Add -foffload.
11531 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
11533 * gcc.c (close_at_file, execute): Replace alloca by XALLOCAVEC.
11534 (check_offload_target_name): Fix splitting OFFLOAD_TARGETS into
11535 a candidate list; better inform no offload target is configured
11536 and fix hint extraction when passed target is not '\0' at [len].
11537 * common.opt (foffload): Add tailing '.'.
11538 (foffload-options): Likewise; fix flag name in the help string.
11540 2021-06-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
11543 * config/arm/arm_neon.h: Move vabs intrinsics before vcage_f32.
11544 (vcage_f32): Gate comparison on __FAST_MATH__.
11545 (vcageq_f32): Likewise.
11546 (vcale_f32): Likewise.
11547 (vcaleq_f32): Likewise.
11548 (vcagt_f32): Likewise.
11549 (vcagtq_f32): Likewise.
11550 (vcalt_f32): Likewise.
11551 (vcaltq_f32): Likewise.
11552 (vcage_f16): Likewise.
11553 (vcageq_f16): Likewise.
11554 (vcale_f16): Likewise.
11555 (vcaleq_f16): Likewise.
11556 (vcagt_f16): Likewise.
11557 (vcagtq_f16): Likewise.
11558 (vcalt_f16): Likewise.
11559 (vcaltq_f16): Likewise.
11561 2021-06-30 Richard Biener <rguenther@suse.de>
11563 PR tree-optimization/101264
11564 * tree-vect-slp.c (vect_optimize_slp): Propagate the
11565 computed perm_in to all "any" permute successors
11566 we cannot de-duplicate immediately.
11568 2021-06-30 liuhongt <hongtao.liu@intel.com>
11571 * config/i386/sse.md
11572 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>):
11574 (avx512f_sfixupimm<mode><maskz_scalar_name><round_saeonly_name>):
11576 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>"): Refined.
11577 * config/i386/subst.md (maskz_scalar): New define_subst.
11578 (maskz_scalar_name): New subst_attr.
11579 (maskz_scalar_op5): Ditto.
11580 (round_saeonly_maskz_scalar_op5): Ditto.
11581 (round_saeonly_maskz_scalar_operand5): Ditto.
11583 2021-06-30 David Edelsohn <dje.gcc@gmail.com>
11585 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
11586 Increase code CSECT alignment to at least 32 bytes.
11587 * config/rs6000/xcoff.h (TEXT_SECTION_ASM_OP): Add 32 byte
11588 alignment designation.
11590 2021-06-29 Sergei Trofimovich <siarheit@google.com>
11592 * doc/generic.texi: Fix s/net yet/not yet/ typo.
11594 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
11596 PR tree-optimization/101254
11597 * range-op.cc (operator_minus::op1_op2_relation_effect): Check for
11598 wrapping/non-wrapping when setting the result range.
11600 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
11602 * value-query.cc (gimple_range_global): Allow phis.
11604 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
11606 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): Use stmt.
11607 (simplify_using_ranges::op_with_boolean_value_range_p): Add a
11608 statement for location context.
11609 (check_for_binary_op_overflow): Ditto.
11610 (simplify_using_ranges::get_vr_for_comparison): Ditto.
11611 (simplify_using_ranges::compare_name_with_value): Ditto.
11612 (simplify_using_ranges::compare_names): Ditto.
11613 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Ditto.
11614 (simplify_using_ranges::simplify_truth_ops_using_ranges): Ditto.
11615 (simplify_using_ranges::simplify_min_or_max_using_ranges): Ditto.
11616 (simplify_using_ranges::simplify_internal_call_using_ranges): Ditto.
11617 (simplify_using_ranges::two_valued_val_range_p): Ditto.
11618 (simplify_using_ranges::simplify): Ditto.
11619 * vr-values.h: Adjust prototypes.
11621 2021-06-29 Uroš Bizjak <ubizjak@gmail.com>
11624 * config/i386/mmx.md (vec_addsubv2sf3): New insn pattern.
11626 2021-06-29 Julian Brown <julian@codesourcery.com>
11628 * config/gcn/gcn.c (gcn_init_libfuncs): New function.
11629 (TARGET_INIT_LIBFUNCS): Define target hook using above function.
11630 * config/gcn/gcn.h (UNITS_PER_WORD): Define to 8 for IN_LIBGCC2, 4
11632 (LIBGCC2_UNITS_PER_WORD, BITS_PER_WORD): Remove definitions.
11633 (MAX_FIXED_MODE_SIZE): Change to 128.
11635 2021-06-29 Julian Brown <julian@codesourcery.com>
11637 * config/gcn/gcn.md (UNSPEC_FLBIT_INT): New unspec constant.
11638 (s_mnemonic): Add clrsb.
11639 (gcn_flbit<mode>_int): Add insn pattern for SImode/DImode.
11640 (clrsb<mode>2): Add expander for SImode/DImode.
11642 2021-06-29 Julian Brown <julian@codesourcery.com>
11644 * config/gcn/gcn.md (<su>mulsidi3, <su>mulsidi3_reg, <su>mulsidi3_imm,
11645 muldi3): Add patterns.
11647 2021-06-29 Julian Brown <julian@codesourcery.com>
11649 * config/gcn/gcn.md (<su>mulsi3_highpart): Change to expander.
11650 (<su>mulsi3_highpart_reg, <su>mulsi3_highpart_imm): New patterns.
11652 2021-06-29 Julian Brown <julian@codesourcery.com>
11654 * config/gcn/gcn.md (mulsi3): Make s_mulk_i32 variant clobber SCC.
11656 2021-06-29 Joseph Myers <joseph@codesourcery.com>
11658 * btfout.c, ctfout.c: Include "memmodel.h".
11660 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
11662 * gcc.c (check_offload_target_name): Cast len argument to
11663 %q.*s to 'int'; avoid -Wstringop-truncation warning.
11665 2021-06-29 Richard Biener <rguenther@suse.de>
11667 * tree-vect-slp.c (vect_optimize_slp): Forward propagate
11668 to "any" permute nodes and relax "any" permute proapgation
11669 during iterative backward propagation.
11671 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
11674 * common.opt (-foffload=): Update description.
11675 (-foffload-options=): New.
11676 * doc/invoke.texi (C Language Options): Document
11677 -foffload and -foffload-options.
11678 * gcc.c (check_offload_target_name): New, split off from
11679 handle_foffload_option.
11680 (check_foffload_target_names): New.
11681 (handle_foffload_option): Handle -foffload=default.
11682 (driver_handle_option): Update for -foffload-options.
11683 * lto-opts.c (lto_write_options): Use -foffload-options
11684 instead of -foffload.
11685 * lto-wrapper.c (merge_and_complain, append_offload_options):
11687 * opts.c (common_handle_option): Likewise.
11689 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
11691 * doc/invoke.texi (C Language Options): Sort options
11692 alphabetically in optlist and also the description itself.
11693 Remove leftover -fallow-single-precision from and add missing
11694 -fgnu-tm to the optlist.
11696 2021-06-29 Richard Biener <rguenther@suse.de>
11698 * tree-vect-slp.c (slpg_vertex::visited): Remove.
11699 (vect_slp_perms_eq): Handle -1 permutes.
11700 (vect_optimize_slp): Rewrite permute propagation.
11702 2021-06-29 Jakub Jelinek <jakub@redhat.com>
11705 * match.pd ((intptr_t)x eq/ne CST to x eq/ne (typeof x) CST): Don't
11706 perform the optimization in GENERIC when sanitizing and x has a
11709 2021-06-29 Richard Biener <rguenther@suse.de>
11711 PR tree-optimization/101242
11712 * tree-vect-slp.c (vect_slp_build_vertices): Force-add
11713 PHIs with not represented initial values as leafs.
11715 2021-06-29 Jan-Benedict Glaw <jbglaw@getslash.de>
11717 * config/pdp11/pdp11.h (ASM_OUTPUT_SKIP): Fix signedness warning.
11718 * config/pdp11/pdp11.c (pdp11_asm_print_operand_punct_valid_p): Remove
11719 "register" keyword.
11720 (pdp11_initial_elimination_offset) Remove unused variable.
11721 (pdp11_cmp_length) Ditto.
11722 (pdp11_insn_cost): Ditto, and fix signedness warning.
11724 2021-06-29 David Edelsohn <dje.gcc@gmail.com>
11726 * btfout.c: Include tm_p.h.
11729 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
11731 * config/bpf/bpf.c (bpf_expand_prologue): Do not mark insns as
11733 (bpf_expand_epilogue): Likewise.
11734 * config/bpf/bpf.h (DWARF2_FRAME_INFO): Define to 0.
11735 Do not define DBX_DEBUGGING_INFO.
11737 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
11739 * doc/invoke.texi: Document the CTF and BTF debug info options.
11741 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
11742 David Faust <david.faust@oracle.com>
11743 Jose E. Marchesi <jose.marchesi@oracle.com>
11744 Weimin Pan <weimin.pan@oracle.com>
11746 * Makefile.in: Add ctfc.*, ctfout.c and btfout.c files to
11747 GTFILES. Add new object files.
11748 * common.opt: Add CTF and BTF debug info options.
11749 * btfout.c: New file.
11750 * ctfc.c: Likewise.
11751 * ctfc.h: Likewise.
11752 * ctfout.c: Likewise.
11753 * dwarf2ctf.c: Likewise.
11754 * dwarf2ctf.h: Likewise.
11755 * dwarf2cfi.c (dwarf2out_do_frame): Acknowledge CTF_DEBUG and
11757 * dwarf2out.c (dwarf2out_source_line): Likewise.
11758 (dwarf2out_finish): Skip emitting DWARF if CTF or BTF are to
11760 (debug_format_do_cu): New function.
11761 (dwarf2out_early_finish): Traverse DIEs and emit CTF/BTF for
11763 Include dwarf2ctf.c.
11764 * final.c (dwarf2_debug_info_emitted_p): Acknowledge DWARF-based debug
11766 * flag-types.h (enum debug_info_type): Add CTF_DEBUG and BTF_DEBUG.
11767 (CTF_DEBUG): New bitmask.
11768 (BTF_DEBUG): Likewise.
11769 (enum ctf_debug_info_levels): New enum.
11770 * gengtype.c (open_base_files): Handle ctfc.h.
11771 (main): Handle uint32_t type.
11772 * flags.h (btf_debuginfo_p): New definition.
11773 (dwarf_based_debuginfo_p): Likewise.
11774 * opts.c (debug_type_names): Add entries for CTF and BTF.
11775 (btf_debuginfo_p): New function.
11776 (dwarf_based_debuginfo_p): Likewise.
11777 (common_handle_option): Handle -gctfN and -gbtf options.
11778 (set_debug_level): Set CTF_DEBUG, BTF_DEBUG whenever appropriate.
11779 * toplev.c (process_options): Inform the user and ignore -gctfLEVEL if
11782 2021-06-28 Jose E. Marchesi <jose.marchesi@oracle.com>
11784 * dwarf2out.c (AT_class): Function is no longer static.
11785 (AT_int): Likewise.
11786 (AT_unsigned): Likewise.
11787 (AT_loc): Likewise.
11788 (get_AT): Likewise.
11789 (get_AT_string): Likewise.
11790 (get_AT_flag): Likewise.
11791 (get_AT_unsigned): Likewise.
11792 (get_AT_ref): Likewise.
11793 (new_die_raw): Likewise.
11794 (lookup_decl_die): Likewise.
11795 (base_type_die): Likewise.
11796 (add_name_attribute): Likewise.
11797 (add_AT_int): Likewise.
11798 (add_AT_unsigned): Likewise.
11799 (add_AT_loc): Likewise.
11800 (dw_get_die_tag): New function.
11801 (dw_get_die_child): Likewise.
11802 (dw_get_die_sib): Likewise.
11803 (struct dwarf_file_data): Move from here to dwarf2out.h
11804 (struct dw_attr_struct): Likewise.
11805 * dwarf2out.h: Analogous changes.
11807 2021-06-28 Martin Jambor <mjambor@suse.cz>
11810 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
11811 members m_dead_stmts and m_dead_ssas.
11812 * ipa-param-manipulation.c
11813 (ipa_param_body_adjustments::mark_dead_statements): New function.
11814 (ipa_param_body_adjustments::common_initialization): Call it on
11815 all removed but not split parameters.
11816 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
11818 (ipa_param_body_adjustments::modify_call_stmt): Remove arguments that
11820 * tree-inline.c (remap_gimple_stmt): Do not copy dead statements, reset
11821 dead debug statements.
11822 (copy_phis_for_bb): Do not copy dead PHI nodes.
11824 2021-06-28 Martin Jambor <mjambor@suse.cz>
11827 * symtab-clones.h (clone_info): Removed member param_adjustments.
11828 * ipa-param-manipulation.h: Adjust initial comment to reflect how we
11829 deal with pass-through splits now.
11830 (ipa_param_performed_split): Removed.
11831 (ipa_param_adjustments::modify_call): Adjusted parameters.
11832 (class ipa_param_body_adjustments): Adjusted parameters of
11833 register_replacement, modify_gimple_stmt and modify_call_stmt.
11834 (ipa_verify_edge_has_no_modifications): Declare.
11835 (ipa_edge_modifications_finalize): Declare.
11836 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Remove
11837 performed_splits processing, pas only edge to padjs->modify_call,
11838 check that call arguments were not modified if they should not have
11840 * cgraphclones.c (cgraph_node::create_clone): Do not copy performed
11842 * ipa-param-manipulation.c (struct pass_through_split_map): New type.
11843 (ipa_edge_modification_info): Likewise.
11844 (ipa_edge_modification_sum): Likewise.
11845 (ipa_edge_modifications): New edge summary.
11846 (ipa_verify_edge_has_no_modifications): New function.
11847 (transitive_split_p): Removed.
11848 (transitive_split_map): Likewise.
11849 (init_transitive_splits): Likewise.
11850 (ipa_param_adjustments::modify_call): Adjusted to use the new edge
11851 summary instead of performed_splits.
11852 (ipa_param_body_adjustments::register_replacement): Drop dummy
11853 parameter, set base_index of the created ipa_param_body_replacement.
11854 (phi_arg_will_live_p): New function.
11855 (ipa_param_body_adjustments::common_initialization): Do not create
11856 IPA_SRA dummy decls.
11857 (simple_tree_swap_info): Removed.
11858 (remap_split_decl_to_dummy): Likewise.
11859 (record_argument_state_1): New function.
11860 (record_argument_state): Likewise.
11861 (ipa_param_body_adjustments::modify_call_stmt): New parameter
11862 orig_stmt. Do not work with dummy decls, save necessary info about
11863 changes to ipa_edge_modifications.
11864 (ipa_param_body_adjustments::modify_gimple_stmt): New parameter
11865 orig_stmt, pass it to modify_call_stmt.
11866 (ipa_param_body_adjustments::modify_cfun_body): Adjust call to
11867 modify_gimple_stmt.
11868 (ipa_edge_modifications_finalize): New function.
11869 * tree-inline.c (remap_gimple_stmt): Pass original statement to
11870 modify_gimple_stmt.
11871 (copy_phis_for_bb): Do not copy dead PHI nodes.
11872 (expand_call_inline): Do not remap performed_splits.
11873 (update_clone_info): Likewise.
11874 * toplev.c: Include ipa-param-manipulation.h.
11875 (toplev::finalize): Call ipa_edge_modifications_finalize.
11877 2021-06-28 Andrew Pinski <apinski@marvell.com>
11879 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Duplicate range
11880 info if we're the only things setting the target PHI.
11881 (value_replacement): Don't duplicate range here.
11882 (minmax_replacement): Likewise.
11884 2021-06-28 Richard Biener <rguenther@suse.de>
11886 PR tree-optimization/101229
11887 * gimple-walk.c (gimple_walk_op): Handle PHIs.
11889 2021-06-28 Martin Liska <mliska@suse.cz>
11891 * config/v850/v850.c (construct_dispose_instruction): Allocate
11893 (construct_prepare_instruction): Likewise.
11895 2021-06-28 Martin Liska <mliska@suse.cz>
11897 * config/v850/v850.c (v850_option_override): Build default
11899 (v850_can_inline_p): New. Allow MASK_PROLOG_FUNCTION to be
11900 ignored for inlining.
11901 (TARGET_CAN_INLINE_P): New.
11903 2021-06-28 Richard Biener <rguenther@suse.de>
11905 PR tree-optimization/101207
11906 * tree-vect-slp.c (vect_optimize_slp): Do BB reduction
11907 permute eliding for load permutations properly.
11909 2021-06-28 Richard Biener <rguenther@suse.de>
11911 PR tree-optimization/101173
11912 * gimple-loop-interchange.cc
11913 (tree_loop_interchange::valid_data_dependences): Disallow outer
11914 loop dependence distance of zero.
11916 2021-06-28 liuhongt <hongtao.liu@intel.com>
11919 * config/i386/sse.md (*avx_cmp<mode>3_lt): New
11920 define_insn_and_split.
11921 (*avx_cmp<mode>3_ltint): Ditto.
11922 (*avx2_pcmp<mode>3_3): Ditto.
11923 (*avx2_pcmp<mode>3_4): Ditto.
11924 (*avx2_pcmp<mode>3_5): Ditto.
11926 2021-06-28 liuhongt <hongtao.liu@intel.com>
11928 * config/i386/i386-builtin.def (IX86_BUILTIN_BLENDVPD256,
11929 IX86_BUILTIN_BLENDVPS256, IX86_BUILTIN_PBLENDVB256,
11930 IX86_BUILTIN_BLENDVPD, IX86_BUILTIN_BLENDVPS,
11931 IX86_BUILTIN_PBLENDVB128): Replace icode with
11933 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold blendv
11935 * config/i386/sse.md (*<sse4_1_avx2>_pblendvb_lt_subreg_not):
11936 New pre_reload splitter.
11938 2021-06-27 Andrew Pinski <apinski@marvell.com>
11940 PR middle-end/101230
11941 * fold-const.c (fold_ternary_loc): Check
11942 the return value of invert_tree_comparison.
11944 2021-06-27 David Edelsohn <dje.gcc@gmail.com>
11946 * config.gcc: Add SPDX License Identifier.
11947 (powerpc-ibm-aix789): Default to aix73.h.
11948 (powerpc-ibm-aix7.2.*.*): New stanza.
11949 * config/rs6000/aix72.h: Add SPDX License Identifier.
11950 * config/rs6000/aix73.h: New file.
11952 2021-06-26 Jason Merrill <jason@redhat.com>
11954 * except.c: #include "dwarf2.h" instead of "dwarf2out.h".
11956 2021-06-26 Andrew Pinski <apinski@marvell.com>
11958 * genmatch.c (lower_cond): Copy for_subst_vec
11959 for the simplify also.
11960 (lower): Swap the order for lower_for and lower_cond.
11962 2021-06-26 Andrew Pinski <apinski@marvell.com>
11964 * tree-ssa-phiopt.c (match_simplify_replacement): Reset
11965 flow senatitive info on the moved ssa set.
11967 2021-06-26 Andrew Pinski <apinski@marvell.com>
11969 * fold-const.c (fold_cond_expr_with_comparison):
11970 Exand arg0 into comp_code, arg00, and arg01.
11971 (fold_ternary_loc): Use invert_tree_comparison
11972 instead of fold_invert_truthvalue for the case
11973 where we have A CMP B ? C : A.
11975 2021-06-25 Martin Sebor <msebor@redhat.com>
11977 PR middle-end/101216
11978 * calls.c (maybe_warn_rdwr_sizes): Use the no_warning constant.
11980 2021-06-25 Jeff Law <jeffreyalaw@gmail.com>
11982 * config/h8300/h8300.c (select_cc_mode): Handle ASHIFTRT and LSHIFTRT.
11984 2021-06-25 Richard Biener <rguenther@suse.de>
11986 PR tree-optimization/101202
11987 * tree-vect-slp.c (vect_optimize_slp): Explicitely handle
11990 2021-06-25 Richard Biener <rguenther@suse.de>
11992 * tree-vect-slp-patterns.c (addsub_pattern::build): Copy
11993 STMT_VINFO_REDUC_DEF from the original representative.
11995 2021-06-25 Martin Sebor <msebor@redhat.com>
11997 * builtins.c (warn_string_no_nul): Replace uses of TREE_NO_WARNING,
11998 gimple_no_warning_p and gimple_set_no_warning with
11999 warning_suppressed_p, and suppress_warning.
12001 (maybe_warn_for_bound): Same.
12002 (warn_for_access): Same.
12003 (check_access): Same.
12004 (expand_builtin_strncmp): Same.
12005 (fold_builtin_varargs): Same.
12006 * calls.c (maybe_warn_nonstring_arg): Same.
12007 (maybe_warn_rdwr_sizes): Same.
12008 * cfgexpand.c (expand_call_stmt): Same.
12009 * cgraphunit.c (check_global_declaration): Same.
12010 * fold-const.c (fold_undefer_overflow_warnings): Same.
12011 (fold_truth_not_expr): Same.
12012 (fold_unary_loc): Same.
12013 (fold_checksum_tree): Same.
12014 * gimple-array-bounds.cc (array_bounds_checker::check_array_ref): Same.
12015 (array_bounds_checker::check_mem_ref): Same.
12016 (array_bounds_checker::check_addr_expr): Same.
12017 (array_bounds_checker::check_array_bounds): Same.
12018 * gimple-expr.c (copy_var_decl): Same.
12019 * gimple-fold.c (gimple_fold_builtin_strcpy): Same.
12020 (gimple_fold_builtin_strncat): Same.
12021 (gimple_fold_builtin_stxcpy_chk): Same.
12022 (gimple_fold_builtin_stpcpy): Same.
12023 (gimple_fold_builtin_sprintf): Same.
12024 (fold_stmt_1): Same.
12025 * gimple-ssa-isolate-paths.c (diag_returned_locals): Same.
12026 * gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Same.
12027 * gimple-ssa-sprintf.c (handle_printf_call): Same.
12028 * gimple-ssa-store-merging.c (imm_store_chain_info::output_merged_store): Same.
12029 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
12030 * gimple-ssa-warn-restrict.h: Adjust declarations.
12031 (maybe_diag_access_bounds): Replace uses of TREE_NO_WARNING,
12032 gimple_no_warning_p and gimple_set_no_warning with
12033 warning_suppressed_p, and suppress_warning.
12034 (check_call): Same.
12035 (check_bounds_or_overlap): Same.
12036 * gimple.c (gimple_build_call_from_tree): Same.
12037 * gimplify.c (gimplify_return_expr): Same.
12038 (gimplify_cond_expr): Same.
12039 (gimplify_modify_expr_complex_part): Same.
12040 (gimplify_modify_expr): Same.
12041 (gimple_push_cleanup): Same.
12042 (gimplify_expr): Same.
12043 * omp-expand.c (expand_omp_for_generic): Same.
12044 (expand_omp_taskloop_for_outer): Same.
12045 * omp-low.c (lower_rec_input_clauses): Same.
12046 (lower_lastprivate_clauses): Same.
12047 (lower_send_clauses): Same.
12048 (lower_omp_target): Same.
12049 * tree-cfg.c (pass_warn_function_return::execute): Same.
12050 * tree-complex.c (create_one_component_var): Same.
12051 * tree-inline.c (remap_gimple_op_r): Same.
12052 (copy_tree_body_r): Same.
12053 (declare_return_variable): Same.
12054 (expand_call_inline): Same.
12055 * tree-nested.c (lookup_field_for_decl): Same.
12056 * tree-sra.c (create_access_replacement): Same.
12057 (generate_subtree_copies): Same.
12058 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Same.
12059 * tree-ssa-forwprop.c (combine_cond_expr_cond): Same.
12060 * tree-ssa-loop-ch.c (ch_base::copy_headers): Same.
12061 * tree-ssa-loop-im.c (execute_sm): Same.
12062 * tree-ssa-phiopt.c (cond_store_replacement): Same.
12063 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
12064 (handle_builtin_strcpy): Same.
12065 (maybe_diag_stxncpy_trunc): Same.
12066 (handle_builtin_stxncpy_strncat): Same.
12067 (handle_builtin_strcat): Same.
12068 * tree-ssa-uninit.c (get_no_uninit_warning): Same.
12069 (set_no_uninit_warning): Same.
12070 (uninit_undefined_value_p): Same.
12071 (warn_uninit): Same.
12072 (maybe_warn_operand): Same.
12073 * tree-vrp.c (compare_values_warnv): Same.
12074 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr): Same.
12075 (test_for_singularity): Same.
12076 * gimple.h (warning_suppressed_p): New function.
12077 (suppress_warning): Same.
12078 (copy_no_warning): Same.
12079 (gimple_set_block): Call gimple_set_location.
12080 (gimple_set_location): Call copy_warning.
12082 2021-06-25 Martin Sebor <msebor@redhat.com>
12084 * tree.h (warning_suppressed_at, copy_warning,
12085 warning_suppressed_p, suppress_warning): New functions.
12087 2021-06-25 Martin Sebor <msebor@redhat.com>
12089 * Makefile.in (OBJS-libcommon): Add diagnostic-spec.o.
12090 * gengtype.c (open_base_files): Add diagnostic-spec.h.
12091 * diagnostic-spec.c: New file.
12092 * diagnostic-spec.h: New file.
12093 * tree.h (no_warning, all_warnings, suppress_warning_at): New
12095 * warning-control.cc: New file.
12097 2021-06-25 liuhongt <hongtao.liu@intel.com>
12100 * config/i386/i386.c (x86_order_regs_for_local_alloc):
12103 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
12105 PR tree-optimization/101189
12106 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Pass
12107 LHS range of condition to postfold routine.
12108 (fold_using_range::postfold_gcond_edges): Only process the TRUE or
12109 FALSE edge if the LHS range supports it being taken.
12110 * gimple-range-fold.h (postfold_gcond_edges): Add range parameter.
12112 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
12114 * value-relation.cc (equiv_oracle::dump): Do not dump NULL blocks.
12115 (relation_oracle::find_relation_block): Check correct bitmap.
12116 (relation_oracle::dump): Do not dump NULL blocks.
12118 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
12120 * gimple-range-cache.cc (ranger_cache::propagate_cache): Call
12121 range_on_edge instead of manually calculating.
12123 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
12125 * range-op.cc: Fix comment.
12127 2021-06-24 Uroš Bizjak <ubizjak@gmail.com>
12130 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
12131 Handle V8QI and V4HI modes.
12132 * config/i386/mmx.md (sse4_1_<any_extend:code>v4qiv4hi2):
12134 (sse4_1_<any_extend:code>v4qiv4hi2): Ditto.
12135 (mmxpackmode): New mode attribute.
12136 (vec_pack_trunc_<mmxpackmode:mode>): New expander.
12137 (mmxunpackmode): New mode attribute.
12138 (vec_unpacks_lo_<mmxunpackmode:mode>): New expander.
12139 (vec_unpacks_hi_<mmxunpackmode:mode>): Ditto.
12140 (vec_unpacku_lo_<mmxunpackmode:mode>): Ditto.
12141 (vec_unpacku_hi_<mmxunpackmode:mode>): Ditto.
12142 * config/i386/i386.md (extsuffix): Move from ...
12143 * config/i386/sse.md: ... here.
12145 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
12147 * dwarf2out.c (dwarf2out_assembly_start): Emit .file 0 marker here..
12148 (dwarf2out_finish): ...instead of here.
12150 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
12152 * configure.ac (--gdwarf-5 option): Use objdump instead of readelf.
12153 (working --gdwarf-4/--gdwarf-5 for all sources): Likewise.
12154 (--gdwarf-4 not refusing generated .debug_line): Adjust for Windows.
12155 * configure: Regenerate.
12157 2021-06-24 Richard Biener <rguenther@suse.de>
12159 * config/i386/sse.md (vec_addsubv4df3, vec_addsubv2df3,
12160 vec_addsubv8sf3, vec_addsubv4sf3): Merge into ...
12161 (vec_addsub<mode>3): ... using a new addsub_cst mode attribute.
12163 2021-06-24 Richard Biener <rguenther@suse.de>
12165 * config/i386/sse.md (avx_addsubv4df3): Rename to
12167 (avx_addsubv8sf3): Rename to vec_addsubv8sf3.
12168 (sse3_addsubv2df3): Rename to vec_addsubv2df3.
12169 (sse3_addsubv4sf3): Rename to vec_addsubv4sf3.
12170 * config/i386/i386-builtin.def: Adjust.
12171 * internal-fn.def (VEC_ADDSUB): New internal optab fn.
12172 * optabs.def (vec_addsub_optab): New optab.
12173 * tree-vect-slp-patterns.c (class addsub_pattern): New.
12174 (slp_patterns): Add addsub_pattern.
12175 * tree-vect-slp.c (vect_optimize_slp): Disable propagation
12176 across CFN_VEC_ADDSUB.
12177 * tree-vectorizer.h (vect_pattern::vect_pattern): Make
12179 * doc/md.texi (vec_addsub<mode>3): Document.
12181 2021-06-24 Jakub Jelinek <jakub@redhat.com>
12183 PR middle-end/101170
12184 * df-scan.c (df_ref_record): For paradoxical big-endian SUBREGs
12185 where regno + subreg_regno_offset wraps around use 0 as starting
12188 2021-06-24 Jakub Jelinek <jakub@redhat.com>
12190 PR middle-end/101172
12191 * stor-layout.c (finish_bitfield_representative): If nextf has
12192 error_mark_node type, set repr type to error_mark_node too.
12194 2021-06-24 Ilya Leoshkevich <iii@linux.ibm.com>
12196 * config/s390/s390.c (s390_function_profiler): Ignore labelno
12198 * config/s390/s390.h (NO_PROFILE_COUNTERS): Define.
12200 2021-06-24 Richard Biener <rguenther@suse.de>
12202 * tree-vect-slp.c (vect_optimize_slp): Do not propagate
12203 across operations that have different semantics on different
12206 2021-06-24 Jakub Jelinek <jakub@redhat.com>
12208 * tree.h (OMP_CLAUSE_MAP_IN_REDUCTION): Document meaning for OpenMP.
12209 * gimplify.c (gimplify_scan_omp_clauses): For OpenMP map clauses
12210 with OMP_CLAUSE_MAP_IN_REDUCTION flag partially defer gimplification
12211 of non-decl OMP_CLAUSE_DECL. For OMP_CLAUSE_IN_REDUCTION on
12212 OMP_TARGET user outer_ctx instead of ctx for placeholders and
12213 initializer/combiner gimplification.
12214 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_MAP_IN_REDUCTION
12215 on target constructs.
12216 (lower_rec_input_clauses): Likewise.
12217 (lower_omp_target): Likewise.
12218 * omp-expand.c (expand_omp_target): Temporarily ignore nowait clause
12219 on target if in_reduction is present.
12221 2021-06-24 Kewen Lin <linkw@linux.ibm.com>
12223 * tree-predcom.c (class pcom_worker): New class.
12224 (release_chain): Renamed to...
12225 (pcom_worker::release_chain): ...this.
12226 (release_chains): Renamed to...
12227 (pcom_worker::release_chains): ...this.
12228 (aff_combination_dr_offset): Renamed to...
12229 (pcom_worker::aff_combination_dr_offset): ...this.
12230 (determine_offset): Renamed to...
12231 (pcom_worker::determine_offset): ...this.
12232 (class comp_ptrs): New class.
12233 (split_data_refs_to_components): Renamed to...
12234 (pcom_worker::split_data_refs_to_components): ...this,
12235 and update with class comp_ptrs.
12236 (suitable_component_p): Renamed to...
12237 (pcom_worker::suitable_component_p): ...this.
12238 (filter_suitable_components): Renamed to...
12239 (pcom_worker::filter_suitable_components): ...this.
12240 (valid_initializer_p): Renamed to...
12241 (pcom_worker::valid_initializer_p): ...this.
12242 (find_looparound_phi): Renamed to...
12243 (pcom_worker::find_looparound_phi): ...this.
12244 (add_looparound_copies): Renamed to...
12245 (pcom_worker::add_looparound_copies): ...this.
12246 (determine_roots_comp): Renamed to...
12247 (pcom_worker::determine_roots_comp): ...this.
12248 (determine_roots): Renamed to...
12249 (pcom_worker::determine_roots): ...this.
12250 (single_nonlooparound_use): Renamed to...
12251 (pcom_worker::single_nonlooparound_use): ...this.
12252 (remove_stmt): Renamed to...
12253 (pcom_worker::remove_stmt): ...this.
12254 (execute_pred_commoning_chain): Renamed to...
12255 (pcom_worker::execute_pred_commoning_chain): ...this.
12256 (execute_pred_commoning): Renamed to...
12257 (pcom_worker::execute_pred_commoning): ...this.
12258 (struct epcc_data): New member worker.
12259 (execute_pred_commoning_cbck): Call execute_pred_commoning
12260 with pcom_worker pointer.
12261 (find_use_stmt): Renamed to...
12262 (pcom_worker::find_use_stmt): ...this.
12263 (find_associative_operation_root): Renamed to...
12264 (pcom_worker::find_associative_operation_root): ...this.
12265 (find_common_use_stmt): Renamed to...
12266 (pcom_worker::find_common_use_stmt): ...this.
12267 (combinable_refs_p): Renamed to...
12268 (pcom_worker::combinable_refs_p): ...this.
12269 (reassociate_to_the_same_stmt): Renamed to...
12270 (pcom_worker::reassociate_to_the_same_stmt): ...this.
12271 (stmt_combining_refs): Renamed to...
12272 (pcom_worker::stmt_combining_refs): ...this.
12273 (combine_chains): Renamed to...
12274 (pcom_worker::combine_chains): ...this.
12275 (try_combine_chains): Renamed to...
12276 (pcom_worker::try_combine_chains): ...this.
12277 (prepare_initializers_chain): Renamed to...
12278 (pcom_worker::prepare_initializers_chain): ...this.
12279 (prepare_initializers): Renamed to...
12280 (pcom_worker::prepare_initializers): ...this.
12281 (prepare_finalizers_chain): Renamed to...
12282 (pcom_worker::prepare_finalizers_chain): ...this.
12283 (prepare_finalizers): Renamed to...
12284 (pcom_worker::prepare_finalizers): ...this.
12285 (tree_predictive_commoning_loop): Renamed to...
12286 (pcom_worker::tree_predictive_commoning_loop): ...this, adjust
12287 some calls and remove some cleanup code.
12288 (tree_predictive_commoning): Adjusted to use pcom_worker instance.
12289 (static variable looparound_phis): Remove.
12290 (static variable name_expansions): Remove.
12292 2021-06-24 Richard Biener <rguenther@suse.de>
12294 * tree-vect-slp.c (slpg_vertex): New struct.
12295 (vect_slp_build_vertices): Adjust.
12296 (vect_optimize_slp): Likewise. Maintain an outgoing permute
12297 and a materialized one.
12299 2021-06-24 Richard Biener <rguenther@suse.de>
12301 PR tree-optimization/101105
12302 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
12303 Only ignore steps when they are equal or scalar order is preserved.
12305 2021-06-24 liuhongt <hongtao.liu@intel.com>
12308 * config/i386/i386-expand.c (ix86_expand_vec_interleave):
12309 Adjust comments for ix86_expand_vecop_qihi2.
12310 (ix86_expand_vecmul_qihi): Renamed to ..
12311 (ix86_expand_vecop_qihi2): Adjust function prototype to
12312 support shift operation, add static to definition.
12313 (ix86_expand_vec_shift_qihi_constant): Add static to definition.
12314 (ix86_expand_vecop_qihi): Call ix86_expand_vecop_qihi2 and
12315 ix86_expand_vec_shift_qihi_constant.
12316 * config/i386/i386-protos.h (ix86_expand_vecmul_qihi): Deleted.
12317 (ix86_expand_vec_shift_qihi_constant): Deleted.
12318 * config/i386/sse.md (VI12_256_512_AVX512VL): New mode
12320 (mulv8qi3): Call ix86_expand_vecop_qihi directly, add
12321 condition TARGET_64BIT.
12322 (mul<mode>3): Ditto.
12323 (<insn><mode>3): Ditto.
12324 (vlshr<mode>3): Extend to support avx512 vlshr.
12325 (v<insn><mode>3): New expander for
12327 (v<insn>v8qi3): Ditto.
12328 (vashrv8hi3<mask_name>): Renamed to ..
12329 (vashr<mode>3): And extend to support V16QImode for avx512.
12330 (vashrv16qi3): Deleted.
12331 (vashrv2di3<mask_name>): Extend expander to support avx512
12334 2021-06-23 Dimitar Dimitrov <dimitar@dinux.eu>
12336 * doc/lto.texi (Design Overview): Update that slim objects are
12339 2021-06-23 Aaron Sawdey <acsawdey@linux.ibm.com>
12341 * config/rs6000/rs6000-cpus.def: Take OPTION_MASK_PCREL_OPT out
12342 of OTHER_POWER10_MASKS so it will not be enabled by default.
12344 2021-06-23 Richard Biener <rguenther@suse.de>
12345 Martin Jambor <mjambor@suse.cz>
12347 * tree-inline.c (setup_one_parameter): Set TREE_READONLY of the
12348 param replacement unconditionally. Adjust comment.
12350 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
12352 * Makefile.in (OBJS): Add gimple-range-fold.o
12353 * gimple-range-fold.cc: New.
12354 * gimple-range-fold.h: New.
12355 * gimple-range-gori.cc (gimple_range_calc_op1): Move to here.
12356 (gimple_range_calc_op2): Ditto.
12357 * gimple-range-gori.h: Move prototypes to here.
12358 * gimple-range.cc: Adjust include files.
12359 (fur_source:fur_source): Relocate to gimple-range-fold.cc.
12360 (fur_source::get_operand): Ditto.
12361 (fur_source::get_phi_operand): Ditto.
12362 (fur_source::query_relation): Ditto.
12363 (fur_source::register_relation): Ditto.
12364 (class fur_edge): Ditto.
12365 (fur_edge::fur_edge): Ditto.
12366 (fur_edge::get_operand): Ditto.
12367 (fur_edge::get_phi_operand): Ditto.
12368 (fur_stmt::fur_stmt): Ditto.
12369 (fur_stmt::get_operand): Ditto.
12370 (fur_stmt::get_phi_operand): Ditto.
12371 (fur_stmt::query_relation): Ditto.
12372 (class fur_depend): Relocate to gimple-range-fold.h.
12373 (fur_depend::fur_depend): Relocate to gimple-range-fold.cc.
12374 (fur_depend::register_relation): Ditto.
12375 (fur_depend::register_relation): Ditto.
12376 (class fur_list): Ditto.
12377 (fur_list::fur_list): Ditto.
12378 (fur_list::get_operand): Ditto.
12379 (fur_list::get_phi_operand): Ditto.
12380 (fold_range): Ditto.
12381 (adjust_pointer_diff_expr): Ditto.
12382 (gimple_range_adjustment): Ditto.
12383 (gimple_range_base_of_assignment): Ditto.
12384 (gimple_range_operand1): Ditto.
12385 (gimple_range_operand2): Ditto.
12386 (gimple_range_calc_op1): Relocate to gimple-range-gori.cc.
12387 (gimple_range_calc_op2): Ditto.
12388 (fold_using_range::fold_stmt): Relocate to gimple-range-fold.cc.
12389 (fold_using_range::range_of_range_op): Ditto.
12390 (fold_using_range::range_of_address): Ditto.
12391 (fold_using_range::range_of_phi): Ditto.
12392 (fold_using_range::range_of_call): Ditto.
12393 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
12394 (fold_using_range::range_of_builtin_call): Ditto.
12395 (fold_using_range::range_of_cond_expr): Ditto.
12396 (fold_using_range::range_of_ssa_name_with_loop_info): Ditto.
12397 (fold_using_range::relation_fold_and_or): Ditto.
12398 (fold_using_range::postfold_gcond_edges): Ditto.
12399 * gimple-range.h: Add gimple-range-fold.h to include files. Change
12400 GIMPLE_RANGE_STMT_H to GIMPLE_RANGE_H.
12401 (gimple_range_handler): Relocate to gimple-range-fold.h.
12402 (gimple_range_ssa_p): Ditto.
12403 (range_compatible_p): Ditto.
12404 (class fur_source): Ditto.
12405 (class fur_stmt): Ditto.
12406 (class fold_using_range): Ditto.
12407 (gimple_range_calc_op1): Relocate to gimple-range-gori.h
12408 (gimple_range_calc_op2): Ditto.
12410 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
12412 PR tree-optimization/101148
12413 PR tree-optimization/101014
12414 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
12415 (ranger_cache::~ranger_cache): Adjust.
12416 (ranger_cache::block_range): Check if propagation disallowed.
12417 (ranger_cache::propagate_cache): Disallow propagation if new value
12418 can't be stored properly.
12419 * gimple-range-cache.h (ranger_cache::m_propfail): New member.
12421 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
12423 * gimple-range-cache.cc (class ssa_block_ranges): Adjust prototype.
12424 (sbr_vector::set_bb_range): Return true.
12425 (class sbr_sparse_bitmap): Adjust.
12426 (sbr_sparse_bitmap::set_bb_range): Return value.
12427 (block_range_cache::set_bb_range): Return value.
12428 (ranger_cache::propagate_cache): Use return value to print msg.
12429 * gimple-range-cache.h (class block_range_cache): Adjust.
12431 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
12433 * gimple-range.cc (dump_bb): Use range_on_edge from the cache.
12435 2021-06-23 Jeff Law <jeffreyalaw@gmail.com>
12437 * config/h8300/logical.md (<code><mode>3<ccnz>): Use <cczn>
12438 so this pattern can be used for test/compare removal. Pass
12439 current insn to compute_logical_op_length and output_logical_op.
12440 * config/h8300/h8300.c (compute_logical_op_cc): Remove.
12441 (h8300_and_costs): Add argument to compute_logical_op_length.
12442 (output_logical_op): Add new argument. Use it to determine if the
12443 condition codes are used and adjust the output accordingly.
12444 (compute_logical_op_length): Add new argument and update length
12445 computations when condition codes are used.
12446 * config/h8300/h8300-protos.h (compute_logical_op_length): Update
12448 (output_logical_op): Likewise.
12450 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
12453 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
12454 Handle 64bit modes for TARGET_XOP. Use indirect gen_* functions.
12455 * config/i386/mmx.md (mmx_ppermv64): New insn pattern.
12456 * config/i386/i386.md (unspec): Move UNSPEC_XOP_PERMUTE from ...
12457 * config/i386/sse.md (unspec): ... here.
12459 2021-06-23 Martin Liska <mliska@suse.cz>
12462 * optc-save-gen.awk: Put back arm_fp16_format to
12465 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
12468 * config/i386/i386.md (bsr_rex64): Add zero-flag setting RTX.
12471 (clz<mode>2): Update RTX pattern for additions.
12473 2021-06-23 Jakub Jelinek <jakub@redhat.com>
12475 PR middle-end/101167
12476 * omp-low.c (lower_omp_regimplify_p): Regimplify also PARM_DECLs
12477 and RESULT_DECLs that have DECL_HAS_VALUE_EXPR_P set.
12479 2021-06-22 Sergei Trofimovich <siarheit@google.com>
12481 * doc/rtl.texi: drop unbalanced parenthesis.
12483 2021-06-22 Richard Biener <rguenther@suse.de>
12485 PR middle-end/101156
12486 * gimplify.c (gimplify_expr): Remove premature incorrect
12489 2021-06-22 Jakub Jelinek <jakub@redhat.com>
12491 PR tree-optimization/101159
12492 * tree-vect-patterns.c (vect_recog_popcount_pattern): Fix some
12495 2021-06-22 Jakub Jelinek <jakub@redhat.com>
12497 PR middle-end/101160
12498 * function.c (assign_parms): For decl_result with TYPE_EMPTY_P type
12499 clear crtl->return_rtx instead of keeping it referencing a pseudo.
12501 2021-06-22 Jakub Jelinek <jakub@redhat.com>
12502 Andrew Pinski <apinski@marvell.com>
12504 PR tree-optimization/101162
12505 * fold-const.c (range_check_type): Handle OFFSET_TYPE like pointer
12508 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12510 * range-op.cc (range_relational_tests): New.
12511 (range_op_tests): Call range_relational_tests.
12513 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12515 * range-op.cc (operator_cast::lhs_op1_relation): New.
12516 (operator_identity::lhs_op1_relation): Mew.
12518 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12520 * range-op.cc (operator_minus::op1_op2_relation_effect): New.
12522 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12524 * range-op.cc (operator_plus::lhs_op1_relation): New.
12525 (operator_plus::lhs_op2_relation): New.
12527 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12529 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a
12530 relation_oracle if dominators exist.
12531 (ranger_cache::~ranger_cache): Dispose of oracle.
12532 (ranger_cache::dump_bb): Dump oracle.
12533 * gimple-range.cc (fur_source::fur_source): New.
12534 (fur_source::get_operand): Use mmeber query.
12535 (fur_source::get_phi_operand): Use member_query.
12536 (fur_source::query_relation): New.
12537 (fur_source::register_dependency): Delete.
12538 (fur_source::register_relation): New.
12539 (fur_edge::fur_edge): Adjust.
12540 (fur_edge::get_phi_operand): Fix comment.
12541 (fur_edge::query): Delete.
12542 (fur_stmt::fur_stmt): Adjust.
12543 (fur_stmt::query): Delete.
12544 (fur_depend::fur_depend): Adjust.
12545 (fur_depend::register_relation): New.
12546 (fur_depend::register_relation): New.
12547 (fur_list::fur_list): Adjust.
12548 (fur_list::get_operand): Use member query.
12549 (fold_using_range::range_of_range_op): Process and query relations.
12550 (fold_using_range::range_of_address): Adjust dependency call.
12551 (fold_using_range::range_of_phi): Ditto.
12552 (gimple_ranger::gimple_ranger): New. Use ranger_ache oracle.
12553 (fold_using_range::relation_fold_and_or): New.
12554 (fold_using_range::postfold_gcond_edges): New.
12555 * gimple-range.h (class gimple_ranger): Adjust.
12556 (class fur_source): Adjust members.
12557 (class fur_stmt): Ditto.
12558 (class fold_using_range): Ditto.
12560 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12562 * range-op.cc (range_operator::wi_fold): Apply relation effect.
12563 (range_operator::fold_range): Adjust and apply relation effect.
12564 (*::fold_range): Add relation parameters.
12565 (*::op1_range): Ditto.
12566 (*::op2_range): Ditto.
12567 (range_operator::lhs_op1_relation): New.
12568 (range_operator::lhs_op2_relation): New.
12569 (range_operator::op1_op2_relation): New.
12570 (range_operator::op1_op2_relation_effect): New.
12571 (relop_early_resolve): New.
12572 (operator_equal::op1_op2_relation): New.
12573 (operator_equal::fold_range): Call relop_early_resolve.
12574 (operator_not_equal::op1_op2_relation): New.
12575 (operator_not_equal::fold_range): Call relop_early_resolve.
12576 (operator_lt::op1_op2_relation): New.
12577 (operator_lt::fold_range): Call relop_early_resolve.
12578 (operator_le::op1_op2_relation): New.
12579 (operator_le::fold_range): Call relop_early_resolve.
12580 (operator_gt::op1_op2_relation): New.
12581 (operator_gt::fold_range): Call relop_early_resolve.
12582 (operator_ge::op1_op2_relation): New.
12583 (operator_ge::fold_range): Call relop_early_resolve.
12584 * range-op.h (class range_operator): Adjust parameters and methods.
12586 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12588 * Makefile.in (OBJS): Add value-relation.o.
12589 * gimple-range.h: Adjust include files.
12590 * tree-data-ref.c: Adjust include file order.
12591 * value-query.cc (range_query::get_value_range): Default to no oracle.
12592 (range_query::query_relation): New.
12593 (range_query::query_relation): New.
12594 * value-query.h (class range_query): Adjust.
12595 * value-relation.cc: New.
12596 * value-relation.h: New.
12598 2021-06-22 Richard Biener <rguenther@suse.de>
12600 PR tree-optimization/101151
12601 * tree-ssa-sink.c (statement_sink_location): Expand irreducible
12604 2021-06-22 Jojo R <rjiejie@linux.alibaba.com>
12606 * config/riscv/riscv.c (thead_c906_tune_info): New.
12607 (riscv_tune_info_table): Use new tune.
12609 2021-06-22 Richard Biener <rguenther@suse.de>
12611 PR tree-optimization/101158
12612 * tree-vect-slp.c (vect_build_slp_tree_1): Move same operand
12613 checking after checking for matching operation.
12615 2021-06-22 Richard Biener <rguenther@suse.de>
12617 PR tree-optimization/101159
12618 * tree-vect-patterns.c (vect_recog_popcount_pattern): Add
12619 missing NULL vectype check.
12621 2021-06-22 Richard Biener <rguenther@suse.de>
12623 PR tree-optimization/101154
12624 * tree-vect-slp.c (vect_build_slp_tree_2): Fix out-of-bound access.
12626 2021-06-22 Jakub Jelinek <jakub@redhat.com>
12629 * config/i386/i386-protos.h (ix86_last_zero_store_uid): Declare.
12630 * config/i386/i386-expand.c (ix86_last_zero_store_uid): New variable.
12631 * config/i386/i386.c (ix86_expand_prologue): Clear it.
12632 * config/i386/i386.md (peephole2s for 1/2/4 stores of const0_rtx):
12633 Remove "" from match_operand. Emit new insns using emit_move_insn and
12634 set ix86_last_zero_store_uid to INSN_UID of the last store.
12635 Add peephole2s for 1/2/4 stores of const0_rtx following previous
12638 2021-06-22 Martin Liska <mliska@suse.cz>
12640 * auto-profile.c (AUTO_PROFILE_VERSION): Bump as string format
12643 2021-06-22 Martin Liska <mliska@suse.cz>
12645 * gcov-io.h: Remove padding entries.
12647 2021-06-22 liuhongt <hongtao.liu@intel.com>
12649 PR tree-optimization/97770
12650 * tree-vect-patterns.c (vect_recog_popcount_pattern):
12652 (vect_recog_func vect_vect_recog_func_ptrs): Add new pattern.
12654 2021-06-22 liuhongt <hongtao.liu@intel.com>
12657 * config/i386/i386-builtin.def (BDESC): Adjust builtin name.
12658 * config/i386/sse.md (<avx512>_expand<mode>_mask): Rename to ..
12659 (expand<mode>_mask): this ..
12660 (*expand<mode>_mask): New pre_reload splitter to transform
12661 v{,p}expand* to vmov* when mask is zero, all ones, or has all
12662 ones in it's lower part, otherwise still generate
12665 2021-06-22 liuhongt <hongtao.liu@intel.com>
12668 * config/i386/i386-expand.c
12669 (ix86_expand_special_args_builtin): Keep constm1_operand only
12670 if it satisfies insn's operand predicate.
12672 2021-06-21 Jason Merrill <jason@redhat.com>
12675 * df-scan.c (df_ref_record): Check that regno < endregno.
12676 * function.c (assign_parms, expand_function_end): Do nothing with a
12677 TYPE_EMPTY_P result.
12679 2021-06-21 Richard Biener <rguenther@suse.de>
12681 PR tree-optimization/101120
12682 * tree-vect-data-refs.c (bump_vector_ptr): Fold the
12684 * tree-vect-slp.c (vect_transform_slp_perm_load): Add
12685 DR chain DCE capability.
12686 * tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
12687 * tree-vect-stmts.c (vectorizable_load): Remove unused
12688 loads in the DR chain for SLP.
12690 2021-06-21 Jakub Jelinek <jakub@redhat.com>
12692 PR inline-asm/100785
12693 * gimplify.c (gimplify_asm_expr): Don't diagnose errors if
12694 output or input operands were already error_mark_node.
12695 * cfgexpand.c (expand_asm_stmt): If errors are emitted,
12696 remove all inputs, outputs and clobbers from the asm and
12697 set template to "".
12699 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
12701 * config/arm/arm_neon.h (vceq_s8): Replace builtin with __a == __b.
12702 (vceq_s16): Likewise.
12703 (vceq_s32): Likewise.
12704 (vceq_u8): Likewise.
12705 (vceq_u16): Likewise.
12706 (vceq_u32): Likewise.
12707 (vceq_p8): Likewise.
12708 (vceqq_s8): Likewise.
12709 (vceqq_s16): Likewise.
12710 (vceqq_s32): Likewise.
12711 (vceqq_u8): Likewise.
12712 (vceqq_u16): Likewise.
12713 (vceqq_u32): Likewise.
12714 (vceqq_p8): Likewise.
12715 (vceq_f32): Gate __a == __b on __FAST_MATH__.
12716 (vceqq_f32): Likewise.
12717 (vceq_f16): Likewise.
12718 (vceqq_f16): Likewise.
12720 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
12723 * config/arm/iterators.md (NEON_VACMP): Remove.
12724 * config/arm/neon.md (neon_vca<cmp_op><mode>): Use GLTE instead of GTGE
12726 (neon_vca<cmp_op><mode>_insn): Likewise.
12727 (neon_vca<cmp_op_unsp><mode>_insn_unspec): Use NEON_VAGLTE instead of
12730 2021-06-21 Richard Biener <rguenther@suse.de>
12732 PR tree-optimization/101121
12733 * tree-vect-slp.c (vect_build_slp_tree_2): To not fail fatally
12734 when we just lack a stmt with the desired op when doing permutation.
12735 (vect_build_slp_tree): When caching a failed SLP build attempt
12736 assert that at least one lane is marked as not matching.
12738 2021-06-21 liuhongt <hongtao.liu@intel.com>
12741 * config/i386/i386.md: (*anddi_1): Disparage slightly the mask
12742 register alternative.
12743 (*and<mode>_1): Ditto.
12745 (*andn<mode>_1): Ditto.
12746 (*<code><mode>_1): Ditto.
12747 (*<code>qi_1): Ditto.
12748 (*one_cmpl<mode>2_1): Ditto.
12749 (*one_cmplsi2_1_zext): Ditto.
12750 (*one_cmplqi2_1): Ditto.
12751 * config/i386/i386.c (x86_order_regs_for_local_alloc): Change
12752 the order of mask registers to be before general registers.
12754 2021-06-21 Roger Sayle <roger@nextmovesoftware.com>
12757 * config/i386/i386.md: New define_peephole2s to shrink writing
12758 1, 2 or 4 consecutive zeros to memory when optimizing for size.
12760 2021-06-18 Jeff Law <jeffreyalaw@gmail.com>
12762 * config/h8300/h8300.c (h8300_select_cc_mode): Handle SYMBOL_REF.
12763 * config/h8300/logical.md (<code><mode>3 logcial expander): Generate
12764 more efficient code when the source can be trivially simplified.
12766 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
12768 * gimple-range-cache.cc (ranger_cache::range_of_def): Calculate
12769 a range if global is not available.
12770 (ranger_cache::entry_range): Fallback to range_of_def.
12771 * gimple-range-cache.h (range_of_def): Adjust prototype.
12773 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
12775 PR tree-optimization/101014
12776 * gimple-range-cache.cc (ranger_cache::ranger_cache): Remove poor
12778 (ranger_cache::~ranger_cache): Ditto.
12779 (ranger_cache::enable_new_values): Delete.
12780 (ranger_cache::push_poor_value): Delete.
12781 (ranger_cache::range_of_def): Remove poor value processing.
12782 (ranger_cache::entry_range): Ditto.
12783 (ranger_cache::fill_block_cache): Ditto.
12784 * gimple-range-cache.h (class ranger_cache): Remove poor value members.
12785 * gimple-range.cc (gimple_ranger::range_of_expr): Remove call.
12786 * gimple-range.h (class gimple_ranger): Adjust.
12788 2021-06-18 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
12791 * common/config/arm/arm-common.c (arm_canon_arch_option_1): New function
12792 derived from arm_canon_arch.
12793 (arm_canon_arch_option): Call it.
12794 (arm_canon_arch_multilib_option): New function.
12795 * config/arm/arm-cpus.in (IGNORE_FOR_MULTILIB): New fgroup.
12796 * config/arm/arm.h (arm_canon_arch_multilib_option): New prototype.
12797 (CANON_ARCH_MULTILIB_SPEC_FUNCTION): New macro.
12798 (MULTILIB_ARCH_CANONICAL_SPECS): New macro.
12799 (DRIVER_SELF_SPECS): Add MULTILIB_ARCH_CANONICAL_SPECS.
12800 * config/arm/arm.opt (mlibarch): New option.
12801 * config/arm/t-rmprofile (MULTILIB_MATCHES): For armv8*-m, replace use
12802 of march on RHS with mlibarch.
12804 2021-06-18 Marcel Vollweiler <marcel@codesourcery.com>
12806 * config.in: Regenerate.
12807 * config/gcn/gcn.c (print_operand_address): Fix for global_load assembler
12809 * configure: Regenerate.
12810 * configure.ac: Fix for global_load assembler functions.
12812 2021-06-18 Richard Biener <rguenther@suse.de>
12814 PR tree-optimization/101112
12815 * tree-vect-slp.c (vect_slp_linearize_chain): Fix condition
12816 to lookup a pattern stmt def.
12818 2021-06-18 Jakub Jelinek <jakub@redhat.com>
12820 PR middle-end/101062
12821 * stor-layout.c (finish_bitfield_layout): Don't add bitfield
12822 representatives in QUAL_UNION_TYPE.
12824 2021-06-18 Andrew Pinski <apinski@marvell.com>
12826 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
12827 Add counting of how many times it is done.
12828 (factor_out_conditional_conversion): Likewise.
12829 (match_simplify_replacement): Likewise.
12830 (value_replacement): Likewise.
12831 (spaceship_replacement): Likewise.
12832 (cond_store_replacement): Likewise.
12833 (cond_if_else_store_replacement_1): Likewise.
12834 (hoist_adjacent_loads): Likewise.
12836 2021-06-18 Andrew Pinski <apinski@marvell.com>
12838 * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
12839 types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and CNONJ_EXPR.
12840 (verify_gimple_assign_binary): Reject point and offset types on
12841 MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
12842 FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
12843 FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.
12845 2021-06-18 Michael Meissner <meissner@linux.ibm.com>
12847 * config/rs6000/rs6000.c (rs6000_emit_minmax): Add support for ISA
12848 3.1 IEEE 128-bit floating point xsmaxcqp/xsmincqp instructions.
12849 * config/rs6000/rs6000.md (s<minmax><mode>3, IEEE128 iterator):
12852 2021-06-17 Aaron Sawdey <acsawdey@linux.ibm.com>
12854 * config/rs6000/genfusion.pl (gen_logical_addsubf): Add
12855 earlyclobber to alts 0/1.
12856 (gen_addadd): Add earlyclobber to alts 0/1.
12857 * config/rs6000/fusion.md: Regenerate file.
12859 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12861 * cfgloopanal.c (get_loop_hot_path): Make path an auto_vec.
12863 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
12865 * gimple-range-cache.cc: Comment cleanups.
12866 * gimple-range-gori.cc: Comment cleanups.
12867 * gimple-range.cc: Comment/spacing cleanups
12868 * value-range.h: Comment cleanups.
12870 2021-06-17 H.J. Lu <hjl.tools@gmail.com>
12873 * calls.c (expand_call): Replace PUSH_ARGS with
12874 targetm.calls.push_argument (0).
12875 (emit_library_call_value_1): Likewise.
12876 * defaults.h (PUSH_ARGS): Removed.
12877 (PUSH_ARGS_REVERSED): Replace PUSH_ARGS with
12878 targetm.calls.push_argument (0).
12879 * expr.c (block_move_libcall_safe_for_call_parm): Likewise.
12880 (emit_push_insn): Pass the number bytes to push to
12881 targetm.calls.push_argument and pass 0 if ARGS_ADDR is 0.
12882 * hooks.c (hook_bool_uint_true): New.
12883 * hooks.h (hook_bool_uint_true): Likewise.
12884 * rtlanal.c (nonzero_bits1): Replace PUSH_ARGS with
12885 targetm.calls.push_argument (0).
12886 * target.def (push_argument): Add a targetm.calls hook.
12887 * targhooks.c (default_push_argument): New.
12888 * targhooks.h (default_push_argument): Likewise.
12889 * config/bpf/bpf.h (PUSH_ARGS): Removed.
12890 * config/cr16/cr16.c (TARGET_PUSH_ARGUMENT): New.
12891 * config/cr16/cr16.h (PUSH_ARGS): Removed.
12892 * config/i386/i386.c (ix86_push_argument): New.
12893 (TARGET_PUSH_ARGUMENT): Likewise.
12894 * config/i386/i386.h (PUSH_ARGS): Removed.
12895 * config/m32c/m32c.c (TARGET_PUSH_ARGUMENT): New.
12896 * config/m32c/m32c.h (PUSH_ARGS): Removed.
12897 * config/nios2/nios2.h (PUSH_ARGS): Likewise.
12898 * config/pru/pru.h (PUSH_ARGS): Likewise.
12899 * doc/tm.texi.in: Remove PUSH_ARGS documentation. Add
12900 TARGET_PUSH_ARGUMENT hook.
12901 * doc/tm.texi: Regenerated.
12903 2021-06-17 Uroš Bizjak <ubizjak@gmail.com>
12906 * config/i386/i386-expand.c (expand_vector_set_var):
12907 Handle V2FS mode remapping. Pass TARGET_MMX_WITH_SSE to
12908 ix86_expand_vector_init_duplicate.
12909 (ix86_expand_vector_init_duplicate): Emit insv_1 for
12910 QImode for !TARGET_PARTIAL_REG_STALL.
12911 * config/i386/predicates.md (vec_setm_mmx_operand): New predicate.
12912 * config/i386/mmx.md (vec_setv2sf): Use vec_setm_mmx_operand
12913 as operand 2 predicate. Call ix86_expand_vector_set_var
12914 for non-constant index operand.
12915 (vec_setv2si): Ditto.
12916 (vec_setv4hi): Ditto.
12917 (vec_setv8qi): ditto.
12919 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
12921 PR tree-optimization/100790
12922 * gimple-range.cc (range_of_builtin_call): Cleanup clz and ctz
12925 2021-06-17 Martin Liska <mliska@suse.cz>
12927 * doc/invoke.texi: Use consistently -O1 instead of -O.
12929 2021-06-17 Martin Liska <mliska@suse.cz>
12931 * gcov-io.h: Update documentation entry about string format.
12933 2021-06-17 Marius Hillenbrand <mhillen@linux.ibm.com>
12936 * config/s390/vecintrin.h (vec_doublee): Fix to use
12937 __builtin_s390_vflls.
12938 (vec_floate): Fix to use __builtin_s390_vflrd.
12940 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12942 * dominance.c (get_dominated_to_depth): Return auto_vec<basic_block>.
12943 * dominance.h (get_dominated_to_depth): Likewise.
12944 (get_all_dominated_blocks): Likewise.
12945 * cfgcleanup.c (delete_unreachable_blocks): Adjust.
12946 * gcse.c (hoist_code): Likewise.
12947 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
12948 * tree-parloops.c (oacc_entry_exit_ok): Likewise.
12949 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Likewise.
12950 * tree-ssa-phiprop.c (pass_phiprop::execute): Likewise.
12952 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12954 * dominance.c (get_dominated_by_region): Return auto_vec<basic_block>.
12955 * dominance.h (get_dominated_by_region): Likewise.
12956 * tree-cfg.c (gimple_duplicate_sese_region): Adjust.
12957 (gimple_duplicate_sese_tail): Likewise.
12958 (move_sese_region_to_fn): Likewise.
12960 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12962 * dominance.c (get_dominated_by): Return auto_vec<basic_block>.
12963 * dominance.h (get_dominated_by): Likewise.
12964 * auto-profile.c (afdo_find_equiv_class): Adjust.
12965 * cfgloopmanip.c (duplicate_loop_to_header_edge): Likewise.
12966 * loop-unroll.c (unroll_loop_runtime_iterations): Likewise.
12967 * tree-cfg.c (test_linear_chain): Likewise.
12968 (test_diamond): Likewise.
12970 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12972 * cfgloop.h (get_loop_hot_path): Return auto_vec<basic_block>.
12973 * cfgloopanal.c (get_loop_hot_path): Likewise.
12974 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
12976 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12978 * cgraph.c (cgraph_node::collect_callers): Return
12979 auto_vec<cgraph_edge *>.
12980 * cgraph.h (cgraph_node::collect_callers): Likewise.
12981 * ipa-cp.c (create_specialized_node): Adjust.
12982 (decide_about_value): Likewise.
12983 (decide_whether_version_node): Likewise.
12984 * ipa-sra.c (process_isra_node_results): Likewise.
12986 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12988 * vec.h (vl_ptr>::using_auto_storage): Handle null m_vec.
12989 (auto_vec<T, 0>::auto_vec): Define move constructor, and delete copy
12991 (auto_vec<T, 0>::operator=): Define move assignment and delete copy
12994 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
12996 * gimple-range.cc (debug_seed_ranger): New.
12997 (dump_ranger): New.
12998 (debug_ranger): New.
13000 2021-06-17 Richard Biener <rguenther@suse.de>
13002 PR tree-optimization/54400
13003 * tree-vectorizer.h (enum slp_instance_kind): Add
13004 slp_inst_kind_bb_reduc.
13005 (reduction_fn_for_scalar_code): Declare.
13006 * tree-vect-data-refs.c (vect_slp_analyze_instance_dependence):
13007 Check SLP_INSTANCE_KIND instead of looking at the
13009 (vect_slp_analyze_instance_alignment): Likewise.
13010 * tree-vect-loop.c (reduction_fn_for_scalar_code): Export.
13011 * tree-vect-slp.c (vect_slp_linearize_chain): Split out
13012 chain linearization from vect_build_slp_tree_2 and generalize
13013 for the use of BB reduction vectorization.
13014 (vect_build_slp_tree_2): Adjust accordingly.
13015 (vect_optimize_slp): Elide permutes at the root of BB reduction
13017 (vectorizable_bb_reduc_epilogue): New function.
13018 (vect_slp_prune_covered_roots): Likewise.
13019 (vect_slp_analyze_operations): Use them.
13020 (vect_slp_check_for_constructors): Recognize associatable
13021 chains for BB reduction vectorization.
13022 (vectorize_slp_instance_root_stmt): Generate code for the
13023 BB reduction epilogue.
13025 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
13027 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Check with
13029 (gori_compute::may_recompute_p): New.
13030 (gori_compute::outgoing_edge_range_p): Perform recomputations.
13031 * gimple-range-gori.h (class gori_compute): Add prototype.
13033 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
13035 * gimple-range-cache.cc (ranger_cache::range_on_edge): Always return
13036 true when a range can be calculated.
13037 * gimple-range.cc (gimple_ranger::dump_bb): Check has_edge_range_p.
13039 2021-06-16 Martin Sebor <msebor@redhat.com>
13041 * doc/invoke.texi (-Wmismatched-dealloc, -Wmismatched-new-delete):
13042 Correct documented defaults.
13044 2021-06-16 Andrew MacLeod <amacleod@redhat.com>
13046 * gimple-range-cache.cc (ranger_cache::ranger_cache): Initialize
13047 m_new_value_p directly.
13049 2021-06-16 Uroš Bizjak <ubizjak@gmail.com>
13052 * config/i386/i386-expand.c (expand_vec_perm_2perm_pblendv):
13053 Handle 64bit modes for TARGET_SSE4_1.
13054 (expand_vec_perm_pshufb2): Handle 64bit modes for TARGET_SSSE3.
13055 (expand_vec_perm_even_odd_pack): Handle V4HI mode.
13056 (expand_vec_perm_even_odd_1) <case E_V4HImode>: Expand via
13057 expand_vec_perm_pshufb2 for TARGET_SSSE3 and via
13058 expand_vec_perm_even_odd_pack for TARGET_SSE4_1.
13059 * config/i386/mmx.md (mmx_packusdw): New insn pattern.
13061 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
13063 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn<mode>):
13064 Change to an expander that emits the correct instruction
13065 depending on endianness.
13066 (aarch64_<sur><addsub>hn<mode>_insn_le): Define.
13067 (aarch64_<sur><addsub>hn<mode>_insn_be): Define.
13069 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
13071 * config/aarch64/aarch64-simd-builtins.def: Split generator
13072 for aarch64_<su>qmovn builtins into scalar and vector
13074 * config/aarch64/aarch64-simd.md (aarch64_<su>qmovn<mode>_insn_le):
13076 (aarch64_<su>qmovn<mode>_insn_be): Define.
13077 (aarch64_<su>qmovn<mode>): Split into scalar and vector
13078 variants. Change vector variant to an expander that emits the
13079 correct instruction depending on endianness.
13081 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
13083 * config/aarch64/aarch64-simd-builtins.def: Split generator
13084 for aarch64_sqmovun builtins into scalar and vector variants.
13085 * config/aarch64/aarch64-simd.md (aarch64_sqmovun<mode>):
13086 Split into scalar and vector variants. Change vector variant
13087 to an expander that emits the correct instruction depending
13089 (aarch64_sqmovun<mode>_insn_le): Define.
13090 (aarch64_sqmovun<mode>_insn_be): Define.
13092 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
13094 * config/aarch64/aarch64-simd.md (aarch64_xtn<mode>_insn_le):
13095 Define - modeling zero-high-half semantics.
13096 (aarch64_xtn<mode>): Change to an expander that emits the
13097 appropriate instruction depending on endianness.
13098 (aarch64_xtn<mode>_insn_be): Define - modeling zero-high-half
13100 (aarch64_xtn2<mode>_le): Rename to...
13101 (aarch64_xtn2<mode>_insn_le): This.
13102 (aarch64_xtn2<mode>_be): Rename to...
13103 (aarch64_xtn2<mode>_insn_be): This.
13104 (vec_pack_trunc_<mode>): Emit truncation instruction instead
13106 * config/aarch64/iterators.md (Vnarrowd): Add Vnarrowd mode
13107 attribute iterator.
13109 2021-06-16 Martin Jambor <mjambor@suse.cz>
13111 PR tree-optimization/100453
13112 * tree-sra.c (create_access): Disqualify any const candidates
13113 which are written to.
13114 (sra_modify_expr): Do not store sub-replacements back to a const base.
13115 (handle_unscalarized_data_in_subtree): Likewise.
13116 (sra_modify_assign): Likewise. Earlier, use TREE_READONLy test
13117 instead of constant_decl_p.
13119 2021-06-16 Jakub Jelinek <jakub@redhat.com>
13121 PR middle-end/101062
13122 * stor-layout.c (finish_bitfield_representative): For fields in unions
13123 assume nextf is always NULL.
13124 (finish_bitfield_layout): Compute bit field representatives also in
13125 unions, but handle it as if each bitfield was the only field in the
13128 2021-06-16 Richard Biener <rguenther@suse.de>
13130 PR tree-optimization/101088
13131 * tree-ssa-loop-im.c (sm_seq_valid_bb): Only look for
13132 supported refs on edges. Do not assert same ref but
13133 different kind stores are unsuported but mark them so.
13134 (hoist_memory_references): Only look for supported refs
13137 2021-06-16 Roger Sayle <roger@nextmovesoftware.com>
13139 PR rtl-optimization/46235
13140 * config/i386/i386.md: New define_split for bt followed by cmov.
13141 (*bt<mode>_setcqi): New define_insn_and_split for bt followed by setc.
13142 (*bt<mode>_setncqi): New define_insn_and_split for bt then setnc.
13143 (*bt<mode>_setnc<mode>): New define_insn_and_split for bt followed
13144 by setnc with zero extension.
13146 2021-06-16 Richard Biener <rguenther@suse.de>
13148 PR tree-optimization/101083
13149 * tree-vect-slp.c (vect_slp_build_two_operator_nodes): Get
13150 vectype as argument.
13151 (vect_build_slp_tree_2): Adjust.
13153 2021-06-15 Martin Sebor <msebor@redhat.com>
13155 PR middle-end/100876
13156 * builtins.c: (gimple_call_return_array): Account for size_t
13157 mangling as either unsigned int or unsigned long
13159 2021-06-15 Jeff Law <jeffreyalaw@gmail.com>
13161 * compare-elim.c (try_eliminate_compare): Run DCE to clean things
13162 up before eliminating comparisons.
13164 2021-06-15 Aldy Hernandez <aldyh@redhat.com>
13166 * range-op.cc (operator_bitwise_or::wi_fold): Make sure
13167 nonzero|X is nonzero.
13168 (range_op_bitwise_and_tests): Add tests for above.
13170 2021-06-15 Carl Love <cel@us.ibm.com>
13173 * config/rs6000/rs6000-builtin.def (VCMPEQUT): Fix the ICODE for the
13175 (VRLQ, VSLQ, VSRQ, VSRAQ): Remove unused BU_P10_OVERLOAD_2
13178 2021-06-15 Tobias Burnus <tobias@codesourcery.com>
13181 * gimplify.c (enum gimplify_defaultmap_kind): Add GDMK_SCALAR_TARGET.
13182 (struct gimplify_omp_ctx): Extend defaultmap array by one.
13183 (new_omp_context): Init defaultmap[GDMK_SCALAR_TARGET].
13184 (omp_notice_variable): Update type classification for Fortran.
13185 (gimplify_scan_omp_clauses): Update calls for new argument; handle
13186 GDMK_SCALAR_TARGET; for Fortran, GDMK_POINTER avoid GOVD_MAP_0LEN_ARRAY.
13187 * langhooks-def.h (lhd_omp_scalar_p): Add 'ptr_ok' argument.
13188 * langhooks.c (lhd_omp_scalar_p): Likewise.
13189 (LANG_HOOKS_OMP_ALLOCATABLE_P, LANG_HOOKS_OMP_SCALAR_TARGET_P): New.
13190 (LANG_HOOKS_DECLS): Add them.
13191 * langhooks.h (struct lang_hooks_for_decls): Add new hooks, update
13192 omp_scalar_p pointer type to include the new bool argument.
13194 2021-06-15 David Malcolm <dmalcolm@redhat.com>
13196 * doc/analyzer.texi
13197 (Special Functions for Debugging the Analyzer): Add
13198 __analyzer_dump_capacity.
13200 2021-06-15 Jakub Jelinek <jakub@redhat.com>
13203 * expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
13204 case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.
13206 2021-06-15 Richard Biener <rguenther@suse.de>
13208 * cfgloopanal.c (mark_irreducible_loops): Use a dominance
13209 check to identify loop latches.
13210 * cfgloop.c (verify_loop_structure): Likewise.
13211 * loop-init.c (apply_loop_flags): Allow marked irreducible
13212 regions even with multiple latches.
13213 * predict.c (rebuild_frequencies): Simplify.
13215 2021-06-15 Richard Biener <rguenther@suse.de>
13217 * tree-ssa-threadupdate.c
13218 (jump_thread_path_registry::mark_threaded_blocks): Assert we
13219 have marked irreducible regions.
13221 2021-06-14 Martin Sebor <msebor@redhat.com>
13224 * builtins.c (gimple_call_return_array): Check for attribute fn spec.
13225 Handle calls to placement new.
13226 (ndecl_dealloc_argno): Avoid placement delete.
13228 2021-06-14 Peter Bergner <bergner@linux.ibm.com>
13231 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
13232 create_tmp_reg_or_ssa_name().
13234 2021-06-14 Andrew MacLeod <amacleod@redhat.com>
13236 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
13237 (ranger_cache::enable_new_values): Set to specified value and
13238 return the old value.
13239 (ranger_cache::disable_new_values): Delete.
13240 (ranger_cache::fill_block_cache): Disable non 1st order derived
13242 * gimple-range-cache.h (ranger_cache): Adjust prototypes.
13243 * gimple-range.cc (gimple_ranger::range_of_expr): Adjust.
13245 2021-06-14 Uroš Bizjak <ubizjak@gmail.com>
13248 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
13249 Return true early when testing with V2HImode.
13250 * config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.
13252 2021-06-14 Christophe Lyon <christophe.lyon@linaro.org>
13254 * config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
13255 (mve_vec_unpack<US>_hi_<mode>): New pattern.
13256 (@mve_vec_pack_trunc_lo_<mode>): New pattern.
13257 (mve_vmovntq_<supf><mode>): Prefix with '@'.
13258 * config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
13260 (vec_unpack<US>_lo_<mode>): Likewise.
13261 (vec_pack_trunc_<mode>): Rename to
13262 neon_quad_vec_pack_trunc_<mode>.
13263 * config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
13265 (vec_unpack<US>_lo_<mode>): New.
13266 (vec_pack_trunc_<mode>): New.
13268 2021-06-14 Richard Biener <rguenther@suse.de>
13270 PR tree-optimization/100934
13271 * tree-ssa-dom.c (pass_dominator::execute): Properly
13272 mark irreducible regions.
13274 2021-06-14 Martin Liska <mliska@suse.cz>
13276 * doc/invoke.texi: Put r{...} on the same line as @item.
13278 2021-06-14 Martin Liska <mliska@suse.cz>
13280 * doc/invoke.texi: Add missing newline.
13282 2021-06-14 Martin Liska <mliska@suse.cz>
13284 * doc/invoke.texi: Remove '+' charasters.
13286 2021-06-14 Claudiu Zissulescu <claziss@synopsys.com>
13288 * config.gcc (arc): Add support for with_cpu option.
13289 * config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.
13291 2021-06-14 Richard Biener <rguenther@suse.de>
13293 PR tree-optimization/101031
13294 * tree-ssa-strlen.c (maybe_invalidate): Increment max_size
13295 instead of size when accounting for a possibly string
13298 2021-06-14 Martin Liska <mliska@suse.cz>
13300 * gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].
13302 2021-06-14 Aldy Hernandez <aldyh@redhat.com>
13304 * value-query.cc (gimple_range_global): Call get_range_global
13305 if called after inlining.
13307 2021-06-13 Uroš Bizjak <ubizjak@gmail.com>
13310 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
13311 Emit constant permutation insn directly from here.
13313 2021-06-13 Trevor Saunders <tbsaunde@tbsaunde.org>
13315 * attribs.c (find_attribute_namespace): Iterate over vec<> with
13317 * auto-profile.c (afdo_find_equiv_class): Likewise.
13318 * gcc.c (do_specs_vec): Likewise.
13319 (do_spec_1): Likewise.
13320 (driver::set_up_specs): Likewise.
13321 * gimple-loop-jam.c (any_access_function_variant_p): Likewise.
13322 * gimple-ssa-store-merging.c (compatible_load_p): Likewise.
13323 (imm_store_chain_info::try_coalesce_bswap): Likewise.
13324 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
13325 (get_location_for_stmts): Likewise.
13326 * graphite-poly.c (print_iteration_domains): Likewise.
13327 (free_poly_bb): Likewise.
13328 (remove_gbbs_in_scop): Likewise.
13329 (free_scop): Likewise.
13330 (dump_gbb_cases): Likewise.
13331 (dump_gbb_conditions): Likewise.
13332 (print_pdrs): Likewise.
13333 (print_scop): Likewise.
13334 * ifcvt.c (cond_move_process_if_block): Likewise.
13335 * lower-subreg.c (decompose_multiword_subregs): Likewise.
13336 * regcprop.c (pass_cprop_hardreg::execute): Likewise.
13337 * sanopt.c (sanitize_rewrite_addressable_params): Likewise.
13338 * sel-sched-dump.c (dump_insn_vector): Likewise.
13339 * store-motion.c (store_ops_ok): Likewise.
13340 (store_killed_in_insn): Likewise.
13341 * timevar.c (timer::named_items::print): Likewise.
13342 * tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
13343 (cleanup_tree_cfg_noloop): Likewise.
13344 * tree-data-ref.c (dump_data_references): Likewise.
13345 (print_dir_vectors): Likewise.
13346 (print_dist_vectors): Likewise.
13347 (dump_data_dependence_relations): Likewise.
13348 (dump_dist_dir_vectors): Likewise.
13349 (dump_ddrs): Likewise.
13350 (create_runtime_alias_checks): Likewise.
13351 (free_subscripts): Likewise.
13352 (save_dist_v): Likewise.
13353 (save_dir_v): Likewise.
13354 (invariant_access_functions): Likewise.
13355 (same_access_functions): Likewise.
13356 (access_functions_are_affine_or_constant_p): Likewise.
13357 (find_data_references_in_stmt): Likewise.
13358 (graphite_find_data_references_in_stmt): Likewise.
13359 (free_dependence_relations): Likewise.
13360 (free_data_refs): Likewise.
13361 * tree-inline.c (copy_debug_stmts): Likewise.
13362 * tree-into-ssa.c (dump_currdefs): Likewise.
13363 (rewrite_update_phi_arguments): Likewise.
13364 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
13365 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
13367 (vect_slp_analyze_node_dependences): Likewise.
13368 (vect_slp_analyze_instance_dependence): Likewise.
13369 (vect_record_base_alignments): Likewise.
13370 (vect_get_peeling_costs_all_drs): Likewise.
13371 (vect_peeling_supportable): Likewise.
13372 * tree-vectorizer.c (vec_info::~vec_info): Likewise.
13373 (vec_info::free_stmt_vec_infos): Likewise.
13375 2021-06-13 Jeff Law <jeffreyalaw@gmail.com>
13377 * config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
13378 (andqi3_1<cczn>): Removed.
13379 (<ors>qi3_1): Do not split for IOR/XOR a single bit.
13380 (H8/SX bit logicals): Split out from other patterns.
13381 * config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
13382 mulqihi3_const_clobber_flags.
13383 (mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly
13385 2021-06-13 H.J. Lu <hjl.tools@gmail.com>
13388 * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
13389 to true if red zone is used.
13390 (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
13391 ix86_red_zone_used.
13392 * config/i386/i386.h (machine_function): Add red_zone_used.
13393 (ix86_red_zone_size): Removed.
13394 (ix86_red_zone_used): New.
13395 * config/i386/i386.md (peephole2 patterns): Replace
13396 ix86_red_zone_size with ix86_red_zone_used.
13398 2021-06-12 Jason Merrill <jason@redhat.com>
13400 * doc/extend.texi (unused variable attribute): Applies to
13401 structure fields as well.
13403 2021-06-12 Eugene Rozenfeld <erozen@microsoft.com>
13405 * auto-profile.c (read_profile): fix a typo in an error string
13407 2021-06-11 Thomas Schwinge <thomas@codesourcery.com>
13409 * tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
13411 * tree-pretty-print.c (dump_omp_clauses): Update.
13412 (dump_generic_node) <OMP_CLAUSE>: Use it.
13414 2021-06-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
13417 * config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
13418 int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
13419 the polymorphic variants matching code.
13420 (__arm_vld1q_z): Likewise.
13421 (__arm_vld2q): Likewise.
13422 (__arm_vld4q): Likewise.
13423 (__arm_vldrbq_gather_offset): Likewise.
13424 (__arm_vldrbq_gather_offset_z): Likewise.
13426 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
13428 PR tree-optimization/96392
13429 * fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.
13431 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
13433 PR tree-optimization/96392
13434 * fold-const.c (fold_real_zero_addition_p): Take both arguments
13435 of the addition or subtraction, not just the zero. Use this
13436 other argument in tests for signaling NaNs and signed zeros.
13437 (tree_expr_maybe_real_minus_zero_p): New predicate.
13438 * fold-const.h (fold_real_zero_addition_p): Update prototype.
13439 (tree_expr_maybe_real_minus_zero_p): New function prototype.
13440 * match.pd: Update calls to fold_real_zero_addition_p.
13441 Replace HONOR_NANS with tree_expr_maybe_nan_p.
13442 Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
13443 Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
13444 * tree-ssa-reassoc.c (eliminate_using_constants): Update
13445 call to fold_real_zero_addition_p.
13447 2021-06-11 Richard Biener <rguenther@suse.de>
13449 PR tree-optimization/101025
13450 * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
13451 all refs that require dependence checking.
13453 2021-06-11 Richard Biener <rguenther@suse.de>
13455 PR tree-optimization/101028
13456 * tree-vect-slp.c (vect_build_slp_tree_2): When SLP
13457 reassoc discovery fails fatally, mark appropriate lanes
13460 2021-06-11 Richard Biener <rguenther@suse.de>
13462 PR tree-optimization/101026
13463 * tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
13464 have a representative for the associated chain nodes.
13466 2021-06-11 Jakub Jelinek <jakub@redhat.com>
13468 PR rtl-optimization/101008
13469 * simplify-rtx.c (relational_result): New function.
13470 (simplify_logical_relational_operation,
13471 simplify_relational_operation): Use it.
13473 2021-06-11 Jakub Jelinek <jakub@redhat.com>
13476 * config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.
13478 2021-06-11 Uroš Bizjak <ubizjak@gmail.com>
13481 * config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
13482 false if the permutation can be implemented with constant
13483 permutation instruction in wider mode.
13484 (canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
13485 Handle V8QImode and V4HImode.
13487 2021-06-11 Martin Liska <mliska@suse.cz>
13489 PR gcov-profile/100788
13490 * common.opt: Add new option.
13491 * coverage.c (coverage_begin_function): Emit warning instead on
13492 the internal compiler error.
13493 * doc/invoke.texi: Document the option.
13494 * toplev.c (process_options): Enable it by default.
13496 2021-06-11 Richard Biener <rguenther@suse.de>
13498 PR middle-end/101009
13499 * tree-data-ref.c (build_classic_dist_vector_1): Make sure
13500 to set *init_b to true when we encounter a constant equal
13502 (compute_affine_dependence): Also dump the actual DR_REF.
13504 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
13506 PR tree-optimization/100984
13507 * gimple-ssa-evrp.c (ssa_equiv_stack): Use auto_vec for
13508 replacements table.
13509 (ssa_equiv_stack::~ssa_equiv_stack): Remove.
13511 2021-06-11 Kewen Lin <linkw@linux.ibm.com>
13513 * config/rs6000/rs6000.md
13514 (floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
13515 define_insn_and_split.
13517 2021-06-11 Richard Biener <rguenther@suse.de>
13519 * tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
13520 to sort operands of the associative chain.
13522 2021-06-11 Richard Biener <rguenther@suse.de>
13524 * system.h (gcc_stablesort_r): Declare.
13525 * sort.cc (gcc_sort_r): Support stable sort.
13526 (gcc_stablesort_r): Define.
13527 * vec.h (vec<>::stablesort): Add.
13529 2021-06-10 Uroš Bizjak <ubizjak@gmail.com>
13532 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
13533 Handle V2SF mode. Emit SHUFPS to fixup unpack-high for V2SF mode.
13534 (expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
13535 (expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
13536 (expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
13537 (expand_vec_perm_interleave2): Handle 64bit modes.
13538 (expand_vec_perm_even_odd_pack): Handle V8QI mode.
13539 (expand_vec_perm_even_odd_1): Ditto.
13540 (ix86_vectorize_vec_perm_const): Ditto.
13541 * config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
13542 * config/i386/sse.md: ... here.
13543 * config/i386/mmx.md (*vec_interleave_lowv2sf):
13544 New insn_and_split pattern.
13545 (*vec_interleave_highv2sf): Ditto.
13546 (mmx_pshufbv8qi3): New insn pattern.
13547 (*mmx_pblendw): Ditto.
13549 2021-06-10 Peter Bergner <bergner@linux.ibm.com>
13551 * config/rs6000/rs6000-builtin.def (build_pair): New built-in.
13552 (build_acc): Likewise.
13553 * config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
13554 source operands in little-endian mode.
13555 (rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
13556 (mma_init_builtins): Likewise.
13557 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
13558 ordering for the MMA assemble and build source operands.
13559 * doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
13561 (__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
13564 2021-06-10 Jeff Law <jeffreyalaw@gmail.com>
13566 * config/h8300/h8300.c (select_cc_mode): Handle MEM. Use
13568 * config/h8300/extensions.md: Replace _clobber_flags patterns
13571 2021-06-10 Robin Dapp <rdapp@linux.ibm.com>
13573 * config/s390/vector.md (vcond_mask_<mode><mode>): Change to
13574 (vcond_mask_<mode><tointvec>): this.
13576 2021-06-10 Andrew Stubbs <ams@codesourcery.com>
13577 Thomas Schwinge <thomas@codesourcery.com>
13579 * omp-builtins.def (BUILT_IN_GOACC_ENTER_EXIT_DATA): Split into...
13580 (BUILT_IN_GOACC_ENTER_DATA, BUILT_IN_GOACC_EXIT_DATA): ... these.
13581 * gimple.h (enum gf_mask): Split
13582 'GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA' into
13583 'GF_OMP_TARGET_KIND_OACC_ENTER_DATA' and
13584 'GF_OMP_TARGET_KIND_OACC_EXIT_DATA'.
13585 (is_gimple_omp_oacc): Update.
13586 * gimple-pretty-print.c (dump_gimple_omp_target): Likewise.
13587 * gimplify.c (gimplify_omp_target_update): Likewise.
13588 * omp-expand.c (expand_omp_target, build_omp_regions_1)
13589 (omp_make_gimple_edges): Likewise.
13590 * omp-low.c (check_omp_nesting_restrictions, lower_omp_target):
13593 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
13595 * value-query.cc (value_query::value_on_edge): Rename name to
13597 (range_query::range_on_edge): Same.
13598 (range_query::value_of_expr): Same.
13599 (range_query::value_on_edge): Same.
13600 * value-query.h (class value_query): Same.
13601 (class range_query): Same.
13603 2021-06-10 Richard Biener <rguenther@suse.de>
13605 PR tree-optimization/101003
13606 * tree-vect-slp.c (vect_build_slp_tree_2): Appropriately
13607 use the pattern stmt defs when linearizing a chain.
13609 2021-06-10 Jakub Jelinek <jakub@redhat.com>
13612 * ifcvt.c (noce_get_alt_condition, noce_try_abs): Use
13613 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
13615 2021-06-10 Clement Chigot <clement.chigot@atos.net>
13617 * config/rs6000/aix71.h (ASM_CPU_SPEC): Add Power10 directive.
13618 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
13620 2021-06-09 Andrew Pinski <apinski@marvell.com>
13622 PR tree-optimization/100925
13623 * match.pd (a ? CST1 : CST2): Limit transformations
13624 that would produce a negative to integeral types only.
13625 Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.
13627 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
13630 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
13632 * doc/tm.texi: Correctly update.
13634 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
13636 * doc/tm.texi: Correctly update.
13638 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
13641 * doc/tm.texi.in (Trampolines): Add a missing blank line.
13643 2021-06-09 Paul Eggert <eggert@cs.ucla.edu>
13646 * doc/invoke.texi (Code Gen Options); Document that -fno-trampolines
13647 and -ftrampolines work only with Ada.
13648 * doc/tm.texi.in (Trampolines): Likewise.
13649 * doc/tm.texi: Regenerated.
13651 2021-06-09 Carl Love <cel@us.ibm.com>
13653 * config/rs6000/altivec.h (vec_signextll, vec_signexti, vec_signextq):
13654 Add define for new builtins.
13655 * config/rs6000/altivec.md(altivec_vreveti2): Add define_expand.
13656 * config/rs6000/rs6000-builtin.def (VSIGNEXTI, VSIGNEXTLL): Add
13657 overloaded builtin definitions.
13658 (VSIGNEXTSB2W, VSIGNEXTSH2W, VSIGNEXTSB2D, VSIGNEXTSH2D,VSIGNEXTSW2D,
13659 VSIGNEXTSD2Q): Add builtin expansions.
13660 (SIGNEXT): Add P10 overload definition.
13661 * config/rs6000/rs6000-call.c (P9V_BUILTIN_VEC_VSIGNEXTI, P9V_BUILTIN_VEC_VSIGNEXTLL,
13662 P10_BUILTIN_VEC_SIGNEXT): Add overloaded argument definitions.
13663 * config/rs6000/vsx.md (vsx_sign_extend_v2di_v1ti): Add define_insn.
13664 (vsignextend_v2di_v1ti, vsignextend_qi_<mode>, vsignextend_hi_<mode>,
13665 vsignextend_si_v2di)[VIlong]: Add define_expand.
13666 Make define_insn vsx_sign_extend_si_v2di visible.
13667 * doc/extend.texi: Add documentation for the vec_signexti,
13668 vec_signextll builtins and vec_signextq.
13670 2021-06-09 Carl Love <cel@us.ibm.com>
13672 * config/rs6000/rs6000.c (__fixkfti, __fixunskfti, __floattikf,
13673 __floatuntikf): Names changed to __fixkfti_sw, __fixunskfti_sw,
13674 __floattikf_sw, __floatuntikf_sw respectively.
13675 * config/rs6000/rs6000.md (floatti<mode>2, floatunsti<mode>2,
13676 fix_trunc<mode>ti2, fixuns_trunc<mode>ti2): Add
13677 define_insn for mode IEEE 128.
13679 2021-06-09 Carl Love <cel@us.ibm.com>
13681 * config/rs6000/altivec.md (altivec_vslq, altivec_vsrq):
13682 Rename to altivec_vslq_<mode>, altivec_vsrq_<mode>, mode VEC_TI.
13683 * config/rs6000/vector.md (VEC_TI): Was named VSX_TI in vsx.md.
13684 (vashlv1ti3): Change to vashl<mode>3, mode VEC_TI.
13685 (vlshrv1ti3): Change to vlshr<mode>3, mode VEC_TI.
13686 * config/rs6000/vsx.md (VSX_TI): Remove define_mode_iterator. Update
13687 uses of VSX_TI to VEC_TI.
13689 2021-06-09 Carl Love <cel@us.ibm.com>
13691 * config/rs6000/dfp.md (floattitd2, fixtdti2): New define_insns.
13693 2021-06-09 Carl Love <cel@us.ibm.com>
13695 * config/rs6000/altivec.h (vec_dive, vec_mod): Add define for new
13697 * config/rs6000/altivec.md (UNSPEC_VMULEUD, UNSPEC_VMULESD,
13698 UNSPEC_VMULOUD, UNSPEC_VMULOSD): New unspecs.
13699 (altivec_eqv1ti, altivec_gtv1ti, altivec_gtuv1ti, altivec_vmuleud,
13700 altivec_vmuloud, altivec_vmulesd, altivec_vmulosd, altivec_vrlq,
13701 altivec_vrlqmi, altivec_vrlqmi_inst, altivec_vrlqnm,
13702 altivec_vrlqnm_inst, altivec_vslq, altivec_vsrq, altivec_vsraq,
13703 altivec_vcmpequt_p, altivec_vcmpgtst_p, altivec_vcmpgtut_p): New
13705 (vec_widen_umult_even_v2di, vec_widen_smult_even_v2di,
13706 vec_widen_umult_odd_v2di, vec_widen_smult_odd_v2di, altivec_vrlqmi,
13707 altivec_vrlqnm): New define_expands.
13708 * config/rs6000/rs6000-builtin.def (VCMPEQUT_P, VCMPGTST_P,
13709 VCMPGTUT_P): Add macro expansions.
13710 (BU_P10V_AV_P): Add builtin predicate definition.
13711 (VCMPGTUT, VCMPGTST, VCMPEQUT, CMPNET, CMPGE_1TI,
13712 CMPGE_U1TI, CMPLE_1TI, CMPLE_U1TI, VNOR_V1TI_UNS, VNOR_V1TI, VCMPNET_P,
13713 VCMPAET_P, VMULEUD, VMULESD, VMULOUD, VMULOSD, VRLQ,
13714 VSLQ, VSRQ, VSRAQ, VRLQNM, DIV_V1TI, UDIV_V1TI, DIVES_V1TI, DIVEU_V1TI,
13715 MODS_V1TI, MODU_V1TI, VRLQMI): New macro expansions.
13716 (VRLQ, VSLQ, VSRQ, VSRAQ, DIVE, MOD): New overload expansions.
13717 * config/rs6000/rs6000-call.c (P10_BUILTIN_VCMPEQUT,
13718 P10V_BUILTIN_CMPGE_1TI, P10V_BUILTIN_CMPGE_U1TI,
13719 P10V_BUILTIN_VCMPGTUT, P10V_BUILTIN_VCMPGTST,
13720 P10V_BUILTIN_CMPLE_1TI, P10V_BUILTIN_VCMPLE_U1TI,
13721 P10V_BUILTIN_DIV_V1TI, P10V_BUILTIN_UDIV_V1TI,
13722 P10V_BUILTIN_VMULESD, P10V_BUILTIN_VMULEUD,
13723 P10V_BUILTIN_VMULOSD, P10V_BUILTIN_VMULOUD,
13724 P10V_BUILTIN_VNOR_V1TI, P10V_BUILTIN_VNOR_V1TI_UNS,
13725 P10V_BUILTIN_VRLQ, P10V_BUILTIN_VRLQMI,
13726 P10V_BUILTIN_VRLQNM, P10V_BUILTIN_VSLQ,
13727 P10V_BUILTIN_VSRQ, P10V_BUILTIN_VSRAQ,
13728 P10V_BUILTIN_VCMPGTUT_P, P10V_BUILTIN_VCMPGTST_P,
13729 P10V_BUILTIN_VCMPEQUT_P, P10V_BUILTIN_VCMPGTUT_P,
13730 P10V_BUILTIN_VCMPGTST_P, P10V_BUILTIN_CMPNET,
13731 P10V_BUILTIN_VCMPNET_P, P10V_BUILTIN_VCMPAET_P,
13732 P10V_BUILTIN_DIVES_V1TI, P10V_BUILTIN_MODS_V1TI,
13733 P10V_BUILTIN_MODU_V1TI):
13734 New overloaded definitions.
13735 (rs6000_gimple_fold_builtin) [P10V_BUILTIN_VCMPEQUT,
13736 P10V_BUILTIN_CMPNET, P10V_BUILTIN_CMPGE_1TI,
13737 P10V_BUILTIN_CMPGE_U1TI, P10V_BUILTIN_VCMPGTUT,
13738 P10V_BUILTIN_VCMPGTST, P10V_BUILTIN_CMPLE_1TI,
13739 P10V_BUILTIN_CMPLE_U1TI]: New case statements.
13740 (rs6000_init_builtins) [bool_V1TI_type_node, int_ftype_int_v1ti_v1ti]:
13742 (altivec_init_builtins): New E_V1TImode case statement.
13743 (builtin_function_type)[P10_BUILTIN_128BIT_VMULEUD,
13744 P10_BUILTIN_128BIT_VMULOUD, P10_BUILTIN_128BIT_DIVEU_V1TI,
13745 P10_BUILTIN_128BIT_MODU_V1TI, P10_BUILTIN_CMPGE_U1TI,
13746 P10_BUILTIN_VCMPGTUT, P10_BUILTIN_VCMPEQUT]: New case statements.
13747 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute) [E_TImode,
13748 E_V1TImode]: New case statements.
13749 * config/rs6000/rs6000.h (rs6000_builtin_type_index): New enum
13750 value RS6000_BTI_bool_V1TI.
13751 * config/rs6000/vector.md (vector_gtv1ti,vector_nltv1ti,
13752 vector_gtuv1ti, vector_nltuv1ti, vector_ngtv1ti, vector_ngtuv1ti,
13753 vector_eq_v1ti_p, vector_ne_v1ti_p, vector_ae_v1ti_p,
13754 vector_gt_v1ti_p, vector_gtu_v1ti_p, vrotlv1ti3, vashlv1ti3,
13755 vlshrv1ti3, vashrv1ti3): New define_expands.
13756 * config/rs6000/vsx.md (UNSPEC_VSX_DIVSQ, UNSPEC_VSX_DIVUQ,
13757 UNSPEC_VSX_DIVESQ, UNSPEC_VSX_DIVEUQ, UNSPEC_VSX_MODSQ,
13758 UNSPEC_VSX_MODUQ): New unspecs.
13759 (mulv2di3, vsx_div_v1ti, vsx_udiv_v1ti, vsx_dives_v1ti,
13760 vsx_diveu_v1ti, vsx_mods_v1ti, vsx_modu_v1ti, xxswapd_v1ti): New
13762 (vcmpnet): New define_expand.
13763 * doc/extend.texi: Add documentation for the new builtins vec_rl,
13764 vec_rlmi, vec_rlnm, vec_sl, vec_sr, vec_sra, vec_mule, vec_mulo,
13765 vec_div, vec_dive, vec_mod, vec_cmpeq, vec_cmpne, vec_cmpgt, vec_cmplt,
13766 vec_cmpge, vec_cmple, vec_all_eq, vec_all_ne, vec_all_gt, vec_all_lt,
13767 vec_all_ge, vec_all_le, vec_any_eq, vec_any_ne, vec_any_gt, vec_any_lt,
13768 vec_any_ge, vec_any_le.
13770 2021-06-09 Carl Love <cel@us.ibm.com>
13772 * config/rs6000/altivec.md (altivec_vrl<VI_char>mi): Fix
13773 bug in argument generation.
13775 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
13777 * config/arm/iterators.md (<supf>): Remove VCLZQ_U, VCLZQ_S.
13779 * config/arm/mve.md (mve_vclzq_<supf><mode>): Add '@' prefix,
13780 remove <supf> iterator.
13781 (mve_vclzq_u<mode>): New.
13782 * config/arm/neon.md (clz<mode>2): Rename to neon_vclz<mode>.
13783 (neon_vclz<mode): Move to ...
13784 * config/arm/unspecs.md (VCLZQ_U, VCLZQ_S): Remove.
13785 * config/arm/vec-common.md: ... here. Add support for MVE.
13787 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
13789 * config/arm/mve.md (mve_vhaddq_<supf><mode>): Prefix with '@'.
13790 (@mve_vrhaddq_<supf><mode): Likewise.
13791 * config/arm/neon.md (neon_v<r>hadd<sup><mode>): Likewise.
13792 * config/arm/vec-common.md (avg<mode>3_floor, uavg<mode>3_floor)
13793 (avg<mode>3_ceil", uavg<mode>3_ceil): New patterns.
13795 2021-06-09 imba-tjd <109224573@qq.com>
13797 * doc/invoke.texi: Fix typo.
13799 2021-06-09 Roger Sayle <roger@nextmovesoftware.com>
13801 PR middle-end/53267
13802 * fold-const-call.c (fold_const_call_sss) [CASE_CFN_FMOD]:
13803 Support evaluation of fmod/fmodf/fmodl at compile-time.
13805 2021-06-09 Richard Biener <rguenther@suse.de>
13807 PR tree-optimization/100981
13808 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
13809 gimple_get_lhs to also handle calls.
13810 * tree-vect-slp-patterns.c (complex_pattern::build): Transfer
13813 2021-06-09 Richard Biener <rguenther@suse.de>
13815 PR tree-optimization/97832
13816 * tree-vectorizer.h (_slp_tree::failed): New.
13817 * tree-vect-slp.c (_slp_tree::_slp_tree): Initialize
13819 (_slp_tree::~_slp_tree): Free failed.
13820 (vect_build_slp_tree): Retain failed nodes and record
13821 matches in them, copying that back out when running
13822 into a cached fail. Dump start and end of discovery.
13823 (dt_sort_cmp): New.
13824 (vect_build_slp_tree_2): Handle associatable chains
13825 together doing more aggressive operand swapping.
13827 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
13830 * config.gcc (gcc_cv_initfini_array): Set to yes for Linux and
13832 * doc/install.texi: Require glibc 2.1 and binutils 2.12 for
13833 Linux and GNU targets.
13835 2021-06-09 Richard Biener <rguenther@suse.de>
13837 * tree-vect-stmts.c (vect_is_simple_use): Always get dt
13840 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
13842 * config/arc/arc.md (loop_end): Change it to
13843 define_insn_and_split.
13845 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
13847 * config/arc/arc.md (maddhisi4): Use VMAC2H instruction.
13848 (machi): New pattern.
13849 (umaddhisi4): Use VMAC2HU instruction.
13850 (umachi): New pattern.
13852 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
13854 * config/arc/arc-protos.h (arc_split_move_p): New prototype.
13855 * config/arc/arc.c (arc_split_move_p): New function.
13856 (arc_split_move): Clean up.
13857 * config/arc/arc.md (movdi_insn): Clean up, use arc_split_move_p.
13858 (movdf_insn): Likewise.
13859 * config/arc/simdext.md (mov<VWH>_insn): Likewise.
13861 2021-06-09 Uroš Bizjak <ubizjak@gmail.com>
13864 * config/i386/i386.c (print_operand_address_as): Rename "no_rip"
13865 argument to "raw". Do not emit segment overrides when "raw" is true.
13867 2021-06-09 Martin Liska <mliska@suse.cz>
13869 * doc/gcov.texi: Create a proper JSON files.
13870 * doc/invoke.texi: Remove dots in order to make it a valid
13873 2021-06-09 Xionghu Luo <luoxhu@linux.ibm.com>
13875 * config/rs6000/rs6000-p8swap.c (pattern_is_rotate64): New.
13876 (insn_is_load_p): Use pattern_is_rotate64.
13877 (insn_is_swap_p): Likewise.
13878 (quad_aligned_load_p): Likewise.
13879 (const_load_sequence_p): Likewise.
13880 (replace_swapped_aligned_load): Likewise.
13881 (recombine_lvx_pattern): Likewise.
13882 (recombine_stvx_pattern): Likewise.
13884 2021-06-09 Andrew MacLeod <amacleod@redhat.com>
13886 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Use a
13887 fur_stmt source record.
13888 * gimple-range.cc (fur_source::get_operand): Generic range query.
13889 (fur_source::get_phi_operand): New.
13890 (fur_source::register_dependency): New.
13891 (fur_source::query): New.
13892 (class fur_edge): New. Edge source for operands.
13893 (fur_edge::fur_edge): New.
13894 (fur_edge::get_operand): New.
13895 (fur_edge::get_phi_operand): New.
13896 (fur_edge::query): New.
13897 (fur_stmt::fur_stmt): New.
13898 (fur_stmt::get_operand): New.
13899 (fur_stmt::get_phi_operand): New.
13900 (fur_stmt::query): New.
13901 (class fur_depend): New. Statement source and process dependencies.
13902 (fur_depend::fur_depend): New.
13903 (fur_depend::register_dependency): New.
13904 (class fur_list): New. List source for operands.
13905 (fur_list::fur_list): New.
13906 (fur_list::get_operand): New.
13907 (fur_list::get_phi_operand): New.
13908 (fold_range): New. Instantiate appropriate fur_source class and fold.
13909 (fold_using_range::range_of_range_op): Use new API.
13910 (fold_using_range::range_of_address): Ditto.
13911 (fold_using_range::range_of_phi): Ditto.
13912 (imple_ranger::fold_range_internal): Use fur_depend class.
13913 (fold_using_range::range_of_ssa_name_with_loop_info): Use new API.
13914 * gimple-range.h (class fur_source): Now a base class.
13915 (class fur_stmt): New.
13916 (fold_range): New prototypes.
13917 (fur_source::fur_source): Delete.
13919 2021-06-08 Andrew Pinski <apinski@marvell.com>
13921 PR tree-optimization/25290
13922 * tree-ssa-phiopt.c (xor_replacement): Delete.
13923 (tree_ssa_phiopt_worker): Delete use of xor_replacement.
13924 (match_simplify_replacement): Allow one cheap preparation
13925 statement that can be moved to before the if.
13927 2021-06-08 Pat Haugen <pthaugen@linux.ibm.com>
13929 * config/rs6000/power10.md (power10-fused-load, power10-fused-store,
13930 power10-fused_alu, power10-fused-vec, power10-fused-branch): New.
13932 2021-06-08 Jeff Law <jeffreyalaw@gmail.com>
13934 * config/h8300/logical.md (andqi3_1): Move BCLR case into define_insn_and_split.
13935 Create length attribute on define_insn_and_split. Only split for cases which we
13937 (andqi3_1<cczn>): Renamed from andqi3_1_clobber_flags. Only handle AND here and
13938 fix length computation.
13939 (b<code><mode>msx): Combine QImode and HImode H8/SX patterns using iterator.
13941 2021-06-08 Richard Biener <rguenther@suse.de>
13943 PR tree-optimization/100923
13944 * tree-ssa-sccvn.c (valueize_refs_1): Take a pointer to
13945 the operand vector to be valueized.
13946 (valueize_refs): Likewise.
13947 (valueize_shared_reference_ops_from_ref): Adjust.
13948 (valueize_shared_reference_ops_from_call): Likewise.
13949 (vn_reference_lookup_3): Likewise.
13950 (vn_reference_lookup_pieces): Likewise. Re-valueize
13951 with honoring availability when we are about to create
13952 the ao_ref and valueized before.
13953 (vn_reference_lookup): Likewise.
13954 (vn_reference_insert_pieces): Adjust.
13956 2021-06-08 Richard Biener <rguenther@suse.de>
13958 * tree-vectorizer.h (_slp_instance::root_stmt): Change to...
13959 (_slp_instance::root_stmts): ... a vector.
13960 (SLP_INSTANCE_ROOT_STMT): Rename to ...
13961 (SLP_INSTANCE_ROOT_STMTS): ... this.
13962 (slp_root::root): Change to...
13963 (slp_root::roots): ... a vector.
13964 (slp_root::slp_root): Adjust.
13965 * tree-vect-slp.c (_slp_instance::location): Adjust.
13966 (vect_free_slp_instance): Release the root stmt vector.
13967 (vect_build_slp_instance): Adjust.
13968 (vect_analyze_slp): Likewise.
13969 (_bb_vec_info::~_bb_vec_info): Likewise.
13970 (vect_slp_analyze_operations): Likewise.
13971 (vect_bb_vectorization_profitable_p): Likewise. Adjust
13972 costs for the root stmt.
13973 (vect_slp_check_for_constructors): Gather all BIT_INSERT_EXPRs
13975 (vect_slp_analyze_bb_1): Simplify by marking all root stmts
13977 (vectorize_slp_instance_root_stmt): Adjust.
13978 (vect_schedule_slp): Likewise.
13980 2021-06-08 Aldy Hernandez <aldyh@redhat.com>
13982 * gimple-ssa-evrp.c (class ssa_equiv_stack): New.
13983 (ssa_equiv_stack::ssa_equiv_stack): New.
13984 (ssa_equiv_stack::~ssa_equiv_stack): New.
13985 (ssa_equiv_stack::enter): New.
13986 (ssa_equiv_stack::leave): New.
13987 (ssa_equiv_stack::push_replacement): New.
13988 (ssa_equiv_stack::get_replacement): New.
13989 (is_pointer_ssa): New.
13990 (class pointer_equiv_analyzer): New.
13991 (pointer_equiv_analyzer::pointer_equiv_analyzer): New.
13992 (pointer_equiv_analyzer::~pointer_equiv_analyzer): New.
13993 (pointer_equiv_analyzer::set_global_equiv): New.
13994 (pointer_equiv_analyzer::set_cond_equiv): New.
13995 (pointer_equiv_analyzer::get_equiv): New.
13996 (pointer_equiv_analyzer::enter): New.
13997 (pointer_equiv_analyzer::leave): New.
13998 (pointer_equiv_analyzer::get_equiv_expr): New.
13999 (pta_valueize): New.
14000 (pointer_equiv_analyzer::visit_stmt): New.
14001 (pointer_equiv_analyzer::visit_edge): New.
14002 (hybrid_folder::value_of_expr): Call PTA.
14003 (hybrid_folder::value_on_edge): Same.
14004 (hybrid_folder::pre_fold_bb): New.
14005 (hybrid_folder::post_fold_bb): New.
14006 (hybrid_folder::pre_fold_stmt): New.
14007 (rvrp_folder::pre_fold_bb): New.
14008 (rvrp_folder::post_fold_bb): New.
14009 (rvrp_folder::pre_fold_stmt): New.
14010 (rvrp_folder::value_of_expr): Call PTA.
14011 (rvrp_folder::value_on_edge): Same.
14013 2021-06-08 Jakub Jelinek <jakub@redhat.com>
14016 * tree-inline.c (copy_tree_body_r): For OMP_CLAUSE_DEPEND don't
14017 check TREE_CODE if OMP_CLAUSE_DECL is NULL.
14019 2021-06-08 Richard Biener <rguenther@suse.de>
14021 PR middle-end/100951
14022 * tree-vect-generic.c (expand_vector_piecewise): Build a
14023 VECTOR_CST if all elements are constant.
14024 (expand_vector_condition): Likewise.
14025 (lower_vec_perm): Likewise.
14026 (expand_vector_conversion): Likewise.
14028 2021-06-08 Martin Liska <mliska@suse.cz>
14030 * doc/invoke.texi: Document new param evrp-sparse-threshold.
14032 2021-06-08 Martin Liska <mliska@suse.cz>
14034 * genautomata.c (create_automata): Fix typo.
14036 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
14038 PR tree-optimization/100794
14039 * tree-predcom.c (tree_predictive_commoning_loop): Add parameter
14040 allow_unroll_p and only allow unrolling when it's true.
14041 (tree_predictive_commoning): Add parameter allow_unroll_p and
14043 (run_tree_predictive_commoning): Likewise.
14044 (pass_predcom::gate): Check flag_tree_loop_vectorize and
14045 global_options_set.x_flag_predictive_commoning.
14046 (pass_predcom::execute): Adjust for allow_unroll_p.
14048 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
14050 * tree-predcom.c (execute_pred_commoning): Remove update_ssa call.
14051 (tree_predictive_commoning_loop): Factor some cleanup stuffs into
14052 lambda function cleanup, remove scev_reset call, and adjust return
14054 (tree_predictive_commoning): Adjust for different changed values,
14055 only set flag TODO_update_ssa_only_virtuals if changed.
14056 (pass_data pass_data_predcom): Remove TODO_update_ssa_only_virtuals
14057 from todo_flags_finish.
14059 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
14061 * gimple-range-cache.cc (class sbr_sparse_bitmap): New.
14062 (sbr_sparse_bitmap::sbr_sparse_bitmap): New.
14063 (sbr_sparse_bitmap::bitmap_set_quad): New.
14064 (sbr_sparse_bitmap::bitmap_get_quad): New.
14065 (sbr_sparse_bitmap::set_bb_range): New.
14066 (sbr_sparse_bitmap::get_bb_range): New.
14067 (sbr_sparse_bitmap::bb_range_p): New.
14068 (block_range_cache::block_range_cache): initialize bitmap obstack.
14069 (block_range_cache::~block_range_cache): Destruct obstack.
14070 (block_range_cache::set_bb_range): Decide when to utilze the
14071 sparse on entry cache.
14072 * gimple-range-cache.h (block_range_cache): Add bitmap obstack.
14073 * params.opt (-param=evrp-sparse-threshold): New.
14075 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
14077 * bitmap.c (bitmap_set_aligned_chunk): New.
14078 (bitmap_get_aligned_chunk): New.
14079 (test_aligned_chunk): New.
14080 (bitmap_c_tests): Call test_aligned_chunk.
14081 * bitmap.h (bitmap_set_aligned_chunk, bitmap_get_aligned_chunk): New.
14083 2021-06-07 Uroš Bizjak <ubizjak@gmail.com>
14086 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
14088 (ix86_expand_vector_init_one_nonzero): Ditto.
14089 (ix86_expand_vector_init_one_var): Ditto.
14090 (ix86_expand_vector_init_general): Ditto.
14091 * config/i386/mmx.md (vec_initv4qiqi): New expander.
14093 2021-06-07 Jeff Law <jeffreyalaw@gmail.com>
14095 * config/h8300/movepush.md: Change most _clobber_flags
14096 patterns to instead use <cczn> subst.
14097 (movsi_cczn): New pattern with usable CC cases split out.
14098 (movsi_h8sx_cczn): Likewise.
14100 2021-06-07 Martin Liska <mliska@suse.cz>
14102 * common/common-target.def: Split long lines and replace them
14104 * target.def: Likewise.
14105 * doc/tm.texi: Re-generated.
14107 2021-06-07 Jakub Jelinek <jakub@redhat.com>
14110 * fold-const.c (fold_read_from_vector): Return NULL if trying to
14111 read from a CONSTRUCTOR with vector type elements.
14113 2021-06-07 Jakub Jelinek <jakub@redhat.com>
14115 PR middle-end/100898
14116 * tree-inline.c (copy_bb): Only use gimple_call_arg_ptr if memcpy
14117 should copy any arguments. Don't call gimple_call_num_args
14118 on id->call_stmt or call_stmt more than once.
14120 2021-06-07 liuhongt <hongtao.liu@intel.com>
14123 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): Refine
14125 (<insn>v4siv4di2): Delete constraints for define_expand.
14127 2021-06-07 liuhongt <hongtao.liu@intel.com>
14130 * config/i386/i386-expand.c (ix86_expand_builtin): Remove
14131 assignment of cfun->machine->has_explicit_vzeroupper.
14132 * config/i386/i386-features.c
14133 (ix86_add_reg_usage_to_vzerouppers): Delete.
14134 (ix86_add_reg_usage_to_vzeroupper): Ditto.
14135 (rest_of_handle_insert_vzeroupper): Remove
14136 ix86_add_reg_usage_to_vzerouppers, add df_analyze at the end
14138 (gate): Remove cfun->machine->has_explicit_vzeroupper.
14139 * config/i386/i386-protos.h (ix86_expand_avx_vzeroupper):
14141 * config/i386/i386.c (ix86_insn_callee_abi): New function.
14142 (ix86_initialize_callee_abi): Ditto.
14143 (ix86_expand_avx_vzeroupper): Ditto.
14144 (ix86_hard_regno_call_part_clobbered): Adjust for vzeroupper
14146 (TARGET_INSN_CALLEE_ABI): Define as ix86_insn_callee_abi.
14147 (ix86_emit_mode_set): Call ix86_expand_avx_vzeroupper
14149 * config/i386/i386.h (struct GTY(()) machine_function): Delete
14150 has_explicit_vzeroupper.
14151 * config/i386/i386.md (enum unspec): New member
14153 (ABI_DEFAULT,ABI_VZEROUPPER,ABI_UNKNOWN): New
14154 define_constants for insn callee abi index.
14155 * config/i386/predicates.md (vzeroupper_pattern): Adjust.
14156 * config/i386/sse.md (UNSPECV_VZEROUPPER): Deleted.
14157 (avx_vzeroupper): Call ix86_expand_avx_vzeroupper.
14158 (*avx_vzeroupper): Rename to ..
14159 (avx_vzeroupper_callee_abi): .. this, and adjust pattern as
14160 call_insn which has a special vzeroupper ABI.
14161 (*avx_vzeroupper_1): Deleted.
14163 2021-06-07 liuhongt <hongtao.liu@intel.com>
14166 * df-scan.c (df_get_call_refs): When call_insn is a fake call,
14167 it won't use stack pointer reg.
14168 * final.c (leaf_function_p): When call_insn is a fake call, it
14169 won't affect caller as a leaf function.
14170 * reg-stack.c (callee_clobbers_any_stack_reg): New.
14171 (subst_stack_regs): When call_insn doesn't clobber any stack
14172 reg, don't clear the arguments.
14173 * rtl.c (shallow_copy_rtx): Don't clear flag used when orig is
14175 * shrink-wrap.c (requires_stack_frame_p): No need for stack
14176 frame for a fake call.
14177 * rtl.h (FAKE_CALL_P): New macro.
14179 2021-06-06 Eric Botcazou <ebotcazou@adacore.com>
14181 * config/sparc/sparc-protos.h (order_regs_for_local_alloc): Rename
14183 (sparc_order_regs_for_local_alloc): ...this.
14184 (sparc_leaf_reg_remap): Declare.
14185 * config/sparc/sparc.h (ADJUST_REG_ALLOC_ORDER): Adjust.
14186 (LEAF_REG_REMAP): Reimplement as call to sparc_leaf_reg_remap.
14187 * config/sparc/sparc.c (leaf_reg_remap): Delete.
14188 (order_regs_for_local_alloc): Rename to...
14189 (sparc_order_regs_for_local_alloc): ...this.
14190 (sparc_leaf_reg_remap): New function.
14191 (sparc_conditional_register_usage): Do not modify leaf_reg_remap.
14193 2021-06-06 David Edelsohn <dje.gcc@gmail.com>
14195 * config/rs6000/rs6000.c (rs6000_xcoff_asm_output_aligned_decl_common):
14196 Use assemble_name to output BSS section name.
14198 2021-06-06 Uroš Bizjak <ubizjak@gmail.com>
14200 * config/i386/constraints.md (Bs):
14201 Remove boolean operators from match_test RTX.
14204 (M): Use "mode" variable instead of GET_MODE (op) in match_test RTX.
14207 2021-06-06 Martin Liska <mliska@suse.cz>
14209 * doc/extend.texi: Add missing @headitem.
14210 * doc/invoke.texi: Likewise.
14211 * doc/objc.texi: Likewise.
14213 2021-06-06 Martin Liska <mliska@suse.cz>
14215 * genhooks.c (emit_findices): Remove unused function.
14216 (emit_documentation): Do not call emit_findices
14217 and do not search for @Fcode directives.
14219 2021-06-06 Martin Liska <mliska@suse.cz>
14221 * doc/invoke.texi: Remove extra character.
14223 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
14225 * config/sh/sh.md (doloop_end_split): Fix empty split condition.
14227 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
14229 * config/sparc/sparc.md (*snedi<W:mode>_zero_vis3,
14230 *neg_snedi<W:mode>_zero_subxc, *plus_snedi<W:mode>_zero,
14231 *plus_plus_snedi<W:mode>_zero, *minus_snedi<W:mode>_zero,
14232 *minus_minus_snedi<W:mode>_zero): Fix empty split condition.
14234 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
14236 * config/or1k/or1k.md (*movdi): Fix empty split condition.
14238 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
14240 * config/mips/mips.md (<anonymous>, bswapsi2, bswapdi2): Fix empty
14243 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
14245 * config/m68k/m68k.md (*zero_extend_inc, *zero_extend_dec,
14246 *zero_extendsidi2): Fix empty split condition.
14248 2021-06-05 Jeff Law <jeffreyalaw@gmail.com>
14250 * config/h8300/addsub.md: Fix split condition in define_insn_and_split
14252 * config/h8300/bitfield.md: Likewise.
14253 * config/h8300/combiner.md: Likewise.
14254 * config/h8300/divmod.md: Likewise.
14255 * config/h8300/extensions.md: Likewise.
14256 * config/h8300/jumpcall.md: Likewise.
14257 * config/h8300/movepush.md: Likewise.
14258 * config/h8300/multiply.md: Likewise.
14259 * config/h8300/other.md: Likewise.
14260 * config/h8300/shiftrotate.md: Likewise.
14261 * config/h8300/logical.md: Likewise. Fix split pattern to use
14262 code iterator that somehow slipped through.
14264 2021-06-04 Tobias Burnus <tobias@codesourcery.com>
14266 PR middle-end/100905
14267 * tree-nested.c (convert_nonlocal_omp_clauses,
14268 convert_local_omp_clauses): Handle OMP_CLAUSE_BIND.
14270 2021-06-04 Martin Sebor <msebor@redhat.com>
14272 PR middle-end/100732
14273 * gimple-fold.c (gimple_fold_builtin_sprintf): Avoid folding calls
14274 with either source or destination argument of invalid type.
14275 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Avoid checking
14276 calls with arguments of invalid type.
14278 2021-06-04 Martin Sebor <msebor@redhat.com>
14280 * attribs.c (init_attr_rdwr_indices): Use VLA bounds in the expected
14282 (attr_access::vla_bounds): Also handle VLA bounds.
14284 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
14286 * config/i386/predicates.md (GOT_memory_operand):
14287 Implement using match_code RTXes.
14288 (GOT32_symbol_operand): Ditto.
14290 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
14293 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
14295 (ix86_expand_vector_init_general): Ditto.
14296 Use SImode instead of word_mode for logic operations
14297 when GET_MODE_SIZE (mode) < UNITS_PER_WORD.
14298 (expand_vec_perm_even_odd_1): Assert that V2HI mode should be
14299 implemented by expand_vec_perm_1.
14300 (expand_vec_perm_broadcast_1): Assert that V2HI and V4HI modes
14301 should be implemented using standard shuffle patterns.
14302 (ix86_vectorize_vec_perm_const): Handle V2HImode. Add V4HI and
14303 V2HI modes to modes, implementable with shuffle for one operand.
14304 * config/i386/mmx.md (*punpckwd): New insn_and_split pattern.
14305 (*pshufw_1): New insn pattern.
14306 (*vec_dupv2hi): Ditto.
14307 (vec_initv2hihi): New expander.
14309 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
14311 * config/arm/vfp.md (no_literal_pool_df_immediate,
14312 no_literal_pool_sf_immediate): Fix empty split condition.
14314 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
14316 * config/i386/i386.md (*load_tp_x32_zext, *add_tp_x32_zext,
14317 *tls_dynamic_gnu2_combine_32): Fix empty split condition.
14318 * config/i386/sse.md (*<sse2_avx2>_pmovmskb_lt,
14319 *<sse2_avx2>_pmovmskb_zext_lt, *sse2_pmovmskb_ext_lt,
14320 *<sse4_1_avx2>_pblendvb_lt): Likewise.
14322 2021-06-04 Jakub Jelinek <jakub@redhat.com>
14325 * config/i386/i386-expand.c (ix86_expand_vector_init): Handle
14326 concatenation from half-sized modes with TImode elements.
14328 2021-06-04 Claudiu Zissulescu <claziss@synopsys.com>
14330 * config/arc/arc.c (arc_override_options): Disable millicode
14331 thunks when RF16 is on.
14333 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
14335 * config/rs6000/rs6000.h (PROMOTE_MODE): Remove.
14337 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
14339 * config/rs6000/rs6000-call.c (rs6000_promote_function_mode):
14340 Replace PROMOTE_MODE marco with its content.
14342 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
14344 * config/cris/cris.md (*addi_reload): Fix empty split condition.
14346 2021-06-03 Jim Wilson <jimw@sifive.com>
14348 * config.gcc (riscv*-*-*): If --with-riscv-attribute not used,
14349 turn it on for all riscv targets.
14351 2021-06-03 Uroš Bizjak <ubizjak@gmail.com>
14354 * config/i386/i386-expand.c (ix86_expand_vector_set):
14355 Handle V2HI and V4QI modes.
14356 (ix86_expand_vector_extract): Ditto.
14357 * config/i386/mmx.md (*pinsrw): New insn pattern.
14360 (*pextrw_zext): Ditto.
14362 (*pextrb_zext): Ditto.
14363 (vec_setv2hi): New expander.
14364 (vec_extractv2hihi): Ditto.
14365 (vec_setv4qi): Ditto.
14366 (vec_extractv4qiqi): Ditto.
14367 (vec_setv8qi): Enable only for TARGET_SSE4_1.
14368 (vec_extractv8qiqi): Ditto.
14370 2021-06-03 Aaron Sawdey <acsawdey@linux.ibm.com>
14372 * config/rs6000/genfusion.pl (gen_logical_addsubf): Fix input
14373 order to subf instruction.
14374 * config/rs6000/fusion.md: Regenerate.
14376 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
14378 * calls.c (get_size_range): Use range_of_expr instead of
14379 determine_value_range.
14380 * tree-affine.c (expr_to_aff_combination): Same.
14381 * tree-data-ref.c (split_constant_offset): Same.
14382 * tree-vrp.c (determine_value_range_1): Remove.
14383 (determine_value_range): Remove.
14384 * tree-vrp.h (determine_value_range): Remove.
14386 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
14388 * function-tests.c (test_ranges): Call gimple_range_tests.
14389 * gimple-range-cache.cc (ranger_cache::range_of_expr): Pass stmt
14391 * gimple-range.cc (fur_source::get_operand): Do not call
14392 get_tree_range or gimple_range_global.
14394 (get_tree_range): Move to value-query.cc.
14395 Call get_arith_expr_range.
14396 (gimple_ranger::range_of_expr): Add argument to get_tree_range.
14397 Include gimple-range-tests.cc.
14398 * gimple-range.h (fold_range): Add argument.
14399 (get_tree_range): Remove.
14400 * selftest.h (gimple_range_tests): New.
14401 * value-query.cc (global_range_query::range_of_expr): Add
14403 (range_query::get_tree_range): Move from gimple-range.cc.
14404 * value-query.h (class range_query): Add get_tree_range and
14405 get_arith_expr_range. Make fur_source a friend.
14406 * vr-values.c (vr_values::range_of_expr): Pass stmt to
14408 * gimple-range-tests.cc: New file.
14410 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
14412 * gimple-range.cc (gimple_ranger::export_global_ranges): Call
14413 update_global_range.
14414 * value-query.cc (update_global_range): New.
14415 * value-query.h (update_global_range): New.
14417 2021-06-03 David Malcolm <dmalcolm@redhat.com>
14419 * diagnostic-show-locus.c (diagnostic_show_locus): Don't reject
14420 printing the same location twice if there are fix-it hints,
14421 multiple locations, or a label.
14423 2021-06-03 Andre Vieira <andre.simoesdiasvieira@arm.com>
14425 * tree-vect-loop.c (vect_transform_loop): Use main loop's various'
14426 thresholds to narrow the upper bound on epilogue iterations.
14428 2021-06-03 Christophe Lyon <christophe.lyon@linaro.org>
14430 * config/arm/mve.md (mve_vabsq_f<mode>): Use 'abs' instead of unspec.
14431 (mve_vabsq_s<mode>): Likewise.
14432 * config/arm/neon.md (abs<mode>2): Rename to neon_abs<mode>2.
14433 * config/arm/unspecs.md (VABSQ_F, VABSQ_S): Delete.
14434 * config/arm/vec-common.md (neg<mode>2): Rename to
14435 <absneg_str><mode>2.
14437 2021-06-03 Claudiu Zissulescu <claziss@synopsys.com>
14439 * common/config/arc/arc-common.c (arc_option_optimization_table):
14440 Remove malign-call.
14441 * config/arc/arc.c (arc_unalign_branch_p): Remove unused function.
14442 * config/arc/arc.h (TARGET_MIXED_CODE): Remove macro.
14443 (INDEX_REG_CLASS): Only refer to GENERAL_REGS.
14444 * config/arc/arc.md (abssi2_mixed): Remove pattern.
14445 * config/arc/arc.opt (munalign-prob-threshold): Mark it obsolete.
14446 (malign-call): Likewise.
14447 (mmixed-code): Likewise.
14448 * doc/invoke.texi (ARC): Update doc.
14450 2021-06-03 Martin Liska <mliska@suse.cz>
14452 * common.opt: Use proper Enum values.
14453 * opts.c (COVERAGE_SANITIZER_OPT): Remove.
14454 (parse_sanitizer_options): Handle only sanitizer_opts.
14455 (common_handle_option): Just assign value.
14457 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
14460 * tree-inline.c (inline_forbidden_p): Remove test on return type.
14462 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
14464 * dwarf2out.c (loc_list_from_tree_1) <FUNCTION_DECL>: Also generate
14465 DW_OP_GNU_variable_value referencing an existing DIE at file scope.
14466 (type_byte_size): Inline into...
14467 (add_byte_size_attribute): ...this and call add_scalar_info.
14469 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
14471 * dwarf2out.c (mem_loc_descriptor) <UDIV>: Fix typo.
14472 (typed_binop_from_tree): New function.
14473 (loc_list_from_tree_1) <EXACT_DIV_EXPR>: For an unsigned type,
14474 turn a divide by a power of 2 into a shift.
14475 <CEIL_DIV_EXPR>: For an unsigned type, use a signed divide if the
14476 size of the mode is lower than DWARF2_ADDR_SIZE; otherwise, do a
14477 typed divide by calling typed_binop_from_tree.
14479 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
14481 * dwarf2out.c (scompare_loc_descriptor): Fix head comment.
14482 (is_handled_procedure_type): Likewise.
14483 (struct loc_descr_context): Add strict_signedness field.
14484 (resolve_args_picking_1): Deal with DW_OP_[GNU_]deref_type,
14485 DW_OP_[GNU_]convert and DW_OP_[GNU_]reinterpret.
14486 (resolve_args_picking): Minor tweak.
14487 (function_to_dwarf_procedure): Initialize strict_signedness field.
14488 (type_byte_size): Likewise.
14489 (field_byte_offset): Likewise.
14490 (gen_descr_array_type_die): Likewise.
14491 (gen_variant_part): Likewise.
14492 (loc_list_from_tree_1) <CALL_EXPR>: Tidy up and set strict_signedness
14493 to true when a context is present before evaluating the arguments.
14494 <COND_EXPR>: Do not generate a useless comparison with zero.
14495 When dereferencing an address, if strict_signedness is true and the
14496 type is small and signed, use DW_OP_deref_type to do the dereference
14497 and then DW_OP_convert to convert back to the generic type.
14499 2021-06-03 Jakub Jelinek <jakub@redhat.com>
14502 * tree-inline.c (copy_tree_body_r): Handle iterators on
14503 OMP_CLAUSE_AFFINITY or OMP_CLAUSE_DEPEND.
14505 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
14507 * config/arc/arc.md (*bbit_di): Remove.
14509 2021-06-02 Christoph Muellner <cmuellner@gcc.gnu.org>
14511 PR rtl-optimization/100264
14512 * ree.c (get_sub_rtx): Ignore SET expressions without register
14513 destinations and remove assertion, as it is not valid anymore
14514 with this new behaviour.
14515 (merge_def_and_ext): Eliminate destination check for register
14516 as such SET expressions can't occur anymore.
14517 (combine_reaching_defs): Likewise.
14519 2021-06-02 Jakub Jelinek <jakub@redhat.com>
14522 * config/xtensa/xtensa.h (LEAF_REG_REMAP): Cast REGNO to int to avoid
14523 -Wtype-limits warnings.
14524 (DWARF_FRAME_REGISTER): Rewrite into ternary operator with addition
14525 in operands to avoid -Wsign-compare warnings.
14527 2021-06-02 Pat Haugen <pthaugen@linux.ibm.com>
14529 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): Use
14532 2021-06-02 Vineet Gupta <vgupta@synopsys.com>
14534 * config/arc/arc.h (TARGET_CPU_DEFAULT): Change to hs38_linux.
14536 2021-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
14538 * config/s390/s390.md(*ashrdi3_31<setcc><cconly>): Use a single
14540 * config/s390/subst.md(cconly_subst): Use a single constraint
14541 in (match_scratch).
14543 2021-06-02 Martin Liska <mliska@suse.cz>
14545 * ipa-icf.h: Use auto_vec for memory_access_types.
14547 2021-06-02 Jeff Law <jeffreyalaw@gmail.com>
14549 * config/h8300/h8300-protos.h (compute_a_shift_length): Drop unused
14550 argument from prototype.
14551 (output_logical_op): Add rtx_code argument.
14552 (compute_logical_op_length): Likewise.
14553 * config/h8300/h8300.c (h8300_and_costs): Pass additional argument
14554 to compute_a_shift_length.
14555 (output_logical_op); New argument with the rtx code rather than
14556 extracting it from an operand. Handle QImode too.
14557 (compute_logical_op_length): Similary.
14558 (compute_a_shift_length): Drop unused argument.
14559 * config/h8300/h8300.md (logicals): New code iterator.
14560 * config/h8300/logical.md (<code><mode>3 expander): Combine
14561 the "and" expander with the "ior"/"xor" expander.
14562 (bclr<mode>msx): Combine the QI/HI mode patterns.
14563 (<logical><mode>3 insns): Use code iterator rather than match_operator.
14564 Handle QImode as well. Update call to output_logical_op and
14565 compute_logical_op_length to pass in rtx_code
14566 Fix split condition on all define_insn_and_split patterns.
14567 (one_cmpl<mode>2<cczn>): Use <cczn> to support both clobbering
14568 the flags and setting ZN via existing define_subst.
14569 * config/h8300/shiftrotate.md: Drop unused argument from
14570 calls to compute_a_shift_length.
14571 Signed-off-by: Jeff Law <jeffreyalaw@gmail.com>
14573 2021-06-01 Andrew Pinski <apinski@marvell.com>
14575 PR tree-optimization/25290
14576 * tree-ssa-phiopt.c (match_simplify_replacement):
14578 (tree_ssa_phiopt_worker): Use match_simplify_replacement.
14579 (two_value_replacement): Change the comment about
14580 conditional_replacement.
14581 (conditional_replacement): Delete.
14583 2021-06-01 Andrew Pinski <apinski@marvell.com>
14585 PR tree-optimization/95481
14586 * tree-tailcall.c (find_tail_calls): Handle empty typed
14589 2021-06-01 Andrew Pinski <apinski@marvell.com>
14591 * gimplify.c (zero_sized_field_decl): Delete
14592 (zero_sized_type): Delete
14593 (gimplify_init_ctor_eval): Use is_empty_type instead
14594 of zero_sized_field_decl.
14595 (gimplify_modify_expr): Use is_empty_type instead of
14598 2021-06-01 Jason Merrill <jason@redhat.com>
14601 * tree.h (CALL_FROM_NEW_OR_DELETE_P): Adjust comment.
14603 2021-06-01 Jason Merrill <jason@redhat.com>
14606 * diagnostic.h (warning_enabled_at): Declare.
14607 * diagnostic.c (diagnostic_enabled): Factor out from...
14608 (diagnostic_report_diagnostic): ...here.
14609 (warning_enabled_at): New.
14611 2021-06-01 Aldy Hernandez <aldyh@redhat.com>
14613 * gimple-ssa-evrp.c: Enable exporting of global ranges.
14615 2021-06-01 Martin Liska <mliska@suse.cz>
14618 * doc/invoke.texi: Mention that -fgcse-after-reload
14619 is enabled with -O3.
14621 2021-06-01 liuhongt <hongtao.liu@intel.com>
14623 PR tree-optimization/98365
14624 * tree-if-conv.c (strip_nop_cond_scalar_reduction): New function.
14625 (is_cond_scalar_reduction): Handle nop_expr in cond scalar reduction.
14626 (convert_scalar_cond_reduction): Ditto.
14627 (predicate_scalar_phi): Ditto.
14629 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
14631 PR tree-optimization/100781
14632 * gimple-range-cache.cc (ranger_cache::ranger_cache): Enable new
14633 value calculation by default.
14634 (ranger_cache::enable_new_values): New.
14635 (ranger_cache::disable_new_values): New.
14636 (ranger_cache::push_poor_value): Check if new values are allowed.
14637 * gimple-range-cache.h (class ranger_cache): New member/methods.
14638 * gimple-range.cc (gimple_ranger::range_of_expr): Check for debug
14639 statement, and disable/renable new value calculation.
14641 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
14643 * gimple-range-cache.cc (ranger_cache::ssa_range_in_bb): Delete.
14644 (ranger_cache::range_of_def): New.
14645 (ranger_cache::entry_range): New.
14646 (ranger_cache::exit_range): New.
14647 (ranger_cache::range_of_expr): Adjust.
14648 (ranger_cache::range_on_edge): Adjust.
14649 (ranger_cache::propagate_cache): Call exit_range directly.
14650 * gimple-range-cache.h (class ranger_cache): Adjust.
14652 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
14654 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust for
14655 gori_compute being a member rather than base class.
14656 dervied call to member call.
14657 (ranger_cache::dump): No longer dump gori_map.
14658 (ranger_cache::dump_bb): New.
14659 (ranger_cache::get_non_stale_global_range): Adjust for gori_compute
14660 being a member rather than base class.
14661 (ranger_cache::set_global_range): Ditto.
14662 (ranger_cache::ssa_range_in_bb): Ditto.
14663 (ranger_cache::range_of_expr): New.
14664 (ranger_cache::range_on_edge): New.
14665 (ranger_cache::block_range): Adjust for gori_computes. Debug changes.
14666 (ranger_cache::propagate_cache): Adjust debugging output.
14667 (ranger_cache::fill_block_cache): Adjust for gori_computes. Debug
14669 * gimple-range-cache.h (class ranger_cache): Make gori_compute a
14670 member, and inherit from range_query instead.
14671 (ranger_cache::dump_bb): New. split from dump.
14672 * gimple-range-gori.cc (gori_compute::ssa_range_in_bb): Delete.
14673 (gori_compute::expr_range_at_stmt): Delete.
14674 (gori_compute::compute_name_range_op): Delete.
14675 (gori_compute::compute_operand_range_switch): Add fur_source.
14676 (gori_compute::compute_operand_range): Add fur_source param, inline
14677 old compute_name_range_op and optimize_logical_operands.
14678 (struct tf_range): Delete.
14679 (gori_compute::logical_combine): Adjust
14680 (gori_compute::optimize_logical_operands): Delete.
14681 (gori_compute::compute_logical_operands_in_chain): Delete.
14682 (gori_compute::compute_logical_operands): Adjust.
14683 (gori_compute::compute_operand1_range): Adjust to fur_source.
14684 (gori_compute::compute_operand2_range): Ditto.
14685 (gori_compute::compute_operand1_and_operand2_range): Ditto.
14686 (gori_compute::outgoing_edge_range_p): Add range_query parameter,
14687 and adjust to fur_source.
14688 * gimple-range-gori.h (class gori_compute): Simplify and adjust to
14689 range_query and fur_source.
14690 * gimple-range.cc (gimple_ranger::range_on_edge): Query range_on_edge
14691 from the ranger_cache..
14692 (gimple_ranger::fold_range_internal): Adjust to base class change of
14694 (gimple_ranger::dump_bb): Adjust dump.
14695 * gimple-range.h (gimple_ranger):export gori computes object.
14697 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
14699 PR tree-optimization/100774
14700 * gimple-range-cache.cc (ranger_cache::get_non_stale_global_range):
14701 Constant values are also not stale.
14702 (ranger_cache::set_global_range): Range invariant values should also
14703 have the correct timestamp.
14705 2021-05-31 Martin Liska <mliska@suse.cz>
14707 * tree-streamer-in.c (unpack_ts_function_decl_value_fields):
14708 Unpack FUNCTION_DECL_DECL_TYPE.
14709 * tree-streamer-out.c (pack_ts_function_decl_value_fields):
14710 Stream FUNCTION_DECL_DECL_TYPE instead of
14711 DECL_IS_OPERATOR_NEW_P.
14712 * tree.h (set_function_decl_type): Use FUNCTION_DECL_DECL_TYPE
14714 (DECL_IS_OPERATOR_NEW_P): Likewise.
14715 (DECL_IS_OPERATOR_DELETE_P): Likewise.
14716 (DECL_LAMBDA_FUNCTION_P): Likewise.
14718 2021-05-31 Richard Biener <rguenther@suse.de>
14721 * internal-fn.c (expand_SHUFFLEVECTOR): Define.
14722 * internal-fn.def (SHUFFLEVECTOR): New.
14723 * internal-fn.h (expand_SHUFFLEVECTOR): Declare.
14724 * doc/extend.texi: Document __builtin_shufflevector.
14726 2021-05-31 Peter Bergner <bergner@linux.ibm.com>
14729 * config/rs6000/predicates.md(mma_assemble_input_operand): Allow
14730 indexed form addresses.
14732 2021-05-29 Jeff Law <jlaw@tachyum.com>
14734 * config/h8300/h8300.c (h8300_emit_stack_adjustment): Drop unused
14735 parameter. Call callers fixed.
14737 (output_plussi): Add FALLTHRU markers.
14738 (h8300_shift_needs_scratch_p): Add gcc_unreachable marker.
14740 2021-05-29 Jakub Jelinek <jakub@redhat.com>
14742 PR middle-end/99928
14743 * gimplify.c (gimplify_scan_omp_clauses): For taskloop simd
14744 combined with parallel, make sure to add shared clause to
14745 parallel for explicit linear clause.
14747 2021-05-29 Aldy Hernandez <aldyh@redhat.com>
14749 PR tree-optimization/100787
14750 * gimple-ssa-evrp.c: Disable exporting of global ranges.
14752 2021-05-28 Jason Merrill <jason@redhat.com>
14754 * tree-iterator.h (struct tree_stmt_iterator): Add operator++,
14755 operator--, operator*, operator==, and operator!=.
14756 (class tsi_range): New.
14758 2021-05-28 Richard Biener <rguenther@suse.de>
14760 PR tree-optimization/100778
14761 * tree-vect-slp.c (vect_build_slp_tree_1): Prevent possibly
14762 trapping ops in different BBs.
14764 2021-05-28 Richard Biener <rguenther@suse.de>
14767 * tree-inline.c (copy_bb): When processing __builtin_va_arg_pack
14768 copy fntype from original call.
14770 2021-05-28 Martin Liska <mliska@suse.cz>
14772 PR gcov-profile/100751
14773 * doc/gcov.texi: Revert partially a hunk that was wrong.
14775 2021-05-28 Cooper Qu <cooper.qu@linux.alibaba.com>
14777 * config/csky/csky-linux-elf.h (HAVE_sync_compare_and_swapqi):
14779 (HAVE_sync_compare_and_swaphi): Likewise.
14780 (HAVE_sync_compare_and_swapsi): Likewise.
14782 2021-05-28 Jakub Jelinek <jakub@redhat.com>
14784 PR middle-end/99928
14785 * tree.h (OMP_CLAUSE_MAP_IMPLICIT): Define.
14787 2021-05-28 Tobias Burnus <tobias@codesourcery.com>
14789 * gimplify.c (gimplify_omp_affinity): New.
14790 (gimplify_scan_omp_clauses): Call it; remove affinity clause afterwards.
14791 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_AFFINITY.
14792 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_AFFINITY.
14793 * tree.c (omp_clause_num_ops, omp_clause_code_name): Add clause.
14794 (walk_tree_1): Handle OMP_CLAUSE_AFFINITY.
14796 2021-05-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
14797 Richard Biener <rguenther@suse.de>
14799 * match.pd <popcount & / + pattern matching>:
14800 When generating popcount directly fails, try doing it in two halves.
14802 2021-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
14804 * Makefile.in (generated_files): Add gimple-match.c and
14807 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
14809 * gensupport.c (alter_predicate_for_insn): Handle MATCH_DUP.
14811 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
14813 * gensupport.c (alter_constraints): Add MATCH_SCRATCH case.
14815 2021-05-28 Kewen Lin <linkw@linux.ibm.com>
14817 PR tree-optimization/99398
14818 * tree-ssa-forwprop.c (simplify_permutation): Optimize some cases
14819 where the fed operands are CTOR/CST and propagated through
14820 VIEW_CONVERT_EXPR. Call vec_perm_indices::new_shrunk_vector.
14821 * vec-perm-indices.c (vec_perm_indices::new_shrunk_vector): New
14823 * vec-perm-indices.h (vec_perm_indices::new_shrunk_vector): New
14826 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
14828 * config/i386/mmx.md (addv2sf3): Do not call
14829 ix86_fixup_binary_operands_no_copy.
14832 (<smaxmin:code>v2sf3): Ditto.
14833 (<plusminus:insn><MMXMODEI:mode>3): Ditto.
14834 (<plusminus:insn><VI_32:mode>3): Remove expander.
14835 (<plusminus:insn><VI_32:mode>3): Rename from
14836 "*<plusminus:insn><VI_32:mode>3".
14837 (mulv4hi): Do not call ix86_fixup_binary_operands_no_copy.
14838 (mulv2hi3): Remove expander.
14839 (mulv2hi3): Rename from *mulv2hi3.
14840 (<s>mulv2hi3_highpart): Remove expander.
14841 (<s>mulv2hi3_highpart): Rename from *<s>mulv2hi3_highpart.
14842 (<smaxmin:code><MMXMODE14:mode>3): Rename from
14843 "*mmx_<smaxmin:code><MMXMODE14:mode>3".
14844 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Remove expander.
14845 (SMAXMIN_MMXMODEI): Remove mode iterator.
14846 (<smaxmin:code>v4hi3): New expander.
14847 (<smaxmin:code>v4qi3): Rename from *<smaxmin:code>v4qi3.
14848 (<smaxmin:code>v2hi3): Rename from *<smaxmin:code>v2hi3.
14849 (<smaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
14850 (SMAXMIN_VI_32): Remove mode iterator.
14851 (<umaxmin:code><MMXMODE24:mode>3): Rename from
14852 "*mmx_<umaxmin:code><MMXMODE24:mode>3".
14853 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Remove expander.
14854 (UMAXMIN_MMXMODEI): Remove mode iterator.
14855 (<umaxmin:code>v8qi3): New expander.
14856 (<umaxmin:code>v4qi3): Rename from *<umaxmin:code>v4qi3.
14857 (<umaxmin:code>v2hi3): Rename from *<umaxmin:code>v2hi3.
14858 (<umaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
14859 (UMAXMIN_VI_32): Remove mode iterator.
14860 (<any_shift:insn>v2hi3): Remove expander.
14861 (<any_shift:insn>v2hi3): Rename from *<any_shift:insn>v2hi3.
14862 (<any_logic:code><MMXMODEI:mode>3): Do not call
14863 ix86_fixup_binary_operands_no_copy.
14864 (<any_logic:code><VI_32:mode>3): Remove expander.
14865 (<any_logic:code><VI_32:mode>3): Rename from
14866 "*<any_logic:code><VI_32:mode>3".
14867 (uavg<mode>3_ceil): Do not call ix86_fixup_binary_operands_no_copy.
14868 * config/i386/sse.md (div<VF2:mode>3): Do not call
14869 ix86_fixup_binary_operands_no_copy.
14870 (div<VF1:mode>3): Ditto.
14871 (<maxmin:code><VI8_AVX2_AVX512F:mode>3): Ditto.
14872 (smulhrsv4hi3): Ditto.
14873 (smulhrsv2hi3): Ditto.
14875 2021-05-27 Martin Sebor <msebor@redhat.com>
14877 * ggc.h (gt_ggc_mx): Add overloads for all integers.
14879 * hash-map.h (class hash_map): Add pch_nx_helper overloads for all
14881 (hash_map::operator==): New function.
14883 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
14886 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
14887 For TARGET_XOP bypass SSE comparisons for all supported vector modes.
14888 * config/i386/mmx.md (*xop_maskcmp<MMXMODEI:mode>3): New insn pattern.
14889 (*xop_maskcmp<VI_32:mode>3): Ditto.
14890 (*xop_maskcmp_uns<MMXMODEI:mode>3): Ditto.
14891 (*xop_maskcmp_uns<VI_32:mode>3): Ditto.
14893 2021-05-27 Richard Earnshaw <rearnsha@arm.com>
14896 * config/arm/arm.c (arm_configure_build_target): Remove parameter
14897 opts_set, directly check opts parameters for being non-null.
14898 (arm_option_restore): Update call to arm_configure_build_target.
14899 (arm_option_override): Likewise.
14900 (arm_can_inline_p): Likewise.
14901 (arm_valid_target_attribute_tree): Likewise.
14902 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
14903 * config/arm/arm-protos.h (arm_configure_build_target): Adjust
14906 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
14908 * vr-values.c (simplify_conversion_using_ranges): Use
14909 get_range_query instead of get_global_range_query.
14911 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
14913 * gimple-range.cc (get_range_global): Move to value-query.cc.
14914 (gimple_range_global): Same.
14915 (get_global_range_query): Same.
14916 (global_range_query::range_of_expr): Same.
14917 * gimple-range.h (class global_range_query): Move to
14919 (gimple_range_global): Same.
14920 * tree-ssanames.c (get_range_info): Move to value-query.cc.
14921 (get_ptr_nonnull): Same.
14922 * tree-ssanames.h (get_range_info): Remove.
14923 (get_ptr_nonnull): Remove.
14924 * value-query.cc (get_ssa_name_range_info): Move from
14926 (get_ssa_name_ptr_info_nonnull): Same.
14927 (get_range_global): Move from gimple-range.cc.
14928 (gimple_range_global): Same.
14929 (get_global_range_query): Same.
14930 (global_range_query::range_of_expr): Same.
14931 * value-query.h (class global_range_query): Move from
14933 (gimple_range_global): Same.
14935 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
14938 * config/i386/mmx.md (uavgv4qi3_ceil): New insn pattern.
14939 (uavgv2hi3_ceil): Ditto.
14941 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
14944 * doc/extend.texi (scalar_storage_order): Rephrase slightly.
14946 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14948 * tree-ssanames.c (get_range_info): Merge both copies of
14949 get_range_info into one that works with irange.
14950 * tree-ssanames.h (get_range_info): Remove version that works on
14953 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14955 * builtins.c (check_nul_terminated_array): Convert to get_range_query.
14956 (expand_builtin_strnlen): Same.
14957 (determine_block_size): Same.
14958 * fold-const.c (expr_not_equal_to): Same.
14959 * gimple-fold.c (size_must_be_zero_p): Same.
14960 * gimple-match-head.c: Include gimple-range.h.
14961 * gimple-pretty-print.c (dump_ssaname_info): Convert to get_range_query.
14962 * gimple-ssa-warn-restrict.c
14963 (builtin_memref::extend_offset_range): Same.
14964 * graphite-sese-to-poly.c (add_param_constraints): Same.
14965 * internal-fn.c (get_min_precision): Same.
14966 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): Same.
14967 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Same.
14969 * tree-data-ref.c (split_constant_offset): Same.
14970 (dr_step_indicator): Same.
14971 * tree-dfa.c (get_ref_base_and_extent): Same.
14972 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
14973 * tree-ssa-loop-niter.c (refine_value_range_using_guard): Same.
14974 (determine_value_range): Same.
14975 (record_nonwrapping_iv): Same.
14976 (infer_loop_bounds_from_signedness): Same.
14977 (scev_var_range_cant_overflow): Same.
14978 * tree-ssa-phiopt.c (two_value_replacement): Same.
14979 * tree-ssa-pre.c (insert_into_preds_of_block): Same.
14980 * tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Same.
14981 * tree-ssa-strlen.c (handle_builtin_stxncpy_strncat): Same.
14983 (dump_strlen_info): Same.
14984 (set_strlen_range): Same.
14985 (maybe_diag_stxncpy_trunc): Same.
14986 (get_len_or_size): Same.
14987 (handle_integral_assign): Same.
14988 * tree-ssa-structalias.c (find_what_p_points_to): Same.
14989 * tree-ssa-uninit.c (find_var_cmp_const): Same.
14990 * tree-switch-conversion.c (bit_test_cluster::emit): Same.
14991 * tree-vect-patterns.c (vect_get_range_info): Same.
14992 (vect_recog_divmod_pattern): Same.
14993 * tree-vrp.c (intersect_range_with_nonzero_bits): Same.
14994 (register_edge_assert_for_2): Same.
14995 (determine_value_range_1): Same.
14996 * tree.c (get_range_pos_neg): Same.
14997 * vr-values.c (vr_values::get_lattice_entry): Same.
14998 (vr_values::update_value_range): Same.
14999 (simplify_conversion_using_ranges): Same.
15001 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
15003 * gimple-ssa-warn-alloca.c (alloca_call_type): Use
15004 get_range_query instead of query argument.
15005 (pass_walloca::execute): Enable and disable global ranger.
15007 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
15009 * gimple-ssa-evrp.c (rvrp_folder::rvrp_folder): Call
15011 (rvrp_folder::~rvrp_folder): Call disable_ranger.
15012 (hybrid_folder::hybrid_folder): Call enable_ranger.
15013 (hybrid_folder::~hybrid_folder): Call disable_ranger.
15015 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
15017 * function.c (allocate_struct_function): Set cfun->x_range_query.
15018 * function.h (struct function): Declare x_range_query.
15019 (get_range_query): New.
15020 (get_global_range_query): New.
15021 * gimple-range-cache.cc (ssa_global_cache::ssa_global_cache):
15022 Remove call to safe_grow_cleared.
15023 * gimple-range.cc (get_range_global): New.
15024 (gimple_range_global): Move from gimple-range.h.
15025 (get_global_range_query): New.
15026 (global_range_query::range_of_expr): New.
15027 (enable_ranger): New.
15028 (disable_ranger): New.
15029 * gimple-range.h (gimple_range_global): Move to gimple-range.cc.
15030 (class global_range_query): New.
15031 (enable_ranger): New.
15032 (disable_ranger): New.
15033 * gimple-ssa-evrp.c (evrp_folder::~evrp_folder): Rename
15034 dump_all_value_ranges to dump.
15035 * tree-vrp.c (vrp_prop::finalize): Same.
15036 * value-query.cc (range_query::dump): New.
15037 * value-query.h (range_query::dump): New.
15038 * vr-values.c (vr_values::dump_all_value_ranges): Rename to...
15039 (vr_values::dump): ...this.
15040 * vr-values.h (class vr_values): Rename dump_all_value_ranges to
15041 dump and make virtual.
15043 2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
15045 * config/i386/i386.c (ix86_autovectorize_vector_modes):
15046 Add V4QImode and V16QImode for TARGET_SSE2.
15047 * doc/sourcebuild.texi (Vector-specific attributes):
15048 Add vect64 and vect32 description.
15050 2021-05-26 Bernd Edlinger <bernd.edlinger@hotmail.de>
15052 * gimple-range-gori.cc (range_def_chain::register_dependency):
15053 Resize m_def_chain when needed.
15055 2021-05-26 Christophe Lyon <christophe.lyon@linaro.org>
15057 * config/arm/mve.md (mve_vaddvq_<supf><mode>): Prefix with '@'.
15058 * config/arm/neon.md (reduc_plus_scal_<mode>): Move to ..
15059 * config/arm/vec-common.md: .. here. Add support for MVE.
15061 2021-05-26 Jakub Jelinek <jakub@redhat.com>
15063 * config/epiphany/epiphany.c (epiphany_print_operand_address): Remove
15065 * config/microblaze/microblaze.c (microblaze_legitimize_address,
15067 microblaze_option_override, print_operand): Likewise.
15068 * config/microblaze/microblaze.md (call_internal_plt,
15069 call_value_intern_plt, call_value_intern): Likewise.
15070 * config/arm/aout.h (ASM_OUTPUT_ALIGN): Likewise.
15071 * config/iq2000/iq2000.md (call_internal1, call_value_internal1,
15072 call_value_multiple_internal1): Likewise.
15073 * config/bfin/bfin.c (symbolic_reference_mentioned_p): Likewise.
15075 2021-05-26 Jan-Benedict Glaw <jbglaw@lug-owl.de>
15077 * config/arc/arc.c (arc_address_cost, arc_print_operand_address,
15078 arc_ccfsm_advance, symbolic_reference_mentioned_p,
15079 arc_raw_symbolic_reference_mentioned_p): Remove register
15082 2021-05-26 Jakub Jelinek <jakub@redhat.com>
15085 * omp-low.c: Include omp-offload.h.
15086 (create_omp_child_function): If current_function_decl has
15087 "omp declare target" attribute and is_gimple_omp_offloaded,
15088 remove that attribute from the copy of attribute list and
15089 add "omp target entrypoint" attribute instead.
15090 (lower_omp_target): Mark .omp_data_sizes.* and .omp_data_kinds.*
15091 variables for offloading if in omp_maybe_offloaded_ctx.
15092 * omp-offload.c (pass_omp_target_link::execute): Nullify second
15093 argument to GOMP_target_data_ext in offloaded code.
15095 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
15097 * config/csky/csky.c (csky_can_change_mode_class): Delete.
15098 For csky, HF/SF mode use the low bits of VREGS.
15100 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
15102 * gimplify.c (gimplify_decl_expr): Do not clear TREE_READONLY on a
15103 DECL which is a reference for OMP.
15105 2021-05-26 Martin Liska <mliska@suse.cz>
15107 PR gcov-profile/100751
15108 * doc/gcov.texi: Document that __gcov_dump can be called just
15109 once and that __gcov_reset resets run-time counters.
15111 2021-05-26 Martin Liska <mliska@suse.cz>
15113 * doc/install.texi: Port relevant part from install-old.texi
15114 and re-generate list of CPUs and systems.
15116 2021-05-26 Martin Liska <mliska@suse.cz>
15118 * Makefile.in: Remove it.
15119 * doc/include/fdl.texi: Update next/previous chapters.
15120 * doc/install.texi: Likewise.
15121 * doc/install-old.texi: Removed.
15123 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
15125 * config/csky/csky.c (ck810_legitimate_index_p): Support
15126 "base + index" with DF mode.
15127 * config/csky/constraints.md ("Y"): New constraint for memory operands
15128 without index register.
15129 * config/csky/csky_insn_fpuv2.md (fpuv3_movdf): Use "Y" instead of "m"
15130 when mov between memory and general registers, and lower their priority.
15131 * config/csky/csky_insn_fpuv3.md (fpuv2_movdf): Likewise.
15133 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
15135 * config/csky/csky.c (TARGET_PROMOTE_PROTOTYPES): Delete.
15137 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
15139 * config/csky/csky.md (untyped_call): Emit clobber for return
15140 registers to mark them used.
15142 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
15144 * config/csky/csky.md (cskyv2_sextend_ldbs): New.
15146 2021-05-26 Andrew Pinski <apinski@marvell.com>
15148 * match.pd (x < 0 ? ~y : y): New patterns.
15150 2021-05-26 Andrew Pinski <apinski@marvell.com>
15152 * match.pd (A?CST1:CST2): Add simplifcations for A?0:+-1, A?+-1:0,
15153 A?POW2:0 and A?0:POW2.
15155 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15157 * gimple-range-gori.cc (class logical_stmt_cache): Delete
15158 (logical_stmt_cache::logical_stmt_cache ): Delete.
15159 (logical_stmt_cache::~logical_stmt_cache): Delete.
15160 (logical_stmt_cache::cache_entry::dump): Delete.
15161 (logical_stmt_cache::get_range): Delete.
15162 (logical_stmt_cache::cached_name ): Delete.
15163 (logical_stmt_cache::same_cached_name): Delete.
15164 (logical_stmt_cache::cacheable_p): Delete.
15165 (logical_stmt_cache::slot_diagnostics ): Delete.
15166 (logical_stmt_cache::dump): Delete.
15167 (gori_compute_cache::gori_compute_cache): Delete.
15168 (gori_compute_cache::~gori_compute_cache): Delete.
15169 (gori_compute_cache::compute_operand_range): Delete.
15170 (gori_compute_cache::cache_stmt): Delete.
15171 * gimple-range-gori.h (gori_compute::compute_operand_range): Remove
15173 (class gori_compute_cache): Delete.
15175 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15177 * gimple-range.cc (fold_using_range::range_of_range_op): Use m_gori
15179 (fold_using_range::range_of_address): Adjust.
15180 (fold_using_range::range_of_phi): Adjust.
15181 * gimple-range.h (class fur_source): Adjust.
15182 (fur_source::fur_source): Adjust.
15184 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15186 * gimple-range-gori.cc (gori_compute::expr_range_at_stmt): Rename
15187 from expr_range_in_bb and adjust.
15188 (gori_compute::compute_name_range_op): Adjust.
15189 (gori_compute::optimize_logical_operands): Adjust.
15190 (gori_compute::compute_logical_operands_in_chain): Adjust.
15191 (gori_compute::compute_operand1_range): Adjust.
15192 (gori_compute::compute_operand2_range): Adjust.
15193 (ori_compute_cache::cache_stmt): Adjust.
15194 * gimple-range-gori.h (gori_compute): Rename prototype.
15196 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15198 * gimple-range.cc (gimple_ranger::range_of_expr): Non-null should be
15199 checked only after range_of_stmt, not range_on_entry.
15200 (gimple_ranger::range_on_entry): Check for non-null in any
15201 predecessor block, if it is not already non-null.
15202 (gimple_ranger::range_on_exit): DOnt check for non-null after
15203 range on entry call.
15204 (gimple_ranger::dump_bb): New. Split from dump.
15205 (gimple_ranger::dump): Adjust.
15206 * gimple-range.h (class gimple_ranger): Adjust.
15208 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15210 * gimple-range-cache.cc (struct range_timestamp): Delete.
15211 (class temporal_cache): Adjust.
15212 (temporal_cache::get_timestamp): Delete.
15213 (temporal_cache::set_dependency): Delete.
15214 (temporal_cache::temporal_value): Adjust.
15215 (temporal_cache::current_p): Take dependencies as params.
15216 (temporal_cache::set_timestamp): Adjust.
15217 (temporal_cache::set_always_current): Adjust.
15218 (ranger_cache::get_non_stale_global_range): Adjust.
15219 (ranger_cache::register_dependency): Delete.
15220 * gimple-range-cache.h (class range_cache): Adjust.
15222 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15224 * gimple-range-gori.cc (range_def_chain::range_def_chain): init
15226 (range_def_chain::~range_def_chain): Dispose of obstack rather than
15227 each individual bitmap.
15228 (range_def_chain::set_import): New.
15229 (range_def_chain::get_imports): New.
15230 (range_def_chain::chain_import_p): New.
15231 (range_def_chain::register_dependency): Rename from build_def_chain
15233 (range_def_chain::def_chain_in_bitmap_p): New.
15234 (range_def_chain::add_def_chain_to_bitmap): New.
15235 (range_def_chain::has_def_chain): Just check first depenedence.
15236 (range_def_chain::get_def_chain): Process imports, use generic
15237 register_dependency routine.
15238 (range_def_chain::dump): New.
15239 (gori_map::gori_map): Allocate import list.
15240 (gori_map::~gori_map): Release imports.
15241 (gori_map::exports): Check for past allocated block size.
15242 (gori_map::imports): New.
15243 (gori_map::def_chain_in_export_p): Delete.
15244 (gori_map::is_import_p): New.
15245 (gori_map::maybe_add_gori): Handle imports.
15246 (gori_map::dump): Adjust output, add imports.
15247 (gori_compute::has_edge_range_p): Remove def_chain_in_export call.
15248 (gori_export_iterator::gori_export_iterator): New.
15249 (gori_export_iterator::next): New.
15250 (gori_export_iterator::get_name): New.
15251 * gimple-range-gori.h (range_def_chain): Add imports and direct
15252 dependecies via struct rdc.
15253 (range_def_chain::depend1): New.
15254 (range_def_chain::depend2): New.
15255 (class gori_map): Adjust.
15256 (FOR_EACH_GORI_IMPORT_NAME): New.
15257 (FOR_EACH_GORI_EXPORT_NAME): New.
15258 (class gori_export_iterator): New.
15260 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15262 * gimple-range-cache.cc (ranger_cache::ranger_cache): Move initial
15263 export cache filling to here.
15264 * gimple-range-gori.cc (gori_compute::gori_compute) : From Here.
15266 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
15268 * gimple-range-gori.cc (range_def_chain): Move to gimple-range-gori.h.
15269 (gori_map): Move to gimple-range-gori.h.
15270 (gori_compute::gori_compute): Adjust.
15271 (gori_compute::~gori_compute): Delete.
15272 (gori_compute::compute_operand_range_switch): Adjust.
15273 (gori_compute::compute_operand_range): Adjust.
15274 (gori_compute::compute_logical_operands): Adjust.
15275 (gori_compute::has_edge_range_p ): Adjust.
15276 (gori_compute::set_range_invariant): Delete.
15277 (gori_compute::dump): Adjust.
15278 (gori_compute::outgoing_edge_range_p): Adjust.
15279 * gimple-range-gori.h (class range_def_chain): Relocate here.
15280 (class gori_map): Relocate here.
15281 (class gori_compute): Inherit from gori_map, and adjust.
15283 2021-05-25 Aldy Hernandez <aldyh@redhat.com>
15285 * value-range.cc (range_tests_legacy): Use
15286 build_nonstandard_integer_type instead of int and short.
15288 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
15290 * gimplify.c (gimplify_decl_expr): Clear TREE_READONLY on the DECL
15291 when really creating an initialization statement for it.
15293 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
15295 * tree-inline.c (setup_one_parameter): Fix thinko in new condition.
15297 2021-05-25 Kito Cheng <kito.cheng@sifive.com>
15299 * config/riscv/riscv.h (ASM_SPEC): Pass -mno-relax.
15301 2021-05-25 Martin Liska <mliska@suse.cz>
15303 PR tree-optimization/92860
15305 * optc-save-gen.awk: Remove exceptions.
15307 2021-05-25 Martin Liska <mliska@suse.cz>
15309 * asan.h (sanitize_coverage_p): New function.
15310 * doc/extend.texi: Document it.
15311 * fold-const.c (fold_range_test): Use sanitize_flags_p
15312 instead of flag_sanitize_coverage.
15313 (fold_truth_andor): Likewise.
15314 * sancov.c: Likewise.
15315 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
15316 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Handle
15317 -fsanitize-coverage when inlining.
15319 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
15321 * config/csky/csky-modes.def : Fix copyright.
15323 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
15325 * config/csky/csky-modes.def : Amend copyright.
15326 * config/csky/csky_insn_fpuv2.md : Likewise.
15327 * config/csky/csky_insn_fpuv3.md : Likewise.
15329 2021-05-25 Richard Biener <rguenther@suse.de>
15331 PR middle-end/100727
15332 * calls.c (initialize_argument_information): Explicitely test
15333 for WITH_SIZE_EXPR.
15334 * gimple-expr.c (mark_addressable): Skip outer WITH_SIZE_EXPR.
15336 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
15338 * config/csky/csky.h (FRAME_POINTER_REGNUM): Use
15339 HARD_FRAME_POINTER_REGNUM and FRAME_POINTER_REGNUM instead of
15340 the signle definition. The signle definition may not work well
15341 at simplify_subreg_regno().
15342 (HARD_FRAME_POINTER_REGNUM): New.
15343 (ELIMINABLE_REGS): Add for HARD_FRAME_POINTER_REGNUM.
15344 * config/csky/csky.c (get_csky_live_regs, csky_can_eliminate,
15345 csky_initial_elimination_offset, csky_expand_prologue,
15346 csky_expand_epilogue): Add for HARD_FRAME_POINTER_REGNUM.
15348 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
15350 * config/csky/csky.c (csky_option_override):
15351 Init csky_arch_isa_features[] in advance, so TARGET_DSP
15352 and TARGET_DIV can be set well.
15354 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
15356 * config/csky/constraints.md ("l", "h"): Delete.
15357 * config/csky/csky.h (reg_class, REG_CLASS_NAMES,
15358 REG_CLASS_CONTENTS): Delete LO_REGS and HI_REGS.
15359 * config/csky/csky.c (regno_reg_classm,
15360 csky_secondary_reload, csky_register_move_cost):
15361 Use HILO_REGS instead of LO_REGS and HI_REGS.
15363 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
15365 * config/csky/constraints.md ("W"): New constriant for mem operand
15366 with base reg, index register.
15367 ("Q"): Renamed and modified "csky_valid_fpuv2_mem_operand" to
15368 "csky_valid_mem_constraint_operand" to deal with both "Q" and "W"
15370 ("Dv"): New constraint for const double value that can be used at
15372 * config/csky/csky-modes.def (HFmode): New mode.
15373 * config/csky/csky-protos.h (csky_valid_fpuv2_mem_operand): Rename
15374 to "csky_valid_mem_constraint_operand" and support new constraint
15376 (csky_get_movedouble_length): New.
15377 (fpuv3_output_move): New.
15378 (fpuv3_const_double): New.
15379 * config/csky/csky.c (csky_option_override): New arch CK860 with fpv3.
15380 (decompose_csky_address): Refine.
15381 (csky_print_operand): New "CONST_DOUBLE" operand.
15382 (csky_output_move): Support fpv3 instructions.
15383 (csky_get_movedouble_length): New.
15384 (fpuv3_output_move): New.
15385 (fpuv3_const_double): New.
15386 (csky_emit_compare): Cover float comparsion.
15387 (csky_emit_compare_float): Refine.
15388 (csky_vaild_fpuv2_mem_operand): Rename to
15389 "csky_valid_mem_constraint_operand" and support new constraint "W".
15390 (ck860_rtx_costs): New.
15391 (csky_rtx_costs): Add the cost calculation of CK860.
15392 (regno_reg_class): New vregs for fpuv3.
15393 (csky_dbx_regno): Likewise.
15394 (csky_cpu_cpp_builtins): New builtin macro for fpuv3.
15395 (csky_conditional_register_usage): Suporrot fpuv3.
15396 (csky_dwarf_register_span): Suporrot fpuv3.
15397 (csky_init_builtins, csky_mangle_type): Support "__fp16" type.
15398 (ck810_legitimate_index_p): Support fp16.
15399 * config/csky/csky.h (TARGET_TLS): ADD CK860.
15400 (CSKY_VREG_P, CSKY_VREG_LO_P, CSKY_VREG_HI_P): Support fpuv3.
15401 (TARGET_SINGLE_FPU): Support fpuv3.
15402 (TARGET_SUPPORT_FPV3): New.
15403 (FIRST_PSEUDO_REGISTER): Change to 202 to hold the new fpuv3 registers.
15404 (FIXED_REGISTERS, CALL_REALLY_USED_REGISTERS, REGISTER_NAMES,
15405 REG_CLASS_CONTENTS): Support fpuv3.
15406 * config/csky/csky.md (movsf): Move to cksy_insn_fpu.md and refine.
15407 (csky_movsf_fpv2): Likewise.
15408 (ck801_movsf): Likewise.
15409 (csky_movsf): Likewise.
15411 (csky_movdf_fpv2): Likewise.
15412 (ck801_movdf): Likewise.
15413 (csky_movdf): Likewise.
15414 (movsicc): Refine. Use "comparison_operatior" instead of
15415 "ordered_comparison_operatior".
15416 (addsicc): Likewise.
15417 (CSKY_FIRST_VFP3_REGNUM, CSKY_LAST_VFP3_REGNUM): New constant.
15418 (call_value_internal_vh): New.
15419 * config/csky/csky_cores.def (CK860): New arch and cpu.
15424 * config/csky/csky_insn_fpu.md: Refactor. Separate all float patterns
15425 into emit-patterns and match-patterns, remain the emit-patterns here,
15426 and move the match-patterns to csky_insn_fpuv2.md or
15427 csky_insn_fpuv3.md.
15428 * config/csky/csky_insn_fpuv2.md: New file for fpuv2 instructions.
15429 * config/csky/csky_insn_fpuv3.md: New file and new patterns for fpuv3
15431 * config/csky/csky_isa.def (fcr): New.
15436 (CK860): New definition for ck860.
15437 * config/csky/csky_tables.opt (ck860): New processors ck860,
15438 ck860f. And new arch ck860.
15443 * config/csky/predicates.md (csky_float_comparsion_operator): Delete
15444 "geu", "gtu", "leu", "ltu", which will never appear at float comparison.
15445 * config/csky/t-csky-elf: Support 860.
15446 * config/csky/t-csky-linux: Likewise.
15447 * doc/md.texi: Add "Q" and "W" constraints for C-SKY.
15449 2021-05-24 Aaron Sawdey <acsawdey@linux.ibm.com>
15451 * config/rs6000/genfusion.pl (gen_logical_addsubf): Refactor to
15452 add generation of logical-add and add-logical fusion pairs.
15453 * config/rs6000/rs6000-cpus.def: Add new fusion to ISA 3.1 mask
15455 * config/rs6000/rs6000.c (rs6000_option_override_internal): Turn on
15456 logical-add and add-logical fusion by default.
15457 * config/rs6000/rs6000.opt: Add -mpower10-fusion-logical-add and
15458 -mpower10-fusion-add-logical options.
15459 * config/rs6000/fusion.md: Regenerate file.
15461 2021-05-24 Aldy Hernandez <aldyh@redhat.com>
15463 * value-range.cc (irange::legacy_equal_p): Check type when
15464 comparing VR_VARYING types.
15465 (range_tests_legacy): Test comparing VARYING ranges of different
15468 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
15470 * config/aarch64/aarch64.c (neoversen1_tunings):
15471 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
15473 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
15475 * config/aarch64/aarch64.c (aarch64_classify_symbol): Use GOT for
15476 extern weak symbols. Limit symbol offsets for non-GOT symbols with
15479 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
15481 * config/arm/neon.md (vec_load_lanesxi<mode>)
15482 (vec_store_lanexoi<mode>): Move ...
15483 * config/arm/vec-common.md: here.
15485 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
15487 * config/arm/neon.md (vec_load_lanesoi<mode>)
15488 (vec_store_lanesoi<mode>): Move ...
15489 * config/arm/vec-common.md: here.
15491 2021-05-24 liuhongt <hongtao.liu@intel.com>
15494 * config/i386/i386.c (ix86_gimple_fold_builtin): Replacing
15495 stmt with GIMPLE_NOP when lhs doesn't exist.
15497 2021-05-23 Uroš Bizjak <ubizjak@gmail.com>
15500 * config/i386/mmx.md (*push<VI_32:mode>2_rex64):
15501 New instruction pattern.
15502 (*push<VI_32:mode>2): Ditto.
15503 (push splitter for SSE registers): New splitter.
15505 2021-05-23 Andrew Pinski <apinski@marvell.com>
15507 * match.pd ((A & C) != 0 ? D : 0): Limit to non pointer types.
15509 2021-05-22 Aaron Sawdey <acsawdey@linux.ibm.com>
15511 * config/rs6000/genfusion.pl (gen_addadd): Fix incorrect attr types.
15512 * config/rs6000/fusion.md: Regenerate file.
15514 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
15516 * config/rs6000/genfusion.pl (gen_addadd): New function.
15517 * config/rs6000/fusion.md: Regenerate file.
15518 * config/rs6000/rs6000-cpus.def: Add
15519 OPTION_MASK_P10_FUSION_2ADD to masks.
15520 * config/rs6000/rs6000.c (rs6000_option_override_internal):
15521 Handle default value of OPTION_MASK_P10_FUSION_2ADD.
15522 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2add.
15524 2021-05-21 Jakub Jelinek <jakub@redhat.com>
15526 PR middle-end/99928
15527 * tree.h (OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET): Define.
15528 * gimplify.c (enum gimplify_omp_var_data): Fix up
15529 GOVD_MAP_HAS_ATTACHMENTS value, add GOVD_FIRSTPRIVATE_IMPLICIT.
15530 (omp_lastprivate_for_combined_outer_constructs): If combined target
15531 has GOVD_FIRSTPRIVATE_IMPLICIT set for the decl, change it to
15532 GOVD_MAP | GOVD_SEEN.
15533 (gimplify_scan_omp_clauses): Set GOVD_FIRSTPRIVATE_IMPLICIT for
15534 firstprivate clauses with OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT.
15535 (gimplify_adjust_omp_clauses): For firstprivate clauses with
15536 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT either clear that bit and
15537 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET too, or remove it and
15538 let it be replaced by implicit map clause.
15540 2021-05-21 Jakub Jelinek <jakub@redhat.com>
15542 PR middle-end/99928
15543 * gimplify.c (omp_lastprivate_for_combined_outer_constructs): New
15545 (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LASTPRIVATE>: Use it.
15546 (gimplify_omp_for): Likewise.
15548 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15550 PR middle-end/90115
15551 * omp-low.c (oacc_privatization_candidate_p): Reject 'static',
15552 'external' in blocks.
15554 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15556 PR middle-end/90115
15557 * flag-types.h (enum openacc_privatization): New.
15558 * params.opt (-param=openacc-privatization): New.
15559 * doc/invoke.texi (openacc-privatization): Document it.
15560 * omp-general.h (get_openacc_privatization_dump_flags): New
15562 * omp-low.c (oacc_privatization_candidate_p): Add diagnostics.
15563 * omp-offload.c (execute_oacc_device_lower)
15564 <IFN_UNIQUE_OACC_PRIVATE>: Re-work diagnostics.
15565 * target.def (goacc.adjust_private_decl): Add 'location_t'
15567 * doc/tm.texi: Regenerate.
15568 * config/gcn/gcn-protos.h (gcn_goacc_adjust_private_decl): Adjust.
15569 * config/gcn/gcn-tree.c (gcn_goacc_adjust_private_decl): Likewise.
15570 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl):
15571 Likewise. Preserve it for...
15572 (nvptx_goacc_expand_var_decl): ... use here.
15574 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15576 * doc/sourcebuild.texi (Other attributes): Document '__OPTIMIZE__'
15579 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15581 PR middle-end/90115
15582 * omp-low.c (oacc_privatization_candidate_p): New function.
15583 (oacc_privatization_scan_clause_chain)
15584 (oacc_privatization_scan_decl_chain): Use it. Also
15585 'gcc_checking_assert' that we're not seeing duplicates.
15587 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15589 PR middle-end/90115
15590 * omp-offload.c (execute_oacc_device_lower): Skip processing if no
15593 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15595 PR middle-end/90115
15596 * omp-offload.c (execute_oacc_device_lower): Explain.
15598 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15600 PR middle-end/90115
15601 * omp-offload.c (execute_oacc_device_lower)
15602 <IFN_UNIQUE_OACC_PRIVATE>: Diagnose and handle for 'level == -1'
15604 * internal-fn.c (expand_UNIQUE): Don't expect
15605 'IFN_UNIQUE_OACC_PRIVATE'.
15607 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15609 PR middle-end/90115
15610 * omp-low.c (lower_omp_for): Don't evaluate OpenMP 'for' clauses.
15612 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15614 PR middle-end/90115
15615 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl)
15616 (nvptx_goacc_expand_var_decl): Tighten.
15618 2021-05-21 Julian Brown <julian@codesourcery.com>
15619 Chung-Lin Tang <cltang@codesourcery.com>
15620 Thomas Schwinge <thomas@codesourcery.com>
15622 PR middle-end/90115
15623 * doc/tm.texi.in (TARGET_GOACC_EXPAND_VAR_DECL)
15624 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Add documentation hooks.
15625 * doc/tm.texi: Regenerate.
15626 * expr.c (expand_expr_real_1): Expand decls using the
15627 expand_var_decl OpenACC hook if defined.
15628 * internal-fn.c (expand_UNIQUE): Handle IFN_UNIQUE_OACC_PRIVATE.
15629 * internal-fn.h (IFN_UNIQUE_CODES): Add OACC_PRIVATE.
15630 * omp-low.c (omp_context): Add oacc_privatization_candidates
15632 (lower_oacc_reductions): Add PRIVATE_MARKER parameter. Insert
15634 (lower_oacc_head_tail): Add PRIVATE_MARKER parameter. Modify
15635 private marker's gimple call arguments, and pass it to
15636 lower_oacc_reductions.
15637 (oacc_privatization_scan_clause_chain)
15638 (oacc_privatization_scan_decl_chain, lower_oacc_private_marker):
15640 (lower_omp_for, lower_omp_target, lower_omp_1): Use these.
15641 * omp-offload.c (convert.h): Include.
15642 (oacc_loop_xform_head_tail): Treat private-variable markers like
15643 fork/join when transforming head/tail sequences.
15644 (struct var_decl_rewrite_info): Add struct.
15645 (oacc_rewrite_var_decl, is_sync_builtin_call): New functions.
15646 (execute_oacc_device_lower): Support rewriting gang-private
15647 variables using target hook, and fix up addr_expr and var_decl
15649 * target.def (adjust_private_decl, expand_var_decl): New hooks.
15650 * config/gcn/gcn-protos.h (gcn_goacc_adjust_gangprivate_decl):
15652 (gcn_goacc_adjust_private_decl): ...this.
15653 * config/gcn/gcn-tree.c (gcn_goacc_adjust_gangprivate_decl):
15655 (gcn_goacc_adjust_private_decl): ...this. Add LEVEL parameter.
15656 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_GANGPRIVATE_DECL): Rename
15657 definition using gcn_goacc_adjust_gangprivate_decl...
15658 (TARGET_GOACC_ADJUST_PRIVATE_DECL): ...to this, using
15659 gcn_goacc_adjust_private_decl.
15660 * config/nvptx/nvptx.c (tree-pretty-print.h): Include.
15661 (gang_private_shared_size): New global variable.
15662 (gang_private_shared_align): Likewise.
15663 (gang_private_shared_sym): Likewise.
15664 (gang_private_shared_hmap): Likewise.
15665 (nvptx_option_override): Initialize these.
15666 (nvptx_file_end): Output gang_private_shared_sym.
15667 (nvptx_goacc_adjust_private_decl, nvptx_goacc_expand_var_decl):
15669 (nvptx_set_current_function): Clear gang_private_shared_hmap.
15670 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Define hook.
15671 (TARGET_GOACC_EXPAND_VAR_DECL): Likewise.
15673 2021-05-21 H.J. Lu <hjl.tools@gmail.com>
15675 * config/i386/i386-modes.def (MAX_BITSIZE_MODE_ANY_INT): Removed.
15677 2021-05-21 Richard Biener <rguenther@suse.de>
15678 H.J. Lu <hjl.tools@gmail.com>
15680 PR middle-end/90773
15681 * expr.c (expand_constructor): Elide expand_constructor if
15682 move by pieces is preferred.
15684 2021-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15686 * config/aarch64/aarch64-builtins.c (aarch64_call_properties):
15687 Take a flag and mode value as arguments.
15688 (aarch64_modifies_global_state_p): Likewise.
15689 (aarch64_reads_global_state_p): Likewise.
15690 (aarch64_could_trap_p): Likewise.
15691 (aarch64_get_attributes): Likewise.
15692 (aarch64_init_simd_builtins): Adjust callsite of above.
15693 (aarch64_init_fcmla_laneq_builtins): Use aarch64_get_attributes to get
15694 function attributes to apply to builtins.
15695 (aarch64_init_crc32_builtins): Likewise.
15696 (aarch64_init_builtin_rsqrt): Likewise.
15698 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
15700 * config/rs6000/rs6000.md (define_attr "type"): Add types for fusion.
15701 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10): Use new fusion types.
15702 (gen_2logical): Use new fusion types.
15703 * config/rs6000/fusion.md: Regenerate.
15705 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
15708 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
15709 Handle V4QI and V2HI modes.
15710 (ix86_expand_sse_movcc): Ditto.
15711 * config/i386/mmx.md (*<sat_plusminus:insn><VI_32:mode>3):
15712 New instruction pattern.
15713 (*eq<VI_32:mode>3): Ditto.
15714 (*gt<VI_32:mode>3): Ditto.
15715 (*xop_pcmov_<VI_32:mode>): Ditto.
15716 (mmx_pblendvb32): Ditto.
15717 (mmx_pblendvb64): Rename from mmx_pblendvb.
15718 (vec_cmp<VI_32:mode><VI_32:mode>): New expander.
15719 (vec_cmpu<VI_32:mode><VI_32:mode>): Ditto.
15720 (vcond<VI_32:mode><VI_32:mode>): Ditto.
15721 (vcondu<VI_32:mode><VI_32:mode>): Ditto.
15722 (vcond_mask_<VI_32:mode><VI_32:mode>): Ditto.
15724 2021-05-21 Jakub Jelinek <jakub@redhat.com>
15726 PR tree-optimization/94589
15727 * tree-ssa-phiopt.c (spaceship_replacement): For integral rhs1 and
15728 rhs2, treat x <= 4 equivalently to x < 5 etc. In cmp1 and cmp2 (if
15729 not the same as cmp3) treat <= the same as < and >= the same as >.
15730 Don't require that cond2_phi_edge is true edge, instead take
15731 false/true edges into account based on cmp1/cmp2 comparison kinds.
15733 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
15736 * config/i386/mmx.md (SMAXMIN_MMXMODEI): New mode iterator.
15737 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Macroize expander
15738 from <smaxmin:code>v4hi3> and <smaxmin:code><MMXMODE14:mode>3
15739 using SMAXMIN_MMXMODEI mode iterator.
15740 (*<smaxmin:code>v4qi3): New insn pattern.
15741 (*<smaxmin:code>v2hi3): Ditto.
15742 (SMAXMIN_VI_32): New mode iterator.
15743 (<smaxmin:code><SMAXMIN_VI_32>mode3): New expander.
15744 (UMAXMIN_MMXMODEI): New mode iterator.
15745 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Macroize expander
15746 from <umaxmin:code>v8qi3> and <umaxmin:code><MMXMODE24:mode>3
15747 using UMAXMIN_MMXMODEI mode iterator.
15748 (*<umaxmin:code>v4qi3): New insn pattern.
15749 (*<umaxmin:code>v2hi3): Ditto.
15750 (UMAXMIN_VI_32): New mode iterator.
15751 (<umaxmin:code><UMAXMIN_VI_32>mode3): New expander.
15752 (abs<VI_32:mode>2): New insn pattern.
15753 (ssse3_abs<MMXMODEI:mode>2, abs<MMXMODEI:mode>2): Move from ...
15754 * config/i386/sse.md: ... here.
15756 2021-05-20 Clement Chigot <clement.chigot@atos.net>
15757 David Edelsohn <dje.gcc@gmail.com>
15759 * collect2.c (scan_prog_file): Issue non-fatal warning for
15762 2021-05-20 Jonathan Wakely <jwakely@redhat.com>
15764 * doc/invoke.texi (-Wno-c++11-extensions)
15765 (-Wno-c++14-extensions, -Wno-c++17-extensions)
15766 (-Wno-c++20-extensions, -Wno-c++23-extensions): Document
15769 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
15771 * config/c6x/c6x.c (c6x_output_file_unwind): Use dwarf_debuginfo_p.
15772 * config/darwin.c (darwin_override_options): Likewise.
15773 * config/i386/cygming.h (DBX_REGISTER_NUMBER): Likewise.
15774 * config/i386/darwin.h (DBX_REGISTER_NUMBER): Likewise.
15775 (DWARF2_FRAME_REG_OUT): Likewise.
15776 * config/mips/mips.c (mips_output_filename): Likewise.
15777 * config/rs6000/rs6000.c (rs6000_xcoff_declare_function_name):
15779 (rs6000_dbx_register_number): Likewise.
15780 * dbxout.c: Include flags.h.
15781 * dwarf2cfi.c (cfi_label_required_p): Likewise.
15782 (dwarf2out_do_frame): Likewise.
15783 * except.c: Include flags.h.
15784 * final.c (dwarf2_debug_info_emitted_p): Likewise.
15785 (final_scan_insn_1): Likewise.
15786 * flags.h (dwarf_debuginfo_p): New function declaration.
15787 * opts.c (dwarf_debuginfo_p): New function definition.
15788 * targhooks.c (default_debug_unwind_info): Use dwarf_debuginfo_p.
15789 * toplev.c (process_options): Likewise.
15791 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
15793 * common.opt: Change type to support bitmasks.
15794 * flag-types.h (enum debug_info_type): Rename enumerator constants.
15795 (NO_DEBUG): New bitmask.
15796 (DBX_DEBUG): Likewise.
15797 (DWARF2_DEBUG): Likewise.
15798 (XCOFF_DEBUG): Likewise.
15799 (VMS_DEBUG): Likewise.
15800 (VMS_AND_DWARF2_DEBUG): Likewise.
15801 * flags.h (debug_set_to_format): New function declaration.
15802 (debug_set_count): Likewise.
15803 (debug_set_names): Likewise.
15804 * opts.c (debug_type_masks): Array of bitmasks for debug formats.
15805 (debug_set_to_format): New function definition.
15806 (debug_set_count): Likewise.
15807 (debug_set_names): Likewise.
15808 (set_debug_level): Update access to debug_type_names.
15809 * toplev.c: Likewise.
15811 2021-05-20 Martin Sebor <msebor@redhat.com>
15813 PR middle-end/100684
15814 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Handle C++ lambda.
15816 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
15819 * config/i386/i386.md (isa): Remove x64_bmi.
15820 (enabled): Remove x64_bmi.
15821 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
15822 Remove general register alternative.
15823 (*andnot<VI_32:mode>3): Ditto.
15824 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
15825 (*<any_logic:code><VI_32:mode>3): Ditto.
15827 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
15829 * config/arm/arm.c: Include head files tree-vectorizer.h and
15832 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
15835 * config/i386/mmx.md (Yv_Yw): Revert adding V4QI and V2HI modes.
15836 (*<plusminus:insn><VI_32:mode>3): Use Yw instad of <Yv_Yw> constrint.
15837 (<s>mulv4hi3_highpart): New expander.
15838 (*<s>mulv2hi3_highpart): New insn pattern.
15839 (<s>mulv2hi3_higpart): New expander.
15840 (*<any_shift:insn>v2hi3): New insn pattern.
15841 (<any_shift:insn>v2hi3): New expander.
15842 * config/i386/sse.md (smulhrsv2hi3): New expander.
15843 (*smulhrsv2hi3): New insn pattern.
15845 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
15847 * doc/invoke.texi (vect-inner-loop-cost-factor): Document new
15849 * params.opt (vect-inner-loop-cost-factor): New.
15850 * targhooks.c (default_add_stmt_cost): Replace hardcoded factor
15851 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR, include head file
15852 tree-vectorizer.h and its required ones.
15853 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Replace
15854 hardcoded factor 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR.
15855 * config/arm/arm.c (arm_add_stmt_cost): Likewise.
15856 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
15857 * config/rs6000/rs6000.c (rs6000_add_stmt_cost): Likewise.
15858 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
15860 (_loop_vec_info::_loop_vec_info): Init inner_loop_cost_factor.
15861 * tree-vectorizer.h (_loop_vec_info): Add inner_loop_cost_factor.
15862 (LOOP_VINFO_INNER_LOOP_COST_FACTOR): New macro.
15864 2021-05-20 Christophe Lyon <christophe.lyon@linaro.org>
15865 Torbjörn Svensson <torbjorn.svensson@st.com>
15868 * doc/cpp.texi (Common Predefined Macros): Document __FILE_NAME__.
15870 2021-05-20 Jakub Jelinek <jakub@redhat.com>
15872 PR middle-end/99928
15873 * gimplify.c (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LINEAR>: For
15874 explicit linear clause when combined with target, make it map(tofrom:)
15875 instead of no clause or firstprivate.
15877 2021-05-20 Jakub Jelinek <jakub@redhat.com>
15879 PR tree-optimization/94589
15880 * match.pd ((X & Y) == X -> (X & ~Y) == 0): Simplify even in presence
15881 of integral conversions.
15883 2021-05-19 Andrew MacLeod <amacleod@redhat.com>
15885 * gimple-range.cc (fur_source::get_operand): New.
15886 (gimple_range_fold): Delete.
15887 (fold_using_range::fold_stmt): Move from gimple_ranger::calc_stmt.
15888 (fold_using_range::range_of_range_op): Move from gimple_ranger.
15889 (fold_using_range::range_of_address): Ditto.
15890 (fold_using_range::range_of_phi): Ditto.
15891 (fold_using_range::range_of_call): Ditto.
15892 (fold_using_range::range_of_builtin_ubsan_call): Move from
15893 range_of_builtin_ubsan_call.
15894 (fold_using_range::range_of_builtin_call): Move from
15895 range_of_builtin_call.
15896 (gimple_ranger::range_of_builtin_call): Delete.
15897 (fold_using_range::range_of_cond_expr): Move from gimple_ranger.
15898 (gimple_ranger::fold_range_internal): New.
15899 (gimple_ranger::range_of_stmt): Use new fold_using_range API.
15900 (fold_using_range::range_of_ssa_name_with_loop_info): Move from
15901 gimple_ranger. Improve ranges of SSA_NAMES when possible.
15902 * gimple-range.h (gimple_ranger): Remove various range_of routines.
15903 (class fur_source): New.
15904 (class fold_using_range): New.
15905 (fur_source::fur_source): New.
15907 * vr-values.c (vr_values::extract_range_basic): Use fold_using_range
15908 instead of range_of_builtin_call.
15910 2021-05-19 Jonathan Wakely <jwakely@redhat.com>
15912 * doc/cpp.texi (Common Predefined Macros): Update documentation
15913 for the __GXX_EXPERIMENTAL_CXX0X__ macro.
15915 2021-05-19 Alex Coplan <alex.coplan@arm.com>
15918 * config/arm/arm.md (nonsecure_call_internal): Always ensure
15919 callee's address is in a register.
15921 2021-05-19 Geng Qi <gengqi@linux.alibaba.com>
15923 * common/config/riscv/riscv-common.c
15924 (riscv_subset_list::parsing_subset_version): Properly parse the letter
15926 (riscv_subset_list::parse_std_ext,
15927 riscv_subset_list::parse_multiletter_ext): To handle errors generated
15928 in riscv_subset_list::parsing_subset_version.
15930 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15932 * config/aarch64/aarch64-simd.md: Use "neon_move_narrow_q"
15933 type attribute in patterns generating XTN(2).
15935 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15937 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_pack_trunc_<mode>):
15938 Remove as duplicate of...
15939 (aarch64_xtn<mode>): This.
15940 (aarch64_xtn2<mode>_le): Move position in file.
15941 (aarch64_xtn2<mode>_be): Move position in file.
15942 (aarch64_xtn2<mode>): Move position in file.
15943 (vec_pack_trunc_<mode>): Define as an expander.
15945 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15947 * config/aarch64/aarch64-simd-builtins.def: Split builtin
15948 generation for aarch64_<sur>q<r>shr<u>n_n<mode> pattern into
15949 separate scalar and vector generators.
15950 * config/aarch64/aarch64-simd.md
15951 (aarch64_<sur>q<r>shr<u>n_n<mode>): Define as an expander and
15953 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_le): This and...
15954 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_be): This.
15955 * config/aarch64/iterators.md: Define SD_HSDI iterator.
15957 2021-05-19 Jonathn Wright <jonathan.wright@arm.com>
15959 * config/aarch64/aarch64-simd.md: Use UNSPEC_SQXTUN instead
15961 * config/aarch64/iterators.md: Remove UNSPEC_SQXTUN2.
15963 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15965 * config/aarch64/aarch64-simd.md (aarch64_<sur>q<r>shr<u>n2_n<mode>):
15966 Implement as an expand emitting a big/little endian
15967 instruction pattern.
15968 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_le): Define.
15969 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_be): Define.
15971 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15973 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn2<mode>):
15974 Implement as an expand emitting a big/little endian
15975 instruction pattern.
15976 (aarch64_<sur><addsub>hn2<mode>_insn_le): Define.
15977 (aarch64_<sur><addsub>hn2<mode>_insn_be): Define.
15978 * config/aarch64/iterators.md: Remove UNSPEC_[R]ADDHN2 and
15979 UNSPEC_[R]SUBHN2 unspecs and ADDSUBHN2 iterator.
15981 2021-05-19 Richard Biener <rguenther@suse.de>
15983 PR middle-end/100672
15984 * fold-const.c (fold_negate_expr_1): Use element_precision.
15985 (negate_expr_p): Likewise.
15987 2021-05-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
15989 * config/aarch64/iterators.md (SVE_PRED_LOAD): New iterator.
15990 (pred_load): New int attribute.
15991 * config/aarch64/aarch64-sve.md
15992 (aarch64_load_<ANY_EXTEND:optab><SVE_HSDI:mode><SVE_PARTIAL_I:mode>): Use
15993 SVE_PRED_LOAD enum iterator and corresponding pred_load attribute.
15994 * config/aarch64/aarch64-sve-builtins-base.cc (expand): Update call to
15995 code_for_aarch64_load.
15997 2021-05-19 Richard Biener <rguenther@suse.de>
15999 * cfgexpand.c (discover_nonconstant_array_refs_r): Make
16000 sure TARGET_MEM_REF bases are expanded as memory.
16001 * tree-ssa-operands.c (operands_scanner::get_tmr_operands):
16002 Do not mark TARGET_MEM_REF bases addressable.
16003 * tree-ssa.c (non_rewritable_mem_ref_base): Handle
16004 TARGET_MEM_REF bases as never rewritable.
16005 * gimple-walk.c (walk_stmt_load_store_addr_ops): Do not
16006 walk TARGET_MEM_REF bases as address-takens.
16007 * tree-ssa-dce.c (ref_may_be_aliased): Handle TARGET_MEM_REF.
16009 2021-05-19 Richard Biener <rguenther@suse.de>
16011 * builtins.c (get_object_alignment_1): Strip outer
16013 * tree-dfa.c (get_ref_base_and_extent): Handle outer
16014 WITH_SIZE_EXPR for size processing and process the
16016 * tree-ssa-alias.c (ao_ref_base_alias_set): Strip
16017 outer WITH_SIZE_EXPR.
16018 (ao_ref_base_alias_ptr_type): Likewise.
16019 (refs_may_alias_p_2): Allow WITH_SIZE_EXPR in ref->ref
16020 and handle that accordingly, stripping it for the
16021 core alias workers.
16022 * tree.c (get_base_address): Handle WITH_SIZE_EXPR by
16023 looking through it instead of returning NULL.
16025 2021-05-19 Jakub Jelinek <jakub@redhat.com>
16027 PR middle-end/100576
16028 * builtins.c (check_read_access): Convert bound to size_type_node if
16031 2021-05-19 Richard Biener <rguenther@suse.de>
16033 * tree-cfg.c (verify_types_in_gimple_min_lval): Inline...
16034 (verify_types_in_gimple_reference): ... here. Sanitize.
16035 (verify_gimple_call): Verify references in LHS and arguments.
16036 (verify_gimple_assign_single): Reject WITH_SIZE_EXPR.
16038 2021-05-19 Uroš Bizjak <ubizjak@gmail.com>
16040 * config/i386/i386.h (VALID_INT_MODE_P):
16041 Add V8QI, V4HI and V2SI modes for TARGET_64BIT.
16042 * config/i386/i386.md (isa): Add x64_bmi.
16043 (enabled): Handle x64_bmi.
16044 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
16045 Add alternative using 64bit general registers.
16046 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
16048 2021-05-19 Jakub Jelinek <jakub@redhat.com>
16050 PR middle-end/99928
16051 * tree.h (OMP_MASTER_COMBINED): Define.
16052 * gimplify.c (gimplify_scan_omp_clauses): Rewrite lastprivate
16053 handling for outer combined/composite constructs to a loop.
16054 Handle lastprivate on combined target.
16055 (gimplify_expr): Formatting fix.
16057 2021-05-19 Xionghu Luo <luoxhu@linux.ibm.com>
16059 * passes.def: Add sink_code pass before store_merging.
16060 * tree-ssa-sink.c (pass_sink_code:clone): New.
16062 2021-05-18 Bill Schmidt <wschmidt@linux.ibm.com>
16064 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Remove call to
16065 rs6000_special_adjust_field_align_p.
16066 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
16067 * config/rs6000/rs6000-call.c (rs6000_function_arg_boundary):
16068 Remove ABI warning.
16069 (rs6000_function_arg): Likewise.
16070 * config/rs6000/rs6000-protos.h
16071 (rs6000_special_adjust_field_align_p): Remove prototype.
16072 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
16074 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Remove call to
16075 rs6000_special_adjust_field_align_p.
16077 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
16080 * config/i386/i386.h (VALID_SSE2_REG_MODE):
16081 Add V4QI and V2HI modes.
16082 (VALID_INT_MODE_P): Ditto.
16083 * config/i386/mmx.md (VI_32): New mode iterator.
16084 (mmxvecsize): Handle V4QI and V2HI.
16086 (mov<VI_32:mode>): New expander.
16087 (*mov<mode>_internal): New insn pattern.
16088 (movmisalign<VI_32:mode>): New expander.
16089 (neg<VI_32:mode>): New expander.
16090 (<plusminus:insn><VI_32:mode>3): New expander.
16091 (*<plusminus:insn><VI_32:mode>3): New insn pattern.
16092 (mulv2hi3): New expander.
16093 (*mulv2hi3): New insn pattern.
16094 (one_cmpl<VI_32:mode>2): New expander.
16095 (*andnot<VI_32:mode>3): New insn pattern.
16096 (<any_logic:code><VI_32:mode>3): New expander.
16097 (*<any_logic:code><VI_32:mode>3): New insn pattern.
16099 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
16101 * config/i386/sse.md (<any_extend:insn>v4qiv4di2):
16102 Fix a mode mismatch with operand 1.
16104 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
16107 * config/i386/i386-expand.c (split_double_mode): Return
16108 temporary register when simplify_gen_subreg fails with
16109 the high half od the paradoxical subreg.
16111 2021-05-18 Richard Biener <rguenther@suse.de>
16113 * cfgexpand.c (expand_one_var): Pass in forced_stack_var
16114 and honor it when expanding.
16115 (expand_used_vars_for_block): Pass through forced_stack_var.
16116 (expand_used_vars): Likewise.
16117 (discover_nonconstant_array_refs_r): Set bits in
16118 forced_stack_vars instead of marking vars TREE_ADDRESSABLE.
16119 (avoid_type_punning_on_regs): Likewise.
16120 (discover_nonconstant_array_refs): Likewise.
16121 (pass_expand::execute): Create and pass down forced_stack_var
16122 bitmap. For parameters and returns temporarily set
16123 TREE_ADDRESSABLE when expand_function_start.
16125 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
16127 * doc/sourcebuild.texi: Document 'dg-note'.
16129 2021-05-18 Tobias Burnus <tobias@codesourcery.com>
16132 * configure: Regenerate.
16133 * configure.ac (BUILD_CFLAG, BUILD_CXXFLAGS): Add $(CFLAGS-$@).
16135 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
16137 * gimple.h (is_gimple_omp_oacc): Tighten.
16138 * omp-low.c (check_omp_nesting_restrictions): Adjust.
16140 2021-05-18 Richard Biener <rguenther@suse.de>
16142 * tree-ssa-operands.c (mark_address_taken): Simplify.
16144 2021-05-18 Martin Liska <mliska@suse.cz>
16146 * config/gcn/mkoffload.c (STR): Redefine.
16147 * config/i386/intelmic-mkoffload.c (STR): Likewise.
16148 * config/nvptx/mkoffload.c (STR): Likewise.
16150 2021-05-18 Martin Liska <mliska@suse.cz>
16152 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
16153 Use startswith function instead of strncmp.
16154 * common/config/bfin/bfin-common.c (bfin_handle_option): Likewise.
16155 * common/config/riscv/riscv-common.c (riscv_subset_list::parse): Likewise.
16156 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): Likewise.
16157 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Likewise.
16158 * config/alpha/alpha.c (alpha_elf_section_type_flags): Likewise.
16159 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
16160 * config/arm/arm.c (arm_file_start): Likewise.
16161 (arm_valid_target_attribute_rec): Likewise.
16162 (thumb1_md_asm_adjust): Likewise.
16163 * config/arm/driver-arm.c (host_detect_local_cpu): Likewise.
16164 * config/avr/avr.c (STR_PREFIX_P): Likewise.
16165 (avr_set_current_function): Likewise.
16166 (avr_handle_addr_attribute): Likewise.
16167 (avr_asm_output_aligned_decl_common): Likewise.
16168 (avr_asm_named_section): Likewise.
16169 (avr_section_type_flags): Likewise.
16170 (avr_asm_select_section): Likewise.
16171 * config/c6x/c6x.c (c6x_in_small_data_p): Likewise.
16172 (c6x_section_type_flags): Likewise.
16173 * config/darwin-c.c (darwin_cfstring_ref_p): Likewise.
16174 (darwin_objc_declare_unresolved_class_reference): Likewise.
16175 (darwin_objc_declare_class_definition): Likewise.
16176 * config/darwin.c (indirect_data): Likewise.
16177 (darwin_encode_section_info): Likewise.
16178 (darwin_objc2_section): Likewise.
16179 (darwin_objc1_section): Likewise.
16180 (machopic_select_section): Likewise.
16181 (darwin_globalize_label): Likewise.
16182 (darwin_label_is_anonymous_local_objc_name): Likewise.
16183 (darwin_asm_named_section): Likewise.
16184 (darwin_asm_output_dwarf_offset): Likewise.
16185 * config/frv/frv.c (frv_string_begins_with): Likewise.
16186 (frv_in_small_data_p): Likewise.
16187 * config/gcn/mkoffload.c (STR): Likewise.
16189 * config/i386/i386-builtins.c (get_builtin_code_for_version): Likewise.
16190 * config/i386/i386-options.c (ix86_option_override_internal): Likewise.
16191 * config/i386/i386.c (x86_64_elf_section_type_flags): Likewise.
16192 (ix86_md_asm_adjust): Likewise.
16193 * config/i386/intelmic-mkoffload.c (STR): Likewise.
16194 * config/i386/winnt.c (i386_pe_asm_named_section): Likewise.
16195 (i386_pe_file_end): Likewise.
16196 * config/ia64/ia64.c (ia64_in_small_data_p): Likewise.
16197 (ia64_section_type_flags): Likewise.
16198 * config/mips/driver-native.c (host_detect_local_cpu): Likewise.
16199 * config/mips/mips.c (mips_handle_interrupt_attr): Likewise.
16200 (mips16_stub_function_p): Likewise.
16201 (mips_function_rodata_section): Likewise.
16202 * config/msp430/msp430.c (msp430_mcu_name): Likewise.
16203 (msp430_function_section): Likewise.
16204 (msp430_section_type_flags): Likewise.
16205 (msp430_expand_helper): Likewise.
16206 * config/nios2/nios2.c (nios2_small_section_name_p): Likewise.
16207 (nios2_valid_target_attribute_rec): Likewise.
16208 * config/nvptx/mkoffload.c (process): Likewise.
16210 * config/pa/som.h: Likewise.
16211 * config/pdp11/pdp11.c (pdp11_output_ident): Likewise.
16212 * config/riscv/riscv.c (riscv_elf_select_rtx_section): Likewise.
16213 * config/rs6000/rs6000.c (VTABLE_NAME_P): Likewise.
16214 (rs6000_inner_target_options): Likewise.
16215 * config/s390/driver-native.c (s390_host_detect_local_cpu): Likewise.
16216 * config/sparc/driver-sparc.c (host_detect_local_cpu): Likewise.
16217 * config/vax/vax.c (vax_output_int_move): Likewise.
16218 * config/vms/vms-ld.c (startswith): Likewise.
16219 (process_args): Likewise.
16221 * config/vms/vms.c: Likewise.
16223 2021-05-18 Jakub Jelinek <jakub@redhat.com>
16225 PR rtl-optimization/100590
16226 * regcprop.c (copyprop_hardreg_forward_1): Only DCE dead sets if
16227 they are NONJUMP_INSN_P.
16229 2021-05-18 Jakub Jelinek <jakub@redhat.com>
16232 * function.c (push_dummy_function): Set DECL_ARTIFICIAL and
16233 DECL_ASSEMBLER_NAME on the fn_decl.
16235 2021-05-18 Jakub Jelinek <jakub@redhat.com>
16237 PR tree-optimization/94589
16238 * tree-ssa-phiopt.c (spaceship_replacement): Pattern match
16239 phi result used in (res & ~1) == 0 comparison as res >= 0 as
16240 res == 2 would be UB with -ffinite-math-only.
16242 2021-05-18 Martin Liska <mliska@suse.cz>
16244 * Makefile.in: genversion.o should depend on DATESTAMP.
16246 2021-05-18 Claudiu Zissulescu <claziss@synopsys.com>
16248 * config/arc/simdext.md (negv2si2): Remove round bracket.
16250 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
16252 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Define
16253 _Bool as macro expanding to _Bool.
16255 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
16258 * tree.c (build_reference_type_for_mode)
16259 (build_pointer_type_for_mode): Pick pointer mode if MODE argument
16261 (build_reference_type, build_pointer_type): Invoke
16262 build_*_type_for_mode with VOIDmode.
16264 2021-05-17 Andrew MacLeod <amacleod@redhat.com>
16266 PR tree-optimization/100512
16267 * gimple-range-cache.cc (ranger_cache::set_global_range): Mark const
16268 and non-zero pointer ranges as invariant.
16269 * gimple-range.cc (gimple_ranger::range_of_stmt): Remove pointer
16270 processing from here.
16272 2021-05-17 Tom de Vries <tdevries@suse.de>
16275 * config/nvptx/nvptx-protos.h (nvptx_output_atomic_insn): Declare
16276 * config/nvptx/nvptx.c (nvptx_output_barrier)
16277 (nvptx_output_atomic_insn): New function.
16278 (nvptx_print_operand): Add support for 'B'.
16279 * config/nvptx/nvptx.md: Use nvptx_output_atomic_insn for atomic
16282 2021-05-17 Aldy Hernandez <aldyh@redhat.com>
16284 PR tree-optimization/100349
16285 * vr-values.c (bounds_of_var_in_loop): Bail if scev returns
16288 2021-05-17 Tamar Christina <tamar.christina@arm.com>
16290 * config/aarch64/driver-aarch64.c (DEFAULT_ARCH): New.
16291 (host_detect_local_cpu): Use it.
16293 2021-05-17 Martin Liska <mliska@suse.cz>
16295 * doc/invoke.texi: Add 2 missing dots.
16297 2021-05-17 Marius Hillenbrand <mhillen@linux.ibm.com>
16299 PR bootstrap/100552
16300 * configure.ac: Replace pattern substitution with call to sed.
16301 * configure: Regenerate.
16303 2021-05-17 Richard Biener <rguenther@suse.de>
16305 PR middle-end/100582
16306 * tree.c (array_at_struct_end_p): Get to the base of the
16307 reference before looking for the underlying decl.
16309 2021-05-17 Joern Rennecke <joern.rennecke@embecosm.com>
16311 * genoutput.c (validate_insn_alternatives) Make "wrong number of
16312 alternatives" message more specific, and remove assumption on where
16315 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
16317 * config/arm/iterators.md (V16): New iterator.
16318 (VH_cvtto): New iterator.
16319 (v_cmp_result): Added V4HF and V8HF support.
16320 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result>): Use VDQWH.
16321 (vcond<mode><mode>): Likewise.
16322 (vcond_mask_<mode><v_cmp_result>): Likewise.
16323 (vcond<VH_cvtto><mode>): New expander.
16325 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
16327 * config/arm/arm-protos.h (arm_expand_vector_compare): Update
16329 * config/arm/arm.c (arm_expand_vector_compare): Add support for
16331 (arm_expand_vcond): Likewise.
16332 * config/arm/iterators.md (supf): Remove VCMPNEQ_S, VCMPEQQ_S,
16333 VCMPEQQ_N_S, VCMPNEQ_N_S.
16334 (VCMPNEQ, VCMPEQQ, VCMPEQQ_N, VCMPNEQ_N): Remove.
16335 * config/arm/mve.md (@mve_vcmp<mve_cmp_op>q_<mode>): Add '@' prefix.
16336 (@mve_vcmp<mve_cmp_op>q_f<mode>): Likewise.
16337 (@mve_vcmp<mve_cmp_op>q_n_f<mode>): Likewise.
16338 (@mve_vpselq_<supf><mode>): Likewise.
16339 (@mve_vpselq_f<mode>"): Likewise.
16340 * config/arm/neon.md (vec_cmp<mode><v_cmp_result): Enable for MVE
16341 and move to vec-common.md.
16342 (vec_cmpu<mode><mode>): Likewise.
16343 (vcond<mode><mode>): Likewise.
16344 (vcond<V_cvtto><mode>): Likewise.
16345 (vcondu<mode><v_cmp_result>): Likewise.
16346 (vcond_mask_<mode><v_cmp_result>): Likewise.
16347 * config/arm/unspecs.md (VCMPNEQ_U, VCMPNEQ_S, VCMPEQQ_S)
16348 (VCMPEQQ_N_S, VCMPNEQ_N_S, VCMPEQQ_U, CMPEQQ_N_U, VCMPNEQ_N_U)
16349 (VCMPGEQ_N_S, VCMPGEQ_S, VCMPGTQ_N_S, VCMPGTQ_S, VCMPLEQ_N_S)
16350 (VCMPLEQ_S, VCMPLTQ_N_S, VCMPLTQ_S, VCMPCSQ_N_U, VCMPCSQ_U)
16351 (VCMPHIQ_N_U, VCMPHIQ_U): Remove.
16352 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result): Moved
16354 (vec_cmpu<mode><mode>): Likewise.
16355 (vcond<mode><mode>): Likewise.
16356 (vcond<V_cvtto><mode>): Likewise.
16357 (vcondu<mode><v_cmp_result>): Likewise.
16358 (vcond_mask_<mode><v_cmp_result>): Likewise. Added unsafe math
16361 2021-05-17 liuhongt <hongtao.liu@intel.com>
16364 * config/i386/i386.c (ix86_gimple_fold_builtin): Use
16365 gsi_insert_seq_before instead.
16367 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
16369 * doc/sourcebuild.texi (arm_qbit_ok): Rename into...
16370 (arm_sat_ok): ...this.
16372 2021-05-17 Martin Liska <mliska@suse.cz>
16374 * lto-wrapper.c (merge_flto_options): Factor out a new function.
16375 (merge_and_complain): Use it.
16376 (run_gcc): Merge also linker command line -flto=foo argument
16379 2021-05-16 Christophe Lyon <christophe.lyon@linaro.org>
16381 * config/arm/arm.h (CPP_SPEC): Remove error message about
16382 -mlittle-endian/-mbig-endian conflict.
16384 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
16386 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
16387 __ROP_PROTECT__ if -mrop-protect is selected.
16389 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
16391 * config/rs6000/rs6000-internal.h (rs6000_stack): Add
16392 rop_hash_save_offset and rop_hash_size.
16393 * config/rs6000/rs6000-logue.c (rs6000_stack_info): Compute
16394 rop_hash_size and rop_hash_save_offset.
16395 (debug_stack_info): Dump rop_hash_save_offset and rop_hash_size.
16396 (rs6000_emit_prologue): Emit hashst[p] in prologue.
16397 (rs6000_emit_epilogue): Emit hashchk[p] in epilogue.
16398 * config/rs6000/rs6000.md (unspec): Add UNSPEC_HASHST and
16400 (hashst): New define_insn.
16401 (hashchk): Likewise.
16403 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
16405 * config/rs6000/rs6000.c (rs6000_option_override_internal):
16406 Disable shrink wrap when inserting ROP-protect instructions.
16407 * config/rs6000/rs6000.opt (mrop-protect): New option.
16408 (mprivileged): Likewise.
16409 * doc/invoke.texi: Document mrop-protect and mprivileged.
16411 2021-05-15 Hans-Peter Nilsson <hp@axis.com>
16413 * reorg.c (fill_slots_from_thread): Reinstate code typoed out in
16416 2021-05-15 Martin Jambor <mjambor@suse.cz>
16419 2021-05-13 Martin Jambor <mjambor@suse.cz>
16421 PR tree-optimization/100453
16422 * tree-sra.c (sra_modify_assign): All const base accesses do not
16423 need refreshing, not just those from decl_pool.
16424 (sra_modify_assign): Do not refresh into a const base decl.
16426 2021-05-15 Jakub Jelinek <jakub@redhat.com>
16428 PR rtl-optimization/100342
16429 * regcprop.c (copy_value): When copying a source reg in a wider
16430 mode than it has recorded for the value, adjust recorded destination
16431 mode too or punt if !REG_CAN_CHANGE_MODE_P.
16433 2021-05-14 Jason Merrill <jason@redhat.com>
16435 * intl.h: Add comments.
16437 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16439 * config/aarch64/aarch64-simd.md
16440 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Split into...
16441 (aarch64_sqdmlsl2_lane<mode>_internal): ... This...
16442 (aarch64_sqdmlal2_lane<mode>_internal): ... And this.
16443 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Split into ...
16444 (aarch64_sqdmlsl2_laneq<mode>_internal): ... This...
16445 (aarch64_sqdmlal2_laneq<mode>_internal): ... And this.
16446 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>_internal): Split into...
16447 (aarch64_sqdmlsl2_n<mode>_internal): ... This...
16448 (aarch64_sqdmlal2_n<mode>_internal): ... And this.
16450 2021-05-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16453 * config/arm/arm_neon.h (vtst_s8): Replace call to vtst builtin with it's
16454 boolean logic equivalent.
16455 (vtst_s16): Likewise.
16456 (vtst_s32): Likewise.
16457 (vtst_u8): Likewise.
16458 (vtst_u16): Likewise.
16459 (vtst_u32): Likewise.
16460 (vtst_p8): Likewise.
16461 (vtst_p16): Likewise.
16462 (vtstq_s8): Likewise.
16463 (vtstq_s16): Likewise.
16464 (vtstq_s32): Likewise.
16465 (vtstq_u8): Likewise.
16466 (vtstq_u16): Likewise.
16467 (vtstq_u32): Likewise.
16468 (vtstq_p8): Likewise.
16469 (vtstq_p16): Likewise.
16470 * config/arm/arm_neon_builtins.def: Remove entry for vtst.
16471 * config/arm/neon.md (neon_vtst<mode>): Remove pattern.
16473 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16475 * config/aarch64/aarch64-simd.md (aarch64_sqdmlal2<mode>): Merge into...
16476 (aarch64_sqdml<SBINQOPS:as>l2<mode>): ... This.
16477 (aarch64_sqdmlsl2<mode>): Delete.
16478 (aarch64_sqdmlal2_lane<mode>): Merge this...
16479 (aarch64_sqdmlsl2_lane<mode>): ... And this...
16480 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>): ... Into this.
16481 (aarch64_sqdmlal2_laneq<mode>): Merge this...
16482 (aarch64_sqdmlsl2_laneq<mode>): ... And this...
16483 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>): ... Into this.
16484 (aarch64_sqdmlal2_n<mode>): Merge this...
16485 (aarch64_sqdmlsl2_n<mode>): ... And this...
16486 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>): ... Into this.
16488 2021-05-13 Martin Sebor <msebor@redhat.com>
16490 PR middle-end/100574
16491 * builtins.c (access_ref::get_ref): Improve detection of PHIs with
16492 all null arguments.
16494 2021-05-13 Martin Sebor <msebor@redhat.com>
16496 PR tree-optimization/93100
16497 PR middle-end/98583
16498 * tree-ssa-uninit.c (check_defs): Exclude intrinsic functions that
16499 don't modify referenced objects.
16501 2021-05-13 Martin Jambor <mjambor@suse.cz>
16503 PR tree-optimization/100453
16504 * tree-sra.c (sra_modify_assign): All const base accesses do not
16505 need refreshing, not just those from decl_pool.
16506 (sra_modify_assign): Do not refresh into a const base decl.
16508 2021-05-13 Martin Liska <mliska@suse.cz>
16510 * tree-ssa-dom.c: Remove m_simplifier.
16512 2021-05-13 Richard Earnshaw <rearnsha@arm.com>
16515 * config/arm/arm.c (arm_canonicalize_comparison): Correctly
16516 canonicalize DImode inequality comparisons against the
16517 maximum integral value.
16519 2021-05-13 Jakub Jelinek <jakub@redhat.com>
16521 PR tree-optimization/98856
16522 * config/i386/i386.c (ix86_shift_rotate_cost): Add CODE argument.
16523 Expect V2DI and V4DI arithmetic right shifts to be emulated.
16524 (ix86_rtx_costs, ix86_add_stmt_cost): Adjust ix86_shift_rotate_cost
16526 * config/i386/i386-expand.c (expand_vec_perm_2perm_interleave,
16527 expand_vec_perm_2perm_pblendv): New functions.
16528 (ix86_expand_vec_perm_const_1): Use them.
16529 * config/i386/sse.md (ashr<mode>3<mask_name>): Rename to ...
16530 (<mask_codefor>ashr<mode>3<mask_name>): ... this.
16531 (ashr<mode>3): New define_expand with VI248_AVX512BW iterator.
16532 (ashrv4di3): New define_expand.
16533 (ashrv2di3): Change condition to TARGET_SSE2, handle !TARGET_XOP
16534 and !TARGET_AVX512VL expansion.
16536 2021-05-13 Uroš Bizjak <ubizjak@gmail.com>
16539 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Force mode
16540 sizes < 16 to a register when constructing vpcmov pattern.
16541 * config/i386/mmx.md (*xop_pcmov_<mode>): Use MMXMODE124 mode.
16543 2021-05-13 Martin Liska <mliska@suse.cz>
16545 * gcov-io.c (gcov_write_block): Remove.
16546 (gcov_write_words): Likewise.
16547 (gcov_read_words): Re-implement using gcov_read_bytes.
16548 (gcov_allocate): Remove.
16549 (GCOV_BLOCK_SIZE): Likewise.
16550 (struct gcov_var): Remove most of the fields.
16551 (gcov_position): Implement with ftell.
16552 (gcov_rewrite): Remove setting of start and offset fields.
16553 (from_file): Re-format.
16554 (gcov_open): Remove setbuf call. It should not be needed.
16555 (gcov_close): Remove internal buffer handling.
16556 (gcov_magic): Use __builtin_bswap32.
16557 (gcov_write_counter): Use directly gcov_write_unsigned.
16558 (gcov_write_string): Use direct fwrite and do not round
16560 (gcov_seek): Use directly fseek.
16561 (gcov_write_tag): Use gcov_write_unsigned directly.
16562 (gcov_write_length): Likewise.
16563 (gcov_write_tag_length): Likewise.
16564 (gcov_read_bytes): Use directly fread.
16565 (gcov_read_unsigned): Use gcov_read_words.
16566 (gcov_read_counter): Likewise.
16567 (gcov_read_string): Use gcov_read_bytes.
16568 * gcov-io.h (GCOV_WORD_SIZE): Adjust to reflect
16569 that size is not in bytes, but words (4B).
16570 (GCOV_TAG_FUNCTION_LENGTH): Likewise.
16571 (GCOV_TAG_ARCS_LENGTH): Likewise.
16572 (GCOV_TAG_ARCS_NUM): Likewise.
16573 (GCOV_TAG_COUNTER_LENGTH): Likewise.
16574 (GCOV_TAG_COUNTER_NUM): Likewise.
16575 (GCOV_TAG_SUMMARY_LENGTH): Likewise.
16577 2021-05-13 liuhongt <hongtao.liu@intel.com>
16580 * config/i386/sse.md (ssedoublevecmode): Add attribute for
16581 V64QI/V32HI/V16SI/V4DI.
16582 (ssehalfvecmode): Add attribute for V2DI/V2DF.
16583 (*vec_concatv4si_0): Extend to VI124_128.
16584 (*vec_concat<mode>_0): New pre-reload splitter.
16585 * config/i386/predicates.md (movq_parallel): New predicate.
16587 2021-05-13 Alexandre Oliva <oliva@adacore.com>
16589 * targhooks.c (default_zero_call_used_regs): Retry using
16590 successfully-zeroed registers as sources.
16592 2021-05-12 Tobias Burnus <tobias@codesourcery.com>
16594 * omp-low.c (finish_taskreg_scan): Use the proper detach decl.
16596 2021-05-12 Aldy Hernandez <aldyh@redhat.com>
16599 * gimple-range.cc (range_of_builtin_call): Skip out on
16600 processing __builtin_clz when varying.
16602 2021-05-12 Tom de Vries <tdevries@suse.de>
16605 * config/nvptx/nvptx-opts.h (enum ptx_version): New enum.
16606 * config/nvptx/nvptx.c (nvptx_file_start): Print .version according
16607 to ptx_version_option.
16608 * config/nvptx/nvptx.h (TARGET_PTX_6_3): Define.
16609 * config/nvptx/nvptx.md (define_insn "nvptx_shuffle<mode>")
16610 (define_insn "nvptx_vote_ballot"): Use sync variant for
16612 * config/nvptx/nvptx.opt (ptx_version): Add enum.
16613 (mptx): Add option.
16614 * doc/invoke.texi (Nvidia PTX Options): Add mptx item.
16616 2021-05-12 Richard Biener <rguenther@suse.de>
16618 PR tree-optimization/100566
16619 * tree-ssa-sccvn.c (dominated_by_p_w_unex): Properly handle
16620 allow_back for all edge queries.
16622 2021-05-12 liuhongt <hongtao.liu@intel.com>
16625 * config/i386/sse.md (<sse4_1_avx2>_pblendvb): Add
16626 splitters for pblendvb of NOT mask register.
16628 2021-05-12 Richard Biener <rguenther@suse.de>
16630 PR tree-optimization/100519
16631 * tree-ssa-reassoc.c (can_associate_p): Split into...
16632 (can_associate_op_p): ... this
16633 (can_associate_type_p): ... and this.
16634 (is_reassociable_op): Call can_associate_op_p.
16635 (break_up_subtract_bb): Call the appropriate predicates.
16636 (reassociate_bb): Likewise.
16638 2021-05-12 Martin Liska <mliska@suse.cz>
16640 * lto-wrapper.c (merge_and_complain): Merge -flto=arg options.
16641 (run_gcc): Use -flto argument detection for merged
16644 2021-05-12 Martin Liska <mliska@suse.cz>
16646 * lto-wrapper.c (print_lto_docs_link): New function.
16647 (run_gcc): Print warning about missing job server detection
16648 after we know NR of partitions. Do the same for -flto{,=1}.
16649 * opts.c (get_option_html_page): Support -flto option.
16651 2021-05-12 Martin Liska <mliska@suse.cz>
16653 * lto-wrapper.c (get_options_from_collect_gcc_options): Change
16655 (append_option): Remove.
16656 (find_option): Rework to use the vector type.
16657 (remove_option): Remove.
16658 (merge_and_complain): Use vectors for cl_decoded_option data
16660 (append_compiler_options): Likewise.
16661 (append_diag_options): Likewise.
16662 (append_linker_options): Likewise.
16663 (append_offload_options): Likewise.
16664 (compile_offload_image): Likewise.
16665 (compile_images_for_offload_targets): Likewise.
16666 (find_and_merge_options): Likewise.
16667 (run_gcc): Likewise.
16669 2021-05-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
16672 * dwarf2out.c (dwarf2out_finish): Set
16673 have_multiple_function_sections with multi-range text_section.
16675 2021-05-12 Martin Liska <mliska@suse.cz>
16677 PR bootstrap/100560
16678 * Makefile.in: Remove version.h from linker command line.
16680 2021-05-12 Richard Biener <rguenther@suse.de>
16682 PR middle-end/100547
16683 * rtl.h (rtvec_alloc): Make argument size_t.
16684 * rtl.c (rtvec_alloc): Verify the count is less than INT_MAX.
16686 2021-05-12 Jakub Jelinek <jakub@redhat.com>
16688 PR middle-end/100508
16689 * cfgexpand.c (expand_debug_expr): For DEBUG_EXPR_DECL with vector
16690 type, don't reuse DECL_RTL if it has different mode, instead force
16691 creation of a new DEBUG_EXPR.
16693 2021-05-12 Jakub Jelinek <jakub@redhat.com>
16694 Marc Glisse <marc.glisse@inria.fr>
16696 PR tree-optimization/94589
16697 * match.pd ((X & Y) == X -> (X & ~Y) == 0,
16698 (X | Y) == Y -> (X & ~Y) == 0): New GIMPLE simplifications.
16700 2021-05-12 Uroš Bizjak <ubizjak@gmail.com>
16703 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Handle V2SF mode.
16704 * config/i386/mmx.md (MMXMODE124): New mode iterator.
16706 (mmxintvecmode): New mode attribute.
16707 (mmxintvecmodelower): Ditto.
16708 (*mmx_maskcmpv2sf3_comm): New insn pattern.
16709 (*mmx_maskcmpv2sf3): Ditto.
16710 (vec_cmpv2sfv2si): New expander.
16711 (vcond<V2FI:mode>v2si): Ditto.
16712 (mmx_vlendvps): New insn pattern.
16713 (vcond<MMXMODE124:mode><MMXMODEI:mode>): Also handle V2SFmode.
16714 (vcondu<MMXMODE124:mode><MMXMODEI:mode>): Ditto.
16715 (vcond_mask_<mode><mmxintvecmodelower>): Ditto.
16717 2021-05-11 Martin Sebor <msebor@redhat.com>
16719 PR middle-end/21433
16720 * expr.c (expand_expr_real_1): Replace unreachable code with an assert.
16722 2021-05-11 Richard Biener <rguenther@suse.de>
16724 * gimple-fold.c (gimple_fold_call): Do not call
16725 maybe_fold_reference on call arguments or the static chain.
16726 (fold_stmt_1): Do not call maybe_fold_reference on GIMPLE_ASM
16729 2021-05-11 Martin Liska <mliska@suse.cz>
16731 * builtins.def (DEF_HSAIL_BUILTIN): Remove.
16732 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
16733 (DEF_HSAIL_SAT_BUILTIN): Likewise.
16734 (DEF_HSAIL_INTR_BUILTIN): Likewise.
16735 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
16736 * doc/frontends.texi: Remove BRIG.
16737 * doc/install.texi: Likewise.
16738 * doc/invoke.texi: Likewise.
16739 * doc/standards.texi: Likewise.
16740 * brig-builtins.def: Removed.
16741 * brig/ChangeLog: Removed.
16742 * brig/Make-lang.in: Removed.
16743 * brig/brig-builtins.h: Removed.
16744 * brig/brig-c.h: Removed.
16745 * brig/brig-lang.c: Removed.
16746 * brig/brigfrontend/brig-arg-block-handler.cc: Removed.
16747 * brig/brigfrontend/brig-atomic-inst-handler.cc: Removed.
16748 * brig/brigfrontend/brig-basic-inst-handler.cc: Removed.
16749 * brig/brigfrontend/brig-branch-inst-handler.cc: Removed.
16750 * brig/brigfrontend/brig-cmp-inst-handler.cc: Removed.
16751 * brig/brigfrontend/brig-code-entry-handler.cc: Removed.
16752 * brig/brigfrontend/brig-code-entry-handler.h: Removed.
16753 * brig/brigfrontend/brig-comment-handler.cc: Removed.
16754 * brig/brigfrontend/brig-control-handler.cc: Removed.
16755 * brig/brigfrontend/brig-copy-move-inst-handler.cc: Removed.
16756 * brig/brigfrontend/brig-cvt-inst-handler.cc: Removed.
16757 * brig/brigfrontend/brig-fbarrier-handler.cc: Removed.
16758 * brig/brigfrontend/brig-function-handler.cc: Removed.
16759 * brig/brigfrontend/brig-function.cc: Removed.
16760 * brig/brigfrontend/brig-function.h: Removed.
16761 * brig/brigfrontend/brig-inst-mod-handler.cc: Removed.
16762 * brig/brigfrontend/brig-label-handler.cc: Removed.
16763 * brig/brigfrontend/brig-lane-inst-handler.cc: Removed.
16764 * brig/brigfrontend/brig-machine.c: Removed.
16765 * brig/brigfrontend/brig-machine.h: Removed.
16766 * brig/brigfrontend/brig-mem-inst-handler.cc: Removed.
16767 * brig/brigfrontend/brig-module-handler.cc: Removed.
16768 * brig/brigfrontend/brig-queue-inst-handler.cc: Removed.
16769 * brig/brigfrontend/brig-seg-inst-handler.cc: Removed.
16770 * brig/brigfrontend/brig-signal-inst-handler.cc: Removed.
16771 * brig/brigfrontend/brig-to-generic.cc: Removed.
16772 * brig/brigfrontend/brig-to-generic.h: Removed.
16773 * brig/brigfrontend/brig-util.cc: Removed.
16774 * brig/brigfrontend/brig-util.h: Removed.
16775 * brig/brigfrontend/brig-variable-handler.cc: Removed.
16776 * brig/brigfrontend/hsa-brig-format.h: Removed.
16777 * brig/brigfrontend/phsa.h: Removed.
16778 * brig/brigspec.c: Removed.
16779 * brig/config-lang.in: Removed.
16780 * brig/gccbrig.texi: Removed.
16781 * brig/lang-specs.h: Removed.
16782 * brig/lang.opt: Removed.
16784 2021-05-11 Richard Biener <rguenther@suse.de>
16787 * ipa-param-manipulation.c
16788 (ipa_param_body_adjustments::modify_call_stmt): Avoid
16789 altering SSA_NAME_DEF_STMT by adjusting the calls LHS
16790 via gimple_call_lhs_ptr.
16792 2021-05-11 Alex Coplan <alex.coplan@arm.com>
16795 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear):
16796 Avoid emitting CFA adjusts on the sp if we have the fp.
16798 2021-05-11 Richard Sandiford <richard.sandiford@arm.com>
16800 * config/aarch64/iterators.md (VMUL_CHANGE_NLANES): Delete.
16801 (VMULD): New iterator.
16802 (VCOND): Handle V4HF and V8HF.
16803 (VCONQ): Fix entry for V2SF.
16804 * config/aarch64/aarch64-simd.md (mul_lane<mode>3): Use VMULD
16805 instead of VMUL. Use a 64-bit vector mode for the indexed operand.
16806 (*aarch64_mul3_elt_<vswap_width_name><mode>): Merge with...
16807 (mul_laneq<mode>3): ...this define_insn. Use VMUL instead of VDQSF.
16808 Use a 128-bit vector mode for the indexed operand. Use stype for
16809 the scheduling type.
16811 2021-05-11 Richard Biener <rguenther@suse.de>
16813 * gimple-fold.c (maybe_fold_reference): Only return
16814 is_gimple_min_invariant values.
16816 2021-05-11 Richard Biener <rguenther@suse.de>
16818 PR middle-end/100509
16819 * gimple-fold.c (fold_gimple_assign): Only call
16820 get_symbol_constant_value on register type symbols.
16822 2021-05-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
16823 Joe Ramsay <joe.ramsay@arm.com>
16826 * config/arm/arm_mve.h (__arm_vstrwq_scatter_offset): Fix wrong arguments.
16827 (__arm_vcmpneq): Remove duplicate definition.
16828 (__arm_vstrwq_scatter_offset_p): Likewise.
16829 (__arm_vmaxq_x): Likewise.
16830 (__arm_vmlsdavaq): Likewise.
16831 (__arm_vmlsdavaxq): Likewise.
16832 (__arm_vmlsdavq_p): Likewise.
16833 (__arm_vmlsdavxq_p): Likewise.
16834 (__arm_vrmlaldavhaq): Likewise.
16835 (__arm_vstrbq_p): Likewise.
16836 (__arm_vstrbq_scatter_offset): Likewise.
16837 (__arm_vstrbq_scatter_offset_p): Likewise.
16838 (__arm_vstrdq_scatter_offset): Likewise.
16839 (__arm_vstrdq_scatter_offset_p): Likewise.
16840 (__arm_vstrdq_scatter_shifted_offset): Likewise.
16841 (__arm_vstrdq_scatter_shifted_offset_p): Likewise.
16843 2021-05-11 Jakub Jelinek <jakub@redhat.com>
16845 PR middle-end/100471
16846 * omp-low.c (lower_omp_task_reductions): For OMP_TASKLOOP, if data
16847 is 0, bypass the reduction loop including
16848 GOMP_taskgroup_reduction_unregister call.
16850 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
16852 * config/rs6000/rs6000.c (struct rs6000_cost_data): New member
16853 costing_for_scalar.
16854 (rs6000_density_test): Early return if costing_for_scalar is true.
16855 (rs6000_init_cost): Init costing_for_scalar of rs6000_cost_data.
16857 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
16859 * doc/tm.texi: Regenerated.
16860 * target.def (init_cost): Add new parameter costing_for_scalar.
16861 * targhooks.c (default_init_cost): Adjust for new parameter.
16862 * targhooks.h (default_init_cost): Likewise.
16863 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
16864 (vect_compute_single_scalar_iteration_cost): Likewise.
16865 (vect_analyze_loop_2): Likewise.
16866 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
16867 (vect_bb_vectorization_profitable_p): Likewise.
16868 * tree-vectorizer.h (init_cost): Likewise.
16869 * config/aarch64/aarch64.c (aarch64_init_cost): Likewise.
16870 * config/i386/i386.c (ix86_init_cost): Likewise.
16871 * config/rs6000/rs6000.c (rs6000_init_cost): Likewise.
16873 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
16875 * config/rs6000/rs6000.c (rs6000_vect_nonmem): Renamed to
16876 vect_nonmem and moved into...
16877 (struct rs6000_cost_data): ...here.
16878 (rs6000_init_cost): Use vect_nonmem of cost_data instead.
16879 (rs6000_add_stmt_cost): Likewise.
16880 (rs6000_finish_cost): Likewise.
16882 2021-05-10 Eric Botcazou <ebotcazou@adacore.com>
16884 * range-op.cc (get_bool_state): Adjust head comment.
16885 (operator_not_equal::op1_range): Fix comment.
16886 (operator_bitwise_xor::op1_range): Remove call to gcc_unreachable.
16888 2021-05-10 Martin Sebor <msebor@redhat.com>
16890 PR middle-end/100425
16891 PR middle-end/100510
16892 * gimple-ssa-warn-alloca.c (pass_walloca::firast_time_p): Rename...
16893 (pass_walloca::xlimit_certain_p): ...to this.
16894 (pass_walloca::gate): Execute for any kind of handled warning.
16895 (pass_walloca::execute): Avoid issuing "maybe" and "unbounded"
16896 warnings when xlimit_certain_p is set.
16898 2021-05-10 Pat Haugen <pthaugen@linux.ibm.com>
16900 * config/rs6000/rs6000.c (rs6000_ira_change_pseudo_allocno_class):
16901 Return ALTIVEC_REGS if that is best_class.
16902 (rs6000_compute_pressure_classes): Add ALTIVEC_REGS.
16904 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16906 * config/arm/arm.h (CPP_SPEC): Remove error message about
16909 2021-05-10 Martin Jambor <mjambor@suse.cz>
16911 * ipa-prop.h (IPA_NODE_REF): Removed.
16912 (IPA_NODE_REF_GET_CREATE): Likewise.
16913 (IPA_EDGE_REF): Likewise.
16914 (IPA_EDGE_REF_GET_CREATE): Likewise.
16915 (IS_VALID_JUMP_FUNC_INDEX): Likewise.
16916 * ipa-cp.c (print_all_lattices): Replaced IPA_NODE_REF with a direct
16917 use of ipa_node_params_sum.
16918 (ipcp_versionable_function_p): Likewise.
16919 (push_node_to_stack): Likewise.
16920 (pop_node_from_stack): Likewise.
16921 (set_single_call_flag): Replaced two IPA_NODE_REF with one single
16922 direct use of ipa_node_params_sum.
16923 (initialize_node_lattices): Replaced IPA_NODE_REF with a direct use of
16924 ipa_node_params_sum.
16925 (ipa_context_from_jfunc): Replaced IPA_EDGE_REF with a direct use of
16927 (ipcp_verify_propagated_values): Replaced IPA_NODE_REF with a direct
16928 use of ipa_node_params_sum.
16929 (self_recursively_generated_p): Likewise.
16930 (propagate_scalar_across_jump_function): Likewise.
16931 (propagate_context_across_jump_function): Replaced IPA_EDGE_REF with a
16932 direct use of ipa_edge_args_sum, moved the lookup after the early
16933 exit. Replaced IPA_NODE_REF with a direct use of ipa_node_params_sum.
16934 (propagate_bits_across_jump_function): Replaced IPA_NODE_REF with
16935 direct uses of ipa_node_params_sum.
16936 (propagate_vr_across_jump_function): Likewise.
16937 (propagate_aggregate_lattice): Likewise.
16938 (propagate_aggs_across_jump_function): Likewise.
16939 (propagate_constants_across_call): Likewise, also replaced
16940 IPA_EDGE_REF with a direct use of ipa_edge_args_sum.
16941 (good_cloning_opportunity_p): Replaced IPA_NODE_REF with a direct use
16942 of ipa_node_params_sum.
16943 (estimate_local_effects): Likewise.
16944 (add_all_node_vals_to_toposort): Likewise.
16945 (propagate_constants_topo): Likewise.
16946 (ipcp_propagate_stage): Likewise.
16947 (ipcp_discover_new_direct_edges): Likewise.
16948 (calls_same_node_or_its_all_contexts_clone_p): Likewise.
16949 (cgraph_edge_brings_value_p): Likewise (in both overloaded functions).
16950 (get_info_about_necessary_edges): Likewise.
16951 (want_remove_some_param_p): Likewise.
16952 (create_specialized_node): Likewise.
16953 (self_recursive_pass_through_p): Likewise.
16954 (self_recursive_agg_pass_through_p): Likewise.
16955 (find_more_scalar_values_for_callers_subset): Likewise and also
16956 replaced IPA_EDGE_REF with direct uses of ipa_edge_args_sum, in one
16957 case replacing two of those with a single query.
16958 (find_more_contexts_for_caller_subset): Likewise for the
16959 ipa_polymorphic_call_context overload.
16960 (intersect_aggregates_with_edge): Replaced IPA_EDGE_REF with a direct
16961 use of ipa_edge_args_sum. Replaced IPA_NODE_REF with direct uses of
16962 ipa_node_params_sum.
16963 (find_aggregate_values_for_callers_subset): Likewise, also reusing
16964 results of ipa_edge_args_sum->get.
16965 (cgraph_edge_brings_all_scalars_for_node): Replaced IPA_NODE_REF with
16966 direct uses of ipa_node_params_sum, replaced IPA_EDGE_REF with a
16967 direct use of ipa_edge_args_sum.
16968 (cgraph_edge_brings_all_agg_vals_for_node): Likewise, moved node
16969 summary query after the early exit and reused the result later.
16970 (decide_about_value): Replaced IPA_NODE_REF with a direct use of
16971 ipa_node_params_sum.
16972 (decide_whether_version_node): Likewise. Removed re-querying for
16973 summaries after cloning.
16974 (spread_undeadness): Replaced IPA_NODE_REF with a direct use of
16975 ipa_node_params_sum.
16976 (has_undead_caller_from_outside_scc_p): Likewise, reusing results of
16978 (identify_dead_nodes): Likewise.
16979 (ipcp_store_bits_results): Replaced IPA_NODE_REF with direct uses of
16980 ipa_node_params_sum.
16981 (ipcp_store_vr_results): Likewise.
16982 * ipa-fnsummary.c (evaluate_properties_for_edge): Likewise.
16983 (ipa_fn_summary_t::duplicate): Likewise.
16984 (analyze_function_body): Likewise.
16985 (estimate_calls_size_and_time): Likewise.
16986 (ipa_cached_call_context::duplicate_from): Likewise.
16987 (ipa_call_context::equal_to): Likewise.
16988 (remap_edge_params): Likewise.
16989 (ipa_merge_fn_summary_after_inlining): Likewise.
16990 (inline_read_section): Likewise.
16991 * ipa-icf.c (sem_function::param_used_p): Likewise.
16992 * ipa-modref.c (compute_parm_map): Likewise.
16993 (compute_parm_map): Replaced IPA_EDGE_REF with a direct use of
16995 (get_access_for_fnspec): Replaced IPA_NODE_REF with a direct use of
16996 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
16998 * ipa-profile.c (check_argument_count): Likewise.
16999 * ipa-prop.c (ipa_alloc_node_params): Replaced IPA_NODE_REF_GET_CREATE
17000 with a direct use of ipa_node_params_sum.
17001 (ipa_initialize_node_params): Likewise.
17002 (ipa_print_node_jump_functions_for_edge): Replaced IPA_EDGE_REF with a
17003 direct use of ipa_edge_args_sum and reused the query result.
17004 (ipa_compute_jump_functions_for_edge): Replaced IPA_NODE_REF with a
17005 direct use of ipa_node_params_sum and replaced IPA_EDGE_REF with a
17006 direct use of ipa_edge_args_sum.
17007 (ipa_note_param_call): Replaced IPA_NODE_REF with a direct use of
17008 ipa_node_params_sum and reused the result of the query.
17009 (ipa_analyze_node): Likewise.
17010 (ipa_analyze_controlled_uses): Replaced IPA_NODE_REF with a direct use
17011 of ipa_node_params_sum.
17012 (update_jump_functions_after_inlining): Replaced IPA_EDGE_REF with
17013 direct uses of ipa_edge_args_sum.
17014 (update_indirect_edges_after_inlining): Replaced IPA_NODE_REF with
17015 direct uses of ipa_node_params_sum and replaced IPA_EDGE_REF with a
17016 direct use of ipa_edge_args_sum. Removed superficial re-querying the
17018 (propagate_controlled_uses): Replaced IPA_NODE_REF with direct uses of
17019 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
17021 (ipa_propagate_indirect_call_infos): Replaced IPA_EDGE_REF with a
17022 direct use of ipa_edge_args_sum.
17023 (ipa_edge_args_sum_t::duplicate): Replaced IPA_NODE_REF with a direct
17024 use of ipa_node_params_sum.
17025 (ipa_print_node_params): Likewise.
17026 (ipa_write_node_info): Likewise and also replaced IPA_EDGE_REF with
17027 direct uses of ipa_edge_args_sum.
17028 (ipa_read_edge_info): Replaced IPA_EDGE_REF with a direct use of
17030 (ipa_read_node_info): Replaced IPA_NODE_REF with a direct use of
17031 ipa_node_params_sum.
17032 (ipa_prop_write_jump_functions): Likewise. Move variable node to the
17033 scopes where it is used.
17035 2021-05-10 Uroš Bizjak <ubizjak@gmail.com>
17037 * config/i386/i386-expand.c (ix86_expand_sse_movcc)
17038 <case E_V2SImode>: Force op_true to register.
17040 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17042 * config/arm/iterators.md (MVE_FP_COMPARISONS): New.
17043 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_f<mode>)
17044 (mve_vcmp<mve_cmp_op>q_n_f<mode>): New, merge all vcmp_*f*
17046 (mve_vcmpeqq_f<mode>, mve_vcmpeqq_n_f<mode>, mve_vcmpgeq_f<mode>)
17047 (mve_vcmpgeq_n_f<mode>, mve_vcmpgtq_f<mode>)
17048 (mve_vcmpgtq_n_f<mode>, mve_vcmpleq_f<mode>)
17049 (mve_vcmpleq_n_f<mode>, mve_vcmpltq_f<mode>)
17050 (mve_vcmpltq_n_f<mode>, mve_vcmpneq_f<mode>)
17051 (mve_vcmpneq_n_f<mode>): Remove.
17052 * config/arm/unspecs.md (VCMPEQQ_F, VCMPEQQ_N_F, VCMPGEQ_F)
17053 (VCMPGEQ_N_F, VCMPGTQ_F, VCMPGTQ_N_F, VCMPLEQ_F, VCMPLEQ_N_F)
17054 (VCMPLTQ_F, VCMPLTQ_N_F, VCMPNEQ_F, VCMPNEQ_N_F): Remove.
17056 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17058 * config/arm/iterators.md (MVE_COMPARISONS): New.
17060 (mve_cmp_type): New.
17061 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_<mode>): New, merge all
17063 (mve_vcmpneq_<mode>, mve_vcmpcsq_n_<mode>, mve_vcmpcsq_<mode>)
17064 (mve_vcmpeqq_n_<mode>, mve_vcmpeqq_<mode>, mve_vcmpgeq_n_<mode>)
17065 (mve_vcmpgeq_<mode>, mve_vcmpgtq_n_<mode>, mve_vcmpgtq_<mode>)
17066 (mve_vcmphiq_n_<mode>, mve_vcmphiq_<mode>, mve_vcmpleq_n_<mode>)
17067 (mve_vcmpleq_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
17068 (mve_vcmpneq_n_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
17069 (mve_vcmpneq_n_<mode>): Remove.
17071 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17073 * config/arm/arm_mve.h (__arm_vcmp*): Remove 's' suffix.
17074 * config/arm/arm_mve_builtins.def (vcmp*): Remove 's' suffix.
17075 * config/arm/mve.md (mve_vcmp*): Remove 's' suffix in pattern
17078 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17080 * config/arm/arm_mve_builtins.def (vcmpneq_u): Remove.
17081 (vcmpneq_n_u): Likewise.
17082 (vcmpeqq_u,): Likewise.
17083 (vcmpeqq_n_u): Likewise.
17084 * config/arm/iterators.md (supf): Remove VCMPNEQ_U, VCMPEQQ_U,
17085 VCMPEQQ_N_U and VCMPNEQ_N_U.
17086 * config/arm/mve.md (mve_vcmpneq): Remove <supf> iteration.
17087 (mve_vcmpeqq_n): Likewise.
17088 (mve_vcmpeqq): Likewise.
17089 (mve_vcmpneq_n): Likewise.
17091 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
17093 * config/arm/arm_mve.h (__arm_vcmpeq*u*, __arm_vcmpne*u*): Call
17094 the 's' version of the builtin.
17096 2021-05-10 Richard Biener <rguenther@suse.de>
17098 PR tree-optimization/100492
17099 * tree-loop-distribution.c (find_seed_stmts_for_distribution):
17100 Find nothing when the loop contains an irreducible region.
17102 2021-05-10 Richard Biener <rguenther@suse.de>
17104 PR middle-end/100464
17106 * gimple-fold.c (canonicalize_constructor_val): Do not set
17109 2021-05-10 Richard Biener <rguenther@suse.de>
17111 PR tree-optimization/100434
17112 * tree-ssa-dse.c (initialize_ao_ref_for_dse): Handle
17114 (dse_optimize_stmt): Handle call LHS by dropping the
17115 LHS or the whole call if it doesn't have other
17117 (pass_dse::execute): Adjust.
17119 2021-05-10 Martin Liska <mliska@suse.cz>
17121 * Makefile.in: Add missing genversion rule.
17123 2021-05-10 Alex Coplan <alex.coplan@arm.com>
17126 * config/arm/mve.md (*mve_mov<mode>): Simplify output code. Use
17127 vldrw.u32 and vstrw.32 for V2D[IF]mode loads and stores.
17129 2021-05-10 Martin Liska <mliska@suse.cz>
17131 * builtins.c (is_builtin_name): Use startswith
17132 function instead of strncmp.
17133 * collect2.c (main): Likewise.
17134 (has_lto_section): Likewise.
17135 (scan_libraries): Likewise.
17136 * coverage.c (coverage_checksum_string): Likewise.
17137 (coverage_init): Likewise.
17138 * dwarf2out.c (is_cxx): Likewise.
17139 (gen_compile_unit_die): Likewise.
17140 * gcc-ar.c (main): Likewise.
17141 * gcc.c (init_spec): Likewise.
17142 (read_specs): Likewise.
17143 (execute): Likewise.
17144 (check_live_switch): Likewise.
17145 * genattrtab.c (write_attr_case): Likewise.
17146 (IS_ATTR_GROUP): Likewise.
17147 * gencfn-macros.c (main): Likewise.
17148 * gengtype.c (type_for_name): Likewise.
17149 (gen_rtx_next): Likewise.
17150 (get_file_langdir): Likewise.
17151 (write_local): Likewise.
17152 * genmatch.c (get_operator): Likewise.
17153 (get_operand_type): Likewise.
17154 (expr::gen_transform): Likewise.
17155 * genoutput.c (validate_optab_operands): Likewise.
17156 * incpath.c (add_sysroot_to_chain): Likewise.
17157 * langhooks.c (lang_GNU_C): Likewise.
17158 (lang_GNU_CXX): Likewise.
17159 (lang_GNU_Fortran): Likewise.
17160 (lang_GNU_OBJC): Likewise.
17161 * lto-wrapper.c (run_gcc): Likewise.
17162 * omp-general.c (omp_max_simt_vf): Likewise.
17163 * omp-low.c (omp_runtime_api_call): Likewise.
17164 * opts-common.c (parse_options_from_collect_gcc_options): Likewise.
17165 * read-rtl-function.c (function_reader::read_rtx_operand_r): Likewise.
17166 * real.c (real_from_string): Likewise.
17167 * selftest.c (assert_str_startswith): Likewise.
17168 * timevar.c (timer::validate_phases): Likewise.
17169 * tree.c (get_file_function_name): Likewise.
17170 * ubsan.c (ubsan_use_new_style_p): Likewise.
17171 * varasm.c (default_function_rodata_section): Likewise.
17172 (incorporeal_function_p): Likewise.
17173 (default_section_type_flags): Likewise.
17174 * system.h (startswith): Define startswith.
17176 2021-05-10 Martin Liska <mliska@suse.cz>
17178 * bitmap.h (class auto_bitmap): Remove
17179 __cplusplus >= 201103.
17180 * config/aarch64/aarch64.c: Likewise.
17181 * gimple-ssa-store-merging.c (store_immediate_info::store_immediate_info):
17183 * sbitmap.h: Likewise.
17185 2021-05-10 Martin Liska <mliska@suse.cz>
17187 * Makefile.in: Rename gcov-iov to genversion and depend
17188 on version.h (instead of gcov-iov.h).
17189 * gcov-io.h: Include version.h instread of gcov-iov.h.
17190 * gengtype-state.c (read_state_version): Likewise.
17191 * gcov-iov.c: Moved to...
17192 * genversion.c: ...here.
17193 * lto-streamer.h (LTO_major_version): Define it with
17195 * version.c: Removed.
17196 * version.h: Removed.
17198 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17200 * config/arc/arc.md (UNSPEC_ARC_DMPYWH): Define.
17201 * config/arc/simdext.md (VCT): Add predicates for iterator
17204 (voptab): Likewise.
17205 (vec_widen_<V_US>mult_hi_v4hi): Change pattern predicate.
17206 (<voptab>v2si3): New patterns.
17208 (reduc_plus_scal_v4hi): Likewise.
17209 (reduc_plus_scal_v2si): Likewise.
17210 (vec_duplicatev2si): Likewise.
17211 (vec_duplicatev4hi): Likewise.
17213 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17215 * config/arc/simdext.md: Format and cleanup file.
17217 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17219 * config/arc/simdext.md (movmisalignv2hi): Allow misaligned access
17220 only when munaligned-access option is on.
17221 (movmisalign<mode>): Likewise.
17223 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17225 * common/config/arc/arc-common.c (arc_handle_option): Remove dot
17227 * config/arc/arc.c (arc_reorg): Remove underscore from string.
17229 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17231 * config/arc/arc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
17232 (CTZ_DEFINED_VALUE_AT_ZERO): Likewise.
17233 * config/arc/arc.md (clrsbsi2): Cleanup pattern.
17234 (norm_f): Likewise.
17237 (clzsi2): Use fls instruction when available.
17238 (arc_clzsi2): Likewise.
17240 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17242 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add r26 and r27.
17244 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
17246 * doc/extend.texi (__builtin_arc_sr): Swap arguments.
17248 2021-05-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
17250 PR middle-end/100467
17251 * toplev.c (compile_file): Call insn_locations_init before
17252 targetm.asm_out.code_end.
17254 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
17257 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
17259 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
17261 2021-05-07 Jakub Jelinek <jakub@redhat.com>
17262 Andrew Stubbs <amd@codesourcery.com>
17265 * builtins.c (try_store_by_multiple_pieces): Use force_operand for
17266 emit_move_insn operands.
17268 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
17270 * cfgexpand.c (expand_gimple_basic_block): Do not inherit a current
17271 location for the outgoing edges of an empty block.
17272 * dwarf2out.c (add_subscript_info): Retrieve the bounds and index
17273 type by means of the get_array_descr_info langhook, if it is set and
17274 returns true. Remove obsolete code dealing with unnamed subtypes.
17276 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
17278 * gimple-range-cache.cc (ssa_block_ranges): Virtualize.
17279 (sbr_vector): Renamed from ssa_block_cache.
17280 (sbr_vector::sbr_vector): Allocate from obstack abd initialize.
17281 (ssa_block_ranges::~ssa_block_ranges): Remove.
17282 (sbr_vector::set_bb_range): Use varying and undefined cached values.
17283 (ssa_block_ranges::set_bb_varying): Remove.
17284 (sbr_vector::get_bb_range): Adjust assert.
17285 (sbr_vector::bb_range_p): Adjust assert.
17286 (~block_range_cache): No freeing loop required.
17287 (block_range_cache::get_block_ranges): Remove.
17288 (block_range_cache::set_bb_range): Inline get_block_ranges.
17289 (block_range_cache::set_bb_varying): Remove.
17290 * gimple-range-cache.h (set_bb_varying): Remove prototype.
17291 * value-range.h (irange_allocator::get_memory): New.
17293 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
17295 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Search
17296 dominator tree is available and requested.
17297 (ranger_cache::ssa_range_in_bb): Don't search dom tree here.
17298 (ranger_cache::fill_block_cache): Don't search dom tree here either.
17299 * gimple-range-cache.h (non_null_deref_p): Add dom_search param.
17301 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
17303 * gimple-range.cc (gimple_ranger::range_on_exit): Handle block with
17304 only PHI nodes better.
17306 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
17308 * gimple-range-edge.h (gimple_outgoing_range): Rename from
17310 (gcond_edge_range): Export prototype.
17311 * gimple-range-edge.cc (gcond_edge_range): New.
17312 (gimple_outgoing_range::edge_range_p): Use gcond_edge_range.
17313 * gimple-range-gori.h (gori_compute): Use gimple_outgoing_range.
17315 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
17317 * gimple-range-edge.cc (outgoing_range::calc_switch_ranges): Compute
17318 default range into a temp and allocate only what is needed.
17320 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
17322 * range-op.cc (operator_trunc_mod::wi_fold): x % 0 is UNDEFINED.
17324 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
17326 * gimple-range.h (gimple_range_global): Pick up parameter initial
17327 values, and use-before defined locals are UNDEFINED.
17329 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
17331 * doc/extend.texi (scalar_storage_order): Mention effect on pointer
17333 * tree.h (reverse_storage_order_for_component_p): Return false if
17334 the type is a pointer.
17336 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
17338 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
17340 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
17343 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
17344 Handle V8QI, V4HI and V2SI modes.
17345 * config/i386/mmx.md (mmx_pblendvb): New insn pattern.
17346 * config/i386/sse.md (unspec): Move UNSPEC_BLENDV ...
17347 * config/i386/i386.md (unspec): ... here.
17349 2021-05-07 Tobias Burnus <tobias@codesourcery.com>
17350 Tom de Vries <tdevries@suse.de>
17352 * omp-low.c (lower_rec_simd_input_clauses): Set max_vf = 1 if
17353 a truth_value_p reduction variable is nonintegral.
17355 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
17358 * config/i386/i386-expand.c (ix86_use_mask_cmp_p):
17359 Return false for mode sizes < 16.
17361 2021-05-07 Jakub Jelinek <jakub@redhat.com>
17364 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn.
17366 2021-05-06 Martin Jambor <mjambor@suse.cz>
17368 * ipa-sra.c (ipa_sra_dump_all_summaries): Dump edge summaries even
17369 when there is no function summary.
17370 (ipa_sra_summarize_function): produce edge summaries even when
17373 2021-05-06 Tom Tromey <tom@tromey.com>
17375 * godump.c (string_hash_eq): Remove.
17376 (go_finish): Use htab_eq_string.
17378 2021-05-06 Tom Tromey <tom@tromey.com>
17380 * gengtype-state.c (read_state): Use htab_eq_string.
17381 (string_eq): Remove.
17383 2021-05-06 Tom Tromey <tom@tromey.com>
17385 * gensupport.c (htab_eq_string): Remove.
17387 2021-05-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
17390 * debug.h (gcc_debug_hooks): Add set_ignored_loc function pointer.
17391 * dwarf2out.h (dw_fde_node::ignored_debug): New data item.
17392 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Add dummy
17393 set_ignored_loc callbacks.
17394 * debug.c (do_nothing_debug_hooks): Likewise.
17395 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
17396 * dwarf2out.c (text_section_used, cold_text_section_used): Remove.
17397 (in_text_section_p, last_text_label, last_cold_label,
17398 switch_text_ranges, switch_cold_ranges): New data items.
17399 (dwarf2out_note_section_used): Remove.
17400 (dwarf2out_begin_prologue): Set fde->ignored_debug and
17402 (mark_ignored_debug_section): New helper function.
17403 (dwarf2out_end_epilogue, dwarf2out_switch_text_section): Call
17404 mark_ignored_debug_section.
17405 (dwarf2_debug_hooks): Use dwarf2out_set_ignored_loc.
17406 (dwarf2_lineno_debug_hooks): Use dummy for set_ignored_loc.
17407 (size_of_aranges): Adjust formula for multi-part text ranges size.
17408 (output_aranges): Output multi-part text ranges.
17409 (dwarf2out_set_ignored_loc): New callback function.
17410 (dwarf2out_finish): Output multi-part text ranges.
17411 (dwarf2out_c_finalize): Clear new data items.
17412 * final.c (final_start_function_1): Call set_ignored_loc callback.
17413 (final_scan_insn_1): Likewise.
17414 * ggc-page.c (gt_ggc_mx): New helper function.
17415 * stringpool.c (gt_pch_nx): Likewise.
17417 2021-05-06 Richard Biener <rguenther@suse.de>
17419 * timevar.def (TV_TREE_INSERT_PHI_NODES): Remove.
17420 (TV_TREE_SSA_REWRITE_BLOCKS): Likewise.
17421 (TV_TREE_INTO_SSA): New.
17422 * tree-into-ssa.c (insert_phi_nodes): Do not account separately.
17423 (rewrite_blocks): Likewise.
17424 (pass_data_build_ssa): Account to TV_TREE_INTO_SSA.
17426 2021-05-06 Jakub Jelinek <jakub@redhat.com>
17428 * tree-ssa-phiopt.c (value_replacement, minmax_replacement,
17429 abs_replacement, xor_replacement,
17430 cond_removal_in_popcount_clz_ctz_pattern,
17431 replace_phi_edge_with_variable): Change type of phi argument from
17432 gimple * to gphi *.
17434 2021-05-06 Richard Biener <rguenther@suse.de>
17436 * tree-ssa-loop-split.c (split_loop): Delay updating SSA form.
17437 Output an opt-info message.
17438 (do_split_loop_on_cond): Likewise.
17439 (tree_ssa_split_loops): Update SSA form here.
17441 2021-05-06 Richard Biener <rguenther@suse.de>
17443 * tree-inline.c (tree_function_versioning): Fix DECL_BY_REFERENCE
17444 return variable removal.
17446 2021-05-06 Marius Hillenbrand <mhillen@linux.ibm.com>
17448 * config/s390/s390-builtins.def (O_M5, O1_M5, ...): Remove unused macros.
17449 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
17450 (s390_vec_permi_dbl, s390_vpdi): Use the O3_U2 type for the immediate
17452 * config/s390/s390.c (s390_const_operand_ok): Remove unused
17455 2021-05-06 Jakub Jelinek <jakub@redhat.com>
17457 PR tree-optimization/94589
17458 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Call
17459 spaceship_replacement.
17460 (cond_only_block_p, spaceship_replacement): New functions.
17462 2021-05-06 Richard Biener <rguenther@suse.de>
17465 * tree-emutls.c (gen_emutls_addr): Pass in whether we're
17466 dealing with a debug use and only query existing addresses
17468 (lower_emutls_1): Avoid splitting out addresses for debug
17469 stmts, reset the debug stmt when we fail to find existing
17471 (lower_emutls_phi_arg): Set wi.stmt.
17473 2021-05-06 Christoph Muellner <cmuellner@gcc.gnu.org>
17476 * config/riscv/riscv.c (riscv_block_move_loop): Use cbranch helper.
17477 * config/riscv/riscv.md (cbranch<mode>4): Generate helpers.
17478 (stack_protect_test): Use cbranch helper.
17480 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
17483 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
17484 always return the establisher frame for __builtin_frame_address (0).
17486 2021-05-05 Ivan Sorokin <vanyacpp@gmail.com>
17489 * config/i386/i386-builtins.c (ix86_cpu_model_type_node): New.
17490 (ix86_cpu_model_var): Likewise.
17491 (ix86_cpu_features2_type_node): Likewise.
17492 (ix86_cpu_features2_var): Likewise.
17493 (fold_builtin_cpu): Cache __cpu_model and __cpu_features2 with
17496 2021-05-05 Martin Sebor <msebor@redhat.com>
17498 * passes.def (pass_warn_printf): Run after SSA.
17500 2021-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17502 * config/arm/neon.md (neon_vtst_combine<mode>): New pattern.
17503 * config/arm/predicates.md (minus_one_operand): New predicate.
17505 2021-05-05 Jeff Law <jlaw@tachyum.com>
17507 * config/avr/avr.md: Remove references to CC_STATUS_INIT.
17509 2021-05-05 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
17511 PR rtl-optimization/100263
17512 * postreload.c (move2add_valid_value_p): Ensure register can
17515 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
17517 PR rtl-optimization/100411
17518 * cfgcleanup.c (try_crossjump_to_edge): Also skip end of prologue
17519 and beginning of function markers.
17521 2021-05-05 Jeff Law <jlaw@tachyum.com>
17523 * config/cr16/cr16.h (NOTICE_UPDATE_CC): Remove.
17524 * config/cr16/cr16.c (notice_update_cc): Remove.
17525 * config/cr16/cr16-protos.h (notice_update_cc): Remove.
17527 2021-05-05 Uroš Bizjak <ubizjak@gmail.com>
17530 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
17531 Handle V8QI, V4HI and V2SI modes.
17532 * config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
17533 (ix86_build_signbit_mask): Ditto.
17534 * config/i386/mmx.md (MMXMODE14): New mode iterator.
17535 (<smaxmin:code><MMXMODE14:mode>3): New expander.
17536 (*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
17537 (<umaxmin:code><MMXMODE24:mode>3): New expander.
17538 (*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
17539 (vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
17540 (vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
17541 (vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
17542 (vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
17543 (vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
17545 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
17547 * dwarf2out.c (loc_list_from_tree_1) <DECL>: During early DWARF, do
17548 not expand the VALUE_EXPR of variables put in the non-local frame.
17549 * gimplify.c (gimplify_type_sizes) <RECORD_TYPE>: If the type is not
17550 to be ignored for debug info, ensure its variable offsets are not.
17552 2021-05-05 Richard Biener <rguenther@suse.de>
17554 PR tree-optimization/79333
17555 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
17556 Fold stmt following SSA edges.
17558 2021-05-05 Richard Biener <rguenther@suse.de>
17560 PR middle-end/100394
17561 * calls.c (expand_call): Preserve possibly throwing calls.
17562 * cfgexpand.c (expand_call_stmt): When a call can throw signal
17563 RTL expansion there are side-effects.
17564 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Simplify,
17565 mark all possibly throwing stmts necessary unless we can elide
17567 * tree-ssa-dse.c (pass_dse::execute): Preserve exceptions unless
17568 -fdelete-dead-exceptions.
17569 * tree.h (DECL_PURE_P): Add note about exceptions.
17571 2021-05-05 Alexandre Oliva <oliva@adacore.com>
17573 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Make it
17576 2021-05-04 David Edelsohn <dje.gcc@gmail.com>
17578 * config/rs6000/rs6000-call.c (rs6000_output_mi_thunk): Use
17579 get_fnname_from_decl for name of thunk.
17580 * config/rs6000/rs6000.c (rs6000_declare_alias): Use assemble_name
17581 and ASM_OUTPUT_LABEL.
17582 (rs6000_xcoff_declare_function_name): Use assemble_name and
17584 (rs6000_xcoff_declare_object_name): Use ASM_OUTPUT_LABEL.
17585 (rs6000_xcoff_encode_section_info): Don't add mapping class
17586 for aliases. Always add [DS] mapping class to primary
17588 (rs6000_asm_weaken_decl): Don't explicitly add [DS].
17590 2021-05-04 Martin Sebor <msebor@redhat.com>
17592 PR middle-end/100307
17593 * builtins.c (compute_objsize_r): Clear base0 for pointers.
17595 2021-05-04 Jeff Law <jlaw@tachyum.com>
17597 * config/bfin/bfin.h (NOTICE_UPDATE_CC): Remove.
17599 2021-05-04 Segher Boessenkool <segher@kernel.crashing.org>
17601 * caller-save.c: Remove CC0.
17602 * cfgcleanup.c: Remove CC0.
17603 * cfgrtl.c: Remove CC0.
17604 * combine.c: Remove CC0.
17605 * compare-elim.c: Remove CC0.
17606 * conditions.h: Remove CC0.
17607 * config/h8300/h8300.h: Remove CC0.
17608 * config/h8300/h8300-protos.h: Remove CC0.
17609 * config/h8300/peepholes.md: Remove CC0.
17610 * config/i386/x86-tune-sched.c: Remove CC0.
17611 * config/m68k/m68k.c: Remove CC0.
17612 * config/rl78/rl78.c: Remove CC0.
17613 * config/sparc/sparc.c: Remove CC0.
17614 * config/xtensa/xtensa.c: Remove CC0.
17615 (gen_conditional_move): Use pc_rtx instead of cc0_rtx in a piece of
17616 RTL where that is used as a placeholder only.
17617 * cprop.c: Remove CC0.
17618 * cse.c: Remove CC0.
17619 * cselib.c: Remove CC0.
17620 * df-problems.c: Remove CC0.
17621 * df-scan.c: Remove CC0.
17622 * doc/md.texi: Remove CC0. Adjust an example.
17623 * doc/rtl.texi: Remove CC0. Adjust an example.
17624 * doc/tm.texi: Regenerate.
17625 * doc/tm.texi.in: Remove CC0.
17626 * emit-rtl.c: Remove CC0.
17627 * final.c: Remove CC0.
17628 * fwprop.c: Remove CC0.
17629 * gcse-common.c: Remove CC0.
17630 * gcse.c: Remove CC0.
17631 * genattrtab.c: Remove CC0.
17632 * genconfig.c: Remove CC0.
17633 * genemit.c: Remove CC0.
17634 * genextract.c: Remove CC0.
17635 * gengenrtl.c: Remove CC0.
17636 * genrecog.c: Remove CC0.
17637 * haifa-sched.c: Remove CC0.
17638 * ifcvt.c: Remove CC0.
17639 * ira-costs.c: Remove CC0.
17640 * ira.c: Remove CC0.
17641 * jump.c: Remove CC0.
17642 * loop-invariant.c: Remove CC0.
17643 * lra-constraints.c: Remove CC0.
17644 * lra-eliminations.c: Remove CC0.
17645 * optabs.c: Remove CC0.
17646 * postreload-gcse.c: Remove CC0.
17647 * postreload.c: Remove CC0.
17648 * print-rtl.c: Remove CC0.
17649 * read-rtl-function.c: Remove CC0.
17650 * reg-notes.def: Remove CC0.
17651 * reg-stack.c: Remove CC0.
17652 * reginfo.c: Remove CC0.
17653 * regrename.c: Remove CC0.
17654 * reload.c: Remove CC0.
17655 * reload1.c: Remove CC0.
17656 * reorg.c: Remove CC0.
17657 * resource.c: Remove CC0.
17658 * rtl.c: Remove CC0.
17659 * rtl.def: Remove CC0.
17660 * rtl.h: Remove CC0.
17661 * rtlanal.c: Remove CC0.
17662 * sched-deps.c: Remove CC0.
17663 * sched-rgn.c: Remove CC0.
17664 * shrink-wrap.c: Remove CC0.
17665 * simplify-rtx.c: Remove CC0.
17666 * system.h: Remove CC0. Poison NOTICE_UPDATE_CC, CC_STATUS_MDEP_INIT,
17667 CC_STATUS_MDEP, and CC_STATUS.
17668 * target.def: Remove CC0.
17669 * valtrack.c: Remove CC0.
17670 * var-tracking.c: Remove CC0.
17672 2021-05-04 Richard Biener <rguenther@suse.de>
17674 PR tree-optimization/100414
17675 * tree-ssa-phiopt.c (get_non_trapping): Do not compute dominance
17677 (tree_ssa_phiopt_worker): But unconditionally here.
17679 2021-05-04 Tobias Burnus <tobias@codesourcery.com>
17681 * omp-low.c (lower_rec_input_clauses, lower_reduction_clauses): Handle
17682 && and || with floating-point and complex arguments.
17684 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
17686 * tree-inline.c (insert_debug_decl_map): Delete.
17687 (copy_debug_stmt): Minor tweak.
17688 (setup_one_parameter): Do not use a variable if the value is either
17689 a read-only DECL or a non-addressable local variable in the caller.
17690 In this case, insert the debug-only variable in the map manually.
17691 (expand_call_inline): Do not generate a CLOBBER for these values.
17692 * tree-inline.h (debug_map): Minor tweak.
17694 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
17696 * builtins.c (builtin_with_linkage_p): Return true for stp[n]cpy.
17697 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Tidy up.
17699 2021-05-04 Richard Biener <rguenther@suse.de>
17701 PR tree-optimization/100329
17702 * tree-ssa-reassoc.c (can_reassociate_p): Do not reassociate
17704 (insert_stmt_after): Assert we're not running into asm goto.
17706 2021-05-04 Richard Biener <rguenther@suse.de>
17708 PR tree-optimization/100398
17709 * tree-ssa-dse.c (pass_dse::execute): Preserve control
17712 2021-05-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17714 * builtins.c (try_store_by_multiple_pieces): Fix constfun's prototype.
17716 2021-05-04 Alexandre Oliva <oliva@adacore.com>
17718 * builtins.c (try_store_by_multiple_pieces): New.
17719 (expand_builtin_memset_args): Use it. If target_char_cast
17720 fails, proceed as for non-constant val. Pass len's ctz to...
17721 * expr.c (clear_storage_hints): ... this. Try store by
17722 multiple pieces after setmem.
17723 (clear_storage): Adjust.
17724 * expr.h (clear_storage_hints): Likewise.
17725 (try_store_by_multiple_pieces): Declare.
17726 * passes.def: Replace the last copy_prop with ccp.
17728 2021-05-03 Tom de Vries <tdevries@suse.de>
17731 * omp-low.c (lower_rec_input_clauses): Disable SIMT for user-defined
17734 2021-05-03 Richard Biener <rguenther@suse.de>
17736 * tree-ssa-dse.c (dse_classify_store): Track two PHI defs.
17738 2021-05-03 Richard Biener <rguenther@suse.de>
17740 * tree-ssa-dse.c: Do not include domwalk.h but cfganal.h.
17741 (dse_dom_walker): Remove.
17742 (dse_dom_walker::dse_optimize_stmt): Rename...
17743 (dse_optimize_stmt): ... to this, pass in live_bytes sbitmap.
17744 (dse_dom_walker::before_dom_children): Inline ...
17745 (pass_dse::execute): ... here. Perform a reverse program
17748 2021-05-03 H.J. Lu <hjl.tools@gmail.com>
17751 * configure: Regenerated.
17753 2021-05-03 Ilya Leoshkevich <iii@linux.ibm.com>
17756 * config/s390/s390.c (s390_hard_fp_reg_p): New function.
17757 (s390_md_asm_adjust): Handle hard registers.
17759 2021-05-03 Jakub Jelinek <jakub@redhat.com>
17761 PR tree-optimization/100382
17762 * tree-ssa-dse.c: Include tree-eh.h.
17763 (dse_dom_walker::before_dom_children): Don't remove stmts if
17764 stmt_unremovable_because_of_non_call_eh_p is true.
17766 2021-05-02 David Edelsohn <dje.gcc@gmail.com>
17768 * varasm.c (compute_reloc_for_var): Split out from...
17769 (get_variable_section): Use it.
17770 * output.h (compute_reloc_for_var): Declare.
17771 * config/rs6000/rs6000-protos.h
17772 (rs6000_xcoff_asm_output_aligned_decl_common): Change alignment to
17774 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address_aix):
17775 Don't append storage mapping class to symbol.
17776 (rs6000_xcoff_asm_named_section): Add BS and UL mapping classes.
17777 Don't convert TLS BSS to common.
17778 (rs6000_xcoff_unique_section): Don't fall back to select_secton.
17779 (rs6000_xcoff_section_type_flags): Add SECTION_BSS if DECL is
17781 (rs6000_xcoff_asm_globalize_decl_name): Don't strip storage
17783 (rs6000_xcoff_asm_output_aligned_decl_common): Align is unsigned int.
17784 If align is 0 from TLS class, use the same rules as varasm.c
17785 If not common, switch to BSS section manually.
17786 If common, emit appropriate comm or lcomm directive.
17787 (rs6000_xcoff_encode_section_info): Add logic to append all
17788 storage mapping classes.
17789 (rs6000_asm_weaken_decl): Adjust for qualname symbols.
17790 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
17791 rs6000_xcoff_asm_output_aligned_decl_common.
17792 (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
17793 rs6000_xcoff_asm_output_aligned_decl_common.
17794 (ASM_OUTPUT_TLS_COMMON): Use
17795 rs6000_xcoff_asm_output_aligned_decl_common.
17797 2021-05-02 Jakub Jelinek <jakub@redhat.com>
17800 * config/nvptx/nvptx.c (nvptx_sese_pseudo): Use nullptr instead of 0
17801 as first argument of pseudo_node_t constructors.
17803 2021-05-02 Jakub Jelinek <jakub@redhat.com>
17806 * config/i386/t-i386 (TM_H): Add $(srcdir)/config/i386/i386-isa.def.
17808 2021-05-01 Aldy Hernandez <aldyh@redhat.com>
17810 * value-range.cc (DEFINE_INT_RANGE_GC_STUBS): Remove.
17811 (gt_pch_nx (int_range<1> *&)): New.
17812 (gt_ggc_mx (int_range<1> *&)): New.
17813 * value-range.h (class irange): Add GTY support for
17816 2021-05-01 Geng Qi <gengqi@linux.alibaba.com>
17818 * doc/options.texi (Negative): Change either or to both and.
17820 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17822 * config/aarch64/aarch64-simd-builtins.def: Add
17823 float_ml[as][q]_laneq builtin generator macros.
17824 * config/aarch64/aarch64-simd.md (mul_laneq<mode>3): Define.
17825 (aarch64_float_mla_laneq<mode>): Define.
17826 (aarch64_float_mls_laneq<mode>): Define.
17827 * config/aarch64/arm_neon.h (vmla_laneq_f32): Use RTL builtin
17828 instead of GCC vector extensions.
17829 (vmlaq_laneq_f32): Likewise.
17830 (vmls_laneq_f32): Likewise.
17831 (vmlsq_laneq_f32): Likewise.
17833 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17835 * config/aarch64/aarch64-simd-builtins.def: Add
17836 float_ml[as]_lane builtin generator macros.
17837 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt<mode>):
17839 (mul_lane<mode>3): This, and re-order arguments.
17840 (aarch64_float_mla_lane<mode>): Define.
17841 (aarch64_float_mls_lane<mode>): Define.
17842 * config/aarch64/arm_neon.h (vmla_lane_f32): Use RTL builtin
17843 instead of GCC vector extensions.
17844 (vmlaq_lane_f32): Likewise.
17845 (vmls_lane_f32): Likewise.
17846 (vmlsq_lane_f32): Likewise.
17848 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17850 * config/aarch64/aarch64-simd-builtins.def: Add float_ml[as]
17851 builtin generator macros.
17852 * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>):
17854 (aarch64_float_mls<mode>): Define.
17855 * config/aarch64/arm_neon.h (vmla_f32): Use RTL builtin
17856 instead of relying on GCC vector extensions.
17857 (vmla_f64): Likewise.
17858 (vmlaq_f32): Likewise.
17859 (vmlaq_f64): Likewise.
17860 (vmls_f32): Likewise.
17861 (vmls_f64): Likewise.
17862 (vmlsq_f32): Likewise.
17863 (vmlsq_f64): Likewise.
17864 * config/aarch64/iterators.md: Define VDQF_DF mode iterator.
17866 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17868 * config/aarch64/aarch64-simd-builtins.def: Add
17869 float_ml[as]_n_builtin generator macros.
17870 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt_from_dup<mode>):
17872 (mul_n<mode>3): This, and re-order arguments.
17873 (aarch64_float_mla_n<mode>): Define.
17874 (aarch64_float_mls_n<mode>): Define.
17875 * config/aarch64/arm_neon.h (vmla_n_f32): Use RTL builtin
17876 instead of inline asm.
17877 (vmlaq_n_f32): Likewise.
17878 (vmls_n_f32): Likewise.
17879 (vmlsq_n_f32): Likewise.
17881 2021-04-30 Jonathan Wright <joanthan.wright@arm.com>
17883 * config/aarch64/aarch64-simd-builtins.def: Add pmull[2]
17884 builtin generator macros.
17885 * config/aarch64/aarch64-simd.md (aarch64_pmullv8qi): Define.
17886 (aarch64_pmull_hiv16qi_insn): Define.
17887 (aarch64_pmull_hiv16qi): Define.
17888 * config/aarch64/arm_neon.h (vmull_high_p8): Use RTL builtin
17889 instead of inline asm.
17890 (vmull_p8): Likewise.
17892 2021-04-30 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
17894 * config/avr/avr.md: Adjust peepholes to match and
17895 generate parallels with clobber of REG_CC.
17896 (mov<mode>_insn): Rename to mov<mode>_insn_split.
17897 (*mov<mode>_insn): Rename to mov<mode>_insn.
17899 2021-04-30 David Edelsohn <dje.gcc@gmail.com>
17901 * varasm.c (use_blocks_for_decl_p): Don't use section anchors
17902 for VAR_DECLs if -fdata-sections enabled.
17904 2021-04-30 Michael Meissner <meissner@linux.ibm.com>
17906 PR bootstrap/100327
17907 * config/rs6000/rs6000.c
17908 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Define.
17909 (rs6000_libgcc_floating_mode_supported_p): New target hook.
17911 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
17913 * tree-ssa-threadbackward.c (class thread_jumps): Split out code
17915 (class back_threader_registry): ...to here...
17916 (class back_threader_profitability): ...and here...
17917 (thread_jumps::thread_through_all_blocks): Remove argument.
17918 (back_threader_registry::back_threader_registry): New.
17919 (back_threader_registry::~back_threader_registry): New.
17920 (back_threader_registry::thread_through_all_blocks): New.
17921 (thread_jumps::profitable_jump_thread_path): Move from here...
17922 (back_threader_profitability::profitable_path_p): ...to here.
17923 (thread_jumps::find_taken_edge): New.
17924 (thread_jumps::convert_and_register_current_path): Move...
17925 (back_threader_registry::register_path): ...to here.
17926 (thread_jumps::register_jump_thread_path_if_profitable): Move...
17927 (thread_jumps::maybe_register_path): ...to here.
17928 (thread_jumps::handle_phi): Call find_taken_edge and
17929 maybe_register_path.
17930 (thread_jumps::handle_assignment): Same.
17931 (thread_jumps::fsm_find_control_statement_thread_paths): Remove
17932 tree argument to handle_phi and handle_assignment.
17933 (thread_jumps::find_jump_threads_backwards): Set m_name. Remove
17934 set of m_speed_p and m_max_threaded_paths.
17935 (pass_thread_jumps::execute): Remove second argument from
17936 find_jump_threads_backwards.
17937 (pass_early_thread_jumps::execute): Same.
17939 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
17941 * tree-ssa-dom.c (class dom_jump_threader_simplifier): New.
17942 (class dom_opt_dom_walker): Initialize some class variables.
17943 (pass_dominator::execute): Pass evrp_range_analyzer and
17944 dom_jump_threader_simplifier to dom_opt_dom_walker.
17945 Adjust for some functions moving into classes.
17946 (simplify_stmt_for_jump_threading): Adjust and move to...
17947 (jump_threader_simplifier::simplify): ...here.
17948 (dom_opt_dom_walker::before_dom_children): Adjust for
17949 m_evrp_range_analyzer.
17950 (dom_opt_dom_walker::after_dom_children): Remove x_vr_values hack.
17951 (test_for_singularity): Place in dom_opt_dom_walker class.
17952 (dom_opt_dom_walker::optimize_stmt): The argument
17953 evrp_range_analyzer is now a class field.
17954 * tree-ssa-threadbackward.c (class thread_jumps): Add m_registry.
17955 (thread_jumps::thread_through_all_blocks): New.
17956 (thread_jumps::convert_and_register_current_path): Use m_registry.
17957 (pass_thread_jumps::execute): Adjust for thread_through_all_blocks
17958 being in the threader class.
17959 (pass_early_thread_jumps::execute): Same.
17960 * tree-ssa-threadedge.c (threadedge_initialize_values): Move...
17961 (jump_threader::jump_threader): ...here.
17962 (threadedge_finalize_values): Move...
17963 (jump_threader::~jump_threader): ...here.
17964 (jump_threader::remove_jump_threads_including): New.
17965 (jump_threader::thread_through_all_blocks): New.
17966 (record_temporary_equivalences_from_phis): Move...
17967 (jump_threader::record_temporary_equivalences_from_phis): ...here.
17968 (record_temporary_equivalences_from_stmts_at_dest): Move...
17969 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
17971 (simplify_control_stmt_condition_1): Move to jump_threader class.
17972 (simplify_control_stmt_condition): Move...
17973 (jump_threader::simplify_control_stmt_condition): ...here.
17974 (thread_around_empty_blocks): Move...
17975 (jump_threader::thread_around_empty_blocks): ...here.
17976 (thread_through_normal_block): Move...
17977 (jump_threader::thread_through_normal_block): ...here.
17978 (thread_across_edge): Move...
17979 (jump_threader::thread_across_edge): ...here.
17980 (thread_outgoing_edges): Move...
17981 (jump_threader::thread_outgoing_edges): ...here.
17982 * tree-ssa-threadedge.h: Move externally facing functings...
17983 (class jump_threader): ...here...
17984 (class jump_threader_simplifier): ...and here.
17985 * tree-ssa-threadupdate.c (struct redirection_data): Remove comment.
17986 (jump_thread_path_allocator::jump_thread_path_allocator): New.
17987 (jump_thread_path_allocator::~jump_thread_path_allocator): New.
17988 (jump_thread_path_allocator::allocate_thread_edge): New.
17989 (jump_thread_path_allocator::allocate_thread_path): New.
17990 (jump_thread_path_registry::jump_thread_path_registry): New.
17991 (jump_thread_path_registry::~jump_thread_path_registry): New.
17992 (jump_thread_path_registry::allocate_thread_edge): New.
17993 (jump_thread_path_registry::allocate_thread_path): New.
17994 (dump_jump_thread_path): Make extern.
17995 (debug (const vec<jump_thread_edge *> &path)): New.
17996 (struct removed_edges): Move to tree-ssa-threadupdate.h.
17997 (struct thread_stats_d): Remove.
17998 (remove_ctrl_stmt_and_useless_edges): Make static.
17999 (lookup_redirection_data): Move...
18000 (jump_thread_path_registry::lookup_redirection_data): ...here.
18001 (ssa_redirect_edges): Make static.
18002 (thread_block_1): Move...
18003 (jump_thread_path_registry::thread_block_1): ...here.
18004 (thread_block): Move...
18005 (jump_thread_path_registry::thread_block): ...here.
18006 (thread_through_loop_header): Move...
18007 (jump_thread_path_registry::thread_through_loop_header): ...here.
18008 (mark_threaded_blocks): Move...
18009 (jump_thread_path_registry::mark_threaded_blocks): ...here.
18010 (debug_path): Move...
18011 (jump_thread_path_registry::debug_path): ...here.
18012 (debug_all_paths): Move...
18013 (jump_thread_path_registry::dump): ..here.
18014 (rewire_first_differing_edge): Move...
18015 (jump_thread_path_registry::rewire_first_differing_edge): ...here.
18016 (adjust_paths_after_duplication): Move...
18017 (jump_thread_path_registry::adjust_paths_after_duplication): ...here.
18018 (duplicate_thread_path): Move...
18019 (jump_thread_path_registry::duplicate_thread_path): ..here.
18020 (remove_jump_threads_including): Move...
18021 (jump_thread_path_registry::remove_jump_threads_including): ...here.
18022 (thread_through_all_blocks): Move to...
18023 (jump_thread_path_registry::thread_through_all_blocks): ...here.
18024 (delete_jump_thread_path): Remove.
18025 (register_jump_thread): Move...
18026 (jump_thread_path_registry::register_jump_thread): ...here.
18027 * tree-ssa-threadupdate.h: Move externally facing functions...
18028 (class jump_thread_path_allocator): ...here...
18029 (class jump_thread_path_registry): ...and here.
18030 (thread_through_all_blocks): Remove.
18031 (struct removed_edges): New.
18032 (register_jump_thread): Remove.
18033 (remove_jump_threads_including): Remove.
18034 (delete_jump_thread_path): Remove.
18035 (remove_ctrl_stmt_and_useless_edges): Remove.
18036 (free_dom_edge_info): New prototype.
18037 * tree-vrp.c: Remove x_vr_values hack.
18038 (class vrp_jump_threader_simplifier): New.
18039 (vrp_jump_threader_simplifier::simplify): New.
18040 (vrp_jump_threader::vrp_jump_threader): Adjust method signature.
18041 Remove m_dummy_cond.
18042 Instantiate m_simplifier and m_threader.
18043 (vrp_jump_threader::thread_through_all_blocks): New.
18044 (vrp_jump_threader::simplify_stmt): Remove.
18045 (vrp_jump_threader::after_dom_children): Do not set m_dummy_cond.
18046 Remove x_vr_values hack.
18047 (execute_vrp): Adjust for thread_through_all_blocks being in a
18050 2021-04-30 Christophe Lyon <christophe.lyon@linaro.org>
18052 * genflags.c (gen_insn): Print failed expansion string.
18054 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
18056 * expr.c (alignment_for_piecewise_move): Call mode_for_size
18057 without limit to MAX_FIXED_MODE_SIZE.
18059 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
18061 PR middle-end/90773
18062 * builtins.c (builtin_memset_gen_str): Don't use return from
18063 simplify_gen_subreg.
18065 2021-04-30 Uroš Bizjak <ubizjak@gmail.com>
18068 * config/i386/i386.md (*add<mode>3_carry_0r): New insn pattern.
18069 (*addsi3_carry_zext_0r): Ditto.
18070 (*sub<mode>3_carry_0): Ditto.
18071 (*subsi3_carry_zext_0r): Ditto.
18072 * config/i386/predicates.md (ix86_carry_flag_unset_operator):
18074 * config/i386/i386.c (ix86_rtx_costs) <case PLUS, case MINUS>:
18075 Also consider ix86_carry_flag_unset_operator to calculate
18076 the cost of adc/sbb insn.
18078 2021-04-30 Roman Zhuykov <zhroma@ispras.ru>
18080 PR rtl-optimization/100225
18081 PR rtl-optimization/84878
18082 * modulo-sched.c (sms_schedule): Use note_stores to skip loops
18083 where we have an instruction which touches (writes) any hard
18084 register from df->regular_block_artificial_uses set.
18085 Allow not-single-set instruction only right before basic block
18088 2021-04-30 Geng Qi <gengqi@linux.alibaba.com>
18090 * config/riscv/riscv.opt (march=,mabi=): Negative itself.
18092 2021-04-30 LevyHsu <admin@levyhsu.com>
18094 * config/riscv/riscv.c (riscv_min_arithmetic_precision): New.
18095 * config/riscv/riscv.h (TARGET_MIN_ARITHMETIC_PRECISION): New.
18096 * config/riscv/riscv.md (addv<mode>4, uaddv<mode>4): New.
18097 (subv<mode>4, usubv<mode>4, mulv<mode>4, umulv<mode>4): New.
18099 2021-04-29 Alexandre Oliva <oliva@adacore.com>
18101 * config.gcc: Merged x86 and x86_64 cpu_type-setting cases.
18103 2021-04-29 Alexandre Oliva <oliva@adacore.com>
18105 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Rename to...
18106 (ASM_OUTPUT_MAX_SKIP_ALIGN): ... this. Enclose in do/while(0).
18107 * config/i386/i386.c: Adjust.
18108 * config/i386/i386.md: Adjust.
18109 * config/i386/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Drop.
18110 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18111 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18112 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18113 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18114 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18115 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18116 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18117 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18118 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
18119 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
18121 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
18123 * config/i386/i386-expand.c (ix86_expand_int_compare):
18124 Swap operands of GTU and LEU comparison to emit carry flag comparison.
18125 * config/i386/i386.md (*add<mode>3_carry_0): Change insn
18126 predicate to allow more combine opportunities with memory operands.
18127 (*sub<mode>3_carry_0): Ditto.
18129 2021-04-29 Richard Sandiford <richard.sandiford@arm.com>
18131 PR rtl-optimization/100303
18132 * rtl-ssa/accesses.cc (function_info::make_use_available): Take a
18133 boolean that indicates whether the use will only be used in
18134 debug instructions. Treat it in the same way that existing
18135 cross-EBB debug references would be handled if so.
18136 (function_info::make_uses_available): Likewise.
18137 * rtl-ssa/functions.h (function_info::make_uses_available): Update
18138 prototype accordingly.
18139 (function_info::make_uses_available): Likewise.
18140 * fwprop.c (try_fwprop_subst): Update call accordingly.
18142 2021-04-29 Jeff Law <jlaw@tachyum.com>
18144 * config/nios2/nios2-protos.h (nios2_fpu_insn_enabled): Move outside
18147 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
18148 Richard Biener <rguenther@suse.de>
18151 * config/i386/i386-builtin.def (IX86_BUILTIN_MASKLOADPD)
18152 (IX86_BUILTIN_MASKLOADPS, IX86_BUILTIN_MASKLOADPD256)
18153 (IX86_BUILTIN_MASKLOADPS256, IX86_BUILTIN_MASKLOADD)
18154 (IX86_BUILTIN_MASKLOADQ, IX86_BUILTIN_MASKLOADD256)
18155 (IX86_BUILTIN_MASKLOADQ256): Move from SPECIAL_ARGS
18156 to PURE_ARGS category.
18157 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
18158 Handle PURE_ARGS category.
18159 * config/i386/i386-expand.c (ix86_expand_builtin): Ditto.
18161 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
18163 * configure.ac: Check for the presence of sys/locking.h header and
18164 for whether _LK_LOCK is supported by _locking.
18165 * configure: Regenerate.
18166 * config.in: Likewise.
18167 * gcov-io.h: Define GCOV_LOCKED_WITH_LOCKING if HOST_HAS_LK_LOCK.
18168 * gcov-io.c (gcov_open): Add support for GCOV_LOCKED_WITH_LOCKING.
18169 * system.h: Include <sys/locking.h> if HAVE_SYS_LOCKING_H.
18171 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
18173 * config/i386/predicates.md (fcmov_comparison_operator):
18174 Do not check for trivial FP comparison operator.
18175 <case GEU, case LTU>: Allow CCGZmode.
18176 <case GTU, case LEU>: Do not allow CCCmode.
18177 (ix86_comparison_operator) <case GTU, case LEU>: Allow only CCmode.
18178 (ix86_carry_flag_operator): Match only LTU and UNLT code.
18179 Do not check for trivial FP comparison operator. Allow CCGZmode.
18181 2021-04-29 Tom de Vries <tdevries@suse.de>
18183 * omp-expand.c (expand_omp_simd): Add step_orig, and replace uses of
18184 fd->loop.step by either step or orig_step.
18186 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
18188 * config/sparc/sparc.c (gen_load_pcrel_sym): Delete.
18189 (load_got_register): Do the PIC dance here.
18190 (sparc_legitimize_tls_address): Simplify.
18191 (sparc_emit_probe_stack_range): Likewise.
18192 (sparc32_initialize_trampoline): Likewise.
18193 (sparc64_initialize_trampoline): Likewise.
18194 * config/sparc/sparc.md (load_pcrel_sym<P:mode>): Add @ marker.
18195 (probe_stack_range<P:mode>): Likewise.
18196 (flush<P:mode>): Likewise.
18197 (tgd_hi22<P:mode>): Likewise.
18198 (tgd_lo10<P:mode>): Likewise.
18199 (tgd_add<P:mode>): Likewise.
18200 (tgd_call<P:mode>): Likewise.
18201 (tldm_hi22<P:mode>): Likewise.
18202 (tldm_lo10<P:mode>): Likewise.
18203 (tldm_add<P:mode>): Likewise.
18204 (tldm_call<P:mode>): Likewise.
18205 (tldo_hix22<P:mode>): Likewise.
18206 (tldo_lox10<P:mode>): Likewise.
18207 (tldo_add<P:mode>): Likewise.
18208 (tie_hi22<P:mode>): Likewise.
18209 (tie_lo10<P:mode>): Likewise.
18210 (tie_add<P:mode>): Likewise.
18211 (tle_hix22<P:mode>): Likewise.
18212 (tle_lox10<P:mode>): Likewise.
18213 (stack_protect_setsi): Rename to...
18214 (stack_protect_set32): ...this.
18215 (stack_protect_setdi): Rename to...
18216 (stack_protect_set64): ...this.
18217 (stack_protect_set): Adjust calls to above.
18218 (stack_protect_testsi): Rename to...
18219 (stack_protect_test32): ...this.
18220 (stack_protect_testdi): Rename to...
18221 (stack_protect_test64): ...this.
18222 (stack_protect_test): Adjust calls to above.
18224 2021-04-29 H.J. Lu <hjl.tools@gmail.com>
18226 PR middle-end/90773
18227 * builtins.c (builtin_memcpy_read_str): Add a dummy argument.
18228 (builtin_strncpy_read_str): Likewise.
18229 (builtin_memset_read_str): Add an argument for the previous RTL
18230 information and generate the new RTL from the previous RTL info.
18231 (builtin_memset_gen_str): Likewise.
18232 * builtins.h (builtin_strncpy_read_str): Update the prototype.
18233 (builtin_memset_read_str): Likewise.
18234 * expr.c (by_pieces_ninsns): If targetm.overlap_op_by_pieces_p()
18235 returns true, round up size and alignment to the widest integer
18236 mode for maximum size.
18237 (pieces_addr::adjust): Add a pointer to by_pieces_prev argument
18238 and pass it to m_constfn.
18239 (op_by_pieces_d): Add m_push and m_overlap_op_by_pieces.
18240 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
18241 initialize m_push. Initialize m_overlap_op_by_pieces with
18242 targetm.overlap_op_by_pieces_p ().
18243 (op_by_pieces_d::run): Pass the previous RTL information to
18244 pieces_addr::adjust and generate overlapping operations if
18245 m_overlap_op_by_pieces is true.
18247 (move_by_pieces_d::move_by_pieces_d): Updated for op_by_pieces_d
18249 (store_by_pieces_d::store_by_pieces_d): Updated for op_by_pieces_d
18251 (can_store_by_pieces): Use by_pieces_constfn on constfun.
18252 (store_by_pieces): Use by_pieces_constfn on constfun. Updated
18253 for op_by_pieces_d change.
18254 (clear_by_pieces_1): Add a dummy argument.
18255 (clear_by_pieces): Updated for op_by_pieces_d change.
18256 (compare_by_pieces_d::compare_by_pieces_d): Likewise.
18257 (string_cst_read_str): Add a dummy argument.
18258 * expr.h (by_pieces_constfn): Add a dummy argument.
18259 (by_pieces_prev): New.
18260 * target.def (overlap_op_by_pieces_p): New target hook.
18261 * config/i386/i386.c (TARGET_OVERLAP_OP_BY_PIECES_P): New.
18262 * doc/tm.texi.in: Add TARGET_OVERLAP_OP_BY_PIECES_P.
18263 * doc/tm.texi: Regenerated.
18265 2021-04-29 Richard Biener <rguenther@suse.de>
18267 PR tree-optimization/100253
18268 * tree-vect-stmts.c (vectorizable_load): Do not assume
18269 element alignment when DR_MISALIGNMENT is -1.
18270 (vectorizable_store): Likewise.
18272 2021-04-29 Jakub Jelinek <jakub@redhat.com>
18275 * config/aarch64/aarch64.c (aarch64_add_offset_1_temporaries): Use
18276 absu_hwi instead of abs_hwi.
18278 2021-04-29 Richard Biener <rguenther@suse.de>
18280 PR middle-end/38474
18281 * tree-ssa-structalias.c (add_graph_edge): Avoid direct
18282 forwarding when indirect forwarding through ESCAPED
18285 2021-04-29 Tom de Vries <tdevries@suse.de>
18288 * internal-fn.c (expand_GOMP_SIMT_ENTER_ALLOC)
18289 (expand_GOMP_SIMT_LAST_LANE, expand_GOMP_SIMT_ORDERED_PRED)
18290 (expand_GOMP_SIMT_VOTE_ANY, expand_GOMP_SIMT_XCHG_BFLY)
18291 (expand_GOMP_SIMT_XCHG_IDX): Ensure target is assigned to.
18293 2021-04-29 Richard Biener <rguenther@suse.de>
18295 PR tree-optimization/99912
18296 * tree-ssa-dse.c (dse_dom_walker::m_need_cfg_cleanup): New.
18297 (dse_dom_walker::todo): Likewise.
18298 (dse_dom_walker::dse_optimize_stmt): Move VDEF check to the
18300 (dse_dom_walker::before_dom_children): Remove trivially
18301 dead SSA defs and schedule CFG cleanup if we removed all
18303 (pass_dse::execute): Get TODO as computed by the DOM walker
18304 and return it. Wipe dominator info earlier.
18306 2021-04-29 Richard Biener <rguenther@suse.de>
18309 * ipa-prop.c (ipcp_modif_dom_walker::before_dom_children):
18310 Track blocks to cleanup EH in new m_need_eh_cleanup.
18311 (ipcp_modif_dom_walker::cleanup_eh): New.
18312 (ipcp_transform_function): Release dominator info before
18315 2021-04-29 Martin Sebor <msebor@redhat.com>
18317 PR middle-end/100250
18318 * attribs.c (attr_access::array_as_string): Avoid dereferencing
18319 a pointer when it's null.
18321 2021-04-29 Martin Sebor <msebor@redhat.com>
18323 * Makefile.in (OBJS): Add ipa-free-lang-data.o.
18324 * ipa-free-lang-data.cc: New file.
18325 * tree.c: Move pass free_lang_data to file above.
18326 (build_array_type_1): Declare extern.
18327 * tree.h (build_array_type_1): Declare.
18329 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18331 * config/aarch64/aarch64-simd-builtins.def: Modify comment to
18332 make consistent with updated RTL pattern.
18333 * config/aarch64/aarch64-simd.md (aarch64_<sur>qmovn<mode>):
18334 Implement using ss_truncate and us_truncate rather than
18336 * config/aarch64/iterators.md: Remove redundant unspecs and
18337 iterator: UNSPEC_[SU]QXTN and SUQMOVN respectively.
18339 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18341 * config/aarch64/arm_acle.h (__attribute__): Make intrinsic
18342 attributes consistent with those defined in arm_neon.h.
18344 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18346 * config/aarch64/arm_fp16.h (__attribute__): Make intrinsic
18347 attributes consistent with those defined in arm_neon.h.
18349 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18351 * config/aarch64/aarch64-simd-builtins.def: Add
18352 float_trunc_rodd builtin generator macros.
18353 * config/aarch64/aarch64-simd.md (aarch64_float_trunc_rodd_df):
18355 (aarch64_float_trunc_rodd_lo_v2sf): Define.
18356 (aarch64_float_trunc_rodd_hi_v4sf_le): Define.
18357 (aarch64_float_trunc_rodd_hi_v4sf_be): Define.
18358 (aarch64_float_trunc_rodd_hi_v4sf): Define.
18359 * config/aarch64/arm_neon.h (vcvtx_f32_f64): Use RTL builtin
18360 instead of inline asm.
18361 (vcvtx_high_f32_f64): Likewise.
18362 (vcvtxd_f32_f64): Likewise.
18363 * config/aarch64/iterators.md: Add FCVTXN unspec.
18365 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18367 * config/aarch64/aarch64-simd-builtins.def: Add tbx1 builtin
18369 * config/aarch64/aarch64-simd.md (aarch64_tbx1<mode>):
18371 * config/aarch64/arm_neon.h (vqtbx1_s8): USE RTL builtin
18372 instead of inline asm.
18373 (vqtbx1_u8): Likewise.
18374 (vqtbx1_p8): Likewise.
18375 (vqtbx1q_s8): Likewise.
18376 (vqtbx1q_u8): Likewise.
18377 (vqtbx1q_p8): Likewise.
18378 (vtbx2_s8): Likewise.
18379 (vtbx2_u8): Likewise.
18380 (vtbx2_p8): Likewise.
18382 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18384 * config/aarch64/aarch64-simd-builtins.def: Add tbl1 builtin
18386 * config/aarch64/arm_neon.h (vqtbl1_p8): Use RTL builtin
18387 instead of inline asm.
18388 (vqtbl1_s8): Likewise.
18389 (vqtbl1_u8): Likewise.
18390 (vqtbl1q_p8): Likewise.
18391 (vqtbl1q_s8): Likewise.
18392 (vqtbl1q_u8): Likewise.
18393 (vtbl1_s8): Likewise.
18394 (vtbl1_u8): Likewise.
18395 (vtbl1_p8): Likewise.
18396 (vtbl2_s8): Likewise.
18397 (vtbl2_u8): Likewise.
18398 (vtbl2_p8): Likewise.
18400 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18402 * config/aarch64/aarch64-simd-builtins.def: Add polynomial
18403 ssri_n buitin generator macro.
18404 * config/aarch64/arm_neon.h (vsri_n_p8): Use RTL builtin
18405 instead of inline asm.
18406 (vsri_n_p16): Likewise.
18407 (vsri_n_p64): Likewise.
18408 (vsriq_n_p8): Likewise.
18409 (vsriq_n_p16): Likewise.
18410 (vsriq_n_p64): Likewise.
18412 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18414 * config/aarch64/aarch64-simd-builtins.def: Use VALLP mode
18415 iterator for polynomial ssli_n builtin generator macro.
18416 * config/aarch64/arm_neon.h (vsli_n_p8): Use RTL builtin
18417 instead of inline asm.
18418 (vsli_n_p16): Likewise.
18419 (vsliq_n_p8): Likewise.
18420 (vsliq_n_p16): Likewise.
18421 * config/aarch64/iterators.md: Define VALLP mode iterator.
18423 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18425 * config/aarch64/aarch64-simd-builtins.def: Use VDQV_L
18426 iterator to generate [su]adalp RTL builtins.
18427 * config/aarch64/aarch64-simd.md: Use VDQV_L iterator in
18428 [su]adalp RTL pattern.
18429 * config/aarch64/arm_neon.h (vpadal_s32): Use RTL builtin
18430 instead of inline asm.
18431 (vpadal_u32): Likewise.
18433 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18435 * config/aarch64/aarch64-simd-builtins.def: Add [su]addlp
18436 builtin generator macros.
18437 * config/aarch64/aarch64-simd.md (aarch64_<su>addlp<mode>):
18439 * config/aarch64/arm_neon.h (vpaddl_s8): Use RTL builtin
18440 instead of inline asm.
18441 (vpaddl_s16): Likewise.
18442 (vpaddl_s32): Likewise.
18443 (vpaddl_u8): Likewise.
18444 (vpaddl_u16): Likewise.
18445 (vpaddl_u32): Likewise.
18446 (vpaddlq_s8): Likewise.
18447 (vpaddlq_s16): Likewise.
18448 (vpaddlq_s32): Likewise.
18449 (vpaddlq_u8): Likewise.
18450 (vpaddlq_u16): Likewise.
18451 (vpaddlq_u32): Liwewise.
18452 * config/aarch64/iterators.md: Define [SU]ADDLP unspecs with
18453 appropriate attributes.
18455 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18457 * config/aarch64/aarch64-simd-builtins.def: Use VDQ_I iterator
18458 for aarch64_addp<mode> builtin macro generator.
18459 * config/aarch64/aarch64-simd.md: Use VDQ_I iterator in
18460 aarch64_addp<mode> RTL pattern.
18461 * config/aarch64/arm_neon.h (vpaddq_s8): Use RTL builtin
18462 instead of inline asm.
18463 (vpaddq_s16): Likewise.
18464 (vpaddq_s32): Likewise.
18465 (vpaddq_s64): Likewise.
18466 (vpaddq_u8): Likewise.
18467 (vpaddq_u16): Likewise.
18468 (vpaddq_u32): Likewise.
18469 (vpaddq_u64): Likewise.
18471 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
18473 * config/aarch64/aarch64-simd-builtins.def: Add sq[r]dmulh_n
18474 builtin generator macros.
18475 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_n<mode>):
18477 * config/aarch64/arm_neon.h (vqdmulh_n_s16): Use RTL builtin
18478 instead of inline asm.
18479 (vqdmulh_n_s32): Likewise.
18480 (vqdmulhq_n_s16): Likewise.
18481 (vqdmulhq_n_s32): Likewise.
18482 (vqrdmulh_n_s16): Likewise.
18483 (vqrdmulh_n_s32): Likewise.
18484 (vqrdmulhq_n_s16): Likewise.
18485 (vqrdmulhq_n_s32): Likewise.
18487 2021-04-28 Tobias Burnus <tobias@codesourcery.com>
18489 * doc/install.texi (--enable-offload-defaulted): Document.
18491 2021-04-28 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
18493 * config/avr/avr-dimode.md: Turn existing patterns into
18494 define_insn_and_split style patterns where the splitter
18495 adds a clobber of the condition code register. Drop "cc"
18496 attribute. Add new patterns to match output of
18498 * config/avr/avr-fixed.md: Likewise.
18499 * config/avr/avr.c (cc_reg_rtx): New.
18500 (avr_parallel_insn_from_insns): Adjust insn count
18501 for removal of set of cc0.
18502 (avr_is_casesi_sequence): Likewise.
18503 (avr_casei_sequence_check_operands): Likewise.
18504 (avr_optimize_casesi): Likewise. Also insert
18505 new insns after jump_insn.
18506 (avr_pass_casesi::avr_rest_of_handle_casesi): Adjust
18507 for removal of set of cc0.
18508 (avr_init_expanders): Initialize cc_reg_rtx.
18509 (avr_regno_reg_class): Handle REG_CC.
18510 (cond_string): Remove usage of CC_OVERFLOW_UNUSABLE.
18511 (avr_notice_update_cc): Remove function.
18512 (ret_cond_branch): Remove usage of CC_OVERFLOW_UNUSABLE.
18513 (compare_condition): Adjust for PARALLEL with
18515 (out_shift_with_cnt): Likewise.
18516 (ashlhi3_out): Likewise.
18517 (ashrhi3_out): Likewise.
18518 (lshrhi3_out): Likewise.
18519 (avr_class_max_nregs): Return single reg for REG_CC.
18520 (avr_compare_pattern): Check for REG_CC instead
18522 (avr_reorg_remove_redundant_compare): Likewise.
18523 (avr_reorg):Adjust for PARALLEL with REG_CC clobber.
18524 (avr_hard_regno_nregs): Return single reg for REG_CC.
18525 (avr_hard_regno_mode_ok): Allow only CCmode for REG_CC.
18526 (avr_md_asm_adjust): Clobber REG_CC.
18527 (TARGET_HARD_REGNO_NREGS): Define.
18528 (TARGET_CLASS_MAX_NREGS): Define.
18529 (TARGET_MD_ASM_ADJUST): Define.
18530 * config/avr/avr.h (FIRST_PSEUDO_REGISTER): Adjust
18532 (enum reg_class): Add CC_REG class.
18533 (NOTICE_UPDATE_CC): Remove.
18534 (CC_OVERFLOW_UNUSABLE): Remove.
18535 (CC_NO_CARRY): Remove.
18536 * config/avr/avr.md: Turn existing patterns into
18537 define_insn_and_split style patterns where the splitter
18538 adds a clobber of the condition code register. Drop "cc"
18539 attribute. Add new patterns to match output of
18541 (sez): Remove unused pattern.
18543 2021-04-28 Richard Earnshaw <rearnsha@arm.com>
18546 * config/arm/arm.c (arm_hard_regno_mode_ok): Only allow VPR to be
18549 2021-04-28 Richard Sandiford <richard.sandiford@arm.com>
18552 * config/aarch64/constraints.md (Utq): Require the address to
18553 be valid for both the element mode and for V2DImode.
18555 2021-04-28 Jakub Jelinek <jakub@redhat.com>
18556 Tobias Burnus <tobias@codesourcery.com>
18558 * configure.ac (OFFLOAD_DEFAULTED): AC_DEFINE if offload-defaulted.
18559 * gcc.c (process_command): New variable.
18560 (driver::maybe_putenv_OFFLOAD_TARGETS): If OFFLOAD_DEFAULTED,
18561 set it if -foffload is defaulted.
18562 * lto-wrapper.c (OFFLOAD_TARGET_DEFAULT_ENV): Define.
18563 (compile_offload_image): If OFFLOAD_DEFAULTED and
18564 OFFLOAD_TARGET_DEFAULT is in the environment, don't fail
18565 if corresponding mkoffload can't be found.
18566 (compile_images_for_offload_targets): Likewise. Free and clear
18567 offload_names if no valid offload is found.
18568 * config.in: Regenerate.
18569 * configure: Regenerate.
18571 2021-04-28 Richard Biener <rguenther@suse.de>
18573 PR tree-optimization/100292
18574 * tree-vect-generic.c (expand_vector_condition): Do not fold
18577 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
18579 * config/rs6000/aix.h (SUBTARGET_DRIVER_SELF_SPECS): New.
18580 * config/rs6000/aix64.opt (m64): New.
18583 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
18585 * config/vax/vax.c (print_operand_address, vax_address_cost_1)
18586 (index_term_p): Handle ASHIFT too.
18588 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
18590 * config/vax/builtins.md (jbb<ccss>i<mode>): Remove operand #3.
18591 (sync_lock_test_and_set<mode>): Adjust accordingly.
18592 (sync_lock_release<mode>): Likewise.
18594 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
18596 * config/vax/vax-protos.h (adjacent_operands_p): Remove
18598 * config/vax/vax.c (adjacent_operands_p): Remove.
18600 2021-04-27 Maciej W. Rozycki <macro@linux-mips.org>
18602 * ifcvt.c (dead_or_predicable) [!IFCVT_MODIFY_TESTS]: Fall
18603 through to the non-conditional execution case if getting the
18604 condition for conditional execution has failed.
18606 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
18608 PR middle-end/100284
18609 * gimple.c (gimple_could_trap_p_1): Remove VEC_COND_EXPR test.
18610 * tree-eh.c (operation_could_trap_p): Handle VEC_COND_EXPR rather
18611 than asserting on it.
18613 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
18615 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): Protect
18616 with TARGET_AIX_OS.
18618 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
18621 * calls.c (precompute_register_parameters): Additionally test
18622 targetm.precompute_tls_p to pre-compute argument.
18623 * config/rs6000/aix.h (TARGET_PRECOMPUTE_TLS_P): Define.
18624 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): New.
18625 * target.def (precompute_tls_p): New.
18626 * doc/tm.texi.in (TARGET_PRECOMPUTE_TLS_P): Add hook documentation.
18627 * doc/tm.texi: Regenerated.
18629 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18632 * config/aarch64/aarch64.c (aarch64_print_operand): Cast -UINTVAL
18633 back to HOST_WIDE_INT.
18635 2021-04-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
18638 * simplify-rtx.c (simplify_context::simplify_subreg): Check the
18639 memory alignment for the outer mode.
18641 2021-04-27 H.J. Lu <hjl.tools@gmail.com>
18643 PR middle-end/90773
18644 * expr.c (op_by_pieces_d::get_usable_mode): New member function.
18645 (op_by_pieces_d::run): Cange a while loop to a do-while loop.
18647 2021-04-27 Alex Coplan <alex.coplan@arm.com>
18650 * config/arm/arm.c (arm_split_compare_and_swap): Fix up codegen
18651 with negative immediates: ensure we expand cbranchsi4_scratch
18652 correctly and ensure we satisfy its constraints.
18653 * config/arm/sync.md
18654 (@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): Don't
18655 attempt to tie two output operands together with constraints;
18656 collapse two alternatives.
18657 (@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
18658 * config/arm/thumb1.md (cbranchsi4_neg_late): New.
18660 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18663 * config/aarch64/predicates.md (aarch64_sub_immediate,
18664 aarch64_plus_immediate): Use -UINTVAL instead of -INTVAL.
18665 * config/aarch64/aarch64.md (casesi, rotl<mode>3): Likewise.
18666 * config/aarch64/aarch64.c (aarch64_print_operand,
18667 aarch64_split_atomic_op, aarch64_expand_subvti): Likewise.
18669 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18671 PR tree-optimization/100239
18672 * tree-vect-generic.c (lower_vec_perm): Don't accept constant
18673 permutations with all indices from the first zero element as vec_shl.
18675 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18677 PR rtl-optimization/100254
18678 * cfgcleanup.c (outgoing_edges_match): Check REG_EH_REGION on
18679 last1 and last2 insns rather than BB_END (bb1) and BB_END (bb2) insns.
18681 2021-04-27 Richard Biener <rguenther@suse.de>
18683 PR tree-optimization/99912
18684 * passes.def: Add comment about new TODO_remove_unused_locals.
18685 * tree-stdarg.c (pass_data_stdarg): Run TODO_remove_unused_locals
18688 2021-04-27 Richard Biener <rguenther@suse.de>
18690 PR tree-optimization/99912
18691 * passes.def (pass_all_optimizations): Add pass_dse before
18692 the first pass_dce, move the first pass_dse before the
18693 pass_dce following pass_pre.
18695 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18697 PR tree-optimization/95527
18698 * generic-match-head.c: Include tm.h.
18699 * gimple-match-head.c: Include tm.h.
18700 * match.pd (CLZ == INTEGER_CST): Don't use
18701 #ifdef CLZ_DEFINED_VALUE_AT_ZERO, only test CLZ_DEFINED_VALUE_AT_ZERO
18702 if clz == CFN_CLZ. Add missing val declaration.
18703 (CTZ cmp CST): New simplifications.
18705 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18707 PR tree-optimization/96696
18708 * expr.c (expand_expr_divmod): New function.
18709 (expand_expr_real_2) <case TRUNC_DIV_EXPR>: Use it for truncations and
18710 divisions. Formatting fixes.
18711 <case MULT_EXPR>: Optimize x / y * y as x - x % y if the latter is
18714 2021-04-27 Martin Jambor <mjambor@suse.cz>
18717 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
18718 If removing a call statement LHS SSA name, release it.
18720 2021-04-27 Richard Earnshaw <rearnsha@arm.com>
18723 * config/arm/arm.c (THUMB2_WORK_REGS): Check PIC_OFFSET_TABLE_REGNUM
18724 is valid before including it in the mask.
18726 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
18729 * config/aarch64/aarch64.c (aarch64_comp_type_attributes): Handle
18732 2021-04-27 Richard Biener <rguenther@suse.de>
18734 PR tree-optimization/100051
18735 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Add
18736 disambiguator based on access size vs. decl size.
18738 2021-04-27 Richard Biener <rguenther@suse.de>
18740 PR tree-optimization/100278
18741 * tree-ssa-pre.c (compute_avail): Give up when we cannot
18742 adjust TBAA beacuse of mismatching bases.
18744 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18747 * config/i386/i386.md (*<insn><mode>3_mask, *<insn><mode>3_mask_1):
18748 For any_rotate define_insn_split and following splitters, use
18749 SWI iterator instead of SWI48.
18751 2021-04-27 Richard Biener <rguenther@suse.de>
18753 PR tree-optimization/99776
18754 * match.pd (bit_field_ref (ctor)): Relax element extract
18755 type compatibility checks.
18757 2021-04-27 Cui,Lili <lili.cui@intel.com>
18759 * common/config/i386/i386-common.c (processor_names):
18760 Sync processor_names with processor_type.
18761 * config/i386/i386-options.c (processor_cost_table):
18762 Sync processor_cost_table with processor_type.
18764 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18766 * value-range.cc (irange::irange_set_1bit_anti_range): Add assert.
18767 (irange::set): Call irange_set_1bit_anti_range for handling all
18768 1-bit ranges. Fall through on ~[MIN,MAX].
18770 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18772 * value-range.cc (irange::legacy_num_pairs): Remove.
18773 (irange::invert): Change gcc_assert to gcc_checking_assert.
18774 * value-range.h (irange::num_pairs): Adjust for a cached
18775 num_pairs(). Also, rename all gcc_assert's to
18776 gcc_checking_assert's.
18778 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18780 * value-range.cc (irange::operator=): Set m_kind.
18781 (irange::copy_to_legacy): Handle varying and undefined sources
18782 as a legacy copy since they can be easily copied.
18783 (irange::irange_set): Set m_kind.
18784 (irange::irange_set_anti_range): Same.
18785 (irange::set): Rename normalize_min_max to normalize_kind.
18786 (irange::verify_range): Adjust for multi-ranges having the
18788 (irange::irange_union): Set m_kind.
18789 (irange::irange_intersect): Same.
18790 (irange::invert): Same.
18791 * value-range.h (irange::kind): Always return m_kind.
18792 (irange::varying_p): Rename to...
18793 (irange::varying_comptaible_p): ...this.
18794 (irange::undefined_p): Only look at m_kind.
18795 (irange::irange): Always set VR_UNDEFINED if applicable.
18796 (irange::set_undefined): Always set VR_UNDEFINED.
18797 (irange::set_varying): Always set m_kind to VR_VARYING.
18798 (irange::normalize_min_max): Rename to...
18799 (irange::normalize_kind): ...this.
18801 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18803 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
18804 Adjust for constant_p including varying_p.
18805 * tree-vrp.c (vrp_prop::finalize): Same.
18806 (determine_value_range): Same.
18807 * vr-values.c (vr_values::range_of_expr): Same.
18808 * value-range.cc (irange::symbolic_p): Do not check varying_p.
18809 (irange::constant_p): Same.
18811 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18813 * value-range.cc (irange::legacy_lower_bound): Replace
18814 !undefined_p check with num_ranges > 0.
18815 (irange::legacy_upper_bound): Same.
18816 * value-range.h (irange::type): Same.
18817 (irange::lower_bound): Same.
18818 (irange::upper_bound): Same.
18820 2021-04-26 Richard Biener <rguenther@suse.de>
18822 PR tree-optimization/99956
18823 * gimple-loop-interchange.cc (compute_access_stride):
18824 Try instantiating the access in a shallower loop nest
18825 if instantiating failed.
18826 (compute_access_strides): Pass adjustable loop_nest
18827 to compute_access_stride.
18829 2021-04-26 Christophe Lyon <christophe.lyon@linaro.org>
18831 * doc/sourcebuild.texi (arm_cmse_hw): Document.
18833 2021-04-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18835 * config/aarch64/iterators.md (vwcore): Handle V4BF, V8BF.
18837 2021-04-26 Thomas Schwinge <thomas@codesourcery.com>
18838 Nathan Sidwell <nathan@codesourcery.com>
18839 Tom de Vries <vries@codesourcery.com>
18840 Julian Brown <julian@codesourcery.com>
18841 Kwok Cheung Yeung <kcy@codesourcery.com>
18843 * omp-offload.c (oacc_validate_dims): Implement
18844 '-Wopenacc-parallelism'.
18845 * doc/invoke.texi (-Wopenacc-parallelism): Document.
18847 2021-04-26 Richard Biener <rguenther@suse.de>
18849 * tree-cfg.h (gimplify_build1): Remove.
18850 (gimplify_build2): Likewise.
18851 (gimplify_build3): Likewise.
18852 * tree-cfg.c (gimplify_build1): Move to tree-vect-generic.c.
18853 (gimplify_build2): Likewise.
18854 (gimplify_build3): Likewise.
18855 * tree-vect-generic.c (gimplify_build1): Move from tree-cfg.c.
18857 (gimplify_build2): Likewise.
18858 (gimplify_build3): Likewise.
18859 (tree_vec_extract): Use resimplify with following SSA edges.
18860 (expand_vector_parallel): Avoid passing NULL size/bitpos
18861 to tree_vec_extract.
18862 * expr.c (store_constructor): Deal with zero-element CTORs.
18863 * match.pd (bit_field_ref <vector CTOR>): Make sure to
18864 produce vector constants when possible.
18866 2021-04-26 Richard Biener <rguenther@suse.de>
18868 * tree-complex.c: Include gimple-fold.h.
18869 (expand_complex_addition): Use gimple_build.
18870 (expand_complex_multiplication_components): Likewise.
18871 (expand_complex_multiplication): Likewise.
18872 (expand_complex_div_straight): Likewise.
18873 (expand_complex_div_wide): Likewise.
18874 (expand_complex_division): Likewise.
18875 (expand_complex_conjugate): Likewise.
18876 (expand_complex_comparison): Likewise.
18878 2021-04-26 Richard Biener <rguenther@suse.de>
18880 * tree-ssa-phiopt.c (two_value_replacement): Remove use
18881 of legacy gimplify_buildN API.
18883 2021-04-26 Richard Biener <rguenther@suse.de>
18885 PR tree-optimization/99473
18886 * tree-ssa-phiopt.c (cond_store_replacement): Handle all
18889 2021-04-26 Richard Biener <rguenther@suse.de>
18891 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin):
18892 Use replace_call_with_value.
18894 2021-04-26 Richard Biener <rguenther@suse.de>
18896 * tree-ssa-propagate.h (valid_gimple_rhs_p): Remove.
18897 (update_gimple_call): Likewise.
18898 (update_call_from_tree): Likewise.
18899 * tree-ssa-propagate.c (valid_gimple_rhs_p): Remove.
18900 (valid_gimple_call_p): Likewise.
18901 (move_ssa_defining_stmt_for_defs): Likewise.
18902 (finish_update_gimple_call): Likewise.
18903 (update_gimple_call): Likewise.
18904 (update_call_from_tree): Likewise.
18905 (propagate_tree_value_into_stmt): Use replace_call_with_value.
18906 * gimple-fold.h (update_gimple_call): Declare.
18907 * gimple-fold.c (valid_gimple_rhs_p): Move here from
18908 tree-ssa-propagate.c.
18909 (update_gimple_call): Likewise.
18910 (valid_gimple_call_p): Likewise.
18911 (finish_update_gimple_call): Likewise, and simplify.
18912 (gimplify_and_update_call_from_tree): Implement
18913 update_call_from_tree functionality, avoid excessive
18914 push/pop_gimplify_context.
18915 (gimple_fold_builtin): Use only gimplify_and_update_call_from_tree.
18916 (gimple_fold_call): Likewise.
18917 * gimple-ssa-sprintf.c (try_substitute_return_value): Likewise.
18918 * tree-ssa-ccp.c (ccp_folder::fold_stmt): Likewise.
18919 (pass_fold_builtins::execute): Likewise.
18920 (optimize_stack_restore): Use replace_call_with_value.
18921 * tree-cfg.c (fold_loop_internal_call): Likewise.
18922 * tree-ssa-dce.c (maybe_optimize_arith_overflow): Use
18923 only gimplify_and_update_call_from_tree.
18924 * tree-ssa-strlen.c (handle_builtin_strlen): Likewise.
18925 (handle_builtin_strchr): Likewise.
18926 * tsan.c: Include gimple-fold.h instead of tree-ssa-propagate.h.
18928 2021-04-26 Jakub Jelinek <jakub@redhat.com>
18931 * vmsdbgout.c (ASM_OUTPUT_DEBUG_STRING, vmsdbgout_begin_block,
18932 vmsdbgout_end_block, lookup_filename, vmsdbgout_source_line): Remove
18935 2021-04-25 liuhongt <hongtao.liu@intel.com>
18938 * config/i386/i386-builtin.def (BDESC): Change the icode of
18939 the following builtins to CODE_FOR_nothing.
18940 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
18941 IX86_BUILTIN_PCMPEQB128, IX86_BUILTIN_PCMPEQW128,
18942 IX86_BUILTIN_PCMPEQD128, IX86_BUILTIN_PCMPEQQ,
18943 IX86_BUILTIN_PCMPEQB256, IX86_BUILTIN_PCMPEQW256,
18944 IX86_BUILTIN_PCMPEQD256, IX86_BUILTIN_PCMPEQQ256,
18945 IX86_BUILTIN_PCMPGTB128, IX86_BUILTIN_PCMPGTW128,
18946 IX86_BUILTIN_PCMPGTD128, IX86_BUILTIN_PCMPGTQ,
18947 IX86_BUILTIN_PCMPGTB256, IX86_BUILTIN_PCMPGTW256,
18948 IX86_BUILTIN_PCMPGTD256, IX86_BUILTIN_PCMPGTQ256.
18949 * config/i386/sse.md (avx2_eq<mode>3): Deleted.
18950 (sse2_eq<mode>3): Ditto.
18951 (sse4_1_eqv2di3): Ditto.
18952 (sse2_gt<mode>3): Rename to ..
18953 (*sse2_gt<mode>3): .. this.
18955 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
18958 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
18961 * config/darwin.c (darwin_binds_local_p): Assume that any
18962 public symbol might be interposed for PIC code. Update function
18963 header comment to reflect current Darwin capability.
18965 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
18968 * config/darwin.c (darwin_binds_local_p): Assume that any
18969 public symbol might be interposed for PIC code. Update function
18970 header comment to reflect current Darwin capability.
18972 2021-04-24 Richard Sandiford <richard.sandiford@arm.com>
18974 * doc/sourcebuild.texi: Document no-opts and any-opts target
18977 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
18979 * config/bpf/bpf.h (ASM_OUTPUT_ALIGNED_BSS): Use .type and .lcomm.
18981 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
18983 * config/bpf/bpf.h (FUNCTION_BOUNDARY): Set to 64.
18985 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
18988 * config/i386/i386-options.c (ix86_option_override_internal):
18989 Error out when -m96bit-long-double is used with 64bit targets.
18990 * config/i386/i386.md (*pushxf_rounded): Remove pattern.
18992 2021-04-23 Martin Liska <mliska@suse.cz>
18994 * lto-wrapper.c: Remove FIXME about usage of
18995 hardware_concurrency. The function is not on par with
18998 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
19001 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
19002 Copy operand 3 to operand 4. Use sse_reg_operand
19003 as operand 3 predicate.
19004 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): Ditto.
19005 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
19006 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
19007 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2):
19008 Copy operand 1 to operand 0.
19009 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): Ditto.
19010 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
19011 (LDX_ATOMIC/LDX_ATOMIC FP store peephole2 with mem blockage): Ditto.
19013 2021-04-23 Alex Coplan <alex.coplan@arm.com>
19015 PR rtl-optimization/100230
19016 * early-remat.c (early_remat::sort_candidates): Use delete[]
19017 instead of delete for array allocated with new[].
19019 2021-04-23 Richard Biener <rguenther@suse.de>
19021 * genmatch.c (lower_cond): Remove VEC_COND_EXPR special-casing.
19022 (capture_info::capture_info): Likewise.
19023 (capture_info::walk_match): Likewise.
19024 (expr::gen_transform): Likewise.
19025 (dt_simplify::gen_1): Likewise.
19026 * gimple-match-head.c (maybe_resimplify_conditional_op):
19027 Remove VEC_COND_EXPR special-casing.
19028 (gimple_simplify): Likewise.
19029 * gimple.c (gimple_could_trap_p_1): Adjust.
19030 * tree-ssa-pre.c (compute_avail): Allow VEC_COND_EXPR
19031 to participate in PRE.
19033 2021-04-23 Richard Biener <rguenther@suse.de>
19035 * cfganal.c (connect_infinite_loops_to_exit): First call
19036 add_noreturn_fake_exit_edges.
19037 * ipa-sra.c (process_scan_results): Do not call the now redundant
19038 add_noreturn_fake_exit_edges.
19039 * predict.c (tree_estimate_probability): Likewise.
19040 (rebuild_frequencies): Likewise.
19041 * store-motion.c (one_store_motion_pass): Likewise.
19043 2021-04-23 Richard Biener <rguenther@suse.de>
19045 PR tree-optimization/100222
19046 * predict.c (pass_profile::execute): Remove redundant call to
19047 mark_irreducible_loops.
19048 (report_predictor_hitrates): Likewise.
19050 2021-04-23 Richard Biener <rguenther@suse.de>
19052 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Avoid
19053 valid_gimple_rhs_p by instead gimplifying to one.
19055 2021-04-23 Richard Biener <rguenther@suse.de>
19057 PR tree-optimization/99971
19058 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
19059 Always use TBAA for loads.
19061 2021-04-23 liuhongt <hongtao.liu@intel.com>
19064 * config/i386/i386-options.c (ix86_option_override_internal):
19065 Clear MASK_AVX256_SPLIT_UNALIGNED_LOAD/STORE in x_target_flags
19066 when X86_TUNE_AVX256_UNALIGNED_LOAD/STORE_OPTIMAL is enabled
19067 by target attribute.
19069 2021-04-23 David Edelsohn <dje.gcc@gmail.com>
19071 * config/rs6000/aix71.h (PREFERRED_DEBUGGING_TYPE): Change to
19073 * config/rs6000/aix72.h (PREFERRED_DEBUGGING_TYPE): Same.
19075 2021-04-22 David Edelsohn <dje.gcc@gmail.com>
19077 * config.gcc (powerpc-ibm-aix6.*): Remove.
19078 * config/rs6000/aix61.h: Delete.
19080 2021-04-22 Martin Liska <mliska@suse.cz>
19082 PR testsuite/100159
19083 PR testsuite/100192
19084 * builtins.c (expand_builtin): Fix typos and missing comments.
19085 * dwarf2out.c (gen_subprogram_die): Likewise.
19086 (gen_struct_or_union_type_die): Likewise.
19088 2021-04-22 Uroš Bizjak <ubizjak@gmail.com>
19091 * config/i386/i386-expand.c (ix86_expand_convert_uns_sidf_sse):
19092 Remove the sign with FE_DOWNWARD, where x - x = -0.0.
19094 2021-04-21 Iain Sandoe <iain@sandoe.co.uk>
19096 * config/i386/darwin.h (TARGET_64BIT): Remove definition
19097 based on TARGET_ISA_64BIT.
19098 (TARGET_64BIT_P): Remove definition based on
19099 TARGET_ISA_64BIT_P().
19101 2021-04-21 Martin Liska <mliska@suse.cz>
19104 2021-04-21 Martin Liska <mliska@suse.cz>
19106 * lto-wrapper.c (cpuset_popcount): Remove.
19107 (init_num_threads): Remove and use hardware_concurrency.
19109 2021-04-21 Martin Liska <mliska@suse.cz>
19112 * main.c (main): Call toplev::finalize in CHECKING_P mode.
19113 * ipa-modref.c (ipa_modref_c_finalize): summaries are NULL
19114 when incremental LTO linking happens.
19116 2021-04-21 Martin Liska <mliska@suse.cz>
19118 * lto-wrapper.c (run_gcc): When -flto=jobserver is used, but the
19119 makeserver cannot be detected, then use -flto=N fallback.
19121 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
19123 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): When cross-compiling,
19124 default to yes for aarch64-linux-gnu.
19125 * configure: Regenerate.
19127 2021-04-21 Martin Liska <mliska@suse.cz>
19129 * lto-wrapper.c (cpuset_popcount): Remove.
19130 (init_num_threads): Remove and use hardware_concurrency.
19132 2021-04-21 Martin Liska <mliska@suse.cz>
19134 * config/i386/i386.c: Remove superfluous || TARGET_MACHO
19135 which remains to be '(... || 0)' and clang complains about it.
19136 * dwarf2out.c (AT_vms_delta): Declare conditionally.
19137 (add_AT_vms_delta): Likewise.
19138 * tree.c (fld_simplified_type): Use rather more common pattern
19139 for disabling of something (#if 0).
19140 (get_tree_code_name): Likewise.
19141 (verify_type_variant): Likewise.
19143 2021-04-21 Martin Liska <mliska@suse.cz>
19145 * config/i386/i386-expand.c (decide_alignment): Use newly named
19146 macro TARGET_CPU_P.
19147 * config/i386/i386.c (ix86_decompose_address): Likewise.
19148 (ix86_address_cost): Likewise.
19149 (ix86_lea_outperforms): Likewise.
19150 (ix86_avoid_lea_for_addr): Likewise.
19151 (ix86_add_stmt_cost): Likewise.
19152 * config/i386/i386.h (TARGET_*): Remove.
19153 (TARGET_CPU_P): New macro.
19154 * config/i386/i386.md: Use newly named macro TARGET_CPU_P.
19155 * config/i386/x86-tune-sched-atom.c (do_reorder_for_imul): Likewise.
19156 (swap_top_of_ready_list): Likewise.
19157 (ix86_atom_sched_reorder): Likewise.
19158 * config/i386/x86-tune-sched-bd.c (ix86_bd_has_dispatch): Likewise.
19159 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Likewise.
19161 2021-04-21 Martin Liska <mliska@suse.cz>
19163 * config/i386/i386-options.c (TARGET_EXPLICIT_NO_SAHF_P):
19165 (SET_TARGET_NO_SAHF): Likewise.
19166 (TARGET_EXPLICIT_PREFETCH_SSE_P): Likewise.
19167 (SET_TARGET_PREFETCH_SSE): Likewise.
19168 (TARGET_EXPLICIT_NO_TUNE_P): Likewise.
19169 (SET_TARGET_NO_TUNE): Likewise.
19170 (TARGET_EXPLICIT_NO_80387_P): Likewise.
19171 (SET_TARGET_NO_80387): Likewise.
19173 * config/i386/i386.h (TARGET_*): Remove.
19174 * opth-gen.awk: Generate new used macros.
19176 2021-04-21 Martin Liska <mliska@suse.cz>
19178 * config/i386/i386.h (PTA_*): Remove.
19179 (enum pta_flag): New.
19180 (DEF_PTA): Generate PTA_* values from i386-isa.def.
19181 * config/i386/i386-isa.def: New file.
19183 2021-04-21 Alex Coplan <alex.coplan@arm.com>
19186 * config/aarch64/aarch64-bti-insert.c (aarch64_bti_j_insn_p): New.
19187 (rest_of_insert_bti): Avoid inserting duplicate bti j insns for
19188 jump table targets.
19190 2021-04-21 H.J. Lu <hjl.tools@gmail.com>
19192 * config.gcc: Install mwaitintrin.h for i[34567]86-*-* and
19193 x86_64-*-* targets.
19194 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_MWAIT_SET):
19196 (OPTION_MASK_ISA2_MWAIT_UNSET): Likewise.
19197 (ix86_handle_option): Handle -mmwait.
19198 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
19199 Replace OPTION_MASK_ISA_SSE3 with OPTION_MASK_ISA2_MWAIT on
19200 __builtin_ia32_monitor and __builtin_ia32_mwait.
19201 * config/i386/i386-options.c (isa2_opts): Add -mmwait.
19202 (ix86_valid_target_attribute_inner_p): Likewise.
19203 (ix86_option_override_internal): Enable mwait/monitor
19204 instructions for -msse3.
19205 * config/i386/i386.h (TARGET_MWAIT): New.
19206 (TARGET_MWAIT_P): Likewise.
19207 * config/i386/i386.opt: Add -mmwait.
19208 * config/i386/mwaitintrin.h: New file.
19209 * config/i386/pmmintrin.h: Include <mwaitintrin.h>.
19210 * config/i386/sse.md (sse3_mwait): Replace TARGET_SSE3 with
19212 (@sse3_monitor_<mode>): Likewise.
19213 * config/i386/x86gprintrin.h: Include <mwaitintrin.h>.
19214 * doc/extend.texi: Document mwait target attribute.
19215 * doc/invoke.texi: Document -mmwait.
19217 2021-04-21 Martin Liska <mliska@suse.cz>
19219 * config/i386/i386-options.c (DEF_ENUM): Remove it.
19220 * config/i386/i386-opts.h (DEF_ENUM): Likewise.
19221 * config/i386/stringop.def (DEF_ENUM): Likewise.
19223 2021-04-21 Martin Liska <mliska@suse.cz>
19225 * tree-cfg.c (gimple_verify_flow_info): Use qD instead
19226 of print_generic_expr.
19228 2021-04-21 Jakub Jelinek <jakub@redhat.com>
19230 PR rtl-optimization/100148
19231 * cprop.c (constprop_register): Use next_nondebug_insn instead of
19234 2021-04-21 Martin Liska <mliska@suse.cz>
19237 * cgraphunit.c (cgraph_node::analyze): Remove duplicate
19238 free_dominance_info calls.
19240 2021-04-21 Richard Biener <rguenther@suse.de>
19242 * gimple-fold.c (maybe_fold_reference): Remove is_lhs
19243 parameter (and assume it to be false).
19244 (fold_gimple_assign): Adjust, remove all callers of
19245 maybe_fold_reference calling it with is_lhs true.
19246 (gimple_fold_call): Likewise.
19247 (fold_stmt_1): Likewise.
19249 2021-04-21 Richard Biener <rguenther@suse.de>
19251 * fold-const.c (pedantic_non_lvalue_loc): Remove.
19252 (fold_binary_loc): Adjust.
19253 (fold_ternary_loc): Likewise.
19255 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
19257 PR middle-end/100130
19258 * varasm.c (get_block_for_decl): Make sure that any use of the
19259 retain attribute matches the section's retain flag.
19260 (switch_to_section): Check for retain mismatches even when
19261 changing sections, but do not warn if the given decl is the
19262 section's named.decl.
19263 (output_object_block): Pass the first decl in the block (if any)
19264 to switch_to_section.
19266 2021-04-20 H.J. Lu <hjl.tools@gmail.com>
19268 * config/i386/i386-c.c (ix86_target_macros_internal): Define
19269 __CRC32__ for -mcrc32.
19270 * config/i386/i386-options.c (ix86_option_override_internal):
19271 Enable crc32 instruction for -msse4.2.
19272 * config/i386/i386.md (sse4_2_crc32<mode>): Remove TARGET_SSE4_2
19274 (sse4_2_crc32di): Likewise.
19275 * config/i386/ia32intrin.h: Use crc32 target option for CRC32
19278 2021-04-20 Segher Boessenkool <segher@kernel.crashing.org>
19281 * config/rs6000/rs6000.c (rs6000_machine_from_flags): Do not consider
19284 2021-04-20 Martin Liska <mliska@suse.cz>
19286 * doc/invoke.texi: Fix typo.
19287 * params.opt: Likewise.
19289 2021-04-20 Martin Liska <mliska@suse.cz>
19291 * doc/invoke.texi: Document new param.
19293 2021-04-19 Andrew MacLeod <amacleod@redhat.com>
19295 PR tree-optimization/100081
19296 * gimple-range-cache.h (ranger_cache): Inherit from gori_compute
19297 rather than gori_compute_cache.
19298 * gimple-range-gori.cc (is_gimple_logical_p): Move to top of file.
19299 (range_def_chain::m_logical_depth): New member.
19300 (range_def_chain::range_def_chain): Initialize m_logical_depth.
19301 (range_def_chain::get_def_chain): Don't build defchains through more
19302 than LOGICAL_LIMIT logical expressions.
19303 * params.opt (param_ranger_logical_depth): New.
19305 2021-04-19 Richard Earnshaw <rearnsha@arm.com>
19308 * config/arm/arm.c (arm_configure_build_target): Do not strip
19309 extended FPU/SIMD feature bits from the target ISA when -mfpu
19310 is specified (partial revert of r11-8168).
19312 2021-04-19 Thomas Schwinge <thomas@codesourcery.com>
19314 * params.opt (-param=openacc-kernels=): Add.
19315 * omp-oacc-kernels-decompose.cc
19316 (pass_omp_oacc_kernels_decompose::gate): Use it.
19317 * doc/invoke.texi (-fopenacc-kernels=@var{mode}): Move...
19318 (--param): ... here, 'openacc-kernels'.
19320 2021-04-19 Martin Liska <mliska@suse.cz>
19323 * gengtype.c (finish_root_table): Align function arguments
19324 in between declaration and definition.
19326 2021-04-19 Eric Botcazou <ebotcazou@adacore.com>
19328 * config/i386/winnt.c (i386_pe_seh_cold_init): Properly deal with
19329 frames larger than the SEH maximum frame size.
19331 2021-04-18 Segher Boessenkool <segher@kernel.crashing.org>
19333 PR rtl-optimization/99927
19334 * combine.c (distribute_notes) [REG_UNUSED]: If the register already
19335 is dead, just drop it.
19337 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
19340 * config/i386/winnt-d.c (TARGET_D_TEMPLATES_ALWAYS_COMDAT): Define.
19341 * doc/tm.texi: Regenerate.
19342 * doc/tm.texi.in (D language and ABI): Add @hook for
19343 TARGET_D_TEMPLATES_ALWAYS_COMDAT.
19345 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
19347 * config/darwin-d.c (darwin_d_handle_target_object_format): New
19349 (darwin_d_register_target_info): New function.
19350 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19351 * config/dragonfly-d.c (dragonfly_d_handle_target_object_format): New
19353 (dragonfly_d_register_target_info): New function.
19354 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19355 * config/freebsd-d.c (freebsd_d_handle_target_object_format): New
19357 (freebsd_d_register_target_info): New function.
19358 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19359 * config/glibc-d.c (glibc_d_handle_target_object_format): New
19361 (glibc_d_register_target_info): New function.
19362 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19363 * config/i386/i386-d.c (ix86_d_handle_target_object_format): New
19365 (ix86_d_register_target_info): Add ix86_d_handle_target_object_format
19366 as handler for objectFormat key.
19367 * config/i386/winnt-d.c (winnt_d_handle_target_object_format): New
19369 (winnt_d_register_target_info): New function.
19370 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19371 * config/netbsd-d.c (netbsd_d_handle_target_object_format): New
19373 (netbsd_d_register_target_info): New function.
19374 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19375 * config/openbsd-d.c (openbsd_d_handle_target_object_format): New
19377 (openbsd_d_register_target_info): New function.
19378 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19379 * config/pa/pa-d.c (pa_d_handle_target_object_format): New function.
19380 (pa_d_register_target_info): Add pa_d_handle_target_object_format as
19381 handler for objectFormat key.
19382 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_object_format): New
19384 (rs6000_d_register_target_info): Add
19385 rs6000_d_handle_target_object_format as handler for objectFormat key.
19386 * config/sol2-d.c (solaris_d_handle_target_object_format): New
19388 (solaris_d_register_target_info): New function.
19389 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
19391 2021-04-16 Jakub Jelinek <jakub@redhat.com>
19394 * config/aarch64/aarch64.c (aarch64_function_arg_alignment): Change
19395 abi_break argument from bool * to unsigned *, store there the pre-GCC 9
19397 (aarch64_layout_arg, aarch64_gimplify_va_arg_expr): Adjust callers.
19398 (aarch64_function_arg_regno_p): Likewise. Only emit -Wpsabi note if
19399 the old and new alignment after applying MIN/MAX to it is different.
19401 2021-04-16 Tamar Christina <tamar.christina@arm.com>
19404 * config/aarch64/aarch64-sve.md (@aarch64_sve_trn1_conv<mode>): New.
19405 * config/aarch64/aarch64.c (aarch64_expand_sve_const_pred_trn): Use new
19407 * config/aarch64/iterators.md (UNSPEC_TRN1_CONV): New.
19409 2021-04-16 Bill Schmidt <wschmidt@linux.ibm.com>
19411 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Revise
19412 this section and its subsections.
19414 2021-04-16 Jakub Jelinek <jakub@redhat.com>
19417 * config/aarch64/aarch64.md (*neg_asr_si2_extr, *extrsi5_insn_di): New
19418 define_insn patterns.
19420 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
19422 PR rtl-optimization/98689
19423 * reg-notes.def (UNTYPED_CALL): New note.
19424 * combine.c (distribute_notes): Handle it.
19425 * emit-rtl.c (try_split): Likewise.
19426 * rtlanal.c (rtx_properties::try_to_add_insn): Likewise. Assume
19427 that calls with the note implicitly set all return value registers.
19428 * builtins.c (expand_builtin_apply): Add a REG_UNTYPED_CALL
19431 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
19433 PR rtl-optimization/99596
19434 * rtlanal.c (rtx_properties::try_to_add_insn): Don't add global
19435 register accesses for const calls. Assume that pure functions
19436 can only read from global registers. Ignore cases in which
19437 the stack pointer has been marked global.
19439 2021-04-16 Jakub Jelinek <jakub@redhat.com>
19442 * tree-vect-loop.c (vect_transform_loop): Don't remove just
19443 dead scalar .MASK_LOAD calls, but also dead .COND_* calls - replace
19444 them by their last argument.
19446 2021-04-15 Martin Liska <mliska@suse.cz>
19448 * doc/invoke.texi: Other params don't use it, remove it.
19450 2021-04-15 Richard Biener <rguenther@suse.de>
19452 * gimple-builder.h: Add deprecation note.
19454 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
19457 * attribs.h (restrict_type_identity_attributes_to): Declare.
19458 * attribs.c (restrict_type_identity_attributes_to): New function.
19460 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
19463 * attribs.h (affects_type_identity_attributes): Declare.
19464 * attribs.c (remove_attributes_matching): New function.
19465 (affects_type_identity_attributes): Likewise.
19467 2021-04-15 Jakub Jelinek <jakub@redhat.com>
19470 * config/aarch64/aarch64.md (*<LOGICAL:optab>_<SHIFT:optab><mode>3):
19471 Add combine splitters for *<LOGICAL:optab>_ashl<mode>3 with
19472 ZERO_EXTEND, SIGN_EXTEND or AND.
19474 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
19476 PR rtl-optimization/99929
19477 * rtl.h (same_vector_encodings_p): New function.
19478 * cse.c (exp_equiv_p): Check that CONST_VECTORs have the same encoding.
19479 * cselib.c (rtx_equal_for_cselib_1): Likewise.
19480 * jump.c (rtx_renumbered_equal_p): Likewise.
19481 * lra-constraints.c (operands_match_p): Likewise.
19482 * reload.c (operands_match_p): Likewise.
19483 * rtl.c (rtx_equal_p_cb, rtx_equal_p): Likewise.
19485 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
19487 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Print
19488 more information about variable-length CONST_VECTORs.
19490 2021-04-14 Vladimir N. Makarov <vmakarov@redhat.com>
19492 PR rtl-optimization/100066
19493 * lra-constraints.c (split_reg): Check paradoxical_subreg_p for
19494 ordered modes when choosing splitting mode for hard reg.
19496 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
19499 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector_sel):
19501 (aarch64_expand_sve_const_vector): Use it for nelts_per_pattern==2.
19503 2021-04-14 Andreas Krebbel <krebbel@linux.ibm.com>
19505 * config/s390/s390-builtins.def (O_M5, O_M12, ...): Add new macros
19506 for mask operand types.
19507 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
19508 (s390_vec_permi_dbl, s390_vpdi): Use the M5 type for the immediate
19510 (s390_vec_msum_u128, s390_vmslg): Use the M12 type for the
19512 * config/s390/s390.c (s390_const_operand_ok): Check the new
19513 operand types and generate a list of valid values.
19515 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
19517 * doc/tm.texi: Regenerate.
19518 * doc/tm.texi.in (D language and ABI): Add @hook for
19519 TARGET_D_REGISTER_OS_TARGET_INFO.
19521 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
19523 * config/aarch64/aarch64-d.c (aarch64_d_handle_target_float_abi): New
19525 (aarch64_d_register_target_info): New function.
19526 * config/aarch64/aarch64-protos.h (aarch64_d_register_target_info):
19528 * config/aarch64/aarch64.h (TARGET_D_REGISTER_CPU_TARGET_INFO):
19530 * config/arm/arm-d.c (arm_d_handle_target_float_abi): New function.
19531 (arm_d_register_target_info): New function.
19532 * config/arm/arm-protos.h (arm_d_register_target_info): Declare.
19533 * config/arm/arm.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19534 * config/i386/i386-d.c (ix86_d_handle_target_float_abi): New function.
19535 (ix86_d_register_target_info): New function.
19536 * config/i386/i386-protos.h (ix86_d_register_target_info): Declare.
19537 * config/i386/i386.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19538 * config/mips/mips-d.c (mips_d_handle_target_float_abi): New function.
19539 (mips_d_register_target_info): New function.
19540 * config/mips/mips-protos.h (mips_d_register_target_info): Declare.
19541 * config/mips/mips.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19542 * config/pa/pa-d.c (pa_d_handle_target_float_abi): New function.
19543 (pa_d_register_target_info): New function.
19544 * config/pa/pa-protos.h (pa_d_register_target_info): Declare.
19545 * config/pa/pa.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19546 * config/riscv/riscv-d.c (riscv_d_handle_target_float_abi): New
19548 (riscv_d_register_target_info): New function.
19549 * config/riscv/riscv-protos.h (riscv_d_register_target_info): Declare.
19550 * config/riscv/riscv.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19551 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_float_abi): New
19553 (rs6000_d_register_target_info): New function.
19554 * config/rs6000/rs6000-protos.h (rs6000_d_register_target_info):
19556 * config/rs6000/rs6000.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19557 * config/s390/s390-d.c (s390_d_handle_target_float_abi): New function.
19558 (s390_d_register_target_info): New function.
19559 * config/s390/s390-protos.h (s390_d_register_target_info): Declare.
19560 * config/s390/s390.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19561 * config/sparc/sparc-d.c (sparc_d_handle_target_float_abi): New
19563 (sparc_d_register_target_info): New function.
19564 * config/sparc/sparc-protos.h (sparc_d_register_target_info): Declare.
19565 * config/sparc/sparc.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19566 * doc/tm.texi: Regenerate.
19567 * doc/tm.texi.in (D language and ABI): Add @hook for
19568 TARGET_D_REGISTER_CPU_TARGET_INFO.
19570 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
19572 * config/i386/i386-d.c (ix86_d_has_stdcall_convention): New function.
19573 * config/i386/i386-protos.h (ix86_d_has_stdcall_convention): Declare.
19574 * config/i386/i386.h (TARGET_D_HAS_STDCALL_CONVENTION): Define.
19575 * doc/tm.texi: Regenerate.
19576 * doc/tm.texi.in (D language and ABI): Add @hook for
19577 TARGET_D_HAS_STDCALL_CONVENTION.
19579 2021-04-14 Richard Biener <rguenther@suse.de>
19581 * tree-cfg.c (verify_gimple_assign_ternary): Verify that
19582 VEC_COND_EXPRs have a gimple_val condition.
19583 * tree-ssa-propagate.c (valid_gimple_rhs_p): VEC_COND_EXPR
19584 can no longer have a GENERIC condition.
19586 2021-04-14 Richard Earnshaw <rearnsha@arm.com>
19589 * config/arm/arm.c (arm_configure_build_target): Strip isa_all_fpbits
19590 from the isa_delta when -mfpu has been used.
19591 (arm_options_perform_arch_sanity_checks): It's the architecture that
19592 lacks an FPU not the processor.
19594 2021-04-13 Richard Biener <rguenther@suse.de>
19596 PR tree-optimization/100053
19597 * tree-ssa-sccvn.c (vn_nary_op_get_predicated_value): Do
19598 not use optimistic dominance queries for backedges to validate
19600 (dominated_by_p_w_unex): Add parameter to ignore executable
19601 state on backedges.
19602 (rpo_elim::eliminate_avail): Adjust.
19604 2021-04-13 Jakub Jelinek <jakub@redhat.com>
19607 * config/aarch64/aarch64.md (*aarch64_bfxil<mode>_extr,
19608 *aarch64_bfxilsi_extrdi): New define_insn patterns.
19610 2021-04-13 Jakub Jelinek <jakub@redhat.com>
19613 * simplify-rtx.c (simplify_immed_subreg): For MODE_COMPOSITE_P
19614 outermode, return NULL if the result doesn't encode back to the
19615 original byte sequence.
19616 (simplify_gen_subreg): Don't create SUBREGs from constants to
19617 MODE_COMPOSITE_P outermode.
19619 2021-04-12 Jakub Jelinek <jakub@redhat.com>
19621 PR rtl-optimization/99905
19622 * combine.c (expand_compound_operation): If pos + len > modewidth,
19623 perform the right shift by pos in inner_mode and then convert to mode,
19624 instead of trying to simplify a shift of rtx with inner_mode by pos
19625 as if it was a shift in mode.
19627 2021-04-12 Jakub Jelinek <jakub@redhat.com>
19630 * combine.c (simplify_and_const_int_1): Don't optimize varop
19631 away if it has side-effects.
19633 2021-04-12 Martin Liska <mliska@suse.cz>
19635 * doc/extend.texi: Escape @smallexample content.
19637 2021-04-12 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
19639 * config/s390/s390.md ("*movdi_31", "*movdi_64"): Add
19640 alternative in order to load a DFP zero.
19642 2021-04-12 Martin Liska <mliska@suse.cz>
19644 * doc/extend.texi: Be more precise in documentation
19645 of symver attribute.
19647 2021-04-12 Martin Liska <mliska@suse.cz>
19650 * gimplify.c (gimplify_expr): Right now, we unpoison all
19651 variables before a goto <dest>. We should not do it if we are
19654 2021-04-12 Cui,Lili <lili.cui@intel.com>
19656 * common/config/i386/cpuinfo.h (get_intel_cpu): Handle
19658 * common/config/i386/i386-common.c (processor_names): Add
19660 (processor_alias_table): Add rocketlake.
19661 * common/config/i386/i386-cpuinfo.h (processor_subtypes): Add
19662 INTEL_COREI7_ROCKETLAKE.
19663 * config.gcc: Add -march=rocketlake.
19664 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19666 * config/i386/i386-options.c (m_ROCKETLAKE) : Define.
19667 (processor_cost_table): Add rocketlake cost.
19668 * config/i386/i386.h (ix86_size_cost) : Define
19670 (processor_type) : Add PROCESSOR_ROCKETLAKE.
19671 (PTA_ROCKETLAKE): Ditto.
19672 * doc/extend.texi: Add rocketlake.
19673 * doc/invoke.texi: Add rocketlake.
19675 2021-04-12 Cui,Lili <lili.cui@intel.com>
19677 * config/i386/i386.h (PTA_ALDERLAKE): Change alderlake ISA list.
19678 * config/i386/i386-options.c (m_CORE_AVX2): Add m_ALDERLAKE.
19679 * common/config/i386/cpuinfo.h (get_intel_cpu): Add AlderLake model.
19680 * doc/invoke.texi: Change alderlake ISA list.
19682 2021-04-11 Hafiz Abid Qadeer <abidh@codesourcery.com>
19684 PR middle-end/98088
19685 * omp-expand.c (expand_oacc_collapse_init): Update condition in
19688 2021-04-10 H.J. Lu <hjl.tools@gmail.com>
19691 * config/i386/serializeintrin.h (_serialize): Defined as macro.
19693 2021-04-10 Jakub Jelinek <jakub@redhat.com>
19696 * expr.c (expand_expr_addr_expr_1): Test is_global_var rather than
19697 just TREE_STATIC on COMPOUND_LITERAL_EXPR_DECLs.
19699 2021-04-10 Jakub Jelinek <jakub@redhat.com>
19701 PR middle-end/99989
19702 * gimple-ssa-warn-alloca.c
19703 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit to
19704 0 with integer precision unconditionally.
19706 2021-04-10 Jakub Jelinek <jakub@redhat.com>
19708 PR rtl-optimization/98601
19709 * rtlanal.c (rtx_addr_can_trap_p_1): Allow in assert unknown size
19710 not just for BLKmode, but also for VOIDmode. For STRICT_ALIGNMENT
19711 unaligned_mems handle VOIDmode like BLKmode.
19713 2021-04-10 Jan Hubicka <hubicka@ucw.cz>
19716 * tree.c (free_lang_data_in_decl): Do not release body of
19717 declare_variant_alt.
19719 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
19721 * config/aarch64/aarch64.c (aarch64_option_restore): If the
19722 architecture was specified explicitly and the tuning wasn't,
19723 tune for the architecture rather than the configured default CPU.
19725 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
19727 * config/aarch64/aarch64.md (tlsdesc_small_sve_<mode>): Use X30
19728 as the temporary register.
19730 2021-04-09 Martin Liska <mliska@suse.cz>
19732 * doc/extend.texi: Move non-target attributes on the top level.
19734 2021-04-09 Martin Liska <mliska@suse.cz>
19736 * doc/invoke.texi: Document minimum and maximum value of the
19737 argument for both supported compression algorithms.
19739 2021-04-08 David Edelsohn <dje.gcc@gmail.com>
19741 * config/rs6000/rs6000.c (rs6000_xcoff_select_section): Select
19742 TLS BSS before TLS data.
19743 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .comm.
19745 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
19747 * doc/sourcebuild.texi (stdint_types_mbig_endian): Document.
19749 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
19751 * match.pd: Extend vec_cond folds to handle shifts.
19753 2021-04-08 Maciej W. Rozycki <macro@orcam.me.uk>
19755 * config/vax/vax.md: Fix comment for `*bit<mode>' pattern's
19758 2021-04-08 Alex Coplan <alex.coplan@arm.com>
19761 * config/arm/iterators.md (MVE_vecs): New.
19762 (V_elem): Also handle V2DF.
19763 * config/arm/mve.md (*mve_mov<mode>): Rename to ...
19764 (*mve_vdup<mode>): ... this. Remove second alternative since
19765 vec_duplicate of const_int is not canonical RTL, and we don't
19766 want to match symbol_refs.
19767 (*mve_vec_duplicate<mode>): Delete (pattern is redundant).
19769 2021-04-08 Xionghu Luo <luoxhu@linux.ibm.com>
19771 * fold-const.c (fold_single_bit_test): Fix typo.
19772 * print-rtl.c (print_rtx_insn_vec): Call print_rtl_single
19775 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
19777 PR tree-optimization/97513
19778 * tree-vect-slp.c (vect_add_slp_permutation): New function,
19780 (vectorizable_slp_permutation): ...here. Detect cases in which
19781 all VEC_PERM_EXPRs are guaranteed to have the same stepped
19782 permute vector and only generate one permute vector for that case.
19783 Extend that case to handle variable-length vectors.
19785 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
19787 PR tree-optimization/99873
19788 * tree-vect-slp.c (vect_slp_prefer_store_lanes_p): New function.
19789 (vect_build_slp_instance): Don't split store groups that could
19790 use IFN_STORE_LANES.
19792 2021-04-07 Jakub Jelinek <jakub@redhat.com>
19795 * varasm.c (output_constant_pool_contents): Don't strip name encoding
19796 from XSTR (desc->sym, 0) or from label before passing those to
19799 2021-04-07 Richard Biener <rguenther@suse.de>
19801 PR tree-optimization/99954
19802 * tree-loop-distribution.c: Include tree-affine.h.
19803 (generate_memcpy_builtin): Try using tree-affine to prove
19805 (loop_distribution::classify_builtin_ldst): Always classify
19808 2021-04-07 Richard Biener <rguenther@suse.de>
19810 PR tree-optimization/99947
19811 * tree-vect-loop.c (vectorizable_induction): Pre-allocate
19812 steps vector to avoid pushing elements from the reallocated
19815 2021-04-07 Richard Biener <rguenther@suse.de>
19817 * tree-ssa-sccvn.h (print_vn_reference_ops): Declare.
19818 * tree-ssa-pre.c (print_pre_expr): Factor out VN reference operand
19820 * tree-ssa-sccvn.c (print_vn_reference_ops): ... into this new
19822 (debug_vn_reference_ops): New.
19824 2021-04-07 Bin Cheng <bin.cheng@linux.alibaba.com>
19826 PR tree-optimization/98736
19827 * tree-loop-distribution.c
19828 * (loop_distribution::bb_top_order_init):
19829 Compute RPO with programing order preserved by calling function
19830 rev_post_order_and_mark_dfs_back_seme.
19832 2021-04-06 Vladimir N. Makarov <vmakarov@redhat.com>
19835 * lra-constraints.c (split_reg): Don't check paradoxical_subreg_p.
19836 * lra-lives.c (clear_sparseset_regnos, regnos_in_sparseset_p): New
19838 (process_bb_lives): Don't update biggest mode of hard reg for
19839 implicit in multi-register group. Use the new functions for
19840 updating dead_set and unused_set by register notes.
19842 2021-04-06 Xianmiao Qu <xianmiao_qu@c-sky.com>
19844 * config/csky/csky_pipeline_ck802.md : Use insn reservation name
19847 2021-04-06 H.J. Lu <hjl.tools@gmail.com>
19849 * config/i386/x86-tune-costs.h (skylake_memcpy): Updated.
19850 (skylake_memset): Likewise.
19851 (skylake_cost): Change CLEAR_RATIO to 17.
19852 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
19853 Replace m_CANNONLAKE, m_ICELAKE_CLIENT, m_ICELAKE_SERVER,
19854 m_TIGERLAKE and m_SAPPHIRERAPIDS with m_SKYLAKE and m_CORE_AVX512.
19856 2021-04-06 Richard Biener <rguenther@suse.de>
19858 PR tree-optimization/99880
19859 * tree-vect-loop.c (maybe_set_vectorized_backedge_value): Only
19860 set vectorized defs of relevant PHIs.
19862 2021-04-06 Richard Biener <rguenther@suse.de>
19864 PR tree-optimization/99924
19865 * tree-vect-slp.c (vect_bb_partition_graph_r): Do not mark
19866 nodes w/o scalar stmts as visited.
19868 2021-04-06 Alex Coplan <alex.coplan@arm.com>
19871 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Also use base
19872 PCS for [su]fix_optab.
19874 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
19876 * config/darwin.c (machopic_legitimize_pic_address): Check
19877 that the current pic register is one of the hard reg set
19878 before setting liveness.
19880 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
19882 * config/darwin.c (machopic_legitimize_pic_address): Fix
19883 whitespace, remove unused code.
19885 2021-04-03 Jakub Jelinek <jakub@redhat.com>
19887 PR tree-optimization/99882
19888 * gimple-ssa-store-merging.c (bswap_view_convert): Handle val with
19891 2021-04-03 Jakub Jelinek <jakub@redhat.com>
19893 PR rtl-optimization/99863
19894 * dse.c (replace_read): Drop regs_live argument. Instead of
19895 regs_live, use store_insn->fixed_regs_live if non-NULL,
19896 otherwise punt if insns sequence clobbers or sets any hard
19899 2021-04-03 Jakub Jelinek <jakub@redhat.com>
19902 * targhooks.h (default_print_patchable_function_entry_1): Declare.
19903 * targhooks.c (default_print_patchable_function_entry_1): New function,
19904 copied from default_print_patchable_function_entry with an added flags
19906 (default_print_patchable_function_entry): Rewritten into a small
19907 wrapper around default_print_patchable_function_entry_1.
19908 * config/rs6000/rs6000.c (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
19910 (rs6000_print_patchable_function_entry): New function.
19912 2021-04-02 Eric Botcazou <ebotcazou@adacore.com>
19914 * doc/invoke.texi (fdelete-dead-exceptions): Minor tweak.
19916 2021-04-01 Jason Merrill <jason@redhat.com>
19919 * common.opt: Document v15 and v16.
19921 2021-04-01 Richard Biener <rguenther@suse.de>
19923 PR tree-optimization/99863
19924 * gimplify.c (gimplify_init_constructor): Recompute vector
19927 2021-04-01 Jakub Jelinek <jakub@redhat.com>
19929 * doc/extend.texi (symver attribute): Fix up syntax errors
19932 2021-04-01 Jakub Jelinek <jakub@redhat.com>
19934 PR tree-optimization/96573
19935 * gimple-ssa-store-merging.c (init_symbolic_number): Handle
19936 also pointer types.
19938 2021-04-01 Richard Biener <rguenther@suse.de>
19940 PR tree-optimization/99856
19941 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Promote
19942 precision to vector element precision.
19944 2021-04-01 Martin Jambor <mjambor@suse.cz>
19946 PR tree-optimization/97009
19947 * tree-sra.c (access_or_its_child_written): New function.
19948 (propagate_subaccesses_from_rhs): Use it instead of a simple grp_write
19951 2021-03-31 Jan Hubicka <hubicka@ucw.cz>
19954 * cif-code.def (USES_COMDAT_LOCAL): Make CIF_FINAL_NORMAL.
19956 2021-03-31 Pat Haugen <pthaugen@linux.ibm.com>
19959 * config/rs6000/altivec.md (xxspltiw_v4si, xxspltiw_v4sf_inst,
19960 xxspltidp_v2df_inst, xxsplti32dx_v4si_inst, xxsplti32dx_v4sf_inst,
19961 xxblend_<mode>, xxpermx_inst, xxeval): Mark prefixed.
19962 * config/rs6000/mma.md (mma_<vvi4i4i8>, mma_<avvi4i4i8>,
19963 mma_<vvi4i4i2>, mma_<avvi4i4i2>, mma_<vvi4i4>, mma_<avvi4i4>,
19964 mma_<pvi4i2>, mma_<apvi4i2>, mma_<vvi4i4i4>, mma_<avvi4i4i4>):
19966 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Adjust test.
19967 * config/rs6000/rs6000.md (define_attr "maybe_prefixed"): New.
19968 (define_attr "prefixed"): Update initializer.
19970 2021-03-31 Jakub Jelinek <jakub@redhat.com>
19973 * dwarf2out.c (debug_ranges_dwo_section): New variable.
19974 (DW_RANGES_IDX_SKELETON): Define.
19975 (struct dw_ranges): Add begin_entry and end_entry members.
19976 (DEBUG_DWO_RNGLISTS_SECTION): Define.
19977 (add_ranges_num): Adjust r initializer for addition of *_entry
19979 (add_ranges_by_labels): For -gsplit-dwarf and force_direct,
19980 set idx to DW_RANGES_IDX_SKELETON.
19981 (use_distinct_base_address_for_range): New function.
19982 (index_rnglists): Don't set r->idx if it is equal to
19983 DW_RANGES_IDX_SKELETON. Initialize r->begin_entry and
19984 r->end_entry for -gsplit-dwarf if those will be needed by
19986 (output_rnglists): Add DWO argument. If true, switch to
19987 debug_ranges_dwo_section rather than debug_ranges_section.
19988 Adjust l1/l2 label indexes. Only output the offset table when
19989 dwo is true and don't include in there the skeleton range
19990 entry if present. For -gsplit-dwarf, skip ranges that belong
19991 to the other rnglists section. Change return type from void
19992 to bool and return true if there are any range entries for
19993 the other section. For dwarf_split_debug_info use
19994 DW_RLE_startx_endx, DW_RLE_startx_length and DW_RLE_base_addressx
19995 entries instead of DW_RLE_start_end, DW_RLE_start_length and
19996 DW_RLE_base_address. Use use_distinct_base_address_for_range.
19997 (init_sections_and_labels): Initialize debug_ranges_dwo_section
19998 if -gsplit-dwarf and DWARF >= 5. Adjust ranges_section_label
19999 and range_base_label indexes.
20000 (dwarf2out_finish): Call index_rnglists earlier before finalizing
20001 .debug_addr. Never emit DW_AT_rnglists_base attribute. For
20002 -gsplit-dwarf and DWARF >= 5 call output_rnglists up to twice
20003 with different dwo arguments.
20004 (dwarf2out_c_finalize): Clear debug_ranges_dwo_section.
20006 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
20008 PR tree-optimization/98268
20009 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Call
20010 recompute_tree_invariant_for_addr_expr after successfully
20011 folding a TARGET_MEM_REF that occurs inside an ADDR_EXPR.
20013 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
20015 PR tree-optimization/99726
20016 * tree-data-ref.c (create_intersect_range_checks_index): Bail
20017 out if there is more than one access function SCEV for the loop
20020 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
20022 PR rtl-optimization/97141
20023 PR rtl-optimization/98726
20024 * emit-rtl.c (valid_for_const_vector_p): Return true for
20026 * rtx-vector-builder.h (rtx_vector_builder::step): Return a
20027 poly_wide_int instead of a wide_int.
20028 (rtx_vector_builder::apply_set): Take a poly_wide_int instead
20030 * rtx-vector-builder.c (rtx_vector_builder::apply_set): Likewise.
20031 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Return
20032 false for CONST_VECTORs that cannot be forced to memory.
20033 * config/aarch64/aarch64-simd.md (mov<mode>): If a CONST_VECTOR
20034 is too complex to force to memory, build it up from individual
20037 2021-03-31 Jan Hubicka <jh@suse.cz>
20040 * cgraph.c (cgraph_node::release_body): Fix overactive check.
20042 2021-03-31 Christophe Lyon <christophe.lyon@linaro.org>
20045 * config/arm/vec-common.md (mul<mode>3): Disable on iwMMXT, expect
20048 2021-03-31 H.J. Lu <hjl.tools@gmail.com>
20050 * config/i386/i386-expand.c (expand_set_or_cpymem_via_rep):
20051 For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, don't convert QImode
20053 (decide_alg): For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, use
20054 "rep movsb/stosb" only for known sizes.
20055 * config/i386/i386-options.c (processor_cost_table): Use Ice
20056 Lake cost for Cannon Lake, Ice Lake, Tiger Lake, Sapphire
20057 Rapids and Alder Lake.
20058 * config/i386/i386.h (TARGET_PREFER_KNOWN_REP_MOVSB_STOSB): New.
20059 * config/i386/x86-tune-costs.h (icelake_memcpy): New.
20060 (icelake_memset): Likewise.
20061 (icelake_cost): Likewise.
20062 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
20065 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
20068 * config/aarch64/aarch64.c
20069 (aarch64_vectorize_preferred_vector_alignment): Query the size
20070 of the provided SVE vector; do not assume that all SVE vectors
20071 have the same size.
20073 2021-03-31 Jan Hubicka <jh@suse.cz>
20076 * cgraph.c (cgraph_node::release_body): Remove all callers and
20078 * cgraphclones.c (cgraph_node::materialize_clone): Do not do it here.
20079 * cgraphunit.c (cgraph_node::expand): And here.
20081 2021-03-31 Martin Liska <mliska@suse.cz>
20083 * ipa-modref.c (analyze_ssa_name_flags): Fix coding style
20084 and one negated condition.
20086 2021-03-31 Jakub Jelinek <jakub@redhat.com>
20087 Richard Sandiford <richard.sandiford@arm.com>
20090 * config/aarch64/aarch64.md (*add<mode>3_poly_1): Swap Uai and Uav
20091 constraints on operands[2] and similarly 0 and rk constraints
20092 on operands[1] corresponding to that.
20094 2021-03-31 Jakub Jelinek <jakub@redhat.com>
20097 * configure.ac (HAVE_LD_BROKEN_PE_DWARF5): New AC_DEFINE if PECOFF
20098 linker doesn't support DWARF sections new in DWARF5.
20099 * config/i386/i386-options.c (ix86_option_override_internal): Default
20100 to dwarf_version 4 if HAVE_LD_BROKEN_PE_DWARF5 for TARGET_PECOFF
20102 * config.in: Regenerated.
20103 * configure: Regenerated.
20105 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20108 * config/aarch64/aarch64.c (aarch64_analyze_loop_vinfo): Check for
20109 available issue_info before using it.
20111 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20114 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): Do not allow zero
20117 2021-03-30 Xionghu Luo <luoxhu@linux.ibm.com>
20120 * config/rs6000/altivec.md (altivec_lvsl_reg): Change to ...
20121 (altivec_lvsl_reg_<mode>): ... this.
20122 (altivec_lvsr_reg): Change to ...
20123 (altivec_lvsr_reg_<mode>): ... this.
20124 * config/rs6000/predicates.md (vec_set_index_operand): New.
20125 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20126 Enable 32bit variable vec_insert for all TARGET_VSX.
20127 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
20128 Enable 32bit variable vec_insert for p9 and above.
20129 (rs6000_expand_vector_set_var_p8): Rename to ...
20130 (rs6000_expand_vector_set_var_p7): ... this.
20131 (rs6000_expand_vector_set): Use TARGET_VSX and adjust assert
20133 * config/rs6000/vector.md (vec_set<mode>): Use vec_set_index_operand.
20134 * config/rs6000/vsx.md (xl_len_r): Use gen_altivec_lvsl_reg_di and
20135 gen_altivec_lvsr_reg_di.
20137 2021-03-30 H.J. Lu <hjl.tools@gmail.com>
20140 * config/i386/ia32intrin.h (__rdtsc): Defined as macro.
20141 (__rdtscp): Likewise.
20143 2021-03-30 Tamar Christina <tamar.christina@arm.com>
20145 PR tree-optimization/99825
20146 * tree-vect-slp-patterns.c (vect_check_evenodd_blend):
20147 Reject non-mult 2 lanes.
20149 2021-03-30 Richard Earnshaw <rearnsha@arm.com>
20152 * config/arm/arm.c (arm_file_start): Fix emission of
20153 Tag_ABI_VFP_args attribute.
20155 2021-03-30 Richard Biener <rguenther@suse.de>
20157 PR tree-optimization/99824
20158 * stor-layout.c (set_min_and_max_values_for_integral_type):
20159 Assert the precision is within the bounds of
20160 WIDE_INT_MAX_PRECISION.
20161 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
20162 the outermost component ref only to lower the access size
20163 and initialize that from the access type.
20165 2021-03-30 Richard Sandiford <richard.sandiford@arm.com>
20168 * config/aarch64/aarch64.md (mov<mode>): Pass multi-instruction
20169 CONST_INTs to aarch64_expand_mov_immediate when called after RA.
20171 2021-03-30 Mihailo Stojanovic <mihailo.stojanovic@typhoon-hil.com>
20173 * config/aarch64/aarch64.md
20174 (<optab>_trunc<fcvt_target><GPI:mode>2): Set the "arch"
20175 attribute to disambiguate between SIMD and FP variants of the
20178 2021-03-29 Jan Hubicka <hubicka@ucw.cz>
20180 * ipa-modref.c (merge_call_lhs_flags): Correct handling of deref.
20181 (analyze_ssa_name_flags): Fix typo in comment.
20183 2021-03-29 Alex Coplan <alex.coplan@arm.com>
20186 * config/aarch64/aarch64-sve-builtins.cc
20187 (function_builder::add_function): Add placeholder_p argument, use
20188 placeholder decls if this is set.
20189 (function_builder::add_unique_function): Instead of conditionally adding
20190 direct overloads, unconditionally add either a direct overload or a
20192 (function_builder::add_overloaded_function): Set placeholder_p if we're
20193 using C++ overloads. Use the obstack for string storage instead
20194 of relying on the tree nodes.
20195 (function_builder::add_overloaded_functions): Don't return early for
20196 m_direct_overloads: we need to add placeholders.
20197 * config/aarch64/aarch64-sve-builtins.h
20198 (function_builder::add_function): Add placeholder_p argument.
20200 2021-03-29 Richard Biener <rguenther@suse.de>
20202 PR tree-optimization/99807
20203 * tree-vect-slp.c (vect_slp_analyze_node_operations_1): Move
20204 assert below VEC_PERM handling.
20206 2021-03-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20209 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>): Use
20210 aarch64_simd_or_scalar_imm_zero to match zeroes. Remove pattern
20211 matching const_int 0.
20212 (move_lo_quad_internal_be_<mode>): Likewise.
20213 (move_lo_quad_<mode>): Update for the above.
20214 * config/aarch64/iterators.md (VQ_2E): Delete.
20216 2021-03-29 Jakub Jelinek <jakub@redhat.com>
20218 PR tree-optimization/99777
20219 * fold-const.c (extract_muldiv_1): For conversions, punt on casts from
20220 types other than scalar integral types.
20222 2021-03-28 David Edelsohn <dje.gcc@gmail.com>
20224 * config/rs6000/rs6000.c (rs6000_output_dwarf_dtprel): Do not add
20225 XCOFF TLS reloc decorations.
20227 2021-03-28 Gerald Pfeifer <gerald@pfeifer.com>
20229 * doc/analyzer.texi (Analyzer Internals): Update link to
20230 "A Memory Model for Static Analysis of C Programs".
20232 2021-03-26 David Edelsohn <dje.gcc@gmail.com>
20234 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Call function.
20235 * config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align):
20237 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align): New.
20238 (rs6000_special_round_type_align): Recursively check innermost first
20241 2021-03-26 Jakub Jelinek <jakub@redhat.com>
20244 * dwarf2out.h (struct dw_fde_node): Add rule18 member.
20245 * dwarf2cfi.c (dwarf2out_frame_debug_expr): When handling (set hfp sp)
20246 assignment with drap_reg active, queue reg save for hfp with offset 0
20247 and flush queued reg saves. When handling a push with rule18,
20248 defer queueing reg save for hfp and just assert the offset is 0.
20249 (scan_trace): Assert that fde->rule18 is false.
20251 2021-03-26 Vladimir Makarov <vmakarov@redhat.com>
20254 * ira-costs.c (record_reg_classes): Put case with
20255 CT_RELAXED_MEMORY adjacent to one with CT_MEMORY.
20256 * ira.c (ira_setup_alts): Ditto.
20257 * lra-constraints.c (process_alt_operands): Ditto.
20258 * recog.c (asm_operand_ok): Ditto.
20259 * reload.c (find_reloads): Ditto.
20261 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20263 * config/aarch64/aarch64-protos.h
20264 (cpu_addrcost_table::post_modify_ld3_st3): New member variable.
20265 (cpu_addrcost_table::post_modify_ld4_st4): Likewise.
20266 * config/aarch64/aarch64.c (generic_addrcost_table): Update
20267 accordingly, using the same costs as for post_modify.
20268 (exynosm1_addrcost_table, xgene1_addrcost_table): Likewise.
20269 (thunderx2t99_addrcost_table, thunderx3t110_addrcost_table):
20270 (tsv110_addrcost_table, qdf24xx_addrcost_table): Likewise.
20271 (a64fx_addrcost_table): Likewise.
20272 (neoversev1_addrcost_table): New.
20273 (neoversev1_tunings): Use neoversev1_addrcost_table.
20274 (aarch64_address_cost): Use the new post_modify costs for CImode
20277 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20279 * config/aarch64/aarch64.opt
20280 (-param=aarch64-loop-vect-issue-rate-niters=): New parameter.
20281 * doc/invoke.texi: Document it.
20282 * config/aarch64/aarch64-protos.h (aarch64_base_vec_issue_info)
20283 (aarch64_scalar_vec_issue_info, aarch64_simd_vec_issue_info)
20284 (aarch64_advsimd_vec_issue_info, aarch64_sve_vec_issue_info)
20285 (aarch64_vec_issue_info): New structures.
20286 (cpu_vector_cost): Write comments above the variables rather
20288 (cpu_vector_cost::issue_info): New member variable.
20289 * config/aarch64/aarch64.c: Include gimple-pretty-print.h
20290 and tree-ssa-loop-niter.h.
20291 (generic_vector_cost, a64fx_vector_cost, qdf24xx_vector_cost)
20292 (thunderx_vector_cost, tsv110_vector_cost, cortexa57_vector_cost)
20293 (exynosm1_vector_cost, xgene1_vector_cost, thunderx2t99_vector_cost)
20294 (thunderx3t110_vector_cost): Initialize issue_info to null.
20295 (neoversev1_scalar_issue_info, neoversev1_advsimd_issue_info)
20296 (neoversev1_sve_issue_info, neoversev1_vec_issue_info): New structures.
20297 (neoversev1_vector_cost): Use them.
20298 (aarch64_vec_op_count, aarch64_sve_op_count): New structures.
20299 (aarch64_vector_costs::saw_sve_only_op): New member variable.
20300 (aarch64_vector_costs::num_vector_iterations): Likewise.
20301 (aarch64_vector_costs::scalar_ops): Likewise.
20302 (aarch64_vector_costs::advsimd_ops): Likewise.
20303 (aarch64_vector_costs::sve_ops): Likewise.
20304 (aarch64_vector_costs::seen_loads): Likewise.
20305 (aarch64_simd_vec_costs_for_flags): New function.
20306 (aarch64_analyze_loop_vinfo): Initialize num_vector_iterations.
20307 Count the number of predicate operations required by SVE WHILE
20309 (aarch64_comparison_type, aarch64_multiply_add_p): New functions.
20310 (aarch64_sve_only_stmt_p, aarch64_in_loop_reduction_latency): Likewise.
20311 (aarch64_count_ops): Likewise.
20312 (aarch64_add_stmt_cost): Record whether see an SVE operation
20313 that cannot currently be implementing using Advanced SIMD.
20314 Record issue information about the scalar, Advanced SIMD
20315 and (where relevant) SVE versions of a loop.
20316 (aarch64_vec_op_count::dump): New function.
20317 (aarch64_sve_op_count::dump): Likewise.
20318 (aarch64_estimate_min_cycles_per_iter): Likewise.
20319 (aarch64_adjust_body_cost): If issue information is available,
20320 try to compare the issue rates of the various loop implementations
20321 and increase or decrease the vector body cost accordingly.
20323 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20325 * config/aarch64/aarch64.c (aarch64_detect_vector_stmt_subtype):
20326 Assume a zero cost for induction phis.
20328 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20330 * config/aarch64/aarch64.c (aarch64_embedded_comparison_type): New
20332 (aarch64_adjust_stmt_cost): Add the costs of embedded scalar and
20333 vector comparisons.
20335 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20337 * config/aarch64/aarch64.c (aarch64_detect_scalar_stmt_subtype):
20339 (aarch64_add_stmt_cost): Call it.
20341 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20343 * config/aarch64/aarch64-tuning-flags.def (matched_vector_throughput):
20344 New tuning parameter.
20345 * config/aarch64/aarch64.c (neoversev1_tunings): Use it.
20346 (aarch64_estimated_sve_vq): New function.
20347 (aarch64_vector_costs::analyzed_vinfo): New member variable.
20348 (aarch64_vector_costs::is_loop): Likewise.
20349 (aarch64_vector_costs::unrolled_advsimd_niters): Likewise.
20350 (aarch64_vector_costs::unrolled_advsimd_stmts): Likewise.
20351 (aarch64_record_potential_advsimd_unrolling): New function.
20352 (aarch64_analyze_loop_vinfo, aarch64_analyze_bb_vinfo): Likewise.
20353 (aarch64_add_stmt_cost): Call aarch64_analyze_loop_vinfo or
20354 aarch64_analyze_bb_vinfo on the first use of a costs structure.
20355 Detect whether we're vectorizing a loop for SVE that might be
20356 completely unrolled if it used Advanced SIMD instead.
20357 (aarch64_adjust_body_cost_for_latency): New function.
20358 (aarch64_finish_cost): Call it.
20360 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20362 * config/aarch64/aarch64.c (aarch64_vector_costs): New structure.
20363 (aarch64_init_cost): New function.
20364 (aarch64_add_stmt_cost): Use aarch64_vector_costs instead of
20365 the default unsigned[3].
20366 (aarch64_finish_cost, aarch64_destroy_cost_data): New functions.
20367 (TARGET_VECTORIZE_INIT_COST): Override.
20368 (TARGET_VECTORIZE_FINISH_COST): Likewise.
20369 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
20371 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20373 * config/aarch64/aarch64.c (neoversev1_advsimd_vector_cost)
20374 (neoversev1_sve_vector_cost): New cost structures.
20375 (neoversev1_vector_cost): Likewise.
20376 (neoversev1_tunings): Use them. Enable use_new_vector_costs.
20378 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20380 * config/aarch64/aarch64-protos.h
20381 (sve_vec_cost::scatter_store_elt_cost): New member variable.
20382 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
20383 accordingly, taking the cost from the cost of a scalar_store.
20384 (a64fx_sve_vector_cost): Likewise.
20385 (aarch64_detect_vector_stmt_subtype): Detect scatter stores.
20387 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20389 * config/aarch64/aarch64-protos.h
20390 (simd_vec_cost::store_elt_extra_cost): New member variable.
20391 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
20392 accordingly, using the vec_to_scalar cost for the new field.
20393 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
20394 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
20395 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
20396 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
20397 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
20398 (thunderx3t110_advsimd_vector_cost): Likewise.
20399 (aarch64_detect_vector_stmt_subtype): Detect single-element stores.
20401 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20403 * config/aarch64/aarch64-protos.h (simd_vec_cost::ld2_st2_permute_cost)
20404 (simd_vec_cost::ld3_st3_permute_cost): New member variables.
20405 (simd_vec_cost::ld4_st4_permute_cost): Likewise.
20406 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
20407 accordingly, using zero for the new costs.
20408 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
20409 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
20410 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
20411 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
20412 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
20413 (thunderx3t110_advsimd_vector_cost): Likewise.
20414 (aarch64_ld234_st234_vectors): New function.
20415 (aarch64_adjust_stmt_cost): Likewise.
20416 (aarch64_add_stmt_cost): Call aarch64_adjust_stmt_cost if using
20417 the new vector costs.
20419 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20421 * config/aarch64/aarch64-protos.h (sve_vec_cost): Turn into a
20422 derived class of simd_vec_cost. Add information about CLAST[AB]
20423 and FADDA instructions.
20424 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
20425 accordingly, using the vec_to_scalar costs for the new fields.
20426 (a64fx_sve_vector_cost): Likewise.
20427 (aarch64_reduc_type): New function.
20428 (aarch64_sve_in_loop_reduction_latency): Likewise.
20429 (aarch64_detect_vector_stmt_subtype): Take a vinfo parameter.
20430 Use aarch64_sve_in_loop_reduction_latency to handle SVE reductions
20431 that occur in the loop body.
20432 (aarch64_add_stmt_cost): Update call accordingly.
20434 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
20436 * config/aarch64/aarch64-tuning-flags.def (use_new_vector_costs):
20438 * config/aarch64/aarch64-protos.h (simd_vec_cost): Put comments
20439 above the fields rather than to the right.
20440 (simd_vec_cost::reduc_i8_cost): New member variable.
20441 (simd_vec_cost::reduc_i16_cost): Likewise.
20442 (simd_vec_cost::reduc_i32_cost): Likewise.
20443 (simd_vec_cost::reduc_i64_cost): Likewise.
20444 (simd_vec_cost::reduc_f16_cost): Likewise.
20445 (simd_vec_cost::reduc_f32_cost): Likewise.
20446 (simd_vec_cost::reduc_f64_cost): Likewise.
20447 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
20448 accordingly, using the vec_to_scalar_cost for the new fields.
20449 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
20450 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
20451 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
20452 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
20453 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
20454 (thunderx3t110_advsimd_vector_cost): Likewise.
20455 (aarch64_use_new_vector_costs_p): New function.
20456 (aarch64_simd_vec_costs): New function, split out from...
20457 (aarch64_builtin_vectorization_cost): ...here.
20458 (aarch64_is_reduction): New function.
20459 (aarch64_detect_vector_stmt_subtype): Likewise.
20460 (aarch64_add_stmt_cost): Call aarch64_detect_vector_stmt_subtype if
20461 using the new vector costs.
20463 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
20466 * tree-emutls.c (get_emutls_init_templ_addr): Mark initializer of weak
20467 TLS declarations as public.
20469 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
20471 * config/aarch64/aarch64-d.c (IN_TARGET_CODE): Define.
20472 * config/arm/arm-d.c (IN_TARGET_CODE): Likewise.
20473 * config/i386/i386-d.c (IN_TARGET_CODE): Likewise.
20474 * config/mips/mips-d.c (IN_TARGET_CODE): Likewise.
20475 * config/pa/pa-d.c (IN_TARGET_CODE): Likewise.
20476 * config/riscv/riscv-d.c (IN_TARGET_CODE): Likewise.
20477 * config/rs6000/rs6000-d.c (IN_TARGET_CODE): Likewise.
20478 * config/s390/s390-d.c (IN_TARGET_CODE): Likewise.
20479 * config/sparc/sparc-d.c (IN_TARGET_CODE): Likewise.
20481 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
20484 * config.gcc (*-*-cygwin*): Add winnt-d.o
20485 (*-*-mingw*): Likewise.
20486 * config/i386/cygwin.h (EXTRA_TARGET_D_OS_VERSIONS): New macro.
20487 * config/i386/mingw32.h (EXTRA_TARGET_D_OS_VERSIONS): Likewise.
20488 * config/i386/t-cygming: Add winnt-d.o.
20489 * config/i386/winnt-d.c: New file.
20491 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
20493 * config/freebsd-d.c: Include memmodel.h.
20495 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
20498 * config.gcc (*-*-openbsd*): Add openbsd-d.o.
20499 * config/t-openbsd: Add openbsd-d.o.
20500 * config/openbsd-d.c: New file.
20502 2021-03-25 Stam Markianos-Wright <stam.markianos-wright@arm.com>
20504 PR tree-optimization/96974
20505 * tree-vect-stmts.c (vect_get_vector_types_for_stmt): Replace assert
20506 with graceful exit.
20508 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
20511 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
20515 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
20516 always_inline in system headers.
20518 2021-03-25 Kewen Lin <linkw@linux.ibm.com>
20520 * tree-vect-loop.c (vect_model_reduction_cost): Init inside_cost.
20522 2021-03-25 Jakub Jelinek <jakub@redhat.com>
20525 * tree-core.h (enum operand_equal_flag): Add OEP_ADDRESS_OF_SAME_FIELD.
20526 * fold-const.c (operand_compare::operand_equal_p): Don't compare
20527 field offsets if OEP_ADDRESS_OF_SAME_FIELD.
20529 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
20533 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
20534 always_inline in system headers.
20536 2021-03-25 Richard Biener <rguenther@suse.de>
20538 PR tree-optimization/99746
20539 * tree-vect-slp-patterns.c (complex_pattern::build): Do not mark
20540 the scalar stmt as patterned. Instead set up required things
20543 2021-03-25 Xionghu Luo <luoxhu@linux.ibm.com>
20545 * config/rs6000/rs6000.c (power8_costs): Change l2 cache
20548 2021-03-24 Martin Liska <mliska@suse.cz>
20551 * common/config/i386/i386-common.c (ARRAY_SIZE): Fix off-by-one
20553 * config/i386/i386-options.c (ix86_option_override_internal):
20554 Add run-time assert.
20556 2021-03-24 Martin Jambor <mjambor@suse.cz>
20559 * ipa-cp.c (initialize_node_lattices): Mark as bottom all
20560 parameters with unknown type.
20561 (ipacp_value_safe_for_type): New function.
20562 (propagate_vals_across_arith_jfunc): Verify that the constant type
20563 can be used for a type of the formal parameter.
20564 (propagate_vals_across_ancestor): Likewise.
20565 (propagate_scalar_across_jump_function): Likewise. Pass the type
20566 also to propagate_vals_across_ancestor.
20568 2021-03-24 Christophe Lyon <christophe.lyon@linaro.org>
20571 * config/arm/mve.md (movmisalign<mode>_mve_store): Use Ux
20573 (movmisalign<mode>_mve_load): Likewise.
20575 2021-03-24 Jakub Jelinek <jakub@redhat.com>
20578 * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
20579 movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.
20581 2021-03-24 Alexandre Oliva <oliva@adacore.com>
20583 * doc/sourcebuild.texi (sysconf): New effective target.
20585 2021-03-24 Alexandre Oliva <oliva@adacore.com>
20587 * config/i386/predicates.md (reg_or_const_vec_operand): New.
20588 * config/i386/sse.md (ssse3_pshufbv8qi3): Add an expander for
20589 the now *-prefixed insn_and_split, turn the splitter const vec
20590 into an input for the insn, making it an ignored immediate for
20591 non-split cases, and loaded into the scratch register
20594 2021-03-23 Vladimir N. Makarov <vmakarov@redhat.com>
20597 * config/aarch64/constraints.md (Utq, UOb, UOh, UOw, UOd, UOty):
20598 Use define_relaxed_memory_constraint for them.
20600 2021-03-23 Iain Sandoe <iain@sandoe.co.uk>
20603 * config/host-darwin.c (darwin_gt_pch_use_address): Add a
20604 colon to the diagnostic message.
20606 2021-03-23 Ilya Leoshkevich <iii@linux.ibm.com>
20608 * fwprop.c (fwprop_propagation::fwprop_propagation): Look at
20610 (try_fwprop_subst_note): Use set_info instead of insn_info.
20611 (try_fwprop_subst_pattern): Likewise.
20612 (try_fwprop_subst_notes): Likewise.
20613 (try_fwprop_subst): Likewise.
20614 (forward_propagate_subreg): Likewise.
20615 (forward_propagate_and_simplify): Likewise.
20616 (forward_propagate_into): Likewise.
20617 * rtl-ssa/accesses.h (set_info::single_nondebug_use) New
20619 (set_info::single_nondebug_insn_use): Likewise.
20620 (set_info::single_phi_use): Likewise.
20621 * rtl-ssa/member-fns.inl (set_info::single_nondebug_use) New
20623 (set_info::single_nondebug_insn_use): Likewise.
20624 (set_info::single_phi_use): Likewise.
20626 2021-03-23 Christophe Lyon <christophe.lyon@linaro.org>
20628 * doc/sourcebuild.texi (arm_dsp_ok, arm_dsp): Document.
20630 2021-03-23 Jakub Jelinek <jakub@redhat.com>
20633 * config/aarch64/aarch64.c (aarch64_add_offset): Tell
20634 expand_mult to perform an unsigned rather than a signed
20637 2021-03-23 H.J. Lu <hjl.tools@gmail.com>
20640 * config/i386/cpuid.h (__cpuid): Add __volatile__.
20641 (__cpuid_count): Likewise.
20643 2021-03-23 Richard Biener <rguenther@suse.de>
20645 PR tree-optimization/99721
20646 * tree-vect-slp.c (vect_slp_analyze_node_operations):
20647 Make sure we can schedule the node.
20649 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20651 * config/riscv/riscv.c (riscv_subword): Take endianness into
20652 account when calculating the byte offset.
20654 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20656 * config/riscv/predicates.md (subreg_lowpart_operator): New predicate
20657 * config/riscv/riscv.md (*addsi3_extended2, *subsi3_extended2)
20658 (*negsi2_extended2, *mulsi3_extended2, *<optab>si3_mask)
20659 (*<optab>si3_mask_1, *<optab>di3_mask, *<optab>di3_mask_1)
20660 (*<optab>si3_extend_mask, *<optab>si3_extend_mask_1): Use
20661 new predicate "subreg_lowpart_operator"
20663 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20665 * config/riscv/riscv.c (riscv_swap_instruction): New function
20666 to byteswap an SImode rtx containing an instruction.
20667 (riscv_trampoline_init): Byteswap the generated instructions
20670 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20672 * common/config/riscv/riscv-common.c
20673 (TARGET_DEFAULT_TARGET_FLAGS): Set default endianness.
20674 * config.gcc (riscv32be-*, riscv64be-*): Set
20675 TARGET_BIG_ENDIAN_DEFAULT to 1.
20676 * config/riscv/elf.h (LINK_SPEC): Change -melf* value
20677 depending on default endianness.
20678 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
20679 * config/riscv/linux.h (LINK_SPEC): Likewise.
20680 * config/riscv/riscv.c (TARGET_DEFAULT_TARGET_FLAGS): Set
20681 default endianness.
20682 * config/riscv/riscv.h (DEFAULT_ENDIAN_SPEC): New macro.
20684 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20686 * config/riscv/elf.h (LINK_SPEC): Pass linker endianness flag.
20687 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
20688 * config/riscv/linux.h (LINK_SPEC): Likewise.
20689 * config/riscv/riscv.h (ASM_SPEC): Pass -mbig-endian and
20691 (BYTES_BIG_ENDIAN): Handle big endian.
20692 (WORDS_BIG_ENDIAN): Define to BYTES_BIG_ENDIAN.
20693 * config/riscv/riscv.opt (-mbig-endian, -mlittle-endian): New
20695 * doc/invoke.texi (-mbig-endian, -mlittle-endian): Document.
20697 2021-03-23 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
20699 * regcprop.c (find_oldest_value_reg): Ask target whether
20700 different mode is fine for replacement register.
20702 2021-03-23 Aldy Hernandez <aldyh@redhat.com>
20704 PR tree-optimization/99296
20705 * value-range.cc (irange::irange_set_1bit_anti_range): New.
20706 (irange::irange_set_anti_range): Call irange_set_1bit_anti_range
20707 * value-range.h (irange::irange_set_1bit_anti_range): New.
20709 2021-03-22 Vladimir N. Makarov <vmakarov@redhat.com>
20712 * config/aarch64/constraints.md (UtQ): Use
20713 define_relaxed_memory_constraint for it.
20714 * doc/md.texi (define_relaxed_memory_constraint): Describe it.
20715 * genoutput.c (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
20716 * genpreds.c (constraint_data): Add bitfield is_relaxed_memory.
20717 (have_relaxed_memory_constraints): New static var.
20718 (relaxed_memory_start, relaxed_memory_end): Ditto.
20719 (add_constraint): Add arg is_relaxed_memory. Check name for
20720 relaxed memory. Set up is_relaxed_memory in constraint_data and
20721 have_relaxed_memory_constraints. Adjust calls.
20722 (choose_enum_order): Process relaxed memory.
20723 (write_tm_preds_h): Ditto.
20724 (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
20725 * gensupport.c (process_rtx): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
20726 * ira-costs.c (record_reg_classes): Process CT_RELAXED_MEMORY.
20727 * ira-lives.c (single_reg_class): Use
20728 insn_extra_relaxed_memory_constraint.
20729 * ira.c (ira_setup_alts): CT_RELAXED_MEMORY.
20730 * lra-constraints.c (valid_address_p): Use
20731 insn_extra_relaxed_memory_constraint instead of other memory
20733 (process_alt_operands): Process CT_RELAXED_MEMORY.
20734 (curr_insn_transform): Use insn_extra_relaxed_memory_constraint.
20735 * recog.c (asm_operand_ok, preprocess_constraints): Process
20737 * reload.c (find_reloads): Ditto.
20738 * rtl.def (DEFINE_RELAXED_MEMORY_CONSTRAINT): New.
20739 * stmt.c (parse_input_constraint): Use
20740 insn_extra_relaxed_memory_constraint.
20742 2021-03-22 Segher Boessenkool <segher@kernel.crashing.org>
20745 * ubsan.c (ubsan_instrument_float_cast): Don't test for unordered if
20748 2021-03-22 Alex Coplan <alex.coplan@arm.com>
20751 * config/arm/arm-protos.h (neon_make_constant): Add generate
20752 argument to guard emitting insns, default to true.
20753 * config/arm/arm.c (arm_legitimate_constant_p_1): Reject
20754 CONST_VECTORs which neon_make_constant can't handle.
20755 (neon_vdup_constant): Add generate argument, avoid emitting
20756 insns if it's not set.
20757 (neon_make_constant): Plumb new generate argument through.
20758 * config/arm/constraints.md (Ui): New. Use it...
20759 * config/arm/mve.md (*mve_mov<mode>): ... here.
20760 * config/arm/vec-common.md (movv8hf): Use neon_make_constant to
20761 synthesize constants.
20763 2021-03-22 Richard Biener <rguenther@suse.de>
20765 * debug.h: Add deprecation warning.
20767 2021-03-22 Richard Biener <rguenther@suse.de>
20769 PR tree-optimization/99694
20770 * tree-ssa-sccvn.c (visit_phi): Ignore edges with the
20773 2021-03-22 Kito Cheng <kito.cheng@sifive.com>
20776 * config/riscv/riscv.c (riscv_expand_block_move): Get RTL value
20777 after type checking.
20779 2021-03-22 Jakub Jelinek <jakub@redhat.com>
20783 * dwarf2out.c (get_full_len): Use get_precision rather than
20785 (add_const_value_attribute): Make sure add_AT_wide argument has
20786 precision prec rather than some very wide one.
20788 2021-03-22 Kewen Lin <linkw@linux.ibm.com>
20790 * config/rs6000/rs6000.md (*rotldi3_insert_sf,
20791 *mov<SFDF:mode><SFDF2:mode>cc_p9, floatsi<mode>2_lfiwax,
20792 floatsi<mode>2_lfiwax_mem, floatunssi<mode>2_lfiwzx,
20793 floatunssi<mode>2_lfiwzx_mem, *floatsidf2_internal,
20794 *floatunssidf2_internal, fix_trunc<mode>si2_stfiwx,
20795 fix_trunc<mode>si2_internal, fixuns_trunc<mode>si2_stfiwx,
20796 *round32<mode>2_fprs, *roundu32<mode>2_fprs,
20797 *fix_trunc<mode>si2_internal): Fix empty split condition.
20798 * config/rs6000/vsx.md (*vsx_le_undo_permute_<mode>,
20799 vsx_reduc_<VEC_reduc_name>_v2df, vsx_reduc_<VEC_reduc_name>_v4sf,
20800 *vsx_reduc_<VEC_reduc_name>_v2df_scalar,
20801 *vsx_reduc_<VEC_reduc_name>_v4sf_scalar): Likewise.
20803 2021-03-22 Xionghu Luo <luoxhu@linux.ibm.com>
20806 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
20807 Convert idx to DImode.
20808 (rs6000_expand_vector_set_var_p8): Likewise.
20810 2021-03-21 Jakub Jelinek <jakub@redhat.com>
20813 * dwarf2out.c (insert_float): Change return type from void to
20814 unsigned, handle GET_MODE_SIZE (mode) == 2 and return element size.
20815 (mem_loc_descriptor, loc_descriptor, add_const_value_attribute):
20818 2021-03-20 H.J. Lu <hjl.tools@gmail.com>
20821 * config/i386/i386.c (construct_container): Check cfun != NULL
20822 before accessing silent_p.
20824 2021-03-20 Ahamed Husni <ahamedhusni73@gmail.com>
20826 * asan.c: Fix typos in comments.
20828 2021-03-20 Vladimir N. Makarov <vmakarov@redhat.com>
20830 PR rtl-optimization/99680
20831 * lra-constraints.c (skip_contraint_modifiers): Rename to skip_constraint_modifiers.
20832 (process_address_1): Check empty constraint before using
20835 2021-03-19 Pat Haugen <pthaugen@linux.ibm.com>
20837 * config/rs6000/rs6000.c (power10_cost): New.
20838 (rs6000_option_override_internal): Set Power10 costs.
20839 (rs6000_issue_rate): Set Power10 issue rate.
20840 * config/rs6000/power10.md: Rewrite for Power10.
20842 2021-03-19 Vladimir N. Makarov <vmakarov@redhat.com>
20845 * lra-constraints.c (process_address_1): Don't use unknown
20846 constraint for address constraint.
20848 2021-03-19 Iain Sandoe <iain@sandoe.co.uk>
20851 * config.gcc (powerpc-*-darwin8): Delete the reference to
20852 the now removed darwin8.h.
20854 2021-03-19 Olivier Hainque <hainque@adacore.com>
20857 * config/vxworksae.h (VX_CPU_PREFIX): Define.
20859 2021-03-19 John David Anglin <danglin@gcc.gnu.org>
20861 * config/pa/pa.c (import_milli): Use memcpy instead of strncpy.
20863 2021-03-19 Tamar Christina <tamar.christina@arm.com>
20865 PR tree-optimization/99656
20866 * tree-vect-slp-patterns.c (linear_loads_p,
20867 complex_add_pattern::matches, is_eq_or_top,
20868 vect_validate_multiplication, complex_mul_pattern::matches,
20869 complex_fms_pattern::matches): Remove complex_perm_kinds_t.
20870 * tree-vectorizer.h: (complex_load_perm_t): Removed.
20871 (slp_tree_to_load_perm_map_t): Use complex_perm_kinds_t instead of
20872 complex_load_perm_t.
20874 2021-03-19 H.J. Lu <hjl.tools@gmail.com>
20877 * config/i386/i386-options.c (ix86_init_machine_status): Set
20879 * config/i386/i386.c (init_cumulative_args): Set silent_p to
20881 (construct_container): Return early for return and argument
20882 errors if silent_p is true.
20883 * config/i386/i386.h (machine_function): Add silent_p.
20885 2021-03-19 Jakub Jelinek <jakub@redhat.com>
20888 * config/arm/constraints.md (Ds): New constraint.
20889 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): Use w,Ds
20890 constraint instead of w,Dm.
20892 2021-03-19 Andrew Stubbs <ams@codesourcery.com>
20894 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Fix quotes
20897 2021-03-19 Eric Botcazou <ebotcazou@adacore.com>
20899 PR middle-end/99641
20900 * fold-const.c (native_encode_initializer) <CONSTRUCTOR>: For an
20901 array type, do the computation of the current position in sizetype.
20903 2021-03-18 Vladimir N. Makarov <vmakarov@redhat.com>
20906 * lra-constraints.c (process_address_1): Use lookup_constraint
20907 only for a single constraint.
20909 2021-03-18 Martin Sebor <msebor@redhat.com>
20911 PR middle-end/99502
20912 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): Rename...
20913 (inbounds_memaccess_p): ...to this. Check the ending offset of
20914 the accessed member.
20916 2021-03-18 Andrew Stubbs <ams@codesourcery.com>
20918 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Add %< and
20919 %> quote markers to error messages.
20920 (gcn_goacc_validate_dims): Likewise.
20921 (gcn_conditional_register_usage): Remove exclaimation mark from error
20923 (gcn_vectorize_vec_perm_const): Ensure perm is fully uninitialized.
20925 2021-03-18 Jan Hubicka <hubicka@ucw.cz>
20927 * config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
20930 2021-03-18 Sinan Lin <sinan@isrc.iscas.ac.cn>
20931 Kito Cheng <kito.cheng@sifive.com>
20933 * config/riscv/riscv.c (riscv_block_move_straight): Change type
20934 to unsigned HOST_WIDE_INT for parameter and local variable with
20935 HOST_WIDE_INT type.
20936 (riscv_adjust_block_mem): Ditto.
20937 (riscv_block_move_loop): Ditto.
20938 (riscv_expand_block_move): Ditto.
20940 2021-03-18 Nick Clifton <nickc@redhat.com>
20942 * config/v850/v850.c (construct_restore_jr): Increase static
20944 (construct_save_jarl): Likewise.
20945 * config/v850/v850.h (DWARF2_DEBUGGING_INFO): Define.
20947 2021-03-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20949 * config/aarch64/aarch64.c (aarch64_adjust_generic_arch_tuning): Define.
20950 (aarch64_override_options_internal): Use it.
20951 (generic_tunings): Add AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS to
20954 2021-03-17 Sandra Loosemore <sandra@codesourcery.com>
20956 * config/nios2/nios2.c (nios2_custom_check_insns): Clean up
20957 error message format issues.
20958 (nios2_option_override): Likewise.
20959 (nios2_expand_fpu_builtin): Likewise.
20960 (nios2_init_custom_builtins): Adjust to avoid bogus strncpy
20961 truncation warning.
20962 (nios2_expand_custom_builtin): More error message format fixes.
20963 (nios2_expand_rdwrctl_builtin): Likewise.
20964 (nios2_expand_rdprs_builtin): Likewise.
20965 (nios2_expand_eni_builtin): Likewise.
20966 (nios2_expand_builtin): Likewise.
20967 (nios2_register_custom_code): Likewise.
20968 (nios2_valid_target_attribute_rec): Likewise.
20969 (nios2_add_insn_asm): Fix uninitialized variable warning.
20971 2021-03-17 Jan Hubicka <jh@suse.cz>
20973 * config/i386/x86-tune-costs.h (struct processor_costs): Update costs
20974 of gather to match reality.
20975 * config/i386/x86-tune.def (X86_TUNE_USE_GATHER): Enable for znver3.
20977 2021-03-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20979 * config/aarch64/aarch64-builtins.c (aarch64_expand_rng_builtin): Use EQ
20980 to compare against CC_REG rather than NE.
20982 2021-03-17 H.J. Lu <hjl.tools@gmail.com>
20985 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support
20986 inline assembly statements.
20987 (ix86_print_operand): Update 'P' handling for -fno-plt.
20989 2021-03-17 Tamar Christina <tamar.christina@arm.com>
20992 * config/aarch64/aarch64.c
20993 (aarch64_simd_clone_compute_vecsize_and_simdlen): Remove unused var.
20995 2021-03-16 Segher Boessenkool <segher@kernel.crashing.org>
20998 * config/rs6000/predicates.md (branch_comparison_operator): Allow
20999 ordered and unordered for CCFPmode, if flag_finite_math_only.
21001 2021-03-16 Jakub Jelinek <jakub@redhat.com>
21004 * config/i386/i386-expand.c (ix86_split_lea_for_addr): Emit a MULT
21005 rather than ASHIFT.
21006 * config/i386/i386.md (mult by 1248 into ashift): New splitter.
21008 2021-03-16 Martin Liska <mliska@suse.cz>
21011 * optc-save-gen.awk: Add flag_ipa_ra to exceptions for
21012 cl_optimization_compare function.
21014 2021-03-16 Ilya Leoshkevich <iii@linux.ibm.com>
21016 * config/s390/s390.c (f_constraint_p): Treat "fv" constraints
21019 2021-03-16 Jakub Jelinek <jakub@redhat.com>
21022 * config/i386/i386.h (struct machine_function): Add
21023 has_explicit_vzeroupper bitfield.
21024 * config/i386/i386-expand.c (ix86_expand_builtin): Set
21025 cfun->machine->has_explicit_vzeroupper when expanding
21026 IX86_BUILTIN_VZEROUPPER.
21027 * config/i386/i386-features.c (rest_of_handle_insert_vzeroupper):
21028 Do the mode switching only when TARGET_VZEROUPPER, expensive
21029 optimizations turned on and not optimizing for size.
21030 (pass_insert_vzeroupper::gate): Enable even when
21031 cfun->machine->has_explicit_vzeroupper is set.
21033 2021-03-16 Jakub Jelinek <jakub@redhat.com>
21036 * config/aarch64/aarch64.c
21037 (aarch64_simd_clone_compute_vecsize_and_simdlen): If not a function
21038 definition, walk TYPE_ARG_TYPES list if non-NULL for argument types
21039 instead of DECL_ARGUMENTS. Ignore types for uniform arguments.
21041 2021-03-15 Richard Biener <rguenther@suse.de>
21043 PR tree-optimization/98834
21044 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle missing
21045 subsetting by truncating the access size.
21047 2021-03-15 Jan Hubicka <hubicka@ucw.cz>
21049 * config/i386/i386-options.c (processor_cost_table): Add znver3_cost.
21050 * config/i386/x86-tune-costs.h (znver3_cost): New gobal variable; copy
21053 2021-03-15 Martin Liska <mliska@suse.cz>
21055 * spellcheck.c: Add missing comma in initialization.
21057 2021-03-14 Uroš Bizjak <ubizjak@gmail.com>
21059 * config/i386/sse.md (*vec_extract<mode>): Merge alternative 0 with
21060 alternative 2 and alternative 1 with alternative 3 using
21061 YW register constraint.
21062 (*vec_extract<PEXTR_MODE12:mode>_zext): Merge alternatives
21063 using YW register constraint.
21064 (*vec_extractv16qi_zext): Ditto.
21065 (*vec_extractv4si): Merge alternatives 4 and 5
21066 using Yw register constraint.
21067 (*ssse3_palignr<mode>_perm): Use Yw instead of v for alternative 3.
21069 2021-03-13 Martin Sebor <msebor@redhat.com>
21071 PR tree-optimization/99489
21072 * builtins.c (gimple_call_alloc_size): Fail gracefully when argument
21073 is not a call statement.
21075 2021-03-13 Jakub Jelinek <jakub@redhat.com>
21077 PR tree-optimization/99544
21078 * match.pd (X + (X << C) -> X * (1 + (1 << C))): Don't simplify
21079 if for vector types multiplication can't be done in type's mode.
21081 2021-03-12 Eric Botcazou <ebotcazou@adacore.com>
21084 * config/sparc/constraints.md (w): Rename to...
21085 (W): ... this and ditch previous implementation.
21086 * config/sparc/sparc.md (*movdi_insn_sp64): Replace W with m.
21087 (*movdf_insn_sp64): Likewise.
21088 (*mov<VM64:mode>_insn_sp64): Likewise.
21089 * config/sparc/sync.md (*atomic_compare_and_swap<mode>_1): Replace
21091 (atomic_compare_and_swap_leon3_1): Likewise.
21092 (*atomic_compare_and_swapdi_v8plus): Likewise.
21093 * config/sparc/sparc.c (memory_ok_for_ldd): Remove useless test on
21094 architecture and add missing address validity check during LRA.
21096 2021-03-12 Tobias Burnus <tobias@codesourcery.com>
21099 * gimplify.c (omp_add_variable): Handle NULL_TREE as size
21100 occuring for assumed-size arrays in use_device_{ptr,addr}.
21102 2021-03-12 Jakub Jelinek <jakub@redhat.com>
21105 * config/i386/constraints.md (YW): New internal constraint.
21106 * config/i386/sse.md (v_Yw): Add V4TI, V2TI, V1TI and TI cases.
21107 (*<sse2_avx2>_<insn><mode>3<mask_name>,
21108 *<sse2_avx2>_uavg<mode>3<mask_name>, *abs<mode>2,
21109 *<s>mul<mode>3_highpart<mask_name>): Use <v_Yw> instead of v in
21111 (<sse2_avx2>_psadbw): Use YW instead of v in constraints.
21112 (*avx2_pmaddwd, *sse2_pmaddwd, *<code>v8hi3, *<code>v16qi3,
21113 avx2_pmaddubsw256, ssse3_pmaddubsw128): Merge last two alternatives
21114 into one, use Yw instead of former x,v.
21115 (ashr<mode>3, <insn><mode>3): Use <v_Yw> instead of x in constraints of
21116 the last alternative.
21117 (<sse2_avx2>_packsswb<mask_name>, <sse2_avx2>_packssdw<mask_name>,
21118 <sse2_avx2>_packuswb<mask_name>, <sse4_1_avx2>_packusdw<mask_name>,
21119 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, <ssse3_avx2>_palignr<mode>,
21120 <ssse3_avx2>_pshufb<mode>3<mask_name>): Merge last two alternatives
21121 into one, use <v_Yw> instead of former x,v.
21122 (avx2_interleave_highv32qi<mask_name>,
21123 vec_interleave_highv16qi<mask_name>): Use Yw instead of v in
21124 constraints. Add && <mask_avx512bw_condition> to condition.
21125 (avx2_interleave_lowv32qi<mask_name>,
21126 vec_interleave_lowv16qi<mask_name>,
21127 avx2_interleave_highv16hi<mask_name>,
21128 vec_interleave_highv8hi<mask_name>,
21129 avx2_interleave_lowv16hi<mask_name>, vec_interleave_lowv8hi<mask_name>,
21130 avx2_pshuflw_1<mask_name>, sse2_pshuflw_1<mask_name>,
21131 avx2_pshufhw_1<mask_name>, sse2_pshufhw_1<mask_name>,
21132 avx2_<code>v16qiv16hi2<mask_name>, sse4_1_<code>v8qiv8hi2<mask_name>,
21133 *sse4_1_<code>v8qiv8hi2<mask_name>_1, <sse2_avx2>_<insn><mode>3): Use
21134 Yw instead of v in constraints.
21135 * config/i386/mmx.md (Yv_Yw): New define_mode_attr.
21136 (*mmx_<insn><mode>3, mmx_ashr<mode>3, mmx_<insn><mode>3): Use <Yv_Yw>
21137 instead of Yv in constraints.
21138 (*mmx_<insn><mode>3, *mmx_mulv4hi3, *mmx_smulv4hi3_highpart,
21139 *mmx_umulv4hi3_highpart, *mmx_pmaddwd, *mmx_<code>v4hi3,
21140 *mmx_<code>v8qi3, mmx_pack<s_trunsuffix>swb, mmx_packssdw,
21141 mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
21142 *mmx_uavgv8qi3, *mmx_uavgv4hi3, mmx_psadbw): Use Yw instead of Yv in
21144 (*mmx_pinsrw, *mmx_pinsrb, *mmx_pextrw, *mmx_pextrw_zext, *mmx_pextrb,
21145 *mmx_pextrb_zext): Use YW instead of Yv in constraints.
21146 (*mmx_eq<mode>3, mmx_gt<mode>3): Use x instead of Yv in constraints.
21147 (mmx_andnot<mode>3, *mmx_<code><mode>3): Split last alternative into
21148 two, one with just x, another isa avx512vl with v.
21150 2021-03-12 Martin Liska <mliska@suse.cz>
21152 * doc/invoke.texi: Add missing param documentation.
21154 2021-03-11 David Malcolm <dmalcolm@redhat.com>
21157 * Makefile.in (ANALYZER_OBJS): Add analyzer/feasible-graph.o and
21158 analyzer/trimmed-graph.o.
21159 * doc/analyzer.texi (Analyzer Paths): Rewrite description of
21160 feasibility checking to reflect new implementation.
21161 * doc/invoke.texi (-fdump-analyzer-feasibility): Document new
21163 * shortest-paths.h (shortest_paths::get_shortest_distance): New.
21165 2021-03-11 David Malcolm <dmalcolm@redhat.com>
21167 * digraph.cc (selftest::test_shortest_paths): Update
21168 shortest_paths init for new param. Add test of
21169 SPS_TO_GIVEN_TARGET.
21170 * shortest-paths.h (enum shortest_path_sense): New.
21171 (shortest_paths::shortest_paths): Add "sense" param.
21172 Update for renamings. Generalize to use "sense" param.
21173 (shortest_paths::get_shortest_path): Rename param.
21174 (shortest_paths::m_sense): New field.
21175 (shortest_paths::m_prev): Rename...
21176 (shortest_paths::m_best_edge): ...to this.
21177 (shortest_paths::get_shortest_path): Update for renamings.
21178 Conditionalize flipping of path on sense of traversal.
21180 2021-03-11 David Malcolm <dmalcolm@redhat.com>
21182 * digraph.cc (selftest::test_shortest_paths): Add test coverage
21183 for paths from B and C.
21184 * shortest-paths.h (shortest_paths::shortest_paths): Handle
21185 unreachable nodes, rather than asserting.
21187 2021-03-11 David Edelsohn <dje.gcc@gmail.com>
21190 * config/rs6000/rs6000.c (rs6000_xcoff_file_start): Don't create
21191 xcoff_tbss_section_name.
21192 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .lcomm.
21193 * xcoffout.c (xcoff_tbss_section_name): Delete.
21194 * xcoffout.h (xcoff_tbss_section_name): Delete.
21196 2021-03-11 Richard Biener <rguenther@suse.de>
21198 PR tree-optimization/99523
21199 * tree-cfg.c (dump_function_to_file): Dump SSA names
21200 w/o identifier to the decls section as well, not only those
21201 without a VAR_DECL.
21203 2021-03-11 Jakub Jelinek <jakub@redhat.com>
21206 * ipa-icf-gimple.c (func_checker::compare_gimple_call): For internal
21207 function calls with lhs fail if the lhs don't have compatible types.
21209 2021-03-11 Hans-Peter Nilsson <hp@axis.com>
21211 * config/cris/cris.h (HARD_FRAME_POINTER_REGNUM): Define.
21212 Change FRAME_POINTER_REGNUM to correspond to a new faked
21213 register faked_fp, part of GENNONACR_REGS like faked_ap.
21214 (CRIS_FAKED_REGS_CONTENTS): New helper macro.
21215 (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS, CALL_USED_REGISTERS):
21216 (REG_ALLOC_ORDER, REG_CLASS_CONTENTS, REGNO_OK_FOR_BASE_P)
21217 (ELIMINABLE_REGS, REGISTER_NAMES): Adjust accordingly.
21218 * config/cris/cris.md (CRIS_FP_REGNUM): Renumber to new faked
21220 (CRIS_REAL_FP_REGNUM): New constant.
21221 * config/cris/cris.c (cris_reg_saved_in_regsave_area): Check
21222 for HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM.
21223 (cris_initial_elimination_offset): Handle elimination changes
21224 to HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM
21225 and add one from FRAME_POINTER_REGNUM to
21226 HARD_FRAME_POINTER_REGNUM.
21227 (cris_expand_prologue, cris_expand_epilogue): Emit code for
21228 hard_frame_pointer_rtx instead of frame_pointer_rtx.
21230 2021-03-10 David Edelsohn <dje.gcc@gmail.com>
21233 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Add check for DCmode.
21234 * config/rs6000/rs6000.c (rs6000_special_round_type_align): Same.
21236 2021-03-10 Vladimir N. Makarov <vmakarov@redhat.com>
21239 * lra-constraints.c (process_address_1): Don't check unknown
21240 constraint, use X for empty constraint.
21242 2021-03-10 Alex Coplan <alex.coplan@arm.com>
21244 * config/aarch64/aarch64.c (aarch64_vfp_is_call_or_return_candidate):
21245 Fix typo in comment describing "is_ha" argument.
21247 2021-03-10 John David Anglin <danglin@gcc.gnu.org>
21249 * doc/sourcebuild.texi: Document LRA target selector.
21251 2021-03-10 David Malcolm <dmalcolm@redhat.com>
21253 * doc/ux.texi: Add subsection contrasting interactive versus
21254 batch usage of GCC.
21256 2021-03-10 Joel Hutton <joel.hutton@arm.com>
21259 * tree-vect-stmts.c (vectorizable_store): Fix scatter store mask
21261 (vectorizable_load): Fix gather load mask check condition.
21263 2021-03-10 Richard Biener <rguenther@suse.de>
21265 PR tree-optimization/99510
21266 * tree.c (check_aligned_type): Check that the candidate
21267 has TYPE_USER_ALIGN set instead of matching with the
21270 2021-03-10 Eric Botcazou <ebotcazou@adacore.com>
21272 * config/sparc/sparc.c (sparc_regmode_natural_size): Return 4 for
21273 float and vector integer modes only if the mode is not larger.
21275 2021-03-10 Hans-Peter Nilsson <hp@axis.com>
21277 * config/cris/cris.h (DWARF_FRAME_REGISTERS): Define.
21279 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
21281 * ira.c (ira_setup_alts, ira_get_dup_out_num): Process digital
21283 * ira-lives.c (single_reg_class): Ditto.
21285 2021-03-09 Sebastian Huber <sebastian.huber@embedded-brains.de>
21287 * config.gcc (aarch64-*-rtems*): Include general rtems.h after
21288 the architecture-specific rtems.h.
21289 (aarch64-*-rtems*): Likewise.
21290 (arm*-*-rtems*): Likewise.
21291 (epiphany-*-rtems*): Likewise.
21292 (riscv*-*-rtems*): Likewise.
21294 2021-03-09 Jakub Jelinek <jakub@redhat.com>
21296 PR tree-optimization/99305
21297 * tree-ssa-phiopt.c (conditional_replacement): Test integer_pow2p
21298 before integer_all_onesp instead of vice versa.
21300 2021-03-09 Richard Earnshaw <rearnsha@arm.com>
21302 * common/config/arm/arm-common.c (arm_config_default): Change type
21303 of 'i' to unsigned.
21305 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
21308 * lra-constraints.c (process_address_1): Process constraint 'g'
21309 separately and digital constraints containing more one digit.
21311 2021-03-09 Nick Clifton <nickc@redhat.com>
21313 * config/rx/rx.h (DBX_DEBUGGING_INFO): Define.
21314 (DWARF"_DEBUGGING_INFO): Define.
21316 2021-03-09 Eric Botcazou <ebotcazou@adacore.com>
21319 * calls.c (initialize_argument_information): When the argument
21320 is passed by reference, do not make a copy in a thunk only if
21321 the argument is already in memory. Remove redundant test for
21322 the case of callee copy.
21324 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
21327 * lra-constraints.c (process_address_1): Process 0..9 constraints
21328 in process_address_1.
21330 2021-03-09 Andreas Krebbel <krebbel@linux.ibm.com>
21332 * config/s390/s390.c (struct s390_processor processor_table):
21333 Binutils name string must not be empty.
21335 2021-03-09 Claudiu Zissulescu <claziss@synopsys.com>
21337 * config/arc/arc.c (arc_attr_type): Remove function.
21339 2021-03-09 Martin Liska <mliska@suse.cz>
21342 * config/i386/i386-options.c (ix86_option_override_internal):
21343 Set isa_flags for OPTS argument and not for the global
21346 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
21348 * config/rs6000/predicates.md (ds_form_mem_operand): Check
21351 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
21354 * config/rs6000/predicates.md (ds_form_mem_operand) New
21356 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10) Use
21357 ds_form_mem_operand in ld/lwa patterns.
21358 * config/rs6000/fusion.md: Regenerate file.
21360 2021-03-08 Martin Sebor <msebor@redhat.com>
21362 PR middle-end/98266
21363 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): New function.
21364 (array_bounds_checker::check_array_bounds): Call it.
21366 2021-03-08 Martin Sebor <msebor@redhat.com>
21368 PR middle-end/97631
21369 * tree-ssa-strlen.c (maybe_warn_overflow): Test rawmem.
21370 (handle_builtin_stxncpy_strncat): Rename locals. Determine
21371 destination size from allocation calls. Issue a more appropriate
21373 (handle_builtin_memcpy): Pass true as rawmem to maybe_warn_overflow.
21374 (handle_builtin_memset): Same.
21376 2021-03-08 Peter Bergner <bergner@linux.ibm.com>
21379 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_permute): Add an assert
21380 to ensure we do not have an Altivec style address.
21381 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): Disable if passed
21382 an Altivec style address.
21383 (*vsx_le_perm_store_<mode>): Likewise.
21384 (splitters after *vsx_le_perm_store_<mode>): Likewise.
21385 (vsx_load_<mode>): Disable special expander if passed an Altivec
21387 (vsx_store_<mode>): Likewise.
21389 2021-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21392 * config/aarch64/predicates.md (aarch64_simd_shift_imm_vec_qi): Define.
21393 (aarch64_simd_shift_imm_vec_hi): Likewise.
21394 (aarch64_simd_shift_imm_vec_si): Likewise.
21395 (aarch64_simd_shift_imm_vec_di): Likewise.
21396 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le): Use
21397 predicate from above.
21398 (aarch64_shrn<mode>_insn_be): Likewise.
21399 (aarch64_rshrn<mode>_insn_le): Likewise.
21400 (aarch64_rshrn<mode>_insn_be): Likewise.
21401 (aarch64_shrn2<mode>_insn_le): Likewise.
21402 (aarch64_shrn2<mode>_insn_be): Likewise.
21403 (aarch64_rshrn2<mode>_insn_le): Likewise.
21404 (aarch64_rshrn2<mode>_insn_be): Likewise.
21406 2021-03-08 Vladimir N. Makarov <vmakarov@redhat.com>
21409 * lra-constraints.c (skip_contraint_modifiers): New function.
21410 (process_address_1): Use it before lookup_constraint call.
21412 2021-03-08 Martin Liska <mliska@suse.cz>
21415 * config/i386/i386-options.c (ix86_option_override_internal):
21416 Enable UINTR and HRESET for -march that supports it.
21418 2021-03-08 Ilya Leoshkevich <iii@linux.ibm.com>
21420 * config/s390/s390.c (f_constraint_p): New function.
21421 (s390_md_asm_adjust): Implement TARGET_MD_ASM_ADJUST.
21422 (TARGET_MD_ASM_ADJUST): Likewise.
21424 2021-03-08 Tobias Burnus <tobias@codesourcery.com>
21427 * tree-nested.c (convert_local_reference_stmt): Avoid calling
21428 lookup_field_for_decl for Fortran module (= namespace context).
21430 2021-03-08 Andreas Krebbel <krebbel@linux.ibm.com>
21432 * config/s390/s390.c (s390_expand_vec_compare): Implement <0
21433 comparison with arithmetic right shift.
21434 (s390_expand_vcond): No need for a force_reg anymore.
21435 s390_vec_compare will do it.
21436 * config/s390/vector.md ("vec_cmp<mode><tointvec>"): Accept also
21437 immediate operands.
21439 2021-03-07 Jakub Jelinek <jakub@redhat.com>
21442 * config/i386/constraints.md (Yw): Use SSE_REGS if TARGET_SSE
21443 but TARGET_AVX512BW or TARGET_AVX512VL is not set. Adjust description
21445 * config/i386/sse.md (v_Yw): New define_mode_attr.
21446 (*<insn><mode>3, *mul<mode>3<mask_name>, *avx2_<code><mode>3,
21447 *sse4_1_<code><mode>3<mask_name>): Use <v_Yw> instead of v
21449 * config/i386/mmx.md (mmx_pshufw_1, *vec_dupv4hi): Use Yw instead of
21450 xYw in constraints.
21452 2021-03-06 Julian Brown <julian@codesourcery.com>
21454 * tree-pretty-print.c (dump_generic_node): Emit non-generic
21455 address space info for aggregates.
21457 2021-03-06 Hans-Peter Nilsson <hp@axis.com>
21459 * config/cris/cris.h (MAX_FIXED_MODE_SIZE): Don't define.
21461 2021-03-05 Jakub Jelinek <jakub@redhat.com>
21463 PR middle-end/99322
21464 * tree-cfg.c (bb_to_omp_idx): New variable.
21465 (execute_build_cfg): Release the bb_to_omp_idx vector after
21466 cleanup_tree_cfg returns.
21467 (handle_abnormal_edges): Remove bb_to_omp_idx argument, adjust
21468 for bb_to_omp_idx being a vec<int> instead of pointer to array
21470 (make_edges): Remove bb_to_omp_idx local variable, don't pass
21471 it to handle_abnormal_edges, adjust for bb_to_omp_idx being a
21472 vec<int> instead of pointer to array of ints and don't free/release
21474 (remove_bb): When removing a bb and placing forced label somewhere
21475 else, ensure it is put into the same OpenMP region during cfg
21476 pass if possible or to entry successor as fallback. Unregister
21477 bb from bb_to_omp_idx.
21479 2021-03-05 Vladimir N. Makarov <vmakarov@redhat.com>
21482 * lra-constraints.c (process_address_1): Skip decomposing address
21483 for asm insn operand with unknown constraint.
21485 2021-03-05 Martin Jambor <mjambor@suse.cz>
21488 * cgraph.c (cgraph_edge::set_call_stmt): Do not update all
21489 corresponding speculative edges if we are about to resolve
21490 sepculation. Make edge direct (and so resolve speculations) before
21491 removing it from call_site_hash.
21492 (cgraph_edge::make_direct): Relax the initial assert to allow calling
21493 the function on speculative direct edges.
21495 2021-03-05 Eric Botcazou <ebotcazou@adacore.com>
21497 PR rtl-optimization/99376
21498 * rtlanal.c (nonzero_bits1) <arithmetic operators>: If the number
21499 of low-order zero bits is too large, set the result to 0 directly.
21501 2021-03-04 Jakub Jelinek <jakub@redhat.com>
21503 PR middle-end/93235
21504 * expmed.c (store_bit_field_using_insv): Return false of xop0 is a
21505 SUBREG and a SUBREG to op_mode can't be created.
21507 2021-03-04 Alex Coplan <alex.coplan@arm.com>
21510 * config/aarch64/aarch64-sve-builtins.cc
21511 (function_resolver::require_vector_type): Handle error_mark_node.
21513 2021-03-04 Ilya Leoshkevich <iii@linux.ibm.com>
21515 * cfgexpand.c (expand_asm_loc): Pass new parameter.
21516 (expand_asm_stmt): Likewise.
21517 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Add new
21519 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
21520 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
21521 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
21522 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
21523 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
21524 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
21525 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
21526 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
21527 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
21528 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
21529 * doc/tm.texi (md_asm_adjust): Likewise.
21530 * target.def (md_asm_adjust): Likewise.
21532 2021-03-04 Richard Biener <rguenther@suse.de>
21534 PR middle-end/97855
21535 * tree-pretty-print.c: Poison pp_printf.
21536 (dump_decl_name): Avoid use of pp_printf.
21537 (dump_block_node): Likewise.
21538 (dump_generic_node): Likewise.
21540 2021-03-04 Martin Sebor <msebor@redhat.com>
21542 PR middle-end/96963
21543 PR middle-end/94655
21544 * builtins.c (handle_array_ref): New helper.
21545 (handle_mem_ref): New helper.
21546 (compute_objsize_r): Factor out ARRAY_REF and MEM_REF handling
21547 into new helper functions. Correct a workaround for vectorized
21550 2021-03-03 Pat Haugen <pthaugen@linux.ibm.com>
21552 * config/rs6000/dfp.md (extendddtd2, trunctddd2, *cmp<mode>_internal1,
21553 floatditd2, ftrunc<mode>2, fix<mode>di2, dfp_ddedpd_<mode>,
21554 dfp_denbcd_<mode>, dfp_dxex_<mode>, dfp_diex_<mode>,
21555 *dfp_sgnfcnc_<mode>, dfp_dscli_<mode>, dfp_dscri_<mode>): Update size
21556 attribute for Power10.
21557 * config/rs6000/mma.md (*movoo): Likewise.
21558 * config/rs6000/rs6000.md (define_attr "size"): Add 256.
21559 (define_mode_attr bits): Add DD/TD modes.
21560 * config/rs6000/sync.md (load_quadpti, store_quadpti, load_lockedpti,
21561 store_conditionalpti): Update size attribute for Power10.
21563 2021-03-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21566 * config/sparc/t-sparc (tree-ssanames.o-warn): Don't error for
21567 -Wuninitialized, -Wmaybe-uninitialized.
21568 (wide-int.o-warn): Likewise.
21570 2021-03-03 Richard Earnshaw <rearnsha@arm.com>
21572 * common/config/arm/arm-common.c: Include configargs.h.
21573 (arm_config_default): New function.
21574 (arm_target_mode): Renamed from arm_target_thumb_only. Handle
21575 processors that do not support Thumb. Take into account the
21576 --with-mode configuration setting for selecting the default.
21577 * config/arm/arm.h (OPTION_DEFAULT_SPECS): Remove entry for 'mode'.
21578 (TARGET_MODE_SPEC_FUNCTIONS): Update for function name change.
21580 2021-03-03 Martin Liska <mliska@suse.cz>
21582 PR gcov-profile/97461
21583 * gcov-io.h (GCOV_PREALLOCATED_KVP): Remove.
21585 2021-03-03 Eric Botcazou <ebotcazou@adacore.com>
21588 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
21589 point back the hard frame pointer to its default location when the
21590 frame is larger than SEH_MAX_FRAME_SIZE.
21592 2021-03-03 Jakub Jelinek <jakub@redhat.com>
21595 * config/i386/predicates.md (logic_operator): New define_predicate.
21596 * config/i386/i386.md (mov + mem using comm arith peephole2):
21597 Punt if operands[1] is EXT_REX_SSE_REGNO_P, AVX512BW is not enabled
21598 and the inner mode is [QH]Imode.
21600 2021-03-03 Jakub Jelinek <jakub@redhat.com>
21603 * dwarf2out.c (dw_loc_list_struct): Add end_entry member.
21604 (new_loc_list): Clear end_entry.
21605 (output_loc_list): Only use DW_LLE_startx_length for -gsplit-dwarf
21606 if HAVE_AS_LEB128, otherwise use DW_LLE_startx_endx. Fix comment
21608 (index_location_lists): For dwarf_version >= 5 without HAVE_AS_LEB128,
21609 initialize also end_entry.
21611 2021-03-03 Jakub Jelinek <jakub@redhat.com>
21614 * cfgrtl.c (fixup_partitions): When changing some bbs from hot to cold
21615 partitions, if in non-layout mode after reorder_blocks also move
21616 affected blocks to ensure a single partition transition.
21618 2021-03-03 Jason Merrill <jason@redhat.com>
21621 * cgraphunit.c (process_function_and_variable_attributes): Don't
21622 warn about flatten on an alias if the target also has it.
21623 * cgraph.h (symtab_node::get_alias_target_tree): New.
21625 2021-03-02 David Edelsohn <dje.gcc@gmail.com>
21627 * config/rs6000/rs6000.md (tls_get_tpointer_internal): Prepend
21628 period to symbol name.
21629 (tls_get_addr_internal<mode>): Same.
21631 2021-03-02 David Malcolm <dmalcolm@redhat.com>
21634 * diagnostic-show-locus.c
21635 (selftest::test_one_liner_many_fixits_2): Fix accidental usage of
21638 2021-03-02 Martin Sebor <msebor@redhat.com>
21640 PR middle-end/99276
21641 * builtins.c (warn_for_access): Remove stray warning text.
21643 2021-03-02 Martin Sebor <msebor@redhat.com>
21645 PR middle-end/99295
21646 * doc/extend.texi (attribute malloc): Reword and clarify nonaliasing
21649 2021-03-02 Jakub Jelinek <jakub@redhat.com>
21652 * dwarf2out.c (output_macinfo_op): Use DW_MACRO_*_str* even with
21653 -gdwarf-5 -gstrict-dwarf. For -gsplit-dwarf -gdwarf-5 use
21654 DW_MACRO_*_strx instead of DW_MACRO_*_strp. Handle
21655 DW_MACRO_define_strx and DW_MACRO_undef_strx.
21656 (save_macinfo_strings): Use DW_MACRO_*_str* even with
21657 -gdwarf-5 -gstrict-dwarf. Handle DW_MACRO_define_strx and
21658 DW_MACRO_undef_strx.
21660 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
21662 * config/s390/s390-builtin-types.def (BT_FN_V4SF_V8HI_UINT): New
21664 (BT_FN_V8HI_V8HI_UINT): Likewise.
21665 (BT_FN_V8HI_V4SF_V4SF_UINT): Likewise.
21666 * config/s390/s390-builtins.def (B_NNPA): New macro definition.
21667 (s390_vclfnhs, s390_vclfnls, s390_vcrnfs, s390_vcfn, s390_vcnf):
21668 New builtin definitions.
21669 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Bump
21670 vector extension version.
21671 * config/s390/s390.c (s390_expand_builtin): Check if builtins are
21672 available with current -march level.
21673 * config/s390/s390.md (UNSPEC_NNPA_VCLFNHS_V8HI)
21674 (UNSPEC_NNPA_VCLFNLS_V8HI, UNSPEC_NNPA_VCRNFS_V8HI)
21675 (UNSPEC_NNPA_VCFN_V8HI, UNSPEC_NNPA_VCNF_V8HI): New constants.
21676 * config/s390/vecintrin.h (vec_extend_to_fp32_hi): New macro.
21677 (vec_extend_to_fp32_lo): Likewise.
21678 (vec_round_from_fp32): Likewise.
21679 (vec_convert_to_fp16): Likewise.
21680 (vec_convert_from_fp16): Likewise.
21681 * config/s390/vx-builtins.md (vclfnhs_v8hi): New insn pattern.
21682 (vclfnls_v8hi): Likewise.
21683 (vcrnfs_v8hi): Likewise.
21684 (vcfn_v8hi): Likewise.
21685 (vcnf_v8hi): Likewise.
21687 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
21689 * common/config/s390/s390-common.c (processor_flags_table): New entry.
21690 * config.gcc: Enable arch14 for --with-arch and --with-tune.
21691 * config/s390/driver-native.c (s390_host_detect_local_cpu): Pick
21692 arch14 for unknown CPU models.
21693 * config/s390/s390-opts.h (enum processor_type): Add PROCESSOR_ARCH14.
21694 * config/s390/s390.c (s390_issue_rate): Add case for PROCESSOR_ARCH14.
21695 (s390_get_sched_attrmask): Likewise.
21696 (s390_get_unit_mask): Likewise.
21697 * config/s390/s390.h (enum processor_flags): Add PF_NNPA and PF_ARCH14.
21698 (TARGET_CPU_ARCH14, TARGET_CPU_ARCH14_P, TARGET_CPU_NNPA)
21699 (TARGET_CPU_NNPA_P, TARGET_ARCH14, TARGET_ARCH14_P, TARGET_NNPA)
21700 (TARGET_NNPA_P): New macro definitions.
21701 * config/s390/s390.md ("cpu_facility", "enabled"): Add arch14 and nnpa.
21702 * config/s390/s390.opt: Add PROCESSOR_ARCH14.
21704 2021-03-02 Jakub Jelinek <jakub@redhat.com>
21706 PR middle-end/95757
21707 * tree-vrp.c (register_edge_assert_for): Remove superfluous ()s around
21708 condition. Call register_edge_assert_for_1 for == 0, != 0, == 1 and
21709 != 1 comparisons if name is lhs of a comparison.
21711 2021-03-01 Iain Sandoe <iain@sandoe.co.uk>
21715 * config/darwin-protos.h (darwin_should_restore_cfa_state): New.
21716 * config/darwin.c (darwin_should_restore_cfa_state): New.
21717 * config/darwin.h (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New.
21718 * doc/tm.texi: Regenerated.
21719 * doc/tm.texi.in: Document TARGET_ASM_SHOULD_RESTORE_CFA_STATE.
21720 * dwarf2cfi.c (connect_traces): If the target requests, restore
21721 the CFA expression after a DW_CFA_restore.
21722 * target.def (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New hook.
21724 2021-03-01 Martin Liska <mliska@suse.cz>
21727 * optc-save-gen.awk: Add 4 more exceptions.
21729 2021-03-01 Nathan Sidwell <nathan@acm.org>
21732 * tree.h (TYPE_ALIGN_RAW): New accessor.
21733 (TYPE_ALIGN): Use it.
21735 2021-03-01 Jan Hubicka <jh@suse.cz>
21738 * ipa-fnsummary.c (compute_fn_summary): Fix sanity check.
21740 2021-03-01 Eric Botcazou <ebotcazou@adacore.com>
21743 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
21744 point the hard frame pointer to the SSE register save area instead
21745 of the general register save area. Perform only minimal adjustment
21746 for small frames if it is initially not correctly aligned.
21747 (ix86_expand_prologue): Remove early saves for a SEH target.
21748 * config/i386/winnt.c (struct seh_frame_state): Document constraint.
21750 2021-02-28 Jakub Jelinek <jakub@redhat.com>
21753 * ipa.c (symbol_table::remove_unreachable_nodes): Fix a comment
21754 typo - referneced -> referenced.
21755 * tree.c (component_ref_size): Fix comment typo -
21756 refernce -> reference.
21757 * tree-ssa-alias.c (access_path_may_continue_p): Fix comment typo -
21758 traling -> trailing.
21759 (aliasing_component_refs_p): Fix comment typos -
21760 refernce -> reference and refernece -> reference and
21761 traling -> trailing.
21762 (nonoverlapping_refs_since_match_p): Fix comment typo -
21763 referneces -> references.
21764 * doc/invoke.texi (--param modref-max-bases): Fix a typo -
21765 referneces -> references.
21767 2021-02-27 Iain Sandoe <iain@sandoe.co.uk>
21769 * config/host-darwin.c (darwin_gt_pch_use_address): Modify
21770 diagnostic message to avoid use of a contraction and format
21773 2021-02-27 Jakub Jelinek <jakub@redhat.com>
21776 * gcse.c (gcse_or_cprop_is_too_expensive): Use %wu instead of
21777 HOST_WIDE_INT_PRINT_UNSIGNED in warning format string.
21778 * ipa-devirt.c (ipa_odr_read_section): Use %wd instead of
21779 HOST_WIDE_INT_PRINT_DEC in inform format string. Fix comment
21782 2021-02-26 Richard Biener <rguenther@suse.de>
21784 PR middle-end/99281
21785 * expr.c (store_field): For calls with return-slot optimization
21786 and addressable return type expand the store directly.
21788 2021-02-26 Richard Biener <rguenther@suse.de>
21791 * builtins.c (warn_string_no_nul): Fix diagnostic formatting.
21793 2021-02-26 Peter Bergner <bergner@linux.ibm.com>
21796 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Replace assert
21799 2021-02-26 Aaron Sawdey <acsawdey@linux.ibm.com>
21801 * config.gcc: Add rs6000-pcrel-opt.o.
21802 * config/rs6000/rs6000-pcrel-opt.c: New file.
21803 * config/rs6000/pcrel-opt.md: New file.
21804 * config/rs6000/predicates.md: Add d_form_memory predicate.
21805 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_PCREL_OPT.
21806 * config/rs6000/rs6000-passes.def: Add pass_pcrel_opt.
21807 * config/rs6000/rs6000-protos.h: Add reg_to_non_prefixed(),
21808 pcrel_opt_valid_mem_p(), output_pcrel_opt_reloc(),
21809 and make_pass_pcrel_opt().
21810 * config/rs6000/rs6000.c (reg_to_non_prefixed): Make global.
21811 (rs6000_option_override_internal): Add pcrel-opt.
21812 (rs6000_delegitimize_address): Support pcrel-opt.
21813 (rs6000_opt_masks): Add pcrel-opt.
21814 (pcrel_opt_valid_mem_p): New function.
21815 (reg_to_non_prefixed): Make global.
21816 (rs6000_asm_output_opcode): Reset prepend_p_to_next_insn.
21817 (output_pcrel_opt_reloc): New function.
21818 * config/rs6000/rs6000.md (loads_extern_addr): New attr.
21819 (pcrel_extern_addr): Set loads_extern_addr.
21820 Add include for pcrel-opt.md.
21821 * config/rs6000/rs6000.opt: Add -mpcrel-opt.
21822 * config/rs6000/t-rs6000: Add rules for pcrel-opt.c and
21825 2021-02-26 YunQiang Su <yunqiang.su@cipunited.com>
21828 * config/mips/mips.c (mips_expand_ext_as_unaligned_load):
21829 If TARGET_64BIT and dest is SUBREG, we check the width, if it
21830 equal to SImode, we use SImode operation, just like what we are
21833 2021-02-26 Marek Polacek <polacek@redhat.com>
21835 * builtins.c (warn_for_access): Fix typos.
21837 2021-02-25 Iain Sandoe <iain@sandoe.co.uk>
21839 * config/aarch64/aarch64.md (<optab>_rol<mode>3): Add a '#'
21840 mark in front of the immediate quantity.
21841 (<optab>_rolsi3_uxtw): Likewise.
21843 2021-02-25 Richard Earnshaw <rearnsha@arm.com>
21846 * config/arm/thumb2.md (nonsecure_call_reg_thumb2_fpcxt): New pattern.
21847 (nonsecure_call_value_reg_thumb2_fpcxt): Likewise.
21848 (nonsecure_call_reg_thumb2): Restrict to using r4 for the callee
21849 address and disable when the FPCXT is not available.
21850 (nonsecure_call_value_reg_thumb2): Likewise.
21852 2021-02-25 Nathan Sidwell <nathan@acm.org>
21855 * doc/invoke.texi (flang-info-module-cmi): Renamed option.
21857 2021-02-25 Tamar Christina <tamar.christina@arm.com>
21859 * tree-vect-slp.c (optimize_load_redistribution_1): Abort on NULL nodes.
21861 2021-02-25 Richard Biener <rguenther@suse.de>
21863 PR tree-optimization/99253
21864 * tree-vect-loop.c (check_reduction_path): First compute
21865 code, then verify out-of-loop uses.
21867 2021-02-25 Jakub Jelinek <jakub@redhat.com>
21870 * match.pd ((T)(A) + CST -> (T)(A + CST)): Add :s to convert.
21872 2021-02-25 Jakub Jelinek <jakub@redhat.com>
21874 PR tree-optimization/80635
21875 * tree-vrp.c (vrp_simplify_cond_using_ranges): Also handle
21876 VIEW_CONVERT_EXPR if modes are the same, innerop is integral and
21877 has mode precision.
21879 2021-02-25 Richard Biener <rguenther@suse.de>
21881 * tree-vect-slp.c (optimize_load_redistribution_1): Delay
21882 load_map population.
21883 (vect_match_slp_patterns_2): Revert part of last change.
21884 (vect_analyze_slp): Do not interleave optimize_load_redistribution
21885 with pattern detection but do it afterwards. Dump the
21886 whole SLP graph after pattern recognition and load
21887 redistribution optimization finished.
21889 2021-02-24 Jakub Jelinek <jakub@redhat.com>
21892 * omp-low.c (struct omp_context): Add teams_nested_p and
21893 nonteams_nested_p members.
21894 (scan_omp_target): Diagnose teams nested inside of target with other
21895 directives strictly nested inside of the same target.
21896 (check_omp_nesting_restrictions): Set ctx->teams_nested_p or
21897 ctx->nonteams_nested_p as needed.
21899 2021-02-24 Vladimir N. Makarov <vmakarov@redhat.com>
21901 PR inline-asm/99123
21902 * lra-constraints.c (uses_hard_regs_p): Don't use decompose_mem_address.
21904 2021-02-24 Hans-Peter Nilsson <hp@axis.com>
21906 * config/cris/cris.c (cris_expand_prologue): Set
21907 current_function_static_stack_size, if flag_stack_usage_info.
21909 2021-02-24 Pat Haugen <pthaugen@linux.ibm.com>
21911 * config/rs6000/rs6000.c (next_insn_prefixed_p): Rename.
21912 (rs6000_final_prescan_insn): Adjust.
21913 (rs6000_asm_output_opcode): Likewise.
21915 2021-02-24 Martin Sebor <msebor@redhat.com>
21917 PR middle-end/97172
21918 * attribs.c (attr_access::free_lang_data): Clear attribute arg spec
21919 from function arguments.
21921 2021-02-24 Tamar Christina <tamar.christina@arm.com>
21923 PR tree-optimization/99220
21924 * tree-vect-slp.c (optimize_load_redistribution_1): Remove
21925 node from cache when it's about to be deleted.
21927 2021-02-24 Jakub Jelinek <jakub@redhat.com>
21929 PR tree-optimization/99225
21930 * fold-const.c (fold_binary_loc) <case NE_EXPR>: In (x & (1 << y)) != 0
21931 to ((x >> y) & 1) != 0 simplifications use build_one_cst instead of
21932 build_int_cst (..., 1). Formatting fixes.
21934 2021-02-24 Tamar Christina <tamar.christina@arm.com>
21936 PR tree-optimization/99149
21937 * tree-vect-slp-patterns.c (vect_detect_pair_op): Don't recreate the
21939 (vect_slp_reset_pattern): Remove.
21940 (complex_fma_pattern::matches): Remove call to vect_slp_reset_pattern.
21941 (complex_mul_pattern::build, complex_fma_pattern::build,
21942 complex_fms_pattern::build): Fix ref counts.
21943 * tree-vect-slp.c (vect_free_slp_tree): Undo SLP only pattern relevancy
21944 when node is being deleted.
21945 (vect_match_slp_patterns_2): Correct result of cache hit on patterns.
21946 (vect_schedule_slp): Invalidate SLP_TREE_REPRESENTATIVE of removed
21948 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize value.
21950 2021-02-24 Matthias Klose <doko@ubuntu.com>
21953 2020-12-07 Matthias Klose <doko@ubuntu.com>
21955 * genextract.c (print_header): Undefine ENABLE_RTL_CHECKING
21956 and ENABLE_RTL_FLAG_CHECKING.
21958 2021-02-24 Richard Biener <rguenther@suse.de>
21961 * builtins.c (fold_builtin_next_arg): Avoid NULL arg.
21963 2021-02-23 Peter Bergner <bergner@linux.ibm.com>
21965 * config/rs6000/mma.md (mma_assemble_pair): Rename from this...
21966 (vsx_assemble_pair): ...to this.
21967 (*mma_assemble_pair): Rename from this...
21968 (*vsx_assemble_pair): ...to this.
21969 (mma_disassemble_pair): Rename from this...
21970 (vsx_disassemble_pair): ...to this.
21971 (*mma_disassemble_pair): Rename from this...
21972 (*vsx_disassemble_pair): ...to this.
21973 * config/rs6000/rs6000-builtin.def (BU_MMA_V2, BU_MMA_V3,
21974 BU_COMPAT): New macros.
21975 (mma_assemble_pair): Rename from this...
21976 (vsx_assemble_pair): ...to this.
21977 (mma_disassemble_pair): Rename from this...
21978 (vsx_disassemble_pair): ...to this.
21979 (mma_assemble_pair): New compatibility built-in.
21980 (mma_disassemble_pair): Likewise.
21981 * config/rs6000/rs6000-call.c (struct builtin_compatibility): New.
21982 (RS6000_BUILTIN_COMPAT): Define.
21983 (bdesc_compat): New.
21984 (mma_expand_builtin): Use VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
21985 (rs6000_gimple_fold_mma_builtin): Use MMA_BUILTIN_DISASSEMBLE_PAIR
21986 and VSX_BUILTIN_ASSEMBLE_PAIR.
21987 (rs6000_init_builtins): Register compatibility built-ins.
21988 (mma_init_builtins): Use VSX_BUILTIN_ASSEMBLE_PAIR,
21989 VSX_BUILTIN_ASSEMBLE_PAIR_INTERNAL, VSX_BUILTIN_DISASSEMBLE_PAIR and
21990 VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
21991 * doc/extend.texi (__builtin_mma_assemble_pair): Rename from this...
21992 (__builtin_vsx_assemble_pair): ...to this.
21993 (__builtin_mma_disassemble_pair): Rename from this...
21994 (__builtin_vsx_disassemble_pair): ...to this.
21996 2021-02-23 Martin Liska <mliska@suse.cz>
21999 * ipa-icf.c (sem_variable::merge): Do not merge 2 variables
22000 with different alignment. That leads to an invalid red zone
22001 size allocated in runtime.
22003 2021-02-23 Jakub Jelinek <jakub@redhat.com>
22005 PR tree-optimization/99204
22006 * fold-const.c (fold_read_from_constant_string): Check that
22007 tree_fits_uhwi_p (index) rather than just that index is INTEGER_CST.
22009 2021-02-23 Segher Boessenkool <segher@kernel.crashing.org>
22010 Kewen Lin <linkw@gcc.gnu.org>
22012 * config/rs6000/rs6000.md (*rotl<mode>3_insert_3): Renamed to...
22013 (rotl<mode>3_insert_3): ...this.
22014 (plus_ior_xor): New code_iterator.
22015 (define_split for GPR rl*imi): New splitter.
22016 * config/rs6000/vsx.md (vsx_init_v4si): Use gen_rotldi3_insert_3
22017 for integer merging.
22019 2021-02-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22021 * config/aarch64/aarch64-tuning-flags.def (cse_sve_vl_constants):
22023 * config/aarch64/aarch64.md (add<mode>3): Force CONST_POLY_INT immediates
22024 into a register when the above is enabled.
22025 * config/aarch64/aarch64.c (neoversev1_tunings):
22026 AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
22027 (aarch64_rtx_costs): Use AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
22029 2021-02-22 Hans-Peter Nilsson <hp@axis.com>
22031 * config/cris/cris.c (cris_print_operand) <'T'>: Change
22032 valid operand from is now an addi mult-value to shift-value.
22033 * config/cris/cris.md (*addi): Change expression of scaled
22034 operand from mult to ashift.
22035 * config/cris/cris.md (*addi_reload): New insn_and_split.
22037 2021-02-22 John David Anglin <danglin@gcc.gnu.org>
22040 * config/pa/pa.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Define as
22041 hook_bool_const_tree_hwi_hwi_const_tree_true.
22042 (pa_asm_output_mi_thunk): Add support for nonzero vcall_offset.
22044 2021-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
22046 PR rtl-optimization/98791
22047 * ira-conflicts.c (process_regs_for_copy): Don't create allocno copies
22048 for unordered modes.
22050 2021-02-22 Martin Liska <mliska@suse.cz>
22052 * tree-inline.c (inline_forbidden_p): Set
22053 inline_forbidden_reason.
22055 2021-02-22 Richard Biener <rguenther@suse.de>
22057 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Dump
22060 2021-02-22 Richard Biener <rguenther@suse.de>
22062 PR tree-optimization/99165
22063 * gimple-ssa-store-merging.c (pass_store_merging::process_store):
22064 Accumulate changed to ret.
22066 2021-02-21 Uros Bizjak <ubizjak@gmail.com>
22069 2020-12-09 Uroš Bizjak <ubizjak@gmail.com>
22071 * config/i386/i386.h (REG_ALLOC_ORDER): Remove
22073 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
22076 * config/s390/vector.md (trunctf<DFP_ALL:mode>2_vr): New
22078 (trunctf<DFP_ALL:mode>2): Likewise.
22079 (trunctdtf2_vr): Likewise.
22080 (trunctdtf2): Likewise.
22081 (extend<DFP_ALL:mode>tf2_vr): Likewise.
22082 (extend<DFP_ALL:mode>tf2): Likewise.
22083 (extendtftd2_vr): Likewise.
22084 (extendtftd2): Likewise.
22086 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
22088 * config/s390/vector.md (*fprx2_to_tf): Rename to fprx2_to_tf,
22089 add memory alternative.
22090 (tf_to_fprx2): New pattern.
22092 2021-02-19 Martin Sebor <msebor@redhat.com>
22095 * attribs.c (init_attr_rdwr_indices): Guard vblist use.
22096 (attr_access::free_lang_data): Remove a spurious test.
22098 2021-02-19 Nathan Sidwell <nathan@acm.org>
22100 * doc/invoke.texi (flang-info-module-read): Document.
22102 2021-02-19 Martin Liska <mliska@suse.cz>
22104 PR translation/99167
22105 * params.opt: Fix typo.
22107 2021-02-19 Richard Biener <rguenther@suse.de>
22109 PR middle-end/99122
22110 * tree-inline.c (inline_forbidden_p): Do not inline functions
22111 with VLA arguments or return value.
22113 2021-02-19 Jakub Jelinek <jakub@redhat.com>
22116 * config/arm/arm.md (*stack_protect_combined_set_insn,
22117 *stack_protect_combined_test_insn): If force_const_mem result
22118 is not valid general operand, force its address into the destination
22121 2021-02-19 Jakub Jelinek <jakub@redhat.com>
22124 * tree-cfg.c (gimple_merge_blocks): If bb a starts with eh landing
22125 pad or non-local label, put FORCED_LABELs from bb b after that label
22126 rather than before it.
22128 2021-02-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
22131 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3): Use
22132 expand_vector_broadcast' to emit the vec_duplicate operand.
22134 2021-02-18 Vladimir N. Makarov <vmakarov@redhat.com>
22136 PR rtl-optimization/96264
22137 * lra-remat.c (reg_overlap_for_remat_p): Check also output insn
22140 2021-02-18 H.J. Lu <hjl.tools@gmail.com>
22143 * varasm.c (get_section): Replace SUPPORTS_SHF_GNU_RETAIN with
22144 looking up the retain attribute.
22145 (resolve_unique_section): Likewise.
22146 (get_variable_section): Likewise.
22147 (switch_to_section): Likewise. Warn when a symbol without the
22148 retain attribute and a symbol with the retain attribute are
22149 placed in the section with the same name, instead of the used
22151 * doc/extend.texi: Document the "retain" attribute.
22153 2021-02-18 Nathan Sidwell <nathan@acm.org>
22156 * doc/invoke.texi (flang-info-include-translate): Document header
22159 2021-02-18 Richard Biener <rguenther@suse.de>
22161 PR middle-end/99122
22162 * ipa-fnsummary.c (analyze_function_body): Set
22163 CIF_FUNCTION_NOT_INLINABLE for VLA parameter calls.
22164 * tree-inline.c (insert_init_debug_bind): Pass NULL for
22165 error_mark_node values.
22166 (force_value_to_type): Do not build V_C_Es for WITH_SIZE_EXPR
22168 (setup_one_parameter): Delay force_value_to_type until when
22171 2021-02-18 Hans-Peter Nilsson <hp@axis.com>
22173 PR tree-optimization/99142
22174 * match.pd (clz cmp 0): Gate replacement on single_use of clz result.
22176 2021-02-18 Jakub Jelinek <jakub@redhat.com>
22178 * wide-int-bitmask.h (wide_int_bitmask::wide_int_bitmask (),
22179 wide_int_bitmask::wide_int_bitmask (uint64_t),
22180 wide_int_bitmask::wide_int_bitmask (uint64_t, uint64_t),
22181 wide_int_bitmask::operator ~ () const,
22182 wide_int_bitmask::operator | (wide_int_bitmask) const,
22183 wide_int_bitmask::operator & (wide_int_bitmask) const): Use constexpr
22185 * config/i386/i386.h (PTA_3DNOW, PTA_3DNOW_A, PTA_64BIT, PTA_ABM,
22186 PTA_AES, PTA_AVX, PTA_BMI, PTA_CX16, PTA_F16C, PTA_FMA, PTA_FMA4,
22187 PTA_FSGSBASE, PTA_LWP, PTA_LZCNT, PTA_MMX, PTA_MOVBE, PTA_NO_SAHF,
22188 PTA_PCLMUL, PTA_POPCNT, PTA_PREFETCH_SSE, PTA_RDRND, PTA_SSE, PTA_SSE2,
22189 PTA_SSE3, PTA_SSE4_1, PTA_SSE4_2, PTA_SSE4A, PTA_SSSE3, PTA_TBM,
22190 PTA_XOP, PTA_AVX2, PTA_BMI2, PTA_RTM, PTA_HLE, PTA_PRFCHW, PTA_RDSEED,
22191 PTA_ADX, PTA_FXSR, PTA_XSAVE, PTA_XSAVEOPT, PTA_AVX512F, PTA_AVX512ER,
22192 PTA_AVX512PF, PTA_AVX512CD, PTA_NO_TUNE, PTA_SHA, PTA_PREFETCHWT1,
22193 PTA_CLFLUSHOPT, PTA_XSAVEC, PTA_XSAVES, PTA_AVX512DQ, PTA_AVX512BW,
22194 PTA_AVX512VL, PTA_AVX512IFMA, PTA_AVX512VBMI, PTA_CLWB, PTA_MWAITX,
22195 PTA_CLZERO, PTA_NO_80387, PTA_PKU, PTA_AVX5124VNNIW, PTA_AVX5124FMAPS,
22196 PTA_AVX512VPOPCNTDQ, PTA_SGX, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES,
22197 PTA_AVX512VBMI2, PTA_VPCLMULQDQ, PTA_AVX512BITALG, PTA_RDPID,
22198 PTA_PCONFIG, PTA_WBNOINVD, PTA_AVX512VP2INTERSECT, PTA_PTWRITE,
22199 PTA_AVX512BF16, PTA_WAITPKG, PTA_MOVDIRI, PTA_MOVDIR64B, PTA_ENQCMD,
22200 PTA_CLDEMOTE, PTA_SERIALIZE, PTA_TSXLDTRK, PTA_AMX_TILE, PTA_AMX_INT8,
22201 PTA_AMX_BF16, PTA_UINTR, PTA_HRESET, PTA_KL, PTA_WIDEKL, PTA_AVXVNNI,
22202 PTA_X86_64_BASELINE, PTA_X86_64_V2, PTA_X86_64_V3, PTA_X86_64_V4,
22203 PTA_CORE2, PTA_NEHALEM, PTA_WESTMERE, PTA_SANDYBRIDGE, PTA_IVYBRIDGE,
22204 PTA_HASWELL, PTA_BROADWELL, PTA_SKYLAKE, PTA_SKYLAKE_AVX512,
22205 PTA_CASCADELAKE, PTA_COOPERLAKE, PTA_CANNONLAKE, PTA_ICELAKE_CLIENT,
22206 PTA_ICELAKE_SERVER, PTA_TIGERLAKE, PTA_SAPPHIRERAPIDS, PTA_ALDERLAKE,
22207 PTA_KNL, PTA_BONNELL, PTA_SILVERMONT, PTA_GOLDMONT, PTA_GOLDMONT_PLUS,
22208 PTA_TREMONT, PTA_KNM): Use constexpr instead of const.
22210 2021-02-18 Jakub Jelinek <jakub@redhat.com>
22212 PR middle-end/99109
22213 * gimple-array-bounds.cc (build_zero_elt_array_type): Rename to ...
22214 (build_printable_array_type): ... this. Add nelts argument. For
22215 overaligned eltype, use TYPE_MAIN_VARIANT (eltype) instead. If
22216 nelts, call build_array_type_nelts.
22217 (array_bounds_checker::check_mem_ref): Use build_printable_array_type
22218 instead of build_zero_elt_array_type and build_array_type_nelts.
22220 2021-02-18 Jakub Jelinek <jakub@redhat.com>
22223 * config/i386/i386.c (distance_non_agu_define): Don't call
22224 extract_insn_cached here.
22225 (ix86_lea_outperforms): Save and restore recog_data around call
22226 to distance_non_agu_define and distance_agu_use.
22227 (ix86_ok_to_clobber_flags): Remove.
22228 (ix86_avoid_lea_for_add): Don't call ix86_ok_to_clobber_flags.
22229 (ix86_avoid_lea_for_addr): Likewise. Adjust function comment.
22230 * config/i386/i386.md (*lea<mode>): Change from define_insn_and_split
22231 into define_insn. Move the splitting to define_peephole2 and
22232 check there using peep2_regno_dead_p if FLAGS_REG is dead.
22234 2021-02-17 Julian Brown <julian@codesourcery.com>
22236 * gimplify.c (gimplify_scan_omp_clauses): Handle ATTACH_DETACH
22239 2021-02-17 Xi Ruoyao <xry111@mengyan1223.wang>
22242 * config/mips/mips.c (mips_symbol_insns): Do not use
22243 MSA_SUPPORTED_MODE_P if mode is MAX_MACHINE_MODE.
22245 2021-02-16 Vladimir N. Makarov <vmakarov@redhat.com>
22247 PR inline-asm/98096
22248 * stmt.c (resolve_operand_name_1): Take inout operands into account
22249 for access to labels by names.
22250 * doc/extend.texi: Describe counting operands for accessing labels.
22252 2021-02-16 Richard Biener <rguenther@suse.de>
22254 PR tree-optimization/38474
22255 * tree-ssa-structalias.c (variable_info::address_taken): New.
22256 (new_var_info): Initialize address_taken.
22257 (process_constraint): Set address_taken.
22258 (solve_constraints): Use the new address_taken flag rather
22259 than is_reg_var for sorting variables.
22260 (dump_constraint): Dump the variable number if the name
22263 2021-02-16 Jakub Jelinek <jakub@redhat.com>
22266 * tree-vect-stmts.c (vectorizable_simd_clone_call): For num_calls != 1
22267 multiply by 4096 and for inbranch by 8192.
22268 * config/i386/i386.c (ix86_simd_clone_usable): For TARGET_AVX512F,
22269 return 3, 2 or 1 for mangle letters 'b', 'c' or 'd'.
22271 2021-02-15 Maya Rashish <coypu@sdf.org>
22273 * config/aarch64/aarch64.c (aarch64_init_builtins):
22274 Call SUBTARGET_INIT_BUILTINS.
22276 2021-02-15 Peter Bergner <bergner@linux.ibm.com>
22278 PR rtl-optimization/98872
22279 * init-regs.c (initialize_uninitialized_regs): Skip initialization
22280 if CONST0_RTX is NULL.
22282 2021-02-15 Richard Sandiford <richard.sandiford@arm.com>
22284 PR rtl-optimization/98863
22285 * rtl-ssa/functions.h (function_info::bb_live_out_info): Delete.
22286 (function_info::build_info): Turn into a declaration, moving the
22287 definition to internals.h.
22288 (function_info::bb_walker): Declare.
22289 (function_info::create_reg_use): Likewise.
22290 (function_info::calculate_potential_phi_regs): Take a build_info
22292 (function_info::place_phis, function_info::create_ebbs): Declare.
22293 (function_info::calculate_ebb_live_in_for_debug): Likewise.
22294 (function_info::populate_backedge_phis): Delete.
22295 (function_info::start_block, function_info::end_block): Declare.
22296 (function_info::populate_phi_inputs): Delete.
22297 (function_info::m_potential_phi_regs): Move information to build_info.
22298 * rtl-ssa/internals.h: New file.
22299 (function_info::bb_phi_info): New class.
22300 (function_info::build_info): Moved from functions.h.
22301 Add a constructor and destructor.
22302 (function_info::build_info::ebb_use): Delete.
22303 (function_info::build_info::ebb_def): Likewise.
22304 (function_info::build_info::bb_live_out): Likewise.
22305 (function_info::build_info::tmp_ebb_live_in_for_debug): New variable.
22306 (function_info::build_info::potential_phi_regs): Likewise.
22307 (function_info::build_info::potential_phi_regs_for_debug): Likewise.
22308 (function_info::build_info::ebb_def_regs): Likewise.
22309 (function_info::build_info::bb_phis): Likewise.
22310 (function_info::build_info::bb_mem_live_out): Likewise.
22311 (function_info::build_info::bb_to_rpo): Likewise.
22312 (function_info::build_info::def_stack): Likewise.
22313 (function_info::build_info::old_def_stack_limit): Likewise.
22314 * rtl-ssa/internals.inl (function_info::build_info::record_reg_def):
22315 Remove the regno argument. Push the previous definition onto the
22316 definition stack where necessary.
22317 * rtl-ssa/accesses.cc: Include internals.h.
22318 * rtl-ssa/changes.cc: Likewise.
22319 * rtl-ssa/blocks.cc: Likewise.
22320 (function_info::build_info::build_info): Define.
22321 (function_info::build_info::~build_info): Likewise.
22322 (function_info::bb_walker): New class.
22323 (function_info::bb_walker::bb_walker): Define.
22324 (function_info::add_live_out_use): Convert a logarithmic-complexity
22325 test into a linear one. Allow the same definition to be passed
22327 (function_info::calculate_potential_phi_regs): Moved from
22328 functions.cc. Take a build_info parameter and store the
22329 information there instead.
22330 (function_info::place_phis): New function.
22331 (function_info::add_entry_block_defs): Update call to record_reg_def.
22332 (function_info::calculate_ebb_live_in_for_debug): New function.
22333 (function_info::add_phi_nodes): Use bb_phis to decide which
22334 registers need phi nodes and initialize ebb_def_regs accordingly.
22335 Do not add degenerate phis here.
22336 (function_info::add_artificial_accesses): Use create_reg_use.
22337 Assert that all definitions are listed in the DF LR sets.
22338 Update call to record_reg_def.
22339 (function_info::record_block_live_out): Record live-out register
22340 values in the phis of successor blocks. Use the live-out set
22341 when processing the last block in an EBB, instead of always
22342 using the live-in sets of successor blocks. AND the live sets
22343 with the set of registers that have been defined in the EBB,
22344 rather than with all potential phi registers. Cope correctly
22345 with branches back to the start of the current EBB.
22346 (function_info::start_block): New function.
22347 (function_info::end_block): Likewise.
22348 (function_info::populate_phi_inputs): Likewise.
22349 (function_info::create_ebbs): Likewise.
22350 (function_info::process_all_blocks): Rewrite into a multi-phase
22352 * rtl-ssa/functions.cc: Include internals.h.
22353 (function_info::calculate_potential_phi_regs): Move to blocks.cc.
22354 (function_info::init_function_data): Remove caller.
22355 * rtl-ssa/insns.cc: Include internals.h
22356 (function_info::create_reg_use): New function. Lazily any
22357 degenerate phis needed by the linear RPO view.
22358 (function_info::record_use): Use create_reg_use. When processing
22359 debug uses, use potential_phi_regs and test it before checking
22360 whether the register is live on entry to the current EBB. Lazily
22361 calculate ebb_live_in_for_debug.
22362 (function_info::record_call_clobbers): Update call to record_reg_def.
22363 (function_info::record_def): Likewise.
22365 2021-02-15 Martin Liska <mliska@suse.cz>
22367 * toplev.c (init_asm_output): Free output of
22368 gen_command_line_string function.
22369 (process_options): Likewise.
22371 2021-02-15 Martin Liska <mliska@suse.cz>
22373 * params.opt: Add 2 missing Param keywords.
22375 2021-02-15 Eric Botcazou <ebotcazou@adacore.com>
22377 * df-core.c (df_worklist_dataflow_doublequeue): Use proper cast.
22379 2021-02-15 Jakub Jelinek <jakub@redhat.com>
22381 PR tree-optimization/99079
22382 * match.pd (A % (pow2pcst << N) -> A & ((pow2pcst << N) - 1)): Remove
22383 useless tree_nop_conversion_p (type, TREE_TYPE (@3)) check. Instead
22384 require both type and TREE_TYPE (@1) to be integral types and either
22385 type having smaller or equal precision, or TREE_TYPE (@1) being
22386 unsigned type, or type being signed type. If TREE_TYPE (@1)
22387 doesn't have wrapping overflow, perform the subtraction of one in
22390 2021-02-14 Jan Hubicka <hubicka@ucw.cz>
22391 Richard Biener <rguether@suse.de>
22394 * ipa-reference.c (ipa_init): Only conditinally initialize
22395 reference_vars_to_consider.
22396 (propagate): Conditionally deninitialize reference_vars_to_consider.
22397 (ipa_reference_write_optimization_summary): Sanity check that
22398 reference_vars_to_consider is not allocated.
22400 2021-02-13 Levy Hsu <admin@levyhsu.com>
22403 * config/riscv/riscv-shorten-memrefs.c (pass_shorten_memrefs): Add
22404 extend parameter to get_si_mem_base_reg declaration.
22405 (get_si_mem_base_reg): Add extend parameter. Set it.
22406 (analyze): Pass extend arg to get_si_mem_base_reg.
22407 (transform): Likewise. Use it when rewriting mems.
22408 * config/riscv/riscv.c (riscv_legitimize_move): Check for subword
22409 loads and emit sign/zero extending load followed by subreg move.
22411 2021-02-13 Jim Wilson <jimw@sifive.com>
22414 * config/riscv/riscv.c (riscv_compressed_lw_address_p): Drop early
22415 exit when !reload_completed. Only perform check for compressed reg
22416 if reload_completed.
22417 (riscv_rtx_costs): In MEM case, when optimizing for size and
22418 shorten memrefs, if not compressible, then increase cost.
22420 2021-02-13 Jakub Jelinek <jakub@redhat.com>
22422 PR rtl-optimization/98439
22423 * recog.c (pass_split_before_regstack::gate): Enable even when
22424 pass_split_before_sched2 is enabled if -fselective-scheduling2 is
22427 2021-02-13 Jakub Jelinek <jakub@redhat.com>
22430 * config/i386/mmx.md (*mmx_pshufd_1): Add a combine splitter for
22431 swap of V2SImode elements in memory into DImode memory rotate by 32.
22433 2021-02-12 Martin Sebor <msebor@redhat.com>
22435 * tree-pretty-print.c (print_generic_expr_to_str): Update comment.
22437 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
22439 * rtl-ssa/accesses.cc (function_info::make_use_available): Use
22440 m_temp_obstack rather than m_obstack to allocate the temporary use.
22442 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
22444 * df-problems.c (df_lr_bb_local_compute): Treat partial definitions
22445 as read-modify operations.
22447 2021-02-12 Richard Biener <rguenther@suse.de>
22449 PR middle-end/38474
22450 * ipa-fnsummary.c (unmodified_parm_1): Only walk when
22451 fbi->aa_walk_budget is bigger than zero. Update
22452 fbi->aa_walk_budget.
22453 (param_change_prob): Likewise.
22454 * ipa-prop.c (detect_type_change_from_memory_writes):
22455 Properly account walk_aliased_vdefs.
22456 (parm_preserved_before_stmt_p): Canonicalize updates.
22457 (parm_ref_data_preserved_p): Likewise.
22458 (parm_ref_data_pass_through_p): Likewise.
22459 (determine_known_aggregate_parts): Account own alias queries.
22461 2021-02-12 Martin Liska <mliska@suse.cz>
22463 * opts-common.c (decode_cmdline_option): Release werror_arg.
22464 * opts.c (gen_producer_string): Release output of
22465 gen_command_line_string.
22467 2021-02-12 Richard Biener <rguenther@suse.de>
22469 PR tree-optimization/38474
22470 * params.opt (-param=max-store-chains-to-track=): New param.
22471 (-param=max-stores-to-track=): Likewise.
22472 * doc/invoke.texi (max-store-chains-to-track): Document.
22473 (max-stores-to-track): Likewise.
22474 * gimple-ssa-store-merging.c (pass_store_merging::m_n_chains):
22476 (pass_store_merging::m_n_stores): Likewise.
22477 (pass_store_merging::terminate_and_process_chain): Update
22478 m_n_stores and m_n_chains.
22479 (pass_store_merging::process_store): Likewise. Terminate
22480 oldest chains if the number of stores or chains get too large.
22481 (imm_store_chain_info::terminate_and_process_chain): Dump
22484 2021-02-11 Eric Botcazou <ebotcazou@adacore.com>
22486 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
22487 the cold section, emit a nop before the directive if the previous
22488 active instruction can throw.
22490 2021-02-11 Peter Bergner <bergner@linux.ibm.com>
22493 * config/rs6000/predicates.md (mma_assemble_input_operand): Restrict
22494 memory addresses that are legal for quad word accesses.
22496 2021-02-11 Andrea Corallo <andrea.corallo@arm.com>
22499 * config/arm/thumb2.md (*doloop_end_internal): Generate
22500 alternative sequence to handle long range branches.
22502 2021-02-11 Joel Hutton <joel.hutton@arm.com>
22504 PR tree-optimization/98772
22505 * optabs-tree.c (supportable_half_widening_operation): New function
22506 to check for supportable V8QI->V8HI widening patterns.
22507 * optabs-tree.h (supportable_half_widening_operation): New function.
22508 * tree-vect-stmts.c (vect_create_half_widening_stmts): New function
22509 to create promotion stmts for V8QI->V8HI widening patterns.
22510 (vectorizable_conversion): Add case for V8QI->V8HI.
22512 2021-02-11 Richard Biener <rguenther@suse.de>
22514 * sparseset.h (SPARSESET_ELT_BITS): Remove.
22515 (SPARSESET_ELT_TYPE): Use unsigned int.
22516 * fwprop.c: Do not include sparseset.h.
22518 2021-02-10 Jakub Jelinek <jakub@redhat.com>
22521 * varasm.c (declare_weak): For -fsyntax-only, allow even
22522 TREE_ASM_WRITTEN function decls.
22524 2021-02-10 Jakub Jelinek <jakub@redhat.com>
22527 * config/i386/sse.md (fix<fixunssuffix>_truncv2sfv2di2,
22528 <insn>v8qiv8hi2, <insn>v8qiv8si2, <insn>v4qiv4si2, <insn>v4hiv4si2,
22529 <insn>v8qiv8di2, <insn>v4qiv4di2, <insn>v2qiv2di2, <insn>v4hiv4di2,
22530 <insn>v2hiv2di2, <insn>v2siv2di2): Force operands[1] into REG before
22531 calling simplify_gen_subreg on it.
22533 2021-02-10 Martin Liska <mliska@suse.cz>
22535 * config/nvptx/nvptx.c (nvptx_option_override): Use
22536 flag_patchable_function_entry instead of the removed
22537 function_entry_patch_area_size.
22539 2021-02-10 Martin Liska <mliska@suse.cz>
22541 PR tree-optimization/99002
22542 PR tree-optimization/99026
22543 * gimple-if-to-switch.cc (if_chain::is_beneficial): Fix memory
22544 leak when adjacent cases are merged.
22545 * tree-switch-conversion.c (switch_decision_tree::analyze_switch_statement): Use
22547 (make_pass_lower_switch): Remove trailing whitespace.
22548 * tree-switch-conversion.h (release_clusters): New.
22550 2021-02-10 Richard Biener <rguenther@suse.de>
22552 PR rtl-optimization/99054
22553 * cfgrtl.c (rtl-optimization/99054): Return an auto_vec.
22554 (fixup_partitions): Adjust.
22555 (rtl_verify_edges): Likewise.
22557 2021-02-10 Jakub Jelinek <jakub@redhat.com>
22559 PR middle-end/99007
22560 * gimplify.c (gimplify_scan_omp_clauses): For MEM_REF on reductions,
22561 temporarily disable gimplify_ctxp->into_ssa around gimplify_expr
22564 2021-02-10 Richard Biener <rguenther@suse.de>
22567 * ipa-pure-const.c (propagate_malloc): Use an auto_vec<>
22570 2021-02-10 Richard Biener <rguenther@suse.de>
22572 PR tree-optimization/99024
22573 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Only
22574 clear loop->aux if it is associated with the destroyed loop_vinfo.
22576 2021-02-10 Martin Liska <mliska@suse.cz>
22578 PR tree-optimization/99002
22579 * gimple-if-to-switch.cc (find_conditions): Fix memory leak
22582 2021-02-10 Martin Liska <mliska@suse.cz>
22585 * ipa-icf.c (sem_item::add_reference): Fix memory leak when
22586 a reference exists.
22588 2021-02-10 Jakub Jelinek <jakub@redhat.com>
22591 * dwarf2out.c (prune_unused_types_walk): Mark DW_TAG_variable DIEs
22592 at class scope for DWARF5+.
22594 2021-02-09 Eric Botcazou <ebotcazou@adacore.com>
22596 PR rtl-optimization/96015
22597 * reorg.c (skip_consecutive_labels): Minor comment tweaks.
22598 (relax_delay_slots): When deleting a jump to the next active
22599 instruction over a barrier, first delete the barrier if the
22600 jump is the only way to reach the target label.
22602 2021-02-09 Andre Vieira <andre.simoesdiasvieira@arm.com>
22604 * config/aarch64/aarch64-cost-tables.h: Add entries for vect.mul.
22605 * config/aarch64/aarch64.c (aarch64_rtx_mult_cost): Use vect.mul for
22606 vector multiplies and vect.alu for SSRA.
22607 * config/arm/aarch-common-protos.h (struct vector_cost_table): Define
22608 vect.mul cost field.
22609 * config/arm/aarch-cost-tables.h: Add entries for vect.mul.
22610 * config/arm/arm.c: Likewise.
22612 2021-02-09 Richard Biener <rguenther@suse.de>
22614 PR tree-optimization/98863
22615 * tree-ssa-sccvn.h (vn_avail::next_undo): Add.
22616 * tree-ssa-sccvn.c (last_pushed_avail): New global.
22617 (rpo_elim::eliminate_push_avail): Chain pushed avails.
22618 (unwind_state::avail_top): Add.
22619 (do_unwind): Rewrite unwinding of avail entries.
22620 (do_rpo_vn): Initialize last_pushed_avail and
22621 avail_top of the undo state.
22623 2021-02-09 Jakub Jelinek <jakub@redhat.com>
22625 PR middle-end/99004
22626 * calls.c (maybe_warn_rdwr_sizes): Change s0 and s1 type from
22627 const char * to char * and free those pointers after use.
22629 2021-02-09 Richard Biener <rguenther@suse.de>
22631 PR tree-optimization/99017
22632 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Allow
22633 zero vector cost entries.
22635 2021-02-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
22637 PR middle-end/98974
22638 * tree-vect-stmts.c (vectorizable_condition): Remove shadow vec_num
22639 parameter in vectorizable_condition.
22641 2021-02-08 Richard Biener <rguenther@suse.de>
22644 * tree.c (walk_tree_1): Walk VECTOR_CST elements.
22646 2021-02-08 Martin Liska <mliska@suse.cz>
22649 * cfgexpand.c (pass_expand::execute): Parse per-function option
22650 flag_patchable_function_entry and use it.
22651 * common.opt: Remove function_entry_patch_area_size and
22652 function_entry_patch_area_start global variables.
22653 * opts.c (parse_and_check_patch_area): New function.
22654 (common_handle_option): Use it.
22655 * opts.h (parse_and_check_patch_area): New function.
22656 * toplev.c (process_options): Parse and use
22657 function_entry_patch_area_size.
22659 2021-02-08 Martin Sebor <msebor@redhat.com>
22661 * doc/extend.texi (attribute malloc): Correct typos.
22663 2021-02-05 Nathan Sidwell <nathan@acm.org>
22666 * gcc.c (driver::maybe_run_linker): Check for input file
22667 accessibility if not linking.
22669 2021-02-05 Richard Biener <rguenther@suse.de>
22671 PR tree-optimization/98855
22672 * tree-vectorizer.h (add_stmt_cost): New overload.
22673 * tree-vect-slp.c (li_cost_vec_cmp): New.
22674 (vect_bb_slp_scalar_cost): Cost individual loop regions
22675 separately. Account for the scalar instance root stmt.
22677 2021-02-05 Tom de Vries <tdevries@suse.de>
22680 * tree-switch-conversion.c (jump_table_cluster::emit): Add loc
22682 (bit_test_cluster::emit): Reuse location_t for newly created
22684 (switch_decision_tree::try_switch_expansion): Preserve
22686 * tree-switch-conversion.h: Change function signatures.
22688 2021-02-05 Jakub Jelinek <jakub@redhat.com>
22691 * config/i386/i386-options.c (m_NONE, m_ALL): Define.
22692 * config/i386/x86-tune.def (X86_TUNE_BRANCH_PREDICTION_HINTS,
22693 X86_TUNE_PROMOTE_QI_REGS): Use m_NONE instead of 0U.
22694 (X86_TUNE_QIMODE_MATH): Use m_ALL instead of ~0U.
22696 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22698 * config/aarch64/aarch64-simd-builtins.def (get_high): Define builtin.
22699 * config/aarch64/aarch64-simd.md (aarch64_get_high<mode>): Define.
22700 * config/aarch64/arm_neon.h (__GET_HIGH): Delete.
22701 (vget_high_f16): Reimplement using new builtin.
22702 (vget_high_f32): Likewise.
22703 (vget_high_f64): Likewise.
22704 (vget_high_p8): Likewise.
22705 (vget_high_p16): Likewise.
22706 (vget_high_p64): Likewise.
22707 (vget_high_s8): Likewise.
22708 (vget_high_s16): Likewise.
22709 (vget_high_s32): Likewise.
22710 (vget_high_s64): Likewise.
22711 (vget_high_u8): Likewise.
22712 (vget_high_u16): Likewise.
22713 (vget_high_u32): Likewise.
22714 (vget_high_u64): Likewise.
22716 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22718 * config/aarch64/aarch64-simd-builtins.def (get_low): Define builtin.
22719 * config/aarch64/aarch64-simd.md (aarch64_get_low<mode>): Define.
22720 * config/aarch64/arm_neon.h (__GET_LOW): Delete.
22721 (vget_low_f16): Reimplement using new builtin.
22722 (vget_low_f32): Likewise.
22723 (vget_low_f64): Likewise.
22724 (vget_low_p8): Likewise.
22725 (vget_low_p16): Likewise.
22726 (vget_low_p64): Likewise.
22727 (vget_low_s8): Likewise.
22728 (vget_low_s16): Likewise.
22729 (vget_low_s32): Likewise.
22730 (vget_low_s64): Likewise.
22731 (vget_low_u8): Likewise.
22732 (vget_low_u16): Likewise.
22733 (vget_low_u32): Likewise.
22734 (vget_low_u64): Likewise.
22736 2021-02-05 Kito Cheng <kito.cheng@sifive.com>
22738 * gcc.c (print_multilib_info): Check all required argument is provided
22741 2021-02-05 liuhongt <hongtao.liu@intel.com>
22744 * config/i386/i386-expand.c (ix86_expand_sse_cmp): Don't
22745 generate integer mask comparison for 128/256-bits vector when
22746 op_true/op_false is NULL_RTX or CONSTM1_RTX/CONST0_RTX. Also
22747 delete redundant !maskcmp condition.
22748 (ix86_expand_int_vec_cmp): Ditto but no redundant deletion
22750 (ix86_expand_sse_movcc): Delete definition of maskcmp, add the
22751 condition directly to if (maskcmp), add extra check for
22752 cmpmode, it should be MODE_INT.
22753 (ix86_expand_fp_vec_cmp): Pass NULL to ix86_expand_sse_cmp's
22754 parameters op_true/op_false.
22755 (ix86_use_mask_cmp_p): New.
22757 2021-02-05 liuhongt <hongtao.liu@intel.com>
22760 * config/i386/x86-tune.def (X86_TUNE_AVX256_UNALIGNED_LOAD_OPTIMAL):
22761 Remove m_GENERIC from ~list.
22762 (X86_TUNE_AVX256_UNALIGNED_STORE_OPTIMAL): Ditto.
22764 2021-02-04 David Malcolm <dmalcolm@redhat.com>
22767 * diagnostic-show-locus.c (compatible_locations_p): Require
22768 locations in the same macro map to be either both from the
22769 macro definition, or both from the macro arguments.
22771 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
22773 * config/aarch64/aarch64-simd-builtins.def: Add
22774 [su]mull_hi_lane[q] builtin generator macros.
22775 * config/aarch64/aarch64-simd.md
22776 (aarch64_<su>mull_hi_lane<mode>_insn): Define.
22777 (aarch64_<su>mull_hi_lane<mode>): Define.
22778 (aarch64_<su>mull_hi_laneq<mode>_insn): Define.
22779 (aarch64_<su>mull_hi_laneq<mode>): Define.
22780 * config/aarch64/arm_neon.h (vmull_high_lane_s16): Use RTL
22781 builtin instead of inline asm.
22782 (vmull_high_lane_s32): Likewise.
22783 (vmull_high_lane_u16): Likewise.
22784 (vmull_high_lane_u32): Likewise.
22785 (vmull_high_laneq_s16): Likewise.
22786 (vmull_high_laneq_s32): Likewise.
22787 (vmull_high_laneq_u16): Likewise.
22788 (vmull_high_laneq_u32): Liekwise.
22790 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
22792 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_hi_n
22793 builtin generator macros.
22794 * config/aarch64/aarch64-simd.md
22795 (aarch64_<su>mull_hi_n<mode>_insn): Define.
22796 (aarch64_<su>mull_hi_n<mode>): Define.
22797 * config/aarch64/arm_neon.h (vmull_high_n_s16): Use RTL builtin
22798 instead of inline asm.
22799 (vmull_high_n_s32): Likewise.
22800 (vmull_high_n_u16): Likewise.
22801 (vmull_high_n_u32): Likewise.
22803 2021-02-04 Richard Biener <rguenther@suse.de>
22805 PR tree-optimization/98855
22806 * tree-vect-loop.c (vectorizable_phi): Do not cost
22807 single-argument PHIs.
22808 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Likewise.
22809 * tree-vect-stmts.c (vectorizable_bswap): Also perform
22810 costing for SLP operation.
22812 2021-02-04 Martin Liska <mliska@suse.cz>
22814 * doc/extend.texi: Mention -mprefer-vector-width in target
22817 2021-02-03 Martin Sebor <msebor@redhat.com>
22819 PR tree-optimization/98937
22820 * tree-ssa-strlen.c (strlen_dom_walker::~strlen_dom_walker): Define.
22821 Flush pointer_query cache.
22823 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
22825 * config/rs6000/genfusion.pl (gen_2logical): Add missing
22826 fixes based on patch review.
22827 * config/rs6000/fusion.md: Regenerate file.
22829 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
22831 * config/rs6000/t-rs6000: Comment out auto generation of
22834 2021-02-03 Andrew Stubbs <ams@codesourcery.com>
22836 * config/gcn/gcn-opts.h (enum processor_type): Add PROCESSOR_GFX908.
22837 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Add gfx908.
22838 (output_file_start): Add gfx908.
22839 * config/gcn/gcn.opt (gpu_type): Add gfx908.
22840 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Add march=gfx908.
22841 (MULTILIB_DIRNAMES): Add gfx908.
22842 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): New define.
22843 (main): Recognize gfx908.
22844 * config/gcn/t-omp-device: Add gfx908.
22846 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22848 * config/aarch64/aarch64-simd-builtins.def: Add
22849 [su]mlsl_hi_lane[q] builtin macro generators.
22850 * config/aarch64/aarch64-simd.md
22851 (aarch64_<su>mlsl_hi_lane<mode>_insn): Define.
22852 (aarch64_<su>mlsl_hi_lane<mode>): Define.
22853 (aarch64_<su>mlsl_hi_laneq<mode>_insn): Define.
22854 (aarch64_<su>mlsl_hi_laneq<mode>): Define.
22855 * config/aarch64/arm_neon.h (vmlsl_high_lane_s16): Use RTL
22856 builtin instead of inline asm.
22857 (vmlsl_high_lane_s32): Likewise.
22858 (vmlsl_high_lane_u16): Likewise.
22859 (vmlsl_high_lane_u32): Likewise.
22860 (vmlsl_high_laneq_s16): Likewise.
22861 (vmlsl_high_laneq_s32): Likewise.
22862 (vmlsl_high_laneq_u16): Likewise.
22863 (vmlsl_high_laneq_u32): Likewise.
22864 (vmlal_high_laneq_u32): Likewise.
22866 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22868 * config/aarch64/aarch64-simd-builtins.def: Add
22869 [su]mlal_hi_lane[q] builtin generator macros.
22870 * config/aarch64/aarch64-simd.md
22871 (aarch64_<su>mlal_hi_lane<mode>_insn): Define.
22872 (aarch64_<su>mlal_hi_lane<mode>): Define.
22873 (aarch64_<su>mlal_hi_laneq<mode>_insn): Define.
22874 (aarch64_<su>mlal_hi_laneq<mode>): Define.
22875 * config/aarch64/arm_neon.h (vmlal_high_lane_s16): Use RTL
22876 builtin instead of inline asm.
22877 (vmlal_high_lane_s32): Likewise.
22878 (vmlal_high_lane_u16): Likewise.
22879 (vmlal_high_lane_u32): Likewise.
22880 (vmlal_high_laneq_s16): Likewise.
22881 (vmlal_high_laneq_s32): Likewise.
22882 (vmlal_high_laneq_u16): Likewise.
22883 (vmlal_high_laneq_u32): Likewise.
22885 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22887 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_hi_n
22888 builtin generator macros.
22889 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_hi_n<mode>_insn):
22891 (aarch64_<su>mlsl_hi_n<mode>): Define.
22892 * config/aarch64/arm_neon.h (vmlsl_high_n_s16): Use RTL builtin
22893 instead of inline asm.
22894 (vmlsl_high_n_s32): Likewise.
22895 (vmlsl_high_n_u16): Likewise.
22896 (vmlsl_high_n_u32): Likewise.
22898 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22900 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_hi_n
22901 builtin generator macros.
22902 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_hi_n<mode>_insn):
22904 (aarch64_<su>mlal_hi_n<mode>): Define.
22905 * config/aarch64/arm_neon.h (vmlal_high_n_s16): Use RTL builtin
22906 instead of inline asm.
22907 (vmlal_high_n_s32): Likewise.
22908 (vmlal_high_n_u16): Likewise.
22909 (vmlal_high_n_u32): Likewise.
22911 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22913 * config/aarch64/aarch64-simd-builtins.def: Add RTL builtin
22915 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal_hi<mode>):
22917 (aarch64_<su>mlal_hi<mode>_insn): This.
22918 (aarch64_<su>mlal_hi<mode>): Define.
22919 * config/aarch64/arm_neon.h (vmlal_high_s8): Use RTL builtin
22920 instead of inline asm.
22921 (vmlal_high_s16): Likewise.
22922 (vmlal_high_s32): Likewise.
22923 (vmlal_high_u8): Likewise.
22924 (vmlal_high_u16): Likewise.
22925 (vmlal_high_u32): Likewise.
22927 2021-02-03 Ilya Leoshkevich <iii@linux.ibm.com>
22929 * lra-spills.c (remove_pseudos): Call lra_update_insn_recog_data()
22930 after calling alter_subreg() on a (mem).
22932 2021-02-03 Martin Liska <mliska@suse.cz>
22935 * lto-streamer-out.c (produce_lto_section): Fill up missing
22937 * lto-streamer.h (struct lto_section): Add _padding field.
22939 2021-02-03 Richard Biener <rguenther@suse.de>
22941 * lto-streamer.c (lto_get_section_name): Free temporary
22943 * tree-loop-distribution.c
22944 (loop_distribution::merge_dep_scc_partitions): Free edge data.
22946 2021-02-03 Jakub Jelinek <jakub@redhat.com>
22948 PR middle-end/97487
22949 * ifcvt.c (noce_can_force_operand): New function.
22950 (noce_emit_move_insn): Use it.
22951 (noce_try_sign_mask): Likewise. Formatting fix.
22953 2021-02-03 Jakub Jelinek <jakub@redhat.com>
22955 PR middle-end/97971
22956 * lra-constraints.c (process_alt_operands): For inline asm, don't call
22957 fatal_insn, but instead return false.
22959 2021-02-03 Jakub Jelinek <jakub@redhat.com>
22961 PR tree-optimization/98287
22962 * config/i386/mmx.md (<insn><mode>3): For shifts don't enable expander
22965 2021-02-03 Tamar Christina <tamar.christina@arm.com>
22967 PR tree-optimization/98928
22968 * tree-vect-loop.c (vect_analyze_loop_2): Change
22969 STMT_VINFO_SLP_VECT_ONLY to STMT_VINFO_SLP_VECT_ONLY_PATTERN.
22970 * tree-vect-slp-patterns.c (complex_pattern::build): Likewise.
22971 * tree-vectorizer.h (STMT_VINFO_SLP_VECT_ONLY_PATTERN): New.
22972 (class _stmt_vec_info): Add slp_vect_pattern_only_p.
22974 2021-02-02 Richard Biener <rguenther@suse.de>
22976 * gimple-loop-interchange.cc (prepare_data_references):
22978 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
22979 * tree-ssa-loop-im.c (hoist_memory_references): Likewise.
22980 * tree-vect-stmts.c (vectorizable_condition): Do not
22982 (vectorizable_comparison): Likewise.
22984 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22986 * config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
22987 * config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
22988 * config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
22989 (vrsqrteq_u32): Likewise.
22991 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22993 * config/aarch64/aarch64-simd-builtins.def (sqxtun2): Define builtin.
22994 * config/aarch64/aarch64-simd.md (aarch64_sqxtun2<mode>_le): Define.
22995 (aarch64_sqxtun2<mode>_be): Likewise.
22996 (aarch64_sqxtun2<mode>): Likewise.
22997 * config/aarch64/arm_neon.h (vqmovun_high_s16): Reimplement using builtin.
22998 (vqmovun_high_s32): Likewise.
22999 (vqmovun_high_s64): Likewise.
23000 * config/aarch64/iterators.md (UNSPEC_SQXTUN2): Define.
23002 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23004 * config/aarch64/aarch64-simd-builtins.def (bfdot_lane, bfdot_laneq): Use
23006 (bfmlalb_lane, bfmlalt_lane, bfmlalb_lane_q, bfmlalt_lane_q): Use FP flags.
23008 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23010 * config/aarch64/aarch64-simd-builtins.def (fcmla_lane0, fcmla_lane90,
23011 fcmla_lane180, fcmla_lane270, fcmlaq_lane0, fcmlaq_lane90, fcmlaq_lane180,
23012 fcmlaq_lane270, scvtf, ucvtf, fcvtzs, fcvtzu, scvtfsi, scvtfdi, ucvtfsi,
23013 ucvtfdi, fcvtzshf, fcvtzuhf, fmlal_lane_low, fmlsl_lane_low,
23014 fmlal_laneq_low, fmlsl_laneq_low, fmlalq_lane_low, fmlslq_lane_low,
23015 fmlalq_laneq_low, fmlslq_laneq_low, fmlal_lane_high, fmlsl_lane_high,
23016 fmlal_laneq_high, fmlsl_laneq_high, fmlalq_lane_high, fmlslq_lane_high,
23017 fmlalq_laneq_high, fmlslq_laneq_high): Use FP flags.
23019 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23021 * config/aarch64/aarch64-builtins.c (FLAG_LOAD): Define.
23022 * config/aarch64/aarch64-simd-builtins.def (ld1x2, ld2, ld3, ld4, ld2r,
23023 ld3r, ld4r, ld1, ld1x3, ld1x4): Use LOAD flags.
23025 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23027 * config/aarch64/aarch64-simd-builtins.def (combine, zip1, zip2,
23028 uzp1, uzp2, trn1, trn2, simd_bsl): Use AUTO_FP flags.
23030 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23032 * config/aarch64/aarch64-simd-builtins.def (clrsb, clz, ctz, popcount,
23033 vec_smult_lane_, vec_smlal_lane_, vec_smult_laneq_, vec_smlal_laneq_,
23034 vec_umult_lane_, vec_umlal_lane_, vec_umult_laneq_, vec_umlal_laneq_,
23035 ashl, sshl, ushl, srshl, urshl, sdot_lane, udot_lane, sdot_laneq,
23036 udot_laneq, usdot_lane, usdot_laneq, sudot_lane, sudot_laneq, ashr,
23037 ashr_simd, lshr, lshr_simd, srshr_n, urshr_n, ssra_n, usra_n, srsra_n,
23038 ursra_n, sshll_n, ushll_n, sshll2_n, ushll2_n, ssri_n, usri_n, ssli_n,
23039 ssli_n, usli_n, bswap, rbit, simd_bsl, eor3q, rax1q, xarq, bcaxq): Use
23040 NONE builtin flags.
23042 2021-02-02 Jakub Jelinek <jakub@redhat.com>
23044 PR tree-optimization/98848
23045 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
23046 STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
23048 2021-02-02 Kito Cheng <kito.cheng@sifive.com>
23051 * expr.c: Check mode before calling store_expr.
23053 2021-02-02 Christophe Lyon <christophe.lyon@linaro.org>
23055 * config/arm/iterators.md (supf): Remove VORNQ_S and VORNQ_U.
23057 * config/arm/mve.md (mve_vornq_s<mode>): New entry for vorn
23058 instruction using expression ior.
23059 (mve_vornq_u<mode>): New expander.
23060 (mve_vornq_f<mode>): Use ior code instead of unspec.
23061 * config/arm/unspecs.md (VORNQ_S, VORNQ_U, VORNQ_F): Remove.
23063 2021-02-02 Alexandre Oliva <oliva@adacore.com>
23065 * tree-nested.c (convert_nonlocal_reference_op): Move
23066 current_function_decl restore after re-gimplification.
23067 (convert_local_reference_op): Likewise.
23069 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23071 * config/aarch64/aarch64-simd-builtins.def (rshrn, rshrn2):
23073 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>_insn_le):
23075 (aarch64_rshrn<mode>_insn_be): Likewise.
23076 (aarch64_rshrn<mode>): Likewise.
23077 (aarch64_rshrn2<mode>_insn_le): Likewise.
23078 (aarch64_rshrn2<mode>_insn_be): Likewise.
23079 (aarch64_rshrn2<mode>): Likewise.
23080 * config/aarch64/aarch64.md (unspec): Add UNSPEC_RSHRN.
23081 * config/aarch64/arm_neon.h (vrshrn_high_n_s16): Reimplement
23083 (vrshrn_high_n_s32): Likewise.
23084 (vrshrn_high_n_s64): Likewise.
23085 (vrshrn_high_n_u16): Likewise.
23086 (vrshrn_high_n_u32): Likewise.
23087 (vrshrn_high_n_u64): Likewise.
23088 (vrshrn_n_s16): Likewise.
23089 (vrshrn_n_s32): Likewise.
23090 (vrshrn_n_s64): Likewise.
23091 (vrshrn_n_u16): Likewise.
23092 (vrshrn_n_u32): Likewise.
23093 (vrshrn_n_u64): Likewise.
23095 2021-02-01 Sergei Trofimovich <siarheit@google.com>
23097 PR tree-optimization/98499
23098 * ipa-modref.c (analyze_ssa_name_flags): treat RVO
23099 conservatively and assume all possible side-effects.
23101 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23103 * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi,
23104 vec_unpacku_hi_): Define builtins.
23105 * config/aarch64/arm_neon.h (vmovl_high_s8): Reimplement using
23107 (vmovl_high_s16): Likewise.
23108 (vmovl_high_s32): Likewise.
23109 (vmovl_high_u8): Likewise.
23110 (vmovl_high_u16): Likewise.
23111 (vmovl_high_u32): Likewise.
23113 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23115 * config/aarch64/aarch64-simd-builtins.def (sabdl, uabdl):
23117 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl<mode>): New
23119 * config/aarch64/aarch64.md (unspec): Define UNSPEC_SABDL,
23121 * config/aarch64/arm_neon.h (vabdl_s8): Reimplemet using
23123 (vabdl_s16): Likewise.
23124 (vabdl_s32): Likewise.
23125 (vabdl_u8): Likewise.
23126 (vabdl_u16): Likewise.
23127 (vabdl_u32): Likewise.
23128 * config/aarch64/iterators.md (ABDL): New int iterator.
23129 (sur): Handle UNSPEC_SABDL, UNSPEC_UABDL.
23131 2021-02-01 Martin Sebor <msebor@redhat.com>
23133 * tree.h (BLOCK_VARS): Add comment.
23134 (BLOCK_SUBBLOCKS): Same.
23135 (BLOCK_SUPERCONTEXT): Same.
23136 (BLOCK_ABSTRACT_ORIGIN): Same.
23137 (inlined_function_outer_scope_p): Same.
23139 2021-02-01 Martin Sebor <msebor@redhat.com>
23141 PR middle-end/97172
23142 * attribs.c (attr_access::free_lang_data): Define new function.
23143 * attribs.h (attr_access::free_lang_data): Declare new function.
23145 2021-02-01 Richard Biener <rguenther@suse.de>
23147 * vec.h (auto_vec::auto_vec): Add memory stat parameters
23149 * bitmap.h (auto_bitmap::auto_bitmap): Likewise.
23151 2021-02-01 Tamar Christina <tamar.christina@arm.com>
23153 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>,
23154 aarch64_<su>mlsl<mode>, aarch64_<su>mlsl_n<mode>): Flip mult operands.
23156 2021-02-01 Richard Biener <rguenther@suse.de>
23158 PR rtl-optimization/98863
23159 * config/i386/i386-features.c (convert_scalars_to_vector):
23160 Set DF_RD_PRUNE_DEAD_DEFS.
23162 2021-01-31 Eric Botcazou <ebotcazou@adacore.com>
23164 * system.h (SIZE_MAX): Define if not already defined.
23166 2021-01-30 Aaron Sawdey <acsawdey@linux.ibm.com>
23168 * config/rs6000/genfusion.pl (gen_2logical): New function to
23169 generate patterns for logical-logical fusion.
23170 * config/rs6000/fusion.md: Regenerated patterns.
23171 * config/rs6000/rs6000-cpus.def: Add
23172 OPTION_MASK_P10_FUSION_2LOGICAL.
23173 * config/rs6000/rs6000.c (rs6000_option_override_internal):
23174 Enable logical-logical fusion for p10.
23175 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2logical.
23177 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
23179 * config/rs6000/rs6000.opt: Add periods to new AIX options.
23181 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
23183 * config/rs6000/rs6000.opt (mabi=vec-extabi): New.
23184 (mabi=vec-default): New.
23185 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
23186 __EXTABI__ for AIX Vector extended ABI.
23187 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print AIX Vector
23189 (conditional_register_usage): If AIX vec_extabi enabled, vs20-vs31
23191 * doc/invoke.texi (PowerPC mabi): Add AIX vec-extabi and vec-default.
23193 2021-01-30 Jakub Jelinek <jakub@redhat.com>
23195 * config/i386/i386-features.c (remove_partial_avx_dependency): Clear
23196 DF_DEFER_INSN_RESCAN after calling df_process_deferred_rescans.
23198 2021-01-29 Vladimir N. Makarov <vmakarov@redhat.com>
23201 * lra-constraints.c (in_class_p): Don't narrow class only for REG
23204 2021-01-29 Will Schmidt <will_schmidt@vnet.ibm.com>
23206 * config/rs6000/rs6000-call.c (rs6000_expand_binup_builtin): Add
23207 clauses for CODE_FOR_vsx_xvcvuxddp_scale and
23208 CODE_FOR_vsx_xvcvsxddp_scale to the parameter checking code.
23210 2021-01-29 Andrew MacLeod <amacleod@redhat.com>
23212 PR tree-optimization/98866
23213 * gimple-range-gori.h (gori_compute:set_range_invariant): New.
23214 * gimple-range-gori.cc (gori_map::set_range_invariant): New.
23215 (gori_map::m_maybe_invariant): Rename from all_outgoing.
23216 (gori_map::gori_map): Rename all_outgoing to m_maybe_invariant.
23217 (gori_map::is_export_p): Ditto.
23218 (gori_map::calculate_gori): Ditto.
23219 (gori_compute::set_range_invariant): New.
23220 * gimple-range.cc (gimple_ranger::range_of_stmt): Set range
23221 invariant for pointers evaluating to [1, +INF].
23223 2021-01-29 Richard Biener <rguenther@suse.de>
23225 PR rtl-optimization/98863
23226 * config/i386/i386-features.c (remove_partial_avx_dependency):
23227 Do not perform DF analysis.
23228 (pass_data_remove_partial_avx_dependency): Remove
23231 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
23233 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_n
23234 builtin generator macros.
23235 * config/aarch64/aarch64-simd.md (aarch64_<su>mull_n<mode>):
23237 * config/aarch64/arm_neon.h (vmull_n_s16): Use RTL builtin
23238 instead of inline asm.
23239 (vmull_n_s32): Likewise.
23240 (vmull_n_u16): Likewise.
23241 (vmull_n_u32): Likewise.
23243 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23245 * config/aarch64/aarch64-simd-builtins.def (sabdl2, uabdl2):
23247 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
23249 (aarch64_<sur>abdl2<mode>): ... This.
23250 (<sur>sadv16qi): Adjust use of above.
23251 * config/aarch64/arm_neon.h (vabdl_high_s8): Reimplement using
23253 (vabdl_high_s16): Likewise.
23254 (vabdl_high_s32): Likewise.
23255 (vabdl_high_u8): Likewise.
23256 (vabdl_high_u16): Likewise.
23257 (vabdl_high_u32): Likewise.
23259 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23261 * config/aarch64/aarch64-simd-builtins.def (sabal2): Define
23263 (uabal2): Likewise.
23264 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal2<mode>): New
23266 * config/aarch64/aarch64.md (unspec): Add UNSPEC_SABAL2 and
23268 * config/aarch64/arm_neon.h (vabal_high_s8): Reimplement using
23270 (vabal_high_s16): Likewise.
23271 (vabal_high_s32): Likewise.
23272 (vabal_high_u8): Likewise.
23273 (vabal_high_u16): Likewise.
23274 (vabal_high_u32): Likewise.
23275 * config/aarch64/iterators.md (ABAL2): New mode iterator.
23276 (sur): Handle UNSPEC_SABAL2, UNSPEC_UABAL2.
23278 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23280 * config/aarch64/aarch64-simd-builtins.def (sabal): Define
23283 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal<mode>_4):
23285 (aarch64_<sur>abal<mode>): ... This
23286 (<sur>sadv16qi): Adust use of the above.
23287 * config/aarch64/arm_neon.h (vabal_s8): Reimplement using
23289 (vabal_s16): Likewise.
23290 (vabal_s32): Likewise.
23291 (vabal_u8): Likewise.
23292 (vabal_u16): Likewise.
23293 (vabal_u32): Likewise.
23295 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23297 * config/aarch64/aarch64-simd-builtins.def (saddlv, uaddlv):
23299 * config/aarch64/aarch64-simd.md (aarch64_<su>addlv<mode>):
23301 * config/aarch64/arm_neon.h (vaddlv_s8): Reimplement using
23303 (vaddlv_s16): Likewise.
23304 (vaddlv_u8): Likewise.
23305 (vaddlv_u16): Likewise.
23306 (vaddlvq_s8): Likewise.
23307 (vaddlvq_s16): Likewise.
23308 (vaddlvq_s32): Likewise.
23309 (vaddlvq_u8): Likewise.
23310 (vaddlvq_u16): Likewise.
23311 (vaddlvq_u32): Likewise.
23312 (vaddlv_s32): Likewise.
23313 (vaddlv_u32): Likewise.
23314 * config/aarch64/iterators.md (VDQV_L): New mode iterator.
23315 (unspec): Add UNSPEC_SADDLV, UNSPEC_UADDLV.
23316 (Vwstype): New mode attribute.
23318 (VWIDE_S): Likewise.
23319 (USADDLV): New int iterator.
23320 (su): Handle UNSPEC_SADDLV, UNSPEC_UADDLV.
23322 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
23324 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_lane[q]
23325 builtin generator macros.
23326 * config/aarch64/aarch64-simd.md (aarch64_vec_<su>mlsl_lane<Qlane>):
23328 * config/aarch64/arm_neon.h (vmlsl_lane_s16): Use RTL builtin
23329 instead of inline asm.
23330 (vmlsl_lane_s32): Likewise.
23331 (vmlsl_lane_u16): Likewise.
23332 (vmlsl_lane_u32): Likewise.
23333 (vmlsl_laneq_s16): Likewise.
23334 (vmlsl_laneq_s32): Likewise.
23335 (vmlsl_laneq_u16): Likewise.
23336 (vmlsl_laneq_u32): Likewise.
23338 2021-01-29 Richard Biener <rguenther@suse.de>
23340 * doc/invoke.texi (--param max-gcse-memory): Document unit
23342 * gcse.c (gcse_or_cprop_is_too_expensive): Adjust.
23343 * params.opt (--param max-gcse-memory): Adjust default and
23344 document unit of size.
23346 2021-01-29 Richard Biener <rguenther@suse.de>
23348 PR rtl-optimization/98863
23349 * gcse.c (gcse_or_cprop_is_too_expensive): Use unsigned
23350 HOST_WIDE_INT for the memory estimate.
23352 2021-01-29 Bin Cheng <bin.cheng@linux.alibaba.com>
23353 Richard Biener <rguenther@suse.de>
23355 PR tree-optimization/97627
23356 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
23357 Do not analyze fake edges.
23359 2021-01-29 Richard Biener <rguenther@suse.de>
23361 PR rtl-optimization/98144
23362 * df.h (df_mir_bb_info): Add con_visited member.
23363 * df-problems.c (df_mir_alloc): Initialize con_visited,
23364 do not fully populate IN and OUT.
23365 (df_mir_reset): Likewise.
23366 (df_mir_confluence_0): Set con_visited.
23367 (df_mir_confluence_n): Properly handle implicitely
23368 fully populated IN and OUT as designated by con_visited
23369 and update con_visited accordingly.
23371 2021-01-29 Jakub Jelinek <jakub@redhat.com>
23374 * config/arm/vec-common.md (mve_vshlq_<supf><mode>,
23375 vashl<mode>3, vashr<mode>3, vlshr<mode>3): Add
23376 && !TARGET_REALLY_IWMMXT to conditions.
23378 2021-01-29 Jakub Jelinek <jakub@redhat.com>
23381 * cfgbuild.c (find_bb_boundaries): Reset debug_insn when seeing
23384 2021-01-28 Marek Polacek <polacek@redhat.com>
23387 * stor-layout.c (finalize_type_size): If we reset TYPE_USER_ALIGN in
23388 the main variant, maybe reset it in its variants too.
23389 * tree.c (check_base_type): Return true only if TYPE_USER_ALIGN match.
23390 (check_aligned_type): Check if TYPE_USER_ALIGN match.
23392 2021-01-28 Christophe Lyon <christophe.lyon@linaro.org>
23395 * config/arm/arm.c (arm_rtx_costs_internal): Adjust cost of vector
23396 of constant zero for comparisons.
23398 2021-01-28 Michael Meissner <meissner@linux.ibm.com>
23400 * config/rs6000/rs6000.c (rs6000_mangle_decl_assembler_name): Add
23401 support for mapping built-in function names for long double
23402 built-in functions if long double is IEEE 128-bit.
23404 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
23406 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_n
23407 builtin generator macros.
23408 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_n<mode>):
23410 * config/aarch64/arm_neon.h (vmlsl_n_s16): Use RTL builtin
23411 instead of inline asm.
23412 (vmlsl_n_s32): Likewise.
23413 (vmlsl_n_u16): Likewise.
23414 (vmlsl_n_u32): Likewise.
23416 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
23418 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_n
23419 builtin generator macros.
23420 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>):
23422 * config/aarch64/arm_neon.h (vmlal_n_s16): Use RTL builtin
23423 instead of inline asm.
23424 (vmlal_n_s32): Likewise.
23425 (vmlal_n_u16): Likewise.
23426 (vmlal_n_u32): Likewise.
23428 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23430 * config/aarch64/aarch64-simd-builtins.def (shrn2): Define
23432 * config/aarch64/aarch64-simd.md (aarch64_shrn2<mode>_insn_le):
23434 (aarch64_shrn2<mode>_insn_be): Likewise.
23435 (aarch64_shrn2<mode>): Likewise.
23436 * config/aarch64/arm_neon.h (vshrn_high_n_s16): Reimlplement
23438 (vshrn_high_n_s32): Likewise.
23439 (vshrn_high_n_s64): Likewise.
23440 (vshrn_high_n_u16): Likewise.
23441 (vshrn_high_n_u32): Likewise.
23442 (vshrn_high_n_u64): Likewise.
23444 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23446 * config/aarch64/aarch64-simd-builtins.def (shrn): Define
23448 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le):
23450 (aarch64_shrn<mode>_insn_be): Likewise.
23451 (aarch64_shrn<mode>): Likewise.
23452 * config/aarch64/arm_neon.h (vshrn_n_s16): Reimplement using
23454 (vshrn_n_s32): Likewise.
23455 (vshrn_n_s64): Likewise.
23456 (vshrn_n_u16): Likewise.
23457 (vshrn_n_u32): Likewise.
23458 (vshrn_n_u64): Likewise.
23459 * config/aarch64/iterators.md (vn_mode): New mode attribute.
23461 2021-01-28 Richard Biener <rguenther@suse.de>
23463 PR rtl-optimization/80960
23464 * dse.c (check_mem_read_rtx): Call get_addr on the
23467 2021-01-28 Xionghu Luo <luoxhu@linux.ibm.com>
23468 David Edelsohn <dje.gcc@gmail.com>
23471 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
23472 Don't generate VIEW_CONVERT_EXPR for fcode ALTIVEC_BUILTIN_VEC_INSERT
23474 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
23476 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Remove the
23477 wrapper call rs6000_expand_vector_set_var for cleanup. Call
23478 rs6000_expand_vector_set_var_p9 and rs6000_expand_vector_set_var_p8
23480 (rs6000_expand_vector_set_var): Delete.
23481 (rs6000_expand_vector_set_var_p9): Make static.
23482 (rs6000_expand_vector_set_var_p8): Make static.
23484 2021-01-28 Xing GUO <higuoxing@gmail.com>
23486 * common/config/riscv/riscv-common.c
23487 (riscv_subset_list::parsing_subset_version): Fix -march option parsing
23488 when `p` extension exists.
23490 2021-01-27 Vladimir N. Makarov <vmakarov@redhat.com>
23492 PR rtl-optimization/97684
23493 * ira.c (ira): Call ira_set_pseudo_classes before
23494 update_equiv_regs when it is necessary.
23496 2021-01-27 Jakub Jelinek <jakub@redhat.com>
23499 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Use
23500 %w0, %w1 and %2 instead of %0, %1 and %2.
23502 2021-01-27 Aaron Sawdey <acsawdey@linux.ibm.com>
23504 * config/rs6000/genfusion.pl: New script to generate
23505 define_insn_and_split patterns so combine can arrange fused
23506 instructions next to each other.
23507 * config/rs6000/fusion.md: New file, generated fused instruction
23508 patterns for combine.
23509 * config/rs6000/predicates.md (const_m1_to_1_operand): New predicate.
23510 (non_update_memory_operand): New predicate.
23511 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_P10_FUSION and
23512 OPTION_MASK_P10_FUSION_LD_CMPI to ISA_3_1_MASKS_SERVER and
23514 * config/rs6000/rs6000-protos.h (address_is_non_pfx_d_or_x): Add
23516 * config/rs6000/rs6000.c (rs6000_option_override_internal):
23517 Automatically set OPTION_MASK_P10_FUSION and
23518 OPTION_MASK_P10_FUSION_LD_CMPI if target is power10.
23519 (rs600_opt_masks): Allow -mpower10-fusion
23520 in function attributes.
23521 (address_is_non_pfx_d_or_x): New function.
23522 * config/rs6000/rs6000.h: Add MASK_P10_FUSION.
23523 * config/rs6000/rs6000.md: Include fusion.md.
23524 * config/rs6000/rs6000.opt: Add -mpower10-fusion
23525 and -mpower10-fusion-ld-cmpi.
23526 * config/rs6000/t-rs6000: Add dependencies involving fusion.md.
23528 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
23530 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal
23531 builtin generator macros.
23532 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal<mode>):
23534 (aarch64_<su>mlal<mode>): This.
23535 * config/aarch64/arm_neon.h (vmlal_s8): Use RTL builtin
23536 instead of inline asm.
23537 (vmlal_s16): Likewise.
23538 (vmlal_s32): Likewise.
23539 (vmlal_u8): Likewise.
23540 (vmlal_u16): Likewise.
23541 (vmlal_u32): Likewise.
23543 2021-01-27 Richard Biener <rguenther@suse.de>
23545 PR tree-optimization/98854
23546 * tree-vect-slp.c (vect_build_slp_tree_2): Also build
23547 PHIs from scalars when the number of CTORs matches the
23548 number of children.
23550 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
23552 * config/aarch64/aarch64-simd-builtins.def: Add mls_n builtin
23554 * config/aarch64/aarch64-simd.md (*aarch64_mls_elt_merge<mode>):
23556 (aarch64_mls_n<mode>): This.
23557 * config/aarch64/arm_neon.h (vmls_n_s16): Use RTL builtin
23559 (vmls_n_s32): Likewise.
23560 (vmls_n_u16): Likewise.
23561 (vmls_n_u32): Likewise.
23562 (vmlsq_n_s16): Likewise.
23563 (vmlsq_n_s32): Likewise.
23564 (vmlsq_n_u16): Likewise.
23565 (vmlsq_n_u32): Likewise.
23567 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
23569 * config/aarch64/aarch64-simd-builtins.def: Add mls builtin
23571 * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather
23573 (vmls_s16): Likewise.
23574 (vmls_s32): Likewise.
23575 (vmls_u8): Likewise.
23576 (vmls_u16): Likewise.
23577 (vmls_u32): Likewise.
23578 (vmlsq_s8): Likewise.
23579 (vmlsq_s16): Likewise.
23580 (vmlsq_s32): Likewise.
23581 (vmlsq_u8): Likewise.
23582 (vmlsq_u16): Likewise.
23583 (vmlsq_u32): Likewise.
23585 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
23587 * config/aarch64/aarch64-simd-builtins.def: Add mla_n builtin
23589 * config/aarch64/aarch64-simd.md (*aarch64_mla_elt_merge<mode>):
23591 (aarch64_mla_n<mode>): This.
23592 * config/aarch64/arm_neon.h (vmla_n_s16): Use RTL builtin
23594 (vmla_n_s32): Likewise.
23595 (vmla_n_u16): Likewise.
23596 (vmla_n_u32): Likewise.
23597 (vmlaq_n_s16): Likewise.
23598 (vmlaq_n_s32): Likewise.
23599 (vmlaq_n_u16): Likewise.
23600 (vmlaq_n_u32): Likewise.
23602 2021-01-27 liuhongt <hongtao.liu@intel.com>
23605 * config/i386/sse.md (sse2_gt<mode>3): Drop !TARGET_XOP in condition.
23606 (*sse2_eq<mode>3): Ditto.
23608 2021-01-27 Jakub Jelinek <jakub@redhat.com>
23610 * tree-pass.h (PROP_trees): Rename to ...
23611 (PROP_gimple): ... this.
23612 * cfgexpand.c (pass_data_expand): Replace PROP_trees with PROP_gimple.
23613 * passes.c (execute_function_dump, execute_function_todo,
23614 execute_one_ipa_transform_pass, execute_one_pass): Likewise.
23615 * varpool.c (ctor_for_folding): Likewise.
23617 2021-01-27 Jakub Jelinek <jakub@redhat.com>
23619 PR tree-optimization/97260
23620 * varpool.c: Include tree-pass.h.
23621 (ctor_for_folding): In GENERIC return DECL_INITIAL for TREE_READONLY
23622 non-TREE_SIDE_EFFECTS automatic variables.
23624 2021-01-26 Paul Fee <paul.f.fee@gmail.com>
23626 * doc/cpp.texi (__cplusplus): Document value for -std=c++23
23628 * doc/invoke.texi: Document -std=c++23 and -std=gnu++23.
23629 * dwarf2out.c (highest_c_language): Recognise C++20 and C++23.
23630 (gen_compile_unit_die): Recognise C++23.
23632 2021-01-26 Jakub Jelinek <jakub@redhat.com>
23635 * dwarf2asm.c (dw2_assemble_integer): Cast DWARF2_ADDR_SIZE to int
23638 2021-01-26 Jakub Jelinek <jakub@redhat.com>
23641 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
23642 Use UINTVAL (shft_amnt) and UINTVAL (mask) instead of INTVAL (shft_amnt)
23643 and INTVAL (mask). Add && INTVAL (mask) > 0 condition.
23645 2021-01-26 Richard Biener <rguenther@suse.de>
23647 * gimple-pretty-print.c (dump_binary_rhs): Handle
23648 VEC_WIDEN_{PLUS,MINUS}_{LO,HI}_EXPR.
23650 2021-01-26 Richard Biener <rguenther@suse.de>
23652 PR middle-end/98726
23653 * tree.h (vector_cst_int_elt): Remove.
23654 * tree.c (vector_cst_int_elt): Use poly_wide_int for computations,
23657 2021-01-26 Andrew Stubbs <ams@codesourcery.com>
23659 * config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
23660 for V64DFmode min/max reductions.
23662 2021-01-26 Jakub Jelinek <jakub@redhat.com>
23664 * dwarf2asm.c (dw2_assemble_integer): Handle size twice as large
23665 as DWARF2_ADDR_SIZE if x is not a scalar int by emitting it as
23666 two halves, one with x and the other with const0_rtx, ordered
23667 depending on endianity.
23669 2021-01-26 Alexandre Oliva <oliva@adacore.com>
23671 * gimplify.c (gimplify_decl_expr): Skip asan marking calls for
23672 temporaries not seen in binding block, and not about to be
23673 added as gimple variables.
23675 2021-01-25 Martin Sebor <msebor@redhat.com>
23678 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust warning text.
23680 2021-01-25 Martin Liska <mliska@suse.cz>
23682 * value-prof.c (get_nth_most_common_value): Use %s instead
23685 2021-01-25 Jakub Jelinek <jakub@redhat.com>
23688 * configure.ac (HAVE_AS_GDWARF_5_DEBUG_FLAG): Only define if
23689 readelf -wi is able to read the emitted .debug_info back.
23690 * configure: Regenerated.
23692 2021-01-25 Martin Liska <mliska@suse.cz>
23694 PR gcov-profile/98739
23695 * common.opt: Add missing sign symbol.
23696 * value-prof.c (get_nth_most_common_value): Restore handling
23697 of PROFILE_REPRODUCIBILITY_PARALLEL_RUNS and
23698 PROFILE_REPRODUCIBILITY_MULTITHREADED.
23700 2021-01-25 Richard Biener <rguenther@suse.de>
23702 PR middle-end/98807
23703 * tree.c (vector_element_bits): Always use precision of
23704 the element type for boolean vectors.
23706 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
23708 * config/rtems.h (STARTFILE_SPEC): Remove qnolinkcmds.
23709 (ENDFILE_SPEC): Evaluate qnolinkcmds.
23711 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
23713 * config/rtems.h (STARTFILE_SPEC): Remove nostdlib and
23714 nostartfiles handling since this is already done by
23715 LINK_COMMAND_SPEC. Evaluate qnolinkcmds.
23716 (ENDFILE_SPEC): Remove nostdlib and nostartfiles handling since this
23717 is already done by LINK_COMMAND_SPEC.
23718 (LIB_SPECS): Remove nostdlib and nodefaultlibs handling since
23719 this is already done by LINK_COMMAND_SPEC. Remove qnolinkcmds
23722 2021-01-25 Jakub Jelinek <jakub@redhat.com>
23725 * fold-const-call.c (host_size_t_cst_p): Renamed to ...
23726 (size_t_cst_p): ... this. Check and store unsigned HOST_WIDE_INT
23727 value rather than host size_t.
23728 (fold_const_call): Change type of s2 from size_t to
23729 unsigned HOST_WIDE_INT. Use size_t_cst_p instead of
23730 host_size_t_cst_p. For strncmp calls, pass MIN (s2, SIZE_MAX)
23731 instead of s2 as last argument.
23733 2021-01-25 Tamar Christina <tamar.christina@arm.com>
23735 * config/arm/iterators.md (rotsplit1, rotsplit2, conj_op, fcmac1,
23736 VCMLA_OP, VCMUL_OP): New.
23737 * config/arm/mve.md (mve_vcmlaq<mve_rot><mode>): Support vec_dup 0.
23738 * config/arm/neon.md (cmul<conj_op><mode>3): New.
23739 * config/arm/unspecs.md (UNSPEC_VCMLA_CONJ, UNSPEC_VCMLA180_CONJ,
23740 UNSPEC_VCMUL_CONJ): New.
23741 * config/arm/vec-common.md (cmul<conj_op><mode>3, arm_vcmla<rot><mode>,
23742 cml<fcmac1><conj_op><mode>4): New.
23744 2021-01-23 Jakub Jelinek <jakub@redhat.com>
23747 * config/rs6000/mmintrin.h (__m64): Add __may_alias__ attribute.
23749 2021-01-22 Jonathan Wright <jonathan.wright@arm.com>
23751 * config/aarch64/aarch64-simd-builtins.def: Add mla builtin
23753 * config/aarch64/arm_neon.h (vmla_s8): Use RTL builtin rather
23755 (vmla_s16): Likewise.
23756 (vmla_s32): Likewise.
23757 (vmla_u8): Likewise.
23758 (vmla_u16): Likewise.
23759 (vmla_u32): Likewise.
23760 (vmlaq_s8): Likewise.
23761 (vmlaq_s16): Likewise.
23762 (vmlaq_s32): Likewise.
23763 (vmlaq_u8): Likewise.
23764 (vmlaq_u16): Likewise.
23765 (vmlaq_u32): Likewise.
23767 2021-01-22 David Malcolm <dmalcolm@redhat.com>
23769 * doc/invoke.texi (GCC_EXTRA_DIAGNOSTIC_OUTPUT): Add @findex
23772 2021-01-22 Jakub Jelinek <jakub@redhat.com>
23775 * dwarf2out.c (output_file_names): For -gdwarf-5, if there are no
23776 filenames to emit, still emit the required 0 index directory and
23777 filename entries that match DW_AT_comp_dir and DW_AT_name of the
23780 2021-01-22 Marek Polacek <polacek@redhat.com>
23783 * doc/invoke.texi: Update C++ ABI Version 15 description.
23785 2021-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23787 PR tree-optimization/98766
23788 * tree-ssa-math-opts.c (convert_mult_to_fma): Use maybe_le when
23789 comparing against type size with param_avoid_fma_max_bits.
23791 2021-01-22 Richard Biener <rguenther@suse.de>
23793 PR middle-end/98793
23794 * tree.c (vector_element_bits): Key single-bit bool vector on
23795 integer mode rather than not vector mode.
23797 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
23800 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
23801 Generate ARRAY_REF(VIEW_CONVERT_EXPR) for P8 and later
23803 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): Update
23804 to call different path for P8 and P9.
23805 (rs6000_expand_vector_set_var_p9): New function.
23806 (rs6000_expand_vector_set_var_p8): New function.
23808 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
23812 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
23813 Ajdust variable index vec_insert from address dereference to
23814 ARRAY_REF(VIEW_CONVERT_EXPR) tree expression.
23815 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
23817 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): New function.
23819 2021-01-22 Martin Liska <mliska@suse.cz>
23821 PR gcov-profile/98739
23822 * profile.c (compute_value_histograms): Drop time profile for
23823 -fprofile-reproducible=multithreaded.
23825 2021-01-22 Nathan Sidwell <nathan@acm.org>
23827 * gcc.c (process_command): Don't check OPT_SPECIAL_input_file
23830 2021-01-22 Richard Biener <rguenther@suse.de>
23832 PR middle-end/98773
23833 * tree-data-ref.c (initalize_matrix_A): Revert previous
23834 change, retaining failing on HOST_WIDE_INT_MIN CHREC_RIGHT.
23836 2021-01-22 Jakub Jelinek <jakub@redhat.com>
23838 PR tree-optimization/90248
23839 * match.pd (X cmp 0.0 ? 1.0 : -1.0 -> copysign(1, +-X),
23840 X cmp 0.0 ? -1.0 : +1.0 -> copysign(1, -+X)): Remove
23842 (X * (X cmp 0.0 ? 1.0 : -1.0) -> +-abs(X),
23843 X * (X cmp 0.0 ? -1.0 : 1.0) -> +-abs(X)): New simplifications.
23845 2021-01-22 Jakub Jelinek <jakub@redhat.com>
23847 PR tree-optimization/98255
23848 * tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
23849 extend index - low_bound from sizetype's precision rather than index
23851 (get_addr_base_and_unit_offset_1): Likewise.
23852 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
23853 * gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
23855 2021-01-22 Richard Biener <rguenther@suse.de>
23857 PR tree-optimization/98786
23858 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Avoid
23859 adding new uses of abnormals. Verify we deal with a conditional
23862 2021-01-22 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23865 * optc-save-gen.awk: Add arm_fp16_format to checked_options.
23867 2021-01-22 liuhongt <hongtao.liu@intel.com>
23871 * config/i386/sse.md (VI_128_256): New mode iterator.
23872 (*avx_cmp<mode>3_1, *avx_cmp<mode>3_2, *avx_cmp<mode>3_3,
23873 *avx_cmp<mode>3_4, *avx2_eq<mode>3, *avx2_pcmp<mode>3_1,
23874 *avx2_pcmp<mode>3_2, *avx2_gt<mode>3): New
23875 define_insn_and_split to lower avx512 vector comparison to avx
23876 version when dest is vector.
23877 (*<avx512>_cmp<mode>3,*<avx512>_cmp<mode>3,*<avx512>_ucmp<mode>3):
23878 define_insn_and_split for negating the comparison result.
23879 * config/i386/predicates.md (float_vector_all_ones_operand):
23881 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Use
23882 general NOT operator without UNSPEC_MASKOP.
23884 2021-01-21 Vladimir N. Makarov <vmakarov@redhat.com>
23886 PR rtl-optimization/98777
23887 * lra-int.h (lra_pmode_pseudo): New extern.
23888 * lra.c (lra_pmode_pseudo): New global.
23890 * lra-eliminations.c (eliminate_regs_in_insn): Use it.
23892 2021-01-21 Ilya Leoshkevich <iii@linux.ibm.com>
23894 * fwprop.c (fwprop_propagation::classify_result): Allow
23895 (subreg (mem)) simplifications.
23897 2021-01-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23899 * config/aarch64/aarch64-simd.md (aarch64_sqdml<SBINQOPS:as>l<mode>):
23901 (aarch64_sqdmlal<mode>): ... This...
23902 (aarch64_sqdmlsl<mode>): ... And this.
23903 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Split into...
23904 (aarch64_sqdmlal_lane<mode>): ... This...
23905 (aarch64_sqdmlsl_lane<mode>): ... And this.
23906 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Split into...
23907 (aarch64_sqdmlsl_laneq<mode>): ... This...
23908 (aarch64_sqdmlal_laneq<mode>): ... And this.
23909 (aarch64_sqdml<SBINQOPS:as>l_n<mode>): Split into...
23910 (aarch64_sqdmlsl_n<mode>): ... This...
23911 (aarch64_sqdmlal_n<mode>): ... And this.
23912 (aarch64_sqdml<SBINQOPS:as>l2<mode>_internal): Split into...
23913 (aarch64_sqdmlal2<mode>_internal): ... This...
23914 (aarch64_sqdmlsl2<mode>_internal): ... And this.
23916 2021-01-21 Christophe Lyon <christophe.lyon@linaro.org>
23918 * config/arm/arm_mve.h (__arm_vcmpneq_s8): Fix return type.
23920 2021-01-21 Andrea Corallo <andrea.corallo@arm.com>
23923 * doc/sourcebuild.texi (arm_thumb2_no_arm_v8_1_lob): Document.
23925 2021-01-21 liuhongt <hongtao.liu@intel.com>
23927 PR rtl-optimization/98694
23928 * regcprop.c (copy_value): If SRC had been assigned a mode
23929 narrower than the copy, we can't link DEST into the chain even
23930 they have same hard_regno_nregs(i.e. HImode/SImode in i386
23933 2021-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23935 * config/aarch64/aarch64-simd.md (aarch64_get_lane<mode>):
23936 Convert to define_insn_and_split. Split into simple move when moving
23939 2021-01-20 Segher Boessenkool <segher@kernel.crashing.org>
23941 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Change assert.
23942 Adjust comment. Simplify code.
23944 2021-01-20 Jakub Jelinek <jakub@redhat.com>
23947 * dwarf2out.c (reset_indirect_string): Also reset indirect strings
23948 with DW_FORM_line_strp form.
23949 (prune_unused_types_update_strings): Don't add into debug_str_hash
23950 indirect strings with DW_FORM_line_strp form.
23951 (adjust_name_comp_dir): New function.
23952 (dwarf2out_finish): Call it on CU DIEs after resetting
23953 debug_line_str_hash.
23955 2021-01-20 Vladimir N. Makarov <vmakarov@redhat.com>
23957 PR rtl-optimization/98722
23958 * lra-eliminations.c (eliminate_regs_in_insn): Check that target
23959 has no 3-op add insn to transform insns containing two pluses.
23961 2021-01-20 Richard Biener <rguenther@suse.de>
23963 * hwint.h (add_hwi): New function.
23964 (mul_hwi): Likewise.
23965 * tree-data-ref.c (initialize_matrix_A): Properly translate
23966 tree constants and avoid HOST_WIDE_INT_MIN.
23967 (lambda_matrix_row_add): Avoid undefined integer overflow
23968 and return true on such overflow.
23969 (lambda_matrix_right_hermite): Handle overflow from
23970 lambda_matrix_row_add gracefully. Simplify previous fix.
23971 (analyze_subscript_affine_affine): Likewise.
23973 2021-01-20 Eugene Rozenfeld <erozen@microsoft.com>
23975 PR tree-optimization/96674
23976 * match.pd: New patterns: x < y || y == XXX_MIN --> x <= y - 1
23977 x >= y && y != XXX_MIN --> x > y - 1
23979 2021-01-20 Richard Sandiford <richard.sandiford@arm.com>
23981 PR tree-optimization/98535
23982 * tree-vect-slp.c (duplicate_and_interleave): Use quick_grow_cleared.
23983 If the high and low permutes are the same, remove the high permutes
23984 from the working set and only continue with the low ones.
23986 2021-01-20 Jakub Jelinek <jakub@redhat.com>
23988 PR tree-optimization/98721
23989 * builtins.c (access_ref::inform_access): Don't assume
23990 SSA_NAME_IDENTIFIER must be non-NULL. Print messages about
23991 object whenever allocfn is NULL, rather than only when DECL_P
23992 is true. Use %qE instead of %qD for that. Formatting fixes.
23994 2021-01-20 Richard Biener <rguenther@suse.de>
23996 PR tree-optimization/98758
23997 * tree-data-ref.c (int_divides_p): Use lambda_int arguments.
23998 (lambda_matrix_right_hermite): Avoid undefinedness with
23999 signed integer abs and multiplication.
24000 (analyze_subscript_affine_affine): Use lambda_int.
24002 2021-01-20 David Malcolm <dmalcolm@redhat.com>
24005 * dwarf2out.c (output_line_info): Rename static variable
24006 "generation", moving it out of the function to...
24007 (output_line_info_generation): New.
24008 (init_sections_and_labels): Likewise, renaming the variable to...
24009 (init_sections_and_labels_generation): New.
24010 (dwarf2out_c_finalize): Reset the new variables.
24012 2021-01-19 Martin Sebor <msebor@redhat.com>
24014 PR middle-end/98664
24015 * tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
24016 all functions, even if they're not declared artificial or inline.
24017 * tree.c (tree_inlined_location): Use macro expansion location
24018 only if scope traversal fails to expose one.
24020 2021-01-19 Richard Sandiford <richard.sandiford@arm.com>
24022 PR rtl-optimization/92294
24023 * alias.c (compare_base_symbol_refs): Take an extra parameter
24024 and add the distance between two symbols to it. Enshrine in
24025 comments that -1 means "either 0 or 1, but we can't tell
24026 which at compile time".
24027 (memrefs_conflict_p): Update call accordingly.
24028 (rtx_equal_for_memref_p): Likewise. Take the distance between symbols
24031 2021-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24033 * config/aarch64/aarch64-simd-builtins.def (sqshl, uqshl,
24034 sqrshl, uqrshl, sqadd, uqadd, sqsub, uqsub, suqadd, usqadd, sqmovn,
24035 uqmovn, sqxtn2, uqxtn2, sqabs, sqneg, sqdmlal, sqdmlsl, sqdmlal_lane,
24036 sqdmlsl_lane, sqdmlal_laneq, sqdmlsl_laneq, sqdmlal_n, sqdmlsl_n,
24037 sqdmlal2, sqdmlsl2, sqdmlal2_lane, sqdmlsl2_lane, sqdmlal2_laneq,
24038 sqdmlsl2_laneq, sqdmlal2_n, sqdmlsl2_n, sqdmull, sqdmull_lane,
24039 sqdmull_laneq, sqdmull_n, sqdmull2, sqdmull2_lane, sqdmull2_laneq,
24040 sqdmull2_n, sqdmulh, sqrdmulh, sqdmulh_lane, sqdmulh_laneq,
24041 sqrdmulh_lane, sqrdmulh_laneq, sqshrun_n, sqrshrun_n, sqshrn_n,
24042 uqshrn_n, sqrshrn_n, uqrshrn_n, sqshlu_n, sqshl_n, uqshl_n, sqrdmlah,
24043 sqrdmlsh, sqrdmlah_lane, sqrdmlsh_lane, sqrdmlah_laneq, sqrdmlsh_laneq,
24044 sqmovun): Use NONE flags.
24046 2021-01-19 Richard Biener <rguenther@suse.de>
24049 * ipa-modref.c (analyze_stmt): Only record a summary for a
24052 2021-01-19 Richard Biener <rguenther@suse.de>
24054 PR middle-end/98638
24055 * tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
24057 2021-01-19 Daniel Hellstrom <daniel@gaisler.com>
24059 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
24060 built-in define __FIX_LEON3FT_TN0018.
24062 2021-01-19 Richard Biener <rguenther@suse.de>
24065 * tree-inline.c (tree_function_versioning): Set input_location
24066 to UNKNOWN_LOCATION throughout the function.
24068 2021-01-19 Tobias Burnus <tobias@codesourcery.com>
24071 * omp-low.c (lower_omp_target): Handle nonpointer is_device_ptr.
24073 2021-01-19 Martin Jambor <mjambor@suse.cz>
24076 * ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
24077 whether non-call exceptions allow removal of a statement.
24078 (isra_analyze_call): Pass the appropriate function to
24079 ssa_name_only_returned_p.
24081 2021-01-19 Geng Qi <gengqi@linux.alibaba.com>
24083 * config/riscv/arch-canonicalize (longext_sort): New function for
24084 sorting 'multi-letter'.
24085 * config/riscv/multilib-generator: Adjusting the loop of 'alt' in
24086 'alts'. The 'arch' may not be the first of 'alts'.
24087 (_expand_combination): Add underline for the 'ext' without '*'.
24088 This is because, a single-letter extension can always be treated well
24089 with a '_' prefix, but it cannot be separated out if it is appended
24092 2021-01-18 Vladimir N. Makarov <vmakarov@redhat.com>
24095 * ira.c (ira): Skip abnormal critical edge splitting.
24097 2021-01-18 Jakub Jelinek <jakub@redhat.com>
24099 PR tree-optimization/98727
24100 * tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
24101 second .MUL_OVERFLOW operand for signed multiplication with overflow
24102 checking if the second operand of multiplication is not constant.
24104 2021-01-18 David Edelsohn <dje.gcc@gmail.com>
24106 * doc/invoke.texi (-gdwarf): TPF defaults to version 2 and AIX
24107 defaults to version 4.
24109 2021-01-18 David Malcolm <dmalcolm@redhat.com>
24111 * attribs.h (fndecl_dealloc_argno): New decl.
24112 * builtins.c (call_dealloc_argno): Split out second half of
24114 (fndecl_dealloc_argno): New.
24115 * doc/extend.texi (Common Function Attributes): Document the
24116 interaction between the analyzer and the malloc attribute.
24117 * doc/invoke.texi (Static Analyzer Options): Likewise.
24119 2021-01-17 David Edelsohn <dje.gcc@gmail.com>
24121 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Override
24122 dwarf_version to 4.
24123 * config/rs6000/aix72.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
24125 2021-01-17 Martin Jambor <mjambor@suse.cz>
24128 * cgraph.c (clone_of_p): Check also former_clone_of as we climb
24131 2021-01-17 Mark Wielaard <mark@klomp.org>
24133 * common.opt (gdwarf-): Init(5).
24134 * doc/invoke.texi (-gdwarf): Document default to 5.
24136 2021-01-16 Kwok Cheung Yeung <kcy@codesourcery.com>
24138 * builtin-types.def
24139 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
24141 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
24142 ...this. Add extra argument.
24143 * gimplify.c (omp_default_clause): Ensure that event handle is
24144 firstprivate in a task region.
24145 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_DETACH.
24146 (gimplify_adjust_omp_clauses): Likewise.
24147 * omp-builtins.def (BUILT_IN_GOMP_TASK): Change function type to
24148 BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR.
24149 * omp-expand.c (expand_task_call): Add GOMP_TASK_FLAG_DETACH to flags
24150 if detach clause specified. Add detach argument when generating
24152 * omp-low.c (scan_sharing_clauses): Setup data environment for detach
24154 (finish_taskreg_scan): Move field for variable containing the event
24155 handle to the front of the struct.
24156 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_DETACH. Fix
24158 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
24159 OMP_CLAUSE_DETACH clause.
24160 (convert_local_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
24161 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_DETACH.
24162 * tree.c (omp_clause_num_ops): Add entry for OMP_CLAUSE_DETACH.
24164 (omp_clause_code_name): Add entry for OMP_CLAUSE_DETACH. Fix
24166 (walk_tree_1): Handle OMP_CLAUSE_DETACH.
24168 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
24170 * config/nios2/t-rtems: Reset all MULTILIB_* variables. Shorten
24171 multilib directory names. Use MULTILIB_REQUIRED instead of
24172 MULTILIB_EXCEPTIONS. Add -mhw-mul -mhw-mulx -mhw-div
24173 -mcustom-fpu-cfg=fph2 multilib.
24175 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
24177 * config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
24178 (nios2_init_fpu_configs): Provide register values for new
24179 -mcustom-fpu-cfg=fph2 option variant.
24180 * doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
24183 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
24185 * config/nios2/nios2.c (nios2_custom_check_insns): Remove
24186 custom instruction warnings.
24188 2021-01-16 Jakub Jelinek <jakub@redhat.com>
24190 PR tree-optimization/96669
24191 * match.pd ((CST << x) & 1 -> x == 0): New simplification.
24193 2021-01-16 Jakub Jelinek <jakub@redhat.com>
24195 PR tree-optimization/96271
24196 * passes.def: Pass false argument to first two pass_cd_dce
24197 instances and true to last instance. Add comment that
24198 last instance rewrites no longer addressed locals.
24199 * tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
24201 (pass_cd_dce::set_pass_param): New method.
24202 (pass_cd_dce::execute): Return TODO_update_address_taken from
24203 last cd_dce instance.
24205 2021-01-15 Carl Love <cel@us.ibm.com>
24207 * config/rs6000/altivec.h (vec_mulh, vec_div, vec_dive, vec_mod):
24209 * config/rs6000/altivec.md (VIlong): Move define to file vsx.md.
24210 * config/rs6000/rs6000-builtin.def (DIVES_V4SI, DIVES_V2DI,
24211 DIVEU_V4SI, DIVEU_V2DI, DIVS_V4SI, DIVS_V2DI, DIVU_V4SI,
24212 DIVU_V2DI, MODS_V2DI, MODS_V4SI, MODU_V2DI, MODU_V4SI,
24213 MULHS_V2DI, MULHS_V4SI, MULHU_V2DI, MULHU_V4SI, MULLD_V2DI):
24214 Add builtin define.
24215 (MULH, DIVE, MOD): Add new BU_P10_OVERLOAD_2 definitions.
24216 * config/rs6000/rs6000-call.c (VSX_BUILTIN_VEC_DIV,
24217 VSX_BUILTIN_VEC_DIVE, P10_BUILTIN_VEC_MOD, P10_BUILTIN_VEC_MULH):
24218 New overloaded definitions.
24219 (builtin_function_type) [P10V_BUILTIN_DIVEU_V4SI,
24220 P10V_BUILTIN_DIVEU_V2DI, P10V_BUILTIN_DIVU_V4SI,
24221 P10V_BUILTIN_DIVU_V2DI, P10V_BUILTIN_MODU_V2DI,
24222 P10V_BUILTIN_MODU_V4SI, P10V_BUILTIN_MULHU_V2DI,
24223 P10V_BUILTIN_MULHU_V4SI]: Add case
24224 statement for builtins.
24225 * config/rs6000/rs6000.md (bits): Add new attribute sizes V4SI, V2DI.
24226 * config/rs6000/vsx.md (VIlong): Moved from config/rs6000/altivec.md.
24227 (UNSPEC_VDIVES, UNSPEC_VDIVEU): New unspec definitions.
24228 (vsx_mul_v2di): Add if TARGET_POWER10 statement.
24229 (vsx_udiv_v2di): Add if TARGET_POWER10 statement.
24230 (dives_<mode>, diveu_<mode>, div<mode>3, uvdiv<mode>3,
24231 mods_<mode>, modu_<mode>, mulhs_<mode>, mulhu_<mode>, mulv2di3):
24232 Add define_insn, mode is VIlong.
24233 * doc/extend.texi (vec_mulh, vec_mul, vec_div, vec_dive, vec_mod):
24234 Add builtin descriptions.
24236 2021-01-15 Eric Botcazou <ebotcazou@adacore.com>
24238 * final.c (final_start_function_1): Reset force_source_line.
24240 2021-01-15 Jakub Jelinek <jakub@redhat.com>
24242 PR tree-optimization/96669
24243 * match.pd (((1 << A) & 1) != 0 -> A == 0,
24244 ((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
24245 possibly different power of two constants and to right shift too.
24247 2021-01-15 Jakub Jelinek <jakub@redhat.com>
24249 PR tree-optimization/96681
24250 * match.pd ((x < 0) ^ (y < 0) to (x ^ y) < 0): New simplification.
24251 ((x >= 0) ^ (y >= 0) to (x ^ y) < 0): Likewise.
24252 ((x < 0) ^ (y >= 0) to (x ^ y) >= 0): Likewise.
24253 ((x >= 0) ^ (y < 0) to (x ^ y) >= 0): Likewise.
24255 2021-01-15 Alexandre Oliva <oliva@adacore.com>
24257 * opts.c (gen_command_line_string): Exclude -dumpbase-ext.
24259 2021-01-15 Tamar Christina <tamar.christina@arm.com>
24261 * config/aarch64/aarch64-simd.md (cml<fcmac1><conj_op><mode>4,
24262 cmul<conj_op><mode>3): New.
24263 * config/aarch64/iterators.md (UNSPEC_FCMUL,
24264 UNSPEC_FCMUL180, UNSPEC_FCMLA_CONJ, UNSPEC_FCMLA180_CONJ,
24265 UNSPEC_CMLA_CONJ, UNSPEC_CMLA180_CONJ, UNSPEC_CMUL, UNSPEC_CMUL180,
24266 FCMLA_OP, FCMUL_OP, conj_op, rotsplit1, rotsplit2, fcmac1, sve_rot1,
24267 sve_rot2, SVE2_INT_CMLA_OP, SVE2_INT_CMUL_OP, SVE2_INT_CADD_OP): New.
24268 (rot): Add UNSPEC_FCMUL, UNSPEC_FCMUL180.
24269 (rot_op): Renamed to conj_op.
24270 * config/aarch64/aarch64-sve.md (cml<fcmac1><conj_op><mode>4,
24271 cmul<conj_op><mode>3): New.
24272 * config/aarch64/aarch64-sve2.md (cml<fcmac1><conj_op><mode>4,
24273 cmul<conj_op><mode>3): New.
24275 2021-01-15 David Malcolm <dmalcolm@redhat.com>
24279 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
24280 Escape the tempfile name when constructing the expected output.
24282 2021-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24284 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlsl_hi<mode>):
24286 (aarch64_<su>mlsl_hi<mode>): ... This.
24287 (aarch64_<su>mlsl_hi<mode>): Define.
24288 (*aarch64_<su>mlsl<mode): Rename to...
24289 (aarch64_<su>mlsl<mode): ... This.
24290 * config/aarch64/aarch64-simd-builtins.def (smlsl, umlsl,
24291 smlsl_hi, umlsl_hi): Define builtins.
24292 * config/aarch64/arm_neon.h (vmlsl_high_s8, vmlsl_high_s16,
24293 vmlsl_high_s32, vmlsl_high_u8, vmlsl_high_u16, vmlsl_high_u32,
24294 vmlsl_s8, vmlsl_s16, vmlsl_s32, vmlsl_u8,
24295 vmlsl_u16, vmlsl_u32): Reimplement with builtins.
24297 2021-01-15 Uroš Bizjak <ubizjak@gmail.com>
24299 * config/i386/i386-c.c (ix86_target_macros):
24300 Use cpp_define_formatted for __SIZEOF_FLOAT80__ definition.
24302 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
24305 * config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
24306 * Makefile.in (RTL_SSA_H): New variable.
24307 * config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
24308 * config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
24309 * config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
24311 * config/aarch64/aarch64-cc-fusion.cc: New file.
24313 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
24315 * recog.h (insn_change_watermark::~insn_change_watermark): Avoid
24316 calling cancel_changes for changes that no longer exist.
24318 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
24320 * rtl-ssa/functions.h (function_info::ref_defs): Rename to...
24321 (function_info::reg_defs): ...this.
24322 * rtl-ssa/member-fns.inl (function_info::ref_defs): Rename to...
24323 (function_info::reg_defs): ...this.
24325 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
24328 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
24330 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
24333 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
24336 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
24338 2021-01-15 Richard Biener <rguenther@suse.de>
24340 PR tree-optimization/96376
24341 * tree-vect-stmts.c (get_load_store_type): Disregard alignment
24342 for VMAT_INVARIANT.
24344 2021-01-15 Martin Liska <mliska@suse.cz>
24346 * doc/install.texi: Document that some tests need pytest module.
24347 * doc/sourcebuild.texi: Likewise.
24349 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
24352 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
24354 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
24356 * config/arm/mve.md (mve_vshrq_n_s<mode>_imm): New entry.
24357 (mve_vshrq_n_u<mode>_imm): Likewise.
24358 * config/arm/neon.md (vashr<mode>3, vlshr<mode>3): Move to ...
24359 * config/arm/vec-common.md: ... here.
24361 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
24363 * config/arm/mve.md (mve_vshlq_<supf><mode>): Move to
24365 * config/arm/neon.md (vashl<mode>3): Delete.
24366 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): New.
24367 (vasl<mode>3): New expander.
24369 2021-01-15 Richard Biener <rguenther@suse.de>
24371 PR tree-optimization/98685
24372 * tree-vect-slp.c (vect_schedule_slp_node): Refactor handling
24373 of vector extern defs.
24375 2021-01-14 David Malcolm <dmalcolm@redhat.com>
24378 * diagnostic.c (diagnostic_kind_text): Break out this array
24380 (diagnostic_build_prefix): ...here.
24381 (fancy_abort): Detect when diagnostic_initialize has not yet been
24382 called and fall back to a minimal implementation of printing the
24383 ICE, rather than segfaulting in internal_error.
24385 2021-01-14 David Malcolm <dmalcolm@redhat.com>
24387 * diagnostic.c (diagnostic_initialize): Eliminate
24388 parseable_fixits_p in favor of initializing extra_output_kind from
24389 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
24390 (convert_column_unit): New function, split out from...
24391 (diagnostic_converted_column): ...this.
24392 (print_parseable_fixits): Add "column_unit" and "tabstop" params.
24393 Use them to call convert_column_unit on the column values.
24394 (diagnostic_report_diagnostic): Eliminate conditional on
24395 parseable_fixits_p in favor of a switch statement on
24396 extra_output_kind, passing the appropriate values to the new
24397 params of print_parseable_fixits.
24398 (selftest::test_print_parseable_fixits_none): Update for new
24399 params of print_parseable_fixits.
24400 (selftest::test_print_parseable_fixits_insert): Likewise.
24401 (selftest::test_print_parseable_fixits_remove): Likewise.
24402 (selftest::test_print_parseable_fixits_replace): Likewise.
24403 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
24405 (selftest::diagnostic_c_tests): Call it.
24406 * diagnostic.h (enum diagnostics_extra_output_kind): New.
24407 (diagnostic_context::parseable_fixits_p): Delete field in favor
24409 (diagnostic_context::extra_output_kind): ...this new field.
24410 * doc/invoke.texi (Environment Variables): Add
24411 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
24412 * opts.c (common_handle_option): Update handling of
24413 OPT_fdiagnostics_parseable_fixits for change to diagnostic_context
24416 2021-01-14 Tamar Christina <tamar.christina@arm.com>
24418 * tree-vect-slp-patterns.c (class complex_operations_pattern,
24419 complex_operations_pattern::matches,
24420 complex_operations_pattern::recognize,
24421 complex_operations_pattern::build): New.
24422 (slp_patterns): Use it.
24424 2021-01-14 Tamar Christina <tamar.christina@arm.com>
24426 * internal-fn.def (COMPLEX_FMS, COMPLEX_FMS_CONJ): New.
24427 * optabs.def (cmls_optab, cmls_conj_optab): New.
24428 * doc/md.texi: Document them.
24429 * tree-vect-slp-patterns.c (class complex_fms_pattern,
24430 complex_fms_pattern::matches, complex_fms_pattern::recognize,
24431 complex_fms_pattern::build): New.
24433 2021-01-14 Tamar Christina <tamar.christina@arm.com>
24435 * internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
24436 * optabs.def (cmla_optab, cmla_conj_optab): New.
24437 * doc/md.texi: Document them.
24438 * tree-vect-slp-patterns.c (vect_match_call_p,
24439 class complex_fma_pattern, vect_slp_reset_pattern,
24440 complex_fma_pattern::matches, complex_fma_pattern::recognize,
24441 complex_fma_pattern::build): New.
24443 2021-01-14 Tamar Christina <tamar.christina@arm.com>
24445 * internal-fn.def (COMPLEX_MUL, COMPLEX_MUL_CONJ): New.
24446 * optabs.def (cmul_optab, cmul_conj_optab): New.
24447 * doc/md.texi: Document them.
24448 * tree-vect-slp-patterns.c (vect_match_call_complex_mla,
24449 vect_normalize_conj_loc, is_eq_or_top, vect_validate_multiplication,
24450 vect_build_combine_node, class complex_mul_pattern,
24451 complex_mul_pattern::matches, complex_mul_pattern::recognize,
24452 complex_mul_pattern::build): New.
24454 2021-01-14 Tamar Christina <tamar.christina@arm.com>
24456 * tree-vect-slp.c (optimize_load_redistribution_1): New.
24457 (optimize_load_redistribution, vect_is_slp_load_node): New.
24458 (vect_match_slp_patterns): Use it.
24460 2021-01-14 Tamar Christina <tamar.christina@arm.com>
24462 * tree-vect-slp-patterns.c (complex_add_pattern::build):
24465 2021-01-14 Thomas Schwinge <thomas@codesourcery.com>
24467 * config/gcn/mkoffload.c (main): Create an offload image only in
24468 64-bit configurations.
24470 2021-01-14 H.J. Lu <hjl.tools@gmail.com>
24473 * config/i386/i386-options.c (ix86_option_override_internal):
24474 Issue an error for -fcf-protection with CF_BRANCH when compiling
24475 for 32-bit non-TARGET_CMOV targets.
24477 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
24480 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p):
24481 Remove declaration and initialization of shadow variable "ret".
24482 (ix86_option_override_internal): Remove delcaration of
24483 shadow variable "i". Redeclare shadowed variable to unsigned.
24484 * common/config/i386/i386-common.c (pta_size): Redeclare to unsigned.
24485 * config/i386/i386-builtins.c (get_builtin_code_for_version):
24486 Update for redeclaration.
24487 * config/i386/i386.h (pta_size): Ditto.
24489 2021-01-14 Richard Biener <rguenther@suse.de>
24491 PR tree-optimization/98674
24492 * tree-data-ref.c (base_supports_access_fn_components_p): New.
24493 (initialize_data_dependence_relation): For two bases without
24494 possible access fns resort to type size equality when determining
24495 shape compatibility.
24497 2021-01-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24500 * config/arm/arm_neon.h: Replace calls to __builtin_vcge* by
24501 <=, >= operators in vcle and vcge intrinsics respectively.
24502 * config/arm/arm_neon_builtins.def: Remove entry for
24505 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
24508 * config/i386/i386-options.c (ix86_function_specific_save):
24509 Remove redundant assignment to opts->x_ix86_branch_cost.
24510 * config/i386/i386.c (ix86_prefetch_sse):
24511 Rename from x86_prefetch_sse. Update all uses.
24512 * config/i386/i386.h: Update for rename.
24513 * config/i386/i386-options.h: Ditto.
24515 2021-01-14 Jakub Jelinek <jakub@redhat.com>
24518 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3,
24519 *sse4_1_zero_extendv4hiv4si2_3, *sse4_1_zero_extendv2siv2di2_3):
24520 Use Bm instead of m for non-avx. Add isa attribute.
24522 2021-01-14 Jakub Jelinek <jakub@redhat.com>
24524 PR tree-optimization/96688
24525 * match.pd (~(X >> Y) -> ~X >> Y): New simplification if
24526 ~X can be simplified.
24528 2021-01-14 Richard Sandiford <richard.sandiford@arm.com>
24530 * tree-vect-stmts.c (vect_model_load_cost): Account for unused
24531 IFN_LOAD_LANES results.
24533 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24535 * config/aarch64/aarch64-simd.md (aarch64_<su>xtl<mode>):
24537 (aarch64_xtn<mode>): Likewise.
24538 * config/aarch64/aarch64-simd-builtins.def (sxtl, uxtl, xtn):
24541 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
24543 (vmovl_s16): Likewise.
24544 (vmovl_s32): Likewise.
24545 (vmovl_u8): Likewise.
24546 (vmovl_u16): Likewise.
24547 (vmovl_u32): Likewise.
24548 (vmovn_s16): Likewise.
24549 (vmovn_s32): Likewise.
24550 (vmovn_s64): Likewise.
24551 (vmovn_u16): Likewise.
24552 (vmovn_u32): Likewise.
24553 (vmovn_u64): Likewise.
24555 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24557 * config/aarch64/aarch64-simd.md (aarch64_<su>qxtn2<mode>_le):
24559 (aarch64_<su>qxtn2<mode>_be): Likewise.
24560 (aarch64_<su>qxtn2<mode>): Likewise.
24561 * config/aarch64/aarch64-simd-builtins.def (sqxtn2, uqxtn2):
24563 * config/aarch64/iterators.md (SAT_TRUNC): Define code_iterator.
24564 (su): Handle ss_truncate and us_truncate.
24565 * config/aarch64/arm_neon.h (vqmovn_high_s16): Reimplement using
24567 (vqmovn_high_s32): Likewise.
24568 (vqmovn_high_s64): Likewise.
24569 (vqmovn_high_u16): Likewise.
24570 (vqmovn_high_u32): Likewise.
24571 (vqmovn_high_u64): Likewise.
24573 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24575 * config/aarch64/aarch64-simd.md (aarch64_xtn2<mode>_le):
24577 (aarch64_xtn2<mode>_be): Likewise.
24578 (aarch64_xtn2<mode>): Likewise.
24579 * config/aarch64/aarch64-simd-builtins.def (xtn2): Define
24581 * config/aarch64/arm_neon.h (vmovn_high_s16): Reimplement using
24583 (vmovn_high_s32): Likewise.
24584 (vmovn_high_s64): Likewise.
24585 (vmovn_high_u16): Likewise.
24586 (vmovn_high_u32): Likewise.
24587 (vmovn_high_u64): Likewise.
24589 2021-01-13 Stafford Horne <shorne@gmail.com>
24591 * config/or1k/or1k.h (ASM_PREFERRED_EH_DATA_FORMAT): New macro.
24593 2021-01-13 Stafford Horne <shorne@gmail.com>
24595 * config/or1k/linux.h (TARGET_ASM_FILE_END): Define macro.
24597 2021-01-13 Stafford Horne <shorne@gmail.com>
24599 * config/or1k/or1k.h (TARGET_CPU_CPP_BUILTINS): Add builtin
24600 define for __or1k_hard_float__.
24602 2021-01-13 Stafford Horne <shorne@gmail.com>
24604 * config/or1k/or1k.h (NO_PROFILE_COUNTERS): Define as 1.
24605 (PROFILE_HOOK): Define to call _mcount.
24606 (FUNCTION_PROFILER): Change from abort to no-op.
24608 2021-01-13 Jakub Jelinek <jakub@redhat.com>
24610 PR tree-optimization/96691
24611 * match.pd ((~X | C) ^ D -> (X | C) ^ (~D ^ C),
24612 (~X & C) ^ D -> (X & C) ^ (D ^ C)): New simplifications if
24613 (~D ^ C) or (D ^ C) can be simplified.
24615 2021-01-13 Richard Biener <rguenther@suse.de>
24617 PR tree-optimization/92645
24618 * match.pd (BIT_FIELD_REF to conversion): Delay canonicalization
24619 until after vector lowering.
24621 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
24623 * config/aarch64/aarch64-sve.md (fnma<mode>4): Extend from SVE_FULL_I
24625 (@aarch64_pred_fnma<mode>, cond_fnma<mode>, *cond_fnma<mode>_2)
24626 (*cond_fnma<mode>_4, *cond_fnma<mode>_any): Likewise.
24628 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
24630 * config/aarch64/aarch64-sve.md (fma<mode>4): Extend from SVE_FULL_I
24632 (@aarch64_pred_fma<mode>, cond_fma<mode>, *cond_fma<mode>_2)
24633 (*cond_fma<mode>_4, *cond_fma<mode>_any): Likewise.
24635 2021-01-13 Richard Biener <rguenther@suse.de>
24637 PR tree-optimization/92645
24638 * tree-vect-slp.c (vect_build_slp_tree_1): Relax supported
24639 BIT_FIELD_REF argument.
24640 (vect_build_slp_tree_2): Record the desired vector type
24641 on the external vector def.
24642 (vectorizable_slp_permutation): Handle required punning
24643 of existing vector defs.
24645 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
24647 * rtl-ssa/accesses.h (def_lookup): Fix order of comparison results.
24649 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
24651 * config/sh/sh.md (movsf_ie): Remove operands[2] test.
24653 2021-01-13 Samuel Thibault <samuel.thibault@ens-lyon.org>
24655 * config.gcc [$target == *-*-gnu*]: Enable
24656 'default_gnu_indirect_function'.
24658 2021-01-13 Jakub Jelinek <jakub@redhat.com>
24661 * optabs.c (expand_vec_perm_const): Don't force v0 and v1 into
24662 registers before calling targetm.vectorize.vec_perm_const, only after
24664 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const): Handle
24665 two argument permutation when one operand is zero vector and only
24666 after that force operands into registers.
24667 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_1): New
24668 define_insn_and_split pattern.
24669 (*avx512bw_zero_extendv32qiv32hi2_1): Likewise.
24670 (*avx512f_zero_extendv16hiv16si2_1): Likewise.
24671 (*avx2_zero_extendv8hiv8si2_1): Likewise.
24672 (*avx512f_zero_extendv8siv8di2_1): Likewise.
24673 (*avx2_zero_extendv4siv4di2_1): Likewise.
24674 * config/mips/mips.c (mips_vectorize_vec_perm_const): Force operands
24676 * config/arm/arm.c (arm_vectorize_vec_perm_const): Likewise.
24677 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Likewise.
24678 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const): Likewise.
24679 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const): Likewise.
24680 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const): Likewise.
24681 * config/gcn/gcn.c (gcn_vectorize_vec_perm_const): Likewise. Use std::swap.
24683 2021-01-13 Martin Liska <mliska@suse.cz>
24685 PR tree-optimization/98455
24686 * gimple-if-to-switch.cc (condition_info::record_phi_mapping):
24687 Record also virtual PHIs.
24688 (pass_if_to_switch::execute): Return TODO_cleanup_cfg only
24691 2021-01-13 Jonathan Wakely <jwakely@redhat.com>
24693 * doc/invoke.texi (C++ Modules): Fix typos.
24695 2021-01-13 Richard Biener <rguenther@suse.de>
24697 PR tree-optimization/98640
24698 * tree-ssa-sccvn.c (visit_nary_op): Do not try to
24699 handle plus or minus from a truncated operand to be
24702 2021-01-13 Jakub Jelinek <jakub@redhat.com>
24705 * config/i386/i386.md (*btr<mode>_1, *btr<mode>_2): New
24706 define_insn_and_split patterns.
24707 (splitter after *btr<mode>_2): New splitter.
24709 2021-01-13 Martin Liska <mliska@suse.cz>
24712 * cgraphunit.c (analyze_functions): Remove dead code.
24714 2021-01-13 Qian Jianhua <qianjh@cn.fujitsu.com>
24716 * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
24717 * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
24718 (a64fx_regmove_cost, a64fx_vector_cost): New.
24719 (a64fx_tunings): Use the new added cost tables.
24721 2021-01-13 Jakub Jelinek <jakub@redhat.com>
24724 * config/i386/predicates.md (pmovzx_parallel): New predicate.
24725 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): New
24726 define_insn_and_split pattern.
24727 (*sse4_1_zero_extendv4hiv4si2_3): Likewise.
24728 (*sse4_1_zero_extendv2siv2di2_3): Likewise.
24730 2021-01-13 Julian Brown <julian@codesourcery.com>
24732 * config/gcn/gcn.c (gcn_conditional_register_usage): Remove dead code
24733 to fix v0 register.
24735 2021-01-13 Julian Brown <julian@codesourcery.com>
24737 * config/gcn/gcn.c (gcn_md_reorg): Fix case where EXEC reg is live
24740 2021-01-13 Julian Brown <julian@codesourcery.com>
24742 * config/gcn/gcn-valu.md (recip<mode>2<exec>, recip<mode>2): Use unspec
24743 for reciprocal-approximation instructions.
24744 (div<mode>3): Use fused multiply-accumulate operations for reciprocal
24745 refinement and division result.
24746 * config/gcn/gcn.md (UNSPEC_RCP): New unspec constant.
24748 2021-01-13 Julian Brown <julian@codesourcery.com>
24750 * config/gcn/gcn-valu.md (subdf): Rename to...
24753 2021-01-12 Martin Liska <mliska@suse.cz>
24755 * gcov.c (source_info::debug): Fix printf format for 32-bit hosts.
24757 2021-01-12 Andrea Corallo <andrea.corallo@arm.com>
24759 * function-abi.h: Fix typo.
24761 2021-01-12 Christophe Lyon <christophe.lyon@linaro.org>
24765 * config/arm/arm.h (ARM_HAVE_NEON_V8QI_LDST): New macro.
24766 (ARM_HAVE_NEON_V16QI_LDST, ARM_HAVE_NEON_V4HI_LDST): Likewise.
24767 (ARM_HAVE_NEON_V8HI_LDST, ARM_HAVE_NEON_V2SI_LDST): Likewise.
24768 (ARM_HAVE_NEON_V4SI_LDST, ARM_HAVE_NEON_V4HF_LDST): Likewise.
24769 (ARM_HAVE_NEON_V8HF_LDST, ARM_HAVE_NEON_V4BF_LDST): Likewise.
24770 (ARM_HAVE_NEON_V8BF_LDST, ARM_HAVE_NEON_V2SF_LDST): Likewise.
24771 (ARM_HAVE_NEON_V4SF_LDST, ARM_HAVE_NEON_DI_LDST): Likewise.
24772 (ARM_HAVE_NEON_V2DI_LDST): Likewise.
24773 (ARM_HAVE_V8QI_LDST, ARM_HAVE_V16QI_LDST): Likewise.
24774 (ARM_HAVE_V4HI_LDST, ARM_HAVE_V8HI_LDST): Likewise.
24775 (ARM_HAVE_V2SI_LDST, ARM_HAVE_V4SI_LDST, ARM_HAVE_V4HF_LDST): Likewise.
24776 (ARM_HAVE_V8HF_LDST, ARM_HAVE_V4BF_LDST, ARM_HAVE_V8BF_LDST): Likewise.
24777 (ARM_HAVE_V2SF_LDST, ARM_HAVE_V4SF_LDST, ARM_HAVE_DI_LDST): Likewise.
24778 (ARM_HAVE_V2DI_LDST): Likewise.
24779 * config/arm/mve.md (*movmisalign<mode>_mve_store): New pattern.
24780 (*movmisalign<mode>_mve_load): New pattern.
24781 * config/arm/neon.md (movmisalign<mode>): Move to ...
24782 * config/arm/vec-common.md: ... here.
24784 2021-01-12 Vladimir N. Makarov <vmakarov@redhat.com>
24787 * lra-eliminations.c (eliminate_regs_in_insn): Add transformation
24788 of pattern 'plus (plus (hard reg, const), pseudo)'.
24790 2021-01-12 Richard Biener <rguenther@suse.de>
24792 PR tree-optimization/98550
24793 * tree-vect-slp.c (vect_record_max_nunits): Check whether
24794 the group size is a multiple of the vector element count.
24795 (vect_build_slp_tree_1): When we need to fail because
24796 the vector type choosen causes unrolling do so lazily
24797 without affecting matches only at the end to guide group splitting.
24799 2021-01-12 Martin Liska <mliska@suse.cz>
24802 * optc-save-gen.awk: Compare also n_target_save vars with
24805 2021-01-12 Martin Liska <mliska@suse.cz>
24807 * gcov.c (source_info::debug): New.
24808 (print_usage): Add --debug (-D) option.
24809 (process_args): Likewise.
24810 (generate_results): Call src->debug after
24811 accumulate_line_counts.
24812 (read_graph_file): Properly assign id for EXIT_BLOCK.
24813 * profile.c (branch_prob): Dump function body before it is
24816 2021-01-12 Jakub Jelinek <jakub@redhat.com>
24818 PR tree-optimization/98629
24819 * tree-ssa-math-opts.c (arith_overflow_check_p): Don't update use_stmt
24820 unless returning non-zero.
24822 2021-01-12 Jakub Jelinek <jakub@redhat.com>
24824 PR tree-optimization/95731
24825 * tree-ssa-reassoc.c (optimize_range_tests_cmp_bitwise): Also optimize
24826 x < 0 && y < 0 && z < 0 into (x | y | z) < 0 for signed x, y, z.
24827 (optimize_range_tests): Call optimize_range_tests_cmp_bitwise
24828 only after optimize_range_tests_var_bound.
24830 2021-01-12 Jakub Jelinek <jakub@redhat.com>
24832 * configure.ac: Ensure c/Make-lang.in comes first in @all_lang_makefrags@.
24833 * configure: Regenerated.
24835 2021-01-12 liuhongt <hongtao.liu@intel.com>
24838 * config/i386/i386-builtins.h (BUILTIN_DESC_SWAP_OPERANDS):
24840 * config/i386/i386-expand.c (ix86_expand_sse_comi): Delete
24843 2021-01-12 Alexandre Oliva <oliva@adacore.com>
24845 * ssa-iterators.h (end_imm_use_stmt_traverse): Forward
24847 (auto_end_imm_use_stmt_traverse): New struct.
24848 (FOR_EACH_IMM_USE_STMT): Use it.
24849 (BREAK_FROM_IMM_USE_STMT, RETURN_FROM_IMM_USE_STMT): Remove,
24851 * gimple-ssa-strength-reduction.c: ... here, ...
24852 * graphite-scop-detection.c: ... here, ...
24853 * ipa-modref.c, ipa-pure-const.c, ipa-sra.c: ... here, ...
24854 * tree-predcom.c, tree-ssa-ccp.c: ... here, ...
24855 * tree-ssa-dce.c, tree-ssa-dse.c: ... here, ...
24856 * tree-ssa-loop-ivopts.c, tree-ssa-math-opts.c: ... here, ...
24857 * tree-ssa-phiprop.c, tree-ssa.c: ... here, ...
24858 * tree-vect-slp.c: ... and here, ...
24859 * doc/tree-ssa.texi: ... and the example here.
24861 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24863 * config/aarch64/aarch64-sve.md (sdiv_pow2<mode>3): Extend from
24864 SVE_FULL_I to SVE_I. Generate an UNSPEC_PRED_X.
24865 (*sdiv_pow2<mode>3): New pattern.
24866 (@cond_<sve_int_op><mode>): Extend from SVE_FULL_I to SVE_I.
24867 Wrap the ASRD in an UNSPEC_PRED_X.
24868 (*cond_<sve_int_op><mode>_2): Likewise. Replace the UNSPEC_PRED_X
24869 predicate with a constant PTRUE, if it isn't already.
24870 (*cond_<sve_int_op><mode>_z): Replace with...
24871 (*cond_<sve_int_op><mode>_any): ...this new pattern.
24873 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24875 * config/aarch64/aarch64-sve.md (*cond_bic<mode>_2): Extend from
24876 SVE_FULL_I to SVE_I.
24877 (*cond_bic<mode>_any): Likewise.
24879 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24881 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart)
24882 (@aarch64_pred_<MUL_HIGHPART:optab><mode>): Extend from SVE_FULL_I
24885 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24887 * config/aarch64/aarch64-sve.md (<su>abd<mode>_3): Extend from
24888 SVE_FULL_I to SVE_I.
24889 (*aarch64_cond_<su>abd<mode>_2): Likewise.
24890 (*aarch64_cond_<su>abd<mode>_any): Likewise.
24891 (@aarch64_pred_<su>abd<mode>): Likewise. Use UNSPEC_PRED_X
24892 for the max and min but not for the minus.
24893 (*aarch64_cond_<su>abd<mode>_3): New pattern.
24895 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24897 * config/aarch64/iterators.md (SVE_24I): New iterator.
24898 * config/aarch64/aarch64-sve.md (*aarch64_adr<mode>_shift): Extend from
24899 SVE_FULL_SDI to SVE_24I. Use containers rather than elements.
24901 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24903 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_BINARY:optab><mode>)
24904 (*cond_<SVE_INT_BINARY:optab><mode>_2): Extend from SVE_FULL_I
24906 (*cond_<SVE_INT_BINARY:optab><mode>_3): Likewise.
24907 (*cond_<SVE_INT_BINARY:optab><mode>_any): Likewise.
24908 (*cond_<SVE_INT_BINARY:optab><mode>_2_const): Likewise.
24909 (*cond_<SVE_INT_BINARY:optab><mode>_any_const): Likewise.
24911 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24913 * config/aarch64/aarch64-sve.md (<SVE_INT_BINARY_IMM:optab><mode>3)
24914 (@aarch64_pred_<SVE_INT_BINARY_IMM:optab><mode>)
24915 (*post_ra_<SVE_INT_BINARY_IMM:optab><mode>3): Extend from SVE_FULL_I
24918 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24920 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3)
24921 (v<ASHIFT:optab><mode>3, @aarch64_pred_<optab><mode>)
24922 (*post_ra_v<ASHIFT:optab><mode>3): Extend from SVE_FULL_I to SVE_I.
24924 2021-01-11 Martin Liska <mliska@suse.cz>
24927 * symtab-clones.h (clone_info::release): Release
24928 symtab::m_clones with ggc_delete as it's a GGC memory.
24930 2021-01-11 Matthias Klose <doko@ubuntu.com>
24932 * Makefile.in (LINK_PROGRESS): Show the link target.
24934 2021-01-11 Richard Biener <rguenther@suse.de>
24936 PR tree-optimization/91403
24937 * tree-vect-data-refs.c (vect_analyze_group_access_1): Cap
24938 single-element interleaving group size at 4096 elements.
24940 2021-01-11 Richard Biener <rguenther@suse.de>
24942 PR tree-optimization/98526
24943 * tree-vect-loop.c (vect_model_reduction_cost): Remove costing
24944 of the actual reduction op for the regular case.
24945 (vectorizable_reduction): Cost the stmts
24946 vect_transform_reduction produces here.
24948 2021-01-11 Andreas Krebbel <krebbel@linux.ibm.com>
24950 * tree-ssa-forwprop.c (simplify_vector_constructor): For
24951 big-endian, use UNPACK[_FLOAT]_HI.
24953 2021-01-11 Tamar Christina <tamar.christina@arm.com>
24955 * tree-vect-slp-patterns.c (class complex_pattern,
24956 class complex_add_pattern): Add parameters to matches.
24957 (complex_add_pattern::build): Free memory.
24958 (complex_add_pattern::matches): Move validation end of match.
24959 (complex_add_pattern::recognize): Likewise.
24961 2021-01-11 Tamar Christina <tamar.christina@arm.com>
24963 * tree-vect-slp-patterns.c (linear_loads_p): Fix externals.
24965 2021-01-11 Tamar Christina <tamar.christina@arm.com>
24967 * tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
24969 2021-01-11 Jakub Jelinek <jakub@redhat.com>
24971 PR tree-optimization/95867
24972 * tree-ssa-math-opts.h: New header.
24973 * tree-ssa-math-opts.c: Include tree-ssa-math-opts.h.
24974 (powi_as_mults): No longer static. Use build_one_cst instead of
24975 build_real. Formatting fix.
24976 * tree-ssa-reassoc.c: Include tree-ssa-math-opts.h.
24977 (attempt_builtin_powi): Handle multiplication reassociation without
24978 powi_fndecl using powi_as_mults.
24979 (reassociate_bb): For integral types don't require
24980 -funsafe-math-optimizations to call attempt_builtin_powi.
24982 2021-01-11 Jakub Jelinek <jakub@redhat.com>
24984 PR tree-optimization/95852
24985 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): Change
24986 mul_stmts parameter type to vec<gimple *> &. Before cond_stmt
24987 allow in the bb any of the stmts in that vector, div_stmt and
24988 up to 3 cast stmts.
24989 (arith_cast_equal_p): New function.
24990 (arith_overflow_check_p): Add cast_stmt argument, handle signed
24991 multiply overflow checks.
24992 (match_arith_overflow): Adjust caller. Handle signed multiply
24995 2021-01-11 Jakub Jelinek <jakub@redhat.com>
24997 PR tree-optimization/95852
24998 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): New function.
24999 (uaddsub_overflow_check_p): Renamed to ...
25000 (arith_overflow_check_p): ... this. Handle also multiplication
25001 with overflow check.
25002 (match_uaddsub_overflow): Renamed to ...
25003 (match_arith_overflow): ... this. Add cfg_changed argument. Handle
25004 also multiplication with overflow check. Adjust function comment.
25005 (math_opts_dom_walker::after_dom_children): Adjust callers. Call
25006 match_arith_overflow also for MULT_EXPR.
25008 2021-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25010 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
25011 __builtin_convertvector.
25012 (vmovl_s16): Likewise.
25013 (vmovl_s32): Likewise.
25014 (vmovl_u8): Likewise.
25015 (vmovl_u16): Likewise.
25016 (vmovl_u32): Likewise.
25017 (vmovn_s16): Likewise.
25018 (vmovn_s32): Likewise.
25019 (vmovn_s64): Likewise.
25020 (vmovn_u16): Likewise.
25021 (vmovn_u32): Likewise.
25022 (vmovn_u64): Likewise.
25024 2021-01-11 Martin Liska <mliska@suse.cz>
25026 * gimple-if-to-switch.cc (struct condition_info): Use auto_var.
25027 (if_chain::is_beneficial): Delete clusters
25028 (find_conditions): Make second argument of conditions_in_bbs a
25029 pointer so that we control over it's lifetime.
25030 (pass_if_to_switch::execute): Delete them.
25032 2021-01-11 Kewen Lin <linkw@linux.ibm.com>
25034 * ira.c (move_unallocated_pseudos): Check other_reg and skip if
25037 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
25039 * config/vax/vax.md (cc): Remove mode attribute.
25040 (subst_<cc>, subst_f<cc>): Rename to...
25041 (subst_<mode>, subst_f<VAXccnz:mode>): ... these respectively.
25042 (*cbranch<VAXint:mode>4_<VAXcc:mode>): Update for `cc' removal.
25043 (*cbranch<VAXfp:mode>4_<VAXccnz:mode>): Likewise.
25044 (*branch_<mode>, *branch_<mode>_reversed): Likewise.
25046 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
25048 * config/vax/vax.md (subst_f<cc>): Add mode to operands and
25049 `const_double_zero'.
25051 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
25053 * config/pdp11/pdp11.md (PDPfp): New mode iterator.
25054 (fcc_cc, fcc_ccnz): Use it. Add mode to `const_double_zero' and
25057 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
25059 * genemit.c (gen_exp) <CONST_DOUBLE>: Handle `const_double_zero'
25061 * read-rtl.c (rtx_reader::read_rtx_code): Handle machine mode
25062 with `const_double_zero'.
25063 * doc/rtl.texi (Constant Expression Types): Document it.
25065 2021-01-09 Jakub Jelinek <jakub@redhat.com>
25068 * tree-cfg.c (verify_gimple_assign_binary): Allow lhs of
25069 POINTER_DIFF_EXPR to be any integral type.
25071 2021-01-09 Jakub Jelinek <jakub@redhat.com>
25073 PR rtl-optimization/98603
25074 * function.c (instantiate_virtual_regs_in_insn): For asm goto
25075 with impossible constraints, drop all SETs, CLOBBERs, drop PARALLEL
25076 if any, set ASM_OPERANDS mode to VOIDmode and change
25077 ASM_OPERANDS_OUTPUT_CONSTRAINT and ASM_OPERANDS_OUTPUT_IDX.
25079 2021-01-09 Alexandre Oliva <oliva@gnu.org>
25082 * final.c (notice_source_line): Narrow down the condition to
25083 skip a line-0 marker.
25085 2021-01-08 Sergei Trofimovich <siarheit@google.com>
25087 * ipa-modref.c (merge_call_side_effects): Fix
25088 linebreak split by reordering two print calls.
25090 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
25092 * config/s390/vector.md (*tf_to_fprx2_0): Rename from
25093 "*mov_tf_to_fprx2_0" for consistency, fix constraint.
25094 (*tf_to_fprx2_1): Rename from "*mov_tf_to_fprx2_1" for
25095 consistency, fix constraint.
25097 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
25099 * config/s390/s390-c.c (s390_def_or_undef_macro): Accept
25100 callables instead of mask values.
25101 (struct target_flag_set_p): New predicate.
25102 (s390_cpu_cpp_builtins_internal): Define or undefine
25103 __LONG_DOUBLE_VX__ macro.
25105 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
25108 * config/i386/i386.c (x86_function_profiler): Use R10 and R11
25109 to call mcount in large model with PIC for NO_PROFILE_COUNTERS
25112 2021-01-08 Richard Biener <rguenther@suse.de>
25114 * tree-ssa-sccvn.c (pass_fre::execute): Reset the SCEV hash table.
25116 2021-01-08 Richard Biener <rguenther@suse.de>
25118 * tree-vect-slp.c (scalar_stmts_to_slp_tree_map_t): Fix.
25119 (vect_build_slp_tree): On cache hit release the matched
25120 scalar stmts vector.
25121 * tree-vect-stmts.c (vectorizable_store): Properly free
25122 vec_oprnds before possibly gathering them again.
25124 2021-01-08 Richard Biener <rguenther@suse.de>
25126 PR tree-optimization/98544
25127 * tree-vect-slp.c (vect_optimize_slp): Always materialize
25128 permutes at a permute node.
25130 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
25133 * config/i386/i386.c (x86_function_profiler): Use R10 to call
25134 mcount in large model. Sorry for large model with PIC.
25136 2021-01-08 Jakub Jelinek <jakub@redhat.com>
25139 * config/i386/i386.opt (ix86_cmodel, ix86_incoming_stack_boundary_arg,
25140 ix86_pmode, ix86_preferred_stack_boundary_arg, ix86_regparm,
25141 ix86_veclibabi_type): Remove x_ prefix, use TargetVariable instead of
25142 TargetSave and initialize for variables with enum types.
25143 (mfentry, mstack-protector-guard-reg=, mstack-protector-guard-offset=,
25144 mstack-protector-guard-symbol=): Add Save.
25145 * config/i386/i386-options.c (ix86_function_specific_save,
25146 ix86_function_specific_restore): Don't save or restore x_ix86_cmodel,
25147 x_ix86_incoming_stack_boundary_arg, x_ix86_pmode,
25148 x_ix86_preferred_stack_boundary_arg, x_ix86_regparm,
25149 x_ix86_veclibabi_type.
25151 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
25153 * config/aarch64/aarch64-sve.md (*cnot<mode>): Extend from
25154 SVE_FULL_I to SVE_I.
25155 (*cond_cnot<mode>_2, *cond_cnot<mode>_any): Likewise.
25157 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
25159 * config/aarch64/aarch64-sve.md (*cond_uxt<mode>_2): Extend from
25160 SVE_FULL_I to SVE_I.
25161 (*cond_uxt<mode>_any): Likewise.
25163 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25165 * config/aarch64/iterators.md (Vwhalf): New iterator.
25166 * config/aarch64/aarch64-simd.md (aarch64_<sur>adalp<mode>_3):
25168 (aarch64_<sur>adalp<mode>): ... This. Make more
25170 (<sur>sadv16qi): Adjust callsite of the above.
25171 * config/aarch64/aarch64-simd-builtins.def (sadalp, uadalp): New
25173 * config/aarch64/arm_neon.h (vpadal_s8): Reimplement using
25175 (vpadal_s16): Likewise.
25176 (vpadal_u8): Likewise.
25177 (vpadal_u16): Likewise.
25178 (vpadalq_s8): Likewise.
25179 (vpadalq_s16): Likewise.
25180 (vpadalq_s32): Likewise.
25181 (vpadalq_u8): Likewise.
25182 (vpadalq_u16): Likewise.
25183 (vpadalq_u32): Likewise.
25185 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25187 * config/aarch64/aarch64-simd.md (aarch64_<su>abd<mode>_3):
25189 (aarch64_<su>abd<mode>): ... This.
25190 (<sur>sadv16qi): Adjust callsite of the above.
25191 * config/aarch64/aarch64-simd-builtins.def (sabd, uabd): Define
25193 * config/aarch64/arm_neon.h (vabd_s8): Reimplement using
25195 (vabd_s16): Likewise.
25196 (vabd_s32): Likewise.
25197 (vabd_u8): Likewise.
25198 (vabd_u16): Likewise.
25199 (vabd_u32): Likewise.
25200 (vabdq_s8): Likewise.
25201 (vabdq_s16): Likewise.
25202 (vabdq_s32): Likewise.
25203 (vabdq_u8): Likewise.
25204 (vabdq_u16): Likewise.
25205 (vabdq_u32): Likewise.
25207 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25209 * config/aarch64/aarch64-simd-builtins.def (saba, uaba): Define
25211 * config/aarch64/arm_neon.h (vaba_s8): Implement using builtin.
25212 (vaba_s16): Likewise.
25213 (vaba_s32): Likewise.
25214 (vaba_u8): Likewise.
25215 (vaba_u16): Likewise.
25216 (vaba_u32): Likewise.
25217 (vabaq_s8): Likewise.
25218 (vabaq_s16): Likewise.
25219 (vabaq_s32): Likewise.
25220 (vabaq_u8): Likewise.
25221 (vabaq_u16): Likewise.
25222 (vabaq_u32): Likewise.
25224 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25226 * config/aarch64/aarch64-simd.md (aba<mode>_3): Rename to...
25227 (aarch64_<su>aba<mode>): ... This. Handle uaba as well.
25228 Change RTL pattern to match.
25230 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
25232 * common/config/riscv/riscv-common.c (riscv_current_subset_list): New.
25233 * config/riscv/riscv-c.c (riscv-subset.h): New.
25234 (INCLUDE_STRING): Define.
25235 (riscv_cpu_cpp_builtins): Add new style architecture extension
25237 * config/riscv/riscv-subset.h (riscv_subset_list::begin): New.
25238 (riscv_subset_list::end): New.
25239 (riscv_current_subset_list): New.
25241 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
25243 * common/config/riscv/riscv-common.c (RISCV_DONT_CARE_VERSION):
25244 Move to riscv-subset.h.
25245 (struct riscv_subset_t): Ditto.
25246 (class riscv_subset_list): Ditto.
25247 * config/riscv/riscv-subset.h (RISCV_DONT_CARE_VERSION): Move
25248 from riscv-common.c.
25249 (struct riscv_subset_t): Ditto.
25250 (class riscv_subset_list): Ditto.
25251 * config/riscv/t-riscv ($(common_out_file)): Add file
25254 2021-01-07 Jakub Jelinek <jakub@redhat.com>
25257 * config/i386/i386.md (*bmi_blsi_<mode>_cmp, *bmi_blsi_<mode>_ccno):
25258 New define_insn patterns.
25260 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
25262 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_UNARY:optab><mode>)
25263 (*cond_<SVE_INT_UNARY:optab><mode>_2): Extend from SVE_FULL_I to SVE_I.
25264 (*cond_<SVE_INT_UNARY:optab><mode>_any): Likewise.
25266 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
25268 PR tree-optimization/98560
25269 * internal-fn.def (IFN_VCONDU, IFN_VCONDEQ): Use type vec_cond.
25270 * internal-fn.c (vec_cond_mask_direct): Get the data mode from
25272 (vec_cond_direct): Likewise argument 2.
25273 (vec_condu_direct, vec_condeq_direct): Delete.
25274 (expand_vect_cond_optab_fn): Rename to...
25275 (expand_vec_cond_optab_fn): ...this, replacing old macro.
25276 (expand_vec_condu_optab_fn, expand_vec_condeq_optab_fn): Delete.
25277 (expand_vect_cond_mask_optab_fn): Rename to...
25278 (expand_vec_cond_mask_optab_fn): ...this, replacing old macro.
25279 (direct_vec_cond_mask_optab_supported_p): Treat the optab as a
25281 (direct_vec_cond_optab_supported_p): Likewise.
25282 (direct_vec_condu_optab_supported_p): Delete.
25283 (direct_vec_condeq_optab_supported_p): Delete.
25284 * gimple-isel.cc: Include internal-fn.h.
25285 (gimple_expand_vec_cond_expr): Check that IFN_VCONDEQ is supported
25288 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
25290 PR tree-optimization/98560
25291 * gimple-isel.cc (gimple_expand_vec_cond_expr): If we fail to use
25292 IFN_VCOND{,U,EQ}, fall back on IFN_VCOND_MASK.
25294 2021-01-07 Uroš Bizjak <ubizjak@gmail.com>
25296 * config/i386/i386.md (insn): Merge from plusminus_insn, shift_insn,
25297 rotate_insn and optab code attributes.
25298 Update all uses to merged code attribute.
25299 * config/i386/sse.md: Update all uses to merged code attribute.
25300 * config/i386/mmx.md: Update all uses to merged code attribute.
25302 2021-01-07 Jakub Jelinek <jakub@redhat.com>
25304 PR tree-optimization/98568
25305 * gimple-ssa-store-merging.c (bswap_view_convert): New function.
25306 (bswap_replace): Use it.
25308 2021-01-06 Vladimir N. Makarov <vmakarov@redhat.com>
25310 PR rtl-optimization/97978
25311 * lra-int.h (lra_hard_reg_split_p): New external.
25312 * lra.c (lra_hard_reg_split_p): New global.
25313 (lra): Set up lra_hard_reg_split_p after splitting a hard reg.
25314 * lra-assigns.c (lra_assign): Don't check allocation correctness
25315 after hard reg splitting.
25317 2021-01-06 Martin Sebor <msebor@redhat.com>
25320 * builtins.c (new_delete_mismatch_p): New overload.
25321 (new_delete_mismatch_p (tree, tree)): Call it.
25323 2021-01-06 Alexandre Oliva <oliva@adacore.com>
25325 * Makefile.in (T_GLIMITS_H): New.
25326 (stmp-int-hdrs): Depend on it, use it.
25327 * config/t-vxworks (T_GLIMITS_H): Override it.
25328 (vxw-glimits.h): New.
25330 2021-01-06 Richard Biener <rguenther@suse.de>
25332 PR tree-optimization/98513
25333 * value-range.cc (intersect_ranges): Compare the upper bounds
25334 for the expected relation.
25336 2021-01-06 Gerald Pfeifer <gerald@pfeifer.com>
25339 2020-12-28 Gerald Pfeifer <gerald@pfeifer.com>
25341 * doc/standards.texi (HSAIL): Remove section.
25343 2021-01-05 Samuel Thibault <samuel.thibault@ens-lyon.org>
25345 * configure: Re-generate.
25347 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25349 * doc/invoke.texi (-std=c++20): Adjust for the publication of
25350 ISO 14882:2020 standard.
25351 * doc/standards.texi: Likewise.
25353 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25355 PR tree-optimization/94802
25356 * expr.h (maybe_optimize_sub_cmp_0): Declare.
25357 * expr.c: Include tree-pretty-print.h and flags.h.
25358 (maybe_optimize_sub_cmp_0): New function.
25359 (do_store_flag): Use it.
25360 * cfgexpand.c (expand_gimple_cond): Likewise.
25362 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
25364 * mux-utils.h (pointer_mux::m_ptr): Tweak description of contents.
25365 * rtlanal.c (simple_regno_set): Tweak description to clarify the
25368 2021-01-05 Richard Biener <rguenther@suse.de>
25370 PR tree-optimization/98516
25371 * tree-vect-slp.c (vect_optimize_slp): Permute the incoming
25372 lanes when materializing on a VEC_PERM node.
25373 (vectorizable_slp_permutation): Dump the permute properly.
25375 2021-01-05 Richard Biener <rguenther@suse.de>
25377 * tree-vect-slp.c (vect_slp_region): Move debug counter
25378 to cover individual subgraphs.
25380 2021-01-05 Richard Biener <rguenther@suse.de>
25382 PR tree-optimization/98428
25383 * tree-vect-slp.c (vect_build_slp_tree_1): Properly reject
25384 vector lane extracts for loop vectorization.
25386 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25388 PR tree-optimization/98514
25389 * tree-ssa-reassoc.c (bb_rank): Change type from long * to
25391 (operand_rank): Change type from hash_map<tree, long> to
25392 hash_map<tree, int64_t>.
25393 (phi_rank): Change return type from long to int64_t.
25394 (loop_carried_phi): Change block_rank variable type from long to
25396 (propagate_rank): Change return type, rank parameter type and
25397 op_rank variable type from long to int64_t.
25398 (find_operand_rank): Change return type from long to int64_t
25399 and change slot variable type from long * to int64_t *.
25400 (insert_operand_rank): Change rank parameter type from long to
25402 (get_rank): Change return type and rank variable type from long to
25403 int64_t. Use PRId64 instead of ld to print the rank.
25404 (init_reassoc): Change rank variable type from long to int64_t
25405 and adjust correspondingly bb_rank and operand_rank initialization.
25407 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25409 PR tree-optimization/96928
25410 * tree-ssa-phiopt.c (xor_replacement): New function.
25411 (tree_ssa_phiopt_worker): Call it.
25413 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25415 PR tree-optimization/96930
25416 * match.pd ((A / (1 << B)) -> (A >> B)): If A is extended
25417 from narrower value which has the same type as 1 << B, perform
25418 the right shift on the narrower value followed by extension.
25420 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25422 PR tree-optimization/96239
25423 * gimple-ssa-store-merging.c (maybe_optimize_vector_constructor): New
25425 (get_status_for_store_merging): Don't return BB_INVALID for blocks
25426 with potential bswap optimizable CONSTRUCTORs.
25427 (pass_store_merging::execute): Optimize vector CONSTRUCTORs with bswap
25430 2021-01-05 Richard Biener <rguenther@suse.de>
25432 PR tree-optimization/98381
25433 * tree.c (vector_element_bits): Properly compute bool vector
25435 * tree-vect-loop.c (vectorizable_live_operation): Properly
25436 compute the last lane bit offset.
25438 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
25441 * config/i386/sse.md (sse_cvtps2pi): Redefine as define_insn_and_split.
25442 Clear the top 64 bytes of the input XMM register.
25443 (sse_cvttps2pi): Ditto.
25445 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
25448 * config/i386/xopintrin.h (_mm256_cmov_si256): New.
25450 2021-01-05 H.J. Lu <hjl.tools@gmail.com>
25453 * config/i386/xmmintrin.h (_mm_extract_pi16): Cast to unsigned
25456 2021-01-05 Claudiu Zissulescu <claziss@synopsys.com>
25458 * config/arc/arc.md (maddsidi4_split): Use ACC_REG_FIRST.
25459 (umaddsidi4_split): Likewise.
25461 2021-01-05 liuhongt <hongtao.liu@intel.com>
25464 * config/i386/sse.md (*sse2_pmovskb_zexthisi): New
25465 define_insn_and_split for zero_extend of subreg HI of pmovskb
25467 (*sse2_pmovskb_zexthisi): Add new combine splitters for
25468 zero_extend of not of subreg HI of pmovskb result.
25470 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
25473 * explow.c (convert_memory_address_addr_space_1): Handle UNSPECs
25475 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate): Use
25476 convert_memory_address to convert symbolic immediates to ptr_mode
25477 before forcing them to memory.
25479 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
25481 PR rtl-optimization/97144
25482 * recog.c (constrain_operands): Initialize matching_operand
25483 for each alternative, rather than only doing it once.
25485 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
25487 PR rtl-optimization/98403
25488 * rtl-ssa/changes.cc (function_info::finalize_new_accesses): Explain
25489 why we don't remove call clobbers.
25490 (function_info::apply_changes_to_insn): Don't attempt to add
25491 call clobbers here.
25493 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
25495 PR tree-optimization/98371
25496 * tree-vect-loop.c (vect_reanalyze_as_main_loop): New function.
25497 (vect_analyze_loop): If an epilogue loop appears to be cheaper
25498 than the main loop, re-analyze it as a main loop before adopting
25501 2021-01-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
25504 * configure.ac (NETLIBS): Determine using AX_LIB_SOCKET_NSL.
25505 * aclocal.m4, configure: Regenerate.
25506 * Makefile.in (NETLIBS): Define.
25507 (BACKEND): Remove $(CODYLIB).
25509 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25511 PR rtl-optimization/98334
25512 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
25513 Optimize (X - 1) * Y + Y to X * Y or (X + 1) * Y - Y to X * Y.
25515 2021-01-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
25517 * tree-inline.c (expand_call_inline): Restore input_location.
25518 Return result from recursive call.
25520 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
25522 PR tree-optimization/95401
25523 * config/aarch64/aarch64-sve-builtins.cc
25524 (gimple_folder::load_store_cookie): Use bits rather than bytes
25525 for the alignment argument to IFN_MASK_LOAD and IFN_MASK_STORE.
25526 * gimple-fold.c (gimple_fold_mask_load_store_mem_ref): Likewise.
25527 * tree-vect-stmts.c (vectorizable_store): Likewise.
25528 (vectorizable_load): Likewise.
25530 2021-01-04 Richard Biener <rguenther@suse.de>
25532 PR tree-optimization/98308
25533 * tree-vect-stmts.c (vectorizable_load): Set invariant mask
25536 2021-01-04 Jakub Jelinek <jakub@redhat.com>
25538 PR tree-optimization/95771
25539 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Handle types
25540 with precision smaller than int's precision and types with precision
25541 twice as large as long long. Formatting fixes.
25543 2021-01-04 Richard Biener <rguenther@suse.de>
25545 PR tree-optimization/98464
25546 * tree-ssa-sccvn.c (vn_valueize_for_srt): Rename from ...
25547 (vn_valueize_wrapper): ... this. Temporarily adjust vn_context_bb.
25548 (process_bb): Adjust.
25550 2021-01-04 Matthew Malcomson <matthew.malcomson@arm.com>
25553 * doc/invoke.texi (-fsanitize=address): Fix wording describing
25554 clash with -fsanitize=hwaddress.
25556 2021-01-04 Richard Biener <rguenther@suse.de>
25558 PR tree-optimization/98282
25559 * tree-ssa-sccvn.c (vn_get_stmt_kind): Classify tcc_reference on
25560 invariants as VN_NARY.
25562 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
25565 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Accept
25566 aarch64_simd_reg_or_zero for operand 2. Use the combinez patterns
25567 to handle zero operands.
25569 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
25571 * config/aarch64/aarch64.c (offset_6bit_signed_scaled_p): New function.
25572 (offset_6bit_unsigned_scaled_p): Fix typo in comment.
25573 (aarch64_sve_prefetch_operand_p): Accept MUL VLs in the range
25576 2021-01-04 Richard Biener <rguenther@suse.de>
25578 PR tree-optimization/98393
25579 * tree-vect-slp.c (vect_build_slp_tree): Properly zero matches
25580 when hitting the limit.
25582 2021-01-04 Richard Biener <rguenther@suse.de>
25584 PR tree-optimization/98291
25585 * tree-vect-loop.c (vectorizable_reduction): Bypass
25586 associativity check for SLP reductions with VF 1.
25588 2021-01-04 Jakub Jelinek <jakub@redhat.com>
25590 PR tree-optimization/96782
25591 * match.pd (x == ~x -> false, x != ~x -> true): New simplifications.
25593 2021-01-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
25595 * collect-utils.c (collect_execute): Check dumppfx.
25596 * collect2.c (maybe_run_lto_and_relink, do_link): Pass atsuffix
25597 to collect_execute.
25598 (do_link): Add new parameter atsuffix.
25599 (main): Handle -dumpdir option. Skip one argument for
25600 -o, -isystem and -B options.
25601 * gcc.c (make_at_file): New helper function.
25602 (close_at_file): Use it.
25604 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25606 * config/darwin.h (MIN_LD64_NO_COAL_SECTS): Adjust.
25607 Amend handling for LD64_VERSION fallback defaults.
25609 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25611 * config.gcc: Compute default version information
25612 from the configured target. Likewise defaults for
25614 * config/darwin10.h: Removed.
25615 * config/darwin12.h: Removed.
25616 * config/darwin9.h: Removed.
25617 * config/rs6000/darwin8.h: Removed.
25619 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25621 * config/darwin9.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
25623 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25625 * config/darwin9.h (STACK_CHECK_STATIC_BUILTIN): Move from here..
25626 * config/darwin.h (STACK_CHECK_STATIC_BUILTIN): .. to here.
25628 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25630 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move from
25632 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): ... to here.
25634 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25636 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move the spec
25637 for the Darwin10 unwinder stub from here ...
25638 * config/darwin.h (LINK_COMMAND_SPEC_A): ... to here.
25640 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25642 * config/darwin.h (DSYMUTIL_SPEC): Default to DWARF
25643 (ASM_DEBUG_SPEC):Only define if the assembler supports
25645 (PREFERRED_DEBUGGING_TYPE): Default to DWARF.
25646 (DARWIN_PREFER_DWARF): Define.
25647 * config/darwin9.h (PREFERRED_DEBUGGING_TYPE): Remove.
25648 (DARWIN_PREFER_DWARF): Likewise
25649 (DSYMUTIL_SPEC): Likewise.
25650 (COLLECT_RUN_DSYMUTIL): Likewise.
25651 (ASM_DEBUG_SPEC): Likewise.
25652 (ASM_DEBUG_OPTION_SPEC): Likewise.
25654 2021-01-02 Jan Hubicka <jh@suse.cz>
25656 * cfg.c (free_block): ggc_free bb.
25658 2021-01-01 Jakub Jelinek <jakub@redhat.com>
25660 * gcc.c (process_command): Update copyright notice dates.
25661 * gcov-dump.c (print_version): Ditto.
25662 * gcov.c (print_version): Ditto.
25663 * gcov-tool.c (print_version): Ditto.
25664 * gengtype.c (create_file): Ditto.
25665 * doc/cpp.texi: Bump @copying's copyright year.
25666 * doc/cppinternals.texi: Ditto.
25667 * doc/gcc.texi: Ditto.
25668 * doc/gccint.texi: Ditto.
25669 * doc/gcov.texi: Ditto.
25670 * doc/install.texi: Ditto.
25671 * doc/invoke.texi: Ditto.
25673 2021-01-01 Jakub Jelinek <jakub@redhat.com>
25675 * ChangeLog-2020: Rotate ChangeLog. New file.
25678 Copyright (C) 2021 Free Software Foundation, Inc.
25680 Copying and distribution of this file, with or without modification,
25681 are permitted in any medium without royalty provided the copyright
25682 notice and this notice are preserved.