1 2021-12-24 Uroš Bizjak <ubizjak@gmail.com>
5 * config/i386/mmx.md (divv2sf3): New instruction pattern.
7 2021-12-24 Iain Sandoe <iain@sandoe.co.uk>
9 * config/darwin.c (darwin_override_options): Make a comment
12 2021-12-24 Iain Sandoe <iain@sandoe.co.uk>
14 * config.gcc: Emit L2_MAX_OFILE_ALIGNMENT with suitable
16 * config/darwin.c (darwin_emit_common): Error for alignment
18 * config/darwin.h (MAX_OFILE_ALIGNMENT): Rework to use the
19 configured L2_MAX_OFILE_ALIGNMENT.
21 2021-12-24 Iain Sandoe <iain@sandoe.co.uk>
23 * config/darwin.c (darwin_override_options): When checking for the
24 flag-reorder-and-partition case, also check that it is set on.
26 2021-12-24 Iain Sandoe <iain@sandoe.co.uk>
28 * config/darwin.h (OBJECT_FORMAT_MACHO): New.
30 2021-12-23 Roger Sayle <roger@nextmovesoftware.com>
31 Uroš Bizjak <ubizjak@gmail.com>
34 * config/i386/i386.md (*mov<mode>_and): New define_insn for
35 writing a zero to memory using AND.
36 (*mov<mode>_or): Extend to allow memory destination and HImode.
37 (*movdi_internal): Remove -Oz push/pop optimization from here.
38 (*movsi_internal): Likewise.
39 (peephole2): Perform -Oz push/pop optimization here, only for
40 register destinations, values other than zero, and in functions
41 that don't used the red zone.
42 (peephole2): With -Oz, convert writes of 0 or -1 to memory into
43 their clobber forms, i.e. *mov<mode>_and and *mov<mode>_or resp.
45 2021-12-23 konglin1 <lingling.kong@intel.com>
47 * config/i386/avx512bf16intrin.h (_mm_cvtsbh_ss): Add new intrinsic.
48 (_mm512_cvtpbh_ps): Likewise.
49 (_mm512_maskz_cvtpbh_ps): Likewise.
50 (_mm512_mask_cvtpbh_ps): Likewise.
51 * config/i386/avx512bf16vlintrin.h (_mm_cvtness_sbh): Likewise.
52 (_mm_cvtpbh_ps): Likewise.
53 (_mm256_cvtpbh_ps): Likewise.
54 (_mm_maskz_cvtpbh_ps): Likewise.
55 (_mm256_maskz_cvtpbh_ps): Likewise.
56 (_mm_mask_cvtpbh_ps): Likewise.
57 (_mm256_mask_cvtpbh_ps): Likewise.
59 2021-12-23 Feng Xue <fxue@os.amperecomputing.com>
62 * tree.c (verify_type): Fix typo.
64 2021-12-23 liuhongt <hongtao.liu@intel.com>
68 (*<avx512>_cmp<V48H_AVX512VL:mode>3_zero_extend<SWI248x:mode>):
69 New pre_reload define_insn_and_split.
70 (*<avx512>_cmp<VI12_AVX512VL:mode>3_zero_extend<SWI248x:mode>):
72 (*<avx512>_ucmp<VI12_AVX512VL:mode>3_zero_extend<SWI248x:mode>):
74 (*<avx512>_ucmp<VI48_AVX512VL:mode>3_zero_extend<SWI248x:mode>):
76 (*<avx512>_cmp<V48H_AVX512VL:mode>3_zero_extend<SWI248x:mode>_2):
78 (*<avx512>_cmp<VI12_AVX512VL:mode>3_zero_extend<SWI248x:mode>_2):
80 (*<avx512>_ucmp<VI12_AVX512VL:mode>3_zero_extend<SWI248x:mode>_2):
82 (*<avx512>_ucmp<VI48_AVX512VL:mode>3_zero_extend<SWI248x:mode>_2):
85 2021-12-22 Murray Steele <murray.steele@arm.com>
87 * config.gcc (arm*-*-*): Add arm-mve-builtins.o to extra_objs.
88 * config/arm/arm-c.c (arm_pragma_arm): Handle "#pragma GCC arm".
89 (arm_register_target_pragmas): Register it.
90 * config/arm/arm-protos.h: (arm_mve::arm_handle_mve_types_h): New
92 * config/arm/arm_mve_types.h: Replace MVE type definitions with
94 * config/arm/t-arm: (arm-mve-builtins.o): New target rule.
95 * config/arm/arm-mve-builtins.cc: New file.
96 * config/arm/arm-mve-builtins.def: New file.
97 * config/arm/arm-mve-builtins.h: New file.
99 2021-12-22 Murray Steele <murray.steele@arm.com>
101 * config/arm/arm-builtins.c (enum arm_type_qualifiers): Move to
103 (enum arm_simd_type): Move to arm-builtins.h.
104 (struct arm_simd_type_info): Move to arm-builtins.h.
105 * config/arm/arm-builtins.h (enum arm_simd_type): Move from
107 (enum arm_type_qualifiers): Move from arm-builtins.c.
108 (struct arm_simd_type_info): Move from arm-builtins.c.
110 2021-12-22 Martin Liska <mliska@suse.cz>
112 * doc/extend.texi: Unify all function declarations in examples
113 where some miss trailing ';'.
115 2021-12-22 Martin Liska <mliska@suse.cz>
117 * doc/extend.texi: Unify all function declarations in examples
118 where some miss trailing ';'.
120 2021-12-22 Martin Liska <mliska@suse.cz>
122 * doc/extend.texi: Unify all function declarations in examples
123 where some miss trailing ';'.
125 2021-12-22 Martin Liska <mliska@suse.cz>
127 * doc/extend.texi: Use uppercase letters for SSEx.
129 2021-12-21 Jiang Haochen <haochen.jiang@intel.com>
131 * config/i386/bmiintrin.h (_tzcnt_u16): New intrinsic.
135 2021-12-21 Xionghu Luo <luoxhu@linux.ibm.com>
137 * tree-ssa-loop-im.c (bb_colder_than_loop_preheader): New
139 (get_coldest_out_loop): New function.
140 (determine_max_movement): Use get_coldest_out_loop.
141 (move_computations_worker): Adjust and fix iteration udpate.
142 (class ref_in_loop_hot_body): New functor.
143 (ref_in_loop_hot_body::operator): New.
144 (can_sm_ref_p): Use for_all_locs_in_loop.
145 (fill_coldest_and_hotter_out_loop): New.
146 (tree_ssa_lim_finalize): Free coldest_outermost_loop and
147 hotter_than_inner_loop.
148 (loop_invariant_motion_in_fun): Call fill_coldest_and_hotter_out_loop.
150 2021-12-21 Xionghu Luo <luoxhu@linux.ibm.com>
152 * tree-ssa-loop-split.c (split_loop): Fix incorrect
153 profile_count and probability.
154 (do_split_loop_on_cond): Likewise.
156 2021-12-21 Xionghu Luo <luoxhu@linux.ibm.com>
159 * predict.c (predict_extra_loop_exits): Add loop parameter.
160 (predict_loops): Call with loop argument.
162 2021-12-21 Xionghu Luo <luoxhu@linux.ibm.com>
164 * config/rs6000/altivec.md (altivec_vaddu<VI_char>s): Replace
165 UNSPEC_VADDU with us_plus.
166 (altivec_vadds<VI_char>s): Replace UNSPEC_VADDS with ss_plus.
167 (altivec_vsubu<VI_char>s): Replace UNSPEC_VSUBU with us_minus.
168 (altivec_vsubs<VI_char>s): Replace UNSPEC_VSUBS with ss_minus.
169 (altivec_abss_<mode>): Likewise.
171 2021-12-20 Uroš Bizjak <ubizjak@gmail.com>
174 * config/i386/sse.md (<sse2p4_1>_pinsr<ssemodesuffix>): Add
175 earlyclobber to (x,x,x,i) alternative.
176 (<sse2p4_1>_pinsr<ssemodesuffix> peephole2): Remove.
177 (<sse2p4_1>_pinsr<ssemodesuffix> splitter): Use output
178 operand as a temporary register. Split after reload_completed.
180 2021-12-20 Roger Sayle <roger@nextmovesoftware.com>
181 Uroš Bizjak <ubizjak@gmail.com>
183 * config/i386/i386.md (any_mul_highpart): New code iterator.
184 (sgnprefix, s): Add attribute support for [su]mul_highpart.
185 (<s>mul<mode>3_highpart): Delete expander.
186 (<s>mul<mode>3_highpart, <s>mulsi32_highpart_zext):
187 New define_insn patterns.
188 (define_peephole2): Tweak the register allocation for the above
189 instructions after reload.
191 2021-12-20 Joel Sherrill <joel@rtems.org>
193 * config.gcc: Obsolete m32c-*-rtems* target.
195 2021-12-20 Martin Liska <mliska@suse.cz>
197 * opts.c (default_options_optimization): Support -Oz in -Ox option hints.
199 2021-12-20 Jan Hubicka <hubicka@ucw.cz>
202 * ipa-modref.c (modref_eaf_analysis::analyze_ssa_name): Add deferred
204 (modref_eaf_analysis::propagate): Use it.
206 2021-12-20 liuhongt <hongtao.liu@intel.com>
209 * config/i386/sse.md (*bit_and_float_vector_all_ones): New
212 2021-12-19 Jan Hubicka <hubicka@ucw.cz>
215 * ipa-modref.c (modref_merge_call_site_flags): Fix early exit condition
217 2021-12-19 Patrick Palka <ppalka@redhat.com>
219 * print-tree.c (print_node) <case tcc_declaration>: Dump
222 2021-12-19 Andrew Pinski <apinski@marvell.com>
224 * tree-ssa-phiopt.c (gimple_simplify_phiopt): Annotate the
225 new sequence with the location of the conditional statement.
227 2021-12-18 Roger Sayle <roger@nextmovesoftware.com>
228 Uroš Bizjak <ubizjak@gmail.com>
231 * config/i386/i386.md (any_or_plus): New code iterator.
232 (define_split): Split (HI<<32)|zext(LO) into piece-wise
233 move instructions on !TARGET_64BIT.
234 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4):
235 Restrict to TARGET_64BIT.
237 2021-12-18 Roger Sayle <roger@nextmovesoftware.com>
240 * common.opt (Oz): New command line option.
241 * doc/invoke.texi: Document the new -Oz option.
242 * lto-wrapper.c (merge_and_complain, append_compiler_options):
243 Treat OPT_Oz as synonymous with OPT_Os.
244 * optc-save-gen.awk: Increase maximum value of optimize_size to 2.
245 * opts.c (default_options_optimization) [OPT_Oz]: Handle OPT_Oz
246 just like OPT_Os, except set opt->x_optimize_size to 2.
247 (common_handle_option): Skip OPT_Oz just like OPT_Os.
248 * config/i386/i386.md (*movdi_internal): Use a push/pop sequence
249 for suitable SImode TYPE_IMOV moves when optimize_size > 1.
250 (*movsi_internal): Likewise.
252 2021-12-18 Siddhesh Poyarekar <siddhesh@gotplt.org>
254 PR tree-optimization/103759
255 * tree-object-size.c (unknown, initval): Remove functions.
256 (size_unknown, size_initval, size_unknown_p): Operate directly
259 2021-12-18 François-Xavier Coudert <fxcoudert@gcc.gnu.org>
261 * config/darwin-driver.c: Make version code more future-proof.
262 * config.gcc: Homogeneize darwin versions.
263 * configure.ac: Homogeneize darwin versions.
264 * configure: Regenerate.
266 2021-12-17 Marek Polacek <polacek@redhat.com>
269 * attribs.c (handle_ignored_attributes_option): Create the fake
270 attribute with max_length == -2.
271 (attribute_ignored_p): New overloads.
272 * attribs.h (attribute_ignored_p): Declare them.
273 * tree-core.h (struct attribute_spec): Document that max_length
276 2021-12-17 Segher Boessenkool <segher@kernel.crashing.org>
279 * config/rs6000/rs6000-builtins.def (__builtin_darn): Expand to
280 darn_64_di. Add {32bit} attribute. Return long.
281 (__builtin_darn_32): Expand to darn_32_di. Add {32bit} attribute.
283 (__builtin_darn_raw): Expand to darn_raw_di. Add {32bit} attribute.
285 * config/rs6000/rs6000-call.c (rs6000_expand_builtin): Expand the darn
286 builtins to the _si variants for -m32.
287 * config/rs6000/rs6000.md (UNSPECV_DARN_32, UNSPECV_DARN_RAW): Delete.
288 (UNSPECV_DARN): Update comment.
289 (darn_32, darn_raw, darn): Delete.
290 (darn_32_<mode>, darn_64_<mode>, darn_raw_<mode> for GPR): New.
291 (@darn<mode> for GPR): New.
293 2021-12-17 Bill Schmidt <wschmidt@linux.ibm.com>
295 * config/rs6000/rs6000-overload.def (__builtin_vec_promote): Add second
298 2021-12-17 Richard Sandiford <richard.sandiford@arm.com>
300 PR tree-optimization/103744
301 * tree-vect-stmts.c (vectorizable_load): Handle multi-vector
304 2021-12-17 Martin Liska <mliska@suse.cz>
306 * doc/invoke.texi: Rename to -fstack-protector.
308 2021-12-17 Martin Liska <mliska@suse.cz>
310 * doc/invoke.texi: Fix spelling issues.
312 2021-12-17 Tamar Christina <tamar.christina@arm.com>
314 PR tree-optimization/103741
315 * tree-vect-stmts.c (vectorizable_operation): Check for boolean.
317 2021-12-17 Iain Sandoe <iain@sandoe.co.uk>
319 * config/darwin-driver.c (darwin_driver_init): Exit from the
320 option handling early if the command line is definitely enpty.
321 * config/darwin.h (SUBTARGET_DRIVER_SELF_SPECS): Remove
322 setting for the default content of weak_reference_mismatches.
324 2021-12-17 Iain Sandoe <iain@sandoe.co.uk>
326 * config/rs6000/darwin.h: Drop trailing _x from the
327 builtin_decls array name.
329 2021-12-17 Haochen Jiang <haochen.jiang@intel.com>
332 * config/i386/sse.md (*avx_cmp<mode>3_lt, *avx_cmp<mode>3_ltint):
333 Remove MEM_P restriction and add force_reg for operands[2].
334 (*avx_cmp<mode>3_ltint_not): Add new define_insn_and_split.
336 2021-12-17 Siddhesh Poyarekar <siddhesh@gotplt.org>
338 * builtins.def (BUILT_IN_DYNAMIC_OBJECT_SIZE): New builtin.
339 * tree-object-size.h: Move object size type bits enum from
340 tree-object-size.c and add new value OST_DYNAMIC.
341 * builtins.c (expand_builtin, fold_builtin_2): Handle it.
342 (fold_builtin_object_size): Handle new builtin and adjust for
343 change to compute_builtin_object_size.
344 * tree-object-size.c: Include builtins.h.
345 (compute_builtin_object_size): Adjust.
346 (early_object_sizes_execute_one,
347 dynamic_object_sizes_execute_one): New functions.
348 (object_sizes_execute): Rename insert_min_max_p argument to
349 early. Handle BUILT_IN_DYNAMIC_OBJECT_SIZE and call the new
351 * doc/extend.texi (__builtin_dynamic_object_size): Document new
354 2021-12-17 Siddhesh Poyarekar <siddhesh@gotplt.org>
356 * tree-object-size.h (compute_builtin_object_size): Return tree
357 instead of HOST_WIDE_INT.
358 * builtins.c (fold_builtin_object_size): Adjust.
359 * gimple-fold.c (gimple_fold_builtin_strncat): Likewise.
360 * ubsan.c (instrument_object_size): Likewise.
361 * tree-object-size.c (object_size): New structure.
362 (object_sizes): Change type to vec<object_size>.
363 (initval): New function.
365 (size_unknown_p, size_initval, size_unknown): New functions.
366 (object_sizes_unknown_p): Use it.
367 (object_sizes_get): Return tree.
368 (object_sizes_initialize): Rename from object_sizes_set_force
369 and set VAL parameter type as tree. Add new parameter WHOLEVAL.
370 (object_sizes_set): Set VAL parameter type as tree and adjust
371 implementation. Add new parameter WHOLEVAL.
372 (size_for_offset): New function.
373 (decl_init_size): Adjust comment.
374 (addr_object_size): Change PSIZE parameter to tree and adjust
375 implementation. Add new parameter PWHOLESIZE.
376 (alloc_object_size): Return tree.
377 (compute_builtin_object_size): Return tree in PSIZE.
378 (expr_object_size, call_object_size, unknown_object_size):
379 Adjust for object_sizes_set change.
380 (merge_object_sizes): Drop OFFSET parameter and adjust
381 implementation for tree change.
382 (plus_stmt_object_size): Call collect_object_sizes_for directly
383 instead of merge_object_size and call size_for_offset to get net
385 (cond_expr_object_size, collect_object_sizes_for,
386 object_sizes_execute): Adjust for change of type from
387 HOST_WIDE_INT to tree.
388 (check_for_plus_in_loops_1): Likewise and skip non-positive
391 2021-12-17 Jason Merrill <jason@redhat.com>
394 * common.opt (fabi-version): Add v17.
396 2021-12-16 Uroš Bizjak <ubizjak@gmail.com>
399 * config/i386/i386.h (VALID_AVX256_REG_MODE): Add V16HFmode.
400 (VALID_AVX256_REG_OR_OI_VHF_MODE): Replace with ...
401 (VALID_AVX256_REG_OR_OI_MODE): ... this. Remove V16HFmode.
402 (VALID_AVX512F_SCALAR_MODE): Remove HImode and HFmode.
403 (VALID_AVX512FP16_SCALAR_MODE): New.
404 (VALID_AVX512F_REG_MODE): Add V32HFmode.
405 (VALID_SSE2_REG_MODE): Add V8HFmode, V4HFmode and V2HFmode.
406 (VALID_SSE2_REG_VHF_MODE): Remove.
407 (VALID_INT_MODE_P): Add V2HFmode.
408 * config/i386/i386.c (function_arg_advance_64):
409 Remove explicit mention of V16HFmode and V32HFmode.
410 (ix86_hard_regno_mode_ok): Remove explicit mention of XImode
411 and V32HFmode, use VALID_AVX512F_REG_OR_XI_MODE instead.
412 Use VALID_AVX512FP_SCALAR_MODE for TARGET_aVX512FP16.
413 Use VALID_AVX256_REG_OR_OI_MODE instead of
414 VALID_AVX256_REG_OR_OI_VHF_MODE and VALID_SSE2_REG_MODE instead
415 of VALID_SSE2_REG_VHF_MODE.
416 (ix86_set_reg_reg_cost): Remove usge of VALID_AVX512FP16_REG_MODE.
417 (ix86_vector_mode_supported): Ditto.
419 2021-12-16 Martin Liska <mliska@suse.cz>
422 * attribs.c (decl_attributes): Check if
423 target_option_current_node is changed.
425 2021-12-16 Martin Liska <mliska@suse.cz>
427 * doc/invoke.texi: Add missing dash.
429 2021-12-16 Bill Schmidt <wschmidt@linux.ibm.com>
431 * config/rs6000/rs6000-c.c (resolution): New enum.
432 (resolve_vec_mul): New function.
433 (resolve_vec_cmpne): Likewise.
434 (resolve_vec_adde_sube): Likewise.
435 (resolve_vec_addec_subec): Likewise.
436 (resolve_vec_splats): Likewise.
437 (resolve_vec_extract): Likewise.
438 (resolve_vec_insert): Likewise.
439 (resolve_vec_step): Likewise.
440 (find_instance): Likewise.
441 (altivec_resolve_overloaded_builtin): Many cleanups. Call factored-out
442 functions. Move variable declarations closer to uses. Add commentary.
443 Remove unnecessary levels of braces. Avoid use of gotos. Change
444 misleading variable names. Use switches over if-else-if chains.
446 2021-12-16 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
449 * config/aarch64/aarch64-simd.md (aarch64_movv8di): Allow big endian
450 targets to move V8DI.
452 2021-12-16 Tobias Burnus <tobias@codesourcery.com>
455 * gcc.c (check_offload_target_name): Add 'default' and 'disable'
456 to the candidate list.
458 2021-12-15 Iain Sandoe <iain@sandoe.co.uk>
460 * configure: Regenerate.
462 2021-12-15 Roger Sayle <roger@nextmovesoftware.com>
464 * config/nvptx/nvptx-opts.h (ptx_isa): PTX_ISA_SM75 and PTX_ISA_SM80
466 * config/nvptx/nvptx.opt: Add sm_75 and sm_80 to -misa.
467 * config/nvptx/nvptx.h (TARGET_SM75, TARGET_SM80):
468 New helper macros to conditionalize functionality on target ISA.
469 * config/nvptx/nvptx-c.c (nvptx_cpu_cpp_builtins): Add __PTX_SM__
470 support for the new ISA levels.
471 * config/nvptx/nvptx.c (nvptx_file_start): Add support for TARGET_SM75
473 * config/nvptx/nvptx.md (define_c_enum "unspec"): New UNSPEC_TANH.
474 (define_mode_iterator HSFM): New iterator for HFmode and SFmode.
475 (exp2hf2): New define_insn controlled by TARGET_SM75.
476 (tanh<mode>2): New define_insn controlled by TARGET_SM75.
477 (sminhf3, smaxhf3): New define_isnns controlled by TARGET_SM80.
479 2021-12-15 Tom de Vries <tdevries@suse.de>
481 * config/nvptx/nvptx-opts.h (enum ptx_version): Add PTX_VERSION_7_0.
482 * config/nvptx/nvptx.c (nvptx_file_start): Handle TARGET_PTX_7_0.
483 * config/nvptx/nvptx.h (TARGET_PTX_7_0): New macro.
484 * config/nvptx/nvptx.opt (ptx_version): Add 7.0.
486 2021-12-15 Richard Sandiford <richard.sandiford@arm.com>
487 Tamar Christina <tamar.christina@arm.com>
490 * config/aarch64/aarch64.c (aarch64_short_vector_p): Return false
491 for structure modes, rather than ignoring the type in that case.
493 2021-12-15 Tamar Christina <tamar.christina@arm.com>
495 PR rtl-optimization/103350
496 * ree.c (add_removable_extension): Don't stop at first definition but
499 2021-12-15 Jakub Jelinek <jakub@redhat.com>
502 * dwarf2cfi.c (dwf_cfa_reg): Remove gcc_assert.
503 (operator==, operator!=): New overloaded operators.
504 (dwarf2out_frame_debug_adjust_cfa, dwarf2out_frame_debug_cfa_offset,
505 dwarf2out_frame_debug_expr): Compare vars with cfa_reg type directly
506 with REG rtxes rather than with dwf_cfa_reg results on those REGs.
507 (create_cie_data): Use stack_pointer_rtx instead of
508 gen_rtx_REG (Pmode, STACK_POINTER_REGNUM).
509 (execute_dwarf2_frame): Use hard_frame_pointer_rtx instead of
510 gen_rtx_REG (Pmode, HARD_FRAME_POINTER_REGNUM).
512 2021-12-15 Martin Liska <mliska@suse.cz>
515 * config/i386/i386-builtins.c (fold_builtin_cpu): Compare to 0
516 as API expects that non-zero values are returned (do that
518 For "avx512vbmi2" argument, we return now 1 << 31, which is a
519 negative integer value.
521 2021-12-15 Haochen Jiang <haochen.jiang@intel.com>
524 * config/i386/predicates.md (const_vector_operand):
526 * config/i386/sse.md(<insn><mode>3<mask_name>):
527 Add new define_split below.
529 2021-12-15 Michael Meissner <meissner@the-meissners.org>
531 * config/rs6000/rs6000.md (UNSPEC_XXSPLTIDP_CONST): New unspec.
532 (UNSPEC_XXSPLTIW_CONST): New unspec.
533 (movsf_hardfloat): Add support for generating XXSPLTIDP.
534 (mov<mode>_hardfloat32): Likewise.
535 (mov<mode>_hardfloat64): Likewise.
536 (xxspltidp_<mode>_internal): New insns.
537 (xxspltiw_<mode>_internal): New insns.
538 (splitters for SF/DFmode): Add new splitters for XXSPLTIDP.
540 2021-12-15 Michael Meissner <meissner@the-meissners.org>
542 * config/rs6000/predicates.md (easy_fp_constant): Add support for
543 generating XXSPLTIDP.
544 (vsx_prefixed_constant): Likewise.
545 (easy_vector_constant): Likewise.
546 * config/rs6000/rs6000-protos.h (constant_generates_xxspltidp):
548 * config/rs6000/rs6000.c (output_vec_const_move): Add support for
549 generating XXSPLTIDP.
550 (prefixed_xxsplti_p): Likewise.
551 (constant_generates_xxspltidp): New function.
552 * config/rs6000/rs6000.opt (-msplat-float-constant): New debug option.
554 2021-12-15 Michael Meissner <meissner@linux.ibm.com>
556 * config/rs6000/constraints.md (eP): Update comment.
557 * config/rs6000/predicates.md (easy_fp_constant): Add support for
559 (vsx_prefixed_constant): New predicate.
560 (easy_vector_constant): Add support for
562 * config/rs6000/rs6000-protos.h (prefixed_xxsplti_p): New
564 (constant_generates_xxspltiw): Likewise.
565 * config/rs6000/rs6000.c (xxspltib_constant_p): Generate XXSPLTIW
566 if possible instead of XXSPLTIB and sign extending the constant.
567 (output_vec_const_move): Add support for XXSPLTIW.
568 (prefixed_xxsplti_p): New function.
569 (constant_generates_xxspltiw): New function.
570 * config/rs6000/rs6000.md (prefixed attribute): Add support to
571 mark XXSPLTI* instructions as being prefixed.
572 * config/rs6000/rs6000.opt (-msplat-word-constant): New debug
574 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Add support for
575 generating XXSPLTIW or XXSPLTIDP.
576 (vsx_mov<mode>_32bit): Likewise.
577 * doc/md.texi (PowerPC and IBM RS6000 constraints): Document the
580 2021-12-15 Michael Meissner <meissner@the-meissners.org>
582 * config/rs6000/constraints.md (eQ): New constraint.
583 * config/rs6000/predicates.md (easy_fp_constant): Add support for
584 generating the LXVKQ instruction.
585 (easy_vector_constant_ieee128): New predicate.
586 (easy_vector_constant): Add support for generating the LXVKQ
588 * config/rs6000/rs6000-protos.h (constant_generates_lxvkq): New
590 * config/rs6000/rs6000.c (output_vec_const_move): Add support for
592 (constant_generates_lxvkq): New function.
593 * config/rs6000/rs6000.opt (-mieee128-constant): New debug
595 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Add support for
597 (vsx_mov<mode>_32bit): Likewise.
598 * doc/md.texi (PowerPC and IBM RS6000 constraints): Document the
601 2021-12-15 Michael Meissner <meissner@the-meissners.org>
603 * config/rs6000/rs6000-protos.h (VECTOR_128BIT_BITS): New macro.
604 (VECTOR_128BIT_BYTES): Likewise.
605 (VECTOR_128BIT_HALF_WORDS): Likewise.
606 (VECTOR_128BIT_WORDS): Likewise.
607 (VECTOR_128BIT_DOUBLE_WORDS): Likewise.
608 (vec_const_128bit_type): New structure type.
609 (vec_const_128bit_to_bytes): New declaration.
610 * config/rs6000/rs6000.c (constant_int_to_128bit_vector): New
612 (constant_fp_to_128bit_vector): New helper function.
613 (vec_const_128bit_to_bytes): New function.
615 2021-12-15 Alexandre Oliva <oliva@adacore.com>
618 * builtins.c (try_store_by_multiple_pieces): Drop address
619 conversion to ptr_mode.
621 2021-12-15 Alexandre Oliva <oliva@adacore.com>
624 * builtins.c (try_store_by_multiple_pieces): Fail if min_len
625 is greater than max_len.
627 2021-12-14 liuhongt <hongtao.liu@intel.com>
630 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Check
631 is_gimple_assign before gimple_assign_rhs_code.
633 2021-12-14 Sören Tempel <soeren@soeren-tempel.net>
635 * ginclude/stddef.h (__DEFINED_ptrdiff_t): Add support for musl
636 libc typedef macro guard.
637 (__DEFINED_size_t): Ditto.
638 (__DEFINED_wchar_t): Ditto.
640 2021-12-14 JoJo R <rjiejie@linux.alibaba.com>
642 * regrename.c (find_rename_reg): Return satisfied regno
643 if instruction is noop move.
645 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
647 * config/rs6000/rs6000-c.c (altivec_build_resolved_builtin): Rename
648 rs6000_builtin_decls_x to rs6000_builtin_decls.
649 (altivec_resolve_overloaded_builtin): Likewise. Also rename
650 rs6000_builtin_info_x to rs6000_builtin_info.
651 * config/rs6000/rs6000-call.c (rs6000_invalid_builtin): Rename
652 rs6000_builtin_info_x to rs6000_builtin_info.
653 (rs6000_builtin_is_supported): Likewise.
654 (rs6000_gimple_fold_mma_builtin): Likewise. Also rename
655 rs6000_builtin_decls_x to rs6000_builtin_decls.
656 (rs6000_gimple_fold_builtin): Rename rs6000_builtin_info_x to
658 (cpu_expand_builtin): Likewise.
659 (rs6000_expand_builtin): Likewise.
660 (rs6000_init_builtins): Likewise. Also rename rs6000_builtin_decls_x
661 to rs6000_builtin_decls.
662 (rs6000_builtin_decl): Rename rs6000_builtin_decls_x to
663 rs6000_builtin_decls.
664 * config/rs6000/rs6000-gen-builtins.c (write_decls): In generated code,
665 rename rs6000_builtin_decls_x to rs6000_builtin_decls, and rename
666 rs6000_builtin_info_x to rs6000_builtin_info.
667 (write_bif_static_init): In generated code, rename
668 rs6000_builtin_info_x to rs6000_builtin_info.
669 (write_init_bif_table): In generated code, rename
670 rs6000_builtin_decls_x to rs6000_builtin_decls, and rename
671 rs6000_builtin_info_x to rs6000_builtin_info.
672 (write_init_ovld_table): In generated code, rename
673 rs6000_builtin_decls_x to rs6000_builtin_decls.
674 (write_init_file): Likewise.
675 * config/rs6000/rs6000.c (rs6000_builtin_vectorized_function):
677 (rs6000_builtin_md_vectorized_function): Likewise.
678 (rs6000_builtin_reciprocal): Likewise.
679 (add_condition_to_bb): Likewise.
680 (rs6000_atomic_assign_expand_fenv): Likewise.
682 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
684 * config/rs6000/rs6000-c.c (altivec_resolve_new_overloaded_builtin):
685 Remove forward declaration.
686 (rs6000_new_builtin_type_compatible): Rename to
687 rs6000_builtin_type_compatible.
688 (rs6000_builtin_type_compatible): Remove.
689 (altivec_resolve_overloaded_builtin): Remove.
690 (altivec_build_new_resolved_builtin): Rename to
691 altivec_build_resolved_builtin.
692 (altivec_resolve_new_overloaded_builtin): Rename to
693 altivec_resolve_overloaded_builtin. Remove static keyword. Adjust
694 called function names.
695 * config/rs6000/rs6000-call.c (rs6000_expand_new_builtin): Remove
697 (rs6000_gimple_fold_new_builtin): Likewise.
698 (rs6000_invalid_new_builtin): Rename to rs6000_invalid_builtin.
699 (rs6000_gimple_fold_builtin): Remove.
700 (rs6000_new_builtin_valid_without_lhs): Rename to
701 rs6000_builtin_valid_without_lhs.
702 (rs6000_new_builtin_is_supported): Rename to
703 rs6000_builtin_is_supported.
704 (rs6000_gimple_fold_new_mma_builtin): Rename to
705 rs6000_gimple_fold_mma_builtin.
706 (rs6000_gimple_fold_new_builtin): Rename to
707 rs6000_gimple_fold_builtin. Remove static keyword. Adjust called
709 (rs6000_expand_builtin): Remove.
710 (new_cpu_expand_builtin): Rename to cpu_expand_builtin.
711 (new_mma_expand_builtin): Rename to mma_expand_builtin.
712 (new_htm_spr_num): Rename to htm_spr_num.
713 (new_htm_expand_builtin): Rename to htm_expand_builtin. Change name
715 (rs6000_expand_new_builtin): Rename to rs6000_expand_builtin. Remove
716 static keyword. Adjust called function names.
717 (rs6000_new_builtin_decl): Rename to rs6000_builtin_decl. Remove
719 (rs6000_builtin_decl): Remove.
720 * config/rs6000/rs6000-gen-builtins.c (write_decls): In gnerated code,
721 rename rs6000_new_builtin_is_supported to rs6000_builtin_is_supported.
722 * config/rs6000/rs6000-internal.h (rs6000_invalid_new_builtin): Rename
723 to rs6000_invalid_builtin.
724 * config/rs6000/rs6000.c (rs6000_new_builtin_vectorized_function):
725 Rename to rs6000_builtin_vectorized_function.
726 (rs6000_new_builtin_md_vectorized_function): Rename to
727 rs6000_builtin_md_vectorized_function.
728 (rs6000_builtin_vectorized_function): Remove.
729 (rs6000_builtin_md_vectorized_function): Remove.
731 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
733 * config/rs6000/rs6000-builtin.def: Delete.
734 * config/rs6000/rs6000-call.c (builtin_compatibility): Delete.
735 (builtin_description): Delete.
736 (builtin_hash_struct): Delete.
737 (builtin_hasher): Delete.
738 (builtin_hash_table): Delete.
739 (builtin_hasher::hash): Delete.
740 (builtin_hasher::equal): Delete.
741 (rs6000_builtin_info_type): Delete.
742 (rs6000_builtin_info): Delete.
743 (bdesc_compat): Delete.
744 (bdesc_3arg): Delete.
745 (bdesc_4arg): Delete.
747 (bdesc_2arg): Delete.
748 (bdesc_altivec_preds): Delete.
750 (bdesc_1arg): Delete.
751 (bdesc_0arg): Delete.
754 (rs6000_overloaded_builtin_p): Delete.
755 (rs6000_overloaded_builtin_name): Delete.
756 (htm_spr_num): Delete.
757 (rs6000_builtin_is_supported_p): Delete.
758 (rs6000_gimple_fold_mma_builtin): Delete.
759 (gt-rs6000-call.h): Remove include directive.
760 * config/rs6000/rs6000-protos.h (rs6000_overloaded_builtin_p): Delete.
761 (rs6000_builtin_is_supported_p): Delete.
762 (rs6000_overloaded_builtin_name): Delete.
763 * config/rs6000/rs6000.c (rs6000_builtin_decls): Delete.
764 (rs6000_debug_reg_global): Remove reference to RS6000_BUILTIN_COUNT.
765 * config/rs6000/rs6000.h (rs6000_builtins): Delete.
766 (altivec_builtin_types): Delete.
767 (rs6000_builtin_decls): Delete.
768 * config/rs6000/t-rs6000 (TM_H): Don't add rs6000-builtin.def.
770 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
772 * config/rs6000/rs6000-builtin-new.def: Rename to...
773 * config/rs6000/rs6000-builtins.def: ...this.
774 * config/rs6000/rs6000-gen-builtins.c: Adjust header commentary.
775 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Rename
776 rs6000-builtin-new.def to rs6000-builtins.def.
777 (rs6000-builtins.c): Likewise.
779 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
781 * config/rs6000/rs6000-call.c (altivec_overloaded_builtins): Remove.
782 * config/rs6000/rs6000.h (altivec_overloaded_builtins): Remove.
784 2021-12-14 Peter Bergner <bergner@linux.ibm.com>
787 * config/rs6000/mma.md (UNSPEC_MMA_ASSEMBLE): Rename unspec from this...
788 (UNSPEC_VSX_ASSEMBLE): ...to this.
789 (UNSPECV_MMA_ASSEMBLE): New unspecv.
790 (vsx_assemble_pair): Use UNSPEC_VSX_ASSEMBLE.
791 (*vsx_assemble_pair): Likewise.
792 (mma_assemble_acc): Use UNSPECV_MMA_ASSEMBLE.
793 (*mma_assemble_acc): Likewise.
794 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle
795 UNSPEC_VOLATILE. Use UNSPEC_VSX_ASSEMBLE and UNSPECV_MMA_ASSEMBLE.
797 2021-12-14 Uroš Bizjak <ubizjak@gmail.com>
800 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate)
801 <case E_V8HFmode>: Implement for TARGET_SSE2.
802 <case E_V16HFmode>: Implement for TARGET_AVX.
803 <case E_V32HFmode>: Implement for TARGET_AVX512F.
804 (ix86_expand_vector_set_var): Handle V32HFmode
805 without TARGET_AVX512BW.
806 (ix86_expand_vector_extract)
807 <case E_V8HFmode>: Implement for TARGET_SSE2.
808 <case E_V16HFmode>: Implement for TARGET_AVX.
809 <case E_V32HFmode>: Implement for TARGET_AVX512BW.
810 (expand_vec_perm_broadcast_1) <case E_V8HFmode>: New.
811 * config/i386/sse.md (VI12HF_AVX512VL): Remove
812 TARGET_AVX512FP16 condition.
815 (avx_vbroadcastf128_<mode>): Use V_256H mode iterator.
817 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
819 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Remove
820 test for new_builtins_are_live and simplify.
821 * config/rs6000/rs6000-c.c (altivec_build_resolved_builtin): Remove
823 (altivec_resolve_overloaded_builtin): Remove test for
824 new_builtins_are_live and simplify.
825 * config/rs6000/rs6000-call.c (altivec_init_builtins): Remove forward
827 (builtin_function_type): Likewise.
828 (rs6000_common_init_builtins): Likewise.
829 (htm_init_builtins): Likewise.
830 (mma_init_builtins): Likewise.
831 (def_builtin): Remove dead function.
832 (rs6000_expand_zeroop_builtin): Likewise.
833 (rs6000_expand_mtfsf_builtin): Likewise.
834 (rs6000_expand_mtfsb_builtin): Likewise.
835 (rs6000_expand_set_fpscr_rn_builtin): Likewise.
836 (rs6000_expand_set_fpscr_drn_builtin): Likewise.
837 (rs6000_expand_unop_builtin): Likewise.
838 (altivec_expand_abs_builtin): Likewise.
839 (rs6000_expand_binop_builtin): Likewise.
840 (altivec_expand_lxvr_builtin): Likewise.
841 (altivec_expand_lv_builtin): Likewise.
842 (altivec_expand_stxvl_builtin): Likewise.
843 (altivec_expand_stv_builtin): Likewise.
844 (mma_expand_builtin): Likewise.
845 (htm_expand_builtin): Likewise.
846 (cpu_expand_builtin): Likewise.
847 (rs6000_expand_quaternop_builtin): Likewise.
848 (rs6000_expand_ternop_builtin): Likewise.
849 (altivec_expand_dst_builtin): Likewise.
850 (altivec_expand_vec_sel_builtin): Likewise.
851 (altivec_expand_builtin): Likewise.
852 (rs6000_invalid_builtin): Likewise.
853 (rs6000_builtin_valid_without_lhs): Likewise.
854 (rs6000_gimple_fold_builtin): Remove test for new_builtins_are_live and
856 (rs6000_expand_builtin): Likewise.
857 (rs6000_init_builtins): Remove tests for new_builtins_are_live and
859 (rs6000_builtin_decl): Likewise.
860 (altivec_init_builtins): Remove dead function.
861 (mma_init_builtins): Likewise.
862 (htm_init_builtins): Likewise.
863 (builtin_quaternary_function_type): Likewise.
864 (builtin_function_type): Likewise.
865 (rs6000_common_init_builtins): Likewise.
866 * config/rs6000/rs6000-gen-builtins.c (write_header_file): Don't
867 declare new_builtins_are_live.
868 (write_init_bif_table): In generated code, remove test for
869 new_builtins_are_live and simplify.
870 (write_init_ovld_table): Likewise.
871 (write_init_file): Don't initialize new_builtins_are_live.
872 * config/rs6000/rs6000.c (rs6000_builtin_vectorized_function): Remove
873 test for new_builtins_are_live and simplify.
874 (rs6000_builtin_md_vectorized_function): Likewise.
875 (rs6000_builtin_reciprocal): Likewise.
876 (add_condition_to_bb): Likewise.
877 (rs6000_atomic_assign_expand_fenv): Likewise.
879 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
882 * config/rs6000/rs6000-builtin-new.def (__builtin_altivec_vcmpequd):
883 Move to power8-vector stanza.
884 (__builtin_altivec_vcmpequd_p): Likewise.
885 (__builtin_altivec_vcmpgtsd): Likewise.
886 (__builtin_altivec_vcmpgtsd_p): Likewise.
887 (__builtin_altivec_vcmpgtud): Likewise.
888 (__builtin_altivec_vcmpgtud_p): Likewise.
890 2021-12-14 Bill Schmidt <wschmidt@linux.ibm.com>
893 * config/rs6000/rs6000-builtin-new.def (__builtin_pack_longdouble): Add
895 (__builtin_unpack_longdouble): Likewise.
896 * config/rs6000/rs6000-call.c (rs6000_expand_new_builtin): Add special
897 handling for ibmld attribute.
898 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add isibmld.
899 (parse_bif_attrs): Handle ibmld.
900 (write_decls): Likewise.
901 (write_bif_static_init): Likewise.
903 2021-12-14 Jan Hubicka <hubicka@ucw.cz>
906 * ipa-modref-tree.c (modref_access_node::range_info_useful_p): Handle
907 MODREF_GLOBAL_MEMORY_PARM.
908 (modref_access_node::dump): Likewise.
909 (modref_access_node::get_call_arg): Likewise.
910 * ipa-modref-tree.h (enum modref_special_parms): Add
911 MODREF_GLOBAL_MEMORY_PARM.
912 (modref_access_node::useful_for_kill): Handle
913 MODREF_GLOBAL_MEMORY_PARM.
914 (modref:tree::merge): Add promote_unknown_to_global.
915 * ipa-modref.c (verify_arg):New function.
916 (may_access_nonescaping_parm_p): New function.
917 (modref_access_analysis::record_global_memory_load): New member
919 (modref_access_analysis::record_global_memory_store): Likewise.
920 (modref_access_analysis::process_fnspec): Distingush global and local
922 (modref_access_analysis::analyze_call): Likewise.
923 * tree-ssa-alias.c (ref_may_access_global_memory_p): New function.
924 (modref_may_conflict): Use it.
926 2021-12-14 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
928 * config/aarch64/aarch64-builtins.c (enum aarch64_builtins):
929 Define AARCH64_LS64_BUILTIN_LD64B, AARCH64_LS64_BUILTIN_ST64B,
930 AARCH64_LS64_BUILTIN_ST64BV, AARCH64_LS64_BUILTIN_ST64BV0.
931 (aarch64_init_ls64_builtin_decl): Helper function.
932 (aarch64_init_ls64_builtins): Helper function.
933 (aarch64_init_ls64_builtins_types): Helper function.
934 (aarch64_general_init_builtins): Init LS64 intrisics for
936 (aarch64_expand_builtin_ls64): LS64 intrinsics expander.
937 (aarch64_general_expand_builtin): Handle aarch64_expand_builtin_ls64.
938 (ls64_builtins_data): New helper struct.
939 (v8di_UP): New define.
940 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
942 * config/aarch64/aarch64.c (aarch64_classify_address): Enforce the
943 V8DI range (7-bit signed scaled) for both ends of the range.
944 * config/aarch64/aarch64-simd.md (movv8di): New pattern.
945 (aarch64_movv8di): New pattern.
946 * config/aarch64/aarch64.h (AARCH64_ISA_LS64): New define.
947 (TARGET_LS64): New define.
948 * config/aarch64/aarch64.md: Add UNSPEC_LD64B, UNSPEC_ST64B,
949 UNSPEC_ST64BV and UNSPEC_ST64BV0.
950 (ld64b): New define_insn.
951 (st64b): New define_insn.
952 (st64bv): New define_insn.
953 (st64bv0): New define_insn.
954 * config/aarch64/arm_acle.h (data512_t): New type derived from
956 (__arm_data512_t): New internal type.
957 (__arm_ld64b): New intrinsic.
958 (__arm_st64b): New intrinsic.
959 (__arm_st64bv): New intrinsic.
960 (__arm_st64bv0): New intrinsic.
961 * config/arm/types.md: Add new type ls64.
963 2021-12-14 Olivier Hainque <hainque@adacore.com>
965 * config/i386/t-vxworks: Drop the fPIC multilibs.
967 2021-12-14 Fred Konrad <konrad@adacore.com>
969 * config/rs6000/t-vxworks: Drop the fPIC multilib.
971 2021-12-13 Vladimir N. Makarov <vmakarov@redhat.com>
974 * ira-costs.c (record_operand_costs): Do not take pseudo class
975 calculated on the 1st iteration into account when processing move
976 involving the pseudo and a hard register.
978 2021-12-13 Roger Sayle <roger@nextmovesoftware.com>
980 * config/i386/i386.md (define_split any_or:SWI248 -> orb %?h):
981 Optimize the case where the integer constant operand is zero.
983 2021-12-13 Doug Rupp <rupp@adacore.com>
984 Olivier Hainque <hainque@adacore.com>
986 * config/vxworks.h (VXWORKS_LINK_OS_SPEC): New spec.
987 (VXWORKS_BASE_LINK_SPEC): New spec, using the former.
988 (VXWORKS_EXTRA_LINK_SPEC): New spec for old and new VxWorks.
989 (VXWORKS_LINK_SPEC): Combo of BASE and EXTRA specs.
990 * config/rs6000/vxworks.h (VXWORKS_LINK_OS_SPEC): Empty.
991 (LINK_OS_EXTRA_SPEC32): Use VXWORKS_LINK_SPEC.
992 (LINK_OS_EXTRA_SPEC64): Likewise.
994 2021-12-13 Fred Konrad <konrad@adacore.com>
995 Olivier Hainque <hainque@adacore.com>
997 * config/vxworks.h (VXWORKS_BASE_LIBS_RTP): Guard -lc_internal
998 on !shared+!non-static and document.
999 (VXWORKS_LIB_SPEC): Remove the bits intended to drag the
1000 init/fini functions from libc_internal in the shared lib case.
1001 (VX_CRTBEGIN_SPEC/VX_CRTEND_SPEC): Use vxcrtstuff objects also in
1002 configurations with shared lib and INITFINI_ARRAY support.
1004 2021-12-13 Fred Konrad <konrad@adacore.com>
1005 Olivier Hainque <hainque@adacore.com>
1007 * config/vx-common.h: Define REAL_LIBGCC_SPEC since the
1008 '-non-static' option is not standard.
1009 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Implement the LIBGCC_SPEC
1010 since REAL_LIBGCC_SPEC is used now.
1011 (STARTFILE_PREFIX_SPEC): Use the PIC VSB when building shared libraries
1012 or non-static binaries.
1014 2021-12-13 Jan Hubicka <hubicka@ucw.cz>
1016 * common.opt: Add -fipa-strict-aliasing.
1017 * doc/invoke.texi: Document -fipa-strict-aliasing.
1018 * ipa-modref.c (modref_access_analysis::record_access): Honor
1019 -fipa-strict-aliasing.
1020 (modref_access_analysis::record_access_lto): Likewise.
1022 2021-12-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1024 * config/aarch64/aarch64-arches.def (armv8.8-a): Define.
1025 * config/aarch64/aarch64.h (AARCH64_FL_V8_8): Define.
1026 (AARCH64_FL_FOR_ARCH8_8): Define.
1027 * doc/invoke.texi: Document -march=armv8.8-a.
1029 2021-12-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1031 * config/aarch64/aarch64.c (aarch64_expand_setmem_mops): Define.
1032 (aarch64_expand_setmem): Adjust for TARGET_MOPS.
1033 * config/aarch64/aarch64.h (CLEAR_RATIO): Adjust for TARGET_MOPS.
1034 (SET_RATIO): Likewise.
1035 * config/aarch64/aarch64.md ("unspec"): Add UNSPEC_SETMEM.
1036 (aarch64_setmemdi): Define.
1037 (setmemdi): Adjust for TARGET_MOPS.
1038 * config/aarch64/aarch64.opt (aarch64-mops-memset-size-threshold):
1041 2021-12-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1043 * config/aarch64/aarch64.md (aarch64_movmemdi): Define.
1045 (unspec): Add UNSPEC_MOVMEM.
1046 * config/aarch64/aarch64.opt (aarch64-mops-memmove-size-threshold):
1049 2021-12-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1051 * config/aarch64/aarch64-option-extensions.def (mops): Define.
1052 * config/aarch64/aarch64.c (aarch64_expand_cpymem_mops): Define.
1053 (aarch64_expand_cpymem): Define.
1054 * config/aarch64/aarch64.h (AARCH64_FL_MOPS): Define.
1055 (AARCH64_ISA_MOPS): Define.
1056 (TARGET_MOPS): Define.
1057 (MOVE_RATIO): Adjust for TARGET_MOPS.
1058 * config/aarch64/aarch64.md ("unspec"): Add UNSPEC_CPYMEM.
1059 (aarch64_cpymemdi): New pattern.
1060 (cpymemdi): Adjust for TARGET_MOPS.
1061 * config/aarch64/aarch64.opt (aarch64-mops-memcpy-size-threshol):
1063 * doc/invoke.texi (AArch64 Options): Document +mops.
1065 2021-12-13 Martin Liska <mliska@suse.cz>
1068 * ipa-inline.c (can_inline_edge_p): Move logic checking
1069 no_profile_instrument_function logic to ...
1070 (can_early_inline_edge_p): ... here.
1072 2021-12-13 Olivier Hainque <hainque@adacore.com>
1074 * config/vxworks/_yvals.h: #include yvals.h also if
1077 2021-12-13 Olivier Hainque <hainque@adacore.com>
1079 * config/vxworks.h (VXWORKS_OS_CPP_BUILTINS): Define
1082 2021-12-13 Olivier Hainque <hainque@adacore.com>
1084 * config/t-vxworks: Clear NATIVE_SYSTEM_HEADER_DIR.
1085 * config/vxworks.h (SYSROOT_HEADERS_SUFFIX_SPEC): Define, for
1086 VxWorks 7 and earlier.
1087 (VXWORKS_ADDITIONAL_CPP_SPEC): Simplify accordingly.
1088 (STARTFILE_PREFIX_SPEC): Adjust accordingly.
1089 * config/rs6000/vxworks.h (STARTFILE_PREFIX_SPEC): Adjust.
1091 2021-12-13 Martin Liska <mliska@suse.cz>
1093 * doc/extend.texi: Use @item for the first @itemx entry.
1095 2021-12-13 Jakub Jelinek <jakub@redhat.com>
1097 * machmode.h (gt_pch_nx): Use gt_pointer_operator as type of second
1098 argument instead of equivalent void (*) (void *, void *, void *).
1099 * poly-int.h (gt_pch_nx): Likewise.
1100 * wide-int.h (gt_pch_nx): Likewise.
1101 * config/aarch64/aarch64-sve-builtins.cc (gt_pch_nx): Likewise.
1103 2021-12-13 Jan Hubicka <hubicka@ucw.cz>
1106 * ipa-fnsummary.c (evaluate_conditions_for_known_args): Do not ICE
1107 on ternary expression.
1109 2021-12-13 Kewen Lin <linkw@linux.ibm.com>
1112 * attribs.c (decl_attributes): Check if target options change and
1113 create one node if so.
1115 2021-12-12 Jonathan Wakely <jwakely@redhat.com>
1117 * Makefile.in: Remove unique-ptr-tests.o.
1118 * selftest-run-tests.c (selftest::run_tests): Remove
1119 unique_ptr_tests_cc_tests.
1120 * selftest.h (unique_ptr_tests_cc_tests): Remove.
1121 * system.h: Check INCLUDE_MEMORY instead of INCLUDE_UNIQUE_PTR
1122 and include <memory> instead of "unique-ptr.h".
1123 * unique-ptr-tests.cc: Removed.
1125 2021-12-12 Roger Sayle <roger@nextmovesoftware.com>
1127 * config/nvptx/nvptx-opts.h (ptx_isa): Add PTX_ISA_SM53 ISA level
1129 * config/nvptx/nvptx.opt: Add sm_53 to -misa.
1130 * config/nvptx/nvptx-modes.def: Add support for HFmode.
1131 * config/nvptx/nvptx.h (TARGET_SM53):
1132 New helper macro to conditionalize functionality on target ISA.
1133 * config/nvptx/nvptx-c.c (nvptx_cpu_cpp_builtins): Add __PTX_SM__
1134 support for the new ISA levels.
1135 * config/nvptx/nvptx.c (nvtx_ptx_type_from_mode): Support new HFmode
1136 with the ".f16" suffix/qualifier.
1137 (nvptx_file_start): Add support for TARGET_SM53.
1138 (nvptx_omp_device_kind_arch_isa): Add support for TARGET_SM53
1139 and tweak TARGET_SM35.
1140 (nvptx_scalar_mode_supported_p): Target hook with conditional
1141 HFmode support on TARGET_SM53 and higher.
1142 (nvptx_libgcc_floating_mode_supported_p): Likewise.
1143 (TARGET_SCALAR_MODE_SUPPORTED_P): Use nvptx_scalar_mode_supported_p.
1144 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Likewise, use new hook.
1145 * config/nvptx/nvptx.md (*movhf_insn): New define_insn.
1146 (movhf): New define_expand for HFmode moves.
1147 (addhf3, subhf3, mulhf, extendhf<mode>2, trunc<mode>hf2): New
1148 instructions conditional on TARGET_SM53 (i.e. -misa=sm_53).
1150 2021-12-12 Jan Hubicka <hubicka@ucw.cz>
1153 * ipa-modref.c (modref_access_analysis::analyze): Terminate BB
1154 analysis on NULL memory access.
1155 * ipa-pure-const.c (analyze_function): Likewise.
1157 2021-12-11 Jan Hubicka <hubicka@ucw.cz>
1159 * ipa-profile.c (ipa_profile): Do not update hot bb threshold.
1161 2021-12-11 Jan Hubicka <hubicka@ucw.cz>
1163 * ipa-modref.c (get_modref_function_summary): Use ultimate_alias_target.
1164 (ignore_edge): Likewise.
1165 (compute_parm_map): Likewise.
1166 (modref_propagate_in_scc): Likewise.
1167 (modref_propagate_flags_in_scc): Likewise.
1169 2021-12-10 Jason Merrill <jason@redhat.com>
1171 * symtab.c (symtab_node::equal_address_to): Fix comment typo.
1173 2021-12-10 Doug Rupp <rupp@adacore.com>
1175 * config/vxworks.h (LINK_SPEC): Remove %(link_target).
1176 Change %{v:-v} to %{v:-V}.
1178 2021-12-10 Olivier Hainque <hainque@adacore.com>
1180 * config/t-vxworks: Remove assignment to STMP_FIXINC.
1182 2021-12-10 Martin Liska <mliska@suse.cz>
1184 * params.opt: Add missing dot.
1186 2021-12-10 Roger Sayle <roger@nextmovesoftware.com>
1189 * ipa-modref-tree.h (useful_for_kill_p): Zero width accesses aren't
1190 useful for kill tracking.
1192 2021-12-10 Andrew Stubbs <ams@codesourcery.com>
1194 * config/gcn/mkoffload.c (process_asm): Process the variable table
1195 completely differently.
1196 (process_obj): Encode the varaible data differently.
1198 2021-12-10 Joel Hutton <joel.hutton@arm.com>
1200 PR tree-optimization/103523
1201 * tree-vect-loop.c (vectorizable_induction): Check for
1202 PLUS_EXPR/MINUS_EXPR support.
1204 2021-12-10 Cui,Lili <lili.cui@intel.com>
1206 * config/i386/i386.c (ix86_vector_costs::add_stmt_cost): Remove Tremont.
1208 2021-12-09 Jan Hubicka <hubicka@ucw.cz>
1210 * doc/invoke.texi (max-inline-functions-called-once-loop-depth,
1211 max-inline-functions-called-once-insns): New parameters.
1212 * ipa-inline.c (check_callers): Handle
1213 param_inline_functions_called_once_loop_depth and
1214 param_inline_functions_called_once_insns.
1215 (edge_badness): Fix linebreaks.
1216 * params.opt (param=max-inline-functions-called-once-loop-depth,
1217 param=max-inline-functions-called-once-insn): New params.
1219 2021-12-09 Martin Sebor <msebor@redhat.com>
1221 PR tree-optimization/103215
1222 * pointer-query.cc (access_ref::merge_ref): Extend the offset and
1223 size of the merged object instead of using the larger.
1225 2021-12-09 Martin Sebor <msebor@redhat.com>
1227 PR middle-end/101751
1228 * doc/extend.texi (attribute access): Adjust.
1229 * gimple-ssa-warn-access.cc (pass_waccess::maybe_check_access_sizes):
1230 Treat access mode none on a void* argument as expecting as few as
1233 2021-12-09 Jakub Jelinek <jakub@redhat.com>
1236 * config/aarch64/aarch64-sve-builtins.cc (gt_pch_nx): Change type of
1237 second argument from function with 2 pointer arguments to function
1238 with 3 pointer arguments.
1240 2021-12-09 Olivier Hainque <hainque@adacore.com>
1242 * config/aarch64/aarch64-vxworks.h (TARGET_OS_CPP_BUILTINS):
1243 Use VX_CPU_PREFIX in CPU definitions.
1245 2021-12-09 Martin Sebor <msebor@redhat.com>
1247 * pointer-query.cc (access_ref::dump): Define new function
1248 (pointer_query::dump): Call it.
1249 * pointer-query.h (access_ref::dump): Declare new function.
1251 2021-12-09 Martin Sebor <msebor@redhat.com>
1253 * pointer-query.cc (compute_objsize_r): Add an argument.
1254 (gimple_call_return_array): Pass a new argument to compute_objsize_r.
1255 (access_ref::merge_ref): Same.
1256 (access_ref::inform_access): Add an argument and use it.
1257 (access_data::access_data): Initialize new member.
1258 (handle_min_max_size): Pass a new argument to compute_objsize_r.
1259 (handle_decl): New function.
1260 (handle_array_ref): Pass a new argument to compute_objsize_r.
1261 Avoid incrementing deref.
1262 (set_component_ref_size): New function.
1263 (handle_component_ref): New function.
1264 (handle_mem_ref): Pass a new argument to compute_objsize_r.
1265 Only increment deref after successfully computing object size.
1266 (handle_ssa_name): New function.
1267 (compute_objsize_r): Move code into helpers and call them.
1268 (compute_objsize): Pass a new argument to compute_objsize_r.
1269 * pointer-query.h (access_ref::inform_access): Add an argument.
1270 (access_data::ostype): New member.
1272 2021-12-09 Martin Sebor <msebor@redhat.com>
1274 * pointer-query.cc (access_ref::merge_ref): Define new function.
1275 (access_ref::get_ref): Move code into merge_ref and call it.
1276 * pointer-query.h (access_ref::merge_ref): Declare new function.
1278 2021-12-09 Martin Sebor <msebor@redhat.com>
1280 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Pass
1281 GIMPLE statement to compute_objsize.
1282 * pointer-query.cc (compute_objsize): Add a statement argument.
1283 * pointer-query.h (compute_objsize): Define a new overload.
1285 2021-12-09 Martin Sebor <msebor@redhat.com>
1287 * gimple-ssa-warn-access.cc (check_access): Adjust to member name
1289 (pass_waccess::check_strncmp): Same.
1290 * pointer-query.cc (access_ref::access_ref): Remove arguments.
1292 (access_data::access_data): Define new ctors.
1293 (access_data::set_bound): Define new member function.
1294 (compute_objsize_r): Remove unnecessary code.
1295 * pointer-query.h (struct access_ref): Remove ctor arguments.
1296 (struct access_data): Declare ctor overloads.
1297 (access_data::dst_bndrng): New member.
1298 (access_data::src_bndrng): New member.
1300 2021-12-09 Martin Sebor <msebor@redhat.com>
1302 PR middle-end/103143
1303 * pointer-query.cc (gimple_call_return_array): Call compute_objsize_r.
1305 2021-12-09 Olivier Hainque <hainque@adacore.com>
1306 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1308 * Makefile.in (T_STDINT_GCC_H): New variable, path to
1309 stdint-gcc.h that a target configuration may override when
1310 use_gcc_stdint is "provide".
1311 (stmp-int-hdrs): Depend on it and copy that for
1312 USE_GCC_INT=provide.
1313 * config.gcc (vxworks): Revert to use_gcc_stdint=provide.
1314 * config/t-vxworks (T_STDINT_GCC_H): Define, as vxw-stdint-gcc.h.
1315 (vxw-stdint-gcc.h): New target, produced from the original
1317 (vxw-glimits.h): Use an automatic variable to designate the
1318 first and only prerequisite.
1319 * config/vxworks/stdint.h: Remove.
1321 2021-12-09 Iain Sandoe <iain@sandoe.co.uk>
1324 * config/host-darwin.c (SAFE_ALLOC_SIZE): Remove.
1325 (darwin_gt_pch_get_address): Rework for relocatable PCH.
1326 (darwin_gt_pch_use_address): Likewise.
1328 2021-12-09 Jakub Jelinek <jakub@redhat.com>
1331 * config/host-darwin.c (darwin_gt_pch_use_address): When reading
1332 manually the file into mapped area, update mapped_addr as
1333 an automatic variable rather than addr which is a reference parameter.
1334 * config/host-hpux.c (hpux_gt_pch_use_address): When reading
1335 manually the file into mapped area, update addr as
1336 an automatic variable rather than base which is a reference parameter.
1338 2021-12-09 Jakub Jelinek <jakub@redhat.com>
1341 * coretypes.h (gt_pointer_operator): Use 3 pointer arguments instead
1343 * gengtype.c (struct walk_type_data): Add in_nested_ptr argument.
1344 (walk_type): Temporarily set d->in_nested_ptr around nested_ptr
1346 (write_types_local_user_process_field): Pass a new middle pointer
1347 to gt_pointer_operator op calls, if d->in_nested_ptr pass there
1348 address of d->prev_val[2], otherwise NULL.
1349 (write_types_local_process_field): Likewise.
1350 * ggc-common.c (relocate_ptrs): Add real_ptr_p argument. If equal
1351 to ptr_p, do nothing, otherwise if NULL remember ptr_p's
1352 or if non-NULL real_ptr_p's corresponding new address in
1354 (reloc_addrs_vec): New variable.
1355 (compare_ptr, read_uleb128, write_uleb128): New functions.
1356 (gt_pch_save): When iterating over objects through relocate_ptrs,
1357 save current i into state.ptrs_i. Sort reloc_addrs_vec and emit
1358 it as uleb128 of differences between pointer addresses into the
1360 (gt_pch_restore): Allow restoring of PCH to a different address
1361 than the preferred one, in that case adjust global pointers by bias
1362 and also adjust by bias addresses read from the relocation table
1363 as uleb128 differences. Otherwise fseek over it. Perform
1364 gt_pch_restore_stringpool only after adjusting callbacks and for
1365 callback adjustments also take into account the bias.
1366 (default_gt_pch_use_address): Change type of first argument from
1368 (mmap_gt_pch_use_address): Likewise.
1369 * ggc-tests.c (gt_pch_nx): Pass NULL as new middle argument to op.
1370 * hash-map.h (hash_map::pch_nx_helper): Likewise.
1371 (gt_pch_nx): Likewise.
1372 * hash-set.h (gt_pch_nx): Likewise.
1373 * hash-table.h (gt_pch_nx): Likewise.
1374 * hash-traits.h (ggc_remove::pch_nx): Likewise.
1375 * hosthooks-def.h (default_gt_pch_use_address): Change type of first
1376 argument from void * to void *&.
1377 (mmap_gt_pch_use_address): Likewise.
1378 * hosthooks.h (struct host_hooks): Change type of first argument of
1379 gt_pch_use_address hook from void * to void *&.
1380 * machmode.h (gt_pch_nx): Expect a callback with 3 pointers instead of
1381 two in the middle argument.
1382 * poly-int.h (gt_pch_nx): Likewise.
1383 * stringpool.c (gt_pch_nx): Pass NULL as new middle argument to op.
1384 * tree-cfg.c (gt_pch_nx): Likewise, except for LOCATION_BLOCK pass
1385 the same &(block) twice.
1386 * value-range.h (gt_pch_nx): Pass NULL as new middle argument to op.
1387 * vec.h (gt_pch_nx): Likewise.
1388 * wide-int.h (gt_pch_nx): Likewise.
1389 * config/host-darwin.c (darwin_gt_pch_use_address): Change type of
1390 first argument from void * to void *&.
1391 * config/host-darwin.h (darwin_gt_pch_use_address): Likewise.
1392 * config/host-hpux.c (hpux_gt_pch_use_address): Likewise.
1393 * config/host-linux.c (linux_gt_pch_use_address): Likewise. If
1394 it couldn't succeed to mmap at the preferred location, set base
1395 to the actual one. Update addr in the manual reading loop instead of
1397 * config/host-netbsd.c (netbsd_gt_pch_use_address): Change type of
1398 first argument from void * to void *&.
1399 * config/host-openbsd.c (openbsd_gt_pch_use_address): Likewise.
1400 * config/host-solaris.c (sol_gt_pch_use_address): Likewise.
1401 * config/i386/host-mingw32.c (mingw32_gt_pch_use_address): Likewise.
1402 * config/rs6000/rs6000-gen-builtins.c (write_init_file): Pass NULL
1403 as new middle argument to op in the generated code.
1404 * doc/gty.texi: Adjust samples for the addition of middle pointer
1405 to gt_pointer_operator callback.
1407 2021-12-09 Alexandre Oliva <oliva@adacore.com>
1410 * reg-stack.c (convert_regs_1): Move any_malformed_asm
1412 (reg_to_stack): ... here.
1414 2021-12-09 Alexandre Oliva <oliva@adacore.com>
1417 * expr.c (emit_move_multi_word): Skip clobber during lra.
1419 2021-12-09 Alexandre Oliva <oliva@adacore.com>
1421 PR tree-optimization/103024
1422 PR middle-end/103530
1423 * gimple-harden-conditionals.cc (non_eh_succ_edge): New.
1424 (pass_harden_compares::execute): Accept 1-bit integral types,
1425 and cope with throwing compares.
1427 2021-12-08 Iain Sandoe <iain@sandoe.co.uk>
1429 * config/darwin.h (DARWIN_PIE_SPEC): Add -no_pie when
1430 linking mdynamic-no-pic code on macOS > 10.7.
1432 2021-12-08 Dimitar Dimitrov <dimitar@dinux.eu>
1434 * config/pru/pru.c (pru_section_type_flags): New function.
1435 (TARGET_SECTION_TYPE_FLAGS): Wire it.
1437 2021-12-08 David Faust <david.faust@oracle.com>
1439 * config/bpf/bpf.c (handle_attr_preserve): Avoid calling
1440 is_gimple_assign with a NULL pointer.
1442 2021-12-08 Martin Liska <mliska@suse.cz>
1444 * profile-count.c (profile_count::dump): Add function
1445 that can dump to a provided buffer.
1446 (profile_probability::dump): Likewise.
1447 * profile-count.h: Likewise.
1448 * tree-ssa-loop-unswitch.c (tree_unswitch_single_loop):
1449 Use dump_printf_loc infrastructure.
1450 (tree_unswitch_outer_loop): Likewise.
1451 (find_loop_guard): Likewise.
1452 (hoist_guard): Likewise.
1454 2021-12-08 Chung-Lin Tang <cltang@codesourcery.com>
1456 * gimplify.c (extract_base_bit_offset): Add 'tree *offsetp' parameter,
1457 accomodate case where 'offset' return of get_inner_reference is
1459 (is_or_contains_p): Further robustify conditions.
1460 (omp_target_reorder_clauses): In alloc/to/from sorting phase, also
1461 move following GOMP_MAP_ALWAYS_POINTER maps along. Add new sorting
1462 phase where we make sure pointers with an attach/detach map are ordered
1464 (gimplify_scan_omp_clauses): Add modifications to avoid creating
1465 GOMP_MAP_STRUCT and associated alloc map for attach/detach maps.
1467 2021-12-08 Roger Sayle <roger@nextmovesoftware.com>
1469 * config/nvptx/nvptx.md (*extend_trunc_<mode>2_qi,
1470 *extend_trunc_<mode>2_hi, *extend_trunc_di2_si): New insns.
1471 Use cvt to perform sign-extension of truncation in one step.
1473 2021-12-08 Chung-Lin Tang <cltang@codesourcery.com>
1476 * gimplify.c ("tree-hash-traits.h"): Add include.
1477 (gimplify_scan_omp_clauses): Change struct_map_to_clause to type
1478 hash_map<tree_operand, tree> *. Adjust struct map handling to handle
1479 cases of *A and A->B expressions. Under !DECL_P case of
1480 GOMP_CLAUSE_MAP handling, add STRIP_NOPS for indir_p case, add to
1481 struct_deref_set for map(*ptr_to_struct) cases. Add MEM_REF case when
1482 handling component_ref_p case. Add unshare_expr and gimplification
1483 when created GOMP_MAP_STRUCT is not a DECL. Add code to add
1484 firstprivate pointer for *pointer-to-struct case.
1485 (gimplify_adjust_omp_clauses): Move GOMP_MAP_STRUCT removal code for
1486 exit data directives code to earlier position.
1487 * omp-low.c (lower_omp_target):
1488 Handle GOMP_MAP_ATTACH_ZERO_LENGTH_ARRAY_SECTION, and
1489 GOMP_MAP_POINTER_TO_ZERO_LENGTH_ARRAY_SECTION map kinds.
1490 * tree-pretty-print.c (dump_omp_clause): Likewise.
1492 2021-12-08 Andrew Stubbs <ams@codesourcery.com>
1493 Hafiz Abid Qadeer <abidh@codesourcery.com>
1495 * dwarf2cfi.c (dw_stack_pointer_regnum): Change type to struct cfa_reg.
1496 (dw_frame_pointer_regnum): Likewise.
1497 (new_cfi_row): Use set_by_dwreg.
1498 (get_cfa_from_loc_descr): Use set_by_dwreg. Support register spans.
1499 handle DW_OP_bregx with DW_OP_breg{0-31}. Support DW_OP_lit*,
1500 DW_OP_const*, DW_OP_minus, DW_OP_shl and DW_OP_plus.
1501 (lookup_cfa_1): Use set_by_dwreg.
1502 (def_cfa_0): Update for cfa_reg and support register spans.
1503 (reg_save): Change sreg parameter to struct cfa_reg. Support register
1505 (dwf_cfa_reg): New function.
1506 (dwarf2out_flush_queued_reg_saves): Use dwf_cfa_reg instead of
1508 (dwarf2out_frame_debug_def_cfa): Likewise.
1509 (dwarf2out_frame_debug_adjust_cfa): Likewise.
1510 (dwarf2out_frame_debug_cfa_offset): Likewise. Update reg_save usage.
1511 (dwarf2out_frame_debug_cfa_register): Likewise.
1512 (dwarf2out_frame_debug_expr): Likewise.
1513 (create_pseudo_cfg): Use set_by_dwreg.
1514 (initial_return_save): Use set_by_dwreg and dwf_cfa_reg,
1515 (create_cie_data): Use dwf_cfa_reg.
1516 (execute_dwarf2_frame): Use dwf_cfa_reg.
1517 (dump_cfi_row): Use set_by_dwreg.
1518 * dwarf2out.c (build_span_loc, build_breg_loc): New function.
1519 (build_cfa_loc): Support register spans.
1520 (build_cfa_aligned_loc): Update cfa_reg usage.
1521 (convert_cfa_to_fb_loc_list): Use set_by_dwreg.
1522 * dwarf2out.h (struct cfa_reg): New type.
1523 (struct dw_cfa_location): Use struct cfa_reg.
1524 (build_span_loc): New prototype.
1526 2021-12-08 Haochen Jiang <haochen.jiang@intel.com>
1529 * config/i386/sse.md
1530 (*<sse4_1>_blendv<ssefltmodesuffix><avxsizesuffix>_not_ltint):
1531 Add new define_insn_and_split.
1533 2021-12-08 Alexandre Oliva <oliva@adacore.com>
1535 PR middle-end/103149
1536 * gimple-harden-conditionals.cc (detach_value): Use memory if
1537 general regs won't do.
1539 2021-12-07 Martin Liska <mliska@suse.cz>
1541 PR middle-end/103438
1542 * config/s390/s390.c (s390_valid_target_attribute_inner_p):
1543 Use new enum CLVC_INTEGER.
1544 * opt-functions.awk: Use new CLVC_INTEGER.
1545 * opts-common.c (set_option): Likewise.
1546 (option_enabled): Return -1,0,1 for CLVC_INTEGER.
1547 (get_option_state): Use new CLVC_INTEGER.
1548 (control_warning_option): Likewise.
1549 * opts.h (enum cl_var_type): Likewise.
1551 2021-12-07 H.J. Lu <hjl.tools@gmail.com>
1554 * config/i386/i386.c (ix86_call_use_plt_p): Check FUNCTION_DECL
1555 before calling cgraph_node::get.
1557 2021-12-07 Richard Biener <rguenther@suse.de>
1559 PR tree-optimization/103596
1560 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
1561 Note we are not propagating into a PHI argument to may_propagate_copy.
1562 * tree-ssa-propagate.h (may_propagate_copy): Add
1563 argument specifying whether we propagate into a PHI arg.
1564 * tree-ssa-propagate.c (may_propagate_copy): Likewise.
1565 When not doing so we can replace an abnormal with
1567 (may_propagate_into_stmt): Update may_propagate_copy calls.
1568 (replace_exp_1): Move propagation checking code to
1569 propagate_value and rename to ...
1570 (replace_exp): ... this and elide previous wrapper.
1571 (propagate_value): Perform checking with adjusted
1572 may_propagate_copy call and dispatch to replace_exp.
1574 2021-12-07 Matthias Kretz <m.kretz@gsi.de>
1576 * hash-map.h (hash_map::traverse): Let both overloads behave the
1578 * predict.c (assert_is_empty): Return true, thus not changing
1581 2021-12-07 YunQiang Su <yunqiang.su@cipunited.com>
1583 * config/mips/mips.h (ISA_HAS_UNALIGNED_ACCESS, STRICT_ALIGNMENT):
1584 R6 can unaligned access.
1585 * config/mips/mips.md (movmisalign<mode>): Likewise.
1586 * config/mips/mips.opt: add -m(no-)unaligned-access
1587 * doc/invoke.texi: Likewise.
1589 2021-12-07 Eugene Rozenfeld <erozen@microsoft.com>
1591 * auto-profile.c (afdo_propagate_edge): Improve count propagation algorithm.
1593 2021-12-06 Paul A. Clarke <pc@us.ibm.com>
1596 * config/rs6000/xmmintrin.h (_mm_movemask_ps): Replace "vector" with
1599 2021-12-06 Jose E. Marchesi <jose.marchesi@oracle.com>
1601 * config/bpf/bpf.c (bpf_handle_preserve_access_index_attribute):
1602 Mark arguments `args' and flags' as unused.
1603 (bpf_core_newdecl): Remove unused local `newdecl'.
1604 (bpf_core_newdecl): Remove unused argument `loc'.
1605 (ctfc_debuginfo_early_finish_p): Remove unused function.
1606 (TARGET_CTFC_DEBUGINFO_EARLY_FINISH_P): Remove definition.
1607 (bpf_core_walk): Do not pass a location to bpf_core_newdecl.
1609 2021-12-06 Richard Sandiford <richard.sandiford@arm.com>
1611 * gimple-range-edge.cc (gimple_outgoing_range::edge_range_p): Add
1612 a shortcut for blocks with single successors.
1613 * gimple-range-gori.cc (gori_map::calculate_gori): Likewise.
1615 2021-12-06 Richard Sandiford <richard.sandiford@arm.com>
1617 * value-range.cc (irange::irange_union): Use quick_push rather
1618 than safe_push. Use widest_int rather than wide_int. Avoid
1619 assigning wi::to_* results to wide*_int temporaries.
1621 2021-12-06 Andrew MacLeod <amacleod@redhat.com>
1623 * gimple-range-cache.cc (ranger_cache::fill_block_cache): Check for
1624 a range from dominators before filling the cache.
1625 (ranger_cache::range_from_dom): New.
1626 * gimple-range-cache.h (ranger_cache::range_from_dom): Add prototype.
1628 2021-12-06 Andrew MacLeod <amacleod@redhat.com>
1630 * gimple-range-gori.h (class gori_compute):: Add prototypes.
1631 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Add alternate
1632 API for basic block. Call for edge alterantive.
1633 (gori_compute::may_recompute_p): Ditto.
1635 2021-12-06 Richard Biener <rguenther@suse.de>
1637 PR tree-optimization/103581
1638 * tree-vect-stmts.c (vect_build_gather_load_calls): Properly
1639 guard all the AVX512 mask cases.
1641 2021-12-06 Richard Biener <rguenther@suse.de>
1643 PR tree-optimization/103544
1644 * tree-vect-slp.c (vect_analyze_slp): Only add a SLP reduction
1645 opportunity if the stmt in question is the reduction root.
1646 (dot_slp_tree): Add missing check for NULL child.
1648 2021-12-06 Jakub Jelinek <jakub@redhat.com>
1651 * config/avr/avr.c (avr_output_data_section_asm_op,
1652 avr_output_bss_section_asm_op): Change argument type from const void *
1655 2021-12-06 Tamar Christina <tamar.christina@arm.com>
1657 PR rtl-optimization/103404
1658 * cse.c (find_sets_in_insn): Don't select elements out of a V1 mode
1661 2021-12-06 Hongtao Liu <Hongtao.liu@intel.com>
1662 Uroš Bizjak <ubizjak@gmail.com>
1665 * config/i386/i386.c (ix86_preferred_reload_class): Allow
1666 integer regs when moves between register units are cheap.
1667 * config/i386/i386.h (INT_SSE_CLASS_P): New.
1669 2021-12-05 David Edelsohn <dje.gcc@gmail.com>
1671 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
1672 AIX math builtin initialization before new_builtins_are_live.
1674 2021-12-04 Jakub Jelinek <jakub@redhat.com>
1676 * config/i386/x86-tune.def (X86_TUNE_PARTIAL_REG_DEPENDENCY): Fix
1677 comment typo, Preffer -> prefer.
1678 * ipa-modref-tree.c (modref_access_node::closer_pair_p): Likewise.
1680 2021-12-04 Alexandre Oliva <oliva@adacore.com>
1682 PR rtl-optimization/103028
1683 * ifcvt.c (find_cond_trap): Validate new insns more strictly
1686 2021-12-03 Martin Liska <mliska@suse.cz>
1687 Peter Bergner <bergner@linux.ibm.com>
1690 * config/rs6000/rs6000.c (rs6000_option_override_internal): Move the
1691 disabling of shrink-wrapping when using -mrop-protect from here...
1692 (rs6000_override_options_after_change): ...to here.
1694 2021-12-03 H.J. Lu <hjl.tools@gmail.com>
1697 * config/i386/i386-expand.c (ix86_expand_builtin): Pass PVW_NONE
1698 and PVW_NONE to ix86_target_string.
1699 * config/i386/i386-options.c (ix86_target_string): Add arguments
1700 for move_max and store_max.
1701 (ix86_target_string::add_vector_width): New lambda.
1702 (ix86_debug_options): Pass ix86_move_max and ix86_store_max to
1704 (ix86_function_specific_print): Pass ptr->x_ix86_move_max and
1705 ptr->x_ix86_store_max to ix86_target_string.
1706 (ix86_valid_target_attribute_tree): Handle x_ix86_move_max and
1708 (ix86_option_override_internal): Set the default x_ix86_move_max
1709 and x_ix86_store_max.
1710 * config/i386/i386-options.h (ix86_target_string): Add
1711 prefer_vector_width and prefer_vector_width.
1712 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): Removed.
1713 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
1714 (MOVE_MAX): Use 64 if ix86_move_max or ix86_store_max ==
1715 PVW_AVX512. Use 32 if ix86_move_max or ix86_store_max >=
1717 (STORE_MAX_PIECES): Use 64 if ix86_store_max == PVW_AVX512.
1718 Use 32 if ix86_store_max >= PVW_AVX256.
1719 * config/i386/i386.opt: Add -mmove-max=bits and -mstore-max=bits.
1720 * config/i386/x86-tune.def (X86_TUNE_AVX512_MOVE_BY_PIECES): New.
1721 (X86_TUNE_AVX512_STORE_BY_PIECES): Likewise.
1722 * doc/invoke.texi: Document -mmove-max=bits and -mstore-max=bits.
1724 2021-12-03 Bill Schmidt <wschmidt@linux.ibm.com>
1726 * config/rs6000/rs6000.c (rs6000_builtin_reciprocal): Fix builtin
1729 2021-12-03 SiYu Wu <siyu@isrc.iscas.ac.cn>
1731 * common/config/riscv/riscv-common.c (riscv_implied_info):
1732 Add K-ext related entry.
1733 (riscv_supported_std_ext): Add 'k'.
1734 * config/riscv/arch-canonicalize (CANONICAL_ORDER): Add 'k'.
1735 (IMPLIED_EXT): Add K-ext related entry.
1737 2021-12-03 SiYu Wu <siyu@isrc.iscas.ac.cn>
1739 * common/config/riscv/riscv-common.c (riscv_ext_version_table):
1741 * config/riscv/riscv-opts.h (MASK_ZBKB): New.
1748 (MASK_ZKSED): Ditto.
1751 (TARGET_ZBKB): Ditto.
1752 (TARGET_ZBKC): Ditto.
1753 (TARGET_ZBKX): Ditto.
1754 (TARGET_ZKNE): Ditto.
1755 (TARGET_ZKND): Ditto.
1756 (TARGET_ZKNH): Ditto.
1757 (TARGET_ZKR): Ditto.
1758 (TARGET_ZKSED): Ditto.
1759 (TARGET_ZKSH): Ditto.
1760 (TARGET_ZKT): Ditto.
1761 * config/riscv/riscv.opt (riscv_zk_subext): New.
1763 2021-12-03 Tamar Christina <tamar.christina@arm.com>
1765 * tree-vect-stmts.c (prepare_load_store_mask): Rename to...
1766 (prepare_vec_mask): ...This and record operations that have already been
1768 (vectorizable_call): Use it.
1769 (vectorizable_operation): Likewise.
1770 (vectorizable_store): Likewise.
1771 (vectorizable_load): Likewise.
1772 * tree-vectorizer.h (class _loop_vec_info): Add vec_cond_masked_set.
1773 (vec_cond_masked_set_type, tree_cond_mask_hash): New.
1775 2021-12-03 H.J. Lu <hjl.tools@gmail.com>
1779 * target.def (ifunc_ref_local_ok): Add a target hook.
1780 * varasm.c (default_binds_local_p_3): Force indirect function
1781 resolver non-local only if targetm.ifunc_ref_local_ok returns
1783 * config/i386/i386-expand.c (ix86_expand_call): Call
1784 ix86_call_use_plt_p to check if PLT should be used.
1785 * config/i386/i386-protos.h (ix86_call_use_plt_p): New.
1786 * config/i386/i386.c (output_pic_addr_const): Call
1787 ix86_call_use_plt_p to check if "@PLT" is needed.
1788 (ix86_call_use_plt_p): New.
1789 (TARGET_IFUNC_REF_LOCAL_OK): New.
1790 * doc/tm.texi.in: Add TARGET_IFUNC_REF_LOCAL_OK.
1791 * doc/tm.texi: Regenerated.
1793 2021-12-03 Jakub Jelinek <jakub@redhat.com>
1795 * attribs.h (simple_cst_list_equal): Declare.
1796 * attribs.c (simple_cst_list_equal): No longer static.
1797 * config/i386/i386-options.c (target_attribute_cache): New variable.
1798 (ix86_valid_target_attribute_p): Cache DECL_FUNCTION_SPECIFIC_TARGET
1799 and DECL_FUNCTION_SPECIFIC_OPTIMIZATION based on args.
1801 2021-12-03 Jakub Jelinek <jakub@redhat.com>
1804 * ggc.h (gt_pch_note_callback): Declare.
1805 * gengtype.h (enum typekind): Add TYPE_CALLBACK.
1806 (callback_type): Declare.
1807 * gengtype.c (dbgprint_count_type_at): Handle TYPE_CALLBACK.
1808 (callback_type): New variable.
1809 (process_gc_options): Add CALLBACK argument, handle callback
1811 (set_gc_used_type): Adjust process_gc_options caller, if callback,
1812 set type to &callback_type.
1813 (output_mangled_typename): Handle TYPE_CALLBACK.
1814 (walk_type): Likewise. Handle callback option.
1815 (write_types_process_field): Handle TYPE_CALLBACK.
1816 (write_types_local_user_process_field): Likewise.
1817 (write_types_local_process_field): Likewise.
1818 (write_root): Likewise.
1819 (dump_typekind): Likewise.
1820 (dump_type): Likewise.
1821 * gengtype-state.c (type_lineloc): Handle TYPE_CALLBACK.
1822 (state_writer::write_state_callback_type): New method.
1823 (state_writer::write_state_type): Handle TYPE_CALLBACK.
1824 (read_state_callback_type): New function.
1825 (read_state_type): Handle TYPE_CALLBACK.
1826 * ggc-common.c (callback_vec): New variable.
1827 (gt_pch_note_callback): New function.
1828 (gt_pch_save): Stream out gt_pch_save function address and relocation
1830 (gt_pch_restore): Stream in saved gt_pch_save function address and
1831 relocation table and apply relocations if needed.
1832 * doc/gty.texi (callback): Document new GTY option.
1833 * varasm.c (get_unnamed_section): Change callback argument's type and
1834 last argument's type from const void * to const char *.
1835 (output_section_asm_op): Change argument's type from const void *
1836 to const char *, remove unnecessary cast.
1837 * tree-core.h (struct tree_translation_unit_decl): Drop GTY((skip))
1838 from language member.
1839 * output.h (unnamed_section_callback): Change argument type from
1840 const void * to const char *.
1841 (struct unnamed_section): Use GTY((callback)) instead of GTY((skip))
1842 for callback member. Change data member type from const void *
1844 (struct noswitch_section): Use GTY((callback)) instead of GTY((skip))
1845 for callback member.
1846 (get_unnamed_section): Change callback argument's type and
1847 last argument's type from const void * to const char *.
1848 (output_section_asm_op): Change argument's type from const void *
1850 * config/avr/avr.c (avr_output_progmem_section_asm_op): Likewise.
1851 Remove unneeded cast.
1852 * config/darwin.c (output_objc_section_asm_op): Change argument's type
1853 from const void * to const char *.
1854 * config/pa/pa.c (som_output_text_section_asm_op): Likewise.
1855 (som_output_comdat_data_section_asm_op): Likewise.
1856 * config/rs6000/rs6000.c (rs6000_elf_output_toc_section_asm_op):
1858 (rs6000_xcoff_output_readonly_section_asm_op): Likewise. Instead
1859 of dereferencing directive hardcode variable names and decide based on
1860 whether directive is NULL or not.
1861 (rs6000_xcoff_output_readwrite_section_asm_op): Change argument's type
1862 from const void * to const char *.
1863 (rs6000_xcoff_output_tls_section_asm_op): Likewise. Instead
1864 of dereferencing directive hardcode variable names and decide based on
1865 whether directive is NULL or not.
1866 (rs6000_xcoff_output_toc_section_asm_op): Change argument's type
1867 from const void * to const char *.
1868 (rs6000_xcoff_asm_init_sections): Adjust get_unnamed_section callers.
1870 2021-12-03 Richard Purdie <richard.purdie@linuxfoundation.org>
1872 * Makefile.in: Fix "argument list too long" from install-plugins.
1874 2021-12-03 Matt Jacobson <mhjacobson@me.com>
1876 * config.gcc: For the AVR target, populate TM_MULTILIB_CONFIG.
1877 * config/avr/genmultilib.awk: Add ability to filter generated multilib
1879 * config/avr/t-avr: Pass TM_MULTILIB_CONFIG to genmultilib.awk.
1880 * configure.ac: Update help string for --with-multilib-list.
1881 * configure: Regenerate.
1883 2021-12-02 Ilya Lipnitskiy <ilya.lipnitskiy@gmail.com>
1885 * configure: Regenerate.
1886 * configure.ac: Define TARGET_LIBC_GNUSTACK on musl.
1888 2021-12-02 Iain Sandoe <iain@sandoe.co.uk>
1891 * config/host-darwin.c (TRY_EMPTY_VM_SPACE,
1892 SAFE_ALLOC_SIZE): New.
1893 (darwin_gt_pch_get_address): Rewrite to use nominated
1894 memory segments rather than part of the compiler __DATA
1896 (darwin_gt_pch_use_address): Likewise.
1898 2021-12-02 Gerald Pfeifer <gerald@pfeifer.com>
1900 * doc/install.texi (*-*-freebsd*): Remove references to
1901 FreeBSD 1 and FreeBSD 2.
1903 2021-12-02 Vladimir N. Makarov <vmakarov@redhat.com>
1905 PR rtl-optimization/103437
1906 * ira-color.c (setup_allocno_priorities): Switch off backup code
1907 for overflow if compiler has __builtin_smul_overflow. Use <
1908 for comparison with -INT_MAX.
1910 2021-12-02 Bill Schmidt <wschmidt@linux.ibm.com>
1912 * config/rs6000/altivec.h: Delete a number of #defines that are now
1913 superfluous. Alphabetize. Include rs6000-vecdefines.h. Include some
1915 * config/rs6000/rs6000-builtin-new.def (CMPB): Flag as no32bit.
1916 (BPERMD): Flag as 32bit (needing special handling for 32-bit).
1917 (UNPACK_TD): Return unsigned long long instead of unsigned long.
1918 (GET_TEXASR): Return unsigned long instead of unsigned long long.
1919 (GET_TEXASRU): Likewise.
1920 (GET_TFHAR): Likewise.
1921 (GET_TFIAR): Likewise.
1922 (SET_TEXASR): Pass unsigned long instead of unsigned long long.
1923 (SET_TEXASRU): Likewise.
1924 (SET_TFHAR): Likewise.
1925 (SET_TFIAR): Likewise.
1926 (TABORTDC): Likewise.
1927 (TABORTDCI): Likewise.
1928 * config/rs6000/rs6000-call.c (rs6000_expand_new_builtin): Fix error
1929 handling for no32bit. Add 32bit handling for RS6000_BIF_BPERMD.
1930 * config/rs6000/rs6000-gen-builtins.c (write_init_file): Initialize
1931 new_builtins_are_live to 1.
1933 2021-12-02 Vladimir N. Makarov <vmakarov@redhat.com>
1935 PR rtl-optimization/103437
1936 * ira-color.c (setup_allocno_priorities): Use long long
1937 multiplication as backup for overflow processing.
1939 2021-12-02 Tamar Christina <tamar.christina@arm.com>
1941 * config/aarch64/aarch64-protos.h (aarch64_gen_shareable_zero): New.
1942 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>,
1943 aarch64_rshrn2<mode>): Generate rounding half-ing add when appropriate.
1944 * config/aarch64/aarch64.c (aarch64_gen_shareable_zero): New.
1946 2021-12-02 Vladimir N. Makarov <vmakarov@redhat.com>
1948 PR rtl-optimization/103437
1949 * ira-color.c (setup_allocno_priorities): Process multiplication
1952 2021-12-02 Martin Liska <mliska@suse.cz>
1954 PR gcov-profile/96092
1955 * common.opt: New option.
1956 * coverage.c (coverage_begin_function): Emit filename with
1957 remap_profile_filename.
1958 * doc/invoke.texi: Document the new option.
1959 * file-prefix-map.c (add_profile_prefix_map): New.
1960 (remap_profile_filename): Likewise.
1961 * file-prefix-map.h (add_profile_prefix_map): Likewise.
1962 (remap_profile_filename): Likewise.
1963 * lto-opts.c (lto_write_options): Handle
1964 OPT_fprofile_prefix_map_.
1965 * opts-global.c (handle_common_deferred_options): Likewise.
1966 * opts.c (common_handle_option): Likewise.
1967 (gen_command_line_string): Likewise.
1968 * profile.c (output_location): Emit filename with
1969 remap_profile_filename.
1971 2021-12-02 Richard Biener <rguenther@suse.de>
1973 PR middle-end/103271
1974 * internal-fn.c (expand_DEFERRED_INIT): When the base
1975 of the LHS is a decl with matching constant size use
1976 that as the initialization target instead of an
1977 eventual VLA typed one.
1979 2021-12-02 Richard Biener <rguenther@suse.de>
1981 PR tree-optimization/103527
1982 * tree-vect-stmts.c (vect_build_gather_load_calls): Always
1983 use a truth type for building the vector mask.
1985 2021-12-02 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
1987 * config/vxworks.h (LTO_PLUGIN_SPEC): Adapt to corrected
1988 spelling of -linker-output-auto-nolto-rel.
1990 2021-12-01 Uroš Bizjak <ubizjak@gmail.com>
1993 * config/i386/sse.md (VI2F): Remove mode iterator.
1994 (VI2F_256_512): New mode iterator.
1995 (vec_set<V8_128:mode>_0): New insn pattern.
1996 (vec_set<VI2F_256_512:mode>_0>): Rename from vec_set<VI2F:mode>mode.
1997 Use VI2F_256_512 mode iterator instead of VI2F.
1998 (*axv512fp16_movsh): Remove.
1999 (<sse2p4_1>_pinsr<ssemodesuffix>): Add (x,x,x) AVX2 alternative.
2000 Do not disable V8HF mode insn on AVX2 targets.
2001 (pinsrw -> pbroadcast + pblendw peephole2): New peephole.
2002 (pinsrw -> pbroadcast + pblendw splitter): New post-reload splitter.
2003 * config/i386/i386.md (extendhfsf): Call gen_vec_setv8hf_0.
2004 * config/i386/i386-expand.c (ix86_expand_vector_set)
2005 <case E_V8HFmode>: Use vec_merge path for TARGET_AVX2.
2007 2021-12-01 Roger Sayle <roger@nextmovesoftware.com>
2008 Richard Biener <rguenther@suse.de>
2010 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
2011 Check if simplify_using_initial_conditions allows us to
2012 simplify the expression for may_be_zero.
2013 * match.pd (X != C ? -X : -C -> -X): New transform.
2014 (X != C ? ~X : ~C -> ~X): Likewise.
2015 ((X+1) > Y ? -X : 1 -> X >= Y ? -X : 1): Likewise.
2017 2021-12-01 Peter Bergner <bergner@linux.ibm.com>
2019 PR middle-end/103127
2020 * gimplify.c (is_var_need_auto_init): Handle opaque types.
2022 2021-12-01 Iain Buclaw <ibuclaw@gdcproject.org>
2024 * common.opt (static-libphobos): Add option.
2025 * config/darwin.h (LINK_SPEC): Substitute -lgphobos with libgphobos.a
2026 when linking statically.
2027 * gcc.c (driver_handle_option): Set -static-libphobos as always valid.
2029 2021-12-01 Jason Merrill <jason@redhat.com>
2032 * fold-const.c (maybe_nonzero_address): Use get_create or get
2033 depending on folding_initializer.
2034 (fold_binary_initializer_loc): New.
2035 * fold-const.h (fold_binary_initializer_loc): Declare.
2036 * varasm.c (mark_weak): Don't use the decl location.
2037 * doc/invoke.texi: Document -fconstexpr-fp-except.
2039 2021-12-01 Bill Schmidt <wschmidt@linux.ibm.com>
2042 * config/rs6000/rs6000-c.c (rs6000-builtins.h): Stop including.
2043 (rs6000-internal.h): Include.
2044 (altivec_resolve_new_overloaded_builtin): Move call to
2045 rs6000_invalid_new_builtin here from rs6000_new_builtin_decl.
2046 * config/rs6000/rs6000-call.c (rs6000-builtins.h): Stop including.
2047 (rs6000_invalid_new_builtin): Remove static qualifier.
2048 (rs6000_new_builtin_decl): Remove test for supported builtin.
2049 * config/rs6000/rs6000-internal.h (rs6000-builtins.h): Include.
2050 (rs6000_invalid_new_builtin): Declare.
2051 * config/rs6000/rs6000.c (rs6000-builtins.h): Don't include.
2053 2021-12-01 Aldy Hernandez <aldyh@redhat.com>
2055 PR tree-optimization/103409
2056 * gimple-range-path.cc (path_range_query::compute_ranges_in_phis):
2057 Do all the work with just one ssa_global_cache.
2058 * gimple-range-path.h: Remove m_tmp_phi_cache.
2060 2021-12-01 Richard Sandiford <richard.sandiford@arm.com>
2062 PR tree-optimization/103517
2063 * tree-vect-slp.c (vect_build_slp_tree_1): When allowing two
2064 different component references, check the codes of both them,
2065 rather than just the first.
2067 2021-12-01 Andrew MacLeod <amacleod@redhat.com>
2069 PR tree-optimization/103464
2070 * gimple-range.cc (gimple_ranger::prefill_name): Process phis also.
2071 (gimple_ranger::prefill_stmt_dependencies): Ditto.
2073 2021-12-01 Iain Buclaw <ibuclaw@gdcproject.org>
2075 * doc/install.texi (Prerequisites): Add note that D front end now
2076 requires GDC installed in order to bootstrap.
2077 (Building): Add D compiler section, referencing prerequisites.
2079 2021-12-01 Jakub Jelinek <jakub@redhat.com>
2081 * config/i386/x86-tune.def (X86_TUNE_SCHEDULE,
2082 X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY,
2083 X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Formatting fixes.
2084 (X86_TUNE_USE_GATHER): Put m_GENERIC last for consistency.
2086 2021-12-01 Jakub Jelinek <jakub@redhat.com>
2088 PR rtl-optimization/102356
2089 * rtl.h (simplify_context): Add assoc_count member and
2090 max_assoc_count static member.
2091 * simplify-rtx.c (simplify_associative_operation): Don't reassociate
2092 more than max_assoc_count times within one outermost simplify_* call.
2093 * dwarf2out.c (mem_loc_descriptor): Optimize binary operation
2094 with both operands the same using DW_OP_dup.
2096 2021-12-01 Tamar Christina <tamar.christina@arm.com>
2098 * match.pd: Move below pattern that rewrites to EQ, NE.
2099 * tree.c (bitmask_inv_cst_vector_p): Correct do .. while indentation.
2101 2021-12-01 Siddhesh Poyarekar <siddhesh@gotplt.org>
2102 Martin Liška <mliska@suse.cz>
2104 PR tree-optimization/103456
2105 * tree-object-size.c (merge_object_sizes): Update osi->changed
2106 only if object_sizes_set succeeded.
2108 2021-11-30 liuhongt <hongtao.liu@intel.com>
2112 * config/i386/i386.md (*x86_64_shld_1): Set_attr
2113 length_immediate to 1.
2114 (*x86_shld_1): Ditto.
2115 (*x86_64_shrd_1): Ditto.
2116 (*x86_shrd_1): Ditto.
2118 2021-11-30 Bill Schmidt <wschmidt@linux.ibm.com>
2120 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2122 (altivec_resolve_new_overloaded_builtin): Likewise.
2124 2021-11-30 Martin Jambor <mjambor@suse.cz>
2127 * ipa-sra.c (scan_function): Also check ECF_LOOPING_CONST_OR_PURE flag.
2129 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2131 PR tree-optimization/103494
2132 * tree-vect-stmts.c (vect_get_gather_scatter_ops): Remove ncopies
2133 argument and calculate ncopies from gs_info->offset_vectype
2135 (vectorizable_store, vectorizable_load): Update accordingly.
2137 2021-11-30 Iain Buclaw <ibuclaw@gdcproject.org>
2139 * Makefile.in (GDC): New variable.
2140 (GDCFLAGS): New variable.
2141 * configure: Regenerate.
2142 * configure.ac: Add call to ACX_PROG_GDC. Substitute GDCFLAGS.
2144 2021-11-30 Martin Liska <mliska@suse.cz>
2145 Martin Jambor <mjambor@suse.cz>
2148 * ipa-param-manipulation.c
2149 (ipa_param_body_adjustments::prepare_debug_expressions): Be
2150 careful about hash_map reallocating itself. Simpify a return
2151 which always returns true.
2153 2021-11-30 Andrew MacLeod <amacleod@redhat.com>
2155 PR tree-optimization/103440
2156 * gimple-range-fold.cc (fold_using_range::range_of_phi): Continue
2157 normal param processing for equiv params.
2159 2021-11-30 Richard Biener <rguenther@suse.de>
2161 * config/i386/i386.c (ix86_shift_rotate_cost): Remove
2163 * tree-chrec.c (evolution_function_is_invariant_rec_p):
2165 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
2166 * tree-ssa-pre.c (fully_constant_expression): Likewise.
2167 * tree-vrp.c (operand_less_p): Likewise.
2168 * reload.c (reg_overlap_mentioned_for_reload_p): Remove
2169 unreachable gcc_unreachable ().
2170 * sel-sched-ir.h (bb_next_bb): Likewise.
2171 * varasm.c (compare_constant): Likewise.
2173 2021-11-30 Richard Biener <rguenther@suse.de>
2175 PR tree-optimization/103489
2176 * tree-vect-loop.c (vectorizable_phi): Verify argument
2177 vector type compatibility to mitigate bool pattern recog
2180 2021-11-30 Thomas Schwinge <thomas@codesourcery.com>
2182 * omp-offload.c (oacc_loop_auto_partitions): Remove erroneous
2183 "Orphan reductions cannot have gang partitioning" handling.
2185 2021-11-30 Thomas Schwinge <thomas@codesourcery.com>
2187 * omp-offload.c (oacc_loop_process): Implement "gang reduction on
2188 an orphan loop" checking.
2190 2021-11-30 Cesar Philippidis <cesar@codesourcery.com>
2191 Thomas Schwinge <thomas@codesourcery.com>
2193 * omp-general.h (enum oacc_loop_flags): Add OLF_REDUCTION enum.
2194 * omp-low.c (lower_oacc_head_mark): Use it to mark OpenACC
2196 * omp-offload.c (oacc_loop_auto_partitions): Don't assign gang
2197 level parallelism to orphan reductions.
2199 2021-11-30 Aldy Hernandez <aldyh@redhat.com>
2201 PR tree-optimization/103451
2202 * range-op.cc (operator_div::wi_fold): Remove
2203 can_throw_non_call_exceptions special case.
2204 * tree-ssa-sink.c (sink_code_in_bb): Same.
2206 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2208 * tree-vect-slp.c (arg1_arg4_map): New variable.
2209 (vect_get_operand_map): Handle IFN_MASK_GATHER_LOAD.
2210 (vect_build_slp_tree_1): Likewise.
2211 (vect_build_slp_tree_2): Likewise.
2212 * tree-vect-stmts.c (vectorizable_load): Expect the mask to be
2213 the last SLP child node rather than the first.
2215 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2217 * tree-if-conv.c: Include tree-eh.h.
2218 (predicate_statements): Remove pe argument. Don't hoist
2220 (combine_blocks): Remove pe argument.
2221 (ifcvt_available_on_edge_p, ifcvt_can_hoist): New functions.
2222 (ifcvt_hoist_invariants): Likewise.
2223 (tree_if_conversion): Update call to combine_blocks. Call
2224 ifcvt_hoist_invariants after VN.
2226 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2228 PR tree-optimization/102467
2229 * doc/sourcebuild.texi (vect_gather_load_ifn): Document.
2230 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
2231 Commonize safelen handling. Punt for anything involving
2232 gathers and scatters unless safelen says otherwise.
2233 * tree-vect-slp.c (arg1_map): New variable.
2234 (vect_get_operand_map): Handle IFN_GATHER_LOAD.
2235 (vect_build_slp_tree_1): Likewise.
2236 (vect_build_slp_tree_2): Likewise.
2237 (compatible_calls_p): If vect_get_operand_map returns nonnull,
2238 check that any skipped arguments are equal.
2239 (vect_slp_analyze_node_operations_1): Tighten reduction check.
2240 * tree-vect-stmts.c (check_load_store_for_partial_vectors): Take
2241 an ncopies argument.
2242 (vect_get_gather_scatter_ops): Take slp_node and ncopies arguments.
2244 (vectorizable_store, vectorizable_load): Adjust accordingly.
2246 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2248 * gimple.h (gimple_num_args, gimple_arg, gimple_arg_ptr): New
2250 * tree-vect-slp.c (cond_expr_maps, arg2_map): New variables.
2251 (vect_get_operand_map): New function.
2252 (vect_get_and_check_slp_defs): Fix outdated comment.
2253 Use vect_get_operand_map and new gimple argument accessors.
2254 (vect_build_slp_tree_2): Likewise.
2256 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2258 * tree-vect-slp.c (vect_build_slp_tree_1): Use code_helper
2259 to record the operations performed by statements, only using
2260 CALL_EXPR for things that don't map to built-in or internal
2261 functions. For shifts, require all shift amounts to be equal
2262 if optab_vector is not supported but optab_scalar is.
2264 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2266 * tree-vect-data-refs.c (vect_check_gather_scatter): Continue
2267 processing conversions if the current offset is a pointer.
2269 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2271 * tree-vectorizer.h (vect_is_reduction): Use STMT_VINFO_REDUC_IDX.
2273 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2275 * genopinit.c (main): Turn supports_vec_gather_load and
2276 supports_vec_scatter_store into signed char arrays and remove
2277 supports_vec_gather_load_cached and supports_vec_scatter_store_cached.
2278 * optabs-query.c (supports_vec_convert_optab_p): Add a mode parameter.
2279 If the mode is not VOIDmode, test only for that mode.
2280 (supports_vec_gather_load_p): Likewise.
2281 (supports_vec_scatter_store_p): Likewise.
2282 * optabs-query.h (supports_vec_gather_load_p): Likewise.
2283 (supports_vec_scatter_store_p): Likewise.
2284 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
2285 vector mode to supports_vec_gather_load_p and
2286 supports_vec_scatter_store_p.
2288 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2290 * internal-fn.c (commutative_binary_fn_p): Handle IFN_ADD_OVERFLOW
2291 and IFN_MUL_OVERFLOW.
2293 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2295 * internal-fn.c (commutative_binary_fn_p): Handle IFN_UBSAN_CHECK_ADD
2296 and IFN_UBSAN_CHECK_MUL.
2298 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2300 * internal-fn.c (commutative_binary_fn_p): Handle IFN_COMPLEX_MUL.
2302 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2304 * gimple-fold.c: Include internal-fn.h.
2305 (fold_stmt_1): If a function maps to an internal one, use
2306 first_commutative_argument to canonicalize the order of
2307 commutative arguments.
2308 * gimple-match-head.c (gimple_resimplify2, gimple_resimplify3)
2309 (gimple_resimplify4, gimple_resimplify5): Extend commutativity
2310 checks to functions.
2312 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2314 * doc/md.texi (reduc_fmin_scal_@var{m}): Document.
2315 (reduc_fmax_scal_@var{m}): Likewise.
2316 * optabs.def (reduc_fmax_scal_optab): New optab.
2317 (reduc_fmin_scal_optab): Likewise
2318 * internal-fn.def (REDUC_FMAX, REDUC_FMIN): New functions.
2319 * tree-vect-loop.c (reduction_fn_for_scalar_code): Handle
2320 CASE_CFN_FMAX and CASE_CFN_FMIN.
2321 (neutral_op_for_reduction): Likewise.
2322 (needs_fold_left_reduction_p): Likewise.
2323 * config/aarch64/iterators.md (FMAXMINV): New iterator.
2324 (fmaxmin): Handle UNSPEC_FMAXNMV and UNSPEC_FMINNMV.
2325 * config/aarch64/aarch64-simd.md (reduc_<optab>_scal_<mode>): Fix
2327 (reduc_<fmaxmin>_scal_<mode>): New pattern.
2328 * config/aarch64/aarch64-sve.md (reduc_<fmaxmin>_scal_<mode>):
2331 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2333 * builtins.h (associated_internal_fn): Declare overload that
2334 takes a (combined_cfn, return type) pair.
2335 * builtins.c (associated_internal_fn): Split new overload out
2336 of original fndecl version. Also provide an overload that takes
2337 a (combined_cfn, return type) pair.
2338 * internal-fn.h (commutative_binary_fn_p): Declare.
2339 (commutative_ternary_fn_p): Likewise.
2340 (associative_binary_fn_p): Likewise.
2341 * internal-fn.c (commutative_binary_fn_p, commutative_ternary_fn_p):
2342 New functions, split out from...
2343 (first_commutative_argument): ...here.
2344 (associative_binary_fn_p): New function.
2345 * gimple-match.h (code_helper): Add a constructor that takes
2347 (commutative_binary_op_p): Declare.
2348 (commutative_ternary_op_p): Likewise.
2349 (first_commutative_argument): Likewise.
2350 (associative_binary_op_p): Likewise.
2351 (canonicalize_code): Likewise.
2352 (directly_supported_p): Likewise.
2353 (get_conditional_internal_fn): Likewise.
2354 (gimple_build): New overloads that takes a code_helper.
2355 * gimple-fold.c (gimple_build): Likewise.
2356 * gimple-match-head.c (commutative_binary_op_p): New function.
2357 (commutative_ternary_op_p): Likewise.
2358 (first_commutative_argument): Likewise.
2359 (associative_binary_op_p): Likewise.
2360 (canonicalize_code): Likewise.
2361 (directly_supported_p): Likewise.
2362 (get_conditional_internal_fn): Likewise.
2363 * tree-vectorizer.h: Include gimple-match.h.
2364 (neutral_op_for_reduction): Take a code_helper instead of a tree_code.
2365 (needs_fold_left_reduction_p): Likewise.
2366 (reduction_fn_for_scalar_code): Likewise.
2367 (vect_can_vectorize_without_simd_p): Declare a nNew overload that takes
2369 * tree-vect-loop.c: Include case-cfn-macros.h.
2370 (fold_left_reduction_fn): Take a code_helper instead of a tree_code.
2371 (reduction_fn_for_scalar_code): Likewise.
2372 (neutral_op_for_reduction): Likewise.
2373 (needs_fold_left_reduction_p): Likewise.
2374 (use_mask_by_cond_expr_p): Likewise.
2375 (build_vect_cond_expr): Likewise.
2376 (vect_create_partial_epilog): Likewise. Use gimple_build rather
2377 than gimple_build_assign.
2378 (check_reduction_path): Handle calls and operate on code_helpers
2379 rather than tree_codes.
2380 (vect_is_simple_reduction): Likewise.
2381 (vect_model_reduction_cost): Likewise.
2382 (vect_find_reusable_accumulator): Likewise.
2383 (vect_create_epilog_for_reduction): Likewise.
2384 (vect_transform_cycle_phi): Likewise.
2385 (vectorizable_reduction): Likewise. Make more use of
2387 (vect_transform_reduction): Use gimple_extract_op but expect
2388 a tree_code for now.
2389 (vect_can_vectorize_without_simd_p): New overload that takes
2391 * tree-vect-stmts.c (vectorizable_call): Handle reductions in
2393 * tree-vect-patterns.c (vect_mark_pattern_stmts): Use
2394 gimple_extract_op when updating STMT_VINFO_REDUC_IDX.
2396 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2398 * gimple-match.h (code_helper): Provide == and != overloads.
2399 (code_helper::operator tree_code): Make explicit.
2400 (code_helper::operator combined_fn): Likewise.
2401 * gimple-match-head.c (convert_conditional_op): Use explicit
2402 conversions where necessary.
2403 (gimple_resimplify1, gimple_resimplify2, gimple_resimplify3): Likewise.
2404 (maybe_push_res_to_seq, gimple_simplify): Likewise.
2405 * gimple-fold.c (replace_stmt_with_simplification): Likewise.
2407 2021-11-30 Richard Sandiford <richard.sandiford@arm.com>
2409 * gimple-match.h (code_helper): Add functions for querying whether
2410 the code represents an internal_fn or a built_in_function.
2411 Provide explicit conversion operators for both cases.
2412 (gimple_extract_op): Declare.
2413 * gimple-match-head.c (gimple_extract): New function, extracted from...
2414 (gimple_simplify): ...here.
2415 (gimple_extract_op): New function.
2417 2021-11-30 Eric Botcazou <ebotcazou@adacore.com>
2420 * config/i386/i386.c (ix86_output_call_insn): Beef up comment about
2421 nops emitted with SEH.
2422 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
2423 the cold section, emit a nop before the directive if the previous
2424 active instruction is a call.
2426 2021-11-30 Roger Sayle <roger@nextmovesoftware.com>
2427 Uroš Bizjak <ubizjak@gmail.com>
2430 * config/i386/sse.md (define_split): New splitters to simplify
2431 not;vec_duplicate;and as vec_duplicate;andn.
2433 2021-11-30 Richard Biener <rguenther@suse.de>
2435 PR middle-end/103485
2436 * match.pd (-((int)x >> (prec - 1)) to (unsigned)x >> (prec - 1)):
2437 Use VIEW_CONVERT_EXPR for vectors.
2439 2021-11-30 Richard Biener <rguenther@suse.de>
2441 * cfgrtl.c (skip_insns_after_block): Refactor code to
2442 be more easily readable.
2443 * expr.c (op_by_pieces_d::run): Remove unreachable
2445 * sched-deps.c (sched_analyze): Remove unreachable
2447 * sel-sched-ir.c (in_same_ebb_p): Likewise.
2448 * tree-ssa-alias.c (nonoverlapping_refs_since_match_p):
2449 Remove unreachable code.
2450 * tree-vect-slp.c (vectorize_slp_instance_root_stmt):
2451 Refactor to avoid unreachable loop iteration.
2452 * tree.c (walk_tree_1): Remove unreachable break.
2453 * vec-perm-indices.c (vec_perm_indices::series_p): Remove
2456 2021-11-30 Kewen Lin <linkw@linux.ibm.com>
2459 * config/rs6000/rs6000-call.c (rs6000_builtin_decl): Remove builtin mask
2462 2021-11-30 Kewen Lin <linkw@linux.ibm.com>
2464 * config/rs6000/rs6000.c
2465 (rs6000_cost_data::update_target_cost_per_stmt): Adjust the way to
2466 compute extra penalized cost. Remove useless parameter.
2467 (rs6000_cost_data::rs6000_add_stmt_cost): Adjust the call to function
2468 update_target_cost_per_stmt.
2470 2021-11-30 Kewen Lin <linkw@linux.ibm.com>
2473 2021-11-17 Kewen Lin <linkw@linux.ibm.com>
2475 * config/visium/visium.md (*add<mode>3_insn, *addsi3_insn, *addi3_insn,
2476 *sub<mode>3_insn, *subsi3_insn, *subdi3_insn, *neg<mode>2_insn,
2477 *negdi2_insn, *and<mode>3_insn, *ior<mode>3_insn, *xor<mode>3_insn,
2478 *one_cmpl<mode>2_insn, *ashl<mode>3_insn, *ashr<mode>3_insn,
2479 *lshr<mode>3_insn, *trunchiqi2_insn, *truncsihi2_insn,
2480 *truncdisi2_insn, *extendqihi2_insn, *extendqisi2_insn,
2481 *extendhisi2_insn, *extendsidi2_insn, *zero_extendqihi2_insn,
2482 *zero_extendqisi2_insn, *zero_extendsidi2_insn): Fix split condition.
2484 2021-11-30 Andrew MacLeod <amacleod@redhat.com>
2486 PR tree-optimization/103467
2487 * gimple-range-gori.cc (range_def_chain::register_dependency): Don't
2488 use an object reference after a potential resize.
2490 2021-11-29 Uroš Bizjak <ubizjak@gmail.com>
2493 * config/i386/i386.md (*movhi_internal): Introduce (*v,C) alternative.
2494 Do not allocate non-GPR registers. Optimize xmm->xmm moves when
2495 optimizing for size. Fix vpinsrw insn template.
2496 (*movhf_internal): Fix pinsrw and pextrw insn templates for
2497 AVX targets. Use sselog1 type instead of sselog. Optimize GPR moves.
2498 Optimize xmm->xmm moves for TARGET_SSE_PARTIAL_REG_DEPENDENCY
2499 and TARGET_SSE_SPLIT_REGS targets.
2501 2021-11-29 Eric Gallager <egallager@gcc.gnu.org>
2503 * configure: Re-regenerate.
2505 2021-11-29 Eric Gallager <egallager@gcc.gnu.org>
2508 * Makefile.in: Substitute CTAGS, ETAGS, and CSCOPE
2509 variables. Use ETAGS variable in TAGS target.
2510 * configure: Regenerate.
2511 * configure.ac: Allow CTAGS, ETAGS, and CSCOPE
2512 variables to be overridden.
2514 2021-11-29 Paul A. Clarke <pc@us.ibm.com>
2516 * config/rs6000/xmmintrin.h (_mm_movemask_ps): Use vec_extractm
2518 * config/rs6000/emmintrin.h (_mm_movemask_pd): Likewise.
2519 (_mm_movemask_epi8): Likewise.
2521 2021-11-29 Richard Biener <rguenther@suse.de>
2523 * read-rtl-function.c (function_reader::read_rtx_operand):
2524 Return only after resetting m_in_call_function_usage.
2526 2021-11-29 Roger Sayle <roger@nextmovesoftware.com>
2528 * config/i386/i386-expand.c (ix86_expand_v1ti_to_ti): Perform the
2529 conversion via V2DImode using vec_extractv2didi on TARGET_SSE2.
2530 * config/i386/sse.md (rotlv1ti3, rotrv1ti3): Change constraint
2531 on QImode shift amounts from const_int_operand to general_operand.
2533 2021-11-29 Richard Biener <rguenther@suse.de>
2535 * tree.h (reverse_storage_order_for_component_p): Remove
2536 spurious gcc_unreachable.
2537 * cfganal.c (dfs_find_deadend): Likewise.
2538 * fold-const-call.c (fold_const_logb): Likewise.
2539 (fold_const_significand): Likewise.
2540 * gimple-ssa-store-merging.c (lhs_valid_for_store_merging_p):
2543 2021-11-29 Richard Biener <rguenther@suse.de>
2545 * vec.c (qsort_chk): Do not return the void return value
2546 from the noreturn qsort_chk_error.
2547 * ccmp.c (expand_ccmp_expr_1): Remove unreachable return.
2548 * df-scan.c (df_ref_equal_p): Likewise.
2549 * dwarf2out.c (is_base_type): Likewise.
2550 (add_const_value_attribute): Likewise.
2551 * fixed-value.c (fixed_arithmetic): Likewise.
2552 * gimple-fold.c (gimple_fold_builtin_fputs): Likewise.
2553 * gimple-ssa-strength-reduction.c (stmt_cost): Likewise.
2554 * graphite-isl-ast-to-gimple.c
2555 (gcc_expression_from_isl_expr_op): Likewise.
2556 (gcc_expression_from_isl_expression): Likewise.
2557 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate):
2559 * lto-streamer-in.c (lto_input_mode_table): Likewise.
2561 2021-11-29 liuhongt <hongtao.liu@intel.com>
2564 * config/i386/i386.c (inline_secondary_memory_needed): HImode
2565 move between GPR and SSE registers is supported under
2566 TARGET_SSE2 and above.
2567 * config/i386/i386.md (extendhfsf2): Optimize expander.
2568 (truncsfhf2): Ditto.
2569 * config/i386/sse.md (sse2p4_1): Adjust attr for V8HFmode to
2570 align with V8HImode.
2572 2021-11-29 liuhongt <hongtao.liu@intel.com>
2576 * config/i386/i386.c (ix86_secondary_reload): Without
2577 TARGET_SSE4_1, General register is needed to move HImode from
2578 sse register to memory.
2579 * config/i386/sse.md (*vec_extrachf): Use %vpextrw instead of
2580 pextrw in output templates.
2581 * config/i386/i386.md (movhi_internal): Ditto, also fix typo of
2582 MEM_P (operands[1]) and adjust mode/prefix/type attribute for
2583 alternatives related to sse register.
2585 2021-11-29 Richard Biener <rguenther@suse.de>
2587 PR tree-optimization/103458
2588 * tree-ssa-dce.c (make_forwarders_with_degenerate_phis): Do not
2589 create forwarders for blocks with abnormal predecessors.
2591 2021-11-29 Richard Biener <rguenther@suse.de>
2593 * gimple-predicate-analysis.cc (can_be_invalidated_p):
2594 Restore semantics to the one before the split from
2597 2021-11-28 Jan Hubicka <hubicka@ucw.cz>
2599 * profile.c: Include sreal.h
2600 (struct bb_stats): New.
2601 (cmp_stats): New function.
2602 (compute_branch_probabilities): Output bb stats.
2604 2021-11-28 Jan Hubicka <hubicka@ucw.cz>
2606 * cfghooks.c: Include sreal.h, profile.h.
2607 (profile_record_check_consistency): Fix checking of count counsistency;
2608 record also dynamic mismatches.
2609 * cfgrtl.c (rtl_account_profile_record): Similarly.
2610 * tree-cfg.c (gimple_account_profile_record): Likewise.
2611 * cfghooks.h (struct profile_record): Remove num_mismatched_freq_in,
2612 num_mismatched_freq_out, turn time to double, add
2613 dyn_mismatched_prob_out, dyn_mismatched_count_in,
2614 num_mismatched_prob_out; remove num_mismatched_count_out.
2615 * passes.c (account_profile_1): New function.
2616 (account_profile_in_list): New function.
2617 (pass_manager::dump_profile_report): Rewrite.
2618 (execute_one_ipa_transform_pass): Check profile consistency after
2620 (execute_all_ipa_transforms): Remove cfun test; record all transform
2622 (execute_one_pass): Fix collecting of profile stats.
2624 2021-11-27 Jakub Jelinek <jakub@redhat.com>
2626 PR middle-end/103431
2627 * config/i386/i386.md (x86_64_shld, x86_shld, x86_64_shrd, x86_shrd):
2628 Change insn pattern to accurately describe the instructions.
2629 (*x86_64_shld_1, *x86_shld_1, *x86_64_shrd_1, *x86_shrd_1): New
2630 define_insn patterns.
2631 (*x86_64_shld_2, *x86_shld_2, *x86_64_shrd_2, *x86_shrd_2): New
2632 define_insn_and_split patterns.
2633 (*ashl<dwi>3_doubleword_mask, *ashl<dwi>3_doubleword_mask_1,
2634 *<insn><dwi>3_doubleword_mask, *<insn><dwi>3_doubleword_mask_1,
2635 ix86_rotl<dwi>3_doubleword, ix86_rotr<dwi>3_doubleword): Adjust
2636 splitters for x86_{,64_}sh{l,r}d pattern changes.
2638 2021-11-27 Jakub Jelinek <jakub@redhat.com>
2640 PR tree-optimization/103435
2641 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Avoid UB if
2642 n->range - rsize == 8, just clear both *cmpnop and *cmpxchg in that
2645 2021-11-27 Martin Jambor <mjambor@suse.cz>
2648 * ipa-prop.c (ipcp_transform_function): Call
2649 delete_unreachable_blocks_update_callgraph instead of returning
2652 2021-11-26 Siddhesh Poyarekar <siddhesh@gotplt.org>
2654 * tree-object-size.c (object_sizes_grow, object_sizes_release,
2655 object_sizes_unknown_p, object_sizes_get, object_size_set_force,
2656 object_sizes_set): New functions.
2657 (addr_object_size, compute_builtin_object_size,
2658 expr_object_size, call_object_size, unknown_object_size,
2659 merge_object_sizes, plus_stmt_object_size,
2660 cond_expr_object_size, collect_object_sizes_for,
2661 check_for_plus_in_loops_1, init_object_sizes,
2662 fini_object_sizes): Adjust.
2664 2021-11-26 Siddhesh Poyarekar <siddhesh@gotplt.org>
2666 * tree-object-size.c: New enum.
2667 (object_sizes, computed, addr_object_size,
2668 compute_builtin_object_size, expr_object_size, call_object_size,
2669 merge_object_sizes, plus_stmt_object_size,
2670 collect_object_sizes_for, init_object_sizes, fini_object_sizes,
2671 object_sizes_execute): Replace magic numbers with enums.
2673 2021-11-26 Roger Sayle <roger@nextmovesoftware.com>
2675 * tree-ssa-loop-ivopts.c (cand_value_at): Take a class
2676 tree_niter_desc* argument instead of just a tree for NITER.
2677 If we require the iv candidate value at the end of the final
2678 loop iteration, try using the original loop bound as the
2679 NITER for sufficiently simple loops.
2680 (may_eliminate_iv): Update (only) call to cand_value_at.
2682 2021-11-26 Jan Hubicka <hubicka@ucw.cz>
2684 * ipa-modref.c (analyze_function): Drop parameter F and dump
2685 cgraph node name rather than cfun name.
2686 (modref_generate): Update.
2687 (modref_summaries::insert):Update.
2688 (modref_summaries_lto::insert):Update.
2689 (pass_modref::execute):Update.
2690 (ipa_merge_modref_summary_after_inlining): Improve combining of
2693 2021-11-26 Jan Hubicka <jh@suse.cz>
2696 * ipa-modref.c (update_escape_summary_1): Fix handling of min_flags.
2698 2021-11-26 konglin1 <lingling.kong@intel.com>
2701 * config/i386/i386.c (ix86_can_change_mode_class): Allow 16 bit data in XMM register
2703 * config/i386/i386.md (extendhfsf2): Add extenndhfsf2 for TARGET_F16C.
2704 (extendhfdf2): Restrict extendhfdf for TARGET_AVX512FP16 only.
2705 (*extendhf<mode>2): Rename from extendhf<mode>2.
2706 (truncsfhf2): Likewise.
2707 (truncdfhf2): Likewise.
2708 (*trunc<mode>2): Likewise.
2710 2021-11-26 liuhongt <hongtao.liu@intel.com>
2712 PR middle-end/103419
2713 * match.pd: Fix typo, use the type of second parameter, not
2716 2021-11-25 Jan Hubicka <jh@suse.cz>
2718 * ipa-cp.c (ipa_value_range_from_jfunc): Remove forgotten early return.
2720 2021-11-25 Roger Sayle <roger@nextmovesoftware.com>
2722 PR middle-end/103406
2723 * match.pd (minus @0 @0): Check tree_expr_maybe_infinite_p.
2725 2021-11-25 Martin Jambor <mjambor@suse.cz>
2728 * ipa-prop.h (ipa_get_param): New overload. Move bits of the existing
2730 * ipa-param-manipulation.h (ipa_param_adjustments): New member
2731 function get_updated_index_or_split.
2732 * ipa-param-manipulation.c
2733 (ipa_param_adjustments::get_updated_index_or_split): New function.
2734 * ipa-prop.c (adjust_agg_replacement_values): Reimplement, add
2735 capability to identify scalarized parameters and perform substitution
2737 (ipcp_transform_function): Create descriptors earlier, handle new
2738 return values of adjust_agg_replacement_values.
2740 2021-11-25 Aldy Hernandez <aldyh@redhat.com>
2742 * gimple-range-path.cc (path_range_query::compute_ranges_defined): Remove.
2743 (path_range_query::compute_ranges_in_block): Revert to bitmap order.
2744 * gimple-range-path.h: Remove compute_ranges_defined.
2746 2021-11-25 Andrew Stubbs <ams@codesourcery.com>
2749 * config/gcn/gcn.c (move_callee_saved_registers): Ensure that the
2750 number of spilled registers is counted correctly.
2752 2021-11-25 Jan Hubicka <hubicka@ucw.cz>
2754 * ipa-prop.h (ipa_node_params::ipa_node_params): Initialize
2757 2021-11-25 Andrew MacLeod <amacleod@redhat.com>
2759 PR tree-optimization/103359
2760 * gimple-range-fold.cc (fold_using_range::range_of_phi): If arg is
2761 equivalent to def, don't initially include it's range.
2763 2021-11-25 Jan Hubicka <hubicka@ucw.cz>
2765 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Do not check
2766 gimple_static_chain.
2768 2021-11-25 Richard Biener <rguenther@suse.de>
2770 * cgraphunit.c (symbol_table::output_weakrefs): Remove
2772 (get_alias_symbol): Remove now unused function.
2774 2021-11-25 Richard Biener <rguenther@suse.de>
2776 * cfgrtl.c (rtl_verify_fallthru): Do not stop verifying
2778 (skip_insns_after_block): Remove unreachable break and continue.
2780 2021-11-25 Richard Biener <rguenther@suse.de>
2782 * cfgexpand.c (label_rtx_for_bb): Remove dead loop construct.
2784 2021-11-25 Richard Biener <rguenther@suse.de>
2786 * regset.h (REG_SET_EMPTY_P): New macro.
2787 * cfgcleanup.c (thread_jump): Use REG_SET_EMPTY_P.
2789 2021-11-25 Martin Liska <mliska@suse.cz>
2791 * doc/invoke.texi: Use @option for -Wuninitialized.
2793 2021-11-25 Aldy Hernandez <aldyh@redhat.com>
2795 PR tree-optimization/103254
2796 * gimple-range-path.cc (path_range_query::compute_ranges): Move
2797 exported boolean code...
2798 (path_range_query::compute_imports): ...here.
2800 2021-11-25 Aldy Hernandez <aldyh@redhat.com>
2802 PR tree-optimization/103254
2803 * gimple-range-path.cc (path_range_query::compute_ranges_defined): New
2804 (path_range_query::compute_ranges_in_block): Move to
2805 compute_ranges_defined.
2806 * gimple-range-path.h (compute_ranges_defined): New.
2808 2021-11-25 Jakub Jelinek <jakub@redhat.com>
2810 PR tree-optimization/103417
2811 * match.pd ((X & Y) CMP 0): Only handle eq and ne. Commonalize
2814 2021-11-25 Jakub Jelinek <jakub@redhat.com>
2816 PR tree-optimization/103376
2817 * gimple-ssa-store-merging.c (perform_symbolic_merge): For
2818 BIT_IOR_EXPR, if masked1 && masked2 && masked1 != masked2, don't
2819 punt, but set the corresponding result byte to MARKER_BYTE_UNKNOWN.
2820 For BIT_XOR_EXPR similarly and if masked1 == masked2 and the
2821 byte isn't MARKER_BYTE_UNKNOWN, set the corresponding result byte to
2824 2021-11-25 Jakub Jelinek <jakub@redhat.com>
2827 * doc/invoke.texi (-Wcomma-subscript): Document that for
2828 -std=c++20 the option isn't enabled by default with -Wno-deprecated
2829 but for -std=c++23 it is.
2831 2021-11-24 Jan Hubicka <hubicka@ucw.cz>
2833 * ipa-modref.c (implicit_eaf_flags_for_edge_and_arg): Break out from...
2834 (modref_merge_call_site_flags): ... here.
2835 (ipa_merge_modref_summary_after_inlining): Use it.
2837 2021-11-24 Thomas Schwinge <thomas@codesourcery.com>
2839 * cfgloop.c (verify_loop_structure): Reduce scope of
2840 'class loop *loop' variable.
2841 * ipa-fnsummary.c (analyze_function_body): Likewise.
2842 * loop-init.c (fix_loop_structure): Likewise.
2843 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
2844 * predict.c (predict_loops): Likewise.
2845 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
2846 * tree-vectorizer.c (pass_vectorize::execute): Likewise.
2848 2021-11-24 Andrew MacLeod <amacleod@redhat.com>
2850 PR tree-optimization/103231
2851 * gimple-range.cc (gimple_ranger::gimple_ranger): Create stmt stack.
2852 (gimple_ranger::gimple_ranger): Delete stmt stack.
2853 (gimple_ranger::range_of_stmt): Process depenedencies if they have no
2855 (gimple_ranger::prefill_name): New.
2856 (gimple_ranger::prefill_stmt_dependencies): New.
2857 * gimple-range.h (class gimple_ranger): Add prototypes.
2859 2021-11-24 Andrew MacLeod <amacleod@redhat.com>
2861 * gimple-range-cache.cc (ranger_cache::get_global_range): Always
2862 return a range, return if it came from the cache or not.
2863 (get_non_stale_global_range): Rename to get_global_range, and return
2864 the temporal state in a flag.
2865 * gimple-range-cache.h (get_non_stale_global_range): Rename and adjust.
2866 * gimple-range.cc (gimple_ranger::range_of_expr): No need to query
2868 (gimple_ranger::range_of_stmt): Adjust for global cache temporal state
2871 2021-11-24 Andrew MacLeod <amacleod@redhat.com>
2873 * gimple-range.cc (gimple_ranger::range_on_edge): Call trailer when
2874 a constant is encountered to terminate the trace.
2876 2021-11-24 Maciej W. Rozycki <macro@embecosm.com>
2878 PR middle-end/103059
2879 * reload.c (find_reloads_address_1): Also accept the ASHIFT form
2880 of indexed addressing.
2881 (find_reloads): Adjust accordingly.
2883 2021-11-24 Richard Biener <rguenther@suse.de>
2884 Jan Hubicka <jh@suse.cz>
2886 PR tree-optimization/103168
2887 * ipa-modref.h (struct modref_summary): Add load_accesses.
2888 * ipa-modref.c (modref_summary::finalize): Initialize load_accesses.
2889 * tree-ssa-sccvn.c (visit_reference_op_call): Use modref
2890 info to walk the virtual use->def chain to CSE const/pure
2891 function calls possibly reading from memory.
2893 2021-11-24 Thomas Schwinge <thomas@codesourcery.com>
2895 PR middle-end/103244
2896 * gimplify.c (gimplify_adjust_omp_clauses): Restore previous
2899 2021-11-24 Richard Biener <rguenther@suse.de>
2901 PR middle-end/103193
2902 * match.pd: Avoid canonicalizing (le/ge @0 @0) to (eq @0 @0)
2903 with NaNs and -ftrapping-math.
2905 2021-11-24 Jakub Jelinek <jakub@redhat.com>
2907 PR middle-end/103384
2908 * omp-general.c (omp_context_selector_matches): For ACCEL_COMPILER,
2909 return 0 for kind(host) and continue for kind(nohost).
2911 2021-11-24 Jakub Jelinek <jakub@redhat.com>
2913 PR middle-end/103365
2914 * attribs.h (lookup_attribute): Allow attr_name to start with
2915 underscore, as long as canonicalize_attr_name returns false.
2916 (lookup_attribute_by_prefix): Don't call get_attribute_name twice.
2917 * attribs.c (extract_attribute_substring): Reimplement using
2918 canonicalize_attr_name.
2919 (register_scoped_attribute): Change gcc_assert into
2920 gcc_checking_assert, verify !canonicalize_attr_name rather than
2921 that str.str doesn't start with '_'.
2923 2021-11-24 Jakub Jelinek <jakub@redhat.com>
2925 PR tree-optimization/103376
2926 * gimple-ssa-store-merging.c (perform_symbolic_merge): Add CODE
2927 argument. If CODE is not BIT_IOR_EXPR, ensure that one of masked1
2929 (find_bswap_or_nop_1, find_bswap_or_nop,
2930 imm_store_chain_info::try_coalesce_bswap): Adjust
2931 perform_symbolic_merge callers.
2933 2021-11-24 Richard Biener <rguenther@suse.de>
2935 * tree-ssa-loop-ivopts.c (find_givs): Take loop body as
2936 argument instead of re-computing it.
2937 (find_interesting_uses): Likewise.
2938 (find_induction_variables): Pass through loop body.
2939 (tree_ssa_iv_optimize_loop): Pass down loop body.
2941 2021-11-24 Tamar Christina <tamar.christina@arm.com>
2943 * tree-ssa-phiopt.c (spaceship_replacement): Handle new canonical
2946 2021-11-24 Tamar Christina <tamar.christina@arm.com>
2948 * tree.c (bitmask_inv_cst_vector_p): New.
2949 * tree.h (bitmask_inv_cst_vector_p): New.
2950 * match.pd: Use it in new bitmask compare pattern.
2952 2021-11-24 Jason Merrill <jason@redhat.com>
2954 * timevar.h (class auto_cond_timevar): New.
2956 2021-11-24 Hongtao Liu <hongtao.liu@intel.com>
2957 H.J. Lu <hongjiu.lu@intel.com>
2959 PR tree-optimization/103194
2960 * match.pd (gimple_nop_atomic_bit_test_and_p): Extended to
2962 * tree-ssa-ccp.c (gimple_nop_convert): Declare.
2963 (optimize_atomic_bit_test_and): Enhance
2964 optimize_atomic_bit_test_and to handle truncation.
2966 2021-11-23 Martin Sebor <msebor@redhat.com>
2969 * Makefile.in (OBJS): Add gimple-warn-recursion.o.
2970 * common.opt: Add -Winfinite-recursion.
2971 * doc/invoke.texi (-Winfinite-recursion): Document.
2972 * passes.def (pass_warn_recursion): Schedule a new pass.
2973 * tree-pass.h (make_pass_warn_recursion): Declare.
2974 * gimple-warn-recursion.c: New file.
2976 2021-11-23 Bill Schmidt <wschmidt@linux.ibm.com>
2978 * config/rs6000/rs6000-call.c (rs6000_invalid_new_builtin): Change
2979 error messages for ENB_P8V and ENB_P9V.
2981 2021-11-23 Bill Schmidt <wschmidt@linux.ibm.com>
2983 * config/rs6000/rs6000-builtin-new.def: Add power6-64 stanza. Move
2984 CMPB to power6-64 stanza.
2985 * config/rs6000/rs6000-call.c (rs6000_invalid_new_builtin): Handle
2987 (rs6000_new_builtin_is_supported): Likewise.
2988 (rs6000_expand_new_builtin): Likewise. Clean up formatting.
2989 (rs6000_init_builtins): Handle ENB_P6_64 case.
2990 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): Add BSTZ_P6_64.
2991 (stanza_map): Add entry mapping power6-64 to BSTZ_P6_64.
2992 (enable_string): Add "ENB_P6_64".
2993 (write_decls): Add ENB_P6_64 to bif_enable enum.
2995 2021-11-23 Jan Hubicka <jh@suse.cz>
2997 * ipa-modref-tree.h (struct modref_tree): Remove max_bases, max_refs
2999 (modref_tree::modref_tree): Remove parametr.
3000 (modref_tree::insert_base): Add max_bases parameter.
3001 (modref_tree::insert): Add max_bases, max_refs, max_accesses
3003 (modref_tree::insert): New member function.
3004 (modref_tree::merge): Add max_bases, max_refs, max_accesses
3006 (modref_tree::insert): New member function.
3007 * ipa-modref-tree.c (test_insert_search_collapse): Update.
3008 (test_merge): Update.
3009 * ipa-modref.c (dump_records): Don't dump max_refs and max_bases.
3010 (dump_lto_records): Likewise.
3011 (modref_summary::finalize): Fix whitespace.
3012 (get_modref_function_summary): Likewise.
3013 (modref_access_analysis::record_access): Update.
3014 (modref_access_analysis::record_access_lto): Update.
3015 (modref_access_analysis::process_fnspec): Update.
3016 (analyze_function): Update.
3017 (modref_summaries::duplicate): Update.
3018 (modref_summaries_lto::duplicate): Update.
3019 (write_modref_records): Update.
3020 (read_modref_records): Update.
3021 (read_section): Update.
3022 (propagate_unknown_call): Update.
3023 (modref_propagate_in_scc): Update.
3024 (ipa_merge_modref_summary_after_inlining): Update.
3026 2021-11-23 Martin Liska <mliska@suse.cz>
3028 * doc/invoke.texi: Remove 2 more duplicite param descriptions.
3030 2021-11-23 Richard Biener <rguenther@suse.de>
3032 PR tree-optimization/103361
3033 * gimple-loop-jam.c (adjust_unroll_factor): Use lambda_int
3034 for the dependence distance.
3035 * tree-data-ref.c (print_lambda_vector): Properly print a lambda_int.
3037 2021-11-23 Jakub Jelinek <jakub@redhat.com>
3039 * tree-inline.h (struct copy_body_data): Remove
3040 transform_lang_insert_block member.
3041 * tree-inline.c (remap_block): Don't call
3042 id->transform_lang_insert_block.
3043 (optimize_inline_calls, copy_gimple_seq_and_replace_locals,
3044 tree_function_versioning, maybe_inline_call_in_expr,
3045 copy_fn): Don't initialize id.transform_lang_insert_block.
3046 * gimplify.c (gimplify_omp_loop): Likewise.
3048 2021-11-23 Jan Hubicka <hubicka@ucw.cz>
3050 PR tree-optimization/103335
3051 * tree-ssa-dse.c (valid_ao_ref_for_dse): Rename to ...
3052 (valid_ao_ref_kill_for_dse): ... this; do not check that boundaries
3053 are divisible by BITS_PER_UNIT.
3054 (get_byte_aligned_range_containing_ref): New function.
3055 (get_byte_aligned_range_contained_in_ref): New function.
3056 (normalize_ref): Rename to ...
3057 (get_byte_range): ... this one; handle accesses not aligned to byte
3058 boundary; return range in bytes rater than updating ao_ref.
3059 (clear_live_bytes_for_ref): Take write ref by reference; simplify using
3061 (setup_live_bytes_from_ref): Likewise.
3062 (clear_bytes_written_by): Update.
3063 (live_bytes_read): Update.
3064 (dse_classify_store): Simplify tech before live_bytes_read checks.
3066 2021-11-23 Andrew Pinski <apinski@marvell.com>
3068 PR tree-optimization/102216
3069 * gimple-fold.c (fold_stmt_1): Add canonicalization
3070 of "&MEM[ssa_n, CST]" to "ssa_n p+ CST", note this
3071 can only be done if !in_place.
3073 2021-11-23 Jakub Jelinek <jakub@redhat.com>
3075 PR middle-end/102431
3076 * gimplify.c (replace_reduction_placeholders): Remove.
3077 (note_no_context_vars): New function.
3078 (gimplify_omp_loop): For OMP_PARALLEL's BIND_EXPR create a new
3079 BLOCK. Use copy_tree_body_r with walk_tree instead of unshare_expr
3080 and replace_reduction_placeholders for duplication of
3081 OMP_CLAUSE_REDUCTION_{INIT,MERGE} expressions. Ensure all mentioned
3082 automatic vars have DECL_CONTEXT set to non-NULL before doing so
3083 and reset it afterwards for those vars and their corresponding
3086 2021-11-23 Haochen Gui <guihaoc@gcc.gnu.org>
3089 * config/rs6000/altivec.md (altivec_vreve<mode>2 for VEC_K): Use
3090 xxbrq for v16qi, xxbrq + xxbrh for v8hi and xxbrq + xxbrw for v4si
3091 or v4sf when p9_vector is set.
3092 (altivec_vreve<mode>2 for VEC_64): Defined. Implemented by xxswapd.
3094 2021-11-23 Navid Rahimi <navidrahimi@microsoft.com>
3096 PR tree-optimization/102232
3097 * match.pd (x * (1 + y / x) - y) -> (x - y % x): New optimization.
3099 2021-11-23 Navid Rahimi <navidrahimi@microsoft.com>
3101 PR tree-optimization/96779
3102 * match.pd (-x == x) -> (x == 0): New optimization.
3104 2021-11-22 Roger Sayle <roger@nextmovesoftware.com>
3106 PR tree-optimization/98953
3107 PR tree-optimization/103345
3108 * gimple-ssa-store-merging.c (find_bswap_or_nop_1): Handle
3109 BIT_XOR_EXPR and PLUS_EXPR the same as BIT_IOR_EXPR.
3110 (pass_optimize_bswap::execute): Likewise.
3112 2021-11-22 Martin Liska <mliska@suse.cz>
3114 * doc/invoke.texi: Remove duplicate documentation for 3 params.
3116 2021-11-22 Richard Biener <rguenther@suse.de>
3118 PR tree-optimization/103351
3119 * tree-ssa-dce.c (sort_phi_args): Sort after e->dest_idx as
3122 2021-11-22 liuhongt <hongtao.liu@intel.com>
3125 * config/i386/constraints.md (Bk): New
3126 define_memory_constraint.
3127 * config/i386/i386-protos.h (ix86_gpr_tls_address_pattern_p):
3129 * config/i386/i386.c (ix86_gpr_tls_address_pattern_p): New
3131 * config/i386/i386.md (*movsi_internal): Don't allow
3132 mask/sse/mmx move in TLS code sequences.
3133 (*movdi_internal): Ditto.
3135 2021-11-22 Kewen Lin <linkw@linux.ibm.com>
3137 * config/xtensa/xtensa.md (movdi_internal, movdf_internal): Fix split
3140 2021-11-21 Jakub Jelinek <jakub@redhat.com>
3143 * config/i386/i386-options.c (ix86_valid_target_attribute_p): If
3144 fndecl already has DECL_FUNCTION_SPECIFIC_TARGET, use that as base
3145 instead of target_option_default_node.
3147 2021-11-21 Jan Hubicka <hubicka@ucw.cz>
3150 * ipa-modref.c (parm_map_for_arg): Rename to ...
3151 (parm_map_for_ptr): .. this one; handle static chain and calls to
3153 (modref_access_analysis::get_access): Use parm_map_for_ptr.
3154 (modref_access_analysis::process_fnspec): Update.
3155 (modref_access_analysis::analyze_load): Update.
3156 (modref_access_analysis::analyze_store): Update.
3158 2021-11-21 Jan Hubicka <hubicka@ucw.cz>
3160 * ipa-modref.c (ignore_nondeterminism_p): Move earlier in source
3162 (ignore_retval_p): Likewise.
3163 (ignore_stores_p): Likewise.
3164 (parm_map_for_arg): Likewise.
3165 (class modref_access_analysis): New class.
3166 (modref_access_analysis::set_side_effects): New member function.
3167 (modref_access_analysis::set_nondeterministic): New member function.
3168 (get_access): Turn to ...
3169 (modref_access_analysis::get_access): ... this one.
3170 (record_access): Turn to ...
3171 (modref_access_analysis::record_access): ... this one.
3172 (record_access_lto): Turn to ...
3173 (modref_access_analysis::record_access_lto): ... This one.
3174 (record_access_p): Turn to ...
3175 (modref_access_analysis::record_access_p): ... This one
3176 (modref_access_analysis::record_unknown_load): New member function.
3177 (modref_access_analysis::record_unknown_store): New member function.
3178 (get_access_for_fnspec): Turn to ...
3179 (modref_access_analysis::get_access_for_fnspec): ... this one.
3180 (merge_call_side_effects): Turn to ...
3181 (moderf_access_analysis::merge_call_side_effects): Turn to ...
3182 (collapse_loads): Move later in source code.
3183 (collapse_stores): Move later in source code.
3184 (process_fnspec): Turn to ...
3185 (modref_access_analysis::process_fnspec): ... this one.
3186 (analyze_call): Turn to ...
3187 (modref_access_analysis::analyze_call): ... this one.
3188 (struct summary_ptrs): Remove.
3189 (analyze_load): Turn to ...
3190 (modref_access_analysis::analyze_load): ... this one.
3191 (analyze_store): Turn to ...
3192 (modref_access_analysis::analyze_store): ... this one.
3193 (analyze_stmt): Turn to ...
3194 (modref_access_analysis::analyze_stmt): ... This one.
3195 (remove_summary): Remove.
3196 (modref_access_analysis::propagate): Break out from ...
3197 (modref_access_analysis::analyze): Break out from ...
3198 (analyze_function): ... here.
3200 2021-11-21 Roger Sayle <roger@nextmovesoftware.com>
3201 Robin Dapp <rdapp@linux.ibm.com>
3204 * tree-ssa-math-opts.c (convert_mult_to_widen): Recognize
3205 signed WIDEN_MULT_EXPR if the target supports umul_widen_optab.
3207 2021-11-20 Jan Hubicka <hubicka@ucw.cz>
3210 * ipa-modref.c (ignore_nondeterminism_p): Allow looping pure/cont.
3211 (merge_call_side_effects): Improve debug output.
3213 2021-11-20 Jan Hubicka <hubicka@ucw.cz>
3216 * ipa-pure-const.c (propagate_pure_const): Fix merging of loping flag.
3218 2021-11-20 Jeff Law <jeffreyalaw@gmail.com>
3220 PR tree-optimization/103226
3221 * config/bfin/bfin.md (doloop pattern, splitter and expander): Clobber
3224 2021-11-20 Andrew Pinski <apinski@marvell.com>
3226 PR tree-optimization/103220
3227 * match.pd ((type) X bitop CST): Don't check if CST
3228 fits into the type if only the sign changes.
3230 2021-11-20 Alexandre Oliva <oliva@adacore.com>
3232 PR tree-optimization/102988
3233 * gimple-harden-conditionals.cc (detach_value): Copy SSA_NAME
3234 without decl sharing.
3236 2021-11-19 Iain Sandoe <iain@sandoe.co.uk>
3238 * doc/invoke.texi: Remove whitespace after an @option.
3240 2021-11-19 Paul A. Clarke <pc@us.ibm.com>
3242 * config/rs6000/emmintrin.h (_mm_sad_epu8): Use vec_absd when
3243 _ARCH_PWR9, optimize vec_sum2s when LE.
3245 2021-11-19 Iain Sandoe <iain@sandoe.co.uk>
3248 * config/darwin-driver.c (darwin_driver_init): Handle exported
3249 symbols and symbol lists (suppress automatic export of the TLS
3251 * config/darwin.c (darwin_rename_builtins): Remove workaround.
3252 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
3253 (REAL_LIBGCC_SPEC): Handle revised library uses.
3254 * config/darwin.opt (nodefaultexport): New.
3255 * config/i386/darwin.h (PR80556_WORKAROUND): Remove.
3256 * config/i386/darwin32-biarch.h (PR80556_WORKAROUND): Likewise.
3257 * config/i386/darwin64-biarch.h (PR80556_WORKAROUND): Likewise.
3259 2021-11-19 Martin Jambor <mjambor@suse.cz>
3261 * opts.c (default_options_table): Switch off
3262 flag_semantic_interposition at Ofast.
3263 * doc/invoke.texi (Optimize Options): Document that Ofast switches off
3264 -fsemantic-interposition.
3266 2021-11-19 Jan Hubicka <hubicka@ucw.cz>
3268 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Do not guard modref
3269 by !gimple_call_chain.
3271 2021-11-19 Martin Sebor <msebor@redhat.com>
3275 * doc/invoke.texi (-Waddress): Update.
3277 2021-11-19 Andrew MacLeod <amacleod@redhat.com>
3279 PR tree-optimization/103254
3280 * gimple-range-gori.cc (range_def_chain::get_def_chain): Limit the
3281 depth for all statements with multple ssa names.
3283 2021-11-19 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
3285 * config/s390/s390.md (define_peephole2): Variable insn points
3286 to the first matched insn. Use peep2_next_insn(1) to refer to
3287 the second matched insn.
3289 2021-11-19 Tamar Christina <tamar.christina@arm.com>
3291 PR tree-optimization/103311
3293 * tree-vect-slp-patterns.c (vect_validate_multiplication): Fix CONJ
3294 test to new codegen.
3295 (complex_mul_pattern::matches): Move check downwards.
3297 2021-11-19 Martin Liska <mliska@suse.cz>
3300 2021-11-19 Martin Liska <mliska@suse.cz>
3302 * cfgexpand.c (pass_expand::execute): Use option directly.
3303 * function.c (allocate_struct_function): Likewise.
3304 * gimple-low.c (lower_function_body): Likewise.
3305 (lower_stmt): Likewise.
3306 * gimple-ssa-backprop.c (backprop::prepare_change): Likewise.
3307 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call): Likewise.
3308 * ipa-split.c (split_function): Likewise.
3309 * lto-streamer-in.c (input_function): Likewise.
3310 * sese.c (sese_insert_phis_for_liveouts): Likewise.
3311 * ssa-iterators.h (num_imm_uses): Likewise.
3312 * tree-cfg.c (make_blocks): Likewise.
3313 (gimple_merge_blocks): Likewise.
3314 * tree-inline.c (tree_function_versioning): Likewise.
3315 * tree-loop-distribution.c (generate_loops_for_partition): Likewise.
3316 * tree-sra.c (analyze_access_subtree): Likewise.
3317 * tree-ssa-dce.c (remove_dead_stmt): Likewise.
3318 * tree-ssa-loop-ivopts.c (remove_unused_ivs): Likewise.
3319 * tree-ssa-phiopt.c (spaceship_replacement): Likewise.
3320 * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
3321 * tree-ssa-tail-merge.c (tail_merge_optimize): Likewise.
3322 * tree-ssa-threadedge.c (propagate_threaded_block_debug_into): Likewise.
3323 * tree-ssa.c (gimple_replace_ssa_lhs): Likewise.
3324 (target_for_debug_bind): Likewise.
3325 (insert_debug_temp_for_var_def): Likewise.
3326 (insert_debug_temps_for_defs): Likewise.
3327 (reset_debug_uses): Likewise.
3328 * tree-ssanames.c (release_ssa_name_fn): Likewise.
3329 * tree-vect-loop-manip.c (adjust_vec_debug_stmts): Likewise.
3330 (adjust_debug_stmts): Likewise.
3331 (adjust_phi_and_debug_stmts): Likewise.
3332 (vect_do_peeling): Likewise.
3333 * tree-vect-loop.c (vect_transform_loop_stmt): Likewise.
3334 (vect_transform_loop): Likewise.
3335 * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Remove
3336 (MAY_HAVE_DEBUG_BIND_STMTS): Remove.
3337 (MAY_HAVE_DEBUG_STMTS): Use options directly.
3339 2021-11-19 Giuliano Belinassi <gbelinassi@suse.de>
3341 * gcc.c (process_command): Skip dumpdir override if file is a
3343 * doc/invoke.texi: Update -dumpdir documentation.
3345 2021-11-19 Andrew Pinski <apinski@marvell.com>
3347 PR tree-optimization/103314
3348 * match.pd ((type) X op CST): Restrict the equal
3349 TYPE_PRECISION case to GIMPLE only.
3351 2021-11-19 Martin Liska <mliska@suse.cz>
3354 * ipa-modref-tree.h (struct modref_parm_map): Add default
3356 * ipa-modref.c (ipa_merge_modref_summary_after_inlining): Use it.
3358 2021-11-19 Richard Biener <rguenther@suse.de>
3360 PR middle-end/103248
3361 * tree-eh.c (operation_could_trap_helper_p): Properly handle
3362 fixed-point RDIV_EXPR.
3364 2021-11-19 Richard Biener <rguenther@suse.de>
3366 PR tree-optimization/102436
3367 * tree-ssa-loop-im.c (execute_sm_if_changed): Add mode
3368 to just create the if structure and return the then block.
3369 (execute_sm): Add flag to indicate the var will re-use
3371 (hoist_memory_references): Support a single conditional
3372 block with all stores as special case.
3374 2021-11-19 Andrew Pinski <apinski@marvell.com>
3376 PR tree-optimization/103317
3377 * tree-ssa-phiopt.c (minmax_replacement): For the non empty
3378 middle bb case, check to make sure it has a single predecessor.
3380 2021-11-19 Andrew Pinski <apinski@marvell.com>
3382 PR tree-optimization/103257
3384 ((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0):
3385 Disable until !canonicalize_math_p ().
3387 2021-11-19 Marek Polacek <polacek@redhat.com>
3391 * doc/invoke.texi: Update documentation for -Wuninitialized.
3392 * tree.c (stabilize_reference): Set location.
3394 2021-11-19 liuhongt <hongtao.liu@intel.com>
3397 * config/i386/x86-tune-costs.h (skylake_cost): Reduce cost of
3398 storing 256/512-bit SSE register to be equal to cost of
3399 unaligned store to avoid odd alignment peeling.
3400 (icelake_cost): Ditto.
3402 2021-11-18 David Edelsohn <dje.gcc@gmail.com>
3404 * config/rs6000/predicates.md (current_file_function_operand):
3405 Add flag_semantic_interposition to call of decl_replaceable_p.
3407 2021-11-18 Martin Liska <mliska@suse.cz>
3409 * ipa-modref.c (analyze_function): Do not execute the code
3410 only if dump_file != NULL.
3412 2021-11-18 Martin Liska <mliska@suse.cz>
3414 * ipa-modref.c (analyze_function): Use fnode instead of repeated
3415 cgraph_node::get (current_function_decl).
3417 2021-11-18 Jan Hubicka <hubicka@ucw.cz>
3419 * cgraph.c (cgraph_node::get_availability): Update call of
3421 (cgraph_node::verify_node): Verify that semantic_interposition flag
3423 * cgraph.h: (symtab_node): Add semantic_interposition flag.
3424 * cgraphclones.c (set_new_clone_decl_and_node_flags): Clear
3425 semantic_interposition flag.
3426 * cgraphunit.c (cgraph_node::finalize_function): Set
3427 semantic_interposition flag.
3428 (cgraph_node::add_new_function): Likewise.
3429 (varpool_node::finalize_decl): Likewise.
3430 (cgraph_node::create_wrapper): Likewise.
3431 * common.opt (fsemantic-interposition): Turn to optimization node.
3432 * lto-cgraph.c (lto_output_node): Stream semantic_interposition.
3433 (lto_output_varpool_node): Likewise.
3434 (input_overwrite_node): Likewise.
3435 (input_varpool_node): Likewise.
3436 * symtab.c (symtab_node::dump_base): Dump new flag.
3437 * varasm.c (decl_replaceable_p): Add semantic_interposition_p
3439 * varasm.h (decl_replaceable_p): Update declaration.
3440 * varpool.c (varpool_node::ctor_useable_for_folding_p):
3441 Use semantic_interposition flag.
3442 (varpool_node::get_availability): Likewise.
3443 (varpool_node::create_alias): Copy semantic_interposition flag.
3445 2021-11-18 Jan Hubicka <hubicka@ucw.cz>
3448 * ipa-modref.c (modref_eaf_analysis::merge_call_lhs_flags): Unused
3449 parameter may still be returned.
3450 (modref_eaf_analysis::analyze_ssa_name): Call merge_call_lhs_flags
3451 even for unused function args.
3453 2021-11-18 Tamar Christina <tamar.christina@arm.com>
3455 PR tree-optimization/103311
3456 * tree-vect-slp-patterns.c (complex_mul_pattern::matches,
3457 complex_fms_pattern::matches): Check for multiplications.
3459 2021-11-18 H.J. Lu <hjl.tools@gmail.com>
3462 * config/i386/i386.c (ix86_output_jmp_thunk_or_indirect): Emit
3463 CS prefix for -mindirect-branch-cs-prefix.
3464 (ix86_output_indirect_branch_via_reg): Likewise.
3465 * config/i386/i386.opt: Add -mindirect-branch-cs-prefix.
3466 * doc/invoke.texi: Document -mindirect-branch-cs-prefix.
3468 2021-11-18 David Edelsohn <dje.gcc@gmail.com>
3470 * config/rs6000/driver-rs6000.c (detect_processor_aix): Add
3473 2021-11-18 Tamar Christina <tamar.christina@arm.com>
3475 PR tree-optimization/103253
3476 * tree-ssa-math-opts.c (convert_mult_to_fma): Check for LHS.
3478 2021-11-18 Matthias Kretz <m.kretz@gsi.de>
3480 * doc/extend.texi: Document __builtin_assoc_barrier.
3482 2021-11-18 Martin Liska <mliska@suse.cz>
3484 * cfgexpand.c (pass_expand::execute): Use option directly.
3485 * function.c (allocate_struct_function): Likewise.
3486 * gimple-low.c (lower_function_body): Likewise.
3487 (lower_stmt): Likewise.
3488 * gimple-ssa-backprop.c (backprop::prepare_change): Likewise.
3489 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call): Likewise.
3490 * ipa-split.c (split_function): Likewise.
3491 * lto-streamer-in.c (input_function): Likewise.
3492 * sese.c (sese_insert_phis_for_liveouts): Likewise.
3493 * ssa-iterators.h (num_imm_uses): Likewise.
3494 * tree-cfg.c (make_blocks): Likewise.
3495 (gimple_merge_blocks): Likewise.
3496 * tree-inline.c (tree_function_versioning): Likewise.
3497 * tree-loop-distribution.c (generate_loops_for_partition): Likewise.
3498 * tree-sra.c (analyze_access_subtree): Likewise.
3499 * tree-ssa-dce.c (remove_dead_stmt): Likewise.
3500 * tree-ssa-loop-ivopts.c (remove_unused_ivs): Likewise.
3501 * tree-ssa-phiopt.c (spaceship_replacement): Likewise.
3502 * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
3503 * tree-ssa-tail-merge.c (tail_merge_optimize): Likewise.
3504 * tree-ssa-threadedge.c (propagate_threaded_block_debug_into): Likewise.
3505 * tree-ssa.c (gimple_replace_ssa_lhs): Likewise.
3506 (target_for_debug_bind): Likewise.
3507 (insert_debug_temp_for_var_def): Likewise.
3508 (insert_debug_temps_for_defs): Likewise.
3509 (reset_debug_uses): Likewise.
3510 * tree-ssanames.c (release_ssa_name_fn): Likewise.
3511 * tree-vect-loop-manip.c (adjust_vec_debug_stmts): Likewise.
3512 (adjust_debug_stmts): Likewise.
3513 (adjust_phi_and_debug_stmts): Likewise.
3514 (vect_do_peeling): Likewise.
3515 * tree-vect-loop.c (vect_transform_loop_stmt): Likewise.
3516 (vect_transform_loop): Likewise.
3517 * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Remove
3518 (MAY_HAVE_DEBUG_BIND_STMTS): Remove.
3519 (MAY_HAVE_DEBUG_STMTS): Use options directly.
3521 2021-11-18 Richard Biener <rguenther@suse.de>
3523 PR tree-optimization/103277
3524 * tree-ssa-dse.c (need_ab_cleanup): New.
3525 (dse_optimize_redundant_stores): Adjust.
3526 (delete_dead_or_redundant_assignment): Get extra
3527 need_ab_cleanup argument and set when abnormal cleanup is
3529 (dse_optimize_call): Adjust.
3530 (dse_optimize_stmt): Likewise.
3531 (pass_dse::execute): Allocate and deallocate need_ab_cleanup.
3532 Perform abnormal cleanup.
3533 * tree-ssa-dse.h (delete_dead_or_redundant_assignment): Adjust.
3535 2021-11-18 Hongyu Wang <hongyu.wang@intel.com>
3537 * config/i386/i386-expand.c (ix86_expand_atomic_fetch_op_loop):
3538 Adjust generated cfg to avoid infinite loop.
3540 2021-11-18 konglin1 <lingling.kong@intel.com>
3542 * config/i386/avx512fp16intrin.h (_mm512_mul_pch): Add alias for _mm512_fmul_pch.
3543 (_mm512_mask_mul_pch): Likewise.
3544 (_mm512_maskz_mul_pch): Likewise.
3545 (_mm512_mul_round_pch): Likewise.
3546 (_mm512_mask_mul_round_pch): Likewise.
3547 (_mm512_maskz_mul_round_pch): Likewise.
3548 (_mm512_cmul_pch): Likewise.
3549 (_mm512_mask_cmul_pch): Likewise.
3550 (_mm512_maskz_cmul_pch): Likewise.
3551 (_mm512_cmul_round_pch): Likewise.
3552 (_mm512_mask_cmul_round_pch): Likewise.
3553 (_mm512_maskz_cmul_round_pch): Likewise.
3554 (_mm_mul_sch): Likewise.
3555 (_mm_mask_mul_sch): Likewise.
3556 (_mm_maskz_mul_sch): Likewise.
3557 (_mm_mul_round_sch): Likewise.
3558 (_mm_mask_mul_round_sch): Likewise.
3559 (_mm_maskz_mul_round_sch): Likewise.
3560 (_mm_cmul_sch): Likewise.
3561 (_mm_mask_cmul_sch): Likewise.
3562 (_mm_maskz_cmul_sch): Likewise.
3563 (_mm_cmul_round_sch): Likewise.
3564 (_mm_mask_cmul_round_sch): Likewise.
3565 (_mm_maskz_cmul_round_sch): Likewise.
3566 * config/i386/avx512fp16vlintrin.h (_mm_mul_pch): Likewise.
3567 (_mm_mask_mul_pch): Likewise.
3568 (_mm_maskz_mul_pch): Likewise.
3569 (_mm256_mul_pch): Likewise.
3570 (_mm256_mask_mul_pch): Likewise.
3571 (_mm256_maskz_mul_pch): Likewise.
3572 (_mm_cmul_pch): Likewise.
3573 (_mm_mask_cmul_pch): Likewise.
3574 (_mm_maskz_cmul_pch): Likewise.
3575 (_mm256_cmul_pch): Likewise.
3576 (_mm256_mask_cmul_pch): Likewise.
3577 (_mm256_maskz_cmul_pch): Likewise.
3579 2021-11-17 Andrew Pinski <apinski@marvell.com>
3581 PR tree-optimization/103228
3582 PR tree-optimization/55177
3583 * match.pd ((type) X bitop CST): Also do this
3584 transformation for nop conversions.
3586 2021-11-17 Martin Sebor <msebor@redhat.com>
3588 PR tree-optimization/102759
3589 * gimple-array-bounds.cc (build_printable_array_type): Move...
3590 * gimple-ssa-warn-access.cc (build_printable_array_type): Avoid
3591 pathological function redeclarations that remove a previously
3593 Improve formatting of function arguments in informational notes.
3594 * pointer-query.cc (build_printable_array_type): ...to here.
3595 * pointer-query.h (build_printable_array_type): Declared.
3597 2021-11-17 H.J. Lu <hjl.tools@gmail.com>
3600 * config/i386/i386-opts.h (harden_sls): New enum.
3601 * config/i386/i386.c (output_indirect_thunk): Mitigate against
3602 SLS for function return.
3603 (ix86_output_function_return): Likewise.
3604 (ix86_output_jmp_thunk_or_indirect): Mitigate against indirect
3606 (ix86_output_indirect_jmp): Likewise.
3607 (ix86_output_call_insn): Likewise.
3608 * config/i386/i386.opt: Add -mharden-sls=.
3609 * doc/invoke.texi: Document -mharden-sls=.
3611 2021-11-17 H.J. Lu <hjl.tools@gmail.com>
3614 * config/i386/i386.c (ix86_code_end): Remove "%!" before ret.
3615 (ix86_output_function_return): Likewise.
3616 * config/i386/i386.md (simple_return_pop_internal): Likewise.
3618 2021-11-17 Jan Hubicka <hubicka@ucw.cz>
3621 * ipa-modref.c (read_modref_records): Fix streaminig in of every_access
3624 2021-11-17 Uroš Bizjak <ubizjak@gmail.com>
3626 * config/i386/i386.c (indirect_thunks_used): Redefine as HARD_REG_SET.
3627 (ix86_code_end): Use TEST_HARD_REG_BIT on indirect_thunks_used.
3628 (ix86_output_indirect_branch_via_reg): Use SET_HARD_REG_BIT
3629 on indirect_thunks_used.
3630 (ix86_output_indirect_function_return): Ditto.
3632 2021-11-17 Jan Hubicka <hubicka@ucw.cz>
3634 * ipa-modref-tree.c: Include cgraph.h and tree-streamer.h.
3635 (modref_access_node::stream_out): New member function.
3636 (modref_access_node::stream_in): New member function.
3637 * ipa-modref-tree.h (modref_access_node::stream_out,
3638 modref_access_node::stream_in): Declare.
3639 * ipa-modref.c (modref_summary_lto::useful_p): Free useless kills.
3640 (modref_summary_lto::dump): Dump kills.
3641 (analyze_store): Record kills for LTO
3642 (analyze_stmt): Likewise.
3643 (modref_summaries_lto::duplicate): Duplicate kills.
3644 (write_modref_records): Use new stream_out member function.
3645 (read_modref_records): Likewise.
3646 (modref_write): Stream out kills.
3647 (read_section): Stream in kills
3648 (remap_kills): New function.
3649 (update_signature): Use it.
3651 2021-11-17 Uroš Bizjak <ubizjak@gmail.com>
3653 * config/i386/i386.h (LEGACY_SSE_REGNO_P): New predicate.
3654 (SSE_REGNO_P): Use LEGACY_SSE_REGNO_P predicate.
3655 * config/i386/i386.c (zero_all_vector_registers):
3656 Use LEGACY_SSE_REGNO_P predicate.
3657 (ix86_register_priority): Use REX_INT_REGNO_P, REX_SSE_REGNO_P
3658 and EXT_REG_SSE_REGNO_P predicates.
3659 (ix86_hard_regno_call_part_clobbered): Use REX_SSE_REGNO_P
3660 and LEGACY_SSE_REGNO_P predicates.
3662 2021-11-17 Jason Merrill <jason@redhat.com>
3664 * doc/invoke.texi (C++ Dialect Options): Document
3665 -fimplicit-constexpr.
3667 2021-11-17 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
3669 * config/aarch64/aarch64-modes.def (VECTOR_MODE): New V8DI mode.
3670 * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Handle
3672 * config/aarch64/iterators.md (define_mode_attr nunits): Add entry
3675 2021-11-17 Martin Uecker <uecker@gcc.gnu.org>
3679 * gimplify.c (gimplify_var_or_parm_decl): Update comment.
3680 (gimplify_compound_lval): Gimplify base expression first.
3681 (gimplify_target_expr): Add comment.
3683 2021-11-17 Jakub Jelinek <jakub@redhat.com>
3685 PR tree-optimization/103192
3686 * tree-ssa-loop-im.c (move_computations_worker): Use
3687 reset_flow_sensitive_info instead of manually clearing
3688 SSA_NAME_RANGE_INFO and do it for all SSA_NAMEs, not just ones
3689 with integral types.
3691 2021-11-17 Jakub Jelinek <jakub@redhat.com>
3693 PR tree-optimization/103255
3694 * gimple-range-fold.cc (fold_using_range::range_of_address): Return
3695 range_nonzero rather than unadjusted base's range. Formatting fixes.
3697 2021-11-17 Richard Sandiford <richard.sandiford@arm.com>
3699 * doc/md.texi (cond_fmin@var{mode}, cond_fmax@var{mode}): Document.
3700 * optabs.def (cond_fmin_optab, cond_fmax_optab): New optabs.
3701 * internal-fn.def (COND_FMIN, COND_FMAX): New functions.
3702 * internal-fn.c (first_commutative_argument): Handle them.
3703 (FOR_EACH_COND_FN_PAIR): Likewise.
3704 * match.pd (UNCOND_BINARY, COND_BINARY): Likewise.
3705 * config/aarch64/aarch64-sve.md (cond_<fmaxmin><mode>): New
3708 2021-11-17 Kewen Lin <linkw@linux.ibm.com>
3710 * config/i386/i386.md (*add<dwi>3_doubleword, *addv<dwi>4_doubleword,
3711 *addv<dwi>4_doubleword_1, *sub<dwi>3_doubleword,
3712 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1,
3713 *add<dwi>3_doubleword_cc_overflow_1, *divmodsi4_const,
3714 *neg<dwi>2_doubleword, *tls_dynamic_gnu2_combine_64_<mode>): Fix split
3717 2021-11-17 Andrew Pinski <apinski@marvell.com>
3719 PR tree-optimization/103288
3720 * tree-ssa-phiopt.c (value_replacement): Return early if middle
3721 block has more than one pred.
3723 2021-11-17 Kewen Lin <linkw@linux.ibm.com>
3725 * config/visium/visium.md (*add<mode>3_insn, *addsi3_insn, *addi3_insn,
3726 *sub<mode>3_insn, *subsi3_insn, *subdi3_insn, *neg<mode>2_insn,
3727 *negdi2_insn, *and<mode>3_insn, *ior<mode>3_insn, *xor<mode>3_insn,
3728 *one_cmpl<mode>2_insn, *ashl<mode>3_insn, *ashr<mode>3_insn,
3729 *lshr<mode>3_insn, *trunchiqi2_insn, *truncsihi2_insn,
3730 *truncdisi2_insn, *extendqihi2_insn, *extendqisi2_insn,
3731 *extendhisi2_insn, *extendsidi2_insn, *zero_extendqihi2_insn,
3732 *zero_extendqisi2_insn, *zero_extendsidi2_insn): Fix split condition.
3734 2021-11-17 Marek Polacek <polacek@redhat.com>
3736 PR preprocessor/103026
3737 * doc/invoke.texi: Document -Wbidi-chars.
3739 2021-11-17 Jan Hubicka <hubicka@ucw.cz>
3742 * ipa-modref.c (ipa_merge_modref_summary_after_inlining): Fix clearing
3745 2021-11-16 Jan Hubicka <jh@suse.cz>
3747 * ipa-modref.c (get_modref_function_summary): Declare.
3748 * ipa-modref.h (get_modref_function_summary): New function.
3749 * tree-ssa-dse.c (clear_live_bytes_for_ref): Break out from ...
3750 (clear_bytes_written_by): ... here; also clear memory killed by
3753 2021-11-16 Iain Sandoe <iain@sandoe.co.uk>
3755 * ggc-common.c (gt_pch_save): If we cannot find a suitable
3756 memory segment for save, then error-out, do not try to
3758 (gt_pch_restore): Save the existing line table, and when
3759 the replacement is being read, use that when constructing
3762 2021-11-16 Peter Bergner <bergner@linux.ibm.com>
3765 * config/rs6000/mma.md (*vsx_assemble_pair): Add early-clobber for
3767 (*mma_assemble_acc): Likewise.
3769 2021-11-16 Martin Sebor <msebor@redhat.com>
3771 PR tree-optimization/102960
3772 * gimple-fold.c (get_range_strlen): Take bitmap as an argument rather
3773 than a pointer to it.
3774 (get_range_strlen_tree): Same. Remove bitmap allocation. Use
3776 (get_maxval_strlen): Use an auto_bitmap.
3777 * tree-ssa-strlen.c (get_range_strlen_dynamic): Factor out PHI
3779 (get_range_strlen_phi): ...into this function.
3780 Avoid assuming maximum string length is constant
3781 (printf_strlen_execute): Dump pointer query cache contents when
3782 details are requisted.
3784 2021-11-16 Jason Merrill <jason@redhat.com>
3786 * langhooks.h (struct lang_hooks): Adjust comment.
3787 * print-tree.c (print_node): Also call print_xnode hook for
3790 2021-11-16 Andrew Pinski <apinski@marvell.com>
3792 PR tree-optimization/103218
3793 * match.pd: New pattern for "((type)(a<0)) << SIGNBITOFA".
3795 2021-11-16 Claudiu Zissulescu <claziss@synopsys.com>
3797 * config/arc/arc.md (maddhisi4): Use a single move to accumulator.
3798 (umaddhisi4): Likewise.
3799 (machi): Update pattern.
3802 2021-11-16 Richard Biener <rguenther@suse.de>
3804 PR tree-optimization/102880
3805 * tree-ssa-dce.c (sort_phi_args): New function.
3806 (make_forwarders_with_degenerate_phis): Likewise.
3807 (perform_tree_ssa_dce): Call
3808 make_forwarders_with_degenerate_phis.
3810 2021-11-16 Richard Biener <rguenther@suse.de>
3812 PR tree-optimization/102880
3813 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Push
3814 single_pred (bb1) condition to places that really need it.
3815 (match_simplify_replacement): Likewise.
3816 (value_replacement): Likewise.
3817 (replace_phi_edge_with_variable): Deal with extra edges
3820 2021-11-16 Martin Jambor <mjambor@suse.cz>
3822 * cfgexpand.c (expand_gimple_basic_block): Use build_debug_expr_decl,
3823 add a fixme note about the mode assignment perhaps being unnecessary.
3824 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
3826 (ipa_param_body_adjustments::mark_dead_statements): Likewise.
3827 (ipa_param_body_adjustments::reset_debug_stmts): Likewise.
3828 * tree-inline.c (remap_ssa_name): Likewise.
3829 (tree_function_versioning): Likewise.
3830 * tree-into-ssa.c (rewrite_debug_stmt_uses): Likewise.
3831 * tree-ssa-loop-ivopts.c (remove_unused_ivs): Likewise.
3832 * tree-ssa.c (insert_debug_temp_for_var_def): Likewise.
3834 2021-11-16 Jakub Jelinek <jakub@redhat.com>
3836 PR tree-optimization/103208
3837 * omp-expand.c (expand_omp_build_cond): New function.
3838 (expand_omp_for_init_counts, expand_omp_for_init_vars,
3839 expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Use it.
3841 2021-11-16 Jakub Jelinek <jakub@redhat.com>
3843 PR tree-optimization/102009
3844 * gimple-ssa-warn-access.cc (pass_waccess::check_alloc_size_call):
3845 Punt if any of alloc_size arguments is out of bounds vs. number of
3848 2021-11-16 Roger Sayle <roger@nextmovesoftware.com>
3850 * config/i386/i386.md (*bmi2_rorx<mode3>_1): Make conditional
3851 on !optimize_function_for_size_p.
3852 (*<any_rotate><mode>3_1): Add preferred_for_size attribute.
3853 (define_splits): Conditionalize on !optimize_function_for_size_p.
3854 (*bmi2_rorxsi3_1_zext): Likewise.
3855 (*<any_rotate>si2_1_zext): Add preferred_for_size attribute.
3856 (define_splits): Conditionalize on !optimize_function_for_size_p.
3858 2021-11-16 Jan Hubicka <jh@suse.cz>
3861 * ipa-modref.c (merge_call_side_effects): Fix uninitialized
3864 2021-11-16 Andrew Pinski <apinski@marvell.com>
3866 PR tree-optimization/103245
3867 * match.pd: Combine the abs pattern matching using multiplication.
3868 Adding optional nop_convert too.
3870 2021-11-16 H.J. Lu <hjl.tools@gmail.com>
3872 PR middle-end/103268
3873 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Add a missing
3876 2021-11-15 Siddhesh Poyarekar <siddhesh@gotplt.org>
3878 * gimple-fold.c (gimple_fold_builtin_strncat): Use ranges to
3879 determine if it is safe to transform to strcat.
3880 (gimple_fold_builtin_snprintf): Likewise.
3882 2021-11-15 Siddhesh Poyarekar <siddhesh@gotplt.org>
3884 * gimple-fold.c (known_lower): New function.
3885 (gimple_fold_builtin_strncat_chk,
3886 gimple_fold_builtin_memory_chk, gimple_fold_builtin_stxcpy_chk,
3887 gimple_fold_builtin_stxncpy_chk,
3888 gimple_fold_builtin_snprintf_chk,
3889 gimple_fold_builtin_sprintf_chk): Use it.
3891 2021-11-15 Siddhesh Poyarekar <siddhesh@gotplt.org>
3893 * gimple-fold.c (dump_transformation): New function.
3894 (gimple_fold_builtin_stxcpy_chk,
3895 gimple_fold_builtin_stxncpy_chk): Use it. Simplify to
3896 BUILT_IN_STRNCPY if return value is not used.
3898 2021-11-15 H.J. Lu <hjl.tools@gmail.com>
3900 PR middle-end/103184
3901 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Check optab
3902 before transforming equivalent, but slighly different cases to
3903 their canonical forms.
3905 2021-11-15 Iain Sandoe <iain@sandoe.co.uk>
3908 * config/darwin.h (TARGET_DTORS_FROM_CXA_ATEXIT): New.
3909 * doc/tm.texi: Regenerated.
3910 * doc/tm.texi.in: Add TARGET_DTORS_FROM_CXA_ATEXIT hook.
3911 * ipa.c (cgraph_build_static_cdtor_1): Return the built
3913 (build_cxa_atexit_decl): New.
3914 (build_dso_handle_decl): New.
3915 (build_cxa_dtor_registrations): New.
3916 (compare_cdtor_tu_order): New.
3917 (build_cxa_atexit_fns): New.
3918 (ipa_cdtor_merge): If dtors_from_cxa_atexit is set,
3919 process the DTORs/CTORs accordingly.
3920 (pass_ipa_cdtor_merge::gate): Also run if
3921 dtors_from_cxa_atexit is set.
3922 * target.def (dtors_from_cxa_atexit): New hook.
3924 2021-11-15 Iain Sandoe <iain@sandoe.co.uk>
3926 * config.in: Regenerate.
3927 * configure: Regenerate.
3928 * configure.ac: Test ld64 for -platform-version support.
3930 2021-11-15 Thomas Schwinge <thomas@codesourcery.com>
3932 * diagnostic-spec.h (typedef xint_hash_t)
3933 (typedef xint_hash_map_t): Replace with...
3934 (typedef nowarn_map_t): ... this.
3935 (nowarn_map): Adjust.
3936 * diagnostic-spec.c (nowarn_map, suppress_warning_at): Likewise.
3938 2021-11-15 Thomas Schwinge <thomas@codesourcery.com>
3940 * profile.c (branch_prob): Use 'location_hash' for
3943 2021-11-15 Aldy Hernandez <aldyh@redhat.com>
3945 PR tree-optimization/103207
3946 * value-range.cc (irange::set): Drop overflow.
3948 2021-11-15 Jakub Jelinek <jakub@redhat.com>
3950 * gimplify.c (optimize_target_teams): Only add OMP_CLAUSE_THREAD_LIMIT
3951 to OMP_TARGET_CLAUSES if it isn't there already.
3953 2021-11-15 Aldy Hernandez <aldyh@redhat.com>
3955 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
3956 Remove useless code.
3957 (path_range_query::ssa_defined_in_bb): New.
3958 (path_range_query::ssa_range_in_phi): Avoid fold_range call that
3959 could trigger additional lookups.
3960 Do not use the cache for ARGs defined in this block.
3961 (path_range_query::compute_ranges_in_block): Use ssa_defined_in_bb.
3962 (path_range_query::maybe_register_phi_relation): Same.
3963 (path_range_query::range_of_stmt): Adjust comment.
3964 * gimple-range-path.h (ssa_defined_in_bb): New.
3966 2021-11-15 Aldy Hernandez <aldyh@redhat.com>
3968 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
3969 Default to global range if nothing found.
3971 2021-11-15 Richard Biener <rguenther@suse.de>
3973 PR tree-optimization/103237
3974 * tree-vect-loop.c (vect_is_simple_reduction): Fail for
3975 double reductions with multiple inner loop LC PHI nodes.
3977 2021-11-15 Hongyu Wang <hongyu.wang@intel.com>
3980 * config/i386/i386-expand.c (ix86_expand_atomic_fetch_op_loop):
3981 New expand function.
3982 * config/i386/i386-options.c (ix86_target_string): Add
3983 -mrelax-cmpxchg-loop flag.
3984 (ix86_valid_target_attribute_inner_p): Likewise.
3985 * config/i386/i386-protos.h (ix86_expand_atomic_fetch_op_loop):
3986 New expand function prototype.
3987 * config/i386/i386.opt: Add -mrelax-cmpxchg-loop.
3988 * config/i386/sync.md (atomic_fetch_<logic><mode>): New expander
3990 (atomic_<logic>_fetch<mode>): Likewise.
3991 (atomic_fetch_nand<mode>): Likewise.
3992 (atomic_nand_fetch<mode>): Likewise.
3993 (atomic_fetch_<logic><mode>): New expander for DI,TI modes.
3994 (atomic_<logic>_fetch<mode>): Likewise.
3995 (atomic_fetch_nand<mode>): Likewise.
3996 (atomic_nand_fetch<mode>): Likewise.
3997 * doc/invoke.texi: Document -mrelax-cmpxchg-loop.
3999 2021-11-15 Richard Biener <rguenther@suse.de>
4001 PR tree-optimization/103219
4002 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Use single_exit
4003 to determine the exit for the VN region.
4005 2021-11-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4007 * tree-ssa-loop.c (pass_vectorize): Move to tree-vectorizer.c.
4008 (pass_data_vectorize): Likewise.
4009 (make_pass_vectorize): Likewise.
4010 * tree-vectorizer.c (vectorize_loops): Merge with
4011 pass_vectorize::execute and replace cfun occurences with fun param.
4012 (adjust_simduid_builtins): Add fun param, replace cfun occurences with
4013 fun, and adjust callers approrpiately.
4014 (note_simd_array_uses): Likewise.
4015 (vect_loop_dist_alias_call): Likewise.
4016 (set_uid_loop_bbs): Likewise.
4017 (vect_transform_loops): Likewise.
4018 (try_vectorize_loop_1): Likewise.
4019 (try_vectorize_loop): Likewise.
4021 2021-11-15 Jakub Jelinek <jakub@redhat.com>
4024 * config/i386/sync.md (atomic_bit_test_and_set<mode>,
4025 atomic_bit_test_and_complement<mode>,
4026 atomic_bit_test_and_reset<mode>): Use OPTAB_WIDEN instead of
4029 2021-11-15 Maciej W. Rozycki <macro@embecosm.com>
4031 * config/vax/vax.c (vax_lra_p): New prototype and function.
4032 (TARGET_LRA_P): Wire it.
4033 * config/vax/vax.opt (mlra): New option.
4034 * doc/invoke.texi (Option Summary, VAX Options): Document the
4037 2021-11-14 Jan Hubicka <jh@suse.cz>
4039 * ipa-modref.h (struct modref_summary): Add nondeterministic
4040 and calls_interposable flags.
4041 * ipa-modref.c (modref_summary::modref_summary): Initialize new flags.
4042 (modref_summary::useful_p): Check new flags.
4043 (struct modref_summary_lto): Add nondeterministic and
4044 calls_interposable flags.
4045 (modref_summary_lto::modref_summary_lto): Initialize new flags.
4046 (modref_summary_lto::useful_p): Check new flags.
4047 (modref_summary::dump): Dump new flags.
4048 (modref_summary_lto::dump): Dump new flags.
4049 (ignore_nondeterminism_p): New function.
4050 (merge_call_side_effects): Merge new flags.
4051 (process_fnspec): Likewise.
4052 (analyze_load): Volatile access is nondeterministic.
4053 (analyze_store): Liekwise.
4054 (analyze_stmt): Volatile ASM is nondeterministic.
4055 (analyze_function): Clear new flags.
4056 (modref_summaries::duplicate): Duplicate new flags.
4057 (modref_summaries_lto::duplicate): Duplicate new flags.
4058 (modref_write): Stream new flags.
4059 (read_section): Stream new flags.
4060 (propagate_unknown_call): Update new flags.
4061 (modref_propagate_in_scc): Propagate new flags.
4062 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Check
4064 * tree-ssa-structalias.c (determine_global_memory_access):
4067 2021-11-14 Maciej W. Rozycki <macro@embecosm.com>
4069 * config/vax/vax.h (SET_RATIO): New macro.
4070 * config/vax/vax.md (UNSPEC_SETMEM_FILL): New constant.
4071 (setmemhi): New expander.
4072 (setmemhi1): New insn and splitter.
4073 (*setmemhi1): New insn.
4075 2021-11-14 Jan Hubicka <hubicka@ucw.cz>
4077 * ipa-modref-tree.c (modref_access_node::update_for_kills): New
4079 (modref_access_node::merge_for_kills): Likewise.
4080 (modref_access_node::insert_kill): Likewise.
4081 * ipa-modref-tree.h (modref_access_node::update_for_kills,
4082 modref_access_node::merge_for_kills, modref_access_node::insert_kill):
4084 (modref_access_node::useful_for_kill): New member function.
4085 * ipa-modref.c (modref_summary::useful_p): Release useless kills.
4086 (lto_modref_summary): Add kills.
4087 (modref_summary::dump): Dump kills.
4088 (record_access): Add mdoref_access_node parameter.
4089 (record_access_lto): Likewise.
4090 (merge_call_side_effects): Merge kills.
4091 (analyze_call): Add ALWAYS_EXECUTED param and pass it around.
4092 (struct summary_ptrs): Add always_executed filed.
4093 (analyze_load): Update.
4094 (analyze_store): Update; record kills.
4095 (analyze_stmt): Add always_executed; record kills in clobbers.
4096 (analyze_function): Track always_executed.
4097 (modref_summaries::duplicate): Duplicate kills.
4098 (update_signature): Release kills.
4099 * ipa-modref.h (struct modref_summary): Add kills.
4100 * tree-ssa-alias.c (alias_stats): Add kill stats.
4101 (dump_alias_stats): Dump kill stats.
4102 (store_kills_ref_p): Break out from ...
4103 (stmt_kills_ref_p): Use it; handle modref info based kills.
4105 2021-11-14 Aldy Hernandez <aldyh@redhat.com>
4107 PR tree-optimization/103229
4108 * gimple-range-cache.cc (ssa_global_cache::clear): Do not pass
4109 null value to memset.
4111 2021-11-14 Jan Hubicka <hubicka@ucw.cz>
4113 * ipa-modref-tree.c (modref_access_node::get_call_arg): New member
4115 (modref_access_node::get_ao_ref): Likewise.
4116 * ipa-modref-tree.h (modref_access_node::get_call_arg): Declare.
4117 (modref_access_node::get_ao_ref): Declare.
4118 * tree-ssa-alias.c (modref_may_conflict): Use new accessors.
4119 * tree-ssa-dse.c (dse_optimize_call): Use new accessors.
4121 2021-11-13 Jan Hubicka <jh@suse.cz>
4124 * dbgcnt.def (ipa_attr): New counters.
4125 * ipa-pure-const.c: Include dbgcnt.c
4126 (ipa_make_function_const): Use debug counter.
4127 (ipa_make_function_pure): Likewise.
4128 (propagate_pure_const): Fix bug in my previous change.
4130 2021-11-13 Jan Hubicka <hubicka@ucw.cz>
4132 * ipa-modref-tree.c (modref_access_node::range_info_useful_p):
4133 Offline from ipa-modref-tree.h.
4134 (modref_access_node::dump): Move from ipa-modref.c; make member
4136 * ipa-modref-tree.h (modref_access_node::range_info_useful_p.
4137 modref_access_node::dump): Declare.
4138 * ipa-modref.c (dump_access): Remove.
4139 (dump_records): Update.
4140 (dump_lto_records): Update.
4141 (record_access): Update.
4142 (record_access_lto): Update.
4144 2021-11-13 Jan Hubicka <hubicka@ucw.cz>
4146 * ipa-modref.c (modref_summary::modref_summary): Clear new flags.
4147 (modref_summary::dump): Dump try_dse.
4148 (modref_summary::finalize): Add FUN attribute; compute try-dse.
4149 (analyze_function): Update.
4150 (read_section): Update.
4151 (update_signature): Update.
4152 (pass_ipa_modref::execute): Update.
4153 * ipa-modref.h (struct modref_summary):
4154 * tree-ssa-alias.c (ao_ref_init_from_ptr_and_range): Export.
4155 * tree-ssa-alias.h (ao_ref_init_from_ptr_and_range): Declare.
4156 * tree-ssa-dse.c (dse_optimize_call): New function.
4157 (dse_optimize_stmt): Use it.
4159 2021-11-13 Jan Hubicka <hubicka@ucw.cz>
4161 * ipa-modref-tree.c: Move #if CHECKING_P to proper place.
4163 2021-11-13 Jan Hubicka <jh@suse.cz>
4166 (struct modref_access_node): Move longer member functions to
4168 (modref_ref_node::try_merge_with): Turn into modreef_acces_node member
4170 * ipa-modref-tree.c (modref_access_node::contains): Move here
4171 from ipa-modref-tree.h.
4172 (modref_access_node::update): Likewise.
4173 (modref_access_node::merge): Likewise.
4174 (modref_access_node::closer_pair_p): Likewise.
4175 (modref_access_node::forced_merge): Likewise.
4176 (modref_access_node::update2): Likewise.
4177 (modref_access_node::combined_offsets): Likewise.
4178 (modref_access_node::try_merge_with): Likewise.
4179 (modref_access_node::insert): Likewise.
4181 2021-11-13 Jan Hubicka <jh@suse.cz>
4183 * ipa-modref.c (modref_summary::global_memory_read_p): Remove.
4184 (modref_summary::global_memory_written_p): Remove.
4185 (modref_summary::dump): Dump new flags.
4186 (modref_summary::finalize): New member function.
4187 (analyze_function): Call it.
4188 (read_section): Call it.
4189 (update_signature): Call it.
4190 (pass_ipa_modref::execute): Call it.
4191 * ipa-modref.h (struct modref_summary): Remove
4192 global_memory_read_p and global_memory_written_p.
4193 Add global_memory_read, global_memory_written.
4194 * tree-ssa-structalias.c (determine_global_memory_access):
4197 2021-11-13 Jan Hubicka <jh@suse.cz>
4199 * ipa-fnsummary.c (compute_fn_summary): Use type_attribut_allowed_p
4200 * ipa-param-manipulation.c
4201 (ipa_param_adjustments::type_attribute_allowed_p):
4202 New member function.
4203 (drop_type_attribute_if_params_changed_p): New function.
4204 (build_adjusted_function_type): Use it.
4205 * ipa-param-manipulation.h: Add type_attribute_allowed_p.
4207 2021-11-13 David Malcolm <dmalcolm@redhat.com>
4209 * doc/invoke.texi (Static Analyzer Options): Add
4210 -Wno-analyzer-tainted-allocation-size,
4211 -Wno-analyzer-tainted-divisor, -Wno-analyzer-tainted-offset, and
4212 -Wno-analyzer-tainted-size to list. Add
4213 -Wanalyzer-tainted-allocation-size, -Wanalyzer-tainted-divisor,
4214 -Wanalyzer-tainted-offset, and -Wanalyzer-tainted-size to list
4215 of options effectively enabled by -fanalyzer.
4216 (-Wanalyzer-tainted-allocation-size): New.
4217 (-Wanalyzer-tainted-array-index): Tweak wording; add link to CWE.
4218 (-Wanalyzer-tainted-divisor): New.
4219 (-Wanalyzer-tainted-offset): New.
4220 (-Wanalyzer-tainted-size): New.
4222 2021-11-13 Jan Hubicka <jh@suse.cz>
4224 * attr-fnspec.h (attr_fnspec::arg_eaf_flags): Break out from ...
4225 * gimple.c (gimple_call_arg_flags): ... here.
4226 * ipa-modref.c (analyze_parms): Record flags known from fnspec.
4227 (modref_merge_call_site_flags): Use arg_eaf_flags.
4229 2021-11-13 Aldy Hernandez <aldyh@redhat.com>
4231 PR tree-optimization/103222
4232 * gimple-range-path.cc (path_range_query::compute_ranges_in_phis):
4234 (path_range_query::compute_ranges_in_block): Call
4235 compute_ranges_in_phis.
4236 * gimple-range-path.h (path_range_query::compute_ranges_in_phis):
4239 2021-11-13 Jan Hubicka <jh@suse.cz>
4241 * ipa-fnsummary.c (compute_fn_summary): Do not give up on signature
4242 changes on "fn spec" attribute; give up on varadic types.
4243 * ipa-param-manipulation.c: Include attribs.h.
4244 (build_adjusted_function_type): New parameter ARG_MODIFIED; if it is
4245 true remove "fn spec" attribute.
4246 (ipa_param_adjustments::build_new_function_type): Update.
4247 (ipa_param_body_adjustments::modify_formal_parameters): update.
4248 * ipa-sra.c: Include attribs.h.
4249 (ipa_sra_preliminary_function_checks): Do not check for TYPE_ATTRIBUTES.
4251 2021-11-13 Aldy Hernandez <aldyh@redhat.com>
4253 * gimple-range-path.cc (path_range_query::path_range_query): Merge
4255 (path_range_query::import_p): Move from header file.
4256 (path_range_query::~path_range_query): Adjust for combined ctors.
4257 * gimple-range-path.h: Merge ctors.
4258 (path_range_query::import_p): Move to .cc file.
4260 2021-11-13 Jan Hubicka <jh@suse.cz>
4262 PR tree-optimization/103182
4263 * ipa-modref.c (callee_to_caller_flags): Fix merging of flags.
4264 (modref_eaf_analysis::analyze_ssa_name): Fix merging of flags.
4266 2021-11-12 Stafford Horne <shorne@gmail.com>
4268 * config/or1k/or1k-protos.h (or1k_profile_hook): New function.
4269 * config/or1k/or1k.h (PROFILE_HOOK): Change macro to reference
4270 new function or1k_profile_hook.
4271 * config/or1k/or1k.c (struct machine_function): Add new field
4272 set_mcount_arg_insn.
4273 (or1k_profile_hook): New function.
4274 (or1k_init_pic_reg): Update to inject pic rtx after _mcount arg
4276 (or1k_frame_pointer_required): Frame pointer no longer needed
4279 2021-11-12 Jan Hubicka <jh@suse.cz>
4281 PR tree-optimization/103209
4282 * tree-ssa-structalias.c (find_func_aliases_for_call): Fix
4283 use of handle_rhs_call
4285 2021-11-12 Aldy Hernandez <aldyh@redhat.com>
4287 PR tree-optimization/103202
4288 * gimple-range-path.cc
4289 (path_range_query::compute_ranges_in_block): Solve PHI imports first.
4291 2021-11-12 Jan Hubicka <jh@suse.cz>
4293 * ipa-pure-const.c (propagate_pure_const): Remove redundant check;
4294 fix call of ipa_make_function_const and ipa_make_function_pure.
4296 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4298 * config/aarch64/aarch64.c (aarch64_vector_op::n_advsimd_ops): Delete.
4299 (aarch64_vector_op::m_seen_loads): Likewise.
4300 (aarch64_vector_costs::aarch64_vector_costs): Don't push to
4302 (aarch64_vector_op::count_ops): Remove vectype and factor parameters.
4303 Remove code that tries to predict different vec_flags from the
4305 (aarch64_vector_costs::add_stmt_cost): Update accordingly.
4306 Remove m_advsimd_ops handling.
4308 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4310 * config/aarch64/aarch64.c (aarch64_vector_costs::m_saw_sve_only_op)
4311 (aarch64_sve_only_stmt_p): Delete.
4312 (aarch64_vector_costs::prefer_unrolled_loop): New function,
4313 extracted from adjust_body_cost.
4314 (aarch64_vector_costs::better_main_loop_than_p): New function,
4315 using heuristics extracted from adjust_body_cost and
4316 adjust_body_cost_sve.
4317 (aarch64_vector_costs::adjust_body_cost_sve): Remove
4318 advsimd_cycles_per_iter and could_use_advsimd parameters.
4319 Update after changes above.
4320 (aarch64_vector_costs::adjust_body_cost): Update after changes above.
4322 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4324 * config/aarch64/aarch64.c (aarch64_vec_op_count::m_vf_factor):
4325 New member variable.
4326 (aarch64_vec_op_count::aarch64_vec_op_count): Add a parameter for it.
4327 (aarch64_vec_op_count::vf_factor): New function.
4328 (aarch64_vector_costs::aarch64_vector_costs): When costing for
4329 neoverse-512tvb, pass a vf_factor of 2 for the Neoverse V1 version
4331 (aarch64_vector_costs::adjust_body_cost): Read the vf factor
4332 instead of hard-coding 2.
4334 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4336 * config/aarch64/aarch64.c
4337 (aarch64_vec_op_count::rename_cycles_per_iter): New function.
4338 (aarch64_vec_op_count::min_nonpred_cycles_per_iter): Likewise.
4339 (aarch64_vec_op_count::min_pred_cycles_per_iter): Likewise.
4340 (aarch64_vec_op_count::min_cycles_per_iter): Likewise.
4341 (aarch64_vec_op_count::dump): Move earlier in file. Dump the
4342 above properties too.
4343 (aarch64_estimate_min_cycles_per_iter): Delete.
4344 (adjust_body_cost): Use aarch64_vec_op_count::min_cycles_per_iter
4345 instead of aarch64_estimate_min_cycles_per_iter. Rely on the dump
4346 routine to print CPI estimates.
4347 (adjust_body_cost_sve): Likewise. Use the other functions above
4348 instead of doing the work inline.
4350 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4352 * config/aarch64/aarch64.c (aarch64_vec_op_count): Allow default
4354 (aarch64_vec_op_count::base_issue_info): Remove handling of null
4356 (aarch64_vec_op_count::simd_issue_info): Likewise.
4357 (aarch64_vec_op_count::sve_issue_info): Likewise.
4358 (aarch64_vector_costs::m_ops): Turn into a vector.
4359 (aarch64_vector_costs::m_advsimd_ops): Likewise.
4360 (aarch64_vector_costs::aarch64_vector_costs): Add entries to
4361 the vectors based on aarch64_tune_params.
4362 (aarch64_vector_costs::analyze_loop_vinfo): Update the pred_ops
4363 of all entries in m_ops.
4364 (aarch64_vector_costs::add_stmt_cost): Call count_ops for all
4366 (aarch64_estimate_min_cycles_per_iter): Remove issue_info
4367 parameter and get the information from the ops instead.
4368 (aarch64_vector_costs::adjust_body_cost_sve): Take a
4369 aarch64_vec_issue_info instead of a aarch64_vec_op_count.
4370 (aarch64_vector_costs::adjust_body_cost): Update call accordingly.
4371 Exit earlier if m_ops is empty for either cost structure.
4373 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4375 * config/aarch64/aarch64.c (aarch64_vector_costs::m_scalar_ops)
4376 (aarch64_vector_costs::m_sve_ops): Replace with...
4377 (aarch64_vector_costs::m_ops): ...this.
4378 (aarch64_vector_costs::analyze_loop_vinfo): Update accordingly.
4379 (aarch64_vector_costs::adjust_body_cost_sve): Likewise.
4380 (aarch64_vector_costs::aarch64_vector_costs): Likewise.
4381 Initialize m_vec_flags here rather than in add_stmt_cost.
4382 (aarch64_vector_costs::count_ops): Test for scalar reductions too.
4383 Allow vectype to be null.
4384 (aarch64_vector_costs::add_stmt_cost): Call count_ops for scalar
4385 code too. Don't require vectype to be nonnull.
4386 (aarch64_vector_costs::adjust_body_cost): Take the loop_vec_info
4387 and scalar costs as parameters. Use the scalar costs to determine
4388 the cycles per iteration of the scalar loop, then multiply it
4389 by the estimated VF.
4390 (aarch64_vector_costs::finish_cost): Update call accordingly.
4392 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4394 * config/aarch64/aarch64.c (aarch64_dr_type): New function.
4395 (aarch64_vector_costs::count_ops): Use it rather than the
4396 vectype to determine floatness.
4398 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4400 * config/aarch64/aarch64.c (aarch64_sve_in_loop_reduction_latency):
4401 Remove vectype parameter and get floatness from the type of the
4403 (arch64_in_loop_reduction_latency): Likewise.
4404 (aarch64_detect_vector_stmt_subtype): Update caller.
4405 (aarch64_vector_costs::count_ops): Likewise.
4407 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4409 * config/aarch64/aarch64.c (aarch64_sve_op_count): Fold into...
4410 (aarch64_vec_op_count): ...this. Add a constructor.
4411 (aarch64_vec_op_count::vec_flags): New function.
4412 (aarch64_vec_op_count::base_issue_info): Likewise.
4413 (aarch64_vec_op_count::simd_issue_info): Likewise.
4414 (aarch64_vec_op_count::sve_issue_info): Likewise.
4415 (aarch64_vec_op_count::m_issue_info): New member variable.
4416 (aarch64_vec_op_count::m_vec_flags): Likewise.
4417 (aarch64_vector_costs): Add a constructor.
4418 (aarch64_vector_costs::m_sve_ops): Change type to aarch64_vec_op_count.
4419 (aarch64_vector_costs::aarch64_vector_costs): New function.
4420 Initialize m_scalar_ops, m_advsimd_ops and m_sve_ops.
4421 (aarch64_vector_costs::count_ops): Remove vec_flags and
4422 issue_info parameters, using the new aarch64_vec_op_count
4424 (aarch64_vector_costs::add_stmt_cost): Update call accordingly.
4425 (aarch64_sve_op_count::dump): Fold into...
4426 (aarch64_vec_op_count::dump): ..here.
4428 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
4430 * config/aarch64/aarch64.c: Include tree-dfa.h.
4431 (aarch64_check_consecutive_mems): New function that takes MEM_EXPR
4432 and MEM_OFFSET into account.
4433 (aarch64_swap_ldrstr_operands): Use it.
4434 (aarch64_operands_ok_for_ldpstp): Likewise. Check that the
4435 address of the second memory doesn't depend on the result of
4438 2021-11-12 Jan Hubicka <hubicka@ucw.cz>
4440 * ipa-pure-const.c (ipa_make_function_pure): Fix exit condition.
4442 2021-11-12 Jan Hubicka <jh@suse.cz>
4444 PR tree-optimization/103175
4445 * ipa-modref.c (modref_lattice::merge): Add sanity check.
4446 (callee_to_caller_flags): Make flags adjustment sane.
4447 (modref_eaf_analysis::analyze_ssa_name): Likewise.
4449 2021-11-12 Jan Hubicka <jh@suse.cz>
4452 * ipa-modref.c (analyze_function, modref_propagate_in_scc): Do
4453 not mark pure/const function if there are side-effects.
4455 2021-11-12 Chung-Lin Tang <cltang@codesourcery.com>
4457 * tree.h (OMP_CLAUSE_MAP_RUNTIME_IMPLICIT_P): New access macro for
4458 'implicit' bit, using 'base.deprecated_flag' field of tree_node.
4459 * tree-pretty-print.c (dump_omp_clause): Add support for printing
4460 implicit attribute in tree dumping.
4461 * gimplify.c (gimplify_adjust_omp_clauses_1):
4462 Set OMP_CLAUSE_MAP_RUNTIME_IMPLICIT_P to 1 if map clause is implicitly
4464 (gimplify_adjust_omp_clauses): Adjust place of adding implicitly created
4465 clauses, from simple append, to starting of list, after non-map clauses.
4466 * omp-low.c (lower_omp_target): Add GOMP_MAP_IMPLICIT bits into kind
4467 values passed to libgomp for implicit maps.
4469 2021-11-12 Jakub Jelinek <jakub@redhat.com>
4471 * omp-builtins.def (BUILT_IN_GOMP_TEAMS): Remove.
4472 (BUILT_IN_GOMP_TEAMS4): New.
4473 * builtin-types.def (BT_FN_VOID_UINT_UINT): Remove.
4474 (BT_FN_BOOL_UINT_UINT_UINT_BOOL): New.
4475 * omp-low.c (lower_omp_teams): Use GOMP_teams4 instead of
4476 GOMP_teams, pass to it also num_teams lower-bound expression
4477 or a dup of upper-bound if it is missing and a flag whether
4478 it is the first call or not.
4480 2021-11-12 Martin Liska <mliska@suse.cz>
4482 PR tree-optimization/102497
4483 * gimple-predicate-analysis.cc (add_pred): Remove unused
4486 2021-11-12 Richard Biener <rguenther@suse.de>
4488 PR tree-optimization/103204
4489 * tree-ssa-sccvn.c (valueize_refs_1): Re-valueize the
4490 top operand after folding in an address.
4492 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
4494 * gimple-range-path.cc (path_range_query::path_range_query): New
4495 ctor without a ranger.
4496 (path_range_query::~path_range_query): Free ranger if necessary.
4497 (path_range_query::range_on_path_entry): Adjust m_ranger for pointer.
4498 (path_range_query::ssa_range_in_phi): Same.
4499 (path_range_query::compute_ranges_in_block): Same.
4500 (path_range_query::compute_imports): Same.
4501 (path_range_query::compute_ranges): Same.
4502 (path_range_query::range_of_stmt): Same.
4503 (path_range_query::compute_outgoing_relations): Same.
4504 * gimple-range-path.h (class path_range_query): New ctor.
4505 * tree-ssa-loop-ch.c (ch_base::copy_headers): Remove gimple_ranger
4506 as path_range_query allocates one.
4507 * tree-ssa-threadbackward.c (class back_threader): Remove m_ranger.
4508 (back_threader::~back_threader): Same.
4510 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
4512 * tree-ssa-threadbackward.c
4513 (back_threader_profitability::profitable_path_p): Remove loop
4514 crossing restriction.
4516 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
4518 * ipa-modref.c (analyze_function): Do pure/const discovery, return
4520 (pass_modref::execute): If pure/const is discovered fixup cfg.
4521 (ignore_edge): Do not ignore pure/const edges.
4522 (modref_propagate_in_scc): Do pure/const discovery, return true if
4523 cdtor was promoted pure/const.
4524 (pass_ipa_modref::execute): If needed remove unreachable functions.
4525 * ipa-pure-const.c (warn_function_noreturn): Fix whitespace.
4526 (warn_function_cold): Likewise.
4527 (skip_function_for_local_pure_const): Move earlier.
4528 (ipa_make_function_const): Break out from ...
4529 (ipa_make_function_pure): Break out from ...
4530 (propagate_pure_const): ... here.
4531 (pass_local_pure_const::execute): Use it.
4532 * ipa-utils.h (ipa_make_function_const): Declare.
4533 (ipa_make_function_pure): Declare.
4534 * passes.def: Move early modref after pure-const.
4536 2021-11-11 David Malcolm <dmalcolm@redhat.com>
4539 * diagnostic-show-locus.c (def_policy): Use def_tabstop.
4541 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4543 * config/aarch64/aarch64-builtins.c (TYPES_COMBINE): Delete.
4544 (TYPES_COMBINEP): Delete.
4545 * config/aarch64/aarch64-simd-builtins.def: Declare type-
4546 qualified builtins for vcombine_* intrinsics.
4547 * config/aarch64/arm_neon.h (vcombine_s8): Remove unnecessary
4549 (vcombine_s16): Likewise.
4550 (vcombine_s32): Likewise.
4551 (vcombine_f32): Likewise.
4552 (vcombine_u8): Use type-qualified builtin and remove casts.
4553 (vcombine_u16): Likewise.
4554 (vcombine_u32): Likewise.
4555 (vcombine_u64): Likewise.
4556 (vcombine_p8): Likewise.
4557 (vcombine_p16): Likewise.
4558 (vcombine_p64): Likewise.
4559 (vcombine_bf16): Remove unnecessary cast.
4560 * config/aarch64/iterators.md (VD_I): New mode iterator.
4561 (VDC_P): New mode iterator.
4563 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4565 * config/aarch64/aarch64-builtins.c (TYPES_LOAD1_U): Define.
4566 (TYPES_LOAD1_P): Define.
4567 (TYPES_STORE1_U): Define.
4568 (TYPES_STORE1P): Rename to...
4569 (TYPES_STORE1_P): This.
4570 (get_mem_type_for_load_store): Add unsigned and poly types.
4571 (aarch64_general_gimple_fold_builtin): Add unsigned and poly
4572 type-qualified builtin declarations.
4573 * config/aarch64/aarch64-simd-builtins.def: Declare type-
4574 qualified builtins for LD1/ST1.
4575 * config/aarch64/arm_neon.h (vld1_p8): Use type-qualified
4576 builtin and remove cast.
4577 (vld1_p16): Likewise.
4578 (vld1_u8): Likewise.
4579 (vld1_u16): Likewise.
4580 (vld1_u32): Likewise.
4581 (vld1q_p8): Likewise.
4582 (vld1q_p16): Likewise.
4583 (vld1q_p64): Likewise.
4584 (vld1q_u8): Likewise.
4585 (vld1q_u16): Likewise.
4586 (vld1q_u32): Likewise.
4587 (vld1q_u64): Likewise.
4588 (vst1_p8): Likewise.
4589 (vst1_p16): Likewise.
4590 (vst1_u8): Likewise.
4591 (vst1_u16): Likewise.
4592 (vst1_u32): Likewise.
4593 (vst1q_p8): Likewise.
4594 (vst1q_p16): Likewise.
4595 (vst1q_p64): Likewise.
4596 (vst1q_u8): Likewise.
4597 (vst1q_u16): Likewise.
4598 (vst1q_u32): Likewise.
4599 (vst1q_u64): Likewise.
4600 * config/aarch64/iterators.md (VALLP_NO_DI): New iterator.
4602 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4604 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
4605 builtins for vector reduction.
4606 * config/aarch64/arm_neon.h (vaddv_u8): Use type-qualified
4607 builtin and remove casts.
4608 (vaddv_u16): Likewise.
4609 (vaddv_u32): Likewise.
4610 (vaddvq_u8): Likewise.
4611 (vaddvq_u16): Likewise.
4612 (vaddvq_u32): Likewise.
4613 (vaddvq_u64): Likewise.
4615 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4617 * config/aarch64/aarch64-simd-builtins.def:
4618 * config/aarch64/arm_neon.h (vpaddq_u8): Use type-qualified
4619 builtin and remove casts.
4620 (vpaddq_u16): Likewise.
4621 (vpaddq_u32): Likewise.
4622 (vpaddq_u64): Likewise.
4623 (vpadd_u8): Likewise.
4624 (vpadd_u16): Likewise.
4625 (vpadd_u32): Likewise.
4626 (vpaddd_u64): Likewise.
4628 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4630 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
4631 builtins for [r]subhn[2].
4632 * config/aarch64/arm_neon.h (vsubhn_s16): Remove unnecessary
4634 (vsubhn_s32): Likewise.
4635 (vsubhn_s64): Likewise.
4636 (vsubhn_u16): Use type-qualified builtin and remove casts.
4637 (vsubhn_u32): Likewise.
4638 (vsubhn_u64): Likewise.
4639 (vrsubhn_s16): Remove unnecessary cast.
4640 (vrsubhn_s32): Likewise.
4641 (vrsubhn_s64): Likewise.
4642 (vrsubhn_u16): Use type-qualified builtin and remove casts.
4643 (vrsubhn_u32): Likewise.
4644 (vrsubhn_u64): Likewise.
4645 (vrsubhn_high_s16): Remove unnecessary cast.
4646 (vrsubhn_high_s32): Likewise.
4647 (vrsubhn_high_s64): Likewise.
4648 (vrsubhn_high_u16): Use type-qualified builtin and remove
4650 (vrsubhn_high_u32): Likewise.
4651 (vrsubhn_high_u64): Likewise.
4652 (vsubhn_high_s16): Remove unnecessary cast.
4653 (vsubhn_high_s32): Likewise.
4654 (vsubhn_high_s64): Likewise.
4655 (vsubhn_high_u16): Use type-qualified builtin and remove
4657 (vsubhn_high_u32): Likewise.
4658 (vsubhn_high_u64): Likewise.
4660 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4662 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
4663 builtins for [r]addhn[2].
4664 * config/aarch64/arm_neon.h (vaddhn_s16): Remove unnecessary
4666 (vaddhn_s32): Likewise.
4667 (vaddhn_s64): Likewise.
4668 (vaddhn_u16): Use type-qualified builtin and remove casts.
4669 (vaddhn_u32): Likewise.
4670 (vaddhn_u64): Likewise.
4671 (vraddhn_s16): Remove unnecessary cast.
4672 (vraddhn_s32): Likewise.
4673 (vraddhn_s64): Likewise.
4674 (vraddhn_u16): Use type-qualified builtin and remove casts.
4675 (vraddhn_u32): Likewise.
4676 (vraddhn_u64): Likewise.
4677 (vaddhn_high_s16): Remove unnecessary cast.
4678 (vaddhn_high_s32): Likewise.
4679 (vaddhn_high_s64): Likewise.
4680 (vaddhn_high_u16): Use type-qualified builtin and remove
4682 (vaddhn_high_u32): Likewise.
4683 (vaddhn_high_u64): Likewise.
4684 (vraddhn_high_s16): Remove unnecessary cast.
4685 (vraddhn_high_s32): Likewise.
4686 (vraddhn_high_s64): Likewise.
4687 (vraddhn_high_u16): Use type-qualified builtin and remove
4689 (vraddhn_high_u32): Likewise.
4690 (vraddhn_high_u64): Likewise.
4692 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4694 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
4695 qualifiers in generator macros for uhsub builtins.
4696 * config/aarch64/arm_neon.h (vhsub_s8): Remove unnecessary
4698 (vhsub_s16): Likewise.
4699 (vhsub_s32): Likewise.
4700 (vhsub_u8): Use type-qualified builtin and remove casts.
4701 (vhsub_u16): Likewise.
4702 (vhsub_u32): Likewise.
4703 (vhsubq_s8): Remove unnecessary cast.
4704 (vhsubq_s16): Likewise.
4705 (vhsubq_s32): Likewise.
4706 (vhsubq_u8): Use type-qualified builtin and remove casts.
4707 (vhsubq_u16): Likewise.
4708 (vhsubq_u32): Likewise.
4710 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4712 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
4713 qualifiers in generator macros for u[r]hadd builtins.
4714 * config/aarch64/arm_neon.h (vhadd_s8): Remove unnecessary
4716 (vhadd_s16): Likewise.
4717 (vhadd_s32): Likewise.
4718 (vhadd_u8): Use type-qualified builtin and remove casts.
4719 (vhadd_u16): Likewise.
4720 (vhadd_u32): Likewise.
4721 (vhaddq_s8): Remove unnecessary cast.
4722 (vhaddq_s16): Likewise.
4723 (vhaddq_s32): Likewise.
4724 (vhaddq_u8): Use type-qualified builtin and remove casts.
4725 (vhaddq_u16): Likewise.
4726 (vhaddq_u32): Likewise.
4727 (vrhadd_s8): Remove unnecessary cast.
4728 (vrhadd_s16): Likewise.
4729 (vrhadd_s32): Likewise.
4730 (vrhadd_u8): Use type-qualified builtin and remove casts.
4731 (vrhadd_u16): Likewise.
4732 (vrhadd_u32): Likewise.
4733 (vrhaddq_s8): Remove unnecessary cast.
4734 (vrhaddq_s16): Likewise.
4735 (vrhaddq_s32): Likewise.
4736 (vrhaddq_u8): Use type-wualified builtin and remove casts.
4737 (vrhaddq_u16): Likewise.
4738 (vrhaddq_u32): Likewise.
4740 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4742 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
4743 qualifiers in generator macros for usub[lw][2] builtins.
4744 * config/aarch64/arm_neon.h (vsubl_s8): Remove unnecessary
4746 (vsubl_s16): Likewise.
4747 (vsubl_s32): Likewise.
4748 (vsubl_u8): Use type-qualified builtin and remove casts.
4749 (vsubl_u16): Likewise.
4750 (vsubl_u32): Likewise.
4751 (vsubl_high_s8): Remove unnecessary cast.
4752 (vsubl_high_s16): Likewise.
4753 (vsubl_high_s32): Likewise.
4754 (vsubl_high_u8): Use type-qualified builtin and remove casts.
4755 (vsubl_high_u16): Likewise.
4756 (vsubl_high_u32): Likewise.
4757 (vsubw_s8): Remove unnecessary casts.
4758 (vsubw_s16): Likewise.
4759 (vsubw_s32): Likewise.
4760 (vsubw_u8): Use type-qualified builtin and remove casts.
4761 (vsubw_u16): Likewise.
4762 (vsubw_u32): Likewise.
4763 (vsubw_high_s8): Remove unnecessary cast.
4764 (vsubw_high_s16): Likewise.
4765 (vsubw_high_s32): Likewise.
4766 (vsubw_high_u8): Use type-qualified builtin and remove casts.
4767 (vsubw_high_u16): Likewise.
4768 (vsubw_high_u32): Likewise.
4770 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4772 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
4773 qualifiers in generator macros for uadd[lw][2] builtins.
4774 * config/aarch64/arm_neon.h (vaddl_s8): Remove unnecessary
4776 (vaddl_s16): Likewise.
4777 (vaddl_s32): Likewise.
4778 (vaddl_u8): Use type-qualified builtin and remove casts.
4779 (vaddl_u16): Likewise.
4780 (vaddl_u32): Likewise.
4781 (vaddl_high_s8): Remove unnecessary cast.
4782 (vaddl_high_s16): Likewise.
4783 (vaddl_high_s32): Likewise.
4784 (vaddl_high_u8): Use type-qualified builtin and remove casts.
4785 (vaddl_high_u16): Likewise.
4786 (vaddl_high_u32): Likewise.
4787 (vaddw_s8): Remove unnecessary cast.
4788 (vaddw_s16): Likewise.
4789 (vaddw_s32): Likewise.
4790 (vaddw_u8): Use type-qualified builtin and remove casts.
4791 (vaddw_u16): Likewise.
4792 (vaddw_u32): Likewise.
4793 (vaddw_high_s8): Remove unnecessary cast.
4794 (vaddw_high_s16): Likewise.
4795 (vaddw_high_s32): Likewise.
4796 (vaddw_high_u8): Use type-qualified builtin and remove casts.
4797 (vaddw_high_u16): Likewise.
4798 (vaddw_high_u32): Likewise.
4800 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4802 * config/aarch64/aarch64-simd-builtins.def: Declare type-
4803 qualified builtins for [R]SHRN[2].
4804 * config/aarch64/arm_neon.h (vshrn_n_u16): Use type-qualified
4805 builtin and remove casts.
4806 (vshrn_n_u32): Likewise.
4807 (vshrn_n_u64): Likewise.
4808 (vrshrn_high_n_u16): Likewise.
4809 (vrshrn_high_n_u32): Likewise.
4810 (vrshrn_high_n_u64): Likewise.
4811 (vrshrn_n_u16): Likewise.
4812 (vrshrn_n_u32): Likewise.
4813 (vrshrn_n_u64): Likewise.
4814 (vshrn_high_n_u16): Likewise.
4815 (vshrn_high_n_u32): Likewise.
4816 (vshrn_high_n_u64): Likewise.
4818 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4820 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
4821 type-qualified builtins for XTN[2].
4822 * config/aarch64/arm_neon.h (vmovn_high_u16): Use type-
4823 qualified builtin and remove casts.
4824 (vmovn_high_u32): Likewise.
4825 (vmovn_high_u64): Likewise.
4826 (vmovn_u16): Likewise.
4827 (vmovn_u32): Likewise.
4828 (vmovn_u64): Likewise.
4830 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4832 * config/aarch64/aarch64-simd-builtins.def: Use poly type
4833 qualifier in builtin generator macros.
4834 * config/aarch64/arm_neon.h (vmul_p8): Use type-qualified
4835 builtin and remove casts.
4836 (vmulq_p8): Likewise.
4837 (vmull_high_p8): Likewise.
4838 (vmull_p8): Likewise.
4840 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
4842 * config/aarch64/aarch64-simd-builtins.def: Declare type-
4843 qualified builtin generators for unsigned MLA/MLS intrinsics.
4844 * config/aarch64/arm_neon.h (vmla_n_u16): Use type-qualified
4846 (vmla_n_u32): Likewise.
4847 (vmla_u8): Likewise.
4848 (vmla_u16): Likewise.
4849 (vmla_u32): Likewise.
4850 (vmlaq_n_u16): Likewise.
4851 (vmlaq_n_u32): Likewise.
4852 (vmlaq_u8): Likewise.
4853 (vmlaq_u16): Likewise.
4854 (vmlaq_u32): Likewise.
4855 (vmls_n_u16): Likewise.
4856 (vmls_n_u32): Likewise.
4857 (vmls_u8): Likewise.
4858 (vmls_u16): Likewise.
4859 (vmls_u32): Likewise.
4860 (vmlsq_n_u16): Likewise.
4861 (vmlsq_n_u32): Likewise.
4862 (vmlsq_u8): Likewise.
4863 (vmlsq_u16): Likewise.
4864 (vmlsq_u32): Likewise.
4866 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
4868 * ipa-modref.c (modref_summary::useful_p): Check also for side-effects
4869 with looping const/pure.
4870 (modref_summary_lto::useful_p): Likewise.
4871 (merge_call_side_effects): Merge side effects before early exit
4873 (process_fnspec): Also handle pure functions.
4874 (analyze_call): Do not early exit on looping pure const.
4875 (propagate_unknown_call): Also handle nontrivial SCC as side-effect.
4876 (modref_propagate_in_scc): Update.
4878 2021-11-11 Richard Biener <rguenther@suse.de>
4880 PR tree-optimization/103190
4881 * tree-ssa-reassoc.c (insert_stmt_after): Only assert on asm goto.
4883 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
4885 * gimple-range-path.cc (path_range_query::add_copies_to_imports):
4887 (path_range_query::compute_imports): ...this. Adapt it so it can
4888 be passed the imports bitmap instead of working on m_imports.
4889 (path_range_query::compute_ranges): Call compute_imports in all
4890 cases unless an imports bitmap is passed.
4891 * gimple-range-path.h (path_range_query::compute_imports): New.
4892 (path_range_query::add_copies_to_imports): Remove.
4893 * tree-ssa-threadbackward.c (back_threader::resolve_def): Remove.
4894 (back_threader::find_paths_to_names): Inline resolve_def.
4895 (back_threader::find_paths): Call compute_imports.
4896 (back_threader::resolve_phi): Adjust comment.
4898 2021-11-11 Richard Biener <rguenther@suse.de>
4900 PR tree-optimization/103188
4901 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
4902 Remove query parameter, split out check for size
4904 (ch_base::m_ranger, cb_base::m_query): Remove.
4905 (ch_base::copy_headers): Split processing loop into
4906 analysis around which we allocate and use ranger and
4907 transform where we do not.
4908 (pass_ch::execute): Do not allocate/free ranger here.
4909 (pass_ch_vect::execute): Likewise.
4911 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
4913 * ipa-pure-const.c (propagate_pure_const): Self recursion is
4916 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
4918 * cgraph.c (set_noreturn_flag_1): New function.
4919 (cgraph_node::set_noreturn_flag): New member function
4920 * cgraph.h (cgraph_node::set_noreturn_flags): Declare.
4921 * ipa-pure-const.c (pass_local_pure_const::execute): Use it.
4923 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
4924 Richard Biener <rguenther@suse.de>
4926 * tree-ssa-loop-ch.c (entry_loop_condition_is_static): Resolve
4927 statically to the edge remaining in the loop.
4929 2021-11-11 Richard Biener <rguenther@suse.de>
4931 PR middle-end/103181
4932 * tree-eh.c (operation_could_trap_helper_p): Properly
4933 check vector constants for a zero element for integer
4934 division. Separate floating point and integer division code.
4936 2021-11-11 Jakub Jelinek <jakub@redhat.com>
4939 * dwarf2out.c (field_byte_offset): Do the PCC_BITFIELD_TYPE_MATTERS
4940 handling only for DECL_BIT_FIELD_TYPE decls.
4942 2021-11-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4945 * config/aarch64/aarch64.c (aarch64_process_target_attr): Check if
4946 token is arch extension without leading '+' and emit appropriate
4947 diagnostic for the same.
4949 2021-11-11 Jakub Jelinek <jakub@redhat.com>
4951 * tree.h (OMP_CLAUSE_NUM_TEAMS_EXPR): Rename to ...
4952 (OMP_CLAUSE_NUM_TEAMS_UPPER_EXPR): ... this.
4953 (OMP_CLAUSE_NUM_TEAMS_LOWER_EXPR): Define.
4954 * tree.c (omp_clause_num_ops): Increase num ops for
4955 OMP_CLAUSE_NUM_TEAMS to 2.
4956 * tree-pretty-print.c (dump_omp_clause): Print optional lower bound
4957 for OMP_CLAUSE_NUM_TEAMS.
4958 * gimplify.c (gimplify_scan_omp_clauses): Gimplify
4959 OMP_CLAUSE_NUM_TEAMS_LOWER_EXPR if non-NULL.
4960 (optimize_target_teams): Use OMP_CLAUSE_NUM_TEAMS_UPPER_EXPR instead
4961 of OMP_CLAUSE_NUM_TEAMS_EXPR. Handle OMP_CLAUSE_NUM_TEAMS_LOWER_EXPR.
4962 * omp-low.c (lower_omp_teams): Use OMP_CLAUSE_NUM_TEAMS_UPPER_EXPR
4963 instead of OMP_CLAUSE_NUM_TEAMS_EXPR.
4964 * omp-expand.c (expand_teams_call, get_target_arguments): Likewise.
4966 2021-11-11 Richard Biener <rguenther@suse.de>
4968 * cfganal.c (find_pdom): Remove.
4969 (control_dependences::find_control_dependence): Remove
4970 special-casing of entry block, call get_immediate_dominator
4972 * gimple-predicate-analysis.cc (find_pdom): Remove.
4973 (find_dom): Likewise.
4974 (find_control_equiv_block): Call get_immediate_dominator
4976 (compute_control_dep_chain): Likewise.
4977 (predicate::init_from_phi_def): Likewise.
4979 2021-11-11 Richard Biener <rguenther@suse.de>
4981 * cfganal.h (control_dependences::control_dependence_map):
4983 (control_dependences::m_bitmaps): New.
4984 * cfganal.c (control_dependences::set_control_dependence_map_bit):
4986 (control_dependences::clear_control_dependence_bitmap):
4988 (control_dependences::find_control_dependence): Do not
4989 find_edge for the abnormal edge test.
4990 (control_dependences::control_dependences): Instead do not
4991 add abnormal edges to the edge list. Adjust.
4992 (control_dependences::~control_dependences): Likewise.
4993 (control_dependences::get_edges_dependent_on): Likewise.
4994 * function-tests.c: Include bitmap.h.
4996 2021-11-11 Kewen Lin <linkw@linux.ibm.com>
4998 * doc/invoke.texi: Change references to "future cpu" to "power10",
4999 "-mcpu=future" to "-mcpu=power10". Adjust words for float128.
5001 2021-11-11 Cui,Lili <lili.cui@intel.com>
5003 * config/i386/i386-options.c (m_CORE_AVX2): Remove Alderlake
5005 (processor_cost_table): Use alderlake_cost for Alderlake.
5006 * config/i386/i386.c (ix86_sched_init_global): Handle Alderlake.
5007 * config/i386/x86-tune-costs.h (struct processor_costs): Add alderlake
5009 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Alderlake
5011 (ix86_adjust_cost): Handle Alderlake.
5012 * config/i386/x86-tune.def (X86_TUNE_SCHEDULE): Enable for Alderlake.
5013 (X86_TUNE_PARTIAL_REG_DEPENDENCY): Likewise.
5014 (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY): Likewise.
5015 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): Likewise.
5016 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
5017 (X86_TUNE_MEMORY_MISMATCH_STALL): Likewise.
5018 (X86_TUNE_USE_LEAVE): Likewise.
5019 (X86_TUNE_PUSH_MEMORY): Likewise.
5020 (X86_TUNE_USE_INCDEC): Likewise.
5021 (X86_TUNE_INTEGER_DFMODE_MOVES): Likewise.
5022 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
5023 (X86_TUNE_USE_SAHF): Likewise.
5024 (X86_TUNE_USE_BT): Likewise.
5025 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
5026 (X86_TUNE_ONE_IF_CONV_INSN): Likewise.
5027 (X86_TUNE_AVOID_MFENCE): Likewise.
5028 (X86_TUNE_USE_SIMODE_FIOP): Likewise.
5029 (X86_TUNE_EXT_80387_CONSTANTS): Likewise.
5030 (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL): Likewise.
5031 (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Likewise.
5032 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
5033 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
5034 (X86_TUNE_AVOID_4BYTE_PREFIXES): Likewise.
5035 (X86_TUNE_USE_GATHER): Disable for Alderlake.
5036 (X86_TUNE_AVX256_MOVE_BY_PIECES): Likewise.
5037 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
5039 2021-11-11 liuhongt <hongtao.liu@intel.com>
5042 * config/i386/sse.md (V_128_256): Extend to V8HF/V16HF.
5043 (avxsizesuffix): Ditto.
5045 2021-11-11 Kito Cheng <kito.cheng@sifive.com>
5047 * common/config/riscv/riscv-common.c (riscv_subset_list::to_string): Fix
5048 wrong marco checking.
5050 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
5052 PR tree-optimization/102906
5053 * tree-ssa-loop-ch.c (entry_loop_condition_is_static): New.
5054 (should_duplicate_loop_header_p): Call entry_loop_condition_is_static.
5055 (class ch_base): Add m_ranger and m_query.
5056 (ch_base::copy_headers): Pass m_query to
5057 entry_loop_condition_is_static.
5058 (pass_ch::execute): Allocate and deallocate m_ranger and
5060 (pass_ch_vect::execute): Same.
5062 2021-11-10 Andrew Pinski <apinski@marvell.com>
5065 * config/aarch64/aarch64-simd.md (aarch64_simd_dup<mode>):
5066 Use vwcore iterator for the r constraint output string.
5068 2021-11-10 qing zhao <qing.zhao@oracle.com>
5070 * internal-fn.c (expand_DEFERRED_INIT): Apply pattern initialization
5071 only when have_insn_for return true for the mode. Fix a memory leak.
5073 2021-11-10 Christophe Lyon <christophe.lyon@foss.st.com>
5075 * config/arm/arm.c (cortexa9_extra_costs, cortexa8_extra_costs,
5076 cortexa5_extra_costs, cortexa7_extra_costs,
5077 cortexa12_extra_costs, cortexa15_extra_costs, v7m_extra_costs):
5078 Initialize movi, dup and extract costing fields.
5080 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
5082 * gimple-range-path.cc (path_range_query::path_range_query): Do
5084 (path_range_query::dump): Change m_path uses to non-pointer.
5085 (path_range_query::defined_outside_path): Same.
5086 (path_range_query::set_path): Same.
5087 (path_range_query::add_copies_to_imports): Same.
5088 (path_range_query::range_of_stmt): Same.
5089 (path_range_query::compute_outgoing_relations): Same.
5090 (path_range_query::compute_ranges): Imports are now optional.
5091 Implement overload that takes an edge.
5092 * gimple-range-path.h (class path_range_query): Make imports
5093 optional for compute_ranges. Add compute_ranges(edge) overload.
5094 Make m_path an auto_vec instead of a pointer and adjust
5097 2021-11-10 Tamar Christina <tamar.christina@arm.com>
5099 * tree-vectorizer.h (struct scalar_cond_masked_key): Add inverted_p.
5100 (default_hash_traits<scalar_conf_masked_key>): Likewise.
5101 * tree-vect-stmts.c (vectorizable_condition): Check if inverse of mask
5103 * tree-vectorizer.c (scalar_cond_masked_key::get_cond_ops_from_tree):
5104 Register mask inverses.
5106 2021-11-10 Tamar Christina <tamar.christina@arm.com>
5108 * tree-vectorizer.c (vectorize_loops): Do local CSE through RPVN upon
5109 successful vectorization.
5111 2021-11-10 Andrew MacLeod <amacleod@redhat.com>
5113 * gimple-range-cache.cc (sbr_vector::grow): New.
5114 (sbr_vector::set_bb_range): Call grow.
5115 (sbr_vector::get_bb_range): Same.
5116 (sbr_vector::bb_range_p): Remove assert.
5118 2021-11-10 Tamar Christina <tamar.christina@arm.com>
5120 * config/aarch64/aarch64-simd.md (*aarch64_topbits_shuffle<mode>_le
5121 ,*aarch64_topbits_shuffle<mode>_be): Remove.
5123 2021-11-10 Jan Hubicka <jh@suse.cz>
5125 * ipa-modref.c: Include tree-eh.h
5126 (modref_summary::modref_summary): Initialize side_effects.
5127 (struct modref_summary_lto): New bool field side_effects.
5128 (modref_summary_lto::modref_summary_lto): Initialize side_effects.
5129 (modref_summary::dump): Dump side_effects.
5130 (modref_summary_lto::dump): Dump side_effects.
5131 (merge_call_side_effects): Merge side effects.
5132 (process_fnspec): Calls to non-const/pure or looping
5133 function is a side effect.
5134 (analyze_call): Self-recursion is a side-effect; handle
5136 (analyze_load): Watch for volatile and throwing memory.
5137 (analyze_store): Likewise.
5138 (analyze_stmt): Watch for volatitle asm.
5139 (analyze_function): Handle side_effects.
5140 (modref_summaries::duplicate): Duplicate side_effects.
5141 (modref_summaries_lto::duplicate): Likewise.
5142 (modref_write): Stream side_effects.
5143 (read_section): Likewise.
5144 (update_signature): Update.
5145 (propagate_unknown_call): Handle side_effects.
5146 (modref_propagate_in_scc): Likewise.
5147 * ipa-modref.h (struct modref_summary): Add side_effects.
5148 * ipa-pure-const.c (special_builtin_state): Rename to ...
5149 (builtin_safe_for_const_function_p): ... this one.
5150 (check_call): Update.
5151 (finite_function_p): Break out from ...
5152 (propagate_pure_const): ... here
5153 * ipa-utils.h (finite_function): Declare.
5155 2021-11-10 Lucas A. M. Magalhães <lamm@linux.ibm.com>
5157 * config.gcc (powerpc*-*-*): Remove -rpath from
5158 --with-advance-toolchain.
5160 2021-11-10 Marek Polacek <polacek@redhat.com>
5163 * attribs.c (struct scoped_attributes): Add a bool member.
5164 (lookup_scoped_attribute_spec): Forward declare.
5165 (register_scoped_attributes): New bool parameter, defaulted to
5167 (handle_ignored_attributes_option): New function.
5168 (free_attr_data): New function.
5169 (init_attributes): Call handle_ignored_attributes_option.
5170 (attr_namespace_ignored_p): New function.
5171 (decl_attributes): Check attr_namespace_ignored_p before
5173 * attribs.h (free_attr_data): Declare.
5174 (register_scoped_attributes): Adjust declaration.
5175 (handle_ignored_attributes_option): Declare.
5176 (canonicalize_attr_name): New function template.
5177 (canonicalize_attr_name): Use it.
5178 * common.opt (Wattributes=): New option with a variable.
5179 * doc/extend.texi: Document #pragma GCC diagnostic ignored_attributes.
5180 * doc/invoke.texi: Document -Wno-attributes=.
5181 * opts.c (common_handle_option) <case OPT_Wattributes_>: Handle.
5182 * plugin.h (register_scoped_attributes): Adjust declaration.
5183 * toplev.c (compile_file): Call free_attr_data.
5185 2021-11-10 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
5187 * config/arm/arm-cpus.in (cortex-a710): New CPU.
5188 * config/arm/arm-tables.opt: Regenerate.
5189 * config/arm/arm-tune.md: Regenerate.
5190 * doc/invoke.texi: Update docs.
5192 2021-11-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
5194 * config/aarch64/aarch64-builtins.c
5195 (aarch64_general_gimple_fold_builtin): Mark argument as unused.
5197 2021-11-10 Martin Liska <mliska@suse.cz>
5199 * lto-wrapper.c (merge_and_complain): Make the first argument
5202 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
5204 * config/aarch64/iterators.md (optab): Use fmax_nan instead of
5205 smax_nan and fmin_nan instead of smin_nan.
5206 (maxmin_uns): Rename to...
5207 (fmaxmin): ...this and make the same changes. Remove entries
5208 unrelated to fmax* and fmin*.
5209 * config/aarch64/aarch64.md (<maxmin_uns><mode>3): Rename to...
5210 (<fmaxmin><mode>3): ...this.
5211 * config/aarch64/aarch64-simd.md (aarch64_<maxmin_uns>p<mode>):
5213 (aarch64_<optab>p<mode>): ...this.
5214 (<maxmin_uns><mode>3): Rename to...
5215 (<fmaxmin><mode>3): ...this.
5216 (reduc_<maxmin_uns>_scal_<mode>): Rename to...
5217 (reduc_<optab>_scal_<mode>): ...this and update gen* call.
5218 (aarch64_reduc_<maxmin_uns>_internal<mode>): Rename to...
5219 (aarch64_reduc_<optab>_internal<mode>): ...this.
5220 (aarch64_reduc_<maxmin_uns>_internalv2si): Rename to...
5221 (aarch64_reduc_<optab>_internalv2si): ...this.
5222 * config/aarch64/aarch64-sve.md (<maxmin_uns><mode>3): Rename to...
5223 (<fmaxmin><mode>3): ...this.
5224 * config/aarch64/aarch64-simd-builtins.def (smax_nan, smin_nan)
5226 (fmax_nan, fmin_nan): ...this.
5227 * config/aarch64/arm_neon.h (vmax_f32, vmax_f64, vmaxq_f32, vmaxq_f64)
5228 (vmin_f32, vmin_f64, vminq_f32, vminq_f64, vmax_f16, vmaxq_f16)
5229 (vmin_f16, vminq_f16): Update accordingly.
5231 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
5233 * tree-vectorizer.h (vector_costs::finish_cost): Take the
5234 corresponding scalar costs as a parameter.
5235 (finish_cost): Likewise.
5236 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
5237 (vect_estimate_min_profitable_iters): Update accordingly.
5238 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Likewise.
5239 * tree-vectorizer.c (vector_costs::finish_cost): Likewise.
5240 * config/aarch64/aarch64.c (aarch64_vector_costs::finish_cost):
5242 * config/rs6000/rs6000.c (rs6000_cost_data::finish_cost): Likewise.
5244 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
5246 * tree-vectorizer.h (_loop_vec_info::scalar_costs): New member
5248 (_loop_vec_info::single_scalar_iteration_cost): Delete.
5249 (LOOP_VINFO_SINGLE_SCALAR_ITERATION_COST): Delete.
5250 (vector_costs::total_cost): New function.
5251 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
5252 after above changes.
5253 (_loop_vec_info::~_loop_vec_info): Delete scalar_costs.
5254 (vect_compute_single_scalar_iteration_cost): Store the costs
5255 in loop_vinfo->scalar_costs.
5256 (vect_estimate_min_profitable_iters): Get the scalar cost from
5257 loop_vinfo->scalar_costs.
5259 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
5261 * tree-vectorizer.h (vector_costs::better_main_loop_than_p)
5262 (vector_costs::better_epilogue_loop_than_p)
5263 (vector_costs::compare_inside_loop_cost)
5264 (vector_costs::compare_outside_loop_cost): Likewise.
5265 * tree-vectorizer.c (vector_costs::better_main_loop_than_p)
5266 (vector_costs::better_epilogue_loop_than_p)
5267 (vector_costs::compare_inside_loop_cost)
5268 (vector_costs::compare_outside_loop_cost): New functions,
5269 containing code moved from...
5270 * tree-vect-loop.c (vect_better_loop_vinfo_p): ...here.
5272 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
5274 * tree-vectorizer.h (_loop_vec_info): Remove vec_outside_cost
5275 and vec_inside_cost.
5276 (vector_costs::outside_cost): New function.
5277 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
5279 (vect_estimate_min_profitable_iters): Likewise.
5280 (vect_better_loop_vinfo_p): Get the inside and outside costs
5281 from the loop_vec_infos' vector_costs.
5283 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
5285 * tree-vectorizer.h (vec_info::target_cost_data): Replace with...
5286 (_loop_vec_info::vector_costs): ...this.
5287 (LOOP_VINFO_TARGET_COST_DATA): Delete.
5288 * tree-vectorizer.c (vec_info::vec_info): Remove target_cost_data
5290 (vec_info::~vec_info): Remove corresponding delete.
5291 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5292 vector_costs to null.
5293 (_loop_vec_info::~_loop_vec_info): Delete vector_costs.
5294 (vect_analyze_loop_operations): Update after above changes.
5295 (vect_analyze_loop_2): Likewise.
5296 (vect_estimate_min_profitable_iters): Likewise.
5297 * tree-vect-slp.c (vect_slp_analyze_operations): Likewise.
5299 2021-11-10 Jan Hubicka <hubicka@ucw.cz>
5301 * tree-core.h (EAF_DIRECT): Remove.
5302 (EAF_NOCLOBBER): Remove.
5303 (EAF_UNUSED): Remove.
5304 (EAF_NOESCAPE): Remove.
5305 (EAF_NO_DIRECT_CLOBBER): New.
5306 (EAF_NO_INDIRECT_CLOBBER): New.
5307 (EAF_NODIRECTESCAPE): Remove.
5308 (EAF_NO_DIRECT_ESCAPE): New.
5309 (EAF_NO_INDIRECT_ESCAPE): New.
5310 (EAF_NOT_RETURNED): Remove.
5311 (EAF_NOT_RETURNED_INDIRECTLY): New.
5312 (EAF_NOREAD): Remove.
5313 (EAF_NO_DIRECT_READ): New.
5314 (EAF_NO_INDIRECT_READ): New.
5315 * gimple.c (gimple_call_arg_flags): Update for new flags.
5316 (gimple_call_retslot_flags): Update for new flags.
5317 * ipa-modref.c (dump_eaf_flags): Likewise.
5318 (remove_useless_eaf_flags): Likewise.
5319 (deref_flags): Likewise.
5320 (modref_lattice::init): Likewise.
5321 (modref_lattice::merge): Likewise.
5322 (modref_lattice::merge_direct_load): Likewise.
5323 (modref_lattice::merge_direct_store): Likewise.
5324 (modref_eaf_analysis::merge_call_lhs_flags): Likewise.
5325 (callee_to_caller_flags): Likewise.
5326 (modref_eaf_analysis::analyze_ssa_name): Likewise.
5327 (modref_eaf_analysis::propagate): Likewise.
5328 (modref_merge_call_site_flags): Likewise.
5329 * ipa-modref.h (interposable_eaf_flags): Likewise.
5330 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1) Likewise.
5331 * tree-ssa-structalias.c (handle_call_arg): Likewise.
5332 (handle_rhs_call): Likewise.
5333 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Likewise.
5335 2021-11-10 Tamar Christina <tamar.christina@arm.com>
5337 * tree-vect-slp-patterns.c (complex_mul_pattern::matches): Remove l1node.
5339 2021-11-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
5341 * config/aarch64/aarch64-builtins.c
5342 (aarch64_general_gimple_fold_builtin): Change pointer alignment and
5345 2021-11-10 Jan Hubicka <jh@suse.cz>
5347 * ipa-modref-tree.h (modref_tree::remap_params): Fix off-by-one error.
5349 2021-11-10 H.J. Lu <hongjiu.lu@intel.com>
5350 Hongtao Liu <hongtao.liu@intel.com>
5352 PR middle-end/102566
5353 * match.pd (nop_atomic_bit_test_and_p): New match.
5354 * tree-ssa-ccp.c (convert_atomic_bit_not): New function.
5355 (gimple_nop_atomic_bit_test_and_p): New prototype.
5356 (optimize_atomic_bit_test_and): Transform equivalent, but slighly
5357 different cases to their canonical forms.
5359 2021-11-10 liuhongt <hongtao.liu@intel.com>
5361 PR tree-optimization/103126
5362 * tree-vect-loop.c (neutral_op_for_reduction): Remove static.
5363 * tree-vectorizer.h (neutral_op_for_reduction): Declare.
5364 * tree-if-conv.c : Include tree-vectorizer.h.
5365 (is_cond_scalar_reduction): Handle
5366 BIT_XOR_EXPR/BIT_IOR_EXPR/BIT_AND_EXPR.
5367 (convert_scalar_cond_reduction): Ditto.
5369 2021-11-10 konglin1 <lingling.kong@intel.com>
5371 * config/i386/sse.md (cmul<conj_op><mode>3): add new define_expand.
5372 (cmla<conj_op><mode>4): Likewise
5374 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
5376 * tree-ssa-threadedge.c: Do not include
5377 gimple-ssa-evrp-analyze.h.
5378 * value-pointer-equiv.cc: Same.
5380 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
5382 * tree-ssa-threadbackward.c
5383 (back_threader::maybe_register_path_dump): Abstract path dumping...
5384 (dump_path): ...here.
5385 (back_threader::resolve_phi): Call dump_path.
5388 2021-11-10 konglin1 <lingling.kong@intel.com>
5390 * config/i386/sse.md (fma_<complexpairopname>_<mode>_pair):
5391 Add new define_insn.
5392 (fma_<mode>_fmaddc_bcst): Add new define_insn_and_split.
5393 (fma_<mode>_fcmaddc_bcst): Likewise
5395 2021-11-10 liuhongt <hongtao.liu@intel.com>
5398 * match.pd: Simplify (trunc)fmax/fmin((extend)a, (extend)b) to
5401 2021-11-10 Andrew Pinski <apinski@marvell.com>
5404 * config/aarch64/aarch64.c (aarch64_evpc_ins): Don't use target
5405 as an input, use original one.
5407 2021-11-10 Sandra Loosemore <sandra@codesourcery.com>
5409 * config/nios2/nios2.c (nios2_can_inline_p): New.
5410 (TARGET_CAN_INLINE_P): Define.
5412 2021-11-09 Bill Schmidt <wschmidt@linux.ibm.com>
5414 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin):
5415 Disable gimple fold for RS6000_BIF_{XVMINDP,XVMINSP,VMINFP} and
5416 RS6000_BIF_{XVMAXDP,XVMAXSP,VMAXFP} when fast-math is not set.
5417 (lxvrse_expand_builtin): Modify the expansion for sign extension.
5418 All extensions are done within VSX registers.
5420 2021-11-09 Andrew Pinski <apinski@marvell.com>
5422 PR tree-optimization/10352
5423 * match.pd: Remove check of TYPE_PRECISION for
5426 2021-11-09 Andrew MacLeod <amacleod@redhat.com>
5428 * function.c (allocate_struct_function): Don't set x_range_query.
5429 * function.h (get_range_query): Move to value-query.h.
5430 * gimple-range.cc (enable_ranger): Check that query is currently NULL.
5431 (disable_ranger): Clear function current query field.
5432 * value-query.cc (get_global_range_query): Relocate to:
5433 * value-query.h (get_global_range_query): Here and inline.
5434 (get_range_query): Relocate here from function.h.
5436 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
5438 * tree-ssa-threadbackward.c
5439 (back_threader::maybe_register_path_dump): New.
5440 (back_threader::maybe_register_path): Call maybe_register_path_dump.
5442 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
5444 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
5445 Return NULL when unprofitable.
5447 2021-11-09 Martin Jambor <mjambor@suse.cz>
5449 * tree.h (build_debug_expr_decl): Declare.
5450 * tree.c (build_debug_expr_decl): New function.
5451 * cfgexpand.c (avoid_deep_ter_for_debug): Use build_debug_expr_decl
5452 instead of building a DEBUG_EXPR_DECL.
5453 * ipa-param-manipulation.c
5454 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
5455 * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Likewise.
5456 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
5457 * tree-ssa-phiopt.c (spaceship_replacement): Likewise.
5458 * tree-ssa-reassoc.c (make_new_ssa_for_def): Likewise.
5460 2021-11-09 Bill Schmidt <wschmidt@linux.ibm.com>
5462 * config/rs6000/rs6000-call.c (rs6000_type_string): New function.
5463 (def_builtin): Change debug formatting for easier parsing and
5464 include more information.
5465 (rs6000_init_builtins): Add dump of autogenerated builtins.
5466 (altivec_init_builtins): Dump __builtin_altivec_mask_for_load for
5469 2021-11-09 Marek Polacek <polacek@redhat.com>
5471 * ipa-modref.c (analyze_parms): Fix a typo.
5473 2021-11-09 Kito Cheng <kito.cheng@sifive.com>
5476 * common/config/riscv/riscv-common.c (multi_letter_subset_rank): Remove
5477 assertion for Z*-ext.
5479 2021-11-09 Jan Hubicka <hubicka@ucw.cz>
5481 * ipa-modref.c (analyze_parms): Add past_flags, past_retslot_flags
5482 and past_static_chain; merge past summary with current one.
5483 (analyze_function): Update.
5485 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
5487 * doc/invoke.texi (Invoking GCC): Document --param=threader-debug.
5489 2021-11-09 Martin Liska <mliska@suse.cz>
5491 * print-rtl.c (rtx_writer::rtx_writer): Make the compilation
5492 conditional based on
5493 * print-rtl.h (class rtx_writer): Likewise.
5495 2021-11-09 Thomas Schwinge <thomas@codesourcery.com>
5497 * input.h (location_hash): Use 'BUILTINS_LOCATION' as spare value
5498 for 'Deleted'. Turn into a '#define'.
5500 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
5502 * gimple-range-path.cc (path_range_query::dump): Clean up.
5503 (path_range_query::compute_ranges): Same.
5504 * value-relation.cc (path_oracle::dump): Same.
5506 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
5508 * dumpfile.c (dump_options): Remove TDF_THREADING entry.
5509 * dumpfile.h (enum dump_flag): Remove TDF_THREADING and adjust
5511 * flag-types.h (enum threader_debug): New.
5512 * gimple-range-path.cc (DEBUG_SOLVER): Use param_threader_debug.
5513 * params.opt: Add entry for --param=threader-debug=.
5515 2021-11-09 Thomas Schwinge <thomas@codesourcery.com>
5517 PR middle-end/101204
5519 * diagnostic-spec.h (typedef xint_hash_t): Turn into...
5520 (struct xint_hash_t): ... this.
5521 * doc/gty.texi: Update.
5523 2021-11-09 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
5525 * config/arm/arm-cpus.in (armv9): New define.
5526 (ARMv9a): New group.
5527 (armv9-a): New arch definition.
5528 * config/arm/arm-tables.opt: Regenerate.
5529 * config/arm/arm.h (BASE_ARCH_9A): New arch enum value.
5530 * config/arm/t-aprofile: Added armv9-a and armv9+simd.
5531 * config/arm/t-arm-elf: Added arm9-a, v9_fps and all_v9_archs
5532 to MULTILIB_MATCHES.
5533 * config/arm/t-multilib: Added v9_a_nosimd_variants and
5534 v9_a_simd_variants to MULTILIB_MATCHES.
5535 * doc/invoke.texi: Update docs.
5537 2021-11-09 Martin Jambor <mjambor@suse.cz>
5540 * ipa-param-manipulation.c (replace_with_mapped_expr): Early
5541 return with error_mark_mode when part of expression is mapped to
5543 (ipa_param_body_adjustments::remap_with_debug_expressions): Set
5544 mapped value to NULL if walk_tree returns error_mark_mode.
5546 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
5548 * tree-ssa-strlen.c (compare_nonzero_chars): Pass statement
5550 (get_addr_stridx): Same.
5552 (get_range_strlen_dynamic): Same.
5553 (handle_builtin_strlen): Same.
5554 (handle_builtin_strchr): Same.
5555 (handle_builtin_strcpy): Same.
5556 (maybe_diag_stxncpy_trunc): Same.
5557 (handle_builtin_stxncpy_strncat): Same.
5558 (handle_builtin_memcpy): Same.
5559 (handle_builtin_strcat): Same.
5560 (handle_alloc_call): Same.
5561 (handle_builtin_memset): Same.
5562 (handle_builtin_string_cmp): Same.
5563 (handle_pointer_plus): Same.
5564 (count_nonzero_bytes_addr): Same.
5565 (count_nonzero_bytes): Same.
5566 (handle_store): Same.
5567 (fold_strstr_to_strncmp): Same.
5568 (handle_integral_assign): Same.
5569 (check_and_optimize_stmt): Same.
5570 (class strlen_dom_walker): Replace evrp with ranger.
5571 (strlen_dom_walker::before_dom_children): Remove evrp.
5572 (strlen_dom_walker::after_dom_children): Remove evrp.
5573 * gimple-ssa-warn-access.cc (maybe_check_access_sizes):
5574 Restrict sprintf output.
5576 2021-11-09 Martin Liska <mliska@suse.cz>
5578 * genconditions.c (write_one_condition): Add const qualifier
5581 2021-11-09 Jeff Law <jeffreyalaw@gmail.com>
5583 * match.pd: New pattern to simplify (1 << n) & M ==/!= 0 for M
5586 2021-11-08 Bill Schmidt <wschmidt@linux.ibm.com>
5588 * config/rs6000/rs6000-builtin-new.def (VEC_INIT_V16QI): Use
5589 escape-newline support.
5590 (VEC_INIT_V4SI): Likewise.
5591 (VEC_INIT_V8HI): Likewise.
5592 (PACK_V1TI): Likewise.
5594 (VFIRSTMISMATCHOREOSINDEX_V16QI): Likewise.
5595 (VFIRSTMISMATCHOREOSINDEX_V8HI): Likewise.
5596 (VFIRSTMISMATCHOREOSINDEX_V4SI): Likewise.
5600 (FMAF128_ODD): Likewise.
5601 (VSCEQPUO): Likewise.
5603 (VSIEQPF): Likewise.
5604 (VSTDCQP): Likewise.
5605 (PACK_TD): Likewise.
5606 (TABORTDC): Likewise.
5607 (TABORTDCI): Likewise.
5608 (SE_LXVRBX): Likewise.
5609 (SE_LXVRHX): Likewise.
5610 (SE_LXVRWX): Likewise.
5611 (SE_LXVRDX): Likewise.
5612 (VREPLACE_UN_UV2DI): Likewise.
5613 (VREPLACE_UN_UV4SI): Likewise.
5614 (VREPLACE_UN_V2DI): Likewise.
5615 (VREPLACE_ELT_UV2DI): Likewise.
5616 (VREPLACE_ELT_V2DI): Likewise.
5617 (ZE_LXVRBX): Likewise.
5618 (ZE_LXVRHX): Likewise.
5619 (ZE_LXVRWX): Likewise.
5620 (ZE_LXVRDX): Likewise.
5622 (CNTLZDM): Likewise.
5623 (CNTTZDM): Likewise.
5626 (PMXVBF16GER2): Likewise.
5627 (PMXVBF16GER2_INTERNAL): Likewise.
5628 (PMXVBF16GER2NN): Likewise.
5629 (PMXVBF16GER2NN_INTERNAL): Likewise.
5630 (PMXVBF16GER2NP): Likewise.
5631 (PMXVBF16GER2NP_INTERNAL): Likewise.
5632 (PMXVBF16GER2PN): Likewise.
5633 (PMXVBF16GER2PN_INTERNAL): Likewise.
5634 (PMXVBF16GER2PP): Likewise.
5635 (PMXVBF16GER2PP_INTERNAL): Likewise.
5636 (PMXVF16GER2): Likewise.
5637 (PMXVF16GER2_INTERNAL): Likewise.
5638 (PMXVF16GER2NN): Likewise.
5639 (PMXVF16GER2NN_INTERNAL): Likewise.
5640 (PMXVF16GER2NP): Likewise.
5641 (PMXVF16GER2NP_INTERNAL): Likewise.
5642 (PMXVF16GER2PN): Likewise.
5643 (PMXVF16GER2PN_INTERNAL): Likewise.
5644 (PMXVF16GER2PP): Likewise.
5645 (PMXVF16GER2PP_INTERNAL): Likewise.
5646 (PMXVF32GER_INTERNAL): Likewise.
5647 (PMXVF32GERNN): Likewise.
5648 (PMXVF32GERNN_INTERNAL): Likewise.
5649 (PMXVF32GERNP): Likewise.
5650 (PMXVF32GERNP_INTERNAL): Likewise.
5651 (PMXVF32GERPN): Likewise.
5652 (PMXVF32GERPN_INTERNAL): Likewise.
5653 (PMXVF32GERPP): Likewise.
5654 (PMXVF32GERPP_INTERNAL): Likewise.
5655 (PMXVF64GER): Likewise.
5656 (PMXVF64GER_INTERNAL): Likewise.
5657 (PMXVF64GERNN): Likewise.
5658 (PMXVF64GERNN_INTERNAL): Likewise.
5659 (PMXVF64GERNP): Likewise.
5660 (PMXVF64GERNP_INTERNAL): Likewise.
5661 (PMXVF64GERPN): Likewise.
5662 (PMXVF64GERPN_INTERNAL): Likewise.
5663 (PMXVF64GERPP): Likewise.
5664 (PMXVF64GERPP_INTERNAL): Likewise.
5665 (PMXVI16GER2): Likewise.
5666 (PMXVI16GER2_INTERNAL): Likewise.
5667 (PMXVI16GER2PP): Likewise.
5668 (PMXVI16GER2PP_INTERNAL): Likewise.
5669 (PMXVI16GER2S): Likewise.
5670 (PMXVI16GER2S_INTERNAL): Likewise.
5671 (PMXVI16GER2SPP): Likewise.
5672 (PMXVI16GER2SPP_INTERNAL): Likewise.
5673 (PMXVI4GER8): Likewise.
5674 (PMXVI4GER8_INTERNAL): Likewise.
5675 (PMXVI4GER8PP): Likewise.
5676 (PMXVI4GER8PP_INTERNAL): Likewise.
5677 (PMXVI8GER4): Likewise.
5678 (PMXVI8GER4_INTERNAL): Likewise.
5679 (PMXVI8GER4PP): Likewise.
5680 (PMXVI8GER4PP_INTERNAL): Likewise.
5681 (PMXVI8GER4SPP): Likewise.
5682 (PMXVI8GER4SPP_INTERNAL): Likewise.
5683 * config/rs6000/rs6000-gen-builtins.c (MAXLINES): New macro.
5684 (linebuf): Increase size.
5685 (lines): New variable.
5686 (lastline): Likewise.
5687 (real_line_pos): New function.
5688 (diag): Change signature.
5689 (bif_diag): Change signature; support escape-newline handling.
5690 (ovld_diag): Likewise.
5691 (fatal): Move earlier.
5692 (consume_whitespace): Adjust diag call.
5693 (advance_line): Add escape-newline handling; call fatal.
5694 (safe_inc_pos): Adjust diag call.
5695 (match_identifier): Likewise.
5696 (match_integer): Likewise.
5697 (match_to_right_bracket): Call fatal instead of diag; adjust diag
5699 (match_basetype): Adjust diag calls.
5700 (match_bracketed_pair): Likewise.
5701 (match_const_restriction): Likewise.
5702 (match_type): Likewise.
5703 (parse_args): Likewise.
5704 (parse_bif_attrs): Likewise.
5705 (complete_vector_type): Likewise.
5706 (complete_base_type): Likewise.
5707 (parse_prototype): Likewise.
5708 (parse_bif_entry): Likewise.
5709 (parse_bif_stanza): Likewise.
5710 (parse_ovld_entry): Likewise.
5711 (parse_ovld_stanza): Likewise.
5712 (main): Allocate buffers for lines[].
5714 2021-11-08 Eric Botcazou <ebotcazou@adacore.com>
5716 * config/i386/i386.c (legitimize_pic_address): Adjust comment and
5717 use the REG argument on the CM_LARGE_PIC code path as well.
5718 * config/i386/predicates.md (gotoff_operand): Do not treat VxWorks
5719 specially with the large code models.
5721 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
5723 * gimple.c (gimple_call_static_chain_flags): Revert the workaround
5724 allowing interposition since issues with binds_to_local_def were
5727 2021-11-08 Andrew MacLeod <amacleod@redhat.com>
5729 PR tree-optimization/103122
5730 * gimple-range.cc (gimple_ranger::range_of_expr): Request the cache
5731 entry with "calulate new values" set to false.
5733 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
5735 * builtins.c (is_simple_builtin): Add builitin_dwarf_cfa
5736 and builtin_return_address.
5737 (builtin_fnspec): Annotate builtin_return,
5738 bulitin_eh_pointer, builtin_eh_filter, builtin_unwind_resume,
5739 builtin_cxa_end_cleanup, builtin_eh_copy_values,
5740 builtin_frame_address, builtin_apply_args,
5741 builtin_asan_before_dynamic_init, builtin_asan_after_dynamic_init,
5742 builtin_prefetch, builtin_dwarf_cfa, builtin_return_addrss
5744 * ipa-pure-const.c (special_builtin_state): Add builtin_dwarf_cfa
5745 and builtin_return_address.
5747 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
5749 PR tree-optimization/103177
5750 * passes.def: Move uncprop after pure/const and modref.
5752 2021-11-08 Martin Jambor <mjambor@suse.cz>
5756 * tree-inline.c (remap_gimple_stmt): Unshare the expression without
5757 location before invoking remap_with_debug_expressions on it.
5758 * ipa-param-manipulation.c
5759 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
5761 2021-11-08 David Edelsohn <dje.gcc@gmail.com>
5763 * config/rs6000/vsx.md (vsx_splat_v4si_di): Revert "wa"
5766 2021-11-08 Richard Biener <rguenther@suse.de>
5768 * cfgloop.c (verify_loop_structure): Use a temporary BB flag
5769 instead of an sbitmap to cache irreducible state.
5771 2021-11-08 Aldy Hernandez <aldyh@redhat.com>
5773 PR tree-optimization/103120
5774 * gimple-range-path.cc (path_range_query::range_defined_in_block):
5775 Bail if there's a cache entry.
5777 2021-11-08 Bill Schmidt <wschmidt@linux.ibm.com>
5779 * config/rs6000/rs6000.c (rs6000_builtin_reciprocal): Use
5780 rs6000_builtin_decls_x when appropriate.
5781 (add_condition_to_bb): Likewise.
5782 (rs6000_atomic_assign_expand_fenv): Likewise.
5784 2021-11-08 Bill Schmidt <wschmidt@linux.ibm.com>
5786 * config/rs6000/rs6000-call.c (rs6000_new_builtin_decl): New function.
5787 (rs6000_builtin_decl): Call it.
5789 2021-11-08 Martin Liska <mliska@suse.cz>
5791 * cgraph.c (cgraph_node::dump): Dump it from decl.
5793 2021-11-08 Martin Liska <mliska@suse.cz>
5795 PR gcov-profile/100520
5796 * coverage.c (coverage_compute_profile_id): Strip .gk when
5797 compare debug is used.
5798 * system.h (endswith): New function.
5800 2021-11-08 Martin Liska <mliska@suse.cz>
5802 * cgraph.c (cgraph_node::dump): Dump static_chain_decl.
5804 2021-11-08 Thomas Schwinge <thomas@codesourcery.com>
5806 * config/rs6000/rbtree.c: Fix 'Copyright (C) 2020-21' into '2020-2021'
5807 * config/rs6000/rbtree.h: Likewise.
5808 * config/rs6000/rs6000-builtin-new.def: Likewise.
5809 * config/rs6000/rs6000-gen-builtins.c: Likewise.
5810 * config/rs6000/rs6000-overload.def: Likewise.
5812 2021-11-08 Richard Sandiford <richard.sandiford@arm.com>
5814 * config/aarch64/aarch64.c (aarch64_ldn_stn_vectors): New function.
5815 (aarch64_address_cost): Use it instead of testing for CImode and
5818 2021-11-08 Richard Sandiford <richard.sandiford@arm.com>
5820 * genmodes.c (emit_insn_modes_h): Define NUM_MODE_* macros.
5821 * expmed.h (NUM_MODE_INT): Delete in favor of genmodes definitions.
5822 (NUM_MODE_PARTIAL_INT, NUM_MODE_VECTOR_INT): Likewise.
5823 * real.h (real_format_for_mode): Use NUM_MODE_FLOAT and
5824 NUM_MODE_DECIMAL_FLOAT.
5825 (REAL_MODE_FORMAT): Likewise.
5827 2021-11-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
5829 * tree-vect-loop.c (vect_better_loop_vinfo_p): Change how epilogue loop
5832 2021-11-08 Richard Biener <rguenther@suse.de>
5834 * tree-vectorizer.h (vect_create_loop_vinfo): Add main_loop_info
5836 * tree-vect-loop.c (vect_create_loop_vinfo): Likewise. Set
5837 LOOP_VINFO_ORIG_LOOP_INFO and conditionalize set of
5838 LOOP_VINFO_NITERS_ASSUMPTIONS.
5839 (vect_analyze_loop_1): Adjust.
5840 (vect_analyze_loop): Move loop constraint setting and
5841 SCEV/niter reset here from vect_create_loop_vinfo to perform
5843 (vect_analyze_loop_form): Move dumping of symbolic niters
5844 here from vect_create_loop_vinfo.
5846 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
5848 * ipa-modref-tree.h (enum modref_special_parms): New enum.
5849 (struct modref_access_node): update for special parms.
5850 (struct modref_ref_node): Likewise.
5851 (struct modref_parm_map): Likewise.
5852 (struct modref_tree): Likewise.
5853 * ipa-modref.c (dump_access): Likewise.
5854 (get_access): Detect static chain.
5855 (parm_map_for_arg): Take tree as arg instead of
5857 (merge_call_side_effects): Compute map for static chain.
5858 (process_fnspec): Update.
5859 (struct escape_point): Remove retslot_arg and static_chain_arg.
5860 (analyze_parms): Update.
5861 (compute_parm_map): Update.
5862 (propagate_unknown_call): Update.
5863 (modref_propagate_in_scc): Update.
5864 (modref_merge_call_site_flags): Update.
5865 (ipa_merge_modref_summary_after_inlining): Update.
5866 * tree-ssa-alias.c (modref_may_conflict): Handle static chain.
5867 * ipa-modref-tree.c (test_merge): Update.
5869 2021-11-08 Haochen Gui <guihaoc@gcc.gnu.org>
5871 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin): Disable
5872 gimple fold for VSX_BUILTIN_XVMINDP, ALTIVEC_BUILTIN_VMINFP,
5873 VSX_BUILTIN_XVMAXDP, ALTIVEC_BUILTIN_VMAXFP when fast-math is not
5876 2021-11-08 liuhongt <hongtao.liu@intel.com>
5878 PR tree-optimization/103077
5879 * doc/invoke.texi (Options That Control Optimization):
5880 Update documentation for -ftree-loop-vectorize and
5881 -ftree-slp-vectorize which are enabled by default at -02.
5883 2021-11-08 liuhongt <hongtao.liu@intel.com>
5886 * match.pd (Simplifcation (trunc)copysign((extend)a, (extend)b)
5887 to .COPYSIGN (a, b)): Add !HONOR_SNANS.
5889 2021-11-08 liuhongt <hongtao.liu@intel.com>
5892 * match.pd: Simplify
5893 (trunc)fma ((extend)a, (extend)b, (extend)c) to IFN_FMA (a, b,
5894 c) under flag_unsafe_math_optimizations.
5896 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
5899 * params.opt (modref-max-depth): Add range.
5900 (modref-max-adjustments): Fix range.
5902 2021-11-07 Aldy Hernandez <aldyh@redhat.com>
5904 * tree-pass.h (make_pass_vrp_threader): Remove.
5905 * tree-ssa-threadbackward.c
5906 (back_threader_profitability::profitable_path_p): Remove
5907 ASSERT_EXPR references.
5908 * tree-ssa-threadedge.c (jt_state::register_equivs_stmt): Same.
5909 * tree-vrp.c (vrp_folder::simplify_casted_conds): Same.
5910 (execute_vrp): Same.
5911 (class hybrid_threader): Remove.
5912 (hybrid_threader::hybrid_threader): Remove.
5913 (hybrid_threader::~hybrid_threader): Remove.
5914 (hybrid_threader::before_dom_children): Remove.
5915 (hybrid_threader::after_dom_children): Remove.
5916 (execute_vrp_threader): Remove.
5917 (class pass_vrp_threader): Remove.
5918 (make_pass_vrp_threader): Remove.
5920 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
5922 * gimple.c (gimple_call_arg_flags): Use interposable_eaf_flags.
5923 (gimple_call_retslot_flags): Likewise.
5924 (gimple_call_static_chain_flags): Likewise.
5925 * ipa-modref.c (remove_useless_eaf_flags): Do not remove everything for
5927 (modref_summary::useful_p): Likewise.
5928 (modref_summary_lto::useful_p): Likewise.
5929 (analyze_parms): Do not give up on NOVOPS.
5930 (analyze_function): When dumping report chnages in EAF flags
5931 between IPA and local pass.
5932 (modref_merge_call_site_flags): Compute implicit eaf flags
5933 based on callee ecf_flags and fnspec; if the function does not
5934 bind to current defs use interposable_eaf_flags.
5935 (modref_propagate_flags_in_scc): Update.
5936 * ipa-modref.h (interposable_eaf_flags): New function.
5938 2021-11-07 Bill Schmidt <wschmidt@linux.ibm.com>
5940 * config/rs6000/rs6000-call.c (rs6000_expand_new_builtin): New
5942 (rs6000_invalid_new_builtin): New function.
5943 (rs6000_expand_builtin): Call rs6000_expand_new_builtin.
5944 (rs6000_expand_ldst_mask): New function.
5945 (new_cpu_expand_builtin): Likewise.
5946 (elemrev_icode): Likewise.
5947 (ldv_expand_builtin): Likewise.
5948 (lxvrse_expand_builtin): Likewise.
5949 (lxvrze_expand_builtin): Likewise.
5950 (stv_expand_builtin): Likewise.
5951 (new_mma_expand_builtin): Likewise.
5952 (new_htm_spr_num): Likewise.
5953 (new_htm_expand_builtin): Likewise.
5954 (rs6000_expand_new_builtin): Likewise.
5955 (rs6000_init_builtins): Initialize altivec_builtin_mask_for_load.
5957 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
5959 * ipa-modref.c (modref_lattice): Add do_dataflow,
5960 changed and propagate_to fields.
5961 (modref_lattice::release): Free propagate_to
5962 (modref_lattice::merge): Do not give up early on unknown
5964 (modref_lattice::merge_deref): Likewise.
5965 (modref_eaf_analysis): Update toplevel comment.
5966 (modref_eaf_analysis::analyze_ssa_name): Record postponned ssa names;
5967 do optimistic dataflow initialization.
5968 (modref_eaf_analysis::merge_with_ssa_name): Build dataflow graph.
5969 (modref_eaf_analysis::propagate): New member function.
5970 (analyze_parms): Update to new API of modref_eaf_analysis.
5972 2021-11-06 Jan Hubicka <hubicka@ucw.cz>
5974 * cgraph.h (cgraph_node::can_be_discarded_p): Do not
5975 return true on functions from other partition.
5977 2021-11-06 David Edelsohn <dje.gcc@gmail.com>
5979 * config/rs6000/vsx.md (vsx_splat_v4si): Change constraints to "wa".
5980 (vsx_splat_v4si_di): Change constraint to "wa".
5982 2021-11-06 Aldy Hernandez <aldyh@redhat.com>
5984 PR tree-optimization/103061
5985 * value-relation.cc (path_oracle::path_oracle): Initialize
5987 (path_oracle::killing_def): Set m_killed_defs.
5988 (path_oracle::query_relation): Do not look at the root oracle for
5990 * value-relation.h (class path_oracle): Add m_killed_defs.
5992 2021-11-06 Aldy Hernandez <aldyh@redhat.com>
5994 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
5995 Remove gotos and other cleanups.
5997 2021-11-05 Jan Hubicka <hubicka@ucw.cz>
6000 * ipa-modref-tree.h (modref_tree::insert): Do nothing for
6001 paradoxical and zero sized accesses.
6003 2021-11-05 Jan Hubicka <hubicka@ucw.cz>
6006 * ipa-modref-tree.h (struct modref_access_node): Avoid left shift
6009 2021-11-05 Iain Sandoe <iain@sandoe.co.uk>
6011 * config/darwin.h (ASM_GENERATE_INTERNAL_LABEL): Add LTRAMP
6012 to the list of symbol prefixes that must be made linker-
6015 2021-11-05 Iain Sandoe <iain@sandoe.co.uk>
6017 * config.host: Add support for aarch64-*-darwin.
6018 * config/aarch64/host-aarch64-darwin.c: New file.
6019 * config/aarch64/x-darwin: New file.
6021 2021-11-05 Andrew MacLeod <amacleod@redhat.com>
6023 PR tree-optimization/103093
6024 * gimple-range-gori.cc (range_def_chain::get_imports): Remove assert.
6026 2021-11-05 Andrew MacLeod <amacleod@redhat.com>
6028 PR tree-optimization/102943
6029 * gimple-range-cache.cc (class update_list): New.
6030 (update_list::add): Replace add_to_update.
6031 (update_list::pop): New.
6032 (ranger_cache::ranger_cache): Adjust.
6033 (ranger_cache::~ranger_cache): Adjust.
6034 (ranger_cache::add_to_update): Delete.
6035 (ranger_cache::propagate_cache): Adjust to new class.
6036 (ranger_cache::propagate_updated_value): Ditto.
6037 (ranger_cache::fill_block_cache): Ditto.
6038 * gimple-range-cache.h (class ranger_cache): Adjust to update class.
6040 2021-11-05 Richard Biener <rguenther@suse.de>
6042 * tree-vect-loop.c (vect_analyze_loop): Remove obsolete
6043 comment and expand on another one. Combine nested if.
6045 2021-11-05 John David Anglin <danglin@gcc.gnu.org>
6048 * config/pa/pa-modes.def: Add OImode integer type.
6049 * config/pa/pa.c (pa_scalar_mode_supported_p): Allow TImode
6051 * config/pa/pa.h (MIN_UNITS_PER_WORD) Define to MIN_UNITS_PER_WORD
6052 to UNITS_PER_WORD if IN_LIBGCC2.
6053 * config/pa/pa.md (addti3, addvti3, subti3, subvti3, negti2,
6054 negvti2, ashlti3, shrpd_internal): New patterns.
6055 Change some multi instruction types to multi.
6057 2021-11-05 Jakub Jelinek <jakub@redhat.com>
6060 * config/i386/i386.h
6061 (stringop_algs::stringop_strategy::stringop_strategy): Make the ctor
6064 2021-11-05 Wilco Dijkstra <wdijkstr@arm.com>
6067 * config/aarch64/aarch64.c (aarch64_mov_operand_p): Strip the salt
6069 * config/aarch64/constraints.md: Support const in Usw.
6071 2021-11-05 John David Anglin <danglin@gcc.gnu.org>
6073 * config/pa/pa.h (PREFERRED_DEBUGGING_TYPE): Define to DWARF2_DEBUG.
6074 * config/pa/pa64-hpux.h (PREFERRED_DEBUGGING_TYPE): Remove define.
6076 2021-11-05 Richard Biener <rguenther@suse.de>
6078 * tree-vectorizer.h (vec_info_shared::n_stmts): Add.
6079 (LOOP_VINFO_N_STMTS): Likewise.
6080 (vec_info_for_bb): Remove unused function.
6081 * tree-vectorizer.c (vec_info_shared::vec_info_shared):
6082 Initialize n_stmts member.
6083 * tree-vect-loop.c: Remove INCLUDE_FUNCTIONAL.
6084 (vect_create_loop_vinfo): Do not set loop->aux.
6085 (vect_analyze_loop_2): Do not get n_stmts as argument,
6086 instead use LOOP_VINFO_N_STMTS. Set LOOP_VINFO_VECTORIZABLE_P
6088 (vect_analyze_loop_1): Remove callback, get the mode iterator
6089 and autodetected_vector_mode as argument, advancing the
6090 iterator and initializing autodetected_vector_mode here.
6091 (vect_analyze_loop): Split analysis loop into two, first
6092 processing main loops only and then epilogues.
6094 2021-11-05 Martin Jambor <mjambor@suse.cz>
6096 * ipa-prop.c (compute_complex_assign_jump_func): Remove
6097 unnecessary check for RECORD_TYPE.
6099 2021-11-05 Gerald Pfeifer <gerald@pfeifer.com>
6101 * doc/install.texi2html: Do not generate old.html any longer.
6103 2021-11-05 Martin Liska <mliska@suse.cz>
6106 * opts.c (finish_options): Reset flag_gtoggle when it is used.
6108 2021-11-05 Jakub Jelinek <jakub@redhat.com>
6111 * dwarf2out.c (add_const_value_attribute): Add MODE argument, use it
6112 in CONST_WIDE_INT handling. Adjust recursive calls.
6113 (add_location_or_const_value_attribute): Pass DECL_MODE (decl) to
6114 new add_const_value_attribute argument.
6115 (tree_add_const_value_attribute): Pass TYPE_MODE (type) to new
6116 add_const_value_attribute argument.
6118 2021-11-05 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
6120 * config/vx-common.h: Test value of TARGET_VXWORKS7 rather
6123 2021-11-05 Richard Biener <rguenther@suse.de>
6125 * tree-vectorizer.h (struct vect_loop_form_info): New.
6126 (vect_analyze_loop_form): Adjust.
6127 (vect_create_loop_vinfo): New.
6128 * tree-parloops.c (gather_scalar_reductions): Adjust for
6129 vect_analyze_loop_form API change.
6130 * tree-vect-loop.c: Include <functional>.
6131 (vect_analyze_loop_form_1): Rename to vect_analyze_loop_form,
6132 take struct vect_loop_form_info as output parameter and adjust.
6133 (vect_analyze_loop_form): Rename to vect_create_loop_vinfo and
6134 split out call to the original vect_analyze_loop_form_1.
6135 (vect_reanalyze_as_main_loop): Rename to...
6136 (vect_analyze_loop_1): ... this, factor out the call to
6137 vect_analyze_loop_form and generalize to be able to use it twice ...
6138 (vect_analyze_loop): ... here. Perform vect_analyze_loop_form
6141 2021-11-05 Xionghu Luo <luoxhu@linux.ibm.com>
6144 * config/rs6000/fusion.md: Regenerate.
6145 * config/rs6000/genfusion.pl: Fix incorrect clobber constraint.
6147 2021-11-04 Andreas Krebbel <krebbel@linux.ibm.com>
6149 * config/s390/s390.h (STACK_CHECK_MOVING_SP): New macro
6152 2021-11-04 Tamar Christina <tamar.christina@arm.com>
6154 * config/aarch64/aarch64-builtins.c
6155 (aarch64_general_gimple_fold_builtin): Add ashl, sshl, ushl, ashr,
6156 ashr_simd, lshr, lshr_simd.
6157 * config/aarch64/aarch64-simd-builtins.def (lshr): Use USHIFTIMM.
6158 * config/aarch64/arm_neon.h (vshr_n_u8, vshr_n_u16, vshr_n_u32,
6159 vshrq_n_u8, vshrq_n_u16, vshrq_n_u32, vshrq_n_u64): Fix type hack.
6161 2021-11-04 Tamar Christina <tamar.christina@arm.com>
6163 * match.pd: New negate+shift pattern.
6165 2021-11-04 Andrew MacLeod <amacleod@redhat.com>
6167 PR tree-optimization/103079
6168 * gimple-range-gori.cc (gimple_range_calc_op1): Treat undefined as
6170 (gimple_range_calc_op2): Ditto.
6172 2021-11-04 Martin Jambor <mjambor@suse.cz>
6175 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
6176 members remap_with_debug_expressions, m_dead_ssa_debug_equiv,
6177 m_dead_stmt_debug_equiv and prepare_debug_expressions. Added
6178 parameter to mark_dead_statements.
6179 * ipa-param-manipulation.c: Include tree-phinodes.h and cfgexpand.h.
6180 (ipa_param_body_adjustments::mark_dead_statements): New parameter
6181 debugstack, push into it all SSA names used in debug statements,
6182 produce m_dead_ssa_debug_equiv mapping for the removed param.
6183 (replace_with_mapped_expr): New function.
6184 (ipa_param_body_adjustments::remap_with_debug_expressions): Likewise.
6185 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
6186 (ipa_param_body_adjustments::common_initialization): Gather and
6187 procecc SSA which will be removed but are in debug statements. Simplify.
6188 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
6190 * tree-inline.c (remap_gimple_stmt): Create a debug bind when possible
6191 when avoiding a copy of an unnecessary statement. Remap removed SSA
6192 names in existing debug statements.
6193 (tree_function_versioning): Do not create DEBUG_EXPR_DECL for removed
6194 parameters if we have already done so.
6196 2021-11-04 Jan Hubicka <hubicka@ucw.cz>
6199 * gimple.c (gimple_call_static_chain_flags): Handle case when
6200 nested function does not bind locally.
6202 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
6204 * config/aarch64/aarch64.c (aarch64_function_value): Generate
6205 a register rtx for Neon vector-tuple modes.
6206 (aarch64_layout_arg): Likewise.
6208 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
6210 * lower-subreg.c (simple_move): Prevent decomposition if
6211 modes are not tieable.
6213 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
6214 Richard Sandiford <richard.sandiford@arm.com>
6216 * config/aarch64/aarch64-builtins.c (v2x8qi_UP): Define.
6217 (v2x4hi_UP): Likewise.
6218 (v2x4hf_UP): Likewise.
6219 (v2x4bf_UP): Likewise.
6220 (v2x2si_UP): Likewise.
6221 (v2x2sf_UP): Likewise.
6222 (v2x1di_UP): Likewise.
6223 (v2x1df_UP): Likewise.
6224 (v2x16qi_UP): Likewise.
6225 (v2x8hi_UP): Likewise.
6226 (v2x8hf_UP): Likewise.
6227 (v2x8bf_UP): Likewise.
6228 (v2x4si_UP): Likewise.
6229 (v2x4sf_UP): Likewise.
6230 (v2x2di_UP): Likewise.
6231 (v2x2df_UP): Likewise.
6232 (v3x8qi_UP): Likewise.
6233 (v3x4hi_UP): Likewise.
6234 (v3x4hf_UP): Likewise.
6235 (v3x4bf_UP): Likewise.
6236 (v3x2si_UP): Likewise.
6237 (v3x2sf_UP): Likewise.
6238 (v3x1di_UP): Likewise.
6239 (v3x1df_UP): Likewise.
6240 (v3x16qi_UP): Likewise.
6241 (v3x8hi_UP): Likewise.
6242 (v3x8hf_UP): Likewise.
6243 (v3x8bf_UP): Likewise.
6244 (v3x4si_UP): Likewise.
6245 (v3x4sf_UP): Likewise.
6246 (v3x2di_UP): Likewise.
6247 (v3x2df_UP): Likewise.
6248 (v4x8qi_UP): Likewise.
6249 (v4x4hi_UP): Likewise.
6250 (v4x4hf_UP): Likewise.
6251 (v4x4bf_UP): Likewise.
6252 (v4x2si_UP): Likewise.
6253 (v4x2sf_UP): Likewise.
6254 (v4x1di_UP): Likewise.
6255 (v4x1df_UP): Likewise.
6256 (v4x16qi_UP): Likewise.
6257 (v4x8hi_UP): Likewise.
6258 (v4x8hf_UP): Likewise.
6259 (v4x8bf_UP): Likewise.
6260 (v4x4si_UP): Likewise.
6261 (v4x4sf_UP): Likewise.
6262 (v4x2di_UP): Likewise.
6263 (v4x2df_UP): Likewise.
6264 (TYPES_GETREGP): Delete.
6265 (TYPES_SETREGP): Likewise.
6266 (TYPES_LOADSTRUCT_U): Define.
6267 (TYPES_LOADSTRUCT_P): Likewise.
6268 (TYPES_LOADSTRUCT_LANE_U): Likewise.
6269 (TYPES_LOADSTRUCT_LANE_P): Likewise.
6270 (TYPES_STORE1P): Move for consistency.
6271 (TYPES_STORESTRUCT_U): Define.
6272 (TYPES_STORESTRUCT_P): Likewise.
6273 (TYPES_STORESTRUCT_LANE_U): Likewise.
6274 (TYPES_STORESTRUCT_LANE_P): Likewise.
6275 (aarch64_simd_tuple_types): Define.
6276 (aarch64_lookup_simd_builtin_type): Handle tuple type lookup.
6277 (aarch64_init_simd_builtin_functions): Update frontend lookup
6278 for builtin functions after handling arm_neon.h pragma.
6279 (register_tuple_type): Manually set modes of single-integer
6280 tuple types. Record tuple types.
6281 * config/aarch64/aarch64-modes.def
6282 (ADV_SIMD_D_REG_STRUCT_MODES): Define D-register tuple modes.
6283 (ADV_SIMD_Q_REG_STRUCT_MODES): Define Q-register tuple modes.
6284 (SVE_MODES): Give single-vector modes priority over vector-
6286 (VECTOR_MODES_WITH_PREFIX): Set partial-vector mode order to
6287 be after all single-vector modes.
6288 * config/aarch64/aarch64-simd-builtins.def: Update builtin
6289 generator macros to reflect modifications to the backend
6291 * config/aarch64/aarch64-simd.md (aarch64_simd_ld2<mode>):
6292 Use vector-tuple mode iterator and rename to...
6293 (aarch64_simd_ld2<vstruct_elt>): This.
6294 (aarch64_simd_ld2r<mode>): Use vector-tuple mode iterator and
6296 (aarch64_simd_ld2r<vstruct_elt>): This.
6297 (aarch64_vec_load_lanesoi_lane<mode>): Use vector-tuple mode
6298 iterator and rename to...
6299 (aarch64_vec_load_lanes<mode>_lane<vstruct_elt>): This.
6300 (vec_load_lanesoi<mode>): Use vector-tuple mode iterator and
6302 (vec_load_lanes<mode><vstruct_elt>): This.
6303 (aarch64_simd_st2<mode>): Use vector-tuple mode iterator and
6305 (aarch64_simd_st2<vstruct_elt>): This.
6306 (aarch64_vec_store_lanesoi_lane<mode>): Use vector-tuple mode
6307 iterator and rename to...
6308 (aarch64_vec_store_lanes<mode>_lane<vstruct_elt>): This.
6309 (vec_store_lanesoi<mode>): Use vector-tuple mode iterator and
6311 (vec_store_lanes<mode><vstruct_elt>): This.
6312 (aarch64_simd_ld3<mode>): Use vector-tuple mode iterator and
6314 (aarch64_simd_ld3<vstruct_elt>): This.
6315 (aarch64_simd_ld3r<mode>): Use vector-tuple mode iterator and
6317 (aarch64_simd_ld3r<vstruct_elt>): This.
6318 (aarch64_vec_load_lanesci_lane<mode>): Use vector-tuple mode
6319 iterator and rename to...
6320 (vec_load_lanesci<mode>): This.
6321 (aarch64_simd_st3<mode>): Use vector-tuple mode iterator and
6323 (aarch64_simd_st3<vstruct_elt>): This.
6324 (aarch64_vec_store_lanesci_lane<mode>): Use vector-tuple mode
6325 iterator and rename to...
6326 (vec_store_lanesci<mode>): This.
6327 (aarch64_simd_ld4<mode>): Use vector-tuple mode iterator and
6329 (aarch64_simd_ld4<vstruct_elt>): This.
6330 (aarch64_simd_ld4r<mode>): Use vector-tuple mode iterator and
6332 (aarch64_simd_ld4r<vstruct_elt>): This.
6333 (aarch64_vec_load_lanesxi_lane<mode>): Use vector-tuple mode
6334 iterator and rename to...
6335 (vec_load_lanesxi<mode>): This.
6336 (aarch64_simd_st4<mode>): Use vector-tuple mode iterator and
6338 (aarch64_simd_st4<vstruct_elt>): This.
6339 (aarch64_vec_store_lanesxi_lane<mode>): Use vector-tuple mode
6340 iterator and rename to...
6341 (vec_store_lanesxi<mode>): This.
6342 (mov<mode>): Define for Neon vector-tuple modes.
6343 (aarch64_ld1x3<VALLDIF:mode>): Use vector-tuple mode iterator
6345 (aarch64_ld1x3<vstruct_elt>): This.
6346 (aarch64_ld1_x3_<mode>): Use vector-tuple mode iterator and
6348 (aarch64_ld1_x3_<vstruct_elt>): This.
6349 (aarch64_ld1x4<VALLDIF:mode>): Use vector-tuple mode iterator
6351 (aarch64_ld1x4<vstruct_elt>): This.
6352 (aarch64_ld1_x4_<mode>): Use vector-tuple mode iterator and
6354 (aarch64_ld1_x4_<vstruct_elt>): This.
6355 (aarch64_st1x2<VALLDIF:mode>): Use vector-tuple mode iterator
6357 (aarch64_st1x2<vstruct_elt>): This.
6358 (aarch64_st1_x2_<mode>): Use vector-tuple mode iterator and
6360 (aarch64_st1_x2_<vstruct_elt>): This.
6361 (aarch64_st1x3<VALLDIF:mode>): Use vector-tuple mode iterator
6363 (aarch64_st1x3<vstruct_elt>): This.
6364 (aarch64_st1_x3_<mode>): Use vector-tuple mode iterator and
6366 (aarch64_st1_x3_<vstruct_elt>): This.
6367 (aarch64_st1x4<VALLDIF:mode>): Use vector-tuple mode iterator
6369 (aarch64_st1x4<vstruct_elt>): This.
6370 (aarch64_st1_x4_<mode>): Use vector-tuple mode iterator and
6372 (aarch64_st1_x4_<vstruct_elt>): This.
6373 (*aarch64_mov<mode>): Define for vector-tuple modes.
6374 (*aarch64_be_mov<mode>): Likewise.
6375 (aarch64_ld<VSTRUCT:nregs>r<VALLDIF:mode>): Use vector-tuple
6376 mode iterator and rename to...
6377 (aarch64_ld<nregs>r<vstruct_elt>): This.
6378 (aarch64_ld2<mode>_dreg): Use vector-tuple mode iterator and
6380 (aarch64_ld2<vstruct_elt>_dreg): This.
6381 (aarch64_ld3<mode>_dreg): Use vector-tuple mode iterator and
6383 (aarch64_ld3<vstruct_elt>_dreg): This.
6384 (aarch64_ld4<mode>_dreg): Use vector-tuple mode iterator and
6386 (aarch64_ld4<vstruct_elt>_dreg): This.
6387 (aarch64_ld<VSTRUCT:nregs><VDC:mode>): Use vector-tuple mode
6388 iterator and rename to...
6389 (aarch64_ld<nregs><vstruct_elt>): Use vector-tuple mode
6390 iterator and rename to...
6391 (aarch64_ld<VSTRUCT:nregs><VQ:mode>): Use vector-tuple mode
6392 (aarch64_ld1x2<VQ:mode>): Delete.
6393 (aarch64_ld1x2<VDC:mode>): Use vector-tuple mode iterator and
6395 (aarch64_ld1x2<vstruct_elt>): This.
6396 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use vector-
6397 tuple mode iterator and rename to...
6398 (aarch64_ld<nregs>_lane<vstruct_elt>): This.
6399 (aarch64_get_dreg<VSTRUCT:mode><VDC:mode>): Delete.
6400 (aarch64_get_qreg<VSTRUCT:mode><VQ:mode>): Likewise.
6401 (aarch64_st2<mode>_dreg): Use vector-tuple mode iterator and
6403 (aarch64_st2<vstruct_elt>_dreg): This.
6404 (aarch64_st3<mode>_dreg): Use vector-tuple mode iterator and
6406 (aarch64_st3<vstruct_elt>_dreg): This.
6407 (aarch64_st4<mode>_dreg): Use vector-tuple mode iterator and
6409 (aarch64_st4<vstruct_elt>_dreg): This.
6410 (aarch64_st<VSTRUCT:nregs><VDC:mode>): Use vector-tuple mode
6411 iterator and rename to...
6412 (aarch64_st<nregs><vstruct_elt>): This.
6413 (aarch64_st<VSTRUCT:nregs><VQ:mode>): Use vector-tuple mode
6414 iterator and rename to aarch64_st<nregs><vstruct_elt>.
6415 (aarch64_st<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use vector-
6416 tuple mode iterator and rename to...
6417 (aarch64_st<nregs>_lane<vstruct_elt>): This.
6418 (aarch64_set_qreg<VSTRUCT:mode><VQ:mode>): Delete.
6419 (aarch64_simd_ld1<mode>_x2): Use vector-tuple mode iterator
6421 (aarch64_simd_ld1<vstruct_elt>_x2): This.
6422 * config/aarch64/aarch64.c (aarch64_advsimd_struct_mode_p):
6423 Refactor to include new vector-tuple modes.
6424 (aarch64_classify_vector_mode): Add cases for new vector-
6426 (aarch64_advsimd_partial_struct_mode_p): Define.
6427 (aarch64_advsimd_full_struct_mode_p): Likewise.
6428 (aarch64_advsimd_vector_array_mode): Likewise.
6429 (aarch64_sve_data_mode): Change location in file.
6430 (aarch64_array_mode): Handle case of Neon vector-tuple modes.
6431 (aarch64_hard_regno_nregs): Handle case of partial Neon
6433 (aarch64_classify_address): Refactor to include handling of
6434 Neon vector-tuple modes.
6435 (aarch64_print_operand): Print "d" for "%R" for a partial
6436 Neon vector structure.
6437 (aarch64_expand_vec_perm_1): Use new vector-tuple mode.
6438 (aarch64_modes_tieable_p): Prevent tieing Neon partial struct
6439 modes with scalar machines modes larger than 8 bytes.
6440 (aarch64_can_change_mode_class): Don't allow changes between
6441 partial and full Neon vector-structure modes.
6442 * config/aarch64/arm_neon.h (vst2_lane_f16): Use updated
6443 builtin and remove boiler-plate code for opaque mode.
6444 (vst2_lane_f32): Likewise.
6445 (vst2_lane_f64): Likewise.
6446 (vst2_lane_p8): Likewise.
6447 (vst2_lane_p16): Likewise.
6448 (vst2_lane_p64): Likewise.
6449 (vst2_lane_s8): Likewise.
6450 (vst2_lane_s16): Likewise.
6451 (vst2_lane_s32): Likewise.
6452 (vst2_lane_s64): Likewise.
6453 (vst2_lane_u8): Likewise.
6454 (vst2_lane_u16): Likewise.
6455 (vst2_lane_u32): Likewise.
6456 (vst2_lane_u64): Likewise.
6457 (vst2q_lane_f16): Likewise.
6458 (vst2q_lane_f32): Likewise.
6459 (vst2q_lane_f64): Likewise.
6460 (vst2q_lane_p8): Likewise.
6461 (vst2q_lane_p16): Likewise.
6462 (vst2q_lane_p64): Likewise.
6463 (vst2q_lane_s8): Likewise.
6464 (vst2q_lane_s16): Likewise.
6465 (vst2q_lane_s32): Likewise.
6466 (vst2q_lane_s64): Likewise.
6467 (vst2q_lane_u8): Likewise.
6468 (vst2q_lane_u16): Likewise.
6469 (vst2q_lane_u32): Likewise.
6470 (vst2q_lane_u64): Likewise.
6471 (vst3_lane_f16): Likewise.
6472 (vst3_lane_f32): Likewise.
6473 (vst3_lane_f64): Likewise.
6474 (vst3_lane_p8): Likewise.
6475 (vst3_lane_p16): Likewise.
6476 (vst3_lane_p64): Likewise.
6477 (vst3_lane_s8): Likewise.
6478 (vst3_lane_s16): Likewise.
6479 (vst3_lane_s32): Likewise.
6480 (vst3_lane_s64): Likewise.
6481 (vst3_lane_u8): Likewise.
6482 (vst3_lane_u16): Likewise.
6483 (vst3_lane_u32): Likewise.
6484 (vst3_lane_u64): Likewise.
6485 (vst3q_lane_f16): Likewise.
6486 (vst3q_lane_f32): Likewise.
6487 (vst3q_lane_f64): Likewise.
6488 (vst3q_lane_p8): Likewise.
6489 (vst3q_lane_p16): Likewise.
6490 (vst3q_lane_p64): Likewise.
6491 (vst3q_lane_s8): Likewise.
6492 (vst3q_lane_s16): Likewise.
6493 (vst3q_lane_s32): Likewise.
6494 (vst3q_lane_s64): Likewise.
6495 (vst3q_lane_u8): Likewise.
6496 (vst3q_lane_u16): Likewise.
6497 (vst3q_lane_u32): Likewise.
6498 (vst3q_lane_u64): Likewise.
6499 (vst4_lane_f16): Likewise.
6500 (vst4_lane_f32): Likewise.
6501 (vst4_lane_f64): Likewise.
6502 (vst4_lane_p8): Likewise.
6503 (vst4_lane_p16): Likewise.
6504 (vst4_lane_p64): Likewise.
6505 (vst4_lane_s8): Likewise.
6506 (vst4_lane_s16): Likewise.
6507 (vst4_lane_s32): Likewise.
6508 (vst4_lane_s64): Likewise.
6509 (vst4_lane_u8): Likewise.
6510 (vst4_lane_u16): Likewise.
6511 (vst4_lane_u32): Likewise.
6512 (vst4_lane_u64): Likewise.
6513 (vst4q_lane_f16): Likewise.
6514 (vst4q_lane_f32): Likewise.
6515 (vst4q_lane_f64): Likewise.
6516 (vst4q_lane_p8): Likewise.
6517 (vst4q_lane_p16): Likewise.
6518 (vst4q_lane_p64): Likewise.
6519 (vst4q_lane_s8): Likewise.
6520 (vst4q_lane_s16): Likewise.
6521 (vst4q_lane_s32): Likewise.
6522 (vst4q_lane_s64): Likewise.
6523 (vst4q_lane_u8): Likewise.
6524 (vst4q_lane_u16): Likewise.
6525 (vst4q_lane_u32): Likewise.
6526 (vst4q_lane_u64): Likewise.
6527 (vtbl3_s8): Likewise.
6528 (vtbl3_u8): Likewise.
6529 (vtbl3_p8): Likewise.
6530 (vtbl4_s8): Likewise.
6531 (vtbl4_u8): Likewise.
6532 (vtbl4_p8): Likewise.
6533 (vld1_u8_x3): Likewise.
6534 (vld1_s8_x3): Likewise.
6535 (vld1_u16_x3): Likewise.
6536 (vld1_s16_x3): Likewise.
6537 (vld1_u32_x3): Likewise.
6538 (vld1_s32_x3): Likewise.
6539 (vld1_u64_x3): Likewise.
6540 (vld1_s64_x3): Likewise.
6541 (vld1_f16_x3): Likewise.
6542 (vld1_f32_x3): Likewise.
6543 (vld1_f64_x3): Likewise.
6544 (vld1_p8_x3): Likewise.
6545 (vld1_p16_x3): Likewise.
6546 (vld1_p64_x3): Likewise.
6547 (vld1q_u8_x3): Likewise.
6548 (vld1q_s8_x3): Likewise.
6549 (vld1q_u16_x3): Likewise.
6550 (vld1q_s16_x3): Likewise.
6551 (vld1q_u32_x3): Likewise.
6552 (vld1q_s32_x3): Likewise.
6553 (vld1q_u64_x3): Likewise.
6554 (vld1q_s64_x3): Likewise.
6555 (vld1q_f16_x3): Likewise.
6556 (vld1q_f32_x3): Likewise.
6557 (vld1q_f64_x3): Likewise.
6558 (vld1q_p8_x3): Likewise.
6559 (vld1q_p16_x3): Likewise.
6560 (vld1q_p64_x3): Likewise.
6561 (vld1_u8_x2): Likewise.
6562 (vld1_s8_x2): Likewise.
6563 (vld1_u16_x2): Likewise.
6564 (vld1_s16_x2): Likewise.
6565 (vld1_u32_x2): Likewise.
6566 (vld1_s32_x2): Likewise.
6567 (vld1_u64_x2): Likewise.
6568 (vld1_s64_x2): Likewise.
6569 (vld1_f16_x2): Likewise.
6570 (vld1_f32_x2): Likewise.
6571 (vld1_f64_x2): Likewise.
6572 (vld1_p8_x2): Likewise.
6573 (vld1_p16_x2): Likewise.
6574 (vld1_p64_x2): Likewise.
6575 (vld1q_u8_x2): Likewise.
6576 (vld1q_s8_x2): Likewise.
6577 (vld1q_u16_x2): Likewise.
6578 (vld1q_s16_x2): Likewise.
6579 (vld1q_u32_x2): Likewise.
6580 (vld1q_s32_x2): Likewise.
6581 (vld1q_u64_x2): Likewise.
6582 (vld1q_s64_x2): Likewise.
6583 (vld1q_f16_x2): Likewise.
6584 (vld1q_f32_x2): Likewise.
6585 (vld1q_f64_x2): Likewise.
6586 (vld1q_p8_x2): Likewise.
6587 (vld1q_p16_x2): Likewise.
6588 (vld1q_p64_x2): Likewise.
6589 (vld1_s8_x4): Likewise.
6590 (vld1q_s8_x4): Likewise.
6591 (vld1_s16_x4): Likewise.
6592 (vld1q_s16_x4): Likewise.
6593 (vld1_s32_x4): Likewise.
6594 (vld1q_s32_x4): Likewise.
6595 (vld1_u8_x4): Likewise.
6596 (vld1q_u8_x4): Likewise.
6597 (vld1_u16_x4): Likewise.
6598 (vld1q_u16_x4): Likewise.
6599 (vld1_u32_x4): Likewise.
6600 (vld1q_u32_x4): Likewise.
6601 (vld1_f16_x4): Likewise.
6602 (vld1q_f16_x4): Likewise.
6603 (vld1_f32_x4): Likewise.
6604 (vld1q_f32_x4): Likewise.
6605 (vld1_p8_x4): Likewise.
6606 (vld1q_p8_x4): Likewise.
6607 (vld1_p16_x4): Likewise.
6608 (vld1q_p16_x4): Likewise.
6609 (vld1_s64_x4): Likewise.
6610 (vld1_u64_x4): Likewise.
6611 (vld1_p64_x4): Likewise.
6612 (vld1q_s64_x4): Likewise.
6613 (vld1q_u64_x4): Likewise.
6614 (vld1q_p64_x4): Likewise.
6615 (vld1_f64_x4): Likewise.
6616 (vld1q_f64_x4): Likewise.
6617 (vld2_s64): Likewise.
6618 (vld2_u64): Likewise.
6619 (vld2_f64): Likewise.
6620 (vld2_s8): Likewise.
6621 (vld2_p8): Likewise.
6622 (vld2_p64): Likewise.
6623 (vld2_s16): Likewise.
6624 (vld2_p16): Likewise.
6625 (vld2_s32): Likewise.
6626 (vld2_u8): Likewise.
6627 (vld2_u16): Likewise.
6628 (vld2_u32): Likewise.
6629 (vld2_f16): Likewise.
6630 (vld2_f32): Likewise.
6631 (vld2q_s8): Likewise.
6632 (vld2q_p8): Likewise.
6633 (vld2q_s16): Likewise.
6634 (vld2q_p16): Likewise.
6635 (vld2q_p64): Likewise.
6636 (vld2q_s32): Likewise.
6637 (vld2q_s64): Likewise.
6638 (vld2q_u8): Likewise.
6639 (vld2q_u16): Likewise.
6640 (vld2q_u32): Likewise.
6641 (vld2q_u64): Likewise.
6642 (vld2q_f16): Likewise.
6643 (vld2q_f32): Likewise.
6644 (vld2q_f64): Likewise.
6645 (vld3_s64): Likewise.
6646 (vld3_u64): Likewise.
6647 (vld3_f64): Likewise.
6648 (vld3_s8): Likewise.
6649 (vld3_p8): Likewise.
6650 (vld3_s16): Likewise.
6651 (vld3_p16): Likewise.
6652 (vld3_s32): Likewise.
6653 (vld3_u8): Likewise.
6654 (vld3_u16): Likewise.
6655 (vld3_u32): Likewise.
6656 (vld3_f16): Likewise.
6657 (vld3_f32): Likewise.
6658 (vld3_p64): Likewise.
6659 (vld3q_s8): Likewise.
6660 (vld3q_p8): Likewise.
6661 (vld3q_s16): Likewise.
6662 (vld3q_p16): Likewise.
6663 (vld3q_s32): Likewise.
6664 (vld3q_s64): Likewise.
6665 (vld3q_u8): Likewise.
6666 (vld3q_u16): Likewise.
6667 (vld3q_u32): Likewise.
6668 (vld3q_u64): Likewise.
6669 (vld3q_f16): Likewise.
6670 (vld3q_f32): Likewise.
6671 (vld3q_f64): Likewise.
6672 (vld3q_p64): Likewise.
6673 (vld4_s64): Likewise.
6674 (vld4_u64): Likewise.
6675 (vld4_f64): Likewise.
6676 (vld4_s8): Likewise.
6677 (vld4_p8): Likewise.
6678 (vld4_s16): Likewise.
6679 (vld4_p16): Likewise.
6680 (vld4_s32): Likewise.
6681 (vld4_u8): Likewise.
6682 (vld4_u16): Likewise.
6683 (vld4_u32): Likewise.
6684 (vld4_f16): Likewise.
6685 (vld4_f32): Likewise.
6686 (vld4_p64): Likewise.
6687 (vld4q_s8): Likewise.
6688 (vld4q_p8): Likewise.
6689 (vld4q_s16): Likewise.
6690 (vld4q_p16): Likewise.
6691 (vld4q_s32): Likewise.
6692 (vld4q_s64): Likewise.
6693 (vld4q_u8): Likewise.
6694 (vld4q_u16): Likewise.
6695 (vld4q_u32): Likewise.
6696 (vld4q_u64): Likewise.
6697 (vld4q_f16): Likewise.
6698 (vld4q_f32): Likewise.
6699 (vld4q_f64): Likewise.
6700 (vld4q_p64): Likewise.
6701 (vld2_dup_s8): Likewise.
6702 (vld2_dup_s16): Likewise.
6703 (vld2_dup_s32): Likewise.
6704 (vld2_dup_f16): Likewise.
6705 (vld2_dup_f32): Likewise.
6706 (vld2_dup_f64): Likewise.
6707 (vld2_dup_u8): Likewise.
6708 (vld2_dup_u16): Likewise.
6709 (vld2_dup_u32): Likewise.
6710 (vld2_dup_p8): Likewise.
6711 (vld2_dup_p16): Likewise.
6712 (vld2_dup_p64): Likewise.
6713 (vld2_dup_s64): Likewise.
6714 (vld2_dup_u64): Likewise.
6715 (vld2q_dup_s8): Likewise.
6716 (vld2q_dup_p8): Likewise.
6717 (vld2q_dup_s16): Likewise.
6718 (vld2q_dup_p16): Likewise.
6719 (vld2q_dup_s32): Likewise.
6720 (vld2q_dup_s64): Likewise.
6721 (vld2q_dup_u8): Likewise.
6722 (vld2q_dup_u16): Likewise.
6723 (vld2q_dup_u32): Likewise.
6724 (vld2q_dup_u64): Likewise.
6725 (vld2q_dup_f16): Likewise.
6726 (vld2q_dup_f32): Likewise.
6727 (vld2q_dup_f64): Likewise.
6728 (vld2q_dup_p64): Likewise.
6729 (vld3_dup_s64): Likewise.
6730 (vld3_dup_u64): Likewise.
6731 (vld3_dup_f64): Likewise.
6732 (vld3_dup_s8): Likewise.
6733 (vld3_dup_p8): Likewise.
6734 (vld3_dup_s16): Likewise.
6735 (vld3_dup_p16): Likewise.
6736 (vld3_dup_s32): Likewise.
6737 (vld3_dup_u8): Likewise.
6738 (vld3_dup_u16): Likewise.
6739 (vld3_dup_u32): Likewise.
6740 (vld3_dup_f16): Likewise.
6741 (vld3_dup_f32): Likewise.
6742 (vld3_dup_p64): Likewise.
6743 (vld3q_dup_s8): Likewise.
6744 (vld3q_dup_p8): Likewise.
6745 (vld3q_dup_s16): Likewise.
6746 (vld3q_dup_p16): Likewise.
6747 (vld3q_dup_s32): Likewise.
6748 (vld3q_dup_s64): Likewise.
6749 (vld3q_dup_u8): Likewise.
6750 (vld3q_dup_u16): Likewise.
6751 (vld3q_dup_u32): Likewise.
6752 (vld3q_dup_u64): Likewise.
6753 (vld3q_dup_f16): Likewise.
6754 (vld3q_dup_f32): Likewise.
6755 (vld3q_dup_f64): Likewise.
6756 (vld3q_dup_p64): Likewise.
6757 (vld4_dup_s64): Likewise.
6758 (vld4_dup_u64): Likewise.
6759 (vld4_dup_f64): Likewise.
6760 (vld4_dup_s8): Likewise.
6761 (vld4_dup_p8): Likewise.
6762 (vld4_dup_s16): Likewise.
6763 (vld4_dup_p16): Likewise.
6764 (vld4_dup_s32): Likewise.
6765 (vld4_dup_u8): Likewise.
6766 (vld4_dup_u16): Likewise.
6767 (vld4_dup_u32): Likewise.
6768 (vld4_dup_f16): Likewise.
6769 (vld4_dup_f32): Likewise.
6770 (vld4_dup_p64): Likewise.
6771 (vld4q_dup_s8): Likewise.
6772 (vld4q_dup_p8): Likewise.
6773 (vld4q_dup_s16): Likewise.
6774 (vld4q_dup_p16): Likewise.
6775 (vld4q_dup_s32): Likewise.
6776 (vld4q_dup_s64): Likewise.
6777 (vld4q_dup_u8): Likewise.
6778 (vld4q_dup_u16): Likewise.
6779 (vld4q_dup_u32): Likewise.
6780 (vld4q_dup_u64): Likewise.
6781 (vld4q_dup_f16): Likewise.
6782 (vld4q_dup_f32): Likewise.
6783 (vld4q_dup_f64): Likewise.
6784 (vld4q_dup_p64): Likewise.
6785 (vld2_lane_u8): Likewise.
6786 (vld2_lane_u16): Likewise.
6787 (vld2_lane_u32): Likewise.
6788 (vld2_lane_u64): Likewise.
6789 (vld2_lane_s8): Likewise.
6790 (vld2_lane_s16): Likewise.
6791 (vld2_lane_s32): Likewise.
6792 (vld2_lane_s64): Likewise.
6793 (vld2_lane_f16): Likewise.
6794 (vld2_lane_f32): Likewise.
6795 (vld2_lane_f64): Likewise.
6796 (vld2_lane_p8): Likewise.
6797 (vld2_lane_p16): Likewise.
6798 (vld2_lane_p64): Likewise.
6799 (vld2q_lane_u8): Likewise.
6800 (vld2q_lane_u16): Likewise.
6801 (vld2q_lane_u32): Likewise.
6802 (vld2q_lane_u64): Likewise.
6803 (vld2q_lane_s8): Likewise.
6804 (vld2q_lane_s16): Likewise.
6805 (vld2q_lane_s32): Likewise.
6806 (vld2q_lane_s64): Likewise.
6807 (vld2q_lane_f16): Likewise.
6808 (vld2q_lane_f32): Likewise.
6809 (vld2q_lane_f64): Likewise.
6810 (vld2q_lane_p8): Likewise.
6811 (vld2q_lane_p16): Likewise.
6812 (vld2q_lane_p64): Likewise.
6813 (vld3_lane_u8): Likewise.
6814 (vld3_lane_u16): Likewise.
6815 (vld3_lane_u32): Likewise.
6816 (vld3_lane_u64): Likewise.
6817 (vld3_lane_s8): Likewise.
6818 (vld3_lane_s16): Likewise.
6819 (vld3_lane_s32): Likewise.
6820 (vld3_lane_s64): Likewise.
6821 (vld3_lane_f16): Likewise.
6822 (vld3_lane_f32): Likewise.
6823 (vld3_lane_f64): Likewise.
6824 (vld3_lane_p8): Likewise.
6825 (vld3_lane_p16): Likewise.
6826 (vld3_lane_p64): Likewise.
6827 (vld3q_lane_u8): Likewise.
6828 (vld3q_lane_u16): Likewise.
6829 (vld3q_lane_u32): Likewise.
6830 (vld3q_lane_u64): Likewise.
6831 (vld3q_lane_s8): Likewise.
6832 (vld3q_lane_s16): Likewise.
6833 (vld3q_lane_s32): Likewise.
6834 (vld3q_lane_s64): Likewise.
6835 (vld3q_lane_f16): Likewise.
6836 (vld3q_lane_f32): Likewise.
6837 (vld3q_lane_f64): Likewise.
6838 (vld3q_lane_p8): Likewise.
6839 (vld3q_lane_p16): Likewise.
6840 (vld3q_lane_p64): Likewise.
6841 (vld4_lane_u8): Likewise.
6842 (vld4_lane_u16): Likewise.
6843 (vld4_lane_u32): Likewise.
6844 (vld4_lane_u64): Likewise.
6845 (vld4_lane_s8): Likewise.
6846 (vld4_lane_s16): Likewise.
6847 (vld4_lane_s32): Likewise.
6848 (vld4_lane_s64): Likewise.
6849 (vld4_lane_f16): Likewise.
6850 (vld4_lane_f32): Likewise.
6851 (vld4_lane_f64): Likewise.
6852 (vld4_lane_p8): Likewise.
6853 (vld4_lane_p16): Likewise.
6854 (vld4_lane_p64): Likewise.
6855 (vld4q_lane_u8): Likewise.
6856 (vld4q_lane_u16): Likewise.
6857 (vld4q_lane_u32): Likewise.
6858 (vld4q_lane_u64): Likewise.
6859 (vld4q_lane_s8): Likewise.
6860 (vld4q_lane_s16): Likewise.
6861 (vld4q_lane_s32): Likewise.
6862 (vld4q_lane_s64): Likewise.
6863 (vld4q_lane_f16): Likewise.
6864 (vld4q_lane_f32): Likewise.
6865 (vld4q_lane_f64): Likewise.
6866 (vld4q_lane_p8): Likewise.
6867 (vld4q_lane_p16): Likewise.
6868 (vld4q_lane_p64): Likewise.
6869 (vqtbl2_s8): Likewise.
6870 (vqtbl2_u8): Likewise.
6871 (vqtbl2_p8): Likewise.
6872 (vqtbl2q_s8): Likewise.
6873 (vqtbl2q_u8): Likewise.
6874 (vqtbl2q_p8): Likewise.
6875 (vqtbl3_s8): Likewise.
6876 (vqtbl3_u8): Likewise.
6877 (vqtbl3_p8): Likewise.
6878 (vqtbl3q_s8): Likewise.
6879 (vqtbl3q_u8): Likewise.
6880 (vqtbl3q_p8): Likewise.
6881 (vqtbl4_s8): Likewise.
6882 (vqtbl4_u8): Likewise.
6883 (vqtbl4_p8): Likewise.
6884 (vqtbl4q_s8): Likewise.
6885 (vqtbl4q_u8): Likewise.
6886 (vqtbl4q_p8): Likewise.
6887 (vqtbx2_s8): Likewise.
6888 (vqtbx2_u8): Likewise.
6889 (vqtbx2_p8): Likewise.
6890 (vqtbx2q_s8): Likewise.
6891 (vqtbx2q_u8): Likewise.
6892 (vqtbx2q_p8): Likewise.
6893 (vqtbx3_s8): Likewise.
6894 (vqtbx3_u8): Likewise.
6895 (vqtbx3_p8): Likewise.
6896 (vqtbx3q_s8): Likewise.
6897 (vqtbx3q_u8): Likewise.
6898 (vqtbx3q_p8): Likewise.
6899 (vqtbx4_s8): Likewise.
6900 (vqtbx4_u8): Likewise.
6901 (vqtbx4_p8): Likewise.
6902 (vqtbx4q_s8): Likewise.
6903 (vqtbx4q_u8): Likewise.
6904 (vqtbx4q_p8): Likewise.
6905 (vst1_s64_x2): Likewise.
6906 (vst1_u64_x2): Likewise.
6907 (vst1_f64_x2): Likewise.
6908 (vst1_s8_x2): Likewise.
6909 (vst1_p8_x2): Likewise.
6910 (vst1_s16_x2): Likewise.
6911 (vst1_p16_x2): Likewise.
6912 (vst1_s32_x2): Likewise.
6913 (vst1_u8_x2): Likewise.
6914 (vst1_u16_x2): Likewise.
6915 (vst1_u32_x2): Likewise.
6916 (vst1_f16_x2): Likewise.
6917 (vst1_f32_x2): Likewise.
6918 (vst1_p64_x2): Likewise.
6919 (vst1q_s8_x2): Likewise.
6920 (vst1q_p8_x2): Likewise.
6921 (vst1q_s16_x2): Likewise.
6922 (vst1q_p16_x2): Likewise.
6923 (vst1q_s32_x2): Likewise.
6924 (vst1q_s64_x2): Likewise.
6925 (vst1q_u8_x2): Likewise.
6926 (vst1q_u16_x2): Likewise.
6927 (vst1q_u32_x2): Likewise.
6928 (vst1q_u64_x2): Likewise.
6929 (vst1q_f16_x2): Likewise.
6930 (vst1q_f32_x2): Likewise.
6931 (vst1q_f64_x2): Likewise.
6932 (vst1q_p64_x2): Likewise.
6933 (vst1_s64_x3): Likewise.
6934 (vst1_u64_x3): Likewise.
6935 (vst1_f64_x3): Likewise.
6936 (vst1_s8_x3): Likewise.
6937 (vst1_p8_x3): Likewise.
6938 (vst1_s16_x3): Likewise.
6939 (vst1_p16_x3): Likewise.
6940 (vst1_s32_x3): Likewise.
6941 (vst1_u8_x3): Likewise.
6942 (vst1_u16_x3): Likewise.
6943 (vst1_u32_x3): Likewise.
6944 (vst1_f16_x3): Likewise.
6945 (vst1_f32_x3): Likewise.
6946 (vst1_p64_x3): Likewise.
6947 (vst1q_s8_x3): Likewise.
6948 (vst1q_p8_x3): Likewise.
6949 (vst1q_s16_x3): Likewise.
6950 (vst1q_p16_x3): Likewise.
6951 (vst1q_s32_x3): Likewise.
6952 (vst1q_s64_x3): Likewise.
6953 (vst1q_u8_x3): Likewise.
6954 (vst1q_u16_x3): Likewise.
6955 (vst1q_u32_x3): Likewise.
6956 (vst1q_u64_x3): Likewise.
6957 (vst1q_f16_x3): Likewise.
6958 (vst1q_f32_x3): Likewise.
6959 (vst1q_f64_x3): Likewise.
6960 (vst1q_p64_x3): Likewise.
6961 (vst1_s8_x4): Likewise.
6962 (vst1q_s8_x4): Likewise.
6963 (vst1_s16_x4): Likewise.
6964 (vst1q_s16_x4): Likewise.
6965 (vst1_s32_x4): Likewise.
6966 (vst1q_s32_x4): Likewise.
6967 (vst1_u8_x4): Likewise.
6968 (vst1q_u8_x4): Likewise.
6969 (vst1_u16_x4): Likewise.
6970 (vst1q_u16_x4): Likewise.
6971 (vst1_u32_x4): Likewise.
6972 (vst1q_u32_x4): Likewise.
6973 (vst1_f16_x4): Likewise.
6974 (vst1q_f16_x4): Likewise.
6975 (vst1_f32_x4): Likewise.
6976 (vst1q_f32_x4): Likewise.
6977 (vst1_p8_x4): Likewise.
6978 (vst1q_p8_x4): Likewise.
6979 (vst1_p16_x4): Likewise.
6980 (vst1q_p16_x4): Likewise.
6981 (vst1_s64_x4): Likewise.
6982 (vst1_u64_x4): Likewise.
6983 (vst1_p64_x4): Likewise.
6984 (vst1q_s64_x4): Likewise.
6985 (vst1q_u64_x4): Likewise.
6986 (vst1q_p64_x4): Likewise.
6987 (vst1_f64_x4): Likewise.
6988 (vst1q_f64_x4): Likewise.
6989 (vst2_s64): Likewise.
6990 (vst2_u64): Likewise.
6991 (vst2_f64): Likewise.
6992 (vst2_s8): Likewise.
6993 (vst2_p8): Likewise.
6994 (vst2_s16): Likewise.
6995 (vst2_p16): Likewise.
6996 (vst2_s32): Likewise.
6997 (vst2_u8): Likewise.
6998 (vst2_u16): Likewise.
6999 (vst2_u32): Likewise.
7000 (vst2_f16): Likewise.
7001 (vst2_f32): Likewise.
7002 (vst2_p64): Likewise.
7003 (vst2q_s8): Likewise.
7004 (vst2q_p8): Likewise.
7005 (vst2q_s16): Likewise.
7006 (vst2q_p16): Likewise.
7007 (vst2q_s32): Likewise.
7008 (vst2q_s64): Likewise.
7009 (vst2q_u8): Likewise.
7010 (vst2q_u16): Likewise.
7011 (vst2q_u32): Likewise.
7012 (vst2q_u64): Likewise.
7013 (vst2q_f16): Likewise.
7014 (vst2q_f32): Likewise.
7015 (vst2q_f64): Likewise.
7016 (vst2q_p64): Likewise.
7017 (vst3_s64): Likewise.
7018 (vst3_u64): Likewise.
7019 (vst3_f64): Likewise.
7020 (vst3_s8): Likewise.
7021 (vst3_p8): Likewise.
7022 (vst3_s16): Likewise.
7023 (vst3_p16): Likewise.
7024 (vst3_s32): Likewise.
7025 (vst3_u8): Likewise.
7026 (vst3_u16): Likewise.
7027 (vst3_u32): Likewise.
7028 (vst3_f16): Likewise.
7029 (vst3_f32): Likewise.
7030 (vst3_p64): Likewise.
7031 (vst3q_s8): Likewise.
7032 (vst3q_p8): Likewise.
7033 (vst3q_s16): Likewise.
7034 (vst3q_p16): Likewise.
7035 (vst3q_s32): Likewise.
7036 (vst3q_s64): Likewise.
7037 (vst3q_u8): Likewise.
7038 (vst3q_u16): Likewise.
7039 (vst3q_u32): Likewise.
7040 (vst3q_u64): Likewise.
7041 (vst3q_f16): Likewise.
7042 (vst3q_f32): Likewise.
7043 (vst3q_f64): Likewise.
7044 (vst3q_p64): Likewise.
7045 (vst4_s64): Likewise.
7046 (vst4_u64): Likewise.
7047 (vst4_f64): Likewise.
7048 (vst4_s8): Likewise.
7049 (vst4_p8): Likewise.
7050 (vst4_s16): Likewise.
7051 (vst4_p16): Likewise.
7052 (vst4_s32): Likewise.
7053 (vst4_u8): Likewise.
7054 (vst4_u16): Likewise.
7055 (vst4_u32): Likewise.
7056 (vst4_f16): Likewise.
7057 (vst4_f32): Likewise.
7058 (vst4_p64): Likewise.
7059 (vst4q_s8): Likewise.
7060 (vst4q_p8): Likewise.
7061 (vst4q_s16): Likewise.
7062 (vst4q_p16): Likewise.
7063 (vst4q_s32): Likewise.
7064 (vst4q_s64): Likewise.
7065 (vst4q_u8): Likewise.
7066 (vst4q_u16): Likewise.
7067 (vst4q_u32): Likewise.
7068 (vst4q_u64): Likewise.
7069 (vst4q_f16): Likewise.
7070 (vst4q_f32): Likewise.
7071 (vst4q_f64): Likewise.
7072 (vst4q_p64): Likewise.
7073 (vtbx4_s8): Likewise.
7074 (vtbx4_u8): Likewise.
7075 (vtbx4_p8): Likewise.
7076 (vld1_bf16_x2): Likewise.
7077 (vld1q_bf16_x2): Likewise.
7078 (vld1_bf16_x3): Likewise.
7079 (vld1q_bf16_x3): Likewise.
7080 (vld1_bf16_x4): Likewise.
7081 (vld1q_bf16_x4): Likewise.
7082 (vld2_bf16): Likewise.
7083 (vld2q_bf16): Likewise.
7084 (vld2_dup_bf16): Likewise.
7085 (vld2q_dup_bf16): Likewise.
7086 (vld3_bf16): Likewise.
7087 (vld3q_bf16): Likewise.
7088 (vld3_dup_bf16): Likewise.
7089 (vld3q_dup_bf16): Likewise.
7090 (vld4_bf16): Likewise.
7091 (vld4q_bf16): Likewise.
7092 (vld4_dup_bf16): Likewise.
7093 (vld4q_dup_bf16): Likewise.
7094 (vst1_bf16_x2): Likewise.
7095 (vst1q_bf16_x2): Likewise.
7096 (vst1_bf16_x3): Likewise.
7097 (vst1q_bf16_x3): Likewise.
7098 (vst1_bf16_x4): Likewise.
7099 (vst1q_bf16_x4): Likewise.
7100 (vst2_bf16): Likewise.
7101 (vst2q_bf16): Likewise.
7102 (vst3_bf16): Likewise.
7103 (vst3q_bf16): Likewise.
7104 (vst4_bf16): Likewise.
7105 (vst4q_bf16): Likewise.
7106 (vld2_lane_bf16): Likewise.
7107 (vld2q_lane_bf16): Likewise.
7108 (vld3_lane_bf16): Likewise.
7109 (vld3q_lane_bf16): Likewise.
7110 (vld4_lane_bf16): Likewise.
7111 (vld4q_lane_bf16): Likewise.
7112 (vst2_lane_bf16): Likewise.
7113 (vst2q_lane_bf16): Likewise.
7114 (vst3_lane_bf16): Likewise.
7115 (vst3q_lane_bf16): Likewise.
7116 (vst4_lane_bf16): Likewise.
7117 (vst4q_lane_bf16): Likewise.
7118 * config/aarch64/geniterators.sh: Modify iterator regex to
7119 match new vector-tuple modes.
7120 * config/aarch64/iterators.md (insn_count): Extend mode
7121 attribute with vector-tuple type information.
7123 (Vendreg): Likewise.
7126 (VSTRUCT_2D): New mode iterator.
7127 (VSTRUCT_2DNX): Likewise.
7128 (VSTRUCT_2DX): Likewise.
7129 (VSTRUCT_2Q): Likewise.
7130 (VSTRUCT_2QD): Likewise.
7131 (VSTRUCT_3D): Likewise.
7132 (VSTRUCT_3DNX): Likewise.
7133 (VSTRUCT_3DX): Likewise.
7134 (VSTRUCT_3Q): Likewise.
7135 (VSTRUCT_3QD): Likewise.
7136 (VSTRUCT_4D): Likewise.
7137 (VSTRUCT_4DNX): Likewise.
7138 (VSTRUCT_4DX): Likewise.
7139 (VSTRUCT_4Q): Likewise.
7140 (VSTRUCT_4QD): Likewise.
7141 (VSTRUCT_D): Likewise.
7142 (VSTRUCT_Q): Likewise.
7143 (VSTRUCT_QD): Likewise.
7144 (VSTRUCT_ELT): New mode attribute.
7145 (vstruct_elt): Likewise.
7146 * genmodes.c (VECTOR_MODE): Add default prefix and order
7148 (VECTOR_MODE_WITH_PREFIX): Define.
7149 (make_vector_mode): Add mode prefix and order parameters.
7151 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
7153 * expmed.c (extract_bit_field_1): Ensure modes are tieable.
7155 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
7157 * expr.c (emit_group_load_1): Remove historic workaround.
7159 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
7161 * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
7162 Factor out main loop to...
7163 (aarch64_init_simd_builtin_functions): This new function.
7164 (register_tuple_type): Define.
7165 (aarch64_scalar_builtin_type_p): Define.
7166 (handle_arm_neon_h): Define.
7167 * config/aarch64/aarch64-c.c (aarch64_pragma_aarch64): Handle
7168 pragma for arm_neon.h.
7169 * config/aarch64/aarch64-protos.h (aarch64_advsimd_struct_mode_p):
7171 (handle_arm_neon_h): Likewise.
7172 * config/aarch64/aarch64.c (aarch64_advsimd_struct_mode_p):
7173 Remove static modifier.
7174 * config/aarch64/arm_neon.h (target): Remove Neon vector
7175 structure type definitions.
7177 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
7179 PR tree-optimization/102943
7180 * gimple-range-path.cc (path_range_query::range_on_path_entry):
7181 Prefer range_of_expr unless there are no statements in the BB.
7183 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
7185 PR tree-optimization/102943
7186 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
7187 Avoid duplicate calculation of paths.
7189 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
7191 PR tree-optimization/102943
7192 * gimple-range-path.cc (path_range_query::compute_phi_relations):
7193 Only compute relations for SSA names in the import list.
7194 (path_range_query::compute_outgoing_relations): Same.
7195 * gimple-range-path.h (path_range_query::import_p): New.
7197 2021-11-04 Richard Biener <rguenther@suse.de>
7199 PR rtl-optimization/103075
7200 * simplify-rtx.c (exact_int_to_float_conversion_p): Return
7201 false for a VOIDmode operand.
7203 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
7205 * config/aarch64/aarch64.c (aarch64_vector_costs): Make member
7206 variables private and add "m_" to their names. Remove is_loop.
7207 (aarch64_record_potential_advsimd_unrolling): Replace with...
7208 (aarch64_vector_costs::record_potential_advsimd_unrolling): ...this.
7209 (aarch64_analyze_loop_vinfo): Replace with...
7210 (aarch64_vector_costs::analyze_loop_vinfo): ...this.
7211 Move initialization of (m_)vec_flags to add_stmt_cost.
7212 (aarch64_analyze_bb_vinfo): Delete.
7213 (aarch64_count_ops): Replace with...
7214 (aarch64_vector_costs::count_ops): ...this.
7215 (aarch64_vector_costs::add_stmt_cost): Set m_vec_flags,
7216 using m_costing_for_scalar to test whether we're costing
7217 scalar or vector code.
7218 (aarch64_adjust_body_cost_sve): Replace with...
7219 (aarch64_vector_costs::adjust_body_cost_sve): ...this.
7220 (aarch64_adjust_body_cost): Replace with...
7221 (aarch64_vector_costs::adjust_body_cost): ...this.
7222 (aarch64_vector_costs::finish_cost): Use m_vinfo instead of is_loop.
7224 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
7226 * target.def (targetm.vectorize.init_cost): Replace with...
7227 (targetm.vectorize.create_costs): ...this.
7228 (targetm.vectorize.add_stmt_cost): Delete.
7229 (targetm.vectorize.finish_cost): Likewise.
7230 (targetm.vectorize.destroy_cost_data): Likewise.
7231 * doc/tm.texi.in (TARGET_VECTORIZE_INIT_COST): Replace with...
7232 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
7233 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
7234 (TARGET_VECTORIZE_FINISH_COST): Likewise.
7235 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
7236 * doc/tm.texi: Regenerate.
7237 * tree-vectorizer.h (vec_info::vec_info): Remove target_cost_data
7239 (vec_info::target_cost_data): Change from a void * to a vector_costs *.
7240 (vector_costs): New class.
7241 (init_cost): Take a vec_info and return a vector_costs.
7242 (dump_stmt_cost): Remove data parameter.
7243 (add_stmt_cost): Replace vinfo and data parameters with a vector_costs.
7244 (add_stmt_costs): Likewise.
7245 (finish_cost): Replace data parameter with a vector_costs.
7246 (destroy_cost_data): Delete.
7247 * tree-vectorizer.c (dump_stmt_cost): Remove data argument and
7249 (vec_info::vec_info): Remove the target_cost_data parameter and
7250 initialize the member variable to null instead.
7251 (vec_info::~vec_info): Delete target_cost_data instead of calling
7253 (vector_costs::add_stmt_cost): New function.
7254 (vector_costs::finish_cost): Likewise.
7255 (vector_costs::record_stmt_cost): Likewise.
7256 (vector_costs::adjust_cost_for_freq): Likewise.
7257 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
7258 call to vec_info::vec_info.
7259 (vect_compute_single_scalar_iteration_cost): Update after above
7260 changes to costing interface.
7261 (vect_analyze_loop_operations): Likewise.
7262 (vect_estimate_min_profitable_iters): Likewise.
7263 (vect_analyze_loop_2): Initialize LOOP_VINFO_TARGET_COST_DATA
7264 at the start_over point, where it needs to be recreated after
7265 trying without slp. Update retry code accordingly.
7266 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Update call
7267 to vec_info::vec_info.
7268 (vect_slp_analyze_operation): Update after above changes to costing
7270 (vect_bb_vectorization_profitable_p): Likewise.
7271 * targhooks.h (default_init_cost): Replace with...
7272 (default_vectorize_create_costs): ...this.
7273 (default_add_stmt_cost): Delete.
7274 (default_finish_cost, default_destroy_cost_data): Likewise.
7275 * targhooks.c (default_init_cost): Replace with...
7276 (default_vectorize_create_costs): ...this.
7277 (default_add_stmt_cost): Delete, moving logic to vector_costs instead.
7278 (default_finish_cost, default_destroy_cost_data): Delete.
7279 * config/aarch64/aarch64.c (aarch64_vector_costs): Inherit from
7280 vector_costs. Add a constructor.
7281 (aarch64_init_cost): Replace with...
7282 (aarch64_vectorize_create_costs): ...this.
7283 (aarch64_add_stmt_cost): Replace with...
7284 (aarch64_vector_costs::add_stmt_cost): ...this. Use record_stmt_cost
7285 to adjust the cost for inner loops.
7286 (aarch64_finish_cost): Replace with...
7287 (aarch64_vector_costs::finish_cost): ...this.
7288 (aarch64_destroy_cost_data): Delete.
7289 (TARGET_VECTORIZE_INIT_COST): Replace with...
7290 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
7291 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
7292 (TARGET_VECTORIZE_FINISH_COST): Likewise.
7293 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
7294 * config/i386/i386.c (ix86_vector_costs): New structure.
7295 (ix86_init_cost): Replace with...
7296 (ix86_vectorize_create_costs): ...this.
7297 (ix86_add_stmt_cost): Replace with...
7298 (ix86_vector_costs::add_stmt_cost): ...this. Use adjust_cost_for_freq
7299 to adjust the cost for inner loops.
7300 (ix86_finish_cost, ix86_destroy_cost_data): Delete.
7301 (TARGET_VECTORIZE_INIT_COST): Replace with...
7302 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
7303 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
7304 (TARGET_VECTORIZE_FINISH_COST): Likewise.
7305 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
7306 * config/rs6000/rs6000.c (TARGET_VECTORIZE_INIT_COST): Replace with...
7307 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
7308 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
7309 (TARGET_VECTORIZE_FINISH_COST): Likewise.
7310 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
7311 (rs6000_cost_data): Inherit from vector_costs.
7312 Add a constructor. Drop loop_info, cost and costing_for_scalar
7313 in favor of the corresponding vector_costs member variables.
7314 Add "m_" to the names of the remaining member variables and
7316 (rs6000_density_test): Replace with...
7317 (rs6000_cost_data::density_test): ...this.
7318 (rs6000_init_cost): Replace with...
7319 (rs6000_vectorize_create_costs): ...this.
7320 (rs6000_update_target_cost_per_stmt): Replace with...
7321 (rs6000_cost_data::update_target_cost_per_stmt): ...this.
7322 (rs6000_add_stmt_cost): Replace with...
7323 (rs6000_cost_data::add_stmt_cost): ...this. Use adjust_cost_for_freq
7324 to adjust the cost for inner loops.
7325 (rs6000_adjust_vect_cost_per_loop): Replace with...
7326 (rs6000_cost_data::adjust_vect_cost_per_loop): ...this.
7327 (rs6000_finish_cost): Replace with...
7328 (rs6000_cost_data::finish_cost): ...this. Group loop code
7329 into a single if statement and pass the loop_vinfo down to
7331 (rs6000_destroy_cost_data): Delete.
7333 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
7335 PR tree-optimization/103062
7336 PR tree-optimization/103062
7337 * value-pointer-equiv.cc (ssa_equiv_stack::ssa_equiv_stack):
7338 Increase size of allocation by 1.
7339 (ssa_equiv_stack::push_replacement): Grow as needed.
7340 (ssa_equiv_stack::get_replacement): Same.
7341 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
7342 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Remove delete.
7343 (pointer_equiv_analyzer::set_global_equiv): Grow as needed.
7344 (pointer_equiv_analyzer::get_equiv): Same.
7345 (pointer_equiv_analyzer::get_equiv_expr): Remove const.
7346 * value-pointer-equiv.h (class pointer_equiv_analyzer): Remove
7347 const markers. Use auto_vec instead of tree *.
7349 2021-11-04 Richard Biener <rguenther@suse.de>
7351 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Remove always
7352 true parameter and inline valueization.
7353 (vn_nary_op_lookup_1): Inline valueization from ...
7354 (vn_nary_op_compute_hash): ... here and remove it here.
7355 * tree-ssa-pre.c (phi_translate_1): Do not valueize
7356 before vn_nary_lookup_pieces.
7357 (get_representative_for): Mark created SSA representatives
7360 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
7362 * simplify-rtx.c (simplify_context::simplify_gen_vec_select): Assert
7363 that the operand has a vector mode. Use subreg_lowpart_offset
7364 to test whether an index corresponds to the low part.
7366 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
7368 * read-rtl.c: Remove dead !GENERATOR_FILE block.
7369 * read-rtl-function.c (function_reader::consolidate_singletons):
7370 Generate canonical CONST_VECTORs.
7372 2021-11-04 liuhongt <hongtao.liu@intel.com>
7375 * config/i386/predicates.md (reg_or_notreg_operand): Rename to ..
7376 (regmem_or_bitnot_regmem_operand): .. and extend to handle
7378 * config/i386/sse.md (*<avx512>_vpternlog<mode>_1): Force_reg
7379 the operands which are required to be register_operand.
7380 (*<avx512>_vpternlog<mode>_2): Ditto.
7381 (*<avx512>_vpternlog<mode>_3): Ditto.
7382 (*<avx512>_vternlog<mode>_all): Disallow embeded broadcast for
7383 vector HFmodes since it's not a real AVX512FP16 instruction.
7385 2021-11-04 liuhongt <hongtao.liu@intel.com>
7388 * match.pd: simplify (trunc)copysign((extend)a, (extend)b) to
7389 .COPYSIGN (a,b) when a and b are same type as the truncation
7390 type and has less precision than extend type.
7392 2021-11-04 Richard Biener <rguenther@suse.de>
7394 * doc/generic.texi: Update TARGET_MEM_REF and MEM_REF
7397 2021-11-04 Hongyu Wang <hongyu.wang@intel.com>
7399 * config/i386/sse.md (VI2_AVX512VNNIBW): New mode iterator.
7400 (VI1_AVX512VNNI): Likewise.
7401 (SDOT_VPDP_SUF): New mode_attr.
7404 (sdot_prod<mode>): Use VI2_AVX512F iterator, expand to
7405 vpdpwssd when VNNI targets available.
7406 (usdot_prod<mode>): New expander for vector QImode.
7408 2021-11-04 Hongyu Wang <hongyu.wang@intel.com>
7410 * config/i386/amxtileintrin.h (_tile_loadd_internal): Add
7411 parentheses to base and stride.
7412 (_tile_stream_loadd_internal): Likewise.
7413 (_tile_stored_internal): Likewise.
7415 2021-11-03 Maciej W. Rozycki <macro@embecosm.com>
7417 * config/riscv/riscv.c (riscv_class_max_nregs): Swap the
7418 arguments to `reg_class_subset_p'.
7420 2021-11-03 Joseph Myers <joseph@codesourcery.com>
7423 * fold-const.c (fold_init): New function.
7424 * fold-const.h (fold_init): New prototype.
7426 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
7428 * params.opt (param_vrp2_mode): Make ranger the default for VRP2.
7430 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
7432 * gimple-range-fold.cc (fold_using_range::range_of_phi): Don't import
7433 a range from edge if arg == phidef.
7435 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
7437 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Test
7438 for constant before any other processing.
7440 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
7442 * flag-types.h (RANGER_DEBUG_ALL): Fix values.
7444 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
7446 * gimple-range.cc (gimple_ranger::gimple_ranger): Initialize current_bb.
7447 (gimple_ranger::range_of_expr): Pick up range_on_entry when there is
7448 no explcit context and current_bb is set.
7449 (gimple_ranger::fold_stmt): New.
7450 * gimple-range.h (current_bb, fold_stmt): New.
7451 * tree-vrp.c (rvrp_folder::fold_stmt): Call ranger's fold_stmt.
7453 2021-11-03 Richard Biener <rguenther@suse.de>
7455 PR tree-optimization/102970
7456 * tree-ssa-pre.c (phi_translate_1): Drop clique and base
7457 when translating a MEM_REF over a backedge.
7459 2021-11-03 Philipp Tomsich <philipp.tomsich@vrull.eu>
7461 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New Ampere-1 core.
7462 * config/aarch64/aarch64-tune.md: Regenerate.
7463 * config/aarch64/aarch64-cost-tables.h: Add extra costs for Ampere-1.
7464 * config/aarch64/aarch64.c: Add tuning structures for Ampere-1.
7465 * doc/invoke.texi: Add documentation for Ampere-1 core.
7467 2021-11-03 Wilco Dijkstra <wdijkstr@arm.com>
7469 * config/aarch64/aarch64.md (movsi): Add alternative for GOT accesses.
7471 (ldr_got_small_<mode>): Remove pattern.
7472 (ldr_got_small_sidi): Likewise.
7473 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately): Keep
7474 GOT accesses as moves.
7475 (aarch64_print_operand): Correctly print got_lo12 in L specifier.
7476 (aarch64_mov_operand_p): Make GOT accesses valid move operands.
7477 * config/aarch64/constraints.md: Add new constraint Usw for GOT access.
7479 2021-11-03 Martin Liska <mliska@suse.cz>
7481 * gcov.c (read_line): Remove dead variable.
7483 2021-11-03 Martin Liska <mliska@suse.cz>
7486 * ipa-fnsummary.c (edge_predicate_pool): Rename predicate class to ipa_predicate.
7487 (ipa_fn_summary::account_size_time): Likewise.
7488 (edge_set_predicate): Likewise.
7489 (set_hint_predicate): Likewise.
7490 (add_freqcounting_predicate): Likewise.
7491 (evaluate_conditions_for_known_args): Likewise.
7492 (evaluate_properties_for_edge): Likewise.
7493 (remap_freqcounting_preds_after_dup): Likewise.
7494 (ipa_fn_summary_t::duplicate): Likewise.
7495 (set_cond_stmt_execution_predicate): Likewise.
7496 (set_switch_stmt_execution_predicate): Likewise.
7497 (compute_bb_predicates): Likewise.
7498 (will_be_nonconstant_expr_predicate): Likewise.
7499 (will_be_nonconstant_predicate): Likewise.
7500 (phi_result_unknown_predicate): Likewise.
7501 (predicate_for_phi_result): Likewise.
7502 (analyze_function_body): Likewise.
7503 (compute_fn_summary): Likewise.
7504 (summarize_calls_size_and_time): Likewise.
7505 (estimate_calls_size_and_time): Likewise.
7506 (ipa_call_context::estimate_size_and_time): Likewise.
7507 (remap_edge_summaries): Likewise.
7508 (remap_freqcounting_predicate): Likewise.
7509 (ipa_merge_fn_summary_after_inlining): Likewise.
7510 (ipa_update_overall_fn_summary): Likewise.
7511 (read_ipa_call_summary): Likewise.
7512 (inline_read_section): Likewise.
7513 * ipa-fnsummary.h (struct ipa_freqcounting_predicate): Likewise.
7514 * ipa-predicate.c (predicate::add_clause): Likewise.
7515 (ipa_predicate::add_clause): Likewise.
7516 (predicate::or_with): Likewise.
7517 (ipa_predicate::or_with): Likewise.
7518 (predicate::evaluate): Likewise.
7519 (ipa_predicate::evaluate): Likewise.
7520 (predicate::probability): Likewise.
7521 (ipa_predicate::probability): Likewise.
7522 (dump_condition): Likewise.
7523 (dump_clause): Likewise.
7524 (predicate::dump): Likewise.
7525 (ipa_predicate::dump): Likewise.
7526 (predicate::debug): Likewise.
7527 (ipa_predicate::debug): Likewise.
7528 (predicate::remap_after_duplication): Likewise.
7529 (ipa_predicate::remap_after_duplication): Likewise.
7530 (predicate::remap_after_inlining): Likewise.
7531 (ipa_predicate::remap_after_inlining): Likewise.
7532 (predicate::stream_in): Likewise.
7533 (ipa_predicate::stream_in): Likewise.
7534 (predicate::stream_out): Likewise.
7535 (ipa_predicate::stream_out): Likewise.
7536 (add_condition): Likewise.
7537 * ipa-predicate.h (class predicate): Likewise.
7538 (class ipa_predicate): Likewise.
7539 (add_condition): Likewise.
7541 2021-11-03 Richard Biener <rguenther@suse.de>
7543 * bitmap.h (bitmap_bit_p): Change the return type to bool.
7544 * bitmap.c (bitmap_bit_p): Likewise.
7545 * sbitmap.h (bitmap_bit_p): Likewise.
7546 (bitmap_set_bit): Return whether the bit changed.
7547 (bitmap_clear_bit): Likewise.
7548 * tree-ssa.c (verify_vssa): Make use of the changed state
7549 from bitmap_set_bit.
7551 2021-11-03 Richard Biener <rguenther@suse.de>
7553 PR middle-end/103033
7554 * internal-fn.c (expand_DEFERRED_INIT): Elide the
7555 native_interpret_expr path in favor of folding the
7556 VIEW_CONVERT_EXPR generated when punning the RHS.
7558 2021-11-03 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
7560 * config/s390/s390.c (s390_loop_unroll_adjust): In case of early
7563 2021-11-03 Jan Hubicka <hubicka@ucw.cz>
7566 * ipa-modref.c (callee_to_caller_flags): New function.
7567 (modref_eaf_analysis::analyze_ssa_name): Use it.
7568 (ipa_merge_modref_summary_after_inlining): Fix whitespace.
7570 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
7572 * ipa-modref.c (modref_eaf_analysis::analyze_ssa_name): Revert
7575 2021-11-02 Roger Sayle <roger@nextmovesoftware.com>
7576 Uroš Bizjak <ubizjak@gmail.com>
7578 * config/i386/i386.md (<any_rotate>ti3): Provide expansion for
7579 rotations by non-constant amounts.
7581 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
7583 * ipa-modref.c: Fix anonymous namespace placement.
7584 (class modref_eaf_analysis): New class.
7585 (analyze_ssa_name_flags): Turn to ...
7586 (modref_eaf_analysis::analyze_ssa_name): ... this one.
7587 (merge_call_lhs_flags): Turn to ...
7588 (modref_eaf_analysis::merge_call_lhs_flags): .. this one
7589 (modref_eaf_analysis::merge_with_ssa_name): New member function.
7590 (record_escape_points): Turn to ...
7591 (modref_eaf_analysis::record_escape_points): ... this one.
7592 (analyze_parms): Updat
7593 (ipa_merge_modref_summary_after_inlining): Move to the end of file.
7595 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
7597 * gimple.c (gimple_call_static_chain_flags): New function.
7598 * gimple.h (gimple_call_static_chain_flags): Declare
7599 * ipa-modref.c (modref_summary::modref_summary): Initialize
7601 (modref_summary_lto::modref_summary_lto): Likewise.
7602 (modref_summary::useful_p): Test static_chain_flags.
7603 (modref_summary_lto::useful_p): Likewise.
7604 (struct modref_summary_lto): Add static_chain_flags.
7605 (modref_summary::dump): Dump static_chain_flags.
7606 (modref_summary_lto::dump): Likewise.
7607 (struct escape_point): Add static_cahin_arg.
7608 (analyze_ssa_name_flags): Use gimple_call_static_chain_flags.
7609 (analyze_parms): Handle static chains.
7610 (modref_summaries::duplicate): Duplicate static_chain_flags.
7611 (modref_summaries_lto::duplicate): Likewise.
7612 (modref_write): Stream static_chain_flags.
7613 (read_section): Likewise.
7614 (modref_merge_call_site_flags): Handle static_chain_flags.
7615 * ipa-modref.h (struct modref_summary): Add static_chain_flags.
7616 * tree-ssa-structalias.c (handle_rhs_call): Use
7617 gimple_static_chain_flags.
7619 2021-11-02 Richard Biener <rguenther@suse.de>
7621 PR tree-optimization/103029
7622 * tree-vect-loop-manip.c (vect_loop_versioning): Ensure
7623 the PHI nodes in the loop maintain their original operand
7626 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
7628 * tree-core.h (EAF_NOT_RETURNED_DIRECTLY): New flag.
7629 (EAF_NOREAD): Renumber.
7630 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOT_RETURNED_DIRECTLY.
7631 (remove_useless_eaf_flags): Handle EAF_NOT_RETURNED_DIRECTLY
7632 (deref_flags): Likewise.
7633 (modref_lattice::init): Likewise.
7634 (modref_lattice::merge): Likewise.
7635 (merge_call_lhs_flags): Likewise.
7636 (analyze_ssa_name_flags): Likewise.
7637 (modref_merge_call_site_flags): Likewise.
7638 * tree-ssa-structalias.c (handle_call_arg): Likewise.
7640 2021-11-02 Maciej W. Rozycki <macro@embecosm.com>
7642 * config/riscv/riscv.c (riscv_rtx_costs): Correct a CONST_INT_P
7643 check and remove an unused local variable with shNadd/shNadd.uw
7646 2021-11-02 Tamar Christina <tamar.christina@arm.com>
7648 PR tree-optimization/103007
7649 * tree-vect-slp-patterns.c (complex_fms_pattern::matches): Add elem
7652 2021-11-02 Richard Biener <rguenther@suse.de>
7654 PR middle-end/103038
7655 * fold-const.c (native_interpret_expr): Handle OFFSET_TYPE.
7656 (can_native_interpret_type_p): Likewise.
7657 * internal-fn.c (expand_DEFERRED_INIT): View-convert the
7658 RHS if the LHS is an SSA name.
7660 2021-11-02 Richard Sandiford <richard.sandiford@arm.com>
7662 * langhooks.h (lang_hooks_for_types::simulate_record_decl): New hook.
7663 * langhooks-def.h (lhd_simulate_record_decl): Declare.
7664 (LANG_HOOKS_SIMULATE_RECORD_DECL): Define.
7665 (LANG_HOOKS_FOR_TYPES_INITIALIZER): Include it.
7666 * langhooks.c (lhd_simulate_record_decl): New function.
7668 2021-11-02 Jakub Jelinek <jakub@redhat.com>
7671 * config/i386/i386.c (ix86_vector_mode_supported_p): Reject vector
7672 modes with TImode inner mode if 32-bit.
7674 2021-11-02 liuhongt <hongtao.liu@intel.com>
7676 * doc/sourcebuild.texi (vect_slp_v4qi_store_unalign,
7677 vect_slp_v2hi_store_unalign, vect_slp_v4hi_store_unalign,
7678 vect_slp_v4si_store_unalign): Document efficient target.
7679 (vect_slp_v4qi_store_unalign_1, vect_slp_v8qi_store_unalign_1,
7680 vect_slp_v16qi_store_unalign_1): Ditto.
7681 (vect_slp_v2hi_store_align,vect_slp_v2qi_store_align,
7682 vect_slp_v2si_store_align, vect_slp_v4qi_store_align): Ditto.
7683 (struct_4char_block_move, struct_8char_block_move,
7684 struct_16char_block_move): Ditto.
7686 2021-11-02 Roger Sayle <roger@nextmovesoftware.com>
7687 Jakub Jelinek <jakub@redhat.com>
7690 * config/i386/i386-expand.c (ix86_expand_v1ti_to_ti,
7691 ix86_expand_ti_to_v1ti): New helper functions.
7692 (ix86_expand_v1ti_shift): Check if the amount operand is an
7693 integer constant, and expand as a TImode shift if it isn't.
7694 (ix86_expand_v1ti_rotate): Check if the amount operand is an
7695 integer constant, and expand as a TImode rotate if it isn't.
7696 (ix86_expand_v1ti_ashiftrt): New function to expand arithmetic
7697 right shifts of V1TImode quantities.
7698 * config/i386/i386-protos.h (ix86_expand_v1ti_ashift): Prototype.
7699 * config/i386/sse.md (ashlv1ti3, lshrv1ti3): Change constraints
7700 to QImode general_operand, and let the helper functions lower
7701 shifts by non-constant operands, as TImode shifts. Make
7702 conditional on TARGET_64BIT.
7703 (ashrv1ti3): New expander calling ix86_expand_v1ti_ashiftrt.
7704 (rotlv1ti3, rotrv1ti3): Change shift operand to QImode.
7705 Make conditional on TARGET_64BIT.
7707 2021-11-02 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
7709 * config/s390/s390.md ("*cc_to_int", "tabort", "*tabort_1",
7710 "*tabort_1_plus"): Remove operands non-null check.
7712 2021-11-01 Martin Liska <mliska@suse.cz>
7714 * opt-functions.awk: Add new sanity checking.
7715 * optc-gen.awk: Add new argument to integer_range_info.
7716 * params.opt: Update 2 params which have negative IntegerRange.
7718 2021-11-01 qing zhao <qing.zhao@oracle.com>
7720 * gimplify.c (gimplify_decl_expr): Do not add call to
7721 __builtin_clear_padding when a variable is a gimple register
7722 or it might not have padding.
7723 (gimplify_init_constructor): Likewise.
7725 2021-11-01 Tamar Christina <tamar.christina@arm.com>
7727 * config/arm/aarch-common-protos.h (struct vector_cost_table): Add
7728 movi, dup and extract costing fields.
7729 * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs,
7730 thunderx_extra_costs, thunderx2t99_extra_costs,
7731 thunderx3t110_extra_costs, tsv110_extra_costs, a64fx_extra_costs): Use
7733 * config/arm/aarch-cost-tables.h (generic_extra_costs,
7734 cortexa53_extra_costs, cortexa57_extra_costs, cortexa76_extra_costs,
7735 exynosm1_extra_costs, xgene1_extra_costs): Likewise
7736 * config/aarch64/aarch64-simd.md (aarch64_simd_dup<mode>): Add r->w dup.
7737 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add extra costs.
7739 2021-11-01 Tamar Christina <tamar.christina@arm.com>
7741 * cse.c (add_to_set): New.
7742 (find_sets_in_insn): Register constants in sets.
7743 (canonicalize_insn): Use auto_vec instead.
7744 (cse_insn): Try materializing using vec_dup.
7745 * rtl.h (simplify_context::simplify_gen_vec_select,
7746 simplify_gen_vec_select): New.
7747 * simplify-rtx.c (simplify_context::simplify_gen_vec_select): New.
7749 2021-11-01 David Malcolm <dmalcolm@redhat.com>
7751 * common.opt (fdiagnostics-escape-format=): New.
7752 (diagnostics_escape_format): New enum.
7753 (DIAGNOSTICS_ESCAPE_FORMAT_UNICODE): New enum value.
7754 (DIAGNOSTICS_ESCAPE_FORMAT_BYTES): Likewise.
7755 * diagnostic-format-json.cc (json_end_diagnostic): Add
7756 "escape-source" attribute.
7757 * diagnostic-show-locus.c
7758 (exploc_with_display_col::exploc_with_display_col): Replace
7759 "tabstop" param with a cpp_char_column_policy and add an "aspect"
7760 param. Use these to compute m_display_col accordingly.
7761 (struct char_display_policy): New struct.
7762 (layout::m_policy): New field.
7763 (layout::m_escape_on_output): New field.
7764 (def_policy): New function.
7765 (make_range): Update for changes to exploc_with_display_col ctor.
7766 (default_print_decoded_ch): New.
7767 (width_per_escaped_byte): New.
7768 (escape_as_bytes_width): New.
7769 (escape_as_bytes_print): New.
7770 (escape_as_unicode_width): New.
7771 (escape_as_unicode_print): New.
7773 (layout::layout): Initialize new fields. Update m_exploc ctor
7774 call for above change to ctor.
7775 (layout::maybe_add_location_range): Update for changes to
7776 exploc_with_display_col ctor.
7777 (layout::calculate_x_offset_display): Update for change to
7779 (layout::print_source_line): Pass policy
7780 to cpp_display_width_computation. Capture cpp_decoded_char when
7781 calling process_next_codepoint. Move printing of source code to
7782 m_policy.m_print_cb.
7783 (line_label::line_label): Pass in policy rather than context.
7784 (layout::print_any_labels): Update for change to line_label ctor.
7785 (get_affected_range): Pass in policy rather than context, updating
7786 calls to location_compute_display_column accordingly.
7787 (get_printed_columns): Likewise, also for cpp_display_width.
7788 (correction::correction): Pass in policy rather than tabstop.
7789 (correction::compute_display_cols): Pass m_policy rather than
7790 m_tabstop to cpp_display_width.
7791 (correction::m_tabstop): Replace with...
7792 (correction::m_policy): ...this.
7793 (line_corrections::line_corrections): Pass in policy rather than
7795 (line_corrections::m_context): Replace with...
7796 (line_corrections::m_policy): ...this.
7797 (line_corrections::add_hint): Update to use m_policy rather than
7799 (line_corrections::add_hint): Likewise.
7800 (layout::print_trailing_fixits): Likewise.
7801 (selftest::test_display_widths): New.
7802 (selftest::test_layout_x_offset_display_utf8): Update to use
7803 policy rather than tabstop.
7804 (selftest::test_one_liner_labels_utf8): Add test of escaping
7806 (selftest::test_diagnostic_show_locus_one_liner_utf8): Update to
7807 use policy rather than tabstop.
7808 (selftest::test_overlapped_fixit_printing): Likewise.
7809 (selftest::test_overlapped_fixit_printing_utf8): Likewise.
7810 (selftest::test_overlapped_fixit_printing_2): Likewise.
7811 (selftest::test_tab_expansion): Likewise.
7812 (selftest::test_escaping_bytes_1): New.
7813 (selftest::test_escaping_bytes_2): New.
7814 (selftest::diagnostic_show_locus_c_tests): Call the new tests.
7815 * diagnostic.c (diagnostic_initialize): Initialize
7816 context->escape_format.
7817 (convert_column_unit): Update to use default character width policy.
7818 (selftest::test_diagnostic_get_location_text): Likewise.
7819 * diagnostic.h (enum diagnostics_escape_format): New enum.
7820 (diagnostic_context::escape_format): New field.
7821 * doc/invoke.texi (-fdiagnostics-escape-format=): New option.
7822 (-fdiagnostics-format=): Add "escape-source" attribute to examples
7823 of JSON output, and document it.
7824 * input.c (location_compute_display_column): Pass in "policy"
7825 rather than "tabstop", passing to
7826 cpp_byte_column_to_display_column.
7827 (selftest::test_cpp_utf8): Update to use cpp_char_column_policy.
7828 * input.h (class cpp_char_column_policy): New forward decl.
7829 (location_compute_display_column): Pass in "policy" rather than
7831 * opts.c (common_handle_option): Handle
7832 OPT_fdiagnostics_escape_format_.
7833 * selftest.c (temp_source_file::temp_source_file): New ctor
7834 overload taking a size_t.
7835 * selftest.h (temp_source_file::temp_source_file): Likewise.
7837 2021-11-01 Aldy Hernandez <aldyh@redhat.com>
7839 * dbgcnt.def: Add debug counter for back_thread[12] and
7840 back_threadfull[12].
7841 * passes.def: Pass "first" argument to each back threading pass.
7842 * tree-ssa-threadbackward.c (back_threader::back_threader): Add
7844 (back_threader::debug_counter): New.
7845 (back_threader::maybe_register_path): Call debug_counter.
7847 2021-11-01 Aldy Hernandez <aldyh@redhat.com>
7849 * tree-ssa-threadbackward.c (BT_NONE): New.
7852 (back_threader::back_threader): Add flags.
7853 Move loop initialization here.
7854 (back_threader::~back_threader): New.
7855 (back_threader::find_taken_edge_switch): Change solver and ranger
7857 (back_threader::find_taken_edge_cond): Same.
7858 (back_threader::find_paths_to_names): Same.
7859 (back_threader::find_paths): Same.
7860 (back_threader::dump): Same.
7861 (try_thread_blocks): Merge into thread_blocks.
7862 (back_threader::thread_blocks): New.
7863 (do_early_thread_jumps): Merge into thread_blocks.
7864 (do_thread_jumps): Merge into thread_blocks.
7865 (back_threader::thread_through_all_blocks): Remove.
7867 2021-11-01 Andrew MacLeod <amacleod@redhat.com>
7869 PR tree-optimization/103003
7870 * value-relation.cc (dom_oracle::register_relation): If the 2
7871 ssa names are the same, don't register any relation.
7873 2021-11-01 Dan Li <ashimida@linux.alibaba.com>
7875 * config/aarch64/aarch64.c (aarch64_expand_epilogue): Remove
7876 redundant check for calls_eh_return.
7877 * config/aarch64/aarch64.md (*do_return): Likewise.
7879 2021-11-01 Xionghu Luo <luoxhu@linux.ibm.com>
7881 * cfghooks.c (cfg_hook_duplicate_loop_to_header_edge): Rename
7882 duplicate_loop_to_header_edge to
7883 duplicate_loop_body_to_header_edge.
7884 (cfg_hook_duplicate_loop_body_to_header_edge): Likewise.
7885 * cfghooks.h (struct cfg_hooks): Likewise.
7886 (cfg_hook_duplicate_loop_body_to_header_edge): Likewise.
7887 * cfgloopmanip.c (duplicate_loop_body_to_header_edge): Likewise.
7888 (clone_loop_to_header_edge): Likewise.
7889 * cfgloopmanip.h (duplicate_loop_body_to_header_edge): Likewise.
7890 * cfgrtl.c (struct cfg_hooks): Likewise.
7891 * doc/loop.texi: Likewise.
7892 * loop-unroll.c (unroll_loop_constant_iterations): Likewise.
7893 (unroll_loop_runtime_iterations): Likewise.
7894 (unroll_loop_stupid): Likewise.
7895 (apply_opt_in_copies): Likewise.
7896 * tree-cfg.c (struct cfg_hooks): Likewise.
7897 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
7898 (try_peel_loop): Likewise.
7899 * tree-ssa-loop-manip.c (copy_phi_node_args): Likewise.
7900 (gimple_duplicate_loop_body_to_header_edge): Likewise.
7901 (tree_transform_and_unroll_loop): Likewise.
7902 * tree-ssa-loop-manip.h (gimple_duplicate_loop_body_to_header_edge):
7905 2021-11-01 Xionghu Luo <luoxhu@linux.ibm.com>
7907 * cfgloopmanip.c (loop_version): Refactor loopify to
7908 loop_version. Move condition generation after loopify.
7910 * cfgloopmanip.h (loopify): Delete.
7912 2021-10-31 Jan Hubicka <hubicka@ucw.cz>
7914 * ipa-fnsummary.c: Include tree-dfa.h.
7915 (points_to_local_or_readonly_memory_p): Return true on return
7917 * ipa-modref.c (analyze_ssa_name_flags): Fix handling of copy
7920 2021-10-30 Tobias Burnus <tobias@codesourcery.com>
7922 PR middle-end/102972
7923 * omp-low.c (omp_runtime_api_call): Use DECL_ASSEMBLER_NAME to get
7924 internal Fortran name; new permit_num_teams arg to permit
7925 omp_get_num_teams and omp_get_team_num.
7926 (scan_omp_1_stmt): Update call to it, add missing call for
7927 reverse offload, and check for strictly nested API calls in teams.
7929 2021-10-30 Jakub Jelinek <jakub@redhat.com>
7931 * gimplify.c (gimplify_omp_for): Diagnose threadprivate iterators.
7933 2021-10-29 David Malcolm <dmalcolm@redhat.com>
7935 * selftest.c (assert_streq): Add newlines when emitting non-equal
7938 2021-10-29 David Malcolm <dmalcolm@redhat.com>
7940 * Makefile.in: Fix syntax for reference to LIBDEPS in
7943 2021-10-29 Bernhard Reutner-Fischer <aldot@gcc.gnu.org>
7945 * doc/install.texi: Bump required minimum DejaGnu version.
7947 2021-10-29 Aldy Hernandez <aldyh@redhat.com>
7948 Andrew MacLeod <amacleod@redhat.com>
7950 * value-relation.cc (path_oracle::killing_def): Add a
7951 self-equivalence so we don't look to the root oracle.
7953 2021-10-29 Aldy Hernandez <aldyh@redhat.com>
7955 * passes.def: Replace the pass_thread_jumps before VRP* with
7956 pass_thread_jumps_full. Remove all pass_vrp_threader instances.
7957 * tree-ssa-threadbackward.c (pass_data_thread_jumps_full):
7958 Remove hyphen from "thread-full" name.
7960 2021-10-29 Andrew MacLeod <amacleod@redhat.com>
7962 PR tree-optimization/102983
7963 * gimple-range-cache.h (propagate_updated_value): Make public.
7964 * gimple-range.cc (gimple_ranger::range_of_stmt): Propagate exports
7965 when processing gcond stmts.
7967 2021-10-29 Jan Hubicka <hubicka@ucw.cz>
7969 * gimple.c (gimple_call_retslot_flags): New function.
7970 * gimple.h (gimple_call_retslot_flags): Declare.
7971 * ipa-modref.c: Include tree-cfg.h.
7972 (struct escape_entry): Turn parm_index to signed.
7973 (modref_summary_lto::modref_summary_lto): Add retslot_flags.
7974 (modref_summary::modref_summary): Initialize retslot_flags.
7975 (struct modref_summary_lto): Likewise.
7976 (modref_summary::useful_p): Check retslot_flags.
7977 (modref_summary_lto::useful_p): Likewise.
7978 (modref_summary::dump): Dump retslot_flags.
7979 (modref_summary_lto::dump): Likewise.
7980 (struct escape_point): Add hidden_args enum.
7981 (analyze_ssa_name_flags): Ignore return slot return;
7982 use gimple_call_retslot_flags.
7983 (record_escape_points): Break out from ...
7984 (analyze_parms): ... here; handle retslot_flags.
7985 (modref_summaries::duplicate): Duplicate retslot_flags.
7986 (modref_summaries_lto::duplicate): Likewise.
7987 (modref_write_escape_summary): Stream parm_index as signed.
7988 (modref_read_escape_summary): Likewise.
7989 (modref_write): Stream retslot_flags.
7990 (read_section): Likewise.
7991 (struct escape_map): Fix typo in comment.
7992 (update_escape_summary_1): Fix whitespace.
7993 (ipa_merge_modref_summary_after_inlining): Drop retslot_flags.
7994 (modref_merge_call_site_flags): Merge retslot_flags.
7995 * ipa-modref.h (struct modref_summary): Add retslot_flags.
7996 * tree-ssa-structalias.c (handle_rhs_call): Handle retslot_flags.
7998 2021-10-29 Tamar Christina <tamar.christina@arm.com>
8000 PR tree-optimization/102977
8001 * tree-vect-slp-patterns.c (vect_match_call_p): Remove.
8002 (vect_detect_pair_op): Add crosslane check.
8003 (vect_match_call_complex_mla): Remove.
8004 (class complex_mul_pattern): Update comment.
8005 (complex_mul_pattern::matches): Update detection.
8006 (class complex_fma_pattern): Remove.
8007 (complex_fma_pattern::matches): Remove.
8008 (complex_fma_pattern::recognize): Remove.
8009 (complex_fma_pattern::build): Remove.
8010 (class complex_fms_pattern): Update comment.
8011 (complex_fms_pattern::matches): Remove.
8012 (complex_operations_pattern::recognize): Remove complex_fma_pattern
8014 2021-10-29 Jakub Jelinek <jakub@redhat.com>
8016 * gimple-fold.c (gimple_fold_builtin_memset): Copy over location from
8019 2021-10-29 Xionghu Luo <luoxhu@linux.ibm.com>
8022 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Add
8023 patterns match and emit for VSX xxpermdi.
8025 2021-10-29 liuhongt <hongtao.liu@intel.com>
8028 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF): New
8030 (V16HF_FTYPE_V16HF): Ditto.
8031 (V32HF_FTYPE_V32HF): Ditto.
8032 (V8HF_FTYPE_V8HF_ROUND): Ditto.
8033 (V16HF_FTYPE_V16HF_ROUND): Ditto.
8034 (V32HF_FTYPE_V32HF_ROUND): Ditto.
8035 * config/i386/i386-builtin.def ( IX86_BUILTIN_FLOORPH,
8036 IX86_BUILTIN_CEILPH, IX86_BUILTIN_TRUNCPH,
8037 IX86_BUILTIN_FLOORPH256, IX86_BUILTIN_CEILPH256,
8038 IX86_BUILTIN_TRUNCPH256, IX86_BUILTIN_FLOORPH512,
8039 IX86_BUILTIN_CEILPH512, IX86_BUILTIN_TRUNCPH512): New builtin.
8040 * config/i386/i386-builtins.c
8041 (ix86_builtin_vectorized_function): Enable vectorization for
8042 HFmode FLOOR/CEIL/TRUNC operation.
8043 * config/i386/i386-expand.c (ix86_expand_args_builtin): Handle
8045 * config/i386/sse.md (rint<mode>2, nearbyint<mode>2): Extend
8048 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
8049 Andrew MacLeod <amacleod@redhat.com>
8051 * value-relation.cc (path_oracle::killing_def): Walk the
8052 equivalency list and remove SSA from any equivalencies.
8054 2021-10-28 Stafford Horne <shorne@gmail.com>
8056 * config/or1k/or1k.h (PROFILE_HOOK): Add return address argument
8059 2021-10-28 Jakub Jelinek <jakub@redhat.com>
8061 PR tree-optimization/102951
8062 * fold-const.h (address_compare): Declare.
8063 * fold-const.c (address_compare): New function.
8064 * match.pd (cmp (convert1?@2 addr@0) (convert2? addr@1)): Use
8065 address_compare helper.
8066 (minmax cmp (convert1?@2 addr@0) (convert2?@3 addr@1)): New
8069 2021-10-28 Andrew MacLeod <amacleod@redhat.com>
8071 * vr-values.c (simplify_using_ranges::fold_cond): Change fold message.
8073 2021-10-28 Andrew MacLeod <amacleod@redhat.com>
8075 PR tree-optimization/102940
8076 * tree-vrp.c (execute_ranger_vrp): Reset scev.
8078 2021-10-28 Richard Purdie <richard.purdie@linuxfoundation.org>
8080 * config/nios2/linux.h (MUSL_DYNAMIC_LINKER): Add musl linker
8082 2021-10-28 Richard Purdie <richard.purdie@linuxfoundation.org>
8084 * configure: Regenerate.
8085 * configure.ac: Use CPPFLAGS_FOR_BUILD for GMPINC
8087 2021-10-28 Eric Botcazou <ebotcazou@adacore.com>
8089 * doc/invoke.texi (%X): Remove obsolete reference to -Wl.
8091 2021-10-28 Richard Biener <rguenther@suse.de>
8094 * fold-const.c (fold_convert_const): Avoid int to float
8095 constant folding with -frounding-math and inexact result.
8096 * simplify-rtx.c (simplify_const_unary_operation): Likewise
8097 for both float and unsigned_float.
8099 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
8101 * tree-ssa-threadbackward.c
8102 (back_threader::find_taken_edge_switch): Use find_case_label_range
8103 instead of find_taken_edge.
8105 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
8107 * tree-ssa-threadbackward.c (class back_threader_registry):
8108 Inherit from back_jt_path_registry.
8109 (back_threader_registry::thread_through_all_blocks): Remove.
8110 (back_threader_registry::register_path): Remove
8111 m_lowlevel_registry prefix.
8113 2021-10-28 Richard Biener <rguenther@suse.de>
8116 * fold-const.c (fold_convert_const_real_from_real): Honor
8117 -frounding-math if the conversion is not exact.
8118 * simplify-rtx.c (simplify_const_unary_operation): Do not
8119 simplify FLOAT_TRUNCATE with sign dependent rounding.
8121 2021-10-28 Richard Biener <rguenther@suse.de>
8123 PR tree-optimization/102949
8124 * tree-vect-stmts.c (ensure_base_align): Look at the
8125 dr_info of a group leader and assert we are looking at
8126 one with analyzed alignment.
8128 2021-10-28 Kewen Lin <linkw@linux.ibm.com>
8131 * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Consider
8132 V1T1 mode for unaligned load and store.
8134 2021-10-28 Kito Cheng <kito.cheng@sifive.com>
8136 * config/riscv/riscv.md (zero_extendsidi2_internal): Allow ZBB
8139 2021-10-28 Kito Cheng <kito.cheng@sifive.com>
8141 * config/riscv/arch-canonicalize (CANONICAL_ORDER): Add `i` to
8144 2021-10-28 Alexandre Oliva <oliva@adacore.com>
8146 * common.opt (fharden-compares): New.
8147 (fharden-conditional-branches): New.
8148 * doc/invoke.texi: Document new options.
8149 * gimple-harden-conditionals.cc: New.
8150 * Makefile.in (OBJS): Build it.
8151 * passes.def: Add new passes.
8152 * tree-pass.h (make_pass_harden_compares): Declare.
8153 (make_pass_harden_conditional_branches): Declare.
8155 2021-10-28 Xionghu Luo <luoxhu@linux.ibm.com>
8158 * config/rs6000/altivec.md: Add vsx register constraints.
8159 * config/rs6000/vsx.md (vsx_xxsel<mode>): Delete.
8160 (vsx_xxsel<mode>2): Likewise.
8161 (vsx_xxsel<mode>3): Likewise.
8162 (vsx_xxsel<mode>4): Likewise.
8164 2021-10-28 Xionghu Luo <luoxhu@linux.ibm.com>
8167 * config/rs6000/altivec.md (*altivec_vsel<mode>): Change to ...
8168 (altivec_vsel<mode>): ... this and update define.
8169 (*altivec_vsel<mode>_uns): Delete.
8170 (altivec_vsel<mode>2): New define_insn.
8171 (altivec_vsel<mode>3): Likewise.
8172 (altivec_vsel<mode>4): Likewise.
8173 * config/rs6000/rs6000-call.c (altivec_expand_vec_sel_builtin): New.
8174 (altivec_expand_builtin): Call altivec_expand_vec_sel_builtin to expand
8176 * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Use bit-wise
8177 selection instead of per element.
8178 * config/rs6000/vector.md:
8179 * config/rs6000/vsx.md (*vsx_xxsel<mode>): Change to ...
8180 (vsx_xxsel<mode>): ... this and update define.
8181 (*vsx_xxsel<mode>_uns): Delete.
8182 (vsx_xxsel<mode>2): New define_insn.
8183 (vsx_xxsel<mode>3): Likewise.
8184 (vsx_xxsel<mode>4): Likewise.
8186 2021-10-28 Hongyu Wang <hongyu.wang@intel.com>
8188 * config/i386/i386.c (use_rsqrt_p): Add mode parameter, enable
8189 HFmode rsqrt without TARGET_SSE_MATH.
8190 (ix86_optab_supported_p): Refactor rint, adjust floor, ceil,
8191 btrunc condition to be restricted by -ftrapping-math, adjust
8192 use_rsqrt_p function call.
8193 * config/i386/i386.md (rcphf2): New define_insn.
8194 (rsqrthf2): Likewise.
8195 * config/i386/sse.md (div<mode>3): Change VF2H to VF2.
8196 (div<mode>3): New expander for HF mode.
8197 (rsqrt<mode>2): Likewise.
8198 (*avx512fp16_vmrcpv8hf2): New define_insn for rpad pass.
8199 (*avx512fp16_vmrsqrtv8hf2): Likewise.
8201 2021-10-27 Saagar Jha <saagar@saagarjha.com>
8203 * config.gcc: Adjust for Darwin21.
8204 * config/darwin-c.c (macosx_version_as_macro): Likewise.
8205 * config/darwin-driver.c (validate_macosx_version_min):
8207 (darwin_find_version_from_kernel): Likewise.
8209 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
8211 * gimple-range-path.cc
8212 (path_range_query::range_defined_in_block): Call killing_def.
8214 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
8216 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Dump
8217 operands as well as relation.
8218 * gimple-range-path.cc
8219 (path_range_query::compute_ranges_in_block): Compute PHI relations
8220 first. Compute outgoing relations at the end.
8221 (path_range_query::compute_ranges): Remove call to compute_relations.
8222 (path_range_query::compute_relations): Remove.
8223 (path_range_query::maybe_register_phi_relation): New.
8224 (path_range_query::compute_phi_relations): Abstract out
8225 registering one PHI relation to...
8226 (path_range_query::compute_outgoing_relations): ...here.
8227 * gimple-range-path.h (class path_range_query): Remove
8229 Add maybe_register_phi_relation.
8231 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
8232 Andrew MacLeod <amacleod@redhat.com>
8234 * value-relation.cc (path_oracle::killing_def): Kill second
8237 2021-10-27 Martin Jambor <mjambor@suse.cz>
8239 * ipa-cp.c (good_cloning_opportunity_p): Decide whether to use
8240 profile feedback depending on their local availability.
8242 2021-10-27 Martin Jambor <mjambor@suse.cz>
8244 * params.opt (param_ipa_cp_profile_count_base): New parameter.
8245 * doc/invoke.texi (Optimize Options): Add entry for
8246 ipa-cp-profile-count-base.
8247 * ipa-cp.c (max_count): Replace with base_count, replace all
8248 occurrences too, unless otherwise stated.
8249 (ipcp_cloning_candidate_p): identify mostly-directly called
8250 functions based on their counts, not max_count.
8251 (compare_edge_profile_counts): New function.
8252 (ipcp_propagate_stage): Instead of setting max_count, find the
8253 appropriate edge count in a sorted vector of counts of eligible
8254 edges and make it the base_count.
8256 2021-10-27 Martin Jambor <mjambor@suse.cz>
8258 * ipa-cp.c (struct caller_statistics): New fields rec_count_sum,
8259 n_nonrec_calls and itself, document all fields.
8260 (init_caller_stats): Initialize the above new fields.
8261 (gather_caller_stats): Gather self-recursive counts and calls number.
8262 (get_info_about_necessary_edges): Gather counts of self-recursive and
8263 other edges bringing in the requested value separately.
8264 (dump_profile_updates): Rework to dump info about a single node only.
8265 (lenient_count_portion_handling): New function.
8266 (struct gather_other_count_struct): New type.
8267 (gather_count_of_non_rec_edges): New function.
8268 (struct desc_incoming_count_struct): New type.
8269 (analyze_clone_icoming_counts): New function.
8270 (adjust_clone_incoming_counts): Likewise.
8271 (update_counts_for_self_gen_clones): Likewise.
8272 (update_profiling_info): Rewritten.
8273 (update_specialized_profile): Adjust call to dump_profile_updates.
8274 (create_specialized_node): Do not update profiling info.
8275 (decide_about_value): New parameter self_gen_clones, either push new
8276 clones into it or updat their profile counts. For self-recursively
8277 generated values, use a portion of the node count instead of count
8278 from self-recursive edges to estimate goodness.
8279 (decide_whether_version_node): Gather clones for self-generated values
8280 in a new vector, update their profiles at once at the end.
8282 2021-10-27 Richard Biener <rguenther@suse.de>
8284 * tree-vectorizer.c (vect_transform_loops): New function,
8286 (try_vectorize_loop_1): ... here. Simplify as epilogues
8287 are now fully handled in the split part.
8289 2021-10-27 Jakub Jelinek <jakub@redhat.com>
8291 * omp-expand.c (expand_omp_for_init_counts): Handle non-rectangular
8292 iterators with pointer types.
8293 (expand_omp_for_init_vars, extract_omp_for_update_vars): Likewise.
8295 2021-10-26 Martin Sebor <msebor@redhat.com>
8297 PR tree-optimization/102238
8298 PR tree-optimization/102919
8299 * gimple-ssa-sprintf.c (get_string_length): Add an argument.
8300 (array_elt_at_offset): Move to pointer-query.
8301 (set_aggregate_size_and_offset): New function.
8302 (field_at_offset): Move to pointer-query.
8303 (get_origin_and_offset): Rename...
8304 (get_origin_and_offset_r): this. Add an argument. Make aggregate
8305 handling more robust.
8306 (get_origin_and_offset): New.
8307 (alias_offset): Add an argument.
8308 (format_string): Use subobject size determined by get_origin_and_offset.
8309 * pointer-query.cc (field_at_offset): Move from gimple-ssa-sprintf.c.
8310 Improve/correct handling of aggregates.
8311 (array_elt_at_offset): Same.
8312 * pointer-query.h (field_at_offset): Declare.
8313 (array_elt_at_offset): Declare.
8315 2021-10-26 Martin Sebor <msebor@redhat.com>
8317 * builtins.c (check_strncat_sizes): Pass access_data ctor additional
8319 (expand_builtin_memcmp): Move code to gimple-ssa-warn-access.cc.
8320 (expand_builtin_fork_or_exec): Same.
8321 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Pass
8322 compute_objsize additional arguments.
8323 (inbounds_memaccess_p): Same.
8324 (array_bounds_checker::check_array_bounds): Add an assert. Stash
8325 statement in a member.
8326 (check_array_bounds_dom_walker::before_dom_children): Same.
8327 * gimple-array-bounds.h (array_bounds_checker::m_stmt): New member.
8328 * gimple-ssa-sprintf.c (get_destination_size): Add an argument.
8329 (handle_printf_call): Pass a new argument.
8330 * gimple-ssa-warn-access.cc (get_size_range): Add an argument.
8331 (check_access): Add an argument and pass it along to callees.
8332 (check_read_access): Make a member function.
8333 (pass_waccess::check_strcat): Pass access_data ctor additional
8335 (pass_waccess::check_strncat): Same.
8336 (pass_waccess::check_stxcpy): Same.
8337 (pass_waccess::check_stxncpy): Same.
8338 (pass_waccess::check_strncmp): Same.
8339 (pass_waccess::check_read_access): Same.
8340 (pass_waccess::check_builtin): Same.
8341 (pass_waccess::maybe_check_access_sizes): Same.
8342 (pass_waccess::maybe_check_dealloc_call): Same.
8343 * gimple-ssa-warn-access.h (check_read_access): Declare a new
8345 * pointer-query.cc (compute_objsize_r): Add an argument.
8346 (gimple_call_return_array): Same.
8347 (gimple_call_alloc_size): Same.
8348 (access_ref::access_ref): Same.
8349 (access_ref::get_ref): Same.
8350 (pointer_query::get_ref): Same.
8351 (handle_min_max_size): Pass an arguments to callees.
8352 (handle_array_ref): Add an argument.
8353 (handle_mem_ref): Same.
8354 (compute_objsize): Same.
8355 * pointer-query.h (struct access_ref): Adjust signatures.
8356 (struct access_data): Same.
8357 (gimple_call_alloc_size): Add an argument.
8358 (gimple_parm_array_size): Same.
8359 (compute_objsize): Same.
8360 * tree-ssa-strlen.c (strlen_pass::adjust_last_stmt): Pass an additional
8361 argument to compute_objsize.
8362 (strlen_pass::maybe_warn_overflow): Same.
8363 (maybe_diag_stxncpy_trunc): Same.
8365 2021-10-26 Martin Sebor <msebor@redhat.com>
8367 PR middle-end/102453
8368 * gimple-ssa-warn-access.cc (pass_waccess::check_atomic_builtin): New.
8369 (pass_waccess::check_atomic_builtin): Call it.
8371 2021-10-26 Vladimir N. Makarov <vmakarov@redhat.com>
8373 PR rtl-optimization/102842
8374 * lra-constraints.c (match_reload): Ignore out in checking values
8376 (curr_insn_transform): Collect outputs before doing reloads of operands.
8378 2021-10-26 Paul A. Clarke <pc@us.ibm.com>
8381 * config/rs6000/x86intrin.h: Move some included headers to new
8382 headers. Include new immintrin.h instead of those headers.
8383 * config/rs6000/immintrin.h: New.
8384 * config/rs6000/x86gprintrin.h: New.
8385 * config.gcc (powerpc*-*-*): Add new headers to extra_headers.
8387 2021-10-26 Richard Biener <rguenther@suse.de>
8389 * tree-vectorizer.h (vect_create_addr_base_for_vector_ref):
8390 Remove byte_offset parameter.
8391 (vect_create_data_ref_ptr): Likewise.
8392 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
8394 (vect_create_data_ref_ptr): Likewise.
8395 * tree-vect-stmts.c (vectorizable_store): Adjust.
8396 (vectorizable_load): Likewise.
8398 2021-10-26 Richard Biener <rguenther@suse.de>
8400 PR tree-optimization/96109
8401 * tree-vectorizer.h (dr_misalignment): Add optional offset
8403 * tree-vect-data-refs.c (dr_misalignment): Likewise. Remove
8404 offset applied for negative stride accesses.
8405 (vect_enhance_data_refs_alignment): Compute negative stride
8406 access offset and pass it to dr_misalignment.
8407 * tree-vect-stmts.c (get_negative_load_store_type): Pass
8408 negative offset to dr_misalignment.
8409 (get_group_load_store_type): Likewise.
8410 (get_load_store_type): Likewise.
8411 (vectorizable_store): Remove asserts about alignment.
8412 (vectorizable_load): Likewise.
8414 2021-10-26 Kewen Lin <linkw@linux.ibm.com>
8416 PR tree-optimization/102897
8417 * tree-ssa-forwprop.c (simplify_permutation): Remove a wrong assertion.
8419 2021-10-26 Richard Biener <rguenther@suse.de>
8421 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
8422 Take offset in bytes.
8423 (vect_create_data_ref_ptr): Likewise.
8424 * tree-vect-loop-manip.c (get_misalign_in_elems): Multiply
8425 offset by element size.
8426 (vect_create_cond_for_align_checks): Likewise.
8427 * tree-vect-stmts.c (get_negative_load_store_type): Likewise.
8428 (vectorizable_load): Remove duplicate leftover from merge
8431 2021-10-26 Roger Sayle <roger@nextmovesoftware.com>
8433 * config/i386/i386-expand.c (ix86_expand_v1ti_shift): New helper
8434 function to expand V1TI mode logical shifts by integer constants.
8435 (ix86_expand_v1ti_rotate): New helper function to expand V1TI
8436 mode rotations by integer constants.
8437 * config/i386/i386-protos.h (ix86_expand_v1ti_shift,
8438 ix86_expand_v1ti_rotate): Prototype new functions here.
8439 * config/i386/sse.md (ashlv1ti3, lshrv1ti3, rotlv1ti3, rotrv1ti3):
8440 New TARGET_SSE2 expanders to implement V1TI shifts and rotations.
8442 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
8444 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
8445 Avoid threading circular paths.
8447 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
8449 * tree-ssa-threadbackward.c (back_threader::resolve_phi):
8450 Attempt to resolve all incoming paths to a PHI.
8451 (back_threader::resolve_def): Always return true for PHIs.
8453 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
8455 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
8456 Always try to resolve path without looking back.
8457 * tree-ssa-threadupdate.c (dump_jump_thread): Indidicate whether
8458 edge is a back edge.
8460 2021-10-26 Kewen Lin <linkw@linux.ibm.com>
8462 PR tree-optimization/102789
8463 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
8464 update inits of simd_lane_access.
8466 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
8468 * tree-vrp.c (vrp_simplify_cond_using_ranges): Add return type and
8469 move to vr-values.c.
8470 (simplify_casted_conds): Move to vrp_folder class.
8471 (execute_vrp): Call via vrp_folder now.
8472 * vr-values.c (simplify_cond_using_ranges_1): Call simplify_casted_cond.
8473 (simplify_using_ranges::simplify_casted_cond): Relocate from tree-vrp.c.
8474 * vr-values.h (simplify_casted_cond): Add prototype.
8476 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
8478 * tree-vrp.c (rvrp_folder::fold_stmt): If simplification fails, try
8481 2021-10-25 Paul A. Clarke <pc@us.ibm.com>
8483 * config/rs6000/smmintrin.h (_mm_testz_si128): Add "extern" to
8485 (_mm_testc_si128): Likewise.
8486 (_mm_testnzc_si128): Likewise.
8487 (_mm_blend_ps): Likewise.
8488 (_mm_blendv_ps): Likewise.
8489 (_mm_blend_pd): Likewise.
8490 (_mm_blendv_pd): Likewise.
8491 (_mm_ceil_pd): Likewise.
8492 (_mm_ceil_sd): Likewise.
8493 (_mm_ceil_ps): Likewise.
8494 (_mm_ceil_ss): Likewise.
8495 (_mm_floor_pd): Likewise.
8496 (_mm_floor_sd): Likewise.
8497 (_mm_floor_ps): Likewise.
8498 (_mm_floor_ss): Likewise.
8499 (_mm_minpos_epu16): Likewise.
8500 (_mm_mul_epi32): Likewise.
8501 (_mm_cvtepi8_epi16): Likewise.
8502 (_mm_packus_epi32): Likewise.
8503 (_mm_cmpgt_epi64): Likewise.
8505 2021-10-25 Roger Sayle <roger@nextmovesoftware.com>
8507 * simplify-rtx.c (simplify_binary_operation_1) [SS_ASHIFT]: Simplify
8508 shifts of the mode's smin_value and smax_value when the bit count
8509 operand doesn't have side-effects.
8510 [US_ASHIFT]: Likewise, simplify shifts of the mode's umax_value
8511 when the bit count operand doesn't have side-effects.
8512 (simplify_const_binary_operation) [SS_ASHIFT, US_ASHIFT]: Perform
8513 compile-time evaluation of saturating left shifts with constant
8516 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
8518 * gimple-range.cc (gimple_ranger::export_global_ranges): Remove check
8521 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
8523 * flag-types.h (enum ranger_debug): Adjust values.
8524 * params.opt (ranger_debug): Ditto.
8526 2021-10-25 Martin Jambor <mjambor@suse.cz>
8528 PR tree-optimization/102886
8529 * tree-sra.c (totally_scalarize_subtree): Fix the out of
8532 2021-10-25 Andrew Pinski <apinski@marvell.com>
8534 * tree-ssa-dce.c (simple_dce_from_worklist):
8535 Check stmt_unremovable_because_of_non_call_eh_p also
8536 before removing the statement.
8538 2021-10-25 Richard Biener <rguenther@suse.de>
8540 PR tree-optimization/102905
8541 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
8542 Use vect_supportable_dr_alignment again to determine whether
8543 an access is supported when not aligned.
8545 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
8547 * config/riscv/riscv.c (riscv_rtx_costs): Handle cost model
8550 2021-10-25 Jim Wilson <jimw@sifive.com>
8551 Kito Cheng <kito.cheng@sifive.com>
8552 Jia-Wei Chen <jiawei@iscas.ac.cn>
8553 Shi-Hua Liao <shihua@iscas.ac.cn>
8555 * config/riscv/bitmanip.md (shiftm1): New.
8556 (*bset<mode>): Ditto.
8557 (*bset<mode>_mask): Ditto.
8558 (*bset<mode>_1): Ditto.
8559 (*bset<mode>_1_mask): Ditto.
8560 (*bseti<mode>): Ditto.
8561 (*bclr<mode>): Ditto.
8562 (*bclri<mode>): Ditto.
8563 (*binv<mode>): Ditto.
8564 (*binvi<mode>): Ditto.
8565 (*bext<mode>): Ditto.
8567 * config/riscv/predicates.md (splittable_const_int_operand):
8569 (single_bit_mask_operand): New.
8570 (not_single_bit_mask_operand): Ditto.
8571 (const31_operand): Ditto.
8572 (const63_operand): Ditto.
8573 * config/riscv/riscv.c (riscv_build_integer_1): Handle bseti.
8574 (riscv_output_move): Ditto.
8575 (riscv_print_operand): Handle new operand type: T and S.
8576 * config/riscv/riscv.h (SINGLE_BIT_MASK_OPERAND): New.
8578 2021-10-25 Jim Wilson <jimw@sifive.com>
8580 * config/riscv/riscv.c (riscv_build_integer_1): Build integer
8583 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
8585 * config/riscv/riscv.c (riscv_extend_cost): Handle cost model
8587 (riscv_rtx_costs): Ditto.
8589 2021-10-25 Jim Wilson <jimw@sifive.com>
8590 Kito Cheng <kito.cheng@sifive.com>
8591 Jia-Wei Chen <jiawei@iscas.ac.cn>
8593 * config/riscv/bitmanip.md (bitmanip_bitwise): New.
8594 (bitmanip_minmax): New.
8595 (clz_ctz_pcnt): New.
8596 (bitmanip_optab): New.
8597 (bitmanip_insn): New.
8598 (*<optab>_not<mode>): New.
8599 (*xor_not<mode>): New.
8600 (<bitmanip_optab>si2): New.
8601 (*<bitmanip_optab>disi2): New.
8602 (<bitmanip_optab>di2): New.
8603 (*zero_extendhi<GPR:mode>2_bitmanip): New.
8604 (*extend<SHORT:mode><SUPERQI:mode>2_zbb): New.
8605 (*zero_extendhi<GPR:mode>2_zbb): New.
8608 (rotrsi3_sext): New.
8611 (rotlsi3_sext): New.
8612 (bswap<mode>2): New.
8613 (<bitmanip_optab><mode>3): New.
8614 * config/riscv/riscv.md (type): Add rotate.
8615 (zero_extendhi<GPR:mode>2): Change to define_expand pattern.
8616 (*zero_extendhi<GPR:mode>2): New.
8617 (extend<SHORT:mode><SUPERQI:mode>2): Change to define_expand pattern.
8618 (*extend<SHORT:mode><SUPERQI:mode>2): New.
8620 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
8622 * config/riscv/riscv.c (riscv_extend_cost): Handle cost model
8624 (riscv_rtx_costs): Ditto.
8626 2021-10-25 Jim Wilson <jimw@sifive.com>
8627 Kito Cheng <kito.cheng@sifive.com>
8628 Jia-Wei Chen <jiawei@iscas.ac.cn>
8630 * config/riscv/bitmanip.md (*zero_extendsidi2_bitmanip): New.
8635 (riscv_rtx_costs): Ditto.
8636 * config/riscv/riscv.md: Include bitmanip.md
8637 (type): Add bitmanip bype.
8638 (zero_extendsidi2): Change to define_expand pattern.
8639 (*zero_extendsidi2_internal): New.
8640 (zero_extendsidi2_shifted): Disable for ZBA.
8642 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
8644 * common/config/riscv/riscv-common.c (riscv_ext_version_table):
8645 Add zba, zbb, zbc and zbs.
8646 (riscv_ext_flag_table): Ditto.
8647 * config/riscv/riscv-opts.h (MASK_ZBA): New.
8651 (TARGET_ZBA): Ditto.
8652 (TARGET_ZBB): Ditto.
8653 (TARGET_ZBC): Ditto.
8654 (TARGET_ZBS): Ditto.
8655 * config/riscv/riscv.opt (riscv_zb_subext): New.
8657 2021-10-25 liuhongt <hongtao.liu@intel.com>
8660 * match.pd: Simplify (_Float16) sqrtf((float) a) to .SQRT(a)
8661 when direct_internal_fn_supported_p, similar for sqrt/sqrtl.
8663 2021-10-25 Richard Biener <rguenther@suse.de>
8665 PR tree-optimization/102920
8666 * tree-ssa-sccvn.h (expressions_equal_p): Add argument
8667 controlling VN_TOP matching behavior.
8668 * tree-ssa-sccvn.c (expressions_equal_p): Likewise.
8669 (vn_phi_eq): Do not optimistically match VN_TOP.
8671 2021-10-25 konglin1 <lingling.kong@intel.com>
8673 * config/i386/sse.md (fma_<mode>_fadd_fmul): Add new
8674 define_insn_and_split.
8675 (fma_<mode>_fadd_fcmul):Likewise
8676 (fma_<complexopname>_<mode>_fma_zero):Likewise
8678 2021-10-24 John David Anglin <danglin@gcc.gnu.org>
8680 * config/pa/pa-d.c (pa_d_handle_target_float_abi): Don't check
8681 TARGET_DISABLE_FPREGS.
8682 * config/pa/pa.c (fix_range): Use MASK_SOFT_FLOAT instead of
8683 MASK_DISABLE_FPREGS.
8684 (hppa_rtx_costs): Don't check TARGET_DISABLE_FPREGS. Adjust
8685 cost of hardware integer multiplication.
8686 (pa_conditional_register_usage): Don't check TARGET_DISABLE_FPREGS.
8687 * config/pa/pa.h (INT14_OK_STRICT): Likewise.
8688 * config/pa/pa.md: Don't check TARGET_DISABLE_FPREGS. Check
8689 TARGET_SOFT_FLOAT in patterns that use xmpyu instruction.
8690 * config/pa/pa.opt (mdisable-fpregs): Change target mask to
8691 SOFT_FLOAT. Revise comment.
8692 (msoft-float): New option.
8694 2021-10-24 John David Anglin <danglin@gcc.gnu.org>
8696 * config/pa/pa.md: Don't use 'G' constraint in integer move patterns.
8698 2021-10-24 Gerald Pfeifer <gerald@pfeifer.com>
8700 * doc/install.texi (Specific): Remove obsolete details
8701 around GNU/Linux on Itanium.
8702 (Specific): Remove reference to Windows for Itanium.
8704 2021-10-23 Bernhard Reutner-Fischer <aldot@gcc.gnu.org>
8706 * config/i386/x86-tune-sched-bd.c (dispatch_group): Commentary
8709 2021-10-23 Jan Hubicka <hubicka@ucw.cz>
8711 * tree-ssa-structalias.c (compute_points_to_sets): Cleanup.
8713 2021-10-23 Roger Sayle <roger@nextmovesoftware.com>
8715 * config/i386/sse.md (<any_logic>v1ti3): New define_insn to
8716 implement V1TImode AND, IOR and XOR on TARGET_SSE2 (and above).
8717 (one_cmplv1ti2): New define expand.
8719 2021-10-22 Eric Gallager <egallager@gcc.gnu.org>
8722 * Makefile.in: Handle dvidir and install-dvi target.
8723 * configure: Regenerate.
8724 * configure.ac: Add install-dvi to target_list.
8726 2021-10-22 Gerald Pfeifer <gerald@pfeifer.com>
8728 * doc/install.texi (Binaries): Convert mingw-w64.org to https.
8731 2021-10-22 Richard Biener <rguenther@suse.de>
8733 PR tree-optimization/102893
8734 * tree-ssa-dce.c (find_obviously_necessary_stmts): Fix the
8735 test for an exit edge.
8737 2021-10-22 Aldy Hernandez <aldyh@redhat.com>
8738 Andrew MacLeod <amacleod@redhat.com>
8740 * gimple-range-path.cc (path_range_query::compute_phi_relations):
8741 Kill any global relations we may know before registering a new
8743 * value-relation.cc (path_oracle::killing_def): New.
8744 * value-relation.h (path_oracle::killing_def): New.
8746 2021-10-22 Richard Biener <rguenther@suse.de>
8749 * tree-ssa-sccvn.c (vn_phi_insert): For undefined SSA args
8751 (vn_phi_lookup): Likewise.
8753 2021-10-21 H.J. Lu <hjl.tools@gmail.com>
8756 * doc/invoke.texi: Document -fcf-protection requires i686 or
8759 2021-10-21 Eric Botcazou <ebotcazou@adacore.com>
8761 PR middle-end/102764
8762 * cfgexpand.c (expand_gimple_basic_block): Robustify latest change.
8764 2021-10-21 Jonathan Wright <jonathan.wright@arm.com>
8766 * config/aarch64/arm_neon.h (__STRUCTN): Delete function
8767 macro and all invocations.
8769 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
8771 * doc/invoke.texi (ranger-debug): Document.
8772 * flag-types.h (enum ranger_debug): New.
8773 (enum evrp_mode): Remove debug values.
8774 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Use new debug flag.
8775 * gimple-range-gori.cc (gori_compute::gori_compute): Ditto.
8776 * gimple-range.cc (gimple_ranger::gimple_ranger): Ditto.
8777 * gimple-ssa-evrp.c (hybrid_folder::choose_value): Ditto.
8778 (execute_early_vrp): Use evrp-mode directly.
8779 * params.opt (enum evrp_mode): Remove debug values.
8780 (ranger-debug): New.
8781 (ranger-logical-depth): Relocate to be in alphabetical order.
8783 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
8785 * doc/invoke.texi: (vrp1-mode, vrp2-mode): Document.
8786 * flag-types.h: (enum vrp_mode): New.
8787 * params.opt: (vrp1-mode, vrp2-mode): New.
8788 * tree-vrp.c (vrp_pass_num): New.
8789 (pass_vrp::pass_vrp): Set pass number.
8790 (pass_vrp::execute): Choose which VRP mode to execute.
8792 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
8794 * gimple-ssa-evrp.c (class rvrp_folder): Move to tree-vrp.c.
8795 (execute_early_vrp): For ranger only mode, invoke ranger_vrp.
8796 * tree-vrp.c (class rvrp_folder): Relocate here.
8797 (execute_ranger_vrp): New.
8798 * tree-vrp.h (execute_ranger_vrp): Export.
8800 2021-10-21 Martin Liska <mliska@suse.cz>
8804 * opts.c (finish_options): Process flag_var_tracking* options
8805 here as they can be adjusted by optimize attribute.
8806 Process also flag_syntax_only and flag_gtoggle.
8807 * toplev.c (process_options): Remove it here.
8808 * common.opt: Make debug_nonbind_markers_p as PerFunction
8809 attribute as it depends on optimization level.
8811 2021-10-21 Martin Jambor <mjambor@suse.cz>
8813 PR tree-optimization/102505
8814 * tree-sra.c (totally_scalarize_subtree): Check that the
8815 encountered field fits within the acces we would like to put it
8818 2021-10-21 Aldy Hernandez <aldyh@redhat.com>
8820 * tree-ssa-threadbackward.c
8821 (back_threader::maybe_register_path): Remove circular paths check.
8823 2021-10-21 Richard Biener <rguenther@suse.de>
8825 * toplev.c (process_options): Move the initial debug_hooks
8827 (toplev::main): ... before the call of the post_options
8830 2021-10-21 Richard Biener <rguenther@suse.de>
8832 PR tree-optimization/102847
8833 * tree-vect-stmts.c (vect_model_load_cost): Add the scalar
8834 load cost in the prologue for VMAT_INVARIANT.
8836 2021-10-21 Richard Biener <rguenther@suse.de>
8838 PR tree-optimization/102847
8839 * tree-vect-stmts.c (vect_model_load_cost): Explicitely
8840 handle VMAT_INVARIANT as a splat in the prologue.
8842 2021-10-21 Hongyu Wang <hongyu.wang@intel.com>
8845 * config/i386/i386.c (ix86_get_ssemov): Adjust HFmode vector
8846 move to use the same logic as HImode.
8848 2021-10-21 Richard Biener <rguenther@suse.de>
8850 * tree-vect-slp.c (vect_build_slp_tree_1): Remove
8851 superfluous gimple_call_nothrow_p check.
8853 2021-10-21 Andrew Pinski <apinski@marvell.com>
8855 * tree-cfg.c (maybe_remove_writeonly_store): Add dce_ssa_names argument.
8856 Mark the ssa-name of the rhs as one to be removed.
8857 (execute_fixup_cfg): Update call to maybe_remove_writeonly_store.
8858 Call simple_dce_from_worklist at the end to a simple dce.
8860 2021-10-21 Andrew Pinski <apinski@marvell.com>
8862 * tree-cfg.c (maybe_remove_writeonly_store): New function
8863 factored out from ...
8864 (execute_fixup_cfg): Here. Call maybe_remove_writeonly_store.
8866 2021-10-21 Andrew Pinski <apinski@marvell.com>
8868 * tree-cfg.c (execute_fixup_cfg): Remove comment
8869 about standalone pass.
8871 2021-10-21 Andrew Pinski <apinski@marvell.com>
8873 * tree-cfg.c (execute_fixup_cfg): Output when the statement
8874 is removed when it is a write only var.
8876 2021-10-21 Aldy Hernandez <aldyh@redhat.com>
8878 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
8879 Avoid threading circular paths.
8881 2021-10-20 Alex Coplan <alex.coplan@arm.com>
8883 * calls.c (initialize_argument_information): Remove some dead
8884 code, remove handling for function_arg returning const_int.
8885 * doc/tm.texi: Delete documentation for unused target hooks.
8886 * doc/tm.texi.in: Likewise.
8887 * target.def (load_bounds_for_arg): Delete.
8888 (store_bounds_for_arg): Delete.
8889 (load_returned_bounds): Delete.
8890 (store_returned_bounds): Delete.
8891 * targhooks.c (default_load_bounds_for_arg): Delete.
8892 (default_store_bounds_for_arg): Delete.
8893 (default_load_returned_bounds): Delete.
8894 (default_store_returned_bounds): Delete.
8895 * targhooks.h (default_load_bounds_for_arg): Delete.
8896 (default_store_bounds_for_arg): Delete.
8897 (default_load_returned_bounds): Delete.
8898 (default_store_returned_bounds): Delete.
8900 2021-10-20 Jonathan Wakely <jwakely@redhat.com>
8902 * doc/extend.texi (Basic Asm): Clarify that asm is not an
8904 * doc/invoke.texi (-fno-asm): Fix description for C++.
8906 2021-10-20 Jonathan Wakely <jwakely@redhat.com>
8908 * doc/install.texi: Remove link to old.html
8910 2021-10-20 Tamar Christina <tamar.christina@arm.com>
8912 * config/aarch64/aarch64-simd.md (*aarch64_cmtst_same_<mode>): New.
8914 2021-10-20 Tamar Christina <tamar.christina@arm.com>
8916 * config/aarch64/aarch64-simd.md (*aarch64_narrow_trunc<mode>): New.
8918 2021-10-20 Tamar Christina <tamar.christina@arm.com>
8920 * config/aarch64/aarch64-simd.md (aarch64_simd_ashr<mode>): Add case cmp
8922 * config/aarch64/constraints.md (D1): New.
8924 2021-10-20 Tamar Christina <tamar.christina@arm.com>
8926 * config/aarch64/aarch64-simd.md
8927 (*aarch64_<srn_op>topbits_shuffle<mode>_le): New.
8928 (*aarch64_topbits_shuffle<mode>_le): New.
8929 (*aarch64_<srn_op>topbits_shuffle<mode>_be): New.
8930 (*aarch64_topbits_shuffle<mode>_be): New.
8931 * config/aarch64/predicates.md
8932 (aarch64_simd_shift_imm_vec_exact_top): New.
8934 2021-10-20 Tamar Christina <tamar.christina@arm.com>
8936 * config/aarch64/aarch64-simd.md (*aarch64_<srn_op>shrn<mode>_vect,
8937 *aarch64_<srn_op>shrn<mode>2_vect_le,
8938 *aarch64_<srn_op>shrn<mode>2_vect_be): New.
8939 * config/aarch64/iterators.md (srn_op): New.
8941 2021-10-20 Chung-Lin Tang <cltang@codesourcery.com>
8943 * omp-low.c (omp_copy_decl_2): For !ctx, use record_vars to add new copy
8945 (scan_sharing_clauses): Place copy of OMP_CLAUSE_IN_REDUCTION decl in
8946 ctx->outer instead of ctx.
8948 2021-10-20 Martin Liska <mliska@suse.cz>
8951 2021-10-19 Martin Liska <mliska@suse.cz>
8954 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
8955 * system.h (strip_whilespaces): New function.
8957 2021-10-20 Martin Liska <mliska@suse.cz>
8960 2021-10-19 Martin Liska <mliska@suse.cz>
8963 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
8966 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
8968 * config/aarch64/aarch64.c (aarch64_case_values_threshold):
8969 Change to 8 with -Os, 11 otherwise.
8971 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
8973 * config/aarch64/aarch64.c (neoversev1_tunings):
8974 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
8975 (neoversen2_tunings): Likewise.
8977 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
8980 * config/aarch64/aarch64.md (frint_pattern): Update comment.
8981 * config/aarch64/aarch64-simd-builtins.def: Change frintn to roundeven.
8982 * config/aarch64/arm_fp16.h: Change frintn to roundeven.
8983 * config/aarch64/arm_neon.h: Likewise.
8984 * config/aarch64/iterators.md (frint_pattern): Use roundeven for FRINTN.
8986 2021-10-20 Martin Liska <mliska@suse.cz>
8988 * config/arm/arm.c (arm_unwind_emit_sequence): Do not declare
8989 already declared global variable.
8990 (arm_unwind_emit_set): Use out_file as function argument.
8991 (arm_unwind_emit): Likewise.
8992 * config/darwin.c (machopic_output_data_section_indirection): Likewise.
8993 (machopic_output_stub_indirection): Likewise.
8994 (machopic_output_indirection): Likewise.
8995 (machopic_finish): Likewise.
8996 * config/i386/i386.c (ix86_asm_output_function_label): Likewise.
8997 * config/i386/winnt.c (i386_pe_seh_unwind_emit): Likewise.
8998 * config/ia64/ia64.c (process_epilogue): Likewise.
8999 (process_cfa_adjust_cfa): Likewise.
9000 (process_cfa_register): Likewise.
9001 (process_cfa_offset): Likewise.
9002 (ia64_asm_unwind_emit): Likewise.
9003 * config/s390/s390.c (s390_asm_output_function_label): Likewise.
9005 2021-10-20 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
9007 * config/aarch64/aarch64-builtins.c
9008 (aarch64_general_gimple_fold_builtin):
9009 lower vld1 and vst1 variants of the neon builtins
9010 * config/aarch64/aarch64-protos.h:
9011 (aarch64_general_gimple_fold_builtin): Add gsi parameter.
9012 * config/aarch64/aarch64.c (aarch64_general_gimple_fold_builtin):
9015 2021-10-20 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
9017 * match.pd: Generate IFN_TRUNC.
9019 2021-10-20 Richard Biener <rguenther@suse.de>
9021 PR tree-optimization/102853
9022 * tree-data-ref.c (split_constant_offset_1): Bail out
9023 immediately if the expression traps on overflow.
9025 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
9027 * tree-ssa-threadbackward.c (back_threader::~back_threader): Remove.
9029 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
9031 * tree-ssa-threadupdate.c (back_jt_path_registry::adjust_paths_after_duplication):
9032 Remove superflous debugging message.
9033 (back_jt_path_registry::duplicate_thread_path): Same.
9035 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
9037 * tree-ssa-threadbackward.c (back_threader_registry::back_threader_registry):
9039 (back_threader_registry::register_path): Remove m_threaded_paths.
9041 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
9043 PR tree-optimization/102814
9044 * doc/invoke.texi: Document --param=max-fsm-thread-length.
9045 * params.opt: Add --param=max-fsm-thread-length.
9046 * tree-ssa-threadbackward.c
9047 (back_threader_profitability::profitable_path_p): Fail on paths
9048 longer than max-fsm-thread-length.
9050 2021-10-20 Eric Botcazou <ebotcazou@adacore.com>
9052 PR middle-end/102764
9053 * cfgexpand.c (expand_gimple_basic_block): Disregard a final debug
9054 statement to reset the current location for the outgoing edges.
9056 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
9057 Richard Biener <rguenther@suse.de>
9059 * tree-ssa-threadupdate.c (cancel_thread): Dump threading reason
9060 on the same line as the threading cancellation.
9061 (jt_path_registry::cancel_invalid_paths): Avoid rotating loops.
9062 Avoid threading through loop headers where the path remains in the
9065 2021-10-20 Siddhesh Poyarekar <siddhesh@gotplt.org>
9067 * tree-object-size.c (unknown): Make into a function. Adjust
9069 (unknown_object_size): Simplify implementation.
9071 2021-10-20 Hongtao Liu <hongtao.liu@intel.com>
9072 Kewen Lin <linkw@linux.ibm.com>
9074 * doc/sourcebuild.texi (Effective-Target Keywords): Document
9075 vect_slp_v2qi_store, vect_slp_v4qi_store, vect_slp_v8qi_store,
9076 vect_slp_v16qi_store, vect_slp_v2hi_store,
9077 vect_slp_v4hi_store, vect_slp_v2si_store, vect_slp_v4si_store.
9079 2021-10-19 Jonathan Wakely <jwakely@redhat.com>
9081 * doc/extend.texi (Basic PowerPC Built-in Functions): Fix typo.
9083 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
9087 * config/rs6000/emmintrin.h: Guard POWER8 intrinsics.
9088 * config/rs6000/pmmintrin.h: Same.
9089 * config/rs6000/smmintrin.h: Same.
9090 * config/rs6000/tmmintrin.h: Same.
9092 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
9094 * config.gcc (extra_headers): Add nmmintrin.h.
9096 2021-10-19 Richard Biener <rguenther@suse.de>
9098 * tree-vectorizer.h (vect_supportable_dr_alignment): Add
9099 misalignment parameter.
9100 * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
9101 Do not change DR_MISALIGNMENT in place, instead pass the
9102 adjusted misalignment to vect_supportable_dr_alignment.
9103 (vect_peeling_supportable): Likewise.
9104 (vect_peeling_hash_get_lowest_cost): Adjust.
9105 (vect_enhance_data_refs_alignment): Likewise.
9106 (vect_vfa_access_size): Likewise.
9107 (vect_supportable_dr_alignment): Add misalignment
9108 parameter and simplify.
9109 * tree-vect-stmts.c (get_negative_load_store_type): Adjust.
9110 (get_group_load_store_type): Likewise.
9111 (get_load_store_type): Likewise.
9113 2021-10-19 Clément Chigot <clement.chigot@atos.net>
9115 * config/rs6000/rs6000.c (rs6000_xcoff_file_end): Move
9116 __tls_get_addr reference to .text csect.
9118 2021-10-19 Martin Liska <mliska@suse.cz>
9121 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
9124 2021-10-19 Richard Biener <rguenther@suse.de>
9126 * tree-vectorizer.h (vect_get_store_cost): Adjust signature.
9127 (vect_get_load_cost): Likewise.
9128 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
9129 alignment support scheme and misalignment as arguments
9131 (vect_get_peeling_costs_all_drs): Compute that info here
9132 and note that we shouldn't need to.
9133 * tree-vect-stmts.c (vect_model_store_cost): Get
9134 alignment support scheme and misalignment as arguments.
9135 (vect_get_store_cost): Likewise.
9136 (vect_model_load_cost): Likewise.
9137 (vect_get_load_cost): Likewise.
9138 (vectorizable_store): Pass down alignment support scheme
9139 and misalignment to costing.
9140 (vectorizable_load): Likewise.
9142 2021-10-19 Richard Biener <rguenther@suse.de>
9144 * tree-vect-stmts.c (get_negative_load_store_type): Add
9145 offset output parameter and initialize it.
9146 (get_group_load_store_type): Likewise.
9147 (get_load_store_type): Likewise.
9148 (vectorizable_store): Use offset as computed by
9149 get_load_store_type.
9150 (vectorizable_load): Likewise.
9152 2021-10-19 Richard Biener <rguenther@suse.de>
9154 PR tree-optimization/102827
9155 * tree-if-conv.c (predicate_statements): Add pe parameter
9156 and use that edge to insert invariant stmts on.
9157 (combine_blocks): Pass through pe.
9158 (tree_if_conversion): Compute the edge to insert invariant
9159 stmts on and pass it along.
9161 2021-10-19 Roger Sayle <roger@nextmovesoftware.com>
9164 * config/bfin/bfin.md (addsubv2hi3, subaddv2hi3, ssaddsubv2hi3,
9165 sssubaddv2hi3): Swap the order of operators in vec_concat.
9167 2021-10-19 Xionghu Luo <luoxhu@linux.ibm.com>
9169 * config/rs6000/altivec.md (*altivec_vmrghb_internal): Delete.
9170 (altivec_vmrghb_direct): New.
9171 (*altivec_vmrghh_internal): Delete.
9172 (altivec_vmrghh_direct): New.
9173 (*altivec_vmrghw_internal): Delete.
9174 (altivec_vmrghw_direct_<mode>): New.
9175 (altivec_vmrghw_direct): Delete.
9176 (*altivec_vmrglb_internal): Delete.
9177 (altivec_vmrglb_direct): New.
9178 (*altivec_vmrglh_internal): Delete.
9179 (altivec_vmrglh_direct): New.
9180 (*altivec_vmrglw_internal): Delete.
9181 (altivec_vmrglw_direct_<mode>): New.
9182 (altivec_vmrglw_direct): Delete.
9183 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
9184 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):
9186 * config/rs6000/vsx.md (vsx_xxmrghw_<mode>): Adjust.
9187 (vsx_xxmrglw_<mode>): Adjust.
9189 2021-10-19 Aldy Hernandez <aldyh@redhat.com>
9191 * passes.def: Change threading comment before pass_ccp pass.
9193 2021-10-19 Haochen Gui <guihaoc@gcc.gnu.org>
9195 * config/rs6000/rs6000-call.c (altivec_expand_lxvr_builtin):
9196 Modify the expansion for sign extension. All extensions are done
9197 within VSX registers.
9199 2021-10-19 Richard Biener <rguenther@suse.de>
9201 * tree-vect-stmts.c (get_group_load_store_type): Add
9202 misalignment output parameter and initialize it.
9203 (get_group_load_store_type): Likewise.
9204 (vectorizable_store): Remove now redundant queries.
9205 (vectorizable_load): Likewise.
9207 2021-10-19 Richard Biener <rguenther@suse.de>
9209 * tree-vectorizer.h (vect_supportable_dr_alignment): Remove
9210 check_aligned argument.
9211 * tree-vect-data-refs.c (vect_supportable_dr_alignment):
9213 (vect_peeling_hash_insert): Add supportable_if_not_aligned
9214 argument and do not call vect_supportable_dr_alignment here.
9215 (vect_peeling_supportable): Adjust.
9216 (vect_enhance_data_refs_alignment): Compute whether the
9217 access is supported with different alignment here and
9218 pass that down to vect_peeling_hash_insert.
9219 (vect_vfa_access_size): Adjust.
9220 * tree-vect-stmts.c (vect_get_store_cost): Likewise.
9221 (vect_get_load_cost): Likewise.
9222 (get_negative_load_store_type): Likewise.
9223 (get_group_load_store_type): Likewise.
9224 (get_load_store_type): Likewise.
9226 2021-10-19 Martin Liska <mliska@suse.cz>
9229 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
9230 * system.h (strip_whilespaces): New function.
9232 2021-10-19 dianhong xu <dianhong.xu@intel.com>
9234 * config/i386/avx512fp16intrin.h:
9235 (_mm512_set1_pch): New intrinsic.
9236 * config/i386/avx512fp16vlintrin.h:
9237 (_mm256_set1_pch): New intrinsic.
9238 (_mm_set1_pch): Ditto.
9240 2021-10-18 Andrew MacLeod <amacleod@redhat.com>
9242 PR tree-optimization/102796
9243 * gimple-range.cc (gimple_ranger::range_on_edge): Process EH edges
9244 normally. Return get_tree_range for non gimple_range_ssa_p names.
9245 (gimple_ranger::range_of_stmt): Use get_tree_range for non
9246 gimple_range_ssa_p names.
9248 2021-10-18 Uroš Bizjak <ubizjak@gmail.com>
9251 * config/i386/i386.c (ix86_print_operand_address):
9252 Error out for non-address_operand asm operands.
9254 2021-10-18 Richard Biener <rguenther@suse.de>
9256 * tree-vect-data-refs.c (vect_peeling_hash_insert): Do
9257 not auto-convert dr_alignment_support to bool.
9258 (vect_peeling_supportable): Likewise.
9259 (vect_enhance_data_refs_alignment): Likewise.
9260 (vect_supportable_dr_alignment): Commonize read/write case.
9261 * tree-vect-stmts.c (vect_get_store_cost): Use
9262 dr_alignment_support, not int, for the vect_supportable_dr_alignment
9264 (vect_get_load_cost): Likewise.
9266 2021-10-18 Siddhesh Poyarekar <siddhesh@gotplt.org>
9268 * tree-object-size.c (object_sizes_execute): Consolidate LHS
9269 null check and do it early.
9271 2021-10-18 Richard Biener <rguenther@suse.de>
9273 * tree-vect-stmts.c (vectorizable_store): Use the
9274 computed alignment scheme instead of querying
9277 2021-10-18 Richard Biener <rguenther@suse.de>
9279 * tree-vect-stmts.c (vectorizable_store): Do not recompute
9280 alignment scheme already determined by get_load_store_type.
9282 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
9284 * tree-ssa-threadbackward.c (class pass_thread_jumps_full):
9285 Clone corresponding pass.
9287 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
9289 * combine.c (recog_for_combine): For an unrecognized move/set of
9290 a constant, try force_const_mem to place it in the constant pool.
9291 * cse.c (constant_pool_entries_cost, constant_pool_entries_regcost):
9292 Delete global variables (that are no longer assigned a cost value).
9293 (cse_insn): Simplify logic for deciding whether to place a folded
9294 constant in the constant pool using force_const_mem.
9295 (cse_main): Remove zero initialization of constant_pool_entries_cost
9296 and constant_pool_entries_regcost.
9297 * config/i386/i386.c (ix86_rtx_costs): Make memory accesses
9298 fractionally more expensive, when optimizing for speed.
9300 2021-10-18 Martin Liska <mliska@suse.cz>
9302 PR gcov-profile/102746
9303 PR gcov-profile/102747
9304 * gcov.c (main): Return return_code.
9305 (output_gcov_file): Mark return_code when error happens.
9306 (generate_results): Likewise.
9307 (read_graph_file): Likewise.
9308 (read_count_file): Likewise.
9310 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
9312 * config/bfin/bfin.md (define_constants): Remove UNSPEC_ONES.
9313 (define_insn "ones"): Replace UNSPEC_ONES with a truncate of
9314 a popcount, allowing compile-time evaluation/simplification.
9315 (popcountsi2, popcounthi2): New expanders using a "ones" insn.
9317 2021-10-18 Richard Biener <rguenther@suse.de>
9319 PR tree-optimization/102788
9320 * tree-vect-patterns.c (vect_init_pattern_stmt): Allow
9322 (vect_pattern_recog_1): Likewise.
9323 (vect_recog_bool_pattern): Continue matching the pattern
9324 even if we do not have a vector type for a conversion
9327 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
9329 * simplify-rtx.c (simplify_const_unary_operation) [SS_NEG, SS_ABS]:
9330 Evalute SS_NEG and SS_ABS of a constant argument.
9332 2021-10-18 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
9335 * gimple-match-head.c (try_conditional_simplification): Add case for single operand.
9336 * internal-fn.def: Add entry for COND_NEG internal function.
9337 * internal-fn.c (FOR_EACH_CODE_MAPPING): Add entry for
9338 NEGATE_EXPR, COND_NEG mapping.
9339 * optabs.def: Add entry for cond_neg_optab.
9340 * match.pd (UNCOND_UNARY, COND_UNARY): New operator lists.
9341 (vec_cond COND (foo A) B) -> (IFN_COND_FOO COND A B): New pattern.
9342 (vec_cond COND B (foo A)) -> (IFN_COND_FOO ~COND A B): Likewise.
9344 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
9346 * tree-ssa-strlen.c (class strlen_pass): Rename from
9348 (handle_builtin_stxncpy_strncat): Move to strlen_pass.
9349 (handle_assign): Same.
9350 (adjust_last_stmt): Same.
9351 (maybe_warn_overflow): Same.
9352 (handle_builtin_strlen): Same.
9353 (handle_builtin_strchr): Same.
9354 (handle_builtin_strcpy): Same.
9355 (handle_builtin_strncat): Same.
9356 (handle_builtin_stxncpy_strncat): Same.
9357 (handle_builtin_memcpy): Same.
9358 (handle_builtin_strcat): Same.
9359 (handle_alloc_call): Same.
9360 (handle_builtin_memset): Same.
9361 (handle_builtin_memcmp): Same.
9362 (get_len_or_size): Same.
9363 (strxcmp_eqz_result): Same.
9364 (handle_builtin_string_cmp): Same.
9365 (handle_pointer_plus): Same.
9366 (count_nonzero_bytes_addr): Same.
9367 (count_nonzero_bytes): Same.
9368 (handle_store): Same.
9369 (strlen_check_and_optimize_call): Same.
9370 (handle_integral_assign): Same.
9371 (check_and_optimize_stmt): Same.
9372 (printf_strlen_execute): Rename strlen_dom_walker to strlen_pass.
9374 2021-10-18 Richard Biener <rguenther@suse.de>
9376 PR tree-optimization/102798
9377 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
9378 Only copy points-to info to newly generated SSA names.
9380 2021-10-18 Martin Liska <mliska@suse.cz>
9382 * dbgcnt.c (dbg_cnt_process_opt): Remove unused but set variable.
9383 * gcov.c (get_cycles_count): Likewise.
9384 * lto-compress.c (lto_compression_zlib): Likewise.
9385 (lto_uncompression_zlib): Likewise.
9386 * targhooks.c (default_pch_valid_p): Likewise.
9388 2021-10-17 Aldy Hernandez <aldyh@redhat.com>
9390 * tree-pass.h (make_pass_thread_jumps_full): New.
9391 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Inline.
9392 (try_thread_blocks): Add resolve and speed arguments.
9393 (pass_thread_jumps::execute): Inline.
9394 (do_early_thread_jumps): New.
9395 (do_thread_jumps): New.
9396 (make_pass_thread_jumps): Move.
9397 (pass_early_thread_jumps::gate): Inline.
9398 (pass_early_thread_jumps::execute): Inline.
9399 (class pass_thread_jumps_full): New.
9401 2021-10-16 Piotr Kubaj <pkubaj@FreeBSD.org>
9403 * configure.ac: Treat powerpc64*-*-freebsd* the same as
9404 powerpc64-*-freebsd*.
9405 * configure: Regenerate.
9407 2021-10-16 H.J. Lu <hjl.tools@gmail.com>
9409 * value-query.cc (get_ssa_name_ptr_info_nonnull): Change
9410 set_ptr_nonull to set_ptr_nonnull in comments.
9412 2021-10-16 Jan Hubicka <hubicka@ucw.cz>
9414 PR tree-optimization/102720
9415 * tree-ssa-structalias.c (compute_points_to_sets): Fix producing
9416 of call used and clobbered sets.
9418 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
9420 * config/darwin.h (LINK_COMMAND_SPEC_A): Update 'r' handling to
9421 skip gomp and itm when r or nodefaultlibs is given.
9422 (DSYMUTIL_SPEC): Do not call dsymutil for '-r' link lines.
9423 Update ordering of exclusions, remove duplicate 'v' addition
9424 (collect2 will add this from the main command line).
9426 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
9428 * config/darwin-driver.c (darwin_driver_init): Revise comments, handle
9429 filelist and framework options in specs instead of code.
9430 * config/darwin.h (SUBTARGET_DRIVER_SELF_SPECS): Update to handle link
9431 specs that are really driver ones.
9432 (DARWIN_CC1_SPEC): Likewise.
9433 (CPP_SPEC): Likewise.
9434 (SYSROOT_SPEC): Append space.
9435 (LINK_SYSROOT_SPEC): Remove most driver link specs.
9436 (STANDARD_STARTFILE_PREFIX_2): Update link-related specs.
9437 (STARTFILE_SPEC): Likewise.
9438 (ASM_MMACOSX_VERSION_MIN_SPEC): Fix line wrap.
9439 (ASM_SPEC): Update driver-related specs.
9440 (ASM_FINAL_SPEC): Likewise.
9441 * config/darwin.opt: Remove now unused option aliases.
9442 * config/i386/darwin.h (EXTRA_ASM_OPTS): Ensure space after opt.
9443 (ASM_SPEC): Update driver-related specs.
9445 2021-10-15 Roger Sayle <roger@nextmovesoftware.com>
9447 * config/i386/i386.c (ix86_hardreg_mov_ok): For vector modes,
9448 allow standard_sse_constant_p immediate constants.
9450 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
9452 * config.gcc: Add tm-dwarf2.h to tm_d-file.
9454 2021-10-15 Andrew MacLeod <amacleod@redhat.com>
9456 * gimple-range-fold.h (gimple_range_ssa_p): Don't process names
9457 that occur in abnormal phis.
9458 * gimple-range.cc (gimple_ranger::range_on_edge): Return false for
9459 abnormal and EH edges.
9460 * gimple-ssa-evrp.c (rvrp_folder::value_of_expr): Ditto.
9461 (rvrp_folder::value_on_edge): Ditto.
9462 (rvrp_folder::value_of_stmt): Ditto.
9463 (hybrid_folder::value_of_expr): Ditto for ranger queries.
9464 (hybrid_folder::value_on_edge): Ditto.
9465 (hybrid_folder::value_of_stmt): Ditto.
9466 * value-query.cc (gimple_range_global): Always return a range if
9467 the type is supported.
9469 2021-10-15 John David Anglin <danglin@gcc.gnu.org>
9471 * config/pa/pa.md: Consistently use "rG" constraint for copy
9472 instruction in move patterns.
9474 2021-10-15 Andrew Stubbs <ams@codesourcery.com>
9476 * config/gcn/gcn-hsa.h (S_FIJI): Set unconditionally.
9479 * config/gcn/gcn.c: Hard code SRAM ECC settings for old architectures.
9480 * config/gcn/mkoffload.c (ELFABIVERSION_AMDGPU_HSA): Rename to ...
9481 (ELFABIVERSION_AMDGPU_HSA_V3): ... this.
9482 (ELFABIVERSION_AMDGPU_HSA_V4): New.
9483 (SET_SRAM_ECC_UNSUPPORTED): New.
9484 (copy_early_debug_info): Create elf flags to match the other objects.
9485 (main): Just let the attribute flags pass through.
9487 2021-10-15 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
9489 * tree-loop-distribution.c (reduction_var_overflows_first):
9490 Pass the type of reduction_var as first argument as it is also
9491 done for the load type.
9492 (loop_distribution::transform_reduction_loop): Add missing
9493 TREE_TYPE while determining precission of reduction_var.
9495 2021-10-15 Richard Biener <rguenther@suse.de>
9497 * defaults.h (PREFERRED_DEBUGGING_TYPE): Choose DWARF2_DEBUG
9499 * toplev.c (process_options): Warn when STABS debugging is
9500 enabled but not the preferred format.
9501 * config/pa/som.h (PREFERRED_DEBUGGING_TYPE): Define to
9503 * config/pdp11/pdp11.h (PREFERRED_DEBUGGING_TYPE): Likewise.
9505 2021-10-15 Richard Biener <rguenther@suse.de>
9508 * tree-inline.c (copy_bb): Avoid underflowing nargs.
9510 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
9512 * config/i386/i386-expand.c (ix86_expand_vec_perm): Convert
9513 HFmode input operand to HImode.
9514 (ix86_vectorize_vec_perm_const): Likewise.
9515 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1_hf):
9517 (*avx512f_permvar_truncv8siv8hi_1_hf):
9520 2021-10-15 Richard Biener <rguenther@suse.de>
9522 PR middle-end/102682
9523 * expmed.c (store_bit_field_1): Ensure a LHS subreg would
9524 not create a paradoxical subreg.
9526 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
9528 * config/i386/i386-expand.c (ix86_expand_vector_init):
9529 For half_vector concat for HFmode, handle them like HImode.
9531 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
9533 * tree-ssa-threadbackward.c (class back_threader): Add m_resolve.
9534 (back_threader::back_threader): Same.
9535 (back_threader::resolve_phi): Try to solve without looking back if
9537 (back_threader::find_paths_to_names): Same.
9538 (try_thread_blocks): Pass resolve argument to back threader.
9539 (pass_early_thread_jumps::execute): Same.
9541 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
9543 * doc/invoke.texi: Remove max-fsm-thread-length,
9544 max-fsm-thread-paths, and fsm-maximum-phi-arguments.
9546 * tree-ssa-threadbackward.c (back_threader::back_threader): Remove
9548 (back_threader_registry::back_threader_registry): Same.
9549 (back_threader_profitability::profitable_path_p): Remove
9550 param_max_fsm_thread-length.
9551 (back_threader_registry::register_path): Remove
9552 m_max_allowable_paths.
9554 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
9556 * tree-ssa-threadbackward.c (class back_threader): Make m_imports
9558 (back_threader::~back_threader): Do not release m_path.
9560 2021-10-14 Andrew MacLeod <amacleod@redhat.com>
9562 PR tree-optimization/102738
9563 * vr-values.c (simplify_using_ranges::simplify): Handle RSHIFT_EXPR.
9565 2021-10-14 Kwok Cheung Yeung <kcy@codesourcery.com>
9567 * omp-general.c (omp_check_context_selector): Move from c-omp.c.
9568 (omp_mark_declare_variant): Move from c-omp.c.
9569 (omp_context_name_list_prop): Update for Fortran strings.
9570 * omp-general.h (omp_check_context_selector): New prototype.
9571 (omp_mark_declare_variant): New prototype.
9573 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
9575 * config/rs6000/rs6000.c (rs6000_density_test): Move early
9576 exit test further up the function.
9578 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
9580 * config/arm/arm.c (arm_add_stmt_cost): Delete.
9581 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
9583 2021-10-14 Martin Jambor <mjambor@suse.cz>
9585 * doc/invoke.texi (Optimize Options): Add entry for
9586 ipa-cp-recursive-freq-factor.
9588 2021-10-14 Tamar Christina <tamar.christina@arm.com>
9590 * match.pd: New rule.
9592 2021-10-14 Jan Hubicka <hubicka@ucw.cz>
9595 * ipa-modref-tree.h (modref_access_node::update2):
9596 Also check that parm_offset is unchanged.
9597 (modref_ref_node::insert_access): Fix updating of
9600 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
9602 * tree-ssa-threadbackward.c (back_threader::resolve_phi): Add
9605 2021-10-14 Richard Biener <rguenther@suse.de>
9607 PR tree-optimization/102659
9608 * tree-if-conv.c (if_convertible_gimple_assign_stmt_p): Also
9609 rewrite pointer typed undefined overflow operations.
9610 (predicate_statements): Likewise. Make sure to emit invariant
9611 conversions in the preheader.
9612 * tree-vectorizer.c (vect_loop_vectorized_call): Look through
9613 non-empty preheaders.
9614 * tree-data-ref.c (dr_analyze_indices): Strip useless
9615 conversions to the MEM_REF base type.
9617 2021-10-14 Martin Liska <mliska@suse.cz>
9619 * common.opt: Stop using AUTODETECT_VALUE
9620 and use EnabledBy where possible.
9621 * opts.c: Enable OPT_fvar_tracking with optimize >= 1.
9622 * toplev.c (AUTODETECT_VALUE): Remove macro.
9623 (process_options): Simplify by using EnabledBy and
9624 OPT_fvar_tracking. Use OPTION_SET_P macro instead of
9627 2021-10-14 Jonathan Wright <jonathan.wright@arm.com>
9629 * config/aarch64/arm_neon.h (vld1_s8_x3): Use signed type for
9631 (vld1_s32_x3): Likewise.
9633 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
9635 PR tree-optimization/102736
9636 PR tree-optimization/102736
9637 * gimple-range-path.cc (path_range_query::range_on_path_entry):
9638 Assert that the requested range is defined outside the path.
9639 (path_range_query::ssa_range_in_phi): Do not call
9640 range_on_path_entry for SSA names that are defined within the
9643 2021-10-14 Iain Sandoe <iain@sandoe.co.uk>
9645 * config/darwin-driver.c (darwin_find_version_from_kernel):
9646 Quote internal identifiers and avoid contractions in
9648 (darwin_default_min_version): Likewise.
9649 (darwin_driver_init): Likewise.
9651 2021-10-14 Martin Jambor <mjambor@suse.cz>
9653 * params.opt (ipa-cp-recursive-freq-factor): New.
9654 * ipa-cp.c (ipcp_value): Switch to inline initialization. New members
9655 scc_no, self_recursion_generated_level, same_scc and
9656 self_recursion_generated_p.
9657 (ipcp_lattice::add_value): Replaced parameter unlimited with
9658 same_lat_gen_level, usit it determine limit of values and store it to
9660 (ipcp_lattice<valtype>::print): Dump the new fileds.
9661 (allocate_and_init_ipcp_value): Take same_lat_gen_level as a new
9662 parameter and store it to the new value.
9663 (self_recursively_generated_p): Removed.
9664 (propagate_vals_across_arith_jfunc): Use self_recursion_generated_p
9665 instead of self_recursively_generated_p, store self generation level
9667 (value_topo_info<valtype>::add_val): Set scc_no.
9668 (value_topo_info<valtype>::propagate_effects): Multiply frequencies of
9669 recursively feeding values and self generated values by appropriate
9672 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
9674 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Remove
9675 redundant test for flag_vect_cost_model.
9677 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
9679 * bitmap.c (debug): New overloaded function for auto_bitmaps.
9680 * bitmap.h (debug): Same.
9682 2021-10-14 Hongyu Wang <hongyu.wang@intel.com>
9684 * config/i386/avx512fp16intrin.h (_mm512_mask_fcmadd_pch):
9685 Adjust builtin call.
9686 (_mm512_mask3_fcmadd_pch): Likewise.
9687 (_mm512_mask_fmadd_pch): Likewise
9688 (_mm512_mask3_fmadd_pch): Likewise
9689 (_mm512_mask_fcmadd_round_pch): Likewise
9690 (_mm512_mask3_fcmadd_round_pch): Likewise
9691 (_mm512_mask_fmadd_round_pch): Likewise
9692 (_mm512_mask3_fmadd_round_pch): Likewise
9693 (_mm_mask_fcmadd_sch): Likewise
9694 (_mm_mask3_fcmadd_sch): Likewise
9695 (_mm_mask_fmadd_sch): Likewise
9696 (_mm_mask3_fmadd_sch): Likewise
9697 (_mm_mask_fcmadd_round_sch): Likewise
9698 (_mm_mask3_fcmadd_round_sch): Likewise
9699 (_mm_mask_fmadd_round_sch): Likewise
9700 (_mm_mask3_fmadd_round_sch): Likewise
9701 (_mm_fcmadd_round_sch): Likewise
9702 * config/i386/avx512fp16vlintrin.h (_mm_mask_fmadd_pch):
9703 Adjust builtin call.
9704 (_mm_mask3_fmadd_pch): Likewise
9705 (_mm256_mask_fmadd_pch): Likewise
9706 (_mm256_mask3_fmadd_pch): Likewise
9707 (_mm_mask_fcmadd_pch): Likewise
9708 (_mm_mask3_fcmadd_pch): Likewise
9709 (_mm256_mask_fcmadd_pch): Likewise
9710 (_mm256_mask3_fcmadd_pch): Likewise
9711 * config/i386/i386-builtin.def: Add mask3 builtin for complex
9712 fma, and adjust mask_builtin to corresponding expander.
9713 * config/i386/i386-expand.c (ix86_expand_round_builtin):
9714 Skip eraseing embedded rounding for expanders that emits
9716 * config/i386/sse.md (complexmove): New mode_attr.
9717 (<avx512>_fmaddc_<mode>_mask1<round_expand_name>): New expander.
9718 (<avx512>_fcmaddc_<mode>_mask1<round_expand_name>): Likewise.
9719 (avx512fp16_fmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
9720 (avx512fp16_fcmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
9721 (avx512fp16_fcmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
9722 (avx512fp16_fmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
9723 * config/i386/subst.md (round_embedded_complex): New subst.
9725 2021-10-14 John David Anglin <danglin@gcc.gnu.org>
9727 * config/pa/pa.md (cbranchsf4): Disable if TARGET_SOFT_FLOAT.
9728 (cbranchdf4): Likewise.
9729 Add missing move patterns for TARGET_SOFT_FLOAT.
9731 2021-10-13 Roger Sayle <roger@nextmovesoftware.com>
9733 * config/i386/i386-expand.c (ix86_expand_vector_move): Use a
9734 pseudo intermediate when moving a SUBREG into a hard register,
9735 by checking ix86_hardreg_mov_ok.
9736 (ix86_expand_vector_extract): Store zero-extended SImode
9737 intermediate in a pseudo, then set target using a SUBREG_PROMOTED
9739 * config/i386/sse.md (mov<VMOVE>_internal): Prevent CSE creating
9740 complex (SUBREG) sets of (vector) hard registers before reload, by
9741 checking ix86_hardreg_mov_ok.
9743 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
9745 * ctfc.h (enum ctf_dtu_d_union_enum): Remove redundant comma.
9747 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
9749 * dwarf2ctf.c (gen_ctf_array_type): Fix typo in comment.
9751 2021-10-13 Martin Sebor <msebor@redhat.com>
9753 PR middle-end/102630
9754 * pointer-query.cc (compute_objsize_r): Handle named address spaces.
9756 2021-10-13 Iain Sandoe <iain@sandoe.co.uk>
9758 * collect2.c (is_lto_object_file): Release simple-object
9759 resources, close files.
9761 2021-10-13 Andre Vieira <andre.simoesdiasvieira@arm.com>
9763 * config/arm/arm.c (thumb2_legitimate_address_p): Use VALID_MVE_MODE
9764 when checking mve addressing modes.
9765 (mve_vector_mem_operand): Fix the way we handle pre, post and offset
9767 (arm_print_operand): Fix printing of POST_ and PRE_MODIFY.
9768 * config/arm/mve.md: Use mve_memory_operand predicate everywhere where
9769 there is a single Ux constraint.
9771 2021-10-13 John David Anglin <danglin@gcc.gnu.org>
9773 * config/pa/pa.md (muldi3): Add support for inlining 64-bit
9774 multiplication on 32-bit PA 1.1 and 2.0 targets.
9776 2021-10-13 Martin Liska <mliska@suse.cz>
9778 PR gcov-profile/90364
9779 * coverage.c (build_info): Emit checksum to the global variable.
9780 (build_info_type): Add new field for checksum.
9781 (coverage_obj_finish): Pass object_checksum.
9782 (coverage_init): Use 0 as checksum for .gcno files.
9783 * gcov-dump.c (dump_gcov_file): Dump also new checksum field.
9784 * gcov.c (read_graph_file): Read also checksum.
9785 * doc/invoke.texi: Document the behaviour change.
9787 2021-10-13 Richard Biener <rguenther@suse.de>
9789 * gimple-iterator.h (gsi_iterator_update): Add GSI_LAST_NEW_STMT,
9790 start at integer value 2.
9791 * gimple-iterator.c (gsi_insert_seq_nodes_before): Update
9792 the iterator for GSI_LAST_NEW_STMT.
9793 (gsi_insert_seq_nodes_after): Likewise.
9794 * tree-if-conv.c (predicate_statements): Use GSI_LAST_NEW_STMT.
9795 * tree-ssa.c (execute_update_addresses_taken): Correct bogus
9796 arguments to gsi_replace.
9798 2021-10-13 Martin Liska <mliska@suse.cz>
9801 * common.opt: Use EnabledBy instead of detection in
9802 finish_options and process_options.
9803 * opts.c (finish_options): Remove handling of
9804 x_flag_unroll_all_loops.
9805 * toplev.c (process_options): Likewise for flag_web and
9806 flag_rename_registers.
9808 2021-10-13 Richard Biener <rguenther@suse.de>
9810 PR tree-optimization/102659
9811 * tree-if-conv.c (need_to_rewrite_undefined): New flag.
9812 (if_convertible_gimple_assign_stmt_p): Mark the loop for
9813 rewrite when stmts with undefined behavior on integer
9815 (combine_blocks): Predicate also when we need to rewrite stmts.
9816 (predicate_statements): Rewrite affected stmts to something
9817 with well-defined behavior on overflow.
9818 (tree_if_conversion): Initialize need_to_rewrite_undefined.
9820 2021-10-13 Richard Biener <rguenther@suse.de>
9823 * ipa-sra.c (ptr_parm_has_nonarg_uses): Fix volatileness
9826 2021-10-13 Jose E. Marchesi <jose.marchesi@oracle.com>
9828 * dwarf2ctf.c: Fix typo in comment.
9830 2021-10-12 Bill Schmidt <wschmidt@linux.ibm.com>
9833 * config/rs6000/altivec.h (vec_cpsgn): Swap operand order.
9834 * config/rs6000/rs6000-overload.def (VEC_COPYSIGN): Use SKIP to
9835 avoid generating an automatic #define of vec_cpsgn. Use the
9836 correct built-in for V4SFmode that doesn't depend on VSX.
9838 2021-10-12 Uroš Bizjak <ubizjak@gmail.com>
9842 * config/i386/i386.md (*add<mode>_1_slp): Rewrite as
9843 define_insn_and_split pattern. Add alternative 1 and split it
9844 post reload to insert operand 1 into the low part of operand 0.
9845 (*sub<mode>_1_slp): Ditto.
9846 (*and<mode>_1_slp): Ditto.
9847 (*<any_or:code><mode>_1_slp): Ditto.
9848 (*ashl<mode>3_1_slp): Ditto.
9849 (*<any_shiftrt:insn><mode>3_1_slp): Ditto.
9850 (*<any_rotate:insn><mode>3_1_slp): Ditto.
9851 (*neg<mode>_1_slp): New insn_and_split pattern.
9852 (*one_cmpl<mode>_1_slp): Ditto.
9854 2021-10-12 David Edelsohn <dje.gcc@gmail.com>
9856 * doc/install.texi: Update MinGW and mingw-64 Binaries
9859 2021-10-12 Daniel Le Duc Khoi Nguyen <greenrecyclebin@gmail.com>
9861 * doc/extend.texi (Common Variable Attributes): Fix typos in
9862 alloc_size documentation.
9864 2021-10-12 Richard Biener <rguenther@suse.de>
9866 PR tree-optimization/102696
9867 * tree-vect-slp.c (vect_build_slp_tree_2): Properly mark
9868 the tree fatally failed when we reject a BIT_FIELD_REF.
9870 2021-10-12 Richard Biener <rguenther@suse.de>
9872 PR tree-optimization/102572
9873 * tree-vect-stmts.c (vect_build_gather_load_calls): When
9874 gathering the vectorized defs for the mask pass in the
9875 desired mask vector type so invariants will be handled
9878 2021-10-12 Tamar Christina <tamar.christina@arm.com>
9880 * config/aarch64/aarch64-sve.md (*fcm<cmp_op><mode>_bic_combine,
9881 *fcm<cmp_op><mode>_nor_combine, *fcmuo<mode>_bic_combine,
9882 *fcmuo<mode>_nor_combine): New.
9884 2021-10-12 Eric Botcazou <ebotcazou@adacore.com>
9887 * config/sparc/sparc-modes.def (OI): New integer mode.
9889 2021-10-12 Jakub Jelinek <jakub@redhat.com>
9891 * gimple-fold.h (clear_padding_type_may_have_padding_p): Declare.
9892 * gimple-fold.c (clear_padding_type_may_have_padding_p): No longer
9895 2021-10-12 Jakub Jelinek <jakub@redhat.com>
9897 * tree-vectorizer.h (loop_cost_model): New function.
9898 (unlimited_cost_model): Use it.
9899 * tree-vect-loop.c (vect_analyze_loop_costing): Use loop_cost_model
9900 call instead of flag_vect_cost_model.
9901 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9902 (vect_prune_runtime_alias_test_list): Likewise. Also use it instead
9903 of flag_simd_cost_model.
9905 2021-10-12 liuhongt <hongtao.liu@intel.com>
9908 * config/i386/i386-expand.c (emit_reduc_half): Handle
9910 * config/i386/mmx.md (reduc_<code>_scal_v4qi): New expander.
9911 (reduc_plus_scal_v4qi): Ditto.
9913 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
9915 * config/rs6000/smmintrin.h (_mm_cmpeq_epi64, _mm_cmpgt_epi64,
9916 _mm_mullo_epi32, _mm_mul_epi32, _mm_packus_epi32): New.
9917 * config/rs6000/nmmintrin.h: Copy from i386, tweak to suit.
9919 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
9921 * config/rs6000/smmintrin.h (_mm_cvtepi8_epi16, _mm_cvtepi8_epi32,
9922 _mm_cvtepi8_epi64, _mm_cvtepi16_epi32, _mm_cvtepi16_epi64,
9923 _mm_cvtepi32_epi64, _mm_cvtepu8_epi16, _mm_cvtepu8_epi32,
9924 _mm_cvtepu8_epi64, _mm_cvtepu16_epi32, _mm_cvtepu16_epi64,
9925 _mm_cvtepu32_epi64): New.
9927 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
9929 * config/rs6000/smmintrin.h (_mm_test_all_zeros,
9930 _mm_test_all_ones, _mm_test_mix_ones_zeros): Rewrite as macro.
9932 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
9934 * config/rs6000/smmintrin.h (_mm_min_epi8, _mm_min_epu16,
9935 _mm_min_epi32, _mm_min_epu32, _mm_max_epi8, _mm_max_epu16,
9936 _mm_max_epi32, _mm_max_epu32): New.
9938 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
9940 * ipa-modref-tree.h (struct modref_access_node): Revert
9942 (struct modref_ref_node): Likewise.
9944 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
9946 * ipa-modref-tree.h (modref_tree::global_access_p): New member
9949 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
9950 ignore_stores_eaf_flags): Move to ipa-modref.h
9951 (remove_useless_eaf_flags): Remove early exit on NOCLOBBER.
9952 (modref_summary::global_memory_read_p): New member function.
9953 (modref_summary::global_memory_written_p): New member function.
9954 * ipa-modref.h (modref_summary::global_memory_read_p,
9955 modref_summary::global_memory_written_p): Declare.
9956 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
9957 ignore_stores_eaf_flags): move here.
9958 * tree-ssa-structalias.c: Include ipa-modref-tree.h, ipa-modref.h
9960 (handle_rhs_call): Rewrite.
9961 (handle_call_arg): New function.
9962 (determine_global_memory_access): New function.
9963 (handle_const_call): Remove
9964 (handle_pure_call): Remove
9965 (find_func_aliases_for_call): Update use of handle_rhs_call.
9966 (compute_points_to_sets): Handle global memory acccesses
9969 2021-10-11 Diane Meirowitz <diane.meirowitz@oracle.com>
9971 * doc/invoke.texi: Add link to UndefinedBehaviorSanitizer
9972 documentation, mention UBSAN_OPTIONS, similar to what is done
9973 for AddressSanitizer.
9975 2021-10-11 Richard Biener <rguenther@suse.de>
9977 PR middle-end/102683
9978 * internal-fn.c (expand_DEFERRED_INIT): Check for mode
9979 availability before building an integer type for storage
9982 2021-10-11 Richard Biener <rguenther@suse.de>
9984 PR middle-end/101480
9985 * gimple.c (gimple_call_fnspec): Do not mark operator new/delete
9988 2021-10-11 Martin Liska <mliska@suse.cz>
9990 * common.opt: Remove Init(2) for some options.
9991 * toplev.c (process_options): Do not use AUTODETECT_VALUE, but
9992 use rather OPTION_SET_P.
9994 2021-10-11 Martin Liska <mliska@suse.cz>
9996 * common.opt: Remove usage of IRA_REGION_AUTODETECT.
9997 * flag-types.h (enum ira_region): Likewise.
9998 * toplev.c (process_options): Use OPTION_SET_P instead of
9999 IRA_REGION_AUTODETECT.
10001 2021-10-11 Jakub Jelinek <jakub@redhat.com>
10003 * omp-low.c (omp_runtime_api_call): Handle omp_get_max_teams,
10004 omp_[sg]et_teams_thread_limit and omp_set_num_teams.
10006 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
10008 * config/s390/s390-protos.h (s390_rawmemchr): Add prototype.
10009 * config/s390/s390.c (s390_rawmemchr): New function.
10010 * config/s390/s390.md (rawmemchr<SINT:mode>): New expander.
10011 * config/s390/vector.md (@vec_vfees<mode>): Basically a copy of
10012 the pattern vfees<mode> from vx-builtins.md.
10013 * config/s390/vx-builtins.md (*vfees<mode>): Remove.
10015 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
10017 * builtins.c (get_memory_rtx): Change to external linkage.
10018 * builtins.h (get_memory_rtx): Add function prototype.
10019 * doc/md.texi (rawmemchr<mode>): Document.
10020 * internal-fn.c (expand_RAWMEMCHR): Define.
10021 * internal-fn.def (RAWMEMCHR): Add.
10022 * optabs.def (rawmemchr_optab): Add.
10023 * tree-loop-distribution.c (find_single_drs): Change return code
10024 behaviour by also returning true if no single store was found
10026 (loop_distribution::classify_partition): Respect the new return
10027 code behaviour of function find_single_drs.
10028 (loop_distribution::execute): Call new function
10029 transform_reduction_loop in order to replace rawmemchr or strlen
10030 like loops by calls into builtins.
10031 (generate_reduction_builtin_1): New function.
10032 (generate_rawmemchr_builtin): New function.
10033 (generate_strlen_builtin_1): New function.
10034 (generate_strlen_builtin): New function.
10035 (generate_strlen_builtin_using_rawmemchr): New function.
10036 (reduction_var_overflows_first): New function.
10037 (determine_reduction_stmt_1): New function.
10038 (determine_reduction_stmt): New function.
10039 (loop_distribution::transform_reduction_loop): New function.
10041 2021-10-11 Martin Liska <mliska@suse.cz>
10043 * tree.c (cl_option_hasher::hash): Use cl_optimization_hash
10044 and remove legacy hashing code.
10046 2021-10-11 Kito Cheng <kito.cheng@sifive.com>
10049 * builtins.c (maybe_emit_call_builtin___clear_cache): Allow
10050 CONST_INT for BEGIN and END, and use gcc_assert rather than
10053 2021-10-10 Jakub Jelinek <jakub@redhat.com>
10056 * var-tracking.c (add_stores): For cselib_sp_derived_value_p values
10057 use MO_VAL_SET if loc is not sp.
10059 2021-10-10 Andrew Pinski <apinski@marvell.com>
10061 PR tree-optimization/102622
10062 * match.pd: Swap the order of a?pow2cst:0 and a?-1:0 transformations.
10063 Swap the order of a?0:pow2cst and a?0:-1 transformations.
10065 2021-10-09 liuhongt <hongtao.liu@intel.com>
10068 * config/i386/i386-expand.c (ix86_valid_mask_cmp_mode): Handle
10070 (ix86_use_mask_cmp_p): Ditto.
10071 (ix86_expand_sse_movcc): Ditto.
10072 * config/i386/i386.md (setcc_hf_mask): New define_insn.
10073 (movhf_mask): Ditto.
10074 (UNSPEC_MOVCC_MASK): New unspec.
10075 * config/i386/sse.md (UNSPEC_PCMP): Move to i386.md.
10077 2021-10-08 Vladimir N. Makarov <vmakarov@redhat.com>
10079 PR rtl-optimization/102627
10080 * lra-constraints.c (split_reg): Use at least natural mode of hard reg.
10082 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
10084 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Grow
10087 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
10089 * value-range.cc (irange::debug): New.
10090 * value-range.h (irange::debug): New.
10092 2021-10-08 Richard Sandiford <richard.sandiford@arm.com>
10094 PR tree-optimization/102385
10095 * predict.h (change_edge_frequency): Declare.
10096 * predict.c (change_edge_frequency): New function.
10097 * tree-ssa-loop-manip.h (tree_transform_and_unroll_loop): Remove
10099 (tree_unroll_loop): Likewise.
10100 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Update accordingly.
10101 * tree-predcom.c (pcom_worker::tree_predictive_commoning_loop):
10103 * tree-ssa-loop-prefetch.c (loop_prefetch_arrays): Likewise.
10104 * tree-ssa-loop-manip.c (tree_unroll_loop): Likewise.
10105 (tree_transform_and_unroll_loop): Likewise. Use single_dom_exit
10106 to retrieve the exit edges. Make all the old profile update code
10107 conditional on !single_loop_p -- the case it was written for --
10108 and use a different approach for the single-loop case.
10110 2021-10-08 Martin Liska <mliska@suse.cz>
10112 * config/alpha/alpha.c (alpha_option_override): Use new macro
10114 * config/arc/arc.c (arc_override_options): Likewise.
10115 * config/arm/arm.c (arm_option_override): Likewise.
10116 * config/bfin/bfin.c (bfin_load_pic_reg): Likewise.
10117 * config/c6x/c6x.c (c6x_option_override): Likewise.
10118 * config/csky/csky.c: Likewise.
10119 * config/darwin.c (darwin_override_options): Likewise.
10120 * config/frv/frv.c (frv_option_override): Likewise.
10121 * config/i386/djgpp.h: Likewise.
10122 * config/i386/i386.c (ix86_stack_protect_guard): Likewise.
10123 (ix86_max_noce_ifcvt_seq_cost): Likewise.
10124 * config/ia64/ia64.c (ia64_option_override): Likewise.
10125 (ia64_override_options_after_change): Likewise.
10126 * config/m32c/m32c.c (m32c_option_override): Likewise.
10127 * config/m32r/m32r.c (m32r_init): Likewise.
10128 * config/m68k/m68k.c (m68k_option_override): Likewise.
10129 * config/microblaze/microblaze.c (microblaze_option_override): Likewise.
10130 * config/mips/mips.c (mips_option_override): Likewise.
10131 * config/nios2/nios2.c (nios2_option_override): Likewise.
10132 * config/nvptx/nvptx.c (nvptx_option_override): Likewise.
10133 * config/pa/pa.c (pa_option_override): Likewise.
10134 * config/riscv/riscv.c (riscv_option_override): Likewise.
10135 * config/rs6000/aix71.h: Likewise.
10136 * config/rs6000/aix72.h: Likewise.
10137 * config/rs6000/aix73.h: Likewise.
10138 * config/rs6000/rs6000.c (darwin_rs6000_override_options): Likewise.
10139 (rs6000_override_options_after_change): Likewise.
10140 (rs6000_linux64_override_options): Likewise.
10141 (glibc_supports_ieee_128bit): Likewise.
10142 (rs6000_option_override_internal): Likewise.
10143 (rs6000_file_start): Likewise.
10144 (rs6000_darwin_file_start): Likewise.
10145 * config/rs6000/rtems.h: Likewise.
10146 * config/rs6000/sysv4.h: Likewise.
10147 * config/rs6000/vxworks.h (SUB3TARGET_OVERRIDE_OPTIONS): Likewise.
10148 * config/s390/s390.c (s390_option_override): Likewise.
10149 * config/sh/linux.h: Likewise.
10150 * config/sh/netbsd-elf.h (while): Likewise.
10151 * config/sh/sh.c (sh_option_override): Likewise.
10152 * config/sol2.c (solaris_override_options): Likewise.
10153 * config/sparc/sparc.c (sparc_option_override): Likewise.
10154 * config/tilegx/tilegx.c (tilegx_option_override): Likewise.
10155 * config/visium/visium.c (visium_option_override): Likewise.
10156 * config/vxworks.c (vxworks_override_options): Likewise.
10157 * lto-opts.c (lto_write_options): Likewise.
10158 * omp-expand.c (expand_omp_simd): Likewise.
10159 * omp-general.c (omp_max_vf): Likewise.
10160 * omp-offload.c (oacc_xform_loop): Likewise.
10161 * opts.h (OPTION_SET_P): Likewise.
10162 * targhooks.c (default_max_noce_ifcvt_seq_cost): Likewise.
10163 * toplev.c (process_options): Likewise.
10164 * tree-predcom.c: Likewise.
10165 * tree-sra.c (analyze_all_variable_accesses): Likewise.
10167 2021-10-08 liuhongt <hongtao.liu@intel.com>
10170 * config/i386/i386.c (ix86_optab_supported_p):
10171 Return true for HFmode.
10172 * match.pd: Simplify (_Float16) ceil ((double) x) to
10173 __builtin_ceilf16 (a) when a is _Float16 type and
10174 direct_internal_fn_supported_p.
10176 2021-10-08 liuhongt <hongtao.liu@intel.com>
10179 * config/i386/i386-expand.c (emit_reduc_half): Hanlde V4HImode.
10180 * config/i386/mmx.md (reduc_plus_scal_v4hi): New.
10181 (reduc_<code>_scal_v4hi): New.
10183 2021-10-08 liuhongt <hongtao.liu@intel.com>
10185 * common.opt (ftree-vectorize): Add Var(flag_tree_vectorize).
10186 * doc/invoke.texi (Options That Control Optimization): Update
10188 * opts.c (default_options_table): Enable auto-vectorization at
10189 O2 with very-cheap cost model.
10190 (finish_options): Use cheap cost model for
10191 explicit -ftree{,-loop}-vectorize.
10193 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
10195 * ctfc.c (ctfc_delete_container): Free hash table contents.
10197 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
10199 * toplev.c (process_options): Do not warn for GNU GIMPLE.
10201 2021-10-07 Siddhesh Poyarekar <siddhesh@gotplt.org>
10203 * tree-object-size.c (addr_object_size,
10204 compute_builtin_object_size): Drop PDECL and POFF arguments.
10205 (addr_object_size): Adjust calls.
10206 * tree-object-size.h (compute_builtin_object_size): Drop PDECL
10207 and POFF arguments.
10209 2021-10-07 Roger Sayle <roger@nextmovesoftware.com>
10211 * rtl.def (SMUL_HIGHPART, UMUL_HIGHPART): New RTX codes for
10212 representing signed and unsigned high-part multiplication resp.
10213 * simplify-rtx.c (simplify_binary_operation_1) [SMUL_HIGHPART,
10214 UMUL_HIGHPART]: Simplify high-part multiplications by zero.
10215 [SS_PLUS, US_PLUS, SS_MINUS, US_MINUS, SS_MULT, US_MULT,
10216 SS_DIV, US_DIV]: Similar simplifications for saturating
10218 (simplify_const_binary_operation) [SS_PLUS, US_PLUS, SS_MINUS,
10219 US_MINUS, SS_MULT, US_MULT, SMUL_HIGHPART, UMUL_HIGHPART]:
10220 Implement compile-time evaluation for constant operands.
10221 * dwarf2out.c (mem_loc_descriptor): Skip SMUL_HIGHPART and
10223 * doc/rtl.texi (smul_highpart, umul_highpart): Document RTX codes.
10224 * doc/md.texi (smul@var{m}3_highpart, umul@var{m3}_highpart):
10225 Mention the new smul_highpart and umul_highpart RTX codes.
10226 * doc/invoke.texi: Silence @xref "compilation" warnings.
10228 2021-10-07 Martin Jambor <mjambor@suse.cz>
10231 * ipa-prop.c (ipa_edge_args_sum_t::duplicate): Also handle the
10232 case when the source reference description corresponds to a
10233 referance taken in a function src->caller is inlined to.
10235 2021-10-07 Jan Hubicka <hubicka@ucw.cz>
10238 * ipa-modref-tree.h (modref_access_node::contains_p): Handle offsets
10240 (modref_access_node::try_merge_with): Add sanity check that there
10241 are no redundant entries in the list.
10243 2021-10-07 Richard Biener <rguenther@suse.de>
10245 PR tree-optimization/102608
10246 * tree-ssa-sccvn.c (visit_stmt): Drop .DEFERRED_INIT to
10249 2021-10-07 Martin Liska <mliska@suse.cz>
10251 * toplev.c (toplev::main): Make
10252 save_opt_decoded_options a pointer type
10253 * toplev.h: Likewise.
10255 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
10257 * config/gcn/gcn-valu.md (gather<mode>_insn_2offsets<exec>): Apply
10258 HAVE_GCN_ASM_GLOBAL_LOAD_FIXED.
10259 (scatter<mode>_insn_2offsets<exec_scatter>): Likewise.
10261 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
10263 * config/gcn/gcn-hsa.h (SRAMOPT): Include the whole option string.
10264 Adjust for new -msram-ecc=any behaviour.
10265 (ASM_SPEC): Adjust -mxnack and -msram-ecc usage.
10266 * config/gcn/gcn.c (output_file_start): Implement -msram-ecc=any.
10267 * config/gcn/mkoffload.c (EF_AMDGPU_XNACK): Rename to ...
10268 (EF_AMDGPU_XNACK_V3): ... this.
10269 (EF_AMDGPU_SRAM_ECC): Rename to ...
10270 (EF_AMDGPU_SRAM_ECC_V3): ... this.
10271 (EF_AMDGPU_FEATURE_XNACK_V4): New.
10272 (EF_AMDGPU_FEATURE_XNACK_UNSUPPORTED_V4): New.
10273 (EF_AMDGPU_FEATURE_XNACK_ANY_V4): New.
10274 (EF_AMDGPU_FEATURE_XNACK_OFF_V4): New.
10275 (EF_AMDGPU_FEATURE_XNACK_ON_V4): New.
10276 (EF_AMDGPU_FEATURE_SRAMECC_V4): New.
10277 (EF_AMDGPU_FEATURE_SRAMECC_UNSUPPORTED_V4): New.
10278 (EF_AMDGPU_FEATURE_SRAMECC_ANY_V4): New.
10279 (EF_AMDGPU_FEATURE_SRAMECC_OFF_V4): New.
10280 (EF_AMDGPU_FEATURE_SRAMECC_ON_V4): New.
10281 (SET_XNACK_ON): New.
10282 (SET_XNACK_OFF): New.
10284 (SET_SRAM_ECC_ON): New.
10285 (SET_SRAM_ECC_ANY): New.
10286 (SET_SRAM_ECC_OFF): New.
10287 (TEST_SRAM_ECC_ANY): New.
10288 (TEST_SRAM_ECC_ON): New.
10289 (main): Implement HSACOv4 and -msram-ecc=any.
10291 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
10293 * config.in: Regenerate.
10294 * config/gcn/gcn-hsa.h (X_FIJI): New macro.
10295 (X_900): New macro.
10296 (X_906): New macro.
10297 (X_908): New macro.
10298 (A_FIJI): Rename to ...
10299 (S_FIJI): ... this.
10300 (A_900): Rename to ...
10302 (A_906): Rename to ...
10304 (A_908): Rename to ...
10306 (SRAMOPT): New macro.
10307 (ASM_SPEC): Adjust xnack option usage.
10308 * config/gcn/gcn.c (output_file_start): Adjust amdgcn_target usage.
10309 * configure: Regenerate.
10310 * configure.ac: Detect LLVM assembler dialect.
10312 2021-10-07 Richard Biener <rguenther@suse.de>
10314 * tree-pretty-print.c (dump_generic_node): Do not elide
10315 printing '&' when dumping with -gimple.
10317 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
10319 * gimple-range-cache.cc (non_null_ref::adjust_range): Call new
10321 * gimple-range-fold.cc (adjust_pointer_diff_expr): Ditto.
10322 (adjust_imagpart_expr): Ditto.
10323 * value-range.cc (irange::irange_intersect): Call new routine if
10324 RHS is a single pair.
10325 (irange::intersect): New wide_int version.
10326 * value-range.h (class irange): New prototype.
10328 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
10330 * gimple-range-edge.cc (gimple_outgoing_range::gimple_outgoing_range):
10331 Add parameter to limit size when recognizing switches.
10332 (gimple_outgoing_range::edge_range_p): Check size limit.
10333 * gimple-range-edge.h (gimple_outgoing_range): Add size field.
10334 * gimple-range-gori.cc (gori_map::calculate_gori): Ignore switches
10335 that exceed the size limit.
10336 (gori_compute::gori_compute): Add initializer.
10337 * params.opt (evrp-switch-limit): New.
10338 * doc/invoke.texi: Update docs.
10340 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
10342 * value-range.h (irange::set_varying): Use TYPE_MIN_VALUE and
10343 TYPE_MAX_VALUE instead of creating new trees when possible.
10345 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
10347 * gimple-range-cache.cc (non_null_ref::adjust_range): Check for
10348 zero and non-zero more efficently.
10350 2021-10-06 Richard Biener <rguenther@suse.de>
10353 * dumpfile.h (TDF_GIMPLE_VAL): New.
10354 (dump_flag): Re-order and adjust TDF_* flags. Make
10355 the enum uint32_t. Use std::underlying_type in the
10356 operator overloads.
10357 (optgroup_flag): Likewise for the operator overloads.
10358 * tree-pretty-print.c (dump_generic_node): Wrap ADDR_EXPR
10359 in _Literal if TDF_GIMPLE_VAL.
10360 * gimple-pretty-print.c (dump_gimple_assign): Add
10361 TDF_GIMPLE_VAL to flags when dumping operands where only
10362 is_gimple_val are allowed.
10363 (dump_gimple_cond): Likewise.
10365 2021-10-06 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
10367 * gimple-isel.cc (gimple_expand_vec_cond_expr): Remove redundant if
10370 2021-10-05 qing zhao <qing.zhao@oracle.com>
10372 PR middle-end/102359
10373 * gimplify.c (gimplify_decl_expr): Not add initialization for an
10374 auto variable when it has been initialized by frontend.
10376 2021-10-05 Aldy Hernandez <aldyh@redhat.com>
10378 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
10379 Loosen restrictions
10381 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
10383 * common/config/avr/avr-common.c (avr_handle_option): Mark
10384 argument as ATTRIBUTE_UNUSED.
10386 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
10388 * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC):
10389 Undefine before redefinition.
10391 2021-10-05 Richard Biener <rguenther@suse.de>
10393 * toplev.c (no_backend): Remove global var.
10394 (process_options): Pass in no_backend, move post_options
10395 langhook call to toplev::main.
10396 (do_compile): Pass in no_backend, move process_options call
10398 (toplev::run_self_tests): Check no_backend at the caller.
10399 (toplev::main): Call post_options and process_options
10400 split out from do_compile, do self-tests only if
10401 no_backend is initialized.
10403 2021-10-05 Richard Biener <rguenther@suse.de>
10405 * tree-cfg.c (dump_function_to_file): Dump the UID of the
10406 function as part of the name when requested.
10407 * tree-pretty-print.c (dump_function_name): Dump the UID when
10408 requested and the langhook produced the actual name.
10410 2021-10-05 Richard Biener <rguenther@suse.de>
10412 PR middle-end/102587
10413 PR middle-end/102285
10414 * internal-fn.c (expand_DEFERRED_INIT): Fall back to
10415 zero-initialization as last resort, use the constant
10416 size as given by the DEFERRED_INIT argument to build
10419 2021-10-04 Marek Polacek <polacek@redhat.com>
10422 * doc/invoke.texi: Document -Warray-compare.
10424 2021-10-04 Richard Biener <rguenther@suse.de>
10426 * gimplify.c (is_var_need_auto_init): DECL_HARD_REGISTER
10427 variables are not to be initialized.
10429 2021-10-04 Richard Biener <rguenther@suse.de>
10431 * expr.h (non_mem_decl_p): Declare.
10432 (mem_ref_refers_to_non_mem_p): Likewise.
10433 * expr.c (non_mem_decl_p): Export.
10434 (mem_ref_refers_to_non_mem_p): Likewise.
10435 * internal-fn.c (expand_DEFERRED_INIT): Do not expand the LHS
10436 but check the base with mem_ref_refers_to_non_mem_p
10437 and non_mem_decl_p.
10439 2021-10-04 Richard Biener <rguenther@suse.de>
10441 PR tree-optimization/102570
10442 * tree-ssa-sccvn.h (vn_reference_op_struct): Document
10443 we are using clique for the internal function code.
10444 * tree-ssa-sccvn.c (vn_reference_op_eq): Compare the
10445 internal function code.
10446 (print_vn_reference_ops): Print the internal function code.
10447 (vn_reference_op_compute_hash): Hash it.
10448 (copy_reference_ops_from_call): Record it.
10449 (visit_stmt): Remove the restriction around internal function
10451 (fully_constant_vn_reference_p): Use fold_const_call and handle
10452 internal functions.
10453 (vn_reference_eq): Compare call return types.
10454 * tree-ssa-pre.c (create_expression_by_pieces): Handle
10455 generating calls to internal functions.
10456 (compute_avail): Remove the restriction around internal function
10459 2021-10-04 Aldy Hernandez <aldyh@redhat.com>
10461 PR tree-optimization/102560
10462 * gimple-ssa-warn-alloca.c (alloca_call_type): Remove static
10463 marker for invalid_range.
10465 2021-10-04 Richard Biener <rguenther@suse.de>
10467 PR middle-end/102587
10468 * internal-fn.c (expand_DEFERRED_INIT): Guard register
10469 initialization path an avoid initializing VLA registers
10472 2021-10-04 Eric Botcazou <ebotcazou@adacore.com>
10474 * config/rs6000/vxworks.h (TARGET_INIT_LIBFUNCS): Delete.
10476 2021-10-03 Martin Liska <mliska@suse.cz>
10478 * toplev.c (toplev::main): Check opt_index if it is a part
10481 2021-10-02 Aldy Hernandez <aldyh@redhat.com>
10483 PR tree-optimization/102563
10484 * range-op.cc (operator_lshift::op1_range): Do not clobber
10487 2021-10-02 Martin Liska <mliska@suse.cz>
10489 * toplev.c (toplev::main): save_decoded_options[0] is program
10490 name and so it should be skipped.
10492 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
10494 PR tree-optimization/102546
10495 * range-op.cc (operator_lshift::op1_range): Teach range-ops that
10496 X << Y is non-zero implies X is also non-zero.
10498 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
10500 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
10502 * config/aarch64/aarch64-tune.md: Regenerate.
10503 * doc/invoke.texi: Update docs.
10505 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
10507 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
10509 * config/aarch64/aarch64-tune.md: Regenerate.
10510 * doc/invoke.texi: Update docs.
10512 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
10514 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
10516 * config/aarch64/aarch64-tune.md: Regenerate.
10517 * doc/invoke.texi: Update docs.
10519 2021-10-01 Martin Sebor <msebor@redhat.com>
10522 * doc/invoke.texi (-Waddress): Update.
10523 * gengtype.c (write_types): Avoid -Waddress.
10524 * poly-int.h (POLY_SET_COEFF): Avoid using null.
10526 2021-10-01 John David Anglin <danglin@gcc.gnu.org>
10529 * config/pa/pa.c (pa_option_override): Default to dwarf version 4
10532 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
10534 * config/aarch64/aarch64.h (AARCH64_FL_V9): Update value.
10536 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
10538 * gimple-range-path.cc (path_range_query::compute_ranges): Use
10540 * gimple-range-path.h (class path_range_query): Remove shadowed
10542 (path_range_query::get_path_oracle): New.
10544 2021-10-01 Jakub Jelinek <jakub@redhat.com>
10545 Richard Biener <rguenther@suse.de>
10547 PR sanitizer/102515
10548 * doc/invoke.texi (-fsanitize=integer-divide-by-zero): Remove
10549 INT_MIN / -1 division detection from here ...
10550 (-fsanitize=signed-integer-overflow): ... and add it here.
10552 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
10554 * config/aarch64/aarch64-arches.def (AARCH64_ARCH): Added
10556 * config/aarch64/aarch64.h (AARCH64_FL_V9): New.
10557 (AARCH64_FL_FOR_ARCH9): New flags for Armv9-A.
10558 (AARCH64_ISA_V9): New ISA flag.
10559 * doc/invoke.texi: Update docs.
10561 2021-10-01 Martin Liska <mliska@suse.cz>
10563 * toplev.c (toplev::main): Save decoded optimization options.
10564 * toplev.h (save_opt_decoded_options): New.
10565 * doc/extend.texi: Be more clear about optimize and target
10568 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
10570 * explow.c: Include langhooks.h.
10571 (set_stack_check_libfunc): Build a proper function type.
10573 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
10576 * config/i386/i386.c (legitimate_pic_address_disp_p): For PE-COFF do
10577 not return true for external weak function symbols in medium model.
10579 2021-10-01 Jakub Jelinek <jakub@redhat.com>
10581 * tree.h (OMP_CLAUSE_ORDER_REPRODUCIBLE): Define.
10582 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_ORDER>: Print
10583 reproducible: for OMP_CLAUSE_ORDER_REPRODUCIBLE.
10584 * omp-general.c (omp_extract_for_data): If OMP_CLAUSE_ORDER is seen
10585 without OMP_CLAUSE_ORDER_UNCONSTRAINED, overwrite sched_kind to
10586 OMP_CLAUSE_SCHEDULE_STATIC.
10588 2021-10-01 Richard Biener <rguenther@suse.de>
10590 PR middle-end/102518
10591 * tree-inline.c (setup_one_parameter): Avoid substituting
10592 an invariant into contexts where a GIMPLE register is not valid.
10594 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
10596 * config/arm/arm-cpus.in: Add Cortex-R52+ CPU.
10597 * config/arm/arm-tables.opt: Regenerate.
10598 * config/arm/arm-tune.md: Regenerate.
10599 * doc/invoke.texi: Update docs.
10601 2021-09-30 Uroš Bizjak <ubizjak@gmail.com>
10604 * config/i386/i386.md
10605 (sign_extend:WIDE (any_logic:NARROW (memory, immediate)) splitters):
10608 2021-09-30 Tobias Burnus <tobias@codesourcery.com>
10610 * omp-low.c (omp_runtime_api_call): Add omp_aligned_{,c}alloc and
10611 omp_{c,re}alloc, fix omp_alloc/omp_free.
10613 2021-09-30 Martin Liska <mliska@suse.cz>
10615 * defaults.h (ASM_OUTPUT_ASCII): Do not hide global variable
10616 asm_out_file and stream directly to MYFILE.
10618 2021-09-30 Richard Biener <rguenther@suse.de>
10620 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
10621 Restore and fix condition under which we apply npeel to
10622 the DRs misalignment value.
10624 2021-09-30 Richard Biener <rguenther@suse.de>
10626 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
10627 Fix npeel check for variable amount of peeling.
10629 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
10631 * lto-wrapper.c (run_gcc): Plug snprintf overflow.
10633 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
10635 * gimple-range.cc (gimple_ranger::debug): New.
10636 * gimple-range.h (class gimple_ranger): Add debug.
10638 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
10640 PR middle-end/102519
10641 * tree-vrp.c (hybrid_threader::~hybrid_threader): Free m_query.
10643 2021-09-29 Indu Bhagat <indu.bhagat@oracle.com>
10646 * btfout.c (GTY): Add GTY (()) albeit for cosmetic only purpose.
10647 (btf_finalize): Empty the hash_map btf_var_ids.
10649 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
10651 * tree-vrp.c (thread_through_all_blocks): Return bool.
10652 (execute_vrp_threader): Return TODO_* flags.
10653 (pass_data_vrp_threader): Set todo_flags_finish to 0.
10655 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
10657 * timevar.def (TV_TREE_VRP_THREADER): New.
10658 * tree-vrp.c: Use TV_TREE_VRP_THREADER for VRP threader pass.
10660 2021-09-29 David Faust <david.faust@oracle.com>
10662 * config.gcc (bpf-*-*): Do not overwrite extra_headers.
10664 2021-09-29 Jonathan Wright <jonathan.wright@arm.com>
10666 * config/aarch64/aarch64-builtins.c (TYPES_BINOP_PPU): Define
10667 new type qualifier enum.
10668 (TYPES_TERNOP_SSSU): Likewise.
10669 (TYPES_TERNOP_PPPU): Likewise.
10670 * config/aarch64/aarch64-simd-builtins.def: Define PPU, SSU,
10671 PPPU and SSSU builtin generator macros for qtbl1 and qtbx1
10673 * config/aarch64/arm_neon.h (vqtbl1_p8): Use type-qualified
10674 builtin and remove casts.
10675 (vqtbl1_s8): Likewise.
10676 (vqtbl1q_p8): Likewise.
10677 (vqtbl1q_s8): Likewise.
10678 (vqtbx1_s8): Likewise.
10679 (vqtbx1_p8): Likewise.
10680 (vqtbx1q_s8): Likewise.
10681 (vqtbx1q_p8): Likewise.
10682 (vtbl1_p8): Likewise.
10683 (vtbl2_p8): Likewise.
10684 (vtbx2_p8): Likewise.
10686 2021-09-29 Richard Biener <rguenther@suse.de>
10688 * tree-vect-data-refs.c (vect_dr_misalign_for_aligned_access):
10690 (vect_update_misalignment_for_peel): Use it to update
10691 misaligned to the value necessary for an aligned access.
10692 (vect_get_peeling_costs_all_drs): Likewise.
10693 (vect_enhance_data_refs_alignment): Likewise.
10695 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10697 * config/aarch64/aarch64.c (aarch64_expand_cpymem): Count number of
10698 emitted operations and adjust heuristic for code size.
10700 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10702 * config/aarch64/aarch64.c (aarch64_expand_setmem): Count number of
10703 emitted operations and adjust heuristic for code size.
10705 2021-09-29 Jakub Jelinek <jakub@redhat.com>
10707 PR middle-end/102504
10708 * gimplify.c (gimplify_scan_omp_clauses): Use omp_check_private even
10709 in OMP_SCOPE clauses, not just on worksharing construct clauses.
10711 2021-09-28 Geng Qi <gengqi@linux.alibaba.com>
10713 * config/riscv/riscv.md (mulv<mode>4): Call gen_smul<mode>3_highpart.
10714 (<u>mulditi3): Call <su>muldi3_highpart.
10715 (<u>muldi3_highpart): Rename to <su>muldi3_highpart.
10716 (<u>mulsidi3): Call <su>mulsi3_highpart.
10717 (<u>mulsi3_highpart): Rename to <su>mulsi3_highpart.
10719 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
10721 * config/darwin.h (DSYMUTIL_SPEC): Recognize D sources.
10723 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
10725 * config/rs6000/darwin.h (FIXED_R13): Add for PPC64.
10726 (FIRST_SAVED_GP_REGNO): Save from R13 even when it is one
10729 2021-09-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10731 * config/aarch64/aarch64.h (AARCH64_FL_LS64): Define
10732 (AARCH64_FL_V8_7): Likewise.
10733 (AARCH64_FL_FOR_ARCH8_7): Likewise.
10734 * config/aarch64/aarch64-arches.def (armv8.7-a): Define.
10735 * config/aarch64/aarch64-option-extensions.def (ls64): Define.
10736 * doc/invoke.texi: Document the above.
10738 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
10740 * dbgcnt.c (dbg_cnt_counter): New.
10741 * dbgcnt.h (dbg_cnt_counter): New.
10742 * dumpfile.c (dump_options): Add entry for TDF_THREADING.
10743 * dumpfile.h (enum dump_flag): Add TDF_THREADING.
10744 * gimple-range-path.cc (DEBUG_SOLVER): Use TDF_THREADING.
10745 * tree-ssa-threadupdate.c (dump_jump_thread_path): Dump out
10748 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
10750 * cfgcleanup.c (pass_jump::execute): Check
10751 flag_expensive_optimizations.
10752 (pass_jump_after_combine::gate): Same.
10753 * doc/invoke.texi (-fthread-jumps): Enable for -O1.
10754 * opts.c (default_options_table): Enable -fthread-jumps at -O1.
10755 * tree-ssa-threadupdate.c
10756 (fwd_jt_path_registry::remove_jump_threads_including): Bail unless
10759 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
10761 * tree-ssa-reassoc.c (biased_names): New global.
10762 (propagate_bias_p): New function.
10763 (loop_carried_phi): Remove.
10764 (propagate_rank): Propagate bias along single uses.
10765 (get_rank): Update biased_names when needed.
10767 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
10769 * passes.def (pass_reassoc): Rename parameter to early_p.
10770 * tree-ssa-reassoc.c (reassoc_bias_loop_carried_phi_ranks_p):
10772 (phi_rank): Don't bias loop-carried phi ranks
10773 before vectorization pass.
10774 (execute_reassoc): Add bias_loop_carried_phi_ranks_p parameter.
10775 (pass_reassoc::pass_reassoc): Add bias_loop_carried_phi_ranks_p
10777 (pass_reassoc::set_param): Set bias_loop_carried_phi_ranks_p
10779 (pass_reassoc::execute): Pass bias_loop_carried_phi_ranks_p to
10781 (pass_reassoc::bias_loop_carried_phi_ranks_p): New member.
10783 2021-09-28 Jakub Jelinek <jakub@redhat.com>
10786 * config/i386/i386.c (standard_80387_constant_p): Don't recognize
10787 special 80387 instruction XFmode constants if flag_rounding_math.
10789 2021-09-28 Richard Biener <rguenther@suse.de>
10791 PR tree-optimization/100112
10792 * tree-ssa-sccvn.c (visit_reference_op_load): Record the
10793 referece into the hashtable twice in case last_vuse is
10794 different from the original vuse on the stmt.
10796 2021-09-28 Jakub Jelinek <jakub@redhat.com>
10798 PR middle-end/102492
10799 * gimplify.c (gimplify_adjust_omp_clauses_1): Don't call the
10800 omp_finish_clause langhook on implicitly added OMP_CLAUSE_PRIVATE
10801 clauses on SIMD constructs.
10803 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
10805 PR tree-optimization/102511
10806 * gimple-range-path.cc (path_range_query::range_on_path_entry):
10807 Return VARYING when nothing found.
10809 2021-09-28 Hongyu Wang <hongyu.wang@intel.com>
10812 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): Add
10814 (VALID_SSE2_REG_VHF_MODE): Add V4HFmode and V2HFmode.
10815 (VALID_MMX_REG_MODE): Add V4HFmode.
10816 (SSE_REG_MODE_P): Replace VALID_AVX512FP16_REG_MODE with
10817 vector mode condition.
10818 * config/i386/i386.c (classify_argument): Parse V4HF/V2HF
10820 (function_arg_32): Add V4HFmode.
10821 (function_arg_advance_32): Likewise.
10822 * config/i386/i386.md (mode): Add V4HF/V2HF.
10823 (MODE_SIZE): Likewise.
10824 * config/i386/mmx.md (MMXMODE): Add V4HF mode.
10825 (V_32): Add V2HF mode.
10826 (VHF_32_64): New mode iterator.
10827 (*mov<mode>_internal): Adjust sse alternatives to support
10829 (*mov<mode>_internal): Adjust sse alternatives to support
10831 (<insn><mode>3): New define_insn for add/sub/mul/div.
10833 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
10835 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Check
10837 (pass_early_thread_jumps::gate): Same.
10838 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
10839 Return if !flag_thread_jumps.
10840 * tree-ssa-threadupdate.c
10841 (jt_path_registry::register_jump_thread): Assert that
10842 flag_thread_jumps is true.
10844 2021-09-28 liuhongt <hongtao.liu@intel.com>
10847 (simplify_context::simplify_binary_operation_1): Relax
10848 condition of simplifying (vec_concat:M (vec_select op0
10849 index0)(vec_select op1 index1)) to allow different modes
10850 between op0 and M, but have same inner mode.
10852 2021-09-28 liuhongt <hongtao.liu@intel.com>
10854 * config/i386/i386-expand.c (emit_reduc_half): Handle
10855 V8HF/V16HF/V32HFmode.
10856 * config/i386/sse.md (REDUC_SSE_PLUS_MODE): Add V8HF.
10857 (REDUC_SSE_SMINMAX_MODE): Ditto.
10858 (REDUC_PLUS_MODE): Add V16HF and V32HF.
10859 (REDUC_SMINMAX_MODE): Ditto.
10861 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
10863 * gimple-range-path.cc
10864 (path_range_query::precompute_ranges_in_block): Rename to...
10865 (path_range_query::compute_ranges_in_block): ...this.
10866 (path_range_query::precompute_ranges): Rename to...
10867 (path_range_query::compute_ranges): ...this.
10868 (path_range_query::precompute_relations): Rename to...
10869 (path_range_query::compute_relations): ...this.
10870 (path_range_query::precompute_phi_relations): Rename to...
10871 (path_range_query::compute_phi_relations): ...this.
10872 * gimple-range-path.h: Rename precompute* to compute*.
10873 * tree-ssa-threadbackward.c
10874 (back_threader::find_taken_edge_switch): Same.
10875 (back_threader::find_taken_edge_cond): Same.
10876 * tree-ssa-threadedge.c
10877 (hybrid_jt_simplifier::compute_ranges_from_state): Same.
10878 (hybrid_jt_state::register_equivs_stmt): Inline...
10879 * tree-ssa-threadedge.h: ...here.
10881 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
10883 * tree-vrp.c (lhs_of_dominating_assert): Remove.
10884 (class vrp_jt_state): Remove.
10885 (class vrp_jt_simplifier): Remove.
10886 (vrp_jt_simplifier::simplify): Remove.
10887 (class vrp_jump_threader): Remove.
10888 (vrp_jump_threader::vrp_jump_threader): Remove.
10889 (vrp_jump_threader::~vrp_jump_threader): Remove.
10890 (vrp_jump_threader::before_dom_children): Remove.
10891 (vrp_jump_threader::after_dom_children): Remove.
10893 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
10895 * passes.def (pass_vrp_threader): New.
10896 * tree-pass.h (make_pass_vrp_threader): Add make_pass_vrp_threader.
10897 * tree-ssa-threadedge.c (hybrid_jt_state::register_equivs_stmt): New.
10898 (hybrid_jt_simplifier::hybrid_jt_simplifier): New.
10899 (hybrid_jt_simplifier::simplify): New.
10900 (hybrid_jt_simplifier::compute_ranges_from_state): New.
10901 * tree-ssa-threadedge.h (class hybrid_jt_state): New.
10902 (class hybrid_jt_simplifier): New.
10903 * tree-vrp.c (execute_vrp): Remove ASSERT_EXPR based jump
10905 (class hybrid_threader): New.
10906 (hybrid_threader::hybrid_threader): New.
10907 (hybrid_threader::~hybrid_threader): New.
10908 (hybrid_threader::before_dom_children): New.
10909 (hybrid_threader::after_dom_children): New.
10910 (execute_vrp_threader): New.
10911 (class pass_vrp_threader): New.
10912 (make_pass_vrp_threader): New.
10914 2021-09-27 Martin Liska <mliska@suse.cz>
10916 * output.h (enum section_flag): New.
10917 (SECTION_FORGET): Remove.
10918 (SECTION_ENTSIZE): Make it (1UL << 8) - 1.
10919 (SECTION_STYLE_MASK): Define it based on other enum
10921 * varasm.c (switch_to_section): Remove unused handling of
10924 2021-09-27 Martin Liska <mliska@suse.cz>
10926 * common.opt: Add new variable flag_default_complex_method.
10927 * opts.c (finish_options): Handle flags related to
10928 x_flag_complex_method.
10929 * toplev.c (process_options): Remove option handling related
10930 to flag_complex_method.
10932 2021-09-27 Richard Biener <rguenther@suse.de>
10934 PR middle-end/102450
10935 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid using
10936 type_for_size, instead use int_mode_for_size.
10938 2021-09-27 Andrew Pinski <apinski@marvell.com>
10941 * gimplify.c (gimplify_save_expr): Return early
10942 if the type of val is error_mark_node.
10944 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
10946 * tree-ssanames.c (ssa_name_has_boolean_range): Use
10949 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
10951 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
10952 vrp_visit_cond_stmt.
10953 * tree-ssa-dom.c (cprop_operand): Convert to range_query API.
10954 (cprop_into_stmt): Same.
10955 (dom_opt_dom_walker::optimize_stmt): Same.
10957 2021-09-27 Richard Biener <rguenther@suse.de>
10959 PR tree-optimization/97351
10960 PR tree-optimization/97352
10961 PR tree-optimization/82426
10962 * tree-vectorizer.h (dr_misalignment): Add vector type
10964 (aligned_access_p): Likewise.
10965 (known_alignment_for_access_p): Likewise.
10966 (vect_supportable_dr_alignment): Likewise.
10967 (vect_known_alignment_in_bytes): Likewise. Refactor.
10968 (DR_MISALIGNMENT): Remove.
10969 (vect_update_shared_vectype): Likewise.
10970 * tree-vect-data-refs.c (dr_misalignment): Refactor, handle
10971 a vector type with larger alignment requirement and apply
10972 the negative step adjustment here.
10973 (vect_calculate_target_alignment): Remove.
10974 (vect_compute_data_ref_alignment): Get explicit vector type
10975 argument, do not apply a negative step alignment adjustment
10977 (vect_slp_analyze_node_alignment): Re-analyze alignment
10978 when we re-visit the DR with a bigger desired alignment but
10979 keep more precise results from smaller alignments.
10980 * tree-vect-slp.c (vect_update_shared_vectype): Remove.
10981 (vect_slp_analyze_node_operations_1): Do not update the
10982 shared vector type on stmts.
10983 * tree-vect-stmts.c (vect_analyze_stmt): Push/pop the
10984 vector type of an SLP node to the representative stmt-info.
10985 (vect_transform_stmt): Likewise.
10987 2021-09-27 liuhongt <hongtao.liu@intel.com>
10990 2021-09-09 liuhongt <hongtao.liu@intel.com>
10993 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
10994 (reduc_plus_scal_v4sf): .. this, New define_expand.
10995 (reduc_plus_scal_v2df): .. and this, New define_expand.
10997 2021-09-26 liuhongt <hongtao.liu@intel.com>
10999 * doc/extend.texi (Half-Precision): Remove storage only
11000 description for _Float16 w/o avx512fp16.
11002 2021-09-25 Dimitar Dimitrov <dimitar@dinux.eu>
11004 * config/pru/constraints.md (Rrio): New constraint.
11005 * config/pru/predicates.md (regio_operand): New predicate.
11006 * config/pru/pru-pragma.c (pru_register_pragmas): Register
11007 the __regio_symbol address space.
11008 * config/pru/pru-protos.h (pru_symref2ioregno): Declaration.
11009 * config/pru/pru.c (pru_symref2ioregno): New helper function.
11010 (pru_legitimate_address_p): Remove.
11011 (pru_addr_space_legitimate_address_p): Use the address space
11012 aware hook variant.
11013 (pru_nongeneric_pointer_addrspace): New helper function.
11014 (pru_insert_attributes): New function to validate __regio_symbol
11016 (TARGET_INSERT_ATTRIBUTES): New macro.
11017 (TARGET_LEGITIMATE_ADDRESS_P): Remove.
11018 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): New macro.
11019 * config/pru/pru.h (enum reg_class): Add REGIO_REGS class.
11020 * config/pru/pru.md (*regio_readsi): New pattern to read I/O
11022 (*regio_nozext_writesi): New pattern to write to I/O registers.
11023 (*regio_zext_write_r30<EQS0:mode>): Ditto.
11024 * doc/extend.texi: Document the new PRU Named Address Space.
11026 2021-09-24 Patrick Palka <ppalka@redhat.com>
11030 * real.c (encode_ieee_double): Avoid unwanted sign extension.
11031 (encode_ieee_quad): Likewise.
11033 2021-09-24 Vladimir Makarov <vmakarov@redhat.com>
11035 PR rtl-optimization/102147
11036 * ira-build.c (ira_conflict_vector_profitable_p): Make
11037 profitability calculation independent of host compiler pointer and
11038 IRA_INT_BITS sizes.
11040 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
11042 * gimple-range-path.cc (path_range_query::path_range_query):
11043 Move debugging header...
11044 (path_range_query::precompute_ranges): ...here.
11045 (path_range_query::internal_range_of_expr): Do not call
11046 range_on_path_entry if NAME is defined in the current block.
11048 2021-09-24 Richard Biener <rguenther@suse.de>
11050 * cfghooks.c (verify_flow_info): Verify unallocated BB and
11051 edge flags are not set.
11053 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
11055 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
11057 (jt_path_registry::register_jump_thread): Call
11058 cancel_invalid_paths.
11059 * tree-ssa-threadupdate.h (class jt_path_registry): Add
11060 cancel_invalid_paths.
11062 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
11064 PR tree-optimization/102400
11065 * tree-ssa-sccvn.c (vn_reference_insert_pieces): Initialize
11066 result_vdef to zero value.
11068 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
11070 PR tree-optimization/102451
11071 * tree-ssa-dse.c (delete_dead_or_redundant_call): Record bb of stmt
11074 2021-09-24 Hongyu Wang <hongyu.wang@intel.com>
11076 * config/i386/sse.md (cond_<insn><mode>): Extend to support
11078 (cond_mul<mode>): Likewise.
11079 (cond_div<mode>): Likewise.
11080 (cond_<code><mode>): Likewise.
11081 (cond_fma<mode>): Likewise.
11082 (cond_fms<mode>): Likewise.
11083 (cond_fnma<mode>): Likewise.
11084 (cond_fnms<mode>): Likewise.
11086 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
11088 PR tree-optimization/102463
11089 * gimple-range-fold.cc (fold_using_range::relation_fold_and_or): If
11090 there is no range-ops handler, don't look for a relation.
11092 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
11094 * gimple-range-cache.cc (ranger_cache::ranger_cache): Take
11095 non-executable_edge flag as parameter.
11096 * gimple-range-cache.h (ranger_cache): Adjust prototype.
11097 * gimple-range-gori.cc (gori_compute::gori_compute): Take
11098 non-executable_edge flag as parameter.
11099 (gori_compute::outgoing_edge_range_p): Check new flag.
11100 * gimple-range-gori.h (gori_compute): Adjust prototype.
11101 * gimple-range.cc (gimple_ranger::gimple_ranger): Create new flag.
11102 (gimple_ranger::range_on_edge): Check new flag.
11103 * gimple-range.h (gimple_ranger::non_executable_edge_flag): New.
11104 * gimple-ssa-evrp.c (rvrp_folder): Pass ranger flag to simplifer.
11105 (hybrid_folder::hybrid_folder): Set ranger non-executable flag value.
11106 (hybrid_folder::fold_stmt): Set flag value in the simplifer.
11107 * vr-values.c (simplify_using_ranges::set_and_propagate_unexecutable):
11108 Use not_executable flag if provided inmstead of EDGE_EXECUTABLE.
11109 (simplify_using_ranges::simplify_switch_using_ranges): Clear
11110 EDGE_EXECUTABLE like it originally did.
11111 (simplify_using_ranges::cleanup_edges_and_switches): Clear any
11112 NON_EXECUTABLE flags.
11113 (simplify_using_ranges::simplify_using_ranges): Adjust.
11114 * vr-values.h (class simplify_using_ranges): Adjust.
11115 (simplify_using_ranges::set_range_query): Add non-executable flag param.
11117 2021-09-23 Bill Schmidt <wschmidt@linux.ibm.com>
11120 * config/rs6000/rs6000-call.c (rs6000_aggregate_candidate): Detect
11121 zero-width bit fields and return indicator.
11122 (rs6000_discover_homogeneous_aggregate): Diagnose when the
11123 presence of a zero-width bit field changes parameter passing in
11126 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
11128 * gimple-range-fold.cc (fold_using_range::range_of_phi):
11129 Remove dominator check.
11131 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
11133 * gimple-range-path.cc (path_range_query::precompute_relations):
11134 Hoist edge calculations before using EDGE_SUCC.
11136 2021-09-23 Jonathan Wakely <jwakely@redhat.com>
11138 * configure.ac: Fix --with-multilib-list description.
11139 * configure: Regenerate.
11141 2021-09-23 Richard Biener <rguenther@suse.de>
11143 PR tree-optimization/102448
11144 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
11145 Clear alignment info copied from DR_PTR_INFO.
11147 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
11149 * config/i386/i386-expand.c (ix86_use_mask_cmp_p): Enable
11151 * config/i386/sse.md (sseintvecmodelower): Add HF vector modes.
11152 (<avx512>_store<mode>_mask): Extend to support HF vector modes.
11153 (vec_cmp<mode><avx512fmaskmodelower>): Likewise.
11154 (vcond_mask_<mode><avx512fmaskmodelower>): Likewise.
11155 (vcond<mode><mode>): New expander.
11156 (vcond<mode><sseintvecmodelower>): Likewise.
11157 (vcond<sseintvecmodelower><mode>): Likewise.
11158 (vcondu<mode><sseintvecmodelower>): Likewise.
11160 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
11162 * config/i386/sse.md (extend<ssePHmodelower><mode>2):
11164 (extendv4hf<mode>2): Likewise.
11165 (extendv2hfv2df2): Likewise.
11166 (trunc<mode><ssePHmodelower>2): Likewise.
11167 (avx512fp16_vcvt<castmode>2ph_<mode>): Rename to ...
11168 (trunc<mode>v4hf2): ... this, and drop constraints.
11169 (avx512fp16_vcvtpd2ph_v2df): Rename to ...
11170 (truncv2dfv2hf2): ... this, and likewise.
11172 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
11174 * config/i386/sse.md (float<floatunssuffix><mode><ssePHmodelower>2):
11176 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>):
11178 (float<floatunssuffix><mode>v4hf2): ... this, and drop constraints.
11179 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Rename to ...
11180 (float<floatunssuffix>v2div2hf2): ... this, and likewise.
11182 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
11184 * config/i386/i386.md (fix<fixunssuffix>_trunchf<mode>2): New expander.
11185 (fixuns_trunchfhi2): Likewise.
11186 (*fixuns_trunchfsi2zext): New define_insn.
11187 * config/i386/sse.md (ssePHmodelower): New mode_attr.
11188 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
11189 New expander for same element vector fix_truncate.
11190 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
11191 Likewise for V4HF to V4SI/V4DI fix_truncate.
11192 (fix<fixunssuffix>_truncv2hfv2di2):
11193 Likeise for V2HF to V2DI fix_truncate.
11195 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
11197 * config/i386/i386.md (<code>hf3): New expander.
11199 2021-09-23 liuhongt <hongtao.liu@intel.com>
11201 * config/i386/sse.md (FMAMODEM): extend to handle FP16.
11202 (VFH_SF_AVX512VL): Extend to handle HFmode.
11203 (VF_SF_AVX512VL): Deleted.
11205 2021-09-23 liuhongt <hongtao.liu@intel.com>
11207 * config/i386/i386.md (rinthf2): New expander.
11208 (nearbyinthf2): New expander.
11210 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
11212 * tree-ssa-dom.c (class dom_jump_threader_simplifier): Rename...
11213 (class dom_jt_state): ...this and provide virtual overrides.
11214 (dom_jt_state::register_equiv): New.
11215 (class dom_jt_simplifier): Rename from
11216 dom_jump_threader_simplifier.
11217 (dom_jump_threader_simplifier::simplify): Rename...
11218 (dom_jt_simplifier::simplify): ...to this.
11219 (pass_dominator::execute): Use dom_jt_simplifier and
11221 * tree-ssa-threadedge.c (jump_threader::jump_threader):
11223 (jt_state::register_equivs_stmt): Abstract out...
11224 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
11226 (jump_threader::thread_around_empty_blocks): Update state.
11227 (jump_threader::thread_through_normal_block): Same.
11228 (jt_state::jt_state): Remove.
11229 (jt_state::push): Remove pass specific bits. Keep block vector
11231 (jt_state::append_path): New.
11232 (jt_state::pop): Remove pass specific bits.
11233 (jt_state::register_equiv): Same.
11234 (jt_state::record_ranges_from_stmt): Same.
11235 (jt_state::register_equivs_on_edge): Same. Rename...
11236 (jt_state::register_equivs_edge): ...to this.
11237 (jt_state::dump): New.
11238 (jt_state::debug): New.
11239 (jump_threader_simplifier::simplify): Remove.
11240 (jt_state::get_path): New.
11241 * tree-ssa-threadedge.h (class jt_simplifier): Make into a base
11242 class. Expose common functionality as virtual methods.
11243 (class jump_threader_simplifier): Same. Rename...
11244 (class jt_simplifier): ...to this.
11245 * tree-vrp.c (class vrp_jump_threader_simplifier): Rename...
11246 (class vrp_jt_simplifier): ...to this. Provide pass specific
11248 (class vrp_jt_state): New.
11249 (vrp_jump_threader_simplifier::simplify): Rename...
11250 (vrp_jt_simplifier::simplify): ...to this. Inline code from
11251 what used to be the base class.
11252 (vrp_jump_threader::vrp_jump_threader): Use vrp_jt_state and
11255 2021-09-22 Tobias Burnus <tobias@codesourcery.com>
11258 * doc/invoke.texi (-Wno-missing-include-dirs.): Document Fortran
11261 2021-09-22 Roger Sayle <roger@nextmovesoftware.com>
11262 Richard Biener <rguenther@suse.de>
11264 * match.pd (negation simplifications): Implement some negation
11265 folding transformations from fold-const.c's fold_negate_expr.
11266 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Add a SIMPLIFY
11267 argument, to control whether the op should be simplified prior
11268 to looking up/assigning a value number.
11269 (vn_nary_build_or_lookup): Update call to vn_nary_build_or_lookup_1.
11270 (vn_nary_simplify): Likewise.
11271 (visit_nary_op): Likewise, but when constructing a NEGATE_EXPR
11272 now call vn_nary_build_or_lookup_1 disabling simplification.
11274 2021-09-22 Jiufu Guo <guojiufu@linux.ibm.com>
11276 PR tree-optimization/102087
11277 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
11278 Update bound/cmp/control for niter.
11280 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
11282 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
11283 Move check for non-empty BB here.
11284 (fur_source::register_outgoing_edges): ...from here.
11286 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
11288 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
11289 Remove call to improve_range_with_equivs.
11290 (path_range_query::improve_range_with_equivs): Remove
11291 * gimple-range-path.h: Remove improve_range_with_equivs.
11293 2021-09-22 dianhong xu <dianhong.xu@intel.com>
11295 * config/i386/avx512fp16intrin.h:
11296 (_mm512_mask_blend_ph): New intrinsic.
11297 (_mm512_permutex2var_ph): Ditto.
11298 (_mm512_permutexvar_ph): Ditto.
11299 * config/i386/avx512fp16vlintrin.h:
11300 (_mm256_mask_blend_ph): New intrinsic.
11301 (_mm256_permutex2var_ph): Ditto.
11302 (_mm256_permutexvar_ph): Ditto.
11303 (_mm_mask_blend_ph): Ditto.
11304 (_mm_permutex2var_ph): Ditto.
11305 (_mm_permutexvar_ph): Ditto.
11307 2021-09-22 dianhong xu <dianhong.xu@intel.com>
11309 * config/i386/avx512fp16intrin.h: Add new intrinsics.
11310 (_mm512_conj_pch): New intrinsic.
11311 (_mm512_mask_conj_pch): Ditto.
11312 (_mm512_maskz_conj_pch): Ditto.
11313 * config/i386/avx512fp16vlintrin.h: Add new intrinsics.
11314 (_mm256_conj_pch): New intrinsic.
11315 (_mm256_mask_conj_pch): Ditto.
11316 (_mm256_maskz_conj_pch): Ditto.
11317 (_mm_conj_pch): Ditto.
11318 (_mm_mask_conj_pch): Ditto.
11319 (_mm_maskz_conj_pch): Ditto.
11321 2021-09-22 dianhong xu <dianhong.xu@intel.com>
11323 * config/i386/avx512fp16intrin.h (_MM512_REDUCE_OP): New macro
11324 (_mm512_reduce_add_ph): New intrinsic.
11325 (_mm512_reduce_mul_ph): Ditto.
11326 (_mm512_reduce_min_ph): Ditto.
11327 (_mm512_reduce_max_ph): Ditto.
11328 * config/i386/avx512fp16vlintrin.h
11329 (_MM256_REDUCE_OP/_MM_REDUCE_OP): New macro.
11330 (_mm256_reduce_add_ph): New intrinsic.
11331 (_mm256_reduce_mul_ph): Ditto.
11332 (_mm256_reduce_min_ph): Ditto.
11333 (_mm256_reduce_max_ph): Ditto.
11334 (_mm_reduce_add_ph): Ditto.
11335 (_mm_reduce_mul_ph): Ditto.
11336 (_mm_reduce_min_ph): Ditto.
11337 (_mm_reduce_max_ph): Ditto.
11339 2021-09-22 dianhong xu <dianhong.xu@intel.com>
11341 * config/i386/avx512fp16intrin.h (__m512h_u, __m256h_u,
11342 __m128h_u): New typedef.
11343 (_mm512_load_ph): New intrinsic.
11344 (_mm256_load_ph): Ditto.
11345 (_mm_load_ph): Ditto.
11346 (_mm512_loadu_ph): Ditto.
11347 (_mm256_loadu_ph): Ditto.
11348 (_mm_loadu_ph): Ditto.
11349 (_mm512_store_ph): Ditto.
11350 (_mm256_store_ph): Ditto.
11351 (_mm_store_ph): Ditto.
11352 (_mm512_storeu_ph): Ditto.
11353 (_mm256_storeu_ph): Ditto.
11354 (_mm_storeu_ph): Ditto.
11355 (_mm512_abs_ph): Ditto.
11356 * config/i386/avx512fp16vlintrin.h
11357 (_mm_abs_ph): Ditto.
11358 (_mm256_abs_ph): Ditto.
11360 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
11362 * config/s390/tpf.md (prologue_tpf, epilogue_tpf): Add cc clobber.
11364 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
11367 * config/s390/s390.c (s390_expand_insv): Emit a normal move if it
11368 is actually a full copy of the source operand into the target.
11369 Don't emit a strict low part move if source and target mode match.
11371 2021-09-22 Jakub Jelinek <jakub@redhat.com>
11373 PR middle-end/102415
11374 * omp-expand.c (expand_omp_single): If region->exit is NULL,
11375 assert region->entry is GIMPLE_OMP_SCOPE region and return.
11377 2021-09-22 Jakub Jelinek <jakub@redhat.com>
11379 * tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
11380 * tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
11381 arguments from 2 to 3.
11382 * tree-pretty-print.c (dump_omp_clause): Print allocator() around
11383 allocate clause allocator and print align if present.
11384 * omp-low.c (scan_sharing_clauses): Force allocate_map entry even
11385 for omp_default_mem_alloc if align modifier is present. If align
11386 modifier is present, use TREE_LIST to encode both allocator and
11388 (lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
11389 Handle align modifier on allocator clause if present.
11391 2021-09-22 liuhongt <hongtao.liu@intel.com>
11393 * config/i386/i386.md (define_attr "isa"): Add
11395 (define_attr "enabled"): Correspond fma_or_avx512vl to
11396 TARGET_FMA || TARGET_AVX512VL.
11397 * config/i386/mmx.md (fmav2sf4): Extend to AVX512 fma.
11399 (fnmav2sf4): Ditto.
11400 (fnmsv2sf4): Ditto.
11402 2021-09-22 liuhongt <hongtao.liu@intel.com>
11404 * config/i386/i386.md (cstorehf3): New define_expand.
11406 2021-09-22 liuhongt <hongtao.liu@intel.com>
11408 * config/i386/i386.md (<rounding_insn>hf2): New expander.
11409 (sse4_1_round<mode>2): Extend from MODEF to MODEFH.
11410 * config/i386/sse.md (*sse4_1_round<ssescalarmodesuffix>):
11411 Extend from VF_128 to VFH_128.
11413 2021-09-22 liuhongt <hongtao.liu@intel.com>
11415 * config/i386/i386-features.c (i386-features.c): Handle
11417 * config/i386/i386.md (sqrthf2): New expander.
11418 (*sqrthf2): New define_insn.
11419 * config/i386/sse.md
11420 (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
11423 2021-09-22 liuhongt <hongtao.liu@intel.com>
11425 * config/i386/avx512fp16intrin.h (_mm_mask_fcmadd_sch):
11427 (_mm_mask3_fcmadd_sch): Likewise.
11428 (_mm_maskz_fcmadd_sch): Likewise.
11429 (_mm_fcmadd_sch): Likewise.
11430 (_mm_mask_fmadd_sch): Likewise.
11431 (_mm_mask3_fmadd_sch): Likewise.
11432 (_mm_maskz_fmadd_sch): Likewise.
11433 (_mm_fmadd_sch): Likewise.
11434 (_mm_mask_fcmadd_round_sch): Likewise.
11435 (_mm_mask3_fcmadd_round_sch): Likewise.
11436 (_mm_maskz_fcmadd_round_sch): Likewise.
11437 (_mm_fcmadd_round_sch): Likewise.
11438 (_mm_mask_fmadd_round_sch): Likewise.
11439 (_mm_mask3_fmadd_round_sch): Likewise.
11440 (_mm_maskz_fmadd_round_sch): Likewise.
11441 (_mm_fmadd_round_sch): Likewise.
11442 (_mm_fcmul_sch): Likewise.
11443 (_mm_mask_fcmul_sch): Likewise.
11444 (_mm_maskz_fcmul_sch): Likewise.
11445 (_mm_fmul_sch): Likewise.
11446 (_mm_mask_fmul_sch): Likewise.
11447 (_mm_maskz_fmul_sch): Likewise.
11448 (_mm_fcmul_round_sch): Likewise.
11449 (_mm_mask_fcmul_round_sch): Likewise.
11450 (_mm_maskz_fcmul_round_sch): Likewise.
11451 (_mm_fmul_round_sch): Likewise.
11452 (_mm_mask_fmul_round_sch): Likewise.
11453 (_mm_maskz_fmul_round_sch): Likewise.
11454 * config/i386/i386-builtin.def: Add corresponding new builtins.
11455 * config/i386/sse.md
11456 (avx512fp16_fmaddcsh_v8hf_maskz<round_expand_name>): New expander.
11457 (avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>): Ditto.
11458 (avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>):
11460 (avx512fp16_<complexopname>sh_v8hf_mask<round_name>): Ditto.
11461 (avx512fp16_<complexopname>sh_v8hf<mask_scalarc_name><round_scalarcz_name>):
11463 * config/i386/subst.md (mask_scalarcz_name): New.
11464 (mask_scalarc_name): Ditto.
11465 (mask_scalarc_operand3): Ditto.
11466 (mask_scalarcz_operand4): Ditto.
11467 (round_scalarcz_name): Ditto.
11468 (round_scalarc_mask_operand3): Ditto.
11469 (round_scalarcz_mask_operand4): Ditto.
11470 (round_scalarc_mask_op3): Ditto.
11471 (round_scalarcz_mask_op4): Ditto.
11472 (round_scalarcz_constraint): Ditto.
11473 (round_scalarcz_nimm_predicate): Ditto.
11474 (mask_scalarcz): Ditto.
11475 (mask_scalarc): Ditto.
11476 (round_scalarcz): Ditto.
11478 2021-09-22 liuhongt <hongtao.liu@intel.com>
11480 * config/i386/avx512fp16intrin.h (_mm512_fcmadd_pch):
11482 (_mm512_mask_fcmadd_pch): Likewise.
11483 (_mm512_mask3_fcmadd_pch): Likewise.
11484 (_mm512_maskz_fcmadd_pch): Likewise.
11485 (_mm512_fmadd_pch): Likewise.
11486 (_mm512_mask_fmadd_pch): Likewise.
11487 (_mm512_mask3_fmadd_pch): Likewise.
11488 (_mm512_maskz_fmadd_pch): Likewise.
11489 (_mm512_fcmadd_round_pch): Likewise.
11490 (_mm512_mask_fcmadd_round_pch): Likewise.
11491 (_mm512_mask3_fcmadd_round_pch): Likewise.
11492 (_mm512_maskz_fcmadd_round_pch): Likewise.
11493 (_mm512_fmadd_round_pch): Likewise.
11494 (_mm512_mask_fmadd_round_pch): Likewise.
11495 (_mm512_mask3_fmadd_round_pch): Likewise.
11496 (_mm512_maskz_fmadd_round_pch): Likewise.
11497 (_mm512_fcmul_pch): Likewise.
11498 (_mm512_mask_fcmul_pch): Likewise.
11499 (_mm512_maskz_fcmul_pch): Likewise.
11500 (_mm512_fmul_pch): Likewise.
11501 (_mm512_mask_fmul_pch): Likewise.
11502 (_mm512_maskz_fmul_pch): Likewise.
11503 (_mm512_fcmul_round_pch): Likewise.
11504 (_mm512_mask_fcmul_round_pch): Likewise.
11505 (_mm512_maskz_fcmul_round_pch): Likewise.
11506 (_mm512_fmul_round_pch): Likewise.
11507 (_mm512_mask_fmul_round_pch): Likewise.
11508 (_mm512_maskz_fmul_round_pch): Likewise.
11509 * config/i386/avx512fp16vlintrin.h (_mm_fmadd_pch):
11511 (_mm_mask_fmadd_pch): Likewise.
11512 (_mm_mask3_fmadd_pch): Likewise.
11513 (_mm_maskz_fmadd_pch): Likewise.
11514 (_mm256_fmadd_pch): Likewise.
11515 (_mm256_mask_fmadd_pch): Likewise.
11516 (_mm256_mask3_fmadd_pch): Likewise.
11517 (_mm256_maskz_fmadd_pch): Likewise.
11518 (_mm_fcmadd_pch): Likewise.
11519 (_mm_mask_fcmadd_pch): Likewise.
11520 (_mm_mask3_fcmadd_pch): Likewise.
11521 (_mm_maskz_fcmadd_pch): Likewise.
11522 (_mm256_fcmadd_pch): Likewise.
11523 (_mm256_mask_fcmadd_pch): Likewise.
11524 (_mm256_mask3_fcmadd_pch): Likewise.
11525 (_mm256_maskz_fcmadd_pch): Likewise.
11526 (_mm_fmul_pch): Likewise.
11527 (_mm_mask_fmul_pch): Likewise.
11528 (_mm_maskz_fmul_pch): Likewise.
11529 (_mm256_fmul_pch): Likewise.
11530 (_mm256_mask_fmul_pch): Likewise.
11531 (_mm256_maskz_fmul_pch): Likewise.
11532 (_mm_fcmul_pch): Likewise.
11533 (_mm_mask_fcmul_pch): Likewise.
11534 (_mm_maskz_fcmul_pch): Likewise.
11535 (_mm256_fcmul_pch): Likewise.
11536 (_mm256_mask_fcmul_pch): Likewise.
11537 (_mm256_maskz_fcmul_pch): Likewise.
11538 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF_V8HF_V8HF,
11539 V8HF_FTYPE_V16HF_V16HF_V16HF, V16HF_FTYPE_V16HF_V16HF_V16HF_UQI,
11540 V32HF_FTYPE_V32HF_V32HF_V32HF_INT,
11541 V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT): Add new builtin types.
11542 * config/i386/i386-builtin.def: Add new builtins.
11543 * config/i386/i386-expand.c: Handle new builtin types.
11544 * config/i386/subst.md (SUBST_CV): New.
11545 (maskc_name): Ditto.
11546 (maskc_operand3): Ditto.
11548 (sdc_maskz_name): Ditto.
11549 (sdc_mask_op4): Ditto.
11550 (sdc_mask_op5): Ditto.
11551 (sdc_mask_mode512bit_condition): Ditto.
11553 (round_maskc_operand3): Ditto.
11554 (round_sdc_mask_operand4): Ditto.
11555 (round_maskc_op3): Ditto.
11556 (round_sdc_mask_op4): Ditto.
11557 (round_saeonly_sdc_mask_operand5): Ditto.
11558 * config/i386/sse.md (unspec): Add complex fma unspecs.
11559 (avx512fmaskcmode): New.
11560 (UNSPEC_COMPLEX_F_C_MA): Ditto.
11561 (UNSPEC_COMPLEX_F_C_MUL): Ditto.
11562 (complexopname): Ditto.
11563 (<avx512>_fmaddc_<mode>_maskz<round_expand_name>): New expander.
11564 (<avx512>_fcmaddc_<mode>_maskz<round_expand_name>): Ditto.
11565 (fma_<complexopname>_<mode><sdc_maskz_name><round_name>): New
11567 (<avx512>_<complexopname>_<mode>_mask<round_name>): Ditto.
11568 (<avx512>_<complexopname>_<mode><maskc_name><round_name>): Ditto.
11570 2021-09-22 Kewen Lin <linkw@linux.ibm.com>
11572 * config/rs6000/rs6000.opt (rs6000-density-pct-threshold,
11573 rs6000-density-size-threshold, rs6000-density-penalty,
11574 rs6000-density-load-pct-threshold,
11575 rs6000-density-load-num-threshold): New parameter.
11576 * config/rs6000/rs6000.c (rs6000_density_test): Adjust with
11577 corresponding parameters.
11579 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11581 * gimple-range-path.cc (path_range_query::defined_outside_path):
11583 (path_range_query::range_on_path_entry): New.
11584 (path_range_query::internal_range_of_expr): Resolve unknowns
11586 (path_range_query::improve_range_with_equivs): New.
11587 (path_range_query::ssa_range_in_phi): Resolve unknowns with
11589 * gimple-range-path.h (class path_range_query): Add
11590 defined_outside_path, range_on_path_entry, and
11591 improve_range_with_equivs.
11593 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11595 * gimple-range-path.cc (path_range_query::add_to_imports): New.
11596 (path_range_query::add_copies_to_imports): New.
11597 (path_range_query::precompute_ranges): Call
11598 add_copies_to_imports.
11599 * gimple-range-path.h (class path_range_query): Add prototypes
11600 for add_copies_to_imports and add_to_imports.
11602 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11604 * gimple-range-path.cc (path_range_query::range_defined_in_block):
11605 Remove useless code.
11607 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11609 * gimple-range-fold.h (class fur_source): Make oracle protected.
11610 * gimple-range-path.cc (path_range_query::path_range_query): Add
11611 resolve argument. Initialize oracle.
11612 (path_range_query::~path_range_query): Delete oracle.
11613 (path_range_query::range_of_stmt): Adapt to use relations.
11614 (path_range_query::precompute_ranges): Pre-compute relations.
11615 (class jt_fur_source): New
11616 (jt_fur_source::jt_fur_source): New.
11617 (jt_fur_source::register_relation): New.
11618 (jt_fur_source::query_relation): New.
11619 (path_range_query::precompute_relations): New.
11620 (path_range_query::precompute_phi_relations): New.
11621 * gimple-range-path.h (path_range_query): Add resolve argument.
11622 Add oracle, precompute_relations, precompute_phi_relations.
11623 * tree-ssa-threadbackward.c (back_threader::back_threader): Pass
11624 resolve argument to solver.
11626 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11628 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
11629 Rename postfold_gcond_edges to register_outgoing_edges and
11631 (fold_using_range::postfold_gcond_edges): Rename...
11632 (fur_source::register_outgoing_edges): ...to this.
11633 * gimple-range-fold.h (postfold_gcond_edges): Rename to
11634 register_outgoing_edges and move to fur_source.
11636 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11638 * gimple-range-fold.cc (fold_using_range::range_of_phi): Check
11639 dom_info_available_p.
11641 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
11643 * gimple-range-cache.cc (non_null_ref::non_null_ref): Use create
11644 and quick_grow_cleared instead of safe_grow_cleared.
11646 2021-09-21 Thomas Schwinge <thomas@codesourcery.com>
11649 * omp-oacc-neuter-broadcast.cc (oacc_do_neutering): Evaluate
11650 'random ()' to '0'.
11652 2021-09-21 Richard Earnshaw <rearnsha@arm.com>
11654 * configure.ac: Detect when the assembler supports new-style
11655 architecture extensions.
11656 * common/config/arm/arm-common.c (arm_rewrite_mcpu): Return
11657 the full CPU string if the assembler can grok it.
11658 (arm_rewrite_march): Likewise but for the architecture.
11659 * config.in: Regenerate.
11660 * configure: Regenerate.
11662 2021-09-21 Richard Biener <rguenther@suse.de>
11664 PR tree-optimization/102421
11665 * tree-vect-loop.c (vect_dissolve_slp_only_groups): Copy and
11666 adjust alignment info.
11668 2021-09-21 Kewen Lin <linkw@linux.ibm.com>
11670 * ipa-fnsummary.c (ipa_fn_summary_write): Remove inconsistent
11671 bitfield stream out.
11673 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
11675 * gimple-range-fold.cc (fold_using_range::range_of_phi): Ignore
11676 undefined edges, apply an equivalence if appropriate.
11677 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Return
11678 UNDEFINED if EDGE_EXECUTABLE is not set.
11679 * gimple-range.cc (gimple_ranger::gimple_ranger): Set all edges
11680 as EXECUTABLE upon startup.
11681 (gimple_ranger::range_on_edge): Return UNDEFINED for edges without
11682 EDGE_EXECUTABLE set.
11683 * vr-values.c (set_and_propagate_unexecutable): New.
11684 (simplify_using_ranges::fold_cond): Call set_and_propagate.
11685 (simplify_using_ranges::simplify_switch_using_ranges): Ditto.
11686 * vr-values.h: Add prototype.
11688 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
11690 * value-relation.cc (equiv_oracle::register_initial_def): New.
11691 (equiv_oracle::register_relation): Call register_initial_def.
11692 (equiv_oracle::add_equiv_to_block): New. Split register_relation.
11693 (relation_oracle::register_stmt): Check def block of PHI arguments.
11694 * value-relation.h (equiv_oracle): Add new prototypes.
11696 2021-09-20 Matthias Kretz <m.kretz@gsi.de>
11698 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
11699 Define __RECIPROCAL_MATH__, __NO_SIGNED_ZEROS__,
11700 __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__, and
11701 __ROUNDING_MATH__ according to their corresponding flags.
11702 * doc/cpp.texi: Document __RECIPROCAL_MATH__,
11703 __NO_SIGNED_ZEROS__, __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__,
11704 and __ROUNDING_MATH__.
11706 2021-09-20 Richard Biener <rguenther@suse.de>
11708 * tree-vect-stmts.c (vectorizable_load): Use the vectype
11711 2021-09-20 Richard Biener <rguenther@suse.de>
11713 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
11714 Do not compute alignment of the vectorized access here.
11716 2021-09-20 Richard Biener <rguenther@suse.de>
11718 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
11719 Store -1 for runtime alias peeling iterations.
11721 2021-09-20 Richard Biener <rguenther@suse.de>
11723 * config.gcc: Obsolete hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11*.
11725 2021-09-20 Thomas Schwinge <thomas@codesourcery.com>
11727 * input.c (string_concat_db::record_string_concatenation)
11728 (string_concat_db::get_string_concatenation): Skip for
11729 'RESERVED_LOCATION_P'.
11731 2021-09-20 Richard Biener <rguenther@suse.de>
11733 PR tree-optimization/65206
11734 * tree-data-ref.h (struct data_reference): Add alt_indices,
11736 * tree-data-ref.c (free_data_ref): Release alt_indices.
11737 (dr_analyze_indices): Work on struct indices and get DR_REF as tree.
11738 (create_data_ref): Adjust.
11739 (initialize_data_dependence_relation): Split into head
11740 and tail. When the base objects fail to match up try
11741 again with pointer-based analysis of indices.
11742 * tree-vectorizer.c (vec_info_shared::check_datarefs): Do
11743 not compare the lazily computed alternate set of indices.
11745 2021-09-20 Iain Sandoe <iain@sandoe.co.uk>
11747 * gcc.c: Test for execute OK when we find the
11748 programs for assembler linker and dsymutil and those
11749 were specified at configure-time.
11751 2021-09-19 Martin Sebor <msebor@redhat.com>
11753 PR middle-end/102403
11754 * gimple-predicate-analysis.cc (predicate::init_from_control_deps):
11755 Correct a function pre/postcondition.
11757 2021-09-19 Martin Sebor <msebor@redhat.com>
11759 PR middle-end/102243
11760 * tree-ssa-strlen.c (get_range): Handle null cfun.
11762 2021-09-19 Iain Sandoe <iain@sandoe.co.uk>
11764 * config/darwin.h (LINK_COMMAND_SPEC_A): Use Darwin10
11765 unwinder shim as a convenience library.
11767 2021-09-19 Andrew Pinski <apinski@marvell.com>
11769 * doc/install.texi: Add note about
11770 binutils 2.35 is required for LTO usage.
11772 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
11774 * tree-ssa-threadbackward.c
11775 (back_threader_registry::register_path): Use push_edge.
11776 * tree-ssa-threadedge.c
11777 (jump_threader::thread_around_empty_blocks): Same.
11778 (jump_threader::thread_through_normal_block): Same.
11779 (jump_threader::thread_across_edge): Same. Also, use auto_bitmap.
11781 * tree-ssa-threadupdate.c
11782 (jt_path_registry::allocate_thread_edge): Remove.
11783 (jt_path_registry::push_edge): New.
11784 (dump_jump_thread_path): Make static.
11785 * tree-ssa-threadupdate.h (allocate_thread_edge): Remove.
11788 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
11790 * gimple-range-path.cc (path_range_query::path_range_query): Add
11792 (path_range_query::dump): Remove extern declaration of dump_ranger.
11793 * gimple-range-trace.cc (dump_ranger): Add DEBUG_FUNCTION marker.
11794 * gimple-range-trace.h (dump_ranger): Add prototype.
11796 2021-09-19 John Ericson <git@JohnEricson.me>
11798 * gcc.c (find_a_program): New function, factored out of...
11799 (find_a_file): Here.
11800 (execute): Use find_a_program when looking for programs rather
11803 2021-09-19 Matwey V. Kornilov <matwey.kornilov@gmail.com>
11805 * config/avr/avr-mcus.def: Add atmega324pb.
11806 * doc/avr-mmcu.texi: Corresponding changes.
11808 2021-09-19 Roger Sayle <roger@nextmovesoftware.com>
11810 PR middle-end/88173
11811 * match.pd (cmp @0 REAL_CST@1): When @0 is also REAL_CST, apply
11812 the same transformations as to @1. For comparisons against NaN,
11813 don't check HONOR_SNANS but confirm that neither operand is a
11816 2021-09-19 Benjamin Peterson <benjamin@locrian.net>
11818 * attribs.c (make_unique_name): Delete.
11819 * attribs.h (make_unique_name): Delete.
11821 2021-09-19 Andrew Pinski <apinski@marvell.com>
11823 * lra-constraints.c (check_and_process_move): Assert
11824 that dclass and sclass are greater than or equal to NO_REGS.
11826 2021-09-18 Jakub Jelinek <jakub@redhat.com>
11828 * tree.h (OMP_CLAUSE_ORDER_UNCONSTRAINED): Define.
11829 * tree-pretty-print.c (dump_omp_clause): Print unconstrained:
11830 for OMP_CLAUSE_ORDER_UNCONSTRAINED.
11832 2021-09-18 liuhongt <hongtao.liu@intel.com>
11834 * config/i386/i386-features.c (remove_partial_avx_dependency):
11835 Restrict TARGET_USE_VECTOR_FP_CONVERTS and
11836 TARGET_USE_VECTOR_CONVERTS to conversion instructions only.
11838 2021-09-18 Jakub Jelinek <jakub@redhat.com>
11840 * gimplify.c (omp_default_clause): For C/C++ default({,first}private),
11841 if file/namespace scope variable doesn't have predetermined sharing,
11842 treat it as if there was default(none).
11844 2021-09-18 liuhongt <hongtao.liu@intel.com>
11846 * config/i386/avx512fp16intrin.h (_mm_fmadd_sh):
11848 (_mm_mask_fmadd_sh): Likewise.
11849 (_mm_mask3_fmadd_sh): Likewise.
11850 (_mm_maskz_fmadd_sh): Likewise.
11851 (_mm_fmadd_round_sh): Likewise.
11852 (_mm_mask_fmadd_round_sh): Likewise.
11853 (_mm_mask3_fmadd_round_sh): Likewise.
11854 (_mm_maskz_fmadd_round_sh): Likewise.
11855 (_mm_fnmadd_sh): Likewise.
11856 (_mm_mask_fnmadd_sh): Likewise.
11857 (_mm_mask3_fnmadd_sh): Likewise.
11858 (_mm_maskz_fnmadd_sh): Likewise.
11859 (_mm_fnmadd_round_sh): Likewise.
11860 (_mm_mask_fnmadd_round_sh): Likewise.
11861 (_mm_mask3_fnmadd_round_sh): Likewise.
11862 (_mm_maskz_fnmadd_round_sh): Likewise.
11863 (_mm_fmsub_sh): Likewise.
11864 (_mm_mask_fmsub_sh): Likewise.
11865 (_mm_mask3_fmsub_sh): Likewise.
11866 (_mm_maskz_fmsub_sh): Likewise.
11867 (_mm_fmsub_round_sh): Likewise.
11868 (_mm_mask_fmsub_round_sh): Likewise.
11869 (_mm_mask3_fmsub_round_sh): Likewise.
11870 (_mm_maskz_fmsub_round_sh): Likewise.
11871 (_mm_fnmsub_sh): Likewise.
11872 (_mm_mask_fnmsub_sh): Likewise.
11873 (_mm_mask3_fnmsub_sh): Likewise.
11874 (_mm_maskz_fnmsub_sh): Likewise.
11875 (_mm_fnmsub_round_sh): Likewise.
11876 (_mm_mask_fnmsub_round_sh): Likewise.
11877 (_mm_mask3_fnmsub_round_sh): Likewise.
11878 (_mm_maskz_fnmsub_round_sh): Likewise.
11879 * config/i386/i386-builtin-types.def
11880 (V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT): New builtin type.
11881 * config/i386/i386-builtin.def: Add new builtins.
11882 * config/i386/i386-expand.c: Handle new builtin type.
11883 * config/i386/sse.md (fmai_vmfmadd_<mode><round_name>):
11884 Ajdust to support FP16.
11885 (fmai_vmfmsub_<mode><round_name>): Ditto.
11886 (fmai_vmfnmadd_<mode><round_name>): Ditto.
11887 (fmai_vmfnmsub_<mode><round_name>): Ditto.
11888 (*fmai_fmadd_<mode>): Ditto.
11889 (*fmai_fmsub_<mode>): Ditto.
11890 (*fmai_fnmadd_<mode><round_name>): Ditto.
11891 (*fmai_fnmsub_<mode><round_name>): Ditto.
11892 (avx512f_vmfmadd_<mode>_mask<round_name>): Ditto.
11893 (avx512f_vmfmadd_<mode>_mask3<round_name>): Ditto.
11894 (avx512f_vmfmadd_<mode>_maskz<round_expand_name>): Ditto.
11895 (avx512f_vmfmadd_<mode>_maskz_1<round_name>): Ditto.
11896 (*avx512f_vmfmsub_<mode>_mask<round_name>): Ditto.
11897 (avx512f_vmfmsub_<mode>_mask3<round_name>): Ditto.
11898 (*avx512f_vmfmsub_<mode>_maskz_1<round_name>): Ditto.
11899 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
11900 (*avx512f_vmfnmsub_<mode>_mask3<round_name>): Ditto.
11901 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
11902 (*avx512f_vmfnmadd_<mode>_mask<round_name>): Renamed to ...
11903 (avx512f_vmfnmadd_<mode>_mask<round_name>) ... this, and
11904 adjust to support FP16.
11905 (avx512f_vmfnmadd_<mode>_mask3<round_name>): Ditto.
11906 (avx512f_vmfnmadd_<mode>_maskz_1<round_name>): Ditto.
11907 (avx512f_vmfnmadd_<mode>_maskz<round_expand_name>): New
11910 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
11912 * config/i386/sse.md (avx512fmaskmodelower): Extend to support
11914 (maskload<mode><avx512fmaskmodelower>): Ditto.
11915 (maskstore<mode><avx512fmaskmodelower>): Ditto.
11917 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
11919 * config/i386/i386-expand.c (ix86_expand_fp_absneg_operator):
11921 (ix86_expand_copysign): Ditto.
11922 (ix86_expand_xorsign): Ditto.
11923 * config/i386/i386.c (ix86_build_const_vector): Handle HF vector
11925 (ix86_build_signbit_mask): Ditto.
11926 (ix86_can_change_mode_class): Ditto.
11927 * config/i386/i386.md
11928 (SSEMODEF): Add HFmode.
11929 (ssevecmodef): Ditto.
11930 (<code>hf2): New define_expand.
11931 (*<code>hf2_1): New define_insn_and_split.
11932 (copysign<mode>): Extend to support HFmode under AVX512FP16.
11933 (xorsign<mode>): Ditto.
11934 * config/i386/sse.md (VFB): New mode iterator.
11935 (VFB_128_256): Ditto.
11937 (sseintvecmode2): Support HF vector mode.
11938 (<code><mode>2): Use new mode iterator.
11939 (*<code><mode>2): Ditto.
11940 (copysign<mode>3): Ditto.
11941 (xorsign<mode>3): Ditto.
11942 (<code><mode>3<mask_name>): Ditto.
11943 (<code><mode>3<mask_name>): Ditto.
11944 (<sse>_andnot<mode>3<mask_name>): Adjust for HF vector mode.
11945 (<sse>_andnot<mode>3<mask_name>): Ditto.
11946 (*<code><mode>3<mask_name>): Ditto.
11947 (*<code><mode>3<mask_name>): Ditto.
11949 2021-09-18 liuhongt <hongtao.liu@intel.com>
11951 * config/i386/avx512fp16intrin.h (_mm512_mask_fmadd_ph):
11953 (_mm512_mask3_fmadd_ph): Likewise.
11954 (_mm512_maskz_fmadd_ph): Likewise.
11955 (_mm512_fmadd_round_ph): Likewise.
11956 (_mm512_mask_fmadd_round_ph): Likewise.
11957 (_mm512_mask3_fmadd_round_ph): Likewise.
11958 (_mm512_maskz_fmadd_round_ph): Likewise.
11959 (_mm512_fnmadd_ph): Likewise.
11960 (_mm512_mask_fnmadd_ph): Likewise.
11961 (_mm512_mask3_fnmadd_ph): Likewise.
11962 (_mm512_maskz_fnmadd_ph): Likewise.
11963 (_mm512_fnmadd_round_ph): Likewise.
11964 (_mm512_mask_fnmadd_round_ph): Likewise.
11965 (_mm512_mask3_fnmadd_round_ph): Likewise.
11966 (_mm512_maskz_fnmadd_round_ph): Likewise.
11967 (_mm512_fmsub_ph): Likewise.
11968 (_mm512_mask_fmsub_ph): Likewise.
11969 (_mm512_mask3_fmsub_ph): Likewise.
11970 (_mm512_maskz_fmsub_ph): Likewise.
11971 (_mm512_fmsub_round_ph): Likewise.
11972 (_mm512_mask_fmsub_round_ph): Likewise.
11973 (_mm512_mask3_fmsub_round_ph): Likewise.
11974 (_mm512_maskz_fmsub_round_ph): Likewise.
11975 (_mm512_fnmsub_ph): Likewise.
11976 (_mm512_mask_fnmsub_ph): Likewise.
11977 (_mm512_mask3_fnmsub_ph): Likewise.
11978 (_mm512_maskz_fnmsub_ph): Likewise.
11979 (_mm512_fnmsub_round_ph): Likewise.
11980 (_mm512_mask_fnmsub_round_ph): Likewise.
11981 (_mm512_mask3_fnmsub_round_ph): Likewise.
11982 (_mm512_maskz_fnmsub_round_ph): Likewise.
11983 * config/i386/avx512fp16vlintrin.h (_mm256_fmadd_ph):
11985 (_mm256_mask_fmadd_ph): Likewise.
11986 (_mm256_mask3_fmadd_ph): Likewise.
11987 (_mm256_maskz_fmadd_ph): Likewise.
11988 (_mm_fmadd_ph): Likewise.
11989 (_mm_mask_fmadd_ph): Likewise.
11990 (_mm_mask3_fmadd_ph): Likewise.
11991 (_mm_maskz_fmadd_ph): Likewise.
11992 (_mm256_fnmadd_ph): Likewise.
11993 (_mm256_mask_fnmadd_ph): Likewise.
11994 (_mm256_mask3_fnmadd_ph): Likewise.
11995 (_mm256_maskz_fnmadd_ph): Likewise.
11996 (_mm_fnmadd_ph): Likewise.
11997 (_mm_mask_fnmadd_ph): Likewise.
11998 (_mm_mask3_fnmadd_ph): Likewise.
11999 (_mm_maskz_fnmadd_ph): Likewise.
12000 (_mm256_fmsub_ph): Likewise.
12001 (_mm256_mask_fmsub_ph): Likewise.
12002 (_mm256_mask3_fmsub_ph): Likewise.
12003 (_mm256_maskz_fmsub_ph): Likewise.
12004 (_mm_fmsub_ph): Likewise.
12005 (_mm_mask_fmsub_ph): Likewise.
12006 (_mm_mask3_fmsub_ph): Likewise.
12007 (_mm_maskz_fmsub_ph): Likewise.
12008 (_mm256_fnmsub_ph): Likewise.
12009 (_mm256_mask_fnmsub_ph): Likewise.
12010 (_mm256_mask3_fnmsub_ph): Likewise.
12011 (_mm256_maskz_fnmsub_ph): Likewise.
12012 (_mm_fnmsub_ph): Likewise.
12013 (_mm_mask_fnmsub_ph): Likewise.
12014 (_mm_mask3_fnmsub_ph): Likewise.
12015 (_mm_maskz_fnmsub_ph): Likewise.
12016 * config/i386/i386-builtin.def: Add corresponding new builtins.
12017 * config/i386/sse.md
12018 (<avx512>_fmadd_<mode>_maskz<round_expand_name>): Adjust to
12019 support HF vector modes.
12020 (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
12022 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
12023 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
12024 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
12025 (<avx512>_fmadd_<mode>_mask<round_name>): Ditto.
12026 (<avx512>_fmadd_<mode>_mask3<round_name>): Ditto.
12027 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Ditto.
12028 (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>):
12030 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
12031 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
12032 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
12033 (<avx512>_fmsub_<mode>_mask<round_name>): Ditto.
12034 (<avx512>_fmsub_<mode>_mask3<round_name>): Ditto.
12035 (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>):
12037 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
12038 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
12039 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
12040 (<avx512>_fnmadd_<mode>_mask<round_name>): Ditto.
12041 (<avx512>_fnmadd_<mode>_mask3<round_name>): Ditto.
12042 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Ditto.
12043 (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>):
12045 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
12046 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
12047 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
12048 (<avx512>_fnmsub_<mode>_mask<round_name>): Ditto.
12049 (<avx512>_fnmsub_<mode>_mask3<round_name>): Ditto.
12051 2021-09-18 liuhongt <hongtao.liu@intel.com>
12053 * config/i386/avx512fp16intrin.h (_mm512_fmaddsub_ph):
12055 (_mm512_mask_fmaddsub_ph): Likewise.
12056 (_mm512_mask3_fmaddsub_ph): Likewise.
12057 (_mm512_maskz_fmaddsub_ph): Likewise.
12058 (_mm512_fmaddsub_round_ph): Likewise.
12059 (_mm512_mask_fmaddsub_round_ph): Likewise.
12060 (_mm512_mask3_fmaddsub_round_ph): Likewise.
12061 (_mm512_maskz_fmaddsub_round_ph): Likewise.
12062 (_mm512_mask_fmsubadd_ph): Likewise.
12063 (_mm512_mask3_fmsubadd_ph): Likewise.
12064 (_mm512_maskz_fmsubadd_ph): Likewise.
12065 (_mm512_fmsubadd_round_ph): Likewise.
12066 (_mm512_mask_fmsubadd_round_ph): Likewise.
12067 (_mm512_mask3_fmsubadd_round_ph): Likewise.
12068 (_mm512_maskz_fmsubadd_round_ph): Likewise.
12069 * config/i386/avx512fp16vlintrin.h (_mm256_fmaddsub_ph):
12071 (_mm256_mask_fmaddsub_ph): Likewise.
12072 (_mm256_mask3_fmaddsub_ph): Likewise.
12073 (_mm256_maskz_fmaddsub_ph): Likewise.
12074 (_mm_fmaddsub_ph): Likewise.
12075 (_mm_mask_fmaddsub_ph): Likewise.
12076 (_mm_mask3_fmaddsub_ph): Likewise.
12077 (_mm_maskz_fmaddsub_ph): Likewise.
12078 (_mm256_fmsubadd_ph): Likewise.
12079 (_mm256_mask_fmsubadd_ph): Likewise.
12080 (_mm256_mask3_fmsubadd_ph): Likewise.
12081 (_mm256_maskz_fmsubadd_ph): Likewise.
12082 (_mm_fmsubadd_ph): Likewise.
12083 (_mm_mask_fmsubadd_ph): Likewise.
12084 (_mm_mask3_fmsubadd_ph): Likewise.
12085 (_mm_maskz_fmsubadd_ph): Likewise.
12086 * config/i386/i386-builtin.def: Add corresponding new builtins.
12087 * config/i386/sse.md (VFH_SF_AVX512VL): New mode iterator.
12088 * (<avx512>_fmsubadd_<mode>_maskz<round_expand_name>): New expander.
12089 * (<avx512>_fmaddsub_<mode>_maskz<round_expand_name>): Use
12091 * (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
12093 * (<avx512>_fmaddsub_<mode>_mask<round_name>): Ditto.
12094 * (<avx512>_fmaddsub_<mode>_mask3<round_name>): Ditto.
12095 * (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
12097 * (<avx512>_fmsubadd_<mode>_mask<round_name>): Ditto.
12098 * (<avx512>_fmsubadd_<mode>_mask3<round_name>): Ditto.
12100 2021-09-18 liuhongt <hongtao.liu@intel.com>
12103 * config/i386/i386.c (ix86_print_operand): Handle
12104 V8HF/V16HF/V32HFmode.
12105 * config/i386/i386.h (VALID_BCST_MODE_P): Add HFmode.
12106 * config/i386/sse.md (avx512bcst): Remove.
12108 2021-09-17 Martin Sebor <msebor@redhat.com>
12110 * Makefile.in (OBJS): Add gimple-predicate-analysis.o.
12111 * tree-ssa-uninit.c (max_phi_args): Move to gimple-predicate-analysis.
12112 (MASK_SET_BIT, MASK_TEST_BIT, MASK_EMPTY): Same.
12113 (check_defs): Add comment.
12114 (can_skip_redundant_opnd): Update comment.
12115 (compute_uninit_opnds_pos): Adjust to namespace change.
12116 (find_pdom): Move to gimple-predicate-analysis.cc.
12118 (struct uninit_undef_val_t): New.
12119 (is_non_loop_exit_postdominating): Move to gimple-predicate-analysis.cc.
12120 (find_control_equiv_block): Same.
12121 (MAX_NUM_CHAINS, MAX_CHAIN_LEN, MAX_POSTDOM_CHECK): Same.
12122 (MAX_SWITCH_CASES): Same.
12123 (compute_control_dep_chain): Same.
12124 (find_uninit_use): Use predicate analyzer.
12125 (struct pred_info): Move to gimple-predicate-analysis.
12126 (convert_control_dep_chain_into_preds): Same.
12127 (find_predicates): Same.
12128 (collect_phi_def_edges): Same.
12129 (warn_uninitialized_phi): Use predicate analyzer.
12130 (find_def_preds): Move to gimple-predicate-analysis.
12131 (dump_pred_info): Same.
12132 (dump_pred_chain): Same.
12133 (dump_predicates): Same.
12134 (destroy_predicate_vecs): Remove.
12135 (execute_late_warn_uninitialized): New.
12136 (get_cmp_code): Move to gimple-predicate-analysis.
12137 (is_value_included_in): Same.
12138 (value_sat_pred_p): Same.
12139 (find_matching_predicate_in_rest_chains): Same.
12140 (is_use_properly_guarded): Same.
12141 (prune_uninit_phi_opnds): Same.
12142 (find_var_cmp_const): Same.
12143 (use_pred_not_overlap_with_undef_path_pred): Same.
12144 (pred_equal_p): Same.
12145 (is_neq_relop_p): Same.
12146 (is_neq_zero_form_p): Same.
12147 (pred_expr_equal_p): Same.
12148 (is_pred_expr_subset_of): Same.
12149 (is_pred_chain_subset_of): Same.
12150 (is_included_in): Same.
12151 (is_superset_of): Same.
12152 (pred_neg_p): Same.
12153 (simplify_pred): Same.
12154 (simplify_preds_2): Same.
12155 (simplify_preds_3): Same.
12156 (simplify_preds_4): Same.
12157 (simplify_preds): Same.
12159 (push_to_worklist): Same.
12160 (get_pred_info_from_cmp): Same.
12161 (is_degenerated_phi): Same.
12162 (normalize_one_pred_1): Same.
12163 (normalize_one_pred): Same.
12164 (normalize_one_pred_chain): Same.
12165 (normalize_preds): Same.
12166 (can_one_predicate_be_invalidated_p): Same.
12167 (can_chain_union_be_invalidated_p): Same.
12168 (uninit_uses_cannot_happen): Same.
12169 (pass_late_warn_uninitialized::execute): Define.
12170 * gimple-predicate-analysis.cc: New file.
12171 * gimple-predicate-analysis.h: New file.
12173 2021-09-17 Julian Brown <julian@codesourcery.com>
12175 * config/gcn/gcn.c (gimple.h): Include.
12176 (gcn_fork_join): Emit barrier for worker-level joins.
12177 * omp-oacc-neuter-broadcast.cc (find_local_vars_to_propagate): Add
12178 writes_gang_private bitmap parameter. Set bit for blocks
12179 containing gang-private variable writes.
12180 (worker_single_simple): Don't emit barrier after predicated block.
12181 (worker_single_copy): Don't emit barrier if we're not broadcasting
12182 anything and the block contains no gang-private writes.
12183 (neuter_worker_single): Don't predicate blocks that only contain
12184 NOPs or internal marker functions. Pass has_gang_private_write
12185 argument to worker_single_copy.
12186 (oacc_do_neutering): Add writes_gang_private bitmap handling.
12188 2021-09-17 Julian Brown <julian@codesourcery.com>
12190 * config/gcn/gcn-protos.h
12191 (gcn_goacc_create_worker_broadcast_record): Update prototype.
12192 * config/gcn/gcn-tree.c (gcn_goacc_get_worker_red_decl): Use
12193 preallocated block of LDS memory. Do not cache/share decls for
12194 reduction temporaries between invocations.
12195 (gcn_goacc_reduction_teardown): Unshare VAR on second use.
12196 (gcn_goacc_create_worker_broadcast_record): Add OFFSET parameter
12197 and return temporary LDS space at that offset. Return pointer in
12199 * config/gcn/gcn.c (acc_lds_size, gang_private_hwm, lds_allocs):
12201 (ACC_LDS_SIZE): Define as acc_lds_size.
12202 (gcn_init_machine_status): Don't initialise lds_allocated,
12203 lds_allocs, reduc_decls fields of machine function struct.
12204 (gcn_option_override): Handle default size for gang-private
12205 variables and -mgang-private-size option.
12206 (gcn_expand_prologue): Use LDS_SIZE instead of LDS_SIZE-1 when
12207 initialising M0_REG.
12208 (gcn_shared_mem_layout): New function.
12209 (gcn_print_lds_decl): Update comment. Use global lds_allocs map and
12210 gang_private_hwm variable.
12211 (TARGET_GOACC_SHARED_MEM_LAYOUT): Define target hook.
12212 * config/gcn/gcn.h (machine_function): Remove lds_allocated,
12213 lds_allocs, reduc_decls. Add reduction_base, reduction_limit.
12214 * config/gcn/gcn.opt (gang_private_size_opt): New global.
12215 (mgang-private-size=): New option.
12216 * doc/tm.texi.in (TARGET_GOACC_SHARED_MEM_LAYOUT): Place
12217 documentation hook.
12218 * doc/tm.texi: Regenerate.
12219 * omp-oacc-neuter-broadcast.cc (targhooks.h, diagnostic-core.h):
12221 (build_sender_ref): Handle sender_decl being pointer.
12222 (worker_single_copy): Add PLACEMENT and ISOLATE_BROADCASTS
12223 parameters. Pass placement argument to
12224 create_worker_broadcast_record hook invocations. Handle
12225 sender_decl being pointer and isolate_broadcasts inserting extra
12227 (blk_offset_map_t): Add typedef.
12228 (neuter_worker_single): Add BLK_OFFSET_MAP parameter. Pass
12229 preallocated range to worker_single_copy call.
12230 (dfs_broadcast_reachable_1): New function.
12231 (idx_decl_pair_t, used_range_vec_t): New typedefs.
12232 (sort_size_descending): New function.
12233 (addr_range): New class.
12234 (splay_tree_compare_addr_range, splay_tree_free_key)
12235 (first_fit_range, merge_ranges_1, merge_ranges): New functions.
12236 (execute_omp_oacc_neuter_broadcast): Rename to...
12237 (oacc_do_neutering): ... this. Add BOUNDS_LO, BOUNDS_HI
12238 parameters. Arrange layout of shared memory for broadcast
12240 (execute_omp_oacc_neuter_broadcast): New function.
12241 (pass_omp_oacc_neuter_broadcast::gate): Remove num_workers==1
12242 handling from here. Enable pass for all OpenACC routines in order
12243 to call shared memory-layout hook.
12244 * target.def (create_worker_broadcast_record): Add OFFSET
12246 (shared_mem_layout): New hook.
12248 2021-09-17 Julian Brown <julian@codesourcery.com>
12249 Thomas Schwinge <thomas@codesourcery.com>
12251 * omp-oacc-neuter-broadcast.cc
12252 (pass_omp_oacc_neuter_broadcast::gate): Disable if num_workers is
12254 (execute_omp_oacc_neuter_broadcast): Adjust.
12256 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
12258 * value-relation.cc (class equiv_chain): Move to header file.
12259 (path_oracle::path_oracle): New.
12260 (path_oracle::~path_oracle): New.
12261 (path_oracle::register_relation): New.
12262 (path_oracle::query_relation): New.
12263 (path_oracle::reset_path): New.
12264 (path_oracle::dump): New.
12265 * value-relation.h (class equiv_chain): Move to here.
12266 (class path_oracle): New.
12268 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
12270 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a DOM
12272 * gimple-range-fold.cc (fur_depend::register_relation): Use
12273 register_stmt/edge routines.
12274 * value-relation.cc (equiv_chain::find): Relocate from equiv_oracle.
12275 (equiv_oracle::equiv_oracle): Create self equivalence cache.
12276 (equiv_oracle::~equiv_oracle): Release same.
12277 (equiv_oracle::equiv_set): Return entry from self equiv cache if there
12278 are no equivalences.
12279 (equiv_oracle::find_equiv_block): Move list find to equiv_chain.
12280 (equiv_oracle::register_relation): Rename from register_equiv.
12281 (relation_chain_head::find_relation): Relocate from dom_oracle.
12282 (relation_oracle::register_stmt): New.
12283 (relation_oracle::register_edge): New.
12284 (dom_oracle::*): Rename from relation_oracle.
12285 (dom_oracle::register_relation): Adjust to call equiv_oracle.
12286 (dom_oracle::set_one_relation): Split from register_relation.
12287 (dom_oracle::register_transitives): Consolidate 2 methods.
12288 (dom_oracle::find_relation_block): Move core to relation_chain.
12289 (dom_oracle::query_relation): Rename from find_relation_dom and adjust.
12290 * value-relation.h (class relation_oracle): New pure virtual base.
12291 (class equiv_oracle): Inherit from relation_oracle and adjust.
12292 (class dom_oracle): Rename from old relation_oracle and adjust.
12294 2021-09-17 Martin Sebor <msebor@redhat.com>
12296 PR middle-end/102200
12297 * pointer-query.cc (access_ref::inform_access): Handle MIN/MAX_EXPR.
12298 (handle_min_max_size): Change argument. Store original SSA_NAME for
12299 operands to potentially distinct (sub)objects.
12300 (compute_objsize_r): Adjust call to the above.
12302 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
12304 * config/rs6000/rs6000.c (rs6000-builtins.h): New include.
12305 (rs6000_new_builtin_vectorized_function): New function.
12306 (rs6000_new_builtin_md_vectorized_function): Likewise.
12307 (rs6000_builtin_vectorized_function): Call
12308 rs6000_new_builtin_vectorized_function.
12309 (rs6000_builtin_md_vectorized_function): Call
12310 rs6000_new_builtin_md_vectorized_function.
12312 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
12314 * config/rs6000/rs6000-builtin-new.def (ASSEMBLE_ACC): Add mmaint flag.
12315 (ASSEMBLE_PAIR): Likewise.
12316 (BUILD_ACC): Likewise.
12317 (DISASSEMBLE_ACC): Likewise.
12318 (DISASSEMBLE_PAIR): Likewise.
12319 (PMXVBF16GER2): Likewise.
12320 (PMXVBF16GER2NN): Likewise.
12321 (PMXVBF16GER2NP): Likewise.
12322 (PMXVBF16GER2PN): Likewise.
12323 (PMXVBF16GER2PP): Likewise.
12324 (PMXVF16GER2): Likewise.
12325 (PMXVF16GER2NN): Likewise.
12326 (PMXVF16GER2NP): Likewise.
12327 (PMXVF16GER2PN): Likewise.
12328 (PMXVF16GER2PP): Likewise.
12329 (PMXVF32GER): Likewise.
12330 (PMXVF32GERNN): Likewise.
12331 (PMXVF32GERNP): Likewise.
12332 (PMXVF32GERPN): Likewise.
12333 (PMXVF32GERPP): Likewise.
12334 (PMXVF64GER): Likewise.
12335 (PMXVF64GERNN): Likewise.
12336 (PMXVF64GERNP): Likewise.
12337 (PMXVF64GERPN): Likewise.
12338 (PMXVF64GERPP): Likewise.
12339 (PMXVI16GER2): Likewise.
12340 (PMXVI16GER2PP): Likewise.
12341 (PMXVI16GER2S): Likewise.
12342 (PMXVI16GER2SPP): Likewise.
12343 (PMXVI4GER8): Likewise.
12344 (PMXVI4GER8PP): Likewise.
12345 (PMXVI8GER4): Likewise.
12346 (PMXVI8GER4PP): Likewise.
12347 (PMXVI8GER4SPP): Likewise.
12348 (XVBF16GER2): Likewise.
12349 (XVBF16GER2NN): Likewise.
12350 (XVBF16GER2NP): Likewise.
12351 (XVBF16GER2PN): Likewise.
12352 (XVBF16GER2PP): Likewise.
12353 (XVF16GER2): Likewise.
12354 (XVF16GER2NN): Likewise.
12355 (XVF16GER2NP): Likewise.
12356 (XVF16GER2PN): Likewise.
12357 (XVF16GER2PP): Likewise.
12358 (XVF32GER): Likewise.
12359 (XVF32GERNN): Likewise.
12360 (XVF32GERNP): Likewise.
12361 (XVF32GERPN): Likewise.
12362 (XVF32GERPP): Likewise.
12363 (XVF64GER): Likewise.
12364 (XVF64GERNN): Likewise.
12365 (XVF64GERNP): Likewise.
12366 (XVF64GERPN): Likewise.
12367 (XVF64GERPP): Likewise.
12368 (XVI16GER2): Likewise.
12369 (XVI16GER2PP): Likewise.
12370 (XVI16GER2S): Likewise.
12371 (XVI16GER2SPP): Likewise.
12372 (XVI4GER8): Likewise.
12373 (XVI4GER8PP): Likewise.
12374 (XVI8GER4): Likewise.
12375 (XVI8GER4PP): Likewise.
12376 (XVI8GER4SPP): Likewise.
12377 (XXMFACC): Likewise.
12378 (XXMTACC): Likewise.
12379 (XXSETACCZ): Likewise.
12380 (ASSEMBLE_PAIR_V): Likewise.
12381 (BUILD_PAIR): Likewise.
12382 (DISASSEMBLE_PAIR_V): Likewise.
12385 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_mma_builtin):
12386 Handle RS6000_BIF_LXVP and RS6000_BIF_STXVP.
12387 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add ismmaint.
12388 (parse_bif_attrs): Handle ismmaint.
12389 (write_decls): Add bif_mmaint_bit and bif_is_mmaint.
12390 (write_bif_static_init): Handle ismmaint.
12392 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
12394 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin): New
12396 (rs6000_gimple_fold_builtin): Call rs6000_gimple_fold_new_builtin.
12397 (rs6000_new_builtin_valid_without_lhs): New function.
12398 (rs6000_gimple_fold_new_mma_builtin): Likewise.
12399 (rs6000_gimple_fold_new_builtin): Likewise.
12401 2021-09-17 Thomas Schwinge <thomas@codesourcery.com>
12403 * hash-table.h (hash_table<Descriptor, Lazy, Allocator>::expand):
12404 Destruct stale Value objects.
12405 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
12408 2021-09-17 Roger Sayle <roger@nextmovesoftware.com>
12411 * match.pd (shift optimizations): Disable recent sign-changing
12412 optimization for shifts by zero, these will be folded later.
12414 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
12416 * config/rs6000/rs6000-builtin-new.def (__builtin_mffsl): Move from
12417 [power9] to [always].
12419 2021-09-17 Richard Biener <rguenther@suse.de>
12421 * tree-vect-stmts.c (vectorizable_load): Do not frob
12424 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
12426 * config/i386/i386-features.c (remove_partial_avx_dependency):
12427 Also check TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY and
12428 and TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY before generating
12430 * config/i386/i386.h (TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY):
12432 (TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
12433 * config/i386/i386.md (SSE FP to FP splitters): Replace
12434 TARGET_SSE_PARTIAL_REG_DEPENDENCY with
12435 TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY.
12436 (SSE INT to FP splitter): Replace TARGET_SSE_PARTIAL_REG_DEPENDENCY
12437 with TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY.
12438 * config/i386/x86-tune.def
12439 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): New.
12440 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
12442 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
12445 * config/i386/i386-features.c (remove_partial_avx_dependency):
12446 Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
12447 before generating vxorps.
12449 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
12451 * config/i386/i386-options.c (processor_cost_table): Use
12452 tremont_cost for Tremont.
12453 * config/i386/x86-tune-costs.h (tremont_memcpy): New.
12454 (tremont_memset): Likewise.
12455 (tremont_cost): Likewise.
12456 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
12457 Enable for Tremont.
12459 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
12461 * common/config/i386/i386-common.c: Use Haswell scheduling model
12463 * config/i386/i386.c (ix86_sched_init_global): Prepare for Tremont
12465 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Tremont
12467 (ix86_adjust_cost): Handle Tremont.
12468 * config/i386/x86-tune.def (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY):
12469 Enable for Tremont.
12470 (X86_TUNE_USE_LEAVE): Likewise.
12471 (X86_TUNE_PUSH_MEMORY): Likewise.
12472 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
12473 (X86_TUNE_USE_CLTD): Likewise.
12474 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
12475 (X86_TUNE_AVOID_MFENCE): Likewise.
12476 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
12477 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
12478 (X86_TUNE_ACCUMULATE_OUTGOING_ARGS): Disable for Tremont.
12479 (X86_TUNE_FOUR_JUMP_LIMIT): Likewise.
12480 (X86_TUNE_OPT_AGU): Likewise.
12481 (X86_TUNE_AVOID_LEA_FOR_ADDR): Likewise.
12482 (X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE): Likewise.
12483 (X86_TUNE_EXPAND_ABS): Likewise.
12484 (X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS): Likewise.
12485 (X86_TUNE_SLOW_PSHUFB): Likewise.
12487 2021-09-17 Eric Botcazou <ebotcazou@adacore.com>
12489 PR rtl-optimization/102306
12490 * combine.c (try_combine): Abort the combination if we are about to
12491 duplicate volatile references.
12493 2021-09-17 liuhongt <hongtao.liu@intel.com>
12495 * config/i386/avx512fp16intrin.h (_mm_undefined_ph):
12497 (_mm256_undefined_ph): Likewise.
12498 (_mm512_undefined_ph): Likewise.
12499 (_mm_cvtsh_h): Likewise.
12500 (_mm256_cvtsh_h): Likewise.
12501 (_mm512_cvtsh_h): Likewise.
12502 (_mm512_castph_ps): Likewise.
12503 (_mm512_castph_pd): Likewise.
12504 (_mm512_castph_si512): Likewise.
12505 (_mm512_castph512_ph128): Likewise.
12506 (_mm512_castph512_ph256): Likewise.
12507 (_mm512_castph128_ph512): Likewise.
12508 (_mm512_castph256_ph512): Likewise.
12509 (_mm512_zextph128_ph512): Likewise.
12510 (_mm512_zextph256_ph512): Likewise.
12511 (_mm512_castps_ph): Likewise.
12512 (_mm512_castpd_ph): Likewise.
12513 (_mm512_castsi512_ph): Likewise.
12514 * config/i386/avx512fp16vlintrin.h (_mm_castph_ps):
12516 (_mm256_castph_ps): Likewise.
12517 (_mm_castph_pd): Likewise.
12518 (_mm256_castph_pd): Likewise.
12519 (_mm_castph_si128): Likewise.
12520 (_mm256_castph_si256): Likewise.
12521 (_mm_castps_ph): Likewise.
12522 (_mm256_castps_ph): Likewise.
12523 (_mm_castpd_ph): Likewise.
12524 (_mm256_castpd_ph): Likewise.
12525 (_mm_castsi128_ph): Likewise.
12526 (_mm256_castsi256_ph): Likewise.
12527 (_mm256_castph256_ph128): Likewise.
12528 (_mm256_castph128_ph256): Likewise.
12529 (_mm256_zextph128_ph256): Likewise.
12531 2021-09-17 liuhongt <hongtao.liu@intel.com>
12533 * config/i386/avx512fp16intrin.h (_mm_cvtsh_ss):
12535 (_mm_mask_cvtsh_ss): Likewise.
12536 (_mm_maskz_cvtsh_ss): Likewise.
12537 (_mm_cvtsh_sd): Likewise.
12538 (_mm_mask_cvtsh_sd): Likewise.
12539 (_mm_maskz_cvtsh_sd): Likewise.
12540 (_mm_cvt_roundsh_ss): Likewise.
12541 (_mm_mask_cvt_roundsh_ss): Likewise.
12542 (_mm_maskz_cvt_roundsh_ss): Likewise.
12543 (_mm_cvt_roundsh_sd): Likewise.
12544 (_mm_mask_cvt_roundsh_sd): Likewise.
12545 (_mm_maskz_cvt_roundsh_sd): Likewise.
12546 (_mm_cvtss_sh): Likewise.
12547 (_mm_mask_cvtss_sh): Likewise.
12548 (_mm_maskz_cvtss_sh): Likewise.
12549 (_mm_cvtsd_sh): Likewise.
12550 (_mm_mask_cvtsd_sh): Likewise.
12551 (_mm_maskz_cvtsd_sh): Likewise.
12552 (_mm_cvt_roundss_sh): Likewise.
12553 (_mm_mask_cvt_roundss_sh): Likewise.
12554 (_mm_maskz_cvt_roundss_sh): Likewise.
12555 (_mm_cvt_roundsd_sh): Likewise.
12556 (_mm_mask_cvt_roundsd_sh): Likewise.
12557 (_mm_maskz_cvt_roundsd_sh): Likewise.
12558 * config/i386/i386-builtin-types.def
12559 (V8HF_FTYPE_V2DF_V8HF_V8HF_UQI_INT,
12560 V8HF_FTYPE_V4SF_V8HF_V8HF_UQI_INT,
12561 V2DF_FTYPE_V8HF_V2DF_V2DF_UQI_INT,
12562 V4SF_FTYPE_V8HF_V4SF_V4SF_UQI_INT): Add new builtin types.
12563 * config/i386/i386-builtin.def: Add corrresponding new builtins.
12564 * config/i386/i386-expand.c: Handle new builtin types.
12565 * config/i386/sse.md (VF48_128): New mode iterator.
12566 (avx512fp16_vcvtsh2<ssescalarmodesuffix><mask_scalar_name><round_saeonly_scalar_name>):
12568 (avx512fp16_vcvt<ssescalarmodesuffix>2sh<mask_scalar_name><round_scalar_name>):
12571 2021-09-17 liuhongt <hongtao.liu@intel.com>
12573 * config/i386/avx512fp16intrin.h (_mm512_cvtph_pd):
12575 (_mm512_mask_cvtph_pd): Likewise.
12576 (_mm512_maskz_cvtph_pd): Likewise.
12577 (_mm512_cvt_roundph_pd): Likewise.
12578 (_mm512_mask_cvt_roundph_pd): Likewise.
12579 (_mm512_maskz_cvt_roundph_pd): Likewise.
12580 (_mm512_cvtxph_ps): Likewise.
12581 (_mm512_mask_cvtxph_ps): Likewise.
12582 (_mm512_maskz_cvtxph_ps): Likewise.
12583 (_mm512_cvtx_roundph_ps): Likewise.
12584 (_mm512_mask_cvtx_roundph_ps): Likewise.
12585 (_mm512_maskz_cvtx_roundph_ps): Likewise.
12586 (_mm512_cvtxps_ph): Likewise.
12587 (_mm512_mask_cvtxps_ph): Likewise.
12588 (_mm512_maskz_cvtxps_ph): Likewise.
12589 (_mm512_cvtx_roundps_ph): Likewise.
12590 (_mm512_mask_cvtx_roundps_ph): Likewise.
12591 (_mm512_maskz_cvtx_roundps_ph): Likewise.
12592 (_mm512_cvtpd_ph): Likewise.
12593 (_mm512_mask_cvtpd_ph): Likewise.
12594 (_mm512_maskz_cvtpd_ph): Likewise.
12595 (_mm512_cvt_roundpd_ph): Likewise.
12596 (_mm512_mask_cvt_roundpd_ph): Likewise.
12597 (_mm512_maskz_cvt_roundpd_ph): Likewise.
12598 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_pd):
12600 (_mm_mask_cvtph_pd): Likewise.
12601 (_mm_maskz_cvtph_pd): Likewise.
12602 (_mm256_cvtph_pd): Likewise.
12603 (_mm256_mask_cvtph_pd): Likewise.
12604 (_mm256_maskz_cvtph_pd): Likewise.
12605 (_mm_cvtxph_ps): Likewise.
12606 (_mm_mask_cvtxph_ps): Likewise.
12607 (_mm_maskz_cvtxph_ps): Likewise.
12608 (_mm256_cvtxph_ps): Likewise.
12609 (_mm256_mask_cvtxph_ps): Likewise.
12610 (_mm256_maskz_cvtxph_ps): Likewise.
12611 (_mm_cvtxps_ph): Likewise.
12612 (_mm_mask_cvtxps_ph): Likewise.
12613 (_mm_maskz_cvtxps_ph): Likewise.
12614 (_mm256_cvtxps_ph): Likewise.
12615 (_mm256_mask_cvtxps_ph): Likewise.
12616 (_mm256_maskz_cvtxps_ph): Likewise.
12617 (_mm_cvtpd_ph): Likewise.
12618 (_mm_mask_cvtpd_ph): Likewise.
12619 (_mm_maskz_cvtpd_ph): Likewise.
12620 (_mm256_cvtpd_ph): Likewise.
12621 (_mm256_mask_cvtpd_ph): Likewise.
12622 (_mm256_maskz_cvtpd_ph): Likewise.
12623 * config/i386/i386-builtin.def: Add corresponding new builtins.
12624 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
12625 * config/i386/i386-expand.c: Handle new builtin types.
12626 * config/i386/sse.md
12627 (VF4_128_8_256): New.
12628 (VF48H_AVX512VL): Ditto.
12629 (ssePHmode): Add HF vector modes.
12630 (castmode): Add new convertable modes.
12631 (qq2phsuff): Ditto.
12632 (ph2pssuffix): New.
12633 (avx512fp16_vcvt<castmode>2ph_<mode><mask_name><round_name>): Ditto.
12634 (avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
12635 (*avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
12636 (avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
12637 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
12638 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask_1): Ditto.
12639 (avx512fp16_float_extend_ph<mode>2<mask_name><round_saeonly_name>):
12641 (avx512fp16_float_extend_ph<mode>2<mask_name>): Ditto.
12642 (*avx512fp16_float_extend_ph<mode>2_load<mask_name>): Ditto.
12643 (avx512fp16_float_extend_phv2df2<mask_name>): Ditto.
12644 (*avx512fp16_float_extend_phv2df2_load<mask_name>): Ditto.
12646 2021-09-17 liuhongt <hongtao.liu@intel.com>
12648 * config/i386/avx512fp16intrin.h (_mm_cvttsh_i32):
12650 (_mm_cvttsh_u32): Likewise.
12651 (_mm_cvtt_roundsh_i32): Likewise.
12652 (_mm_cvtt_roundsh_u32): Likewise.
12653 (_mm_cvttsh_i64): Likewise.
12654 (_mm_cvttsh_u64): Likewise.
12655 (_mm_cvtt_roundsh_i64): Likewise.
12656 (_mm_cvtt_roundsh_u64): Likewise.
12657 * config/i386/i386-builtin.def: Add corresponding new builtins.
12658 * config/i386/sse.md
12659 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<round_saeonly_name>):
12662 2021-09-17 liuhongt <hongtao.liu@intel.com>
12664 * config/i386/avx512fp16intrin.h (_mm512_cvttph_epi32):
12666 (_mm512_mask_cvttph_epi32): Likewise.
12667 (_mm512_maskz_cvttph_epi32): Likewise.
12668 (_mm512_cvtt_roundph_epi32): Likewise.
12669 (_mm512_mask_cvtt_roundph_epi32): Likewise.
12670 (_mm512_maskz_cvtt_roundph_epi32): Likewise.
12671 (_mm512_cvttph_epu32): Likewise.
12672 (_mm512_mask_cvttph_epu32): Likewise.
12673 (_mm512_maskz_cvttph_epu32): Likewise.
12674 (_mm512_cvtt_roundph_epu32): Likewise.
12675 (_mm512_mask_cvtt_roundph_epu32): Likewise.
12676 (_mm512_maskz_cvtt_roundph_epu32): Likewise.
12677 (_mm512_cvttph_epi64): Likewise.
12678 (_mm512_mask_cvttph_epi64): Likewise.
12679 (_mm512_maskz_cvttph_epi64): Likewise.
12680 (_mm512_cvtt_roundph_epi64): Likewise.
12681 (_mm512_mask_cvtt_roundph_epi64): Likewise.
12682 (_mm512_maskz_cvtt_roundph_epi64): Likewise.
12683 (_mm512_cvttph_epu64): Likewise.
12684 (_mm512_mask_cvttph_epu64): Likewise.
12685 (_mm512_maskz_cvttph_epu64): Likewise.
12686 (_mm512_cvtt_roundph_epu64): Likewise.
12687 (_mm512_mask_cvtt_roundph_epu64): Likewise.
12688 (_mm512_maskz_cvtt_roundph_epu64): Likewise.
12689 (_mm512_cvttph_epi16): Likewise.
12690 (_mm512_mask_cvttph_epi16): Likewise.
12691 (_mm512_maskz_cvttph_epi16): Likewise.
12692 (_mm512_cvtt_roundph_epi16): Likewise.
12693 (_mm512_mask_cvtt_roundph_epi16): Likewise.
12694 (_mm512_maskz_cvtt_roundph_epi16): Likewise.
12695 (_mm512_cvttph_epu16): Likewise.
12696 (_mm512_mask_cvttph_epu16): Likewise.
12697 (_mm512_maskz_cvttph_epu16): Likewise.
12698 (_mm512_cvtt_roundph_epu16): Likewise.
12699 (_mm512_mask_cvtt_roundph_epu16): Likewise.
12700 (_mm512_maskz_cvtt_roundph_epu16): Likewise.
12701 * config/i386/avx512fp16vlintrin.h (_mm_cvttph_epi32):
12703 (_mm_mask_cvttph_epi32): Likewise.
12704 (_mm_maskz_cvttph_epi32): Likewise.
12705 (_mm256_cvttph_epi32): Likewise.
12706 (_mm256_mask_cvttph_epi32): Likewise.
12707 (_mm256_maskz_cvttph_epi32): Likewise.
12708 (_mm_cvttph_epu32): Likewise.
12709 (_mm_mask_cvttph_epu32): Likewise.
12710 (_mm_maskz_cvttph_epu32): Likewise.
12711 (_mm256_cvttph_epu32): Likewise.
12712 (_mm256_mask_cvttph_epu32): Likewise.
12713 (_mm256_maskz_cvttph_epu32): Likewise.
12714 (_mm_cvttph_epi64): Likewise.
12715 (_mm_mask_cvttph_epi64): Likewise.
12716 (_mm_maskz_cvttph_epi64): Likewise.
12717 (_mm256_cvttph_epi64): Likewise.
12718 (_mm256_mask_cvttph_epi64): Likewise.
12719 (_mm256_maskz_cvttph_epi64): Likewise.
12720 (_mm_cvttph_epu64): Likewise.
12721 (_mm_mask_cvttph_epu64): Likewise.
12722 (_mm_maskz_cvttph_epu64): Likewise.
12723 (_mm256_cvttph_epu64): Likewise.
12724 (_mm256_mask_cvttph_epu64): Likewise.
12725 (_mm256_maskz_cvttph_epu64): Likewise.
12726 (_mm_cvttph_epi16): Likewise.
12727 (_mm_mask_cvttph_epi16): Likewise.
12728 (_mm_maskz_cvttph_epi16): Likewise.
12729 (_mm256_cvttph_epi16): Likewise.
12730 (_mm256_mask_cvttph_epi16): Likewise.
12731 (_mm256_maskz_cvttph_epi16): Likewise.
12732 (_mm_cvttph_epu16): Likewise.
12733 (_mm_mask_cvttph_epu16): Likewise.
12734 (_mm_maskz_cvttph_epu16): Likewise.
12735 (_mm256_cvttph_epu16): Likewise.
12736 (_mm256_mask_cvttph_epu16): Likewise.
12737 (_mm256_maskz_cvttph_epu16): Likewise.
12738 * config/i386/i386-builtin.def: Add new builtins.
12739 * config/i386/sse.md
12740 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name><round_saeonly_name>):
12742 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name>): Ditto.
12743 (*avx512fp16_fix<fixunssuffix>_trunc<mode>2_load<mask_name>): Ditto.
12744 (avx512fp16_fix<fixunssuffix>_truncv2di2<mask_name>): Ditto.
12745 (avx512fp16_fix<fixunssuffix>_truncv2di2_load<mask_name>): Ditto.
12747 2021-09-17 liuhongt <hongtao.liu@intel.com>
12749 * config/i386/avx512fp16intrin.h (_mm_cvtsh_i32): New intrinsic.
12750 (_mm_cvtsh_u32): Likewise.
12751 (_mm_cvt_roundsh_i32): Likewise.
12752 (_mm_cvt_roundsh_u32): Likewise.
12753 (_mm_cvtsh_i64): Likewise.
12754 (_mm_cvtsh_u64): Likewise.
12755 (_mm_cvt_roundsh_i64): Likewise.
12756 (_mm_cvt_roundsh_u64): Likewise.
12757 (_mm_cvti32_sh): Likewise.
12758 (_mm_cvtu32_sh): Likewise.
12759 (_mm_cvt_roundi32_sh): Likewise.
12760 (_mm_cvt_roundu32_sh): Likewise.
12761 (_mm_cvti64_sh): Likewise.
12762 (_mm_cvtu64_sh): Likewise.
12763 (_mm_cvt_roundi64_sh): Likewise.
12764 (_mm_cvt_roundu64_sh): Likewise.
12765 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
12766 * config/i386/i386-builtin.def: Add corresponding new builtins.
12767 * config/i386/i386-expand.c (ix86_expand_round_builtin):
12768 Handle new builtin types.
12769 * config/i386/sse.md
12770 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix><round_name>):
12772 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix>_2): Likewise.
12773 (avx512fp16_vcvt<floatsuffix>si2sh<rex64namesuffix><round_name>): Likewise.
12775 2021-09-16 Bill Schmidt <wschmidt@linux.ibm.com>
12777 * config/rs6000/rs6000-c.c (rs6000-builtins.h): New include.
12778 (altivec_resolve_new_overloaded_builtin): New forward decl.
12779 (rs6000_new_builtin_type_compatible): New function.
12780 (altivec_resolve_overloaded_builtin): Call
12781 altivec_resolve_new_overloaded_builtin.
12782 (altivec_build_new_resolved_builtin): New function.
12783 (altivec_resolve_new_overloaded_builtin): Likewise.
12784 * config/rs6000/rs6000-call.c (rs6000_new_builtin_is_supported):
12786 * config/rs6000/rs6000-gen-builtins.c (write_decls): Remove _p from
12787 name of rs6000_new_builtin_is_supported.
12789 2021-09-16 Uroš Bizjak <ubizjak@gmail.com>
12791 * config/i386/i386-protos.h (ix86_decompose_address):
12792 Change return type to bool.
12793 * config/i386/i386.c (ix86_decompose_address): Ditto.
12795 2021-09-16 Tobias Burnus <tobias@codesourcery.com>
12798 * config/rs6000/t-rs6000 (build/rs6000-gen-builtins.o, build/rbtree.o):
12799 Added 'build/' to target, use build/%.o rule.
12800 (build/rs6000-gen-builtins$(build_exeext)): Add 'build/' and
12801 '$(build_exeext)' to target and 'build/' for the *.o files.
12802 (rs6000-builtins.c): Update for those changes; run rs6000-gen-builtins
12805 2021-09-16 Martin Jambor <mjambor@suse.cz>
12807 * cgraph.c (cgraph_node::dump): Do not check caller count sums if
12808 the body has been removed. Remove trailing whitespace.
12810 2021-09-16 Richard Biener <rguenther@suse.de>
12812 PR middle-end/102360
12813 * internal-fn.c (expand_DEFERRED_INIT): Make pattern-init
12814 of non-memory more robust.
12816 2021-09-16 Daniel Cederman <cederman@gaisler.com>
12818 * config/sparc/sparc-opts.h (enum sparc_processor_type): Add LEON5
12819 * config/sparc/sparc.c (struct processor_costs): Add LEON5 costs
12820 (leon5_adjust_cost): Increase cost of store with data dependency
12821 on ALU instruction and FPU anti-dependencies.
12822 (sparc_option_override): Add LEON5 costs
12823 (sparc_adjust_cost): Add LEON5 cost adjustments
12824 * config/sparc/sparc.h: Add LEON5
12825 * config/sparc/sparc.md: Include LEON5 scheduling information
12826 * config/sparc/sparc.opt: Add LEON5
12827 * doc/invoke.texi: Add LEON5
12828 * config/sparc/leon5.md: New file.
12830 2021-09-16 Daniel Cederman <cederman@gaisler.com>
12832 * config/sparc/sparc.md (stack_protect_set32): Add NOP to prevent
12833 sensitive sequence for B2BST errata workaround.
12835 2021-09-16 Daniel Cederman <cederman@gaisler.com>
12837 * config/sparc/sparc.c (sparc_do_work_around_errata): Do not begin
12838 functions with atomic instruction in the UT700 errata workaround.
12840 2021-09-16 Daniel Cederman <cederman@gaisler.com>
12842 * config/sparc/sparc.c (next_active_non_empty_insn): New function
12843 that returns next active non empty assembly instruction.
12844 (sparc_do_work_around_errata): Use new function.
12846 2021-09-16 Daniel Cederman <cederman@gaisler.com>
12848 * config/sparc/sparc.c (store_insn_p): Add predicate for store
12850 (load_insn_p): Add predicate for load attributes.
12851 (sparc_do_work_around_errata): Use new predicates.
12853 2021-09-16 Andreas Larsson <andreas@gaisler.com>
12855 * config/sparc/sparc.c (dump_target_flag_bits): Print bit names for
12858 2021-09-16 Martin Liska <mliska@suse.cz>
12860 * config/mips/netbsd.h: Fix typo in name of a macro.
12862 2021-09-16 liuhongt <hongtao.liu@intel.com>
12864 PR middle-end/102080
12865 * match.pd: Check mask type when doing cond_op related gimple
12867 * tree.c (is_truth_type_for): New function.
12868 * tree.h (is_truth_type_for): New declaration.
12870 2021-09-16 liuhongt <hongtao.liu@intel.com>
12872 * config/i386/avx512fp16intrin.h (_mm512_cvtepi32_ph): New
12874 (_mm512_mask_cvtepi32_ph): Likewise.
12875 (_mm512_maskz_cvtepi32_ph): Likewise.
12876 (_mm512_cvt_roundepi32_ph): Likewise.
12877 (_mm512_mask_cvt_roundepi32_ph): Likewise.
12878 (_mm512_maskz_cvt_roundepi32_ph): Likewise.
12879 (_mm512_cvtepu32_ph): Likewise.
12880 (_mm512_mask_cvtepu32_ph): Likewise.
12881 (_mm512_maskz_cvtepu32_ph): Likewise.
12882 (_mm512_cvt_roundepu32_ph): Likewise.
12883 (_mm512_mask_cvt_roundepu32_ph): Likewise.
12884 (_mm512_maskz_cvt_roundepu32_ph): Likewise.
12885 (_mm512_cvtepi64_ph): Likewise.
12886 (_mm512_mask_cvtepi64_ph): Likewise.
12887 (_mm512_maskz_cvtepi64_ph): Likewise.
12888 (_mm512_cvt_roundepi64_ph): Likewise.
12889 (_mm512_mask_cvt_roundepi64_ph): Likewise.
12890 (_mm512_maskz_cvt_roundepi64_ph): Likewise.
12891 (_mm512_cvtepu64_ph): Likewise.
12892 (_mm512_mask_cvtepu64_ph): Likewise.
12893 (_mm512_maskz_cvtepu64_ph): Likewise.
12894 (_mm512_cvt_roundepu64_ph): Likewise.
12895 (_mm512_mask_cvt_roundepu64_ph): Likewise.
12896 (_mm512_maskz_cvt_roundepu64_ph): Likewise.
12897 (_mm512_cvtepi16_ph): Likewise.
12898 (_mm512_mask_cvtepi16_ph): Likewise.
12899 (_mm512_maskz_cvtepi16_ph): Likewise.
12900 (_mm512_cvt_roundepi16_ph): Likewise.
12901 (_mm512_mask_cvt_roundepi16_ph): Likewise.
12902 (_mm512_maskz_cvt_roundepi16_ph): Likewise.
12903 (_mm512_cvtepu16_ph): Likewise.
12904 (_mm512_mask_cvtepu16_ph): Likewise.
12905 (_mm512_maskz_cvtepu16_ph): Likewise.
12906 (_mm512_cvt_roundepu16_ph): Likewise.
12907 (_mm512_mask_cvt_roundepu16_ph): Likewise.
12908 (_mm512_maskz_cvt_roundepu16_ph): Likewise.
12909 * config/i386/avx512fp16vlintrin.h (_mm_cvtepi32_ph): New
12911 (_mm_mask_cvtepi32_ph): Likewise.
12912 (_mm_maskz_cvtepi32_ph): Likewise.
12913 (_mm256_cvtepi32_ph): Likewise.
12914 (_mm256_mask_cvtepi32_ph): Likewise.
12915 (_mm256_maskz_cvtepi32_ph): Likewise.
12916 (_mm_cvtepu32_ph): Likewise.
12917 (_mm_mask_cvtepu32_ph): Likewise.
12918 (_mm_maskz_cvtepu32_ph): Likewise.
12919 (_mm256_cvtepu32_ph): Likewise.
12920 (_mm256_mask_cvtepu32_ph): Likewise.
12921 (_mm256_maskz_cvtepu32_ph): Likewise.
12922 (_mm_cvtepi64_ph): Likewise.
12923 (_mm_mask_cvtepi64_ph): Likewise.
12924 (_mm_maskz_cvtepi64_ph): Likewise.
12925 (_mm256_cvtepi64_ph): Likewise.
12926 (_mm256_mask_cvtepi64_ph): Likewise.
12927 (_mm256_maskz_cvtepi64_ph): Likewise.
12928 (_mm_cvtepu64_ph): Likewise.
12929 (_mm_mask_cvtepu64_ph): Likewise.
12930 (_mm_maskz_cvtepu64_ph): Likewise.
12931 (_mm256_cvtepu64_ph): Likewise.
12932 (_mm256_mask_cvtepu64_ph): Likewise.
12933 (_mm256_maskz_cvtepu64_ph): Likewise.
12934 (_mm_cvtepi16_ph): Likewise.
12935 (_mm_mask_cvtepi16_ph): Likewise.
12936 (_mm_maskz_cvtepi16_ph): Likewise.
12937 (_mm256_cvtepi16_ph): Likewise.
12938 (_mm256_mask_cvtepi16_ph): Likewise.
12939 (_mm256_maskz_cvtepi16_ph): Likewise.
12940 (_mm_cvtepu16_ph): Likewise.
12941 (_mm_mask_cvtepu16_ph): Likewise.
12942 (_mm_maskz_cvtepu16_ph): Likewise.
12943 (_mm256_cvtepu16_ph): Likewise.
12944 (_mm256_mask_cvtepu16_ph): Likewise.
12945 (_mm256_maskz_cvtepu16_ph): Likewise.
12946 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
12947 * config/i386/i386-builtin.def: Add corresponding new builtins.
12948 * config/i386/i386-expand.c
12949 (ix86_expand_args_builtin): Handle new builtin types.
12950 (ix86_expand_round_builtin): Ditto.
12951 * config/i386/i386-modes.def: Declare V2HF and V6HF.
12952 * config/i386/sse.md (VI2H_AVX512VL): New.
12953 (qq2phsuff): Ditto.
12954 (sseintvecmode): Add HF vector modes.
12955 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode><mask_name><round_name>):
12957 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
12958 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
12959 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
12960 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
12961 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask_1): Ditto.
12962 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
12963 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
12964 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
12965 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
12966 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask_1): Ditto.
12967 * config/i386/subst.md (round_qq2phsuff): New subst_attr.
12969 2021-09-16 liuhongt <hongtao.liu@intel.com>
12971 * config/i386/avx512fp16intrin.h (_mm512_cvtph_epi32):
12973 (_mm512_mask_cvtph_epi32): Likewise.
12974 (_mm512_maskz_cvtph_epi32): Likewise.
12975 (_mm512_cvt_roundph_epi32): Likewise.
12976 (_mm512_mask_cvt_roundph_epi32): Likewise.
12977 (_mm512_maskz_cvt_roundph_epi32): Likewise.
12978 (_mm512_cvtph_epu32): Likewise.
12979 (_mm512_mask_cvtph_epu32): Likewise.
12980 (_mm512_maskz_cvtph_epu32): Likewise.
12981 (_mm512_cvt_roundph_epu32): Likewise.
12982 (_mm512_mask_cvt_roundph_epu32): Likewise.
12983 (_mm512_maskz_cvt_roundph_epu32): Likewise.
12984 (_mm512_cvtph_epi64): Likewise.
12985 (_mm512_mask_cvtph_epi64): Likewise.
12986 (_mm512_maskz_cvtph_epi64): Likewise.
12987 (_mm512_cvt_roundph_epi64): Likewise.
12988 (_mm512_mask_cvt_roundph_epi64): Likewise.
12989 (_mm512_maskz_cvt_roundph_epi64): Likewise.
12990 (_mm512_cvtph_epu64): Likewise.
12991 (_mm512_mask_cvtph_epu64): Likewise.
12992 (_mm512_maskz_cvtph_epu64): Likewise.
12993 (_mm512_cvt_roundph_epu64): Likewise.
12994 (_mm512_mask_cvt_roundph_epu64): Likewise.
12995 (_mm512_maskz_cvt_roundph_epu64): Likewise.
12996 (_mm512_cvtph_epi16): Likewise.
12997 (_mm512_mask_cvtph_epi16): Likewise.
12998 (_mm512_maskz_cvtph_epi16): Likewise.
12999 (_mm512_cvt_roundph_epi16): Likewise.
13000 (_mm512_mask_cvt_roundph_epi16): Likewise.
13001 (_mm512_maskz_cvt_roundph_epi16): Likewise.
13002 (_mm512_cvtph_epu16): Likewise.
13003 (_mm512_mask_cvtph_epu16): Likewise.
13004 (_mm512_maskz_cvtph_epu16): Likewise.
13005 (_mm512_cvt_roundph_epu16): Likewise.
13006 (_mm512_mask_cvt_roundph_epu16): Likewise.
13007 (_mm512_maskz_cvt_roundph_epu16): Likewise.
13008 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_epi32):
13010 (_mm_mask_cvtph_epi32): Likewise.
13011 (_mm_maskz_cvtph_epi32): Likewise.
13012 (_mm256_cvtph_epi32): Likewise.
13013 (_mm256_mask_cvtph_epi32): Likewise.
13014 (_mm256_maskz_cvtph_epi32): Likewise.
13015 (_mm_cvtph_epu32): Likewise.
13016 (_mm_mask_cvtph_epu32): Likewise.
13017 (_mm_maskz_cvtph_epu32): Likewise.
13018 (_mm256_cvtph_epu32): Likewise.
13019 (_mm256_mask_cvtph_epu32): Likewise.
13020 (_mm256_maskz_cvtph_epu32): Likewise.
13021 (_mm_cvtph_epi64): Likewise.
13022 (_mm_mask_cvtph_epi64): Likewise.
13023 (_mm_maskz_cvtph_epi64): Likewise.
13024 (_mm256_cvtph_epi64): Likewise.
13025 (_mm256_mask_cvtph_epi64): Likewise.
13026 (_mm256_maskz_cvtph_epi64): Likewise.
13027 (_mm_cvtph_epu64): Likewise.
13028 (_mm_mask_cvtph_epu64): Likewise.
13029 (_mm_maskz_cvtph_epu64): Likewise.
13030 (_mm256_cvtph_epu64): Likewise.
13031 (_mm256_mask_cvtph_epu64): Likewise.
13032 (_mm256_maskz_cvtph_epu64): Likewise.
13033 (_mm_cvtph_epi16): Likewise.
13034 (_mm_mask_cvtph_epi16): Likewise.
13035 (_mm_maskz_cvtph_epi16): Likewise.
13036 (_mm256_cvtph_epi16): Likewise.
13037 (_mm256_mask_cvtph_epi16): Likewise.
13038 (_mm256_maskz_cvtph_epi16): Likewise.
13039 (_mm_cvtph_epu16): Likewise.
13040 (_mm_mask_cvtph_epu16): Likewise.
13041 (_mm_maskz_cvtph_epu16): Likewise.
13042 (_mm256_cvtph_epu16): Likewise.
13043 (_mm256_mask_cvtph_epu16): Likewise.
13044 (_mm256_maskz_cvtph_epu16): Likewise.
13045 * config/i386/i386-builtin-types.def: Add new builtin types.
13046 * config/i386/i386-builtin.def: Add new builtins.
13047 * config/i386/i386-expand.c
13048 (ix86_expand_args_builtin): Handle new builtin types.
13049 (ix86_expand_round_builtin): Ditto.
13050 * config/i386/sse.md (sseintconvert): New.
13051 (ssePHmode): Ditto.
13052 (UNSPEC_US_FIX_NOTRUNC): Ditto.
13053 (sseintconvertsignprefix): Ditto.
13054 (avx512fp16_vcvtph2<sseintconvertsignprefix><sseintconvert>_<mode><mask_name><round_name>):
13057 2021-09-16 liuhongt <hongtao.liu@intel.com>
13059 * config/i386/avx512fp16intrin.h: (_mm_cvtsi16_si128):
13061 (_mm_cvtsi128_si16): Likewise.
13062 (_mm_mask_load_sh): Likewise.
13063 (_mm_maskz_load_sh): Likewise.
13064 (_mm_mask_store_sh): Likewise.
13065 (_mm_move_sh): Likewise.
13066 (_mm_mask_move_sh): Likewise.
13067 (_mm_maskz_move_sh): Likewise.
13068 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
13069 * config/i386/i386-builtin.def: Add corresponding new builtins.
13070 * config/i386/i386-expand.c
13071 (ix86_expand_special_args_builtin): Handle new builtin types.
13072 (ix86_expand_vector_init_one_nonzero): Adjust for FP16 target.
13073 * config/i386/sse.md (VI2F): New mode iterator.
13074 (vec_set<mode>_0): Use new mode iterator.
13075 (avx512f_mov<ssescalarmodelower>_mask): Adjust for HF vector mode.
13076 (avx512f_store<mode>_mask): Ditto.
13078 2021-09-16 Kewen Lin <linkw@linux.ibm.com>
13080 * config/rs6000/rs6000.opt (-mtoc-fusion): Remove.
13082 2021-09-15 David Edelsohn <dje.gcc@gmail.com>
13084 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
13085 Proceed if no symbol summary or the symbol alias flag is false.
13087 2021-09-15 Jakub Jelinek <jakub@redhat.com>
13091 * varasm.c (output_constructor_regular_field): Instead of assertion
13092 that array_size_for_constructor result is equal to size of
13093 TREE_TYPE (local->val) in bytes, assert that the type size is greater
13094 or equal to array_size_for_constructor result and use type size as
13097 2021-09-15 Martin Liska <mliska@suse.cz>
13100 * config/i386/vxworks.h: Use new macro TARGET_CPU_P.
13102 2021-09-15 Martin Liska <mliska@suse.cz>
13105 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
13106 Check that we have a symbol summary for a symbol.
13108 2021-09-15 Richard Biener <rguenther@suse.de>
13111 * config/rs6000/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
13112 to inherit from elfos.h
13114 2021-09-15 liuhongt <hongtao.liu@intel.com>
13117 * config/i386/i386-expand.c
13118 (ix86_expand_vector_init_interleave): Use puncklwd to pack 2
13120 (ix86_expand_vector_set): Use blendw instead of pinsrw.
13121 * config/i386/i386.c (ix86_can_change_mode_class): Adjust for
13122 AVX512FP16 which supports 16bit vector load.
13123 * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>):
13125 (avx512bw_interleave_high<mode><mask_name>): .. this, and
13126 extend to V32HFmode.
13127 (avx2_interleave_highv16hi<mask_name>): Rename to ..
13128 (avx2_interleave_high<mode><mask_name>): .. this, and extend
13130 (vec_interleave_highv8hi<mask_name>): Rename to ..
13131 (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode.
13132 (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>):
13134 (<mask_codefor>avx512bw_interleave_low<mode><mask_name>):
13135 this, and extend to V32HFmode.
13136 (avx2_interleave_lowv16hi<mask_name>): Rename to ..
13137 (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode.
13138 (vec_interleave_lowv8hi<mask_name>): Rename to ..
13139 (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode.
13140 (sse4_1_pblendw): Rename to ..
13141 (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode.
13142 (avx2_pblendph): New define_expand.
13143 (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use
13144 sseintmodesuffix instead of ssemodesuffix.
13145 (blendsuf): New mode attr.
13147 2021-09-15 Richard Biener <rguenther@suse.de>
13149 * tree-vectorizer.h (dr_misalignment): Move out of line.
13150 (dr_target_alignment): New.
13151 (DR_TARGET_ALIGNMENT): Wrap dr_target_alignment.
13152 (set_dr_target_alignment): New.
13153 (SET_DR_TARGET_ALIGNMENT): Wrap set_dr_target_alignment.
13154 * tree-vect-data-refs.c (dr_misalignment): Compute and
13155 return the group members misalignment.
13156 (vect_compute_data_ref_alignment): Use SET_DR_TARGET_ALIGNMENT.
13157 (vect_analyze_data_refs_alignment): Compute alignment only
13158 for the first element of a DR group.
13159 (vect_slp_analyze_node_alignment): Likewise.
13161 2021-09-15 Hongyu Wang <hongyu.wang@intel.com>
13163 * config/i386/avx512fp16intrin.h: Adjust all builtin calls.
13164 * config/i386/avx512fp16vlintrin.h: Likewise.
13165 * config/i386/i386-builtin.def: Adjust builtin name and
13166 enumeration to match AVX512F style.
13168 2021-09-15 Richard Biener <rguenther@suse.de>
13170 PR tree-optimization/102318
13171 * tree-vect-loop.c (vect_transform_cycle_phi): Revert
13172 previous change and do the mode conversion separately from
13173 the sign conversion.
13175 2021-09-15 Hongtao Liu <hongtao.liu@intel.com>
13176 Peter Cordes <peter@cordes.ca>
13179 * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
13180 (*vec_extract<mode><ssescalarmodelower>_valign): Output
13181 vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
13184 2021-09-15 Richard Biener <rguenther@suse.de>
13186 * config.gcc: Remove vax-*-openbsd* configuration.
13188 2021-09-15 Richard Biener <rguenther@suse.de>
13190 * config.gcc: Remove m68k-openbsd.
13192 2021-09-15 Max Filippov <jcmvbkbc@gmail.com>
13195 * config/xtensa/t-xtensa (TM_H): Add include/xtensa-config.h.
13197 2021-09-14 Peter Bergner <bergner@linux.ibm.com>
13199 * config/rs6000/mma.md (unspec): Delete UNSPEC_MMA_XXSETACCZ.
13200 (unspecv): Add UNSPECV_MMA_XXSETACCZ.
13201 (*mma_xxsetaccz): Delete.
13202 (mma_xxsetaccz): Change to define_insn. Remove operand 1.
13203 Use UNSPECV_MMA_XXSETACCZ. Update comment.
13204 * config/rs6000/rs6000.c (rs6000_rtx_costs): Use UNSPECV_MMA_XXSETACCZ.
13206 2021-09-14 Iain Sandoe <iain@sandoe.co.uk>
13208 * Makefile.in: Remove variables related to applying no-PIE
13209 to the exes on $build.
13210 * configure: Regenerate.
13211 * configure.ac: Remove configuration related to applying
13212 no-PIE to the exes on $build.
13214 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
13216 * config/arc/arc.md (doloop_end): Add missing mode.
13217 (loop_end): Likewise.
13219 2021-09-14 Jakub Jelinek <jakub@redhat.com>
13221 * gimplify.c (goa_stabilize_expr): Add depth argument, propagate
13222 it to recursive calls, for depth above 7 just gimplify or return.
13223 Perform a test even for MODIFY_EXPR, ADDR_EXPR, COMPOUND_EXPR with
13224 __builtin_clear_padding and TARGET_EXPR.
13225 (gimplify_omp_atomic): Adjust goa_stabilize_expr callers.
13227 2021-09-14 liuhongt <hongtao.liu@intel.com>
13229 * config/i386/avx512fp16intrin.h (_mm_fpclass_sh_mask):
13231 (_mm_mask_fpclass_sh_mask): Likewise.
13232 (_mm512_mask_fpclass_ph_mask): Likewise.
13233 (_mm512_fpclass_ph_mask): Likewise.
13234 (_mm_getexp_sh): Likewise.
13235 (_mm_mask_getexp_sh): Likewise.
13236 (_mm_maskz_getexp_sh): Likewise.
13237 (_mm512_getexp_ph): Likewise.
13238 (_mm512_mask_getexp_ph): Likewise.
13239 (_mm512_maskz_getexp_ph): Likewise.
13240 (_mm_getexp_round_sh): Likewise.
13241 (_mm_mask_getexp_round_sh): Likewise.
13242 (_mm_maskz_getexp_round_sh): Likewise.
13243 (_mm512_getexp_round_ph): Likewise.
13244 (_mm512_mask_getexp_round_ph): Likewise.
13245 (_mm512_maskz_getexp_round_ph): Likewise.
13246 (_mm_getmant_sh): Likewise.
13247 (_mm_mask_getmant_sh): Likewise.
13248 (_mm_maskz_getmant_sh): Likewise.
13249 (_mm512_getmant_ph): Likewise.
13250 (_mm512_mask_getmant_ph): Likewise.
13251 (_mm512_maskz_getmant_ph): Likewise.
13252 (_mm_getmant_round_sh): Likewise.
13253 (_mm_mask_getmant_round_sh): Likewise.
13254 (_mm_maskz_getmant_round_sh): Likewise.
13255 (_mm512_getmant_round_ph): Likewise.
13256 (_mm512_mask_getmant_round_ph): Likewise.
13257 (_mm512_maskz_getmant_round_ph): Likewise.
13258 * config/i386/avx512fp16vlintrin.h (_mm_mask_fpclass_ph_mask):
13260 (_mm_fpclass_ph_mask): Likewise.
13261 (_mm256_mask_fpclass_ph_mask): Likewise.
13262 (_mm256_fpclass_ph_mask): Likewise.
13263 (_mm256_getexp_ph): Likewise.
13264 (_mm256_mask_getexp_ph): Likewise.
13265 (_mm256_maskz_getexp_ph): Likewise.
13266 (_mm_getexp_ph): Likewise.
13267 (_mm_mask_getexp_ph): Likewise.
13268 (_mm_maskz_getexp_ph): Likewise.
13269 (_mm256_getmant_ph): Likewise.
13270 (_mm256_mask_getmant_ph): Likewise.
13271 (_mm256_maskz_getmant_ph): Likewise.
13272 (_mm_getmant_ph): Likewise.
13273 (_mm_mask_getmant_ph): Likewise.
13274 (_mm_maskz_getmant_ph): Likewise.
13275 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
13276 * config/i386/i386-builtin.def: Add corresponding new builtins.
13277 * config/i386/i386-expand.c
13278 (ix86_expand_args_builtin): Handle new builtin types.
13279 (ix86_expand_round_builtin): Ditto.
13280 * config/i386/sse.md (vecmemsuffix): Add HF vector modes.
13281 (<avx512>_getexp<mode><mask_name><round_saeonly_name>): Adjust
13282 to support HF vector modes.
13283 (avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name):
13285 (avx512dq_fpclass<mode><mask_scalar_merge_name>): Ditto.
13286 (avx512dq_vmfpclass<mode><mask_scalar_merge_name>): Ditto.
13287 (<avx512>_getmant<mode><mask_name><round_saeonly_name>): Ditto.
13288 (avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
13291 2021-09-14 liuhongt <hongtao.liu@intel.com>
13293 * config/i386/avx512fp16intrin.h (_mm512_reduce_ph):
13295 (_mm512_mask_reduce_ph): Likewise.
13296 (_mm512_maskz_reduce_ph): Likewise.
13297 (_mm512_reduce_round_ph): Likewise.
13298 (_mm512_mask_reduce_round_ph): Likewise.
13299 (_mm512_maskz_reduce_round_ph): Likewise.
13300 (_mm_reduce_sh): Likewise.
13301 (_mm_mask_reduce_sh): Likewise.
13302 (_mm_maskz_reduce_sh): Likewise.
13303 (_mm_reduce_round_sh): Likewise.
13304 (_mm_mask_reduce_round_sh): Likewise.
13305 (_mm_maskz_reduce_round_sh): Likewise.
13306 (_mm512_roundscale_ph): Likewise.
13307 (_mm512_mask_roundscale_ph): Likewise.
13308 (_mm512_maskz_roundscale_ph): Likewise.
13309 (_mm512_roundscale_round_ph): Likewise.
13310 (_mm512_mask_roundscale_round_ph): Likewise.
13311 (_mm512_maskz_roundscale_round_ph): Likewise.
13312 (_mm_roundscale_sh): Likewise.
13313 (_mm_mask_roundscale_sh): Likewise.
13314 (_mm_maskz_roundscale_sh): Likewise.
13315 (_mm_roundscale_round_sh): Likewise.
13316 (_mm_mask_roundscale_round_sh): Likewise.
13317 (_mm_maskz_roundscale_round_sh): Likewise.
13318 * config/i386/avx512fp16vlintrin.h: (_mm_reduce_ph):
13320 (_mm_mask_reduce_ph): Likewise.
13321 (_mm_maskz_reduce_ph): Likewise.
13322 (_mm256_reduce_ph): Likewise.
13323 (_mm256_mask_reduce_ph): Likewise.
13324 (_mm256_maskz_reduce_ph): Likewise.
13325 (_mm_roundscale_ph): Likewise.
13326 (_mm_mask_roundscale_ph): Likewise.
13327 (_mm_maskz_roundscale_ph): Likewise.
13328 (_mm256_roundscale_ph): Likewise.
13329 (_mm256_mask_roundscale_ph): Likewise.
13330 (_mm256_maskz_roundscale_ph): Likewise.
13331 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
13332 * config/i386/i386-builtin.def: Add corresponding new builtins.
13333 * config/i386/i386-expand.c
13334 (ix86_expand_args_builtin): Handle new builtin types.
13335 (ix86_expand_round_builtin): Ditto.
13336 * config/i386/sse.md (<mask_codefor>reducep<mode><mask_name>):
13338 (<mask_codefor>reducep<mode><mask_name><round_saeonly_name>):
13339 ... this, and adjust for round operands.
13340 (reduces<mode><mask_scalar_name>): Likewise, with ...
13341 (reduces<mode><mask_scalar_name><round_saeonly_scalar_name):
13343 (<avx512>_rndscale<mode><mask_name><round_saeonly_name>):
13344 Adjust for HF vector modes.
13345 (avx512f_rndscale<mode><mask_scalar_name><round_saeonly_scalar_name>):
13347 (*avx512f_rndscale<mode><round_saeonly_name>): Ditto.
13349 2021-09-14 liuhongt <hongtao.liu@intel.com>
13351 * config/i386/avx512fp16intrin.h: (_mm512_rcp_ph):
13353 (_mm512_mask_rcp_ph): Likewise.
13354 (_mm512_maskz_rcp_ph): Likewise.
13355 (_mm_rcp_sh): Likewise.
13356 (_mm_mask_rcp_sh): Likewise.
13357 (_mm_maskz_rcp_sh): Likewise.
13358 (_mm512_scalef_ph): Likewise.
13359 (_mm512_mask_scalef_ph): Likewise.
13360 (_mm512_maskz_scalef_ph): Likewise.
13361 (_mm512_scalef_round_ph): Likewise.
13362 (_mm512_mask_scalef_round_ph): Likewise.
13363 (_mm512_maskz_scalef_round_ph): Likewise.
13364 (_mm_scalef_sh): Likewise.
13365 (_mm_mask_scalef_sh): Likewise.
13366 (_mm_maskz_scalef_sh): Likewise.
13367 (_mm_scalef_round_sh): Likewise.
13368 (_mm_mask_scalef_round_sh): Likewise.
13369 (_mm_maskz_scalef_round_sh): Likewise.
13370 * config/i386/avx512fp16vlintrin.h (_mm_rcp_ph):
13372 (_mm256_rcp_ph): Likewise.
13373 (_mm_mask_rcp_ph): Likewise.
13374 (_mm256_mask_rcp_ph): Likewise.
13375 (_mm_maskz_rcp_ph): Likewise.
13376 (_mm256_maskz_rcp_ph): Likewise.
13377 (_mm_scalef_ph): Likewise.
13378 (_mm256_scalef_ph): Likewise.
13379 (_mm_mask_scalef_ph): Likewise.
13380 (_mm256_mask_scalef_ph): Likewise.
13381 (_mm_maskz_scalef_ph): Likewise.
13382 (_mm256_maskz_scalef_ph): Likewise.
13383 * config/i386/i386-builtin.def: Add new builtins.
13384 * config/i386/sse.md (VFH_AVX512VL): New.
13385 (avx512fp16_rcp<mode>2<mask_name>): Ditto.
13386 (avx512fp16_vmrcpv8hf2<mask_scalar_name>): Ditto.
13387 (avx512f_vmscalef<mode><mask_scalar_name><round_scalar_name>):
13388 Adjust to support HF vector modes.
13389 (<avx512>_scalef<mode><mask_name><round_name>): Ditto.
13391 2021-09-14 liuhongt <hongtao.liu@intel.com>
13393 * config/i386/avx512fp16intrin.h: (_mm512_sqrt_ph):
13395 (_mm512_mask_sqrt_ph): Likewise.
13396 (_mm512_maskz_sqrt_ph): Likewise.
13397 (_mm512_sqrt_round_ph): Likewise.
13398 (_mm512_mask_sqrt_round_ph): Likewise.
13399 (_mm512_maskz_sqrt_round_ph): Likewise.
13400 (_mm512_rsqrt_ph): Likewise.
13401 (_mm512_mask_rsqrt_ph): Likewise.
13402 (_mm512_maskz_rsqrt_ph): Likewise.
13403 (_mm_rsqrt_sh): Likewise.
13404 (_mm_mask_rsqrt_sh): Likewise.
13405 (_mm_maskz_rsqrt_sh): Likewise.
13406 (_mm_sqrt_sh): Likewise.
13407 (_mm_mask_sqrt_sh): Likewise.
13408 (_mm_maskz_sqrt_sh): Likewise.
13409 (_mm_sqrt_round_sh): Likewise.
13410 (_mm_mask_sqrt_round_sh): Likewise.
13411 (_mm_maskz_sqrt_round_sh): Likewise.
13412 * config/i386/avx512fp16vlintrin.h (_mm_sqrt_ph): New intrinsic.
13413 (_mm256_sqrt_ph): Likewise.
13414 (_mm_mask_sqrt_ph): Likewise.
13415 (_mm256_mask_sqrt_ph): Likewise.
13416 (_mm_maskz_sqrt_ph): Likewise.
13417 (_mm256_maskz_sqrt_ph): Likewise.
13418 (_mm_rsqrt_ph): Likewise.
13419 (_mm256_rsqrt_ph): Likewise.
13420 (_mm_mask_rsqrt_ph): Likewise.
13421 (_mm256_mask_rsqrt_ph): Likewise.
13422 (_mm_maskz_rsqrt_ph): Likewise.
13423 (_mm256_maskz_rsqrt_ph): Likewise.
13424 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
13425 * config/i386/i386-builtin.def: Add corresponding new builtins.
13426 * config/i386/i386-expand.c
13427 (ix86_expand_args_builtin): Handle new builtins.
13428 (ix86_expand_round_builtin): Ditto.
13429 * config/i386/sse.md (VF_AVX512FP16VL): New.
13430 (sqrt<mode>2): Adjust for HF vector modes.
13431 (<sse>_sqrt<mode>2<mask_name><round_name>): Likewise.
13432 (<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
13434 (<sse>_rsqrt<mode>2<mask_name>): New.
13435 (avx512fp16_vmrsqrtv8hf2<mask_scalar_name>): Likewise.
13437 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
13439 PR bootstrap/101574
13440 * diagnostic-spec.c (warning_suppressed_at, copy_warning): Handle
13441 'RESERVED_LOCATION_P' locations.
13442 * warning-control.cc (get_nowarn_spec, suppress_warning)
13443 (copy_warning): Likewise.
13445 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
13447 * diagnostic-spec.h (typedef xint_hash_t): Use 'location_t' instead of...
13448 (typedef key_type_t): ... this. Remove.
13449 (nowarn_map): Document.
13450 * diagnostic-spec.c (nowarn_map): Likewise.
13451 * warning-control.cc (convert_to_key): Evolve functions into...
13452 (get_location): ... these. Adjust all users.
13454 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
13456 * warning-control.cc (copy_warning): Remove 'nowarn_map' setup.
13458 2021-09-13 Jason Merrill <jason@redhat.com>
13460 * params.opt: Add destructive-interference-size and
13461 constructive-interference-size.
13462 * doc/invoke.texi: Document them.
13463 * config/aarch64/aarch64.c (aarch64_override_options_internal):
13465 * config/arm/arm.c (arm_option_override): Set them.
13466 * config/i386/i386-options.c (ix86_option_override_internal):
13469 2021-09-13 Martin Liska <mliska@suse.cz>
13470 H.J. Lu <hjl.tools@gmail.com>
13473 * common/config/i386/cpuinfo.h (cpu_indicator_init): Add support
13474 for x86-64 micro levels for __builtin_cpu_supports.
13475 * common/config/i386/i386-cpuinfo.h (enum feature_priority):
13476 Add priorities for the micro-arch levels.
13477 (enum processor_features): Add new features.
13478 * common/config/i386/i386-isas.h: Add micro-arch features.
13479 * config/i386/i386-builtins.c (get_builtin_code_for_version):
13480 Support the micro-arch levels by callsing
13481 __builtin_cpu_supports.
13482 * doc/extend.texi: Document that the levels are support by
13483 __builtin_cpu_supports.
13485 2021-09-13 Andrew Pinski <apinski@marvell.com>
13488 * config/aarch64/aarch64-builtins.c (aarch64_fold_builtin_lane_check):
13490 (aarch64_general_fold_builtin): Handle AARCH64_SIMD_BUILTIN_LANE_CHECK.
13491 (aarch64_general_gimple_fold_builtin): Likewise.
13493 2021-09-13 Andrew Pinski <apinski@marvell.com>
13495 * config.gcc: Add m32r-*-linux* and m32rle-*-linux*
13496 to the Unsupported targets list.
13497 Remove support for m32r-*-linux* and m32rle-*-linux*.
13498 * config/m32r/linux.h: Removed.
13499 * config/m32r/t-linux: Removed.
13501 2021-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13504 * config/aarch64/aarch64.c (aarch64_classify_address): Don't allow
13505 register index for SVE predicate modes.
13507 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
13509 * tree-ssa-threadbackward.c
13510 (back_threader_profitability::profitable_path_p): Remove FSM
13512 (back_threader_registry::register_path): Same.
13513 * tree-ssa-threadedge.c
13514 (jump_threader::simplify_control_stmt_condition): Same.
13515 * tree-ssa-threadupdate.c (jt_path_registry::jt_path_registry):
13516 Add backedge_threads argument.
13517 (fwd_jt_path_registry::fwd_jt_path_registry): Pass
13518 backedge_threads argument.
13519 (back_jt_path_registry::back_jt_path_registry): Same.
13520 (dump_jump_thread_path): Adjust for FSM removal.
13521 (back_jt_path_registry::rewire_first_differing_edge): Same.
13522 (back_jt_path_registry::adjust_paths_after_duplication): Same.
13523 (back_jt_path_registry::update_cfg): Same.
13524 (jt_path_registry::register_jump_thread): Same.
13525 * tree-ssa-threadupdate.h (enum jump_thread_edge_type): Remove
13527 (class back_jt_path_registry): Add backedge_threads to
13530 2021-09-13 Martin Liska <mliska@suse.cz>
13533 * asan.h (sanitize_coverage_p): Handle when fn == NULL.
13535 2021-09-13 H.J. Lu <hjl.tools@gmail.com>
13538 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): New.
13539 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
13540 (MOVE_MAX): Check TARGET_AVX256_MOVE_BY_PIECES and
13541 TARGET_AVX256_STORE_BY_PIECES instead of
13542 TARGET_AVX256_SPLIT_UNALIGNED_LOAD and
13543 TARGET_AVX256_SPLIT_UNALIGNED_STORE.
13544 (STORE_MAX_PIECES): Check TARGET_AVX256_STORE_BY_PIECES instead
13545 of TARGET_AVX256_SPLIT_UNALIGNED_STORE.
13546 * config/i386/x86-tune.def (X86_TUNE_AVX256_MOVE_BY_PIECES): New.
13547 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
13549 2021-09-13 liuhongt <hongtao.liu@intel.com>
13551 PR bootstrap/102302
13552 * expmed.c (extract_bit_field_using_extv): Use
13553 gen_lowpart_if_possible instead of gen_lowpart to avoid ICE.
13555 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
13557 * Makefile.in (OBJS): Add value-pointer-equiv.o.
13558 * gimple-ssa-evrp.c (class ssa_equiv_stack): Move to
13559 value-pointer-equiv.*.
13560 (ssa_equiv_stack::ssa_equiv_stack): Same.
13561 (ssa_equiv_stack::enter): Same.
13562 (ssa_equiv_stack::leave): Same.
13563 (ssa_equiv_stack::push_replacement): Same.
13564 (ssa_equiv_stack::get_replacement): Same.
13565 (is_pointer_ssa): Same.
13566 (class pointer_equiv_analyzer): Same.
13567 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
13568 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Same.
13569 (pointer_equiv_analyzer::set_global_equiv): Same.
13570 (pointer_equiv_analyzer::set_cond_equiv): Same.
13571 (pointer_equiv_analyzer::get_equiv): Same.
13572 (pointer_equiv_analyzer::enter): Same.
13573 (pointer_equiv_analyzer::leave): Same.
13574 (pointer_equiv_analyzer::get_equiv_expr): Same.
13575 (pta_valueize): Same.
13576 (pointer_equiv_analyzer::visit_stmt): Same.
13577 (pointer_equiv_analyzer::visit_edge): Same.
13578 (hybrid_folder::value_of_expr): Same.
13579 (hybrid_folder::value_on_edge): Same.
13580 * value-pointer-equiv.cc: New file.
13581 * value-pointer-equiv.h: New file.
13583 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
13586 * gimple-fold.c (gimple_fold_builtin_memory_op): Allow folding
13587 memcpy if the size is not more than MOVE_MAX * MOVE_RATIO.
13589 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
13592 * config/arm/arm.md (movmisaligndi): New define_expand.
13593 * config/arm/vec-common.md (movmisalign<mode>): Iterate over VDQ mode.
13595 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
13598 * emit-rtl.c (gen_highpart): Use adjust_address to handle
13599 MEM rather than calling simplify_gen_subreg.
13601 2021-09-13 Jan-Benedict Glaw <jbglaw@ług-owl.de>
13603 * config/alpha/vms.h (INIT_CUMULATIVE_ARGS): Wrap multi-statment
13604 define into a block.
13606 2021-09-13 Richard Biener <rguenther@suse.de>
13608 * config/darwin.h (DARWIN_PREFER_DWARF): Do not define.
13609 * config/i386/darwin.h (PREFERRED_DEBUGGING_TYPE): Do not
13610 change based on DARWIN_PREFER_DWARF not being defined.
13612 2021-09-13 Richard Biener <rguenther@suse.de>
13614 * config/i386/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
13615 to inherit from elfos.h
13617 2021-09-13 Richard Biener <rguenther@suse.de>
13619 * config.gcc: Add cr16-*-* to the list of obsoleted targets.
13621 2021-09-13 Richard Biener <rguenther@suse.de>
13623 * config/avr/elf.h (PREFERRED_DEBUGGING_TYPE): Remove
13624 override, pick up DWARF2_DEBUG define from elfos.h
13626 2021-09-13 Richard Biener <rguenther@suse.de>
13628 * config/rx/rx.h (PREFERRED_DEBUGGING_TYPE): Always define to
13631 2021-09-13 Richard Biener <rguenther@suse.de>
13633 * config/alpha/vms.h (PREFERRED_DEBUGGING_TYPE): Define to
13636 2021-09-13 Richard Biener <rguenther@suse.de>
13638 * config/i386/cygming.h: Always default to DWARF2 debugging.
13639 Do not define DBX_DEBUGGING_INFO, that's done via dbxcoff.h
13641 * doc/install.texi: Document binutils 2.16 as minimum
13642 requirement for mingw.
13644 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
13646 * config/rs6000/rs6000.c (struct rs6000_cost_data): New members
13647 nstmts, nloads and extra_ctor_cost.
13648 (rs6000_density_test): Add load density related heuristics. Do
13649 extra costing on vector construction statements if need.
13650 (rs6000_init_cost): Init new members.
13651 (rs6000_update_target_cost_per_stmt): New function.
13652 (rs6000_add_stmt_cost): Factor vect_nonmem hunk out to function
13653 rs6000_update_target_cost_per_stmt and call it.
13655 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
13657 * config/rs6000/rs6000.c (struct rs6000_cost_data): Remove typedef.
13658 (rs6000_init_cost): Adjust.
13660 2021-09-13 liuhongt <hongtao.liu@intel.com>
13662 * config/i386/i386.md: (UNSPEC_COPYSIGN): Remove.
13663 (UNSPEC_XORSIGN): Ditto.
13665 2021-09-12 Roger Sayle <roger@nextmovesoftware.com>
13667 * expr.c (convert_move): Preserve SUBREG_PROMOTED_VAR_P when
13668 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
13671 2021-09-11 Aldy Hernandez <aldyh@redhat.com>
13673 * tree-ssa-threadbackward.c (class back_threader_registry): Use
13674 back_jt_path_registry.
13675 * tree-ssa-threadedge.c (jump_threader::jump_threader): Use
13676 fwd_jt_path_registry.
13677 * tree-ssa-threadedge.h (class jump_threader): Same..
13678 * tree-ssa-threadupdate.c
13679 (jump_thread_path_registry::jump_thread_path_registry): Rename...
13680 (jt_path_registry::jt_path_registry): ...to this.
13681 (jump_thread_path_registry::~jump_thread_path_registry): Rename...
13682 (jt_path_registry::~jt_path_registry): ...this.
13683 (fwd_jt_path_registry::fwd_jt_path_registry): New.
13684 (fwd_jt_path_registry::~fwd_jt_path_registry): New.
13685 (jump_thread_path_registry::allocate_thread_edge): Rename...
13686 (jt_path_registry::allocate_thread_edge): ...to this.
13687 (jump_thread_path_registry::allocate_thread_path): Rename...
13688 (jt_path_registry::allocate_thread_path): ...to this.
13689 (jump_thread_path_registry::lookup_redirection_data): Rename...
13690 (fwd_jt_path_registry::lookup_redirection_data): ...to this.
13691 (jump_thread_path_registry::thread_block_1): Rename...
13692 (fwd_jt_path_registry::thread_block_1): ...to this.
13693 (jump_thread_path_registry::thread_block): Rename...
13694 (fwd_jt_path_registry::thread_block): ...to this.
13695 (jt_path_registry::thread_through_loop_header): Rename...
13696 (fwd_jt_path_registry::thread_through_loop_header): ...to this.
13697 (jump_thread_path_registry::mark_threaded_blocks): Rename...
13698 (fwd_jt_path_registry::mark_threaded_blocks): ...to this.
13699 (jump_thread_path_registry::debug_path): Rename...
13700 (jt_path_registry::debug_path): ...to this.
13701 (jump_thread_path_registry::dump): Rename...
13702 (jt_path_registry::debug): ...to this.
13703 (jump_thread_path_registry::rewire_first_differing_edge): Rename...
13704 (back_jt_path_registry::rewire_first_differing_edge): ...to this.
13705 (jump_thread_path_registry::adjust_paths_after_duplication): Rename...
13706 (back_jt_path_registry::adjust_paths_after_duplication): ...to this.
13707 (jump_thread_path_registry::duplicate_thread_path): Rename...
13708 (back_jt_path_registry::duplicate_thread_path): ...to this. Also,
13709 drop ill-formed candidates.
13710 (jump_thread_path_registry::remove_jump_threads_including): Rename...
13711 (fwd_jt_path_registry::remove_jump_threads_including): ...to this.
13712 (jt_path_registry::thread_through_all_blocks): New.
13713 (back_jt_path_registry::update_cfg): New.
13714 (fwd_jt_path_registry::update_cfg): New.
13715 (jump_thread_path_registry::register_jump_thread): Rename...
13716 (jt_path_registry::register_jump_thread): ...to this.
13717 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
13719 (class jt_path_registry): ...here.
13720 (class fwd_jt_path_registry): New.
13721 (class back_jt_path_registry): New.
13723 2021-09-10 liuhongt <hongtao.liu@intel.com>
13726 2021-09-01 liuhongt <hongtao.liu@intel.com>
13728 * emit-rtl.c (validate_subreg): Get rid of all float-int
13731 2021-09-10 Jakub Jelinek <jakub@redhat.com>
13733 * tree-core.h (enum omp_memory_order): Add OMP_MEMORY_ORDER_MASK,
13734 OMP_FAIL_MEMORY_ORDER_UNSPECIFIED, OMP_FAIL_MEMORY_ORDER_RELAXED,
13735 OMP_FAIL_MEMORY_ORDER_ACQUIRE, OMP_FAIL_MEMORY_ORDER_RELEASE,
13736 OMP_FAIL_MEMORY_ORDER_ACQ_REL, OMP_FAIL_MEMORY_ORDER_SEQ_CST and
13737 OMP_FAIL_MEMORY_ORDER_MASK enumerators.
13738 (OMP_FAIL_MEMORY_ORDER_SHIFT): Define.
13739 * gimple-pretty-print.c (dump_gimple_omp_atomic_load,
13740 dump_gimple_omp_atomic_store): Print [weak] for weak atomic
13742 * gimple.h (enum gf_mask): Change GF_OMP_ATOMIC_MEMORY_ORDER
13743 to 6-bit mask, adjust GF_OMP_ATOMIC_NEED_VALUE value and add
13744 GF_OMP_ATOMIC_WEAK.
13745 (gimple_omp_atomic_weak_p, gimple_omp_atomic_set_weak): New inline
13747 * tree.h (OMP_ATOMIC_WEAK): Define.
13748 * tree-pretty-print.c (dump_omp_atomic_memory_order): Adjust for
13749 fail memory order being encoded in the same enum and also print
13750 fail clause if present.
13751 (dump_generic_node): Print weak clause if OMP_ATOMIC_WEAK.
13752 * gimplify.c (goa_stabilize_expr): Add target_expr and rhs arguments,
13753 handle pre_p == NULL case as a test mode that only returns value
13754 but doesn't change gimplify nor change anything otherwise, adjust
13755 recursive calls, add MODIFY_EXPR, ADDR_EXPR, COND_EXPR, TARGET_EXPR
13756 and CALL_EXPR handling, adjust COMPOUND_EXPR handling for
13757 __builtin_clear_padding calls, for !rhs gimplify as lvalue rather
13759 (gimplify_omp_atomic): Adjust goa_stabilize_expr caller. Handle
13760 COND_EXPR rhs. Set weak flag on gimple load/store for
13762 * omp-expand.c (omp_memory_order_to_fail_memmodel): New function.
13763 (omp_memory_order_to_memmodel): Adjust for fail clause encoded
13765 (expand_omp_atomic_cas): New function.
13766 (expand_omp_atomic_pipeline): Use omp_memory_order_to_fail_memmodel
13768 (expand_omp_atomic): Attempt to optimize atomic compare and exchange
13769 using expand_omp_atomic_cas.
13771 2021-09-10 Aldy Hernandez <aldyh@redhat.com>
13772 Michael Matz <matz@suse.de>
13774 * tree-pass.h (PROP_loop_opts_done): New.
13775 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
13776 Intersect with global range.
13777 * tree-ssa-loop.c (tree_ssa_loop_done): Set PROP_loop_opts_done.
13778 * tree-ssa-threadbackward.c
13779 (back_threader_profitability::profitable_path_p): Disable
13780 threading through latches until after loop optimizations have run.
13782 2021-09-10 David Faust <david.faust@oracle.com>
13784 * doc/invoke.texi: Document BPF -mcpu, -mjmpext, -mjmp32 and -malu32
13787 2021-09-10 David Faust <david.faust@oracle.com>
13789 * config/bpf/bpf-opts.h (bpf_isa_version): New enum.
13790 * config/bpf/bpf-protos.h (bpf_expand_cbranch): New.
13791 * config/bpf/bpf.c (bpf_option_override): Handle -mcpu option.
13792 (bpf_expand_cbranch): New function.
13793 * config/bpf/bpf.md (AM mode iterator): Conditionalize support for SI
13795 (zero_extendsidi2): Only use mov32 instruction if it is available.
13796 (SIM mode iterator): Conditionalize support for SI mode.
13797 (JM mode iterator): New.
13798 (cbranchdi4): Update name, use new JM iterator. Use bpf_expand_cbranch.
13799 (*branch_on_di): Update name, use new JM iterator.
13800 * config/bpf/bpf.opt: (mjmpext): New option.
13801 (malu32): Likewise.
13802 (mjmp32): Likewise.
13804 (bpf_isa): New enum.
13806 2021-09-10 David Faust <david.faust@oracle.com>
13808 * config/bpf/bpf.md (zero_extendhidi2): Add new output template
13809 for register-to-register extensions.
13810 (zero_extendqidi2): Likewise.
13812 2021-09-10 Richard Biener <rguenther@suse.de>
13814 PR middle-end/102273
13815 * internal-fn.c (expand_DEFERRED_INIT): Always expand non-SSA vars.
13817 2021-09-10 Richard Biener <rguenther@suse.de>
13819 PR middle-end/102269
13820 * gimplify.c (is_var_need_auto_init): Empty types do not need
13823 2021-09-10 Richard Biener <rguenther@suse.de>
13825 * configure.ac (--with-stabs): Remove.
13826 * configure: Regenerate.
13827 * doc/install.texi: Remove --with-stabs documentation.
13829 2021-09-10 liuhongt <hongtao.liu@intel.com>
13831 * config/i386/avx512fp16intrin.h: (_mm512_cmp_ph_mask):
13833 (_mm512_mask_cmp_ph_mask): Likewise.
13834 (_mm512_cmp_round_ph_mask): Likewise.
13835 (_mm512_mask_cmp_round_ph_mask): Likewise.
13836 (_mm_cmp_sh_mask): Likewise.
13837 (_mm_mask_cmp_sh_mask): Likewise.
13838 (_mm_cmp_round_sh_mask): Likewise.
13839 (_mm_mask_cmp_round_sh_mask): Likewise.
13840 (_mm_comieq_sh): Likewise.
13841 (_mm_comilt_sh): Likewise.
13842 (_mm_comile_sh): Likewise.
13843 (_mm_comigt_sh): Likewise.
13844 (_mm_comige_sh): Likewise.
13845 (_mm_comineq_sh): Likewise.
13846 (_mm_ucomieq_sh): Likewise.
13847 (_mm_ucomilt_sh): Likewise.
13848 (_mm_ucomile_sh): Likewise.
13849 (_mm_ucomigt_sh): Likewise.
13850 (_mm_ucomige_sh): Likewise.
13851 (_mm_ucomineq_sh): Likewise.
13852 (_mm_comi_round_sh): Likewise.
13853 (_mm_comi_sh): Likewise.
13854 * config/i386/avx512fp16vlintrin.h (_mm_cmp_ph_mask): New intrinsic.
13855 (_mm_mask_cmp_ph_mask): Likewise.
13856 (_mm256_cmp_ph_mask): Likewise.
13857 (_mm256_mask_cmp_ph_mask): Likewise.
13858 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
13859 * config/i386/i386-builtin.def: Add corresponding new builtins.
13860 * config/i386/i386-expand.c
13861 (ix86_expand_args_builtin): Handle new builtin types.
13862 (ix86_expand_round_builtin): Ditto.
13863 * config/i386/i386.md (ssevecmode): Add HF mode.
13864 (MODEFH): New mode iterator.
13865 * config/i386/sse.md
13866 (V48H_AVX512VL): New mode iterator to support HF vector modes.
13867 Ajdust corresponding description.
13868 (ssecmpintprefix): New.
13869 (VI12_AVX512VL): Adjust to support HF vector modes.
13870 (cmp_imm_predicate): Likewise.
13871 (<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
13873 (avx512f_vmcmp<mode>3<round_saeonly_name>): Likewise.
13874 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Likewise.
13875 (<sse>_<unord>comi<round_saeonly_name>): Likewise.
13877 2021-09-10 liuhongt <hongtao.liu@intel.com>
13879 * config/i386/avx512fp16intrin.h: (_mm512_max_ph): New intrinsic.
13880 (_mm512_mask_max_ph): Likewise.
13881 (_mm512_maskz_max_ph): Likewise.
13882 (_mm512_min_ph): Likewise.
13883 (_mm512_mask_min_ph): Likewise.
13884 (_mm512_maskz_min_ph): Likewise.
13885 (_mm512_max_round_ph): Likewise.
13886 (_mm512_mask_max_round_ph): Likewise.
13887 (_mm512_maskz_max_round_ph): Likewise.
13888 (_mm512_min_round_ph): Likewise.
13889 (_mm512_mask_min_round_ph): Likewise.
13890 (_mm512_maskz_min_round_ph): Likewise.
13891 (_mm_max_sh): Likewise.
13892 (_mm_mask_max_sh): Likewise.
13893 (_mm_maskz_max_sh): Likewise.
13894 (_mm_min_sh): Likewise.
13895 (_mm_mask_min_sh): Likewise.
13896 (_mm_maskz_min_sh): Likewise.
13897 (_mm_max_round_sh): Likewise.
13898 (_mm_mask_max_round_sh): Likewise.
13899 (_mm_maskz_max_round_sh): Likewise.
13900 (_mm_min_round_sh): Likewise.
13901 (_mm_mask_min_round_sh): Likewise.
13902 (_mm_maskz_min_round_sh): Likewise.
13903 * config/i386/avx512fp16vlintrin.h (_mm_max_ph): New intrinsic.
13904 (_mm256_max_ph): Likewise.
13905 (_mm_mask_max_ph): Likewise.
13906 (_mm256_mask_max_ph): Likewise.
13907 (_mm_maskz_max_ph): Likewise.
13908 (_mm256_maskz_max_ph): Likewise.
13909 (_mm_min_ph): Likewise.
13910 (_mm256_min_ph): Likewise.
13911 (_mm_mask_min_ph): Likewise.
13912 (_mm256_mask_min_ph): Likewise.
13913 (_mm_maskz_min_ph): Likewise.
13914 (_mm256_maskz_min_ph): Likewise.
13915 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
13916 * config/i386/i386-builtin.def: Add corresponding new builtins.
13917 * config/i386/i386-expand.c
13918 (ix86_expand_args_builtin): Handle new builtin types.
13919 * config/i386/sse.md
13920 (<code><mode>3<mask_name><round_saeonly_name>): Adjust to
13921 support HF vector modes.
13922 (*<code><mode>3<mask_name><round_saeonly_name>): Likewise.
13923 (ieee_<ieee_maxmin><mode>3<mask_name><round_saeonly_name>):
13925 (<sse>_vm<code><mode>3<mask_scalar_name><round_saeonly_scalar_name>):
13927 * config/i386/subst.md (round_saeonly_mode512bit_condition):
13928 Adjust for HF vector modes.
13930 2021-09-10 Liu, Hongtao <hongtao.liu@intel.com>
13932 * config/i386/avx512fp16intrin.h (_mm_add_sh): New intrinsic.
13933 (_mm_mask_add_sh): Likewise.
13934 (_mm_maskz_add_sh): Likewise.
13935 (_mm_sub_sh): Likewise.
13936 (_mm_mask_sub_sh): Likewise.
13937 (_mm_maskz_sub_sh): Likewise.
13938 (_mm_mul_sh): Likewise.
13939 (_mm_mask_mul_sh): Likewise.
13940 (_mm_maskz_mul_sh): Likewise.
13941 (_mm_div_sh): Likewise.
13942 (_mm_mask_div_sh): Likewise.
13943 (_mm_maskz_div_sh): Likewise.
13944 (_mm_add_round_sh): Likewise.
13945 (_mm_mask_add_round_sh): Likewise.
13946 (_mm_maskz_add_round_sh): Likewise.
13947 (_mm_sub_round_sh): Likewise.
13948 (_mm_mask_sub_round_sh): Likewise.
13949 (_mm_maskz_sub_round_sh): Likewise.
13950 (_mm_mul_round_sh): Likewise.
13951 (_mm_mask_mul_round_sh): Likewise.
13952 (_mm_maskz_mul_round_sh): Likewise.
13953 (_mm_div_round_sh): Likewise.
13954 (_mm_mask_div_round_sh): Likewise.
13955 (_mm_maskz_div_round_sh): Likewise.
13956 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
13957 * config/i386/i386-builtin.def: Add corresponding new builtins.
13958 * config/i386/i386-expand.c
13959 (ix86_expand_round_builtin): Handle new builtins.
13960 * config/i386/sse.md (VF_128): Change description.
13961 (<sse>_vm<plusminus_insn><mode>3<mask_scalar_name><round_scalar_name>):
13962 Adjust to support HF vector modes.
13963 (<sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name><round_scalar_name>):
13966 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
13968 * config/i386/i386-expand.c
13969 (ix86_avx256_split_vector_move_misalign): Handle V16HF mode.
13970 * config/i386/i386.c
13971 (ix86_preferred_simd_mode): Handle HF mode.
13972 * config/i386/sse.md (V_256H): New mode iterator.
13973 (avx_vextractf128<mode>): Use it.
13974 (VEC_INIT_MODE): Align vector HFmode condition to vector
13975 HImodes since there're no real HF instruction used.
13976 (VEC_INIT_HALF_MODE): Ditto.
13978 (VIHF_AVX512BW): Ditto.
13979 (*vec_extracthf): Ditto.
13980 (VEC_EXTRACT_MODE): Ditto.
13982 2021-09-10 Richard Biener <rguenther@suse.de>
13985 * config/dbx.h: Remove.
13986 * config/dbxcoff.h: Do not define PREFERRED_DEBUGGING_TYPE.
13987 * config/lynx.h: Likewise.
13989 2021-09-10 liuhongt <hongtao.liu@intel.com>
13991 * config/i386/i386-expand.c (ix86_expand_copysign): Expand
13992 right into ANDNOT + AND + IOR, using paradoxical subregs.
13993 (ix86_split_copysign_const): Remove.
13994 (ix86_split_copysign_var): Ditto.
13995 * config/i386/i386-protos.h (ix86_split_copysign_const): Dotto.
13996 (ix86_split_copysign_var): Ditto.
13997 * config/i386/i386.md (@copysign<mode>3_const): Ditto.
13998 (@copysign<mode>3_var): Ditto.
14000 2021-09-09 qing zhao <qing.zhao@oracle.com>
14002 * builtins.c (expand_builtin_memset): Make external visible.
14003 * builtins.h (expand_builtin_memset): Declare extern.
14004 * common.opt (ftrivial-auto-var-init=): New option.
14005 * doc/extend.texi: Document the uninitialized attribute.
14006 * doc/invoke.texi: Document -ftrivial-auto-var-init.
14007 * flag-types.h (enum auto_init_type): New enumerated type
14009 * gimple-fold.c (clear_padding_type): Add one new parameter.
14010 (clear_padding_union): Likewise.
14011 (clear_padding_emit_loop): Likewise.
14012 (clear_type_padding_in_mask): Likewise.
14013 (gimple_fold_builtin_clear_padding): Handle this new parameter.
14014 * gimplify.c (gimple_add_init_for_auto_var): New function.
14015 (gimple_add_padding_init_for_auto_var): New function.
14016 (is_var_need_auto_init): New function.
14017 (gimplify_decl_expr): Add initialization to automatic variables per
14019 (gimplify_call_expr): Add one new parameter for call to
14020 __builtin_clear_padding.
14021 (gimplify_init_constructor): Add padding initialization in the end.
14022 * internal-fn.c (INIT_PATTERN_VALUE): New macro.
14023 (expand_DEFERRED_INIT): New function.
14024 * internal-fn.def (DEFERRED_INIT): New internal function.
14025 * tree-cfg.c (verify_gimple_call): Verify calls to .DEFERRED_INIT.
14026 * tree-sra.c (generate_subtree_deferred_init): New function.
14027 (scan_function): Avoid setting cannot_scalarize_away_bitmap for
14028 calls to .DEFERRED_INIT.
14029 (sra_modify_deferred_init): New function.
14030 (sra_modify_function_body): Handle calls to DEFERRED_INIT specially.
14031 * tree-ssa-structalias.c (find_func_aliases_for_call): Likewise.
14032 * tree-ssa-uninit.c (warn_uninit): Handle calls to DEFERRED_INIT
14034 (check_defs): Likewise.
14035 (warn_uninitialized_vars): Likewise.
14036 * tree-ssa.c (ssa_undefined_value_p): Likewise.
14037 * tree.c (build_common_builtin_nodes): Build tree node for
14038 BUILT_IN_CLEAR_PADDING when needed.
14040 2021-09-09 Richard Biener <rguenther@suse.de>
14042 * tree-ssa-loop-im.c (fill_always_executed_in_1): Walk
14045 2021-09-09 Richard Biener <rguenther@suse.de>
14047 * tree-ssa-loop-im.c (fill_always_executed_in_1): Integrate
14048 DOM walk from get_loop_body_in_dom_order using a worklist
14051 2021-09-09 liuhongt <hongtao.liu@intel.com>
14053 * config.gcc: Add avx512fp16vlintrin.h.
14054 * config/i386/avx512fp16intrin.h: (_mm512_add_ph): New intrinsic.
14055 (_mm512_mask_add_ph): Likewise.
14056 (_mm512_maskz_add_ph): Likewise.
14057 (_mm512_sub_ph): Likewise.
14058 (_mm512_mask_sub_ph): Likewise.
14059 (_mm512_maskz_sub_ph): Likewise.
14060 (_mm512_mul_ph): Likewise.
14061 (_mm512_mask_mul_ph): Likewise.
14062 (_mm512_maskz_mul_ph): Likewise.
14063 (_mm512_div_ph): Likewise.
14064 (_mm512_mask_div_ph): Likewise.
14065 (_mm512_maskz_div_ph): Likewise.
14066 (_mm512_add_round_ph): Likewise.
14067 (_mm512_mask_add_round_ph): Likewise.
14068 (_mm512_maskz_add_round_ph): Likewise.
14069 (_mm512_sub_round_ph): Likewise.
14070 (_mm512_mask_sub_round_ph): Likewise.
14071 (_mm512_maskz_sub_round_ph): Likewise.
14072 (_mm512_mul_round_ph): Likewise.
14073 (_mm512_mask_mul_round_ph): Likewise.
14074 (_mm512_maskz_mul_round_ph): Likewise.
14075 (_mm512_div_round_ph): Likewise.
14076 (_mm512_mask_div_round_ph): Likewise.
14077 (_mm512_maskz_div_round_ph): Likewise.
14078 * config/i386/avx512fp16vlintrin.h: New header.
14079 * config/i386/i386-builtin-types.def (V16HF, V8HF, V32HF):
14080 Add new builtin types.
14081 * config/i386/i386-builtin.def: Add corresponding builtins.
14082 * config/i386/i386-expand.c
14083 (ix86_expand_args_builtin): Handle new builtin types.
14084 (ix86_expand_round_builtin): Likewise.
14085 * config/i386/immintrin.h: Include avx512fp16vlintrin.h
14086 * config/i386/sse.md (VFH): New mode_iterator.
14088 (avx512fmaskmode): Add HF vector modes.
14089 (avx512fmaskhalfmode): Likewise.
14090 (<plusminus_insn><mode>3<mask_name><round_name>): Adjust to for
14092 (*<plusminus_insn><mode>3<mask_name><round_name>): Likewise.
14093 (mul<mode>3<mask_name><round_name>): Likewise.
14094 (*mul<mode>3<mask_name><round_name>): Likewise.
14095 (div<mode>3): Likewise.
14096 (<sse>_div<mode>3<mask_name><round_name>): Likewise.
14097 * config/i386/subst.md (SUBST_V): Add HF vector modes.
14098 (SUBST_A): Likewise.
14099 (round_mode512bit_condition): Adjust for V32HFmode.
14101 2021-09-09 liuhongt <hongtao.liu@intel.com>
14104 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
14105 (reduc_plus_scal_v4sf): .. this, New define_expand.
14106 (reduc_plus_scal_v2df): .. and this, New define_expand.
14108 2021-09-09 liuhongt <hongtao.liu@intel.com>
14111 * config/i386/sse.md (*vec_extract<mode><ssescalarmodelower>_valign):
14114 2021-09-08 Jonathan Wakely <jwakely@redhat.com>
14117 * doc/trouble.texi (Copy Assignment): Fix description of
14118 behaviour and fix code in example.
14120 2021-09-08 Segher Boessenkool <segher@kernel.crashing.org>
14123 * config/rs6000/rs6000-logue.c (rs6000_emit_epilogue): For ELFv2 use
14124 r11 instead of r12 for restoring CR.
14126 2021-09-08 Jakub Jelinek <jakub@redhat.com>
14127 liuhongt <hongtao.liu@intel.com>
14130 * config/i386/i386.md (@xorsign<mode>3_1): Remove.
14131 * config/i386/i386-expand.c (ix86_expand_xorsign): Expand right away
14132 into AND with mask and XOR, using paradoxical subregs.
14133 (ix86_split_xorsign): Remove.
14134 * config/i386/i386-protos.h (ix86_split_xorsign): Remove.
14136 2021-09-08 Di Zhao <dizhao@os.amperecomputing.com>
14138 * tree-ssa-sccvn.c (vn_nary_op_insert_into): fix result compare
14140 2021-09-08 Jakub Jelinek <jakub@redhat.com>
14143 * config/i386/i386.md (xorsign<mode>3): If operands[1] is equal to
14144 operands[2], emit abs<mode>2 instead.
14145 (@xorsign<mode>3_1): Add early-clobbers for output operand, enable
14146 first alternative even for avx, add another alternative with
14147 =&Yv <- 0, Yv, Yvm constraints.
14148 * config/i386/i386-expand.c (ix86_split_xorsign): If op0 is equal
14149 to op1, emit vpandn instead.
14151 2021-09-08 liuhongt <hongtao.liu@intel.com>
14153 * config/i386/avx512fp16intrin.h (_mm_set_ph): New intrinsic.
14154 (_mm256_set_ph): Likewise.
14155 (_mm512_set_ph): Likewise.
14156 (_mm_setr_ph): Likewise.
14157 (_mm256_setr_ph): Likewise.
14158 (_mm512_setr_ph): Likewise.
14159 (_mm_set1_ph): Likewise.
14160 (_mm256_set1_ph): Likewise.
14161 (_mm512_set1_ph): Likewise.
14162 (_mm_setzero_ph): Likewise.
14163 (_mm256_setzero_ph): Likewise.
14164 (_mm512_setzero_ph): Likewise.
14165 (_mm_set_sh): Likewise.
14166 (_mm_load_sh): Likewise.
14167 (_mm_store_sh): Likewise.
14168 * config/i386/i386-builtin-types.def (V8HF): New type.
14169 (DEF_FUNCTION_TYPE (V8HF, V8HI)): New builtin function type
14170 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
14171 Support vector HFmodes.
14172 (ix86_expand_vector_init_one_nonzero): Likewise.
14173 (ix86_expand_vector_init_one_var): Likewise.
14174 (ix86_expand_vector_init_interleave): Likewise.
14175 (ix86_expand_vector_init_general): Likewise.
14176 (ix86_expand_vector_set): Likewise.
14177 (ix86_expand_vector_extract): Likewise.
14178 (ix86_expand_vector_init_concat): Likewise.
14179 (ix86_expand_sse_movcc): Handle vector HFmodes.
14180 (ix86_expand_vector_set_var): Ditto.
14181 * config/i386/i386-modes.def: Add HF vector modes in comment.
14182 * config/i386/i386.c (classify_argument): Add HF vector modes.
14183 (ix86_hard_regno_mode_ok): Allow HF vector modes for AVX512FP16.
14184 (ix86_vector_mode_supported_p): Likewise.
14185 (ix86_set_reg_reg_cost): Handle vector HFmode.
14186 (ix86_get_ssemov): Handle vector HFmode.
14187 (function_arg_advance_64): Pass unamed V16HFmode and V32HFmode
14189 (function_arg_advance_32): Pass V8HF/V16HF/V32HF by sse reg for 32bit
14191 (function_arg_advance_32): Ditto.
14192 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): New.
14193 (VALID_AVX256_REG_OR_OI_MODE): Rename to ..
14194 (VALID_AVX256_REG_OR_OI_VHF_MODE): .. this, and add V16HF.
14195 (VALID_SSE2_REG_VHF_MODE): New.
14196 (VALID_AVX512VL_128_REG_MODE): Add V8HF and TImode.
14197 (SSE_REG_MODE_P): Add vector HFmode.
14198 * config/i386/i386.md (mode): Add HF vector modes.
14199 (MODE_SIZE): Likewise.
14200 (ssemodesuffix): Add ph suffix for HF vector modes.
14201 * config/i386/sse.md (VFH_128): New mode iterator.
14202 (VMOVE): Adjust for HF vector modes.
14204 (V_256_512): Likewise.
14205 (avx512): Likewise.
14206 (avx512fmaskmode): Likewise.
14207 (shuffletype): Likewise.
14208 (sseinsnmode): Likewise.
14209 (ssedoublevecmode): Likewise.
14210 (ssehalfvecmode): Likewise.
14211 (ssehalfvecmodelower): Likewise.
14212 (ssePScmode): Likewise.
14213 (ssescalarmode): Likewise.
14214 (ssescalarmodelower): Likewise.
14215 (sseintprefix): Likewise.
14217 (bcstscalarsuff): Likewise.
14218 (xtg_mode): Likewise.
14219 (VI12HF_AVX512VL): New mode_iterator.
14220 (VF_AVX512FP16): Likewise.
14222 (VIHF_256): Likewise.
14223 (VIHF_AVX512BW): Likewise.
14224 (V16_256): Likewise.
14225 (V32_512): Likewise.
14226 (sseintmodesuffix): New mode_attr.
14227 (sse): Add scalar and vector HFmodes.
14228 (ssescalarmode): Add vector HFmode mapping.
14229 (ssescalarmodesuffix): Add sh suffix for HFmode.
14230 (*<sse>_vm<insn><mode>3): Use VFH_128.
14231 (*<sse>_vm<multdiv_mnemonic><mode>3): Likewise.
14232 (*ieee_<ieee_maxmin><mode>3): Likewise.
14233 (<avx512>_blendm<mode>): New define_insn.
14234 (vec_setv8hf): New define_expand.
14235 (vec_set<mode>_0): New define_insn for HF vector set.
14236 (*avx512fp16_movsh): Likewise.
14237 (avx512fp16_movsh): Likewise.
14238 (vec_extract_lo_v32hi): Rename to ...
14239 (vec_extract_lo_<mode>): ... this, and adjust to allow HF
14241 (vec_extract_hi_v32hi): Likewise.
14242 (vec_extract_hi_<mode>): Likewise.
14243 (vec_extract_lo_v16hi): Likewise.
14244 (vec_extract_lo_<mode>): Likewise.
14245 (vec_extract_hi_v16hi): Likewise.
14246 (vec_extract_hi_<mode>): Likewise.
14247 (vec_set_hi_v16hi): Likewise.
14248 (vec_set_hi_<mode>): Likewise.
14249 (vec_set_lo_v16hi): Likewise.
14250 (vec_set_lo_<mode>): Likewise.
14251 (*vec_extract<mode>_0): New define_insn_and_split for HF
14253 (*vec_extracthf): New define_insn.
14254 (VEC_EXTRACT_MODE): Add HF vector modes.
14255 (PINSR_MODE): Add V8HF.
14256 (sse2p4_1): Likewise.
14257 (pinsr_evex_isa): Likewise.
14258 (<sse2p4_1>_pinsr<ssemodesuffix>): Adjust to support
14259 insert for V8HFmode.
14260 (pbroadcast_evex_isa): Add HF vector modes.
14261 (AVX2_VEC_DUP_MODE): Likewise.
14262 (VEC_INIT_MODE): Likewise.
14263 (VEC_INIT_HALF_MODE): Likewise.
14264 (avx2_pbroadcast<mode>): Adjust to support HF vector mode
14266 (avx2_pbroadcast<mode>_1): Likewise.
14267 (<avx512>_vec_dup<mode>_1): Likewise.
14268 (<avx512>_vec_dup<mode><mask_name>): Likewise.
14269 (<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>):
14272 2021-09-08 Guo, Xuepeng <xuepeng.guo@intel.com>
14273 H.J. Lu <hongjiu.lu@intel.com>
14274 Liu Hongtao <hongtao.liu@intel.com>
14275 Wang Hongyu <hongyu.wang@intel.com>
14276 Xu Dianhong <dianhong.xu@intel.com>
14278 * common/config/i386/cpuinfo.h (get_available_features):
14279 Detect FEATURE_AVX512FP16.
14280 * common/config/i386/i386-common.c
14281 (OPTION_MASK_ISA_AVX512FP16_SET,
14282 OPTION_MASK_ISA_AVX512FP16_UNSET,
14283 OPTION_MASK_ISA2_AVX512FP16_SET,
14284 OPTION_MASK_ISA2_AVX512FP16_UNSET): New.
14285 (OPTION_MASK_ISA2_AVX512BW_UNSET,
14286 OPTION_MASK_ISA2_AVX512BF16_UNSET): Add AVX512FP16.
14287 (ix86_handle_option): Handle -mavx512fp16.
14288 * common/config/i386/i386-cpuinfo.h (enum processor_features):
14289 Add FEATURE_AVX512FP16.
14290 * common/config/i386/i386-isas.h: Add entry for AVX512FP16.
14291 * config.gcc: Add avx512fp16intrin.h.
14292 * config/i386/avx512fp16intrin.h: New intrinsic header.
14293 * config/i386/cpuid.h: Add bit_AVX512FP16.
14294 * config/i386/i386-builtin-types.def: (FLOAT16): New primitive type.
14295 * config/i386/i386-builtins.c: Support _Float16 type for i386
14297 (ix86_register_float16_builtin_type): New function.
14298 (ix86_float16_type_node): New.
14299 * config/i386/i386-c.c (ix86_target_macros_internal): Define
14301 * config/i386/i386-expand.c (ix86_expand_branch): Support
14303 (ix86_prepare_fp_compare_args): Adjust TARGET_SSE_MATH &&
14304 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
14305 (ix86_expand_fp_movcc): Ditto.
14306 * config/i386/i386-isa.def: Add PTA define for AVX512FP16.
14307 * config/i386/i386-options.c (isa2_opts): Add -mavx512fp16.
14308 (ix86_valid_target_attribute_inner_p): Add avx512fp16 attribute.
14309 * config/i386/i386.c (ix86_get_ssemov): Use
14310 vmovdqu16/vmovw/vmovsh for HFmode/HImode scalar or vector.
14311 (ix86_get_excess_precision): Use
14312 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when TARGET_AVX512FP16
14314 (sse_store_index): Use SFmode cost for HFmode cost.
14315 (inline_memory_move_cost): Add HFmode, and perfer SSE cost over
14316 GPR cost for HFmode.
14317 (ix86_hard_regno_mode_ok): Allow HImode in sse register.
14318 (ix86_mangle_type): Add manlging for _Float16 type.
14319 (inline_secondary_memory_needed): No memory is needed for
14320 16bit movement between gpr and sse reg under
14322 (ix86_multiplication_cost): Adjust TARGET_SSE_MATH &&
14323 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
14324 (ix86_division_cost): Ditto.
14325 (ix86_rtx_costs): Ditto.
14326 (ix86_add_stmt_cost): Ditto.
14327 (ix86_optab_supported_p): Ditto.
14328 * config/i386/i386.h (VALID_AVX512F_SCALAR_MODE): Add HFmode.
14329 (SSE_FLOAT_MODE_SSEMATH_OR_HF_P): Add HFmode.
14330 (PTA_SAPPHIRERAPIDS): Add PTA_AVX512FP16.
14331 * config/i386/i386.md (mode): Add HFmode.
14332 (MODE_SIZE): Add HFmode.
14333 (isa): Add avx512fp16.
14334 (enabled): Handle avx512fp16.
14335 (ssemodesuffix): Add sh suffix for HFmode.
14336 (comm): Add mult, div.
14337 (plusminusmultdiv): New code iterator.
14338 (insn): Add mult, div.
14339 (*movhf_internal): Adjust for avx512fp16 instruction.
14340 (*movhi_internal): Ditto.
14341 (*cmpi<unord>hf): New define_insn for HFmode.
14342 (*ieee_s<ieee_maxmin>hf3): Likewise.
14343 (extendhf<mode>2): Likewise.
14344 (trunc<mode>hf2): Likewise.
14345 (float<floatunssuffix><mode>hf2): Likewise.
14346 (*<insn>hf): Likewise.
14347 (cbranchhf4): New expander.
14348 (movhfcc): Likewise.
14349 (<insn>hf3): Likewise.
14350 (mulhf3): Likewise.
14351 (divhf3): Likewise.
14352 * config/i386/i386.opt: Add mavx512fp16.
14353 * config/i386/immintrin.h: Include avx512fp16intrin.h.
14354 * doc/invoke.texi: Add mavx512fp16.
14355 * doc/extend.texi: Add avx512fp16 Usage Notes.
14357 2021-09-08 liuhongt <hongtao.liu@intel.com>
14359 * common.opt: Support -fexcess-precision=16.
14360 * config/aarch64/aarch64.c (aarch64_excess_precision): Return
14361 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when
14362 EXCESS_PRECISION_TYPE_FLOAT16.
14363 * config/arm/arm.c (arm_excess_precision): Ditto.
14364 * config/i386/i386.c (ix86_get_excess_precision): Ditto.
14365 * config/m68k/m68k.c (m68k_excess_precision): Issue an error
14366 when EXCESS_PRECISION_TYPE_FLOAT16.
14367 * config/s390/s390.c (s390_excess_precision): Ditto.
14368 * coretypes.h (enum excess_precision_type): Add
14369 EXCESS_PRECISION_TYPE_FLOAT16.
14370 * doc/tm.texi (TARGET_C_EXCESS_PRECISION): Update documents.
14371 * doc/tm.texi.in (TARGET_C_EXCESS_PRECISION): Ditto.
14372 * doc/extend.texi (Half-Precision): Document
14373 -fexcess-precision=16.
14374 * flag-types.h (enum excess_precision): Add
14375 EXCESS_PRECISION_FLOAT16.
14376 * target.def (excess_precision): Update document.
14377 * tree.c (excess_precision_type): Set excess_precision_type to
14378 EXCESS_PRECISION_FLOAT16 when -fexcess-precision=16.
14380 2021-09-08 liuhongt <hongtao.liu@intel.com>
14382 * doc/extend.texi: (@node Floating Types): Adjust the wording.
14383 (@node Half-Precision): Ditto.
14385 2021-09-07 Takayuki 'January June' Suwa <jjsuwa_sys3175@yahoo.co.jp>
14388 * config/xtensa/xtensa.c (xtensa_emit_move_sequence): Add
14389 'CONST_INT_P (src)' to the condition of the block that tries to
14390 eliminate literal when loading integer contant.
14392 2021-09-07 David Faust <david.faust@oracle.com>
14394 * doc/extend.texi (BPF Type Attributes) New node.
14395 Document new preserve_access_index attribute.
14396 Document new preserve_access_index builtin.
14397 * doc/invoke.texi: Document -mco-re and -mno-co-re options.
14399 2021-09-07 David Faust <david.faust@oracle.com>
14401 * config/bpf/bpf.c: Adjust includes.
14402 (bpf_handle_preserve_access_index_attribute): New function.
14403 (bpf_attribute_table): Use it here.
14404 (bpf_builtins): Add BPF_BUILTIN_PRESERVE_ACCESS_INDEX.
14405 (bpf_option_override): Handle "-mco-re" option.
14406 (bpf_asm_init_sections): New.
14407 (TARGET_ASM_INIT_SECTIONS): Redefine.
14408 (bpf_file_end): New.
14409 (TARGET_ASM_FILE_END): Redefine.
14410 (bpf_init_builtins): Add "__builtin_preserve_access_index".
14411 (bpf_core_compute, bpf_core_get_index): New.
14412 (is_attr_preserve_access): New.
14413 (bpf_expand_builtin): Handle new builtins.
14414 (bpf_core_newdecl, bpf_core_is_maybe_aggregate_access): New.
14415 (bpf_core_walk): New.
14416 (bpf_resolve_overloaded_builtin): New.
14417 (TARGET_RESOLVE_OVERLOADED_BUILTIN): Redefine.
14418 (handle_attr): New.
14419 (pass_bpf_core_attr): New RTL pass.
14420 * config/bpf/bpf-passes.def: New file.
14421 * config/bpf/bpf-protos.h (make_pass_bpf_core_attr): New.
14422 * config/bpf/coreout.c: New file.
14423 * config/bpf/coreout.h: Likewise.
14424 * config/bpf/t-bpf (TM_H): Add $(srcdir)/config/bpf/coreout.h.
14425 (coreout.o): New rule.
14426 (PASSES_EXTRA): Add $(srcdir)/config/bpf/bpf-passes.def.
14427 * config.gcc (bpf): Add coreout.h to extra_headers.
14428 Add coreout.o to extra_objs.
14429 Add $(srcdir)/config/bpf/coreout.c to target_gtfiles.
14431 2021-09-07 David Faust <david.faust@oracle.com>
14433 * btfout.c (get_btf_id): Function is no longer static.
14434 * ctfc.h: Expose it here.
14436 2021-09-07 David Faust <david.faust@oracle.com>
14438 * ctfc.c (ctf_lookup_tree_type): New function.
14439 * ctfc.h: Likewise.
14441 2021-09-07 David Faust <david.faust@oracle.com>
14443 * ctfc.c (ctf_dtd_lookup): Function is no longer static.
14444 * ctfc.h: Analogous change.
14446 2021-09-07 David Faust <david.faust@oracle.com>
14448 * dwarf2out.c (lookup_type_die): Function is no longer static.
14449 * dwarf2out.h: Expose it here.
14451 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
14453 * dwarf2ctf.c (ctf_debug_finalize): Make it static.
14454 (ctf_debug_early_finish): New definition.
14455 (ctf_debug_finish): Likewise.
14456 * dwarf2ctf.h (ctf_debug_finalize): Remove declaration.
14457 (ctf_debug_early_finish): New declaration.
14458 (ctf_debug_finish): Likewise.
14459 * dwarf2out.c (dwarf2out_finish): Invoke ctf_debug_finish.
14460 (dwarf2out_early_finish): Invoke ctf_debug_early_finish.
14462 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
14464 * config/bpf/bpf.c (bpf_option_override): For BPF backend, disable LTO
14465 support when compiling for CO-RE.
14466 * config/bpf/bpf.opt: Add new command line option -mco-re.
14468 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
14470 * flag-types.h (enum debug_info_type): Add new enum
14471 DINFO_TYPE_BTF_WITH_CORE.
14472 (BTF_WITH_CORE_DEBUG): New bitmask.
14473 * flags.h (btf_with_core_debuginfo_p): New declaration.
14474 * opts.c (btf_with_core_debuginfo_p): New definition.
14476 2021-09-07 Jason Merrill <jason@redhat.com>
14478 * tree.h (error_operand_p): Change to inline function.
14480 2021-09-07 Aldy Hernandez <aldyh@redhat.com>
14482 * tree-ssa-threadedge.c (forwarder_block_p): Rename to...
14483 (empty_block_with_phis_p): ...this.
14484 (potentially_threadable_block): Same.
14485 (jump_threader::thread_through_normal_block): Same.
14487 2021-09-07 Eric Botcazou <ebotcazou@adacore.com>
14490 * dwarf2out.c (mark_base_types): New overloaded function.
14491 (dwarf2out_early_finish): Invoke it on the COMDAT type list as well
14492 as the compilation unit, and call move_marked_base_types afterward.
14494 2021-09-07 H.J. Lu <hjl.tools@gmail.com>
14497 * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
14499 (ix86_expand_vector_convert_uns_vsivsf): Likewise.
14501 2021-09-07 Richard Biener <rguenther@suse.de>
14503 PR tree-optimization/102226
14504 * tree-vect-loop.c (vect_transform_cycle_phi): Record
14505 the converted value for the epilogue PHI use.
14507 2021-09-07 Martin Liska <mliska@suse.cz>
14509 PR gcov-profile/80223
14510 * ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
14511 options, do not inline when no_profile_instrument_function
14512 attributes are different in early inliner. It's fine to inline
14513 it after PGO instrumentation.
14515 2021-09-07 Richard Biener <rguenther@suse.de>
14517 PR tree-optimization/101555
14518 * tree-ssa-pre.c (translate_vuse_through_block): Do not
14519 perform an alias walk to determine the validity of the
14520 mem at the start of the block which is already guaranteed
14521 by means of prune_clobbered_mems.
14522 (phi_translate_1): Pass edge to translate_vuse_through_block.
14524 2021-09-07 Xionghu Luo <luoxhu@linux.ibm.com>
14527 * config/rs6000/rs6000.md (fmod<mode>3): New define_expand.
14528 (remainder<mode>3): Likewise.
14530 2021-09-07 YunQiang Su <yunqiang.su@cipunited.com>
14532 * config/mips/mips.c (mips_file_start): add .module for
14535 2021-09-06 Roger Sayle <roger@nextmovesoftware.com>
14537 * wide-int.cc (wi::clz): Reorder tests to ensure the result
14538 is zero for all negative values.
14540 2021-09-06 Tobias Burnus <tobias@codesourcery.com>
14542 * doc/invoke.texi (-foffload-options): Fix @opindex.
14544 2021-09-06 H.J. Lu <hjl.tools@gmail.com>
14547 * config/i386/i386-expand.c (ix86_split_xorsign): Use operands[2].
14548 * config/i386/i386.md (@xorsign<mode>3_1): Add non-destructive
14549 source alternative for AVX.
14551 2021-09-06 liuhongt <hongtao.liu@intel.com>
14553 PR middle-end/102182
14554 * optabs.c (expand_fix): Add from1 to avoid from being
14557 2021-09-06 Eric Botcazou <ebotcazou@adacore.com>
14559 * dwarf2out.c (modified_type_die): Deal with all array types earlier
14560 and use local variable consistently throughout the function.
14562 2021-09-06 Jakub Jelinek <jakub@redhat.com>
14564 PR tree-optimization/102207
14565 * match.pd: Don't demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
14566 were promoted from signed to wider unsigned type.
14568 2021-09-06 Andrew Pinski <apinski@marvell.com>
14570 PR tree-optimization/63184
14571 * match.pd: Add simplification of pointer_diff of two pointer_plus
14572 with addr_expr in the first operand of each pointer_plus.
14573 Add simplificatoin of ne/eq of two pointer_plus with addr_expr
14574 in the first operand of each pointer_plus.
14576 2021-09-06 Richard Biener <rguenther@suse.de>
14578 PR tree-optimization/102176
14579 * tree-vect-slp.c (vect_slp_gather_vectorized_scalar_stmts):
14581 (vect_bb_slp_scalar_cost): Use the computed set of
14582 vectorized scalar stmts instead of relying on the out-of-date
14583 and not accurate PURE_SLP_STMT.
14584 (vect_bb_vectorization_profitable_p): Compute the set
14585 of vectorized scalar stmts.
14587 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
14589 * gimple-range-path.cc (path_range_query::range_of_stmt): Remove
14590 GIMPLE_COND special casing.
14591 (path_range_query::range_defined_in_block): Use range_of_stmt
14592 instead of calling fold_range directly.
14594 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
14596 * gimple-range-path.cc (path_range_query::range_of_expr): Set
14597 m_undefined_path when appropriate.
14598 (path_range_query::internal_range_of_expr): Copy from range_of_expr.
14599 (path_range_query::unreachable_path_p): New.
14600 (path_range_query::precompute_ranges): Set m_undefined_path.
14601 * gimple-range-path.h (path_range_query::unreachable_path_p): New.
14602 (path_range_query::internal_range_of_expr): New.
14603 * tree-ssa-threadbackward.c (back_threader::find_taken_edge_cond):
14604 Use unreachable_path_p.
14606 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
14608 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
14609 Remove argument and call find_taken_edge.
14610 (back_threader::resolve_phi): Do not calculate taken edge before
14611 calling maybe_register_path.
14612 (back_threader::find_paths_to_names): Same.
14614 2021-09-05 Jeff Law <jlaw@localhost.localdomain>
14616 * config/h8300/h8300.md (QHSI2 mode iterator): New mode iterator.
14617 * config/h8300/testcompare.md (store_c): Update name, use new
14619 (store_neg_c, store_shifted_c): New patterns.
14621 2021-09-03 Segher Boessenkool <segher@kernel.crashing.org>
14624 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): On ELFv2 use r11
14625 instead of r12 for CR save, in all cases.
14627 2021-09-03 Andrew Pinski <apinski@marvell.com>
14629 * config/aarch64/aarch64-sve-builtins.cc (register_vector_type):
14630 Handle error_mark_node as the type of the type_decl.
14632 2021-09-03 Andrew Pinski <apinski@marvell.com>
14634 * config/aarch64/aarch64-builtins.c (struct aarch64_simd_type_info):
14636 (aarch64_simd_types): Likewise.
14637 (aarch64_simd_intOI_type_node): Likewise.
14638 (aarch64_simd_intCI_type_node): Likewise.
14639 (aarch64_simd_intXI_type_node): Likewise.
14640 * config/aarch64/aarch64.h (aarch64_fp16_type_node): Likewise.
14641 (aarch64_fp16_ptr_type_node): Likewise.
14642 (aarch64_bf16_type_node): Likewise.
14643 (aarch64_bf16_ptr_type_node): Likewise.
14645 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14647 * range-op.cc (operator_minus::op1_op2_relation_effect): Abstract
14649 (minus_op1_op2_relation_effect): ...here.
14650 (class operator_pointer_diff): New.
14651 (operator_pointer_diff::op1_op2_relation_effect): Call
14652 minus_op1_op2_relation_effect.
14653 (integral_table::integral_table): Add entry for POINTER_DIFF_EXPR.
14655 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14657 * tree-ssa-threadbackward.c (back_threader::thread_through_all_blocks):
14658 Add may_peel_loop_headers.
14659 (back_threader_registry::thread_through_all_blocks): Same.
14660 (try_thread_blocks): Pass may_peel_loop_headers argument.
14661 (pass_early_thread_jumps::execute): Same.
14663 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14665 * tree-ssa-threadedge.c (has_phis_p): New.
14666 (forwarder_block_p): New.
14667 (potentially_threadable_block): Call forwarder_block_p.
14668 (jump_threader::thread_around_empty_blocks): Call has_phis_p.
14669 (jump_threader::thread_through_normal_block): Call
14672 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14674 * tree-ssa-threadbackward.c (back_threader::dump): New.
14675 (back_threader::debug): New.
14676 (back_threader_profitability::profitable_path_p): Dump blocks
14677 even if we are bailing early.
14679 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14681 * tree-ssa-threadupdate.c (cancel_thread): New.
14682 (jump_thread_path_registry::thread_block_1): Use cancel_thread.
14683 (jump_thread_path_registry::mark_threaded_blocks): Same.
14684 (jump_thread_path_registry::register_jump_thread): Same.
14686 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14688 * tree-ssa-threadedge.c (jt_state::push): Only call methods for
14689 which objects are available.
14690 (jt_state::pop): Same.
14691 (jt_state::register_equiv): Same.
14692 (jt_state::register_equivs_on_edge): Same.
14694 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14696 * tree-ssa-threadedge.c (jump_threader::thread_across_edge):
14697 Move pop until after a thread is registered.
14699 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14701 * tree-ssa-threadupdate.c (debug): New.
14703 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14705 * gimple-range-trace.cc (push_dump_file::push_dump_file): New.
14706 (push_dump_file::~push_dump_file): New.
14707 (dump_ranger): Change dump_file temporarily while dumping
14709 * gimple-range-trace.h (class push_dump_file): New.
14711 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14713 * gimple-range-trace.cc (debug_seed_ranger): Remove static.
14714 (dump_ranger): Dump function name.
14716 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14718 * gimple-range-path.cc (path_range_query::range_defined_in_block):
14719 Adjust for non-null.
14720 (path_range_query::adjust_for_non_null_uses): New.
14721 (path_range_query::precompute_ranges): Call
14722 adjust_for_non_null_uses.
14723 * gimple-range-path.h: Add m_non_null and
14724 adjust_for_non_null_uses.
14726 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14728 * gimple-range-path.cc (path_range_query::dump): Dump path
14730 (path_range_query::precompute_ranges): Dump entire path.
14732 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14734 * value-relation.cc (relation_oracle::debug): New.
14735 * value-relation.h (relation_oracle::debug): New.
14737 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14739 * tree-ssa-loop-ch.c: Remove unnecessary include file.
14741 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14743 * gimple-range-fold.cc (fold_using_range::postfold_gcond_edges):
14744 Skip statements with no defining BB.
14745 * gimple-range-path.cc (path_range_query::range_defined_in_block):
14746 Do not get confused by statements with no defining BB.
14748 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
14750 * gimple-range-fold.cc (adjust_imagpart_expr): Move from
14751 gimple_range_adjustment. Add support for constants.
14752 (adjust_realpart_expr): New.
14753 (gimple_range_adjustment): Move IMAGPART_EXPR code to
14754 adjust_imagpart_expr.
14755 * range-op.cc (integral_table::integral_table): Add entry for
14758 2021-09-03 Jakub Jelinek <jakub@redhat.com>
14760 * omp-expand.c (expand_omp_atomic_pipeline): Use
14761 IFN_ATOMIC_COMPARE_EXCHANGE instead of
14762 BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_? so that memory order
14765 2021-09-03 Jakub Jelinek <jakub@redhat.com>
14768 * tree.h (DECL_FIELD_ABI_IGNORED): Changed into rvalue only macro
14769 that is false if DECL_BIT_FIELD.
14770 (SET_DECL_FIELD_ABI_IGNORED, DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD,
14771 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD): Define.
14772 * tree-streamer-out.c (pack_ts_decl_common_value_fields): For
14773 DECL_BIT_FIELD stream DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead
14774 of DECL_FIELD_ABI_IGNORED.
14775 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
14776 SET_DECL_FIELD_ABI_IGNORED instead of writing to
14777 DECL_FIELD_ABI_IGNORED and for DECL_BIT_FIELD use
14778 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead.
14779 * lto-streamer-out.c (hash_tree): For DECL_BIT_FIELD hash
14780 DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead of DECL_FIELD_ABI_IGNORED.
14782 2021-09-03 liuhongt <hongtao.liu@intel.com>
14785 * config/i386/amxbf16intrin.h : Remove macro check for __AMX_BF16__.
14786 * config/i386/amxint8intrin.h : Remove macro check for __AMX_INT8__.
14787 * config/i386/amxtileintrin.h : Remove macro check for __AMX_TILE__.
14789 2021-09-02 Martin Sebor <msebor@redhat.com>
14791 PR tree-optimization/17506
14793 * tree-ssa-uninit.c (warn_uninit): Remove conditional guarding note.
14795 2021-09-02 Richard Biener <rguenther@suse.de>
14797 * tree-ssa-loop-im.c (fill_always_executed_in_1): Refine
14798 fix for PR78185 and continue processing when leaving
14799 finite inner loops.
14801 2021-09-02 Jakub Jelinek <jakub@redhat.com>
14803 PR tree-optimization/99591
14804 * match.pd: Demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
14807 2021-09-02 Richard Biener <rguenther@suse.de>
14810 2021-09-02 Richard Biener <rguenther@suse.de>
14812 PR tree-optimization/102155
14813 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
14814 over a part of the RPO array and do not recurse here.
14815 Dump blocks marked as always executed.
14816 (fill_always_executed_in): Walk over the RPO array and
14817 process loops whose header we run into.
14818 (loop_invariant_motion_in_fun): Compute the first RPO
14819 using rev_post_order_and_mark_dfs_back_seme in iteration
14820 order and pass that to fill_always_executed_in.
14822 2021-09-02 liuhongt <hongtao.liu@intel.com>
14824 * config/i386/i386-modes.def (FLOAT_MODE): Define ieee HFmode.
14825 * config/i386/i386.c (enum x86_64_reg_class): Add
14826 X86_64_SSEHF_CLASS.
14827 (merge_classes): Handle X86_64_SSEHF_CLASS.
14828 (examine_argument): Ditto.
14829 (construct_container): Ditto.
14830 (classify_argument): Ditto, and set HFmode/HCmode to
14831 X86_64_SSEHF_CLASS.
14832 (function_value_32): Return _FLoat16/Complex Float16 by
14834 (function_value_64): Return _Float16/Complex Float16 by SSE
14836 (ix86_print_operand): Handle CONST_DOUBLE HFmode.
14837 (ix86_secondary_reload): Require gpr as intermediate register
14838 to store _Float16 from sse register when sse4 is not
14840 (ix86_libgcc_floating_mode_supported_p): Enable _FLoat16 under
14842 (ix86_scalar_mode_supported_p): Ditto.
14843 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Defined.
14844 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add HFmode.
14845 (VALID_INT_MODE_P): Add HFmode and HCmode.
14846 * config/i386/i386.md (*pushhf_rex64): New define_insn.
14848 (*movhf_internal): Ditto.
14849 * doc/extend.texi (Half-Precision Floating Point): Documemt
14852 2021-09-02 Richard Biener <rguenther@suse.de>
14854 PR tree-optimization/102155
14855 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
14856 over a part of the RPO array and do not recurse here.
14857 Dump blocks marked as always executed.
14858 (fill_always_executed_in): Walk over the RPO array and
14859 process loops whose header we run into.
14860 (loop_invariant_motion_in_fun): Compute the first RPO
14861 using rev_post_order_and_mark_dfs_back_seme in iteration
14862 order and pass that to fill_always_executed_in.
14864 2021-09-02 YunQiang Su <syq@debian.org>
14867 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
14869 * config/mips/mips.c (mips_module_isa_name): New.
14870 mips_file_start: add .module mipsREV to all asm output
14872 2021-09-01 Jeff Law <jlaw@localhost.localdomain>
14874 PR tree-optimization/102152
14875 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Reduce a vector
14876 comparison to a scalar comparison before calling
14877 update_stmt_if_modified.
14879 2021-09-01 Andrew Pinski <apinski@marvell.com>
14882 * config/aarch64/aarch64.c (aarch64_expand_setmem):
14883 Check STRICT_ALIGNMENT before creating an overlapping
14886 2021-09-01 Martin Sebor <msebor@redhat.com>
14888 * gimple-ssa-warn-access.cc (get_size_range): Add argument.
14889 (check_access): Pass additional argument.
14890 (check_memop_access): Remove template and make a member function.
14891 (maybe_check_dealloc_call): Make a pass_waccess member function.
14892 (class pass_waccess): Add, rename, and remove members.
14893 (pass_waccess::pass_waccess): Adjust to name change.
14894 (pass_waccess::~pass_waccess): Same.
14895 (check_alloca): Make a member function.
14896 (check_alloc_size_call): Same.
14897 (check_strcat): Same.
14898 (check_strncat): Same.
14899 (check_stxcpy): Same.
14900 (check_stxncpy): Same.
14901 (check_strncmp): Same.
14902 (maybe_warn_rdwr_sizes): Rename...
14903 (pass_waccess::maybe_check_access_sizes): ...to this.
14904 (pass_waccess::check_call): Adjust to name changes.
14905 (pass_waccess::maybe_check_dealloc_call): Make a pass_waccess member
14907 (pass_waccess::execute): Adjust to name changes.
14908 * gimple-ssa-warn-access.h (check_memop_access): Remove.
14909 * pointer-query.cc (access_ref::phi): Handle null pointer.
14910 (access_ref::inform_access): Same.
14911 (pointer_query::put_ref): Modify a cached value, not a copy of it.
14912 (pointer_query::dump): New function.
14913 (compute_objsize_r): Avoid overwriting access_ref::bndrng. Cache
14915 * pointer-query.h (pointer_query::dump): Declare.
14916 * tree-ssa-strlen.c (get_range): Simplify. Use function query.
14917 (dump_strlen_info): Use function query.
14918 (printf_strlen_execute): Factor code out into pointer_query::put_ref.
14920 2021-09-01 Thomas Schwinge <thomas@codesourcery.com>
14922 * tree.c (walk_tree_1) <OMP_CLAUSE>: Simplify.
14924 2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
14926 * doc/extend.texi: Document unavailable attribute.
14927 * print-tree.c (print_node): Handle unavailable attribute.
14928 * tree-core.h (struct tree_base): Add a bit to carry unavailability.
14929 * tree.c (error_unavailable_use): New.
14930 * tree.h (TREE_UNAVAILABLE): New.
14931 (error_unavailable_use): New.
14933 2021-09-01 Jakub Jelinek <jakub@redhat.com>
14935 PR tree-optimization/102124
14936 * tree-vect-patterns.c (vect_recog_widen_op_pattern): For ORIG_CODE
14937 MINUS_EXPR, if itype is unsigned with smaller precision than type,
14938 add an extra cast to signed variant of itype to ensure sign-extension.
14940 2021-09-01 Martin Liska <mliska@suse.cz>
14942 * graph.c (draw_cfg_node_succ_edges): Do not color fallthru
14943 edges and rather use colors for TRUE and FALSE edges.
14945 2021-09-01 Richard Biener <rguenther@suse.de>
14947 PR tree-optimization/93491
14948 * tree-ssa-pre.c (compute_avail): Set BB_MAY_NOTRETURN
14949 after processing the stmt itself. Do not consider
14950 pure functions possibly not returning. Properly avoid
14951 adding possibly trapping calls to EXP_GEN when there's
14952 a preceeding possibly not returning call.
14953 * tree-ssa-sccvn.c (vn_reference_may_trap): Conservatively
14956 2021-09-01 Richard Biener <rguenther@suse.de>
14958 PR tree-optimization/102139
14959 * tree-vectorizer.h (vec_base_alignments): Adjust hash-map
14960 type to record a std::pair of the stmt-info and the innermost
14962 (dr_vec_info::group): New member.
14963 * tree-vect-data-refs.c (vect_record_base_alignment): Adjust.
14964 (vect_compute_data_ref_alignment): Verify the recorded
14965 base alignment can be used.
14966 (data_ref_pair): Remove.
14967 (dr_group_sort_cmp): Adjust.
14968 (vect_analyze_data_ref_accesses): Store the group-ID in the
14969 dr_vec_info and operate on a vector of dr_vec_infos.
14971 2021-09-01 YunQiang Su <yunqiang.su@cipunited.com>
14973 * read-md.c (md_reader::handle_enum): support value assignation.
14974 * doc/md.texi: record define_c_enum value assignation support.
14976 2021-09-01 Jakub Jelinek <jakub@redhat.com>
14978 PR tree-optimization/102141
14979 * gimple-ssa-store-merging.c (bswap_view_convert): Add BEFORE
14980 argument. If false, emit stmts after gsi instead of before, and
14982 (bswap_replace): Adjust callers. When converting output of bswap,
14983 emit VIEW_CONVERT prepratation stmts after a copy of gsi instead
14986 2021-09-01 liuhongt <hongtao.liu@intel.com>
14988 * emit-rtl.c (validate_subreg): Get rid of all float-int
14991 2021-09-01 liuhongt <hongtao.liu@intel.com>
14994 2021-08-30 liuhongt <hongtao.liu@intel.com>
14996 * expmed.c (extract_bit_field_1): Make sure we're playing with
14997 integral modes before call extract_integral_bit_field.
14998 (extract_integral_bit_field): Add a parameter of type
14999 scalar_int_mode which corresponds to of tmode.
15000 And call extract_and_convert_fixed_bit_field instead of
15001 extract_fixed_bit_field and convert_extracted_bit_field.
15002 (extract_and_convert_fixed_bit_field): New function, it's a
15003 combination of extract_fixed_bit_field and
15004 convert_extracted_bit_field.
15006 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
15008 * tree.c (walk_tree_1) <OMP_CLAUSE_TILE>: Handle three operands.
15010 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
15012 * omp-general.h (omp_is_reference): Rename to...
15013 (omp_privatize_by_reference): ... this. Adjust all users...
15014 * omp-general.c: ... here, ...
15015 * gimplify.c: ... here, ...
15016 * omp-expand.c: ... here, ...
15017 * omp-low.c: ... here.
15019 2021-08-31 Martin Sebor <msebor@redhat.com>
15021 * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Test
15022 pointer element for equality to zero, not that of the cotaining
15025 2021-08-31 Martin Sebor <msebor@redhat.com>
15027 * gcc-rich-location.h (gcc_rich_location): Make ctor explicit.
15029 2021-08-31 Martin Sebor <msebor@redhat.com>
15031 * function.h (function): Add comments.
15032 (get_range_query): Same. Add attribute returns nonnull.
15034 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
15036 * expr.c (convert_modes): Don't use subreg_promoted_mode on a
15037 SUBREG if it can't be guaranteed to a SUBREG_PROMOTED_VAR_P set.
15038 Instead use the standard (safer) is_a <scalar_int_mode> idiom.
15040 2021-08-31 Jeff Law <jlaw@localhost.localdomain>
15042 * config.gcc (cris-*-elf, cris-*-none): Remove dbxelf.h from
15044 (m32r-*-elf, m32rle-*-elf, m32r-*-linux): Likewise.
15045 (mn10300-*-*, am33_2.0-*-linux*): Likewise.
15046 (xtensa*-*-elf, xtensa*-*-linux, xtensa*-*-uclinux): Likewise.
15047 (m32c-*-elf*, m32c-*-rtems*): Likewise.
15048 * config/cris/cris.h (DBX_NO_XREFS): Remove.
15049 (DBX_CONTIN_LENGTH, DBX_CONTIN_CHAR): Likewise.
15050 * config/m32r/m32r.h (DBXOUT_SOURCE_LINE): Likewise.
15051 (DBX_DEBUGGING_INFO, DBX_CONTIN_LENGTH): Likewise.
15052 * config/mn10300/mn10300.h (DEFAULT_GDB_EXTENSIONS): Likewise.
15053 * config/mn10300/linux.h (DBX_REGISTER_NAMES): Likewise.
15055 2021-08-31 Marcel Vollweiler <marcel@codesourcery.com>
15057 * gimplify.c (gimplify_scan_omp_clauses): Error handling. 'ancestor' only
15058 allowed on target constructs and only with particular other clauses.
15059 * omp-expand.c (expand_omp_target): Output of 'sorry, not supported' if
15060 'ancestor' is used.
15061 * omp-low.c (check_omp_nesting_restrictions): Error handling. No nested OpenMP
15062 structs when 'ancestor' is used.
15063 (scan_omp_1_stmt): No usage of OpenMP runtime routines in a target region when
15064 'ancestor' is used.
15065 * tree-pretty-print.c (dump_omp_clause): Append 'ancestor'.
15066 * tree.h (OMP_CLAUSE_DEVICE_ANCESTOR): Define macro.
15068 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
15070 * expr.c (convert_modes): Preserve SUBREG_PROMOTED_VAR_P when
15071 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
15073 * simplify-rtx.c (simplify_unary_operation_1) [SIGN_EXTEND]:
15074 Likewise, preserve SUBREG_PROMOTED_VAR_P when creating a (wider)
15075 partial subreg from a SUBREG_PROMOTED_VAR_P subreg. Generate
15076 SIGN_EXTEND of the SUBREG_REG when a subreg would be paradoxical.
15077 [ZERO_EXTEND]: Likewise, preserve SUBREG_PROMOTED_VAR_P when
15078 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
15079 subreg. Generate ZERO_EXTEND of the SUBREG_REG when a subreg
15080 would be paradoxical.
15082 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
15084 * combine.c (combine_simplify_rtx): Avoid converting an explicit
15085 TRUNCATE into a lowpart SUBREG on !TRULY_NOOP_TRUNCATION targets.
15086 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
15088 2021-08-31 Richard Biener <rguenther@suse.de>
15090 PR tree-optimization/102142
15091 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Fix
15092 condition under which to unset the visited flag.
15094 2021-08-31 Richard Biener <rguenther@suse.de>
15096 PR middle-end/102129
15097 * tree-ssa-ter.c (find_replaceable_in_bb): Do not move
15098 possibly trapping expressions across calls.
15100 2021-08-31 Jakub Jelinek <jakub@redhat.com>
15102 PR tree-optimization/102134
15103 * tree-ssa-ccp.c (bit_value_binop) <case RSHIFT_EXPR>: If sgn is
15104 UNSIGNED and r1val | r1mask has MSB set, ensure lzcount doesn't
15107 2021-08-31 Andrew Pinski <apinski@marvell.com>
15110 * collect-utils.c (setup_signals): New declaration.
15111 * collect-utils.h (setup_signals): New function.
15112 * collect2.c (handler): Delete.
15113 (main): Instead of manually setting up the signals,
15114 just call setup_signals.
15115 * lto-wrapper.c (main): Likewise.
15117 2021-08-31 Andrew Pinski <apinski@marvell.com>
15120 * config/i386/i386-protos.h (x86_output_aligned_bss):
15121 Change align argument to unsigned type.
15122 (x86_elf_aligned_decl_common): Likewise.
15123 * config/i386/i386.c (x86_elf_aligned_decl_common): Likewise.
15124 (x86_output_aligned_bss): Likewise.
15126 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
15128 * config/mips/mips.c (mips_module_isa_name): New.
15129 mips_file_start: add .module mipsREV to all asm output
15131 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
15133 * config/mips/mips.h (struct mips_cpu_info): define enum mips_isa;
15134 use enum instead of int for 'isa' member.
15135 * config.gcc, config/mips/mips.c, config/mips/mips-cpus.def,
15136 config/mips/netbsd.h: replace hardcoded numbers with enum.
15138 2021-08-31 liuhongt <hongtao.liu@intel.com>
15140 * config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
15141 define_split to define_insn_and_split.
15142 (*avx2_eq<mode>3): Removed.
15143 (<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
15144 (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
15145 (*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
15147 (*avx2_gt<mode>3): Removed.
15148 (<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
15149 define_insn to define_expand, and adjust pattern.
15150 (UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.
15152 2021-08-30 David Malcolm <dmalcolm@redhat.com>
15155 * Makefile.in (ANALYZER_OBJS): Add analyzer/call-info.o.
15157 2021-08-30 Jason Merrill <jason@redhat.com>
15159 * doc/invoke.texi: Document -Wmissing-requires.
15161 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
15163 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Remove
15164 TARGET_EXTRA_BUILTINS guard.
15166 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
15168 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Change
15169 initialization of V2DI_type_node and unsigned_V2DI_type_node.
15171 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
15173 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Use the new
15174 decl when new_builtins_are_live.
15175 * config/rs6000/rs6000-builtin-new.def (__builtin_cfstring): New
15178 2021-08-30 Pat Haugen <pthaugen@linux.ibm.com>
15180 * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add
15181 OPTION_MASK_P10_FUSION_2STORE.
15182 (POWERPC_MASKS): Likewise.
15183 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
15184 store fusion for Power10.
15185 (is_fusable_store): New.
15186 (power10_sched_reorder): Likewise.
15187 (rs6000_sched_reorder): Do Power10 specific reordering.
15188 (rs6000_sched_reorder2): Likewise.
15189 * config/rs6000/rs6000.opt: Add new option.
15191 2021-08-30 Richard Biener <rguenther@suse.de>
15193 PR tree-optimization/102128
15194 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
15195 Move scanning for if-converted scalar code to the caller
15196 and instead delay clearing the visited flag for profitable
15198 (vect_slp_region): Cost all subgraphs before scheduling.
15199 For if-converted BB vectorization scan for scalar COND_EXPRs
15200 and do not vectorize if any found and the cost model is
15203 2021-08-30 Richard Biener <rguenther@suse.de>
15205 * common.opt (fexceptions): Mark
15206 EnabledBy(fnon-call-exceptions).
15207 * doc/invoke.texi (fnon-call-exceptions): Document this
15208 enables -fexceptions.
15210 2021-08-30 Sebastian Huber <sebastian.huber@embedded-brains.de>
15212 * tsystem.h (abort): Define abort() if inhibit_libc is defined and it
15213 is not already defined.
15215 2021-08-30 liuhongt <hongtao.liu@intel.com>
15217 * expmed.c (extract_bit_field_1): Make sure we're playing with
15218 integral modes before call extract_integral_bit_field.
15219 (extract_integral_bit_field): Add a parameter of type
15220 scalar_int_mode which corresponds to of tmode.
15221 And call extract_and_convert_fixed_bit_field instead of
15222 extract_fixed_bit_field and convert_extracted_bit_field.
15223 (extract_and_convert_fixed_bit_field): New function, it's a
15224 combination of extract_fixed_bit_field and
15225 convert_extracted_bit_field.
15227 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
15229 * config/darwin.c (darwin_libc_has_function): Do not run
15230 the checks for x86 or modern Darwin. Make sure that there
15231 is a value set for darwin_macosx_version_min before testing.
15233 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
15235 * config/i386/darwin.h (CLEAR_INSN_CACHE): New.
15237 2021-08-28 Jan Hubicka <hubicka@ucw.cz>
15239 * ipa-modref-tree.h (modref_access_node::merge): Break out
15240 logic combining offsets and logic merging ranges to ...
15241 (modref_access_node::combined_offsets): ... here
15242 (modref_access_node::update2): ... here
15243 (modref_access_node::closer_pair_p): New member function.
15244 (modref_access_node::forced_merge): New member function.
15245 (modre_ref_node::insert): Do merging when table is full.
15247 2021-08-28 YunQiang Su <yunqiang.su@cipunited.com>
15250 * config.gcc: MIPS: use N64 ABI by default if the triple end
15251 with -gnuabi64, which is used by Debian since 2013.
15253 2021-08-28 Alexandre Oliva <oliva@adacore.com>
15255 * ipa-modref.c (analyze_function): Skip debug stmts.
15256 * tree-inline.c (estimate_num_insn): Consider builtins even
15257 without a cgraph_node.
15259 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
15261 * config/h8300/bitfield.md (cstore<mode>4): Remove expander.
15262 * config/h8300/h8300.c (h8300_expand_branch): Remove function.
15263 * config/h8300/h8300-protos.h (h8300_expadn_branch): Remove prototype.
15264 * config/h8300/h8300.md (eqne): New code iterator.
15265 (geultu, geultu_to_c): Similarly.
15266 * config/h8300/testcompare.md (cstore<mode>4): Dummy expander.
15267 (store_c_<mode>, store_c_i_<mode>): New define_insn_and_splits
15268 (cmp<mode>_c): New pattern
15270 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
15272 * tree-ssa-dom.c (reduce_vector_comparison_to_scalar_comparison): New
15274 (dom_opt_dom_walker::optimize_stmt): Use it.
15276 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
15278 * config/darwin.c (finalize_ctors): Add a section-start linker-
15280 (finalize_dtors): Likewise.
15281 * config/darwin.h (MIN_LD64_INIT_TERM_START_LABELS): New.
15283 2021-08-27 Bill Schmidt <wschmidt@linux.ibm.com>
15285 * config/rs6000/rs6000-call.c (rs6000-builtins.h): New #include.
15286 (rs6000_init_builtins): Call rs6000_init_generated_builtins. Skip the
15287 old initialization logic when new builtins are enabled.
15288 * config/rs6000/rs6000-gen-builtins.c (write_decls): Rename
15289 rs6000_autoinit_builtins to rs6000_init_generated_builtins.
15290 (write_init_file): Likewise.
15292 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
15294 * configure.ac (darwin2[[0-9]]* | darwin19*): Alter use of
15295 gcc_GAS_CHECK_FEATURE to remove an extraneous parameter.
15296 (amdgcn-* | gcn-*) Likewise.
15298 2021-08-27 Anthony Sharp <anthonysharp15@gmail.com>
15300 * symbol-summary.h: Added missing template keyword.
15302 2021-08-27 Richard Biener <rguenther@suse.de>
15304 PR tree-optimization/45178
15305 * tree-ssa-dce.c (find_obviously_necessary_stmts): For
15306 infinite loops without exit do not mark control dependent
15307 edges of the latch necessary.
15309 2021-08-27 konglin1 <lingling.kong@intel.com>
15312 * config/i386/sse.md: (<avx512>scattersi<mode>): Add mask operand to
15314 (<avx512>scattersi<mode>): Likewise.
15315 (*avx512f_scattersi<VI48F:mode>): Merge mask operand to set_dest.
15316 (*avx512f_scatterdi<VI48F:mode>): Likewise
15318 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
15320 * config/rs6000/rs6000.c (rs6000_builtin_md_vectorized_function): Add
15321 support for built-in functions MISC_BUILTIN_DIVWE, MISC_BUILTIN_DIVWEU,
15322 MISC_BUILTIN_DIVDE, MISC_BUILTIN_DIVDEU, P10_BUILTIN_CFUGED,
15323 P10_BUILTIN_CNTLZDM, P10_BUILTIN_CNTTZDM, P10_BUILTIN_PDEPD and
15324 P10_BUILTIN_PEXTD on Power10.
15326 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
15328 * config/rs6000/rs6000-call.c (builtin_function_type): Add unsigned
15329 signedness for some Power10 bifs.
15331 2021-08-27 David Edelsohn <dje.gcc@gmail.com>
15334 * config/rs6000/rs6000.c (rs6000_adjust_field_align): Use
15335 computed alignment if the entire struct has attribute packed.
15337 2021-08-27 liuhongt <hongtao.liu@intel.com>
15341 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
15342 IX86_BUILTIN_SHUFPD512, IX86_BUILTIN_SHUFPS512,
15343 IX86_BUILTIN_SHUFPD256, IX86_BUILTIN_SHUFPS,
15344 IX86_BUILTIN_SHUFPS256.
15345 (ix86_masked_all_ones): New function.
15347 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
15349 * config/i386/i386.md (*btr<mode>_1): Call force_reg unconditionally.
15350 (conditional moves with memory inputs splitters): Ditto.
15351 * config/i386/sse.md (one_cmpl<mode>2): Simplify.
15353 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
15355 * ipa-modref-tree.h (modref_access_node::try_merge_with): Restart
15356 search after merging.
15358 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
15360 * config/rs6000/rs6000-overload.def: Add remaining overloads.
15362 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
15364 * config/rs6000/rs6000-builtin-new.def: Add cell stanza.
15366 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
15368 * config/rs6000/rs6000-builtin-new.def: Add ieee128-hw, dfp,
15369 crypto, and htm stanzas.
15371 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
15373 * config/rs6000/rs6000-builtin-new.def: Add mma stanza.
15375 2021-08-26 Martin Sebor <msebor@redhat.com>
15377 * tree-ssa-uninit.c (warn_uninit): Refactor and simplify.
15378 (warn_uninit_phi_uses): Remove argument from calls to warn_uninit.
15379 (warn_uninitialized_vars): Same. Reduce visibility of locals.
15380 (warn_uninitialized_phi): Same.
15382 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
15384 * tree-ssa-ccp.c (get_individual_bits): Helper function to
15385 extract the individual bits from a widest_int constant (mask).
15386 (gray_code_bit_flips): New read-only table for effiently
15387 enumerating permutations/combinations of bits.
15388 (bit_value_binop) [LROTATE_EXPR, RROTATE_EXPR]: Handle rotates
15389 by unknown counts that are guaranteed less than the target
15390 precision and four or fewer unknown bits by enumeration.
15391 [LSHIFT_EXPR, RSHIFT_EXPR]: Likewise, also handle shifts by
15392 enumeration under the same conditions. Handle remaining
15393 shifts as a mask based upon the minimum possible shift value.
15395 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
15396 Richard Biener <rguenther@suse.de>
15398 * match.pd (shift transformations): Remove a redundant
15399 !POINTER_TYPE_P check.
15401 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
15404 * config/i386/i386.md (cmove reg-reg move elimination peephole2s):
15405 Set all_regs to true in the call to replace_rtx.
15407 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
15409 * ipa-modref-tree.c (test_insert_search_collapse): Update test.
15410 * ipa-modref-tree.h (modref_base_node::insert): Be smarter when
15411 hiting --param modref-max-refs limit.
15412 (modref_tree:insert_base): Be smarter when hitting
15413 --param modref-max-bases limit. Add new parameter REF.
15414 (modref_tree:insert): Update.
15415 (modref_tree:merge): Update.
15416 * ipa-modref.c (read_modref_records): Update.
15418 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
15420 * params.opt: (modref-max-adjustments): Add full stop.
15422 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
15424 * ipa-modref-tree.h (modref_ref_node::verify): New member
15426 (modref_ref_node::insert): Use it.
15427 (modref_ref_node::try_mere_with): Fix off by one error.
15429 2021-08-26 Martin Liska <mliska@suse.cz>
15430 Stefan Kneifel <stefan.kneifel@bluewin.ch>
15432 * cgraph.h (create_version_clone_with_body): Add new parameter.
15433 * cgraphclones.c: Likewise.
15434 * multiple_target.c (create_dispatcher_calls): Do not use
15436 (create_target_clone): Likewise here.
15438 2021-08-26 Jonathan Yong <10walls@gmail.com>
15440 * doc/extend.texi: Add note about reserved priorities
15441 to the constructor attribute.
15443 2021-08-25 Martin Sebor <msebor@redhat.com>
15445 * gimple-range-cache.cc (ssa_global_cache::dump): Avoid printing
15446 range table header alone.
15447 * gimple-range.cc (gimple_ranger::export_global_ranges): Same.
15449 2021-08-25 Jan Hubicka <hubicka@ucw.cz>
15451 * doc/invoke.texi: Document --param modref-max-adjustments.
15452 * ipa-modref-tree.c (test_insert_search_collapse): Update.
15453 (test_merge): Update.
15454 * ipa-modref-tree.h (struct modref_access_node): Add adjustments;
15455 (modref_access_node::operator==): Fix handling of access ranges.
15456 (modref_access_node::contains): Constify parameter; handle also
15457 mismatched parm offsets.
15458 (modref_access_node::update): New function.
15459 (modref_access_node::merge): New function.
15460 (unspecified_modref_access_node): Update constructor.
15461 (modref_ref_node::insert_access): Add record_adjustments parameter;
15463 (modref_ref_node::try_merge_with): New private function.
15464 (modref_tree::insert): New record_adjustments parameter.
15465 (modref_tree::merge): New record_adjustments parameter.
15466 (modref_tree::copy_from): Update.
15467 * ipa-modref.c (dump_access): Dump adjustments field.
15468 (get_access): Update constructor.
15469 (record_access): Update call of insert.
15470 (record_access_lto): Update call of insert.
15471 (merge_call_side_effects): Add record_adjustments parameter.
15472 (get_access_for_fnspec): Update.
15473 (process_fnspec): Update.
15474 (analyze_call): Update.
15475 (analyze_function): Update.
15476 (read_modref_records): Update.
15477 (ipa_merge_modref_summary_after_inlining): Update.
15478 (propagate_unknown_call): Update.
15479 (modref_propagate_in_scc): Update.
15480 * params.opt (param-max-modref-adjustments=): New.
15482 2021-08-25 Michael Meissner <meissner@linux.ibm.com>
15484 * config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
15486 (xxspltiw_v4si): Use vecperm type attribute.
15487 (xxspltiw_v4si_inst): Use vecperm type attribute.
15488 (xxspltiw_v4sf_inst): Likewise.
15489 (xxspltidp_v2df): Use vecperm type attribute. Use
15490 UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
15491 (xxspltidp_v2df_inst): Likewise.
15492 (xxsplti32dx_v4si): Use vecperm type attribute.
15493 (xxsplti32dx_v4si_inst): Likewise.
15494 (xxsplti32dx_v4sf_inst): Likewise.
15495 (xxblend_<mode>): Likewise.
15496 (xxpermx): Likewise.
15497 (xxpermx_inst): Likewise.
15498 (xxeval): Likewise.
15500 2021-08-25 Lewis Hyatt <lhyatt@gmail.com>
15503 * coretypes.h (typedef diagnostic_input_charset_callback): Declare.
15504 * diagnostic.c (diagnostic_initialize_input_context): New function.
15505 * diagnostic.h (diagnostic_initialize_input_context): Declare.
15506 * input.c (default_charset_callback): New function.
15507 (file_cache::initialize_input_context): New function.
15508 (file_cache_slot::create): Added ability to convert the input
15509 according to the input context.
15510 (file_cache::file_cache): Initialize the new input context.
15511 (class file_cache_slot): Added new m_alloc_offset member.
15512 (file_cache_slot::file_cache_slot): Initialize the new member.
15513 (file_cache_slot::~file_cache_slot): Handle potentially offset buffer.
15514 (file_cache_slot::maybe_grow): Likewise.
15515 (file_cache_slot::needs_read_p): Handle NULL fp, which is now possible.
15516 (file_cache_slot::get_next_line): Likewise.
15517 * input.h (class file_cache): Added input context member.
15519 2021-08-25 Richard Biener <rguenther@suse.de>
15521 PR tree-optimization/102046
15522 * tree-vect-slp.c (vect_build_slp_tree_2): Conservatively
15523 update ->any_pattern when swapping operands.
15525 2021-08-25 Hongyu Wang <hongyu.wang@intel.com>
15528 * config/i386/i386.c (ix86_live_on_entry): Adjust comment.
15529 (ix86_decompose_address): Remove retval check for ASHIFT,
15530 allow non-canonical zero extend if AND mask covers ASHIFT
15532 (ix86_legitimate_address_p): Adjust condition for decompose.
15533 (ix86_rtx_costs): Adjust cost for lea with non-canonical
15535 Co-Authored by: Uros Bizjak <ubizjak@gmail.com>
15537 2021-08-25 Jiufu Guo <guojiufu@linux.ibm.com>
15539 PR tree-optimization/101145
15540 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
15542 (number_of_iterations_lt): Invoke above function.
15543 (adjust_cond_for_loop_until_wrap):
15544 Merge to number_of_iterations_until_wrap.
15545 (number_of_iterations_cond): Update invokes for
15546 adjust_cond_for_loop_until_wrap and number_of_iterations_lt.
15548 2021-08-25 konglin1 <lingling.kong@intel.com>
15551 * config/i386/avx512dqintrin.h (_mm512_fpclass_ps_mask): Fix
15552 macro define in O0.
15553 (_mm512_mask_fpclass_ps_mask): Ditto.
15555 2021-08-25 Kewen Lin <linkw@linux.ibm.com>
15557 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Remove.
15558 (vec_unpacku_hi_v8hi): Likewise.
15559 (vec_unpacku_lo_v16qi): Likewise.
15560 (vec_unpacku_lo_v8hi): Likewise.
15561 (vec_unpacku_hi_<VP_small_lc>): New define_expand.
15562 (vec_unpacku_lo_<VP_small_lc>): Likewise.
15564 2021-08-24 David Edelsohn <dje.gcc@gmail.com>
15566 * config/rs6000/aix.h (SYSTEM_IMPLICIT_EXTERN_C): Delete.
15567 * config/rs6000/aix71.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
15568 * config/rs6000/aix72.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
15569 * config/rs6000/aix73.h (TARGET_AIX_VERSION): Increase to 73.
15571 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
15573 PR middle-end/102031
15574 * simplify-rtx.c (simplify_truncation): When comparing precisions
15575 use "subreg_prec" variable, not "subreg_mode".
15577 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
15579 * config/rs6000/rs6000-builtin-new.def: Add power10 and power10-64
15582 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
15584 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Initialize
15585 various pointer type nodes.
15586 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add enum
15587 values for various pointer types.
15588 (ptr_V16QI_type_node): New macro.
15589 (ptr_V1TI_type_node): New macro.
15590 (ptr_V2DI_type_node): New macro.
15591 (ptr_V2DF_type_node): New macro.
15592 (ptr_V4SI_type_node): New macro.
15593 (ptr_V4SF_type_node): New macro.
15594 (ptr_V8HI_type_node): New macro.
15595 (ptr_unsigned_V16QI_type_node): New macro.
15596 (ptr_unsigned_V1TI_type_node): New macro.
15597 (ptr_unsigned_V8HI_type_node): New macro.
15598 (ptr_unsigned_V4SI_type_node): New macro.
15599 (ptr_unsigned_V2DI_type_node): New macro.
15600 (ptr_bool_V16QI_type_node): New macro.
15601 (ptr_bool_V8HI_type_node): New macro.
15602 (ptr_bool_V4SI_type_node): New macro.
15603 (ptr_bool_V2DI_type_node): New macro.
15604 (ptr_bool_V1TI_type_node): New macro.
15605 (ptr_pixel_type_node): New macro.
15606 (ptr_intQI_type_node): New macro.
15607 (ptr_uintQI_type_node): New macro.
15608 (ptr_intHI_type_node): New macro.
15609 (ptr_uintHI_type_node): New macro.
15610 (ptr_intSI_type_node): New macro.
15611 (ptr_uintSI_type_node): New macro.
15612 (ptr_intDI_type_node): New macro.
15613 (ptr_uintDI_type_node): New macro.
15614 (ptr_intTI_type_node): New macro.
15615 (ptr_uintTI_type_node): New macro.
15616 (ptr_long_integer_type_node): New macro.
15617 (ptr_long_unsigned_type_node): New macro.
15618 (ptr_float_type_node): New macro.
15619 (ptr_double_type_node): New macro.
15620 (ptr_long_double_type_node): New macro.
15621 (ptr_dfloat64_type_node): New macro.
15622 (ptr_dfloat128_type_node): New macro.
15623 (ptr_ieee128_type_node): New macro.
15624 (ptr_ibm128_type_node): New macro.
15625 (ptr_vector_pair_type_node): New macro.
15626 (ptr_vector_quad_type_node): New macro.
15627 (ptr_long_long_integer_type_node): New macro.
15628 (ptr_long_long_unsigned_type_node): New macro.
15630 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
15632 * config/rs6000/rs6000-builtin-new.def: Add power9-vector, power9,
15633 and power9-64 stanzas.
15635 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
15636 Tom de Vries <tdevries@suse.de>
15638 * config.gcc (nvptx-*-*): Define {c,c++}_target_objs.
15639 * config/nvptx/nvptx-protos.h (nvptx_cpu_cpp_builtins): Prototype.
15640 * config/nvptx/nvptx.h (TARGET_CPU_CPP_BUILTINS): Implement with
15641 a call to the new nvptx_cpu_cpp_builtins function in nvptx-c.c.
15642 * config/nvptx/t-nvptx (nvptx-c.o): New rule.
15643 * config/nvptx/nvptx-c.c: New source file.
15644 (nvptx_cpu_cpp_builtins): Move implementation here.
15646 2021-08-24 Martin Sebor <msebor@redhat.com>
15648 PR middle-end/101600
15649 PR middle-end/101977
15650 * gimple-ssa-warn-access.cc (maybe_warn_for_bound): Tighten up
15651 the phrasing of a warning.
15652 (check_access): Use the remaining size after subtracting any offset
15653 rather than the whole object size.
15654 * pointer-query.cc (access_ref::get_ref): Clear BASE0 flag if it's
15655 clear for any nonnull PHI argument.
15656 (compute_objsize): Clear argument.
15658 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
15660 * config/rs6000/rs6000-builtin-new.def: Add power8-vector stanza.
15662 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
15664 * config/rs6000/rs6000-builtin-new.def: Add power7 and power7-64
15667 2021-08-24 Andrew MacLeod <amacleod@redhat.com>
15669 * value-relation.cc (rr_transitive_table): New.
15670 (relation_transitive): New.
15671 (value_relation::swap): Remove.
15672 (value_relation::apply_transitive): New.
15673 (relation_oracle::relation_oracle): Allocate a new tmp bitmap.
15674 (relation_oracle::register_relation): Call register_transitives.
15675 (relation_oracle::register_transitives): New.
15676 * value-relation.h (relation_oracle): Add new temporary bitmap and
15679 2021-08-24 H.J. Lu <hjl.tools@gmail.com>
15682 * config/i386/i386-expand.c (ix86_expand_vector_move): Broadcast
15683 from integer to a pseudo vector register.
15685 2021-08-24 Richard Biener <rguenther@suse.de>
15687 PR tree-optimization/100089
15688 * tree-vectorizer.h (vect_slp_bb): Rename to ...
15689 (vect_slp_if_converted_bb): ... this and get the original
15690 loop as new argument.
15691 * tree-vectorizer.c (try_vectorize_loop_1): Revert previous fix,
15692 pass original loop to vect_slp_if_converted_bb.
15693 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
15694 If orig_loop was passed scan the not vectorized stmts
15695 for COND_EXPRs and force not profitable if found.
15696 (vect_slp_region): Pass down all SLP instances to costing
15697 if orig_loop was specified.
15698 (vect_slp_bbs): Pass through orig_loop.
15699 (vect_slp_bb): Rename to ...
15700 (vect_slp_if_converted_bb): ... this and get the original
15701 loop as new argument.
15702 (vect_slp_function): Adjust.
15704 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
15707 * config/arm/arm.md (attribute arch): Add fix_vlldm.
15708 (arch_enabled): Use it.
15709 * config/arm/vfp.md (lazy_store_multiple_insn): Add alternative to
15710 use when erratum mitigation is needed.
15712 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
15715 * config/arm/arm.opt (mfix-cmse-cve-2021-35465): New option.
15716 * doc/invoke.texi (Arm Options): Document it.
15717 * config/arm/arm-cpus.in (quirk_vlldm): New feature bit.
15718 (ALL_QUIRKS): Add quirk_vlldm.
15719 (cortex-m33): Add quirk_vlldm.
15720 (cortex-m35p, cortex-m55): Likewise.
15721 * config/arm/arm.c (arm_option_override): Enable fix_vlldm if
15722 targetting an affected CPU and not explicitly controlled on
15725 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
15727 * config/arm/vfp.md (lazy_store_multiple_insn): Rewrite as valid RTL.
15728 (lazy_load_multiple_insn): Likewise.
15730 2021-08-24 liuhongt <hongtao.liu@intel.com>
15733 * config/i386/sse.md (<avx512>_vternlog<mode><sd_maskz_name>):
15734 Enable avx512 embedded broadcast.
15735 (*<avx512>_vternlog<mode>_all): Ditto.
15736 (<avx512>_vternlog<mode>_mask): Ditto.
15738 2021-08-24 liuhongt <hongtao.liu@intel.com>
15741 * config/i386/i386.c (ix86_rtx_costs): Define cost for
15743 * config/i386/i386.h (STRIP_UNARY): New macro.
15744 * config/i386/predicates.md (reg_or_notreg_operand): New
15746 * config/i386/sse.md (*<avx512>_vternlog<mode>_all): New define_insn.
15747 (*<avx512>_vternlog<mode>_1): New pre_reload
15748 define_insn_and_split.
15749 (*<avx512>_vternlog<mode>_2): Ditto.
15750 (*<avx512>_vternlog<mode>_3): Ditto.
15751 (any_logic1,any_logic2): New code iterator.
15752 (logic_op): New code attribute.
15753 (ternlogsuffix): Extend to VNxDF and VNxSF.
15755 2021-08-24 Richard Biener <rguenther@suse.de>
15757 * doc/invoke.texi (vect-inner-loop-cost-factor): Adjust.
15758 * params.opt (--param vect-inner-loop-cost-factor): Adjust
15760 * tree-vect-loop.c (vect_analyze_loop_form): Initialize
15761 inner_loop_cost_factor to the minimum of the estimated number
15762 of iterations of the inner loop and vect-inner-loop-cost-factor.
15764 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
15765 Richard Biener <rguenther@suse.de>
15767 * config/i386/i386-features.c (compute_convert_gain): Provide
15768 more accurate values for CONST_INT, when optimizing for size.
15769 * config/i386/i386.c (COSTS_N_BYTES): Move definition from here...
15770 * config/i386/i386.h (COSTS_N_BYTES): to here.
15772 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
15773 Jakub Jelinek <jakub@redhat.com>
15775 PR middle-end/102029
15776 * match.pd (shift transformations): Add an additional check for
15777 !POINTER_TYPE_P in the recently added left shift transformation.
15779 2021-08-24 liuhongt <hongtao.liu@intel.com>
15781 PR tree-optimization/100089
15782 * tree-vectorizer.c (try_vectorize_loop_1): Disable slp in
15783 loop vectorizer when cost model is very-cheap.
15785 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
15787 * config/rs6000/rs6000-gen-builtins.c (parse_bif_entry): Don't call
15788 asprintf, which is not available on AIX.
15790 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
15792 * config.gcc (target_gtfiles): Add ./rs6000-builtins.h.
15793 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Set.
15795 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
15797 * config.gcc (powerpc*-*-*): Add rs6000-builtins.o to extra_objs.
15798 * config/rs6000/rs6000-gen-builtins.c (main): Close init_file
15800 * config/rs6000/t-rs6000 (rs6000-gen-builtins.o): New target.
15801 (rbtree.o): Likewise.
15802 (rs6000-gen-builtins): Likewise.
15803 (rs6000-builtins.c): Likewise.
15804 (rs6000-builtins.h): Likewise.
15805 (rs6000.o): Add dependency.
15806 (EXTRA_HEADERS): Add rs6000-vecdefines.h.
15807 (rs6000-vecdefines.h): New target.
15808 (rs6000-builtins.o): Likewise.
15809 (rs6000-call.o): Add rs6000-builtins.h as a dependency.
15810 (rs6000-c.o): Likewise.
15812 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
15815 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace):
15816 Diagnose buffer overrun.
15817 (safe_inc_pos): Fix overrun detection.
15818 (match_identifier): Diagnose buffer overrun.
15819 (match_integer): Likewise.
15820 (match_to_right_bracket): Likewise.
15822 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
15824 * ipa-modref-tree.h (modref_access_node::range_info_useful_p):
15825 Improve range compare.
15826 (modref_access_node::contains): New member function.
15827 (modref_access_node::search): Remove.
15828 (modref_access_node::insert): Be smarter about subaccesses.
15830 2021-08-23 Thomas Schwinge <thomas@codesourcery.com>
15832 * config/i386/i386-options.c (ix86_omp_device_kind_arch_isa)
15833 <omp_device_arch> [ACCEL_COMPILER]: Match "intel_mic".
15834 * config/i386/t-omp-device (omp-device-properties-i386) <arch>:
15837 2021-08-23 Jeff Law <jlaw@localhost.localdomain>
15839 * config/h8300/h8300-protos.h (h8300_expand_epilogue): Add new
15841 * config/h8300/jumpcall.md (call, call_value): Restrict to
15842 !SIBLING_CALL_P cases.
15843 (subcall, sibcall_value): New patterns & expanders.
15844 * config/h8300/proepi.md (epilogue): Pass new argument to
15845 h8300_expand_epilogue.
15846 (sibcall_epilogue): New expander.
15847 * config/h8300/h8300.c (h8300_expand_epilogue): Handle sibcall
15849 (h8300_ok_for_sibcall_p): New function.
15850 (TARGET_FUNCTION_OK_FOR_SIBCALL): define.
15852 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
15854 * simplify-rtx.c (simplify_unary_operation_1): [TRUNCATE]:
15855 Handle case where the operand is already the desired mode.
15857 2021-08-23 Richard Biener <rguenther@suse.de>
15860 * tree-ssa-structalias.c (ipa_pta_execute): Check in_other_partition
15861 in addition to has_gimple_body.
15863 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
15865 PR middle-end/101949
15866 * ipa-modref.c (analyze_ssa_name_flags): Fix merging of
15869 2021-08-23 Martin Liska <mliska@suse.cz>
15871 * doc/invoke.texi: Put the option out of -mxl-mode-app-model
15874 2021-08-23 Richard Biener <rguenther@suse.de>
15876 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
15877 Properly scale the inner loop cost only once.
15879 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
15881 * tree-ssa-ccp.c (bit_value_binop) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]:
15882 Provide bounds for unsigned (and signed with non-negative operands)
15883 division and modulus.
15885 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
15887 * simplify-rtx.c (simplify_truncation): Generalize simplification
15888 of (truncate:A (subreg:B X)).
15889 (simplify_unary_operation_1) [FLOAT_TRUNCATE, FLOAT_EXTEND,
15890 SIGN_EXTEND, ZERO_EXTEND]: Handle cases where the operand
15891 already has the desired machine mode.
15892 (test_scalar_int_ops): Add tests that useless extensions and
15893 truncations are optimized away.
15894 (test_scalar_int_ext_ops): New self-test function to confirm
15895 that truncations of extensions are correctly simplified.
15896 (test_scalar_int_ext_ops2): New self-test function to check
15897 truncations of truncations, extensions of extensions, and
15898 truncations of extensions.
15899 (test_scalar_ops): Call the above two functions with a
15900 representative sampling of integer machine modes.
15902 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
15904 * match.pd (shift transformations): Change the sign of an
15905 LSHIFT_EXPR if it reduces the number of explicit conversions.
15907 2021-08-23 Jakub Jelinek <jakub@redhat.com>
15909 PR tree-optimization/86723
15910 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
15911 cast64_to_32 argument, set *cast64_to_32 to false, unless n is
15912 non-memory permutation of 64-bit src which only has bytes of
15913 0 or [5..8] and n->range is 4.
15914 (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
15915 find_bswap_or_nop_finalize caller, support bswap with some bytes
15916 zeroed, as long as at least two bytes are not zeroed.
15917 (bswap_replace): Add mask argument and handle masking of bswap
15919 (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
15920 caller, punt if cast64_to_32 or mask is not all ones.
15921 (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
15922 caller, for now punt if cast64_to_32.
15924 2021-08-23 Richard Biener <rguenther@suse.de>
15926 PR tree-optimization/79334
15927 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Record
15928 a type also for COMPONENT_REFs.
15929 (vn_reference_may_trap): Check ARRAY_REF with constant index
15930 against the array domain.
15932 2021-08-23 liuhongt <hongtao.liu@intel.com>
15935 * config/i386/sse.md (*avx512f_pshufb_truncv8hiv8qi_1): Add
15936 TARGET_AVX512BW to condition.
15938 2021-08-23 Jakub Jelinek <jakub@redhat.com>
15941 * dwarf2out.c (gen_variable_die): Add DW_AT_location for global
15942 register variables already during early_dwarf if possible.
15944 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
15946 * config/arm/arm_mve.h: Fix __arm_vctp16q return type.
15948 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
15951 * config/arm/arm.opt: Fix typo.
15952 * config/arm/t-rmprofile: Fix typo.
15954 2021-08-23 Jakub Jelinek <jakub@redhat.com>
15956 * tree.h (OMP_CLAUSE_GRAINSIZE_STRICT): Define.
15957 (OMP_CLAUSE_NUM_TASKS_STRICT): Define.
15958 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_GRAINSIZE,
15959 case OMP_CLAUSE_NUM_TASKS>: Print strict: modifier.
15960 * omp-expand.c (expand_task_call): Use GOMP_TASK_FLAG_STRICT in iflags
15961 if either grainsize or num_tasks clause has the strict modifier.
15963 2021-08-23 Martin Liska <mliska@suse.cz>
15965 * dbgcnt.def (DEBUG_COUNTER): New counter.
15966 * gimple.c (gimple_call_arg_flags): Use it in IPA PTA.
15968 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
15970 * ipa-modref.c (analyze_ssa_name_flags): Improve handling of return slot.
15972 2021-08-23 Xi Ruoyao <xry111@mengyan1223.wang>
15975 * config/mips/mips-protos.h (mips_msa_output_shift_immediate):
15977 * config/mips/mips.c (mips_msa_output_shift_immediate): New
15979 * config/mips/mips-msa.md (vashl<mode>3, vashr<mode>3,
15980 vlshr<mode>3): Call it.
15982 2021-08-22 Jan Hubicka <hubicka@ucw.cz>
15983 Martin Liska <mliska@suse.cz>
15985 PR middle-end/101949
15986 * ipa-modref.c (analyze_ssa_name_flags): Indirect call implies
15989 2021-08-21 Dragan Mladjenovic <OT_Dragan.Mladjenovic@mediatek.com>
15991 * config/mips/mips.c (mips_function_rodata_section,
15992 TARGET_ASM_FUNCTION_RODATA_SECTION): Removed.
15994 2021-08-21 John David Anglin <danglin@gcc.gnu.org>
15996 * config/pa/pa.c (pa_asm_output_aligned_common): Remove warning.
15998 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
16000 * configure.ac (thread-local storage support): Remove tls_first_major
16001 and tls_first_minor. Use "$conftest_s" to check support.
16002 * configure: Regenerate.
16004 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
16006 * configure.ac: Fixup formatting.
16008 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
16010 * acinclude.m4 (gcc_GAS_CHECK_FEATURE): Remove third argument and ...
16011 * configure.ac: ... update all callers.
16013 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
16016 * acinclude.m4 (_gcc_COMPUTE_GAS_VERSION, _gcc_GAS_VERSION_GTE_IFELSE)
16017 (gcc_GAS_VERSION_GTE_IFELSE): Remove.
16018 (gcc_GAS_CHECK_FEATURE): Do not handle in-tree case specially.
16019 * configure.ac: Remove gcc_cv_gas_major_version, gcc_cv_gas_minor_version.
16020 Remove remaining checks for in-tree assembler.
16021 * configure: Regenerate.
16023 2021-08-20 Jeff Law <jlaw@localhost.localdomain>
16025 * config/h8300/h8300.c (shift_alg_hi): Improve arithmetic shift right
16026 by 15 bits for H8/300H and H8/S. Improve logical shifts by 12
16028 (shift_alg_si): Improve arithmetic right shift by 28-30 bits for
16029 H8/300H. Improve arithmetic shift right by 15 bits for H8/S.
16030 Improve logical shifts by 27 bits for H8/S.
16031 (get_shift_alg): Corresponding changes.
16032 (h8300_option_override): Revert to loops for -Os when profitable.
16034 2021-08-20 Richard Biener <rguenther@suse.de>
16036 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
16038 (vect_analyze_data_ref_accesses): Likewise. Assign the BB
16039 index as group_id when dataref_groups were not computed.
16040 * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
16041 we advace to the next BB.
16043 2021-08-20 Jakub Jelinek <jakub@redhat.com>
16045 * omp-builtins.def (BUILT_IN_GOMP_WARNING, BUILT_IN_GOMP_ERROR): New
16048 2021-08-20 Martin Liska <mliska@suse.cz>
16050 PR gcov-profile/89961
16051 * gcov.c (make_gcov_file_name): Rewrite using std::string.
16052 (mangle_name): Simplify, do not used the second argument.
16053 (strip_extention): New function.
16054 (get_md5sum): Likewise.
16055 (get_gcov_intermediate_filename): Handle properly -p and -x
16057 (output_gcov_file): Use string type.
16058 (generate_results): Likewise.
16059 (md5sum_to_hex): Remove.
16061 2021-08-20 Michael Meissner <meissner@linux.ibm.com>
16063 * config/rs6000/altivec.md (UNSPEC_XXEVAL): Move to vsx.md.
16064 (UNSPEC_XXSPLTIW): Move to vsx.md.
16065 (UNSPEC_XXSPLTID): Move to vsx.md.
16066 (UNSPEC_XXSPLTI32DX): Move to vsx.md.
16067 (UNSPEC_XXBLEND): Move to vsx.md.
16068 (UNSPEC_XXPERMX): Move to vsx.md.
16069 (VM3): Move to vsx.md.
16070 (VM3_char): Move to vsx.md.
16071 (xxspltiw_v4si): Move to vsx.md.
16072 (xxspltiw_v4sf): Move to vsx.md.
16073 (xxspltiw_v4sf_inst): Move to vsx.md.
16074 (xxspltidp_v2df): Move to vsx.md.
16075 (xxspltidp_v2df_inst): Move to vsx.md.
16076 (xxsplti32dx_v4si_inst): Move to vsx.md.
16077 (xxsplti32dx_v4sf): Move to vsx.md.
16078 (xxsplti32dx_v4sf_inst): Move to vsx.md.
16079 (xxblend_<mode>): Move to vsx.md.
16080 (xxpermx): Move to vsx.md.
16081 (xxpermx_inst): Move to vsx.md.
16082 * config/rs6000/vsx.md (UNSPEC_XXEVAL): Move from altivec.md.
16083 (UNSPEC_XXSPLTIW): Move from altivec.md.
16084 (UNSPEC_XXSPLTID): Move from altivec.md.
16085 (UNSPEC_XXSPLTI32DX): Move from altivec.md.
16086 (UNSPEC_XXBLEND): Move from altivec.md.
16087 (UNSPEC_XXPERMX): Move from altivec.md.
16088 (VM3): Move from altivec.md.
16089 (VM3_char): Move from altivec.md.
16090 (xxspltiw_v4si): Move from altivec.md.
16091 (xxspltiw_v4sf): Move from altivec.md.
16092 (xxspltiw_v4sf_inst): Move from altivec.md.
16093 (xxspltidp_v2df): Move from altivec.md.
16094 (xxspltidp_v2df_inst): Move from altivec.md.
16095 (xxsplti32dx_v4si_inst): Move from altivec.md.
16096 (xxsplti32dx_v4sf): Move from altivec.md.
16097 (xxsplti32dx_v4sf_inst): Move from altivec.md.
16098 (xxblend_<mode>): Move from altivec.md.
16099 (xxpermx): Move from altivec.md.
16100 (xxpermx_inst): Move from altivec.md.
16102 2021-08-19 Roger Sayle <roger@nextmovesoftware.com>
16104 * tree-vect-generic.c (expand_vector_operations_1): Use either
16105 gimplify_build1 or gimplify_build2 instead of gimple_build_assign
16106 when constructing scalar splat expressions.
16108 2021-08-19 Peter Bergner <bergner@linux.ibm.com>
16111 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Cast
16112 pointer to __vector_pair *.
16114 2021-08-19 Martin Sebor <msebor@redhat.com>
16116 * gimple-range.cc: Add comments.
16117 * gimple-range.h: Same.
16119 2021-08-19 Martin Sebor <msebor@redhat.com>
16121 PR middle-end/101984
16122 * gimple-ssa-warn-access.cc (pass_waccess::execute): Also call
16125 2021-08-19 Jeff Law <jlaw@localhost.localdomain>
16127 * config.gcc (h8300-*-elf*): Do not include dbxelf.h.
16128 (h8300-*-linux*, v850-*-rtems*, v850*-elf*): Likewise.
16129 * config/v850/v850.h (DEFAULT_GDB_EXTENSIONS): Remove.
16131 2021-08-19 Jakub Jelinek <jakub@redhat.com>
16133 PR middle-end/101950
16134 * optabs.c (expand_clrsb_using_clz): New function.
16135 (expand_unop): Use it as another clrsb expansion fallback.
16137 2021-08-19 liuhongt <hongtao.liu@intel.com>
16140 2021-07-28 liuhongt <hongtao.liu@intel.com>
16143 * config/i386/i386.h (processor_costs): Add new member
16145 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
16146 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
16147 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
16148 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
16149 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
16150 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
16151 generic_cost, core_cost): Initialize integer_to_sse same value
16153 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
16154 * config/i386/i386.c (ix86_builtin_vectorization_cost):
16155 Use integer_to_sse instead of sse_op to calculate the cost of
16158 2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
16160 * config.gcc: Include rpath.opt for Darwin.
16161 * config/darwin.h (DRIVER_SELF_SPECS): Handle -rpath.
16163 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
16165 PR bootstrap/101959
16166 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
16169 2021-08-18 Jonathan Wright <jonathan.wright@arm.com>
16171 * config/aarch64/arm_neon.h (vld3_lane_f64): Use float RTL
16172 pattern and type cast.
16173 (vld4_lane_f32): Use float RTL pattern.
16174 (vld4q_lane_f64): Use float type cast.
16176 2021-08-18 Jan Hubicka <hubicka@ucw.cz>
16178 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Check also
16181 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
16183 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor): Extend.
16184 (test_map_of_type_with_ctor_and_dtor_expand): Add function.
16185 (hash_map_tests_c_tests): Call it.
16187 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
16189 * ggc.h (enum ggc_collect): New.
16190 (ggc_collect): Use it.
16191 * ggc-page.c: Adjust.
16192 * ggc-common.c: Likewise.
16193 * ggc-tests.c: Likewise.
16194 * read-rtl-function.c: Likewise.
16195 * selftest-run-tests.c: Likewise.
16196 * doc/gty.texi (Invoking the garbage collector): Likewise.
16198 2021-08-18 liuhongt <hongtao.liu@intel.com>
16201 * config/i386/i386.h (TARGET_V2DF_REDUCTION_PREFER_HADDPD):
16203 * config/i386/sse.md (*sse3_haddv2df3_low): Add
16204 TARGET_V2DF_REDUCTION_PREFER_HADDPD.
16205 (*sse3_hsubv2df3_low): Ditto.
16206 * config/i386/x86-tune.def
16207 (X86_TUNE_V2DF_REDUCTION_PREFER_HADDPD): New tune.
16209 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
16211 * gimple-range-gori.cc (gori_compute::gori_compute): Enable tracing.
16212 (gori_compute::compute_operand_range): Add tracing.
16213 (gori_compute::logical_combine): Ditto.
16214 (gori_compute::compute_logical_operands): Ditto.
16215 (gori_compute::compute_operand1_range): Ditto.
16216 (gori_compute::compute_operand2_range): Ditto.
16217 (gori_compute::outgoing_edge_range_p): Ditto.
16218 * gimple-range-gori.h (class gori_compute): Add range_tracer.
16220 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
16222 * flag-types.h (enum evrp_mode): Adjust evrp-mode values.
16223 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Relocate from.
16224 * gimple-range-trace.h (DEBUG_RANGE_CACHE): Here.
16225 * params.opt (--param=evrp-mode): Adjust options.
16227 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
16229 * Makefile.in (OBJS): Add gimple-range-trace.o.
16230 * gimple-range-cache.h (enable_new_values): Remove unused prototype.
16231 * gimple-range-fold.cc: Adjust headers.
16232 * gimple-range-trace.cc: New.
16233 * gimple-range-trace.h: New.
16234 * gimple-range.cc (gimple_ranger::gimple_ranger): Enable tracer.
16235 (gimple_ranger::range_of_expr): Add tracing.
16236 (gimple_ranger::range_on_entry): Ditto.
16237 (gimple_ranger::range_on_exit): Ditto.
16238 (gimple_ranger::range_on_edge): Ditto.
16239 (gimple_ranger::fold_range_internal): Ditto.
16240 (gimple_ranger::dump_bb): Do not calculate edge range twice.
16241 (trace_ranger::*): Remove.
16242 (enable_ranger): Never create a trace_ranger.
16243 (debug_seed_ranger): Move to gimple-range-trace.cc.
16244 (dump_ranger): Ditto.
16245 (debug_ranger): Ditto.
16246 * gimple-range.h: Include gimple-range-trace.h.
16247 (range_on_entry, range_on_exit): No longer virtual.
16248 (class trace_ranger): Remove.
16249 (DEBUG_RANGE_CACHE): Move to gimple-range-trace.h.
16251 2021-08-17 Martin Sebor <msebor@redhat.com>
16253 PR middle-end/101854
16254 * builtins.c (expand_builtin_alloca): Move warning code to check_alloca
16255 in gimple-ssa-warn-access.cc.
16256 * calls.c (alloc_max_size): Move code to check_alloca.
16257 (get_size_range): Move to pointer-query.cc.
16258 (maybe_warn_alloc_args_overflow): Move to gimple-ssa-warn-access.cc.
16259 (get_attr_nonstring_decl): Move to tree.c.
16260 (fntype_argno_type): Move to gimple-ssa-warn-access.cc.
16261 (append_attrname): Same.
16262 (maybe_warn_rdwr_sizes): Same.
16263 (initialize_argument_information): Move code to
16264 gimple-ssa-warn-access.cc.
16265 * calls.h (maybe_warn_alloc_args_overflow): Move to
16266 gimple-ssa-warn-access.h.
16267 (get_attr_nonstring_decl): Move to tree.h.
16268 (maybe_warn_nonstring_arg): Move to gimple-ssa-warn-access.h.
16269 (enum size_range_flags): Move to pointer-query.h.
16270 (get_size_range): Same.
16271 * gimple-ssa-warn-access.cc (has_location): Remove unused overload
16272 to avoid Clang -Wunused-function.
16273 (get_size_range): Declare static.
16274 (maybe_emit_free_warning): Rename...
16275 (maybe_check_dealloc_call): ...to this for consistency.
16276 (class pass_waccess): Add members.
16277 (pass_waccess::~pass_waccess): Defined.
16278 (alloc_max_size): Move here from calls.c.
16279 (maybe_warn_alloc_args_overflow): Same.
16280 (check_alloca): New function.
16281 (check_alloc_size_call): New function.
16282 (check_strncat): Handle another warning flag.
16283 (pass_waccess::check_builtin): Handle alloca.
16284 (fntype_argno_type): Move here from calls.c.
16285 (append_attrname): Same.
16286 (maybe_warn_rdwr_sizes): Same.
16287 (pass_waccess::check_call): Define.
16288 (check_nonstring_args): New function.
16289 (pass_waccess::check): Call new member functions.
16290 (pass_waccess::execute): Enable ranger.
16291 * gimple-ssa-warn-access.h (get_size_range): Move here from calls.h.
16292 (maybe_warn_nonstring_arg): Same.
16293 * gimple-ssa-warn-restrict.c: Remove #include.
16294 * pointer-query.cc (get_size_range): Move here from calls.c.
16295 * pointer-query.h (enum size_range_flags): Same.
16296 (get_size_range): Same.
16297 * tree.c (get_attr_nonstring_decl): Move here from calls.c.
16298 * tree.h (get_attr_nonstring_decl): Move here from calls.h.
16300 2021-08-17 Thomas Schwinge <thomas@codesourcery.com>
16302 * ggc.h (ggc_collect): Add 'force_collect' parameter.
16303 * ggc-page.c (ggc_collect): Use that one instead of global
16304 'ggc_force_collect'. Adjust all users.
16305 * doc/gty.texi (Invoking the garbage collector): Update.
16306 * ggc-internal.h (ggc_force_collect): Remove.
16307 * ggc-common.c (ggc_force_collect): Likewise.
16308 * selftest.h (forcibly_ggc_collect): Remove.
16309 * ggc-tests.c (selftest::forcibly_ggc_collect): Likewise.
16310 * read-rtl-function.c (test_loading_labels): Adjust.
16311 * selftest-run-tests.c (run_tests): Likewise.
16313 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
16315 * config/darwin.c (darwin_file_end): Reset and reclaim the
16316 section names table at the end of compile.
16318 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
16321 * config.in: Regenerate.
16322 * config/i386/darwin.h (EXTRA_ASM_OPTS): New
16323 (ASM_SPEC): Pass options to disable branch shortening where
16325 * configure: Regenerate.
16326 * configure.ac: Detect versions of 'as' that support the
16327 optimisation which has the bug.
16329 2021-08-17 Richard Biener <rguenther@suse.de>
16331 * optabs-query.c (supports_vec_gather_load_p): Also check
16333 (supports_vec_scatter_store_p): Likewise.
16334 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Fall
16335 back to masked variants if non-masked are not supported.
16336 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
16337 When we need to use masked gather/scatter but do not have
16338 a mask set up a constant true one.
16339 * tree-vect-stmts.c (vect_check_scalar_mask): Also allow
16340 non-SSA_NAME masks.
16342 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
16344 * tree-ssa-ccp.c (bit_value_binop) [MINUS_EXPR]: Use same
16345 algorithm as PLUS_EXPR to improve subtraction bit bounds.
16346 [POINTER_DIFF_EXPR]: Treat as synonymous with MINUS_EXPR.
16348 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
16350 * tree-ssa-ccp.c (bit_value_mult_const): New helper function to
16351 calculate the mask-value pair result of a multiplication by an
16353 (bit_value_binop) [MULT_EXPR]: Call it from here for
16354 multiplications by (sparse) non-negative constants.
16356 2021-08-17 Christophe Lyon <christophe.lyon@foss.st.com>
16359 * config.gcc (gcc_cv_initfini_array): Leave undefined for
16360 uclinuxfdpiceabi targets.
16362 2021-08-17 Alexandre Oliva <oliva@adacore.com>
16364 * tree-inline.c (maybe_move_debug_stmts_to_successors): Don't
16365 reverse debug stmts.
16367 2021-08-17 Alexandre Oliva <oliva@adacore.com>
16369 * tree-cfg.c (dump_function_to_file): Use fun, not cfun.
16371 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
16373 * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete.
16374 (__LD4Q_LANE_FUNC): Likewise.
16375 (vld4_lane_u8): Define without macro.
16376 (vld4_lane_u16): Likewise.
16377 (vld4_lane_u32): Likewise.
16378 (vld4_lane_u64): Likewise.
16379 (vld4_lane_s8): Likewise.
16380 (vld4_lane_s16): Likewise.
16381 (vld4_lane_s32): Likewise.
16382 (vld4_lane_s64): Likewise.
16383 (vld4_lane_f16): Likewise.
16384 (vld4_lane_f32): Likewise.
16385 (vld4_lane_f64): Likewise.
16386 (vld4_lane_p8): Likewise.
16387 (vld4_lane_p16): Likewise.
16388 (vld4_lane_p64): Likewise.
16389 (vld4q_lane_u8): Likewise.
16390 (vld4q_lane_u16): Likewise.
16391 (vld4q_lane_u32): Likewise.
16392 (vld4q_lane_u64): Likewise.
16393 (vld4q_lane_s8): Likewise.
16394 (vld4q_lane_s16): Likewise.
16395 (vld4q_lane_s32): Likewise.
16396 (vld4q_lane_s64): Likewise.
16397 (vld4q_lane_f16): Likewise.
16398 (vld4q_lane_f32): Likewise.
16399 (vld4q_lane_f64): Likewise.
16400 (vld4q_lane_p8): Likewise.
16401 (vld4q_lane_p16): Likewise.
16402 (vld4q_lane_p64): Likewise.
16403 (vld4_lane_bf16): Likewise.
16404 (vld4q_lane_bf16): Likewise.
16406 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
16408 * config/aarch64/arm_neon.h (__LD3_LANE_FUNC): Delete.
16409 (__LD3Q_LANE_FUNC): Delete.
16410 (vld3_lane_u8): Define without macro.
16411 (vld3_lane_u16): Likewise.
16412 (vld3_lane_u32): Likewise.
16413 (vld3_lane_u64): Likewise.
16414 (vld3_lane_s8): Likewise.
16415 (vld3_lane_s16): Likewise.
16416 (vld3_lane_s32): Likewise.
16417 (vld3_lane_s64): Likewise.
16418 (vld3_lane_f16): Likewise.
16419 (vld3_lane_f32): Likewise.
16420 (vld3_lane_f64): Likewise.
16421 (vld3_lane_p8): Likewise.
16422 (vld3_lane_p16): Likewise.
16423 (vld3_lane_p64): Likewise.
16424 (vld3q_lane_u8): Likewise.
16425 (vld3q_lane_u16): Likewise.
16426 (vld3q_lane_u32): Likewise.
16427 (vld3q_lane_u64): Likewise.
16428 (vld3q_lane_s8): Likewise.
16429 (vld3q_lane_s16): Likewise.
16430 (vld3q_lane_s32): Likewise.
16431 (vld3q_lane_s64): Likewise.
16432 (vld3q_lane_f16): Likewise.
16433 (vld3q_lane_f32): Likewise.
16434 (vld3q_lane_f64): Likewise.
16435 (vld3q_lane_p8): Likewise.
16436 (vld3q_lane_p16): Likewise.
16437 (vld3q_lane_p64): Likewise.
16438 (vld3_lane_bf16): Likewise.
16439 (vld3q_lane_bf16): Likewise.
16441 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
16443 * config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Delete.
16444 (__LD2Q_LANE_FUNC): Likewise.
16445 (vld2_lane_u8): Define without macro.
16446 (vld2_lane_u16): Likewise.
16447 (vld2_lane_u32): Likewise.
16448 (vld2_lane_u64): Likewise.
16449 (vld2_lane_s8): Likewise.
16450 (vld2_lane_s16): Likewise.
16451 (vld2_lane_s32): Likewise.
16452 (vld2_lane_s64): Likewise.
16453 (vld2_lane_f16): Likewise.
16454 (vld2_lane_f32): Likewise.
16455 (vld2_lane_f64): Likewise.
16456 (vld2_lane_p8): Likewise.
16457 (vld2_lane_p16): Likewise.
16458 (vld2_lane_p64): Likewise.
16459 (vld2q_lane_u8): Likewise.
16460 (vld2q_lane_u16): Likewise.
16461 (vld2q_lane_u32): Likewise.
16462 (vld2q_lane_u64): Likewise.
16463 (vld2q_lane_s8): Likewise.
16464 (vld2q_lane_s16): Likewise.
16465 (vld2q_lane_s32): Likewise.
16466 (vld2q_lane_s64): Likewise.
16467 (vld2q_lane_f16): Likewise.
16468 (vld2q_lane_f32): Likewise.
16469 (vld2q_lane_f64): Likewise.
16470 (vld2q_lane_p8): Likewise.
16471 (vld2q_lane_p16): Likewise.
16472 (vld2q_lane_p64): Likewise.
16473 (vld2_lane_bf16): Likewise.
16474 (vld2q_lane_bf16): Likewise.
16476 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
16478 * haifa-sched.c (advance_one_cycle): Output more context-synchronization
16481 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
16483 * haifa-sched.c (enum rfs_decision, rfs_str): Add RFS_AUTOPREF.
16484 (rank_for_schedule): Use it.
16486 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
16488 PR rtl-optimization/91598
16489 * haifa-sched.c (autopref_rank_for_schedule): Prioritize "irrelevant"
16490 insns after memory reads and before memory writes.
16492 2021-08-17 Alistair_Lee <alistair.lee@arm.com>
16494 * rtl.h (CONST_VECTOR_P): New macro.
16495 * config/aarch64/aarch64.c (aarch64_get_sve_pred_bits): Use RTL
16496 code testing macros.
16497 (aarch64_ptrue_all_mode): Likewise.
16498 (aarch64_expand_mov_immediate): Likewise.
16499 (aarch64_const_vec_all_in_range_p): Likewise.
16500 (aarch64_rtx_costs): Likewise.
16501 (aarch64_legitimate_constant_p): Likewise.
16502 (aarch64_simd_valid_immediate): Likewise.
16503 (aarch64_simd_make_constant): Likewise.
16504 (aarch64_convert_mult_to_shift): Likewise.
16505 (aarch64_expand_sve_vec_perm): Likewise.
16506 (aarch64_vec_fpconst_pow_of_2): Likewise.
16508 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
16510 PR tree-optimization/101938
16511 * range-op.cc (operator_abs::op1_range): Special case
16512 -TYPE_MIN_VALUE for flag_wrapv.
16514 2021-08-17 Kewen Lin <linkw@linux.ibm.com>
16516 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Add the cost for
16519 2021-08-17 Jakub Jelinek <jakub@redhat.com>
16521 * tree.def (OMP_SCOPE): New tree code.
16522 * tree.h (OMP_SCOPE_BODY, OMP_SCOPE_CLAUSES): Define.
16523 * tree-nested.c (convert_nonlocal_reference_stmt,
16524 convert_local_reference_stmt, convert_gimple_call): Handle
16526 * tree-pretty-print.c (dump_generic_node): Handle OMP_SCOPE.
16527 * gimple.def (GIMPLE_OMP_SCOPE): New gimple code.
16528 * gimple.c (gimple_build_omp_scope): New function.
16529 (gimple_copy): Handle GIMPLE_OMP_SCOPE.
16530 * gimple.h (gimple_build_omp_scope): Declare.
16531 (gimple_has_substatements): Handle GIMPLE_OMP_SCOPE.
16532 (gimple_omp_scope_clauses, gimple_omp_scope_clauses_ptr,
16533 gimple_omp_scope_set_clauses): New inline functions.
16534 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_SCOPE.
16535 * gimple-pretty-print.c (dump_gimple_omp_scope): New function.
16536 (pp_gimple_stmt_1): Handle GIMPLE_OMP_SCOPE.
16537 * gimple-walk.c (walk_gimple_stmt): Likewise.
16538 * gimple-low.c (lower_stmt): Likewise.
16539 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
16540 (gimplify_scan_omp_clauses): For task reductions, handle OMP_SCOPE
16541 like ORT_WORKSHARE constructs. Adjust diagnostics for %<scope%>
16542 allowing task reductions. Reject inscan reductions on scope.
16543 (omp_find_stores_stmt): Handle GIMPLE_OMP_SCOPE.
16544 (gimplify_omp_workshare, gimplify_expr): Handle OMP_SCOPE.
16545 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_SCOPE.
16546 (estimate_num_insns): Likewise.
16547 * omp-low.c (build_outer_var_ref): Look through GIMPLE_OMP_SCOPE
16548 contexts if var isn't privatized there.
16549 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_SCOPE.
16550 (scan_omp_1_stmt): Likewise.
16551 (maybe_add_implicit_barrier_cancel): Look through outer
16553 (lower_omp_scope): New function.
16554 (lower_omp_task_reductions): Handle OMP_SCOPE.
16555 (lower_omp_1): Handle GIMPLE_OMP_SCOPE.
16556 (diagnose_sb_1, diagnose_sb_2): Likewise.
16557 * omp-expand.c (expand_omp_single): Support also GIMPLE_OMP_SCOPE.
16558 (expand_omp): Handle GIMPLE_OMP_SCOPE.
16559 (omp_make_gimple_edges): Likewise.
16560 * omp-builtins.def (BUILT_IN_GOMP_SCOPE_START): New built-in.
16562 2021-08-17 Richard Biener <rguenther@suse.de>
16564 PR tree-optimization/101925
16565 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Set
16566 reverse on COMPONENT_REF and ARRAY_REF according to
16567 what reverse_storage_order_for_component_p does.
16568 (vn_reference_eq): Compare reversed on reference ops.
16569 (reverse_storage_order_for_component_p): New overload.
16570 (vn_reference_lookup_3): Check reverse_storage_order_for_component_p
16571 on the reference looked up.
16573 2021-08-17 Jeff Law <jlaw@localhost.localdomain>
16575 * config/h8300/h8300.c (shift_alg_si): Avoid loops for most SImode
16576 shifts on the H8/S.
16577 (h8300_option_override): Use loops on H8/S more often when optimizing
16579 (get_shift_alg): Handle new "special" cases on H8/S. Simplify
16580 accordingly. Handle various arithmetic right shifts with special
16581 sequences that we couldn't handle before.
16583 2021-08-16 Jeff Law <jlaw@localhost.localdomain>
16585 * config.gcc (rl78-*-elf*): Do not include dbxelf.h.
16587 2021-08-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
16589 * config/sparc/rtemself.h (SPARC_GCOV_TYPE_SIZE): Define.
16590 * config/sparc/sparc.c (sparc_gcov_type_size): New.
16591 (TARGET_GCOV_TYPE_SIZE): Redefine if SPARC_GCOV_TYPE_SIZE is defined.
16592 * coverage.c (get_gcov_type): Use targetm.gcov_type_size().
16593 * doc/tm.texi (TARGET_GCOV_TYPE_SIZE): Add hook under "Misc".
16594 * doc/tm.texi.in: Regenerate.
16595 * target.def (gcov_type_size): New target hook.
16596 * targhooks.c (default_gcov_type_size): New.
16597 * targhooks.h (default_gcov_type_size): Declare.
16598 * tree-profile.c (gimple_gen_edge_profiler): Use precision of
16600 (gimple_gen_time_profiler): Likewise.
16602 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
16604 * dwarf2out.c (add_scalar_info): Deal with DW_AT_data_bit_offset.
16606 2021-08-16 Tobias Burnus <tobias@codesourcery.com>
16608 PR middle-end/101931
16609 * omp-low.c (omp_runtime_api_call): Update for routines
16610 added in the meanwhile.
16612 2021-08-16 Martin Liska <mliska@suse.cz>
16614 PR tree-optimization/100393
16615 * tree-switch-conversion.c (group_cluster::dump): Use
16616 get_comparison_count.
16617 (jump_table_cluster::find_jump_tables): Pre-compute number of
16618 comparisons and then decrement it. Cache also max_ratio.
16619 (jump_table_cluster::can_be_handled): Change signature.
16620 * tree-switch-conversion.h (get_comparison_count): New.
16622 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
16624 * dwarf2out.c (add_data_member_location_attribute): Use GNAT
16625 encodings only when -fgnat-encodings=all is specified.
16626 (add_bound_info): Likewise.
16627 (add_byte_size_attribute): Likewise.
16628 (gen_member_die): Likewise.
16630 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
16632 * omp-oacc-neuter-broadcast.cc
16633 (execute_omp_oacc_neuter_broadcast): Plug 'par' memory leak.
16635 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
16637 * omp-oacc-neuter-broadcast.cc
16638 (execute_omp_oacc_neuter_broadcast): Clarify memory management for
16641 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
16643 * omp-oacc-neuter-broadcast.cc (field_map): Move variable into...
16644 (execute_omp_oacc_neuter_broadcast): ... here.
16645 (install_var_field, build_receiver_ref, build_sender_ref): Take
16646 'field_map_t *' parameter. Adjust all users.
16647 (worker_single_copy, neuter_worker_single): Take a
16648 'record_field_map_t *' parameter. Adjust all users.
16650 2021-08-16 liuhongt <hongtao.liu@intel.com>
16653 * config/i386/i386.md (ldexp<mode>3): Force operands[1] to
16656 2021-08-16 Martin Liska <mliska@suse.cz>
16659 * multiple_target.c (create_dispatcher_calls): Make default
16660 function local only if it is a definition.
16662 2021-08-16 Martin Liska <mliska@suse.cz>
16665 * ipa-icf-gimple.c (func_checker::compare_ssa_name): Do not
16666 consider equal SSA_NAMEs when one is a param.
16668 2021-08-16 liuhongt <hongtao.liu@intel.com>
16671 * config/i386/i386-expand.c (ix86_expand_vec_perm_vpermt2):
16672 Support vpermi2b for V32QI/V16QImode.
16673 (ix86_extract_perm_from_pool_constant): New function.
16674 (ix86_expand_vec_one_operand_perm_avx512): Support
16675 vpermw/vpermb under TARGET_AVX512BW/TARGET_AVX512VBMI.
16676 (expand_vec_perm_1): Adjust comments for upper.
16677 * config/i386/i386-protos.h (ix86_extract_perm_from_pool_constant):
16679 * config/i386/predicates.md (permvar_truncate_operand): New predicate.
16680 (pshufb_truncv4siv4hi_operand): Ditto.
16681 (pshufb_truncv8hiv8qi_operand): Ditto.
16682 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1):
16683 New pre_reload define_insn_and_split.
16684 (*avx512f_permvar_truncv8siv8hi_1): Ditto.
16685 (*avx512f_vpermvar_truncv8div8si_1): Ditto.
16686 (*avx512f_permvar_truncv32hiv32qi_1): Ditto.
16687 (*avx512f_permvar_truncv16hiv16qi_1): Ditto.
16688 (*avx512f_permvar_truncv4div4si_1): Ditto.
16689 (*avx512f_pshufb_truncv8hiv8qi_1): Ditto.
16690 (*avx512f_pshufb_truncv4siv4hi_1): Ditto.
16691 (*avx512f_pshufd_truncv2div2si_1): Ditto.
16693 2021-08-16 Kito Cheng <kito.cheng@sifive.com>
16695 * config/riscv/multilib-generator: Support code model option for
16697 * doc/install.texi: Add document of new option for
16698 --with-multilib-generator.
16700 2021-08-15 Clément Chigot <clement.chigot@atos.net>
16702 * config/rs6000/rs6000.c (xcoff_tls_exec_model_detected): New.
16703 (rs6000_legitimize_tls_address_aix): Use it.
16704 (rs6000_xcoff_file_end): Add ".ref __tls_get_addr" when
16705 xcoff_tls_exec_model_detected is true.
16707 2021-08-15 Jeff Law <jlaw@localhost.localdomain>
16709 * config/h8300/h8300.c (shift_alg_si): Retune H8/300H shifts
16710 to allow a bit more code growth, saving many dozens of cycles.
16711 (h8300_option_override): Adjus shift_alg_si if optimizing for
16713 (get_shift_alg): Use special + inline shifts for residuals
16716 2021-08-14 Stafford Horne <shorne@gmail.com>
16719 * config/or1k/or1k-opts.h: New file.
16720 * config/or1k/or1k.c (or1k_legitimize_address_1, print_reloc):
16721 Support generating gotha relocations if -mcmodel=large is
16723 * config/or1k/or1k.h (TARGET_CMODEL_SMALL, TARGET_CMODEL_LARGE):
16725 * config/or1k/or1k.opt (mcmodel=): New option.
16726 * doc/invoke.texi (OpenRISC Options): Document mcmodel.
16728 2021-08-14 Martin Sebor <msebor@redhat.com>
16730 PR middle-end/101791
16731 * gimple-ssa-warn-access.cc (new_delete_mismatch_p): Use new argument
16732 to valid_new_delete_pair_p.
16733 * tree.c (valid_new_delete_pair_p): Add argument.
16734 * tree.h (valid_new_delete_pair_p): Same.
16736 2021-08-14 Jakub Jelinek <jakub@redhat.com>
16739 * config/i386/i386-expand.c (expand_vec_perm_broadcast_1)
16740 <case E_V64QImode>: For this mode assert
16741 !TARGET_AVX512BW || d->perm[0] rather than !TARGET_AVX2 || d->perm[0].
16743 2021-08-13 Michael Meissner <meissner@linux.ibm.com>
16746 * config/rs6000/altivec.md (xxeval): Use register_predicate
16747 instead of altivec_register_predicate.
16749 2021-08-13 Martin Sebor <msebor@redhat.com>
16751 PR middle-end/101734
16752 * tree-ssa-uninit.c (maybe_warn_read_write_only): New function.
16753 (maybe_warn_operand): Call it.
16755 2021-08-13 Martin Liska <mliska@suse.cz>
16758 * attribs.c (decl_attributes): Make naked functions "noipa"
16761 2021-08-13 Martin Liska <mliska@suse.cz>
16764 * symtab.c (symtab_node::noninterposable_alias): Do not create
16765 local aliases for target_clone functions as the clonning pass
16768 2021-08-13 Martin Liska <mliska@suse.cz>
16770 * opts.c (LIVE_PATCHING_OPTION): Define.
16771 (control_options_for_live_patching): Use it in error messages.
16773 2021-08-13 Jan Hubicka <hubicka@ucw.cz>
16775 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOREAD.
16776 (implicit_const_eaf_flags, implicit_pure_eaf_flags,
16777 ignore_stores_eaf_flags): New constants.
16778 (remove_useless_eaf_flags): New function.
16779 (eaf_flags_useful_p): Use it.
16780 (deref_flags): Add EAF_NOT_RETURNED if flag is unused;
16782 (modref_lattice::init): Add EAF_NOREAD.
16783 (modref_lattice::add_escape_point): Do not reacord escape point if
16785 (modref_lattice::merge): EAF_NOESCAPE implies EAF_NODIRECTESCAPE;
16786 use remove_useless_eaf_flags.
16787 (modref_lattice::merge_deref): Use ignore_stores_eaf_flags.
16788 (modref_lattice::merge_direct_load): Add EAF_NOREAD
16789 (analyze_ssa_name_flags): Fix handling EAF_NOT_RETURNED
16790 (analyze_parms): Use remove_useless_eaf_flags.
16791 (ipa_merge_modref_summary_after_inlining): Use ignore_stores_eaf_flags.
16792 (modref_merge_call_site_flags): Add caller and ecf_flags parameter;
16793 use remove_useless_eaf_flags.
16794 (modref_propagate_flags_in_scc): Update.
16795 * ipa-modref.h: Turn eaf_flags_t back to char.
16796 * tree-core.h (EAF_NOT_RETURNED): Fix.
16797 (EAF_NOREAD): New constant
16798 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1): Check for
16800 * tree-ssa-structalias.c (handle_rhs_call): Handle new flags.
16801 (handle_pure_call): Likewise.
16803 2021-08-12 Jakub Jelinek <jakub@redhat.com>
16805 * tree.def (OMP_MASKED): New tree code.
16806 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_FILTER.
16807 * tree.h (OMP_MASKED_BODY, OMP_MASKED_CLAUSES, OMP_MASKED_COMBINED,
16808 OMP_CLAUSE_FILTER_EXPR): Define.
16809 * tree.c (omp_clause_num_ops): Add OMP_CLAUSE_FILTER entry.
16810 (omp_clause_code_name): Likewise.
16811 (walk_tree_1): Handle OMP_CLAUSE_FILTER.
16812 * tree-nested.c (convert_nonlocal_omp_clauses,
16813 convert_local_omp_clauses): Handle OMP_CLAUSE_FILTER.
16814 (convert_nonlocal_reference_stmt, convert_local_reference_stmt,
16815 convert_gimple_call): Handle GIMPLE_OMP_MASTER.
16816 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_FILTER.
16817 (dump_generic_node): Handle OMP_MASTER.
16818 * gimple.def (GIMPLE_OMP_MASKED): New gimple code.
16819 * gimple.c (gimple_build_omp_masked): New function.
16820 (gimple_copy): Handle GIMPLE_OMP_MASKED.
16821 * gimple.h (gimple_build_omp_masked): Declare.
16822 (gimple_has_substatements): Handle GIMPLE_OMP_MASKED.
16823 (gimple_omp_masked_clauses, gimple_omp_masked_clauses_ptr,
16824 gimple_omp_masked_set_clauses): New inline functions.
16825 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_MASKED.
16826 * gimple-pretty-print.c (dump_gimple_omp_masked): New function.
16827 (pp_gimple_stmt_1): Handle GIMPLE_OMP_MASKED.
16828 * gimple-walk.c (walk_gimple_stmt): Likewise.
16829 * gimple-low.c (lower_stmt): Likewise.
16830 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
16831 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_FILTER. For clauses
16832 that take one expression rather than decl or constant, force
16833 gimplification of that into a SSA_NAME or temporary unless min
16835 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_FILTER.
16836 (gimplify_expr): Handle OMP_MASKED.
16837 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_MASKED.
16838 (estimate_num_insns): Likewise.
16839 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_FILTER.
16840 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_MASKED. Adjust
16841 diagnostics for existence of masked construct.
16842 (scan_omp_1_stmt, lower_omp_master, lower_omp_1, diagnose_sb_1,
16843 diagnose_sb_2): Handle GIMPLE_OMP_MASKED.
16844 * omp-expand.c (expand_omp_synch, expand_omp, omp_make_gimple_edges):
16847 2021-08-12 Uroš Bizjak <ubizjak@gmail.com>
16850 * config/i386/i386.md (avx512f_scalef<mode>2): New insn pattern.
16851 (ldexp<mode>3): Use avx512f_scalef<mode>2.
16852 (UNSPEC_SCALEF): Move from ...
16853 * config/i386/sse.md (UNSPEC_SCALEF): ... here.
16855 2021-08-12 Jan Hubicka <hubicka@ucw.cz>
16857 * ipa-split.c (consider_split): Fix condition testing void functions.
16859 2021-08-12 Aldy Hernandez <aldyh@redhat.com>
16861 * doc/invoke.texi: Remove docs for threader-mode param.
16862 * flag-types.h (enum threader_mode): Remove.
16863 * params.opt: Remove threader-mode param.
16864 * tree-ssa-threadbackward.c (class back_threader): Remove
16865 path_is_unreachable_p.
16866 Make find_paths private.
16867 Add maybe_thread and thread_through_all_blocks.
16868 Remove reference marker for m_registry.
16869 Remove reference marker for m_profit.
16870 (back_threader::back_threader): Adjust for registry and profit not
16872 (dump_path): Move down.
16873 (debug): Move down.
16874 (class thread_jumps): Remove.
16875 (class back_threader_registry): Remove m_all_paths.
16877 (thread_jumps::thread_through_all_blocks): Move to back_threader
16879 (fsm_find_thread_path): Remove
16880 (back_threader::maybe_thread): New.
16881 (back_threader::thread_through_all_blocks): Move from
16883 (back_threader_registry::back_threader_registry): Remove
16885 (back_threader_registry::~back_threader_registry): Remove.
16886 (thread_jumps::find_taken_edge): Remove.
16887 (thread_jumps::check_subpath_and_update_thread_path): Remove.
16888 (thread_jumps::maybe_register_path): Remove.
16889 (thread_jumps::handle_phi): Remove.
16890 (handle_assignment_p): Remove.
16891 (thread_jumps::handle_assignment): Remove.
16892 (thread_jumps::fsm_find_control_statement_thread_paths): Remove.
16893 (thread_jumps::find_jump_threads_backwards): Remove.
16894 (thread_jumps::find_jump_threads_backwards_with_ranger): Remove.
16895 (try_thread_blocks): Rename find_jump_threads_backwards to
16897 (pass_early_thread_jumps::execute): Same.
16899 2021-08-12 Tobias Burnus <tobias@codesourcery.com>
16901 * tree-core.h (omp_clause_proc_bind_kind): Add
16902 OMP_CLAUSE_PROC_BIND_PRIMARY.
16903 * tree-pretty-print.c (dump_omp_clause): Add TODO comment to
16904 change 'master' to 'primary' in proc_bind for OpenMP 5.1.
16906 2021-08-12 Claudiu Zissulescu <claziss@synopsys.com>
16908 * common/config/arc/arc-common.c (arc_option_init_struct): Remove
16909 fno-common reference.
16910 * config/arc/arc.c (arc_override_options): Remove overriding of
16913 2021-08-12 Jakub Jelinek <jakub@redhat.com>
16916 * config/i386/i386-expand.c (ix86_expand_vec_one_operand_perm_avx512):
16917 If d->testing_p, return true after performing checks instead of
16918 actually expanding the insn.
16919 (expand_vec_perm_broadcast_1): Handle V32HImode - assert
16920 !TARGET_AVX512BW and return false.
16922 2021-08-12 Eric Botcazou <ebotcazou@gcc.gnu.org>
16924 * configure.ac (PE linker --disable-dynamicbase support): New check.
16925 * configure: Regenerate.
16926 * config.in: Likewise.
16927 * config/i386/mingw32.h (LINK_SPEC_DISABLE_DYNAMICBASE): New define.
16928 (LINK_SPEC): Use it.
16929 * config/i386/mingw-w64.h (LINK_SPEC_DISABLE_DYNAMICBASE): Likewise.
16930 (LINK_SPEC): Likewise.
16932 2021-08-12 liuhongt <hongtao.liu@intel.com>
16935 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_2): New
16936 post_reload define_insn_and_split.
16937 (*avx512bw_zero_extendv32qiv32hi2_2): Ditto.
16938 (*sse4_1_zero_extendv8qiv8hi2_4): Ditto.
16939 (*avx512f_zero_extendv16hiv16si2_2): Ditto.
16940 (*avx2_zero_extendv8hiv8si2_2): Ditto.
16941 (*sse4_1_zero_extendv4hiv4si2_4): Ditto.
16942 (*avx512f_zero_extendv8siv8di2_2): Ditto.
16943 (*avx2_zero_extendv4siv4di2_2): Ditto.
16944 (*sse4_1_zero_extendv2siv2di2_4): Ditto.
16945 (VI248_256, VI248_512, VI148_512, VI148_256, VI148_128): New
16948 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
16950 * config/rs6000/rs6000-builtin-new.def: Add always, power5, and
16953 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
16955 * config/rs6000/rs6000-builtin-new.def: Add vsx stanza.
16957 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
16959 * config/rs6000/rs6000-builtin-new.def: Finish altivec stanza.
16960 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
16961 initialization of pcvoid_type_node here...
16962 (altivec_init_builtins): ...from here.
16963 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add
16964 RS6000_BTI_const_ptr_void.
16965 (pcvoid_type_node): New macro.
16967 2021-08-11 Richard Biener <rguenther@suse.de>
16970 * tree-ssa-forwprop.c (pass_forwprop::execute): Do not decompose
16971 hard-register accesses.
16973 2021-08-11 Richard Biener <rguenther@suse.de>
16975 * tree-ssa-operands.c (operands_scanner::get_expr_operands):
16976 Do not look at COMPONENT_REF FIELD_DECLs TREE_THIS_VOLATILE
16977 to determine has_volatile_ops.
16979 2021-08-11 Eric Botcazou <ebotcazou@gcc.gnu.org>
16981 * cfgexpand.c (expand_used_vars): Reuse attribs local variable.
16983 2021-08-11 Jan Hubicka <hubicka@ucw.cz>
16984 Alexandre Oliva <oliva@adacore.com>
16986 * ipa-modref.c (modref_lattice::dump): Fix escape_point's min_flags
16988 (modref_lattice::merge_deref): Fix handling of indirect scape points.
16989 (update_escape_summary_1): Likewise.
16990 (update_escape_summary): Likewise.
16991 (ipa_merge_modref_summary_after_inlining): Likewise.
16993 2021-08-11 Richard Biener <rguenther@suse.de>
16995 PR middle-end/101858
16996 * fold-const.c (fold_binary_loc): Guard simplification
16997 of X < (cast) (1 << Y) to integer types.
16999 2021-08-11 Richard Biener <rguenther@suse.de>
17001 PR tree-optimization/101861
17002 * tree-vect-stmts.c (vectorizable_load): Fix error in
17003 previous change with regard to gather vectorization.
17005 2021-08-11 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
17008 * config/arm/arm_neon.h (vdup_n_s8): Replace call to builtin
17010 (vdup_n_s16): Likewise.
17011 (vdup_n_s32): Likewise.
17012 (vdup_n_s64): Likewise.
17013 (vdup_n_u8): Likewise.
17014 (vdup_n_u16): Likewise.
17015 (vdup_n_u32): Likewise.
17016 (vdup_n_u64): Likewise.
17017 (vdup_n_p8): Likewise.
17018 (vdup_n_p16): Likewise.
17019 (vdup_n_p64): Likewise.
17020 (vdup_n_f16): Likewise.
17021 (vdup_n_f32): Likewise.
17022 (vdupq_n_s8): Likewise.
17023 (vdupq_n_s16): Likewise.
17024 (vdupq_n_s32): Likewise.
17025 (vdupq_n_s64): Likewise.
17026 (vdupq_n_u8): Likewise.
17027 (vdupq_n_u16): Likewise.
17028 (vdupq_n_u32): Likewise.
17029 (vdupq_n_u64): Likewise.
17030 (vdupq_n_p8): Likewise.
17031 (vdupq_n_p16): Likewise.
17032 (vdupq_n_p64): Likewise.
17033 (vdupq_n_f16): Likewise.
17034 (vdupq_n_f32): Likewise.
17035 (vmov_n_s8): Replace call to builtin with call to corresponding
17037 (vmov_n_s16): Likewise.
17038 (vmov_n_s32): Likewise.
17039 (vmov_n_s64): Likewise.
17040 (vmov_n_u8): Likewise.
17041 (vmov_n_u16): Likewise.
17042 (vmov_n_u32): Likewise.
17043 (vmov_n_u64): Likewise.
17044 (vmov_n_p8): Likewise.
17045 (vmov_n_p16): Likewise.
17046 (vmov_n_f16): Likewise.
17047 (vmov_n_f32): Likewise.
17048 (vmovq_n_s8): Likewise.
17049 (vmovq_n_s16): Likewise.
17050 (vmovq_n_s32): Likewise.
17051 (vmovq_n_s64): Likewise.
17052 (vmovq_n_u8): Likewise.
17053 (vmovq_n_u16): Likewise.
17054 (vmovq_n_u32): Likewise.
17055 (vmovq_n_u64): Likewise.
17056 (vmovq_n_p8): Likewise.
17057 (vmovq_n_p16): Likewise.
17058 (vmovq_n_f16): Likewise.
17059 (vmovq_n_f32): Likewise.
17060 * config/arm/arm_neon_builtins.def: Remove entries for vdup_n.
17062 2021-08-11 liuhongt <hongtao.liu@intel.com>
17065 * config/i386/i386.md (ldexp<mode>3): Extend to vscalefs[sd]
17066 when TARGET_AVX512F and TARGET_SSE_MATH.
17068 2021-08-10 Jakub Jelinek <jakub@redhat.com>
17071 * config/i386/i386-expand.c (expand_vec_perm_even_odd): Return false
17072 for V32HImode if !TARGET_AVX512BW.
17073 (ix86_vectorize_vec_perm_const) <case E_V32HImode, case E_V64QImode>:
17074 If !TARGET_AVX512BW and TARGET_AVX512F and d.testing_p, don't fail
17075 early, but actually check the permutation.
17077 2021-08-10 Richard Biener <rguenther@suse.de>
17079 PR tree-optimization/101809
17080 * tree-vect-stmts.c (get_load_store_type): Allow emulated
17081 gathers with offset vector nunits being a constant multiple
17082 of the data vector nunits.
17083 (vect_get_gather_scatter_ops): Use the appropriate nunits
17084 for the offset vector defs.
17085 (vectorizable_store): Adjust call to
17086 vect_get_gather_scatter_ops.
17087 (vectorizable_load): Likewise. Handle the case of less
17088 offset vectors than data vectors.
17090 2021-08-10 Jakub Jelinek <jakub@redhat.com>
17093 * config/i386/sse.md (*avx512f_shuf_<shuffletype>64x2_1<mask_name>_1,
17094 *avx512f_shuf_<shuffletype>32x4_1<mask_name>_1): New define_insn
17097 2021-08-10 Richard Biener <rguenther@suse.de>
17099 PR tree-optimization/101801
17100 PR tree-optimization/101819
17101 * tree-vectorizer.h (vect_emulated_vector_p): Declare.
17102 * tree-vect-loop.c (vect_emulated_vector_p): New function.
17103 (vectorizable_reduction): Re-instantiate a check for emulated
17105 * tree-vect-stmts.c (vectorizable_shift): Likewise.
17106 (vectorizable_operation): Likewise. Cost emulated vector
17107 operations according to the scalar sequence synthesized by
17110 2021-08-10 Richard Biener <rguenther@suse.de>
17112 PR middle-end/101824
17113 * tree-nested.c (get_frame_field): Mark the COMPONENT_REF as
17114 volatile in case the variable was.
17116 2021-08-10 H.J. Lu <hjl.tools@gmail.com>
17119 * config/i386/constraints.md (BC): Document for integer SSE
17120 constant all bits set operand.
17121 (BF): New constraint for const floating-point all bits set
17123 * config/i386/i386.c (standard_sse_constant_p): Likewise.
17124 (standard_sse_constant_opcode): Likewise.
17125 * config/i386/sse.md (sseconstm1): New mode attribute.
17126 (mov<mode>_internal): Replace BC with <sseconstm1>.
17128 2021-08-10 liuhongt <hongtao.liu@intel.com>
17130 * config/i386/sse.md (cond_<insn><mode>): New expander.
17131 (VI248_AVX512VLBW): New mode iterator.
17132 * config/i386/predicates.md
17133 (nonimmediate_or_const_vec_dup_operand): New predicate.
17135 2021-08-09 Andrew MacLeod <amacleod@redhat.com>
17137 PR tree-optimization/101741
17138 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Check
17139 type of parameter for toupper/tolower.
17141 2021-08-09 Martin Jambor <mjambor@suse.cz>
17143 PR testsuite/101654
17144 * ipa-prop.c (propagate_controlled_uses): Removed a spurious space.
17146 2021-08-09 Pat Haugen <pthaugen@linux.ibm.com>
17148 * config/rs6000/rs6000.c (is_load_insn1): Verify destination is a
17150 (is_store_insn1): Verify source is a register.
17152 2021-08-09 Uroš Bizjak <ubizjak@gmail.com>
17155 * config/i386/mmx.md (<any_logic:code>v2sf3):
17156 Rename from *mmx_<any_logic:code>v2sf3
17158 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
17160 * config/nvptx/nvptx.c: Cross-reference parts adapted in
17161 'gcc/omp-oacc-neuter-broadcast.cc'.
17162 * omp-low.c: Likewise.
17163 * omp-oacc-neuter-broadcast.cc: Cross-reference parts adapted from
17166 2021-08-09 Julian Brown <julian@codesourcery.com>
17167 Kwok Cheung Yeung <kcy@codesourcery.com>
17168 Thomas Schwinge <thomas@codesourcery.com>
17170 * config/gcn/gcn.c (gcn_init_builtins): Override decls for
17171 BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START,
17172 BUILT_IN_GOACC_SINGLE_COPY_END and BUILT_IN_GOACC_BARRIER.
17173 (gcn_goacc_validate_dims): Turn on worker partitioning unconditionally.
17174 (gcn_fork_join): Update comment.
17175 * config/gcn/gcn.opt (flag_worker_partitioning): Remove.
17176 (macc_experimental_workers): Remove unused option.
17178 2021-08-09 Julian Brown <julian@codesourcery.com>
17179 Nathan Sidwell <nathan@codesourcery.com> (via 'gcc/config/nvptx/nvptx.c' master)
17180 Kwok Cheung Yeung <kcy@codesourcery.com>
17181 Thomas Schwinge <thomas@codesourcery.com>
17183 * Makefile.in (OBJS): Add omp-oacc-neuter-broadcast.o.
17184 * doc/tm.texi.in (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD):
17185 Add documentation hook.
17186 * doc/tm.texi: Regenerate.
17187 * omp-oacc-neuter-broadcast.cc: New file.
17188 * omp-builtins.def (BUILT_IN_GOACC_BARRIER)
17189 (BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START)
17190 (BUILT_IN_GOACC_SINGLE_COPY_END): New builtins.
17191 * passes.def (pass_omp_oacc_neuter_broadcast): Add pass.
17192 * target.def (goacc.create_worker_broadcast_record): Add target
17194 * tree-pass.h (make_pass_omp_oacc_neuter_broadcast): Add
17196 * config/gcn/gcn-protos.h (gcn_goacc_adjust_propagation_record):
17197 Rename prototype to...
17198 (gcn_goacc_create_worker_broadcast_record): ... this.
17199 * config/gcn/gcn-tree.c (gcn_goacc_adjust_propagation_record): Rename
17201 (gcn_goacc_create_worker_broadcast_record): ... this.
17202 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_PROPAGATION_RECORD):
17204 (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD): ... this.
17206 2021-08-09 Tejas Belagod <tejas.belagod@arm.com>
17209 * config/aarch64/aarch64-simd.md (vlshr<mode>3, vashr<mode>3): Use
17210 the right iterator.
17212 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
17214 * Makefile.in (GTFILES): Remove '$(srcdir)/omp-offload.c'.
17216 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
17218 * builtins.def (DEF_GOACC_BUILTIN, DEF_GOMP_BUILTIN): Don't
17219 consider '-foffload-abi'.
17220 * common.opt (-foffload-abi): Remove 'Var', 'Init'.
17221 * opts.c (common_handle_option) <-foffload-abi> [ACCEL_COMPILER]:
17224 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
17226 * optc-gen.awk: Sanity check that 'Init' doesn't appear without
17229 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
17231 * omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE): Remove.
17233 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
17235 * doc/gty.texi (Files): Update.
17237 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
17239 * doc/gty.texi (Files): Fix GTY header file example.
17241 2021-08-09 Roger Sayle <roger@nextmovesoftware.com>
17243 * tree-ssa-ccp.c (value_mask_to_min_max): Helper function to
17244 determine the upper and lower bounds from a mask-value pair.
17245 (bit_value_unop) [ABS_EXPR, ABSU_EXPR]: Add support for
17246 absolute value and unsigned absolute value expressions.
17247 (bit_value_binop): Initialize *VAL's precision.
17248 [LT_EXPR, LE_EXPR]: Use value_mask_to_min_max to determine
17249 upper and lower bounds of operands. Add LE_EXPR/GE_EXPR
17250 support when the operands are unknown but potentially equal.
17251 [MIN_EXPR, MAX_EXPR]: Support minimum/maximum expressions.
17253 2021-08-09 Bin Cheng <bin.cheng@linux.alibaba.com>
17255 * config/aarch64/aarch64.md
17256 (*extend<SHORT:mode><GPI:mode>2_aarch64): Use %<GPI:w>0.
17258 2021-08-08 Sergei Trofimovich <siarheit@google.com>
17260 * lra-constraints.c: Fix s/otput/output/ typo.
17262 2021-08-06 Martin Sebor <msebor@redhat.com>
17264 * builtins.c (expand_builtin_memchr): Move to gimple-ssa-warn-access.cc.
17265 (expand_builtin_strcat): Same.
17266 (expand_builtin_stpncpy): Same.
17267 (expand_builtin_strncat): Same.
17268 (check_read_access): Same.
17269 (check_memop_access): Same.
17270 (expand_builtin_strlen): Move checks to gimple-ssa-warn-access.cc.
17271 (expand_builtin_strnlen): Same.
17272 (expand_builtin_memcpy): Same.
17273 (expand_builtin_memmove): Same.
17274 (expand_builtin_mempcpy): Same.
17275 (expand_builtin_strcpy): Same.
17276 (expand_builtin_strcpy_args): Same.
17277 (expand_builtin_stpcpy_1): Same.
17278 (expand_builtin_strncpy): Same.
17279 (expand_builtin_memset): Same.
17280 (expand_builtin_bzero): Same.
17281 (expand_builtin_strcmp): Same.
17282 (expand_builtin_strncmp): Same.
17283 (expand_builtin): Remove handlers.
17284 (fold_builtin_strlen): Add a comment.
17285 * builtins.h (check_access): Move to gimple-ssa-warn-access.cc.
17286 * calls.c (maybe_warn_nonstring_arg): Same.
17287 * diagnostic-spec.c (nowarn_spec_t::nowarn_spec_t): Add warning option.
17288 * gimple-fold.c (gimple_fold_builtin_strcpy): Pass argument to callee.
17289 (gimple_fold_builtin_stpcpy): Same.
17290 * gimple-ssa-warn-access.cc (has_location): New function.
17291 (get_location): Same.
17292 (get_callee_fndecl): Same.
17293 (call_nargs): Same.
17295 (warn_string_no_nul): Define.
17296 (unterminated_array): Same.
17297 (check_nul_terminated_array): Same.
17298 (maybe_warn_nonstring_arg): Same.
17299 (maybe_warn_for_bound): Same.
17300 (warn_for_access): Same.
17301 (check_access): Same.
17302 (check_memop_access): Same.
17303 (check_read_access): Same.
17304 (warn_dealloc_offset): Use helper functions.
17305 (maybe_emit_free_warning): Same.
17306 (class pass_waccess): Add members.
17307 (check_strcat): New function.
17308 (check_strncat): New function.
17309 (check_stxcpy): New function.
17310 (check_stxncpy): New function.
17311 (check_strncmp): New function.
17312 (pass_waccess::check_builtin): New function.
17313 (pass_waccess::check): Call it.
17314 * gimple-ssa-warn-access.h (warn_string_no_nul): Move here from
17316 (maybe_warn_for_bound): Same.
17317 (check_access): Same.
17318 (check_memop_access): Same.
17319 (check_read_access): Same.
17320 * pointer-query.h (struct access_data): Define a ctor overload.
17322 2021-08-06 Richard Biener <rguenther@suse.de>
17324 PR tree-optimization/101801
17325 * tree-vectorizer.h (vect_worthwhile_without_simd_p): Rename...
17326 (vect_can_vectorize_without_simd_p): ... to this.
17327 * tree-vect-loop.c (vect_worthwhile_without_simd_p): Rename...
17328 (vect_can_vectorize_without_simd_p): ... to this and fold
17329 in vect_min_worthwhile_factor.
17330 (vect_min_worthwhile_factor): Remove.
17331 (vectorizable_reduction): Adjust and remove the cost part.
17332 * tree-vect-stmts.c (vectorizable_shift): Likewise.
17333 (vectorizable_operation): Likewise.
17335 2021-08-06 Uroš Bizjak <ubizjak@gmail.com>
17338 * config/i386/i386.md (cmove reg-to-reg move elimination peephole2s):
17339 Add general_gr_operand predicate to operand 3.
17341 2021-08-06 Roger Sayle <roger@nextmovesoftware.com>
17343 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern): Use
17344 CFN_BUILT_IN_CLRSB* instead of BUILT_IN_CLRSB* for consistency.
17346 2021-08-06 Tamar Christina <tamar.christina@arm.com>
17348 * config/aarch64/aarch64-sve-builtins.cc (register_svpattern,
17349 register_svprfop): Pass vec<> by pointer.
17350 * langhooks-def.h (lhd_simulate_enum_decl): Likewise.
17351 * langhooks.c (lhd_simulate_enum_decl): Likewise.
17352 * langhooks.h (struct lang_hooks_for_types): Likewise.
17354 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
17356 * config/aarch64/arm_neon.h (vst1_bf16_x2): Use
17357 __builtin_memcpy instead of constructing an additional
17358 __builtin_aarch64_simd_oi one vector at a time.
17359 (vst1q_bf16_x2): Likewise.
17360 (vst1_bf16_x3): Use __builtin_memcpy instead of constructing
17361 an additional __builtin_aarch64_simd_ci one vector at a time.
17362 (vst1q_bf16_x3): Likewise.
17363 (vst1_bf16_x4): Use __builtin_memcpy instead of a union.
17364 (vst1q_bf16_x4): Likewise.
17365 (vst2_bf16): Use __builtin_memcpy instead of constructing an
17366 additional __builtin_aarch64_simd_oi one vector at a time.
17367 (vst2q_bf16): Likewise.
17368 (vst3_bf16): Use __builtin_memcpy instead of constructing an
17369 additional __builtin_aarch64_simd_ci mode one vector at a
17371 (vst3q_bf16): Likewise.
17372 (vst4_bf16): Use __builtin_memcpy instead of constructing an
17373 additional __builtin_aarch64_simd_xi one vector at a time.
17374 (vst4q_bf16): Likewise.
17376 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
17378 * config/aarch64/arm_neon.h (__ST2_LANE_FUNC): Delete.
17379 (__ST2Q_LANE_FUNC): Delete.
17380 (vst2_lane_f16): Use __builtin_memcpy to copy vector
17381 structure instead of constructing __builtin_aarch64_simd_oi
17382 one vector at a time.
17383 (vst2_lane_f32): Likewise.
17384 (vst2_lane_f64): Likewise.
17385 (vst2_lane_p8): Likewise.
17386 (vst2_lane_p16): Likewise.
17387 (vst2_lane_p64): Likewise.
17388 (vst2_lane_s8): Likewise.
17389 (vst2_lane_s16): Likewise.
17390 (vst2_lane_s32): Likewise.
17391 (vst2_lane_s64): Likewise.
17392 (vst2_lane_u8): Likewise.
17393 (vst2_lane_u16): Likewise.
17394 (vst2_lane_u32): Likewise.
17395 (vst2_lane_u64): Likewise.
17396 (vst2_lane_bf16): Likewise.
17397 (vst2q_lane_f16): Use __builtin_memcpy to copy vector
17398 structure instead of using a union.
17399 (vst2q_lane_f32): Likewise.
17400 (vst2q_lane_f64): Likewise.
17401 (vst2q_lane_p8): Likewise.
17402 (vst2q_lane_p16): Likewise.
17403 (vst2q_lane_p64): Likewise.
17404 (vst2q_lane_s8): Likewise.
17405 (vst2q_lane_s16): Likewise.
17406 (vst2q_lane_s32): Likewise.
17407 (vst2q_lane_s64): Likewise.
17408 (vst2q_lane_u8): Likewise.
17409 (vst2q_lane_u16): Likewise.
17410 (vst2q_lane_u32): Likewise.
17411 (vst2q_lane_u64): Likewise.
17412 (vst2q_lane_bf16): Likewise.
17414 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
17416 * config/aarch64/arm_neon.h (__ST3_LANE_FUNC): Delete.
17417 (__ST3Q_LANE_FUNC): Delete.
17418 (vst3_lane_f16): Use __builtin_memcpy to copy vector
17419 structure instead of constructing __builtin_aarch64_simd_ci
17420 one vector at a time.
17421 (vst3_lane_f32): Likewise.
17422 (vst3_lane_f64): Likewise.
17423 (vst3_lane_p8): Likewise.
17424 (vst3_lane_p16): Likewise.
17425 (vst3_lane_p64): Likewise.
17426 (vst3_lane_s8): Likewise.
17427 (vst3_lane_s16): Likewise.
17428 (vst3_lane_s32): Likewise.
17429 (vst3_lane_s64): Likewise.
17430 (vst3_lane_u8): Likewise.
17431 (vst3_lane_u16): Likewise.
17432 (vst3_lane_u32): Likewise.
17433 (vst3_lane_u64): Likewise.
17434 (vst3_lane_bf16): Likewise.
17435 (vst3q_lane_f16): Use __builtin_memcpy to copy vector
17436 structure instead of using a union.
17437 (vst3q_lane_f32): Likewise.
17438 (vst3q_lane_f64): Likewise.
17439 (vst3q_lane_p8): Likewise.
17440 (vst3q_lane_p16): Likewise.
17441 (vst3q_lane_p64): Likewise.
17442 (vst3q_lane_s8): Likewise.
17443 (vst3q_lane_s16): Likewise.
17444 (vst3q_lane_s32): Likewise.
17445 (vst3q_lane_s64): Likewise.
17446 (vst3q_lane_u8): Likewise.
17447 (vst3q_lane_u16): Likewise.
17448 (vst3q_lane_u32): Likewise.
17449 (vst3q_lane_u64): Likewise.
17450 (vst3q_lane_bf16): Likewise.
17452 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
17454 * config/aarch64/arm_neon.h (__ST4_LANE_FUNC): Delete.
17455 (__ST4Q_LANE_FUNC): Delete.
17456 (vst4_lane_f16): Use __builtin_memcpy to copy vector
17457 structure instead of constructing __builtin_aarch64_simd_xi
17458 one vector at a time.
17459 (vst4_lane_f32): Likewise.
17460 (vst4_lane_f64): Likewise.
17461 (vst4_lane_p8): Likewise.
17462 (vst4_lane_p16): Likewise.
17463 (vst4_lane_p64): Likewise.
17464 (vst4_lane_s8): Likewise.
17465 (vst4_lane_s16): Likewise.
17466 (vst4_lane_s32): Likewise.
17467 (vst4_lane_s64): Likewise.
17468 (vst4_lane_u8): Likewise.
17469 (vst4_lane_u16): Likewise.
17470 (vst4_lane_u32): Likewise.
17471 (vst4_lane_u64): Likewise.
17472 (vst4_lane_bf16): Likewise.
17473 (vst4q_lane_f16): Use __builtin_memcpy to copy vector
17474 structure instead of using a union.
17475 (vst4q_lane_f32): Likewise.
17476 (vst4q_lane_f64): Likewise.
17477 (vst4q_lane_p8): Likewise.
17478 (vst4q_lane_p16): Likewise.
17479 (vst4q_lane_p64): Likewise.
17480 (vst4q_lane_s8): Likewise.
17481 (vst4q_lane_s16): Likewise.
17482 (vst4q_lane_s32): Likewise.
17483 (vst4q_lane_s64): Likewise.
17484 (vst4q_lane_u8): Likewise.
17485 (vst4q_lane_u16): Likewise.
17486 (vst4q_lane_u32): Likewise.
17487 (vst4q_lane_u64): Likewise.
17488 (vst4q_lane_bf16): Likewise.
17490 2021-08-06 Martin Liska <mliska@suse.cz>
17492 * config/rs6000/rs6000.c (rs6000_option_override_internal): When
17493 a target option is restored, it can have
17494 rs6000_long_double_type_size set to FLOAT_PRECISION_TFmode
17495 and error should not be emitted.
17497 2021-08-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
17499 * gcov-io.h (gcov_write): Declare.
17500 * gcov-io.c (gcov_write): New.
17501 (gcov_write_counter): Remove.
17502 (gcov_write_tag_length): Likewise.
17503 (gcov_write_summary): Replace gcov_write_tag_length() with calls to
17504 gcov_write_unsigned().
17505 * doc/invoke.texi (fprofile-info-section): Mention
17506 __gcov_info_to_gdca().
17508 2021-08-06 Martin Sebor <msebor@redhat.com>
17510 * dominance.c (prune_bbs_to_update_dominators): Adjust by-value vec
17511 arguments to by-reference.
17512 (iterate_fix_dominators): Same.
17513 * dominance.h (iterate_fix_dominators): Same.
17514 * ipa-prop.h: Call auto_vec::to_vec_legacy.
17515 * tree-data-ref.c (dump_data_dependence_relation): Adjust by-value vec
17516 arguments to by-reference.
17517 (debug_data_dependence_relation): Same.
17518 (dump_data_dependence_relations): Same.
17519 * tree-data-ref.h (debug_data_dependence_relation): Same.
17520 (dump_data_dependence_relations): Same.
17521 * tree-predcom.c (dump_chains): Same.
17522 (initialize_root_vars_lm): Same.
17523 (determine_unroll_factor): Same.
17524 (replace_phis_by_defined_names): Same.
17525 (insert_init_seqs): Same.
17526 (pcom_worker::tree_predictive_commoning_loop): Call
17527 auto_vec::to_vec_legacy.
17528 * tree-ssa-pre.c (insert_into_preds_of_block): Adjust by-value vec
17529 arguments to by-reference.
17530 * tree-ssa-threadbackward.c (populate_worklist): Same.
17531 (back_threader::resolve_def): Same.
17532 * tree-vect-data-refs.c (vect_check_nonzero_value): Same.
17533 (vect_enhance_data_refs_alignment): Same.
17534 (vect_check_lower_bound): Same.
17535 (vect_prune_runtime_alias_test_list): Same.
17536 (vect_permute_store_chain): Same.
17537 * tree-vect-slp-patterns.c (vect_normalize_conj_loc): Same.
17538 * tree-vect-stmts.c (vect_create_vectorized_demotion_stmts): Same.
17539 * tree-vectorizer.h (vect_permute_store_chain): Same.
17540 * vec.c (test_init): New function.
17541 (vec_c_tests): Call new function.
17542 * vec.h (vec): Declare ctors, dtor, and assignment.
17543 (auto_vec::vec_to_legacy): New function.
17544 (vec::copy): Adjust initialization.
17546 2021-08-05 H.J. Lu <hjl.tools@gmail.com>
17549 * config/i386/i386.c (ix86_can_inline_p): Ignore MASK_80387 if
17550 callee only uses GPRs.
17551 * config/i386/ia32intrin.h: Revert commit 5463cee2770.
17552 * config/i386/serializeintrin.h: Revert commit 71958f740f1.
17553 * config/i386/x86gprintrin.h: Add
17554 #pragma GCC target("general-regs-only") and #pragma GCC pop_options
17555 to disable non-GPR ISAs.
17557 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
17559 PR middle-end/101787
17560 * doc/md.texi (cond_ashl, cond_ashr, cond_lshr): Document.
17562 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
17564 * tree-vectorizer.h (vect_is_store_elt_extraction, vect_is_reduction)
17565 (vect_reduc_type, vect_embedded_comparison_type, vect_comparison_type)
17566 (vect_is_extending_load, vect_is_integer_truncation): New functions,
17567 moved from aarch64.c but given different names.
17568 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction)
17569 (aarch64_is_reduction, aarch64_reduc_type)
17570 (aarch64_embedded_comparison_type, aarch64_comparison_type)
17571 (aarch64_extending_load_p, aarch64_integer_truncation_p): Delete
17572 in favor of the above. Update callers accordingly.
17574 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
17577 * config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
17578 writing .cpu directive in asm output.
17579 * config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
17580 (arm_last_printed_arch_string): Delete.
17581 (arm_last-printed_fpu_string): Delete.
17582 (arm_configure_build_target): If use of floating-point/SIMD is
17583 disabled, remove all fp/simd related features from the target ISA.
17584 (last_arm_targ_options): New variable.
17585 (arm_print_asm_arch_directives): Add new parameters. Change order
17586 of emitted directives and handle all cases here.
17587 (arm_file_start): Always call arm_print_asm_arch_directives, move
17588 all generation of .arch/.arch_extension here.
17589 (arm_file_end): Call arm_print_asm_arch.
17590 (arm_declare_function_name): Call arm_print_asm_arch_directives
17591 instead of printing .arch/.fpu directives directly.
17593 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
17595 * config/arm/arm.c (arm_configure_build_target): Don't call
17596 arm_option_reconfigure_globals.
17597 (arm_option_restore): Call arm_option_reconfigure_globals after
17598 reconfiguring the target.
17599 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
17601 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
17603 * config/arm/arm.c (arm_configure_build_target): Ensure the target's
17604 arch_name is always set.
17606 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
17608 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
17609 of vec_select high-half from being added into Neon subtract
17612 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
17614 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
17615 of vec_select high-half from being added into Neon add cost.
17617 2021-08-05 Kewen Lin <linkw@linux.ibm.com>
17619 * cfgloop.h (loops_list::loops_list): Add one optional argument
17620 root and adjust accordingly, update loop tree walking and factor
17622 * cfgloop.c (loops_list::walk_loop_tree): ... this. New function.
17624 2021-08-05 Eric Botcazou <ebotcazou@gcc.gnu.org>
17626 PR tree-optimization/101626
17627 * tree-sra.c (propagate_subaccesses_from_rhs): Do not set the
17628 reverse scalar storage order on a pointer or vector component.
17630 2021-08-05 liuhongt <hongtao.liu@intel.com>
17632 * config/i386/sse.md (cond_<code><mode>): New expander.
17634 2021-08-05 liuhongt <hongtao.liu@intel.com>
17636 * config/i386/sse.md (cond_<code><mode>): New expander.
17638 2021-08-05 liuhongt <hongtao.liu@intel.com>
17640 * config/i386/sse.md (cond_<code><mode>): New expander.
17642 2021-08-04 David Malcolm <dmalcolm@redhat.com>
17645 * Makefile.in (ANALYZER_OBJS): Add analyzer/region-model-asm.o.
17647 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
17650 * config/i386/i386.h (STORE_MAX_PIECES): Allow 16/32/64 bytes
17651 only if TARGET_INTER_UNIT_MOVES_TO_VEC is true.
17653 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
17656 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
17657 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
17658 data with SSE register from one memory location to another.
17660 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
17662 * config/s390/s390.c (expand_perm_with_vpdi): New function.
17663 (vectorize_vec_perm_const_1): Call expand_perm_with_vpdi.
17664 * config/s390/vector.md (*vpdi1<mode>, @vpdi1<mode>): Enable a
17665 parameterized expander.
17666 (*vpdi4<mode>, @vpdi4<mode>): Likewise.
17668 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
17670 * config/s390/s390.c (MAX_VECT_LEN): Define macro.
17671 (struct expand_vec_perm_d): Define struct.
17672 (expand_perm_with_merge): New function.
17673 (vectorize_vec_perm_const_1): New function.
17674 (s390_vectorize_vec_perm_const): New function.
17675 (TARGET_VECTORIZE_VEC_PERM_CONST): Define target macro.
17677 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
17679 * config/s390/vector.md (V_HW_64): Remove mode iterator.
17680 (*vec_load_pair<mode>): Use V_HW_2 instead of V_HW_64.
17681 * config/s390/vx-builtins.md
17682 (vec_scatter_element<V_HW_2:mode>_SI): Use V_HW_2 instead of
17685 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
17687 * config/s390/s390.md (UNSPEC_VEC_PERMI): Remove constant
17689 * config/s390/vector.md (*vpdi1<mode>, *vpdi4<mode>): New pattern
17691 * config/s390/vx-builtins.md (*vec_permi<mode>): Emit generic rtx
17692 instead of an unspec.
17694 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
17696 * config/s390/s390-modes.def: Add more vector modes to support
17697 concatenation of two vectors.
17698 * config/s390/s390-protos.h (s390_expand_merge_perm_const): Add
17700 (s390_expand_merge): Likewise.
17701 * config/s390/s390.c (s390_expand_merge_perm_const): New function.
17702 (s390_expand_merge): New function.
17703 * config/s390/s390.md (UNSPEC_VEC_MERGEH, UNSPEC_VEC_MERGEL):
17704 Remove constant definitions.
17705 * config/s390/vector.md (V_HW_2): Add mode iterators.
17706 (VI_HW_4, V_HW_4): Rename VI_HW_4 to V_HW_4.
17707 (vec_2x_nelts, vec_2x_wide): New mode attributes.
17708 (*vmrhb, *vmrlb, *vmrhh, *vmrlh, *vmrhf, *vmrlf, *vmrhg, *vmrlg):
17709 New pattern definitions.
17710 (vec_widen_umult_lo_<mode>, vec_widen_umult_hi_<mode>)
17711 (vec_widen_smult_lo_<mode>, vec_widen_smult_hi_<mode>)
17712 (vec_unpacks_lo_v4sf, vec_unpacks_hi_v4sf, vec_unpacks_lo_v2df)
17713 (vec_unpacks_hi_v2df): Adjust expanders to emit non-unspec RTX for
17715 * config/s390/vx-builtins.md (V_HW_4): Remove mode iterator. Now
17717 (vec_mergeh<mode>, vec_mergel<mode>): Use s390_expand_merge to
17718 emit vec merge pattern.
17720 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
17722 * config/aarch64/aarch64.c (aarch64_strip_extend_vec_half):
17724 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent cost of
17725 vec_select high-half from being added into Neon multiply
17727 * rtlanal.c (vec_series_highpart_p): Define.
17728 * rtlanal.h (vec_series_highpart_p): Declare.
17730 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
17732 * config/aarch64/aarch64.c (aarch64_strip_duplicate_vec_elt):
17734 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent
17735 vec_select cost from being added into Neon multiply cost.
17737 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
17739 * tree-vect-loop.c (vect_better_loop_vinfo_p): Detect cases in
17740 which old_loop_vinfo is an epilogue loop that handles a constant
17741 number of iterations.
17743 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
17745 * tree-vect-loop.c (vect_analyze_loop): Print a dump message
17746 when a reanalyzed loop fails to be cheaper than the current
17749 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
17751 * config/aarch64/aarch64.c: Fix a typo.
17753 2021-08-04 Vincent Lefèvre <vincent-gcc@vinc17.net>
17755 PR gcov-profile/101773
17756 * gcov-io.c (gcov_close): Check return code of a fclose.
17758 2021-08-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
17761 * dwarf2out.c (dwarf2out_assembly_start): Emit a dummy
17762 .file statement when needed.
17764 2021-08-04 Richard Biener <rguenther@suse.de>
17766 * tree-vect-data-refs.c (vect_check_gather_scatter):
17767 Include widening conversions only when the result is
17768 still handed by native gather or the current offset
17769 size not already matches the data size.
17770 Also succeed analysis in case there's no native support,
17771 noted by a IFN_LAST ifn and a NULL decl.
17772 (vect_analyze_data_refs): Always consider gathers.
17773 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
17774 Test for no IFN gather rather than decl gather.
17775 * tree-vect-stmts.c (vect_model_load_cost): Pass in the
17776 gather-scatter info and cost emulated gathers accordingly.
17777 (vect_truncate_gather_scatter_offset): Properly test for
17779 (vect_use_strided_gather_scatters_p): Likewise.
17780 (get_load_store_type): Handle emulated gathers and its
17782 (vectorizable_load): Likewise. Emulate them by extracting
17783 scalar offsets, doing scalar loads and a vector construct.
17785 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
17788 * expr.c (op_by_pieces_d::op_by_pieces_d): Add a max_pieces
17789 argument to set m_max_size.
17790 (move_by_pieces_d): Pass MOVE_MAX_PIECES to op_by_pieces_d.
17791 (store_by_pieces_d): Pass STORE_MAX_PIECES to op_by_pieces_d.
17792 (compare_by_pieces_d): Pass COMPARE_MAX_PIECES to op_by_pieces_d.
17794 2021-08-04 Roger Sayle <roger@nextmovesoftware.com>
17795 Marc Glisse <marc.glisse@inria.fr>
17797 * match.pd (bit_ior, bit_xor): Canonicalize (X*C1)|(X*C2) and
17798 (X*C1)^(X*C2) as X*(C1+C2), and related variants, using
17799 tree_nonzero_bits to ensure that operands are bit-wise disjoint.
17801 2021-08-04 Richard Biener <rguenther@suse.de>
17803 * tree-ssa-forwprop.c (pass_forwprop::execute): Split
17804 out code to decompose vector loads ...
17805 (optimize_vector_load): ... here. Generalize it to
17806 handle intermediate widening and TARGET_MEM_REF loads
17807 and apply it to loads with a supported vector mode as well.
17809 2021-08-04 Richard Biener <rguenther@suse.de>
17811 PR tree-optimization/101756
17812 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Make sure
17813 the result of the reduction epilogue is compatible to the original
17816 2021-08-04 liuhongt <hongtao.liu@intel.com>
17819 * config/i386/i386.md (peephole2): Refine predicate from
17820 register_operand to general_reg_operand.
17822 2021-08-04 Aldy Hernandez <aldyh@redhat.com>
17824 * gimple-range-path.h (path_range_query::dump): Mark override.
17826 2021-08-04 Richard Biener <rguenther@suse.de>
17828 PR tree-optimization/101769
17829 * tree-tailcall.c (eliminate_tail_call): Add the created loop
17830 for the first recursion and return it via the new output parameter.
17831 (optimize_tail_call): Pass through new output param.
17832 (tree_optimize_tail_calls_1): After creating all latches,
17833 add the created loop to the loop tree. Do not mark loops for fixup.
17835 2021-08-04 Martin Liska <mliska@suse.cz>
17837 * doc/invoke.texi: Document threader-mode param.
17839 2021-08-04 liuhongt <hongtao.liu@intel.com>
17841 * config/i386/sse.md (cond_fma<mode>): New expander.
17842 (cond_fms<mode>): Ditto.
17843 (cond_fnma<mode>): Ditto.
17844 (cond_fnms<mode>): Ditto.
17846 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
17848 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode>): Use && instead of &.
17850 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
17852 * config/rs6000/constraints.md: Remove "e" from the list of available
17853 constraint characters.
17855 2021-08-03 Eugene Rozenfeld <erozen@microsoft.com>
17857 PR gcov-profile/71672
17858 * auto-profile.c (afdo_indirect_call): Fix setup of the historgram value for indirect calls.
17860 2021-08-03 Paul A. Clarke <pc@us.ibm.com>
17862 * config/rs6000/smmintrin.h (_mm_minpos_epu16): New.
17864 2021-08-03 H.J. Lu <hjl.tools@gmail.com>
17866 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): In 64-bit mode,
17867 try XMM31 to avoid vzeroupper.
17869 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17871 * doc/invoke.texi: Document -mtune=neoverse-512tvb and
17872 -mcpu=neoverse-512tvb.
17873 * config/aarch64/aarch64-cores.def (neoverse-512tvb): New entry.
17874 * config/aarch64/aarch64-tune.md: Regenerate.
17875 * config/aarch64/aarch64.c (neoverse512tvb_sve_vector_cost)
17876 (neoverse512tvb_sve_issue_info, neoverse512tvb_vec_issue_info)
17877 (neoverse512tvb_vector_cost, neoverse512tvb_tunings): New structures.
17878 (aarch64_adjust_body_cost_sve): Handle -mtune=neoverse-512tvb.
17879 (aarch64_adjust_body_cost): Likewise.
17881 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17883 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Only
17884 record issue information for operations that occur in the
17887 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17889 * config/aarch64/aarch64.c (aarch64_multiply_add_p): Add a vec_flags
17890 parameter. Detect cases in which an Advanced SIMD MLA would almost
17891 certainly require a MOV.
17892 (aarch64_count_ops): Update accordingly.
17894 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17896 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction): New
17897 function, split out from...
17898 (aarch64_detect_vector_stmt_subtype): ...here.
17899 (aarch64_add_stmt_cost): Treat extracting element 0 as free.
17901 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17903 * config/aarch64/aarch64-protos.h (sve_vec_cost):
17904 Add gather_load_x32_cost and gather_load_x64_cost.
17905 * config/aarch64/aarch64.c (generic_sve_vector_cost)
17906 (a64fx_sve_vector_cost, neoversev1_sve_vector_cost): Update
17907 accordingly, using the values given by the scalar_load * number
17908 of elements calculation that we used previously.
17909 (aarch64_detect_vector_stmt_subtype): Use the new fields.
17911 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17913 * config/aarch64/aarch64.c (aarch64_adjust_body_cost_sve): New
17914 function, split out from...
17915 (aarch64_adjust_body_cost): ...here.
17917 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17919 * config/aarch64/fractional-cost.h: New file.
17920 * config/aarch64/aarch64.c: Include <algorithm> (indirectly)
17921 and cost_fraction.h.
17922 (vec_cost_fraction): New typedef.
17923 (aarch64_detect_scalar_stmt_subtype): Use it for statement costs.
17924 (aarch64_detect_vector_stmt_subtype): Likewise.
17925 (aarch64_sve_adjust_stmt_cost, aarch64_adjust_stmt_cost): Likewise.
17926 (aarch64_estimate_min_cycles_per_iter): Use vec_cost_fraction
17928 (aarch64_adjust_body_cost): Likewise.
17929 (aarch64_test_cost_fraction): New function.
17930 (aarch64_run_selftests): Call it.
17932 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
17934 * config/aarch64/aarch64-protos.h (tune_params::sve_width): Turn
17936 * config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Update
17938 (aarch64_estimated_poly_value): Likewise. Use the least significant
17939 set bit for the minimum and likely values. Use the most significant
17940 set bit for the maximum value.
17942 2021-08-03 liuhongt <hongtao.liu@intel.com>
17944 * config/i386/sse.md (cond_<insn><mode>): New expander.
17945 (cond_mul<mode>): Ditto.
17947 2021-08-03 Kewen Lin <linkw@linux.ibm.com>
17949 * tree-cfg.c (move_sese_region_to_fn): Fix typos on dloop.
17951 2021-08-03 liuhongt <hongtao.liu@intel.com>
17953 * config/i386/sse.md (cond_<insn><mode>):New expander.
17954 (cond_mul<mode>): Ditto.
17955 (cond_div<mode>): Ditto.
17957 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
17959 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Also
17960 check stack_realign_needed for stack realignment.
17961 (ix86_legitimate_constant_p): Always allow CONST_WIDE_INT smaller
17962 than the largest integer supported by vector register.
17963 * config/i386/i386.h (MAX_MOVE_MAX): New. Set to 64.
17964 (MOVE_MAX): Set to bytes of the largest integer supported by
17966 (STORE_MAX_PIECES): New.
17968 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
17970 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
17971 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
17972 data from one memory location to another.
17974 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
17976 PR middle-end/90773
17977 * config/i386/i386.c (TARGET_GEN_MEMSET_SCRATCH_RTX): New.
17979 2021-08-02 Aldy Hernandez <aldyh@redhat.com>
17981 PR tree-optimization/101724
17982 * params.opt: Remove --param=threader-iterative.
17983 * tree-ssa-threadbackward.c (pass_thread_jumps::execute): Remove
17986 2021-08-02 Tom de Vries <tdevries@suse.de>
17988 PR middle-end/101665
17989 * doc/extend.texi (nonnull attribute): Improve documentation.
17991 2021-08-02 Andrew Pinski <apinski@marvell.com>
17993 PR rtl-optimization/101683
17994 * rtlanal.c (may_trap_p_1): Handle UNSIGNED_FIX.
17996 2021-08-02 Roger Sayle <roger@nextmovesoftware.com>
17998 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern):
17999 Renamed from cond_removal_in_popcount_clz_ctz_pattern.
18000 Add support for BSWAP, FFS, PARITY and CLRSB builtins.
18001 (tree_ssa_phiop_worker): Update call to function above.
18003 2021-08-01 H.J. Lu <hjl.tools@gmail.com>
18006 * config/i386/i386.md (bsr_rex64_1_zext): New.
18007 (combine splitter for constant - clzll): Replace gen_bsr_rex64_1
18008 with gen_bsr_rex64_1_zext.
18010 2021-07-31 Jakub Jelinek <jakub@redhat.com>
18013 * config/i386/i386.md (bsr_rex64_1, bsr_1, bsr_zext_1): New
18014 define_insn patterns.
18015 (*bsr_rex64_2, *bsr_2): New define_insn_and_split patterns.
18016 Add combine splitters for constant - clz.
18017 (clz<mode>2): Use a temporary pseudo for bsr result.
18019 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
18021 * config/rs6000/smmintrin.h (_mm_floor_pd, _mm_floor_ps,
18022 _mm_floor_sd, _mm_floor_ss): New.
18024 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
18026 * config/rs6000/smmintrin.h (_mm_ceil_pd, _mm_ceil_ps,
18027 _mm_ceil_sd, _mm_ceil_ss): New.
18029 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
18031 * config/rs6000/smmintrin.h (_mm_blend_pd, _mm_blendv_pd,
18032 _mm_blend_ps, _mm_blendv_ps): New.
18034 2021-07-30 Roger Sayle <roger@nextmovesoftware.com>
18035 Uroš Bizjak <ubizjak@gmail.com>
18037 * config/i386/i386.md (*dec_cmov<mode>): New define_insn_and_split
18038 to generate a conditional move using the carry flag after sub $1.
18039 (peephole2): Eliminate a register-to-register move by inverting
18040 the condition of a conditional move.
18042 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
18044 * config/mmix/mmix.md ("call", "call_value", "*call_real")
18045 ("*call_value_real"): Don't generate rtx mentioning the generic
18046 operands 1 and 2 to "call", and similarly for "call_value".
18047 * config/mmix/mmix.c (mmix_print_operand_punct_valid_p)
18048 (mmix_print_operand): Use '!' instead of 'p'.
18050 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
18052 * doc/md.texi (call): Correct information about operand 2.
18053 * config/mmix/mmix.md ("call", "call_value"): Remove fixed FIXMEs.
18055 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
18057 * range-op.cc (operator_trunc_mod::wi_fold): Fold constants.
18059 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
18061 * range-op.cc (operator_div::wi_fold): Return UNDEFINED for [0, 0] divisor.
18063 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
18065 * gimple-range-cache.cc (*::set_bb_range): Change const basic_block to
18066 const_basic_block..
18067 (*::get_bb_range): Ditto.
18068 (*::bb_range_p): Ditto.
18069 * gimple-range-cache.h: Change prototypes.
18071 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
18073 PR middle-end/90773
18074 * builtins.c (builtin_memcpy_read_str): Change the mode argument
18075 from scalar_int_mode to fixed_size_mode.
18076 (builtin_strncpy_read_str): Likewise.
18077 (gen_memset_value_from_prev): New function.
18078 (builtin_memset_read_str): Change the mode argument from
18079 scalar_int_mode to fixed_size_mode. Use gen_memset_value_from_prev
18080 and support CONST_VECTOR.
18081 (builtin_memset_gen_str): Likewise.
18082 (try_store_by_multiple_pieces): Use by_pieces_constfn to declare
18084 * builtins.h (builtin_strncpy_read_str): Replace scalar_int_mode
18085 with fixed_size_mode.
18086 (builtin_memset_read_str): Likewise.
18087 * expr.c (widest_int_mode_for_size): Renamed to ...
18088 (widest_fixed_size_mode_for_size): Add a bool argument to
18089 indicate if QI vector mode can be used.
18090 (by_pieces_ninsns): Call widest_fixed_size_mode_for_size
18091 instead of widest_int_mode_for_size.
18092 (pieces_addr::adjust): Change the mode argument from
18093 scalar_int_mode to fixed_size_mode.
18094 (op_by_pieces_d): Make m_len read-only. Add a bool member,
18095 m_qi_vector_mode, to indicate that QI vector mode can be used.
18096 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
18097 initialize m_qi_vector_mode. Call widest_fixed_size_mode_for_size
18098 instead of widest_int_mode_for_size.
18099 (op_by_pieces_d::get_usable_mode): Change the mode argument from
18100 scalar_int_mode to fixed_size_mode. Call
18101 widest_fixed_size_mode_for_size instead of
18102 widest_int_mode_for_size.
18103 (op_by_pieces_d::smallest_fixed_size_mode_for_size): New member
18104 function to return the smallest integer or QI vector mode.
18105 (op_by_pieces_d::run): Call widest_fixed_size_mode_for_size
18106 instead of widest_int_mode_for_size. Call
18107 smallest_fixed_size_mode_for_size instead of
18108 smallest_int_mode_for_size.
18109 (store_by_pieces_d::store_by_pieces_d): Add a bool argument to
18110 indicate that QI vector mode can be used and pass it to
18111 op_by_pieces_d::op_by_pieces_d.
18112 (can_store_by_pieces): Call widest_fixed_size_mode_for_size
18113 instead of widest_int_mode_for_size. Pass memsetp to
18114 widest_fixed_size_mode_for_size to support QI vector mode.
18115 Allow all CONST_VECTORs for memset if vec_duplicate is supported.
18116 (store_by_pieces): Pass memsetp to
18117 store_by_pieces_d::store_by_pieces_d.
18118 (clear_by_pieces_1): Removed.
18119 (clear_by_pieces): Replace clear_by_pieces_1 with
18120 builtin_memset_read_str and pass true to store_by_pieces_d to
18121 support vector mode broadcast.
18122 (string_cst_read_str): Change the mode argument from
18123 scalar_int_mode to fixed_size_mode.
18124 * expr.h (by_pieces_constfn): Change scalar_int_mode to
18126 (by_pieces_prev): Likewise.
18127 * rtl.h (lowpart_subreg_regno): New.
18128 * rtlanal.c (lowpart_subreg_regno): New. A wrapper around
18129 simplify_subreg_regno.
18130 * target.def (gen_memset_scratch_rtx): New hook.
18131 * doc/tm.texi.in: Add TARGET_GEN_MEMSET_SCRATCH_RTX.
18132 * doc/tm.texi: Regenerated.
18134 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
18137 * config/mips/mips.c (mips_atomic_assign_expand_fenv): Use
18138 TARGET_EXPR instead of MODIFY_EXPR.
18140 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
18143 * config/mips/mips-protos.h (mips_expand_vec_cmp_expr): Declare.
18144 * config/mips/mips.c (mips_expand_vec_cmp_expr): New function.
18145 * config/mips/mips-msa.md (vec_cmp<MSA:mode><mode_i>): New
18147 (vec_cmpu<IMSA:mode><mode_i>): New expander.
18149 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
18152 * config/i386/i386-options.c (ix86_option_override_internal):
18153 Don't enable LZCNT/POPCNT if they have been disabled explicitly.
18155 2021-07-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
18158 * config/arm/arm_neon.h (vld1_p64): Replace call to builtin by
18159 explicitly dereferencing __a.
18160 (vld1_s64): Likewise.
18161 (vld1_u64): Likewise.
18162 * config/arm/arm_neon_builtins.def (vld1): Remove entry for di
18163 and change to VAR13.
18165 2021-07-30 Aldy Hernandez <aldyh@redhat.com>
18167 * gimple-loop-versioning.cc (lv_dom_walker::lv_dom_walker): Remove
18168 use of m_range_analyzer.
18169 (loop_versioning::lv_dom_walker::before_dom_children): Same.
18170 (loop_versioning::lv_dom_walker::after_dom_children): Remove.
18171 (loop_versioning::prune_loop_conditions): Replace vr_values use
18172 with range_query interface.
18173 (pass_loop_versioning::execute): Use ranger.
18175 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
18178 * ipa-devirt.c (ipa_odr_read_section): Compare the precision of
18179 enum values, and emit a warning if they mismatch.
18181 2021-07-30 Kewen Lin <linkw@linux.ibm.com>
18183 * cfgloop.h (as_const): New function.
18184 (class loop_iterator): Rename to ...
18185 (class loops_list): ... this.
18186 (loop_iterator::next): Rename to ...
18187 (loops_list::Iter::fill_curr_loop): ... this and adjust.
18188 (loop_iterator::loop_iterator): Rename to ...
18189 (loops_list::loops_list): ... this and adjust.
18190 (loops_list::Iter): New class.
18191 (loops_list::iterator): New type.
18192 (loops_list::const_iterator): New type.
18193 (loops_list::begin): New function.
18194 (loops_list::end): Likewise.
18195 (loops_list::begin const): Likewise.
18196 (loops_list::end const): Likewise.
18197 (FOR_EACH_LOOP): Remove.
18198 (FOR_EACH_LOOP_FN): Remove.
18199 * cfgloop.c (flow_loops_dump): Adjust FOR_EACH_LOOP* with range-based
18200 for loop with loops_list instance.
18201 (sort_sibling_loops): Likewise.
18202 (disambiguate_loops_with_multiple_latches): Likewise.
18203 (verify_loop_structure): Likewise.
18204 * cfgloopmanip.c (create_preheaders): Likewise.
18205 (force_single_succ_latches): Likewise.
18206 * config/aarch64/falkor-tag-collision-avoidance.c
18207 (execute_tag_collision_avoidance): Likewise.
18208 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Likewise.
18209 * config/s390/s390.c (s390_adjust_loops): Likewise.
18210 * doc/loop.texi: Likewise.
18211 * gimple-loop-interchange.cc (pass_linterchange::execute): Likewise.
18212 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
18213 * gimple-loop-versioning.cc (loop_versioning::analyze_blocks): Likewise.
18214 (loop_versioning::make_versioning_decisions): Likewise.
18215 * gimple-ssa-split-paths.c (split_paths): Likewise.
18216 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
18217 * graphite.c (canonicalize_loop_form): Likewise.
18218 (graphite_transform_loops): Likewise.
18219 * ipa-fnsummary.c (analyze_function_body): Likewise.
18220 * ipa-pure-const.c (analyze_function): Likewise.
18221 * loop-doloop.c (doloop_optimize_loops): Likewise.
18222 * loop-init.c (loop_optimizer_finalize): Likewise.
18223 (fix_loop_structure): Likewise.
18224 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
18225 (move_loop_invariants): Likewise.
18226 * loop-unroll.c (decide_unrolling): Likewise.
18227 (unroll_loops): Likewise.
18228 * modulo-sched.c (sms_schedule): Likewise.
18229 * predict.c (predict_loops): Likewise.
18230 (pass_profile::execute): Likewise.
18231 * profile.c (branch_prob): Likewise.
18232 * sel-sched-ir.c (sel_finish_pipelining): Likewise.
18233 (sel_find_rgns): Likewise.
18234 * tree-cfg.c (replace_loop_annotate): Likewise.
18235 (replace_uses_by): Likewise.
18236 (move_sese_region_to_fn): Likewise.
18237 * tree-if-conv.c (pass_if_conversion::execute): Likewise.
18238 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
18239 * tree-parloops.c (parallelize_loops): Likewise.
18240 * tree-predcom.c (tree_predictive_commoning): Likewise.
18241 * tree-scalar-evolution.c (scev_initialize): Likewise.
18242 (scev_reset): Likewise.
18243 * tree-ssa-dce.c (find_obviously_necessary_stmts): Likewise.
18244 * tree-ssa-live.c (remove_unused_locals): Likewise.
18245 * tree-ssa-loop-ch.c (ch_base::copy_headers): Likewise.
18246 * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
18247 (tree_ssa_lim_initialize): Likewise.
18248 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Likewise.
18249 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize): Likewise.
18250 * tree-ssa-loop-manip.c (get_loops_exits): Likewise.
18251 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations): Likewise.
18252 (free_numbers_of_iterations_estimates): Likewise.
18253 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Likewise.
18254 * tree-ssa-loop-split.c (tree_ssa_split_loops): Likewise.
18255 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Likewise.
18256 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
18257 (pass_scev_cprop::execute): Likewise.
18258 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
18259 * tree-ssa-sccvn.c (do_rpo_vn): Likewise.
18260 * tree-ssa-threadupdate.c
18261 (jump_thread_path_registry::thread_through_all_blocks): Likewise.
18262 * tree-vectorizer.c (vectorize_loops): Likewise.
18263 * tree-vrp.c (vrp_asserts::find_assert_locations): Likewise.
18265 2021-07-29 Hans-Peter Nilsson <hp@bitrange.com>
18267 * config/mmix/mmix.c (mmix_function_arg_1): Avoid
18268 generating a VOIDmode register for e.g the
18269 function_arg_info::end_marker.
18271 2021-07-29 Jeff Law <jeffreyalaw@gmail.com>
18273 * config/h8300/h8300-modes.def: Add CCZ, CCV and CCC, drop CCZNV.
18274 * config/h8300/h8300.md (H8cc mode iterator): Add CCZ.
18275 (cc mode_attr): Similarly.
18276 (ccz subst_attr): Similarly.
18277 * config/h8300/jumpcall.md: Add new patterns for branch-on-bit.
18278 * config/h8300/testcompare.md: Remove various cc0 based patterns
18279 that had been commented out. Add pattern to set CCZ from a bit
18282 2021-07-29 Thomas Schwinge <thomas@codesourcery.com>
18283 Julian Brown <julian@codesourcery.com>
18284 Kwok Cheung Yeung <kcy@codesourcery.com>
18286 * omp-offload.c (oacc_loop_xform_head_tail, oacc_loop_process):
18287 'update_stmt' after modification.
18288 (pass_oacc_loop_designation): New function, extracted out of...
18289 (pass_oacc_device_lower): ... this.
18290 (pass_data_oacc_loop_designation, pass_oacc_loop_designation)
18291 (make_pass_oacc_loop_designation): New
18292 * passes.def: Add it.
18293 * tree-parloops.c (create_parallel_loop): Adjust.
18294 * tree-pass.h (make_pass_oacc_loop_designation): New.
18296 2021-07-29 Aldy Hernandez <aldyh@redhat.com>
18298 * flag-types.h (enum threader_mode): New.
18299 * params.opt: Add entry for --param=threader-mode.
18300 * tree-ssa-threadbackward.c (THREADER_ITERATIVE_MODE): New.
18301 (class back_threader): New.
18302 (back_threader::back_threader): New.
18303 (back_threader::~back_threader): New.
18304 (back_threader::maybe_register_path): New.
18305 (back_threader::find_taken_edge): New.
18306 (back_threader::find_taken_edge_switch): New.
18307 (back_threader::find_taken_edge_cond): New.
18308 (back_threader::resolve_def): New.
18309 (back_threader::resolve_phi): New.
18310 (back_threader::find_paths_to_names): New.
18311 (back_threader::find_paths): New.
18314 (thread_jumps::find_jump_threads_backwards): Call ranger threader.
18315 (thread_jumps::find_jump_threads_backwards_with_ranger): New.
18316 (pass_thread_jumps::execute): Abstract out code...
18317 (try_thread_blocks): ...here.
18318 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
18319 Abstract out threading candidate code to...
18320 (single_succ_to_potentially_threadable_block): ...here.
18321 * tree-ssa-threadedge.h (single_succ_to_potentially_threadable_block):
18323 * tree-ssa-threadupdate.c (register_jump_thread): Return boolean.
18324 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
18325 Return bool from register_jump_thread.
18327 2021-07-29 Andreas Krebbel <krebbel@linux.ibm.com>
18329 * target.def: in0 and in1 do not need to be registers.
18330 * doc/tm.texi: Regenerate.
18332 2021-07-29 liuhongt <hongtao.liu@intel.com>
18335 * config/i386/i386.c (ix86_widen_mult_cost): New function.
18336 (ix86_add_stmt_cost): Use ix86_widen_mult_cost for
18339 2021-07-29 Jiufu Guo <guojiufu@linux.ibm.com>
18342 * config/rs6000/rs6000.c (TARGET_PREFERRED_DOLOOP_MODE): New hook.
18343 (rs6000_preferred_doloop_mode): New hook.
18344 * doc/tm.texi: Regenerate.
18345 * doc/tm.texi.in: Add hook preferred_doloop_mode.
18346 * target.def (preferred_doloop_mode): New hook.
18347 * targhooks.c (default_preferred_doloop_mode): New hook.
18348 * targhooks.h (default_preferred_doloop_mode): New hook.
18349 * tree-ssa-loop-ivopts.c (compute_doloop_base_on_mode): New function.
18350 (add_iv_candidate_for_doloop): Call targetm.preferred_doloop_mode
18351 and compute_doloop_base_on_mode.
18353 2021-07-28 Martin Sebor <msebor@redhat.com>
18355 PR middle-end/101494
18356 * tree-ssa-uninit.c (maybe_warn_operand): Correct object offset
18357 and size computation.
18359 2021-07-28 Martin Sebor <msebor@redhat.com>
18361 PR middle-end/101601
18362 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Remove
18364 Handle pointers to functions.
18366 2021-07-28 Martin Sebor <msebor@redhat.com>
18368 * Makefile.in (OBJS): Add gimple-ssa-warn-access.o and pointer-query.o.
18369 * attribs.h (fndecl_dealloc_argno): Move fndecl_dealloc_argno to tree.h.
18370 * builtins.c (compute_objsize_r): Move to pointer-query.cc.
18371 (access_ref::access_ref): Same.
18372 (access_ref::phi): Same.
18373 (access_ref::get_ref): Same.
18374 (access_ref::size_remaining): Same.
18375 (access_ref::offset_in_range): Same.
18376 (access_ref::add_offset): Same.
18377 (access_ref::inform_access): Same.
18378 (ssa_name_limit_t::visit_phi): Same.
18379 (ssa_name_limit_t::leave_phi): Same.
18380 (ssa_name_limit_t::next): Same.
18381 (ssa_name_limit_t::next_phi): Same.
18382 (ssa_name_limit_t::~ssa_name_limit_t): Same.
18383 (pointer_query::pointer_query): Same.
18384 (pointer_query::get_ref): Same.
18385 (pointer_query::put_ref): Same.
18386 (pointer_query::flush_cache): Same.
18387 (warn_string_no_nul): Move to gimple-ssa-warn-access.cc.
18388 (check_nul_terminated_array): Same.
18389 (unterminated_array): Same.
18390 (maybe_warn_for_bound): Same.
18391 (check_read_access): Same.
18392 (warn_for_access): Same.
18393 (get_size_range): Same.
18394 (check_access): Same.
18395 (gimple_call_alloc_size): Move to tree.c.
18396 (gimple_parm_array_size): Move to pointer-query.cc.
18397 (get_offset_range): Same.
18398 (gimple_call_return_array): Same.
18399 (handle_min_max_size): Same.
18400 (handle_array_ref): Same.
18401 (handle_mem_ref): Same.
18402 (compute_objsize): Same.
18403 (gimple_call_alloc_p): Move to gimple-ssa-warn-access.cc.
18404 (call_dealloc_argno): Same.
18405 (fndecl_dealloc_argno): Same.
18406 (new_delete_mismatch_p): Same.
18407 (matching_alloc_calls_p): Same.
18408 (warn_dealloc_offset): Same.
18409 (maybe_emit_free_warning): Same.
18410 * builtins.h (check_nul_terminated_array): Move to
18411 gimple-ssa-warn-access.h.
18412 (check_nul_terminated_array): Same.
18413 (warn_string_no_nul): Same.
18414 (unterminated_array): Same.
18415 (class ssa_name_limit_t): Same.
18416 (class pointer_query): Same.
18417 (struct access_ref): Same.
18418 (class range_query): Same.
18419 (struct access_data): Same.
18420 (gimple_call_alloc_size): Same.
18421 (gimple_parm_array_size): Same.
18422 (compute_objsize): Same.
18423 (class access_data): Same.
18424 (maybe_emit_free_warning): Same.
18425 * calls.c (initialize_argument_information): Remove call to
18426 maybe_emit_free_warning.
18427 * gimple-array-bounds.cc: Include new header..
18428 * gimple-fold.c: Same.
18429 * gimple-ssa-sprintf.c: Same.
18430 * gimple-ssa-warn-restrict.c: Same.
18431 * passes.def: Add pass_warn_access.
18432 * tree-pass.h (make_pass_warn_access): Declare.
18433 * tree-ssa-strlen.c: Include new headers.
18434 * tree.c (fndecl_dealloc_argno): Move here from builtins.c.
18435 * tree.h (fndecl_dealloc_argno): Move here from attribs.h.
18436 * gimple-ssa-warn-access.cc: New file.
18437 * gimple-ssa-warn-access.h: New file.
18438 * pointer-query.cc: New file.
18439 * pointer-query.h: New file.
18441 2021-07-28 Jakub Jelinek <jakub@redhat.com>
18443 PR middle-end/101624
18444 * ubsan.c (maybe_instrument_pointer_overflow,
18445 instrument_object_size): Only test DECL_REGISTER on VAR_DECLs,
18446 PARM_DECLs or RESULT_DECLs.
18447 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
18449 2021-07-28 Jakub Jelinek <jakub@redhat.com>
18451 PR middle-end/101642
18452 * match.pd (bswap16 (x) == bswap16 (y)): Cast both operands
18453 to type of bswap16 for comparison.
18454 (bswap16 (x) == cst): Cast bswap16 operand to type of cst.
18456 2021-07-28 Richard Biener <rguenther@suse.de>
18458 PR tree-optimization/101615
18459 * tree-vect-slp.c (vect_optimize_slp): Materialize permutes
18460 at CTOR SLP graph entries.
18462 2021-07-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18464 * config/aarch64/aarch64.md (*extend<SHORT:mode><GPI:mode>2_aarch64):
18465 Add "r,w" alternative.
18467 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
18470 * config/i386/i386.c (ix86_avx_u128_mode_needed): Don't set
18471 AVX_U128_DIRTY when all bits are zero.
18473 2021-07-28 Richard Biener <rguenther@suse.de>
18475 PR tree-optimization/101615
18476 * tree-vect-slp.c (vect_optimize_slp): Pre-existing vector
18477 external nodes cannot be permuted so make them perm_out 0.
18479 2021-07-28 Andrew Stubbs <ams@codesourcery.com>
18482 * config.in: Regenerate.
18483 * config/gcn/gcn-hsa.h (A_FIJI): New define.
18484 (A_900): New define.
18485 (A_906): New define.
18486 (A_908): New define.
18487 (ASM_SPEC): Use A_FIJI, A_900, A_906 and A_908.
18488 * config/gcn/gcn.c (output_file_start): Adjust attributes according
18489 to the assembler capabilities.
18490 * config/gcn/mkoffload.c (main): Likewise.
18491 * configure: Regenerate.
18492 * configure.ac: Add tests for LLVM assembler attribute features.
18494 2021-07-28 Andrew MacLeod <amacleod@redhat.com>
18496 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Check for
18497 cond_false and cond_true on branches.
18499 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
18501 * config/aarch64/aarch64.c (aarch64_gen_adjusted_ldpstp): use
18504 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
18506 * alias.c (init_alias_analysis): Don't skip prologue/epilogue.
18508 2021-07-28 Jakub Jelinek <jakub@redhat.com>
18511 * config/i386/sse.md (vashr<mode>3): Split into vashrv8di3 expander
18512 and vashrv4di3 expander, where the latter requires just TARGET_AVX2
18513 and has special !TARGET_AVX512VL expansion.
18514 (vashrv2di3<mask_name>): Rename to ...
18515 (vashrv2di3): ... this. Change condition to TARGET_XOP || TARGET_AVX2
18516 and add special !TARGET_XOP && !TARGET_AVX512VL expansion.
18518 2021-07-28 Martin Uecker <muecker@gwdg.de>
18520 * calls.c (maybe_warn_rdwr_sizes): Correct argument
18521 numbers in warning that were switched.
18523 2021-07-28 Kewen Lin <linkw@linux.ibm.com>
18525 PR tree-optimization/101596
18526 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Fix wrong check
18527 by using new_type's precision instead.
18529 2021-07-28 liuhongt <hongtao.liu@intel.com>
18532 * config/i386/i386.h (processor_costs): Add new member
18534 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
18535 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
18536 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
18537 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
18538 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
18539 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
18540 generic_cost, core_cost): Initialize integer_to_sse same value
18542 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
18543 * config/i386/i386.c (ix86_builtin_vectorization_cost):
18544 Use integer_to_sse instead of sse_op to calculate the cost of
18547 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
18549 * config/rs6000/rs6000-gen-builtins.c (write_ovld_static_init): New
18551 (write_init_file): Call write_ovld_static_init.
18553 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
18555 * config/rs6000/rs6000-gen-builtins.c (write_bif_static_init): New
18557 (write_init_file): Call write_bif_static_init.
18559 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
18561 * config/rs6000/rs6000-gen-builtins.c (typemap): New struct.
18562 (TYPE_MAP_SIZE): New macro.
18563 (type_map): New initialized variable.
18564 (typemap_cmp): New function.
18565 (write_type_node): Likewise.
18566 (write_fntype_init): Implement.
18568 2021-07-27 Martin Sebor <msebor@redhat.com>
18570 PR tree-optimization/101584
18571 * tree-ssa-uninit.c (builtin_call_nomodifying_p): New function.
18572 (check_defs): Call it.
18574 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
18576 * tree-ssa-dom.c (dom_jump_threader_simplifier):
18577 Put avail_exprs_stack in the class, instead of passing it to
18578 jump_threader_simplifier.
18579 (dom_jump_threader_simplifier::simplify): Add state argument.
18580 (dom_opt_dom_walker): Add state.
18581 (pass_dominator::execute): Pass state to threader.
18582 (dom_opt_dom_walker::before_dom_children): Use state.
18583 * tree-ssa-threadedge.c (jump_threader::jump_threader): Replace
18584 arguments by state.
18585 (jump_threader::record_temporary_equivalences_from_phis):
18586 Register equivalences through the state variable.
18587 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
18588 Record ranges in a statement through the state variable.
18589 (jump_threader::simplify_control_stmt_condition): Pass state to
18591 (jump_threader::simplify_control_stmt_condition_1): Same.
18592 (jump_threader::thread_around_empty_blocks): Remove obsolete
18594 (jump_threader::thread_through_normal_block): Record equivalences
18595 on edge through the state variable.
18596 (jump_threader::thread_across_edge): Abstract state pushing.
18597 (jt_state::jt_state): New.
18598 (jt_state::push): New.
18599 (jt_state::pop): New.
18600 (jt_state::register_equiv): New.
18601 (jt_state::record_ranges_from_stmt): New.
18602 (jt_state::register_equivs_on_edge): New.
18603 (jump_threader_simplifier::jump_threader_simplifier): Move from
18605 (jump_threader_simplifier::simplify): Add state argument.
18606 * tree-ssa-threadedge.h (class jt_state): New.
18607 (class jump_threader): Add state to constructor.
18608 (class jump_threader_simplifier): Add state to simplify. Remove
18609 avail_exprs_stack from class.
18610 * tree-vrp.c (vrp_jump_threader_simplifier::simplify): Add state
18612 (vrp_jump_threader::vrp_jump_threader): Add state.
18613 (vrp_jump_threader::~vrp_jump_threader): Cleanup state.
18615 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
18617 * Makefile.in (OBJS): Add gimple-range-path.o.
18618 * gimple-range-path.cc: New file.
18619 * gimple-range-path.h: New file.
18621 2021-07-27 Jonathan Wright <jonathan.wright@arm.com>
18623 * config/aarch64/aarch64-simd.md: Push sign/zero-extension
18624 inside vec_duplicate for all patterns.
18625 * simplify-rtx.c (simplify_context::simplify_unary_operation_1):
18626 Push sign/zero-extension inside vec_duplicate.
18628 2021-07-27 Richard Biener <rguenther@suse.de>
18630 PR tree-optimization/101573
18631 * tree-ssa-uninit.c (warn_uninit_phi_uses): New function
18632 looking at uninitialized PHI arg defs in some constrained cases.
18633 (warn_uninitialized_vars): Call it.
18634 (execute_early_warn_uninitialized): Calculate dominators.
18636 2021-07-27 Richard Biener <rguenther@suse.de>
18638 PR tree-optimization/39821
18639 * tree-vect-stmts.c (vect_model_promotion_demotion_cost): Use
18640 vector_stmt for widening arithmetic.
18641 (vectorizable_conversion): Adjust.
18643 2021-07-27 Martin Jambor <mjambor@suse.cz>
18645 * cgraph.h (ipa_replace_map): New field force_load_ref.
18646 * ipa-prop.h (ipa_param_descriptor): Reduce precision of move_cost,
18647 aded new flag load_dereferenced, adjusted comments.
18648 (ipa_get_param_dereferenced): New function.
18649 (ipa_set_param_dereferenced): Likewise.
18650 * cgraphclones.c (cgraph_node::create_virtual_clone): Follow it.
18651 * ipa-cp.c: Include gimple.h.
18652 (ipcp_discover_new_direct_edges): Take into account dereferenced flag.
18653 (get_replacement_map): New parameter force_load_ref, set the
18654 appropriate flag in ipa_replace_map if set.
18655 (struct symbol_and_index_together): New type.
18656 (adjust_refs_in_act_callers): New function.
18657 (adjust_references_in_caller): Likewise.
18658 (create_specialized_node): When appropriate, call
18659 adjust_references_in_caller and force only load references.
18660 * ipa-prop.c (load_from_dereferenced_name): New function.
18661 (ipa_analyze_controlled_uses): Also detect loads from a
18662 dereference, harden testing of call statements.
18663 (ipa_write_node_info): Stream the dereferenced flag.
18664 (ipa_read_node_info): Likewise.
18665 (ipa_set_jf_constant): Also create refdesc when jump function
18666 references a variable.
18667 (cgraph_node_for_jfunc): Rename to symtab_node_for_jfunc, work
18668 also on references of variables and return a symtab_node. Adjust
18670 (propagate_controlled_uses): Also remove references to VAR_DECLs.
18672 2021-07-27 Jakub Jelinek <jakub@redhat.com>
18674 PR middle-end/101586
18675 * gimple-fold.c (clear_padding_type): Ignore FIELD_DECLs with byte
18676 positions above or equal to sz except for diagnostics of flexible
18679 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
18681 PR tree-optimization/78888
18682 * gimple-range-fold.cc (get_letter_range): New.
18683 (fold_using_range::range_of_builtin_call): Call get_letter_range.
18685 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
18687 PR tree-optimization/78888
18688 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Add cases
18689 for CFN_BUILT_IN_TOUPPER and CFN_BUILT_IN_TOLOWER.
18691 2021-07-26 Roger Sayle <roger@nextmovesoftware.com>
18692 Marc Glisse <marc.glisse@inria.fr>
18694 * match.pd (rotate): Simplify equality/inequality of rotations.
18695 (bswap): Simplify equality/inequality tests of byte swapping.
18697 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
18699 * range-op.cc (operator_bitwise_xor::op1_op2_relation_effect):
18702 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
18704 * range-op.cc (operator_lshift::fold_range): Pass rel to
18705 base class fold_range.
18706 (operator_rshift::fold_range): Same.
18708 2021-07-26 Ashimida <ashimida@linux.alibaba.com>
18711 * toplev.h (min_align_loops_log): Remove declaration.
18712 (min_align_jumps_log, min_align_labels_log): Likewise.
18713 (min_align_functions_log): Likewise.
18715 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
18717 * tree-vrp.c (vrp_simplify_cond_using_ranges): Rename vr_values
18719 (execute_vrp): Abstract out simplification of conditionals...
18720 (simplify_casted_conds): ...here.
18722 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
18724 * gimple-array-bounds.cc (array_bounds_checker::get_value_range):
18725 Add gimple argument.
18726 (array_bounds_checker::check_array_ref): Same.
18727 (array_bounds_checker::check_addr_expr): Same.
18728 (array_bounds_checker::check_array_bounds): Pass statement to
18729 check_array_bounds and check_addr_expr.
18730 * gimple-array-bounds.h (check_array_bounds): Add gimple argument.
18731 (check_addr_expr): Same.
18732 (get_value_range): Same.
18734 2021-07-26 Tamar Christina <tamar.christina@arm.com>
18736 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): Rename to..
18737 (sdot_prod, udot_prod): ... This.
18738 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): Merged
18740 (<sur>dot_prod<vsi2qi>): ... this.
18741 (aarch64_<sur>dot_lane<vsi2qi>, aarch64_<sur>dot_laneq<vsi2qi>):
18742 Change operands order.
18743 (<sur>sadv16qi): Use new operands order.
18744 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32,
18745 vdotq_s32): Use new RTL ordering.
18747 2021-07-26 Tamar Christina <tamar.christina@arm.com>
18749 * config/aarch64/aarch64-builtins.c (TYPES_TERNOP_SUSS,
18750 aarch64_types_ternop_suss_qualifiers): New.
18751 * config/aarch64/aarch64-simd-builtins.def (usdot_prod): Use it.
18752 * config/aarch64/aarch64-simd.md (usdot_prod<vsi2qi>): Re-organize RTL.
18753 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Use it.
18755 2021-07-23 Jakub Jelinek <jakub@redhat.com>
18757 PR rtl-optimization/101562
18758 * expmed.c (store_integral_bit_field): Only use movstrict_optab
18759 if the operand isn't paradoxical.
18761 2021-07-23 Aldy Hernandez <aldyh@redhat.com>
18763 * gimple-array-bounds.h (class array_bounds_checker): Change
18764 ranges type to range_query.
18766 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18768 * config/aarch64/arm_neon.h (vst1_s64_x2): Use
18769 __builtin_memcpy instead of constructing
18770 __builtin_aarch64_simd_oi one vector at a time.
18771 (vst1_u64_x2): Likewise.
18772 (vst1_f64_x2): Likewise.
18773 (vst1_s8_x2): Likewise.
18774 (vst1_p8_x2): Likewise.
18775 (vst1_s16_x2): Likewise.
18776 (vst1_p16_x2): Likewise.
18777 (vst1_s32_x2): Likewise.
18778 (vst1_u8_x2): Likewise.
18779 (vst1_u16_x2): Likewise.
18780 (vst1_u32_x2): Likewise.
18781 (vst1_f16_x2): Likewise.
18782 (vst1_f32_x2): Likewise.
18783 (vst1_p64_x2): Likewise.
18784 (vst1q_s8_x2): Likewise.
18785 (vst1q_p8_x2): Likewise.
18786 (vst1q_s16_x2): Likewise.
18787 (vst1q_p16_x2): Likewise.
18788 (vst1q_s32_x2): Likewise.
18789 (vst1q_s64_x2): Likewise.
18790 (vst1q_u8_x2): Likewise.
18791 (vst1q_u16_x2): Likewise.
18792 (vst1q_u32_x2): Likewise.
18793 (vst1q_u64_x2): Likewise.
18794 (vst1q_f16_x2): Likewise.
18795 (vst1q_f32_x2): Likewise.
18796 (vst1q_f64_x2): Likewise.
18797 (vst1q_p64_x2): Likewise.
18799 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18801 * config/aarch64/arm_neon.h (vst1_s64_x3): Use
18802 __builtin_memcpy instead of constructing
18803 __builtin_aarch64_simd_ci one vector at a time.
18804 (vst1_u64_x3): Likewise.
18805 (vst1_f64_x3): Likewise.
18806 (vst1_s8_x3): Likewise.
18807 (vst1_p8_x3): Likewise.
18808 (vst1_s16_x3): Likewise.
18809 (vst1_p16_x3): Likewise.
18810 (vst1_s32_x3): Likewise.
18811 (vst1_u8_x3): Likewise.
18812 (vst1_u16_x3): Likewise.
18813 (vst1_u32_x3): Likewise.
18814 (vst1_f16_x3): Likewise.
18815 (vst1_f32_x3): Likewise.
18816 (vst1_p64_x3): Likewise.
18817 (vst1q_s8_x3): Likewise.
18818 (vst1q_p8_x3): Likewise.
18819 (vst1q_s16_x3): Likewise.
18820 (vst1q_p16_x3): Likewise.
18821 (vst1q_s32_x3): Likewise.
18822 (vst1q_s64_x3): Likewise.
18823 (vst1q_u8_x3): Likewise.
18824 (vst1q_u16_x3): Likewise.
18825 (vst1q_u32_x3): Likewise.
18826 (vst1q_u64_x3): Likewise.
18827 (vst1q_f16_x3): Likewise.
18828 (vst1q_f32_x3): Likewise.
18829 (vst1q_f64_x3): Likewise.
18830 (vst1q_p64_x3): Likewise.
18832 2021-07-23 H.J. Lu <hjl.tools@gmail.com>
18835 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): Don't return
18836 hard register when LRA is in progress.
18838 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18840 * config/aarch64/arm_neon.h (vst1_s8_x4): Use
18841 __builtin_memcpy instead of using a union.
18842 (vst1q_s8_x4): Likewise.
18843 (vst1_s16_x4): Likewise.
18844 (vst1q_s16_x4): Likewise.
18845 (vst1_s32_x4): Likewise.
18846 (vst1q_s32_x4): Likewise.
18847 (vst1_u8_x4): Likewise.
18848 (vst1q_u8_x4): Likewise.
18849 (vst1_u16_x4): Likewise.
18850 (vst1q_u16_x4): Likewise.
18851 (vst1_u32_x4): Likewise.
18852 (vst1q_u32_x4): Likewise.
18853 (vst1_f16_x4): Likewise.
18854 (vst1q_f16_x4): Likewise.
18855 (vst1_f32_x4): Likewise.
18856 (vst1q_f32_x4): Likewise.
18857 (vst1_p8_x4): Likewise.
18858 (vst1q_p8_x4): Likewise.
18859 (vst1_p16_x4): Likewise.
18860 (vst1q_p16_x4): Likewise.
18861 (vst1_s64_x4): Likewise.
18862 (vst1_u64_x4): Likewise.
18863 (vst1_p64_x4): Likewise.
18864 (vst1q_s64_x4): Likewise.
18865 (vst1q_u64_x4): Likewise.
18866 (vst1q_p64_x4): Likewise.
18867 (vst1_f64_x4): Likewise.
18868 (vst1q_f64_x4): Likewise.
18870 2021-07-23 Jonathan Wrightt <jonathan.wright@arm.com>
18872 * config/aarch64/arm_neon.h (vst2_s64): Use __builtin_memcpy
18873 instead of constructing __builtin_aarch64_simd_oi one vector
18875 (vst2_u64): Likewise.
18876 (vst2_f64): Likewise.
18877 (vst2_s8): Likewise.
18878 (vst2_p8): Likewise.
18879 (vst2_s16): Likewise.
18880 (vst2_p16): Likewise.
18881 (vst2_s32): Likewise.
18882 (vst2_u8): Likewise.
18883 (vst2_u16): Likewise.
18884 (vst2_u32): Likewise.
18885 (vst2_f16): Likewise.
18886 (vst2_f32): Likewise.
18887 (vst2_p64): Likewise.
18888 (vst2q_s8): Likewise.
18889 (vst2q_p8): Likewise.
18890 (vst2q_s16): Likewise.
18891 (vst2q_p16): Likewise.
18892 (vst2q_s32): Likewise.
18893 (vst2q_s64): Likewise.
18894 (vst2q_u8): Likewise.
18895 (vst2q_u16): Likewise.
18896 (vst2q_u32): Likewise.
18897 (vst2q_u64): Likewise.
18898 (vst2q_f16): Likewise.
18899 (vst2q_f32): Likewise.
18900 (vst2q_f64): Likewise.
18901 (vst2q_p64): Likewise.
18903 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18905 * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy
18906 instead of constructing __builtin_aarch64_simd_ci one vector
18908 (vst3_u64): Likewise.
18909 (vst3_f64): Likewise.
18910 (vst3_s8): Likewise.
18911 (vst3_p8): Likewise.
18912 (vst3_s16): Likewise.
18913 (vst3_p16): Likewise.
18914 (vst3_s32): Likewise.
18915 (vst3_u8): Likewise.
18916 (vst3_u16): Likewise.
18917 (vst3_u32): Likewise.
18918 (vst3_f16): Likewise.
18919 (vst3_f32): Likewise.
18920 (vst3_p64): Likewise.
18921 (vst3q_s8): Likewise.
18922 (vst3q_p8): Likewise.
18923 (vst3q_s16): Likewise.
18924 (vst3q_p16): Likewise.
18925 (vst3q_s32): Likewise.
18926 (vst3q_s64): Likewise.
18927 (vst3q_u8): Likewise.
18928 (vst3q_u16): Likewise.
18929 (vst3q_u32): Likewise.
18930 (vst3q_u64): Likewise.
18931 (vst3q_f16): Likewise.
18932 (vst3q_f32): Likewise.
18933 (vst3q_f64): Likewise.
18934 (vst3q_p64): Likewise.
18936 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18938 * config/aarch64/arm_neon.h (vst4_s64): Use __builtin_memcpy
18939 instead of constructing __builtin_aarch64_simd_xi one vector
18941 (vst4_u64): Likewise.
18942 (vst4_f64): Likewise.
18943 (vst4_s8): Likewise.
18944 (vst4_p8): Likewise.
18945 (vst4_s16): Likewise.
18946 (vst4_p16): Likewise.
18947 (vst4_s32): Likewise.
18948 (vst4_u8): Likewise.
18949 (vst4_u16): Likewise.
18950 (vst4_u32): Likewise.
18951 (vst4_f16): Likewise.
18952 (vst4_f32): Likewise.
18953 (vst4_p64): Likewise.
18954 (vst4q_s8): Likewise.
18955 (vst4q_p8): Likewise.
18956 (vst4q_s16): Likewise.
18957 (vst4q_p16): Likewise.
18958 (vst4q_s32): Likewise.
18959 (vst4q_s64): Likewise.
18960 (vst4q_u8): Likewise.
18961 (vst4q_u16): Likewise.
18962 (vst4q_u32): Likewise.
18963 (vst4q_u64): Likewise.
18964 (vst4q_f16): Likewise.
18965 (vst4q_f32): Likewise.
18966 (vst4q_f64): Likewise.
18967 (vst4q_p64): Likewise.
18969 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18971 * config/aarch64/arm_neon.h (vtbx4_s8): Use __builtin_memcpy
18972 instead of constructing __builtin_aarch64_simd_oi one vector
18974 (vtbx4_u8): Likewise.
18975 (vtbx4_p8): Likewise.
18977 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18979 * config/aarch64/arm_neon.h (vtbl3_s8): Use __builtin_memcpy
18980 instead of constructing __builtin_aarch64_simd_oi one vector
18982 (vtbl3_u8): Likewise.
18983 (vtbl3_p8): Likewise.
18984 (vtbl4_s8): Likewise.
18985 (vtbl4_u8): Likewise.
18986 (vtbl4_p8): Likewise.
18988 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
18990 * config/aarch64/arm_neon.h (vqtbx2_s8): Use __builtin_memcpy
18991 instead of constructing __builtin_aarch64_simd_oi one vector
18993 (vqtbx2_u8): Likewise.
18994 (vqtbx2_p8): Likewise.
18995 (vqtbx2q_s8): Likewise.
18996 (vqtbx2q_u8): Likewise.
18997 (vqtbx2q_p8): Likewise.
18998 (vqtbx3_s8): Use __builtin_memcpy instead of constructing
18999 __builtin_aarch64_simd_ci one vector at a time.
19000 (vqtbx3_u8): Likewise.
19001 (vqtbx3_p8): Likewise.
19002 (vqtbx3q_s8): Likewise.
19003 (vqtbx3q_u8): Likewise.
19004 (vqtbx3q_p8): Likewise.
19005 (vqtbx4_s8): Use __builtin_memcpy instead of constructing
19006 __builtin_aarch64_simd_xi one vector at a time.
19007 (vqtbx4_u8): Likewise.
19008 (vqtbx4_p8): Likewise.
19009 (vqtbx4q_s8): Likewise.
19010 (vqtbx4q_u8): Likewise.
19011 (vqtbx4q_p8): Likewise.
19013 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
19015 * config/aarch64/arm_neon.h (vqtbl2_s8): Use __builtin_memcpy
19016 instead of constructing __builtin_aarch64_simd_oi one vector
19018 (vqtbl2_u8): Likewise.
19019 (vqtbl2_p8): Likewise.
19020 (vqtbl2q_s8): Likewise.
19021 (vqtbl2q_u8): Likewise.
19022 (vqtbl2q_p8): Likewise.
19023 (vqtbl3_s8): Use __builtin_memcpy instead of constructing
19024 __builtin_aarch64_simd_ci one vector at a time.
19025 (vqtbl3_u8): Likewise.
19026 (vqtbl3_p8): Likewise.
19027 (vqtbl3q_s8): Likewise.
19028 (vqtbl3q_u8): Likewise.
19029 (vqtbl3q_p8): Likewise.
19030 (vqtbl4_s8): Use __builtin_memcpy instead of constructing
19031 __builtin_aarch64_simd_xi one vector at a time.
19032 (vqtbl4_u8): Likewise.
19033 (vqtbl4_p8): Likewise.
19034 (vqtbl4q_s8): Likewise.
19035 (vqtbl4q_u8): Likewise.
19036 (vqtbl4q_p8): Likewise.
19038 2021-07-23 Haochen Gui <guihaoc@gcc.gnu.org>
19041 * config/rs6000/rs6000.md (cstore<mode>4): Fix wrong fall through.
19043 2021-07-22 Andrew Pinski <apinski@marvell.com>
19045 PR tree-optimization/10153
19046 * tree-tailcall.c (create_tailcall_accumulator):
19047 Don't call fold_convert as the type should be correct already.
19048 (tree_optimize_tail_calls_1): Use build_{one,zero}_cst instead
19049 of integer_{one,zero}_node for the call of create_tailcall_accumulator.
19051 2021-07-22 Aldy Hernandez <aldyh@redhat.com>
19053 * gimple-range-cache.cc (non_null_ref::adjust_range): Replace
19054 varying_p check for null/non-null check.
19056 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
19058 PR tree-optimization/101511
19059 * value-relation.cc (relation_oracle::query_relation): Check if ssa1
19060 is in ssa2's equiv set, and don't trap if so.
19062 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
19064 PR tree-optimization/101497
19065 * gimple-range-fold.cc (fold_using_range::range_of_cond_expr): Check
19068 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
19070 PR tree-optimization/101496
19071 * vr-values.c (simplify_using_ranges::fold_cond): Call range_of_stmt
19072 first, then vrp_visit_cond_Stmt.
19074 2021-07-22 liuhongt <hongtao.liu@intel.com>
19076 * config/i386/i386-expand.c
19077 (ix86_broadcast_from_integer_constant): Rename to ..
19078 (ix86_broadcast_from_constant): .. this, and extend it to
19080 (ix86_expand_vector_move): Extend to float mode.
19081 * config/i386/i386-features.c
19082 (replace_constant_pool_with_broadcast): Remove.
19083 (remove_partial_avx_dependency_gate): Ditto.
19084 (constant_pool_broadcast): Ditto.
19085 (class pass_constant_pool_broadcast): Ditto.
19086 (make_pass_constant_pool_broadcast): Ditto.
19087 (remove_partial_avx_dependency): Adjust gate.
19088 * config/i386/i386-passes.def: Remove pass_constant_pool_broadcast.
19089 * config/i386/i386-protos.h
19090 (make_pass_constant_pool_broadcast): Remove.
19092 2021-07-22 liuhongt <hongtao.liu@intel.com>
19094 * config/i386/constraints.md (Wb): New constraint.
19096 * config/i386/i386.md (*ashlhi3_1): Extend to avx512 mask
19098 (*ashlqi3_1): Ditto.
19099 (*<insn><mode>3_1): Split to ..
19100 (*ashr<mode>3_1): this, ...
19101 (*lshr<mode>3_1): and this, also extend this pattern to avx512
19103 (*<insn><mode>3_1): Split to ..
19104 (*ashr<mode>3_1): this, ...
19105 (*lshrqi3_1): and this, also extend this pattern to avx512
19107 (*lshrhi3_1): And this, also extend this pattern to avx512
19109 * config/i386/sse.md (k<code><mode>): New define_split after
19110 it to convert generic shift pattern to mask shift ones.
19112 2021-07-21 Thomas Schwinge <thomas@codesourcery.com>
19113 Joseph Myers <joseph@codesourcery.com>
19114 Cesar Philippidis <cesar@codesourcery.com>
19116 * tree-core.h (omp_clause_code): Add 'OMP_CLAUSE_NOHOST'.
19117 * tree.c (omp_clause_num_ops, omp_clause_code_name, walk_tree_1):
19119 * tree-pretty-print.c (dump_omp_clause): Likewise.
19120 * omp-general.c (oacc_verify_routine_clauses): Likewise.
19121 * gimplify.c (gimplify_scan_omp_clauses)
19122 (gimplify_adjust_omp_clauses): Likewise.
19123 * tree-nested.c (convert_nonlocal_omp_clauses)
19124 (convert_local_omp_clauses): Likewise.
19125 * omp-low.c (scan_sharing_clauses): Likewise.
19126 * omp-offload.c (execute_oacc_device_lower): Update.
19128 2021-07-21 Martin Sebor <msebor@redhat.com>
19130 * tree-ssa-alias.c (walk_aliased_vdefs_1): Fix typos in a comment.
19132 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19134 * config/rs6000/rs6000-gen-builtins.c (write_init_bif_table):
19137 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19139 * config/rs6000/rs6000-gen-builtins.c (write_fntype): New
19141 (write_fntype_init): New stub function.
19142 (write_init_bif_table): Likewise.
19143 (write_init_ovld_table): New function.
19144 (write_init_file): Implement.
19146 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19148 * config/rs6000/rs6000-gen-builtins.c
19149 (write_autogenerated_header): New function.
19150 (write_decls): Likewise.
19151 (write_extern_fntype): New callback function.
19152 (write_header_file): Implement.
19154 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19156 * config/rs6000/rs6000-gen-builtins.c (write_defines_file):
19159 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19161 * config/rs6000/rs6000-gen-builtins.c (complete_vector_type): New
19163 (complete_base_type): Likewise.
19164 (construct_fntype_id): Likewise.
19165 (parse_bif_entry): Call contruct_fntype_id.
19166 (parse_ovld_entry): Likewise.
19168 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19170 * config/rs6000/rs6000-gen-builtins.c (ovld_stanza): New struct.
19171 (MAXOVLDSTANZAS): New macro.
19172 (ovld_stanzas): New variable.
19173 (curr_ovld_stanza): Likewise.
19174 (MAXOVLDS): New macro.
19175 (ovlddata): New struct.
19176 (ovlds): New variable.
19177 (curr_ovld): Likewise.
19178 (max_ovld_args): Likewise.
19179 (parse_ovld_entry): New function.
19180 (parse_ovld_stanza): Likewise.
19181 (parse_ovld): Implement.
19183 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19185 * config/rs6000/rs6000-gen-builtins.c (parse_bif_attrs):
19188 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19190 * config/rs6000/rs6000-gen-builtins.c (parse_args): New function.
19191 (parse_prototype): Implement.
19193 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19195 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): New enum.
19196 (curr_bif_stanza): New variable.
19197 (stanza_entry): New struct.
19198 (stanza_map): New initialized variable.
19199 (enable_string): Likewise.
19200 (fnkinds): New enum.
19201 (typelist): New struct.
19202 (attrinfo): Likewise.
19203 (MAXRESTROPNDS): New macro.
19204 (prototype): New struct.
19205 (MAXBIFS): New macro.
19206 (bifdata): New struct.
19207 (bifs): New variable.
19208 (curr_bif): Likewise.
19209 (bif_order): Likewise.
19210 (bif_index): Likewise.
19211 (fatal): New function.
19212 (stanza_name_to_stanza): Likewise.
19213 (parse_bif_attrs): New stub function.
19214 (parse_prototype): Likewise.
19215 (parse_bif_entry): New function.
19216 (parse_bif_stanza): Likewise.
19217 (parse_bif): Implement.
19218 (set_bif_order): New function.
19219 (create_bif_order): Implement.
19221 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
19223 * config/rs6000/rs6000-gen-builtins.c (rbtree.h): New #include.
19224 (num_bifs): New variable.
19225 (num_ovld_stanzas): Likewise.
19226 (num_ovlds): Likewise.
19227 (parse_codes): New enum.
19228 (bif_rbt): New variable.
19229 (ovld_rbt): Likewise.
19230 (fntype_rbt): Likewise.
19231 (bifo_rbt): Likewise.
19232 (parse_bif): New stub function.
19233 (create_bif_order): Likewise.
19234 (parse_ovld): Likewise.
19235 (write_header_file): Likewise.
19236 (write_init_file): Likewise.
19237 (write_defines_file): Likewise.
19238 (delete_output_files): New function.
19241 2021-07-21 H.J. Lu <hjl.tools@gmail.com>
19244 * config/i386/i386-builtin.def: Remove OPTION_MASK_ISA_SSE4_2
19245 from CRC32 _builtin functions.
19247 2021-07-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
19249 * coverage.c (build_gcov_info_var_registration): Mark the object placed
19250 in the linker set as referenced so that it does not get optimized away.
19252 2021-07-21 Kito Cheng <kito.cheng@sifive.com>
19255 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
19257 * config.gcc (riscv*-*-*): Detect which python is available.
19259 2021-07-21 Jakub Jelinek <jakub@redhat.com>
19261 PR middle-end/101535
19262 * gimplify.c (omp_check_private): Properly skip ORT_TARGET_DATA
19263 contexts in which decl isn't privatized and for ORT_TARGET return
19264 false if decl is mapped.
19266 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
19268 * gimple-loop-jam.c: Include tree-ssa-sccvn.h.
19269 (tree_loop_unroll_and_jam): Run value-numbering on a loop that
19270 has been successfully unrolled.
19272 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
19274 * tree-ssa-loop-manip.c (determine_exit_conditions): Return a null
19275 exit condition if no tail loop is needed, and if the original exit
19276 condition should therefore be kept as-is.
19277 (tree_transform_and_unroll_loop): Handle that case here too.
19279 2021-07-21 Kewen Lin <linkw@linux.ibm.com>
19281 * tree-data-ref.c (free_dependence_relations): Adjust to pass vec
19283 (free_data_refs): Likewise.
19284 * tree-data-ref.h (free_dependence_relations): Likewise.
19285 (free_data_refs): Likewise.
19286 * tree-predcom.c (struct chain): Use auto_vec instead of vec for
19288 (struct component): Likewise.
19289 (pcom_worker::pcom_worker): Adjust for auto_vec and renaming changes.
19290 (pcom_worker::~pcom_worker): Likewise.
19291 (pcom_worker::release_chain): Adjust as auto_vec changes.
19292 (pcom_worker::loop): Rename to ...
19293 (pcom_worker::m_loop): ... this.
19294 (pcom_worker::datarefs): Rename to ...
19295 (pcom_worker::m_datarefs): ... this. Use auto_vec instead of vec.
19296 (pcom_worker::dependences): Rename to ...
19297 (pcom_worker::m_dependences): ... this. Use auto_vec instead of vec.
19298 (pcom_worker::chains): Rename to ...
19299 (pcom_worker::m_chains): ... this. Use auto_vec instead of vec.
19300 (pcom_worker::looparound_phis): Rename to ...
19301 (pcom_worker::m_looparound_phis): ... this. Use auto_vec instead of
19303 (pcom_worker::cache): Rename to ...
19304 (pcom_worker::m_cache): ... this. Use auto_vec instead of vec.
19305 (pcom_worker::release_chain): Adjust for auto_vec changes.
19306 (pcom_worker::release_chains): Adjust for auto_vec and renaming
19308 (release_component): Remove.
19309 (release_components): Adjust for release_component removal.
19310 (component_of): Adjust to use vec.
19311 (merge_comps): Likewise.
19312 (pcom_worker::aff_combination_dr_offset): Adjust for renaming changes.
19313 (pcom_worker::determine_offset): Likewise.
19314 (class comp_ptrs): Remove.
19315 (pcom_worker::split_data_refs_to_components): Adjust for renaming
19316 changes, for comp_ptrs removal with auto_vec.
19317 (pcom_worker::suitable_component_p): Adjust for renaming changes.
19318 (pcom_worker::filter_suitable_components): Adjust for release_component
19320 (pcom_worker::valid_initializer_p): Adjust for renaming changes.
19321 (pcom_worker::find_looparound_phi): Likewise.
19322 (pcom_worker::add_looparound_copies): Likewise.
19323 (pcom_worker::determine_roots_comp): Likewise.
19324 (pcom_worker::single_nonlooparound_use): Likewise.
19325 (pcom_worker::execute_pred_commoning_chain): Likewise.
19326 (pcom_worker::execute_pred_commoning): Likewise.
19327 (pcom_worker::try_combine_chains): Likewise.
19328 (pcom_worker::prepare_initializers_chain): Likewise.
19329 (pcom_worker::prepare_initializers): Likewise.
19330 (pcom_worker::prepare_finalizers_chain): Likewise.
19331 (pcom_worker::prepare_finalizers): Likewise.
19332 (pcom_worker::tree_predictive_commoning_loop): Likewise.
19334 2021-07-20 Martin Sebor <msebor@redhat.com>
19336 PR middle-end/101397
19337 * builtins.c (gimple_call_return_array): Add argument. Correct
19338 offsets for memchr, mempcpy, stpcpy, and stpncpy.
19339 (compute_objsize_r): Adjust offset computation for argument returning
19342 2021-07-20 Martin Sebor <msebor@redhat.com>
19344 PR middle-end/101300
19345 * tree-ssa-uninit.c (check_defs): Handle UBSAN built-ins.
19347 2021-07-20 Jeff Law <jlaw@localhost.localdomain>
19349 * function.c (assign_parm_setup_block): Use adjust_address instead
19350 of change_address to preserve MEM_EXPR and friends.
19352 2021-07-20 Martin Sebor <msebor@redhat.com>
19354 * cfgloop.h (single_likely_exit): Adjust by-value argument to
19355 by-const-reference.
19356 * cfgloopanal.c (single_likely_exit): Same.
19357 * cgraph.h (struct cgraph_node): Same.
19358 * cgraphclones.c (cgraph_node::create_virtual_clone): Same.
19359 * genautomata.c (merge_states): Same.
19360 * genextract.c (VEC_char_to_string): Same.
19361 * genmatch.c (dt_node::gen_kids_1): Same.
19362 (walk_captures): Adjust by-value argument to by-reference.
19363 * gimple-ssa-store-merging.c (check_no_overlap): Adjust by-value argument
19364 to by-const-reference.
19365 * gimple.c (gimple_build_call_vec): Same.
19366 (gimple_build_call_internal_vec): Same.
19367 (gimple_build_switch): Same.
19368 (sort_case_labels): Same.
19369 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
19371 * gimple.h (gimple_build_call_vec): Adjust by-value argument to
19372 by-const-reference.
19373 (gimple_build_call_internal_vec): Same.
19374 (gimple_build_switch): Same.
19375 (sort_case_labels): Same.
19376 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
19378 * haifa-sched.c (calc_priorities): Adjust by-value argument to
19379 by-const-reference.
19380 (sched_init_luids): Same.
19381 (haifa_init_h_i_d): Same.
19382 * ipa-cp.c (ipa_get_indirect_edge_target_1): Same.
19383 (adjust_callers_for_value_intersection): Adjust by-value argument to
19385 (find_more_scalar_values_for_callers_subset): Adjust by-value argument to
19386 by-const-reference.
19387 (find_more_contexts_for_caller_subset): Same.
19388 (find_aggregate_values_for_callers_subset): Same.
19389 (copy_useful_known_contexts): Same.
19390 * ipa-fnsummary.c (remap_edge_summaries): Same.
19391 (remap_freqcounting_predicate): Same.
19392 * ipa-inline.c (add_new_edges_to_heap): Adjust by-value argument to
19394 * ipa-predicate.c (predicate::remap_after_inlining): Adjust by-value argument
19395 to by-const-reference.
19396 * ipa-predicate.h (predicate::remap_after_inlining): Same.
19397 * ipa-prop.c (ipa_find_agg_cst_for_param): Same.
19398 * ipa-prop.h (ipa_find_agg_cst_for_param): Same.
19399 * ira-build.c (ira_loop_tree_body_rev_postorder): Same.
19400 * read-rtl.c (add_overload_instance): Same.
19401 * rtl.h (native_decode_rtx): Same.
19402 (native_decode_vector_rtx): Same.
19403 * sched-int.h (sched_init_luids): Same.
19404 (haifa_init_h_i_d): Same.
19405 * simplify-rtx.c (native_decode_vector_rtx): Same.
19406 (native_decode_rtx): Same.
19407 * tree-call-cdce.c (gen_shrink_wrap_conditions): Same.
19408 (shrink_wrap_one_built_in_call_with_conds): Same.
19409 (shrink_wrap_conditional_dead_built_in_calls): Same.
19410 * tree-data-ref.c (create_runtime_alias_checks): Same.
19411 (compute_all_dependences): Same.
19412 * tree-data-ref.h (compute_all_dependences): Same.
19413 (create_runtime_alias_checks): Same.
19414 (index_in_loop_nest): Same.
19415 * tree-if-conv.c (mask_exists): Same.
19416 * tree-loop-distribution.c (class loop_distribution): Same.
19417 (loop_distribution::create_rdg_vertices): Same.
19418 (dump_rdg_partitions): Same.
19419 (debug_rdg_partitions): Same.
19420 (partition_contains_all_rw): Same.
19421 (loop_distribution::distribute_loop): Same.
19422 * tree-parloops.c (oacc_entry_exit_ok_1): Same.
19423 (oacc_entry_exit_single_gang): Same.
19424 * tree-ssa-loop-im.c (hoist_memory_references): Same.
19425 (loop_suitable_for_sm): Same.
19426 * tree-ssa-loop-niter.c (bound_index): Same.
19427 * tree-ssa-reassoc.c (update_ops): Same.
19428 (swap_ops_for_binary_stmt): Same.
19429 (rewrite_expr_tree): Same.
19430 (rewrite_expr_tree_parallel): Same.
19431 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Same.
19432 * tree-ssa-sccvn.h (ao_ref_init_from_vn_reference): Same.
19433 * tree-ssa-structalias.c (process_all_all_constraints): Same.
19434 (make_constraints_to): Same.
19435 (handle_lhs_call): Same.
19436 (find_func_aliases_for_builtin_call): Same.
19437 (sort_fieldstack): Same.
19438 (check_for_overlaps): Same.
19439 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Same.
19440 (vect_create_cond_for_unequal_addrs): Same.
19441 (vect_create_cond_for_lower_bounds): Same.
19442 (vect_create_cond_for_alias_checks): Same.
19443 * tree-vect-slp-patterns.c (vect_validate_multiplication): Same.
19444 * tree-vect-slp.c (vect_analyze_slp_instance): Same.
19445 (vect_make_slp_decision): Same.
19446 (vect_slp_bbs): Same.
19447 (duplicate_and_interleave): Same.
19448 (vect_transform_slp_perm_load): Same.
19449 (vect_schedule_slp): Same.
19450 * tree-vectorizer.h (vect_transform_slp_perm_load): Same.
19451 (vect_schedule_slp): Same.
19452 (duplicate_and_interleave): Same.
19453 * tree.c (build_vector_from_ctor): Same.
19454 (build_vector): Same.
19455 (check_vector_cst): Same.
19456 (check_vector_cst_duplicate): Same.
19457 (check_vector_cst_fill): Same.
19458 (check_vector_cst_stepped): Same.
19459 * tree.h (build_vector_from_ctor): Same.
19461 2021-07-20 Jakub Jelinek <jakub@redhat.com>
19464 * config/rs6000/rs6000-protos.h (easy_altivec_constant): Change return
19465 type from bool to int.
19466 * config/rs6000/rs6000.c (vspltis_constant): Fix up handling the
19467 EASY_VECTOR_MSB case if either step or copies is not 1.
19468 (vspltis_shifted): Fix comment typo.
19469 (easy_altivec_constant): Change return type from bool to int, instead
19470 of returning true return byte size of the element mode that should be
19471 used to synthetize the constant.
19472 * config/rs6000/predicates.md (easy_vector_constant_msb): Require
19473 that vspltis_shifted is 0, handle the case where easy_altivec_constant
19474 assumes using different vector mode from CONST_VECTOR's mode.
19475 * config/rs6000/altivec.md (easy_vector_constant_msb splitter): Use
19476 easy_altivec_constant to determine mode in which -1 >> -1 should be
19477 performed, use rs6000_expand_vector_init instead of gen_vec_initv4sisi.
19479 2021-07-20 Richard Biener <rguenther@suse.de>
19482 * dwarf2out.h (dwarf_file_data): Add key member.
19483 * dwarf2out.c (dwarf_file_hasher::equal): Compare key.
19484 (dwarf_file_hasher::hash): Hash key.
19485 (lookup_filename): Remap the filename and store it in the
19486 filename member of dwarf_file_data when creating a new
19488 (file_name_acquire): Do not remap the filename again.
19489 (maybe_emit_file): Likewise.
19491 2021-07-20 Jonathan Wright <jonathan.wright@arm.com>
19493 * config/aarch64/aarch64-simd-builtins.def: Use two variant
19494 generators for all TBL/TBX intrinsics and rename to
19495 consistent forms: qtbl[1234] or qtbx[1234].
19496 * config/aarch64/aarch64-simd.md (aarch64_tbl1<mode>):
19498 (aarch64_qtbl1<mode>): This.
19499 (aarch64_tbx1<mode>): Rename to...
19500 (aarch64_qtbx1<mode>): This.
19501 (aarch64_tbl2v16qi): Delete.
19502 (aarch64_tbl3<mode>): Rename to...
19503 (aarch64_qtbl2<mode>): This.
19504 (aarch64_tbx4<mode>): Rename to...
19505 (aarch64_qtbx2<mode>): This.
19506 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_1): Use
19507 renamed qtbl1 and qtbl2 RTL patterns.
19508 * config/aarch64/arm_neon.h (vqtbl1_p8): Use renamed qtbl1
19510 (vqtbl1_s8): Likewise.
19511 (vqtbl1_u8): Likewise.
19512 (vqtbl1q_p8): Likewise.
19513 (vqtbl1q_s8): Likewise.
19514 (vqtbl1q_u8): Likewise.
19515 (vqtbx1_s8): Use renamed qtbx1 RTL pattern.
19516 (vqtbx1_u8): Likewise.
19517 (vqtbx1_p8): Likewise.
19518 (vqtbx1q_s8): Likewise.
19519 (vqtbx1q_u8): Likewise.
19520 (vqtbx1q_p8): Likewise.
19521 (vtbl1_s8): Use renamed qtbl1 RTL pattern.
19522 (vtbl1_u8): Likewise.
19523 (vtbl1_p8): Likewise.
19524 (vtbl2_s8): Likewise
19525 (vtbl2_u8): Likewise.
19526 (vtbl2_p8): Likewise.
19527 (vtbl3_s8): Use renamed qtbl2 RTL pattern.
19528 (vtbl3_u8): Likewise.
19529 (vtbl3_p8): Likewise.
19530 (vtbl4_s8): Likewise.
19531 (vtbl4_u8): Likewise.
19532 (vtbl4_p8): Likewise.
19533 (vtbx2_s8): Use renamed qtbx2 RTL pattern.
19534 (vtbx2_u8): Likewise.
19535 (vtbx2_p8): Likewise.
19536 (vqtbl2_s8): Use renamed qtbl2 RTL pattern.
19537 (vqtbl2_u8): Likewise.
19538 (vqtbl2_p8): Likewise.
19539 (vqtbl2q_s8): Likewise.
19540 (vqtbl2q_u8): Likewise.
19541 (vqtbl2q_p8): Likewise.
19542 (vqtbx2_s8): Use renamed qtbx2 RTL pattern.
19543 (vqtbx2_u8): Likewise.
19544 (vqtbx2_p8): Likewise.
19545 (vqtbx2q_s8): Likewise.
19546 (vqtbx2q_u8): Likewise.
19547 (vqtbx2q_p8): Likewise.
19548 (vtbx4_s8): Likewise.
19549 (vtbx4_u8): Likewise.
19550 (vtbx4_p8): Likewise.
19552 2021-07-20 Uroš Bizjak <ubizjak@gmail.com>
19555 * config/i386/sync.md (define_peephole2 atomic_storedi_fpu):
19557 (define_peephole2 atomic_loaddi_fpu): Ditto.
19559 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
19561 * config.gcc (riscv*-*-*): Detect which python is available.
19563 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
19565 * config/rs6000/vsx.md (mulhs_<mode>): Rename to...
19566 (smul<mode>3_highpart): ... this.
19567 (mulhu_<mode>): Rename to...
19568 (umul<mode>3_highpart): ... this.
19569 * config/rs6000/rs6000-builtin.def (MULHS_V2DI, MULHS_V4SI,
19570 MULHU_V2DI, MULHU_V4SI): Adjust.
19572 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
19574 PR tree-optimization/100696
19575 * internal-fn.c (first_commutative_argument): Add info for IFN_MULH.
19576 * internal-fn.def (IFN_MULH): New internal function.
19577 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Add support to
19578 recog normal multiply highpart as IFN_MULH.
19579 * config/i386/i386.c (ix86_add_stmt_cost): Adjust for combined
19582 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
19584 * config/elfos.h (CTF_DEBUGGING_INFO): New definition.
19585 (BTF_DEBUGGING_INFO): Likewise.
19586 * doc/tm.texi.in: Document the new macros.
19587 * doc/tm.texi: Regenerated.
19588 * toplev.c: Guard initialization of debug hooks.
19590 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
19592 * flags.h (ctf_debuginfo_p): New function declaration.
19593 * opts.c (ctf_debuginfo_p): New function definition.
19595 2021-07-19 Andrew Stubbs <ams@codesourcery.com>
19598 * config/gcn/gcn-hsa.h (DRIVER_SELF_SPECS): New.
19599 (ASM_SPEC): Set -mattr for xnack and sram-ecc.
19600 * config/gcn/gcn-opts.h (enum sram_ecc_type): New.
19601 * config/gcn/gcn-valu.md: Add a warning comment.
19602 * config/gcn/gcn.c (gcn_option_override): Add "sorry" for -mxnack.
19603 (output_file_start): Add xnack and sram-ecc state to ".amdgcn_target".
19604 * config/gcn/gcn.md: Add a warning comment.
19605 * config/gcn/gcn.opt: Add -mxnack and -msram-ecc.
19606 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): Remove
19608 (EF_AMDGPU_XNACK): New.
19609 (EF_AMDGPU_SRAM_ECC): New.
19611 (copy_early_debug_info): Use elf_flags.
19612 (main): Handle -mxnack and -msram-ecc options.
19613 * doc/invoke.texi: Document -mxnack and -msram-ecc.
19615 2021-07-19 Andrew Pinski <apinski@marvell.com>
19618 * config/aarch64/aarch64.md (csneg3_uxtw_insn): Rename to ...
19619 (*cs<neg_not_cs>3_uxtw_insn4): and extend to NEG_NOT.
19621 2021-07-19 Richard Biener <rguenther@suse.de>
19623 PR tree-optimization/101505
19624 * tree-vect-patterns.c (vect_determine_precisions): Walk
19625 PHIs also for loop vectorization.
19627 2021-07-19 Richard Biener <rguenther@suse.de>
19629 * gimple.h (gimple_expr_type): Remove.
19630 * doc/gimple.texi: Remove gimple_expr_type documentation.
19632 2021-07-19 Richard Biener <rguenther@suse.de>
19634 * tree-ssa-sccvn.c (vn_reference_eq): Handle NULL vr->type.
19635 (ao_ref_init_from_vn_reference): Likewise.
19636 (fully_constant_reference): Likewise.
19637 (vn_reference_lookup_call): Do not set vr->type to random
19639 * tree-ssa-pre.c (compute_avail): Do not try to PRE calls
19641 * tree-vect-generic.c (expand_vector_piecewise): Pass in
19642 whether we expanded parallel.
19643 (expand_vector_parallel): Adjust.
19644 (expand_vector_addition): Likewise.
19645 (expand_vector_comparison): Likewise.
19646 (expand_vector_operation): Likewise.
19647 (expand_vector_scalar_condition): Likewise.
19648 (expand_vector_conversion): Likewise.
19650 2021-07-19 Richard Biener <rguenther@suse.de>
19652 * tree-vrp.c (register_edge_assert_for_2): Use the
19654 (vrp_folder::fold_predicate_in): Likewise.
19655 * vr-values.c (gimple_assign_nonzero_p): Likewise.
19656 (vr_values::extract_range_from_comparison): Likewise.
19657 (vr_values::extract_range_from_ubsan_builtin): Use the
19658 type of the first operand.
19659 (vr_values::extract_range_basic): Push down type
19660 computation, use the appropriate LHS.
19661 (vr_values::extract_range_from_assignment): Use the
19664 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
19667 * common/config/i386/i386-common.c (ix86_handle_option): For
19668 -mgeneral-regs-only, enable the GPR only instructions which are
19669 enabled implicitly by SSE ISAs unless they have been disabled
19672 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
19675 * config/i386/i386.c (ix86_check_avx_upper_stores): Moved before
19676 ix86_avx_u128_mode_needed.
19677 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY if callee
19678 returns AVX register.
19680 2021-07-17 Jan Hubicka <hubicka@ucw.cz>
19682 * tree-ssa-structalias.c (handle_rhs_call): Support EAF_NOT_RETURNED.
19683 (handle_const_call): Liekise
19684 (handle_pure_call): Liekise
19686 2021-07-17 Andrew MacLeod <amacleod@redhat.com>
19688 PR tree-optimization/96542
19689 * range-op.cc (range_operator::wi_fold_in_parts): New.
19690 (range_operator::fold_range): Call wi_fold_in_parts.
19691 (operator_lshift::wi_fold): Fix broken lshift by [0,0].
19692 * range-op.h (wi_fold_in_parts): Add prototype.
19694 2021-07-16 David Malcolm <dmalcolm@redhat.com>
19696 * doc/analyzer.texi: Add __analyzer_dump_state.
19698 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19700 * config/rs6000/rbtree.c: New file.
19701 * config/rs6000/rbtree.h: New file.
19703 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19705 * config/rs6000/rs6000-gen-builtins.c (restriction): New enum.
19706 (typeinfo): Add restr field.
19707 (match_bracketed_pair): New function.
19708 (match_const_restriction): Implement.
19710 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19712 * config/rs6000/rs6000-gen-builtins.c (match_basetype): Implement.
19714 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19716 * config/rs6000/rs6000-gen-builtins.c (void_status): New enum.
19717 (basetype): Likewise.
19718 (typeinfo): Likewise.
19719 (handle_pointer): New function.
19720 (match_basetype): New stub function.
19721 (match_const_restriction): Likewise.
19722 (match_type): New function.
19724 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19726 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace): New
19728 (advance_line): Likewise.
19729 (safe_inc_pos): Likewise.
19730 (match_identifier): Likewise.
19731 (match_integer): Likewise.
19732 (match_to_right_bracket): Likewise.
19734 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19736 * config/rs6000/rs6000-gen-builtins.c (bif_file): New variable.
19737 (ovld_file): Likewise.
19738 (header_file): Likewise.
19739 (init_file): Likewise.
19740 (defines_file): Likewise.
19741 (pgm_path): Likewise.
19742 (bif_path): Likewise.
19743 (ovld_path): Likewise.
19744 (header_path): Likewise.
19745 (init_path): Likewise.
19746 (defines_path): Likewise.
19747 (LINELEN): New macro.
19748 (linebuf): New variable.
19752 (bif_diag): New function.
19753 (ovld_diag): Likewise.
19755 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19757 * config/rs6000/rs6000-builtin-new.def: New.
19758 * config/rs6000/rs6000-overload.def: New.
19760 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19762 * config/rs6000/rs6000-gen-builtins.c: New.
19764 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
19766 * Makefile.in (EXTRA_GTYPE_DEPS): New variable.
19767 (s-gtype): Depend on EXTRA_GTYPE_DEPS.
19768 * gengtype-state.c (state_writer::write_state_file_list): Add a
19769 parameter to the fileslist expression for the number of build
19771 (read_state_files_list): Detect build headers and strip the
19772 initial "./" or ".\" from their names.
19773 * gengtype.c (build_headers): New global variable.
19774 (num_build_headers): Likewise.
19775 (open_base_files): Emit #include for each build header.
19776 (main): Detect and count build headers.
19777 * gengtype.h (build_headers): New extern variable.
19778 (num_build_headers): Likewise.
19780 2021-07-16 Richard Biener <rguenther@suse.de>
19782 * gimple-ssa-store-merging.c (verify_symbolic_number_p): Use
19783 the type of the LHS.
19784 (find_bswap_or_nop_1): Likewise.
19785 (find_bswap_or_nop): Likewise.
19786 * tree-vectorizer.h (vect_get_smallest_scalar_type): Adjust
19788 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
19789 Remove unused parameters, pass in the scalar type. Fix
19790 internal store function handling.
19791 * tree-vect-stmts.c (vect_analyze_stmt): Remove assert.
19792 (vect_get_vector_types_for_stmt): Move down check for
19793 existing vector stmt after we've determined a scalar type.
19794 Pass down the used scalar type to vect_get_smallest_scalar_type.
19795 * tree-vect-generic.c (expand_vector_condition): Use
19796 the type of the LHS.
19797 (expand_vector_scalar_condition): Likewise.
19798 (expand_vector_operations_1): Likewise.
19799 * tree-vect-patterns.c (vect_widened_op_tree): Likewise.
19800 (vect_recog_dot_prod_pattern): Likewise.
19801 (vect_recog_sad_pattern): Likewise.
19802 (vect_recog_widen_op_pattern): Likewise.
19803 (vect_recog_widen_sum_pattern): Likewise.
19804 (vect_recog_mixed_size_cond_pattern): Likewise.
19806 2021-07-16 Jan Hubicka <hubicka@ucw.cz>
19808 * ipa-modref.c (struct escape_entry): Use eaf_fleags_t.
19809 (dump_eaf_flags): Dump EAF_NOT_RETURNED
19810 (eaf_flags_useful_p): Use eaf_fleags_t; handle const functions
19811 and EAF_NOT_RETURNED.
19812 (modref_summary::useful_p): Likewise.
19813 (modref_summary_lto::useful_p): Likewise.
19814 (struct) modref_summary_lto: Use eaf_fleags_t.
19815 (deref_flags): Handle EAF_NOT_RETURNED.
19816 (struct escape_point): Use min_flags.
19817 (modref_lattice::init): Add EAF_NOT_RETURNED.
19818 (merge_call_lhs_flags): Ignore EAF_NOT_RETURNED functions
19819 (analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
19821 (analyze_parms): Also analyze const functions; update conition on
19823 (modref_write): Update streaming.
19824 (read_section): Update streaming.
19825 (remap_arg_flags): Use eaf_flags_t.
19826 (modref_merge_call_site_flags): Hanlde EAF_NOT_RETURNED.
19827 * ipa-modref.h: (eaf_flags_t): New typedef.
19828 (struct modref_summary): Use eaf_flags_t.
19829 * tree-core.h (EAF_NOT_RETURNED): New constant.
19831 2021-07-16 Richard Biener <rguenther@suse.de>
19833 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Use
19834 the type of the LHS.
19835 (gimple_assign_nonnegative_warnv_p): Likewise.
19836 (gimple_call_nonnegative_warnv_p): Likewise. Return false
19837 if the call has no LHS.
19838 * gimple.c (gimple_could_trap_p_1): Use the type of the LHS.
19839 * tree-eh.c (stmt_could_throw_1_p): Likewise.
19840 * tree-inline.c (insert_init_stmt): Likewise.
19841 * tree-ssa-loop-niter.c (get_val_for): Likewise.
19842 * tree-outof-ssa.c (ssa_is_replaceable_p): Use the type of
19844 * tree-ssa-sccvn.c (init_vn_nary_op_from_stmt): Take a
19845 gassign *. Use the type of the lhs.
19846 (vn_nary_op_lookup_stmt): Adjust.
19847 (vn_nary_op_insert_stmt): Likewise.
19849 2021-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
19851 * config/s390/predicates.md (bras_sym_operand): Accept all
19852 functions in 64-bit mode, use UNSPEC_PLT31.
19853 (larl_operand): Use UNSPEC_PLT31.
19854 * config/s390/s390.c (s390_loadrelative_operand_p): Likewise.
19855 (legitimize_pic_address): Likewise.
19856 (s390_emit_tls_call_insn): Mark __tls_get_offset as function,
19858 (s390_delegitimize_address): Use UNSPEC_PLT31.
19859 (s390_output_addr_const_extra): Likewise.
19860 (print_operand): Add @PLT to TLS calls, handle %K.
19861 (s390_function_profiler): Mark __fentry__/_mcount as function,
19862 use %K, use UNSPEC_PLT31.
19863 (s390_output_mi_thunk): Use only UNSPEC_GOT, use %K.
19864 (s390_emit_call): Use UNSPEC_PLT31.
19865 (s390_emit_tpf_eh_return): Mark __tpf_eh_return as function.
19866 * config/s390/s390.md (UNSPEC_PLT31): Rename from UNSPEC_PLT.
19867 (*movdi_64): Use %K.
19868 (reload_base_64): Likewise.
19869 (*sibcall_brc): Likewise.
19870 (*sibcall_brcl): Likewise.
19871 (*sibcall_value_brc): Likewise.
19872 (*sibcall_value_brcl): Likewise.
19874 (*brasl): Likewise.
19875 (*bras_r): Likewise.
19876 (*brasl_r): Likewise.
19877 (*bras_tls): Likewise.
19878 (*brasl_tls): Likewise.
19879 (main_base_64): Likewise.
19880 (reload_base_64): Likewise.
19881 (@split_stack_call<mode>): Likewise.
19883 2021-07-16 Richard Biener <rguenther@suse.de>
19885 PR tree-optimization/101467
19886 * tree-vect-stmts.c (vect_gen_while): Properly guard
19887 make_temp_ssa_name usage.
19889 2021-07-16 Cooper Qu <cooper.qu@linux.alibaba.com>
19891 * config.gcc: Don't use forked print-sysroot-suffix.sh and
19892 t-sysroot-suffix for C-SKY.
19893 * config/csky/print-sysroot-suffix.sh: Delete.
19894 * config/csky/t-csky-linux: Delete.
19895 * config/csky/t-sysroot-suffix: Define MULTILIB_DIRNAMES
19896 instead of CSKY_MULTILIB_DIRNAMES.
19898 2021-07-16 Richard Biener <rguenther@suse.de>
19900 * tree-vect-loop.c (vect_transform_cycle_phi): Correct sign
19901 conversion issues with the partial reduction of the reused
19902 vector accumulator.
19904 2021-07-16 Richard Biener <rguenther@suse.de>
19906 * config/i386/i386-options.c (ix86_option_override_internal): Set
19907 param_vect_partial_vector_usage to zero if not set.
19909 2021-07-15 Uroš Bizjak <ubizjak@gmail.com>
19912 * config/i386/i386.h (VALID_SSE_REG_MODE): Add TDmode.
19913 (VALID_INT_MODE_P): Add SDmode and DDmode.
19914 Add TDmode for TARGET_64BIT.
19915 (VALID_DFP_MODE_P): Remove.
19916 * config/i386/i386.c (ix86_hard_regno_mode_ok):
19917 Do not use VALID_DFP_MODE_P.
19919 2021-07-15 Andrew MacLeod <amacleod@redhat.com>
19921 * gimple-range-fold.cc (adjust_pointer_diff_expr): Use
19923 (fold_using_range::fold_stmt): Ditto.
19924 (fold_using_range::range_of_range_op): Ditto.
19925 (fold_using_range::range_of_phi): Ditto.
19926 (fold_using_range::range_of_call): Ditto.
19927 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
19928 (fold_using_range::range_of_builtin_call): Ditto.
19929 (fold_using_range::range_of_cond_expr): Ditto.
19930 * gimple-range-fold.h (gimple_range_type): New.
19932 2021-07-15 Martin Sebor <msebor@redhat.com>
19934 PR middle-end/97027
19935 * tree-ssa-strlen.c (handle_assign): New function.
19936 (maybe_warn_overflow): Add argument.
19937 (nonzero_bytes_for_type): New function.
19938 (count_nonzero_bytes): Handle more tree types. Call
19939 nonzero_bytes_for_tye.
19940 (count_nonzero_bytes): Handle types.
19941 (handle_store): Handle stores from function calls.
19942 (strlen_check_and_optimize_call): Move code to handle_assign. Call
19943 it for assignments from function calls.
19945 2021-07-15 David Malcolm <dmalcolm@redhat.com>
19950 * doc/invoke.texi: Add -Wanalyzer-use-of-uninitialized-value.
19952 2021-07-15 David Malcolm <dmalcolm@redhat.com>
19954 * doc/invoke.texi (-fdump-analyzer-exploded-paths): New.
19956 2021-07-15 Martin Sebor <msebor@redhat.com>
19960 * fold-const.c (operand_compare::operand_equal_p): Handle OEP_DECL_NAME.
19961 (operand_compare::verify_hash_value): Same.
19962 * tree-core.h (OEP_DECL_NAME): New.
19964 2021-07-15 Martin Jambor <mjambor@suse.cz>
19966 * profile-count.h (profile_count::value): Change the return type to
19968 * gimple-pretty-print.c (dump_gimple_bb_header): Adjust print
19970 * tree-cfg.c (dump_function_to_file): Likewise.
19972 2021-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
19975 * config/rs6000/rs6000-p8swap.c (has_part_mult): New.
19976 (rs6000_analyze_swaps): Insns containing a subreg of a mult are
19979 2021-07-15 Richard Biener <rguenther@suse.de>
19981 * tree-vectorizer.h (vect_gen_while): Match up with
19982 vect_gen_while_not.
19983 * tree-vect-stmts.c (vect_gen_while): Adjust API to that
19984 of vect_gen_while_not.
19985 (vect_gen_while_not): Adjust.
19986 * tree-vect-loop-manip.c (vect_set_loop_controls_directly): Likewise.
19988 2021-07-15 Aldy Hernandez <aldyh@redhat.com>
19990 * gimple-range-cache.cc (non_null_ref::adjust_range): New.
19991 (ranger_cache::range_of_def): Call adjust_range.
19992 (ranger_cache::entry_range): Same.
19993 * gimple-range-cache.h (non_null_ref::adjust_range): New.
19994 * gimple-range.cc (gimple_ranger::range_of_expr): Call
19996 (gimple_ranger::range_on_entry): Same.
19998 2021-07-15 Tamar Christina <tamar.christina@arm.com>
20001 2021-07-14 Tamar Christina <tamar.christina@arm.com>
20003 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
20005 2021-07-15 Tamar Christina <tamar.christina@arm.com>
20008 2021-07-14 Tamar Christina <tamar.christina@arm.com>
20010 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
20011 (sdot_prod, udot_prod): ...These.
20012 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
20013 (aarch64_<sur>dot<vsi2qi>): Rename to...
20014 (<sur>dot_prod<vsi2qi>): ...This.
20015 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
20018 2021-07-15 Jakub Jelinek <jakub@redhat.com>
20020 PR middle-end/101437
20021 * gimplify.c (gimplify_expr): Throw away volatile reads from empty
20022 types even if they have non-BLKmode TYPE_MODE.
20024 2021-07-15 Richard Biener <rguenther@suse.de>
20027 * gcc.c (process_command): Process -gtoggle like process_options
20028 would after parsing options.
20030 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
20032 * cfgexpand.c (expand_asm_loc): Adjust.
20033 (expand_asm_stmt): Likewise.
20034 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Likewise.
20035 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
20036 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
20037 * config/avr/avr.c (avr_md_asm_adjust): Likewise.
20038 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
20039 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
20040 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
20041 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
20042 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
20043 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
20044 * config/s390/s390.c (s390_md_asm_adjust): Likewise.
20045 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
20046 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
20047 * doc/tm.texi: Regenerate.
20048 * target.def: Add location argument to md_asm_adjust.
20050 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
20052 * tree-diagnostic.c (diagnostic_report_current_function): Use the
20053 diagnostic's location, not input_location.
20055 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
20057 * cfgexpand.c (tree_conflicts_with_clobbers_p): Pass location to
20059 (expand_asm_stmt): Likewise.
20061 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
20063 * config/rs6000/rs6000.c (adjacent_mem_locations): Return the lower
20064 addressed memory rtx, if any.
20065 (rs6000_split_multireg_move): Fix code formatting.
20066 Handle MMA build built-ins with operands in adjacent memory locations.
20068 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
20070 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Move to later
20073 2021-07-14 Jason Merrill <jason@redhat.com>
20075 * sel-sched-ir.h (get_all_loop_exits): Use auto_vec.
20077 2021-07-14 Jason Merrill <jason@redhat.com>
20079 * doc/invoke.texi: -fdelete-dead-exceptions is on by default for
20082 2021-07-14 Tamar Christina <tamar.christina@arm.com>
20084 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
20085 Remove erroneous line.
20087 2021-07-14 Andrew MacLeod <amacleod@redhat.com>
20089 * params.opt (param_evrp_mode): Change default.
20091 2021-07-14 Tamar Christina <tamar.christina@arm.com>
20093 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
20094 (sdot_prod, udot_prod): ...These.
20095 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
20096 (aarch64_<sur>dot<vsi2qi>): Rename to...
20097 (<sur>dot_prod<vsi2qi>): ...This.
20098 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
20101 2021-07-14 Tamar Christina <tamar.christina@arm.com>
20103 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
20105 2021-07-14 Tamar Christina <tamar.christina@arm.com>
20107 * doc/sourcebuild.texi (arm_v8_2a_i8mm_neon_hw): Document.
20109 2021-07-14 Tamar Christina <tamar.christina@arm.com>
20111 * config/arm/neon.md (usdot_prod<vsi2qi>): New.
20113 2021-07-14 Tamar Christina <tamar.christina@arm.com>
20115 * config/aarch64/aarch64-simd.md (aarch64_usdot<vsi2qi>): Rename to...
20116 (usdot_prod<vsi2qi>): ... This.
20117 * config/aarch64/aarch64-simd-builtins.def (usdot): Rename to...
20118 (usdot_prod): ...This.
20119 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Likewise.
20120 * config/aarch64/aarch64-sve.md (@aarch64_<sur>dot_prod<vsi2qi>):
20122 (@<sur>dot_prod<vsi2qi>): ...This.
20123 * config/aarch64/aarch64-sve-builtins-base.cc
20124 (svusdot_impl::expand): Use it.
20126 2021-07-14 Tamar Christina <tamar.christina@arm.com>
20128 * optabs.def (usdot_prod_optab): New.
20129 * doc/md.texi: Document it and clarify other dot prod optabs.
20130 * optabs-tree.h (enum optab_subtype): Add optab_vector_mixed_sign.
20131 * optabs-tree.c (optab_for_tree_code): Support usdot_prod_optab.
20132 * optabs.c (expand_widen_pattern_expr): Likewise.
20133 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
20134 * tree-vect-loop.c (vectorizable_reduction): Query dot-product kind.
20135 * tree-vect-patterns.c (vect_supportable_direct_optab_p): Take optional
20137 (vect_widened_op_tree): Optionally ignore
20139 (vect_recog_dot_prod_pattern): Support usdot_prod_optab.
20141 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
20144 * config/i386/driver-i386.c (host_detect_local_cpu): Check
20145 "arch [32|64]" and "tune [32|64]" for 32-bit and 64-bit codegen.
20146 Enable UINTR only for 64-bit codegen.
20147 * config/i386/i386-options.c
20148 (ix86_option_override_internal::DEF_PTA): Skip PTA_UINTR if not
20150 * config/i386/i386.h (ARCH_ARG): New.
20151 (CC1_CPU_SPEC): Pass "[arch|tune] 32" for 32-bit codegen and
20152 "[arch|tune] 64" for 64-bit codegen.
20154 2021-07-14 Richard Biener <rguenther@suse.de>
20156 PR tree-optimization/101445
20157 * tree-vect-stmts.c (vectorizable_load): Do the gap adjustment
20158 of the IV in the correct direction for negative stride
20161 2021-07-14 Jakub Jelinek <jakub@redhat.com>
20164 * godump.c (godump_str_hash): New type.
20165 (godump_container::pot_dummy_types): Use string_hash instead of
20166 ptr_hash in the hash_set.
20168 2021-07-14 Richard Biener <rguenther@suse.de>
20170 * tree-vect-loop.c (vect_find_reusable_accumulator): Handle
20171 vector types where the old vector type has a multiple of
20172 the new vector type elements.
20173 (vect_create_partial_epilog): New function, split out from...
20174 (vect_create_epilog_for_reduction): ... here.
20175 (vect_transform_cycle_phi): Reduce the re-used accumulator
20176 to the new vector type.
20178 2021-07-14 Alexandre Oliva <oliva@adacore.com>
20180 * tree-ssa-alias.c (attr_fnspec::verify): Fix index in
20181 non-'t'-sized arg check.
20183 2021-07-14 Alexandre Oliva <oliva@adacore.com>
20185 * tree-cfg.c (cleanup_dead_labels_eh): Update
20186 post_landing_pad label upon change of landing pad block's
20188 (cleanup_dead_labels): Check that a removed label is not that
20191 2021-07-13 Jonathan Wright <jonathan.wright@arm.com>
20193 * combine.c (combine_simplify_rtx): Add vec_select -> subreg
20195 * config/aarch64/aarch64.md (*zero_extend<SHORT:mode><GPI:mode>2_aarch64):
20196 Add Neon to general purpose register case for zero-extend
20198 * config/arm/vfp.md (*arm_movsi_vfp): Remove "*" from *t -> r
20199 case to prevent some cases opting to go through memory.
20200 * cse.c (fold_rtx): Add vec_select -> subreg simplification.
20201 * rtl.c (rtvec_series_p): Define predicate to determine
20202 whether a vector contains a linear series of integers.
20203 * rtl.h (rtvec_series_p): Define.
20204 * rtlanal.c (vec_series_lowpart_p): Define predicate to
20205 determine if a vector selection is equivalent to the low part
20207 * rtlanal.h (vec_series_lowpart_p): Define.
20208 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
20209 Add vec_select -> subreg simplification.
20211 2021-07-13 Paul A. Clarke <pc@us.ibm.com>
20213 * config/rs6000/smmintrin.h (_mm_testz_si128, _mm_testc_si128,
20214 _mm_testnzc_si128, _mm_test_all_ones, _mm_test_all_zeros,
20215 _mm_test_mix_ones_zeros): New.
20217 2021-07-13 Roger Sayle <roger@nextmovesoftware.com>
20218 Richard Biener <rguenther@suse.de>
20220 * gimple.c (gimple_could_trap_p_1): Make S argument a
20221 "const gimple*". Preserve constness in call to
20222 gimple_asm_volatile_p.
20223 (gimple_could_trap_p): Make S argument a "const gimple*".
20224 * gimple.h (gimple_could_trap_p_1, gimple_could_trap_p):
20225 Update function prototypes.
20227 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20229 * tree-vectorizer.h (vect_reusable_accumulator): New structure.
20230 (_loop_vec_info::main_loop_edge): New field.
20231 (_loop_vec_info::skip_main_loop_edge): Likewise.
20232 (_loop_vec_info::skip_this_loop_edge): Likewise.
20233 (_loop_vec_info::reusable_accumulators): Likewise.
20234 (_stmt_vec_info::reduc_scalar_results): Likewise.
20235 (_stmt_vec_info::reused_accumulator): Likewise.
20236 (vect_get_main_loop_result): Declare.
20237 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize
20238 reduc_scalar_inputs.
20239 (vec_info::free_stmt_vec_info): Free reduc_scalar_inputs.
20240 * tree-vect-loop-manip.c (vect_get_main_loop_result): New function.
20241 (vect_do_peeling): Fill an epilogue loop's main_loop_edge,
20242 skip_main_loop_edge and skip_this_loop_edge fields.
20243 * tree-vect-loop.c (INCLUDE_ALGORITHM): Define.
20244 (vect_emit_reduction_init_stmts): New function.
20245 (get_initial_def_for_reduction): Use it.
20246 (get_initial_defs_for_reduction): Likewise. Change the vinfo
20247 parameter to a loop_vec_info.
20248 (vect_create_epilog_for_reduction): Store the scalar results
20249 in the reduc_info. If an epilogue loop is reusing an accumulator
20250 from the main loop, and if the epilogue loop can also be skipped,
20251 try to place the reduction code in the join block. Record
20252 accumulators that could potentially be reused by epilogue loops.
20253 (vect_transform_cycle_phi): When vectorizing epilogue loops,
20254 try to reuse accumulators from the main loop. Record the initial
20255 value in reduc_info for non-SLP reductions too.
20257 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20259 * tree-vect-loop.c (get_initial_def_for_reduction): Remove
20260 adjustment handling. Take the neutral value as an argument,
20261 in place of the code argument.
20262 (vect_transform_cycle_phi): Update accordingly. Handle the
20263 initial values of cond reductions separately from code reductions.
20264 Choose the adjustment here rather than in
20265 get_initial_def_for_reduction. Sink the splat of vec_initial_def.
20267 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20269 * tree-vect-loop.c (neutral_op_for_slp_reduction): Replace with...
20270 (neutral_op_for_reduction): ...this, providing a more general
20272 (vect_create_epilog_for_reduction): Update accordingly.
20273 (vectorizable_reduction): Likewise.
20274 (vect_transform_cycle_phi): Likewise.
20276 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20278 * tree-vect-loop.c (get_initial_def_for_reduction): Take the
20279 reduc_info instead of the original stmt_vec_info.
20280 (vect_transform_cycle_phi): Update accordingly.
20282 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20284 * tree-vect-loop.c (get_initial_defs_for_reduction): Take the
20285 reduc_info as an additional parameter.
20286 (vect_transform_cycle_phi): Update accordingly.
20288 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20290 * tree-vectorizer.h: Include tree-ssa-operands.h.
20291 (vect_phi_initial_value): New function.
20292 * tree-vect-loop.c (neutral_op_for_slp_reduction): Use it.
20293 (get_initial_defs_for_reduction, info_for_reduction): Likewise.
20294 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
20295 (vect_transform_cycle_phi, vectorizable_induction): Likewise.
20297 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20299 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
20300 the phi results to vectype after creating them. Remove later
20301 conversion code that thus becomes redundant.
20303 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20305 * tree-vect-loop.c (vect_create_epilog_for_reduction): Replace
20306 the new_phis vector with a reduc_inputs vector. Combine handling
20307 of reduction chains and ncopies > 1.
20309 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20311 * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
20312 scalar_results to group_size elements after reducing down from
20313 N*group_size elements. Construct an array_slice of the live-out
20314 stmts and assert that there is one stmt per scalar result.
20316 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20318 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
20319 nested_in_vect_loop and use double_reduc everywhere. Remove dead
20320 assignment to "loop".
20322 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
20324 * internal-fn.c (vectorized_internal_fn_supported_p): Handle
20325 vector types first. For scalar types, consider both the preferred
20326 vector mode and the alternative vector modes.
20327 * optabs-query.c (can_vec_mask_load_store_p): Use the same
20328 structure as above, in particular using related_vector_mode
20329 for modes provided by autovectorize_vector_modes.
20331 2021-07-13 Jakub Jelinek <jakub@redhat.com>
20332 Richard Biener <rguenther@suse.de>
20334 PR tree-optimization/101419
20335 * tree-pass.h (PROP_objsz): Define.
20336 (make_pass_early_object_sizes): Declare.
20337 * passes.def (pass_all_early_optimizations): Rename pass_object_sizes
20338 there to pass_early_object_sizes, drop parameter.
20339 (pass_all_optimizations): Move pass_object_sizes right after pass_ccp,
20340 drop parameter, move pass_post_ipa_warn right after that.
20341 * tree-object-size.c (pass_object_sizes::execute): Rename to...
20342 (object_sizes_execute): ... this. Add insert_min_max_p argument.
20343 (pass_data_object_sizes): Move after object_sizes_execute.
20344 (pass_object_sizes): Likewise. In execute method call
20345 object_sizes_execute, drop set_pass_param method and insert_min_max_p
20346 non-static data member and its initializer in the ctor.
20347 (pass_data_early_object_sizes, pass_early_object_sizes,
20348 make_pass_early_object_sizes): New.
20349 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Use
20350 (cfun->curr_properties & PROP_objsz) instead of cfun->after_inlining.
20352 2021-07-13 Kito Cheng <kito.cheng@sifive.com>
20355 * config/riscv/constraints.md ("S"): Update description and remove
20357 * doc/md.texi (Machine Constraints): Document the 'S' constraints
20360 2021-07-13 Richard Biener <rguenther@suse.de>
20363 2021-07-12 Richard Biener <rguenther@suse.de>
20365 * tree-vect-slp.c (vect_slp_region): Show the number of
20366 SLP graph entries in the optimization message.
20368 2021-07-13 Michael Meissner <meissner@linux.ibm.com>
20370 * config/rs6000/altivec.md (xxspltiw_v4sf): Change local variable
20372 * config/rs6000/rs6000-protos.h (rs6000_const_f32_to_i32): Change
20373 return type to long.
20374 * config/rs6000/rs6000.c (rs6000_const_f32_to_i32): Change return
20377 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
20379 * gimple-range-fold.cc (fold_using_range::range_of_builtin_ubsan_call):
20380 Query relation between the 2 operands and use it.
20382 2021-07-12 Sergei Trofimovich <siarheit@google.com>
20384 * doc/cfg.texi: Fix s/ei_safe_safe/ei_safe_edge/ typo.
20386 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
20389 * config/i386/predicates.md (vec_setm_sse41_operand):
20390 Rename from vec_setm_operand.
20391 (vec_setm_avx2_operand): New predicate.
20392 * config/i386/sse.md (vec_set<V_128:mode>): Use V_128 mode iterator.
20393 Use vec_setm_sse41_operand as operand 2 predicate.
20394 (vec_set<V_256_512:mode): New expander.
20395 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_sse41_operand
20396 as operand 2 predicate.
20398 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
20400 PR tree-optimization/101335
20401 * range-op.cc (operator_cast::lhs_op1_relation): Delete.
20403 2021-07-12 Andrew Pinski <apinski@marvell.com>
20405 * tree-ssa-phiopt.c (match_simplify_replacement): Move
20406 insert of the sequence before the movement of the
20407 statement. Check if to see if the statement is used
20408 outside of the original phi to see if we should move it.
20410 2021-07-12 Richard Biener <rguenther@suse.de>
20412 * dump-context.h (debug_dump_context::debug_dump_context):
20413 Add FILE * parameter defaulted to stderr.
20414 * dumpfile.c (debug_dump_context::debug_dump_context): Adjust.
20415 * tree-vect-slp.c (dot_slp_tree): New functions.
20417 2021-07-12 Richard Biener <rguenther@suse.de>
20419 PR tree-optimization/101373
20420 * tree-ssa-pre.c (prune_clobbered_mems): Also prune trapping
20421 references when the BB may not return.
20422 (compute_avail): Pass in the function we're working on and
20423 replace cfun references with it. Externally throwing
20424 const calls also possibly terminate the function.
20425 (pass_pre::execute): Pass down the function we're working on.
20426 * gcse.c (compute_hash_table_work): Externally throwing
20427 const/pure calls also need record_last_mem_set_info.
20428 * postreload-gcse.c (record_opr_changes): Looping or externally
20429 throwing const/pure calls also need record_last_mem_set_info.
20431 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
20433 * recog.c (memory_address_addr_space_p): Change the type to bool.
20434 Return true/false instead of 1/0.
20435 (offsettable_memref_p): Ditto.
20436 (offsettable_nonstrict_memref_p): Ditto.
20437 (offsettable_address_addr_space_p): Ditto.
20438 Change the type of addressp indirect function to bool.
20439 * recog.h (memory_address_addr_space_p): Change the type to bool.
20440 (strict_memory_address_addr_space_p): Ditto.
20441 (offsettable_memref_p): Ditto.
20442 (offsettable_nonstrict_memref_p): Ditto.
20443 (offsettable_address_addr_space_p): Ditto.
20444 * reload.c (maybe_memory_address_addr_space_p): Ditto.
20445 (strict_memory_address_addr_space_p): Change the type to bool.
20446 Return true/false instead of 1/0.
20447 (maybe_memory_address_addr_space_p): Change the type to bool.
20449 2021-07-12 Richard Biener <rguenther@suse.de>
20451 * tree-vect-slp.c (vect_slp_region): Show the number of
20452 SLP graph entries in the optimization message.
20454 2021-07-12 Richard Biener <rguenther@suse.de>
20456 PR tree-optimization/101394
20457 * tree-ssa-pre.c (do_pre_regular_insertion): Avoid inserting
20458 copies from abnormals for a full redundancy.
20460 2021-07-12 Richard Biener <rguenther@suse.de>
20462 PR middle-end/101423
20463 * gimple.c (gimple_could_trap_p_1): Internal function calls
20465 * tree-eh.c (tree_could_trap_p): Likewise.
20467 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
20470 * config/arm/arm_neon.h (vmul_n_u32): Replace call to builtin with
20472 (vmulq_n_u32): Likewise.
20473 (vmul_n_f32): Gate __a * __b on __FAST_MATH__.
20474 (vmulq_n_f32): Likewise.
20475 (vmul_n_f16): Likewise.
20476 (vmulq_n_f16): Likewise.
20478 2021-07-12 Martin Liska <mliska@suse.cz>
20480 PR sanitizer/101425
20481 * gcc.c (check_offload_target_name): Call
20482 candidates_list_and_hint only if we have a candidate.
20484 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
20487 * config/arm/neon.md (vec_init): Move to ...
20488 * config/arm/vec-common.md (vec_init): ... here.
20489 Change the pattern's mode to VDQX and gate it on VALID_MVE_MODE.
20491 2021-07-12 Roger Sayle <roger@nextmovesoftware.com>
20493 PR tree-optimization/101403
20494 * match.pd ((T)bswap(X)>>C): Correctly handle cases where
20495 signedness of the shift is not the same as the signedness of
20496 the type extension.
20498 2021-07-09 Roger Sayle <roger@nextmovesoftware.com>
20499 Uroš Bizjak <ubizjak@gmail.com>
20501 * config/i386/i386.md (*divmodsi4_const): Optimize SImode
20502 divmod of a constant numerator with new define_insn_and_split.
20504 2021-07-09 Iain Sandoe <iain@sandoe.co.uk>
20507 * config/i386/i386-expand.c (ix86_expand_call): If a call is
20508 to a non-local-binding, or local but to a public symbol, then
20509 assume that it might be indirected via the lazy symbol binder.
20510 Mark R10 and R10 as clobbered in that case.
20512 2021-07-09 Eric Botcazou <ebotcazou@adacore.com>
20515 * gcc.c (ASM_DEBUG_DWARF_OPTION): Set again to --gdwarf2 in
20516 the case where HAVE_AS_WORKING_DWARF_N_FLAG is not defined
20517 and HAVE_LD_BROKEN_PE_DWARF5 is defined.
20519 2021-07-09 Uroš Bizjak <ubizjak@gmail.com>
20521 * config/i386/i386.md (*udivmodsi4_pow2_zext_1): Limit the
20522 log2 range of operands[3] to [1,31].
20523 (*udivmodsi4_pow2_zext_2): Ditto. Correct insn RTX pattern.
20525 2021-07-09 Sergei Trofimovich <siarheit@google.com>
20527 * doc/md.texi: Don't split @smallexample in multiple @groups.
20529 2021-07-09 Sergei Trofimovich <siarheit@google.com>
20531 * doc/md.texi: Add missing 'see' word.
20533 2021-07-09 Andrew Pinski <apinski@marvell.com>
20535 * tree-ssa-phiopt.c (phiopt_early_allow): Change arguments
20536 to take sequence and gimple_match_op. Accept the case where
20537 op is a SSA_NAME and one statement in the sequence.
20538 Also allow constants.
20539 (gimple_simplify_phiopt): Always pass a sequence to resimplify.
20540 Update call to phiopt_early_allow. Discard the sequence if not
20543 2021-07-09 Xi Ruoyao <xry111@mengyan1223.wang>
20548 * config/mips/mips.c (mips_const_insns): Use MSA_SUPPORTED_MODE_P
20549 instead of ISA_HAS_MSA.
20550 (mips_expand_vec_unpack): Likewise.
20551 (mips_expand_vector_init): Likewise.
20553 2021-07-09 Kewen Lin <linkw@linux.ibm.com>
20555 * config/rs6000/vsx.md (mods_<mode>): Rename to...
20556 (mod<mode>3): ... this.
20557 (modu_<mode>): Rename to...
20558 (umod<mode>3): ... this.
20559 * config/rs6000/rs6000-builtin.def (MODS_V2DI, MODS_V4SI, MODU_V2DI,
20560 MODU_V4SI): Adjust.
20562 2021-07-08 Jeff Law <jeffreyalaw@gmail.com>
20564 * config/h8300/shiftrotate.md (variable shifts): Expose condition
20565 code handling for the test before the loop.
20567 2021-07-08 Martin Jambor <mjambor@suse.cz>
20570 * ipa-sra.c (class isra_call_summary): New member
20571 m_before_any_store, initialize it in the constructor.
20572 (isra_call_summary::dump): Dump the new field.
20573 (ipa_sra_call_summaries::duplicate): Copy it.
20574 (process_scan_results): Set it.
20575 (isra_write_edge_summary): Stream it.
20576 (isra_read_edge_summary): Likewise.
20577 (param_splitting_across_edge): Only override
20578 safe_to_import_accesses if m_before_any_store is set.
20580 2021-07-08 Martin Sebor <msebor@redhat.com>
20582 PR bootstrap/101374
20583 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref):
20584 Use Object Size Type 0 instead of 1.
20586 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
20588 * tree-vect-loop.c (vectorizable_reduction): Remove always-true
20591 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
20593 * match.pd: Simplify an extend-operate-truncate sequence involving
20596 2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
20597 Richard Biener <rguenther@suse.de>
20599 PR tree-optimization/40210
20600 * match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
20601 (x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
20602 when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
20604 2021-07-08 Uroš Bizjak <ubizjak@gmail.com>
20607 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
20609 * config/i386/mmx.md (V_32): New mode iterator.
20610 (mov<V_32:mode>): Use V_32 mode iterator.
20611 (*mov<V_32:mode>_internal): Ditto.
20612 (*push<V_32:mode>2_rex64): Ditto.
20613 (*push<V_32:mode>2): Ditto.
20614 (movmisalign<V_32:mode>): Ditto.
20615 (mmx_<any_shiftrt:insn>v1si3): New insn pattern.
20616 (sse4_1_<any_extend:code>v2qiv2hi2): Ditto.
20617 (vec_unpacks_lo_v4qi): New expander.
20618 (vec_unpacks_hi_v4qi): Ditto.
20619 (vec_unpacku_lo_v4qi): Ditto.
20620 (vec_unpacku_hi_v4qi): Ditto.
20621 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add V1SImode.
20622 (VALID_INT_MODE_P): Ditto.
20624 2021-07-08 Michael Meissner <meissner@linux.ibm.com>
20627 * config/rs6000/rs6000.md (udivti3): New insn.
20628 (divti3): New insn.
20629 (umodti3): New insn.
20630 (modti3): New insn.
20632 2021-07-07 Martin Sebor <msebor@redhat.com>
20634 PR tree-optimization/100137
20635 PR tree-optimization/99121
20636 PR tree-optimization/97027
20637 * builtins.c (access_ref::access_ref): Also set offmax.
20638 (access_ref::offset_in_range): Define new function.
20639 (access_ref::add_offset): Set offmax.
20640 (access_ref::inform_access): Handle access_none.
20641 (handle_mem_ref): Clear ostype.
20642 (compute_objsize_r): Handle ASSERT_EXPR.
20643 * builtins.h (struct access_ref): Add offmax member.
20644 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Use
20645 compute_objsize() and simplify.
20647 2021-07-07 Peter Bergner <bergner@linux.ibm.com>
20649 * config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
20650 and VSX_BUILTIN_STXVP.
20652 2021-07-07 Martin Sebor <msebor@redhat.com>
20655 * config/aarch64/aarch64.c (aarch64_simd_lane_bounds): Remove
20656 a stray %K from error_at() missed in r12-2088.
20658 2021-07-07 Richard Biener <rguenther@suse.de>
20660 PR tree-optimization/99728
20661 * tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
20663 (mem_refs_may_alias_p): Add assert we handled aggregate
20665 (sm_seq_valid_bb): Give up when running into aggregate copies.
20666 (ref_indep_loop_p): Handle aggregate copies as never
20667 being invariant themselves but allow other refs to be
20668 disambiguated against them.
20669 (can_sm_ref_p): Do not try to apply store-motion to aggregate
20672 2021-07-06 Indu Bhagat <indu.bhagat@oracle.com>
20675 * dwarf2ctf.c (ctf_get_AT_data_member_location): Multiply by 8 to get
20678 2021-07-06 Martin Sebor <msebor@redhat.com>
20680 * gimple-pretty-print.c (percent_G_format): Remove.
20681 * tree-diagnostic.c (default_tree_printer): Remove calls.
20682 * tree-pretty-print.c (percent_K_format): Remove.
20683 * tree-pretty-print.h (percent_K_format): Remove.
20685 2021-07-06 Martin Sebor <msebor@redhat.com>
20687 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_builtin):
20688 Remove %K and use error_at.
20689 (aarch64_expand_fcmla_builtin): Same.
20690 (aarch64_expand_builtin_tme): Same.
20691 (aarch64_expand_builtin_memtag): Same.
20692 * config/arm/arm-builtins.c (arm_expand_acle_builtin): Same.
20693 (arm_expand_builtin): Same.
20694 * config/arm/arm.c (bounds_check): Same.
20696 2021-07-06 Martin Sebor <msebor@redhat.com>
20698 * builtins.c (warn_string_no_nul): Remove %G.
20699 (maybe_warn_for_bound): Same.
20700 (warn_for_access): Same.
20701 (check_access): Same.
20702 (check_strncat_sizes): Same.
20703 (expand_builtin_strncat): Same.
20704 (expand_builtin_strncmp): Same.
20705 (expand_builtin): Same.
20706 (expand_builtin_object_size): Same.
20707 (warn_dealloc_offset): Same.
20708 (maybe_emit_free_warning): Same.
20709 * calls.c (maybe_warn_alloc_args_overflow): Same.
20710 (maybe_warn_nonstring_arg): Same.
20711 (maybe_warn_rdwr_sizes): Same.
20712 * expr.c (expand_expr_real_1): Remove %K.
20713 * gimple-fold.c (gimple_fold_builtin_strncpy): Remove %G.
20714 (gimple_fold_builtin_strncat): Same.
20715 * gimple-ssa-sprintf.c (format_directive): Same.
20716 (handle_printf_call): Same.
20717 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Same.
20718 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
20719 (maybe_diag_access_bounds): Same. Call gimple_location.
20720 (check_bounds_or_overlap): Same.
20721 * trans-mem.c (ipa_tm_scan_irr_block): Remove %K. Simplify.
20722 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Remove %G.
20723 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
20724 (maybe_diag_stxncpy_trunc): Same.
20725 (handle_builtin_stxncpy_strncat): Same.
20726 (maybe_warn_pointless_strcmp): Same.
20727 * tree-ssa-uninit.c (maybe_warn_operand): Same.
20729 2021-07-06 Uroš Bizjak <ubizjak@gmail.com>
20732 * config/i386/predicates.md (vec_setm_operand): Enable
20733 register_operand for TARGET_SSE4_1.
20734 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_operand
20735 as operand 2 predicate. Call ix86_expand_vector_set_var
20736 for non-constant index operand.
20737 (vec_setv4qi): Use vec_setm_mmx_operand as operand 2 predicate.
20738 Call ix86_expand_vector_set_var for non-constant index operand.
20740 2021-07-06 Jeff Law <jeffreyalaw@gmail.com>
20742 * config/h8300/jumpcall.md (*branch): When possible, generate
20743 the comparison in CCZN mode.
20744 * config/h8300/predicates.md (simple_memory_operand): Reject all
20745 auto-increment addressing modes.
20747 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
20749 PR bootstrap/100246
20750 * config/i386/i386.h (struct stringop_algs): Define a CTOR for
20753 2021-07-06 Richard Biener <rguenther@suse.de>
20755 * doc/md.texi (vec_fmaddsub<mode>4): Document.
20756 (vec_fmsubadd<mode>4): Likewise.
20757 * optabs.def (vec_fmaddsub$a4): Add.
20758 (vec_fmsubadd$a4): Likewise.
20759 * internal-fn.def (IFN_VEC_FMADDSUB): Add.
20760 (IFN_VEC_FMSUBADD): Likewise.
20761 * tree-vect-slp-patterns.c (addsub_pattern::recognize):
20762 Refactor to handle IFN_VEC_FMADDSUB and IFN_VEC_FMSUBADD.
20763 (addsub_pattern::build): Likewise.
20764 * tree-vect-slp.c (vect_optimize_slp): CFN_VEC_FMADDSUB
20765 and CFN_VEC_FMSUBADD are not transparent for permutes.
20766 * config/i386/sse.md (vec_fmaddsub<mode>4): New expander.
20767 (vec_fmsubadd<mode>4): Likewise.
20769 2021-07-06 Richard Biener <rguenther@suse.de>
20771 * doc/invoke.texi (fmove-loop-stores): Document.
20772 * common.opt (fmove-loop-stores): New option.
20773 * opts.c (default_options_table): Enable -fmove-loop-stores
20774 at -O1 but not -Og.
20775 * tree-ssa-loop-im.c (pass_lim::execute): Pass
20776 flag_move_loop_stores instead of true to
20777 loop_invariant_motion_in_fun.
20779 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
20781 * doc/install.texi: Document --with-dsymutil.
20783 2021-07-06 Andrew Pinski <apinski@marvell.com>
20785 PR tree-optimization/101256
20786 * dbgcnt.def (phiopt_edge_range): New counter.
20787 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
20788 Check to make sure the new name is defined in the same
20789 bb as the conditional before duplicating range info.
20790 Also add debug counter.
20792 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
20794 PR rtl-optimization/100328
20795 * config/i386/i386-options.c (ix86_option_override_internal):
20796 Set param_ira_consider_dup_in_all_alts to 0.
20798 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
20800 PR rtl-optimization/100328
20801 * doc/invoke.texi (ira-consider-dup-in-all-alts): Document new
20803 * ira.c (ira_get_dup_out_num): Adjust as parameter
20804 param_ira_consider_dup_in_all_alts.
20805 * params.opt (ira-consider-dup-in-all-alts): New.
20806 * ira-conflicts.c (process_regs_for_copy): Add one parameter
20807 single_input_op_has_cstr_p.
20808 (get_freq_for_shuffle_copy): New function.
20809 (add_insn_allocno_copies): Adjust as single_input_op_has_cstr_p.
20810 * ira-int.h (ira_get_dup_out_num): Add one bool parameter.
20812 2021-07-05 Jeff Law <jeffreyalaw@gmail.com>
20814 * config/h8300/shiftrotate.md (shift-by-variable patterns): Update to
20815 generate condition code aware RTL directly.
20817 2021-07-05 Andrew Pinski <apinski@marvell.com>
20819 PR tree-optimization/101039
20820 * match.pd (A CMP 0 ? A : -A): New patterns.
20821 * tree-ssa-phiopt.c (abs_replacement): Delete function.
20822 (tree_ssa_phiopt_worker): Don't call abs_replacement.
20823 Update comment about abs_replacement.
20825 2021-07-05 Andrew Pinski <apinski@marvell.com>
20827 * tree-ssa-phiopt.c (gimple_simplify_phiopt):
20828 If "A ? B : C" fails to simplify, try "(!A) ? C : B".
20830 2021-07-05 Andrew Pinski <apinski@marvell.com>
20832 * tree-ssa-phiopt.c (match_simplify_replacement):
20833 Add early_p argument. Call gimple_simplify_phiopt
20834 instead of gimple_simplify.
20835 (tree_ssa_phiopt_worker): Update call to
20836 match_simplify_replacement and allow unconditionally.
20837 (phiopt_early_allow): New function.
20838 (gimple_simplify_phiopt): New function.
20840 2021-07-05 Andrew Pinski <apinski@marvell.com>
20842 PR middle-end/101237
20843 * fold-const.c (negate_expr_p): Remove call to element_mode
20844 and TREE_MODE/TREE_TYPE when calling HONOR_SIGNED_ZEROS,
20845 HONOR_SIGN_DEPENDENT_ROUNDING, and HONOR_SNANS.
20846 (fold_negate_expr_1): Likewise.
20847 (const_unop): Likewise.
20848 (fold_cond_expr_with_comparison): Likewise.
20849 (fold_binary_loc): Likewise.
20850 (fold_ternary_loc): Likewise.
20851 (tree_call_nonnegative_warnv_p): Likewise.
20852 * match.pd (-(A + B) -> (-B) - A): Likewise.
20854 2021-07-05 Iain Sandoe <iain@sandoe.co.uk>
20856 * configure.ac: Handle --with-dsymutil in the same way as we
20857 do for the assembler and linker. (DEFAULT_DSYMUTIL): New.
20858 Extract the type and version for the dsymutil configured or
20859 found by the default searches.
20860 * config.in: Regenerated.
20861 * configure: Regenerated.
20862 * collect2.c (do_dsymutil): Handle locating dsymutil in the
20863 same way as for the assembler and linker.
20864 * config/darwin.h (DSYMUTIL): Delete.
20865 * gcc.c: Report a configured dsymutil correctly.
20866 * exec-tool.in: Allow for dsymutil.
20868 2021-07-05 Uroš Bizjak <ubizjak@gmail.com>
20870 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
20871 Handle V4QI and V2HI modes.
20872 (expand_vec_perm_blend): Allow 4-byte vector modes with TARGET_SSE4_1.
20873 Handle V4QI mode. Emit mmx_pblendvb32 for 4-byte modes.
20874 (expand_vec_perm_pshufb): Rewrite to use switch statemets.
20875 Handle 4-byte dual operands with TARGET_XOP and single operands
20876 with TARGET_SSSE3. Emit mmx_ppermv32 for TARGET_XOP and
20877 mmx_pshufbv4qi3 for TARGET_SSSE3.
20878 (expand_vec_perm_pblendv): Allow 4-byte vector modes with TARGET_SSE4_1.
20879 (expand_vec_perm_interleave2): Allow 4-byte vector modes.
20880 (expand_vec_perm_pshufb2): Allow 4-byte vector modes with TARGET_SSSE3.
20881 (expand_vec_perm_even_odd_1): Handle V4QI mode.
20882 (expand_vec_perm_broadcast_1): Handle V4QI mode.
20883 (ix86_vectorize_vec_perm_const): Handle V4QI mode.
20884 * config/i386/mmx.md (mmx_ppermv32): New insn pattern.
20885 (mmx_pshufbv4qi3): Ditto.
20886 (*mmx_pblendw32): Ditto.
20887 (*mmx_pblendw64): Rename from *mmx_pblendw.
20888 (mmx_punpckhbw_low): New insn_and_split pattern.
20889 (mmx_punpcklbw_low): Ditto.
20891 2021-07-05 Richard Biener <rguenther@suse.de>
20893 * tree-vect-loop-manip.c (vect_loop_versioning): Do not
20894 set LOOP_C_INFINITE on the vectorized loop.
20896 2021-07-05 Richard Biener <rguenther@suse.de>
20898 PR middle-end/101291
20899 * cfgloopmanip.c (loop_version): Set the loop copy of the
20900 versioned loop to the new loop.
20902 2021-07-04 Iain Sandoe <iain@sandoe.co.uk>
20905 * config.gcc: Ensure that Darwin biarch definitions are
20906 added before i386.h.
20907 * config/i386/darwin.h (TARGET_64BIT): Remove.
20908 (PR80556_WORKAROUND): New.
20909 (REAL_LIBGCC_SPEC): Amend to use PR80556_WORKAROUND.
20910 (DARWIN_SUBARCH_SPEC): New.
20911 * config/i386/darwin32-biarch.h (TARGET_64BIT_DEFAULT,
20912 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
20913 (REAL_LIBGCC_SPEC): Remove.
20914 * config/i386/darwin64-biarch.h (TARGET_64BIT_DEFAULT,
20915 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
20916 (REAL_LIBGCC_SPEC): Remove.
20918 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
20920 PR middle-end/101294
20921 * expr.c (store_constructor): Don't use vec_duplicate on vector.
20923 2021-07-02 Martin Sebor <msebor@redhat.com>
20925 PR middle-end/98871
20926 PR middle-end/98512
20927 * diagnostic.c (get_any_inlining_info): New.
20928 (update_effective_level_from_pragmas): Handle inlining context.
20929 (diagnostic_enabled): Same.
20930 (diagnostic_report_diagnostic): Same.
20931 * diagnostic.h (struct diagnostic_info): Add ctor.
20932 (struct diagnostic_context): Add new member.
20933 * tree-diagnostic.c (set_inlining_locations): New.
20934 (tree_diagnostics_defaults): Set new callback pointer.
20936 2021-07-02 Peter Bergner <bergner@linux.ibm.com>
20938 * config/rs6000/rs6000-builtin.def (BU_MMA_PAIR_LD, BU_MMA_PAIR_ST):
20940 (__builtin_vsx_lxvp, __builtin_vsx_stxvp): New built-ins.
20941 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Expand
20942 lxvp and stxvp built-ins.
20943 (mma_init_builtins): Handle lxvp and stxvp built-ins.
20944 (builtin_function_type): Likewise.
20945 * doc/extend.texi (__builtin_vsx_lxvp, __builtin_mma_stxvp): Document.
20947 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
20949 * config/h8300/h8300-protos.h (compute_a_shift_cc): Accept
20950 additional argument for the code.
20951 * config/h8300/h8300.c (compute_a_shift_cc): Accept additional
20952 argument for the code. Just return if the ZN bits are useful or
20953 not rather than the old style CC_* enums.
20954 * config/h8300/shiftrotate.md (shiftqi_noscratch): Move before
20955 more generic shiftqi patterns.
20956 (shifthi_noscratch, shiftsi_noscratch): Similarly.
20957 (shiftqi_noscratch_set_flags): New pattern.
20958 (shifthi_noscratch_set_flags, shiftsi_noscratch_set_flags): Likewise.
20960 2021-07-02 Andrew MacLeod <amacleod@redhat.com>
20962 PR tree-optimization/101223
20963 * range-op.cc (build_lt): Add -1 for signed values.
20964 (built_gt): Subtract -1 for signed values.
20966 2021-07-02 David Faust <david.faust@oracle.com>
20968 * btfout.c (get_btf_kind): Support BTF_KIND_FLOAT.
20969 (btf_asm_type): Likewise.
20971 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
20973 * config/h8300/h8300-protos.h (output_a_shift): Make first argument
20974 an array of rtx rather than a pointer to rtx. Add code argument.
20975 (compute_a_shift_length): Similarly.
20976 * config/h8300/h8300.c (h8300_shift_costs): Adjust now that the
20977 shift itself isn't an operand. Create dummy operand[0] to carry
20978 a mode and pass a suitable rtx code to compute_a_shift_length.
20979 (get_shift_alg): Adjust operand number of clobber in output templates.
20980 (output_a_shift): Make first argument an array of rtx rather than
20981 a pointer to rtx. Add code argument for the type of shift.
20982 Adjust now that the shift itself is no longer an operand.
20983 (compute_a_shift_length): Similarly.
20984 * config/h8300/shiftrotate.md (shiftqi, shifthi, shiftsi): Use an
20985 iterator rather than nshift_operator.
20986 (shiftqi_noscratch, shifthi_noscratch, shiftsi_noscratch): Likewise.
20987 (shiftqi_clobber_flags): Adjust to API changes in output_a_shift
20988 and compute_a_shift_length.
20989 (shiftqi_noscratch_clobber_flags): Likewise.
20990 (shifthi_noscratch_clobber_flags): Likewise.
20991 (shiftsi_noscratch_clobber_flags): Likewise.
20993 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
20996 * config/darwin.h (DSYMUTIL_SPEC): Do not try to run
20997 dsymutil for BTF/CTF.
20999 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
21002 * config/darwin.h (CTF_INFO_SECTION_NAME): Update the
21003 segment to include BTF.
21004 (BTF_INFO_SECTION_NAME): New.
21006 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
21008 * config/m32r/m32r-protos.h (call_operand): Adjust return type.
21009 (small_data_operand, memreg_operand, small_insn_p): Likewise.
21010 * config/m32r/m32r.c (call_operand): Adjust return type.
21011 (small_data_operand, memreg_operand): Likewise.
21013 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
21015 * config/frv/frv-protos.h (integer_register_operand): Adjust return
21017 (frv_load_operand, gpr_or_fpr_operand, gpr_no_subreg_operand): Likewise.
21018 (fpr_or_int6_operand, gpr_or_int_operand); Likewise.
21019 (gpr_or_int12_operand, gpr_or_int10_operand); Likewise.
21020 (move_source_operand, move_destination_operand): Likewise.
21021 (condexec_source_operand, condexec_dest_operand): Likewise.
21022 (lr_operand, gpr_or_memory_operand, fpr_or_memory_operand): Likewise.
21023 (reg_or_0_operand, fcc_operand, icc_operand, cc_operand): Likewise.
21024 (fcr_operand, icr_operand, cr_operand, call_operand): Likewise.
21025 (fpr_operand, even_reg_operand, odd_reg_operand): Likewise.
21026 (even_gpr_operand, odd_gpr_operand, quad_fpr_operand): Likewise.
21027 (even_fpr_operand, odd_fpr_operand): Likewise.
21028 (dbl_memory_one_insn_operand, dbl_memory_two_insn_operand): Likewise.
21029 (int12_operand, int6_operand, int5_operand, uint5_operand): Likewise.
21030 (uint4_operand, uint1_operand, int_2word_operand): Likewise
21031 (upper_int16_operand, uint16_operand, symbolic_operand): Likewise.
21032 (relational_operator, float_relational_operator): Likewise.
21033 (ccr_eqne_operator, minmax_operator): Likewise.
21034 (condexec_si_binary_operator, condexec_si_media_operator): Likewise.
21035 (condexec_si_divide_operator, condexec_si_unary_operator): Likewise.
21036 (condexec_sf_conv_operator, condexec_sf_add_operator): Likewise.
21037 (intop_compare_operator, acc_operand, even_acc_operand): Likewise.
21038 (quad_acc_operand, accg_operand): Likewise.
21040 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
21042 * config/stormy16/stormy16-protos.h (xstormy16_below_100_symbol): Change
21043 return type to a bool.
21044 (nonimmediate_nonstack_operand): Likewise.
21045 (xstormy16_splittable_below100_operand): Likewise.
21046 * config/stormy16/stormy16.c (xstormy16_below_100_symbol): Fix
21048 (xstormy16_splittable_below100_operand): Likewise.
21050 2021-07-02 Richard Biener <rguenther@suse.de>
21052 PR tree-optimization/101293
21053 * tree-ssa-loop-im.c (mem_ref_hasher::equal): Compare MEM_REF bases
21054 with combined offsets.
21055 (gather_mem_refs_stmt): Hash MEM_REFs as if their offset were
21056 combined with the rest of the offset.
21058 2021-07-02 Eric Botcazou <ebotcazou@adacore.com>
21060 * config/i386/i386.c (asm_preferred_eh_data_format): Always use the
21061 PIC encodings for PE-COFF targets.
21063 2021-07-02 Jakub Jelinek <jakub@redhat.com>
21066 * config/i386/i386-expand.c (ix86_broadcast_from_integer_constant):
21067 Return nullptr for TImode inner mode.
21069 2021-07-02 Richard Biener <rguenther@suse.de>
21071 PR tree-optimization/101280
21072 PR tree-optimization/101173
21073 * gimple-loop-interchange.cc
21074 (tree_loop_interchange::valid_data_dependences): Properly
21075 guard all dependence checks with DDR_REVERSED_P or its
21078 2021-07-02 Hongyu Wang <hongyu.wang@intel.com>
21080 * config/i386/i386-expand.c (ix86_expand_builtin):
21081 Add branch to clear odata when ZF is set for asedecenc_expand
21082 and wideaesdecenc_expand.
21084 2021-07-02 Eugene Rozenfeld <erozen@microsoft.com>
21086 * config/i386/gcc-auto-profile: regenerate
21088 2021-07-02 liuhongt <hongtao.liu@intel.com>
21090 * config/i386/sse.md (trunc<mode><pmov_dst_4>2): Refined to ..
21091 (trunc<mode><pmov_dst_4_lower>2): this.
21093 2021-07-01 David Malcolm <dmalcolm@redhat.com>
21095 * diagnostic.h (diagnostic_context::m_file_cache): New field.
21096 * input.c (class fcache): Rename to...
21097 (class file_cache_slot): ...this, making most members private and
21098 prefixing fields with "m_".
21099 (file_cache_slot::get_file_path): New accessor.
21100 (file_cache_slot::get_use_count): New accessor.
21101 (file_cache_slot::missing_trailing_newline_p): New accessor.
21102 (file_cache_slot::inc_use_count): New.
21103 (fcache_buffer_size): Move to...
21104 (file_cache_slot::buffer_size): ...here.
21105 (fcache_line_record_size): Move to...
21106 (file_cache_slot::line_record_size): ...here.
21107 (fcache_tab): Delete, in favor of global_dc->m_file_cache.
21108 (fcache_tab_size): Move to file_cache::num_file_slots.
21109 (diagnostic_file_cache_init): Update for move of fcache_tab
21110 to global_dc->m_file_cache.
21111 (diagnostic_file_cache_fini): Likewise.
21112 (lookup_file_in_cache_tab): Convert to...
21113 (file_cache::lookup_file): ...this.
21114 (diagnostics_file_cache_forcibly_evict_file): Update for move of
21115 fcache_tab to global_dc->m_file_cache, moving most of
21116 implementation to...
21117 (file_cache::forcibly_evict_file): ...this new function and...
21118 (file_cache_slot::evict): ...this new function.
21119 (evicted_cache_tab_entry): Convert to...
21120 (file_cache::evicted_cache_tab_entry): ...this.
21121 (add_file_to_cache_tab): Convert to...
21122 (file_cache::add_file): ...this, moving bulk of implementation
21124 (file_cache_slot::create): ..this new function.
21125 (file_cache::file_cache): New.
21126 (file_cache::~file_cache): New.
21127 (lookup_or_add_file_to_cache_tab): Convert to...
21128 (file_cache::lookup_or_add_file): ..this new function.
21129 (fcache::fcache): Rename to...
21130 (file_cache_slot::file_cache_slot): ...this, adding "m_" prefixes
21132 (fcache::~fcache): Rename to...
21133 (file_cache_slot::~file_cache_slot): ...this, adding "m_" prefixes
21135 (needs_read): Convert to...
21136 (file_cache_slot::needs_read_p): ...this.
21137 (needs_grow): Convert to...
21138 (file_cache_slot::needs_grow_p): ...this.
21139 (maybe_grow): Convert to...
21140 (file_cache_slot::maybe_grow): ...this.
21141 (read_data): Convert to...
21142 (file_cache_slot::read_data): ...this.
21143 (maybe_read_data): Convert to...
21144 (file_cache_slot::maybe_read_data): ...this.
21145 (get_next_line): Convert to...
21146 (file_cache_slot::get_next_line): ...this.
21147 (goto_next_line): Convert to...
21148 (file_cache_slot::goto_next_line): ...this.
21149 (read_line_num): Convert to...
21150 (file_cache_slot::read_line_num): ...this.
21151 (location_get_source_line): Update for moving of globals to
21152 global_dc->m_file_cache.
21153 (location_missing_trailing_newline): Likewise.
21154 * input.h (class file_cache_slot): New forward decl.
21155 (class file_cache): New.
21157 2021-07-01 Michael Meissner <meissner@linux.ibm.com>
21159 * config/rs6000/rs6000.c (rs6000_maybe_emit_fp_cmove): Add IEEE
21160 128-bit floating point conditional move support.
21161 (have_compare_and_set_mask): Add IEEE 128-bit floating point
21163 * config/rs6000/rs6000.md (mov<mode>cc, IEEE128 iterator): New insn.
21164 (mov<mode>cc_p10, IEEE128 iterator): New insn.
21165 (mov<mode>cc_invert_p10, IEEE128 iterator): New insn.
21166 (fpmask<mode>, IEEE128 iterator): New insn.
21167 (xxsel<mode>, IEEE128 iterator): New insn.
21169 2021-07-01 Iain Sandoe <iain@sandoe.co.uk>
21172 * config/darwin.h (CTF_INFO_SECTION_NAME): New.
21174 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
21176 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
21178 * config/i386/i386-protos.h (ix86_expand_vector_init_duplicate):
21180 * config/i386/sse.md (INT_BROADCAST_MODE): New mode iterator.
21181 (vec_duplicate<mode>): New expander.
21183 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
21186 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
21188 (ix86_byte_broadcast): New function.
21189 (ix86_convert_const_wide_int_to_broadcast): Likewise.
21190 (ix86_expand_move): Convert CONST_WIDE_INT to broadcast if mode
21191 size is 16 bytes or bigger.
21192 (ix86_broadcast_from_integer_constant): New function.
21193 (ix86_expand_vector_move): Convert CONST_WIDE_INT and CONST_VECTOR
21194 to broadcast if mode size is 16 bytes or bigger.
21195 * config/i386/i386-protos.h (ix86_gen_scratch_sse_rtx): New
21197 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): New function.
21199 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
21201 * config/i386/predicates.md (ix86_endbr_immediate_operand):
21202 Return true/false instead of 1/0.
21203 (movq_parallel): Ditto.
21205 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
21207 * recog.c (general_operand): Return true/false instead of 1/0.
21208 (register_operand): Ditto.
21209 (immediate_operand): Ditto.
21210 (const_int_operand): Ditto.
21211 (const_scalar_int_operand): Ditto.
21212 (const_double_operand): Ditto.
21213 (push_operand): Ditto.
21214 (pop_operand): Ditto.
21215 (memory_operand): Ditto.
21216 (indirect_operand): Ditto.
21218 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
21220 * genpreds.c (write_predicate_subfunction):
21221 Change the type of written subfunction to bool.
21222 (write_one_predicate_function):
21223 Change the type of written function to bool.
21224 (write_tm_preds_h): Ditto.
21225 * recog.h (*insn_operand_predicate_fn): Change the type to bool.
21226 * recog.c (general_operand): Change the type to bool.
21227 (address_operand): Ditto.
21228 (register_operand): Ditto.
21229 (pmode_register_operand): Ditto.
21230 (scratch_operand): Ditto.
21231 (immediate_operand): Ditto.
21232 (const_int_operand): Ditto.
21233 (const_scalar_int_operand): Ditto.
21234 (const_double_operand): Ditto.
21235 (nonimmediate_operand): Ditto.
21236 (nonmemory_operand): Ditto.
21237 (push_operand): Ditto.
21238 (pop_operand): Ditto.
21239 (memory_operand): Ditto.
21240 (indirect_operand): Ditto.
21241 (ordered_comparison_operator): Ditto.
21242 (comparison_operator): Ditto.
21243 * config/i386/i386-expand.c (ix86_expand_sse_cmp):
21244 Change the type of indirect predicate function to bool.
21245 * config/rs6000/rs6000.c (easy_vector_constant):
21246 Change the type to bool.
21247 * config/mips/mips-protos.h (m16_based_address_p):
21248 Change the type of operand 3 to bool.
21250 2021-07-01 Richard Biener <rguenther@suse.de>
21252 PR tree-optimization/101280
21253 PR tree-optimization/101173
21254 * gimple-loop-interchange.cc
21255 (tree_loop_interchange::valid_data_dependences): Revert
21256 previous change and instead correctly handle DDR_REVERSED_P
21259 2021-07-01 Richard Biener <rguenther@suse.de>
21261 PR tree-optimization/101278
21262 * tree-ssa-dse.c (dse_classify_store): First check for
21263 uses, then ignore stmt for chaining purposes.
21265 2021-07-01 Richard Biener <rguenther@suse.de>
21267 PR tree-optimization/100778
21268 * tree-vect-slp.c (vect_schedule_slp_node): Do not place trapping
21269 vectorized ops ahead of their scalar BB.
21271 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
21274 * config/i386/i386.md (*nabs<dwi>2_doubleword):
21275 New insn_and_split pattern.
21276 (*nabs<dwi>2_1): Ditto.
21277 * config/i386/i386-features.c
21278 (general_scalar_chain::compute_convert_gain):
21279 Handle (NEG (ABS (...))) RTX. Rewrite src code
21280 scanner as switch statement.
21281 (general_scalar_chain::convert_insn):
21282 Handle (NEG (ABS (...))) RTX.
21283 (general_scalar_to_vector_candidate_p):
21284 Detect (NEG (ABS (...))) RTX. Reorder case statements
21285 for (AND (NOT (...) ...)) fallthrough.
21287 2021-07-01 Richard Biener <rguenther@suse.de>
21289 PR tree-optimization/101178
21290 * tree-vect-slp.c (slpg_vertex::materialize): Remove.
21291 (slpg::perm_in): Add.
21292 (slpg::get_perm_in): Remove.
21293 (slpg::get_perm_materialized): Add.
21294 (vect_optimize_slp): Handle VEC_PERM nodes more optimally
21295 during permute propagation and materialization.
21297 2021-07-01 Jakub Jelinek <jakub@redhat.com>
21300 * dwarf2out.c (loc_list_from_tree_1): Handle COMPOUND_LITERAL_EXPR.
21302 2021-07-01 Jakub Jelinek <jakub@redhat.com>
21304 PR middle-end/94366
21305 * omp-low.c (lower_rec_input_clauses): Rename is_fp_and_or to
21306 is_truth_op, set it for TRUTH_*IF_EXPR regardless of new_var's type,
21307 use boolean_type_node instead of integer_type_node as NE_EXPR type.
21308 (lower_reduction_clauses): Likewise.
21310 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
21312 * config/gcn/gcn.c: Include dwarf2.h.
21313 (gcn_addr_space_debug): New function.
21314 (TARGET_ADDR_SPACE_DEBUG): New hook.
21316 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
21318 * common/config/gcn/gcn-common.c
21319 (gcn_option_optimization_table): Change OPT_fomit_frame_pointer to -O3.
21320 * config/gcn/gcn.c (gcn_expand_prologue): Prefer the frame pointer
21322 (gcn_expand_prologue): Prefer the frame pointer when emitting CFI.
21323 (gcn_frame_pointer_rqd): New function.
21324 (TARGET_FRAME_POINTER_REQUIRED): New hook.
21326 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
21328 * config/gcn/gcn.c (move_callee_saved_registers): Emit CFI notes for
21329 prologue register saves.
21330 (gcn_debug_unwind_info): Use UI_DWARF2.
21331 (gcn_dwarf_register_number): Map DWARF_LINK_REGISTER to DWARF PC.
21332 (gcn_dwarf_register_span): DWARF_LINK_REGISTER doesn't span.
21333 * config/gcn/gcn.h: (DWARF_FRAME_RETURN_COLUMN): New define.
21334 (DWARF_LINK_REGISTER): New define.
21335 (FIRST_PSEUDO_REGISTER): Increment.
21336 (FIXED_REGISTERS): Add entry for DWARF_LINK_REGISTER.
21337 (CALL_USED_REGISTERS): Likewise.
21338 (REGISTER_NAMES): Likewise.
21340 2021-06-30 Richard Biener <rguenther@suse.de>
21342 PR tree-optimization/101267
21343 * tree-vect-stmts.c (vect_check_scalar_mask): Adjust
21344 API and use SLP compatible interface of vect_is_simple_use.
21345 Reject not vectorized SLP defs for callers that do not support
21347 (vect_check_store_rhs): Handle masked stores and pass down
21348 the appropriate operator index.
21349 (vectorizable_call): Adjust.
21350 (vectorizable_store): Likewise.
21351 (vectorizable_load): Likewise. Handle SLP pecularity of
21353 (vect_is_simple_use): Remove special-casing of masked stores.
21355 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
21357 * common.opt (foffload): Remove help as Driver only.
21358 * gcc.c (display_help): Add -foffload.
21360 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
21362 * gcc.c (close_at_file, execute): Replace alloca by XALLOCAVEC.
21363 (check_offload_target_name): Fix splitting OFFLOAD_TARGETS into
21364 a candidate list; better inform no offload target is configured
21365 and fix hint extraction when passed target is not '\0' at [len].
21366 * common.opt (foffload): Add tailing '.'.
21367 (foffload-options): Likewise; fix flag name in the help string.
21369 2021-06-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
21372 * config/arm/arm_neon.h: Move vabs intrinsics before vcage_f32.
21373 (vcage_f32): Gate comparison on __FAST_MATH__.
21374 (vcageq_f32): Likewise.
21375 (vcale_f32): Likewise.
21376 (vcaleq_f32): Likewise.
21377 (vcagt_f32): Likewise.
21378 (vcagtq_f32): Likewise.
21379 (vcalt_f32): Likewise.
21380 (vcaltq_f32): Likewise.
21381 (vcage_f16): Likewise.
21382 (vcageq_f16): Likewise.
21383 (vcale_f16): Likewise.
21384 (vcaleq_f16): Likewise.
21385 (vcagt_f16): Likewise.
21386 (vcagtq_f16): Likewise.
21387 (vcalt_f16): Likewise.
21388 (vcaltq_f16): Likewise.
21390 2021-06-30 Richard Biener <rguenther@suse.de>
21392 PR tree-optimization/101264
21393 * tree-vect-slp.c (vect_optimize_slp): Propagate the
21394 computed perm_in to all "any" permute successors
21395 we cannot de-duplicate immediately.
21397 2021-06-30 liuhongt <hongtao.liu@intel.com>
21400 * config/i386/sse.md
21401 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>):
21403 (avx512f_sfixupimm<mode><maskz_scalar_name><round_saeonly_name>):
21405 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>"): Refined.
21406 * config/i386/subst.md (maskz_scalar): New define_subst.
21407 (maskz_scalar_name): New subst_attr.
21408 (maskz_scalar_op5): Ditto.
21409 (round_saeonly_maskz_scalar_op5): Ditto.
21410 (round_saeonly_maskz_scalar_operand5): Ditto.
21412 2021-06-30 David Edelsohn <dje.gcc@gmail.com>
21414 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
21415 Increase code CSECT alignment to at least 32 bytes.
21416 * config/rs6000/xcoff.h (TEXT_SECTION_ASM_OP): Add 32 byte
21417 alignment designation.
21419 2021-06-29 Sergei Trofimovich <siarheit@google.com>
21421 * doc/generic.texi: Fix s/net yet/not yet/ typo.
21423 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
21425 PR tree-optimization/101254
21426 * range-op.cc (operator_minus::op1_op2_relation_effect): Check for
21427 wrapping/non-wrapping when setting the result range.
21429 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
21431 * value-query.cc (gimple_range_global): Allow phis.
21433 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
21435 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): Use stmt.
21436 (simplify_using_ranges::op_with_boolean_value_range_p): Add a
21437 statement for location context.
21438 (check_for_binary_op_overflow): Ditto.
21439 (simplify_using_ranges::get_vr_for_comparison): Ditto.
21440 (simplify_using_ranges::compare_name_with_value): Ditto.
21441 (simplify_using_ranges::compare_names): Ditto.
21442 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Ditto.
21443 (simplify_using_ranges::simplify_truth_ops_using_ranges): Ditto.
21444 (simplify_using_ranges::simplify_min_or_max_using_ranges): Ditto.
21445 (simplify_using_ranges::simplify_internal_call_using_ranges): Ditto.
21446 (simplify_using_ranges::two_valued_val_range_p): Ditto.
21447 (simplify_using_ranges::simplify): Ditto.
21448 * vr-values.h: Adjust prototypes.
21450 2021-06-29 Uroš Bizjak <ubizjak@gmail.com>
21453 * config/i386/mmx.md (vec_addsubv2sf3): New insn pattern.
21455 2021-06-29 Julian Brown <julian@codesourcery.com>
21457 * config/gcn/gcn.c (gcn_init_libfuncs): New function.
21458 (TARGET_INIT_LIBFUNCS): Define target hook using above function.
21459 * config/gcn/gcn.h (UNITS_PER_WORD): Define to 8 for IN_LIBGCC2, 4
21461 (LIBGCC2_UNITS_PER_WORD, BITS_PER_WORD): Remove definitions.
21462 (MAX_FIXED_MODE_SIZE): Change to 128.
21464 2021-06-29 Julian Brown <julian@codesourcery.com>
21466 * config/gcn/gcn.md (UNSPEC_FLBIT_INT): New unspec constant.
21467 (s_mnemonic): Add clrsb.
21468 (gcn_flbit<mode>_int): Add insn pattern for SImode/DImode.
21469 (clrsb<mode>2): Add expander for SImode/DImode.
21471 2021-06-29 Julian Brown <julian@codesourcery.com>
21473 * config/gcn/gcn.md (<su>mulsidi3, <su>mulsidi3_reg, <su>mulsidi3_imm,
21474 muldi3): Add patterns.
21476 2021-06-29 Julian Brown <julian@codesourcery.com>
21478 * config/gcn/gcn.md (<su>mulsi3_highpart): Change to expander.
21479 (<su>mulsi3_highpart_reg, <su>mulsi3_highpart_imm): New patterns.
21481 2021-06-29 Julian Brown <julian@codesourcery.com>
21483 * config/gcn/gcn.md (mulsi3): Make s_mulk_i32 variant clobber SCC.
21485 2021-06-29 Joseph Myers <joseph@codesourcery.com>
21487 * btfout.c, ctfout.c: Include "memmodel.h".
21489 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
21491 * gcc.c (check_offload_target_name): Cast len argument to
21492 %q.*s to 'int'; avoid -Wstringop-truncation warning.
21494 2021-06-29 Richard Biener <rguenther@suse.de>
21496 * tree-vect-slp.c (vect_optimize_slp): Forward propagate
21497 to "any" permute nodes and relax "any" permute proapgation
21498 during iterative backward propagation.
21500 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
21503 * common.opt (-foffload=): Update description.
21504 (-foffload-options=): New.
21505 * doc/invoke.texi (C Language Options): Document
21506 -foffload and -foffload-options.
21507 * gcc.c (check_offload_target_name): New, split off from
21508 handle_foffload_option.
21509 (check_foffload_target_names): New.
21510 (handle_foffload_option): Handle -foffload=default.
21511 (driver_handle_option): Update for -foffload-options.
21512 * lto-opts.c (lto_write_options): Use -foffload-options
21513 instead of -foffload.
21514 * lto-wrapper.c (merge_and_complain, append_offload_options):
21516 * opts.c (common_handle_option): Likewise.
21518 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
21520 * doc/invoke.texi (C Language Options): Sort options
21521 alphabetically in optlist and also the description itself.
21522 Remove leftover -fallow-single-precision from and add missing
21523 -fgnu-tm to the optlist.
21525 2021-06-29 Richard Biener <rguenther@suse.de>
21527 * tree-vect-slp.c (slpg_vertex::visited): Remove.
21528 (vect_slp_perms_eq): Handle -1 permutes.
21529 (vect_optimize_slp): Rewrite permute propagation.
21531 2021-06-29 Jakub Jelinek <jakub@redhat.com>
21534 * match.pd ((intptr_t)x eq/ne CST to x eq/ne (typeof x) CST): Don't
21535 perform the optimization in GENERIC when sanitizing and x has a
21538 2021-06-29 Richard Biener <rguenther@suse.de>
21540 PR tree-optimization/101242
21541 * tree-vect-slp.c (vect_slp_build_vertices): Force-add
21542 PHIs with not represented initial values as leafs.
21544 2021-06-29 Jan-Benedict Glaw <jbglaw@getslash.de>
21546 * config/pdp11/pdp11.h (ASM_OUTPUT_SKIP): Fix signedness warning.
21547 * config/pdp11/pdp11.c (pdp11_asm_print_operand_punct_valid_p): Remove
21548 "register" keyword.
21549 (pdp11_initial_elimination_offset) Remove unused variable.
21550 (pdp11_cmp_length) Ditto.
21551 (pdp11_insn_cost): Ditto, and fix signedness warning.
21553 2021-06-29 David Edelsohn <dje.gcc@gmail.com>
21555 * btfout.c: Include tm_p.h.
21558 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
21560 * config/bpf/bpf.c (bpf_expand_prologue): Do not mark insns as
21562 (bpf_expand_epilogue): Likewise.
21563 * config/bpf/bpf.h (DWARF2_FRAME_INFO): Define to 0.
21564 Do not define DBX_DEBUGGING_INFO.
21566 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
21568 * doc/invoke.texi: Document the CTF and BTF debug info options.
21570 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
21571 David Faust <david.faust@oracle.com>
21572 Jose E. Marchesi <jose.marchesi@oracle.com>
21573 Weimin Pan <weimin.pan@oracle.com>
21575 * Makefile.in: Add ctfc.*, ctfout.c and btfout.c files to
21576 GTFILES. Add new object files.
21577 * common.opt: Add CTF and BTF debug info options.
21578 * btfout.c: New file.
21579 * ctfc.c: Likewise.
21580 * ctfc.h: Likewise.
21581 * ctfout.c: Likewise.
21582 * dwarf2ctf.c: Likewise.
21583 * dwarf2ctf.h: Likewise.
21584 * dwarf2cfi.c (dwarf2out_do_frame): Acknowledge CTF_DEBUG and
21586 * dwarf2out.c (dwarf2out_source_line): Likewise.
21587 (dwarf2out_finish): Skip emitting DWARF if CTF or BTF are to
21589 (debug_format_do_cu): New function.
21590 (dwarf2out_early_finish): Traverse DIEs and emit CTF/BTF for
21592 Include dwarf2ctf.c.
21593 * final.c (dwarf2_debug_info_emitted_p): Acknowledge DWARF-based debug
21595 * flag-types.h (enum debug_info_type): Add CTF_DEBUG and BTF_DEBUG.
21596 (CTF_DEBUG): New bitmask.
21597 (BTF_DEBUG): Likewise.
21598 (enum ctf_debug_info_levels): New enum.
21599 * gengtype.c (open_base_files): Handle ctfc.h.
21600 (main): Handle uint32_t type.
21601 * flags.h (btf_debuginfo_p): New definition.
21602 (dwarf_based_debuginfo_p): Likewise.
21603 * opts.c (debug_type_names): Add entries for CTF and BTF.
21604 (btf_debuginfo_p): New function.
21605 (dwarf_based_debuginfo_p): Likewise.
21606 (common_handle_option): Handle -gctfN and -gbtf options.
21607 (set_debug_level): Set CTF_DEBUG, BTF_DEBUG whenever appropriate.
21608 * toplev.c (process_options): Inform the user and ignore -gctfLEVEL if
21611 2021-06-28 Jose E. Marchesi <jose.marchesi@oracle.com>
21613 * dwarf2out.c (AT_class): Function is no longer static.
21614 (AT_int): Likewise.
21615 (AT_unsigned): Likewise.
21616 (AT_loc): Likewise.
21617 (get_AT): Likewise.
21618 (get_AT_string): Likewise.
21619 (get_AT_flag): Likewise.
21620 (get_AT_unsigned): Likewise.
21621 (get_AT_ref): Likewise.
21622 (new_die_raw): Likewise.
21623 (lookup_decl_die): Likewise.
21624 (base_type_die): Likewise.
21625 (add_name_attribute): Likewise.
21626 (add_AT_int): Likewise.
21627 (add_AT_unsigned): Likewise.
21628 (add_AT_loc): Likewise.
21629 (dw_get_die_tag): New function.
21630 (dw_get_die_child): Likewise.
21631 (dw_get_die_sib): Likewise.
21632 (struct dwarf_file_data): Move from here to dwarf2out.h
21633 (struct dw_attr_struct): Likewise.
21634 * dwarf2out.h: Analogous changes.
21636 2021-06-28 Martin Jambor <mjambor@suse.cz>
21639 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
21640 members m_dead_stmts and m_dead_ssas.
21641 * ipa-param-manipulation.c
21642 (ipa_param_body_adjustments::mark_dead_statements): New function.
21643 (ipa_param_body_adjustments::common_initialization): Call it on
21644 all removed but not split parameters.
21645 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
21647 (ipa_param_body_adjustments::modify_call_stmt): Remove arguments that
21649 * tree-inline.c (remap_gimple_stmt): Do not copy dead statements, reset
21650 dead debug statements.
21651 (copy_phis_for_bb): Do not copy dead PHI nodes.
21653 2021-06-28 Martin Jambor <mjambor@suse.cz>
21656 * symtab-clones.h (clone_info): Removed member param_adjustments.
21657 * ipa-param-manipulation.h: Adjust initial comment to reflect how we
21658 deal with pass-through splits now.
21659 (ipa_param_performed_split): Removed.
21660 (ipa_param_adjustments::modify_call): Adjusted parameters.
21661 (class ipa_param_body_adjustments): Adjusted parameters of
21662 register_replacement, modify_gimple_stmt and modify_call_stmt.
21663 (ipa_verify_edge_has_no_modifications): Declare.
21664 (ipa_edge_modifications_finalize): Declare.
21665 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Remove
21666 performed_splits processing, pas only edge to padjs->modify_call,
21667 check that call arguments were not modified if they should not have
21669 * cgraphclones.c (cgraph_node::create_clone): Do not copy performed
21671 * ipa-param-manipulation.c (struct pass_through_split_map): New type.
21672 (ipa_edge_modification_info): Likewise.
21673 (ipa_edge_modification_sum): Likewise.
21674 (ipa_edge_modifications): New edge summary.
21675 (ipa_verify_edge_has_no_modifications): New function.
21676 (transitive_split_p): Removed.
21677 (transitive_split_map): Likewise.
21678 (init_transitive_splits): Likewise.
21679 (ipa_param_adjustments::modify_call): Adjusted to use the new edge
21680 summary instead of performed_splits.
21681 (ipa_param_body_adjustments::register_replacement): Drop dummy
21682 parameter, set base_index of the created ipa_param_body_replacement.
21683 (phi_arg_will_live_p): New function.
21684 (ipa_param_body_adjustments::common_initialization): Do not create
21685 IPA_SRA dummy decls.
21686 (simple_tree_swap_info): Removed.
21687 (remap_split_decl_to_dummy): Likewise.
21688 (record_argument_state_1): New function.
21689 (record_argument_state): Likewise.
21690 (ipa_param_body_adjustments::modify_call_stmt): New parameter
21691 orig_stmt. Do not work with dummy decls, save necessary info about
21692 changes to ipa_edge_modifications.
21693 (ipa_param_body_adjustments::modify_gimple_stmt): New parameter
21694 orig_stmt, pass it to modify_call_stmt.
21695 (ipa_param_body_adjustments::modify_cfun_body): Adjust call to
21696 modify_gimple_stmt.
21697 (ipa_edge_modifications_finalize): New function.
21698 * tree-inline.c (remap_gimple_stmt): Pass original statement to
21699 modify_gimple_stmt.
21700 (copy_phis_for_bb): Do not copy dead PHI nodes.
21701 (expand_call_inline): Do not remap performed_splits.
21702 (update_clone_info): Likewise.
21703 * toplev.c: Include ipa-param-manipulation.h.
21704 (toplev::finalize): Call ipa_edge_modifications_finalize.
21706 2021-06-28 Andrew Pinski <apinski@marvell.com>
21708 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Duplicate range
21709 info if we're the only things setting the target PHI.
21710 (value_replacement): Don't duplicate range here.
21711 (minmax_replacement): Likewise.
21713 2021-06-28 Richard Biener <rguenther@suse.de>
21715 PR tree-optimization/101229
21716 * gimple-walk.c (gimple_walk_op): Handle PHIs.
21718 2021-06-28 Martin Liska <mliska@suse.cz>
21720 * config/v850/v850.c (construct_dispose_instruction): Allocate
21722 (construct_prepare_instruction): Likewise.
21724 2021-06-28 Martin Liska <mliska@suse.cz>
21726 * config/v850/v850.c (v850_option_override): Build default
21728 (v850_can_inline_p): New. Allow MASK_PROLOG_FUNCTION to be
21729 ignored for inlining.
21730 (TARGET_CAN_INLINE_P): New.
21732 2021-06-28 Richard Biener <rguenther@suse.de>
21734 PR tree-optimization/101207
21735 * tree-vect-slp.c (vect_optimize_slp): Do BB reduction
21736 permute eliding for load permutations properly.
21738 2021-06-28 Richard Biener <rguenther@suse.de>
21740 PR tree-optimization/101173
21741 * gimple-loop-interchange.cc
21742 (tree_loop_interchange::valid_data_dependences): Disallow outer
21743 loop dependence distance of zero.
21745 2021-06-28 liuhongt <hongtao.liu@intel.com>
21748 * config/i386/sse.md (*avx_cmp<mode>3_lt): New
21749 define_insn_and_split.
21750 (*avx_cmp<mode>3_ltint): Ditto.
21751 (*avx2_pcmp<mode>3_3): Ditto.
21752 (*avx2_pcmp<mode>3_4): Ditto.
21753 (*avx2_pcmp<mode>3_5): Ditto.
21755 2021-06-28 liuhongt <hongtao.liu@intel.com>
21757 * config/i386/i386-builtin.def (IX86_BUILTIN_BLENDVPD256,
21758 IX86_BUILTIN_BLENDVPS256, IX86_BUILTIN_PBLENDVB256,
21759 IX86_BUILTIN_BLENDVPD, IX86_BUILTIN_BLENDVPS,
21760 IX86_BUILTIN_PBLENDVB128): Replace icode with
21762 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold blendv
21764 * config/i386/sse.md (*<sse4_1_avx2>_pblendvb_lt_subreg_not):
21765 New pre_reload splitter.
21767 2021-06-27 Andrew Pinski <apinski@marvell.com>
21769 PR middle-end/101230
21770 * fold-const.c (fold_ternary_loc): Check
21771 the return value of invert_tree_comparison.
21773 2021-06-27 David Edelsohn <dje.gcc@gmail.com>
21775 * config.gcc: Add SPDX License Identifier.
21776 (powerpc-ibm-aix789): Default to aix73.h.
21777 (powerpc-ibm-aix7.2.*.*): New stanza.
21778 * config/rs6000/aix72.h: Add SPDX License Identifier.
21779 * config/rs6000/aix73.h: New file.
21781 2021-06-26 Jason Merrill <jason@redhat.com>
21783 * except.c: #include "dwarf2.h" instead of "dwarf2out.h".
21785 2021-06-26 Andrew Pinski <apinski@marvell.com>
21787 * genmatch.c (lower_cond): Copy for_subst_vec
21788 for the simplify also.
21789 (lower): Swap the order for lower_for and lower_cond.
21791 2021-06-26 Andrew Pinski <apinski@marvell.com>
21793 * tree-ssa-phiopt.c (match_simplify_replacement): Reset
21794 flow senatitive info on the moved ssa set.
21796 2021-06-26 Andrew Pinski <apinski@marvell.com>
21798 * fold-const.c (fold_cond_expr_with_comparison):
21799 Exand arg0 into comp_code, arg00, and arg01.
21800 (fold_ternary_loc): Use invert_tree_comparison
21801 instead of fold_invert_truthvalue for the case
21802 where we have A CMP B ? C : A.
21804 2021-06-25 Martin Sebor <msebor@redhat.com>
21806 PR middle-end/101216
21807 * calls.c (maybe_warn_rdwr_sizes): Use the no_warning constant.
21809 2021-06-25 Jeff Law <jeffreyalaw@gmail.com>
21811 * config/h8300/h8300.c (select_cc_mode): Handle ASHIFTRT and LSHIFTRT.
21813 2021-06-25 Richard Biener <rguenther@suse.de>
21815 PR tree-optimization/101202
21816 * tree-vect-slp.c (vect_optimize_slp): Explicitely handle
21819 2021-06-25 Richard Biener <rguenther@suse.de>
21821 * tree-vect-slp-patterns.c (addsub_pattern::build): Copy
21822 STMT_VINFO_REDUC_DEF from the original representative.
21824 2021-06-25 Martin Sebor <msebor@redhat.com>
21826 * builtins.c (warn_string_no_nul): Replace uses of TREE_NO_WARNING,
21827 gimple_no_warning_p and gimple_set_no_warning with
21828 warning_suppressed_p, and suppress_warning.
21830 (maybe_warn_for_bound): Same.
21831 (warn_for_access): Same.
21832 (check_access): Same.
21833 (expand_builtin_strncmp): Same.
21834 (fold_builtin_varargs): Same.
21835 * calls.c (maybe_warn_nonstring_arg): Same.
21836 (maybe_warn_rdwr_sizes): Same.
21837 * cfgexpand.c (expand_call_stmt): Same.
21838 * cgraphunit.c (check_global_declaration): Same.
21839 * fold-const.c (fold_undefer_overflow_warnings): Same.
21840 (fold_truth_not_expr): Same.
21841 (fold_unary_loc): Same.
21842 (fold_checksum_tree): Same.
21843 * gimple-array-bounds.cc (array_bounds_checker::check_array_ref): Same.
21844 (array_bounds_checker::check_mem_ref): Same.
21845 (array_bounds_checker::check_addr_expr): Same.
21846 (array_bounds_checker::check_array_bounds): Same.
21847 * gimple-expr.c (copy_var_decl): Same.
21848 * gimple-fold.c (gimple_fold_builtin_strcpy): Same.
21849 (gimple_fold_builtin_strncat): Same.
21850 (gimple_fold_builtin_stxcpy_chk): Same.
21851 (gimple_fold_builtin_stpcpy): Same.
21852 (gimple_fold_builtin_sprintf): Same.
21853 (fold_stmt_1): Same.
21854 * gimple-ssa-isolate-paths.c (diag_returned_locals): Same.
21855 * gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Same.
21856 * gimple-ssa-sprintf.c (handle_printf_call): Same.
21857 * gimple-ssa-store-merging.c (imm_store_chain_info::output_merged_store): Same.
21858 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
21859 * gimple-ssa-warn-restrict.h: Adjust declarations.
21860 (maybe_diag_access_bounds): Replace uses of TREE_NO_WARNING,
21861 gimple_no_warning_p and gimple_set_no_warning with
21862 warning_suppressed_p, and suppress_warning.
21863 (check_call): Same.
21864 (check_bounds_or_overlap): Same.
21865 * gimple.c (gimple_build_call_from_tree): Same.
21866 * gimplify.c (gimplify_return_expr): Same.
21867 (gimplify_cond_expr): Same.
21868 (gimplify_modify_expr_complex_part): Same.
21869 (gimplify_modify_expr): Same.
21870 (gimple_push_cleanup): Same.
21871 (gimplify_expr): Same.
21872 * omp-expand.c (expand_omp_for_generic): Same.
21873 (expand_omp_taskloop_for_outer): Same.
21874 * omp-low.c (lower_rec_input_clauses): Same.
21875 (lower_lastprivate_clauses): Same.
21876 (lower_send_clauses): Same.
21877 (lower_omp_target): Same.
21878 * tree-cfg.c (pass_warn_function_return::execute): Same.
21879 * tree-complex.c (create_one_component_var): Same.
21880 * tree-inline.c (remap_gimple_op_r): Same.
21881 (copy_tree_body_r): Same.
21882 (declare_return_variable): Same.
21883 (expand_call_inline): Same.
21884 * tree-nested.c (lookup_field_for_decl): Same.
21885 * tree-sra.c (create_access_replacement): Same.
21886 (generate_subtree_copies): Same.
21887 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Same.
21888 * tree-ssa-forwprop.c (combine_cond_expr_cond): Same.
21889 * tree-ssa-loop-ch.c (ch_base::copy_headers): Same.
21890 * tree-ssa-loop-im.c (execute_sm): Same.
21891 * tree-ssa-phiopt.c (cond_store_replacement): Same.
21892 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
21893 (handle_builtin_strcpy): Same.
21894 (maybe_diag_stxncpy_trunc): Same.
21895 (handle_builtin_stxncpy_strncat): Same.
21896 (handle_builtin_strcat): Same.
21897 * tree-ssa-uninit.c (get_no_uninit_warning): Same.
21898 (set_no_uninit_warning): Same.
21899 (uninit_undefined_value_p): Same.
21900 (warn_uninit): Same.
21901 (maybe_warn_operand): Same.
21902 * tree-vrp.c (compare_values_warnv): Same.
21903 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr): Same.
21904 (test_for_singularity): Same.
21905 * gimple.h (warning_suppressed_p): New function.
21906 (suppress_warning): Same.
21907 (copy_no_warning): Same.
21908 (gimple_set_block): Call gimple_set_location.
21909 (gimple_set_location): Call copy_warning.
21911 2021-06-25 Martin Sebor <msebor@redhat.com>
21913 * tree.h (warning_suppressed_at, copy_warning,
21914 warning_suppressed_p, suppress_warning): New functions.
21916 2021-06-25 Martin Sebor <msebor@redhat.com>
21918 * Makefile.in (OBJS-libcommon): Add diagnostic-spec.o.
21919 * gengtype.c (open_base_files): Add diagnostic-spec.h.
21920 * diagnostic-spec.c: New file.
21921 * diagnostic-spec.h: New file.
21922 * tree.h (no_warning, all_warnings, suppress_warning_at): New
21924 * warning-control.cc: New file.
21926 2021-06-25 liuhongt <hongtao.liu@intel.com>
21929 * config/i386/i386.c (x86_order_regs_for_local_alloc):
21932 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
21934 PR tree-optimization/101189
21935 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Pass
21936 LHS range of condition to postfold routine.
21937 (fold_using_range::postfold_gcond_edges): Only process the TRUE or
21938 FALSE edge if the LHS range supports it being taken.
21939 * gimple-range-fold.h (postfold_gcond_edges): Add range parameter.
21941 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
21943 * value-relation.cc (equiv_oracle::dump): Do not dump NULL blocks.
21944 (relation_oracle::find_relation_block): Check correct bitmap.
21945 (relation_oracle::dump): Do not dump NULL blocks.
21947 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
21949 * gimple-range-cache.cc (ranger_cache::propagate_cache): Call
21950 range_on_edge instead of manually calculating.
21952 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
21954 * range-op.cc: Fix comment.
21956 2021-06-24 Uroš Bizjak <ubizjak@gmail.com>
21959 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
21960 Handle V8QI and V4HI modes.
21961 * config/i386/mmx.md (sse4_1_<any_extend:code>v4qiv4hi2):
21963 (sse4_1_<any_extend:code>v4qiv4hi2): Ditto.
21964 (mmxpackmode): New mode attribute.
21965 (vec_pack_trunc_<mmxpackmode:mode>): New expander.
21966 (mmxunpackmode): New mode attribute.
21967 (vec_unpacks_lo_<mmxunpackmode:mode>): New expander.
21968 (vec_unpacks_hi_<mmxunpackmode:mode>): Ditto.
21969 (vec_unpacku_lo_<mmxunpackmode:mode>): Ditto.
21970 (vec_unpacku_hi_<mmxunpackmode:mode>): Ditto.
21971 * config/i386/i386.md (extsuffix): Move from ...
21972 * config/i386/sse.md: ... here.
21974 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
21976 * dwarf2out.c (dwarf2out_assembly_start): Emit .file 0 marker here..
21977 (dwarf2out_finish): ...instead of here.
21979 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
21981 * configure.ac (--gdwarf-5 option): Use objdump instead of readelf.
21982 (working --gdwarf-4/--gdwarf-5 for all sources): Likewise.
21983 (--gdwarf-4 not refusing generated .debug_line): Adjust for Windows.
21984 * configure: Regenerate.
21986 2021-06-24 Richard Biener <rguenther@suse.de>
21988 * config/i386/sse.md (vec_addsubv4df3, vec_addsubv2df3,
21989 vec_addsubv8sf3, vec_addsubv4sf3): Merge into ...
21990 (vec_addsub<mode>3): ... using a new addsub_cst mode attribute.
21992 2021-06-24 Richard Biener <rguenther@suse.de>
21994 * config/i386/sse.md (avx_addsubv4df3): Rename to
21996 (avx_addsubv8sf3): Rename to vec_addsubv8sf3.
21997 (sse3_addsubv2df3): Rename to vec_addsubv2df3.
21998 (sse3_addsubv4sf3): Rename to vec_addsubv4sf3.
21999 * config/i386/i386-builtin.def: Adjust.
22000 * internal-fn.def (VEC_ADDSUB): New internal optab fn.
22001 * optabs.def (vec_addsub_optab): New optab.
22002 * tree-vect-slp-patterns.c (class addsub_pattern): New.
22003 (slp_patterns): Add addsub_pattern.
22004 * tree-vect-slp.c (vect_optimize_slp): Disable propagation
22005 across CFN_VEC_ADDSUB.
22006 * tree-vectorizer.h (vect_pattern::vect_pattern): Make
22008 * doc/md.texi (vec_addsub<mode>3): Document.
22010 2021-06-24 Jakub Jelinek <jakub@redhat.com>
22012 PR middle-end/101170
22013 * df-scan.c (df_ref_record): For paradoxical big-endian SUBREGs
22014 where regno + subreg_regno_offset wraps around use 0 as starting
22017 2021-06-24 Jakub Jelinek <jakub@redhat.com>
22019 PR middle-end/101172
22020 * stor-layout.c (finish_bitfield_representative): If nextf has
22021 error_mark_node type, set repr type to error_mark_node too.
22023 2021-06-24 Ilya Leoshkevich <iii@linux.ibm.com>
22025 * config/s390/s390.c (s390_function_profiler): Ignore labelno
22027 * config/s390/s390.h (NO_PROFILE_COUNTERS): Define.
22029 2021-06-24 Richard Biener <rguenther@suse.de>
22031 * tree-vect-slp.c (vect_optimize_slp): Do not propagate
22032 across operations that have different semantics on different
22035 2021-06-24 Jakub Jelinek <jakub@redhat.com>
22037 * tree.h (OMP_CLAUSE_MAP_IN_REDUCTION): Document meaning for OpenMP.
22038 * gimplify.c (gimplify_scan_omp_clauses): For OpenMP map clauses
22039 with OMP_CLAUSE_MAP_IN_REDUCTION flag partially defer gimplification
22040 of non-decl OMP_CLAUSE_DECL. For OMP_CLAUSE_IN_REDUCTION on
22041 OMP_TARGET user outer_ctx instead of ctx for placeholders and
22042 initializer/combiner gimplification.
22043 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_MAP_IN_REDUCTION
22044 on target constructs.
22045 (lower_rec_input_clauses): Likewise.
22046 (lower_omp_target): Likewise.
22047 * omp-expand.c (expand_omp_target): Temporarily ignore nowait clause
22048 on target if in_reduction is present.
22050 2021-06-24 Kewen Lin <linkw@linux.ibm.com>
22052 * tree-predcom.c (class pcom_worker): New class.
22053 (release_chain): Renamed to...
22054 (pcom_worker::release_chain): ...this.
22055 (release_chains): Renamed to...
22056 (pcom_worker::release_chains): ...this.
22057 (aff_combination_dr_offset): Renamed to...
22058 (pcom_worker::aff_combination_dr_offset): ...this.
22059 (determine_offset): Renamed to...
22060 (pcom_worker::determine_offset): ...this.
22061 (class comp_ptrs): New class.
22062 (split_data_refs_to_components): Renamed to...
22063 (pcom_worker::split_data_refs_to_components): ...this,
22064 and update with class comp_ptrs.
22065 (suitable_component_p): Renamed to...
22066 (pcom_worker::suitable_component_p): ...this.
22067 (filter_suitable_components): Renamed to...
22068 (pcom_worker::filter_suitable_components): ...this.
22069 (valid_initializer_p): Renamed to...
22070 (pcom_worker::valid_initializer_p): ...this.
22071 (find_looparound_phi): Renamed to...
22072 (pcom_worker::find_looparound_phi): ...this.
22073 (add_looparound_copies): Renamed to...
22074 (pcom_worker::add_looparound_copies): ...this.
22075 (determine_roots_comp): Renamed to...
22076 (pcom_worker::determine_roots_comp): ...this.
22077 (determine_roots): Renamed to...
22078 (pcom_worker::determine_roots): ...this.
22079 (single_nonlooparound_use): Renamed to...
22080 (pcom_worker::single_nonlooparound_use): ...this.
22081 (remove_stmt): Renamed to...
22082 (pcom_worker::remove_stmt): ...this.
22083 (execute_pred_commoning_chain): Renamed to...
22084 (pcom_worker::execute_pred_commoning_chain): ...this.
22085 (execute_pred_commoning): Renamed to...
22086 (pcom_worker::execute_pred_commoning): ...this.
22087 (struct epcc_data): New member worker.
22088 (execute_pred_commoning_cbck): Call execute_pred_commoning
22089 with pcom_worker pointer.
22090 (find_use_stmt): Renamed to...
22091 (pcom_worker::find_use_stmt): ...this.
22092 (find_associative_operation_root): Renamed to...
22093 (pcom_worker::find_associative_operation_root): ...this.
22094 (find_common_use_stmt): Renamed to...
22095 (pcom_worker::find_common_use_stmt): ...this.
22096 (combinable_refs_p): Renamed to...
22097 (pcom_worker::combinable_refs_p): ...this.
22098 (reassociate_to_the_same_stmt): Renamed to...
22099 (pcom_worker::reassociate_to_the_same_stmt): ...this.
22100 (stmt_combining_refs): Renamed to...
22101 (pcom_worker::stmt_combining_refs): ...this.
22102 (combine_chains): Renamed to...
22103 (pcom_worker::combine_chains): ...this.
22104 (try_combine_chains): Renamed to...
22105 (pcom_worker::try_combine_chains): ...this.
22106 (prepare_initializers_chain): Renamed to...
22107 (pcom_worker::prepare_initializers_chain): ...this.
22108 (prepare_initializers): Renamed to...
22109 (pcom_worker::prepare_initializers): ...this.
22110 (prepare_finalizers_chain): Renamed to...
22111 (pcom_worker::prepare_finalizers_chain): ...this.
22112 (prepare_finalizers): Renamed to...
22113 (pcom_worker::prepare_finalizers): ...this.
22114 (tree_predictive_commoning_loop): Renamed to...
22115 (pcom_worker::tree_predictive_commoning_loop): ...this, adjust
22116 some calls and remove some cleanup code.
22117 (tree_predictive_commoning): Adjusted to use pcom_worker instance.
22118 (static variable looparound_phis): Remove.
22119 (static variable name_expansions): Remove.
22121 2021-06-24 Richard Biener <rguenther@suse.de>
22123 * tree-vect-slp.c (slpg_vertex): New struct.
22124 (vect_slp_build_vertices): Adjust.
22125 (vect_optimize_slp): Likewise. Maintain an outgoing permute
22126 and a materialized one.
22128 2021-06-24 Richard Biener <rguenther@suse.de>
22130 PR tree-optimization/101105
22131 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
22132 Only ignore steps when they are equal or scalar order is preserved.
22134 2021-06-24 liuhongt <hongtao.liu@intel.com>
22137 * config/i386/i386-expand.c (ix86_expand_vec_interleave):
22138 Adjust comments for ix86_expand_vecop_qihi2.
22139 (ix86_expand_vecmul_qihi): Renamed to ..
22140 (ix86_expand_vecop_qihi2): Adjust function prototype to
22141 support shift operation, add static to definition.
22142 (ix86_expand_vec_shift_qihi_constant): Add static to definition.
22143 (ix86_expand_vecop_qihi): Call ix86_expand_vecop_qihi2 and
22144 ix86_expand_vec_shift_qihi_constant.
22145 * config/i386/i386-protos.h (ix86_expand_vecmul_qihi): Deleted.
22146 (ix86_expand_vec_shift_qihi_constant): Deleted.
22147 * config/i386/sse.md (VI12_256_512_AVX512VL): New mode
22149 (mulv8qi3): Call ix86_expand_vecop_qihi directly, add
22150 condition TARGET_64BIT.
22151 (mul<mode>3): Ditto.
22152 (<insn><mode>3): Ditto.
22153 (vlshr<mode>3): Extend to support avx512 vlshr.
22154 (v<insn><mode>3): New expander for
22156 (v<insn>v8qi3): Ditto.
22157 (vashrv8hi3<mask_name>): Renamed to ..
22158 (vashr<mode>3): And extend to support V16QImode for avx512.
22159 (vashrv16qi3): Deleted.
22160 (vashrv2di3<mask_name>): Extend expander to support avx512
22163 2021-06-23 Dimitar Dimitrov <dimitar@dinux.eu>
22165 * doc/lto.texi (Design Overview): Update that slim objects are
22168 2021-06-23 Aaron Sawdey <acsawdey@linux.ibm.com>
22170 * config/rs6000/rs6000-cpus.def: Take OPTION_MASK_PCREL_OPT out
22171 of OTHER_POWER10_MASKS so it will not be enabled by default.
22173 2021-06-23 Richard Biener <rguenther@suse.de>
22174 Martin Jambor <mjambor@suse.cz>
22176 * tree-inline.c (setup_one_parameter): Set TREE_READONLY of the
22177 param replacement unconditionally. Adjust comment.
22179 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
22181 * Makefile.in (OBJS): Add gimple-range-fold.o
22182 * gimple-range-fold.cc: New.
22183 * gimple-range-fold.h: New.
22184 * gimple-range-gori.cc (gimple_range_calc_op1): Move to here.
22185 (gimple_range_calc_op2): Ditto.
22186 * gimple-range-gori.h: Move prototypes to here.
22187 * gimple-range.cc: Adjust include files.
22188 (fur_source:fur_source): Relocate to gimple-range-fold.cc.
22189 (fur_source::get_operand): Ditto.
22190 (fur_source::get_phi_operand): Ditto.
22191 (fur_source::query_relation): Ditto.
22192 (fur_source::register_relation): Ditto.
22193 (class fur_edge): Ditto.
22194 (fur_edge::fur_edge): Ditto.
22195 (fur_edge::get_operand): Ditto.
22196 (fur_edge::get_phi_operand): Ditto.
22197 (fur_stmt::fur_stmt): Ditto.
22198 (fur_stmt::get_operand): Ditto.
22199 (fur_stmt::get_phi_operand): Ditto.
22200 (fur_stmt::query_relation): Ditto.
22201 (class fur_depend): Relocate to gimple-range-fold.h.
22202 (fur_depend::fur_depend): Relocate to gimple-range-fold.cc.
22203 (fur_depend::register_relation): Ditto.
22204 (fur_depend::register_relation): Ditto.
22205 (class fur_list): Ditto.
22206 (fur_list::fur_list): Ditto.
22207 (fur_list::get_operand): Ditto.
22208 (fur_list::get_phi_operand): Ditto.
22209 (fold_range): Ditto.
22210 (adjust_pointer_diff_expr): Ditto.
22211 (gimple_range_adjustment): Ditto.
22212 (gimple_range_base_of_assignment): Ditto.
22213 (gimple_range_operand1): Ditto.
22214 (gimple_range_operand2): Ditto.
22215 (gimple_range_calc_op1): Relocate to gimple-range-gori.cc.
22216 (gimple_range_calc_op2): Ditto.
22217 (fold_using_range::fold_stmt): Relocate to gimple-range-fold.cc.
22218 (fold_using_range::range_of_range_op): Ditto.
22219 (fold_using_range::range_of_address): Ditto.
22220 (fold_using_range::range_of_phi): Ditto.
22221 (fold_using_range::range_of_call): Ditto.
22222 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
22223 (fold_using_range::range_of_builtin_call): Ditto.
22224 (fold_using_range::range_of_cond_expr): Ditto.
22225 (fold_using_range::range_of_ssa_name_with_loop_info): Ditto.
22226 (fold_using_range::relation_fold_and_or): Ditto.
22227 (fold_using_range::postfold_gcond_edges): Ditto.
22228 * gimple-range.h: Add gimple-range-fold.h to include files. Change
22229 GIMPLE_RANGE_STMT_H to GIMPLE_RANGE_H.
22230 (gimple_range_handler): Relocate to gimple-range-fold.h.
22231 (gimple_range_ssa_p): Ditto.
22232 (range_compatible_p): Ditto.
22233 (class fur_source): Ditto.
22234 (class fur_stmt): Ditto.
22235 (class fold_using_range): Ditto.
22236 (gimple_range_calc_op1): Relocate to gimple-range-gori.h
22237 (gimple_range_calc_op2): Ditto.
22239 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
22241 PR tree-optimization/101148
22242 PR tree-optimization/101014
22243 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
22244 (ranger_cache::~ranger_cache): Adjust.
22245 (ranger_cache::block_range): Check if propagation disallowed.
22246 (ranger_cache::propagate_cache): Disallow propagation if new value
22247 can't be stored properly.
22248 * gimple-range-cache.h (ranger_cache::m_propfail): New member.
22250 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
22252 * gimple-range-cache.cc (class ssa_block_ranges): Adjust prototype.
22253 (sbr_vector::set_bb_range): Return true.
22254 (class sbr_sparse_bitmap): Adjust.
22255 (sbr_sparse_bitmap::set_bb_range): Return value.
22256 (block_range_cache::set_bb_range): Return value.
22257 (ranger_cache::propagate_cache): Use return value to print msg.
22258 * gimple-range-cache.h (class block_range_cache): Adjust.
22260 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
22262 * gimple-range.cc (dump_bb): Use range_on_edge from the cache.
22264 2021-06-23 Jeff Law <jeffreyalaw@gmail.com>
22266 * config/h8300/logical.md (<code><mode>3<ccnz>): Use <cczn>
22267 so this pattern can be used for test/compare removal. Pass
22268 current insn to compute_logical_op_length and output_logical_op.
22269 * config/h8300/h8300.c (compute_logical_op_cc): Remove.
22270 (h8300_and_costs): Add argument to compute_logical_op_length.
22271 (output_logical_op): Add new argument. Use it to determine if the
22272 condition codes are used and adjust the output accordingly.
22273 (compute_logical_op_length): Add new argument and update length
22274 computations when condition codes are used.
22275 * config/h8300/h8300-protos.h (compute_logical_op_length): Update
22277 (output_logical_op): Likewise.
22279 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
22282 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
22283 Handle 64bit modes for TARGET_XOP. Use indirect gen_* functions.
22284 * config/i386/mmx.md (mmx_ppermv64): New insn pattern.
22285 * config/i386/i386.md (unspec): Move UNSPEC_XOP_PERMUTE from ...
22286 * config/i386/sse.md (unspec): ... here.
22288 2021-06-23 Martin Liska <mliska@suse.cz>
22291 * optc-save-gen.awk: Put back arm_fp16_format to
22294 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
22297 * config/i386/i386.md (bsr_rex64): Add zero-flag setting RTX.
22300 (clz<mode>2): Update RTX pattern for additions.
22302 2021-06-23 Jakub Jelinek <jakub@redhat.com>
22304 PR middle-end/101167
22305 * omp-low.c (lower_omp_regimplify_p): Regimplify also PARM_DECLs
22306 and RESULT_DECLs that have DECL_HAS_VALUE_EXPR_P set.
22308 2021-06-22 Sergei Trofimovich <siarheit@google.com>
22310 * doc/rtl.texi: drop unbalanced parenthesis.
22312 2021-06-22 Richard Biener <rguenther@suse.de>
22314 PR middle-end/101156
22315 * gimplify.c (gimplify_expr): Remove premature incorrect
22318 2021-06-22 Jakub Jelinek <jakub@redhat.com>
22320 PR tree-optimization/101159
22321 * tree-vect-patterns.c (vect_recog_popcount_pattern): Fix some
22324 2021-06-22 Jakub Jelinek <jakub@redhat.com>
22326 PR middle-end/101160
22327 * function.c (assign_parms): For decl_result with TYPE_EMPTY_P type
22328 clear crtl->return_rtx instead of keeping it referencing a pseudo.
22330 2021-06-22 Jakub Jelinek <jakub@redhat.com>
22331 Andrew Pinski <apinski@marvell.com>
22333 PR tree-optimization/101162
22334 * fold-const.c (range_check_type): Handle OFFSET_TYPE like pointer
22337 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
22339 * range-op.cc (range_relational_tests): New.
22340 (range_op_tests): Call range_relational_tests.
22342 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
22344 * range-op.cc (operator_cast::lhs_op1_relation): New.
22345 (operator_identity::lhs_op1_relation): Mew.
22347 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
22349 * range-op.cc (operator_minus::op1_op2_relation_effect): New.
22351 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
22353 * range-op.cc (operator_plus::lhs_op1_relation): New.
22354 (operator_plus::lhs_op2_relation): New.
22356 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
22358 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a
22359 relation_oracle if dominators exist.
22360 (ranger_cache::~ranger_cache): Dispose of oracle.
22361 (ranger_cache::dump_bb): Dump oracle.
22362 * gimple-range.cc (fur_source::fur_source): New.
22363 (fur_source::get_operand): Use mmeber query.
22364 (fur_source::get_phi_operand): Use member_query.
22365 (fur_source::query_relation): New.
22366 (fur_source::register_dependency): Delete.
22367 (fur_source::register_relation): New.
22368 (fur_edge::fur_edge): Adjust.
22369 (fur_edge::get_phi_operand): Fix comment.
22370 (fur_edge::query): Delete.
22371 (fur_stmt::fur_stmt): Adjust.
22372 (fur_stmt::query): Delete.
22373 (fur_depend::fur_depend): Adjust.
22374 (fur_depend::register_relation): New.
22375 (fur_depend::register_relation): New.
22376 (fur_list::fur_list): Adjust.
22377 (fur_list::get_operand): Use member query.
22378 (fold_using_range::range_of_range_op): Process and query relations.
22379 (fold_using_range::range_of_address): Adjust dependency call.
22380 (fold_using_range::range_of_phi): Ditto.
22381 (gimple_ranger::gimple_ranger): New. Use ranger_ache oracle.
22382 (fold_using_range::relation_fold_and_or): New.
22383 (fold_using_range::postfold_gcond_edges): New.
22384 * gimple-range.h (class gimple_ranger): Adjust.
22385 (class fur_source): Adjust members.
22386 (class fur_stmt): Ditto.
22387 (class fold_using_range): Ditto.
22389 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
22391 * range-op.cc (range_operator::wi_fold): Apply relation effect.
22392 (range_operator::fold_range): Adjust and apply relation effect.
22393 (*::fold_range): Add relation parameters.
22394 (*::op1_range): Ditto.
22395 (*::op2_range): Ditto.
22396 (range_operator::lhs_op1_relation): New.
22397 (range_operator::lhs_op2_relation): New.
22398 (range_operator::op1_op2_relation): New.
22399 (range_operator::op1_op2_relation_effect): New.
22400 (relop_early_resolve): New.
22401 (operator_equal::op1_op2_relation): New.
22402 (operator_equal::fold_range): Call relop_early_resolve.
22403 (operator_not_equal::op1_op2_relation): New.
22404 (operator_not_equal::fold_range): Call relop_early_resolve.
22405 (operator_lt::op1_op2_relation): New.
22406 (operator_lt::fold_range): Call relop_early_resolve.
22407 (operator_le::op1_op2_relation): New.
22408 (operator_le::fold_range): Call relop_early_resolve.
22409 (operator_gt::op1_op2_relation): New.
22410 (operator_gt::fold_range): Call relop_early_resolve.
22411 (operator_ge::op1_op2_relation): New.
22412 (operator_ge::fold_range): Call relop_early_resolve.
22413 * range-op.h (class range_operator): Adjust parameters and methods.
22415 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
22417 * Makefile.in (OBJS): Add value-relation.o.
22418 * gimple-range.h: Adjust include files.
22419 * tree-data-ref.c: Adjust include file order.
22420 * value-query.cc (range_query::get_value_range): Default to no oracle.
22421 (range_query::query_relation): New.
22422 (range_query::query_relation): New.
22423 * value-query.h (class range_query): Adjust.
22424 * value-relation.cc: New.
22425 * value-relation.h: New.
22427 2021-06-22 Richard Biener <rguenther@suse.de>
22429 PR tree-optimization/101151
22430 * tree-ssa-sink.c (statement_sink_location): Expand irreducible
22433 2021-06-22 Jojo R <rjiejie@linux.alibaba.com>
22435 * config/riscv/riscv.c (thead_c906_tune_info): New.
22436 (riscv_tune_info_table): Use new tune.
22438 2021-06-22 Richard Biener <rguenther@suse.de>
22440 PR tree-optimization/101158
22441 * tree-vect-slp.c (vect_build_slp_tree_1): Move same operand
22442 checking after checking for matching operation.
22444 2021-06-22 Richard Biener <rguenther@suse.de>
22446 PR tree-optimization/101159
22447 * tree-vect-patterns.c (vect_recog_popcount_pattern): Add
22448 missing NULL vectype check.
22450 2021-06-22 Richard Biener <rguenther@suse.de>
22452 PR tree-optimization/101154
22453 * tree-vect-slp.c (vect_build_slp_tree_2): Fix out-of-bound access.
22455 2021-06-22 Jakub Jelinek <jakub@redhat.com>
22458 * config/i386/i386-protos.h (ix86_last_zero_store_uid): Declare.
22459 * config/i386/i386-expand.c (ix86_last_zero_store_uid): New variable.
22460 * config/i386/i386.c (ix86_expand_prologue): Clear it.
22461 * config/i386/i386.md (peephole2s for 1/2/4 stores of const0_rtx):
22462 Remove "" from match_operand. Emit new insns using emit_move_insn and
22463 set ix86_last_zero_store_uid to INSN_UID of the last store.
22464 Add peephole2s for 1/2/4 stores of const0_rtx following previous
22467 2021-06-22 Martin Liska <mliska@suse.cz>
22469 * auto-profile.c (AUTO_PROFILE_VERSION): Bump as string format
22472 2021-06-22 Martin Liska <mliska@suse.cz>
22474 * gcov-io.h: Remove padding entries.
22476 2021-06-22 liuhongt <hongtao.liu@intel.com>
22478 PR tree-optimization/97770
22479 * tree-vect-patterns.c (vect_recog_popcount_pattern):
22481 (vect_recog_func vect_vect_recog_func_ptrs): Add new pattern.
22483 2021-06-22 liuhongt <hongtao.liu@intel.com>
22486 * config/i386/i386-builtin.def (BDESC): Adjust builtin name.
22487 * config/i386/sse.md (<avx512>_expand<mode>_mask): Rename to ..
22488 (expand<mode>_mask): this ..
22489 (*expand<mode>_mask): New pre_reload splitter to transform
22490 v{,p}expand* to vmov* when mask is zero, all ones, or has all
22491 ones in it's lower part, otherwise still generate
22494 2021-06-22 liuhongt <hongtao.liu@intel.com>
22497 * config/i386/i386-expand.c
22498 (ix86_expand_special_args_builtin): Keep constm1_operand only
22499 if it satisfies insn's operand predicate.
22501 2021-06-21 Jason Merrill <jason@redhat.com>
22504 * df-scan.c (df_ref_record): Check that regno < endregno.
22505 * function.c (assign_parms, expand_function_end): Do nothing with a
22506 TYPE_EMPTY_P result.
22508 2021-06-21 Richard Biener <rguenther@suse.de>
22510 PR tree-optimization/101120
22511 * tree-vect-data-refs.c (bump_vector_ptr): Fold the
22513 * tree-vect-slp.c (vect_transform_slp_perm_load): Add
22514 DR chain DCE capability.
22515 * tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
22516 * tree-vect-stmts.c (vectorizable_load): Remove unused
22517 loads in the DR chain for SLP.
22519 2021-06-21 Jakub Jelinek <jakub@redhat.com>
22521 PR inline-asm/100785
22522 * gimplify.c (gimplify_asm_expr): Don't diagnose errors if
22523 output or input operands were already error_mark_node.
22524 * cfgexpand.c (expand_asm_stmt): If errors are emitted,
22525 remove all inputs, outputs and clobbers from the asm and
22526 set template to "".
22528 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
22530 * config/arm/arm_neon.h (vceq_s8): Replace builtin with __a == __b.
22531 (vceq_s16): Likewise.
22532 (vceq_s32): Likewise.
22533 (vceq_u8): Likewise.
22534 (vceq_u16): Likewise.
22535 (vceq_u32): Likewise.
22536 (vceq_p8): Likewise.
22537 (vceqq_s8): Likewise.
22538 (vceqq_s16): Likewise.
22539 (vceqq_s32): Likewise.
22540 (vceqq_u8): Likewise.
22541 (vceqq_u16): Likewise.
22542 (vceqq_u32): Likewise.
22543 (vceqq_p8): Likewise.
22544 (vceq_f32): Gate __a == __b on __FAST_MATH__.
22545 (vceqq_f32): Likewise.
22546 (vceq_f16): Likewise.
22547 (vceqq_f16): Likewise.
22549 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
22552 * config/arm/iterators.md (NEON_VACMP): Remove.
22553 * config/arm/neon.md (neon_vca<cmp_op><mode>): Use GLTE instead of GTGE
22555 (neon_vca<cmp_op><mode>_insn): Likewise.
22556 (neon_vca<cmp_op_unsp><mode>_insn_unspec): Use NEON_VAGLTE instead of
22559 2021-06-21 Richard Biener <rguenther@suse.de>
22561 PR tree-optimization/101121
22562 * tree-vect-slp.c (vect_build_slp_tree_2): To not fail fatally
22563 when we just lack a stmt with the desired op when doing permutation.
22564 (vect_build_slp_tree): When caching a failed SLP build attempt
22565 assert that at least one lane is marked as not matching.
22567 2021-06-21 liuhongt <hongtao.liu@intel.com>
22570 * config/i386/i386.md: (*anddi_1): Disparage slightly the mask
22571 register alternative.
22572 (*and<mode>_1): Ditto.
22574 (*andn<mode>_1): Ditto.
22575 (*<code><mode>_1): Ditto.
22576 (*<code>qi_1): Ditto.
22577 (*one_cmpl<mode>2_1): Ditto.
22578 (*one_cmplsi2_1_zext): Ditto.
22579 (*one_cmplqi2_1): Ditto.
22580 * config/i386/i386.c (x86_order_regs_for_local_alloc): Change
22581 the order of mask registers to be before general registers.
22583 2021-06-21 Roger Sayle <roger@nextmovesoftware.com>
22586 * config/i386/i386.md: New define_peephole2s to shrink writing
22587 1, 2 or 4 consecutive zeros to memory when optimizing for size.
22589 2021-06-18 Jeff Law <jeffreyalaw@gmail.com>
22591 * config/h8300/h8300.c (h8300_select_cc_mode): Handle SYMBOL_REF.
22592 * config/h8300/logical.md (<code><mode>3 logcial expander): Generate
22593 more efficient code when the source can be trivially simplified.
22595 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
22597 * gimple-range-cache.cc (ranger_cache::range_of_def): Calculate
22598 a range if global is not available.
22599 (ranger_cache::entry_range): Fallback to range_of_def.
22600 * gimple-range-cache.h (range_of_def): Adjust prototype.
22602 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
22604 PR tree-optimization/101014
22605 * gimple-range-cache.cc (ranger_cache::ranger_cache): Remove poor
22607 (ranger_cache::~ranger_cache): Ditto.
22608 (ranger_cache::enable_new_values): Delete.
22609 (ranger_cache::push_poor_value): Delete.
22610 (ranger_cache::range_of_def): Remove poor value processing.
22611 (ranger_cache::entry_range): Ditto.
22612 (ranger_cache::fill_block_cache): Ditto.
22613 * gimple-range-cache.h (class ranger_cache): Remove poor value members.
22614 * gimple-range.cc (gimple_ranger::range_of_expr): Remove call.
22615 * gimple-range.h (class gimple_ranger): Adjust.
22617 2021-06-18 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
22620 * common/config/arm/arm-common.c (arm_canon_arch_option_1): New function
22621 derived from arm_canon_arch.
22622 (arm_canon_arch_option): Call it.
22623 (arm_canon_arch_multilib_option): New function.
22624 * config/arm/arm-cpus.in (IGNORE_FOR_MULTILIB): New fgroup.
22625 * config/arm/arm.h (arm_canon_arch_multilib_option): New prototype.
22626 (CANON_ARCH_MULTILIB_SPEC_FUNCTION): New macro.
22627 (MULTILIB_ARCH_CANONICAL_SPECS): New macro.
22628 (DRIVER_SELF_SPECS): Add MULTILIB_ARCH_CANONICAL_SPECS.
22629 * config/arm/arm.opt (mlibarch): New option.
22630 * config/arm/t-rmprofile (MULTILIB_MATCHES): For armv8*-m, replace use
22631 of march on RHS with mlibarch.
22633 2021-06-18 Marcel Vollweiler <marcel@codesourcery.com>
22635 * config.in: Regenerate.
22636 * config/gcn/gcn.c (print_operand_address): Fix for global_load assembler
22638 * configure: Regenerate.
22639 * configure.ac: Fix for global_load assembler functions.
22641 2021-06-18 Richard Biener <rguenther@suse.de>
22643 PR tree-optimization/101112
22644 * tree-vect-slp.c (vect_slp_linearize_chain): Fix condition
22645 to lookup a pattern stmt def.
22647 2021-06-18 Jakub Jelinek <jakub@redhat.com>
22649 PR middle-end/101062
22650 * stor-layout.c (finish_bitfield_layout): Don't add bitfield
22651 representatives in QUAL_UNION_TYPE.
22653 2021-06-18 Andrew Pinski <apinski@marvell.com>
22655 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
22656 Add counting of how many times it is done.
22657 (factor_out_conditional_conversion): Likewise.
22658 (match_simplify_replacement): Likewise.
22659 (value_replacement): Likewise.
22660 (spaceship_replacement): Likewise.
22661 (cond_store_replacement): Likewise.
22662 (cond_if_else_store_replacement_1): Likewise.
22663 (hoist_adjacent_loads): Likewise.
22665 2021-06-18 Andrew Pinski <apinski@marvell.com>
22667 * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
22668 types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and CNONJ_EXPR.
22669 (verify_gimple_assign_binary): Reject point and offset types on
22670 MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
22671 FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
22672 FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.
22674 2021-06-18 Michael Meissner <meissner@linux.ibm.com>
22676 * config/rs6000/rs6000.c (rs6000_emit_minmax): Add support for ISA
22677 3.1 IEEE 128-bit floating point xsmaxcqp/xsmincqp instructions.
22678 * config/rs6000/rs6000.md (s<minmax><mode>3, IEEE128 iterator):
22681 2021-06-17 Aaron Sawdey <acsawdey@linux.ibm.com>
22683 * config/rs6000/genfusion.pl (gen_logical_addsubf): Add
22684 earlyclobber to alts 0/1.
22685 (gen_addadd): Add earlyclobber to alts 0/1.
22686 * config/rs6000/fusion.md: Regenerate file.
22688 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22690 * cfgloopanal.c (get_loop_hot_path): Make path an auto_vec.
22692 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
22694 * gimple-range-cache.cc: Comment cleanups.
22695 * gimple-range-gori.cc: Comment cleanups.
22696 * gimple-range.cc: Comment/spacing cleanups
22697 * value-range.h: Comment cleanups.
22699 2021-06-17 H.J. Lu <hjl.tools@gmail.com>
22702 * calls.c (expand_call): Replace PUSH_ARGS with
22703 targetm.calls.push_argument (0).
22704 (emit_library_call_value_1): Likewise.
22705 * defaults.h (PUSH_ARGS): Removed.
22706 (PUSH_ARGS_REVERSED): Replace PUSH_ARGS with
22707 targetm.calls.push_argument (0).
22708 * expr.c (block_move_libcall_safe_for_call_parm): Likewise.
22709 (emit_push_insn): Pass the number bytes to push to
22710 targetm.calls.push_argument and pass 0 if ARGS_ADDR is 0.
22711 * hooks.c (hook_bool_uint_true): New.
22712 * hooks.h (hook_bool_uint_true): Likewise.
22713 * rtlanal.c (nonzero_bits1): Replace PUSH_ARGS with
22714 targetm.calls.push_argument (0).
22715 * target.def (push_argument): Add a targetm.calls hook.
22716 * targhooks.c (default_push_argument): New.
22717 * targhooks.h (default_push_argument): Likewise.
22718 * config/bpf/bpf.h (PUSH_ARGS): Removed.
22719 * config/cr16/cr16.c (TARGET_PUSH_ARGUMENT): New.
22720 * config/cr16/cr16.h (PUSH_ARGS): Removed.
22721 * config/i386/i386.c (ix86_push_argument): New.
22722 (TARGET_PUSH_ARGUMENT): Likewise.
22723 * config/i386/i386.h (PUSH_ARGS): Removed.
22724 * config/m32c/m32c.c (TARGET_PUSH_ARGUMENT): New.
22725 * config/m32c/m32c.h (PUSH_ARGS): Removed.
22726 * config/nios2/nios2.h (PUSH_ARGS): Likewise.
22727 * config/pru/pru.h (PUSH_ARGS): Likewise.
22728 * doc/tm.texi.in: Remove PUSH_ARGS documentation. Add
22729 TARGET_PUSH_ARGUMENT hook.
22730 * doc/tm.texi: Regenerated.
22732 2021-06-17 Uroš Bizjak <ubizjak@gmail.com>
22735 * config/i386/i386-expand.c (expand_vector_set_var):
22736 Handle V2FS mode remapping. Pass TARGET_MMX_WITH_SSE to
22737 ix86_expand_vector_init_duplicate.
22738 (ix86_expand_vector_init_duplicate): Emit insv_1 for
22739 QImode for !TARGET_PARTIAL_REG_STALL.
22740 * config/i386/predicates.md (vec_setm_mmx_operand): New predicate.
22741 * config/i386/mmx.md (vec_setv2sf): Use vec_setm_mmx_operand
22742 as operand 2 predicate. Call ix86_expand_vector_set_var
22743 for non-constant index operand.
22744 (vec_setv2si): Ditto.
22745 (vec_setv4hi): Ditto.
22746 (vec_setv8qi): ditto.
22748 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
22750 PR tree-optimization/100790
22751 * gimple-range.cc (range_of_builtin_call): Cleanup clz and ctz
22754 2021-06-17 Martin Liska <mliska@suse.cz>
22756 * doc/invoke.texi: Use consistently -O1 instead of -O.
22758 2021-06-17 Martin Liska <mliska@suse.cz>
22760 * gcov-io.h: Update documentation entry about string format.
22762 2021-06-17 Marius Hillenbrand <mhillen@linux.ibm.com>
22765 * config/s390/vecintrin.h (vec_doublee): Fix to use
22766 __builtin_s390_vflls.
22767 (vec_floate): Fix to use __builtin_s390_vflrd.
22769 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22771 * dominance.c (get_dominated_to_depth): Return auto_vec<basic_block>.
22772 * dominance.h (get_dominated_to_depth): Likewise.
22773 (get_all_dominated_blocks): Likewise.
22774 * cfgcleanup.c (delete_unreachable_blocks): Adjust.
22775 * gcse.c (hoist_code): Likewise.
22776 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
22777 * tree-parloops.c (oacc_entry_exit_ok): Likewise.
22778 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Likewise.
22779 * tree-ssa-phiprop.c (pass_phiprop::execute): Likewise.
22781 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22783 * dominance.c (get_dominated_by_region): Return auto_vec<basic_block>.
22784 * dominance.h (get_dominated_by_region): Likewise.
22785 * tree-cfg.c (gimple_duplicate_sese_region): Adjust.
22786 (gimple_duplicate_sese_tail): Likewise.
22787 (move_sese_region_to_fn): Likewise.
22789 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22791 * dominance.c (get_dominated_by): Return auto_vec<basic_block>.
22792 * dominance.h (get_dominated_by): Likewise.
22793 * auto-profile.c (afdo_find_equiv_class): Adjust.
22794 * cfgloopmanip.c (duplicate_loop_to_header_edge): Likewise.
22795 * loop-unroll.c (unroll_loop_runtime_iterations): Likewise.
22796 * tree-cfg.c (test_linear_chain): Likewise.
22797 (test_diamond): Likewise.
22799 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22801 * cfgloop.h (get_loop_hot_path): Return auto_vec<basic_block>.
22802 * cfgloopanal.c (get_loop_hot_path): Likewise.
22803 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
22805 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22807 * cgraph.c (cgraph_node::collect_callers): Return
22808 auto_vec<cgraph_edge *>.
22809 * cgraph.h (cgraph_node::collect_callers): Likewise.
22810 * ipa-cp.c (create_specialized_node): Adjust.
22811 (decide_about_value): Likewise.
22812 (decide_whether_version_node): Likewise.
22813 * ipa-sra.c (process_isra_node_results): Likewise.
22815 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
22817 * vec.h (vl_ptr>::using_auto_storage): Handle null m_vec.
22818 (auto_vec<T, 0>::auto_vec): Define move constructor, and delete copy
22820 (auto_vec<T, 0>::operator=): Define move assignment and delete copy
22823 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
22825 * gimple-range.cc (debug_seed_ranger): New.
22826 (dump_ranger): New.
22827 (debug_ranger): New.
22829 2021-06-17 Richard Biener <rguenther@suse.de>
22831 PR tree-optimization/54400
22832 * tree-vectorizer.h (enum slp_instance_kind): Add
22833 slp_inst_kind_bb_reduc.
22834 (reduction_fn_for_scalar_code): Declare.
22835 * tree-vect-data-refs.c (vect_slp_analyze_instance_dependence):
22836 Check SLP_INSTANCE_KIND instead of looking at the
22838 (vect_slp_analyze_instance_alignment): Likewise.
22839 * tree-vect-loop.c (reduction_fn_for_scalar_code): Export.
22840 * tree-vect-slp.c (vect_slp_linearize_chain): Split out
22841 chain linearization from vect_build_slp_tree_2 and generalize
22842 for the use of BB reduction vectorization.
22843 (vect_build_slp_tree_2): Adjust accordingly.
22844 (vect_optimize_slp): Elide permutes at the root of BB reduction
22846 (vectorizable_bb_reduc_epilogue): New function.
22847 (vect_slp_prune_covered_roots): Likewise.
22848 (vect_slp_analyze_operations): Use them.
22849 (vect_slp_check_for_constructors): Recognize associatable
22850 chains for BB reduction vectorization.
22851 (vectorize_slp_instance_root_stmt): Generate code for the
22852 BB reduction epilogue.
22854 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
22856 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Check with
22858 (gori_compute::may_recompute_p): New.
22859 (gori_compute::outgoing_edge_range_p): Perform recomputations.
22860 * gimple-range-gori.h (class gori_compute): Add prototype.
22862 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
22864 * gimple-range-cache.cc (ranger_cache::range_on_edge): Always return
22865 true when a range can be calculated.
22866 * gimple-range.cc (gimple_ranger::dump_bb): Check has_edge_range_p.
22868 2021-06-16 Martin Sebor <msebor@redhat.com>
22870 * doc/invoke.texi (-Wmismatched-dealloc, -Wmismatched-new-delete):
22871 Correct documented defaults.
22873 2021-06-16 Andrew MacLeod <amacleod@redhat.com>
22875 * gimple-range-cache.cc (ranger_cache::ranger_cache): Initialize
22876 m_new_value_p directly.
22878 2021-06-16 Uroš Bizjak <ubizjak@gmail.com>
22881 * config/i386/i386-expand.c (expand_vec_perm_2perm_pblendv):
22882 Handle 64bit modes for TARGET_SSE4_1.
22883 (expand_vec_perm_pshufb2): Handle 64bit modes for TARGET_SSSE3.
22884 (expand_vec_perm_even_odd_pack): Handle V4HI mode.
22885 (expand_vec_perm_even_odd_1) <case E_V4HImode>: Expand via
22886 expand_vec_perm_pshufb2 for TARGET_SSSE3 and via
22887 expand_vec_perm_even_odd_pack for TARGET_SSE4_1.
22888 * config/i386/mmx.md (mmx_packusdw): New insn pattern.
22890 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
22892 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn<mode>):
22893 Change to an expander that emits the correct instruction
22894 depending on endianness.
22895 (aarch64_<sur><addsub>hn<mode>_insn_le): Define.
22896 (aarch64_<sur><addsub>hn<mode>_insn_be): Define.
22898 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
22900 * config/aarch64/aarch64-simd-builtins.def: Split generator
22901 for aarch64_<su>qmovn builtins into scalar and vector
22903 * config/aarch64/aarch64-simd.md (aarch64_<su>qmovn<mode>_insn_le):
22905 (aarch64_<su>qmovn<mode>_insn_be): Define.
22906 (aarch64_<su>qmovn<mode>): Split into scalar and vector
22907 variants. Change vector variant to an expander that emits the
22908 correct instruction depending on endianness.
22910 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
22912 * config/aarch64/aarch64-simd-builtins.def: Split generator
22913 for aarch64_sqmovun builtins into scalar and vector variants.
22914 * config/aarch64/aarch64-simd.md (aarch64_sqmovun<mode>):
22915 Split into scalar and vector variants. Change vector variant
22916 to an expander that emits the correct instruction depending
22918 (aarch64_sqmovun<mode>_insn_le): Define.
22919 (aarch64_sqmovun<mode>_insn_be): Define.
22921 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
22923 * config/aarch64/aarch64-simd.md (aarch64_xtn<mode>_insn_le):
22924 Define - modeling zero-high-half semantics.
22925 (aarch64_xtn<mode>): Change to an expander that emits the
22926 appropriate instruction depending on endianness.
22927 (aarch64_xtn<mode>_insn_be): Define - modeling zero-high-half
22929 (aarch64_xtn2<mode>_le): Rename to...
22930 (aarch64_xtn2<mode>_insn_le): This.
22931 (aarch64_xtn2<mode>_be): Rename to...
22932 (aarch64_xtn2<mode>_insn_be): This.
22933 (vec_pack_trunc_<mode>): Emit truncation instruction instead
22935 * config/aarch64/iterators.md (Vnarrowd): Add Vnarrowd mode
22936 attribute iterator.
22938 2021-06-16 Martin Jambor <mjambor@suse.cz>
22940 PR tree-optimization/100453
22941 * tree-sra.c (create_access): Disqualify any const candidates
22942 which are written to.
22943 (sra_modify_expr): Do not store sub-replacements back to a const base.
22944 (handle_unscalarized_data_in_subtree): Likewise.
22945 (sra_modify_assign): Likewise. Earlier, use TREE_READONLy test
22946 instead of constant_decl_p.
22948 2021-06-16 Jakub Jelinek <jakub@redhat.com>
22950 PR middle-end/101062
22951 * stor-layout.c (finish_bitfield_representative): For fields in unions
22952 assume nextf is always NULL.
22953 (finish_bitfield_layout): Compute bit field representatives also in
22954 unions, but handle it as if each bitfield was the only field in the
22957 2021-06-16 Richard Biener <rguenther@suse.de>
22959 PR tree-optimization/101088
22960 * tree-ssa-loop-im.c (sm_seq_valid_bb): Only look for
22961 supported refs on edges. Do not assert same ref but
22962 different kind stores are unsuported but mark them so.
22963 (hoist_memory_references): Only look for supported refs
22966 2021-06-16 Roger Sayle <roger@nextmovesoftware.com>
22968 PR rtl-optimization/46235
22969 * config/i386/i386.md: New define_split for bt followed by cmov.
22970 (*bt<mode>_setcqi): New define_insn_and_split for bt followed by setc.
22971 (*bt<mode>_setncqi): New define_insn_and_split for bt then setnc.
22972 (*bt<mode>_setnc<mode>): New define_insn_and_split for bt followed
22973 by setnc with zero extension.
22975 2021-06-16 Richard Biener <rguenther@suse.de>
22977 PR tree-optimization/101083
22978 * tree-vect-slp.c (vect_slp_build_two_operator_nodes): Get
22979 vectype as argument.
22980 (vect_build_slp_tree_2): Adjust.
22982 2021-06-15 Martin Sebor <msebor@redhat.com>
22984 PR middle-end/100876
22985 * builtins.c: (gimple_call_return_array): Account for size_t
22986 mangling as either unsigned int or unsigned long
22988 2021-06-15 Jeff Law <jeffreyalaw@gmail.com>
22990 * compare-elim.c (try_eliminate_compare): Run DCE to clean things
22991 up before eliminating comparisons.
22993 2021-06-15 Aldy Hernandez <aldyh@redhat.com>
22995 * range-op.cc (operator_bitwise_or::wi_fold): Make sure
22996 nonzero|X is nonzero.
22997 (range_op_bitwise_and_tests): Add tests for above.
22999 2021-06-15 Carl Love <cel@us.ibm.com>
23002 * config/rs6000/rs6000-builtin.def (VCMPEQUT): Fix the ICODE for the
23004 (VRLQ, VSLQ, VSRQ, VSRAQ): Remove unused BU_P10_OVERLOAD_2
23007 2021-06-15 Tobias Burnus <tobias@codesourcery.com>
23010 * gimplify.c (enum gimplify_defaultmap_kind): Add GDMK_SCALAR_TARGET.
23011 (struct gimplify_omp_ctx): Extend defaultmap array by one.
23012 (new_omp_context): Init defaultmap[GDMK_SCALAR_TARGET].
23013 (omp_notice_variable): Update type classification for Fortran.
23014 (gimplify_scan_omp_clauses): Update calls for new argument; handle
23015 GDMK_SCALAR_TARGET; for Fortran, GDMK_POINTER avoid GOVD_MAP_0LEN_ARRAY.
23016 * langhooks-def.h (lhd_omp_scalar_p): Add 'ptr_ok' argument.
23017 * langhooks.c (lhd_omp_scalar_p): Likewise.
23018 (LANG_HOOKS_OMP_ALLOCATABLE_P, LANG_HOOKS_OMP_SCALAR_TARGET_P): New.
23019 (LANG_HOOKS_DECLS): Add them.
23020 * langhooks.h (struct lang_hooks_for_decls): Add new hooks, update
23021 omp_scalar_p pointer type to include the new bool argument.
23023 2021-06-15 David Malcolm <dmalcolm@redhat.com>
23025 * doc/analyzer.texi
23026 (Special Functions for Debugging the Analyzer): Add
23027 __analyzer_dump_capacity.
23029 2021-06-15 Jakub Jelinek <jakub@redhat.com>
23032 * expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
23033 case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.
23035 2021-06-15 Richard Biener <rguenther@suse.de>
23037 * cfgloopanal.c (mark_irreducible_loops): Use a dominance
23038 check to identify loop latches.
23039 * cfgloop.c (verify_loop_structure): Likewise.
23040 * loop-init.c (apply_loop_flags): Allow marked irreducible
23041 regions even with multiple latches.
23042 * predict.c (rebuild_frequencies): Simplify.
23044 2021-06-15 Richard Biener <rguenther@suse.de>
23046 * tree-ssa-threadupdate.c
23047 (jump_thread_path_registry::mark_threaded_blocks): Assert we
23048 have marked irreducible regions.
23050 2021-06-14 Martin Sebor <msebor@redhat.com>
23053 * builtins.c (gimple_call_return_array): Check for attribute fn spec.
23054 Handle calls to placement new.
23055 (ndecl_dealloc_argno): Avoid placement delete.
23057 2021-06-14 Peter Bergner <bergner@linux.ibm.com>
23060 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
23061 create_tmp_reg_or_ssa_name().
23063 2021-06-14 Andrew MacLeod <amacleod@redhat.com>
23065 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
23066 (ranger_cache::enable_new_values): Set to specified value and
23067 return the old value.
23068 (ranger_cache::disable_new_values): Delete.
23069 (ranger_cache::fill_block_cache): Disable non 1st order derived
23071 * gimple-range-cache.h (ranger_cache): Adjust prototypes.
23072 * gimple-range.cc (gimple_ranger::range_of_expr): Adjust.
23074 2021-06-14 Uroš Bizjak <ubizjak@gmail.com>
23077 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
23078 Return true early when testing with V2HImode.
23079 * config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.
23081 2021-06-14 Christophe Lyon <christophe.lyon@linaro.org>
23083 * config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
23084 (mve_vec_unpack<US>_hi_<mode>): New pattern.
23085 (@mve_vec_pack_trunc_lo_<mode>): New pattern.
23086 (mve_vmovntq_<supf><mode>): Prefix with '@'.
23087 * config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
23089 (vec_unpack<US>_lo_<mode>): Likewise.
23090 (vec_pack_trunc_<mode>): Rename to
23091 neon_quad_vec_pack_trunc_<mode>.
23092 * config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
23094 (vec_unpack<US>_lo_<mode>): New.
23095 (vec_pack_trunc_<mode>): New.
23097 2021-06-14 Richard Biener <rguenther@suse.de>
23099 PR tree-optimization/100934
23100 * tree-ssa-dom.c (pass_dominator::execute): Properly
23101 mark irreducible regions.
23103 2021-06-14 Martin Liska <mliska@suse.cz>
23105 * doc/invoke.texi: Put r{...} on the same line as @item.
23107 2021-06-14 Martin Liska <mliska@suse.cz>
23109 * doc/invoke.texi: Add missing newline.
23111 2021-06-14 Martin Liska <mliska@suse.cz>
23113 * doc/invoke.texi: Remove '+' charasters.
23115 2021-06-14 Claudiu Zissulescu <claziss@synopsys.com>
23117 * config.gcc (arc): Add support for with_cpu option.
23118 * config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.
23120 2021-06-14 Richard Biener <rguenther@suse.de>
23122 PR tree-optimization/101031
23123 * tree-ssa-strlen.c (maybe_invalidate): Increment max_size
23124 instead of size when accounting for a possibly string
23127 2021-06-14 Martin Liska <mliska@suse.cz>
23129 * gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].
23131 2021-06-14 Aldy Hernandez <aldyh@redhat.com>
23133 * value-query.cc (gimple_range_global): Call get_range_global
23134 if called after inlining.
23136 2021-06-13 Uroš Bizjak <ubizjak@gmail.com>
23139 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
23140 Emit constant permutation insn directly from here.
23142 2021-06-13 Trevor Saunders <tbsaunde@tbsaunde.org>
23144 * attribs.c (find_attribute_namespace): Iterate over vec<> with
23146 * auto-profile.c (afdo_find_equiv_class): Likewise.
23147 * gcc.c (do_specs_vec): Likewise.
23148 (do_spec_1): Likewise.
23149 (driver::set_up_specs): Likewise.
23150 * gimple-loop-jam.c (any_access_function_variant_p): Likewise.
23151 * gimple-ssa-store-merging.c (compatible_load_p): Likewise.
23152 (imm_store_chain_info::try_coalesce_bswap): Likewise.
23153 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
23154 (get_location_for_stmts): Likewise.
23155 * graphite-poly.c (print_iteration_domains): Likewise.
23156 (free_poly_bb): Likewise.
23157 (remove_gbbs_in_scop): Likewise.
23158 (free_scop): Likewise.
23159 (dump_gbb_cases): Likewise.
23160 (dump_gbb_conditions): Likewise.
23161 (print_pdrs): Likewise.
23162 (print_scop): Likewise.
23163 * ifcvt.c (cond_move_process_if_block): Likewise.
23164 * lower-subreg.c (decompose_multiword_subregs): Likewise.
23165 * regcprop.c (pass_cprop_hardreg::execute): Likewise.
23166 * sanopt.c (sanitize_rewrite_addressable_params): Likewise.
23167 * sel-sched-dump.c (dump_insn_vector): Likewise.
23168 * store-motion.c (store_ops_ok): Likewise.
23169 (store_killed_in_insn): Likewise.
23170 * timevar.c (timer::named_items::print): Likewise.
23171 * tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
23172 (cleanup_tree_cfg_noloop): Likewise.
23173 * tree-data-ref.c (dump_data_references): Likewise.
23174 (print_dir_vectors): Likewise.
23175 (print_dist_vectors): Likewise.
23176 (dump_data_dependence_relations): Likewise.
23177 (dump_dist_dir_vectors): Likewise.
23178 (dump_ddrs): Likewise.
23179 (create_runtime_alias_checks): Likewise.
23180 (free_subscripts): Likewise.
23181 (save_dist_v): Likewise.
23182 (save_dir_v): Likewise.
23183 (invariant_access_functions): Likewise.
23184 (same_access_functions): Likewise.
23185 (access_functions_are_affine_or_constant_p): Likewise.
23186 (find_data_references_in_stmt): Likewise.
23187 (graphite_find_data_references_in_stmt): Likewise.
23188 (free_dependence_relations): Likewise.
23189 (free_data_refs): Likewise.
23190 * tree-inline.c (copy_debug_stmts): Likewise.
23191 * tree-into-ssa.c (dump_currdefs): Likewise.
23192 (rewrite_update_phi_arguments): Likewise.
23193 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
23194 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
23196 (vect_slp_analyze_node_dependences): Likewise.
23197 (vect_slp_analyze_instance_dependence): Likewise.
23198 (vect_record_base_alignments): Likewise.
23199 (vect_get_peeling_costs_all_drs): Likewise.
23200 (vect_peeling_supportable): Likewise.
23201 * tree-vectorizer.c (vec_info::~vec_info): Likewise.
23202 (vec_info::free_stmt_vec_infos): Likewise.
23204 2021-06-13 Jeff Law <jeffreyalaw@gmail.com>
23206 * config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
23207 (andqi3_1<cczn>): Removed.
23208 (<ors>qi3_1): Do not split for IOR/XOR a single bit.
23209 (H8/SX bit logicals): Split out from other patterns.
23210 * config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
23211 mulqihi3_const_clobber_flags.
23212 (mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly
23214 2021-06-13 H.J. Lu <hjl.tools@gmail.com>
23217 * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
23218 to true if red zone is used.
23219 (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
23220 ix86_red_zone_used.
23221 * config/i386/i386.h (machine_function): Add red_zone_used.
23222 (ix86_red_zone_size): Removed.
23223 (ix86_red_zone_used): New.
23224 * config/i386/i386.md (peephole2 patterns): Replace
23225 ix86_red_zone_size with ix86_red_zone_used.
23227 2021-06-12 Jason Merrill <jason@redhat.com>
23229 * doc/extend.texi (unused variable attribute): Applies to
23230 structure fields as well.
23232 2021-06-12 Eugene Rozenfeld <erozen@microsoft.com>
23234 * auto-profile.c (read_profile): fix a typo in an error string
23236 2021-06-11 Thomas Schwinge <thomas@codesourcery.com>
23238 * tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
23240 * tree-pretty-print.c (dump_omp_clauses): Update.
23241 (dump_generic_node) <OMP_CLAUSE>: Use it.
23243 2021-06-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
23246 * config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
23247 int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
23248 the polymorphic variants matching code.
23249 (__arm_vld1q_z): Likewise.
23250 (__arm_vld2q): Likewise.
23251 (__arm_vld4q): Likewise.
23252 (__arm_vldrbq_gather_offset): Likewise.
23253 (__arm_vldrbq_gather_offset_z): Likewise.
23255 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
23257 PR tree-optimization/96392
23258 * fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.
23260 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
23262 PR tree-optimization/96392
23263 * fold-const.c (fold_real_zero_addition_p): Take both arguments
23264 of the addition or subtraction, not just the zero. Use this
23265 other argument in tests for signaling NaNs and signed zeros.
23266 (tree_expr_maybe_real_minus_zero_p): New predicate.
23267 * fold-const.h (fold_real_zero_addition_p): Update prototype.
23268 (tree_expr_maybe_real_minus_zero_p): New function prototype.
23269 * match.pd: Update calls to fold_real_zero_addition_p.
23270 Replace HONOR_NANS with tree_expr_maybe_nan_p.
23271 Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
23272 Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
23273 * tree-ssa-reassoc.c (eliminate_using_constants): Update
23274 call to fold_real_zero_addition_p.
23276 2021-06-11 Richard Biener <rguenther@suse.de>
23278 PR tree-optimization/101025
23279 * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
23280 all refs that require dependence checking.
23282 2021-06-11 Richard Biener <rguenther@suse.de>
23284 PR tree-optimization/101028
23285 * tree-vect-slp.c (vect_build_slp_tree_2): When SLP
23286 reassoc discovery fails fatally, mark appropriate lanes
23289 2021-06-11 Richard Biener <rguenther@suse.de>
23291 PR tree-optimization/101026
23292 * tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
23293 have a representative for the associated chain nodes.
23295 2021-06-11 Jakub Jelinek <jakub@redhat.com>
23297 PR rtl-optimization/101008
23298 * simplify-rtx.c (relational_result): New function.
23299 (simplify_logical_relational_operation,
23300 simplify_relational_operation): Use it.
23302 2021-06-11 Jakub Jelinek <jakub@redhat.com>
23305 * config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.
23307 2021-06-11 Uroš Bizjak <ubizjak@gmail.com>
23310 * config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
23311 false if the permutation can be implemented with constant
23312 permutation instruction in wider mode.
23313 (canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
23314 Handle V8QImode and V4HImode.
23316 2021-06-11 Martin Liska <mliska@suse.cz>
23318 PR gcov-profile/100788
23319 * common.opt: Add new option.
23320 * coverage.c (coverage_begin_function): Emit warning instead on
23321 the internal compiler error.
23322 * doc/invoke.texi: Document the option.
23323 * toplev.c (process_options): Enable it by default.
23325 2021-06-11 Richard Biener <rguenther@suse.de>
23327 PR middle-end/101009
23328 * tree-data-ref.c (build_classic_dist_vector_1): Make sure
23329 to set *init_b to true when we encounter a constant equal
23331 (compute_affine_dependence): Also dump the actual DR_REF.
23333 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
23335 PR tree-optimization/100984
23336 * gimple-ssa-evrp.c (ssa_equiv_stack): Use auto_vec for
23337 replacements table.
23338 (ssa_equiv_stack::~ssa_equiv_stack): Remove.
23340 2021-06-11 Kewen Lin <linkw@linux.ibm.com>
23342 * config/rs6000/rs6000.md
23343 (floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
23344 define_insn_and_split.
23346 2021-06-11 Richard Biener <rguenther@suse.de>
23348 * tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
23349 to sort operands of the associative chain.
23351 2021-06-11 Richard Biener <rguenther@suse.de>
23353 * system.h (gcc_stablesort_r): Declare.
23354 * sort.cc (gcc_sort_r): Support stable sort.
23355 (gcc_stablesort_r): Define.
23356 * vec.h (vec<>::stablesort): Add.
23358 2021-06-10 Uroš Bizjak <ubizjak@gmail.com>
23361 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
23362 Handle V2SF mode. Emit SHUFPS to fixup unpack-high for V2SF mode.
23363 (expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
23364 (expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
23365 (expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
23366 (expand_vec_perm_interleave2): Handle 64bit modes.
23367 (expand_vec_perm_even_odd_pack): Handle V8QI mode.
23368 (expand_vec_perm_even_odd_1): Ditto.
23369 (ix86_vectorize_vec_perm_const): Ditto.
23370 * config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
23371 * config/i386/sse.md: ... here.
23372 * config/i386/mmx.md (*vec_interleave_lowv2sf):
23373 New insn_and_split pattern.
23374 (*vec_interleave_highv2sf): Ditto.
23375 (mmx_pshufbv8qi3): New insn pattern.
23376 (*mmx_pblendw): Ditto.
23378 2021-06-10 Peter Bergner <bergner@linux.ibm.com>
23380 * config/rs6000/rs6000-builtin.def (build_pair): New built-in.
23381 (build_acc): Likewise.
23382 * config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
23383 source operands in little-endian mode.
23384 (rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
23385 (mma_init_builtins): Likewise.
23386 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
23387 ordering for the MMA assemble and build source operands.
23388 * doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
23390 (__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
23393 2021-06-10 Jeff Law <jeffreyalaw@gmail.com>
23395 * config/h8300/h8300.c (select_cc_mode): Handle MEM. Use
23397 * config/h8300/extensions.md: Replace _clobber_flags patterns
23400 2021-06-10 Robin Dapp <rdapp@linux.ibm.com>
23402 * config/s390/vector.md (vcond_mask_<mode><mode>): Change to
23403 (vcond_mask_<mode><tointvec>): this.
23405 2021-06-10 Andrew Stubbs <ams@codesourcery.com>
23406 Thomas Schwinge <thomas@codesourcery.com>
23408 * omp-builtins.def (BUILT_IN_GOACC_ENTER_EXIT_DATA): Split into...
23409 (BUILT_IN_GOACC_ENTER_DATA, BUILT_IN_GOACC_EXIT_DATA): ... these.
23410 * gimple.h (enum gf_mask): Split
23411 'GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA' into
23412 'GF_OMP_TARGET_KIND_OACC_ENTER_DATA' and
23413 'GF_OMP_TARGET_KIND_OACC_EXIT_DATA'.
23414 (is_gimple_omp_oacc): Update.
23415 * gimple-pretty-print.c (dump_gimple_omp_target): Likewise.
23416 * gimplify.c (gimplify_omp_target_update): Likewise.
23417 * omp-expand.c (expand_omp_target, build_omp_regions_1)
23418 (omp_make_gimple_edges): Likewise.
23419 * omp-low.c (check_omp_nesting_restrictions, lower_omp_target):
23422 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
23424 * value-query.cc (value_query::value_on_edge): Rename name to
23426 (range_query::range_on_edge): Same.
23427 (range_query::value_of_expr): Same.
23428 (range_query::value_on_edge): Same.
23429 * value-query.h (class value_query): Same.
23430 (class range_query): Same.
23432 2021-06-10 Richard Biener <rguenther@suse.de>
23434 PR tree-optimization/101003
23435 * tree-vect-slp.c (vect_build_slp_tree_2): Appropriately
23436 use the pattern stmt defs when linearizing a chain.
23438 2021-06-10 Jakub Jelinek <jakub@redhat.com>
23441 * ifcvt.c (noce_get_alt_condition, noce_try_abs): Use
23442 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
23444 2021-06-10 Clement Chigot <clement.chigot@atos.net>
23446 * config/rs6000/aix71.h (ASM_CPU_SPEC): Add Power10 directive.
23447 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
23449 2021-06-09 Andrew Pinski <apinski@marvell.com>
23451 PR tree-optimization/100925
23452 * match.pd (a ? CST1 : CST2): Limit transformations
23453 that would produce a negative to integeral types only.
23454 Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.
23456 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
23459 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
23461 * doc/tm.texi: Correctly update.
23463 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
23465 * doc/tm.texi: Correctly update.
23467 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
23470 * doc/tm.texi.in (Trampolines): Add a missing blank line.
23472 2021-06-09 Paul Eggert <eggert@cs.ucla.edu>
23475 * doc/invoke.texi (Code Gen Options); Document that -fno-trampolines
23476 and -ftrampolines work only with Ada.
23477 * doc/tm.texi.in (Trampolines): Likewise.
23478 * doc/tm.texi: Regenerated.
23480 2021-06-09 Carl Love <cel@us.ibm.com>
23482 * config/rs6000/altivec.h (vec_signextll, vec_signexti, vec_signextq):
23483 Add define for new builtins.
23484 * config/rs6000/altivec.md(altivec_vreveti2): Add define_expand.
23485 * config/rs6000/rs6000-builtin.def (VSIGNEXTI, VSIGNEXTLL): Add
23486 overloaded builtin definitions.
23487 (VSIGNEXTSB2W, VSIGNEXTSH2W, VSIGNEXTSB2D, VSIGNEXTSH2D,VSIGNEXTSW2D,
23488 VSIGNEXTSD2Q): Add builtin expansions.
23489 (SIGNEXT): Add P10 overload definition.
23490 * config/rs6000/rs6000-call.c (P9V_BUILTIN_VEC_VSIGNEXTI, P9V_BUILTIN_VEC_VSIGNEXTLL,
23491 P10_BUILTIN_VEC_SIGNEXT): Add overloaded argument definitions.
23492 * config/rs6000/vsx.md (vsx_sign_extend_v2di_v1ti): Add define_insn.
23493 (vsignextend_v2di_v1ti, vsignextend_qi_<mode>, vsignextend_hi_<mode>,
23494 vsignextend_si_v2di)[VIlong]: Add define_expand.
23495 Make define_insn vsx_sign_extend_si_v2di visible.
23496 * doc/extend.texi: Add documentation for the vec_signexti,
23497 vec_signextll builtins and vec_signextq.
23499 2021-06-09 Carl Love <cel@us.ibm.com>
23501 * config/rs6000/rs6000.c (__fixkfti, __fixunskfti, __floattikf,
23502 __floatuntikf): Names changed to __fixkfti_sw, __fixunskfti_sw,
23503 __floattikf_sw, __floatuntikf_sw respectively.
23504 * config/rs6000/rs6000.md (floatti<mode>2, floatunsti<mode>2,
23505 fix_trunc<mode>ti2, fixuns_trunc<mode>ti2): Add
23506 define_insn for mode IEEE 128.
23508 2021-06-09 Carl Love <cel@us.ibm.com>
23510 * config/rs6000/altivec.md (altivec_vslq, altivec_vsrq):
23511 Rename to altivec_vslq_<mode>, altivec_vsrq_<mode>, mode VEC_TI.
23512 * config/rs6000/vector.md (VEC_TI): Was named VSX_TI in vsx.md.
23513 (vashlv1ti3): Change to vashl<mode>3, mode VEC_TI.
23514 (vlshrv1ti3): Change to vlshr<mode>3, mode VEC_TI.
23515 * config/rs6000/vsx.md (VSX_TI): Remove define_mode_iterator. Update
23516 uses of VSX_TI to VEC_TI.
23518 2021-06-09 Carl Love <cel@us.ibm.com>
23520 * config/rs6000/dfp.md (floattitd2, fixtdti2): New define_insns.
23522 2021-06-09 Carl Love <cel@us.ibm.com>
23524 * config/rs6000/altivec.h (vec_dive, vec_mod): Add define for new
23526 * config/rs6000/altivec.md (UNSPEC_VMULEUD, UNSPEC_VMULESD,
23527 UNSPEC_VMULOUD, UNSPEC_VMULOSD): New unspecs.
23528 (altivec_eqv1ti, altivec_gtv1ti, altivec_gtuv1ti, altivec_vmuleud,
23529 altivec_vmuloud, altivec_vmulesd, altivec_vmulosd, altivec_vrlq,
23530 altivec_vrlqmi, altivec_vrlqmi_inst, altivec_vrlqnm,
23531 altivec_vrlqnm_inst, altivec_vslq, altivec_vsrq, altivec_vsraq,
23532 altivec_vcmpequt_p, altivec_vcmpgtst_p, altivec_vcmpgtut_p): New
23534 (vec_widen_umult_even_v2di, vec_widen_smult_even_v2di,
23535 vec_widen_umult_odd_v2di, vec_widen_smult_odd_v2di, altivec_vrlqmi,
23536 altivec_vrlqnm): New define_expands.
23537 * config/rs6000/rs6000-builtin.def (VCMPEQUT_P, VCMPGTST_P,
23538 VCMPGTUT_P): Add macro expansions.
23539 (BU_P10V_AV_P): Add builtin predicate definition.
23540 (VCMPGTUT, VCMPGTST, VCMPEQUT, CMPNET, CMPGE_1TI,
23541 CMPGE_U1TI, CMPLE_1TI, CMPLE_U1TI, VNOR_V1TI_UNS, VNOR_V1TI, VCMPNET_P,
23542 VCMPAET_P, VMULEUD, VMULESD, VMULOUD, VMULOSD, VRLQ,
23543 VSLQ, VSRQ, VSRAQ, VRLQNM, DIV_V1TI, UDIV_V1TI, DIVES_V1TI, DIVEU_V1TI,
23544 MODS_V1TI, MODU_V1TI, VRLQMI): New macro expansions.
23545 (VRLQ, VSLQ, VSRQ, VSRAQ, DIVE, MOD): New overload expansions.
23546 * config/rs6000/rs6000-call.c (P10_BUILTIN_VCMPEQUT,
23547 P10V_BUILTIN_CMPGE_1TI, P10V_BUILTIN_CMPGE_U1TI,
23548 P10V_BUILTIN_VCMPGTUT, P10V_BUILTIN_VCMPGTST,
23549 P10V_BUILTIN_CMPLE_1TI, P10V_BUILTIN_VCMPLE_U1TI,
23550 P10V_BUILTIN_DIV_V1TI, P10V_BUILTIN_UDIV_V1TI,
23551 P10V_BUILTIN_VMULESD, P10V_BUILTIN_VMULEUD,
23552 P10V_BUILTIN_VMULOSD, P10V_BUILTIN_VMULOUD,
23553 P10V_BUILTIN_VNOR_V1TI, P10V_BUILTIN_VNOR_V1TI_UNS,
23554 P10V_BUILTIN_VRLQ, P10V_BUILTIN_VRLQMI,
23555 P10V_BUILTIN_VRLQNM, P10V_BUILTIN_VSLQ,
23556 P10V_BUILTIN_VSRQ, P10V_BUILTIN_VSRAQ,
23557 P10V_BUILTIN_VCMPGTUT_P, P10V_BUILTIN_VCMPGTST_P,
23558 P10V_BUILTIN_VCMPEQUT_P, P10V_BUILTIN_VCMPGTUT_P,
23559 P10V_BUILTIN_VCMPGTST_P, P10V_BUILTIN_CMPNET,
23560 P10V_BUILTIN_VCMPNET_P, P10V_BUILTIN_VCMPAET_P,
23561 P10V_BUILTIN_DIVES_V1TI, P10V_BUILTIN_MODS_V1TI,
23562 P10V_BUILTIN_MODU_V1TI):
23563 New overloaded definitions.
23564 (rs6000_gimple_fold_builtin) [P10V_BUILTIN_VCMPEQUT,
23565 P10V_BUILTIN_CMPNET, P10V_BUILTIN_CMPGE_1TI,
23566 P10V_BUILTIN_CMPGE_U1TI, P10V_BUILTIN_VCMPGTUT,
23567 P10V_BUILTIN_VCMPGTST, P10V_BUILTIN_CMPLE_1TI,
23568 P10V_BUILTIN_CMPLE_U1TI]: New case statements.
23569 (rs6000_init_builtins) [bool_V1TI_type_node, int_ftype_int_v1ti_v1ti]:
23571 (altivec_init_builtins): New E_V1TImode case statement.
23572 (builtin_function_type)[P10_BUILTIN_128BIT_VMULEUD,
23573 P10_BUILTIN_128BIT_VMULOUD, P10_BUILTIN_128BIT_DIVEU_V1TI,
23574 P10_BUILTIN_128BIT_MODU_V1TI, P10_BUILTIN_CMPGE_U1TI,
23575 P10_BUILTIN_VCMPGTUT, P10_BUILTIN_VCMPEQUT]: New case statements.
23576 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute) [E_TImode,
23577 E_V1TImode]: New case statements.
23578 * config/rs6000/rs6000.h (rs6000_builtin_type_index): New enum
23579 value RS6000_BTI_bool_V1TI.
23580 * config/rs6000/vector.md (vector_gtv1ti,vector_nltv1ti,
23581 vector_gtuv1ti, vector_nltuv1ti, vector_ngtv1ti, vector_ngtuv1ti,
23582 vector_eq_v1ti_p, vector_ne_v1ti_p, vector_ae_v1ti_p,
23583 vector_gt_v1ti_p, vector_gtu_v1ti_p, vrotlv1ti3, vashlv1ti3,
23584 vlshrv1ti3, vashrv1ti3): New define_expands.
23585 * config/rs6000/vsx.md (UNSPEC_VSX_DIVSQ, UNSPEC_VSX_DIVUQ,
23586 UNSPEC_VSX_DIVESQ, UNSPEC_VSX_DIVEUQ, UNSPEC_VSX_MODSQ,
23587 UNSPEC_VSX_MODUQ): New unspecs.
23588 (mulv2di3, vsx_div_v1ti, vsx_udiv_v1ti, vsx_dives_v1ti,
23589 vsx_diveu_v1ti, vsx_mods_v1ti, vsx_modu_v1ti, xxswapd_v1ti): New
23591 (vcmpnet): New define_expand.
23592 * doc/extend.texi: Add documentation for the new builtins vec_rl,
23593 vec_rlmi, vec_rlnm, vec_sl, vec_sr, vec_sra, vec_mule, vec_mulo,
23594 vec_div, vec_dive, vec_mod, vec_cmpeq, vec_cmpne, vec_cmpgt, vec_cmplt,
23595 vec_cmpge, vec_cmple, vec_all_eq, vec_all_ne, vec_all_gt, vec_all_lt,
23596 vec_all_ge, vec_all_le, vec_any_eq, vec_any_ne, vec_any_gt, vec_any_lt,
23597 vec_any_ge, vec_any_le.
23599 2021-06-09 Carl Love <cel@us.ibm.com>
23601 * config/rs6000/altivec.md (altivec_vrl<VI_char>mi): Fix
23602 bug in argument generation.
23604 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
23606 * config/arm/iterators.md (<supf>): Remove VCLZQ_U, VCLZQ_S.
23608 * config/arm/mve.md (mve_vclzq_<supf><mode>): Add '@' prefix,
23609 remove <supf> iterator.
23610 (mve_vclzq_u<mode>): New.
23611 * config/arm/neon.md (clz<mode>2): Rename to neon_vclz<mode>.
23612 (neon_vclz<mode): Move to ...
23613 * config/arm/unspecs.md (VCLZQ_U, VCLZQ_S): Remove.
23614 * config/arm/vec-common.md: ... here. Add support for MVE.
23616 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
23618 * config/arm/mve.md (mve_vhaddq_<supf><mode>): Prefix with '@'.
23619 (@mve_vrhaddq_<supf><mode): Likewise.
23620 * config/arm/neon.md (neon_v<r>hadd<sup><mode>): Likewise.
23621 * config/arm/vec-common.md (avg<mode>3_floor, uavg<mode>3_floor)
23622 (avg<mode>3_ceil", uavg<mode>3_ceil): New patterns.
23624 2021-06-09 imba-tjd <109224573@qq.com>
23626 * doc/invoke.texi: Fix typo.
23628 2021-06-09 Roger Sayle <roger@nextmovesoftware.com>
23630 PR middle-end/53267
23631 * fold-const-call.c (fold_const_call_sss) [CASE_CFN_FMOD]:
23632 Support evaluation of fmod/fmodf/fmodl at compile-time.
23634 2021-06-09 Richard Biener <rguenther@suse.de>
23636 PR tree-optimization/100981
23637 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
23638 gimple_get_lhs to also handle calls.
23639 * tree-vect-slp-patterns.c (complex_pattern::build): Transfer
23642 2021-06-09 Richard Biener <rguenther@suse.de>
23644 PR tree-optimization/97832
23645 * tree-vectorizer.h (_slp_tree::failed): New.
23646 * tree-vect-slp.c (_slp_tree::_slp_tree): Initialize
23648 (_slp_tree::~_slp_tree): Free failed.
23649 (vect_build_slp_tree): Retain failed nodes and record
23650 matches in them, copying that back out when running
23651 into a cached fail. Dump start and end of discovery.
23652 (dt_sort_cmp): New.
23653 (vect_build_slp_tree_2): Handle associatable chains
23654 together doing more aggressive operand swapping.
23656 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
23659 * config.gcc (gcc_cv_initfini_array): Set to yes for Linux and
23661 * doc/install.texi: Require glibc 2.1 and binutils 2.12 for
23662 Linux and GNU targets.
23664 2021-06-09 Richard Biener <rguenther@suse.de>
23666 * tree-vect-stmts.c (vect_is_simple_use): Always get dt
23669 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
23671 * config/arc/arc.md (loop_end): Change it to
23672 define_insn_and_split.
23674 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
23676 * config/arc/arc.md (maddhisi4): Use VMAC2H instruction.
23677 (machi): New pattern.
23678 (umaddhisi4): Use VMAC2HU instruction.
23679 (umachi): New pattern.
23681 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
23683 * config/arc/arc-protos.h (arc_split_move_p): New prototype.
23684 * config/arc/arc.c (arc_split_move_p): New function.
23685 (arc_split_move): Clean up.
23686 * config/arc/arc.md (movdi_insn): Clean up, use arc_split_move_p.
23687 (movdf_insn): Likewise.
23688 * config/arc/simdext.md (mov<VWH>_insn): Likewise.
23690 2021-06-09 Uroš Bizjak <ubizjak@gmail.com>
23693 * config/i386/i386.c (print_operand_address_as): Rename "no_rip"
23694 argument to "raw". Do not emit segment overrides when "raw" is true.
23696 2021-06-09 Martin Liska <mliska@suse.cz>
23698 * doc/gcov.texi: Create a proper JSON files.
23699 * doc/invoke.texi: Remove dots in order to make it a valid
23702 2021-06-09 Xionghu Luo <luoxhu@linux.ibm.com>
23704 * config/rs6000/rs6000-p8swap.c (pattern_is_rotate64): New.
23705 (insn_is_load_p): Use pattern_is_rotate64.
23706 (insn_is_swap_p): Likewise.
23707 (quad_aligned_load_p): Likewise.
23708 (const_load_sequence_p): Likewise.
23709 (replace_swapped_aligned_load): Likewise.
23710 (recombine_lvx_pattern): Likewise.
23711 (recombine_stvx_pattern): Likewise.
23713 2021-06-09 Andrew MacLeod <amacleod@redhat.com>
23715 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Use a
23716 fur_stmt source record.
23717 * gimple-range.cc (fur_source::get_operand): Generic range query.
23718 (fur_source::get_phi_operand): New.
23719 (fur_source::register_dependency): New.
23720 (fur_source::query): New.
23721 (class fur_edge): New. Edge source for operands.
23722 (fur_edge::fur_edge): New.
23723 (fur_edge::get_operand): New.
23724 (fur_edge::get_phi_operand): New.
23725 (fur_edge::query): New.
23726 (fur_stmt::fur_stmt): New.
23727 (fur_stmt::get_operand): New.
23728 (fur_stmt::get_phi_operand): New.
23729 (fur_stmt::query): New.
23730 (class fur_depend): New. Statement source and process dependencies.
23731 (fur_depend::fur_depend): New.
23732 (fur_depend::register_dependency): New.
23733 (class fur_list): New. List source for operands.
23734 (fur_list::fur_list): New.
23735 (fur_list::get_operand): New.
23736 (fur_list::get_phi_operand): New.
23737 (fold_range): New. Instantiate appropriate fur_source class and fold.
23738 (fold_using_range::range_of_range_op): Use new API.
23739 (fold_using_range::range_of_address): Ditto.
23740 (fold_using_range::range_of_phi): Ditto.
23741 (imple_ranger::fold_range_internal): Use fur_depend class.
23742 (fold_using_range::range_of_ssa_name_with_loop_info): Use new API.
23743 * gimple-range.h (class fur_source): Now a base class.
23744 (class fur_stmt): New.
23745 (fold_range): New prototypes.
23746 (fur_source::fur_source): Delete.
23748 2021-06-08 Andrew Pinski <apinski@marvell.com>
23750 PR tree-optimization/25290
23751 * tree-ssa-phiopt.c (xor_replacement): Delete.
23752 (tree_ssa_phiopt_worker): Delete use of xor_replacement.
23753 (match_simplify_replacement): Allow one cheap preparation
23754 statement that can be moved to before the if.
23756 2021-06-08 Pat Haugen <pthaugen@linux.ibm.com>
23758 * config/rs6000/power10.md (power10-fused-load, power10-fused-store,
23759 power10-fused_alu, power10-fused-vec, power10-fused-branch): New.
23761 2021-06-08 Jeff Law <jeffreyalaw@gmail.com>
23763 * config/h8300/logical.md (andqi3_1): Move BCLR case into define_insn_and_split.
23764 Create length attribute on define_insn_and_split. Only split for cases which we
23766 (andqi3_1<cczn>): Renamed from andqi3_1_clobber_flags. Only handle AND here and
23767 fix length computation.
23768 (b<code><mode>msx): Combine QImode and HImode H8/SX patterns using iterator.
23770 2021-06-08 Richard Biener <rguenther@suse.de>
23772 PR tree-optimization/100923
23773 * tree-ssa-sccvn.c (valueize_refs_1): Take a pointer to
23774 the operand vector to be valueized.
23775 (valueize_refs): Likewise.
23776 (valueize_shared_reference_ops_from_ref): Adjust.
23777 (valueize_shared_reference_ops_from_call): Likewise.
23778 (vn_reference_lookup_3): Likewise.
23779 (vn_reference_lookup_pieces): Likewise. Re-valueize
23780 with honoring availability when we are about to create
23781 the ao_ref and valueized before.
23782 (vn_reference_lookup): Likewise.
23783 (vn_reference_insert_pieces): Adjust.
23785 2021-06-08 Richard Biener <rguenther@suse.de>
23787 * tree-vectorizer.h (_slp_instance::root_stmt): Change to...
23788 (_slp_instance::root_stmts): ... a vector.
23789 (SLP_INSTANCE_ROOT_STMT): Rename to ...
23790 (SLP_INSTANCE_ROOT_STMTS): ... this.
23791 (slp_root::root): Change to...
23792 (slp_root::roots): ... a vector.
23793 (slp_root::slp_root): Adjust.
23794 * tree-vect-slp.c (_slp_instance::location): Adjust.
23795 (vect_free_slp_instance): Release the root stmt vector.
23796 (vect_build_slp_instance): Adjust.
23797 (vect_analyze_slp): Likewise.
23798 (_bb_vec_info::~_bb_vec_info): Likewise.
23799 (vect_slp_analyze_operations): Likewise.
23800 (vect_bb_vectorization_profitable_p): Likewise. Adjust
23801 costs for the root stmt.
23802 (vect_slp_check_for_constructors): Gather all BIT_INSERT_EXPRs
23804 (vect_slp_analyze_bb_1): Simplify by marking all root stmts
23806 (vectorize_slp_instance_root_stmt): Adjust.
23807 (vect_schedule_slp): Likewise.
23809 2021-06-08 Aldy Hernandez <aldyh@redhat.com>
23811 * gimple-ssa-evrp.c (class ssa_equiv_stack): New.
23812 (ssa_equiv_stack::ssa_equiv_stack): New.
23813 (ssa_equiv_stack::~ssa_equiv_stack): New.
23814 (ssa_equiv_stack::enter): New.
23815 (ssa_equiv_stack::leave): New.
23816 (ssa_equiv_stack::push_replacement): New.
23817 (ssa_equiv_stack::get_replacement): New.
23818 (is_pointer_ssa): New.
23819 (class pointer_equiv_analyzer): New.
23820 (pointer_equiv_analyzer::pointer_equiv_analyzer): New.
23821 (pointer_equiv_analyzer::~pointer_equiv_analyzer): New.
23822 (pointer_equiv_analyzer::set_global_equiv): New.
23823 (pointer_equiv_analyzer::set_cond_equiv): New.
23824 (pointer_equiv_analyzer::get_equiv): New.
23825 (pointer_equiv_analyzer::enter): New.
23826 (pointer_equiv_analyzer::leave): New.
23827 (pointer_equiv_analyzer::get_equiv_expr): New.
23828 (pta_valueize): New.
23829 (pointer_equiv_analyzer::visit_stmt): New.
23830 (pointer_equiv_analyzer::visit_edge): New.
23831 (hybrid_folder::value_of_expr): Call PTA.
23832 (hybrid_folder::value_on_edge): Same.
23833 (hybrid_folder::pre_fold_bb): New.
23834 (hybrid_folder::post_fold_bb): New.
23835 (hybrid_folder::pre_fold_stmt): New.
23836 (rvrp_folder::pre_fold_bb): New.
23837 (rvrp_folder::post_fold_bb): New.
23838 (rvrp_folder::pre_fold_stmt): New.
23839 (rvrp_folder::value_of_expr): Call PTA.
23840 (rvrp_folder::value_on_edge): Same.
23842 2021-06-08 Jakub Jelinek <jakub@redhat.com>
23845 * tree-inline.c (copy_tree_body_r): For OMP_CLAUSE_DEPEND don't
23846 check TREE_CODE if OMP_CLAUSE_DECL is NULL.
23848 2021-06-08 Richard Biener <rguenther@suse.de>
23850 PR middle-end/100951
23851 * tree-vect-generic.c (expand_vector_piecewise): Build a
23852 VECTOR_CST if all elements are constant.
23853 (expand_vector_condition): Likewise.
23854 (lower_vec_perm): Likewise.
23855 (expand_vector_conversion): Likewise.
23857 2021-06-08 Martin Liska <mliska@suse.cz>
23859 * doc/invoke.texi: Document new param evrp-sparse-threshold.
23861 2021-06-08 Martin Liska <mliska@suse.cz>
23863 * genautomata.c (create_automata): Fix typo.
23865 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
23867 PR tree-optimization/100794
23868 * tree-predcom.c (tree_predictive_commoning_loop): Add parameter
23869 allow_unroll_p and only allow unrolling when it's true.
23870 (tree_predictive_commoning): Add parameter allow_unroll_p and
23872 (run_tree_predictive_commoning): Likewise.
23873 (pass_predcom::gate): Check flag_tree_loop_vectorize and
23874 global_options_set.x_flag_predictive_commoning.
23875 (pass_predcom::execute): Adjust for allow_unroll_p.
23877 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
23879 * tree-predcom.c (execute_pred_commoning): Remove update_ssa call.
23880 (tree_predictive_commoning_loop): Factor some cleanup stuffs into
23881 lambda function cleanup, remove scev_reset call, and adjust return
23883 (tree_predictive_commoning): Adjust for different changed values,
23884 only set flag TODO_update_ssa_only_virtuals if changed.
23885 (pass_data pass_data_predcom): Remove TODO_update_ssa_only_virtuals
23886 from todo_flags_finish.
23888 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
23890 * gimple-range-cache.cc (class sbr_sparse_bitmap): New.
23891 (sbr_sparse_bitmap::sbr_sparse_bitmap): New.
23892 (sbr_sparse_bitmap::bitmap_set_quad): New.
23893 (sbr_sparse_bitmap::bitmap_get_quad): New.
23894 (sbr_sparse_bitmap::set_bb_range): New.
23895 (sbr_sparse_bitmap::get_bb_range): New.
23896 (sbr_sparse_bitmap::bb_range_p): New.
23897 (block_range_cache::block_range_cache): initialize bitmap obstack.
23898 (block_range_cache::~block_range_cache): Destruct obstack.
23899 (block_range_cache::set_bb_range): Decide when to utilze the
23900 sparse on entry cache.
23901 * gimple-range-cache.h (block_range_cache): Add bitmap obstack.
23902 * params.opt (-param=evrp-sparse-threshold): New.
23904 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
23906 * bitmap.c (bitmap_set_aligned_chunk): New.
23907 (bitmap_get_aligned_chunk): New.
23908 (test_aligned_chunk): New.
23909 (bitmap_c_tests): Call test_aligned_chunk.
23910 * bitmap.h (bitmap_set_aligned_chunk, bitmap_get_aligned_chunk): New.
23912 2021-06-07 Uroš Bizjak <ubizjak@gmail.com>
23915 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
23917 (ix86_expand_vector_init_one_nonzero): Ditto.
23918 (ix86_expand_vector_init_one_var): Ditto.
23919 (ix86_expand_vector_init_general): Ditto.
23920 * config/i386/mmx.md (vec_initv4qiqi): New expander.
23922 2021-06-07 Jeff Law <jeffreyalaw@gmail.com>
23924 * config/h8300/movepush.md: Change most _clobber_flags
23925 patterns to instead use <cczn> subst.
23926 (movsi_cczn): New pattern with usable CC cases split out.
23927 (movsi_h8sx_cczn): Likewise.
23929 2021-06-07 Martin Liska <mliska@suse.cz>
23931 * common/common-target.def: Split long lines and replace them
23933 * target.def: Likewise.
23934 * doc/tm.texi: Re-generated.
23936 2021-06-07 Jakub Jelinek <jakub@redhat.com>
23939 * fold-const.c (fold_read_from_vector): Return NULL if trying to
23940 read from a CONSTRUCTOR with vector type elements.
23942 2021-06-07 Jakub Jelinek <jakub@redhat.com>
23944 PR middle-end/100898
23945 * tree-inline.c (copy_bb): Only use gimple_call_arg_ptr if memcpy
23946 should copy any arguments. Don't call gimple_call_num_args
23947 on id->call_stmt or call_stmt more than once.
23949 2021-06-07 liuhongt <hongtao.liu@intel.com>
23952 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): Refine
23954 (<insn>v4siv4di2): Delete constraints for define_expand.
23956 2021-06-07 liuhongt <hongtao.liu@intel.com>
23959 * config/i386/i386-expand.c (ix86_expand_builtin): Remove
23960 assignment of cfun->machine->has_explicit_vzeroupper.
23961 * config/i386/i386-features.c
23962 (ix86_add_reg_usage_to_vzerouppers): Delete.
23963 (ix86_add_reg_usage_to_vzeroupper): Ditto.
23964 (rest_of_handle_insert_vzeroupper): Remove
23965 ix86_add_reg_usage_to_vzerouppers, add df_analyze at the end
23967 (gate): Remove cfun->machine->has_explicit_vzeroupper.
23968 * config/i386/i386-protos.h (ix86_expand_avx_vzeroupper):
23970 * config/i386/i386.c (ix86_insn_callee_abi): New function.
23971 (ix86_initialize_callee_abi): Ditto.
23972 (ix86_expand_avx_vzeroupper): Ditto.
23973 (ix86_hard_regno_call_part_clobbered): Adjust for vzeroupper
23975 (TARGET_INSN_CALLEE_ABI): Define as ix86_insn_callee_abi.
23976 (ix86_emit_mode_set): Call ix86_expand_avx_vzeroupper
23978 * config/i386/i386.h (struct GTY(()) machine_function): Delete
23979 has_explicit_vzeroupper.
23980 * config/i386/i386.md (enum unspec): New member
23982 (ABI_DEFAULT,ABI_VZEROUPPER,ABI_UNKNOWN): New
23983 define_constants for insn callee abi index.
23984 * config/i386/predicates.md (vzeroupper_pattern): Adjust.
23985 * config/i386/sse.md (UNSPECV_VZEROUPPER): Deleted.
23986 (avx_vzeroupper): Call ix86_expand_avx_vzeroupper.
23987 (*avx_vzeroupper): Rename to ..
23988 (avx_vzeroupper_callee_abi): .. this, and adjust pattern as
23989 call_insn which has a special vzeroupper ABI.
23990 (*avx_vzeroupper_1): Deleted.
23992 2021-06-07 liuhongt <hongtao.liu@intel.com>
23995 * df-scan.c (df_get_call_refs): When call_insn is a fake call,
23996 it won't use stack pointer reg.
23997 * final.c (leaf_function_p): When call_insn is a fake call, it
23998 won't affect caller as a leaf function.
23999 * reg-stack.c (callee_clobbers_any_stack_reg): New.
24000 (subst_stack_regs): When call_insn doesn't clobber any stack
24001 reg, don't clear the arguments.
24002 * rtl.c (shallow_copy_rtx): Don't clear flag used when orig is
24004 * shrink-wrap.c (requires_stack_frame_p): No need for stack
24005 frame for a fake call.
24006 * rtl.h (FAKE_CALL_P): New macro.
24008 2021-06-06 Eric Botcazou <ebotcazou@adacore.com>
24010 * config/sparc/sparc-protos.h (order_regs_for_local_alloc): Rename
24012 (sparc_order_regs_for_local_alloc): ...this.
24013 (sparc_leaf_reg_remap): Declare.
24014 * config/sparc/sparc.h (ADJUST_REG_ALLOC_ORDER): Adjust.
24015 (LEAF_REG_REMAP): Reimplement as call to sparc_leaf_reg_remap.
24016 * config/sparc/sparc.c (leaf_reg_remap): Delete.
24017 (order_regs_for_local_alloc): Rename to...
24018 (sparc_order_regs_for_local_alloc): ...this.
24019 (sparc_leaf_reg_remap): New function.
24020 (sparc_conditional_register_usage): Do not modify leaf_reg_remap.
24022 2021-06-06 David Edelsohn <dje.gcc@gmail.com>
24024 * config/rs6000/rs6000.c (rs6000_xcoff_asm_output_aligned_decl_common):
24025 Use assemble_name to output BSS section name.
24027 2021-06-06 Uroš Bizjak <ubizjak@gmail.com>
24029 * config/i386/constraints.md (Bs):
24030 Remove boolean operators from match_test RTX.
24033 (M): Use "mode" variable instead of GET_MODE (op) in match_test RTX.
24036 2021-06-06 Martin Liska <mliska@suse.cz>
24038 * doc/extend.texi: Add missing @headitem.
24039 * doc/invoke.texi: Likewise.
24040 * doc/objc.texi: Likewise.
24042 2021-06-06 Martin Liska <mliska@suse.cz>
24044 * genhooks.c (emit_findices): Remove unused function.
24045 (emit_documentation): Do not call emit_findices
24046 and do not search for @Fcode directives.
24048 2021-06-06 Martin Liska <mliska@suse.cz>
24050 * doc/invoke.texi: Remove extra character.
24052 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
24054 * config/sh/sh.md (doloop_end_split): Fix empty split condition.
24056 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
24058 * config/sparc/sparc.md (*snedi<W:mode>_zero_vis3,
24059 *neg_snedi<W:mode>_zero_subxc, *plus_snedi<W:mode>_zero,
24060 *plus_plus_snedi<W:mode>_zero, *minus_snedi<W:mode>_zero,
24061 *minus_minus_snedi<W:mode>_zero): Fix empty split condition.
24063 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
24065 * config/or1k/or1k.md (*movdi): Fix empty split condition.
24067 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
24069 * config/mips/mips.md (<anonymous>, bswapsi2, bswapdi2): Fix empty
24072 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
24074 * config/m68k/m68k.md (*zero_extend_inc, *zero_extend_dec,
24075 *zero_extendsidi2): Fix empty split condition.
24077 2021-06-05 Jeff Law <jeffreyalaw@gmail.com>
24079 * config/h8300/addsub.md: Fix split condition in define_insn_and_split
24081 * config/h8300/bitfield.md: Likewise.
24082 * config/h8300/combiner.md: Likewise.
24083 * config/h8300/divmod.md: Likewise.
24084 * config/h8300/extensions.md: Likewise.
24085 * config/h8300/jumpcall.md: Likewise.
24086 * config/h8300/movepush.md: Likewise.
24087 * config/h8300/multiply.md: Likewise.
24088 * config/h8300/other.md: Likewise.
24089 * config/h8300/shiftrotate.md: Likewise.
24090 * config/h8300/logical.md: Likewise. Fix split pattern to use
24091 code iterator that somehow slipped through.
24093 2021-06-04 Tobias Burnus <tobias@codesourcery.com>
24095 PR middle-end/100905
24096 * tree-nested.c (convert_nonlocal_omp_clauses,
24097 convert_local_omp_clauses): Handle OMP_CLAUSE_BIND.
24099 2021-06-04 Martin Sebor <msebor@redhat.com>
24101 PR middle-end/100732
24102 * gimple-fold.c (gimple_fold_builtin_sprintf): Avoid folding calls
24103 with either source or destination argument of invalid type.
24104 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Avoid checking
24105 calls with arguments of invalid type.
24107 2021-06-04 Martin Sebor <msebor@redhat.com>
24109 * attribs.c (init_attr_rdwr_indices): Use VLA bounds in the expected
24111 (attr_access::vla_bounds): Also handle VLA bounds.
24113 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
24115 * config/i386/predicates.md (GOT_memory_operand):
24116 Implement using match_code RTXes.
24117 (GOT32_symbol_operand): Ditto.
24119 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
24122 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
24124 (ix86_expand_vector_init_general): Ditto.
24125 Use SImode instead of word_mode for logic operations
24126 when GET_MODE_SIZE (mode) < UNITS_PER_WORD.
24127 (expand_vec_perm_even_odd_1): Assert that V2HI mode should be
24128 implemented by expand_vec_perm_1.
24129 (expand_vec_perm_broadcast_1): Assert that V2HI and V4HI modes
24130 should be implemented using standard shuffle patterns.
24131 (ix86_vectorize_vec_perm_const): Handle V2HImode. Add V4HI and
24132 V2HI modes to modes, implementable with shuffle for one operand.
24133 * config/i386/mmx.md (*punpckwd): New insn_and_split pattern.
24134 (*pshufw_1): New insn pattern.
24135 (*vec_dupv2hi): Ditto.
24136 (vec_initv2hihi): New expander.
24138 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
24140 * config/arm/vfp.md (no_literal_pool_df_immediate,
24141 no_literal_pool_sf_immediate): Fix empty split condition.
24143 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
24145 * config/i386/i386.md (*load_tp_x32_zext, *add_tp_x32_zext,
24146 *tls_dynamic_gnu2_combine_32): Fix empty split condition.
24147 * config/i386/sse.md (*<sse2_avx2>_pmovmskb_lt,
24148 *<sse2_avx2>_pmovmskb_zext_lt, *sse2_pmovmskb_ext_lt,
24149 *<sse4_1_avx2>_pblendvb_lt): Likewise.
24151 2021-06-04 Jakub Jelinek <jakub@redhat.com>
24154 * config/i386/i386-expand.c (ix86_expand_vector_init): Handle
24155 concatenation from half-sized modes with TImode elements.
24157 2021-06-04 Claudiu Zissulescu <claziss@synopsys.com>
24159 * config/arc/arc.c (arc_override_options): Disable millicode
24160 thunks when RF16 is on.
24162 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
24164 * config/rs6000/rs6000.h (PROMOTE_MODE): Remove.
24166 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
24168 * config/rs6000/rs6000-call.c (rs6000_promote_function_mode):
24169 Replace PROMOTE_MODE marco with its content.
24171 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
24173 * config/cris/cris.md (*addi_reload): Fix empty split condition.
24175 2021-06-03 Jim Wilson <jimw@sifive.com>
24177 * config.gcc (riscv*-*-*): If --with-riscv-attribute not used,
24178 turn it on for all riscv targets.
24180 2021-06-03 Uroš Bizjak <ubizjak@gmail.com>
24183 * config/i386/i386-expand.c (ix86_expand_vector_set):
24184 Handle V2HI and V4QI modes.
24185 (ix86_expand_vector_extract): Ditto.
24186 * config/i386/mmx.md (*pinsrw): New insn pattern.
24189 (*pextrw_zext): Ditto.
24191 (*pextrb_zext): Ditto.
24192 (vec_setv2hi): New expander.
24193 (vec_extractv2hihi): Ditto.
24194 (vec_setv4qi): Ditto.
24195 (vec_extractv4qiqi): Ditto.
24196 (vec_setv8qi): Enable only for TARGET_SSE4_1.
24197 (vec_extractv8qiqi): Ditto.
24199 2021-06-03 Aaron Sawdey <acsawdey@linux.ibm.com>
24201 * config/rs6000/genfusion.pl (gen_logical_addsubf): Fix input
24202 order to subf instruction.
24203 * config/rs6000/fusion.md: Regenerate.
24205 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
24207 * calls.c (get_size_range): Use range_of_expr instead of
24208 determine_value_range.
24209 * tree-affine.c (expr_to_aff_combination): Same.
24210 * tree-data-ref.c (split_constant_offset): Same.
24211 * tree-vrp.c (determine_value_range_1): Remove.
24212 (determine_value_range): Remove.
24213 * tree-vrp.h (determine_value_range): Remove.
24215 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
24217 * function-tests.c (test_ranges): Call gimple_range_tests.
24218 * gimple-range-cache.cc (ranger_cache::range_of_expr): Pass stmt
24220 * gimple-range.cc (fur_source::get_operand): Do not call
24221 get_tree_range or gimple_range_global.
24223 (get_tree_range): Move to value-query.cc.
24224 Call get_arith_expr_range.
24225 (gimple_ranger::range_of_expr): Add argument to get_tree_range.
24226 Include gimple-range-tests.cc.
24227 * gimple-range.h (fold_range): Add argument.
24228 (get_tree_range): Remove.
24229 * selftest.h (gimple_range_tests): New.
24230 * value-query.cc (global_range_query::range_of_expr): Add
24232 (range_query::get_tree_range): Move from gimple-range.cc.
24233 * value-query.h (class range_query): Add get_tree_range and
24234 get_arith_expr_range. Make fur_source a friend.
24235 * vr-values.c (vr_values::range_of_expr): Pass stmt to
24237 * gimple-range-tests.cc: New file.
24239 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
24241 * gimple-range.cc (gimple_ranger::export_global_ranges): Call
24242 update_global_range.
24243 * value-query.cc (update_global_range): New.
24244 * value-query.h (update_global_range): New.
24246 2021-06-03 David Malcolm <dmalcolm@redhat.com>
24248 * diagnostic-show-locus.c (diagnostic_show_locus): Don't reject
24249 printing the same location twice if there are fix-it hints,
24250 multiple locations, or a label.
24252 2021-06-03 Andre Vieira <andre.simoesdiasvieira@arm.com>
24254 * tree-vect-loop.c (vect_transform_loop): Use main loop's various'
24255 thresholds to narrow the upper bound on epilogue iterations.
24257 2021-06-03 Christophe Lyon <christophe.lyon@linaro.org>
24259 * config/arm/mve.md (mve_vabsq_f<mode>): Use 'abs' instead of unspec.
24260 (mve_vabsq_s<mode>): Likewise.
24261 * config/arm/neon.md (abs<mode>2): Rename to neon_abs<mode>2.
24262 * config/arm/unspecs.md (VABSQ_F, VABSQ_S): Delete.
24263 * config/arm/vec-common.md (neg<mode>2): Rename to
24264 <absneg_str><mode>2.
24266 2021-06-03 Claudiu Zissulescu <claziss@synopsys.com>
24268 * common/config/arc/arc-common.c (arc_option_optimization_table):
24269 Remove malign-call.
24270 * config/arc/arc.c (arc_unalign_branch_p): Remove unused function.
24271 * config/arc/arc.h (TARGET_MIXED_CODE): Remove macro.
24272 (INDEX_REG_CLASS): Only refer to GENERAL_REGS.
24273 * config/arc/arc.md (abssi2_mixed): Remove pattern.
24274 * config/arc/arc.opt (munalign-prob-threshold): Mark it obsolete.
24275 (malign-call): Likewise.
24276 (mmixed-code): Likewise.
24277 * doc/invoke.texi (ARC): Update doc.
24279 2021-06-03 Martin Liska <mliska@suse.cz>
24281 * common.opt: Use proper Enum values.
24282 * opts.c (COVERAGE_SANITIZER_OPT): Remove.
24283 (parse_sanitizer_options): Handle only sanitizer_opts.
24284 (common_handle_option): Just assign value.
24286 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
24289 * tree-inline.c (inline_forbidden_p): Remove test on return type.
24291 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
24293 * dwarf2out.c (loc_list_from_tree_1) <FUNCTION_DECL>: Also generate
24294 DW_OP_GNU_variable_value referencing an existing DIE at file scope.
24295 (type_byte_size): Inline into...
24296 (add_byte_size_attribute): ...this and call add_scalar_info.
24298 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
24300 * dwarf2out.c (mem_loc_descriptor) <UDIV>: Fix typo.
24301 (typed_binop_from_tree): New function.
24302 (loc_list_from_tree_1) <EXACT_DIV_EXPR>: For an unsigned type,
24303 turn a divide by a power of 2 into a shift.
24304 <CEIL_DIV_EXPR>: For an unsigned type, use a signed divide if the
24305 size of the mode is lower than DWARF2_ADDR_SIZE; otherwise, do a
24306 typed divide by calling typed_binop_from_tree.
24308 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
24310 * dwarf2out.c (scompare_loc_descriptor): Fix head comment.
24311 (is_handled_procedure_type): Likewise.
24312 (struct loc_descr_context): Add strict_signedness field.
24313 (resolve_args_picking_1): Deal with DW_OP_[GNU_]deref_type,
24314 DW_OP_[GNU_]convert and DW_OP_[GNU_]reinterpret.
24315 (resolve_args_picking): Minor tweak.
24316 (function_to_dwarf_procedure): Initialize strict_signedness field.
24317 (type_byte_size): Likewise.
24318 (field_byte_offset): Likewise.
24319 (gen_descr_array_type_die): Likewise.
24320 (gen_variant_part): Likewise.
24321 (loc_list_from_tree_1) <CALL_EXPR>: Tidy up and set strict_signedness
24322 to true when a context is present before evaluating the arguments.
24323 <COND_EXPR>: Do not generate a useless comparison with zero.
24324 When dereferencing an address, if strict_signedness is true and the
24325 type is small and signed, use DW_OP_deref_type to do the dereference
24326 and then DW_OP_convert to convert back to the generic type.
24328 2021-06-03 Jakub Jelinek <jakub@redhat.com>
24331 * tree-inline.c (copy_tree_body_r): Handle iterators on
24332 OMP_CLAUSE_AFFINITY or OMP_CLAUSE_DEPEND.
24334 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
24336 * config/arc/arc.md (*bbit_di): Remove.
24338 2021-06-02 Christoph Muellner <cmuellner@gcc.gnu.org>
24340 PR rtl-optimization/100264
24341 * ree.c (get_sub_rtx): Ignore SET expressions without register
24342 destinations and remove assertion, as it is not valid anymore
24343 with this new behaviour.
24344 (merge_def_and_ext): Eliminate destination check for register
24345 as such SET expressions can't occur anymore.
24346 (combine_reaching_defs): Likewise.
24348 2021-06-02 Jakub Jelinek <jakub@redhat.com>
24351 * config/xtensa/xtensa.h (LEAF_REG_REMAP): Cast REGNO to int to avoid
24352 -Wtype-limits warnings.
24353 (DWARF_FRAME_REGISTER): Rewrite into ternary operator with addition
24354 in operands to avoid -Wsign-compare warnings.
24356 2021-06-02 Pat Haugen <pthaugen@linux.ibm.com>
24358 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): Use
24361 2021-06-02 Vineet Gupta <vgupta@synopsys.com>
24363 * config/arc/arc.h (TARGET_CPU_DEFAULT): Change to hs38_linux.
24365 2021-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
24367 * config/s390/s390.md(*ashrdi3_31<setcc><cconly>): Use a single
24369 * config/s390/subst.md(cconly_subst): Use a single constraint
24370 in (match_scratch).
24372 2021-06-02 Martin Liska <mliska@suse.cz>
24374 * ipa-icf.h: Use auto_vec for memory_access_types.
24376 2021-06-02 Jeff Law <jeffreyalaw@gmail.com>
24378 * config/h8300/h8300-protos.h (compute_a_shift_length): Drop unused
24379 argument from prototype.
24380 (output_logical_op): Add rtx_code argument.
24381 (compute_logical_op_length): Likewise.
24382 * config/h8300/h8300.c (h8300_and_costs): Pass additional argument
24383 to compute_a_shift_length.
24384 (output_logical_op); New argument with the rtx code rather than
24385 extracting it from an operand. Handle QImode too.
24386 (compute_logical_op_length): Similary.
24387 (compute_a_shift_length): Drop unused argument.
24388 * config/h8300/h8300.md (logicals): New code iterator.
24389 * config/h8300/logical.md (<code><mode>3 expander): Combine
24390 the "and" expander with the "ior"/"xor" expander.
24391 (bclr<mode>msx): Combine the QI/HI mode patterns.
24392 (<logical><mode>3 insns): Use code iterator rather than match_operator.
24393 Handle QImode as well. Update call to output_logical_op and
24394 compute_logical_op_length to pass in rtx_code
24395 Fix split condition on all define_insn_and_split patterns.
24396 (one_cmpl<mode>2<cczn>): Use <cczn> to support both clobbering
24397 the flags and setting ZN via existing define_subst.
24398 * config/h8300/shiftrotate.md: Drop unused argument from
24399 calls to compute_a_shift_length.
24400 Signed-off-by: Jeff Law <jeffreyalaw@gmail.com>
24402 2021-06-01 Andrew Pinski <apinski@marvell.com>
24404 PR tree-optimization/25290
24405 * tree-ssa-phiopt.c (match_simplify_replacement):
24407 (tree_ssa_phiopt_worker): Use match_simplify_replacement.
24408 (two_value_replacement): Change the comment about
24409 conditional_replacement.
24410 (conditional_replacement): Delete.
24412 2021-06-01 Andrew Pinski <apinski@marvell.com>
24414 PR tree-optimization/95481
24415 * tree-tailcall.c (find_tail_calls): Handle empty typed
24418 2021-06-01 Andrew Pinski <apinski@marvell.com>
24420 * gimplify.c (zero_sized_field_decl): Delete
24421 (zero_sized_type): Delete
24422 (gimplify_init_ctor_eval): Use is_empty_type instead
24423 of zero_sized_field_decl.
24424 (gimplify_modify_expr): Use is_empty_type instead of
24427 2021-06-01 Jason Merrill <jason@redhat.com>
24430 * tree.h (CALL_FROM_NEW_OR_DELETE_P): Adjust comment.
24432 2021-06-01 Jason Merrill <jason@redhat.com>
24435 * diagnostic.h (warning_enabled_at): Declare.
24436 * diagnostic.c (diagnostic_enabled): Factor out from...
24437 (diagnostic_report_diagnostic): ...here.
24438 (warning_enabled_at): New.
24440 2021-06-01 Aldy Hernandez <aldyh@redhat.com>
24442 * gimple-ssa-evrp.c: Enable exporting of global ranges.
24444 2021-06-01 Martin Liska <mliska@suse.cz>
24447 * doc/invoke.texi: Mention that -fgcse-after-reload
24448 is enabled with -O3.
24450 2021-06-01 liuhongt <hongtao.liu@intel.com>
24452 PR tree-optimization/98365
24453 * tree-if-conv.c (strip_nop_cond_scalar_reduction): New function.
24454 (is_cond_scalar_reduction): Handle nop_expr in cond scalar reduction.
24455 (convert_scalar_cond_reduction): Ditto.
24456 (predicate_scalar_phi): Ditto.
24458 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
24460 PR tree-optimization/100781
24461 * gimple-range-cache.cc (ranger_cache::ranger_cache): Enable new
24462 value calculation by default.
24463 (ranger_cache::enable_new_values): New.
24464 (ranger_cache::disable_new_values): New.
24465 (ranger_cache::push_poor_value): Check if new values are allowed.
24466 * gimple-range-cache.h (class ranger_cache): New member/methods.
24467 * gimple-range.cc (gimple_ranger::range_of_expr): Check for debug
24468 statement, and disable/renable new value calculation.
24470 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
24472 * gimple-range-cache.cc (ranger_cache::ssa_range_in_bb): Delete.
24473 (ranger_cache::range_of_def): New.
24474 (ranger_cache::entry_range): New.
24475 (ranger_cache::exit_range): New.
24476 (ranger_cache::range_of_expr): Adjust.
24477 (ranger_cache::range_on_edge): Adjust.
24478 (ranger_cache::propagate_cache): Call exit_range directly.
24479 * gimple-range-cache.h (class ranger_cache): Adjust.
24481 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
24483 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust for
24484 gori_compute being a member rather than base class.
24485 dervied call to member call.
24486 (ranger_cache::dump): No longer dump gori_map.
24487 (ranger_cache::dump_bb): New.
24488 (ranger_cache::get_non_stale_global_range): Adjust for gori_compute
24489 being a member rather than base class.
24490 (ranger_cache::set_global_range): Ditto.
24491 (ranger_cache::ssa_range_in_bb): Ditto.
24492 (ranger_cache::range_of_expr): New.
24493 (ranger_cache::range_on_edge): New.
24494 (ranger_cache::block_range): Adjust for gori_computes. Debug changes.
24495 (ranger_cache::propagate_cache): Adjust debugging output.
24496 (ranger_cache::fill_block_cache): Adjust for gori_computes. Debug
24498 * gimple-range-cache.h (class ranger_cache): Make gori_compute a
24499 member, and inherit from range_query instead.
24500 (ranger_cache::dump_bb): New. split from dump.
24501 * gimple-range-gori.cc (gori_compute::ssa_range_in_bb): Delete.
24502 (gori_compute::expr_range_at_stmt): Delete.
24503 (gori_compute::compute_name_range_op): Delete.
24504 (gori_compute::compute_operand_range_switch): Add fur_source.
24505 (gori_compute::compute_operand_range): Add fur_source param, inline
24506 old compute_name_range_op and optimize_logical_operands.
24507 (struct tf_range): Delete.
24508 (gori_compute::logical_combine): Adjust
24509 (gori_compute::optimize_logical_operands): Delete.
24510 (gori_compute::compute_logical_operands_in_chain): Delete.
24511 (gori_compute::compute_logical_operands): Adjust.
24512 (gori_compute::compute_operand1_range): Adjust to fur_source.
24513 (gori_compute::compute_operand2_range): Ditto.
24514 (gori_compute::compute_operand1_and_operand2_range): Ditto.
24515 (gori_compute::outgoing_edge_range_p): Add range_query parameter,
24516 and adjust to fur_source.
24517 * gimple-range-gori.h (class gori_compute): Simplify and adjust to
24518 range_query and fur_source.
24519 * gimple-range.cc (gimple_ranger::range_on_edge): Query range_on_edge
24520 from the ranger_cache..
24521 (gimple_ranger::fold_range_internal): Adjust to base class change of
24523 (gimple_ranger::dump_bb): Adjust dump.
24524 * gimple-range.h (gimple_ranger):export gori computes object.
24526 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
24528 PR tree-optimization/100774
24529 * gimple-range-cache.cc (ranger_cache::get_non_stale_global_range):
24530 Constant values are also not stale.
24531 (ranger_cache::set_global_range): Range invariant values should also
24532 have the correct timestamp.
24534 2021-05-31 Martin Liska <mliska@suse.cz>
24536 * tree-streamer-in.c (unpack_ts_function_decl_value_fields):
24537 Unpack FUNCTION_DECL_DECL_TYPE.
24538 * tree-streamer-out.c (pack_ts_function_decl_value_fields):
24539 Stream FUNCTION_DECL_DECL_TYPE instead of
24540 DECL_IS_OPERATOR_NEW_P.
24541 * tree.h (set_function_decl_type): Use FUNCTION_DECL_DECL_TYPE
24543 (DECL_IS_OPERATOR_NEW_P): Likewise.
24544 (DECL_IS_OPERATOR_DELETE_P): Likewise.
24545 (DECL_LAMBDA_FUNCTION_P): Likewise.
24547 2021-05-31 Richard Biener <rguenther@suse.de>
24550 * internal-fn.c (expand_SHUFFLEVECTOR): Define.
24551 * internal-fn.def (SHUFFLEVECTOR): New.
24552 * internal-fn.h (expand_SHUFFLEVECTOR): Declare.
24553 * doc/extend.texi: Document __builtin_shufflevector.
24555 2021-05-31 Peter Bergner <bergner@linux.ibm.com>
24558 * config/rs6000/predicates.md(mma_assemble_input_operand): Allow
24559 indexed form addresses.
24561 2021-05-29 Jeff Law <jlaw@tachyum.com>
24563 * config/h8300/h8300.c (h8300_emit_stack_adjustment): Drop unused
24564 parameter. Call callers fixed.
24566 (output_plussi): Add FALLTHRU markers.
24567 (h8300_shift_needs_scratch_p): Add gcc_unreachable marker.
24569 2021-05-29 Jakub Jelinek <jakub@redhat.com>
24571 PR middle-end/99928
24572 * gimplify.c (gimplify_scan_omp_clauses): For taskloop simd
24573 combined with parallel, make sure to add shared clause to
24574 parallel for explicit linear clause.
24576 2021-05-29 Aldy Hernandez <aldyh@redhat.com>
24578 PR tree-optimization/100787
24579 * gimple-ssa-evrp.c: Disable exporting of global ranges.
24581 2021-05-28 Jason Merrill <jason@redhat.com>
24583 * tree-iterator.h (struct tree_stmt_iterator): Add operator++,
24584 operator--, operator*, operator==, and operator!=.
24585 (class tsi_range): New.
24587 2021-05-28 Richard Biener <rguenther@suse.de>
24589 PR tree-optimization/100778
24590 * tree-vect-slp.c (vect_build_slp_tree_1): Prevent possibly
24591 trapping ops in different BBs.
24593 2021-05-28 Richard Biener <rguenther@suse.de>
24596 * tree-inline.c (copy_bb): When processing __builtin_va_arg_pack
24597 copy fntype from original call.
24599 2021-05-28 Martin Liska <mliska@suse.cz>
24601 PR gcov-profile/100751
24602 * doc/gcov.texi: Revert partially a hunk that was wrong.
24604 2021-05-28 Cooper Qu <cooper.qu@linux.alibaba.com>
24606 * config/csky/csky-linux-elf.h (HAVE_sync_compare_and_swapqi):
24608 (HAVE_sync_compare_and_swaphi): Likewise.
24609 (HAVE_sync_compare_and_swapsi): Likewise.
24611 2021-05-28 Jakub Jelinek <jakub@redhat.com>
24613 PR middle-end/99928
24614 * tree.h (OMP_CLAUSE_MAP_IMPLICIT): Define.
24616 2021-05-28 Tobias Burnus <tobias@codesourcery.com>
24618 * gimplify.c (gimplify_omp_affinity): New.
24619 (gimplify_scan_omp_clauses): Call it; remove affinity clause afterwards.
24620 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_AFFINITY.
24621 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_AFFINITY.
24622 * tree.c (omp_clause_num_ops, omp_clause_code_name): Add clause.
24623 (walk_tree_1): Handle OMP_CLAUSE_AFFINITY.
24625 2021-05-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
24626 Richard Biener <rguenther@suse.de>
24628 * match.pd <popcount & / + pattern matching>:
24629 When generating popcount directly fails, try doing it in two halves.
24631 2021-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
24633 * Makefile.in (generated_files): Add gimple-match.c and
24636 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
24638 * gensupport.c (alter_predicate_for_insn): Handle MATCH_DUP.
24640 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
24642 * gensupport.c (alter_constraints): Add MATCH_SCRATCH case.
24644 2021-05-28 Kewen Lin <linkw@linux.ibm.com>
24646 PR tree-optimization/99398
24647 * tree-ssa-forwprop.c (simplify_permutation): Optimize some cases
24648 where the fed operands are CTOR/CST and propagated through
24649 VIEW_CONVERT_EXPR. Call vec_perm_indices::new_shrunk_vector.
24650 * vec-perm-indices.c (vec_perm_indices::new_shrunk_vector): New
24652 * vec-perm-indices.h (vec_perm_indices::new_shrunk_vector): New
24655 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
24657 * config/i386/mmx.md (addv2sf3): Do not call
24658 ix86_fixup_binary_operands_no_copy.
24661 (<smaxmin:code>v2sf3): Ditto.
24662 (<plusminus:insn><MMXMODEI:mode>3): Ditto.
24663 (<plusminus:insn><VI_32:mode>3): Remove expander.
24664 (<plusminus:insn><VI_32:mode>3): Rename from
24665 "*<plusminus:insn><VI_32:mode>3".
24666 (mulv4hi): Do not call ix86_fixup_binary_operands_no_copy.
24667 (mulv2hi3): Remove expander.
24668 (mulv2hi3): Rename from *mulv2hi3.
24669 (<s>mulv2hi3_highpart): Remove expander.
24670 (<s>mulv2hi3_highpart): Rename from *<s>mulv2hi3_highpart.
24671 (<smaxmin:code><MMXMODE14:mode>3): Rename from
24672 "*mmx_<smaxmin:code><MMXMODE14:mode>3".
24673 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Remove expander.
24674 (SMAXMIN_MMXMODEI): Remove mode iterator.
24675 (<smaxmin:code>v4hi3): New expander.
24676 (<smaxmin:code>v4qi3): Rename from *<smaxmin:code>v4qi3.
24677 (<smaxmin:code>v2hi3): Rename from *<smaxmin:code>v2hi3.
24678 (<smaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
24679 (SMAXMIN_VI_32): Remove mode iterator.
24680 (<umaxmin:code><MMXMODE24:mode>3): Rename from
24681 "*mmx_<umaxmin:code><MMXMODE24:mode>3".
24682 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Remove expander.
24683 (UMAXMIN_MMXMODEI): Remove mode iterator.
24684 (<umaxmin:code>v8qi3): New expander.
24685 (<umaxmin:code>v4qi3): Rename from *<umaxmin:code>v4qi3.
24686 (<umaxmin:code>v2hi3): Rename from *<umaxmin:code>v2hi3.
24687 (<umaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
24688 (UMAXMIN_VI_32): Remove mode iterator.
24689 (<any_shift:insn>v2hi3): Remove expander.
24690 (<any_shift:insn>v2hi3): Rename from *<any_shift:insn>v2hi3.
24691 (<any_logic:code><MMXMODEI:mode>3): Do not call
24692 ix86_fixup_binary_operands_no_copy.
24693 (<any_logic:code><VI_32:mode>3): Remove expander.
24694 (<any_logic:code><VI_32:mode>3): Rename from
24695 "*<any_logic:code><VI_32:mode>3".
24696 (uavg<mode>3_ceil): Do not call ix86_fixup_binary_operands_no_copy.
24697 * config/i386/sse.md (div<VF2:mode>3): Do not call
24698 ix86_fixup_binary_operands_no_copy.
24699 (div<VF1:mode>3): Ditto.
24700 (<maxmin:code><VI8_AVX2_AVX512F:mode>3): Ditto.
24701 (smulhrsv4hi3): Ditto.
24702 (smulhrsv2hi3): Ditto.
24704 2021-05-27 Martin Sebor <msebor@redhat.com>
24706 * ggc.h (gt_ggc_mx): Add overloads for all integers.
24708 * hash-map.h (class hash_map): Add pch_nx_helper overloads for all
24710 (hash_map::operator==): New function.
24712 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
24715 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
24716 For TARGET_XOP bypass SSE comparisons for all supported vector modes.
24717 * config/i386/mmx.md (*xop_maskcmp<MMXMODEI:mode>3): New insn pattern.
24718 (*xop_maskcmp<VI_32:mode>3): Ditto.
24719 (*xop_maskcmp_uns<MMXMODEI:mode>3): Ditto.
24720 (*xop_maskcmp_uns<VI_32:mode>3): Ditto.
24722 2021-05-27 Richard Earnshaw <rearnsha@arm.com>
24725 * config/arm/arm.c (arm_configure_build_target): Remove parameter
24726 opts_set, directly check opts parameters for being non-null.
24727 (arm_option_restore): Update call to arm_configure_build_target.
24728 (arm_option_override): Likewise.
24729 (arm_can_inline_p): Likewise.
24730 (arm_valid_target_attribute_tree): Likewise.
24731 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
24732 * config/arm/arm-protos.h (arm_configure_build_target): Adjust
24735 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
24737 * vr-values.c (simplify_conversion_using_ranges): Use
24738 get_range_query instead of get_global_range_query.
24740 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
24742 * gimple-range.cc (get_range_global): Move to value-query.cc.
24743 (gimple_range_global): Same.
24744 (get_global_range_query): Same.
24745 (global_range_query::range_of_expr): Same.
24746 * gimple-range.h (class global_range_query): Move to
24748 (gimple_range_global): Same.
24749 * tree-ssanames.c (get_range_info): Move to value-query.cc.
24750 (get_ptr_nonnull): Same.
24751 * tree-ssanames.h (get_range_info): Remove.
24752 (get_ptr_nonnull): Remove.
24753 * value-query.cc (get_ssa_name_range_info): Move from
24755 (get_ssa_name_ptr_info_nonnull): Same.
24756 (get_range_global): Move from gimple-range.cc.
24757 (gimple_range_global): Same.
24758 (get_global_range_query): Same.
24759 (global_range_query::range_of_expr): Same.
24760 * value-query.h (class global_range_query): Move from
24762 (gimple_range_global): Same.
24764 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
24767 * config/i386/mmx.md (uavgv4qi3_ceil): New insn pattern.
24768 (uavgv2hi3_ceil): Ditto.
24770 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
24773 * doc/extend.texi (scalar_storage_order): Rephrase slightly.
24775 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
24777 * tree-ssanames.c (get_range_info): Merge both copies of
24778 get_range_info into one that works with irange.
24779 * tree-ssanames.h (get_range_info): Remove version that works on
24782 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
24784 * builtins.c (check_nul_terminated_array): Convert to get_range_query.
24785 (expand_builtin_strnlen): Same.
24786 (determine_block_size): Same.
24787 * fold-const.c (expr_not_equal_to): Same.
24788 * gimple-fold.c (size_must_be_zero_p): Same.
24789 * gimple-match-head.c: Include gimple-range.h.
24790 * gimple-pretty-print.c (dump_ssaname_info): Convert to get_range_query.
24791 * gimple-ssa-warn-restrict.c
24792 (builtin_memref::extend_offset_range): Same.
24793 * graphite-sese-to-poly.c (add_param_constraints): Same.
24794 * internal-fn.c (get_min_precision): Same.
24795 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): Same.
24796 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Same.
24798 * tree-data-ref.c (split_constant_offset): Same.
24799 (dr_step_indicator): Same.
24800 * tree-dfa.c (get_ref_base_and_extent): Same.
24801 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
24802 * tree-ssa-loop-niter.c (refine_value_range_using_guard): Same.
24803 (determine_value_range): Same.
24804 (record_nonwrapping_iv): Same.
24805 (infer_loop_bounds_from_signedness): Same.
24806 (scev_var_range_cant_overflow): Same.
24807 * tree-ssa-phiopt.c (two_value_replacement): Same.
24808 * tree-ssa-pre.c (insert_into_preds_of_block): Same.
24809 * tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Same.
24810 * tree-ssa-strlen.c (handle_builtin_stxncpy_strncat): Same.
24812 (dump_strlen_info): Same.
24813 (set_strlen_range): Same.
24814 (maybe_diag_stxncpy_trunc): Same.
24815 (get_len_or_size): Same.
24816 (handle_integral_assign): Same.
24817 * tree-ssa-structalias.c (find_what_p_points_to): Same.
24818 * tree-ssa-uninit.c (find_var_cmp_const): Same.
24819 * tree-switch-conversion.c (bit_test_cluster::emit): Same.
24820 * tree-vect-patterns.c (vect_get_range_info): Same.
24821 (vect_recog_divmod_pattern): Same.
24822 * tree-vrp.c (intersect_range_with_nonzero_bits): Same.
24823 (register_edge_assert_for_2): Same.
24824 (determine_value_range_1): Same.
24825 * tree.c (get_range_pos_neg): Same.
24826 * vr-values.c (vr_values::get_lattice_entry): Same.
24827 (vr_values::update_value_range): Same.
24828 (simplify_conversion_using_ranges): Same.
24830 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
24832 * gimple-ssa-warn-alloca.c (alloca_call_type): Use
24833 get_range_query instead of query argument.
24834 (pass_walloca::execute): Enable and disable global ranger.
24836 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
24838 * gimple-ssa-evrp.c (rvrp_folder::rvrp_folder): Call
24840 (rvrp_folder::~rvrp_folder): Call disable_ranger.
24841 (hybrid_folder::hybrid_folder): Call enable_ranger.
24842 (hybrid_folder::~hybrid_folder): Call disable_ranger.
24844 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
24846 * function.c (allocate_struct_function): Set cfun->x_range_query.
24847 * function.h (struct function): Declare x_range_query.
24848 (get_range_query): New.
24849 (get_global_range_query): New.
24850 * gimple-range-cache.cc (ssa_global_cache::ssa_global_cache):
24851 Remove call to safe_grow_cleared.
24852 * gimple-range.cc (get_range_global): New.
24853 (gimple_range_global): Move from gimple-range.h.
24854 (get_global_range_query): New.
24855 (global_range_query::range_of_expr): New.
24856 (enable_ranger): New.
24857 (disable_ranger): New.
24858 * gimple-range.h (gimple_range_global): Move to gimple-range.cc.
24859 (class global_range_query): New.
24860 (enable_ranger): New.
24861 (disable_ranger): New.
24862 * gimple-ssa-evrp.c (evrp_folder::~evrp_folder): Rename
24863 dump_all_value_ranges to dump.
24864 * tree-vrp.c (vrp_prop::finalize): Same.
24865 * value-query.cc (range_query::dump): New.
24866 * value-query.h (range_query::dump): New.
24867 * vr-values.c (vr_values::dump_all_value_ranges): Rename to...
24868 (vr_values::dump): ...this.
24869 * vr-values.h (class vr_values): Rename dump_all_value_ranges to
24870 dump and make virtual.
24872 2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
24874 * config/i386/i386.c (ix86_autovectorize_vector_modes):
24875 Add V4QImode and V16QImode for TARGET_SSE2.
24876 * doc/sourcebuild.texi (Vector-specific attributes):
24877 Add vect64 and vect32 description.
24879 2021-05-26 Bernd Edlinger <bernd.edlinger@hotmail.de>
24881 * gimple-range-gori.cc (range_def_chain::register_dependency):
24882 Resize m_def_chain when needed.
24884 2021-05-26 Christophe Lyon <christophe.lyon@linaro.org>
24886 * config/arm/mve.md (mve_vaddvq_<supf><mode>): Prefix with '@'.
24887 * config/arm/neon.md (reduc_plus_scal_<mode>): Move to ..
24888 * config/arm/vec-common.md: .. here. Add support for MVE.
24890 2021-05-26 Jakub Jelinek <jakub@redhat.com>
24892 * config/epiphany/epiphany.c (epiphany_print_operand_address): Remove
24894 * config/microblaze/microblaze.c (microblaze_legitimize_address,
24896 microblaze_option_override, print_operand): Likewise.
24897 * config/microblaze/microblaze.md (call_internal_plt,
24898 call_value_intern_plt, call_value_intern): Likewise.
24899 * config/arm/aout.h (ASM_OUTPUT_ALIGN): Likewise.
24900 * config/iq2000/iq2000.md (call_internal1, call_value_internal1,
24901 call_value_multiple_internal1): Likewise.
24902 * config/bfin/bfin.c (symbolic_reference_mentioned_p): Likewise.
24904 2021-05-26 Jan-Benedict Glaw <jbglaw@lug-owl.de>
24906 * config/arc/arc.c (arc_address_cost, arc_print_operand_address,
24907 arc_ccfsm_advance, symbolic_reference_mentioned_p,
24908 arc_raw_symbolic_reference_mentioned_p): Remove register
24911 2021-05-26 Jakub Jelinek <jakub@redhat.com>
24914 * omp-low.c: Include omp-offload.h.
24915 (create_omp_child_function): If current_function_decl has
24916 "omp declare target" attribute and is_gimple_omp_offloaded,
24917 remove that attribute from the copy of attribute list and
24918 add "omp target entrypoint" attribute instead.
24919 (lower_omp_target): Mark .omp_data_sizes.* and .omp_data_kinds.*
24920 variables for offloading if in omp_maybe_offloaded_ctx.
24921 * omp-offload.c (pass_omp_target_link::execute): Nullify second
24922 argument to GOMP_target_data_ext in offloaded code.
24924 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
24926 * config/csky/csky.c (csky_can_change_mode_class): Delete.
24927 For csky, HF/SF mode use the low bits of VREGS.
24929 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
24931 * gimplify.c (gimplify_decl_expr): Do not clear TREE_READONLY on a
24932 DECL which is a reference for OMP.
24934 2021-05-26 Martin Liska <mliska@suse.cz>
24936 PR gcov-profile/100751
24937 * doc/gcov.texi: Document that __gcov_dump can be called just
24938 once and that __gcov_reset resets run-time counters.
24940 2021-05-26 Martin Liska <mliska@suse.cz>
24942 * doc/install.texi: Port relevant part from install-old.texi
24943 and re-generate list of CPUs and systems.
24945 2021-05-26 Martin Liska <mliska@suse.cz>
24947 * Makefile.in: Remove it.
24948 * doc/include/fdl.texi: Update next/previous chapters.
24949 * doc/install.texi: Likewise.
24950 * doc/install-old.texi: Removed.
24952 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
24954 * config/csky/csky.c (ck810_legitimate_index_p): Support
24955 "base + index" with DF mode.
24956 * config/csky/constraints.md ("Y"): New constraint for memory operands
24957 without index register.
24958 * config/csky/csky_insn_fpuv2.md (fpuv3_movdf): Use "Y" instead of "m"
24959 when mov between memory and general registers, and lower their priority.
24960 * config/csky/csky_insn_fpuv3.md (fpuv2_movdf): Likewise.
24962 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
24964 * config/csky/csky.c (TARGET_PROMOTE_PROTOTYPES): Delete.
24966 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
24968 * config/csky/csky.md (untyped_call): Emit clobber for return
24969 registers to mark them used.
24971 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
24973 * config/csky/csky.md (cskyv2_sextend_ldbs): New.
24975 2021-05-26 Andrew Pinski <apinski@marvell.com>
24977 * match.pd (x < 0 ? ~y : y): New patterns.
24979 2021-05-26 Andrew Pinski <apinski@marvell.com>
24981 * match.pd (A?CST1:CST2): Add simplifcations for A?0:+-1, A?+-1:0,
24982 A?POW2:0 and A?0:POW2.
24984 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
24986 * gimple-range-gori.cc (class logical_stmt_cache): Delete
24987 (logical_stmt_cache::logical_stmt_cache ): Delete.
24988 (logical_stmt_cache::~logical_stmt_cache): Delete.
24989 (logical_stmt_cache::cache_entry::dump): Delete.
24990 (logical_stmt_cache::get_range): Delete.
24991 (logical_stmt_cache::cached_name ): Delete.
24992 (logical_stmt_cache::same_cached_name): Delete.
24993 (logical_stmt_cache::cacheable_p): Delete.
24994 (logical_stmt_cache::slot_diagnostics ): Delete.
24995 (logical_stmt_cache::dump): Delete.
24996 (gori_compute_cache::gori_compute_cache): Delete.
24997 (gori_compute_cache::~gori_compute_cache): Delete.
24998 (gori_compute_cache::compute_operand_range): Delete.
24999 (gori_compute_cache::cache_stmt): Delete.
25000 * gimple-range-gori.h (gori_compute::compute_operand_range): Remove
25002 (class gori_compute_cache): Delete.
25004 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
25006 * gimple-range.cc (fold_using_range::range_of_range_op): Use m_gori
25008 (fold_using_range::range_of_address): Adjust.
25009 (fold_using_range::range_of_phi): Adjust.
25010 * gimple-range.h (class fur_source): Adjust.
25011 (fur_source::fur_source): Adjust.
25013 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
25015 * gimple-range-gori.cc (gori_compute::expr_range_at_stmt): Rename
25016 from expr_range_in_bb and adjust.
25017 (gori_compute::compute_name_range_op): Adjust.
25018 (gori_compute::optimize_logical_operands): Adjust.
25019 (gori_compute::compute_logical_operands_in_chain): Adjust.
25020 (gori_compute::compute_operand1_range): Adjust.
25021 (gori_compute::compute_operand2_range): Adjust.
25022 (ori_compute_cache::cache_stmt): Adjust.
25023 * gimple-range-gori.h (gori_compute): Rename prototype.
25025 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
25027 * gimple-range.cc (gimple_ranger::range_of_expr): Non-null should be
25028 checked only after range_of_stmt, not range_on_entry.
25029 (gimple_ranger::range_on_entry): Check for non-null in any
25030 predecessor block, if it is not already non-null.
25031 (gimple_ranger::range_on_exit): DOnt check for non-null after
25032 range on entry call.
25033 (gimple_ranger::dump_bb): New. Split from dump.
25034 (gimple_ranger::dump): Adjust.
25035 * gimple-range.h (class gimple_ranger): Adjust.
25037 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
25039 * gimple-range-cache.cc (struct range_timestamp): Delete.
25040 (class temporal_cache): Adjust.
25041 (temporal_cache::get_timestamp): Delete.
25042 (temporal_cache::set_dependency): Delete.
25043 (temporal_cache::temporal_value): Adjust.
25044 (temporal_cache::current_p): Take dependencies as params.
25045 (temporal_cache::set_timestamp): Adjust.
25046 (temporal_cache::set_always_current): Adjust.
25047 (ranger_cache::get_non_stale_global_range): Adjust.
25048 (ranger_cache::register_dependency): Delete.
25049 * gimple-range-cache.h (class range_cache): Adjust.
25051 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
25053 * gimple-range-gori.cc (range_def_chain::range_def_chain): init
25055 (range_def_chain::~range_def_chain): Dispose of obstack rather than
25056 each individual bitmap.
25057 (range_def_chain::set_import): New.
25058 (range_def_chain::get_imports): New.
25059 (range_def_chain::chain_import_p): New.
25060 (range_def_chain::register_dependency): Rename from build_def_chain
25062 (range_def_chain::def_chain_in_bitmap_p): New.
25063 (range_def_chain::add_def_chain_to_bitmap): New.
25064 (range_def_chain::has_def_chain): Just check first depenedence.
25065 (range_def_chain::get_def_chain): Process imports, use generic
25066 register_dependency routine.
25067 (range_def_chain::dump): New.
25068 (gori_map::gori_map): Allocate import list.
25069 (gori_map::~gori_map): Release imports.
25070 (gori_map::exports): Check for past allocated block size.
25071 (gori_map::imports): New.
25072 (gori_map::def_chain_in_export_p): Delete.
25073 (gori_map::is_import_p): New.
25074 (gori_map::maybe_add_gori): Handle imports.
25075 (gori_map::dump): Adjust output, add imports.
25076 (gori_compute::has_edge_range_p): Remove def_chain_in_export call.
25077 (gori_export_iterator::gori_export_iterator): New.
25078 (gori_export_iterator::next): New.
25079 (gori_export_iterator::get_name): New.
25080 * gimple-range-gori.h (range_def_chain): Add imports and direct
25081 dependecies via struct rdc.
25082 (range_def_chain::depend1): New.
25083 (range_def_chain::depend2): New.
25084 (class gori_map): Adjust.
25085 (FOR_EACH_GORI_IMPORT_NAME): New.
25086 (FOR_EACH_GORI_EXPORT_NAME): New.
25087 (class gori_export_iterator): New.
25089 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
25091 * gimple-range-cache.cc (ranger_cache::ranger_cache): Move initial
25092 export cache filling to here.
25093 * gimple-range-gori.cc (gori_compute::gori_compute) : From Here.
25095 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
25097 * gimple-range-gori.cc (range_def_chain): Move to gimple-range-gori.h.
25098 (gori_map): Move to gimple-range-gori.h.
25099 (gori_compute::gori_compute): Adjust.
25100 (gori_compute::~gori_compute): Delete.
25101 (gori_compute::compute_operand_range_switch): Adjust.
25102 (gori_compute::compute_operand_range): Adjust.
25103 (gori_compute::compute_logical_operands): Adjust.
25104 (gori_compute::has_edge_range_p ): Adjust.
25105 (gori_compute::set_range_invariant): Delete.
25106 (gori_compute::dump): Adjust.
25107 (gori_compute::outgoing_edge_range_p): Adjust.
25108 * gimple-range-gori.h (class range_def_chain): Relocate here.
25109 (class gori_map): Relocate here.
25110 (class gori_compute): Inherit from gori_map, and adjust.
25112 2021-05-25 Aldy Hernandez <aldyh@redhat.com>
25114 * value-range.cc (range_tests_legacy): Use
25115 build_nonstandard_integer_type instead of int and short.
25117 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
25119 * gimplify.c (gimplify_decl_expr): Clear TREE_READONLY on the DECL
25120 when really creating an initialization statement for it.
25122 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
25124 * tree-inline.c (setup_one_parameter): Fix thinko in new condition.
25126 2021-05-25 Kito Cheng <kito.cheng@sifive.com>
25128 * config/riscv/riscv.h (ASM_SPEC): Pass -mno-relax.
25130 2021-05-25 Martin Liska <mliska@suse.cz>
25132 PR tree-optimization/92860
25134 * optc-save-gen.awk: Remove exceptions.
25136 2021-05-25 Martin Liska <mliska@suse.cz>
25138 * asan.h (sanitize_coverage_p): New function.
25139 * doc/extend.texi: Document it.
25140 * fold-const.c (fold_range_test): Use sanitize_flags_p
25141 instead of flag_sanitize_coverage.
25142 (fold_truth_andor): Likewise.
25143 * sancov.c: Likewise.
25144 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
25145 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Handle
25146 -fsanitize-coverage when inlining.
25148 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
25150 * config/csky/csky-modes.def : Fix copyright.
25152 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
25154 * config/csky/csky-modes.def : Amend copyright.
25155 * config/csky/csky_insn_fpuv2.md : Likewise.
25156 * config/csky/csky_insn_fpuv3.md : Likewise.
25158 2021-05-25 Richard Biener <rguenther@suse.de>
25160 PR middle-end/100727
25161 * calls.c (initialize_argument_information): Explicitely test
25162 for WITH_SIZE_EXPR.
25163 * gimple-expr.c (mark_addressable): Skip outer WITH_SIZE_EXPR.
25165 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
25167 * config/csky/csky.h (FRAME_POINTER_REGNUM): Use
25168 HARD_FRAME_POINTER_REGNUM and FRAME_POINTER_REGNUM instead of
25169 the signle definition. The signle definition may not work well
25170 at simplify_subreg_regno().
25171 (HARD_FRAME_POINTER_REGNUM): New.
25172 (ELIMINABLE_REGS): Add for HARD_FRAME_POINTER_REGNUM.
25173 * config/csky/csky.c (get_csky_live_regs, csky_can_eliminate,
25174 csky_initial_elimination_offset, csky_expand_prologue,
25175 csky_expand_epilogue): Add for HARD_FRAME_POINTER_REGNUM.
25177 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
25179 * config/csky/csky.c (csky_option_override):
25180 Init csky_arch_isa_features[] in advance, so TARGET_DSP
25181 and TARGET_DIV can be set well.
25183 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
25185 * config/csky/constraints.md ("l", "h"): Delete.
25186 * config/csky/csky.h (reg_class, REG_CLASS_NAMES,
25187 REG_CLASS_CONTENTS): Delete LO_REGS and HI_REGS.
25188 * config/csky/csky.c (regno_reg_classm,
25189 csky_secondary_reload, csky_register_move_cost):
25190 Use HILO_REGS instead of LO_REGS and HI_REGS.
25192 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
25194 * config/csky/constraints.md ("W"): New constriant for mem operand
25195 with base reg, index register.
25196 ("Q"): Renamed and modified "csky_valid_fpuv2_mem_operand" to
25197 "csky_valid_mem_constraint_operand" to deal with both "Q" and "W"
25199 ("Dv"): New constraint for const double value that can be used at
25201 * config/csky/csky-modes.def (HFmode): New mode.
25202 * config/csky/csky-protos.h (csky_valid_fpuv2_mem_operand): Rename
25203 to "csky_valid_mem_constraint_operand" and support new constraint
25205 (csky_get_movedouble_length): New.
25206 (fpuv3_output_move): New.
25207 (fpuv3_const_double): New.
25208 * config/csky/csky.c (csky_option_override): New arch CK860 with fpv3.
25209 (decompose_csky_address): Refine.
25210 (csky_print_operand): New "CONST_DOUBLE" operand.
25211 (csky_output_move): Support fpv3 instructions.
25212 (csky_get_movedouble_length): New.
25213 (fpuv3_output_move): New.
25214 (fpuv3_const_double): New.
25215 (csky_emit_compare): Cover float comparsion.
25216 (csky_emit_compare_float): Refine.
25217 (csky_vaild_fpuv2_mem_operand): Rename to
25218 "csky_valid_mem_constraint_operand" and support new constraint "W".
25219 (ck860_rtx_costs): New.
25220 (csky_rtx_costs): Add the cost calculation of CK860.
25221 (regno_reg_class): New vregs for fpuv3.
25222 (csky_dbx_regno): Likewise.
25223 (csky_cpu_cpp_builtins): New builtin macro for fpuv3.
25224 (csky_conditional_register_usage): Suporrot fpuv3.
25225 (csky_dwarf_register_span): Suporrot fpuv3.
25226 (csky_init_builtins, csky_mangle_type): Support "__fp16" type.
25227 (ck810_legitimate_index_p): Support fp16.
25228 * config/csky/csky.h (TARGET_TLS): ADD CK860.
25229 (CSKY_VREG_P, CSKY_VREG_LO_P, CSKY_VREG_HI_P): Support fpuv3.
25230 (TARGET_SINGLE_FPU): Support fpuv3.
25231 (TARGET_SUPPORT_FPV3): New.
25232 (FIRST_PSEUDO_REGISTER): Change to 202 to hold the new fpuv3 registers.
25233 (FIXED_REGISTERS, CALL_REALLY_USED_REGISTERS, REGISTER_NAMES,
25234 REG_CLASS_CONTENTS): Support fpuv3.
25235 * config/csky/csky.md (movsf): Move to cksy_insn_fpu.md and refine.
25236 (csky_movsf_fpv2): Likewise.
25237 (ck801_movsf): Likewise.
25238 (csky_movsf): Likewise.
25240 (csky_movdf_fpv2): Likewise.
25241 (ck801_movdf): Likewise.
25242 (csky_movdf): Likewise.
25243 (movsicc): Refine. Use "comparison_operatior" instead of
25244 "ordered_comparison_operatior".
25245 (addsicc): Likewise.
25246 (CSKY_FIRST_VFP3_REGNUM, CSKY_LAST_VFP3_REGNUM): New constant.
25247 (call_value_internal_vh): New.
25248 * config/csky/csky_cores.def (CK860): New arch and cpu.
25253 * config/csky/csky_insn_fpu.md: Refactor. Separate all float patterns
25254 into emit-patterns and match-patterns, remain the emit-patterns here,
25255 and move the match-patterns to csky_insn_fpuv2.md or
25256 csky_insn_fpuv3.md.
25257 * config/csky/csky_insn_fpuv2.md: New file for fpuv2 instructions.
25258 * config/csky/csky_insn_fpuv3.md: New file and new patterns for fpuv3
25260 * config/csky/csky_isa.def (fcr): New.
25265 (CK860): New definition for ck860.
25266 * config/csky/csky_tables.opt (ck860): New processors ck860,
25267 ck860f. And new arch ck860.
25272 * config/csky/predicates.md (csky_float_comparsion_operator): Delete
25273 "geu", "gtu", "leu", "ltu", which will never appear at float comparison.
25274 * config/csky/t-csky-elf: Support 860.
25275 * config/csky/t-csky-linux: Likewise.
25276 * doc/md.texi: Add "Q" and "W" constraints for C-SKY.
25278 2021-05-24 Aaron Sawdey <acsawdey@linux.ibm.com>
25280 * config/rs6000/genfusion.pl (gen_logical_addsubf): Refactor to
25281 add generation of logical-add and add-logical fusion pairs.
25282 * config/rs6000/rs6000-cpus.def: Add new fusion to ISA 3.1 mask
25284 * config/rs6000/rs6000.c (rs6000_option_override_internal): Turn on
25285 logical-add and add-logical fusion by default.
25286 * config/rs6000/rs6000.opt: Add -mpower10-fusion-logical-add and
25287 -mpower10-fusion-add-logical options.
25288 * config/rs6000/fusion.md: Regenerate file.
25290 2021-05-24 Aldy Hernandez <aldyh@redhat.com>
25292 * value-range.cc (irange::legacy_equal_p): Check type when
25293 comparing VR_VARYING types.
25294 (range_tests_legacy): Test comparing VARYING ranges of different
25297 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
25299 * config/aarch64/aarch64.c (neoversen1_tunings):
25300 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
25302 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
25304 * config/aarch64/aarch64.c (aarch64_classify_symbol): Use GOT for
25305 extern weak symbols. Limit symbol offsets for non-GOT symbols with
25308 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
25310 * config/arm/neon.md (vec_load_lanesxi<mode>)
25311 (vec_store_lanexoi<mode>): Move ...
25312 * config/arm/vec-common.md: here.
25314 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
25316 * config/arm/neon.md (vec_load_lanesoi<mode>)
25317 (vec_store_lanesoi<mode>): Move ...
25318 * config/arm/vec-common.md: here.
25320 2021-05-24 liuhongt <hongtao.liu@intel.com>
25323 * config/i386/i386.c (ix86_gimple_fold_builtin): Replacing
25324 stmt with GIMPLE_NOP when lhs doesn't exist.
25326 2021-05-23 Uroš Bizjak <ubizjak@gmail.com>
25329 * config/i386/mmx.md (*push<VI_32:mode>2_rex64):
25330 New instruction pattern.
25331 (*push<VI_32:mode>2): Ditto.
25332 (push splitter for SSE registers): New splitter.
25334 2021-05-23 Andrew Pinski <apinski@marvell.com>
25336 * match.pd ((A & C) != 0 ? D : 0): Limit to non pointer types.
25338 2021-05-22 Aaron Sawdey <acsawdey@linux.ibm.com>
25340 * config/rs6000/genfusion.pl (gen_addadd): Fix incorrect attr types.
25341 * config/rs6000/fusion.md: Regenerate file.
25343 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
25345 * config/rs6000/genfusion.pl (gen_addadd): New function.
25346 * config/rs6000/fusion.md: Regenerate file.
25347 * config/rs6000/rs6000-cpus.def: Add
25348 OPTION_MASK_P10_FUSION_2ADD to masks.
25349 * config/rs6000/rs6000.c (rs6000_option_override_internal):
25350 Handle default value of OPTION_MASK_P10_FUSION_2ADD.
25351 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2add.
25353 2021-05-21 Jakub Jelinek <jakub@redhat.com>
25355 PR middle-end/99928
25356 * tree.h (OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET): Define.
25357 * gimplify.c (enum gimplify_omp_var_data): Fix up
25358 GOVD_MAP_HAS_ATTACHMENTS value, add GOVD_FIRSTPRIVATE_IMPLICIT.
25359 (omp_lastprivate_for_combined_outer_constructs): If combined target
25360 has GOVD_FIRSTPRIVATE_IMPLICIT set for the decl, change it to
25361 GOVD_MAP | GOVD_SEEN.
25362 (gimplify_scan_omp_clauses): Set GOVD_FIRSTPRIVATE_IMPLICIT for
25363 firstprivate clauses with OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT.
25364 (gimplify_adjust_omp_clauses): For firstprivate clauses with
25365 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT either clear that bit and
25366 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET too, or remove it and
25367 let it be replaced by implicit map clause.
25369 2021-05-21 Jakub Jelinek <jakub@redhat.com>
25371 PR middle-end/99928
25372 * gimplify.c (omp_lastprivate_for_combined_outer_constructs): New
25374 (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LASTPRIVATE>: Use it.
25375 (gimplify_omp_for): Likewise.
25377 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25379 PR middle-end/90115
25380 * omp-low.c (oacc_privatization_candidate_p): Reject 'static',
25381 'external' in blocks.
25383 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25385 PR middle-end/90115
25386 * flag-types.h (enum openacc_privatization): New.
25387 * params.opt (-param=openacc-privatization): New.
25388 * doc/invoke.texi (openacc-privatization): Document it.
25389 * omp-general.h (get_openacc_privatization_dump_flags): New
25391 * omp-low.c (oacc_privatization_candidate_p): Add diagnostics.
25392 * omp-offload.c (execute_oacc_device_lower)
25393 <IFN_UNIQUE_OACC_PRIVATE>: Re-work diagnostics.
25394 * target.def (goacc.adjust_private_decl): Add 'location_t'
25396 * doc/tm.texi: Regenerate.
25397 * config/gcn/gcn-protos.h (gcn_goacc_adjust_private_decl): Adjust.
25398 * config/gcn/gcn-tree.c (gcn_goacc_adjust_private_decl): Likewise.
25399 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl):
25400 Likewise. Preserve it for...
25401 (nvptx_goacc_expand_var_decl): ... use here.
25403 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25405 * doc/sourcebuild.texi (Other attributes): Document '__OPTIMIZE__'
25408 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25410 PR middle-end/90115
25411 * omp-low.c (oacc_privatization_candidate_p): New function.
25412 (oacc_privatization_scan_clause_chain)
25413 (oacc_privatization_scan_decl_chain): Use it. Also
25414 'gcc_checking_assert' that we're not seeing duplicates.
25416 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25418 PR middle-end/90115
25419 * omp-offload.c (execute_oacc_device_lower): Skip processing if no
25422 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25424 PR middle-end/90115
25425 * omp-offload.c (execute_oacc_device_lower): Explain.
25427 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25429 PR middle-end/90115
25430 * omp-offload.c (execute_oacc_device_lower)
25431 <IFN_UNIQUE_OACC_PRIVATE>: Diagnose and handle for 'level == -1'
25433 * internal-fn.c (expand_UNIQUE): Don't expect
25434 'IFN_UNIQUE_OACC_PRIVATE'.
25436 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25438 PR middle-end/90115
25439 * omp-low.c (lower_omp_for): Don't evaluate OpenMP 'for' clauses.
25441 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
25443 PR middle-end/90115
25444 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl)
25445 (nvptx_goacc_expand_var_decl): Tighten.
25447 2021-05-21 Julian Brown <julian@codesourcery.com>
25448 Chung-Lin Tang <cltang@codesourcery.com>
25449 Thomas Schwinge <thomas@codesourcery.com>
25451 PR middle-end/90115
25452 * doc/tm.texi.in (TARGET_GOACC_EXPAND_VAR_DECL)
25453 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Add documentation hooks.
25454 * doc/tm.texi: Regenerate.
25455 * expr.c (expand_expr_real_1): Expand decls using the
25456 expand_var_decl OpenACC hook if defined.
25457 * internal-fn.c (expand_UNIQUE): Handle IFN_UNIQUE_OACC_PRIVATE.
25458 * internal-fn.h (IFN_UNIQUE_CODES): Add OACC_PRIVATE.
25459 * omp-low.c (omp_context): Add oacc_privatization_candidates
25461 (lower_oacc_reductions): Add PRIVATE_MARKER parameter. Insert
25463 (lower_oacc_head_tail): Add PRIVATE_MARKER parameter. Modify
25464 private marker's gimple call arguments, and pass it to
25465 lower_oacc_reductions.
25466 (oacc_privatization_scan_clause_chain)
25467 (oacc_privatization_scan_decl_chain, lower_oacc_private_marker):
25469 (lower_omp_for, lower_omp_target, lower_omp_1): Use these.
25470 * omp-offload.c (convert.h): Include.
25471 (oacc_loop_xform_head_tail): Treat private-variable markers like
25472 fork/join when transforming head/tail sequences.
25473 (struct var_decl_rewrite_info): Add struct.
25474 (oacc_rewrite_var_decl, is_sync_builtin_call): New functions.
25475 (execute_oacc_device_lower): Support rewriting gang-private
25476 variables using target hook, and fix up addr_expr and var_decl
25478 * target.def (adjust_private_decl, expand_var_decl): New hooks.
25479 * config/gcn/gcn-protos.h (gcn_goacc_adjust_gangprivate_decl):
25481 (gcn_goacc_adjust_private_decl): ...this.
25482 * config/gcn/gcn-tree.c (gcn_goacc_adjust_gangprivate_decl):
25484 (gcn_goacc_adjust_private_decl): ...this. Add LEVEL parameter.
25485 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_GANGPRIVATE_DECL): Rename
25486 definition using gcn_goacc_adjust_gangprivate_decl...
25487 (TARGET_GOACC_ADJUST_PRIVATE_DECL): ...to this, using
25488 gcn_goacc_adjust_private_decl.
25489 * config/nvptx/nvptx.c (tree-pretty-print.h): Include.
25490 (gang_private_shared_size): New global variable.
25491 (gang_private_shared_align): Likewise.
25492 (gang_private_shared_sym): Likewise.
25493 (gang_private_shared_hmap): Likewise.
25494 (nvptx_option_override): Initialize these.
25495 (nvptx_file_end): Output gang_private_shared_sym.
25496 (nvptx_goacc_adjust_private_decl, nvptx_goacc_expand_var_decl):
25498 (nvptx_set_current_function): Clear gang_private_shared_hmap.
25499 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Define hook.
25500 (TARGET_GOACC_EXPAND_VAR_DECL): Likewise.
25502 2021-05-21 H.J. Lu <hjl.tools@gmail.com>
25504 * config/i386/i386-modes.def (MAX_BITSIZE_MODE_ANY_INT): Removed.
25506 2021-05-21 Richard Biener <rguenther@suse.de>
25507 H.J. Lu <hjl.tools@gmail.com>
25509 PR middle-end/90773
25510 * expr.c (expand_constructor): Elide expand_constructor if
25511 move by pieces is preferred.
25513 2021-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25515 * config/aarch64/aarch64-builtins.c (aarch64_call_properties):
25516 Take a flag and mode value as arguments.
25517 (aarch64_modifies_global_state_p): Likewise.
25518 (aarch64_reads_global_state_p): Likewise.
25519 (aarch64_could_trap_p): Likewise.
25520 (aarch64_get_attributes): Likewise.
25521 (aarch64_init_simd_builtins): Adjust callsite of above.
25522 (aarch64_init_fcmla_laneq_builtins): Use aarch64_get_attributes to get
25523 function attributes to apply to builtins.
25524 (aarch64_init_crc32_builtins): Likewise.
25525 (aarch64_init_builtin_rsqrt): Likewise.
25527 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
25529 * config/rs6000/rs6000.md (define_attr "type"): Add types for fusion.
25530 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10): Use new fusion types.
25531 (gen_2logical): Use new fusion types.
25532 * config/rs6000/fusion.md: Regenerate.
25534 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
25537 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
25538 Handle V4QI and V2HI modes.
25539 (ix86_expand_sse_movcc): Ditto.
25540 * config/i386/mmx.md (*<sat_plusminus:insn><VI_32:mode>3):
25541 New instruction pattern.
25542 (*eq<VI_32:mode>3): Ditto.
25543 (*gt<VI_32:mode>3): Ditto.
25544 (*xop_pcmov_<VI_32:mode>): Ditto.
25545 (mmx_pblendvb32): Ditto.
25546 (mmx_pblendvb64): Rename from mmx_pblendvb.
25547 (vec_cmp<VI_32:mode><VI_32:mode>): New expander.
25548 (vec_cmpu<VI_32:mode><VI_32:mode>): Ditto.
25549 (vcond<VI_32:mode><VI_32:mode>): Ditto.
25550 (vcondu<VI_32:mode><VI_32:mode>): Ditto.
25551 (vcond_mask_<VI_32:mode><VI_32:mode>): Ditto.
25553 2021-05-21 Jakub Jelinek <jakub@redhat.com>
25555 PR tree-optimization/94589
25556 * tree-ssa-phiopt.c (spaceship_replacement): For integral rhs1 and
25557 rhs2, treat x <= 4 equivalently to x < 5 etc. In cmp1 and cmp2 (if
25558 not the same as cmp3) treat <= the same as < and >= the same as >.
25559 Don't require that cond2_phi_edge is true edge, instead take
25560 false/true edges into account based on cmp1/cmp2 comparison kinds.
25562 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
25565 * config/i386/mmx.md (SMAXMIN_MMXMODEI): New mode iterator.
25566 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Macroize expander
25567 from <smaxmin:code>v4hi3> and <smaxmin:code><MMXMODE14:mode>3
25568 using SMAXMIN_MMXMODEI mode iterator.
25569 (*<smaxmin:code>v4qi3): New insn pattern.
25570 (*<smaxmin:code>v2hi3): Ditto.
25571 (SMAXMIN_VI_32): New mode iterator.
25572 (<smaxmin:code><SMAXMIN_VI_32>mode3): New expander.
25573 (UMAXMIN_MMXMODEI): New mode iterator.
25574 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Macroize expander
25575 from <umaxmin:code>v8qi3> and <umaxmin:code><MMXMODE24:mode>3
25576 using UMAXMIN_MMXMODEI mode iterator.
25577 (*<umaxmin:code>v4qi3): New insn pattern.
25578 (*<umaxmin:code>v2hi3): Ditto.
25579 (UMAXMIN_VI_32): New mode iterator.
25580 (<umaxmin:code><UMAXMIN_VI_32>mode3): New expander.
25581 (abs<VI_32:mode>2): New insn pattern.
25582 (ssse3_abs<MMXMODEI:mode>2, abs<MMXMODEI:mode>2): Move from ...
25583 * config/i386/sse.md: ... here.
25585 2021-05-20 Clement Chigot <clement.chigot@atos.net>
25586 David Edelsohn <dje.gcc@gmail.com>
25588 * collect2.c (scan_prog_file): Issue non-fatal warning for
25591 2021-05-20 Jonathan Wakely <jwakely@redhat.com>
25593 * doc/invoke.texi (-Wno-c++11-extensions)
25594 (-Wno-c++14-extensions, -Wno-c++17-extensions)
25595 (-Wno-c++20-extensions, -Wno-c++23-extensions): Document
25598 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
25600 * config/c6x/c6x.c (c6x_output_file_unwind): Use dwarf_debuginfo_p.
25601 * config/darwin.c (darwin_override_options): Likewise.
25602 * config/i386/cygming.h (DBX_REGISTER_NUMBER): Likewise.
25603 * config/i386/darwin.h (DBX_REGISTER_NUMBER): Likewise.
25604 (DWARF2_FRAME_REG_OUT): Likewise.
25605 * config/mips/mips.c (mips_output_filename): Likewise.
25606 * config/rs6000/rs6000.c (rs6000_xcoff_declare_function_name):
25608 (rs6000_dbx_register_number): Likewise.
25609 * dbxout.c: Include flags.h.
25610 * dwarf2cfi.c (cfi_label_required_p): Likewise.
25611 (dwarf2out_do_frame): Likewise.
25612 * except.c: Include flags.h.
25613 * final.c (dwarf2_debug_info_emitted_p): Likewise.
25614 (final_scan_insn_1): Likewise.
25615 * flags.h (dwarf_debuginfo_p): New function declaration.
25616 * opts.c (dwarf_debuginfo_p): New function definition.
25617 * targhooks.c (default_debug_unwind_info): Use dwarf_debuginfo_p.
25618 * toplev.c (process_options): Likewise.
25620 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
25622 * common.opt: Change type to support bitmasks.
25623 * flag-types.h (enum debug_info_type): Rename enumerator constants.
25624 (NO_DEBUG): New bitmask.
25625 (DBX_DEBUG): Likewise.
25626 (DWARF2_DEBUG): Likewise.
25627 (XCOFF_DEBUG): Likewise.
25628 (VMS_DEBUG): Likewise.
25629 (VMS_AND_DWARF2_DEBUG): Likewise.
25630 * flags.h (debug_set_to_format): New function declaration.
25631 (debug_set_count): Likewise.
25632 (debug_set_names): Likewise.
25633 * opts.c (debug_type_masks): Array of bitmasks for debug formats.
25634 (debug_set_to_format): New function definition.
25635 (debug_set_count): Likewise.
25636 (debug_set_names): Likewise.
25637 (set_debug_level): Update access to debug_type_names.
25638 * toplev.c: Likewise.
25640 2021-05-20 Martin Sebor <msebor@redhat.com>
25642 PR middle-end/100684
25643 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Handle C++ lambda.
25645 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
25648 * config/i386/i386.md (isa): Remove x64_bmi.
25649 (enabled): Remove x64_bmi.
25650 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
25651 Remove general register alternative.
25652 (*andnot<VI_32:mode>3): Ditto.
25653 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
25654 (*<any_logic:code><VI_32:mode>3): Ditto.
25656 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
25658 * config/arm/arm.c: Include head files tree-vectorizer.h and
25661 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
25664 * config/i386/mmx.md (Yv_Yw): Revert adding V4QI and V2HI modes.
25665 (*<plusminus:insn><VI_32:mode>3): Use Yw instad of <Yv_Yw> constrint.
25666 (<s>mulv4hi3_highpart): New expander.
25667 (*<s>mulv2hi3_highpart): New insn pattern.
25668 (<s>mulv2hi3_higpart): New expander.
25669 (*<any_shift:insn>v2hi3): New insn pattern.
25670 (<any_shift:insn>v2hi3): New expander.
25671 * config/i386/sse.md (smulhrsv2hi3): New expander.
25672 (*smulhrsv2hi3): New insn pattern.
25674 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
25676 * doc/invoke.texi (vect-inner-loop-cost-factor): Document new
25678 * params.opt (vect-inner-loop-cost-factor): New.
25679 * targhooks.c (default_add_stmt_cost): Replace hardcoded factor
25680 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR, include head file
25681 tree-vectorizer.h and its required ones.
25682 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Replace
25683 hardcoded factor 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR.
25684 * config/arm/arm.c (arm_add_stmt_cost): Likewise.
25685 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
25686 * config/rs6000/rs6000.c (rs6000_add_stmt_cost): Likewise.
25687 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
25689 (_loop_vec_info::_loop_vec_info): Init inner_loop_cost_factor.
25690 * tree-vectorizer.h (_loop_vec_info): Add inner_loop_cost_factor.
25691 (LOOP_VINFO_INNER_LOOP_COST_FACTOR): New macro.
25693 2021-05-20 Christophe Lyon <christophe.lyon@linaro.org>
25694 Torbjörn Svensson <torbjorn.svensson@st.com>
25697 * doc/cpp.texi (Common Predefined Macros): Document __FILE_NAME__.
25699 2021-05-20 Jakub Jelinek <jakub@redhat.com>
25701 PR middle-end/99928
25702 * gimplify.c (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LINEAR>: For
25703 explicit linear clause when combined with target, make it map(tofrom:)
25704 instead of no clause or firstprivate.
25706 2021-05-20 Jakub Jelinek <jakub@redhat.com>
25708 PR tree-optimization/94589
25709 * match.pd ((X & Y) == X -> (X & ~Y) == 0): Simplify even in presence
25710 of integral conversions.
25712 2021-05-19 Andrew MacLeod <amacleod@redhat.com>
25714 * gimple-range.cc (fur_source::get_operand): New.
25715 (gimple_range_fold): Delete.
25716 (fold_using_range::fold_stmt): Move from gimple_ranger::calc_stmt.
25717 (fold_using_range::range_of_range_op): Move from gimple_ranger.
25718 (fold_using_range::range_of_address): Ditto.
25719 (fold_using_range::range_of_phi): Ditto.
25720 (fold_using_range::range_of_call): Ditto.
25721 (fold_using_range::range_of_builtin_ubsan_call): Move from
25722 range_of_builtin_ubsan_call.
25723 (fold_using_range::range_of_builtin_call): Move from
25724 range_of_builtin_call.
25725 (gimple_ranger::range_of_builtin_call): Delete.
25726 (fold_using_range::range_of_cond_expr): Move from gimple_ranger.
25727 (gimple_ranger::fold_range_internal): New.
25728 (gimple_ranger::range_of_stmt): Use new fold_using_range API.
25729 (fold_using_range::range_of_ssa_name_with_loop_info): Move from
25730 gimple_ranger. Improve ranges of SSA_NAMES when possible.
25731 * gimple-range.h (gimple_ranger): Remove various range_of routines.
25732 (class fur_source): New.
25733 (class fold_using_range): New.
25734 (fur_source::fur_source): New.
25736 * vr-values.c (vr_values::extract_range_basic): Use fold_using_range
25737 instead of range_of_builtin_call.
25739 2021-05-19 Jonathan Wakely <jwakely@redhat.com>
25741 * doc/cpp.texi (Common Predefined Macros): Update documentation
25742 for the __GXX_EXPERIMENTAL_CXX0X__ macro.
25744 2021-05-19 Alex Coplan <alex.coplan@arm.com>
25747 * config/arm/arm.md (nonsecure_call_internal): Always ensure
25748 callee's address is in a register.
25750 2021-05-19 Geng Qi <gengqi@linux.alibaba.com>
25752 * common/config/riscv/riscv-common.c
25753 (riscv_subset_list::parsing_subset_version): Properly parse the letter
25755 (riscv_subset_list::parse_std_ext,
25756 riscv_subset_list::parse_multiletter_ext): To handle errors generated
25757 in riscv_subset_list::parsing_subset_version.
25759 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
25761 * config/aarch64/aarch64-simd.md: Use "neon_move_narrow_q"
25762 type attribute in patterns generating XTN(2).
25764 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
25766 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_pack_trunc_<mode>):
25767 Remove as duplicate of...
25768 (aarch64_xtn<mode>): This.
25769 (aarch64_xtn2<mode>_le): Move position in file.
25770 (aarch64_xtn2<mode>_be): Move position in file.
25771 (aarch64_xtn2<mode>): Move position in file.
25772 (vec_pack_trunc_<mode>): Define as an expander.
25774 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
25776 * config/aarch64/aarch64-simd-builtins.def: Split builtin
25777 generation for aarch64_<sur>q<r>shr<u>n_n<mode> pattern into
25778 separate scalar and vector generators.
25779 * config/aarch64/aarch64-simd.md
25780 (aarch64_<sur>q<r>shr<u>n_n<mode>): Define as an expander and
25782 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_le): This and...
25783 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_be): This.
25784 * config/aarch64/iterators.md: Define SD_HSDI iterator.
25786 2021-05-19 Jonathn Wright <jonathan.wright@arm.com>
25788 * config/aarch64/aarch64-simd.md: Use UNSPEC_SQXTUN instead
25790 * config/aarch64/iterators.md: Remove UNSPEC_SQXTUN2.
25792 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
25794 * config/aarch64/aarch64-simd.md (aarch64_<sur>q<r>shr<u>n2_n<mode>):
25795 Implement as an expand emitting a big/little endian
25796 instruction pattern.
25797 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_le): Define.
25798 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_be): Define.
25800 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
25802 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn2<mode>):
25803 Implement as an expand emitting a big/little endian
25804 instruction pattern.
25805 (aarch64_<sur><addsub>hn2<mode>_insn_le): Define.
25806 (aarch64_<sur><addsub>hn2<mode>_insn_be): Define.
25807 * config/aarch64/iterators.md: Remove UNSPEC_[R]ADDHN2 and
25808 UNSPEC_[R]SUBHN2 unspecs and ADDSUBHN2 iterator.
25810 2021-05-19 Richard Biener <rguenther@suse.de>
25812 PR middle-end/100672
25813 * fold-const.c (fold_negate_expr_1): Use element_precision.
25814 (negate_expr_p): Likewise.
25816 2021-05-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
25818 * config/aarch64/iterators.md (SVE_PRED_LOAD): New iterator.
25819 (pred_load): New int attribute.
25820 * config/aarch64/aarch64-sve.md
25821 (aarch64_load_<ANY_EXTEND:optab><SVE_HSDI:mode><SVE_PARTIAL_I:mode>): Use
25822 SVE_PRED_LOAD enum iterator and corresponding pred_load attribute.
25823 * config/aarch64/aarch64-sve-builtins-base.cc (expand): Update call to
25824 code_for_aarch64_load.
25826 2021-05-19 Richard Biener <rguenther@suse.de>
25828 * cfgexpand.c (discover_nonconstant_array_refs_r): Make
25829 sure TARGET_MEM_REF bases are expanded as memory.
25830 * tree-ssa-operands.c (operands_scanner::get_tmr_operands):
25831 Do not mark TARGET_MEM_REF bases addressable.
25832 * tree-ssa.c (non_rewritable_mem_ref_base): Handle
25833 TARGET_MEM_REF bases as never rewritable.
25834 * gimple-walk.c (walk_stmt_load_store_addr_ops): Do not
25835 walk TARGET_MEM_REF bases as address-takens.
25836 * tree-ssa-dce.c (ref_may_be_aliased): Handle TARGET_MEM_REF.
25838 2021-05-19 Richard Biener <rguenther@suse.de>
25840 * builtins.c (get_object_alignment_1): Strip outer
25842 * tree-dfa.c (get_ref_base_and_extent): Handle outer
25843 WITH_SIZE_EXPR for size processing and process the
25845 * tree-ssa-alias.c (ao_ref_base_alias_set): Strip
25846 outer WITH_SIZE_EXPR.
25847 (ao_ref_base_alias_ptr_type): Likewise.
25848 (refs_may_alias_p_2): Allow WITH_SIZE_EXPR in ref->ref
25849 and handle that accordingly, stripping it for the
25850 core alias workers.
25851 * tree.c (get_base_address): Handle WITH_SIZE_EXPR by
25852 looking through it instead of returning NULL.
25854 2021-05-19 Jakub Jelinek <jakub@redhat.com>
25856 PR middle-end/100576
25857 * builtins.c (check_read_access): Convert bound to size_type_node if
25860 2021-05-19 Richard Biener <rguenther@suse.de>
25862 * tree-cfg.c (verify_types_in_gimple_min_lval): Inline...
25863 (verify_types_in_gimple_reference): ... here. Sanitize.
25864 (verify_gimple_call): Verify references in LHS and arguments.
25865 (verify_gimple_assign_single): Reject WITH_SIZE_EXPR.
25867 2021-05-19 Uroš Bizjak <ubizjak@gmail.com>
25869 * config/i386/i386.h (VALID_INT_MODE_P):
25870 Add V8QI, V4HI and V2SI modes for TARGET_64BIT.
25871 * config/i386/i386.md (isa): Add x64_bmi.
25872 (enabled): Handle x64_bmi.
25873 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
25874 Add alternative using 64bit general registers.
25875 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
25877 2021-05-19 Jakub Jelinek <jakub@redhat.com>
25879 PR middle-end/99928
25880 * tree.h (OMP_MASTER_COMBINED): Define.
25881 * gimplify.c (gimplify_scan_omp_clauses): Rewrite lastprivate
25882 handling for outer combined/composite constructs to a loop.
25883 Handle lastprivate on combined target.
25884 (gimplify_expr): Formatting fix.
25886 2021-05-19 Xionghu Luo <luoxhu@linux.ibm.com>
25888 * passes.def: Add sink_code pass before store_merging.
25889 * tree-ssa-sink.c (pass_sink_code:clone): New.
25891 2021-05-18 Bill Schmidt <wschmidt@linux.ibm.com>
25893 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Remove call to
25894 rs6000_special_adjust_field_align_p.
25895 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
25896 * config/rs6000/rs6000-call.c (rs6000_function_arg_boundary):
25897 Remove ABI warning.
25898 (rs6000_function_arg): Likewise.
25899 * config/rs6000/rs6000-protos.h
25900 (rs6000_special_adjust_field_align_p): Remove prototype.
25901 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
25903 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Remove call to
25904 rs6000_special_adjust_field_align_p.
25906 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
25909 * config/i386/i386.h (VALID_SSE2_REG_MODE):
25910 Add V4QI and V2HI modes.
25911 (VALID_INT_MODE_P): Ditto.
25912 * config/i386/mmx.md (VI_32): New mode iterator.
25913 (mmxvecsize): Handle V4QI and V2HI.
25915 (mov<VI_32:mode>): New expander.
25916 (*mov<mode>_internal): New insn pattern.
25917 (movmisalign<VI_32:mode>): New expander.
25918 (neg<VI_32:mode>): New expander.
25919 (<plusminus:insn><VI_32:mode>3): New expander.
25920 (*<plusminus:insn><VI_32:mode>3): New insn pattern.
25921 (mulv2hi3): New expander.
25922 (*mulv2hi3): New insn pattern.
25923 (one_cmpl<VI_32:mode>2): New expander.
25924 (*andnot<VI_32:mode>3): New insn pattern.
25925 (<any_logic:code><VI_32:mode>3): New expander.
25926 (*<any_logic:code><VI_32:mode>3): New insn pattern.
25928 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
25930 * config/i386/sse.md (<any_extend:insn>v4qiv4di2):
25931 Fix a mode mismatch with operand 1.
25933 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
25936 * config/i386/i386-expand.c (split_double_mode): Return
25937 temporary register when simplify_gen_subreg fails with
25938 the high half od the paradoxical subreg.
25940 2021-05-18 Richard Biener <rguenther@suse.de>
25942 * cfgexpand.c (expand_one_var): Pass in forced_stack_var
25943 and honor it when expanding.
25944 (expand_used_vars_for_block): Pass through forced_stack_var.
25945 (expand_used_vars): Likewise.
25946 (discover_nonconstant_array_refs_r): Set bits in
25947 forced_stack_vars instead of marking vars TREE_ADDRESSABLE.
25948 (avoid_type_punning_on_regs): Likewise.
25949 (discover_nonconstant_array_refs): Likewise.
25950 (pass_expand::execute): Create and pass down forced_stack_var
25951 bitmap. For parameters and returns temporarily set
25952 TREE_ADDRESSABLE when expand_function_start.
25954 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
25956 * doc/sourcebuild.texi: Document 'dg-note'.
25958 2021-05-18 Tobias Burnus <tobias@codesourcery.com>
25961 * configure: Regenerate.
25962 * configure.ac (BUILD_CFLAG, BUILD_CXXFLAGS): Add $(CFLAGS-$@).
25964 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
25966 * gimple.h (is_gimple_omp_oacc): Tighten.
25967 * omp-low.c (check_omp_nesting_restrictions): Adjust.
25969 2021-05-18 Richard Biener <rguenther@suse.de>
25971 * tree-ssa-operands.c (mark_address_taken): Simplify.
25973 2021-05-18 Martin Liska <mliska@suse.cz>
25975 * config/gcn/mkoffload.c (STR): Redefine.
25976 * config/i386/intelmic-mkoffload.c (STR): Likewise.
25977 * config/nvptx/mkoffload.c (STR): Likewise.
25979 2021-05-18 Martin Liska <mliska@suse.cz>
25981 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
25982 Use startswith function instead of strncmp.
25983 * common/config/bfin/bfin-common.c (bfin_handle_option): Likewise.
25984 * common/config/riscv/riscv-common.c (riscv_subset_list::parse): Likewise.
25985 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): Likewise.
25986 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Likewise.
25987 * config/alpha/alpha.c (alpha_elf_section_type_flags): Likewise.
25988 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
25989 * config/arm/arm.c (arm_file_start): Likewise.
25990 (arm_valid_target_attribute_rec): Likewise.
25991 (thumb1_md_asm_adjust): Likewise.
25992 * config/arm/driver-arm.c (host_detect_local_cpu): Likewise.
25993 * config/avr/avr.c (STR_PREFIX_P): Likewise.
25994 (avr_set_current_function): Likewise.
25995 (avr_handle_addr_attribute): Likewise.
25996 (avr_asm_output_aligned_decl_common): Likewise.
25997 (avr_asm_named_section): Likewise.
25998 (avr_section_type_flags): Likewise.
25999 (avr_asm_select_section): Likewise.
26000 * config/c6x/c6x.c (c6x_in_small_data_p): Likewise.
26001 (c6x_section_type_flags): Likewise.
26002 * config/darwin-c.c (darwin_cfstring_ref_p): Likewise.
26003 (darwin_objc_declare_unresolved_class_reference): Likewise.
26004 (darwin_objc_declare_class_definition): Likewise.
26005 * config/darwin.c (indirect_data): Likewise.
26006 (darwin_encode_section_info): Likewise.
26007 (darwin_objc2_section): Likewise.
26008 (darwin_objc1_section): Likewise.
26009 (machopic_select_section): Likewise.
26010 (darwin_globalize_label): Likewise.
26011 (darwin_label_is_anonymous_local_objc_name): Likewise.
26012 (darwin_asm_named_section): Likewise.
26013 (darwin_asm_output_dwarf_offset): Likewise.
26014 * config/frv/frv.c (frv_string_begins_with): Likewise.
26015 (frv_in_small_data_p): Likewise.
26016 * config/gcn/mkoffload.c (STR): Likewise.
26018 * config/i386/i386-builtins.c (get_builtin_code_for_version): Likewise.
26019 * config/i386/i386-options.c (ix86_option_override_internal): Likewise.
26020 * config/i386/i386.c (x86_64_elf_section_type_flags): Likewise.
26021 (ix86_md_asm_adjust): Likewise.
26022 * config/i386/intelmic-mkoffload.c (STR): Likewise.
26023 * config/i386/winnt.c (i386_pe_asm_named_section): Likewise.
26024 (i386_pe_file_end): Likewise.
26025 * config/ia64/ia64.c (ia64_in_small_data_p): Likewise.
26026 (ia64_section_type_flags): Likewise.
26027 * config/mips/driver-native.c (host_detect_local_cpu): Likewise.
26028 * config/mips/mips.c (mips_handle_interrupt_attr): Likewise.
26029 (mips16_stub_function_p): Likewise.
26030 (mips_function_rodata_section): Likewise.
26031 * config/msp430/msp430.c (msp430_mcu_name): Likewise.
26032 (msp430_function_section): Likewise.
26033 (msp430_section_type_flags): Likewise.
26034 (msp430_expand_helper): Likewise.
26035 * config/nios2/nios2.c (nios2_small_section_name_p): Likewise.
26036 (nios2_valid_target_attribute_rec): Likewise.
26037 * config/nvptx/mkoffload.c (process): Likewise.
26039 * config/pa/som.h: Likewise.
26040 * config/pdp11/pdp11.c (pdp11_output_ident): Likewise.
26041 * config/riscv/riscv.c (riscv_elf_select_rtx_section): Likewise.
26042 * config/rs6000/rs6000.c (VTABLE_NAME_P): Likewise.
26043 (rs6000_inner_target_options): Likewise.
26044 * config/s390/driver-native.c (s390_host_detect_local_cpu): Likewise.
26045 * config/sparc/driver-sparc.c (host_detect_local_cpu): Likewise.
26046 * config/vax/vax.c (vax_output_int_move): Likewise.
26047 * config/vms/vms-ld.c (startswith): Likewise.
26048 (process_args): Likewise.
26050 * config/vms/vms.c: Likewise.
26052 2021-05-18 Jakub Jelinek <jakub@redhat.com>
26054 PR rtl-optimization/100590
26055 * regcprop.c (copyprop_hardreg_forward_1): Only DCE dead sets if
26056 they are NONJUMP_INSN_P.
26058 2021-05-18 Jakub Jelinek <jakub@redhat.com>
26061 * function.c (push_dummy_function): Set DECL_ARTIFICIAL and
26062 DECL_ASSEMBLER_NAME on the fn_decl.
26064 2021-05-18 Jakub Jelinek <jakub@redhat.com>
26066 PR tree-optimization/94589
26067 * tree-ssa-phiopt.c (spaceship_replacement): Pattern match
26068 phi result used in (res & ~1) == 0 comparison as res >= 0 as
26069 res == 2 would be UB with -ffinite-math-only.
26071 2021-05-18 Martin Liska <mliska@suse.cz>
26073 * Makefile.in: genversion.o should depend on DATESTAMP.
26075 2021-05-18 Claudiu Zissulescu <claziss@synopsys.com>
26077 * config/arc/simdext.md (negv2si2): Remove round bracket.
26079 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
26081 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Define
26082 _Bool as macro expanding to _Bool.
26084 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
26087 * tree.c (build_reference_type_for_mode)
26088 (build_pointer_type_for_mode): Pick pointer mode if MODE argument
26090 (build_reference_type, build_pointer_type): Invoke
26091 build_*_type_for_mode with VOIDmode.
26093 2021-05-17 Andrew MacLeod <amacleod@redhat.com>
26095 PR tree-optimization/100512
26096 * gimple-range-cache.cc (ranger_cache::set_global_range): Mark const
26097 and non-zero pointer ranges as invariant.
26098 * gimple-range.cc (gimple_ranger::range_of_stmt): Remove pointer
26099 processing from here.
26101 2021-05-17 Tom de Vries <tdevries@suse.de>
26104 * config/nvptx/nvptx-protos.h (nvptx_output_atomic_insn): Declare
26105 * config/nvptx/nvptx.c (nvptx_output_barrier)
26106 (nvptx_output_atomic_insn): New function.
26107 (nvptx_print_operand): Add support for 'B'.
26108 * config/nvptx/nvptx.md: Use nvptx_output_atomic_insn for atomic
26111 2021-05-17 Aldy Hernandez <aldyh@redhat.com>
26113 PR tree-optimization/100349
26114 * vr-values.c (bounds_of_var_in_loop): Bail if scev returns
26117 2021-05-17 Tamar Christina <tamar.christina@arm.com>
26119 * config/aarch64/driver-aarch64.c (DEFAULT_ARCH): New.
26120 (host_detect_local_cpu): Use it.
26122 2021-05-17 Martin Liska <mliska@suse.cz>
26124 * doc/invoke.texi: Add 2 missing dots.
26126 2021-05-17 Marius Hillenbrand <mhillen@linux.ibm.com>
26128 PR bootstrap/100552
26129 * configure.ac: Replace pattern substitution with call to sed.
26130 * configure: Regenerate.
26132 2021-05-17 Richard Biener <rguenther@suse.de>
26134 PR middle-end/100582
26135 * tree.c (array_at_struct_end_p): Get to the base of the
26136 reference before looking for the underlying decl.
26138 2021-05-17 Joern Rennecke <joern.rennecke@embecosm.com>
26140 * genoutput.c (validate_insn_alternatives) Make "wrong number of
26141 alternatives" message more specific, and remove assumption on where
26144 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
26146 * config/arm/iterators.md (V16): New iterator.
26147 (VH_cvtto): New iterator.
26148 (v_cmp_result): Added V4HF and V8HF support.
26149 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result>): Use VDQWH.
26150 (vcond<mode><mode>): Likewise.
26151 (vcond_mask_<mode><v_cmp_result>): Likewise.
26152 (vcond<VH_cvtto><mode>): New expander.
26154 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
26156 * config/arm/arm-protos.h (arm_expand_vector_compare): Update
26158 * config/arm/arm.c (arm_expand_vector_compare): Add support for
26160 (arm_expand_vcond): Likewise.
26161 * config/arm/iterators.md (supf): Remove VCMPNEQ_S, VCMPEQQ_S,
26162 VCMPEQQ_N_S, VCMPNEQ_N_S.
26163 (VCMPNEQ, VCMPEQQ, VCMPEQQ_N, VCMPNEQ_N): Remove.
26164 * config/arm/mve.md (@mve_vcmp<mve_cmp_op>q_<mode>): Add '@' prefix.
26165 (@mve_vcmp<mve_cmp_op>q_f<mode>): Likewise.
26166 (@mve_vcmp<mve_cmp_op>q_n_f<mode>): Likewise.
26167 (@mve_vpselq_<supf><mode>): Likewise.
26168 (@mve_vpselq_f<mode>"): Likewise.
26169 * config/arm/neon.md (vec_cmp<mode><v_cmp_result): Enable for MVE
26170 and move to vec-common.md.
26171 (vec_cmpu<mode><mode>): Likewise.
26172 (vcond<mode><mode>): Likewise.
26173 (vcond<V_cvtto><mode>): Likewise.
26174 (vcondu<mode><v_cmp_result>): Likewise.
26175 (vcond_mask_<mode><v_cmp_result>): Likewise.
26176 * config/arm/unspecs.md (VCMPNEQ_U, VCMPNEQ_S, VCMPEQQ_S)
26177 (VCMPEQQ_N_S, VCMPNEQ_N_S, VCMPEQQ_U, CMPEQQ_N_U, VCMPNEQ_N_U)
26178 (VCMPGEQ_N_S, VCMPGEQ_S, VCMPGTQ_N_S, VCMPGTQ_S, VCMPLEQ_N_S)
26179 (VCMPLEQ_S, VCMPLTQ_N_S, VCMPLTQ_S, VCMPCSQ_N_U, VCMPCSQ_U)
26180 (VCMPHIQ_N_U, VCMPHIQ_U): Remove.
26181 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result): Moved
26183 (vec_cmpu<mode><mode>): Likewise.
26184 (vcond<mode><mode>): Likewise.
26185 (vcond<V_cvtto><mode>): Likewise.
26186 (vcondu<mode><v_cmp_result>): Likewise.
26187 (vcond_mask_<mode><v_cmp_result>): Likewise. Added unsafe math
26190 2021-05-17 liuhongt <hongtao.liu@intel.com>
26193 * config/i386/i386.c (ix86_gimple_fold_builtin): Use
26194 gsi_insert_seq_before instead.
26196 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
26198 * doc/sourcebuild.texi (arm_qbit_ok): Rename into...
26199 (arm_sat_ok): ...this.
26201 2021-05-17 Martin Liska <mliska@suse.cz>
26203 * lto-wrapper.c (merge_flto_options): Factor out a new function.
26204 (merge_and_complain): Use it.
26205 (run_gcc): Merge also linker command line -flto=foo argument
26208 2021-05-16 Christophe Lyon <christophe.lyon@linaro.org>
26210 * config/arm/arm.h (CPP_SPEC): Remove error message about
26211 -mlittle-endian/-mbig-endian conflict.
26213 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
26215 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
26216 __ROP_PROTECT__ if -mrop-protect is selected.
26218 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
26220 * config/rs6000/rs6000-internal.h (rs6000_stack): Add
26221 rop_hash_save_offset and rop_hash_size.
26222 * config/rs6000/rs6000-logue.c (rs6000_stack_info): Compute
26223 rop_hash_size and rop_hash_save_offset.
26224 (debug_stack_info): Dump rop_hash_save_offset and rop_hash_size.
26225 (rs6000_emit_prologue): Emit hashst[p] in prologue.
26226 (rs6000_emit_epilogue): Emit hashchk[p] in epilogue.
26227 * config/rs6000/rs6000.md (unspec): Add UNSPEC_HASHST and
26229 (hashst): New define_insn.
26230 (hashchk): Likewise.
26232 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
26234 * config/rs6000/rs6000.c (rs6000_option_override_internal):
26235 Disable shrink wrap when inserting ROP-protect instructions.
26236 * config/rs6000/rs6000.opt (mrop-protect): New option.
26237 (mprivileged): Likewise.
26238 * doc/invoke.texi: Document mrop-protect and mprivileged.
26240 2021-05-15 Hans-Peter Nilsson <hp@axis.com>
26242 * reorg.c (fill_slots_from_thread): Reinstate code typoed out in
26245 2021-05-15 Martin Jambor <mjambor@suse.cz>
26248 2021-05-13 Martin Jambor <mjambor@suse.cz>
26250 PR tree-optimization/100453
26251 * tree-sra.c (sra_modify_assign): All const base accesses do not
26252 need refreshing, not just those from decl_pool.
26253 (sra_modify_assign): Do not refresh into a const base decl.
26255 2021-05-15 Jakub Jelinek <jakub@redhat.com>
26257 PR rtl-optimization/100342
26258 * regcprop.c (copy_value): When copying a source reg in a wider
26259 mode than it has recorded for the value, adjust recorded destination
26260 mode too or punt if !REG_CAN_CHANGE_MODE_P.
26262 2021-05-14 Jason Merrill <jason@redhat.com>
26264 * intl.h: Add comments.
26266 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26268 * config/aarch64/aarch64-simd.md
26269 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Split into...
26270 (aarch64_sqdmlsl2_lane<mode>_internal): ... This...
26271 (aarch64_sqdmlal2_lane<mode>_internal): ... And this.
26272 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Split into ...
26273 (aarch64_sqdmlsl2_laneq<mode>_internal): ... This...
26274 (aarch64_sqdmlal2_laneq<mode>_internal): ... And this.
26275 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>_internal): Split into...
26276 (aarch64_sqdmlsl2_n<mode>_internal): ... This...
26277 (aarch64_sqdmlal2_n<mode>_internal): ... And this.
26279 2021-05-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
26282 * config/arm/arm_neon.h (vtst_s8): Replace call to vtst builtin with it's
26283 boolean logic equivalent.
26284 (vtst_s16): Likewise.
26285 (vtst_s32): Likewise.
26286 (vtst_u8): Likewise.
26287 (vtst_u16): Likewise.
26288 (vtst_u32): Likewise.
26289 (vtst_p8): Likewise.
26290 (vtst_p16): Likewise.
26291 (vtstq_s8): Likewise.
26292 (vtstq_s16): Likewise.
26293 (vtstq_s32): Likewise.
26294 (vtstq_u8): Likewise.
26295 (vtstq_u16): Likewise.
26296 (vtstq_u32): Likewise.
26297 (vtstq_p8): Likewise.
26298 (vtstq_p16): Likewise.
26299 * config/arm/arm_neon_builtins.def: Remove entry for vtst.
26300 * config/arm/neon.md (neon_vtst<mode>): Remove pattern.
26302 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26304 * config/aarch64/aarch64-simd.md (aarch64_sqdmlal2<mode>): Merge into...
26305 (aarch64_sqdml<SBINQOPS:as>l2<mode>): ... This.
26306 (aarch64_sqdmlsl2<mode>): Delete.
26307 (aarch64_sqdmlal2_lane<mode>): Merge this...
26308 (aarch64_sqdmlsl2_lane<mode>): ... And this...
26309 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>): ... Into this.
26310 (aarch64_sqdmlal2_laneq<mode>): Merge this...
26311 (aarch64_sqdmlsl2_laneq<mode>): ... And this...
26312 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>): ... Into this.
26313 (aarch64_sqdmlal2_n<mode>): Merge this...
26314 (aarch64_sqdmlsl2_n<mode>): ... And this...
26315 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>): ... Into this.
26317 2021-05-13 Martin Sebor <msebor@redhat.com>
26319 PR middle-end/100574
26320 * builtins.c (access_ref::get_ref): Improve detection of PHIs with
26321 all null arguments.
26323 2021-05-13 Martin Sebor <msebor@redhat.com>
26325 PR tree-optimization/93100
26326 PR middle-end/98583
26327 * tree-ssa-uninit.c (check_defs): Exclude intrinsic functions that
26328 don't modify referenced objects.
26330 2021-05-13 Martin Jambor <mjambor@suse.cz>
26332 PR tree-optimization/100453
26333 * tree-sra.c (sra_modify_assign): All const base accesses do not
26334 need refreshing, not just those from decl_pool.
26335 (sra_modify_assign): Do not refresh into a const base decl.
26337 2021-05-13 Martin Liska <mliska@suse.cz>
26339 * tree-ssa-dom.c: Remove m_simplifier.
26341 2021-05-13 Richard Earnshaw <rearnsha@arm.com>
26344 * config/arm/arm.c (arm_canonicalize_comparison): Correctly
26345 canonicalize DImode inequality comparisons against the
26346 maximum integral value.
26348 2021-05-13 Jakub Jelinek <jakub@redhat.com>
26350 PR tree-optimization/98856
26351 * config/i386/i386.c (ix86_shift_rotate_cost): Add CODE argument.
26352 Expect V2DI and V4DI arithmetic right shifts to be emulated.
26353 (ix86_rtx_costs, ix86_add_stmt_cost): Adjust ix86_shift_rotate_cost
26355 * config/i386/i386-expand.c (expand_vec_perm_2perm_interleave,
26356 expand_vec_perm_2perm_pblendv): New functions.
26357 (ix86_expand_vec_perm_const_1): Use them.
26358 * config/i386/sse.md (ashr<mode>3<mask_name>): Rename to ...
26359 (<mask_codefor>ashr<mode>3<mask_name>): ... this.
26360 (ashr<mode>3): New define_expand with VI248_AVX512BW iterator.
26361 (ashrv4di3): New define_expand.
26362 (ashrv2di3): Change condition to TARGET_SSE2, handle !TARGET_XOP
26363 and !TARGET_AVX512VL expansion.
26365 2021-05-13 Uroš Bizjak <ubizjak@gmail.com>
26368 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Force mode
26369 sizes < 16 to a register when constructing vpcmov pattern.
26370 * config/i386/mmx.md (*xop_pcmov_<mode>): Use MMXMODE124 mode.
26372 2021-05-13 Martin Liska <mliska@suse.cz>
26374 * gcov-io.c (gcov_write_block): Remove.
26375 (gcov_write_words): Likewise.
26376 (gcov_read_words): Re-implement using gcov_read_bytes.
26377 (gcov_allocate): Remove.
26378 (GCOV_BLOCK_SIZE): Likewise.
26379 (struct gcov_var): Remove most of the fields.
26380 (gcov_position): Implement with ftell.
26381 (gcov_rewrite): Remove setting of start and offset fields.
26382 (from_file): Re-format.
26383 (gcov_open): Remove setbuf call. It should not be needed.
26384 (gcov_close): Remove internal buffer handling.
26385 (gcov_magic): Use __builtin_bswap32.
26386 (gcov_write_counter): Use directly gcov_write_unsigned.
26387 (gcov_write_string): Use direct fwrite and do not round
26389 (gcov_seek): Use directly fseek.
26390 (gcov_write_tag): Use gcov_write_unsigned directly.
26391 (gcov_write_length): Likewise.
26392 (gcov_write_tag_length): Likewise.
26393 (gcov_read_bytes): Use directly fread.
26394 (gcov_read_unsigned): Use gcov_read_words.
26395 (gcov_read_counter): Likewise.
26396 (gcov_read_string): Use gcov_read_bytes.
26397 * gcov-io.h (GCOV_WORD_SIZE): Adjust to reflect
26398 that size is not in bytes, but words (4B).
26399 (GCOV_TAG_FUNCTION_LENGTH): Likewise.
26400 (GCOV_TAG_ARCS_LENGTH): Likewise.
26401 (GCOV_TAG_ARCS_NUM): Likewise.
26402 (GCOV_TAG_COUNTER_LENGTH): Likewise.
26403 (GCOV_TAG_COUNTER_NUM): Likewise.
26404 (GCOV_TAG_SUMMARY_LENGTH): Likewise.
26406 2021-05-13 liuhongt <hongtao.liu@intel.com>
26409 * config/i386/sse.md (ssedoublevecmode): Add attribute for
26410 V64QI/V32HI/V16SI/V4DI.
26411 (ssehalfvecmode): Add attribute for V2DI/V2DF.
26412 (*vec_concatv4si_0): Extend to VI124_128.
26413 (*vec_concat<mode>_0): New pre-reload splitter.
26414 * config/i386/predicates.md (movq_parallel): New predicate.
26416 2021-05-13 Alexandre Oliva <oliva@adacore.com>
26418 * targhooks.c (default_zero_call_used_regs): Retry using
26419 successfully-zeroed registers as sources.
26421 2021-05-12 Tobias Burnus <tobias@codesourcery.com>
26423 * omp-low.c (finish_taskreg_scan): Use the proper detach decl.
26425 2021-05-12 Aldy Hernandez <aldyh@redhat.com>
26428 * gimple-range.cc (range_of_builtin_call): Skip out on
26429 processing __builtin_clz when varying.
26431 2021-05-12 Tom de Vries <tdevries@suse.de>
26434 * config/nvptx/nvptx-opts.h (enum ptx_version): New enum.
26435 * config/nvptx/nvptx.c (nvptx_file_start): Print .version according
26436 to ptx_version_option.
26437 * config/nvptx/nvptx.h (TARGET_PTX_6_3): Define.
26438 * config/nvptx/nvptx.md (define_insn "nvptx_shuffle<mode>")
26439 (define_insn "nvptx_vote_ballot"): Use sync variant for
26441 * config/nvptx/nvptx.opt (ptx_version): Add enum.
26442 (mptx): Add option.
26443 * doc/invoke.texi (Nvidia PTX Options): Add mptx item.
26445 2021-05-12 Richard Biener <rguenther@suse.de>
26447 PR tree-optimization/100566
26448 * tree-ssa-sccvn.c (dominated_by_p_w_unex): Properly handle
26449 allow_back for all edge queries.
26451 2021-05-12 liuhongt <hongtao.liu@intel.com>
26454 * config/i386/sse.md (<sse4_1_avx2>_pblendvb): Add
26455 splitters for pblendvb of NOT mask register.
26457 2021-05-12 Richard Biener <rguenther@suse.de>
26459 PR tree-optimization/100519
26460 * tree-ssa-reassoc.c (can_associate_p): Split into...
26461 (can_associate_op_p): ... this
26462 (can_associate_type_p): ... and this.
26463 (is_reassociable_op): Call can_associate_op_p.
26464 (break_up_subtract_bb): Call the appropriate predicates.
26465 (reassociate_bb): Likewise.
26467 2021-05-12 Martin Liska <mliska@suse.cz>
26469 * lto-wrapper.c (merge_and_complain): Merge -flto=arg options.
26470 (run_gcc): Use -flto argument detection for merged
26473 2021-05-12 Martin Liska <mliska@suse.cz>
26475 * lto-wrapper.c (print_lto_docs_link): New function.
26476 (run_gcc): Print warning about missing job server detection
26477 after we know NR of partitions. Do the same for -flto{,=1}.
26478 * opts.c (get_option_html_page): Support -flto option.
26480 2021-05-12 Martin Liska <mliska@suse.cz>
26482 * lto-wrapper.c (get_options_from_collect_gcc_options): Change
26484 (append_option): Remove.
26485 (find_option): Rework to use the vector type.
26486 (remove_option): Remove.
26487 (merge_and_complain): Use vectors for cl_decoded_option data
26489 (append_compiler_options): Likewise.
26490 (append_diag_options): Likewise.
26491 (append_linker_options): Likewise.
26492 (append_offload_options): Likewise.
26493 (compile_offload_image): Likewise.
26494 (compile_images_for_offload_targets): Likewise.
26495 (find_and_merge_options): Likewise.
26496 (run_gcc): Likewise.
26498 2021-05-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
26501 * dwarf2out.c (dwarf2out_finish): Set
26502 have_multiple_function_sections with multi-range text_section.
26504 2021-05-12 Martin Liska <mliska@suse.cz>
26506 PR bootstrap/100560
26507 * Makefile.in: Remove version.h from linker command line.
26509 2021-05-12 Richard Biener <rguenther@suse.de>
26511 PR middle-end/100547
26512 * rtl.h (rtvec_alloc): Make argument size_t.
26513 * rtl.c (rtvec_alloc): Verify the count is less than INT_MAX.
26515 2021-05-12 Jakub Jelinek <jakub@redhat.com>
26517 PR middle-end/100508
26518 * cfgexpand.c (expand_debug_expr): For DEBUG_EXPR_DECL with vector
26519 type, don't reuse DECL_RTL if it has different mode, instead force
26520 creation of a new DEBUG_EXPR.
26522 2021-05-12 Jakub Jelinek <jakub@redhat.com>
26523 Marc Glisse <marc.glisse@inria.fr>
26525 PR tree-optimization/94589
26526 * match.pd ((X & Y) == X -> (X & ~Y) == 0,
26527 (X | Y) == Y -> (X & ~Y) == 0): New GIMPLE simplifications.
26529 2021-05-12 Uroš Bizjak <ubizjak@gmail.com>
26532 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Handle V2SF mode.
26533 * config/i386/mmx.md (MMXMODE124): New mode iterator.
26535 (mmxintvecmode): New mode attribute.
26536 (mmxintvecmodelower): Ditto.
26537 (*mmx_maskcmpv2sf3_comm): New insn pattern.
26538 (*mmx_maskcmpv2sf3): Ditto.
26539 (vec_cmpv2sfv2si): New expander.
26540 (vcond<V2FI:mode>v2si): Ditto.
26541 (mmx_vlendvps): New insn pattern.
26542 (vcond<MMXMODE124:mode><MMXMODEI:mode>): Also handle V2SFmode.
26543 (vcondu<MMXMODE124:mode><MMXMODEI:mode>): Ditto.
26544 (vcond_mask_<mode><mmxintvecmodelower>): Ditto.
26546 2021-05-11 Martin Sebor <msebor@redhat.com>
26548 PR middle-end/21433
26549 * expr.c (expand_expr_real_1): Replace unreachable code with an assert.
26551 2021-05-11 Richard Biener <rguenther@suse.de>
26553 * gimple-fold.c (gimple_fold_call): Do not call
26554 maybe_fold_reference on call arguments or the static chain.
26555 (fold_stmt_1): Do not call maybe_fold_reference on GIMPLE_ASM
26558 2021-05-11 Martin Liska <mliska@suse.cz>
26560 * builtins.def (DEF_HSAIL_BUILTIN): Remove.
26561 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
26562 (DEF_HSAIL_SAT_BUILTIN): Likewise.
26563 (DEF_HSAIL_INTR_BUILTIN): Likewise.
26564 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
26565 * doc/frontends.texi: Remove BRIG.
26566 * doc/install.texi: Likewise.
26567 * doc/invoke.texi: Likewise.
26568 * doc/standards.texi: Likewise.
26569 * brig-builtins.def: Removed.
26570 * brig/ChangeLog: Removed.
26571 * brig/Make-lang.in: Removed.
26572 * brig/brig-builtins.h: Removed.
26573 * brig/brig-c.h: Removed.
26574 * brig/brig-lang.c: Removed.
26575 * brig/brigfrontend/brig-arg-block-handler.cc: Removed.
26576 * brig/brigfrontend/brig-atomic-inst-handler.cc: Removed.
26577 * brig/brigfrontend/brig-basic-inst-handler.cc: Removed.
26578 * brig/brigfrontend/brig-branch-inst-handler.cc: Removed.
26579 * brig/brigfrontend/brig-cmp-inst-handler.cc: Removed.
26580 * brig/brigfrontend/brig-code-entry-handler.cc: Removed.
26581 * brig/brigfrontend/brig-code-entry-handler.h: Removed.
26582 * brig/brigfrontend/brig-comment-handler.cc: Removed.
26583 * brig/brigfrontend/brig-control-handler.cc: Removed.
26584 * brig/brigfrontend/brig-copy-move-inst-handler.cc: Removed.
26585 * brig/brigfrontend/brig-cvt-inst-handler.cc: Removed.
26586 * brig/brigfrontend/brig-fbarrier-handler.cc: Removed.
26587 * brig/brigfrontend/brig-function-handler.cc: Removed.
26588 * brig/brigfrontend/brig-function.cc: Removed.
26589 * brig/brigfrontend/brig-function.h: Removed.
26590 * brig/brigfrontend/brig-inst-mod-handler.cc: Removed.
26591 * brig/brigfrontend/brig-label-handler.cc: Removed.
26592 * brig/brigfrontend/brig-lane-inst-handler.cc: Removed.
26593 * brig/brigfrontend/brig-machine.c: Removed.
26594 * brig/brigfrontend/brig-machine.h: Removed.
26595 * brig/brigfrontend/brig-mem-inst-handler.cc: Removed.
26596 * brig/brigfrontend/brig-module-handler.cc: Removed.
26597 * brig/brigfrontend/brig-queue-inst-handler.cc: Removed.
26598 * brig/brigfrontend/brig-seg-inst-handler.cc: Removed.
26599 * brig/brigfrontend/brig-signal-inst-handler.cc: Removed.
26600 * brig/brigfrontend/brig-to-generic.cc: Removed.
26601 * brig/brigfrontend/brig-to-generic.h: Removed.
26602 * brig/brigfrontend/brig-util.cc: Removed.
26603 * brig/brigfrontend/brig-util.h: Removed.
26604 * brig/brigfrontend/brig-variable-handler.cc: Removed.
26605 * brig/brigfrontend/hsa-brig-format.h: Removed.
26606 * brig/brigfrontend/phsa.h: Removed.
26607 * brig/brigspec.c: Removed.
26608 * brig/config-lang.in: Removed.
26609 * brig/gccbrig.texi: Removed.
26610 * brig/lang-specs.h: Removed.
26611 * brig/lang.opt: Removed.
26613 2021-05-11 Richard Biener <rguenther@suse.de>
26616 * ipa-param-manipulation.c
26617 (ipa_param_body_adjustments::modify_call_stmt): Avoid
26618 altering SSA_NAME_DEF_STMT by adjusting the calls LHS
26619 via gimple_call_lhs_ptr.
26621 2021-05-11 Alex Coplan <alex.coplan@arm.com>
26624 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear):
26625 Avoid emitting CFA adjusts on the sp if we have the fp.
26627 2021-05-11 Richard Sandiford <richard.sandiford@arm.com>
26629 * config/aarch64/iterators.md (VMUL_CHANGE_NLANES): Delete.
26630 (VMULD): New iterator.
26631 (VCOND): Handle V4HF and V8HF.
26632 (VCONQ): Fix entry for V2SF.
26633 * config/aarch64/aarch64-simd.md (mul_lane<mode>3): Use VMULD
26634 instead of VMUL. Use a 64-bit vector mode for the indexed operand.
26635 (*aarch64_mul3_elt_<vswap_width_name><mode>): Merge with...
26636 (mul_laneq<mode>3): ...this define_insn. Use VMUL instead of VDQSF.
26637 Use a 128-bit vector mode for the indexed operand. Use stype for
26638 the scheduling type.
26640 2021-05-11 Richard Biener <rguenther@suse.de>
26642 * gimple-fold.c (maybe_fold_reference): Only return
26643 is_gimple_min_invariant values.
26645 2021-05-11 Richard Biener <rguenther@suse.de>
26647 PR middle-end/100509
26648 * gimple-fold.c (fold_gimple_assign): Only call
26649 get_symbol_constant_value on register type symbols.
26651 2021-05-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
26652 Joe Ramsay <joe.ramsay@arm.com>
26655 * config/arm/arm_mve.h (__arm_vstrwq_scatter_offset): Fix wrong arguments.
26656 (__arm_vcmpneq): Remove duplicate definition.
26657 (__arm_vstrwq_scatter_offset_p): Likewise.
26658 (__arm_vmaxq_x): Likewise.
26659 (__arm_vmlsdavaq): Likewise.
26660 (__arm_vmlsdavaxq): Likewise.
26661 (__arm_vmlsdavq_p): Likewise.
26662 (__arm_vmlsdavxq_p): Likewise.
26663 (__arm_vrmlaldavhaq): Likewise.
26664 (__arm_vstrbq_p): Likewise.
26665 (__arm_vstrbq_scatter_offset): Likewise.
26666 (__arm_vstrbq_scatter_offset_p): Likewise.
26667 (__arm_vstrdq_scatter_offset): Likewise.
26668 (__arm_vstrdq_scatter_offset_p): Likewise.
26669 (__arm_vstrdq_scatter_shifted_offset): Likewise.
26670 (__arm_vstrdq_scatter_shifted_offset_p): Likewise.
26672 2021-05-11 Jakub Jelinek <jakub@redhat.com>
26674 PR middle-end/100471
26675 * omp-low.c (lower_omp_task_reductions): For OMP_TASKLOOP, if data
26676 is 0, bypass the reduction loop including
26677 GOMP_taskgroup_reduction_unregister call.
26679 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
26681 * config/rs6000/rs6000.c (struct rs6000_cost_data): New member
26682 costing_for_scalar.
26683 (rs6000_density_test): Early return if costing_for_scalar is true.
26684 (rs6000_init_cost): Init costing_for_scalar of rs6000_cost_data.
26686 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
26688 * doc/tm.texi: Regenerated.
26689 * target.def (init_cost): Add new parameter costing_for_scalar.
26690 * targhooks.c (default_init_cost): Adjust for new parameter.
26691 * targhooks.h (default_init_cost): Likewise.
26692 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
26693 (vect_compute_single_scalar_iteration_cost): Likewise.
26694 (vect_analyze_loop_2): Likewise.
26695 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
26696 (vect_bb_vectorization_profitable_p): Likewise.
26697 * tree-vectorizer.h (init_cost): Likewise.
26698 * config/aarch64/aarch64.c (aarch64_init_cost): Likewise.
26699 * config/i386/i386.c (ix86_init_cost): Likewise.
26700 * config/rs6000/rs6000.c (rs6000_init_cost): Likewise.
26702 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
26704 * config/rs6000/rs6000.c (rs6000_vect_nonmem): Renamed to
26705 vect_nonmem and moved into...
26706 (struct rs6000_cost_data): ...here.
26707 (rs6000_init_cost): Use vect_nonmem of cost_data instead.
26708 (rs6000_add_stmt_cost): Likewise.
26709 (rs6000_finish_cost): Likewise.
26711 2021-05-10 Eric Botcazou <ebotcazou@adacore.com>
26713 * range-op.cc (get_bool_state): Adjust head comment.
26714 (operator_not_equal::op1_range): Fix comment.
26715 (operator_bitwise_xor::op1_range): Remove call to gcc_unreachable.
26717 2021-05-10 Martin Sebor <msebor@redhat.com>
26719 PR middle-end/100425
26720 PR middle-end/100510
26721 * gimple-ssa-warn-alloca.c (pass_walloca::firast_time_p): Rename...
26722 (pass_walloca::xlimit_certain_p): ...to this.
26723 (pass_walloca::gate): Execute for any kind of handled warning.
26724 (pass_walloca::execute): Avoid issuing "maybe" and "unbounded"
26725 warnings when xlimit_certain_p is set.
26727 2021-05-10 Pat Haugen <pthaugen@linux.ibm.com>
26729 * config/rs6000/rs6000.c (rs6000_ira_change_pseudo_allocno_class):
26730 Return ALTIVEC_REGS if that is best_class.
26731 (rs6000_compute_pressure_classes): Add ALTIVEC_REGS.
26733 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
26735 * config/arm/arm.h (CPP_SPEC): Remove error message about
26738 2021-05-10 Martin Jambor <mjambor@suse.cz>
26740 * ipa-prop.h (IPA_NODE_REF): Removed.
26741 (IPA_NODE_REF_GET_CREATE): Likewise.
26742 (IPA_EDGE_REF): Likewise.
26743 (IPA_EDGE_REF_GET_CREATE): Likewise.
26744 (IS_VALID_JUMP_FUNC_INDEX): Likewise.
26745 * ipa-cp.c (print_all_lattices): Replaced IPA_NODE_REF with a direct
26746 use of ipa_node_params_sum.
26747 (ipcp_versionable_function_p): Likewise.
26748 (push_node_to_stack): Likewise.
26749 (pop_node_from_stack): Likewise.
26750 (set_single_call_flag): Replaced two IPA_NODE_REF with one single
26751 direct use of ipa_node_params_sum.
26752 (initialize_node_lattices): Replaced IPA_NODE_REF with a direct use of
26753 ipa_node_params_sum.
26754 (ipa_context_from_jfunc): Replaced IPA_EDGE_REF with a direct use of
26756 (ipcp_verify_propagated_values): Replaced IPA_NODE_REF with a direct
26757 use of ipa_node_params_sum.
26758 (self_recursively_generated_p): Likewise.
26759 (propagate_scalar_across_jump_function): Likewise.
26760 (propagate_context_across_jump_function): Replaced IPA_EDGE_REF with a
26761 direct use of ipa_edge_args_sum, moved the lookup after the early
26762 exit. Replaced IPA_NODE_REF with a direct use of ipa_node_params_sum.
26763 (propagate_bits_across_jump_function): Replaced IPA_NODE_REF with
26764 direct uses of ipa_node_params_sum.
26765 (propagate_vr_across_jump_function): Likewise.
26766 (propagate_aggregate_lattice): Likewise.
26767 (propagate_aggs_across_jump_function): Likewise.
26768 (propagate_constants_across_call): Likewise, also replaced
26769 IPA_EDGE_REF with a direct use of ipa_edge_args_sum.
26770 (good_cloning_opportunity_p): Replaced IPA_NODE_REF with a direct use
26771 of ipa_node_params_sum.
26772 (estimate_local_effects): Likewise.
26773 (add_all_node_vals_to_toposort): Likewise.
26774 (propagate_constants_topo): Likewise.
26775 (ipcp_propagate_stage): Likewise.
26776 (ipcp_discover_new_direct_edges): Likewise.
26777 (calls_same_node_or_its_all_contexts_clone_p): Likewise.
26778 (cgraph_edge_brings_value_p): Likewise (in both overloaded functions).
26779 (get_info_about_necessary_edges): Likewise.
26780 (want_remove_some_param_p): Likewise.
26781 (create_specialized_node): Likewise.
26782 (self_recursive_pass_through_p): Likewise.
26783 (self_recursive_agg_pass_through_p): Likewise.
26784 (find_more_scalar_values_for_callers_subset): Likewise and also
26785 replaced IPA_EDGE_REF with direct uses of ipa_edge_args_sum, in one
26786 case replacing two of those with a single query.
26787 (find_more_contexts_for_caller_subset): Likewise for the
26788 ipa_polymorphic_call_context overload.
26789 (intersect_aggregates_with_edge): Replaced IPA_EDGE_REF with a direct
26790 use of ipa_edge_args_sum. Replaced IPA_NODE_REF with direct uses of
26791 ipa_node_params_sum.
26792 (find_aggregate_values_for_callers_subset): Likewise, also reusing
26793 results of ipa_edge_args_sum->get.
26794 (cgraph_edge_brings_all_scalars_for_node): Replaced IPA_NODE_REF with
26795 direct uses of ipa_node_params_sum, replaced IPA_EDGE_REF with a
26796 direct use of ipa_edge_args_sum.
26797 (cgraph_edge_brings_all_agg_vals_for_node): Likewise, moved node
26798 summary query after the early exit and reused the result later.
26799 (decide_about_value): Replaced IPA_NODE_REF with a direct use of
26800 ipa_node_params_sum.
26801 (decide_whether_version_node): Likewise. Removed re-querying for
26802 summaries after cloning.
26803 (spread_undeadness): Replaced IPA_NODE_REF with a direct use of
26804 ipa_node_params_sum.
26805 (has_undead_caller_from_outside_scc_p): Likewise, reusing results of
26807 (identify_dead_nodes): Likewise.
26808 (ipcp_store_bits_results): Replaced IPA_NODE_REF with direct uses of
26809 ipa_node_params_sum.
26810 (ipcp_store_vr_results): Likewise.
26811 * ipa-fnsummary.c (evaluate_properties_for_edge): Likewise.
26812 (ipa_fn_summary_t::duplicate): Likewise.
26813 (analyze_function_body): Likewise.
26814 (estimate_calls_size_and_time): Likewise.
26815 (ipa_cached_call_context::duplicate_from): Likewise.
26816 (ipa_call_context::equal_to): Likewise.
26817 (remap_edge_params): Likewise.
26818 (ipa_merge_fn_summary_after_inlining): Likewise.
26819 (inline_read_section): Likewise.
26820 * ipa-icf.c (sem_function::param_used_p): Likewise.
26821 * ipa-modref.c (compute_parm_map): Likewise.
26822 (compute_parm_map): Replaced IPA_EDGE_REF with a direct use of
26824 (get_access_for_fnspec): Replaced IPA_NODE_REF with a direct use of
26825 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
26827 * ipa-profile.c (check_argument_count): Likewise.
26828 * ipa-prop.c (ipa_alloc_node_params): Replaced IPA_NODE_REF_GET_CREATE
26829 with a direct use of ipa_node_params_sum.
26830 (ipa_initialize_node_params): Likewise.
26831 (ipa_print_node_jump_functions_for_edge): Replaced IPA_EDGE_REF with a
26832 direct use of ipa_edge_args_sum and reused the query result.
26833 (ipa_compute_jump_functions_for_edge): Replaced IPA_NODE_REF with a
26834 direct use of ipa_node_params_sum and replaced IPA_EDGE_REF with a
26835 direct use of ipa_edge_args_sum.
26836 (ipa_note_param_call): Replaced IPA_NODE_REF with a direct use of
26837 ipa_node_params_sum and reused the result of the query.
26838 (ipa_analyze_node): Likewise.
26839 (ipa_analyze_controlled_uses): Replaced IPA_NODE_REF with a direct use
26840 of ipa_node_params_sum.
26841 (update_jump_functions_after_inlining): Replaced IPA_EDGE_REF with
26842 direct uses of ipa_edge_args_sum.
26843 (update_indirect_edges_after_inlining): Replaced IPA_NODE_REF with
26844 direct uses of ipa_node_params_sum and replaced IPA_EDGE_REF with a
26845 direct use of ipa_edge_args_sum. Removed superficial re-querying the
26847 (propagate_controlled_uses): Replaced IPA_NODE_REF with direct uses of
26848 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
26850 (ipa_propagate_indirect_call_infos): Replaced IPA_EDGE_REF with a
26851 direct use of ipa_edge_args_sum.
26852 (ipa_edge_args_sum_t::duplicate): Replaced IPA_NODE_REF with a direct
26853 use of ipa_node_params_sum.
26854 (ipa_print_node_params): Likewise.
26855 (ipa_write_node_info): Likewise and also replaced IPA_EDGE_REF with
26856 direct uses of ipa_edge_args_sum.
26857 (ipa_read_edge_info): Replaced IPA_EDGE_REF with a direct use of
26859 (ipa_read_node_info): Replaced IPA_NODE_REF with a direct use of
26860 ipa_node_params_sum.
26861 (ipa_prop_write_jump_functions): Likewise. Move variable node to the
26862 scopes where it is used.
26864 2021-05-10 Uroš Bizjak <ubizjak@gmail.com>
26866 * config/i386/i386-expand.c (ix86_expand_sse_movcc)
26867 <case E_V2SImode>: Force op_true to register.
26869 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
26871 * config/arm/iterators.md (MVE_FP_COMPARISONS): New.
26872 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_f<mode>)
26873 (mve_vcmp<mve_cmp_op>q_n_f<mode>): New, merge all vcmp_*f*
26875 (mve_vcmpeqq_f<mode>, mve_vcmpeqq_n_f<mode>, mve_vcmpgeq_f<mode>)
26876 (mve_vcmpgeq_n_f<mode>, mve_vcmpgtq_f<mode>)
26877 (mve_vcmpgtq_n_f<mode>, mve_vcmpleq_f<mode>)
26878 (mve_vcmpleq_n_f<mode>, mve_vcmpltq_f<mode>)
26879 (mve_vcmpltq_n_f<mode>, mve_vcmpneq_f<mode>)
26880 (mve_vcmpneq_n_f<mode>): Remove.
26881 * config/arm/unspecs.md (VCMPEQQ_F, VCMPEQQ_N_F, VCMPGEQ_F)
26882 (VCMPGEQ_N_F, VCMPGTQ_F, VCMPGTQ_N_F, VCMPLEQ_F, VCMPLEQ_N_F)
26883 (VCMPLTQ_F, VCMPLTQ_N_F, VCMPNEQ_F, VCMPNEQ_N_F): Remove.
26885 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
26887 * config/arm/iterators.md (MVE_COMPARISONS): New.
26889 (mve_cmp_type): New.
26890 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_<mode>): New, merge all
26892 (mve_vcmpneq_<mode>, mve_vcmpcsq_n_<mode>, mve_vcmpcsq_<mode>)
26893 (mve_vcmpeqq_n_<mode>, mve_vcmpeqq_<mode>, mve_vcmpgeq_n_<mode>)
26894 (mve_vcmpgeq_<mode>, mve_vcmpgtq_n_<mode>, mve_vcmpgtq_<mode>)
26895 (mve_vcmphiq_n_<mode>, mve_vcmphiq_<mode>, mve_vcmpleq_n_<mode>)
26896 (mve_vcmpleq_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
26897 (mve_vcmpneq_n_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
26898 (mve_vcmpneq_n_<mode>): Remove.
26900 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
26902 * config/arm/arm_mve.h (__arm_vcmp*): Remove 's' suffix.
26903 * config/arm/arm_mve_builtins.def (vcmp*): Remove 's' suffix.
26904 * config/arm/mve.md (mve_vcmp*): Remove 's' suffix in pattern
26907 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
26909 * config/arm/arm_mve_builtins.def (vcmpneq_u): Remove.
26910 (vcmpneq_n_u): Likewise.
26911 (vcmpeqq_u,): Likewise.
26912 (vcmpeqq_n_u): Likewise.
26913 * config/arm/iterators.md (supf): Remove VCMPNEQ_U, VCMPEQQ_U,
26914 VCMPEQQ_N_U and VCMPNEQ_N_U.
26915 * config/arm/mve.md (mve_vcmpneq): Remove <supf> iteration.
26916 (mve_vcmpeqq_n): Likewise.
26917 (mve_vcmpeqq): Likewise.
26918 (mve_vcmpneq_n): Likewise.
26920 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
26922 * config/arm/arm_mve.h (__arm_vcmpeq*u*, __arm_vcmpne*u*): Call
26923 the 's' version of the builtin.
26925 2021-05-10 Richard Biener <rguenther@suse.de>
26927 PR tree-optimization/100492
26928 * tree-loop-distribution.c (find_seed_stmts_for_distribution):
26929 Find nothing when the loop contains an irreducible region.
26931 2021-05-10 Richard Biener <rguenther@suse.de>
26933 PR middle-end/100464
26935 * gimple-fold.c (canonicalize_constructor_val): Do not set
26938 2021-05-10 Richard Biener <rguenther@suse.de>
26940 PR tree-optimization/100434
26941 * tree-ssa-dse.c (initialize_ao_ref_for_dse): Handle
26943 (dse_optimize_stmt): Handle call LHS by dropping the
26944 LHS or the whole call if it doesn't have other
26946 (pass_dse::execute): Adjust.
26948 2021-05-10 Martin Liska <mliska@suse.cz>
26950 * Makefile.in: Add missing genversion rule.
26952 2021-05-10 Alex Coplan <alex.coplan@arm.com>
26955 * config/arm/mve.md (*mve_mov<mode>): Simplify output code. Use
26956 vldrw.u32 and vstrw.32 for V2D[IF]mode loads and stores.
26958 2021-05-10 Martin Liska <mliska@suse.cz>
26960 * builtins.c (is_builtin_name): Use startswith
26961 function instead of strncmp.
26962 * collect2.c (main): Likewise.
26963 (has_lto_section): Likewise.
26964 (scan_libraries): Likewise.
26965 * coverage.c (coverage_checksum_string): Likewise.
26966 (coverage_init): Likewise.
26967 * dwarf2out.c (is_cxx): Likewise.
26968 (gen_compile_unit_die): Likewise.
26969 * gcc-ar.c (main): Likewise.
26970 * gcc.c (init_spec): Likewise.
26971 (read_specs): Likewise.
26972 (execute): Likewise.
26973 (check_live_switch): Likewise.
26974 * genattrtab.c (write_attr_case): Likewise.
26975 (IS_ATTR_GROUP): Likewise.
26976 * gencfn-macros.c (main): Likewise.
26977 * gengtype.c (type_for_name): Likewise.
26978 (gen_rtx_next): Likewise.
26979 (get_file_langdir): Likewise.
26980 (write_local): Likewise.
26981 * genmatch.c (get_operator): Likewise.
26982 (get_operand_type): Likewise.
26983 (expr::gen_transform): Likewise.
26984 * genoutput.c (validate_optab_operands): Likewise.
26985 * incpath.c (add_sysroot_to_chain): Likewise.
26986 * langhooks.c (lang_GNU_C): Likewise.
26987 (lang_GNU_CXX): Likewise.
26988 (lang_GNU_Fortran): Likewise.
26989 (lang_GNU_OBJC): Likewise.
26990 * lto-wrapper.c (run_gcc): Likewise.
26991 * omp-general.c (omp_max_simt_vf): Likewise.
26992 * omp-low.c (omp_runtime_api_call): Likewise.
26993 * opts-common.c (parse_options_from_collect_gcc_options): Likewise.
26994 * read-rtl-function.c (function_reader::read_rtx_operand_r): Likewise.
26995 * real.c (real_from_string): Likewise.
26996 * selftest.c (assert_str_startswith): Likewise.
26997 * timevar.c (timer::validate_phases): Likewise.
26998 * tree.c (get_file_function_name): Likewise.
26999 * ubsan.c (ubsan_use_new_style_p): Likewise.
27000 * varasm.c (default_function_rodata_section): Likewise.
27001 (incorporeal_function_p): Likewise.
27002 (default_section_type_flags): Likewise.
27003 * system.h (startswith): Define startswith.
27005 2021-05-10 Martin Liska <mliska@suse.cz>
27007 * bitmap.h (class auto_bitmap): Remove
27008 __cplusplus >= 201103.
27009 * config/aarch64/aarch64.c: Likewise.
27010 * gimple-ssa-store-merging.c (store_immediate_info::store_immediate_info):
27012 * sbitmap.h: Likewise.
27014 2021-05-10 Martin Liska <mliska@suse.cz>
27016 * Makefile.in: Rename gcov-iov to genversion and depend
27017 on version.h (instead of gcov-iov.h).
27018 * gcov-io.h: Include version.h instread of gcov-iov.h.
27019 * gengtype-state.c (read_state_version): Likewise.
27020 * gcov-iov.c: Moved to...
27021 * genversion.c: ...here.
27022 * lto-streamer.h (LTO_major_version): Define it with
27024 * version.c: Removed.
27025 * version.h: Removed.
27027 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
27029 * config/arc/arc.md (UNSPEC_ARC_DMPYWH): Define.
27030 * config/arc/simdext.md (VCT): Add predicates for iterator
27033 (voptab): Likewise.
27034 (vec_widen_<V_US>mult_hi_v4hi): Change pattern predicate.
27035 (<voptab>v2si3): New patterns.
27037 (reduc_plus_scal_v4hi): Likewise.
27038 (reduc_plus_scal_v2si): Likewise.
27039 (vec_duplicatev2si): Likewise.
27040 (vec_duplicatev4hi): Likewise.
27042 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
27044 * config/arc/simdext.md: Format and cleanup file.
27046 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
27048 * config/arc/simdext.md (movmisalignv2hi): Allow misaligned access
27049 only when munaligned-access option is on.
27050 (movmisalign<mode>): Likewise.
27052 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
27054 * common/config/arc/arc-common.c (arc_handle_option): Remove dot
27056 * config/arc/arc.c (arc_reorg): Remove underscore from string.
27058 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
27060 * config/arc/arc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
27061 (CTZ_DEFINED_VALUE_AT_ZERO): Likewise.
27062 * config/arc/arc.md (clrsbsi2): Cleanup pattern.
27063 (norm_f): Likewise.
27066 (clzsi2): Use fls instruction when available.
27067 (arc_clzsi2): Likewise.
27069 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
27071 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add r26 and r27.
27073 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
27075 * doc/extend.texi (__builtin_arc_sr): Swap arguments.
27077 2021-05-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
27079 PR middle-end/100467
27080 * toplev.c (compile_file): Call insn_locations_init before
27081 targetm.asm_out.code_end.
27083 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
27086 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
27088 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
27090 2021-05-07 Jakub Jelinek <jakub@redhat.com>
27091 Andrew Stubbs <amd@codesourcery.com>
27094 * builtins.c (try_store_by_multiple_pieces): Use force_operand for
27095 emit_move_insn operands.
27097 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
27099 * cfgexpand.c (expand_gimple_basic_block): Do not inherit a current
27100 location for the outgoing edges of an empty block.
27101 * dwarf2out.c (add_subscript_info): Retrieve the bounds and index
27102 type by means of the get_array_descr_info langhook, if it is set and
27103 returns true. Remove obsolete code dealing with unnamed subtypes.
27105 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
27107 * gimple-range-cache.cc (ssa_block_ranges): Virtualize.
27108 (sbr_vector): Renamed from ssa_block_cache.
27109 (sbr_vector::sbr_vector): Allocate from obstack abd initialize.
27110 (ssa_block_ranges::~ssa_block_ranges): Remove.
27111 (sbr_vector::set_bb_range): Use varying and undefined cached values.
27112 (ssa_block_ranges::set_bb_varying): Remove.
27113 (sbr_vector::get_bb_range): Adjust assert.
27114 (sbr_vector::bb_range_p): Adjust assert.
27115 (~block_range_cache): No freeing loop required.
27116 (block_range_cache::get_block_ranges): Remove.
27117 (block_range_cache::set_bb_range): Inline get_block_ranges.
27118 (block_range_cache::set_bb_varying): Remove.
27119 * gimple-range-cache.h (set_bb_varying): Remove prototype.
27120 * value-range.h (irange_allocator::get_memory): New.
27122 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
27124 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Search
27125 dominator tree is available and requested.
27126 (ranger_cache::ssa_range_in_bb): Don't search dom tree here.
27127 (ranger_cache::fill_block_cache): Don't search dom tree here either.
27128 * gimple-range-cache.h (non_null_deref_p): Add dom_search param.
27130 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
27132 * gimple-range.cc (gimple_ranger::range_on_exit): Handle block with
27133 only PHI nodes better.
27135 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
27137 * gimple-range-edge.h (gimple_outgoing_range): Rename from
27139 (gcond_edge_range): Export prototype.
27140 * gimple-range-edge.cc (gcond_edge_range): New.
27141 (gimple_outgoing_range::edge_range_p): Use gcond_edge_range.
27142 * gimple-range-gori.h (gori_compute): Use gimple_outgoing_range.
27144 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
27146 * gimple-range-edge.cc (outgoing_range::calc_switch_ranges): Compute
27147 default range into a temp and allocate only what is needed.
27149 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
27151 * range-op.cc (operator_trunc_mod::wi_fold): x % 0 is UNDEFINED.
27153 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
27155 * gimple-range.h (gimple_range_global): Pick up parameter initial
27156 values, and use-before defined locals are UNDEFINED.
27158 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
27160 * doc/extend.texi (scalar_storage_order): Mention effect on pointer
27162 * tree.h (reverse_storage_order_for_component_p): Return false if
27163 the type is a pointer.
27165 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
27167 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
27169 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
27172 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
27173 Handle V8QI, V4HI and V2SI modes.
27174 * config/i386/mmx.md (mmx_pblendvb): New insn pattern.
27175 * config/i386/sse.md (unspec): Move UNSPEC_BLENDV ...
27176 * config/i386/i386.md (unspec): ... here.
27178 2021-05-07 Tobias Burnus <tobias@codesourcery.com>
27179 Tom de Vries <tdevries@suse.de>
27181 * omp-low.c (lower_rec_simd_input_clauses): Set max_vf = 1 if
27182 a truth_value_p reduction variable is nonintegral.
27184 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
27187 * config/i386/i386-expand.c (ix86_use_mask_cmp_p):
27188 Return false for mode sizes < 16.
27190 2021-05-07 Jakub Jelinek <jakub@redhat.com>
27193 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn.
27195 2021-05-06 Martin Jambor <mjambor@suse.cz>
27197 * ipa-sra.c (ipa_sra_dump_all_summaries): Dump edge summaries even
27198 when there is no function summary.
27199 (ipa_sra_summarize_function): produce edge summaries even when
27202 2021-05-06 Tom Tromey <tom@tromey.com>
27204 * godump.c (string_hash_eq): Remove.
27205 (go_finish): Use htab_eq_string.
27207 2021-05-06 Tom Tromey <tom@tromey.com>
27209 * gengtype-state.c (read_state): Use htab_eq_string.
27210 (string_eq): Remove.
27212 2021-05-06 Tom Tromey <tom@tromey.com>
27214 * gensupport.c (htab_eq_string): Remove.
27216 2021-05-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
27219 * debug.h (gcc_debug_hooks): Add set_ignored_loc function pointer.
27220 * dwarf2out.h (dw_fde_node::ignored_debug): New data item.
27221 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Add dummy
27222 set_ignored_loc callbacks.
27223 * debug.c (do_nothing_debug_hooks): Likewise.
27224 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
27225 * dwarf2out.c (text_section_used, cold_text_section_used): Remove.
27226 (in_text_section_p, last_text_label, last_cold_label,
27227 switch_text_ranges, switch_cold_ranges): New data items.
27228 (dwarf2out_note_section_used): Remove.
27229 (dwarf2out_begin_prologue): Set fde->ignored_debug and
27231 (mark_ignored_debug_section): New helper function.
27232 (dwarf2out_end_epilogue, dwarf2out_switch_text_section): Call
27233 mark_ignored_debug_section.
27234 (dwarf2_debug_hooks): Use dwarf2out_set_ignored_loc.
27235 (dwarf2_lineno_debug_hooks): Use dummy for set_ignored_loc.
27236 (size_of_aranges): Adjust formula for multi-part text ranges size.
27237 (output_aranges): Output multi-part text ranges.
27238 (dwarf2out_set_ignored_loc): New callback function.
27239 (dwarf2out_finish): Output multi-part text ranges.
27240 (dwarf2out_c_finalize): Clear new data items.
27241 * final.c (final_start_function_1): Call set_ignored_loc callback.
27242 (final_scan_insn_1): Likewise.
27243 * ggc-page.c (gt_ggc_mx): New helper function.
27244 * stringpool.c (gt_pch_nx): Likewise.
27246 2021-05-06 Richard Biener <rguenther@suse.de>
27248 * timevar.def (TV_TREE_INSERT_PHI_NODES): Remove.
27249 (TV_TREE_SSA_REWRITE_BLOCKS): Likewise.
27250 (TV_TREE_INTO_SSA): New.
27251 * tree-into-ssa.c (insert_phi_nodes): Do not account separately.
27252 (rewrite_blocks): Likewise.
27253 (pass_data_build_ssa): Account to TV_TREE_INTO_SSA.
27255 2021-05-06 Jakub Jelinek <jakub@redhat.com>
27257 * tree-ssa-phiopt.c (value_replacement, minmax_replacement,
27258 abs_replacement, xor_replacement,
27259 cond_removal_in_popcount_clz_ctz_pattern,
27260 replace_phi_edge_with_variable): Change type of phi argument from
27261 gimple * to gphi *.
27263 2021-05-06 Richard Biener <rguenther@suse.de>
27265 * tree-ssa-loop-split.c (split_loop): Delay updating SSA form.
27266 Output an opt-info message.
27267 (do_split_loop_on_cond): Likewise.
27268 (tree_ssa_split_loops): Update SSA form here.
27270 2021-05-06 Richard Biener <rguenther@suse.de>
27272 * tree-inline.c (tree_function_versioning): Fix DECL_BY_REFERENCE
27273 return variable removal.
27275 2021-05-06 Marius Hillenbrand <mhillen@linux.ibm.com>
27277 * config/s390/s390-builtins.def (O_M5, O1_M5, ...): Remove unused macros.
27278 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
27279 (s390_vec_permi_dbl, s390_vpdi): Use the O3_U2 type for the immediate
27281 * config/s390/s390.c (s390_const_operand_ok): Remove unused
27284 2021-05-06 Jakub Jelinek <jakub@redhat.com>
27286 PR tree-optimization/94589
27287 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Call
27288 spaceship_replacement.
27289 (cond_only_block_p, spaceship_replacement): New functions.
27291 2021-05-06 Richard Biener <rguenther@suse.de>
27294 * tree-emutls.c (gen_emutls_addr): Pass in whether we're
27295 dealing with a debug use and only query existing addresses
27297 (lower_emutls_1): Avoid splitting out addresses for debug
27298 stmts, reset the debug stmt when we fail to find existing
27300 (lower_emutls_phi_arg): Set wi.stmt.
27302 2021-05-06 Christoph Muellner <cmuellner@gcc.gnu.org>
27305 * config/riscv/riscv.c (riscv_block_move_loop): Use cbranch helper.
27306 * config/riscv/riscv.md (cbranch<mode>4): Generate helpers.
27307 (stack_protect_test): Use cbranch helper.
27309 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
27312 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
27313 always return the establisher frame for __builtin_frame_address (0).
27315 2021-05-05 Ivan Sorokin <vanyacpp@gmail.com>
27318 * config/i386/i386-builtins.c (ix86_cpu_model_type_node): New.
27319 (ix86_cpu_model_var): Likewise.
27320 (ix86_cpu_features2_type_node): Likewise.
27321 (ix86_cpu_features2_var): Likewise.
27322 (fold_builtin_cpu): Cache __cpu_model and __cpu_features2 with
27325 2021-05-05 Martin Sebor <msebor@redhat.com>
27327 * passes.def (pass_warn_printf): Run after SSA.
27329 2021-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
27331 * config/arm/neon.md (neon_vtst_combine<mode>): New pattern.
27332 * config/arm/predicates.md (minus_one_operand): New predicate.
27334 2021-05-05 Jeff Law <jlaw@tachyum.com>
27336 * config/avr/avr.md: Remove references to CC_STATUS_INIT.
27338 2021-05-05 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
27340 PR rtl-optimization/100263
27341 * postreload.c (move2add_valid_value_p): Ensure register can
27344 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
27346 PR rtl-optimization/100411
27347 * cfgcleanup.c (try_crossjump_to_edge): Also skip end of prologue
27348 and beginning of function markers.
27350 2021-05-05 Jeff Law <jlaw@tachyum.com>
27352 * config/cr16/cr16.h (NOTICE_UPDATE_CC): Remove.
27353 * config/cr16/cr16.c (notice_update_cc): Remove.
27354 * config/cr16/cr16-protos.h (notice_update_cc): Remove.
27356 2021-05-05 Uroš Bizjak <ubizjak@gmail.com>
27359 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
27360 Handle V8QI, V4HI and V2SI modes.
27361 * config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
27362 (ix86_build_signbit_mask): Ditto.
27363 * config/i386/mmx.md (MMXMODE14): New mode iterator.
27364 (<smaxmin:code><MMXMODE14:mode>3): New expander.
27365 (*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
27366 (<umaxmin:code><MMXMODE24:mode>3): New expander.
27367 (*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
27368 (vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
27369 (vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
27370 (vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
27371 (vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
27372 (vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
27374 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
27376 * dwarf2out.c (loc_list_from_tree_1) <DECL>: During early DWARF, do
27377 not expand the VALUE_EXPR of variables put in the non-local frame.
27378 * gimplify.c (gimplify_type_sizes) <RECORD_TYPE>: If the type is not
27379 to be ignored for debug info, ensure its variable offsets are not.
27381 2021-05-05 Richard Biener <rguenther@suse.de>
27383 PR tree-optimization/79333
27384 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
27385 Fold stmt following SSA edges.
27387 2021-05-05 Richard Biener <rguenther@suse.de>
27389 PR middle-end/100394
27390 * calls.c (expand_call): Preserve possibly throwing calls.
27391 * cfgexpand.c (expand_call_stmt): When a call can throw signal
27392 RTL expansion there are side-effects.
27393 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Simplify,
27394 mark all possibly throwing stmts necessary unless we can elide
27396 * tree-ssa-dse.c (pass_dse::execute): Preserve exceptions unless
27397 -fdelete-dead-exceptions.
27398 * tree.h (DECL_PURE_P): Add note about exceptions.
27400 2021-05-05 Alexandre Oliva <oliva@adacore.com>
27402 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Make it
27405 2021-05-04 David Edelsohn <dje.gcc@gmail.com>
27407 * config/rs6000/rs6000-call.c (rs6000_output_mi_thunk): Use
27408 get_fnname_from_decl for name of thunk.
27409 * config/rs6000/rs6000.c (rs6000_declare_alias): Use assemble_name
27410 and ASM_OUTPUT_LABEL.
27411 (rs6000_xcoff_declare_function_name): Use assemble_name and
27413 (rs6000_xcoff_declare_object_name): Use ASM_OUTPUT_LABEL.
27414 (rs6000_xcoff_encode_section_info): Don't add mapping class
27415 for aliases. Always add [DS] mapping class to primary
27417 (rs6000_asm_weaken_decl): Don't explicitly add [DS].
27419 2021-05-04 Martin Sebor <msebor@redhat.com>
27421 PR middle-end/100307
27422 * builtins.c (compute_objsize_r): Clear base0 for pointers.
27424 2021-05-04 Jeff Law <jlaw@tachyum.com>
27426 * config/bfin/bfin.h (NOTICE_UPDATE_CC): Remove.
27428 2021-05-04 Segher Boessenkool <segher@kernel.crashing.org>
27430 * caller-save.c: Remove CC0.
27431 * cfgcleanup.c: Remove CC0.
27432 * cfgrtl.c: Remove CC0.
27433 * combine.c: Remove CC0.
27434 * compare-elim.c: Remove CC0.
27435 * conditions.h: Remove CC0.
27436 * config/h8300/h8300.h: Remove CC0.
27437 * config/h8300/h8300-protos.h: Remove CC0.
27438 * config/h8300/peepholes.md: Remove CC0.
27439 * config/i386/x86-tune-sched.c: Remove CC0.
27440 * config/m68k/m68k.c: Remove CC0.
27441 * config/rl78/rl78.c: Remove CC0.
27442 * config/sparc/sparc.c: Remove CC0.
27443 * config/xtensa/xtensa.c: Remove CC0.
27444 (gen_conditional_move): Use pc_rtx instead of cc0_rtx in a piece of
27445 RTL where that is used as a placeholder only.
27446 * cprop.c: Remove CC0.
27447 * cse.c: Remove CC0.
27448 * cselib.c: Remove CC0.
27449 * df-problems.c: Remove CC0.
27450 * df-scan.c: Remove CC0.
27451 * doc/md.texi: Remove CC0. Adjust an example.
27452 * doc/rtl.texi: Remove CC0. Adjust an example.
27453 * doc/tm.texi: Regenerate.
27454 * doc/tm.texi.in: Remove CC0.
27455 * emit-rtl.c: Remove CC0.
27456 * final.c: Remove CC0.
27457 * fwprop.c: Remove CC0.
27458 * gcse-common.c: Remove CC0.
27459 * gcse.c: Remove CC0.
27460 * genattrtab.c: Remove CC0.
27461 * genconfig.c: Remove CC0.
27462 * genemit.c: Remove CC0.
27463 * genextract.c: Remove CC0.
27464 * gengenrtl.c: Remove CC0.
27465 * genrecog.c: Remove CC0.
27466 * haifa-sched.c: Remove CC0.
27467 * ifcvt.c: Remove CC0.
27468 * ira-costs.c: Remove CC0.
27469 * ira.c: Remove CC0.
27470 * jump.c: Remove CC0.
27471 * loop-invariant.c: Remove CC0.
27472 * lra-constraints.c: Remove CC0.
27473 * lra-eliminations.c: Remove CC0.
27474 * optabs.c: Remove CC0.
27475 * postreload-gcse.c: Remove CC0.
27476 * postreload.c: Remove CC0.
27477 * print-rtl.c: Remove CC0.
27478 * read-rtl-function.c: Remove CC0.
27479 * reg-notes.def: Remove CC0.
27480 * reg-stack.c: Remove CC0.
27481 * reginfo.c: Remove CC0.
27482 * regrename.c: Remove CC0.
27483 * reload.c: Remove CC0.
27484 * reload1.c: Remove CC0.
27485 * reorg.c: Remove CC0.
27486 * resource.c: Remove CC0.
27487 * rtl.c: Remove CC0.
27488 * rtl.def: Remove CC0.
27489 * rtl.h: Remove CC0.
27490 * rtlanal.c: Remove CC0.
27491 * sched-deps.c: Remove CC0.
27492 * sched-rgn.c: Remove CC0.
27493 * shrink-wrap.c: Remove CC0.
27494 * simplify-rtx.c: Remove CC0.
27495 * system.h: Remove CC0. Poison NOTICE_UPDATE_CC, CC_STATUS_MDEP_INIT,
27496 CC_STATUS_MDEP, and CC_STATUS.
27497 * target.def: Remove CC0.
27498 * valtrack.c: Remove CC0.
27499 * var-tracking.c: Remove CC0.
27501 2021-05-04 Richard Biener <rguenther@suse.de>
27503 PR tree-optimization/100414
27504 * tree-ssa-phiopt.c (get_non_trapping): Do not compute dominance
27506 (tree_ssa_phiopt_worker): But unconditionally here.
27508 2021-05-04 Tobias Burnus <tobias@codesourcery.com>
27510 * omp-low.c (lower_rec_input_clauses, lower_reduction_clauses): Handle
27511 && and || with floating-point and complex arguments.
27513 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
27515 * tree-inline.c (insert_debug_decl_map): Delete.
27516 (copy_debug_stmt): Minor tweak.
27517 (setup_one_parameter): Do not use a variable if the value is either
27518 a read-only DECL or a non-addressable local variable in the caller.
27519 In this case, insert the debug-only variable in the map manually.
27520 (expand_call_inline): Do not generate a CLOBBER for these values.
27521 * tree-inline.h (debug_map): Minor tweak.
27523 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
27525 * builtins.c (builtin_with_linkage_p): Return true for stp[n]cpy.
27526 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Tidy up.
27528 2021-05-04 Richard Biener <rguenther@suse.de>
27530 PR tree-optimization/100329
27531 * tree-ssa-reassoc.c (can_reassociate_p): Do not reassociate
27533 (insert_stmt_after): Assert we're not running into asm goto.
27535 2021-05-04 Richard Biener <rguenther@suse.de>
27537 PR tree-optimization/100398
27538 * tree-ssa-dse.c (pass_dse::execute): Preserve control
27541 2021-05-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
27543 * builtins.c (try_store_by_multiple_pieces): Fix constfun's prototype.
27545 2021-05-04 Alexandre Oliva <oliva@adacore.com>
27547 * builtins.c (try_store_by_multiple_pieces): New.
27548 (expand_builtin_memset_args): Use it. If target_char_cast
27549 fails, proceed as for non-constant val. Pass len's ctz to...
27550 * expr.c (clear_storage_hints): ... this. Try store by
27551 multiple pieces after setmem.
27552 (clear_storage): Adjust.
27553 * expr.h (clear_storage_hints): Likewise.
27554 (try_store_by_multiple_pieces): Declare.
27555 * passes.def: Replace the last copy_prop with ccp.
27557 2021-05-03 Tom de Vries <tdevries@suse.de>
27560 * omp-low.c (lower_rec_input_clauses): Disable SIMT for user-defined
27563 2021-05-03 Richard Biener <rguenther@suse.de>
27565 * tree-ssa-dse.c (dse_classify_store): Track two PHI defs.
27567 2021-05-03 Richard Biener <rguenther@suse.de>
27569 * tree-ssa-dse.c: Do not include domwalk.h but cfganal.h.
27570 (dse_dom_walker): Remove.
27571 (dse_dom_walker::dse_optimize_stmt): Rename...
27572 (dse_optimize_stmt): ... to this, pass in live_bytes sbitmap.
27573 (dse_dom_walker::before_dom_children): Inline ...
27574 (pass_dse::execute): ... here. Perform a reverse program
27577 2021-05-03 H.J. Lu <hjl.tools@gmail.com>
27580 * configure: Regenerated.
27582 2021-05-03 Ilya Leoshkevich <iii@linux.ibm.com>
27585 * config/s390/s390.c (s390_hard_fp_reg_p): New function.
27586 (s390_md_asm_adjust): Handle hard registers.
27588 2021-05-03 Jakub Jelinek <jakub@redhat.com>
27590 PR tree-optimization/100382
27591 * tree-ssa-dse.c: Include tree-eh.h.
27592 (dse_dom_walker::before_dom_children): Don't remove stmts if
27593 stmt_unremovable_because_of_non_call_eh_p is true.
27595 2021-05-02 David Edelsohn <dje.gcc@gmail.com>
27597 * varasm.c (compute_reloc_for_var): Split out from...
27598 (get_variable_section): Use it.
27599 * output.h (compute_reloc_for_var): Declare.
27600 * config/rs6000/rs6000-protos.h
27601 (rs6000_xcoff_asm_output_aligned_decl_common): Change alignment to
27603 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address_aix):
27604 Don't append storage mapping class to symbol.
27605 (rs6000_xcoff_asm_named_section): Add BS and UL mapping classes.
27606 Don't convert TLS BSS to common.
27607 (rs6000_xcoff_unique_section): Don't fall back to select_secton.
27608 (rs6000_xcoff_section_type_flags): Add SECTION_BSS if DECL is
27610 (rs6000_xcoff_asm_globalize_decl_name): Don't strip storage
27612 (rs6000_xcoff_asm_output_aligned_decl_common): Align is unsigned int.
27613 If align is 0 from TLS class, use the same rules as varasm.c
27614 If not common, switch to BSS section manually.
27615 If common, emit appropriate comm or lcomm directive.
27616 (rs6000_xcoff_encode_section_info): Add logic to append all
27617 storage mapping classes.
27618 (rs6000_asm_weaken_decl): Adjust for qualname symbols.
27619 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
27620 rs6000_xcoff_asm_output_aligned_decl_common.
27621 (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
27622 rs6000_xcoff_asm_output_aligned_decl_common.
27623 (ASM_OUTPUT_TLS_COMMON): Use
27624 rs6000_xcoff_asm_output_aligned_decl_common.
27626 2021-05-02 Jakub Jelinek <jakub@redhat.com>
27629 * config/nvptx/nvptx.c (nvptx_sese_pseudo): Use nullptr instead of 0
27630 as first argument of pseudo_node_t constructors.
27632 2021-05-02 Jakub Jelinek <jakub@redhat.com>
27635 * config/i386/t-i386 (TM_H): Add $(srcdir)/config/i386/i386-isa.def.
27637 2021-05-01 Aldy Hernandez <aldyh@redhat.com>
27639 * value-range.cc (DEFINE_INT_RANGE_GC_STUBS): Remove.
27640 (gt_pch_nx (int_range<1> *&)): New.
27641 (gt_ggc_mx (int_range<1> *&)): New.
27642 * value-range.h (class irange): Add GTY support for
27645 2021-05-01 Geng Qi <gengqi@linux.alibaba.com>
27647 * doc/options.texi (Negative): Change either or to both and.
27649 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
27651 * config/aarch64/aarch64-simd-builtins.def: Add
27652 float_ml[as][q]_laneq builtin generator macros.
27653 * config/aarch64/aarch64-simd.md (mul_laneq<mode>3): Define.
27654 (aarch64_float_mla_laneq<mode>): Define.
27655 (aarch64_float_mls_laneq<mode>): Define.
27656 * config/aarch64/arm_neon.h (vmla_laneq_f32): Use RTL builtin
27657 instead of GCC vector extensions.
27658 (vmlaq_laneq_f32): Likewise.
27659 (vmls_laneq_f32): Likewise.
27660 (vmlsq_laneq_f32): Likewise.
27662 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
27664 * config/aarch64/aarch64-simd-builtins.def: Add
27665 float_ml[as]_lane builtin generator macros.
27666 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt<mode>):
27668 (mul_lane<mode>3): This, and re-order arguments.
27669 (aarch64_float_mla_lane<mode>): Define.
27670 (aarch64_float_mls_lane<mode>): Define.
27671 * config/aarch64/arm_neon.h (vmla_lane_f32): Use RTL builtin
27672 instead of GCC vector extensions.
27673 (vmlaq_lane_f32): Likewise.
27674 (vmls_lane_f32): Likewise.
27675 (vmlsq_lane_f32): Likewise.
27677 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
27679 * config/aarch64/aarch64-simd-builtins.def: Add float_ml[as]
27680 builtin generator macros.
27681 * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>):
27683 (aarch64_float_mls<mode>): Define.
27684 * config/aarch64/arm_neon.h (vmla_f32): Use RTL builtin
27685 instead of relying on GCC vector extensions.
27686 (vmla_f64): Likewise.
27687 (vmlaq_f32): Likewise.
27688 (vmlaq_f64): Likewise.
27689 (vmls_f32): Likewise.
27690 (vmls_f64): Likewise.
27691 (vmlsq_f32): Likewise.
27692 (vmlsq_f64): Likewise.
27693 * config/aarch64/iterators.md: Define VDQF_DF mode iterator.
27695 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
27697 * config/aarch64/aarch64-simd-builtins.def: Add
27698 float_ml[as]_n_builtin generator macros.
27699 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt_from_dup<mode>):
27701 (mul_n<mode>3): This, and re-order arguments.
27702 (aarch64_float_mla_n<mode>): Define.
27703 (aarch64_float_mls_n<mode>): Define.
27704 * config/aarch64/arm_neon.h (vmla_n_f32): Use RTL builtin
27705 instead of inline asm.
27706 (vmlaq_n_f32): Likewise.
27707 (vmls_n_f32): Likewise.
27708 (vmlsq_n_f32): Likewise.
27710 2021-04-30 Jonathan Wright <joanthan.wright@arm.com>
27712 * config/aarch64/aarch64-simd-builtins.def: Add pmull[2]
27713 builtin generator macros.
27714 * config/aarch64/aarch64-simd.md (aarch64_pmullv8qi): Define.
27715 (aarch64_pmull_hiv16qi_insn): Define.
27716 (aarch64_pmull_hiv16qi): Define.
27717 * config/aarch64/arm_neon.h (vmull_high_p8): Use RTL builtin
27718 instead of inline asm.
27719 (vmull_p8): Likewise.
27721 2021-04-30 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
27723 * config/avr/avr.md: Adjust peepholes to match and
27724 generate parallels with clobber of REG_CC.
27725 (mov<mode>_insn): Rename to mov<mode>_insn_split.
27726 (*mov<mode>_insn): Rename to mov<mode>_insn.
27728 2021-04-30 David Edelsohn <dje.gcc@gmail.com>
27730 * varasm.c (use_blocks_for_decl_p): Don't use section anchors
27731 for VAR_DECLs if -fdata-sections enabled.
27733 2021-04-30 Michael Meissner <meissner@linux.ibm.com>
27735 PR bootstrap/100327
27736 * config/rs6000/rs6000.c
27737 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Define.
27738 (rs6000_libgcc_floating_mode_supported_p): New target hook.
27740 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
27742 * tree-ssa-threadbackward.c (class thread_jumps): Split out code
27744 (class back_threader_registry): ...to here...
27745 (class back_threader_profitability): ...and here...
27746 (thread_jumps::thread_through_all_blocks): Remove argument.
27747 (back_threader_registry::back_threader_registry): New.
27748 (back_threader_registry::~back_threader_registry): New.
27749 (back_threader_registry::thread_through_all_blocks): New.
27750 (thread_jumps::profitable_jump_thread_path): Move from here...
27751 (back_threader_profitability::profitable_path_p): ...to here.
27752 (thread_jumps::find_taken_edge): New.
27753 (thread_jumps::convert_and_register_current_path): Move...
27754 (back_threader_registry::register_path): ...to here.
27755 (thread_jumps::register_jump_thread_path_if_profitable): Move...
27756 (thread_jumps::maybe_register_path): ...to here.
27757 (thread_jumps::handle_phi): Call find_taken_edge and
27758 maybe_register_path.
27759 (thread_jumps::handle_assignment): Same.
27760 (thread_jumps::fsm_find_control_statement_thread_paths): Remove
27761 tree argument to handle_phi and handle_assignment.
27762 (thread_jumps::find_jump_threads_backwards): Set m_name. Remove
27763 set of m_speed_p and m_max_threaded_paths.
27764 (pass_thread_jumps::execute): Remove second argument from
27765 find_jump_threads_backwards.
27766 (pass_early_thread_jumps::execute): Same.
27768 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
27770 * tree-ssa-dom.c (class dom_jump_threader_simplifier): New.
27771 (class dom_opt_dom_walker): Initialize some class variables.
27772 (pass_dominator::execute): Pass evrp_range_analyzer and
27773 dom_jump_threader_simplifier to dom_opt_dom_walker.
27774 Adjust for some functions moving into classes.
27775 (simplify_stmt_for_jump_threading): Adjust and move to...
27776 (jump_threader_simplifier::simplify): ...here.
27777 (dom_opt_dom_walker::before_dom_children): Adjust for
27778 m_evrp_range_analyzer.
27779 (dom_opt_dom_walker::after_dom_children): Remove x_vr_values hack.
27780 (test_for_singularity): Place in dom_opt_dom_walker class.
27781 (dom_opt_dom_walker::optimize_stmt): The argument
27782 evrp_range_analyzer is now a class field.
27783 * tree-ssa-threadbackward.c (class thread_jumps): Add m_registry.
27784 (thread_jumps::thread_through_all_blocks): New.
27785 (thread_jumps::convert_and_register_current_path): Use m_registry.
27786 (pass_thread_jumps::execute): Adjust for thread_through_all_blocks
27787 being in the threader class.
27788 (pass_early_thread_jumps::execute): Same.
27789 * tree-ssa-threadedge.c (threadedge_initialize_values): Move...
27790 (jump_threader::jump_threader): ...here.
27791 (threadedge_finalize_values): Move...
27792 (jump_threader::~jump_threader): ...here.
27793 (jump_threader::remove_jump_threads_including): New.
27794 (jump_threader::thread_through_all_blocks): New.
27795 (record_temporary_equivalences_from_phis): Move...
27796 (jump_threader::record_temporary_equivalences_from_phis): ...here.
27797 (record_temporary_equivalences_from_stmts_at_dest): Move...
27798 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
27800 (simplify_control_stmt_condition_1): Move to jump_threader class.
27801 (simplify_control_stmt_condition): Move...
27802 (jump_threader::simplify_control_stmt_condition): ...here.
27803 (thread_around_empty_blocks): Move...
27804 (jump_threader::thread_around_empty_blocks): ...here.
27805 (thread_through_normal_block): Move...
27806 (jump_threader::thread_through_normal_block): ...here.
27807 (thread_across_edge): Move...
27808 (jump_threader::thread_across_edge): ...here.
27809 (thread_outgoing_edges): Move...
27810 (jump_threader::thread_outgoing_edges): ...here.
27811 * tree-ssa-threadedge.h: Move externally facing functings...
27812 (class jump_threader): ...here...
27813 (class jump_threader_simplifier): ...and here.
27814 * tree-ssa-threadupdate.c (struct redirection_data): Remove comment.
27815 (jump_thread_path_allocator::jump_thread_path_allocator): New.
27816 (jump_thread_path_allocator::~jump_thread_path_allocator): New.
27817 (jump_thread_path_allocator::allocate_thread_edge): New.
27818 (jump_thread_path_allocator::allocate_thread_path): New.
27819 (jump_thread_path_registry::jump_thread_path_registry): New.
27820 (jump_thread_path_registry::~jump_thread_path_registry): New.
27821 (jump_thread_path_registry::allocate_thread_edge): New.
27822 (jump_thread_path_registry::allocate_thread_path): New.
27823 (dump_jump_thread_path): Make extern.
27824 (debug (const vec<jump_thread_edge *> &path)): New.
27825 (struct removed_edges): Move to tree-ssa-threadupdate.h.
27826 (struct thread_stats_d): Remove.
27827 (remove_ctrl_stmt_and_useless_edges): Make static.
27828 (lookup_redirection_data): Move...
27829 (jump_thread_path_registry::lookup_redirection_data): ...here.
27830 (ssa_redirect_edges): Make static.
27831 (thread_block_1): Move...
27832 (jump_thread_path_registry::thread_block_1): ...here.
27833 (thread_block): Move...
27834 (jump_thread_path_registry::thread_block): ...here.
27835 (thread_through_loop_header): Move...
27836 (jump_thread_path_registry::thread_through_loop_header): ...here.
27837 (mark_threaded_blocks): Move...
27838 (jump_thread_path_registry::mark_threaded_blocks): ...here.
27839 (debug_path): Move...
27840 (jump_thread_path_registry::debug_path): ...here.
27841 (debug_all_paths): Move...
27842 (jump_thread_path_registry::dump): ..here.
27843 (rewire_first_differing_edge): Move...
27844 (jump_thread_path_registry::rewire_first_differing_edge): ...here.
27845 (adjust_paths_after_duplication): Move...
27846 (jump_thread_path_registry::adjust_paths_after_duplication): ...here.
27847 (duplicate_thread_path): Move...
27848 (jump_thread_path_registry::duplicate_thread_path): ..here.
27849 (remove_jump_threads_including): Move...
27850 (jump_thread_path_registry::remove_jump_threads_including): ...here.
27851 (thread_through_all_blocks): Move to...
27852 (jump_thread_path_registry::thread_through_all_blocks): ...here.
27853 (delete_jump_thread_path): Remove.
27854 (register_jump_thread): Move...
27855 (jump_thread_path_registry::register_jump_thread): ...here.
27856 * tree-ssa-threadupdate.h: Move externally facing functions...
27857 (class jump_thread_path_allocator): ...here...
27858 (class jump_thread_path_registry): ...and here.
27859 (thread_through_all_blocks): Remove.
27860 (struct removed_edges): New.
27861 (register_jump_thread): Remove.
27862 (remove_jump_threads_including): Remove.
27863 (delete_jump_thread_path): Remove.
27864 (remove_ctrl_stmt_and_useless_edges): Remove.
27865 (free_dom_edge_info): New prototype.
27866 * tree-vrp.c: Remove x_vr_values hack.
27867 (class vrp_jump_threader_simplifier): New.
27868 (vrp_jump_threader_simplifier::simplify): New.
27869 (vrp_jump_threader::vrp_jump_threader): Adjust method signature.
27870 Remove m_dummy_cond.
27871 Instantiate m_simplifier and m_threader.
27872 (vrp_jump_threader::thread_through_all_blocks): New.
27873 (vrp_jump_threader::simplify_stmt): Remove.
27874 (vrp_jump_threader::after_dom_children): Do not set m_dummy_cond.
27875 Remove x_vr_values hack.
27876 (execute_vrp): Adjust for thread_through_all_blocks being in a
27879 2021-04-30 Christophe Lyon <christophe.lyon@linaro.org>
27881 * genflags.c (gen_insn): Print failed expansion string.
27883 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
27885 * expr.c (alignment_for_piecewise_move): Call mode_for_size
27886 without limit to MAX_FIXED_MODE_SIZE.
27888 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
27890 PR middle-end/90773
27891 * builtins.c (builtin_memset_gen_str): Don't use return from
27892 simplify_gen_subreg.
27894 2021-04-30 Uroš Bizjak <ubizjak@gmail.com>
27897 * config/i386/i386.md (*add<mode>3_carry_0r): New insn pattern.
27898 (*addsi3_carry_zext_0r): Ditto.
27899 (*sub<mode>3_carry_0): Ditto.
27900 (*subsi3_carry_zext_0r): Ditto.
27901 * config/i386/predicates.md (ix86_carry_flag_unset_operator):
27903 * config/i386/i386.c (ix86_rtx_costs) <case PLUS, case MINUS>:
27904 Also consider ix86_carry_flag_unset_operator to calculate
27905 the cost of adc/sbb insn.
27907 2021-04-30 Roman Zhuykov <zhroma@ispras.ru>
27909 PR rtl-optimization/100225
27910 PR rtl-optimization/84878
27911 * modulo-sched.c (sms_schedule): Use note_stores to skip loops
27912 where we have an instruction which touches (writes) any hard
27913 register from df->regular_block_artificial_uses set.
27914 Allow not-single-set instruction only right before basic block
27917 2021-04-30 Geng Qi <gengqi@linux.alibaba.com>
27919 * config/riscv/riscv.opt (march=,mabi=): Negative itself.
27921 2021-04-30 LevyHsu <admin@levyhsu.com>
27923 * config/riscv/riscv.c (riscv_min_arithmetic_precision): New.
27924 * config/riscv/riscv.h (TARGET_MIN_ARITHMETIC_PRECISION): New.
27925 * config/riscv/riscv.md (addv<mode>4, uaddv<mode>4): New.
27926 (subv<mode>4, usubv<mode>4, mulv<mode>4, umulv<mode>4): New.
27928 2021-04-29 Alexandre Oliva <oliva@adacore.com>
27930 * config.gcc: Merged x86 and x86_64 cpu_type-setting cases.
27932 2021-04-29 Alexandre Oliva <oliva@adacore.com>
27934 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Rename to...
27935 (ASM_OUTPUT_MAX_SKIP_ALIGN): ... this. Enclose in do/while(0).
27936 * config/i386/i386.c: Adjust.
27937 * config/i386/i386.md: Adjust.
27938 * config/i386/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Drop.
27939 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27940 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27941 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27942 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27943 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27944 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27945 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27946 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27947 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
27948 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
27950 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
27952 * config/i386/i386-expand.c (ix86_expand_int_compare):
27953 Swap operands of GTU and LEU comparison to emit carry flag comparison.
27954 * config/i386/i386.md (*add<mode>3_carry_0): Change insn
27955 predicate to allow more combine opportunities with memory operands.
27956 (*sub<mode>3_carry_0): Ditto.
27958 2021-04-29 Richard Sandiford <richard.sandiford@arm.com>
27960 PR rtl-optimization/100303
27961 * rtl-ssa/accesses.cc (function_info::make_use_available): Take a
27962 boolean that indicates whether the use will only be used in
27963 debug instructions. Treat it in the same way that existing
27964 cross-EBB debug references would be handled if so.
27965 (function_info::make_uses_available): Likewise.
27966 * rtl-ssa/functions.h (function_info::make_uses_available): Update
27967 prototype accordingly.
27968 (function_info::make_uses_available): Likewise.
27969 * fwprop.c (try_fwprop_subst): Update call accordingly.
27971 2021-04-29 Jeff Law <jlaw@tachyum.com>
27973 * config/nios2/nios2-protos.h (nios2_fpu_insn_enabled): Move outside
27976 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
27977 Richard Biener <rguenther@suse.de>
27980 * config/i386/i386-builtin.def (IX86_BUILTIN_MASKLOADPD)
27981 (IX86_BUILTIN_MASKLOADPS, IX86_BUILTIN_MASKLOADPD256)
27982 (IX86_BUILTIN_MASKLOADPS256, IX86_BUILTIN_MASKLOADD)
27983 (IX86_BUILTIN_MASKLOADQ, IX86_BUILTIN_MASKLOADD256)
27984 (IX86_BUILTIN_MASKLOADQ256): Move from SPECIAL_ARGS
27985 to PURE_ARGS category.
27986 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
27987 Handle PURE_ARGS category.
27988 * config/i386/i386-expand.c (ix86_expand_builtin): Ditto.
27990 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
27992 * configure.ac: Check for the presence of sys/locking.h header and
27993 for whether _LK_LOCK is supported by _locking.
27994 * configure: Regenerate.
27995 * config.in: Likewise.
27996 * gcov-io.h: Define GCOV_LOCKED_WITH_LOCKING if HOST_HAS_LK_LOCK.
27997 * gcov-io.c (gcov_open): Add support for GCOV_LOCKED_WITH_LOCKING.
27998 * system.h: Include <sys/locking.h> if HAVE_SYS_LOCKING_H.
28000 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
28002 * config/i386/predicates.md (fcmov_comparison_operator):
28003 Do not check for trivial FP comparison operator.
28004 <case GEU, case LTU>: Allow CCGZmode.
28005 <case GTU, case LEU>: Do not allow CCCmode.
28006 (ix86_comparison_operator) <case GTU, case LEU>: Allow only CCmode.
28007 (ix86_carry_flag_operator): Match only LTU and UNLT code.
28008 Do not check for trivial FP comparison operator. Allow CCGZmode.
28010 2021-04-29 Tom de Vries <tdevries@suse.de>
28012 * omp-expand.c (expand_omp_simd): Add step_orig, and replace uses of
28013 fd->loop.step by either step or orig_step.
28015 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
28017 * config/sparc/sparc.c (gen_load_pcrel_sym): Delete.
28018 (load_got_register): Do the PIC dance here.
28019 (sparc_legitimize_tls_address): Simplify.
28020 (sparc_emit_probe_stack_range): Likewise.
28021 (sparc32_initialize_trampoline): Likewise.
28022 (sparc64_initialize_trampoline): Likewise.
28023 * config/sparc/sparc.md (load_pcrel_sym<P:mode>): Add @ marker.
28024 (probe_stack_range<P:mode>): Likewise.
28025 (flush<P:mode>): Likewise.
28026 (tgd_hi22<P:mode>): Likewise.
28027 (tgd_lo10<P:mode>): Likewise.
28028 (tgd_add<P:mode>): Likewise.
28029 (tgd_call<P:mode>): Likewise.
28030 (tldm_hi22<P:mode>): Likewise.
28031 (tldm_lo10<P:mode>): Likewise.
28032 (tldm_add<P:mode>): Likewise.
28033 (tldm_call<P:mode>): Likewise.
28034 (tldo_hix22<P:mode>): Likewise.
28035 (tldo_lox10<P:mode>): Likewise.
28036 (tldo_add<P:mode>): Likewise.
28037 (tie_hi22<P:mode>): Likewise.
28038 (tie_lo10<P:mode>): Likewise.
28039 (tie_add<P:mode>): Likewise.
28040 (tle_hix22<P:mode>): Likewise.
28041 (tle_lox10<P:mode>): Likewise.
28042 (stack_protect_setsi): Rename to...
28043 (stack_protect_set32): ...this.
28044 (stack_protect_setdi): Rename to...
28045 (stack_protect_set64): ...this.
28046 (stack_protect_set): Adjust calls to above.
28047 (stack_protect_testsi): Rename to...
28048 (stack_protect_test32): ...this.
28049 (stack_protect_testdi): Rename to...
28050 (stack_protect_test64): ...this.
28051 (stack_protect_test): Adjust calls to above.
28053 2021-04-29 H.J. Lu <hjl.tools@gmail.com>
28055 PR middle-end/90773
28056 * builtins.c (builtin_memcpy_read_str): Add a dummy argument.
28057 (builtin_strncpy_read_str): Likewise.
28058 (builtin_memset_read_str): Add an argument for the previous RTL
28059 information and generate the new RTL from the previous RTL info.
28060 (builtin_memset_gen_str): Likewise.
28061 * builtins.h (builtin_strncpy_read_str): Update the prototype.
28062 (builtin_memset_read_str): Likewise.
28063 * expr.c (by_pieces_ninsns): If targetm.overlap_op_by_pieces_p()
28064 returns true, round up size and alignment to the widest integer
28065 mode for maximum size.
28066 (pieces_addr::adjust): Add a pointer to by_pieces_prev argument
28067 and pass it to m_constfn.
28068 (op_by_pieces_d): Add m_push and m_overlap_op_by_pieces.
28069 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
28070 initialize m_push. Initialize m_overlap_op_by_pieces with
28071 targetm.overlap_op_by_pieces_p ().
28072 (op_by_pieces_d::run): Pass the previous RTL information to
28073 pieces_addr::adjust and generate overlapping operations if
28074 m_overlap_op_by_pieces is true.
28076 (move_by_pieces_d::move_by_pieces_d): Updated for op_by_pieces_d
28078 (store_by_pieces_d::store_by_pieces_d): Updated for op_by_pieces_d
28080 (can_store_by_pieces): Use by_pieces_constfn on constfun.
28081 (store_by_pieces): Use by_pieces_constfn on constfun. Updated
28082 for op_by_pieces_d change.
28083 (clear_by_pieces_1): Add a dummy argument.
28084 (clear_by_pieces): Updated for op_by_pieces_d change.
28085 (compare_by_pieces_d::compare_by_pieces_d): Likewise.
28086 (string_cst_read_str): Add a dummy argument.
28087 * expr.h (by_pieces_constfn): Add a dummy argument.
28088 (by_pieces_prev): New.
28089 * target.def (overlap_op_by_pieces_p): New target hook.
28090 * config/i386/i386.c (TARGET_OVERLAP_OP_BY_PIECES_P): New.
28091 * doc/tm.texi.in: Add TARGET_OVERLAP_OP_BY_PIECES_P.
28092 * doc/tm.texi: Regenerated.
28094 2021-04-29 Richard Biener <rguenther@suse.de>
28096 PR tree-optimization/100253
28097 * tree-vect-stmts.c (vectorizable_load): Do not assume
28098 element alignment when DR_MISALIGNMENT is -1.
28099 (vectorizable_store): Likewise.
28101 2021-04-29 Jakub Jelinek <jakub@redhat.com>
28104 * config/aarch64/aarch64.c (aarch64_add_offset_1_temporaries): Use
28105 absu_hwi instead of abs_hwi.
28107 2021-04-29 Richard Biener <rguenther@suse.de>
28109 PR middle-end/38474
28110 * tree-ssa-structalias.c (add_graph_edge): Avoid direct
28111 forwarding when indirect forwarding through ESCAPED
28114 2021-04-29 Tom de Vries <tdevries@suse.de>
28117 * internal-fn.c (expand_GOMP_SIMT_ENTER_ALLOC)
28118 (expand_GOMP_SIMT_LAST_LANE, expand_GOMP_SIMT_ORDERED_PRED)
28119 (expand_GOMP_SIMT_VOTE_ANY, expand_GOMP_SIMT_XCHG_BFLY)
28120 (expand_GOMP_SIMT_XCHG_IDX): Ensure target is assigned to.
28122 2021-04-29 Richard Biener <rguenther@suse.de>
28124 PR tree-optimization/99912
28125 * tree-ssa-dse.c (dse_dom_walker::m_need_cfg_cleanup): New.
28126 (dse_dom_walker::todo): Likewise.
28127 (dse_dom_walker::dse_optimize_stmt): Move VDEF check to the
28129 (dse_dom_walker::before_dom_children): Remove trivially
28130 dead SSA defs and schedule CFG cleanup if we removed all
28132 (pass_dse::execute): Get TODO as computed by the DOM walker
28133 and return it. Wipe dominator info earlier.
28135 2021-04-29 Richard Biener <rguenther@suse.de>
28138 * ipa-prop.c (ipcp_modif_dom_walker::before_dom_children):
28139 Track blocks to cleanup EH in new m_need_eh_cleanup.
28140 (ipcp_modif_dom_walker::cleanup_eh): New.
28141 (ipcp_transform_function): Release dominator info before
28144 2021-04-29 Martin Sebor <msebor@redhat.com>
28146 PR middle-end/100250
28147 * attribs.c (attr_access::array_as_string): Avoid dereferencing
28148 a pointer when it's null.
28150 2021-04-29 Martin Sebor <msebor@redhat.com>
28152 * Makefile.in (OBJS): Add ipa-free-lang-data.o.
28153 * ipa-free-lang-data.cc: New file.
28154 * tree.c: Move pass free_lang_data to file above.
28155 (build_array_type_1): Declare extern.
28156 * tree.h (build_array_type_1): Declare.
28158 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28160 * config/aarch64/aarch64-simd-builtins.def: Modify comment to
28161 make consistent with updated RTL pattern.
28162 * config/aarch64/aarch64-simd.md (aarch64_<sur>qmovn<mode>):
28163 Implement using ss_truncate and us_truncate rather than
28165 * config/aarch64/iterators.md: Remove redundant unspecs and
28166 iterator: UNSPEC_[SU]QXTN and SUQMOVN respectively.
28168 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28170 * config/aarch64/arm_acle.h (__attribute__): Make intrinsic
28171 attributes consistent with those defined in arm_neon.h.
28173 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28175 * config/aarch64/arm_fp16.h (__attribute__): Make intrinsic
28176 attributes consistent with those defined in arm_neon.h.
28178 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28180 * config/aarch64/aarch64-simd-builtins.def: Add
28181 float_trunc_rodd builtin generator macros.
28182 * config/aarch64/aarch64-simd.md (aarch64_float_trunc_rodd_df):
28184 (aarch64_float_trunc_rodd_lo_v2sf): Define.
28185 (aarch64_float_trunc_rodd_hi_v4sf_le): Define.
28186 (aarch64_float_trunc_rodd_hi_v4sf_be): Define.
28187 (aarch64_float_trunc_rodd_hi_v4sf): Define.
28188 * config/aarch64/arm_neon.h (vcvtx_f32_f64): Use RTL builtin
28189 instead of inline asm.
28190 (vcvtx_high_f32_f64): Likewise.
28191 (vcvtxd_f32_f64): Likewise.
28192 * config/aarch64/iterators.md: Add FCVTXN unspec.
28194 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28196 * config/aarch64/aarch64-simd-builtins.def: Add tbx1 builtin
28198 * config/aarch64/aarch64-simd.md (aarch64_tbx1<mode>):
28200 * config/aarch64/arm_neon.h (vqtbx1_s8): USE RTL builtin
28201 instead of inline asm.
28202 (vqtbx1_u8): Likewise.
28203 (vqtbx1_p8): Likewise.
28204 (vqtbx1q_s8): Likewise.
28205 (vqtbx1q_u8): Likewise.
28206 (vqtbx1q_p8): Likewise.
28207 (vtbx2_s8): Likewise.
28208 (vtbx2_u8): Likewise.
28209 (vtbx2_p8): Likewise.
28211 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28213 * config/aarch64/aarch64-simd-builtins.def: Add tbl1 builtin
28215 * config/aarch64/arm_neon.h (vqtbl1_p8): Use RTL builtin
28216 instead of inline asm.
28217 (vqtbl1_s8): Likewise.
28218 (vqtbl1_u8): Likewise.
28219 (vqtbl1q_p8): Likewise.
28220 (vqtbl1q_s8): Likewise.
28221 (vqtbl1q_u8): Likewise.
28222 (vtbl1_s8): Likewise.
28223 (vtbl1_u8): Likewise.
28224 (vtbl1_p8): Likewise.
28225 (vtbl2_s8): Likewise.
28226 (vtbl2_u8): Likewise.
28227 (vtbl2_p8): Likewise.
28229 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28231 * config/aarch64/aarch64-simd-builtins.def: Add polynomial
28232 ssri_n buitin generator macro.
28233 * config/aarch64/arm_neon.h (vsri_n_p8): Use RTL builtin
28234 instead of inline asm.
28235 (vsri_n_p16): Likewise.
28236 (vsri_n_p64): Likewise.
28237 (vsriq_n_p8): Likewise.
28238 (vsriq_n_p16): Likewise.
28239 (vsriq_n_p64): Likewise.
28241 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28243 * config/aarch64/aarch64-simd-builtins.def: Use VALLP mode
28244 iterator for polynomial ssli_n builtin generator macro.
28245 * config/aarch64/arm_neon.h (vsli_n_p8): Use RTL builtin
28246 instead of inline asm.
28247 (vsli_n_p16): Likewise.
28248 (vsliq_n_p8): Likewise.
28249 (vsliq_n_p16): Likewise.
28250 * config/aarch64/iterators.md: Define VALLP mode iterator.
28252 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28254 * config/aarch64/aarch64-simd-builtins.def: Use VDQV_L
28255 iterator to generate [su]adalp RTL builtins.
28256 * config/aarch64/aarch64-simd.md: Use VDQV_L iterator in
28257 [su]adalp RTL pattern.
28258 * config/aarch64/arm_neon.h (vpadal_s32): Use RTL builtin
28259 instead of inline asm.
28260 (vpadal_u32): Likewise.
28262 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28264 * config/aarch64/aarch64-simd-builtins.def: Add [su]addlp
28265 builtin generator macros.
28266 * config/aarch64/aarch64-simd.md (aarch64_<su>addlp<mode>):
28268 * config/aarch64/arm_neon.h (vpaddl_s8): Use RTL builtin
28269 instead of inline asm.
28270 (vpaddl_s16): Likewise.
28271 (vpaddl_s32): Likewise.
28272 (vpaddl_u8): Likewise.
28273 (vpaddl_u16): Likewise.
28274 (vpaddl_u32): Likewise.
28275 (vpaddlq_s8): Likewise.
28276 (vpaddlq_s16): Likewise.
28277 (vpaddlq_s32): Likewise.
28278 (vpaddlq_u8): Likewise.
28279 (vpaddlq_u16): Likewise.
28280 (vpaddlq_u32): Liwewise.
28281 * config/aarch64/iterators.md: Define [SU]ADDLP unspecs with
28282 appropriate attributes.
28284 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28286 * config/aarch64/aarch64-simd-builtins.def: Use VDQ_I iterator
28287 for aarch64_addp<mode> builtin macro generator.
28288 * config/aarch64/aarch64-simd.md: Use VDQ_I iterator in
28289 aarch64_addp<mode> RTL pattern.
28290 * config/aarch64/arm_neon.h (vpaddq_s8): Use RTL builtin
28291 instead of inline asm.
28292 (vpaddq_s16): Likewise.
28293 (vpaddq_s32): Likewise.
28294 (vpaddq_s64): Likewise.
28295 (vpaddq_u8): Likewise.
28296 (vpaddq_u16): Likewise.
28297 (vpaddq_u32): Likewise.
28298 (vpaddq_u64): Likewise.
28300 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
28302 * config/aarch64/aarch64-simd-builtins.def: Add sq[r]dmulh_n
28303 builtin generator macros.
28304 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_n<mode>):
28306 * config/aarch64/arm_neon.h (vqdmulh_n_s16): Use RTL builtin
28307 instead of inline asm.
28308 (vqdmulh_n_s32): Likewise.
28309 (vqdmulhq_n_s16): Likewise.
28310 (vqdmulhq_n_s32): Likewise.
28311 (vqrdmulh_n_s16): Likewise.
28312 (vqrdmulh_n_s32): Likewise.
28313 (vqrdmulhq_n_s16): Likewise.
28314 (vqrdmulhq_n_s32): Likewise.
28316 2021-04-28 Tobias Burnus <tobias@codesourcery.com>
28318 * doc/install.texi (--enable-offload-defaulted): Document.
28320 2021-04-28 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
28322 * config/avr/avr-dimode.md: Turn existing patterns into
28323 define_insn_and_split style patterns where the splitter
28324 adds a clobber of the condition code register. Drop "cc"
28325 attribute. Add new patterns to match output of
28327 * config/avr/avr-fixed.md: Likewise.
28328 * config/avr/avr.c (cc_reg_rtx): New.
28329 (avr_parallel_insn_from_insns): Adjust insn count
28330 for removal of set of cc0.
28331 (avr_is_casesi_sequence): Likewise.
28332 (avr_casei_sequence_check_operands): Likewise.
28333 (avr_optimize_casesi): Likewise. Also insert
28334 new insns after jump_insn.
28335 (avr_pass_casesi::avr_rest_of_handle_casesi): Adjust
28336 for removal of set of cc0.
28337 (avr_init_expanders): Initialize cc_reg_rtx.
28338 (avr_regno_reg_class): Handle REG_CC.
28339 (cond_string): Remove usage of CC_OVERFLOW_UNUSABLE.
28340 (avr_notice_update_cc): Remove function.
28341 (ret_cond_branch): Remove usage of CC_OVERFLOW_UNUSABLE.
28342 (compare_condition): Adjust for PARALLEL with
28344 (out_shift_with_cnt): Likewise.
28345 (ashlhi3_out): Likewise.
28346 (ashrhi3_out): Likewise.
28347 (lshrhi3_out): Likewise.
28348 (avr_class_max_nregs): Return single reg for REG_CC.
28349 (avr_compare_pattern): Check for REG_CC instead
28351 (avr_reorg_remove_redundant_compare): Likewise.
28352 (avr_reorg):Adjust for PARALLEL with REG_CC clobber.
28353 (avr_hard_regno_nregs): Return single reg for REG_CC.
28354 (avr_hard_regno_mode_ok): Allow only CCmode for REG_CC.
28355 (avr_md_asm_adjust): Clobber REG_CC.
28356 (TARGET_HARD_REGNO_NREGS): Define.
28357 (TARGET_CLASS_MAX_NREGS): Define.
28358 (TARGET_MD_ASM_ADJUST): Define.
28359 * config/avr/avr.h (FIRST_PSEUDO_REGISTER): Adjust
28361 (enum reg_class): Add CC_REG class.
28362 (NOTICE_UPDATE_CC): Remove.
28363 (CC_OVERFLOW_UNUSABLE): Remove.
28364 (CC_NO_CARRY): Remove.
28365 * config/avr/avr.md: Turn existing patterns into
28366 define_insn_and_split style patterns where the splitter
28367 adds a clobber of the condition code register. Drop "cc"
28368 attribute. Add new patterns to match output of
28370 (sez): Remove unused pattern.
28372 2021-04-28 Richard Earnshaw <rearnsha@arm.com>
28375 * config/arm/arm.c (arm_hard_regno_mode_ok): Only allow VPR to be
28378 2021-04-28 Richard Sandiford <richard.sandiford@arm.com>
28381 * config/aarch64/constraints.md (Utq): Require the address to
28382 be valid for both the element mode and for V2DImode.
28384 2021-04-28 Jakub Jelinek <jakub@redhat.com>
28385 Tobias Burnus <tobias@codesourcery.com>
28387 * configure.ac (OFFLOAD_DEFAULTED): AC_DEFINE if offload-defaulted.
28388 * gcc.c (process_command): New variable.
28389 (driver::maybe_putenv_OFFLOAD_TARGETS): If OFFLOAD_DEFAULTED,
28390 set it if -foffload is defaulted.
28391 * lto-wrapper.c (OFFLOAD_TARGET_DEFAULT_ENV): Define.
28392 (compile_offload_image): If OFFLOAD_DEFAULTED and
28393 OFFLOAD_TARGET_DEFAULT is in the environment, don't fail
28394 if corresponding mkoffload can't be found.
28395 (compile_images_for_offload_targets): Likewise. Free and clear
28396 offload_names if no valid offload is found.
28397 * config.in: Regenerate.
28398 * configure: Regenerate.
28400 2021-04-28 Richard Biener <rguenther@suse.de>
28402 PR tree-optimization/100292
28403 * tree-vect-generic.c (expand_vector_condition): Do not fold
28406 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
28408 * config/rs6000/aix.h (SUBTARGET_DRIVER_SELF_SPECS): New.
28409 * config/rs6000/aix64.opt (m64): New.
28412 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
28414 * config/vax/vax.c (print_operand_address, vax_address_cost_1)
28415 (index_term_p): Handle ASHIFT too.
28417 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
28419 * config/vax/builtins.md (jbb<ccss>i<mode>): Remove operand #3.
28420 (sync_lock_test_and_set<mode>): Adjust accordingly.
28421 (sync_lock_release<mode>): Likewise.
28423 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
28425 * config/vax/vax-protos.h (adjacent_operands_p): Remove
28427 * config/vax/vax.c (adjacent_operands_p): Remove.
28429 2021-04-27 Maciej W. Rozycki <macro@linux-mips.org>
28431 * ifcvt.c (dead_or_predicable) [!IFCVT_MODIFY_TESTS]: Fall
28432 through to the non-conditional execution case if getting the
28433 condition for conditional execution has failed.
28435 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
28437 PR middle-end/100284
28438 * gimple.c (gimple_could_trap_p_1): Remove VEC_COND_EXPR test.
28439 * tree-eh.c (operation_could_trap_p): Handle VEC_COND_EXPR rather
28440 than asserting on it.
28442 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
28444 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): Protect
28445 with TARGET_AIX_OS.
28447 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
28450 * calls.c (precompute_register_parameters): Additionally test
28451 targetm.precompute_tls_p to pre-compute argument.
28452 * config/rs6000/aix.h (TARGET_PRECOMPUTE_TLS_P): Define.
28453 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): New.
28454 * target.def (precompute_tls_p): New.
28455 * doc/tm.texi.in (TARGET_PRECOMPUTE_TLS_P): Add hook documentation.
28456 * doc/tm.texi: Regenerated.
28458 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28461 * config/aarch64/aarch64.c (aarch64_print_operand): Cast -UINTVAL
28462 back to HOST_WIDE_INT.
28464 2021-04-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
28467 * simplify-rtx.c (simplify_context::simplify_subreg): Check the
28468 memory alignment for the outer mode.
28470 2021-04-27 H.J. Lu <hjl.tools@gmail.com>
28472 PR middle-end/90773
28473 * expr.c (op_by_pieces_d::get_usable_mode): New member function.
28474 (op_by_pieces_d::run): Cange a while loop to a do-while loop.
28476 2021-04-27 Alex Coplan <alex.coplan@arm.com>
28479 * config/arm/arm.c (arm_split_compare_and_swap): Fix up codegen
28480 with negative immediates: ensure we expand cbranchsi4_scratch
28481 correctly and ensure we satisfy its constraints.
28482 * config/arm/sync.md
28483 (@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): Don't
28484 attempt to tie two output operands together with constraints;
28485 collapse two alternatives.
28486 (@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
28487 * config/arm/thumb1.md (cbranchsi4_neg_late): New.
28489 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28492 * config/aarch64/predicates.md (aarch64_sub_immediate,
28493 aarch64_plus_immediate): Use -UINTVAL instead of -INTVAL.
28494 * config/aarch64/aarch64.md (casesi, rotl<mode>3): Likewise.
28495 * config/aarch64/aarch64.c (aarch64_print_operand,
28496 aarch64_split_atomic_op, aarch64_expand_subvti): Likewise.
28498 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28500 PR tree-optimization/100239
28501 * tree-vect-generic.c (lower_vec_perm): Don't accept constant
28502 permutations with all indices from the first zero element as vec_shl.
28504 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28506 PR rtl-optimization/100254
28507 * cfgcleanup.c (outgoing_edges_match): Check REG_EH_REGION on
28508 last1 and last2 insns rather than BB_END (bb1) and BB_END (bb2) insns.
28510 2021-04-27 Richard Biener <rguenther@suse.de>
28512 PR tree-optimization/99912
28513 * passes.def: Add comment about new TODO_remove_unused_locals.
28514 * tree-stdarg.c (pass_data_stdarg): Run TODO_remove_unused_locals
28517 2021-04-27 Richard Biener <rguenther@suse.de>
28519 PR tree-optimization/99912
28520 * passes.def (pass_all_optimizations): Add pass_dse before
28521 the first pass_dce, move the first pass_dse before the
28522 pass_dce following pass_pre.
28524 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28526 PR tree-optimization/95527
28527 * generic-match-head.c: Include tm.h.
28528 * gimple-match-head.c: Include tm.h.
28529 * match.pd (CLZ == INTEGER_CST): Don't use
28530 #ifdef CLZ_DEFINED_VALUE_AT_ZERO, only test CLZ_DEFINED_VALUE_AT_ZERO
28531 if clz == CFN_CLZ. Add missing val declaration.
28532 (CTZ cmp CST): New simplifications.
28534 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28536 PR tree-optimization/96696
28537 * expr.c (expand_expr_divmod): New function.
28538 (expand_expr_real_2) <case TRUNC_DIV_EXPR>: Use it for truncations and
28539 divisions. Formatting fixes.
28540 <case MULT_EXPR>: Optimize x / y * y as x - x % y if the latter is
28543 2021-04-27 Martin Jambor <mjambor@suse.cz>
28546 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
28547 If removing a call statement LHS SSA name, release it.
28549 2021-04-27 Richard Earnshaw <rearnsha@arm.com>
28552 * config/arm/arm.c (THUMB2_WORK_REGS): Check PIC_OFFSET_TABLE_REGNUM
28553 is valid before including it in the mask.
28555 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
28558 * config/aarch64/aarch64.c (aarch64_comp_type_attributes): Handle
28561 2021-04-27 Richard Biener <rguenther@suse.de>
28563 PR tree-optimization/100051
28564 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Add
28565 disambiguator based on access size vs. decl size.
28567 2021-04-27 Richard Biener <rguenther@suse.de>
28569 PR tree-optimization/100278
28570 * tree-ssa-pre.c (compute_avail): Give up when we cannot
28571 adjust TBAA beacuse of mismatching bases.
28573 2021-04-27 Jakub Jelinek <jakub@redhat.com>
28576 * config/i386/i386.md (*<insn><mode>3_mask, *<insn><mode>3_mask_1):
28577 For any_rotate define_insn_split and following splitters, use
28578 SWI iterator instead of SWI48.
28580 2021-04-27 Richard Biener <rguenther@suse.de>
28582 PR tree-optimization/99776
28583 * match.pd (bit_field_ref (ctor)): Relax element extract
28584 type compatibility checks.
28586 2021-04-27 Cui,Lili <lili.cui@intel.com>
28588 * common/config/i386/i386-common.c (processor_names):
28589 Sync processor_names with processor_type.
28590 * config/i386/i386-options.c (processor_cost_table):
28591 Sync processor_cost_table with processor_type.
28593 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
28595 * value-range.cc (irange::irange_set_1bit_anti_range): Add assert.
28596 (irange::set): Call irange_set_1bit_anti_range for handling all
28597 1-bit ranges. Fall through on ~[MIN,MAX].
28599 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
28601 * value-range.cc (irange::legacy_num_pairs): Remove.
28602 (irange::invert): Change gcc_assert to gcc_checking_assert.
28603 * value-range.h (irange::num_pairs): Adjust for a cached
28604 num_pairs(). Also, rename all gcc_assert's to
28605 gcc_checking_assert's.
28607 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
28609 * value-range.cc (irange::operator=): Set m_kind.
28610 (irange::copy_to_legacy): Handle varying and undefined sources
28611 as a legacy copy since they can be easily copied.
28612 (irange::irange_set): Set m_kind.
28613 (irange::irange_set_anti_range): Same.
28614 (irange::set): Rename normalize_min_max to normalize_kind.
28615 (irange::verify_range): Adjust for multi-ranges having the
28617 (irange::irange_union): Set m_kind.
28618 (irange::irange_intersect): Same.
28619 (irange::invert): Same.
28620 * value-range.h (irange::kind): Always return m_kind.
28621 (irange::varying_p): Rename to...
28622 (irange::varying_comptaible_p): ...this.
28623 (irange::undefined_p): Only look at m_kind.
28624 (irange::irange): Always set VR_UNDEFINED if applicable.
28625 (irange::set_undefined): Always set VR_UNDEFINED.
28626 (irange::set_varying): Always set m_kind to VR_VARYING.
28627 (irange::normalize_min_max): Rename to...
28628 (irange::normalize_kind): ...this.
28630 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
28632 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
28633 Adjust for constant_p including varying_p.
28634 * tree-vrp.c (vrp_prop::finalize): Same.
28635 (determine_value_range): Same.
28636 * vr-values.c (vr_values::range_of_expr): Same.
28637 * value-range.cc (irange::symbolic_p): Do not check varying_p.
28638 (irange::constant_p): Same.
28640 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
28642 * value-range.cc (irange::legacy_lower_bound): Replace
28643 !undefined_p check with num_ranges > 0.
28644 (irange::legacy_upper_bound): Same.
28645 * value-range.h (irange::type): Same.
28646 (irange::lower_bound): Same.
28647 (irange::upper_bound): Same.
28649 2021-04-26 Richard Biener <rguenther@suse.de>
28651 PR tree-optimization/99956
28652 * gimple-loop-interchange.cc (compute_access_stride):
28653 Try instantiating the access in a shallower loop nest
28654 if instantiating failed.
28655 (compute_access_strides): Pass adjustable loop_nest
28656 to compute_access_stride.
28658 2021-04-26 Christophe Lyon <christophe.lyon@linaro.org>
28660 * doc/sourcebuild.texi (arm_cmse_hw): Document.
28662 2021-04-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28664 * config/aarch64/iterators.md (vwcore): Handle V4BF, V8BF.
28666 2021-04-26 Thomas Schwinge <thomas@codesourcery.com>
28667 Nathan Sidwell <nathan@codesourcery.com>
28668 Tom de Vries <vries@codesourcery.com>
28669 Julian Brown <julian@codesourcery.com>
28670 Kwok Cheung Yeung <kcy@codesourcery.com>
28672 * omp-offload.c (oacc_validate_dims): Implement
28673 '-Wopenacc-parallelism'.
28674 * doc/invoke.texi (-Wopenacc-parallelism): Document.
28676 2021-04-26 Richard Biener <rguenther@suse.de>
28678 * tree-cfg.h (gimplify_build1): Remove.
28679 (gimplify_build2): Likewise.
28680 (gimplify_build3): Likewise.
28681 * tree-cfg.c (gimplify_build1): Move to tree-vect-generic.c.
28682 (gimplify_build2): Likewise.
28683 (gimplify_build3): Likewise.
28684 * tree-vect-generic.c (gimplify_build1): Move from tree-cfg.c.
28686 (gimplify_build2): Likewise.
28687 (gimplify_build3): Likewise.
28688 (tree_vec_extract): Use resimplify with following SSA edges.
28689 (expand_vector_parallel): Avoid passing NULL size/bitpos
28690 to tree_vec_extract.
28691 * expr.c (store_constructor): Deal with zero-element CTORs.
28692 * match.pd (bit_field_ref <vector CTOR>): Make sure to
28693 produce vector constants when possible.
28695 2021-04-26 Richard Biener <rguenther@suse.de>
28697 * tree-complex.c: Include gimple-fold.h.
28698 (expand_complex_addition): Use gimple_build.
28699 (expand_complex_multiplication_components): Likewise.
28700 (expand_complex_multiplication): Likewise.
28701 (expand_complex_div_straight): Likewise.
28702 (expand_complex_div_wide): Likewise.
28703 (expand_complex_division): Likewise.
28704 (expand_complex_conjugate): Likewise.
28705 (expand_complex_comparison): Likewise.
28707 2021-04-26 Richard Biener <rguenther@suse.de>
28709 * tree-ssa-phiopt.c (two_value_replacement): Remove use
28710 of legacy gimplify_buildN API.
28712 2021-04-26 Richard Biener <rguenther@suse.de>
28714 PR tree-optimization/99473
28715 * tree-ssa-phiopt.c (cond_store_replacement): Handle all
28718 2021-04-26 Richard Biener <rguenther@suse.de>
28720 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin):
28721 Use replace_call_with_value.
28723 2021-04-26 Richard Biener <rguenther@suse.de>
28725 * tree-ssa-propagate.h (valid_gimple_rhs_p): Remove.
28726 (update_gimple_call): Likewise.
28727 (update_call_from_tree): Likewise.
28728 * tree-ssa-propagate.c (valid_gimple_rhs_p): Remove.
28729 (valid_gimple_call_p): Likewise.
28730 (move_ssa_defining_stmt_for_defs): Likewise.
28731 (finish_update_gimple_call): Likewise.
28732 (update_gimple_call): Likewise.
28733 (update_call_from_tree): Likewise.
28734 (propagate_tree_value_into_stmt): Use replace_call_with_value.
28735 * gimple-fold.h (update_gimple_call): Declare.
28736 * gimple-fold.c (valid_gimple_rhs_p): Move here from
28737 tree-ssa-propagate.c.
28738 (update_gimple_call): Likewise.
28739 (valid_gimple_call_p): Likewise.
28740 (finish_update_gimple_call): Likewise, and simplify.
28741 (gimplify_and_update_call_from_tree): Implement
28742 update_call_from_tree functionality, avoid excessive
28743 push/pop_gimplify_context.
28744 (gimple_fold_builtin): Use only gimplify_and_update_call_from_tree.
28745 (gimple_fold_call): Likewise.
28746 * gimple-ssa-sprintf.c (try_substitute_return_value): Likewise.
28747 * tree-ssa-ccp.c (ccp_folder::fold_stmt): Likewise.
28748 (pass_fold_builtins::execute): Likewise.
28749 (optimize_stack_restore): Use replace_call_with_value.
28750 * tree-cfg.c (fold_loop_internal_call): Likewise.
28751 * tree-ssa-dce.c (maybe_optimize_arith_overflow): Use
28752 only gimplify_and_update_call_from_tree.
28753 * tree-ssa-strlen.c (handle_builtin_strlen): Likewise.
28754 (handle_builtin_strchr): Likewise.
28755 * tsan.c: Include gimple-fold.h instead of tree-ssa-propagate.h.
28757 2021-04-26 Jakub Jelinek <jakub@redhat.com>
28760 * vmsdbgout.c (ASM_OUTPUT_DEBUG_STRING, vmsdbgout_begin_block,
28761 vmsdbgout_end_block, lookup_filename, vmsdbgout_source_line): Remove
28764 2021-04-25 liuhongt <hongtao.liu@intel.com>
28767 * config/i386/i386-builtin.def (BDESC): Change the icode of
28768 the following builtins to CODE_FOR_nothing.
28769 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
28770 IX86_BUILTIN_PCMPEQB128, IX86_BUILTIN_PCMPEQW128,
28771 IX86_BUILTIN_PCMPEQD128, IX86_BUILTIN_PCMPEQQ,
28772 IX86_BUILTIN_PCMPEQB256, IX86_BUILTIN_PCMPEQW256,
28773 IX86_BUILTIN_PCMPEQD256, IX86_BUILTIN_PCMPEQQ256,
28774 IX86_BUILTIN_PCMPGTB128, IX86_BUILTIN_PCMPGTW128,
28775 IX86_BUILTIN_PCMPGTD128, IX86_BUILTIN_PCMPGTQ,
28776 IX86_BUILTIN_PCMPGTB256, IX86_BUILTIN_PCMPGTW256,
28777 IX86_BUILTIN_PCMPGTD256, IX86_BUILTIN_PCMPGTQ256.
28778 * config/i386/sse.md (avx2_eq<mode>3): Deleted.
28779 (sse2_eq<mode>3): Ditto.
28780 (sse4_1_eqv2di3): Ditto.
28781 (sse2_gt<mode>3): Rename to ..
28782 (*sse2_gt<mode>3): .. this.
28784 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
28787 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
28790 * config/darwin.c (darwin_binds_local_p): Assume that any
28791 public symbol might be interposed for PIC code. Update function
28792 header comment to reflect current Darwin capability.
28794 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
28797 * config/darwin.c (darwin_binds_local_p): Assume that any
28798 public symbol might be interposed for PIC code. Update function
28799 header comment to reflect current Darwin capability.
28801 2021-04-24 Richard Sandiford <richard.sandiford@arm.com>
28803 * doc/sourcebuild.texi: Document no-opts and any-opts target
28806 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
28808 * config/bpf/bpf.h (ASM_OUTPUT_ALIGNED_BSS): Use .type and .lcomm.
28810 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
28812 * config/bpf/bpf.h (FUNCTION_BOUNDARY): Set to 64.
28814 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
28817 * config/i386/i386-options.c (ix86_option_override_internal):
28818 Error out when -m96bit-long-double is used with 64bit targets.
28819 * config/i386/i386.md (*pushxf_rounded): Remove pattern.
28821 2021-04-23 Martin Liska <mliska@suse.cz>
28823 * lto-wrapper.c: Remove FIXME about usage of
28824 hardware_concurrency. The function is not on par with
28827 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
28830 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
28831 Copy operand 3 to operand 4. Use sse_reg_operand
28832 as operand 3 predicate.
28833 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): Ditto.
28834 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
28835 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
28836 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2):
28837 Copy operand 1 to operand 0.
28838 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): Ditto.
28839 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
28840 (LDX_ATOMIC/LDX_ATOMIC FP store peephole2 with mem blockage): Ditto.
28842 2021-04-23 Alex Coplan <alex.coplan@arm.com>
28844 PR rtl-optimization/100230
28845 * early-remat.c (early_remat::sort_candidates): Use delete[]
28846 instead of delete for array allocated with new[].
28848 2021-04-23 Richard Biener <rguenther@suse.de>
28850 * genmatch.c (lower_cond): Remove VEC_COND_EXPR special-casing.
28851 (capture_info::capture_info): Likewise.
28852 (capture_info::walk_match): Likewise.
28853 (expr::gen_transform): Likewise.
28854 (dt_simplify::gen_1): Likewise.
28855 * gimple-match-head.c (maybe_resimplify_conditional_op):
28856 Remove VEC_COND_EXPR special-casing.
28857 (gimple_simplify): Likewise.
28858 * gimple.c (gimple_could_trap_p_1): Adjust.
28859 * tree-ssa-pre.c (compute_avail): Allow VEC_COND_EXPR
28860 to participate in PRE.
28862 2021-04-23 Richard Biener <rguenther@suse.de>
28864 * cfganal.c (connect_infinite_loops_to_exit): First call
28865 add_noreturn_fake_exit_edges.
28866 * ipa-sra.c (process_scan_results): Do not call the now redundant
28867 add_noreturn_fake_exit_edges.
28868 * predict.c (tree_estimate_probability): Likewise.
28869 (rebuild_frequencies): Likewise.
28870 * store-motion.c (one_store_motion_pass): Likewise.
28872 2021-04-23 Richard Biener <rguenther@suse.de>
28874 PR tree-optimization/100222
28875 * predict.c (pass_profile::execute): Remove redundant call to
28876 mark_irreducible_loops.
28877 (report_predictor_hitrates): Likewise.
28879 2021-04-23 Richard Biener <rguenther@suse.de>
28881 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Avoid
28882 valid_gimple_rhs_p by instead gimplifying to one.
28884 2021-04-23 Richard Biener <rguenther@suse.de>
28886 PR tree-optimization/99971
28887 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
28888 Always use TBAA for loads.
28890 2021-04-23 liuhongt <hongtao.liu@intel.com>
28893 * config/i386/i386-options.c (ix86_option_override_internal):
28894 Clear MASK_AVX256_SPLIT_UNALIGNED_LOAD/STORE in x_target_flags
28895 when X86_TUNE_AVX256_UNALIGNED_LOAD/STORE_OPTIMAL is enabled
28896 by target attribute.
28898 2021-04-23 David Edelsohn <dje.gcc@gmail.com>
28900 * config/rs6000/aix71.h (PREFERRED_DEBUGGING_TYPE): Change to
28902 * config/rs6000/aix72.h (PREFERRED_DEBUGGING_TYPE): Same.
28904 2021-04-22 David Edelsohn <dje.gcc@gmail.com>
28906 * config.gcc (powerpc-ibm-aix6.*): Remove.
28907 * config/rs6000/aix61.h: Delete.
28909 2021-04-22 Martin Liska <mliska@suse.cz>
28911 PR testsuite/100159
28912 PR testsuite/100192
28913 * builtins.c (expand_builtin): Fix typos and missing comments.
28914 * dwarf2out.c (gen_subprogram_die): Likewise.
28915 (gen_struct_or_union_type_die): Likewise.
28917 2021-04-22 Uroš Bizjak <ubizjak@gmail.com>
28920 * config/i386/i386-expand.c (ix86_expand_convert_uns_sidf_sse):
28921 Remove the sign with FE_DOWNWARD, where x - x = -0.0.
28923 2021-04-21 Iain Sandoe <iain@sandoe.co.uk>
28925 * config/i386/darwin.h (TARGET_64BIT): Remove definition
28926 based on TARGET_ISA_64BIT.
28927 (TARGET_64BIT_P): Remove definition based on
28928 TARGET_ISA_64BIT_P().
28930 2021-04-21 Martin Liska <mliska@suse.cz>
28933 2021-04-21 Martin Liska <mliska@suse.cz>
28935 * lto-wrapper.c (cpuset_popcount): Remove.
28936 (init_num_threads): Remove and use hardware_concurrency.
28938 2021-04-21 Martin Liska <mliska@suse.cz>
28941 * main.c (main): Call toplev::finalize in CHECKING_P mode.
28942 * ipa-modref.c (ipa_modref_c_finalize): summaries are NULL
28943 when incremental LTO linking happens.
28945 2021-04-21 Martin Liska <mliska@suse.cz>
28947 * lto-wrapper.c (run_gcc): When -flto=jobserver is used, but the
28948 makeserver cannot be detected, then use -flto=N fallback.
28950 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
28952 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): When cross-compiling,
28953 default to yes for aarch64-linux-gnu.
28954 * configure: Regenerate.
28956 2021-04-21 Martin Liska <mliska@suse.cz>
28958 * lto-wrapper.c (cpuset_popcount): Remove.
28959 (init_num_threads): Remove and use hardware_concurrency.
28961 2021-04-21 Martin Liska <mliska@suse.cz>
28963 * config/i386/i386.c: Remove superfluous || TARGET_MACHO
28964 which remains to be '(... || 0)' and clang complains about it.
28965 * dwarf2out.c (AT_vms_delta): Declare conditionally.
28966 (add_AT_vms_delta): Likewise.
28967 * tree.c (fld_simplified_type): Use rather more common pattern
28968 for disabling of something (#if 0).
28969 (get_tree_code_name): Likewise.
28970 (verify_type_variant): Likewise.
28972 2021-04-21 Martin Liska <mliska@suse.cz>
28974 * config/i386/i386-expand.c (decide_alignment): Use newly named
28975 macro TARGET_CPU_P.
28976 * config/i386/i386.c (ix86_decompose_address): Likewise.
28977 (ix86_address_cost): Likewise.
28978 (ix86_lea_outperforms): Likewise.
28979 (ix86_avoid_lea_for_addr): Likewise.
28980 (ix86_add_stmt_cost): Likewise.
28981 * config/i386/i386.h (TARGET_*): Remove.
28982 (TARGET_CPU_P): New macro.
28983 * config/i386/i386.md: Use newly named macro TARGET_CPU_P.
28984 * config/i386/x86-tune-sched-atom.c (do_reorder_for_imul): Likewise.
28985 (swap_top_of_ready_list): Likewise.
28986 (ix86_atom_sched_reorder): Likewise.
28987 * config/i386/x86-tune-sched-bd.c (ix86_bd_has_dispatch): Likewise.
28988 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Likewise.
28990 2021-04-21 Martin Liska <mliska@suse.cz>
28992 * config/i386/i386-options.c (TARGET_EXPLICIT_NO_SAHF_P):
28994 (SET_TARGET_NO_SAHF): Likewise.
28995 (TARGET_EXPLICIT_PREFETCH_SSE_P): Likewise.
28996 (SET_TARGET_PREFETCH_SSE): Likewise.
28997 (TARGET_EXPLICIT_NO_TUNE_P): Likewise.
28998 (SET_TARGET_NO_TUNE): Likewise.
28999 (TARGET_EXPLICIT_NO_80387_P): Likewise.
29000 (SET_TARGET_NO_80387): Likewise.
29002 * config/i386/i386.h (TARGET_*): Remove.
29003 * opth-gen.awk: Generate new used macros.
29005 2021-04-21 Martin Liska <mliska@suse.cz>
29007 * config/i386/i386.h (PTA_*): Remove.
29008 (enum pta_flag): New.
29009 (DEF_PTA): Generate PTA_* values from i386-isa.def.
29010 * config/i386/i386-isa.def: New file.
29012 2021-04-21 Alex Coplan <alex.coplan@arm.com>
29015 * config/aarch64/aarch64-bti-insert.c (aarch64_bti_j_insn_p): New.
29016 (rest_of_insert_bti): Avoid inserting duplicate bti j insns for
29017 jump table targets.
29019 2021-04-21 H.J. Lu <hjl.tools@gmail.com>
29021 * config.gcc: Install mwaitintrin.h for i[34567]86-*-* and
29022 x86_64-*-* targets.
29023 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_MWAIT_SET):
29025 (OPTION_MASK_ISA2_MWAIT_UNSET): Likewise.
29026 (ix86_handle_option): Handle -mmwait.
29027 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
29028 Replace OPTION_MASK_ISA_SSE3 with OPTION_MASK_ISA2_MWAIT on
29029 __builtin_ia32_monitor and __builtin_ia32_mwait.
29030 * config/i386/i386-options.c (isa2_opts): Add -mmwait.
29031 (ix86_valid_target_attribute_inner_p): Likewise.
29032 (ix86_option_override_internal): Enable mwait/monitor
29033 instructions for -msse3.
29034 * config/i386/i386.h (TARGET_MWAIT): New.
29035 (TARGET_MWAIT_P): Likewise.
29036 * config/i386/i386.opt: Add -mmwait.
29037 * config/i386/mwaitintrin.h: New file.
29038 * config/i386/pmmintrin.h: Include <mwaitintrin.h>.
29039 * config/i386/sse.md (sse3_mwait): Replace TARGET_SSE3 with
29041 (@sse3_monitor_<mode>): Likewise.
29042 * config/i386/x86gprintrin.h: Include <mwaitintrin.h>.
29043 * doc/extend.texi: Document mwait target attribute.
29044 * doc/invoke.texi: Document -mmwait.
29046 2021-04-21 Martin Liska <mliska@suse.cz>
29048 * config/i386/i386-options.c (DEF_ENUM): Remove it.
29049 * config/i386/i386-opts.h (DEF_ENUM): Likewise.
29050 * config/i386/stringop.def (DEF_ENUM): Likewise.
29052 2021-04-21 Martin Liska <mliska@suse.cz>
29054 * tree-cfg.c (gimple_verify_flow_info): Use qD instead
29055 of print_generic_expr.
29057 2021-04-21 Jakub Jelinek <jakub@redhat.com>
29059 PR rtl-optimization/100148
29060 * cprop.c (constprop_register): Use next_nondebug_insn instead of
29063 2021-04-21 Martin Liska <mliska@suse.cz>
29066 * cgraphunit.c (cgraph_node::analyze): Remove duplicate
29067 free_dominance_info calls.
29069 2021-04-21 Richard Biener <rguenther@suse.de>
29071 * gimple-fold.c (maybe_fold_reference): Remove is_lhs
29072 parameter (and assume it to be false).
29073 (fold_gimple_assign): Adjust, remove all callers of
29074 maybe_fold_reference calling it with is_lhs true.
29075 (gimple_fold_call): Likewise.
29076 (fold_stmt_1): Likewise.
29078 2021-04-21 Richard Biener <rguenther@suse.de>
29080 * fold-const.c (pedantic_non_lvalue_loc): Remove.
29081 (fold_binary_loc): Adjust.
29082 (fold_ternary_loc): Likewise.
29084 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
29086 PR middle-end/100130
29087 * varasm.c (get_block_for_decl): Make sure that any use of the
29088 retain attribute matches the section's retain flag.
29089 (switch_to_section): Check for retain mismatches even when
29090 changing sections, but do not warn if the given decl is the
29091 section's named.decl.
29092 (output_object_block): Pass the first decl in the block (if any)
29093 to switch_to_section.
29095 2021-04-20 H.J. Lu <hjl.tools@gmail.com>
29097 * config/i386/i386-c.c (ix86_target_macros_internal): Define
29098 __CRC32__ for -mcrc32.
29099 * config/i386/i386-options.c (ix86_option_override_internal):
29100 Enable crc32 instruction for -msse4.2.
29101 * config/i386/i386.md (sse4_2_crc32<mode>): Remove TARGET_SSE4_2
29103 (sse4_2_crc32di): Likewise.
29104 * config/i386/ia32intrin.h: Use crc32 target option for CRC32
29107 2021-04-20 Segher Boessenkool <segher@kernel.crashing.org>
29110 * config/rs6000/rs6000.c (rs6000_machine_from_flags): Do not consider
29113 2021-04-20 Martin Liska <mliska@suse.cz>
29115 * doc/invoke.texi: Fix typo.
29116 * params.opt: Likewise.
29118 2021-04-20 Martin Liska <mliska@suse.cz>
29120 * doc/invoke.texi: Document new param.
29122 2021-04-19 Andrew MacLeod <amacleod@redhat.com>
29124 PR tree-optimization/100081
29125 * gimple-range-cache.h (ranger_cache): Inherit from gori_compute
29126 rather than gori_compute_cache.
29127 * gimple-range-gori.cc (is_gimple_logical_p): Move to top of file.
29128 (range_def_chain::m_logical_depth): New member.
29129 (range_def_chain::range_def_chain): Initialize m_logical_depth.
29130 (range_def_chain::get_def_chain): Don't build defchains through more
29131 than LOGICAL_LIMIT logical expressions.
29132 * params.opt (param_ranger_logical_depth): New.
29134 2021-04-19 Richard Earnshaw <rearnsha@arm.com>
29137 * config/arm/arm.c (arm_configure_build_target): Do not strip
29138 extended FPU/SIMD feature bits from the target ISA when -mfpu
29139 is specified (partial revert of r11-8168).
29141 2021-04-19 Thomas Schwinge <thomas@codesourcery.com>
29143 * params.opt (-param=openacc-kernels=): Add.
29144 * omp-oacc-kernels-decompose.cc
29145 (pass_omp_oacc_kernels_decompose::gate): Use it.
29146 * doc/invoke.texi (-fopenacc-kernels=@var{mode}): Move...
29147 (--param): ... here, 'openacc-kernels'.
29149 2021-04-19 Martin Liska <mliska@suse.cz>
29152 * gengtype.c (finish_root_table): Align function arguments
29153 in between declaration and definition.
29155 2021-04-19 Eric Botcazou <ebotcazou@adacore.com>
29157 * config/i386/winnt.c (i386_pe_seh_cold_init): Properly deal with
29158 frames larger than the SEH maximum frame size.
29160 2021-04-18 Segher Boessenkool <segher@kernel.crashing.org>
29162 PR rtl-optimization/99927
29163 * combine.c (distribute_notes) [REG_UNUSED]: If the register already
29164 is dead, just drop it.
29166 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
29169 * config/i386/winnt-d.c (TARGET_D_TEMPLATES_ALWAYS_COMDAT): Define.
29170 * doc/tm.texi: Regenerate.
29171 * doc/tm.texi.in (D language and ABI): Add @hook for
29172 TARGET_D_TEMPLATES_ALWAYS_COMDAT.
29174 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
29176 * config/darwin-d.c (darwin_d_handle_target_object_format): New
29178 (darwin_d_register_target_info): New function.
29179 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
29180 * config/dragonfly-d.c (dragonfly_d_handle_target_object_format): New
29182 (dragonfly_d_register_target_info): New function.
29183 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
29184 * config/freebsd-d.c (freebsd_d_handle_target_object_format): New
29186 (freebsd_d_register_target_info): New function.
29187 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
29188 * config/glibc-d.c (glibc_d_handle_target_object_format): New
29190 (glibc_d_register_target_info): New function.
29191 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
29192 * config/i386/i386-d.c (ix86_d_handle_target_object_format): New
29194 (ix86_d_register_target_info): Add ix86_d_handle_target_object_format
29195 as handler for objectFormat key.
29196 * config/i386/winnt-d.c (winnt_d_handle_target_object_format): New
29198 (winnt_d_register_target_info): New function.
29199 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
29200 * config/netbsd-d.c (netbsd_d_handle_target_object_format): New
29202 (netbsd_d_register_target_info): New function.
29203 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
29204 * config/openbsd-d.c (openbsd_d_handle_target_object_format): New
29206 (openbsd_d_register_target_info): New function.
29207 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
29208 * config/pa/pa-d.c (pa_d_handle_target_object_format): New function.
29209 (pa_d_register_target_info): Add pa_d_handle_target_object_format as
29210 handler for objectFormat key.
29211 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_object_format): New
29213 (rs6000_d_register_target_info): Add
29214 rs6000_d_handle_target_object_format as handler for objectFormat key.
29215 * config/sol2-d.c (solaris_d_handle_target_object_format): New
29217 (solaris_d_register_target_info): New function.
29218 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
29220 2021-04-16 Jakub Jelinek <jakub@redhat.com>
29223 * config/aarch64/aarch64.c (aarch64_function_arg_alignment): Change
29224 abi_break argument from bool * to unsigned *, store there the pre-GCC 9
29226 (aarch64_layout_arg, aarch64_gimplify_va_arg_expr): Adjust callers.
29227 (aarch64_function_arg_regno_p): Likewise. Only emit -Wpsabi note if
29228 the old and new alignment after applying MIN/MAX to it is different.
29230 2021-04-16 Tamar Christina <tamar.christina@arm.com>
29233 * config/aarch64/aarch64-sve.md (@aarch64_sve_trn1_conv<mode>): New.
29234 * config/aarch64/aarch64.c (aarch64_expand_sve_const_pred_trn): Use new
29236 * config/aarch64/iterators.md (UNSPEC_TRN1_CONV): New.
29238 2021-04-16 Bill Schmidt <wschmidt@linux.ibm.com>
29240 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Revise
29241 this section and its subsections.
29243 2021-04-16 Jakub Jelinek <jakub@redhat.com>
29246 * config/aarch64/aarch64.md (*neg_asr_si2_extr, *extrsi5_insn_di): New
29247 define_insn patterns.
29249 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
29251 PR rtl-optimization/98689
29252 * reg-notes.def (UNTYPED_CALL): New note.
29253 * combine.c (distribute_notes): Handle it.
29254 * emit-rtl.c (try_split): Likewise.
29255 * rtlanal.c (rtx_properties::try_to_add_insn): Likewise. Assume
29256 that calls with the note implicitly set all return value registers.
29257 * builtins.c (expand_builtin_apply): Add a REG_UNTYPED_CALL
29260 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
29262 PR rtl-optimization/99596
29263 * rtlanal.c (rtx_properties::try_to_add_insn): Don't add global
29264 register accesses for const calls. Assume that pure functions
29265 can only read from global registers. Ignore cases in which
29266 the stack pointer has been marked global.
29268 2021-04-16 Jakub Jelinek <jakub@redhat.com>
29271 * tree-vect-loop.c (vect_transform_loop): Don't remove just
29272 dead scalar .MASK_LOAD calls, but also dead .COND_* calls - replace
29273 them by their last argument.
29275 2021-04-15 Martin Liska <mliska@suse.cz>
29277 * doc/invoke.texi: Other params don't use it, remove it.
29279 2021-04-15 Richard Biener <rguenther@suse.de>
29281 * gimple-builder.h: Add deprecation note.
29283 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
29286 * attribs.h (restrict_type_identity_attributes_to): Declare.
29287 * attribs.c (restrict_type_identity_attributes_to): New function.
29289 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
29292 * attribs.h (affects_type_identity_attributes): Declare.
29293 * attribs.c (remove_attributes_matching): New function.
29294 (affects_type_identity_attributes): Likewise.
29296 2021-04-15 Jakub Jelinek <jakub@redhat.com>
29299 * config/aarch64/aarch64.md (*<LOGICAL:optab>_<SHIFT:optab><mode>3):
29300 Add combine splitters for *<LOGICAL:optab>_ashl<mode>3 with
29301 ZERO_EXTEND, SIGN_EXTEND or AND.
29303 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
29305 PR rtl-optimization/99929
29306 * rtl.h (same_vector_encodings_p): New function.
29307 * cse.c (exp_equiv_p): Check that CONST_VECTORs have the same encoding.
29308 * cselib.c (rtx_equal_for_cselib_1): Likewise.
29309 * jump.c (rtx_renumbered_equal_p): Likewise.
29310 * lra-constraints.c (operands_match_p): Likewise.
29311 * reload.c (operands_match_p): Likewise.
29312 * rtl.c (rtx_equal_p_cb, rtx_equal_p): Likewise.
29314 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
29316 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Print
29317 more information about variable-length CONST_VECTORs.
29319 2021-04-14 Vladimir N. Makarov <vmakarov@redhat.com>
29321 PR rtl-optimization/100066
29322 * lra-constraints.c (split_reg): Check paradoxical_subreg_p for
29323 ordered modes when choosing splitting mode for hard reg.
29325 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
29328 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector_sel):
29330 (aarch64_expand_sve_const_vector): Use it for nelts_per_pattern==2.
29332 2021-04-14 Andreas Krebbel <krebbel@linux.ibm.com>
29334 * config/s390/s390-builtins.def (O_M5, O_M12, ...): Add new macros
29335 for mask operand types.
29336 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
29337 (s390_vec_permi_dbl, s390_vpdi): Use the M5 type for the immediate
29339 (s390_vec_msum_u128, s390_vmslg): Use the M12 type for the
29341 * config/s390/s390.c (s390_const_operand_ok): Check the new
29342 operand types and generate a list of valid values.
29344 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
29346 * doc/tm.texi: Regenerate.
29347 * doc/tm.texi.in (D language and ABI): Add @hook for
29348 TARGET_D_REGISTER_OS_TARGET_INFO.
29350 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
29352 * config/aarch64/aarch64-d.c (aarch64_d_handle_target_float_abi): New
29354 (aarch64_d_register_target_info): New function.
29355 * config/aarch64/aarch64-protos.h (aarch64_d_register_target_info):
29357 * config/aarch64/aarch64.h (TARGET_D_REGISTER_CPU_TARGET_INFO):
29359 * config/arm/arm-d.c (arm_d_handle_target_float_abi): New function.
29360 (arm_d_register_target_info): New function.
29361 * config/arm/arm-protos.h (arm_d_register_target_info): Declare.
29362 * config/arm/arm.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29363 * config/i386/i386-d.c (ix86_d_handle_target_float_abi): New function.
29364 (ix86_d_register_target_info): New function.
29365 * config/i386/i386-protos.h (ix86_d_register_target_info): Declare.
29366 * config/i386/i386.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29367 * config/mips/mips-d.c (mips_d_handle_target_float_abi): New function.
29368 (mips_d_register_target_info): New function.
29369 * config/mips/mips-protos.h (mips_d_register_target_info): Declare.
29370 * config/mips/mips.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29371 * config/pa/pa-d.c (pa_d_handle_target_float_abi): New function.
29372 (pa_d_register_target_info): New function.
29373 * config/pa/pa-protos.h (pa_d_register_target_info): Declare.
29374 * config/pa/pa.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29375 * config/riscv/riscv-d.c (riscv_d_handle_target_float_abi): New
29377 (riscv_d_register_target_info): New function.
29378 * config/riscv/riscv-protos.h (riscv_d_register_target_info): Declare.
29379 * config/riscv/riscv.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29380 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_float_abi): New
29382 (rs6000_d_register_target_info): New function.
29383 * config/rs6000/rs6000-protos.h (rs6000_d_register_target_info):
29385 * config/rs6000/rs6000.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29386 * config/s390/s390-d.c (s390_d_handle_target_float_abi): New function.
29387 (s390_d_register_target_info): New function.
29388 * config/s390/s390-protos.h (s390_d_register_target_info): Declare.
29389 * config/s390/s390.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29390 * config/sparc/sparc-d.c (sparc_d_handle_target_float_abi): New
29392 (sparc_d_register_target_info): New function.
29393 * config/sparc/sparc-protos.h (sparc_d_register_target_info): Declare.
29394 * config/sparc/sparc.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
29395 * doc/tm.texi: Regenerate.
29396 * doc/tm.texi.in (D language and ABI): Add @hook for
29397 TARGET_D_REGISTER_CPU_TARGET_INFO.
29399 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
29401 * config/i386/i386-d.c (ix86_d_has_stdcall_convention): New function.
29402 * config/i386/i386-protos.h (ix86_d_has_stdcall_convention): Declare.
29403 * config/i386/i386.h (TARGET_D_HAS_STDCALL_CONVENTION): Define.
29404 * doc/tm.texi: Regenerate.
29405 * doc/tm.texi.in (D language and ABI): Add @hook for
29406 TARGET_D_HAS_STDCALL_CONVENTION.
29408 2021-04-14 Richard Biener <rguenther@suse.de>
29410 * tree-cfg.c (verify_gimple_assign_ternary): Verify that
29411 VEC_COND_EXPRs have a gimple_val condition.
29412 * tree-ssa-propagate.c (valid_gimple_rhs_p): VEC_COND_EXPR
29413 can no longer have a GENERIC condition.
29415 2021-04-14 Richard Earnshaw <rearnsha@arm.com>
29418 * config/arm/arm.c (arm_configure_build_target): Strip isa_all_fpbits
29419 from the isa_delta when -mfpu has been used.
29420 (arm_options_perform_arch_sanity_checks): It's the architecture that
29421 lacks an FPU not the processor.
29423 2021-04-13 Richard Biener <rguenther@suse.de>
29425 PR tree-optimization/100053
29426 * tree-ssa-sccvn.c (vn_nary_op_get_predicated_value): Do
29427 not use optimistic dominance queries for backedges to validate
29429 (dominated_by_p_w_unex): Add parameter to ignore executable
29430 state on backedges.
29431 (rpo_elim::eliminate_avail): Adjust.
29433 2021-04-13 Jakub Jelinek <jakub@redhat.com>
29436 * config/aarch64/aarch64.md (*aarch64_bfxil<mode>_extr,
29437 *aarch64_bfxilsi_extrdi): New define_insn patterns.
29439 2021-04-13 Jakub Jelinek <jakub@redhat.com>
29442 * simplify-rtx.c (simplify_immed_subreg): For MODE_COMPOSITE_P
29443 outermode, return NULL if the result doesn't encode back to the
29444 original byte sequence.
29445 (simplify_gen_subreg): Don't create SUBREGs from constants to
29446 MODE_COMPOSITE_P outermode.
29448 2021-04-12 Jakub Jelinek <jakub@redhat.com>
29450 PR rtl-optimization/99905
29451 * combine.c (expand_compound_operation): If pos + len > modewidth,
29452 perform the right shift by pos in inner_mode and then convert to mode,
29453 instead of trying to simplify a shift of rtx with inner_mode by pos
29454 as if it was a shift in mode.
29456 2021-04-12 Jakub Jelinek <jakub@redhat.com>
29459 * combine.c (simplify_and_const_int_1): Don't optimize varop
29460 away if it has side-effects.
29462 2021-04-12 Martin Liska <mliska@suse.cz>
29464 * doc/extend.texi: Escape @smallexample content.
29466 2021-04-12 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
29468 * config/s390/s390.md ("*movdi_31", "*movdi_64"): Add
29469 alternative in order to load a DFP zero.
29471 2021-04-12 Martin Liska <mliska@suse.cz>
29473 * doc/extend.texi: Be more precise in documentation
29474 of symver attribute.
29476 2021-04-12 Martin Liska <mliska@suse.cz>
29479 * gimplify.c (gimplify_expr): Right now, we unpoison all
29480 variables before a goto <dest>. We should not do it if we are
29483 2021-04-12 Cui,Lili <lili.cui@intel.com>
29485 * common/config/i386/cpuinfo.h (get_intel_cpu): Handle
29487 * common/config/i386/i386-common.c (processor_names): Add
29489 (processor_alias_table): Add rocketlake.
29490 * common/config/i386/i386-cpuinfo.h (processor_subtypes): Add
29491 INTEL_COREI7_ROCKETLAKE.
29492 * config.gcc: Add -march=rocketlake.
29493 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
29495 * config/i386/i386-options.c (m_ROCKETLAKE) : Define.
29496 (processor_cost_table): Add rocketlake cost.
29497 * config/i386/i386.h (ix86_size_cost) : Define
29499 (processor_type) : Add PROCESSOR_ROCKETLAKE.
29500 (PTA_ROCKETLAKE): Ditto.
29501 * doc/extend.texi: Add rocketlake.
29502 * doc/invoke.texi: Add rocketlake.
29504 2021-04-12 Cui,Lili <lili.cui@intel.com>
29506 * config/i386/i386.h (PTA_ALDERLAKE): Change alderlake ISA list.
29507 * config/i386/i386-options.c (m_CORE_AVX2): Add m_ALDERLAKE.
29508 * common/config/i386/cpuinfo.h (get_intel_cpu): Add AlderLake model.
29509 * doc/invoke.texi: Change alderlake ISA list.
29511 2021-04-11 Hafiz Abid Qadeer <abidh@codesourcery.com>
29513 PR middle-end/98088
29514 * omp-expand.c (expand_oacc_collapse_init): Update condition in
29517 2021-04-10 H.J. Lu <hjl.tools@gmail.com>
29520 * config/i386/serializeintrin.h (_serialize): Defined as macro.
29522 2021-04-10 Jakub Jelinek <jakub@redhat.com>
29525 * expr.c (expand_expr_addr_expr_1): Test is_global_var rather than
29526 just TREE_STATIC on COMPOUND_LITERAL_EXPR_DECLs.
29528 2021-04-10 Jakub Jelinek <jakub@redhat.com>
29530 PR middle-end/99989
29531 * gimple-ssa-warn-alloca.c
29532 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit to
29533 0 with integer precision unconditionally.
29535 2021-04-10 Jakub Jelinek <jakub@redhat.com>
29537 PR rtl-optimization/98601
29538 * rtlanal.c (rtx_addr_can_trap_p_1): Allow in assert unknown size
29539 not just for BLKmode, but also for VOIDmode. For STRICT_ALIGNMENT
29540 unaligned_mems handle VOIDmode like BLKmode.
29542 2021-04-10 Jan Hubicka <hubicka@ucw.cz>
29545 * tree.c (free_lang_data_in_decl): Do not release body of
29546 declare_variant_alt.
29548 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
29550 * config/aarch64/aarch64.c (aarch64_option_restore): If the
29551 architecture was specified explicitly and the tuning wasn't,
29552 tune for the architecture rather than the configured default CPU.
29554 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
29556 * config/aarch64/aarch64.md (tlsdesc_small_sve_<mode>): Use X30
29557 as the temporary register.
29559 2021-04-09 Martin Liska <mliska@suse.cz>
29561 * doc/extend.texi: Move non-target attributes on the top level.
29563 2021-04-09 Martin Liska <mliska@suse.cz>
29565 * doc/invoke.texi: Document minimum and maximum value of the
29566 argument for both supported compression algorithms.
29568 2021-04-08 David Edelsohn <dje.gcc@gmail.com>
29570 * config/rs6000/rs6000.c (rs6000_xcoff_select_section): Select
29571 TLS BSS before TLS data.
29572 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .comm.
29574 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
29576 * doc/sourcebuild.texi (stdint_types_mbig_endian): Document.
29578 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
29580 * match.pd: Extend vec_cond folds to handle shifts.
29582 2021-04-08 Maciej W. Rozycki <macro@orcam.me.uk>
29584 * config/vax/vax.md: Fix comment for `*bit<mode>' pattern's
29587 2021-04-08 Alex Coplan <alex.coplan@arm.com>
29590 * config/arm/iterators.md (MVE_vecs): New.
29591 (V_elem): Also handle V2DF.
29592 * config/arm/mve.md (*mve_mov<mode>): Rename to ...
29593 (*mve_vdup<mode>): ... this. Remove second alternative since
29594 vec_duplicate of const_int is not canonical RTL, and we don't
29595 want to match symbol_refs.
29596 (*mve_vec_duplicate<mode>): Delete (pattern is redundant).
29598 2021-04-08 Xionghu Luo <luoxhu@linux.ibm.com>
29600 * fold-const.c (fold_single_bit_test): Fix typo.
29601 * print-rtl.c (print_rtx_insn_vec): Call print_rtl_single
29604 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
29606 PR tree-optimization/97513
29607 * tree-vect-slp.c (vect_add_slp_permutation): New function,
29609 (vectorizable_slp_permutation): ...here. Detect cases in which
29610 all VEC_PERM_EXPRs are guaranteed to have the same stepped
29611 permute vector and only generate one permute vector for that case.
29612 Extend that case to handle variable-length vectors.
29614 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
29616 PR tree-optimization/99873
29617 * tree-vect-slp.c (vect_slp_prefer_store_lanes_p): New function.
29618 (vect_build_slp_instance): Don't split store groups that could
29619 use IFN_STORE_LANES.
29621 2021-04-07 Jakub Jelinek <jakub@redhat.com>
29624 * varasm.c (output_constant_pool_contents): Don't strip name encoding
29625 from XSTR (desc->sym, 0) or from label before passing those to
29628 2021-04-07 Richard Biener <rguenther@suse.de>
29630 PR tree-optimization/99954
29631 * tree-loop-distribution.c: Include tree-affine.h.
29632 (generate_memcpy_builtin): Try using tree-affine to prove
29634 (loop_distribution::classify_builtin_ldst): Always classify
29637 2021-04-07 Richard Biener <rguenther@suse.de>
29639 PR tree-optimization/99947
29640 * tree-vect-loop.c (vectorizable_induction): Pre-allocate
29641 steps vector to avoid pushing elements from the reallocated
29644 2021-04-07 Richard Biener <rguenther@suse.de>
29646 * tree-ssa-sccvn.h (print_vn_reference_ops): Declare.
29647 * tree-ssa-pre.c (print_pre_expr): Factor out VN reference operand
29649 * tree-ssa-sccvn.c (print_vn_reference_ops): ... into this new
29651 (debug_vn_reference_ops): New.
29653 2021-04-07 Bin Cheng <bin.cheng@linux.alibaba.com>
29655 PR tree-optimization/98736
29656 * tree-loop-distribution.c
29657 * (loop_distribution::bb_top_order_init):
29658 Compute RPO with programing order preserved by calling function
29659 rev_post_order_and_mark_dfs_back_seme.
29661 2021-04-06 Vladimir N. Makarov <vmakarov@redhat.com>
29664 * lra-constraints.c (split_reg): Don't check paradoxical_subreg_p.
29665 * lra-lives.c (clear_sparseset_regnos, regnos_in_sparseset_p): New
29667 (process_bb_lives): Don't update biggest mode of hard reg for
29668 implicit in multi-register group. Use the new functions for
29669 updating dead_set and unused_set by register notes.
29671 2021-04-06 Xianmiao Qu <xianmiao_qu@c-sky.com>
29673 * config/csky/csky_pipeline_ck802.md : Use insn reservation name
29676 2021-04-06 H.J. Lu <hjl.tools@gmail.com>
29678 * config/i386/x86-tune-costs.h (skylake_memcpy): Updated.
29679 (skylake_memset): Likewise.
29680 (skylake_cost): Change CLEAR_RATIO to 17.
29681 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
29682 Replace m_CANNONLAKE, m_ICELAKE_CLIENT, m_ICELAKE_SERVER,
29683 m_TIGERLAKE and m_SAPPHIRERAPIDS with m_SKYLAKE and m_CORE_AVX512.
29685 2021-04-06 Richard Biener <rguenther@suse.de>
29687 PR tree-optimization/99880
29688 * tree-vect-loop.c (maybe_set_vectorized_backedge_value): Only
29689 set vectorized defs of relevant PHIs.
29691 2021-04-06 Richard Biener <rguenther@suse.de>
29693 PR tree-optimization/99924
29694 * tree-vect-slp.c (vect_bb_partition_graph_r): Do not mark
29695 nodes w/o scalar stmts as visited.
29697 2021-04-06 Alex Coplan <alex.coplan@arm.com>
29700 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Also use base
29701 PCS for [su]fix_optab.
29703 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
29705 * config/darwin.c (machopic_legitimize_pic_address): Check
29706 that the current pic register is one of the hard reg set
29707 before setting liveness.
29709 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
29711 * config/darwin.c (machopic_legitimize_pic_address): Fix
29712 whitespace, remove unused code.
29714 2021-04-03 Jakub Jelinek <jakub@redhat.com>
29716 PR tree-optimization/99882
29717 * gimple-ssa-store-merging.c (bswap_view_convert): Handle val with
29720 2021-04-03 Jakub Jelinek <jakub@redhat.com>
29722 PR rtl-optimization/99863
29723 * dse.c (replace_read): Drop regs_live argument. Instead of
29724 regs_live, use store_insn->fixed_regs_live if non-NULL,
29725 otherwise punt if insns sequence clobbers or sets any hard
29728 2021-04-03 Jakub Jelinek <jakub@redhat.com>
29731 * targhooks.h (default_print_patchable_function_entry_1): Declare.
29732 * targhooks.c (default_print_patchable_function_entry_1): New function,
29733 copied from default_print_patchable_function_entry with an added flags
29735 (default_print_patchable_function_entry): Rewritten into a small
29736 wrapper around default_print_patchable_function_entry_1.
29737 * config/rs6000/rs6000.c (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
29739 (rs6000_print_patchable_function_entry): New function.
29741 2021-04-02 Eric Botcazou <ebotcazou@adacore.com>
29743 * doc/invoke.texi (fdelete-dead-exceptions): Minor tweak.
29745 2021-04-01 Jason Merrill <jason@redhat.com>
29748 * common.opt: Document v15 and v16.
29750 2021-04-01 Richard Biener <rguenther@suse.de>
29752 PR tree-optimization/99863
29753 * gimplify.c (gimplify_init_constructor): Recompute vector
29756 2021-04-01 Jakub Jelinek <jakub@redhat.com>
29758 * doc/extend.texi (symver attribute): Fix up syntax errors
29761 2021-04-01 Jakub Jelinek <jakub@redhat.com>
29763 PR tree-optimization/96573
29764 * gimple-ssa-store-merging.c (init_symbolic_number): Handle
29765 also pointer types.
29767 2021-04-01 Richard Biener <rguenther@suse.de>
29769 PR tree-optimization/99856
29770 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Promote
29771 precision to vector element precision.
29773 2021-04-01 Martin Jambor <mjambor@suse.cz>
29775 PR tree-optimization/97009
29776 * tree-sra.c (access_or_its_child_written): New function.
29777 (propagate_subaccesses_from_rhs): Use it instead of a simple grp_write
29780 2021-03-31 Jan Hubicka <hubicka@ucw.cz>
29783 * cif-code.def (USES_COMDAT_LOCAL): Make CIF_FINAL_NORMAL.
29785 2021-03-31 Pat Haugen <pthaugen@linux.ibm.com>
29788 * config/rs6000/altivec.md (xxspltiw_v4si, xxspltiw_v4sf_inst,
29789 xxspltidp_v2df_inst, xxsplti32dx_v4si_inst, xxsplti32dx_v4sf_inst,
29790 xxblend_<mode>, xxpermx_inst, xxeval): Mark prefixed.
29791 * config/rs6000/mma.md (mma_<vvi4i4i8>, mma_<avvi4i4i8>,
29792 mma_<vvi4i4i2>, mma_<avvi4i4i2>, mma_<vvi4i4>, mma_<avvi4i4>,
29793 mma_<pvi4i2>, mma_<apvi4i2>, mma_<vvi4i4i4>, mma_<avvi4i4i4>):
29795 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Adjust test.
29796 * config/rs6000/rs6000.md (define_attr "maybe_prefixed"): New.
29797 (define_attr "prefixed"): Update initializer.
29799 2021-03-31 Jakub Jelinek <jakub@redhat.com>
29802 * dwarf2out.c (debug_ranges_dwo_section): New variable.
29803 (DW_RANGES_IDX_SKELETON): Define.
29804 (struct dw_ranges): Add begin_entry and end_entry members.
29805 (DEBUG_DWO_RNGLISTS_SECTION): Define.
29806 (add_ranges_num): Adjust r initializer for addition of *_entry
29808 (add_ranges_by_labels): For -gsplit-dwarf and force_direct,
29809 set idx to DW_RANGES_IDX_SKELETON.
29810 (use_distinct_base_address_for_range): New function.
29811 (index_rnglists): Don't set r->idx if it is equal to
29812 DW_RANGES_IDX_SKELETON. Initialize r->begin_entry and
29813 r->end_entry for -gsplit-dwarf if those will be needed by
29815 (output_rnglists): Add DWO argument. If true, switch to
29816 debug_ranges_dwo_section rather than debug_ranges_section.
29817 Adjust l1/l2 label indexes. Only output the offset table when
29818 dwo is true and don't include in there the skeleton range
29819 entry if present. For -gsplit-dwarf, skip ranges that belong
29820 to the other rnglists section. Change return type from void
29821 to bool and return true if there are any range entries for
29822 the other section. For dwarf_split_debug_info use
29823 DW_RLE_startx_endx, DW_RLE_startx_length and DW_RLE_base_addressx
29824 entries instead of DW_RLE_start_end, DW_RLE_start_length and
29825 DW_RLE_base_address. Use use_distinct_base_address_for_range.
29826 (init_sections_and_labels): Initialize debug_ranges_dwo_section
29827 if -gsplit-dwarf and DWARF >= 5. Adjust ranges_section_label
29828 and range_base_label indexes.
29829 (dwarf2out_finish): Call index_rnglists earlier before finalizing
29830 .debug_addr. Never emit DW_AT_rnglists_base attribute. For
29831 -gsplit-dwarf and DWARF >= 5 call output_rnglists up to twice
29832 with different dwo arguments.
29833 (dwarf2out_c_finalize): Clear debug_ranges_dwo_section.
29835 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
29837 PR tree-optimization/98268
29838 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Call
29839 recompute_tree_invariant_for_addr_expr after successfully
29840 folding a TARGET_MEM_REF that occurs inside an ADDR_EXPR.
29842 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
29844 PR tree-optimization/99726
29845 * tree-data-ref.c (create_intersect_range_checks_index): Bail
29846 out if there is more than one access function SCEV for the loop
29849 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
29851 PR rtl-optimization/97141
29852 PR rtl-optimization/98726
29853 * emit-rtl.c (valid_for_const_vector_p): Return true for
29855 * rtx-vector-builder.h (rtx_vector_builder::step): Return a
29856 poly_wide_int instead of a wide_int.
29857 (rtx_vector_builder::apply_set): Take a poly_wide_int instead
29859 * rtx-vector-builder.c (rtx_vector_builder::apply_set): Likewise.
29860 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Return
29861 false for CONST_VECTORs that cannot be forced to memory.
29862 * config/aarch64/aarch64-simd.md (mov<mode>): If a CONST_VECTOR
29863 is too complex to force to memory, build it up from individual
29866 2021-03-31 Jan Hubicka <jh@suse.cz>
29869 * cgraph.c (cgraph_node::release_body): Fix overactive check.
29871 2021-03-31 Christophe Lyon <christophe.lyon@linaro.org>
29874 * config/arm/vec-common.md (mul<mode>3): Disable on iwMMXT, expect
29877 2021-03-31 H.J. Lu <hjl.tools@gmail.com>
29879 * config/i386/i386-expand.c (expand_set_or_cpymem_via_rep):
29880 For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, don't convert QImode
29882 (decide_alg): For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, use
29883 "rep movsb/stosb" only for known sizes.
29884 * config/i386/i386-options.c (processor_cost_table): Use Ice
29885 Lake cost for Cannon Lake, Ice Lake, Tiger Lake, Sapphire
29886 Rapids and Alder Lake.
29887 * config/i386/i386.h (TARGET_PREFER_KNOWN_REP_MOVSB_STOSB): New.
29888 * config/i386/x86-tune-costs.h (icelake_memcpy): New.
29889 (icelake_memset): Likewise.
29890 (icelake_cost): Likewise.
29891 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
29894 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
29897 * config/aarch64/aarch64.c
29898 (aarch64_vectorize_preferred_vector_alignment): Query the size
29899 of the provided SVE vector; do not assume that all SVE vectors
29900 have the same size.
29902 2021-03-31 Jan Hubicka <jh@suse.cz>
29905 * cgraph.c (cgraph_node::release_body): Remove all callers and
29907 * cgraphclones.c (cgraph_node::materialize_clone): Do not do it here.
29908 * cgraphunit.c (cgraph_node::expand): And here.
29910 2021-03-31 Martin Liska <mliska@suse.cz>
29912 * ipa-modref.c (analyze_ssa_name_flags): Fix coding style
29913 and one negated condition.
29915 2021-03-31 Jakub Jelinek <jakub@redhat.com>
29916 Richard Sandiford <richard.sandiford@arm.com>
29919 * config/aarch64/aarch64.md (*add<mode>3_poly_1): Swap Uai and Uav
29920 constraints on operands[2] and similarly 0 and rk constraints
29921 on operands[1] corresponding to that.
29923 2021-03-31 Jakub Jelinek <jakub@redhat.com>
29926 * configure.ac (HAVE_LD_BROKEN_PE_DWARF5): New AC_DEFINE if PECOFF
29927 linker doesn't support DWARF sections new in DWARF5.
29928 * config/i386/i386-options.c (ix86_option_override_internal): Default
29929 to dwarf_version 4 if HAVE_LD_BROKEN_PE_DWARF5 for TARGET_PECOFF
29931 * config.in: Regenerated.
29932 * configure: Regenerated.
29934 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29937 * config/aarch64/aarch64.c (aarch64_analyze_loop_vinfo): Check for
29938 available issue_info before using it.
29940 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29943 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): Do not allow zero
29946 2021-03-30 Xionghu Luo <luoxhu@linux.ibm.com>
29949 * config/rs6000/altivec.md (altivec_lvsl_reg): Change to ...
29950 (altivec_lvsl_reg_<mode>): ... this.
29951 (altivec_lvsr_reg): Change to ...
29952 (altivec_lvsr_reg_<mode>): ... this.
29953 * config/rs6000/predicates.md (vec_set_index_operand): New.
29954 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
29955 Enable 32bit variable vec_insert for all TARGET_VSX.
29956 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
29957 Enable 32bit variable vec_insert for p9 and above.
29958 (rs6000_expand_vector_set_var_p8): Rename to ...
29959 (rs6000_expand_vector_set_var_p7): ... this.
29960 (rs6000_expand_vector_set): Use TARGET_VSX and adjust assert
29962 * config/rs6000/vector.md (vec_set<mode>): Use vec_set_index_operand.
29963 * config/rs6000/vsx.md (xl_len_r): Use gen_altivec_lvsl_reg_di and
29964 gen_altivec_lvsr_reg_di.
29966 2021-03-30 H.J. Lu <hjl.tools@gmail.com>
29969 * config/i386/ia32intrin.h (__rdtsc): Defined as macro.
29970 (__rdtscp): Likewise.
29972 2021-03-30 Tamar Christina <tamar.christina@arm.com>
29974 PR tree-optimization/99825
29975 * tree-vect-slp-patterns.c (vect_check_evenodd_blend):
29976 Reject non-mult 2 lanes.
29978 2021-03-30 Richard Earnshaw <rearnsha@arm.com>
29981 * config/arm/arm.c (arm_file_start): Fix emission of
29982 Tag_ABI_VFP_args attribute.
29984 2021-03-30 Richard Biener <rguenther@suse.de>
29986 PR tree-optimization/99824
29987 * stor-layout.c (set_min_and_max_values_for_integral_type):
29988 Assert the precision is within the bounds of
29989 WIDE_INT_MAX_PRECISION.
29990 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
29991 the outermost component ref only to lower the access size
29992 and initialize that from the access type.
29994 2021-03-30 Richard Sandiford <richard.sandiford@arm.com>
29997 * config/aarch64/aarch64.md (mov<mode>): Pass multi-instruction
29998 CONST_INTs to aarch64_expand_mov_immediate when called after RA.
30000 2021-03-30 Mihailo Stojanovic <mihailo.stojanovic@typhoon-hil.com>
30002 * config/aarch64/aarch64.md
30003 (<optab>_trunc<fcvt_target><GPI:mode>2): Set the "arch"
30004 attribute to disambiguate between SIMD and FP variants of the
30007 2021-03-29 Jan Hubicka <hubicka@ucw.cz>
30009 * ipa-modref.c (merge_call_lhs_flags): Correct handling of deref.
30010 (analyze_ssa_name_flags): Fix typo in comment.
30012 2021-03-29 Alex Coplan <alex.coplan@arm.com>
30015 * config/aarch64/aarch64-sve-builtins.cc
30016 (function_builder::add_function): Add placeholder_p argument, use
30017 placeholder decls if this is set.
30018 (function_builder::add_unique_function): Instead of conditionally adding
30019 direct overloads, unconditionally add either a direct overload or a
30021 (function_builder::add_overloaded_function): Set placeholder_p if we're
30022 using C++ overloads. Use the obstack for string storage instead
30023 of relying on the tree nodes.
30024 (function_builder::add_overloaded_functions): Don't return early for
30025 m_direct_overloads: we need to add placeholders.
30026 * config/aarch64/aarch64-sve-builtins.h
30027 (function_builder::add_function): Add placeholder_p argument.
30029 2021-03-29 Richard Biener <rguenther@suse.de>
30031 PR tree-optimization/99807
30032 * tree-vect-slp.c (vect_slp_analyze_node_operations_1): Move
30033 assert below VEC_PERM handling.
30035 2021-03-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30038 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>): Use
30039 aarch64_simd_or_scalar_imm_zero to match zeroes. Remove pattern
30040 matching const_int 0.
30041 (move_lo_quad_internal_be_<mode>): Likewise.
30042 (move_lo_quad_<mode>): Update for the above.
30043 * config/aarch64/iterators.md (VQ_2E): Delete.
30045 2021-03-29 Jakub Jelinek <jakub@redhat.com>
30047 PR tree-optimization/99777
30048 * fold-const.c (extract_muldiv_1): For conversions, punt on casts from
30049 types other than scalar integral types.
30051 2021-03-28 David Edelsohn <dje.gcc@gmail.com>
30053 * config/rs6000/rs6000.c (rs6000_output_dwarf_dtprel): Do not add
30054 XCOFF TLS reloc decorations.
30056 2021-03-28 Gerald Pfeifer <gerald@pfeifer.com>
30058 * doc/analyzer.texi (Analyzer Internals): Update link to
30059 "A Memory Model for Static Analysis of C Programs".
30061 2021-03-26 David Edelsohn <dje.gcc@gmail.com>
30063 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Call function.
30064 * config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align):
30066 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align): New.
30067 (rs6000_special_round_type_align): Recursively check innermost first
30070 2021-03-26 Jakub Jelinek <jakub@redhat.com>
30073 * dwarf2out.h (struct dw_fde_node): Add rule18 member.
30074 * dwarf2cfi.c (dwarf2out_frame_debug_expr): When handling (set hfp sp)
30075 assignment with drap_reg active, queue reg save for hfp with offset 0
30076 and flush queued reg saves. When handling a push with rule18,
30077 defer queueing reg save for hfp and just assert the offset is 0.
30078 (scan_trace): Assert that fde->rule18 is false.
30080 2021-03-26 Vladimir Makarov <vmakarov@redhat.com>
30083 * ira-costs.c (record_reg_classes): Put case with
30084 CT_RELAXED_MEMORY adjacent to one with CT_MEMORY.
30085 * ira.c (ira_setup_alts): Ditto.
30086 * lra-constraints.c (process_alt_operands): Ditto.
30087 * recog.c (asm_operand_ok): Ditto.
30088 * reload.c (find_reloads): Ditto.
30090 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30092 * config/aarch64/aarch64-protos.h
30093 (cpu_addrcost_table::post_modify_ld3_st3): New member variable.
30094 (cpu_addrcost_table::post_modify_ld4_st4): Likewise.
30095 * config/aarch64/aarch64.c (generic_addrcost_table): Update
30096 accordingly, using the same costs as for post_modify.
30097 (exynosm1_addrcost_table, xgene1_addrcost_table): Likewise.
30098 (thunderx2t99_addrcost_table, thunderx3t110_addrcost_table):
30099 (tsv110_addrcost_table, qdf24xx_addrcost_table): Likewise.
30100 (a64fx_addrcost_table): Likewise.
30101 (neoversev1_addrcost_table): New.
30102 (neoversev1_tunings): Use neoversev1_addrcost_table.
30103 (aarch64_address_cost): Use the new post_modify costs for CImode
30106 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30108 * config/aarch64/aarch64.opt
30109 (-param=aarch64-loop-vect-issue-rate-niters=): New parameter.
30110 * doc/invoke.texi: Document it.
30111 * config/aarch64/aarch64-protos.h (aarch64_base_vec_issue_info)
30112 (aarch64_scalar_vec_issue_info, aarch64_simd_vec_issue_info)
30113 (aarch64_advsimd_vec_issue_info, aarch64_sve_vec_issue_info)
30114 (aarch64_vec_issue_info): New structures.
30115 (cpu_vector_cost): Write comments above the variables rather
30117 (cpu_vector_cost::issue_info): New member variable.
30118 * config/aarch64/aarch64.c: Include gimple-pretty-print.h
30119 and tree-ssa-loop-niter.h.
30120 (generic_vector_cost, a64fx_vector_cost, qdf24xx_vector_cost)
30121 (thunderx_vector_cost, tsv110_vector_cost, cortexa57_vector_cost)
30122 (exynosm1_vector_cost, xgene1_vector_cost, thunderx2t99_vector_cost)
30123 (thunderx3t110_vector_cost): Initialize issue_info to null.
30124 (neoversev1_scalar_issue_info, neoversev1_advsimd_issue_info)
30125 (neoversev1_sve_issue_info, neoversev1_vec_issue_info): New structures.
30126 (neoversev1_vector_cost): Use them.
30127 (aarch64_vec_op_count, aarch64_sve_op_count): New structures.
30128 (aarch64_vector_costs::saw_sve_only_op): New member variable.
30129 (aarch64_vector_costs::num_vector_iterations): Likewise.
30130 (aarch64_vector_costs::scalar_ops): Likewise.
30131 (aarch64_vector_costs::advsimd_ops): Likewise.
30132 (aarch64_vector_costs::sve_ops): Likewise.
30133 (aarch64_vector_costs::seen_loads): Likewise.
30134 (aarch64_simd_vec_costs_for_flags): New function.
30135 (aarch64_analyze_loop_vinfo): Initialize num_vector_iterations.
30136 Count the number of predicate operations required by SVE WHILE
30138 (aarch64_comparison_type, aarch64_multiply_add_p): New functions.
30139 (aarch64_sve_only_stmt_p, aarch64_in_loop_reduction_latency): Likewise.
30140 (aarch64_count_ops): Likewise.
30141 (aarch64_add_stmt_cost): Record whether see an SVE operation
30142 that cannot currently be implementing using Advanced SIMD.
30143 Record issue information about the scalar, Advanced SIMD
30144 and (where relevant) SVE versions of a loop.
30145 (aarch64_vec_op_count::dump): New function.
30146 (aarch64_sve_op_count::dump): Likewise.
30147 (aarch64_estimate_min_cycles_per_iter): Likewise.
30148 (aarch64_adjust_body_cost): If issue information is available,
30149 try to compare the issue rates of the various loop implementations
30150 and increase or decrease the vector body cost accordingly.
30152 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30154 * config/aarch64/aarch64.c (aarch64_detect_vector_stmt_subtype):
30155 Assume a zero cost for induction phis.
30157 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30159 * config/aarch64/aarch64.c (aarch64_embedded_comparison_type): New
30161 (aarch64_adjust_stmt_cost): Add the costs of embedded scalar and
30162 vector comparisons.
30164 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30166 * config/aarch64/aarch64.c (aarch64_detect_scalar_stmt_subtype):
30168 (aarch64_add_stmt_cost): Call it.
30170 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30172 * config/aarch64/aarch64-tuning-flags.def (matched_vector_throughput):
30173 New tuning parameter.
30174 * config/aarch64/aarch64.c (neoversev1_tunings): Use it.
30175 (aarch64_estimated_sve_vq): New function.
30176 (aarch64_vector_costs::analyzed_vinfo): New member variable.
30177 (aarch64_vector_costs::is_loop): Likewise.
30178 (aarch64_vector_costs::unrolled_advsimd_niters): Likewise.
30179 (aarch64_vector_costs::unrolled_advsimd_stmts): Likewise.
30180 (aarch64_record_potential_advsimd_unrolling): New function.
30181 (aarch64_analyze_loop_vinfo, aarch64_analyze_bb_vinfo): Likewise.
30182 (aarch64_add_stmt_cost): Call aarch64_analyze_loop_vinfo or
30183 aarch64_analyze_bb_vinfo on the first use of a costs structure.
30184 Detect whether we're vectorizing a loop for SVE that might be
30185 completely unrolled if it used Advanced SIMD instead.
30186 (aarch64_adjust_body_cost_for_latency): New function.
30187 (aarch64_finish_cost): Call it.
30189 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30191 * config/aarch64/aarch64.c (aarch64_vector_costs): New structure.
30192 (aarch64_init_cost): New function.
30193 (aarch64_add_stmt_cost): Use aarch64_vector_costs instead of
30194 the default unsigned[3].
30195 (aarch64_finish_cost, aarch64_destroy_cost_data): New functions.
30196 (TARGET_VECTORIZE_INIT_COST): Override.
30197 (TARGET_VECTORIZE_FINISH_COST): Likewise.
30198 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
30200 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30202 * config/aarch64/aarch64.c (neoversev1_advsimd_vector_cost)
30203 (neoversev1_sve_vector_cost): New cost structures.
30204 (neoversev1_vector_cost): Likewise.
30205 (neoversev1_tunings): Use them. Enable use_new_vector_costs.
30207 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30209 * config/aarch64/aarch64-protos.h
30210 (sve_vec_cost::scatter_store_elt_cost): New member variable.
30211 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
30212 accordingly, taking the cost from the cost of a scalar_store.
30213 (a64fx_sve_vector_cost): Likewise.
30214 (aarch64_detect_vector_stmt_subtype): Detect scatter stores.
30216 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30218 * config/aarch64/aarch64-protos.h
30219 (simd_vec_cost::store_elt_extra_cost): New member variable.
30220 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
30221 accordingly, using the vec_to_scalar cost for the new field.
30222 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
30223 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
30224 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
30225 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
30226 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
30227 (thunderx3t110_advsimd_vector_cost): Likewise.
30228 (aarch64_detect_vector_stmt_subtype): Detect single-element stores.
30230 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30232 * config/aarch64/aarch64-protos.h (simd_vec_cost::ld2_st2_permute_cost)
30233 (simd_vec_cost::ld3_st3_permute_cost): New member variables.
30234 (simd_vec_cost::ld4_st4_permute_cost): Likewise.
30235 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
30236 accordingly, using zero for the new costs.
30237 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
30238 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
30239 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
30240 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
30241 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
30242 (thunderx3t110_advsimd_vector_cost): Likewise.
30243 (aarch64_ld234_st234_vectors): New function.
30244 (aarch64_adjust_stmt_cost): Likewise.
30245 (aarch64_add_stmt_cost): Call aarch64_adjust_stmt_cost if using
30246 the new vector costs.
30248 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30250 * config/aarch64/aarch64-protos.h (sve_vec_cost): Turn into a
30251 derived class of simd_vec_cost. Add information about CLAST[AB]
30252 and FADDA instructions.
30253 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
30254 accordingly, using the vec_to_scalar costs for the new fields.
30255 (a64fx_sve_vector_cost): Likewise.
30256 (aarch64_reduc_type): New function.
30257 (aarch64_sve_in_loop_reduction_latency): Likewise.
30258 (aarch64_detect_vector_stmt_subtype): Take a vinfo parameter.
30259 Use aarch64_sve_in_loop_reduction_latency to handle SVE reductions
30260 that occur in the loop body.
30261 (aarch64_add_stmt_cost): Update call accordingly.
30263 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
30265 * config/aarch64/aarch64-tuning-flags.def (use_new_vector_costs):
30267 * config/aarch64/aarch64-protos.h (simd_vec_cost): Put comments
30268 above the fields rather than to the right.
30269 (simd_vec_cost::reduc_i8_cost): New member variable.
30270 (simd_vec_cost::reduc_i16_cost): Likewise.
30271 (simd_vec_cost::reduc_i32_cost): Likewise.
30272 (simd_vec_cost::reduc_i64_cost): Likewise.
30273 (simd_vec_cost::reduc_f16_cost): Likewise.
30274 (simd_vec_cost::reduc_f32_cost): Likewise.
30275 (simd_vec_cost::reduc_f64_cost): Likewise.
30276 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
30277 accordingly, using the vec_to_scalar_cost for the new fields.
30278 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
30279 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
30280 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
30281 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
30282 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
30283 (thunderx3t110_advsimd_vector_cost): Likewise.
30284 (aarch64_use_new_vector_costs_p): New function.
30285 (aarch64_simd_vec_costs): New function, split out from...
30286 (aarch64_builtin_vectorization_cost): ...here.
30287 (aarch64_is_reduction): New function.
30288 (aarch64_detect_vector_stmt_subtype): Likewise.
30289 (aarch64_add_stmt_cost): Call aarch64_detect_vector_stmt_subtype if
30290 using the new vector costs.
30292 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
30295 * tree-emutls.c (get_emutls_init_templ_addr): Mark initializer of weak
30296 TLS declarations as public.
30298 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
30300 * config/aarch64/aarch64-d.c (IN_TARGET_CODE): Define.
30301 * config/arm/arm-d.c (IN_TARGET_CODE): Likewise.
30302 * config/i386/i386-d.c (IN_TARGET_CODE): Likewise.
30303 * config/mips/mips-d.c (IN_TARGET_CODE): Likewise.
30304 * config/pa/pa-d.c (IN_TARGET_CODE): Likewise.
30305 * config/riscv/riscv-d.c (IN_TARGET_CODE): Likewise.
30306 * config/rs6000/rs6000-d.c (IN_TARGET_CODE): Likewise.
30307 * config/s390/s390-d.c (IN_TARGET_CODE): Likewise.
30308 * config/sparc/sparc-d.c (IN_TARGET_CODE): Likewise.
30310 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
30313 * config.gcc (*-*-cygwin*): Add winnt-d.o
30314 (*-*-mingw*): Likewise.
30315 * config/i386/cygwin.h (EXTRA_TARGET_D_OS_VERSIONS): New macro.
30316 * config/i386/mingw32.h (EXTRA_TARGET_D_OS_VERSIONS): Likewise.
30317 * config/i386/t-cygming: Add winnt-d.o.
30318 * config/i386/winnt-d.c: New file.
30320 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
30322 * config/freebsd-d.c: Include memmodel.h.
30324 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
30327 * config.gcc (*-*-openbsd*): Add openbsd-d.o.
30328 * config/t-openbsd: Add openbsd-d.o.
30329 * config/openbsd-d.c: New file.
30331 2021-03-25 Stam Markianos-Wright <stam.markianos-wright@arm.com>
30333 PR tree-optimization/96974
30334 * tree-vect-stmts.c (vect_get_vector_types_for_stmt): Replace assert
30335 with graceful exit.
30337 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
30340 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
30344 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
30345 always_inline in system headers.
30347 2021-03-25 Kewen Lin <linkw@linux.ibm.com>
30349 * tree-vect-loop.c (vect_model_reduction_cost): Init inside_cost.
30351 2021-03-25 Jakub Jelinek <jakub@redhat.com>
30354 * tree-core.h (enum operand_equal_flag): Add OEP_ADDRESS_OF_SAME_FIELD.
30355 * fold-const.c (operand_compare::operand_equal_p): Don't compare
30356 field offsets if OEP_ADDRESS_OF_SAME_FIELD.
30358 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
30362 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
30363 always_inline in system headers.
30365 2021-03-25 Richard Biener <rguenther@suse.de>
30367 PR tree-optimization/99746
30368 * tree-vect-slp-patterns.c (complex_pattern::build): Do not mark
30369 the scalar stmt as patterned. Instead set up required things
30372 2021-03-25 Xionghu Luo <luoxhu@linux.ibm.com>
30374 * config/rs6000/rs6000.c (power8_costs): Change l2 cache
30377 2021-03-24 Martin Liska <mliska@suse.cz>
30380 * common/config/i386/i386-common.c (ARRAY_SIZE): Fix off-by-one
30382 * config/i386/i386-options.c (ix86_option_override_internal):
30383 Add run-time assert.
30385 2021-03-24 Martin Jambor <mjambor@suse.cz>
30388 * ipa-cp.c (initialize_node_lattices): Mark as bottom all
30389 parameters with unknown type.
30390 (ipacp_value_safe_for_type): New function.
30391 (propagate_vals_across_arith_jfunc): Verify that the constant type
30392 can be used for a type of the formal parameter.
30393 (propagate_vals_across_ancestor): Likewise.
30394 (propagate_scalar_across_jump_function): Likewise. Pass the type
30395 also to propagate_vals_across_ancestor.
30397 2021-03-24 Christophe Lyon <christophe.lyon@linaro.org>
30400 * config/arm/mve.md (movmisalign<mode>_mve_store): Use Ux
30402 (movmisalign<mode>_mve_load): Likewise.
30404 2021-03-24 Jakub Jelinek <jakub@redhat.com>
30407 * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
30408 movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.
30410 2021-03-24 Alexandre Oliva <oliva@adacore.com>
30412 * doc/sourcebuild.texi (sysconf): New effective target.
30414 2021-03-24 Alexandre Oliva <oliva@adacore.com>
30416 * config/i386/predicates.md (reg_or_const_vec_operand): New.
30417 * config/i386/sse.md (ssse3_pshufbv8qi3): Add an expander for
30418 the now *-prefixed insn_and_split, turn the splitter const vec
30419 into an input for the insn, making it an ignored immediate for
30420 non-split cases, and loaded into the scratch register
30423 2021-03-23 Vladimir N. Makarov <vmakarov@redhat.com>
30426 * config/aarch64/constraints.md (Utq, UOb, UOh, UOw, UOd, UOty):
30427 Use define_relaxed_memory_constraint for them.
30429 2021-03-23 Iain Sandoe <iain@sandoe.co.uk>
30432 * config/host-darwin.c (darwin_gt_pch_use_address): Add a
30433 colon to the diagnostic message.
30435 2021-03-23 Ilya Leoshkevich <iii@linux.ibm.com>
30437 * fwprop.c (fwprop_propagation::fwprop_propagation): Look at
30439 (try_fwprop_subst_note): Use set_info instead of insn_info.
30440 (try_fwprop_subst_pattern): Likewise.
30441 (try_fwprop_subst_notes): Likewise.
30442 (try_fwprop_subst): Likewise.
30443 (forward_propagate_subreg): Likewise.
30444 (forward_propagate_and_simplify): Likewise.
30445 (forward_propagate_into): Likewise.
30446 * rtl-ssa/accesses.h (set_info::single_nondebug_use) New
30448 (set_info::single_nondebug_insn_use): Likewise.
30449 (set_info::single_phi_use): Likewise.
30450 * rtl-ssa/member-fns.inl (set_info::single_nondebug_use) New
30452 (set_info::single_nondebug_insn_use): Likewise.
30453 (set_info::single_phi_use): Likewise.
30455 2021-03-23 Christophe Lyon <christophe.lyon@linaro.org>
30457 * doc/sourcebuild.texi (arm_dsp_ok, arm_dsp): Document.
30459 2021-03-23 Jakub Jelinek <jakub@redhat.com>
30462 * config/aarch64/aarch64.c (aarch64_add_offset): Tell
30463 expand_mult to perform an unsigned rather than a signed
30466 2021-03-23 H.J. Lu <hjl.tools@gmail.com>
30469 * config/i386/cpuid.h (__cpuid): Add __volatile__.
30470 (__cpuid_count): Likewise.
30472 2021-03-23 Richard Biener <rguenther@suse.de>
30474 PR tree-optimization/99721
30475 * tree-vect-slp.c (vect_slp_analyze_node_operations):
30476 Make sure we can schedule the node.
30478 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
30480 * config/riscv/riscv.c (riscv_subword): Take endianness into
30481 account when calculating the byte offset.
30483 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
30485 * config/riscv/predicates.md (subreg_lowpart_operator): New predicate
30486 * config/riscv/riscv.md (*addsi3_extended2, *subsi3_extended2)
30487 (*negsi2_extended2, *mulsi3_extended2, *<optab>si3_mask)
30488 (*<optab>si3_mask_1, *<optab>di3_mask, *<optab>di3_mask_1)
30489 (*<optab>si3_extend_mask, *<optab>si3_extend_mask_1): Use
30490 new predicate "subreg_lowpart_operator"
30492 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
30494 * config/riscv/riscv.c (riscv_swap_instruction): New function
30495 to byteswap an SImode rtx containing an instruction.
30496 (riscv_trampoline_init): Byteswap the generated instructions
30499 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
30501 * common/config/riscv/riscv-common.c
30502 (TARGET_DEFAULT_TARGET_FLAGS): Set default endianness.
30503 * config.gcc (riscv32be-*, riscv64be-*): Set
30504 TARGET_BIG_ENDIAN_DEFAULT to 1.
30505 * config/riscv/elf.h (LINK_SPEC): Change -melf* value
30506 depending on default endianness.
30507 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
30508 * config/riscv/linux.h (LINK_SPEC): Likewise.
30509 * config/riscv/riscv.c (TARGET_DEFAULT_TARGET_FLAGS): Set
30510 default endianness.
30511 * config/riscv/riscv.h (DEFAULT_ENDIAN_SPEC): New macro.
30513 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
30515 * config/riscv/elf.h (LINK_SPEC): Pass linker endianness flag.
30516 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
30517 * config/riscv/linux.h (LINK_SPEC): Likewise.
30518 * config/riscv/riscv.h (ASM_SPEC): Pass -mbig-endian and
30520 (BYTES_BIG_ENDIAN): Handle big endian.
30521 (WORDS_BIG_ENDIAN): Define to BYTES_BIG_ENDIAN.
30522 * config/riscv/riscv.opt (-mbig-endian, -mlittle-endian): New
30524 * doc/invoke.texi (-mbig-endian, -mlittle-endian): Document.
30526 2021-03-23 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
30528 * regcprop.c (find_oldest_value_reg): Ask target whether
30529 different mode is fine for replacement register.
30531 2021-03-23 Aldy Hernandez <aldyh@redhat.com>
30533 PR tree-optimization/99296
30534 * value-range.cc (irange::irange_set_1bit_anti_range): New.
30535 (irange::irange_set_anti_range): Call irange_set_1bit_anti_range
30536 * value-range.h (irange::irange_set_1bit_anti_range): New.
30538 2021-03-22 Vladimir N. Makarov <vmakarov@redhat.com>
30541 * config/aarch64/constraints.md (UtQ): Use
30542 define_relaxed_memory_constraint for it.
30543 * doc/md.texi (define_relaxed_memory_constraint): Describe it.
30544 * genoutput.c (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
30545 * genpreds.c (constraint_data): Add bitfield is_relaxed_memory.
30546 (have_relaxed_memory_constraints): New static var.
30547 (relaxed_memory_start, relaxed_memory_end): Ditto.
30548 (add_constraint): Add arg is_relaxed_memory. Check name for
30549 relaxed memory. Set up is_relaxed_memory in constraint_data and
30550 have_relaxed_memory_constraints. Adjust calls.
30551 (choose_enum_order): Process relaxed memory.
30552 (write_tm_preds_h): Ditto.
30553 (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
30554 * gensupport.c (process_rtx): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
30555 * ira-costs.c (record_reg_classes): Process CT_RELAXED_MEMORY.
30556 * ira-lives.c (single_reg_class): Use
30557 insn_extra_relaxed_memory_constraint.
30558 * ira.c (ira_setup_alts): CT_RELAXED_MEMORY.
30559 * lra-constraints.c (valid_address_p): Use
30560 insn_extra_relaxed_memory_constraint instead of other memory
30562 (process_alt_operands): Process CT_RELAXED_MEMORY.
30563 (curr_insn_transform): Use insn_extra_relaxed_memory_constraint.
30564 * recog.c (asm_operand_ok, preprocess_constraints): Process
30566 * reload.c (find_reloads): Ditto.
30567 * rtl.def (DEFINE_RELAXED_MEMORY_CONSTRAINT): New.
30568 * stmt.c (parse_input_constraint): Use
30569 insn_extra_relaxed_memory_constraint.
30571 2021-03-22 Segher Boessenkool <segher@kernel.crashing.org>
30574 * ubsan.c (ubsan_instrument_float_cast): Don't test for unordered if
30577 2021-03-22 Alex Coplan <alex.coplan@arm.com>
30580 * config/arm/arm-protos.h (neon_make_constant): Add generate
30581 argument to guard emitting insns, default to true.
30582 * config/arm/arm.c (arm_legitimate_constant_p_1): Reject
30583 CONST_VECTORs which neon_make_constant can't handle.
30584 (neon_vdup_constant): Add generate argument, avoid emitting
30585 insns if it's not set.
30586 (neon_make_constant): Plumb new generate argument through.
30587 * config/arm/constraints.md (Ui): New. Use it...
30588 * config/arm/mve.md (*mve_mov<mode>): ... here.
30589 * config/arm/vec-common.md (movv8hf): Use neon_make_constant to
30590 synthesize constants.
30592 2021-03-22 Richard Biener <rguenther@suse.de>
30594 * debug.h: Add deprecation warning.
30596 2021-03-22 Richard Biener <rguenther@suse.de>
30598 PR tree-optimization/99694
30599 * tree-ssa-sccvn.c (visit_phi): Ignore edges with the
30602 2021-03-22 Kito Cheng <kito.cheng@sifive.com>
30605 * config/riscv/riscv.c (riscv_expand_block_move): Get RTL value
30606 after type checking.
30608 2021-03-22 Jakub Jelinek <jakub@redhat.com>
30612 * dwarf2out.c (get_full_len): Use get_precision rather than
30614 (add_const_value_attribute): Make sure add_AT_wide argument has
30615 precision prec rather than some very wide one.
30617 2021-03-22 Kewen Lin <linkw@linux.ibm.com>
30619 * config/rs6000/rs6000.md (*rotldi3_insert_sf,
30620 *mov<SFDF:mode><SFDF2:mode>cc_p9, floatsi<mode>2_lfiwax,
30621 floatsi<mode>2_lfiwax_mem, floatunssi<mode>2_lfiwzx,
30622 floatunssi<mode>2_lfiwzx_mem, *floatsidf2_internal,
30623 *floatunssidf2_internal, fix_trunc<mode>si2_stfiwx,
30624 fix_trunc<mode>si2_internal, fixuns_trunc<mode>si2_stfiwx,
30625 *round32<mode>2_fprs, *roundu32<mode>2_fprs,
30626 *fix_trunc<mode>si2_internal): Fix empty split condition.
30627 * config/rs6000/vsx.md (*vsx_le_undo_permute_<mode>,
30628 vsx_reduc_<VEC_reduc_name>_v2df, vsx_reduc_<VEC_reduc_name>_v4sf,
30629 *vsx_reduc_<VEC_reduc_name>_v2df_scalar,
30630 *vsx_reduc_<VEC_reduc_name>_v4sf_scalar): Likewise.
30632 2021-03-22 Xionghu Luo <luoxhu@linux.ibm.com>
30635 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
30636 Convert idx to DImode.
30637 (rs6000_expand_vector_set_var_p8): Likewise.
30639 2021-03-21 Jakub Jelinek <jakub@redhat.com>
30642 * dwarf2out.c (insert_float): Change return type from void to
30643 unsigned, handle GET_MODE_SIZE (mode) == 2 and return element size.
30644 (mem_loc_descriptor, loc_descriptor, add_const_value_attribute):
30647 2021-03-20 H.J. Lu <hjl.tools@gmail.com>
30650 * config/i386/i386.c (construct_container): Check cfun != NULL
30651 before accessing silent_p.
30653 2021-03-20 Ahamed Husni <ahamedhusni73@gmail.com>
30655 * asan.c: Fix typos in comments.
30657 2021-03-20 Vladimir N. Makarov <vmakarov@redhat.com>
30659 PR rtl-optimization/99680
30660 * lra-constraints.c (skip_contraint_modifiers): Rename to skip_constraint_modifiers.
30661 (process_address_1): Check empty constraint before using
30664 2021-03-19 Pat Haugen <pthaugen@linux.ibm.com>
30666 * config/rs6000/rs6000.c (power10_cost): New.
30667 (rs6000_option_override_internal): Set Power10 costs.
30668 (rs6000_issue_rate): Set Power10 issue rate.
30669 * config/rs6000/power10.md: Rewrite for Power10.
30671 2021-03-19 Vladimir N. Makarov <vmakarov@redhat.com>
30674 * lra-constraints.c (process_address_1): Don't use unknown
30675 constraint for address constraint.
30677 2021-03-19 Iain Sandoe <iain@sandoe.co.uk>
30680 * config.gcc (powerpc-*-darwin8): Delete the reference to
30681 the now removed darwin8.h.
30683 2021-03-19 Olivier Hainque <hainque@adacore.com>
30686 * config/vxworksae.h (VX_CPU_PREFIX): Define.
30688 2021-03-19 John David Anglin <danglin@gcc.gnu.org>
30690 * config/pa/pa.c (import_milli): Use memcpy instead of strncpy.
30692 2021-03-19 Tamar Christina <tamar.christina@arm.com>
30694 PR tree-optimization/99656
30695 * tree-vect-slp-patterns.c (linear_loads_p,
30696 complex_add_pattern::matches, is_eq_or_top,
30697 vect_validate_multiplication, complex_mul_pattern::matches,
30698 complex_fms_pattern::matches): Remove complex_perm_kinds_t.
30699 * tree-vectorizer.h: (complex_load_perm_t): Removed.
30700 (slp_tree_to_load_perm_map_t): Use complex_perm_kinds_t instead of
30701 complex_load_perm_t.
30703 2021-03-19 H.J. Lu <hjl.tools@gmail.com>
30706 * config/i386/i386-options.c (ix86_init_machine_status): Set
30708 * config/i386/i386.c (init_cumulative_args): Set silent_p to
30710 (construct_container): Return early for return and argument
30711 errors if silent_p is true.
30712 * config/i386/i386.h (machine_function): Add silent_p.
30714 2021-03-19 Jakub Jelinek <jakub@redhat.com>
30717 * config/arm/constraints.md (Ds): New constraint.
30718 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): Use w,Ds
30719 constraint instead of w,Dm.
30721 2021-03-19 Andrew Stubbs <ams@codesourcery.com>
30723 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Fix quotes
30726 2021-03-19 Eric Botcazou <ebotcazou@adacore.com>
30728 PR middle-end/99641
30729 * fold-const.c (native_encode_initializer) <CONSTRUCTOR>: For an
30730 array type, do the computation of the current position in sizetype.
30732 2021-03-18 Vladimir N. Makarov <vmakarov@redhat.com>
30735 * lra-constraints.c (process_address_1): Use lookup_constraint
30736 only for a single constraint.
30738 2021-03-18 Martin Sebor <msebor@redhat.com>
30740 PR middle-end/99502
30741 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): Rename...
30742 (inbounds_memaccess_p): ...to this. Check the ending offset of
30743 the accessed member.
30745 2021-03-18 Andrew Stubbs <ams@codesourcery.com>
30747 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Add %< and
30748 %> quote markers to error messages.
30749 (gcn_goacc_validate_dims): Likewise.
30750 (gcn_conditional_register_usage): Remove exclaimation mark from error
30752 (gcn_vectorize_vec_perm_const): Ensure perm is fully uninitialized.
30754 2021-03-18 Jan Hubicka <hubicka@ucw.cz>
30756 * config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
30759 2021-03-18 Sinan Lin <sinan@isrc.iscas.ac.cn>
30760 Kito Cheng <kito.cheng@sifive.com>
30762 * config/riscv/riscv.c (riscv_block_move_straight): Change type
30763 to unsigned HOST_WIDE_INT for parameter and local variable with
30764 HOST_WIDE_INT type.
30765 (riscv_adjust_block_mem): Ditto.
30766 (riscv_block_move_loop): Ditto.
30767 (riscv_expand_block_move): Ditto.
30769 2021-03-18 Nick Clifton <nickc@redhat.com>
30771 * config/v850/v850.c (construct_restore_jr): Increase static
30773 (construct_save_jarl): Likewise.
30774 * config/v850/v850.h (DWARF2_DEBUGGING_INFO): Define.
30776 2021-03-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30778 * config/aarch64/aarch64.c (aarch64_adjust_generic_arch_tuning): Define.
30779 (aarch64_override_options_internal): Use it.
30780 (generic_tunings): Add AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS to
30783 2021-03-17 Sandra Loosemore <sandra@codesourcery.com>
30785 * config/nios2/nios2.c (nios2_custom_check_insns): Clean up
30786 error message format issues.
30787 (nios2_option_override): Likewise.
30788 (nios2_expand_fpu_builtin): Likewise.
30789 (nios2_init_custom_builtins): Adjust to avoid bogus strncpy
30790 truncation warning.
30791 (nios2_expand_custom_builtin): More error message format fixes.
30792 (nios2_expand_rdwrctl_builtin): Likewise.
30793 (nios2_expand_rdprs_builtin): Likewise.
30794 (nios2_expand_eni_builtin): Likewise.
30795 (nios2_expand_builtin): Likewise.
30796 (nios2_register_custom_code): Likewise.
30797 (nios2_valid_target_attribute_rec): Likewise.
30798 (nios2_add_insn_asm): Fix uninitialized variable warning.
30800 2021-03-17 Jan Hubicka <jh@suse.cz>
30802 * config/i386/x86-tune-costs.h (struct processor_costs): Update costs
30803 of gather to match reality.
30804 * config/i386/x86-tune.def (X86_TUNE_USE_GATHER): Enable for znver3.
30806 2021-03-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30808 * config/aarch64/aarch64-builtins.c (aarch64_expand_rng_builtin): Use EQ
30809 to compare against CC_REG rather than NE.
30811 2021-03-17 H.J. Lu <hjl.tools@gmail.com>
30814 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support
30815 inline assembly statements.
30816 (ix86_print_operand): Update 'P' handling for -fno-plt.
30818 2021-03-17 Tamar Christina <tamar.christina@arm.com>
30821 * config/aarch64/aarch64.c
30822 (aarch64_simd_clone_compute_vecsize_and_simdlen): Remove unused var.
30824 2021-03-16 Segher Boessenkool <segher@kernel.crashing.org>
30827 * config/rs6000/predicates.md (branch_comparison_operator): Allow
30828 ordered and unordered for CCFPmode, if flag_finite_math_only.
30830 2021-03-16 Jakub Jelinek <jakub@redhat.com>
30833 * config/i386/i386-expand.c (ix86_split_lea_for_addr): Emit a MULT
30834 rather than ASHIFT.
30835 * config/i386/i386.md (mult by 1248 into ashift): New splitter.
30837 2021-03-16 Martin Liska <mliska@suse.cz>
30840 * optc-save-gen.awk: Add flag_ipa_ra to exceptions for
30841 cl_optimization_compare function.
30843 2021-03-16 Ilya Leoshkevich <iii@linux.ibm.com>
30845 * config/s390/s390.c (f_constraint_p): Treat "fv" constraints
30848 2021-03-16 Jakub Jelinek <jakub@redhat.com>
30851 * config/i386/i386.h (struct machine_function): Add
30852 has_explicit_vzeroupper bitfield.
30853 * config/i386/i386-expand.c (ix86_expand_builtin): Set
30854 cfun->machine->has_explicit_vzeroupper when expanding
30855 IX86_BUILTIN_VZEROUPPER.
30856 * config/i386/i386-features.c (rest_of_handle_insert_vzeroupper):
30857 Do the mode switching only when TARGET_VZEROUPPER, expensive
30858 optimizations turned on and not optimizing for size.
30859 (pass_insert_vzeroupper::gate): Enable even when
30860 cfun->machine->has_explicit_vzeroupper is set.
30862 2021-03-16 Jakub Jelinek <jakub@redhat.com>
30865 * config/aarch64/aarch64.c
30866 (aarch64_simd_clone_compute_vecsize_and_simdlen): If not a function
30867 definition, walk TYPE_ARG_TYPES list if non-NULL for argument types
30868 instead of DECL_ARGUMENTS. Ignore types for uniform arguments.
30870 2021-03-15 Richard Biener <rguenther@suse.de>
30872 PR tree-optimization/98834
30873 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle missing
30874 subsetting by truncating the access size.
30876 2021-03-15 Jan Hubicka <hubicka@ucw.cz>
30878 * config/i386/i386-options.c (processor_cost_table): Add znver3_cost.
30879 * config/i386/x86-tune-costs.h (znver3_cost): New gobal variable; copy
30882 2021-03-15 Martin Liska <mliska@suse.cz>
30884 * spellcheck.c: Add missing comma in initialization.
30886 2021-03-14 Uroš Bizjak <ubizjak@gmail.com>
30888 * config/i386/sse.md (*vec_extract<mode>): Merge alternative 0 with
30889 alternative 2 and alternative 1 with alternative 3 using
30890 YW register constraint.
30891 (*vec_extract<PEXTR_MODE12:mode>_zext): Merge alternatives
30892 using YW register constraint.
30893 (*vec_extractv16qi_zext): Ditto.
30894 (*vec_extractv4si): Merge alternatives 4 and 5
30895 using Yw register constraint.
30896 (*ssse3_palignr<mode>_perm): Use Yw instead of v for alternative 3.
30898 2021-03-13 Martin Sebor <msebor@redhat.com>
30900 PR tree-optimization/99489
30901 * builtins.c (gimple_call_alloc_size): Fail gracefully when argument
30902 is not a call statement.
30904 2021-03-13 Jakub Jelinek <jakub@redhat.com>
30906 PR tree-optimization/99544
30907 * match.pd (X + (X << C) -> X * (1 + (1 << C))): Don't simplify
30908 if for vector types multiplication can't be done in type's mode.
30910 2021-03-12 Eric Botcazou <ebotcazou@adacore.com>
30913 * config/sparc/constraints.md (w): Rename to...
30914 (W): ... this and ditch previous implementation.
30915 * config/sparc/sparc.md (*movdi_insn_sp64): Replace W with m.
30916 (*movdf_insn_sp64): Likewise.
30917 (*mov<VM64:mode>_insn_sp64): Likewise.
30918 * config/sparc/sync.md (*atomic_compare_and_swap<mode>_1): Replace
30920 (atomic_compare_and_swap_leon3_1): Likewise.
30921 (*atomic_compare_and_swapdi_v8plus): Likewise.
30922 * config/sparc/sparc.c (memory_ok_for_ldd): Remove useless test on
30923 architecture and add missing address validity check during LRA.
30925 2021-03-12 Tobias Burnus <tobias@codesourcery.com>
30928 * gimplify.c (omp_add_variable): Handle NULL_TREE as size
30929 occuring for assumed-size arrays in use_device_{ptr,addr}.
30931 2021-03-12 Jakub Jelinek <jakub@redhat.com>
30934 * config/i386/constraints.md (YW): New internal constraint.
30935 * config/i386/sse.md (v_Yw): Add V4TI, V2TI, V1TI and TI cases.
30936 (*<sse2_avx2>_<insn><mode>3<mask_name>,
30937 *<sse2_avx2>_uavg<mode>3<mask_name>, *abs<mode>2,
30938 *<s>mul<mode>3_highpart<mask_name>): Use <v_Yw> instead of v in
30940 (<sse2_avx2>_psadbw): Use YW instead of v in constraints.
30941 (*avx2_pmaddwd, *sse2_pmaddwd, *<code>v8hi3, *<code>v16qi3,
30942 avx2_pmaddubsw256, ssse3_pmaddubsw128): Merge last two alternatives
30943 into one, use Yw instead of former x,v.
30944 (ashr<mode>3, <insn><mode>3): Use <v_Yw> instead of x in constraints of
30945 the last alternative.
30946 (<sse2_avx2>_packsswb<mask_name>, <sse2_avx2>_packssdw<mask_name>,
30947 <sse2_avx2>_packuswb<mask_name>, <sse4_1_avx2>_packusdw<mask_name>,
30948 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, <ssse3_avx2>_palignr<mode>,
30949 <ssse3_avx2>_pshufb<mode>3<mask_name>): Merge last two alternatives
30950 into one, use <v_Yw> instead of former x,v.
30951 (avx2_interleave_highv32qi<mask_name>,
30952 vec_interleave_highv16qi<mask_name>): Use Yw instead of v in
30953 constraints. Add && <mask_avx512bw_condition> to condition.
30954 (avx2_interleave_lowv32qi<mask_name>,
30955 vec_interleave_lowv16qi<mask_name>,
30956 avx2_interleave_highv16hi<mask_name>,
30957 vec_interleave_highv8hi<mask_name>,
30958 avx2_interleave_lowv16hi<mask_name>, vec_interleave_lowv8hi<mask_name>,
30959 avx2_pshuflw_1<mask_name>, sse2_pshuflw_1<mask_name>,
30960 avx2_pshufhw_1<mask_name>, sse2_pshufhw_1<mask_name>,
30961 avx2_<code>v16qiv16hi2<mask_name>, sse4_1_<code>v8qiv8hi2<mask_name>,
30962 *sse4_1_<code>v8qiv8hi2<mask_name>_1, <sse2_avx2>_<insn><mode>3): Use
30963 Yw instead of v in constraints.
30964 * config/i386/mmx.md (Yv_Yw): New define_mode_attr.
30965 (*mmx_<insn><mode>3, mmx_ashr<mode>3, mmx_<insn><mode>3): Use <Yv_Yw>
30966 instead of Yv in constraints.
30967 (*mmx_<insn><mode>3, *mmx_mulv4hi3, *mmx_smulv4hi3_highpart,
30968 *mmx_umulv4hi3_highpart, *mmx_pmaddwd, *mmx_<code>v4hi3,
30969 *mmx_<code>v8qi3, mmx_pack<s_trunsuffix>swb, mmx_packssdw,
30970 mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
30971 *mmx_uavgv8qi3, *mmx_uavgv4hi3, mmx_psadbw): Use Yw instead of Yv in
30973 (*mmx_pinsrw, *mmx_pinsrb, *mmx_pextrw, *mmx_pextrw_zext, *mmx_pextrb,
30974 *mmx_pextrb_zext): Use YW instead of Yv in constraints.
30975 (*mmx_eq<mode>3, mmx_gt<mode>3): Use x instead of Yv in constraints.
30976 (mmx_andnot<mode>3, *mmx_<code><mode>3): Split last alternative into
30977 two, one with just x, another isa avx512vl with v.
30979 2021-03-12 Martin Liska <mliska@suse.cz>
30981 * doc/invoke.texi: Add missing param documentation.
30983 2021-03-11 David Malcolm <dmalcolm@redhat.com>
30986 * Makefile.in (ANALYZER_OBJS): Add analyzer/feasible-graph.o and
30987 analyzer/trimmed-graph.o.
30988 * doc/analyzer.texi (Analyzer Paths): Rewrite description of
30989 feasibility checking to reflect new implementation.
30990 * doc/invoke.texi (-fdump-analyzer-feasibility): Document new
30992 * shortest-paths.h (shortest_paths::get_shortest_distance): New.
30994 2021-03-11 David Malcolm <dmalcolm@redhat.com>
30996 * digraph.cc (selftest::test_shortest_paths): Update
30997 shortest_paths init for new param. Add test of
30998 SPS_TO_GIVEN_TARGET.
30999 * shortest-paths.h (enum shortest_path_sense): New.
31000 (shortest_paths::shortest_paths): Add "sense" param.
31001 Update for renamings. Generalize to use "sense" param.
31002 (shortest_paths::get_shortest_path): Rename param.
31003 (shortest_paths::m_sense): New field.
31004 (shortest_paths::m_prev): Rename...
31005 (shortest_paths::m_best_edge): ...to this.
31006 (shortest_paths::get_shortest_path): Update for renamings.
31007 Conditionalize flipping of path on sense of traversal.
31009 2021-03-11 David Malcolm <dmalcolm@redhat.com>
31011 * digraph.cc (selftest::test_shortest_paths): Add test coverage
31012 for paths from B and C.
31013 * shortest-paths.h (shortest_paths::shortest_paths): Handle
31014 unreachable nodes, rather than asserting.
31016 2021-03-11 David Edelsohn <dje.gcc@gmail.com>
31019 * config/rs6000/rs6000.c (rs6000_xcoff_file_start): Don't create
31020 xcoff_tbss_section_name.
31021 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .lcomm.
31022 * xcoffout.c (xcoff_tbss_section_name): Delete.
31023 * xcoffout.h (xcoff_tbss_section_name): Delete.
31025 2021-03-11 Richard Biener <rguenther@suse.de>
31027 PR tree-optimization/99523
31028 * tree-cfg.c (dump_function_to_file): Dump SSA names
31029 w/o identifier to the decls section as well, not only those
31030 without a VAR_DECL.
31032 2021-03-11 Jakub Jelinek <jakub@redhat.com>
31035 * ipa-icf-gimple.c (func_checker::compare_gimple_call): For internal
31036 function calls with lhs fail if the lhs don't have compatible types.
31038 2021-03-11 Hans-Peter Nilsson <hp@axis.com>
31040 * config/cris/cris.h (HARD_FRAME_POINTER_REGNUM): Define.
31041 Change FRAME_POINTER_REGNUM to correspond to a new faked
31042 register faked_fp, part of GENNONACR_REGS like faked_ap.
31043 (CRIS_FAKED_REGS_CONTENTS): New helper macro.
31044 (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS, CALL_USED_REGISTERS):
31045 (REG_ALLOC_ORDER, REG_CLASS_CONTENTS, REGNO_OK_FOR_BASE_P)
31046 (ELIMINABLE_REGS, REGISTER_NAMES): Adjust accordingly.
31047 * config/cris/cris.md (CRIS_FP_REGNUM): Renumber to new faked
31049 (CRIS_REAL_FP_REGNUM): New constant.
31050 * config/cris/cris.c (cris_reg_saved_in_regsave_area): Check
31051 for HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM.
31052 (cris_initial_elimination_offset): Handle elimination changes
31053 to HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM
31054 and add one from FRAME_POINTER_REGNUM to
31055 HARD_FRAME_POINTER_REGNUM.
31056 (cris_expand_prologue, cris_expand_epilogue): Emit code for
31057 hard_frame_pointer_rtx instead of frame_pointer_rtx.
31059 2021-03-10 David Edelsohn <dje.gcc@gmail.com>
31062 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Add check for DCmode.
31063 * config/rs6000/rs6000.c (rs6000_special_round_type_align): Same.
31065 2021-03-10 Vladimir N. Makarov <vmakarov@redhat.com>
31068 * lra-constraints.c (process_address_1): Don't check unknown
31069 constraint, use X for empty constraint.
31071 2021-03-10 Alex Coplan <alex.coplan@arm.com>
31073 * config/aarch64/aarch64.c (aarch64_vfp_is_call_or_return_candidate):
31074 Fix typo in comment describing "is_ha" argument.
31076 2021-03-10 John David Anglin <danglin@gcc.gnu.org>
31078 * doc/sourcebuild.texi: Document LRA target selector.
31080 2021-03-10 David Malcolm <dmalcolm@redhat.com>
31082 * doc/ux.texi: Add subsection contrasting interactive versus
31083 batch usage of GCC.
31085 2021-03-10 Joel Hutton <joel.hutton@arm.com>
31088 * tree-vect-stmts.c (vectorizable_store): Fix scatter store mask
31090 (vectorizable_load): Fix gather load mask check condition.
31092 2021-03-10 Richard Biener <rguenther@suse.de>
31094 PR tree-optimization/99510
31095 * tree.c (check_aligned_type): Check that the candidate
31096 has TYPE_USER_ALIGN set instead of matching with the
31099 2021-03-10 Eric Botcazou <ebotcazou@adacore.com>
31101 * config/sparc/sparc.c (sparc_regmode_natural_size): Return 4 for
31102 float and vector integer modes only if the mode is not larger.
31104 2021-03-10 Hans-Peter Nilsson <hp@axis.com>
31106 * config/cris/cris.h (DWARF_FRAME_REGISTERS): Define.
31108 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
31110 * ira.c (ira_setup_alts, ira_get_dup_out_num): Process digital
31112 * ira-lives.c (single_reg_class): Ditto.
31114 2021-03-09 Sebastian Huber <sebastian.huber@embedded-brains.de>
31116 * config.gcc (aarch64-*-rtems*): Include general rtems.h after
31117 the architecture-specific rtems.h.
31118 (aarch64-*-rtems*): Likewise.
31119 (arm*-*-rtems*): Likewise.
31120 (epiphany-*-rtems*): Likewise.
31121 (riscv*-*-rtems*): Likewise.
31123 2021-03-09 Jakub Jelinek <jakub@redhat.com>
31125 PR tree-optimization/99305
31126 * tree-ssa-phiopt.c (conditional_replacement): Test integer_pow2p
31127 before integer_all_onesp instead of vice versa.
31129 2021-03-09 Richard Earnshaw <rearnsha@arm.com>
31131 * common/config/arm/arm-common.c (arm_config_default): Change type
31132 of 'i' to unsigned.
31134 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
31137 * lra-constraints.c (process_address_1): Process constraint 'g'
31138 separately and digital constraints containing more one digit.
31140 2021-03-09 Nick Clifton <nickc@redhat.com>
31142 * config/rx/rx.h (DBX_DEBUGGING_INFO): Define.
31143 (DWARF"_DEBUGGING_INFO): Define.
31145 2021-03-09 Eric Botcazou <ebotcazou@adacore.com>
31148 * calls.c (initialize_argument_information): When the argument
31149 is passed by reference, do not make a copy in a thunk only if
31150 the argument is already in memory. Remove redundant test for
31151 the case of callee copy.
31153 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
31156 * lra-constraints.c (process_address_1): Process 0..9 constraints
31157 in process_address_1.
31159 2021-03-09 Andreas Krebbel <krebbel@linux.ibm.com>
31161 * config/s390/s390.c (struct s390_processor processor_table):
31162 Binutils name string must not be empty.
31164 2021-03-09 Claudiu Zissulescu <claziss@synopsys.com>
31166 * config/arc/arc.c (arc_attr_type): Remove function.
31168 2021-03-09 Martin Liska <mliska@suse.cz>
31171 * config/i386/i386-options.c (ix86_option_override_internal):
31172 Set isa_flags for OPTS argument and not for the global
31175 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
31177 * config/rs6000/predicates.md (ds_form_mem_operand): Check
31180 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
31183 * config/rs6000/predicates.md (ds_form_mem_operand) New
31185 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10) Use
31186 ds_form_mem_operand in ld/lwa patterns.
31187 * config/rs6000/fusion.md: Regenerate file.
31189 2021-03-08 Martin Sebor <msebor@redhat.com>
31191 PR middle-end/98266
31192 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): New function.
31193 (array_bounds_checker::check_array_bounds): Call it.
31195 2021-03-08 Martin Sebor <msebor@redhat.com>
31197 PR middle-end/97631
31198 * tree-ssa-strlen.c (maybe_warn_overflow): Test rawmem.
31199 (handle_builtin_stxncpy_strncat): Rename locals. Determine
31200 destination size from allocation calls. Issue a more appropriate
31202 (handle_builtin_memcpy): Pass true as rawmem to maybe_warn_overflow.
31203 (handle_builtin_memset): Same.
31205 2021-03-08 Peter Bergner <bergner@linux.ibm.com>
31208 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_permute): Add an assert
31209 to ensure we do not have an Altivec style address.
31210 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): Disable if passed
31211 an Altivec style address.
31212 (*vsx_le_perm_store_<mode>): Likewise.
31213 (splitters after *vsx_le_perm_store_<mode>): Likewise.
31214 (vsx_load_<mode>): Disable special expander if passed an Altivec
31216 (vsx_store_<mode>): Likewise.
31218 2021-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31221 * config/aarch64/predicates.md (aarch64_simd_shift_imm_vec_qi): Define.
31222 (aarch64_simd_shift_imm_vec_hi): Likewise.
31223 (aarch64_simd_shift_imm_vec_si): Likewise.
31224 (aarch64_simd_shift_imm_vec_di): Likewise.
31225 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le): Use
31226 predicate from above.
31227 (aarch64_shrn<mode>_insn_be): Likewise.
31228 (aarch64_rshrn<mode>_insn_le): Likewise.
31229 (aarch64_rshrn<mode>_insn_be): Likewise.
31230 (aarch64_shrn2<mode>_insn_le): Likewise.
31231 (aarch64_shrn2<mode>_insn_be): Likewise.
31232 (aarch64_rshrn2<mode>_insn_le): Likewise.
31233 (aarch64_rshrn2<mode>_insn_be): Likewise.
31235 2021-03-08 Vladimir N. Makarov <vmakarov@redhat.com>
31238 * lra-constraints.c (skip_contraint_modifiers): New function.
31239 (process_address_1): Use it before lookup_constraint call.
31241 2021-03-08 Martin Liska <mliska@suse.cz>
31244 * config/i386/i386-options.c (ix86_option_override_internal):
31245 Enable UINTR and HRESET for -march that supports it.
31247 2021-03-08 Ilya Leoshkevich <iii@linux.ibm.com>
31249 * config/s390/s390.c (f_constraint_p): New function.
31250 (s390_md_asm_adjust): Implement TARGET_MD_ASM_ADJUST.
31251 (TARGET_MD_ASM_ADJUST): Likewise.
31253 2021-03-08 Tobias Burnus <tobias@codesourcery.com>
31256 * tree-nested.c (convert_local_reference_stmt): Avoid calling
31257 lookup_field_for_decl for Fortran module (= namespace context).
31259 2021-03-08 Andreas Krebbel <krebbel@linux.ibm.com>
31261 * config/s390/s390.c (s390_expand_vec_compare): Implement <0
31262 comparison with arithmetic right shift.
31263 (s390_expand_vcond): No need for a force_reg anymore.
31264 s390_vec_compare will do it.
31265 * config/s390/vector.md ("vec_cmp<mode><tointvec>"): Accept also
31266 immediate operands.
31268 2021-03-07 Jakub Jelinek <jakub@redhat.com>
31271 * config/i386/constraints.md (Yw): Use SSE_REGS if TARGET_SSE
31272 but TARGET_AVX512BW or TARGET_AVX512VL is not set. Adjust description
31274 * config/i386/sse.md (v_Yw): New define_mode_attr.
31275 (*<insn><mode>3, *mul<mode>3<mask_name>, *avx2_<code><mode>3,
31276 *sse4_1_<code><mode>3<mask_name>): Use <v_Yw> instead of v
31278 * config/i386/mmx.md (mmx_pshufw_1, *vec_dupv4hi): Use Yw instead of
31279 xYw in constraints.
31281 2021-03-06 Julian Brown <julian@codesourcery.com>
31283 * tree-pretty-print.c (dump_generic_node): Emit non-generic
31284 address space info for aggregates.
31286 2021-03-06 Hans-Peter Nilsson <hp@axis.com>
31288 * config/cris/cris.h (MAX_FIXED_MODE_SIZE): Don't define.
31290 2021-03-05 Jakub Jelinek <jakub@redhat.com>
31292 PR middle-end/99322
31293 * tree-cfg.c (bb_to_omp_idx): New variable.
31294 (execute_build_cfg): Release the bb_to_omp_idx vector after
31295 cleanup_tree_cfg returns.
31296 (handle_abnormal_edges): Remove bb_to_omp_idx argument, adjust
31297 for bb_to_omp_idx being a vec<int> instead of pointer to array
31299 (make_edges): Remove bb_to_omp_idx local variable, don't pass
31300 it to handle_abnormal_edges, adjust for bb_to_omp_idx being a
31301 vec<int> instead of pointer to array of ints and don't free/release
31303 (remove_bb): When removing a bb and placing forced label somewhere
31304 else, ensure it is put into the same OpenMP region during cfg
31305 pass if possible or to entry successor as fallback. Unregister
31306 bb from bb_to_omp_idx.
31308 2021-03-05 Vladimir N. Makarov <vmakarov@redhat.com>
31311 * lra-constraints.c (process_address_1): Skip decomposing address
31312 for asm insn operand with unknown constraint.
31314 2021-03-05 Martin Jambor <mjambor@suse.cz>
31317 * cgraph.c (cgraph_edge::set_call_stmt): Do not update all
31318 corresponding speculative edges if we are about to resolve
31319 sepculation. Make edge direct (and so resolve speculations) before
31320 removing it from call_site_hash.
31321 (cgraph_edge::make_direct): Relax the initial assert to allow calling
31322 the function on speculative direct edges.
31324 2021-03-05 Eric Botcazou <ebotcazou@adacore.com>
31326 PR rtl-optimization/99376
31327 * rtlanal.c (nonzero_bits1) <arithmetic operators>: If the number
31328 of low-order zero bits is too large, set the result to 0 directly.
31330 2021-03-04 Jakub Jelinek <jakub@redhat.com>
31332 PR middle-end/93235
31333 * expmed.c (store_bit_field_using_insv): Return false of xop0 is a
31334 SUBREG and a SUBREG to op_mode can't be created.
31336 2021-03-04 Alex Coplan <alex.coplan@arm.com>
31339 * config/aarch64/aarch64-sve-builtins.cc
31340 (function_resolver::require_vector_type): Handle error_mark_node.
31342 2021-03-04 Ilya Leoshkevich <iii@linux.ibm.com>
31344 * cfgexpand.c (expand_asm_loc): Pass new parameter.
31345 (expand_asm_stmt): Likewise.
31346 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Add new
31348 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
31349 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
31350 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
31351 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
31352 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
31353 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
31354 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
31355 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
31356 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
31357 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
31358 * doc/tm.texi (md_asm_adjust): Likewise.
31359 * target.def (md_asm_adjust): Likewise.
31361 2021-03-04 Richard Biener <rguenther@suse.de>
31363 PR middle-end/97855
31364 * tree-pretty-print.c: Poison pp_printf.
31365 (dump_decl_name): Avoid use of pp_printf.
31366 (dump_block_node): Likewise.
31367 (dump_generic_node): Likewise.
31369 2021-03-04 Martin Sebor <msebor@redhat.com>
31371 PR middle-end/96963
31372 PR middle-end/94655
31373 * builtins.c (handle_array_ref): New helper.
31374 (handle_mem_ref): New helper.
31375 (compute_objsize_r): Factor out ARRAY_REF and MEM_REF handling
31376 into new helper functions. Correct a workaround for vectorized
31379 2021-03-03 Pat Haugen <pthaugen@linux.ibm.com>
31381 * config/rs6000/dfp.md (extendddtd2, trunctddd2, *cmp<mode>_internal1,
31382 floatditd2, ftrunc<mode>2, fix<mode>di2, dfp_ddedpd_<mode>,
31383 dfp_denbcd_<mode>, dfp_dxex_<mode>, dfp_diex_<mode>,
31384 *dfp_sgnfcnc_<mode>, dfp_dscli_<mode>, dfp_dscri_<mode>): Update size
31385 attribute for Power10.
31386 * config/rs6000/mma.md (*movoo): Likewise.
31387 * config/rs6000/rs6000.md (define_attr "size"): Add 256.
31388 (define_mode_attr bits): Add DD/TD modes.
31389 * config/rs6000/sync.md (load_quadpti, store_quadpti, load_lockedpti,
31390 store_conditionalpti): Update size attribute for Power10.
31392 2021-03-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
31395 * config/sparc/t-sparc (tree-ssanames.o-warn): Don't error for
31396 -Wuninitialized, -Wmaybe-uninitialized.
31397 (wide-int.o-warn): Likewise.
31399 2021-03-03 Richard Earnshaw <rearnsha@arm.com>
31401 * common/config/arm/arm-common.c: Include configargs.h.
31402 (arm_config_default): New function.
31403 (arm_target_mode): Renamed from arm_target_thumb_only. Handle
31404 processors that do not support Thumb. Take into account the
31405 --with-mode configuration setting for selecting the default.
31406 * config/arm/arm.h (OPTION_DEFAULT_SPECS): Remove entry for 'mode'.
31407 (TARGET_MODE_SPEC_FUNCTIONS): Update for function name change.
31409 2021-03-03 Martin Liska <mliska@suse.cz>
31411 PR gcov-profile/97461
31412 * gcov-io.h (GCOV_PREALLOCATED_KVP): Remove.
31414 2021-03-03 Eric Botcazou <ebotcazou@adacore.com>
31417 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
31418 point back the hard frame pointer to its default location when the
31419 frame is larger than SEH_MAX_FRAME_SIZE.
31421 2021-03-03 Jakub Jelinek <jakub@redhat.com>
31424 * config/i386/predicates.md (logic_operator): New define_predicate.
31425 * config/i386/i386.md (mov + mem using comm arith peephole2):
31426 Punt if operands[1] is EXT_REX_SSE_REGNO_P, AVX512BW is not enabled
31427 and the inner mode is [QH]Imode.
31429 2021-03-03 Jakub Jelinek <jakub@redhat.com>
31432 * dwarf2out.c (dw_loc_list_struct): Add end_entry member.
31433 (new_loc_list): Clear end_entry.
31434 (output_loc_list): Only use DW_LLE_startx_length for -gsplit-dwarf
31435 if HAVE_AS_LEB128, otherwise use DW_LLE_startx_endx. Fix comment
31437 (index_location_lists): For dwarf_version >= 5 without HAVE_AS_LEB128,
31438 initialize also end_entry.
31440 2021-03-03 Jakub Jelinek <jakub@redhat.com>
31443 * cfgrtl.c (fixup_partitions): When changing some bbs from hot to cold
31444 partitions, if in non-layout mode after reorder_blocks also move
31445 affected blocks to ensure a single partition transition.
31447 2021-03-03 Jason Merrill <jason@redhat.com>
31450 * cgraphunit.c (process_function_and_variable_attributes): Don't
31451 warn about flatten on an alias if the target also has it.
31452 * cgraph.h (symtab_node::get_alias_target_tree): New.
31454 2021-03-02 David Edelsohn <dje.gcc@gmail.com>
31456 * config/rs6000/rs6000.md (tls_get_tpointer_internal): Prepend
31457 period to symbol name.
31458 (tls_get_addr_internal<mode>): Same.
31460 2021-03-02 David Malcolm <dmalcolm@redhat.com>
31463 * diagnostic-show-locus.c
31464 (selftest::test_one_liner_many_fixits_2): Fix accidental usage of
31467 2021-03-02 Martin Sebor <msebor@redhat.com>
31469 PR middle-end/99276
31470 * builtins.c (warn_for_access): Remove stray warning text.
31472 2021-03-02 Martin Sebor <msebor@redhat.com>
31474 PR middle-end/99295
31475 * doc/extend.texi (attribute malloc): Reword and clarify nonaliasing
31478 2021-03-02 Jakub Jelinek <jakub@redhat.com>
31481 * dwarf2out.c (output_macinfo_op): Use DW_MACRO_*_str* even with
31482 -gdwarf-5 -gstrict-dwarf. For -gsplit-dwarf -gdwarf-5 use
31483 DW_MACRO_*_strx instead of DW_MACRO_*_strp. Handle
31484 DW_MACRO_define_strx and DW_MACRO_undef_strx.
31485 (save_macinfo_strings): Use DW_MACRO_*_str* even with
31486 -gdwarf-5 -gstrict-dwarf. Handle DW_MACRO_define_strx and
31487 DW_MACRO_undef_strx.
31489 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
31491 * config/s390/s390-builtin-types.def (BT_FN_V4SF_V8HI_UINT): New
31493 (BT_FN_V8HI_V8HI_UINT): Likewise.
31494 (BT_FN_V8HI_V4SF_V4SF_UINT): Likewise.
31495 * config/s390/s390-builtins.def (B_NNPA): New macro definition.
31496 (s390_vclfnhs, s390_vclfnls, s390_vcrnfs, s390_vcfn, s390_vcnf):
31497 New builtin definitions.
31498 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Bump
31499 vector extension version.
31500 * config/s390/s390.c (s390_expand_builtin): Check if builtins are
31501 available with current -march level.
31502 * config/s390/s390.md (UNSPEC_NNPA_VCLFNHS_V8HI)
31503 (UNSPEC_NNPA_VCLFNLS_V8HI, UNSPEC_NNPA_VCRNFS_V8HI)
31504 (UNSPEC_NNPA_VCFN_V8HI, UNSPEC_NNPA_VCNF_V8HI): New constants.
31505 * config/s390/vecintrin.h (vec_extend_to_fp32_hi): New macro.
31506 (vec_extend_to_fp32_lo): Likewise.
31507 (vec_round_from_fp32): Likewise.
31508 (vec_convert_to_fp16): Likewise.
31509 (vec_convert_from_fp16): Likewise.
31510 * config/s390/vx-builtins.md (vclfnhs_v8hi): New insn pattern.
31511 (vclfnls_v8hi): Likewise.
31512 (vcrnfs_v8hi): Likewise.
31513 (vcfn_v8hi): Likewise.
31514 (vcnf_v8hi): Likewise.
31516 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
31518 * common/config/s390/s390-common.c (processor_flags_table): New entry.
31519 * config.gcc: Enable arch14 for --with-arch and --with-tune.
31520 * config/s390/driver-native.c (s390_host_detect_local_cpu): Pick
31521 arch14 for unknown CPU models.
31522 * config/s390/s390-opts.h (enum processor_type): Add PROCESSOR_ARCH14.
31523 * config/s390/s390.c (s390_issue_rate): Add case for PROCESSOR_ARCH14.
31524 (s390_get_sched_attrmask): Likewise.
31525 (s390_get_unit_mask): Likewise.
31526 * config/s390/s390.h (enum processor_flags): Add PF_NNPA and PF_ARCH14.
31527 (TARGET_CPU_ARCH14, TARGET_CPU_ARCH14_P, TARGET_CPU_NNPA)
31528 (TARGET_CPU_NNPA_P, TARGET_ARCH14, TARGET_ARCH14_P, TARGET_NNPA)
31529 (TARGET_NNPA_P): New macro definitions.
31530 * config/s390/s390.md ("cpu_facility", "enabled"): Add arch14 and nnpa.
31531 * config/s390/s390.opt: Add PROCESSOR_ARCH14.
31533 2021-03-02 Jakub Jelinek <jakub@redhat.com>
31535 PR middle-end/95757
31536 * tree-vrp.c (register_edge_assert_for): Remove superfluous ()s around
31537 condition. Call register_edge_assert_for_1 for == 0, != 0, == 1 and
31538 != 1 comparisons if name is lhs of a comparison.
31540 2021-03-01 Iain Sandoe <iain@sandoe.co.uk>
31544 * config/darwin-protos.h (darwin_should_restore_cfa_state): New.
31545 * config/darwin.c (darwin_should_restore_cfa_state): New.
31546 * config/darwin.h (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New.
31547 * doc/tm.texi: Regenerated.
31548 * doc/tm.texi.in: Document TARGET_ASM_SHOULD_RESTORE_CFA_STATE.
31549 * dwarf2cfi.c (connect_traces): If the target requests, restore
31550 the CFA expression after a DW_CFA_restore.
31551 * target.def (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New hook.
31553 2021-03-01 Martin Liska <mliska@suse.cz>
31556 * optc-save-gen.awk: Add 4 more exceptions.
31558 2021-03-01 Nathan Sidwell <nathan@acm.org>
31561 * tree.h (TYPE_ALIGN_RAW): New accessor.
31562 (TYPE_ALIGN): Use it.
31564 2021-03-01 Jan Hubicka <jh@suse.cz>
31567 * ipa-fnsummary.c (compute_fn_summary): Fix sanity check.
31569 2021-03-01 Eric Botcazou <ebotcazou@adacore.com>
31572 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
31573 point the hard frame pointer to the SSE register save area instead
31574 of the general register save area. Perform only minimal adjustment
31575 for small frames if it is initially not correctly aligned.
31576 (ix86_expand_prologue): Remove early saves for a SEH target.
31577 * config/i386/winnt.c (struct seh_frame_state): Document constraint.
31579 2021-02-28 Jakub Jelinek <jakub@redhat.com>
31582 * ipa.c (symbol_table::remove_unreachable_nodes): Fix a comment
31583 typo - referneced -> referenced.
31584 * tree.c (component_ref_size): Fix comment typo -
31585 refernce -> reference.
31586 * tree-ssa-alias.c (access_path_may_continue_p): Fix comment typo -
31587 traling -> trailing.
31588 (aliasing_component_refs_p): Fix comment typos -
31589 refernce -> reference and refernece -> reference and
31590 traling -> trailing.
31591 (nonoverlapping_refs_since_match_p): Fix comment typo -
31592 referneces -> references.
31593 * doc/invoke.texi (--param modref-max-bases): Fix a typo -
31594 referneces -> references.
31596 2021-02-27 Iain Sandoe <iain@sandoe.co.uk>
31598 * config/host-darwin.c (darwin_gt_pch_use_address): Modify
31599 diagnostic message to avoid use of a contraction and format
31602 2021-02-27 Jakub Jelinek <jakub@redhat.com>
31605 * gcse.c (gcse_or_cprop_is_too_expensive): Use %wu instead of
31606 HOST_WIDE_INT_PRINT_UNSIGNED in warning format string.
31607 * ipa-devirt.c (ipa_odr_read_section): Use %wd instead of
31608 HOST_WIDE_INT_PRINT_DEC in inform format string. Fix comment
31611 2021-02-26 Richard Biener <rguenther@suse.de>
31613 PR middle-end/99281
31614 * expr.c (store_field): For calls with return-slot optimization
31615 and addressable return type expand the store directly.
31617 2021-02-26 Richard Biener <rguenther@suse.de>
31620 * builtins.c (warn_string_no_nul): Fix diagnostic formatting.
31622 2021-02-26 Peter Bergner <bergner@linux.ibm.com>
31625 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Replace assert
31628 2021-02-26 Aaron Sawdey <acsawdey@linux.ibm.com>
31630 * config.gcc: Add rs6000-pcrel-opt.o.
31631 * config/rs6000/rs6000-pcrel-opt.c: New file.
31632 * config/rs6000/pcrel-opt.md: New file.
31633 * config/rs6000/predicates.md: Add d_form_memory predicate.
31634 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_PCREL_OPT.
31635 * config/rs6000/rs6000-passes.def: Add pass_pcrel_opt.
31636 * config/rs6000/rs6000-protos.h: Add reg_to_non_prefixed(),
31637 pcrel_opt_valid_mem_p(), output_pcrel_opt_reloc(),
31638 and make_pass_pcrel_opt().
31639 * config/rs6000/rs6000.c (reg_to_non_prefixed): Make global.
31640 (rs6000_option_override_internal): Add pcrel-opt.
31641 (rs6000_delegitimize_address): Support pcrel-opt.
31642 (rs6000_opt_masks): Add pcrel-opt.
31643 (pcrel_opt_valid_mem_p): New function.
31644 (reg_to_non_prefixed): Make global.
31645 (rs6000_asm_output_opcode): Reset prepend_p_to_next_insn.
31646 (output_pcrel_opt_reloc): New function.
31647 * config/rs6000/rs6000.md (loads_extern_addr): New attr.
31648 (pcrel_extern_addr): Set loads_extern_addr.
31649 Add include for pcrel-opt.md.
31650 * config/rs6000/rs6000.opt: Add -mpcrel-opt.
31651 * config/rs6000/t-rs6000: Add rules for pcrel-opt.c and
31654 2021-02-26 YunQiang Su <yunqiang.su@cipunited.com>
31657 * config/mips/mips.c (mips_expand_ext_as_unaligned_load):
31658 If TARGET_64BIT and dest is SUBREG, we check the width, if it
31659 equal to SImode, we use SImode operation, just like what we are
31662 2021-02-26 Marek Polacek <polacek@redhat.com>
31664 * builtins.c (warn_for_access): Fix typos.
31666 2021-02-25 Iain Sandoe <iain@sandoe.co.uk>
31668 * config/aarch64/aarch64.md (<optab>_rol<mode>3): Add a '#'
31669 mark in front of the immediate quantity.
31670 (<optab>_rolsi3_uxtw): Likewise.
31672 2021-02-25 Richard Earnshaw <rearnsha@arm.com>
31675 * config/arm/thumb2.md (nonsecure_call_reg_thumb2_fpcxt): New pattern.
31676 (nonsecure_call_value_reg_thumb2_fpcxt): Likewise.
31677 (nonsecure_call_reg_thumb2): Restrict to using r4 for the callee
31678 address and disable when the FPCXT is not available.
31679 (nonsecure_call_value_reg_thumb2): Likewise.
31681 2021-02-25 Nathan Sidwell <nathan@acm.org>
31684 * doc/invoke.texi (flang-info-module-cmi): Renamed option.
31686 2021-02-25 Tamar Christina <tamar.christina@arm.com>
31688 * tree-vect-slp.c (optimize_load_redistribution_1): Abort on NULL nodes.
31690 2021-02-25 Richard Biener <rguenther@suse.de>
31692 PR tree-optimization/99253
31693 * tree-vect-loop.c (check_reduction_path): First compute
31694 code, then verify out-of-loop uses.
31696 2021-02-25 Jakub Jelinek <jakub@redhat.com>
31699 * match.pd ((T)(A) + CST -> (T)(A + CST)): Add :s to convert.
31701 2021-02-25 Jakub Jelinek <jakub@redhat.com>
31703 PR tree-optimization/80635
31704 * tree-vrp.c (vrp_simplify_cond_using_ranges): Also handle
31705 VIEW_CONVERT_EXPR if modes are the same, innerop is integral and
31706 has mode precision.
31708 2021-02-25 Richard Biener <rguenther@suse.de>
31710 * tree-vect-slp.c (optimize_load_redistribution_1): Delay
31711 load_map population.
31712 (vect_match_slp_patterns_2): Revert part of last change.
31713 (vect_analyze_slp): Do not interleave optimize_load_redistribution
31714 with pattern detection but do it afterwards. Dump the
31715 whole SLP graph after pattern recognition and load
31716 redistribution optimization finished.
31718 2021-02-24 Jakub Jelinek <jakub@redhat.com>
31721 * omp-low.c (struct omp_context): Add teams_nested_p and
31722 nonteams_nested_p members.
31723 (scan_omp_target): Diagnose teams nested inside of target with other
31724 directives strictly nested inside of the same target.
31725 (check_omp_nesting_restrictions): Set ctx->teams_nested_p or
31726 ctx->nonteams_nested_p as needed.
31728 2021-02-24 Vladimir N. Makarov <vmakarov@redhat.com>
31730 PR inline-asm/99123
31731 * lra-constraints.c (uses_hard_regs_p): Don't use decompose_mem_address.
31733 2021-02-24 Hans-Peter Nilsson <hp@axis.com>
31735 * config/cris/cris.c (cris_expand_prologue): Set
31736 current_function_static_stack_size, if flag_stack_usage_info.
31738 2021-02-24 Pat Haugen <pthaugen@linux.ibm.com>
31740 * config/rs6000/rs6000.c (next_insn_prefixed_p): Rename.
31741 (rs6000_final_prescan_insn): Adjust.
31742 (rs6000_asm_output_opcode): Likewise.
31744 2021-02-24 Martin Sebor <msebor@redhat.com>
31746 PR middle-end/97172
31747 * attribs.c (attr_access::free_lang_data): Clear attribute arg spec
31748 from function arguments.
31750 2021-02-24 Tamar Christina <tamar.christina@arm.com>
31752 PR tree-optimization/99220
31753 * tree-vect-slp.c (optimize_load_redistribution_1): Remove
31754 node from cache when it's about to be deleted.
31756 2021-02-24 Jakub Jelinek <jakub@redhat.com>
31758 PR tree-optimization/99225
31759 * fold-const.c (fold_binary_loc) <case NE_EXPR>: In (x & (1 << y)) != 0
31760 to ((x >> y) & 1) != 0 simplifications use build_one_cst instead of
31761 build_int_cst (..., 1). Formatting fixes.
31763 2021-02-24 Tamar Christina <tamar.christina@arm.com>
31765 PR tree-optimization/99149
31766 * tree-vect-slp-patterns.c (vect_detect_pair_op): Don't recreate the
31768 (vect_slp_reset_pattern): Remove.
31769 (complex_fma_pattern::matches): Remove call to vect_slp_reset_pattern.
31770 (complex_mul_pattern::build, complex_fma_pattern::build,
31771 complex_fms_pattern::build): Fix ref counts.
31772 * tree-vect-slp.c (vect_free_slp_tree): Undo SLP only pattern relevancy
31773 when node is being deleted.
31774 (vect_match_slp_patterns_2): Correct result of cache hit on patterns.
31775 (vect_schedule_slp): Invalidate SLP_TREE_REPRESENTATIVE of removed
31777 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize value.
31779 2021-02-24 Matthias Klose <doko@ubuntu.com>
31782 2020-12-07 Matthias Klose <doko@ubuntu.com>
31784 * genextract.c (print_header): Undefine ENABLE_RTL_CHECKING
31785 and ENABLE_RTL_FLAG_CHECKING.
31787 2021-02-24 Richard Biener <rguenther@suse.de>
31790 * builtins.c (fold_builtin_next_arg): Avoid NULL arg.
31792 2021-02-23 Peter Bergner <bergner@linux.ibm.com>
31794 * config/rs6000/mma.md (mma_assemble_pair): Rename from this...
31795 (vsx_assemble_pair): ...to this.
31796 (*mma_assemble_pair): Rename from this...
31797 (*vsx_assemble_pair): ...to this.
31798 (mma_disassemble_pair): Rename from this...
31799 (vsx_disassemble_pair): ...to this.
31800 (*mma_disassemble_pair): Rename from this...
31801 (*vsx_disassemble_pair): ...to this.
31802 * config/rs6000/rs6000-builtin.def (BU_MMA_V2, BU_MMA_V3,
31803 BU_COMPAT): New macros.
31804 (mma_assemble_pair): Rename from this...
31805 (vsx_assemble_pair): ...to this.
31806 (mma_disassemble_pair): Rename from this...
31807 (vsx_disassemble_pair): ...to this.
31808 (mma_assemble_pair): New compatibility built-in.
31809 (mma_disassemble_pair): Likewise.
31810 * config/rs6000/rs6000-call.c (struct builtin_compatibility): New.
31811 (RS6000_BUILTIN_COMPAT): Define.
31812 (bdesc_compat): New.
31813 (mma_expand_builtin): Use VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
31814 (rs6000_gimple_fold_mma_builtin): Use MMA_BUILTIN_DISASSEMBLE_PAIR
31815 and VSX_BUILTIN_ASSEMBLE_PAIR.
31816 (rs6000_init_builtins): Register compatibility built-ins.
31817 (mma_init_builtins): Use VSX_BUILTIN_ASSEMBLE_PAIR,
31818 VSX_BUILTIN_ASSEMBLE_PAIR_INTERNAL, VSX_BUILTIN_DISASSEMBLE_PAIR and
31819 VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
31820 * doc/extend.texi (__builtin_mma_assemble_pair): Rename from this...
31821 (__builtin_vsx_assemble_pair): ...to this.
31822 (__builtin_mma_disassemble_pair): Rename from this...
31823 (__builtin_vsx_disassemble_pair): ...to this.
31825 2021-02-23 Martin Liska <mliska@suse.cz>
31828 * ipa-icf.c (sem_variable::merge): Do not merge 2 variables
31829 with different alignment. That leads to an invalid red zone
31830 size allocated in runtime.
31832 2021-02-23 Jakub Jelinek <jakub@redhat.com>
31834 PR tree-optimization/99204
31835 * fold-const.c (fold_read_from_constant_string): Check that
31836 tree_fits_uhwi_p (index) rather than just that index is INTEGER_CST.
31838 2021-02-23 Segher Boessenkool <segher@kernel.crashing.org>
31839 Kewen Lin <linkw@gcc.gnu.org>
31841 * config/rs6000/rs6000.md (*rotl<mode>3_insert_3): Renamed to...
31842 (rotl<mode>3_insert_3): ...this.
31843 (plus_ior_xor): New code_iterator.
31844 (define_split for GPR rl*imi): New splitter.
31845 * config/rs6000/vsx.md (vsx_init_v4si): Use gen_rotldi3_insert_3
31846 for integer merging.
31848 2021-02-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31850 * config/aarch64/aarch64-tuning-flags.def (cse_sve_vl_constants):
31852 * config/aarch64/aarch64.md (add<mode>3): Force CONST_POLY_INT immediates
31853 into a register when the above is enabled.
31854 * config/aarch64/aarch64.c (neoversev1_tunings):
31855 AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
31856 (aarch64_rtx_costs): Use AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
31858 2021-02-22 Hans-Peter Nilsson <hp@axis.com>
31860 * config/cris/cris.c (cris_print_operand) <'T'>: Change
31861 valid operand from is now an addi mult-value to shift-value.
31862 * config/cris/cris.md (*addi): Change expression of scaled
31863 operand from mult to ashift.
31864 * config/cris/cris.md (*addi_reload): New insn_and_split.
31866 2021-02-22 John David Anglin <danglin@gcc.gnu.org>
31869 * config/pa/pa.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Define as
31870 hook_bool_const_tree_hwi_hwi_const_tree_true.
31871 (pa_asm_output_mi_thunk): Add support for nonzero vcall_offset.
31873 2021-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
31875 PR rtl-optimization/98791
31876 * ira-conflicts.c (process_regs_for_copy): Don't create allocno copies
31877 for unordered modes.
31879 2021-02-22 Martin Liska <mliska@suse.cz>
31881 * tree-inline.c (inline_forbidden_p): Set
31882 inline_forbidden_reason.
31884 2021-02-22 Richard Biener <rguenther@suse.de>
31886 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Dump
31889 2021-02-22 Richard Biener <rguenther@suse.de>
31891 PR tree-optimization/99165
31892 * gimple-ssa-store-merging.c (pass_store_merging::process_store):
31893 Accumulate changed to ret.
31895 2021-02-21 Uros Bizjak <ubizjak@gmail.com>
31898 2020-12-09 Uroš Bizjak <ubizjak@gmail.com>
31900 * config/i386/i386.h (REG_ALLOC_ORDER): Remove
31902 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
31905 * config/s390/vector.md (trunctf<DFP_ALL:mode>2_vr): New
31907 (trunctf<DFP_ALL:mode>2): Likewise.
31908 (trunctdtf2_vr): Likewise.
31909 (trunctdtf2): Likewise.
31910 (extend<DFP_ALL:mode>tf2_vr): Likewise.
31911 (extend<DFP_ALL:mode>tf2): Likewise.
31912 (extendtftd2_vr): Likewise.
31913 (extendtftd2): Likewise.
31915 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
31917 * config/s390/vector.md (*fprx2_to_tf): Rename to fprx2_to_tf,
31918 add memory alternative.
31919 (tf_to_fprx2): New pattern.
31921 2021-02-19 Martin Sebor <msebor@redhat.com>
31924 * attribs.c (init_attr_rdwr_indices): Guard vblist use.
31925 (attr_access::free_lang_data): Remove a spurious test.
31927 2021-02-19 Nathan Sidwell <nathan@acm.org>
31929 * doc/invoke.texi (flang-info-module-read): Document.
31931 2021-02-19 Martin Liska <mliska@suse.cz>
31933 PR translation/99167
31934 * params.opt: Fix typo.
31936 2021-02-19 Richard Biener <rguenther@suse.de>
31938 PR middle-end/99122
31939 * tree-inline.c (inline_forbidden_p): Do not inline functions
31940 with VLA arguments or return value.
31942 2021-02-19 Jakub Jelinek <jakub@redhat.com>
31945 * config/arm/arm.md (*stack_protect_combined_set_insn,
31946 *stack_protect_combined_test_insn): If force_const_mem result
31947 is not valid general operand, force its address into the destination
31950 2021-02-19 Jakub Jelinek <jakub@redhat.com>
31953 * tree-cfg.c (gimple_merge_blocks): If bb a starts with eh landing
31954 pad or non-local label, put FORCED_LABELs from bb b after that label
31955 rather than before it.
31957 2021-02-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
31960 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3): Use
31961 expand_vector_broadcast' to emit the vec_duplicate operand.
31963 2021-02-18 Vladimir N. Makarov <vmakarov@redhat.com>
31965 PR rtl-optimization/96264
31966 * lra-remat.c (reg_overlap_for_remat_p): Check also output insn
31969 2021-02-18 H.J. Lu <hjl.tools@gmail.com>
31972 * varasm.c (get_section): Replace SUPPORTS_SHF_GNU_RETAIN with
31973 looking up the retain attribute.
31974 (resolve_unique_section): Likewise.
31975 (get_variable_section): Likewise.
31976 (switch_to_section): Likewise. Warn when a symbol without the
31977 retain attribute and a symbol with the retain attribute are
31978 placed in the section with the same name, instead of the used
31980 * doc/extend.texi: Document the "retain" attribute.
31982 2021-02-18 Nathan Sidwell <nathan@acm.org>
31985 * doc/invoke.texi (flang-info-include-translate): Document header
31988 2021-02-18 Richard Biener <rguenther@suse.de>
31990 PR middle-end/99122
31991 * ipa-fnsummary.c (analyze_function_body): Set
31992 CIF_FUNCTION_NOT_INLINABLE for VLA parameter calls.
31993 * tree-inline.c (insert_init_debug_bind): Pass NULL for
31994 error_mark_node values.
31995 (force_value_to_type): Do not build V_C_Es for WITH_SIZE_EXPR
31997 (setup_one_parameter): Delay force_value_to_type until when
32000 2021-02-18 Hans-Peter Nilsson <hp@axis.com>
32002 PR tree-optimization/99142
32003 * match.pd (clz cmp 0): Gate replacement on single_use of clz result.
32005 2021-02-18 Jakub Jelinek <jakub@redhat.com>
32007 * wide-int-bitmask.h (wide_int_bitmask::wide_int_bitmask (),
32008 wide_int_bitmask::wide_int_bitmask (uint64_t),
32009 wide_int_bitmask::wide_int_bitmask (uint64_t, uint64_t),
32010 wide_int_bitmask::operator ~ () const,
32011 wide_int_bitmask::operator | (wide_int_bitmask) const,
32012 wide_int_bitmask::operator & (wide_int_bitmask) const): Use constexpr
32014 * config/i386/i386.h (PTA_3DNOW, PTA_3DNOW_A, PTA_64BIT, PTA_ABM,
32015 PTA_AES, PTA_AVX, PTA_BMI, PTA_CX16, PTA_F16C, PTA_FMA, PTA_FMA4,
32016 PTA_FSGSBASE, PTA_LWP, PTA_LZCNT, PTA_MMX, PTA_MOVBE, PTA_NO_SAHF,
32017 PTA_PCLMUL, PTA_POPCNT, PTA_PREFETCH_SSE, PTA_RDRND, PTA_SSE, PTA_SSE2,
32018 PTA_SSE3, PTA_SSE4_1, PTA_SSE4_2, PTA_SSE4A, PTA_SSSE3, PTA_TBM,
32019 PTA_XOP, PTA_AVX2, PTA_BMI2, PTA_RTM, PTA_HLE, PTA_PRFCHW, PTA_RDSEED,
32020 PTA_ADX, PTA_FXSR, PTA_XSAVE, PTA_XSAVEOPT, PTA_AVX512F, PTA_AVX512ER,
32021 PTA_AVX512PF, PTA_AVX512CD, PTA_NO_TUNE, PTA_SHA, PTA_PREFETCHWT1,
32022 PTA_CLFLUSHOPT, PTA_XSAVEC, PTA_XSAVES, PTA_AVX512DQ, PTA_AVX512BW,
32023 PTA_AVX512VL, PTA_AVX512IFMA, PTA_AVX512VBMI, PTA_CLWB, PTA_MWAITX,
32024 PTA_CLZERO, PTA_NO_80387, PTA_PKU, PTA_AVX5124VNNIW, PTA_AVX5124FMAPS,
32025 PTA_AVX512VPOPCNTDQ, PTA_SGX, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES,
32026 PTA_AVX512VBMI2, PTA_VPCLMULQDQ, PTA_AVX512BITALG, PTA_RDPID,
32027 PTA_PCONFIG, PTA_WBNOINVD, PTA_AVX512VP2INTERSECT, PTA_PTWRITE,
32028 PTA_AVX512BF16, PTA_WAITPKG, PTA_MOVDIRI, PTA_MOVDIR64B, PTA_ENQCMD,
32029 PTA_CLDEMOTE, PTA_SERIALIZE, PTA_TSXLDTRK, PTA_AMX_TILE, PTA_AMX_INT8,
32030 PTA_AMX_BF16, PTA_UINTR, PTA_HRESET, PTA_KL, PTA_WIDEKL, PTA_AVXVNNI,
32031 PTA_X86_64_BASELINE, PTA_X86_64_V2, PTA_X86_64_V3, PTA_X86_64_V4,
32032 PTA_CORE2, PTA_NEHALEM, PTA_WESTMERE, PTA_SANDYBRIDGE, PTA_IVYBRIDGE,
32033 PTA_HASWELL, PTA_BROADWELL, PTA_SKYLAKE, PTA_SKYLAKE_AVX512,
32034 PTA_CASCADELAKE, PTA_COOPERLAKE, PTA_CANNONLAKE, PTA_ICELAKE_CLIENT,
32035 PTA_ICELAKE_SERVER, PTA_TIGERLAKE, PTA_SAPPHIRERAPIDS, PTA_ALDERLAKE,
32036 PTA_KNL, PTA_BONNELL, PTA_SILVERMONT, PTA_GOLDMONT, PTA_GOLDMONT_PLUS,
32037 PTA_TREMONT, PTA_KNM): Use constexpr instead of const.
32039 2021-02-18 Jakub Jelinek <jakub@redhat.com>
32041 PR middle-end/99109
32042 * gimple-array-bounds.cc (build_zero_elt_array_type): Rename to ...
32043 (build_printable_array_type): ... this. Add nelts argument. For
32044 overaligned eltype, use TYPE_MAIN_VARIANT (eltype) instead. If
32045 nelts, call build_array_type_nelts.
32046 (array_bounds_checker::check_mem_ref): Use build_printable_array_type
32047 instead of build_zero_elt_array_type and build_array_type_nelts.
32049 2021-02-18 Jakub Jelinek <jakub@redhat.com>
32052 * config/i386/i386.c (distance_non_agu_define): Don't call
32053 extract_insn_cached here.
32054 (ix86_lea_outperforms): Save and restore recog_data around call
32055 to distance_non_agu_define and distance_agu_use.
32056 (ix86_ok_to_clobber_flags): Remove.
32057 (ix86_avoid_lea_for_add): Don't call ix86_ok_to_clobber_flags.
32058 (ix86_avoid_lea_for_addr): Likewise. Adjust function comment.
32059 * config/i386/i386.md (*lea<mode>): Change from define_insn_and_split
32060 into define_insn. Move the splitting to define_peephole2 and
32061 check there using peep2_regno_dead_p if FLAGS_REG is dead.
32063 2021-02-17 Julian Brown <julian@codesourcery.com>
32065 * gimplify.c (gimplify_scan_omp_clauses): Handle ATTACH_DETACH
32068 2021-02-17 Xi Ruoyao <xry111@mengyan1223.wang>
32071 * config/mips/mips.c (mips_symbol_insns): Do not use
32072 MSA_SUPPORTED_MODE_P if mode is MAX_MACHINE_MODE.
32074 2021-02-16 Vladimir N. Makarov <vmakarov@redhat.com>
32076 PR inline-asm/98096
32077 * stmt.c (resolve_operand_name_1): Take inout operands into account
32078 for access to labels by names.
32079 * doc/extend.texi: Describe counting operands for accessing labels.
32081 2021-02-16 Richard Biener <rguenther@suse.de>
32083 PR tree-optimization/38474
32084 * tree-ssa-structalias.c (variable_info::address_taken): New.
32085 (new_var_info): Initialize address_taken.
32086 (process_constraint): Set address_taken.
32087 (solve_constraints): Use the new address_taken flag rather
32088 than is_reg_var for sorting variables.
32089 (dump_constraint): Dump the variable number if the name
32092 2021-02-16 Jakub Jelinek <jakub@redhat.com>
32095 * tree-vect-stmts.c (vectorizable_simd_clone_call): For num_calls != 1
32096 multiply by 4096 and for inbranch by 8192.
32097 * config/i386/i386.c (ix86_simd_clone_usable): For TARGET_AVX512F,
32098 return 3, 2 or 1 for mangle letters 'b', 'c' or 'd'.
32100 2021-02-15 Maya Rashish <coypu@sdf.org>
32102 * config/aarch64/aarch64.c (aarch64_init_builtins):
32103 Call SUBTARGET_INIT_BUILTINS.
32105 2021-02-15 Peter Bergner <bergner@linux.ibm.com>
32107 PR rtl-optimization/98872
32108 * init-regs.c (initialize_uninitialized_regs): Skip initialization
32109 if CONST0_RTX is NULL.
32111 2021-02-15 Richard Sandiford <richard.sandiford@arm.com>
32113 PR rtl-optimization/98863
32114 * rtl-ssa/functions.h (function_info::bb_live_out_info): Delete.
32115 (function_info::build_info): Turn into a declaration, moving the
32116 definition to internals.h.
32117 (function_info::bb_walker): Declare.
32118 (function_info::create_reg_use): Likewise.
32119 (function_info::calculate_potential_phi_regs): Take a build_info
32121 (function_info::place_phis, function_info::create_ebbs): Declare.
32122 (function_info::calculate_ebb_live_in_for_debug): Likewise.
32123 (function_info::populate_backedge_phis): Delete.
32124 (function_info::start_block, function_info::end_block): Declare.
32125 (function_info::populate_phi_inputs): Delete.
32126 (function_info::m_potential_phi_regs): Move information to build_info.
32127 * rtl-ssa/internals.h: New file.
32128 (function_info::bb_phi_info): New class.
32129 (function_info::build_info): Moved from functions.h.
32130 Add a constructor and destructor.
32131 (function_info::build_info::ebb_use): Delete.
32132 (function_info::build_info::ebb_def): Likewise.
32133 (function_info::build_info::bb_live_out): Likewise.
32134 (function_info::build_info::tmp_ebb_live_in_for_debug): New variable.
32135 (function_info::build_info::potential_phi_regs): Likewise.
32136 (function_info::build_info::potential_phi_regs_for_debug): Likewise.
32137 (function_info::build_info::ebb_def_regs): Likewise.
32138 (function_info::build_info::bb_phis): Likewise.
32139 (function_info::build_info::bb_mem_live_out): Likewise.
32140 (function_info::build_info::bb_to_rpo): Likewise.
32141 (function_info::build_info::def_stack): Likewise.
32142 (function_info::build_info::old_def_stack_limit): Likewise.
32143 * rtl-ssa/internals.inl (function_info::build_info::record_reg_def):
32144 Remove the regno argument. Push the previous definition onto the
32145 definition stack where necessary.
32146 * rtl-ssa/accesses.cc: Include internals.h.
32147 * rtl-ssa/changes.cc: Likewise.
32148 * rtl-ssa/blocks.cc: Likewise.
32149 (function_info::build_info::build_info): Define.
32150 (function_info::build_info::~build_info): Likewise.
32151 (function_info::bb_walker): New class.
32152 (function_info::bb_walker::bb_walker): Define.
32153 (function_info::add_live_out_use): Convert a logarithmic-complexity
32154 test into a linear one. Allow the same definition to be passed
32156 (function_info::calculate_potential_phi_regs): Moved from
32157 functions.cc. Take a build_info parameter and store the
32158 information there instead.
32159 (function_info::place_phis): New function.
32160 (function_info::add_entry_block_defs): Update call to record_reg_def.
32161 (function_info::calculate_ebb_live_in_for_debug): New function.
32162 (function_info::add_phi_nodes): Use bb_phis to decide which
32163 registers need phi nodes and initialize ebb_def_regs accordingly.
32164 Do not add degenerate phis here.
32165 (function_info::add_artificial_accesses): Use create_reg_use.
32166 Assert that all definitions are listed in the DF LR sets.
32167 Update call to record_reg_def.
32168 (function_info::record_block_live_out): Record live-out register
32169 values in the phis of successor blocks. Use the live-out set
32170 when processing the last block in an EBB, instead of always
32171 using the live-in sets of successor blocks. AND the live sets
32172 with the set of registers that have been defined in the EBB,
32173 rather than with all potential phi registers. Cope correctly
32174 with branches back to the start of the current EBB.
32175 (function_info::start_block): New function.
32176 (function_info::end_block): Likewise.
32177 (function_info::populate_phi_inputs): Likewise.
32178 (function_info::create_ebbs): Likewise.
32179 (function_info::process_all_blocks): Rewrite into a multi-phase
32181 * rtl-ssa/functions.cc: Include internals.h.
32182 (function_info::calculate_potential_phi_regs): Move to blocks.cc.
32183 (function_info::init_function_data): Remove caller.
32184 * rtl-ssa/insns.cc: Include internals.h
32185 (function_info::create_reg_use): New function. Lazily any
32186 degenerate phis needed by the linear RPO view.
32187 (function_info::record_use): Use create_reg_use. When processing
32188 debug uses, use potential_phi_regs and test it before checking
32189 whether the register is live on entry to the current EBB. Lazily
32190 calculate ebb_live_in_for_debug.
32191 (function_info::record_call_clobbers): Update call to record_reg_def.
32192 (function_info::record_def): Likewise.
32194 2021-02-15 Martin Liska <mliska@suse.cz>
32196 * toplev.c (init_asm_output): Free output of
32197 gen_command_line_string function.
32198 (process_options): Likewise.
32200 2021-02-15 Martin Liska <mliska@suse.cz>
32202 * params.opt: Add 2 missing Param keywords.
32204 2021-02-15 Eric Botcazou <ebotcazou@adacore.com>
32206 * df-core.c (df_worklist_dataflow_doublequeue): Use proper cast.
32208 2021-02-15 Jakub Jelinek <jakub@redhat.com>
32210 PR tree-optimization/99079
32211 * match.pd (A % (pow2pcst << N) -> A & ((pow2pcst << N) - 1)): Remove
32212 useless tree_nop_conversion_p (type, TREE_TYPE (@3)) check. Instead
32213 require both type and TREE_TYPE (@1) to be integral types and either
32214 type having smaller or equal precision, or TREE_TYPE (@1) being
32215 unsigned type, or type being signed type. If TREE_TYPE (@1)
32216 doesn't have wrapping overflow, perform the subtraction of one in
32219 2021-02-14 Jan Hubicka <hubicka@ucw.cz>
32220 Richard Biener <rguether@suse.de>
32223 * ipa-reference.c (ipa_init): Only conditinally initialize
32224 reference_vars_to_consider.
32225 (propagate): Conditionally deninitialize reference_vars_to_consider.
32226 (ipa_reference_write_optimization_summary): Sanity check that
32227 reference_vars_to_consider is not allocated.
32229 2021-02-13 Levy Hsu <admin@levyhsu.com>
32232 * config/riscv/riscv-shorten-memrefs.c (pass_shorten_memrefs): Add
32233 extend parameter to get_si_mem_base_reg declaration.
32234 (get_si_mem_base_reg): Add extend parameter. Set it.
32235 (analyze): Pass extend arg to get_si_mem_base_reg.
32236 (transform): Likewise. Use it when rewriting mems.
32237 * config/riscv/riscv.c (riscv_legitimize_move): Check for subword
32238 loads and emit sign/zero extending load followed by subreg move.
32240 2021-02-13 Jim Wilson <jimw@sifive.com>
32243 * config/riscv/riscv.c (riscv_compressed_lw_address_p): Drop early
32244 exit when !reload_completed. Only perform check for compressed reg
32245 if reload_completed.
32246 (riscv_rtx_costs): In MEM case, when optimizing for size and
32247 shorten memrefs, if not compressible, then increase cost.
32249 2021-02-13 Jakub Jelinek <jakub@redhat.com>
32251 PR rtl-optimization/98439
32252 * recog.c (pass_split_before_regstack::gate): Enable even when
32253 pass_split_before_sched2 is enabled if -fselective-scheduling2 is
32256 2021-02-13 Jakub Jelinek <jakub@redhat.com>
32259 * config/i386/mmx.md (*mmx_pshufd_1): Add a combine splitter for
32260 swap of V2SImode elements in memory into DImode memory rotate by 32.
32262 2021-02-12 Martin Sebor <msebor@redhat.com>
32264 * tree-pretty-print.c (print_generic_expr_to_str): Update comment.
32266 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
32268 * rtl-ssa/accesses.cc (function_info::make_use_available): Use
32269 m_temp_obstack rather than m_obstack to allocate the temporary use.
32271 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
32273 * df-problems.c (df_lr_bb_local_compute): Treat partial definitions
32274 as read-modify operations.
32276 2021-02-12 Richard Biener <rguenther@suse.de>
32278 PR middle-end/38474
32279 * ipa-fnsummary.c (unmodified_parm_1): Only walk when
32280 fbi->aa_walk_budget is bigger than zero. Update
32281 fbi->aa_walk_budget.
32282 (param_change_prob): Likewise.
32283 * ipa-prop.c (detect_type_change_from_memory_writes):
32284 Properly account walk_aliased_vdefs.
32285 (parm_preserved_before_stmt_p): Canonicalize updates.
32286 (parm_ref_data_preserved_p): Likewise.
32287 (parm_ref_data_pass_through_p): Likewise.
32288 (determine_known_aggregate_parts): Account own alias queries.
32290 2021-02-12 Martin Liska <mliska@suse.cz>
32292 * opts-common.c (decode_cmdline_option): Release werror_arg.
32293 * opts.c (gen_producer_string): Release output of
32294 gen_command_line_string.
32296 2021-02-12 Richard Biener <rguenther@suse.de>
32298 PR tree-optimization/38474
32299 * params.opt (-param=max-store-chains-to-track=): New param.
32300 (-param=max-stores-to-track=): Likewise.
32301 * doc/invoke.texi (max-store-chains-to-track): Document.
32302 (max-stores-to-track): Likewise.
32303 * gimple-ssa-store-merging.c (pass_store_merging::m_n_chains):
32305 (pass_store_merging::m_n_stores): Likewise.
32306 (pass_store_merging::terminate_and_process_chain): Update
32307 m_n_stores and m_n_chains.
32308 (pass_store_merging::process_store): Likewise. Terminate
32309 oldest chains if the number of stores or chains get too large.
32310 (imm_store_chain_info::terminate_and_process_chain): Dump
32313 2021-02-11 Eric Botcazou <ebotcazou@adacore.com>
32315 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
32316 the cold section, emit a nop before the directive if the previous
32317 active instruction can throw.
32319 2021-02-11 Peter Bergner <bergner@linux.ibm.com>
32322 * config/rs6000/predicates.md (mma_assemble_input_operand): Restrict
32323 memory addresses that are legal for quad word accesses.
32325 2021-02-11 Andrea Corallo <andrea.corallo@arm.com>
32328 * config/arm/thumb2.md (*doloop_end_internal): Generate
32329 alternative sequence to handle long range branches.
32331 2021-02-11 Joel Hutton <joel.hutton@arm.com>
32333 PR tree-optimization/98772
32334 * optabs-tree.c (supportable_half_widening_operation): New function
32335 to check for supportable V8QI->V8HI widening patterns.
32336 * optabs-tree.h (supportable_half_widening_operation): New function.
32337 * tree-vect-stmts.c (vect_create_half_widening_stmts): New function
32338 to create promotion stmts for V8QI->V8HI widening patterns.
32339 (vectorizable_conversion): Add case for V8QI->V8HI.
32341 2021-02-11 Richard Biener <rguenther@suse.de>
32343 * sparseset.h (SPARSESET_ELT_BITS): Remove.
32344 (SPARSESET_ELT_TYPE): Use unsigned int.
32345 * fwprop.c: Do not include sparseset.h.
32347 2021-02-10 Jakub Jelinek <jakub@redhat.com>
32350 * varasm.c (declare_weak): For -fsyntax-only, allow even
32351 TREE_ASM_WRITTEN function decls.
32353 2021-02-10 Jakub Jelinek <jakub@redhat.com>
32356 * config/i386/sse.md (fix<fixunssuffix>_truncv2sfv2di2,
32357 <insn>v8qiv8hi2, <insn>v8qiv8si2, <insn>v4qiv4si2, <insn>v4hiv4si2,
32358 <insn>v8qiv8di2, <insn>v4qiv4di2, <insn>v2qiv2di2, <insn>v4hiv4di2,
32359 <insn>v2hiv2di2, <insn>v2siv2di2): Force operands[1] into REG before
32360 calling simplify_gen_subreg on it.
32362 2021-02-10 Martin Liska <mliska@suse.cz>
32364 * config/nvptx/nvptx.c (nvptx_option_override): Use
32365 flag_patchable_function_entry instead of the removed
32366 function_entry_patch_area_size.
32368 2021-02-10 Martin Liska <mliska@suse.cz>
32370 PR tree-optimization/99002
32371 PR tree-optimization/99026
32372 * gimple-if-to-switch.cc (if_chain::is_beneficial): Fix memory
32373 leak when adjacent cases are merged.
32374 * tree-switch-conversion.c (switch_decision_tree::analyze_switch_statement): Use
32376 (make_pass_lower_switch): Remove trailing whitespace.
32377 * tree-switch-conversion.h (release_clusters): New.
32379 2021-02-10 Richard Biener <rguenther@suse.de>
32381 PR rtl-optimization/99054
32382 * cfgrtl.c (rtl-optimization/99054): Return an auto_vec.
32383 (fixup_partitions): Adjust.
32384 (rtl_verify_edges): Likewise.
32386 2021-02-10 Jakub Jelinek <jakub@redhat.com>
32388 PR middle-end/99007
32389 * gimplify.c (gimplify_scan_omp_clauses): For MEM_REF on reductions,
32390 temporarily disable gimplify_ctxp->into_ssa around gimplify_expr
32393 2021-02-10 Richard Biener <rguenther@suse.de>
32396 * ipa-pure-const.c (propagate_malloc): Use an auto_vec<>
32399 2021-02-10 Richard Biener <rguenther@suse.de>
32401 PR tree-optimization/99024
32402 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Only
32403 clear loop->aux if it is associated with the destroyed loop_vinfo.
32405 2021-02-10 Martin Liska <mliska@suse.cz>
32407 PR tree-optimization/99002
32408 * gimple-if-to-switch.cc (find_conditions): Fix memory leak
32411 2021-02-10 Martin Liska <mliska@suse.cz>
32414 * ipa-icf.c (sem_item::add_reference): Fix memory leak when
32415 a reference exists.
32417 2021-02-10 Jakub Jelinek <jakub@redhat.com>
32420 * dwarf2out.c (prune_unused_types_walk): Mark DW_TAG_variable DIEs
32421 at class scope for DWARF5+.
32423 2021-02-09 Eric Botcazou <ebotcazou@adacore.com>
32425 PR rtl-optimization/96015
32426 * reorg.c (skip_consecutive_labels): Minor comment tweaks.
32427 (relax_delay_slots): When deleting a jump to the next active
32428 instruction over a barrier, first delete the barrier if the
32429 jump is the only way to reach the target label.
32431 2021-02-09 Andre Vieira <andre.simoesdiasvieira@arm.com>
32433 * config/aarch64/aarch64-cost-tables.h: Add entries for vect.mul.
32434 * config/aarch64/aarch64.c (aarch64_rtx_mult_cost): Use vect.mul for
32435 vector multiplies and vect.alu for SSRA.
32436 * config/arm/aarch-common-protos.h (struct vector_cost_table): Define
32437 vect.mul cost field.
32438 * config/arm/aarch-cost-tables.h: Add entries for vect.mul.
32439 * config/arm/arm.c: Likewise.
32441 2021-02-09 Richard Biener <rguenther@suse.de>
32443 PR tree-optimization/98863
32444 * tree-ssa-sccvn.h (vn_avail::next_undo): Add.
32445 * tree-ssa-sccvn.c (last_pushed_avail): New global.
32446 (rpo_elim::eliminate_push_avail): Chain pushed avails.
32447 (unwind_state::avail_top): Add.
32448 (do_unwind): Rewrite unwinding of avail entries.
32449 (do_rpo_vn): Initialize last_pushed_avail and
32450 avail_top of the undo state.
32452 2021-02-09 Jakub Jelinek <jakub@redhat.com>
32454 PR middle-end/99004
32455 * calls.c (maybe_warn_rdwr_sizes): Change s0 and s1 type from
32456 const char * to char * and free those pointers after use.
32458 2021-02-09 Richard Biener <rguenther@suse.de>
32460 PR tree-optimization/99017
32461 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Allow
32462 zero vector cost entries.
32464 2021-02-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
32466 PR middle-end/98974
32467 * tree-vect-stmts.c (vectorizable_condition): Remove shadow vec_num
32468 parameter in vectorizable_condition.
32470 2021-02-08 Richard Biener <rguenther@suse.de>
32473 * tree.c (walk_tree_1): Walk VECTOR_CST elements.
32475 2021-02-08 Martin Liska <mliska@suse.cz>
32478 * cfgexpand.c (pass_expand::execute): Parse per-function option
32479 flag_patchable_function_entry and use it.
32480 * common.opt: Remove function_entry_patch_area_size and
32481 function_entry_patch_area_start global variables.
32482 * opts.c (parse_and_check_patch_area): New function.
32483 (common_handle_option): Use it.
32484 * opts.h (parse_and_check_patch_area): New function.
32485 * toplev.c (process_options): Parse and use
32486 function_entry_patch_area_size.
32488 2021-02-08 Martin Sebor <msebor@redhat.com>
32490 * doc/extend.texi (attribute malloc): Correct typos.
32492 2021-02-05 Nathan Sidwell <nathan@acm.org>
32495 * gcc.c (driver::maybe_run_linker): Check for input file
32496 accessibility if not linking.
32498 2021-02-05 Richard Biener <rguenther@suse.de>
32500 PR tree-optimization/98855
32501 * tree-vectorizer.h (add_stmt_cost): New overload.
32502 * tree-vect-slp.c (li_cost_vec_cmp): New.
32503 (vect_bb_slp_scalar_cost): Cost individual loop regions
32504 separately. Account for the scalar instance root stmt.
32506 2021-02-05 Tom de Vries <tdevries@suse.de>
32509 * tree-switch-conversion.c (jump_table_cluster::emit): Add loc
32511 (bit_test_cluster::emit): Reuse location_t for newly created
32513 (switch_decision_tree::try_switch_expansion): Preserve
32515 * tree-switch-conversion.h: Change function signatures.
32517 2021-02-05 Jakub Jelinek <jakub@redhat.com>
32520 * config/i386/i386-options.c (m_NONE, m_ALL): Define.
32521 * config/i386/x86-tune.def (X86_TUNE_BRANCH_PREDICTION_HINTS,
32522 X86_TUNE_PROMOTE_QI_REGS): Use m_NONE instead of 0U.
32523 (X86_TUNE_QIMODE_MATH): Use m_ALL instead of ~0U.
32525 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32527 * config/aarch64/aarch64-simd-builtins.def (get_high): Define builtin.
32528 * config/aarch64/aarch64-simd.md (aarch64_get_high<mode>): Define.
32529 * config/aarch64/arm_neon.h (__GET_HIGH): Delete.
32530 (vget_high_f16): Reimplement using new builtin.
32531 (vget_high_f32): Likewise.
32532 (vget_high_f64): Likewise.
32533 (vget_high_p8): Likewise.
32534 (vget_high_p16): Likewise.
32535 (vget_high_p64): Likewise.
32536 (vget_high_s8): Likewise.
32537 (vget_high_s16): Likewise.
32538 (vget_high_s32): Likewise.
32539 (vget_high_s64): Likewise.
32540 (vget_high_u8): Likewise.
32541 (vget_high_u16): Likewise.
32542 (vget_high_u32): Likewise.
32543 (vget_high_u64): Likewise.
32545 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32547 * config/aarch64/aarch64-simd-builtins.def (get_low): Define builtin.
32548 * config/aarch64/aarch64-simd.md (aarch64_get_low<mode>): Define.
32549 * config/aarch64/arm_neon.h (__GET_LOW): Delete.
32550 (vget_low_f16): Reimplement using new builtin.
32551 (vget_low_f32): Likewise.
32552 (vget_low_f64): Likewise.
32553 (vget_low_p8): Likewise.
32554 (vget_low_p16): Likewise.
32555 (vget_low_p64): Likewise.
32556 (vget_low_s8): Likewise.
32557 (vget_low_s16): Likewise.
32558 (vget_low_s32): Likewise.
32559 (vget_low_s64): Likewise.
32560 (vget_low_u8): Likewise.
32561 (vget_low_u16): Likewise.
32562 (vget_low_u32): Likewise.
32563 (vget_low_u64): Likewise.
32565 2021-02-05 Kito Cheng <kito.cheng@sifive.com>
32567 * gcc.c (print_multilib_info): Check all required argument is provided
32570 2021-02-05 liuhongt <hongtao.liu@intel.com>
32573 * config/i386/i386-expand.c (ix86_expand_sse_cmp): Don't
32574 generate integer mask comparison for 128/256-bits vector when
32575 op_true/op_false is NULL_RTX or CONSTM1_RTX/CONST0_RTX. Also
32576 delete redundant !maskcmp condition.
32577 (ix86_expand_int_vec_cmp): Ditto but no redundant deletion
32579 (ix86_expand_sse_movcc): Delete definition of maskcmp, add the
32580 condition directly to if (maskcmp), add extra check for
32581 cmpmode, it should be MODE_INT.
32582 (ix86_expand_fp_vec_cmp): Pass NULL to ix86_expand_sse_cmp's
32583 parameters op_true/op_false.
32584 (ix86_use_mask_cmp_p): New.
32586 2021-02-05 liuhongt <hongtao.liu@intel.com>
32589 * config/i386/x86-tune.def (X86_TUNE_AVX256_UNALIGNED_LOAD_OPTIMAL):
32590 Remove m_GENERIC from ~list.
32591 (X86_TUNE_AVX256_UNALIGNED_STORE_OPTIMAL): Ditto.
32593 2021-02-04 David Malcolm <dmalcolm@redhat.com>
32596 * diagnostic-show-locus.c (compatible_locations_p): Require
32597 locations in the same macro map to be either both from the
32598 macro definition, or both from the macro arguments.
32600 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
32602 * config/aarch64/aarch64-simd-builtins.def: Add
32603 [su]mull_hi_lane[q] builtin generator macros.
32604 * config/aarch64/aarch64-simd.md
32605 (aarch64_<su>mull_hi_lane<mode>_insn): Define.
32606 (aarch64_<su>mull_hi_lane<mode>): Define.
32607 (aarch64_<su>mull_hi_laneq<mode>_insn): Define.
32608 (aarch64_<su>mull_hi_laneq<mode>): Define.
32609 * config/aarch64/arm_neon.h (vmull_high_lane_s16): Use RTL
32610 builtin instead of inline asm.
32611 (vmull_high_lane_s32): Likewise.
32612 (vmull_high_lane_u16): Likewise.
32613 (vmull_high_lane_u32): Likewise.
32614 (vmull_high_laneq_s16): Likewise.
32615 (vmull_high_laneq_s32): Likewise.
32616 (vmull_high_laneq_u16): Likewise.
32617 (vmull_high_laneq_u32): Liekwise.
32619 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
32621 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_hi_n
32622 builtin generator macros.
32623 * config/aarch64/aarch64-simd.md
32624 (aarch64_<su>mull_hi_n<mode>_insn): Define.
32625 (aarch64_<su>mull_hi_n<mode>): Define.
32626 * config/aarch64/arm_neon.h (vmull_high_n_s16): Use RTL builtin
32627 instead of inline asm.
32628 (vmull_high_n_s32): Likewise.
32629 (vmull_high_n_u16): Likewise.
32630 (vmull_high_n_u32): Likewise.
32632 2021-02-04 Richard Biener <rguenther@suse.de>
32634 PR tree-optimization/98855
32635 * tree-vect-loop.c (vectorizable_phi): Do not cost
32636 single-argument PHIs.
32637 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Likewise.
32638 * tree-vect-stmts.c (vectorizable_bswap): Also perform
32639 costing for SLP operation.
32641 2021-02-04 Martin Liska <mliska@suse.cz>
32643 * doc/extend.texi: Mention -mprefer-vector-width in target
32646 2021-02-03 Martin Sebor <msebor@redhat.com>
32648 PR tree-optimization/98937
32649 * tree-ssa-strlen.c (strlen_dom_walker::~strlen_dom_walker): Define.
32650 Flush pointer_query cache.
32652 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
32654 * config/rs6000/genfusion.pl (gen_2logical): Add missing
32655 fixes based on patch review.
32656 * config/rs6000/fusion.md: Regenerate file.
32658 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
32660 * config/rs6000/t-rs6000: Comment out auto generation of
32663 2021-02-03 Andrew Stubbs <ams@codesourcery.com>
32665 * config/gcn/gcn-opts.h (enum processor_type): Add PROCESSOR_GFX908.
32666 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Add gfx908.
32667 (output_file_start): Add gfx908.
32668 * config/gcn/gcn.opt (gpu_type): Add gfx908.
32669 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Add march=gfx908.
32670 (MULTILIB_DIRNAMES): Add gfx908.
32671 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): New define.
32672 (main): Recognize gfx908.
32673 * config/gcn/t-omp-device: Add gfx908.
32675 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
32677 * config/aarch64/aarch64-simd-builtins.def: Add
32678 [su]mlsl_hi_lane[q] builtin macro generators.
32679 * config/aarch64/aarch64-simd.md
32680 (aarch64_<su>mlsl_hi_lane<mode>_insn): Define.
32681 (aarch64_<su>mlsl_hi_lane<mode>): Define.
32682 (aarch64_<su>mlsl_hi_laneq<mode>_insn): Define.
32683 (aarch64_<su>mlsl_hi_laneq<mode>): Define.
32684 * config/aarch64/arm_neon.h (vmlsl_high_lane_s16): Use RTL
32685 builtin instead of inline asm.
32686 (vmlsl_high_lane_s32): Likewise.
32687 (vmlsl_high_lane_u16): Likewise.
32688 (vmlsl_high_lane_u32): Likewise.
32689 (vmlsl_high_laneq_s16): Likewise.
32690 (vmlsl_high_laneq_s32): Likewise.
32691 (vmlsl_high_laneq_u16): Likewise.
32692 (vmlsl_high_laneq_u32): Likewise.
32693 (vmlal_high_laneq_u32): Likewise.
32695 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
32697 * config/aarch64/aarch64-simd-builtins.def: Add
32698 [su]mlal_hi_lane[q] builtin generator macros.
32699 * config/aarch64/aarch64-simd.md
32700 (aarch64_<su>mlal_hi_lane<mode>_insn): Define.
32701 (aarch64_<su>mlal_hi_lane<mode>): Define.
32702 (aarch64_<su>mlal_hi_laneq<mode>_insn): Define.
32703 (aarch64_<su>mlal_hi_laneq<mode>): Define.
32704 * config/aarch64/arm_neon.h (vmlal_high_lane_s16): Use RTL
32705 builtin instead of inline asm.
32706 (vmlal_high_lane_s32): Likewise.
32707 (vmlal_high_lane_u16): Likewise.
32708 (vmlal_high_lane_u32): Likewise.
32709 (vmlal_high_laneq_s16): Likewise.
32710 (vmlal_high_laneq_s32): Likewise.
32711 (vmlal_high_laneq_u16): Likewise.
32712 (vmlal_high_laneq_u32): Likewise.
32714 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
32716 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_hi_n
32717 builtin generator macros.
32718 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_hi_n<mode>_insn):
32720 (aarch64_<su>mlsl_hi_n<mode>): Define.
32721 * config/aarch64/arm_neon.h (vmlsl_high_n_s16): Use RTL builtin
32722 instead of inline asm.
32723 (vmlsl_high_n_s32): Likewise.
32724 (vmlsl_high_n_u16): Likewise.
32725 (vmlsl_high_n_u32): Likewise.
32727 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
32729 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_hi_n
32730 builtin generator macros.
32731 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_hi_n<mode>_insn):
32733 (aarch64_<su>mlal_hi_n<mode>): Define.
32734 * config/aarch64/arm_neon.h (vmlal_high_n_s16): Use RTL builtin
32735 instead of inline asm.
32736 (vmlal_high_n_s32): Likewise.
32737 (vmlal_high_n_u16): Likewise.
32738 (vmlal_high_n_u32): Likewise.
32740 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
32742 * config/aarch64/aarch64-simd-builtins.def: Add RTL builtin
32744 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal_hi<mode>):
32746 (aarch64_<su>mlal_hi<mode>_insn): This.
32747 (aarch64_<su>mlal_hi<mode>): Define.
32748 * config/aarch64/arm_neon.h (vmlal_high_s8): Use RTL builtin
32749 instead of inline asm.
32750 (vmlal_high_s16): Likewise.
32751 (vmlal_high_s32): Likewise.
32752 (vmlal_high_u8): Likewise.
32753 (vmlal_high_u16): Likewise.
32754 (vmlal_high_u32): Likewise.
32756 2021-02-03 Ilya Leoshkevich <iii@linux.ibm.com>
32758 * lra-spills.c (remove_pseudos): Call lra_update_insn_recog_data()
32759 after calling alter_subreg() on a (mem).
32761 2021-02-03 Martin Liska <mliska@suse.cz>
32764 * lto-streamer-out.c (produce_lto_section): Fill up missing
32766 * lto-streamer.h (struct lto_section): Add _padding field.
32768 2021-02-03 Richard Biener <rguenther@suse.de>
32770 * lto-streamer.c (lto_get_section_name): Free temporary
32772 * tree-loop-distribution.c
32773 (loop_distribution::merge_dep_scc_partitions): Free edge data.
32775 2021-02-03 Jakub Jelinek <jakub@redhat.com>
32777 PR middle-end/97487
32778 * ifcvt.c (noce_can_force_operand): New function.
32779 (noce_emit_move_insn): Use it.
32780 (noce_try_sign_mask): Likewise. Formatting fix.
32782 2021-02-03 Jakub Jelinek <jakub@redhat.com>
32784 PR middle-end/97971
32785 * lra-constraints.c (process_alt_operands): For inline asm, don't call
32786 fatal_insn, but instead return false.
32788 2021-02-03 Jakub Jelinek <jakub@redhat.com>
32790 PR tree-optimization/98287
32791 * config/i386/mmx.md (<insn><mode>3): For shifts don't enable expander
32794 2021-02-03 Tamar Christina <tamar.christina@arm.com>
32796 PR tree-optimization/98928
32797 * tree-vect-loop.c (vect_analyze_loop_2): Change
32798 STMT_VINFO_SLP_VECT_ONLY to STMT_VINFO_SLP_VECT_ONLY_PATTERN.
32799 * tree-vect-slp-patterns.c (complex_pattern::build): Likewise.
32800 * tree-vectorizer.h (STMT_VINFO_SLP_VECT_ONLY_PATTERN): New.
32801 (class _stmt_vec_info): Add slp_vect_pattern_only_p.
32803 2021-02-02 Richard Biener <rguenther@suse.de>
32805 * gimple-loop-interchange.cc (prepare_data_references):
32807 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
32808 * tree-ssa-loop-im.c (hoist_memory_references): Likewise.
32809 * tree-vect-stmts.c (vectorizable_condition): Do not
32811 (vectorizable_comparison): Likewise.
32813 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32815 * config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
32816 * config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
32817 * config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
32818 (vrsqrteq_u32): Likewise.
32820 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32822 * config/aarch64/aarch64-simd-builtins.def (sqxtun2): Define builtin.
32823 * config/aarch64/aarch64-simd.md (aarch64_sqxtun2<mode>_le): Define.
32824 (aarch64_sqxtun2<mode>_be): Likewise.
32825 (aarch64_sqxtun2<mode>): Likewise.
32826 * config/aarch64/arm_neon.h (vqmovun_high_s16): Reimplement using builtin.
32827 (vqmovun_high_s32): Likewise.
32828 (vqmovun_high_s64): Likewise.
32829 * config/aarch64/iterators.md (UNSPEC_SQXTUN2): Define.
32831 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32833 * config/aarch64/aarch64-simd-builtins.def (bfdot_lane, bfdot_laneq): Use
32835 (bfmlalb_lane, bfmlalt_lane, bfmlalb_lane_q, bfmlalt_lane_q): Use FP flags.
32837 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32839 * config/aarch64/aarch64-simd-builtins.def (fcmla_lane0, fcmla_lane90,
32840 fcmla_lane180, fcmla_lane270, fcmlaq_lane0, fcmlaq_lane90, fcmlaq_lane180,
32841 fcmlaq_lane270, scvtf, ucvtf, fcvtzs, fcvtzu, scvtfsi, scvtfdi, ucvtfsi,
32842 ucvtfdi, fcvtzshf, fcvtzuhf, fmlal_lane_low, fmlsl_lane_low,
32843 fmlal_laneq_low, fmlsl_laneq_low, fmlalq_lane_low, fmlslq_lane_low,
32844 fmlalq_laneq_low, fmlslq_laneq_low, fmlal_lane_high, fmlsl_lane_high,
32845 fmlal_laneq_high, fmlsl_laneq_high, fmlalq_lane_high, fmlslq_lane_high,
32846 fmlalq_laneq_high, fmlslq_laneq_high): Use FP flags.
32848 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32850 * config/aarch64/aarch64-builtins.c (FLAG_LOAD): Define.
32851 * config/aarch64/aarch64-simd-builtins.def (ld1x2, ld2, ld3, ld4, ld2r,
32852 ld3r, ld4r, ld1, ld1x3, ld1x4): Use LOAD flags.
32854 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32856 * config/aarch64/aarch64-simd-builtins.def (combine, zip1, zip2,
32857 uzp1, uzp2, trn1, trn2, simd_bsl): Use AUTO_FP flags.
32859 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32861 * config/aarch64/aarch64-simd-builtins.def (clrsb, clz, ctz, popcount,
32862 vec_smult_lane_, vec_smlal_lane_, vec_smult_laneq_, vec_smlal_laneq_,
32863 vec_umult_lane_, vec_umlal_lane_, vec_umult_laneq_, vec_umlal_laneq_,
32864 ashl, sshl, ushl, srshl, urshl, sdot_lane, udot_lane, sdot_laneq,
32865 udot_laneq, usdot_lane, usdot_laneq, sudot_lane, sudot_laneq, ashr,
32866 ashr_simd, lshr, lshr_simd, srshr_n, urshr_n, ssra_n, usra_n, srsra_n,
32867 ursra_n, sshll_n, ushll_n, sshll2_n, ushll2_n, ssri_n, usri_n, ssli_n,
32868 ssli_n, usli_n, bswap, rbit, simd_bsl, eor3q, rax1q, xarq, bcaxq): Use
32869 NONE builtin flags.
32871 2021-02-02 Jakub Jelinek <jakub@redhat.com>
32873 PR tree-optimization/98848
32874 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
32875 STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
32877 2021-02-02 Kito Cheng <kito.cheng@sifive.com>
32880 * expr.c: Check mode before calling store_expr.
32882 2021-02-02 Christophe Lyon <christophe.lyon@linaro.org>
32884 * config/arm/iterators.md (supf): Remove VORNQ_S and VORNQ_U.
32886 * config/arm/mve.md (mve_vornq_s<mode>): New entry for vorn
32887 instruction using expression ior.
32888 (mve_vornq_u<mode>): New expander.
32889 (mve_vornq_f<mode>): Use ior code instead of unspec.
32890 * config/arm/unspecs.md (VORNQ_S, VORNQ_U, VORNQ_F): Remove.
32892 2021-02-02 Alexandre Oliva <oliva@adacore.com>
32894 * tree-nested.c (convert_nonlocal_reference_op): Move
32895 current_function_decl restore after re-gimplification.
32896 (convert_local_reference_op): Likewise.
32898 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32900 * config/aarch64/aarch64-simd-builtins.def (rshrn, rshrn2):
32902 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>_insn_le):
32904 (aarch64_rshrn<mode>_insn_be): Likewise.
32905 (aarch64_rshrn<mode>): Likewise.
32906 (aarch64_rshrn2<mode>_insn_le): Likewise.
32907 (aarch64_rshrn2<mode>_insn_be): Likewise.
32908 (aarch64_rshrn2<mode>): Likewise.
32909 * config/aarch64/aarch64.md (unspec): Add UNSPEC_RSHRN.
32910 * config/aarch64/arm_neon.h (vrshrn_high_n_s16): Reimplement
32912 (vrshrn_high_n_s32): Likewise.
32913 (vrshrn_high_n_s64): Likewise.
32914 (vrshrn_high_n_u16): Likewise.
32915 (vrshrn_high_n_u32): Likewise.
32916 (vrshrn_high_n_u64): Likewise.
32917 (vrshrn_n_s16): Likewise.
32918 (vrshrn_n_s32): Likewise.
32919 (vrshrn_n_s64): Likewise.
32920 (vrshrn_n_u16): Likewise.
32921 (vrshrn_n_u32): Likewise.
32922 (vrshrn_n_u64): Likewise.
32924 2021-02-01 Sergei Trofimovich <siarheit@google.com>
32926 PR tree-optimization/98499
32927 * ipa-modref.c (analyze_ssa_name_flags): treat RVO
32928 conservatively and assume all possible side-effects.
32930 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32932 * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi,
32933 vec_unpacku_hi_): Define builtins.
32934 * config/aarch64/arm_neon.h (vmovl_high_s8): Reimplement using
32936 (vmovl_high_s16): Likewise.
32937 (vmovl_high_s32): Likewise.
32938 (vmovl_high_u8): Likewise.
32939 (vmovl_high_u16): Likewise.
32940 (vmovl_high_u32): Likewise.
32942 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32944 * config/aarch64/aarch64-simd-builtins.def (sabdl, uabdl):
32946 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl<mode>): New
32948 * config/aarch64/aarch64.md (unspec): Define UNSPEC_SABDL,
32950 * config/aarch64/arm_neon.h (vabdl_s8): Reimplemet using
32952 (vabdl_s16): Likewise.
32953 (vabdl_s32): Likewise.
32954 (vabdl_u8): Likewise.
32955 (vabdl_u16): Likewise.
32956 (vabdl_u32): Likewise.
32957 * config/aarch64/iterators.md (ABDL): New int iterator.
32958 (sur): Handle UNSPEC_SABDL, UNSPEC_UABDL.
32960 2021-02-01 Martin Sebor <msebor@redhat.com>
32962 * tree.h (BLOCK_VARS): Add comment.
32963 (BLOCK_SUBBLOCKS): Same.
32964 (BLOCK_SUPERCONTEXT): Same.
32965 (BLOCK_ABSTRACT_ORIGIN): Same.
32966 (inlined_function_outer_scope_p): Same.
32968 2021-02-01 Martin Sebor <msebor@redhat.com>
32970 PR middle-end/97172
32971 * attribs.c (attr_access::free_lang_data): Define new function.
32972 * attribs.h (attr_access::free_lang_data): Declare new function.
32974 2021-02-01 Richard Biener <rguenther@suse.de>
32976 * vec.h (auto_vec::auto_vec): Add memory stat parameters
32978 * bitmap.h (auto_bitmap::auto_bitmap): Likewise.
32980 2021-02-01 Tamar Christina <tamar.christina@arm.com>
32982 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>,
32983 aarch64_<su>mlsl<mode>, aarch64_<su>mlsl_n<mode>): Flip mult operands.
32985 2021-02-01 Richard Biener <rguenther@suse.de>
32987 PR rtl-optimization/98863
32988 * config/i386/i386-features.c (convert_scalars_to_vector):
32989 Set DF_RD_PRUNE_DEAD_DEFS.
32991 2021-01-31 Eric Botcazou <ebotcazou@adacore.com>
32993 * system.h (SIZE_MAX): Define if not already defined.
32995 2021-01-30 Aaron Sawdey <acsawdey@linux.ibm.com>
32997 * config/rs6000/genfusion.pl (gen_2logical): New function to
32998 generate patterns for logical-logical fusion.
32999 * config/rs6000/fusion.md: Regenerated patterns.
33000 * config/rs6000/rs6000-cpus.def: Add
33001 OPTION_MASK_P10_FUSION_2LOGICAL.
33002 * config/rs6000/rs6000.c (rs6000_option_override_internal):
33003 Enable logical-logical fusion for p10.
33004 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2logical.
33006 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
33008 * config/rs6000/rs6000.opt: Add periods to new AIX options.
33010 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
33012 * config/rs6000/rs6000.opt (mabi=vec-extabi): New.
33013 (mabi=vec-default): New.
33014 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
33015 __EXTABI__ for AIX Vector extended ABI.
33016 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print AIX Vector
33018 (conditional_register_usage): If AIX vec_extabi enabled, vs20-vs31
33020 * doc/invoke.texi (PowerPC mabi): Add AIX vec-extabi and vec-default.
33022 2021-01-30 Jakub Jelinek <jakub@redhat.com>
33024 * config/i386/i386-features.c (remove_partial_avx_dependency): Clear
33025 DF_DEFER_INSN_RESCAN after calling df_process_deferred_rescans.
33027 2021-01-29 Vladimir N. Makarov <vmakarov@redhat.com>
33030 * lra-constraints.c (in_class_p): Don't narrow class only for REG
33033 2021-01-29 Will Schmidt <will_schmidt@vnet.ibm.com>
33035 * config/rs6000/rs6000-call.c (rs6000_expand_binup_builtin): Add
33036 clauses for CODE_FOR_vsx_xvcvuxddp_scale and
33037 CODE_FOR_vsx_xvcvsxddp_scale to the parameter checking code.
33039 2021-01-29 Andrew MacLeod <amacleod@redhat.com>
33041 PR tree-optimization/98866
33042 * gimple-range-gori.h (gori_compute:set_range_invariant): New.
33043 * gimple-range-gori.cc (gori_map::set_range_invariant): New.
33044 (gori_map::m_maybe_invariant): Rename from all_outgoing.
33045 (gori_map::gori_map): Rename all_outgoing to m_maybe_invariant.
33046 (gori_map::is_export_p): Ditto.
33047 (gori_map::calculate_gori): Ditto.
33048 (gori_compute::set_range_invariant): New.
33049 * gimple-range.cc (gimple_ranger::range_of_stmt): Set range
33050 invariant for pointers evaluating to [1, +INF].
33052 2021-01-29 Richard Biener <rguenther@suse.de>
33054 PR rtl-optimization/98863
33055 * config/i386/i386-features.c (remove_partial_avx_dependency):
33056 Do not perform DF analysis.
33057 (pass_data_remove_partial_avx_dependency): Remove
33060 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
33062 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_n
33063 builtin generator macros.
33064 * config/aarch64/aarch64-simd.md (aarch64_<su>mull_n<mode>):
33066 * config/aarch64/arm_neon.h (vmull_n_s16): Use RTL builtin
33067 instead of inline asm.
33068 (vmull_n_s32): Likewise.
33069 (vmull_n_u16): Likewise.
33070 (vmull_n_u32): Likewise.
33072 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33074 * config/aarch64/aarch64-simd-builtins.def (sabdl2, uabdl2):
33076 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
33078 (aarch64_<sur>abdl2<mode>): ... This.
33079 (<sur>sadv16qi): Adjust use of above.
33080 * config/aarch64/arm_neon.h (vabdl_high_s8): Reimplement using
33082 (vabdl_high_s16): Likewise.
33083 (vabdl_high_s32): Likewise.
33084 (vabdl_high_u8): Likewise.
33085 (vabdl_high_u16): Likewise.
33086 (vabdl_high_u32): Likewise.
33088 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33090 * config/aarch64/aarch64-simd-builtins.def (sabal2): Define
33092 (uabal2): Likewise.
33093 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal2<mode>): New
33095 * config/aarch64/aarch64.md (unspec): Add UNSPEC_SABAL2 and
33097 * config/aarch64/arm_neon.h (vabal_high_s8): Reimplement using
33099 (vabal_high_s16): Likewise.
33100 (vabal_high_s32): Likewise.
33101 (vabal_high_u8): Likewise.
33102 (vabal_high_u16): Likewise.
33103 (vabal_high_u32): Likewise.
33104 * config/aarch64/iterators.md (ABAL2): New mode iterator.
33105 (sur): Handle UNSPEC_SABAL2, UNSPEC_UABAL2.
33107 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33109 * config/aarch64/aarch64-simd-builtins.def (sabal): Define
33112 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal<mode>_4):
33114 (aarch64_<sur>abal<mode>): ... This
33115 (<sur>sadv16qi): Adust use of the above.
33116 * config/aarch64/arm_neon.h (vabal_s8): Reimplement using
33118 (vabal_s16): Likewise.
33119 (vabal_s32): Likewise.
33120 (vabal_u8): Likewise.
33121 (vabal_u16): Likewise.
33122 (vabal_u32): Likewise.
33124 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33126 * config/aarch64/aarch64-simd-builtins.def (saddlv, uaddlv):
33128 * config/aarch64/aarch64-simd.md (aarch64_<su>addlv<mode>):
33130 * config/aarch64/arm_neon.h (vaddlv_s8): Reimplement using
33132 (vaddlv_s16): Likewise.
33133 (vaddlv_u8): Likewise.
33134 (vaddlv_u16): Likewise.
33135 (vaddlvq_s8): Likewise.
33136 (vaddlvq_s16): Likewise.
33137 (vaddlvq_s32): Likewise.
33138 (vaddlvq_u8): Likewise.
33139 (vaddlvq_u16): Likewise.
33140 (vaddlvq_u32): Likewise.
33141 (vaddlv_s32): Likewise.
33142 (vaddlv_u32): Likewise.
33143 * config/aarch64/iterators.md (VDQV_L): New mode iterator.
33144 (unspec): Add UNSPEC_SADDLV, UNSPEC_UADDLV.
33145 (Vwstype): New mode attribute.
33147 (VWIDE_S): Likewise.
33148 (USADDLV): New int iterator.
33149 (su): Handle UNSPEC_SADDLV, UNSPEC_UADDLV.
33151 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
33153 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_lane[q]
33154 builtin generator macros.
33155 * config/aarch64/aarch64-simd.md (aarch64_vec_<su>mlsl_lane<Qlane>):
33157 * config/aarch64/arm_neon.h (vmlsl_lane_s16): Use RTL builtin
33158 instead of inline asm.
33159 (vmlsl_lane_s32): Likewise.
33160 (vmlsl_lane_u16): Likewise.
33161 (vmlsl_lane_u32): Likewise.
33162 (vmlsl_laneq_s16): Likewise.
33163 (vmlsl_laneq_s32): Likewise.
33164 (vmlsl_laneq_u16): Likewise.
33165 (vmlsl_laneq_u32): Likewise.
33167 2021-01-29 Richard Biener <rguenther@suse.de>
33169 * doc/invoke.texi (--param max-gcse-memory): Document unit
33171 * gcse.c (gcse_or_cprop_is_too_expensive): Adjust.
33172 * params.opt (--param max-gcse-memory): Adjust default and
33173 document unit of size.
33175 2021-01-29 Richard Biener <rguenther@suse.de>
33177 PR rtl-optimization/98863
33178 * gcse.c (gcse_or_cprop_is_too_expensive): Use unsigned
33179 HOST_WIDE_INT for the memory estimate.
33181 2021-01-29 Bin Cheng <bin.cheng@linux.alibaba.com>
33182 Richard Biener <rguenther@suse.de>
33184 PR tree-optimization/97627
33185 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
33186 Do not analyze fake edges.
33188 2021-01-29 Richard Biener <rguenther@suse.de>
33190 PR rtl-optimization/98144
33191 * df.h (df_mir_bb_info): Add con_visited member.
33192 * df-problems.c (df_mir_alloc): Initialize con_visited,
33193 do not fully populate IN and OUT.
33194 (df_mir_reset): Likewise.
33195 (df_mir_confluence_0): Set con_visited.
33196 (df_mir_confluence_n): Properly handle implicitely
33197 fully populated IN and OUT as designated by con_visited
33198 and update con_visited accordingly.
33200 2021-01-29 Jakub Jelinek <jakub@redhat.com>
33203 * config/arm/vec-common.md (mve_vshlq_<supf><mode>,
33204 vashl<mode>3, vashr<mode>3, vlshr<mode>3): Add
33205 && !TARGET_REALLY_IWMMXT to conditions.
33207 2021-01-29 Jakub Jelinek <jakub@redhat.com>
33210 * cfgbuild.c (find_bb_boundaries): Reset debug_insn when seeing
33213 2021-01-28 Marek Polacek <polacek@redhat.com>
33216 * stor-layout.c (finalize_type_size): If we reset TYPE_USER_ALIGN in
33217 the main variant, maybe reset it in its variants too.
33218 * tree.c (check_base_type): Return true only if TYPE_USER_ALIGN match.
33219 (check_aligned_type): Check if TYPE_USER_ALIGN match.
33221 2021-01-28 Christophe Lyon <christophe.lyon@linaro.org>
33224 * config/arm/arm.c (arm_rtx_costs_internal): Adjust cost of vector
33225 of constant zero for comparisons.
33227 2021-01-28 Michael Meissner <meissner@linux.ibm.com>
33229 * config/rs6000/rs6000.c (rs6000_mangle_decl_assembler_name): Add
33230 support for mapping built-in function names for long double
33231 built-in functions if long double is IEEE 128-bit.
33233 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
33235 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_n
33236 builtin generator macros.
33237 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_n<mode>):
33239 * config/aarch64/arm_neon.h (vmlsl_n_s16): Use RTL builtin
33240 instead of inline asm.
33241 (vmlsl_n_s32): Likewise.
33242 (vmlsl_n_u16): Likewise.
33243 (vmlsl_n_u32): Likewise.
33245 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
33247 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_n
33248 builtin generator macros.
33249 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>):
33251 * config/aarch64/arm_neon.h (vmlal_n_s16): Use RTL builtin
33252 instead of inline asm.
33253 (vmlal_n_s32): Likewise.
33254 (vmlal_n_u16): Likewise.
33255 (vmlal_n_u32): Likewise.
33257 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33259 * config/aarch64/aarch64-simd-builtins.def (shrn2): Define
33261 * config/aarch64/aarch64-simd.md (aarch64_shrn2<mode>_insn_le):
33263 (aarch64_shrn2<mode>_insn_be): Likewise.
33264 (aarch64_shrn2<mode>): Likewise.
33265 * config/aarch64/arm_neon.h (vshrn_high_n_s16): Reimlplement
33267 (vshrn_high_n_s32): Likewise.
33268 (vshrn_high_n_s64): Likewise.
33269 (vshrn_high_n_u16): Likewise.
33270 (vshrn_high_n_u32): Likewise.
33271 (vshrn_high_n_u64): Likewise.
33273 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33275 * config/aarch64/aarch64-simd-builtins.def (shrn): Define
33277 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le):
33279 (aarch64_shrn<mode>_insn_be): Likewise.
33280 (aarch64_shrn<mode>): Likewise.
33281 * config/aarch64/arm_neon.h (vshrn_n_s16): Reimplement using
33283 (vshrn_n_s32): Likewise.
33284 (vshrn_n_s64): Likewise.
33285 (vshrn_n_u16): Likewise.
33286 (vshrn_n_u32): Likewise.
33287 (vshrn_n_u64): Likewise.
33288 * config/aarch64/iterators.md (vn_mode): New mode attribute.
33290 2021-01-28 Richard Biener <rguenther@suse.de>
33292 PR rtl-optimization/80960
33293 * dse.c (check_mem_read_rtx): Call get_addr on the
33296 2021-01-28 Xionghu Luo <luoxhu@linux.ibm.com>
33297 David Edelsohn <dje.gcc@gmail.com>
33300 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
33301 Don't generate VIEW_CONVERT_EXPR for fcode ALTIVEC_BUILTIN_VEC_INSERT
33303 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
33305 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Remove the
33306 wrapper call rs6000_expand_vector_set_var for cleanup. Call
33307 rs6000_expand_vector_set_var_p9 and rs6000_expand_vector_set_var_p8
33309 (rs6000_expand_vector_set_var): Delete.
33310 (rs6000_expand_vector_set_var_p9): Make static.
33311 (rs6000_expand_vector_set_var_p8): Make static.
33313 2021-01-28 Xing GUO <higuoxing@gmail.com>
33315 * common/config/riscv/riscv-common.c
33316 (riscv_subset_list::parsing_subset_version): Fix -march option parsing
33317 when `p` extension exists.
33319 2021-01-27 Vladimir N. Makarov <vmakarov@redhat.com>
33321 PR rtl-optimization/97684
33322 * ira.c (ira): Call ira_set_pseudo_classes before
33323 update_equiv_regs when it is necessary.
33325 2021-01-27 Jakub Jelinek <jakub@redhat.com>
33328 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Use
33329 %w0, %w1 and %2 instead of %0, %1 and %2.
33331 2021-01-27 Aaron Sawdey <acsawdey@linux.ibm.com>
33333 * config/rs6000/genfusion.pl: New script to generate
33334 define_insn_and_split patterns so combine can arrange fused
33335 instructions next to each other.
33336 * config/rs6000/fusion.md: New file, generated fused instruction
33337 patterns for combine.
33338 * config/rs6000/predicates.md (const_m1_to_1_operand): New predicate.
33339 (non_update_memory_operand): New predicate.
33340 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_P10_FUSION and
33341 OPTION_MASK_P10_FUSION_LD_CMPI to ISA_3_1_MASKS_SERVER and
33343 * config/rs6000/rs6000-protos.h (address_is_non_pfx_d_or_x): Add
33345 * config/rs6000/rs6000.c (rs6000_option_override_internal):
33346 Automatically set OPTION_MASK_P10_FUSION and
33347 OPTION_MASK_P10_FUSION_LD_CMPI if target is power10.
33348 (rs600_opt_masks): Allow -mpower10-fusion
33349 in function attributes.
33350 (address_is_non_pfx_d_or_x): New function.
33351 * config/rs6000/rs6000.h: Add MASK_P10_FUSION.
33352 * config/rs6000/rs6000.md: Include fusion.md.
33353 * config/rs6000/rs6000.opt: Add -mpower10-fusion
33354 and -mpower10-fusion-ld-cmpi.
33355 * config/rs6000/t-rs6000: Add dependencies involving fusion.md.
33357 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
33359 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal
33360 builtin generator macros.
33361 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal<mode>):
33363 (aarch64_<su>mlal<mode>): This.
33364 * config/aarch64/arm_neon.h (vmlal_s8): Use RTL builtin
33365 instead of inline asm.
33366 (vmlal_s16): Likewise.
33367 (vmlal_s32): Likewise.
33368 (vmlal_u8): Likewise.
33369 (vmlal_u16): Likewise.
33370 (vmlal_u32): Likewise.
33372 2021-01-27 Richard Biener <rguenther@suse.de>
33374 PR tree-optimization/98854
33375 * tree-vect-slp.c (vect_build_slp_tree_2): Also build
33376 PHIs from scalars when the number of CTORs matches the
33377 number of children.
33379 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
33381 * config/aarch64/aarch64-simd-builtins.def: Add mls_n builtin
33383 * config/aarch64/aarch64-simd.md (*aarch64_mls_elt_merge<mode>):
33385 (aarch64_mls_n<mode>): This.
33386 * config/aarch64/arm_neon.h (vmls_n_s16): Use RTL builtin
33388 (vmls_n_s32): Likewise.
33389 (vmls_n_u16): Likewise.
33390 (vmls_n_u32): Likewise.
33391 (vmlsq_n_s16): Likewise.
33392 (vmlsq_n_s32): Likewise.
33393 (vmlsq_n_u16): Likewise.
33394 (vmlsq_n_u32): Likewise.
33396 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
33398 * config/aarch64/aarch64-simd-builtins.def: Add mls builtin
33400 * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather
33402 (vmls_s16): Likewise.
33403 (vmls_s32): Likewise.
33404 (vmls_u8): Likewise.
33405 (vmls_u16): Likewise.
33406 (vmls_u32): Likewise.
33407 (vmlsq_s8): Likewise.
33408 (vmlsq_s16): Likewise.
33409 (vmlsq_s32): Likewise.
33410 (vmlsq_u8): Likewise.
33411 (vmlsq_u16): Likewise.
33412 (vmlsq_u32): Likewise.
33414 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
33416 * config/aarch64/aarch64-simd-builtins.def: Add mla_n builtin
33418 * config/aarch64/aarch64-simd.md (*aarch64_mla_elt_merge<mode>):
33420 (aarch64_mla_n<mode>): This.
33421 * config/aarch64/arm_neon.h (vmla_n_s16): Use RTL builtin
33423 (vmla_n_s32): Likewise.
33424 (vmla_n_u16): Likewise.
33425 (vmla_n_u32): Likewise.
33426 (vmlaq_n_s16): Likewise.
33427 (vmlaq_n_s32): Likewise.
33428 (vmlaq_n_u16): Likewise.
33429 (vmlaq_n_u32): Likewise.
33431 2021-01-27 liuhongt <hongtao.liu@intel.com>
33434 * config/i386/sse.md (sse2_gt<mode>3): Drop !TARGET_XOP in condition.
33435 (*sse2_eq<mode>3): Ditto.
33437 2021-01-27 Jakub Jelinek <jakub@redhat.com>
33439 * tree-pass.h (PROP_trees): Rename to ...
33440 (PROP_gimple): ... this.
33441 * cfgexpand.c (pass_data_expand): Replace PROP_trees with PROP_gimple.
33442 * passes.c (execute_function_dump, execute_function_todo,
33443 execute_one_ipa_transform_pass, execute_one_pass): Likewise.
33444 * varpool.c (ctor_for_folding): Likewise.
33446 2021-01-27 Jakub Jelinek <jakub@redhat.com>
33448 PR tree-optimization/97260
33449 * varpool.c: Include tree-pass.h.
33450 (ctor_for_folding): In GENERIC return DECL_INITIAL for TREE_READONLY
33451 non-TREE_SIDE_EFFECTS automatic variables.
33453 2021-01-26 Paul Fee <paul.f.fee@gmail.com>
33455 * doc/cpp.texi (__cplusplus): Document value for -std=c++23
33457 * doc/invoke.texi: Document -std=c++23 and -std=gnu++23.
33458 * dwarf2out.c (highest_c_language): Recognise C++20 and C++23.
33459 (gen_compile_unit_die): Recognise C++23.
33461 2021-01-26 Jakub Jelinek <jakub@redhat.com>
33464 * dwarf2asm.c (dw2_assemble_integer): Cast DWARF2_ADDR_SIZE to int
33467 2021-01-26 Jakub Jelinek <jakub@redhat.com>
33470 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
33471 Use UINTVAL (shft_amnt) and UINTVAL (mask) instead of INTVAL (shft_amnt)
33472 and INTVAL (mask). Add && INTVAL (mask) > 0 condition.
33474 2021-01-26 Richard Biener <rguenther@suse.de>
33476 * gimple-pretty-print.c (dump_binary_rhs): Handle
33477 VEC_WIDEN_{PLUS,MINUS}_{LO,HI}_EXPR.
33479 2021-01-26 Richard Biener <rguenther@suse.de>
33481 PR middle-end/98726
33482 * tree.h (vector_cst_int_elt): Remove.
33483 * tree.c (vector_cst_int_elt): Use poly_wide_int for computations,
33486 2021-01-26 Andrew Stubbs <ams@codesourcery.com>
33488 * config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
33489 for V64DFmode min/max reductions.
33491 2021-01-26 Jakub Jelinek <jakub@redhat.com>
33493 * dwarf2asm.c (dw2_assemble_integer): Handle size twice as large
33494 as DWARF2_ADDR_SIZE if x is not a scalar int by emitting it as
33495 two halves, one with x and the other with const0_rtx, ordered
33496 depending on endianity.
33498 2021-01-26 Alexandre Oliva <oliva@adacore.com>
33500 * gimplify.c (gimplify_decl_expr): Skip asan marking calls for
33501 temporaries not seen in binding block, and not about to be
33502 added as gimple variables.
33504 2021-01-25 Martin Sebor <msebor@redhat.com>
33507 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust warning text.
33509 2021-01-25 Martin Liska <mliska@suse.cz>
33511 * value-prof.c (get_nth_most_common_value): Use %s instead
33514 2021-01-25 Jakub Jelinek <jakub@redhat.com>
33517 * configure.ac (HAVE_AS_GDWARF_5_DEBUG_FLAG): Only define if
33518 readelf -wi is able to read the emitted .debug_info back.
33519 * configure: Regenerated.
33521 2021-01-25 Martin Liska <mliska@suse.cz>
33523 PR gcov-profile/98739
33524 * common.opt: Add missing sign symbol.
33525 * value-prof.c (get_nth_most_common_value): Restore handling
33526 of PROFILE_REPRODUCIBILITY_PARALLEL_RUNS and
33527 PROFILE_REPRODUCIBILITY_MULTITHREADED.
33529 2021-01-25 Richard Biener <rguenther@suse.de>
33531 PR middle-end/98807
33532 * tree.c (vector_element_bits): Always use precision of
33533 the element type for boolean vectors.
33535 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
33537 * config/rtems.h (STARTFILE_SPEC): Remove qnolinkcmds.
33538 (ENDFILE_SPEC): Evaluate qnolinkcmds.
33540 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
33542 * config/rtems.h (STARTFILE_SPEC): Remove nostdlib and
33543 nostartfiles handling since this is already done by
33544 LINK_COMMAND_SPEC. Evaluate qnolinkcmds.
33545 (ENDFILE_SPEC): Remove nostdlib and nostartfiles handling since this
33546 is already done by LINK_COMMAND_SPEC.
33547 (LIB_SPECS): Remove nostdlib and nodefaultlibs handling since
33548 this is already done by LINK_COMMAND_SPEC. Remove qnolinkcmds
33551 2021-01-25 Jakub Jelinek <jakub@redhat.com>
33554 * fold-const-call.c (host_size_t_cst_p): Renamed to ...
33555 (size_t_cst_p): ... this. Check and store unsigned HOST_WIDE_INT
33556 value rather than host size_t.
33557 (fold_const_call): Change type of s2 from size_t to
33558 unsigned HOST_WIDE_INT. Use size_t_cst_p instead of
33559 host_size_t_cst_p. For strncmp calls, pass MIN (s2, SIZE_MAX)
33560 instead of s2 as last argument.
33562 2021-01-25 Tamar Christina <tamar.christina@arm.com>
33564 * config/arm/iterators.md (rotsplit1, rotsplit2, conj_op, fcmac1,
33565 VCMLA_OP, VCMUL_OP): New.
33566 * config/arm/mve.md (mve_vcmlaq<mve_rot><mode>): Support vec_dup 0.
33567 * config/arm/neon.md (cmul<conj_op><mode>3): New.
33568 * config/arm/unspecs.md (UNSPEC_VCMLA_CONJ, UNSPEC_VCMLA180_CONJ,
33569 UNSPEC_VCMUL_CONJ): New.
33570 * config/arm/vec-common.md (cmul<conj_op><mode>3, arm_vcmla<rot><mode>,
33571 cml<fcmac1><conj_op><mode>4): New.
33573 2021-01-23 Jakub Jelinek <jakub@redhat.com>
33576 * config/rs6000/mmintrin.h (__m64): Add __may_alias__ attribute.
33578 2021-01-22 Jonathan Wright <jonathan.wright@arm.com>
33580 * config/aarch64/aarch64-simd-builtins.def: Add mla builtin
33582 * config/aarch64/arm_neon.h (vmla_s8): Use RTL builtin rather
33584 (vmla_s16): Likewise.
33585 (vmla_s32): Likewise.
33586 (vmla_u8): Likewise.
33587 (vmla_u16): Likewise.
33588 (vmla_u32): Likewise.
33589 (vmlaq_s8): Likewise.
33590 (vmlaq_s16): Likewise.
33591 (vmlaq_s32): Likewise.
33592 (vmlaq_u8): Likewise.
33593 (vmlaq_u16): Likewise.
33594 (vmlaq_u32): Likewise.
33596 2021-01-22 David Malcolm <dmalcolm@redhat.com>
33598 * doc/invoke.texi (GCC_EXTRA_DIAGNOSTIC_OUTPUT): Add @findex
33601 2021-01-22 Jakub Jelinek <jakub@redhat.com>
33604 * dwarf2out.c (output_file_names): For -gdwarf-5, if there are no
33605 filenames to emit, still emit the required 0 index directory and
33606 filename entries that match DW_AT_comp_dir and DW_AT_name of the
33609 2021-01-22 Marek Polacek <polacek@redhat.com>
33612 * doc/invoke.texi: Update C++ ABI Version 15 description.
33614 2021-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33616 PR tree-optimization/98766
33617 * tree-ssa-math-opts.c (convert_mult_to_fma): Use maybe_le when
33618 comparing against type size with param_avoid_fma_max_bits.
33620 2021-01-22 Richard Biener <rguenther@suse.de>
33622 PR middle-end/98793
33623 * tree.c (vector_element_bits): Key single-bit bool vector on
33624 integer mode rather than not vector mode.
33626 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
33629 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
33630 Generate ARRAY_REF(VIEW_CONVERT_EXPR) for P8 and later
33632 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): Update
33633 to call different path for P8 and P9.
33634 (rs6000_expand_vector_set_var_p9): New function.
33635 (rs6000_expand_vector_set_var_p8): New function.
33637 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
33641 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
33642 Ajdust variable index vec_insert from address dereference to
33643 ARRAY_REF(VIEW_CONVERT_EXPR) tree expression.
33644 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
33646 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): New function.
33648 2021-01-22 Martin Liska <mliska@suse.cz>
33650 PR gcov-profile/98739
33651 * profile.c (compute_value_histograms): Drop time profile for
33652 -fprofile-reproducible=multithreaded.
33654 2021-01-22 Nathan Sidwell <nathan@acm.org>
33656 * gcc.c (process_command): Don't check OPT_SPECIAL_input_file
33659 2021-01-22 Richard Biener <rguenther@suse.de>
33661 PR middle-end/98773
33662 * tree-data-ref.c (initalize_matrix_A): Revert previous
33663 change, retaining failing on HOST_WIDE_INT_MIN CHREC_RIGHT.
33665 2021-01-22 Jakub Jelinek <jakub@redhat.com>
33667 PR tree-optimization/90248
33668 * match.pd (X cmp 0.0 ? 1.0 : -1.0 -> copysign(1, +-X),
33669 X cmp 0.0 ? -1.0 : +1.0 -> copysign(1, -+X)): Remove
33671 (X * (X cmp 0.0 ? 1.0 : -1.0) -> +-abs(X),
33672 X * (X cmp 0.0 ? -1.0 : 1.0) -> +-abs(X)): New simplifications.
33674 2021-01-22 Jakub Jelinek <jakub@redhat.com>
33676 PR tree-optimization/98255
33677 * tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
33678 extend index - low_bound from sizetype's precision rather than index
33680 (get_addr_base_and_unit_offset_1): Likewise.
33681 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
33682 * gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
33684 2021-01-22 Richard Biener <rguenther@suse.de>
33686 PR tree-optimization/98786
33687 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Avoid
33688 adding new uses of abnormals. Verify we deal with a conditional
33691 2021-01-22 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
33694 * optc-save-gen.awk: Add arm_fp16_format to checked_options.
33696 2021-01-22 liuhongt <hongtao.liu@intel.com>
33700 * config/i386/sse.md (VI_128_256): New mode iterator.
33701 (*avx_cmp<mode>3_1, *avx_cmp<mode>3_2, *avx_cmp<mode>3_3,
33702 *avx_cmp<mode>3_4, *avx2_eq<mode>3, *avx2_pcmp<mode>3_1,
33703 *avx2_pcmp<mode>3_2, *avx2_gt<mode>3): New
33704 define_insn_and_split to lower avx512 vector comparison to avx
33705 version when dest is vector.
33706 (*<avx512>_cmp<mode>3,*<avx512>_cmp<mode>3,*<avx512>_ucmp<mode>3):
33707 define_insn_and_split for negating the comparison result.
33708 * config/i386/predicates.md (float_vector_all_ones_operand):
33710 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Use
33711 general NOT operator without UNSPEC_MASKOP.
33713 2021-01-21 Vladimir N. Makarov <vmakarov@redhat.com>
33715 PR rtl-optimization/98777
33716 * lra-int.h (lra_pmode_pseudo): New extern.
33717 * lra.c (lra_pmode_pseudo): New global.
33719 * lra-eliminations.c (eliminate_regs_in_insn): Use it.
33721 2021-01-21 Ilya Leoshkevich <iii@linux.ibm.com>
33723 * fwprop.c (fwprop_propagation::classify_result): Allow
33724 (subreg (mem)) simplifications.
33726 2021-01-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33728 * config/aarch64/aarch64-simd.md (aarch64_sqdml<SBINQOPS:as>l<mode>):
33730 (aarch64_sqdmlal<mode>): ... This...
33731 (aarch64_sqdmlsl<mode>): ... And this.
33732 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Split into...
33733 (aarch64_sqdmlal_lane<mode>): ... This...
33734 (aarch64_sqdmlsl_lane<mode>): ... And this.
33735 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Split into...
33736 (aarch64_sqdmlsl_laneq<mode>): ... This...
33737 (aarch64_sqdmlal_laneq<mode>): ... And this.
33738 (aarch64_sqdml<SBINQOPS:as>l_n<mode>): Split into...
33739 (aarch64_sqdmlsl_n<mode>): ... This...
33740 (aarch64_sqdmlal_n<mode>): ... And this.
33741 (aarch64_sqdml<SBINQOPS:as>l2<mode>_internal): Split into...
33742 (aarch64_sqdmlal2<mode>_internal): ... This...
33743 (aarch64_sqdmlsl2<mode>_internal): ... And this.
33745 2021-01-21 Christophe Lyon <christophe.lyon@linaro.org>
33747 * config/arm/arm_mve.h (__arm_vcmpneq_s8): Fix return type.
33749 2021-01-21 Andrea Corallo <andrea.corallo@arm.com>
33752 * doc/sourcebuild.texi (arm_thumb2_no_arm_v8_1_lob): Document.
33754 2021-01-21 liuhongt <hongtao.liu@intel.com>
33756 PR rtl-optimization/98694
33757 * regcprop.c (copy_value): If SRC had been assigned a mode
33758 narrower than the copy, we can't link DEST into the chain even
33759 they have same hard_regno_nregs(i.e. HImode/SImode in i386
33762 2021-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33764 * config/aarch64/aarch64-simd.md (aarch64_get_lane<mode>):
33765 Convert to define_insn_and_split. Split into simple move when moving
33768 2021-01-20 Segher Boessenkool <segher@kernel.crashing.org>
33770 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Change assert.
33771 Adjust comment. Simplify code.
33773 2021-01-20 Jakub Jelinek <jakub@redhat.com>
33776 * dwarf2out.c (reset_indirect_string): Also reset indirect strings
33777 with DW_FORM_line_strp form.
33778 (prune_unused_types_update_strings): Don't add into debug_str_hash
33779 indirect strings with DW_FORM_line_strp form.
33780 (adjust_name_comp_dir): New function.
33781 (dwarf2out_finish): Call it on CU DIEs after resetting
33782 debug_line_str_hash.
33784 2021-01-20 Vladimir N. Makarov <vmakarov@redhat.com>
33786 PR rtl-optimization/98722
33787 * lra-eliminations.c (eliminate_regs_in_insn): Check that target
33788 has no 3-op add insn to transform insns containing two pluses.
33790 2021-01-20 Richard Biener <rguenther@suse.de>
33792 * hwint.h (add_hwi): New function.
33793 (mul_hwi): Likewise.
33794 * tree-data-ref.c (initialize_matrix_A): Properly translate
33795 tree constants and avoid HOST_WIDE_INT_MIN.
33796 (lambda_matrix_row_add): Avoid undefined integer overflow
33797 and return true on such overflow.
33798 (lambda_matrix_right_hermite): Handle overflow from
33799 lambda_matrix_row_add gracefully. Simplify previous fix.
33800 (analyze_subscript_affine_affine): Likewise.
33802 2021-01-20 Eugene Rozenfeld <erozen@microsoft.com>
33804 PR tree-optimization/96674
33805 * match.pd: New patterns: x < y || y == XXX_MIN --> x <= y - 1
33806 x >= y && y != XXX_MIN --> x > y - 1
33808 2021-01-20 Richard Sandiford <richard.sandiford@arm.com>
33810 PR tree-optimization/98535
33811 * tree-vect-slp.c (duplicate_and_interleave): Use quick_grow_cleared.
33812 If the high and low permutes are the same, remove the high permutes
33813 from the working set and only continue with the low ones.
33815 2021-01-20 Jakub Jelinek <jakub@redhat.com>
33817 PR tree-optimization/98721
33818 * builtins.c (access_ref::inform_access): Don't assume
33819 SSA_NAME_IDENTIFIER must be non-NULL. Print messages about
33820 object whenever allocfn is NULL, rather than only when DECL_P
33821 is true. Use %qE instead of %qD for that. Formatting fixes.
33823 2021-01-20 Richard Biener <rguenther@suse.de>
33825 PR tree-optimization/98758
33826 * tree-data-ref.c (int_divides_p): Use lambda_int arguments.
33827 (lambda_matrix_right_hermite): Avoid undefinedness with
33828 signed integer abs and multiplication.
33829 (analyze_subscript_affine_affine): Use lambda_int.
33831 2021-01-20 David Malcolm <dmalcolm@redhat.com>
33834 * dwarf2out.c (output_line_info): Rename static variable
33835 "generation", moving it out of the function to...
33836 (output_line_info_generation): New.
33837 (init_sections_and_labels): Likewise, renaming the variable to...
33838 (init_sections_and_labels_generation): New.
33839 (dwarf2out_c_finalize): Reset the new variables.
33841 2021-01-19 Martin Sebor <msebor@redhat.com>
33843 PR middle-end/98664
33844 * tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
33845 all functions, even if they're not declared artificial or inline.
33846 * tree.c (tree_inlined_location): Use macro expansion location
33847 only if scope traversal fails to expose one.
33849 2021-01-19 Richard Sandiford <richard.sandiford@arm.com>
33851 PR rtl-optimization/92294
33852 * alias.c (compare_base_symbol_refs): Take an extra parameter
33853 and add the distance between two symbols to it. Enshrine in
33854 comments that -1 means "either 0 or 1, but we can't tell
33855 which at compile time".
33856 (memrefs_conflict_p): Update call accordingly.
33857 (rtx_equal_for_memref_p): Likewise. Take the distance between symbols
33860 2021-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33862 * config/aarch64/aarch64-simd-builtins.def (sqshl, uqshl,
33863 sqrshl, uqrshl, sqadd, uqadd, sqsub, uqsub, suqadd, usqadd, sqmovn,
33864 uqmovn, sqxtn2, uqxtn2, sqabs, sqneg, sqdmlal, sqdmlsl, sqdmlal_lane,
33865 sqdmlsl_lane, sqdmlal_laneq, sqdmlsl_laneq, sqdmlal_n, sqdmlsl_n,
33866 sqdmlal2, sqdmlsl2, sqdmlal2_lane, sqdmlsl2_lane, sqdmlal2_laneq,
33867 sqdmlsl2_laneq, sqdmlal2_n, sqdmlsl2_n, sqdmull, sqdmull_lane,
33868 sqdmull_laneq, sqdmull_n, sqdmull2, sqdmull2_lane, sqdmull2_laneq,
33869 sqdmull2_n, sqdmulh, sqrdmulh, sqdmulh_lane, sqdmulh_laneq,
33870 sqrdmulh_lane, sqrdmulh_laneq, sqshrun_n, sqrshrun_n, sqshrn_n,
33871 uqshrn_n, sqrshrn_n, uqrshrn_n, sqshlu_n, sqshl_n, uqshl_n, sqrdmlah,
33872 sqrdmlsh, sqrdmlah_lane, sqrdmlsh_lane, sqrdmlah_laneq, sqrdmlsh_laneq,
33873 sqmovun): Use NONE flags.
33875 2021-01-19 Richard Biener <rguenther@suse.de>
33878 * ipa-modref.c (analyze_stmt): Only record a summary for a
33881 2021-01-19 Richard Biener <rguenther@suse.de>
33883 PR middle-end/98638
33884 * tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
33886 2021-01-19 Daniel Hellstrom <daniel@gaisler.com>
33888 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
33889 built-in define __FIX_LEON3FT_TN0018.
33891 2021-01-19 Richard Biener <rguenther@suse.de>
33894 * tree-inline.c (tree_function_versioning): Set input_location
33895 to UNKNOWN_LOCATION throughout the function.
33897 2021-01-19 Tobias Burnus <tobias@codesourcery.com>
33900 * omp-low.c (lower_omp_target): Handle nonpointer is_device_ptr.
33902 2021-01-19 Martin Jambor <mjambor@suse.cz>
33905 * ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
33906 whether non-call exceptions allow removal of a statement.
33907 (isra_analyze_call): Pass the appropriate function to
33908 ssa_name_only_returned_p.
33910 2021-01-19 Geng Qi <gengqi@linux.alibaba.com>
33912 * config/riscv/arch-canonicalize (longext_sort): New function for
33913 sorting 'multi-letter'.
33914 * config/riscv/multilib-generator: Adjusting the loop of 'alt' in
33915 'alts'. The 'arch' may not be the first of 'alts'.
33916 (_expand_combination): Add underline for the 'ext' without '*'.
33917 This is because, a single-letter extension can always be treated well
33918 with a '_' prefix, but it cannot be separated out if it is appended
33921 2021-01-18 Vladimir N. Makarov <vmakarov@redhat.com>
33924 * ira.c (ira): Skip abnormal critical edge splitting.
33926 2021-01-18 Jakub Jelinek <jakub@redhat.com>
33928 PR tree-optimization/98727
33929 * tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
33930 second .MUL_OVERFLOW operand for signed multiplication with overflow
33931 checking if the second operand of multiplication is not constant.
33933 2021-01-18 David Edelsohn <dje.gcc@gmail.com>
33935 * doc/invoke.texi (-gdwarf): TPF defaults to version 2 and AIX
33936 defaults to version 4.
33938 2021-01-18 David Malcolm <dmalcolm@redhat.com>
33940 * attribs.h (fndecl_dealloc_argno): New decl.
33941 * builtins.c (call_dealloc_argno): Split out second half of
33943 (fndecl_dealloc_argno): New.
33944 * doc/extend.texi (Common Function Attributes): Document the
33945 interaction between the analyzer and the malloc attribute.
33946 * doc/invoke.texi (Static Analyzer Options): Likewise.
33948 2021-01-17 David Edelsohn <dje.gcc@gmail.com>
33950 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Override
33951 dwarf_version to 4.
33952 * config/rs6000/aix72.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
33954 2021-01-17 Martin Jambor <mjambor@suse.cz>
33957 * cgraph.c (clone_of_p): Check also former_clone_of as we climb
33960 2021-01-17 Mark Wielaard <mark@klomp.org>
33962 * common.opt (gdwarf-): Init(5).
33963 * doc/invoke.texi (-gdwarf): Document default to 5.
33965 2021-01-16 Kwok Cheung Yeung <kcy@codesourcery.com>
33967 * builtin-types.def
33968 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
33970 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
33971 ...this. Add extra argument.
33972 * gimplify.c (omp_default_clause): Ensure that event handle is
33973 firstprivate in a task region.
33974 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_DETACH.
33975 (gimplify_adjust_omp_clauses): Likewise.
33976 * omp-builtins.def (BUILT_IN_GOMP_TASK): Change function type to
33977 BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR.
33978 * omp-expand.c (expand_task_call): Add GOMP_TASK_FLAG_DETACH to flags
33979 if detach clause specified. Add detach argument when generating
33981 * omp-low.c (scan_sharing_clauses): Setup data environment for detach
33983 (finish_taskreg_scan): Move field for variable containing the event
33984 handle to the front of the struct.
33985 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_DETACH. Fix
33987 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
33988 OMP_CLAUSE_DETACH clause.
33989 (convert_local_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
33990 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_DETACH.
33991 * tree.c (omp_clause_num_ops): Add entry for OMP_CLAUSE_DETACH.
33993 (omp_clause_code_name): Add entry for OMP_CLAUSE_DETACH. Fix
33995 (walk_tree_1): Handle OMP_CLAUSE_DETACH.
33997 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
33999 * config/nios2/t-rtems: Reset all MULTILIB_* variables. Shorten
34000 multilib directory names. Use MULTILIB_REQUIRED instead of
34001 MULTILIB_EXCEPTIONS. Add -mhw-mul -mhw-mulx -mhw-div
34002 -mcustom-fpu-cfg=fph2 multilib.
34004 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
34006 * config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
34007 (nios2_init_fpu_configs): Provide register values for new
34008 -mcustom-fpu-cfg=fph2 option variant.
34009 * doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
34012 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
34014 * config/nios2/nios2.c (nios2_custom_check_insns): Remove
34015 custom instruction warnings.
34017 2021-01-16 Jakub Jelinek <jakub@redhat.com>
34019 PR tree-optimization/96669
34020 * match.pd ((CST << x) & 1 -> x == 0): New simplification.
34022 2021-01-16 Jakub Jelinek <jakub@redhat.com>
34024 PR tree-optimization/96271
34025 * passes.def: Pass false argument to first two pass_cd_dce
34026 instances and true to last instance. Add comment that
34027 last instance rewrites no longer addressed locals.
34028 * tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
34030 (pass_cd_dce::set_pass_param): New method.
34031 (pass_cd_dce::execute): Return TODO_update_address_taken from
34032 last cd_dce instance.
34034 2021-01-15 Carl Love <cel@us.ibm.com>
34036 * config/rs6000/altivec.h (vec_mulh, vec_div, vec_dive, vec_mod):
34038 * config/rs6000/altivec.md (VIlong): Move define to file vsx.md.
34039 * config/rs6000/rs6000-builtin.def (DIVES_V4SI, DIVES_V2DI,
34040 DIVEU_V4SI, DIVEU_V2DI, DIVS_V4SI, DIVS_V2DI, DIVU_V4SI,
34041 DIVU_V2DI, MODS_V2DI, MODS_V4SI, MODU_V2DI, MODU_V4SI,
34042 MULHS_V2DI, MULHS_V4SI, MULHU_V2DI, MULHU_V4SI, MULLD_V2DI):
34043 Add builtin define.
34044 (MULH, DIVE, MOD): Add new BU_P10_OVERLOAD_2 definitions.
34045 * config/rs6000/rs6000-call.c (VSX_BUILTIN_VEC_DIV,
34046 VSX_BUILTIN_VEC_DIVE, P10_BUILTIN_VEC_MOD, P10_BUILTIN_VEC_MULH):
34047 New overloaded definitions.
34048 (builtin_function_type) [P10V_BUILTIN_DIVEU_V4SI,
34049 P10V_BUILTIN_DIVEU_V2DI, P10V_BUILTIN_DIVU_V4SI,
34050 P10V_BUILTIN_DIVU_V2DI, P10V_BUILTIN_MODU_V2DI,
34051 P10V_BUILTIN_MODU_V4SI, P10V_BUILTIN_MULHU_V2DI,
34052 P10V_BUILTIN_MULHU_V4SI]: Add case
34053 statement for builtins.
34054 * config/rs6000/rs6000.md (bits): Add new attribute sizes V4SI, V2DI.
34055 * config/rs6000/vsx.md (VIlong): Moved from config/rs6000/altivec.md.
34056 (UNSPEC_VDIVES, UNSPEC_VDIVEU): New unspec definitions.
34057 (vsx_mul_v2di): Add if TARGET_POWER10 statement.
34058 (vsx_udiv_v2di): Add if TARGET_POWER10 statement.
34059 (dives_<mode>, diveu_<mode>, div<mode>3, uvdiv<mode>3,
34060 mods_<mode>, modu_<mode>, mulhs_<mode>, mulhu_<mode>, mulv2di3):
34061 Add define_insn, mode is VIlong.
34062 * doc/extend.texi (vec_mulh, vec_mul, vec_div, vec_dive, vec_mod):
34063 Add builtin descriptions.
34065 2021-01-15 Eric Botcazou <ebotcazou@adacore.com>
34067 * final.c (final_start_function_1): Reset force_source_line.
34069 2021-01-15 Jakub Jelinek <jakub@redhat.com>
34071 PR tree-optimization/96669
34072 * match.pd (((1 << A) & 1) != 0 -> A == 0,
34073 ((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
34074 possibly different power of two constants and to right shift too.
34076 2021-01-15 Jakub Jelinek <jakub@redhat.com>
34078 PR tree-optimization/96681
34079 * match.pd ((x < 0) ^ (y < 0) to (x ^ y) < 0): New simplification.
34080 ((x >= 0) ^ (y >= 0) to (x ^ y) < 0): Likewise.
34081 ((x < 0) ^ (y >= 0) to (x ^ y) >= 0): Likewise.
34082 ((x >= 0) ^ (y < 0) to (x ^ y) >= 0): Likewise.
34084 2021-01-15 Alexandre Oliva <oliva@adacore.com>
34086 * opts.c (gen_command_line_string): Exclude -dumpbase-ext.
34088 2021-01-15 Tamar Christina <tamar.christina@arm.com>
34090 * config/aarch64/aarch64-simd.md (cml<fcmac1><conj_op><mode>4,
34091 cmul<conj_op><mode>3): New.
34092 * config/aarch64/iterators.md (UNSPEC_FCMUL,
34093 UNSPEC_FCMUL180, UNSPEC_FCMLA_CONJ, UNSPEC_FCMLA180_CONJ,
34094 UNSPEC_CMLA_CONJ, UNSPEC_CMLA180_CONJ, UNSPEC_CMUL, UNSPEC_CMUL180,
34095 FCMLA_OP, FCMUL_OP, conj_op, rotsplit1, rotsplit2, fcmac1, sve_rot1,
34096 sve_rot2, SVE2_INT_CMLA_OP, SVE2_INT_CMUL_OP, SVE2_INT_CADD_OP): New.
34097 (rot): Add UNSPEC_FCMUL, UNSPEC_FCMUL180.
34098 (rot_op): Renamed to conj_op.
34099 * config/aarch64/aarch64-sve.md (cml<fcmac1><conj_op><mode>4,
34100 cmul<conj_op><mode>3): New.
34101 * config/aarch64/aarch64-sve2.md (cml<fcmac1><conj_op><mode>4,
34102 cmul<conj_op><mode>3): New.
34104 2021-01-15 David Malcolm <dmalcolm@redhat.com>
34108 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
34109 Escape the tempfile name when constructing the expected output.
34111 2021-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34113 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlsl_hi<mode>):
34115 (aarch64_<su>mlsl_hi<mode>): ... This.
34116 (aarch64_<su>mlsl_hi<mode>): Define.
34117 (*aarch64_<su>mlsl<mode): Rename to...
34118 (aarch64_<su>mlsl<mode): ... This.
34119 * config/aarch64/aarch64-simd-builtins.def (smlsl, umlsl,
34120 smlsl_hi, umlsl_hi): Define builtins.
34121 * config/aarch64/arm_neon.h (vmlsl_high_s8, vmlsl_high_s16,
34122 vmlsl_high_s32, vmlsl_high_u8, vmlsl_high_u16, vmlsl_high_u32,
34123 vmlsl_s8, vmlsl_s16, vmlsl_s32, vmlsl_u8,
34124 vmlsl_u16, vmlsl_u32): Reimplement with builtins.
34126 2021-01-15 Uroš Bizjak <ubizjak@gmail.com>
34128 * config/i386/i386-c.c (ix86_target_macros):
34129 Use cpp_define_formatted for __SIZEOF_FLOAT80__ definition.
34131 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
34134 * config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
34135 * Makefile.in (RTL_SSA_H): New variable.
34136 * config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
34137 * config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
34138 * config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
34140 * config/aarch64/aarch64-cc-fusion.cc: New file.
34142 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
34144 * recog.h (insn_change_watermark::~insn_change_watermark): Avoid
34145 calling cancel_changes for changes that no longer exist.
34147 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
34149 * rtl-ssa/functions.h (function_info::ref_defs): Rename to...
34150 (function_info::reg_defs): ...this.
34151 * rtl-ssa/member-fns.inl (function_info::ref_defs): Rename to...
34152 (function_info::reg_defs): ...this.
34154 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
34157 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
34159 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
34162 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
34165 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
34167 2021-01-15 Richard Biener <rguenther@suse.de>
34169 PR tree-optimization/96376
34170 * tree-vect-stmts.c (get_load_store_type): Disregard alignment
34171 for VMAT_INVARIANT.
34173 2021-01-15 Martin Liska <mliska@suse.cz>
34175 * doc/install.texi: Document that some tests need pytest module.
34176 * doc/sourcebuild.texi: Likewise.
34178 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
34181 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
34183 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
34185 * config/arm/mve.md (mve_vshrq_n_s<mode>_imm): New entry.
34186 (mve_vshrq_n_u<mode>_imm): Likewise.
34187 * config/arm/neon.md (vashr<mode>3, vlshr<mode>3): Move to ...
34188 * config/arm/vec-common.md: ... here.
34190 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
34192 * config/arm/mve.md (mve_vshlq_<supf><mode>): Move to
34194 * config/arm/neon.md (vashl<mode>3): Delete.
34195 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): New.
34196 (vasl<mode>3): New expander.
34198 2021-01-15 Richard Biener <rguenther@suse.de>
34200 PR tree-optimization/98685
34201 * tree-vect-slp.c (vect_schedule_slp_node): Refactor handling
34202 of vector extern defs.
34204 2021-01-14 David Malcolm <dmalcolm@redhat.com>
34207 * diagnostic.c (diagnostic_kind_text): Break out this array
34209 (diagnostic_build_prefix): ...here.
34210 (fancy_abort): Detect when diagnostic_initialize has not yet been
34211 called and fall back to a minimal implementation of printing the
34212 ICE, rather than segfaulting in internal_error.
34214 2021-01-14 David Malcolm <dmalcolm@redhat.com>
34216 * diagnostic.c (diagnostic_initialize): Eliminate
34217 parseable_fixits_p in favor of initializing extra_output_kind from
34218 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
34219 (convert_column_unit): New function, split out from...
34220 (diagnostic_converted_column): ...this.
34221 (print_parseable_fixits): Add "column_unit" and "tabstop" params.
34222 Use them to call convert_column_unit on the column values.
34223 (diagnostic_report_diagnostic): Eliminate conditional on
34224 parseable_fixits_p in favor of a switch statement on
34225 extra_output_kind, passing the appropriate values to the new
34226 params of print_parseable_fixits.
34227 (selftest::test_print_parseable_fixits_none): Update for new
34228 params of print_parseable_fixits.
34229 (selftest::test_print_parseable_fixits_insert): Likewise.
34230 (selftest::test_print_parseable_fixits_remove): Likewise.
34231 (selftest::test_print_parseable_fixits_replace): Likewise.
34232 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
34234 (selftest::diagnostic_c_tests): Call it.
34235 * diagnostic.h (enum diagnostics_extra_output_kind): New.
34236 (diagnostic_context::parseable_fixits_p): Delete field in favor
34238 (diagnostic_context::extra_output_kind): ...this new field.
34239 * doc/invoke.texi (Environment Variables): Add
34240 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
34241 * opts.c (common_handle_option): Update handling of
34242 OPT_fdiagnostics_parseable_fixits for change to diagnostic_context
34245 2021-01-14 Tamar Christina <tamar.christina@arm.com>
34247 * tree-vect-slp-patterns.c (class complex_operations_pattern,
34248 complex_operations_pattern::matches,
34249 complex_operations_pattern::recognize,
34250 complex_operations_pattern::build): New.
34251 (slp_patterns): Use it.
34253 2021-01-14 Tamar Christina <tamar.christina@arm.com>
34255 * internal-fn.def (COMPLEX_FMS, COMPLEX_FMS_CONJ): New.
34256 * optabs.def (cmls_optab, cmls_conj_optab): New.
34257 * doc/md.texi: Document them.
34258 * tree-vect-slp-patterns.c (class complex_fms_pattern,
34259 complex_fms_pattern::matches, complex_fms_pattern::recognize,
34260 complex_fms_pattern::build): New.
34262 2021-01-14 Tamar Christina <tamar.christina@arm.com>
34264 * internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
34265 * optabs.def (cmla_optab, cmla_conj_optab): New.
34266 * doc/md.texi: Document them.
34267 * tree-vect-slp-patterns.c (vect_match_call_p,
34268 class complex_fma_pattern, vect_slp_reset_pattern,
34269 complex_fma_pattern::matches, complex_fma_pattern::recognize,
34270 complex_fma_pattern::build): New.
34272 2021-01-14 Tamar Christina <tamar.christina@arm.com>
34274 * internal-fn.def (COMPLEX_MUL, COMPLEX_MUL_CONJ): New.
34275 * optabs.def (cmul_optab, cmul_conj_optab): New.
34276 * doc/md.texi: Document them.
34277 * tree-vect-slp-patterns.c (vect_match_call_complex_mla,
34278 vect_normalize_conj_loc, is_eq_or_top, vect_validate_multiplication,
34279 vect_build_combine_node, class complex_mul_pattern,
34280 complex_mul_pattern::matches, complex_mul_pattern::recognize,
34281 complex_mul_pattern::build): New.
34283 2021-01-14 Tamar Christina <tamar.christina@arm.com>
34285 * tree-vect-slp.c (optimize_load_redistribution_1): New.
34286 (optimize_load_redistribution, vect_is_slp_load_node): New.
34287 (vect_match_slp_patterns): Use it.
34289 2021-01-14 Tamar Christina <tamar.christina@arm.com>
34291 * tree-vect-slp-patterns.c (complex_add_pattern::build):
34294 2021-01-14 Thomas Schwinge <thomas@codesourcery.com>
34296 * config/gcn/mkoffload.c (main): Create an offload image only in
34297 64-bit configurations.
34299 2021-01-14 H.J. Lu <hjl.tools@gmail.com>
34302 * config/i386/i386-options.c (ix86_option_override_internal):
34303 Issue an error for -fcf-protection with CF_BRANCH when compiling
34304 for 32-bit non-TARGET_CMOV targets.
34306 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
34309 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p):
34310 Remove declaration and initialization of shadow variable "ret".
34311 (ix86_option_override_internal): Remove delcaration of
34312 shadow variable "i". Redeclare shadowed variable to unsigned.
34313 * common/config/i386/i386-common.c (pta_size): Redeclare to unsigned.
34314 * config/i386/i386-builtins.c (get_builtin_code_for_version):
34315 Update for redeclaration.
34316 * config/i386/i386.h (pta_size): Ditto.
34318 2021-01-14 Richard Biener <rguenther@suse.de>
34320 PR tree-optimization/98674
34321 * tree-data-ref.c (base_supports_access_fn_components_p): New.
34322 (initialize_data_dependence_relation): For two bases without
34323 possible access fns resort to type size equality when determining
34324 shape compatibility.
34326 2021-01-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
34329 * config/arm/arm_neon.h: Replace calls to __builtin_vcge* by
34330 <=, >= operators in vcle and vcge intrinsics respectively.
34331 * config/arm/arm_neon_builtins.def: Remove entry for
34334 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
34337 * config/i386/i386-options.c (ix86_function_specific_save):
34338 Remove redundant assignment to opts->x_ix86_branch_cost.
34339 * config/i386/i386.c (ix86_prefetch_sse):
34340 Rename from x86_prefetch_sse. Update all uses.
34341 * config/i386/i386.h: Update for rename.
34342 * config/i386/i386-options.h: Ditto.
34344 2021-01-14 Jakub Jelinek <jakub@redhat.com>
34347 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3,
34348 *sse4_1_zero_extendv4hiv4si2_3, *sse4_1_zero_extendv2siv2di2_3):
34349 Use Bm instead of m for non-avx. Add isa attribute.
34351 2021-01-14 Jakub Jelinek <jakub@redhat.com>
34353 PR tree-optimization/96688
34354 * match.pd (~(X >> Y) -> ~X >> Y): New simplification if
34355 ~X can be simplified.
34357 2021-01-14 Richard Sandiford <richard.sandiford@arm.com>
34359 * tree-vect-stmts.c (vect_model_load_cost): Account for unused
34360 IFN_LOAD_LANES results.
34362 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34364 * config/aarch64/aarch64-simd.md (aarch64_<su>xtl<mode>):
34366 (aarch64_xtn<mode>): Likewise.
34367 * config/aarch64/aarch64-simd-builtins.def (sxtl, uxtl, xtn):
34370 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
34372 (vmovl_s16): Likewise.
34373 (vmovl_s32): Likewise.
34374 (vmovl_u8): Likewise.
34375 (vmovl_u16): Likewise.
34376 (vmovl_u32): Likewise.
34377 (vmovn_s16): Likewise.
34378 (vmovn_s32): Likewise.
34379 (vmovn_s64): Likewise.
34380 (vmovn_u16): Likewise.
34381 (vmovn_u32): Likewise.
34382 (vmovn_u64): Likewise.
34384 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34386 * config/aarch64/aarch64-simd.md (aarch64_<su>qxtn2<mode>_le):
34388 (aarch64_<su>qxtn2<mode>_be): Likewise.
34389 (aarch64_<su>qxtn2<mode>): Likewise.
34390 * config/aarch64/aarch64-simd-builtins.def (sqxtn2, uqxtn2):
34392 * config/aarch64/iterators.md (SAT_TRUNC): Define code_iterator.
34393 (su): Handle ss_truncate and us_truncate.
34394 * config/aarch64/arm_neon.h (vqmovn_high_s16): Reimplement using
34396 (vqmovn_high_s32): Likewise.
34397 (vqmovn_high_s64): Likewise.
34398 (vqmovn_high_u16): Likewise.
34399 (vqmovn_high_u32): Likewise.
34400 (vqmovn_high_u64): Likewise.
34402 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34404 * config/aarch64/aarch64-simd.md (aarch64_xtn2<mode>_le):
34406 (aarch64_xtn2<mode>_be): Likewise.
34407 (aarch64_xtn2<mode>): Likewise.
34408 * config/aarch64/aarch64-simd-builtins.def (xtn2): Define
34410 * config/aarch64/arm_neon.h (vmovn_high_s16): Reimplement using
34412 (vmovn_high_s32): Likewise.
34413 (vmovn_high_s64): Likewise.
34414 (vmovn_high_u16): Likewise.
34415 (vmovn_high_u32): Likewise.
34416 (vmovn_high_u64): Likewise.
34418 2021-01-13 Stafford Horne <shorne@gmail.com>
34420 * config/or1k/or1k.h (ASM_PREFERRED_EH_DATA_FORMAT): New macro.
34422 2021-01-13 Stafford Horne <shorne@gmail.com>
34424 * config/or1k/linux.h (TARGET_ASM_FILE_END): Define macro.
34426 2021-01-13 Stafford Horne <shorne@gmail.com>
34428 * config/or1k/or1k.h (TARGET_CPU_CPP_BUILTINS): Add builtin
34429 define for __or1k_hard_float__.
34431 2021-01-13 Stafford Horne <shorne@gmail.com>
34433 * config/or1k/or1k.h (NO_PROFILE_COUNTERS): Define as 1.
34434 (PROFILE_HOOK): Define to call _mcount.
34435 (FUNCTION_PROFILER): Change from abort to no-op.
34437 2021-01-13 Jakub Jelinek <jakub@redhat.com>
34439 PR tree-optimization/96691
34440 * match.pd ((~X | C) ^ D -> (X | C) ^ (~D ^ C),
34441 (~X & C) ^ D -> (X & C) ^ (D ^ C)): New simplifications if
34442 (~D ^ C) or (D ^ C) can be simplified.
34444 2021-01-13 Richard Biener <rguenther@suse.de>
34446 PR tree-optimization/92645
34447 * match.pd (BIT_FIELD_REF to conversion): Delay canonicalization
34448 until after vector lowering.
34450 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
34452 * config/aarch64/aarch64-sve.md (fnma<mode>4): Extend from SVE_FULL_I
34454 (@aarch64_pred_fnma<mode>, cond_fnma<mode>, *cond_fnma<mode>_2)
34455 (*cond_fnma<mode>_4, *cond_fnma<mode>_any): Likewise.
34457 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
34459 * config/aarch64/aarch64-sve.md (fma<mode>4): Extend from SVE_FULL_I
34461 (@aarch64_pred_fma<mode>, cond_fma<mode>, *cond_fma<mode>_2)
34462 (*cond_fma<mode>_4, *cond_fma<mode>_any): Likewise.
34464 2021-01-13 Richard Biener <rguenther@suse.de>
34466 PR tree-optimization/92645
34467 * tree-vect-slp.c (vect_build_slp_tree_1): Relax supported
34468 BIT_FIELD_REF argument.
34469 (vect_build_slp_tree_2): Record the desired vector type
34470 on the external vector def.
34471 (vectorizable_slp_permutation): Handle required punning
34472 of existing vector defs.
34474 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
34476 * rtl-ssa/accesses.h (def_lookup): Fix order of comparison results.
34478 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
34480 * config/sh/sh.md (movsf_ie): Remove operands[2] test.
34482 2021-01-13 Samuel Thibault <samuel.thibault@ens-lyon.org>
34484 * config.gcc [$target == *-*-gnu*]: Enable
34485 'default_gnu_indirect_function'.
34487 2021-01-13 Jakub Jelinek <jakub@redhat.com>
34490 * optabs.c (expand_vec_perm_const): Don't force v0 and v1 into
34491 registers before calling targetm.vectorize.vec_perm_const, only after
34493 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const): Handle
34494 two argument permutation when one operand is zero vector and only
34495 after that force operands into registers.
34496 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_1): New
34497 define_insn_and_split pattern.
34498 (*avx512bw_zero_extendv32qiv32hi2_1): Likewise.
34499 (*avx512f_zero_extendv16hiv16si2_1): Likewise.
34500 (*avx2_zero_extendv8hiv8si2_1): Likewise.
34501 (*avx512f_zero_extendv8siv8di2_1): Likewise.
34502 (*avx2_zero_extendv4siv4di2_1): Likewise.
34503 * config/mips/mips.c (mips_vectorize_vec_perm_const): Force operands
34505 * config/arm/arm.c (arm_vectorize_vec_perm_const): Likewise.
34506 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Likewise.
34507 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const): Likewise.
34508 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const): Likewise.
34509 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const): Likewise.
34510 * config/gcn/gcn.c (gcn_vectorize_vec_perm_const): Likewise. Use std::swap.
34512 2021-01-13 Martin Liska <mliska@suse.cz>
34514 PR tree-optimization/98455
34515 * gimple-if-to-switch.cc (condition_info::record_phi_mapping):
34516 Record also virtual PHIs.
34517 (pass_if_to_switch::execute): Return TODO_cleanup_cfg only
34520 2021-01-13 Jonathan Wakely <jwakely@redhat.com>
34522 * doc/invoke.texi (C++ Modules): Fix typos.
34524 2021-01-13 Richard Biener <rguenther@suse.de>
34526 PR tree-optimization/98640
34527 * tree-ssa-sccvn.c (visit_nary_op): Do not try to
34528 handle plus or minus from a truncated operand to be
34531 2021-01-13 Jakub Jelinek <jakub@redhat.com>
34534 * config/i386/i386.md (*btr<mode>_1, *btr<mode>_2): New
34535 define_insn_and_split patterns.
34536 (splitter after *btr<mode>_2): New splitter.
34538 2021-01-13 Martin Liska <mliska@suse.cz>
34541 * cgraphunit.c (analyze_functions): Remove dead code.
34543 2021-01-13 Qian Jianhua <qianjh@cn.fujitsu.com>
34545 * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
34546 * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
34547 (a64fx_regmove_cost, a64fx_vector_cost): New.
34548 (a64fx_tunings): Use the new added cost tables.
34550 2021-01-13 Jakub Jelinek <jakub@redhat.com>
34553 * config/i386/predicates.md (pmovzx_parallel): New predicate.
34554 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): New
34555 define_insn_and_split pattern.
34556 (*sse4_1_zero_extendv4hiv4si2_3): Likewise.
34557 (*sse4_1_zero_extendv2siv2di2_3): Likewise.
34559 2021-01-13 Julian Brown <julian@codesourcery.com>
34561 * config/gcn/gcn.c (gcn_conditional_register_usage): Remove dead code
34562 to fix v0 register.
34564 2021-01-13 Julian Brown <julian@codesourcery.com>
34566 * config/gcn/gcn.c (gcn_md_reorg): Fix case where EXEC reg is live
34569 2021-01-13 Julian Brown <julian@codesourcery.com>
34571 * config/gcn/gcn-valu.md (recip<mode>2<exec>, recip<mode>2): Use unspec
34572 for reciprocal-approximation instructions.
34573 (div<mode>3): Use fused multiply-accumulate operations for reciprocal
34574 refinement and division result.
34575 * config/gcn/gcn.md (UNSPEC_RCP): New unspec constant.
34577 2021-01-13 Julian Brown <julian@codesourcery.com>
34579 * config/gcn/gcn-valu.md (subdf): Rename to...
34582 2021-01-12 Martin Liska <mliska@suse.cz>
34584 * gcov.c (source_info::debug): Fix printf format for 32-bit hosts.
34586 2021-01-12 Andrea Corallo <andrea.corallo@arm.com>
34588 * function-abi.h: Fix typo.
34590 2021-01-12 Christophe Lyon <christophe.lyon@linaro.org>
34594 * config/arm/arm.h (ARM_HAVE_NEON_V8QI_LDST): New macro.
34595 (ARM_HAVE_NEON_V16QI_LDST, ARM_HAVE_NEON_V4HI_LDST): Likewise.
34596 (ARM_HAVE_NEON_V8HI_LDST, ARM_HAVE_NEON_V2SI_LDST): Likewise.
34597 (ARM_HAVE_NEON_V4SI_LDST, ARM_HAVE_NEON_V4HF_LDST): Likewise.
34598 (ARM_HAVE_NEON_V8HF_LDST, ARM_HAVE_NEON_V4BF_LDST): Likewise.
34599 (ARM_HAVE_NEON_V8BF_LDST, ARM_HAVE_NEON_V2SF_LDST): Likewise.
34600 (ARM_HAVE_NEON_V4SF_LDST, ARM_HAVE_NEON_DI_LDST): Likewise.
34601 (ARM_HAVE_NEON_V2DI_LDST): Likewise.
34602 (ARM_HAVE_V8QI_LDST, ARM_HAVE_V16QI_LDST): Likewise.
34603 (ARM_HAVE_V4HI_LDST, ARM_HAVE_V8HI_LDST): Likewise.
34604 (ARM_HAVE_V2SI_LDST, ARM_HAVE_V4SI_LDST, ARM_HAVE_V4HF_LDST): Likewise.
34605 (ARM_HAVE_V8HF_LDST, ARM_HAVE_V4BF_LDST, ARM_HAVE_V8BF_LDST): Likewise.
34606 (ARM_HAVE_V2SF_LDST, ARM_HAVE_V4SF_LDST, ARM_HAVE_DI_LDST): Likewise.
34607 (ARM_HAVE_V2DI_LDST): Likewise.
34608 * config/arm/mve.md (*movmisalign<mode>_mve_store): New pattern.
34609 (*movmisalign<mode>_mve_load): New pattern.
34610 * config/arm/neon.md (movmisalign<mode>): Move to ...
34611 * config/arm/vec-common.md: ... here.
34613 2021-01-12 Vladimir N. Makarov <vmakarov@redhat.com>
34616 * lra-eliminations.c (eliminate_regs_in_insn): Add transformation
34617 of pattern 'plus (plus (hard reg, const), pseudo)'.
34619 2021-01-12 Richard Biener <rguenther@suse.de>
34621 PR tree-optimization/98550
34622 * tree-vect-slp.c (vect_record_max_nunits): Check whether
34623 the group size is a multiple of the vector element count.
34624 (vect_build_slp_tree_1): When we need to fail because
34625 the vector type choosen causes unrolling do so lazily
34626 without affecting matches only at the end to guide group splitting.
34628 2021-01-12 Martin Liska <mliska@suse.cz>
34631 * optc-save-gen.awk: Compare also n_target_save vars with
34634 2021-01-12 Martin Liska <mliska@suse.cz>
34636 * gcov.c (source_info::debug): New.
34637 (print_usage): Add --debug (-D) option.
34638 (process_args): Likewise.
34639 (generate_results): Call src->debug after
34640 accumulate_line_counts.
34641 (read_graph_file): Properly assign id for EXIT_BLOCK.
34642 * profile.c (branch_prob): Dump function body before it is
34645 2021-01-12 Jakub Jelinek <jakub@redhat.com>
34647 PR tree-optimization/98629
34648 * tree-ssa-math-opts.c (arith_overflow_check_p): Don't update use_stmt
34649 unless returning non-zero.
34651 2021-01-12 Jakub Jelinek <jakub@redhat.com>
34653 PR tree-optimization/95731
34654 * tree-ssa-reassoc.c (optimize_range_tests_cmp_bitwise): Also optimize
34655 x < 0 && y < 0 && z < 0 into (x | y | z) < 0 for signed x, y, z.
34656 (optimize_range_tests): Call optimize_range_tests_cmp_bitwise
34657 only after optimize_range_tests_var_bound.
34659 2021-01-12 Jakub Jelinek <jakub@redhat.com>
34661 * configure.ac: Ensure c/Make-lang.in comes first in @all_lang_makefrags@.
34662 * configure: Regenerated.
34664 2021-01-12 liuhongt <hongtao.liu@intel.com>
34667 * config/i386/i386-builtins.h (BUILTIN_DESC_SWAP_OPERANDS):
34669 * config/i386/i386-expand.c (ix86_expand_sse_comi): Delete
34672 2021-01-12 Alexandre Oliva <oliva@adacore.com>
34674 * ssa-iterators.h (end_imm_use_stmt_traverse): Forward
34676 (auto_end_imm_use_stmt_traverse): New struct.
34677 (FOR_EACH_IMM_USE_STMT): Use it.
34678 (BREAK_FROM_IMM_USE_STMT, RETURN_FROM_IMM_USE_STMT): Remove,
34680 * gimple-ssa-strength-reduction.c: ... here, ...
34681 * graphite-scop-detection.c: ... here, ...
34682 * ipa-modref.c, ipa-pure-const.c, ipa-sra.c: ... here, ...
34683 * tree-predcom.c, tree-ssa-ccp.c: ... here, ...
34684 * tree-ssa-dce.c, tree-ssa-dse.c: ... here, ...
34685 * tree-ssa-loop-ivopts.c, tree-ssa-math-opts.c: ... here, ...
34686 * tree-ssa-phiprop.c, tree-ssa.c: ... here, ...
34687 * tree-vect-slp.c: ... and here, ...
34688 * doc/tree-ssa.texi: ... and the example here.
34690 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34692 * config/aarch64/aarch64-sve.md (sdiv_pow2<mode>3): Extend from
34693 SVE_FULL_I to SVE_I. Generate an UNSPEC_PRED_X.
34694 (*sdiv_pow2<mode>3): New pattern.
34695 (@cond_<sve_int_op><mode>): Extend from SVE_FULL_I to SVE_I.
34696 Wrap the ASRD in an UNSPEC_PRED_X.
34697 (*cond_<sve_int_op><mode>_2): Likewise. Replace the UNSPEC_PRED_X
34698 predicate with a constant PTRUE, if it isn't already.
34699 (*cond_<sve_int_op><mode>_z): Replace with...
34700 (*cond_<sve_int_op><mode>_any): ...this new pattern.
34702 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34704 * config/aarch64/aarch64-sve.md (*cond_bic<mode>_2): Extend from
34705 SVE_FULL_I to SVE_I.
34706 (*cond_bic<mode>_any): Likewise.
34708 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34710 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart)
34711 (@aarch64_pred_<MUL_HIGHPART:optab><mode>): Extend from SVE_FULL_I
34714 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34716 * config/aarch64/aarch64-sve.md (<su>abd<mode>_3): Extend from
34717 SVE_FULL_I to SVE_I.
34718 (*aarch64_cond_<su>abd<mode>_2): Likewise.
34719 (*aarch64_cond_<su>abd<mode>_any): Likewise.
34720 (@aarch64_pred_<su>abd<mode>): Likewise. Use UNSPEC_PRED_X
34721 for the max and min but not for the minus.
34722 (*aarch64_cond_<su>abd<mode>_3): New pattern.
34724 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34726 * config/aarch64/iterators.md (SVE_24I): New iterator.
34727 * config/aarch64/aarch64-sve.md (*aarch64_adr<mode>_shift): Extend from
34728 SVE_FULL_SDI to SVE_24I. Use containers rather than elements.
34730 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34732 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_BINARY:optab><mode>)
34733 (*cond_<SVE_INT_BINARY:optab><mode>_2): Extend from SVE_FULL_I
34735 (*cond_<SVE_INT_BINARY:optab><mode>_3): Likewise.
34736 (*cond_<SVE_INT_BINARY:optab><mode>_any): Likewise.
34737 (*cond_<SVE_INT_BINARY:optab><mode>_2_const): Likewise.
34738 (*cond_<SVE_INT_BINARY:optab><mode>_any_const): Likewise.
34740 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34742 * config/aarch64/aarch64-sve.md (<SVE_INT_BINARY_IMM:optab><mode>3)
34743 (@aarch64_pred_<SVE_INT_BINARY_IMM:optab><mode>)
34744 (*post_ra_<SVE_INT_BINARY_IMM:optab><mode>3): Extend from SVE_FULL_I
34747 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
34749 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3)
34750 (v<ASHIFT:optab><mode>3, @aarch64_pred_<optab><mode>)
34751 (*post_ra_v<ASHIFT:optab><mode>3): Extend from SVE_FULL_I to SVE_I.
34753 2021-01-11 Martin Liska <mliska@suse.cz>
34756 * symtab-clones.h (clone_info::release): Release
34757 symtab::m_clones with ggc_delete as it's a GGC memory.
34759 2021-01-11 Matthias Klose <doko@ubuntu.com>
34761 * Makefile.in (LINK_PROGRESS): Show the link target.
34763 2021-01-11 Richard Biener <rguenther@suse.de>
34765 PR tree-optimization/91403
34766 * tree-vect-data-refs.c (vect_analyze_group_access_1): Cap
34767 single-element interleaving group size at 4096 elements.
34769 2021-01-11 Richard Biener <rguenther@suse.de>
34771 PR tree-optimization/98526
34772 * tree-vect-loop.c (vect_model_reduction_cost): Remove costing
34773 of the actual reduction op for the regular case.
34774 (vectorizable_reduction): Cost the stmts
34775 vect_transform_reduction produces here.
34777 2021-01-11 Andreas Krebbel <krebbel@linux.ibm.com>
34779 * tree-ssa-forwprop.c (simplify_vector_constructor): For
34780 big-endian, use UNPACK[_FLOAT]_HI.
34782 2021-01-11 Tamar Christina <tamar.christina@arm.com>
34784 * tree-vect-slp-patterns.c (class complex_pattern,
34785 class complex_add_pattern): Add parameters to matches.
34786 (complex_add_pattern::build): Free memory.
34787 (complex_add_pattern::matches): Move validation end of match.
34788 (complex_add_pattern::recognize): Likewise.
34790 2021-01-11 Tamar Christina <tamar.christina@arm.com>
34792 * tree-vect-slp-patterns.c (linear_loads_p): Fix externals.
34794 2021-01-11 Tamar Christina <tamar.christina@arm.com>
34796 * tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
34798 2021-01-11 Jakub Jelinek <jakub@redhat.com>
34800 PR tree-optimization/95867
34801 * tree-ssa-math-opts.h: New header.
34802 * tree-ssa-math-opts.c: Include tree-ssa-math-opts.h.
34803 (powi_as_mults): No longer static. Use build_one_cst instead of
34804 build_real. Formatting fix.
34805 * tree-ssa-reassoc.c: Include tree-ssa-math-opts.h.
34806 (attempt_builtin_powi): Handle multiplication reassociation without
34807 powi_fndecl using powi_as_mults.
34808 (reassociate_bb): For integral types don't require
34809 -funsafe-math-optimizations to call attempt_builtin_powi.
34811 2021-01-11 Jakub Jelinek <jakub@redhat.com>
34813 PR tree-optimization/95852
34814 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): Change
34815 mul_stmts parameter type to vec<gimple *> &. Before cond_stmt
34816 allow in the bb any of the stmts in that vector, div_stmt and
34817 up to 3 cast stmts.
34818 (arith_cast_equal_p): New function.
34819 (arith_overflow_check_p): Add cast_stmt argument, handle signed
34820 multiply overflow checks.
34821 (match_arith_overflow): Adjust caller. Handle signed multiply
34824 2021-01-11 Jakub Jelinek <jakub@redhat.com>
34826 PR tree-optimization/95852
34827 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): New function.
34828 (uaddsub_overflow_check_p): Renamed to ...
34829 (arith_overflow_check_p): ... this. Handle also multiplication
34830 with overflow check.
34831 (match_uaddsub_overflow): Renamed to ...
34832 (match_arith_overflow): ... this. Add cfg_changed argument. Handle
34833 also multiplication with overflow check. Adjust function comment.
34834 (math_opts_dom_walker::after_dom_children): Adjust callers. Call
34835 match_arith_overflow also for MULT_EXPR.
34837 2021-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34839 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
34840 __builtin_convertvector.
34841 (vmovl_s16): Likewise.
34842 (vmovl_s32): Likewise.
34843 (vmovl_u8): Likewise.
34844 (vmovl_u16): Likewise.
34845 (vmovl_u32): Likewise.
34846 (vmovn_s16): Likewise.
34847 (vmovn_s32): Likewise.
34848 (vmovn_s64): Likewise.
34849 (vmovn_u16): Likewise.
34850 (vmovn_u32): Likewise.
34851 (vmovn_u64): Likewise.
34853 2021-01-11 Martin Liska <mliska@suse.cz>
34855 * gimple-if-to-switch.cc (struct condition_info): Use auto_var.
34856 (if_chain::is_beneficial): Delete clusters
34857 (find_conditions): Make second argument of conditions_in_bbs a
34858 pointer so that we control over it's lifetime.
34859 (pass_if_to_switch::execute): Delete them.
34861 2021-01-11 Kewen Lin <linkw@linux.ibm.com>
34863 * ira.c (move_unallocated_pseudos): Check other_reg and skip if
34866 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
34868 * config/vax/vax.md (cc): Remove mode attribute.
34869 (subst_<cc>, subst_f<cc>): Rename to...
34870 (subst_<mode>, subst_f<VAXccnz:mode>): ... these respectively.
34871 (*cbranch<VAXint:mode>4_<VAXcc:mode>): Update for `cc' removal.
34872 (*cbranch<VAXfp:mode>4_<VAXccnz:mode>): Likewise.
34873 (*branch_<mode>, *branch_<mode>_reversed): Likewise.
34875 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
34877 * config/vax/vax.md (subst_f<cc>): Add mode to operands and
34878 `const_double_zero'.
34880 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
34882 * config/pdp11/pdp11.md (PDPfp): New mode iterator.
34883 (fcc_cc, fcc_ccnz): Use it. Add mode to `const_double_zero' and
34886 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
34888 * genemit.c (gen_exp) <CONST_DOUBLE>: Handle `const_double_zero'
34890 * read-rtl.c (rtx_reader::read_rtx_code): Handle machine mode
34891 with `const_double_zero'.
34892 * doc/rtl.texi (Constant Expression Types): Document it.
34894 2021-01-09 Jakub Jelinek <jakub@redhat.com>
34897 * tree-cfg.c (verify_gimple_assign_binary): Allow lhs of
34898 POINTER_DIFF_EXPR to be any integral type.
34900 2021-01-09 Jakub Jelinek <jakub@redhat.com>
34902 PR rtl-optimization/98603
34903 * function.c (instantiate_virtual_regs_in_insn): For asm goto
34904 with impossible constraints, drop all SETs, CLOBBERs, drop PARALLEL
34905 if any, set ASM_OPERANDS mode to VOIDmode and change
34906 ASM_OPERANDS_OUTPUT_CONSTRAINT and ASM_OPERANDS_OUTPUT_IDX.
34908 2021-01-09 Alexandre Oliva <oliva@gnu.org>
34911 * final.c (notice_source_line): Narrow down the condition to
34912 skip a line-0 marker.
34914 2021-01-08 Sergei Trofimovich <siarheit@google.com>
34916 * ipa-modref.c (merge_call_side_effects): Fix
34917 linebreak split by reordering two print calls.
34919 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
34921 * config/s390/vector.md (*tf_to_fprx2_0): Rename from
34922 "*mov_tf_to_fprx2_0" for consistency, fix constraint.
34923 (*tf_to_fprx2_1): Rename from "*mov_tf_to_fprx2_1" for
34924 consistency, fix constraint.
34926 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
34928 * config/s390/s390-c.c (s390_def_or_undef_macro): Accept
34929 callables instead of mask values.
34930 (struct target_flag_set_p): New predicate.
34931 (s390_cpu_cpp_builtins_internal): Define or undefine
34932 __LONG_DOUBLE_VX__ macro.
34934 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
34937 * config/i386/i386.c (x86_function_profiler): Use R10 and R11
34938 to call mcount in large model with PIC for NO_PROFILE_COUNTERS
34941 2021-01-08 Richard Biener <rguenther@suse.de>
34943 * tree-ssa-sccvn.c (pass_fre::execute): Reset the SCEV hash table.
34945 2021-01-08 Richard Biener <rguenther@suse.de>
34947 * tree-vect-slp.c (scalar_stmts_to_slp_tree_map_t): Fix.
34948 (vect_build_slp_tree): On cache hit release the matched
34949 scalar stmts vector.
34950 * tree-vect-stmts.c (vectorizable_store): Properly free
34951 vec_oprnds before possibly gathering them again.
34953 2021-01-08 Richard Biener <rguenther@suse.de>
34955 PR tree-optimization/98544
34956 * tree-vect-slp.c (vect_optimize_slp): Always materialize
34957 permutes at a permute node.
34959 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
34962 * config/i386/i386.c (x86_function_profiler): Use R10 to call
34963 mcount in large model. Sorry for large model with PIC.
34965 2021-01-08 Jakub Jelinek <jakub@redhat.com>
34968 * config/i386/i386.opt (ix86_cmodel, ix86_incoming_stack_boundary_arg,
34969 ix86_pmode, ix86_preferred_stack_boundary_arg, ix86_regparm,
34970 ix86_veclibabi_type): Remove x_ prefix, use TargetVariable instead of
34971 TargetSave and initialize for variables with enum types.
34972 (mfentry, mstack-protector-guard-reg=, mstack-protector-guard-offset=,
34973 mstack-protector-guard-symbol=): Add Save.
34974 * config/i386/i386-options.c (ix86_function_specific_save,
34975 ix86_function_specific_restore): Don't save or restore x_ix86_cmodel,
34976 x_ix86_incoming_stack_boundary_arg, x_ix86_pmode,
34977 x_ix86_preferred_stack_boundary_arg, x_ix86_regparm,
34978 x_ix86_veclibabi_type.
34980 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
34982 * config/aarch64/aarch64-sve.md (*cnot<mode>): Extend from
34983 SVE_FULL_I to SVE_I.
34984 (*cond_cnot<mode>_2, *cond_cnot<mode>_any): Likewise.
34986 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
34988 * config/aarch64/aarch64-sve.md (*cond_uxt<mode>_2): Extend from
34989 SVE_FULL_I to SVE_I.
34990 (*cond_uxt<mode>_any): Likewise.
34992 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34994 * config/aarch64/iterators.md (Vwhalf): New iterator.
34995 * config/aarch64/aarch64-simd.md (aarch64_<sur>adalp<mode>_3):
34997 (aarch64_<sur>adalp<mode>): ... This. Make more
34999 (<sur>sadv16qi): Adjust callsite of the above.
35000 * config/aarch64/aarch64-simd-builtins.def (sadalp, uadalp): New
35002 * config/aarch64/arm_neon.h (vpadal_s8): Reimplement using
35004 (vpadal_s16): Likewise.
35005 (vpadal_u8): Likewise.
35006 (vpadal_u16): Likewise.
35007 (vpadalq_s8): Likewise.
35008 (vpadalq_s16): Likewise.
35009 (vpadalq_s32): Likewise.
35010 (vpadalq_u8): Likewise.
35011 (vpadalq_u16): Likewise.
35012 (vpadalq_u32): Likewise.
35014 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
35016 * config/aarch64/aarch64-simd.md (aarch64_<su>abd<mode>_3):
35018 (aarch64_<su>abd<mode>): ... This.
35019 (<sur>sadv16qi): Adjust callsite of the above.
35020 * config/aarch64/aarch64-simd-builtins.def (sabd, uabd): Define
35022 * config/aarch64/arm_neon.h (vabd_s8): Reimplement using
35024 (vabd_s16): Likewise.
35025 (vabd_s32): Likewise.
35026 (vabd_u8): Likewise.
35027 (vabd_u16): Likewise.
35028 (vabd_u32): Likewise.
35029 (vabdq_s8): Likewise.
35030 (vabdq_s16): Likewise.
35031 (vabdq_s32): Likewise.
35032 (vabdq_u8): Likewise.
35033 (vabdq_u16): Likewise.
35034 (vabdq_u32): Likewise.
35036 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
35038 * config/aarch64/aarch64-simd-builtins.def (saba, uaba): Define
35040 * config/aarch64/arm_neon.h (vaba_s8): Implement using builtin.
35041 (vaba_s16): Likewise.
35042 (vaba_s32): Likewise.
35043 (vaba_u8): Likewise.
35044 (vaba_u16): Likewise.
35045 (vaba_u32): Likewise.
35046 (vabaq_s8): Likewise.
35047 (vabaq_s16): Likewise.
35048 (vabaq_s32): Likewise.
35049 (vabaq_u8): Likewise.
35050 (vabaq_u16): Likewise.
35051 (vabaq_u32): Likewise.
35053 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
35055 * config/aarch64/aarch64-simd.md (aba<mode>_3): Rename to...
35056 (aarch64_<su>aba<mode>): ... This. Handle uaba as well.
35057 Change RTL pattern to match.
35059 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
35061 * common/config/riscv/riscv-common.c (riscv_current_subset_list): New.
35062 * config/riscv/riscv-c.c (riscv-subset.h): New.
35063 (INCLUDE_STRING): Define.
35064 (riscv_cpu_cpp_builtins): Add new style architecture extension
35066 * config/riscv/riscv-subset.h (riscv_subset_list::begin): New.
35067 (riscv_subset_list::end): New.
35068 (riscv_current_subset_list): New.
35070 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
35072 * common/config/riscv/riscv-common.c (RISCV_DONT_CARE_VERSION):
35073 Move to riscv-subset.h.
35074 (struct riscv_subset_t): Ditto.
35075 (class riscv_subset_list): Ditto.
35076 * config/riscv/riscv-subset.h (RISCV_DONT_CARE_VERSION): Move
35077 from riscv-common.c.
35078 (struct riscv_subset_t): Ditto.
35079 (class riscv_subset_list): Ditto.
35080 * config/riscv/t-riscv ($(common_out_file)): Add file
35083 2021-01-07 Jakub Jelinek <jakub@redhat.com>
35086 * config/i386/i386.md (*bmi_blsi_<mode>_cmp, *bmi_blsi_<mode>_ccno):
35087 New define_insn patterns.
35089 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
35091 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_UNARY:optab><mode>)
35092 (*cond_<SVE_INT_UNARY:optab><mode>_2): Extend from SVE_FULL_I to SVE_I.
35093 (*cond_<SVE_INT_UNARY:optab><mode>_any): Likewise.
35095 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
35097 PR tree-optimization/98560
35098 * internal-fn.def (IFN_VCONDU, IFN_VCONDEQ): Use type vec_cond.
35099 * internal-fn.c (vec_cond_mask_direct): Get the data mode from
35101 (vec_cond_direct): Likewise argument 2.
35102 (vec_condu_direct, vec_condeq_direct): Delete.
35103 (expand_vect_cond_optab_fn): Rename to...
35104 (expand_vec_cond_optab_fn): ...this, replacing old macro.
35105 (expand_vec_condu_optab_fn, expand_vec_condeq_optab_fn): Delete.
35106 (expand_vect_cond_mask_optab_fn): Rename to...
35107 (expand_vec_cond_mask_optab_fn): ...this, replacing old macro.
35108 (direct_vec_cond_mask_optab_supported_p): Treat the optab as a
35110 (direct_vec_cond_optab_supported_p): Likewise.
35111 (direct_vec_condu_optab_supported_p): Delete.
35112 (direct_vec_condeq_optab_supported_p): Delete.
35113 * gimple-isel.cc: Include internal-fn.h.
35114 (gimple_expand_vec_cond_expr): Check that IFN_VCONDEQ is supported
35117 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
35119 PR tree-optimization/98560
35120 * gimple-isel.cc (gimple_expand_vec_cond_expr): If we fail to use
35121 IFN_VCOND{,U,EQ}, fall back on IFN_VCOND_MASK.
35123 2021-01-07 Uroš Bizjak <ubizjak@gmail.com>
35125 * config/i386/i386.md (insn): Merge from plusminus_insn, shift_insn,
35126 rotate_insn and optab code attributes.
35127 Update all uses to merged code attribute.
35128 * config/i386/sse.md: Update all uses to merged code attribute.
35129 * config/i386/mmx.md: Update all uses to merged code attribute.
35131 2021-01-07 Jakub Jelinek <jakub@redhat.com>
35133 PR tree-optimization/98568
35134 * gimple-ssa-store-merging.c (bswap_view_convert): New function.
35135 (bswap_replace): Use it.
35137 2021-01-06 Vladimir N. Makarov <vmakarov@redhat.com>
35139 PR rtl-optimization/97978
35140 * lra-int.h (lra_hard_reg_split_p): New external.
35141 * lra.c (lra_hard_reg_split_p): New global.
35142 (lra): Set up lra_hard_reg_split_p after splitting a hard reg.
35143 * lra-assigns.c (lra_assign): Don't check allocation correctness
35144 after hard reg splitting.
35146 2021-01-06 Martin Sebor <msebor@redhat.com>
35149 * builtins.c (new_delete_mismatch_p): New overload.
35150 (new_delete_mismatch_p (tree, tree)): Call it.
35152 2021-01-06 Alexandre Oliva <oliva@adacore.com>
35154 * Makefile.in (T_GLIMITS_H): New.
35155 (stmp-int-hdrs): Depend on it, use it.
35156 * config/t-vxworks (T_GLIMITS_H): Override it.
35157 (vxw-glimits.h): New.
35159 2021-01-06 Richard Biener <rguenther@suse.de>
35161 PR tree-optimization/98513
35162 * value-range.cc (intersect_ranges): Compare the upper bounds
35163 for the expected relation.
35165 2021-01-06 Gerald Pfeifer <gerald@pfeifer.com>
35168 2020-12-28 Gerald Pfeifer <gerald@pfeifer.com>
35170 * doc/standards.texi (HSAIL): Remove section.
35172 2021-01-05 Samuel Thibault <samuel.thibault@ens-lyon.org>
35174 * configure: Re-generate.
35176 2021-01-05 Jakub Jelinek <jakub@redhat.com>
35178 * doc/invoke.texi (-std=c++20): Adjust for the publication of
35179 ISO 14882:2020 standard.
35180 * doc/standards.texi: Likewise.
35182 2021-01-05 Jakub Jelinek <jakub@redhat.com>
35184 PR tree-optimization/94802
35185 * expr.h (maybe_optimize_sub_cmp_0): Declare.
35186 * expr.c: Include tree-pretty-print.h and flags.h.
35187 (maybe_optimize_sub_cmp_0): New function.
35188 (do_store_flag): Use it.
35189 * cfgexpand.c (expand_gimple_cond): Likewise.
35191 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
35193 * mux-utils.h (pointer_mux::m_ptr): Tweak description of contents.
35194 * rtlanal.c (simple_regno_set): Tweak description to clarify the
35197 2021-01-05 Richard Biener <rguenther@suse.de>
35199 PR tree-optimization/98516
35200 * tree-vect-slp.c (vect_optimize_slp): Permute the incoming
35201 lanes when materializing on a VEC_PERM node.
35202 (vectorizable_slp_permutation): Dump the permute properly.
35204 2021-01-05 Richard Biener <rguenther@suse.de>
35206 * tree-vect-slp.c (vect_slp_region): Move debug counter
35207 to cover individual subgraphs.
35209 2021-01-05 Richard Biener <rguenther@suse.de>
35211 PR tree-optimization/98428
35212 * tree-vect-slp.c (vect_build_slp_tree_1): Properly reject
35213 vector lane extracts for loop vectorization.
35215 2021-01-05 Jakub Jelinek <jakub@redhat.com>
35217 PR tree-optimization/98514
35218 * tree-ssa-reassoc.c (bb_rank): Change type from long * to
35220 (operand_rank): Change type from hash_map<tree, long> to
35221 hash_map<tree, int64_t>.
35222 (phi_rank): Change return type from long to int64_t.
35223 (loop_carried_phi): Change block_rank variable type from long to
35225 (propagate_rank): Change return type, rank parameter type and
35226 op_rank variable type from long to int64_t.
35227 (find_operand_rank): Change return type from long to int64_t
35228 and change slot variable type from long * to int64_t *.
35229 (insert_operand_rank): Change rank parameter type from long to
35231 (get_rank): Change return type and rank variable type from long to
35232 int64_t. Use PRId64 instead of ld to print the rank.
35233 (init_reassoc): Change rank variable type from long to int64_t
35234 and adjust correspondingly bb_rank and operand_rank initialization.
35236 2021-01-05 Jakub Jelinek <jakub@redhat.com>
35238 PR tree-optimization/96928
35239 * tree-ssa-phiopt.c (xor_replacement): New function.
35240 (tree_ssa_phiopt_worker): Call it.
35242 2021-01-05 Jakub Jelinek <jakub@redhat.com>
35244 PR tree-optimization/96930
35245 * match.pd ((A / (1 << B)) -> (A >> B)): If A is extended
35246 from narrower value which has the same type as 1 << B, perform
35247 the right shift on the narrower value followed by extension.
35249 2021-01-05 Jakub Jelinek <jakub@redhat.com>
35251 PR tree-optimization/96239
35252 * gimple-ssa-store-merging.c (maybe_optimize_vector_constructor): New
35254 (get_status_for_store_merging): Don't return BB_INVALID for blocks
35255 with potential bswap optimizable CONSTRUCTORs.
35256 (pass_store_merging::execute): Optimize vector CONSTRUCTORs with bswap
35259 2021-01-05 Richard Biener <rguenther@suse.de>
35261 PR tree-optimization/98381
35262 * tree.c (vector_element_bits): Properly compute bool vector
35264 * tree-vect-loop.c (vectorizable_live_operation): Properly
35265 compute the last lane bit offset.
35267 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
35270 * config/i386/sse.md (sse_cvtps2pi): Redefine as define_insn_and_split.
35271 Clear the top 64 bytes of the input XMM register.
35272 (sse_cvttps2pi): Ditto.
35274 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
35277 * config/i386/xopintrin.h (_mm256_cmov_si256): New.
35279 2021-01-05 H.J. Lu <hjl.tools@gmail.com>
35282 * config/i386/xmmintrin.h (_mm_extract_pi16): Cast to unsigned
35285 2021-01-05 Claudiu Zissulescu <claziss@synopsys.com>
35287 * config/arc/arc.md (maddsidi4_split): Use ACC_REG_FIRST.
35288 (umaddsidi4_split): Likewise.
35290 2021-01-05 liuhongt <hongtao.liu@intel.com>
35293 * config/i386/sse.md (*sse2_pmovskb_zexthisi): New
35294 define_insn_and_split for zero_extend of subreg HI of pmovskb
35296 (*sse2_pmovskb_zexthisi): Add new combine splitters for
35297 zero_extend of not of subreg HI of pmovskb result.
35299 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
35302 * explow.c (convert_memory_address_addr_space_1): Handle UNSPECs
35304 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate): Use
35305 convert_memory_address to convert symbolic immediates to ptr_mode
35306 before forcing them to memory.
35308 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
35310 PR rtl-optimization/97144
35311 * recog.c (constrain_operands): Initialize matching_operand
35312 for each alternative, rather than only doing it once.
35314 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
35316 PR rtl-optimization/98403
35317 * rtl-ssa/changes.cc (function_info::finalize_new_accesses): Explain
35318 why we don't remove call clobbers.
35319 (function_info::apply_changes_to_insn): Don't attempt to add
35320 call clobbers here.
35322 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
35324 PR tree-optimization/98371
35325 * tree-vect-loop.c (vect_reanalyze_as_main_loop): New function.
35326 (vect_analyze_loop): If an epilogue loop appears to be cheaper
35327 than the main loop, re-analyze it as a main loop before adopting
35330 2021-01-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
35333 * configure.ac (NETLIBS): Determine using AX_LIB_SOCKET_NSL.
35334 * aclocal.m4, configure: Regenerate.
35335 * Makefile.in (NETLIBS): Define.
35336 (BACKEND): Remove $(CODYLIB).
35338 2021-01-05 Jakub Jelinek <jakub@redhat.com>
35340 PR rtl-optimization/98334
35341 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
35342 Optimize (X - 1) * Y + Y to X * Y or (X + 1) * Y - Y to X * Y.
35344 2021-01-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
35346 * tree-inline.c (expand_call_inline): Restore input_location.
35347 Return result from recursive call.
35349 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
35351 PR tree-optimization/95401
35352 * config/aarch64/aarch64-sve-builtins.cc
35353 (gimple_folder::load_store_cookie): Use bits rather than bytes
35354 for the alignment argument to IFN_MASK_LOAD and IFN_MASK_STORE.
35355 * gimple-fold.c (gimple_fold_mask_load_store_mem_ref): Likewise.
35356 * tree-vect-stmts.c (vectorizable_store): Likewise.
35357 (vectorizable_load): Likewise.
35359 2021-01-04 Richard Biener <rguenther@suse.de>
35361 PR tree-optimization/98308
35362 * tree-vect-stmts.c (vectorizable_load): Set invariant mask
35365 2021-01-04 Jakub Jelinek <jakub@redhat.com>
35367 PR tree-optimization/95771
35368 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Handle types
35369 with precision smaller than int's precision and types with precision
35370 twice as large as long long. Formatting fixes.
35372 2021-01-04 Richard Biener <rguenther@suse.de>
35374 PR tree-optimization/98464
35375 * tree-ssa-sccvn.c (vn_valueize_for_srt): Rename from ...
35376 (vn_valueize_wrapper): ... this. Temporarily adjust vn_context_bb.
35377 (process_bb): Adjust.
35379 2021-01-04 Matthew Malcomson <matthew.malcomson@arm.com>
35382 * doc/invoke.texi (-fsanitize=address): Fix wording describing
35383 clash with -fsanitize=hwaddress.
35385 2021-01-04 Richard Biener <rguenther@suse.de>
35387 PR tree-optimization/98282
35388 * tree-ssa-sccvn.c (vn_get_stmt_kind): Classify tcc_reference on
35389 invariants as VN_NARY.
35391 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
35394 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Accept
35395 aarch64_simd_reg_or_zero for operand 2. Use the combinez patterns
35396 to handle zero operands.
35398 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
35400 * config/aarch64/aarch64.c (offset_6bit_signed_scaled_p): New function.
35401 (offset_6bit_unsigned_scaled_p): Fix typo in comment.
35402 (aarch64_sve_prefetch_operand_p): Accept MUL VLs in the range
35405 2021-01-04 Richard Biener <rguenther@suse.de>
35407 PR tree-optimization/98393
35408 * tree-vect-slp.c (vect_build_slp_tree): Properly zero matches
35409 when hitting the limit.
35411 2021-01-04 Richard Biener <rguenther@suse.de>
35413 PR tree-optimization/98291
35414 * tree-vect-loop.c (vectorizable_reduction): Bypass
35415 associativity check for SLP reductions with VF 1.
35417 2021-01-04 Jakub Jelinek <jakub@redhat.com>
35419 PR tree-optimization/96782
35420 * match.pd (x == ~x -> false, x != ~x -> true): New simplifications.
35422 2021-01-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
35424 * collect-utils.c (collect_execute): Check dumppfx.
35425 * collect2.c (maybe_run_lto_and_relink, do_link): Pass atsuffix
35426 to collect_execute.
35427 (do_link): Add new parameter atsuffix.
35428 (main): Handle -dumpdir option. Skip one argument for
35429 -o, -isystem and -B options.
35430 * gcc.c (make_at_file): New helper function.
35431 (close_at_file): Use it.
35433 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35435 * config/darwin.h (MIN_LD64_NO_COAL_SECTS): Adjust.
35436 Amend handling for LD64_VERSION fallback defaults.
35438 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35440 * config.gcc: Compute default version information
35441 from the configured target. Likewise defaults for
35443 * config/darwin10.h: Removed.
35444 * config/darwin12.h: Removed.
35445 * config/darwin9.h: Removed.
35446 * config/rs6000/darwin8.h: Removed.
35448 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35450 * config/darwin9.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
35452 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35454 * config/darwin9.h (STACK_CHECK_STATIC_BUILTIN): Move from here..
35455 * config/darwin.h (STACK_CHECK_STATIC_BUILTIN): .. to here.
35457 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35459 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move from
35461 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): ... to here.
35463 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35465 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move the spec
35466 for the Darwin10 unwinder stub from here ...
35467 * config/darwin.h (LINK_COMMAND_SPEC_A): ... to here.
35469 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
35471 * config/darwin.h (DSYMUTIL_SPEC): Default to DWARF
35472 (ASM_DEBUG_SPEC):Only define if the assembler supports
35474 (PREFERRED_DEBUGGING_TYPE): Default to DWARF.
35475 (DARWIN_PREFER_DWARF): Define.
35476 * config/darwin9.h (PREFERRED_DEBUGGING_TYPE): Remove.
35477 (DARWIN_PREFER_DWARF): Likewise
35478 (DSYMUTIL_SPEC): Likewise.
35479 (COLLECT_RUN_DSYMUTIL): Likewise.
35480 (ASM_DEBUG_SPEC): Likewise.
35481 (ASM_DEBUG_OPTION_SPEC): Likewise.
35483 2021-01-02 Jan Hubicka <jh@suse.cz>
35485 * cfg.c (free_block): ggc_free bb.
35487 2021-01-01 Jakub Jelinek <jakub@redhat.com>
35489 * gcc.c (process_command): Update copyright notice dates.
35490 * gcov-dump.c (print_version): Ditto.
35491 * gcov.c (print_version): Ditto.
35492 * gcov-tool.c (print_version): Ditto.
35493 * gengtype.c (create_file): Ditto.
35494 * doc/cpp.texi: Bump @copying's copyright year.
35495 * doc/cppinternals.texi: Ditto.
35496 * doc/gcc.texi: Ditto.
35497 * doc/gccint.texi: Ditto.
35498 * doc/gcov.texi: Ditto.
35499 * doc/install.texi: Ditto.
35500 * doc/invoke.texi: Ditto.
35502 2021-01-01 Jakub Jelinek <jakub@redhat.com>
35504 * ChangeLog-2020: Rotate ChangeLog. New file.
35507 Copyright (C) 2021 Free Software Foundation, Inc.
35509 Copying and distribution of this file, with or without modification,
35510 are permitted in any medium without royalty provided the copyright
35511 notice and this notice are preserved.